{
    "2312.02027": {
        "paper_data": {
            "title": "Stochastic Optimal Control Matching",
            "url": "http://arxiv.org/abs/2312.02027v5",
            "arxiv_id": "2312.02027",
            "authors": [
                "Carles Domingo-Enrich",
                "Jiequn Han",
                "Brandon Amos",
                "Joan Bruna",
                "Ricky T. Q. Chen"
            ],
            "abstract": "Stochastic optimal control, which has the goal of driving the behavior of noisy systems, is broadly applicable in science, engineering and artificial intelligence. Our work introduces Stochastic Optimal Control Matching (SOCM), a novel Iterative Diffusion Optimization (IDO) technique for stochastic optimal control that stems from the same philosophy as the conditional score matching loss for diffusion models. That is, the control is learned via a least squares problem by trying to fit a matching vector field. The training loss, which is closely connected to the cross-entropy loss, is optimized with respect to both the control function and a family of reparameterization matrices which appear in the matching vector field. The optimization with respect to the reparameterization matrices aims at minimizing the variance of the matching vector field. Experimentally, our algorithm achieves lower error than all the existing IDO techniques for stochastic optimal control for three out of four control problems, in some cases by an order of magnitude. The key idea underlying SOCM is the path-wise reparameterization trick, a novel technique that may be of independent interest. Code at https://github.com/facebookresearch/SOC-matching",
            "introduction": " Introduction Stochastic optimal control aims to drive the behavior of a noisy system in order to minimize a given cost. It has myriad applications in science and engineering: examples include the simulation of rare events in molecular dynamics (Hartmann et al., 2014; Hartmann and Sch\u00fctte, 2012; Zhang et al., 2014; Holdijk et al., 2023), finance and economics (Pham, 2009; Fleming and Stein, 2004), stochastic filtering and data assimilation (Mitter, 1996; Reich, 2019), nonconvex optimization (Chaudhari et al., 2018), power systems and energy markets (Belloni et al., 2016; Powell and Meisel, 2016), and robotics (Theodorou et al., 2011; Gorodetsky et al., 2018). Stochastic optimal has also been very impactful in neighboring fields such as mean-field games (Carmona et al., 2018), optimal transport (Villani, 2003, 2008), backward stochastic differential equations (BSDEs) (Carmona, 2016) and large deviations (Feng and Kurtz, 2006). Recently, it has been the basis of algorithms to sample from unnormalized densities (Zhang and Chen, 2022; Vargas et al., 2023; Berner et al., 2023; Richter and Berner, 2024). For continuous-time problems with low-dimensional state spaces, the standard approach to learn the optimal control is to solve the Hamilton-Jacobi-Bellman (HJB) partial differential equation (PDE) by gridding the space and using classical numerical methods to solve the Hamilton-Jacobi-Bellman equation for these 1D problems. It is not a problem where we actually have the ground truth control in closed form. Path Integral Sampler on Mixture of Gaussians We set b(x, t) = 0 , f (x, t) = 0 , g (x) = log( \u00b50(x)/\u00b5(x)) =\u2212\u2225x\u22252 2\u2212d 2log(2\u03c0)\u2212log\u00b5(x),(175) where T= 1, and \u00b5is the density of a mixture of two Gaussians with means \u00b1e1, where e1= (1,0, . . . , 0), and variance Id. Note that we take \u00b5to be normalized, i.e.R \u00b5(x) dx= 1, or equivalently, logZ= log\u0000R \u00b5(x) dx\u0001 = 0. In Figure 4, we use the following Monte Carlo estimator of the control objective at the control u: \u02c6Su(X) =RT 0\u00001 2\u2225u(Xu t, t)\u22252+f(Xu t, t)\u0001 dt+g(Xu T) +R \u27e8u(Xu t, t),dBt\u27e9. (176) Note that this estimator is unbiased because E[R \u27e8u(Xu t, t),dBt\u27e9] = 0. This is known as the Sticking the Landing estimator, as it has zero variance when uis the optimal control (Roeder et al., 2017). The fact that E[\u2212\u02c6Su(X)]\u2264logZ= 0with equality when u=u\u2217is stated as (Zhang and Chen, 2022, Thm. 4). E.2 Model architectures As a general guideline, the control function can be thought of as the analog of the score function in diffusion models; hence, a natural choice for the architecture can be U-Nets or diffusion transformers if the control task is on images, audio or video. Other domains may require different architectures. In the experiments), this quantity is zero for the optimal control u\u2217. Hence, the normalized standard deviation of \u03b1is an alternative metric to measure the optimality of the learned control. 37Figure 7 shows an exponential moving average of the norm squared of the gradient for Linear OU and Double Well . For Linear OU , the minimum gradient norm is achieved by the adjoint method, while forDouble Well it is achieved by the cross entropy loss. The training instabilities of the adjoint method become apparent as well. Interestingly, in both settings the algorithms with smallest gradients are not SOCM, which is the algorithm with smallest error as shown in Figure 2. Understanding this phenomenon is outside of the scope of this paper. Figure 8 shows that",
            "references": [
                {
                    "title": "Improved sampling via learned diffusions",
                    "abstract": "Recently, a series of papers proposed deep learning-based approaches to sample from target distributions using controlled diffusion processes, being trained only on the unnormalized target densities without access to samples. Building on previous work, we identify these approaches as special cases of a generalized Schr\\\"odinger bridge problem, seeking a stochastic evolution between a given prior distribution and the specified target. We further generalize this framework by introducing a variational formulation based on divergences between path space measures of time-reversed diffusion processes. This abstract perspective leads to practical losses that can be optimized by gradient-based algorithms and includes previous objectives as special cases. At the same time, it allows us to consider divergences other than the reverse Kullback-Leibler divergence that is known to suffer from mode collapse. In particular, we propose the so-called log-variance loss, which exhibits favorable numerical properties and leads to significantly improved performance across all considered approaches."
                },
                {
                    "title": "Stochastic Interpolants: A Unifying Framework for Flows and Diffusions",
                    "abstract": "A class of generative models that unifies flow-based and diffusion-based methods is introduced. These models extend the framework proposed in Albergo&Vanden-Eijnden (2023), enabling the use of a broad class of continuous-time stochastic processes called `stochastic interpolants' to bridge any two arbitrary probability density functions exactly in finite time. These interpolants are built by combining data from the two prescribed densities with an additional latent variable that shapes the bridge in a flexible way. The time-dependent probability density function of the stochastic interpolant is shown to satisfy a first-order transport equation as well as a family of forward and backward Fokker-Planck equations with tunable diffusion. Upon consideration of the time evolution of an individual sample, this viewpoint immediately leads to both deterministic and stochastic generative models based on probability flow equations or stochastic differential equations with an adjustable level of noise. The drift coefficients entering these models are time-dependent velocity fields characterized as the unique minimizers of simple quadratic objective functions, one of which is a new objective for the score of the interpolant density. Remarkably, we show that minimization of these quadratic objectives leads to control of the likelihood for any of our generative models built upon stochastic dynamics. By contrast, we establish that generative models based upon a deterministic dynamics must, in addition, control the Fisher divergence between the target and the model. We also construct estimators for the likelihood and the cross-entropy of interpolant-based generative models, discuss connections with other stochastic bridges, and demonstrate that such models recover the Schr\\\"odinger bridge between the two target densities when explicitly optimizing over the interpolant."
                },
                {
                    "title": "Denoising Diffusion Samplers",
                    "abstract": "Denoising diffusion models are a popular class of generative models providing state-of-the-art results in many domains. One adds gradually noise to data using a diffusion to transform the data distribution into a Gaussian distribution. Samples from the generative model are then obtained by simulating an approximation of the time-reversal of this diffusion initialized by Gaussian samples. Practically, the intractable score terms appearing in the time-reversed process are approximated using score matching techniques. We explore here a similar idea to sample approximately from unnormalized probability density functions and estimate their normalizing constants. We consider a process where the target density diffuses towards a Gaussian. Denoising Diffusion Samplers (DDS) are obtained by approximating the corresponding time-reversal. While score matching is not applicable in this context, we can leverage many of the ideas introduced in generative modeling for Monte Carlo sampling. Existing theoretical results from denoising diffusion models also provide theoretical guarantees for DDS. We discuss the connections between DDS, optimal control and Schr\\\"odinger bridges and finally demonstrate DDS experimentally on a variety of challenging sampling tasks."
                },
                {
                    "title": "An optimal control perspective on diffusion-based generative modeling",
                    "abstract": "We establish a connection between stochastic optimal control and generative models based on stochastic differential equations (SDEs), such as recently developed diffusion probabilistic models. In particular, we derive a Hamilton-Jacobi-Bellman equation that governs the evolution of the log-densities of the underlying SDE marginals. This perspective allows to transfer methods from optimal control theory to generative modeling. First, we show that the evidence lower bound is a direct consequence of the well-known verification theorem from control theory. Further, we can formulate diffusion-based generative modeling as a minimization of the Kullback-Leibler divergence between suitable measures in path space. Finally, we develop a novel diffusion-based method for sampling from unnormalized densities -- a problem frequently occurring in statistics and computational sciences. We demonstrate that our time-reversed diffusion sampler (DIS) can outperform other diffusion-based sampling approaches on multiple numerical examples."
                },
                {
                    "title": "Building Normalizing Flows with Stochastic Interpolants",
                    "abstract": "A generative model based on a continuous-time normalizing flow between any pair of base and target probability densities is proposed. The velocity field of this flow is inferred from the probability current of a time-dependent density that interpolates between the base and the target in finite time. Unlike conventional normalizing flow inference methods based the maximum likelihood principle, which require costly backpropagation through ODE solvers, our interpolant approach leads to a simple quadratic loss for the velocity itself which is expressed in terms of expectations that are readily amenable to empirical estimation. The flow can be used to generate samples from either the base or target, and to estimate the likelihood at any time along the interpolant. In addition, the flow can be optimized to minimize the path length of the interpolant density, thereby paving the way for building optimal transport maps. In situations where the base is a Gaussian density, we also show that the velocity of our normalizing flow can also be used to construct a diffusion model to sample the target as well as estimate its score. However, our approach shows that we can bypass this diffusion completely and work at the level of the probability flow with greater simplicity, opening an avenue for methods based solely on ordinary differential equations as an alternative to those based on stochastic differential equations. Benchmarking on density estimation tasks illustrates that the learned flow can match and surpass conventional continuous flows at a fraction of the cost, and compares well with diffusions on image generation on CIFAR-10 and ImageNet $32\\times32$. The method scales ab-initio ODE flows to previously unreachable image resolutions, demonstrated up to $128\\times128$."
                },
                {
                    "title": "Stochastic Optimal Control for Collective Variable Free Sampling of Molecular Transition Paths",
                    "abstract": "We consider the problem of sampling transition paths between two given metastable states of a molecular system, e.g. a folded and unfolded protein or products and reactants of a chemical reaction. Due to the existence of high energy barriers separating the states, these transition paths are unlikely to be sampled with standard Molecular Dynamics (MD) simulation. Traditional methods to augment MD with a bias potential to increase the probability of the transition rely on a dimensionality reduction step based on Collective Variables (CVs). Unfortunately, selecting appropriate CVs requires chemical intuition and traditional methods are therefore not always applicable to larger systems. Additionally, when incorrect CVs are used, the bias potential might not be minimal and bias the system along dimensions irrelevant to the transition. Showing a formal relation between the problem of sampling molecular transition paths, the Schr\\\"odinger bridge problem and stochastic optimal control with neural network policies, we propose a machine learning method for sampling said transitions. Unlike previous non-machine learning approaches our method, named PIPS, does not depend on CVs. We show that our method successful generates low energy transitions for Alanine Dipeptide as well as the larger Polyproline and Chignolin proteins."
                },
                {
                    "title": "Path Integral Sampler: a stochastic control approach for sampling",
                    "abstract": "We present Path Integral Sampler~(PIS), a novel algorithm to draw samples from unnormalized probability density functions. The PIS is built on the Schr\\\"odinger bridge problem which aims to recover the most likely evolution of a diffusion process given its initial distribution and terminal distribution. The PIS draws samples from the initial distribution and then propagates the samples through the Schr\\\"odinger bridge to reach the terminal distribution. Applying the Girsanov theorem, with a simple prior diffusion, we formulate the PIS as a stochastic optimal control problem whose running cost is the control energy and terminal cost is chosen according to the target distribution. By modeling the control as a neural network, we establish a sampling algorithm that can be trained end-to-end. We provide theoretical justification of the sampling quality of PIS in terms of Wasserstein distance when sub-optimal control is used. Moreover, the path integrals theory is used to compute importance weights of the samples to compensate for the bias induced by the sub-optimality of the controller and time-discretization. We experimentally demonstrate the advantages of PIS compared with other start-of-the-art sampling methods on a variety of tasks."
                },
                {
                    "title": "A Neural Network Approach for High-Dimensional Optimal Control Applied to Multiagent Path Finding",
                    "abstract": "We propose a neural network (NN) approach that yields approximate solutions for high-dimensional optimal control (OC) problems and demonstrate its effectiveness using examples from multiagent path finding. Our approach yields control in a feedback form, where the policy function is given by an NN. In particular, we fuse the Hamilton\u2013Jacobi\u2013Bellman (HJB) and Pontryagin maximum principle (PMP) approaches by parameterizing the value function with an NN. Our approach enables us to obtain approximately OCs in real time without having to solve an optimization problem. Once the policy function is trained, generating a control at a given space\u2013time location takes milliseconds; in contrast, efficient nonlinear programming methods typically perform the same task in seconds. We train the NN offline using the objective function of the control problem and penalty terms that enforce the HJB equations. Therefore, our training algorithm does not involve data generated by another algorithm. By training on a distribution of initial states, we ensure the controls\u2019 optimality on a large portion of the state space. Our grid-free approach scales efficiently to dimensions where grids become impractical or infeasible. We apply our approach to several multiagent collision-avoidance problems in up to 150 dimensions. Furthermore, we empirically observe that the number of parameters in our approach scales linearly with the dimension of the control problem, thereby mitigating the curse of dimensionality."
                },
                {
                    "title": "Neural SDEs as Infinite-Dimensional GANs",
                    "abstract": "Stochastic differential equations (SDEs) are a staple of mathematical modelling of temporal dynamics. However, a fundamental limitation has been that such models have typically been relatively inflexible, which recent work introducing Neural SDEs has sought to solve. Here, we show that the current classical approach to fitting SDEs may be approached as a special case of (Wasserstein) GANs, and in doing so the neural and classical regimes may be brought together. The input noise is Brownian motion, the output samples are time-evolving paths produced by a numerical solver, and by parameterising a discriminator as a Neural Controlled Differential Equation (CDE), we obtain Neural SDEs as (in modern machine learning parlance) continuous-time generative time series models. Unlike previous work on this problem, this is a direct extension of the classical approach without reference to either prespecified statistics or density functions. Arbitrary drift and diffusions are admissible, so as the Wasserstein loss has a unique global minima, in the infinite data limit any SDE may be learnt. Example code has been made available as part of the \\texttt{torchsde} repository."
                },
                {
                    "title": "Actor-Critic Method for High Dimensional Static Hamilton-Jacobi-Bellman Partial Differential Equations based on Neural Networks",
                    "abstract": "We propose a novel numerical method for high dimensional Hamilton--Jacobi--Bellman (HJB) type elliptic partial differential equations (PDEs). The HJB PDEs, reformulated as optimal control problems, are tackled by the actor-critic framework inspired by reinforcement learning, based on neural network parametrization of the value and control functions. Within the actor-critic framework, we employ a policy gradient approach to improve the control, while for the value function, we derive a variance reduced least-squares temporal difference method using stochastic calculus. To numerically discretize the stochastic control problem, we employ an adaptive step size scheme to improve the accuracy near the domain boundary. Numerical examples up to $20$ spatial dimensions including the linear quadratic regulators, the stochastic Van der Pol oscillators, the diffusive Eikonal equations, and fully nonlinear elliptic PDEs derived from a regulator problem are presented to validate the effectiveness of our proposed method."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "A Semi-Lagrangian Scheme for Hamilton-Jacobi-Bellman Equations on Networks",
                    "abstract": "We present a semi-Lagrangian scheme for the approximation of a class of Hamilton--Jacobi--Bellman (HJB) equations on networks. The scheme is explicit, consistent, and stable for large time steps. W..."
                },
                {
                    "title": "Algorithms for solving high dimensional PDEs: from nonlinear Monte Carlo to machine learning",
                    "abstract": "In recent years, tremendous progress has been made on numerical algorithms for solving partial differential equations (PDEs) in a very high dimension, using ideas from either nonlinear (multilevel) Monte Carlo or deep learning. They are potentially free of the curse of dimensionality for many different applications and have been proven to be so in the case of some nonlinear Monte Carlo methods for nonlinear parabolic PDEs. In this paper, we review these numerical and theoretical advances. In addition to algorithms based on stochastic reformulations of the original problem, such as the multilevel Picard iteration and the deep backward stochastic differential equations method, we also discuss algorithms based on the more traditional Ritz, Galerkin, and least square formulations. We hope to demonstrate to the reader that studying PDEs as well as control and variational problems in very high dimensions might very well be among the most promising new directions in mathematics and scientific computing in the near future."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Scalable Gradients for Stochastic Differential Equations",
                    "abstract": "The adjoint sensitivity method scalably computes gradients of solutions to ordinary differential equations. We generalize this method to stochastic differential equations, allowing time-efficient and constant-memory computation of gradients with high-order adaptive solvers. Specifically, we derive a stochastic differential equation whose solution is the gradient, a memory-efficient algorithm for caching noise, and conditions under which numerical solutions converge. In addition, we combine our method with gradient-based stochastic variational inference for latent stochastic differential equations. We use our method to fit stochastic dynamics defined by neural networks, achieving competitive performance on a 50-dimensional motion capture dataset."
                },
                {
                    "title": "Deep Fictitious Play for Finding Markovian Nash Equilibrium in Multi-Agent Games",
                    "abstract": "We propose a deep neural network-based algorithm to identify the Markovian Nash equilibrium of general large $N$-player stochastic differential games. Following the idea of fictitious play, we recast the $N$-player game into $N$ decoupled decision problems (one for each player) and solve them iteratively. The individual decision problem is characterized by a semilinear Hamilton-Jacobi-Bellman equation, to solve which we employ the recently developed deep BSDE method. The resulted algorithm can solve large $N$-player games for which conventional numerical methods would suffer from the curse of dimensionality. Multiple numerical examples involving identical or heterogeneous agents, with risk-neutral or risk-sensitive objectives, are tested to validate the accuracy of the proposed algorithm in large group games. Even for a fifty-player game with the presence of common noise, the proposed algorithm still finds the approximate Nash equilibrium accurately, which, to our best knowledge, is difficult to achieve by other numerical algorithms."
                },
                {
                    "title": "Overcoming the curse of dimensionality in the numerical approximation of Allen\u2013Cahn partial differential equations via truncated full-history recursive multilevel Picard approximations",
                    "abstract": "Abstract One of the most challenging problems in applied mathematics is the approximate solution of nonlinear partial differential equations (PDEs) in high dimensions. Standard deterministic approximation methods like finite differences or finite elements suffer from the curse of dimensionality in the sense that the computational effort grows exponentially in the dimension. In this work we overcome this difficulty in the case of reaction\u2013diffusion type PDEs with a locally Lipschitz continuous coervice nonlinearity (such as Allen\u2013Cahn PDEs) by introducing and analyzing truncated variants of the recently introduced full-history recursive multilevel Picard approximation schemes."
                },
                {
                    "title": "Convergence Analysis of Machine Learning Algorithms for the Numerical Solution of Mean Field Control and Games: I - The Ergodic Case",
                    "abstract": "We propose two algorithms for the solution of the optimal control of ergodic McKean-Vlasov dynamics. Both algorithms are based on the approximation of the theoretical solutions by neural networks, the latter being characterized by their architecture and a set of parameters. This allows the use of modern machine learning tools, and efficient implementations of stochastic gradient descent. The first algorithm is based on the idiosyncrasies of the ergodic optimal control problem. We provide a mathematical proof of the convergence of the algorithm, and we analyze rigorously the approximation by controlling the different sources of error. The second method is an adaptation of the deep Galerkin method to the system of partial differential equations issued from the optimality condition. We demonstrate the efficiency of these algorithms on several numerical examples, some of them being chosen to show that our algorithms succeed where existing ones failed. We also argue that both methods can easily be applied to problems in dimensions larger than what can be found in the existing literature. Finally, we illustrate the fact that, although the first algorithm is specifically designed for mean field control problems, the second one is more general and can also be applied to the partial differential equation systems arising in the theory of mean field games."
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "Variational approach to rare event simulation using least-squares regression.",
                    "abstract": "We propose an adaptive importance sampling scheme for the simulation of rare events when the underlying dynamics is given by diffusion. The scheme is based on a Gibbs variational principle that is used to determine the optimal (i.e., zero-variance) change of measure and exploits the fact that the latter can be rephrased as a stochastic optimal control problem. The control problem can be solved by a stochastic approximation algorithm, using the Feynman-Kac representation of the associated dynamic programming equations, and we discuss numerical aspects for high-dimensional problems along with simple toy examples."
                },
                {
                    "title": "Data assimilation: The Schr\u00f6dinger perspective",
                    "abstract": "Data assimilation addresses the general problem of how to combine model-based predictions with partial and noisy observations of the process in an optimal manner. This survey focuses on sequential data assimilation techniques using probabilistic particle-based algorithms. In addition to surveying recent developments for discrete- and continuous-time data assimilation, both in terms of mathematical foundations and algorithmic implementations, we also provide a unifying framework from the perspective of coupling of measures, and Schr\u00f6dinger\u2019s boundary value problem for stochastic processes in particular."
                },
                {
                    "title": "Overcoming the curse of dimensionality in the numerical approximation of semilinear parabolic partial differential equations",
                    "abstract": "For a long time it has been well-known that high-dimensional linear parabolic partial differential equations (PDEs) can be approximated by Monte Carlo methods with a computational effort which grows polynomially both in the dimension and in the reciprocal of the prescribed accuracy. In other words, linear PDEs do not suffer from the curse of dimensionality. For general semilinear PDEs with Lipschitz coefficients, however, it remained an open question whether these suffer from the curse of dimensionality. In this paper we partially solve this open problem. More precisely, we prove in the case of semilinear heat equations with gradient-independent and globally Lipschitz continuous nonlinearities that the computational effort of a variant of the recently introduced multilevel Picard approximations grows at most polynomially both in the dimension and in the reciprocal of the required accuracy."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "Solving stochastic differential equations and Kolmogorov equations by means of deep learning",
                    "abstract": "Stochastic differential equations (SDEs) and the Kolmogorov partial differential equations (PDEs) associated to them have been widely used in models from engineering, finance, and the natural sciences. In particular, SDEs and Kolmogorov PDEs, respectively, are highly employed in models for the approximative pricing of financial derivatives. Kolmogorov PDEs and SDEs, respectively, can typically not be solved explicitly and it has been and still is an active topic of research to design and analyze numerical methods which are able to approximately solve Kolmogorov PDEs and SDEs, respectively. Nearly all approximation methods for Kolmogorov PDEs in the literature suffer under the curse of dimensionality or only provide approximations of the solution of the PDE at a single fixed space-time point. In this paper we derive and propose a numerical approximation method which aims to overcome both of the above mentioned drawbacks and intends to deliver a numerical approximation of the Kolmogorov PDE on an entire region $[a,b]^d$ without suffering from the curse of dimensionality. Numerical results on examples including the heat equation, the Black-Scholes model, the stochastic Lorenz equation, and the Heston model suggest that the proposed approximation algorithm is quite effective in high dimensions in terms of both accuracy and speed."
                },
                {
                    "title": "Deep Optimal Stopping",
                    "abstract": "In this paper we develop a deep learning method for optimal stopping problems which directly learns the optimal stopping rule from Monte Carlo samples. As such, it is broadly applicable in situations where the underlying randomness can efficiently be simulated. We test the approach on three problems: the pricing of a Bermudan max-call option, the pricing of a callable multi barrier reverse convertible and the problem of optimally stopping a fractional Brownian motion. In all three cases it produces very accurate results in high-dimensional situations with short computing times."
                },
                {
                    "title": "Variational Characterization of Free Energy: Theory and Algorithms",
                    "abstract": "The article surveys and extends variational formulations of the thermodynamic free energy and discusses their information-theoretic content from the perspective of mathematical statistics. We revisit the well-known Jarzynski equality for nonequilibrium free energy sampling within the framework of importance sampling and Girsanov change-of-measure transformations. The implications of the different variational formulations for designing efficient stochastic optimization and nonequilibrium simulation algorithms for computing free energies are discussed and illustrated."
                },
                {
                    "title": "Multilevel Picard Approximations of High-Dimensional Semilinear Parabolic Differential Equations with Gradient-Dependent Nonlinearities",
                    "abstract": "Parabolic partial differential equations (PDEs) and backward stochastic differential equations (BSDEs) have a wide range of applications. In particular, high-dimensional PDEs with gradient-dependent nonlinearities appear often in the state-of-the-art pricing and hedging of financial derivatives. In this article we prove that semilinear heat equations with gradient-dependent nonlinearities can be approximated under suitable assumptions with computational complexity that grows polynomially both in the dimension and the reciprocal of the accuracy."
                },
                {
                    "title": "Solving high-dimensional partial differential equations using deep learning",
                    "abstract": "Significance Partial differential equations (PDEs) are among the most ubiquitous tools used in modeling problems in nature. However, solving high-dimensional PDEs has been notoriously difficult due to the \u201ccurse of dimensionality.\u201d This paper introduces a practical algorithm for solving nonlinear PDEs in very high (hundreds and potentially thousands of) dimensions. Numerical results suggest that the proposed algorithm is quite effective for a wide variety of problems, in terms of both accuracy and speed. We believe that this opens up a host of possibilities in economics, finance, operational research, and physics, by considering all participating agents, assets, resources, or particles together at the same time, instead of making ad hoc assumptions on their interrelationships. Developing algorithms for solving high-dimensional partial differential equations (PDEs) has been an exceedingly difficult task for a long time, due to the notoriously difficult problem known as the \u201ccurse of dimensionality.\u201d This paper introduces a deep learning-based approach that can handle general high-dimensional parabolic PDEs. To this end, the PDEs are reformulated using backward stochastic differential equations and the gradient of the unknown solution is approximated by neural networks, very much in the spirit of deep reinforcement learning with the gradient acting as the policy function. Numerical results on examples including the nonlinear Black\u2013Scholes equation, the Hamilton\u2013Jacobi\u2013Bellman equation, and the Allen\u2013Cahn equation suggest that the proposed algorithm is quite effective in high dimensions, in terms of both accuracy and cost. This opens up possibilities in economics, finance, operational research, and physics, by considering all participating agents, assets, resources, or particles together at the same time, instead of making ad hoc assumptions on their interrelationships."
                },
                {
                    "title": "Sticking the Landing: Simple, Lower-Variance Gradient Estimators for Variational Inference",
                    "abstract": "We propose a simple and general variant of the standard reparameterized gradient estimator for the variational evidence lower bound. Specifically, we remove a part of the total derivative with respect to the variational parameters that corresponds to the score function. Removing this term produces an unbiased gradient estimator whose variance approaches zero as the approximate posterior approaches the exact posterior. We analyze the behavior of this gradient estimator theoretically and empirically, and generalize it to more complex variational distributions such as mixtures and importance-weighted posteriors."
                },
                {
                    "title": "High-dimensional stochastic optimal control using continuous tensor decompositions",
                    "abstract": "Motion planning and control problems are embedded and essential in almost all robotics applications. These problems are often formulated as stochastic optimal control problems and solved using dynamic programming algorithms. Unfortunately, most existing algorithms that guarantee convergence to optimal solutions suffer from the curse of dimensionality: the run time of the algorithm grows exponentially with the dimension of the state space of the system. We propose novel dynamic programming algorithms that alleviate the curse of dimensionality in problems that exhibit certain low-rank structure. The proposed algorithms are based on continuous tensor decompositions recently developed by the authors. Essentially, the algorithms represent high-dimensional functions (e.g. the value function) in a compressed format, and directly perform dynamic programming computations (e.g. value iteration, policy iteration) in this format. Under certain technical assumptions, the new algorithms guarantee convergence towards optimal solutions with arbitrary precision. Furthermore, the run times of the new algorithms scale polynomially with the state dimension and polynomially with the ranks of the value function. This approach realizes substantial computational savings in \u201ccompressible\u201d problem instances, where value functions admit low-rank approximations. We demonstrate the new algorithms in a wide range of problems, including a simulated six-dimensional agile quadcopter maneuvering example and a seven-dimensional aircraft perching example. In some of these examples, we estimate computational savings of up to 10 orders of magnitude over standard value iteration algorithms. We further demonstrate the algorithms running in real time on board a quadcopter during a flight experiment under motion capture."
                },
                {
                    "title": "A stochastic optimal control solution to the energy management of a microgrid with storage and renewables",
                    "abstract": "The presence of renewable energy generators in a microgrid calls for the usage of storage units so as to smooth the variability in energy production. This work addresses the optimal management of a battery in a microgrid including a wind turbine facility. A Markov chain model is employed to predict the wind power production and the optimal management of the energy storage element is formulated as a stochastic optimal control problem. An approximate dynamic programming approach resting on system abstraction is then proposed for control policy design. Some numerical examples show the effectiveness of the approach."
                },
                {
                    "title": "Safe and Efficient Off-Policy Reinforcement Learning",
                    "abstract": "In this work, we take a fresh look at some old and new algorithms for off-policy, return-based reinforcement learning. Expressing these in a common form, we derive a novel algorithm, Retrace($\\lambda$), with three desired properties: (1) it has low variance; (2) it safely uses samples collected from any behaviour policy, whatever its degree of \"off-policyness\"; and (3) it is efficient as it makes the best use of samples collected from near on-policy behaviour policies. We analyze the contractive nature of the related operator under both off-policy policy evaluation and control settings and derive online sample-based algorithms. We believe this is the first return-based off-policy control algorithm converging a.s. to $Q^*$ without the GLIE assumption (Greedy in the Limit with Infinite Exploration). As a corollary, we prove the convergence of Watkins' Q($\\lambda$), which was an open problem since 1989. We illustrate the benefits of Retrace($\\lambda$) on a standard suite of Atari 2600 games."
                },
                {
                    "title": "Tutorial on Stochastic Optimization in Energy\u2014Part I: Modeling and Policies",
                    "abstract": "There is a wide range of problems in energy systems that require making decisions in the presence of different forms of uncertainty. The fields that address sequential, stochastic decision problems lack a standard canonical modeling framework, with fragmented, competing solution strategies. Recognizing that we will never agree on a single notational system, this two-part tutorial proposes a simple, straightforward canonical model (that is most familiar to people with a control theory background), and introduces four fundamental classes of policies which integrate the competing strategies that have been proposed under names such as control theory, dynamic programming, stochastic programming and robust optimization. Part II of the tutorial illustrates the modeling framework using a simple energy storage problem, where we show that, depending on the problem characteristics, each of the four classes of policies may be best."
                },
                {
                    "title": "Lectures on Bsdes, Stochastic Control, and Stochastic Differential Games with Financial Applications",
                    "abstract": "The goal of this textbook is to introduce students to the stochastic analysis tools that play an increasing role in the probabilistic approach to optimization problems, including stochastic control and stochastic differential games. While optimal control is taught in many graduate programs in applied mathematics and operations research, the author was intrigued by the lack of coverage of the theory of stochastic differential games. This is the first title in SIAM's Financial Mathematics book series and is based on the author s lecture notes. It will be helpful to students who are interested in stochastic differential equations (forward, backward, forward-backward); the probabilistic approach to stochastic control (dynamic programming and the stochastic maximum principle); and mean field games and control of McKean-Vlasov dynamics. The theory is illustrated by applications to models of systemic risk, macroeconomic growth, flocking/schooling, crowd behavior, and predatory trading, among others. Audience: This book is written for young researchers and newcomers to stochastic control and stochastic differential games."
                },
                {
                    "title": "Variational Inference with Normalizing Flows",
                    "abstract": "The choice of approximate posterior distribution is one of the core problems in variational inference. Most applications of variational inference employ simple families of posterior approximations in order to allow for efficient inference, focusing on mean-field or other simple structured approximations. This restriction has a significant impact on the quality of inferences made using variational methods. We introduce a new approach for specifying flexible, arbitrarily complex and scalable approximate posterior distributions. Our approximations are distributions constructed through a normalizing flow, whereby a simple initial density is transformed into a more complex one by applying a sequence of invertible transformations until a desired level of complexity is attained. We use this view of normalizing flows to develop categories of finite and infinitesimal flows and provide a unified view of approaches for constructing rich posterior approximations. We demonstrate that the theoretical advantages of having posteriors that better match the true posterior, combined with the scalability of amortized variational approaches, provides a clear improvement in performance and applicability of variational inference."
                },
                {
                    "title": "Applications of the Cross-Entropy Method to Importance Sampling and Optimal Control of Diffusions",
                    "abstract": "We study the cross-entropy method for diffusions. One of the results is a versatile cross-entropy algorithm that can be used to design efficient importance sampling strategies for rare events or to solve optimal control problems. The approach is based on the minimization of a suitable cross-entropy functional, with a parametric family of exponentially tilted probability distributions. We illustrate the new algorithm with several numerical examples and discuss algorithmic issues and possible extensions of the method."
                },
                {
                    "title": "Characterization of Rare Events in Molecular Dynamics",
                    "abstract": "A good deal of molecular dynamics simulations aims at predicting and quantifying rare events, such as the folding of a protein or a phase transition. Simulating rare events is often prohibitive, especially if the equations of motion are high-dimensional, as is the case in molecular dynamics. Various algorithms have been proposed for efficiently computing mean first passage times, transition rates or reaction pathways. This article surveys and discusses recent developments in the field of rare event simulation and outlines a new approach that combines ideas from optimal control and statistical mechanics. The optimal control approach described in detail resembles the use of Jarzynski's equality for free energy calculations, but with an optimized protocol that speeds up the sampling, while (theoretically) giving variance-free estimators of the rare events statistics. We illustrate the new approach with two numerical examples and discuss its relation to existing methods."
                },
                {
                    "title": "Efficient rare event simulation by optimal nonequilibrium forcing",
                    "abstract": "Rare event simulation and estimation for systems in equilibrium are among the most challenging topics in molecular dynamics. As was shown by Jarzynski and others, nonequilibrium forcing can theoretically be used to obtain equilibrium rare event statistics. The advantage seems to be that the external force can speed up the sampling of the rare events by biasing the equilibrium distribution towards a distribution under which the rare events are no longer rare. Yet algorithmic methods based on Jarzynski\u2019s and related results often fail to be efficient because they are based on sampling in path space. We present a new method that replaces the path sampling problem by minimization of a cross-entropy-like functional which boils down to finding the optimal nonequilibrium forcing. We show how to solve the related optimization problem in an efficient way by using an iterative strategy based on milestoning."
                },
                {
                    "title": "On the Convergence of Finite Element Methods for Hamilton-Jacobi-Bellman Equations",
                    "abstract": "We study the convergence of monotone $P1$ finite element methods on unstructured meshes for fully nonlinear Hamilton--Jacobi--Bellman equations arising from stochastic optimal control problems with possibly degenerate, isotropic diffusions. Using elliptic projection operators we treat discretizations which violate the consistency conditions of the framework by Barles and Souganidis. We obtain strong uniform convergence of the numerical solutions and, under nondegeneracy assumptions, strong $L^2$ convergence of the gradients."
                },
                {
                    "title": "Semi-Lagrangian schemes for linear and fully non-linear diffusion equations",
                    "abstract": "For linear and fully non-linear diffusion equations of Bellman-Isaacs type, we introduce a class of approximation schemes based on differencing and interpolation. As opposed to classical numerical methods, these schemes work for general diffusions with coefficient matrices that may be non-diagonal dominant and arbitrarily degenerate. In general such schemes have to have a wide stencil. Besides providing a unifying framework for several known first order accurate schemes, our class of schemes includes new first and higher order versions. The methods are easy to implement and more efficient than some other known schemes. We prove consistency and stability of the methods, and for the monotone first order methods, we prove convergence in the general case and robust error estimates in the convex case. The methods are extensively tested."
                },
                {
                    "title": "Large Deviations for Stochastic Processes",
                    "abstract": "The notes are devoted to results on large deviations for sequences of Markov processes following closely the book by Feng and Kurtz ([FK06]). We outline how convergence of Fleming's nonlinear semigroups (logarithmically transformed nonlinear semigroups) implies large deviation principles analogous to the use of convergence of linear semigroups in weak convergence. The latter method is based on the range condition for the corresponding generator. Viscosity solution methods however provide applicable conditions for the necessary nonlinear semigroup convergence. Once having established the validity of the large deviation principle one is concerned to obtain more tractable representations for the corresponding rate function. This in turn can be achieved once a variational representation of the limiting generator of Fleming's semigroup can be established. The obtained variational representation of the generator allows for a suitable control representation of the rate function. The notes conclude with a couple of examples to show how the methodology via Fleming's semigroups works. These notes are based on the mini-course 'Large deviations for stochastic processes' the author held during the workshop 'Dynamical Gibs-non-Gibbs transitions' at EURANDOM in Eindhoven, December 2011, and at the Max-Planck institute for mathematics in the sciences in Leipzig, July 2012."
                },
                {
                    "title": "The Cross-Entropy Method: A Unified Approach to Combinatorial Optimization, Monte-Carlo Simulation, and Machine Learning",
                    "abstract": "counterpart as a special case, which leads to the construction of more universal \ufb01lters that expand the capability and the robustness of data processing"
                },
                {
                    "title": "A regression-based Monte Carlo method to solve backward stochastic differential equations",
                    "abstract": "We are concerned with the numerical resolution of backward stochastic differential equations. We propose a new numerical scheme based on iterative regressions on function bases, which coefficients are evaluated using Monte Carlo simulations. A full convergence analysis is derived. Numerical experiments about finance are included, in particular, concerning option pricing with differential interest rates."
                },
                {
                    "title": "Path integrals and symmetry breaking for optimal control theory",
                    "abstract": "This paper considers linear-quadratic control of a non-linear dynamical system subject to arbitrary cost. I show that for this class of stochastic control problems the non-linear Hamilton\u2013Jacobi\u2013Bellman equation can be transformed into a linear equation. The transformation is similar to the transformation used to relate the classical Hamilton\u2013Jacobi equation to the Schr\u00f6dinger equation. As a result of the linearity, the usual backward computation can be replaced by a forward diffusion process that can be computed by stochastic integration or by the evaluation of a path integral. It is shown how in the deterministic limit the Pontryagin minimum principle formalism is recovered. The significance of the path integral approach is that it forms the basis for a number of efficient computational methods, such as Monte Carlo sampling, the Laplace approximation and the variational approximation. We show the effectiveness of the first two methods in a number of examples. Examples are given that show the qualitative difference between stochastic and deterministic control and the occurrence of symmetry breaking as a function of the noise."
                },
                {
                    "title": "A fast algorithm for the two dimensional HJB equation of stochastic control",
                    "abstract": "This paper analyses the implementation of the generalized finite differences method for the HJB equation of stochastic control, introduced by two of the authors in (Bonnans and Zidani, SIAM J. Numer. Anal. 41 (2003) 1008-1021). The computation of coefficients needs to solve at each point of the grid (and for each control) a linear programming problem. We show here that, for two dimensional problems, this linear programming problem can be solved in O(pmax )o perations, wherepmax is the size of the stencil. The method is based on a walk on the Stern-Brocot tree, and on the related filling of the set of positive semidefinite matrices of size two."
                },
                {
                    "title": "A numerical scheme for BSDEs",
                    "abstract": "In this paper we propose a numerical scheme for a class of backward stochastic differential equations (BSDEs) with possible path-dependent terminal values. We prove that our scheme converges in the strong L2 sense and derive its rate of convergence. As an intermediate step we prove an L2-type regularity of the solution to such BSDEs. Such a notion of regularity, which can be thought of as the modulus of continuity of the paths in an L2 sense, is new. Some other features of our scheme include the following: (i) both components of the solution are approximated by step processes (i.e., piecewise constant processes); (ii) the regularity requirements on the coefficients are practically \"minimum\"; (iii) the dimension of the integrals involved in the approximation is independent of the partition size."
                },
                {
                    "title": "Topics in Optimal Transportation",
                    "abstract": "Introduction The Kantorovich duality Geometry of optimal transportation Brenier's polar factorization theorem The Monge-Ampere equation Displacement interpolation and displacement convexity Geometric and Gaussian inequalities The metric side of optimal transportation A differential point of view on optimal transportation Entropy production and transportation inequalities Problems Bibliography Table of short statements Index."
                },
                {
                    "title": "Infinite-Horizon Policy-Gradient Estimation",
                    "abstract": "Gradient-based approaches to direct policy search in reinforcement learning have received much recent attention as a means to solve problems of partial observability and to avoid some of the problems associated with policy degradation in value-function methods. In this paper we introduce GPOMDP, a simulation-based algorithm for generating a biased estimate of the gradient of the average reward in Partially Observable Markov Decision Processes (POMDPs) controlled by parameterized stochastic policies. A similar algorithm was proposed by Kimura, Yamamura, and Kobayashi (1995). The algorithm's chief advantages are that it requires storage of only twice the number of policy parameters, uses one free parameter \u03b2 \u2208 [0, 1] (which has a natural interpretation in terms of bias-variance trade-off), and requires no knowledge of the underlying state. We prove convergence of GPOMDP, and show how the correct choice of the parameter \u03b2 is related to the mixing time of the controlled POMDP. We briefly describe extensions of GPOMDP to controlled Markov chains, continuous state, observation and control spaces, multiple-agents, higher-order derivatives, and a version for training stochastic policies with internal states. In a companion paper (Baxter, Bartlett, & Weaver, 2001) we show how the gradient estimates generated by GPOMDP can be used in both a traditional stochastic gradient algorithm and a conjugate-gradient procedure to find local optima of the average reward."
                },
                {
                    "title": "Filtering and stochastic control: a historical perspective",
                    "abstract": "We attempt to give a historical account of the main ideas leading to the development of nonlinear filtering and stochastic control as we know it today. We present a development of linear filtering theory, beginning with Wiener-Kolmogoroff filtering and ending with Kalman filtering. The linear-quadratic-Gaussian problem of stochastic control is considered and states that for this problem the optimal stochastic control can be constructed by solving separately a state estimation problem and a deterministic optimal control problem. Many of the ideas presented here generalize to the nonlinear situation. A reasonably detailed discussion of nonlinear filtering, again from the innovations viewpoint, is given. Finally, we deal with optimal stochastic control. The general method of discussing these problems is dynamic programming."
                },
                {
                    "title": "Numerical methods for stochastic control problems in continuous time",
                    "abstract": "H. J. Kushner and P. G. Dupuis. Springer-Verlag, New York/Heidelberg, May 1992. 439 pp., DM 98. ISBN 3-540-97834-8"
                },
                {
                    "title": "Some Guidelines and Guarantees for Common Random Numbers",
                    "abstract": "Common random numbers CRN is a widely-used technique for reducing variance in comparing stochastic systems through simulation. Its popularity derives from its intuitive appeal and ease of implementation. However, though CRN has been observed to work well with a broad range of models, the class of systems for which it is provably advantageous has remained rather limited. \n \nThis paper has two purposes: We first discuss the effectiveness and optimality of CRN in a general setting, stressing the roles played by monotonicity and continuity properties. We then present specific, new classes of systems and comparisons for which CRN is beneficial and even optimal. Our conclusions for these systems are largely consistent with simulation practice and lend further theoretical support to folklore. Our results differ from those of previous analyses primarily because we put conditions on the timing of events, rather than the sequence of states, in a discrete-event simulation. \n \nWe formulate our results in three settings corresponding to three applications of CRN: distributional comparisons, structural comparisons, and sensitivity analysis. In each case, we make use of conditions that simultaneously ensure monotonicity and continuity in the timing of events. These properties are established through explicit recursions for event epochs in terms of increasing, continuous functions."
                },
                {
                    "title": "A Monte Carlo method for sensitivity analysis and parametric optimization of nonlinear stochastic systems",
                    "abstract": "For high-dimensional or nonlinear problems there are serious limitations on the power of available computational methods for the optimization or parametric optimization of stochastic systems. The paper develops an effective Monte Carlo method for obtaining good estimators of systems sensitivities with respect to system parameters under quite general conditions on the systems and cost functions. The value of the method is borne out by numerical experiments, and the computational requirements are favorable with respect to competing methods when the dimension is high or the nonlinearities \u201csevere.\u201d The method is a type of \u201cderivative of likelihood ratio\u201d method. Jump-diffusion, functional diffusion, and reflected diffusion models of broad types are covered by the basic technique (e.g., the type of limit model that arises in the analysis of queueing systems under heavy traffic, where the boundary reflection conditions are discontinuous). For a wide class of problems, the cost function or dynamics need not be ..."
                },
                {
                    "title": "On the Convergence Rates of IPA and FDC Derivative Estimators",
                    "abstract": "We show that under the (sufficient) conditions usually given for infinitesimal perturbation analysis (IPA) to apply for derivative estimation, a finite-difference scheme with common random numbers (FDC) has the same order of convergence, namely O ( n \u22121/2 ), provided that the size of the finite-difference interval converges to zero fast enough. This holds for both one- and two-sided FDC. This also holds for different variants of IPA, such as some versions of smoothed perturbation analysis (SPA), which is based on conditional expectation. Finally, this also holds for the estimation of steady-state performance measures by truncated-horizon estimators, under some ergodicity assumptions. Our developments do not involve monotonicity, but are based on continuity and smoothness. We give examples and numerical illustrations which show that the actual difference in mean square error (MSE) between IPA and FDC is typically negligible. We also obtain the order of convergence of that difference, which is faster than the convergence of the MSE to zero."
                },
                {
                    "title": "Sensitivity Analysis for Simulations via Likelihood Ratios",
                    "abstract": "We present a simple method of estimating the sensitivity of quantities obtained from simulation with respect to a class of parameters. Here sensitivity means the derivative of an expectation with respect to a parameter. The class of parameters includes, for example, Poisson rates, discrete probabilities, and the mean and variance of a Normal distribution. The method is extremely well suited to regenerative simulation, and can be implemented on extant simulations with little effort, increase in running time, or memory requirements. It is based on some change-of-measure formulas derived from likelihood ratios. In addition to the theorems that underly the technique, we present some numerical examples."
                },
                {
                    "title": "Stochastic differential equations : an introduction with applications",
                    "abstract": "Some Mathematical Preliminaries.- Ito Integrals.- The Ito Formula and the Martingale Representation Theorem.- Stochastic Differential Equations.- The Filtering Problem.- Diffusions: Basic Properties.- Other Topics in Diffusion Theory.- Applications to Boundary Value Problems.- Application to Optimal Stopping.- Application to Stochastic Control.- Application to Mathematical Finance."
                },
                {
                    "title": "Numerical Methods for Hamilton-Jacobi-Bellman Equations",
                    "abstract": "Numerical Methods for Hamilton-Jacobi-Bellman equation by Constantin Greif The University of Wisconsin Milwaukee, 2017 Under the Supervision of Professor Bruce A. Wade In this work we considered HJB equations, that arise from stochastic optimal control problems with a finite time interval. If the diffusion is allowed to become degenerate, the solution cannot be understood in the classical sense. Therefore one needs the notion of viscosity solutions. With some stability and consistency assumptions, monotone methods provide the convergence to the viscosity solution. In this thesis we looked at monotone finite difference methods, semi lagragian methods and finite element methods for isotropic diffusion. In the last chapter we introduce the vanishing moment method, a method not based on monotonicity. \u00a9Copyright by Constantin Greif, 2017 All Rights Reserved"
                },
                {
                    "title": "An iterative path integral stochastic optimal control approach for learning robotic tasks",
                    "abstract": "Abstract Recent work on path integral stochastic optimal control theory Theodorou et al. (2010a); Theodorou (2011) has shown promising results in planning and control of nonlinear systems in high dimensional state spaces. The path integral control framework relies on the transformation of the nonlinear Hamilton Jacobi Bellman (HJB) partial differential equation (PDE) into a linear PDE and the approximation of its solution via the use of the Feynman Kac lemma. In this work, we are reviewing the generalized version of path integral stochastic optimal control formalism Theodorou et al. (2010a), used for optimal control and planing of stochastic dynamical systems with state dependent control and diffusion matrices. Moreover we present the iterative path integral control approach, the so called P olicy I mprovement with P ath I ntegrals or (PI 2 ) which is capable of scaling in high dimensional robotic control problems. Furthermore we present a convergence analysis of the proposed algorithm and we apply the proposed framework to a variety of robotic tasks. Finally with the goal to perform locomotion the iterative path integral control is applied for learning nonlinear limit cycle attractors with adjustable land scape."
                },
                {
                    "title": "Brownian Motion and Stochastic Calculus",
                    "abstract": "1 Preliminaries of Measure Theory De\u0085nition 1 F P ( ) is said to be an algebra if (1) 2 F (2) A;B 2 F implies A S B 2 F (3) A 2 F implies AC 2 F . F is said to be a semialgebra or semi-ring is (1) ;? 2 F (2) A;B 2 F implies AB 2 F (3) A;B 2 F implies AnB = Pn k=1Ak for some Ak 2 F De\u0085nition 2 -Class: If A;B 2 implies AB 2 De\u0085nition 3 -Class: if (1) 2 ;(2) A;B 2 with A B implies A B 2 ; (3) An 2 ; An \" A implies A 2 De\u0085nition 4 Let L R be such that 2 L implies ; 1 2 L. L-class: (1)1 2 L (2) L is linearly closed (3) n 2 L; 0 n \" ; bounded or 2 L implies 2 L Proposition 5 (i)M being -Class & -Class impliesM is -algebra (ii) If -Class contains the -Class ; then ( )"
                },
                {
                    "title": "LANDMARK PAPER REPRISE - STOCHASTIC APPROXIMATION FOR MONTE CARLO OPTIMIZATION",
                    "abstract": "In In this paper, we introduce two convergent Monte Carlo algorithms for optimizing complex stochastic systems. The first algorithm, which is applicable to to regenerative processes, operates by estimating finite differences. The second method is of Robbins-Monro type and is applicable to to Markov chains. The algorithm is driven by derivative estimates obtained via a likelihood ratio argument."
                },
                {
                    "title": "SENSITIVITY ANALYSIS USING IT\u00d4 \u2013 MALLIAVIN CALCULUS AND",
                    "abstract": "Abstract. We consider a multidimensional diffusion process (X\u03b1 t )0\u2264t\u2264T whose dynamics depends on a parameter \u03b1. Our first purpose is to write as an expectation the sensitivity \u2207\u03b1J(\u03b1) for the expected cost J(\u03b1) = E(f(X\u03b1 T )), in order to evaluate it using Monte Carlo simulations. This issue arises, for example, from stochastic control problems (where the controller is parameterized, which reduces the control problem to a parametric optimization one) or from model misspecifications in finance. Previous evaluations of \u2207\u03b1J(\u03b1) using simulations were limited to smooth cost functions f or to diffusion coefficients not depending on \u03b1 (see Yang and Kushner, SIAM J. Control Optim., 29 (1991), pp. 1216\u20131249). In this paper, we cover the general case, deriving three new approaches to evaluate \u2207\u03b1J(\u03b1), which we call the Malliavin calculus approach, the adjoint approach, and the martingale approach. To accomplish this, we leverage It\u00f4 calculus, Malliavin calculus, and martingale arguments. In the second part of this work, we provide discretization procedures to simulate the relevant random variables; then we analyze their respective errors. This analysis proves that the discretization error is essentially linear with respect to the time step. This result, which was already known in some specific situations, appears to be true in this much wider context. Finally, we provide numerical experiments in random mechanics and finance and compare the different methods in terms of variance, complexity, computational time, and time discretization error."
                }
            ],
            "categories": [
                "math.OC",
                "cs.LG",
                "cs.NA",
                "math.NA",
                "math.PR",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively learn optimal control policies for stochastic systems with high-dimensional state spaces using advanced sampling techniques?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of stochastic optimal control, which has significant implications across various domains such as finance, robotics, and molecular dynamics. By developing more effective methods for learning optimal control policies, we can enhance the performance of systems in uncertain environments, leading to improved decision-making and efficiency in real-world applications. This research could pave the way for future studies that explore more complex systems and contribute to the development of robust algorithms that can handle high-dimensional data, ultimately influencing both theoretical understanding and practical implementations in diverse fields.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the high-dimensional nature of the state spaces involved, which complicates the application of traditional numerical methods like the Hamilton-Jacobi-Bellman equation. Naive approaches may fail due to the curse of dimensionality, where the computational cost and complexity increase exponentially with the number of dimensions. Additionally, the stochastic nature of the systems introduces noise and uncertainty, making it difficult to accurately estimate control policies. Overcoming these technical obstacles requires innovative sampling techniques and robust model architectures that can effectively capture the dynamics of the system.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on low-dimensional problems or relied on simplified models that do not adequately capture the complexities of real-world systems. Limitations in computational resources and the lack of sophisticated sampling methods have hindered progress in high-dimensional stochastic optimal control. Additionally, existing solutions may not generalize well across different applications, leading to a gap in effective methodologies. Our approach aims to bridge this gap by leveraging advanced sampling techniques and model architectures that have not been fully explored in the context of stochastic control.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using a Path Integral Sampler on a mixture of Gaussians to learn optimal control policies. We will utilize a dataset generated from stochastic systems with high-dimensional state spaces and evaluate our approach using metrics such as the normalized standard deviation of the control function and the performance of the learned policies in minimizing the control objective. The expected outcomes include demonstrating the effectiveness of our sampling technique in achieving optimal control and providing insights into the stability and performance of different model architectures, such as"
            }
        },
        "author_data": {
            "526b62e3-2f3f-46eb-a661-6cabdbc68107": {
                "pk": "526b62e3-2f3f-46eb-a661-6cabdbc68107",
                "name": "Carles Domingo-Enrich",
                "collaborators": [
                    "Youssef Mroueh",
                    "Joan Bruna",
                    "Alberto Bietti",
                    "Eric Vanden-Eijnden",
                    "Fabian Pedregosa",
                    "Damien Scieur",
                    "Yair Schiff",
                    "Raaz Dwivedi",
                    "Lester Mackey",
                    "Marylou Gabri\u00e9"
                ],
                "domain": [
                    "Generative Modeling",
                    "Stochastic Optimization",
                    "Differential Privacy",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Depth and Feature Learning are Provably Beneficial for Neural Network Discriminators",
                        "abstract": "We construct pairs of distributions $\\mu_d, \\nu_d$ on $\\mathbb{R}^d$ such that the quantity $|\\mathbb{E}_{x \\sim \\mu_d} [F(x)] - \\mathbb{E}_{x \\sim \\nu_d} [F(x)]|$ decreases as $\\Omega(1/d^2)$ for some three-layer ReLU network $F$ with polynomial width and weights, while declining exponentially in $d$ if $F$ is any two-layer network with polynomial weights. This shows that deep GAN discriminators are able to distinguish distributions that shallow discriminators cannot. Analogously, we build pairs of distributions $\\mu_d, \\nu_d$ on $\\mathbb{R}^d$ such that $|\\mathbb{E}_{x \\sim \\mu_d} [F(x)] - \\mathbb{E}_{x \\sim \\nu_d} [F(x)]|$ decreases as $\\Omega(1/(d\\log d))$ for two-layer ReLU networks with polynomial weights, while declining exponentially for bounded-norm functions in the associated RKHS. This confirms that feature learning is beneficial for discriminators. Our bounds are based on Fourier transforms."
                    },
                    {
                        "title": "Computing the Variance of Shuffling Stochastic Gradient Algorithms via Power Spectral Density Analysis",
                        "abstract": "When solving finite-sum minimization problems, two common alternatives to stochastic gradient descent (SGD) with theoretical benefits are random reshuffling (SGD-RR) and shuffle-once (SGD-SO), in which functions are sampled in cycles without replacement. Under a convenient stochastic noise approximation which holds experimentally, we study the stationary variances of the iterates of SGD, SGD-RR and SGD-SO, whose leading terms decrease in this order, and obtain simple approximations. To obtain our results, we study the power spectral density of the stochastic gradient noise sequences. Our analysis extends beyond SGD to SGD with momentum and to the stochastic Nesterov's accelerated gradient method. We perform experiments on quadratic objective functions to test the validity of our approximation and the correctness of our findings."
                    },
                    {
                        "title": "A Taxonomy of Loss Functions for Stochastic Optimal Control",
                        "abstract": "Stochastic optimal control (SOC) aims to direct the behavior of noisy systems and has widespread applications in science, engineering, and artificial intelligence. In particular, reward fine-tuning of diffusion and flow matching models and sampling from unnormalized methods can be recast as SOC problems. A recent work has introduced Adjoint Matching (Domingo-Enrich et al., 2024), a loss function for SOC problems that vastly outperforms existing loss functions in the reward fine-tuning setup. The goal of this work is to clarify the connections between all the existing (and some new) SOC loss functions. Namely, we show that SOC loss functions can be grouped into classes that share the same gradient in expectation, which means that their optimization landscape is the same; they only differ in their gradient variance. We perform simple SOC experiments to understand the strengths and weaknesses of different loss functions."
                    },
                    {
                        "title": "Auditing Differential Privacy in High Dimensions with the Kernel Quantum R\u00e9nyi Divergence",
                        "abstract": "Differential privacy (DP) is the de facto standard for private data release and private machine learning. Auditing black-box DP algorithms and mechanisms to certify whether they satisfy a certain DP guarantee is challenging, especially in high dimension. We propose relaxations of differential privacy based on new divergences on probability distributions: the kernel R\\'enyi divergence and its regularized version. We show that the regularized kernel R\\'enyi divergence can be estimated from samples even in high dimensions, giving rise to auditing procedures for $\\varepsilon$-DP, $(\\varepsilon,\\delta)$-DP and $(\\alpha,\\varepsilon)$-R\\'enyi DP."
                    },
                    {
                        "title": "Separation Results between Fixed-Kernel and Feature-Learning Probability Metrics",
                        "abstract": "Several works in implicit and explicit generative modeling empirically observed that feature-learning discriminators outperform fixed-kernel discriminators in terms of the sample quality of the models. We provide separation results between probability metrics with fixed-kernel and feature-learning discriminators using the function classes $\\mathcal{F}_2$ and $\\mathcal{F}_1$ respectively, which were developed to study overparametrized two-layer neural networks. In particular, we construct pairs of distributions over hyper-spheres that can not be discriminated by fixed kernel $(\\mathcal{F}_2)$ integral probability metric (IPM) and Stein discrepancy (SD) in high dimensions, but that can be discriminated by their feature learning ($\\mathcal{F}_1$) counterparts. To further study the separation we provide links between the $\\mathcal{F}_1$ and $\\mathcal{F}_2$ IPMs with sliced Wasserstein distances. Our work suggests that fixed-kernel discriminators perform worse than their feature learning counterparts because their corresponding metrics are weaker."
                    },
                    {
                        "title": "Simultaneous Transport Evolution for Minimax Equilibria on Measures",
                        "abstract": "Min-max optimization problems arise in several key machine learning setups, including adversarial learning and generative modeling. In their general form, in absence of convexity/concavity assumptions, finding pure equilibria of the underlying two-player zero-sum game is computationally hard [Daskalakis et al., 2021]. In this work we focus instead in finding mixed equilibria, and consider the associated lifted problem in the space of probability measures. By adding entropic regularization, our main result establishes global convergence towards the global equilibrium by using simultaneous gradient ascent-descent with respect to the Wasserstein metric -- a dynamics that admits efficient particle discretization in high-dimensions, as opposed to entropic mirror descent. We complement this positive result with a related entropy-regularized loss which is not bilinear but still convex-concave in the Wasserstein geometry, and for which simultaneous dynamics do not converge yet timescale separation does. Taken together, these results showcase the benign geometry of bilinear games in the space of measures, enabling particle dynamics with global qualitative convergence guarantees."
                    },
                    {
                        "title": "Tighter Sparse Approximation Bounds for ReLU Neural Networks",
                        "abstract": "A well-known line of work (Barron, 1993; Breiman, 1993; Klusowski & Barron, 2018) provides bounds on the width $n$ of a ReLU two-layer neural network needed to approximate a function $f$ over the ball $\\mathcal{B}_R(\\mathbb{R}^d)$ up to error $\\epsilon$, when the Fourier based quantity $C_f = \\frac{1}{(2\\pi)^{d/2}} \\int_{\\mathbb{R}^d} \\|\\xi\\|^2 |\\hat{f}(\\xi)| \\ d\\xi$ is finite. More recently Ongie et al. (2019) used the Radon transform as a tool for analysis of infinite-width ReLU two-layer networks. In particular, they introduce the concept of Radon-based $\\mathcal{R}$-norms and show that a function defined on $\\mathbb{R}^d$ can be represented as an infinite-width two-layer neural network if and only if its $\\mathcal{R}$-norm is finite. In this work, we extend the framework of Ongie et al. (2019) and define similar Radon-based semi-norms ($\\mathcal{R}, \\mathcal{U}$-norms) such that a function admits an infinite-width neural network representation on a bounded open set $\\mathcal{U} \\subseteq \\mathbb{R}^d$ when its $\\mathcal{R}, \\mathcal{U}$-norm is finite. Building on this, we derive sparse (finite-width) neural network approximation bounds that refine those of Breiman (1993); Klusowski & Barron (2018). Finally, we show that infinite-width neural network representations on bounded open sets are not unique and study their structure, providing a functional view of mode connectivity."
                    },
                    {
                        "title": "Average-case Acceleration for Bilinear Games and Normal Matrices",
                        "abstract": "Advances in generative modeling and adversarial learning have given rise to renewed interest in smooth games. However, the absence of symmetry in the matrix of second derivatives poses challenges that are not present in the classical minimization framework. While a rich theory of average-case analysis has been developed for minimization problems, little is known in the context of smooth games. In this work we take a first step towards closing this gap by developing average-case optimal first-order methods for a subset of smooth games. We make the following three main contributions. First, we show that for zero-sum bilinear games the average-case optimal method is the optimal method for the minimization of the Hamiltonian. Second, we provide an explicit expression for the optimal method corresponding to normal matrices, potentially non-symmetric. Finally, we specialize it to matrices with eigenvalues located in a disk and show a provable speed-up compared to worst-case optimal algorithms. We illustrate our findings through benchmarks with a varying degree of mismatch with our assumptions."
                    },
                    {
                        "title": "On Energy-Based Models with Overparametrized Shallow Neural Networks",
                        "abstract": "Energy-based models (EBMs) are a simple yet powerful framework for generative modeling. They are based on a trainable energy function which defines an associated Gibbs measure, and they can be trained and sampled from via well-established statistical tools, such as MCMC. Neural networks may be used as energy function approximators, providing both a rich class of expressive models as well as a flexible device to incorporate data structure. In this work we focus on shallow neural networks. Building from the incipient theory of overparametrized neural networks, we show that models trained in the so-called \"active\" regime provide a statistical advantage over their associated \"lazy\" or kernel regime, leading to improved adaptivity to hidden low-dimensional structure in the data distribution, as already observed in supervised learning. Our study covers both maximum likelihood and Stein Discrepancy estimators, and we validate our theoretical results with numerical experiments on synthetic data."
                    },
                    {
                        "title": "Learning with Stochastic Orders",
                        "abstract": "Learning high-dimensional distributions is often done with explicit likelihood modeling or implicit modeling via minimizing integral probability metrics (IPMs). In this paper, we expand this learning paradigm to stochastic orders, namely, the convex or Choquet order between probability measures. Towards this end, exploiting the relation between convex orders and optimal transport, we introduce the Choquet-Toland distance between probability measures, that can be used as a drop-in replacement for IPMs. We also introduce the Variational Dominance Criterion (VDC) to learn probability measures with dominance constraints, that encode the desired stochastic order between the learned measure and a known baseline. We analyze both quantities and show that they suffer from the curse of dimensionality and propose surrogates via input convex maxout networks (ICMNs), that enjoy parametric rates. We provide a min-max framework for learning with stochastic orders and validate it experimentally on synthetic and high-dimensional image generation, with promising results. Finally, our ICMNs class of convex functions and its derived Rademacher Complexity are of independent interest beyond their application in convex orders."
                    },
                    {
                        "title": "Compress Then Test: Powerful Kernel Testing in Near-linear Time",
                        "abstract": "Kernel two-sample testing provides a powerful framework for distinguishing any pair of distributions based on $n$ sample points. However, existing kernel tests either run in $n^2$ time or sacrifice undue power to improve runtime. To address these shortcomings, we introduce Compress Then Test (CTT), a new framework for high-powered kernel testing based on sample compression. CTT cheaply approximates an expensive test by compressing each $n$ point sample into a small but provably high-fidelity coreset. For standard kernels and subexponential distributions, CTT inherits the statistical behavior of a quadratic-time test -- recovering the same optimal detection boundary -- while running in near-linear time. We couple these advances with cheaper permutation testing, justified by new power analyses; improved time-vs.-quality guarantees for low-rank approximation; and a fast aggregation procedure for identifying especially discriminating kernels. In our experiments with real and simulated data, CTT and its extensions provide 20--200x speed-ups over state-of-the-art approximate MMD tests with no loss of power."
                    },
                    {
                        "title": "Dual Training of Energy-Based Models with Overparametrized Shallow Neural Networks",
                        "abstract": "Energy-based models (EBMs) are generative models that are usually trained via maximum likelihood estimation. This approach becomes challenging in generic situations where the trained energy is non-convex, due to the need to sample the Gibbs distribution associated with this energy. Using general Fenchel duality results, we derive variational principles dual to maximum likelihood EBMs with shallow overparametrized neural network energies, both in the feature-learning and lazy linearized regimes. In the feature-learning regime, this dual formulation justifies using a two time-scale gradient ascent-descent (GDA) training algorithm in which one updates concurrently the particles in the sample space and the neurons in the parameter space of the energy. We also consider a variant of this algorithm in which the particles are sometimes restarted at random samples drawn from the data set, and show that performing these restarts at every iteration step corresponds to score matching training. These results are illustrated in simple numerical experiments, which indicates that GDA performs best when features and particles are updated using similar time scales."
                    },
                    {
                        "title": "Adjoint Matching: Fine-tuning Flow and Diffusion Generative Models with Memoryless Stochastic Optimal Control",
                        "abstract": "Dynamical generative models that produce samples through an iterative process, such as Flow Matching and denoising diffusion models, have seen widespread use, but there have not been many theoretically-sound methods for improving these models with reward fine-tuning. In this work, we cast reward fine-tuning as stochastic optimal control (SOC). Critically, we prove that a very specific memoryless noise schedule must be enforced during fine-tuning, in order to account for the dependency between the noise variable and the generated samples. We also propose a new algorithm named Adjoint Matching which outperforms existing SOC algorithms, by casting SOC problems as a regression problem. We find that our approach significantly improves over existing methods for reward fine-tuning, achieving better consistency, realism, and generalization to unseen human preference reward models, while retaining sample diversity."
                    }
                ]
            },
            "2a704e2e-afb1-4705-85ae-de45aa2fcd6a": {
                "pk": "2a704e2e-afb1-4705-85ae-de45aa2fcd6a",
                "name": "Jiequn Han",
                "collaborators": [
                    "Weinan E",
                    "Jihao Long",
                    "Ruimeng Hu",
                    "Linfeng Zhang",
                    "Arnulf Jentzen",
                    "Yue Zhao",
                    "Joan Bruna",
                    "Qianxiao Li",
                    "Chao Ma",
                    "Zheng Ma"
                ],
                "domain": [
                    "Stochastic Control",
                    "Reinforcement Learning",
                    "Deep Learning",
                    "Partial Differential Equations"
                ],
                "publications": [
                    {
                        "title": "Deep Learning Approximation for Stochastic Control Problems",
                        "abstract": "Many real world stochastic control problems suffer from the \"curse of dimensionality\". To overcome this difficulty, we develop a deep learning approach that directly solves high-dimensional stochastic control problems based on Monte-Carlo sampling. We approximate the time-dependent controls as feedforward neural networks and stack these networks together through model dynamics. The objective function for the control problem plays the role of the loss function for the deep neural network. We test this approach using examples from the areas of optimal trading and energy storage. Our results suggest that the algorithm presented here achieves satisfactory accuracy and at the same time, can handle rather high dimensional problems."
                    },
                    {
                        "title": "Convergence of the Deep BSDE Method for Coupled FBSDEs",
                        "abstract": "The recently proposed numerical algorithm, deep BSDE method, has shown remarkable performance in solving high-dimensional forward-backward stochastic differential equations (FBSDEs) and parabolic partial differential equations (PDEs). This article lays a theoretical foundation for the deep BSDE method in the general case of coupled FBSDEs. In particular, a posteriori error estimation of the solution is provided and it is proved that the error converges to zero given the universal approximation capability of neural networks. Numerical results are presented to demonstrate the accuracy of the analyzed algorithm in solving high-dimensional coupled FBSDEs."
                    },
                    {
                        "title": "Perturbational Complexity by Distribution Mismatch: A Systematic Analysis of Reinforcement Learning in Reproducing Kernel Hilbert Space",
                        "abstract": "Most existing theoretical analysis of reinforcement learning (RL) is limited to the tabular setting or linear models due to the difficulty in dealing with function approximation in high dimensional space with an uncertain environment. This work offers a fresh perspective into this challenge by analyzing RL in a general reproducing kernel Hilbert space (RKHS). We consider a family of Markov decision processes $\\mathcal{M}$ of which the reward functions lie in the unit ball of an RKHS and transition probabilities lie in a given arbitrary set. We define a quantity called perturbational complexity by distribution mismatch $\\Delta_{\\mathcal{M}}(\\epsilon)$ to characterize the complexity of the admissible state-action distribution space in response to a perturbation in the RKHS with scale $\\epsilon$. We show that $\\Delta_{\\mathcal{M}}(\\epsilon)$ gives both the lower bound of the error of all possible algorithms and the upper bound of two specific algorithms (fitted reward and fitted Q-iteration) for the RL problem. Hence, the decay of $\\Delta_\\mathcal{M}(\\epsilon)$ with respect to $\\epsilon$ measures the difficulty of the RL problem on $\\mathcal{M}$. We further provide some concrete examples and discuss whether $\\Delta_{\\mathcal{M}}(\\epsilon)$ decays fast or not in these examples. As a byproduct, we show that when the reward functions lie in a high dimensional RKHS, even if the transition probability is known and the action space is finite, it is still possible for RL problems to suffer from the curse of dimensionality."
                    },
                    {
                        "title": "Deep Fictitious Play for Finding Markovian Nash Equilibrium in Multi-Agent Games",
                        "abstract": "We propose a deep neural network-based algorithm to identify the Markovian Nash equilibrium of general large $N$-player stochastic differential games. Following the idea of fictitious play, we recast the $N$-player game into $N$ decoupled decision problems (one for each player) and solve them iteratively. The individual decision problem is characterized by a semilinear Hamilton-Jacobi-Bellman equation, to solve which we employ the recently developed deep BSDE method. The resulted algorithm can solve large $N$-player games for which conventional numerical methods would suffer from the curse of dimensionality. Multiple numerical examples involving identical or heterogeneous agents, with risk-neutral or risk-sensitive objectives, are tested to validate the accuracy of the proposed algorithm in large group games. Even for a fifty-player game with the presence of common noise, the proposed algorithm still finds the approximate Nash equilibrium accurately, which, to our best knowledge, is difficult to achieve by other numerical algorithms."
                    },
                    {
                        "title": "Recurrent Neural Networks for Stochastic Control Problems with Delay",
                        "abstract": "Stochastic control problems with delay are challenging due to the path-dependent feature of the system and thus its intrinsic high dimensions. In this paper, we propose and systematically study deep neural networks-based algorithms to solve stochastic control problems with delay features. Specifically, we employ neural networks for sequence modeling (\\emph{e.g.}, recurrent neural networks such as long short-term memory) to parameterize the policy and optimize the objective function. The proposed algorithms are tested on three benchmark examples: a linear-quadratic problem, optimal consumption with fixed finite delay, and portfolio optimization with complete memory. Particularly, we notice that the architecture of recurrent neural networks naturally captures the path-dependent feature with much flexibility and yields better performance with more efficient and stable training of the network compared to feedforward networks. The superiority is even evident in the case of portfolio optimization with complete memory, which features infinite delay."
                    },
                    {
                        "title": "Offline Supervised Learning V.S. Online Direct Policy Optimization: A Comparative Study and A Unified Training Paradigm for Neural Network-Based Optimal Feedback Control",
                        "abstract": "This work is concerned with solving neural network-based feedback controllers efficiently for optimal control problems. We first conduct a comparative study of two prevalent approaches: offline supervised learning and online direct policy optimization. Albeit the training part of the supervised learning approach is relatively easy, the success of the method heavily depends on the optimal control dataset generated by open-loop optimal control solvers. In contrast, direct policy optimization turns the optimal control problem into an optimization problem directly without any requirement of pre-computing, but the dynamics-related objective can be hard to optimize when the problem is complicated. Our results underscore the superiority of offline supervised learning in terms of both optimality and training time. To overcome the main challenges, dataset and optimization, in the two approaches respectively, we complement them and propose the Pre-train and Fine-tune strategy as a unified training paradigm for optimal feedback control, which further improves the performance and robustness significantly. Our code is accessible at https://github.com/yzhao98/DeepOptimalControl."
                    },
                    {
                        "title": "Posterior Sampling with Denoising Oracles via Tilted Transport",
                        "abstract": "Score-based diffusion models have significantly advanced high-dimensional data generation across various domains, by learning a denoising oracle (or score) from datasets. From a Bayesian perspective, they offer a realistic modeling of data priors and facilitate solving inverse problems through posterior sampling. Although many heuristic methods have been developed recently for this purpose, they lack the quantitative guarantees needed in many scientific applications.   In this work, we introduce the \\textit{tilted transport} technique, which leverages the quadratic structure of the log-likelihood in linear inverse problems in combination with the prior denoising oracle to transform the original posterior sampling problem into a new `boosted' posterior that is provably easier to sample from. We quantify the conditions under which this boosted posterior is strongly log-concave, highlighting the dependencies on the condition number of the measurement matrix and the signal-to-noise ratio. The resulting posterior sampling scheme is shown to reach the computational threshold predicted for sampling Ising models [Kunisky'23] with a direct analysis, and is further validated on high-dimensional Gaussian mixture models and scalar field $\\varphi^4$ models."
                    },
                    {
                        "title": "Reinforcement Learning with Function Approximation: From Linear to Nonlinear",
                        "abstract": "Function approximation has been an indispensable component in modern reinforcement learning algorithms designed to tackle problems with large state spaces in high dimensions. This paper reviews recent results on error analysis for these reinforcement learning algorithms in linear or nonlinear approximation settings, emphasizing approximation error and estimation error/sample complexity. We discuss various properties related to approximation error and present concrete conditions on transition probability and reward function under which these properties hold true. Sample complexity analysis in reinforcement learning is more complicated than in supervised learning, primarily due to the distribution mismatch phenomenon. With assumptions on the linear structure of the problem, numerous algorithms in the literature achieve polynomial sample complexity with respect to the number of features, episode length, and accuracy, although the minimax rate has not been achieved yet. These results rely on the $L^\\infty$ and UCB estimation of estimation error, which can handle the distribution mismatch phenomenon. The problem and analysis become substantially more challenging in the setting of nonlinear function approximation, as both $L^\\infty$ and UCB estimation are inadequate for bounding the error with a favorable rate in high dimensions. We discuss additional assumptions necessary to address the distribution mismatch and derive meaningful results for nonlinear RL problems."
                    },
                    {
                        "title": "Empowering Optimal Control with Machine Learning: A Perspective from Model Predictive Control",
                        "abstract": "Solving complex optimal control problems have confronted computational challenges for a long time. Recent advances in machine learning have provided us with new opportunities to address these challenges. This paper takes model predictive control, a popular optimal control method, as the primary example to survey recent progress that leverages machine learning techniques to empower optimal control solvers. We also discuss some of the main challenges encountered when applying machine learning to develop more robust optimal control algorithms."
                    },
                    {
                        "title": "A Mean-Field Optimal Control Formulation of Deep Learning",
                        "abstract": "Recent work linking deep neural networks and dynamical systems opened up new avenues to analyze deep learning. In particular, it is observed that new insights can be obtained by recasting deep learning as an optimal control problem on difference or differential equations. However, the mathematical aspects of such a formulation have not been systematically explored. This paper introduces the mathematical formulation of the population risk minimization problem in deep learning as a mean-field optimal control problem. Mirroring the development of classical optimal control, we state and prove optimality conditions of both the Hamilton-Jacobi-Bellman type and the Pontryagin type. These mean-field results reflect the probabilistic nature of the learning problem. In addition, by appealing to the mean-field Pontryagin's maximum principle, we establish some quantitative relationships between population and empirical learning problems. This serves to establish a mathematical foundation for investigating the algorithmic and theoretical connections between optimal control and deep learning."
                    },
                    {
                        "title": "Solving Many-Electron Schr\u00f6dinger Equation Using Deep Neural Networks",
                        "abstract": "We introduce a new family of trial wave-functions based on deep neural networks to solve the many-electron Schr\\\"odinger equation. The Pauli exclusion principle is dealt with explicitly to ensure that the trial wave-functions are physical. The optimal trial wave-function is obtained through variational Monte Carlo and the computational cost scales quadratically with the number of electrons. The algorithm does not make use of any prior knowledge such as atomic orbitals. Yet it is able to represent accurately the ground-states of the tested systems, including He, H2, Be, B, LiH, and a chain of 10 hydrogen atoms. This opens up new possibilities for solving large-scale many-electron Schr\\\"odinger equation."
                    },
                    {
                        "title": "Convergence of Deep Fictitious Play for Stochastic Differential Games",
                        "abstract": "Stochastic differential games have been used extensively to model agents' competitions in Finance, for instance, in P2P lending platforms from the Fintech industry, the banking system for systemic risk, and insurance markets. The recently proposed machine learning algorithm, deep fictitious play, provides a novel efficient tool for finding Markovian Nash equilibrium of large $N$-player asymmetric stochastic differential games [J. Han and R. Hu, Mathematical and Scientific Machine Learning Conference, pages 221-245, PMLR, 2020]. By incorporating the idea of fictitious play, the algorithm decouples the game into $N$ sub-optimization problems, and identifies each player's optimal strategy with the deep backward stochastic differential equation (BSDE) method parallelly and repeatedly. In this paper, we prove the convergence of deep fictitious play (DFP) to the true Nash equilibrium. We can also show that the strategy based on DFP forms an $\\eps$-Nash equilibrium. We generalize the algorithm by proposing a new approach to decouple the games, and present numerical results of large population games showing the empirical convergence of the algorithm beyond the technical assumptions in the theorems."
                    },
                    {
                        "title": "Algorithms for Solving High Dimensional PDEs: From Nonlinear Monte Carlo to Machine Learning",
                        "abstract": "In recent years, tremendous progress has been made on numerical algorithms for solving partial differential equations (PDEs) in a very high dimension, using ideas from either nonlinear (multilevel) Monte Carlo or deep learning. They are potentially free of the curse of dimensionality for many different applications and have been proven to be so in the case of some nonlinear Monte Carlo methods for nonlinear parabolic PDEs.   In this paper, we review these numerical and theoretical advances. In addition to algorithms based on stochastic reformulations of the original problem, such as the multilevel Picard iteration and the Deep BSDE method, we also discuss algorithms based on the more traditional Ritz, Galerkin, and least square formulations. We hope to demonstrate to the reader that studying PDEs as well as control and variational problems in very high dimensions might very well be among the most promising new directions in mathematics and scientific computing in the near future."
                    },
                    {
                        "title": "Solving high-dimensional partial differential equations using deep learning",
                        "abstract": "Developing algorithms for solving high-dimensional partial differential equations (PDEs) has been an exceedingly difficult task for a long time, due to the notoriously difficult problem known as the \"curse of dimensionality\". This paper introduces a deep learning-based approach that can handle general high-dimensional parabolic PDEs. To this end, the PDEs are reformulated using backward stochastic differential equations and the gradient of the unknown solution is approximated by neural networks, very much in the spirit of deep reinforcement learning with the gradient acting as the policy function. Numerical results on examples including the nonlinear Black-Scholes equation, the Hamilton-Jacobi-Bellman equation, and the Allen-Cahn equation suggest that the proposed algorithm is quite effective in high dimensions, in terms of both accuracy and cost. This opens up new possibilities in economics, finance, operational research, and physics, by considering all participating agents, assets, resources, or particles together at the same time, instead of making ad hoc assumptions on their inter-relationships."
                    },
                    {
                        "title": "Uniformly Accurate Machine Learning Based Hydrodynamic Models for Kinetic Equations",
                        "abstract": "A new framework is introduced for constructing interpretable and truly reliable reduced models for multiscale problems in situations without scale separation. Hydrodynamic approximation to the kinetic equation is used as an example to illustrate the main steps and issues involved. To this end, a set of generalized moments are constructed first to optimally represent the underlying velocity distribution. The well-known closure problem is then solved with the aim of best capturing the associated dynamics of the kinetic equation. The issue of physical constraints such as Galilean invariance is addressed and an active learning procedure is introduced to help ensure that the dataset used is representative enough. The reduced system takes the form of a conventional moment system and works regardless of the numerical discretization used. Numerical results are presented for the BGK (Bhatnagar--Gross--Krook) model and binary collision of Maxwell molecules. We demonstrate that the reduced model achieves a uniform accuracy in a wide range of Knudsen numbers spanning from the hydrodynamic limit to free molecular flow."
                    },
                    {
                        "title": "Solving high-dimensional eigenvalue problems using deep neural networks: A diffusion Monte Carlo like approach",
                        "abstract": "We propose a new method to solve eigenvalue problems for linear and semilinear second order differential operators in high dimensions based on deep neural networks. The eigenvalue problem is reformulated as a fixed point problem of the semigroup flow induced by the operator, whose solution can be represented by Feynman-Kac formula in terms of forward-backward stochastic differential equations. The method shares a similar spirit with diffusion Monte Carlo but augments a direct approximation to the eigenfunction through neural-network ansatz. The criterion of fixed point provides a natural loss function to search for parameters via optimization. Our approach is able to provide accurate eigenvalue and eigenfunction approximations in several numerical examples, including Fokker-Planck operator and the linear and nonlinear Schr\\\"odinger operators in high dimensions."
                    },
                    {
                        "title": "Integrating Machine Learning with Physics-Based Modeling",
                        "abstract": "Machine learning is poised as a very powerful tool that can drastically improve our ability to carry out scientific research. However, many issues need to be addressed before this becomes a reality. This article focuses on one particular issue of broad interest: How can we integrate machine learning with physics-based modeling to develop new interpretable and truly reliable physical models? After introducing the general guidelines, we discuss the two most important issues for developing machine learning-based physical models: Imposing physical constraints and obtaining optimal datasets. We also provide a simple and intuitive explanation for the fundamental reasons behind the success of modern machine learning, as well as an introduction to the concurrent machine learning framework needed for integrating machine learning with physics-based modeling. Molecular dynamics and moment closure of kinetic equations are used as examples to illustrate the main issues discussed. We end with a general discussion on where this integration will lead us to, and where the new frontier will be after machine learning is successfully integrated into scientific modeling."
                    },
                    {
                        "title": "An $L^2$ Analysis of Reinforcement Learning in High Dimensions with Kernel and Neural Network Approximation",
                        "abstract": "Reinforcement learning (RL) algorithms based on high-dimensional function approximation have achieved tremendous empirical success in large-scale problems with an enormous number of states. However, most analysis of such algorithms gives rise to error bounds that involve either the number of states or the number of features. This paper considers the situation where the function approximation is made either using the kernel method or the two-layer neural network model, in the context of a fitted Q-iteration algorithm with explicit regularization. We establish an $\\tilde{O}(H^3|\\mathcal {A}|^{\\frac14}n^{-\\frac14})$ bound for the optimal policy with $Hn$ samples, where $H$ is the length of each episode and $|\\mathcal {A}|$ is the size of action space. Our analysis hinges on analyzing the $L^2$ error of the approximated Q-function using $n$ data points. Even though this result still requires a finite-sized action space, the error bound is independent of the dimensionality of the state space."
                    },
                    {
                        "title": "From microscopic theory to macroscopic theory: a systematic study on static modeling for liquid crystals",
                        "abstract": "In this paper, we propose a systematic way of liquid crystal modeling to build connection between microscopic theory and macroscopic theory. A new Q-tensor theory based on Onsager's molecular theory which leads to liquid crystals with certain shape has been proposed. Making uniaxial assumption, we can recover the Oseen-Frank theory from the derived $Q$-tensor theory, and the Oseen-Frank model coefficients can be examined. In addition, the smectic-A phase can also be characterized by the derived macroscopic model."
                    },
                    {
                        "title": "DeePMD-kit: A deep learning package for many-body potential energy representation and molecular dynamics",
                        "abstract": "Recent developments in many-body potential energy representation via deep learning have brought new hopes to addressing the accuracy-versus-efficiency dilemma in molecular simulations. Here we describe DeePMD-kit, a package written in Python/C++ that has been designed to minimize the effort required to build deep learning based representation of potential energy and force field and to perform molecular dynamics. Potential applications of DeePMD-kit span from finite molecules to extended systems and from metallic systems to chemically bonded systems. DeePMD-kit is interfaced with TensorFlow, one of the most popular deep learning frameworks, making the training process highly automatic and efficient. On the other end, DeePMD-kit is interfaced with high-performance classical molecular dynamics and quantum (path-integral) molecular dynamics packages, i.e., LAMMPS and the i-PI, respectively. Thus, upon training, the potential energy and force field models can be used to perform efficient molecular simulations for different purposes. As an example of the many potential applications of the package, we use DeePMD-kit to learn the interatomic potential energy and forces of a water model using data obtained from density functional theory. We demonstrate that the resulted molecular dynamics model reproduces accurately the structural information contained in the original model."
                    }
                ]
            },
            "babcade6-e395-4322-8cd6-c2c647fd552b": {
                "pk": "babcade6-e395-4322-8cd6-c2c647fd552b",
                "name": "Brandon Amos",
                "collaborators": [
                    "J. Zico Kolter",
                    "Samuel Cohen",
                    "Denis Yarats",
                    "Rajiv Sambharya",
                    "Georgina Hall",
                    "Bartolomeo Stellato",
                    "Ricky T. Q. Chen",
                    "Maximilian Nickel",
                    "Shobha Venkataraman",
                    "Yaron Lipman"
                ],
                "domain": [
                    "Optimization",
                    "Reinforcement Learning",
                    "Neural Networks",
                    "Amortized Learning"
                ],
                "publications": [
                    {
                        "title": "Tutorial on amortized optimization",
                        "abstract": "Optimization is a ubiquitous modeling tool and is often deployed in settings which repeatedly solve similar instances of the same problem. Amortized optimization methods use learning to predict the solutions to problems in these settings, exploiting the shared structure between similar problem instances. These methods have been crucial in variational inference and reinforcement learning and are capable of solving optimization problems many orders of magnitudes times faster than traditional optimization methods that do not use amortization. This tutorial presents an introduction to the amortized optimization foundations behind these advancements and overviews their applications in variational inference, sparse coding, gradient-based meta-learning, control, reinforcement learning, convex optimization, optimal transport, and deep equilibrium networks. The source code for this tutorial is available at https://github.com/facebookresearch/amortized-optimization-tutorial."
                    },
                    {
                        "title": "On amortizing convex conjugates for optimal transport",
                        "abstract": "This paper focuses on computing the convex conjugate operation that arises when solving Euclidean Wasserstein-2 optimal transport problems. This conjugation, which is also referred to as the Legendre-Fenchel conjugate or c-transform,is considered difficult to compute and in practice,Wasserstein-2 methods are limited by not being able to exactly conjugate the dual potentials in continuous space. To overcome this, the computation of the conjugate can be approximated with amortized optimization, which learns a model to predict the conjugate. I show that combining amortized approximations to the conjugate with a solver for fine-tuning significantly improves the quality of transport maps learned for the Wasserstein-2 benchmark by Korotin et al. (2021a) and is able to model many 2-dimensional couplings and flows considered in the literature. All of the baselines, methods, and solvers in this paper are available at http://github.com/facebookresearch/w2ot."
                    },
                    {
                        "title": "The Differentiable Cross-Entropy Method",
                        "abstract": "We study the cross-entropy method (CEM) for the non-convex optimization of a continuous and parameterized objective function and introduce a differentiable variant that enables us to differentiate the output of CEM with respect to the objective function's parameters. In the machine learning setting this brings CEM inside of the end-to-end learning pipeline where this has otherwise been impossible. We show applications in a synthetic energy-based structured prediction task and in non-convex continuous control. In the control setting we show how to embed optimal action sequences into a lower-dimensional space. DCEM enables us to fine-tune CEM-based controllers with policy optimization."
                    },
                    {
                        "title": "Neural Fixed-Point Acceleration for Convex Optimization",
                        "abstract": "Fixed-point iterations are at the heart of numerical computing and are often a computational bottleneck in real-time applications that typically need a fast solution of moderate accuracy. We present neural fixed-point acceleration which combines ideas from meta-learning and classical acceleration methods to automatically learn to accelerate fixed-point problems that are drawn from a distribution. We apply our framework to SCS, the state-of-the-art solver for convex cone programming, and design models and loss functions to overcome the challenges of learning over unrolled optimization and acceleration instabilities. Our work brings neural acceleration into any optimization problem expressible with CVXPY. The source code behind this paper is available at https://github.com/facebookresearch/neural-scs"
                    },
                    {
                        "title": "OptNet: Differentiable Optimization as a Layer in Neural Networks",
                        "abstract": "This paper presents OptNet, a network architecture that integrates optimization problems (here, specifically in the form of quadratic programs) as individual layers in larger end-to-end trainable deep networks. These layers encode constraints and complex dependencies between the hidden states that traditional convolutional and fully-connected layers often cannot capture. We explore the foundations for such an architecture: we show how techniques from sensitivity analysis, bilevel optimization, and implicit differentiation can be used to exactly differentiate through these layers and with respect to layer parameters; we develop a highly efficient solver for these layers that exploits fast GPU-based batch solves within a primal-dual interior point method, and which provides backpropagation gradients with virtually no additional cost on top of the solve; and we highlight the application of these approaches in several problems. In one notable example, the method is learns to play mini-Sudoku (4x4) given just input and output games, with no a-priori information about the rules of the game; this highlights the ability of OptNet to learn hard constraints better than other neural architectures."
                    },
                    {
                        "title": "Riemannian Convex Potential Maps",
                        "abstract": "Modeling distributions on Riemannian manifolds is a crucial component in understanding non-Euclidean data that arises, e.g., in physics and geology. The budding approaches in this space are limited by representational and computational tradeoffs. We propose and study a class of flows that uses convex potentials from Riemannian optimal transport. These are universal and can model distributions on any compact Riemannian manifold without requiring domain knowledge of the manifold to be integrated into the architecture. We demonstrate that these flows can model standard distributions on spheres, and tori, on synthetic and geological data. Our source code is freely available online at http://github.com/facebookresearch/rcpm"
                    },
                    {
                        "title": "Input Convex Neural Networks",
                        "abstract": "This paper presents the input convex neural network architecture. These are scalar-valued (potentially deep) neural networks with constraints on the network parameters such that the output of the network is a convex function of (some of) the inputs. The networks allow for efficient inference via optimization over some inputs to the network given others, and can be applied to settings including structured prediction, data imputation, reinforcement learning, and others. In this paper we lay the basic groundwork for these models, proposing methods for inference, optimization and learning, and analyze their representational power. We show that many existing neural network architectures can be made input-convex with a minor modification, and develop specialized optimization algorithms tailored to this setting. Finally, we highlight the performance of the methods on multi-label prediction, image completion, and reinforcement learning problems, where we show improvement over the existing state of the art in many cases."
                    },
                    {
                        "title": "The Limited Multi-Label Projection Layer",
                        "abstract": "We propose the Limited Multi-Label (LML) projection layer as a new primitive operation for end-to-end learning systems. The LML layer provides a probabilistic way of modeling multi-label predictions limited to having exactly k labels. We derive efficient forward and backward passes for this layer and show how the layer can be used to optimize the top-k recall for multi-label tasks with incomplete label information. We evaluate LML layers on top-k CIFAR-100 classification and scene graph generation. We show that LML layers add a negligible amount of computational overhead, strictly improve the model's representational capacity, and improve accuracy. We also revisit the truncated top-k entropy method as a competitive baseline for top-k classification."
                    },
                    {
                        "title": "Meta Optimal Transport",
                        "abstract": "We study the use of amortized optimization to predict optimal transport (OT) maps from the input measures, which we call Meta OT. This helps repeatedly solve similar OT problems between different measures by leveraging the knowledge and information present from past problems to rapidly predict and solve new problems. Otherwise, standard methods ignore the knowledge of the past solutions and suboptimally re-solve each problem from scratch. We instantiate Meta OT models in discrete and continuous settings between grayscale images, spherical data, classification labels, and color palettes and use them to improve the computational time of standard OT solvers. Our source code is available at http://github.com/facebookresearch/meta-ot"
                    },
                    {
                        "title": "Depth-Limited Solving for Imperfect-Information Games",
                        "abstract": "A fundamental challenge in imperfect-information games is that states do not have well-defined values. As a result, depth-limited search algorithms used in single-agent settings and perfect-information games do not apply. This paper introduces a principled way to conduct depth-limited solving in imperfect-information games by allowing the opponent to choose among a number of strategies for the remainder of the game at the depth limit. Each one of these strategies results in a different set of values for leaf nodes. This forces an agent to be robust to the different strategies an opponent may employ. We demonstrate the effectiveness of this approach by building a master-level heads-up no-limit Texas hold'em poker AI that defeats two prior top agents using only a 4-core CPU and 16 GB of memory. Developing such a powerful agent would have previously required a supercomputer."
                    },
                    {
                        "title": "Cross-Domain Imitation Learning via Optimal Transport",
                        "abstract": "Cross-domain imitation learning studies how to leverage expert demonstrations of one agent to train an imitation agent with a different embodiment or morphology. Comparing trajectories and stationary distributions between the expert and imitation agents is challenging because they live on different systems that may not even have the same dimensionality. We propose Gromov-Wasserstein Imitation Learning (GWIL), a method for cross-domain imitation that uses the Gromov-Wasserstein distance to align and compare states between the different spaces of the agents. Our theory formally characterizes the scenarios where GWIL preserves optimality, revealing its possibilities and limitations. We demonstrate the effectiveness of GWIL in non-trivial continuous control domains ranging from simple rigid transformation of the expert domain to arbitrary transformation of the state-action space."
                    },
                    {
                        "title": "Input Convex Gradient Networks",
                        "abstract": "The gradients of convex functions are expressive models of non-trivial vector fields. For example, Brenier's theorem yields that the optimal transport map between any two measures on Euclidean space under the squared distance is realized as a convex gradient, which is a key insight used in recent generative flow models. In this paper, we study how to model convex gradients by integrating a Jacobian-vector product parameterized by a neural network, which we call the Input Convex Gradient Network (ICGN). We theoretically study ICGNs and compare them to taking the gradient of an Input-Convex Neural Network (ICNN), empirically demonstrating that a single layer ICGN can fit a toy example better than a single layer ICNN. Lastly, we explore extensions to deeper networks and connections to constructions from Riemannian geometry."
                    },
                    {
                        "title": "End-to-End Learning to Warm-Start for Real-Time Quadratic Optimization",
                        "abstract": "First-order methods are widely used to solve convex quadratic programs (QPs) in real-time applications because of their low per-iteration cost. However, they can suffer from slow convergence to accurate solutions. In this paper, we present a framework which learns an effective warm-start for a popular first-order method in real-time applications, Douglas-Rachford (DR) splitting, across a family of parametric QPs. This framework consists of two modules: a feedforward neural network block, which takes as input the parameters of the QP and outputs a warm-start, and a block which performs a fixed number of iterations of DR splitting from this warm-start and outputs a candidate solution. A key feature of our framework is its ability to do end-to-end learning as we differentiate through the DR iterations. To illustrate the effectiveness of our method, we provide generalization bounds (based on Rademacher complexity) that improve with the number of training problems and number of iterations simultaneously. We further apply our method to three real-time applications and observe that, by learning good warm-starts, we are able to significantly reduce the number of iterations required to obtain high-quality solutions."
                    },
                    {
                        "title": "Learning to Warm-Start Fixed-Point Optimization Algorithms",
                        "abstract": "We introduce a machine-learning framework to warm-start fixed-point optimization algorithms. Our architecture consists of a neural network mapping problem parameters to warm starts, followed by a predefined number of fixed-point iterations. We propose two loss functions designed to either minimize the fixed-point residual or the distance to a ground truth solution. In this way, the neural network predicts warm starts with the end-to-end goal of minimizing the downstream loss. An important feature of our architecture is its flexibility, in that it can predict a warm start for fixed-point algorithms run for any number of steps, without being limited to the number of steps it has been trained on. We provide PAC-Bayes generalization bounds on unseen data for common classes of fixed-point operators: contractive, linearly convergent, and averaged. Applying this framework to well-known applications in control, statistics, and signal processing, we observe a significant reduction in the number of iterations and solution time required to solve these problems, through learned warm starts."
                    },
                    {
                        "title": "Semi-Supervised Offline Reinforcement Learning with Action-Free Trajectories",
                        "abstract": "Natural agents can effectively learn from multiple data sources that differ in size, quality, and types of measurements. We study this heterogeneity in the context of offline reinforcement learning (RL) by introducing a new, practically motivated semi-supervised setting. Here, an agent has access to two sets of trajectories: labelled trajectories containing state, action and reward triplets at every timestep, along with unlabelled trajectories that contain only state and reward information. For this setting, we develop and study a simple meta-algorithmic pipeline that learns an inverse dynamics model on the labelled data to obtain proxy-labels for the unlabelled data, followed by the use of any offline RL algorithm on the true and proxy-labelled trajectories. Empirically, we find this simple pipeline to be highly successful -- on several D4RL benchmarks~\\cite{fu2020d4rl}, certain offline RL algorithms can match the performance of variants trained on a fully labelled dataset even when we label only 10\\% of trajectories which are highly suboptimal. To strengthen our understanding, we perform a large-scale controlled empirical study investigating the interplay of data-centric properties of the labelled and unlabelled datasets, with algorithmic design choices (e.g., choice of inverse dynamics, offline RL algorithm) to identify general trends and best practices for training RL agents on semi-supervised offline datasets."
                    },
                    {
                        "title": "On the model-based stochastic value gradient for continuous reinforcement learning",
                        "abstract": "For over a decade, model-based reinforcement learning has been seen as a way to leverage control-based domain knowledge to improve the sample-efficiency of reinforcement learning agents. While model-based agents are conceptually appealing, their policies tend to lag behind those of model-free agents in terms of final reward, especially in non-trivial environments. In response, researchers have proposed model-based agents with increasingly complex components, from ensembles of probabilistic dynamics models, to heuristics for mitigating model error. In a reversal of this trend, we show that simple model-based agents can be derived from existing ideas that not only match, but outperform state-of-the-art model-free agents in terms of both sample-efficiency and final reward. We find that a model-free soft value estimate for policy evaluation and a model-based stochastic value gradient for policy improvement is an effective combination, achieving state-of-the-art results on a high-dimensional humanoid control task, which most model-based agents are unable to solve. Our findings suggest that model-based policy evaluation deserves closer attention."
                    },
                    {
                        "title": "Objective Mismatch in Model-based Reinforcement Learning",
                        "abstract": "Model-based reinforcement learning (MBRL) has been shown to be a powerful framework for data-efficiently learning control of continuous tasks. Recent work in MBRL has mostly focused on using more advanced function approximators and planning schemes, with little development of the general framework. In this paper, we identify a fundamental issue of the standard MBRL framework -- what we call the objective mismatch issue. Objective mismatch arises when one objective is optimized in the hope that a second, often uncorrelated, metric will also be optimized. In the context of MBRL, we characterize the objective mismatch between training the forward dynamics model w.r.t.~the likelihood of the one-step ahead prediction, and the overall goal of improving performance on a downstream control task. For example, this issue can emerge with the realization that dynamics models effective for a specific task do not necessarily need to be globally accurate, and vice versa globally accurate models might not be sufficiently accurate locally to obtain good control performance on a specific task. In our experiments, we study this objective mismatch issue and demonstrate that the likelihood of one-step ahead predictions is not always correlated with control performance. This observation highlights a critical limitation in the MBRL framework which will require further research to be fully understood and addressed. We propose an initial method to mitigate the mismatch issue by re-weighting dynamics model training. Building on it, we conclude with a discussion about other potential directions of research for addressing this issue."
                    },
                    {
                        "title": "Task-based End-to-end Model Learning in Stochastic Optimization",
                        "abstract": "With the increasing popularity of machine learning techniques, it has become common to see prediction algorithms operating within some larger process. However, the criteria by which we train these algorithms often differ from the ultimate criteria on which we evaluate them. This paper proposes an end-to-end approach for learning probabilistic machine learning models in a manner that directly captures the ultimate task-based objective for which they will be used, within the context of stochastic programming. We present three experimental evaluations of the proposed approach: a classical inventory stock problem, a real-world electrical grid scheduling task, and a real-world energy storage arbitrage task. We show that the proposed approach can outperform both traditional modeling and purely black-box policy optimization approaches in these applications."
                    },
                    {
                        "title": "Learning Neural Event Functions for Ordinary Differential Equations",
                        "abstract": "The existing Neural ODE formulation relies on an explicit knowledge of the termination time. We extend Neural ODEs to implicitly defined termination criteria modeled by neural event functions, which can be chained together and differentiated through. Neural Event ODEs are capable of modeling discrete and instantaneous changes in a continuous-time system, without prior knowledge of when these changes should occur or how many such changes should exist. We test our approach in modeling hybrid discrete- and continuous- systems such as switching dynamical systems and collision in multi-body systems, and we propose simulation-based training of point processes with applications in discrete control."
                    },
                    {
                        "title": "Neural Spatio-Temporal Point Processes",
                        "abstract": "We propose a new class of parameterizations for spatio-temporal point processes which leverage Neural ODEs as a computational method and enable flexible, high-fidelity models of discrete events that are localized in continuous time and space. Central to our approach is a combination of continuous-time neural networks with two novel neural architectures, i.e., Jump and Attentive Continuous-time Normalizing Flows. This approach allows us to learn complex distributions for both the spatial and temporal domain and to condition non-trivially on the observed event history. We validate our models on data sets from a wide variety of contexts such as seismology, epidemiology, urban mobility, and neuroscience."
                    }
                ]
            },
            "482f5f59-e947-4d97-a520-6c959b2b8b79": {
                "pk": "482f5f59-e947-4d97-a520-6c959b2b8b79",
                "name": "Joan Bruna",
                "collaborators": [
                    "St\u00e9phane Mallat",
                    "Aaron Zweig",
                    "Arthur Szlam",
                    "Stephane Mallat",
                    "Jiequn Han",
                    "Victor Garcia",
                    "Yann LeCun",
                    "David Folqu\u00e9",
                    "Sainbayar Sukhbaatar",
                    "Shariq Mobin"
                ],
                "domain": [
                    "Signal Processing",
                    "Machine Learning",
                    "Graph Neural Network",
                    "Audio Synthesis"
                ],
                "publications": [
                    {
                        "title": "Geometric Models with Co-occurrence Groups",
                        "abstract": "A geometric model of sparse signal representations is introduced for classes of signals. It is computed by optimizing co-occurrence groups with a maximum likelihood estimate calculated with a Bernoulli mixture model. Applications to face image compression and MNIST digit classification illustrate the applicability of this model."
                    },
                    {
                        "title": "Classification with Scattering Operators",
                        "abstract": "A scattering vector is a local descriptor including multiscale and multi-direction co-occurrence information. It is computed with a cascade of wavelet decompositions and complex modulus. This scattering representation is locally translation invariant and linearizes deformations. A supervised classification algorithm is computed with a PCA model selection on scattering vectors. State of the art results are obtained for handwritten digit recognition and texture classification."
                    },
                    {
                        "title": "Classification with Invariant Scattering Representations",
                        "abstract": "A scattering transform defines a signal representation which is invariant to translations and Lipschitz continuous relatively to deformations. It is implemented with a non-linear convolution network that iterates over wavelet and modulus operators. Lipschitz continuity locally linearizes deformations. Complex classes of signals and textures can be modeled with low-dimensional affine spaces, computed with a PCA in the scattering domain. Classification is performed with a penalized model selection. State of the art results are obtained for handwritten digit recognition over small training sets, and for texture classification."
                    },
                    {
                        "title": "Audio Texture Synthesis with Scattering Moments",
                        "abstract": "We introduce an audio texture synthesis algorithm based on scattering moments. A scattering transform is computed by iteratively decomposing a signal with complex wavelet filter banks and computing their amplitude envelop. Scattering moments provide general representations of stationary processes computed as expected values of scattering coefficients. They are estimated with low variance estimators from single realizations. Audio signals having prescribed scattering moments are synthesized with a gradient descent algorithms. Audio synthesis examples show that scattering representation provide good synthesis of audio textures with much fewer coefficients than the state of the art."
                    },
                    {
                        "title": "Invariant Scattering Convolution Networks",
                        "abstract": "A wavelet scattering network computes a translation invariant image representation, which is stable to deformations and preserves high frequency information for classification. It cascades wavelet transform convolutions with non-linear modulus and averaging operators. The first network layer outputs SIFT-type descriptors whereas the next layers provide complementary invariant information which improves classification. The mathematical analysis of wavelet scattering networks explains important properties of deep convolution networks for classification.   A scattering representation of stationary processes incorporates higher order moments and can thus discriminate textures having the same Fourier power spectrum. State of the art classification results are obtained for handwritten digits and texture discrimination, using a Gaussian kernel SVM and a generative PCA classifier."
                    },
                    {
                        "title": "Provably Efficient Third-Person Imitation from Offline Observation",
                        "abstract": "Domain adaptation in imitation learning represents an essential step towards improving generalizability. However, even in the restricted setting of third-person imitation where transfer is between isomorphic Markov Decision Processes, there are no strong guarantees on the performance of transferred policies. We present problem-dependent, statistical learning guarantees for third-person imitation from observation in an offline setting, and a lower bound on performance in the online setting."
                    },
                    {
                        "title": "Multiscale Sparse Microcanonical Models",
                        "abstract": "We study approximations of non-Gaussian stationary processes having long range correlations with microcanonical models. These models are conditioned by the empirical value of an energy vector, evaluated on a single realization. Asymptotic properties of maximum entropy microcanonical and macrocanonical processes and their convergence to Gibbs measures are reviewed. We show that the Jacobian of the energy vector controls the entropy rate of microcanonical processes.   Sampling maximum entropy processes through MCMC algorithms require too many operations when the number of constraints is large. We define microcanonical gradient descent processes by transporting a maximum entropy measure with a gradient descent algorithm which enforces the energy conditions. Convergence and symmetries are analyzed. Approximations of non-Gaussian processes with long range interactions are defined with multiscale energy vectors computed with wavelet and scattering transforms. Sparsity properties are captured with $\\bf l^1$ norms. Approximations of Gaussian, Ising and point processes are studied, as well as image and audio texture synthesis."
                    },
                    {
                        "title": "Posterior Sampling with Denoising Oracles via Tilted Transport",
                        "abstract": "Score-based diffusion models have significantly advanced high-dimensional data generation across various domains, by learning a denoising oracle (or score) from datasets. From a Bayesian perspective, they offer a realistic modeling of data priors and facilitate solving inverse problems through posterior sampling. Although many heuristic methods have been developed recently for this purpose, they lack the quantitative guarantees needed in many scientific applications.   In this work, we introduce the \\textit{tilted transport} technique, which leverages the quadratic structure of the log-likelihood in linear inverse problems in combination with the prior denoising oracle to transform the original posterior sampling problem into a new `boosted' posterior that is provably easier to sample from. We quantify the conditions under which this boosted posterior is strongly log-concave, highlighting the dependencies on the condition number of the measurement matrix and the signal-to-noise ratio. The resulting posterior sampling scheme is shown to reach the computational threshold predicted for sampling Ising models [Kunisky'23] with a direct analysis, and is further validated on high-dimensional Gaussian mixture models and scalar field $\\varphi^4$ models."
                    },
                    {
                        "title": "Few-Shot Learning with Graph Neural Networks",
                        "abstract": "We propose to study the problem of few-shot learning with the prism of inference on a partially observed graphical model, constructed from a collection of input images whose label can be either observed or not. By assimilating generic message-passing inference algorithms with their neural-network counterparts, we define a graph neural network architecture that generalizes several of the recently proposed few-shot learning models. Besides providing improved numerical performance, our framework is easily extended to variants of few-shot learning, such as semi-supervised or active learning, demonstrating the ability of graph-based models to operate well on 'relational' tasks."
                    },
                    {
                        "title": "Exponential Separations in Symmetric Neural Networks",
                        "abstract": "In this work we demonstrate a novel separation between symmetric neural network architectures. Specifically, we consider the Relational Network~\\parencite{santoro2017simple} architecture as a natural generalization of the DeepSets~\\parencite{zaheer2017deep} architecture, and study their representational gap. Under the restriction to analytic activation functions, we construct a symmetric function acting on sets of size $N$ with elements in dimension $D$, which can be efficiently approximated by the former architecture, but provably requires width exponential in $N$ and $D$ for the latter."
                    },
                    {
                        "title": "Signal Recovery from Pooling Representations",
                        "abstract": "In this work we compute lower Lipschitz bounds of $\\ell_p$ pooling operators for $p=1, 2, \\infty$ as well as $\\ell_p$ pooling operators preceded by half-rectification layers. These give sufficient conditions for the design of invertible neural network layers. Numerical experiments on MNIST and image patches confirm that pooling layers can be inverted with phase recovery algorithms. Moreover, the regularity of the inverse pooling, controlled by the lower Lipschitz constant, is empirically verified with a nearest neighbor regression."
                    },
                    {
                        "title": "Planning with Arithmetic and Geometric Attributes",
                        "abstract": "A desirable property of an intelligent agent is its ability to understand its environment to quickly generalize to novel tasks and compose simpler tasks into more complex ones. If the environment has geometric or arithmetic structure, the agent should exploit these for faster generalization. Building on recent work that augments the environment with user-specified attributes, we show that further equipping these attributes with the appropriate geometric and arithmetic structure brings substantial gains in sample complexity."
                    },
                    {
                        "title": "Voice Conversion using Convolutional Neural Networks",
                        "abstract": "The human auditory system is able to distinguish the vocal source of thousands of speakers, yet not much is known about what features the auditory system uses to do this. Fourier Transforms are capable of capturing the pitch and harmonic structure of the speaker but this alone proves insufficient at identifying speakers uniquely. The remaining structure, often referred to as timbre, is critical to identifying speakers but we understood little about it. In this paper we use recent advances in neural networks in order to manipulate the voice of one speaker into another by transforming not only the pitch of the speaker, but the timbre. We review generative models built with neural networks as well as architectures for creating neural networks that learn analogies. Our preliminary results converting voices from one speaker to another are encouraging."
                    },
                    {
                        "title": "A Functional Perspective on Learning Symmetric Functions with Neural Networks",
                        "abstract": "Symmetric functions, which take as input an unordered, fixed-size set, are known to be universally representable by neural networks that enforce permutation invariance. These architectures only give guarantees for fixed input sizes, yet in many practical applications, including point clouds and particle physics, a relevant notion of generalization should include varying the input size. In this work we treat symmetric functions (of any size) as functions over probability measures, and study the learning and representation of neural networks defined on measures. By focusing on shallow architectures, we establish approximation and generalization bounds under different choices of regularization (such as RKHS and variation norms), that capture a hierarchy of functional spaces with increasing degree of non-linear learning. The resulting models can be learned efficiently and enjoy generalization guarantees that extend across input sizes, as we verify empirically."
                    },
                    {
                        "title": "Approximating Orthogonal Matrices with Effective Givens Factorization",
                        "abstract": "We analyze effective approximation of unitary matrices. In our formulation, a unitary matrix is represented as a product of rotations in two-dimensional subspaces, so-called Givens rotations. Instead of the quadratic dimension dependence when applying a dense matrix, applying such an approximation scales with the number factors, each of which can be implemented efficiently. Consequently, in settings where an approximation is once computed and then applied many times, such a representation becomes advantageous. Although effective Givens factorization is not possible for generic unitary operators, we show that minimizing a sparsity-inducing objective with a coordinate descent algorithm on the unitary group yields good factorizations for structured matrices. Canonical applications of such a setup are orthogonal basis transforms. We demonstrate numerical results of approximating the graph Fourier transform, which is the matrix obtained when diagonalizing a graph Laplacian."
                    },
                    {
                        "title": "Multi-fidelity Stability for Graph Representation Learning",
                        "abstract": "In the problem of structured prediction with graph representation learning (GRL for short), the hypothesis returned by the algorithm maps the set of features in the \\emph{receptive field} of the targeted vertex to its label. To understand the learnability of those algorithms, we introduce a weaker form of uniform stability termed \\emph{multi-fidelity stability} and give learning guarantees for weakly dependent graphs. We testify that ~\\citet{london2016stability}'s claim on the generalization of a single sample holds for GRL when the receptive field is sparse. In addition, we study the stability induced bound for two popular algorithms: \\textbf{(1)} Stochastic gradient descent under convex and non-convex landscape. In this example, we provide non-asymptotic bounds that highly depend on the sparsity of the receptive field constructed by the algorithm. \\textbf{(2)} The constrained regression problem on a 1-layer linear equivariant GNN. In this example, we present lower bounds for the discrepancy between the two types of stability, which justified the multi-fidelity design."
                    },
                    {
                        "title": "Simultaneous Transport Evolution for Minimax Equilibria on Measures",
                        "abstract": "Min-max optimization problems arise in several key machine learning setups, including adversarial learning and generative modeling. In their general form, in absence of convexity/concavity assumptions, finding pure equilibria of the underlying two-player zero-sum game is computationally hard [Daskalakis et al., 2021]. In this work we focus instead in finding mixed equilibria, and consider the associated lifted problem in the space of probability measures. By adding entropic regularization, our main result establishes global convergence towards the global equilibrium by using simultaneous gradient ascent-descent with respect to the Wasserstein metric -- a dynamics that admits efficient particle discretization in high-dimensions, as opposed to entropic mirror descent. We complement this positive result with a related entropy-regularized loss which is not bilinear but still convex-concave in the Wasserstein geometry, and for which simultaneous dynamics do not converge yet timescale separation does. Taken together, these results showcase the benign geometry of bilinear games in the space of measures, enabling particle dynamics with global qualitative convergence guarantees."
                    },
                    {
                        "title": "Towards Antisymmetric Neural Ansatz Separation",
                        "abstract": "We study separations between two fundamental models (or \\emph{Ans\\\"atze}) of antisymmetric functions, that is, functions $f$ of the form $f(x_{\\sigma(1)}, \\ldots, x_{\\sigma(N)}) = \\text{sign}(\\sigma)f(x_1, \\ldots, x_N)$, where $\\sigma$ is any permutation. These arise in the context of quantum chemistry, and are the basic modeling tool for wavefunctions of Fermionic systems. Specifically, we consider two popular antisymmetric Ans\\\"atze: the Slater representation, which leverages the alternating structure of determinants, and the Jastrow ansatz, which augments Slater determinants with a product by an arbitrary symmetric function. We construct an antisymmetric function in $N$ dimensions that can be efficiently expressed in Jastrow form, yet provably cannot be approximated by Slater determinants unless there are exponentially (in $N^2$) many terms. This represents the first explicit quantitative separation between these two Ans\\\"atze."
                    },
                    {
                        "title": "Symmetric Single Index Learning",
                        "abstract": "Few neural architectures lend themselves to provable learning with gradient based methods. One popular model is the single-index model, in which labels are produced by composing an unknown linear projection with a possibly unknown scalar link function. Learning this model with SGD is relatively well-understood, whereby the so-called information exponent of the link function governs a polynomial sample complexity rate. However, extending this analysis to deeper or more complicated architectures remains challenging.   In this work, we consider single index learning in the setting of symmetric neural networks. Under analytic assumptions on the activation and maximum degree assumptions on the link function, we prove that gradient flow recovers the hidden planted direction, represented as a finitely supported vector in the feature space of power sum polynomials. We characterize a notion of information exponent adapted to our setting that controls the efficiency of learning."
                    },
                    {
                        "title": "Understanding Trainable Sparse Coding via Matrix Factorization",
                        "abstract": "Sparse coding is a core building block in many data analysis and machine learning pipelines. Typically it is solved by relying on generic optimization techniques, that are optimal in the class of first-order methods for non-smooth, convex functions, such as the Iterative Soft Thresholding Algorithm and its accelerated version (ISTA, FISTA). However, these methods don't exploit the particular structure of the problem at hand nor the input data distribution. An acceleration using neural networks was proposed in \\cite{Gregor10}, coined LISTA, which showed empirically that one could achieve high quality estimates with few iterations by modifying the parameters of the proximal splitting appropriately.   In this paper we study the reasons for such acceleration. Our mathematical analysis reveals that it is related to a specific matrix factorization of the Gram kernel of the dictionary, which attempts to nearly diagonalise the kernel with a basis that produces a small perturbation of the $\\ell_1$ ball. When this factorization succeeds, we prove that the resulting splitting algorithm enjoys an improved convergence bound with respect to the non-adaptive version. Moreover, our analysis also shows that conditions for acceleration occur mostly at the beginning of the iterative process, consistent with numerical experiments. We further validate our analysis by showing that on dictionaries where this factorization does not exist, adaptive acceleration fails."
                    }
                ]
            },
            "d0f6dff8-67f5-4290-bc0a-3e766d3f8149": {
                "pk": "d0f6dff8-67f5-4290-bc0a-3e766d3f8149",
                "name": "Ricky T. Q. Chen",
                "collaborators": [
                    "David Duvenaud",
                    "Maximilian Nickel",
                    "Brandon Amos",
                    "Xuechen Li",
                    "Yaron Lipman",
                    "Yulia Rubanova",
                    "Will Grathwohl",
                    "Jesse Bettencourt",
                    "Jens Behrmann",
                    "J\u00f6rn-Henrik Jacobsen"
                ],
                "domain": [
                    "Neural Networks",
                    "Generative Models",
                    "Differential Equations",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Neural Networks with Cheap Differential Operators",
                        "abstract": "Gradients of neural networks can be computed efficiently for any architecture, but some applications require differential operators with higher time complexity. We describe a family of restricted neural network architectures that allow efficient computation of a family of differential operators involving dimension-wise derivatives, used in cases such as computing the divergence. Our proposed architecture has a Jacobian matrix composed of diagonal and hollow (non-diagonal) components. We can then modify the backward computation graph to extract dimension-wise derivatives efficiently with automatic differentiation. We demonstrate these cheap differential operators for solving root-finding subproblems in implicit ODE solvers, exact density evaluation for continuous normalizing flows, and evaluating the Fokker--Planck equation for training stochastic differential equation models."
                    },
                    {
                        "title": "Flow Matching on General Geometries",
                        "abstract": "We propose Riemannian Flow Matching (RFM), a simple yet powerful framework for training continuous normalizing flows on manifolds. Existing methods for generative modeling on manifolds either require expensive simulation, are inherently unable to scale to high dimensions, or use approximations for limiting quantities that result in biased training objectives. Riemannian Flow Matching bypasses these limitations and offers several advantages over previous approaches: it is simulation-free on simple geometries, does not require divergence computation, and computes its target vector field in closed-form. The key ingredient behind RFM is the construction of a relatively simple premetric for defining target vector fields, which encompasses the existing Euclidean case. To extend to general geometries, we rely on the use of spectral decompositions to efficiently compute premetrics on the fly. Our method achieves state-of-the-art performance on many real-world non-Euclidean datasets, and we demonstrate tractable training on general geometries, including triangular meshes with highly non-trivial curvature and boundaries."
                    },
                    {
                        "title": "Learning Neural Event Functions for Ordinary Differential Equations",
                        "abstract": "The existing Neural ODE formulation relies on an explicit knowledge of the termination time. We extend Neural ODEs to implicitly defined termination criteria modeled by neural event functions, which can be chained together and differentiated through. Neural Event ODEs are capable of modeling discrete and instantaneous changes in a continuous-time system, without prior knowledge of when these changes should occur or how many such changes should exist. We test our approach in modeling hybrid discrete- and continuous- systems such as switching dynamical systems and collision in multi-body systems, and we propose simulation-based training of point processes with applications in discrete control."
                    },
                    {
                        "title": "Neural Spatio-Temporal Point Processes",
                        "abstract": "We propose a new class of parameterizations for spatio-temporal point processes which leverage Neural ODEs as a computational method and enable flexible, high-fidelity models of discrete events that are localized in continuous time and space. Central to our approach is a combination of continuous-time neural networks with two novel neural architectures, i.e., Jump and Attentive Continuous-time Normalizing Flows. This approach allows us to learn complex distributions for both the spatial and temporal domain and to condition non-trivially on the observed event history. We validate our models on data sets from a wide variety of contexts such as seismology, epidemiology, urban mobility, and neuroscience."
                    },
                    {
                        "title": "Semi-Discrete Normalizing Flows through Differentiable Tessellation",
                        "abstract": "Mapping between discrete and continuous distributions is a difficult task and many have had to resort to heuristical approaches. We propose a tessellation-based approach that directly learns quantization boundaries in a continuous space, complete with exact likelihood evaluations. This is done through constructing normalizing flows on convex polytopes parameterized using a simple homeomorphism with an efficient log determinant Jacobian. We explore this approach in two application settings, mapping from discrete to continuous and vice versa. Firstly, a Voronoi dequantization allows automatically learning quantization boundaries in a multidimensional space. The location of boundaries and distances between regions can encode useful structural relations between the quantized discrete values. Secondly, a Voronoi mixture model has near-constant computation cost for likelihood evaluation regardless of the number of mixture components. Empirically, we show improvements over existing methods across a range of structured data modalities."
                    },
                    {
                        "title": "Neural Conservation Laws: A Divergence-Free Perspective",
                        "abstract": "We investigate the parameterization of deep neural networks that by design satisfy the continuity equation, a fundamental conservation law. This is enabled by the observation that any solution of the continuity equation can be represented as a divergence-free vector field. We hence propose building divergence-free neural networks through the concept of differential forms, and with the aid of automatic differentiation, realize two practical constructions. As a result, we can parameterize pairs of densities and vector fields that always exactly satisfy the continuity equation, foregoing the need for extra penalty methods or expensive numerical simulation. Furthermore, we prove these models are universal and so can be used to represent any divergence-free vector field. Finally, we experimentally validate our approaches by computing neural network-based solutions to fluid equations, solving for the Hodge decomposition, and learning dynamical optimal transport maps."
                    },
                    {
                        "title": "Latent ODEs for Irregularly-Sampled Time Series",
                        "abstract": "Time series with non-uniform intervals occur in many applications, and are difficult to model using standard recurrent neural networks (RNNs). We generalize RNNs to have continuous-time hidden dynamics defined by ordinary differential equations (ODEs), a model we call ODE-RNNs. Furthermore, we use ODE-RNNs to replace the recognition network of the recently-proposed Latent ODE model. Both ODE-RNNs and Latent ODEs can naturally handle arbitrary time gaps between observations, and can explicitly model the probability of observation times using Poisson processes. We show experimentally that these ODE-based models outperform their RNN-based counterparts on irregularly-sampled data."
                    },
                    {
                        "title": "\"Hey, that's not an ODE\": Faster ODE Adjoints via Seminorms",
                        "abstract": "Neural differential equations may be trained by backpropagating gradients via the adjoint method, which is another differential equation typically solved using an adaptive-step-size numerical differential equation solver. A proposed step is accepted if its error, \\emph{relative to some norm}, is sufficiently small; else it is rejected, the step is shrunk, and the process is repeated. Here, we demonstrate that the particular structure of the adjoint equations makes the usual choices of norm (such as $L^2$) unnecessarily stringent. By replacing it with a more appropriate (semi)norm, fewer steps are unnecessarily rejected and the backpropagation is made faster. This requires only minor code modifications. Experiments on a wide range of tasks -- including time series, generative modeling, and physical control -- demonstrate a median improvement of 40% fewer function evaluations. On some problems we see as much as 62% fewer function evaluations, so that the overall training time is roughly halved."
                    },
                    {
                        "title": "Isolating Sources of Disentanglement in Variational Autoencoders",
                        "abstract": "We decompose the evidence lower bound to show the existence of a term measuring the total correlation between latent variables. We use this to motivate our $\\beta$-TCVAE (Total Correlation Variational Autoencoder), a refinement of the state-of-the-art $\\beta$-VAE objective for learning disentangled representations, requiring no additional hyperparameters during training. We further propose a principled classifier-free measure of disentanglement called the mutual information gap (MIG). We perform extensive quantitative and qualitative experiments, in both restricted and non-restricted settings, and show a strong relation between total correlation and disentanglement, when the latent variables model is trained using our framework."
                    },
                    {
                        "title": "FFJORD: Free-form Continuous Dynamics for Scalable Reversible Generative Models",
                        "abstract": "A promising class of generative models maps points from a simple distribution to a complex distribution through an invertible neural network. Likelihood-based training of these models requires restricting their architectures to allow cheap computation of Jacobian determinants. Alternatively, the Jacobian trace can be used if the transformation is specified by an ordinary differential equation. In this paper, we use Hutchinson's trace estimator to give a scalable unbiased estimate of the log-density. The result is a continuous-time invertible generative model with unbiased density estimation and one-pass sampling, while allowing unrestricted neural network architectures. We demonstrate our approach on high-dimensional density estimation, image generation, and variational inference, achieving the state-of-the-art among exact likelihood methods with efficient sampling."
                    },
                    {
                        "title": "Residual Flows for Invertible Generative Modeling",
                        "abstract": "Flow-based generative models parameterize probability distributions through an invertible transformation and can be trained by maximum likelihood. Invertible residual networks provide a flexible family of transformations where only Lipschitz conditions rather than strict architectural constraints are needed for enforcing invertibility. However, prior work trained invertible residual networks for density estimation by relying on biased log-density estimates whose bias increased with the network's expressiveness. We give a tractable unbiased estimate of the log density using a \"Russian roulette\" estimator, and reduce the memory required during training by using an alternative infinite series for the gradient. Furthermore, we improve invertible residual blocks by proposing the use of activation functions that avoid derivative saturation and generalizing the Lipschitz condition to induced mixed norms. The resulting approach, called Residual Flows, achieves state-of-the-art performance on density estimation amongst flow-based models, and outperforms networks that use coupling blocks at joint generative and discriminative modeling."
                    },
                    {
                        "title": "Self-Tuning Stochastic Optimization with Curvature-Aware Gradient Filtering",
                        "abstract": "Standard first-order stochastic optimization algorithms base their updates solely on the average mini-batch gradient, and it has been shown that tracking additional quantities such as the curvature can help de-sensitize common hyperparameters. Based on this intuition, we explore the use of exact per-sample Hessian-vector products and gradients to construct optimizers that are self-tuning and hyperparameter-free. Based on a dynamics model of the gradient, we derive a process which leads to a curvature-corrected, noise-adaptive online gradient estimate. The smoothness of our updates makes it more amenable to simple step size selection schemes, which we also base off of our estimates quantities. We prove that our model-based procedure converges in the noisy quadratic setting. Though we do not see similar gains in deep learning tasks, we can match the performance of well-tuned optimizers and ultimately, this is an interesting step for constructing self-tuning optimizers."
                    },
                    {
                        "title": "Latent Discretization for Continuous-time Sequence Compression",
                        "abstract": "Neural compression offers a domain-agnostic approach to creating codecs for lossy or lossless compression via deep generative models. For sequence compression, however, most deep sequence models have costs that scale with the sequence length rather than the sequence complexity. In this work, we instead treat data sequences as observations from an underlying continuous-time process and learn how to efficiently discretize while retaining information about the full sequence. As a consequence of decoupling sequential information from its temporal discretization, our approach allows for greater compression rates and smaller computational complexity. Moreover, the continuous-time approach naturally allows us to decode at different time intervals. We empirically verify our approach on multiple domains involving compression of video and motion capture sequences, showing that our approaches can automatically achieve reductions in bit rates by learning how to discretize."
                    },
                    {
                        "title": "Neural Ordinary Differential Equations",
                        "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                    },
                    {
                        "title": "Invertible Residual Networks",
                        "abstract": "We show that standard ResNet architectures can be made invertible, allowing the same model to be used for classification, density estimation, and generation. Typically, enforcing invertibility requires partitioning dimensions or restricting network architectures. In contrast, our approach only requires adding a simple normalization step during training, already available in standard frameworks. Invertible ResNets define a generative model which can be trained by maximum likelihood on unlabeled data. To compute likelihoods, we introduce a tractable approximation to the Jacobian log-determinant of a residual block. Our empirical evaluation shows that invertible ResNets perform competitively with both state-of-the-art image classifiers and flow-based generative models, something that has not been previously achieved with a single architecture."
                    },
                    {
                        "title": "Scalable Gradients for Stochastic Differential Equations",
                        "abstract": "The adjoint sensitivity method scalably computes gradients of solutions to ordinary differential equations. We generalize this method to stochastic differential equations, allowing time-efficient and constant-memory computation of gradients with high-order adaptive solvers. Specifically, we derive a stochastic differential equation whose solution is the gradient, a memory-efficient algorithm for caching noise, and conditions under which numerical solutions converge. In addition, we combine our method with gradient-based stochastic variational inference for latent stochastic differential equations. We use our method to fit stochastic dynamics defined by neural networks, achieving competitive performance on a 50-dimensional motion capture dataset."
                    },
                    {
                        "title": "Infinitely Deep Bayesian Neural Networks with Stochastic Differential Equations",
                        "abstract": "We perform scalable approximate inference in continuous-depth Bayesian neural networks. In this model class, uncertainty about separate weights in each layer gives hidden units that follow a stochastic differential equation. We demonstrate gradient-based stochastic variational inference in this infinite-parameter setting, producing arbitrarily-flexible approximate posteriors. We also derive a novel gradient estimator that approaches zero variance as the approximate posterior over weights approaches the true posterior. This approach brings continuous-depth Bayesian neural nets to a competitive comparison against discrete-depth alternatives, while inheriting the memory-efficient training and tunable precision of Neural ODEs."
                    },
                    {
                        "title": "Inverse design of dissipative quantum steady-states with implicit differentiation",
                        "abstract": "Inverse design of a property that depends on the steady-state of an open quantum system is commonly done by grid-search type of methods. In this paper we present a new methodology that allows us to compute the gradient of the steady-state of an open quantum system with respect to any parameter of the Hamiltonian using the implicit differentiation theorem. As an example, we present a simulation of a spin-boson model where the steady-state solution is obtained using Redfield theory."
                    },
                    {
                        "title": "Convex Potential Flows: Universal Probability Distributions with Optimal Transport and Convex Optimization",
                        "abstract": "Flow-based models are powerful tools for designing probabilistic models with tractable density. This paper introduces Convex Potential Flows (CP-Flow), a natural and efficient parameterization of invertible models inspired by the optimal transport (OT) theory. CP-Flows are the gradient map of a strongly convex neural potential function. The convexity implies invertibility and allows us to resort to convex optimization to solve the convex conjugate for efficient inversion. To enable maximum likelihood training, we derive a new gradient estimator of the log-determinant of the Jacobian, which involves solving an inverse-Hessian vector product using the conjugate gradient method. The gradient estimator has constant-memory cost, and can be made effectively unbiased by reducing the error tolerance level of the convex optimization routine. Theoretically, we prove that CP-Flows are universal density approximators and are optimal in the OT sense. Our empirical results show that CP-Flow performs competitively on standard benchmarks of density estimation and variational inference."
                    },
                    {
                        "title": "Unifying Generative Models with GFlowNets and Beyond",
                        "abstract": "There are many frameworks for deep generative modeling, each often presented with their own specific training algorithms and inference methods. Here, we demonstrate the connections between existing deep generative models and the recently introduced GFlowNet framework, a probabilistic inference machine which treats sampling as a decision-making process. This analysis sheds light on their overlapping traits and provides a unifying viewpoint through the lens of learning with Markovian trajectories. Our framework provides a means for unifying training and inference algorithms, and provides a route to shine a unifying light over many generative models. Beyond this, we provide a practical and experimentally verified recipe for improving generative modeling with insights from the GFlowNet perspective."
                    }
                ]
            }
        }
    },
    "2404.09632": {
        "paper_data": {
            "title": "Bridging Vision and Language Spaces with Assignment Prediction",
            "url": "http://arxiv.org/abs/2404.09632v1",
            "arxiv_id": "2404.09632",
            "authors": [
                "Jungin Park",
                "Jiyoung Lee",
                "Kwanghoon Sohn"
            ],
            "abstract": "This paper introduces VLAP, a novel approach that bridges pretrained vision models and large language models (LLMs) to make frozen LLMs understand the visual world. VLAP transforms the embedding space of pretrained vision models into the LLMs' word embedding space using a single linear layer for efficient and general-purpose visual and language understanding. Specifically, we harness well-established word embeddings to bridge two modality embedding spaces. The visual and text representations are simultaneously assigned to a set of word embeddings within pretrained LLMs by formulating the assigning procedure as an optimal transport problem. We predict the assignment of one modality from the representation of another modality data, enforcing consistent assignments for paired multimodal data. This allows vision and language representations to contain the same information, grounding the frozen LLMs' word embedding space in visual data. Moreover, a robust semantic taxonomy of LLMs can be preserved with visual data since the LLMs interpret and reason linguistic information from correlations between word embeddings. Experimental results show that VLAP achieves substantial improvements over the previous linear transformation-based approaches across a range of vision-language tasks, including image captioning, visual question answering, and cross-modal retrieval. We also demonstrate the learned visual representations hold a semantic taxonomy of LLMs, making visual semantic arithmetic possible.",
            "introduction": "   1 Introduction  Vision-language models (VLMs) have achieved significant progress in demonstrating remarkable transfer and zero-shot capabilities on vision-language downstream tasks\u00a0(Tan & Bansal, 2019; Lu et\u00a0al., 2019; Chen et\u00a0al., 2020; Huang et\u00a0al., 2020; Radford et\u00a0al., 2021; Jia et\u00a0al., 2021). Most cutting-edge VLMs have been developed with progressively scaled-up foundation models and datasets. However, this trend underscores the demand for substantial computational resources and an extensive collection of image-text pairs. Concurrently, pretrained unimodal foundation models, such as vision transformers\u00a0(Dosovitskiy et\u00a0al., 2021; Caron et\u00a0al., 2021; Bao et\u00a0al., 2022) and large language models (LLMs)\u00a0(Brown et\u00a0al., 2020; Zhang et\u00a0al., 2022; Raffel et\u00a0al., 2020; Touvron et\u00a0al., 2023; Chiang et\u00a0al., 2023), have been developed with self-supervision, renewing the state-of-the-art in vision and language downstream tasks respectively. This raises the interesting question: Can pretrained unimodal models extend their capabilities beyond the modality of pretraining data?   Recently, intriguing attempts have been made to assemble pretrained vision models and LLMs to seem VLMs\u00a0(Mokady et\u00a0al., 2021; Tsimpoukelli et\u00a0al., 2021; Alayrac et\u00a0al., 2022; Patel & Pavlick, 2022; Eichenberg et\u00a0al., 2022; Li et\u00a0al., 2022; 2023; Guo et\u00a0al., 2023; Liu et\u00a0al., 2023a). Most of these attempts suggested ways to train relatively small-scale connection modules while freezing pretrained parameters. By doing so, it is possible to efficiently make VLMs that leverage discriminative visual representations from pretrained vision models, coupled with powerful language modeling and zero-shot capabilities of pretrained LLMs. While the burden of training costs in those methods has significantly decreased compared to pretraining VLMs from scratch, they still require high computations (both time and memory), which limits applicability (e.g. Flamingo\u00a0(Alayrac et\u00a0al., 2022) requires 15 days with 1535 TPUs to train 10.2B parameters, BLIP-2\u00a0(Li et\u00a0al., 2023) requires 9 days with 16 A100 GPUs to train 188M parameters).   Figure 1: Overview of VLAP. We train a single linear layer following two learning objectives: Assignments prediction to bridge the modality gap between the visual and text representations; and image captioning to yield the generative capability of frozen LLMs.   Linear transformation-based approaches\u00a0(Merullo et\u00a0al., 2023; Koh et\u00a0al., 2023) have facilitated efficient cross-modal alignment with a cost-efficient linear transformation between vision and language spaces. Merullo et\u00a0al. (2023) has shown that visual representations can be linearly transformed into LLMs\u2019 input embedding as soft prompts, demonstrating that well-learned visual and text representations are functionally equivalent. Similarly, Koh et\u00a0al. (2023) has proposed to generate free-form text interleaved with images by learning additional trainable tokens and linear layers. These linear transformation-based cross-modal alignments have been achieved by the image captioning objective\u00a0(Merullo et\u00a0al., 2023) and an additional retrieval objective\u00a0(Koh et\u00a0al., 2023), which directly compares the representations of image-text pairs and enforces instance discrimination as contrastive learning\u00a0(Chopra et\u00a0al., 2005). However, the inherent modality gap between different modality data leads to the inconsistency between the isolated embedding spaces (as their pretraining has been solely on unimodal data) and makes cross-modal alignment for frozen unimodal models difficult with the following challenges: (1) it relies heavily on the amount of linguistic supervision used in the pretraining of vision models, showing limited performance in image-only trained vision models (e.g. CLIP vs BEiT)\u00a0(Merullo et\u00a0al., 2023). (2) the contrastive objective (such as InfoNCE\u00a0(van\u00a0den Oord et\u00a0al., 2018) in Koh et\u00a0al. (2023)) is insufficient to bridge complex embedding spaces because it inherently",
            "references": [
                {
                    "title": "COT: Unsupervised Domain Adaptation with Clustering and Optimal Transport",
                    "abstract": "Unsupervised domain adaptation (UDA) aims to transfer the knowledge from a labeled source domain to an unlabeled target domain. Typically, to guarantee desirable knowledge transfer, aligning the distribution between source and target domain from a global perspective is widely adopted in UDA. Recent researchers further point out the importance of local-level alignment and propose to construct instance-pair alignment by leveraging on Optimal Transport (OT) theory. However, existing OT-based UDA approaches are limited to handling class imbalance challenges and introduce a heavy computation overhead when considering a large-scale training situation. To cope with two aforementioned issues, we propose a Clustering-based Optimal Transport (COT) algorithm, which formulates the alignment procedure as an Optimal Transport problem and constructs a mapping between clustering centers in the source and target domain via an end-to-end manner. With this alignment on clustering centers, our COT eliminates the negative effect caused by class imbalance and reduces the computation cost simultaneously. Empirically, our COT achieves state-of-the-art performance on several authoritative benchmark datasets."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Grounding Language Models to Images for Multimodal Inputs and Outputs",
                    "abstract": "We propose an efficient method to ground pretrained text-only language models to the visual domain, enabling them to process arbitrarily interleaved image-and-text data, and generate text interleaved with retrieved images. Our method leverages the abilities of language models learnt from large scale text-only pretraining, such as in-context learning and free-form text generation. We keep the language model frozen, and finetune input and output linear layers to enable cross-modality interactions. This allows our model to process arbitrarily interleaved image-and-text inputs, and generate free-form text interleaved with retrieved images. We achieve strong zero-shot performance on grounded tasks such as contextual image retrieval and multimodal dialogue, and showcase compelling interactive abilities. Our approach works with any off-the-shelf language model and paves the way towards an effective, general solution for leveraging pretrained language models in visually grounded settings."
                },
                {
                    "title": "Scaling Instruction-Finetuned Language Models",
                    "abstract": "Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models."
                },
                {
                    "title": "Linearly Mapping from Image to Text Space",
                    "abstract": "The extent to which text-only language models (LMs) learn to represent features of the non-linguistic world is an open question. Prior work has shown that pretrained LMs can be taught to caption images when a vision model's parameters are optimized to encode images in the language space. We test a stronger hypothesis: that the conceptual representations learned by frozen text-only models and vision-only models are similar enough that this can be achieved with a linear map. We show that the image representations from vision models can be transferred as continuous prompts to frozen LMs by training only a single linear projection. Using these to prompt the LM achieves competitive performance on captioning and visual question answering tasks compared to models that tune both the image encoder and text decoder (such as the MAGMA model). We compare three image encoders with increasing amounts of linguistic supervision seen during pretraining: BEIT (no linguistic information), NF-ResNET (lexical category information), and CLIP (full natural language descriptions). We find that all three encoders perform equally well at transferring visual property information to the language model (e.g., whether an animal is large or small), but that image encoders pretrained with linguistic supervision more saliently encode category information (e.g., distinguishing hippo vs. elephant) and thus perform significantly better on benchmark language-and-vision tasks. Our results indicate that LMs encode conceptual information structurally similarly to vision-based models, even those that are solely trained on images. Code is available here: https://github.com/jmerullo/limber"
                },
                {
                    "title": "Multimodal Knowledge Alignment with Reinforcement Learning",
                    "abstract": "Large language models readily adapt to novel settings, even without task-specific training data. Can their zero-shot capacity be extended to multimodal inputs? In this work, we propose ESPER which extends language-only zero-shot models to unseen multimodal tasks, like image and audio captioning. Our key novelty is to use reinforcement learning to align multimodal inputs to language model generations without direct supervision: for example, in the image case our reward optimization relies only on cosine similarity derived from CLIP, and thus requires no additional explicitly paired (image, caption) data. Because the parameters of the language model are left unchanged, the model maintains its capacity for zero-shot generalization. Experiments demonstrate that ESPER outperforms baselines and prior work on a variety of zero-shot tasks; these include a new benchmark we collect+release, ESP dataset, which tasks models with generating several diversely-styled captions for each image."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "Flamingo: a Visual Language Model for Few-Shot Learning",
                    "abstract": "Building models that can be rapidly adapted to novel tasks using only a handful of annotated examples is an open challenge for multimodal machine learning research. We introduce Flamingo, a family of Visual Language Models (VLM) with this ability. We propose key architectural innovations to: (i) bridge powerful pretrained vision-only and language-only models, (ii) handle sequences of arbitrarily interleaved visual and textual data, and (iii) seamlessly ingest images or videos as inputs. Thanks to their flexibility, Flamingo models can be trained on large-scale multimodal web corpora containing arbitrarily interleaved text and images, which is key to endow them with in-context few-shot learning capabilities. We perform a thorough evaluation of our models, exploring and measuring their ability to rapidly adapt to a variety of image and video tasks. These include open-ended tasks such as visual question-answering, where the model is prompted with a question which it has to answer; captioning tasks, which evaluate the ability to describe a scene or an event; and close-ended tasks such as multiple-choice visual question-answering. For tasks lying anywhere on this spectrum, a single Flamingo model can achieve a new state of the art with few-shot learning, simply by prompting the model with task-specific examples. On numerous benchmarks, Flamingo outperforms models fine-tuned on thousands of times more task-specific data."
                },
                {
                    "title": "\"This is my unicorn, Fluffy\": Personalizing frozen vision-language representations",
                    "abstract": "Large Vision&Language models pretrained on web-scale data provide representations that are invaluable for numerous V&L problems. However, it is unclear how they can be used for reasoning about user-specific visual concepts in unstructured language. This problem arises in multiple domains, from personalized image retrieval to personalized interaction with smart devices. We introduce a new learning setup called Personalized Vision&Language (PerVL) with two new benchmark datasets for retrieving and segmenting user-specific\"personalized\"concepts\"in the wild\". In PerVL, one should learn personalized concepts (1) independently of the downstream task (2) allowing a pretrained model to reason about them with free language, and (3) does not require personalized negative examples. We propose an architecture for solving PerVL that operates by extending the input vocabulary of a pretrained model with new word embeddings for the new personalized concepts. The model can then reason about them by simply using them in a sentence. We demonstrate that our approach learns personalized visual concepts from a few examples and can effectively apply them in image retrieval and semantic segmentation using rich textual queries."
                },
                {
                    "title": "Mind the Gap: Understanding the Modality Gap in Multi-modal Contrastive Representation Learning",
                    "abstract": "We present modality gap, an intriguing geometric phenomenon of the representation space of multi-modal models. Specifically, we show that different data modalities (e.g. images and text) are embedded at arm's length in their shared representation in multi-modal models such as CLIP. Our systematic analysis demonstrates that this gap is caused by a combination of model initialization and contrastive learning optimization. In model initialization, we show empirically and theoretically that the representation of a common deep neural network is restricted to a narrow cone. As a consequence, in a multi-modal model with two encoders, the representations of the two modalities are clearly apart when the model is initialized. During optimization, contrastive learning keeps the different modalities separate by a certain distance, which is influenced by the temperature parameter in the loss function. Our experiments further demonstrate that varying the modality gap distance has a significant impact in improving the model's downstream zero-shot classification performance and fairness. Our code and data are available at https://modalitygap.readthedocs.io/"
                },
                {
                    "title": "Multi-modal Alignment using Representation Codebook",
                    "abstract": "Aligning signals from different modalities is an important step in vision-language representation learning as it affects the performance of later stages such as cross-modality fusion. Since image and text typically reside in different regions of the feature space, directly aligning them at instance level is challenging especially when features are still evolving during training. In this paper, we propose to align at a higher and more stable level using cluster representation. Specifically, we treat image and text as two \u201cviews\u201d of the same entity, and encode them into a joint vision-language coding space spanned by a dictionary of cluster centers (codebook). We contrast positive and negative samples via their cluster assignments while simultaneously optimizing the cluster centers. To further smooth out the learning process, we adopt a teacher-student distillation paradigm, where the momentum teacher of one view guides the student learning of the other. We evaluated our approach on common vision language benchmarks and obtain new SoTA on zero-shot cross modality retrieval while being competitive on various other transfer tasks."
                },
                {
                    "title": "BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation",
                    "abstract": "Vision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has been largely achieved by scaling up the dataset with noisy image-text pairs collected from the web, which is a suboptimal source of supervision. In this paper, we propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks. BLIP effectively utilizes the noisy web data by bootstrapping the captions, where a captioner generates synthetic captions and a filter removes the noisy ones. We achieve state-of-the-art results on a wide range of vision-language tasks, such as image-text retrieval (+2.7% in average recall@1), image captioning (+2.8% in CIDEr), and VQA (+1.6% in VQA score). BLIP also demonstrates strong generalization ability when directly transferred to video-language tasks in a zero-shot manner. Code, models, and datasets are released at https://github.com/salesforce/BLIP."
                },
                {
                    "title": "VL-ADAPTER: Parameter-Efficient Transfer Learning for Vision-and-Language Tasks",
                    "abstract": "Recently, fine-tuning language models pre-trained on large text corpora have provided huge improvements on vision-and-language (V&L) tasks as well as on pure language tasks. However, fine-tuning the entire parameter set of pre-trained models becomes impractical since the model size is growing rapidly. Hence, in this paper, we introduce adapter-based parameter-efficient transfer learning techniques to V&L models such as VL-BART and VL-T5. We evaluate our methods in a unified multi-task setup on both image-text and video-text benchmarks. For the image-text tasks, we use four diverse V&L datasets: VQAv2, GQA, NLVR2, and MSCOCO image captioning. For video-text tasks, we use TVQA, How2QA, TVC, and YC2C. With careful training and thorough experiments, we benchmark three popular adapter-based methods (Adapter, Hyperformer, Compacter) against the standard full fine-tuning and the recently proposed prompt-tuning approach. We also enhance the efficiency and performance of adapters by sharing their weights to attain knowledge across tasks. Our results demonstrate that training the adapter with the weight-sharing technique (4.18% of total parameters for image-text tasks and 3.39% for video-text tasks) can match the performance of fine-tuning the entire model. Lastly, we present a comprehensive analysis including the combination of adapter and task-specific prompts and the impact of V&L pre-training on adapters. 11The code for our CVPR 2022 paper is available at: https://github.com/ylsung/VL_adapter."
                },
                {
                    "title": "MAGMA - Multimodal Augmentation of Generative Models through Adapter-based Finetuning",
                    "abstract": "Large-scale pretraining is fast becoming the norm in Vision-Language (VL) modeling. However, prevailing VL approaches are limited by the requirement for labeled data and the use of complex multi-step pretraining objectives. We present MAGMA - a simple method for augmenting generative language models with additional modalities using adapter-based finetuning. Building on Frozen, we train a series of VL models that autoregressively generate text from arbitrary combinations of visual and textual input. The pretraining is entirely end-to-end using a single language modeling objective, simplifying optimization compared to previous approaches. Importantly, the language model weights remain unchanged during training, allowing for transfer of encyclopedic knowledge and in-context learning abilities from language pretraining. MAGMA outperforms Frozen on open-ended generative tasks, achieving state of the art results on the OKVQA benchmark and competitive results on a range of other popular VL benchmarks, while pretraining on 0.2% of the number of samples used to train SimVLM."
                },
                {
                    "title": "ZeroCap: Zero-Shot Image-to-Text Generation for Visual-Semantic Arithmetic",
                    "abstract": "Recent text-to-image matching models apply contrastive learning to large corpora of uncurated pairs of images and sentences. While such models can provide a powerful score for matching and subsequent zero-shot tasks, they are not capable of generating caption given an image. In this work, we repurpose such models to generate a descriptive text given an image at inference time, without any further training or tuning step. This is done by combining the visual-semantic model with a large language model, benefiting from the knowledge in both web-scale models. The resulting captions are much less restrictive than those obtained by supervised captioning methods. Moreover, as a zero-shot learning method, it is extremely flexible and we demonstrate its ability to perform image arithmetic in which the inputs can be either images or text and the output is a sentence. This enables novel high-level vision capabilities such as comparing two images or solving visual analogy tests. Our code is available at: https://github.com/YoadTew/zero-shot-image-to-text."
                },
                {
                    "title": "ClipCap: CLIP Prefix for Image Captioning",
                    "abstract": "Image captioning is a fundamental task in vision-language understanding, where the model predicts a textual informative caption to a given input image. In this paper, we present a simple approach to address this task. We use CLIP encoding as a prefix to the caption, by employing a simple mapping network, and then fine-tunes a language model to generate the image captions. The recently proposed CLIP model contains rich semantic features which were trained with textual context, making it best for vision-language perception. Our key idea is that together with a pre-trained language model (GPT2), we obtain a wide understanding of both visual and textual data. Hence, our approach only requires rather quick training to produce a competent captioning model. Without additional annotations or pre-training, it efficiently generates meaningful captions for large-scale and diverse datasets. Surprisingly, our method works well even when only the mapping network is trained, while both CLIP and the language model remain frozen, allowing a lighter architecture with less trainable parameters. Through quantitative evaluation, we demonstrate our model achieves comparable results to state-of-the-art methods on the challenging Conceptual Captions and nocaps datasets, while it is simpler, faster, and lighter. Our code is available in https://github.com/rmokady/CLIP_prefix_caption."
                },
                {
                    "title": "Align before Fuse: Vision and Language Representation Learning with Momentum Distillation",
                    "abstract": "Large-scale vision and language representation learning has shown promising improvements on various vision-language tasks. Most existing methods employ a transformer-based multimodal encoder to jointly model visual tokens (region-based image features) and word tokens. Because the visual tokens and word tokens are unaligned, it is challenging for the multimodal encoder to learn image-text interactions. In this paper, we introduce a contrastive loss to ALign the image and text representations BEfore Fusing (ALBEF) them through cross-modal attention, which enables more grounded vision and language representation learning. Unlike most existing methods, our method does not require bounding box annotations nor high-resolution images. In order to improve learning from noisy web data, we propose momentum distillation, a self-training method which learns from pseudo-targets produced by a momentum model. We provide a theoretical analysis of ALBEF from a mutual information maximization perspective, showing that different training tasks can be interpreted as different ways to generate views for an image-text pair. ALBEF achieves state-of-the-art performance on multiple downstream vision-language tasks. On image-text retrieval, ALBEF outperforms methods that are pre-trained on orders of magnitude larger datasets. On VQA and NLVR$^2$, ALBEF achieves absolute improvements of 2.37% and 3.84% compared to the state-of-the-art, while enjoying faster inference speed. Code and pre-trained models are available at https://github.com/salesforce/ALBEF/."
                },
                {
                    "title": "Multimodal Few-Shot Learning with Frozen Language Models",
                    "abstract": "When trained at sufficient scale, auto-regressive language models exhibit the notable ability to learn a new language task after being prompted with just a few examples. Here, we present a simple, yet effective, approach for transferring this few-shot learning ability to a multimodal setting (vision and language). Using aligned image and caption data, we train a vision encoder to represent each image as a sequence of continuous embeddings, such that a pre-trained, frozen language model prompted with this prefix generates the appropriate caption. The resulting system is a multimodal few-shot learner, with the surprising ability to learn a variety of new tasks when conditioned on examples, represented as a sequence of multiple interleaved image and text embeddings. We demonstrate that it can rapidly learn words for new objects and novel visual categories, do visual question-answering with only a handful of examples, and make use of outside knowledge, by measuring a single model on a variety of established and new benchmarks."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "CLIPScore: A Reference-free Evaluation Metric for Image Captioning",
                    "abstract": "Image captioning has conventionally relied on reference-based automatic evaluations, where machine captions are compared against captions written by humans. This is in contrast to the reference-free manner in which humans assess caption quality. In this paper, we report the surprising empirical finding that CLIP (Radford et al., 2021), a cross-modal model pretrained on 400M image+caption pairs from the web, can be used for robust automatic evaluation of image captioning without the need for references. Experiments spanning several corpora demonstrate that our new reference-free metric, CLIPScore, achieves the highest correlation with human judgements, outperforming existing reference-based metrics like CIDEr and SPICE. Information gain experiments demonstrate that CLIPScore, with its tight focus on image-text compatibility, is complementary to existing reference-based metrics that emphasize text-text similarities. Thus, we also present a reference-augmented version, RefCLIPScore, which achieves even higher correlation. Beyond literal description tasks, several case studies reveal domains where CLIPScore performs well (clip-art images, alt-text rating), but also where it is relatively weaker in comparison to reference-based metrics, e.g., news captions that require richer contextual knowledge."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Conceptual 12M: Pushing Web-Scale Image-Text Pre-Training To Recognize Long-Tail Visual Concepts",
                    "abstract": "The availability of large-scale image captioning and visual question answering datasets has contributed significantly to recent successes in vision-and-language pretraining. However, these datasets are often collected with overrestrictive requirements inherited from their original target tasks (e.g., image caption generation), which limit the resulting dataset scale and diversity. We take a step further in pushing the limits of vision-and-language pretraining data by relaxing the data collection pipeline used in Conceptual Captions 3M (CC3M) [54] and introduce the Conceptual 12M (CC12M), a dataset with 12 million image-text pairs specifically meant to be used for visionand-language pre-training. We perform an analysis of this dataset and benchmark its effectiveness against CC3M on multiple downstream tasks with an emphasis on long-tail visual recognition. Our results clearly illustrate the benefit of scaling up pre-training data for vision-and-language tasks, as indicated by the new state-of-the-art results on both the nocaps and Conceptual Captions benchmarks.1"
                },
                {
                    "title": "Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision",
                    "abstract": "Pre-trained representations are becoming crucial for many NLP and perception tasks. While representation learning in NLP has transitioned to training on raw text without human annotations, visual and vision-language representations still rely heavily on curated training datasets that are expensive or require expert knowledge. For vision applications, representations are mostly learned using datasets with explicit class labels such as ImageNet or OpenImages. For vision-language, popular datasets like Conceptual Captions, MSCOCO, or CLIP all involve a non-trivial data collection (and cleaning) process. This costly curation process limits the size of datasets and hence hinders the scaling of trained models. In this paper, we leverage a noisy dataset of over one billion image alt-text pairs, obtained without expensive filtering or post-processing steps in the Conceptual Captions dataset. A simple dual-encoder architecture learns to align visual and language representations of the image and text pairs using a contrastive loss. We show that the scale of our corpus can make up for its noise and leads to state-of-the-art representations even with such a simple learning scheme. Our visual representation achieves strong performance when transferred to classification tasks such as ImageNet and VTAB. The aligned visual and language representations enables zero-shot image classification and also set new state-of-the-art results on Flickr30K and MSCOCO image-text retrieval benchmarks, even when compared with more sophisticated cross-attention models. The representations also enable cross-modality search with complex text and text + image queries."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Probing Pretrained Language Models for Lexical Semantics",
                    "abstract": "The success of large pretrained language models (LMs) such as BERT and RoBERTa has sparked interest in probing their representations, in order to unveil what types of knowledge they implicitly capture. While prior research focused on morphosyntactic, semantic, and world knowledge, it remains unclear to which extent LMs also derive lexical type-level knowledge from words in context. In this work, we present a systematic empirical analysis across six typologically diverse languages and five different lexical tasks, addressing the following questions: 1) How do different lexical knowledge extraction strategies (monolingual versus multilingual source LM, out-of-context versus in-context encoding, inclusion of special tokens, and layer-wise averaging) impact performance? How consistent are the observed effects across tasks and languages? 2) Is lexical knowledge stored in few parameters, or is it scattered throughout the network? 3) How do these representations fare against traditional static word vectors in lexical tasks? 4) Does the lexical information emerging from independently trained monolingual LMs display latent similarities? Our main results indicate patterns and best practices that hold universally, but also point to prominent variations across languages and tasks. Moreover, we validate the claim that lower Transformer layers carry more type-level lexical knowledge, but also show that this knowledge is distributed across multiple layers."
                },
                {
                    "title": "Parametric UMAP Embeddings for Representation and Semisupervised Learning",
                    "abstract": "Abstract UMAP is a nonparametric graph-based dimensionality reduction algorithm using applied Riemannian geometry and algebraic topology to find low-dimensional embeddings of structured data. The UMAP algorithm consists of two steps: (1) computing a graphical representation of a data set (fuzzy simplicial complex) and (2) through stochastic gradient descent, optimizing a low-dimensional embedding of the graph. Here, we extend the second step of UMAP to a parametric optimization over neural network weights, learning a parametric relationship between data and embedding. We first demonstrate that parametric UMAP performs comparably to its nonparametric counterpart while conferring the benefit of a learned parametric mapping (e.g., fast online embeddings for new data). We then explore UMAP as a regularization, constraining the latent distribution of autoencoders, parametrically varying global structure preservation, and improving classifier accuracy for semisupervised learning by capturing structure in unlabeled data.1"
                },
                {
                    "title": "Labelling unlabelled videos from scratch with multi-modal self-supervision",
                    "abstract": "A large part of the current success of deep learning lies in the effectiveness of data -- more precisely: labelled data. Yet, labelling a dataset with human annotation continues to carry high costs, especially for videos. While in the image domain, recent methods have allowed to generate meaningful (pseudo-) labels for unlabelled datasets without supervision, this development is missing for the video domain where learning feature representations is the current focus. In this work, we a) show that unsupervised labelling of a video dataset does not come for free from strong feature encoders and b) propose a novel clustering method that allows pseudo-labelling of a video dataset without any human annotations, by leveraging the natural correspondence between the audio and visual modalities. An extensive analysis shows that the resulting clusters have high semantic overlap to ground truth human labels. We further introduce the first benchmarking results on unsupervised labelling of common video datasets Kinetics, Kinetics-Sound, VGG-Sound and AVE."
                },
                {
                    "title": "Unsupervised Learning of Visual Features by Contrasting Cluster Assignments",
                    "abstract": "Unsupervised image representations have significantly reduced the gap with supervised pretraining, notably with the recent achievements of contrastive learning methods. These contrastive methods typically work online and rely on a large number of explicit pairwise feature comparisons, which is computationally challenging. In this paper, we propose an online algorithm, SwAV, that takes advantage of contrastive methods without requiring to compute pairwise comparisons. Specifically, our method simultaneously clusters the data while enforcing consistency between cluster assignments produced for different augmentations (or views) of the same image, instead of comparing features directly as in contrastive learning. Simply put, we use a swapped prediction mechanism where we predict the cluster assignment of a view from the representation of another view. Our method can be trained with large and small batches and can scale to unlimited amounts of data. Compared to previous contrastive methods, our method is more memory efficient since it does not require a large memory bank or a special momentum network. In addition, we also propose a new data augmentation strategy, multi-crop, that uses a mix of views with different resolutions in place of two full-resolution views, without increasing the memory or compute requirements much. We validate our findings by achieving 75.3% top-1 accuracy on ImageNet with ResNet-50, as well as surpassing supervised pretraining on all the considered transfer tasks."
                },
                {
                    "title": "Reliable Weighted Optimal Transport for Unsupervised Domain Adaptation",
                    "abstract": "Recently, extensive researches have been proposed to address the UDA problem, which aims to learn transferrable models for the unlabeled target domain. Among them, the optimal transport is a promising metric to align the representations of the source and target domains. However, most existing works based on optimal transport ignore the intra-domain structure, only achieving coarse pair-wise matching. The target samples distributed near the edge of the clusters, or far from their corresponding class centers are easily to be misclassified by the decision boundary learned from the source domain. In this paper, we present Reliable Weighted Optimal Transport (RWOT) for unsupervised domain adaptation, including novel Shrinking Subspace Reliability (SSR) and weighted optimal transport strategy. Specifically, SSR exploits spatial prototypical information and intra-domain structure to dynamically measure the sample-level domain discrepancy across domains. Besides, the weighted optimal transport strategy based on SSR is exploited to achieve the precise-pair-wise optimal transport procedure, which reduces negative transfer brought by the samples near decision boundaries in the target domain. RWOT also equips with the discriminative centroid clustering exploitation strategy to learn transfer features. A thorough evaluation shows that RWOT outperforms existing state-of-the-art method on standard domain adaptation benchmarks."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Pixel-BERT: Aligning Image Pixels with Text by Deep Multi-Modal Transformers",
                    "abstract": "We propose Pixel-BERT to align image pixels with text by deep multi-modal transformers that jointly learn visual and language embedding in a unified end-to-end framework. We aim to build a more accurate and thorough connection between image pixels and language semantics directly from image and sentence pairs instead of using region-based image features as the most recent vision and language tasks. Our Pixel-BERT which aligns semantic connection in pixel and text level solves the limitation of task-specific visual representation for vision and language tasks. It also relieves the cost of bounding box annotations and overcomes the unbalance between semantic labels in visual task and language semantic. To provide a better representation for down-stream tasks, we pre-train a universal end-to-end model with image and sentence pairs from Visual Genome dataset and MS-COCO dataset. We propose to use a random pixel sampling mechanism to enhance the robustness of visual representation and to apply the Masked Language Model and Image-Text Matching as pre-training tasks. Extensive experiments on downstream tasks with our pre-trained model show that our approach makes the most state-of-the-arts in downstream tasks, including Visual Question Answering (VQA), image-text retrieval, Natural Language for Visual Reasoning for Real (NLVR). Particularly, we boost the performance of a single model in VQA task by 2.17 points compared with SOTA under fair comparison."
                },
                {
                    "title": "Self-labelling via simultaneous clustering and representation learning",
                    "abstract": "Combining clustering and representation learning is one of the most promising approaches for unsupervised learning of deep neural networks. However, doing so naively leads to ill posed learning problems with degenerate solutions. In this paper, we propose a novel and principled learning formulation that addresses these issues. The method is obtained by maximizing the information between labels and input data indices. We show that this criterion extends standard crossentropy minimization to an optimal transport problem, which we solve efficiently for millions of input images and thousands of labels using a fast variant of the Sinkhorn-Knopp algorithm. The resulting method is able to self-label visual data so as to train highly competitive image representations without manual labels. Our method achieves state of the art representation learning performance for AlexNet and ResNet-50 on SVHN, CIFAR-10, CIFAR-100 and ImageNet and yields the first self-supervised AlexNet that outperforms the supervised Pascal VOC detection baseline. Code and models are available."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "LXMERT: Learning Cross-Modality Encoder Representations from Transformers",
                    "abstract": "Vision-and-language reasoning requires an understanding of visual concepts, language semantics, and, most importantly, the alignment and relationships between these two modalities. We thus propose the LXMERT (Learning Cross-Modality Encoder Representations from Transformers) framework to learn these vision-and-language connections. In LXMERT, we build a large-scale Transformer model that consists of three encoders: an object relationship encoder, a language encoder, and a cross-modality encoder. Next, to endow our model with the capability of connecting vision and language semantics, we pre-train the model with large amounts of image-and-sentence pairs, via five diverse representative pre-training tasks: masked language modeling, masked object prediction (feature regression and label classification), cross-modality matching, and image question answering. These tasks help in learning both intra-modality and cross-modality relationships. After fine-tuning from our pre-trained parameters, our model achieves the state-of-the-art results on two visual question answering datasets (i.e., VQA and GQA). We also show the generalizability of our pre-trained cross-modality model by adapting it to a challenging visual-reasoning task, NLVR2, and improve the previous best result by 22% absolute (54% to 76%). Lastly, we demonstrate detailed ablation studies to prove that both our novel model components and pre-training strategies significantly contribute to our strong results. Code and pre-trained models publicly available at: https://github.com/airsplay/lxmert"
                },
                {
                    "title": "ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks",
                    "abstract": "We present ViLBERT (short for Vision-and-Language BERT), a model for learning task-agnostic joint representations of image content and natural language. We extend the popular BERT architecture to a multi-modal two-stream model, pro-cessing both visual and textual inputs in separate streams that interact through co-attentional transformer layers. We pretrain our model through two proxy tasks on the large, automatically collected Conceptual Captions dataset and then transfer it to multiple established vision-and-language tasks -- visual question answering, visual commonsense reasoning, referring expressions, and caption-based image retrieval -- by making only minor additions to the base architecture. We observe significant improvements across tasks compared to existing task-specific models -- achieving state-of-the-art on all four tasks. Our work represents a shift away from learning groundings between vision and language only as part of task training and towards treating visual grounding as a pretrainable and transferable capability."
                },
                {
                    "title": "What Does BERT Learn about the Structure of Language?",
                    "abstract": "BERT is a recent language representation model that has surprisingly performed well in diverse language understanding benchmarks. This result indicates the possibility that BERT networks capture structural information about language. In this work, we provide novel support for this claim by performing a series of experiments to unpack the elements of English language structure learned by BERT. Our findings are fourfold. BERT\u2019s phrasal representation captures the phrase-level information in the lower layers. The intermediate layers of BERT compose a rich hierarchy of linguistic information, starting with surface features at the bottom, syntactic features in the middle followed by semantic features at the top. BERT requires deeper layers while tracking subject-verb agreement to handle long-term dependency problem. Finally, the compositional scheme underlying BERT mimics classical, tree-like structures."
                },
                {
                    "title": "A Structural Probe for Finding Syntax in Word Representations",
                    "abstract": "Recent work has improved our ability to detect linguistic knowledge in word representations. However, current methods for detecting syntactic knowledge do not test whether syntax trees are represented in their entirety. In this work, we propose a structural probe, which evaluates whether syntax trees are embedded in a linear transformation of a neural network\u2019s word representation space. The probe identifies a linear transformation under which squared L2 distance encodes the distance between words in the parse tree, and one in which squared L2 norm encodes depth in the parse tree. Using our probe, we show that such transformations exist for both ELMo and BERT but not in baselines, providing evidence that entire syntax trees are embedded implicitly in deep models\u2019 vector geometry."
                },
                {
                    "title": "BERT Rediscovers the Classical NLP Pipeline",
                    "abstract": "Pre-trained text encoders have rapidly advanced the state of the art on many NLP tasks. We focus on one such model, BERT, and aim to quantify where linguistic information is captured within the network. We find that the model represents the steps of the traditional NLP pipeline in an interpretable and localizable way, and that the regions responsible for each step appear in the expected sequence: POS tagging, parsing, NER, semantic roles, then coreference. Qualitative analysis reveals that the model can and often does adjust this pipeline dynamically, revising lower-level decisions on the basis of disambiguating information from higher-level representations."
                },
                {
                    "title": "Linguistic Knowledge and Transferability of Contextual Representations",
                    "abstract": "Contextual word representations derived from large-scale neural language models are successful across a diverse set of NLP tasks, suggesting that they encode useful and transferable features of language. To shed light on the linguistic knowledge they capture, we study the representations produced by several recent pretrained contextualizers (variants of ELMo, the OpenAI transformer language model, and BERT) with a suite of sixteen diverse probing tasks. We find that linear models trained on top of frozen contextual representations are competitive with state-of-the-art task-specific models in many cases, but fail on tasks requiring fine-grained linguistic knowledge (e.g., conjunct identification). To investigate the transferability of contextual word representations, we quantify differences in the transferability of individual layers within contextualizers, especially between recurrent neural networks (RNNs) and transformers. For instance, higher layers of RNNs are more task-specific, while transformer layers do not exhibit the same monotonic trend. In addition, to better understand what makes contextual word representations transferable, we compare language model pretraining with eleven supervised pretraining tasks. For any given task, pretraining on a closely related task yields better performance than language model pretraining (which is better on average) when the pretraining dataset is fixed. However, language model pretraining on more data gives the best results."
                },
                {
                    "title": "Representation Learning with Contrastive Predictive Coding",
                    "abstract": "While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments."
                },
                {
                    "title": "Conceptual Captions: A Cleaned, Hypernymed, Image Alt-text Dataset For Automatic Image Captioning",
                    "abstract": "We present a new dataset of image caption annotations, Conceptual Captions, which contains an order of magnitude more images than the MS-COCO dataset (Lin et al., 2014) and represents a wider variety of both images and image caption styles. We achieve this by extracting and filtering image caption annotations from billions of webpages. We also present quantitative evaluations of a number of image captioning models and show that a model architecture based on Inception-ResNetv2 (Szegedy et al., 2016) for image-feature extraction and Transformer (Vaswani et al., 2017) for sequence modeling achieves the best performance when trained on the Conceptual Captions dataset."
                },
                {
                    "title": "Unsupervised Feature Learning via Non-parametric Instance Discrimination",
                    "abstract": "Neural net classifiers trained on data with annotated class labels can also capture apparent visual similarity among categories without being directed to do so. We study whether this observation can be extended beyond the conventional domain of supervised learning: Can we learn a good feature representation that captures apparent similarity among instances, instead of classes, by merely asking the feature to be discriminative of individual instances? We formulate this intuition as a non-parametric classification problem at the instance-level, and use noise-contrastive estimation to tackle the computational challenges imposed by the large number of instance classes. Our experimental results demonstrate that, under unsupervised learning settings, our method surpasses the state-of-the-art on ImageNet classification by a large margin. Our method is also remarkable for consistently improving test performance with more training data and better network architectures. By fine-tuning the learned feature, we further obtain competitive results for semi-supervised learning and object detection tasks. Our non-parametric model is highly compact: With 128 features per image, our method requires only 600MB storage for a million images, enabling fast nearest neighbour retrieval at the run time."
                },
                {
                    "title": "Visual Dialog",
                    "abstract": "We introduce the task of Visual Dialog, which requires an AI agent to hold a meaningful dialog with humans in natural, conversational language about visual content. Specifically, given an image, a dialog history, and a question about the image, the agent has to ground the question in image, infer context from history, and answer the question accurately. Visual Dialog is disentangled enough from a specific downstream task so as to serve as a general test of machine intelligence, while being grounded in vision enough to allow objective evaluation of individual responses and benchmark progress. We develop a novel two-person chat data-collection protocol to curate a large-scale Visual Dialog dataset (VisDial). VisDial contains 1 dialog (10 question-answer pairs) on ~140k images from the COCO dataset, with a total of ~1.4M dialog question-answer pairs. We introduce a family of neural encoder-decoder models for Visual Dialog with 3 encoders (Late Fusion, Hierarchical Recurrent Encoder and Memory Network) and 2 decoders (generative and discriminative), which outperform a number of sophisticated baselines. We propose a retrieval-based evaluation protocol for Visual Dialog where the AI agent is asked to sort a set of candidate answers and evaluated on metrics such as mean-reciprocal-rank of human response. We quantify gap between machine and human performance on the Visual Dialog task via human studies. Our dataset, code, and trained models will be released publicly at https://visualdialog.org. Putting it all together, we demonstrate the first visual chatbot!."
                },
                {
                    "title": "CIDEr: Consensus-based image description evaluation",
                    "abstract": "Automatically describing an image with a sentence is a long-standing challenge in computer vision and natural language processing. Due to recent progress in object detection, attribute classification, action recognition, etc., there is renewed interest in this area. However, evaluating the quality of descriptions has proven to be challenging. We propose a novel paradigm for evaluating image descriptions that uses human consensus. This paradigm consists of three main parts: a new triplet-based method of collecting human annotations to measure consensus, a new automated metric that captures consensus, and two new datasets: PASCAL-50S and ABSTRACT-50S that contain 50 sentences describing each image. Our simple metric captures human judgment of consensus better than existing metrics across sentences generated by various sources. We also evaluate five state-of-the-art image description approaches using this new protocol and provide a benchmark for future comparisons. A version of CIDEr named CIDEr-D is available as a part of MS COCO evaluation server to enable systematic evaluation and benchmarking."
                },
                {
                    "title": "Sinkhorn Distances: Lightspeed Computation of Optimal Transport",
                    "abstract": "Optimal transportation distances are a fundamental family of parameterized distances for histograms. Despite their appealing theoretical properties, excellent performance in retrieval tasks and intuitive formulation, their computation involves the resolution of a linear program whose cost is prohibitive whenever the histograms' dimension exceeds a few hundreds. We propose in this work a new family of optimal transportation distances that look at transportation problems from a maximum-entropy perspective. We smooth the classical optimal transportation problem with an entropic regularization term, and show that the resulting optimum is also a distance which can be computed through Sinkhorn-Knopp's matrix scaling algorithm at a speed that is several orders of magnitude faster than that of transportation solvers. We also report improved performance over classical optimal transportation distances on the MNIST benchmark problem."
                },
                {
                    "title": "Learning a similarity metric discriminatively, with application to face verification",
                    "abstract": "We present a method for training a similarity metric from data. The method can be used for recognition or verification applications where the number of categories is very large and not known during training, and where the number of training samples for a single category is very small. The idea is to learn a function that maps input patterns into a target space such that the L/sub 1/ norm in the target space approximates the \"semantic\" distance in the input space. The method is applied to a face verification task. The learning process minimizes a discriminative loss function that drives the similarity metric to be small for pairs of faces from the same person, and large for pairs from different persons. The mapping from raw to the target space is a convolutional network whose architecture is designed for robustness to geometric distortions. The system is tested on the Purdue/AR face database which has a very high degree of variability in the pose, lighting, expression, position, and artificial occlusions such as dark glasses and obscuring scarves."
                },
                {
                    "title": "METEOR: An Automatic Metric for MT Evaluation with Improved Correlation with Human Judgments",
                    "abstract": "We describe METEOR, an automatic metric for machine translation evaluation that is based on a generalized concept of unigram matching between the machineproduced translation and human-produced reference translations. Unigrams can be matched based on their surface forms, stemmed forms, and meanings; furthermore, METEOR can be easily extended to include more advanced matching strategies. Once all generalized unigram matches between the two strings have been found, METEOR computes a score for this matching using a combination of unigram-precision, unigram-recall, and a measure of fragmentation that is designed to directly capture how well-ordered the matched words in the machine translation are in relation to the reference. We evaluate METEOR by measuring the correlation between the metric scores and human judgments of translation quality. We compute the Pearson R correlation value between its scores and human quality assessments of the LDC TIDES 2003 Arabic-to-English and Chinese-to-English datasets. We perform segment-bysegment correlation, and show that METEOR gets an R correlation value of 0.347 on the Arabic data and 0.331 on the Chinese data. This is shown to be an improvement on using simply unigramprecision, unigram-recall and their harmonic F1 combination. We also perform experiments to show the relative contributions of the various mapping modules."
                },
                {
                    "title": "ROUGE: A Package for Automatic Evaluation of Summaries",
                    "abstract": "ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST."
                },
                {
                    "title": "Bleu: a Method for Automatic Evaluation of Machine Translation",
                    "abstract": "Human evaluations of machine translation are extensive but expensive. Human evaluations can take months to finish and involve human labor that can not be reused. We propose a method of automatic machine translation evaluation that is quick, inexpensive, and language-independent, that correlates highly with human evaluation, and that has little marginal cost per run. We present this method as an automated understudy to skilled human judges which substitutes for them when there is need for quick or frequent evaluations."
                },
                {
                    "title": "Mapping Language Models to Grounded Conceptual Spaces",
                    "abstract": "A fundamental criticism of text-only language models (LMs) is their lack of grounding \u2014that is, the ability to tie a word for which they have learned a representation to its referent in the non-linguistic world. However, despite this limitation, large pre-trained LMs have been shown to have a remarkable grasp of the conceptual structure of language, as demonstrated by their ability to answer questions, generate \ufb02uent text, or make inferences about entities, objects, and properties that they have never physically observed. In this work we investigate the extent to which the rich conceptual structure that LMs learn indeed re\ufb02ects the conceptual structure of the non-linguistic world\u2014which is something that LMs have never observed. We do this by testing whether the LMs can learn to map an entire conceptual domain (e.g., direction or colour) onto a grounded world representation given only a small number of examples. For example, we show a model what the word \u201cleft\u201d means using a textual depiction of a grid world, and assess how well it can generalise to related concepts, for example, the word \u201cright\u201d , in a similar grid world. We investigate a range of generative language models of varying sizes (including GPT-2 and GPT-3), and see that although the smaller models struggle to perform this mapping, the largest model can not only learn to ground the concepts that it is explicitly taught, but appears to generalise to several instances of unseen concepts as well. Our results suggest an alternative means of building grounded language models: rather than learning grounded representations \u201cfrom scratch\u201d, it is possible that large text-only models learn a suf\ufb01ciently rich conceptual structure that could allow them to be grounded in a data-ef\ufb01cient way."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "nocaps: novel object captioning at scale",
                    "abstract": "Image captioning models have achieved impressive results on datasets containing limited visual concepts and large amounts of paired image-caption training data. However, if these models are to ever function in the wild, a much larger variety of visual concepts must be learned, ideally from less supervision. To encourage the development of image captioning models that can learn visual concepts from alternative data sources, such as object detection datasets, we present the first large-scale benchmark for this task. Dubbed \u2018nocaps\u2019, for novel object captioning at scale, our benchmark consists of 166,100 human-generated captions describing 15,100 images from the Open Images validation and test sets. The associated training data consists of COCO image-caption pairs, plus Open Images image-level labels and object bounding boxes. Since Open Images contains many more classes than COCO, nearly 400 object classes seen in test images have no or very few associated training captions (hence, nocaps). We extend existing novel object captioning models to establish strong baselines for this benchmark and provide analysis to guide future work."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Improving Language Understanding by Generative Pre-Training",
                    "abstract": "Natural language understanding comprises a wide range of diverse tasks such as textual entailment, question answering, semantic similarity assessment, and document classi\ufb01cation. Although large unlabeled text corpora are abundant, labeled data for learning these speci\ufb01c tasks is scarce, making it challenging for discriminatively trained models to perform adequately. We demonstrate that large gains on these tasks can be realized by generative pre-training of a language model on a diverse corpus of unlabeled text, followed by discriminative \ufb01ne-tuning on each speci\ufb01c task. In contrast to previous approaches, we make use of task-aware input transformations during \ufb01ne-tuning to achieve effective transfer while requiring minimal changes to the model architecture. We demonstrate the effectiveness of our approach on a wide range of benchmarks for natural language understanding. Our general task-agnostic model outperforms discriminatively trained models that use architectures speci\ufb01cally crafted for each task, signi\ufb01cantly improving upon the state of the art in 9 out of the 12 tasks studied. For instance, we achieve absolute improvements of 8.9% on commonsense reasoning (Stories Cloze Test), 5.7% on question answering (RACE), and 1.5% on textual entailment (MultiNLI)."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan pretrained unimodal models effectively extend their capabilities to perform vision-language tasks without requiring extensive computational resources and large datasets?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing demand for efficient and scalable vision-language models (VLMs) that can operate with limited resources. By demonstrating that unimodal models can be adapted for cross-modal tasks, this research could lead to significant advancements in the field, enabling broader accessibility and application of VLMs in real-world scenarios. It could also inspire future research to explore novel methods for cross-modal learning, potentially leading to more efficient architectures and training paradigms.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent modality gap between visual and textual data, which complicates the alignment of representations from unimodal models. Naive approaches may fail because they do not adequately account for the differences in embedding spaces created during unimodal pretraining. Key obstacles include the reliance on linguistic supervision during vision model pretraining, which limits performance when using image-only trained models, and the inadequacy of contrastive learning objectives to bridge complex embedding spaces effectively.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on training large-scale VLMs from scratch or on assembling pretrained unimodal models without addressing the fundamental challenges of cross-modal alignment. Limitations in existing solutions include the high computational costs and the insufficient performance of linear transformation-based approaches when applied to unimodal models. My approach differs by proposing a more efficient method that leverages linear transformations and specific learning objectives to bridge the modality gap, thus improving upon prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nMy proposed methodology involves training a single linear layer with two learning objectives: assignments prediction to bridge the modality gap and image captioning to utilize the generative capabilities of frozen language models. I will use a dataset of image-text pairs and evaluate the model's performance using metrics such as accuracy in assignments prediction and quality of generated captions. The expected outcomes include demonstrating that unimodal models can effectively perform vision-language tasks with reduced computational requirements, thereby paving the way for more accessible VLM applications."
            }
        },
        "author_data": {
            "208673b2-52f1-4a04-9b62-6f081218376e": {
                "pk": "208673b2-52f1-4a04-9b62-6f081218376e",
                "name": "Jungin Park",
                "collaborators": [
                    "K. Sohn",
                    "Jiyoung Lee",
                    "Seungryong Kim",
                    "Jin Kim",
                    "Dongbo Min",
                    "Sangryul Jeon",
                    "Minsu Kim",
                    "S. Park",
                    "Kwonyoung Kim",
                    "Ig-Jae Kim"
                ],
                "domain": [
                    "Computer Vision",
                    "Video Understanding",
                    "Domain Generalization",
                    "Action Recognition"
                ],
                "publications": [
                    {
                        "title": "PartMix: Regularization Strategy to Learn Part Discovery for Visible-Infrared Person Re-Identification",
                        "abstract": "Modern data augmentation using a mixture-based technique can regularize the models from overfitting to the training data in various computer vision applications, but a proper data augmentation technique tailored for the part-based Visible-Infrared person Re-IDentification (VI-ReID) models remains unexplored. In this paper, we present a novel data augmentation technique, dubbed PartMix, that synthesizes the augmented samples by mixing the part descriptors across the modalities to improve the performance of part-based VI-ReID models. Especially, we synthesize the positive and negative samples within the same and across different identities and regularize the backbone model through contrastive learning. In addition, we also present an entropy-based mining strategy to weaken the adverse impact of unreliable positive and negative samples. When incorporated into existing part-based VI-ReID model, PartMix consistently boosts the performance. We conduct experiments to demonstrate the effectiveness of our PartMix over the existing VI-ReID methods and provide ablation studies."
                    },
                    {
                        "title": "Dual-Path Adaptation from Image to Video Transformers",
                        "abstract": "In this paper, we efficiently transfer the surpassing representation power of the vision foundation models, such as ViT and Swin, for video understanding with only a few trainable parameters. Previous adaptation methods have simultaneously considered spatial and temporal modeling with a unified learnable module but still suffered from fully leveraging the representative capabilities of image transformers. We argue that the popular dual-path (two-stream) architecture in video models can mitigate this problem. We propose a novel DualPath adaptation separated into spatial and temporal adaptation paths, where a lightweight bottleneck adapter is employed in each transformer block. Especially for temporal dynamic modeling, we incorporate consecutive frames into a grid-like frameset to precisely imitate vision transformers' capability that extrapolates relationships between tokens. In addition, we extensively investigate the multiple baselines from a unified perspective in video understanding and compare them with DualPath. Experimental results on four action recognition benchmarks prove that pretrained image transformers with DualPath can be effectively generalized beyond the data domain."
                    },
                    {
                        "title": "Knowing Where to Focus: Event-aware Transformer for Video Grounding",
                        "abstract": "Recent DETR-based video grounding models have made the model directly predict moment timestamps without any hand-crafted components, such as a pre-defined proposal or non-maximum suppression, by learning moment queries. However, their input-agnostic moment queries inevitably overlook an intrinsic temporal structure of a video, providing limited positional information. In this paper, we formulate an event-aware dynamic moment query to enable the model to take the input-specific content and positional information of the video into account. To this end, we present two levels of reasoning: 1) Event reasoning that captures distinctive event units constituting a given video using a slot attention mechanism; and 2) moment reasoning that fuses the moment queries with a given sentence through a gated fusion transformer layer and learns interactions between the moment queries and video-sentence representations to predict moment timestamps. Extensive experiments demonstrate the effectiveness and efficiency of the event-aware dynamic moment queries, outperforming state-of-the-art approaches on several video grounding benchmarks. The code is publicly available at https://github.com/jinhyunj/EaTR."
                    },
                    {
                        "title": "Pin the Memory: Learning to Generalize Semantic Segmentation",
                        "abstract": "The rise of deep neural networks has led to several break-throughs for semantic segmentation. In spite of this, a model trained on source domain often fails to work properly in new challenging domains, that is directly concerned with the generalization capability of the model. In this paper, we present a novel memory-guided domain generalization method for semantic segmentation based on meta-learning framework. Especially, our method abstracts the conceptual knowledge of semantic classes into categorical memory which is constant beyond the domains. Upon the meta-learning concept, we repeatedly train memory-guided networks and simulate virtual test to 1) learn how to memorize a domain-agnostic and distinct information of classes and 2) offer an externally settled memory as a class-guidance to reduce the ambiguity of representation in the test data of arbitrary unseen domain. To this end, we also propose memory divergence and feature cohesion losses, which encourage to learn memory reading and update processes for category-aware domain generalization. Extensive experiments for semantic segmentation demonstrate the superior generalization capability of our method over state-of-the-art works on various benchmarks.11https://github.com/Genie-Kim/PintheMemory"
                    },
                    {
                        "title": "PointFix: Learning to Fix Domain Bias for Robust Online Stereo Adaptation",
                        "abstract": "Online stereo adaptation tackles the domain shift problem, caused by different environments between synthetic (training) and real (test) datasets, to promptly adapt stereo models in dynamic real-world applications such as autonomous driving. However, previous methods often fail to counteract particular regions related to dynamic objects with more severe environmental changes. To mitigate this issue, we propose to incorporate an auxiliary point-selective network into a meta-learning framework, called PointFix, to provide a robust initialization of stereo models for online stereo adaptation. In a nutshell, our auxiliary network learns to fix local variants intensively by effectively back-propagating local information through the meta-gradient for the robust initialization of the baseline model. This network is model-agnostic, so can be used in any kind of architectures in a plug-and-play manner. We conduct extensive experiments to verify the effectiveness of our method under three adaptation settings such as short-, mid-, and long-term sequences. Experimental results show that the proper initialization of the base stereo model by the auxiliary network enables our learning paradigm to achieve state-of-the-art performance at inference."
                    },
                    {
                        "title": "Language-free Training for Zero-shot Video Grounding",
                        "abstract": "Given an untrimmed video and a language query depicting a specific temporal moment in the video, video grounding aims to localize the time interval by understanding the text and video simultaneously. One of the most challenging issues is an extremely time- and cost-consuming annotation collection, including video captions in a natural language form and their corresponding temporal regions. In this paper, we present a simple yet novel training framework for video grounding in the zero-shot setting, which learns a network with only video data without any annotation. Inspired by the recent language-free paradigm, i.e. training without language data, we train the network without compelling the generation of fake (pseudo) text queries into a natural language form. Specifically, we propose a method for learning a video grounding model by selecting a temporal interval as a hypothetical correct answer and considering the visual feature selected by our method in the interval as a language feature, with the help of the well-aligned visual-language space of CLIP. Extensive experiments demonstrate the prominence of our language-free training framework, outperforming the existing zero-shot video grounding method and even several weakly-supervised approaches with large margins on two standard datasets."
                    },
                    {
                        "title": "Probabilistic Representations for Video Contrastive Learning",
                        "abstract": "This paper presents Probabilistic Video Contrastive Learning, a self-supervised representation learning method that bridges contrastive learning with probabilistic representation. We hypothesize that the clips composing the video have different distributions in short-term duration, but can represent the complicated and sophisticated video distribution through combination in a common embedding space. Thus, the proposed method represents video clips as normal distributions and combines them into a Mixture of Gaussians to model the whole video distribution. By sampling embeddings from the whole video distribution, we can circumvent the careful sampling strategy or transformations to generate augmented views of the clips, unlike previous deterministic methods that have mainly focused on such sample generation strategies for contrastive learning. We further propose a stochastic contrastive loss to learn proper video distributions and handle the inherent uncertainty from the nature of the raw video. Experimental results verify that our probabilistic embedding stands as a state-of-the-art video representation learning for action recognition and video retrieval on the most popular benchmarks, including UCF101 and HMDB51."
                    },
                    {
                        "title": "SimOn: A Simple Framework for Online Temporal Action Localization",
                        "abstract": "Online Temporal Action Localization (On-TAL) aims to immediately provide action instances from untrimmed streaming videos. The model is not allowed to utilize future frames and any processing techniques to modify past predictions, making On-TAL much more challenging. In this paper, we propose a simple yet effective framework, termed SimOn, that learns to predict action instances using the popular Transformer architecture in an end-to-end manner. Speci\ufb01cally, the model takes the current frame feature as a query and a set of past context information as keys and values of the Transformer. Different from the prior work that uses a set of outputs of the model as past contexts, we leverage the past visual context and the learnable context embedding for the current query. Experimental results on the THUMOS14 and ActivityNet1.3 datasets show that our model remarkably outperforms the previous methods, achieving a new state-of-the-art On-TAL performance. In addition, the evaluation for Online Detection of Action Start (ODAS) demonstrates the effectiveness and robustness of our method in the online setting. The code is available at https://github.com/TuanTNG/SimOn ."
                    },
                    {
                        "title": "Bridge to Answer: Structure-aware Graph Interaction Network for Video Question Answering",
                        "abstract": "This paper presents a novel method, termed Bridge to Answer, to infer correct answers for questions about a given video by leveraging adequate graph interactions of heterogeneous crossmodal graphs. To realize this, we learn question conditioned visual graphs by exploiting the relation between video and question to enable each visual node using question-to-visual interactions to encompass both visual and linguistic cues. In addition, we propose bridged visual-to-visual interactions to incorporate two complementary visual information on appearance and motion by placing the question graph as an intermediate bridge. This bridged architecture allows reliable message passing through compositional semantics of the question to generate an appropriate answer. As a result, our method can learn the question conditioned visual representations attributed to appearance and motion that show powerful capability for video question answering. Extensive experiments prove that the proposed method provides effective and superior performance than state-of-the-art methods on several benchmarks."
                    },
                    {
                        "title": "Self-Balanced Learning for Domain Generalization",
                        "abstract": "Domain generalization aims to learn a prediction model on multi-domain source data such that the model can generalize to a target domain with unknown statistics. Most existing approaches have been developed under the assumption that the source data is well-balanced in terms of both domain and class. However, real-world training data collected with different composition biases often exhibits severe distribution gaps for domain and class, leading to substantial performance degradation. In this paper, we propose a self-balanced domain generalization framework that adaptively learns the weights of losses to alleviate the bias caused by different distributions of the multi-domain source data. The self-balanced scheme is based on an auxiliary reweighting network that iteratively updates the weight of loss conditioned on the domain and class information by leveraging balanced meta data. Experimental results demonstrate the effectiveness of our method overwhelming state-of-the-art works for domain generalization."
                    },
                    {
                        "title": "Cross-Domain Grouping and Alignment for Domain Adaptive Semantic Segmentation",
                        "abstract": "Existing techniques to adapt semantic segmentation networks across source and target domains within deep convolutional neural networks (CNNs) deal with all the samples from the two domains in a global or category-aware manner. They do not consider an inter-class variation within the target domain itself or estimated category, providing the limitation to encode the domains having a multi-modal data distribution. To overcome this limitation, we introduce a learnable clustering module, and a novel domain adaptation framework, called cross-domain grouping and alignment. To cluster the samples across domains with an aim to maximize the domain alignment without forgetting precise segmentation ability on the source domain, we present two loss functions, in particular, for encouraging semantic consistency and orthogonality among the clusters. We also present a loss so as to solve a class imbalance problem, which is the other limitation of the previous methods. Our experiments show that our method consistently boosts the adaptation performance in semantic segmentation, outperforming the state-of-the-arts on various domain adaptation settings."
                    },
                    {
                        "title": "Video Summarization by Learning Relationships between Action and Scene",
                        "abstract": "We propose a novel deep architecture for video summarization in untrimmed videos that simultaneously recognizes action and scene classes for every video segments. Our networks accomplish this through a multi-task fusion approach based on two types of attention modules to explore semantic correlations between action and scene in the videos. The proposed networks consist of the feature embedding networks and attention inference networks to stochastically leverage the inferred action and scene feature representations. Additionally, we design a new center loss function that learns the feature representations by enforcing to minimize the intra-class variations and to maximize the inter-class variations. Our model achieves a score of 0.8409 for summarization and accuracy of 0.7294 for action and scene recognition on test set of CoVieW'19 dataset, which is ranked 3rd."
                    },
                    {
                        "title": "Context-Aware Emotion Recognition Networks",
                        "abstract": "Traditional techniques for emotion recognition have focused on the facial expression analysis only, thus providing limited ability to encode context that comprehensively represents the emotional responses. We present deep networks for context-aware emotion recognition, called CAER-Net, that exploit not only human facial expression but also context information in a joint and boosting manner. The key idea is to hide human faces in a visual scene and seek other contexts based on an attention mechanism. Our networks consist of two sub-networks, including two-stream encoding networks to separately extract the features of face and context regions, and adaptive fusion networks to fuse such features in an adaptive fashion. We also introduce a novel benchmark for context-aware emotion recognition, called CAER, that is appropriate than existing benchmarks both qualitatively and quantitatively. On several benchmarks, CAER-Net proves the effect of context for emotion recognition. Our dataset is available at http://caer-dataset.github.io."
                    },
                    {
                        "title": "Graph Regularization Network with Semantic Affinity for Weakly-Supervised Temporal Action Localization",
                        "abstract": "This paper presents a novel deep architecture for weakly-supervised temporal action localization that not only generates segment-level action responses but also propagates segment-level responses to the neighborhood in a form of graph Laplacian regularization. Specifically, our approach consists of two sub-modules; a class activation module to estimate the action score map over time through the action classifiers, and a graph regularization module to refine the estimated action score map by solving a quadratic programming problem with the predicted segment-level semantic affinities. Since these two modules are integrated with fully differentiable layers, the proposed networks can be jointly trained in an end-to-end manner. Experimental results on Thumos14 and ActivityNet1.2 demonstrate that the proposed method provides outstanding performances in weakly-supervised temporal action localization."
                    },
                    {
                        "title": "Learning to Detect, Associate, and Recognize Human Actions and Surrounding Scenes in Untrimmed Videos",
                        "abstract": "While recognizing human actions and surrounding scenes addresses different aspects of video understanding, they have strong correlations that can be used to complement the singular information of each other. In this paper, we propose an approach for joint action and scene recognition that is formulated in an end-to-end learning framework based on temporal attention techniques and the fusion of them. By applying temporal attention modules to the generic feature network, action and scene features are extracted efficiently, and then they are composed to a single feature vector through the proposed fusion module. Our experiments on the CoVieW18 dataset show that our model is able to detect temporal attention with only weak supervision, and remarkably improves multi-task action and scene classification accuracies."
                    }
                ]
            },
            "466a2a49-610e-4072-ada9-eb7717c78231": {
                "pk": "466a2a49-610e-4072-ada9-eb7717c78231",
                "name": "Jiyoung Lee",
                "collaborators": [
                    "K. Sohn",
                    "Jungin Park",
                    "Jin-Hwa Kim",
                    "Junyoung Seo",
                    "Seung Wook Kim",
                    "S. Park",
                    "Dongbo Min",
                    "Sunok Kim",
                    "Hanjae Kim",
                    "Kwanghoon Sohn"
                ],
                "domain": [
                    "Computer Vision",
                    "Video Understanding",
                    "Multi-modal Learning",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Prototype-guided Attention Distillation for Discriminative Person Search.",
                        "abstract": "Person search aims to localize a person of interest in a large image gallery captured by multiple, non-overlapping cameras. Prevalent unified methods have suffered from (1) noisy proposals with mis-detection and occlusion, and (2) large appearance variation within a class, which deteriorates the prototype-based metric learning. To address these problems, we introduce a Prototype-guided Attention Distillation, shortly PAD, which exploits a prototype (a typical representation of an identity) as a guidance to the attention module to consistently highlight identity-inherent regions across different poses. To utilize the knowledge encoded in prototypes for matching unseen IDs, PAD conducts attention distillation to guide student Re-ID queries by deeply mimicking attention maps from the prototype query. Additionally, to address large intra-class variation induced by pose or camera views, we extend PAD with multiple part prototypes representing consistent local regions across different instances. Furthermore, we exploit an adaptive momentum strategy for robust attention distillation in PAD to update more distinct prototypes. Extensive experiments conducted on CUHK-SYSU and PRW demonstrate the effectiveness of PAD, showcasing state-of-the-art performance. Moreover, our distilled attention surprisingly highlights distinguished multiple regions for person search."
                    },
                    {
                        "title": "Let 2D Diffusion Model Know 3D-Consistency for Robust Text-to-3D Generation",
                        "abstract": "Text-to-3D generation has shown rapid progress in recent days with the advent of score distillation, a methodology of using pretrained text-to-2D diffusion models to optimize neural radiance field (NeRF) in the zero-shot setting. However, the lack of 3D awareness in the 2D diffusion models destabilizes score distillation-based methods from reconstructing a plausible 3D scene. To address this issue, we propose 3DFuse, a novel framework that incorporates 3D awareness into pretrained 2D diffusion models, enhancing the robustness and 3D consistency of score distillation-based methods. We realize this by first constructing a coarse 3D structure of a given text prompt and then utilizing projected, view-specific depth map as a condition for the diffusion model. Additionally, we introduce a training strategy that enables the 2D diffusion model learns to handle the errors and sparsity within the coarse 3D structure for robust generation, as well as a method for ensuring semantic consistency throughout all viewpoints of the scene. Our framework surpasses the limitations of prior arts, and has significant implications for 3D consistent generation of 2D diffusion models."
                    },
                    {
                        "title": "Semi-Parametric Video-Grounded Text Generation",
                        "abstract": "Efficient video-language modeling should consider the computational cost because of a large, sometimes intractable, number of video frames. Parametric approaches such as the attention mechanism may not be ideal since its computational cost quadratically increases as the video length increases. Rather, previous studies have relied on offline feature extraction or frame sampling to represent the video efficiently, focusing on cross-modal modeling in short video clips. In this paper, we propose a semi-parametric video-grounded text generation model, SeViT, a novel perspective on scalable video-language modeling toward long untrimmed videos. Treating a video as an external data store, SeViT includes a non-parametric frame retriever to select a few query-relevant frames from the data store for a given query and a parametric generator to effectively aggregate the frames with the query via late fusion methods. Experimental results demonstrate our method has a significant advantage in longer videos and causal video understanding. Moreover, our model achieves the new state of the art on four video-language datasets, iVQA (+4.8), Next-QA (+6.9), and Activitynet-QA (+4.8) in accuracy, and MSRVTT-Caption (+3.6) in CIDEr."
                    },
                    {
                        "title": "Dual-Path Adaptation from Image to Video Transformers",
                        "abstract": "In this paper, we efficiently transfer the surpassing representation power of the vision foundation models, such as ViT and Swin, for video understanding with only a few trainable parameters. Previous adaptation methods have simultaneously considered spatial and temporal modeling with a unified learnable module but still suffered from fully leveraging the representative capabilities of image transformers. We argue that the popular dual-path (two-stream) architecture in video models can mitigate this problem. We propose a novel DualPath adaptation separated into spatial and temporal adaptation paths, where a lightweight bottleneck adapter is employed in each transformer block. Especially for temporal dynamic modeling, we incorporate consecutive frames into a grid-like frameset to precisely imitate vision transformers' capability that extrapolates relationships between tokens. In addition, we extensively investigate the multiple baselines from a unified perspective in video understanding and compare them with DualPath. Experimental results on four action recognition benchmarks prove that pretrained image transformers with DualPath can be effectively generalized beyond the data domain."
                    },
                    {
                        "title": "Dense Text-to-Image Generation with Attention Modulation",
                        "abstract": "Existing text-to-image diffusion models struggle to synthesize realistic images given dense captions, where each text prompt provides a detailed description for a specific image region. To address this, we propose DenseDiffusion, a training-free method that adapts a pre-trained text-to-image model to handle such dense captions while offering control over the scene layout. We first analyze the relationship between generated images\u2019 layouts and the pre-trained model\u2019s intermediate attention maps. Next, we develop an attention modulation method that guides objects to appear in specific regions according to layout guidance. Without requiring additional fine-tuning or datasets, we improve image generation performance given dense captions regarding both automatic and human evaluation scores. In addition, we achieve similar-quality visual results with models specifically trained with layout conditions. Code and data are available at https://github.com/naver-ai/DenseDiffusion."
                    },
                    {
                        "title": "Hierarchical Visual Primitive Experts for Compositional Zero-Shot Learning",
                        "abstract": "Compositional zero-shot learning (CZSL) aims to recognize unseen compositions with prior knowledge of known primitives (attribute and object). Previous works for CZSL often suffer from grasping the contextuality between attribute and object, as well as the discriminability of visual features, and the long-tailed distribution of real-world compositional data. We propose a simple and scalable framework called Composition Transformer (CoT) to address these issues. CoT employs object and attribute experts in distinctive manners to generate representative embeddings, using the visual network hierarchically. The object expert extracts representative object embeddings from the final layer in a bottom-up manner, while the attribute expert makes attribute embeddings in a top-down manner with a proposed object-guided attention module that models contextuality explicitly. To remedy biased prediction caused by imbalanced data distribution, we develop a simple minority attribute augmentation (MAA) that synthesizes virtual samples by mixing two images and oversampling minority attribute classes. Our method achieves SoTA performance on several benchmarks, including MIT-States, C-GQA, and VAW-CZSL. We also demonstrate the effectiveness of CoT in improving visual discrimination and addressing the model bias from the imbalanced data distribution. The code is available at https://github.com/HanjaeKim98/CoT."
                    },
                    {
                        "title": "Pin the Memory: Learning to Generalize Semantic Segmentation",
                        "abstract": "The rise of deep neural networks has led to several break-throughs for semantic segmentation. In spite of this, a model trained on source domain often fails to work properly in new challenging domains, that is directly concerned with the generalization capability of the model. In this paper, we present a novel memory-guided domain generalization method for semantic segmentation based on meta-learning framework. Especially, our method abstracts the conceptual knowledge of semantic classes into categorical memory which is constant beyond the domains. Upon the meta-learning concept, we repeatedly train memory-guided networks and simulate virtual test to 1) learn how to memorize a domain-agnostic and distinct information of classes and 2) offer an externally settled memory as a class-guidance to reduce the ambiguity of representation in the test data of arbitrary unseen domain. To this end, we also propose memory divergence and feature cohesion losses, which encourage to learn memory reading and update processes for category-aware domain generalization. Extensive experiments for semantic segmentation demonstrate the superior generalization capability of our method over state-of-the-art works on various benchmarks.11https://github.com/Genie-Kim/PintheMemory"
                    },
                    {
                        "title": "MIDMs: Matching Interleaved Diffusion Models for Exemplar-based Image Translation",
                        "abstract": "We present a novel method for exemplar-based image translation, called matching interleaved diffusion models (MIDMs). Most existing methods for this task were formulated as GAN-based matching-then-generation framework. However, in this framework, matching errors induced by the difficulty of semantic matching across cross-domain, e.g., sketch and photo, can be easily propagated to the generation step, which in turn leads to the degenerated results. Motivated by the recent success of diffusion models, overcoming the shortcomings of GANs, we incorporate the diffusion models to overcome these limitations. Specifically, we formulate a diffusion-based matching-and-generation framework that interleaves cross-domain matching and diffusion steps in the latent space by iteratively feeding the intermediate warp into the noising process and denoising it to generate a translated image. In addition, to improve the reliability of diffusion process, we design confidence-aware process using cycle-consistency to consider only confident regions during translation. Experimental results show that our MIDMs generate more plausible images than state-of-the-art methods."
                    },
                    {
                        "title": "PointFix: Learning to Fix Domain Bias for Robust Online Stereo Adaptation",
                        "abstract": "Online stereo adaptation tackles the domain shift problem, caused by different environments between synthetic (training) and real (test) datasets, to promptly adapt stereo models in dynamic real-world applications such as autonomous driving. However, previous methods often fail to counteract particular regions related to dynamic objects with more severe environmental changes. To mitigate this issue, we propose to incorporate an auxiliary point-selective network into a meta-learning framework, called PointFix, to provide a robust initialization of stereo models for online stereo adaptation. In a nutshell, our auxiliary network learns to fix local variants intensively by effectively back-propagating local information through the meta-gradient for the robust initialization of the baseline model. This network is model-agnostic, so can be used in any kind of architectures in a plug-and-play manner. We conduct extensive experiments to verify the effectiveness of our method under three adaptation settings such as short-, mid-, and long-term sequences. Experimental results show that the proper initialization of the base stereo model by the auxiliary network enables our learning paradigm to achieve state-of-the-art performance at inference."
                    },
                    {
                        "title": "Multi-Domain Unsupervised Image-to-Image Translation with Appearance Adaptive Convolution",
                        "abstract": "Over the past few years, image-to-image (I2I) translation methods have been proposed to translate a given image into diverse outputs. Despite the impressive results, they mainly focus on the I2I translation between two domains, so the multi-domain I2I translation still remains a challenge. To address this problem, we propose a novel multi-domain unsupervised image-to-image translation (MDUIT) framework that leverages the decomposed content feature and appearance adaptive convolution to translate an image into a target appearance while preserving the given geometric content. We also exploit a contrast learning objective, which improves the disentanglement ability and effectively utilizes multi-domain image data in the training process by pairing the semantically similar images. This allows our method to learn the diverse mappings between multiple visual domains with only a single framework. We show that the proposed method produces visually diverse and plausible results in multiple domains compared to the state-of-the-art methods."
                    },
                    {
                        "title": "Language-free Training for Zero-shot Video Grounding",
                        "abstract": "Given an untrimmed video and a language query depicting a specific temporal moment in the video, video grounding aims to localize the time interval by understanding the text and video simultaneously. One of the most challenging issues is an extremely time- and cost-consuming annotation collection, including video captions in a natural language form and their corresponding temporal regions. In this paper, we present a simple yet novel training framework for video grounding in the zero-shot setting, which learns a network with only video data without any annotation. Inspired by the recent language-free paradigm, i.e. training without language data, we train the network without compelling the generation of fake (pseudo) text queries into a natural language form. Specifically, we propose a method for learning a video grounding model by selecting a temporal interval as a hypothetical correct answer and considering the visual feature selected by our method in the interval as a language feature, with the help of the well-aligned visual-language space of CLIP. Extensive experiments demonstrate the prominence of our language-free training framework, outperforming the existing zero-shot video grounding method and even several weakly-supervised approaches with large margins on two standard datasets."
                    },
                    {
                        "title": "Probabilistic Representations for Video Contrastive Learning",
                        "abstract": "This paper presents Probabilistic Video Contrastive Learning, a self-supervised representation learning method that bridges contrastive learning with probabilistic representation. We hypothesize that the clips composing the video have different distributions in short-term duration, but can represent the complicated and sophisticated video distribution through combination in a common embedding space. Thus, the proposed method represents video clips as normal distributions and combines them into a Mixture of Gaussians to model the whole video distribution. By sampling embeddings from the whole video distribution, we can circumvent the careful sampling strategy or transformations to generate augmented views of the clips, unlike previous deterministic methods that have mainly focused on such sample generation strategies for contrastive learning. We further propose a stochastic contrastive loss to learn proper video distributions and handle the inherent uncertainty from the nature of the raw video. Experimental results verify that our probabilistic embedding stands as a state-of-the-art video representation learning for action recognition and video retrieval on the most popular benchmarks, including UCF101 and HMDB51."
                    },
                    {
                        "title": "Bridge to Answer: Structure-aware Graph Interaction Network for Video Question Answering",
                        "abstract": "This paper presents a novel method, termed Bridge to Answer, to infer correct answers for questions about a given video by leveraging adequate graph interactions of heterogeneous crossmodal graphs. To realize this, we learn question conditioned visual graphs by exploiting the relation between video and question to enable each visual node using question-to-visual interactions to encompass both visual and linguistic cues. In addition, we propose bridged visual-to-visual interactions to incorporate two complementary visual information on appearance and motion by placing the question graph as an intermediate bridge. This bridged architecture allows reliable message passing through compositional semantics of the question to generate an appropriate answer. As a result, our method can learn the question conditioned visual representations attributed to appearance and motion that show powerful capability for video question answering. Extensive experiments prove that the proposed method provides effective and superior performance than state-of-the-art methods on several benchmarks."
                    },
                    {
                        "title": "CausalCity: Complex Simulations with Agency for Causal Discovery and Reasoning",
                        "abstract": "The ability to perform causal and counterfactual reasoning are central properties of human intelligence. Decision-making systems that can perform these types of reasoning have the potential to be more generalizable and interpretable. Simulations have helped advance the state-of-the-art in this domain, by providing the ability to systematically vary parameters (e.g., confounders) and generate examples of the outcomes in the case of counterfactual scenarios. However, simulating complex temporal causal events in multi-agent scenarios, such as those that exist in driving and vehicle navigation, is challenging. To help address this, we present a high-fidelity simulation environment that is designed for developing algorithms for causal discovery and counterfactual reasoning in the safety-critical context. A core component of our work is to introduce \\textit{agency}, such that it is simple to define and create complex scenarios using high-level definitions. The vehicles then operate with agency to complete these objectives, meaning low-level behaviors need only be controlled if necessary. We perform experiments with three state-of-the-art methods to create baselines and highlight the affordances of this environment. Finally, we highlight challenges and opportunities for future work."
                    },
                    {
                        "title": "Looking into Your Speech: Learning Cross-modal Affinity for Audio-visual Speech Separation",
                        "abstract": "In this paper, we address the problem of separating individual speech signals from videos using audio-visual neural processing. Most conventional approaches utilize frame-wise matching criteria to extract shared information between co-occurring audio and video. Thus, their performance heavily depends on the accuracy of audio-visual synchronization and the effectiveness of their representations. To overcome the frame discontinuity problem between two modalities due to transmission delay mismatch or jitter, we propose a cross-modal affinity network (CaffNet) that learns global correspondence as well as locally-varying affinities between audio and visual streams. Given that the global term provides stability over a temporal sequence at the utterance-level, this resolves the label permutation problem characterized by inconsistent assignments. By extending the proposed cross-modal affinity on the complex network, we further improve the separation performance in the complex spectral domain. Experimental results verify that the proposed methods outperform conventional ones on various datasets, demonstrating their advantages in real-world scenarios."
                    },
                    {
                        "title": "Multi-Modal Recurrent Attention Networks for Facial Expression Recognition",
                        "abstract": "Recent deep neural networks based methods have achieved state-of-the-art performance on various facial expression recognition tasks. Despite such progress, previous researches for facial expression recognition have mainly focused on analyzing color recording videos only. However, the complex emotions expressed by people with different skin colors under different lighting conditions through dynamic facial expressions can be fully understandable by integrating information from multi-modal videos. We present a novel method to estimate dimensional emotion states, where color, depth, and thermal recording videos are used as a multi-modal input. Our networks, called multi-modal recurrent attention networks (MRAN), learn spatiotemporal attention volumes to robustly recognize the facial expression based on attention-boosted feature volumes. We leverage the depth and thermal sequences as guidance priors for color sequence to selectively focus on emotional discriminative regions. We also introduce a novel benchmark for multi-modal facial expression recognition, termed as multi-modal arousal-valence facial expression recognition (MAVFER), which consists of color, depth, and thermal recording videos with corresponding continuous arousal-valence scores. The experimental results show that our method can achieve the state-of-the-art results in dimensional facial expression recognition on color recording datasets including RECOLA, SEWA and AFEW, and a multi-modal recording dataset including MAVFER."
                    }
                ]
            },
            "e15d33b7-2d26-455a-a706-ec414b7727aa": {
                "pk": "e15d33b7-2d26-455a-a706-ec414b7727aa",
                "name": "Kwanghoon Sohn",
                "collaborators": [
                    "Hyeongjun Kwon",
                    "Jin Kim",
                    "Ilhoon Yoon",
                    "Junyoung Park",
                    "Jinhyun Jang",
                    "Kwonyoung Kim",
                    "Hanjae Kim",
                    "Jiyoung Lee",
                    "Hyunsung Jang",
                    "S. Park"
                ],
                "domain": [
                    "Computer Vision",
                    "Domain Adaptation",
                    "Semi-Supervised Learning",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Improving Visual Recognition with Hyperbolical Visual Hierarchy Mapping",
                        "abstract": "Visual scenes are naturally organized in a hierarchy, where a coarse semantic is recursively comprised of several fine details. Exploring such a visual hierarchy is crucial to recognize the complex relations of visual elements, leading to a comprehensive scene understanding. In this paper, we propose a Visual Hierarchy Mapper (Hi-Mapper), a novel approach for enhancing the structured understanding of the pre-trained Deep Neural Networks (DNNs). Hi-Mapper investigates the hierarchical organization of the visual scene by 1) pre-defining a hierarchy tree through the encapsulation of probability densities; and 2) learning the hierarchical relations in hyperbolic space with a novel hierarchical contrastive loss. The pre-defined hierarchy tree recursively interacts with the visual features of the pre-trained DNNs through hierarchy decomposition and encoding procedures, thereby effectively identifying the visual hierarchy and enhancing the recognition of an entire scene. Extensive experiments demonstrate that Hi-Mapper significantly enhances the representation capability of DNNs, leading to an improved performance on various tasks, including image classification and dense prediction tasks. The code is available at https://github.com/kwonjunn0l/Hi-Mapper."
                    },
                    {
                        "title": "Prototype-guided Attention Distillation for Discriminative Person Search.",
                        "abstract": "Person search aims to localize a person of interest in a large image gallery captured by multiple, non-overlapping cameras. Prevalent unified methods have suffered from (1) noisy proposals with mis-detection and occlusion, and (2) large appearance variation within a class, which deteriorates the prototype-based metric learning. To address these problems, we introduce a Prototype-guided Attention Distillation, shortly PAD, which exploits a prototype (a typical representation of an identity) as a guidance to the attention module to consistently highlight identity-inherent regions across different poses. To utilize the knowledge encoded in prototypes for matching unseen IDs, PAD conducts attention distillation to guide student Re-ID queries by deeply mimicking attention maps from the prototype query. Additionally, to address large intra-class variation induced by pose or camera views, we extend PAD with multiple part prototypes representing consistent local regions across different instances. Furthermore, we exploit an adaptive momentum strategy for robust attention distillation in PAD to update more distinct prototypes. Extensive experiments conducted on CUHK-SYSU and PRW demonstrate the effectiveness of PAD, showcasing state-of-the-art performance. Moreover, our distilled attention surprisingly highlights distinguished multiple regions for person search."
                    },
                    {
                        "title": "Enhancing Source-Free Domain Adaptive Object Detection with Low-confidence Pseudo Label Distillation",
                        "abstract": "Source-Free domain adaptive Object Detection (SFOD) is a promising strategy for deploying trained detectors to new, unlabeled domains without accessing source data, addressing significant concerns around data privacy and efficiency. Most SFOD methods leverage a Mean-Teacher (MT) self-training paradigm relying heavily on High-confidence Pseudo Labels (HPL). However, these HPL often overlook small instances that undergo significant appearance changes with domain shifts. Additionally, HPL ignore instances with low confidence due to the scarcity of training samples, resulting in biased adaptation toward familiar instances from the source domain. To address this limitation, we introduce the Low-confidence Pseudo Label Distillation (LPLD) loss within the Mean-Teacher based SFOD framework. This novel approach is designed to leverage the proposals from Region Proposal Network (RPN), which potentially encompasses hard-to-detect objects in unfamiliar domains. Initially, we extract HPL using a standard pseudo-labeling technique and mine a set of Low-confidence Pseudo Labels (LPL) from proposals generated by RPN, leaving those that do not overlap significantly with HPL. These LPL are further refined by leveraging class-relation information and reducing the effect of inherent noise for the LPLD loss calculation. Furthermore, we use feature distance to adaptively weight the LPLD loss to focus on LPL containing a larger foreground area. Our method outperforms previous SFOD methods on four cross-domain object detection benchmarks. Extensive experiments demonstrate that our LPLD loss leads to effective adaptation by reducing false negatives and facilitating the use of domain-invariant knowledge from the source model. Code is available at https://github.com/junia3/LPLD."
                    },
                    {
                        "title": "Rethinking Open-World Semi-Supervised Learning: Distribution Mismatch and Inductive Inference",
                        "abstract": "Open-world semi-supervised learning (OWSSL) extends conventional semi-supervised learning to open-world scenarios by taking account of novel categories in unlabeled datasets. Despite the recent advancements in OWSSL, the success often relies on the assumptions that 1) labeled and unlabeled datasets share the same balanced class prior distribution, which does not generally hold in real-world applications, and 2) unlabeled training datasets are utilized for evaluation, where such transductive inference might not adequately address challenges in the wild. In this paper, we aim to generalize OWSSL by addressing them. Our work suggests that practical OWSSL may require different training settings, evaluation methods, and learning strategies compared to those prevalent in the existing literature."
                    },
                    {
                        "title": "Layer-wise Auto-Weighting for Non-Stationary Test-Time Adaptation",
                        "abstract": "Given the inevitability of domain shifts during inference in real-world applications, test-time adaptation (TTA) is essential for model adaptation after deployment. However, the real-world scenario of continuously changing target distributions presents challenges including catastrophic forgetting and error accumulation. Existing TTA methods for non-stationary domain shifts, while effective, incur excessive computational load, making them impractical for on-device settings. In this paper, we introduce a layer-wise auto-weighting algorithm for continual and gradual TTA that autonomously identifies layers for preservation or concentrated adaptation. By leveraging the Fisher Information Matrix (FIM), we first design the learning weight to selectively focus on layers associated with log-likelihood changes while preserving unrelated ones. Then, we further propose an exponential min-max scaler to make certain layers nearly frozen while mitigating outliers. This minimizes forgetting and error accumulation, leading to efficient adaptation to non-stationary target distribution. Experiments on CIFAR-10C, CIFAR-100C, and ImageNet-C show our method outperforms conventional continual and gradual TTA approaches while significantly reducing computational load, highlighting the importance of FIM-based learning weight in adapting to continuously or gradually shifting target domains.1"
                    }
                ]
            }
        }
    },
    "2310.06213": {
        "paper_data": {
            "title": "GeoLLM: Extracting Geospatial Knowledge from Large Language Models",
            "url": "http://arxiv.org/abs/2310.06213v2",
            "arxiv_id": "2310.06213",
            "authors": [
                "Rohin Manvi",
                "Samar Khanna",
                "Gengchen Mai",
                "Marshall Burke",
                "David Lobell",
                "Stefano Ermon"
            ],
            "abstract": "The application of machine learning (ML) in a range of geospatial tasks is increasingly common but often relies on globally available covariates such as satellite imagery that can either be expensive or lack predictive power. Here we explore the question of whether the vast amounts of knowledge found in Internet language corpora, now compressed within large language models (LLMs), can be leveraged for geospatial prediction tasks. We first demonstrate that LLMs embed remarkable spatial information about locations, but naively querying LLMs using geographic coordinates alone is ineffective in predicting key indicators like population density. We then present GeoLLM, a novel method that can effectively extract geospatial knowledge from LLMs with auxiliary map data from OpenStreetMap. We demonstrate the utility of our approach across multiple tasks of central interest to the international community, including the measurement of population density and economic livelihoods. Across these tasks, our method demonstrates a 70% improvement in performance (measured using Pearson's $r^2$) relative to baselines that use nearest neighbors or use information directly from the prompt, and performance equal to or exceeding satellite-based benchmarks in the literature. With GeoLLM, we observe that GPT-3.5 outperforms Llama 2 and RoBERTa by 19% and 51% respectively, suggesting that the performance of our method scales well with the size of the model and its pretraining dataset. Our experiments reveal that LLMs are remarkably sample-efficient, rich in geospatial information, and robust across the globe. Crucially, GeoLLM shows promise in mitigating the limitations of existing geospatial covariates and complementing them well. Code is available on the project website: https://rohinmanvi.github.io/GeoLLM",
            "introduction": " introduction of instruction-finetuned LLMs, directly querying knowledge with prompts is a potentially simpler method. In this paper, we focus on showing the quantity and quality of geospatial knowledge in pre-trained language models. Knowledge Extraction Many works aim to improve knowledge extraction from language models by fine-tuning or prompt tuning. Works on prompt tuning explore mining prompts from the web (Jiang et al., 2019), optimizing prompts in the discrete space of words and tokens (Shin et al., 2020), optimizing prompts in the continuous embedding space (Zhong et al., 2021), and using adapters (Newman et al., 2021). However, prompts are often difficult to craft (Qin & Eisner, 2021; Jiang et al., 2019), and are sensitive to small changes. Fine-tuning language models for downstream tasks has shown to be an effective and simple method to tune and elicit specific knowledge for evaluation (Radford & Narasimhan, 2018; Dong et al., 2019). It is also shown that fine-tuning may result in higher performance gains compared to prompt tuning (Fichtel et al., 2021). Additionally, LoRA and quantization are shown to greatly reduce the computational cost of fine-tuning (Hu et al., 2021; Dettmers et al., 2023). For these reasons, we choose to extract geospatial knowledge with fine- tuning. NLP for Geospatial Tasks Natural language processing (NLP) has been used for many geospatial tasks with various data sources. Hu et al. (2017) extracted semantics relations among cities by lever- aging news articles. Sheehan et al. (2019) utilized Wikipedia to predict poverty. Xu et al. (2016) accessed traffic by using OpenStreetMap data. Signorini et al. (2011); Paul & Dredze (2011) utilized tweets to predict track-levels disease activities. Recently, researchers started to explore LLMs\u2019 ca- pability on various geospatial tasks. Mai et al. (2023a) demonstrated the usability of large language models on various geospatial applications including fine-grained address recognition, time-series dementia record forecasting, urban function prediction, and so on. GeoGPT (Zhang et al., 2023) has been proposed as a GPT-3.5-based autonomous AI tool that can conduct geospatial data collection, processing, and analysis in an autonomous manner with the instruction of only natural language. However, both works mainly relied on the pre-trained LLMs and did not explore the possibility of fine-tuning LLMs for a geo-specific foundation model. Deng et al. (2023) developed an LLM in geoscience, so-called K2, by fine-tuning on a geoscience text corpus which has shown remark- able performances on various NLP tasks in the geoscience domain. However, the K2 geoscience language foundation model is still limited to the common NLP tasks such as question answering, summarization, text classification, etc. In contrast, in this work, we conduct a deep investigation into the possibility of extracting geospatial knowledge from LLMs and whether we can use it for various real-world geospatial tasks like predicting population density. By fine-tuning various LLMs, we demonstrate the quantity, quality, and scalable nature of geospatial knowledge contained in LLMs. 3 M ETHOD Abstractly, we want to map geographic coordinates (latitude, longitude) and additional features to a response variable, such as population density or asset wealth. A naive approach would be to spatially interpolate using the coordinates. However, this is far from ideal, especially for small sample sizes. Features or covariates can help, but such data can be difficult to obtain. Given this challenge, we aim to determine how much LLMs know about these coordinates and how",
            "references": [
                {
                    "title": "Language Modeling Is Compression",
                    "abstract": "It has long been established that predictive models can be transformed into lossless compressors and vice versa. Incidentally, in recent years, the machine learning community has focused on training increasingly large and powerful self-supervised (language) models. Since these large language models exhibit impressive predictive capabilities, they are well-positioned to be strong compressors. In this work, we advocate for viewing the prediction problem through the lens of compression and evaluate the compression capabilities of large (foundation) models. We show that large language models are powerful general-purpose predictors and that the compression viewpoint provides novel insights into scaling laws, tokenization, and in-context learning. For example, Chinchilla 70B, while trained primarily on text, compresses ImageNet patches to 43.4% and LibriSpeech samples to 16.4% of their raw size, beating domain-specific compressors like PNG (58.5%) or FLAC (30.3%), respectively. Finally, we show that the prediction-compression equivalence allows us to use any compressor (like gzip) to build a conditional generative model."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "GeoGPT: Understanding and Processing Geospatial Tasks through An Autonomous GPT",
                    "abstract": "Decision-makers in GIS need to combine a series of spatial algorithms and operations to solve geospatial tasks. For example, in the task of facility siting, the Buffer tool is usually first used to locate areas close or away from some specific entities; then, the Intersect or Erase tool is used to select candidate areas satisfied multiple requirements. Though professionals can easily understand and solve these geospatial tasks by sequentially utilizing relevant tools, it is difficult for non-professionals to handle these problems. Recently, Generative Pre-trained Transformer (e.g., ChatGPT) presents strong performance in semantic understanding and reasoning. Especially, AutoGPT can further extend the capabilities of large language models (LLMs) by automatically reasoning and calling externally defined tools. Inspired by these studies, we attempt to lower the threshold of non-professional users to solve geospatial tasks by integrating the semantic understanding ability inherent in LLMs with mature tools within the GIS community. Specifically, we develop a new framework called GeoGPT that can conduct geospatial data collection, processing, and analysis in an autonomous manner with the instruction of only natural language. In other words, GeoGPT is used to understand the demands of non-professional users merely based on input natural language descriptions, and then think, plan, and execute defined GIS tools to output final effective results. Several cases including geospatial data crawling, spatial query, facility siting, and mapping validate the effectiveness of our framework. Though limited cases are presented in this paper, GeoGPT can be further extended to various tasks by equipping with more GIS tools, and we think the paradigm of\"foundational plus professional\"implied in GeoGPT provides an effective way to develop next-generation GIS in this era of large foundation models."
                },
                {
                    "title": "Sphere2Vec: A General-Purpose Location Representation Learning over a Spherical Surface for Large-Scale Geospatial Predictions",
                    "abstract": "Generating learning-friendly representations for points in space is a fundamental and long-standing problem in ML. Recently, multi-scale encoding schemes (such as Space2Vec and NeRF) were proposed to directly encode any point in 2D/3D Euclidean space as a high-dimensional vector, and has been successfully applied to various geospatial prediction and generative tasks. However, all current 2D and 3D location encoders are designed to model point distances in Euclidean space. So when applied to large-scale real-world GPS coordinate datasets, which require distance metric learning on the spherical surface, both types of models can fail due to the map projection distortion problem (2D) and the spherical-to-Euclidean distance approximation error (3D). To solve these problems, we propose a multi-scale location encoder called Sphere2Vec which can preserve spherical distances when encoding point coordinates on a spherical surface. We developed a unified view of distance-reserving encoding on spheres based on the DFS. We also provide theoretical proof that the Sphere2Vec preserves the spherical surface distance between any two points, while existing encoding schemes do not. Experiments on 20 synthetic datasets show that Sphere2Vec can outperform all baseline models on all these datasets with up to 30.8% error rate reduction. We then apply Sphere2Vec to three geo-aware image classification tasks - fine-grained species recognition, Flickr image recognition, and remote sensing image classification. Results on 7 real-world datasets show the superiority of Sphere2Vec over multiple location encoders on all three tasks. Further analysis shows that Sphere2Vec outperforms other location encoder models, especially in the polar regions and data-sparse areas because of its nature for spherical surface distance preservation. Code and data are available at https://gengchenmai.github.io/sphere2vec-website/."
                },
                {
                    "title": "QLoRA: Efficient Finetuning of Quantized LLMs",
                    "abstract": "We present QLoRA, an efficient finetuning approach that reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU while preserving full 16-bit finetuning task performance. QLoRA backpropagates gradients through a frozen, 4-bit quantized pretrained language model into Low Rank Adapters~(LoRA). Our best model family, which we name Guanaco, outperforms all previous openly released models on the Vicuna benchmark, reaching 99.3% of the performance level of ChatGPT while only requiring 24 hours of finetuning on a single GPU. QLoRA introduces a number of innovations to save memory without sacrificing performance: (a) 4-bit NormalFloat (NF4), a new data type that is information theoretically optimal for normally distributed weights (b) double quantization to reduce the average memory footprint by quantizing the quantization constants, and (c) paged optimziers to manage memory spikes. We use QLoRA to finetune more than 1,000 models, providing a detailed analysis of instruction following and chatbot performance across 8 instruction datasets, multiple model types (LLaMA, T5), and model scales that would be infeasible to run with regular finetuning (e.g. 33B and 65B parameter models). Our results show that QLoRA finetuning on a small high-quality dataset leads to state-of-the-art results, even when using smaller models than the previous SoTA. We provide a detailed analysis of chatbot performance based on both human and GPT-4 evaluations showing that GPT-4 evaluations are a cheap and reasonable alternative to human evaluation. Furthermore, we find that current chatbot benchmarks are not trustworthy to accurately evaluate the performance levels of chatbots. A lemon-picked analysis demonstrates where Guanaco fails compared to ChatGPT. We release all of our models and code, including CUDA kernels for 4-bit training."
                },
                {
                    "title": "CSP: Self-Supervised Contrastive Spatial Pre-Training for Geospatial-Visual Representations",
                    "abstract": "Geo-tagged images are publicly available in large quantities, whereas labels such as object classes are rather scarce and expensive to collect. Meanwhile, contrastive learning has achieved tremendous success in various natural image and language tasks with limited labeled data. However, existing methods fail to fully leverage geospatial information, which can be paramount to distinguishing objects that are visually similar. To directly leverage the abundant geospatial information associated with images in pre-training, fine-tuning, and inference stages, we present Contrastive Spatial Pre-Training (CSP), a self-supervised learning framework for geo-tagged images. We use a dual-encoder to separately encode the images and their corresponding geo-locations, and use contrastive objectives to learn effective location representations from images, which can be transferred to downstream supervised tasks such as image classification. Experiments show that CSP can improve model performance on both iNat2018 and fMoW datasets. Especially, on iNat2018, CSP significantly boosts the model performance with 10-34% relative improvement with various labeled training data sampling ratios."
                },
                {
                    "title": "On the Opportunities and Challenges of Foundation Models for Geospatial Artificial Intelligence",
                    "abstract": "Large pre-trained models, also known as foundation models (FMs), are trained in a task-agnostic manner on large-scale data and can be adapted to a wide range of downstream tasks by fine-tuning, few-shot, or even zero-shot learning. Despite their successes in language and vision tasks, we have yet seen an attempt to develop foundation models for geospatial artificial intelligence (GeoAI). In this work, we explore the promises and challenges of developing multimodal foundation models for GeoAI. We first investigate the potential of many existing FMs by testing their performances on seven tasks across multiple geospatial subdomains including Geospatial Semantics, Health Geography, Urban Geography, and Remote Sensing. Our results indicate that on several geospatial tasks that only involve text modality such as toponym recognition, location description recognition, and US state-level/county-level dementia time series forecasting, these task-agnostic LLMs can outperform task-specific fully-supervised models in a zero-shot or few-shot learning setting. However, on other geospatial tasks, especially tasks that involve multiple data modalities (e.g., POI-based urban function classification, street view image-based urban noise intensity classification, and remote sensing image scene classification), existing foundation models still underperform task-specific models. Based on these observations, we propose that one of the major challenges of developing a FM for GeoAI is to address the multimodality nature of geospatial tasks. After discussing the distinct challenges of each geospatial data modality, we suggest the possibility of a multimodal foundation model which can reason over various types of geospatial data through geospatial alignments. We conclude this paper by discussing the unique risks and challenges to develop such a model for GeoAI."
                },
                {
                    "title": "A Survey of Large Language Models",
                    "abstract": "Language is essentially a complex, intricate system of human expressions governed by grammatical rules. It poses a significant challenge to develop capable AI algorithms for comprehending and grasping a language. As a major approach, language modeling has been widely studied for language understanding and generation in the past two decades, evolving from statistical language models to neural language models. Recently, pre-trained language models (PLMs) have been proposed by pre-training Transformer models over large-scale corpora, showing strong capabilities in solving various NLP tasks. Since researchers have found that model scaling can lead to performance improvement, they further study the scaling effect by increasing the model size to an even larger size. Interestingly, when the parameter scale exceeds a certain level, these enlarged language models not only achieve a significant performance improvement but also show some special abilities that are not present in small-scale language models. To discriminate the difference in parameter scale, the research community has coined the term large language models (LLM) for the PLMs of significant size. Recently, the research on LLMs has been largely advanced by both academia and industry, and a remarkable progress is the launch of ChatGPT, which has attracted widespread attention from society. The technical evolution of LLMs has been making an important impact on the entire AI community, which would revolutionize the way how we develop and use AI algorithms. In this survey, we review the recent advances of LLMs by introducing the background, key findings, and mainstream techniques. In particular, we focus on four major aspects of LLMs, namely pre-training, adaptation tuning, utilization, and capacity evaluation. Besides, we also summarize the available resources for developing LLMs and discuss the remaining issues for future directions."
                },
                {
                    "title": "ChatGPT Is a Knowledgeable but Inexperienced Solver: An Investigation of Commonsense Problem in Large Language Models",
                    "abstract": "Large language models (LLMs) have made significant progress in NLP. However, their ability to memorize, represent, and leverage commonsense knowledge has been a well-known pain point. In this paper, we specifically focus on ChatGPT, a widely used and easily accessible LLM, and ask the following questions: (1) Can ChatGPT effectively answer commonsense questions? (2) Is ChatGPT aware of the underlying commonsense knowledge for answering a specific question? (3) Is ChatGPT knowledgeable in commonsense? (4) Can ChatGPT effectively leverage commonsense for answering questions? We conduct a series of experiments on 11 datasets to evaluate ChatGPT\u2019s commonsense abilities, including answering commonsense questions, identifying necessary knowledge, generating knowledge descriptions, and using knowledge descriptions to answer questions again. Experimental results show that: (1) ChatGPT can achieve good QA accuracies in commonsense tasks, while still struggling with certain domains of datasets. (2) ChatGPT is knowledgeable, and can accurately generate most of the commonsense knowledge using knowledge prompts. (3) Despite its knowledge, ChatGPT is an inexperienced commonsense problem solver, which cannot precisely identify the needed commonsense for answering a specific question. These findings raise the need to explore improved mechanisms for effectively incorporating commonsense into LLMs like ChatGPT, such as better instruction following and commonsense guidance."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "SustainBench: Benchmarks for Monitoring the Sustainable Development Goals with Machine Learning",
                    "abstract": "Progress toward the United Nations Sustainable Development Goals (SDGs) has been hindered by a lack of data on key environmental and socioeconomic indicators, which historically have come from ground surveys with sparse temporal and spatial coverage. Recent advances in machine learning have made it possible to utilize abundant, frequently-updated, and globally available data, such as from satellites or social media, to provide insights into progress toward SDGs. Despite promising early results, approaches to using such data for SDG measurement thus far have largely evaluated on different datasets or used inconsistent evaluation metrics, making it hard to understand whether performance is improving and where additional research would be most fruitful. Furthermore, processing satellite and ground survey data requires domain knowledge that many in the machine learning community lack. In this paper, we introduce SustainBench, a collection of 15 benchmark tasks across 7 SDGs, including tasks related to economic development, agriculture, health, education, water and sanitation, climate action, and life on land. Datasets for 11 of the 15 tasks are released publicly for the first time. Our goals for SustainBench are to (1) lower the barriers to entry for the machine learning community to contribute to measuring and achieving the SDGs; (2) provide standard benchmarks for evaluating machine learning models on tasks across a variety of SDGs; and (3) encourage the development of novel machine learning methods where improved model performance facilitates progress towards the SDGs."
                },
                {
                    "title": "P-Adapters: Robustly Extracting Factual Information from Language Models with Diverse Prompts",
                    "abstract": "Recent work (e.g. LAMA (Petroni et al., 2019)) has found that the quality of the factual information extracted from Large Language Models (LLMs) depends on the prompts used to query them. This inconsistency is problematic because different users will query LLMs for the same information using different wording, but should receive the same, accurate responses regardless. In this work we aim to address this shortcoming by introducing P-Adapters: lightweight models that sit between the embedding layer and first attention layer of LLMs. They take LLM embeddings as input and output continuous prompts that are used to query the LLM. Additionally, we investigate Mixture of Experts (MoE) models that learn a set of continuous prompts (\"experts\") and select one to query the LLM. They require a separate classifier trained on human-annotated data to map natural language prompts to the continuous ones. P-Adapters perform comparably to the more complex MoE models in extracting factual information from BERT and RoBERTa while eliminating the need for additional annotations. P-Adapters show between 12-26% absolute improvement in precision and 36-50% absolute improvement in consistency over a baseline of only using natural language queries. Finally, we investigate what makes P-Adapters successful and conclude that a significant factor is access to the LLM's embeddings of the original natural language prompt, particularly the subject of the entity pair being queried."
                },
                {
                    "title": "Can Language Models be Biomedical Knowledge Bases?",
                    "abstract": "Pre-trained language models (LMs) have become ubiquitous in solving various natural language processing (NLP) tasks. There has been increasing interest in what knowledge these LMs contain and how we can extract that knowledge, treating LMs as knowledge bases (KBs). While there has been much work on probing LMs in the general domain, there has been little attention to whether these powerful LMs can be used as domain-specific KBs. To this end, we create the BioLAMA benchmark, which is comprised of 49K biomedical factual knowledge triples for probing biomedical LMs. We find that biomedical LMs with recently proposed probing methods can achieve up to 18.51% Acc@5 on retrieving biomedical knowledge. Although this seems promising given the task difficulty, our detailed analyses reveal that most predictions are highly correlated with prompt templates without any subjects, hence producing similar results on each relation and hindering their capabilities to be used as domain-specific KBs. We hope that BioLAMA can serve as a challenging benchmark for biomedical factual probing."
                },
                {
                    "title": "On the Opportunities and Risks of Foundation Models",
                    "abstract": "AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature."
                },
                {
                    "title": "Time-Aware Language Models as Temporal Knowledge Bases",
                    "abstract": "Many facts come with an expiration date, from the name of the President to the basketball team Lebron James plays for. However, most language models (LMs) are trained on snapshots of data collected at a specific moment in time. This can limit their utility, especially in the closed-book setting where the pretraining corpus must contain the facts the model should memorize. We introduce a diagnostic dataset aimed at probing LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum\u2014those trained on specific slices of temporal data, as well as those trained on a wide range of temporal data. To mitigate these problems, we propose a simple technique for jointly modeling text with its timestamp. This improves memorization of seen facts from the training time period, as well as calibration on predictions about unseen facts from future time periods. We also show that models trained with temporal context can be efficiently \u201crefreshed\u201d as new data arrives, without the need for retraining from scratch."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Probing Across Time: What Does RoBERTa Know and When?",
                    "abstract": "Models of language trained on very large corpora have been demonstrated useful for NLP. As fixed artifacts, they have become the object of intense study, with many researchers\"probing\"the extent to which linguistic abstractions, factual and commonsense knowledge, and reasoning abilities they acquire and readily demonstrate. Building on this line of work, we consider a new question: for types of knowledge a language model learns, when during (pre)training are they acquired? We plot probing performance across iterations, using RoBERTa as a case study. Among our findings: linguistic knowledge is acquired fast, stably, and robustly across domains. Facts and commonsense are slower and more domain-sensitive. Reasoning abilities are, in general, not stably acquired. As new datasets, pretraining protocols, and probes emerge, we believe that probing-across-time analyses can help researchers understand the complex, intermingled learning that these models undergo and guide us toward more efficient approaches that accomplish necessary learning faster."
                },
                {
                    "title": "Microestimates of wealth for all low- and middle-income countries",
                    "abstract": "Significance Many critical policy decisions rely on data about the geographic distribution of wealth and poverty, yet only half of all countries have access to adequate data on poverty. This paper creates a complete and publicly available set of microestimates of the distribution of relative poverty and wealth across all 135 low- and middle-income countries. We provide extensive evidence of the accuracy and validity of the estimates and also provide confidence intervals for each microestimate to facilitate responsible downstream use. These methods and maps provide a set of tools to study economic development and growth, guide interventions, monitor and evaluate policies, and track the elimination of poverty worldwide. Many critical policy decisions, from strategic investments to the allocation of humanitarian aid, rely on data about the geographic distribution of wealth and poverty. Yet many poverty maps are out of date or exist only at very coarse levels of granularity. Here we develop microestimates of the relative wealth and poverty of the populated surface of all 135 low- and middle-income countries (LMICs) at 2.4 km resolution. The estimates are built by applying machine-learning algorithms to vast and heterogeneous data from satellites, mobile phone networks, and topographic maps, as well as aggregated and deidentified connectivity data from Facebook. We train and calibrate the estimates using nationally representative household survey data from 56 LMICs and then validate their accuracy using four independent sources of household survey data from 18 countries. We also provide confidence intervals for each microestimate to facilitate responsible downstream use. These estimates are provided free for public use in the hope that they enable targeted policy response to the COVID-19 pandemic, provide the foundation for insights into the causes and consequences of economic development and growth, and promote responsible policymaking in support of sustainable development."
                },
                {
                    "title": "Learning How to Ask: Querying LMs with Mixtures of Soft Prompts",
                    "abstract": "Natural-language prompts have recently been used to coax pretrained language models into performing other AI tasks, using a fill-in-the-blank paradigm (Petroni et al., 2019) or a few-shot extrapolation paradigm (Brown et al., 2020). For example, language models retain factual knowledge from their training corpora that can be extracted by asking them to \u201cfill in the blank\u201d in a sentential prompt. However, where does this prompt come from? We explore the idea of learning prompts by gradient descent\u2014either fine-tuning prompts taken from previous work, or starting from random initialization. Our prompts consist of \u201csoft words,\u201d i.e., continuous vectors that are not necessarily word type embeddings from the language model. Furthermore, for each task, we optimize a mixture of prompts, learning which prompts are most effective and how to ensemble them. Across multiple English LMs and tasks, our approach hugely outperforms previous methods, showing that the implicit factual knowledge in language models was previously underestimated. Moreover, this knowledge is cheap to elicit: random initialization is nearly as good as informed initialization."
                },
                {
                    "title": "Factual Probing Is [MASK]: Learning vs. Learning to Recall",
                    "abstract": "Petroni et al. (2019) demonstrated that it is possible to retrieve world facts from a pre-trained language model by expressing them as cloze-style prompts and interpret the model\u2019s prediction accuracy as a lower bound on the amount of factual information it encodes. Subsequent work has attempted to tighten the estimate by searching for better prompts, using a disjoint set of facts as training data. In this work, we make two complementary contributions to better understand these factual probing techniques. First, we propose OptiPrompt, a novel and efficient method which directly optimizes in continuous embedding space. We find this simple method is able to predict an additional 6.4% of facts in the LAMA benchmark. Second, we raise a more important question: Can we really interpret these probing results as a lower bound? Is it possible that these prompt-search methods learn from the training data too? We find, somewhat surprisingly, that the training data used by these methods contains certain regularities of the underlying fact distribution, and all the existing prompt methods, including ours, are able to exploit them for better fact prediction. We conduct a set of control experiments to disentangle \u201clearning\u201d from \u201clearning to recall\u201d, providing a more detailed picture of what different prompts can reveal about pre-trained language models."
                },
                {
                    "title": "Eliciting Knowledge from Language Models Using Automatically Generated Prompts",
                    "abstract": "The remarkable success of pretrained language models has motivated the study of what kinds of knowledge these models learn during pretraining. Reformulating tasks as fill-in-the-blanks problems (e.g., cloze tests) is a natural approach for gauging such knowledge, however, its usage is limited by the manual effort and guesswork required to write suitable prompts. To address this, we develop AutoPrompt, an automated method to create prompts for a diverse set of tasks, based on a gradient-guided search. Using AutoPrompt, we show that masked language models (MLMs) have an inherent capability to perform sentiment analysis and natural language inference without additional parameters or finetuning, sometimes achieving performance on par with recent state-of-the-art supervised models. We also show that our prompts elicit more accurate factual knowledge from MLMs than the manually created prompts on the LAMA benchmark, and that MLMs can be used as relation extractors more effectively than supervised relation extraction models. These results demonstrate that automatically generated prompts are a viable parameter-free alternative to existing probing methods, and as pretrained LMs become more sophisticated and capable, potentially a replacement for finetuning."
                },
                {
                    "title": "Do Language Embeddings capture Scales?",
                    "abstract": "Pretrained Language Models (LMs) have been shown to possess significant linguistic, common sense and factual knowledge. One form of knowledge that has not been studied yet in this context is information about the scalar magnitudes of objects. We show that pretrained language models capture a significant amount of this information but are short of the capability required for general common-sense reasoning. We identify contextual information in pre-training and numeracy as two key factors affecting their performance, and show that a simple method of canonicalizing numbers can have a significant effect on the results."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Birds Have Four Legs?! NumerSense: Probing Numerical Commonsense Knowledge of Pre-trained Language Models",
                    "abstract": "Recent works show that pre-trained language models (PTLMs), such as BERT, possess certain commonsense and factual knowledge. They suggest that it is promising to use PTLMs as \"neural knowledge bases\" via predicting masked words. Surprisingly, we find that this may not work for numerical commonsense knowledge (e.g., a bird usually has two legs). In this paper, we investigate whether and to what extent we can induce numerical commonsense knowledge from PTLMs as well as the robustness of this process. To study this, we introduce a novel probing task with a diagnostic dataset, NumerSense, containing 13.6k masked-word-prediction probes (10.5k for fine-tuning and 3.1k for testing). Our analysis reveals that: (1) BERT and its stronger variant RoBERTa perform poorly on the diagnostic dataset prior to any fine-tuning; (2) fine-tuning with distant supervision brings some improvement; (3) the best supervised model still performs poorly as compared to human performance (54.06% vs 96.3% in accuracy)."
                },
                {
                    "title": "How Much Knowledge Can You Pack into the Parameters of a Language Model?",
                    "abstract": "It has recently been observed that neural language models trained on unstructured text can implicitly store and retrieve knowledge using natural language queries. In this short paper, we measure the practical utility of this approach by fine-tuning pre-trained models to answer questions without access to any external context or knowledge. We show that this approach scales surprisingly well with model size and outperforms models that explicitly look up knowledge on the open-domain variants of Natural Questions and WebQuestions. To facilitate reproducibility and future work, we release our code and trained models."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Using satellite imagery to understand and promote sustainable development",
                    "abstract": "Satellite monitoring of development Recent years have witnessed rapid growth in satellite-based approaches to quantifying aspects of land use, especially those monitoring the outcomes of sustainable development programs. Burke et al. reviewed this recent progress with a particular focus on machine-learning approaches and artificial intelligence methods. Drawing on examples mostly from Africa, they conclude that satellite-based methods enhance rather than replace ground-based data collection, and progress depends on a combined approach. Science, this issue p. eabe8628 BACKGROUND Accurate and comprehensive measurements of a range of sustainable development outcomes are fundamental inputs into both research and policy. For instance, good measures are needed to monitor progress toward sustainability goals and evaluate interventions designed to improve development outcomes. Traditional approaches to measurement of many key outcomes rely on household surveys that are conducted infrequently in many parts of the world and are often of low accuracy. The paucity of ground data stands in contrast to the rapidly growing abundance and quality of satellite imagery. Multiple public and private sensors launched in recent years provide temporal, spatial, and spectral information on changes happening on Earth\u2019s surface. Here we review a rapidly growing scientific literature that seeks to use this satellite imagery to measure and understand various outcomes related to sustainable development. We pay particular attention to recent approaches that use methods from artificial intelligence to extract information from images, as these methods typically outperform earlier approaches and enable new insights. Our focus is on settings and applications where humans themselves, or what they produce, are the outcome of interest and on where these outcomes are being measured using satellite imagery. ADVANCES We describe and synthesize the variety of approaches that have been used to extract information from satellite imagery, with particular attention given to recent machine learning\u2013based approaches and settings in which training data are limited or noisy. We then quantitatively assess predictive performance of these approaches in the domains of smallholder agriculture, economic livelihoods, population, and informal settlements. We show that satellite-based performance in predicting these outcomes is reasonably strong and improving. Performance improvements have come through a combination of more numerous and accurate training data, more abundant and higher-quality imagery, and creative application of advances in computer vision to satellite inputs and sustainability outcomes. Further, our analyses suggest that reported model performance likely understates true performance in many settings, given the noisy data on which predictions are evaluated and the types of noise typically observed in sustainability applications. For multiple outcomes of interest, satellite-based estimates can now equal or exceed the accuracy of traditional approaches to outcome measurement. We describe multiple methods through which the true performance of satellite-based approaches can be better understood. Integration of satellite-based sustainability measurements into research has been broad, and we describe applications in agriculture, fisheries, health, and economics. Documented uses of these measurements in public-sector decision-making are rarer, which we attribute in part to the novelty of the approaches, their lack of interpretability, and the potential benefits to some policy-makers of not having certain outcomes be measured. OUTLOOK The largest constraint to satellite-based model performance is now training data rather than imagery. While imagery has become abundant, the scarcity and frequent unreliability of ground data make both training and validation of satellite-based models difficult. Expanding the quantity and quality of such data will quickly accelerate progress in this field. Other opportunities for advancement include improvements in model interpretability, fusion of satellites with other nontraditional data that provide complementary information, and more-rigorous evaluation of satellite-based approaches (relative to available alternatives) in the context of specific use cases. Nevertheless, despite the current and future promise of satellite-based approaches, we argue that these approaches will amplify rather than replace existing ground-based data collection efforts in most settings. Many outcomes of interest will likely never be accurately estimated with satellites; for outcomes where satellites do have predictive power, high-quality local training data can nearly always improve model performance. Increasing collection of satellite imagery can help measure livelihood outcomes in areas where ground data are sparse. (Left) Interval between nationally representative economic surveys over the past three decades shows long lags in many developing countries. (Middle) Recently added public and private satellites have broken the traditional trade-off between temporal and spatial resolution. (Right) Performance in measuring the presence of informal settlements, crop yields on smallholder agricultural plots, and village-level asset wealth. R2, coefficient of determination. Accurate and comprehensive measurements of a range of sustainable development outcomes are fundamental inputs into both research and policy. We synthesize the growing literature that uses satellite imagery to understand these outcomes, with a focus on approaches that combine imagery with machine learning. We quantify the paucity of ground data on key human-related outcomes and the growing abundance and improving resolution (spatial, temporal, and spectral) of satellite imagery. We then review recent machine learning approaches to model-building in the context of scarce and noisy training data, highlighting how this noise often leads to incorrect assessment of model performance. We quantify recent model performance across multiple sustainable development domains, discuss research and policy applications, explore constraints to future progress, and highlight research directions for the field."
                },
                {
                    "title": "How Can We Know What Language Models Know?",
                    "abstract": "Abstract Recent work has presented intriguing results examining the knowledge contained in language models (LMs) by having the LM fill in the blanks of prompts such as \u201cObama is a __ by profession\u201d. These prompts are usually manually created, and quite possibly sub-optimal; another prompt such as \u201cObama worked as a __ \u201d may result in more accurately predicting the correct profession. Because of this, given an inappropriate prompt, we might fail to retrieve facts that the LM does know, and thus any given prompt only provides a lower bound estimate of the knowledge contained in an LM. In this paper, we attempt to more accurately estimate the knowledge contained in LMs by automatically discovering better prompts to use in this querying process. Specifically, we propose mining-based and paraphrasing-based methods to automatically generate high-quality and diverse prompts, as well as ensemble methods to combine answers from different prompts. Extensive experiments on the LAMA benchmark for extracting relational knowledge from LMs demonstrate that our methods can improve accuracy from 31.1% to 39.6%, providing a tighter lower bound on what LMs know. We have released the code and the resulting LM Prompt And Query Archive (LPAQA) at https://github.com/jzbjyb/LPAQA."
                },
                {
                    "title": "Commonsense Knowledge Mining from Pretrained Models",
                    "abstract": "Inferring commonsense knowledge is a key challenge in machine learning. Due to the sparsity of training data, previous work has shown that supervised methods for commonsense knowledge mining underperform when evaluated on novel data. In this work, we develop a method for generating commonsense knowledge using a large, pre-trained bidirectional language model. By transforming relational triples into masked sentences, we can use this model to rank a triple\u2019s validity by the estimated pointwise mutual information between the two entities. Since we do not update the weights of the bidirectional model, our approach is not biased by the coverage of any one commonsense knowledge base. Though we do worse on a held-out test set than models explicitly trained on a corresponding training set, our approach outperforms these methods when mining commonsense knowledge from new sources, suggesting that our unsupervised technique generalizes better than current supervised approaches."
                },
                {
                    "title": "RoBERTa: A Robustly Optimized BERT Pretraining Approach",
                    "abstract": "Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code."
                },
                {
                    "title": "Development and Delivery of Species Distribution Models to Inform Decision-Making",
                    "abstract": "\n Information on where species occur is an important component of conservation and management decisions, but knowledge of distributions is often coarse or incomplete. Species distribution models provide a tool for mapping habitat and can produce credible, defensible, and repeatable information with which to inform decisions. However, these models are sensitive to data inputs and methodological choices, making it important to assess the reliability and utility of model predictions. We provide a rubric that model developers can use to communicate a model's attributes and its appropriate uses. We emphasize the importance of tailoring model development and delivery to the species of interest and the intended use and the advantages of iterative modeling and validation. We highlight how species distribution models have been used to design surveys for new populations, inform spatial prioritization decisions for management actions, and support regulatory decision-making and compliance, tying these examples back to our model assessment rubric."
                },
                {
                    "title": "Unified Language Model Pre-training for Natural Language Understanding and Generation",
                    "abstract": "This paper presents a new Unified pre-trained Language Model (UniLM) that can be fine-tuned for both natural language understanding and generation tasks. The model is pre-trained using three types of language modeling tasks: unidirectional, bidirectional, and sequence-to-sequence prediction. The unified modeling is achieved by employing a shared Transformer network and utilizing specific self-attention masks to control what context the prediction conditions on. UniLM compares favorably with BERT on the GLUE benchmark, and the SQuAD 2.0 and CoQA question answering tasks. Moreover, UniLM achieves new state-of-the-art results on five natural language generation datasets, including improving the CNN/DailyMail abstractive summarization ROUGE-L to 40.51 (2.04 absolute improvement), the Gigaword abstractive summarization ROUGE-L to 35.75 (0.86 absolute improvement), the CoQA generative question answering F1 score to 82.5 (37.1 absolute improvement), the SQuAD question generation BLEU-4 to 22.12 (3.75 absolute improvement), and the DSTC7 document-grounded dialog response generation NIST-4 to 2.67 (human performance is 2.65). The code and pre-trained models are available at this https URL."
                },
                {
                    "title": "Predicting Economic Development using Geolocated Wikipedia Articles",
                    "abstract": "Progress on the UN Sustainable Development Goals (SDGs) is hampered by a persistent lack of data regarding key social, environmental, and economic indicators, particularly in developing countries. For example, data on poverty - the first of seventeen SDGs - is both spatially sparse and infrequently collected in Sub-Saharan Africa due to the high cost of surveys. Here we propose a novel method for estimating socioeconomic indicators using open-source, geolocated textual information from Wikipedia articles. We demonstrate that modern NLP techniques can be used to predict community-level asset wealth and education outcomes using nearby geolocated Wikipedia articles. When paired with nightlights satellite imagery, our method outperforms all previously published benchmarks for this prediction task, indicating the potential of Wikipedia to inform both research in the social sciences and future policy decisions."
                },
                {
                    "title": "Can Human Development be Measured with Satellite Imagery?",
                    "abstract": "In many developing country environments, it is difficult or impossible to obtain recent, reliable estimates of human development. Nationally representative household surveys, which are the standard instrument for determining development policy and priorities, are typically too expensive to collect with any regularity. Recently, however, researchers have shown the potential for remote sensing technologies to provide a possible solution to this data constraint. In particular, recent work indicates that satellite imagery can be processed with deep neural networks to accurately estimate the sub-regional distribution of wealth in sub-Saharan Africa. In this paper, we explore the extent to which the same approach--- of using convolutional neural networks to process satellite imagery--- can be used to measure a broader set of human development indicators, in a broader range of geographic contexts. Our analysis produces three main results: First, we successfully replicate prior work showing that satellite images can accurately infer a wealth-based index of poverty in sub-Saharan Africa. Second, we show that this approach can generalize to predicting poverty in other countries and continents, but that the performance is sensitive to the hyperparameters used to tune the learning algorithm. Finally, we find that this approach does not trivially generalize to predicting other measures of development such as educational attainment, access to drinking water, and a variety of health-related indicators. We discuss in detail whether these findings represent a fundamental limitation of this approach, or could be fixed through more concerted adaptations of the machine learning environment."
                },
                {
                    "title": "Poverty Prediction with Public Landsat 7 Satellite Imagery and Machine Learning",
                    "abstract": "Obtaining detailed and reliable data about local economic livelihoods in developing countries is expensive, and data are consequently scarce. Previous work has shown that it is possible to measure local-level economic livelihoods using high-resolution satellite imagery. However, such imagery is relatively expensive to acquire, often not updated frequently, and is mainly available for recent years. We train CNN models on free and publicly available multispectral daytime satellite images of the African continent from the Landsat 7 satellite, which has collected imagery with global coverage for almost two decades. We show that despite these images' lower resolution, we can achieve accuracies that exceed previous benchmarks."
                },
                {
                    "title": "VIIRS night-time lights",
                    "abstract": "ABSTRACT The Visible Infrared Imaging Radiometer Suite (VIIRS) Day/Night Band (DNB) collects global low-light imaging data that have significant improvements over comparable data collected for 40 years by the DMSP Operational Linescan System. One of the prominent features of DNB data is the detection of electric lighting present on the Earth\u2019s surface. Most of these lights are from human settlements. VIIRS collects source data that could be used to generate monthly and annual science grade global radiance maps of human settlements with electric lighting. There are a substantial number of steps involved in producing a product that has been cleaned to exclude background noise, solar and lunar contamination, data degraded by cloud cover, and features unrelated to electric lighting (e.g. fires, flares, volcanoes). This article describes the algorithms developed for the production of high-quality global VIIRS night-time lights. There is a broad base of science users for VIIRS night-time lights products, ranging from land-use scientists, urban geographers, ecologists, carbon modellers, astronomers, demographers, economists, and social scientists."
                },
                {
                    "title": "Comprehensive survey of deep learning in remote sensing: theories, tools, and challenges for the community",
                    "abstract": "Abstract. In recent years, deep learning (DL), a rebranding of neural networks (NNs), has risen to the top in numerous areas, namely computer vision (CV), speech recognition, and natural language processing. Whereas remote sensing (RS) possesses a number of unique challenges, primarily related to sensors and applications, inevitably RS draws from many of the same theories as CV, e.g., statistics, fusion, and machine learning, to name a few. This means that the RS community should not only be aware of advancements such as DL, but also be leading researchers in this area. Herein, we provide the most comprehensive survey of state-of-the-art RS DL research. We also review recent new developments in the DL field that can be used in DL for RS. Namely, we focus on theories, tools, and challenges for the RS community. Specifically, we focus on unsolved challenges and opportunities as they relate to (i) inadequate data sets, (ii) human-understandable solutions for modeling physical phenomena, (iii) big data, (iv) nontraditional heterogeneous data sources, (v) DL architectures and learning algorithms for spectral, spatial, and temporal data, (vi) transfer learning, (vii) an improved theoretical understanding of DL systems, (viii) high barriers to entry, and (ix) training and optimizing the DL."
                },
                {
                    "title": "A Deep Learning Approach for Population Estimation from Satellite Imagery",
                    "abstract": "Knowing where people live is a fundamental component of many decision making processes such as urban development, infectious disease containment, evacuation planning, risk management, conservation planning, and more. While bottom-up, survey driven censuses can provide a comprehensive view into the population landscape of a country, they are expensive to realize, are infrequently performed, and only provide population counts over broad areas. Population disaggregation techniques and population projection methods individually address these shortcomings, but also have shortcomings of their own. To jointly answer the questions of \"where do people live\" and \"how many people live there,\" we propose a deep learning model for creating high-resolution population estimations from satellite imagery. Specifically, we train convolutional neural networks to predict population in the USA at a 0.01\u00b0x0.01\u00b0 resolution grid from 1-year composite Landsat imagery. We validate these models in two ways: quantitatively, by comparing our model's grid cell estimates aggregated at a county-level to several US Census county-level population projections, and qualitatively, by directly interpreting the model's predictions in terms of the satellite image inputs. We find that aggregating our model's estimates gives comparable results to the Census county-level population projections and that the predictions made by our model can be directly interpreted, which give it advantages over traditional population disaggregation methods. In general, our model is an example of how machine learning techniques can be an effective tool for extracting information from inherently unstructured, remotely sensed data to provide effective solutions to social problems."
                },
                {
                    "title": "Extracting and analyzing semantic relatedness between cities using news articles",
                    "abstract": "ABSTRACT News articles capture a variety of topics about our society. They reflect not only the socioeconomic activities that happened in our physical world, but also some of the cultures, human interests, and public concerns that exist only in the perceptions of people. Cities are frequently mentioned in news articles, and two or more cities may co-occur in the same article. Such co-occurrence often suggests certain relatedness between the mentioned cities, and the relatedness may be under different topics depending on the contents of the news articles. We consider the relatedness under different topics as semantic relatedness. By reading news articles, one can grasp the general semantic relatedness between cities; yet, given hundreds of thousands of news articles, it is very difficult, if not impossible, for anyone to manually read them. This paper proposes a computational framework which can \u2018read\u2019 a large number of news articles and extract the semantic relatedness between cities. This framework is based on a natural language processing model and employs a machine learning process to identify the main topics of news articles. We describe the overall structure of this framework and its individual modules, and then apply it to an experimental dataset with more than 500,000 news articles covering the top 100 US cities spanning a 10-year period. We perform exploratory visualizations of the extracted semantic relatedness under different topics and over multiple years. We also analyze the impact of geographic distance on semantic relatedness and find varied distance decay effects. The proposed framework can be used to support large-scale content analysis in city network research."
                },
                {
                    "title": "Cross-region traffic prediction for China on OpenStreetMap",
                    "abstract": "OpenStreetMap (OSM) is a free, open-source and popular mapping service. However, due to various reasons, it doesn't offer live traffic information or traffic prediction for China. This paper presents an approach and a system to learn a prediction model from graphical traffic condition data provided by Baidu Map, which is a commercial, close-source map provider in China, and apply the model on OSM so that one can predict the traffic conditions with nearly 90% accuracy in various parts of Shanghai, China, even though no traffic data is available for that area from Baidu Map. This novel system can be useful in urban planning, transportation dispatching as well as personal travel planning."
                },
                {
                    "title": "Combining satellite imagery and machine learning to predict poverty",
                    "abstract": "Measuring consumption and wealth remotely Nighttime lighting is a rough proxy for economic wealth, and nighttime maps of the world show that many developing countries are sparsely illuminated. Jean et al. combined nighttime maps with high-resolution daytime satellite images (see the Perspective by Blumenstock). With a bit of machine-learning wizardry, the combined images can be converted into accurate estimates of household consumption and assets, both of which are hard to measure in poorer countries. Furthermore, the night- and day-time data are publicly available and nonproprietary. Science, this issue p. 790; see also p. 753 Satellites collect data that can be used to measure income and wealth. Reliable data on economic livelihoods remain scarce in the developing world, hampering efforts to study these outcomes and to design policies that improve them. Here we demonstrate an accurate, inexpensive, and scalable method for estimating consumption expenditure and asset wealth from high-resolution satellite imagery. Using survey and satellite data from five African countries\u2014Nigeria, Tanzania, Uganda, Malawi, and Rwanda\u2014we show how a convolutional neural network can be trained to identify image features that can explain up to 75% of the variation in local-level economic outcomes. Our method, which requires only publicly available data, could transform efforts to track and target poverty in developing countries. It also demonstrates how powerful machine learning techniques can be applied in a setting with limited training data, suggesting broad potential application across many scientific domains."
                },
                {
                    "title": "Bag of Tricks for Efficient Text Classification",
                    "abstract": "This paper explores a simple and efficient baseline for text classification. Our experiments show that our fast text classifier fastText is often on par with deep learning classifiers in terms of accuracy, and many orders of magnitude faster for training and evaluation. We can train fastText on more than one billion words in less than ten minutes using a standard multicore CPU, and classify half a million sentences among 312K classes in less than a minute."
                },
                {
                    "title": "XGBoost: A Scalable Tree Boosting System",
                    "abstract": "Tree boosting is a highly effective and widely used machine learning method. In this paper, we describe a scalable end-to-end tree boosting system called XGBoost, which is used widely by data scientists to achieve state-of-the-art results on many machine learning challenges. We propose a novel sparsity-aware algorithm for sparse data and weighted quantile sketch for approximate tree learning. More importantly, we provide insights on cache access patterns, data compression and sharding to build a scalable tree boosting system. By combining these insights, XGBoost scales beyond billions of examples using far fewer resources than existing systems."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Predicting poverty and wealth from mobile phone metadata",
                    "abstract": "Predicting unmeasurable wealth In developing countries, collecting data on basic economic quantities, such as wealth and income, is costly, time-consuming, and unreliable. Taking advantage of the ubiquity of mobile phones in Rwanda, Blumenstock et al. mapped mobile phone metadata inputs to individual phone subscriber wealth. They applied the model to predict wealth throughout Rwanda and show that the predictions matched well with those from detailed boots-on-the-ground surveys of the population. Science, this issue p. 1073 Metadata from individuals\u2019 phones can be used to predict aggregate-level characteristics such as access to electricity. Accurate and timely estimates of population characteristics are a critical input to social and economic research and policy. In industrialized economies, novel sources of data are enabling new approaches to demographic profiling, but in developing countries, fewer sources of big data exist. We show that an individual\u2019s past history of mobile phone use can be used to infer his or her socioeconomic status. Furthermore, we demonstrate that the predicted attributes of millions of individuals can, in turn, accurately reconstruct the distribution of wealth of an entire nation or to infer the asset distribution of microregions composed of just a few households. In resource-constrained environments where censuses and household surveys are rare, this approach creates an option for gathering localized and timely information at a fraction of the cost of traditional methods."
                },
                {
                    "title": "You Are What You Tweet: Analyzing Twitter for Public Health",
                    "abstract": "\n \n Analyzing user messages in social media can measure different population characteristics, including public health measures. For example, recent work has correlated Twitter messages with influenza rates in the United States; but this has largely been the extent of mining Twitter for public health. In this work, we consider a broader range of public health applications for Twitter. We apply the recently introduced Ailment Topic Aspect Model to over one and a half million health related tweets and discover mentions of over a dozen ailments, including allergies, obesity and insomnia. We introduce extensions to incorporate prior knowledge into this model and apply it to several tasks: tracking illnesses over times (syndromic surveillance), measuring behavioral risk factors, localizing illnesses by geographic region, and analyzing symptoms and medication usage. We show quantitative correlations with public health data and qualitative evaluations of model output. Our results suggest that Twitter has broad applicability for public health research.\n \n"
                },
                {
                    "title": "The Use of Twitter to Track Levels of Disease Activity and Public Concern in the U.S. during the Influenza A H1N1 Pandemic",
                    "abstract": "Twitter is a free social networking and micro-blogging service that enables its millions of users to send and read each other's \u201ctweets,\u201d or short, 140-character messages. The service has more than 190 million registered users and processes about 55 million tweets per day. Useful information about news and geopolitical events lies embedded in the Twitter stream, which embodies, in the aggregate, Twitter users' perspectives and reactions to current events. By virtue of sheer volume, content embedded in the Twitter stream may be useful for tracking or even forecasting behavior if it can be extracted in an efficient manner. In this study, we examine the use of information embedded in the Twitter stream to (1) track rapidly-evolving public sentiment with respect to H1N1 or swine flu, and (2) track and measure actual disease activity. We also show that Twitter can be used as a measure of public interest or concern about health-related events. Our results show that estimates of influenza-like illness derived from Twitter chatter accurately track reported disease levels."
                },
                {
                    "title": "Prompt Tuning or Fine-Tuning - Investigating Relational Knowledge in Pre-Trained Language Models",
                    "abstract": "Extracting relational knowledge from large pre-trained language models by a cloze-style sentence serving as a query has shown promising results. In particular, language models can be queried similar to knowledge graphs. The performance of the relational fact extraction task depends signi\ufb01cantly on the query sentence, also known under the term prompt . Tuning these prompts has shown to increase the precision on standard language models by a maximum of around 12% points. However, usually large amounts of data in the form of existing knowledge graph facts and large text corpora are needed to train the required additional model. In this work, we propose using a completely di\ufb00erent approach: Instead of spending resources on training an additional model, we simply perform an adaptive \ufb01ne-tuning of the pre-trained language model on the standard \ufb01ll-mask task using a small training dataset of existing facts from a knowledge graph. We investigate the di\ufb00erences between complex prompting techniques and adaptive \ufb01ne-tuning in an extensive evaluation. Remarkably, adaptive \ufb01ne-tuning outperforms all baselines, even by using signi\ufb01cantly fewer training facts. Additionally, we analyze the transfer learning capabilities of this adapted language model by training on a restricted set of relations to show that even fewer training relations are needed to achieve high knowledge extraction quality."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Improving Language Understanding by Generative Pre-Training",
                    "abstract": "Natural language understanding comprises a wide range of diverse tasks such as textual entailment, question answering, semantic similarity assessment, and document classi\ufb01cation. Although large unlabeled text corpora are abundant, labeled data for learning these speci\ufb01c tasks is scarce, making it challenging for discriminatively trained models to perform adequately. We demonstrate that large gains on these tasks can be realized by generative pre-training of a language model on a diverse corpus of unlabeled text, followed by discriminative \ufb01ne-tuning on each speci\ufb01c task. In contrast to previous approaches, we make use of task-aware input transformations during \ufb01ne-tuning to achieve effective transfer while requiring minimal changes to the model architecture. We demonstrate the effectiveness of our approach on a wide range of benchmarks for natural language understanding. Our general task-agnostic model outperforms discriminatively trained models that use architectures speci\ufb01cally crafted for each task, signi\ufb01cantly improving upon the state of the art in 9 out of the 12 tasks studied. For instance, we achieve absolute improvements of 8.9% on commonsense reasoning (Stories Cloze Test), 5.7% on question answering (RACE), and 1.5% on textual entailment (MultiNLI)."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively extract and utilize geospatial knowledge from instruction-finetuned large language models (LLMs) for real-world geospatial tasks, such as predicting population density?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem has significant implications for the research community as it could enhance the understanding of LLMs' capabilities in geospatial contexts, leading to the development of geo-specific foundation models. This research could pave the way for more accurate and efficient applications in urban planning, disaster response, and resource management, ultimately advancing knowledge in both NLP and geospatial analysis. By demonstrating the potential of LLMs in geospatial tasks, future research can explore novel applications and methodologies that leverage these models for complex geospatial challenges.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the inherent complexity of geospatial data, which often requires integrating various features and covariates that may be difficult to obtain. Naive approaches, such as spatial interpolation based solely on geographic coordinates, may fail due to their inability to account for the underlying relationships and patterns in the data, especially with small sample sizes. Additionally, the sensitivity of prompt crafting and the limitations of existing fine-tuning methods pose technical obstacles that need to be addressed to effectively extract and utilize geospatial knowledge from LLMs.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the capabilities of pre-trained LLMs without delving into the fine-tuning necessary for geospatial applications. Existing solutions have been limited by their focus on common NLP tasks rather than exploring the potential of LLMs for specific geospatial knowledge extraction. Barriers such as the lack of tailored datasets and methodologies for geospatial tasks have hindered progress. Our approach differs by conducting a comprehensive investigation into fine-tuning LLMs specifically for geospatial knowledge extraction, aiming to demonstrate the quantity and quality of this knowledge for practical applications.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves fine-tuning various LLMs on a curated geospatial dataset that includes geographic coordinates and associated features to predict response variables like population density. We will evaluate the models using metrics such as prediction accuracy and model performance on geospatial tasks. The expected outcomes include a demonstration of the quantity and quality of geospatial knowledge contained in LLMs, as well as"
            }
        },
        "author_data": {
            "9dd176af-aebf-4885-b77d-e32659e56763": {
                "pk": "9dd176af-aebf-4885-b77d-e32659e56763",
                "name": "Rohin Manvi",
                "collaborators": [
                    "Stefano Ermon",
                    "Anikait Singh",
                    "Samar Khanna",
                    "Marshall Burke",
                    "David B. Lobell",
                    "Kabir Jolly",
                    "J. Quevedo",
                    "Avash Shrestha",
                    "Vincent Li"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "3D Reconstruction",
                    "Question Answering"
                ],
                "publications": [
                    {
                        "title": "Adaptive Inference-Time Compute: LLMs Can Predict if They Can Do Better, Even Mid-Generation",
                        "abstract": "Inference-time computation is a powerful paradigm to enhance the performance of large language models (LLMs), with Best-of-N sampling being a widely used technique. However, this method is computationally expensive, requiring both (1) an external reward model and (2) the generation of multiple samples. In this work, we introduce a new generative self-evaluation scheme designed to adaptively reduce the number of generated samples while maintaining or even improving performance. We use a generative reward model formulation, allowing the LLM to predict mid-generation the probability that restarting the generation will yield a better response. These predictions are obtained without an external reward model and can be used to decide whether or not to generate more samples, prune unpromising samples early on, or to pick the best sample. This capability is very inexpensive as it involves generating a single predefined token. Trained using a dataset constructed with real unfiltered LMSYS user prompts, Llama 3.1 8B's win rate against GPT-4 on AlpacaEval increases from 21% to 34% with 16 samples and math performance on GSM8K improves from 84% to 91%. By sampling only when the LLM determines that it is beneficial to do so and adaptively adjusting temperature annealing, we demonstrate that 74% of the improvement from using 16 samples can be achieved with only 1.2 samples on average. We further demonstrate that 50-75% of samples can be pruned early in generation with minimal degradation in performance. Overall, our methods enable more efficient and scalable compute utilization during inference for LLMs."
                    },
                    {
                        "title": "Large Language Models are Geographically Biased",
                        "abstract": "Large Language Models (LLMs) inherently carry the biases contained in their training corpora, which can lead to the perpetuation of societal harm. As the impact of these foundation models grows, understanding and evaluating their biases becomes crucial to achieving fairness and accuracy. We propose to study what LLMs know about the world we live in through the lens of geography. This approach is particularly powerful as there is ground truth for the numerous aspects of human life that are meaningfully projected onto geographic space such as culture, race, language, politics, and religion. We show various problematic geographic biases, which we define as systemic errors in geospatial predictions. Initially, we demonstrate that LLMs are capable of making accurate zero-shot geospatial predictions in the form of ratings that show strong monotonic correlation with ground truth (Spearman's $\\rho$ of up to 0.89). We then show that LLMs exhibit common biases across a range of objective and subjective topics. In particular, LLMs are clearly biased against locations with lower socioeconomic conditions (e.g. most of Africa) on a variety of sensitive subjective topics such as attractiveness, morality, and intelligence (Spearman's $\\rho$ of up to 0.70). Finally, we introduce a bias score to quantify this and find that there is significant variation in the magnitude of bias across existing LLMs. Code is available on the project website: https://rohinmanvi.github.io/GeoLLM"
                    },
                    {
                        "title": "Depth-Aware Pixel2Mesh",
                        "abstract": "Currently, there are existing methods that are able to construct 3D meshes of objects when given 2D image inputs. However, there is room to improve upon these models by adding in a component of object recognition that we as humans use to perceive the world around us: depth. Our goal is to build upon the Pixel2Mesh work by expanding its capacity from utilizing RGB input images to RGB-D input images. These RGB-D images were created using the MiDaS model, and were fed into a modified Pixel2Mesh to create 3D meshes. While our initial approach involving leveraging differentiable rendering was initially unsuccessful, we found that when using the depth-aware Pixel2Mesh, training results exhibited faster convergence and validation results indicated stronger total loss minimization and comparable voxel loss performance."
                    },
                    {
                        "title": "Implemention and Analysis of Character-level Embeddings, Self-attention, and R-NET on BiDAF for QA on SQuAD 2.0",
                        "abstract": "Question Answering (QA) is an important task in the field of NLP. From chat bots to social media to medical text analysis, there are countless crucial applications of QA in the real world. In this paper, we aim to improve upon a given Bi-Directional Attention Flow (BiDAF) model for the task of QA on the SQuAD 2.0 dataset, by implementing various features such as character embeddings, self-attention, and another model\u2019s embeddings/encoding layers. We saw general improvement in our model\u2019s performance as we implemented features, with the exception of the other model\u2019s embeddings/encoding layers by themselves, which saw slightly worse performance. We analyzed our results and submitted our best model, which had tweaked hyperparameters, to the Gradescope Default Final Project - SQuAD track test leaderboard, where we got an EM score of 62.06 and an FM score of 65.14, an improvement on the provided baseline model\u2019s performance on the dev leaderboard. Finally, we qualitatively analyzed how our best model performed on different question types."
                    }
                ]
            },
            "53bb4a51-d819-4689-9c74-b819bbb2e2cb": {
                "pk": "53bb4a51-d819-4689-9c74-b819bbb2e2cb",
                "name": "Samar Khanna",
                "collaborators": [
                    "Stefano Ermon",
                    "David B. Lobell",
                    "Marshall Burke",
                    "Linqi Zhou",
                    "Medhanie Irgau",
                    "Louis Foucard",
                    "Yi Shi",
                    "Chi-Kuei Liu",
                    "Quinn Z Shen",
                    "Thuyen Ngo",
                    "Zi-Xiang Xia",
                    "Jeremy Andrew Irvin",
                    "Emily Ruoyu Liu",
                    "Joyce Chuyi Chen",
                    "Ines Dormoy",
                    "Jinyoung Kim",
                    "Zhuo Zheng",
                    "Rohin Manvi",
                    "Aaron Lou",
                    "Patrick Liu",
                    "Chenlin Meng",
                    "Robin Rombach"
                ],
                "domain": [
                    "Computer Vision",
                    "Natural Language Processing",
                    "Generative Models",
                    "Remote Sensing"
                ],
                "publications": [
                    {
                        "title": "ExPLoRA: Parameter-Efficient Extended Pre-Training to Adapt Vision Transformers under Domain Shifts",
                        "abstract": "Parameter-efficient fine-tuning (PEFT) techniques such as low-rank adaptation (LoRA) can effectively adapt large pre-trained foundation models to downstream tasks using only a small fraction (0.1%-10%) of the original trainable weights. An under-explored question of PEFT is in extending the pre-training phase without supervised labels; that is, can we adapt a pre-trained foundation model to a new domain via efficient self-supervised pre-training on this new domain? In this work, we introduce ExPLoRA, a highly effective technique to improve transfer learning of pre-trained vision transformers (ViTs) under domain shifts. Initializing a ViT with pre-trained weights on large, natural-image datasets such as from DinoV2 or MAE, ExPLoRA continues the unsupervised pre-training objective on a new domain, unfreezing 1-2 pre-trained ViT blocks and tuning all other layers with LoRA. We then fine-tune the resulting model only with LoRA on this new domain for supervised learning. Our experiments demonstrate state-of-the-art results on satellite imagery, even outperforming fully pre-training and fine-tuning ViTs. Using the DinoV2 training objective, we demonstrate up to 7.5% improvement in linear probing top-1 accuracy on downstream tasks while using<10% of the number of parameters that are used in prior fully-tuned state-of-the art approaches. Our ablation studies confirm the efficacy of our approach over other baselines, including PEFT and unfreezing more ViT blocks. Code is available on the project website: https://samar-khanna.github.io/ExPLoRA/"
                    },
                    {
                        "title": "SpotNet: An Image Centric, Lidar Anchored Approach To Long Range Perception",
                        "abstract": "In this paper, we propose SpotNet: a fast, single stage, image-centric but LiDAR anchored approach for long range 3D object detection. We demonstrate that our approach to LiDAR/image sensor fusion, combined with the joint learning of 2D and 3D detection tasks, can lead to accurate 3D object detection with very sparse LiDAR support. Unlike more recent bird's-eye-view (BEV) sensor-fusion methods which scale with range $r$ as $O(r^2)$, SpotNet scales as $O(1)$ with range. We argue that such an architecture is ideally suited to leverage each sensor's strength, i.e. semantic understanding from images and accurate range finding from LiDAR data. Finally we show that anchoring detections on LiDAR points removes the need to regress distances, and so the architecture is able to transfer from 2MP to 8MP resolution images without re-training."
                    },
                    {
                        "title": "TEOChat: A Large Vision-Language Assistant for Temporal Earth Observation Data",
                        "abstract": "Large vision and language assistants have enabled new capabilities for interpreting natural images. These approaches have recently been adapted to earth observation data, but they are only able to handle single image inputs, limiting their use for many real-world tasks. In this work, we develop a new vision and language assistant called TEOChat that can engage in conversations about temporal sequences of earth observation data. To train TEOChat, we curate an instruction-following dataset composed of many single image and temporal tasks including building change and damage assessment, semantic change detection, and temporal scene classification. We show that TEOChat can perform a wide variety of spatial and temporal reasoning tasks, substantially outperforming previous vision and language assistants, and even achieving comparable or better performance than specialist models trained to perform these specific tasks. Furthermore, TEOChat achieves impressive zero-shot performance on a change detection and change question answering dataset, outperforms GPT-4o and Gemini 1.5 Pro on multiple temporal tasks, and exhibits stronger single image capabilities than a comparable single EO image instruction-following model. We publicly release our data, models, and code at https://github.com/ermongroup/TEOChat ."
                    },
                    {
                        "title": "Large Language Models are Geographically Biased",
                        "abstract": "Large Language Models (LLMs) inherently carry the biases contained in their training corpora, which can lead to the perpetuation of societal harm. As the impact of these foundation models grows, understanding and evaluating their biases becomes crucial to achieving fairness and accuracy. We propose to study what LLMs know about the world we live in through the lens of geography. This approach is particularly powerful as there is ground truth for the numerous aspects of human life that are meaningfully projected onto geographic space such as culture, race, language, politics, and religion. We show various problematic geographic biases, which we define as systemic errors in geospatial predictions. Initially, we demonstrate that LLMs are capable of making accurate zero-shot geospatial predictions in the form of ratings that show strong monotonic correlation with ground truth (Spearman's $\\rho$ of up to 0.89). We then show that LLMs exhibit common biases across a range of objective and subjective topics. In particular, LLMs are clearly biased against locations with lower socioeconomic conditions (e.g. most of Africa) on a variety of sensitive subjective topics such as attractiveness, morality, and intelligence (Spearman's $\\rho$ of up to 0.70). Finally, we introduce a bias score to quantify this and find that there is significant variation in the magnitude of bias across existing LLMs. Code is available on the project website: https://rohinmanvi.github.io/GeoLLM"
                    },
                    {
                        "title": "Denoising Diffusion Bridge Models",
                        "abstract": "Diffusion models are powerful generative models that map noise to data using stochastic processes. However, for many applications such as image editing, the model input comes from a distribution that is not random noise. As such, diffusion models must rely on cumbersome methods like guidance or projected sampling to incorporate this information in the generative process. In our work, we propose Denoising Diffusion Bridge Models (DDBMs), a natural alternative to this paradigm based on diffusion bridges, a family of processes that interpolate between two paired distributions given as endpoints. Our method learns the score of the diffusion bridge from data and maps from one endpoint distribution to the other by solving a (stochastic) differential equation based on the learned score. Our method naturally unifies several classes of generative models, such as score-based diffusion models and OT-Flow-Matching, allowing us to adapt existing design and architectural choices to our more general problem. Empirically, we apply DDBMs to challenging image datasets in both pixel and latent space. On standard image translation problems, DDBMs achieve significant improvement over baseline methods, and, when we reduce the problem to image generation by setting the source distribution to random noise, DDBMs achieve comparable FID scores to state-of-the-art methods despite being built for a more general task."
                    },
                    {
                        "title": "DiffusionSat: A Generative Foundation Model for Satellite Imagery",
                        "abstract": "Diffusion models have achieved state-of-the-art results on many modalities including images, speech, and video. However, existing models are not tailored to support remote sensing data, which is widely used in important applications including environmental monitoring and crop-yield prediction. Satellite images are significantly different from natural images -- they can be multi-spectral, irregularly sampled across time -- and existing diffusion models trained on images from the Web do not support them. Furthermore, remote sensing data is inherently spatio-temporal, requiring conditional generation tasks not supported by traditional methods based on captions or images. In this paper, we present DiffusionSat, to date the largest generative foundation model trained on a collection of publicly available large, high-resolution remote sensing datasets. As text-based captions are sparsely available for satellite images, we incorporate the associated metadata such as geolocation as conditioning information. Our method produces realistic samples and can be used to solve multiple generative tasks including temporal generation, superresolution given multi-spectral inputs and in-painting. Our method outperforms previous state-of-the-art methods for satellite image generation and is the first large-scale generative foundation model for satellite imagery. The project website can be found here: https://samar-khanna.github.io/DiffusionSat/"
                    }
                ]
            },
            "6e664c87-2e8c-414c-b17c-9c756737d365": {
                "pk": "6e664c87-2e8c-414c-b17c-9c756737d365",
                "name": "Gengchen Mai",
                "collaborators": [
                    "Krzysztof Janowicz",
                    "Tianming Liu",
                    "Ni Lao",
                    "Rui Zhu",
                    "Sheng Li",
                    "Zhangyu Wang",
                    "Kitty Currier",
                    "Zilong Liu",
                    "Jiaming Song",
                    "Stefano Ermon",
                    "Haixing Dai",
                    "Ninghao Liu",
                    "C. Shimizu",
                    "Ling Cai",
                    "Colby K. Fisher",
                    "Pascal Hitzler",
                    "M. Schildhauer",
                    "Meilin Shi",
                    "Yuanyuan Tian",
                    "Jinmeng Rao",
                    "Ziyuan Li",
                    "Song Gao",
                    "Yutong He",
                    "Zheng Liu",
                    "Zihao Wu",
                    "Suhang Song",
                    "Dajiang Zhu",
                    "Xiang Li",
                    "Jin Sun",
                    "Shirly Stephe",
                    "Adrita Barua",
                    "Antrea Christou",
                    "Abhilekha Dalal",
                    "Wenwen Li",
                    "Mohammad Saeid Mahdavinejad",
                    "Dean Rehberger",
                    "Sanaz Saki Norouzi",
                    "Sizhe Wang",
                    "Joseph Zalewski",
                    "Lu Zhou",
                    "Ivan Majic",
                    "Ye Qi",
                    "Michael Zhang",
                    "Weiwei Sun",
                    "Yuchi Ma",
                    "Chenlin Meng",
                    "Hongxu Ma",
                    "Jielu Zhang",
                    "Zhongliang Zhou",
                    "Lan Mu",
                    "Mengxuan Hu",
                    "Shirly Stephen",
                    "Anna Lopez-Carr",
                    "Andrew Schroeder",
                    "Tony Huang",
                    "Yiwei Li",
                    "Lin Zhao",
                    "Ye Shen",
                    "Xiaobai Yao",
                    "Lu Shi",
                    "Quanzheng Li",
                    "Zhuo Chen",
                    "D. Zhang",
                    "E. Latif",
                    "Matthew Nyaaba",
                    "Xuansheng Wu",
                    "Guoyu Lu",
                    "Xiaoming Zhai",
                    "G. Lu",
                    "L. Chai",
                    "Haijian Sun",
                    "Xianqiao Wang",
                    "Ning Liu",
                    "R. Xu",
                    "Daniel Petti",
                    "Changying Li",
                    "Weiming Huang",
                    "Deepak Mishra",
                    "G. Cong",
                    "Yingjie Hu",
                    "Chris Cundy",
                    "Yucheng Shi",
                    "Hehuan Ma",
                    "Wenliang Zhong",
                    "Junzhou Huang",
                    "Saed Rezayi",
                    "Chandra Dhakal",
                    "Bao Ge",
                    "Chen Zhen",
                    "Nina Wiedemann",
                    "J. Verstegen",
                    "Grant McKenzie",
                    "A. Graser",
                    "Yao Xuan",
                    "Wen-Hang Zuo",
                    "K. Janowicz"
                ],
                "domain": [
                    "Geospatial AI",
                    "Machine Learning",
                    "Knowledge Graph",
                    "Remote Sensing"
                ],
                "publications": [
                    {
                        "title": "MC-GTA: Metric-Constrained Model-Based Clustering using Goodness-of-fit Tests with Autocorrelations",
                        "abstract": "A wide range of (multivariate) temporal (1D) and spatial (2D) data analysis tasks, such as grouping vehicle sensor trajectories, can be formulated as clustering with given metric constraints. Existing metric-constrained clustering algorithms overlook the rich correlation between feature similarity and metric distance, i.e., metric autocorrelation. The model-based variations of these clustering algorithms (e.g. TICC and STICC) achieve SOTA performance, yet suffer from computational instability and complexity by using a metric-constrained Expectation-Maximization procedure. In order to address these two problems, we propose a novel clustering algorithm, MC-GTA (Model-based Clustering via Goodness-of-fit Tests with Autocorrelations). Its objective is only composed of pairwise weighted sums of feature similarity terms (square Wasserstein-2 distance) and metric autocorrelation terms (a novel multivariate generalization of classic semivariogram). We show that MC-GTA is effectively minimizing the total hinge loss for intra-cluster observation pairs not passing goodness-of-fit tests, i.e., statistically not originating from the same distribution. Experiments on 1D/2D synthetic and real-world datasets demonstrate that MC-GTA successfully incorporates metric autocorrelation. It outperforms strong baselines by large margins (up to 14.3% in ARI and 32.1% in NMI) with faster and stabler optimization (>10x speedup)."
                    },
                    {
                        "title": "The KnowWhereGraph Ontology",
                        "abstract": "KnowWhereGraph is one of the largest fully publicly available geospatial knowledge graphs. It includes data from 30 layers on natural hazards (e.g., hurricanes, wildfires), climate variables (e.g., air temperature, precipitation), soil properties, crop and land-cover types, demographics, and human health, various place and region identifiers, among other themes. These have been leveraged through the graph by a variety of applications to address challenges in food security and agricultural supply chains; sustainability related to soil conservation practices and farm labor; and delivery of emergency humanitarian aid following a disaster. In this paper, we introduce the ontology that acts as the schema for KnowWhereGraph. This broad overview provides insight into the requirements and design specifications for the graph and its schema, including the development methodology (modular ontology modeling) and the resources utilized to implement, materialize, and deploy KnowWhereGraph with its end-user interfaces and public query SPARQL endpoint."
                    },
                    {
                        "title": "Probing the Information Theoretical Roots of Spatial Dependence Measures",
                        "abstract": "Intuitively, there is a relation between measures of spatial dependence and information theoretical measures of entropy. For instance, we can provide an intuition of why spatial data is special by stating that, on average, spatial data samples contain less than expected information. Similarly, spatial data, e.g., remotely sensed imagery, that is easy to compress is also likely to show significant spatial autocorrelation. Formulating our (highly specific) core concepts of spatial information theory in the widely used language of information theory opens new perspectives on their differences and similarities and also fosters cross-disciplinary collaboration, e.g., with the broader AI/ML communities. Interestingly, however, this intuitive relation is challenging to formalize and generalize, leading prior work to rely mostly on experimental results, e.g., for describing landscape patterns. In this work, we will explore the information theoretical roots of spatial autocorrelation, more specifically Moran's I, through the lens of self-information (also known as surprisal) and provide both formal proofs and experiments."
                    },
                    {
                        "title": "EVKG: An interlinked and interoperable electric vehicle knowledge graph for smart transportation system",
                        "abstract": "Over the past decade, the electric vehicle (EV) industry has experienced unprecedented growth and diversification, resulting in a complex ecosystem. To effectively manage this multifaceted field, we present an EV\u2010centric knowledge graph (EVKG) as a comprehensive, cross\u2010domain, extensible, and open geospatial knowledge management system. The EVKG encapsulates essential EV\u2010related knowledge, including EV adoption, EV supply equipment, and electricity transmission network, to support decision\u2010making related to EV technology development, infrastructure planning, and policy\u2010making by providing timely and accurate information and analysis. To enrich and contextualize the EVKG, we integrate the developed EV\u2010relevant ontology modules from existing well\u2010known knowledge graphs and ontologies. This integration enables interoperability with other knowledge graphs in the Linked Data Open Cloud, enhancing the EVKG's value as a knowledge hub for EV decision\u2010making. Using six competency questions, we demonstrate how the EVKG can be used to answer various types of EV\u2010related questions, providing critical insights into the EV ecosystem. Our EVKG provides an efficient and effective approach for managing the complex and diverse EV industry. By consolidating critical EV\u2010related knowledge into a single, easily accessible resource, the EVKG supports decision\u2010makers in making informed choices about EV technology development, infrastructure planning, and policy\u2010making. As a flexible and extensible platform, the EVKG is capable of accommodating a wide range of data sources, enabling it to evolve alongside the rapidly changing EV landscape."
                    },
                    {
                        "title": "SSIF: Learning Continuous Image Representation for Spatial-Spectral Super-Resolution",
                        "abstract": "Existing digital sensors capture images at fixed spatial and spectral resolutions (e.g., RGB, multispectral, and hyperspectral images), and each combination requires bespoke machine learning models. Neural Implicit Functions partially overcome the spatial resolution challenge by representing an image in a resolution-independent way. However, they still operate at fixed, pre-defined spectral resolutions. To address this challenge, we propose Spatial-Spectral Implicit Function (SSIF), a neural implicit model that represents an image as a function of both continuous pixel coordinates in the spatial domain and continuous wavelengths in the spectral domain. We empirically demonstrate the effectiveness of SSIF on two challenging spatio-spectral super-resolution benchmarks. We observe that SSIF consistently outperforms state-of-the-art baselines even when the baselines are allowed to train separate models at each spectral resolution. We show that SSIF generalizes well to both unseen spatial resolutions and spectral resolutions. Moreover, SSIF can generate high-resolution images that improve the performance of downstream tasks (e.g., land use classification) by 1.7%-7%."
                    },
                    {
                        "title": "Text2Seg: Remote Sensing Image Semantic Segmentation via Text-Guided Visual Foundation Models",
                        "abstract": "Remote sensing imagery has attracted significant attention in recent years due to its instrumental role in global environmental monitoring, land usage monitoring, and more. As image databases grow each year, performing automatic segmentation with deep learning models has gradually become the standard approach for processing the data. Despite the improved performance of current models, certain limitations remain unresolved. Firstly, training deep learning models for segmentation requires per-pixel annotations. Given the large size of datasets, only a small portion is fully annotated and ready for training. Additionally, the high intra-dataset variance in remote sensing data limits the transfer learning ability of such models. Although recently proposed generic segmentation models like SAM have shown promising results in zero-shot instance-level segmentation, adapting them to semantic segmentation is a non-trivial task. To tackle these challenges, we propose a novel method named Text2Seg for remote sensing semantic segmentation. Text2Seg overcomes the dependency on extensive annotations by employing an automatic prompt generation process using different visual foundation models (VFMs), which are trained to understand semantic information in various ways. This approach not only reduces the need for fully annotated datasets but also enhances the model's ability to generalize across diverse datasets. Evaluations on four widely adopted remote sensing datasets demonstrate that Text2Seg significantly improves zero-shot prediction performance compared to the vanilla SAM model, with relative improvements ranging from 31% to 225%. Our code is available at https://github.com/Douglas2Code/Text2Seg."
                    },
                    {
                        "title": "Building Privacy-Preserving and Secure Geospatial Artificial Intelligence Foundation Models (Vision Paper)",
                        "abstract": "In recent years we have seen substantial advances in foundation models for artificial intelligence, including language, vision, and multimodal models. Recent studies have highlighted the potential of using foundation models in geospatial artificial intelligence, known as GeoAI Foundation Models, for geographic question answering, remote sensing image understanding, map generation, and location-based services, among others. However, the development and application of GeoAI foundation models can pose serious privacy and security risks, which have not been fully discussed or addressed to date. This paper introduces the potential privacy and security risks throughout the lifecycle of GeoAI foundation models and proposes a comprehensive blueprint for research directions and preventative and control strategies. Through this vision paper, we hope to draw the attention of researchers and policymakers in geospatial domains to these privacy and security risks inherent in GeoAI foundation models and advocate for the development of privacy-preserving and secure GeoAI foundation models."
                    },
                    {
                        "title": "CSP: Self-Supervised Contrastive Spatial Pre-Training for Geospatial-Visual Representations",
                        "abstract": "Geo-tagged images are publicly available in large quantities, whereas labels such as object classes are rather scarce and expensive to collect. Meanwhile, contrastive learning has achieved tremendous success in various natural image and language tasks with limited labeled data. However, existing methods fail to fully leverage geospatial information, which can be paramount to distinguishing objects that are visually similar. To directly leverage the abundant geospatial information associated with images in pre-training, fine-tuning, and inference stages, we present Contrastive Spatial Pre-Training (CSP), a self-supervised learning framework for geo-tagged images. We use a dual-encoder to separately encode the images and their corresponding geo-locations, and use contrastive objectives to learn effective location representations from images, which can be transferred to downstream supervised tasks such as image classification. Experiments show that CSP can improve model performance on both iNat2018 and fMoW datasets. Especially, on iNat2018, CSP significantly boosts the model performance with 10-34% relative improvement with various labeled training data sampling ratios."
                    },
                    {
                        "title": "The Expertise Ontology: Modeling Expertise in the Context of Emergency Management",
                        "abstract": "It is crucial for emergency management organizations to have rapid access to relevant experts who can advise and assist following a disaster. To improve expert-mining and recommendation capabilities, creating a knowledge graph that links experts to their corresponding topics of expertise and other sources of relevant information is a natural choice to capture an integrated network of people and a rich taxonomy of expertise. In this paper, we present an ontology for modeling experts, their expertise topics and relations between them, and their spatiotemporal scoping. We go on to discuss the primary conceptual components and how they can be instantiated, then present overarching examples related to emergency management operations. The ontology synthesizes three different ways to characterize an expert, based on a) identifiable academic expertise; b) voluntary engagements, work-related responsibilities or experience; and c) organization specializations or affiliations."
                    },
                    {
                        "title": "AD-AutoGPT: An Autonomous GPT for Alzheimer's Disease Infodemiology",
                        "abstract": "In this pioneering study, inspired by AutoGPT, the state-of-the-art open-source application based on the GPT-4 large language model, we develop a novel tool called AD-AutoGPT which can conduct data collection, processing, and analysis about complex health narratives of Alzheimer's Disease in an autonomous manner via users' textual prompts. We collated comprehensive data from a variety of news sources, including the Alzheimer's Association, BBC, Mayo Clinic, and the National Institute on Aging since June 2022, leading to the autonomous execution of robust trend analyses, intertopic distance maps visualization, and identification of salient terms pertinent to Alzheimer's Disease. This approach has yielded not only a quantifiable metric of relevant discourse but also valuable insights into public focus on Alzheimer's Disease. This application of AD-AutoGPT in public health signifies the transformative potential of AI in facilitating a data-rich understanding of complex health narratives like Alzheimer's Disease in an autonomous manner, setting the groundwork for future AI-driven investigations in global health landscapes."
                    },
                    {
                        "title": "AGI: Artificial General Intelligence for Education",
                        "abstract": "Artificial general intelligence (AGI) has gained global recognition as a future technology due to the emergence of breakthrough large language models and chatbots such as GPT-4 and ChatGPT, respectively. Compared to conventional AI models, typically designed for a limited range of tasks, demand significant amounts of domain-specific data for training and may not always consider intricate interpersonal dynamics in education. AGI, driven by the recent large pre-trained models, represents a significant leap in the capability of machines to perform tasks that require human-level intelligence, such as reasoning, problem-solving, decision-making, and even understanding human emotions and social interactions. This position paper reviews AGI's key concepts, capabilities, scope, and potential within future education, including achieving future educational goals, designing pedagogy and curriculum, and performing assessments. It highlights that AGI can significantly improve intelligent tutoring systems, educational assessment, and evaluation procedures. AGI systems can adapt to individual student needs, offering tailored learning experiences. They can also provide comprehensive feedback on student performance and dynamically adjust teaching methods based on student progress. The paper emphasizes that AGI's capabilities extend to understanding human emotions and social interactions, which are critical in educational settings. The paper discusses that ethical issues in education with AGI include data bias, fairness, and privacy and emphasizes the need for codes of conduct to ensure responsible AGI use in academic settings like homework, teaching, and recruitment. We also conclude that the development of AGI necessitates interdisciplinary collaborations between educators and AI engineers to advance research and application efforts."
                    },
                    {
                        "title": "AGI for Agriculture",
                        "abstract": "Artificial General Intelligence (AGI) is poised to revolutionize a variety of sectors, including healthcare, finance, transportation, and education. Within healthcare, AGI is being utilized to analyze clinical medical notes, recognize patterns in patient data, and aid in patient management. Agriculture is another critical sector that impacts the lives of individuals worldwide. It serves as a foundation for providing food, fiber, and fuel, yet faces several challenges, such as climate change, soil degradation, water scarcity, and food security. AGI has the potential to tackle these issues by enhancing crop yields, reducing waste, and promoting sustainable farming practices. It can also help farmers make informed decisions by leveraging real-time data, leading to more efficient and effective farm management. This paper delves into the potential future applications of AGI in agriculture, such as agriculture image processing, natural language processing (NLP), robotics, knowledge graphs, and infrastructure, and their impact on precision livestock and precision crops. By leveraging the power of AGI, these emerging technologies can provide farmers with actionable insights, allowing for optimized decision-making and increased productivity. The transformative potential of AGI in agriculture is vast, and this paper aims to highlight its potential to revolutionize the industry."
                    },
                    {
                        "title": "On the Opportunities and Challenges of Foundation Models for Geospatial Artificial Intelligence",
                        "abstract": "Large pre-trained models, also known as foundation models (FMs), are trained in a task-agnostic manner on large-scale data and can be adapted to a wide range of downstream tasks by fine-tuning, few-shot, or even zero-shot learning. Despite their successes in language and vision tasks, we have yet seen an attempt to develop foundation models for geospatial artificial intelligence (GeoAI). In this work, we explore the promises and challenges of developing multimodal foundation models for GeoAI. We first investigate the potential of many existing FMs by testing their performances on seven tasks across multiple geospatial subdomains including Geospatial Semantics, Health Geography, Urban Geography, and Remote Sensing. Our results indicate that on several geospatial tasks that only involve text modality such as toponym recognition, location description recognition, and US state-level/county-level dementia time series forecasting, these task-agnostic LLMs can outperform task-specific fully-supervised models in a zero-shot or few-shot learning setting. However, on other geospatial tasks, especially tasks that involve multiple data modalities (e.g., POI-based urban function classification, street view image-based urban noise intensity classification, and remote sensing image scene classification), existing foundation models still underperform task-specific models. Based on these observations, we propose that one of the major challenges of developing a FM for GeoAI is to address the multimodality nature of geospatial tasks. After discussing the distinct challenges of each geospatial data modality, we suggest the possibility of a multimodal foundation model which can reason over various types of geospatial data through geospatial alignments. We conclude this paper by discussing the unique risks and challenges to develop such a model for GeoAI."
                    },
                    {
                        "title": "ChatGraph: Interpretable Text Classification by Converting ChatGPT Knowledge to Graphs",
                        "abstract": "ChatGPT, as a recently launched large language model (LLM), has shown superior performance in various natural language processing (NLP) tasks. However, two major limitations hinder its potential applications: 1) the inflexibility of finetuning on downstream tasks, and 2) the lack of interpretability in the decision-making process. To tackle these limitations, we propose a novel framework that leverages the power of ChatGPT for specific tasks, such as text classification, while improving its interpretability. The proposed framework conducts a knowledge graph extraction task to extract refined and structural knowledge from the raw data using ChatGPT. The rich knowledge is then converted into a graph, which is further used to train an interpretable linear classifier to make predictions. To evaluate the effectiveness of our proposed method, we conduct experiments on four benchmark datasets. The results demonstrate that our method can significantly improve the prediction performance compared to directly utilizing ChatGPT for text classification tasks. Furthermore, our method provides a more transparent decision-making process compared with previous text classification methods. The code is available at https://github.com/sycny/ChatGraph."
                    },
                    {
                        "title": "Exploring New Frontiers in Agricultural NLP: Investigating the Potential of Large Language Models for Food Applications",
                        "abstract": "This paper explores new frontiers in agricultural natural language processing by investigating the effectiveness of using food-related text corpora for pretraining transformer-based language models. In particular, we focus on the task of semantic matching, which involves establishing mappings between food descriptions and nutrition data. To accomplish this, we fine-tune a pre-trained transformer-based language model, AgriBERT, on this task, utilizing an external source of knowledge, such as the FoodOn ontology. To advance the field of agricultural NLP, we propose two new avenues of exploration: (1) utilizing GPT-based models as a baseline and (2) leveraging ChatGPT as an external source of knowledge. ChatGPT has shown to be a strong baseline in many NLP tasks, and we believe it has the potential to improve our model in the task of semantic matching and enhance our model's understanding of food-related concepts and relationships. Additionally, we experiment with other applications, such as cuisine prediction based on food ingredients, and expand the scope of our research to include other NLP tasks beyond semantic matching. Overall, this paper provides promising avenues for future research in this field, with potential implications for improving the performance of agricultural NLP applications."
                    },
                    {
                        "title": "Thinking Geographically about AI Sustainability",
                        "abstract": "Abstract. Driven by foundation models, recent progress in AI and machine learning has reached unprecedented complexity. For instance, the GPT-3 language model consists of 175 billion parameters and a training-data size of 570 GB. While it has achieved remarkable performance in generating text that is difficult to distinguish from human-authored content, a single training of the model is estimated to produce over 550 metric tons of CO2 emissions. Likewise, we see advances in GeoAI research improving large-scale prediction tasks like satellite image classification and global climate modeling, to name but a couple. While these models have not yet reached comparable complexity and emissions levels, spatio-temporal models differ from language and image-generation models in several ways that make it necessary to (re)train them more often, with potentially large implications for sustainability. While recent work in the machine learning community has started calling for greener and more energy-efficient AI alongside improvements in model accuracy, this trend has not yet reached the GeoAI community at large. In this work, we bring this issue to not only the attention of the GeoAI community but also present ethical considerations from a geographic perspective that are missing from the broader, ongoing AI-sustainability discussion. To start this discussion, we propose a framework to evaluate models from several sustainability-related angles, including energy efficiency, carbon intensity, transparency, and social implications. We encourage future AI/GeoAI work to acknowledge its environmental impact as a step towards a more resource-conscious society. Similar to the current push for reproducibility, future publications should also report the energy/carbon costs of improvements over prior work. "
                    },
                    {
                        "title": "Sphere2Vec: A General-Purpose Location Representation Learning over a Spherical Surface for Large-Scale Geospatial Predictions",
                        "abstract": "Generating learning-friendly representations for points in space is a fundamental and long-standing problem in ML. Recently, multi-scale encoding schemes (such as Space2Vec and NeRF) were proposed to directly encode any point in 2D/3D Euclidean space as a high-dimensional vector, and has been successfully applied to various geospatial prediction and generative tasks. However, all current 2D and 3D location encoders are designed to model point distances in Euclidean space. So when applied to large-scale real-world GPS coordinate datasets, which require distance metric learning on the spherical surface, both types of models can fail due to the map projection distortion problem (2D) and the spherical-to-Euclidean distance approximation error (3D). To solve these problems, we propose a multi-scale location encoder called Sphere2Vec which can preserve spherical distances when encoding point coordinates on a spherical surface. We developed a unified view of distance-reserving encoding on spheres based on the DFS. We also provide theoretical proof that the Sphere2Vec preserves the spherical surface distance between any two points, while existing encoding schemes do not. Experiments on 20 synthetic datasets show that Sphere2Vec can outperform all baseline models on all these datasets with up to 30.8% error rate reduction. We then apply Sphere2Vec to three geo-aware image classification tasks - fine-grained species recognition, Flickr image recognition, and remote sensing image classification. Results on 7 real-world datasets show the superiority of Sphere2Vec over multiple location encoders on all three tasks. Further analysis shows that Sphere2Vec outperforms other location encoder models, especially in the polar regions and data-sparse areas because of its nature for spherical surface distance preservation. Code and data are available at https://gengchenmai.github.io/sphere2vec-website/."
                    }
                ]
            },
            "f59d925e-892c-43b4-b0f2-d39e445aea6c": {
                "pk": "f59d925e-892c-43b4-b0f2-d39e445aea6c",
                "name": "Marshall Burke",
                "collaborators": [
                    "Samar Khanna",
                    "David B. Lobell",
                    "Stefano Ermon",
                    "Rohin Manvi",
                    "Patrick Liu",
                    "Linqi Zhou",
                    "Chenlin Meng",
                    "Robin Rombach"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Remote Sensing",
                    "Generative Models",
                    "Bias Evaluation"
                ],
                "publications": [
                    {
                        "title": "Large Language Models are Geographically Biased",
                        "abstract": "Large Language Models (LLMs) inherently carry the biases contained in their training corpora, which can lead to the perpetuation of societal harm. As the impact of these foundation models grows, understanding and evaluating their biases becomes crucial to achieving fairness and accuracy. We propose to study what LLMs know about the world we live in through the lens of geography. This approach is particularly powerful as there is ground truth for the numerous aspects of human life that are meaningfully projected onto geographic space such as culture, race, language, politics, and religion. We show various problematic geographic biases, which we define as systemic errors in geospatial predictions. Initially, we demonstrate that LLMs are capable of making accurate zero-shot geospatial predictions in the form of ratings that show strong monotonic correlation with ground truth (Spearman's $\\rho$ of up to 0.89). We then show that LLMs exhibit common biases across a range of objective and subjective topics. In particular, LLMs are clearly biased against locations with lower socioeconomic conditions (e.g. most of Africa) on a variety of sensitive subjective topics such as attractiveness, morality, and intelligence (Spearman's $\\rho$ of up to 0.70). Finally, we introduce a bias score to quantify this and find that there is significant variation in the magnitude of bias across existing LLMs. Code is available on the project website: https://rohinmanvi.github.io/GeoLLM"
                    },
                    {
                        "title": "DiffusionSat: A Generative Foundation Model for Satellite Imagery",
                        "abstract": "Diffusion models have achieved state-of-the-art results on many modalities including images, speech, and video. However, existing models are not tailored to support remote sensing data, which is widely used in important applications including environmental monitoring and crop-yield prediction. Satellite images are significantly different from natural images -- they can be multi-spectral, irregularly sampled across time -- and existing diffusion models trained on images from the Web do not support them. Furthermore, remote sensing data is inherently spatio-temporal, requiring conditional generation tasks not supported by traditional methods based on captions or images. In this paper, we present DiffusionSat, to date the largest generative foundation model trained on a collection of publicly available large, high-resolution remote sensing datasets. As text-based captions are sparsely available for satellite images, we incorporate the associated metadata such as geolocation as conditioning information. Our method produces realistic samples and can be used to solve multiple generative tasks including temporal generation, superresolution given multi-spectral inputs and in-painting. Our method outperforms previous state-of-the-art methods for satellite image generation and is the first large-scale generative foundation model for satellite imagery. The project website can be found here: https://samar-khanna.github.io/DiffusionSat/"
                    }
                ]
            },
            "eb41e013-102b-41f8-809d-04ca6a23b9f3": {
                "pk": "eb41e013-102b-41f8-809d-04ca6a23b9f3",
                "name": "David Lobell",
                "collaborators": [
                    "Samar Khanna",
                    "Stefano Ermon",
                    "Marshall Burke",
                    "J. Deines",
                    "Yi Yang",
                    "David Tilman",
                    "Zhenong Jin",
                    "Peter Smith",
                    "Christopher B Barrett",
                    "Yongguan Zhu",
                    "Jennifer Burney",
                    "Paolo D'Odorico",
                    "P. Fantke",
                    "Joseph E. Fargione",
                    "Jacques C. Finlay",
                    "M. Rulli",
                    "Lindsey Sloat",
                    "Kees Jan van Groenigen",
                    "Paul C. West",
                    "Lewis Ziska",
                    "Anna M. Michalak",
                    "Medhanie Irgau",
                    "S. Archontoulis",
                    "I. Huber",
                    "Rohin Manvi",
                    "Trevor W Cambron",
                    "Bruno Lopez",
                    "Rinkal Patel",
                    "Sang-Zi Liang",
                    "Karli A. Moore",
                    "Liyin He",
                    "L. Rosa",
                    "Yuan Wang",
                    "Y. Yin",
                    "R. Doughty",
                    "Yitong Yao",
                    "J. Berry",
                    "Christian Frankenberg",
                    "Angela Tsao",
                    "Ikenna Nzewi",
                    "Ayodeji Jayeoba",
                    "Uzoma Ayogu",
                    "Dan M. Kluger",
                    "Art B. Owen",
                    "Corisa A Wong",
                    "M. Mauter",
                    "Patrick Liu",
                    "Linqi Zhou",
                    "Chenlin Meng",
                    "Robin Rombach",
                    "J. Dykema",
                    "Edward Miguel",
                    "S. Satyanath"
                ],
                "domain": [
                    "Climate Change",
                    "Agriculture",
                    "Remote Sensing",
                    "Environmental Impact"
                ],
                "publications": [
                    {
                        "title": "Climate change exacerbates the environmental impacts of agriculture",
                        "abstract": "Agriculture\u2019s global environmental impacts are widely expected to continue expanding, driven by population and economic growth and dietary changes. This Review highlights climate change as an additional amplifier of agriculture\u2019s environmental impacts, by reducing agricultural productivity, reducing the efficacy of agrochemicals, increasing soil erosion, accelerating the growth and expanding the range of crop diseases and pests, and increasing land clearing. We identify multiple pathways through which climate change intensifies agricultural greenhouse gas emissions, creating a potentially powerful climate change\u2013reinforcing feedback loop. The challenges raised by climate change underscore the urgent need to transition to sustainable, climate-resilient agricultural systems. This requires investments that both accelerate adoption of proven solutions that provide multiple benefits, and that discover and scale new beneficial processes and food products. Editor\u2019s summary Intensive agriculture creates multiple environmental challenges, including runoff of excess nutrients, use of chemical pesticides, biodiversity loss, and greenhouse gas emissions. Curbing these negative effects of agriculture while feeding a growing global population is already a massive challenge and one that will likely be exacerbated by climate change. Yang et al. reviewed the research on interactions between climate change and agriculture\u2019s environmental impacts and found that most of this work suggests even greater challenges to come. \u2014Bianca Lopez BACKGROUND Modern agriculture has large global environmental impacts, from greenhouse gas (GHG) emissions to water and air pollution to biodiversity loss. It is widely expected that agriculture\u2019s environmental impacts will escalate because a growing and richer global population will likely demand more agricultural products, including foods, feeds, fibers, and biofuels. Climate change, though less studied in this regard, is a second potential amplifier of agricultural environmental impacts. As climate change intensifies, it not only poses great risks to agricultural productivity, but its deviation from the relatively stable state under which modern agriculture evolved and has long operated also has profound implications for how agriculture interacts with the environment, and vice versa. In this Review, we present a synthesis of how climate change could amplify the environmental impacts of agriculture, from increases in GHG emissions, water use and scarcity, soil erosion, nitrogen and phosphorus pollution, pest outbreaks and pesticide use and pollution, and biodiversity loss. We discuss solutions to the challenges raised by how climate change could affect agricultural sustainability. ADVANCES Two major findings emerge from our synthesis. First, climate change will likely exacerbate the already large environmental impacts of agricultural production. It would do so by (i) directly and negatively affecting agricultural productivity; (ii) reducing the efficacy of agrochemicals and increasing their loss to the environment; and (iii) increasing crop pests and soil erosion. Yield reductions and loss of fertile soils would lead to increased land clearing and subsequent species extinctions and GHG emissions, and/or to agrochemical intensification to retain or increase yields. Greater agrochemical inputs mean greater chemical pollution of the environment and greater GHG emissions. Decreased agrochemical efficacy could also lead to compensatory increases in agrochemical use. These climatic impacts can be independent of, hence additive to or multiplicative of, one another, making climate change a potentially important amplifier of agriculture\u2019s environmental footprints. Second, agriculture already accounts for almost a quarter of global GHG emissions, and agriculture\u2019s response to climate change could create a powerful positive GHG feedback loop through multiple pathways. Climate change could intensify agricultural GHG emissions directly\u2014for example, through increasing CH4 emissions from rice paddies, N2O emissions from soil, and CO2 emissions from land clearing and soil tillage. It could also induce more GHG emissions indirectly\u2014for example, if more agrochemicals are needed to compensate for their reduced efficacy; if drier conditions require more energy- and carbon-intensive irrigation; or if greater losses of nutrients from agricultural fields stimulate more biogenic GHG emissions in aquatic systems. These climate-agriculture feedback effects warrant increased attention. OUTLOOK Agricultural reliability and sustainability are of key long-term importance for human and planetary health. The challenges raised by climate change require accelerated adoption of practices and technologies that improve agriculture\u2019s environmental sustainability and climate resilience, especially those that can simultaneously deliver multiple benefits, such as diversification and integrated soil fertility management. However, socioeconomic barriers impede adoption of sustainable practices and technologies. Improved context-specific understanding of these barriers is urgently needed, as are innovative policies to overcome them. Greater investments in the discovery, adaptation, and cost reduction of emerging agricultural technologies could help make agriculture more reliable, sustainable, and climate resilient, as could the adoption of novel foods and healthier diets that would require far less land, water, and chemicals per capita than at present. Our synthesis also identifies many unanswered questions. Although the direction of climate impacts is often clear, the relevant magnitudes are not. Quantitative assessments from local to global scales will be needed, especially for the climate-agriculture feedback loop. Other feedback loops, such as the climate-agriculture-biodiversity loop, are also complicated and require mechanistic and quantitative analyses. The major environmental impacts of agricultural systems and potentially exacerbating effects of climate change (dark red circle in the center). IMAGES: Vectorstock.com"
                    },
                    {
                        "title": "ExPLoRA: Parameter-Efficient Extended Pre-Training to Adapt Vision Transformers under Domain Shifts",
                        "abstract": "Parameter-efficient fine-tuning (PEFT) techniques such as low-rank adaptation (LoRA) can effectively adapt large pre-trained foundation models to downstream tasks using only a small fraction (0.1%-10%) of the original trainable weights. An under-explored question of PEFT is in extending the pre-training phase without supervised labels; that is, can we adapt a pre-trained foundation model to a new domain via efficient self-supervised pre-training on this new domain? In this work, we introduce ExPLoRA, a highly effective technique to improve transfer learning of pre-trained vision transformers (ViTs) under domain shifts. Initializing a ViT with pre-trained weights on large, natural-image datasets such as from DinoV2 or MAE, ExPLoRA continues the unsupervised pre-training objective on a new domain, unfreezing 1-2 pre-trained ViT blocks and tuning all other layers with LoRA. We then fine-tune the resulting model only with LoRA on this new domain for supervised learning. Our experiments demonstrate state-of-the-art results on satellite imagery, even outperforming fully pre-training and fine-tuning ViTs. Using the DinoV2 training objective, we demonstrate up to 7.5% improvement in linear probing top-1 accuracy on downstream tasks while using<10% of the number of parameters that are used in prior fully-tuned state-of-the art approaches. Our ablation studies confirm the efficacy of our approach over other baselines, including PEFT and unfreezing more ViT blocks. Code is available on the project website: https://samar-khanna.github.io/ExPLoRA/"
                    },
                    {
                        "title": "Observational evidence for groundwater influence on crop yields in the United States",
                        "abstract": "Significance The US Corn Belt is a major global producer of both maize and soybeans, with maize output alone worth more than $30 billion per year. Hydrological extremes\u2014both dry and wet\u2014can greatly reduce yields, although most prior work has focused solely on near-surface soil moisture. Here, we demonstrate with region-wide observational data that accessible groundwater increases yield stability over time by providing supplemental water, especially in years with below-average rainfall. Our results suggest that groundwater yield contribution is likely more modest (up to ~7% in dry conditions) than previous work relying on idealistic crop simulations suggests (up to ~24% in dry conditions). Properly understanding and accounting for groundwater\u2019s role will inform climate adaptation in the region."
                    },
                    {
                        "title": "Large Language Models are Geographically Biased",
                        "abstract": "Large Language Models (LLMs) inherently carry the biases contained in their training corpora, which can lead to the perpetuation of societal harm. As the impact of these foundation models grows, understanding and evaluating their biases becomes crucial to achieving fairness and accuracy. We propose to study what LLMs know about the world we live in through the lens of geography. This approach is particularly powerful as there is ground truth for the numerous aspects of human life that are meaningfully projected onto geographic space such as culture, race, language, politics, and religion. We show various problematic geographic biases, which we define as systemic errors in geospatial predictions. Initially, we demonstrate that LLMs are capable of making accurate zero-shot geospatial predictions in the form of ratings that show strong monotonic correlation with ground truth (Spearman's $\\rho$ of up to 0.89). We then show that LLMs exhibit common biases across a range of objective and subjective topics. In particular, LLMs are clearly biased against locations with lower socioeconomic conditions (e.g. most of Africa) on a variety of sensitive subjective topics such as attractiveness, morality, and intelligence (Spearman's $\\rho$ of up to 0.70). Finally, we introduce a bias score to quantify this and find that there is significant variation in the magnitude of bias across existing LLMs. Code is available on the project website: https://rohinmanvi.github.io/GeoLLM"
                    },
                    {
                        "title": "Further adoption of conservation tillage can increase maize yields in the western US Corn Belt",
                        "abstract": "Conservation tillage can reduce soil erosion, increase soil health, and decrease labor and fuel input costs. Despite these benefits, potential yield impacts remain an important concern for farmers considering adoption. Previous research suggests that conservation tillage is likely to have the largest yield benefits in more arid conditions, but a lack of field-level analyses across climatic, management and soil conditions limits confidence in such predictions. Satellite imagery provides the opportunity to monitor agricultural lands at sub-field resolution across large spatial scales and wide environmental gradients. Here we investigate the maize yield impacts of conservation tillage in the semi-arid western US Corn Belt, using sub-field resolution datasets on tillage practices and crop yields derived from satellite data spanning four states (Nebraska, Kansas, South Dakota, and North Dakota) between 2008 and 2020. On these datasets, we estimate heterogenous yield outcomes for several thousand maize fields across gradients in climate, soil quality and irrigation status by using a causal forests analysis, an adaptation of the random forests machine-learning algorithm for causal inference on observational data. We find that long-term adoption of conservation tillage increased rainfed maize yields by an average of 9.9% in the region. Impacts on irrigated yields were small and not statistically significant. These results, along with an analysis of variables related to greater than average yield benefits, indicate that improved water infiltration and retention are the primary reasons for conservation tillage benefits. Despite yield benefits, many fields estimated to see increased yields under long term low till have not adopted the practice. Therefore, we identify specific counties likely to benefit most from increased levels of adoption. Our results strengthen the understanding of the impacts of conservation agriculture on crop yields and help define environments and counties most likely to benefit from conservation tillage."
                    },
                    {
                        "title": "Opportunities and Barriers for Agrivoltaics on Tribal Lands",
                        "abstract": "Recent federal legislation, like the 2021 Infrastructure Investment and Jobs Act and 2022 Inflation Reduction Act, has led to a push for more solar energy on Tribal lands, increasing competition for already limited agricultural land. Agrivoltaics is an innovative technology with the potential to lessen the tradeoffs between agriculture production and solar energy generation. This study investigates the opportunities and barriers for agrivoltaics on Tribal lands through expert qualitative interviews with Tribal agriculture professionals that inform geospatial suitability analysis of physical characteristics. Qualitative results indicate agrivoltaics on Tribal lands could contribute positively to food sovereignty, energy sovereignty, and economic development goals for Tribes; on the other hand, Tribal agriculture professionals have technical, economic, siting, and socioecological concerns that should be addressed through future work. Quantitatively, we find up to 15 million acres of Tribal agricultural land may be feasible for micro-grid agrivoltaics, with 7 million acres in sufficient proximity to existing transmission lines to tie into the grid. The leading states for Tribal agrivoltaics by land area are South Dakota, Montana, and Arizona, each home to Tribes with large land bases and a strong agricultural economy. This work aims to inform Tribal land managers, policymakers, and researchers on the opportunities and barriers for agrivoltaics on Tribal lands."
                    },
                    {
                        "title": "The weekly cycle of photosynthesis in Europe reveals the negative impact of particulate pollution on ecosystem productivity",
                        "abstract": "Significance The impact of particulate pollution on photosynthesis at continental scales remains uncertain. We report evidence of a widespread negative relationship between photosynthesis and aerosol loading across Europe using space-based observations of solar-induced fluorescence (SIF) that are mechanistically linked to plant photosynthesis. Specifically, we find that 64% of regions in Europe show increased SIF on weekends, when there is more plant-available sunlight due to reduced particulate pollution. Using satellite observational datasets, our findings provide compelling evidence that reducing particulate pollution has the potential to enhance ecosystem productivity by absorbing carbon from the atmosphere."
                    },
                    {
                        "title": "Canopy Height Mapping for Plantations in Nigeria Using GEDI, Landsat, and Sentinel-2",
                        "abstract": "Canopy height data from the Global Ecosystem Dynamics Investigation (GEDI) mission has powered the development of global forest height products, but these data and products have not been validated in non-forest tree plantation settings. In this study, we collected field observations of the canopy heights throughout oil palm plantations in Nigeria and evaluated the performance of existing global canopy height map (CHM) products as well as a local model trained on the GEDI and various Landsat and Sentinel-2 feature combinations. We found that existing CHMs fared poorly in the region, with mean absolute errors (MAE) of 4.2\u20136.2 m. However, the locally trained models performed well (MAE = 2.5 m), indicating that using the GEDI and optical satellite data can still be effective, even in a region with relatively sparse GEDI coverage. In addition to improved overall performance, the local model was especially effective at reducing errors for short (<5 m) trees, where the global products struggle to capture the canopy height."
                    },
                    {
                        "title": "Biases in estimates of air pollution impacts: the role of omitted variables and measurement errors",
                        "abstract": "Observational studies often use linear regression to assess the effect of ambient air pollution on outcomes of interest, such as human health indicators or crop yields. Yet pollution datasets are typically noisy and include only a subset of the potentially relevant pollutants, giving rise to both measurement error bias (MEB) and omitted variable bias (OVB). While it is well understood that these biases exist, less is understood about whether these biases tend to be positive or negative, even though it is sometimes falsely claimed that measurement error simply biases regression coefficient estimates towards zero. In this paper, we study the direction of these biases under the realistic assumptions that the concentrations of different types of air pollutants are positively correlated with each other and that each type of pollutant has a nonpositive association with the outcome variable. We demonstrate both theoretically and using simulations that under these two assumptions, the OVB will typically be negative and that more often than not the MEB for null pollutants or for pollutants that are perfectly measured will be negative. We also use a crop yield and air pollution dataset to show that these biases tend to be negative in the setting of our motivating application. We do this by introducing a validation scheme that does not require knowing the true coefficients. While this paper is motivated by studies assessing the effect of air pollutants on crop yields, the findings are also relevant to regression-based studies assessing the effect of air pollutants on human health outcomes. The validation scheme can also be used to empirically study OVB or MEB in other contexts."
                    },
                    {
                        "title": "Multicriteria Suitability Index for Prioritizing Early-Stage Deployments of Wastewater-Derived Fertilizers in Sub-Saharan Africa.",
                        "abstract": "Recycling nutrients from wastewater could simultaneously decrease the carbon intensity of traditional ammonia supply chains and increase the accessibility of local fertilizer. Despite the theoretical potential, techno-economic viability of wastewater nutrient recovery in sub-Saharan Africa has been poorly characterized at subnational scales. This work proposes a multicriteria suitability index to describe techno-economic viability of wastewater-derived fertilizer technologies with district-scale resolution. This index, with a range from 0 to 1 (highest suitability), incorporates key drivers, including population density, soil conditions, sanitation levels, and fertilizer prices. We found that suitability varies widely within and across countries in sub-Saharan Africa and that the primary limiting factor is the absence of sanitation infrastructure. Regions with a minimum of 10% cropland area and a suitability index of at least 0.9 were identified as highly suitable target regions for initial deployment. While they comprise only 1% of the analyzed area, these regions are home to 39 million people and contain up to 3.7 million hectares of cropland. Wastewater-derived fertilizer technologies could deliver an average of 25 kg of nitrogen per hectare of cropland, generating additional food equivalent to the annual consumption of 6 million people. Screening for high suitability can inform selection of effective lighthouse demonstration sites that derisk technology deployment and promote the transition to a more circular nutrient economy."
                    },
                    {
                        "title": "DiffusionSat: A Generative Foundation Model for Satellite Imagery",
                        "abstract": "Diffusion models have achieved state-of-the-art results on many modalities including images, speech, and video. However, existing models are not tailored to support remote sensing data, which is widely used in important applications including environmental monitoring and crop-yield prediction. Satellite images are significantly different from natural images -- they can be multi-spectral, irregularly sampled across time -- and existing diffusion models trained on images from the Web do not support them. Furthermore, remote sensing data is inherently spatio-temporal, requiring conditional generation tasks not supported by traditional methods based on captions or images. In this paper, we present DiffusionSat, to date the largest generative foundation model trained on a collection of publicly available large, high-resolution remote sensing datasets. As text-based captions are sparsely available for satellite images, we incorporate the associated metadata such as geolocation as conditioning information. Our method produces realistic samples and can be used to solve multiple generative tasks including temporal generation, superresolution given multi-spectral inputs and in-painting. Our method outperforms previous state-of-the-art methods for satellite image generation and is the first large-scale generative foundation model for satellite imagery. The project website can be found here: https://samar-khanna.github.io/DiffusionSat/"
                    },
                    {
                        "title": "Incorporating Climate Uncertainty into Estimates of Climate Change Impacts",
                        "abstract": "Abstract Quantitative estimates of the impacts of climate change on economic outcomes are important for public policy. We show that the vast majority of estimates fail to account for well-established uncertainty in future temperature and rainfall changes, leading to potentially misleading projections. We reexamine seven well-cited studies and show that accounting for climate uncertainty leads to a much larger range of projected climate impacts and a greater likelihood of worst-case outcomes, an important policy parameter. Incorporating climate uncertainty into future economic impact assessments will be critical for providing the best possible information on potential impacts."
                    }
                ]
            },
            "9c0193cc-2af3-4acd-bbc2-754fb3504f61": {
                "pk": "9c0193cc-2af3-4acd-bbc2-754fb3504f61",
                "name": "Stefano Ermon",
                "collaborators": [
                    "Minkai Xu",
                    "J. Leskovec",
                    "Chenlin Meng",
                    "Ling Yang",
                    "Bin Cui",
                    "Zhaochen Yu",
                    "K. Azizzadenesheli",
                    "Tailin Wu",
                    "Charles Marx",
                    "Hongxu Ma",
                    "Ni Lao",
                    "Gengchen Mai",
                    "Zhilong Zhang",
                    "Jiaqi Han",
                    "Aaron Lou",
                    "Jean Kossaifi",
                    "Arvind Ramanathan",
                    "A. Anandkumar",
                    "Zhuo Zheng",
                    "Yanfei Zhong",
                    "Liangpei Zhang",
                    "W. Neiswanger",
                    "Hongtao Zheng",
                    "Volodymyr Kuleshov",
                    "Nemin Wu",
                    "Qian Cao",
                    "Zhangyu Wang",
                    "Zeping Liu",
                    "Yanlin Qi",
                    "Jielu Zhang",
                    "Joshua Ni",
                    "Xiaobai Yao",
                    "Lan Mu",
                    "T. Ganu",
                    "A. Nambi",
                    "Hao Li",
                    "Yang Zou",
                    "Ying Wang",
                    "Orchid Majumder",
                    "Yusheng Xie",
                    "R. Manmatha",
                    "Ashwin Swaminathan",
                    "Zhuowen Tu",
                    "Stefano Soatto",
                    "Jingwei Liu",
                    "Zixiang Zhang",
                    "Xingchao Liu",
                    "Wentao Zhang",
                    "Siyi Gu",
                    "Alexander Powers",
                    "Weili Nie",
                    "Tomas Geffner",
                    "Karsten Kreis",
                    "Arash Vahdat",
                    "Rafael Rafailov",
                    "Archit Sharma",
                    "E. Mitchell",
                    "Christopher D. Manning",
                    "Chelsea Finn",
                    "Alexandre Lacoste",
                    "Nils Lehmann",
                    "Pau Rodr\u00edguez L\u00f3pez",
                    "Evan D. Sherwin",
                    "H. Kerner",
                    "Bjorn Lutjens",
                    "J. Irvin",
                    "David Dao",
                    "H. Alemohammad",
                    "Alexandre Drouin",
                    "Mehmet Gunturkun",
                    "Gabriel Huang",
                    "David V\u00e1zquez",
                    "Dava Newman",
                    "Y. Bengio",
                    "Xiao Xiang Zhu",
                    "Xuan Zhang",
                    "Limei Wang",
                    "Jacob Helwig",
                    "Youzhi Luo",
                    "Cong Fu",
                    "Yaochen Xie",
                    "Meng Liu",
                    "Yu-Ching Lin",
                    "Zhao Xu",
                    "Keqiang Yan",
                    "Keir Adams",
                    "Maurice Weiler",
                    "Xiner Li",
                    "Tianfan Fu",
                    "Yucheng Wang",
                    "Haiyang Yu",
                    "Yuqing Xie",
                    "Xiang Fu",
                    "A. Strasser",
                    "Shenglong Xu",
                    "Yi Liu",
                    "Yuanqi Du",
                    "Alexandra Saxton",
                    "Hongyi Ling",
                    "Hannah Lawrence"
                ],
                "domain": [
                    "Diffusion Models",
                    "Machine Learning",
                    "Natural Language Processing",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "Mastering Text-to-Image Diffusion: Recaptioning, Planning, and Generating with Multimodal LLMs",
                        "abstract": "Diffusion models have exhibit exceptional performance in text-to-image generation and editing. However, existing methods often face challenges when handling complex text prompts that involve multiple objects with multiple attributes and relationships. In this paper, we propose a brand new training-free text-to-image generation/editing framework, namely Recaption, Plan and Generate (RPG), harnessing the powerful chain-of-thought reasoning ability of multimodal LLMs to enhance the compositionality of text-to-image diffusion models. Our approach employs the MLLM as a global planner to decompose the process of generating complex images into multiple simpler generation tasks within subregions. We propose complementary regional diffusion to enable region-wise compositional generation. Furthermore, we integrate text-guided image generation and editing within the proposed RPG in a closed-loop fashion, thereby enhancing generalization ability. Extensive experiments demonstrate our RPG outperforms state-of-the-art text-to-image diffusion models, including DALL-E 3 and SDXL, particularly in multi-category object composition and text-image semantic alignment. Notably, our RPG framework exhibits wide compatibility with various MLLM architectures (e.g., MiniGPT-4) and diffusion backbones (e.g., ControlNet). Our code is available at: https://github.com/YangLing0818/RPG-DiffusionMaster"
                    },
                    {
                        "title": "Equivariant Graph Neural Operator for Modeling 3D Dynamics",
                        "abstract": "Modeling the complex three-dimensional (3D) dynamics of relational systems is an important problem in the natural sciences, with applications ranging from molecular simulations to particle mechanics. Machine learning methods have achieved good success by learning graph neural networks to model spatial interactions. However, these approaches do not faithfully capture temporal correlations since they only model next-step predictions. In this work, we propose Equivariant Graph Neural Operator (EGNO), a novel and principled method that directly models dynamics as trajectories instead of just next-step prediction. Different from existing methods, EGNO explicitly learns the temporal evolution of 3D dynamics where we formulate the dynamics as a function over time and learn neural operators to approximate it. To capture the temporal correlations while keeping the intrinsic SE(3)-equivariance, we develop equivariant temporal convolutions parameterized in the Fourier space and build EGNO by stacking the Fourier layers over equivariant networks. EGNO is the first operator learning framework that is capable of modeling solution dynamics functions over time while retaining 3D equivariance. Comprehensive experiments in multiple domains, including particle simulations, human motion capture, and molecular dynamics, demonstrate the significantly superior performance of EGNO against existing methods, thanks to the equivariant temporal modeling. Our code is available at https://github.com/MinkaiXu/egno."
                    },
                    {
                        "title": "Segment Any Change",
                        "abstract": "Visual foundation models have achieved remarkable results in zero-shot image classification and segmentation, but zero-shot change detection remains an open problem. In this paper, we propose the segment any change models (AnyChange), a new type of change detection model that supports zero-shot prediction and generalization on unseen change types and data distributions. AnyChange is built on the segment anything model (SAM) via our training-free adaptation method, bitemporal latent matching. By revealing and exploiting intra-image and inter-image semantic similarities in SAM's latent space, bitemporal latent matching endows SAM with zero-shot change detection capabilities in a training-free way. We also propose a point query mechanism to enable AnyChange's zero-shot object-centric change detection capability. We perform extensive experiments to confirm the effectiveness of AnyChange for zero-shot change detection. AnyChange sets a new record on the SECOND benchmark for unsupervised change detection, exceeding the previous SOTA by up to 4.4% F$_1$ score, and achieving comparable accuracy with negligible manual annotations (1 pixel per image) for supervised change detection."
                    },
                    {
                        "title": "Uncertainty Quantification for Forward and Inverse Problems of PDEs via Latent Global Evolution",
                        "abstract": "Deep learning-based surrogate models have demonstrated remarkable advantages over classical solvers in terms of speed, often achieving speedups of 10 to 1000 times over traditional partial differential equation (PDE) solvers. However, a significant challenge hindering their widespread adoption in both scientific and industrial domains is the lack of understanding about their prediction uncertainties, particularly in scenarios that involve critical decision making. To address this limitation, we propose a method that integrates efficient and precise uncertainty quantification into a deep learning-based surrogate model. Our method, termed Latent Evolution of PDEs with Uncertainty Quantification (LE-PDE-UQ), endows deep learning-based surrogate models with robust and efficient uncertainty quantification capabilities for both forward and inverse problems. LE-PDE-UQ leverages latent vectors within a latent space to evolve both the system's state and its corresponding uncertainty estimation. The latent vectors are decoded to provide predictions for the system's state as well as estimates of its uncertainty. In extensive experiments, we demonstrate the accurate uncertainty quantification performance of our approach, surpassing that of strong baselines including deep ensembles, Bayesian neural network layers, and dropout. Our method excels at propagating uncertainty over extended auto-regressive rollouts, making it suitable for scenarios involving long-term predictions. Our code is available at: https://github.com/AI4Science-WestlakeU/le-pde-uq."
                    },
                    {
                        "title": "Calibrated Probabilistic Forecasts for Arbitrary Sequences",
                        "abstract": "Real-world data streams can change unpredictably due to distribution shifts, feedback loops and adversarial actors, which challenges the validity of forecasts. We present a forecasting framework ensuring valid uncertainty estimates regardless of how data evolves. Leveraging the concept of Blackwell approachability from game theory, we introduce a forecasting framework that guarantees calibrated uncertainties for outcomes in any compact space (e.g., classification or bounded regression). We extend this framework to recalibrate existing forecasters, guaranteeing accurate uncertainties without sacrificing predictive performance. We implement both general-purpose gradient-based algorithms and algorithms optimized for popular special cases of our framework. Empirically, our algorithms improve calibration and downstream decision-making for energy systems."
                    },
                    {
                        "title": "TorchSpatial: A Location Encoding Framework and Benchmark for Spatial Representation Learning",
                        "abstract": "Spatial representation learning (SRL) aims at learning general-purpose neural network representations from various types of spatial data (e.g., points, polylines, polygons, networks, images, etc.) in their native formats. Learning good spatial representations is a fundamental problem for various downstream applications such as species distribution modeling, weather forecasting, trajectory generation, geographic question answering, etc. Even though SRL has become the foundation of almost all geospatial artificial intelligence (GeoAI) research, we have not yet seen significant efforts to develop an extensive deep learning framework and benchmark to support SRL model development and evaluation. To fill this gap, we propose TorchSpatial, a learning framework and benchmark for location (point) encoding, which is one of the most fundamental data types of spatial representation learning. TorchSpatial contains three key components: 1) a unified location encoding framework that consolidates 15 commonly recognized location encoders, ensuring scalability and reproducibility of the implementations; 2) the LocBench benchmark tasks encompassing 7 geo-aware image classification and 4 geo-aware image regression datasets; 3) a comprehensive suite of evaluation metrics to quantify geo-aware models' overall performance as well as their geographic bias, with a novel Geo-Bias Score metric. Finally, we provide a detailed analysis and insights into the model performance and geographic bias of different location encoders. We believe TorchSpatial will foster future advancement of spatial representation learning and spatial fairness in GeoAI research. The TorchSpatial model framework, LocBench, and Geo-Bias Score evaluation framework are available at https://github.com/seai-lab/TorchSpatial."
                    },
                    {
                        "title": "On the Scalability of Diffusion-based Text-to-Image Generation",
                        "abstract": "Scaling up model and data size has been quite successful for the evolution of LLMs. However, the scaling law for the diffusion based text-to-image (T2I) models is not fully explored. It is also unclear how to efficiently scale the model for better performance at reduced cost. The different training settings and expensive training cost make a fair model comparison extremely difficult. In this work, we empirically study the scaling properties of diffusion based T2I models by performing extensive and rigours ablations on scaling both denoising backbones and training set, including training scaled UNet and Transformer variants ranging from 0.4B to 4B parameters on datasets upto 600M images. For model scaling, we find the location and amount of cross attention distinguishes the performance of existing UNet designs. And increasing the transformer blocks is more parameter-efficient for improving text-image alignment than increasing channel numbers. We then identify an efficient UNet variant, which is 45% smaller and 28% faster than SDXL's UNet. On the data scaling side, we show the quality and diversity of the training set matters more than simply dataset size. Increasing caption density and diversity improves text-image alignment performance and the learning efficiency. Finally, we provide scaling functions to predict the text-image alignment performance as functions of the scale of model size, compute and dataset size."
                    },
                    {
                        "title": "Contextualized Diffusion Models for Text-Guided Image and Video Generation",
                        "abstract": "Conditional diffusion models have exhibited superior performance in high-fidelity text-guided visual generation and editing. Nevertheless, prevailing text-guided visual diffusion models primarily focus on incorporating text-visual relationships exclusively into the reverse process, often disregarding their relevance in the forward process. This inconsistency between forward and reverse processes may limit the precise conveyance of textual semantics in visual synthesis results. To address this issue, we propose a novel and general contextualized diffusion model (ContextDiff) by incorporating the cross-modal context encompassing interactions and alignments between text condition and visual sample into forward and reverse processes. We propagate this context to all timesteps in the two processes to adapt their trajectories, thereby facilitating cross-modal conditional modeling. We generalize our contextualized diffusion to both DDPMs and DDIMs with theoretical derivations, and demonstrate the effectiveness of our model in evaluations with two challenging tasks: text-to-image generation, and text-to-video editing. In each task, our ContextDiff achieves new state-of-the-art performance, significantly enhancing the semantic alignment between text condition and generated samples, as evidenced by quantitative and qualitative evaluations. Our code is available at https://github.com/YangLing0818/ContextDiff"
                    },
                    {
                        "title": "Consistency Flow Matching: Defining Straight Flows with Velocity Consistency",
                        "abstract": "Flow matching (FM) is a general framework for defining probability paths via Ordinary Differential Equations (ODEs) to transform between noise and data samples. Recent approaches attempt to straighten these flow trajectories to generate high-quality samples with fewer function evaluations, typically through iterative rectification methods or optimal transport solutions. In this paper, we introduce Consistency Flow Matching (Consistency-FM), a novel FM method that explicitly enforces self-consistency in the velocity field. Consistency-FM directly defines straight flows starting from different times to the same endpoint, imposing constraints on their velocity values. Additionally, we propose a multi-segment training approach for Consistency-FM to enhance expressiveness, achieving a better trade-off between sampling quality and speed. Preliminary experiments demonstrate that our Consistency-FM significantly improves training efficiency by converging 4.4x faster than consistency models and 1.7x faster than rectified flow models while achieving better generation quality. Our code is available at: https://github.com/YangLing0818/consistency_flow_matching"
                    },
                    {
                        "title": "Aligning Target-Aware Molecule Diffusion Models with Exact Energy Optimization",
                        "abstract": "Generating ligand molecules for specific protein targets, known as structure-based drug design, is a fundamental problem in therapeutics development and biological discovery. Recently, target-aware generative models, especially diffusion models, have shown great promise in modeling protein-ligand interactions and generating candidate drugs. However, existing models primarily focus on learning the chemical distribution of all drug candidates, which lacks effective steerability on the chemical quality of model generations. In this paper, we propose a novel and general alignment framework to align pretrained target diffusion models with preferred functional properties, named AliDiff. AliDiff shifts the target-conditioned chemical distribution towards regions with higher binding affinity and structural rationality, specified by user-defined reward functions, via the preference optimization approach. To avoid the overfitting problem in common preference optimization objectives, we further develop an improved Exact Energy Preference Optimization method to yield an exact and efficient alignment of the diffusion models, and provide the closed-form expression for the converged distribution. Empirical studies on the CrossDocked2020 benchmark show that AliDiff can generate molecules with state-of-the-art binding energies with up to -7.07 Avg. Vina Score, while maintaining strong molecular properties. Code is available at https://github.com/MinkaiXu/AliDiff."
                    },
                    {
                        "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                        "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                    },
                    {
                        "title": "GEO-Bench: Toward Foundation Models for Earth Monitoring",
                        "abstract": "Recent progress in self-supervision has shown that pre-training large neural networks on vast amounts of unsupervised data can lead to substantial increases in generalization to downstream tasks. Such models, recently coined foundation models, have been transformational to the field of natural language processing. Variants have also been proposed for image data, but their applicability to remote sensing tasks is limited. To stimulate the development of foundation models for Earth monitoring, we propose a benchmark comprised of six classification and six segmentation tasks, which were carefully curated and adapted to be both relevant to the field and well-suited for model evaluation. We accompany this benchmark with a robust methodology for evaluating models and reporting aggregated results to enable a reliable assessment of progress. Finally, we report results for 20 baselines to gain information about the performance of existing models. We believe that this benchmark will be a driver of progress across a variety of Earth monitoring tasks."
                    },
                    {
                        "title": "Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems",
                        "abstract": "Advances in artificial intelligence (AI) are fueling a new paradigm of discoveries in natural sciences. Today, AI has started to advance natural sciences by improving, accelerating, and enabling our understanding of natural phenomena at a wide range of spatial and temporal scales, giving rise to a new area of research known as AI for science (AI4Science). Being an emerging research paradigm, AI4Science is unique in that it is an enormous and highly interdisciplinary area. Thus, a unified and technical treatment of this field is needed yet challenging. This work aims to provide a technically thorough account of a subarea of AI4Science; namely, AI for quantum, atomistic, and continuum systems. These areas aim at understanding the physical world from the subatomic (wavefunctions and electron density), atomic (molecules, proteins, materials, and interactions), to macro (fluids, climate, and subsurface) scales and form an important subarea of AI4Science. A unique advantage of focusing on these areas is that they largely share a common set of challenges, thereby allowing a unified and foundational treatment. A key common challenge is how to capture physics first principles, especially symmetries, in natural systems by deep learning methods. We provide an in-depth yet intuitive account of techniques to achieve equivariance to symmetry transformations. We also discuss other common technical challenges, including explainability, out-of-distribution generalization, knowledge transfer with foundation and large language models, and uncertainty quantification. To facilitate learning and education, we provide categorized lists of resources that we found to be useful. We strive to be thorough and unified and hope this initial effort may trigger more community interests and efforts to further advance AI4Science."
                    },
                    {
                        "title": "Manifold Preserving Guided Diffusion",
                        "abstract": "Despite the recent advancements, conditional image generation still faces challenges of cost, generalizability, and the need for task-specific training. In this paper, we propose Manifold Preserving Guided Diffusion (MPGD), a training-free conditional generation framework that leverages pretrained diffusion models and off-the-shelf neural networks with minimal additional inference cost for a broad range of tasks. Specifically, we leverage the manifold hypothesis to refine the guided diffusion steps and introduce a shortcut algorithm in the process. We then propose two methods for on-manifold training-free guidance using pre-trained autoencoders and demonstrate that our shortcut inherently preserves the manifolds when applied to latent diffusion models. Our experiments show that MPGD is efficient and effective for solving a variety of conditional generation applications in low-compute settings, and can consistently offer up to 3.8x speed-ups with the same number of diffusion steps while maintaining high sample quality compared to the baselines."
                    },
                    {
                        "title": "But Are You Sure? An Uncertainty-Aware Perspective on Explainable AI",
                        "abstract": "Although black-box models can accurately predict outcomes such as weather patterns, they often lack transparency, making it challenging to extract meaningful insights (such as which atmospheric conditions signal future rainfall). Model explanations attempt to identify the essential features of a model, but these explanations can be inconsistent: two near-optimal models may admit vastly different explanations. In this paper, we propose a solution to this problem by constructing uncertainty sets for explanations of the optimal model(s) in both frequentist and Bayesian settings. Our uncertainty sets are guaranteed to include the explanation of the optimal model with high probability, even though this model is unknown. We demonstrate the effectiveness of our approach in both synthetic and real-world experiments, illustrating how our uncertainty sets can be used to calibrate trust in model explanations."
                    },
                    {
                        "title": "SSIF: Learning Continuous Image Representation for Spatial-Spectral Super-Resolution",
                        "abstract": "Existing digital sensors capture images at fixed spatial and spectral resolutions (e.g., RGB, multispectral, and hyperspectral images), and each combination requires bespoke machine learning models. Neural Implicit Functions partially overcome the spatial resolution challenge by representing an image in a resolution-independent way. However, they still operate at fixed, pre-defined spectral resolutions. To address this challenge, we propose Spatial-Spectral Implicit Function (SSIF), a neural implicit model that represents an image as a function of both continuous pixel coordinates in the spatial domain and continuous wavelengths in the spectral domain. We empirically demonstrate the effectiveness of SSIF on two challenging spatio-spectral super-resolution benchmarks. We observe that SSIF consistently outperforms state-of-the-art baselines even when the baselines are allowed to train separate models at each spectral resolution. We show that SSIF generalizes well to both unseen spatial resolutions and spectral resolutions. Moreover, SSIF can generate high-resolution images that improve the performance of downstream tasks (e.g., land use classification) by 1.7%-7%."
                    },
                    {
                        "title": "MERMAIDE: Learning to Align Learners using Model-Based Meta-Learning",
                        "abstract": "We study how a principal can efficiently and effectively intervene on the rewards of a previously unseen learning agent in order to induce desirable outcomes. This is relevant to many real-world settings like auctions or taxation, where the principal may not know the learning behavior nor the rewards of real people. Moreover, the principal should be few-shot adaptable and minimize the number of interventions, because interventions are often costly. We introduce MERMAIDE, a model-based meta-learning framework to train a principal that can quickly adapt to out-of-distribution agents with different learning strategies and reward functions. We validate this approach step-by-step. First, in a Stackelberg setting with a best-response agent, we show that meta-learning enables quick convergence to the theoretically known Stackelberg equilibrium at test time, although noisy observations severely increase the sample complexity. We then show that our model-based meta-learning approach is cost-effective in intervening on bandit agents with unseen explore-exploit strategies. Finally, we outperform baselines that use either meta-learning or agent behavior modeling, in both $0$-shot and $K=1$-shot settings with partial agent information."
                    },
                    {
                        "title": "MADiff: Offline Multi-agent Learning with Diffusion Models",
                        "abstract": "Diffusion model (DM) recently achieved huge success in various scenarios including offline reinforcement learning, where the diffusion planner learn to generate desired trajectories during online evaluations. However, despite the effectiveness in single-agent learning, it remains unclear how DMs can operate in multi-agent problems, where agents can hardly complete teamwork without good coordination by independently modeling each agent's trajectories. In this paper, we propose MADiff, a novel generative multi-agent learning framework to tackle this problem. MADiff is realized with an attention-based diffusion model to model the complex coordination among behaviors of multiple agents. To the best of our knowledge, MADiff is the first diffusion-based multi-agent learning framework, which behaves as both a decentralized policy and a centralized controller. During decentralized executions, MADiff simultaneously performs teammate modeling, and the centralized controller can also be applied in multi-agent trajectory predictions. Our experiments show the superior performance of MADiff compared to baseline algorithms in a wide range of multi-agent learning tasks, which emphasizes the effectiveness of MADiff in modeling complex multi-agent interactions. Our code is available at https://github.com/zbzhu99/madiff."
                    },
                    {
                        "title": "Holistic Evaluation of Text-To-Image Models",
                        "abstract": "The stunning qualitative improvement of recent text-to-image models has led to their widespread attention and adoption. However, we lack a comprehensive quantitative understanding of their capabilities and risks. To fill this gap, we introduce a new benchmark, Holistic Evaluation of Text-to-Image Models (HEIM). Whereas previous evaluations focus mostly on text-image alignment and image quality, we identify 12 aspects, including text-image alignment, image quality, aesthetics, originality, reasoning, knowledge, bias, toxicity, fairness, robustness, multilinguality, and efficiency. We curate 62 scenarios encompassing these aspects and evaluate 26 state-of-the-art text-to-image models on this benchmark. Our results reveal that no single model excels in all aspects, with different models demonstrating different strengths. We release the generated images and human evaluation results for full transparency at https://crfm.stanford.edu/heim/v1.1.0 and the code at https://github.com/stanford-crfm/helm, which is integrated with the HELM codebase."
                    },
                    {
                        "title": "Understanding and Adopting Rational Behavior by Bellman Score Estimation",
                        "abstract": "We are interested in solving a class of problems that seek to understand and adopt rational behavior from demonstrations. We may broadly classify these problems into four categories of reward identi\ufb01cation, counterfactual analysis, behavior imitation"
                    }
                ]
            }
        }
    },
    "2402.04376": {
        "paper_data": {
            "title": "Scaling laws for learning with real and surrogate data",
            "url": "http://arxiv.org/abs/2402.04376v2",
            "arxiv_id": "2402.04376",
            "authors": [
                "Ayush Jain",
                "Andrea Montanari",
                "Eren Sasoglu"
            ],
            "abstract": "Collecting large quantities of high-quality data can be prohibitively expensive or impractical, and a bottleneck in machine learning. One may instead augment a small set of $n$ data points from the target distribution with data from more accessible sources, e.g. data collected under different circumstances or synthesized by generative models. We refer to such data as `surrogate data.' We introduce a weighted empirical risk minimization (ERM) approach for integrating surrogate data into training. We analyze mathematically this method under several classical statistical models, and validate our findings empirically on datasets from different domains. Our main findings are: $(i)$ Integrating surrogate data can significantly reduce the test error on the original distribution. Surprisingly, this can happen even when the surrogate data is unrelated to the original ones. We trace back this behavior to the classical Stein's paradox. $(ii)$ In order to reap the benefit of surrogate data, it is crucial to use optimally weighted ERM. $(iii)$ The test error of models trained on mixtures of real and surrogate data is approximately described by a scaling law. This scaling law can be used to predict the optimal weighting scheme, and to choose the amount of surrogate data to add.",
            "introduction": "   1 Introduction and overview   1.1 Motivation and formulation  Consider a standard learning setting where we are given\u00a0n\ud835\udc5bnitalic_n i.i.d. points \ud835\udc9bisubscript\ud835\udc9b\ud835\udc56{\\boldsymbol{z}}_{i}bold_italic_z start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT from a target distribution \ud835\udc9f\ud835\udc9f\\mathcal{D}caligraphic_D. Given a family of rich parametric models governed by the parameter \ud835\udf3d\ud835\udf3d{\\boldsymbol{\\theta}}bold_italic_\u03b8, the goal is to find the parameter \ud835\udf3d\ud835\udf3d{\\boldsymbol{\\theta}}bold_italic_\u03b8 that minimizes the expected test loss Rtest\u2062(\ud835\udf3d)subscript\ud835\udc45test\ud835\udf3dR_{\\mbox{\\tiny\\rm test}}({\\boldsymbol{\\theta}})italic_R start_POSTSUBSCRIPT test end_POSTSUBSCRIPT ( bold_italic_\u03b8 ) between the model predictions for a chosen \ud835\udf3d\ud835\udf3d{\\boldsymbol{\\theta}}bold_italic_\u03b8 and the data generated from the distribution \ud835\udc9f\ud835\udc9f\\mathcal{D}caligraphic_D. In many application domains, the available data \ud835\udc81=(\ud835\udc9bi)i\u2264n\ud835\udc81subscriptsubscript\ud835\udc9b\ud835\udc56\ud835\udc56\ud835\udc5b{\\boldsymbol{Z}}=({\\boldsymbol{z}}_{i})_{i\\leq n}bold_italic_Z = ( bold_italic_z start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) start_POSTSUBSCRIPT italic_i \u2264 italic_n end_POSTSUBSCRIPT from the target distribution, referred to as either real or original data, may be difficult or expensive to acquire. One may then attempt to supplement these data with a different, cheaper source. Examples of such cheaper sources are (i)\ud835\udc56(i)( italic_i )\u00a0publicly available datasets; (i\u2062i)\ud835\udc56\ud835\udc56(ii)( italic_i italic_i )\u00a0datasets owned by the same research group or company but acquired in different circumstances, e.g.\u00a0in a different location; (i\u2062i\u2062i)\ud835\udc56\ud835\udc56\ud835\udc56(iii)( italic_i italic_i italic_i )\u00a0synthetic data produced by a generative model.   We will denote the data points obtained from this source by \ud835\udc9bissubscriptsuperscript\ud835\udc9b\ud835\udc60\ud835\udc56{\\boldsymbol{z}}^{s}_{i}bold_italic_z start_POSTSUPERSCRIPT italic_s end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT, and assume we have m\ud835\udc5amitalic_m of them. To formalize, assume \u2018surrogate\u2019 data \ud835\udc81s=(\ud835\udc9bis)i\u2264msuperscript\ud835\udc81\ud835\udc60subscriptsuperscriptsubscript\ud835\udc9b\ud835\udc56\ud835\udc60\ud835\udc56\ud835\udc5a{\\boldsymbol{Z}}^{s}=({\\boldsymbol{z}}_{i}^{s})_{i\\leq m}bold_italic_Z start_POSTSUPERSCRIPT italic_s end_POSTSUPERSCRIPT = ( bold_italic_z start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_s end_POSTSUPERSCRIPT ) start_POSTSUBSCRIPT italic_i \u2264 italic_m end_POSTSUBSCRIPT is a collection of i.i.d. samples from some distribution \ud835\udc9fssuperscript\ud835\udc9f\ud835\udc60\\mathcal{D}^{s}caligraphic_D start_POSTSUPERSCRIPT italic_s end_POSTSUPERSCRIPT. In general, we will not assume the distribution \ud835\udc9fssuperscript\ud835\udc9f\ud835\udc60\\mathcal{D}^{s}caligraphic_D start_POSTSUPERSCRIPT italic_s end_POSTSUPERSCRIPT of synthetic data to be close to the original data distribution \ud835\udc9f\ud835\udc9f\\mathcal{D}caligraphic_D, however we assume that these distributions are over the same domain. A number of questions arise: (i)\ud835\udc56(i)( italic_i )\u00a0How should we use the surrogate data in training? (i\u2062i)\ud835\udc56\ud835\udc56(ii)( italic_i italic_i )\u00a0How many surrogate samples should we add to the original data? (i\u2062i\u2062i)\ud835\udc56\ud835\udc56\ud835\udc56(iii)( italic_i italic_i italic_i )\u00a0Can we predict the improvement in test error achieved by adding surrogate samples to the training?   A natural approach would be to add the surrogate data to the original one in the usual training procedure, and indeed many authors have explored this approach (see Section 1.3). Namely, one attempts to minimize the overall empirical risk R^n+mnaive\u2062(\ud835\udf3d)=\u2211i=1n\u2113\u2062(\ud835\udf3d;\ud835\udc9bi)+\u2211i=1m\u2113\u2062(\ud835\udf3d;\ud835\udc9bis)superscriptsubscript^\ud835\udc45\ud835\udc5b\ud835\udc5anaive\ud835\udf3dsuperscriptsubscript\ud835\udc561\ud835\udc5b\u2113\ud835\udf3dsubscript\ud835\udc9b\ud835\udc56superscriptsubscript\ud835\udc561\ud835\udc5a\u2113\ud835\udf3dsubscriptsuperscript\ud835\udc9b\ud835\udc60\ud835\udc56\\widehat{R}_{n+m}^{\\mbox{\\tiny\\rm naive}}({\\boldsymbol{\\theta}})=\\sum_{i=1}^{n% }\\ell({\\boldsymbol{\\theta}};{\\boldsymbol{z}}_{i})+\\sum_{i=1}^{m}\\ell({% \\boldsymbol{\\theta}};{\\boldsymbol{z}}^{s}_{i})over^ start_ARG italic_R end_ARG start_POSTSUBSCRIPT italic_n + italic_m end_POSTSUBSCRIPT start_POSTSUPERSCRIPT naive end_POSTSUPERSCRIPT ( bold_italic_\u03b8 ) = \u2211 start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_n end_POSTSUPERSCRIPT roman_\u2113 ( bold_italic_\u03b8 ; bold_italic_z start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) + \u2211 start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_m end_POSTSUPERSCRIPT roman_\u2113 ( bold_italic_\u03b8 ; bold_italic_z start_POSTSUPERSCRIPT italic_s end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ), where \u2113\u2062(z,\u03b8)\u2113\ud835\udc67\ud835\udf03\\ell(z,\\theta)roman_\u2113 ( italic_z , italic_\u03b8 ) is a train loss function.   However, a moment of reflection reveals that this approach has serious shortcomings. Consider a simple mean estimation problem, whereby \ud835\udc9bi\u223c\ud835\uddad\u2062(\ud835\udf3d\u2217,\ud835\udc70d)similar-tosubscript\ud835\udc9b\ud835\udc56\ud835\uddadsubscript\ud835\udf3dsubscript\ud835\udc70\ud835\udc51{\\boldsymbol{z}}_{i}\\sim{\\sf N}({\\boldsymbol{\\theta}}_{*},{\\boldsymbol{I}}_{d})bold_italic_z start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT \u223c sansserif_N ( bold_italic_\u03b8 start_POSTSUBSCRIPT \u2217 end_POSTSUBSCRIPT , bold_italic_I start_POSTSUBSCRIPT italic_d end_POSTSUBSCRIPT ), \ud835\udc9bis\u223c\ud835\uddad\u2062(\ud835\udf3d\u2217s,\ud835\udc70d)similar-tosubscriptsuperscript\ud835\udc9b\ud835\udc60\ud835\udc56\ud835\uddadsubscriptsuperscript\ud835\udf3d\ud835\udc60subscript\ud835\udc70\ud835\udc51{\\boldsymbol{z}}^{s}_{i}\\sim{\\sf N}({\\boldsymbol{\\theta}}^{s}_{*},{\\boldsymbol% {I}}_{d})bold_italic_z start_POSTSUPERSCRIPT italic_s end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT \u223c sansserif_N ( bold_italic_\u03b8 start_POSTSUPERSCRIPT italic_s end_POSTSUPERSCRIPT start_POSTSUBSCRIPT \u2217 end_POSTSUBSCRIPT , bold_italic_I start_POSTSUBSCRIPT italic_d end_POSTSUBSCRIPT ), \u2113\u2062(\ud835\udf3d;\ud835\udc9b)=\u2016\ud835\udf3d\u2212\ud835\udc9b\u20162\u2113\ud835\udf3d\ud835\udc9bsuperscriptnorm\ud835\udf3d\ud835\udc9b2\\ell({\\boldsymbol{\\theta}};{\\boldsymbol{z}})=\\|{\\boldsymbol{\\theta}}-{% \\boldsymbol{z}}\\|^{2}roman_\u2113 ( bold_italic_\u03b8 ; bold_italic_z ) = \u2225 bold_italic_\u03b8 - bold_italic_z \u2225 start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT, and Rtest\u2062(\ud835\udf3d)=\u2016\ud835\udf3d\u2212\ud835\udf3d\u2217\u20162subscript\ud835\udc45test\ud835\udf3dsuperscriptnorm\ud835\udf3dsubscript\ud835\udf3d2R_{\\mbox{\\tiny\\rm test}}({{\\boldsymbol{\\theta}}})=\\|{\\boldsymbol{\\theta}}-{% \\boldsymbol{\\theta}}_{*}\\|^{2}italic_R start_POSTSUBSCRIPT test end_POSTSUBSCRIPT ( bold_italic_\u03b8 ) = \u2225 bold_italic_\u03b8 - bold_italic_\u03b8 start_POSTSUBSCRIPT \u2217 end_POSTSUBSCRIPT \u2225 start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT. A straightforward calculation yields that the",
            "references": [
                {
                    "title": "Scaling Laws of Synthetic Images for Model Training \u2026 for Now",
                    "abstract": "Recent significant advances in text-to-image models un-lock the possibility of training vision systems using synthetic images, potentially overcoming the difficulty of collecting curated data at scale. It is unclear, however, how these models behave at scale, as more synthetic data is added to the training set. In this paper we study the scaling laws of synthetic images generated by state of the art text-to-image models, for the training of supervised models: image classifiers with label supervision, and CLIP with language super-vision. We identify several factors, including text prompts, classifier-free guidance scale, and types of text-to-image models, that significantly affect scaling behavior. After tuning these factors, we observe that synthetic images demon-strate a scaling trend similar to, but slightly less effective than, real images in CLIP training, while they significantly underperform in scaling when training supervised image classifiers. Our analysis indicates that the main reason for this underperformance is the inability of off-the-shelf text-to-image models to generate certain concepts, a limitation that significantly impairs the training of image classifiers. Our findings also suggest that scaling synthetic data can be particularly effective in scenarios such as: (1) when there is a limited supply of real images for a supervised problem (e.g., fewer than 0.5 million images in ImageNet), (2) when the evaluation dataset diverges significantly from the training data, indicating the out-of-distribution scenario, or (3) when synthetic data is used in conjunction with real images, as demonstrated in the training of CLIP models."
                },
                {
                    "title": "Scaling Data-Constrained Language Models",
                    "abstract": "The current trend of scaling language models involves increasing both parameter count and training dataset size. Extrapolating this trend suggests that training dataset size may soon be limited by the amount of text data available on the internet. Motivated by this limit, we investigate scaling language models in data-constrained regimes. Specifically, we run a large set of experiments varying the extent of data repetition and compute budget, ranging up to 900 billion training tokens and 9 billion parameter models. We find that with constrained data for a fixed compute budget, training with up to 4 epochs of repeated data yields negligible changes to loss compared to having unique data. However, with more repetition, the value of adding compute eventually decays to zero. We propose and empirically validate a scaling law for compute optimality that accounts for the decreasing value of repeated tokens and excess parameters. Finally, we experiment with approaches mitigating data scarcity, including augmenting the training dataset with code data or removing commonly used filters. Models and datasets from our 400 training runs are freely available at https://github.com/huggingface/datablations."
                },
                {
                    "title": "Dimension free ridge regression",
                    "abstract": "Random matrix theory has become a widely useful tool in high-dimensional statistics and theoretical machine learning. However, random matrix theory is largely focused on the proportional asymptotics in which the number of columns grows proportionally to the number of rows of the data matrix. This is not always the most natural setting in statistics where columns correspond to covariates and rows to samples. With the objective to move beyond the proportional asymptotics, we revisit ridge regression ($\\ell_2$-penalized least squares) on i.i.d. data $(x_i, y_i)$, $i\\le n$, where $x_i$ is a feature vector and $y_i = \\beta^\\top x_i +\\epsilon_i \\in\\mathbb{R}$ is a response. We allow the feature vector to be high-dimensional, or even infinite-dimensional, in which case it belongs to a separable Hilbert space, and assume either $z_i := \\Sigma^{-1/2}x_i$ to have i.i.d. entries, or to satisfy a certain convex concentration property. Within this setting, we establish non-asymptotic bounds that approximate the bias and variance of ridge regression in terms of the bias and variance of an `equivalent' sequence model (a regression model with diagonal design matrix). The approximation is up to multiplicative factors bounded by $(1\\pm \\Delta)$ for some explicitly small $\\Delta$. Previously, such an approximation result was known only in the proportional regime and only up to additive errors: in particular, it did not allow to characterize the behavior of the excess risk when this converges to $0$. Our general theory recovers earlier results in the proportional regime (with better error rates). As a new application, we obtain a completely explicit and sharp characterization of ridge regression for Hilbert covariates with regularly varying spectrum. Finally, we analyze the overparametrized near-interpolation setting and obtain sharp `benign overfitting' guarantees."
                },
                {
                    "title": "Is synthetic data from generative models ready for image recognition?",
                    "abstract": "Recent text-to-image generation models have shown promising results in generating high-fidelity photo-realistic images. Though the results are astonishing to human eyes, how applicable these generated images are for recognition tasks remains under-explored. In this work, we extensively study whether and how synthetic images generated from state-of-the-art text-to-image generation models can be used for image recognition tasks, and focus on two perspectives: synthetic data for improving classification models in data-scarce settings (i.e. zero-shot and few-shot), and synthetic data for large-scale model pre-training for transfer learning. We showcase the powerfulness and shortcomings of synthetic data from existing generative models, and propose strategies for better applying synthetic data for recognition tasks. Code: https://github.com/CVMI-Lab/SyntheticData."
                },
                {
                    "title": "Revisiting Neural Scaling Laws in Language and Vision",
                    "abstract": "The remarkable progress in deep learning in recent years is largely driven by improvements in scale, where bigger models are trained on larger datasets for longer schedules. To predict the benefit of scale empirically, we argue for a more rigorous methodology based on the extrapolation loss, instead of reporting the best-fitting (interpolating) parameters. We then present a recipe for estimating scaling law parameters reliably from learning curves. We demonstrate that it extrapolates more accurately than previous methods in a wide range of architecture families across several domains, including image classification, neural machine translation (NMT) and language modeling, in addition to tasks from the BIG-Bench evaluation benchmark. Finally, we release a benchmark dataset comprising of 90 evaluation tasks to facilitate research in this domain."
                },
                {
                    "title": "Training Compute-Optimal Large Language Models",
                    "abstract": "We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over 400 language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, Chinchilla, that uses the same compute budget as Gopher but with 70B parameters and 4$\\times$ more more data. Chinchilla uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that Chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, Chinchilla reaches a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, greater than a 7% improvement over Gopher."
                },
                {
                    "title": "NeRF-Supervision: Learning Dense Object Descriptors from Neural Radiance Fields",
                    "abstract": "Thin, reflective objects such as forks and whisks are common in our daily lives, but they are particularly chal-lenging for robot perception because it is hard to reconstruct them using commodity RGB-D cameras or multi-view stereo techniques. While traditional pipelines struggle with objects like these, Neural Radiance Fields (NeRFs) have recently been shown to be remarkably effective for performing view synthesis on objects with thin structures or reflective materials. In this paper we explore the use of NeRF as a new source of supervision for robust robot vision systems. In particular, we demonstrate that a NeRF representation of a scene can be used to train dense object descriptors. We use an optimized NeRF to extract dense correspondences between multiple views of an object, and then use these correspondences as training data for learning a view-invariant representation of the object. NeRF's usage of a density field allows us to reformulate the correspondence problem with a novel distribution-of-depths formulation, as opposed to the conventional approach of using a depth map. Dense correspondence models supervised with our method significantly outperform off-the-shelf learned descriptors by 106% (PCK@3px metric, more than doubling performance) and outperform our baseline supervised with multi-view stereo by 29%. Furthermore, we demonstrate the learned dense descriptors enable robots to perform accurate 6-degree of freedom (6-DoF) pick and place of thin and reflective objects."
                },
                {
                    "title": "Generating Training Data with Language Models: Towards Zero-Shot Language Understanding",
                    "abstract": "Pretrained language models (PLMs) have demonstrated remarkable performance in various natural language processing tasks: Unidirectional PLMs (e.g., GPT) are well known for their superior text generation capabilities; bidirectional PLMs (e.g., BERT) have been the prominent choice for natural language understanding (NLU) tasks. While both types of models have achieved promising few-shot learning performance, their potential for zero-shot learning has been underexplored. In this paper, we present a simple approach that uses both types of PLMs for fully zero-shot learning of NLU tasks without requiring any task-specific data: A unidirectional PLM generates class-conditioned texts guided by prompts, which are used as the training data for fine-tuning a bidirectional PLM. With quality training data selected based on the generation probability and regularization techniques (label smoothing and temporal ensembling) applied to the fine-tuning stage for better generalization and stability, our approach demonstrates strong performance across seven classification tasks of the GLUE benchmark (e.g., 72.3/73.8 on MNLI-m/mm and 92.8 on SST-2), significantly outperforming zero-shot prompting methods and achieving even comparable results to strong few-shot approaches using 32 training samples per class."
                },
                {
                    "title": "LENS: Localization enhanced by NeRF synthesis",
                    "abstract": "Neural Radiance Fields (NeRF) have recently demonstrated photo-realistic results for the task of novel view synthesis. In this paper, we propose to apply novel view synthesis to the robot relocalization problem: we demonstrate improvement of camera pose regression thanks to an additional synthetic dataset rendered by the NeRF class of algorithm. To avoid spawning novel views in irrelevant places we selected virtual camera locations from NeRF internal representation of the 3D geometry of the scene. We further improved localization accuracy of pose regressors using synthesized realistic and geometry consistent images as data augmentation during training. At the time of publication, our approach improved state of the art with a 60% lower error on Cambridge Landmarks and 7-scenes datasets. Hence, the resulting accuracy becomes comparable to structure-based methods, without any architecture modification or domain adaptation constraints. Since our method allows almost infinite generation of training data, we investigated limitations of camera pose regression depending on size and distribution of data used for training on public benchmarks. We concluded that pose regression accuracy is mostly bounded by relatively small and biased datasets rather than capacity of the pose regression model to solve the localization task."
                },
                {
                    "title": "Generate, Annotate, and Learn: NLP with Synthetic Text",
                    "abstract": "Abstract This paper studies the use of language models as a source of synthetic unlabeled text for NLP. We formulate a general framework called \u201cgenerate, annotate, and learn (GAL)\u201d to take advantage of synthetic text within knowledge distillation, self-training, and few-shot learning applications. To generate high-quality task-specific text, we either fine-tune LMs on inputs from the task of interest, or prompt large LMs with few examples. We use the best available classifier to annotate synthetic text with soft pseudo labels for knowledge distillation and self-training, and use LMs to obtain hard labels for few-shot learning. We train new supervised models on the combination of labeled and pseudo-labeled data, which results in significant gains across several applications. We investigate key components of GAL and present theoretical and empirical arguments against the use of class-conditional LMs to generate synthetic labeled text instead of unlabeled text. GAL achieves new state-of-the-art knowledge distillation results for 6-layer transformers on the GLUE leaderboard."
                },
                {
                    "title": "Scaling Laws for Transfer",
                    "abstract": "We study empirical scaling laws for transfer learning between distributions in an unsupervised, fine-tuning setting. When we train increasingly large neural networks from-scratch on a fixed-size dataset, they eventually become data-limited and stop improving in performance (cross-entropy loss). When we do the same for models pre-trained on a large language dataset, the slope in performance gains is merely reduced rather than going to zero. We calculate the effective data\"transferred\"from pre-training by determining how much data a transformer of the same size would have required to achieve the same loss when training from scratch. In other words, we focus on units of data while holding everything else fixed. We find that the effective data transferred is described well in the low data regime by a power-law of parameter count and fine-tuning dataset size. We believe the exponents in these power-laws correspond to measures of the generality of a model and proximity of distributions (in a directed rather than symmetric sense). We find that pre-training effectively multiplies the fine-tuning dataset size. Transfer, like overall performance, scales predictably in terms of parameters, data, and compute."
                },
                {
                    "title": "Scaling Laws for Autoregressive Generative Modeling",
                    "abstract": "We identify empirical scaling laws for the cross-entropy loss in four domains: generative image modeling, video modeling, multimodal image$\\leftrightarrow$text models, and mathematical problem solving. In all cases autoregressive Transformers smoothly improve in performance as model size and compute budgets increase, following a power-law plus constant scaling law. The optimal model size also depends on the compute budget through a power-law, with exponents that are nearly universal across all data domains. \nThe cross-entropy loss has an information theoretic interpretation as $S($True$) + D_{\\mathrm{KL}}($True$||$Model$)$, and the empirical scaling laws suggest a prediction for both the true data distribution's entropy and the KL divergence between the true and model distributions. With this interpretation, billion-parameter Transformers are nearly perfect models of the YFCC100M image distribution downsampled to an $8\\times 8$ resolution, and we can forecast the model size needed to achieve any given reducible loss (ie $D_{\\mathrm{KL}}$) in nats/image for other resolutions. \nWe find a number of additional scaling laws in specific domains: (a) we identify a scaling relation for the mutual information between captions and images in multimodal models, and show how to answer the question \"Is a picture worth a thousand words?\"; (b) in the case of mathematical problem solving, we identify scaling laws for model performance when extrapolating beyond the training distribution; (c) we finetune generative image models for ImageNet classification and find smooth scaling of the classification loss and error rate, even as the generative loss levels off. Taken together, these results strengthen the case that scaling laws have important implications for neural network performance, including on downstream tasks."
                },
                {
                    "title": "High-Dimensional Probability: An Introduction with Applications in Data Science",
                    "abstract": "\u00a9 2018, Cambridge University Press Let us summarize our findings. A random projection of a set T in R n onto an m-dimensional subspace approximately preserves the geometry of T if m \u2a86 d ( T ) . For..."
                },
                {
                    "title": "Mixing Real and Synthetic Data to Enhance Neural Network Training - A Review of Current Approaches",
                    "abstract": "Deep neural networks have gained tremendous importance in many computer vision tasks. However, their power comes at the cost of large amounts of annotated data required for supervised training. In this work we review and compare different techniques available in the literature to improve training results without acquiring additional annotated real-world data. This goal is mostly achieved by applying annotation-preserving transformations to existing data or by synthetically creating more data."
                },
                {
                    "title": "On the Theory of Transfer Learning: The Importance of Task Diversity",
                    "abstract": "We provide new statistical guarantees for transfer learning via representation learning--when transfer is achieved by learning a feature representation shared across different tasks. This enables learning on new tasks using far less data than is required to learn them in isolation. Formally, we consider $t+1$ tasks parameterized by functions of the form $f_j \\circ h$ in a general function class $\\mathcal{F} \\circ \\mathcal{H}$, where each $f_j$ is a task-specific function in $\\mathcal{F}$ and $h$ is the shared representation in $\\mathcal{H}$. Letting $C(\\cdot)$ denote the complexity measure of the function class, we show that for diverse training tasks (1) the sample complexity needed to learn the shared representation across the first $t$ training tasks scales as $C(\\mathcal{H}) + t C(\\mathcal{F})$, despite no explicit access to a signal from the feature representation and (2) with an accurate estimate of the representation, the sample complexity needed to learn a new task scales only with $C(\\mathcal{F})$. Our results depend upon a new general notion of task diversity--applicable to models with general tasks, features, and losses--as well as a novel chain rule for Gaussian complexities. Finally, we exhibit the utility of our general framework in several models of importance in the literature."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "A Constructive Prediction of the Generalization Error Across Scales",
                    "abstract": "The dependency of the generalization error of neural networks on model and dataset size is of critical importance both in practice and for understanding the theory of neural networks. Nevertheless, the functional form of this dependency remains elusive. In this work, we present a functional form which approximates well the generalization error in practice. Capitalizing on the successful concept of model scaling (e.g., width, depth), we are able to simultaneously construct such a form and specify the exact models which can attain it across model/data scales. Our construction follows insights obtained from observations conducted over a range of model/data scales, in various model types and datasets, in vision and language tasks. We show that the form both fits the observations well across scales, and provides accurate predictions from small- to large-scale models and data."
                },
                {
                    "title": "Learning Semantic Segmentation From Synthetic Data: A Geometrically Guided Input-Output Adaptation Approach",
                    "abstract": "As an alternative to manual pixel-wise annotation, synthetic data has been increasingly used for training semantic segmentation models. Such synthetic images and semantic labels can be easily generated from virtual 3D environments. In this work, we propose an approach to cross-domain semantic segmentation with the auxiliary geometric information, which can also be easily obtained from virtual environments. The geometric information is utilized on two levels for reducing domain shift: on the input level, we augment the standard image translation network with the geometric information to translate synthetic images into realistic style; on the output level, we build a task network which simultaneously performs semantic segmentation and depth estimation. Meanwhile, adversarial training is applied on the joint output space to preserve the correlation between semantics and depth. The proposed approach is validated on two pairs of synthetic to real dataset: from Virtual KITTI to KITTI, and from SYNTHIA to Cityscapes, where we achieve a clear performance gain compared to the baselines and various competing methods, demonstrating the effectiveness of the geometric information for cross-domain semantic segmentation."
                },
                {
                    "title": "The distribution of the Lasso: Uniform control over sparse balls and adaptive parameter tuning",
                    "abstract": "The Lasso is a popular regression method for high-dimensional problems in which the number of parameters $\\theta_1,\\dots,\\theta_N$, is larger than the number $n$ of samples: $N>n$. A useful heuristics relates the statistical properties of the Lasso estimator to that of a simple soft-thresholding denoiser,in a denoising problem in which the parameters $(\\theta_i)_{i\\le N}$ are observed in Gaussian noise, with a carefully tuned variance. Earlier work confirmed this picture in the limit $n,N\\to\\infty$, pointwise in the parameters $\\theta$, and in the value of the regularization parameter. \nHere, we consider a standard random design model and prove exponential concentration of its empirical distribution around the prediction provided by the Gaussian denoising model. Crucially, our results are uniform with respect to $\\theta$ belonging to $\\ell_q$ balls, $q\\in [0,1]$, and with respect to the regularization parameter. This allows to derive sharp results for the performances of various data-driven procedures to tune the regularization. \nOur proofs make use of Gaussian comparison inequalities, and in particular of a version of Gordon's minimax theorem developed by Thrampoulidis, Oymak, and Hassibi, which controls the optimum value of the Lasso optimization problem. Crucially, we prove a stability property of the minimizer in Wasserstein distance, that allows to characterize properties of the minimizer itself."
                },
                {
                    "title": "Training Deep Networks with Synthetic Data: Bridging the Reality Gap by Domain Randomization",
                    "abstract": "We present a system for training deep neural networks for object detection using synthetic images. To handle the variability in real-world data, the system relies upon the technique of domain randomization, in which the parameters of the simulator-such as lighting, pose, object textures, etc.-are randomized in non-realistic ways to force the neural network to learn the essential features of the object of interest. We explore the importance of these parameters, showing that it is possible to produce a network with compelling performance using only non-artistically-generated synthetic data. With additional fine-tuning on real data, the network yields better performance than using real data alone. This result opens up the possibility of using inexpensive synthetic data for training neural networks while avoiding the need to collect large amounts of hand-annotated real-world data or to generate high-fidelity synthetic worlds-both of which remain bottlenecks for many applications. The approach is evaluated on bounding box detection of cars on the KITTI dataset."
                },
                {
                    "title": "Driving in the Matrix: Can virtual worlds replace human-generated annotations for real world tasks?",
                    "abstract": "Deep learning has rapidly transformed the state of the art algorithms used to address a variety of problems in computer vision and robotics. These breakthroughs have relied upon massive amounts of human annotated training data. This time consuming process has begun impeding the progress of these deep learning efforts. This paper describes a method to incorporate photo-realistic computer images from a simulation engine to rapidly generate annotated data that can be used for the training of machine learning algorithms. We demonstrate that a state of the art architecture, which is trained only using these synthetic annotations, performs better than the identical architecture trained on human annotated real-world data, when tested on the KITTI data set for vehicle detection. By training machine learning algorithms on a rich virtual world, real objects in real scenes can be learned and classified using synthetic data. This approach offers the possibility of accelerating deep learning's application to sensor-based classification problems like those that appear in self-driving cars. The source code and data to train and validate the networks described in this paper are made available for researchers."
                },
                {
                    "title": "The SYNTHIA Dataset: A Large Collection of Synthetic Images for Semantic Segmentation of Urban Scenes",
                    "abstract": "Vision-based semantic segmentation in urban scenarios is a key functionality for autonomous driving. Recent revolutionary results of deep convolutional neural networks (DCNNs) foreshadow the advent of reliable classifiers to perform such visual tasks. However, DCNNs require learning of many parameters from raw images, thus, having a sufficient amount of diverse images with class annotations is needed. These annotations are obtained via cumbersome, human labour which is particularly challenging for semantic segmentation since pixel-level annotations are required. In this paper, we propose to use a virtual world to automatically generate realistic synthetic images with pixel-level annotations. Then, we address the question of how useful such data can be for semantic segmentation - in particular, when using a DCNN paradigm. In order to answer this question we have generated a synthetic collection of diverse urban images, named SYNTHIA, with automatically generated class annotations. We use SYNTHIA in combination with publicly available real-world urban images with manually provided annotations. Then, we conduct experiments with DCNNs that show how the inclusion of SYNTHIA in the training stage significantly improves performance on the semantic segmentation task."
                },
                {
                    "title": "Precise Error Analysis of Regularized  $M$ -Estimators in High Dimensions",
                    "abstract": "A popular approach for estimating an unknown signal <inline-formula> <tex-math notation=\"LaTeX\">$ \\mathbf {x}_{0}\\in \\mathbb {R} ^{n}$ </tex-math></inline-formula> from noisy, linear measurements <inline-formula> <tex-math notation=\"LaTeX\">$ \\mathbf {y}= \\mathbf {A} \\mathbf {x} _{0}+ \\mathbf {z}\\in \\mathbb {R}^{m}$ </tex-math></inline-formula> is via solving a so called <italic>regularized</italic> <inline-formula> <tex-math notation=\"LaTeX\">$M$ </tex-math></inline-formula>-estimator: <inline-formula> <tex-math notation=\"LaTeX\">$\\hat{\\mathbf {x}} :=\\arg \\min _ \\mathbf {x} \\mathcal {L} (\\mathbf {y}- \\mathbf {A} \\mathbf {x})+\\lambda f(\\mathbf {x})$ </tex-math></inline-formula>. Here, <inline-formula> <tex-math notation=\"LaTeX\">$ \\mathcal {L}$ </tex-math></inline-formula> is a convex loss function, <inline-formula> <tex-math notation=\"LaTeX\">$f$ </tex-math></inline-formula> is a convex (typically, non-smooth) regularizer, and <inline-formula> <tex-math notation=\"LaTeX\">$\\lambda > 0$ </tex-math></inline-formula> is a regularizer parameter. We analyze the squared error performance <inline-formula> <tex-math notation=\"LaTeX\">$\\|\\hat{\\mathbf {x}} - \\mathbf {x}_{0}\\|_{2}^{2}$ </tex-math></inline-formula> of such estimators in the <italic>high-dimensional proportional regime</italic> where <inline-formula> <tex-math notation=\"LaTeX\">$m,n\\rightarrow \\infty $ </tex-math></inline-formula> and <inline-formula> <tex-math notation=\"LaTeX\">$m/n\\rightarrow \\delta $ </tex-math></inline-formula>. The design matrix <inline-formula> <tex-math notation=\"LaTeX\">$ \\mathbf {A}$ </tex-math></inline-formula> is assumed to have entries iid Gaussian; only minimal and rather mild regularity conditions are imposed on the loss function, the regularizer, and on the noise and signal distributions. We show that the squared error converges in probability to a nontrivial limit that is given as the solution to a minimax convex-concave optimization problem on four scalar optimization variables. We identify a new summary parameter, termed the <italic>expected Moreau envelope</italic> to play a central role in the error characterization. The <italic>precise</italic> nature of the results permits an accurate performance comparison between different instances of regularized <inline-formula> <tex-math notation=\"LaTeX\">$M$ </tex-math></inline-formula>-estimators and allows to optimally tune the involved parameters (such as the regularizer parameter and the number of measurements). The key ingredient of our proof is the <italic>convex Gaussian min-max theorem</italic> which is a tight and strengthened version of a classical Gaussian comparison inequality that was proved by Gordon in 1988."
                },
                {
                    "title": "Mathematical Foundations of Infinite-Dimensional Statistical Models",
                    "abstract": "1. Nonparametric statistical models 2. Gaussian processes 3. Empirical processes 4. Function spaces and approximation theory 5. Linear nonparametric estimators 6. The minimax paradigm 7. Likelihood-based procedures 8. Adaptive inference."
                },
                {
                    "title": "Regularized Linear Regression: A Precise Analysis of the Estimation Error",
                    "abstract": "Non-smooth regularized convex optimization procedures have emerged as a powerful tool to recover structured signals (sparse, low-rank, etc.) from (possibly compressed) noisy linear measurements. We focus on the problem of linear regression and consider a general class of optimization methods that minimize a loss function measuring the misfit of the model to the observations with an added structured-inducing regularization term. Celebrated instances include the LASSO, GroupLASSO, Least-Absolute Deviations method, etc.. We develop a quite general framework for how to determine precise prediction performance guaranties (e.g. mean-square-error) of such methods for the case of Gaussian measurement ensemble. The machinery builds upon Gordon\u2019s Gaussian min-max theorem under additional convexity assumptions that arise in many practical applications. This theorem associates with a primary optimization (PO) problem a simplified auxiliary optimization (AO) problem from which we can tightly infer properties of the original (PO), such as the optimal cost, the norm of the optimal solution, etc. Our theory applies to general loss functions and regularization and provides guidelines on how to optimally tune the regularizer coefficient when certain structural properties (such as sparsity level, rank, etc.) are known."
                },
                {
                    "title": "The Benefit of Multitask Representation Learning",
                    "abstract": "We discuss a general method to learn data representations from multiple tasks. We provide a justification for this method in both settings of multitask learning and learning-to-learn. The method is illustrated in detail in the special case of linear feature learning. Conditions on the theoretical advantage offered by multitask representation learning over independent task learning are established. In particular, focusing on the important example of half-space learning, we derive the regime in which multitask representation learning is beneficial over independent task learning, as a function of the sample size, the number of tasks and the intrinsic data dimensionality. Other potential applications of our results include multitask feature learning in reproducing kernel Hilbert spaces and multilayer, deep networks."
                },
                {
                    "title": "Asymptotic equivalence for nonparametric regression with multivariate and random design",
                    "abstract": "We show that nonparametric regression is asymptotically equivalent, in Le Cam's sense, to a sequence of Gaussian white noise experiments as the number of observations tends to infinity. We propose a general constructive framework, based on approximation spaces, which allows asymptotic equivalence to be achieved, even in the cases of multivariate and random design."
                },
                {
                    "title": "Asymptotic equivalence of nonparametric regression and white noise",
                    "abstract": "The principal result is that, under conditions, to any nonparametric regression problem there corresponds an asymptotically equivalent sequence of white noise with drift problems, and conversely. This asymptotic equivalence is in a global and uniform sense. Any normalized risk function attainable in one problem is asymptotically attainable in the other, with the difference in normalized risks converging to zero uniformly over the entire parameter space. The results are constructive. A recipe is provided for producing these asymptotically equivalent procedures. Some implications and generalizations of the principal result are also discussed."
                },
                {
                    "title": "NLTK: The Natural Language Toolkit",
                    "abstract": "The Natural Language Toolkit is a suite of program modules, data sets and tutorials supporting research and teaching in computational linguistics and natural language processing. NLTK is written in Python and distributed under the GPL open source license. Over the past year the toolkit has been rewritten, simplifying many linguistic data structures and taking advantage of recent enhancements in the Python language. This paper reports on the simpli-\ufb01ed toolkit and explains how it is used in teaching NLP."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively incorporate surrogate data from different distributions into the training process to improve model performance on a target distribution?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses the challenge of data scarcity in many real-world applications. By effectively utilizing surrogate data, researchers can enhance model training, leading to improved generalization and performance. This advancement could open new avenues for practical applications in fields such as healthcare, finance, and autonomous systems, where acquiring original data is often expensive or impractical. Furthermore, it could stimulate future research on data integration techniques and the development of more robust machine learning models.\n\n### [Question 3] - Why is it hard?\nThe complexity of this problem arises from the potential mismatch between the original and surrogate data distributions. Naive approaches, such as simply combining datasets, may lead to suboptimal model performance due to the introduction of noise or irrelevant information from the surrogate data. Additionally, determining the appropriate amount of surrogate data to include and predicting its impact on test error are non-trivial tasks. Technical challenges include ensuring that the surrogate data contributes positively to the learning process without degrading the model's ability to generalize to the target distribution.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often overlooked the nuances of integrating surrogate data, leading to a lack of effective methodologies. Many existing solutions fail to account for the differences in data distributions, resulting in ineffective training strategies. Barriers such as insufficient theoretical frameworks and a lack of empirical validation have hindered progress in this area. Our approach aims to fill these gaps by providing a structured methodology that explicitly addresses the integration of surrogate data while considering its distributional characteristics.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves a systematic framework for incorporating surrogate data into the training process. We will utilize a combination of theoretical analysis and empirical validation, employing a dataset that includes both original and surrogate samples. The performance will be evaluated using metrics such as test loss and generalization error. We expect that our approach will demonstrate a significant improvement in model performance compared to naive methods, providing insights into the optimal use of surrogate data in machine learning tasks."
            }
        },
        "author_data": {
            "95718dcf-905a-407b-877e-29717556e325": {
                "pk": "95718dcf-905a-407b-877e-29717556e325",
                "name": "Ayush Jain",
                "collaborators": [
                    "Alon Orlitsky",
                    "Ujjwal Guin",
                    "Vaishakh Ravindrakumar",
                    "Yi Hao",
                    "Ehsan Haghighat",
                    "Sai Nelaturi",
                    "Andrew Szot",
                    "Joseph J. Lim",
                    "N. M. Meenachi",
                    "B. Venkatraman"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Robust Learning",
                    "Natural Language Processing",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Detecting Causality with Symplectic Quandles",
                        "abstract": "We investigate the capability of Symplectic quandles to detect causality for (2+1)-dimensional globally hyperbolic spacetimes (X). Allen and Swenberg showed that Alexander-Conway polynomial is insufficient to distinguish connected sum of two Hopf links from the links in the family of Allen-Swenberg 2-sky like links, suggesting that it can not always detect causality in X. We find that symplectic quandles, combined with Alexander-Conway polynomial, can distinguish these two type of links, thereby suggesting their ability to detect causality in X. The fact that symplectic quandles can capture causality in the Allen-Swenberg example is intriguing since the theorem of Chernov and Nemirovski, which states that Legendrian linking equals causality, is proved using Contact Geometry methods."
                    },
                    {
                        "title": "A General Method for Robust Learning from Batches",
                        "abstract": "In many applications, data is collected in batches, some of which are corrupt or even adversarial. Recent work derived optimal robust algorithms for estimating discrete distributions in this setting. We consider a general framework of robust learning from batches, and determine the limits of both classification and distribution estimation over arbitrary, including continuous, domains. Building on these results, we derive the first robust agnostic computationally-efficient learning algorithms for piecewise-interval classification, and for piecewise-polynomial, monotone, log-concave, and gaussian-mixture distribution estimation."
                    },
                    {
                        "title": "Optimal Robust Learning of Discrete Distributions from Batches",
                        "abstract": "Many applications, including natural language processing, sensor networks, collaborative filtering, and federated learning, call for estimating discrete distributions from data collected in batches, some of which may be untrustworthy, erroneous, faulty, or even adversarial.   Previous estimators for this setting ran in exponential time, and for some regimes required a suboptimal number of batches. We provide the first polynomial-time estimator that is optimal in the number of batches and achieves essentially the best possible estimation accuracy."
                    },
                    {
                        "title": "Linear-Sample Learning of Low-Rank Distributions",
                        "abstract": "Many latent-variable applications, including community detection, collaborative filtering, genomic analysis, and NLP, model data as generated by low-rank matrices. Yet despite considerable research, except for very special cases, the number of samples required to efficiently recover the underlying matrices has not been known. We determine the onset of learning in several common latent-variable settings. For all of them, we show that learning $k\\times k$, rank-$r$, matrices to normalized $L_{1}$ distance $\\epsilon$ requires $\\Omega(\\frac{kr}{\\epsilon^2})$ samples, and propose an algorithm that uses ${\\cal O}(\\frac{kr}{\\epsilon^2}\\log^2\\frac r\\epsilon)$ samples, a number linear in the high dimension, and nearly linear in the, typically low, rank. The algorithm improves on existing spectral techniques and runs in polynomial time. The proofs establish new results on the rapid convergence of the spectral distance between the model and observation matrices, and may be of independent interest."
                    },
                    {
                        "title": "A Novel Tampering Attack on AES Cores with Hardware Trojans",
                        "abstract": "The implementation of cryptographic primitives in integrated circuits (ICs) continues to increase over the years due to the recent advancement of semiconductor manufacturing and reduction of cost per transistors. The hardware implementation makes cryptographic operations faster and more energy-efficient. However, various hardware attacks have been proposed aiming to extract the secret key in order to undermine the security of these primitives. In this paper, we focus on the widely used advanced encryption standard (AES) block cipher and demonstrate its vulnerability against tampering attack. Our proposed attack relies on implanting a hardware Trojan in the netlist by an untrusted foundry, which can design and implement such a Trojan as it has access to the design layout and mask information. The hardware Trojan's activation modifies a particular round's input data by preventing the effect of all previous rounds' key-dependent computation. We propose to use a sequential hardware Trojan to deliver the payload at the input of an internal round for achieving this modification of data. All the internal subkeys, and finally, the secret key can be computed from the observed ciphertext once the Trojan is activated. We implement our proposed tampering attack with a sequential hardware Trojan inserted into a 128-bit AES design from OpenCores benchmark suite and report the area overhead to demonstrate the feasibility of the proposed tampering attack."
                    },
                    {
                        "title": "LatticeGraphNet: A two-scale graph neural operator for simulating lattice structures",
                        "abstract": "This study introduces a two-scale Graph Neural Operator (GNO), namely, LatticeGraphNet (LGN), designed as a surrogate model for costly nonlinear finite-element simulations of three-dimensional latticed parts and structures. LGN has two networks: LGN-i, learning the reduced dynamics of lattices, and LGN-ii, learning the mapping from the reduced representation onto the tetrahedral mesh. LGN can predict deformation for arbitrary lattices, therefore the name operator. Our approach significantly reduces inference time while maintaining high accuracy for unseen simulations, establishing the use of GNOs as efficient surrogate models for evaluating mechanical responses of lattices and structures."
                    },
                    {
                        "title": "Generalization to New Actions in Reinforcement Learning",
                        "abstract": "A fundamental trait of intelligence is the ability to achieve goals in the face of novel circumstances, such as making decisions from new action choices. However, standard reinforcement learning assumes a fixed set of actions and requires expensive retraining when given a new action set. To make learning agents more adaptable, we introduce the problem of zero-shot generalization to new actions. We propose a two-stage framework where the agent first infers action representations from action information acquired separately from the task. A policy flexible to varying action sets is then trained with generalization objectives. We benchmark generalization on sequential tasks, such as selecting from an unseen tool-set to solve physical reasoning puzzles and stacking towers with novel 3D shapes. Videos and code are available at https://sites.google.com/view/action-generalization"
                    },
                    {
                        "title": "Robust estimation algorithms don't need to know the corruption level",
                        "abstract": "Real data are rarely pure. Hence the past half-century has seen great interest in robust estimation algorithms that perform well even when part of the data is corrupt. However, their vast majority approach optimal accuracy only when given a tight upper bound on the fraction of corrupt data. Such bounds are not available in practice, resulting in weak guarantees and often poor performance. This brief note abstracts the complex and pervasive robustness problem into a simple geometric puzzle. It then applies the puzzle's solution to derive a universal meta technique that converts any robust estimation algorithm requiring a tight corruption-level upper bound to achieve its optimal accuracy into one achieving essentially the same accuracy without using any upper bounds."
                    },
                    {
                        "title": "NukeBERT: A Pre-trained language model for Low Resource Nuclear Domain",
                        "abstract": "Significant advances have been made in recent years on Natural Language Processing with machines surpassing human performance in many tasks, including but not limited to Question Answering. The majority of deep learning methods for Question Answering targets domains with large datasets and highly matured literature. The area of Nuclear and Atomic energy has largely remained unexplored in exploiting non-annotated data for driving industry viable applications. Due to lack of dataset, a new dataset was created from the 7000 research papers on nuclear domain. This paper contributes to research in understanding nuclear domain knowledge which is then evaluated on Nuclear Question Answering Dataset (NQuAD) created by nuclear domain experts as part of this research. NQuAD contains 612 questions developed on 181 paragraphs randomly selected from the IGCAR research paper corpus. In this paper, the Nuclear Bidirectional Encoder Representational Transformers (NukeBERT) is proposed, which incorporates a novel technique for building BERT vocabulary to make it suitable for tasks with less training data. The experiments evaluated on NQuAD revealed that NukeBERT was able to outperform BERT significantly, thus validating the adopted methodology. Training NukeBERT is computationally expensive and hence we will be open-sourcing the NukeBERT pretrained weights and NQuAD for fostering further research work in the nuclear domain."
                    },
                    {
                        "title": "TAAL: Tampering Attack on Any Key-based Logic Locked Circuits",
                        "abstract": "Due to the globalization of semiconductor manufacturing and test processes, the system-on-a-chip (SoC) designers no longer design the complete SoC and manufacture chips on their own. This outsourcing of the design and manufacturing of Integrated Circuits (ICs) has resulted in several threats, such as overproduction of ICs, sale of out-of-specification/rejected ICs, and piracy of Intellectual Properties (IPs). Logic locking has emerged as a promising defense strategy against these threats. However, various attacks about the extraction of secret keys have undermined the security of logic locking techniques. Over the years, researchers have proposed different techniques to prevent existing attacks. In this paper, we propose a novel attack that can break any logic locking techniques that rely on the stored secret key. This proposed TAAL attack is based on implanting a hardware Trojan in the netlist, which leaks the secret key to an adversary once activated. As an untrusted foundry can extract the netlist of a design from the layout/mask information, it is feasible to implement such a hardware Trojan. All three proposed types of TAAL attacks can be used for extracting secret keys. We have introduced the models for both the combinational and sequential hardware Trojans that evade manufacturing tests. An adversary only needs to choose one hardware Trojan out of a large set of all possible Trojans to launch the TAAL attack."
                    },
                    {
                        "title": "ATPG-Guided Fault Injection Attacks on Logic Locking",
                        "abstract": "Logic Locking is a well-accepted protection technique to enable trust in the outsourced design and fabrication processes of integrated circuits (ICs) where the original design is modified by incorporating additional key gates in the netlist, resulting in a key-dependent functional circuit. The original functionality of the chip is recovered once it is programmed with the secret key, otherwise, it produces incorrect results for some input patterns. Over the past decade, different attacks have been proposed to break logic locking, simultaneously motivating researchers to develop more secure countermeasures. In this paper, we propose a novel stuck-at fault-based differential fault analysis (DFA) attack, which can be used to break logic locking that relies on a stored secret key. This proposed attack is based on self-referencing, where the secret key is determined by injecting faults in the key lines and comparing the response with its fault-free counterpart. A commercial ATPG tool can be used to generate test patterns that detect these faults, which will be used in DFA to determine the secret key. One test pattern is sufficient to determine one key bit, which results in at most |K| test patterns to determine the entire secret key of size |K|. The proposed attack is generic and can be extended to break any logic locked circuits."
                    },
                    {
                        "title": "Bottom Up Top Down Detection Transformers for Language Grounding in Images and Point Clouds",
                        "abstract": "Most models tasked to ground referential utterances in 2D and 3D scenes learn to select the referred object from a pool of object proposals provided by a pre-trained detector. This is limiting because an utterance may refer to visual entities at various levels of granularity, such as the chair, the leg of the chair, or the tip of the front leg of the chair, which may be missed by the detector. We propose a language grounding model that attends on the referential utterance and on the object proposal pool computed from a pre-trained detector to decode referenced objects with a detection head, without selecting them from the pool. In this way, it is helped by powerful pre-trained object detectors without being restricted by their misses. We call our model Bottom Up Top Down DEtection TRansformers (BUTD-DETR) because it uses both language guidance (top down) and objectness guidance (bottom-up) to ground referential utterances in images and point clouds. Moreover, BUTD-DETR casts object detection as referential grounding and uses object labels as language prompts to be grounded in the visual scene, augmenting supervision for the referential grounding task in this way. The proposed model sets a new state-of-the-art across popular 3D language grounding benchmarks with significant performance gains over previous 3D approaches (12.6% on SR3D, 11.6% on NR3D and 6.3% on ScanRefer). When applied in 2D images, it performs on par with the previous state of the art. We ablate the design choices of our model and quantify their contribution to performance. Our code and checkpoints can be found at the project website https://butd-detr.github.io."
                    },
                    {
                        "title": "SURF: A Simple, Universal, Robust, Fast Distribution Learning Algorithm",
                        "abstract": "Sample- and computationally-efficient distribution estimation is a fundamental tenet in statistics and machine learning. We present SURF, an algorithm for approximating distributions by piecewise polynomials. SURF is: simple, replacing prior complex optimization techniques by straight-forward {empirical probability} approximation of each potential polynomial piece {through simple empirical-probability interpolation}, and using plain divide-and-conquer to merge the pieces; universal, as well-known polynomial-approximation results imply that it accurately approximates a large class of common distributions; robust to distribution mis-specification as for any degree $d \\le 8$, it estimates any distribution to an $\\ell_1$ distance $< 3$ times that of the nearest degree-$d$ piecewise polynomial, improving known factor upper bounds of 3 for single polynomials and 15 for polynomials with arbitrarily many pieces; fast, using optimal sample complexity, running in near sample-linear time, and if given sorted samples it may be parallelized to run in sub-linear time. In experiments, SURF outperforms state-of-the art algorithms."
                    },
                    {
                        "title": "TURF: A Two-factor, Universal, Robust, Fast Distribution Learning Algorithm",
                        "abstract": "Approximating distributions from their samples is a canonical statistical-learning problem. One of its most powerful and successful modalities approximates every distribution to an $\\ell_1$ distance essentially at most a constant times larger than its closest $t$-piece degree-$d$ polynomial, where $t\\ge1$ and $d\\ge0$. Letting $c_{t,d}$ denote the smallest such factor, clearly $c_{1,0}=1$, and it can be shown that $c_{t,d}\\ge 2$ for all other $t$ and $d$. Yet current computationally efficient algorithms show only $c_{t,1}\\le 2.25$ and the bound rises quickly to $c_{t,d}\\le 3$ for $d\\ge 9$. We derive a near-linear-time and essentially sample-optimal estimator that establishes $c_{t,d}=2$ for all $(t,d)\\ne(1,0)$. Additionally, for many practical distributions, the lowest approximation distance is achieved by polynomials with vastly varying number of pieces. We provide a method that estimates this number near-optimally, hence helps approach the best possible approximation. Experiments combining the two techniques confirm improved performance over existing methodologies."
                    },
                    {
                        "title": "Variance Penalized On-Policy and Off-Policy Actor-Critic",
                        "abstract": "Reinforcement learning algorithms are typically geared towards optimizing the expected return of an agent. However, in many practical applications, low variance in the return is desired to ensure the reliability of an algorithm. In this paper, we propose on-policy and off-policy actor-critic algorithms that optimize a performance criterion involving both mean and variance in the return. Previous work uses the second moment of return to estimate the variance indirectly. Instead, we use a much simpler recently proposed direct variance estimator which updates the estimates incrementally using temporal difference methods. Using the variance-penalized criterion, we guarantee the convergence of our algorithm to locally optimal policies for finite state action Markov decision processes. We demonstrate the utility of our algorithm in tabular and continuous MuJoCo domains. Our approach not only performs on par with actor-critic and prior variance-penalization baselines in terms of expected return, but also generates trajectories which have lower variance in the return."
                    },
                    {
                        "title": "Deep learning via LSTM models for COVID-19 infection forecasting in India",
                        "abstract": "The COVID-19 pandemic continues to have major impact to health and medical infrastructure, economy, and agriculture. Prominent computational and mathematical models have been unreliable due to the complexity of the spread of infections. Moreover, lack of data collection and reporting makes modelling attempts difficult and unreliable. Hence, we need to re-look at the situation with reliable data sources and innovative forecasting models. Deep learning models such as recurrent neural networks are well suited for modelling spatiotemporal sequences. In this paper, we apply recurrent neural networks such as long short term memory (LSTM), bidirectional LSTM, and encoder-decoder LSTM models for multi-step (short-term) COVID-19 infection forecasting. We select Indian states with COVID-19 hotpots and capture the first (2020) and second (2021) wave of infections and provide two months ahead forecast. Our model predicts that the likelihood of another wave of infections in October and November 2021 is low; however, the authorities need to be vigilant given emerging variants of the virus. The accuracy of the predictions motivate the application of the method in other countries and regions. Nevertheless, the challenges in modelling remain due to the reliability of data and difficulties in capturing factors such as population density, logistics, and social aspects such as culture and lifestyle."
                    },
                    {
                        "title": "Deep Reinforcement Agent for Efficient Instant Search",
                        "abstract": "Instant Search is a paradigm where a search system retrieves answers on the fly while typing. The na\\\"ive implementation of an Instant Search system would hit the search back-end for results each time a user types a key, imposing a very high load on the underlying search system. In this paper, we propose to address the load issue by identifying tokens that are semantically more salient towards retrieving relevant documents and utilize this knowledge to trigger an instant search selectively. We train a reinforcement agent that interacts directly with the search engine and learns to predict the word's importance. Our proposed method treats the underlying search system as a black box and is more universally applicable to a diverse set of architectures. Furthermore, a novel evaluation framework is presented to study the trade-off between the number of triggered searches and the system's performance. We utilize the framework to evaluate and compare the proposed reinforcement method with other intuitive baselines. Experimental results demonstrate the efficacy of the proposed method towards achieving a superior trade-off."
                    }
                ]
            },
            "72265570-e3de-4f32-87a9-811dfc074fa2": {
                "pk": "72265570-e3de-4f32-87a9-811dfc074fa2",
                "name": "Andrea Montanari",
                "collaborators": [
                    "Sergio Caracciolo",
                    "Andrea Pelissetto",
                    "Riccardo Zecchina",
                    "Tommaso Rizzo"
                ],
                "domain": [
                    "Statistical Mechanics",
                    "Graph Theory",
                    "Error-Correcting Codes",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Two Lectures on Iterative Coding and Statistical Mechanics",
                        "abstract": "These are the notes for two lectures delivered at the Les Houches summer school Mathematical Statistical Mechanics, held in July 2005.   I review some basic notions on sparse graph error correcting codes with emphasis on `modern' aspects, such as, iterative belief propagation decoding. Relations with statistical mechanics, inference and random combinatorial optimization are stressed, as well as some general mathematical ideas and open problems."
                    },
                    {
                        "title": "Turbo codes: the phase transition",
                        "abstract": "Turbo codes are a very efficient method for communicating reliably through a noisy channel. There is no theoretical understanding of their effectiveness. In [1] they are mapped onto a class of disordered spin models. The analytical calculations concerning these models are reported here. We prove the existence of a no-error phase and compute its local stability threshold. As a byproduct, we gain some insight into the dynamics of the decoding algorithm."
                    },
                    {
                        "title": "Non-Perturbative Renormalization in Lattice Field Theory",
                        "abstract": "Non-perturbative renormalization of lattice composite operators plays a crucial role in many applications of lattice field theory. We sketch the general problems involved in this task and the methods which are currently used to cope with them. We present a detailed investigation of a new approach based on the operator product expansion. We test the new method on the two-dimensional O(3) sigma-model and discuss its advantages and limitations. Ph.D. thesis at Scuola Normale Superiore, Pisa, Italy."
                    },
                    {
                        "title": "Graphical Models Concepts in Compressed Sensing",
                        "abstract": "This paper surveys recent work in applying ideas from graphical models and message passing algorithms to solve large scale regularized regression problems. In particular, the focus is on compressed sensing reconstruction via ell_1 penalized least-squares (known as LASSO or BPDN). We discuss how to derive fast approximate message passing algorithms to solve this problem. Surprisingly, the analysis of such algorithms allows to prove exact high-dimensional limit results for the LASSO risk.   This paper will appear as a chapter in a book on `Compressed Sensing' edited by Yonina Eldar and Gitta Kutyniok."
                    },
                    {
                        "title": "Computational Implications of Reducing Data to Sufficient Statistics",
                        "abstract": "Given a large dataset and an estimation task, it is common to pre-process the data by reducing them to a set of sufficient statistics. This step is often regarded as straightforward and advantageous (in that it simplifies statistical analysis). I show that -on the contrary- reducing data to sufficient statistics can change a computationally tractable estimation problem into an intractable one. I discuss connections with recent work in theoretical computer science, and implications for some techniques to estimate graphical models."
                    },
                    {
                        "title": "The glassy phase of Gallager codes",
                        "abstract": "Gallager codes are the best error-correcting codes to-date. In this paper we study them by using the tools of statistical mechanics. The corresponding statistical mechanics model is a spin model on a sparse random graph. The model can be solved by elementary methods (i.e. without replicas) in a large connectivity limit. For low enough temperatures it presents a completely frozen glassy phase (q_{EA}=1). The same scenario is shown to hold for finite connectivities. In this case we adopt the replica approach and exhibit a one-step replica symmetry breaking order parameter. We argue that our ansatz yields the exact solution of the model. This allows us to determine the whole phase diagram and to understand the performances of Gallager codes."
                    },
                    {
                        "title": "Tight bounds for LDPC and LDGM codes under MAP decoding",
                        "abstract": "A new method for analyzing low density parity check (LDPC) codes and low density generator matrix (LDGM) codes under bit maximum a posteriori probability (MAP) decoding is introduced. The method is based on a rigorous approach to spin glasses developed by Francesco Guerra. It allows to construct lower bounds on the entropy of the transmitted message conditional to the received one. Based on heuristic statistical mechanics calculations, we conjecture such bounds to be tight. The result holds for standard irregular ensembles when used over binary input output symmetric channels. The method is first developed for Tanner graph ensembles with Poisson left degree distribution. It is then generalized to `multi-Poisson' graphs, and, by a completion procedure, to arbitrary degree distribution."
                    },
                    {
                        "title": "Estimating Random Variables from Random Sparse Observations",
                        "abstract": "Let X_1,...., X_n be a collection of iid discrete random variables, and Y_1,..., Y_m a set of noisy observations of such variables. Assume each observation Y_a to be a random function of some a random subset of the X_i's, and consider the conditional distribution of X_i given the observations, namely \\mu_i(x_i)\\equiv\\prob\\{X_i=x_i|Y\\} (a posteriori probability).   We establish a general relation between the distribution of \\mu_i, and the fixed points of the associated density evolution operator. Such relation holds asymptotically in the large system limit, provided the average number of variables an observation depends on is bounded. We discuss the relevance of our result to a number of applications, ranging from sparse graph codes, to multi-user detection, to group testing."
                    },
                    {
                        "title": "A Grothendieck-type inequality for local maxima",
                        "abstract": "A large number of problems in optimization, machine learning, signal processing can be effectively addressed by suitable semidefinite programming (SDP) relaxations. Unfortunately, generic SDP solvers hardly scale beyond instances with a few hundreds variables (in the underlying combinatorial problem). On the other hand, it has been observed empirically that an effective strategy amounts to introducing a (non-convex) rank constraint, and solving the resulting smooth optimization problem by ascent methods. This non-convex problem has --generically-- a large number of local maxima, and the reason for this success is therefore unclear.   This paper provides rigorous support for this approach. For the problem of maximizing a linear functional over the elliptope, we prove that all local maxima are within a small gap from the SDP optimum. In several problems of interest, arbitrarily small relative error can be achieved by taking the rank constraint $k$ to be of order one, independently of the problem size."
                    },
                    {
                        "title": "Optimization of the Sherrington-Kirkpatrick Hamiltonian",
                        "abstract": "Let ${\\boldsymbol A}\\in{\\mathbb R}^{n\\times n}$ be a symmetric random matrix with independent and identically distributed Gaussian entries above the diagonal. We consider the problem of maximizing $\\langle{\\boldsymbol \\sigma},{\\boldsymbol A}{\\boldsymbol \\sigma}\\rangle$ over binary vectors ${\\boldsymbol \\sigma}\\in\\{+1,-1\\}^n$. In the language of statistical physics, this amounts to finding the ground state of the Sherrington-Kirkpatrick model of spin glasses. The asymptotic value of this optimization problem was characterized by Parisi via a celebrated variational principle, subsequently proved by Talagrand. We give an algorithm that, for any $\\varepsilon>0$, outputs ${\\boldsymbol \\sigma}_*\\in\\{-1,+1\\}^n$ such that $\\langle{\\boldsymbol \\sigma}_*,{\\boldsymbol A}{\\boldsymbol \\sigma}_*\\rangle$ is at least $(1-\\varepsilon)$ of the optimum value, with probability converging to one as $n\\to\\infty$. The algorithm's time complexity is $C(\\varepsilon)\\, n^2$. It is a message-passing algorithm, but the specific structure of its update rules is new.   As a side result, we prove that, at (low) non-zero temperature, the algorithm constructs approximate solutions of the Thouless-Anderson-Palmer equations."
                    },
                    {
                        "title": "Sampling, Diffusions, and Stochastic Localization",
                        "abstract": "Diffusions are a successful technique to sample from high-dimensional distributions can be either explicitly given or learnt from a collection of samples. They implement a diffusion process whose endpoint is a sample from the target distribution and whose drift is typically represented as a neural network. Stochastic localization is a successful technique to prove mixing of Markov Chains and other functional inequalities in high dimension. An algorithmic version of stochastic localization was introduced in [EAMS2022], to obtain an algorithm that samples from certain statistical mechanics models.   This notes have three objectives: (i) Generalize the construction [EAMS2022] to other stochastic localization processes; (ii) Clarify the connection between diffusions and stochastic localization. In particular we show that standard denoising diffusions are stochastic localizations but other examples that are naturally suggested by the proposed viewpoint; (iii) Describe some insights that follow from this viewpoint."
                    },
                    {
                        "title": "Optimization of random cost functions and statistical physics",
                        "abstract": "This is the text of my report presented at the 29th Solvay Conference on Physics on `The Structure and Dynamics of Disordered Systems' held in Bruxelles from October 19 to 21, 2023. I consider the problem of minimizing a random energy function $H(\\sigma)$, where $\\sigma$ is an $N$-dimensional vector, in the high-dimensional regime $N\\gg 1$. Using as a reference point a 1986 paper by Fu and Anderson, I take stock of the progress on this question over the last 40 years. In particular, I focus on the influence and ramifications of ideas originating from statistical physics. My own conclusion is that several of the most fundamental questions in this area (which in 1986 were barely formulated) have now received mathematically rigorous answers, at least in simple -- yet highly nontrivial -- settings. Instrumental to this spectacular progress was the dialogue between different research communities: physics, computer science, mathematics."
                    },
                    {
                        "title": "Finding One Community in a Sparse Graph",
                        "abstract": "We consider a random sparse graph with bounded average degree, in which a subset of vertices has higher connectivity than the background. In particular, the average degree inside this subset of vertices is larger than outside (but still bounded). Given a realization of such graph, we aim at identifying the hidden subset of vertices. This can be regarded as a model for the problem of finding a tightly knitted community in a social network, or a cluster in a relational dataset.   In this paper we present two sets of contributions: $(i)$ We use the cavity method from spin glass theory to derive an exact phase diagram for the reconstruction problem. In particular, as the difference in edge probability increases, the problem undergoes two phase transitions, a static phase transition and a dynamic one. $(ii)$ We establish rigorous bounds on the dynamic phase transition and prove that, above a certain threshold, a local algorithm (belief propagation) correctly identify most of the hidden set. Below the same threshold \\emph{no local algorithm} can achieve this goal. However, in this regime the subset can be identified by exhaustive search.   For small hidden sets and large average degree, the phase transition for local algorithms takes an intriguingly simple form. Local algorithms succeed with high probability for ${\\rm deg}_{\\rm in} - {\\rm deg}_{\\rm out} > \\sqrt{{\\rm deg}_{\\rm out}/e}$ and fail for ${\\rm deg}_{\\rm in} - {\\rm deg}_{\\rm out} < \\sqrt{{\\rm deg}_{\\rm out}/e}$ (with ${\\rm deg}_{\\rm in}$, ${\\rm deg}_{\\rm out}$ the average degrees inside and outside the community). We argue that spectral algorithms are also ineffective in the latter regime.   It is an open problem whether any polynomial time algorithms might succeed for ${\\rm deg}_{\\rm in} - {\\rm deg}_{\\rm out} < \\sqrt{{\\rm deg}_{\\rm out}/e}$."
                    },
                    {
                        "title": "Composite operators from the operator product expansion: what can go wrong?",
                        "abstract": "The operator product expansion is used to compute the matrix elements of composite renormalized operators on the lattice. We study the product of two fundamental fields in the two-dimensional sigma-model and discuss the possible sources of systematic errors. The key problem turns out to be the violation of asymptotic scaling."
                    },
                    {
                        "title": "Operator Product Expansion on the Lattice: a Numerical Test in the Two-Dimensional Non-Linear Sigma-Model",
                        "abstract": "We consider the short-distance behaviour of the product of the Noether O(N) currents in the lattice nonlinear sigma-model. We compare the numerical results with the predictions of the operator product expansion, using one-loop perturbative renormalization-group improved Wilson coefficients. We find that, even on quite small lattices (m a \\approx 1/6), the perturbative operator product expansion describes that data with an error of 5-10% in a large window 2a \\ltapprox x \\ltapprox m^{-1}. We present a detailed discussion of the possible systematic errors."
                    },
                    {
                        "title": "Asymptotically free models and discrete non-Abelian groups",
                        "abstract": "We study the two-dimensional renormalization-group flow induced by perturbations that reduce the global symmetry of the O(3) sigma-model to the discrete symmetries of Platonic solids. We estimate the value of the correlation length at which differences in the behaviour of the various models should be expected. For the icosahedron model, we find xi > 200. We provide an explanation for the recent numerical results of Patrascioiu and Seiler and of Hasenfratz and Niedermayer."
                    },
                    {
                        "title": "Spin models on Platonic solids and asymptotic freedom",
                        "abstract": "We consider a two-dimensional sigma-model with discrete icosahedral/dodecahedral symmetry. We present high-precision finite-size numerical results that show that the continuum limit of this model is different from the continuum limit of the rotationally invariant O(3) sigma-model."
                    },
                    {
                        "title": "Discrete non-Abelian groups and asymptotically free models",
                        "abstract": "We consider a two-dimensional $\\sigma$-model with discrete icosahedral/dodecahedral symmetry. Using the perturbative renormalization group, we argue that this model has a different continuum limit with respect to the O(3) $\\sigma$ model. Such an argument is confirmed by a high-precision numerical simulation."
                    },
                    {
                        "title": "Boosting search by rare events",
                        "abstract": "Randomized search algorithms for hard combinatorial problems exhibit a large variability of performances. We study the different types of rare events which occur in such out-of-equilibrium stochastic processes and we show how they cooperate in determining the final distribution of running times. As a byproduct of our analysis we show how search algorithms are optimized by random restarts."
                    },
                    {
                        "title": "How to Compute Loop Corrections to Bethe Approximation",
                        "abstract": "We introduce a method for computing corrections to Bethe approximation for spin models on arbitrary lattices. Unlike cluster variational methods, the new approach takes into account fluctuations on all length scales.   The derivation of the leading correction is explained and applied to two simple examples: the ferromagnetic Ising model on d-dimensional lattices, and the spin glass on random graphs (both in their high-temperature phases). In the first case we rederive the well-known Ginzburg criterion and the upper critical dimension. In the second, we compute finite-size corrections to the free energy."
                    }
                ]
            },
            "b5521054-a9d9-4fc0-8875-84ea09dec461": {
                "pk": "b5521054-a9d9-4fc0-8875-84ea09dec461",
                "name": "Eren Sasoglu",
                "collaborators": [
                    "Lele Wang",
                    "Young-Han Kim",
                    "Satish Babu Korada",
                    "Emre Telatar",
                    "Bernd Bandemer",
                    "Ido Tal",
                    "Edmund Yeh",
                    "Erdal Arikan",
                    "Rudiger Urbanke",
                    "Fatemeh Arbabjolfaei"
                ],
                "domain": [
                    "Channel Coding",
                    "Polar Codes",
                    "Information Theory",
                    "Interference Channels"
                ],
                "publications": [
                    {
                        "title": "An entropy inequality for q-ary random variables and its application to channel polarization",
                        "abstract": "It is shown that given two copies of a q-ary input channel $W$, where q is prime, it is possible to create two channels $W^-$ and $W^+$ whose symmetric capacities satisfy $I(W^-)\\le I(W)\\le I(W^+)$, where the inequalities are strict except in trivial cases. This leads to a simple proof of channel polarization in the q-ary case."
                    },
                    {
                        "title": "A Class of Transformations that Polarize Symmetric Binary-Input Memoryless Channels",
                        "abstract": "A generalization of Ar\\i kan's polar code construction using transformations of the form $G^{\\otimes n}$ where $G$ is an $\\ell \\times \\ell$ matrix is considered. Necessary and sufficient conditions are given for these transformations to ensure channel polarization. It is shown that a large class of such transformations polarize symmetric binary-input memoryless channels."
                    },
                    {
                        "title": "Polar coding for interference networks",
                        "abstract": "A polar coding scheme for interference networks is introduced. The scheme combines Arikan's monotone chain rules for multiple-access channels and a method by Hassani and Urbanke to 'align' two incompatible polarization processes. It achieves the Han--Kobayashi inner bound for two-user interference channels and generalizes to interference networks."
                    },
                    {
                        "title": "Universal Polarization",
                        "abstract": "A method to polarize channels universally is introduced. The method is based on combining two distinct channels in each polarization step, as opposed to Arikan's original method of combining identical channels. This creates an equal number of only two types of channels, one of which becomes progressively better as the other becomes worse. The locations of the good polarized channels are independent of the underlying channel, guaranteeing universality. Polarizing the good channels further with Arikan's method results in universal polar codes of rate 1/2. The method is generalized to construct codes of arbitrary rates.   It is also shown that the less noisy ordering of channels is preserved under polarization, and thus a good polar code for a given channel will perform well over a less noisy one."
                    },
                    {
                        "title": "Polar Coding for Processes with Memory",
                        "abstract": "We study polar coding for stochastic processes with memory. For example, a process may be defined by the joint distribution of the input and output of a channel. The memory may be present in the channel, the input, or both. We show that $\\psi$-mixing processes polarize under the standard Ar\\i{}kan transform, under a mild condition. We further show that the rate of polarization of the \\emph{low-entropy} synthetic channels is roughly $O(2^{-\\sqrt{N}})$, where $N$ is the blocklength. That is, essentially the same rate as in the memoryless case."
                    },
                    {
                        "title": "Polar codes for the two-user multiple-access channel",
                        "abstract": "Arikan's polar coding method is extended to two-user multiple-access channels. It is shown that if the two users of the channel use the Arikan construction, the resulting channels will polarize to one of five possible extremals, on each of which uncoded transmission is optimal. The sum rate achieved by this coding technique is the one that correponds to uniform input distributions. The encoding and decoding complexities and the error performance of these codes are as in the single-user case: $O(n\\log n)$ for encoding and decoding, and $o(\\exp(-n^{1/2-\\epsilon}))$ for block error probability, where $n$ is the block length."
                    },
                    {
                        "title": "Polarization for arbitrary discrete memoryless channels",
                        "abstract": "Channel polarization, originally proposed for binary-input channels, is generalized to arbitrary discrete memoryless channels. Specifically, it is shown that when the input alphabet size is a prime number, a similar construction to that for the binary case leads to polarization. This method can be extended to channels of composite input alphabet sizes by decomposing such channels into a set of channels with prime input alphabet sizes. It is also shown that all discrete memoryless channels can be polarized by randomized constructions. The introduction of randomness does not change the order of complexity of polar code construction, encoding, and decoding. A previous result on the error probability behavior of polar codes is also extended to the case of arbitrary discrete memoryless channels. The generalization of polarization to channels with arbitrary finite input alphabet sizes leads to polar-coding methods for approaching the true (as opposed to symmetric) channel capacity of arbitrary channels with discrete or continuous input alphabets."
                    },
                    {
                        "title": "Polar Codes: Characterization of Exponent, Bounds, and Constructions",
                        "abstract": "Polar codes were recently introduced by Ar\\i kan. They achieve the capacity of arbitrary symmetric binary-input discrete memoryless channels under a low complexity successive cancellation decoding strategy. The original polar code construction is closely related to the recursive construction of Reed-Muller codes and is based on the $2 \\times 2$ matrix $\\bigl[ 1 &0 1& 1 \\bigr]$. It was shown by Ar\\i kan and Telatar that this construction achieves an error exponent of $\\frac12$, i.e., that for sufficiently large blocklengths the error probability decays exponentially in the square root of the length. It was already mentioned by Ar\\i kan that in principle larger matrices can be used to construct polar codes. A fundamental question then is to see whether there exist matrices with exponent exceeding $\\frac12$. We first show that any $\\ell \\times \\ell$ matrix none of whose column permutations is upper triangular polarizes symmetric channels. We then characterize the exponent of a given square matrix and derive upper and lower bounds on achievable exponents. Using these bounds we show that there are no matrices of size less than 15 with exponents exceeding $\\frac12$. Further, we give a general construction based on BCH codes which for large $n$ achieves exponents arbitrarily close to 1 and which exceeds $\\frac12$ for size 16."
                    },
                    {
                        "title": "A Comparison of Superposition Coding Schemes",
                        "abstract": "There are two variants of superposition coding schemes. Cover's original superposition coding scheme has code clouds of the identical shape, while Bergmans's superposition coding scheme has code clouds of independently generated shapes. These two schemes yield identical achievable rate regions in several scenarios, such as the capacity region for degraded broadcast channels. This paper shows that under the optimal maximum likelihood decoding, these two superposition coding schemes can result in different rate regions. In particular, it is shown that for the two-receiver broadcast channel, Cover's superposition coding scheme can achieve rates strictly larger than Bergmans's scheme."
                    },
                    {
                        "title": "On the Capacity Region for Index Coding",
                        "abstract": "A new inner bound on the capacity region of a general index coding problem is established. Unlike most existing bounds that are based on graph theoretic or algebraic tools, the bound is built on a random coding scheme and optimal decoding, and has a simple polymatroidal single-letter expression. The utility of the inner bound is demonstrated by examples that include the capacity region for all index coding problems with up to five messages (there are 9846 nonisomorphic ones)."
                    },
                    {
                        "title": "Sliding-Window Superposition Coding:Two-User Interference Channels",
                        "abstract": "A low-complexity coding scheme is developed to achieve the rate region of maximum likelihood decoding for interference channels. As in the classical rate-splitting multiple access scheme by Grant, Rimoldi, Urbanke, and Whiting, the proposed coding scheme uses superposition of multiple codewords with successive cancellation decoding, which can be implemented using standard point-to-point encoders and decoders. Unlike rate-splitting multiple access, which is not rate-optimal for multiple receivers, the proposed coding scheme transmits codewords over multiple blocks in a staggered manner and recovers them successively over sliding decoding windows, achieving the single-stream optimal rate region as well as the more general Han--Kobayashi inner bound for the two-user interference channel. The feasibility of this scheme in practice is verified by implementing it using commercial channel codes over the two-user Gaussian interference channel."
                    }
                ]
            }
        }
    },
    "2306.04793": {
        "paper_data": {
            "title": "On the Joint Interaction of Models, Data, and Features",
            "url": "http://arxiv.org/abs/2306.04793v1",
            "arxiv_id": "2306.04793",
            "authors": [
                "Yiding Jiang",
                "Christina Baek",
                "J. Zico Kolter"
            ],
            "abstract": "Learning features from data is one of the defining characteristics of deep learning, but our theoretical understanding of the role features play in deep learning is still rudimentary. To address this gap, we introduce a new tool, the interaction tensor, for empirically analyzing the interaction between data and model through features. With the interaction tensor, we make several key observations about how features are distributed in data and how models with different random seeds learn different features. Based on these observations, we propose a conceptual framework for feature learning. Under this framework, the expected accuracy for a single hypothesis and agreement for a pair of hypotheses can both be derived in closed-form. We demonstrate that the proposed framework can explain empirically observed phenomena, including the recently discovered Generalization Disagreement Equality (GDE) that allows for estimating the generalization error with only unlabeled data. Further, our theory also provides explicit construction of natural data distributions that break the GDE. Thus, we believe this work provides valuable new insight into our understanding of feature learning.",
            "introduction": " Introduction It is commonly said that deep learning performs feature learning, whereby the models extract useful patterns from the data and use the patterns to make predictions. Most successful applications of deep learning today involve first training the models on a large amount of data and then fine-tuning the pre- trained model on downstream tasks [ 13,9,53]. Their success suggests the models are learning useful and transferable knowledge from the data that allows them to solve similar tasks more efficiently. Experimentally, many different works [ 46,63,5,47,48,36] have studied various aspects of the features learned by deep neural networks. These works help the community gain a better intuitive understanding of the mechanisms underpinning deep learning as well as improve the interpretability of deep models. However, to the best of our knowledge, the theoretical understanding of the role features play in deep learning is still under-explored. For example, one prevailing framework for understanding deep learning is the neural tangent kernel (NTK) [ 28]. It is well-known that NTK is insufficient for understanding feature learning since the framework effectively analyzes deep learning as performing kernel regression with features defined by the gradient of the network at initialization , which is independent of any observed data. While it may be intuitive to think of defining features as quantifying the information in data that models use to make predictions, the community has yet to reach a consensus on the exact definition of features in deep learning beyond toy models. Nonetheless, it is undeniable that the models have learned something from the data. In fact, the same models trained with different random seeds would learn different information that leads to different predictions [ 35]. This phenomenon has important downstream consequences for ensembling randomly initialized networks including better generalization [ 1], calibration [ 35], and the Generalization Disagreement Equality (GDE) [ 29] where the expected test accuracy is equal to the expected agreement in deep ensembles. We postulate that a good definition of features should be fine-grained enough to discern the difference in the knowledge Preprint. Under review.arXiv:2306.04793v1  [cs.LG]  7 Jun 2023Figure 1: Visualization of images with the least features (left) and the most features (right) for classes of CIFAR 10 under our feature definition (defined in Section 3.1). Each row corresponds to one class of CIFAR 10 (zoom in for better viewing quality). We can see that the images with the least features are semantically similar to each other whereas the images with the most features are much more diverse and contain unusual instances of the class or objects in rare viewpoints. More examples for all classes can be found in Figure 11 of the appendix. of different models. To this end, we attempt to define such a construct that allows us to analyze the similarity and differences of information learned by different models, while also remaining amenable to quantitative and theoretical analysis. Through this definition, we can gain deeper insights into the behavior of models and the underlying mechanisms of feature learning. Notably, we show that GDE arises immediately as a consequence of how neural networks learn appropriately defined features. This phenomenon was previously explained by assuming calibration of the underlying ensemble, which is often a strong assumption to make [32]. We first begin with an empirical investigation of feature learning,",
            "references": [
                {
                    "title": "Toy Models of Superposition",
                    "abstract": "Neural networks often pack many unrelated concepts into a single neuron \u2013 a puzzling phenomenon known as 'polysemanticity' which makes interpretability much more challenging. This paper provides a toy model where polysemanticity can be fully understood, arising as a result of models storing additional sparse features in \"superposition.\" We demonstrate the existence of a phase change, a surprising connection to the geometry of uniform polytopes, and evidence of a link to adversarial examples. We also discuss potential implications for mechanistic interpretability. We recommend reading this paper as an HTML article. It would be very convenient if the individual neurons of artificial neural networks corresponded to cleanly interpretable features of the input. For example, in an \u201cideal\u201d ImageNet classifier, each neuron would fire only in the presence of a specific visual feature, such as the color red, a leftfacing curve, or a dog snout. Empirically, in models we have studied, some of the neurons do cleanly map to features. But it isn't always the case that features correspond so cleanly to neurons, especially in large language models where it actually seems rare for neurons to correspond to clean features. This brings up many questions. Why is it that neurons sometimes align with features and sometimes don't? Why do some models and tasks have many of these clean neurons, while they're vanishingly rare in others? In this paper, we use toy models \u2014 small ReLU networks trained on synthetic data with sparse input features \u2014 to investigate how and when models represent more features than they have dimensions. We call this phenomenon superposition. When features are sparse, superposition allows compression beyond what a linear model would do, at the cost of \"interference\" that requires nonlinear filtering. Consider a toy model where we train an embedding of five features of varying importance in two dimensions, add a ReLU afterwards for filtering, and vary the sparsity of the features. With dense features, the model learns to represent an orthogonal basis of the most important two features (similar to what Principal Component Analysis might give us), and the other three features are not represented. But if we make the features sparse, this changes: 1 This figure and a few others can be reproduced using the toy model framework Colab notebook in our Github repo Not only can models store additional features in superposition by tolerating some interference, but we'll show that, at least in certain limited cases, models can perform computation while in superposition. (In particular, we'll show that models can put simple circuits computing the absolute value function in superposition.) This leads us to hypothesize that the neural networks we observe in practice are in some sense noisily simulating larger, highly sparse networks. In other words, it's possible that models we train can be thought of as doing \u201cthe same thing as\u201d an imagined muchlarger model, representing the exact same features but with no interference. Feature superposition isn't a novel idea. A number of previous interpretability papers have speculated about it , and it's very closely related to the long-studied topic of compressed sensing in mathematics , as well as the ideas of distributed, dense, and population codes in neuroscience and deep learning . What, then, is the contribution of this paper? For interpretability researchers, our main contribution is providing a direct demonstration that superposition occurs in artificial neural networks given a relatively natural setup, suggesting this may also occur in practice. We offer a theory of when and why this occurs, revealing a phase diagram for superposition. We also discover that, at least in our toy model, superposition exhibits complex geometric structure. But our results may also be of broader interest. We find preliminary evidence that superposition may be linked to adversarial examples and grokking, and might also suggest a theory for the performance of mixture of experts models. More broadly, the toy model we investigate has unexpectedly rich structure, exhibiting phase changes, a geometric structure based on uniform polytopes, \"energy level\"-like jumps during training, and a phenomenon which is qualitatively similar to the fractional quantum Hall effect in physics. We originally investigated the subject to gain understanding of cleanly-interpretable neurons in larger models, but we've found these toy models to be surprisingly interesting in their own right. [1, 2] [3] [4] [5] KEY RESULTS FROM OUR TOY MODELS In our toy models, we are able to demonstrate that: Superposition is a real, observed phenomenon. Both monosemantic and polysemantic neurons can form. At least some kinds of computation can be performed in superposition. Whether features are stored in superposition is governed by a phase change. Superposition organizes features into geometric structures such as digons, triangles, pentagons, and tetrahedrons. Our toy models are simple ReLU networks, so it seems fair to say that neural networks exhibit these properties in at least some regimes, but it's very unclear what to generalize to real networks. Definitions and Motivation: Features, Directions, and Superposition In our work, we often think of neural networks as having features of the input represented as directions in activation space. This isn't a trivial claim. It isn't obvious what kind of structure we should expect neural network representations to have. When we say something like \"word embeddings have a gender direction\" or \"vision models have curve detector neurons\", one is implicitly making strong claims about the structure of network representations. Despite this, we believe this kind of \"linear representation hypothesis\" is supported both by significant empirical findings and theoretical arguments. One might think of this as two separate properties, which we'll explore in more detail shortly: Decomposability: Network representations can be described in terms of independently understandable features. Linearity: Features are represented by direction. If we hope to reverse engineer neural networks, we need a property like decomposability. Decomposability is what allows us to reason about the model without fitting the whole thing in our heads! But it's not enough for things to be decomposable: we need to be able to access the decomposition somehow. In order to do this, we need to identify the individual features within a representation. In a linear representation, this corresponds to determining which directions in activation space correspond to which independent features of the input. Sometimes, identifying feature directions is very easy because features seem to correspond to neurons. For example, many neurons in the early layers of InceptionV1 clearly correspond to features (e.g. curve detector neurons ). Why is it that we sometimes get this extremely helpful property, but in other cases don't? We hypothesize that there are really two countervailing forces driving this: [6] Privileged Basis: Only some representations have a privileged basis which encourages features to align with basis directions (i.e. to correspond to neurons). Superposition: Linear representations can represent more features than dimensions, using a strategy we call superposition. This can be seen as neural networks simulating larger networks. This pushes features away from corresponding to neurons. Superposition has been hypothesized in previous work  . However, we're not aware of feature superposition having been unambiguously demonstrated to occur in neural networks before ( demonstrates a closely related phenomenon of model superposition). The goal of this paper is to change that, demonstrating superposition and exploring how it interacts with privileged bases. If superposition occurs in networks, it deeply influences what approaches to interpretability research make sense, so unambiguous demonstration seems important. [1, 2] [7] The goal of this section will be to motivate these ideas and unpack them in detail. It's worth noting that many of the ideas in this section have close connections to ideas in other lines of interpretability research (especially disentanglement), neuroscience (distributed representations, population codes, etc), compressed sensing, and many other lines of work. This section will focus on articulating our perspective on the problem. We'll discuss these other lines of work in detail in Related Work."
                },
                {
                    "title": "Agreement-on-the-Line: Predicting the Performance of Neural Networks under Distribution Shift",
                    "abstract": "Recently, Miller et al. showed that a model's in-distribution (ID) accuracy has a strong linear correlation with its out-of-distribution (OOD) accuracy on several OOD benchmarks -- a phenomenon they dubbed ''accuracy-on-the-line''. While a useful tool for model selection (i.e., the model most likely to perform the best OOD is the one with highest ID accuracy), this fact does not help estimate the actual OOD performance of models without access to a labeled OOD validation set. In this paper, we show a similar but surprising phenomenon also holds for the agreement between pairs of neural network classifiers: whenever accuracy-on-the-line holds, we observe that the OOD agreement between the predictions of any two pairs of neural networks (with potentially different architectures) also observes a strong linear correlation with their ID agreement. Furthermore, we observe that the slope and bias of OOD vs ID agreement closely matches that of OOD vs ID accuracy. This phenomenon, which we call agreement-on-the-line, has important practical applications: without any labeled data, we can predict the OOD accuracy of classifiers}, since OOD agreement can be estimated with just unlabeled data. Our prediction algorithm outperforms previous methods both in shifts where agreement-on-the-line holds and, surprisingly, when accuracy is not on the line. This phenomenon also provides new insights into deep neural networks: unlike accuracy-on-the-line, agreement-on-the-line appears to only hold for neural network classifiers."
                },
                {
                    "title": "High-dimensional Asymptotics of Feature Learning: How One Gradient Step Improves the Representation",
                    "abstract": "We study the first gradient descent step on the first-layer parameters $\\boldsymbol{W}$ in a two-layer neural network: $f(\\boldsymbol{x}) = \\frac{1}{\\sqrt{N}}\\boldsymbol{a}^\\top\\sigma(\\boldsymbol{W}^\\top\\boldsymbol{x})$, where $\\boldsymbol{W}\\in\\mathbb{R}^{d\\times N}, \\boldsymbol{a}\\in\\mathbb{R}^{N}$ are randomly initialized, and the training objective is the empirical MSE loss: $\\frac{1}{n}\\sum_{i=1}^n (f(\\boldsymbol{x}_i)-y_i)^2$. In the proportional asymptotic limit where $n,d,N\\to\\infty$ at the same rate, and an idealized student-teacher setting, we show that the first gradient update contains a rank-1\"spike\", which results in an alignment between the first-layer weights and the linear component of the teacher model $f^*$. To characterize the impact of this alignment, we compute the prediction risk of ridge regression on the conjugate kernel after one gradient step on $\\boldsymbol{W}$ with learning rate $\\eta$, when $f^*$ is a single-index model. We consider two scalings of the first step learning rate $\\eta$. For small $\\eta$, we establish a Gaussian equivalence property for the trained feature map, and prove that the learned kernel improves upon the initial random features model, but cannot defeat the best linear model on the input. Whereas for sufficiently large $\\eta$, we prove that for certain $f^*$, the same ridge estimator on trained features can go beyond this\"linear regime\"and outperform a wide range of random features and rotationally invariant kernels. Our results demonstrate that even one gradient step can lead to a considerable advantage over random features, and highlight the role of learning rate scaling in the initial phase of training."
                },
                {
                    "title": "A Note on \"Assessing Generalization of SGD via Disagreement\"",
                    "abstract": "Several recent works find empirically that the average test error of deep neural networks can be estimated via the prediction disagreement of models, which does not require labels. In particular, Jiang et al. (2022) show for the disagreement between two separately trained networks that this `Generalization Disagreement Equality' follows from the well-calibrated nature of deep ensembles under the notion of a proposed `class-aggregated calibration.' In this reproduction, we show that the suggested theory might be impractical because a deep ensemble's calibration can deteriorate as prediction disagreement increases, which is precisely when the coupling of test error and disagreement is of interest, while labels are needed to estimate the calibration on new datasets. Further, we simplify the theoretical statements and proofs, showing them to be straightforward within a probabilistic context, unlike the original hypothesis space view employed by Jiang et al. (2022)."
                },
                {
                    "title": "Accuracy on the Line: on the Strong Correlation Between Out-of-Distribution and In-Distribution Generalization",
                    "abstract": "For machine learning systems to be reliable, we must understand their performance in unseen, out-of-distribution environments. In this paper, we empirically show that out-of-distribution performance is strongly correlated with in-distribution performance for a wide range of models and distribution shifts. Specifically, we demonstrate strong correlations between in-distribution and out-of-distribution performance on variants of CIFAR-10&ImageNet, a synthetic pose estimation task derived from YCB objects, satellite imagery classification in FMoW-WILDS, and wildlife classification in iWildCam-WILDS. The strong correlations hold across model architectures, hyperparameters, training set size, and training duration, and are more precise than what is expected from existing domain adaptation theory. To complete the picture, we also investigate cases where the correlation is weaker, for instance some synthetic distribution shifts from CIFAR-10-C and the tissue classification dataset Camelyon17-WILDS. Finally, we provide a candidate theory based on a Gaussian data model that shows how changes in the data covariance arising from distribution shift can affect the observed correlations."
                },
                {
                    "title": "Assessing Generalization of SGD via Disagreement",
                    "abstract": "We empirically show that the test error of deep networks can be estimated by simply training the same architecture on the same training set but with a different run of Stochastic Gradient Descent (SGD), and measuring the disagreement rate between the two networks on unlabeled test data. This builds on -- and is a stronger version of -- the observation in Nakkiran&Bansal '20, which requires the second run to be on an altogether fresh training set. We further theoretically show that this peculiar phenomenon arises from the \\emph{well-calibrated} nature of \\emph{ensembles} of SGD-trained models. This finding not only provides a simple empirical measure to directly predict the test error using unlabeled test data, but also establishes a new conceptual connection between generalization and calibration."
                },
                {
                    "title": "Toward Understanding the Feature Learning Process of Self-supervised Contrastive Learning",
                    "abstract": "How can neural networks trained by contrastive learning extract features from the unlabeled data? Why does contrastive learning usually need much stronger data augmentations than supervised learning to ensure good representations? These questions involve both the optimization and statistical aspects of deep learning, but can hardly be answered by analyzing supervised learning, where the target functions are the highest pursuit. Indeed, in self-supervised learning, it is inevitable to relate to the optimization/generalization of neural networks to how they can encode the latent structures in the data, which we refer to as the feature learning process. In this work, we formally study how contrastive learning learns the feature representations for neural networks by analyzing its feature learning process. We consider the case where our data are comprised of two types of features: the more semantically aligned sparse features which we want to learn from, and the other dense features we want to avoid. Theoretically, we prove that contrastive learning using $\\mathbf{ReLU}$ networks provably learns the desired sparse features if proper augmentations are adopted. We present an underlying principle called $\\textbf{feature decoupling}$ to explain the effects of augmentations, where we theoretically characterize how augmentations can reduce the correlations of dense features between positive samples while keeping the correlations of sparse features intact, thereby forcing the neural networks to learn from the self-supervision of sparse features. Empirically, we verified that the feature decoupling principle matches the underlying mechanism of contrastive learning in practice."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "RegNet: Self-Regulated Network for Image Classification",
                    "abstract": "The ResNet and its variants have achieved remarkable successes in various computer vision tasks. Despite its success in making gradient flow through building blocks, the information communication of intermediate layers of blocks is ignored. To address this issue, in this brief, we propose to introduce a regulator module as a memory mechanism to extract complementary features of the intermediate layers, which are further fed to the ResNet. In particular, the regulator module is composed of convolutional recurrent neural networks (RNNs) [e.g., convolutional long short-term memories (LSTMs) or convolutional gated recurrent units (GRUs)], which are shown to be good at extracting spatio-temporal information. We named the new regulated network as regulated residual network (RegNet). The regulator module can be easily implemented and appended to any ResNet architecture. Experimental results on three image classification datasets have demonstrated the promising performance of the proposed architecture compared with the standard ResNet, squeeze-and-excitation ResNet, and other state-of-the-art architectures."
                },
                {
                    "title": "Towards Understanding Ensemble, Knowledge Distillation and Self-Distillation in Deep Learning",
                    "abstract": "We formally study how Ensemble of deep learning models can improve test accuracy, and how the superior performance of ensemble can be distilled into a single model using Knowledge Distillation. We consider the challenging case where the ensemble is simply an average of the outputs of a few independently trained neural networks with the SAME architecture, trained using the SAME algorithm on the SAME data set, and they only differ by the random seeds used in the initialization. We empirically show that ensemble/knowledge distillation in deep learning works very differently from traditional learning theory, especially differently from ensemble of random feature mappings or the neural-tangent-kernel feature mappings, and is potentially out of the scope of existing theorems. Thus, to properly understand ensemble and knowledge distillation in deep learning, we develop a theory showing that when data has a structure we refer to as \"multi-view\", then ensemble of independently trained neural networks can provably improve test accuracy, and such superior test accuracy can also be provably distilled into a single model by training a single model to match the output of the ensemble instead of the true label. Our result sheds light on how ensemble works in deep learning in a way that is completely different from traditional theorems, and how the \"dark knowledge\" is hidden in the outputs of the ensemble -- that can be used in knowledge distillation -- comparing to the true data labels. In the end, we prove that self-distillation can also be viewed as implicitly combining ensemble and knowledge distillation to improve test accuracy."
                },
                {
                    "title": "Distributional Generalization: A New Kind of Generalization",
                    "abstract": "We introduce a new notion of generalization -- Distributional Generalization -- which roughly states that outputs of a classifier at train and test time are close *as distributions*, as opposed to close in just their average error. For example, if we mislabel 30% of dogs as cats in the train set of CIFAR-10, then a ResNet trained to interpolation will in fact mislabel roughly 30% of dogs as cats on the *test set* as well, while leaving other classes unaffected. This behavior is not captured by classical generalization, which would only consider the average error and not the distribution of errors over the input domain. Our formal conjectures, which are much more general than this example, characterize the form of distributional generalization that can be expected in terms of problem parameters: model architecture, training procedure, number of samples, and data distribution. We give empirical evidence for these conjectures across a variety of domains in machine learning, including neural networks, kernel machines, and decision trees. Our results thus advance our empirical understanding of interpolating classifiers."
                },
                {
                    "title": "Prevalence of neural collapse during the terminal phase of deep learning training",
                    "abstract": "Significance Modern deep neural networks for image classification have achieved superhuman performance. Yet, the complex details of trained networks have forced most practitioners and researchers to regard them as black boxes with little that could be understood. This paper considers in detail a now-standard training methodology: driving the cross-entropy loss to zero, continuing long after the classification error is already zero. Applying this methodology to an authoritative collection of standard deepnets and datasets, we observe the emergence of a simple and highly symmetric geometry of the deepnet features and of the deepnet classifier, and we document important benefits that the geometry conveys\u2014thereby helping us understand an important component of the modern deep learning training paradigm. Modern practice for training classification deepnets involves a terminal phase of training (TPT), which begins at the epoch where training error first vanishes. During TPT, the training error stays effectively zero, while training loss is pushed toward zero. Direct measurements of TPT, for three prototypical deepnet architectures and across seven canonical classification datasets, expose a pervasive inductive bias we call neural collapse (NC), involving four deeply interconnected phenomena. (NC1) Cross-example within-class variability of last-layer training activations collapses to zero, as the individual activations themselves collapse to their class means. (NC2) The class means collapse to the vertices of a simplex equiangular tight frame (ETF). (NC3) Up to rescaling, the last-layer classifiers collapse to the class means or in other words, to the simplex ETF (i.e., to a self-dual configuration). (NC4) For a given activation, the classifier\u2019s decision collapses to simply choosing whichever class has the closest train class mean (i.e., the nearest class center [NCC] decision rule). The symmetric and very simple geometry induced by the TPT confers important benefits, including better generalization performance, better robustness, and better interpretability."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Feature Purification: How Adversarial Training Performs Robust Deep Learning",
                    "abstract": "Despite the empirical success of using adversarial training to defend deep learning models against adversarial perturbations, so far, it still remains rather unclear what the principles are behind the existence of adversarial perturbations, and what adversarial training does to the neural network to remove them. In this paper, we present a principle that we call feature purification, where we show one of the causes of the existence of adversarial examples is the accumulation of certain small dense mixtures in the hidden weights during the training process of a neural network; and more importantly, one of the goals of adversarial training is to remove such mixtures to purify hidden weights. We present both experiments on the CIFAR-10 dataset to illustrate this principle, and a theoretical result proving that for certain natural classification tasks, training a two-layer neural network with ReLU activation using randomly initialized gradient descent indeed satisfies this principle. Technically, we give, to the best of our knowledge, the first re-sult proving that the following two can hold simultaneously for training a neural network with ReLU activation. (1) Training over the original data is indeed non-robust to small adversarial perturbations of some radius. (2) Adversarial training, even with an empirical perturbation algorithm such as FGM, can in fact be provably robust against any perturbations of the same radius. Finally, we also prove a complexity lower bound, showing that low complexity models such as linear classifiers, low-degree polynomials, or even the neural tangent kernel for this network, cannot defend against perturbations of this same radius, no matter what algorithms are used to train them. 11The full version of this paper can be found at https://arxiv.org/abs/-2005.l0190."
                },
                {
                    "title": "GANSpace: Discovering Interpretable GAN Controls",
                    "abstract": "This paper describes a simple technique to analyze Generative Adversarial Networks (GANs) and create interpretable controls for image synthesis, such as change of viewpoint, aging, lighting, and time of day. We identify important latent directions based on Principal Components Analysis (PCA) applied either in latent space or feature space. Then, we show that a large number of interpretable controls can be defined by layer-wise perturbation along the principal directions. Moreover, we show that BigGAN can be controlled with layer-wise inputs in a StyleGAN-like manner. We show results on different GANs trained on various datasets, and demonstrate good qualitative matches to edit directions found through earlier supervised approaches."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "Characterizing Structural Regularities of Labeled Data in Overparameterized Models",
                    "abstract": "Human learners appreciate that observations usually form hierarchies of regularities and sub-regularities. For example, English verbs have irregular cases that must be memorized (e.g., go -> went) and regular cases that generalize well (e.g., kiss -> kissed, miss -> missed). Likewise, deep neural networks have the capacity to memorize rare or irregular forms but nonetheless generalize across instances that share common patterns or structures. We analyze how individual instances are treated by a model via a consistency score. The score is the expected accuracy of a particular architecture for a held-out instance on a training set of a given size sampled from the data distribution. We obtain empirical estimates of this score for individual instances in multiple data sets, and we show that the score identifies out-of-distribution and mislabeled examples at one end of the continuum and regular examples at the other end. We explore two categories of proxies to the consistency score: pairwise distance based proxy and the training statistics based proxies. We conclude with two applications using C-scores to help understand the dynamics of representation learning and filter out outliers, and discussions of other potential applications such as curriculum learning, and active data collection."
                },
                {
                    "title": "Deep Ensembles: A Loss Landscape Perspective",
                    "abstract": "Deep ensembles have been empirically shown to be a promising approach for improving accuracy, uncertainty and out-of-distribution robustness of deep learning models. While deep ensembles were theoretically motivated by the bootstrap, non-bootstrap ensembles trained with just random initialization also perform well in practice, which suggests that there could be other explanations for why deep ensembles work well. Bayesian neural networks, which learn distributions over the parameters of the network, are theoretically well-motivated by Bayesian principles, but do not perform as well as deep ensembles in practice, particularly under dataset shift. One possible explanation for this gap between theory and practice is that popular scalable variational Bayesian methods tend to focus on a single mode, whereas deep ensembles tend to explore diverse modes in function space. We investigate this hypothesis by building on recent work on understanding the loss landscape of neural networks and adding our own exploration to measure the similarity of functions in the space of predictions. Our results show that random initializations explore entirely different modes, while functions along an optimization trajectory or sampled from the subspace thereof cluster within a single mode predictions-wise, while often deviating significantly in the weight space. Developing the concept of the diversity--accuracy plane, we show that the decorrelation power of random initializations is unmatched by popular subspace sampling methods. Finally, we evaluate the relative effects of ensembling, subspace based methods and ensembles of subspace based methods, and the experimental results validate our hypothesis."
                },
                {
                    "title": "Towards Explaining the Regularization Effect of Initial Large Learning Rate in Training Neural Networks",
                    "abstract": "Stochastic gradient descent with a large initial learning rate is a widely adopted method for training modern neural net architectures. Although a small initial learning rate allows for faster training and better test performance initially, the large learning rate achieves better generalization soon after the learning rate is annealed. Towards explaining this phenomenon, we devise a setting in which we can prove that a two layer network trained with large initial learning rate and annealing provably generalizes better than the same network trained with a small learning rate from the start. The key insight in our analysis is that the order of learning different types of patterns is crucial: because the small learning rate model first memorizes low noise, hard-to-fit patterns, it generalizes worse on higher noise, easier-to-fit patterns than its large learning rate counterpart. This concept translates to a larger-scale setting: we demonstrate that one can add a small patch to CIFAR-10 images that is immediately memorizable by a model with small initial learning rate, but ignored by the model with large learning rate until after annealing. Our experiments show that this causes the small learning rate model's accuracy on unmodified images to suffer, as it relies too much on the patch early on."
                },
                {
                    "title": "Does learning require memorization? a short tale about a long tail",
                    "abstract": "State-of-the-art results on image recognition tasks are achieved using over-parameterized learning algorithms that (nearly) perfectly fit the training set and are known to fit well even random labels. This tendency to memorize seemingly useless training data labels is not explained by existing theoretical analyses. Memorization of the training data also presents significant privacy risks when the training data contains sensitive personal information and thus it is important to understand whether such memorization is necessary for accurate learning. We provide a simple conceptual explanation and a theoretical model demonstrating that for natural data distributions memorization of labels is necessary for achieving close-to-optimal generalization error. The model is motivated and supported by the results of several recent empirical works. In our model, data is sampled from a mixture of subpopulations and the frequencies of these subpopulations are chosen from some prior. The model allows to quantify the effect of not fitting the training data on the generalization performance of the learned classifier and demonstrates that memorization is necessary whenever frequencies are long-tailed. Image and text data are known to follow such distributions and therefore our results establish a formal link between these empirical phenomena. Our results also have concrete implications for the cost of ensuring differential privacy in learning."
                },
                {
                    "title": "Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift",
                    "abstract": "Modern machine learning methods including deep learning have achieved great success in predictive accuracy for supervised learning tasks, but may still fall short in giving useful estimates of their predictive {\\em uncertainty}. Quantifying uncertainty is especially critical in real-world settings, which often involve input distributions that are shifted from the training distribution due to a variety of factors including sample bias and non-stationarity. In such settings, well calibrated uncertainty estimates convey information about when a model's output should (or should not) be trusted. Many probabilistic deep learning methods, including Bayesian-and non-Bayesian methods, have been proposed in the literature for quantifying predictive uncertainty, but to our knowledge there has not previously been a rigorous large-scale empirical comparison of these methods under dataset shift. We present a large-scale benchmark of existing state-of-the-art methods on classification problems and investigate the effect of dataset shift on accuracy and calibration. We find that traditional post-hoc calibration does indeed fall short, as do several other previous methods. However, some methods that marginalize over models give surprisingly strong results across a broad spectrum of tasks."
                },
                {
                    "title": "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks",
                    "abstract": "Convolutional Neural Networks (ConvNets) are commonly developed at a fixed resource budget, and then scaled up for better accuracy if more resources are available. In this paper, we systematically study model scaling and identify that carefully balancing network depth, width, and resolution can lead to better performance. Based on this observation, we propose a new scaling method that uniformly scales all dimensions of depth/width/resolution using a simple yet highly effective compound coefficient. We demonstrate the effectiveness of this method on scaling up MobileNets and ResNet. \nTo go even further, we use neural architecture search to design a new baseline network and scale it up to obtain a family of models, called EfficientNets, which achieve much better accuracy and efficiency than previous ConvNets. In particular, our EfficientNet-B7 achieves state-of-the-art 84.4% top-1 / 97.1% top-5 accuracy on ImageNet, while being 8.4x smaller and 6.1x faster on inference than the best existing ConvNet. Our EfficientNets also transfer well and achieve state-of-the-art accuracy on CIFAR-100 (91.7%), Flowers (98.8%), and 3 other transfer learning datasets, with an order of magnitude fewer parameters. Source code is at this https URL."
                },
                {
                    "title": "Similarity of Neural Network Representations Revisited",
                    "abstract": "Recent work has sought to understand the behavior of neural networks by comparing representations between layers and between different trained models. We examine methods for comparing neural network representations based on canonical correlation analysis (CCA). We show that CCA belongs to a family of statistics for measuring multivariate similarity, but that neither CCA nor any other statistic that is invariant to invertible linear transformation can measure meaningful similarities between representations of higher dimension than the number of data points. We introduce a similarity index that measures the relationship between representational similarity matrices and does not suffer from this limitation. This similarity index is equivalent to centered kernel alignment (CKA) and is also closely connected to CCA. Unlike CCA, CKA can reliably identify correspondences between representations in networks trained from different initializations."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "Insights on representational similarity in neural networks with canonical correlation",
                    "abstract": "Comparing different neural network representations and determining how representations evolve over time remain challenging open questions in our understanding of the function of neural networks. Comparing representations in neural networks is fundamentally difficult as the structure of representations varies greatly, even across groups of networks trained on identical tasks, and over the course of training. Here, we develop projection weighted CCA (Canonical Correlation Analysis) as a tool for understanding neural networks, building off of SVCCA, a recently proposed method (Raghu et al., 2017). We first improve the core method, showing how to differentiate between signal and noise, and then apply this technique to compare across a group of CNNs, demonstrating that networks which generalize converge to more similar representations than networks which memorize, that wider networks converge to more similar solutions than narrow networks, and that trained networks with identical topology but different learning rates converge to distinct clusters with diverse representations. We also investigate the representational dynamics of RNNs, across both training and sequential timesteps, finding that RNNs converge in a bottom-up pattern over the course of training and that the hidden state is highly variable over the course of a sequence, even when accounting for linear transforms. Together, these results provide new insights into the function of CNNs and RNNs, and demonstrate the utility of using CCA to understand representations."
                },
                {
                    "title": "Deep Learning Scaling is Predictable, Empirically",
                    "abstract": "Deep learning (DL) creates impactful advances following a virtuous recipe: model architecture search, creating large training data sets, and scaling computation. It is widely believed that growing training sets and models should improve accuracy and result in better products. As DL application domains grow, we would like a deeper understanding of the relationships between training set size, computational scale, and model accuracy improvements to advance the state-of-the-art. \nThis paper presents a large scale empirical characterization of generalization error and model size growth as training sets grow. We introduce a methodology for this measurement and test four machine learning domains: machine translation, language modeling, image processing, and speech recognition. Our empirical results show power-law generalization error scaling across a breadth of factors, resulting in power-law exponents---the \"steepness\" of the learning curve---yet to be explained by theoretical work. Further, model improvements only shift the error but do not appear to affect the power-law exponent. We also show that model size scales sublinearly with data size. These scaling relationships have significant implications on deep learning research, practice, and systems. They can assist model debugging, setting accuracy targets, and decisions about data set growth. They can also guide computing system design and underscore the importance of continued computational scaling."
                },
                {
                    "title": "Squeeze-and-Excitation Networks",
                    "abstract": "Convolutional neural networks are built upon the convolution operation, which extracts informative features by fusing spatial and channel-wise information together within local receptive fields. In order to boost the representational power of a network, several recent approaches have shown the benefit of enhancing spatial encoding. In this work, we focus on the channel relationship and propose a novel architectural unit, which we term the \"Squeeze-and-Excitation\" (SE) block, that adaptively recalibrates channel-wise feature responses by explicitly modelling interdependencies between channels. We demonstrate that by stacking these blocks together, we can construct SENet architectures that generalise extremely well across challenging datasets. Crucially, we find that SE blocks produce significant performance improvements for existing state-of-the-art deep architectures at minimal additional computational cost. SENets formed the foundation of our ILSVRC 2017 classification submission which won first place and significantly reduced the top-5 error to 2.251%, achieving a ~25% relative improvement over the winning entry of 2016. Code and models are available at https://github.com/hujie-frank/SENet."
                },
                {
                    "title": "ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices",
                    "abstract": "We introduce an extremely computation-efficient CNN architecture named ShuffleNet, which is designed specially for mobile devices with very limited computing power (e.g., 10-150 MFLOPs). The new architecture utilizes two new operations, pointwise group convolution and channel shuffle, to greatly reduce computation cost while maintaining accuracy. Experiments on ImageNet classification and MS COCO object detection demonstrate the superior performance of ShuffleNet over other structures, e.g. lower top-1 error (absolute 7.8%) than recent MobileNet [12] on ImageNet classification task, under the computation budget of 40 MFLOPs. On an ARM-based mobile device, ShuffleNet achieves ~13\u00c3\u2014 actual speedup over AlexNet while maintaining comparable accuracy."
                },
                {
                    "title": "SVCCA: Singular Vector Canonical Correlation Analysis for Deep Learning Dynamics and Interpretability",
                    "abstract": "We propose a new technique, Singular Vector Canonical Correlation Analysis (SVCCA), a tool for quickly comparing two representations in a way that is both invariant to affine transform (allowing comparison between different layers and networks) and fast to compute (allowing more comparisons to be calculated than with previous methods). We deploy this tool to measure the intrinsic dimensionality of layers, showing in some cases needless over-parameterization; to probe learning dynamics throughout training, finding that networks converge to final representations from the bottom up; to show where class-specific information in networks is formed; and to suggest new training regimes that simultaneously save computation and overfit less."
                },
                {
                    "title": "Network Dissection: Quantifying Interpretability of Deep Visual Representations",
                    "abstract": "We propose a general framework called Network Dissection for quantifying the interpretability of latent representations of CNNs by evaluating the alignment between individual hidden units and a set of semantic concepts. Given any CNN model, the proposed method draws on a data set of concepts to score the semantics of hidden units at each intermediate convolutional layer. The units with semantics are labeled across a broad range of visual concepts including objects, parts, scenes, textures, materials, and colors. We use the proposed method to test the hypothesis that interpretability is an axis-independent property of the representation space, then we apply the method to compare the latent representations of various networks when trained to solve different classification problems. We further analyze the effect of training iterations, compare networks trained with different initializations, and measure the effect of dropout and batch normalization on the interpretability of deep visual representations. We demonstrate that the proposed method can shed light on characteristics of CNN models and training methods that go beyond measurements of their discriminative power."
                },
                {
                    "title": "MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications",
                    "abstract": "We present a class of efficient models called MobileNets for mobile and embedded vision applications. MobileNets are based on a streamlined architecture that uses depth-wise separable convolutions to build light weight deep neural networks. We introduce two simple global hyper-parameters that efficiently trade off between latency and accuracy. These hyper-parameters allow the model builder to choose the right sized model for their application based on the constraints of the problem. We present extensive experiments on resource and accuracy tradeoffs and show strong performance compared to other popular models on ImageNet classification. We then demonstrate the effectiveness of MobileNets across a wide range of applications and use cases including object detection, finegrain classification, face attributes and large scale geo-localization."
                },
                {
                    "title": "Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles",
                    "abstract": "Deep neural networks (NNs) are powerful black box predictors that have recently achieved impressive performance on a wide spectrum of tasks. Quantifying predictive uncertainty in NNs is a challenging and yet unsolved problem. Bayesian NNs, which learn a distribution over weights, are currently the state-of-the-art for estimating predictive uncertainty; however these require significant modifications to the training procedure and are computationally expensive compared to standard (non-Bayesian) NNs. We propose an alternative to Bayesian NNs that is simple to implement, readily parallelizable, requires very little hyperparameter tuning, and yields high quality predictive uncertainty estimates. Through a series of experiments on classification and regression benchmarks, we demonstrate that our method produces well-calibrated uncertainty estimates which are as good or better than approximate Bayesian NNs. To assess robustness to dataset shift, we evaluate the predictive uncertainty on test examples from known and unknown distributions, and show that our method is able to express higher uncertainty on out-of-distribution examples. We demonstrate the scalability of our method by evaluating predictive uncertainty estimates on ImageNet."
                },
                {
                    "title": "Aggregated Residual Transformations for Deep Neural Networks",
                    "abstract": "We present a simple, highly modularized network architecture for image classification. Our network is constructed by repeating a building block that aggregates a set of transformations with the same topology. Our simple design results in a homogeneous, multi-branch architecture that has only a few hyper-parameters to set. This strategy exposes a new dimension, which we call cardinality (the size of the set of transformations), as an essential factor in addition to the dimensions of depth and width. On the ImageNet-1K dataset, we empirically show that even under the restricted condition of maintaining complexity, increasing cardinality is able to improve classification accuracy. Moreover, increasing cardinality is more effective than going deeper or wider when we increase the capacity. Our models, named ResNeXt, are the foundations of our entry to the ILSVRC 2016 classification task in which we secured 2nd place. We further investigate ResNeXt on an ImageNet-5K set and the COCO detection set, also showing better results than its ResNet counterpart. The code and models are publicly available online."
                },
                {
                    "title": "Densely Connected Convolutional Networks",
                    "abstract": "Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections&#x2014;one between each layer and its subsequent layer&#x2014;our network has L(L+1)/2 direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters. We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet). DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less memory and computation to achieve high performance. Code and pre-trained models are available at https://github.com/liuzhuang13/DenseNet."
                },
                {
                    "title": "Synthesizing the preferred inputs for neurons in neural networks via deep generator networks",
                    "abstract": "Deep neural networks (DNNs) have demonstrated state-of-the-art results on many pattern recognition tasks, especially vision classification problems. Understanding the inner workings of such computational brains is both fascinating basic science that is interesting in its own right - similar to why we study the human brain - and will enable researchers to further improve DNNs. One path to understanding how a neural network functions internally is to study what each of its neurons has learned to detect. One such method is called activation maximization (AM), which synthesizes an input (e.g. an image) that highly activates a neuron. Here we dramatically improve the qualitative state of the art of activation maximization by harnessing a powerful, learned prior: a deep generator network (DGN). The algorithm (1) generates qualitatively state-of-the-art synthetic images that look almost real, (2) reveals the features learned by each neuron in an interpretable way, (3) generalizes well to new datasets and somewhat well to different network architectures without requiring the prior to be relearned, and (4) can be considered as a high-quality generative method (in this case, by generating novel, creative, interesting, recognizable images)."
                },
                {
                    "title": "SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <1MB model size",
                    "abstract": "Recent research on deep neural networks has focused primarily on improving accuracy. For a given accuracy level, it is typically possible to identify multiple DNN architectures that achieve that accuracy level. With equivalent accuracy, smaller DNN architectures offer at least three advantages: (1) Smaller DNNs require less communication across servers during distributed training. (2) Smaller DNNs require less bandwidth to export a new model from the cloud to an autonomous car. (3) Smaller DNNs are more feasible to deploy on FPGAs and other hardware with limited memory. To provide all of these advantages, we propose a small DNN architecture called SqueezeNet. SqueezeNet achieves AlexNet-level accuracy on ImageNet with 50x fewer parameters. Additionally, with model compression techniques we are able to compress SqueezeNet to less than 0.5MB (510x smaller than AlexNet). \nThe SqueezeNet architecture is available for download here: this https URL"
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Convergent Learning: Do different neural networks learn the same representations?",
                    "abstract": "Recent success in training deep neural networks have prompted active investigation into the features learned on their intermediate layers. Such research is difficult because it requires making sense of non-linear computations performed by millions of parameters, but valuable because it increases our ability to understand current models and create improved versions of them. In this paper we investigate the extent to which neural networks exhibit what we call convergent learning, which is when the representations learned by multiple nets converge to a set of features which are either individually similar between networks or where subsets of features span similar low-dimensional spaces. We propose a specific method of probing representations: training multiple networks and then comparing and contrasting their individual, learned representations at the level of neurons or groups of neurons. We begin research into this question using three techniques to approximately align different neural networks on a feature level: a bipartite matching approach that makes one-to-one assignments between neurons, a sparse prediction approach that finds one-to-many mappings, and a spectral clustering approach that finds many-to-many mappings. This initial investigation reveals a few previously unknown properties of neural networks, and we argue that future research into the question of convergent learning will yield many more. The insights described here include (1) that some features are learned reliably in multiple networks, yet other features are not consistently learned; (2) that units learn to span low-dimensional subspaces and, while these subspaces are common to multiple networks, the specific basis vectors learned are not; (3) that the representation codes show evidence of being a mix between a local code and slightly, but not fully, distributed codes across multiple units."
                },
                {
                    "title": "Weight Uncertainty in Neural Network",
                    "abstract": "We introduce a new, efficient, principled and backpropagation-compatible algorithm for learning a probability distribution on the weights of a neural network, called Bayes by Backprop. It regularises the weights by minimising a compression cost, known as the variational free energy or the expected lower bound on the marginal likelihood. We show that this principled kind of regularisation yields comparable performance to dropout on MNIST classification. We then demonstrate how the learnt uncertainty in the weights can be used to improve generalisation in non-linear regression problems, and how this weight uncertainty can be used to drive the exploration-exploitation trade-off in reinforcement learning."
                },
                {
                    "title": "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning",
                    "abstract": "Deep learning tools have gained tremendous attention in applied machine learning. However such tools for regression and classification do not capture model uncertainty. In comparison, Bayesian models offer a mathematically grounded framework to reason about model uncertainty, but usually come with a prohibitive computational cost. In this paper we develop a new theoretical framework casting dropout training in deep neural networks (NNs) as approximate Bayesian inference in deep Gaussian processes. A direct result of this theory gives us tools to model uncertainty with dropout NNs -- extracting information from existing models that has been thrown away so far. This mitigates the problem of representing uncertainty in deep learning without sacrificing either computational complexity or test accuracy. We perform an extensive study of the properties of dropout's uncertainty. Various network architectures and non-linearities are assessed on tasks of regression and classification, using MNIST as an example. We show a considerable improvement in predictive log-likelihood and RMSE compared to existing state-of-the-art methods, and finish by using dropout's uncertainty in deep reinforcement learning."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "Representation Learning: A Review and New Perspectives",
                    "abstract": "The success of machine learning algorithms generally depends on data representation, and we hypothesize that this is because different representations can entangle and hide more or less the different explanatory factors of variation behind the data. Although specific domain knowledge can be used to help design representations, learning with generic priors can also be used, and the quest for AI is motivating the design of more powerful representation-learning algorithms implementing such priors. This paper reviews recent work in the area of unsupervised feature learning and deep learning, covering advances in probabilistic models, autoencoders, manifold learning, and deep networks. This motivates longer term unanswered questions about the appropriate objectives for learning good representations, for computing representations (i.e., inference), and the geometrical connections between representation learning, density estimation, and manifold learning."
                },
                {
                    "title": "Bayesian Learning via Stochastic Gradient Langevin Dynamics",
                    "abstract": "In this paper we propose a new framework for learning from large scale datasets based on iterative learning from small mini-batches. By adding the right amount of noise to a standard stochastic gradient optimization algorithm we show that the iterates will converge to samples from the true posterior distribution as we anneal the stepsize. This seamless transition between optimization and Bayesian posterior sampling provides an inbuilt protection against overfitting. We also propose a practical method for Monte Carlo estimates of posterior statistics which monitors a \"sampling threshold\" and collects samples after it has been surpassed. We apply the method to three models: a mixture of Gaussians, logistic regression and ICA with natural gradients."
                },
                {
                    "title": "Feature selection for ensembles:a hierarchical multi-objective genetic algorithm approach",
                    "abstract": "Feature selection for ensembles has shown to be an effectivestrategy for ensemble creation. In this paper we presentan ensemble feature selection approach based on a hierarchicalmulti-objective genetic algorithm. The first level performsfeature selection in order to generate a set of goodclassifiers while the second one combines them to providea set of powerful ensembles. The proposed method is evaluatedin the context of handwritten digit recognition, usingthree different feature sets and neural networks (MLP) asclassifiers. Experiments conducted on NIST SD19 demonstratedthe effectiveness of the proposed strategy."
                },
                {
                    "title": "Verification of Probabilistic Predictions: A Brief Review",
                    "abstract": "Abstract The evaluation process is considered in some detail with particular reference to probabilistic predictions. The process consists of several ordered steps at each of which elements (of the process) are identified. Consideration of the purposes leads to the identification of two distinct forms of evaluation: operational evaluation concerned with the value of predictions to the user and empirical evaluation, or verification, concerned with the perfection of predictions, i.e., the association between predictions and observations. Attributes, i.e., desirable properties, of predictions are defined with reference to these purposes, and a number of measures of the attributes for empirical evaluation are considered. An artificial example of comparative verification in which different measures appear to yield contradictory results is used to demonstrate the importance of, and need for, a careful analysis of the evaluation process."
                },
                {
                    "title": "Local Signal Adaptivity: Provable Feature Learning in Neural Networks Beyond Kernels",
                    "abstract": "Neural networks have been shown to outperform kernel methods in practice (including neural tangent kernels). Most theoretical explanations of this performance gap focus on learning a complex hypothesis class; in some cases, it is unclear whether this hypothesis class captures realistic data. In this work, we propose a related, but alternative, explanation for this performance gap in the image classi\ufb01cation setting, based on \ufb01nding a sparse signal in the presence of noise. Speci\ufb01cally, we prove that, for a simple data distribution with sparse signal amidst high-variance noise, a simple convolutional neural network trained using stochastic gradient descent simultaneously learns to threshold out the noise and \ufb01nd the signal. On the other hand, the corresponding neural tangent kernel, with a \ufb01xed set of predetermined features, is unable to adapt to the signal in this manner. We supplement our theoretical results by demonstrating this phenomenon empirically: in CIFAR-10 and MNIST images with various backgrounds, as the background noise increases in intensity, a CNN\u2019s performance stays relatively robust, whereas its corresponding neural tangent kernel sees a notable drop in performance. We therefore propose the local signal adaptivity (LSA) phenomenon as one explanation for the superiority of neural networks over kernel methods."
                },
                {
                    "title": "Tensor Programs IV: Feature Learning in Infinite-Width Neural Networks",
                    "abstract": "As its width tends to in\ufb01nity, a deep neural network\u2019s behavior under gradient descent can become simpli\ufb01ed and predictable (e.g. given by the Neural Tangent Kernel (NTK)), if it is parametrized appropriately (e.g. the NTK parametrization). However, we show that the standard and NTK parametrizations of a neural network do not admit in\ufb01nite-width limits that can learn features, which is crucial for pretraining and transfer learning such as with BERT. We propose simple modi\ufb01cations to the standard parametrization to allow for feature learning in the limit. Using the Tensor Programs technique, we derive explicit formulas for such limits. On Word2Vec and few-shot learning on Omniglot via MAML, two canonical tasks that rely crucially on feature learning, we compute these limits exactly. We \ufb01nd that they outperform both NTK baselines and \ufb01nite-width networks, with the latter approaching the in\ufb01nite-width feature learning performance as width increases. See arXiv:2011.14522 for the full version of this paper."
                },
                {
                    "title": "GENERATIVE ADVERSARIAL NETS",
                    "abstract": "Estimating individualized treatment effects (ITE) is a challenging task due to the need for an individual\u2019s potential outcomes to be learned from biased data and without having access to the counterfactuals. We propose a novel method for inferring ITE based on the Generative Adversarial Nets (GANs) framework. Our method, termed Generative Adversarial Nets for inference of Individualized Treatment Effects (GANITE), is motivated by the possibility that we can capture the uncertainty in the counterfactual distributions by attempting to learn them using a GAN. We generate proxies of the counterfactual outcomes using a counterfactual generator, G, and then pass these proxies to an ITE generator, I, in order to train it. By modeling both of these using the GAN framework, we are able to infer based on the factual data, while still accounting for the unseen counterfactuals. We test our method on three real-world datasets (with both binary and multiple treatments) and show that GANITE outperforms state-of-the-art methods."
                },
                {
                    "title": "Reading Digits in Natural Images with Unsupervised Feature Learning",
                    "abstract": "Detecting and reading text from natural images is a hard computer vision task that is central to a variety of emerging applications. Related problems like document character recognition have been widely studied by computer vision and machine learning researchers and are virtually solved for practical applications like reading handwritten digits. Reliably recognizing characters in more complex scenes like photographs, however, is far more difficult: the best existing methods lag well behind human performance on the same tasks. In this paper we attack the problem of recognizing digits in a real application using unsupervised feature learning methods: reading house numbers from street level photos. To this end, we introduce a new benchmark dataset for research use containing over 600,000 labeled digits cropped from Street View images. We then demonstrate the difficulty of recognizing these digits when the problem is approached with hand-designed features. Finally, we employ variants of two recently proposed unsupervised feature learning methods and find that they are convincingly superior on our benchmarks."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nWhat is the theoretical understanding of the role of features in deep learning, and how can we define and analyze these features to better understand the knowledge learned by different models?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant gap in our understanding of deep learning mechanisms. A clearer definition and analysis of features could lead to improved interpretability of models, better generalization, and enhanced performance in downstream tasks. This research could pave the way for future studies on feature learning, potentially leading to practical applications in various domains such as computer vision, natural language processing, and beyond. By advancing our knowledge of feature learning, we can develop more robust models that leverage transferable knowledge effectively.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the lack of consensus on the definition of features in deep learning, particularly beyond simple models. Naive approaches may fail because they do not account for the complexity of feature interactions and the variability in knowledge learned by models trained with different initializations. Technical obstacles include the need for a rigorous theoretical framework that can quantitatively analyze features and their impact on model behavior. Additionally, the empirical investigation of feature learning requires sophisticated methodologies to capture the nuances of how different models learn from data.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on empirical observations and specific aspects of feature learning without establishing a comprehensive theoretical framework. Limitations in existing solutions include an over-reliance on frameworks like the neural tangent kernel (NTK), which do not adequately address feature learning in the context of observed data. Barriers such as the complexity of deep learning models and the diversity of learned features have hindered progress. Our approach differs by proposing a new construct for defining features that allows for both qualitative and quantitative analysis, thereby addressing the shortcomings of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves defining a new construct for features that enables the analysis of the similarities and differences in knowledge learned by various models. We will conduct empirical investigations using datasets such as CIFAR-10, focusing on metrics that quantify feature diversity and model performance. The expected outcomes include a clearer understanding of feature learning dynamics, insights into the Generalization Disagreement Equality (GDE), and a framework that can be applied to enhance model interpretability"
            }
        },
        "author_data": {
            "1f02ccb6-abc1-4f1c-af92-af4c69a452a0": {
                "pk": "1f02ccb6-abc1-4f1c-af92-af4c69a452a0",
                "name": "Yiding Jiang",
                "collaborators": [
                    "J. Z. Kolter",
                    "Chelsea Finn",
                    "H. Mobahi",
                    "Samy Bengio",
                    "Behnam Neyshabur",
                    "Allan Zhou",
                    "Kaien Yang",
                    "Kaylee Burns",
                    "Samuel Sokota",
                    "Zico Kolter"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Neural Networks",
                    "Generalization",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "On the Importance of Exploration for Generalization in Reinforcement Learning",
                        "abstract": "Existing approaches for improving generalization in deep reinforcement learning (RL) have mostly focused on representation learning, neglecting RL-specific aspects such as exploration. We hypothesize that the agent's exploration strategy plays a key role in its ability to generalize to new environments. Through a series of experiments in a tabular contextual MDP, we show that exploration is helpful not only for efficiently finding the optimal policy for the training environments but also for acquiring knowledge that helps decision making in unseen environments. Based on these observations, we propose EDE: Exploration via Distributional Ensemble, a method that encourages exploration of states with high epistemic uncertainty through an ensemble of Q-value distributions. Our algorithm is the first value-based approach to achieve state-of-the-art on both Procgen and Crafter, two benchmarks for generalization in RL with high-dimensional observations. The open-sourced implementation can be found at https://github.com/facebookresearch/ede ."
                    },
                    {
                        "title": "Permutation Equivariant Neural Functionals",
                        "abstract": "This work studies the design of neural networks that can process the weights or gradients of other neural networks, which we refer to as neural functional networks (NFNs). Despite a wide range of potential applications, including learned optimization, processing implicit neural representations, network editing, and policy evaluation, there are few unifying principles for designing effective architectures that process the weights of other networks. We approach the design of neural functionals through the lens of symmetry, in particular by focusing on the permutation symmetries that arise in the weights of deep feedforward networks because hidden layer neurons have no inherent order. We introduce a framework for building permutation equivariant neural functionals, whose architectures encode these symmetries as an inductive bias. The key building blocks of this framework are NF-Layers (neural functional layers) that we constrain to be permutation equivariant through an appropriate parameter sharing scheme. In our experiments, we find that permutation equivariant neural functionals are effective on a diverse set of tasks that require processing the weights of MLPs and CNNs, such as predicting classifier generalization, producing\"winning ticket\"sparsity masks for initializations, and classifying or editing implicit neural representations (INRs). In addition, we provide code for our models and experiments at https://github.com/AllanYangZhou/nfn."
                    },
                    {
                        "title": "Language models are weak learners",
                        "abstract": "A central notion in practical and theoretical machine learning is that of a $\\textit{weak learner}$, classifiers that achieve better-than-random performance (on any given distribution over data), even by a small margin. Such weak learners form the practical basis for canonical machine learning methods such as boosting. In this work, we illustrate that prompt-based large language models can operate effectively as said weak learners. Specifically, we illustrate the use of a large language model (LLM) as a weak learner in a boosting algorithm applied to tabular data. We show that by providing (properly sampled according to the distribution of interest) text descriptions of tabular data samples, LLMs can produce a summary of the samples that serves as a template for classification and achieves the aim of acting as a weak learner on this task. We incorporate these models into a boosting approach, which in some settings can leverage the knowledge within the LLM to outperform traditional tree-based boosting. The model outperforms both few-shot learning and occasionally even more involved fine-tuning procedures, particularly for tasks involving small numbers of data points. The results illustrate the potential for prompt-based LLMs to function not just as few-shot learners themselves, but as components of larger machine learning pipelines."
                    },
                    {
                        "title": "Neural Functional Transformers",
                        "abstract": "The recent success of neural networks as implicit representation of data has driven growing interest in neural functionals: models that can process other neural networks as input by operating directly over their weight spaces. Nevertheless, constructing expressive and efficient neural functional architectures that can handle high-dimensional weight-space objects remains challenging. This paper uses the attention mechanism to define a novel set of permutation equivariant weight-space layers and composes them into deep equivariant models called neural functional Transformers (NFTs). NFTs respect weight-space permutation symmetries while incorporating the advantages of attention, which have exhibited remarkable success across multiple domains. In experiments processing the weights of feedforward MLPs and CNNs, we find that NFTs match or exceed the performance of prior weight-space methods. We also leverage NFTs to develop Inr2Array, a novel method for computing permutation invariant latent representations from the weights of implicit neural representations (INRs). Our proposed method improves INR classification accuracy by up to $+17\\%$ over existing methods. We provide an implementation of our layers at https://github.com/AllanYangZhou/nfn."
                    },
                    {
                        "title": "Learning Options via Compression",
                        "abstract": "Identifying statistical regularities in solutions to some tasks in multi-task reinforcement learning can accelerate the learning of new tasks. Skill learning offers one way of identifying these regularities by decomposing pre-collected experiences into a sequence of skills. A popular approach to skill learning is maximizing the likelihood of the pre-collected experience with latent variable models, where the latent variables represent the skills. However, there are often many solutions that maximize the likelihood equally well, including degenerate solutions. To address this underspecification, we propose a new objective that combines the maximum likelihood objective with a penalty on the description length of the skills. This penalty incentivizes the skills to maximally extract common structures from the experiences. Empirically, our objective learns skills that solve downstream tasks in fewer samples compared to skills learned from only maximizing likelihood. Further, while most prior works in the offline multi-task setting focus on tasks with low-dimensional observations, our objective can scale to challenging tasks with high-dimensional image observations."
                    },
                    {
                        "title": "Agreement-on-the-Line: Predicting the Performance of Neural Networks under Distribution Shift",
                        "abstract": "Recently, Miller et al. showed that a model's in-distribution (ID) accuracy has a strong linear correlation with its out-of-distribution (OOD) accuracy on several OOD benchmarks -- a phenomenon they dubbed ''accuracy-on-the-line''. While a useful tool for model selection (i.e., the model most likely to perform the best OOD is the one with highest ID accuracy), this fact does not help estimate the actual OOD performance of models without access to a labeled OOD validation set. In this paper, we show a similar but surprising phenomenon also holds for the agreement between pairs of neural network classifiers: whenever accuracy-on-the-line holds, we observe that the OOD agreement between the predictions of any two pairs of neural networks (with potentially different architectures) also observes a strong linear correlation with their ID agreement. Furthermore, we observe that the slope and bias of OOD vs ID agreement closely matches that of OOD vs ID accuracy. This phenomenon, which we call agreement-on-the-line, has important practical applications: without any labeled data, we can predict the OOD accuracy of classifiers}, since OOD agreement can be estimated with just unlabeled data. Our prediction algorithm outperforms previous methods both in shifts where agreement-on-the-line holds and, surprisingly, when accuracy is not on the line. This phenomenon also provides new insights into deep neural networks: unlike accuracy-on-the-line, agreement-on-the-line appears to only hold for neural network classifiers."
                    },
                    {
                        "title": "Ask & Explore: Grounded Question Answering for Curiosity-Driven Exploration",
                        "abstract": "In many real-world scenarios where extrinsic rewards to the agent are extremely sparse, curiosity has emerged as a useful concept providing intrinsic rewards that enable the agent to explore its environment and acquire information to achieve its goals. Despite their strong performance on many sparse-reward tasks, existing curiosity approaches rely on an overly holistic view of state transitions, and do not allow for a structured understanding of specific aspects of the environment. In this paper, we formulate curiosity based on grounded question answering by encouraging the agent to ask questions about the environment and be curious when the answers to these questions change. We show that natural language questions encourage the agent to uncover specific knowledge about their environment such as the physical properties of objects as well as their spatial relationships with other objects, which serve as valuable curiosity rewards to solve sparse-reward tasks more efficiently."
                    },
                    {
                        "title": "Assessing Generalization of SGD via Disagreement",
                        "abstract": "We empirically show that the test error of deep networks can be estimated by simply training the same architecture on the same training set but with a different run of Stochastic Gradient Descent (SGD), and measuring the disagreement rate between the two networks on unlabeled test data. This builds on -- and is a stronger version of -- the observation in Nakkiran&Bansal '20, which requires the second run to be on an altogether fresh training set. We further theoretically show that this peculiar phenomenon arises from the \\emph{well-calibrated} nature of \\emph{ensembles} of SGD-trained models. This finding not only provides a simple empirical measure to directly predict the test error using unlabeled test data, but also establishes a new conceptual connection between generalization and calibration."
                    },
                    {
                        "title": "Methods and Analysis of The First Competition in Predicting Generalization of Deep Learning",
                        "abstract": "Deep learning has been recently successfully applied to an ever larger number of problems, ranging from pattern recognition to complex decision making. However, several concerns have been raised, including guarantees of good generalization , which is of foremost importance. Despite numerous attempts, conventional statistical learning approaches fall short of providing a satisfactory explanation on why deep learning works. In a competition hosted at the Thirty-Fourth Conference on Neural Information Processing Systems (NeurIPS 2020)"
                    },
                    {
                        "title": "NeurIPS 2020 Competition: Predicting Generalization in Deep Learning",
                        "abstract": "Understanding generalization in deep learning is arguably one of the most important questions in deep learning. Deep learning has been successfully adopted to a large number of problems ranging from pattern recognition to complex decision making, but many recent researchers have raised many concerns about deep learning, among which the most important is generalization. Despite numerous attempts, conventional statistical learning approaches have yet been able to provide a satisfactory explanation on why deep learning works. A recent line of works aims to address the problem by trying to predict the generalization performance through complexity measures. In this competition, we invite the community to propose complexity measures that can accurately predict generalization of models. A robust and general complexity measure would potentially lead to a better understanding of deep learning's underlying mechanism and behavior of deep models on unseen data, or shed light on better generalization bounds. All these outcomes will be important for making deep learning more robust and reliable."
                    },
                    {
                        "title": "Observational Overfitting in Reinforcement Learning",
                        "abstract": "A major component of overfitting in model-free reinforcement learning (RL) involves the case where the agent may mistakenly correlate reward with certain spurious features from the observations generated by the Markov Decision Process (MDP). We provide a general framework for analyzing this scenario, which we use to design multiple synthetic benchmarks from only modifying the observation space of an MDP. When an agent overfits to different observation spaces even if the underlying MDP dynamics is fixed, we term this observational overfitting. Our experiments expose intriguing properties especially with regards to implicit regularization, and also corroborate results from previous works in RL generalization and supervised learning (SL)."
                    },
                    {
                        "title": "Fantastic Generalization Measures and Where to Find Them",
                        "abstract": "Generalization of deep networks has been of great interest in recent years, resulting in a number of theoretically and empirically motivated complexity measures. However, most papers proposing such measures study only a small set of models, leaving open the question of whether the conclusion drawn from those experiments would remain valid in other settings. We present the first large scale study of generalization in deep networks. We investigate more then 40 complexity measures taken from both theoretical bounds and empirical studies. We train over 10,000 convolutional networks by systematically varying commonly used hyperparameters. Hoping to uncover potentially causal relationships between each measure and generalization, we analyze carefully controlled experiments and show surprising failures of some measures as well as promising measures for further research."
                    },
                    {
                        "title": "Language as an Abstraction for Hierarchical Deep Reinforcement Learning",
                        "abstract": "Solving complex, temporally-extended tasks is a long-standing problem in reinforcement learning (RL). We hypothesize that one critical element of solving such problems is the notion of compositionality. With the ability to learn concepts and sub-skills that can be composed to solve longer tasks, i.e. hierarchical RL, we can acquire temporally-extended behaviors. However, acquiring effective yet general abstractions for hierarchical RL is remarkably challenging. In this paper, we propose to use language as the abstraction, as it provides unique compositional structure, enabling fast learning and combinatorial generalization, while retaining tremendous flexibility, making it suitable for a variety of problems. Our approach learns an instruction-following low-level policy and a high-level policy that can reuse abstractions across tasks, in essence, permitting agents to reason using structured language. To study compositional task learning, we introduce an open-source object interaction environment built using the MuJoCo physics engine and the CLEVR engine. We find that, using our approach, agents can learn to solve to diverse, temporally-extended tasks such as object sorting and multi-object rearrangement, including from raw pixel observations. Our analysis reveals that the compositional nature of language is critical for learning diverse sub-skills and systematically generalizing to new sub-skills in comparison to non-compositional abstractions that use the same supervision."
                    },
                    {
                        "title": "Adversarial Grasp Objects",
                        "abstract": "Learning-based approaches to robust robot grasp planning can grasp a wide variety of objects, but may be prone to failure on some objects. Inspired by recent results in computer vision, we define a class of \u201cadversarial grasp objects that are physically similar to a given object but significantly less \u201dgraspable\u201d in terms of a specified robot grasping policy. We present three algorithms for synthesizing adversarial grasp objects under the grasp reliability measure of Dex-Net 1.0 for parallel-jaw grippers: 1) two analytic algorithms that perturb vertices on antipodal faces (one that uses random perturbations and one that uses systematic perturbations), and 2) a deep-learning-based approach using a variation of the Cross-Entropy Method (CEM) augmented with a generative adversarial network (GAN) to synthesize classes of adversarial grasp objects represented by discrete Signed Distance Functions. The random perturbation algorithm reduces graspability by 32%, 12%, and 32% for intersected cylinders, intersected prisms, and ShapeNet bottles, respectively, while maintaining shape similarity using geometric constraints. The systematic perturbation algorithm reduces graspability by 32%, 11%, and 21%; and the GAN reduces graspability by 22%, 36%, and 17%, on the same objects. We use the algorithms to generate and 3D print adversarial grasp objects. Simulation and physical experiments confirm that all algorithms are effective at reducing graspability."
                    },
                    {
                        "title": "Predicting the Generalization Gap in Deep Networks with Margin Distributions",
                        "abstract": "As shown in recent research, deep neural networks can perfectly fit randomly labeled data, but with very poor accuracy on held out data. This phenomenon indicates that loss functions such as cross-entropy are not a reliable indicator of generalization. This leads to the crucial question of how generalization gap should be predicted from the training data and network parameters. In this paper, we propose such a measure, and conduct extensive empirical studies on how well it can predict the generalization gap. Our measure is based on the concept of margin distribution, which are the distances of training points to the decision boundary. We find that it is necessary to use margin distributions at multiple layers of a deep network. On the CIFAR-10 and the CIFAR-100 datasets, our proposed measure correlates very strongly with the generalization gap. In addition, we find the following other factors to be of importance: normalizing margin values for scale independence, using characterizations of margin distribution rather than just the margin (closest distance to decision boundary), and working in log space instead of linear space (effectively using a product of margins rather than a sum). Our measure can be easily applied to feedforward deep networks with any architecture and may point towards new training loss functions that could enable better generalization."
                    }
                ]
            },
            "2fa89700-416c-46a5-99ab-db9b63438807": {
                "pk": "2fa89700-416c-46a5-99ab-db9b63438807",
                "name": "Christina Baek",
                "collaborators": [
                    "Byoung-Tak Zhang",
                    "Beom-Jin Lee",
                    "Donghyun Kwak",
                    "Yi Ma",
                    "Sang-Woo Lee",
                    "Je-hwan Ryu",
                    "Ji\u2010Hoon Lee",
                    "Yiding Jiang",
                    "Ziyang Wu",
                    "Yuexiang Zhai"
                ],
                "domain": [
                    "Machine Learning",
                    "DNA Computing",
                    "Reinforcement Learning",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Agreement-on-the-Line: Predicting the Performance of Neural Networks under Distribution Shift",
                        "abstract": "Recently, Miller et al. showed that a model's in-distribution (ID) accuracy has a strong linear correlation with its out-of-distribution (OOD) accuracy on several OOD benchmarks -- a phenomenon they dubbed ''accuracy-on-the-line''. While a useful tool for model selection (i.e., the model most likely to perform the best OOD is the one with highest ID accuracy), this fact does not help estimate the actual OOD performance of models without access to a labeled OOD validation set. In this paper, we show a similar but surprising phenomenon also holds for the agreement between pairs of neural network classifiers: whenever accuracy-on-the-line holds, we observe that the OOD agreement between the predictions of any two pairs of neural networks (with potentially different architectures) also observes a strong linear correlation with their ID agreement. Furthermore, we observe that the slope and bias of OOD vs ID agreement closely matches that of OOD vs ID accuracy. This phenomenon, which we call agreement-on-the-line, has important practical applications: without any labeled data, we can predict the OOD accuracy of classifiers}, since OOD agreement can be estimated with just unlabeled data. Our prediction algorithm outperforms previous methods both in shifts where agreement-on-the-line holds and, surprisingly, when accuracy is not on the line. This phenomenon also provides new insights into deep neural networks: unlike accuracy-on-the-line, agreement-on-the-line appears to only hold for neural network classifiers."
                    },
                    {
                        "title": "Efficient Maximal Coding Rate Reduction by Variational Forms",
                        "abstract": "The principle of Maximal Coding Rate Reduction (MCR2) has recently been proposed as a training objective for learning discriminative low-dimensional structures intrinsic to high-dimensional data to allow for more robust training than standard approaches, such as cross-entropy minimization. However, despite the advantages that have been shown for MCR2 training, MCR2 suffers from a significant computational cost due to the need to evaluate and differentiate a significant number of log-determinant terms that grows linearly with the number of classes. By taking advantage of variational forms of spectral functions of a matrix, we reformulate the MCR2 objective to a form that can scale significantly without compromising training accuracy. Experiments in image classification demonstrate that our proposed formulation results in a significant speed up over optimizing the original MCR2 objective directly and often results in higher quality learned representations. Further, our approach may be of independent interest in other models that require computation of log-determinant forms, such as in system identification or normalizing flow models."
                    },
                    {
                        "title": "Computational Benefits of Intermediate Rewards for Hierarchical Planning",
                        "abstract": "Many hierarchical reinforcement learning (RL) applications have empirically veri\ufb01ed that incorporating prior knowledge in reward design improves convergence speed and practical performance. We attempt to quantify the computational bene\ufb01ts of hierarchical RL from a planning perspective under assumptions about the intermediate state and intermediate rewards frequently (but often implicitly) adopted in practice. Our approach reveals a trade-o\ufb00 between computational complexity and the pursuit of the shortest path in hierarchical planning: using intermediate rewards signi\ufb01cantly reduces the computational complexity in \ufb01nding a successful policy but does not guarantee to \ufb01nd the shortest path, whereas using sparse terminal rewards \ufb01nds the shortest path at a signi\ufb01cantly higher computational cost. We also corroborate our theoretical results with extensive experiments on the MiniGrid environments using Q-learning and other popular deep RL algorithms"
                    },
                    {
                        "title": "Assessing Generalization of SGD via Disagreement",
                        "abstract": "We empirically show that the test error of deep networks can be estimated by simply training the same architecture on the same training set but with a different run of Stochastic Gradient Descent (SGD), and measuring the disagreement rate between the two networks on unlabeled test data. This builds on -- and is a stronger version of -- the observation in Nakkiran&Bansal '20, which requires the second run to be on an altogether fresh training set. We further theoretically show that this peculiar phenomenon arises from the \\emph{well-calibrated} nature of \\emph{ensembles} of SGD-trained models. This finding not only provides a simple empirical measure to directly predict the test error using unlabeled test data, but also establishes a new conceptual connection between generalization and calibration."
                    },
                    {
                        "title": "Computational Benefits of Intermediate Rewards for Goal-Reaching Policy Learning",
                        "abstract": "     Many goal-reaching reinforcement learning (RL) tasks have empirically verified that rewarding the agent on subgoals improves convergence speed and practical performance. We attempt to provide a theoretical framework to quantify the computational benefits of rewarding the completion of subgoals, in terms of the number of synchronous value iterations. In particular, we consider subgoals as one-way intermediate states, which can only be visited once per episode and propose two settings that consider these one-way intermediate states: the one-way single-path (OWSP) and the one-way multi-path (OWMP) settings. In both OWSP and OWMP settings, we demonstrate that adding intermediate rewards to subgoals is more computationally efficient than only rewarding the agent once it completes the goal of reaching a terminal state. We also reveal a trade-off between computational complexity and the pursuit of the shortest path in the OWMP setting: adding intermediate rewards significantly reduces the computational complexity of reaching the goal but the agent may not find the shortest path, whereas with sparse terminal rewards, the agent finds the shortest path at a significantly higher computational cost. We also corroborate our theoretical results with extensive experiments on the MiniGrid environments using Q-learning and some popular deep RL algorithms.      "
                    },
                    {
                        "title": "Incremental Learning via Rate Reduction",
                        "abstract": "Current deep learning architectures suffer from catastrophic forgetting, a failure to retain knowledge of previously learned classes when incrementally trained on new classes. The fundamental roadblock faced by deep learning methods is that the models are optimized as \"black boxes,\" making it difficult to properly adjust the model parameters to preserve knowledge about previously seen data. To overcome the problem of catastrophic forgetting, we propose utilizing an alternative \"white box\" architecture derived from the principle of rate reduction, where each layer of the network is explicitly computed without back propagation. Under this paradigm, we demonstrate that, given a pretrained network and new data classes, our approach can provably construct a new network that emulates joint training with all past and new classes. Finally, our experiments show that our proposed learning algorithm observes significantly less decay in classification performance, outperforming state of the art methods on MNIST and CIFAR-10 by a large margin and justifying the use of \"white box\" algorithms for incremental learning even for sufficiently complex image data."
                    },
                    {
                        "title": "Machine Learning Algorithm for DNA-Based Molecular Ligands",
                        "abstract": "Recent research in DNA nanotechnology has demonstrated that biological substrates can be used for computing at a molecular level. However, in vitro demonstrations of DNA computations use preprogrammed, rule-based methods which lack the adaptability that may be essential in developing molecular systems that function in dynamic environments. Here, we introduce an in vitro molecular algorithm that \u2018learns\u2019 molecular models from training data, opening the possibility of \u2018machine learning\u2019 in wet molecular systems. Our algorithm enables enzymatic weight update by targeting internal loop structures in DNA and ensemble learning, based on the hypernetwork model. This novel approach allows massively parallel processing of DNA with enzymes for specific structural selection for learning in an iterative manner. We also introduce an intuitive method of DNA data construction to dramatically reduce the number of unique DNA sequences needed to cover the large search space of feature sets. By combining molecular computing and machine learning the proposed algorithm makes a step closer to developing molecular computing technologies for future access to more intelligent molecular systems."
                    },
                    {
                        "title": "Enzymatic Weight Update Algorithm for DNA-Based Molecular Learning",
                        "abstract": "Recent research in DNA nanotechnology has demonstrated that biological substrates can be used for computing at a molecular level. However, in vitro demonstrations of DNA computations use preprogrammed, rule-based methods which lack the adaptability that may be essential in developing molecular systems that function in dynamic environments. Here, we introduce an in vitro molecular algorithm that \u2018learns\u2019 molecular models from training data, opening the possibility of \u2018machine learning\u2019 in wet molecular systems. Our algorithm enables enzymatic weight update by targeting internal loop structures in DNA and ensemble learning, based on the hypernetwork model. This novel approach allows massively parallel processing of DNA with enzymes for specific structural selection for learning in an iterative manner. We also introduce an intuitive method of DNA data construction to dramatically reduce the number of unique DNA sequences needed to cover the large search space of feature sets. By combining molecular computing and machine learning the proposed algorithm makes a step closer to developing molecular computing technologies for future access to more intelligent molecular systems."
                    },
                    {
                        "title": "Intelligent DNA learning system with multiple iterations of weight update",
                        "abstract": "The limitations of complexity adaptability and scalability in intelligent molecular systems are addressed in this paper with a solid molecular framework for machine learning in vitro with DNA. We use molecular biology methods and enzymatic weight update to implement five iterations during training to classify between handwritten digit data. Although this is a work in progress, encouraging results are achieved in the test run of our training and test stages and fluorescent measurement of outcome visible for pattern classification."
                    },
                    {
                        "title": "Perception-Action-Learning System for Mobile Social-Service Robots Using Deep Learning",
                        "abstract": "    We introduce a novel perception-action-learning system for mobile social-service robots. The state-of-the-art deep learning techniques were incorporated into each module which significantly improves the performance in solving social service tasks. The system not only demonstrated fast and robust performance in a homelike environment but also achieved the highest score in the RoboCup2017@Home Social Standard Platform League (SSPL) held in Nagoya, Japan.   "
                    },
                    {
                        "title": "Robust Human Following by Deep Bayesian Trajectory Prediction for Home Service Robots",
                        "abstract": "The capability of following a person is crucial in service-oriented robots for human assistance and cooperation. Though a vast variety of following systems exist, they lack robustness against dynamic changes of the environment and relocating to continue following a lost target. Here we present a robust human following system that has the extendability to commercial service robot platforms having a RGB-D camera. The proposed framework integrates deep learning methods for perception and variational Bayesian techniques for trajectory prediction. Deep learning modules enable robots to accompany a person by detecting the target, learning the target and following while avoiding collision within the dynamic home environment. The variational Bayesian techniques robustly predict the trajectory of the target by empowering the following ability of the robot when target is lost. We experimentally demonstrate the capability of the deep Bayesian trajectory prediction method on real-time usage, following abilities, collision avoidance and trajectory prediction of the system. The proposed system was deployed at the RoboCup@Home 2017 Social Standard Platform League and successfully demonstrated its robust functions and smooth person following capability resulting in winning the 1st place."
                    },
                    {
                        "title": "In vivo Intelligent Agents with CRISPR-Cas 9 for Molecular Learning",
                        "abstract": "In vivo computing is rapidly evolving how science today is understood and engineered to create smarter agents, capable of co-existing in the complex and dynamic environments of biological systems. Here, we discuss the various breakthroughs in DNA nanotechnology and limitations of current in vivo methods where DNA nanodevices are used for drug delivery. We propose alternative approaches for in vivo agents, which avoid the limitations of unwanted immune responses or unclear location of nanodevices. We discuss the revolutionary CRISPR-Cas9 genome editing technique and how this could be used in conjunction with implementing molecular learning with mammalian cells and model organisms."
                    },
                    {
                        "title": "Random Ensemble Hypernetwork for Pattern Recognition with Enzymatic Weight Update",
                        "abstract": "Pattern recognition is a major division of machine learning which focuses on learning the patterns and regularities in data. It differs to that of pattern matching where only exact matches are found. However in the field of DNA computing, molecular pattern recognition has not been well established due to the lack of control of molecules in liquid state, instability and inaccuracy to solve such problems. Also the cost of designing the mass amount of DNA to represent data is a realistic issue. Here, we propose the random ensemble Hypernetwork as a model for pattern recognition in vitro with handwritten digit data encoded to DNA. For the manipulation of DNA in vitro, this molecular programming model is proposed to build a massively-parallel classification device, with the use of enzymatic weight update. Furthermore, a novel method of encoding vast amounts of data to DNA is introduced, also allowing the production of random hyperedges, a key difference to previous studies or computational implementations which only focused on fixed hyperedges."
                    },
                    {
                        "title": "Online Constructive Machine Learning with Molecular Hypernetworks in DNA Computing",
                        "abstract": "Various benefits of DNA computing such as programmability, immense data storage capacity and massively parallel processing have led to provide fundamental building blocks to motivated goals of building smart in vivo robots with implications in various applications. However, molecular machine learning has not yet been employed to solve more complex tasks such as pattern recognition, due to the lack of control of molecules in liquid state, instability and inaccuracy of in vitro manipulation to carry out learning algorithms experimentally. Here, a potential link between DNA computing and machine learning is proposed through the complementary action of a constructive machine learning model, the hypernetwork for DNA computation. Constructive DNA self-assembly and enzymatic weight update underlies the experimental scheme for wet lab implementation and online constructive learning of hypernetworks. We introduce both theoretical and practical details of the proposed model, present preliminary experimental results for examining the practical viability and discuss further applications."
                    },
                    {
                        "title": "Evolutionary Molecular Learning for Hand-written Digit Classification with Enzymatic Manipulation of DNA sequences",
                        "abstract": "Implementation of machine learning through the manipulation of DNA sequences is an advancing field in DNA Computing. Here, we propose a novel experimental scheme to demonstrate evolutionary molecular learning through the Hypernetwork to classify hand-written digits using enzymatic manipulation of DNA. This algorithm contains many operations, such as selection, amplification, elimination, which must be organized and designed for a realistic experimental approach. However, previous attempts to implement the Hypernetwork using DNA failed to address the time and experimental load required as well as costs to order enough DNA sequences to cover large data sets. It also lacked robustness in that the order of Hyperedges was fixed to 3 and the issue of scalability was not addressed they were only designed to classify 2-class digits. With the use of DNA cleaving enzymes, such as S1 nuclease we devise a method of reducing the total number of steps required to perform each operation and to allow 10-class digit classification in the same number of steps. With this theoretically sound but experimentally concise process, an improved method of employing the Hypernetwork is presented. This study introduces this experimental model and validates the additional step of enzymatic manipulation of DNA. With this a new avenue for exploiting the many benefits of DNA computing for the development of programmable and autonomous systems which function in biological organisms may be discovered. 2015\ub144 \ub3d9\uacc4\ud559\uc220\ubc1c\ud45c\ud68c \ub17c\ubb38\uc9d1"
                    },
                    {
                        "title": "Massively-Parallel Pattern Recognition through Evolutionary Molecular Hypernetwork in DNA Computing",
                        "abstract": "Handwritten digit classification is a classical problem in field of machine learning or artificial intelligence. Here, we propose a DNA computing method to explore the intersection between biology and computation. We implement of the Molecular Evolutionary Hypernetwork in vitro using DNA sequences as data. The pixels of the images are encoded into DNA sequences which are designed to create 3-order hyperedges to proceed as the training or test data to a random DNA library of 3-order hyperedges, ie. The Hypernetwork. This machine learning algorithm consists of the hybridization of training or test data to the Hypernetwork (random DNA library in the initial stage), selection and amplification of best matched sequences, isolation, classification, feedback and update of the Hypernetwork for the next learning phase. However, for this algorithm to be demonstrated experimentally, the creation of a 3-order hyperedge random library is crucial for the success of this learning process using a realistic sized dataset. This paper demonstrates the making of a random single-stranded DNA library through the use of ligation and isolation techniques, and validates the experimental steps required to implement the evolutionary molecular Hypernetwork in DNA Computing."
                    },
                    {
                        "title": "Molecular Computing Simulation of Cognitive Anagram Solving",
                        "abstract": "An anagram is a form of word play to find a new word from a set of given alphabet letters. Good human anagram solvers use the strategy of bigrams. They explore a constraint satisfaction network in parallel and answers consequently pop out quickly. In this paper, we propose a molecular computational algorithm using the same process as this. We encoded letters into DNA sequences and made bigrams and then words by connecting the letter sequences. From letters and bigrams, we performed DNA hybridization, ligation, gel electrophoresis and finally, extraction and separation to extract bigrams. From the matched bigrams and words, we performed the four molecular operations again to distinguish between right and wrong results. Experimental results show that our molecular computer can identify cor rect answers and incorrect answers. Our work shows a new possibility for modeling the cognitive and parallel thinking process of a human."
                    },
                    {
                        "title": "Molecular Evolutionary Learning of DNA Hypernetworks for Hand-written Digit Classification",
                        "abstract": "An intersection between biology and computation is explored by classifying handwritten digits using DNA computing. To do this, we propose employing the Molecular Evolutionary Hypernetwork [1] in vitro using specifically encoded DNA as pixels of images. This machine learning algorithm consists of various stages; matching of training or test data with the Hypernetwork (starting with a random single-stranded library of 3-order hyperedges), selection and amplification of similar sequences, isolation, classification, feedback and update of the Hypernetwork used in the next round of training. A key limiting factor of this design is the issue of scalability, as a very large dataset would need to be provided for demonstrating the learning process [2]. To address this, and for the verification of each experimental step, we design, create and test a random single-stranded DNA library using various techniques, which provides a base of larger scale experiments demonstrating molecular learning."
                    }
                ]
            },
            "49c5e4f6-3d55-4952-9eba-8c0d02144c3e": {
                "pk": "49c5e4f6-3d55-4952-9eba-8c0d02144c3e",
                "name": "J. Zico Kolter",
                "collaborators": [
                    "Zhengyang Geng",
                    "Dylan Sam",
                    "Ashwini Pokle",
                    "Samuel Sokota",
                    "Yutong He",
                    "Naoki Murata",
                    "Yiding Jiang",
                    "J. Williams",
                    "Yuki Mitsufuji",
                    "Ruslan Salakhutdinov"
                ],
                "domain": [
                    "Generative Models",
                    "Reinforcement Learning",
                    "Machine Learning",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "Consistency Models Made Easy",
                        "abstract": "Consistency models (CMs) offer faster sampling than traditional diffusion models, but their training is resource-intensive. For example, as of 2024, training a state-of-the-art CM on CIFAR-10 takes one week on 8 GPUs. In this work, we propose an effective scheme for training CMs that largely improves the efficiency of building such models. Specifically, by expressing CM trajectories via a particular differential equation, we argue that diffusion models can be viewed as a special case of CMs. We can thus fine-tune a consistency model starting from a pretrained diffusion model and progressively approximate the full consistency condition to stronger degrees over the training process. Our resulting method, which we term Easy Consistency Tuning (ECT), achieves vastly reduced training times while improving upon the quality of previous methods: for example, ECT achieves a 2-step FID of 2.73 on CIFAR10 within 1 hour on a single A100 GPU, matching Consistency Distillation trained for hundreds of GPU hours. Owing to this computational efficiency, we investigate the scaling laws of CMs under ECT, showing that they obey the classic power law scaling, hinting at their ability to improve efficiency and performance at larger scales. Our code (https://github.com/locuslab/ect) is publicly available, making CMs more accessible to the broader community."
                    },
                    {
                        "title": "One-Step Diffusion Distillation through Score Implicit Matching",
                        "abstract": "Despite their strong performances on many generative tasks, diffusion models require a large number of sampling steps in order to generate realistic samples. This has motivated the community to develop effective methods to distill pre-trained diffusion models into more efficient models, but these methods still typically require few-step inference or perform substantially worse than the underlying model. In this paper, we present Score Implicit Matching (SIM) a new approach to distilling pre-trained diffusion models into single-step generator models, while maintaining almost the same sample generation ability as the original model as well as being data-free with no need of training samples for distillation. The method rests upon the fact that, although the traditional score-based loss is intractable to minimize for generator models, under certain conditions we can efficiently compute the gradients for a wide class of score-based divergences between a diffusion model and a generator. SIM shows strong empirical performances for one-step generators: on the CIFAR10 dataset, it achieves an FID of 2.06 for unconditional generation and 1.96 for class-conditional generation. Moreover, by applying SIM to a leading transformer-based diffusion model, we distill a single-step generator for text-to-image (T2I) generation that attains an aesthetic score of 6.42 with no performance decline over the original multi-step counterpart, clearly outperforming the other one-step generators including SDXL-TURBO of 5.33, SDXL-LIGHTNING of 5.34 and HYPER-SDXL of 5.85. We will release this industry-ready one-step transformer-based T2I generator along with this paper."
                    },
                    {
                        "title": "Computing Low-Entropy Couplings for Large-Support Distributions",
                        "abstract": "Minimum-entropy coupling (MEC) -- the process of finding a joint distribution with minimum entropy for given marginals -- has applications in areas such as causality and steganography. However, existing algorithms are either computationally intractable for large-support distributions or limited to specific distribution types and sensitive to hyperparameter choices. This work addresses these limitations by unifying a prior family of iterative MEC (IMEC) approaches into a generalized partition-based formalism. From this framework, we derive a novel IMEC algorithm called ARIMEC, capable of handling arbitrary discrete distributions, and introduce a method to make IMEC robust to suboptimal hyperparameter settings. These innovations facilitate the application of IMEC to high-throughput steganography with language models, among other settings. Our codebase is available at https://github.com/ssokota/mec ."
                    },
                    {
                        "title": "Automated Black-box Prompt Engineering for Personalized Text-to-Image Generation",
                        "abstract": "Prompt engineering is effective for controlling the output of text-to-image (T2I) generative models, but it is also laborious due to the need for manually crafted prompts. This challenge has spurred the development of algorithms for automated prompt generation. However, these methods often struggle with transferability across T2I models, require white-box access to the underlying model, and produce non-intuitive prompts. In this work, we introduce PRISM, an algorithm that automatically identifies human-interpretable and transferable prompts that can effectively generate desired concepts given only black-box access to T2I models. Inspired by large language model (LLM) jailbreaking, PRISM leverages the in-context learning ability of LLMs to iteratively refine the candidate prompts distribution for given reference images. Our experiments demonstrate the versatility and effectiveness of PRISM in generating accurate prompts for objects, styles and images across multiple T2I models, including Stable Diffusion, DALL-E, and Midjourney."
                    },
                    {
                        "title": "FUSE-ing Language Models: Zero-Shot Adapter Discovery for Prompt Optimization Across Tokenizers",
                        "abstract": "The widespread use of large language models has resulted in a multitude of tokenizers and embedding spaces, making knowledge transfer in prompt discovery tasks difficult. In this work, we propose FUSE (Flexible Unification of Semantic Embeddings), an inexpensive approach to approximating an adapter layer that maps from one model's textual embedding space to another, even across different tokenizers. We introduce a third-order tensor-based representation of a model's embedding space that aligns semantic embeddings that have been split apart by different tokenizers, and use this representation to derive an approximation of the gradient of one model's outputs with respect to another model's embedding space. We show the efficacy of our approach via multi-objective optimization over vision-language and causal language models for image captioning and sentiment-based image captioning."
                    },
                    {
                        "title": "Massive Activations in Large Language Models",
                        "abstract": "We observe an empirical phenomenon in Large Language Models (LLMs) -- very few activations exhibit significantly larger values than others (e.g., 100,000 times larger). We call them massive activations. First, we demonstrate the widespread existence of massive activations across various LLMs and characterize their locations. Second, we find their values largely stay constant regardless of the input, and they function as indispensable bias terms in LLMs. Third, these massive activations lead to the concentration of attention probabilities to their corresponding tokens, and further, implicit bias terms in the self-attention output. Last, we also study massive activations in Vision Transformers. Code is available at https://github.com/locuslab/massive-activations."
                    },
                    {
                        "title": "Bayesian Neural Networks with Domain Knowledge Priors",
                        "abstract": "Bayesian neural networks (BNNs) have recently gained popularity due to their ability to quantify model uncertainty. However, specifying a prior for BNNs that captures relevant domain knowledge is often extremely challenging. In this work, we propose a framework for integrating general forms of domain knowledge (i.e., any knowledge that can be represented by a loss function) into a BNN prior through variational inference, while enabling computationally efficient posterior inference and sampling. Specifically, our approach results in a prior over neural network weights that assigns high probability mass to models that better align with our domain knowledge, leading to posterior samples that also exhibit this behavior. We show that BNNs using our proposed domain knowledge priors outperform those with standard priors (e.g., isotropic Gaussian, Gaussian process), successfully incorporating diverse types of prior information such as fairness, physics rules, and healthcare knowledge and achieving better predictive performance. We also present techniques for transferring the learned priors across different model architectures, demonstrating their broad utility across various settings."
                    },
                    {
                        "title": "Transformers to SSMs: Distilling Quadratic Knowledge to Subquadratic Models",
                        "abstract": "Transformer architectures have become a dominant paradigm for domains like language modeling but suffer in many inference settings due to their quadratic-time self-attention. Recently proposed subquadratic architectures, such as Mamba, have shown promise, but have been pretrained with substantially less computational resources than the strongest Transformer models. In this work, we present a method that is able to distill a pretrained Transformer architecture into alternative architectures such as state space models (SSMs). The key idea to our approach is that we can view both Transformers and SSMs as applying different forms of mixing matrices over the token sequences. We can thus progressively distill the Transformer architecture by matching different degrees of granularity in the SSM: first matching the mixing matrices themselves, then the hidden units at each block, and finally the end-to-end predictions. Our method, called MOHAWK, is able to distill a Mamba-2 variant based on the Phi-1.5 architecture (Phi-Mamba) using only 3B tokens and a hybrid version (Hybrid Phi-Mamba) using 5B tokens. Despite using less than 1% of the training data typically used to train models from scratch, Phi-Mamba boasts substantially stronger performance compared to all past open-source non-Transformer models. MOHAWK allows models like SSMs to leverage computational resources invested in training Transformer-based architectures, highlighting a new avenue for building such models."
                    },
                    {
                        "title": "AcceleratedLiNGAM: Learning Causal DAGs at the speed of GPUs",
                        "abstract": "Existing causal discovery methods based on combinatorial optimization or search are slow, prohibiting their application on large-scale datasets. In response, more recent methods attempt to address this limitation by formulating causal discovery as structure learning with continuous optimization but such approaches thus far provide no statistical guarantees. In this paper, we show that by efficiently parallelizing existing causal discovery methods, we can in fact scale them to thousands of dimensions, making them practical for substantially larger-scale problems. In particular, we parallelize the LiNGAM method, which is quadratic in the number of variables, obtaining up to a 32-fold speed-up on benchmark datasets when compared with existing sequential implementations. Specifically, we focus on the causal ordering subprocedure in DirectLiNGAM and implement GPU kernels to accelerate it. This allows us to apply DirectLiNGAM to causal inference on large-scale gene expression data with genetic interventions yielding competitive results compared with specialized continuous optimization methods, and Var-LiNGAM for causal discovery on U.S. stock data."
                    },
                    {
                        "title": "Learning with Explanation Constraints",
                        "abstract": "As larger deep learning models are hard to interpret, there has been a recent focus on generating explanations of these black-box models. In contrast, we may have apriori explanations of how models should behave. In this paper, we formalize this notion as learning from explanation constraints and provide a learning theoretic framework to analyze how such explanations can improve the learning of our models. One may naturally ask,\"When would these explanations be helpful?\"Our first key contribution addresses this question via a class of models that satisfies these explanation constraints in expectation over new data. We provide a characterization of the benefits of these models (in terms of the reduction of their Rademacher complexities) for a canonical class of explanations given by gradient information in the settings of both linear models and two layer neural networks. In addition, we provide an algorithmic solution for our framework, via a variational approximation that achieves better performance and satisfies these constraints more frequently, when compared to simpler augmented Lagrangian methods to incorporate these explanations. We demonstrate the benefits of our approach over a large array of synthetic and real-world experiments."
                    },
                    {
                        "title": "Sinkhorn-Flow: Predicting Probability Mass Flow in Dynamical Systems Using Optimal Transport",
                        "abstract": "Predicting how distributions over discrete variables vary over time is a common task in time series forecasting. But whereas most approaches focus on merely predicting the distribution at subsequent time steps, a crucial piece of information in many settings is to determine how this probability mass flows between the different elements over time. We propose a new approach to predicting such mass flow over time using optimal transport. Specifically, we propose a generic approach to predicting transport matrices in end-to-end deep learning systems, replacing the standard softmax operation with Sinkhorn iterations. We apply our approach to the task of predicting how communities will evolve over time in social network settings, and show that the approach improves substantially over alternative prediction methods. We specifically highlight results on the task of predicting faction evolution in Ukrainian parliamentary voting."
                    },
                    {
                        "title": "One-Step Diffusion Distillation via Deep Equilibrium Models",
                        "abstract": "Diffusion models excel at producing high-quality samples but naively require hundreds of iterations, prompting multiple attempts to distill the generation process into a faster network. However, many existing approaches suffer from a variety of challenges: the process for distillation training can be complex, often requiring multiple training stages, and the resulting models perform poorly when utilized in single-step generative applications. In this paper, we introduce a simple yet effective means of distilling diffusion models directly from initial noise to the resulting image. Of particular importance to our approach is to leverage a new Deep Equilibrium (DEQ) model as the distilled architecture: the Generative Equilibrium Transformer (GET). Our method enables fully offline training with just noise/image pairs from the diffusion model while achieving superior performance compared to existing one-step methods on comparable training budgets. We demonstrate that the DEQ architecture is crucial to this capability, as GET matches a $5\\times$ larger ViT in terms of FID scores while striking a critical balance of computational cost and image quality. Code, checkpoints, and datasets are available."
                    },
                    {
                        "title": "Abstracting Imperfect Information Away from Two-Player Zero-Sum Games",
                        "abstract": "In their seminal work, Nayyar et al. (2013) showed that imperfect information can be abstracted away from common-payoff games by having players publicly announce their policies as they play. This insight underpins sound solvers and decision-time planning algorithms for common-payoff games. Unfortunately, a naive application of the same insight to two-player zero-sum games fails because Nash equilibria of the game with public policy announcements may not correspond to Nash equilibria of the original game. As a consequence, existing sound decision-time planning algorithms require complicated additional mechanisms that have unappealing properties. The main contribution of this work is showing that certain regularized equilibria do not possess the aforementioned non-correspondence problem -- thus, computing them can be treated as perfect-information problems. Because these regularized equilibria can be made arbitrarily close to Nash equilibria, our result opens the door to a new perspective to solving two-player zero-sum games and yields a simplified framework for decision-time planning in two-player zero-sum games, void of the unappealing properties that plague existing decision-time planning approaches."
                    },
                    {
                        "title": "Manifold Preserving Guided Diffusion",
                        "abstract": "Despite the recent advancements, conditional image generation still faces challenges of cost, generalizability, and the need for task-specific training. In this paper, we propose Manifold Preserving Guided Diffusion (MPGD), a training-free conditional generation framework that leverages pretrained diffusion models and off-the-shelf neural networks with minimal additional inference cost for a broad range of tasks. Specifically, we leverage the manifold hypothesis to refine the guided diffusion steps and introduce a shortcut algorithm in the process. We then propose two methods for on-manifold training-free guidance using pre-trained autoencoders and demonstrate that our shortcut inherently preserves the manifolds when applied to latent diffusion models. Our experiments show that MPGD is efficient and effective for solving a variety of conditional generation applications in low-compute settings, and can consistently offer up to 3.8x speed-ups with the same number of diffusion steps while maintaining high sample quality compared to the baselines."
                    },
                    {
                        "title": "On the Importance of Exploration for Generalization in Reinforcement Learning",
                        "abstract": "Existing approaches for improving generalization in deep reinforcement learning (RL) have mostly focused on representation learning, neglecting RL-specific aspects such as exploration. We hypothesize that the agent's exploration strategy plays a key role in its ability to generalize to new environments. Through a series of experiments in a tabular contextual MDP, we show that exploration is helpful not only for efficiently finding the optimal policy for the training environments but also for acquiring knowledge that helps decision making in unseen environments. Based on these observations, we propose EDE: Exploration via Distributional Ensemble, a method that encourages exploration of states with high epistemic uncertainty through an ensemble of Q-value distributions. Our algorithm is the first value-based approach to achieve state-of-the-art on both Procgen and Crafter, two benchmarks for generalization in RL with high-dimensional observations. The open-sourced implementation can be found at https://github.com/facebookresearch/ede ."
                    },
                    {
                        "title": "Mimetic Initialization of Self-Attention Layers",
                        "abstract": "It is notoriously difficult to train Transformers on small datasets; typically, large pre-trained models are instead used as the starting point. We explore the weights of such pre-trained Transformers (particularly for vision) to attempt to find reasons for this discrepancy. Surprisingly, we find that simply initializing the weights of self-attention layers so that they\"look\"more like their pre-trained counterparts allows us to train vanilla Transformers faster and to higher final accuracies, particularly on vision tasks such as CIFAR-10 and ImageNet classification, where we see gains in accuracy of over 5% and 4%, respectively. Our initialization scheme is closed form, learning-free, and very simple: we set the product of the query and key weights to be approximately the identity, and the product of the value and projection weights to approximately the negative identity. As this mimics the patterns we saw in pre-trained Transformers, we call the technique\"mimetic initialization\"."
                    },
                    {
                        "title": "Universal and Transferable Adversarial Attacks on Aligned Language Models",
                        "abstract": "Because\"out-of-the-box\"large language models are capable of generating a great deal of objectionable content, recent work has focused on aligning these models in an attempt to prevent undesirable generation. While there has been some success at circumventing these measures -- so-called\"jailbreaks\"against LLMs -- these attacks have required significant human ingenuity and are brittle in practice. In this paper, we propose a simple and effective attack method that causes aligned language models to generate objectionable behaviors. Specifically, our approach finds a suffix that, when attached to a wide range of queries for an LLM to produce objectionable content, aims to maximize the probability that the model produces an affirmative response (rather than refusing to answer). However, instead of relying on manual engineering, our approach automatically produces these adversarial suffixes by a combination of greedy and gradient-based search techniques, and also improves over past automatic prompt generation methods. Surprisingly, we find that the adversarial prompts generated by our approach are quite transferable, including to black-box, publicly released LLMs. Specifically, we train an adversarial attack suffix on multiple prompts (i.e., queries asking for many different types of objectionable content), as well as multiple models (in our case, Vicuna-7B and 13B). When doing so, the resulting attack suffix is able to induce objectionable content in the public interfaces to ChatGPT, Bard, and Claude, as well as open source LLMs such as LLaMA-2-Chat, Pythia, Falcon, and others. In total, this work significantly advances the state-of-the-art in adversarial attacks against aligned language models, raising important questions about how such systems can be prevented from producing objectionable information. Code is available at github.com/llm-attacks/llm-attacks."
                    },
                    {
                        "title": "Practical Critic Gradient based Actor Critic for On-Policy Reinforcement Learning",
                        "abstract": "On-policy reinforcement learning algorithms have been shown to be remarkably efficient at learning policies for continuous control robotics tasks. They are highly parallelizable and hence have benefited tremendously from the recent rise in GPU based parallel simulators. The most widely used on-policy reinforcement learning algorithm is proximal policy optimization (PPO) which was introduced in 2017 and was designed for a somewhat different setting with CPU based serial or less parallelizable simulators. However, suprisingly, it has maintained dominance even on tasks based on the highly parallelizable simulators of today. In this paper, we show that a different class of on-policy algorithms based on estimating the policy gradient using the critic-action gradients are better suited when using highly parallelizable simulators. The primary issues for these algorithms arise from the lack of diversity of the on-policy experiences used for the updates and the instabilities arising from the interaction between the biased critic gradients and the rapidly changing policy distribution. We address the former by simply increasing the number of parallel simulation runs (thanks to the GPU based simulators) along with an appropriate schedule on the policy entropy to ensure diversity of samples. We address the latter by adding a policy averaging step and value averaging step (as in off-policy methods). With these modifications, we observe that the critic gradient based on-policy method (CGAC) consistently achieves higher episode returns compared with existing baselines. Furthermore, in environments with high dimensional action space, CGAC also trains much faster (in wall-clock time) than the corresponding baselines."
                    },
                    {
                        "title": "Projected Off-Policy Q-Learning (POP-QL) for Stabilizing Offline Reinforcement Learning",
                        "abstract": "A key problem in off-policy Reinforcement Learning (RL) is the mismatch, or distribution shift, between the dataset and the distribution over states and actions visited by the learned policy. This problem is exacerbated in the fully offline setting. The main approach to correct this shift has been through importance sampling, which leads to high-variance gradients. Other approaches, such as conservatism or behavior-regularization, regularize the policy at the cost of performance. In this paper, we propose a new approach for stable off-policy Q-Learning. Our method, Projected Off-Policy Q-Learning (POP-QL), is a novel actor-critic algorithm that simultaneously reweights off-policy samples and constrains the policy to prevent divergence and reduce value-approximation error. In our experiments, POP-QL not only shows competitive performance on standard benchmarks, but also out-performs competing methods in tasks where the data-collection policy is significantly sub-optimal."
                    }
                ]
            }
        }
    },
    "2405.17580": {
        "paper_data": {
            "title": "Mixed Dynamics In Linear Networks: Unifying the Lazy and Active Regimes",
            "url": "http://arxiv.org/abs/2405.17580v1",
            "arxiv_id": "2405.17580",
            "authors": [
                "Zhenfeng Tu",
                "Santiago Aranguri",
                "Arthur Jacot"
            ],
            "abstract": "The training dynamics of linear networks are well studied in two distinct setups: the lazy regime and balanced/active regime, depending on the initialization and width of the network. We provide a surprisingly simple unyfing formula for the evolution of the learned matrix that contains as special cases both lazy and balanced regimes but also a mixed regime in between the two. In the mixed regime, a part of the network is lazy while the other is balanced. More precisely the network is lazy along singular values that are below a certain threshold and balanced along those that are above the same threshold. At initialization, all singular values are lazy, allowing for the network to align itself with the task, so that later in time, when some of the singular value cross the threshold and become active they will converge rapidly (convergence in the balanced regime is notoriously difficult in the absence of alignment). The mixed regime is the `best of both worlds': it converges from any random initialization (in contrast to balanced dynamics which require special initialization), and has a low rank bias (absent in the lazy dynamics). This allows us to prove an almost complete phase diagram of training behavior as a function of the variance at initialization and the width, for a MSE training task.",
            "introduction": "   1 Introduction  Whether in linear networks or nonlinear ones, there has been a lot of interest in the distinction between the lazy regime jacot2018neural  and the active regime Chizat2018 ; Rotskoff2018 ; chizat2018note ; yang2020feature ; bordelon2022deep_MF  as the number of neurons grows towards infinity. In the lazy regime the training dynamics become linear, so that they can be easily described in terms of the Neural Tangent Kernel (NTK) jacot2018neural ; exact_arora2019 ; yang2020NTK ; liu2020_NTK-PL-inequ , while the active regime exhibits complex nonlinear dynamics. While our understanding of the active regime remains much more limited, it appears to be characterized by the emergence of feature learninggeiger2019disentangling ; yang2020feature , and of a form of sparsity advani_2017_independent_diag ; bach2017_F1_norm ; abbe_2021_staircase ; abbe_2022_staircase  (the type of sparsity observed depends on the network type bach2017_F1_norm ; dai_2021_repres_cost_DLN ; jacot_2022_BN_rank_short ; jacot_2023_bottleneck2 , but we will focus on fully-connected linear networks which exhibit a rank sparsity in the learned linear map arora_2018_depth_speed_Lp ; li2020towards ; jacot-2021-DLN-Saddle ; wang_2023_bias_SGD_L2 ) which are both absent in the lazy regime .   Note that even though it is common to talk of the \u2018the\u2019 active regime, we do not know yet whether there is only one or multiple active regimes. Indeed the term active regime is usually used to describe any regime that differs from the lazy regime and exhibit some form of feature learning. Though we do not have an complete understanding of where the lazy regimes ends and the active regime(s) start, we know that the lazy regime requires extreme overparametrization (a large number of neurons in comparison to the number of datapoints) Allen-Zhu2018 ; Du2019 , a \u2018large\u2019 initialization of the weights chizat2018note , a small learning rate, and early stopping when using a cross-entropy loss or weight decay . Indeed, active regimes have been observed by breaking either of these requirements: taking limits with mild or no overparametrization arous2022summary_stats , taking smaller or even vanishingly small initializations li2020towards ; jacot-2021-DLN-Saddle , using large learning rates lewkowycz_2020_large_lr  or SGD Pesme_2021_SGD_bias_diag_nets ; wang_2023_bias_SGD_L2 , or studying the late training dynamics with the cross-entropy loss Ji_2018_directional ; chizat_2020_implicit_bias  or weight decay lewkowycz2020_L2_training ; Ongie_2020_repres_bounded_norm_shallow_ReLU_net ; jacot_2022_L2_reformulation ; jacot_2022_BN_rank . Though each of these can lead to active regimes with significantly different dynamics, they often lead to similar types of feature learning and sparsity.   In this paper, we study this transition in the context of linear networks and focus mainly on the effects of the width w\ud835\udc64witalic_w and the variance of the weights at initialization \u03c32superscript\ud835\udf0e2\\sigma^{2}italic_\u03c3 start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT, and give a precise and almost complete phase diagram, showing the transitions between lazy and active regimes. In this setting, we will show that there typically is only \u2018one\u2019 active regime, which is the same (up to approximation) as the already well-studied balanced regime saxe_2014_exact ; arora_2018_depth_speed_Lp ; arora_2019_matrix_factorization .   But our result also paint a more subtle picture than the lazy/active dichotomy. We propose a more granular approach, where at a certain time some part of the network can be in the lazy regime, while others are in the active or balanced regime. More precisely the network is lazy along the singular values of the matrix represented by the network that are smaller than \u03c32\u2062wsuperscript\ud835\udf0e2\ud835\udc64\\sigma^{2}witalic_\u03c3 start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_w, and in the active regime along the singular values larger than",
            "references": [
                {
                    "title": "Exact learning dynamics of deep linear networks with prior knowledge",
                    "abstract": "Learning in deep neural networks is known to depend critically on the knowledge embedded in the initial network weights. However, few theoretical results have precisely linked prior knowledge to learning dynamics. Here we derive exact solutions to the dynamics of learning with rich prior knowledge in deep linear networks by generalising Fukumizu\u2019s matrix Riccati solution (Fukumizu 1998 Gen 1 1E\u201303). We obtain explicit expressions for the evolving network function, hidden representational similarity, and neural tangent kernel over training for a broad class of initialisations and tasks. The expressions reveal a class of task-independent initialisations that radically alter learning dynamics from slow non-linear dynamics to fast exponential trajectories while converging to a global optimum with identical representational similarity, dissociating learning trajectories from the structure of initial internal representations. We characterise how network weights dynamically align with task structure, rigorously justifying why previous solutions successfully described learning from small initial weights without incorporating their fine-scale structure. Finally, we discuss the implications of these findings for continual learning, reversal learning and learning of structured knowledge. Taken together, our results provide a mathematical toolkit for understanding the impact of prior knowledge on deep learning."
                },
                {
                    "title": "How Over-Parameterization Slows Down Gradient Descent in Matrix Sensing: The Curses of Symmetry and Initialization",
                    "abstract": "This paper rigorously shows how over-parameterization changes the convergence behaviors of gradient descent (GD) for the matrix sensing problem, where the goal is to recover an unknown low-rank ground-truth matrix from near-isotropic linear measurements. First, we consider the symmetric setting with the symmetric parameterization where $M^* \\in \\mathbb{R}^{n \\times n}$ is a positive semi-definite unknown matrix of rank $r \\ll n$, and one uses a symmetric parameterization $XX^\\top$ to learn $M^*$. Here $X \\in \\mathbb{R}^{n \\times k}$ with $k>r$ is the factor matrix. We give a novel $\\Omega (1/T^2)$ lower bound of randomly initialized GD for the over-parameterized case ($k>r$) where $T$ is the number of iterations. This is in stark contrast to the exact-parameterization scenario ($k=r$) where the convergence rate is $\\exp (-\\Omega (T))$. Next, we study asymmetric setting where $M^* \\in \\mathbb{R}^{n_1 \\times n_2}$ is the unknown matrix of rank $r \\ll \\min\\{n_1,n_2\\}$, and one uses an asymmetric parameterization $FG^\\top$ to learn $M^*$ where $F \\in \\mathbb{R}^{n_1 \\times k}$ and $G \\in \\mathbb{R}^{n_2 \\times k}$. Building on prior work, we give a global exact convergence result of randomly initialized GD for the exact-parameterization case ($k=r$) with an $\\exp (-\\Omega(T))$ rate. Furthermore, we give the first global exact convergence result for the over-parameterization case ($k>r$) with an $\\exp(-\\Omega(\\alpha^2 T))$ rate where $\\alpha$ is the initialization scale. This linear convergence result in the over-parameterization case is especially significant because one can apply the asymmetric parameterization to the symmetric setting to speed up from $\\Omega (1/T^2)$ to linear convergence. On the other hand, we propose a novel method that only modifies one step of GD and obtains a convergence rate independent of $\\alpha$, recovering the rate in the exact-parameterization case."
                },
                {
                    "title": "Bottleneck Structure in Learned Features: Low-Dimension vs Regularity Tradeoff",
                    "abstract": "Previous work has shown that DNNs with large depth $L$ and $L_{2}$-regularization are biased towards learning low-dimensional representations of the inputs, which can be interpreted as minimizing a notion of rank $R^{(0)}(f)$ of the learned function $f$, conjectured to be the Bottleneck rank. We compute finite depth corrections to this result, revealing a measure $R^{(1)}$ of regularity which bounds the pseudo-determinant of the Jacobian $\\left|Jf(x)\\right|_{+}$ and is subadditive under composition and addition. This formalizes a balance between learning low-dimensional representations and minimizing complexity/irregularity in the feature maps, allowing the network to learn the `right' inner dimension. Finally, we prove the conjectured bottleneck structure in the learned features as $L\\to\\infty$: for large depths, almost all hidden representations are approximately $R^{(0)}(f)$-dimensional, and almost all weight matrices $W_{\\ell}$ have $R^{(0)}(f)$ singular values close to 1 while the others are $O(L^{-\\frac{1}{2}})$. Interestingly, the use of large learning rates is required to guarantee an order $O(L)$ NTK which in turns guarantees infinite depth convergence of the representations of almost all layers."
                },
                {
                    "title": "Implicit bias of SGD in L2-regularized linear DNNs: One-way jumps from high to low rank",
                    "abstract": "The $L_{2}$-regularized loss of Deep Linear Networks (DLNs) with more than one hidden layers has multiple local minima, corresponding to matrices with different ranks. In tasks such as matrix completion, the goal is to converge to the local minimum with the smallest rank that still fits the training data. While rank-underestimating minima can be avoided since they do not fit the data, GD might get stuck at rank-overestimating minima. We show that with SGD, there is always a probability to jump from a higher rank minimum to a lower rank one, but the probability of jumping back is zero. More precisely, we define a sequence of sets $B_{1}\\subset B_{2}\\subset\\cdots\\subset B_{R}$ so that $B_{r}$ contains all minima of rank $r$ or less (and not more) that are absorbing for small enough ridge parameters $\\lambda$ and learning rates $\\eta$: SGD has prob. 0 of leaving $B_{r}$, and from any starting point there is a non-zero prob. for SGD to go in $B_{r}$."
                },
                {
                    "title": "Understanding Incremental Learning of Gradient Descent: A Fine-grained Analysis of Matrix Sensing",
                    "abstract": "It is believed that Gradient Descent (GD) induces an implicit bias towards good generalization in training machine learning models. This paper provides a fine-grained analysis of the dynamics of GD for the matrix sensing problem, whose goal is to recover a low-rank ground-truth matrix from near-isotropic linear measurements. It is shown that GD with small initialization behaves similarly to the greedy low-rank learning heuristics (Li et al., 2020) and follows an incremental learning procedure (Gissin et al., 2019): GD sequentially learns solutions with increasing ranks until it recovers the ground truth matrix. Compared to existing works which only analyze the first learning phase for rank-1 solutions, our result provides characterizations for the whole learning process. Moreover, besides the over-parameterized regime that many prior works focused on, our analysis of the incremental learning procedure also applies to the under-parameterized regime. Finally, we conduct numerical experiments to confirm our theoretical findings."
                },
                {
                    "title": "Infinite-width limit of deep linear neural networks",
                    "abstract": "This paper studies the infinite\u2010width limit of deep linear neural networks (NNs) initialized with random parameters. We obtain that, when the number of parameters diverges, the training dynamics converge (in a precise sense) to the dynamics obtained from a gradient descent on an infinitely wide deterministic linear NN. Moreover, even if the weights remain random, we get their precise law along the training dynamics, and prove a quantitative convergence result of the linear predictor in terms of the number of parameters. We finally study the continuous\u2010time limit obtained for infinitely wide linear NNs and show that the linear predictors of the NN converge at an exponential rate to the minimal \u2010norm minimizer of the\u00a0risk."
                },
                {
                    "title": "Implicit Bias of Large Depth Networks: a Notion of Rank for Nonlinear Functions",
                    "abstract": "We show that the representation cost of fully connected neural networks with homogeneous nonlinearities - which describes the implicit bias in function space of networks with $L_2$-regularization or with losses such as the cross-entropy - converges as the depth of the network goes to infinity to a notion of rank over nonlinear functions. We then inquire under which conditions the global minima of the loss recover the `true' rank of the data: we show that for too large depths the global minimum will be approximately rank 1 (underestimating the rank); we then argue that there is a range of depths which grows with the number of datapoints where the true rank is recovered. Finally, we discuss the effect of the rank of a classifier on the topology of the resulting class boundaries and show that autoencoders with optimal nonlinear rank are naturally denoising."
                },
                {
                    "title": "High\u2010dimensional limit theorems for SGD: Effective dynamics and critical scaling",
                    "abstract": "We study the scaling limits of stochastic gradient descent (SGD) with constant step\u2010size in the high\u2010dimensional regime. We prove limit theorems for the trajectories of summary statistics (i.e., finite\u2010dimensional functions) of SGD as the dimension goes to infinity. Our approach allows one to choose the summary statistics that are tracked, the initialization, and the step\u2010size. It yields both ballistic (ODE) and diffusive (SDE) limits, with the limit depending dramatically on the former choices. We show a critical scaling regime for the step\u2010size, below which the effective ballistic dynamics matches gradient flow for the population loss, but at which, a new correction term appears which changes the phase diagram. About the fixed points of this effective dynamics, the corresponding diffusive limits can be quite complex and even degenerate. We demonstrate our approach on popular examples including estimation for spiked matrix and tensor models and classification via two\u2010layer networks for binary and XOR\u2010type Gaussian mixture models. These examples exhibit surprising phenomena including multimodal timescales to convergence as well as convergence to sub\u2010optimal solutions with probability bounded away from zero from random (e.g., Gaussian) initializations. At the same time, we demonstrate the benefit of overparametrization by showing that the latter probability goes to zero as the second layer width grows."
                },
                {
                    "title": "Feature Learning in L2-regularized DNNs: Attraction/Repulsion and Sparsity",
                    "abstract": "We study the loss surface of DNNs with $L_{2}$ regularization. We show that the loss in terms of the parameters can be reformulated into a loss in terms of the layerwise activations $Z_{\\ell}$ of the training set. This reformulation reveals the dynamics behind feature learning: each hidden representations $Z_{\\ell}$ are optimal w.r.t. to an attraction/repulsion problem and interpolate between the input and output representations, keeping as little information from the input as necessary to construct the activation of the next layer. For positively homogeneous non-linearities, the loss can be further reformulated in terms of the covariances of the hidden representations, which takes the form of a partially convex optimization over a convex cone. This second reformulation allows us to prove a sparsity result for homogeneous DNNs: any local minimum of the $L_{2}$-regularized loss can be achieved with at most $N(N+1)$ neurons in each hidden layer (where $N$ is the size of the training set). We show that this bound is tight by giving an example of a local minimum that requires $N^{2}/4$ hidden neurons. But we also observe numerically that in more traditional settings much less than $N^{2}$ neurons are required to reach the minima."
                },
                {
                    "title": "Self-consistent dynamical field theory of kernel evolution in wide neural networks",
                    "abstract": "We analyze feature learning in infinite-width neural networks trained with gradient flow through a self-consistent dynamical field theory. We construct a collection of deterministic dynamical order parameters which are inner-product kernels for hidden unit activations and gradients in each layer at pairs of time points, providing a reduced description of network activity through training. These kernel order parameters collectively define the hidden layer activation distribution, the evolution of the neural tangent kernel (NTK), and consequently, output predictions. We show that the field theory derivation recovers the recursive stochastic process of infinite-width feature learning networks obtained by Yang and Hu with tensor programs. For deep linear networks, these kernels satisfy a set of algebraic matrix equations. For nonlinear networks, we provide an alternating sampling procedure to self-consistently solve for the kernel order parameters. We provide comparisons of the self-consistent solution to various approximation schemes including the static NTK approximation, gradient independence assumption, and leading order perturbation theory, showing that each of these approximations can break down in regimes where general self-consistent solutions still provide an accurate description. Lastly, we provide experiments in more realistic settings which demonstrate that the loss and kernel dynamics of convolutional neural networks at fixed feature learning strength are preserved across different widths on a image classification task."
                },
                {
                    "title": "The merged-staircase property: a necessary and nearly sufficient condition for SGD learning of sparse functions on two-layer neural networks",
                    "abstract": "It is currently known how to characterize functions that neural networks can learn with SGD for two extremal parameterizations: neural networks in the linear regime, and neural networks with no structural constraints. However, for the main parametrization of interest (non-linear but regular networks) no tight characterization has yet been achieved, despite significant developments. We take a step in this direction by considering depth-2 neural networks trained by SGD in the mean-field regime. We consider functions on binary inputs that depend on a latent low-dimensional subspace (i.e., small number of coordinates). This regime is of interest since it is poorly understood how neural networks routinely tackle high-dimensional datasets and adapt to latent low-dimensional structure without suffering from the curse of dimensionality. Accordingly, we study SGD-learnability with $O(d)$ sample complexity in a large ambient dimension $d$. Our main results characterize a hierarchical property, the\"merged-staircase property\", that is both necessary and nearly sufficient for learning in this setting. We further show that non-linear training is necessary: for this class of functions, linear methods on any feature map (e.g., the NTK) are not capable of learning efficiently. The key tools are a new\"dimension-free\"dynamics approximation result that applies to functions defined on a latent space of low-dimension, a proof of global convergence based on polynomial identity testing, and an improvement of lower bounds against linear methods for non-almost orthogonal functions."
                },
                {
                    "title": "The staircase property: How hierarchical structure can guide deep learning",
                    "abstract": "This paper identi\ufb01es a structural property of data distributions that enables deep neural networks to learn hierarchically. We de\ufb01ne the \u201cstaircase\u201d property for functions over the Boolean hypercube, which posits that high-order Fourier coef\ufb01cients are reachable from lower-order Fourier coef\ufb01cients along increasing chains. We prove that functions satisfying this property can be learned in polynomial time using layerwise stochastic coordinate descent on regular neural networks \u2013 a class of network architectures and initializations that have homogeneity properties. Our analysis shows that for such staircase functions and neural networks, the gradient-based algorithm learns high-level features by greedily combining lower-level features along the depth of the network. We further back our theoretical results with experiments showing that staircase functions are learnable by more standard ResNet architectures with stochastic gradient descent. Both the theoretical and experimental results support the fact that the staircase property has a role to play in understanding the capabilities of gradient-based learning on regular networks, in contrast to general polynomial-size networks that can emulate any Statistical Query or PAC algorithm, as recently shown."
                },
                {
                    "title": "Implicit Bias of SGD for Diagonal Linear Networks: a Provable Benefit of Stochasticity",
                    "abstract": "Understanding the implicit bias of training algorithms is of crucial importance in order to explain the success of overparametrised neural networks. In this paper, we study the dynamics of stochastic gradient descent over diagonal linear networks through its continuous time version, namely stochastic gradient flow. We explicitly characterise the solution chosen by the stochastic flow and prove that it always enjoys better generalisation properties than that of gradient flow. Quite surprisingly, we show that the convergence speed of the training loss controls the magnitude of the biasing effect: the slower the convergence, the better the bias. To fully complete our analysis, we provide convergence guarantees for the dynamics. We also give experimental results which support our theoretical claims. Our findings highlight the fact that structured noise can induce better generalisation and they help explain the greater performances observed in practice of stochastic gradient descent over gradient descent."
                },
                {
                    "title": "Towards Resolving the Implicit Bias of Gradient Descent for Matrix Factorization: Greedy Low-Rank Learning",
                    "abstract": "Matrix factorization is a simple and natural test-bed to investigate the implicit regularization of gradient descent. Gunasekar et al. (2018) conjectured that Gradient Flow with infinitesimal initialization converges to the solution that minimizes the nuclear norm, but a series of recent papers argued that the language of norm minimization is not sufficient to give a full characterization for the implicit regularization. In this work, we provide theoretical and empirical evidence that for depth-2 matrix factorization, gradient flow with infinitesimal initialization is mathematically equivalent to a simple heuristic rank minimization algorithm, Greedy Low-Rank Learning, under some reasonable assumptions. This generalizes the rank minimization view from previous works to a much broader setting and enables us to construct counter-examples to refute the conjecture from Gunasekar et al. (2018). We also extend the results to the case where depth $\\ge 3$, and we show that the benefit of being deeper is that the above convergence has a much weaker dependence over initialization magnitude so that this rank minimization is more likely to take effect for initialization with practical scale."
                },
                {
                    "title": "Feature Learning in Infinite-Width Neural Networks",
                    "abstract": "As its width tends to infinity, a deep neural network's behavior under gradient descent can become simplified and predictable (e.g. given by the Neural Tangent Kernel (NTK)), if it is parametrized appropriately (e.g. the NTK parametrization). However, we show that the standard and NTK parametrizations of a neural network do not admit infinite-width limits that can learn features, which is crucial for pretraining and transfer learning such as with BERT. We propose simple modifications to the standard parametrization to allow for feature learning in the limit. Using the *Tensor Programs* technique, we derive explicit formulas for such limits. On Word2Vec and few-shot learning on Omniglot via MAML, two canonical tasks that rely crucially on feature learning, we compute these limits exactly. We find that they outperform both NTK baselines and finite-width networks, with the latter approaching the infinite-width feature learning performance as width increases. \nMore generally, we classify a natural space of neural network parametrizations that generalizes standard, NTK, and Mean Field parametrizations. We show 1) any parametrization in this space either admits feature learning or has an infinite-width training dynamics given by kernel gradient descent, but not both; 2) any such infinite-width limit can be computed using the Tensor Programs technique."
                },
                {
                    "title": "Tensor Programs II: Neural Tangent Kernel for Any Architecture",
                    "abstract": "We prove that a randomly initialized neural network of *any architecture* has its Tangent Kernel (NTK) converge to a deterministic limit, as the network widths tend to infinity. We demonstrate how to calculate this limit. In prior literature, the heuristic study of neural network gradients often assumes every weight matrix used in forward propagation is independent from its transpose used in backpropagation (Schoenholz et al. 2017). This is known as the *gradient independence assumption (GIA)*. We identify a commonly satisfied condition, which we call *Simple GIA Check*, such that the NTK limit calculation based on GIA is correct. Conversely, when Simple GIA Check fails, we show GIA can result in wrong answers. Our material here presents the NTK results of Yang (2019a) in a friendly manner and showcases the *tensor programs* technique for understanding wide neural networks. We provide reference implementations of infinite-width NTKs of recurrent neural network, transformer, and batch normalization at this https URL."
                },
                {
                    "title": "On the training dynamics of deep networks with L2 regularization",
                    "abstract": "We study the role of $L_2$ regularization in deep learning, and uncover simple relations between the performance of the model, the $L_2$ coefficient, the learning rate, and the number of training steps. These empirical relations hold when the network is overparameterized. They can be used to predict the optimal regularization parameter of a given model. In addition, based on these observations we propose a dynamical schedule for the regularization parameter that improves performance and speeds up training. We test these proposals in modern image classification settings. Finally, we show that these empirical relations can be understood theoretically in the context of infinitely wide networks. We derive the gradient flow dynamics of such networks, and compare the role of $L_2$ regularization in this context with that of linear models."
                },
                {
                    "title": "The large learning rate phase of deep learning: the catapult mechanism",
                    "abstract": "The choice of initial learning rate can have a profound effect on the performance of deep networks. We present a class of neural networks with solvable training dynamics, and confirm their predictions empirically in practical deep learning settings. The networks exhibit sharply distinct behaviors at small and large learning rates. The two regimes are separated by a phase transition. In the small learning rate phase, training can be understood using the existing theory of infinitely wide neural networks. At large learning rates the model captures qualitatively distinct phenomena, including the convergence of gradient descent dynamics to flatter minima. One key prediction of our model is a narrow range of large, stable learning rates. We find good agreement between our model's predictions and training dynamics in realistic deep learning settings. Furthermore, we find that the optimal performance in such settings is often found in the large learning rate phase. We believe our results shed light on characteristics of models trained at different learning rates. In particular, they fill a gap between existing wide neural network theory, and the nonlinear, large learning rate, training dynamics relevant to practice."
                },
                {
                    "title": "On the Global Convergence of Training Deep Linear ResNets",
                    "abstract": "We study the convergence of gradient descent (GD) and stochastic gradient descent (SGD) for training $L$-hidden-layer linear residual networks (ResNets). We prove that for training deep residual networks with certain linear transformations at input and output layers, which are fixed throughout training, both GD and SGD with zero initialization on all hidden weights can converge to the global minimum of the training loss. Moreover, when specializing to appropriate Gaussian random linear transformations, GD and SGD provably optimize wide enough deep linear ResNets. Compared with the global convergence result of GD for training standard deep linear networks (Du & Hu 2019), our condition on the neural network width is sharper by a factor of $O(\\kappa L)$, where $\\kappa$ denotes the condition number of the covariance matrix of the training data. We further propose a modified identity input and output transformations, and show that a $(d+k)$-wide neural network is sufficient to guarantee the global convergence of GD/SGD, where $d,k$ are the input and output dimensions respectively."
                },
                {
                    "title": "Toward a theory of optimization for over-parameterized systems of non-linear equations: the lessons of deep learning",
                    "abstract": "The success of deep learning is due, to a great extent, to the remarkable effectiveness of gradient-based optimization methods applied to large neural networks. In this work we isolate some general mathematical structures allowing for efficient optimization in over-parameterized systems of non-linear equations, a setting that includes deep neural networks. In particular, we show that optimization problems corresponding to such systems are not convex, even locally, but instead satisfy the Polyak-Lojasiewicz (PL) condition allowing for efficient optimization by gradient descent or SGD. We connect the PL condition of these systems to the condition number associated to the tangent kernel and develop a non-linear theory parallel to classical analyses of over-parameterized linear equations. We discuss how these ideas apply to training shallow and deep neural networks. Finally, we point out that tangent kernels associated to certain large system may be far from constant, even locally. Yet, our analysis still allows to demonstrate existence of solutions and convergence of gradient descent and SGD."
                },
                {
                    "title": "Implicit Bias of Gradient Descent for Wide Two-layer Neural Networks Trained with the Logistic Loss",
                    "abstract": "Neural networks trained to minimize the logistic (a.k.a. cross-entropy) loss with gradient-based methods are observed to perform well in many supervised classification tasks. Towards understanding this phenomenon, we analyze the training and generalization behavior of infinitely wide two-layer neural networks with homogeneous activations. We show that the limits of the gradient flow on exponentially tailed losses can be fully characterized as a max-margin classifier in a certain non-Hilbertian space of functions. In presence of hidden low-dimensional structures, the resulting margin is independent of the ambiant dimension, which leads to strong generalization bounds. In contrast, training only the output layer implicitly solves a kernel support vector machine, which a priori does not enjoy such an adaptivity. Our analysis of training is non-quantitative in terms of running time but we prove computational guarantees in simplified settings by showing equivalences with online mirror descent. Finally, numerical experiments suggest that our analysis describes well the practical behavior of two-layer neural networks with ReLU activation and confirm the statistical benefits of this implicit bias."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "A Function Space View of Bounded Norm Infinite Width ReLU Nets: The Multivariate Case",
                    "abstract": "A key element of understanding the efficacy of overparameterized neural networks is characterizing how they represent functions as the number of weights in the network approaches infinity. In this paper, we characterize the norm required to realize a function $f:\\mathbb{R}^d\\rightarrow\\mathbb{R}$ as a single hidden-layer ReLU network with an unbounded number of units (infinite width), but where the Euclidean norm of the weights is bounded, including precisely characterizing which functions can be realized with finite norm. This was settled for univariate univariate functions in Savarese et al. (2019), where it was shown that the required norm is determined by the L1-norm of the second derivative of the function. We extend the characterization to multivariate functions (i.e., networks with d input units), relating the required norm to the L1-norm of the Radon transform of a (d+1)/2-power Laplacian of the function. This characterization allows us to show that all functions in Sobolev spaces $W^{s,1}(\\mathbb{R})$, $s\\geq d+1$, can be represented with bounded norm, to calculate the required norm for several specific functions, and to obtain a depth separation result. These results have important implications for understanding generalization performance and the distinction between neural networks and more traditional kernel learning."
                },
                {
                    "title": "Disentangling feature and lazy training in deep neural networks",
                    "abstract": "Two distinct limits for deep learning have been derived as the network width h \u2192 \u221e, depending on how the weights of the last layer scale with h. In the neural tangent Kernel (NTK) limit, the dynamics becomes linear in the weights and is described by a frozen kernel \u0398 (the NTK). By contrast, in the mean-field limit, the dynamics can be expressed in terms of the distribution of the parameters associated with a neuron, that follows a partial differential equation. In this work we consider deep networks where the weights in the last layer scale as \u03b1h \u22121/2 at initialization. By varying \u03b1 and h, we probe the crossover between the two limits. We observe two the previously identified regimes of \u2018lazy training\u2019 and \u2018feature training\u2019. In the lazy-training regime, the dynamics is almost linear and the NTK barely changes after initialization. The feature-training regime includes the mean-field formulation as a limiting case and is characterized by a kernel that evolves in time, and thus learns some features. We perform numerical experiments on MNIST, Fashion-MNIST, EMNIST and CIFAR10 and consider various architectures. We find that: (i) the two regimes are separated by an \u03b1* that scales as 1h . (ii) Network architecture and data structure play an important role in determining which regime is better: in our tests, fully-connected networks perform generally better in the lazy-training regime, unlike convolutional networks. (iii) In both regimes, the fluctuations \u03b4F induced on the learned function by initial conditions decay as \u03b4F\u223c1/h , leading to a performance that increases with h. The same improvement can also be obtained at an intermediate width by ensemble-averaging several networks that are trained independently. (iv) In the feature-training regime we identify a time scale t1\u223ch\u03b1 , such that for t \u226a t 1 the dynamics is linear. At t \u223c t 1, the output has grown by a magnitude h and the changes of the tangent kernel | |\u0394\u0398| | become significant. Ultimately, it follows ||\u0394\u0398||\u223c(h\u03b1)\u2212a for ReLU and Softplus activation functions, with a < 2 and a \u2192 2 as depth grows. We provide scaling arguments supporting these findings."
                },
                {
                    "title": "Kernel and Rich Regimes in Overparametrized Models",
                    "abstract": "A recent line of work studies overparametrized neural networks in the \"kernel regime,\" i.e. when the network behaves during training as a kernelized linear predictor, and thus training with gradient descent has the effect of finding the minimum RKHS norm solution. This stands in contrast to other studies which demonstrate how gradient descent on overparametrized multilayer networks can induce rich implicit biases that are not RKHS norms. Building on an observation by Chizat and Bach, we show how the scale of the initialization controls the transition between the \"kernel\" (aka lazy) and \"rich\" (aka active) regimes and affects generalization properties in multilayer homogeneous models. We also highlight an interesting role for the width of a model in the case that the predictor is not identically zero at initialization. We provide a complete and detailed analysis for a family of simple depth-$D$ models that already exhibit an interesting and meaningful transition between the kernel and rich regimes, and we also demonstrate this transition empirically for more complex matrix factorization models and multilayer non-linear networks."
                },
                {
                    "title": "Implicit Regularization in Deep Matrix Factorization",
                    "abstract": "Efforts to understand the generalization mystery in deep learning have led to the belief that gradient-based optimization induces a form of implicit regularization, a bias towards models of low \"complexity.\" We study the implicit regularization of gradient descent over deep linear neural networks for matrix completion and sensing, a model referred to as deep matrix factorization. Our first finding, supported by theory and experiments, is that adding depth to a matrix factorization enhances an implicit tendency towards low-rank solutions, oftentimes leading to more accurate recovery. Secondly, we present theoretical and empirical arguments questioning a nascent view by which implicit regularization in matrix factorization can be captured using simple mathematical norms. Our results point to the possibility that the language of standard regularizers may not be rich enough to fully encompass the implicit regularization brought forth by gradient-based optimization."
                },
                {
                    "title": "Implicit Regularization of Discrete Gradient Dynamics in Deep Linear Neural Networks",
                    "abstract": "When optimizing over-parameterized models, such as deep neural networks, a large set of parameters can achieve zero training error. In such cases, the choice of the optimization algorithm and its respective hyper-parameters introduces biases that will lead to convergence to specific minimizers of the objective. Consequently, this choice can be considered as an implicit regularization for the training of over-parametrized models. In this work, we push this idea further by studying the discrete gradient dynamics of the training of a two-layer linear network with the least-squares loss. Using a time rescaling, we show that, with a vanishing initialization and a small enough step size, this dynamics sequentially learns the solutions of a reduced-rank regression with a gradually increasing rank."
                },
                {
                    "title": "On Exact Computation with an Infinitely Wide Neural Net",
                    "abstract": "How well does a classic deep net architecture like AlexNet or VGG19 classify on a standard dataset such as CIFAR-10 when its width --- namely, number of channels in convolutional layers, and number of nodes in fully-connected internal layers --- is allowed to increase to infinity? Such questions have come to the forefront in the quest to theoretically understand deep learning and its mysteries about optimization and generalization. They also connect deep learning to notions such as Gaussian processes and kernels. A recent paper [Jacot et al., 2018] introduced the Neural Tangent Kernel (NTK) which captures the behavior of fully-connected deep nets in the infinite width limit trained by gradient descent; this object was implicit in some other recent papers. An attraction of such ideas is that a pure kernel-based method is used to capture the power of a fully-trained deep net of infinite width. \nThe current paper gives the first efficient exact algorithm for computing the extension of NTK to convolutional neural nets, which we call Convolutional NTK (CNTK), as well as an efficient GPU implementation of this algorithm. This results in a significant new benchmark for the performance of a pure kernel-based method on CIFAR-10, being $10\\%$ higher than the methods reported in [Novak et al., 2019], and only $6\\%$ lower than the performance of the corresponding finite deep net architecture (once batch normalization, etc. are turned off). Theoretically, we also give the first non-asymptotic proof showing that a fully-trained sufficiently wide net is indeed equivalent to the kernel regression predictor using NTK."
                },
                {
                    "title": "A Note on Lazy Training in Supervised Differentiable Programming",
                    "abstract": "In a series of recent theoretical works, it has been shown that strongly over-parameterized neural networks trained with gradient-based methods could converge linearly to zero training loss, with their parameters hardly varying. In this note, our goal is to exhibit the simple structure that is behind these results. In a simplified setting, we prove that \"lazy training\" essentially solves a kernel regression. We also show that this behavior is not so much due to over-parameterization than to a choice of scaling, often implicit, that allows to linearize the model around its initialization. These theoretical results complemented with simple numerical experiments make it seem unlikely that \"lazy training\" is behind the many successes of neural networks in high dimensional tasks."
                },
                {
                    "title": "A mathematical theory of semantic development in deep neural networks",
                    "abstract": "Significance Over the course of development, humans learn myriad facts about items in the world, and naturally group these items into useful categories and structures. This semantic knowledge is essential for diverse behaviors and inferences in adulthood. How is this richly structured semantic knowledge acquired, organized, deployed, and represented by neuronal networks in the brain? We address this question by studying how the nonlinear learning dynamics of deep linear networks acquires information about complex environmental structures. Our results show that this deep learning dynamics can self-organize emergent hidden representations in a manner that recapitulates many empirical phenomena in human semantic development. Such deep networks thus provide a mathematically tractable window into the development of internal neural representations through experience. An extensive body of empirical research has revealed remarkable regularities in the acquisition, organization, deployment, and neural representation of human semantic knowledge, thereby raising a fundamental conceptual question: What are the theoretical principles governing the ability of neural networks to acquire, organize, and deploy abstract knowledge by integrating across many individual experiences? We address this question by mathematically analyzing the nonlinear dynamics of learning in deep linear networks. We find exact solutions to this learning dynamics that yield a conceptual explanation for the prevalence of many disparate phenomena in semantic cognition, including the hierarchical differentiation of concepts through rapid developmental transitions, the ubiquity of semantic illusions between such transitions, the emergence of item typicality and category coherence as factors controlling the speed of semantic processing, changing patterns of inductive projection over development, and the conservation of semantic similarity in neural representations across species. Thus, surprisingly, our simple neural model qualitatively recapitulates many diverse regularities underlying semantic development, while providing analytic insight into how the statistical structure of an environment can interact with nonlinear deep-learning dynamics to give rise to these regularities."
                },
                {
                    "title": "Gradient descent aligns the layers of deep linear networks",
                    "abstract": "This paper establishes risk convergence and asymptotic weight matrix alignment --- a form of implicit regularization --- of gradient flow and gradient descent when applied to deep linear networks on linearly separable data. In more detail, for gradient flow applied to strictly decreasing loss functions (with similar results for gradient descent with particular decreasing step sizes): (i) the risk converges to 0; (ii) the normalized i-th weight matrix asymptotically equals its rank-1 approximation $u_iv_i^{\\top}$; (iii) these rank-1 matrices are aligned across layers, meaning $|v_{i+1}^{\\top}u_i|\\to1$. In the case of the logistic loss (binary cross entropy), more can be said: the linear function induced by the network --- the product of its weight matrices --- converges to the same direction as the maximum margin solution. This last property was identified in prior work, but only under assumptions on gradient descent which here are implied by the alignment phenomenon."
                },
                {
                    "title": "Gradient Descent Provably Optimizes Over-parameterized Neural Networks",
                    "abstract": "One of the mysteries in the success of neural networks is randomly initialized first order methods like gradient descent can achieve zero training loss even though the objective function is non-convex and non-smooth. This paper demystifies this surprising phenomenon for two-layer fully connected ReLU activated neural networks. For an $m$ hidden node shallow neural network with ReLU activation and $n$ training data, we show as long as $m$ is large enough and no two inputs are parallel, randomly initialized gradient descent converges to a globally optimal solution at a linear convergence rate for the quadratic loss function. \nOur analysis relies on the following observation: over-parameterization and random initialization jointly restrict every weight vector to be close to its initialization for all iterations, which allows us to exploit a strong convexity-like property to show that gradient descent converges at a global linear rate to the global optimum. We believe these insights are also useful in analyzing deep models and other first order methods."
                },
                {
                    "title": "A Convergence Analysis of Gradient Descent for Deep Linear Neural Networks",
                    "abstract": "We analyze speed of convergence to global optimum for gradient descent training a deep linear neural network (parameterized as $x \\mapsto W_N W_{N-1} \\cdots W_1 x$) by minimizing the $\\ell_2$ loss over whitened data. Convergence at a linear rate is guaranteed when the following hold: (i) dimensions of hidden layers are at least the minimum of the input and output dimensions; (ii) weight matrices at initialization are approximately balanced; and (iii) the initial loss is smaller than the loss of any rank-deficient solution. The assumptions on initialization (conditions (ii) and (iii)) are necessary, in the sense that violating any one of them may lead to convergence failure. Moreover, in the important case of output dimension 1, i.e. scalar regression, they are met, and thus convergence to global optimum holds, with constant probability under a random initialization scheme. Our results significantly extend previous analyses, e.g., of deep linear residual networks (Bartlett et al., 2018)."
                },
                {
                    "title": "A Convergence Theory for Deep Learning via Over-Parameterization",
                    "abstract": "Deep neural networks (DNNs) have demonstrated dominating performance in many fields; since AlexNet, networks used in practice are going wider and deeper. On the theoretical side, a long line of works has been focusing on training neural networks with one hidden layer. The theory of multi-layer networks remains largely unsettled. \nIn this work, we prove why stochastic gradient descent (SGD) can find $\\textit{global minima}$ on the training objective of DNNs in $\\textit{polynomial time}$. We only make two assumptions: the inputs are non-degenerate and the network is over-parameterized. The latter means the network width is sufficiently large: $\\textit{polynomial}$ in $L$, the number of layers and in $n$, the number of samples. \nOur key technique is to derive that, in a sufficiently large neighborhood of the random initialization, the optimization landscape is almost-convex and semi-smooth even with ReLU activations. This implies an equivalence between over-parameterized neural networks and neural tangent kernel (NTK) in the finite (and polynomial) width setting. \nAs concrete examples, starting from randomly initialized weights, we prove that SGD can attain 100% training accuracy in classification tasks, or minimize regression loss in linear convergence speed, with running time polynomial in $n,L$. Our theory applies to the widely-used but non-smooth ReLU activation, and to any smooth and possibly non-convex loss functions. In terms of network architectures, our theory at least applies to fully-connected neural networks, convolutional neural networks (CNN), and residual neural networks (ResNet)."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "On the Global Convergence of Gradient Descent for Over-parameterized Models using Optimal Transport",
                    "abstract": "Many tasks in machine learning and signal processing can be solved by minimizing a convex function of a measure. This includes sparse spikes deconvolution or training a neural network with a single hidden layer. For these problems, we study a simple minimization method: the unknown measure is discretized into a mixture of particles and a continuous-time gradient descent is performed on their weights and positions. This is an idealization of the usual way to train neural networks with a large hidden layer. We show that, when initialized correctly and in the many-particle limit, this gradient flow, although non-convex, converges to global minimizers. The proof involves Wasserstein gradient flows, a by-product of optimal transport theory. Numerical experiments show that this asymptotic behavior is already at play for a reasonable number of particles, even in high dimension."
                },
                {
                    "title": "On the Optimization of Deep Networks: Implicit Acceleration by Overparameterization",
                    "abstract": "Conventional wisdom in deep learning states that increasing depth improves expressiveness but complicates optimization. This paper suggests that, sometimes, increasing depth can speed up optimization. The effect of depth on optimization is decoupled from expressiveness by focusing on settings where additional layers amount to overparameterization - linear neural networks, a well-studied model. Theoretical analysis, as well as experiments, show that here depth acts as a preconditioner which may accelerate convergence. Even on simple convex problems such as linear regression with $\\ell_p$ loss, $p>2$, gradient descent can benefit from transitioning to a non-convex overparameterized objective, more than it would from some common acceleration schemes. We also prove that it is mathematically impossible to obtain the acceleration effect of overparametrization via gradients of any regularizer."
                },
                {
                    "title": "Breaking the Curse of Dimensionality with Convex Neural Networks",
                    "abstract": "We consider neural networks with a single hidden layer and non-decreasing homogeneous activa-tion functions like the rectified linear units. By letting the number of hidden units grow unbounded and using classical non-Euclidean regularization tools on the output weights, we provide a detailed theoretical analysis of their generalization performance, with a study of both the approximation and the estimation errors. We show in particular that they are adaptive to unknown underlying linear structures, such as the dependence on the projection of the input variables onto a low-dimensional subspace. Moreover, when using sparsity-inducing norms on the input weights, we show that high-dimensional non-linear variable selection may be achieved, without any strong assumption regarding the data and with a total number of variables potentially exponential in the number of ob-servations. In addition, we provide a simple geometric interpretation to the non-convex problem of addition of a new unit, which is the core potentially hard computational element in the framework of learning from continuously many basis functions. We provide simple conditions for convex relaxations to achieve the same generalization error bounds, even when constant-factor approxi-mations cannot be found (e.g., because it is NP-hard such as for the zero-homogeneous activation function). We were not able to find strong enough convex relaxations and leave open the existence or non-existence of polynomial-time algorithms."
                },
                {
                    "title": "A useful variant of the Davis--Kahan theorem for statisticians",
                    "abstract": "The Davis\u2013Kahan theorem is used in the analysis of many statistical procedures to bound the distance between subspaces spanned by population eigenvectors and their sample versions. It relies on an eigenvalue separation condition between certain population and sample eigenvalues. We present a variant of this result that depends only on a population eigenvalue separation condition, making it more natural and convenient for direct application in statistical contexts, and provide an improvement in many cases to the usual bound in the statistical literature. We also give an extension to situations where the matrices under study may be asymmetric or even non-square, and where interest is in the distance between subspaces spanned by corresponding singular vectors."
                },
                {
                    "title": "Exact solutions to the nonlinear dynamics of learning in deep linear neural networks",
                    "abstract": "Despite the widespread practical success of deep learning methods, our theoretical understanding of the dynamics of learning in deep neural networks remains quite sparse. We attempt to bridge the gap between the theory and practice of deep learning by systematically analyzing learning dynamics for the restricted case of deep linear neural networks. Despite the linearity of their input-output map, such networks have nonlinear gradient descent dynamics on weights that change with the addition of each new hidden layer. We show that deep linear networks exhibit nonlinear learning phenomena similar to those seen in simulations of nonlinear networks, including long plateaus followed by rapid transitions to lower error solutions, and faster convergence from greedy unsupervised pretraining initial conditions than from random initial conditions. We provide an analytical description of these phenomena by finding new exact solutions to the nonlinear dynamics of deep learning. Our theoretical analysis also reveals the surprising finding that as the depth of a network approaches infinity, learning speed can nevertheless remain finite: for a special class of initial conditions on the weights, very deep networks incur only a finite, depth independent, delay in learning speed relative to shallow networks. We show that, under certain conditions on the training data, unsupervised pretraining can find this special class of initial conditions, while scaled random Gaussian initializations cannot. We further exhibit a new class of random orthogonal initial conditions on weights that, like unsupervised pre-training, enjoys depth independent learning times. We further show that these initial conditions also lead to faithful propagation of gradients even in deep nonlinear networks, as long as they operate in a special regime known as the edge of chaos."
                },
                {
                    "title": "Isotropic local laws for sample covariance and generalized Wigner matrices",
                    "abstract": "We consider sample covariance matrices of the form $X^*X$, where $X$ is an $M \\times N$ matrix with independent random entries. \u00a0We prove the isotropic local Marchenko-Pastur law, i.e. we prove that the resolvent $(X^* X - z)^{-1}$ converges to a multiple of the identity in the sense of quadratic forms. More precisely, we establish sharp high-probability bounds on the quantity $\\langle v , (X^* X - z)^{-1}w\\rangle - \\langle v , w\\rangle m(z)$, where $m$ is the Stieltjes transform of the Marchenko-Pastur law and $v , w \\in \\mathbb{C}^N$. We require the logarithms of the dimensions $M$ and $N$ to be comparable. Our result holds down to scales $\\Im z \\geq N^{-1+\\varepsilon}$ and throughout the entire spectrum away from 0. We also prove analogous results for generalized Wigner matrices."
                },
                {
                    "title": "On the Convergence of Gradient Flow on Multi-layer Linear Models",
                    "abstract": "In this paper, we analyze the convergence of gradient flow on a multi-layer linear model with a loss function of the form f ( W 1 W 2 \u00b7 \u00b7 \u00b7 W L ) . We show that when f satisfies the gradient dominance property, proper weight initialization leads to exponential convergence of the gradient flow to a global minimum of the loss. Moreover, the convergence rate depends on two trajectory-specific quantities that are controlled by the weight initialization: the imbalance matrices , which measure the difference between the weights of adjacent layers, and the least singular value of the weight product W = W 1 W 2 \u00b7 \u00b7 \u00b7 W L . Our analysis exploits the fact that the gradient of the overparam-eterized loss can be written as the composition of the non-overparametrized gradient with a time-varying (weight-dependent) linear operator whose smallest eigenvalue controls the convergence rate. The key challenge we address is to derive a uniform lower bound for this time-varying eigenvalue that lead to improved rates for several multi-layer network models studied in the literature."
                },
                {
                    "title": "Linear Convergence of Gradient Descent For Finite Width Over-parametrized Linear Networks With General Initialization",
                    "abstract": "Recent theoretical analyses of the convergence of gradient descent (GD) to a global minimum for over-parametrized neural networks make strong assumptions on the step size (infinitesimal), the hidden-layer width (infinite), or the initialization (spectral, balanced). In this work, we relax these assumptions and derive a linear convergence rate for two-layer linear networks trained using GD on the squared loss in the case of finite step size, finite width and general initialization. Despite the generality of our analysis, our rate estimates are significantly tighter than those of prior work. Moreover, we provide a time-varying step size rule that monotonically improves the convergence rate as the loss function decreases to zero. Numerical experiments validate our findings."
                },
                {
                    "title": "Understanding the Dynamics of Gradient Flow in Overparameterized Linear models",
                    "abstract": "We provide a detailed analysis of the dynamics of the gradient flow in overparameterized two-layer linear models. A particularly interesting feature of this model is that its nonlinear dynamics can be exactly solved as a consequence of a large number of conservation laws that constrain the system to follow particular trajectories. More precisely, the gradient flow preserves the difference of the Gramian matrices of the input and output weights, and its convergence to equilibrium depends on both the magnitude of that difference (which is fixed at initialization) and the spectrum of the data. In addition, and generalizing prior work, we prove our results without assuming small, balanced or spectral initialization for the weights. Moreover, we establish interesting mathematical connections between matrix factorization problems and differential equations of the Riccati type."
                },
                {
                    "title": "On the Explicit Role of Initialization on the Convergence and Implicit Bias of Overparametrized Linear Networks",
                    "abstract": "Neural networks trained via gradient descent with random initialization and without any regularization enjoy good generalization performance in practice despite being highly overparametrized. A promising direction to explain this phenomenon is to study how initialization and overparametriza-tion affect convergence and implicit bias of training algorithms. In this paper, we present a novel analysis of single-hidden-layer linear networks trained under gradient \ufb02ow, which connects initialization, optimization, and overparametrization. Firstly, we show that the squared loss converges exponentially to its optimum at a rate that depends on the level of imbalance of the initialization. Secondly, we show that proper initialization constrains the dynamics of the network parameters to lie within an invariant set. In turn, minimizing the loss over this set leads to the min-norm solution. Finally, we show that large hidden layer width, together with (properly scaled) random initialization, ensures proximity to such an invariant set during training, allowing us to derive a novel non-asymptotic upper-bound on the distance between the trained network and the min-norm solution."
                },
                {
                    "title": "Representation Costs of Linear Neural Networks: Analysis and Design",
                    "abstract": "For different parameterizations (mappings from parameters to predictors), we study the regularization cost in predictor space induced by l 2 regularization on the parameters (weights). We focus on linear neural networks as parameterizations of linear predictors. We identify the representation cost of certain sparse linear ConvNets and residual networks. In order to get a better understanding of how the architecture and parameterization affect the representation cost, we also study the reverse problem, identifying which regularizers on linear predictors (e"
                },
                {
                    "title": "Parameters as interacting particles: long time convergence and asymptotic error scaling of neural networks",
                    "abstract": "The performance of neural networks on high-dimensional data distributions suggests that it may be possible to parameterize a representation of a given high-dimensional function with controllably small errors, potentially outperforming standard interpolation methods. We demonstrate, both theoretically and numerically, that this is indeed the case. We map the parameters of a neural network to a system of particles relaxing with an interaction potential determined by the loss function. We show that in the limit that the number of parameters $n$ is large, the landscape of the mean-squared error becomes convex and the representation error in the function scales as $O(n^{-1})$. In this limit, we prove a dynamical variant of the universal approximation theorem showing that the optimal representation can be attained by stochastic gradient descent, the algorithm ubiquitously used for parameter optimization in machine learning. In the asymptotic regime, we study the fluctuations around the optimal representation and show that they arise at a scale $O(n^{-1})$. These fluctuations in the landscape identify the natural scale for the noise in stochastic gradient descent. Our results apply to both single and multi-layer neural networks, as well as standard kernel methods like radial basis functions."
                },
                {
                    "title": "Effect of Batch Learning in Multilayer Neural Networks",
                    "abstract": "This paper discusses batch gradient descent learning in mul-tilayer networks with a large number of statistical training data. We emphasize on the diierence between regular cases, where the prepared model has the same size as the true function , and overrealizable cases, where the model has surplus hidden units to realize the true function. First, experimental study on multilayer perceptrons and linear neural networks (LNN) shows that batch learning induces strong overtrain-ing on both models in overrealizable cases, which means the degrade of generalization error by surplus units can be alleviated. We theoretically analyze the dynamics in LNN, and show that this overtraining is caused by shrinkage of the parameters corresponding to surplus units."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nWhat are the precise conditions and dynamics that characterize the transition between lazy and active regimes in fully-connected linear networks as the number of neurons increases?\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding the transition between lazy and active regimes in neural networks is crucial for advancing the theoretical foundations of machine learning. By clarifying these dynamics, the research can lead to improved training methodologies and architectures that leverage feature learning and sparsity, ultimately enhancing model performance and efficiency. This work could inspire future research into the design of neural networks that optimize learning dynamics, potentially leading to practical applications in various fields such as computer vision, natural language processing, and beyond.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complex interplay between network architecture, initialization, and training dynamics. Naive approaches may fail because they do not account for the nuanced behavior of networks under different conditions, such as varying initialization scales and learning rates. Theoretical obstacles include the need to rigorously define and differentiate between the lazy and active regimes, as well as the difficulty in capturing the dynamics of feature learning and sparsity in high-dimensional spaces.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often treated the lazy and active regimes as distinct and separate, lacking a comprehensive understanding of their transition. Limitations in prior work include insufficient exploration of the effects of network width and weight initialization variance on these regimes. Additionally, many studies have focused on specific cases without providing a unified framework. This research aims to fill these gaps by offering a detailed phase diagram and a more granular view of the regimes, highlighting the coexistence of lazy and active dynamics within the same network.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves analyzing fully-connected linear networks by varying the width (w) and the variance of the weights at initialization (\u03c3\u00b2). The study will utilize a phase diagram to illustrate the transitions between lazy and active regimes, employing metrics that capture the dynamics of feature learning and sparsity. Expected outcomes include a clearer understanding of the conditions under which networks operate in different regimes, as well as insights into the coexistence of these regimes, which could inform future neural network designs and training strategies."
            }
        },
        "author_data": {
            "3e20439a-cee0-470e-9934-58279e00f7e6": {
                "pk": "3e20439a-cee0-470e-9934-58279e00f7e6",
                "name": "Zhenfeng Tu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "5199b94a-631b-48a0-99cf-00bb369c1a85": {
                "pk": "5199b94a-631b-48a0-99cf-00bb369c1a85",
                "name": "Santiago Aranguri",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "2cb86a5e-420e-4d98-b73f-ae146d4c8f69": {
                "pk": "2cb86a5e-420e-4d98-b73f-ae146d4c8f69",
                "name": "Arthur Jacot",
                "collaborators": [
                    "Cl\u00e9ment Hongler",
                    "Franck Gabriel",
                    "Berfin \u015eim\u015fek",
                    "Fran\u00e7ois Ged",
                    "Francesco Spadaro",
                    "Zihan Wang",
                    "Yuxiao Wen",
                    "Alexandre Kaiser",
                    "Seok Hoan Choi",
                    "Benjamin Dupuis"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Optimization",
                    "Theoretical Analysis"
                ],
                "publications": [
                    {
                        "title": "Implicit Bias of Large Depth Networks: a Notion of Rank for Nonlinear Functions",
                        "abstract": "We show that the representation cost of fully connected neural networks with homogeneous nonlinearities - which describes the implicit bias in function space of networks with $L_2$-regularization or with losses such as the cross-entropy - converges as the depth of the network goes to infinity to a notion of rank over nonlinear functions. We then inquire under which conditions the global minima of the loss recover the `true' rank of the data: we show that for too large depths the global minimum will be approximately rank 1 (underestimating the rank); we then argue that there is a range of depths which grows with the number of datapoints where the true rank is recovered. Finally, we discuss the effect of the rank of a classifier on the topology of the resulting class boundaries and show that autoencoders with optimal nonlinear rank are naturally denoising."
                    },
                    {
                        "title": "Bottleneck Structure in Learned Features: Low-Dimension vs Regularity Tradeoff",
                        "abstract": "Previous work has shown that DNNs with large depth $L$ and $L_{2}$-regularization are biased towards learning low-dimensional representations of the inputs, which can be interpreted as minimizing a notion of rank $R^{(0)}(f)$ of the learned function $f$, conjectured to be the Bottleneck rank. We compute finite depth corrections to this result, revealing a measure $R^{(1)}$ of regularity which bounds the pseudo-determinant of the Jacobian $\\left|Jf(x)\\right|_{+}$ and is subadditive under composition and addition. This formalizes a balance between learning low-dimensional representations and minimizing complexity/irregularity in the feature maps, allowing the network to learn the `right' inner dimension. Finally, we prove the conjectured bottleneck structure in the learned features as $L\\to\\infty$: for large depths, almost all hidden representations are approximately $R^{(0)}(f)$-dimensional, and almost all weight matrices $W_{\\ell}$ have $R^{(0)}(f)$ singular values close to 1 while the others are $O(L^{-\\frac{1}{2}})$. Interestingly, the use of large learning rates is required to guarantee an order $O(L)$ NTK which in turns guarantees infinite depth convergence of the representations of almost all layers."
                    },
                    {
                        "title": "Implicit bias of SGD in $L_{2}$-regularized linear DNNs: One-way jumps from high to low rank",
                        "abstract": "The $L_{2}$-regularized loss of Deep Linear Networks (DLNs) with more than one hidden layers has multiple local minima, corresponding to matrices with different ranks. In tasks such as matrix completion, the goal is to converge to the local minimum with the smallest rank that still fits the training data. While rank-underestimating minima can be avoided since they do not fit the data, GD might get stuck at rank-overestimating minima. We show that with SGD, there is always a probability to jump from a higher rank minimum to a lower rank one, but the probability of jumping back is zero. More precisely, we define a sequence of sets $B_{1}\\subset B_{2}\\subset\\cdots\\subset B_{R}$ so that $B_{r}$ contains all minima of rank $r$ or less (and not more) that are absorbing for small enough ridge parameters $\\lambda$ and learning rates $\\eta$: SGD has prob. 0 of leaving $B_{r}$, and from any starting point there is a non-zero prob. for SGD to go in $B_{r}$."
                    },
                    {
                        "title": "Hamiltonian Mechanics of Feature Learning: Bottleneck Structure in Leaky ResNets",
                        "abstract": "We study Leaky ResNets, which interpolate between ResNets ($\\tilde{L}=0$) and Fully-Connected nets ($\\tilde{L}\\to\\infty$) depending on an 'effective depth' hyper-parameter $\\tilde{L}$. In the infinite depth limit, we study 'representation geodesics' $A_{p}$: continuous paths in representation space (similar to NeuralODEs) from input $p=0$ to output $p=1$ that minimize the parameter norm of the network. We give a Lagrangian and Hamiltonian reformulation, which highlight the importance of two terms: a kinetic energy which favors small layer derivatives $\\partial_{p}A_{p}$ and a potential energy that favors low-dimensional representations, as measured by the 'Cost of Identity'. The balance between these two forces offers an intuitive understanding of feature learning in ResNets. We leverage this intuition to explain the emergence of a bottleneck structure, as observed in previous work: for large $\\tilde{L}$ the potential energy dominates and leads to a separation of timescales, where the representation jumps rapidly from the high dimensional inputs to a low-dimensional representation, move slowly inside the space of low-dimensional representations, before jumping back to the potentially high-dimensional outputs. Inspired by this phenomenon, we train with an adaptive layer step-size to adapt to the separation of timescales."
                    },
                    {
                        "title": "The asymptotic spectrum of the Hessian of DNN throughout training",
                        "abstract": "The dynamics of DNNs during gradient descent is described by the so-called Neural Tangent Kernel (NTK). In this article, we show that the NTK allows one to gain precise insight into the Hessian of the cost of DNNs. When the NTK is fixed during training, we obtain a full characterization of the asymptotics of the spectrum of the Hessian, at initialization and during training. In the so-called mean-field limit, where the NTK is not fixed during training, we describe the first two moments of the Hessian at initialization."
                    },
                    {
                        "title": "How DNNs break the Curse of Dimensionality: Compositionality and Symmetry Learning",
                        "abstract": "We show that deep neural networks (DNNs) can efficiently learn any composition of functions with bounded $F_{1}$-norm, which allows DNNs to break the curse of dimensionality in ways that shallow networks cannot. More specifically, we derive a generalization bound that combines a covering number argument for compositionality, and the $F_{1}$-norm (or the related Barron norm) for large width adaptivity. We show that the global minimizer of the regularized loss of DNNs can fit for example the composition of two functions $f^{*}=h\\circ g$ from a small number of observations, assuming $g$ is smooth/regular and reduces the dimensionality (e.g. $g$ could be the modulo map of the symmetries of $f^{*}$), so that $h$ can be learned in spite of its low regularity. The measures of regularity we consider is the Sobolev norm with different levels of differentiability, which is well adapted to the $F_{1}$ norm. We compute scaling laws empirically and observe phase transitions depending on whether $g$ or $h$ is harder to learn, as predicted by our theory."
                    },
                    {
                        "title": "DNN-Based Topology Optimisation: Spatial Invariance and Neural Tangent Kernel",
                        "abstract": "We study the Solid Isotropic Material Penalisation (SIMP) method with a density field generated by a fully-connected neural network, taking the coordinates as inputs. In the large width limit, we show that the use of DNNs leads to a filtering effect similar to traditional filtering techniques for SIMP, with a filter described by the Neural Tangent Kernel (NTK). This filter is however not invariant under translation, leading to visual artifacts and non-optimal shapes. We propose two embeddings of the input coordinates, which lead to (approximate) spatial invariance of the NTK and of the filter. We empirically confirm our theoretical observations and study how the filter size is affected by the architecture of the network. Our solution can easily be applied to any other coordinates-based generation method."
                    },
                    {
                        "title": "Understanding Layer-wise Contributions in Deep Neural Networks through Spectral Analysis",
                        "abstract": "Spectral analysis is a powerful tool, decomposing any function into simpler parts. In machine learning, Mercer's theorem generalizes this idea, providing for any kernel and input distribution a natural basis of functions of increasing frequency. More recently, several works have extended this analysis to deep neural networks through the framework of Neural Tangent Kernel. In this work, we analyze the layer-wise spectral bias of Deep Neural Networks and relate it to the contributions of different layers in the reduction of generalization error for a given target function. We utilize the properties of Hermite polynomials and Spherical Harmonics to prove that initial layers exhibit a larger bias towards high-frequency functions defined on the unit sphere. We further provide empirical results validating our theory in high dimensional datasets for Deep Neural Networks."
                    },
                    {
                        "title": "Which Frequencies do CNNs Need? Emergent Bottleneck Structure in Feature Learning",
                        "abstract": "We describe the emergence of a Convolution Bottleneck (CBN) structure in CNNs, where the network uses its first few layers to transform the input representation into a representation that is supported only along a few frequencies and channels, before using the last few layers to map back to the outputs. We define the CBN rank, which describes the number and type of frequencies that are kept inside the bottleneck, and partially prove that the parameter norm required to represent a function $f$ scales as depth times the CBN rank $f$. We also show that the parameter norm depends at next order on the regularity of $f$. We show that any network with almost optimal parameter norm will exhibit a CBN structure in both the weights and - under the assumption that the network is stable under large learning rate - the activations, which motivates the common practice of down-sampling; and we verify that the CBN results still hold with down-sampling. Finally we use the CBN structure to interpret the functions learned by CNNs on a number of tasks."
                    },
                    {
                        "title": "Order and Chaos: NTK views on DNN Normalization, Checkerboard and Boundary Artifacts",
                        "abstract": "We analyze architectural features of Deep Neural Networks (DNNs) using the so-called Neural Tangent Kernel (NTK), which describes the training and generalization of DNNs in the infinite-width setting. In this setting, we show that for fully-connected DNNs, as the depth grows, two regimes appear: \"order\", where the (scaled) NTK converges to a constant, and \"chaos\", where it converges to a Kronecker delta. Extreme order slows down training while extreme chaos hinders generalization. Using the scaled ReLU as a nonlinearity, we end up in the ordered regime. In contrast, Layer Normalization brings the network into the chaotic regime. We observe a similar effect for Batch Normalization (BN) applied after the last nonlinearity. We uncover the same order and chaos modes in Deep Deconvolutional Networks (DC-NNs). Our analysis explains the appearance of so-called checkerboard patterns and border artifacts. Moving the network into the chaotic regime prevents checkerboard patterns; we propose a graph-based parametrization which eliminates border artifacts; finally, we introduce a new layer-dependent learning rate to improve the convergence of DC-NNs. We illustrate our findings on DCGANs: the ordered regime leads to a collapse of the generator to a checkerboard mode, which can be avoided by tuning the nonlinearity to reach the chaotic regime. As a result, we are able to obtain good quality samples for DCGANs without BN."
                    },
                    {
                        "title": "Feature Learning in $L_{2}$-regularized DNNs: Attraction/Repulsion and Sparsity",
                        "abstract": "We study the loss surface of DNNs with $L_{2}$ regularization. We show that the loss in terms of the parameters can be reformulated into a loss in terms of the layerwise activations $Z_{\\ell}$ of the training set. This reformulation reveals the dynamics behind feature learning: each hidden representations $Z_{\\ell}$ are optimal w.r.t. to an attraction/repulsion problem and interpolate between the input and output representations, keeping as little information from the input as necessary to construct the activation of the next layer. For positively homogeneous non-linearities, the loss can be further reformulated in terms of the covariances of the hidden representations, which takes the form of a partially convex optimization over a convex cone.   This second reformulation allows us to prove a sparsity result for homogeneous DNNs: any local minimum of the $L_{2}$-regularized loss can be achieved with at most $N(N+1)$ neurons in each hidden layer (where $N$ is the size of the training set). We show that this bound is tight by giving an example of a local minimum that requires $N^{2}/4$ hidden neurons. But we also observe numerically that in more traditional settings much less than $N^{2}$ neurons are required to reach the minima."
                    },
                    {
                        "title": "Implicit Regularization of Random Feature Models",
                        "abstract": "Random Feature (RF) models are used as efficient parametric approximations of kernel methods. We investigate, by means of random matrix theory, the connection between Gaussian RF models and Kernel Ridge Regression (KRR). For a Gaussian RF model with $P$ features, $N$ data points, and a ridge $\\lambda$, we show that the average (i.e. expected) RF predictor is close to a KRR predictor with an effective ridge $\\tilde{\\lambda}$. We show that $\\tilde{\\lambda} > \\lambda$ and $\\tilde{\\lambda} \\searrow \\lambda$ monotonically as $P$ grows, thus revealing the implicit regularization effect of finite RF sampling. We then compare the risk (i.e. test error) of the $\\tilde{\\lambda}$-KRR predictor with the average risk of the $\\lambda$-RF predictor and obtain a precise and explicit bound on their difference. Finally, we empirically find an extremely good agreement between the test errors of the average $\\lambda$-RF predictor and $\\tilde{\\lambda}$-KRR predictor."
                    },
                    {
                        "title": "Saddle-to-Saddle Dynamics in Deep Linear Networks: Small Initialization Training, Symmetry, and Sparsity",
                        "abstract": "The dynamics of Deep Linear Networks (DLNs) is dramatically affected by the variance $\\sigma^2$ of the parameters at initialization $\\theta_0$. For DLNs of width $w$, we show a phase transition w.r.t. the scaling $\\gamma$ of the variance $\\sigma^2=w^{-\\gamma}$ as $w\\to\\infty$: for large variance ($\\gamma<1$), $\\theta_0$ is very close to a global minimum but far from any saddle point, and for small variance ($\\gamma>1$), $\\theta_0$ is close to a saddle point and far from any global minimum. While the first case corresponds to the well-studied NTK regime, the second case is less understood. This motivates the study of the case $\\gamma \\to +\\infty$, where we conjecture a Saddle-to-Saddle dynamics: throughout training, gradient descent visits the neighborhoods of a sequence of saddles, each corresponding to linear maps of increasing rank, until reaching a sparse global minimum. We support this conjecture with a theorem for the dynamics between the first two saddles, as well as some numerical experiments."
                    },
                    {
                        "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                        "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function $f_\\theta$ (which maps input vectors to output vectors) follows the kernel gradient of the functional cost (which is convex, in contrast to the parameter cost) w.r.t. a new kernel: the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and it stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We prove the positive-definiteness of the limiting NTK when the data is supported on the sphere and the non-linearity is non-polynomial. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function $f_\\theta$ follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                    },
                    {
                        "title": "Kernel Alignment Risk Estimator: Risk Prediction from Training Data",
                        "abstract": "We study the risk (i.e. generalization error) of Kernel Ridge Regression (KRR) for a kernel $K$ with ridge $\\lambda>0$ and i.i.d. observations. For this, we introduce two objects: the Signal Capture Threshold (SCT) and the Kernel Alignment Risk Estimator (KARE). The SCT $\\vartheta_{K,\\lambda}$ is a function of the data distribution: it can be used to identify the components of the data that the KRR predictor captures, and to approximate the (expected) KRR risk. This then leads to a KRR risk approximation by the KARE $\\rho_{K, \\lambda}$, an explicit function of the training data, agnostic of the true data distribution. We phrase the regression problem in a functional setting. The key results then follow from a finite-size analysis of the Stieltjes transform of general Wishart random matrices. Under a natural universality assumption (that the KRR moments depend asymptotically on the first two moments of the observations) we capture the mean and variance of the KRR predictor. We numerically investigate our findings on the Higgs and MNIST datasets for various classical kernels: the KARE gives an excellent approximation of the risk, thus supporting our universality assumption. Using the KARE, one can compare choices of Kernels and hyperparameters directly from the training set. The KARE thus provides a promising data-dependent procedure to select Kernels that generalize well."
                    },
                    {
                        "title": "Wide Neural Networks Trained with Weight Decay Provably Exhibit Neural Collapse",
                        "abstract": "Deep neural networks (DNNs) at convergence consistently represent the training data in the last layer via a highly symmetric geometric structure referred to as neural collapse. This empirical evidence has spurred a line of theoretical research aimed at proving the emergence of neural collapse, mostly focusing on the unconstrained features model. Here, the features of the penultimate layer are free variables, which makes the model data-agnostic and, hence, puts into question its ability to capture DNN training. Our work addresses the issue, moving away from unconstrained features and studying DNNs that end with at least two linear layers. We first prove generic guarantees on neural collapse that assume (i) low training error and balancedness of the linear layers (for within-class variability collapse), and (ii) bounded conditioning of the features before the linear part (for orthogonality of class-means, as well as their alignment with weight matrices). We then show that such assumptions hold for gradient descent training with weight decay: (i) for networks with a wide first layer, we prove low training error and balancedness, and (ii) for solutions that are either nearly optimal or stable under large learning rates, we additionally prove the bounded conditioning. Taken together, our results are the first to show neural collapse in the end-to-end training of DNNs."
                    },
                    {
                        "title": "Shallow diffusion networks provably learn hidden low-dimensional structure",
                        "abstract": "Diffusion-based generative models provide a powerful framework for learning to sample from a complex target distribution. The remarkable empirical success of these models applied to high-dimensional signals, including images and video, stands in stark contrast to classical results highlighting the curse of dimensionality for distribution recovery. In this work, we take a step towards understanding this gap through a careful analysis of learning diffusion models over the Barron space of single layer neural networks. In particular, we show that these shallow models provably adapt to simple forms of low dimensional structure, thereby avoiding the curse of dimensionality. We combine our results with recent analyses of sampling with diffusion models to provide an end-to-end sample complexity bound for learning to sample from structured distributions. Importantly, our results do not require specialized architectures tailored to particular latent structures, and instead rely on the low-index structure of the Barron space to adapt to the underlying distribution."
                    },
                    {
                        "title": "Disentangling feature and lazy training in deep neural networks",
                        "abstract": "Two distinct limits for deep learning have been derived as the network width $h\\rightarrow \\infty$, depending on how the weights of the last layer scale with $h$. In the Neural Tangent Kernel (NTK) limit, the dynamics becomes linear in the weights and is described by a frozen kernel $\\Theta$. By contrast, in the Mean-Field limit, the dynamics can be expressed in terms of the distribution of the parameters associated with a neuron, that follows a partial differential equation. In this work we consider deep networks where the weights in the last layer scale as $\\alpha h^{-1/2}$ at initialization. By varying $\\alpha$ and $h$, we probe the crossover between the two limits. We observe the previously identified regimes of lazy training and feature training. In the lazy-training regime, the dynamics is almost linear and the NTK barely changes after initialization. The feature-training regime includes the mean-field formulation as a limiting case and is characterized by a kernel that evolves in time, and learns some features. We perform numerical experiments on MNIST, Fashion-MNIST, EMNIST and CIFAR10 and consider various architectures. We find that (i) The two regimes are separated by an $\\alpha^*$ that scales as $h^{-1/2}$. (ii) Network architecture and data structure play an important role in determining which regime is better: in our tests, fully-connected networks perform generally better in the lazy-training regime, unlike convolutional networks. (iii) In both regimes, the fluctuations $\\delta F$ induced on the learned function by initial conditions decay as $\\delta F\\sim 1/\\sqrt{h}$, leading to a performance that increases with $h$. The same improvement can also be obtained at an intermediate width by ensemble-averaging several networks. (iv) In the feature-training regime we identify a time scale $t_1\\sim\\sqrt{h}\\alpha$, such that for $t\\ll t_1$ the dynamics is linear."
                    },
                    {
                        "title": "Geometry of the Loss Landscape in Overparameterized Neural Networks: Symmetries and Invariances",
                        "abstract": "We study how permutation symmetries in overparameterized multi-layer neural networks generate `symmetry-induced' critical points. Assuming a network with $ L $ layers of minimal widths $ r_1^*, \\ldots, r_{L-1}^* $ reaches a zero-loss minimum at $ r_1^*! \\cdots r_{L-1}^*! $ isolated points that are permutations of one another, we show that adding one extra neuron to each layer is sufficient to connect all these previously discrete minima into a single manifold. For a two-layer overparameterized network of width $ r^*+ h =: m $ we explicitly describe the manifold of global minima: it consists of $ T(r^*, m) $ affine subspaces of dimension at least $ h $ that are connected to one another. For a network of width $m$, we identify the number $G(r,m)$ of affine subspaces containing only symmetry-induced critical points that are related to the critical points of a smaller network of width $r<r^*$. Via a combinatorial analysis, we derive closed-form formulas for $ T $ and $ G $ and show that the number of symmetry-induced critical subspaces dominates the number of affine subspaces forming the global minima manifold in the mildly overparameterized regime (small $ h $) and vice versa in the vastly overparameterized regime ($h \\gg r^*$). Our results provide new insights into the minimization of the non-convex loss function of overparameterized neural networks."
                    }
                ]
            }
        }
    },
    "2405.14530": {
        "paper_data": {
            "title": "Multistable Shape from Shading Emerges from Patch Diffusion",
            "url": "http://arxiv.org/abs/2405.14530v1",
            "arxiv_id": "2405.14530",
            "authors": [
                "Xinran Nicole Han",
                "Todd Zickler",
                "Ko Nishino"
            ],
            "abstract": "Models for monocular shape reconstruction of surfaces with diffuse reflection -- shape from shading -- ought to produce distributions of outputs, because there are fundamental mathematical ambiguities of both continuous (e.g., bas-relief) and discrete (e.g., convex/concave) varieties which are also experienced by humans. Yet, the outputs of current models are limited to point estimates or tight distributions around single modes, which prevent them from capturing these effects. We introduce a model that reconstructs a multimodal distribution of shapes from a single shading image, which aligns with the human experience of multistable perception. We train a small denoising diffusion process to generate surface normal fields from $16\\times 16$ patches of synthetic images of everyday 3D objects. We deploy this model patch-wise at multiple scales, with guidance from inter-patch shape consistency constraints. Despite its relatively small parameter count and predominantly bottom-up structure, we show that multistable shape explanations emerge from this model for ''ambiguous'' test images that humans experience as being multistable. At the same time, the model produces veridical shape estimates for object-like images that include distinctive occluding contours and appear less ambiguous. This may inspire new architectures for stochastic 3D shape perception that are more efficient and better aligned with human experience.",
            "introduction": "   1 Introduction  From chiaroscuro in Renaissance paintings to the interplay of light and dark in Ansel Adams\u2019 photographs, humans are masters at perceiving three-dimensional shape from variations of image intensity\u2014shading\u2014from a single image alone. Even though our visual experience is dominated by everyday objects, our perception of shape from shading has been shown to robustly generalize to synthetic \u201cnon-ecological\u201d images invented by vision scientists. Some of these images have ambiguous (e.g., convex/concave) interpretations and lead to multistable perceptions, where one\u2019s impression of 3D shape alternates between two or more competing explanations. Figure\u00a01 shows an example adapted from\u00a0[29], which is alternately interpreted as an indentation in a convex surrounding surface or a protrusion on a concave surrounding surface. Both explanations are physically correct because the same image can be generated from either shape under different lighting conditions.   How can a computational model acquire this human ability to capture multiple underlying shape explanations? An algorithmic suggestion comes from Marr\u2019s principle of least commitment\u00a0[32, 33], which requires not doing anything that may later have to be undone. But this is in contrast to many computer vision methods for shape from shading, including SIRFS [4] and recent neural feed-forward models [50, 53], which are deterministic and produce a single, best estimate of shape. These types of models commit to one explanation based on priors that are either designed or learned from a dataset, and they cannot express multiple interpretations of an ambiguous image. They are unlikely to be good models for the mechanisms that underlie multistable perception.   Figure 1:  Many shapes (left) can explain the same image (middle) under different lighting, including flattened and tilted versions and convex/concave flips. The concave/convex flip in this example is also perceived by humans, often aided by rotating the image clockwise by 90 degrees. Previous methods for inferring surface normals (SIRFS\u00a0[4], Derender3D\u00a0[50], Wonder3D\u00a0[31]) or depth (Marigold\u00a0[24], Depth Anything\u00a0[53]) produce a single shape estimate or a unimodal distribution. Our model produces a multimodal distribution that matches the perceived flip. (Image adapted from [29].)   Instead, we approach monocular shape inference as a conditional generative process, and inspired by a long history of shape from shading with local patches\u00a0[15, 22, 7, 27, 51, 17, 14], we present a bottom-up, patch-based diffusion model that can mimic multistable perception for ambiguous images of diffuse shading. Notably, our model is trained using images of familiar object-like shapes and has no prior experience with the ambiguous images that we use for testing. It is built on a small conditional diffusion process that is pre-trained to predict surface normals from 16\u00d716161616\\times 1616 \u00d7 16 image patches. When this patch process is applied in a non-overlapping manner with simple inter-patch shape consistency constraints, our model coordinates those patches in the sampling process and turns out to capture global ambiguities that are experienced by humans.   An important attribute of our model is the way it handles lighting. It builds on the mathematical observation that shape perception can precede lighting inference\u00a0[28]. It also adheres to the philosophy that inferred lighting cannot and should not be precise because of effects like global illumination\u00a0[47]. Our model achieves this by",
            "references": [
                {
                    "title": "Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data",
                    "abstract": "This work presents Depth Anything11While the grammatical soundness of this name may be questionable, we treat it as a whole and pay homage to Segment Anything [26]., a highly practical solution for robust monocular depth estimation. Without pursuing novel technical modules, we aim to build a simple yet powerful foundation model dealing with any images under any circumstances. To this end, we scale up the dataset by designing a data engine to collect and automatically annotate large-scale unlabeled data (~62M), which significantly enlarges the data coverage and thus is able to reduce the generalization error. We investigate two simple yet effective strategies that make data scaling-up promising. First, a more challenging optimization target is created by leveraging data augmentation tools. It compels the model to actively seek extra visual knowledge and acquire robust representations. Second, an auxiliary supervision is developed to enforce the model to inherit rich semantic priors from pre-trained encoders. We evaluate its zero-shot capabilities extensively, including six public datasets and randomly captured photos. It demonstrates impressive generalization ability (Figure 1). Further, through fine-tuning it with metric depth information from NYUv2 and KITTI, new SOTAs are set. Our better depth model also results in a better depth-conditioned ControlNet. Our models are released here."
                },
                {
                    "title": "Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation",
                    "abstract": "Monocular depth estimation is a fundamental computer vision task. Recovering 3D depth from a single image is geometrically ill-posed and requires scene understanding, so it is not surprising that the rise of deep learning has led to a breakthrough. The impressive progress of monocular depth estimators has mirrored the growth in model capacity, from relatively modest CNNs to large Transformer architectures. Still, monocular depth estimators tend to struggle when presented with images with unfamiliar content and layout, since their knowledge of the visual world is restricted by the data seen during training, and challenged by zero-shot generalization to new domains. This motivates us to explore whether the extensive priors captured in recent generative diffusion models can enable better, more generalizable depth estimation. We introduce Marigold, a method for affine-invariant monocular depth estimation that is derived from Stable Diffusion and retains its rich prior knowledge. The estimator can be fine-tuned in a couple of days on a single GPU using only synthetic training data. It delivers state-of-the-art performance across a wide range of datasets, including over 20% performance gains in specific cases. Project page: https://marigoldmonodepth.github.io."
                },
                {
                    "title": "Wonder3D: Single Image to 3D Using Cross-Domain Diffusion",
                    "abstract": "In this work, we introduce Wonder3D, a novel method for efficiently generating high-fidelity textured meshes from single-view images. Recent methods based on Score Distillation Sampling (SDS) have shown the potential to recover 3D geometry from 2D diffusion priors, but they typically suffer from time-consuming per-shape optimization and inconsistent geometry. In contrast, certain works di-rectly produce 3D information via fast network inferences, but their results are often of low quality and lack geometric details. To holistically improve the quality, consistency, and efficiency of single-view reconstruction tasks, we pro-pose a cross-domain diffusion model that generates multi-view normal maps and the corresponding color images. To ensure the consistency of generation, we employ a multi-view cross-domain attention mechanism that facilitates information exchange across views and modalities. Lastly, we introduce a geometry-aware normal fusion algorithm that extracts high-quality surfaces from the multi-view 2D representations in only 2 r-;\u00bb 3 minutes. Our extensive evaluations demonstrate that our method achieves high-quality reconstruction results, robust generalization, and good efficiency compared to prior works."
                },
                {
                    "title": "Patched Denoising Diffusion Models For High-Resolution Image Synthesis",
                    "abstract": "We propose an effective denoising diffusion model for generating high-resolution images (e.g., 1024$\\times$512), trained on small-size image patches (e.g., 64$\\times$64). We name our algorithm Patch-DM, in which a new feature collage strategy is designed to avoid the boundary artifact when synthesizing large-size images. Feature collage systematically crops and combines partial features of the neighboring patches to predict the features of a shifted image patch, allowing the seamless generation of the entire image due to the overlap in the patch feature space. Patch-DM produces high-quality image synthesis results on our newly collected dataset of nature images (1024$\\times$512), as well as on standard benchmarks of smaller sizes (256$\\times$256), including LSUN-Bedroom, LSUN-Church, and FFHQ. We compare our method with previous patch-based generation methods and achieve state-of-the-art FID scores on all four datasets. Further, Patch-DM also reduces memory complexity compared to the classic diffusion models."
                },
                {
                    "title": "Solving Inverse Problems with Latent Diffusion Models via Hard Data Consistency",
                    "abstract": "Diffusion models have recently emerged as powerful generative priors for solving inverse problems. However, training diffusion models in the pixel space are both data-intensive and computationally demanding, which restricts their applicability as priors for high-dimensional real-world data such as medical images. Latent diffusion models, which operate in a much lower-dimensional space, offer a solution to these challenges. However, incorporating latent diffusion models to solve inverse problems remains a challenging problem due to the nonlinearity of the encoder and decoder. To address these issues, we propose \\textit{ReSample}, an algorithm that can solve general inverse problems with pre-trained latent diffusion models. Our algorithm incorporates data consistency by solving an optimization problem during the reverse sampling process, a concept that we term as hard data consistency. Upon solving this optimization problem, we propose a novel resampling scheme to map the measurement-consistent sample back onto the noisy data manifold and theoretically demonstrate its benefits. Lastly, we apply our algorithm to solve a wide range of linear and nonlinear inverse problems in both natural and medical images, demonstrating that our approach outperforms existing state-of-the-art approaches, including those based on pixel-space diffusion models."
                },
                {
                    "title": "Restart Sampling for Improving Generative Processes",
                    "abstract": "Generative processes that involve solving differential equations, such as diffusion models, frequently necessitate balancing speed and quality. ODE-based samplers are fast but plateau in performance while SDE-based samplers deliver higher sample quality at the cost of increased sampling time. We attribute this difference to sampling errors: ODE-samplers involve smaller discretization errors while stochasticity in SDE contracts accumulated errors. Based on these findings, we propose a novel sampling algorithm called Restart in order to better balance discretization errors and contraction. The sampling method alternates between adding substantial noise in additional forward steps and strictly following a backward ODE. Empirically, Restart sampler surpasses previous SDE and ODE samplers in both speed and accuracy. Restart not only outperforms the previous best SDE results, but also accelerates the sampling speed by 10-fold / 2-fold on CIFAR-10 / ImageNet $64 \\times 64$. In addition, it attains significantly better sample quality than ODE samplers within comparable sampling times. Moreover, Restart better balances text-image alignment/visual quality versus diversity than previous samplers in the large-scale text-to-image Stable Diffusion model pre-trained on LAION $512 \\times 512$. Code is available at https://github.com/Newbeeer/diffusion_restart_sampling"
                },
                {
                    "title": "SyncDiffusion: Coherent Montage via Synchronized Joint Diffusions",
                    "abstract": "The remarkable capabilities of pretrained image diffusion models have been utilized not only for generating fixed-size images but also for creating panoramas. However, naive stitching of multiple images often results in visible seams. Recent techniques have attempted to address this issue by performing joint diffusions in multiple windows and averaging latent features in overlapping regions. However, these approaches, which focus on seamless montage generation, often yield incoherent outputs by blending different scenes within a single image. To overcome this limitation, we propose SyncDiffusion, a plug-and-play module that synchronizes multiple diffusions through gradient descent from a perceptual similarity loss. Specifically, we compute the gradient of the perceptual loss using the predicted denoised images at each denoising step, providing meaningful guidance for achieving coherent montages. Our experimental results demonstrate that our method produces significantly more coherent outputs compared to previous methods (66.35% vs. 33.65% in our user study) while still maintaining fidelity (as assessed by GIQA) and compatibility with the input prompt (as measured by CLIP score). We further demonstrate the versatility of our method across three plug-and-play applications: layout-guided image generation, conditional image generation and 360-degree panorama generation. Our project page is at https://syncdiffusion.github.io."
                },
                {
                    "title": "Patch Diffusion: Faster and More Data-Efficient Training of Diffusion Models",
                    "abstract": "Diffusion models are powerful, but they require a lot of time and data to train. We propose Patch Diffusion, a generic patch-wise training framework, to significantly reduce the training time costs while improving data efficiency, which thus helps democratize diffusion model training to broader users. At the core of our innovations is a new conditional score function at the patch level, where the patch location in the original image is included as additional coordinate channels, while the patch size is randomized and diversified throughout training to encode the cross-region dependency at multiple scales. Sampling with our method is as easy as in the original diffusion model. Through Patch Diffusion, we could achieve $\\mathbf{\\ge 2\\times}$ faster training, while maintaining comparable or better generation quality. Patch Diffusion meanwhile improves the performance of diffusion models trained on relatively small datasets, $e.g.$, as few as 5,000 images to train from scratch. We achieve outstanding FID scores in line with state-of-the-art benchmarks: 1.77 on CelebA-64$\\times$64, 1.93 on AFHQv2-Wild-64$\\times$64, and 2.72 on ImageNet-256$\\times$256. We share our code and pre-trained models at https://github.com/Zhendong-Wang/Patch-Diffusion."
                },
                {
                    "title": "DINOv2: Learning Robust Visual Features without Supervision",
                    "abstract": "The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels."
                },
                {
                    "title": "MultiDiffusion: Fusing Diffusion Paths for Controlled Image Generation",
                    "abstract": "Recent advances in text-to-image generation with diffusion models present transformative capabilities in image quality. However, user controllability of the generated image, and fast adaptation to new tasks still remains an open challenge, currently mostly addressed by costly and long re-training and fine-tuning or ad-hoc adaptations to specific image generation tasks. In this work, we present MultiDiffusion, a unified framework that enables versatile and controllable image generation, using a pre-trained text-to-image diffusion model, without any further training or finetuning. At the center of our approach is a new generation process, based on an optimization task that binds together multiple diffusion generation processes with a shared set of parameters or constraints. We show that MultiDiffusion can be readily applied to generate high quality and diverse images that adhere to user-provided controls, such as desired aspect ratio (e.g., panorama), and spatial guiding signals, ranging from tight segmentation masks to bounding boxes. Project webpage: https://multidiffusion.github.io"
                },
                {
                    "title": "Diffusion Posterior Sampling for General Noisy Inverse Problems",
                    "abstract": "Diffusion models have been recently studied as powerful generative inverse problem solvers, owing to their high quality reconstructions and the ease of combining existing iterative solvers. However, most works focus on solving simple linear inverse problems in noiseless settings, which significantly under-represents the complexity of real-world problems. In this work, we extend diffusion solvers to efficiently handle general noisy (non)linear inverse problems via approximation of the posterior sampling. Interestingly, the resulting posterior sampling scheme is a blended version of diffusion sampling with the manifold constrained gradient without a strict measurement consistency projection step, yielding a more desirable generative path in noisy settings compared to the previous studies. Our method demonstrates that diffusion models can incorporate various measurement noise statistics such as Gaussian and Poisson, and also efficiently handle noisy nonlinear inverse problems such as Fourier phase retrieval and non-uniform deblurring. Code available at https://github.com/DPS2022/diffusion-posterior-sampling"
                },
                {
                    "title": "Restoring Vision in Adverse Weather Conditions With Patch-Based Denoising Diffusion Models",
                    "abstract": "Image restoration under adverse weather conditions has been of significant interest for various computer vision applications. Recent successful methods rely on the current progress in deep neural network architectural designs (e.g., with vision transformers). Motivated by the recent progress achieved with state-of-the-art conditional generative models, we present a novel patch-based image restoration algorithm based on denoising diffusion probabilistic models. Our patch-based diffusion modeling approach enables size-agnostic image restoration by using a guided denoising process with smoothed noise estimates across overlapping patches during inference. We empirically evaluate our model on benchmark datasets for image desnowing, combined deraining and dehazing, and raindrop removal. We demonstrate our approach to achieve state-of-the-art performances on both weather-specific and multi-weather image restoration, and experimentally show strong generalization to real-world test images."
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "Universal Photometric Stereo Network using Global Lighting Contexts",
                    "abstract": "This paper tackles a new photometric stereo task, named universal photometric stereo. Unlike existing tasks that assumed specific physical lighting models; hence, drastically limited their usability, a solution algorithm of this task is supposed to work for objects with diverse shapes and materials under arbitrary lighting variations without assum-ig any specific models. To solve this extremely challenging task, we present a purely data-driven method, which eliminates the prior assumption of lighting by replacing the recovery of physical lighting parameters with the extraction of the generic lighting representation, named global lighting contexts. We use them like lighting parameters in a calibrated photometric stereo network to recover surface normal vectors pixelwisely. To adapt our network to a wide variety of shapes, materials and lightings, it is trained on a new synthetic dataset which simulates the appearance of objects in the wild. Our method is compared with other state-of-the-art uncalibrated photometric stereo methods on our test data to demonstrate the significance of our method."
                },
                {
                    "title": "Google Scanned Objects: A High-Quality Dataset of 3D Scanned Household Items",
                    "abstract": "Interactive 3D simulations have enabled break-throughs in robotics and computer vision, but simulating the broad diversity of environments needed for deep learning requires large corpora of photo-realistic 3D object models. To address this need, we present Google Scanned Objects, an open-source collection of over one thousand 3D-scanned household items released under a Creative Commons license; these models are preprocessed for use in Ignition Gazebo and the Bullet simulation platforms, but are easily adaptable to other simulators. We describe our object scanning and curation pipeline, then provide statistics about the contents of the dataset and its usage. We hope that the diversity, quality, and flexibility of Google Scanned Objects will lead to advances in interactive simulation, synthetic perception, and robotic learning."
                },
                {
                    "title": "De-rendering 3D Objects in the Wild",
                    "abstract": "With increasing focus on augmented and virtual reality (XR) applications comes the demand for algorithms that can lift objects from images into representations that are suitable for a wide variety of related 3D tasks. Large-scale deployment of XR devices and applications means that we cannot solely rely on supervised learning, as collecting and annotating data for the unlimited variety of objects in the real world is infeasible. We present a weakly supervised method that is able to decompose a single image of an object into shape (depth and normals), material (albedo, reflectivity and shininess) and global lighting parameters. For training, the method only relies on a rough initial shape estimate of the training objects to bootstrap the learning process. This shape supervision can come for example from a pretrained depth network or-more generically-from a traditional structure-from-motion pipeline. In our experiments, we show that the method can successfully de-render 2D images into a decomposed 3D representation and generalizes to unseen object categories. Since in-the-wild evaluation is difficult due to the lack of ground truth data, we also introduce a photo-realistic synthetic test set that allows for quantitative evaluation. Please find our project page at: https://github.com/Brummi/derender3d"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "DiffusionCLIP: Text-Guided Diffusion Models for Robust Image Manipulation",
                    "abstract": "Recently, GAN inversion methods combined with Contrastive Language-Image Pretraining (CLIP) enables zeroshot image manipulation guided by text prompts. However, their applications to diverse real images are still difficult due to the limited GAN inversion capability. Specifically, these approaches often have difficulties in reconstructing images with novel poses, views, and highly variable contents compared to the training data, altering object identity, or producing unwanted image artifacts. To mitigate these problems and enable faithful manipulation of real images, we propose a novel method, dubbed DiffusionCLIP, that performs textdriven image manipulation using diffusion models. Based on full inversion capability and high-quality image generation power of recent diffusion models, our method performs zeroshot image manipulation successfully even between unseen domains and takes another step towards general application by manipulating images from a widely varying ImageNet dataset. Furthermore, we propose a novel noise combination method that allows straightforward multi-attribute manipulation. Extensive experiments and human evaluation confirmed robust and superior manipulation performance of our methods compared to the existing baselines. Code is available at https://github.com/gwang-kim/DiffusionCLIP.git"
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "Vision Transformers for Dense Prediction",
                    "abstract": "We introduce dense prediction transformers, an architecture that leverages vision transformers in place of convolutional networks as a backbone for dense prediction tasks. We assemble tokens from various stages of the vision transformer into image-like representations at various resolutions and progressively combine them into full-resolution predictions using a convolutional decoder. The transformer backbone processes representations at a constant and relatively high resolution and has a global receptive field at every stage. These properties allow the dense prediction transformer to provide finer-grained and more globally coherent predictions when compared to fully-convolutional networks. Our experiments show that this architecture yields substantial improvements on dense prediction tasks, especially when a large amount of training data is available. For monocular depth estimation, we observe an improvement of up to 28% in relative performance when compared to a state-of-the-art fully-convolutional network. When applied to semantic segmentation, dense prediction transformers set a new state of the art on ADE20K with 49.02% mIoU. We further show that the architecture can be fine-tuned on smaller datasets such as NYUv2, KITTI, and Pascal Context where it also sets the new state of the art. Our models are available at https://github.com/intel-isl/DPT."
                },
                {
                    "title": "Improved Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) are a class of generative models which have recently been shown to produce excellent samples. We show that with a few simple modifications, DDPMs can also achieve competitive log-likelihoods while maintaining high sample quality. Additionally, we find that learning variances of the reverse diffusion process allows sampling with an order of magnitude fewer forward passes with a negligible difference in sample quality, which is important for the practical deployment of these models. We additionally use precision and recall to compare how well DDPMs and GANs cover the target distribution. Finally, we show that the sample quality and likelihood of these models scale smoothly with model capacity and training compute, making them easily scalable. We release our code at https://github.com/openai/improved-diffusion"
                },
                {
                    "title": "Outdoor Inverse Rendering From a Single Image Using Multiview Self-Supervision",
                    "abstract": "In this paper we show how to perform scene-level inverse rendering to recover shape, reflectance and lighting from a single, uncontrolled image using a fully convolutional neural network. The network takes an RGB image as input, regresses albedo, shadow and normal maps from which we infer least squares optimal spherical harmonic lighting coefficients. Our network is trained using large uncontrolled multiview and timelapse image collections without ground truth. By incorporating a differentiable renderer, our network can learn from self-supervision. Since the problem is ill-posed we introduce additional supervision. Our key insight is to perform offline multiview stereo (MVS) on images containing rich illumination variation. From the MVS pose and depth maps, we can cross project between overlapping views such that Siamese training can be used to ensure consistent estimation of photometric invariants. MVS depth also provides direct coarse supervision for normal map estimation. We believe this is the first attempt to use MVS supervision for learning inverse rendering. In addition, we learn a statistical natural illumination prior. We evaluate performance on inverse rendering, normal map estimation and intrinsic image decomposition benchmarks."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "A Lighting-Invariant Point Processor for Shading",
                    "abstract": "Under the conventional diffuse shading model with unknown directional lighting, the set of quadratic surface shapes that are consistent with the spatial derivatives of intensity at a single image point is a two-dimensional algebraic variety embedded in the five-dimensional space of quadratic shapes. We describe the geometry of this variety, and we introduce a concise feedforward model that computes an explicit, differentiable approximation of the variety from the intensity and its derivatives at any single image point. The result is a parallelizable processor that operates at each image point and produces a lighting-invariant descriptor of the continuous set of compatible surface shapes at the point. We describe two applications of this processor: two-shot uncalibrated photometric stereo and quadratic-surface shape from shading."
                },
                {
                    "title": "From boundaries to bumps: When closed (extremal) contours are critical",
                    "abstract": "Abstract Invariants underlying shape inference are elusive: A variety of shapes can give rise to the same image, and a variety of images can be rendered from the same shape. The occluding contour is a rare exception: It has both image salience, in terms of isophotes, and surface meaning, in terms of surface normal. We relax the notion of occluding contour and, more accurately, the rim on the object that projects to it, to define closed extremal curves. This new shape descriptor is invariant over different renderings. It exists at the topological level, which guarantees an image-based counterpart. It surrounds bumps and dents, as well as common interior shape components, and formalizes the qualitative nature of bump perception. The invariants are biologically computable, unify shape inferences from shading and specular materials, and predict new phenomena in bump and dent perception. Most important, working at the topological level allows us to capture the elusive aspect of bump boundaries."
                },
                {
                    "title": "Efficient Attention: Attention with Linear Complexities",
                    "abstract": "Dot-product attention has wide applications in computer vision and natural language processing. However, its memory and computational costs grow quadratically with the input size. Such growth prohibits its application on high- resolution inputs. To remedy this drawback, this paper proposes a novel efficient attention mechanism equivalent to dot-product attention but with substantially less memory and computational costs. Its resource efficiency allows more widespread and flexible integration of attention modules into a network, which leads to better accuracies. Empirical evaluations demonstrated the effectiveness of its advantages. Efficient attention modules brought significant performance boosts to object detectors and instance segmenters on MS-COCO 2017. Further, the resource efficiency democratizes attention to complex models, where high costs prohibit the use of dot-product attention. As an exemplar, a model with efficient attention achieved state-of- the-art accuracies for stereo depth estimation on the Scene Flow dataset. Code is available at https://github.com/cmsflash/efficient-attention."
                },
                {
                    "title": "InverseRenderNet: Learning Single Image Inverse Rendering",
                    "abstract": "We show how to train a fully convolutional neural network to perform inverse rendering from a single, uncontrolled image. The network takes an RGB image as input, regresses albedo and normal maps from which we compute lighting coefficients. Our network is trained using large uncontrolled image collections without ground truth. By incorporating a differentiable renderer, our network can learn from self-supervision. Since the problem is ill-posed we introduce additional supervision: 1. We learn a statistical natural illumination prior, 2. Our key insight is to perform offline multiview stereo (MVS) on images containing rich illumination variation. From the MVS pose and depth maps, we can cross project between overlapping views such that Siamese training can be used to ensure consistent estimation of photometric invariants. MVS depth also provides direct coarse supervision for normal map estimation. We believe this is the first attempt to use MVS supervision for learning inverse rendering."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "A Multiscale Variable-Grouping Framework for MRF Energy Minimization",
                    "abstract": "We present a multiscale approach for minimizing the energy associated with Markov Random Fields (MRFs) with energy functions that include arbitrary pairwise potentials. The MRF is represented on a hierarchy of successively coarser scales, where the problem on each scale is itself an MRF with suitably defined potentials. These representations are used to construct an efficient multiscale algorithm that seeks a minimal-energy solution to the original problem. The algorithm is iterative and features a bidirectional crosstalk between fine and coarse representations. We use consistency criteria to guarantee that the energy is nonincreasing throughout the iterative process. The algorithm is evaluated on real-world datasets, achieving competitive performance in relatively short run-times."
                },
                {
                    "title": "Shape and Illumination from Shading using the Generic Viewpoint Assumption",
                    "abstract": "The Generic Viewpoint Assumption (GVA) states that the position of the viewer or the light in a scene is not special. Thus, any estimated parameters from an observation should be stable under small perturbations such as object, viewpoint or light positions. The GVA has been analyzed and quantified in previous works, but has not been put to practical use in actual vision tasks. In this paper, we show how to utilize the GVA to estimate shape and illumination from a single shading image, without the use of other priors. We propose a novel linearized Spherical Harmonics (SH) shading model which enables us to obtain a computationally efficient form of the GVA term. Together with a data term, we build a model whose unknowns are shape and SH illumination. The model parameters are estimated using the Alternating Direction Method of Multipliers embedded in a multi-scale estimation framework. In this prior-free framework, we obtain competitive shape and illumination estimation results under a variety of models and lighting conditions, requiring fewer assumptions than competing methods."
                },
                {
                    "title": "Is the perception of 3D shape from shading based on assumed reflectance and illumination?",
                    "abstract": "The research described in the present article was designed to compare three types of image shading: one generated with a Lambertian BRDF and homogeneous illumination such that image intensity was determined entirely by local surface orientation irrespective of position; one that was textured with a linear intensity gradient, such that image intensity was determined entirely by local surface position irrespective of orientation; and another that was generated with a Lambertian BRDF and inhomogeneous illumination such that image intensity was influenced by both position and orientation. A gauge figure adjustment task was used to measure observers' perceptions of local surface orientation on the depicted surfaces, and the probe points included 60 pairs of regions that both had the same orientation. The results show clearly that observers' perceptions of these three types of stimuli were remarkably similar, and that probe regions with similar apparent orientations could have large differences in image intensity. This latter finding is incompatible with any process for computing shape from shading that assumes any plausible reflectance function combined with any possible homogeneous illumination."
                },
                {
                    "title": "How Shading Constrains Surface Patches without Knowledge of Light Sources",
                    "abstract": "Shape-from-shading (SFS) is a classical inverse problem in computer and human vision. This shape reconstruction problem is inherently ill-posed. We show that the isophotes on smooth surfaces with Lambertian reflectance can be directly related to surface properties without consideration of the light source. Using techniques from modern differential geometry, we derive relationships between the curvature of the isophotes and the shape operator for the surface. Neurobiology motivates the geometric approach, and our calculations allow us to characterize the matching local family of surfaces that can result from any given shading patch. We illustrate the local ambiguity in several examples."
                },
                {
                    "title": "From Shading to Local Shape",
                    "abstract": "We develop a framework for extracting a concise representation of the shape information available from diffuse shading in a small image patch. This produces a mid-level scene descriptor, comprised of local shape distributions that are inferred separately at every image patch across multiple scales. The framework is based on a quadratic representation of local shape that, in the absence of noise, has guarantees on recovering accurate local shape and lighting. And when noise is present, the inferred local shape distributions provide useful shape information without over-committing to any particular image explanation. These local shape distributions naturally encode the fact that some smooth diffuse regions are more informative than others, and they enable efficient and robust reconstruction of object-scale shape. Experimental results show that this approach to surface reconstruction compares well against the state-of-art on both synthetic images and captured photographs."
                },
                {
                    "title": "Characterizing Ambiguity in Light Source Invariant Shape from Shading",
                    "abstract": "Shape from shading is a classical inverse problem in computer vision. This shape reconstruction problem is inherently ill-dened; it depends on the assumed light source direction. We introduce a novel mathematical formulation for calculating local surface shape based on covariant derivatives of the shading ow eld, rather than the customary integral minimization or P.D.E approaches. On smooth surfaces, we show second derivatives of brightness are independent of the light sources and can be directly related to surface properties. We use these measurements to dene the matching local family of surfaces that can result from any given shading patch, changing the emphasis to characterizing ambiguity in the problem. We give an example of how these local surface ambiguities collapse along certain image contours and how this can be used for the reconstruction problem."
                },
                {
                    "title": "The shading cue in context",
                    "abstract": "The shading cue is supposed to be a major factor in monocular stereopsis. However, the hypothesis is hardly corroborated by available data. For instance, the conventional stimulus used in perception research, which involves a circular disk with monotonic luminance gradient on a uniform surround, is theoretically \u2018explained\u2019 by any quadric surface, including spherical caps or cups (the conventional response categories), cylindrical ruts or ridges, and saddle surfaces. Whereas cylindrical ruts or ridges are reported when the outline is changed from circular to square, saddle surfaces are never reported. We introduce a method that allows us to differentiate between such possible responses. We report observations on a number of variations of the conventional stimulus, including variations of shape and quality of the boundary, and contexts that allow the observer to infer illumination direction. We find strong and expected influences of outline shape, but, perhaps surprisingly, we fail to find any influence of context, and only partial influence of outline quality. Moreover, we report appreciable differences within the generic population. We trace some of the idiosyncrasies (as compared to shape from shading algorithms) of the human observer to generic properties of the environment, in particular the fact that many objects are limited in size and elliptically convex over most of their boundaries."
                },
                {
                    "title": "Polynomial shape from shading",
                    "abstract": "We examine the shape from shading problem without boundary conditions as a polynomial system. This view allows, in generic cases, a complete solution for ideal polyhedral objects. For the general case we propose a semidefinite programming relaxation procedure, and an exact line search iterative procedure with a new smoothness term that favors folds at edges. We use this numerical technique to inspect shading ambiguities."
                },
                {
                    "title": "Examplar-based Shape from Shading",
                    "abstract": "Traditional Shape-from-Shading (SFS) techniques aim to solve an under-constrained problem: estimating depth map from one single image. The results are usually brittle from real images containing detailed shapes. Inspired by recent advances in texture synthesis, we present an exemplar-based approach to improve the robustness and accuracy of SFS. In essence, we utilize an appearance database synthesized from known 3D models where each image pixel is associated with its ground-truth normal. The input image is compared against the images in the database to find the most likely normals. The prior knowledge from the database is formulated as an additional cost term under an energy minimization framework to solve the depth map. Using a generic small database consisting of 50 spheres with different radius, our approach has demonstrated its capability to obviously improve the reconstruction quality from both synthetic and real images with different shapes, in particular those with small details."
                },
                {
                    "title": "Example Based 3D Reconstruction from Single 2D Images",
                    "abstract": "We present a novel solution to the problem of depth reconstruction from a single image. Single view 3D reconstruction is an ill-posed problem. We address this problem by using an example-based synthesis approach. Our method uses a database of objects from a single class (e.g. hands, human figures) containing example patches of feasible mappings from the appearance to the depth of each object. Given an image of a novel object, we combine the known depths of patches from similar objects to produce a plausible depth estimate. This is achieved by optimizing a global target function representing the likelihood of the candidate depth. We demonstrate how the variability of 3D shapes and their poses can be handled by updating the example database on-the-fly. In addition, we show how we can employ our method for the novel task of recovering an estimate for the occluded backside of the imaged objects. Finally, we present results on a variety of object classes and a range of imaging conditions."
                },
                {
                    "title": "Shape Information From Shading: A Theory About Human Perception",
                    "abstract": "I present evidence that people assume a simple, linear reflectance function when interpreting shading information. Using this reflectance function I derive a closed-form solution to the problem of extracting shape information from image shading. The solution does not employ an assumption about surface smoothness and so is directly applicable to complex natural surfaces such as hair or cloth. A simple biological mechanism is proposed to implement this recovery of shape. It is shown that this simple mechanism can also extract significant shape information from line drawings."
                },
                {
                    "title": "A Method for Enforcing Integrability in Shape from Shading Algorithms",
                    "abstract": "An approach for enforcing integrability, a particular implementation of the approach, an example of its application to extending an existing shape-from-shading algorithm, and experimental results showing the improvement that results from enforcing integrability are presented. A possibly nonintegrable estimate of surface slopes is represented by a finite set of basis functions, and integrability is enforced by calculating the orthogonal projection onto a vector subspace spanning the set of integrable slopes. The integrability projection constraint was applied to extending an iterative shape-from-shading algorithm of M.J. Brooks and B.K.P. Horn (1985). Experimental results show that the extended algorithm converges faster and with less error than the original version. Good surface reconstructions were obtained with and without known boundary conditions and for fairly complicated surfaces. >"
                },
                {
                    "title": "SHAPE FROM SHADING: A METHOD FOR OBTAINING THE SHAPE OF A SMOOTH OPAQUE OBJECT FROM ONE VIEW",
                    "abstract": "A method will be described for finding the shape of a smooth opaque object from a monocular image, given a knowledge of the surface photometry, the position of the light-source and certain auxiliary information to resolve ambiguities. This method is complementary to the use of stereoscopy which relies on matching up sharp detail and will fail on smooth objects. Until now the image processing of single views has been restricted to objects which can meaningfully be considered two-dimensional or bounded by plane surfaces. It is possible to derive a first-order non-linear partial differential equation in two unknowns relating the intensity at the image points to the shape of the object. This equation can be solved by means of an equivalent set of five ordinary differential equations. A curve traced out by solving this set of equations for one set of starting values is called a characteristic strip. Starting one of these strips from each point on some initial curve will produce the whole solution surface. The initial curves can usually be constructed around so-called singular points. A number of applications of this method will be discussed including one to lunar topography and one to the scanning electron microscope. In both of these cases great simplifications occur in the equations. A note on polyhedra follows and a quantitative theory of facial make-up is touched upon. An implementation of some of these ideas on the PDP-6 computer with its attached image-dissector camera at the Artificial Intelligence Laboratory will be described, and also a nose-recognition program."
                },
                {
                    "title": "Early processing of visual information.",
                    "abstract": "An introduction is given to a theory of early visual information processing. The theory has been implemented, and examples are given of images at various stages of analysis. It is argued that the first step of consequence is to compute a primitive but rich description of the grey-level changes present in an image. The description is expressed in a vocabulary of kinds of intensity change (EDGE, SHADING-EDGE, EXTENDED-EDGE, LINE, BLOB etc.). Modifying parameters are bound to the elements in the description, specifying their POSITION, ORIENTATION, TERMINATION points, CONTRAST, SIZE and FUZZINESS. This description is obtained from the intensity array by fixed techniques, and it is called the primal sketch. For most images, the primal sketch is large and unwieldy. The second important step in visual information processing is to group its contents in a way that is appropriate for later recognition. From our ability to interpret drawings with little semantic content, one may infer the presence in our perceptual equipment of symbolic processes that can define \"place-tokens\" in an image in various ways, and can group them according to certain rules. Homomorphic techniques fail to account for many of these grouping phenomena, whose explanations require mechanisms of construction rather than mechanisms of detection. The necessary grouping of elements in the primal sketch may be achieved by a mechanism that has available the processes inferred from above, together with the ability to select items by first order discriminations acting on the elements' parameters. Only occasionally do these mechanisms use downward-flowing information about the contents of the particular image being processed. It is argued that \"non-attentive\" vision is in practice implemented by these grouping operations and first order discriminations acting on the primal sketch. The class of computations so obtained differs slightly from the class of second order operations on the intensity array. The extraction of a form from the primal sketch using these techniques amounts to the separation of figure from ground. It is concluded that most of the separation can be carried out by using techniques that do not depend upon the particular image in question. Therefore, figure-ground separation can normally precede the description of the shape of the extracted form. Up to this point, higher-level knowledge and purpose are brought to bear on only a few of the decisions taken during the processing. This relegates the widespread use of downward-flowing information to a later stage than is found in current machine-vision programs, and implies that such knowledge should influence the control of, rather than interfering with, the actual data-processing that is taking place lower down."
                },
                {
                    "title": "Ieee Transactions on Pattern Analysis and Machine Intelligence Shape, Illumination, and Reflectance from Shading",
                    "abstract": "\u2014A fundamental problem in computer vision is that of inferring the intrinsic, 3D structure of the world from flat, 2D images of that world. Traditional methods for recovering scene properties such as shape, reflectance, or illumination rely on multiple observations of the same scene to overconstrain the problem. Recovering these same properties from a single image seems almost impossible in comparison \u2014 there are an infinite number of shapes, paint, and lights that exactly reproduce a single image. However, certain explanations are more likely than others: surfaces tend to be smooth, paint tends to be uniform, and illumination tends to be natural. We therefore pose this problem as one of statistical inference, and define an optimization problem that searches for the most likely explanation of a single image. Our technique can be viewed as a superset of several classic computer vision problems (shape-from-shading, intrinsic images, color constancy, illumination estimation, etc) and outperforms all previous solutions to those constituent problems."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can a computational model effectively capture multiple underlying shape explanations from ambiguous images, similar to human perception of shape from shading?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of computer vision, particularly in understanding how machines can interpret ambiguous visual information in a manner akin to human perception. This research could lead to significant improvements in applications such as autonomous navigation, augmented reality, and robotics, where accurate shape inference from single images is essential. By developing models that can express multiple interpretations of ambiguous images, future research can explore more nuanced and flexible approaches to visual perception, potentially leading to breakthroughs in how machines understand and interact with the world.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent ambiguity of visual information, where multiple shapes can produce the same shading effects under different lighting conditions. Naive approaches may fail because they typically produce a single, deterministic output, which does not account for the multistable nature of human perception. The technical obstacles include designing a model that can generate multimodal distributions of shape interpretations while maintaining consistency across patches of the image. Additionally, the model must effectively handle the complexities of lighting inference, which is often imprecise due to factors like global illumination.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on deterministic models that yield a single best estimate of shape, which limits their ability to represent the multistable perceptions humans experience. Existing solutions often lack the flexibility to accommodate the ambiguities present in certain images. Barriers to progress include the absence of models that can integrate local patch information while maintaining global consistency. This research proposes a novel approach by utilizing a conditional generative process that allows for the exploration of multiple shape interpretations, thereby addressing the limitations of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves a bottom-up, patch-based diffusion model trained to predict surface normals from small image patches (16\u00d716 pixels). The model incorporates inter-patch shape consistency constraints to coordinate the patches during the sampling process, enabling it to capture global ambiguities. The expected outcome is a multimodal distribution of shape interpretations that aligns with human perception of ambiguous images, demonstrating the model's ability to mimic multistable perception effectively. The evaluation will be based on how well the model's outputs correspond to human interpretations of the"
            }
        },
        "author_data": {
            "d8e92252-3ff0-43ee-b25e-49b3dc4eb253": {
                "pk": "d8e92252-3ff0-43ee-b25e-49b3dc4eb253",
                "name": "Xinran Nicole Han",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "1ac2d7d3-965d-450d-8e0f-f75256b6b88b": {
                "pk": "1ac2d7d3-965d-450d-8e0f-f75256b6b88b",
                "name": "Todd Zickler",
                "collaborators": [
                    "Steven J. Gortler",
                    "Ayan Chakrabarti",
                    "Dor Verbin",
                    "Qi Guo",
                    "Ying Xiong",
                    "Ioannis Gkioulekas",
                    "Louis Theran",
                    "Jialiang Wang",
                    "Stanley H. Chan",
                    "Yue M. Lu"
                ],
                "domain": [
                    "Computer Vision",
                    "Image Processing",
                    "Deep Learning",
                    "Optical Metasurfaces"
                ],
                "publications": [
                    {
                        "title": "Image Restoration with Signal-dependent Camera Noise",
                        "abstract": "This article describes a fast iterative algorithm for image denoising and deconvolution with signal-dependent observation noise. We use an optimization strategy based on variable splitting that adapts traditional Gaussian noise-based restoration algorithms to account for the observed image being corrupted by mixed Poisson-Gaussian noise and quantization errors."
                    },
                    {
                        "title": "Level Set Stereo for Cooperative Grouping with Occlusion",
                        "abstract": "Localizing stereo boundaries is difficult because matching cues are absent in the occluded regions that are adjacent to them. We introduce an energy and level-set optimizer that improves boundaries by encoding the essential geometry of occlusions: The spatial extent of an occlusion must equal the amplitude of the disparity jump that causes it. In a collection of figure-ground scenes from Middlebury and Falling Things stereo datasets, the model provides more accurate boundaries than previous occlusion-handling techniques."
                    },
                    {
                        "title": "Field of Junctions: Extracting Boundary Structure at Low SNR",
                        "abstract": "We introduce a bottom-up model for simultaneously finding many boundary elements in an image, including contours, corners and junctions. The model explains boundary shape in each small patch using a 'generalized M-junction' comprising M angles and a freely-moving vertex. Images are analyzed using non-convex optimization to cooperatively find M+2 junction values at every location, with spatial consistency being enforced by a novel regularizer that reduces curvature while preserving corners and junctions. The resulting 'field of junctions' is simultaneously a contour detector, corner/junction detector, and boundary-aware smoothing of regional appearance. Notably, its unified analysis of contours, corners, junctions and uniform regions allows it to succeed at high noise levels, where other methods for segmentation and boundary detection fail."
                    },
                    {
                        "title": "StegaPos: Preventing Unwanted Crops and Replacements with Imperceptible Positional Embeddings",
                        "abstract": "We present a learned, spatially-varying steganography system that allows detecting when and how images have been altered by cropping, splicing or inpainting after publication. The system comprises a learned encoder that imperceptibly hides distinct positional signatures in every local image region before publication, and an accompanying learned decoder that extracts the steganographic signatures to determine, for each local image region, its 2D positional coordinates within the originally-published image. Crop and replacement edits become detectable by the inconsistencies they cause in the hidden positional signatures. Using a prototype system for small $(400 \\times 400)$ images, we show experimentally that simple CNN encoder and decoder architectures can be trained jointly to achieve detection that is reliable and robust, without introducing perceptible distortion. This approach could help individuals and image-sharing platforms certify that an image was published by a trusted source, and also know which parts of such an image, if any, have been substantially altered since publication."
                    },
                    {
                        "title": "Level Set Binocular Stereo with Occlusions",
                        "abstract": "Localizing stereo boundaries and predicting nearby disparities are difficult because stereo boundaries induce occluded regions where matching cues are absent. Most modern computer vision algorithms treat occlusions secondarily (e.g., via left-right consistency checks after matching) or rely on high-level cues to improve nearby disparities (e.g., via deep networks and large training sets). They ignore the geometry of stereo occlusions, which dictates that the spatial extent of occlusion must equal the amplitude of the disparity jump that causes it. This paper introduces an energy and level-set optimizer that improves boundaries by encoding occlusion geometry. Our model applies to two-layer, figure-ground scenes, and it can be implemented cooperatively using messages that pass predominantly between parents and children in an undecimated hierarchy of multi-scale image patches. In a small collection of figure-ground scenes curated from Middlebury and Falling Things stereo datasets, our model provides more accurate boundaries than previous occlusion-handling stereo techniques. This suggests new directions for creating cooperative stereo systems that incorporate occlusion cues in a human-like manner."
                    },
                    {
                        "title": "Unique Geometry and Texture from Corresponding Image Patches",
                        "abstract": "We present a sufficient condition for recovering unique texture and viewpoints from unknown orthographic projections of a flat texture process. We show that four observations are sufficient in general, and we characterize the ambiguous cases. The results are applicable to shape from texture and texture-based structure from motion."
                    },
                    {
                        "title": "Tackling 3D ToF Artifacts Through Learning and the FLAT Dataset",
                        "abstract": "Scene motion, multiple reflections, and sensor noise introduce artifacts in the depth reconstruction performed by time-of-flight cameras. We propose a two-stage, deep-learning approach to address all of these sources of artifacts simultaneously. We also introduce FLAT, a synthetic dataset of 2000 ToF measurements that capture all of these nonidealities, and allows to simulate different camera hardware. Using the Kinect 2 camera as a baseline, we show improved reconstruction errors over state-of-the-art methods, on both simulated and real data."
                    },
                    {
                        "title": "Low-level Vision by Consensus in a Spatial Hierarchy of Regions",
                        "abstract": "We introduce a multi-scale framework for low-level vision, where the goal is estimating physical scene values from image data---such as depth from stereo image pairs. The framework uses a dense, overlapping set of image regions at multiple scales and a \"local model,\" such as a slanted-plane model for stereo disparity, that is expected to be valid piecewise across the visual field. Estimation is cast as optimization over a dichotomous mixture of variables, simultaneously determining which regions are inliers with respect to the local model (binary variables) and the correct co-ordinates in the local model space for each inlying region (continuous variables). When the regions are organized into a multi-scale hierarchy, optimization can occur in an efficient and parallel architecture, where distributed computational units iteratively perform calculations and share information through sparse connections between parents and children. The framework performs well on a standard benchmark for binocular stereo, and it produces a distributional scene representation that is appropriate for combining with higher-level reasoning and other low-level cues."
                    },
                    {
                        "title": "Linear Symmetries of the Unsquared Measurement Variety",
                        "abstract": "We introduce a new family of algebraic varieties, $L_{d,n}$, which we call the unsquared measurement varieties. This family is parameterized by a number of points $n$ and a dimension $d$. These varieties arise naturally from problems in rigidity theory and distance geometry. In those applications, it can be useful to understand the group of linear automorphisms of $L_{d,n}$. Notably, a result of Regge implies that $L_{2,4}$ has an unexpected linear automorphism. In this paper, we give a complete characterization of the linear automorphisms of $L_{d,n}$ for all $n$ and $d$. We show, that apart from $L_{2,4}$ the unsquared measurement varieties have no unexpected automorphisms. Moreover, for $L_{2,4}$ we characterize the full automorphism group."
                    },
                    {
                        "title": "Monte Carlo non local means: Random sampling for large-scale image filtering",
                        "abstract": "We propose a randomized version of the non-local means (NLM) algorithm for large-scale image filtering. The new algorithm, called Monte Carlo non-local means (MCNLM), speeds up the classical NLM by computing a small subset of image patch distances, which are randomly selected according to a designed sampling pattern. We make two contributions. First, we analyze the performance of the MCNLM algorithm and show that, for large images or large external image databases, the random outcomes of MCNLM are tightly concentrated around the deterministic full NLM result. In particular, our error probability bounds show that, at any given sampling ratio, the probability for MCNLM to have a large deviation from the original NLM solution decays exponentially as the size of the image or database grows. Second, we derive explicit formulas for optimal sampling patterns that minimize the error probability bound by exploiting partial knowledge of the pairwise similarity weights. Numerical experiments show that MCNLM is competitive with other state-of-the-art fast NLM algorithms for single-image denoising. When applied to denoising images using an external database containing ten billion patches, MCNLM returns a randomized solution that is within 0.2 dB of the full NLM solution while reducing the runtime by three orders of magnitude."
                    },
                    {
                        "title": "Understanding Symmetric Smoothing Filters: A Gaussian Mixture Model Perspective",
                        "abstract": "Many patch-based image denoising algorithms can be formulated as applying a smoothing filter to the noisy image. Expressed as matrices, the smoothing filters must be row normalized so that each row sums to unity. Surprisingly, if we apply a column normalization before the row normalization, the performance of the smoothing filter can often be significantly improved. Prior works showed that such performance gain is related to the Sinkhorn-Knopp balancing algorithm, an iterative procedure that symmetrizes a row-stochastic matrix to a doubly-stochastic matrix. However, a complete understanding of the performance gain phenomenon is still lacking.   In this paper, we study the performance gain phenomenon from a statistical learning perspective. We show that Sinkhorn-Knopp is equivalent to an Expectation-Maximization (EM) algorithm of learning a Gaussian mixture model of the image patches. By establishing the correspondence between the steps of Sinkhorn-Knopp and the EM algorithm, we provide a geometrical interpretation of the symmetrization process. This observation allows us to develop a new denoising algorithm called Gaussian mixture model symmetric smoothing filter (GSF). GSF is an extension of the Sinkhorn-Knopp and is a generalization of the original smoothing filters. Despite its simple formulation, GSF outperforms many existing smoothing filters and has a similar performance compared to several state-of-the-art denoising algorithms."
                    },
                    {
                        "title": "Trilateration using Unlabeled Path or Loop Lengths",
                        "abstract": "Let $\\mathbf{p}$ be a configuration of $n$ points in $\\mathbb{R}^d$ for some $n$ and some $d \\ge 2$. Each pair of points defines an edge, which has a Euclideanlength in the configuration. A path is an ordered sequence of the points, and a loop is a path that begins and ends at the same point. A path or loop, as a sequence of edges, also has a Euclidean length, which is simply the sum of its Euclidean edge lengths. We are interested in reconstructing $\\mathbf{p}$ given a set of edge, path and loop lengths. In particular, we consider the unlabeled setting where the lengths are given simply as a set of real numbers, and are not labeled with the combinatorial data describing which paths or loops gave rise to these lengths. In this paper, we study the question of when $\\mathbf{p}$ will be uniquely determined (up to an unknowable Euclidean transform) from some given set of path or loop lengths through an exhaustive trilateration process.   Such a process has already been used for the simpler problem of reconstruction using unlabeled edge lengths. This paper also provides a complete proof that this process must work in that edge-setting when given a sufficiently rich set of edge measurements and assuming that $\\mathbf{p}$ is generic."
                    },
                    {
                        "title": "Determining Generic Point Configurations From Unlabeled Path or Loop Lengths",
                        "abstract": "Let $\\mathbf{p}$ be a configuration of $n$ points in $\\mathbb{R}^d$ for some $n$ and some $d \\ge 2$. Each pair of points defines an edge, which has a Euclidean length in the configuration. A path is an ordered sequence of the points, and a loop is a path that has the same endpoints. A path or loop, as a sequence of edges, also has a Euclidean length. In this paper, we study the question of when $\\mathbf{p}$ will be uniquely determined (up to an unknowable Euclidean transform) from a given set of path or loop lengths. In particular, we consider the setting where the lengths are given simply as a set of real numbers, and are not labeled with the combinatorial data describing the paths or loops that gave rise to the lengths.   Our main result is a condition on the set of paths or loops that is sufficient to guarantee such a unique determination. We also provide an algorithm, under a real computational model, for performing a reconstruction of $\\mathbf{p}$ from such unlabeled lengths.   To obtain our results, we introduce a new family of algebraic varieties which we call the unsquared measurement varieties. The family is parameterized by the number of points $n$ and the dimension $d$, and our results follow from a complete characterization of the linear automorphisms of these varieties for all $n$ and $d$. The linear automorphisms for the special case of $n = 4$ and $d = 2$ correspond to the so-called Regge symmetries of the tetrahedron."
                    },
                    {
                        "title": "Modeling Radiometric Uncertainty for Vision with Tone-mapped Color Images",
                        "abstract": "To produce images that are suitable for display, tone-mapping is widely used in digital cameras to map linear color measurements into narrow gamuts with limited dynamic range. This introduces non-linear distortion that must be undone, through a radiometric calibration process, before computer vision systems can analyze such photographs radiometrically. This paper considers the inherent uncertainty of undoing the effects of tone-mapping. We observe that this uncertainty varies substantially across color space, making some pixels more reliable than others. We introduce a model for this uncertainty and a method for fitting it to a given camera or imaging pipeline. Once fit, the model provides for each pixel in a tone-mapped digital photograph a probability distribution over linear scene colors that could have induced it. We demonstrate how these distributions can be useful for visual inference by incorporating them into estimation algorithms for a representative set of vision tasks."
                    },
                    {
                        "title": "Boundary Attention: Learning curves, corners, junctions and grouping",
                        "abstract": "We present a lightweight network that infers grouping and boundaries, including curves, corners and junctions. It operates in a bottom-up fashion, analogous to classical methods for sub-pixel edge localization and edge-linking, but with a higher-dimensional representation of local boundary structure, and notions of local scale and spatial consistency that are learned instead of designed. Our network uses a mechanism that we call boundary attention: a geometry-aware local attention operation that, when applied densely and repeatedly, progressively refines a pixel-resolution field of variables that specify the boundary structure in every overlapping patch within an image. Unlike many edge detectors that produce rasterized binary edge maps, our model provides a rich, unrasterized representation of the geometric structure in every local region. We find that its intentional geometric bias allows it to be trained on simple synthetic shapes and then generalize to extracting boundaries from noisy low-light photographs."
                    },
                    {
                        "title": "Polarization Multi-Image Synthesis with Birefringent Metasurfaces",
                        "abstract": "Optical metasurfaces composed of precisely engineered nanostructures have gained significant attention for their ability to manipulate light and implement distinct functionalities based on the properties of the incident field. Computational imaging systems have started harnessing this capability to produce sets of coded measurements that benefit certain tasks when paired with digital post-processing. Inspired by these works, we introduce a new system that uses a birefringent metasurface with a polarizer-mosaicked photosensor to capture four optically-coded measurements in a single exposure. We apply this system to the task of incoherent opto-electronic filtering, where digital spatial-filtering operations are replaced by simpler, per-pixel sums across the four polarization channels, independent of the spatial filter size. In contrast to previous work on incoherent opto-electronic filtering that can realize only one spatial filter, our approach can realize a continuous family of filters from a single capture, with filters being selected from the family by adjusting the post-capture digital summation weights. To find a metasurface that can realize a set of user-specified spatial filters, we introduce a form of gradient descent with a novel regularizer that encourages light efficiency and a high signal-to-noise ratio. We demonstrate several examples in simulation and with fabricated prototypes, including some with spatial filters that have prescribed variations with respect to depth and wavelength.   Visit the Project Page at https://deanhazineh.github.io/publications/Multi_Image_Synthesis/MIS_Home.html"
                    },
                    {
                        "title": "From Shading to Local Shape",
                        "abstract": "We develop a framework for extracting a concise representation of the shape information available from diffuse shading in a small image patch. This produces a mid-level scene descriptor, comprised of local shape distributions that are inferred separately at every image patch across multiple scales. The framework is based on a quadratic representation of local shape that, in the absence of noise, has guarantees on recovering accurate local shape and lighting. And when noise is present, the inferred local shape distributions provide useful shape information without over-committing to any particular image explanation. These local shape distributions naturally encode the fact that some smooth diffuse regions are more informative than others, and they enable efficient and robust reconstruction of object-scale shape. Experimental results show that this approach to surface reconstruction compares well against the state-of-art on both synthetic images and captured photographs."
                    },
                    {
                        "title": "Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields",
                        "abstract": "Neural Radiance Fields (NeRF) is a popular view synthesis technique that represents a scene as a continuous volumetric function, parameterized by multilayer perceptrons that provide the volume density and view-dependent emitted radiance at each location. While NeRF-based techniques excel at representing fine geometric structures with smoothly varying view-dependent appearance, they often fail to accurately capture and reproduce the appearance of glossy surfaces. We address this limitation by introducing Ref-NeRF, which replaces NeRF's parameterization of view-dependent outgoing radiance with a representation of reflected radiance and structures this function using a collection of spatially-varying scene properties. We show that together with a regularizer on normal vectors, our model significantly improves the realism and accuracy of specular reflections. Furthermore, we show that our model's internal representation of outgoing radiance is interpretable and useful for scene editing."
                    },
                    {
                        "title": "D-Flat: A Differentiable Flat-Optics Framework for End-to-End Metasurface Visual Sensor Design",
                        "abstract": "Optical metasurfaces are planar substrates with custom-designed, nanoscale features that selectively modulate incident light with respect to direction, wavelength, and polarization. When coupled with photodetectors and appropriate post-capture processing, they provide a means to create computational imagers and sensors that are exceptionally small and have distinctive capabilities. We introduce D-Flat, a framework in TensorFlow that renders physically-accurate images induced by metasurface optical systems. This framework is fully differentiable with respect to metasurface shape and post-capture computational parameters and allows simultaneous optimization with respect to almost any measure of sensor performance. D-Flat enables simulation of millimeter to centimeter diameter metasurfaces on commodity computers, and it is modular in the sense of accommodating a variety of wave optics models for scattering at the metasurface and for propagation to photosensors. We validate D-Flat against symbolic calculations and previous experimental measurements, and we provide simulations that demonstrate its ability to discover novel computational sensor designs for two applications: single-shot depth sensing and single-shot spatial frequency filtering."
                    }
                ]
            },
            "565ae404-8182-4705-98f6-10f729b3d954": {
                "pk": "565ae404-8182-4705-98f6-10f729b3d954",
                "name": "Ko Nishino",
                "collaborators": [
                    "Shohei Nobuhara",
                    "Gabriel Schwartz",
                    "Hang Zhang",
                    "Kristin Dana",
                    "Kohei Yamashita",
                    "Mai Nishimura",
                    "Yuto Enyo",
                    "Tomoki Ichikawa",
                    "Stephen Lombardi",
                    "Genki Kinoshita"
                ],
                "domain": [
                    "Computer Vision",
                    "Material Recognition",
                    "Inverse Rendering",
                    "3D Reconstruction"
                ],
                "publications": [
                    {
                        "title": "Radiometric Scene Decomposition: Scene Reflectance, Illumination, and Geometry from RGB-D Images",
                        "abstract": "Recovering the radiometric properties of a scene (i.e., the reflectance, illumination, and geometry) is a long-sought ability of computer vision that can provide invaluable information for a wide range of applications. Deciphering the radiometric ingredients from the appearance of a real-world scene, as opposed to a single isolated object, is particularly challenging as it generally consists of various objects with different material compositions exhibiting complex reflectance and light interactions that are also part of the illumination. We introduce the first method for radiometric scene decomposition that handles those intricacies. We use RGB-D images to bootstrap geometry recovery and simultaneously recover the complex reflectance and natural illumination while refining the noisy initial geometry and segmenting the scene into different material regions. Most important, we handle real-world scenes consisting of multiple objects of unknown materials, which necessitates the modeling of spatially-varying complex reflectance, natural illumination, texture, interreflection and shadows. We systematically evaluate the effectiveness of our method on synthetic scenes and demonstrate its application to real-world scenes. The results show that rich radiometric information can be recovered from RGB-D images and demonstrate a new role RGB-D sensors can play for general scene understanding tasks."
                    },
                    {
                        "title": "Diffusion Reflectance Map: Single-Image Stochastic Inverse Rendering of Illumination and Reflectance",
                        "abstract": "Reflectance bounds the frequency spectrum of illumination in the object appearance. In this paper, we introduce the first stochastic inverse rendering method, which recovers the attenuated frequency spectrum of an illumination jointly with the reflectance of an object of known geometry from a single image. Our key idea is to solve this blind inverse problem in the reflectance map, an appearance representation invariant to the underlying geometry, by learning to reverse the image formation with a novel diffusion model which we refer to as the Diffusion Reflectance Map Network (DRMNet). Given an observed reflectance map converted and completed from the single input image, DRMNet generates a reflectance map corresponding to a perfect mirror sphere while jointly estimating the reflectance. The forward process can be understood as gradually filtering a natural illumination with lower and lower frequency reflectance and additive Gaussian noise. DRMNet learns to invert this process with two subnetworks, IllNet and RefNet, which work in concert towards this joint estimation. The network is trained on an extensive synthetic dataset and is demonstrated to generalize to real images, showing state-of-the-art accuracy on established datasets."
                    },
                    {
                        "title": "Integrating Local Material Recognition with Large-Scale Perceptual Attribute Discovery",
                        "abstract": "Material attributes have been shown to provide a discriminative intermediate representation for recognizing materials, especially for the challenging task of recognition from local material appearance (i.e., regardless of object and scene context). In the past, however, material attributes have been recognized separately preceding category recognition. In contrast, neuroscience studies on material perception and computer vision research on object and place recognition have shown that attributes are produced as a by-product during the category recognition process. Does the same hold true for material attribute and category recognition? In this paper, we introduce a novel material category recognition network architecture to show that perceptual attributes can, in fact, be automatically discovered inside a local material recognition framework. The novel material-attribute-category convolutional neural network (MAC-CNN) produces perceptual material attributes from the intermediate pooling layers of an end-to-end trained category recognition network using an auxiliary loss function that encodes human material perception. To train this model, we introduce a novel large-scale database of local material appearance organized under a canonical material category taxonomy and careful image patch extraction that avoids unwanted object and scene context. We show that the discovered attributes correspond well with semantically-meaningful visual material traits via Boolean algebra, and enable recognition of previously unseen material categories given only a few examples. These results have strong implications in how perceptually meaningful attributes can be learned in other recognition tasks."
                    },
                    {
                        "title": "Recognizing Material Properties from Images",
                        "abstract": "Humans rely on properties of the materials that make up objects to guide our interactions with them. Grasping smooth materials, for example, requires care, and softness is an ideal property for fabric used in bedding. Even when these properties are not visual (e.g. softness is a physical property), we may still infer their presence visually. We refer to such material properties as visual material attributes. Recognizing these attributes in images can contribute valuable information for general scene understanding and material recognition. Unlike well-known object and scene attributes, visual material attributes are local properties with no fixed shape or spatial extent. We show that given a set of images annotated with known material attributes, we may accurately recognize the attributes from small local image patches. Obtaining such annotations in a consistent fashion at scale, however, is challenging. To address this, we introduce a method that allows us to probe the human visual perception of materials by asking simple yes/no questions comparing pairs of image patches. This provides sufficient weak supervision to build a set of attributes and associated classifiers that, while unnamed, serve the same function as the named attributes we use to describe materials. Doing so allows us to recognize visual material attributes without resorting to exhaustive manual annotation of a fixed set of named attributes. Furthermore, we show that this method may be integrated in the end-to-end learning of a material classification CNN to simultaneously recognize materials and discover their visual attributes. Our experimental results show that visual material attributes, whether named or automatically discovered, provide a useful intermediate representation for known material categories themselves as well as a basis for transfer learning when recognizing previously-unseen categories."
                    },
                    {
                        "title": "Material Recognition from Local Appearance in Global Context",
                        "abstract": "Recognition of materials has proven to be a challenging problem due to the wide variation in appearance within and between categories. Global image context, such as where the material is or what object it makes up, can be crucial to recognizing the material. Existing methods, however, operate on an implicit fusion of materials and context by using large receptive fields as input (i.e., large image patches). Many recent material recognition methods treat materials as yet another set of labels like objects. Materials are, however, fundamentally different from objects as they have no inherent shape or defined spatial extent. Approaches that ignore this can only take advantage of limited implicit context as it appears during training. We instead show that recognizing materials purely from their local appearance and integrating separately recognized global contextual cues including objects and places leads to superior dense, per-pixel, material recognition. We achieve this by training a fully-convolutional material recognition network end-to-end with only material category supervision. We integrate object and place estimates to this network from independent CNNs. This approach avoids the necessity of preparing an impractically-large amount of training data to cover the product space of materials, objects, and scenes, while fully leveraging contextual cues for dense material recognition. Furthermore, we perform a detailed analysis of the effects of context granularity, spatial resolution, and the network level at which we introduce context. On a recently introduced comprehensive and diverse material database \\cite{Schwartz2016}, we confirm that our method achieves state-of-the-art accuracy with significantly less training data compared to past methods."
                    },
                    {
                        "title": "Camera Height Doesn't Change: Unsupervised Training for Metric Monocular Road-Scene Depth Estimation",
                        "abstract": "In this paper, we introduce a novel training method for making any monocular depth network learn absolute scale and estimate metric road-scene depth just from regular training data, i.e., driving videos. We refer to this training framework as FUMET. The key idea is to leverage cars found on the road as sources of scale supervision and to incorporate them in network training robustly. FUMET detects and estimates the sizes of cars in a frame and aggregates scale information extracted from them into an estimate of the camera height whose consistency across the entire video sequence is enforced as scale supervision. This realizes robust unsupervised training of any, otherwise scale-oblivious, monocular depth network so that they become not only scale-aware but also metric-accurate without the need for auxiliary sensors and extra supervision. Extensive experiments on the KITTI and the Cityscapes datasets show the effectiveness of FUMET, which achieves state-of-the-art accuracy. We also show that FUMET enables training on mixed datasets of different camera heights, which leads to larger-scale training and better generalization. Metric depth reconstruction is essential in any road-scene visual modeling, and FUMET democratizes its deployment by establishing the means to convert any model into a metric depth estimator."
                    },
                    {
                        "title": "Reflectance Hashing for Material Recognition",
                        "abstract": "We introduce a novel method for using reflectance to identify materials. Reflectance offers a unique signature of the material but is challenging to measure and use for recognizing materials due to its high-dimensionality. In this work, one-shot reflectance is captured using a unique optical camera measuring {\\it reflectance disks} where the pixel coordinates correspond to surface viewing angles. The reflectance has class-specific stucture and angular gradients computed in this reflectance space reveal the material class.   These reflectance disks encode discriminative information for efficient and accurate material recognition. We introduce a framework called reflectance hashing that models the reflectance disks with dictionary learning and binary hashing. We demonstrate the effectiveness of reflectance hashing for material recognition with a number of real-world materials."
                    },
                    {
                        "title": "Friction from Reflectance: Deep Reflectance Codes for Predicting Physical Surface Properties from One-Shot In-Field Reflectance",
                        "abstract": "Images are the standard input for vision algorithms, but one-shot infield reflectance measurements are creating new opportunities for recognition and scene understanding. In this work, we address the question of what reflectance can reveal about materials in an efficient manner. We go beyond the question of recognition and labeling and ask the question: What intrinsic physical properties of the surface can be estimated using reflectance? We introduce a framework that enables prediction of actual friction values for surfaces using one-shot reflectance measurements. This work is a first of its kind vision-based friction estimation. We develop a novel representation for reflectance disks that capture partial BRDF measurements instantaneously. Our method of deep reflectance codes combines CNN features and fisher vector pooling with optimal binary embedding to create codes that have sufficient discriminatory power and have important properties of illumination and spatial invariance. The experimental results demonstrate that reflectance can play a new role in deciphering the underlying physical properties of real-world scenes."
                    },
                    {
                        "title": "3D-GMNet: Single-View 3D Shape Recovery as A Gaussian Mixture",
                        "abstract": "In this paper, we introduce 3D-GMNet, a deep neural network for 3D object shape reconstruction from a single image. As the name suggests, 3D-GMNet recovers 3D shape as a Gaussian mixture. In contrast to voxels, point clouds, or meshes, a Gaussian mixture representation provides an analytical expression with a small memory footprint while accurately representing the target 3D shape. At the same time, it offers a number of additional advantages including instant pose estimation and controllable level-of-detail reconstruction, while also enabling interpretation as a point cloud, volume, and a mesh model. We train 3D-GMNet end-to-end with single input images and corresponding 3D models by introducing two novel loss functions, a 3D Gaussian mixture loss and a 2D multi-view loss, which collectively enable accurate shape reconstruction as kernel density estimation. We thoroughly evaluate the effectiveness of 3D-GMNet with synthetic and real images of objects. The results show accurate reconstruction with a compact representation that also realizes novel applications of single-image 3D reconstruction."
                    },
                    {
                        "title": "View Birdification in the Crowd: Ground-Plane Localization from Perceived Movements",
                        "abstract": "We introduce view birdification, the problem of recovering ground-plane movements of people in a crowd from an ego-centric video captured from an observer (e.g., a person or a vehicle) also moving in the crowd. Recovered ground-plane movements would provide a sound basis for situational understanding and benefit downstream applications in computer vision and robotics. In this paper, we formulate view birdification as a geometric trajectory reconstruction problem and derive a cascaded optimization method from a Bayesian perspective. The method first estimates the observer's movement and then localizes surrounding pedestrians for each frame while taking into account the local interactions between them. We introduce three datasets by leveraging synthetic and real trajectories of people in crowds and evaluate the effectiveness of our method. The results demonstrate the accuracy of our method and set the ground for further studies of view birdification as an important but challenging visual understanding problem."
                    },
                    {
                        "title": "nLMVS-Net: Deep Non-Lambertian Multi-View Stereo",
                        "abstract": "We introduce a novel multi-view stereo (MVS) method that can simultaneously recover not just per-pixel depth but also surface normals, together with the reflectance of textureless, complex non-Lambertian surfaces captured under known but natural illumination. Our key idea is to formulate MVS as an end-to-end learnable network, which we refer to as nLMVS-Net, that seamlessly integrates radiometric cues to leverage surface normals as view-independent surface features for learned cost volume construction and filtering. It first estimates surface normals as pixel-wise probability densities for each view with a novel shape-from-shading network. These per-pixel surface normal densities and the input multi-view images are then input to a novel cost volume filtering network that learns to recover per-pixel depth and surface normal. The reflectance is also explicitly estimated by alternating with geometry reconstruction. Extensive quantitative evaluations on newly established synthetic and real-world datasets show that nLMVS-Net can robustly and accurately recover the shape and reflectance of complex objects in natural settings."
                    },
                    {
                        "title": "Fresnel Microfacet BRDF: Unification of Polari-Radiometric Surface-Body Reflection",
                        "abstract": "Computer vision applications have heavily relied on the linear combination of Lambertian diffuse and microfacet specular reflection models for representing reflected radiance, which turns out to be physically incompatible and limited in applicability. In this paper, we derive a novel analytical reflectance model, which we refer to as Fresnel Microfacet BRDF model, that is physically accurate and generalizes to various real-world surfaces. Our key idea is to model the Fresnel reflection and transmission of the surface microgeometry with a collection of oriented mirror facets, both for body and surface reflections. We carefully derive the Fresnel reflection and transmission for each microfacet as well as the light transport between them in the subsurface. This physically-grounded modeling also allows us to express the polarimetric behavior of reflected light in addition to its radiometric behavior. That is, FMBRDF unifies not only body and surface reflections but also light reflection in radiometry and polarization and represents them in a single model. Experimental results demonstrate its effectiveness in accuracy, expressive power, and image-based estimation."
                    },
                    {
                        "title": "InCrowdFormer: On-Ground Pedestrian World Model From Egocentric Views",
                        "abstract": "We introduce an on-ground Pedestrian World Model, a computational model that can predict how pedestrians move around an observer in the crowd on the ground plane, but from just the egocentric-views of the observer. Our model, InCrowdFormer, fully leverages the Transformer architecture by modeling pedestrian interaction and egocentric to top-down view transformation with attention, and autoregressively predicts on-ground positions of a variable number of people with an encoder-decoder architecture. We encode the uncertainties arising from unknown pedestrian heights with latent codes to predict the posterior distributions of pedestrian positions. We validate the effectiveness of InCrowdFormer on a novel prediction benchmark of real movements. The results show that InCrowdFormer accurately predicts the future coordination of pedestrians. To the best of our knowledge, InCrowdFormer is the first-of-its-kind pedestrian world model which we believe will benefit a wide range of egocentric-view applications including crowd navigation, tracking, and synthesis."
                    },
                    {
                        "title": "TransPoser: Transformer as an Optimizer for Joint Object Shape and Pose Estimation",
                        "abstract": "We propose a novel method for joint estimation of shape and pose of rigid objects from their sequentially observed RGB-D images. In sharp contrast to past approaches that rely on complex non-linear optimization, we propose to formulate it as a neural optimization that learns to efficiently estimate the shape and pose. We introduce Deep Directional Distance Function (DeepDDF), a neural network that directly outputs the depth image of an object given the camera viewpoint and viewing direction, for efficient error computation in 2D image space. We formulate the joint estimation itself as a Transformer which we refer to as TransPoser. We fully leverage the tokenization and multi-head attention to sequentially process the growing set of observations and to efficiently update the shape and pose with a learned momentum, respectively. Experimental results on synthetic and real data show that DeepDDF achieves high accuracy as a category-level object shape representation and TransPoser achieves state-of-the-art accuracy efficiently for joint shape and pose estimation."
                    },
                    {
                        "title": "DeePoint: Visual Pointing Recognition and Direction Estimation",
                        "abstract": "In this paper, we realize automatic visual recognition and direction estimation of pointing. We introduce the first neural pointing understanding method based on two key contributions. The first is the introduction of a first-of-its-kind large-scale dataset for pointing recognition and direction estimation, which we refer to as the DP Dataset. DP Dataset consists of more than 2 million frames of 33 people pointing in various styles annotated for each frame with pointing timings and 3D directions. The second is DeePoint, a novel deep network model for joint recognition and 3D direction estimation of pointing. DeePoint is a Transformer-based network which fully leverages the spatio-temporal coordination of the body parts, not just the hands. Through extensive experiments, we demonstrate the accuracy and efficiency of DeePoint. We believe DP Dataset and DeePoint will serve as a sound foundation for visual human intention understanding."
                    },
                    {
                        "title": "DeepShaRM: Multi-View Shape and Reflectance Map Recovery Under Unknown Lighting",
                        "abstract": "Geometry reconstruction of textureless, non-Lambertian objects under unknown natural illumination (i.e., in the wild) remains challenging as correspondences cannot be established and the reflectance cannot be expressed in simple analytical forms. We derive a novel multi-view method, DeepShaRM, that achieves state-of-the-art accuracy on this challenging task. Unlike past methods that formulate this as inverse-rendering, i.e., estimation of reflectance, illumination, and geometry from images, our key idea is to realize that reflectance and illumination need not be disentangled and instead estimated as a compound reflectance map. We introduce a novel deep reflectance map estimation network that recovers the camera-view reflectance maps from the surface normals of the current geometry estimate and the input multi-view images. The network also explicitly estimates per-pixel confidence scores to handle global light transport effects. A deep shape-from-shading network then updates the geometry estimate expressed with a signed distance function using the recovered reflectance maps. By alternating between these two, and, most important, by bypassing the ill-posed problem of reflectance and illumination decomposition, the method accurately recovers object geometry in these challenging settings. Extensive experiments on both synthetic and real-world data clearly demonstrate its state-of-the-art accuracy."
                    },
                    {
                        "title": "SPIDeRS: Structured Polarization for Invisible Depth and Reflectance Sensing",
                        "abstract": "Can we capture shape and reflectance in stealth? Such capability would be valuable for many application domains in vision, xR, robotics, and HCI. We introduce structured polarization for invisible depth and reflectance sensing (SPIDeRS), the first depth and reflectance sensing method using patterns of polarized light. The key idea is to modulate the angle of linear polarization (AoLP) of projected light at each pixel. The use of polarization makes it invisible and lets us recover not only depth but also directly surface normals and even reflectance. We implement SPIDeRS with a liquid crystal spatial light modulator (SLM) and a polarimetric camera. We derive a novel method for robustly extracting the projected structured polarization pattern from the polarimetric object appearance. We evaluate the effectiveness of SPIDeRS by applying it to a number of real-world objects. The results show that our method successfully reconstructs object shapes of various materials and is robust to diffuse reflection and ambient light. We also demonstrate relighting using recovered surface normals and reflectance. We believe SPIDeRS opens a new avenue of polarization use in visual sensing."
                    },
                    {
                        "title": "Invertible Neural BRDF for Object Inverse Rendering",
                        "abstract": "We introduce a novel neural network-based BRDF model and a Bayesian framework for object inverse rendering, i.e., joint estimation of reflectance and natural illumination from a single image of an object of known geometry. The BRDF is expressed with an invertible neural network, namely, normalizing flow, which provides the expressive power of a high-dimensional representation, computational simplicity of a compact analytical model, and physical plausibility of a real-world BRDF. We extract the latent space of real-world reflectance by conditioning this model, which directly results in a strong reflectance prior. We refer to this model as the invertible neural BRDF model (iBRDF). We also devise a deep illumination prior by leveraging the structural bias of deep neural networks. By integrating this novel BRDF model and reflectance and illumination priors in a MAP estimation formulation, we show that this joint estimation can be computed efficiently with stochastic gradient descent. We experimentally validate the accuracy of the invertible neural BRDF model on a large number of measured data and demonstrate its use in object inverse rendering on a number of synthetic and real images. The results show new ways in which deep neural networks can help solve challenging radiometric inverse problems."
                    },
                    {
                        "title": "Differential Angular Imaging for Material Recognition",
                        "abstract": "Material recognition for real-world outdoor surfaces has become increasingly important for computer vision to support its operation \"in the wild.\" Computational surface modeling that underlies material recognition has transitioned from reflectance modeling using in-lab controlled radiometric measurements to image-based representations based on internet-mined images of materials captured in the scene. We propose to take a middle-ground approach for material recognition that takes advantage of both rich radiometric cues and flexible image capture. We realize this by developing a framework for differential angular imaging, where small angular variations in image capture provide an enhanced appearance representation and significant recognition improvement. We build a large-scale material database, Ground Terrain in Outdoor Scenes (GTOS) database, geared towards real use for autonomous agents. The database consists of over 30,000 images covering 40 classes of outdoor ground terrain under varying weather and lighting conditions. We develop a novel approach for material recognition called a Differential Angular Imaging Network (DAIN) to fully leverage this large dataset. With this novel network architecture, we extract characteristics of materials encoded in the angular and spatial gradients of their appearance. Our results show that DAIN achieves recognition performance that surpasses single view or coarsely quantized multiview images. These results demonstrate the effectiveness of differential angular imaging as a means for flexible, in-place material recognition."
                    }
                ]
            }
        }
    },
    "2305.17555": {
        "paper_data": {
            "title": "Diffeomorphic Mesh Deformation via Efficient Optimal Transport for Cortical Surface Reconstruction",
            "url": "http://arxiv.org/abs/2305.17555v3",
            "arxiv_id": "2305.17555",
            "authors": [
                "Tung Le",
                "Khai Nguyen",
                "Shanlin Sun",
                "Kun Han",
                "Nhat Ho",
                "Xiaohui Xie"
            ],
            "abstract": "Mesh deformation plays a pivotal role in many 3D vision tasks including dynamic simulations, rendering, and reconstruction. However, defining an efficient discrepancy between predicted and target meshes remains an open problem. A prevalent approach in current deep learning is the set-based approach which measures the discrepancy between two surfaces by comparing two randomly sampled point-clouds from the two meshes with Chamfer pseudo-distance. Nevertheless, the set-based approach still has limitations such as lacking a theoretical guarantee for choosing the number of points in sampled point-clouds, and the pseudo-metricity and the quadratic complexity of the Chamfer divergence. To address these issues, we propose a novel metric for learning mesh deformation. The metric is defined by sliced Wasserstein distance on meshes represented as probability measures that generalize the set-based approach. By leveraging probability measure space, we gain flexibility in encoding meshes using diverse forms of probability measures, such as continuous, empirical, and discrete measures via varifold representation. After having encoded probability measures, we can compare meshes by using the sliced Wasserstein distance which is an effective optimal transport distance with linear computational complexity and can provide a fast statistical rate for approximating the surface of meshes. To the end, we employ a neural ordinary differential equation (ODE) to deform the input surface into the target shape by modeling the trajectories of the points on the surface. Our experiments on cortical surface reconstruction demonstrate that our approach surpasses other competing methods in multiple datasets and metrics.",
            "introduction": "   1 Introduction  Mesh deformation is a fundamental task in 3D computer vision and computer graphics. A wide range of shape reconstruction tasks\u00a0(Chen & Zhang, 2019; Jiang et\u00a0al., 2020; Mescheder et\u00a0al., 2019; Niemeyer et\u00a0al., 2020; Park et\u00a0al., 2019; Sun et\u00a0al., 2023; Yariv et\u00a0al., 2020) and shape registration\u00a0(Ashburner, 2007; Dalca et\u00a0al., 2018; Balakrishnan et\u00a0al., 2019; Le et\u00a0al., 2024; Krebs et\u00a0al., 2019; Dalca et\u00a0al., 2019; Han et\u00a0al., 2024; Sun et\u00a0al., 2022; Han et\u00a0al., 2023) leverages state-of-the-art mesh deformation methodology. One popular approach for mesh deformation is to estimate the vertex displacement vectors (3D offsets) while keeping their connectivity\u00a0(Wang et\u00a0al., 2019; Bongratz et\u00a0al., 2022). However, displacement-based methods cannot guarantee the manifoldness of the resulting mesh and often produce self-intersecting faces. To address this issue, diffeomorphic transformation\u00a0(Ruelle & Sullivan, 1975; Arsigny, 2004) is one effective way to deform a mesh while preserving its topology. As an instance of diffeomorphic surface deformation, Neural Mesh Flow (NMF)\u00a0(Gupta, 2020) learns a sequence of diffeomorphic flows between two meshes and models the trajectories of the mesh vertices as ordinary differential equations (ODEs). However, these methods have limited shape representation capacity and struggle to perform well on complex manifolds. To overcome this limitation, several diffeomorphic mesh deformation models have been proposed, such as CortexODE\u00a0(Ma et\u00a0al., 2022) and CorticalFlow\u00a0(Lebrat et\u00a0al., 2022), which aim to handle hard manifolds, such as the cortical surface. While CorticalFlow\u00a0(Lebrat et\u00a0al., 2022) introduces diffeomorphic mesh deformation (DMD) modules to learn a series of stationary velocity fields, CortexODE\u00a0(Ma et\u00a0al., 2022) encodes spatial information of the MRI images along with vertices features using an MLP model and employs neural ODEs\u00a0(Chen et\u00a0al., 2018) to model the trajectories of the points. Nonetheless, these approaches often rely on Chamfer distance as the objective function, which might have disadvantages.   Selecting an appropriate metric to evaluate the dissimilarity between two meshes is a crucial step in learning deformation mesh models. Recent literature favors the approach of using set-based comparison due to its simplicity. In particular, the set-based approach first samples two sets of points on the surface meshes, and then use Chamfer distance (CD)\u00a0(Deng et\u00a0al., 2018; Duan et\u00a0al., 2019; Groueix et\u00a0al., 2018) to compare two meshes and optimize them. However, the CD loss tends to get trapped in local minima easily\u00a0(Achlioptas et\u00a0al., 2018; Nguyen et\u00a0al., 2021b; 2023; 2024a), failing to distinguish bad samples from the true ones, as demonstrated by our toy example in Fig.\u00a01. Although a weighted CD has been proposed to prioritize fitting local regions with high curvatures in Vox2Cortex\u00a0(Bongratz et\u00a0al., 2022), the issue is only alleviated but not resolved completely, still resulting in suboptimal assignments between two sets of points. Therefore, we propose novel approaches to transform a mesh into a probability measure that generalizes the set-based approach. Furthermore, by relying on the probability measure approach, we can employ geometric measure theory to represent mesh as an oriented varifold\u00a0(Almgren, 1966; Vaillant & Glaunes, 2005; Glaun\u00e8s et\u00a0al., 2008) and get a better approximation of the mesh compared to the random sampling approach. After that, we adopt efficient optimal transport to compare these measures since optimal transport distances are naturally fitted to compare disjoint-support measures.    Figure 1:",
            "references": [
                {
                    "title": "Integrating Efficient Optimal Transport and Functional Maps for Unsupervised Shape Correspondence Learning",
                    "abstract": "In the realm of computer vision and graphics, accurately establishing correspondences between geometric 3D shapes is pivotal for applications like object tracking, registration, texture transfer, and statistical shape analysis. Moving beyond traditional hand-crafted and data-driven feature learning methods, we incorporate spectral methods with deep learning, focusing on functional maps (FMs) and optimal transport (OT). Traditional OT-based approaches, often reliant on entropy regularization OT in learning-based framework, face computational challenges due to their quadratic cost. Our key contribution is to employ the sliced Wasserstein distance (SWD)for OT, which is a valid fast optimal transport metric in an unsupervised shape matching framework. This unsupervised framework integrates functional map regularizers with a novel OT-based loss derived from SWD, enhancing feature alignment between shapes treated as discrete probability measures. We also introduce an adaptive refinement process utilizing entropy regularized OT, further refining feature alignments for accurate point-to-point correspondences. Our method demonstrates superior performance in non-rigid shape matching, including near-isometric and non-isometric scenarios, and excels in downstream tasks like segmentation transfer. The empirical results on diverse datasets highlight our framework's effectiveness and generalization capabilities, setting new standards in non-rigid shape matching with efficient OT metrics and an adaptive refinement module. Code is available at11https://github.com/Tungthanhlee/EOT-Correspondence."
                },
                {
                    "title": "Temporal 3D Shape Modeling for Video-based Cloth-Changing Person Re-Identification",
                    "abstract": "Video-based Cloth-Changing Person Re-ID (VCCRe-ID) refers to a real-world Re-ID problem where texture information like appearance or clothing becomes unreliable in long-term, limiting the applicability of traditional Re-ID methods. VCCRe-ID has not been well studied primarily due to (1) limited public datasets and (2) challenges related to extracting identity-related clothes-invariant cues from videos. Few existing works have heavily focused on gait-based features, which are severely affected under view-point changes and occlusions. In this work, we propose \u201cTemporal 3D ShapE Modeling for VCCRe-ID\u201d (SEMI), a lightweight end-to-end framework that addresses these issues by learning human 3D shape representations. The SEMI framework comprises of a Temporal 3D Shape Modeling branch, which extracts discriminative frame-wise 3D shape features using a temporal encoder, and an identity-aware 3D regressor. This is followed by a novel Attention-based Shape Aggregation (ASA) module that effectively aggregates frame-wise shape features for a fine-grained video-wise shape embedding. ASA leverages an attention mechanism to amplify the contribution of the most important frames while reducing redundancy during the aggregation process. Experiments on two VCCRe-ID datasets demonstrate that our proposed framework outperforms state-of-the-art methods by 10.7% in rank-1 accuracy and 7.4% in mAP in cloth-changing setting."
                },
                {
                    "title": "Quasi-Monte Carlo for 3D Sliced Wasserstein",
                    "abstract": "Monte Carlo (MC) integration has been employed as the standard approximation method for the Sliced Wasserstein (SW) distance, whose analytical expression involves an intractable expectation. However, MC integration is not optimal in terms of absolute approximation error. To provide a better class of empirical SW, we propose quasi-sliced Wasserstein (QSW) approximations that rely on Quasi-Monte Carlo (QMC) methods. For a comprehensive investigation of QMC for SW, we focus on the 3D setting, specifically computing the SW between probability measures in three dimensions. In greater detail, we empirically evaluate various methods to construct QMC point sets on the 3D unit-hypersphere, including the Gaussian-based and equal area mappings, generalized spiral points, and optimizing discrepancy energies. Furthermore, to obtain an unbiased estimator for stochastic optimization, we extend QSW to Randomized Quasi-Sliced Wasserstein (RQSW) by introducing randomness in the discussed point sets. Theoretically, we prove the asymptotic convergence of QSW and the unbiasedness of RQSW. Finally, we conduct experiments on various 3D tasks, such as point-cloud comparison, point-cloud interpolation, image style transfer, and training deep point-cloud autoencoders, to demonstrate the favorable performance of the proposed QSW and RQSW variants."
                },
                {
                    "title": "Hybrid-CSR: Coupling Explicit and Implicit Shape Representation for Cortical Surface Reconstruction",
                    "abstract": "We present Hybrid-CSR, a geometric deep-learning model that combines explicit and implicit shape representations for cortical surface reconstruction. Specifically, Hybrid-CSR begins with explicit deformations of template meshes to obtain coarsely reconstructed cortical surfaces, based on which the oriented point clouds are estimated for the subsequent differentiable poisson surface reconstruction. By doing so, our method unifies explicit (oriented point clouds) and implicit (indicator function) cortical surface reconstruction. Compared to explicit representation-based methods, our hybrid approach is more friendly to capture detailed structures, and when compared with implicit representation-based methods, our method can be topology aware because of end-to-end training with a mesh-based deformation module. In order to address topology defects, we propose a new topology correction pipeline that relies on optimization-based diffeomorphic surface registration. Experimental results on three brain datasets show that our approach surpasses existing implicit and explicit cortical surface reconstruction methods in numeric metrics in terms of accuracy, regularity, and consistency."
                },
                {
                    "title": "Hybrid Neural Diffeomorphic Flow for Shape Representation and Generation via Triplane",
                    "abstract": "Deep Implicit Functions (DIFs) have gained popularity in 3D computer vision due to their compactness and continuous representation capabilities. However, addressing dense correspondences and semantic relationships across DIF-encoded shapes remains a critical challenge, limiting their applications in texture transfer and shape analysis. Moreover, recent endeavors in 3D shape generation using DIFs often neglect correspondence and topology preservation. This paper presents HNDF (Hybrid Neural Diffeomorphic Flow), a method that implicitly learns the underlying representation and decomposes intricate dense correspondences into explicitly axis-aligned triplane features. To avoid suboptimal representations trapped in local minima, we propose hybrid supervision that captures both local and global correspondences. Unlike conventional approaches that directly generate new 3D shapes, we further explore the idea of shape generation with deformed template shape via diffeomorphic flows, where the deformation is encoded by the generated triplane features. Leveraging a pre-existing 2D diffusion model, we produce high-quality and diverse 3D diffeomorphic flows through generated triplanes features, ensuring topological consistency with the template shape. Extensive experiments on medical image organ segmentation datasets evaluate the effectiveness of HNDF in 3D shape representation and generation."
                },
                {
                    "title": "Sliced Wasserstein Estimation with Control Variates",
                    "abstract": "The sliced Wasserstein (SW) distances between two probability measures are defined as the expectation of the Wasserstein distance between two one-dimensional projections of the two measures. The randomness comes from a projecting direction that is used to project the two input measures to one dimension. Due to the intractability of the expectation, Monte Carlo integration is performed to estimate the value of the SW distance. Despite having various variants, there has been no prior work that improves the Monte Carlo estimation scheme for the SW distance in terms of controlling its variance. To bridge the literature on variance reduction and the literature on the SW distance, we propose computationally efficient control variates to reduce the variance of the empirical estimation of the SW distance. The key idea is to first find Gaussian approximations of projected one-dimensional measures, then we utilize the closed-form of the Wasserstein-2 distance between two Gaussian distributions to design the control variates. In particular, we propose using a lower bound and an upper bound of the Wasserstein-2 distance between two fitted Gaussians as two computationally efficient control variates. We empirically show that the proposed control variate estimators can help to reduce the variance considerably when comparing measures over images and point-clouds. Finally, we demonstrate the favorable performance of the proposed control variate estimators in gradient flows to interpolate between two point-clouds and in deep generative modeling on standard image datasets, such as CIFAR10 and CelebA."
                },
                {
                    "title": "Energy-Based Sliced Wasserstein Distance",
                    "abstract": "The sliced Wasserstein (SW) distance has been widely recognized as a statistically effective and computationally efficient metric between two probability measures. A key component of the SW distance is the slicing distribution. There are two existing approaches for choosing this distribution. The first approach is using a fixed prior distribution. The second approach is optimizing for the best distribution which belongs to a parametric family of distributions and can maximize the expected distance. However, both approaches have their limitations. A fixed prior distribution is non-informative in terms of highlighting projecting directions that can discriminate two general probability measures. Doing optimization for the best distribution is often expensive and unstable. Moreover, designing the parametric family of the candidate distribution could be easily misspecified. To address the issues, we propose to design the slicing distribution as an energy-based distribution that is parameter-free and has the density proportional to an energy function of the projected one-dimensional Wasserstein distance. We then derive a novel sliced Wasserstein metric, energy-based sliced Waserstein (EBSW) distance, and investigate its topological, statistical, and computational properties via importance sampling, sampling importance resampling, and Markov Chain methods. Finally, we conduct experiments on point-cloud gradient flow, color transfer, and point-cloud reconstruction to show the favorable performance of the EBSW."
                },
                {
                    "title": "Self-Attention Amortized Distributional Projection Optimization for Sliced Wasserstein Point-Cloud Reconstruction",
                    "abstract": "Max sliced Wasserstein (Max-SW) distance has been widely known as a solution for less discriminative projections of sliced Wasserstein (SW) distance. In applications that have various independent pairs of probability measures, amortized projection optimization is utilized to predict the ``max\"projecting directions given two input measures instead of using projected gradient ascent multiple times. Despite being efficient, Max-SW and its amortized version cannot guarantee metricity property due to the sub-optimality of the projected gradient ascent and the amortization gap. Therefore, we propose to replace Max-SW with distributional sliced Wasserstein distance with von Mises-Fisher (vMF) projecting distribution (v-DSW). Since v-DSW is a metric with any non-degenerate vMF distribution, its amortized version can guarantee the metricity when performing amortization. Furthermore, current amortized models are not permutation invariant and symmetric. To address the issue, we design amortized models based on self-attention architecture. In particular, we adopt efficient self-attention architectures to make the computation linear in the number of supports. With the two improvements, we derive self-attention amortized distributional projection optimization and show its appealing performance in point-cloud reconstruction and its downstream applications."
                },
                {
                    "title": "CorticalFlow++: Boosting Cortical Surface Reconstruction Accuracy, Regularity, and Interoperability",
                    "abstract": ". The problem of Cortical Surface Reconstruction from magnetic resonance imaging has been traditionally addressed using lengthy pipelines of image processing techniques like FreeSurfer, CAT, or CIVET. These frameworks require very long runtimes deemed unfeasible for real-time applications and unpractical for large-scale studies. Recently, supervised deep learning approaches have been introduced to speed up this task cutting down the reconstruction time from hours to seconds. Using the state-of-the-art CorticalFlow model as a blueprint, this paper proposes three modi\ufb01cations to improve its accuracy and interoperability with existing surface analysis tools, while not sacri\ufb01cing its fast inference time and low GPU memory consumption. First, we employ a more accurate ODE solver to reduce the di\ufb00eomorphic mapping approximation error. Second, we devise a routine to produce smoother template meshes avoiding mesh artifacts caused by sharp edges in CorticalFlow\u2019s convex-hull based template. Last, we recast pial surface prediction as the deformation of the predicted white surface leading to a one-to-one mapping between white and pial surface vertices. This mapping is essen-tial to many existing surface analysis tools for cortical morphometry. We name the resulting method CorticalFlow ++ . Using large-scale datasets, we demonstrate the proposed changes provide more geometric accuracy and surface regularity while keeping the reconstruction time and GPU memory requirements almost unchanged."
                },
                {
                    "title": "CorticalFlow: A Diffeomorphic Mesh Deformation Module for Cortical Surface Reconstruction",
                    "abstract": "In this paper we introduce CorticalFlow, a new geometric deep-learning model that, given a 3-dimensional image, learns to deform a reference template towards a targeted object. To conserve the template mesh's topological properties, we train our model over a set of diffeomorphic transformations. This new implementation of a flow Ordinary Differential Equation (ODE) framework benefits from a small GPU memory footprint, allowing the generation of surfaces with several hundred thousand vertices. To reduce topological errors introduced by its discrete resolution, we derive numeric conditions which improve the manifoldness of the predicted triangle mesh. To exhibit the utility of CorticalFlow, we demonstrate its performance for the challenging task of brain cortical surface reconstruction. In contrast to current state-of-the-art, CorticalFlow produces superior surfaces while reducing the computation time from nine and a half minutes to one second. More significantly, CorticalFlow enforces the generation of anatomically plausible surfaces; the absence of which has been a major impediment restricting the clinical relevance of such surface reconstruction methods."
                },
                {
                    "title": "Statistical inference with regularized optimal transport",
                    "abstract": "\n Optimal transport (OT) is a versatile framework for comparing probability measures, with many applications to statistics, machine learning and applied mathematics. However, OT distances suffer from computational and statistical scalability issues to high dimensions, which motivated the study of regularized OT methods like slicing, smoothing and entropic penalty. This work establishes a unified framework for deriving limit distributions of empirical regularized OT distances, semiparametric efficiency of the plug-in empirical estimator and bootstrap consistency. We apply the unified framework to provide a comprehensive statistical treatment of (i) average- and max-sliced $p$-Wasserstein distances, for which several gaps in existing literature are closed; (ii) smooth distances with compactly supported kernels, the analysis of which is motivated by computational considerations; and (iii) entropic OT, for which our method generalizes existing limit distribution results and establishes, for the first time, efficiency and bootstrap consistency. While our focus is on these three regularized OT distances as applications, the flexibility of the proposed framework renders it applicable to broad classes of functionals beyond these examples."
                },
                {
                    "title": "Vox2Cortex: Fast Explicit Reconstruction of Cortical Surfaces from 3D MRI Scans with Geometric Deep Neural Networks",
                    "abstract": "The reconstruction of cortical surfaces from brain magnetic resonance imaging (MRI) scans is essential for quantitative analyses of cortical thickness and sulcal morphology. Although traditional and deep learning-based algorithmic pipelines exist for this purpose, they have two major drawbacks: lengthy runtimes of multiple hours (traditional) or intricate post-processing, such as mesh extraction and topology correction (deep learning-based). In this work, we address both of these issues and propose Vox2Cortex, a deep learning-based algorithm that directly yields topologically correct, three-dimensional meshes of the boundaries of the cortex. Vox2Cortex leverages convolutional and graph convolutional neural networks to deform an initial template to the densely folded geometry of the cortex represented by an input MRI scan. We show in extensive experiments on three brain MRI datasets that our meshes are as accurate as the ones reconstructed by state-of-the-art methods in the field, without the need for time- and resource-intensive post-processing. To accurately reconstruct the tightly folded cortex, we work with meshes containing about 168,000 vertices at test time, scaling deep explicit reconstruction methods to a new level."
                },
                {
                    "title": "Topology-Preserving Shape Reconstruction and Registration via Neural Diffeomorphic Flow",
                    "abstract": "Deep Implicit Functions (DIFs) represent 3D geometry with continuous signed distance functions learned through deep neural nets. Recently DIFs-based methods have been proposed to handle shape reconstruction and dense point correspondences simultaneously, capturing semantic relationships across shapes of the same class by learning a DIFs-modeled shape template. These methods provide great flexibility and accuracy in reconstructing 3D shapes and inferring correspondences. However, the point correspondences built from these methods do not intrinsically preserve the topology of the shapes, unlike mesh-based template matching methods. This limits their applications on 3D geometries where underlying topological structures exist and matter, such as anatomical structures in medical images. In this paper, we propose a new model called Neural Diffeomorphic Flow (NDF) to learn deep implicit shape templates, representing shapes as conditional diffeomorphic deformations of templates, intrinsically preserving shape topologies. The diffeomorphic deformation is realized by an autodecoder consisting of Neural Ordinary Differential Equation (NODE) blocks that progressively map shapes to implicit templates. We conduct extensive experiments on several medical image organ segmentation datasets to evaluate the effectiveness of NDF on reconstructing and aligning shapes. NDF achieves consistently state-of-the-art organ shape reconstruction and registration results in both accuracy and quality. The source code is publicly available at https://github.com/Siwensun/Neural_Diffeomorphic_Flow-NDF."
                },
                {
                    "title": "Diffeomorphic Image Registration with Neural Velocity Field",
                    "abstract": "Diffeomorphic image registration, offering smooth transformation and topology preservation, is required in many medical image analysis tasks. Traditional methods impose certain modeling constraints on the space of admissible transformations and use optimization to find the optimal transformation between two images. Specifying the right space of admissible transformations is challenging: the registration quality can be poor if the space is too restrictive, while the optimization can be hard to solve if the space is too general. Recent learning-based methods, utilizing deep neural networks to learn the transformation directly, achieve fast inference, but face challenges in accuracy due to the difficulties in capturing the small local deformations and generalization ability. Here we propose a new optimization-based method named DNVF (Diffeomorphic Image Registration with Neural Velocity Field) which utilizes deep neural network to model the space of admissible transformations. A multilayer perceptron (MLP) with sinusoidal activation function is used to represent the continuous velocity field and assigns a velocity vector to every point in space, providing the flexibility of modeling complex deformations as well as the convenience of optimization. Moreover, we propose a cascaded image registration framework (Cas-DNVF) by combining the benefits of both optimization and learning based methods, where a fully convolutional neural network (FCN) is trained to predict the initial deformation, followed by DNVF for further refinement. Experiments on two large-scale 3D MR brain scan datasets demonstrate that our proposed methods significantly outperform the state-of-the-art registration methods."
                },
                {
                    "title": "CortexODE: Learning Cortical Surface Reconstruction by Neural ODEs",
                    "abstract": "We present CortexODE, a deep learning framework for cortical surface reconstruction. CortexODE leverages neural ordinary differential equations (ODEs) to deform an input surface into a target shape by learning a diffeomorphic flow. The trajectories of the points on the surface are modeled as ODEs, where the derivatives of their coordinates are parameterized via a learnable Lipschitz-continuous deformation network. This provides theoretical guarantees for the prevention of self-intersections. CortexODE can be integrated to an automatic learning-based pipeline, which reconstructs cortical surfaces efficiently in less than 5 seconds. The pipeline utilizes a 3D U-Net to predict a white matter segmentation from brain Magnetic Resonance Imaging (MRI) scans, and further generates a signed distance function that represents an initial surface. Fast topology correction is introduced to guarantee homeomorphism to a sphere. Following the isosurface extraction step, two CortexODE models are trained to deform the initial surface to white matter and pial surfaces respectively. The proposed pipeline is evaluated on large-scale neuroimage datasets in various age groups including neonates (25-45 weeks), young adults (22-36 years) and elderly subjects (55-90 years). Our experiments demonstrate that the CortexODE-based pipeline can achieve less than 0.2mm average geometric error while being orders of magnitude faster compared to conventional processing pipelines."
                },
                {
                    "title": "Do Neural Optimal Transport Solvers Work? A Continuous Wasserstein-2 Benchmark",
                    "abstract": "Despite the recent popularity of neural network-based solvers for optimal transport (OT), there is no standard quantitative way to evaluate their performance. In this paper, we address this issue for quadratic-cost transport -- specifically, computation of the Wasserstein-2 distance, a commonly-used formulation of optimal transport in machine learning. To overcome the challenge of computing ground truth transport maps between continuous measures needed to assess these solvers, we use input-convex neural networks (ICNN) to construct pairs of measures whose ground truth OT maps can be obtained analytically. This strategy yields pairs of continuous benchmark measures in high-dimensional spaces such as spaces of images. We thoroughly evaluate existing optimal transport solvers using these benchmark measures. Even though these solvers perform well in downstream tasks, many do not faithfully recover optimal transport maps. To investigate the cause of this discrepancy, we further test the solvers in a setting of image generation. Our study reveals crucial limitations of existing solvers and shows that increased OT accuracy does not necessarily correlate to better results downstream."
                },
                {
                    "title": "Point-set Distances for Learning Representations of 3D Point Clouds",
                    "abstract": "Learning an effective representation of 3D point clouds requires a good metric to measure the discrepancy between two 3D point sets, which is non-trivial due to their irregularity. Most of the previous works resort to using the Chamfer discrepancy or Earth Mover\u2019s distance, but those metrics are either ineffective in measuring the differences between point clouds or computationally expensive. In this paper, we conduct a systematic study with extensive experiments on distance metrics for 3D point clouds. From this study, we propose to use sliced Wasserstein distance and its variants for learning representations of 3D point clouds. In addition, we introduce a new algorithm to estimate sliced Wasserstein distance that guarantees that the estimated value is close enough to the true one. Experiments show that the sliced Wasserstein distance and its variants allow the neural network to learn a more efficient representation compared to the Chamfer discrepancy. We demonstrate the efficiency of the sliced Wasserstein metric and its variants on several tasks in 3D computer vision including training a point cloud autoencoder, generative modeling, transfer learning, and point cloud registration."
                },
                {
                    "title": "DeepCSR: A 3D Deep Learning Approach for Cortical Surface Reconstruction",
                    "abstract": "The study of neurodegenerative diseases relies on the reconstruction and analysis of the brain cortex from magnetic resonance imaging (MRI). Traditional frameworks for this task like FreeSurfer demand lengthy runtimes, while its accelerated variant FastSurfer still relies on a voxel-wise segmentation which is limited by its resolution to capture narrow continuous objects as cortical surfaces. Having these limitations in mind, we propose DeepCSR, a 3D deep learning framework for cortical surface reconstruction from MRI. Towards this end, we train a neural network model with hypercolumn features to predict implicit surface representations for points in a brain template space. After training, the cortical surface at a desired level of detail is obtained by evaluating surface representations at specific coordinates, and subsequently applying a topology correction algorithm and an isosurface extraction method. Thanks to the continuous nature of this approach and the efficacy of its hypercolumn features scheme, DeepCSR efficiently reconstructs cortical surfaces at high resolution capturing fine details in the cortical folding. Moreover, DeepCSR is as accurate, more precise, and faster than the widely used FreeSurfer toolbox and its deep learning powered variant FastSurfer on reconstructing cortical surfaces from MRI which should facilitate large-scale medical studies and new healthcare applications."
                },
                {
                    "title": "Neural Mesh Flow: 3D Manifold Mesh Generationvia Diffeomorphic Flows",
                    "abstract": "Meshes are important representations of physical 3D entities in the virtual world. Applications like rendering, simulations and 3D printing require meshes to be manifold so that they can interact with the world like the real objects they represent. Prior methods generate meshes with great geometric accuracy but poor manifoldness. In this work, we propose Neural Mesh Flow (NMF) to generate two-manifold meshes for genus-0 shapes. Specifically, NMF is a shape auto-encoder consisting of several Neural Ordinary Differential Equation (NODE)[1]blocks that learn accurate mesh geometry by progressively deforming a spherical mesh. Training NMF is simpler compared to state-of-the-art methods since it does not require any explicit mesh-based regularization. Our experiments demonstrate that NMF facilitates several applications such as single-view mesh reconstruction, global shape parameterization, texture mapping, shape deformation and correspondence. Importantly, we demonstrate that manifold meshes generated using NMF are better-suited for physically-based rendering and simulation. Code and data will be released."
                },
                {
                    "title": "Multiview Neural Surface Reconstruction by Disentangling Geometry and Appearance",
                    "abstract": "In this work we address the challenging problem of multiview 3D surface reconstruction. We introduce a neural network architecture that simultaneously learns the unknown geometry, camera parameters, and a neural renderer that approximates the light reflected from the surface towards the camera. The geometry is represented as a zero level-set of a neural network, while the neural renderer, derived from the rendering equation, is capable of (implicitly) modeling a wide set of lighting conditions and materials. We trained our network on real world 2D images of objects with different material properties, lighting conditions, and noisy camera initializations from the DTU MVS dataset. We found our model to produce state of the art 3D surface reconstructions with high fidelity, resolution and detail."
                },
                {
                    "title": "Local Implicit Grid Representations for 3D Scenes",
                    "abstract": "Shape priors learned from data are commonly used to reconstruct 3D objects from partial or noisy data. Yet no such shape priors are available for indoor scenes, since typical 3D autoencoders cannot handle their scale, complexity, or diversity. In this paper, we introduce Local Implicit Grid Representations, a new 3D shape representation designed for scalability and generality. The motivating idea is that most 3D surfaces share geometric details at some scale -- i.e., at a scale smaller than an entire object and larger than a small patch. We train an autoencoder to learn an embedding of local crops of 3D shapes at that size. Then, we use the decoder as a component in a shape optimization that solves for a set of latent codes on a regular grid of overlapping crops such that an interpolation of the decoded local shapes matches a partial or noisy observation. We demonstrate the value of this proposed approach for 3D surface reconstruction from sparse point observations, showing significantly better results than alternative approaches."
                },
                {
                    "title": "Distributional Sliced-Wasserstein and Applications to Generative Modeling",
                    "abstract": "Sliced-Wasserstein distance (SWD) and its variation, Max Sliced-Wasserstein distance (Max-SWD), have been widely used in the recent years due to their fast computation and scalability when the probability measures lie in very high dimension. However, these distances still have their weakness, SWD requires a lot of projection samples because it uses the uniform distribution to sample projecting directions, Max-SWD uses only one projection, causing it to lose a large amount of information. In this paper, we propose a novel distance that finds optimal penalized probability measure over the slices, which is named Distributional Sliced-Wasserstein distance (DSWD). We show that the DSWD is a generalization of both SWD and Max-SWD, and the proposed distance could be found by searching for the push-forward measure over a set of measures satisfying some certain constraints. Moreover, similar to SWD, we can extend Generalized Sliced-Wasserstein distance (GSWD) to Distributional Generalized Sliced-Wasserstein distance (DGSWD). Finally, we carry out extensive experiments to demonstrate the favorable generative modeling performances of our distances over the previous sliced-based distances in large-scale real datasets."
                },
                {
                    "title": "Differentiable Volumetric Rendering: Learning Implicit 3D Representations Without 3D Supervision",
                    "abstract": "Learning-based 3D reconstruction methods have shown impressive results. However, most methods require 3D supervision which is often hard to obtain for real-world datasets. Recently, several works have proposed differentiable rendering techniques to train reconstruction models from RGB images. Unfortunately, these approaches are currently restricted to voxel- and mesh-based representations, suffering from discretization or low resolution. In this work, we propose a differentiable rendering formulation for implicit shape and texture representations. Implicit representations have recently gained popularity as they represent shape and texture continuously. Our key insight is that depth gradients can be derived analytically using the concept of implicit differentiation. This allows us to learn implicit shape and texture representations directly from RGB images. We experimentally show that our single-view reconstructions rival those learned with full 3D supervision. Moreover, we find that our method can be used for multi-view 3D reconstruction, directly resulting in watertight meshes."
                },
                {
                    "title": "Accelerating 3D deep learning with PyTorch3D",
                    "abstract": "1. Accelerating 3D Deep Learning with PyTorch3D, arXiv 2007.08501 2. Mesh R-CNN, ICCV 2019 3. SynSin: End-to-end View Synthesis from a Single Image, CVPR 2020 4. Fast Differentiable Raycasting for Neural Rendering using Sphere-based Representations, arXiv 2004.07484"
                },
                {
                    "title": "Sinkhorn Divergences for Unbalanced Optimal Transport",
                    "abstract": "This paper extends the formulation of Sinkhorn divergences to the unbalanced setting of arbitrary positive measures, providing both theoretical and algorithmic advances. Sinkhorn divergences leverage the entropic regularization of Optimal Transport (OT) to define geometric loss functions. They are differentiable, cheap to compute and do not suffer from the curse of dimensionality, while maintaining the geometric properties of OT, in particular they metrize the weak$^*$ convergence. Extending these divergences to the unbalanced setting is of utmost importance since most applications in data sciences require to handle both transportation and creation/destruction of mass. This includes for instance problems as diverse as shape registration in medical imaging, density fitting in statistics, generative modeling in machine learning, and particles flows involving birth/death dynamics. Our first set of contributions is the definition and the theoretical analysis of the unbalanced Sinkhorn divergences. They enjoy the same properties as the balanced divergences (classical OT), which are obtained as a special case. Indeed, we show that they are convex, differentiable and metrize the weak$^*$ convergence. Our second set of contributions studies generalized Sinkkhorn iterations, which enable a fast, stable and massively parallelizable algorithm to compute these divergences. We show, under mild assumptions, a linear rate of convergence, independent of the number of samples, i.e. which can cope with arbitrary input measures. We also highlight the versatility of this method, which takes benefit from the latest advances in term of GPU computing, for instance through the KeOps library for fast and scalable kernel operations."
                },
                {
                    "title": "Deep Mesh Reconstruction From Single RGB Images via Topology Modification Networks",
                    "abstract": "Reconstructing the 3D mesh of a general object from a single image is now possible thanks to the latest advances of deep learning technologies. However, due to the nontrivial difficulty of generating a feasible mesh structure, the state-of-the-art approaches often simplify the problem by learning the displacements of a template mesh that deforms it to the target surface. Though reconstructing a 3D shape with complex topology can be achieved by deforming multiple mesh patches, it remains difficult to stitch the results to ensure a high meshing quality. In this paper, we present an end-to-end single-view mesh reconstruction framework that is able to generate high-quality meshes with complex topologies from a single genus-0 template mesh. The key to our approach is a novel progressive shaping framework that alternates between mesh deformation and topology modification. While a deformation network predicts the per-vertex translations that reduce the gap between the reconstructed mesh and the ground truth, a novel topology modification network is employed to prune the error-prone faces, enabling the evolution of topology. By iterating over the two procedures, one can progressively modify the mesh topology while achieving higher reconstruction accuracy. Moreover, a boundary refinement network is designed to refine the boundary conditions to further improve the visual quality of the reconstructed mesh. Extensive experiments demonstrate that our approach outperforms the current state-of-the-art methods both qualitatively and quantitatively, especially for the shapes with complex topologies."
                },
                {
                    "title": "Statistical bounds for entropic optimal transport: sample complexity and the central limit theorem",
                    "abstract": "We prove several fundamental statistical bounds for entropic OT with the squared Euclidean cost between subgaussian probability measures in arbitrary dimension. First, through a new sample complexity result we establish the rate of convergence of entropic OT for empirical measures. Our analysis improves exponentially on the bound of Genevay et al. (2019) and extends their work to unbounded measures. Second, we establish a central limit theorem for entropic OT, based on techniques developed by Del Barrio and Loubes (2019). Previously, such a result was only known for finite metric spaces. As an application of our results, we develop and analyze a new technique for estimating the entropy of a random variable corrupted by gaussian noise."
                },
                {
                    "title": "DISN: Deep Implicit Surface Network for High-quality Single-view 3D Reconstruction",
                    "abstract": "Reconstructing 3D shapes from single-view images has been a long-standing research problem. In this paper, we present DISN, a Deep Implicit Surface Network which can generate a high-quality detail-rich 3D mesh from an 2D image by predicting the underlying signed distance fields. In addition to utilizing global image features, DISN predicts the projected location for each 3D point on the 2D image, and extracts local features from the image feature maps. Combining global and local features significantly improves the accuracy of the signed distance field prediction, especially for the detail-rich areas. To the best of our knowledge, DISN is the first method that constantly captures details such as holes and thin structures present in 3D shapes from single-view images. DISN achieves the state-of-the-art single-view reconstruction performance on a variety of shape categories reconstructed from both synthetic and real images. Code is available at this https URL The supplementary can be found at this https URL"
                },
                {
                    "title": "3D Point Cloud Denoising via Deep Neural Network Based Local Surface Estimation",
                    "abstract": "We present a neural-network-based architecture for 3D point cloud denoising called neural projection denoising (NPD). In our previous work, we proposed a two-stage denoising algorithm, which first estimates reference planes and follows by projecting noisy points to estimated reference planes. Since the estimated reference planes are inevitably noisy, multi-projection is applied to stabilize the denoising performance. NPD algorithm uses a neural network to estimate reference planes for points in noisy point clouds. With more accurate estimations of reference planes, we are able to achieve better denoising performances with only one-time projection. To the best of our knowledge, NPD is the first work to denoise 3D point clouds with deep learning techniques. To conduct the experiments, we sample 40000 point clouds from the 3D data in ShapeNet to train a network and sample 350 point clouds from the 3D data in ModelNet10 to test. Experimental results show that our algorithm can estimate normal vectors of points in noisy point clouds. Comparing to five competitive methods, the proposed algorithm achieves better denoising performance and produces much smaller variances. Our code is available at https://github.com/chaojingduan/Neural-Projection."
                },
                {
                    "title": "3DN: 3D Deformation Network",
                    "abstract": "Applications in virtual and augmented reality create a demand for rapid creation and easy access to large sets of 3D models. An effective way to address this demand is to edit or deform existing 3D models based on a reference, e.g., a 2D image which is very easy to acquire. Given such a source 3D model and a target which can be a 2D image, 3D model, or a point cloud acquired as a depth scan, we introduce 3DN, an end-to-end network that deforms the source model to resemble the target. Our method infers per-vertex offset displacements while keeping the mesh connectivity of the source model fixed. We present a training strategy which uses a novel differentiable operation, mesh sampling operator, to generalize our method across source and target models with varying mesh densities. Mesh sampling operator can be seamlessly integrated into the network to handle meshes with different topologies. Qualitative and quantitative results show that our method generates higher quality results compared to the state-of-the art learning-based methods for 3D shape generation."
                },
                {
                    "title": "GEOMetrics: Exploiting Geometric Structure for Graph-Encoded Objects",
                    "abstract": "Mesh models are a promising approach for encoding the structure of 3D objects. Current mesh reconstruction systems predict uniformly distributed vertex locations of a predetermined graph through a series of graph convolutions, leading to compromises with respect to performance or resolution. In this paper, we argue that the graph representation of geometric objects allows for additional structure, which should be leveraged for enhanced reconstruction. Thus, we propose a system which properly benefits from the advantages of the geometric structure of graph encoded objects by introducing (1) a graph convolutional update preserving vertex information; (2) an adaptive splitting heuristic allowing detail to emerge; and (3) a training objective operating both on the local surfaces defined by vertices as well as the global structure defined by the mesh. Our proposed method is evaluated on the task of 3D object reconstruction from images with the ShapeNet dataset, where we demonstrate state of the art performance, both visually and numerically, while having far smaller space requirements by generating adaptive meshes"
                },
                {
                    "title": "DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation",
                    "abstract": "Computer graphics, 3D computer vision and robotics communities have produced multiple approaches to representing 3D geometry for rendering and reconstruction. These provide trade-offs across fidelity, efficiency and compression capabilities. In this work, we introduce DeepSDF, a learned continuous Signed Distance Function (SDF) representation of a class of shapes that enables high quality shape representation, interpolation and completion from partial and noisy 3D input data. DeepSDF, like its classical counterpart, represents a shape's surface by a continuous volumetric field: the magnitude of a point in the field represents the distance to the surface boundary and the sign indicates whether the region is inside (-) or outside (+) of the shape, hence our representation implicitly encodes a shape's boundary as the zero-level-set of the learned function while explicitly representing the classification of space as being part of the shapes interior or not. While classical SDF's both in analytical or discretized voxel form typically represent the surface of a single shape, DeepSDF can represent an entire class of shapes. Furthermore, we show state-of-the-art performance for learned 3D shape representation and completion while reducing the model size by an order of magnitude compared with previous work."
                },
                {
                    "title": "Learning a Probabilistic Model for Diffeomorphic Registration",
                    "abstract": "We propose to learn a low-dimensional probabilistic deformation model from data which can be used for the registration and the analysis of deformations. The latent variable model maps similar deformations close to each other in an encoding space. It enables to compare deformations, to generate normal or pathological deformations for any new image, or to transport deformations from one image pair to any other image. Our unsupervised method is based on the variational inference. In particular, we use a conditional variational autoencoder network and constrain transformations to be symmetric and diffeomorphic by applying a differentiable exponentiation layer with a symmetric loss function. We also present a formulation that includes spatial regularization such as the diffusion-based filters. In addition, our framework provides multi-scale velocity field estimations. We evaluated our method on 3-D intra-subject registration using 334 cardiac cine-MRIs. On this dataset, our method showed the state-of-the-art performance with a mean DICE score of 81.2% and a mean Hausdorff distance of 7.3 mm using 32 latent dimensions compared to three state-of-the-art methods while also demonstrating more regular deformation fields. The average time per registration was 0.32 s. Besides, we visualized the learned latent space and showed that the encoded deformations can be used to transport deformations and to cluster diseases with a classification accuracy of 83% after applying a linear projection."
                },
                {
                    "title": "Occupancy Networks: Learning 3D Reconstruction in Function Space",
                    "abstract": "With the advent of deep neural networks, learning-based approaches for 3D reconstruction have gained popularity. However, unlike for images, in 3D there is no canonical representation which is both computationally and memory efficient yet allows for representing high-resolution geometry of arbitrary topology. Many of the state-of-the-art learning-based 3D reconstruction approaches can hence only represent very coarse 3D geometry or are limited to a restricted domain. In this paper, we propose Occupancy Networks, a new representation for learning-based 3D reconstruction methods. Occupancy networks implicitly represent the 3D surface as the continuous decision boundary of a deep neural network classifier. In contrast to existing approaches, our representation encodes a description of the 3D output at infinite resolution without excessive memory footprint. We validate that our representation can efficiently encode 3D structure and can be inferred from various kinds of input. Our experiments demonstrate competitive results, both qualitatively and quantitatively, for the challenging tasks of 3D reconstruction from single images, noisy point clouds and coarse discrete voxel grids. We believe that occupancy networks will become a useful tool in a wide variety of learning-based 3D tasks."
                },
                {
                    "title": "Learning Implicit Fields for Generative Shape Modeling",
                    "abstract": "We advocate the use of implicit fields for learning generative models of shapes and introduce an implicit field decoder, called IM-NET, for shape generation, aimed at improving the visual quality of the generated shapes. An implicit field assigns a value to each point in 3D space, so that a shape can be extracted as an iso-surface. IM-NET is trained to perform this assignment by means of a binary classifier. Specifically, it takes a point coordinate, along with a feature vector encoding a shape, and outputs a value which indicates whether the point is outside the shape or not. By replacing conventional decoders by our implicit decoder for representation learning (via IM-AE) and shape generation (via IM-GAN), we demonstrate superior results for tasks such as generative shape modeling, interpolation, and single-view 3D reconstruction, particularly in terms of visual quality. Code and supplementary material are available at https://github.com/czq142857/implicit-decoder."
                },
                {
                    "title": "Interpolating between Optimal Transport and MMD using Sinkhorn Divergences",
                    "abstract": "Comparing probability distributions is a fundamental problem in data sciences. Simple norms and divergences such as the total variation and the relative entropy only compare densities in a point-wise manner and fail to capture the geometric nature of the problem. In sharp contrast, Maximum Mean Discrepancies (MMD) and Optimal Transport distances (OT) are two classes of distances between measures that take into account the geometry of the underlying space and metrize the convergence in law.\n\nThis paper studies the Sinkhorn divergences, a family of geometric divergences that interpolates between MMD and OT. Relying on a new notion of geometric entropy, we provide theoretical guarantees for these divergences: positivity, convexity and metrization of the convergence in law. On the practical side, we detail a numerical scheme that enables the large scale application of these divergences for machine learning: on the GPU, gradients of the Sinkhorn loss can be computed for batches of a million samples."
                },
                {
                    "title": "VoxelMorph: A Learning Framework for Deformable Medical Image Registration",
                    "abstract": "We present VoxelMorph, a fast learning-based framework for deformable, pairwise medical image registration. Traditional registration methods optimize an objective function for each pair of images, which can be time-consuming for large datasets or rich deformation models. In contrast to this approach and building on recent learning-based methods, we formulate registration as a function that maps an input image pair to a deformation field that aligns these images. We parameterize the function via a convolutional neural network and optimize the parameters of the neural network on a set of images. Given a new pair of scans, VoxelMorph rapidly computes a deformation field by directly evaluating the function. In this paper, we explore two different training strategies. In the first (unsupervised) setting, we train the model to maximize standard image matching objective functions that are based on the image intensities. In the second setting, we leverage auxiliary segmentations available in the training data. We demonstrate that the unsupervised model\u2019s accuracy is comparable to the state-of-the-art methods while operating orders of magnitude faster. We also show that VoxelMorph trained with auxiliary data improves registration accuracy at test time and evaluate the effect of training set size on registration. Our method promises to speed up medical image analysis and processing pipelines while facilitating novel directions in learning-based registration and its applications. Our code is freely available at https://github.com/voxelmorph/voxelmorph."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "A Papier-Mache Approach to Learning 3D Surface Generation",
                    "abstract": "We introduce a method for learning to generate the surface of 3D shapes. Our approach represents a 3D shape as a collection of parametric surface elements and, in contrast to methods generating voxel grids or point clouds, naturally infers a surface representation of the shape. Beyond its novelty, our new shape generation framework, AtlasNet, comes with significant advantages, such as improved precision and generalization capabilities, and the possibility to generate a shape of arbitrary resolution without memory issues. We demonstrate these benefits and compare to strong baselines on the ShapeNet benchmark for two applications: (i) autoencoding shapes, and (ii) single-view reconstruction from a still image. We also provide results showing its potential for other applications, such as morphing, parametrization, super-resolution, matching, and co-segmentation."
                },
                {
                    "title": "Diffeomorphic Registration of Discrete Geometric Distributions",
                    "abstract": "This paper proposes a new framework and algorithms to address the problem of diffeomorphic registration on a general class of geometric objects that can be described as discrete distributions of local direction vectors. It builds on both the large deformation diffeomorphic metric mapping (LDDMM) model and the concept of oriented varifolds introduced in previous works like [Kaltenmark 2017]. Unlike previous approaches in which varifold representations are only used as surrogates to define and evaluate fidelity terms, the specificity of this paper is to derive direct deformation models and corresponding matching algorithms for discrete varifolds. We show that it gives on the one hand an alternative numerical setting for curve and surface matching but that it can also handle efficiently more general shape structures, including multi-directional objects or multi-modal images represented as distributions of unit gradient vectors."
                },
                {
                    "title": "Wasserstein Generative Adversarial Networks",
                    "abstract": "We introduce a new algorithm named WGAN, an alternative to traditional GAN training. In this new model, we show that we can improve the stability of learning, get rid of problems like mode collapse, and provide meaningful learning curves useful for debugging and hyperparameter searches. Furthermore, we show that the corresponding optimization problem is sound, and provide extensive theoretical work highlighting the deep connections to different distances between distributions."
                },
                {
                    "title": "Learning Representations and Generative Models for 3D Point Clouds",
                    "abstract": "Three-dimensional geometric data offer an excellent domain for studying representation learning and generative modeling. In this paper, we look at geometric data represented as point clouds. We introduce a deep AutoEncoder (AE) network with state-of-the-art reconstruction quality and generalization ability. The learned representations outperform existing methods on 3D recognition tasks and enable shape editing via simple algebraic manipulations, such as semantic part editing, shape analogies and shape interpolation, as well as shape completion. We perform a thorough study of different generative models including GANs operating on the raw point clouds, significantly improved GANs trained in the fixed latent space of our AEs, and Gaussian Mixture Models (GMMs). To quantitatively evaluate generative models we introduce measures of sample fidelity and diversity based on matchings between sets of point clouds. Interestingly, our evaluation of generalization, fidelity and diversity reveals that GMMs trained in the latent space of our AEs yield the best results overall."
                },
                {
                    "title": "A General Framework for Curve and Surface Comparison and Registration with Oriented Varifolds",
                    "abstract": "This paper introduces a general setting for the construction of data fidelity metrics between oriented or non-oriented geometric shapes like curves, curve sets or surfaces. These metrics are based on the representation of shapes as distributions of their local tangent or normal vectors and the definition of reproducing kernels on these spaces. The construction, that combines in one common setting and extends the previous frameworks of currents and varifolds, provides a very large class of kernel metrics which can be easily computed without requiring any kind of parametrization of shapes and which are smooth enough to give robustness to certain imperfections that could result e.g. from bad segmentation. We then give a sense, with synthetic examples, of the versatility and potentialities of such metrics when used in various problems like shape comparison, clustering and diffeomorphic registration."
                },
                {
                    "title": "Hierarchical Surface Prediction for 3D Object Reconstruction",
                    "abstract": "Recently, Convolutional Neural Networks have shown promising results for 3D geometry prediction. They can make predictions from very little input data such as a single color image. A major limitation of such approaches is that they only predict a coarse resolution voxel grid, which does not capture the surface of the objects well. We propose a general framework, called hierarchical surface prediction (HSP), which facilitates prediction of high resolution voxel grids. The main insight is that it is sufficient to predict high resolution voxels around the predicted surfaces. The exterior and interior of the objects can be represented with coarse resolution voxels. Our approach is not dependent on a specific input type. We show results for geometry prediction from color images, depth images and shape completion from partial voxel grids. Our analysis shows that our high resolution predictions are more accurate than low resolution predictions."
                },
                {
                    "title": "Octree Generating Networks: Efficient Convolutional Architectures for High-resolution 3D Outputs",
                    "abstract": "We present a deep convolutional decoder architecture that can generate volumetric 3D outputs in a compute- and memory-efficient manner by using an octree representation. The network learns to predict both the structure of the octree, and the occupancy values of individual cells. This makes it a particularly valuable technique for generating 3D shapes. In contrast to standard decoders acting on regular voxel grids, the architecture does not have cubic complexity. This allows representing much higher resolution outputs with a limited memory budget. We demonstrate this in several application domains, including 3D convolutional autoencoders, generation of objects and whole scenes from high-level representations, and shape from a single image."
                },
                {
                    "title": "On parameter estimation with the Wasserstein distance",
                    "abstract": "\n Statistical inference can be performed by minimizing, over the parameter space, the Wasserstein distance between model distributions and the empirical distribution of the data. We study asymptotic properties of such minimum Wasserstein distance estimators, complementing results derived by Bassetti, Bodini and Regazzini in 2006. In particular, our results cover the misspecified setting, in which the data-generating process is not assumed to be part of the family of distributions described by the model. Our results are motivated by recent applications of minimum Wasserstein estimators to complex generative models. We discuss some difficulties arising in the numerical approximation of these estimators. Two of our numerical examples ($g$-and-$\\kappa$ and sum of log-normals) are taken from the literature on approximate Bayesian computation and have likelihood functions that are not analytically tractable. Two other examples involve misspecified models."
                },
                {
                    "title": "Analysis of geometric and functional shapes with extensions of currents : applications to registration and atlas estimation",
                    "abstract": "This thesis addresses several questions related to the recent field of computational anatomy. Broadly speaking, computational anatomy intends to analyse shape variability among populations of anatomical structures. In this work, we are focused, in the first place, on the case of datasets of curves, surfaces and more generally submanifolds. Our goal is to provide a mathematical and numerical setting to build relevant data attachment terms between those objects in the purpose of embedding it into the large diffeomorphic metric mapping (LDDMM) model for shape registration. Previous approaches have been relying on the concept of currents that represents oriented submanifolds. We first propose an extension of these methods to the situation of non-oriented shapes by adapting the concept of varifolds from geometric measure theory. In the second place, we focus on the study of geometrico-functional structures we call 'functional shapes' (or fshapes), which combine varying geometries across individuals with signal functions defined on these shapes. We introduce the new notion of fshape metamorphosis to generalize the idea of deformation groups in the pure geometrical case. In addition, we define the extended setting of 'functional currents' to quantify dissimilarity between fshapes and thus perform geometrico-functional registration between such objects. Finally, in the last part of the thesis, we move on to the issue of analyzing entire groups of individuals (shapes or fshapes) together. In that perspective, we introduce an atlas estimation variational formulation that we prove to be mathematically well-posed and build algorithms to estimate templates and atlases from populations, as well as tools to perform statistical analysis and classification. All these methods are evaluated on several applications to synthetic datasets on the one hand and real datasets from biomedical imaging on the other."
                },
                {
                    "title": "Sinkhorn Distances: Lightspeed Computation of Optimal Transport",
                    "abstract": "Optimal transportation distances are a fundamental family of parameterized distances for histograms. Despite their appealing theoretical properties, excellent performance in retrieval tasks and intuitive formulation, their computation involves the resolution of a linear program whose cost is prohibitive whenever the histograms' dimension exceeds a few hundreds. We propose in this work a new family of optimal transportation distances that look at transportation problems from a maximum-entropy perspective. We smooth the classical optimal transportation problem with an entropic regularization term, and show that the resulting optimum is also a distance which can be computed through Sinkhorn-Knopp's matrix scaling algorithm at a speed that is several orders of magnitude faster than that of transportation solvers. We also report improved performance over classical optimal transportation distances on the MNIST benchmark problem."
                },
                {
                    "title": "The Varifold Representation of Nonoriented Shapes for Diffeomorphic Registration",
                    "abstract": "In this paper, we address the problem of orientation that naturally arises when representing shapes such as curves or surfaces as currents. In the field of computational anatomy, the framework of currents has indeed proved very efficient in modeling a wide variety of shapes. However, in such approaches, orientation of shapes is a fundamental issue that can lead to several drawbacks in treating certain kinds of datasets. More specifically, problems occur with structures like acute pikes because of canceling effects of currents or with data that consists in many disconnected pieces like fiber bundles for which currents require a consistent orientation of all pieces. As a promising alternative to currents, varifolds, introduced in the context of geometric measure theory by Almgren, allow the representation of any nonoriented manifold (more generally any nonoriented rectifiable set). In particular, we explain how varifolds can encode numerically nonoriented objects both from the discrete and the continuous po..."
                },
                {
                    "title": "A Bayesian Generative Model for Surface Template Estimation",
                    "abstract": "3D surfaces are important geometric models for many objects of interest in image analysis and Computational Anatomy. In this paper, we describe a Bayesian inference scheme for estimating a template surface from a set of observed surface data. In order to achieve this, we use the geodesic shooting approach to construct a statistical model for the generation and the observations of random surfaces. We develop a mode approximation EM algorithm to infer the maximum a posteriori estimation of initial momentum \u03bc, which determines the template surface. Experimental results of caudate, thalamus, and hippocampus data are presented."
                },
                {
                    "title": "The Alzheimer's disease neuroimaging initiative (ADNI): MRI methods",
                    "abstract": "The Alzheimer's Disease Neuroimaging Initiative (ADNI) is a longitudinal multisite observational study of healthy elders, mild cognitive impairment (MCI), and Alzheimer's disease. Magnetic resonance imaging (MRI), (18F)\u2010fluorodeoxyglucose positron emission tomography (FDG PET), urine serum, and cerebrospinal fluid (CSF) biomarkers, as well as clinical/psychometric assessments are acquiredat multiple time points. All data will be cross\u2010linked and made available to the general scientific community. The purpose of this report is to describe the MRI methods employed in ADNI. The ADNI MRI core established specifications thatguided protocol development. A major effort was devoted toevaluating 3D T1\u2010weighted sequences for morphometric analyses. Several options for this sequence were optimized for the relevant manufacturer platforms and then compared in a reduced\u2010scale clinical trial. The protocol selected for the ADNI study includes: back\u2010to\u2010back 3D magnetization prepared rapid gradient echo (MP\u2010RAGE) scans; B1\u2010calibration scans when applicable; and an axial proton density\u2010T2 dual contrast (i.e., echo) fast spin echo/turbo spin echo (FSE/TSE) for pathology detection. ADNI MRI methods seek to maximize scientific utility while minimizing the burden placed on participants. The approach taken in ADNI to standardization across sites and platforms of the MRI protocol, postacquisition corrections, and phantom\u2010based monitoring of all scanners could be used as a model for other multisite trials. J. Magn. Reson. Imaging 2008. \u00a9 2008 Wiley\u2010Liss, Inc."
                },
                {
                    "title": "Open Access Series of Imaging Studies (OASIS): Cross-sectional MRI Data in Young, Middle Aged, Nondemented, and Demented Older Adults",
                    "abstract": "The Open Access Series of Imaging Studies is a series of magnetic resonance imaging data sets that is publicly available for study and analysis. The initial data set consists of a cross-sectional collection of 416 subjects aged 18 to 96 years. One hundred of the included subjects older than 60 years have been clinically diagnosed with very mild to moderate Alzheimer's disease. The subjects are all right-handed and include both men and women. For each subject, three or four individual T1-weighted magnetic resonance imaging scans obtained in single imaging sessions are included. Multiple within-session acquisitions provide extremely high contrast-to-noise ratio, making the data amenable to a wide range of analytic approaches including automated computational analysis. Additionally, a reliability data set is included containing 20 subjects without dementia imaged on a subsequent visit within 90 days of their initial session. Automated calculation of whole-brain volume and estimated total intracranial volume are presented to demonstrate use of the data for measuring differences associated with normal aging and Alzheimer's disease."
                },
                {
                    "title": "Processing Data in Lie Groups : An Algebraic Approach. Application to Non-Linear Registration and Diffusion Tensor MRI. (Traitement de donn\u00e9es dans les groupes de Lie : une approche alg\u00e9brique. Application au recalage non-lin\u00e9aire et \u00e0 l'imagerie du tenseur de diffusion)",
                    "abstract": "Recently, the need for rigorous frameworks for the processing of non-linear data has grown considerably in medical imaging. In this thesis, we propose several general frameworks to process various types of non-linear data, which all belong to Lie groups. To this end, we rely on the algebraic properties of these spaces. Thus, we propose a general processing framework for symmetric and positive-definite matrices, named Log-Euclidean, very simple to use and which has excellent theoretical properties. It is particularly well-adapted to the processing of diffusion tensor MRI. We also propose several frameworks, called polyaffine, to parameterize locally rigid or affine transformations, in a way that guarantees their invertibility. Their use is illustrated in the case of the locally rigid registration of histological slices and of the locally affine 3D registration of MRIs of the human brain. This led us to propose two general frameworks for computing statistics in finite-dimensional Lie groups: first the Log-Euclidean one, which generalizes our work on tensors, and second a framework based on the novel notion of bi-invariant mean, whose properties generalize to Lie groups those of the arithmetic mean. Finally, we generalize our Log-Euclidean framework to diffeomorphic geometrical transformations, which opens the way to a general and consistent framework for statistics in computational anatomy."
                },
                {
                    "title": "Diffeomorphic matching of distributions: a new approach for unlabelled point-sets and sub-manifolds matching",
                    "abstract": "In the paper, we study the problem of optimal matching of two generalized functions (distributions) via a diffeomorphic transformation of the ambient space. In the particular case of discrete distributions (weighted sums of Dirac measures), we provide a new algorithm to compare two arbitrary unlabelled sets of points, and show that it behaves properly in limit of continuous distributions on sub-manifolds. As a consequence, the algorithm may apply to various matching problems, such as curve or surface matching (via a sub-sampling), or mixings of landmark and curve data. As the solution forbids high energy solutions, it is also robust towards addition of noise and the technique can be used for nonlinear projection of datasets. We present 2D and 3D experiments."
                },
                {
                    "title": "Marching cubes: A high resolution 3D surface construction algorithm",
                    "abstract": "We present a new algorithm, called marching cubes, that creates triangle models of constant density surfaces from 3D medical data. Using a divide-and-conquer approach to generate inter-slice connectivity, we create a case table that defines triangle topology. The algorithm processes the 3D medical data in scan-line order and calculates triangle vertices using linear interpolation. We find the gradient of the original data, normalize it, and use it as a basis for shading the models. The detail in images produced from the generated surface models is the result of maintaining the inter-slice connectivity, surface data, and gradient information present in the original 3D data. Results from computed tomography (CT), magnetic resonance (MR), and single-photon emission computed tomography (SPECT) illustrate the quality and functionality of marching cubes. We also discuss improvements that decrease processing time and add solid modeling capabilities."
                },
                {
                    "title": "Parametric Correspondence and Chamfer Matching: Two New Techniques for Image Matching",
                    "abstract": "Parametric correspondence is a technique for matching images to a three dimensional symbolic reference map. An analytic camera model is used to predict the location and appearance of landmarks in the image, generating a projection for an assumed viewpoint. Correspondence is achieved by adjusting the parameters of the camera model until the appearances of the landmarks optimally match a symbolic description extracted from the image. \n \nThe matching of image and map features is performed rapidly by a new technique, called \"chamfer matching\", that compares the shapes of two collections of shape fragments, at a cost proportional to linear dimension, rather than area. These two techniques permit the matching of spatially extensive features on the basis of shape, which reduces the risk of ambiguous matches and the dependence on viewing conditions inherent in conventional image based correlation matching."
                },
                {
                    "title": "Plateau's problem : an invitation to varifold geometry",
                    "abstract": "The phenomena of least area problems Integration of differential forms over rectifiable sets Varifolds Variational problems involving varifolds References Additional references Index."
                },
                {
                    "title": "Theory Of Ordinary Differential Equations",
                    "abstract": "Thank you for reading theory of ordinary differential equations. Maybe you have knowledge that, people have search numerous times for their favorite books like this theory of ordinary differential equations, but end up in malicious downloads. Rather than reading a good book with a cup of coffee in the afternoon, instead they cope with some malicious bugs inside their desktop computer. theory of ordinary differential equations is available in our digital library an online access to it is set as public so you can get it instantly. Our digital library hosts in multiple countries, allowing you to get the most less latency time to download any of our books like this one. Kindly say, the theory of ordinary differential equations is universally compatible with any devices to read."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the evaluation and optimization of mesh deformation models to avoid local minima and enhance the representation of complex manifolds?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of 3D computer vision and graphics, as it directly impacts the accuracy and reliability of shape reconstruction and registration tasks. By developing a more effective metric for mesh deformation, we can enhance the performance of existing models, leading to better applications in medical imaging, virtual reality, and computer-aided design. This research could pave the way for future studies to explore more complex shapes and topologies, ultimately enriching the understanding of geometric transformations in various domains.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of accurately representing and comparing 3D shapes, particularly when dealing with complex manifolds. Naive approaches, such as using Chamfer distance, often lead to local minima and fail to distinguish between good and bad samples, resulting in suboptimal mesh deformations. Additionally, the need to preserve the topology of the mesh while allowing for flexible deformations adds a layer of theoretical and practical complexity. Overcoming these obstacles requires innovative methodologies that can effectively capture the geometric properties of the meshes and provide robust comparisons.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily relied on set-based comparison methods, such as Chamfer distance, which have limitations in distinguishing between mesh deformations effectively. While some advancements, like weighted Chamfer distance, have been made, they do not fully resolve the issues of local minima and poor sample differentiation. The lack of a comprehensive approach that utilizes geometric measure theory and optimal transport has prevented a more effective solution from emerging. Our approach differs by transforming the mesh into a probability measure and employing optimal transport, which allows for a more nuanced comparison of disjoint-support measures.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves transforming the mesh into a probability measure that generalizes the set-based approach, utilizing geometric measure theory to represent the mesh as an oriented varifold. We will employ optimal transport to compare these measures, which is particularly suited for disjoint-support measures. The dataset will consist of various complex 3D meshes, and we will evaluate our approach using metrics that assess the quality of mesh deformation and the accuracy of shape representation. The"
            }
        },
        "author_data": {
            "2ce83217-c461-449f-a27b-102f3db0a7b9": {
                "pk": "2ce83217-c461-449f-a27b-102f3db0a7b9",
                "name": "Tung Le",
                "collaborators": [
                    "A. Chernov"
                ],
                "domain": [
                    "Numerical Analysis",
                    "Eigenvalue Problems",
                    "Partial Differential Equations",
                    "Parameter Sensitivity"
                ],
                "publications": [
                    {
                        "title": "Analytic and Gevrey class regularity for parametric elliptic eigenvalue problems",
                        "abstract": "We investigate a class of parametric elliptic eigenvalue problems with homogeneous essential boundary conditions where the coefficients (and hence the solution $u$) may depend on a parameter $y$. For the efficient approximate evaluation of parameter sensitivities of the first eigenpairs on the entire parameter space we propose and analyse Gevrey class and analytic regularity of the solution with respect to the parameters. This is made possible by a novel proof technique which we introduce and demonstrate in this paper. Our regularity result has immediate implications for convergence of various numerical schemes for parametric elliptic eigenvalue problems, in particular, for elliptic eigenvalue problems with infinitely many parameters arising from elliptic differential operators with random coefficients."
                    }
                ]
            },
            "29f62b34-4e13-40ac-9a68-ea57a285a08a": {
                "pk": "29f62b34-4e13-40ac-9a68-ea57a285a08a",
                "name": "Khai Nguyen",
                "collaborators": [
                    "Nhat Ho",
                    "Huy Nguyen",
                    "T. Nguyen",
                    "Nicola Bariletto",
                    "Tongzheng Ren",
                    "Tam Nguyen",
                    "S. Osher",
                    "Manh Luong",
                    "Reza Haf",
                    "D.Q. Phung",
                    "Lizhen Qu",
                    "Shujian Zhang",
                    "Tam Le",
                    "Tung Le",
                    "Shanlin Sun",
                    "Xiaohui Xie",
                    "Hai Nguyen",
                    "TrungTin Nguyen",
                    "Dang Nguyen",
                    "N. Ho",
                    "Dat Do",
                    "Hai Do",
                    "Vishwanath Saragadam",
                    "Minh Pham",
                    "Duy Khuong Nguyen",
                    "Dung D. Le",
                    "Trang Nguyen",
                    "Xing Han",
                    "J. Ghosh"
                ],
                "domain": [
                    "Optimal Transport",
                    "Machine Learning",
                    "Deep Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Revisiting Deep Audio-Text Retrieval Through the Lens of Transportation",
                        "abstract": "The Learning-to-match (LTM) framework proves to be an effective inverse optimal transport approach for learning the underlying ground metric between two sources of data, facilitating subsequent matching. However, the conventional LTM framework faces scalability challenges, necessitating the use of the entire dataset each time the parameters of the ground metric are updated. In adapting LTM to the deep learning context, we introduce the mini-batch Learning-to-match (m-LTM) framework for audio-text retrieval problems. This framework leverages mini-batch subsampling and Mahalanobis-enhanced family of ground metrics. Moreover, to cope with misaligned training data in practice, we propose a variant using partial optimal transport to mitigate the harm of misaligned data pairs in training data. We conduct extensive experiments on audio-text matching problems using three datasets: AudioCaps, Clotho, and ESC-50. Results demonstrate that our proposed method is capable of learning rich and expressive joint embedding space, which achieves SOTA performance. Beyond this, the proposed m-LTM framework is able to close the modality gap across audio and text embedding, which surpasses both triplet and contrastive loss in the zero-shot sound event detection task on the ESC-50 dataset. Notably, our strategy of employing partial optimal transport with m-LTM demonstrates greater noise tolerance than contrastive loss, especially under varying noise ratios in training data on the AudioCaps dataset. Our code is available at https://github.com/v-manhlt3/m-LTM-Audio-Text-Retrieval"
                    },
                    {
                        "title": "On Parameter Estimation in Deviated Gaussian Mixture of Experts",
                        "abstract": "We consider the parameter estimation problem in the deviated Gaussian mixture of experts in which the data are generated from $(1 - \\lambda^{\\ast}) g_0(Y| X)+ \\lambda^{\\ast} \\sum_{i = 1}^{k_{\\ast}} p_{i}^{\\ast} f(Y|(a_{i}^{\\ast})^{\\top}X+b_i^{\\ast},\\sigma_{i}^{\\ast})$, where $X, Y$ are respectively a covariate vector and a response variable, $g_{0}(Y|X)$ is a known function, $\\lambda^{\\ast} \\in [0, 1]$ is true but unknown mixing proportion, and $(p_{i}^{\\ast}, a_{i}^{\\ast}, b_{i}^{\\ast}, \\sigma_{i}^{\\ast})$ for $1 \\leq i \\leq k^{\\ast}$ are unknown parameters of the Gaussian mixture of experts. This problem arises from the goodness-of-fit test when we would like to test whether the data are generated from $g_{0}(Y|X)$ (null hypothesis) or they are generated from the whole mixture (alternative hypothesis). Based on the algebraic structure of the expert functions and the distinguishability between $g_0$ and the mixture part, we construct novel Voronoi-based loss functions to capture the convergence rates of maximum likelihood estimation (MLE) for our models. We further demonstrate that our proposed loss functions characterize the local convergence rates of parameter estimation more accurately than the generalized Wasserstein, a loss function being commonly used for estimating parameters in the Gaussian mixture of experts."
                    },
                    {
                        "title": "Sliced Wasserstein with Random-Path Projecting Directions",
                        "abstract": "Slicing distribution selection has been used as an effective technique to improve the performance of parameter estimators based on minimizing sliced Wasserstein distance in applications. Previous works either utilize expensive optimization to select the slicing distribution or use slicing distributions that require expensive sampling methods. In this work, we propose an optimization-free slicing distribution that provides a fast sampling for the Monte Carlo estimation of expectation. In particular, we introduce the random-path projecting direction (RPD) which is constructed by leveraging the normalized difference between two random vectors following the two input measures. From the RPD, we derive the random-path slicing distribution (RPSD) and two variants of sliced Wasserstein, i.e., the Random-Path Projection Sliced Wasserstein (RPSW) and the Importance Weighted Random-Path Projection Sliced Wasserstein (IWRPSW). We then discuss the topological, statistical, and computational properties of RPSW and IWRPSW. Finally, we showcase the favorable performance of RPSW and IWRPSW in gradient flow and the training of denoising diffusion generative models on images."
                    },
                    {
                        "title": "Hierarchical Hybrid Sliced Wasserstein: A Scalable Metric for Heterogeneous Joint Distributions",
                        "abstract": "Sliced Wasserstein (SW) and Generalized Sliced Wasserstein (GSW) have been widely used in applications due to their computational and statistical scalability. However, the SW and the GSW are only defined between distributions supported on a homogeneous domain. This limitation prevents their usage in applications with heterogeneous joint distributions with marginal distributions supported on multiple different domains. Using SW and GSW directly on the joint domains cannot make a meaningful comparison since their homogeneous slicing operator i.e., Radon Transform (RT) and Generalized Radon Transform (GRT) are not expressive enough to capture the structure of the joint supports set. To address the issue, we propose two new slicing operators i.e., Partial Generalized Radon Transform (PGRT) and Hierarchical Hybrid Radon Transform (HHRT). In greater detail, PGRT is the generalization of Partial Radon Transform (PRT), which transforms a subset of function arguments non-linearly while HHRT is the composition of PRT and multiple domain-specific PGRT on marginal domain arguments. By using HHRT, we extend the SW into Hierarchical Hybrid Sliced Wasserstein (H2SW) distance which is designed specifically for comparing heterogeneous joint distributions. We then discuss the topological, statistical, and computational properties of H2SW. Finally, we demonstrate the favorable performance of H2SW in 3D mesh deformation, deep 3D mesh autoencoders, and datasets comparison."
                    },
                    {
                        "title": "Integrating Efficient Optimal Transport and Functional Maps for Unsupervised Shape Correspondence Learning",
                        "abstract": "In the realm of computer vision and graphics, accurately establishing correspondences between geometric 3D shapes is pivotal for applications like object tracking, registration, texture transfer, and statistical shape analysis. Moving beyond traditional hand-crafted and data-driven feature learning methods, we incorporate spectral methods with deep learning, focusing on functional maps (FMs) and optimal transport (OT). Traditional OT-based approaches, often reliant on entropy regularization OT in learning-based framework, face computational challenges due to their quadratic cost. Our key contribution is to employ the sliced Wasserstein distance (SWD)for OT, which is a valid fast optimal transport metric in an unsupervised shape matching framework. This unsupervised framework integrates functional map regularizers with a novel OT-based loss derived from SWD, enhancing feature alignment between shapes treated as discrete probability measures. We also introduce an adaptive refinement process utilizing entropy regularized OT, further refining feature alignments for accurate point-to-point correspondences. Our method demonstrates superior performance in non-rigid shape matching, including near-isometric and non-isometric scenarios, and excels in downstream tasks like segmentation transfer. The empirical results on diverse datasets highlight our framework's effectiveness and generalization capabilities, setting new standards in non-rigid shape matching with efficient OT metrics and an adaptive refinement module. Code is available at11https://github.com/Tungthanhlee/EOT-Correspondence."
                    },
                    {
                        "title": "Marginal Fairness Sliced Wasserstein Barycenter",
                        "abstract": "The sliced Wasserstein barycenter (SWB) is a widely acknowledged method for efficiently generalizing the averaging operation within probability measure spaces. However, achieving marginal fairness SWB, ensuring approximately equal distances from the barycenter to marginals, remains unexplored. The uniform weighted SWB is not necessarily the optimal choice to obtain the desired marginal fairness barycenter due to the heterogeneous structure of marginals and the non-optimality of the optimization. As the first attempt to tackle the problem, we define the marginal fairness sliced Wasserstein barycenter (MFSWB) as a constrained SWB problem. Due to the computational disadvantages of the formal definition, we propose two hyperparameter-free and computationally tractable surrogate MFSWB problems that implicitly minimize the distances to marginals and encourage marginal fairness at the same time. To further improve the efficiency, we perform slicing distribution selection and obtain the third surrogate definition by introducing a new slicing distribution that focuses more on marginally unfair projecting directions. We discuss the relationship of the three proposed problems and their relationship to sliced multi-marginal Wasserstein distance. Finally, we conduct experiments on finding 3D point-clouds averaging, color harmonization, and training of sliced Wasserstein autoencoder with class-fairness representation to show the favorable performance of the proposed surrogate MFSWB problems."
                    },
                    {
                        "title": "Borrowing Strength in Distributionally Robust Optimization via Hierarchical Dirichlet Processes",
                        "abstract": "This paper presents a novel optimization framework to address key challenges presented by modern machine learning applications: High dimensionality, distributional uncertainty, and data heterogeneity. Our approach unifies regularized estimation, distributionally robust optimization (DRO), and hierarchical Bayesian modeling in a single data-driven criterion. By employing a hierarchical Dirichlet process (HDP) prior, the method effectively handles multi-source data, achieving regularization, distributional robustness, and borrowing strength across diverse yet related data-generating processes. We demonstrate the method's advantages by establishing theoretical performance guarantees and tractable Monte Carlo approximations based on Dirichlet process (DP) theory. Numerical experiments validate the framework's efficacy in improving and stabilizing both prediction and parameter estimation accuracy, showcasing its potential for application in complex data environments."
                    },
                    {
                        "title": "Markovian Sliced Wasserstein Distances: Beyond Independent Projections",
                        "abstract": "Sliced Wasserstein (SW) distance suffers from redundant projections due to independent uniform random projecting directions. To partially overcome the issue, max K sliced Wasserstein (Max-K-SW) distance ($K\\geq 1$), seeks the best discriminative orthogonal projecting directions. Despite being able to reduce the number of projections, the metricity of Max-K-SW cannot be guaranteed in practice due to the non-optimality of the optimization. Moreover, the orthogonality constraint is also computationally expensive and might not be effective. To address the problem, we introduce a new family of SW distances, named Markovian sliced Wasserstein (MSW) distance, which imposes a first-order Markov structure on projecting directions. We discuss various members of MSW by specifying the Markov structure including the prior distribution, the transition distribution, and the burning and thinning technique. Moreover, we investigate the theoretical properties of MSW including topological properties (metricity, weak convergence, and connection to other distances), statistical properties (sample complexity, and Monte Carlo estimation error), and computational properties (computational complexity and memory complexity). Finally, we compare MSW distances with previous SW variants in various applications such as gradient flows, color transfer, and deep generative modeling to demonstrate the favorable performance of MSW."
                    },
                    {
                        "title": "Energy-Based Sliced Wasserstein Distance",
                        "abstract": "The sliced Wasserstein (SW) distance has been widely recognized as a statistically effective and computationally efficient metric between two probability measures. A key component of the SW distance is the slicing distribution. There are two existing approaches for choosing this distribution. The first approach is using a fixed prior distribution. The second approach is optimizing for the best distribution which belongs to a parametric family of distributions and can maximize the expected distance. However, both approaches have their limitations. A fixed prior distribution is non-informative in terms of highlighting projecting directions that can discriminate two general probability measures. Doing optimization for the best distribution is often expensive and unstable. Moreover, designing the parametric family of the candidate distribution could be easily misspecified. To address the issues, we propose to design the slicing distribution as an energy-based distribution that is parameter-free and has the density proportional to an energy function of the projected one-dimensional Wasserstein distance. We then derive a novel sliced Wasserstein metric, energy-based sliced Waserstein (EBSW) distance, and investigate its topological, statistical, and computational properties via importance sampling, sampling importance resampling, and Markov Chain methods. Finally, we conduct experiments on point-cloud gradient flow, color transfer, and point-cloud reconstruction to show the favorable performance of the EBSW."
                    },
                    {
                        "title": "Towards Convergence Rates for Parameter Estimation in Gaussian-gated Mixture of Experts",
                        "abstract": "Originally introduced as a neural network for ensemble learning, mixture of experts (MoE) has recently become a fundamental building block of highly successful modern deep neural networks for heterogeneous data analysis in several applications of machine learning and statistics. Despite its popularity in practice, a satisfactory level of theoretical understanding of the MoE model is far from complete. To shed new light on this problem, we provide a convergence analysis for maximum likelihood estimation (MLE) in the Gaussian-gated MoE model. The main challenge of that analysis comes from the inclusion of covariates in the Gaussian gating functions and expert networks, which leads to their intrinsic interaction via some partial differential equations with respect to their parameters. We tackle these issues by designing novel Voronoi loss functions among parameters to accurately capture the heterogeneity of parameter estimation rates. Our findings reveal that the MLE has distinct behaviors under two complement settings of location parameters of the Gaussian gating functions, namely when all these parameters are non-zero versus when at least one among them vanishes. Notably, these behaviors can be characterized by the solvability of two different systems of polynomial equations. Finally, we conduct a simulation study to empirically verify our theoretical results."
                    },
                    {
                        "title": "Quasi-Monte Carlo for 3D Sliced Wasserstein",
                        "abstract": "Monte Carlo (MC) integration has been employed as the standard approximation method for the Sliced Wasserstein (SW) distance, whose analytical expression involves an intractable expectation. However, MC integration is not optimal in terms of absolute approximation error. To provide a better class of empirical SW, we propose quasi-sliced Wasserstein (QSW) approximations that rely on Quasi-Monte Carlo (QMC) methods. For a comprehensive investigation of QMC for SW, we focus on the 3D setting, specifically computing the SW between probability measures in three dimensions. In greater detail, we empirically evaluate various methods to construct QMC point sets on the 3D unit-hypersphere, including the Gaussian-based and equal area mappings, generalized spiral points, and optimizing discrepancy energies. Furthermore, to obtain an unbiased estimator for stochastic optimization, we extend QSW to Randomized Quasi-Sliced Wasserstein (RQSW) by introducing randomness in the discussed point sets. Theoretically, we prove the asymptotic convergence of QSW and the unbiasedness of RQSW. Finally, we conduct experiments on various 3D tasks, such as point-cloud comparison, point-cloud interpolation, image style transfer, and training deep point-cloud autoencoders, to demonstrate the favorable performance of the proposed QSW and RQSW variants."
                    },
                    {
                        "title": "Sliced Wasserstein Estimation with Control Variates",
                        "abstract": "The sliced Wasserstein (SW) distances between two probability measures are defined as the expectation of the Wasserstein distance between two one-dimensional projections of the two measures. The randomness comes from a projecting direction that is used to project the two input measures to one dimension. Due to the intractability of the expectation, Monte Carlo integration is performed to estimate the value of the SW distance. Despite having various variants, there has been no prior work that improves the Monte Carlo estimation scheme for the SW distance in terms of controlling its variance. To bridge the literature on variance reduction and the literature on the SW distance, we propose computationally efficient control variates to reduce the variance of the empirical estimation of the SW distance. The key idea is to first find Gaussian approximations of projected one-dimensional measures, then we utilize the closed-form of the Wasserstein-2 distance between two Gaussian distributions to design the control variates. In particular, we propose using a lower bound and an upper bound of the Wasserstein-2 distance between two fitted Gaussians as two computationally efficient control variates. We empirically show that the proposed control variate estimators can help to reduce the variance considerably when comparing measures over images and point-clouds. Finally, we demonstrate the favorable performance of the proposed control variate estimators in gradient flows to interpolate between two point-clouds and in deep generative modeling on standard image datasets, such as CIFAR10 and CelebA."
                    },
                    {
                        "title": "Self-Attention Amortized Distributional Projection Optimization for Sliced Wasserstein Point-Cloud Reconstruction",
                        "abstract": "Max sliced Wasserstein (Max-SW) distance has been widely known as a solution for less discriminative projections of sliced Wasserstein (SW) distance. In applications that have various independent pairs of probability measures, amortized projection optimization is utilized to predict the ``max\"projecting directions given two input measures instead of using projected gradient ascent multiple times. Despite being efficient, Max-SW and its amortized version cannot guarantee metricity property due to the sub-optimality of the projected gradient ascent and the amortization gap. Therefore, we propose to replace Max-SW with distributional sliced Wasserstein distance with von Mises-Fisher (vMF) projecting distribution (v-DSW). Since v-DSW is a metric with any non-degenerate vMF distribution, its amortized version can guarantee the metricity when performing amortization. Furthermore, current amortized models are not permutation invariant and symmetric. To address the issue, we design amortized models based on self-attention architecture. In particular, we adopt efficient self-attention architectures to make the computation linear in the number of supports. With the two improvements, we derive self-attention amortized distributional projection optimization and show its appealing performance in point-cloud reconstruction and its downstream applications."
                    },
                    {
                        "title": "Minimax Optimal Rate for Parameter Estimation in Multivariate Deviated Models",
                        "abstract": "We study the maximum likelihood estimation (MLE) in the multivariate deviated model where the data are generated from the density function $(1-\\lambda^{\\ast})h_{0}(x)+\\lambda^{\\ast}f(x|\\mu^{\\ast}, \\Sigma^{\\ast})$ in which $h_{0}$ is a known function, $\\lambda^{\\ast} \\in [0,1]$ and $(\\mu^{\\ast}, \\Sigma^{\\ast})$ are unknown parameters to estimate. The main challenges in deriving the convergence rate of the MLE mainly come from two issues: (1) The interaction between the function $h_{0}$ and the density function $f$; (2) The deviated proportion $\\lambda^{\\ast}$ can go to the extreme points of $[0,1]$ as the sample size tends to infinity. To address these challenges, we develop the \\emph{distinguishability condition} to capture the linear independent relation between the function $h_{0}$ and the density function $f$. We then provide comprehensive convergence rates of the MLE via the vanishing rate of $\\lambda^{\\ast}$ to zero as well as the distinguishability of two functions $h_{0}$ and $f$."
                    },
                    {
                        "title": "Improving Transformer with an Admixture of Attention Heads",
                        "abstract": "Transformers with multi-head self-attention have achieved remarkable success in sequence modeling and beyond. However, they suffer from high computational and memory complexities for computing the attention matrix at each head. Recently, it has been shown that those attention matrices lie on a low-dimensional manifold and, thus, are redundant. We propose the Transformer with a Finite Admixture of Shared Heads (FiSHformers), a novel class of efficient and flexible transformers that allow the sharing of attention matrices between attention heads. At the core of FiSHformer is a novel finite admixture model of shared heads (FiSH) that samples attention matrices from a set of global attention matrices. The number of global attention matrices is much smaller than the number of local attention matrices generated. FiSHformers directly learn these global attention matrices rather than the local ones as in other transformers, thus significantly improving the computational and memory efficiency of the model. We empirically verify the advantages of the FiSHformer over the baseline transformers in a wide range of practical applications including language modeling, machine translation, and image classification. On the WikiText-103, IWSLT\u201914 De-En and WMT\u201914"
                    },
                    {
                        "title": "Transformer with Fourier Integral Attentions",
                        "abstract": "Multi-head attention empowers the recent success of transformers, the state-of-the-art models that have achieved remarkable success in sequence modeling and beyond. These attention mechanisms compute the pairwise dot products between the queries and keys, which results from the use of unnormalized Gaussian kernels with the assumption that the queries follow a mixture of Gaussian distribution. There is no guarantee that this assumption is valid in practice. In response, we first interpret attention in transformers as a nonparametric kernel regression. We then propose the FourierFormer, a new class of transformers in which the dot-product kernels are replaced by the novel generalized Fourier integral kernels. Different from the dot-product kernels, where we need to choose a good covariance matrix to capture the dependency of the features of data, the generalized Fourier integral kernels can automatically capture such dependency and remove the need to tune the covariance matrix. We theoretically prove that our proposed Fourier integral kernels can efficiently approximate any key and query distributions. Compared to the conventional transformers with dot-product attention, FourierFormers attain better accuracy and reduce the redundancy between attention heads. We empirically corroborate the advantages of FourierFormers over the baseline transformers in a variety of practical applications including language modeling and image classification."
                    },
                    {
                        "title": "Fast Approximation of the Generalized Sliced-Wasserstein Distance",
                        "abstract": "Generalized sliced-Wasserstein distance is a variant of slicedWasserstein distance that exploits the power of non-linear projection through a given defining function to better capture the complex structures of probability distributions. Similar to the sliced-Wasserstein distance, generalized slicedWasserstein is defined as an expectation over random projections which can be approximated by the Monte Carlo method. However, the complexity of that approximation can be expensive in high-dimensional settings. To that end, we propose to form deterministic and fast approximations of the generalized sliced-Wasserstein distance by using the concentration of random projections when the defining functions are polynomial function and neural network type function. Our approximations hinge upon an important result that one-dimensional projections of a high-dimensional random vector are approximately Gaussian."
                    },
                    {
                        "title": "Revisiting Sliced Wasserstein on Images: From Vectorization to Convolution",
                        "abstract": "The conventional sliced Wasserstein is defined between two probability measures that have realizations as vectors. When comparing two probability measures over images, practitioners first need to vectorize images and then project them to one-dimensional space by using matrix multiplication between the sample matrix and the projection matrix. After that, the sliced Wasserstein is evaluated by averaging the two corresponding one-dimensional projected probability measures. However, this approach has two limitations. The first limitation is that the spatial structure of images is not captured efficiently by the vectorization step; therefore, the later slicing process becomes harder to gather the discrepancy information. The second limitation is memory inefficiency since each slicing direction is a vector that has the same dimension as the images. To address these limitations, we propose novel slicing methods for sliced Wasserstein between probability measures over images that are based on the convolution operators. We derive convolution sliced Wasserstein (CSW) and its variants via incorporating stride, dilation, and non-linear activation function into the convolution operators. We investigate the metricity of CSW as well as its sample complexity, its computational complexity, and its connection to conventional sliced Wasserstein distances. Finally, we demonstrate the favorable performance of CSW over the conventional sliced Wasserstein in comparing probability measures over images and in training deep generative modeling on images."
                    },
                    {
                        "title": "Robustify Transformers with Robust Kernel Density Estimation",
                        "abstract": "Recent advances in Transformer architecture have empowered its empirical success in various tasks across di\ufb00erent domains. However, existing works mainly focus on improving the standard accuracy and computational cost, without considering the robustness of contaminated samples. Existing work [40] has shown that the self-attention mechanism, which is the center of the Transformer architecture, can be viewed as a non-parametric estimator based on the well-known kernel density estimation (KDE). This motivates us to leverage the robust kernel density estimation (RKDE) in the self-attention mechanism, to alleviate the issue of the contamination of data by down-weighting the weight of bad samples in the estimation process. The modi\ufb01ed self-attention mechanism can be incorporated into di\ufb00erent Transformer variants. Empirical results on language modeling and image classi\ufb01cation tasks demonstrate the e\ufb00ectiveness of this approach."
                    }
                ]
            },
            "9d17e5db-f75c-4ea8-88d8-5ef981bb1602": {
                "pk": "9d17e5db-f75c-4ea8-88d8-5ef981bb1602",
                "name": "Shanlin Sun",
                "collaborators": [
                    "Xiaohui Xie",
                    "Kun Han",
                    "Xiangyi Yan",
                    "Haoyu Ma",
                    "Deying Kong",
                    "Chenyu You",
                    "Hao Tang",
                    "Junayed Naushad",
                    "Thanh-Tung Le",
                    "Yifeng Xiong",
                    "J. Duncan",
                    "Xingwei Liu",
                    "Tung Le",
                    "Khai Nguyen",
                    "Nhat Ho",
                    "Bingbing Zhuang",
                    "Ziyu Jiang",
                    "Buyu Liu",
                    "M. Chandraker",
                    "Gabriel De Araujo",
                    "Tong Zhang",
                    "Pooya Khosravi",
                    "Linguang Zhang",
                    "Liangjian Chen",
                    "Yingxin Cao"
                ],
                "domain": [
                    "Computer Vision",
                    "Medical Imaging",
                    "Deep Learning",
                    "Shape Analysis"
                ],
                "publications": [
                    {
                        "title": "Integrating Efficient Optimal Transport and Functional Maps for Unsupervised Shape Correspondence Learning",
                        "abstract": "In the realm of computer vision and graphics, accurately establishing correspondences between geometric 3D shapes is pivotal for applications like object tracking, registration, texture transfer, and statistical shape analysis. Moving beyond traditional hand-crafted and data-driven feature learning methods, we incorporate spectral methods with deep learning, focusing on functional maps (FMs) and optimal transport (OT). Traditional OT-based approaches, often reliant on entropy regularization OT in learning-based framework, face computational challenges due to their quadratic cost. Our key contribution is to employ the sliced Wasserstein distance (SWD)for OT, which is a valid fast optimal transport metric in an unsupervised shape matching framework. This unsupervised framework integrates functional map regularizers with a novel OT-based loss derived from SWD, enhancing feature alignment between shapes treated as discrete probability measures. We also introduce an adaptive refinement process utilizing entropy regularized OT, further refining feature alignments for accurate point-to-point correspondences. Our method demonstrates superior performance in non-rigid shape matching, including near-isometric and non-isometric scenarios, and excels in downstream tasks like segmentation transfer. The empirical results on diverse datasets highlight our framework's effectiveness and generalization capabilities, setting new standards in non-rigid shape matching with efficient OT metrics and an adaptive refinement module. Code is available at11https://github.com/Tungthanhlee/EOT-Correspondence."
                    },
                    {
                        "title": "LidaRF: Delving into Lidar for Neural Radiance Field on Street Scenes",
                        "abstract": "Photorealistic simulation plays a crucial role in applications such as autonomous driving, where advances in neural radiance fields (NeRFs) may allow better scalability through the automatic creation of digital 3D assets. However, reconstruction quality suffers on street scenes due to largely collinear camera motions and sparser samplings at higher speeds. On the other hand, the application often demands rendering from camera views that deviate from the inputs to accurately simulate behaviors like lane changes. In this paper, we propose several insights that allow a better utilization of Lidar data to improve NeRF quality on street scenes. First, our framework learns a geometric scene representation from Lidar, which are fused with the implicit grid-based representation for radiance decoding, thereby supplying stronger geometric information offered by explicit point cloud. Sec-ond, we put forth a robust occlusion-aware depth supervision scheme, which allows utilizing densified Lidar points by accumulation. Third, we generate augmented training views from Lidar points for further improvement. Our insights translate to largely improved novel view synthesis under real driving scenes."
                    },
                    {
                        "title": "AFTer-SAM: Adapting SAM with Axial Fusion Transformer for Medical Imaging Segmentation",
                        "abstract": "The Segmentation Anything Model (SAM) has demonstrated effectiveness in various segmentation tasks. However, its application to 3D medical data has posed challenges due to its inherent design for both 2D and natural images. While there have been attempts to apply SAM to medical images on a slice-by-slice basis, the outcomes have been less than optimal. In this study, we introduce AFTer-SAM, an adaptation of SAM designed for volumetric medical image segmentation. By incorporating an Axial Fusion Transformer, AFTer-SAM is capable of capturing both intra-slice details and inter-slice contextual information, essential for accurate medical image segmentation. Given the potential computational challenges of training this enhanced model, we utilize Low-Rank Adaptation (LoRA) to efficiently finetune the weights of the Axial Fusion Transformer. This ensures a streamlined training process without compromising on performance. Our results indicate that AFTer-SAM offers significant improvements in volumetric medical image segmentation, suggesting a promising direction for the application of large pre-trained models in medical imaging."
                    },
                    {
                        "title": "Adaptive Image Registration: A Hybrid Approach Integrating Deep Learning and Optimization Functions for Enhanced Precision",
                        "abstract": "Image registration has traditionally been done using two distinct approaches: learning based methods, relying on robust deep neural networks, and optimization-based methods, applying complex mathematical transformations to warp images accordingly. Of course, both paradigms offer advantages and disadvantages, and, in this work, we seek to combine their respective strengths into a single streamlined framework, using the outputs of the learning based method as initial parameters for optimization while prioritizing computational power for the image pairs that offer the greatest loss. Our investigations showed improvements of up to 1.6% in test data, while maintaining the same inference time, and a substantial 1.0% points performance gain in deformation field smoothness."
                    },
                    {
                        "title": "Hybrid Neural Diffeomorphic Flow for Shape Representation and Generation via Triplane",
                        "abstract": "Deep Implicit Functions (DIFs) have gained popularity in 3D computer vision due to their compactness and continuous representation capabilities. However, addressing dense correspondences and semantic relationships across DIF-encoded shapes remains a critical challenge, limiting their applications in texture transfer and shape analysis. Moreover, recent endeavors in 3D shape generation using DIFs often neglect correspondence and topology preservation. This paper presents HNDF (Hybrid Neural Diffeomorphic Flow), a method that implicitly learns the underlying representation and decomposes intricate dense correspondences into explicitly axis-aligned triplane features. To avoid suboptimal representations trapped in local minima, we propose hybrid supervision that captures both local and global correspondences. Unlike conventional approaches that directly generate new 3D shapes, we further explore the idea of shape generation with deformed template shape via diffeomorphic flows, where the deformation is encoded by the generated triplane features. Leveraging a pre-existing 2D diffusion model, we produce high-quality and diverse 3D diffeomorphic flows through generated triplanes features, ensuring topological consistency with the template shape. Extensive experiments on medical image organ segmentation datasets evaluate the effectiveness of HNDF in 3D shape representation and generation."
                    },
                    {
                        "title": "Hybrid-CSR: Coupling Explicit and Implicit Shape Representation for Cortical Surface Reconstruction",
                        "abstract": "We present Hybrid-CSR, a geometric deep-learning model that combines explicit and implicit shape representations for cortical surface reconstruction. Specifically, Hybrid-CSR begins with explicit deformations of template meshes to obtain coarsely reconstructed cortical surfaces, based on which the oriented point clouds are estimated for the subsequent differentiable poisson surface reconstruction. By doing so, our method unifies explicit (oriented point clouds) and implicit (indicator function) cortical surface reconstruction. Compared to explicit representation-based methods, our hybrid approach is more friendly to capture detailed structures, and when compared with implicit representation-based methods, our method can be topology aware because of end-to-end training with a mesh-based deformation module. In order to address topology defects, we propose a new topology correction pipeline that relies on optimization-based diffeomorphic surface registration. Experimental results on three brain datasets show that our approach surpasses existing implicit and explicit cortical surface reconstruction methods in numeric metrics in terms of accuracy, regularity, and consistency."
                    },
                    {
                        "title": "Light Field Diffusion for Single-View Novel View Synthesis",
                        "abstract": "Single-view novel view synthesis (NVS), the task of generating images from new viewpoints based on a single reference image, is important but challenging in computer vision. Recent advancements in NVS have leveraged Denoising Diffusion Probabilistic Models (DDPMs) for their exceptional ability to produce high-fidelity images. However, current diffusion-based methods typically utilize camera pose matrices to globally and implicitly enforce 3D constraints, which can lead to inconsistencies in images generated from varying viewpoints, particularly in regions with complex textures and structures. To address these limitations, we present Light Field Diffusion (LFD), a novel conditional diffusion-based approach that transcends the conventional reliance on camera pose matrices. Starting from the camera pose matrices, LFD transforms them into light field encoding, with the same shape as the reference image, to describe the direction of each ray. By integrating light field encoding with the reference image, our method imposes local pixel-wise constraints within the diffusion process, fostering enhanced view consistency. Our approach not only involves training image LFD on the ShapeNet Car dataset but also includes fine-tuning a pre-trained latent diffusion model on the Objaverse dataset. This enables our latent LFD model to exhibit remarkable zero-shot generalization capabilities across out-of-distribution datasets like RTMV as well as in-the-wild images. Experiments demonstrate that LFD not only produces high-fidelity images but also achieves superior 3D consistency in complex regions, outperforming existing novel view synthesis methods."
                    },
                    {
                        "title": "CVTHead: One-shot Controllable Head Avatar with Vertex-feature Transformer",
                        "abstract": "Reconstructing personalized animatable head avatars has significant implications in the fields of AR/VR. Existing methods for achieving explicit face control of 3D Morphable Models (3DMM) typically rely on multi-view images or videos of a single subject, making the reconstruction process complex. Additionally, the traditional rendering pipeline is time-consuming, limiting real-time animation possibilities. In this paper, we introduce CVTHead, a novel approach that generates controllable neural head avatars from a single reference image using point-based neural rendering. CVT-Head considers the sparse vertices of mesh as the point set and employs the proposed Vertex-feature Transformer to learn local feature descriptors for each vertex. This enables the modeling of long-range dependencies among all the vertices. Experimental results on the VoxCeleb dataset demonstrate that CVTHead achieves comparable performance to state-of-the-art graphics-based methods. Moreover, it enables efficient rendering of novel human heads with various expressions, head poses, and camera views. These attributes can be explicitly controlled using the coefficients of 3DMMs, facilitating versatile and realistic animation in real-time scenarios. Codes and pre-trained model can be found at https://github.com/HowieMa/CVTHead."
                    },
                    {
                        "title": "Localized Region Contrast for Enhancing Self-Supervised Learning in Medical Image Segmentation",
                        "abstract": "Recent advancements in self-supervised learning have demonstrated that effective visual representations can be learned from unlabeled images. This has led to increased interest in applying self-supervised learning to the medical domain, where unlabeled images are abundant and labeled images are difficult to obtain. However, most self-supervised learning approaches are modeled as image level discriminative or generative proxy tasks, which may not capture the finer level representations necessary for dense prediction tasks like multi-organ segmentation. In this paper, we propose a novel contrastive learning framework that integrates Localized Region Contrast (LRC) to enhance existing self-supervised pre-training methods for medical image segmentation. Our approach involves identifying Super-pixels by Felzenszwalb's algorithm and performing local contrastive learning using a novel contrastive sampling loss. Through extensive experiments on three multi-organ segmentation datasets, we demonstrate that integrating LRC to an existing self-supervised method in a limited annotation setting significantly improves segmentation performance. Moreover, we show that LRC can also be applied to fully-supervised pre-training methods to further boost performance."
                    },
                    {
                        "title": "MedGen3D: A Deep Generative Framework for Paired 3D Image and Mask Generation",
                        "abstract": "Acquiring and annotating sufficient labeled data is crucial in developing accurate and robust learning-based models, but obtaining such data can be challenging in many medical image segmentation tasks. One promising solution is to synthesize realistic data with ground-truth mask annotations. However, no prior studies have explored generating complete 3D volumetric images with masks. In this paper, we present MedGen3D, a deep generative framework that can generate paired 3D medical images and masks. First, we represent the 3D medical data as 2D sequences and propose the Multi-Condition Diffusion Probabilistic Model (MC-DPM) to generate multi-label mask sequences adhering to anatomical geometry. Then, we use an image sequence generator and semantic diffusion refiner conditioned on the generated mask sequences to produce realistic 3D medical images that align with the generated masks. Our proposed framework guarantees accurate alignment between synthetic images and segmentation maps. Experiments on 3D thoracic CT and brain MRI datasets show that our synthetic data is both diverse and faithful to the original data, and demonstrate the benefits for downstream segmentation tasks. We anticipate that MedGen3D's ability to synthesize paired 3D medical images and masks will prove valuable in training deep learning models for medical imaging tasks."
                    },
                    {
                        "title": "Representation Recovering for Self-Supervised Pre-training on Medical Images",
                        "abstract": "Advances in self-supervised learning have drawn attention to developing techniques to extract effective visual representations from unlabeled images. Contrastive learning (CL) trains a model to extract consistent features by generating different views. Recent success of Masked Autoencoders (MAE) highlights the benefit of generative modeling in self-supervised learning. The generative approaches encode the input into a compact embedding and empower the model\u2019s ability of recovering the original input. However, in our experiments, we found vanilla MAE mainly recovers coarse high level semantic information and is inadequate in recovering detailed low level information. We show that in dense downstream prediction tasks like multi-organ segmentation, directly applying MAE is not ideal. Here, we propose RepRec, a hybrid visual representation learning framework for self-supervised pre-training on large-scale unlabelled medical datasets, which takes advantage of both contrastive and generative modeling. To solve the aforementioned dilemma that MAE encounters, a convolutional encoder is pre-trained to provide low-level feature information, in a contrastive way; and a transformer encoder is pre-trained to produce high level semantic dependency, in a generative way \u2013 by recovering masked representations from the convolutional encoder. Extensive experiments on three multi-organ segmentation datasets demonstrate that our method outperforms current state-of-the-art methods."
                    },
                    {
                        "title": "Identity-Aware Hand Mesh Estimation and Personalization from RGB Images",
                        "abstract": "Reconstructing 3D hand meshes from monocular RGB images has attracted increasing amount of attention due to its enormous potential applications in the field of AR/VR. Most state-of-the-art methods attempt to tackle this task in an anonymous manner. Specifically, the identity of the subject is ignored even though it is practically available in real applications where the user is unchanged in a continuous recording session. In this paper, we propose an identity-aware hand mesh estimation model, which can incorporate the identity information represented by the intrinsic shape parameters of the subject. We demonstrate the importance of the identity information by comparing the proposed identity-aware model to a baseline which treats subject anonymously. Furthermore, to handle the use case where the test subject is unseen, we propose a novel personalization pipeline to calibrate the intrinsic shape parameters using only a few unlabeled RGB images of the subject. Experiments on two large scale public datasets validate the state-of-the-art performance of our proposed method."
                    },
                    {
                        "title": "Diffeomorphic Image Registration with Neural Velocity Field",
                        "abstract": "Diffeomorphic image registration, offering smooth transformation and topology preservation, is required in many medical image analysis tasks. Traditional methods impose certain modeling constraints on the space of admissible transformations and use optimization to find the optimal transformation between two images. Specifying the right space of admissible transformations is challenging: the registration quality can be poor if the space is too restrictive, while the optimization can be hard to solve if the space is too general. Recent learning-based methods, utilizing deep neural networks to learn the transformation directly, achieve fast inference, but face challenges in accuracy due to the difficulties in capturing the small local deformations and generalization ability. Here we propose a new optimization-based method named DNVF (Diffeomorphic Image Registration with Neural Velocity Field) which utilizes deep neural network to model the space of admissible transformations. A multilayer perceptron (MLP) with sinusoidal activation function is used to represent the continuous velocity field and assigns a velocity vector to every point in space, providing the flexibility of modeling complex deformations as well as the convenience of optimization. Moreover, we propose a cascaded image registration framework (Cas-DNVF) by combining the benefits of both optimization and learning based methods, where a fully convolutional neural network (FCN) is trained to predict the initial deformation, followed by DNVF for further refinement. Experiments on two large-scale 3D MR brain scan datasets demonstrate that our proposed methods significantly outperform the state-of-the-art registration methods."
                    },
                    {
                        "title": "Topology-Preserving Shape Reconstruction and Registration via Neural Diffeomorphic Flow",
                        "abstract": "Deep Implicit Functions (DIFs) represent 3D geometry with continuous signed distance functions learned through deep neural nets. Recently DIFs-based methods have been proposed to handle shape reconstruction and dense point correspondences simultaneously, capturing semantic relationships across shapes of the same class by learning a DIFs-modeled shape template. These methods provide great flexibility and accuracy in reconstructing 3D shapes and inferring correspondences. However, the point correspondences built from these methods do not intrinsically preserve the topology of the shapes, unlike mesh-based template matching methods. This limits their applications on 3D geometries where underlying topological structures exist and matter, such as anatomical structures in medical images. In this paper, we propose a new model called Neural Diffeomorphic Flow (NDF) to learn deep implicit shape templates, representing shapes as conditional diffeomorphic deformations of templates, intrinsically preserving shape topologies. The diffeomorphic deformation is realized by an autodecoder consisting of Neural Ordinary Differential Equation (NODE) blocks that progressively map shapes to implicit templates. We conduct extensive experiments on several medical image organ segmentation datasets to evaluate the effectiveness of NDF on reconstructing and aligning shapes. NDF achieves consistently state-of-the-art organ shape reconstruction and registration results in both accuracy and quality. The source code is publicly available at https://github.com/Siwensun/Neural_Diffeomorphic_Flow-NDF."
                    },
                    {
                        "title": "Self-Supervised Contrastive Learning for Multi-Organ Segmentation",
                        "abstract": "Medical imaging modalities such as computed tomography (CT) and magnetic resonance imaging (MRI) play an important role in clinical workflows because they allow radiologists to analyse a patient\u2019s anatomy in great detail while being minimally invasive. Organ segmentation, in particular, is often performed as a preliminary step for treatment planning, diagnosis"
                    },
                    {
                        "title": "Implicit Optimizer for Diffeomorphic Image Registration",
                        "abstract": ". Di\ufb00eomorphic image registration is the underlying technology in medical image processing which enables the invertibility and point-to-point correspondence. Recently, numerous learning-based meth-ods utilizing convolutional neural networks (CNNs) have been proposed for registration problems. Compared with the speed boosting, accuracy improvement brought by the complicated CNN-based methods is minor. To tackle this problem, we propose a rapid and accurate Implicit Op-timizer for Di\ufb00eomorphic Image Registration (IDIR) which utilize the Deep Implicit Function as the neural velocity \ufb01eld (NVF) whose input is the point coordinate p and output is velocity vector at that point v . To reduce the huge memory consumption brought by NVF for 3D volumes, a sparse sampling is employed to the framework. We evaluate our method on two 3D large-scale MR brain scan datasets, the results show that our proposed method provide faster and better registration results than conventional image registration approaches and outperforms the learning-based methods by a signi\ufb01cant margin, while maintaining the desired di\ufb00eomorphic properties."
                    },
                    {
                        "title": "MIRNF: Medical Image Registration via Neural Fields",
                        "abstract": "Image registration is widely used in medical image analysis to provide spatial correspondences between two images. Recently learning-based methods utilizing convolutional neural networks (CNNs) have been proposed for solving image registration problems. The learning-based methods tend to be much faster than traditional optimization based methods, but the accuracy improvements gained from the complex CNN-based methods are modest. Here we introduce a new deep-neural net based image registration framework, named MIRNF , which represents the correspondence mapping with a continuous function implemented via Neural Fields. MIRNF outputs either a deformation vector or velocity vector given a 3D coordinate as input. To ensure the mapping is diffeomorphic, the velocity vector output from MIRNF is integrated using the Neural ODE solver to derive the correspondences between two images. Furthermore, we propose a hybrid coordinate sampler along with a cascaded architecture to achieve the high-similarity mapping performance and low-distortion deformation \ufb01elds. We conduct experiments on two 3D MR brain scan datasets, showing that our proposed framework provides state-of-art registration performance while maintaining comparable optimization time."
                    }
                ]
            },
            "f63652ed-400c-4794-8d89-a07dd6bb1be3": {
                "pk": "f63652ed-400c-4794-8d89-a07dd6bb1be3",
                "name": "Kun Han",
                "collaborators": [
                    "Shanlin Sun",
                    "Xiaohui Xie",
                    "Xiangyi Yan",
                    "Hao Tang",
                    "Haoyu Ma",
                    "Deying Kong",
                    "Chenyu You",
                    "Yifeng Xiong",
                    "Junayed Naushad",
                    "J. Duncan",
                    "Xingwei Liu",
                    "Tianang Leng",
                    "Thanh-Tung Le",
                    "Tong Zhang",
                    "Pooya Khosravi",
                    "Linguang Zhang",
                    "Liangjian Chen",
                    "Bin Yang",
                    "Haiwei Pan",
                    "Shuning He",
                    "Xuecheng Zhao",
                    "Narisu Bai",
                    "Xuming Chen",
                    "Qian Huang",
                    "Yong Liu"
                ],
                "domain": [
                    "3D Computer Vision",
                    "Medical Image Segmentation",
                    "Deep Learning",
                    "Self-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Hybrid Neural Diffeomorphic Flow for Shape Representation and Generation via Triplane",
                        "abstract": "Deep Implicit Functions (DIFs) have gained popularity in 3D computer vision due to their compactness and continuous representation capabilities. However, addressing dense correspondences and semantic relationships across DIF-encoded shapes remains a critical challenge, limiting their applications in texture transfer and shape analysis. Moreover, recent endeavors in 3D shape generation using DIFs often neglect correspondence and topology preservation. This paper presents HNDF (Hybrid Neural Diffeomorphic Flow), a method that implicitly learns the underlying representation and decomposes intricate dense correspondences into explicitly axis-aligned triplane features. To avoid suboptimal representations trapped in local minima, we propose hybrid supervision that captures both local and global correspondences. Unlike conventional approaches that directly generate new 3D shapes, we further explore the idea of shape generation with deformed template shape via diffeomorphic flows, where the deformation is encoded by the generated triplane features. Leveraging a pre-existing 2D diffusion model, we produce high-quality and diverse 3D diffeomorphic flows through generated triplanes features, ensuring topological consistency with the template shape. Extensive experiments on medical image organ segmentation datasets evaluate the effectiveness of HNDF in 3D shape representation and generation."
                    },
                    {
                        "title": "Self-Sampling Meta SAM: Enhancing Few-shot Medical Image Segmentation with Meta-Learning",
                        "abstract": "While the Segment Anything Model (SAM) excels in semantic segmentation for general-purpose images, its performance significantly deteriorates when applied to medical images, primarily attributable to insufficient representation of medical images in its training dataset. Nonetheless, gathering comprehensive datasets and training models that are universally applicable is particularly challenging due to the long-tail problem common in medical images.To address this gap, here we present a Self-Sampling Meta SAM (SSM-SAM) framework for few-shot medical image segmentation. Our innovation lies in the design of three key modules: 1) An online fast gradient descent optimizer, further optimized by a meta-learner, which ensures swift and robust adaptation to new tasks. 2) A Self-Sampling module designed to provide well-aligned visual prompts for improved attention allocation; and 3) A robust attention-based decoder specifically designed for few-shot medical image segmentation to capture relationship between different slices. Extensive experiments on a popular abdominal CT dataset and an MRI dataset demonstrate that the proposed method achieves significant improvements over state-of-the-art methods in few-shot segmentation, with an average improvements of 10.21% and 1.80% in terms of DSC, respectively. In conclusion, we present a novel approach for rapid online adaptation in interactive image segmentation, adapting to a new organ in just 0.83 minutes. Code is available at https://github.com/DragonDescentZerotsu/SSM-SAM"
                    },
                    {
                        "title": "Hybrid-CSR: Coupling Explicit and Implicit Shape Representation for Cortical Surface Reconstruction",
                        "abstract": "We present Hybrid-CSR, a geometric deep-learning model that combines explicit and implicit shape representations for cortical surface reconstruction. Specifically, Hybrid-CSR begins with explicit deformations of template meshes to obtain coarsely reconstructed cortical surfaces, based on which the oriented point clouds are estimated for the subsequent differentiable poisson surface reconstruction. By doing so, our method unifies explicit (oriented point clouds) and implicit (indicator function) cortical surface reconstruction. Compared to explicit representation-based methods, our hybrid approach is more friendly to capture detailed structures, and when compared with implicit representation-based methods, our method can be topology aware because of end-to-end training with a mesh-based deformation module. In order to address topology defects, we propose a new topology correction pipeline that relies on optimization-based diffeomorphic surface registration. Experimental results on three brain datasets show that our approach surpasses existing implicit and explicit cortical surface reconstruction methods in numeric metrics in terms of accuracy, regularity, and consistency."
                    },
                    {
                        "title": "Light Field Diffusion for Single-View Novel View Synthesis",
                        "abstract": "Single-view novel view synthesis (NVS), the task of generating images from new viewpoints based on a single reference image, is important but challenging in computer vision. Recent advancements in NVS have leveraged Denoising Diffusion Probabilistic Models (DDPMs) for their exceptional ability to produce high-fidelity images. However, current diffusion-based methods typically utilize camera pose matrices to globally and implicitly enforce 3D constraints, which can lead to inconsistencies in images generated from varying viewpoints, particularly in regions with complex textures and structures. To address these limitations, we present Light Field Diffusion (LFD), a novel conditional diffusion-based approach that transcends the conventional reliance on camera pose matrices. Starting from the camera pose matrices, LFD transforms them into light field encoding, with the same shape as the reference image, to describe the direction of each ray. By integrating light field encoding with the reference image, our method imposes local pixel-wise constraints within the diffusion process, fostering enhanced view consistency. Our approach not only involves training image LFD on the ShapeNet Car dataset but also includes fine-tuning a pre-trained latent diffusion model on the Objaverse dataset. This enables our latent LFD model to exhibit remarkable zero-shot generalization capabilities across out-of-distribution datasets like RTMV as well as in-the-wild images. Experiments demonstrate that LFD not only produces high-fidelity images but also achieves superior 3D consistency in complex regions, outperforming existing novel view synthesis methods."
                    },
                    {
                        "title": "CVTHead: One-shot Controllable Head Avatar with Vertex-feature Transformer",
                        "abstract": "Reconstructing personalized animatable head avatars has significant implications in the fields of AR/VR. Existing methods for achieving explicit face control of 3D Morphable Models (3DMM) typically rely on multi-view images or videos of a single subject, making the reconstruction process complex. Additionally, the traditional rendering pipeline is time-consuming, limiting real-time animation possibilities. In this paper, we introduce CVTHead, a novel approach that generates controllable neural head avatars from a single reference image using point-based neural rendering. CVT-Head considers the sparse vertices of mesh as the point set and employs the proposed Vertex-feature Transformer to learn local feature descriptors for each vertex. This enables the modeling of long-range dependencies among all the vertices. Experimental results on the VoxCeleb dataset demonstrate that CVTHead achieves comparable performance to state-of-the-art graphics-based methods. Moreover, it enables efficient rendering of novel human heads with various expressions, head poses, and camera views. These attributes can be explicitly controlled using the coefficients of 3DMMs, facilitating versatile and realistic animation in real-time scenarios. Codes and pre-trained model can be found at https://github.com/HowieMa/CVTHead."
                    },
                    {
                        "title": "Localized Region Contrast for Enhancing Self-Supervised Learning in Medical Image Segmentation",
                        "abstract": "Recent advancements in self-supervised learning have demonstrated that effective visual representations can be learned from unlabeled images. This has led to increased interest in applying self-supervised learning to the medical domain, where unlabeled images are abundant and labeled images are difficult to obtain. However, most self-supervised learning approaches are modeled as image level discriminative or generative proxy tasks, which may not capture the finer level representations necessary for dense prediction tasks like multi-organ segmentation. In this paper, we propose a novel contrastive learning framework that integrates Localized Region Contrast (LRC) to enhance existing self-supervised pre-training methods for medical image segmentation. Our approach involves identifying Super-pixels by Felzenszwalb's algorithm and performing local contrastive learning using a novel contrastive sampling loss. Through extensive experiments on three multi-organ segmentation datasets, we demonstrate that integrating LRC to an existing self-supervised method in a limited annotation setting significantly improves segmentation performance. Moreover, we show that LRC can also be applied to fully-supervised pre-training methods to further boost performance."
                    },
                    {
                        "title": "MedGen3D: A Deep Generative Framework for Paired 3D Image and Mask Generation",
                        "abstract": "Acquiring and annotating sufficient labeled data is crucial in developing accurate and robust learning-based models, but obtaining such data can be challenging in many medical image segmentation tasks. One promising solution is to synthesize realistic data with ground-truth mask annotations. However, no prior studies have explored generating complete 3D volumetric images with masks. In this paper, we present MedGen3D, a deep generative framework that can generate paired 3D medical images and masks. First, we represent the 3D medical data as 2D sequences and propose the Multi-Condition Diffusion Probabilistic Model (MC-DPM) to generate multi-label mask sequences adhering to anatomical geometry. Then, we use an image sequence generator and semantic diffusion refiner conditioned on the generated mask sequences to produce realistic 3D medical images that align with the generated masks. Our proposed framework guarantees accurate alignment between synthetic images and segmentation maps. Experiments on 3D thoracic CT and brain MRI datasets show that our synthetic data is both diverse and faithful to the original data, and demonstrate the benefits for downstream segmentation tasks. We anticipate that MedGen3D's ability to synthesize paired 3D medical images and masks will prove valuable in training deep learning models for medical imaging tasks."
                    },
                    {
                        "title": "Representation Recovering for Self-Supervised Pre-training on Medical Images",
                        "abstract": "Advances in self-supervised learning have drawn attention to developing techniques to extract effective visual representations from unlabeled images. Contrastive learning (CL) trains a model to extract consistent features by generating different views. Recent success of Masked Autoencoders (MAE) highlights the benefit of generative modeling in self-supervised learning. The generative approaches encode the input into a compact embedding and empower the model\u2019s ability of recovering the original input. However, in our experiments, we found vanilla MAE mainly recovers coarse high level semantic information and is inadequate in recovering detailed low level information. We show that in dense downstream prediction tasks like multi-organ segmentation, directly applying MAE is not ideal. Here, we propose RepRec, a hybrid visual representation learning framework for self-supervised pre-training on large-scale unlabelled medical datasets, which takes advantage of both contrastive and generative modeling. To solve the aforementioned dilemma that MAE encounters, a convolutional encoder is pre-trained to provide low-level feature information, in a contrastive way; and a transformer encoder is pre-trained to produce high level semantic dependency, in a generative way \u2013 by recovering masked representations from the convolutional encoder. Extensive experiments on three multi-organ segmentation datasets demonstrate that our method outperforms current state-of-the-art methods."
                    },
                    {
                        "title": "Identity-Aware Hand Mesh Estimation and Personalization from RGB Images",
                        "abstract": "Reconstructing 3D hand meshes from monocular RGB images has attracted increasing amount of attention due to its enormous potential applications in the field of AR/VR. Most state-of-the-art methods attempt to tackle this task in an anonymous manner. Specifically, the identity of the subject is ignored even though it is practically available in real applications where the user is unchanged in a continuous recording session. In this paper, we propose an identity-aware hand mesh estimation model, which can incorporate the identity information represented by the intrinsic shape parameters of the subject. We demonstrate the importance of the identity information by comparing the proposed identity-aware model to a baseline which treats subject anonymously. Furthermore, to handle the use case where the test subject is unseen, we propose a novel personalization pipeline to calibrate the intrinsic shape parameters using only a few unlabeled RGB images of the subject. Experiments on two large scale public datasets validate the state-of-the-art performance of our proposed method."
                    },
                    {
                        "title": "Diffeomorphic Image Registration with Neural Velocity Field",
                        "abstract": "Diffeomorphic image registration, offering smooth transformation and topology preservation, is required in many medical image analysis tasks. Traditional methods impose certain modeling constraints on the space of admissible transformations and use optimization to find the optimal transformation between two images. Specifying the right space of admissible transformations is challenging: the registration quality can be poor if the space is too restrictive, while the optimization can be hard to solve if the space is too general. Recent learning-based methods, utilizing deep neural networks to learn the transformation directly, achieve fast inference, but face challenges in accuracy due to the difficulties in capturing the small local deformations and generalization ability. Here we propose a new optimization-based method named DNVF (Diffeomorphic Image Registration with Neural Velocity Field) which utilizes deep neural network to model the space of admissible transformations. A multilayer perceptron (MLP) with sinusoidal activation function is used to represent the continuous velocity field and assigns a velocity vector to every point in space, providing the flexibility of modeling complex deformations as well as the convenience of optimization. Moreover, we propose a cascaded image registration framework (Cas-DNVF) by combining the benefits of both optimization and learning based methods, where a fully convolutional neural network (FCN) is trained to predict the initial deformation, followed by DNVF for further refinement. Experiments on two large-scale 3D MR brain scan datasets demonstrate that our proposed methods significantly outperform the state-of-the-art registration methods."
                    },
                    {
                        "title": "Topology-Preserving Shape Reconstruction and Registration via Neural Diffeomorphic Flow",
                        "abstract": "Deep Implicit Functions (DIFs) represent 3D geometry with continuous signed distance functions learned through deep neural nets. Recently DIFs-based methods have been proposed to handle shape reconstruction and dense point correspondences simultaneously, capturing semantic relationships across shapes of the same class by learning a DIFs-modeled shape template. These methods provide great flexibility and accuracy in reconstructing 3D shapes and inferring correspondences. However, the point correspondences built from these methods do not intrinsically preserve the topology of the shapes, unlike mesh-based template matching methods. This limits their applications on 3D geometries where underlying topological structures exist and matter, such as anatomical structures in medical images. In this paper, we propose a new model called Neural Diffeomorphic Flow (NDF) to learn deep implicit shape templates, representing shapes as conditional diffeomorphic deformations of templates, intrinsically preserving shape topologies. The diffeomorphic deformation is realized by an autodecoder consisting of Neural Ordinary Differential Equation (NODE) blocks that progressively map shapes to implicit templates. We conduct extensive experiments on several medical image organ segmentation datasets to evaluate the effectiveness of NDF on reconstructing and aligning shapes. NDF achieves consistently state-of-the-art organ shape reconstruction and registration results in both accuracy and quality. The source code is publicly available at https://github.com/Siwensun/Neural_Diffeomorphic_Flow-NDF."
                    },
                    {
                        "title": "Implicit Optimizer for Diffeomorphic Image Registration",
                        "abstract": ". Di\ufb00eomorphic image registration is the underlying technology in medical image processing which enables the invertibility and point-to-point correspondence. Recently, numerous learning-based meth-ods utilizing convolutional neural networks (CNNs) have been proposed for registration problems. Compared with the speed boosting, accuracy improvement brought by the complicated CNN-based methods is minor. To tackle this problem, we propose a rapid and accurate Implicit Op-timizer for Di\ufb00eomorphic Image Registration (IDIR) which utilize the Deep Implicit Function as the neural velocity \ufb01eld (NVF) whose input is the point coordinate p and output is velocity vector at that point v . To reduce the huge memory consumption brought by NVF for 3D volumes, a sparse sampling is employed to the framework. We evaluate our method on two 3D large-scale MR brain scan datasets, the results show that our proposed method provide faster and better registration results than conventional image registration approaches and outperforms the learning-based methods by a signi\ufb01cant margin, while maintaining the desired di\ufb00eomorphic properties."
                    },
                    {
                        "title": "MIRNF: Medical Image Registration via Neural Fields",
                        "abstract": "Image registration is widely used in medical image analysis to provide spatial correspondences between two images. Recently learning-based methods utilizing convolutional neural networks (CNNs) have been proposed for solving image registration problems. The learning-based methods tend to be much faster than traditional optimization based methods, but the accuracy improvements gained from the complex CNN-based methods are modest. Here we introduce a new deep-neural net based image registration framework, named MIRNF , which represents the correspondence mapping with a continuous function implemented via Neural Fields. MIRNF outputs either a deformation vector or velocity vector given a 3D coordinate as input. To ensure the mapping is diffeomorphic, the velocity vector output from MIRNF is integrated using the Neural ODE solver to derive the correspondences between two images. Furthermore, we propose a hybrid coordinate sampler along with a cascaded architecture to achieve the high-similarity mapping performance and low-distortion deformation \ufb01elds. We conduct experiments on two 3D MR brain scan datasets, showing that our proposed framework provides state-of-art registration performance while maintaining comparable optimization time."
                    },
                    {
                        "title": "Forest Fire Thermal Infrared Image Segmentation Based on K-V Model*",
                        "abstract": "With the growing problem of forest fires, thermal infrared imaging technology is gradually applied to monitor and control forest fires. The segmentation of thermal infrared images is of great significance as an important part of this technology. This paper proposes an image segmentation model based on K-means clustering and variational (K-V model), which is used to alleviate the problem that the forest fire thermal infrared image is difficult to be segmented due to the presence of smoke masking, boundary blur of the fire area and regional dispersion of the fire area. Experiments are on a data set obtained by transforming the forest fire thermal infrared images collected on the Internet. This paper tests the running time and qualitative segmentation results of the proposed K-V model, and obtains convincing performance."
                    },
                    {
                        "title": "Spatial Context-Aware Self-Attention Model For Multi-Organ Segmentation",
                        "abstract": "Multi-organ segmentation is one of most successful applications of deep learning in medical image analysis. Deep convolutional neural nets (CNNs) have shown great promise in achieving clinically applicable image segmentation performance on CT or MRI images. State-of-the-art CNN segmentation models apply either 2D or 3D convolutions on input images, with pros and cons associated with each method: 2D convolution is fast, less memory-intensive but inadequate for extracting 3D contextual information from volumetric images, while the opposite is true for 3D convolution. To fit a 3D CNN model on CT or MRI images on commodity GPUs, one usually has to either downsample input images or use cropped local regions as inputs, which limits the utility of 3D models for multi-organ segmentation. In this work, we propose a new framework for combining 3D and 2D models, in which the segmentation is realized through high-resolution 2D convolutions, but guided by spatial contextual information extracted from a low-resolution 3D model. We implement a self-attention mechanism to control which 3D features should be used to guide 2D segmentation. Our model is light on memory usage but fully equipped to take 3D contextual information into account. Experiments on multiple organ segmentation datasets demonstrate that by taking advantage of both 2D and 3D models, our method is consistently outperforms existing 2D and 3D models in organ segmentation accuracy, while being able to directly take raw whole-volume image data as inputs."
                    }
                ]
            },
            "1837716a-d6f2-46be-8818-495bad82d0de": {
                "pk": "1837716a-d6f2-46be-8818-495bad82d0de",
                "name": "Nhat Ho",
                "collaborators": [
                    "Khai Nguyen",
                    "T. Nguyen",
                    "Tam Nguyen",
                    "S. Osher",
                    "Huy Nguyen",
                    "Richard Baraniuk",
                    "Tongzheng Ren",
                    "Hien Dang",
                    "Tho Tran",
                    "Hung The Tran",
                    "TrungTin Nguyen",
                    "Disha Makhija",
                    "J. Ghosh",
                    "Hai Do",
                    "Duy Khuong Nguyen",
                    "Dat Do",
                    "A. Bertozzi",
                    "Hung Tran",
                    "Nicola Bariletto",
                    "Pedram Akbarian",
                    "Fanqi Yan",
                    "Long Bui",
                    "Dung D. Le",
                    "D. M. Nguyen",
                    "N. T. Diep",
                    "T. Pham",
                    "T. Cao",
                    "Binh Duc Nguyen",
                    "P. Swoboda",
                    "Shadi Albarqouni",
                    "P. Xie",
                    "Daniel Sonntag",
                    "Mathias Niepert",
                    "Vishwanath Saragadam",
                    "Minh Pham",
                    "Qiujiang Jin",
                    "Aryan Mokhtari",
                    "X. Nguyen"
                ],
                "domain": [
                    "Machine Learning",
                    "Deep Learning",
                    "Federated Learning",
                    "Wasserstein Distance"
                ],
                "publications": [
                    {
                        "title": "A Primal-Dual Framework for Transformers and Neural Networks",
                        "abstract": "Self-attention is key to the remarkable success of transformers in sequence modeling tasks including many applications in natural language processing and computer vision. Like neural network layers, these attention mechanisms are often developed by heuristics and experience. To provide a principled framework for constructing attention layers in transformers, we show that the self-attention corresponds to the support vector expansion derived from a support vector regression problem, whose primal formulation has the form of a neural network layer. Using our framework, we derive popular attention layers used in practice and propose two new attentions: 1) the Batch Normalized Attention (Attention-BN) derived from the batch normalization layer and 2) the Attention with Scaled Head (Attention-SH) derived from using less training data to fit the SVR model. We empirically demonstrate the advantages of the Attention-BN and Attention-SH in reducing head redundancy, increasing the model's accuracy, and improving the model's efficiency in a variety of practical applications including image and time-series classification."
                    },
                    {
                        "title": "Markovian Sliced Wasserstein Distances: Beyond Independent Projections",
                        "abstract": "Sliced Wasserstein (SW) distance suffers from redundant projections due to independent uniform random projecting directions. To partially overcome the issue, max K sliced Wasserstein (Max-K-SW) distance ($K\\geq 1$), seeks the best discriminative orthogonal projecting directions. Despite being able to reduce the number of projections, the metricity of Max-K-SW cannot be guaranteed in practice due to the non-optimality of the optimization. Moreover, the orthogonality constraint is also computationally expensive and might not be effective. To address the problem, we introduce a new family of SW distances, named Markovian sliced Wasserstein (MSW) distance, which imposes a first-order Markov structure on projecting directions. We discuss various members of MSW by specifying the Markov structure including the prior distribution, the transition distribution, and the burning and thinning technique. Moreover, we investigate the theoretical properties of MSW including topological properties (metricity, weak convergence, and connection to other distances), statistical properties (sample complexity, and Monte Carlo estimation error), and computational properties (computational complexity and memory complexity). Finally, we compare MSW distances with previous SW variants in various applications such as gradient flows, color transfer, and deep generative modeling to demonstrate the favorable performance of MSW."
                    },
                    {
                        "title": "Neural Collapse in Deep Linear Networks: From Balanced to Imbalanced Data",
                        "abstract": "Modern deep neural networks have achieved impressive performance on tasks from image classification to natural language processing. Surprisingly, these complex systems with massive amounts of parameters exhibit the same structural properties in their last-layer features and classifiers across canonical datasets when training until convergence. In particular, it has been observed that the last-layer features collapse to their class-means, and those class-means are the vertices of a simplex Equiangular Tight Frame (ETF). This phenomenon is known as Neural Collapse (NC). Recent papers have theoretically shown that NC emerges in the global minimizers of training problems with the simplified\"unconstrained feature model\". In this context, we take a step further and prove the NC occurrences in deep linear networks for the popular mean squared error (MSE) and cross entropy (CE) losses, showing that global solutions exhibit NC properties across the linear layers. Furthermore, we extend our study to imbalanced data for MSE loss and present the first geometric analysis of NC under bias-free setting. Our results demonstrate the convergence of the last-layer features and classifiers to a geometry consisting of orthogonal vectors, whose lengths depend on the amount of data in their corresponding classes. Finally, we empirically validate our theoretical analyses on synthetic and practical network architectures with both balanced and imbalanced scenarios."
                    },
                    {
                        "title": "Energy-Based Sliced Wasserstein Distance",
                        "abstract": "The sliced Wasserstein (SW) distance has been widely recognized as a statistically effective and computationally efficient metric between two probability measures. A key component of the SW distance is the slicing distribution. There are two existing approaches for choosing this distribution. The first approach is using a fixed prior distribution. The second approach is optimizing for the best distribution which belongs to a parametric family of distributions and can maximize the expected distance. However, both approaches have their limitations. A fixed prior distribution is non-informative in terms of highlighting projecting directions that can discriminate two general probability measures. Doing optimization for the best distribution is often expensive and unstable. Moreover, designing the parametric family of the candidate distribution could be easily misspecified. To address the issues, we propose to design the slicing distribution as an energy-based distribution that is parameter-free and has the density proportional to an energy function of the projected one-dimensional Wasserstein distance. We then derive a novel sliced Wasserstein metric, energy-based sliced Waserstein (EBSW) distance, and investigate its topological, statistical, and computational properties via importance sampling, sampling importance resampling, and Markov Chain methods. Finally, we conduct experiments on point-cloud gradient flow, color transfer, and point-cloud reconstruction to show the favorable performance of the EBSW."
                    },
                    {
                        "title": "Towards Convergence Rates for Parameter Estimation in Gaussian-gated Mixture of Experts",
                        "abstract": "Originally introduced as a neural network for ensemble learning, mixture of experts (MoE) has recently become a fundamental building block of highly successful modern deep neural networks for heterogeneous data analysis in several applications of machine learning and statistics. Despite its popularity in practice, a satisfactory level of theoretical understanding of the MoE model is far from complete. To shed new light on this problem, we provide a convergence analysis for maximum likelihood estimation (MLE) in the Gaussian-gated MoE model. The main challenge of that analysis comes from the inclusion of covariates in the Gaussian gating functions and expert networks, which leads to their intrinsic interaction via some partial differential equations with respect to their parameters. We tackle these issues by designing novel Voronoi loss functions among parameters to accurately capture the heterogeneity of parameter estimation rates. Our findings reveal that the MLE has distinct behaviors under two complement settings of location parameters of the Gaussian gating functions, namely when all these parameters are non-zero versus when at least one among them vanishes. Notably, these behaviors can be characterized by the solvability of two different systems of polynomial equations. Finally, we conduct a simulation study to empirically verify our theoretical results."
                    },
                    {
                        "title": "Privacy Preserving Bayesian Federated Learning in Heterogeneous Settings",
                        "abstract": "In several practical applications of federated learning (FL), the clients are highly heterogeneous in terms of both their data and compute resources, and therefore enforcing the same model architecture for each client is very limiting. Moreover, the need for uncertainty quantification and data privacy constraints are often particularly amplified for clients that have limited local data. This paper presents a unified FL framework to simultaneously address all these constraints and concerns, based on training customized local Bayesian models that learn well even in the absence of large local datasets. A Bayesian framework provides a natural way of incorporating supervision in the form of prior distributions. We use priors in the functional (output) space of the networks to facilitate collaboration across heterogeneous clients. Moreover, formal differential privacy guarantees are provided for this framework. Experiments on standard FL datasets demonstrate that our approach outperforms strong baselines in both homogeneous and heterogeneous settings and under strict privacy constraints, while also providing characterizations of model uncertainties."
                    },
                    {
                        "title": "Quasi-Monte Carlo for 3D Sliced Wasserstein",
                        "abstract": "Monte Carlo (MC) integration has been employed as the standard approximation method for the Sliced Wasserstein (SW) distance, whose analytical expression involves an intractable expectation. However, MC integration is not optimal in terms of absolute approximation error. To provide a better class of empirical SW, we propose quasi-sliced Wasserstein (QSW) approximations that rely on Quasi-Monte Carlo (QMC) methods. For a comprehensive investigation of QMC for SW, we focus on the 3D setting, specifically computing the SW between probability measures in three dimensions. In greater detail, we empirically evaluate various methods to construct QMC point sets on the 3D unit-hypersphere, including the Gaussian-based and equal area mappings, generalized spiral points, and optimizing discrepancy energies. Furthermore, to obtain an unbiased estimator for stochastic optimization, we extend QSW to Randomized Quasi-Sliced Wasserstein (RQSW) by introducing randomness in the discussed point sets. Theoretically, we prove the asymptotic convergence of QSW and the unbiasedness of RQSW. Finally, we conduct experiments on various 3D tasks, such as point-cloud comparison, point-cloud interpolation, image style transfer, and training deep point-cloud autoencoders, to demonstrate the favorable performance of the proposed QSW and RQSW variants."
                    },
                    {
                        "title": "Statistical Perspective of Top-K Sparse Softmax Gating Mixture of Experts",
                        "abstract": "Top-K sparse softmax gating mixture of experts has been widely used for scaling up massive deep-learning architectures without increasing the computational cost. Despite its popularity in real-world applications, the theoretical understanding of that gating function has remained an open problem. The main challenge comes from the structure of the top-K sparse softmax gating function, which partitions the input space into multiple regions with distinct behaviors. By focusing on a Gaussian mixture of experts, we establish theoretical results on the effects of the top-K sparse softmax gating function on both density and parameter estimations. Our results hinge upon defining novel loss functions among parameters to capture different behaviors of the input regions. When the true number of experts $k_{\\ast}$ is known, we demonstrate that the convergence rates of density and parameter estimations are both parametric on the sample size. However, when $k_{\\ast}$ becomes unknown and the true model is over-specified by a Gaussian mixture of $k$ experts where $k>k_{\\ast}$, our findings suggest that the number of experts selected from the top-K sparse softmax gating function must exceed the total cardinality of a certain number of Voronoi cells associated with the true parameters to guarantee the convergence of the density estimation. Moreover, while the density estimation rate remains parametric under this setting, the parameter estimation rates become substantially slow due to an intrinsic interaction between the softmax gating and expert functions."
                    },
                    {
                        "title": "Sliced Wasserstein Estimation with Control Variates",
                        "abstract": "The sliced Wasserstein (SW) distances between two probability measures are defined as the expectation of the Wasserstein distance between two one-dimensional projections of the two measures. The randomness comes from a projecting direction that is used to project the two input measures to one dimension. Due to the intractability of the expectation, Monte Carlo integration is performed to estimate the value of the SW distance. Despite having various variants, there has been no prior work that improves the Monte Carlo estimation scheme for the SW distance in terms of controlling its variance. To bridge the literature on variance reduction and the literature on the SW distance, we propose computationally efficient control variates to reduce the variance of the empirical estimation of the SW distance. The key idea is to first find Gaussian approximations of projected one-dimensional measures, then we utilize the closed-form of the Wasserstein-2 distance between two Gaussian distributions to design the control variates. In particular, we propose using a lower bound and an upper bound of the Wasserstein-2 distance between two fitted Gaussians as two computationally efficient control variates. We empirically show that the proposed control variate estimators can help to reduce the variance considerably when comparing measures over images and point-clouds. Finally, we demonstrate the favorable performance of the proposed control variate estimators in gradient flows to interpolate between two point-clouds and in deep generative modeling on standard image datasets, such as CIFAR10 and CelebA."
                    },
                    {
                        "title": "A Probabilistic Framework for Pruning Transformers Via a Finite Admixture of Keys",
                        "abstract": "Pairwise dot product-based self-attention is key to the success of transformers which achieve state-of-the-art performance across a variety of applications in language and vision, but are costly to compute. It has been shown that most attention scores and keys in transformers are redundant and can be removed without loss of accuracy. In this paper, we develop a novel probabilistic framework for pruning attention scores and keys in transformers. We first formulate an admixture model of attention keys whose input data to be clustered are attention queries. We show that attention scores in self-attention correspond to the posterior distribution of this model when attention keys admit a uniform prior distribution. We then relax this uniform prior constraint and let the model learn these priors from data, resulting in a new Finite Admixture of Keys (FiAK). The learned priors are used for pruning away redundant attention scores and keys in the baseline transformers, improving the diversity of attention patterns that the models capture. We corroborate the efficiency of transformers pruned with FiAK on the ImageNet object classification and WikiText-103 language modeling tasks. Our experiments demonstrate that transformers pruned with FiAK yield similar or better accuracy than the baseline dense transformers while being much more efficient in terms of memory and computational cost."
                    },
                    {
                        "title": "Beyond Vanilla Variational Autoencoders: Detecting Posterior Collapse in Conditional and Hierarchical Variational Autoencoders",
                        "abstract": "The posterior collapse phenomenon in variational autoencoder (VAE), where the variational posterior distribution closely matches the prior distribution, can hinder the quality of the learned latent variables. As a consequence of posterior collapse, the latent variables extracted by the encoder in VAE preserve less information from the input data and thus fail to produce meaningful representations as input to the reconstruction process in the decoder. While this phenomenon has been an actively addressed topic related to VAE performance, the theory for posterior collapse remains underdeveloped, especially beyond the standard VAE. In this work, we advance the theoretical understanding of posterior collapse to two important and prevalent yet less studied classes of VAE: conditional VAE and hierarchical VAE. Specifically, via a non-trivial theoretical analysis of linear conditional VAE and hierarchical VAE with two levels of latent, we prove that the cause of posterior collapses in these models includes the correlation between the input and output of the conditional VAE and the effect of learnable encoder variance in the hierarchical VAE. We empirically validate our theoretical findings for linear conditional and hierarchical VAE and demonstrate that these results are also predictive for non-linear cases with extensive experiments."
                    },
                    {
                        "title": "Minimax Optimal Rate for Parameter Estimation in Multivariate Deviated Models",
                        "abstract": "We study the maximum likelihood estimation (MLE) in the multivariate deviated model where the data are generated from the density function $(1-\\lambda^{\\ast})h_{0}(x)+\\lambda^{\\ast}f(x|\\mu^{\\ast}, \\Sigma^{\\ast})$ in which $h_{0}$ is a known function, $\\lambda^{\\ast} \\in [0,1]$ and $(\\mu^{\\ast}, \\Sigma^{\\ast})$ are unknown parameters to estimate. The main challenges in deriving the convergence rate of the MLE mainly come from two issues: (1) The interaction between the function $h_{0}$ and the density function $f$; (2) The deviated proportion $\\lambda^{\\ast}$ can go to the extreme points of $[0,1]$ as the sample size tends to infinity. To address these challenges, we develop the \\emph{distinguishability condition} to capture the linear independent relation between the function $h_{0}$ and the density function $f$. We then provide comprehensive convergence rates of the MLE via the vanishing rate of $\\lambda^{\\ast}$ to zero as well as the distinguishability of two functions $h_{0}$ and $f$."
                    },
                    {
                        "title": "Demystifying Softmax Gating Function in Gaussian Mixture of Experts",
                        "abstract": "Understanding the parameter estimation of softmax gating Gaussian mixture of experts has remained a long-standing open problem in the literature. It is mainly due to three fundamental theoretical challenges associated with the softmax gating function: (i) the identifiability only up to the translation of parameters; (ii) the intrinsic interaction via partial differential equations between the softmax gating and the expert functions in the Gaussian density; (iii) the complex dependence between the numerator and denominator of the conditional density of softmax gating Gaussian mixture of experts. We resolve these challenges by proposing novel Voronoi loss functions among parameters and establishing the convergence rates of maximum likelihood estimator (MLE) for solving parameter estimation in these models. When the true number of experts is unknown and over-specified, our findings show a connection between the convergence rate of the MLE and a solvability problem of a system of polynomial equations."
                    },
                    {
                        "title": "LVM-Med: Learning Large-Scale Self-Supervised Vision Models for Medical Imaging via Second-order Graph Matching",
                        "abstract": "Obtaining large pre-trained models that can be fine-tuned to new tasks with limited annotated samples has remained an open challenge for medical imaging data. While pre-trained deep networks on ImageNet and vision-language foundation models trained on web-scale data are prevailing approaches, their effectiveness on medical tasks is limited due to the significant domain shift between natural and medical images. To bridge this gap, we introduce LVM-Med, the first family of deep networks trained on large-scale medical datasets. We have collected approximately 1.3 million medical images from 55 publicly available datasets, covering a large number of organs and modalities such as CT, MRI, X-ray, and Ultrasound. We benchmark several state-of-the-art self-supervised algorithms on this dataset and propose a novel self-supervised contrastive learning algorithm using a graph-matching formulation. The proposed approach makes three contributions: (i) it integrates prior pair-wise image similarity metrics based on local and global information; (ii) it captures the structural constraints of feature embeddings through a loss function constructed via a combinatorial graph-matching objective; and (iii) it can be trained efficiently end-to-end using modern gradient-estimation techniques for black-box solvers. We thoroughly evaluate the proposed LVM-Med on 15 downstream medical tasks ranging from segmentation and classification to object detection, and both for the in and out-of-distribution settings. LVM-Med empirically outperforms a number of state-of-the-art supervised, self-supervised, and foundation models. For challenging tasks such as Brain Tumor Classification or Diabetic Retinopathy Grading, LVM-Med improves previous vision-language models trained on 1 billion masks by 6-7% while using only a ResNet-50."
                    },
                    {
                        "title": "Improving Transformer with an Admixture of Attention Heads",
                        "abstract": "Transformers with multi-head self-attention have achieved remarkable success in sequence modeling and beyond. However, they suffer from high computational and memory complexities for computing the attention matrix at each head. Recently, it has been shown that those attention matrices lie on a low-dimensional manifold and, thus, are redundant. We propose the Transformer with a Finite Admixture of Shared Heads (FiSHformers), a novel class of efficient and flexible transformers that allow the sharing of attention matrices between attention heads. At the core of FiSHformer is a novel finite admixture model of shared heads (FiSH) that samples attention matrices from a set of global attention matrices. The number of global attention matrices is much smaller than the number of local attention matrices generated. FiSHformers directly learn these global attention matrices rather than the local ones as in other transformers, thus significantly improving the computational and memory efficiency of the model. We empirically verify the advantages of the FiSHformer over the baseline transformers in a wide range of practical applications including language modeling, machine translation, and image classification. On the WikiText-103, IWSLT\u201914 De-En and WMT\u201914"
                    },
                    {
                        "title": "Transformer with Fourier Integral Attentions",
                        "abstract": "Multi-head attention empowers the recent success of transformers, the state-of-the-art models that have achieved remarkable success in sequence modeling and beyond. These attention mechanisms compute the pairwise dot products between the queries and keys, which results from the use of unnormalized Gaussian kernels with the assumption that the queries follow a mixture of Gaussian distribution. There is no guarantee that this assumption is valid in practice. In response, we first interpret attention in transformers as a nonparametric kernel regression. We then propose the FourierFormer, a new class of transformers in which the dot-product kernels are replaced by the novel generalized Fourier integral kernels. Different from the dot-product kernels, where we need to choose a good covariance matrix to capture the dependency of the features of data, the generalized Fourier integral kernels can automatically capture such dependency and remove the need to tune the covariance matrix. We theoretically prove that our proposed Fourier integral kernels can efficiently approximate any key and query distributions. Compared to the conventional transformers with dot-product attention, FourierFormers attain better accuracy and reduce the redundancy between attention heads. We empirically corroborate the advantages of FourierFormers over the baseline transformers in a variety of practical applications including language modeling and image classification."
                    },
                    {
                        "title": "Federated Self-supervised Learning for Heterogeneous Clients",
                        "abstract": "Federated Learning has become an important learning paradigm due to its privacy and computational benefits. As the field advances, two key challenges that still remain to be addressed are: (1) system heterogeneity - variability in the compute and/or data resources present on each client, and (2) lack of labeled data in certain federated settings. Several recent developments have tried to overcome these challenges independently. In this work, we propose a unified and systematic framework, \\emph{Heterogeneous Self-supervised Federated Learning} (Hetero-SSFL) for enabling self-supervised learning with federation on heterogeneous clients. The proposed framework allows collaborative representation learning across all the clients without imposing architectural constraints or requiring presence of labeled data. The key idea in Hetero-SSFL is to let each client train its unique self-supervised model and enable the joint learning across clients by aligning the lower dimensional representations on a common dataset. The entire training procedure could be viewed as self and peer-supervised as both the local training and the alignment procedures do not require presence of any labeled data. As in conventional self-supervised learning, the obtained client models are task independent and can be used for varied end-tasks. We provide a convergence guarantee of the proposed framework for non-convex objectives in heterogeneous settings and also empirically demonstrate that our proposed approach outperforms the state of the art methods by a significant margin."
                    },
                    {
                        "title": "Statistical and Computational Complexities of BFGS Quasi-Newton Method for Generalized Linear Models",
                        "abstract": "The gradient descent (GD) method has been used widely to solve parameter estimation in generalized linear models (GLMs), a generalization of linear models when the link function can be non-linear. In GLMs with a polynomial link function, it has been shown that in the high signal-to-noise ratio (SNR) regime, due to the problem's strong convexity and smoothness, GD converges linearly and reaches the final desired accuracy in a logarithmic number of iterations. In contrast, in the low SNR setting, where the problem becomes locally convex, GD converges at a slower rate and requires a polynomial number of iterations to reach the desired accuracy. Even though Newton's method can be used to resolve the flat curvature of the loss functions in the low SNR case, its computational cost is prohibitive in high-dimensional settings as it is $\\mathcal{O}(d^3)$, where $d$ the is the problem dimension. To address the shortcomings of GD and Newton's method, we propose the use of the BFGS quasi-Newton method to solve parameter estimation of the GLMs, which has a per iteration cost of $\\mathcal{O}(d^2)$. When the SNR is low, for GLMs with a polynomial link function of degree $p$, we demonstrate that the iterates of BFGS converge linearly to the optimal solution of the population least-square loss function, and the contraction coefficient of the BFGS algorithm is comparable to that of Newton's method. Moreover, the contraction factor of the linear rate is independent of problem parameters and only depends on the degree of the link function $p$. Also, for the empirical loss with $n$ samples, we prove that in the low SNR setting of GLMs with a polynomial link function of degree $p$, the iterates of BFGS reach a final statistical radius of $\\mathcal{O}((d/n)^{\\frac{1}{2p+2}})$ after at most $\\log(n/d)$ iterations."
                    },
                    {
                        "title": "Beyond Black Box Densities: Parameter Learning for the Deviated Components",
                        "abstract": "As we collect additional samples from a data population for which a known density function estimate may have been previously obtained by a black box method, the increased complexity of the data set may result in the true density being deviated from the known estimate by a mixture distribution. To model this phenomenon, we consider the \\emph{deviating mixture model} $(1-\\lambda^{*})h_0 + \\lambda^{*} (\\sum_{i = 1}^{k} p_{i}^{*} f(x|\\theta_{i}^{*}))$, where $h_0$ is a known density function, while the deviated proportion $\\lambda^{*}$ and latent mixing measure $G_{*} = \\sum_{i = 1}^{k} p_{i}^{*} \\delta_{\\theta_i^{*}}$ associated with the mixture distribution are unknown. Via a novel notion of distinguishability between the known density $h_{0}$ and the deviated mixture distribution, we establish rates of convergence for the maximum likelihood estimates of $\\lambda^{*}$ and $G^{*}$ under Wasserstein metric. Simulation studies are carried out to illustrate the theory."
                    }
                ]
            },
            "e0bdb8db-8ca9-44fb-ae9d-e3b669249903": {
                "pk": "e0bdb8db-8ca9-44fb-ae9d-e3b669249903",
                "name": "Xiaohui Xie",
                "collaborators": [
                    "Wei Li",
                    "Yu-Bo Sun",
                    "Yi Zhao",
                    "Qi Chen",
                    "Meiling Yu",
                    "Ying-Yi Wei",
                    "Chao Lian",
                    "Zhiping Zhang",
                    "Meng Feng",
                    "Wenqi Wang",
                    "Xiaomeng Yuan",
                    "Yan-Shuo Shi",
                    "Tingting Liu",
                    "Xiaohui Liu",
                    "Zhiwu Chen",
                    "Yan Wang",
                    "Dake Chen",
                    "Dongsheng Zhang",
                    "Jian-Shui Lin",
                    "Qingqing Wang",
                    "Dongsheng Zhong",
                    "Jinju Zhang",
                    "Tianhui Yuan",
                    "Hui-yi Wu",
                    "Bin Li",
                    "Shuangdi Li",
                    "D. An",
                    "Yue Deng",
                    "Shaoxiang Xian",
                    "X. Xiong",
                    "Kui-wu Yao",
                    "Dong Cui",
                    "L. Shen",
                    "Yihuan Shen",
                    "Guantong Han",
                    "Qianfei Cao",
                    "Jing Wang",
                    "Hao Wei",
                    "Qian-nan Wang",
                    "K. Zheng",
                    "Jiannan Wang",
                    "Shaofeng Li",
                    "Handan Zhang",
                    "Chan-Hsiang Chang",
                    "Tiantian Li",
                    "Jishi Cui",
                    "Liang-Gang Wang",
                    "Meng-Qian Liu",
                    "Minglian Zhang",
                    "Yi-Qu Ding",
                    "Zheng-Min Liang",
                    "Tingjun Hu",
                    "Wan-Wen Liang",
                    "Hang-tong Hu",
                    "Ming Kuang",
                    "Ming-de Lu",
                    "Sui Peng",
                    "W. Wang",
                    "Ying Zhou",
                    "Shouli Yi",
                    "Ying Xie",
                    "Mi Cao",
                    "Ling Zhang",
                    "Ting-Jun Hu",
                    "Wenwen Shen",
                    "Longhui Li",
                    "Yue Liu",
                    "Weisheng Chen",
                    "Huifen Liu",
                    "Wenwu Zhang",
                    "Yu Liu",
                    "Haihang Yu",
                    "Wenhua Zhou",
                    "Biao Gong",
                    "Yiliang Lv",
                    "Zhen Han",
                    "Guoshuai Zhao",
                    "Xueming Qian"
                ],
                "domain": [
                    "Meta-analysis",
                    "Drug Treatment",
                    "Aquaculture",
                    "Knowledge Graph"
                ],
                "publications": [
                    {
                        "title": "Aflibercept for long-term treatment of diabetic macular edema and proliferative diabetic retinopathy: a meta-analysis",
                        "abstract": "Purpose This meta-analysis compared the long-term (12 months or 24 months) efficacy and safety of intravitreal aflibercept injection (IAI) for diabetic macular edema (DME) and proliferative diabetic retinopathy (PDR). Methods We selected 16 randomized controlled trials (RCTs) performed after 2015 that had a minimum of 12 months and up to 24 months of treatment and conducted a meta-analysis with Review Manager version 5.3. Visual acuity (VA), central subfield thickness (CST) and adverse events were the outcomes selected for evaluation from the eligible studies. Results Based on 16 RCTs, we evaluated a total of 7125 patients. For PDR and severe DME with poor baseline vision, after a minimum of 12 months and up to 24 months of treatment, the aflibercept treatment group obtained better VA improvement than the focal/grid laser photocoagulation treatment group (MD=13.30; 95%CI: 13.01~13.58; P<0.001) or other treatments (ranibizumab, focal/grid laser photocoagulation, PRP, et al.) group (MD=1.10; 95%CI: 1.05~1.16; P<0.001). In addition, the aflibercept treatment group got higher CST reduction than the focal/grid laser photocoagulation treatment (MD=-33.76; 95%CI: -45.53 ~ -21.99; P<0.001) or other treatments (ranibizumab, focal/grid laser photocoagulation, et al.) group (MD=-33.76; 95%CI: -45.53 ~ -21.99; P<0.001). There was no significant difference in the overall incidence of ocular and non-ocular adverse events in each treatment group. Conclusions This meta-analysis showed that the advantages of IAI are obvious in the management of DME and PDR with poor baseline vision for long-term observation (a minimum of 12 months and up to 24 months) with both VA improvement and CST reduction. Applied IAI separately trended to be more effective than panretinal photocoagulation separately in VA improvement for PDR. More parameters should be required to assess functional and anatomic outcomes."
                    },
                    {
                        "title": "Pure Inertial Waves Radiating From Low\u2010Frequency Flows Over Large\u2010Scale Topography",
                        "abstract": "Mooring data collected at a flat\u2010topped seamount suggest the generation of pure inertial waves (PIWs; waves with a dominant frequency equal to the local inertial frequency f) by low\u2010frequency flows over large\u2010scale topography. Energetic PIWs were observed within a narrow depth range (\u223c100 m) near the seafloor at the edge of the summit. These waves could be associated with low\u2010frequency flows. A two\u2010dimensional nonhydrostatic model was used to show that the observed PIWs are most likely internal wave beams generated by low\u2010frequency flows over the seamount. Two types of PIWs were identified via observation and model. The first is a PIW that can only travel horizontally. The other propagates upward, with super\u2010inertial intrinsic frequency that is Doppler\u2010shifted by the flows to f. Nonlinear triadic interactions among waves with the frequencies [0, f, f] may transfer energy from mean flows to PIWs, promoting energy decay of geostrophic flows over large\u2010scale topography."
                    },
                    {
                        "title": "Efficacy and safety of Qiangli Dingxuan tablet combined with amlodipine besylate for essential hypertension: a randomized, double-blind, placebo-controlled, parallel-group, multicenter trial",
                        "abstract": "Background: Hypertension, a major cardiovascular risk factor, severely impacts patients\u2019 quality of life. Qiangli Dingxuan tablet (QDT) is a formally approved Chinese patent medicine, which has been widely used as an adjunctive treatment for hypertension. This study aimed to investigate the antihypertensive efficacy and safety of QDT combined with amlodipine besylate in patients with essential hypertension. Methods: In this randomized, double-blind, placebo-controlled, parallel-group, multicenter trial conducted in China, patients diagnosed with grade 1 to 2 essential hypertension were randomly assigned in a 1:1 to the treatment of QDT or placebo for 12 weeks, alongside their ongoing treatment with amlodipine besylate. The primary outcome was the change in office blood pressure (BP) from baseline to 12 weeks. In addition, safety analysis included the assessment of vital signs and laboratory values. Results: At baseline, 269 patients were randomly assigned to the QDT group (n = 133) or the placebo group (n = 136), and there were no significant differences in baseline characteristics between the two groups. The primary outcome based on the full analysis set from baseline to 12 weeks showed that the mean difference in the change of office systolic BP reduction between the two groups was 6.86 mmHg (95%CI, 4.84 to 8.88, p < 0.0001), for office diastolic BP, the mean difference in the change of office diastolic BP reduction between the two groups was 4.64 mmHg (95%CI, 3.10 to 6.18, p < 0.0001). In addition, traditional Chinese medicine symptom scores were significantly decreased in the QDT group compared with the placebo group. No severe adverse events attributable to QDT were reported. Conclusion: The combination of QDT and amlodipine besylate demonstrates superior efficacy compared to amlodipine besylate monotherapy in the management of essential hypertension. QDT shows potential as an adjunctive treatment for essential hypertension. However, further rigorous clinical trials are warranted to validate these findings. Clinical Trial Registration: [https://clinicaltrials.gov/study/NCT05521282?cond=NCT05521282&rank=1]; Identifier: [NCT05521282]"
                    },
                    {
                        "title": "Investigation on the Carbonation Behavior of Alkali-Activated Pastes Served under Windy Environments",
                        "abstract": "Most reinforced concrete structures serve under windy environments, and the carbonation resistance under that circumstance exhibits significant difference from that under the steady (no wind) environment. In this study, a windy environment was simulated using one self-developed wind tunnel, and alkali-activated slag/fly ash paste specimens were adopted for the carbonation under variant windy environments. Meanwhile, to reveal the effect of inner humidity on the carbonation, sliced alkali-activated materials (AAM) were mass-balanced first to variant humidity, and were then carbonated under a 2.5 m/s windy environment. With the assistance of computed tomography (CT), the structure of AAM at variant carbonation ages was rendered. The experimental result showed that wind is capable of promoting the exchange of moisture between the sample inside and the outer atmosphere, leading to faster carbonation as compared to that under no wind environment. When preconditioned to lower inner humidity, the carbonation rate of AAM was faster because the larger gaseous space benefited the intrusion of both CO2 and moisture. Furthermore, when preconditioned to lower humidity, the cracking extent of AAM was severer, which also contributed to the faster carbonation. Moreover, compared with ordinary Portland cement (OPC), the carbonation front on each instant 1D gray-scale value profile was broader, which suggested that the carbonation progress of AAM under windy environments was no longer controlled solely by diffusion. In addition, the gray-scale value on instant 1D profile fluctuated drastically, which verified cracking in AAM carbonated under windy environments. The current work not only deepens the understanding of the carbonation mechanism in-site (mostly under windy environments), but also helps to develop more environment-friendly construction material, with better durability performance."
                    },
                    {
                        "title": "Along-slope bottom currents driven by dissipation of internal tides in the northeastern South China Sea",
                        "abstract": "Recent mooring observations on the continental slope on the east side of the Dongsha Island in the northeastern South China Sea (SCS) showed that an along-slope bottom current can be generated when internal tides obliquely incident to the slope are dissipated near the seafloor. In this study, new mooring data collected on the south side of the Dongsha Island are used to explore the universality of internal wave driven the bottom currents and test the ability of the previous theory in estimating the along-slope current. The data show strong near-bottom energy dissipation due to the critical reflection of diurnal internal tides on the continental slope, with a time-mean depth-integrated dissipation rate of ~4.8\u00d710-3 W/m2. Because of the obliquely incident of diurnal internal tides to the slope, near-bottom dissipation of internal tides generates a southwestward along-slope current, with the maximum velocity exceeding 6 cm/s. By comparing the observations, the previous theory for internal wave induced mean flows developed by Thorpe (1999) shows a good ability to estimate the along-slope bottom current velocity. Based on the theory, as well as modelled energy dissipation on the entire continental slope in the northeastern SCS, a map is obtained to quantitatively describe the along-slope bottom flow caused by internal tide breaking on the slope."
                    },
                    {
                        "title": "Configuration of the active region for the Ge-on-Si photodetector based on carrier mobility",
                        "abstract": "The design of vertical and lateral PIN Ge-on-Si photodetectors was motivated by the disparity in electron and hole mobilities. In the case of vertical PIN junction detectors, configuring the slab region as n-type doping leads to a notable increase in the bandwidth of approximately 20 GHz compared to utilizing p-type doping for the slab. For lateral PIN junction detectors, we determined that setting the length of the n-type slab region to be 2.8 times that of the p-type slab region, based on the carrier saturation drift rate ratio, does not compromise the bandwidth. This configuration enhances the bandwidth while minimizing light absorption loss from the electrode. The proposed design in this study enhances the performance of Ge-on-Si photodetectors without adding complexity to the fabrication process. The principles applied in this study serve as instructive references for the conceptualization of other photonic or electronic devices, reinforcing the widespread applicability of these design strategies."
                    },
                    {
                        "title": "Effects of different water quality regulators on growth performance, immunologic function, and domestic water quality of GIFT tilapia",
                        "abstract": "Water quality regulation is widely recognized as a highly effective strategy for disease prevention in the field of aquaculture, and it holds significant potential for the development of sustainable aquaculture. Herein, four water quality regulators, including potassium monopersulfate (KMPS), tetrakis hydroxymethyl phosphonium sulfate (THPS), bacillus subtilis (BS), and chitosan (CS), were added to the culture water of Oreochromis niloticus (GIFT tilapia) every seven days. Subsequently, the effects of these four water quality regulators on GIFT tilapia were comprehensively evaluated by measuring the water quality index of daily growth-related performance and immune indexes of GIFT tilapia. The findings indicated that implementing the four water quality regulators resulted in a decrease in the content of ammonia nitrogen, active phosphate, nitrite, total organic carbon (TOC), and chemical oxygen demand (COD) in the water. Additionally, these regulators were found to maintain dissolved oxygen (DO) levels and pH of the water effectively. Furthermore, using these regulators demonstrated positive effects on various physiological parameters of GIFT tilapia, including improvements in final body weight, weight gain rate (WGR), specific growth rate (SGR), condition factor (CF), feed conversion ratio (FCR), spleen index (SI), hepato-somatic index (HSI), immune cell count, the activity of antioxidant-related enzymes (Nitric oxide, NO and Superoxide dismutase, SOD), and mRNA expression levels of immunity-related factors (Tumor Necrosis Factor-alpha, TNF-\u03b1 and Interleukin-1 beta, IL-1\u03b2) in the liver and spleen. Notably, the most significant improvements were observed in the groups treated with the BS and CS water quality regulators. Moreover, BS and CS groups exhibited significantly higher serum levels of albumin (ALB) and total protein (TP) (P < 0.05), whereas the other indicators showed no significant difference (P > 0.05) compared to the control group. However, the KMPS and THPS groups of GIFT tilapia exhibited significantly higher serum levels of aspartate aminotransferase (AST), alanine transaminase (ALT), creatinine (CRE) and blood urea nitrogen (BUN) (P < 0.05), whereas they exhibited significantly decreased HSI (P < 0.05). In addition, the partially pathological observations revealed the presence of cell vacuolation, nuclear shrinkage, and pyknosis within the liver. In conclusion, these four water quality regulators, mainly BS and CS, could improve the growth performance and immunity of GIFT tilapia to varying degrees by regulating the water quality and then further increasing the expression levels of immune-related factors or the activity of antioxidant-related enzymes of GIFT tilapia. On the contrary, the prolonged use of KMPS and THPS may gradually diminish their growth-enhancing properties and potentially hinder the growth of GIFT tilapia."
                    },
                    {
                        "title": "IN THE PATIENTS WITH SITUS INVERSUS TOTALIS: A CASE SERIES",
                        "abstract": "objective"
                    },
                    {
                        "title": "RNA-Seq and 16S rRNA Reveals That Tian\u2013Dong\u2013Tang\u2013Gan Powder Alleviates Environmental Stress-Induced Decline in Immune and Antioxidant Function and Gut Microbiota Dysbiosis in Litopenaeus vannami",
                        "abstract": "Ammonia stress and nitrite stress can induce immune depression and oxidative stress in Litopenaeus vannami (L. vannamei). Earlier reports showed that L. vannamei immunity, resistance to ammonia stress, and resistance to nitrite stress improved after Tian\u2013Dong\u2013Tang\u2013Gan Powder (TDTGP) treatment, but the mechanism is not clear. In this study, three thousand L. vannamei were fed different doses of TDTGP for 35 days and then subjected to ammonia and nitrite stress treatments for 72 h. Transcriptome and 16-Seq ribosomal RNA gene sequencing (16S rRNA-seq) were used to analyze hepatopancreas gene expression and changes in gut microbiota abundance in each group. The results showed that after TDTGP treatment, hepatopancreas mRNA expression levels of immunity- and antioxidant-related genes were increased, the abundance of Vibrionaceae in the gut microbiota was decreased, and the abundance of Rhodobacteraceae and Flavobacteriaceae was increased. In addition, after TDTGP treatment, the effects of ammonia and nitrite stress on the mRNA expression of Pu, cat-4, PPAF2, HO, Hsp90b1, etc. were reduced and the disruption of the gut microbiota was alleviated. In short, TDTGP can regulate the immunity and antioxidant of L. vannamei by increasing the expression levels of immunity- and antioxidant-related genes and regulating the abundance of Rhodobacteraceae and Flavobacteriaceae in the gut microbiota."
                    },
                    {
                        "title": "The motivation against change in male methamphetamine users in the compulsory detoxification setting",
                        "abstract": "Aims The study was designed to develop a measurement for the motivation for and against change in methamphetamine users in the compulsory detoxification setting. Design This is a cross-sectional study. Setting The study was carried out in a compulsory detoxification center for male drug users in China. Participants A total of 228 male methamphetamine users who had undergone the program for at least 30 days. Measurements The motivation for/against change relating to compulsory detoxification was carried out using the Likert scale. A series of questionnaires were filled out by the participants, including the Egna Minnen Betr\u00e4ffande Uppfostran for rearing style, the Barratt Impulsiveness Scale-11, the adult ADHD self-report scale, and the Pittsburgh sleep quality index. Participants were also asked to recall the withdrawal symptoms before the program and to rate their current craving levels. Findings Motivations were grouped into three factors, namely, the expectation to use drugs upon the completion of the program (factor 1), the disagreement with the compulsory setting (factor 2), and the motivation to quit drug use (factor 3). Cronbach's alpha values were 0.8037, 0.8049, and 0.6292, respectively. The structural equation model showed that the overall motivation was characterized by motivation against change rather than that for change. The overall motivation was also directly affected by the current craving level and indirectly affected by the severity of addiction, paternal authoritarian upbringing style, and ADHD traits. Conclusion This study provided a measurement of motivation for and against change in subjects with drug misconduct and suggested that the motivation against change may disclose more psychological barriers than the motivation for change."
                    },
                    {
                        "title": "Logic Diffusion for Knowledge Graph Reasoning",
                        "abstract": "Most recent works focus on answering first order logical queries to explore the knowledge graph reasoning via multi-hop logic predictions. However, existing reasoning models are limited by the circumscribed logical paradigms of training samples, which leads to a weak generalization of unseen logic. To address these issues, we propose a plug-in module called Logic Diffusion (LoD) to discover unseen queries from surroundings and achieves dynamical equilibrium between different kinds of patterns. The basic idea of LoD is relation diffusion and sampling sub-logic by random walking as well as a special training mechanism called gradient adaption. Besides, LoD is accompanied by a novel loss function to further achieve the robust logical diffusion when facing noisy data in training or testing sets. Extensive experiments on four public datasets demonstrate the superiority of mainstream knowledge graph reasoning models with LoD over state-of-the-art. Moreover, our ablation study proves the general effectiveness of LoD on the noise-rich knowledge graph."
                    }
                ]
            }
        }
    },
    "2407.10956": {
        "paper_data": {
            "title": "Spider2-V: How Far Are Multimodal Agents From Automating Data Science and Engineering Workflows?",
            "url": "http://arxiv.org/abs/2407.10956v1",
            "arxiv_id": "2407.10956",
            "authors": [
                "Ruisheng Cao",
                "Fangyu Lei",
                "Haoyuan Wu",
                "Jixuan Chen",
                "Yeqiao Fu",
                "Hongcheng Gao",
                "Xinzhuang Xiong",
                "Hanchong Zhang",
                "Yuchen Mao",
                "Wenjing Hu",
                "Tianbao Xie",
                "Hongshen Xu",
                "Danyang Zhang",
                "Sida Wang",
                "Ruoxi Sun",
                "Pengcheng Yin",
                "Caiming Xiong",
                "Ansong Ni",
                "Qian Liu",
                "Victor Zhong",
                "Lu Chen",
                "Kai Yu",
                "Tao Yu"
            ],
            "abstract": "Data science and engineering workflows often span multiple stages, from warehousing to orchestration, using tools like BigQuery, dbt, and Airbyte. As vision language models (VLMs) advance in multimodal understanding and code generation, VLM-based agents could potentially automate these workflows by generating SQL queries, Python code, and GUI operations. This automation can improve the productivity of experts while democratizing access to large-scale data analysis. In this paper, we introduce Spider2-V, the first multimodal agent benchmark focusing on professional data science and engineering workflows, featuring 494 real-world tasks in authentic computer environments and incorporating 20 enterprise-level professional applications. These tasks, derived from real-world use cases, evaluate the ability of a multimodal agent to perform data-related tasks by writing code and managing the GUI in enterprise data software systems. To balance realistic simulation with evaluation simplicity, we devote significant effort to developing automatic configurations for task setup and carefully crafting evaluation metrics for each task. Furthermore, we supplement multimodal agents with comprehensive documents of these enterprise data software systems. Our empirical evaluation reveals that existing state-of-the-art LLM/VLM-based agents do not reliably automate full data workflows (14.0% success). Even with step-by-step guidance, these agents still underperform in tasks that require fine-grained, knowledge-intensive GUI actions (16.2%) and involve remote cloud-hosted workspaces (10.6%). We hope that Spider2-V paves the way for autonomous multimodal agents to transform the automation of data science and engineering workflow. Our code and data are available at https://spider2-v.github.io.",
            "introduction": " Introduction Data science and engineering pipelines usually rely on professional data software systems such asBigQuery ,dbt, and Airbyte to acquire, process, and orchestrate large-scale data. Utilizing these enterprise systems involves writing SQL and Python code, as well as frequent and repetitive graphical user interface (GUI) controls, which can be complex even for experienced data scien- tists and engineers. With rapid advances in large language models (LLMs) and vision language models (VLMs), LLM/VLM-based autonomous agents have the potential to automate these work- \u2217Work done while interning at the University of Hong Kong. Preprint. Under review.arXiv:2407.10956v1  [cs.AI]  15 Jul 2024datawarehousing dataingestion datatransformationdataorchestration datavisualization Task1:Load data underthecurrentGoogle Drive folder into anewtable\u201cdata1\u201doftheopenedBigQuerydataset. GUIcontrolonDriveUI(copydatalink)GUIcontrolonBigQueryUI(createnewtable)GUIcontrolonpopuppanel(fillininformation) Task2:Savetop 20 dramatic movies since 2000 fromSnowflakedatabaseIMDBintofile\"top20movies.csv\"onDesktopwithdetailedrequirementsintheopened.txtfile. GUIcontrolonSnowflakeUI(createnewworksheet) writeSQL GUIcontrolcross apps(rename output file) Figure 1: Spider2-V is a multimodal agent benchmark spanning across complete data science and engineering workflows ( e.g., two task examples in the Figure above). It involves various professional enterprise-level applications and includes intensive GUI controls apart from code writing throughout the real-time multi-turn interaction with an executable computer environment. flows [ 37,32], enhancing productivity for data scientists and engineers [ 38,16] while democratizing access to large-scale data [15, 40]. Previous studies on data agents focused mainly on daily life data processing and analysis by generating code or API calls [ 42,9,4], neglecting other crucial stages of data science and engineering ( e.g., data ingestion and integration) using enterprise applications ( e.g.,Snowflake ,Airflow , and Dagster ). Additionally, to complete data workflows, data scientists and engineers often need to navigate multiple professional data systems, combining code writing with intensive GUI controls, such as navigating web pages and clicking buttons [ 5,45]. However, there is currently no benchmark that integrates both code generation and GUI controls for professional data science and engineering. To address this gap, we propose Spider2-V, the first multimodal agent benchmark covering the entire data science and engineering workflow, involving 494real-world tasks in a real-time executable computer environment and 20professional enterprise data software. Spider2-V aims to evaluate a multimodal agent\u2019s ability to perform professional data-related tasks by writing code and managing the GUI in enterprise data software systems, including data warehousing ( e.g.,BigQuery ), data ingestion and integration ( e.g.,Airbyte ), data transformation ( e.g.,dbt), data analysis and visualization ( e.g., Superset ), and data orchestration ( e.g.,Dagster ). These tasks are derived from real-world practices, such as official tutorials on professional applications and open-source data engineering projects (with two task examples presented in Figure 1). We also supplement retrieval-augmented agents with official documentation and tutorials of these software systems to assess their capability to generalize and learn from these resources. Each task in Spider2-V is defined within an executable computer environment based on OS- WORLD [34], which allows multimodal agents to simulate human actions ( e.g., typing code or clicking buttons) in a realistic setting. Specifically, a multimodal agent can observe real-time image- style screenshots and text-style accessibility tree of professional data applications in the current workflow and execute its predicted actions in dynamic multi-round interaction with the computer. This environment is connected to the real-world Internet, allowing the inclusion of professional software requiring authentic user accounts ( e.g.,Snowflake ). To ensure reproducible and reliable experiments (see Table 7), we compare the performances using different formats of retrieved documents on a subset ( 130task samples) of Spider2-V. And pure text format outperforms the",
            "references": [
                {
                    "title": "WONDERBREAD: A Benchmark for Evaluating Multimodal Foundation Models on Business Process Management Tasks",
                    "abstract": "Existing ML benchmarks lack the depth and diversity of annotations needed for evaluating models on business process management (BPM) tasks. BPM is the practice of documenting, measuring, improving, and automating enterprise workflows. However, research has focused almost exclusively on one task - full end-to-end automation using agents based on multimodal foundation models (FMs) like GPT-4. This focus on automation ignores the reality of how most BPM tools are applied today - simply documenting the relevant workflow takes 60% of the time of the typical process optimization project. To address this gap we present WONDERBREAD, the first benchmark for evaluating multimodal FMs on BPM tasks beyond automation. Our contributions are: (1) a dataset containing 2928 documented workflow demonstrations; (2) 6 novel BPM tasks sourced from real-world applications ranging from workflow documentation to knowledge transfer to process improvement; and (3) an automated evaluation harness. Our benchmark shows that while state-of-the-art FMs can automatically generate documentation (e.g. recalling 88% of the steps taken in a video demonstration of a workflow), they struggle to re-apply that knowledge towards finer-grained validation of workflow completion (F1<0.3). We hope WONDERBREAD encourages the development of more\"human-centered\"AI tooling for enterprise applications and furthers the exploration of multimodal FMs for the broader universe of BPM tasks. We publish our dataset and experiments here: https://github.com/HazyResearch/wonderbread"
                },
                {
                    "title": "SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering",
                    "abstract": "Language model (LM) agents are increasingly being used to automate complicated tasks in digital environments. Just as humans benefit from powerful software applications, such as integrated development environments, for complex tasks like software engineering, we posit that LM agents represent a new category of end users with their own needs and abilities, and would benefit from specially-built interfaces to the software they use. We investigate how interface design affects the performance of language model agents. As a result of this exploration, we introduce SWE-agent: a system that facilitates LM agents to autonomously use computers to solve software engineering tasks. SWE-agent's custom agent-computer interface (ACI) significantly enhances an agent's ability to create and edit code files, navigate entire repositories, and execute tests and other programs. We evaluate SWE-agent on SWE-bench and HumanEvalFix, achieving state-of-the-art performance on both with a pass@1 rate of 12.5% and 87.7%, respectively, far exceeding the previous state-of-the-art achieved with non-interactive LMs. Finally, we provide insight on how the design of the ACI can impact agents' behavior and performance."
                },
                {
                    "title": "WorkArena: How Capable Are Web Agents at Solving Common Knowledge Work Tasks?",
                    "abstract": "We study the use of large language model-based agents for interacting with software via web browsers. Unlike prior work, we focus on measuring the agents' ability to perform tasks that span the typical daily work of knowledge workers utilizing enterprise software systems. To this end, we propose WorkArena, a remote-hosted benchmark of 33 tasks based on the widely-used ServiceNow platform. We also introduce BrowserGym, an environment for the design and evaluation of such agents, offering a rich set of actions as well as multimodal observations. Our empirical evaluation reveals that while current agents show promise on WorkArena, there remains a considerable gap towards achieving full task automation. Notably, our analysis uncovers a significant performance disparity between open and closed-source LLMs, highlighting a critical area for future exploration and development in the field."
                },
                {
                    "title": "Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context",
                    "abstract": "In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content."
                },
                {
                    "title": "OmniACT: A Dataset and Benchmark for Enabling Multimodal Generalist Autonomous Agents for Desktop and Web",
                    "abstract": "For decades, human-computer interaction has fundamentally been manual. Even today, almost all productive work done on the computer necessitates human input at every step. Autonomous virtual agents represent an exciting step in automating many of these menial tasks. Virtual agents would empower users with limited technical proficiency to harness the full possibilities of computer systems. They could also enable the efficient streamlining of numerous computer tasks, ranging from calendar management to complex travel bookings, with minimal human intervention. In this paper, we introduce OmniACT, the first-of-a-kind dataset and benchmark for assessing an agent's capability to generate executable programs to accomplish computer tasks. Our scope extends beyond traditional web automation, covering a diverse range of desktop applications. The dataset consists of fundamental tasks such as\"Play the next song\", as well as longer horizon tasks such as\"Send an email to John Doe mentioning the time and place to meet\". Specifically, given a pair of screen image and a visually-grounded natural language task, the goal is to generate a script capable of fully executing the task. We run several strong baseline language model agents on our benchmark. The strongest baseline, GPT-4, performs the best on our benchmark However, its performance level still reaches only 15% of the human proficiency in generating executable scripts capable of completing the task, demonstrating the challenge of our task for conventional web agents. Our benchmark provides a platform to measure and evaluate the progress of language model agents in automating computer tasks and motivates future work towards building multimodal models that bridge large language models and the visual grounding of computer screens."
                },
                {
                    "title": "ARKS: Active Retrieval in Knowledge Soup for Code Generation",
                    "abstract": "Recently the retrieval-augmented generation (RAG) paradigm has raised much attention for its potential in incorporating external knowledge into large language models (LLMs) without further training. While widely explored in natural language applications, its utilization in code generation remains under-explored. In this paper, we introduce Active Retrieval in Knowledge Soup (ARKS), an advanced strategy for generalizing large language models for code. In contrast to relying on a single source, we construct a knowledge soup integrating web search, documentation, execution feedback, and evolved code snippets. We employ an active retrieval strategy that iteratively refines the query and updates the knowledge soup. To assess the performance of ARKS, we compile a new benchmark comprising realistic coding problems associated with frequently updated libraries and long-tail programming languages. Experimental results on ChatGPT and CodeLlama demonstrate a substantial improvement in the average execution accuracy of ARKS on LLMs. The analysis confirms the effectiveness of our proposed knowledge soup and active retrieval strategies, offering rich insights into the construction of effective retrieval-augmented code generation (RACG) pipelines. Our model, code, and data are available at https://arks-codegen.github.io."
                },
                {
                    "title": "OS-Copilot: Towards Generalist Computer Agents with Self-Improvement",
                    "abstract": "Autonomous interaction with the computer has been a longstanding challenge with great potential, and the recent proliferation of large language models (LLMs) has markedly accelerated progress in building digital agents. However, most of these agents are designed to interact with a narrow domain, such as a specific software or website. This narrow focus constrains their applicability for general computer tasks. To this end, we introduce OS-Copilot, a framework to build generalist agents capable of interfacing with comprehensive elements in an operating system (OS), including the web, code terminals, files, multimedia, and various third-party applications. We use OS-Copilot to create FRIDAY, a self-improving embodied agent for automating general computer tasks. On GAIA, a general AI assistants benchmark, FRIDAY outperforms previous methods by 35%, showcasing strong generalization to unseen applications via accumulated skills from previous tasks. We also present numerical and quantitative evidence that FRIDAY learns to control and self-improve on Excel and Powerpoint with minimal supervision. Our OS-Copilot framework and empirical findings provide infrastructure and insights for future research toward more capable and general-purpose computer agents."
                },
                {
                    "title": "WebLINX: Real-World Website Navigation with Multi-Turn Dialogue",
                    "abstract": "We propose the problem of conversational web navigation, where a digital agent controls a web browser and follows user instructions to solve real-world tasks in a multi-turn dialogue fashion. To support this problem, we introduce WEBLINX - a large-scale benchmark of 100K interactions across 2300 expert demonstrations of conversational web navigation. Our benchmark covers a broad range of patterns on over 150 real-world websites and can be used to train and evaluate agents in diverse scenarios. Due to the magnitude of information present, Large Language Models (LLMs) cannot process entire web pages in real-time. To solve this bottleneck, we design a retrieval-inspired model that efficiently prunes HTML pages by ranking relevant elements. We use the selected elements, along with screenshots and action history, to assess a variety of models for their ability to replicate human behavior when navigating the web. Our experiments span from small text-only to proprietary multimodal LLMs. We find that smaller finetuned decoders surpass the best zero-shot LLMs (including GPT-4V), but also larger finetuned multimodal models which were explicitly pretrained on screenshots. However, all finetuned models struggle to generalize to unseen websites. Our findings highlight the need for large multimodal models that can generalize to novel settings. Our code, data and models are available for research: https://mcgill-nlp.github.io/weblinx"
                },
                {
                    "title": "Mobile-Agent: Autonomous Multi-Modal Mobile Device Agent with Visual Perception",
                    "abstract": "Mobile device agent based on Multimodal Large Language Models (MLLM) is becoming a popular application. In this paper, we introduce Mobile-Agent, an autonomous multi-modal mobile device agent. Mobile-Agent first leverages visual perception tools to accurately identify and locate both the visual and textual elements within the app's front-end interface. Based on the perceived vision context, it then autonomously plans and decomposes the complex operation task, and navigates the mobile Apps through operations step by step. Different from previous solutions that rely on XML files of Apps or mobile system metadata, Mobile-Agent allows for greater adaptability across diverse mobile operating environments in a vision-centric way, thereby eliminating the necessity for system-specific customizations. To assess the performance of Mobile-Agent, we introduced Mobile-Eval, a benchmark for evaluating mobile device operations. Based on Mobile-Eval, we conducted a comprehensive evaluation of Mobile-Agent. The experimental results indicate that Mobile-Agent achieved remarkable accuracy and completion rates. Even with challenging instructions, such as multi-app operations, Mobile-Agent can still complete the requirements. Code and model will be open-sourced at https://github.com/X-PLUG/MobileAgent."
                },
                {
                    "title": "InfiAgent-DABench: Evaluating Agents on Data Analysis Tasks",
                    "abstract": "In this paper, we introduce InfiAgent-DABench, the first benchmark specifically designed to evaluate LLM-based agents on data analysis tasks. These tasks require agents to end-to-end solving complex tasks by interacting with an execution environment. This benchmark contains DAEval, a dataset consisting of 257 data analysis questions derived from 52 CSV files, and an agent framework which incorporates LLMs to serve as data analysis agents for both serving and evaluation. Since data analysis questions are often open-ended and hard to evaluate without human supervision, we adopt a format-prompting technique to convert each question into a closed-form format so that they can be automatically evaluated. Our extensive benchmarking of 34 LLMs uncovers the current challenges encountered in data analysis tasks. In addition, building on top of our agent framework, we develop a specialized agent, DAAgent, which surpasses GPT-3.5 by 3.9% on DABench. Evaluation datasets and toolkits for InfiAgent-DABench are released at https://github.com/InfiAgent/InfiAgent ."
                },
                {
                    "title": "Mixtral of Experts",
                    "abstract": "We introduce Mixtral 8x7B, a Sparse Mixture of Experts (SMoE) language model. Mixtral has the same architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks (i.e. experts). For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, each token has access to 47B parameters, but only uses 13B active parameters during inference. Mixtral was trained with a context size of 32k tokens and it outperforms or matches Llama 2 70B and GPT-3.5 across all evaluated benchmarks. In particular, Mixtral vastly outperforms Llama 2 70B on mathematics, code generation, and multilingual benchmarks. We also provide a model fine-tuned to follow instructions, Mixtral 8x7B - Instruct, that surpasses GPT-3.5 Turbo, Claude-2.1, Gemini Pro, and Llama 2 70B - chat model on human benchmarks. Both the base and instruct models are released under the Apache 2.0 license."
                },
                {
                    "title": "AppAgent: Multimodal Agents as Smartphone Users",
                    "abstract": "Recent advancements in large language models (LLMs) have led to the creation of intelligent agents capable of performing complex tasks. This paper introduces a novel LLM-based multimodal agent framework designed to operate smartphone applications. Our framework enables the agent to operate smartphone applications through a simplified action space, mimicking human-like interactions such as tapping and swiping. This novel approach bypasses the need for system back-end access, thereby broadening its applicability across diverse apps. Central to our agent's functionality is its innovative learning method. The agent learns to navigate and use new apps either through autonomous exploration or by observing human demonstrations. This process generates a knowledge base that the agent refers to for executing complex tasks across different applications. To demonstrate the practicality of our agent, we conducted extensive testing over 50 tasks in 10 different applications, including social media, email, maps, shopping, and sophisticated image editing tools. The results affirm our agent's proficiency in handling a diverse array of high-level tasks."
                },
                {
                    "title": "ASSISTGUI: Task-Oriented Desktop Graphical User Interface Automation",
                    "abstract": "Graphical User Interface (GUI) automation holds significant promise for assisting users with complex tasks, thereby boosting human productivity. Existing works leveraging Large Language Model (LLM) or LLM-based AI agents have shown capabilities in automating tasks on Android and Web platforms. However, these tasks are primarily aimed at simple device usage and entertainment operations. This paper presents a novel benchmark, AssistGUI, to evaluate whether models are capable of manipulating the mouse and keyboard on the Windows platform in response to user-requested tasks. We carefully collected a set of 100 tasks from nine widely-used software applications, such as, After Effects and MS Word, each accompanied by the necessary project files for better evaluation. Moreover, we propose an advanced Actor-Critic Embodied Agent framework, which incorporates a sophisticated GUI parser driven by an LLM-agent and an enhanced reasoning mechanism adept at handling lengthy procedural tasks. Our experimental results reveal that our GUI Parser and Reasoning mechanism outshine existing methods in performance. Nevertheless, the potential remains substantial, with the best model attaining only a 46% success rate on our benchmark. We conclude with a thorough analysis of the current methods' limitations, setting the stage for future breakthroughs in this domain."
                },
                {
                    "title": "Retrieval-Augmented Generation for Large Language Models: A Survey",
                    "abstract": "Large Language Models (LLMs) showcase impressive capabilities but encounter challenges like hallucination, outdated knowledge, and non-transparent, untraceable reasoning processes. Retrieval-Augmented Generation (RAG) has emerged as a promising solution by incorporating knowledge from external databases. This enhances the accuracy and credibility of the generation, particularly for knowledge-intensive tasks, and allows for continuous knowledge updates and integration of domain-specific information. RAG synergistically merges LLMs' intrinsic knowledge with the vast, dynamic repositories of external databases. This comprehensive review paper offers a detailed examination of the progression of RAG paradigms, encompassing the Naive RAG, the Advanced RAG, and the Modular RAG. It meticulously scrutinizes the tripartite foundation of RAG frameworks, which includes the retrieval, the generation and the augmentation techniques. The paper highlights the state-of-the-art technologies embedded in each of these critical components, providing a profound understanding of the advancements in RAG systems. Furthermore, this paper introduces up-to-date evaluation framework and benchmark. At the end, this article delineates the challenges currently faced and points out prospective avenues for research and development."
                },
                {
                    "title": "SWE-bench: Can Language Models Resolve Real-World GitHub Issues?",
                    "abstract": "Language models have outpaced our ability to evaluate them effectively, but for their future development it is essential to study the frontier of their capabilities. We find real-world software engineering to be a rich, sustainable, and challenging testbed for evaluating the next generation of language models. To this end, we introduce SWE-bench, an evaluation framework consisting of $2,294$ software engineering problems drawn from real GitHub issues and corresponding pull requests across $12$ popular Python repositories. Given a codebase along with a description of an issue to be resolved, a language model is tasked with editing the codebase to address the issue. Resolving issues in SWE-bench frequently requires understanding and coordinating changes across multiple functions, classes, and even files simultaneously, calling for models to interact with execution environments, process extremely long contexts and perform complex reasoning that goes far beyond traditional code generation tasks. Our evaluations show that both state-of-the-art proprietary models and our fine-tuned model SWE-Llama can resolve only the simplest issues. The best-performing model, Claude 2, is able to solve a mere $1.96$% of the issues. Advances on SWE-bench represent steps towards LMs that are more practical, intelligent, and autonomous."
                },
                {
                    "title": "Qwen Technical Report",
                    "abstract": "Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural language processing tasks that were previously thought to be exclusive to humans. In this work, we introduce Qwen, the first installment of our large language model series. Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts. It includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques. The base language models consistently demonstrate superior performance across a multitude of downstream tasks, and the chat models, particularly those trained using Reinforcement Learning from Human Feedback (RLHF), are highly competitive. The chat models possess advanced tool-use and planning capabilities for creating agent applications, showcasing impressive performance even when compared to bigger models on complex tasks like utilizing a code interpreter. Furthermore, we have developed coding-specialized models, Code-Qwen and Code-Qwen-Chat, as well as mathematics-focused models, Math-Qwen-Chat, which are built upon base language models. These models demonstrate significantly improved performance in comparison with open-source models, and slightly fall behind the proprietary models."
                },
                {
                    "title": "WebArena: A Realistic Web Environment for Building Autonomous Agents",
                    "abstract": "With advances in generative AI, there is now potential for autonomous agents to manage daily tasks via natural language commands. However, current agents are primarily created and tested in simplified synthetic environments, leading to a disconnect with real-world scenarios. In this paper, we build an environment for language-guided agents that is highly realistic and reproducible. Specifically, we focus on agents that perform tasks on the web, and create an environment with fully functional websites from four common domains: e-commerce, social forum discussions, collaborative software development, and content management. Our environment is enriched with tools (e.g., a map) and external knowledge bases (e.g., user manuals) to encourage human-like task-solving. Building upon our environment, we release a set of benchmark tasks focusing on evaluating the functional correctness of task completions. The tasks in our benchmark are diverse, long-horizon, and designed to emulate tasks that humans routinely perform on the internet. We experiment with several baseline agents, integrating recent techniques such as reasoning before acting. The results demonstrate that solving complex tasks is challenging: our best GPT-4-based agent only achieves an end-to-end task success rate of 14.41%, significantly lower than the human performance of 78.24%. These results highlight the need for further development of robust agents, that current state-of-the-art large language models are far from perfect performance in these real-life tasks, and that WebArena can be used to measure such progress."
                },
                {
                    "title": "Android in the Wild: A Large-Scale Dataset for Android Device Control",
                    "abstract": "There is a growing interest in device-control systems that can interpret human natural language instructions and execute them on a digital device by directly controlling its user interface. We present a dataset for device-control research, Android in the Wild (AITW), which is orders of magnitude larger than current datasets. The dataset contains human demonstrations of device interactions, including the screens and actions, and corresponding natural language instructions. It consists of 715k episodes spanning 30k unique instructions, four versions of Android (v10-13),and eight device types (Pixel 2 XL to Pixel 6) with varying screen resolutions. It contains multi-step tasks that require semantic understanding of language and visual context. This dataset poses a new challenge: actions available through the user interface must be inferred from their visual appearance. And, instead of simple UI element-based actions, the action space consists of precise gestures (e.g., horizontal scrolls to operate carousel widgets). We organize our dataset to encourage robustness analysis of device-control systems, i.e., how well a system performs in the presence of new task descriptions, new applications, or new platform versions. We develop two agents and report performance across the dataset. The dataset is available at https://github.com/google-research/google-research/tree/master/android_in_the_wild."
                },
                {
                    "title": "InterCode: Standardizing and Benchmarking Interactive Coding with Execution Feedback",
                    "abstract": "Humans write code in a fundamentally interactive manner and rely on constant execution feedback to correct errors, resolve ambiguities, and decompose tasks. While LLMs have recently exhibited promising coding capabilities, current coding benchmarks mostly consider a static instruction-to-code sequence transduction process, which has the potential for error propagation and a disconnect between the generated code and its final execution environment. To address this gap, we introduce InterCode, a lightweight, flexible, and easy-to-use framework of interactive coding as a standard reinforcement learning (RL) environment, with code as actions and execution feedback as observations. Our framework is language and platform agnostic, uses self-contained Docker environments to provide safe and reproducible execution, and is compatible out-of-the-box with traditional seq2seq coding methods, while enabling the development of new methods for interactive code generation. We use InterCode to create three interactive code environments with Bash, SQL, and Python as action spaces, leveraging data from the static NL2Bash, Spider, and MBPP datasets. We demonstrate InterCode's viability as a testbed by evaluating multiple state-of-the-art LLMs configured with different prompting strategies such as ReAct and Plan&Solve. Our results showcase the benefits of interactive code generation and demonstrate that InterCode can serve as a challenging benchmark for advancing code understanding and generation capabilities. InterCode is designed to be easily extensible and can even be used to create new tasks such as Capture the Flag, a popular coding puzzle that is inherently multi-step and involves multiple programming languages. Project site with code and data: https://intercode-benchmark.github.io"
                },
                {
                    "title": "Mind2Web: Towards a Generalist Agent for the Web",
                    "abstract": "We introduce Mind2Web, the first dataset for developing and evaluating generalist agents for the web that can follow language instructions to complete complex tasks on any website. Existing datasets for web agents either use simulated websites or only cover a limited set of websites and tasks, thus not suitable for generalist web agents. With over 2,000 open-ended tasks collected from 137 websites spanning 31 domains and crowdsourced action sequences for the tasks, Mind2Web provides three necessary ingredients for building generalist web agents: 1) diverse domains, websites, and tasks, 2) use of real-world websites instead of simulated and simplified ones, and 3) a broad spectrum of user interaction patterns. Based on Mind2Web, we conduct an initial exploration of using large language models (LLMs) for building generalist web agents. While the raw HTML of real-world websites are often too large to be fed to LLMs, we show that first filtering it with a small LM significantly improves the effectiveness and efficiency of LLMs. Our solution demonstrates a decent level of performance, even on websites or entire domains the model has never seen before, but there is still a substantial room to improve towards truly generalizable agents. We open-source our dataset, model implementation, and trained models (https://osu-nlp-group.github.io/Mind2Web) to facilitate further research on building a generalist agent for the web."
                },
                {
                    "title": "SheetCopilot: Bringing Software Productivity to the Next Level through Large Language Models",
                    "abstract": "Computer end users have spent billions of hours completing daily tasks like tabular data processing and project timeline scheduling. Most of these tasks are repetitive and error-prone, yet most end users lack the skill to automate these burdensome works. With the advent of large language models (LLMs), directing software with natural language user requests become a reachable goal. In this work, we propose a SheetCopilot agent that takes natural language task and control spreadsheet to fulfill the requirements. We propose a set of atomic actions as an abstraction of spreadsheet software functionalities. We further design a state machine-based task planning framework for LLMs to robustly interact with spreadsheets. We curate a representative dataset containing 221 spreadsheet control tasks and establish a fully automated evaluation pipeline for rigorously benchmarking the ability of LLMs in software control tasks. Our SheetCopilot correctly completes 44.3\\% of tasks for a single generation, outperforming the strong code generation baseline by a wide margin. Our project page:https://sheetcopilot.github.io/."
                },
                {
                    "title": "On the Tool Manipulation Capability of Open-source Large Language Models",
                    "abstract": "Recent studies on software tool manipulation with large language models (LLMs) mostly rely on closed model APIs. The industrial adoption of these models is substantially constrained due to the security and robustness risks in exposing information to closed LLM API services. In this paper, we ask can we enhance open-source LLMs to be competitive to leading closed LLM APIs in tool manipulation, with practical amount of human supervision. By analyzing common tool manipulation failures, we first demonstrate that open-source LLMs may require training with usage examples, in-context demonstration and generation style regulation to resolve failures. These insights motivate us to revisit classical methods in LLM literature, and demonstrate that we can adapt them as model alignment with programmatic data generation, system prompts and in-context demonstration retrievers to enhance open-source LLMs for tool manipulation. To evaluate these techniques, we create the ToolBench, a tool manipulation benchmark consisting of diverse software tools for real-world tasks. We demonstrate that our techniques can boost leading open-source LLMs by up to 90% success rate, showing capabilities competitive to OpenAI GPT-4 in 4 out of 8 ToolBench tasks. We show that such enhancement typically requires about one developer day to curate data for each tool, rendering a recipe with practical amount of human supervision."
                },
                {
                    "title": "Tool Learning with Foundation Models",
                    "abstract": "Humans possess an extraordinary ability to create and utilize tools, allowing them to overcome physical limitations and explore new frontiers. With the advent of foundation models, AI systems have the potential to be equally adept in tool use as humans. This paradigm, i.e., tool learning with foundation models, combines the strengths of specialized tools and foundation models to achieve enhanced accuracy, efficiency, and automation in problem-solving. Despite its immense potential, there is still a lack of a comprehensive understanding of key challenges, opportunities, and future endeavors in this field. To this end, we present a systematic investigation of tool learning in this paper. We first introduce the background of tool learning, including its cognitive origins, the paradigm shift of foundation models, and the complementary roles of tools and models. Then we recapitulate existing tool learning research into tool-augmented and tool-oriented learning. We formulate a general tool learning framework: starting from understanding the user instruction, models should learn to decompose a complex task into several subtasks, dynamically adjust their plan through reasoning, and effectively conquer each sub-task by selecting appropriate tools. We also discuss how to train models for improved tool-use capabilities and facilitate the generalization in tool learning. Considering the lack of a systematic tool learning evaluation in prior works, we experiment with 18 representative tools and show the potential of current foundation models in skillfully utilizing tools. Finally, we discuss several open problems that require further investigation for tool learning. In general, we hope this paper could inspire future research in integrating tools with foundation models."
                },
                {
                    "title": "Reflexion: language agents with verbal reinforcement learning",
                    "abstract": "Large language models (LLMs) have been increasingly used to interact with external environments (e.g., games, compilers, APIs) as goal-driven agents. However, it remains challenging for these language agents to quickly and efficiently learn from trial-and-error as traditional reinforcement learning methods require extensive training samples and expensive model fine-tuning. We propose Reflexion, a novel framework to reinforce language agents not by updating weights, but instead through linguistic feedback. Concretely, Reflexion agents verbally reflect on task feedback signals, then maintain their own reflective text in an episodic memory buffer to induce better decision-making in subsequent trials. Reflexion is flexible enough to incorporate various types (scalar values or free-form language) and sources (external or internally simulated) of feedback signals, and obtains significant improvements over a baseline agent across diverse tasks (sequential decision-making, coding, language reasoning). For example, Reflexion achieves a 91% pass@1 accuracy on the HumanEval coding benchmark, surpassing the previous state-of-the-art GPT-4 that achieves 80%. We also conduct ablation and analysis studies using different feedback signals, feedback incorporation methods, and agent types, and provide insights into how they affect performance."
                },
                {
                    "title": "Natural Language to Code Generation in Interactive Data Science Notebooks",
                    "abstract": "Computational notebooks, such as Jupyter notebooks, are interactive computing environments that are ubiquitous among data scientists to perform data wrangling and analytic tasks. To measure the performance of AI pair programmers that automatically synthesize programs for those tasks given natural language (NL) intents from users, we build ARCADE, a benchmark of 1078 code generation problems using the pandas data analysis framework in data science notebooks. ARCADE features multiple rounds of NL-to-code problems from the same notebook. It requires a model to understand rich multi-modal contexts, such as existing notebook cells and their execution states as well as previous turns of interaction. To establish a strong baseline on this challenging task, we develop PaChiNCo, a 62B code language model (LM) for Python computational notebooks, which significantly outperforms public code LMs. Finally, we explore few-shot prompting strategies to elicit better code with step-by-step decomposition and NL explanation, showing the potential to improve the diversity and explainability of model predictions. Arcade is publicly available at https://github.com/google-research/arcade-nl2code/."
                },
                {
                    "title": "DS-1000: A Natural and Reliable Benchmark for Data Science Code Generation",
                    "abstract": "We introduce DS-1000, a code generation benchmark with a thousand data science problems spanning seven Python libraries, such as NumPy and Pandas. Compared to prior works, DS-1000 incorporates three core features. First, our problems reflect diverse, realistic, and practical use cases since we collected them from StackOverflow. Second, our automatic evaluation is highly specific (reliable) -- across all Codex-002-predicted solutions that our evaluation accept, only 1.8% of them are incorrect; we achieve this with multi-criteria metrics, checking both functional correctness by running test cases and surface-form constraints by restricting API usages or keywords. Finally, we proactively defend against memorization by slightly modifying our problems to be different from the original StackOverflow source; consequently, models cannot answer them correctly by memorizing the solutions from pre-training. The current best public system (Codex-002) achieves 43.3% accuracy, leaving ample room for improvement. We release our benchmark at https://ds1000-code-gen.github.io."
                },
                {
                    "title": "WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents",
                    "abstract": "Existing benchmarks for grounding language in interactive environments either lack real-world linguistic elements, or prove difficult to scale up due to substantial human involvement in the collection of data or feedback signals. To bridge this gap, we develop WebShop -- a simulated e-commerce website environment with $1.18$ million real-world products and $12,087$ crowd-sourced text instructions. Given a text instruction specifying a product requirement, an agent needs to navigate multiple types of webpages and issue diverse actions to find, customize, and purchase an item. WebShop provides several challenges for language grounding including understanding compositional instructions, query (re-)formulation, comprehending and acting on noisy text in webpages, and performing strategic exploration. We collect over $1,600$ human demonstrations for the task, and train and evaluate a diverse range of agents using reinforcement learning, imitation learning, and pre-trained image and language models. Our best model achieves a task success rate of $29\\%$, which outperforms rule-based heuristics ($9.6\\%$) but is far lower than human expert performance ($59\\%$). We also analyze agent and human trajectories and ablate various model components to provide insights for developing future agents with stronger language understanding and decision making abilities. Finally, we show that agents trained on WebShop exhibit non-trivial sim-to-real transfer when evaluated on amazon.com and ebay.com, indicating the potential value of WebShop in developing practical web-based agents that can operate in the wild."
                },
                {
                    "title": "Nerfies: Deformable Neural Radiance Fields",
                    "abstract": "We present the first method capable of photorealistically reconstructing deformable scenes using photos/videos captured casually from mobile phones. Our approach augments neural radiance fields (NeRF) by optimizing an additional continuous volumetric deformation field that warps each observed point into a canonical 5D NeRF. We observe that these NeRF-like deformation fields are prone to local minima, and propose a coarse-to-fine optimization method for coordinate-based models that allows for more robust optimization. By adapting principles from geometry processing and physical simulation to NeRF-like models, we propose an elastic regularization of the deformation field that further improves robustness. We show that our method can turn casually captured selfie photos/videos into deformable NeRF models that allow for photorealistic renderings of the subject from arbitrary viewpoints, which we dub \"nerfies.\" We evaluate our method by collecting time-synchronized data using a rig with two mobile phones, yielding train/validation images of the same pose at different viewpoints. We show that our method faithfully reconstructs non-rigidly deforming scenes and reproduces unseen views with high fidelity."
                },
                {
                    "title": "Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task",
                    "abstract": "We present Spider, a large-scale complex and cross-domain semantic parsing and text-to-SQL dataset annotated by 11 college students. It consists of 10,181 questions and 5,693 unique complex SQL queries on 200 databases with multiple tables covering 138 different domains. We define a new complex and cross-domain semantic parsing and text-to-SQL task so that different complicated SQL queries and databases appear in train and test sets. In this way, the task requires the model to generalize well to both new SQL queries and new database schemas. Therefore, Spider is distinct from most of the previous semantic parsing tasks because they all use a single database and have the exact same program in the train set and the test set. We experiment with various state-of-the-art models and the best model achieves only 9.7% exact matching accuracy on a database split setting. This shows that Spider presents a strong challenge for future research. Our dataset and task with the most recent updates are publicly available at https://yale-lily.github.io/seq2sql/spider."
                },
                {
                    "title": "Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration",
                    "abstract": "Reinforcement learning (RL) agents improve through trial-and-error, but when reward is sparse and the agent cannot discover successful action sequences, learning stagnates. This has been a notable problem in training deep RL agents to perform web-based tasks, such as booking flights or replying to emails, where a single mistake can ruin the entire sequence of actions. A common remedy is to \"warm-start\" the agent by pre-training it to mimic expert demonstrations, but this is prone to overfitting. Instead, we propose to constrain exploration using demonstrations. From each demonstration, we induce high-level \"workflows\" which constrain the allowable actions at each time step to be similar to those in the demonstration (e.g., \"Step 1: click on a textbox; Step 2: enter some text\"). Our exploration policy then learns to identify successful workflows and samples actions that satisfy these workflows. Workflows prune out bad exploration directions and accelerate the agent's ability to discover rewards. We use our approach to train a novel neural policy designed to handle the semi-structured nature of websites, and evaluate on a suite of web tasks, including the recent World of Bits benchmark. We achieve new state-of-the-art results, and show that workflow-guided exploration improves sample efficiency over behavioral cloning by more than 100x."
                },
                {
                    "title": "World of Bits: An Open-Domain Platform for Web-Based Agents",
                    "abstract": "While simulated game environments have greatly accelerated research in reinforcement learning, existing environments lack the open-domain re-alism of tasks in computer vision or natural language processing, which operate on artifacts created by humans in natural, organic settings. To foster reinforcement learning research in such settings, we introduce the World of Bits (WoB), a platform in which agents complete tasks on the Internet by performing low-level keyboard and mouse actions. The two main challenges are: (i) to curate a diverse set of natural web-based tasks, and (ii) to ensure that these tasks have a well-de\ufb01ned reward structure and are re-producible despite the transience of the web. To tackle this, we develop a methodology in which crowdworkers create tasks de\ufb01ned by natural language questions and provide demonstrations of how to answer the question on real websites us-ing keyboard and mouse; HTTP traf\ufb01c is cached to create a reproducible of\ufb02ine approximation of the website. Finally, we show that agents trained via behavioral cloning and reinforcement learning can complete a range of web-based tasks."
                },
                {
                    "title": "WHO Technical Report",
                    "abstract": "The Feather River Coordinated Resource Management Group (FR-CRM) has been restoring channel/ meadow/ floodplain systems in the Feather River watershed since 1985. Project and watershed-wide monitoring has shown multiple benefits of this type of work. With the concern over global climate change, the group wanted to measure the carbon sequestered in project areas. No protocol was found to measure carbon stores in native Sierra Nevada meadows. Plumas County funded the FR-CRM to conduct a pilot study to develop such a protocol. The sampling protocol included discrete sampling at consistent soil depths to determine the vertical distribution of carbon. A Technical Advisory Committee developed and refined a multi-project sampling protocol for three restored meadows and three un-restored meadows. Data from the un-restored meadows will also provide base-line data for before and after restoration comparisons. Initial data analysis indicates that restored meadows contain twice as much total carbon as degraded meadows; on average approximately 40 tonnes more carbon per acre. Virtually all of the additional carbon in restored meadows occurs in the soil, and is thus protected from loss via grazing, haying, wildfire, etc. Introduction In 1994 the Feather River Coordinated Resource Management (FR-CRM) group shifted its stream restoration approach from bank stabilization to landscape function. Called meadow re-watering, this approach entails returning the incised stream channel to the remnant channel(s) on the historic floodplain and eliminating the incised channel as a feature in the landscape. Historic channel incision resulted in significant land degradation as the adjacent groundwater levels dropped commensurate with the incising stream bed. Vegetation conversion rapidly follows as deep, densely rooted meadow plant communities convert to xeric shrubs and other plants. After a decade of meadow restoration, the FR-CRM recognized the possibility of a significant change in carbon stocks in these restored meadows and valleys. Plumas County has been a leader in advocating for investment in watershed ecosystem services such as water storage and filtering, and now, carbon sequestration. The county provided funding for the FR-CRM to conduct a pilot study of carbon in biomass and soils. Watershed Location and Characteristics The upper Feather River watershed is located in northeastern California encompassing 3,222 square miles that drains west from east of the Sierra crest into Oroville Reservoir and thence to the Sacramento River. Annual runoff produced from this watershed provides over 1,400 MW of hydroelectric power, and represents a significant component of the California State Water Project, annually providing 2.3 millionacre feet of water for urban, industrial and agricultural consumers downstream. The Feather River watershed is primarily comprised of two distinct geologies: the Sierra Nevada granitic batholith of the western third of the watershed; and Basin and Range fault-block meta-volcanics, metasedimentary and recent basalts in the eastern two-thirds. It is the Basin and Range zone (Diamond Mtns.) of the watershed that has been the primary area of restoration. This geologic m\u00e9lange of faulted and weathered rock has resulted in over 390 square miles of expansive meadows and valleys comprised of deep fine grained alluvium, shown as green and yellow in Figure 1. Figure 1. Upper Feather River Watershed Upper watershed meadows and valleys (shown as green/yellow in Figure 1), often dozens of miles in length, once supported a rich ecosystem of meadow and riparian habitats, for coldwater-loving trout, a diversity of wildlife, and indigenous peoples during the dry summers of California\u2019s Mediterranean climate. The densely rooted vegetation, cohesive soils and expansive floodplains all contributed to the sustainability of these meso-scale floodplain meadows, with associated alluvial fans. River system segments are often characterized simplistically as transport and depositional reaches. Depositional reaches feature lower gradients and a more expansive fluvial setting. These landscape attributes, in conjunction with the type and quantity of sediment, debris and nutrients, are what provide for the development and evolution of meso-scale \u201csinks\u201d or \u201cwarehouses\u201d, for the hydrologic products of the basin. Viewed as a macro-hyporheic corridor ( Harvey and Wagner, 2000; Boulton, et.al., 1998; Stanford and Ward, 1993) these features are crucial as a landscape zone of active mass and energy transfer as well as an active storage reservoir for water, sediment and nutrients. The long-term recruitment and evolution of these features involve physical, Figure 2. Typical Alluvial Features biological and chemical synthesis within the natural variability of fluvial processes. Euro-American settlement of the watershed began in 1850 with gold mining in the western portions of the watershed and, soon thereafter, agricultural production in meadows to support the mining communities. Dairy farming, horses (for cavalry mounts), sheep and beef cattle were some of the early intensive disturbances that led to localized channel incision. The resultant lowering of shallow groundwater elevations began to alter and weaken the vegetative structure of the system. Soon, near the burgeoning communities in the mid-elevation valleys, a permanent road system was established with frequent channel manipulation and relocation efforts to simplify drainage and minimize bridge construction, again leading to localized incision. In the early 1900\u2019s both an intercontinental, and numerous local, railroad systems were constructed throughout the watershed. The local railroad networks, for the purpose of both mining and logging, were routed through the long low-gradient valleys for ease of construction. These valleys were still relatively wet at that time so elevated grades were constructed using adjacent borrow ditches. By 1940, the severe morphological changes imposed by the railroad grades, in conjunction with the above referenced land use impacts resulted in rapid, severe systemic incision of many upper watershed meadow systems. In the mid 1980\u2019s numerous watershed stakeholders adopted a statutory authority that allowed for Coordinated Resource Management and Planning (CRMP). Twenty-four federal, state and local, public and private entities now form the Feather River Coordinated Resource Management (FRCRM) group to adopt, support and implement a watershed-wide restoration program. FR-CRM Restoration Approach & Background The FRCRM began an ongoing implementation program to address these watershed issues in 1990. Initially, these projects focused on geomorphic restoration techniques (Rosgen, 1996) to stabilize incised stream channels. While overall success was encouraging, the projects illustrated the concept that any restoration work in the incised channels was subject to elevated stresses even in moderate flood events (510 year return interval). Concurrently, the benefits from this approach were localized and limited to reduced erosion, and incremental improvement of aquatic habitats and water quality. Little overall improvement of watershed conditions was being realized (Wilcox, et al 2001). This led to re-evaluating restoration approach to encompass the entire historic fluvially-evolved valley bottom. Called meadow re-watering, this approach entails returning the incised stream channel to the remnant channel(s) on the historic floodplain and eliminating the incised channel as a water conveyance feature in the landscape (Figures 3 & 4 and photos 1a, 1b, 2a & 2b). Simultaneously, the FRCRM had received a project assistance request from the United States Forest Service, Plumas National Forest (PNF) to develop restoration alternatives for Cottonwood Creek in the Big Flat Meadow (Photos 2a & 2b). FRCRM staff, led by Jim Wilcox, began conducting surveys and data collection that included the entire relic meadow from hillslope to hillslope. This data collection effort quickly pointed to the nascent meadow re-watering technology as a likely restoration alternative. Figure 3. Typical cross-section, showing pre-project incision, post-project plug elevation, and the new channel. Photos 1a and 1b below show this same cross-section, however, the entire gully is not shown in the pre-project photo. Photo 1aClarks Creek Pre-project, July, 2001 Photo 1bClarks Creek Post project, July, 2006 The rocks in the background of photos 1a and 1b can be used for reference. Because the new channel is in a different location, the photo point also moved in order to show the channel in the preand postproject conditions. Figure 4. Typical cross-section, showing pre-project incision, post-project plug elevation, and the new channel. Implemented in 1995, this project quickly validated the fundamental soundness of this approach. The one mile long, 47 acre project produced elevated shallow groundwater levels, eliminated gully wall erosion, filtered sediments delivered from the upper watershed, extended and increased summer baseflows, and reversed the xeric vegetation trends resulting in improved terrestrial, avian and aquatic habitats. These benefits persisted despite withstanding a 100-year RI (return interval) flood in 1997. Photo 2aBig Flat Pre-project, Dec.,1993 Photo 2bBig Flat Post project, May, 2006 The success of this initial project led to the implementation of an additional 18 projects utilizing this technology (Table 1.). Varying in scale and watershed characteristics, these projects have restored another 20 miles of channel and 5,000 acres of meadow/floodplain. Carbon Sequestration Qualitatively, these projects appeared to significantly increase organic carbon stocks through the much increased root mass as well as increased surface growth, and, possibly, through the more effective hyporheic exchange throughout the meadow. The purpose of the following protocol is to quantitatively establish the effe"
                },
                {
                    "title": "Mobile-Env: A Universal Platform for Training and Evaluation of Mobile Interaction",
                    "abstract": "The interaction platform plays a crucial role in the recent advancement of the control and decision domains like game playing and embodied intelligence. However, there is still a lack of a satisfactory platform for the information user interface (InfoUI) interaction. The proposed InfoUI comprises not only the plain text information, but the multimodal contents and a few spatial structures with styles as well. To help the research of InfoUI interaction, a novel platform Mobile-Env is presented in this paper. The Mobile-Env platform is designed to be \ufb02exible, adaptable, and easily-extended. Based on Mobile-Env, an InfoUI task set is then built for a demonstration and evaluation. An agent based on the large-scale language model (LLM) is tested on the task set. The experiment results demonstrate the great potential of the LLM to do text understanding and matching and, meanwhile, reveal the necessity of a better mechanism of interaction feedback and exploration. Several new discussions are conducted as well. A demo video is available at https://youtu.be/gKV6KZYwxGY . The code repository is available at https: //github.com/X-LANCE/Mobile-Env . The proposed WikiHow task set is made public at https://huggingface.co/datasets/zdy023/WikiHow-taskset ."
                },
                {
                    "title": "Overleaf Example",
                    "abstract": "We present Set-of-Mark (SoM), a new visual prompting method, to unleash the visual grounding abilities of large multimodal models (LMMs), such as GPT-4V. As illustrated in Fig. 1 (right), we employ off-the-shelf interactive segmentation models, such as SAM, to partition an image into regions at different levels of granularity, and overlay these regions with a set of marks e.g., alphanumerics, masks, boxes. Using the marked image as input, GPT-4V can answer the questions that require visual grounding. We perform a comprehensive empirical study to validate the effectiveness of SoM on a wide range of fine-grained vision and multimodal tasks. For example, our experiments show that GPT-4V with SoM outperforms the state-of-the-art fully-finetuned referring segmentation model on RefCOCOg in a zero-shot setting. ar X iv :2 31 0. 11 44 1v 1 [ cs .C V ] 1 7 O ct 2 02 3"
                },
                {
                    "title": "The Claude 3 Model Family: Opus, Sonnet, Haiku",
                    "abstract": "We introduce Claude 3, a new family of large multimodal models \u2013 Claude 3 Opus , our most capable offering, Claude 3 Sonnet , which provides a combination of skills and speed, and Claude 3 Haiku , our fastest and least expensive model. All new models have vision capabilities that enable them to process and analyze image data. The Claude 3 family demonstrates strong performance across benchmark evaluations and sets a new standard on measures of reasoning, math, and coding. Claude 3 Opus achieves state-of-the-art results on evaluations like GPQA [1], MMLU [2], MMMU [3] and many more. Claude 3 Haiku performs as well or better than Claude 2 [4] on most pure-text tasks, while Sonnet and Opus significantly outperform it. Additionally, these models exhibit improved fluency in non-English languages, making them more versatile for a global audience. In this report, we provide an in-depth analysis of our evaluations, focusing on core capabilities, safety, societal impacts, and the catastrophic risk assessments we committed to in our Responsible Scaling Policy [5]."
                }
            ],
            "categories": [
                "cs.AI",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we develop a multimodal agent benchmark that integrates both code generation and GUI controls to automate the entire data science and engineering workflow across various professional enterprise data software systems?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses a significant gap in the automation of data science and engineering workflows, which traditionally rely on complex manual processes. By creating a benchmark like Spider2-V, we can enhance productivity for data scientists and engineers, democratize access to large-scale data, and pave the way for future research on multimodal agents. This advancement could lead to practical applications that streamline data workflows, reduce human error, and enable non-experts to engage with data science tasks more effectively.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the need to integrate multiple modalities\u2014code generation and GUI interactions\u2014within a single framework. Naive approaches may fail because they often overlook the complexities of real-time interactions with various enterprise applications, which require both coding skills and the ability to navigate graphical interfaces. Additionally, technical obstacles include ensuring that the multimodal agent can accurately interpret and execute tasks in a dynamic environment, as well as managing the variability in user interfaces across different software systems.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on specific aspects of data processing, such as code generation or API calls, without addressing the full spectrum of data science and engineering tasks that involve intensive GUI controls. Barriers to solving this problem include a lack of comprehensive benchmarks that encompass both coding and GUI interactions, as well as the complexity of simulating human-like behavior in professional data applications. Our approach differs by providing a holistic benchmark (Spider2-V) that evaluates multimodal agents across a wide range of real-world tasks, thus filling the existing gaps in the literature.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves developing the Spider2-V benchmark, which includes 494 real-world tasks across 20 professional enterprise data software systems. The tasks will be executed in a real-time executable computer environment (OS-WORLD) that allows multimodal agents to simulate human actions. We will assess the agents' performance using metrics such as task completion rate and accuracy in both code generation and GUI interactions. The expected outcomes include a validated benchmark that demonstrates the capabilities of multimodal agents in automating complex data workflows, ultimately"
            }
        },
        "author_data": {
            "9cceac30-cf2b-4b4d-ad6e-cd6782e61211": {
                "pk": "9cceac30-cf2b-4b4d-ad6e-cd6782e61211",
                "name": "Ruisheng Cao",
                "collaborators": [
                    "Kai Yu",
                    "Lu Chen",
                    "Su Zhu",
                    "Hongshen Xu",
                    "Hanchong Zhang",
                    "Zhi Chen",
                    "Jieyu Li",
                    "Chen Liu",
                    "Yanbin Zhao",
                    "Da Ma"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Semantic Parsing",
                    "Text-to-SQL",
                    "Few-shot Learning"
                ],
                "publications": [
                    {
                        "title": "Dual Learning for Semi-Supervised Natural Language Understanding",
                        "abstract": "Natural language understanding (NLU) converts sentences into structured semantic forms. The paucity of annotated training samples is still a fundamental challenge of NLU. To solve this data sparsity problem, previous work based on semi-supervised learning mainly focuses on exploiting unlabeled sentences. In this work, we introduce a dual task of NLU, semantic-to-sentence generation (SSG), and propose a new framework for semi-supervised NLU with the corresponding dual model. The framework is composed of dual pseudo-labeling and dual learning method, which enables an NLU model to make full use of data (labeled and unlabeled) through a closed-loop of the primal and dual tasks. By incorporating the dual task, the framework can exploit pure semantic forms as well as unlabeled sentences, and further improve the NLU and SSG models iteratively in the closed-loop. The proposed approaches are evaluated on two public datasets (ATIS and SNIPS). Experiments in the semi-supervised setting show that our methods can outperform various baselines significantly, and extensive ablation studies are conducted to verify the effectiveness of our framework. Finally, our method can also achieve the state-of-the-art performance on the two datasets in the supervised setting. Our code is available at \\url{https://github.com/rhythmcao/slu-dual-learning.git}."
                    },
                    {
                        "title": "Vector Projection Network for Few-shot Slot Tagging in Natural Language Understanding",
                        "abstract": "Few-shot slot tagging becomes appealing for rapid domain transfer and adaptation, motivated by the tremendous development of conversational dialogue systems. In this paper, we propose a vector projection network for few-shot slot tagging, which exploits projections of contextual word embeddings on each target label vector as word-label similarities. Essentially, this approach is equivalent to a normalized linear model with an adaptive bias. The contrastive experiment demonstrates that our proposed vector projection based similarity metric can significantly surpass other variants. Specifically, in the five-shot setting on benchmarks SNIPS and NER, our method outperforms the strongest few-shot learning baseline by $6.30$ and $13.79$ points on F$_1$ score, respectively. Our code will be released at https://github.com/sz128/few_shot_slot_tagging_and_NER."
                    },
                    {
                        "title": "ACT-SQL: In-Context Learning for Text-to-SQL with Automatically-Generated Chain-of-Thought",
                        "abstract": "Recently Large Language Models (LLMs) have been proven to have strong abilities in various domains and tasks. We study the problem of prompt designing in the text-to-SQL task and attempt to improve the LLMs' reasoning ability when generating SQL queries. Besides the trivial few-shot in-context learning setting, we design our chain-of-thought (CoT) prompt with a similar method to schema linking. We provide a method named ACT-SQL to automatically generate auto-CoT exemplars and thus the whole process doesn't need manual labeling. Our approach is cost-saving since we only use the LLMs' API call once when generating one SQL query. Furthermore, we extend our in-context learning method to the multi-turn text-to-SQL task. The experiment results show that the LLMs' performance can benefit from our ACT-SQL approach. Our approach achieves SOTA performance on the Spider dev set among existing in-context learning approaches."
                    },
                    {
                        "title": "CoE-SQL: In-Context Learning for Multi-Turn Text-to-SQL with Chain-of-Editions",
                        "abstract": "Recently, Large Language Models (LLMs) have been demonstrated to possess impressive capabilities in a variety of domains and tasks. We investigate the issue of prompt design in the multi-turn text-to-SQL task and attempt to enhance the LLMs' reasoning capacity when generating SQL queries. In the conversational context, the current SQL query can be modified from the preceding SQL query with only a few operations due to the context dependency. We introduce our method called CoE-SQL which can prompt LLMs to generate the SQL query based on the previously generated SQL query with an edition chain. We also conduct extensive ablation studies to determine the optimal configuration of our approach. Our approach outperforms different in-context learning baselines stably and achieves state-of-the-art performances on two benchmarks SParC and CoSQL using LLMs, which is also competitive to the SOTA fine-tuned models."
                    },
                    {
                        "title": "Semantic Parsing with Dual Learning",
                        "abstract": "Semantic parsing converts natural language queries into structured logical forms. The paucity of annotated training samples is a fundamental challenge in this field. In this work, we develop a semantic parsing framework with the dual learning algorithm, which enables a semantic parser to make full use of data (labeled and even unlabeled) through a dual-learning game. This game between a primal model (semantic parsing) and a dual model (logical form to query) forces them to regularize each other, and can achieve feedback signals from some prior-knowledge. By utilizing the prior-knowledge of logical form structures, we propose a novel reward signal at the surface and semantic levels which tends to generate complete and reasonable logical forms. Experimental results show that our approach achieves new state-of-the-art performance on ATIS dataset and gets competitive performance on Overnight dataset."
                    },
                    {
                        "title": "LGESQL: Line Graph Enhanced Text-to-SQL Model with Mixed Local and Non-Local Relations",
                        "abstract": "This work aims to tackle the challenging heterogeneous graph encoding problem in the text-to-SQL task. Previous methods are typically node-centric and merely utilize different weight matrices to parameterize edge types, which 1) ignore the rich semantics embedded in the topological structure of edges, and 2) fail to distinguish local and non-local relations for each node. To this end, we propose a Line Graph Enhanced Text-to-SQL (LGESQL) model to mine the underlying relational features without constructing meta-paths. By virtue of the line graph, messages propagate more efficiently through not only connections between nodes, but also the topology of directed edges. Furthermore, both local and non-local relations are integrated distinctively during the graph iteration. We also design an auxiliary task called graph pruning to improve the discriminative capability of the encoder. Our framework achieves state-of-the-art results (62.8% with Glove, 72.0% with Electra) on the cross-domain text-to-SQL benchmark Spider at the time of writing."
                    },
                    {
                        "title": "Jointly Encoding Word Confusion Network and Dialogue Context with BERT for Spoken Language Understanding",
                        "abstract": "Spoken Language Understanding (SLU) converts hypotheses from automatic speech recognizer (ASR) into structured semantic representations. ASR recognition errors can severely degenerate the performance of the subsequent SLU module. To address this issue, word confusion networks (WCNs) have been used to encode the input for SLU, which contain richer information than 1-best or n-best hypotheses list. To further eliminate ambiguity, the last system act of dialogue context is also utilized as additional input. In this paper, a novel BERT based SLU model (WCN-BERT SLU) is proposed to encode WCNs and the dialogue context jointly. It can integrate both structural information and ASR posterior probabilities of WCNs in the BERT architecture. Experiments on DSTC2, a benchmark of SLU, show that the proposed method is effective and can outperform previous state-of-the-art models significantly."
                    },
                    {
                        "title": "Few-Shot NLU with Vector Projection Distance and Abstract Triangular CRF",
                        "abstract": "Data sparsity problem is a key challenge of Natural Language Understanding (NLU), especially for a new target domain. By training an NLU model in source domains and applying the model to an arbitrary target domain directly (even without fine-tuning), few-shot NLU becomes crucial to mitigate the data scarcity issue. In this paper, we propose to improve prototypical networks with vector projection distance and abstract triangular Conditional Random Field (CRF) for the few-shot NLU. The vector projection distance exploits projections of contextual word embeddings on label vectors as word-label similarities, which is equivalent to a normalized linear model. The abstract triangular CRF learns domain-agnostic label transitions for joint intent classification and slot filling tasks. Extensive experiments demonstrate that our proposed methods can significantly surpass strong baselines. Specifically, our approach can achieve a new state-of-the-art on two few-shot NLU benchmarks (Few-Joint and SNIPS) in Chinese and English without fine-tuning on target domains."
                    },
                    {
                        "title": "TIE: Topological Information Enhanced Structural Reading Comprehension on Web Pages",
                        "abstract": "Recently, the structural reading comprehension (SRC) task on web pages has attracted increasing research interests. Although previous SRC work has leveraged extra information such as HTML tags or XPaths, the informative topology of web pages is not effectively exploited. In this work, we propose a Topological Information Enhanced model (TIE), which transforms the token-level task into a tag-level task by introducing a two-stage process (i.e. node locating and answer refining). Based on that, TIE integrates Graph Attention Network (GAT) and Pre-trained Language Model (PLM) to leverage the topological information of both logical structures and spatial structures. Experimental results demonstrate that our model outperforms strong baselines and achieves state-of-the-art performances on the web-based SRC benchmark WebSRC at the time of writing. The code of TIE will be publicly available at https://github.com/X-LANCE/TIE."
                    },
                    {
                        "title": "Unsupervised Dual Paraphrasing for Two-stage Semantic Parsing",
                        "abstract": "One daunting problem for semantic parsing is the scarcity of annotation. Aiming to reduce nontrivial human labor, we propose a two-stage semantic parsing framework, where the first stage utilizes an unsupervised paraphrase model to convert an unlabeled natural language utterance into the canonical utterance. The downstream naive semantic parser accepts the intermediate output and returns the target logical form. Furthermore, the entire training process is split into two phases: pre-training and cycle learning. Three tailored self-supervised tasks are introduced throughout training to activate the unsupervised paraphrase model. Experimental results on benchmarks Overnight and GeoGranno demonstrate that our framework is effective and compatible with supervised training."
                    },
                    {
                        "title": "ASTormer: An AST Structure-aware Transformer Decoder for Text-to-SQL",
                        "abstract": "Text-to-SQL aims to generate an executable SQL program given the user utterance and the corresponding database schema. To ensure the well-formedness of output SQLs, one prominent approach adopts a grammar-based recurrent decoder to produce the equivalent SQL abstract syntax tree (AST). However, previous methods mainly utilize an RNN-series decoder, which 1) is time-consuming and inefficient and 2) introduces very few structure priors. In this work, we propose an AST structure-aware Transformer decoder (ASTormer) to replace traditional RNN cells. The structural knowledge, such as node types and positions in the tree, is seamlessly incorporated into the decoder via both absolute and relative position embeddings. Besides, the proposed framework is compatible with different traversing orders even considering adaptive node selection. Extensive experiments on five text-to-SQL benchmarks demonstrate the effectiveness and efficiency of our structured decoder compared to competitive baselines."
                    },
                    {
                        "title": "ShadowGNN: Graph Projection Neural Network for Text-to-SQL Parser",
                        "abstract": "Given a database schema, Text-to-SQL aims to translate a natural language question into the corresponding SQL query. Under the setup of cross-domain, traditional semantic parsing models struggle to adapt to unseen database schemas. To improve the model generalization capability for rare and unseen schemas, we propose a new architecture, ShadowGNN, which processes schemas at abstract and semantic levels. By ignoring names of semantic items in databases, abstract schemas are exploited in a well-designed graph projection neural network to obtain delexicalized representation of question and schema. Based on the domain-independent representations, a relation-aware transformer is utilized to further extract logical linking between question and schema. Finally, a SQL decoder with context-free grammar is applied. On the challenging Text-to-SQL benchmark Spider, empirical results show that ShadowGNN outperforms state-of-the-art models. When the annotated data is extremely limited (only 10\\% training set), ShadowGNN gets over absolute 5\\% performance gain, which shows its powerful generalization ability. Our implementation will be open-sourced at \\url{https://github.com/WowCZ/shadowgnn}."
                    },
                    {
                        "title": "CSS: A Large-scale Cross-schema Chinese Text-to-SQL Medical Dataset",
                        "abstract": "The cross-domain text-to-SQL task aims to build a system that can parse user questions into SQL on complete unseen databases, and the single-domain text-to-SQL task evaluates the performance on identical databases. Both of these setups confront unavoidable difficulties in real-world applications. To this end, we introduce the cross-schema text-to-SQL task, where the databases of evaluation data are different from that in the training data but come from the same domain. Furthermore, we present CSS, a large-scale CrosS-Schema Chinese text-to-SQL dataset, to carry on corresponding studies. CSS originally consisted of 4,340 question/SQL pairs across 2 databases. In order to generalize models to different medical systems, we extend CSS and create 19 new databases along with 29,280 corresponding dataset examples. Moreover, CSS is also a large corpus for single-domain Chinese text-to-SQL studies. We present the data collection approach and a series of analyses of the data statistics. To show the potential and usefulness of CSS, benchmarking baselines have been conducted and reported. Our dataset is publicly available at \\url{https://huggingface.co/datasets/zhanghanchong/css}."
                    },
                    {
                        "title": "A BiRGAT Model for Multi-intent Spoken Language Understanding with Hierarchical Semantic Frames",
                        "abstract": "Previous work on spoken language understanding (SLU) mainly focuses on single-intent settings, where each input utterance merely contains one user intent. This configuration significantly limits the surface form of user utterances and the capacity of output semantics. In this work, we first propose a Multi-Intent dataset which is collected from a realistic in-Vehicle dialogue System, called MIVS. The target semantic frame is organized in a 3-layer hierarchical structure to tackle the alignment and assignment problems in multi-intent cases. Accordingly, we devise a BiRGAT model to encode the hierarchy of ontology items, the backbone of which is a dual relational graph attention network. Coupled with the 3-way pointer-generator decoder, our method outperforms traditional sequence labeling and classification-based schemes by a large margin."
                    },
                    {
                        "title": "Hierarchical Multimodal Pre-training for Visually Rich Webpage Understanding",
                        "abstract": "The growing prevalence of visually rich documents, such as webpages and scanned/digital-born documents (images, PDFs, etc.), has led to increased interest in automatic document understanding and information extraction across academia and industry. Although various document modalities, including image, text, layout, and structure, facilitate human information retrieval, the interconnected nature of these modalities presents challenges for neural networks. In this paper, we introduce WebLM, a multimodal pre-training network designed to address the limitations of solely modeling text and structure modalities of HTML in webpages. Instead of processing document images as unified natural images, WebLM integrates the hierarchical structure of document images to enhance the understanding of markup-language-based documents. Additionally, we propose several pre-training tasks to model the interaction among text, structure, and image modalities effectively. Empirical results demonstrate that the pre-trained WebLM significantly surpasses previous state-of-the-art pre-trained models across several webpage understanding tasks. The pre-trained models and code are available at https://github.com/X-LANCE/weblm."
                    },
                    {
                        "title": "Decoupled Dialogue Modeling and Semantic Parsing for Multi-Turn Text-to-SQL",
                        "abstract": "Recently, Text-to-SQL for multi-turn dialogue has attracted great interest. Here, the user input of the current turn is parsed into the corresponding SQL query of the appropriate database, given all previous dialogue history. Current approaches mostly employ end-to-end models and consequently face two challenges. First, dialogue history modeling and Text-to-SQL parsing are implicitly combined, hence it is hard to carry out interpretable analysis and obtain targeted improvement. Second, SQL annotation of multi-turn dialogue is very expensive, leading to training data sparsity. In this paper, we propose a novel decoupled multi-turn Text-to-SQL framework, where an utterance rewrite model first explicitly solves completion of dialogue context, and then a single-turn Text-to-SQL parser follows. A dual learning approach is also proposed for the utterance rewrite model to address the data sparsity problem. Compared with end-to-end approaches, the proposed decoupled method can achieve excellent performance without any annotated in-domain data. With just a few annotated rewrite cases, the decoupled method outperforms the released state-of-the-art end-to-end models on both SParC and CoSQL datasets."
                    },
                    {
                        "title": "On the Structural Generalization in Text-to-SQL",
                        "abstract": "Exploring the generalization of a text-to-SQL parser is essential for a system to automatically adapt the real-world databases. Previous works provided investigations focusing on lexical diversity, including the influence of the synonym and perturbations in both natural language questions and databases. However, research on the structure variety of database schema~(DS) is deficient. Specifically, confronted with the same input question, the target SQL is probably represented in different ways when the DS comes to a different structure. In this work, we provide in-deep discussions about the structural generalization of text-to-SQL tasks. We observe that current datasets are too templated to study structural generalization. To collect eligible test data, we propose a framework to generate novel text-to-SQL data via automatic and synchronous (DS, SQL) pair altering. In the experiments, significant performance reduction when evaluating well-trained text-to-SQL models on the synthetic samples demonstrates the limitation of current research regarding structural generalization. According to comprehensive analysis, we suggest the practical reason is the overfitting of (NL, SQL) patterns."
                    }
                ]
            },
            "a4577f0b-8b57-443d-a5db-654a66244320": {
                "pk": "a4577f0b-8b57-443d-a5db-654a66244320",
                "name": "Fangyu Lei",
                "collaborators": [
                    "Kang Liu",
                    "Jun Zhao",
                    "Shizhu He",
                    "Yifan Wei",
                    "Tongxu Luo",
                    "Yiming Huang",
                    "Weihao Liu",
                    "Jiahe Lei",
                    "Xiaoyan Yu",
                    "Xiang Li"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Question Answering",
                    "Large Language Models",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Answering Numerical Reasoning Questions in Table-Text Hybrid Contents with Graph-based Encoder and Tree-based Decoder",
                        "abstract": "In the real-world question answering scenarios, hybrid form combining both tabular and textual contents has attracted more and more attention, among which numerical reasoning problem is one of the most typical and challenging problems. Existing methods usually adopt encoder-decoder framework to represent hybrid contents and generate answers. However, it can not capture the rich relationship among numerical value, table schema, and text information on the encoder side. The decoder uses a simple predefined operator classifier which is not flexible enough to handle numerical reasoning processes with diverse expressions. To address these problems, this paper proposes a \\textbf{Re}lational \\textbf{G}raph enhanced \\textbf{H}ybrid table-text \\textbf{N}umerical reasoning model with \\textbf{T}ree decoder (\\textbf{RegHNT}). It models the numerical question answering over table-text hybrid contents as an expression tree generation task. Moreover, we propose a novel relational graph modeling method, which models alignment between questions, tables, and paragraphs. We validated our model on the publicly available table-text hybrid QA benchmark (TAT-QA). The proposed RegHNT significantly outperform the baseline model and achieve state-of-the-art results. We openly released the source code and data at https://github.com/lfy79001/RegHNT (2022-05-05)."
                    },
                    {
                        "title": "Multi-View Graph Representation Learning for Answering Hybrid Numerical Reasoning Question",
                        "abstract": "Hybrid question answering (HybridQA) over the financial report contains both textual and tabular data, and requires the model to select the appropriate evidence for the numerical reasoning task. Existing methods based on encoder-decoder framework employ a expression tree-based decoder to solve numerical reasoning problems. However, encoders rely more on Machine Reading Comprehension (MRC) methods, which take table serialization and text splicing as input, damaging the granularity relationship between table and text as well as the spatial structure information of table itself. In order to solve these problems, the paper proposes a Multi-View Graph (MVG) Encoder to take the relations among the granularity into account and capture the relations from multiple view. By utilizing MVGE as a module, we constuct Tabular View, Relation View and Numerical View which aim to retain the original characteristics of the hybrid data. We validate our model on the publicly available table-text hybrid QA benchmark (TAT-QA) and outperform the state-of-the-art model."
                    },
                    {
                        "title": "S$^3$HQA: A Three-Stage Approach for Multi-hop Text-Table Hybrid Question Answering",
                        "abstract": "Answering multi-hop questions over hybrid factual knowledge from the given text and table (TextTableQA) is a challenging task. Existing models mainly adopt a retriever-reader framework, which have several deficiencies, such as noisy labeling in training retriever, insufficient utilization of heterogeneous information over text and table, and deficient ability for different reasoning operations. In this paper, we propose a three-stage TextTableQA framework S3HQA, which comprises of retriever, selector, and reasoner. We use a retriever with refinement training to solve the noisy labeling problem. Then, a hybrid selector considers the linked relationships between heterogeneous data to select the most relevant factual knowledge. For the final stage, instead of adapting a reading comprehension module like in previous methods, we employ a generation-based reasoner to obtain answers. This includes two approaches: a row-wise generator and an LLM prompting generator~(first time used in this task). The experimental results demonstrate that our method achieves competitive results in the few-shot setting. When trained on the full dataset, our approach outperforms all baseline methods, ranking first on the HybridQA leaderboard."
                    },
                    {
                        "title": "S3Eval: A Synthetic, Scalable, Systematic Evaluation Suite for Large Language Models",
                        "abstract": "The rapid development of Large Language Models (LLMs) has led to great strides in model capabilities like long-context understanding and reasoning. However, as LLMs are able to process longer contexts, it becomes more challenging to evaluate whether they have acquired certain capabilities, since the length of text (e.g., 200K tokens) they can process far exceeds what humans can reliably assess in a reasonable duration. In this paper, we propose using complex synthetic tasks as a proxy evaluation method, and present S3Eval, a Synthetic, Scalable, Systematic evaluation suite for LLMs evaluation. The synthetic nature of S3Eval provides users full control over the dataset, allowing them to systematically probe LLM capabilities by scaling text length and varying task difficulty across diverse scenarios. The strong correlation between S3Eval and real-world benchmarks demonstrates the soundness of using S3Eval for evaluation of LLMs. S3Eval provides a flexible and infinite long-context data generation method. We have generated a comprehensive dataset called S3Eval-Standard, and experimental results have shown that it poses significant challenges for all existing LLMs."
                    },
                    {
                        "title": "Assessing Knowledge Editing in Language Models via Relation Perspective",
                        "abstract": "Knowledge Editing (KE) for modifying factual knowledge in Large Language Models (LLMs) has been receiving increasing attention. However, existing knowledge editing methods are entity-centric, and it is unclear whether this approach is suitable for a relation-centric perspective. To address this gap, this paper constructs a new benchmark named RaKE, which focuses on Relation based Knowledge Editing. In this paper, we establish a suite of innovative metrics for evaluation and conduct comprehensive experiments involving various knowledge editing baselines. We notice that existing knowledge editing methods exhibit the potential difficulty in their ability to edit relations. Therefore, we further explore the role of relations in factual triplets within the transformer. Our research results confirm that knowledge related to relations is not only stored in the FFN network but also in the attention layers. This provides experimental support for future relation-based knowledge editing methods."
                    },
                    {
                        "title": "Neeko: Leveraging Dynamic LoRA for Efficient Multi-Character Role-Playing Agent",
                        "abstract": "Large Language Models (LLMs) have revolutionized open-domain dialogue agents but encounter challenges in multi-character role-playing (MCRP) scenarios. To address the issue, we present Neeko, an innovative framework designed for efficient multiple characters imitation. Unlike existing methods, Neeko employs a dynamic low-rank adapter (LoRA) strategy, enabling it to adapt seamlessly to diverse characters. Our framework breaks down the role-playing process into agent pre-training, multiple characters playing, and character incremental learning, effectively handling both seen and unseen roles. This dynamic approach, coupled with distinct LoRA blocks for each character, enhances Neeko's adaptability to unique attributes, personalities, and speaking patterns. As a result, Neeko demonstrates superior performance in MCRP over most existing methods, offering more engaging and versatile user interaction experiences. Code and data are available at https://github.com/weiyifan1023/Neeko."
                    },
                    {
                        "title": "MMHQA-ICL: Multimodal In-context Learning for Hybrid Question Answering over Text, Tables and Images",
                        "abstract": "In the real world, knowledge often exists in a multimodal and heterogeneous form. Addressing the task of question answering with hybrid data types, including text, tables, and images, is a challenging task (MMHQA). Recently, with the rise of large language models (LLM), in-context learning (ICL) has become the most popular way to solve QA problems. We propose MMHQA-ICL framework for addressing this problems, which includes stronger heterogeneous data retriever and an image caption module. Most importantly, we propose a Type-specific In-context Learning Strategy for MMHQA, enabling LLMs to leverage their powerful performance in this task. We are the first to use end-to-end LLM prompting method for this task. Experimental results demonstrate that our framework outperforms all baselines and methods trained on the full dataset, achieving state-of-the-art results under the few-shot setting on the MultimodalQA dataset."
                    },
                    {
                        "title": "HRoT: Hybrid prompt strategy and Retrieval of Thought for Table-Text Hybrid Question Answering",
                        "abstract": "Answering numerical questions over hybrid contents from the given tables and text(TextTableQA) is a challenging task. Recently, Large Language Models (LLMs) have gained significant attention in the NLP community. With the emergence of large language models, In-Context Learning and Chain-of-Thought prompting have become two particularly popular research topics in this field. In this paper, we introduce a new prompting strategy called Hybrid prompt strategy and Retrieval of Thought for TextTableQA. Through In-Context Learning, we prompt the model to develop the ability of retrieval thinking when dealing with hybrid data. Our method achieves superior performance compared to the fully-supervised SOTA on the MultiHiertt dataset in the few-shot setting."
                    },
                    {
                        "title": "MoELoRA: Contrastive Learning Guided Mixture of Experts on Parameter-Efficient Fine-Tuning for Large Language Models",
                        "abstract": "Fine-tuning is often necessary to enhance the adaptability of Large Language Models (LLM) to downstream tasks. Nonetheless, the process of updating billions of parameters demands significant computational resources and training time, which poses a substantial obstacle to the widespread application of large-scale models in various scenarios. To address this issue, Parameter-Efficient Fine-Tuning (PEFT) has emerged as a prominent paradigm in recent research. However, current PEFT approaches that employ a limited set of global parameters (such as LoRA, which adds low-rank approximation matrices to all weights) face challenges in flexibly combining different computational modules in downstream tasks. In this work, we introduce a novel PEFT method: MoELoRA. We consider LoRA as Mixture of Experts (MoE), and to mitigate the random routing phenomenon observed in MoE, we propose the utilization of contrastive learning to encourage experts to learn distinct features. We conducted experiments on 11 tasks in math reasoning and common-sense reasoning benchmarks. With the same number of parameters, our approach outperforms LoRA significantly. In math reasoning, MoELoRA achieved an average performance that was 4.2% higher than LoRA, and demonstrated competitive performance compared to the 175B GPT-3.5 on several benchmarks."
                    },
                    {
                        "title": "MenatQA: A New Dataset for Testing the Temporal Comprehension and Reasoning Abilities of Large Language Models",
                        "abstract": "Large language models (LLMs) have shown nearly saturated performance on many natural language processing (NLP) tasks. As a result, it is natural for people to believe that LLMs have also mastered abilities such as time understanding and reasoning. However, research on the temporal sensitivity of LLMs has been insufficiently emphasized. To fill this gap, this paper constructs Multiple Sensitive Factors Time QA (MenatQA), which encompasses three temporal factors (scope factor, order factor, counterfactual factor) with total 2,853 samples for evaluating the time comprehension and reasoning abilities of LLMs. This paper tests current mainstream LLMs with different parameter sizes, ranging from billions to hundreds of billions. The results show most LLMs fall behind smaller temporal reasoning models with different degree on these factors. In specific, LLMs show a significant vulnerability to temporal biases and depend heavily on the temporal information provided in questions. Furthermore, this paper undertakes a preliminary investigation into potential improvement strategies by devising specific prompts and leveraging external tools. These approaches serve as valuable baselines or references for future research endeavors."
                    },
                    {
                        "title": "TableQAKit: A Comprehensive and Practical Toolkit for Table-based Question Answering",
                        "abstract": "Table-based question answering (TableQA) is an important task in natural language processing, which requires comprehending tables and employing various reasoning ways to answer the questions. This paper introduces TableQAKit, the first comprehensive toolkit designed specifically for TableQA. The toolkit designs a unified platform that includes plentiful TableQA datasets and integrates popular methods of this task as well as large language models (LLMs). Users can add their datasets and methods according to the friendly interface. Also, pleasantly surprised using the modules in this toolkit achieves new SOTA on some datasets. Finally, \\tableqakit{} also provides an LLM-based TableQA Benchmark for evaluating the role of LLMs in TableQA. TableQAKit is open-source with an interactive interface that includes visual operations, and comprehensive data for ease of use."
                    }
                ]
            },
            "dd6a1d83-12b9-45a3-9b0f-126c1aad8c65": {
                "pk": "dd6a1d83-12b9-45a3-9b0f-126c1aad8c65",
                "name": "Haoyuan Wu",
                "collaborators": [
                    "Bei Yu",
                    "Zhuolun He",
                    "Haisheng Zheng",
                    "Xinyun Zhang",
                    "Xufeng Yao",
                    "Yuan Pu",
                    "Tairu Qiu",
                    "Su Zheng",
                    "Peng Xu",
                    "Peiyu Liao"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Retrieval-Augmented Generation",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Parameter-Efficient Sparsity Crafting from Dense to Mixture-of-Experts for Instruction Tuning on General Tasks",
                        "abstract": "Large language models (LLMs) have demonstrated considerable proficiency in general natural language processing (NLP) tasks. Instruction tuning, a successful paradigm, enhances the ability of LLMs to follow natural language instructions and exhibit robust generalization across general tasks. However, these models often encounter performance limitations across multiple tasks due to constrained model capacity. Expanding this capacity during the instruction tuning phase poses significant challenges. To address this issue, we introduce parameter-efficient sparsity crafting (PESC), which crafts dense models into sparse models using the mixture-of-experts (MoE) architecture. PESC integrates adapters into the MoE layers of sparse models, differentiating experts without altering the individual weights within these layers. This method significantly reduces computational costs and GPU memory requirements, facilitating model capacity expansion through a minimal parameter increase when guaranteeing the quality of approximation in function space compared to original sparse upcycling. Our empirical evaluation demonstrates the effectiveness of the PESC method. Using PESC during instruction tuning, our best sparse model outperforms other sparse and dense models and exhibits superior general capabilities compared to GPT-3.5. Our code is available at https://github.com/wuhy68/Parameter-Efficient-MoE."
                    },
                    {
                        "title": "Customized Retrieval Augmented Generation and Benchmarking for EDA Tool Documentation QA",
                        "abstract": "Retrieval augmented generation (RAG) enhances the accuracy and reliability of generative AI models by sourcing factual information from external databases, which is extensively employed in document-grounded question-answering (QA) tasks. Off-the-shelf RAG flows are well pretrained on general-purpose documents, yet they encounter significant challenges when being applied to knowledge-intensive vertical domains, such as electronic design automation (EDA). This paper addresses such issue by proposing a customized RAG framework along with three domain-specific techniques for EDA tool documentation QA, including a contrastive learning scheme for text embedding model fine-tuning, a reranker distilled from proprietary LLM, and a generative LLM fine-tuned with high-quality domain corpus. Furthermore, we have developed and released a documentation QA evaluation benchmark, ORD-QA, for OpenROAD, an advanced RTL-to-GDSII design platform. Experimental results demonstrate that our proposed RAG flow and techniques have achieved superior performance on ORD-QA as well as on a commercial tool, compared with state-of-the-arts. The ORD-QA benchmark and the training dataset for our customized RAG flow are open-source at https://github.com/lesliepy99/RAG-EDA."
                    },
                    {
                        "title": "ChatEDA: A Large Language Model Powered Autonomous Agent for EDA",
                        "abstract": "The integration of a complex set of Electronic Design Automation (EDA) tools to enhance interoperability is a critical concern for circuit designers. Recent advancements in large language models (LLMs) have showcased their exceptional capabilities in natural language processing and comprehension, offering a novel approach to interfacing with EDA tools. This research paper introduces ChatEDA, an autonomous agent for EDA empowered by an LLM, AutoMage, complemented by EDA tools serving as executors. ChatEDA streamlines the design flow from the Register-Transfer Level (RTL) to the Graphic Data System Version II (GDSII) by effectively managing task decomposition, script generation, and task execution. Through comprehensive experimental evaluations, ChatEDA has demonstrated its proficiency in handling diverse requirements, and our fine-tuned AutoMage model has exhibited superior performance compared to GPT-4 and other similar LLMs."
                    },
                    {
                        "title": "p-Laplacian Adaptation for Generative Pre-trained Vision-Language Models",
                        "abstract": "Vision-Language models (VLMs) pre-trained on large corpora have demonstrated notable success across a range of downstream tasks. In light of the rapidly increasing size of pre-trained VLMs, parameter-efficient transfer learning (PETL) has garnered attention as a viable alternative to full fine-tuning. One such approach is the adapter, which introduces a few trainable parameters into the pre-trained models while preserving the original parameters during adaptation. In this paper, we present a novel modeling framework that recasts adapter tuning after attention as a graph message passing process on attention graphs, where the projected query and value features and attention matrix constitute the node features and the graph adjacency matrix, respectively. Within this framework, tuning adapters in VLMs necessitates handling heterophilic graphs, owing to the disparity between the projected query and value space. To address this challenge, we propose a new adapter architecture, $p$-adapter, which employs $p$-Laplacian message passing in Graph Neural Networks (GNNs). Specifically, the attention weights are re-normalized based on the features, and the features are then aggregated using the calibrated attention matrix, enabling the dynamic exploitation of information with varying frequencies in the heterophilic attention graphs. We conduct extensive experiments on different pre-trained VLMs and multi-modal tasks, including visual question answering, visual entailment, and image captioning. The experimental results validate our method's significant superiority over other PETL methods."
                    },
                    {
                        "title": "ARKS: Active Retrieval in Knowledge Soup for Code Generation",
                        "abstract": "Recently the retrieval-augmented generation (RAG) paradigm has raised much attention for its potential in incorporating external knowledge into large language models (LLMs) without further training. While widely explored in natural language applications, its utilization in code generation remains under-explored. In this paper, we introduce Active Retrieval in Knowledge Soup (ARKS), an advanced strategy for generalizing large language models for code. In contrast to relying on a single source, we construct a knowledge soup integrating web search, documentation, execution feedback, and evolved code snippets. We employ an active retrieval strategy that iteratively refines the query and updates the knowledge soup. To assess the performance of ARKS, we compile a new benchmark comprising realistic coding problems associated with frequently updated libraries and long-tail programming languages. Experimental results on ChatGPT and CodeLlama demonstrate a substantial improvement in the average execution accuracy of ARKS on LLMs. The analysis confirms the effectiveness of our proposed knowledge soup and active retrieval strategies, offering rich insights into the construction of effective retrieval-augmented code generation (RACG) pipelines. Our model, code, and data are available at https://arks-codegen.github.io."
                    }
                ]
            },
            "c90c4056-ca3b-4e31-958d-b4395bb68441": {
                "pk": "c90c4056-ca3b-4e31-958d-b4395bb68441",
                "name": "Jixuan Chen",
                "collaborators": [
                    "Timothy Ossowski",
                    "Danyal Maqbool",
                    "Zefan Cai",
                    "Tyler Bradshaw",
                    "Junjie Hu",
                    "Zhantao Yang",
                    "Ruili Feng",
                    "Keyu Yan",
                    "Huangji Wang",
                    "Zhicai Wang"
                ],
                "domain": [
                    "Multi-Agent Systems",
                    "Language Communication",
                    "Vision Language Models",
                    "Human-Computer Interaction"
                ],
                "publications": [
                    {
                        "title": "COMMA: A Communicative Multimodal Multi-Agent Benchmark",
                        "abstract": "The rapid advances of multi-modal agents built on large foundation models have largely overlooked their potential for language-based communication between agents in collaborative tasks. This oversight presents a critical gap in understanding their effectiveness in real-world deployments, particularly when communicating with humans. Existing agentic benchmarks fail to address key aspects of inter-agent communication and collaboration, particularly in scenarios where agents have unequal access to information and must work together to achieve tasks beyond the scope of individual capabilities. To fill this gap, we introduce a novel benchmark designed to evaluate the collaborative performance of multimodal multi-agent systems through language communication. Our benchmark features a variety of scenarios, providing a comprehensive evaluation across four key categories of agentic capability in a communicative collaboration setting. By testing both agent-agent and agent-human collaborations using open-source and closed-source models, our findings reveal surprising weaknesses in state-of-the-art models, including proprietary models like GPT-4o. These models struggle to outperform even a simple random agent baseline in agent-agent collaboration and only surpass the random baseline when a human is involved."
                    },
                    {
                        "title": "BACON: Supercharge Your VLM with Bag-of-Concept Graph to Mitigate Hallucinations",
                        "abstract": "This paper presents Bag-of-Concept Graph (BACON) to gift models with limited linguistic abilities to taste the privilege of Vision Language Models (VLMs) and boost downstream tasks such as detection, visual question answering (VQA), and image generation. Since the visual scenes in physical worlds are structured with complex relations between objects, BACON breaks down annotations into basic minimum elements and presents them in a graph structure. Element-wise style enables easy understanding, and structural composition liberates difficult locating. Careful prompt design births the BACON captions with the help of public-available VLMs and segmentation methods. In this way, we gather a dataset with 100K annotated images, which endow VLMs with remarkable capabilities, such as accurately generating BACON, transforming prompts into BACON format, envisioning scenarios in the style of BACONr, and dynamically modifying elements within BACON through interactive dialogue and more. Wide representative experiments, including detection, VQA, and image generation tasks, tell BACON as a lifeline to achieve previous out-of-reach tasks or excel in their current cutting-edge solutions."
                    },
                    {
                        "title": "OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments",
                        "abstract": "Autonomous agents that accomplish complex computer tasks with minimal human interventions have the potential to transform human-computer interaction, significantly enhancing accessibility and productivity. However, existing benchmarks either lack an interactive environment or are limited to environments specific to certain applications or domains, failing to reflect the diverse and complex nature of real-world computer use, thereby limiting the scope of tasks and agent scalability. To address this issue, we introduce OSWorld, the first-of-its-kind scalable, real computer environment for multimodal agents, supporting task setup, execution-based evaluation, and interactive learning across various operating systems such as Ubuntu, Windows, and macOS. OSWorld can serve as a unified, integrated computer environment for assessing open-ended computer tasks that involve arbitrary applications. Building upon OSWorld, we create a benchmark of 369 computer tasks involving real web and desktop apps in open domains, OS file I/O, and workflows spanning multiple applications. Each task example is derived from real-world computer use cases and includes a detailed initial state setup configuration and a custom execution-based evaluation script for reliable, reproducible evaluation. Extensive evaluation of state-of-the-art LLM/VLM-based agents on OSWorld reveals significant deficiencies in their ability to serve as computer assistants. While humans can accomplish over 72.36% of the tasks, the best model achieves only 12.24% success, primarily struggling with GUI grounding and operational knowledge. Comprehensive analysis using OSWorld provides valuable insights for developing multimodal generalist agents that were not possible with previous benchmarks. Our code, environment, baseline models, and data are publicly available at https://os-world.github.io."
                    }
                ]
            },
            "7316e01a-c934-4b2a-b4df-eab3887e5519": {
                "pk": "7316e01a-c934-4b2a-b4df-eab3887e5519",
                "name": "Yeqiao Fu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "2c2a2fa0-a779-49c3-8d65-6196d71bd551": {
                "pk": "2c2a2fa0-a779-49c3-8d65-6196d71bd551",
                "name": "Hongcheng Gao",
                "collaborators": [
                    "Zhijie Deng",
                    "Yangyi Chen",
                    "Zhiyuan Liu",
                    "Hao Zhang",
                    "Maosong Sun",
                    "Ganqu Cui",
                    "Baolong Bi",
                    "Shenghua Liu",
                    "Yiwei Wang",
                    "Lingrui Mei"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Adversarial Attacks",
                    "Multimodal Models"
                ],
                "publications": [
                    {
                        "title": "Efficient Detection of LLM-generated Texts with a Bayesian Surrogate Model",
                        "abstract": "The detection of machine-generated text, especially from large language models (LLMs), is crucial in preventing serious social problems resulting from their misuse. Some methods train dedicated detectors on specific datasets but fall short in generalizing to unseen test data, while other zero-shot ones often yield suboptimal performance. Although the recent DetectGPT has shown promising detection performance, it suffers from significant inefficiency issues, as detecting a single candidate requires querying the source LLM with hundreds of its perturbations. This paper aims to bridge this gap. Concretely, we propose to incorporate a Bayesian surrogate model, which allows us to select typical samples based on Bayesian uncertainty and interpolate scores from typical samples to other samples, to improve query efficiency. Empirical results demonstrate that our method significantly outperforms existing approaches under a low query budget. Notably, when detecting the text generated by LLaMA family models, our method with just 2 or 3 queries can outperform DetectGPT with 200 queries."
                    },
                    {
                        "title": "Evaluating the Robustness of Text-to-image Diffusion Models against Real-world Attacks",
                        "abstract": "Text-to-image (T2I) diffusion models (DMs) have shown promise in generating high-quality images from textual descriptions. The real-world applications of these models require particular attention to their safety and fidelity, but this has not been sufficiently explored. One fundamental question is whether existing T2I DMs are robust against variations over input texts. To answer it, this work provides the first robustness evaluation of T2I DMs against real-world attacks. Unlike prior studies that focus on malicious attacks involving apocryphal alterations to the input texts, we consider an attack space spanned by realistic errors (e.g., typo, glyph, phonetic) that humans can make, to ensure semantic consistency. Given the inherent randomness of the generation process, we develop novel distribution-based attack objectives to mislead T2I DMs. We perform attacks in a black-box manner without any knowledge of the model. Extensive experiments demonstrate the effectiveness of our method for attacking popular T2I DMs and simultaneously reveal their non-trivial robustness issues. Moreover, we provide an in-depth analysis of our method to show that it is not designed to attack the text encoder in T2I DMs solely."
                    },
                    {
                        "title": "Textual Backdoor Attacks Can Be More Harmful via Two Simple Tricks",
                        "abstract": "Backdoor attacks are a kind of emergent security threat in deep learning. After being injected with a backdoor, a deep neural model will behave normally on standard inputs but give adversary-specified predictions once the input contains specific backdoor triggers. In this paper, we find two simple tricks that can make existing textual backdoor attacks much more harmful. The first trick is to add an extra training task to distinguish poisoned and clean data during the training of the victim model, and the second one is to use all the clean training data rather than remove the original clean data corresponding to the poisoned data. These two tricks are universally applicable to different attack models. We conduct experiments in three tough situations including clean data fine-tuning, low-poisoning-rate, and label-consistent attacks. Experimental results show that the two tricks can significantly improve attack performance. This paper exhibits the great potential harmfulness of backdoor attacks. All the code and data can be obtained at \\url{https://github.com/thunlp/StyleAttack}."
                    },
                    {
                        "title": "Exploring the Universal Vulnerability of Prompt-based Learning Paradigm",
                        "abstract": "Prompt-based learning paradigm bridges the gap between pre-training and fine-tuning, and works effectively under the few-shot setting. However, we find that this learning paradigm inherits the vulnerability from the pre-training stage, where model predictions can be misled by inserting certain triggers into the text. In this paper, we explore this universal vulnerability by either injecting backdoor triggers or searching for adversarial triggers on pre-trained language models using only plain text. In both scenarios, we demonstrate that our triggers can totally control or severely decrease the performance of prompt-based models fine-tuned on arbitrary downstream tasks, reflecting the universal vulnerability of the prompt-based learning paradigm. Further experiments show that adversarial triggers have good transferability among language models. We also find conventional fine-tuning models are not vulnerable to adversarial triggers constructed from pre-trained language models. We conclude by proposing a potential solution to mitigate our attack methods. Code and data are publicly available at https://github.com/leix28/prompt-universal-vulnerability"
                    },
                    {
                        "title": "Universal Prompt Optimizer for Safe Text-to-Image Generation",
                        "abstract": "Text-to-Image (T2I) models have shown great performance in generating images based on textual prompts. However, these models are vulnerable to unsafe input to generate unsafe content like sexual, harassment and illegal-activity images. Existing studies based on image checker, model fine-tuning and embedding blocking are impractical in real-world applications. Hence, we propose the first universal prompt optimizer for safe T2I (POSI) generation in black-box scenario. We first construct a dataset consisting of toxic-clean prompt pairs by GPT-3.5 Turbo. To guide the optimizer to have the ability of converting toxic prompt to clean prompt while preserving semantic information, we design a novel reward function measuring toxicity and text alignment of generated images and train the optimizer through Proximal Policy Optimization. Experiments show that our approach can effectively reduce the likelihood of various T2I models in generating inappropriate images, with no significant impact on text alignment. It is also flexible to be combined with methods to achieve better performance. Our code is available at https://github.com/wzongyu/POSI."
                    },
                    {
                        "title": "AdaMoE: Token-Adaptive Routing with Null Experts for Mixture-of-Experts Language Models",
                        "abstract": "Mixture of experts (MoE) has become the standard for constructing production-level large language models (LLMs) due to its promise to boost model capacity without causing significant overheads. Nevertheless, existing MoE methods usually enforce a constant top-k routing for all tokens, which is arguably restrictive because various tokens (e.g., \"<EOS>\" vs. \"apple\") may require various numbers of experts for feature abstraction. Lifting such a constraint can help make the most of limited resources and unleash the potential of the model for downstream tasks. In this sense, we introduce AdaMoE to realize token-adaptive routing for MoE, where different tokens are permitted to select a various number of experts. AdaMoE makes minimal modifications to the vanilla MoE with top-k routing -- it simply introduces a fixed number of null experts, which do not consume any FLOPs, to the expert set and increases the value of k. AdaMoE does not force each token to occupy a fixed number of null experts but ensures the average usage of the null experts with a load-balancing loss, leading to an adaptive number of null/true experts used by each token. AdaMoE exhibits a strong resemblance to MoEs with expert choice routing while allowing for trivial auto-regressive modeling. AdaMoE is easy to implement and can be effectively applied to pre-trained (MoE-)LLMs. Extensive studies show that AdaMoE can reduce average expert load (FLOPs) while achieving superior performance. For example, on the ARC-C dataset, applying our method to fine-tuning Mixtral-8x7B can reduce FLOPs by 14.5% while increasing accuracy by 1.69%."
                    },
                    {
                        "title": "Adaptive Token Biaser: Knowledge Editing via Biasing Key Entities",
                        "abstract": "The parametric knowledge memorized by large language models (LLMs) becomes outdated quickly. In-context editing (ICE) is currently the most effective method for updating the knowledge of LLMs. Recent advancements involve enhancing ICE by modifying the decoding strategy, obviating the need for altering internal model structures or adjusting external prompts. However, this enhancement operates across the entire sequence generation, encompassing a plethora of non-critical tokens. In this work, we introduce $\\textbf{A}$daptive $\\textbf{T}$oken $\\textbf{Bias}$er ($\\textbf{ATBias}$), a new decoding technique designed to enhance ICE. It focuses on the tokens that are mostly related to knowledge during decoding, biasing their logits by matching key entities related to new and parametric knowledge. Experimental results show that ATBias significantly enhances ICE performance, achieving up to a 32.3% improvement over state-of-the-art ICE methods while incurring only half the latency. ATBias not only improves the knowledge editing capabilities of ICE but can also be widely applied to LLMs with negligible cost."
                    },
                    {
                        "title": "StruEdit: Structured Outputs Enable the Fast and Accurate Knowledge Editing for Large Language Models",
                        "abstract": "As the modern tool of choice for question answering, large language models (LLMs) are expected to deliver answers with up-to-date knowledge. To achieve such ideal question-answering systems, locating and then editing outdated knowledge in the natural language outputs is a general target of popular knowledge editing methods. However, this target is challenging, as both identifying which tokens to edit in the reasoning steps and ensuring the coherence of the revised reasoning chain are difficult tasks. We argue that these challenges stem from the unstructured nature of natural language outputs. To address the above challenges, we propose $\\textbf{Stru}$ctural $\\textbf{Edit}$ing ($\\textbf{StruEdit}$), an improved baseline for knowledge editing. We first prompt LLMs to produce structured outputs consisting of reasoning triplets. Then, StruEdit removes any potentially outdated knowledge and efficiently refills the structured outputs with up-to-date information in a single step. Experimental results show that StruEdit consistently delivers the highest accuracy with lowest latency compared with other knowledge editing methods."
                    },
                    {
                        "title": "Meta-Unlearning on Diffusion Models: Preventing Relearning Unlearned Concepts",
                        "abstract": "With the rapid progress of diffusion-based content generation, significant efforts are being made to unlearn harmful or copyrighted concepts from pretrained diffusion models (DMs) to prevent potential model misuse. However, it is observed that even when DMs are properly unlearned before release, malicious finetuning can compromise this process, causing DMs to relearn the unlearned concepts. This occurs partly because certain benign concepts (e.g., \"skin\") retained in DMs are related to the unlearned ones (e.g., \"nudity\"), facilitating their relearning via finetuning. To address this, we propose meta-unlearning on DMs. Intuitively, a meta-unlearned DM should behave like an unlearned DM when used as is; moreover, if the meta-unlearned DM undergoes malicious finetuning on unlearned concepts, the related benign concepts retained within it will be triggered to self-destruct, hindering the relearning of unlearned concepts. Our meta-unlearning framework is compatible with most existing unlearning methods, requiring only the addition of an easy-to-implement meta objective. We validate our approach through empirical experiments on meta-unlearning concepts from Stable Diffusion models (SD-v1-4 and SDXL), supported by extensive ablation studies. Our code is available at https://github.com/sail-sg/Meta-Unlearning."
                    },
                    {
                        "title": "Why Should Adversarial Perturbations be Imperceptible? Rethink the Research Paradigm in Adversarial NLP",
                        "abstract": "Textual adversarial samples play important roles in multiple subfields of NLP research, including security, evaluation, explainability, and data augmentation. However, most work mixes all these roles, obscuring the problem definitions and research goals of the security role that aims to reveal the practical concerns of NLP models. In this paper, we rethink the research paradigm of textual adversarial samples in security scenarios. We discuss the deficiencies in previous work and propose our suggestions that the research on the Security-oriented adversarial NLP (SoadNLP) should: (1) evaluate their methods on security tasks to demonstrate the real-world concerns; (2) consider real-world attackers' goals, instead of developing impractical methods. To this end, we first collect, process, and release a security datasets collection Advbench. Then, we reformalize the task and adjust the emphasis on different goals in SoadNLP. Next, we propose a simple method based on heuristic rules that can easily fulfill the actual adversarial goals to simulate real-world attack methods. We conduct experiments on both the attack and the defense sides on Advbench. Experimental results show that our method has higher practical value, indicating that the research paradigm in SoadNLP may start from our new benchmark. All the code and data of Advbench can be obtained at \\url{https://github.com/thunlp/Advbench}."
                    },
                    {
                        "title": "Emu: Generative Pretraining in Multimodality",
                        "abstract": "We present Emu, a Transformer-based multimodal foundation model, which can seamlessly generate images and texts in multimodal context. This omnivore model can take in any single-modality or multimodal data input indiscriminately (e.g., interleaved image, text and video) through a one-model-for-all autoregressive training process. First, visual signals are encoded into embeddings, and together with text tokens form an interleaved input sequence. Emu is then end-to-end trained with a unified objective of classifying the next text token or regressing the next visual embedding in the multimodal sequence. This versatile multimodality empowers the exploration of diverse pretraining data sources at scale, such as videos with interleaved frames and text, webpages with interleaved images and text, as well as web-scale image-text pairs and video-text pairs. Emu can serve as a generalist multimodal interface for both image-to-text and text-to-image tasks, and supports in-context image and text generation. Across a broad range of zero-shot/few-shot tasks including image captioning, visual question answering, video question answering and text-to-image generation, Emu demonstrates superb performance compared to state-of-the-art large multimodal models. Extended capabilities such as multimodal assistants via instruction tuning are also demonstrated with impressive performance."
                    },
                    {
                        "title": "Is Factuality Enhancement a Free Lunch For LLMs? Better Factuality Can Lead to Worse Context-Faithfulness",
                        "abstract": "As the modern tools of choice for text understanding and generation, large language models (LLMs) are expected to accurately output answers by leveraging the input context. This requires LLMs to possess both context-faithfulness and factual accuracy. Extensive efforts have been made to enable better outputs from LLMs by mitigating hallucinations through factuality enhancement methods. However, they also pose risks of hindering context-faithfulness, as factuality enhancement can lead LLMs to become overly confident in their parametric knowledge, causing them to overlook the relevant input context. In this work, we argue that current factuality enhancement methods can significantly undermine the context-faithfulness of LLMs. We first revisit the current factuality enhancement methods and evaluate their effectiveness in enhancing factual accuracy. Next, we evaluate their performance on knowledge editing tasks to assess the potential impact on context-faithfulness. The experimental results reveal that while these methods may yield inconsistent improvements in factual accuracy, they also cause a more severe decline in context-faithfulness, with the largest decrease reaching a striking 69.7\\%. To explain these declines, we analyze the hidden states and logit distributions for the tokens representing new knowledge and parametric knowledge respectively, highlighting the limitations of current approaches. Our finding highlights the complex trade-offs inherent in enhancing LLMs. Therefore, we recommend that more research on LLMs' factuality enhancement make efforts to reduce the sacrifice of context-faithfulness."
                    },
                    {
                        "title": "Revisiting Out-of-distribution Robustness in NLP: Benchmark, Analysis, and LLMs Evaluations",
                        "abstract": "This paper reexamines the research on out-of-distribution (OOD) robustness in the field of NLP. We find that the distribution shift settings in previous studies commonly lack adequate challenges, hindering the accurate evaluation of OOD robustness. To address these issues, we propose a benchmark construction protocol that ensures clear differentiation and challenging distribution shifts. Then we introduce BOSS, a Benchmark suite for Out-of-distribution robustneSS evaluation covering 5 tasks and 20 datasets. Based on BOSS, we conduct a series of experiments on pre-trained language models for analysis and evaluation of OOD robustness. First, for vanilla fine-tuning, we examine the relationship between in-distribution (ID) and OOD performance. We identify three typical types that unveil the inner learning mechanism, which could potentially facilitate the forecasting of OOD robustness, correlating with the advancements on ID datasets. Then, we evaluate 5 classic methods on BOSS and find that, despite exhibiting some effectiveness in specific cases, they do not offer significant improvement compared to vanilla fine-tuning. Further, we evaluate 5 LLMs with various adaptation paradigms and find that when sufficient ID data is available, fine-tuning domain-specific models outperform LLMs on ID examples significantly. However, in the case of OOD instances, prioritizing LLMs with in-context learning yields better results. We identify that both fine-tuned small models and LLMs face challenges in effectively addressing downstream tasks. The code is public at \\url{https://github.com/lifan-yuan/OOD_NLP}."
                    }
                ]
            },
            "69dd7f81-e934-4833-914c-d2837d59cc21": {
                "pk": "69dd7f81-e934-4833-914c-d2837d59cc21",
                "name": "Xinzhuang Xiong",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "c3887e69-b571-4e0b-b7c6-2a1248f9d69c": {
                "pk": "c3887e69-b571-4e0b-b7c6-2a1248f9d69c",
                "name": "Hanchong Zhang",
                "collaborators": [
                    "Ruisheng Cao",
                    "Lu Chen",
                    "Kai Yu",
                    "Hongshen Xu",
                    "Jieyu Li",
                    "Su Zhu",
                    "Peiyuan Feng",
                    "Yichen He",
                    "Guanhua Huang",
                    "Yuan Lin"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Text-to-SQL",
                    "Large Language Models",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "ACT-SQL: In-Context Learning for Text-to-SQL with Automatically-Generated Chain-of-Thought",
                        "abstract": "Recently Large Language Models (LLMs) have been proven to have strong abilities in various domains and tasks. We study the problem of prompt designing in the text-to-SQL task and attempt to improve the LLMs' reasoning ability when generating SQL queries. Besides the trivial few-shot in-context learning setting, we design our chain-of-thought (CoT) prompt with a similar method to schema linking. We provide a method named ACT-SQL to automatically generate auto-CoT exemplars and thus the whole process doesn't need manual labeling. Our approach is cost-saving since we only use the LLMs' API call once when generating one SQL query. Furthermore, we extend our in-context learning method to the multi-turn text-to-SQL task. The experiment results show that the LLMs' performance can benefit from our ACT-SQL approach. Our approach achieves SOTA performance on the Spider dev set among existing in-context learning approaches."
                    },
                    {
                        "title": "CoE-SQL: In-Context Learning for Multi-Turn Text-to-SQL with Chain-of-Editions",
                        "abstract": "Recently, Large Language Models (LLMs) have been demonstrated to possess impressive capabilities in a variety of domains and tasks. We investigate the issue of prompt design in the multi-turn text-to-SQL task and attempt to enhance the LLMs' reasoning capacity when generating SQL queries. In the conversational context, the current SQL query can be modified from the preceding SQL query with only a few operations due to the context dependency. We introduce our method called CoE-SQL which can prompt LLMs to generate the SQL query based on the previously generated SQL query with an edition chain. We also conduct extensive ablation studies to determine the optimal configuration of our approach. Our approach outperforms different in-context learning baselines stably and achieves state-of-the-art performances on two benchmarks SParC and CoSQL using LLMs, which is also competitive to the SOTA fine-tuned models."
                    },
                    {
                        "title": "AGILE: A Novel Framework of LLM Agents",
                        "abstract": "We introduce a novel framework of LLM agents named AGILE (AGent that Interacts and Learns from Environments) designed to perform complex conversational tasks with users, leveraging LLMs, memory, tools, and interactions with experts. The agent's abilities include not only conversation but also reflection, utilization of tools, and consultation with experts. We formulate the construction of such an LLM agent as a reinforcement learning problem, in which the LLM serves as the policy model. We fine-tune the LLM using labeled data of actions and the PPO algorithm. We focus on question answering and release a dataset for agents called ProductQA, comprising challenging questions in online shopping. Our extensive experiments on ProductQA and MedMCQA show that AGILE agents based on 13B and 7B LLMs trained with PPO can outperform GPT-4 agents. Our ablation study highlights the indispensability of memory, tools, consultation, reflection, and reinforcement learning in achieving the agent's strong performance."
                    },
                    {
                        "title": "CSS: A Large-scale Cross-schema Chinese Text-to-SQL Medical Dataset",
                        "abstract": "The cross-domain text-to-SQL task aims to build a system that can parse user questions into SQL on complete unseen databases, and the single-domain text-to-SQL task evaluates the performance on identical databases. Both of these setups confront unavoidable difficulties in real-world applications. To this end, we introduce the cross-schema text-to-SQL task, where the databases of evaluation data are different from that in the training data but come from the same domain. Furthermore, we present CSS, a large-scale CrosS-Schema Chinese text-to-SQL dataset, to carry on corresponding studies. CSS originally consisted of 4,340 question/SQL pairs across 2 databases. In order to generalize models to different medical systems, we extend CSS and create 19 new databases along with 29,280 corresponding dataset examples. Moreover, CSS is also a large corpus for single-domain Chinese text-to-SQL studies. We present the data collection approach and a series of analyses of the data statistics. To show the potential and usefulness of CSS, benchmarking baselines have been conducted and reported. Our dataset is publicly available at \\url{https://huggingface.co/datasets/zhanghanchong/css}."
                    },
                    {
                        "title": "ASTormer: An AST Structure-aware Transformer Decoder for Text-to-SQL",
                        "abstract": "Text-to-SQL aims to generate an executable SQL program given the user utterance and the corresponding database schema. To ensure the well-formedness of output SQLs, one prominent approach adopts a grammar-based recurrent decoder to produce the equivalent SQL abstract syntax tree (AST). However, previous methods mainly utilize an RNN-series decoder, which 1) is time-consuming and inefficient and 2) introduces very few structure priors. In this work, we propose an AST structure-aware Transformer decoder (ASTormer) to replace traditional RNN cells. The structural knowledge, such as node types and positions in the tree, is seamlessly incorporated into the decoder via both absolute and relative position embeddings. Besides, the proposed framework is compatible with different traversing orders even considering adaptive node selection. Extensive experiments on five text-to-SQL benchmarks demonstrate the effectiveness and efficiency of our structured decoder compared to competitive baselines."
                    },
                    {
                        "title": "A BiRGAT Model for Multi-intent Spoken Language Understanding with Hierarchical Semantic Frames",
                        "abstract": "Previous work on spoken language understanding (SLU) mainly focuses on single-intent settings, where each input utterance merely contains one user intent. This configuration significantly limits the surface form of user utterances and the capacity of output semantics. In this work, we first propose a Multi-Intent dataset which is collected from a realistic in-Vehicle dialogue System, called MIVS. The target semantic frame is organized in a 3-layer hierarchical structure to tackle the alignment and assignment problems in multi-intent cases. Accordingly, we devise a BiRGAT model to encode the hierarchy of ontology items, the backbone of which is a dual relational graph attention network. Coupled with the 3-way pointer-generator decoder, our method outperforms traditional sequence labeling and classification-based schemes by a large margin."
                    },
                    {
                        "title": "On the Structural Generalization in Text-to-SQL",
                        "abstract": "Exploring the generalization of a text-to-SQL parser is essential for a system to automatically adapt the real-world databases. Previous works provided investigations focusing on lexical diversity, including the influence of the synonym and perturbations in both natural language questions and databases. However, research on the structure variety of database schema~(DS) is deficient. Specifically, confronted with the same input question, the target SQL is probably represented in different ways when the DS comes to a different structure. In this work, we provide in-deep discussions about the structural generalization of text-to-SQL tasks. We observe that current datasets are too templated to study structural generalization. To collect eligible test data, we propose a framework to generate novel text-to-SQL data via automatic and synchronous (DS, SQL) pair altering. In the experiments, significant performance reduction when evaluating well-trained text-to-SQL models on the synthetic samples demonstrates the limitation of current research regarding structural generalization. According to comprehensive analysis, we suggest the practical reason is the overfitting of (NL, SQL) patterns."
                    }
                ]
            },
            "0efff2c1-b98f-4a02-a16c-a102d7143d15": {
                "pk": "0efff2c1-b98f-4a02-a16c-a102d7143d15",
                "name": "Yuchen Mao",
                "collaborators": [
                    "Lin Chen",
                    "Jiayi Lian",
                    "Guochuan Zhang",
                    "Siu-Wing Cheng",
                    "Zhongkai Tao",
                    "Sung-Jin Oh",
                    "Zhanpeng Zhou",
                    "Mingze Wang",
                    "Bingrui Li",
                    "Junchi Yan"
                ],
                "domain": [
                    "Algorithm Design",
                    "Approximation Algorithms",
                    "General Relativity",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Localized initial data for Einstein equations",
                        "abstract": "We apply a new method with explicit solution operators to construct asymptotically flat initial data sets of the vacuum Einstein equation with new localization properties. Applications include an improvement of the decay rate in Carlotto--Schoen [arXiv:1407.4766] to $\\mathcal{O}(|x|^{-(d-2)})$ and a construction of nontrivial asymptotically flat initial data supported in a degenerate sector $\\{(x',x_d)\\in\\mathbb{R}^d:|x'|\\leq x_d^\\alpha\\}$ for $\\frac{3}{d+1}<\\alpha<1$."
                    },
                    {
                        "title": "A Nearly Quadratic-Time FPTAS for Knapsack",
                        "abstract": "We investigate polynomial-time approximation schemes for the classic 0-1 knapsack problem. The previous algorithm by Deng, Jin, and Mao (SODA'23) has approximation factor $1 + \\eps$ with running time $\\widetilde{O}(n + \\frac{1}{\\eps^{2.2}})$. There is a lower Bound of $(n + \\frac{1}{\\eps})^{2-o(1)}$ conditioned on the hypothesis that $(\\min, +)$ has no truly subquadratic algorithm. We close the gap by proposing an approximation scheme that runs in $\\widetilde{O}(n + \\frac{1}{\\eps^2})$ time."
                    },
                    {
                        "title": "Approximating Partition in Near-Linear Time",
                        "abstract": "We propose an $\\widetilde{O}(n + 1/\\eps)$-time FPTAS (Fully Polynomial-Time Approximation Scheme) for the classical Partition problem. This is the best possible (up to a polylogarithmic factor) assuming SETH (Strong Exponential Time Hypothesis) [Abboud, Bringmann, Hermelin, and Shabtay'22]. Prior to our work, the best known FPTAS for Partition runs in $\\widetilde{O}(n + 1/\\eps^{5/4})$ time [Deng, Jin and Mao'23, Wu and Chen'22]. Our result is obtained by solving a more general problem of weakly approximating Subset Sum."
                    },
                    {
                        "title": "Restricted Max-Min Fair Allocation",
                        "abstract": "The restricted max-min fair allocation problem seeks an allocation of resources to players that maximizes the minimum total value obtained by any player. It is NP-hard to approximate the problem to a ratio less than 2. Comparing the current best algorithm for estimating the optimal value with the current best for constructing an allocation, there is quite a gap between the ratios that can be achieved in polynomial time: roughly 4 for estimation and roughly $6 + 2\\sqrt{10}$ for construction. We propose an algorithm that constructs an allocation with value within a factor of $6 + \\delta$ from the optimum for any constant $\\delta > 0$. The running time is polynomial in the input size for any constant $\\delta$ chosen."
                    },
                    {
                        "title": "Integrality Gap of the Configuration LP for the Restricted Max-Min Fair Allocation",
                        "abstract": "The max-min fair allocation problem seeks an allocation of resources to players that maximizes the minimum total value obtained by any player. Each player $p$ has a non-negative value $v_{pr}$ on resource $r$. In the restricted case, we have $v_{pr}\\in \\{v_r, 0\\}$. That is, a resource $r$ is worth value $v_r$ for the players who desire it and value 0 for the other players. In this paper, we consider the configuration LP, a linear programming relaxation for the restricted problem. The integrality gap of the configuration LP is at least $2$. Asadpour, Feige, and Saberi proved an upper bound of $4$. We improve the upper bound to $23/6$ using the dual of the configuration LP. Since the configuration LP can be solved to any desired accuracy $\\delta$ in polynomial time, our result leads to a polynomial-time algorithm which estimates the optimal value within a factor of $23/6+\\delta$."
                    },
                    {
                        "title": "Restricted Max-Min Allocation: Approximation and Integrality Gap",
                        "abstract": "Asadpour, Feige, and Saberi proved that the integrality gap of the configuration LP for the restricted max-min allocation problem is at most $4$. However, their proof does not give a polynomial-time approximation algorithm. A lot of efforts have been devoted to designing an efficient algorithm whose approximation ratio can match this upper bound for the integrality gap. In ICALP 2018, we present a $(6 + \\delta)$-approximation algorithm where $\\delta$ can be any positive constant, and there is still a gap of roughly $2$. In this paper, we narrow the gap significantly by proposing a $(4+\\delta)$-approximation algorithm where $\\delta$ can be any positive constant. The approximation ratio is with respect to the optimal value of the configuration LP, and the running time is $\\mathit{poly}(m,n)\\cdot n^{\\mathit{poly}(\\frac{1}{\\delta})}$ where $n$ is the number of players and $m$ is the number of resources. We also improve the upper bound for the integrality gap of the configuration LP to $3 + \\frac{21}{26} \\approx 3.808$."
                    },
                    {
                        "title": "An Improved Pseudopolynomial Time Algorithm for Subset Sum",
                        "abstract": "We investigate pseudo-polynomial time algorithms for Subset Sum. Given a multi-set $X$ of $n$ positive integers and a target $t$, Subset Sum asks whether some subset of $X$ sums to $t$. Bringmann proposes an $\\tilde{O}(n + t)$-time algorithm [Bringmann SODA'17], and an open question has naturally arisen: can Subset Sum be solved in $O(n + w)$ time? Here $w$ is the maximum integer in $X$. We make a progress towards resolving the open question by proposing an $\\tilde{O}(n + \\sqrt{wt})$-time algorithm."
                    },
                    {
                        "title": "Faster Algorithms for Bounded Knapsack and Bounded Subset Sum Via Fine-Grained Proximity Results",
                        "abstract": "We investigate pseudopolynomial-time algorithms for Bounded Knapsack and Bounded Subset Sum. Recent years have seen a growing interest in settling their fine-grained complexity with respect to various parameters. For Bounded Knapsack, the number of items $n$ and the maximum item weight $w_{\\max}$ are two of the most natural parameters that have been studied extensively in the literature. The previous best running time in terms of $n$ and $w_{\\max}$ is $O(n + w^3_{\\max})$ [Polak, Rohwedder, Wegrzycki '21]. There is a conditional lower bound of $O((n + w_{\\max})^{2-o(1)})$ based on $(\\min,+)$-convolution hypothesis [Cygan, Mucha, Wegrzycki, Wlodarczyk '17]. We narrow the gap significantly by proposing a $\\tilde{O}(n + w^{12/5}_{\\max})$-time algorithm. Note that in the regime where $w_{\\max} \\approx n$, our algorithm runs in $\\tilde{O}(n^{12/5})$ time, while all the previous algorithms require $\\Omega(n^3)$ time in the worst case.   For Bounded Subset Sum, we give two algorithms running in $\\tilde{O}(nw_{\\max})$ and $\\tilde{O}(n + w^{3/2}_{\\max})$ time, respectively. These results match the currently best running time for 0-1 Subset Sum. Prior to our work, the best running times (in terms of $n$ and $w_{\\max}$) for Bounded Subset Sum is $\\tilde{O}(n + w^{5/3}_{\\max})$ [Polak, Rohwedder, Wegrzycki '21] and $\\tilde{O}(n + \\mu_{\\max}^{1/2}w_{\\max}^{3/2})$ [implied by Bringmann '19 and Bringmann, Wellnitz '21], where $\\mu_{\\max}$ refers to the maximum multiplicity of item weights."
                    },
                    {
                        "title": "Initial data gluing in the asymptotically flat regime via solution operators with prescribed support properties",
                        "abstract": "We give new proofs of general relativistic initial data gluing results on unit-scale annuli based on explicit solution operators for the linearized constraint equation around the flat case with prescribed support properties. These results retrieve and optimize - in terms of positivity, regularity, size and/or spatial decay requirements - a number of known theorems concerning asymptotically flat initial data, including Kerr exterior gluing by Corvino-Schoen and Chru\\'sciel-Delay, interior gluing (or \"fill-in\") by Bieri-Chru\\'sciel, and obstruction-free gluing by Czimek-Rodnianski. In particular, our proof of the strengthened obstruction-free gluing theorem relies on purely spacelike techniques, rather than null gluing as in the original approach."
                    },
                    {
                        "title": "Sharpness-Aware Minimization Efficiently Selects Flatter Minima Late in Training",
                        "abstract": "Sharpness-Aware Minimization (SAM) has substantially improved the generalization of neural networks under various settings. Despite the success, its effectiveness remains poorly understood. In this work, we discover an intriguing phenomenon in the training dynamics of SAM, shedding lights on understanding its implicit bias towards flatter minima over Stochastic Gradient Descent (SGD). Specifically, we find that SAM efficiently selects flatter minima late in training. Remarkably, even a few epochs of SAM applied at the end of training yield nearly the same generalization and solution sharpness as full SAM training. Subsequently, we delve deeper into the underlying mechanism behind this phenomenon. Theoretically, we identify two phases in the learning dynamics after applying SAM late in training: i) SAM first escapes the minimum found by SGD exponentially fast; and ii) then rapidly converges to a flatter minimum within the same valley. Furthermore, we empirically investigate the role of SAM during the early training phase. We conjecture that the optimization method chosen in the late phase is more crucial in shaping the final solution's properties. Based on this viewpoint, we extend our findings from SAM to Adversarial Training."
                    }
                ]
            },
            "30193ed4-daff-405e-83ce-6203b30a441c": {
                "pk": "30193ed4-daff-405e-83ce-6203b30a441c",
                "name": "Wenjing Hu",
                "collaborators": [
                    "Rui Su",
                    "Wenjing Huang",
                    "Haoyu Ma",
                    "Xiaowei Song",
                    "Jinglu Hu",
                    "Jina Park",
                    "Ick Hoon Jin",
                    "Hao Liu",
                    "Yong Zang"
                ],
                "domain": [
                    "Video Object Detection",
                    "Bayesian Statistics",
                    "Machine Learning",
                    "Medical Research"
                ],
                "publications": [
                    {
                        "title": "SGE net: Video object detection with squeezed GRU and information entropy map",
                        "abstract": "Recently, deep learning based video object detection has attracted more and more attention. Compared with object detection of static images, video object detection is more challenging due to the motion of objects, while providing rich temporal information. The RNN-based algorithm is an effective way to enhance detection performance in videos with temporal information. However, most studies in this area only focus on accuracy while ignoring the calculation cost and the number of parameters.   In this paper, we propose an efficient method that combines channel-reduced convolutional GRU (Squeezed GRU), and Information Entropy map for video object detection (SGE-Net). The experimental results validate the accuracy improvement, computational savings of the Squeezed GRU, and superiority of the information entropy attention mechanism on the classification performance. The mAP has increased by 3.7 contrasted with the baseline, and the number of parameters has decreased from 6.33 million to 0.67 million compared with the standard GRU."
                    },
                    {
                        "title": "A Bayesian Precision Response-adaptive Phase II Clinical Trial Design for Radiotherapies with Competing Risk Survival Outcomes",
                        "abstract": "Many phase II clinical trials have used survival outcomes as the primary endpoints in recent decades. Suppose the radiotherapy is evaluated in a phase II trial using survival outcomes. In that case, the competing risk issue often arises because the time to disease progression can be censored by the time to normal tissue complications, and vice versa. Besides, much literature has examined that patients receiving the same radiotherapy dose may yield distinct responses due to their heterogeneous radiation susceptibility statuses. Therefore, the \"one-dose-fit-all\" strategy often fails, and it is more relevant to evaluate the subgroup-specific treatment effect with the subgroup defined by the radiation susceptibility status. In this paper, we propose a Bayesian precision phase II trial design evaluating the subgroup-specific treatment effects of radiotherapy. We use the cause-specific hazard approach to model the competing risk survival outcomes. We propose restricting the candidate radiation doses based on each patient's radiation susceptibility status. Only the clinically feasible personalized dose will be considered, which enhances the benefit for the patients in the trial. In addition, we propose a stratified Bayesian adaptive randomization scheme such that more patients will be randomized to the dose reporting more favorable survival outcomes. Numerical studies have shown that the proposed design performed well and outperformed the conventional design ignoring the competing risk issue."
                    }
                ]
            },
            "30005f3a-fb16-4c19-98fd-852990cb9a99": {
                "pk": "30005f3a-fb16-4c19-98fd-852990cb9a99",
                "name": "Tianbao Xie",
                "collaborators": [
                    "Libo Qin",
                    "Wanxiang Che",
                    "Ting Liu",
                    "Qiguang Chen",
                    "Xiao Xu",
                    "Tao Yu",
                    "Shijue Huang",
                    "Fuxuan Wei",
                    "Yiru Chen",
                    "Ryan Li"
                ],
                "domain": [
                    "Spoken Language Understanding",
                    "Task-oriented Dialogue",
                    "Natural Language Processing",
                    "Zero-shot Learning"
                ],
                "publications": [
                    {
                        "title": "A Survey on Spoken Language Understanding: Recent Advances and New Frontiers",
                        "abstract": "Spoken Language Understanding (SLU) aims to extract the semantics frame of user queries, which is a core component in a task-oriented dialog system. With the burst of deep neural networks and the evolution of pre-trained language models, the research of SLU has obtained significant breakthroughs. However, there remains a lack of a comprehensive survey summarizing existing approaches and recent trends, which motivated the work presented in this article. In this paper, we survey recent advances and new frontiers in SLU. Specifically, we give a thorough review of this research field, covering different aspects including (1) new taxonomy: we provide a new perspective for SLU filed, including single model vs. joint model, implicit joint modeling vs. explicit joint modeling in joint model, non pre-trained paradigm vs. pre-trained paradigm;(2) new frontiers: some emerging areas in complex SLU as well as the corresponding challenges; (3) abundant open-source resources: to help the community, we have collected, organized the related papers, baseline projects and leaderboard on a public website where SLU researchers could directly access to the recent progress. We hope that this survey can shed a light on future research in SLU field."
                    },
                    {
                        "title": "Don't be Contradicted with Anything! CI-ToD: Towards Benchmarking Consistency for Task-oriented Dialogue System",
                        "abstract": "Consistency Identification has obtained remarkable success on open-domain dialogue, which can be used for preventing inconsistent response generation. However, in contrast to the rapid development in open-domain dialogue, few efforts have been made to the task-oriented dialogue direction. In this paper, we argue that consistency problem is more urgent in task-oriented domain. To facilitate the research, we introduce CI-ToD, a novel dataset for Consistency Identification in Task-oriented Dialog system. In addition, we not only annotate the single label to enable the model to judge whether the system response is contradictory, but also provide more fine-grained labels (i.e., Dialogue History Inconsistency, User Query Inconsistency and Knowledge Base Inconsistency) to encourage model to know what inconsistent sources lead to it. Empirical results show that state-of-the-art methods only achieve 51.3%, which is far behind the human performance of 93.2%, indicating that there is ample room for improving consistency identification ability. Finally, we conduct exhaustive experiments and qualitative analysis to comprehend key challenges and provide guidance for future directions. All datasets and models are publicly available at \\url{https://github.com/yizhen20133868/CI-ToD}."
                    },
                    {
                        "title": "GL-GIN: Fast and Accurate Non-Autoregressive Model for Joint Multiple Intent Detection and Slot Filling",
                        "abstract": "Multi-intent SLU can handle multiple intents in an utterance, which has attracted increasing attention. However, the state-of-the-art joint models heavily rely on autoregressive approaches, resulting in two issues: slow inference speed and information leakage. In this paper, we explore a non-autoregressive model for joint multiple intent detection and slot filling, achieving more fast and accurate. Specifically, we propose a Global-Locally Graph Interaction Network (GL-GIN) where a local slot-aware graph interaction layer is proposed to model slot dependency for alleviating uncoordinated slots problem while a global intent-slot graph interaction layer is introduced to model the interaction between multiple intents and all slots in the utterance. Experimental results on two public datasets show that our framework achieves state-of-the-art performance while being 11.5 times faster."
                    },
                    {
                        "title": "NL2INTERFACE: Interactive Visualization Interface Generation from Natural Language Queries",
                        "abstract": "We develop NL2INTERFACE to explore the potential of generating usable interactive multi-visualization interfaces from natural language queries. With NL2INTERFACE, users can directly write natural language queries to automatically generate a fully interactive multi-visualization interface without any extra effort of learning a tool or programming language. Further, users can interact with the interfaces to easily transform the data and quickly see the results in the visualizations."
                    },
                    {
                        "title": "In-Context Learning for Few-Shot Dialogue State Tracking",
                        "abstract": "Collecting and annotating task-oriented dialogues is time-consuming and costly; thus, zero and few shot learning could greatly benefit dialogue state tracking (DST). In this work, we propose an in-context learning (ICL) framework for zero-shot and few-shot learning DST, where a large pre-trained language model (LM) takes a test instance and a few exemplars as input, and directly decodes the dialogue state without any parameter updates. To better leverage a tabular domain description in the LM prompt, we reformulate DST into a text-to-SQL problem. We also propose a novel approach to retrieve annotated dialogues as exemplars. Empirical results on MultiWOZ show that our method IC-DST substantially outperforms previous fine-tuned state-of-the-art models in few-shot settings. In addition, we test IC-DST in zero-shot settings, in which the model only takes a fixed task instruction as input, finding that it outperforms previous zero-shot methods by a large margin."
                    },
                    {
                        "title": "GL-CLeF: A Global-Local Contrastive Learning Framework for Cross-lingual Spoken Language Understanding",
                        "abstract": "Due to high data demands of current methods, attention to zero-shot cross-lingual spoken language understanding (SLU) has grown, as such approaches greatly reduce human annotation effort. However, existing models solely rely on shared parameters, which can only perform implicit alignment across languages. We present Global--Local Contrastive Learning Framework (GL-CLeF) to address this shortcoming. Specifically, we employ contrastive learning, leveraging bilingual dictionaries to construct multilingual views of the same utterance, then encourage their representations to be more similar than negative example pairs, which achieves to explicitly aligned representations of similar sentences across languages. In addition, a key step in GL-CLeF is a proposed Local and Global component, which achieves a fine-grained cross-lingual transfer (i.e., sentence-level Local intent transfer, token-level Local slot transfer, and semantic-level Global transfer across intent and slot). Experiments on MultiATIS++ show that GL-CLeF achieves the best performance and successfully pulls representations of similar sentences across languages closer."
                    },
                    {
                        "title": "AgentStore: Scalable Integration of Heterogeneous Agents As Specialized Generalist Computer Assistant",
                        "abstract": "Digital agents capable of automating complex computer tasks have attracted considerable attention due to their immense potential to enhance human-computer interaction. However, existing agent methods exhibit deficiencies in their generalization and specialization capabilities, especially in handling open-ended computer tasks in real-world environments. Inspired by the rich functionality of the App store, we present AgentStore, a scalable platform designed to dynamically integrate heterogeneous agents for automating computer tasks. AgentStore empowers users to integrate third-party agents, allowing the system to continuously enrich its capabilities and adapt to rapidly evolving operating systems. Additionally, we propose a novel core \\textbf{MetaAgent} with the \\textbf{AgentToken} strategy to efficiently manage diverse agents and utilize their specialized and generalist abilities for both domain-specific and system-wide tasks. Extensive experiments on three challenging benchmarks demonstrate that AgentStore surpasses the limitations of previous systems with narrow capabilities, particularly achieving a significant improvement from 11.21\\% to 23.85\\% on the OSWorld benchmark, more than doubling the previous results. Comprehensive quantitative and qualitative results further demonstrate AgentStore's ability to enhance agent systems in both generalization and specialization, underscoring its potential for developing the specialized generalist computer assistant. All our codes will be made publicly available in https://chengyou-jia.github.io/AgentStore-Home."
                    }
                ]
            },
            "54c0254a-ff04-4ac9-929b-fb117fc33fd4": {
                "pk": "54c0254a-ff04-4ac9-929b-fb117fc33fd4",
                "name": "Hongshen Xu",
                "collaborators": [
                    "Lu Chen",
                    "Kai Yu",
                    "Ruisheng Cao",
                    "Hanchong Zhang",
                    "Da Ma",
                    "Zihan Zhao",
                    "Su Zhu",
                    "Zichen Zhu",
                    "Situo Zhang",
                    "Shuai Fan"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Model Compression",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Multilingual Brain Surgeon: Large Language Models Can be Compressed Leaving No Language Behind",
                        "abstract": "Large Language Models (LLMs) have ushered in a new era in Natural Language Processing, but their massive size demands effective compression techniques for practicality. Although numerous model compression techniques have been investigated, they typically rely on a calibration set that overlooks the multilingual context and results in significant accuracy degradation for low-resource languages. This paper introduces Multilingual Brain Surgeon (MBS), a novel calibration data sampling method for multilingual LLMs compression. MBS overcomes the English-centric limitations of existing methods by sampling calibration data from various languages proportionally to the language distribution of the model training datasets. Our experiments, conducted on the BLOOM multilingual LLM, demonstrate that MBS improves the performance of existing English-centric compression methods, especially for low-resource languages. We also uncover the dynamics of language interaction during compression, revealing that the larger the proportion of a language in the training set and the more similar the language is to the calibration language, the better performance the language retains after compression. In conclusion, MBS presents an innovative approach to compressing multilingual LLMs, addressing the performance disparities and improving the language inclusivity of existing compression techniques."
                    },
                    {
                        "title": "ACT-SQL: In-Context Learning for Text-to-SQL with Automatically-Generated Chain-of-Thought",
                        "abstract": "Recently Large Language Models (LLMs) have been proven to have strong abilities in various domains and tasks. We study the problem of prompt designing in the text-to-SQL task and attempt to improve the LLMs' reasoning ability when generating SQL queries. Besides the trivial few-shot in-context learning setting, we design our chain-of-thought (CoT) prompt with a similar method to schema linking. We provide a method named ACT-SQL to automatically generate auto-CoT exemplars and thus the whole process doesn't need manual labeling. Our approach is cost-saving since we only use the LLMs' API call once when generating one SQL query. Furthermore, we extend our in-context learning method to the multi-turn text-to-SQL task. The experiment results show that the LLMs' performance can benefit from our ACT-SQL approach. Our approach achieves SOTA performance on the Spider dev set among existing in-context learning approaches."
                    },
                    {
                        "title": "CoE-SQL: In-Context Learning for Multi-Turn Text-to-SQL with Chain-of-Editions",
                        "abstract": "Recently, Large Language Models (LLMs) have been demonstrated to possess impressive capabilities in a variety of domains and tasks. We investigate the issue of prompt design in the multi-turn text-to-SQL task and attempt to enhance the LLMs' reasoning capacity when generating SQL queries. In the conversational context, the current SQL query can be modified from the preceding SQL query with only a few operations due to the context dependency. We introduce our method called CoE-SQL which can prompt LLMs to generate the SQL query based on the previously generated SQL query with an edition chain. We also conduct extensive ablation studies to determine the optimal configuration of our approach. Our approach outperforms different in-context learning baselines stably and achieves state-of-the-art performances on two benchmarks SParC and CoSQL using LLMs, which is also competitive to the SOTA fine-tuned models."
                    },
                    {
                        "title": "Rejection Improves Reliability: Training LLMs to Refuse Unknown Questions Using RL from Knowledge Feedback",
                        "abstract": "Large Language Models (LLMs) often generate erroneous outputs, known as hallucinations, due to their limitations in discerning questions beyond their knowledge scope. While addressing hallucination has been a focal point in research, previous efforts primarily concentrate on enhancing correctness without giving due consideration to the significance of rejection mechanisms. In this paper, we conduct a comprehensive examination of the role of rejection, introducing the notion of model reliability along with corresponding metrics. These metrics measure the model's ability to provide accurate responses while adeptly rejecting questions exceeding its knowledge boundaries, thereby minimizing hallucinations. To improve the inherent reliability of LLMs, we present a novel alignment framework called Reinforcement Learning from Knowledge Feedback (RLKF). RLKF leverages knowledge feedback to dynamically determine the model's knowledge boundary and trains a reliable reward model to encourage the refusal of out-of-knowledge questions. Experimental results on mathematical questions affirm the substantial efficacy of RLKF in significantly enhancing LLM reliability."
                    },
                    {
                        "title": "TIE: Topological Information Enhanced Structural Reading Comprehension on Web Pages",
                        "abstract": "Recently, the structural reading comprehension (SRC) task on web pages has attracted increasing research interests. Although previous SRC work has leveraged extra information such as HTML tags or XPaths, the informative topology of web pages is not effectively exploited. In this work, we propose a Topological Information Enhanced model (TIE), which transforms the token-level task into a tag-level task by introducing a two-stage process (i.e. node locating and answer refining). Based on that, TIE integrates Graph Attention Network (GAT) and Pre-trained Language Model (PLM) to leverage the topological information of both logical structures and spatial structures. Experimental results demonstrate that our model outperforms strong baselines and achieves state-of-the-art performances on the web-based SRC benchmark WebSRC at the time of writing. The code of TIE will be publicly available at https://github.com/X-LANCE/TIE."
                    },
                    {
                        "title": "Large Language Models Are Semi-Parametric Reinforcement Learning Agents",
                        "abstract": "Inspired by the insights in cognitive science with respect to human memory and reasoning mechanism, a novel evolvable LLM-based (Large Language Model) agent framework is proposed as REMEMBERER. By equipping the LLM with a long-term experience memory, REMEMBERER is capable of exploiting the experiences from the past episodes even for different task goals, which excels an LLM-based agent with fixed exemplars or equipped with a transient working memory. We further introduce Reinforcement Learning with Experience Memory (RLEM) to update the memory. Thus, the whole system can learn from the experiences of both success and failure, and evolve its capability without fine-tuning the parameters of the LLM. In this way, the proposed REMEMBERER constitutes a semi-parametric RL agent. Extensive experiments are conducted on two RL task sets to evaluate the proposed framework. The average results with different initialization and training sets exceed the prior SOTA by 4% and 2% for the success rate on two task sets and demonstrate the superiority and robustness of REMEMBERER."
                    },
                    {
                        "title": "A BiRGAT Model for Multi-intent Spoken Language Understanding with Hierarchical Semantic Frames",
                        "abstract": "Previous work on spoken language understanding (SLU) mainly focuses on single-intent settings, where each input utterance merely contains one user intent. This configuration significantly limits the surface form of user utterances and the capacity of output semantics. In this work, we first propose a Multi-Intent dataset which is collected from a realistic in-Vehicle dialogue System, called MIVS. The target semantic frame is organized in a 3-layer hierarchical structure to tackle the alignment and assignment problems in multi-intent cases. Accordingly, we devise a BiRGAT model to encode the hierarchy of ontology items, the backbone of which is a dual relational graph attention network. Coupled with the 3-way pointer-generator decoder, our method outperforms traditional sequence labeling and classification-based schemes by a large margin."
                    },
                    {
                        "title": "Hierarchical Multimodal Pre-training for Visually Rich Webpage Understanding",
                        "abstract": "The growing prevalence of visually rich documents, such as webpages and scanned/digital-born documents (images, PDFs, etc.), has led to increased interest in automatic document understanding and information extraction across academia and industry. Although various document modalities, including image, text, layout, and structure, facilitate human information retrieval, the interconnected nature of these modalities presents challenges for neural networks. In this paper, we introduce WebLM, a multimodal pre-training network designed to address the limitations of solely modeling text and structure modalities of HTML in webpages. Instead of processing document images as unified natural images, WebLM integrates the hierarchical structure of document images to enhance the understanding of markup-language-based documents. Additionally, we propose several pre-training tasks to model the interaction among text, structure, and image modalities effectively. Empirical results demonstrate that the pre-trained WebLM significantly surpasses previous state-of-the-art pre-trained models across several webpage understanding tasks. The pre-trained models and code are available at https://github.com/X-LANCE/weblm."
                    },
                    {
                        "title": "ASTormer: An AST Structure-aware Transformer Decoder for Text-to-SQL",
                        "abstract": "Text-to-SQL aims to generate an executable SQL program given the user utterance and the corresponding database schema. To ensure the well-formedness of output SQLs, one prominent approach adopts a grammar-based recurrent decoder to produce the equivalent SQL abstract syntax tree (AST). However, previous methods mainly utilize an RNN-series decoder, which 1) is time-consuming and inefficient and 2) introduces very few structure priors. In this work, we propose an AST structure-aware Transformer decoder (ASTormer) to replace traditional RNN cells. The structural knowledge, such as node types and positions in the tree, is seamlessly incorporated into the decoder via both absolute and relative position embeddings. Besides, the proposed framework is compatible with different traversing orders even considering adaptive node selection. Extensive experiments on five text-to-SQL benchmarks demonstrate the effectiveness and efficiency of our structured decoder compared to competitive baselines."
                    },
                    {
                        "title": "On the Structural Generalization in Text-to-SQL",
                        "abstract": "Exploring the generalization of a text-to-SQL parser is essential for a system to automatically adapt the real-world databases. Previous works provided investigations focusing on lexical diversity, including the influence of the synonym and perturbations in both natural language questions and databases. However, research on the structure variety of database schema~(DS) is deficient. Specifically, confronted with the same input question, the target SQL is probably represented in different ways when the DS comes to a different structure. In this work, we provide in-deep discussions about the structural generalization of text-to-SQL tasks. We observe that current datasets are too templated to study structural generalization. To collect eligible test data, we propose a framework to generate novel text-to-SQL data via automatic and synchronous (DS, SQL) pair altering. In the experiments, significant performance reduction when evaluating well-trained text-to-SQL models on the synthetic samples demonstrates the limitation of current research regarding structural generalization. According to comprehensive analysis, we suggest the practical reason is the overfitting of (NL, SQL) patterns."
                    },
                    {
                        "title": "Sparsity-Accelerated Training for Large Language Models",
                        "abstract": "Large language models (LLMs) have demonstrated proficiency across various natural language processing (NLP) tasks but often require additional training, such as continual pre-training and supervised fine-tuning. However, the costs associated with this, primarily due to their large parameter count, remain high. This paper proposes leveraging \\emph{sparsity} in pre-trained LLMs to expedite this training process. By observing sparsity in activated neurons during forward iterations, we identify the potential for computational speed-ups by excluding inactive neurons. We address associated challenges by extending existing neuron importance evaluation metrics and introducing a ladder omission rate scheduler. Our experiments on Llama-2 demonstrate that Sparsity-Accelerated Training (SAT) achieves comparable or superior performance to standard training while significantly accelerating the process. Specifically, SAT achieves a $45\\%$ throughput improvement in continual pre-training and saves $38\\%$ training time in supervised fine-tuning in practice. It offers a simple, hardware-agnostic, and easily deployable framework for additional LLM training. Our code is available at https://github.com/OpenDFM/SAT."
                    }
                ]
            },
            "57b77e33-783b-4c1c-8147-b580fc908cc6": {
                "pk": "57b77e33-783b-4c1c-8147-b580fc908cc6",
                "name": "Danyang Zhang",
                "collaborators": [
                    "Yansong Tang",
                    "Jiwen Lu",
                    "Jie Zhou",
                    "Xinyi Huang",
                    "Yingyi Wu",
                    "Jiacheng Hu",
                    "Yujian Long",
                    "Lu Chen",
                    "Situo Zhang",
                    "Hongshen Xu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Reinforcement Learning",
                    "Action Recognition",
                    "Uncertainty Learning"
                ],
                "publications": [
                    {
                        "title": "Improving Academic Skills Assessment with NLP and Ensemble Learning",
                        "abstract": "This study addresses the critical challenges of assessing foundational academic skills by leveraging advancements in natural language processing (NLP). Traditional assessment methods often struggle to provide timely and comprehensive feedback on key cognitive and linguistic aspects, such as coherence, syntax, and analytical reasoning. Our approach integrates multiple state-of-the-art NLP models, including BERT, RoBERTa, BART, DeBERTa, and T5, within an ensemble learning framework. These models are combined through stacking techniques using LightGBM and Ridge regression to enhance predictive accuracy. The methodology involves detailed data preprocessing, feature extraction, and pseudo-label learning to optimize model performance. By incorporating sophisticated NLP techniques and ensemble learning, this study significantly improves the accuracy and efficiency of assessments, offering a robust solution that surpasses traditional methods and opens new avenues for educational technology research focused on enhancing core academic competencies."
                    },
                    {
                        "title": "Large Language Models Are Semi-Parametric Reinforcement Learning Agents",
                        "abstract": "Inspired by the insights in cognitive science with respect to human memory and reasoning mechanism, a novel evolvable LLM-based (Large Language Model) agent framework is proposed as REMEMBERER. By equipping the LLM with a long-term experience memory, REMEMBERER is capable of exploiting the experiences from the past episodes even for different task goals, which excels an LLM-based agent with fixed exemplars or equipped with a transient working memory. We further introduce Reinforcement Learning with Experience Memory (RLEM) to update the memory. Thus, the whole system can learn from the experiences of both success and failure, and evolve its capability without fine-tuning the parameters of the LLM. In this way, the proposed REMEMBERER constitutes a semi-parametric RL agent. Extensive experiments are conducted on two RL task sets to evaluate the proposed framework. The average results with different initialization and training sets exceed the prior SOTA by 4% and 2% for the success rate on two task sets and demonstrate the superiority and robustness of REMEMBERER."
                    },
                    {
                        "title": "Uncertainty-aware Score Distribution Learning for Action Quality Assessment",
                        "abstract": "Assessing action quality from videos has attracted growing attention in recent years. Most existing approaches usually tackle this problem based on regression algorithms, which ignore the intrinsic ambiguity in the score labels caused by multiple judges or their subjective appraisals. To address this issue, we propose an uncertainty-aware score distribution learning (USDL) approach for action quality assessment (AQA). Specifically, we regard an action as an instance associated with a score distribution, which describes the probability of different evaluated scores. Moreover, under the circumstance where fine-grained score labels are available (e.g., difficulty degree of an action or multiple scores from different judges), we further devise a multi-path uncertainty-aware score distributions learning (MUSDL) method to explore the disentangled components of a score. We conduct experiments on three AQA datasets containing various Olympic actions and surgical activities, where our approaches set new state-of-the-arts under the Spearman's Rank Correlation."
                    },
                    {
                        "title": "Learning from Temporal Spatial Cubism for Cross-Dataset Skeleton-based Action Recognition",
                        "abstract": "Rapid progress and superior performance have been achieved for skeleton-based action recognition recently. In this article, we investigate this problem under a cross-dataset setting, which is a new, pragmatic, and challenging task in real-world scenarios. Following the unsupervised domain adaptation (UDA) paradigm, the action labels are only available on a source dataset, but unavailable on a target dataset in the training stage. Different from the conventional adversarial learning-based approaches for UDA, we utilize a self-supervision scheme to reduce the domain shift between two skeleton-based action datasets. Our inspiration is drawn from Cubism, an art genre from the early 20th century, which breaks and reassembles the objects to convey a greater context. By segmenting and permuting temporal segments or human body parts, we design two self-supervised learning classification tasks to explore the temporal and spatial dependency of a skeleton-based action and improve the generalization ability of the model. We conduct experiments on six datasets for skeleton-based action recognition, including three large-scale datasets (NTU RGB+D, PKU-MMD, and Kinetics) where new cross-dataset settings and benchmarks are established. Extensive results demonstrate that our method outperforms state-of-the-art approaches. The source codes of our model and all the compared methods are available at https://github.com/shanice-l/st-cubism."
                    }
                ]
            },
            "29dea23d-1cea-4a11-aa8a-7e208c8561ab": {
                "pk": "29dea23d-1cea-4a11-aa8a-7e208c8561ab",
                "name": "Sida Wang",
                "collaborators": [
                    "Stefan Wager",
                    "Percy Liang",
                    "Weiwei Guo",
                    "Huiji Gao",
                    "Bo Long",
                    "Linyuan Gong",
                    "Mostafa Elhoushi",
                    "Alvin Cheung",
                    "William Fithian",
                    "Lili Yu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Query Auto Completion",
                    "Machine Learning",
                    "Regularization Techniques"
                ],
                "publications": [
                    {
                        "title": "Efficient Neural Query Auto Completion",
                        "abstract": "Query Auto Completion (QAC), as the starting point of information retrieval tasks, is critical to user experience. Generally it has two steps: generating completed query candidates according to query prefixes, and ranking them based on extracted features. Three major challenges are observed for a query auto completion system: (1) QAC has a strict online latency requirement. For each keystroke, results must be returned within tens of milliseconds, which poses a significant challenge in designing sophisticated language models for it. (2) For unseen queries, generated candidates are of poor quality as contextual information is not fully utilized. (3) Traditional QAC systems heavily rely on handcrafted features such as the query candidate frequency in search logs, lacking sufficient semantic understanding of the candidate.   In this paper, we propose an efficient neural QAC system with effective context modeling to overcome these challenges. On the candidate generation side, this system uses as much information as possible in unseen prefixes to generate relevant candidates, increasing the recall by a large margin. On the candidate ranking side, an unnormalized language model is proposed, which effectively captures deep semantics of queries. This approach presents better ranking performance over state-of-the-art neural ranking methods and reduces $\\sim$95\\% latency compared to neural language modeling methods. The empirical results on public datasets show that our model achieves a good balance between accuracy and efficiency. This system is served in LinkedIn job search with significant product impact observed."
                    },
                    {
                        "title": "Dropout Training as Adaptive Regularization",
                        "abstract": "Dropout and other feature noising schemes control overfitting by artificially corrupting the training data. For generalized linear models, dropout performs a form of adaptive regularization. Using this viewpoint, we show that the dropout regularizer is first-order equivalent to an L2 regularizer applied after scaling the features by an estimate of the inverse diagonal Fisher information matrix. We also establish a connection to AdaGrad, an online learning algorithm, and find that a close relative of AdaGrad operates by repeatedly solving linear dropout-regularized problems. By casting dropout as regularization, we develop a natural semi-supervised algorithm that uses unlabeled data to create a better adaptive regularizer. We apply this idea to document classification tasks, and show that it consistently boosts the performance of dropout training, improving on state-of-the-art results on the IMDB reviews dataset."
                    },
                    {
                        "title": "Evaluation of LLMs on Syntax-Aware Code Fill-in-the-Middle Tasks",
                        "abstract": "We introduce Syntax-Aware Fill-In-the-Middle (SAFIM), a new benchmark for evaluating Large Language Models (LLMs) on the code Fill-in-the-Middle (FIM) task. This benchmark focuses on syntax-aware completions of program structures such as code blocks and conditional expressions, and includes 17,720 examples from multiple programming languages, sourced from recent code submissions after April 2022 to minimize data contamination. SAFIM provides a robust framework with various prompt designs and novel syntax-aware post-processing techniques, facilitating accurate and fair comparisons across LLMs. Our comprehensive evaluation of 15 LLMs shows that FIM pretraining not only enhances FIM proficiency but also improves Left-to-Right (L2R) inference using LLMs. Our findings challenge conventional beliefs and suggest that pretraining methods and data quality have more impact than model size. SAFIM thus serves as a foundational platform for future research in effective pretraining strategies for code LLMs. The evaluation toolkit and dataset are available at https://github.com/gonglinyuan/safim, and the leaderboard is available at https://safimbenchmark.com."
                    },
                    {
                        "title": "Altitude Training: Strong Bounds for Single-Layer Dropout",
                        "abstract": "Dropout training, originally designed for deep neural networks, has been successful on high-dimensional single-layer natural language tasks. This paper proposes a theoretical explanation for this phenomenon: we show that, under a generative Poisson topic model with long documents, dropout training improves the exponent in the generalization bound for empirical risk minimization. Dropout achieves this gain much like a marathon runner who practices at altitude: once a classifier learns to perform reasonably well on training examples that have been artificially corrupted by dropout, it will do very well on the uncorrupted test set. We also show that, under similar conditions, dropout preserves the Bayes decision boundary and should therefore induce minimal bias in high dimensions."
                    },
                    {
                        "title": "Interactive Classification by Asking Informative Questions",
                        "abstract": "We study the potential for interaction in natural language classification. We add a limited form of interaction for intent classification, where users provide an initial query using natural language, and the system asks for additional information using binary or multi-choice questions. At each turn, our system decides between asking the most informative question or making the final classification prediction.The simplicity of the model allows for bootstrapping of the system without interaction data, instead relying on simple crowdsourcing tasks. We evaluate our approach on two domains, showing the benefit of interaction and the advantage of learning to balance between asking additional questions and making the final prediction."
                    }
                ]
            },
            "e37afecb-cc7d-4d10-9dfa-56cfa347db45": {
                "pk": "e37afecb-cc7d-4d10-9dfa-56cfa347db45",
                "name": "Ruoxi Sun",
                "collaborators": [
                    "Minhui Xue",
                    "Wei Wang",
                    "Hanjun Dai",
                    "Yifan Zhou",
                    "Zhendong Shi",
                    "Matthew Crawford",
                    "Hamish Spencer",
                    "Zewei Shi",
                    "Jieshan Chen",
                    "Jiamou Sun"
                ],
                "domain": [
                    "Machine Learning",
                    "Cybersecurity",
                    "Graph Neural Network",
                    "Privacy"
                ],
                "publications": [
                    {
                        "title": "Quality Assessment of Online Automated Privacy Policy Generators: An Empirical Study",
                        "abstract": "Online Automated Privacy Policy Generators (APPGs) are tools used by app developers to quickly create app privacy policies which are required by privacy regulations to be incorporated to each mobile app. The creation of these tools brings convenience to app developers; however, the quality of these tools puts developers and stakeholders at legal risk. In this paper, we conduct an empirical study to assess the quality of online APPGs. We analyze the completeness of privacy policies, determine what categories and items should be covered in a complete privacy policy, and conduct APPG assessment with boilerplate apps. The results of assessment show that due to the lack of static or dynamic analysis of app's behavior, developers may encounter two types of issues caused by APPGs. First, the generated policies could be incomplete because they do not cover all the essential items required by a privacy policy. Second, some generated privacy policies contain unnecessary personal information collection or arbitrary commitments inconsistent with user input. Ultimately, the defects of APPGs may potentially lead to serious legal issues. We hope that the results and insights developed in this paper can motivate the healthy and ethical development of APPGs towards generating a more complete, accurate, and robust privacy policy."
                    },
                    {
                        "title": "Visualization and Attack Prevention for a Sensor-Based Agricultural Monitoring System",
                        "abstract": "This project proposes a sensor-based visual agricultural monitoring system. Distinguished from traditional agricultural monitoring systems, this system further analyzes basic agricultural data and prevents and monitors common wireless network attacks such as Selective Forwarding, Black Hole Attacks, Sinkhole Attacks, Flooding Attacks and Misdirection Attacks. Experimental verification and evaluation of the attack prevention and monitoring are also conducted."
                    },
                    {
                        "title": "Statically Detecting Adversarial Malware through Randomised Chaining",
                        "abstract": "With the rapid growth of malware attacks, more antivirus developers consider deploying machine learning technologies into their productions. Researchers and developers published various machine learning-based detectors with high precision on malware detection in recent years. Although numerous machine learning-based malware detectors are available, they face various machine learning-targeted attacks, including evasion and adversarial attacks. This project explores how and why adversarial examples evade malware detectors, then proposes a randomised chaining method to defend against adversarial malware statically. This research is crucial for working towards combating the pertinent malware cybercrime."
                    },
                    {
                        "title": "Dissecting Malware in the Wild",
                        "abstract": "With the increasingly rapid development of new malicious computer software by bad faith actors, both commercial and research-oriented antivirus detectors have come to make greater use of machine learning tactics to identify such malware as harmful before end users are exposed to their effects. This, in turn, has spurred the development of tools that allow for known malware to be manipulated such that they can evade being classified as dangerous by these machine learning-based detectors, while retaining their malicious functionality. These manipulations function by applying a set of changes that can be made to Windows programs that result in a different file structure and signature without altering the software's capabilities. Various proposals have been made for the most effective way of applying these alterations to input malware to deceive static malware detectors; the purpose of this research is to examine these proposals and test their implementations to determine which tactics tend to generate the most successful attacks."
                    },
                    {
                        "title": "The Invisible Game on the Internet: A Case Study of Decoding Deceptive Patterns",
                        "abstract": "Deceptive patterns are design practices embedded in digital platforms to manipulate users, representing a widespread and long-standing issue in the web and mobile software development industry. Legislative actions highlight the urgency of globally regulating deceptive patterns. However, despite advancements in detection tools, a significant gap exists in assessing deceptive pattern risks. In this study, we introduce a comprehensive approach involving the interactions between the Adversary, Watchdog (e.g., detection tools), and Challengers (e.g., users) to formalize and decode deceptive pattern threats. Based on this, we propose a quantitative risk assessment system. Representative cases are analyzed to showcase the practicability of the proposed risk scoring system, emphasizing the importance of involving human factors in deceptive pattern risk assessment."
                    },
                    {
                        "title": "Energy-based View of Retrosynthesis",
                        "abstract": "Retrosynthesis -- the process of identifying a set of reactants to synthesize a target molecule -- is of vital importance to material design and drug discovery. Existing machine learning approaches based on language models and graph neural networks have achieved encouraging results. In this paper, we propose a framework that unifies sequence- and graph-based methods as energy-based models (EBMs) with different energy functions. This unified perspective provides critical insights about EBM variants through a comprehensive assessment of performance. Additionally, we present a novel dual variant within the framework that performs consistent training over Bayesian forward- and backward-prediction by constraining the agreement between the two directions. This model improves state-of-the-art performance by 9.6% for template-free approaches where the reaction type is unknown."
                    },
                    {
                        "title": "Does GNN Pretraining Help Molecular Representation?",
                        "abstract": "Extracting informative representations of molecules using Graph neural networks (GNNs) is crucial in AI-driven drug discovery. Recently, the graph research community has been trying to replicate the success of self-supervised pretraining in natural language processing, with several successes claimed. However, we find the benefit brought by self-supervised pretraining on small molecular data can be negligible in many cases. We conduct thorough ablation studies on the key components of GNN pretraining, including pretraining objectives, data splitting methods, input features, pretraining dataset scales, and GNN architectures, to see how they affect the accuracy of the downstream tasks. Our first important finding is, self-supervised graph pretraining do not always have statistically significant advantages over non-pretraining methods in many settings. Secondly, although noticeable improvement can be observed with additional supervised pretraining, the improvement may diminish with richer features or more balanced data splits. Thirdly, hyper-parameters could have larger impacts on accuracy of downstream tasks than the choice of pretraining tasks, especially when the scales of downstream tasks are small. Finally, we provide our conjectures where the complexity of some pretraining methods on small molecules might be insufficient, followed by empirical evidences on different pretraining datasets."
                    },
                    {
                        "title": "M^4I: Multi-modal Models Membership Inference",
                        "abstract": "With the development of machine learning techniques, the attention of research has been moved from single-modal learning to multi-modal learning, as real-world data exist in the form of different modalities. However, multi-modal models often carry more information than single-modal models and they are usually applied in sensitive scenarios, such as medical report generation or disease identification. Compared with the existing membership inference against machine learning classifiers, we focus on the problem that the input and output of the multi-modal models are in different modalities, such as image captioning. This work studies the privacy leakage of multi-modal models through the lens of membership inference attack, a process of determining whether a data record involves in the model training process or not. To achieve this, we propose Multi-modal Models Membership Inference (M^4I) with two attack methods to infer the membership status, named metric-based (MB) M^4I and feature-based (FB) M^4I, respectively. More specifically, MB M^4I adopts similarity metrics while attacking to infer target data membership. FB M^4I uses a pre-trained shadow multi-modal feature extractor to achieve the purpose of data inference attack by comparing the similarities from extracted input and output features. Extensive experimental results show that both attack methods can achieve strong performances. Respectively, 72.5% and 94.83% of attack success rates on average can be obtained under unrestricted scenarios. Moreover, we evaluate multiple defense mechanisms against our attacks. The source code of M^4I attacks is publicly available at https://github.com/MultimodalMI/Multimodal-membership-inference.git."
                    },
                    {
                        "title": "Reverse engineering learned optimizers reveals known and novel mechanisms",
                        "abstract": "Learned optimizers are algorithms that can themselves be trained to solve optimization problems. In contrast to baseline optimizers (such as momentum or Adam) that use simple update rules derived from theoretical principles, learned optimizers use flexible, high-dimensional, nonlinear parameterizations. Although this can lead to better performance in certain settings, their inner workings remain a mystery. How is a learned optimizer able to outperform a well tuned baseline? Has it learned a sophisticated combination of existing optimization techniques, or is it implementing completely new behavior? In this work, we address these questions by careful analysis and visualization of learned optimizers. We study learned optimizers trained from scratch on three disparate tasks, and discover that they have learned interpretable mechanisms, including: momentum, gradient clipping, learning rate schedules, and a new form of learning rate adaptation. Moreover, we show how the dynamics of learned optimizers enables these behaviors. Our results help elucidate the previously murky understanding of how learned optimizers work, and establish tools for interpreting future learned optimizers."
                    }
                ]
            },
            "6906d89e-d71f-4c46-b7e6-5268707fb84b": {
                "pk": "6906d89e-d71f-4c46-b7e6-5268707fb84b",
                "name": "Pengcheng Yin",
                "collaborators": [
                    "Graham Neubig",
                    "Chunting Zhou",
                    "Junxian He",
                    "John Wieting",
                    "Avirup Sil",
                    "Zhengdong Lu",
                    "Hang Li",
                    "Ben Kao",
                    "Xinyi Wang",
                    "Hieu Pham"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Semantic Parsing",
                    "Code Generation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "A Syntactic Neural Model for General-Purpose Code Generation",
                        "abstract": "We consider the problem of parsing natural language descriptions into source code written in a general-purpose programming language like Python. Existing data-driven methods treat this problem as a language generation task without considering the underlying syntax of the target programming language. Informed by previous work in semantic parsing, in this paper we propose a novel neural architecture powered by a grammar model to explicitly capture the target syntax as prior knowledge. Experiments find this an effective way to scale up to generation of complex programs from natural language descriptions, achieving state-of-the-art results that well outperform previous code generation and semantic parsing approaches."
                    },
                    {
                        "title": "TRANX: A Transition-based Neural Abstract Syntax Parser for Semantic Parsing and Code Generation",
                        "abstract": "We present TRANX, a transition-based neural semantic parser that maps natural language (NL) utterances into formal meaning representations (MRs). TRANX uses a transition system based on the abstract syntax description language for the target MR, which gives it two major advantages: (1) it is highly accurate, using information from the syntax of the target MR to constrain the output space and model the information flow, and (2) it is highly generalizable, and can easily be applied to new types of MR by just writing a new abstract syntax description corresponding to the allowable structures in the MR. Experiments on four different semantic parsing and code generation tasks show that our system is generalizable, extensible, and effective, registering strong results compared to existing neural semantic parsers."
                    },
                    {
                        "title": "StructVAE: Tree-structured Latent Variable Models for Semi-supervised Semantic Parsing",
                        "abstract": "Semantic parsing is the task of transducing natural language (NL) utterances into formal meaning representations (MRs), commonly represented as tree structures. Annotating NL utterances with their corresponding MRs is expensive and time-consuming, and thus the limited availability of labeled data often becomes the bottleneck of data-driven, supervised models. We introduce StructVAE, a variational auto-encoding model for semisupervised semantic parsing, which learns both from limited amounts of parallel data, and readily-available unlabeled NL utterances. StructVAE models latent MRs not observed in the unlabeled data as tree-structured latent variables. Experiments on semantic parsing on the ATIS domain and Python code generation show that with extra unlabeled data, StructVAE outperforms strong supervised models."
                    },
                    {
                        "title": "On The Ingredients of an Effective Zero-shot Semantic Parser",
                        "abstract": "Semantic parsers map natural language utterances into meaning representations (e.g., programs). Such models are typically bottlenecked by the paucity of training data due to the required laborious annotation efforts. Recent studies have performed zero-shot learning by synthesizing training examples of canonical utterances and programs from a grammar, and further paraphrasing these utterances to improve linguistic diversity. However, such synthetic examples cannot fully capture patterns in real data. In this paper we analyze zero-shot parsers through the lenses of the language and logical gaps (Herzig and Berant, 2019), which quantify the discrepancy of language and programmatic patterns between the canonical examples and real-world user-issued ones. We propose bridging these gaps using improved grammars, stronger paraphrasers, and efficient learning methods using canonical examples that most likely reflect real user intents. Our model achieves strong performance on two semantic parsing benchmarks (Scholar, Geo) with zero labeled data."
                    },
                    {
                        "title": "Neural Enquirer: Learning to Query Tables with Natural Language",
                        "abstract": "We proposed Neural Enquirer as a neural network architecture to execute a natural language (NL) query on a knowledge-base (KB) for answers. Basically, Neural Enquirer finds the distributed representation of a query and then executes it on knowledge-base tables to obtain the answer as one of the values in the tables. Unlike similar efforts in end-to-end training of semantic parsers, Neural Enquirer is fully \"neuralized\": it not only gives distributional representation of the query and the knowledge-base, but also realizes the execution of compositional queries as a series of differentiable operations, with intermediate results (consisting of annotations of the tables at different levels) saved on multiple layers of memory. Neural Enquirer can be trained with gradient descent, with which not only the parameters of the controlling components and semantic parsing component, but also the embeddings of the tables and query words can be learned from scratch. The training can be done in an end-to-end fashion, but it can take stronger guidance, e.g., the step-by-step supervision for complicated queries, and benefit from it. Neural Enquirer is one step towards building neural network systems which seek to understand language by executing it on real-world. Our experiments show that Neural Enquirer can learn to execute fairly complicated NL queries on tables with rich structures."
                    },
                    {
                        "title": "A Tree-based Decoder for Neural Machine Translation",
                        "abstract": "Recent advances in Neural Machine Translation (NMT) show that adding syntactic information to NMT systems can improve the quality of their translations. Most existing work utilizes some specific types of linguistically-inspired tree structures, like constituency and dependency parse trees. This is often done via a standard RNN decoder that operates on a linearized target tree structure. However, it is an open question of what specific linguistic formalism, if any, is the best structural representation for NMT. In this paper, we (1) propose an NMT model that can naturally generate the topology of an arbitrary tree structure on the target side, and (2) experiment with various target tree structures. Our experiments show the surprising result that our model delivers the best improvements with balanced binary trees constructed without any linguistic knowledge; this model outperforms standard seq2seq models by up to 2.1 BLEU points, and other methods for incorporating target-side syntax by up to 0.7 BLEU."
                    },
                    {
                        "title": "Improving Open Information Extraction via Iterative Rank-Aware Learning",
                        "abstract": "Open information extraction (IE) is the task of extracting open-domain assertions from natural language sentences. A key step in open IE is confidence modeling, ranking the extractions based on their estimated quality to adjust precision and recall of extracted assertions. We found that the extraction likelihood, a confidence measure used by current supervised open IE systems, is not well calibrated when comparing the quality of assertions extracted from different sentences. We propose an additional binary classification loss to calibrate the likelihood to make it more globally comparable, and an iterative learning process, where extractions generated by the open IE model are incrementally included as training samples to help the model learn from trial and error. Experiments on OIE2016 demonstrate the effectiveness of our method. Code and data are available at https://github.com/jzbjyb/oie_rank."
                    },
                    {
                        "title": "Procedures as Programs: Hierarchical Control of Situated Agents through Natural Language",
                        "abstract": "When humans conceive how to perform a particular task, they do so hierarchically: splitting higher-level tasks into smaller sub-tasks. However, in the literature on natural language (NL) command of situated agents, most works have treated the procedures to be executed as flat sequences of simple actions, or any hierarchies of procedures have been shallow at best. In this paper, we propose a formalism of procedures as programs, a powerful yet intuitive method of representing hierarchical procedural knowledge for agent command and control. We further propose a modeling paradigm of hierarchical modular networks, which consist of a planner and reactors that convert NL intents to predictions of executable programs and probe the environment for information necessary to complete the program execution. We instantiate this framework on the IQA and ALFRED datasets for NL instruction following. Our model outperforms reactive baselines by a large margin on both datasets. We also demonstrate that our framework is more data-efficient, and that it allows for fast iterative development."
                    },
                    {
                        "title": "Unsupervised Evaluation of Code LLMs with Round-Trip Correctness",
                        "abstract": "To evaluate code large language models (LLMs), research has relied on a few small manually curated benchmarks, such as HumanEval and MBPP, which represent a narrow part of the real-world software domains. In this work, we introduce round-trip correctness (RTC) as an alternative evaluation method. RTC allows Code LLM evaluation on a broader spectrum of real-world software domains without the need for costly human curation. RTC rests on the idea that we can ask a model to make a prediction (e.g., describe some code using natural language), feed that prediction back (e.g., synthesize code from the predicted description), and check if this round-trip leads to code that is semantically equivalent to the original input. We show how to employ RTC to evaluate code synthesis and editing. We find that RTC strongly correlates with model performance on existing narrow-domain code synthesis benchmarks while allowing us to expand to a much broader set of domains and tasks which was not previously possible without costly human annotations."
                    },
                    {
                        "title": "Merging Weak and Active Supervision for Semantic Parsing",
                        "abstract": "A semantic parser maps natural language commands (NLs) from the users to executable meaning representations (MRs), which are later executed in certain environment to obtain user-desired results. The fully-supervised training of such parser requires NL/MR pairs, annotated by domain experts, which makes them expensive to collect. However, weakly-supervised semantic parsers are learnt only from pairs of NL and expected execution results, leaving the MRs latent. While weak supervision is cheaper to acquire, learning from this input poses difficulties. It demands that parsers search a large space with a very weak learning signal and it is hard to avoid spurious MRs that achieve the correct answer in the wrong way. These factors lead to a performance gap between parsers trained in weakly- and fully-supervised setting. To bridge this gap, we examine the intersection between weak supervision and active learning, which allows the learner to actively select examples and query for manual annotations as extra supervision to improve the model trained under weak supervision. We study different active learning heuristics for selecting examples to query, and various forms of extra supervision for such queries. We evaluate the effectiveness of our method on two different datasets. Experiments on the WikiSQL show that by annotating only 1.8% of examples, we improve over a state-of-the-art weakly-supervised baseline by 6.4%, achieving an accuracy of 79.0%, which is only 1.3% away from the model trained with full supervision. Experiments on WikiTableQuestions with human annotators show that our method can improve the performance with only 100 active queries, especially for weakly-supervised parsers learnt from a cold start."
                    },
                    {
                        "title": "TaBERT: Pretraining for Joint Understanding of Textual and Tabular Data",
                        "abstract": "Recent years have witnessed the burgeoning of pretrained language models (LMs) for text-based natural language (NL) understanding tasks. Such models are typically trained on free-form NL text, hence may not be suitable for tasks like semantic parsing over structured data, which require reasoning over both free-form NL questions and structured tabular data (e.g., database tables). In this paper we present TaBERT, a pretrained LM that jointly learns representations for NL sentences and (semi-)structured tables. TaBERT is trained on a large corpus of 26 million tables and their English contexts. In experiments, neural semantic parsers using TaBERT as feature representation layers achieve new best results on the challenging weakly-supervised semantic parsing benchmark WikiTableQuestions, while performing competitively on the text-to-SQL dataset Spider. Implementation of the model will be available at http://fburl.com/TaBERT ."
                    },
                    {
                        "title": "Learning to Mine Aligned Code and Natural Language Pairs from Stack Overflow",
                        "abstract": "For tasks like code synthesis from natural language, code retrieval, and code summarization, data-driven models have shown great promise. However, creating these models require parallel data between natural language (NL) and code with fine-grained alignments. Stack Overflow (SO) is a promising source to create such a data set: the questions are diverse and most of them have corresponding answers with high-quality code snippets. However, existing heuristic methods (e.g., pairing the title of a post with the code in the accepted answer) are limited both in their coverage and the correctness of the NL-code pairs obtained. In this paper, we propose a novel method to mine high-quality aligned data from SO using two sets of features: hand-crafted features considering the structure of the extracted snippets, and correspondence features obtained by training a probabilistic model to capture the correlation between NL and code using neural networks. These features are fed into a classifier that determines the quality of mined NL-code pairs. Experiments using Python and Java as test beds show that the proposed method greatly expands coverage and accuracy over existing mining methods, even when using only a small number of labeled examples. Further, we find that reasonable results are achieved even when training the classifier on one language and testing on another, showing promise for scaling NL-code mining to a wide variety of programming languages beyond those for which we are able to annotate data."
                    }
                ]
            },
            "7b449035-c4fd-4b9e-9241-66398e052cba": {
                "pk": "7b449035-c4fd-4b9e-9241-66398e052cba",
                "name": "Caiming Xiong",
                "collaborators": [
                    "Richard Socher",
                    "Victor Zhong",
                    "Yingbo Zhou",
                    "James Bradbury",
                    "Huan Wang",
                    "Stephen Merity",
                    "Tianmin Shu",
                    "Huishuai Zhang",
                    "Shayne Longpre",
                    "Sabeek Pradhan"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Neural Networks",
                    "Natural Language Processing",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Hierarchical and Interpretable Skill Acquisition in Multi-task Reinforcement Learning",
                        "abstract": "Learning policies for complex tasks that require multiple different skills is a major challenge in reinforcement learning (RL). It is also a requirement for its deployment in real-world scenarios. This paper proposes a novel framework for efficient multi-task reinforcement learning. Our framework trains agents to employ hierarchical policies that decide when to use a previously learned policy and when to learn a new skill. This enables agents to continually acquire new skills during different stages of training. Each learned task corresponds to a human language description. Because agents can only access previously learned skills through these descriptions, the agent can always provide a human-interpretable description of its choices. In order to help the agent learn the complex temporal dependencies necessary for the hierarchical policy, we provide it with a stochastic temporal grammar that modulates when to rely on previously learned skills and when to execute new skills. We validate our approach on Minecraft games designed to explicitly test the ability to reuse previously learned skills while simultaneously learning new skills."
                    },
                    {
                        "title": "Block-diagonal Hessian-free Optimization for Training Neural Networks",
                        "abstract": "Second-order methods for neural network optimization have several advantages over methods based on first-order gradient descent, including better scaling to large mini-batch sizes and fewer updates needed for convergence. But they are rarely applied to deep learning in practice because of high computational cost and the need for model-dependent algorithmic variations. We introduce a variant of the Hessian-free method that leverages a block-diagonal approximation of the generalized Gauss-Newton matrix. Our method computes the curvature approximation matrix only for pairs of parameters from the same layer or block of the neural network and performs conjugate gradient updates independently for each block. Experiments on deep autoencoders, deep convolutional networks, and multilayer LSTMs demonstrate better convergence and generalization compared to the original Hessian-free approach and the Adam method."
                    },
                    {
                        "title": "A Way out of the Odyssey: Analyzing and Combining Recent Insights for LSTMs",
                        "abstract": "LSTMs have become a basic building block for many deep NLP models. In recent years, many improvements and variations have been proposed for deep sequence models in general, and LSTMs in particular. We propose and analyze a series of augmentations and modifications to LSTM networks resulting in improved performance for text classification datasets. We observe compounding improvements on traditional LSTMs using Monte Carlo test-time model averaging, average pooling, and residual connections, along with four other suggested modifications. Our analysis provides a simple, reliable, and high quality baseline model."
                    },
                    {
                        "title": "Enabling High Data Throughput Reinforcement Learning on GPUs: A Domain Agnostic Framework for Data-Driven Scientific Research",
                        "abstract": "We introduce WarpSci, a domain agnostic framework designed to overcome crucial system bottlenecks encountered in the application of reinforcement learning to intricate environments with vast datasets featuring high-dimensional observation or action spaces. Notably, our framework eliminates the need for data transfer between the CPU and GPU, enabling the concurrent execution of thousands of simulations on a single or multiple GPUs. This high data throughput architecture proves particularly advantageous for data-driven scientific research, where intricate environment models are commonly essential."
                    },
                    {
                        "title": "Dynamic Memory Networks for Visual and Textual Question Answering",
                        "abstract": "Neural network architectures with memory and attention mechanisms exhibit certain reasoning capabilities required for question answering. One such architecture, the dynamic memory network (DMN), obtained high accuracy on a variety of language tasks. However, it was not shown whether the architecture achieves strong results for question answering when supporting facts are not marked during training or whether it could be applied to other modalities such as images. Based on an analysis of the DMN, we propose several improvements to its memory and input modules. Together with these changes we introduce a novel input module for images in order to be able to answer visual questions. Our new DMN+ model improves the state of the art on both the Visual Question Answering dataset and the \\babi-10k text question-answering dataset without supporting fact supervision."
                    },
                    {
                        "title": "Action Understanding with Multiple Classes of Actors",
                        "abstract": "Despite the rapid progress, existing works on action understanding focus strictly on one type of action agent, which we call actor---a human adult, ignoring the diversity of actions performed by other actors. To overcome this narrow viewpoint, our paper marks the first effort in the computer vision community to jointly consider algorithmic understanding of various types of actors undergoing various actions. To begin with, we collect a large annotated Actor-Action Dataset (A2D) that consists of 3782 short videos and 31 temporally untrimmed long videos. We formulate the general actor-action understanding problem and instantiate it at various granularities: video-level single- and multiple-label actor-action recognition, and pixel-level actor-action segmentation. We propose and examine a comprehensive set of graphical models that consider the various types of interplay among actors and actions. Our findings have led us to conclusive evidence that the joint modeling of actor and action improves performance over modeling each of them independently, and further improvement can be obtained by considering the multi-scale natural in video understanding. Hence, our paper concludes the argument of the value of explicit consideration of various actors in comprehensive action understanding and provides a dataset and a benchmark for later works exploring this new problem."
                    },
                    {
                        "title": "DCN+: Mixed Objective and Deep Residual Coattention for Question Answering",
                        "abstract": "Traditional models for question answering optimize using cross entropy loss, which encourages exact answers at the cost of penalizing nearby or overlapping answers that are sometimes equally accurate. We propose a mixed objective that combines cross entropy loss with self-critical policy learning. The objective uses rewards derived from word overlap to solve the misalignment between evaluation metric and optimization objective. In addition to the mixed objective, we improve dynamic coattention networks (DCN) with a deep residual coattention encoder that is inspired by recent work in deep self-attention and residual networks. Our proposals improve model performance across question types and input lengths, especially for long questions that requires the ability to capture long-term dependencies. On the Stanford Question Answering Dataset, our model achieves state-of-the-art results with 75.1% exact match accuracy and 83.1% F1, while the ensemble obtains 78.9% exact match accuracy and 86.0% F1."
                    },
                    {
                        "title": "Improving End-to-End Speech Recognition with Policy Learning",
                        "abstract": "Connectionist temporal classification (CTC) is widely used for maximum likelihood learning in end-to-end speech recognition models. However, there is usually a disparity between the negative maximum likelihood and the performance metric used in speech recognition, e.g., word error rate (WER). This results in a mismatch between the objective function and metric during training. We show that the above problem can be mitigated by jointly training with maximum likelihood and policy gradient. In particular, with policy learning we are able to directly optimize on the (otherwise non-differentiable) performance metric. We show that joint training improves relative performance by 4% to 13% for our end-to-end model as compared to the same model learned through maximum likelihood. The model achieves 5.53% WER on Wall Street Journal dataset, and 5.42% and 14.70% on Librispeech test-clean and test-other set, respectively."
                    },
                    {
                        "title": "Improved Regularization Techniques for End-to-End Speech Recognition",
                        "abstract": "Regularization is important for end-to-end speech models, since the models are highly flexible and easy to overfit. Data augmentation and dropout has been important for improving end-to-end models in other domains. However, they are relatively under explored for end-to-end speech models. Therefore, we investigate the effectiveness of both methods for end-to-end trainable, deep speech recognition models. We augment audio data through random perturbations of tempo, pitch, volume, temporal alignment, and adding random noise.We further investigate the effect of dropout when applied to the inputs of all layers of the network. We show that the combination of data augmentation and dropout give a relative performance improvement on both Wall Street Journal (WSJ) and LibriSpeech dataset of over 20%. Our model performance is also competitive with other end-to-end speech models on both datasets."
                    },
                    {
                        "title": "Interpretable Counting for Visual Question Answering",
                        "abstract": "Questions that require counting a variety of objects in images remain a major challenge in visual question answering (VQA). The most common approaches to VQA involve either classifying answers based on fixed length representations of both the image and question or summing fractional counts estimated from each section of the image. In contrast, we treat counting as a sequential decision process and force our model to make discrete choices of what to count. Specifically, the model sequentially selects from detected objects and learns interactions between objects that influence subsequent selections. A distinction of our approach is its intuitive and interpretable output, as discrete counts are automatically grounded in the image. Furthermore, our method outperforms the state of the art architecture for VQA on multiple metrics that evaluate counting."
                    },
                    {
                        "title": "Efficient and Robust Question Answering from Minimal Context over Documents",
                        "abstract": "Neural models for question answering (QA) over documents have achieved significant performance improvements. Although effective, these models do not scale to large corpora due to their complex modeling of interactions between the document and the question. Moreover, recent work has shown that such models are sensitive to adversarial inputs. In this paper, we study the minimal context required to answer the question, and find that most questions in existing datasets can be answered with a small set of sentences. Inspired by this observation, we propose a simple sentence selector to select the minimal set of sentences to feed into the QA model. Our overall system achieves significant reductions in training (up to 15 times) and inference times (up to 13 times), with accuracy comparable to or better than the state-of-the-art on SQuAD, NewsQA, TriviaQA and SQuAD-Open. Furthermore, our experimental results and analyses show that our approach is more robust to adversarial inputs."
                    },
                    {
                        "title": "Global-Locally Self-Attentive Dialogue State Tracker",
                        "abstract": "Dialogue state tracking, which estimates user goals and requests given the dialogue context, is an essential part of task-oriented dialogue systems. In this paper, we propose the Global-Locally Self-Attentive Dialogue State Tracker (GLAD), which learns representations of the user utterance and previous system actions with global-local modules. Our model uses global modules to share parameters between estimators for different types (called slots) of dialogue states, and uses local modules to learn slot-specific features. We show that this significantly improves tracking of rare states and achieves state-of-the-art performance on the WoZ and DSTC2 state tracking tasks. GLAD obtains 88.1% joint goal accuracy and 97.1% request accuracy on WoZ, outperforming prior work by 3.7% and 5.5%. On DSTC2, our model obtains 74.5% joint goal accuracy and 97.5% request accuracy, outperforming prior work by 1.1% and 1.0%."
                    },
                    {
                        "title": "Improving Abstraction in Text Summarization",
                        "abstract": "Abstractive text summarization aims to shorten long text documents into a human readable form that contains the most important facts from the original document. However, the level of actual abstraction as measured by novel phrases that do not appear in the source document remains low in existing approaches. We propose two techniques to improve the level of abstraction of generated summaries. First, we decompose the decoder into a contextual network that retrieves relevant parts of the source document, and a pretrained language model that incorporates prior knowledge about language generation. Second, we propose a novelty metric that is optimized directly through policy learning to encourage the generation of novel phrases. Our model achieves results comparable to state-of-the-art models, as determined by ROUGE scores and human evaluations, while achieving a significantly higher level of abstraction as measured by n-gram overlap with the source document."
                    },
                    {
                        "title": "Global-to-local Memory Pointer Networks for Task-Oriented Dialogue",
                        "abstract": "End-to-end task-oriented dialogue is challenging since knowledge bases are usually large, dynamic and hard to incorporate into a learning framework. We propose the global-to-local memory pointer (GLMP) networks to address this issue. In our model, a global memory encoder and a local memory decoder are proposed to share external knowledge. The encoder encodes dialogue history, modifies global contextual representation, and generates a global memory pointer. The decoder first generates a sketch response with unfilled slots. Next, it passes the global memory pointer to filter the external knowledge for relevant information, then instantiates the slots via the local memory pointers. We empirically show that our model can improve copy accuracy and mitigate the common out-of-vocabulary problem. As a result, GLMP is able to improve over the previous state-of-the-art models in both simulated bAbI Dialogue dataset and human-human Stanford Multi-domain Dialogue dataset on automatic and human evaluation."
                    },
                    {
                        "title": "Quasi-Recurrent Neural Networks",
                        "abstract": "Recurrent neural networks are a powerful tool for modeling sequential data, but the dependence of each timestep's computation on the previous timestep's output limits parallelism and makes RNNs unwieldy for very long sequences. We introduce quasi-recurrent neural networks (QRNNs), an approach to neural sequence modeling that alternates convolutional layers, which apply in parallel across timesteps, and a minimalist recurrent pooling function that applies in parallel across channels. Despite lacking trainable recurrent layers, stacked QRNNs have better predictive accuracy than stacked LSTMs of the same hidden size. Due to their increased parallelism, they are up to 16 times faster at train and test time. Experiments on language modeling, sentiment classification, and character-level neural machine translation demonstrate these advantages and underline the viability of QRNNs as a basic building block for a variety of sequence tasks."
                    },
                    {
                        "title": "A Joint Many-Task Model: Growing a Neural Network for Multiple NLP Tasks",
                        "abstract": "Transfer and multi-task learning have traditionally focused on either a single source-target pair or very few, similar tasks. Ideally, the linguistic levels of morphology, syntax and semantics would benefit each other by being trained in a single model. We introduce a joint many-task model together with a strategy for successively growing its depth to solve increasingly complex tasks. Higher layers include shortcut connections to lower-level task predictions to reflect linguistic hierarchies. We use a simple regularization term to allow for optimizing all model weights to improve one task's loss without exhibiting catastrophic interference of the other tasks. Our single end-to-end model obtains state-of-the-art or competitive results on five different tasks from tagging, parsing, relatedness, and entailment tasks."
                    },
                    {
                        "title": "Dynamic Coattention Networks For Question Answering",
                        "abstract": "Several deep learning models have been proposed for question answering. However, due to their single-pass nature, they have no way to recover from local maxima corresponding to incorrect answers. To address this problem, we introduce the Dynamic Coattention Network (DCN) for question answering. The DCN first fuses co-dependent representations of the question and the document in order to focus on relevant parts of both. Then a dynamic pointing decoder iterates over potential answer spans. This iterative procedure enables the model to recover from initial local maxima corresponding to incorrect answers. On the Stanford question answering dataset, a single DCN model improves the previous state of the art from 71.0% F1 to 75.9%, while a DCN ensemble obtains 80.4% F1."
                    },
                    {
                        "title": "Knowing When to Look: Adaptive Attention via A Visual Sentinel for Image Captioning",
                        "abstract": "Attention-based neural encoder-decoder frameworks have been widely adopted for image captioning. Most methods force visual attention to be active for every generated word. However, the decoder likely requires little to no visual information from the image to predict non-visual words such as \"the\" and \"of\". Other words that may seem visual can often be predicted reliably just from the language model e.g., \"sign\" after \"behind a red stop\" or \"phone\" following \"talking on a cell\". In this paper, we propose a novel adaptive attention model with a visual sentinel. At each time step, our model decides whether to attend to the image (and if so, to which regions) or to the visual sentinel. The model decides whether to attend to the image and where, in order to extract meaningful information for sequential word generation. We test our method on the COCO image captioning 2015 challenge dataset and Flickr30K. Our approach sets the new state-of-the-art by a significant margin."
                    },
                    {
                        "title": "A Multi-Discriminator CycleGAN for Unsupervised Non-Parallel Speech Domain Adaptation",
                        "abstract": "Domain adaptation plays an important role for speech recognition models, in particular, for domains that have low resources. We propose a novel generative model based on cyclic-consistent generative adversarial network (CycleGAN) for unsupervised non-parallel speech domain adaptation. The proposed model employs multiple independent discriminators on the power spectrogram, each in charge of different frequency bands. As a result we have 1) better discriminators that focus on fine-grained details of the frequency features, and 2) a generator that is capable of generating more realistic domain-adapted spectrogram. We demonstrate the effectiveness of our method on speech recognition with gender adaptation, where the model only has access to supervised data from one gender during training, but is evaluated on the other at test time. Our model is able to achieve an average of $7.41\\%$ on phoneme error rate, and $11.10\\%$ word error rate relative performance improvement as compared to the baseline, on TIMIT and WSJ dataset, respectively. Qualitatively, our model also generates more natural sounding speech, when conditioned on data from the other domain."
                    },
                    {
                        "title": "On the Generalization Gap in Reparameterizable Reinforcement Learning",
                        "abstract": "Understanding generalization in reinforcement learning (RL) is a significant challenge, as many common assumptions of traditional supervised learning theory do not apply. We focus on the special class of reparameterizable RL problems, where the trajectory distribution can be decomposed using the reparametrization trick. For this problem class, estimating the expected return is efficient and the trajectory can be computed deterministically given peripheral random variables, which enables us to study reparametrizable RL using supervised learning and transfer learning theory. Through these relationships, we derive guarantees on the gap between the expected and empirical return for both intrinsic and external errors, based on Rademacher complexity as well as the PAC-Bayes bound. Our bound suggests the generalization capability of reparameterizable RL is related to multiple factors including \"smoothness\" of the environment transition, reward and agent policy function class. We also empirically verify the relationship between the generalization gap and these factors through simulations."
                    }
                ]
            },
            "7a34cf23-8d8f-4c3e-b197-077d7ba83012": {
                "pk": "7a34cf23-8d8f-4c3e-b197-077d7ba83012",
                "name": "Ansong Ni",
                "collaborators": [
                    "Dragomir Radev",
                    "Pengcheng Yin",
                    "Yusen Zhang",
                    "Tao Yu",
                    "Budhaditya Deb",
                    "Chenguang Zhu",
                    "Rui Zhang",
                    "Arman Cohan",
                    "Ahmed H. Awadallah",
                    "Chen Henry Wu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Code Generation",
                    "Machine Learning",
                    "Summarization"
                ],
                "publications": [
                    {
                        "title": "Quantifying Contamination in Evaluating Code Generation Capabilities of Language Models",
                        "abstract": "While large language models have achieved remarkable performance on various code generation benchmarks, there have been growing concerns regarding potential contamination of these benchmarks as they may be leaked into pretraining and finetuning data. While recent work has investigated contamination in natural language generation and understanding tasks, there has been less extensive research into how data contamination impacts the evaluation of code generation, which is critical for understanding the robustness and reliability of LLMs in programming contexts. In this work, we perform a comprehensive study of data contamination of popular code generation benchmarks, and precisely quantify their overlap with pretraining corpus through both surface-level and semantic-level matching. In our experiments, we show that there are substantial overlap between popular code generation benchmarks and open training corpus, and models perform significantly better on the subset of the benchmarks where similar solutions are seen during training. We also conduct extensive analysis on the factors that affects model memorization and generalization, such as model size, problem difficulty, and question length. We release all resulting files from our matching pipeline for future research."
                    },
                    {
                        "title": "Merging Weak and Active Supervision for Semantic Parsing",
                        "abstract": "A semantic parser maps natural language commands (NLs) from the users to executable meaning representations (MRs), which are later executed in certain environment to obtain user-desired results. The fully-supervised training of such parser requires NL/MR pairs, annotated by domain experts, which makes them expensive to collect. However, weakly-supervised semantic parsers are learnt only from pairs of NL and expected execution results, leaving the MRs latent. While weak supervision is cheaper to acquire, learning from this input poses difficulties. It demands that parsers search a large space with a very weak learning signal and it is hard to avoid spurious MRs that achieve the correct answer in the wrong way. These factors lead to a performance gap between parsers trained in weakly- and fully-supervised setting. To bridge this gap, we examine the intersection between weak supervision and active learning, which allows the learner to actively select examples and query for manual annotations as extra supervision to improve the model trained under weak supervision. We study different active learning heuristics for selecting examples to query, and various forms of extra supervision for such queries. We evaluate the effectiveness of our method on two different datasets. Experiments on the WikiSQL show that by annotating only 1.8% of examples, we improve over a state-of-the-art weakly-supervised baseline by 6.4%, achieving an accuracy of 79.0%, which is only 1.3% away from the model trained with full supervision. Experiments on WikiTableQuestions with human annotators show that our method can improve the performance with only 100 active queries, especially for weakly-supervised parsers learnt from a cold start."
                    },
                    {
                        "title": "Mitigating False-Negative Contexts in Multi-document Question Answering with Retrieval Marginalization",
                        "abstract": "Question Answering (QA) tasks requiring information from multiple documents often rely on a retrieval model to identify relevant information for reasoning. The retrieval model is typically trained to maximize the likelihood of the labeled supporting evidence. However, when retrieving from large text corpora such as Wikipedia, the correct answer can often be obtained from multiple evidence candidates. Moreover, not all such candidates are labeled as positive during annotation, rendering the training signal weak and noisy. This problem is exacerbated when the questions are unanswerable or when the answers are Boolean, since the model cannot rely on lexical overlap to make a connection between the answer and supporting evidence. We develop a new parameterization of set-valued retrieval that handles unanswerable queries, and we show that marginalizing over this set during training allows a model to mitigate false negatives in supporting evidence annotations. We test our method on two multi-document QA datasets, IIRC and HotpotQA. On IIRC, we show that joint modeling with marginalization improves model performance by 5.5 F1 points and achieves a new state-of-the-art performance of 50.5 F1. We also show that retrieval marginalization results in 4.1 QA F1 improvement over a non-marginalized baseline on HotpotQA in the fullwiki setting."
                    },
                    {
                        "title": "Explicit Knowledge Transfer for Weakly-Supervised Code Generation",
                        "abstract": "Large language models (LLMs) can acquire strong code-generation capabilities through few-shot learning. In contrast, supervised fine-tuning is still needed for smaller models to achieve good performance. Such fine-tuning demands a large number of task-specific NL-code pairs, which are expensive to obtain. In this paper, we attempt to transfer the code generation ability of an LLM to a smaller model with the aid of weakly-supervised data. More specifically, we propose explicit knowledge transfer (EKT), which uses the few-shot capabilities of a teacher LLM to create NL-code pairs that we then filter for correctness and fine-tune the student on. We evaluate EKT on the task of generating code solutions to math word problems from the GSM8k dataset. We find that EKT not only yields better performance than training with expert iteration, but also outperforms knowledge distillation, another form of knowledge transfer. A GPT-Neo 1.3B model trained using EKT with a GPT-J teacher achieves a 12.4% pass@100 on GSM8k, while the same student and teacher trained with knowledge distillation yield only a 3.7% pass@100. We also show that it is possible for a student model to outperform the teacher using EKT."
                    },
                    {
                        "title": "NExT: Teaching Large Language Models to Reason about Code Execution",
                        "abstract": "A fundamental skill among human developers is the ability to understand and reason about program execution. As an example, a programmer can mentally simulate code execution in natural language to debug and repair code (aka. rubber duck debugging). However, large language models (LLMs) of code are typically trained on the surface textual form of programs, thus may lack a semantic understanding of how programs execute at run-time. To address this issue, we propose NExT, a method to teach LLMs to inspect the execution traces of programs (variable states of executed lines) and reason about their run-time behavior through chain-of-thought (CoT) rationales. Specifically, NExT uses self-training to bootstrap a synthetic training set of execution-aware rationales that lead to correct task solutions (e.g., fixed programs) without laborious manual annotation. Experiments on program repair tasks based on MBPP and HumanEval demonstrate that NExT improves the fix rate of a PaLM 2 model, by 26.1% and 14.3% absolute, respectively, with significantly improved rationale quality as verified by automated metrics and human raters. Our model can also generalize to scenarios where program traces are absent at test-time."
                    },
                    {
                        "title": "SummerTime: Text Summarization Toolkit for Non-experts",
                        "abstract": "Recent advances in summarization provide models that can generate summaries of higher quality. Such models now exist for a number of summarization tasks, including query-based summarization, dialogue summarization, and multi-document summarization. While such models and tasks are rapidly growing in the research field, it has also become challenging for non-experts to keep track of them. To make summarization methods more accessible to a wider audience, we develop SummerTime by rethinking the summarization task from the perspective of an NLP non-expert. SummerTime is a complete toolkit for text summarization, including various models, datasets and evaluation metrics, for a full spectrum of summarization-related tasks. SummerTime integrates with libraries designed for NLP researchers, and enables users with easy-to-use APIs. With SummerTime, users can locate pipeline solutions and search for the best model with their own data, and visualize the differences, all with a few lines of code. We also provide explanations for models and evaluation metrics to help users understand the model behaviors and select models that best suit their needs. Our library, along with a notebook demo, is available at https://github.com/Yale-LILY/SummerTime."
                    },
                    {
                        "title": "Learning Math Reasoning from Self-Sampled Correct and Partially-Correct Solutions",
                        "abstract": "Pretrained language models have shown superior performance on many natural language processing tasks, yet they still struggle at multi-step formal reasoning tasks like grade school math problems. One key challenge of finetuning them to solve such math reasoning problems is that many existing datasets only contain one reference solution for each problem, despite the fact that there are often alternative solutions resembling different reasoning paths to the final answer. This way, the finetuned models are biased towards the limited reference solutions, which limits their generalization to unseen examples. To mitigate this issue, we propose to let the model perform sampling during training and learn from both self-sampled fully-correct solutions, which yield the correct answer upon execution, and partially-correct solutions, whose intermediate state matches an intermediate state of a known correct solution. We show that our use of self-sampled correct and partially-correct solutions can benefit learning and help guide the sampling process, leading to more efficient exploration of the solution space. Additionally, we explore various training objectives to support learning from multiple solutions per example and find they greatly affect the performance. Experiments on two math reasoning datasets show the effectiveness of our method compared to learning from a single reference solution with MLE, where we improve PASS@100 from 35.5% to 44.5% for GSM8K, and 27.6% to 36.2% PASS@80 for MathQA. Such improvements are also consistent across different model sizes. Our code is available at https://github.com/microsoft/TraceCodegen."
                    },
                    {
                        "title": "LEVER: Learning to Verify Language-to-Code Generation with Execution",
                        "abstract": "The advent of large language models trained on code (code LLMs) has led to significant progress in language-to-code generation. State-of-the-art approaches in this area combine LLM decoding with sample pruning and reranking using test cases or heuristics based on the execution results. However, it is challenging to obtain test cases for many real-world language-to-code applications, and heuristics cannot well capture the semantic features of the execution results, such as data type and value range, which often indicates the correctness of the program. In this work, we propose LEVER, a simple approach to improve language-to-code generation by learning to verify the generated programs with their execution results. Specifically, we train verifiers to determine whether a program sampled from the LLMs is correct or not based on the natural language input, the program itself and its execution results. The sampled programs are reranked by combining the verification score with the LLM generation probability, and marginalizing over programs with the same execution results. On four datasets across the domains of table QA, math QA and basic Python programming, LEVER consistently improves over the base code LLMs(4.6% to 10.9% with code-davinci-002) and achieves new state-of-the-art results on all of them."
                    },
                    {
                        "title": "Leveraging Locality in Abstractive Text Summarization",
                        "abstract": "Neural attention models have achieved significant improvements on many natural language processing tasks. However, the quadratic memory complexity of the self-attention module with respect to the input length hinders their applications in long text summarization. Instead of designing more efficient attention modules, we approach this problem by investigating if models with a restricted context can have competitive performance compared with the memory-efficient attention models that maintain a global context by treating the input as a single sequence. Our model is applied to individual pages which contain parts of inputs grouped by the principle of locality during both encoding and decoding. We empirically investigated three kinds of locality in text summarization at different levels of granularity, ranging from sentences to documents. Our experimental results show that our model has a better performance compared with strong baselines with efficient attention modules, and our analysis provides further insights into our locality-aware modeling strategy."
                    },
                    {
                        "title": "SOAR: A Synthesis Approach for Data Science API Refactoring",
                        "abstract": "With the growth of the open-source data science community, both the number of data science libraries and the number of versions for the same library are increasing rapidly. To match the evolving APIs from those libraries, open-source organizations often have to exert manual effort to refactor the APIs used in the code base. Moreover, due to the abundance of similar open-source libraries, data scientists working on a certain application may have an abundance of libraries to choose, maintain and migrate between. The manual refactoring between APIs is a tedious and error-prone task. Although recent research efforts were made on performing automatic API refactoring between different languages, previous work relies on statistical learning with collected pairwise training data for the API matching and migration. Using large statistical data for refactoring is not ideal because such training data will not be available for a new library or a new version of the same library. We introduce Synthesis for Open-Source API Refactoring (SOAR), a novel technique that requires no training data to achieve API migration and refactoring. SOAR relies only on the documentation that is readily available at the release of the library to learn API representations and mapping between libraries. Using program synthesis, SOAR automatically computes the correct configuration of arguments to the APIs and any glue code required to invoke those APIs. SOAR also uses the interpreter's error messages when running refactored code to generate logical constraints that can be used to prune the search space. Our empirical evaluation shows that SOAR can successfully refactor 80% of our benchmarks corresponding to deep learning models with up to 44 layers with an average run time of 97.23 seconds, and 90% of the data wrangling benchmarks with an average run time of 17.31 seconds."
                    },
                    {
                        "title": "L2CEval: Evaluating Language-to-Code Generation Capabilities of Large Language Models",
                        "abstract": "Recently, large language models (LLMs), especially those that are pretrained on code, have demonstrated strong capabilities in generating programs from natural language inputs in a few-shot or even zero-shot manner. Despite promising results, there is a notable lack of a comprehensive evaluation of these models language-to-code generation capabilities. Existing studies often focus on specific tasks, model architectures, or learning paradigms, leading to a fragmented understanding of the overall landscape. In this work, we present L2CEval, a systematic evaluation of the language-to-code generation capabilities of LLMs on 7 tasks across the domain spectrum of semantic parsing, math reasoning and Python programming, analyzing the factors that potentially affect their performance, such as model size, pretraining data, instruction tuning, and different prompting methods. In addition to assessing model performance, we measure confidence calibration for the models and conduct human evaluations of the output programs. This enables us to identify and analyze the typical failure modes across various tasks and models. L2CEval offers a comprehensive understanding of the capabilities and limitations of LLMs in language-to-code generation. We also release the evaluation framework and all model outputs, hoping to lay the groundwork for further future research in this domain."
                    },
                    {
                        "title": "An Exploratory Study on Long Dialogue Summarization: What Works and What's Next",
                        "abstract": "Dialogue summarization helps readers capture salient information from long conversations in meetings, interviews, and TV series. However, real-world dialogues pose a great challenge to current summarization models, as the dialogue length typically exceeds the input limits imposed by recent transformer-based pre-trained models, and the interactive nature of dialogues makes relevant information more context-dependent and sparsely distributed than news articles. In this work, we perform a comprehensive study on long dialogue summarization by investigating three strategies to deal with the lengthy input problem and locate relevant information: (1) extended transformer models such as Longformer, (2) retrieve-then-summarize pipeline models with several dialogue utterance retrieval methods, and (3) hierarchical dialogue encoding models such as HMNet. Our experimental results on three long dialogue datasets (QMSum, MediaSum, SummScreen) show that the retrieve-then-summarize pipeline models yield the best performance. We also demonstrate that the summary quality can be further improved with a stronger retrieval model and pretraining on proper external summarization datasets."
                    },
                    {
                        "title": "DYLE: Dynamic Latent Extraction for Abstractive Long-Input Summarization",
                        "abstract": "Transformer-based models have achieved state-of-the-art performance on short-input summarization. However, they still struggle with summarizing longer text. In this paper, we present DYLE, a novel dynamic latent extraction approach for abstractive long-input summarization. DYLE jointly trains an extractor and a generator and treats the extracted text snippets as the latent variable, allowing dynamic snippet-level attention weights during decoding. To provide adequate supervision, we propose simple yet effective heuristics for oracle extraction as well as a consistency loss term, which encourages the extractor to approximate the averaged dynamic weights predicted by the generator. We evaluate our method on different long-document and long-dialogue summarization tasks: GovReport, QMSum, and arXiv. Experiment results show that DYLE outperforms all existing methods on GovReport and QMSum, with gains up to 6.1 ROUGE, while yielding strong results on arXiv. Further analysis shows that the proposed dynamic weights provide interpretability of our generation process."
                    },
                    {
                        "title": "Summ^N: A Multi-Stage Summarization Framework for Long Input Dialogues and Documents",
                        "abstract": "Text summarization helps readers capture salient information from documents, news, interviews, and meetings. However, most state-of-the-art pretrained language models (LM) are unable to efficiently process long text for many summarization tasks. In this paper, we propose Summ$^N$, a simple, flexible, and effective multi-stage framework for input texts that are longer than the maximum context length of typical pretrained LMs. Summ$^N$ first splits the data samples and generates a coarse summary in multiple stages and then produces the final fine-grained summary based on it. Our framework can process input text of arbitrary length by adjusting the number of stages while keeping the LM input size fixed. Moreover, it can deal with both single-source documents and dialogues, and it can be used on top of different backbone abstractive summarization models. To the best of our knowledge, Summ$^N$ is the first multi-stage split-then-summarize framework for long input summarization. Our experiments demonstrate that Summ$^N$ outperforms previous state-of-the-art methods by improving ROUGE scores on three long meeting summarization datasets AMI, ICSI, and QMSum, two long TV series datasets from SummScreen, and a long document summarization dataset GovReport. Our data and code are available at https://github.com/psunlpgroup/Summ-N."
                    },
                    {
                        "title": "UnifiedSKG: Unifying and Multi-Tasking Structured Knowledge Grounding with Text-to-Text Language Models",
                        "abstract": "Structured knowledge grounding (SKG) leverages structured knowledge to complete user requests, such as semantic parsing over databases and question answering over knowledge bases. Since the inputs and outputs of SKG tasks are heterogeneous, they have been studied separately by different communities, which limits systematic and compatible research on SKG. In this paper, we overcome this limitation by proposing the UnifiedSKG framework, which unifies 21 SKG tasks into a text-to-text format, aiming to promote systematic SKG research, instead of being exclusive to a single task, domain, or dataset. We use UnifiedSKG to benchmark T5 with different sizes and show that T5, with simple modifications when necessary, achieves state-of-the-art performance on almost all of the 21 tasks. We further demonstrate that multi-task prefix-tuning improves the performance on most tasks, largely improving the overall performance. UnifiedSKG also facilitates the investigation of zero-shot and few-shot learning, and we show that T0, GPT-3, and Codex struggle in zero-shot and few-shot learning for SKG. We also use UnifiedSKG to conduct a series of controlled experiments on structured knowledge encoding variants across SKG tasks. UnifiedSKG is easily extensible to more tasks, and it is open-sourced at https://github.com/hkunlp/unifiedskg."
                    }
                ]
            },
            "f7f5a36b-bdec-40ce-806a-3a586e1fbfc6": {
                "pk": "f7f5a36b-bdec-40ce-806a-3a586e1fbfc6",
                "name": "Qian Liu",
                "collaborators": [
                    "Chongchao Huang",
                    "Mei Zhang",
                    "Dazhong He",
                    "Bing Wu"
                ],
                "domain": [
                    "Haptic Interaction",
                    "Quantum Communication",
                    "Tactile Perception",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "The combined equilibrium of business land-use and its congestion pricing principles (working paper)",
                        "abstract": "This working paper is divided into two parts. Firstly, we develop a new combined equilibrium model of business land-use, which puts travelers' traffic equilibrium and business companies' competitive location equilibrium into a unified framework. A variational inequality is presented for the combined equilibrium and the properties of equilibrium solution are investigated. Secondly, the congestion pricing principles associated with the combined equilibrium are studied. From the mathematical point of view, we prove that there exists an optimal road pricing scheme that can minimize the social cost of travelers. This road pricing scheme generalizes the traditional link-based optimal road pricing scheme. Furthermore, when simultaneously imposing charges on travelers and companies is allowed, we prove that there exists an optimal congestion pricing scheme that can derive a combined equilibrium toward an overall system optimum. The economic meaning of every pricing scheme proposed in this paper is discussed in detail. At last, a simple numerical example is used to demonstrate that the optimal congestion pricing scheme may indeed reduce the social cost."
                    },
                    {
                        "title": "Generation and complete nondestructive analysis of hyperentanglement assisted by nitrogen-vacancy centers in resonators",
                        "abstract": "We present two efficient schemes for the deterministic generation and the complete nondestructive analysis of hyperentangled Bell states in both the polarization and spatial-mode degrees of freedom (DOFs) of two-photon systems, assisted by the nitrogen-vacancy (NV) centers in diamonds coupled to microtoroidal resonators as a result of cavity quantum electrodynamics (QED). With the input-output process of photons, two-photon polarization-spatial hyperentangled Bell states can be generated in a deterministic way and their complete nondestructive analysis can be achieved. These schemes can be generalized to generate and analyze hyperentangled Greenberger-Horne-Zeilinger states of multi-photon systems as well. Compared with previous works, these two schemes relax the difficulty of their implementation in experiment as it is not difficult to obtain the $\\pi$ phase shift in single-sided NV-cavity systems. Moreover, our schemes do not require that the transmission for the uncoupled cavity is balanceable with the reflectance for the coupled cavity. Our calculations show that these schemes can reach a high fidelity and efficiency with current technology, which may be a benefit to long-distance high-capacity quantum communication with two DOFs of photon systems."
                    },
                    {
                        "title": "Self-supervised Spatio-Temporal Graph Mask-Passing Attention Network for Perceptual Importance Prediction of Multi-point Tactility",
                        "abstract": "While visual and auditory information are prevalent in modern multimedia systems, haptic interaction, e.g., tactile and kinesthetic interaction, provides a unique form of human perception. However, multimedia technology for contact interaction is less mature than non-contact multimedia technologies and requires further development. Specialized haptic media technologies, requiring low latency and bitrates, are essential to enable haptic interaction, necessitating haptic information compression. Existing vibrotactile signal compression methods, based on the perceptual model, do not consider the characteristics of fused tactile perception at multiple spatially distributed interaction points. In fact, differences in tactile perceptual importance are not limited to conventional frequency and time domains, but also encompass differences in the spatial locations on the skin unique to tactile perception. For the most frequently used tactile information, vibrotactile texture perception, we have developed a model to predict its perceptual importance at multiple points, based on self-supervised learning and Spatio-Temporal Graph Neural Network. Current experimental results indicate that this model can effectively predict the perceptual importance of various points in multi-point tactile perception scenarios."
                    },
                    {
                        "title": "TactileAR: Active Tactile Pattern Reconstruction",
                        "abstract": "High-resolution (HR) contact surface information is essential for robotic grasping and precise manipulation tasks. However, it remains a challenge for current taxel-based sensors to obtain HR tactile information. In this paper, we focus on utilizing low-resolution (LR) tactile sensors to reconstruct the localized, dense, and HR representation of contact surfaces. In particular, we build a Gaussian triaxial tactile sensor degradation model and propose a tactile pattern reconstruction framework based on the Kalman filter. This framework enables the reconstruction of 2-D HR contact surface shapes using collected LR tactile sequences. In addition, we present an active exploration strategy to enhance the reconstruction efficiency. We evaluate the proposed method in real-world scenarios with comparison to existing prior-information-based approaches. Experimental results confirm the efficiency of the proposed approach and demonstrate satisfactory reconstructions of complex contact surface shapes. Code: https://github.com/wmtlab/tactileAR"
                    }
                ]
            },
            "a7399d2f-d2fd-4ece-ac72-650606c65b2c": {
                "pk": "a7399d2f-d2fd-4ece-ac72-650606c65b2c",
                "name": "Victor Zhong",
                "collaborators": [
                    "Luke Zettlemoyer",
                    "Caiming Xiong",
                    "Richard Socher",
                    "Tim Rockt\u00e4schel",
                    "Edward Grefenstette",
                    "Machel Reid",
                    "Sewon Min",
                    "Jesse Mu",
                    "Hannaneh Hajishirzi",
                    "Austin W. Hanjie"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Question Answering",
                    "Reinforcement Learning",
                    "Dialogue Systems"
                ],
                "publications": [
                    {
                        "title": "E3: Entailment-driven Extracting and Editing for Conversational Machine Reading",
                        "abstract": "Conversational machine reading systems help users answer high-level questions (e.g. determine if they qualify for particular government benefits) when they do not know the exact rules by which the determination is made(e.g. whether they need certain income levels or veteran status). The key challenge is that these rules are only provided in the form of a procedural text (e.g. guidelines from government website) which the system must read to figure out what to ask the user. We present a new conversational machine reading model that jointly extracts a set of decision rules from the procedural text while reasoning about which are entailed by the conversational history and which still need to be edited to create questions for the user. On the recently introduced ShARC conversational machine reading dataset, our Entailment-driven Extract and Edit network (E3) achieves a new state-of-the-art, outperforming existing systems as well as a new BERT-based baseline. In addition, by explicitly highlighting which information still needs to be gathered, E3 provides a more explainable alternative to prior work. We release source code for our models and experiments at https://github.com/vzhong/e3."
                    },
                    {
                        "title": "LEWIS: Levenshtein Editing for Unsupervised Text Style Transfer",
                        "abstract": "Many types of text style transfer can be achieved with only small, precise edits (e.g. sentiment transfer from I had a terrible time... to I had a great time...). We propose a coarse-to-fine editor for style transfer that transforms text using Levenshtein edit operations (e.g. insert, replace, delete). Unlike prior single-span edit methods, our method concurrently edits multiple spans in the source text. To train without parallel style text pairs (e.g. pairs of +/- sentiment statements), we propose an unsupervised data synthesis procedure. We first convert text to style-agnostic templates using style classifier attention (e.g. I had a SLOT time...), then fill in slots in these templates using fine-tuned pretrained language models. Our method outperforms existing generation and editing style transfer methods on sentiment (Yelp, Amazon) and politeness (Polite) transfer. In particular, multi-span editing achieves higher performance and more diverse output than single-span editing. Moreover, compared to previous methods on unsupervised data synthesis, our method results in higher quality parallel style pairs and improves model performance."
                    },
                    {
                        "title": "DCN+: Mixed Objective and Deep Residual Coattention for Question Answering",
                        "abstract": "Traditional models for question answering optimize using cross entropy loss, which encourages exact answers at the cost of penalizing nearby or overlapping answers that are sometimes equally accurate. We propose a mixed objective that combines cross entropy loss with self-critical policy learning. The objective uses rewards derived from word overlap to solve the misalignment between evaluation metric and optimization objective. In addition to the mixed objective, we improve dynamic coattention networks (DCN) with a deep residual coattention encoder that is inspired by recent work in deep self-attention and residual networks. Our proposals improve model performance across question types and input lengths, especially for long questions that requires the ability to capture long-term dependencies. On the Stanford Question Answering Dataset, our model achieves state-of-the-art results with 75.1% exact match accuracy and 83.1% F1, while the ensemble obtains 78.9% exact match accuracy and 86.0% F1."
                    },
                    {
                        "title": "Global-Locally Self-Attentive Dialogue State Tracker",
                        "abstract": "Dialogue state tracking, which estimates user goals and requests given the dialogue context, is an essential part of task-oriented dialogue systems. In this paper, we propose the Global-Locally Self-Attentive Dialogue State Tracker (GLAD), which learns representations of the user utterance and previous system actions with global-local modules. Our model uses global modules to share parameters between estimators for different types (called slots) of dialogue states, and uses local modules to learn slot-specific features. We show that this significantly improves tracking of rare states and achieves state-of-the-art performance on the WoZ and DSTC2 state tracking tasks. GLAD obtains 88.1% joint goal accuracy and 97.1% request accuracy on WoZ, outperforming prior work by 3.7% and 5.5%. On DSTC2, our model obtains 74.5% joint goal accuracy and 97.5% request accuracy, outperforming prior work by 1.1% and 1.0%."
                    },
                    {
                        "title": "Dynamic Coattention Networks For Question Answering",
                        "abstract": "Several deep learning models have been proposed for question answering. However, due to their single-pass nature, they have no way to recover from local maxima corresponding to incorrect answers. To address this problem, we introduce the Dynamic Coattention Network (DCN) for question answering. The DCN first fuses co-dependent representations of the question and the document in order to focus on relevant parts of both. Then a dynamic pointing decoder iterates over potential answer spans. This iterative procedure enables the model to recover from initial local maxima corresponding to incorrect answers. On the Stanford question answering dataset, a single DCN model improves the previous state of the art from 71.0% F1 to 75.9%, while a DCN ensemble obtains 80.4% F1."
                    },
                    {
                        "title": "RTFM: Generalising to Novel Environment Dynamics via Reading",
                        "abstract": "Obtaining policies that can generalise to new environments in reinforcement learning is challenging. In this work, we demonstrate that language understanding via a reading policy learner is a promising vehicle for generalisation to new environments. We propose a grounded policy learning problem, Read to Fight Monsters (RTFM), in which the agent must jointly reason over a language goal, relevant dynamics described in a document, and environment observations. We procedurally generate environment dynamics and corresponding language descriptions of the dynamics, such that agents must read to understand new environment dynamics instead of memorising any particular information. In addition, we propose txt2$\\pi$, a model that captures three-way interactions between the goal, document, and observations. On RTFM, txt2$\\pi$ generalises to new environments with dynamics not seen during training via reading. Furthermore, our model outperforms baselines such as FiLM and language-conditioned CNNs on RTFM. Through curriculum learning, txt2$\\pi$ produces policies that excel on complex RTFM tasks requiring several reasoning and coreference steps."
                    },
                    {
                        "title": "Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning",
                        "abstract": "A significant amount of the world's knowledge is stored in relational databases. However, the ability for users to retrieve facts from a database is limited due to a lack of understanding of query languages such as SQL. We propose Seq2SQL, a deep neural network for translating natural language questions to corresponding SQL queries. Our model leverages the structure of SQL queries to significantly reduce the output space of generated queries. Moreover, we use rewards from in-the-loop query execution over the database to learn a policy to generate unordered parts of the query, which we show are less suitable for optimization via cross entropy loss. In addition, we will publish WikiSQL, a dataset of 80654 hand-annotated examples of questions and SQL queries distributed across 24241 tables from Wikipedia. This dataset is required to train our model and is an order of magnitude larger than comparable datasets. By applying policy-based reinforcement learning with a query execution environment to WikiSQL, our model Seq2SQL outperforms attentional sequence to sequence models, improving execution accuracy from 35.9% to 59.4% and logical form accuracy from 23.4% to 48.3%."
                    },
                    {
                        "title": "Efficient and Robust Question Answering from Minimal Context over Documents",
                        "abstract": "Neural models for question answering (QA) over documents have achieved significant performance improvements. Although effective, these models do not scale to large corpora due to their complex modeling of interactions between the document and the question. Moreover, recent work has shown that such models are sensitive to adversarial inputs. In this paper, we study the minimal context required to answer the question, and find that most questions in existing datasets can be answered with a small set of sentences. Inspired by this observation, we propose a simple sentence selector to select the minimal set of sentences to feed into the QA model. Our overall system achieves significant reductions in training (up to 15 times) and inference times (up to 13 times), with accuracy comparable to or better than the state-of-the-art on SQuAD, NewsQA, TriviaQA and SQuAD-Open. Furthermore, our experimental results and analyses show that our approach is more robust to adversarial inputs."
                    },
                    {
                        "title": "Multi-hop Reading Comprehension through Question Decomposition and Rescoring",
                        "abstract": "Multi-hop Reading Comprehension (RC) requires reasoning and aggregation across several paragraphs. We propose a system for multi-hop RC that decomposes a compositional question into simpler sub-questions that can be answered by off-the-shelf single-hop RC models. Since annotations for such decomposition are expensive, we recast sub-question generation as a span prediction problem and show that our method, trained using only 400 labeled examples, generates sub-questions that are as effective as human-authored sub-questions. We also introduce a new global rescoring approach that considers each decomposition (i.e. the sub-questions and their answers) to select the best final answer, greatly improving overall performance. Our experiments on HotpotQA show that this approach achieves the state-of-the-art results, while providing explainable evidence for its decision making in the form of sub-questions."
                    },
                    {
                        "title": "Grounding Language to Entities and Dynamics for Generalization in Reinforcement Learning",
                        "abstract": "We investigate the use of natural language to drive the generalization of control policies and introduce the new multi-task environment Messenger with free-form text manuals describing the environment dynamics. Unlike previous work, Messenger does not assume prior knowledge connecting text and state observations $-$ the control policy must simultaneously ground the game manual to entity symbols and dynamics in the environment. We develop a new model, EMMA (Entity Mapper with Multi-modal Attention) which uses an entity-conditioned attention module that allows for selective focus over relevant descriptions in the manual for each entity in the environment. EMMA is end-to-end differentiable and learns a latent grounding of entities and dynamics from text to observations using only environment rewards. EMMA achieves successful zero-shot generalization to unseen games with new dynamics, obtaining a 40% higher win rate compared to multiple baselines. However, win rate on the hardest stage of Messenger remains low (10%), demonstrating the need for additional work in this direction."
                    },
                    {
                        "title": "M2D2: A Massively Multi-domain Language Modeling Dataset",
                        "abstract": "We present M2D2, a fine-grained, massively multi-domain corpus for studying domain adaptation in language models (LMs). M2D2 consists of 8.5B tokens and spans 145 domains extracted from Wikipedia and Semantic Scholar. Using ontologies derived from Wikipedia and ArXiv categories, we organize the domains in each data source into 22 groups. This two-level hierarchy enables the study of relationships between domains and their effects on in- and out-of-domain performance after adaptation. We also present a number of insights into the nature of effective domain adaptation in LMs, as examples of the new types of studies M2D2 enables. To improve in-domain performance, we show the benefits of adapting the LM along a domain hierarchy; adapting to smaller amounts of fine-grained domain-specific data can lead to larger in-domain performance gains than larger amounts of weakly relevant data. We further demonstrate a trade-off between in-domain specialization and out-of-domain generalization within and across ontologies, as well as a strong correlation between out-of-domain performance and lexical overlap between domains."
                    },
                    {
                        "title": "Coarse-grain Fine-grain Coattention Network for Multi-evidence Question Answering",
                        "abstract": "End-to-end neural models have made significant progress in question answering, however recent studies show that these models implicitly assume that the answer and evidence appear close together in a single document. In this work, we propose the Coarse-grain Fine-grain Coattention Network (CFC), a new question answering model that combines information from evidence across multiple documents. The CFC consists of a coarse-grain module that interprets documents with respect to the query then finds a relevant answer, and a fine-grain module which scores each candidate answer by comparing its occurrences across all of the documents with the query. We design these modules using hierarchies of coattention and self-attention, which learn to emphasize different parts of the input. On the Qangaroo WikiHop multi-evidence question answering task, the CFC obtains a new state-of-the-art result of 70.6% on the blind test set, outperforming the previous best by 3% accuracy despite not using pretrained contextual encoders."
                    },
                    {
                        "title": "Grounded Adaptation for Zero-shot Executable Semantic Parsing",
                        "abstract": "We propose Grounded Adaptation for Zero-shot Executable Semantic Parsing (GAZP) to adapt an existing semantic parser to new environments (e.g. new database schemas). GAZP combines a forward semantic parser with a backward utterance generator to synthesize data (e.g. utterances and SQL queries) in the new environment, then selects cycle-consistent examples to adapt the parser. Unlike data-augmentation, which typically synthesizes unverified examples in the training environment, GAZP synthesizes examples in the new environment whose input-output consistency are verified. On the Spider, Sparc, and CoSQL zero-shot semantic parsing tasks, GAZP improves logical form and execution accuracy of the baseline parser. Our analyses show that GAZP outperforms data-augmentation in the training environment, performance increases with the amount of GAZP-synthesized data, and cycle-consistency is central to successful adaptation."
                    },
                    {
                        "title": "Improving Policy Learning via Language Dynamics Distillation",
                        "abstract": "Recent work has shown that augmenting environments with language descriptions improves policy learning. However, for environments with complex language abstractions, learning how to ground language to observations is difficult due to sparse, delayed rewards. We propose Language Dynamics Distillation (LDD), which pretrains a model to predict environment dynamics given demonstrations with language descriptions, and then fine-tunes these language-aware pretrained representations via reinforcement learning (RL). In this way, the model is trained to both maximize expected reward and retain knowledge about how language relates to environment dynamics. On SILG, a benchmark of five tasks with language descriptions that evaluate distinct generalization challenges on unseen environments (NetHack, ALFWorld, RTFM, Messenger, and Touchdown), LDD outperforms tabula-rasa RL, VAE pretraining, and methods that learn from unlabeled demonstrations in inverse RL and reward shaping with pretrained experts. In our analyses, we show that language descriptions in demonstrations improve sample-efficiency and generalization across environments, and that dynamics modelling with expert demonstrations is more effective than with non-experts."
                    },
                    {
                        "title": "Policy Improvement using Language Feedback Models",
                        "abstract": "We introduce Language Feedback Models (LFMs) that identify desirable behaviour - actions that help achieve tasks specified in the instruction - for imitation learning in instruction following. To train LFMs, we obtain feedback from Large Language Models (LLMs) on visual trajectories verbalized to language descriptions. First, by using LFMs to identify desirable behaviour to imitate, we improve in task-completion rate over strong behavioural cloning baselines on three distinct language grounding environments (Touchdown, ScienceWorld, and ALFWorld). Second, LFMs outperform using LLMs as experts to directly predict actions, when controlling for the number of LLM output tokens. Third, LFMs generalize to unseen environments, improving task-completion rate by 3.5-12.0% through one round of adaptation. Finally, LFM can be modified to provide human-interpretable feedback without performance loss, allowing human verification of desirable behaviour for imitation learning."
                    },
                    {
                        "title": "RoMQA: A Benchmark for Robust, Multi-evidence, Multi-answer Question Answering",
                        "abstract": "We introduce RoMQA, the first benchmark for robust, multi-evidence, multi-answer question answering (QA). RoMQA contains clusters of questions that are derived from related constraints mined from the Wikidata knowledge graph. RoMQA evaluates robustness of QA models to varying constraints by measuring worst-case performance within each question cluster. Compared to prior QA datasets, RoMQA has more human-written questions that require reasoning over more evidence text and have, on average, many more correct answers. In addition, human annotators rate RoMQA questions as more natural or likely to be asked by people. We evaluate state-of-the-art large language models in zero-shot, few-shot, and fine-tuning settings, and find that RoMQA is challenging: zero-shot and few-shot models perform similarly to naive baselines, while supervised retrieval methods perform well below gold evidence upper bounds. Moreover, existing models are not robust to variations in question constraints, but can be made more robust by tuning on clusters of related questions. Our results show that RoMQA is a challenging benchmark for large language models, and provides a quantifiable test to build more robust QA methods."
                    },
                    {
                        "title": "Improving Intrinsic Exploration with Language Abstractions",
                        "abstract": "Reinforcement learning (RL) agents are particularly hard to train when rewards are sparse. One common solution is to use intrinsic rewards to encourage agents to explore their environment. However, recent intrinsic exploration methods often use state-based novelty measures which reward low-level exploration and may not scale to domains requiring more abstract skills. Instead, we explore natural language as a general medium for highlighting relevant abstractions in an environment. Unlike previous work, we evaluate whether language can improve over existing exploration methods by directly extending (and comparing to) competitive intrinsic exploration baselines: AMIGo (Campero et al., 2021) and NovelD (Zhang et al., 2021). These language-based variants outperform their non-linguistic forms by 47-85% across 13 challenging tasks from the MiniGrid and MiniHack environment suites."
                    }
                ]
            },
            "044bd3aa-70ba-4fe4-9416-06dd0d43e373": {
                "pk": "044bd3aa-70ba-4fe4-9416-06dd0d43e373",
                "name": "Lu Chen",
                "collaborators": [],
                "domain": [
                    "Philosophy of Science",
                    "Geometry",
                    "Nonstandard Analysis",
                    "Structural Realism"
                ],
                "publications": [
                    {
                        "title": "Why the Weyl Tile Argument is Wrong",
                        "abstract": "Weyl famously argued that if space were discrete, then Euclidean geometry could not hold even approximately. Since then, many philosophers have responded to this argument by advancing alternative accounts of discrete geometry that recover approximately Euclidean space. However, they have missed an importantly flawed assumption in Weyl's argument: physical geometry is determined by fundamental spacetime structures independently from dynamical laws. In this paper, I aim to show its falsity through two rigorous examples: random walks in statistical physics and quantum mechanics."
                    },
                    {
                        "title": "Infinitesimal Gunk",
                        "abstract": "In this paper, I advance an original view of the structure of space called \\textit{Infinitesimal Gunk}. This view says that every region of space can be further divided and some regions have infinitesimal size, where infinitesimals are understood in the framework of Robinson's (1966) nonstandard analysis. This view, I argue, provides a novel reply to the inconsistency arguments proposed by Arntzenius (2008) and Russell (2008), which have troubled a more familiar gunky approach. Moreover, it has important advantages over the alternative views these authors suggested. Unlike Arntzenius's proposal, it does not introduce regions with no interior. It also has a much richer measure theory than Russell's proposal and does not retreat to mere finite additivity."
                    },
                    {
                        "title": "A partial defense of algebraic relationalism",
                        "abstract": "I defend algebraicism, according to which physical fields can be understood in terms of their structural relations without reference to a spacetime manifold, as a genuine relationalist view against the conventional wisdom that it is equivalent to substantivalism, according to which spacetime exists fundamentally. I criticize the standard version of algebraicism that is considered equivalent to substantivalism. Furthermore, I present alternative examples of algebraicism that better implement relationalism with their conceptual advantages over substantivalism or its standard algebraic counterpart."
                    },
                    {
                        "title": "Symmetries as Isomorphisms",
                        "abstract": "Symmetries and isomorphisms play similar conceptual roles when we consider how models represent physical situations, but they are formally distinct, as two models related by symmetries are not typically isomorphic. I offer a rigorous categorical strategy that formulate symmetries as isomorphisms between models and apply it to classical electromagnetism, and evaluate its philosophical significance in relation to the recent debate between `sophistication' and `reduction'. In addition to traditional spacetime models, I also consider algebraic models, in which case we can use the method of natural operators to address the problem of ontological nonperspicuity faced by the categorical strategy. Finally, I briefly expound on the significance of symmetries as isomorphisms in the framework of Univalent Foundations, in which isomorphic structures are formally identified."
                    },
                    {
                        "title": "Univalence and Ontic Structuralism",
                        "abstract": "The persistent challenge of formulating ontic structuralism in a rigorous manner, which prioritizes structures over the entities they contain, calls for a transformation of traditional logical frameworks. I argue that Univalent Foundations (UF), which feature the axiom that all isomorphic structures are identical, offer such a foundation and are more attractive than other proposed structuralist frameworks. Furthermore, I delve into the significance in the case of the hole argument and, very briefly, the nature of symmetries."
                    },
                    {
                        "title": "KAM theorem with large twist and finite smooth large perturbation",
                        "abstract": "In the present paper, we will discuss the following non-degenerate Hamiltonian system \\begin{equation*} H(\\theta,t,I)=\\frac{H_0(I)}{\\varepsilon^{a}}+\\frac{P(\\theta,t,I)}{\\varepsilon^{b}}, \\end{equation*}   where $(\\theta,t,I)\\in\\mathbf{{T}}^{d+1}\\times[1,2]^d$ ($\\mathbf{{T}}:=\\mathbf{{R}}/{2\\pi \\mathbf{Z}}$), $a,b$ are given positive constants with $a>b$, $H_0: [1,2]^d\\rightarrow \\mathbf R$ is real analytic and $P: \\mathbf T^{d+1}\\times [1,2]^d\\rightarrow \\mathbf R$ is $C^{\\ell}$ with $\\ell=\\frac{2(d+1)(5a-b+2ad)}{a-b}+\\mu$, $0<\\mu\\ll1$. We prove that if $\\varepsilon$ is sufficiently small, there is an invariant torus with given Diophantine frequency vector for the above Hamiltonian system. As for application, we prove that a finite network of Duffing oscillators with periodic exterior forces possesses Lagrangian stability for almost all initial data."
                    },
                    {
                        "title": "Intrinsic Local Distances: A Mixed Solution to Weyl's Tile Argument",
                        "abstract": "Weyl's tile argument purports to show that there are no natural distance functions in atomistic space that approximate Euclidean geometry. I advance a response to this argument that relies on a new account of distance in atomistic space, called \\textit{the mixed account}, according to which \\textit{local distances} are primitive and other distances are derived from them. Under this account, atomistic space can approximate Euclidean space (and continuous space in general) very well. To motivate this account as a genuine solution to Weyl's tile argument, I argue that this account is no less natural than the standard account of distance in continuous space. I also argue that the mixed account has distinctive advantages over Forrest's (1995) account in response to Weyl's tile argument, which can be considered as a restricted version of the mixed account."
                    },
                    {
                        "title": "Smooth Infinitesimals in the Metaphysical Foundation of Spacetime Theories",
                        "abstract": "I propose a theory of space with infinitesimal regions called \\textit{smooth infinitesimal geometry} (SIG) based on certain algebraic objects (i.e., rings), which regiments a mode of reasoning heuristically used by geometricists and physicists (e.g., circle is composed of infinitely many straight lines). I argue that SIG has the following utilities. (1) It provides a simple metaphysics of vector fields and tangent space that are otherwise perplexing. A tangent space can be considered an infinitesimal region of space. (2) It generalizes a standard implementation of spacetime algebraicism (according to which physical fields exist fundamentally without an underlying manifold) called \\textit{Einstein algebras}. (3) It solves the long-standing problem of interpreting \\textit{smooth infinitesimal analysis} (SIA) realistically, an alternative foundation of spacetime theories to real analysis (Lawvere 1980). SIA is formulated in intuitionistic logic and is thought to have no classical reformulations (Hellman 2006). Against this, I argue that SIG is (part of) such a reformulation. But SIG has an unorthodox mereology, in which the principle of supplementation fails."
                    },
                    {
                        "title": "Do Simple Infinitesimal Parts Solve Zeno's Paradox of Measure?",
                        "abstract": "I develop a new view of the structure of space--called infinitesimal atomism--as a reply to Zeno's paradox of measure. According to this view, space is composed of ultimate parts with infinitesimal size, where infinitesimals are understood within the framework of Robinson's nonstandard analysis. Notably, this view satisfies a version of additivity: for every region that has a size, its size is the sum of the sizes of its disjoint parts. In particular, the size of a finite region is the sum of the sizes of its infinitesimal parts. Although this view is a coherent approach to Zeno's paradox and is preferable to Skyrms's (1983) infinitesimal approach, it faces both the main problem for the standard view (the problem of unmeasurable regions) and the main problem for finite atomism (Weyl's tile argument), leaving it with no clear advantage over these familiar alternatives."
                    },
                    {
                        "title": "Can we \"effectivize'' spacetime?",
                        "abstract": "According to \\textit{effective realism}, scientific theories give us knowledge about the unobservable world, but not at the fundamental level. This view is justified by the well-received \\textit{effective-field-theory} (EFT) approach to physics, according to which our best physical theories are only applicable up to a certain energy scale and expected to break down beyond that. In this paper, I explain the motivations for the EFT approach and effective realism and their benefits. I also raise new challenges for this approach. Applying effective realism to \\textit{effective quantum gravity} (EQG) reveals its shortcomings: EQG does not give us a realistic theory of spacetime even within its scope of validity. It also exposes a general interpretative dilemma faced by all EFTs concerning their indispensable references to classical spacetime beyond their scope of validity."
                    },
                    {
                        "title": "On invariant tori in some reversible systems",
                        "abstract": "In the present paper, we consider the following reversible system \\begin{equation*} \\begin{cases} \\dot{x}=\\omega_0+f(x,y),\\\\ \\dot{y}=g(x,y), \\end{cases} \\end{equation*} where $x\\in\\mathbf{T}^{d}$, $y\\backsim0\\in \\mathbf{R}^{d}$, $\\omega_0$ is Diophantine, $f(x,y)=O(y)$, $g(x,y)=O(y^2)$ and $f$, $g$ are reversible with respect to the involution G: $(x,y)\\mapsto(-x,y)$, that is, $f(-x,y)=f(x,y)$, $g(-x,y)=-g(x,y)$. We study the accumulation of an analytic invariant torus $\\Gamma_0$ of the reversible system with Diophantine frequency $\\omega_0$ by other invariant tori. We will prove that if the Birkhoff normal form around $\\Gamma_0$ is 0-degenerate, then $\\Gamma_0$ is accumulated by other analytic invariant tori, the Lebesgue measure of the union of these tori being positive and the density of the union of these tori at $\\Gamma_0$ being one. We will also prove that if the Birkhoff normal form around $\\Gamma_0$ is $j$-degenerate ($1\\leq j\\leq d-1$) and condition (1.6) is satisfied, then through $\\Gamma_0$ there passes an analytic subvariety of dimension $d+j$ foliated into analytic invariant tori with frequency vector $\\omega_0$. If the Birkhoff normal form around $\\Gamma_0$ is $d-1$-degenerate, we will prove a stronger result, that is, a full neighborhood of $\\Gamma_0$ is foliated into analytic invariant tori with frequency vectors proportional to $\\omega_0$."
                    }
                ]
            },
            "daa72483-86fe-4573-b71e-3c35b7c3bb45": {
                "pk": "daa72483-86fe-4573-b71e-3c35b7c3bb45",
                "name": "Kai Yu",
                "collaborators": [
                    "Chenpeng Du",
                    "Su Zhu",
                    "Kai Sun",
                    "Qizhe Xie",
                    "Tong Zhang",
                    "Heinrich Dinkel",
                    "Tao Liu",
                    "Tianxing He",
                    "Yu Zhang",
                    "Jasha Droppo"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Speech Recognition",
                    "Machine Learning",
                    "Semi-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Recurrent Polynomial Network for Dialogue State Tracking",
                        "abstract": "Dialogue state tracking (DST) is a process to estimate the distribution of the dialogue states as a dialogue progresses. Recent studies on constrained Markov Bayesian polynomial (CMBP) framework take the first step towards bridging the gap between rule-based and statistical approaches for DST. In this paper, the gap is further bridged by a novel framework -- recurrent polynomial network (RPN). RPN's unique structure enables the framework to have all the advantages of CMBP including efficiency, portability and interpretability. Additionally, RPN achieves more properties of statistical approaches than CMBP. RPN was evaluated on the data corpora of the second and the third Dialog State Tracking Challenge (DSTC-2/3). Experiments showed that RPN can significantly outperform both traditional rule-based approaches and statistical approaches with similar feature set. Compared with the state-of-the-art statistical DST approaches with a lot richer features, RPN is also competitive."
                    },
                    {
                        "title": "High Dimensional Nonlinear Learning using Local Coordinate Coding",
                        "abstract": "This paper introduces a new method for semi-supervised learning on high dimensional nonlinear manifolds, which includes a phase of unsupervised basis learning and a phase of supervised function learning. The learned bases provide a set of anchor points to form a local coordinate system, such that each data point $x$ on the manifold can be locally approximated by a linear combination of its nearby anchor points, with the linear weights offering a local-coordinate coding of $x$. We show that a high dimensional nonlinear function can be approximated by a global linear function with respect to this coding scheme, and the approximation quality is ensured by the locality of such coding. The method turns a difficult nonlinear learning problem into a simple global linear learning problem, which overcomes some drawbacks of traditional local learning methods. The work also gives a theoretical justification to the empirical success of some biologically-inspired models using sparse coding of sensory data, since a local coding scheme must be sufficiently sparse. However, sparsity does not always satisfy locality conditions, and can thus possibly lead to suboptimal results. The properties and performances of the method are empirically verified on synthetic data, handwritten digit classification, and object recognition tasks."
                    },
                    {
                        "title": "Duration robust weakly supervised sound event detection",
                        "abstract": "Task 4 of the DCASE2018 challenge demonstrated that substantially more research is needed for a real-world application of sound event detection. Analyzing the challenge results it can be seen that most successful models are biased towards predicting long (e.g., over 5s) clips. This work aims to investigate the performance impact of fixed-sized window median filter post-processing and advocate the use of double thresholding as a more robust and predictable post-processing method. Further, four different temporal subsampling methods within the CRNN framework are proposed: mean-max, alpha-mean-max, Lp-norm and convolutional. We show that for this task subsampling the temporal resolution by a neural network enhances the F1 score as well as its robustness towards short, sporadic sound events. Our best single model achieves 30.1% F1 on the evaluation set and the best fusion model 32.5%, while being robust to event length variations."
                    },
                    {
                        "title": "Rich Prosody Diversity Modelling with Phone-level Mixture Density Network",
                        "abstract": "Generating natural speech with diverse and smooth prosody pattern is a challenging task. Although random sampling with phone-level prosody distribution has been investigated to generate different prosody patterns, the diversity of the generated speech is still very limited and far from what can be achieved by human. This is largely due to the use of uni-modal distribution, such as single Gaussian, in the prior works of phone-level prosody modelling. In this work, we propose a novel approach that models phone-level prosodies with GMM based mixture density network (GMM-MDN). Experiments on the LJSpeech dataset demonstrate that phone-level prosodies can precisely control the synthetic speech and GMM-MDN can generate more natural and smooth prosody pattern than a single Gaussian. Subjective evaluations further show that the proposed approach not only achieves better naturalness, but also significantly improves the prosody diversity in synthetic speech without the need of manual control."
                    },
                    {
                        "title": "Concept Transfer Learning for Adaptive Language Understanding",
                        "abstract": "Concept definition is important in language understanding (LU) adaptation since literal definition difference can easily lead to data sparsity even if different data sets are actually semantically correlated. To address this issue, in this paper, a novel concept transfer learning approach is proposed. Here, substructures within literal concept definition are investigated to reveal the relationship between concepts. A hierarchical semantic representation for concepts is proposed, where a semantic slot is represented as a composition of {\\em atomic concepts}. Based on this new hierarchical representation, transfer learning approaches are developed for adaptive LU. The approaches are applied to two tasks: value set mismatch and domain adaptation, and evaluated on two LU benchmarks: ATIS and DSTC 2\\&3. Thorough empirical studies validate both the efficiency and effectiveness of the proposed method. In particular, we achieve state-of-the-art performance ($F_1$-score 96.08\\%) on ATIS by only using lexicon features."
                    },
                    {
                        "title": "Encoder-decoder with Focus-mechanism for Sequence Labelling Based Spoken Language Understanding",
                        "abstract": "This paper investigates the framework of encoder-decoder with attention for sequence labelling based spoken language understanding. We introduce Bidirectional Long Short Term Memory - Long Short Term Memory networks (BLSTM-LSTM) as the encoder-decoder model to fully utilize the power of deep learning. In the sequence labelling task, the input and output sequences are aligned word by word, while the attention mechanism cannot provide the exact alignment. To address this limitation, we propose a novel focus mechanism for encoder-decoder framework. Experiments on the standard ATIS dataset showed that BLSTM-LSTM with focus mechanism defined the new state-of-the-art by outperforming standard BLSTM and attention based encoder-decoder. Further experiments also show that the proposed model is more robust to speech recognition errors."
                    },
                    {
                        "title": "BER: Balanced Error Rate For Speaker Diarization",
                        "abstract": "DER is the primary metric to evaluate diarization performance while facing a dilemma: the errors in short utterances or segments tend to be overwhelmed by longer ones. Short segments, e.g., `yes' or `no,' still have semantic information. Besides, DER overlooks errors in less-talked speakers. Although JER balances speaker errors, it still suffers from the same dilemma. Considering all those aspects, duration error, segment error, and speaker-weighted error constituting a complete diarization evaluation, we propose a Balanced Error Rate (BER) to evaluate speaker diarization. First, we propose a segment-level error rate (SER) via connected sub-graphs and adaptive IoU threshold to get accurate segment matching. Second, to evaluate diarization in a unified way, we adopt a speaker-specific harmonic mean between duration and segment, followed by a speaker-weighted average. Third, we analyze our metric via the modularized system, EEND, and the multi-modal method on real datasets. SER and BER are publicly available at https://github.com/X-LANCE/BER."
                    },
                    {
                        "title": "Phone-Level Prosody Modelling with GMM-Based MDN for Diverse and Controllable Speech Synthesis",
                        "abstract": "Generating natural speech with a diverse and smooth prosody pattern is a challenging task. Although random sampling with phone-level prosody distribution has been investigated to generate different prosody patterns, the diversity of the generated speech is still very limited and far from what can be achieved by humans. This is largely due to the use of uni-modal distribution, such as single Gaussian, in the prior works of phone-level prosody modelling. In this work, we propose a novel approach that models phone-level prosodies with a GMM-based mixture density network(MDN) and then extend it for multi-speaker TTS using speaker adaptation transforms of Gaussian means and variances. Furthermore, we show that we can clone the prosodies from a reference speech by sampling prosodies from the Gaussian components that produce the reference prosodies. Our experiments on LJSpeech and LibriTTS dataset show that the proposed method with GMM-based MDN not only achieves significantly better diversity than using a single Gaussian in both single-speaker and multi-speaker TTS, but also provides better naturalness. The prosody cloning experiments demonstrate that the prosody similarity of the proposed method with GMM-based MDN is comparable to recent proposed fine-grained VAE while the target speaker similarity is better."
                    },
                    {
                        "title": "On Training Bi-directional Neural Network Language Model with Noise Contrastive Estimation",
                        "abstract": "We propose to train bi-directional neural network language model(NNLM) with noise contrastive estimation(NCE). Experiments are conducted on a rescore task on the PTB data set. It is shown that NCE-trained bi-directional NNLM outperformed the one trained by conventional maximum likelihood training. But still(regretfully), it did not out-perform the baseline uni-directional NNLM."
                    },
                    {
                        "title": "Semi-supervised Learning for Code-Switching ASR with Large Language Model Filter",
                        "abstract": "Code-switching (CS) phenomenon occurs when words or phrases from different languages are alternated in a single sentence. Due to data scarcity, building an effective CS Automatic Speech Recognition (ASR) system remains challenging. In this paper, we propose to enhance CS-ASR systems by utilizing rich unsupervised monolingual speech data within a semi-supervised learning framework, particularly when access to CS data is limited. To achieve this, we establish a general paradigm for applying noisy student training (NST) to the CS-ASR task. Specifically, we introduce the LLM-Filter, which leverages well-designed prompt templates to activate the correction capability of large language models (LLMs) for monolingual data selection and pseudo-labels refinement during NST. Our experiments on the supervised ASRU-CS and unsupervised AISHELL-2 and LibriSpeech datasets show that our method not only achieves significant improvements over supervised and semi-supervised learning baselines for the CS task, but also attains better performance compared with the fully-supervised oracle upper-bound on the CS English part. Additionally, we further investigate the influence of accent on AESRC dataset and demonstrate that our method can get achieve additional benefits when the monolingual data contains relevant linguistic characteristic."
                    }
                ]
            },
            "1de1c52b-5dc6-46c9-bd5b-f77213399b2e": {
                "pk": "1de1c52b-5dc6-46c9-bd5b-f77213399b2e",
                "name": "Tao Yu",
                "collaborators": [
                    "Bowen Zeng",
                    "Haitao Cao",
                    "J. W. Rao",
                    "Junlin Xiong",
                    "Chengyuan Cai",
                    "Stavros Garoufalidis",
                    "Shafiq Joty"
                ],
                "domain": [
                    "Mathematics",
                    "Quantum Physics",
                    "Control Systems",
                    "Non-Hermitian Systems"
                ],
                "publications": [
                    {
                        "title": "Left-invariant Ricci collineations associated to Yano connections on three-dimensional Lorentzian Lie groups",
                        "abstract": "In this paper, we classify Left-invariant Ricci collineations associated to Yano connections on three-dimensional Lorentzian Lie groups."
                    },
                    {
                        "title": "Left-invariant Ricci collineations associated to canonical connections on three-dimensional Lorentzian Lie groups",
                        "abstract": "In this paper, we classify Left-invariant Ricci collineations associated to canonical connections and Kobayashi-Nomizu connections on three-dimensional Lorentzian Lie groups."
                    },
                    {
                        "title": "Center of the stated skein algebra",
                        "abstract": "The stated skein algebra is a generalization of the Kauffman bracket skein algebra introduced in the study of quantum trace maps. When the quantum parameter is a root of unity, the stated skein algebra has a big center and is finitely generated as a module over the center. We give the center a simple description and calculate the dimension over center of the stated skein algebra."
                    },
                    {
                        "title": "Explicit representations and Azumaya loci of skein algebras of small surfaces",
                        "abstract": "We construct finite dimensional representations of the Kauffman bracket skein algebra of the one-punctured torus and four-punctured sphere at all roots of unity. The representations are given by explicit formulas. They all have dimensions equal to the PI degrees of the skein algebras, and they realize all classical shadows. We then use the reducibility of these representations to determine the Azumaya loci. In particular, the Azumaya loci of these surfaces contain the smooth loci of the classical shadow varieties, with equality in the case of the one-punctured torus and proper containment in the case of the four-punctured sphere."
                    },
                    {
                        "title": "Measure-theoretic mean equicontinuity and bounded complexity",
                        "abstract": "Let $(X,\\mathcal{B},\\mu,T)$ be a measure preserving system. We say that a function $f\\in L^2(X,\\mu)$ is $\\mu$-mean equicontinuous if for any $\\epsilon>0$ there is $k\\in \\mathbb{N}$ and measurable sets ${A_1,A_2,\\cdots,A_k}$ with $\\mu\\left(\\bigcup\\limits_{i=1}^k A_i\\right)>1-\\epsilon$ such that whenever $x,y\\in A_i$ for some $1\\leq i\\leq k$, one has \\[ \\limsup_{n\\to\\infty}\\frac{1}{n}\\sum_{j=0}^{n-1}|f(T^jx)-f(T^jy)|<\\epsilon. \\] Measure complexity with respect to $f$ is also introduced. It is shown that $f$ is an almost periodic function if and only if $f$ is $\\mu$-mean equicontinuous if and only if $\\mu$ has bounded complexity with respect to $f$.   Ferenczi studied measure-theoretic complexity using $\\alpha$-names of a partition and the Hamming distance. He proved that if a measure preserving system is ergodic, then the complexity function is bounded if and only if the system has discrete spectrum. We show that this result holds without the assumption of ergodicity."
                    },
                    {
                        "title": "Nonreciprocal surface magnetoelastic dynamics",
                        "abstract": "Motivated by recent experiments, we investigate the nonreciprocal magnetoelastic interaction between the surface acoustic phonons of dielectric non-magnetic substrates and magnons of proximity nanomagnets. The magnetization dynamics exerts rotating forces at the edges of the nanomagnet that causes the nonreciprocal interaction with surface phonons due to its rotation-momentum locking. This coupling induces the nonreciprocity of the surface phonon transmission and a nearly complete phonon diode effect by several (tens of) magnetic nanowires of high (ordinary) magnetic quality. Phase-sensitive microwave transmission is also nonreciprocal that can pick up clear signals of the coherent phonons excited by magnetization dynamics. Nonreciprocal pumping of phonons by precessing magnetization is predicted using Landauer-B\\\"uttiker formalism."
                    },
                    {
                        "title": "The Heterogeneous Multiscale Finite Volume Method for Convection-Diffusion-Reaction Problem",
                        "abstract": "In this paper, we employ an finite volume method (FVM) based on the heterogenous multiscale method (HMM), for the multiscale convection-diffusion-reaction problem. The optimal order convergence rate in H^1-norm is given for periodic medias."
                    },
                    {
                        "title": "Perspective on non-Hermitian physics in magnetic systems",
                        "abstract": "A perspective on non-Hermitian physics in magnetic systems is addressed in this short article, including exceptional points, exceptional nodal phases, the non-Hermitian SSH model, and the non-Hermitian skin effect."
                    },
                    {
                        "title": "Distributed Networked Controller Design for Large-scale Systems under Round-Robin Communication Protocol",
                        "abstract": "This paper studies the distributed L2-gain control problem for continuous-time large-scale systems under Round-Robin communication protocol. In this protocol, each sub-controller obtains its own subsystem's state information continuously, while communicating with neighbors at discrete-time instants periodically. Distributed controllers are designed such that the closed-loop system is exponentially stable and that the prescribed L2-gain is satisfied. The design condition is obtained based on a time-delay approach and given in terms of linear matrix inequalities. Finally, three numerical examples are presented to illustrate the efficiency of the proposed scheme."
                    },
                    {
                        "title": "Spin Radiation of Electrons, Excitons, and Phonons",
                        "abstract": "In the celebrated Stern-Gerlach experiment an inhomogeneous static magnetic field separates a beam of charge-neutral atoms with opposite spins, thereby driving a ``spin current\" normal to the propagation direction. Here we generalize it to the dynamic scenario by demonstrating a spin transfer between an AC inhomogeneous magnetic field and intraband electrons or charge-neutral excitons and phonons. We predict that parametric pumping can efficiently radiate their DC spin currents from local AC magnetic sources with van der Waals semiconductors as prototypes. This mechanism brings a unified and efficient paradigm in the spin transport of distinct mobile carriers."
                    },
                    {
                        "title": "A Quantum trace map for 3-manifolds",
                        "abstract": "We define a quantum trace map from the skein module of a 3-manifold with torus boundary components to a module (left and right quotient of a quantum torus) constructed from an ideal triangulation. Our map is a 3-dimensional version of the well-known quantum trace map on surfaces introduced by Bonahon and Wong and further developed by L\\^e."
                    },
                    {
                        "title": "Online Conversation Disentanglement with Pointer Networks",
                        "abstract": "Huge amounts of textual conversations occur online every day, where multiple conversations take place concurrently. Interleaved conversations lead to difficulties in not only following the ongoing discussions but also extracting relevant information from simultaneous messages. Conversation disentanglement aims to separate intermingled messages into detached conversations. However, existing disentanglement methods rely mostly on handcrafted features that are dataset specific, which hinders generalization and adaptability. In this work, we propose an end-to-end online framework for conversation disentanglement that avoids time-consuming domain-specific feature engineering. We design a novel way to embed the whole utterance that comprises timestamp, speaker, and message text, and proposes a custom attention mechanism that models disentanglement as a pointing problem while effectively capturing inter-utterance interactions in an end-to-end fashion. We also introduce a joint-learning objective to better capture contextual information. Our experiments on the Ubuntu IRC dataset show that our method achieves state-of-the-art performance in both link and conversation prediction tasks."
                    },
                    {
                        "title": "Giant Microwave Sensitivity of Magnetic Array by Long-Range Chiral Interaction Driven Skin Effect",
                        "abstract": "Non-Hermitian skin effect was observed in one-dimensional systems with short-range chiral interaction. Long-range chiral interaction mediated by traveling waves also favors the accumulation of energy, but has not yet showed non-Hermitian topology. Here we find that the strong interference brought by the wave propagation is detrimental for accumulation. By suppression of interference via the damping of traveling waves, we predict the non-Hermitian skin effect of magnetic excitation in a periodic array of magnetic nanowires that are coupled chirally via spin waves of thin magnetic films. The local excitation of a wire at one edge by weak microwaves of magnitude $\\sim \\mu{\\rm T}$ leads to a considerable spin-wave amplitude at the other edge, i.e. a remarkable functionality useful for sensitive, non-local, and non-reciprocal detection of microwaves."
                    },
                    {
                        "title": "Radiation-free and non-Hermitian topology inertial defect states of on-chip magnons",
                        "abstract": "Radiative damping is a strong dissipation source for the quantum emitters hybridized with propagating photons, electrons, or phonons, which is not easily avoidable for on-chip magnonic emitters as well that can radiate via the surface acoustic waves of the substrate. Here we demonstrate in an array of on-chip nano-magnets coupled in a long range via exchanging the surface acoustic waves that a point defect in the array, which can be introduced by the local magnon frequency shift by a local biased magnetic field or the absence of a magnetic wire, strongly localizes the magnons, in contrast to the well spreading Bloch-like collective magnon modes in such an array setting. The radiation of the magnon defect states is exponentially suppressed by the distance of the defect to the array edges. Moreover, this defect state is strikingly inertial to the non-Hermitian topology that localizes all the extended states at one boundary. Such configuration robust magnon defect states towards radiation-free limit may be suitable for high-fidelity magnon quantum information storage in the future on-chip magnonic devices."
                    }
                ]
            }
        }
    },
    "2401.11374": {
        "paper_data": {
            "title": "Language Models as Hierarchy Encoders",
            "url": "http://arxiv.org/abs/2401.11374v3",
            "arxiv_id": "2401.11374",
            "authors": [
                "Yuan He",
                "Zhangdie Yuan",
                "Jiaoyan Chen",
                "Ian Horrocks"
            ],
            "abstract": "Interpreting hierarchical structures latent in language is a key limitation of current language models (LMs). While previous research has implicitly leveraged these hierarchies to enhance LMs, approaches for their explicit encoding are yet to be explored. To address this, we introduce a novel approach to re-train transformer encoder-based LMs as Hierarchy Transformer encoders (HiTs), harnessing the expansive nature of hyperbolic space. Our method situates the output embedding space of pre-trained LMs within a Poincar\\'e ball with a curvature that adapts to the embedding dimension, followed by training on hyperbolic clustering and centripetal losses. These losses are designed to effectively cluster related entities (input as texts) and organise them hierarchically. We evaluate HiTs against pre-trained LMs, standard fine-tuned LMs, and several hyperbolic embedding baselines, focusing on their capabilities in simulating transitive inference, predicting subsumptions, and transferring knowledge across hierarchies. The results demonstrate that HiTs consistently outperform all baselines in these tasks, underscoring the effectiveness and transferability of our re-trained hierarchy encoders.",
            "introduction": "   1 Introduction  In the field of Natural Language Processing (NLP) and related areas, the emergence of transformer-based language models (LMs) such as BERT (encoder-based) [1], GPT (decoder-based) [2], and the more recent large language models (LLMs) like GPT-4 [3] and Llama 2 [4], has marked a significant progression. Nonetheless, these models face a notable challenge in effectively encoding and interpreting hierarchical structures latent in language. This limitation has been highlighted by several studies, including those by [5] and [6], which employed prompt-based probes to reveal the limited hierarchical knowledge in pre-trained LMs, and the work by [7], which demonstrated these models\u2019 struggles with capturing the transitivity of hierarchical relationships.   Prior research has explored various methods to infuse hierarchical information into LM training. Common approaches include classification-based fine-tuning using sentence head embedding with a classification layer [8] or few-shot prompting with an answer mapping to classification labels [6]. To further pre-train, or re-train222In this work, the term re-train refers to train LMs on a new corpus without modifying its architecture; it is distinguished from standard fine-tuning that involves adding task-specific layers which lead to additional learnable parameters. LMs on a corpus constructed from hierarchical data, [9] converted structural representations into textual formats to align the masked language modeling objective. Others, like [10] and [11], have focused on extracting analogous and contrasting examples from hierarchical structures for a similarity-based contrastive learning objective. The aforementioned studies leveraged hierarchical information as implicit signals to augment LMs, yet no existing works specifically targeted the explicit encoding of hierarchies with LMs.   Figure 1: Illustration of how hierarchies are explicitly encoded in HiTs. The square (d\ud835\udc51ditalic_d-dimensional hyper-cube) refers to the output embedding space of transformer encoder-based LMs whose final activation function is typically tanh\\tanhroman_tanh, and the circumscribed circle (d\ud835\udc51ditalic_d-dimensional hyper-sphere) refers to the Poincar\u00e9 ball of radius d\ud835\udc51\\sqrt{d}square-root start_ARG italic_d end_ARG. The distance and norm metrics involved in our hyperbolic losses are defined w.r.t.\u00a0this manifold.    To bridge this gap, we introduce a novel approach to re-train transformer encoder-based LMs as Hierarchy Transformer encoders (HiTs). Inspired by the efficacy of hyperbolic geometry in representing hierarchical structures [12, 13], we propose the hyperbolic clustering and centripetal losses tailored for LM re-training. As illustrated in Figure 1, transformer encoder-based LMs typically use a tanh\\tanhroman_tanh activation function in the last layer, which maps each embedding dimension to the range [\u22121,1]11[-1,1][ - 1 , 1 ]. Consequently, the output embeddings of LMs are confined within a unit d\ud835\udc51ditalic_d-dimensional hyper-cube. Leveraging this characteristic, we utilise a Poincar\u00e9 ball of radius d\ud835\udc51\\sqrt{d}square-root start_ARG italic_d end_ARG, whose boundary circumscribes333We ignore the vertices of the hyper-cube, as they are on the boundary and thus are undefined in the open Poincar\u00e9 ball. the output embedding space of LMs. The metrics for distance and norm used in our hyperbolic losses are defined w.r.t.\u00a0this specific manifold. After re-training, entities are not only clustered according to their relatedness but also hierarchically organised.   In evaluating HiTs, we compare their performance against pre-trained LMs, standard fine-tuned LMs, and previous hyperbolic embedding models in the Multi-hop Inference and Mixed-hop Prediction tasks. The Multi-hop Inference task, following the setting in [13], involves",
            "references": [
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "DeepOnto: A Python Package for Ontology Engineering with Deep Learning",
                    "abstract": "Integrating deep learning techniques, particularly language models (LMs), with knowledge representation techniques like ontologies has raised widespread attention, urging the need of a platform that supports both paradigms. Although packages such as OWL API and Jena offer robust support for basic ontology processing features, they lack the capability to transform various types of information within ontologies into formats suitable for downstream deep learning-based applications. Moreover, widely-used ontology APIs are primarily Java-based while deep learning frameworks like PyTorch and Tensorflow are mainly for Python programming. To address the needs, we present DeepOnto, a Python package designed for ontology engineering with deep learning. The package encompasses a core ontology processing module founded on the widely-recognised and reliable OWL API, encapsulating its fundamental features in a more \u201cPythonic\u201d manner and extending its capabilities to incorporate other essential components including reasoning, verbalisation, normalisation, taxonomy, projection, and more. Building on this module, DeepOnto offers a suite of tools, resources, and algorithms that support various ontology engineering tasks, such as ontology alignment and completion, by harnessing deep learning methods, primarily pre-trained LMs. In this paper, we also demonstrate the practical utility of DeepOnto through two use-cases: the Digital Health Coaching in Samsung Research UK and the Bio-ML track of the Ontology Alignment Evaluation Initiative (OAEI)."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Language Model Analysis for Ontology Subsumption Inference",
                    "abstract": "Investigating whether pre-trained language models (LMs) can function as knowledge bases (KBs) has raised wide research interests recently. However, existing works focus on simple, triple-based, relational KBs, but omit more sophisticated, logic-based, conceptualised KBs such as OWL ontologies. To investigate an LM's knowledge of ontologies, we propose OntoLAMA, a set of inference-based probing tasks and datasets from ontology subsumption axioms involving both atomic and complex concepts. We conduct extensive experiments on ontologies of different domains and scales, and our results demonstrate that LMs encode relatively less background knowledge of Subsumption Inference (SI) than traditional Natural Language Inference (NLI) but can improve on SI significantly when a small number of samples are given. We will open-source our code and datasets."
                },
                {
                    "title": "SimCSE: Simple Contrastive Learning of Sentence Embeddings",
                    "abstract": "This paper presents SimCSE, a simple contrastive learning framework that greatly advances the state-of-the-art sentence embeddings. We first describe an unsupervised approach, which takes an input sentence and predicts itself in a contrastive objective, with only standard dropout used as noise. This simple method works surprisingly well, performing on par with previous supervised counterparts. We find that dropout acts as minimal data augmentation and removing it leads to a representation collapse. Then, we propose a supervised approach, which incorporates annotated pairs from natural language inference datasets into our contrastive learning framework, by using \u201centailment\u201d pairs as positives and \u201ccontradiction\u201d pairs as hard negatives. We evaluate SimCSE on standard semantic textual similarity (STS) tasks, and our unsupervised and supervised models using BERT base achieve an average of 76.3% and 81.6% Spearman\u2019s correlation respectively, a 4.2% and 2.2% improvement compared to previous best results. We also show\u2014both theoretically and empirically\u2014that contrastive learning objective regularizes pre-trained embeddings\u2019 anisotropic space to be more uniform, and it better aligns positive pairs when supervised signals are available."
                },
                {
                    "title": "Probing BERT in Hyperbolic Spaces",
                    "abstract": "Recently, a variety of probing tasks are proposed to discover linguistic properties learned in contextualized word embeddings. Many of these works implicitly assume these embeddings lay in certain metric spaces, typically the Euclidean space. This work considers a family of geometrically special spaces, the hyperbolic spaces, that exhibit better inductive biases for hierarchical structures and may better reveal linguistic hierarchies encoded in contextualized representations. We introduce a Poincare probe, a structural probe projecting these embeddings into a Poincare subspace with explicitly defined hierarchies. We focus on two probing objectives: (a) dependency trees where the hierarchy is defined as head-dependent structures; (b) lexical sentiments where the hierarchy is defined as the polarity of words (positivity and negativity). We argue that a key desideratum of a probe is its sensitivity to the existence of linguistic structures. We apply our probes on BERT, a typical contextualized embedding model. In a syntactic subspace, our probe better recovers tree structures than Euclidean probes, revealing the possibility that the geometry of BERT syntax may not necessarily be Euclidean. In a sentiment subspace, we reveal two possible meta-embeddings for positive and negative sentiments and show how lexically-controlled contextualization would change the geometric localization of embeddings. We demonstrate the findings with our Poincare probe via extensive experiments and visualization. Our results can be reproduced at https://github.com/FranxYao/PoincareProbe."
                },
                {
                    "title": "Self-Alignment Pretraining for Biomedical Entity Representations",
                    "abstract": "Despite the widespread success of self-supervised learning via masked language models (MLM), accurately capturing fine-grained semantic relationships in the biomedical domain remains a challenge. This is of paramount importance for entity-level tasks such as entity linking where the ability to model entity relations (especially synonymy) is pivotal. To address this challenge, we propose SapBERT, a pretraining scheme that self-aligns the representation space of biomedical entities. We design a scalable metric learning framework that can leverage UMLS, a massive collection of biomedical ontologies with 4M+ concepts. In contrast with previous pipeline-based hybrid systems, SapBERT offers an elegant one-model-for-all solution to the problem of medical entity linking (MEL), achieving a new state-of-the-art (SOTA) on six MEL benchmarking datasets. In the scientific domain, we achieve SOTA even without task-specific supervision. With substantial improvement over various domain-specific pretrained MLMs such as BioBERT, SciBERTand and PubMedBERT, our pretraining scheme proves to be both effective and robust."
                },
                {
                    "title": "Geoopt: Riemannian Optimization in PyTorch",
                    "abstract": "Geoopt is a research-oriented modular open-source package for Riemannian Optimization in PyTorch. The core of Geoopt is a standard Manifold interface that allows for the generic implementation of optimization algorithms. Geoopt supports basic Riemannian SGD as well as adaptive optimization algorithms. Geoopt also provides several algorithms and arithmetic methods for supported manifolds, which allow composing geometry-aware neural network layers that can be integrated with existing models."
                },
                {
                    "title": "MPNet: Masked and Permuted Pre-training for Language Understanding",
                    "abstract": "BERT adopts masked language modeling (MLM) for pre-training and is one of the most successful pre-training models. Since BERT neglects dependency among predicted tokens, XLNet introduces permuted language modeling (PLM) for pre-training to address this problem. We argue that XLNet does not leverage the full position information of a sentence and thus suffers from position discrepancy between pre-training and fine-tuning. In this paper, we propose MPNet, a novel pre-training method that inherits the advantages of BERT and XLNet and avoids their limitations. MPNet leverages the dependency among predicted tokens through permuted language modeling (vs. MLM in BERT), and takes auxiliary position information as input to make the model see a full sentence and thus reducing the position discrepancy (vs. PLM in XLNet). We pre-train MPNet on a large-scale dataset (over 160GB text corpora) and fine-tune on a variety of down-streaming tasks (GLUE, SQuAD, etc). Experimental results show that MPNet outperforms MLM and PLM by a large margin, and achieves better results on these tasks compared with previous state-of-the-art pre-trained methods (e.g., BERT, XLNet, RoBERTa) under the same model setting. We release the code and pre-trained model in GitHub\\footnote{\\url{this https URL}}."
                },
                {
                    "title": "MiniLM: Deep Self-Attention Distillation for Task-Agnostic Compression of Pre-Trained Transformers",
                    "abstract": "Pre-trained language models (e.g., BERT (Devlin et al., 2018) and its variants) have achieved remarkable success in varieties of NLP tasks. However, these models usually consist of hundreds of millions of parameters which brings challenges for fine-tuning and online serving in real-life applications due to latency and capacity constraints. In this work, we present a simple and effective approach to compress large Transformer (Vaswani et al., 2017) based pre-trained models, termed as deep self-attention distillation. The small model (student) is trained by deeply mimicking the self-attention module, which plays a vital role in Transformer networks, of the large model (teacher). Specifically, we propose distilling the self-attention module of the last Transformer layer of the teacher, which is effective and flexible for the student. Furthermore, we introduce the scaled dot-product between values in the self-attention module as the new deep self-attention knowledge, in addition to the attention distributions (i.e., the scaled dot-product of queries and keys) that have been used in existing works. Moreover, we show that introducing a teacher assistant (Mirzadeh et al., 2019) also helps the distillation of large pre-trained Transformer models. Experimental results demonstrate that our monolingual model outperforms state-of-the-art baselines in different parameter size of student models. In particular, it retains more than 99% accuracy on SQuAD 2.0 and several GLUE benchmark tasks using 50% of the Transformer parameters and computations of the teacher model. We also obtain competitive results in applying deep self-attention distillation to multilingual pre-trained models."
                },
                {
                    "title": "Masked Language Model Scoring",
                    "abstract": "Pretrained masked language models (MLMs) require finetuning for most NLP tasks. Instead, we evaluate MLMs out of the box via their pseudo-log-likelihood scores (PLLs), which are computed by masking tokens one by one. We show that PLLs outperform scores from autoregressive language models like GPT-2 in a variety of tasks. By rescoring ASR and NMT hypotheses, RoBERTa reduces an end-to-end LibriSpeech model\u2019s WER by 30% relative and adds up to +1.7 BLEU on state-of-the-art baselines for low-resource translation pairs, with further gains from domain adaptation. We attribute this success to PLL\u2019s unsupervised expression of linguistic acceptability without a left-to-right bias, greatly improving on scores from GPT-2 (+10 points on island effects, NPI licensing in BLiMP). One can finetune MLMs to give scores without masking, enabling computation in a single inference pass. In all, PLLs and their associated pseudo-perplexities (PPPLs) enable plug-and-play use of the growing number of pretrained MLMs; e.g., we use a single cross-lingual model to rescore translations in multiple languages. We release our library for language model scoring at https://github.com/awslabs/mlm-scoring."
                },
                {
                    "title": "Language Models as Knowledge Bases?",
                    "abstract": "Recent progress in pretraining language models on large textual corpora led to a surge of improvements for downstream NLP tasks. Whilst learning linguistic knowledge, these models may also be storing relational knowledge present in the training data, and may be able to answer queries structured as \u201cfill-in-the-blank\u201d cloze statements. Language models have many advantages over structured knowledge bases: they require no schema engineering, allow practitioners to query about an open class of relations, are easy to extend to more data, and require no human supervision to train. We present an in-depth analysis of the relational knowledge already present (without fine-tuning) in a wide range of state-of-the-art pretrained language models. We find that (i) without fine-tuning, BERT contains relational knowledge competitive with traditional NLP methods that have some access to oracle knowledge, (ii) BERT also does remarkably well on open-domain question answering against a supervised baseline, and (iii) certain types of factual knowledge are learned much more readily than others by standard language model pretraining approaches. The surprisingly strong ability of these models to recall factual knowledge without any fine-tuning demonstrates their potential as unsupervised open-domain QA systems. The code to reproduce our analysis is available at https://github.com/facebookresearch/LAMA."
                },
                {
                    "title": "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
                    "abstract": "BERT (Devlin et al., 2018) and RoBERTa (Liu et al., 2019) has set a new state-of-the-art performance on sentence-pair regression tasks like semantic textual similarity (STS). However, it requires that both sentences are fed into the network, which causes a massive computational overhead: Finding the most similar pair in a collection of 10,000 sentences requires about 50 million inference computations (~65 hours) with BERT. The construction of BERT makes it unsuitable for semantic similarity search as well as for unsupervised tasks like clustering. In this publication, we present Sentence-BERT (SBERT), a modification of the pretrained BERT network that use siamese and triplet network structures to derive semantically meaningful sentence embeddings that can be compared using cosine-similarity. This reduces the effort for finding the most similar pair from 65 hours with BERT / RoBERTa to about 5 seconds with SBERT, while maintaining the accuracy from BERT. We evaluate SBERT and SRoBERTa on common STS tasks and transfer learning tasks, where it outperforms other state-of-the-art sentence embeddings methods."
                },
                {
                    "title": "RoBERTa: A Robustly Optimized BERT Pretraining Approach",
                    "abstract": "Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code."
                },
                {
                    "title": "Poincar\u00e9 GloVe: Hyperbolic Word Embeddings",
                    "abstract": "Words are not created equal. In fact, they form an aristocratic graph with a latent hierarchical structure that the next generation of unsupervised learned word embeddings should reveal. In this paper, justified by the notion of delta-hyperbolicity or tree-likeliness of a space, we propose to embed words in a Cartesian product of hyperbolic spaces which we theoretically connect to the Gaussian word embeddings and their Fisher geometry. This connection allows us to introduce a novel principled hypernymy score for word embeddings. Moreover, we adapt the well-known Glove algorithm to learn unsupervised word embeddings in this type of Riemannian manifolds. We further explain how to solve the analogy task using the Riemannian parallel transport that generalizes vector arithmetics to this new type of geometry. Empirically, based on extensive experiments, we prove that our embeddings, trained unsupervised, are the first to simultaneously outperform strong and popular baselines on the tasks of similarity, analogy and hypernymy detection. In particular, for word hypernymy, we obtain new state-of-the-art on fully unsupervised WBLESS classification accuracy."
                },
                {
                    "title": "Riemannian Adaptive Optimization Methods",
                    "abstract": "Several first order stochastic optimization methods commonly used in the Euclidean domain such as stochastic gradient descent (SGD), accelerated gradient descent or variance reduced methods have already been adapted to certain Riemannian settings. However, some of the most popular of these optimization tools - namely Adam , Adagrad and the more recent Amsgrad - remain to be generalized to Riemannian manifolds. We discuss the difficulty of generalizing such adaptive schemes to the most agnostic Riemannian setting, and then provide algorithms and convergence proofs for geodesically convex objectives in the particular case of a product of Riemannian manifolds, in which adaptivity is implemented across manifolds in the cartesian product. Our generalization is tight in the sense that choosing the Euclidean space as Riemannian manifold yields the same algorithms and regret bounds as those that were already known for the standard algorithms. Experimentally, we show faster convergence and to a lower train loss value for Riemannian adaptive methods over their corresponding baselines on the realistic task of embedding the WordNet taxonomy in the Poincare ball."
                },
                {
                    "title": "Embedding Text in Hyperbolic Spaces",
                    "abstract": "Natural language text exhibits hierarchical structure in a variety of respects. Ideally, we could incorporate our prior knowledge of this hierarchical structure into unsupervised learning algorithms that work on text data. Recent work by Nickel and Kiela (2017) proposed using hyperbolic instead of Euclidean embedding spaces to represent hierarchical data and demonstrated encouraging results when embedding graphs. In this work, we extend their method with a re-parameterization technique that allows us to learn hyperbolic embeddings of arbitrarily parameterized objects. We apply this framework to learn word and sentence embeddings in hyperbolic space in an unsupervised manner from text corpora. The resulting embeddings seem to encode certain intuitive notions of hierarchy, such as word-context frequency and phrase constituency. However, the implicit continuous hierarchy in the learned hyperbolic space makes interrogating the model\u2019s learned hierarchies more difficult than for models that learn explicit edges between items. The learned hyperbolic embeddings show improvements over Euclidean embeddings in some \u2013 but not all \u2013 downstream tasks, suggesting that hierarchical organization is more useful for some tasks than others."
                },
                {
                    "title": "Hyperbolic Neural Networks",
                    "abstract": "Hyperbolic spaces have recently gained momentum in the context of machine learning due to their high capacity and tree-likeliness properties. However, the representational power of hyperbolic geometry is not yet on par with Euclidean geometry, mostly because of the absence of corresponding hyperbolic neural network layers. This makes it hard to use hyperbolic embeddings in downstream tasks. Here, we bridge this gap in a principled manner by combining the formalism of Mobius gyrovector spaces with the Riemannian geometry of the Poincare model of hyperbolic spaces. As a result, we derive hyperbolic versions of important deep learning tools: multinomial logistic regression, feed-forward and recurrent neural networks such as gated recurrent units. This allows to embed sequential data and perform classification in the hyperbolic space. Empirically, we show that, even if hyperbolic optimization tools are limited, hyperbolic sentence embeddings either outperform or are on par with their Euclidean variants on textual entailment and noisy-prefix recognition tasks."
                },
                {
                    "title": "Hyperbolic Entailment Cones for Learning Hierarchical Embeddings",
                    "abstract": "Learning graph representations via low-dimensional embeddings that preserve relevant network properties is an important class of problems in machine learning. We here present a novel method to embed directed acyclic graphs. Following prior work, we first advocate for using hyperbolic spaces which provably model tree-like structures better than Euclidean geometry. Second, we view hierarchical relations as partial orders defined using a family of nested geodesically convex cones. We prove that these entailment cones admit an optimal shape with a closed form expression both in the Euclidean and hyperbolic spaces. Moreover, they canonically define the embedding learning process. Experiments show significant improvements of our method over strong recent baselines both in terms of representational capacity and generalization."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Poincar\u00e9 Embeddings for Learning Hierarchical Representations",
                    "abstract": "Representation learning has become an invaluable approach for learning from symbolic data such as text and graphs. However, while complex symbolic datasets often exhibit a latent hierarchical structure, state-of-the-art methods typically learn embeddings in Euclidean vector spaces, which do not account for this property. For this purpose, we introduce a new approach for learning hierarchical representations of symbolic data by embedding them into hyperbolic space -- or more precisely into an n-dimensional Poincare ball. Due to the underlying hyperbolic geometry, this allows us to learn parsimonious representations of symbolic data by simultaneously capturing hierarchy and similarity. We introduce an efficient algorithm to learn the embeddings based on Riemannian optimization and show experimentally that Poincare embeddings outperform Euclidean embeddings significantly on data with latent hierarchies, both in terms of representation capacity and in terms of generalization ability."
                },
                {
                    "title": "Schema.org: Evolution of Structured Data on the Web",
                    "abstract": "Separation between content and presentation has always been one of the important design aspects of the Web. Historically, however, even though most Web sites were driven off structured databases, they published their content purely in HTML. Services such as Web search, price comparison, reservation engines, etc. that operated on this content had access only to HTML. Applications requiring access to the structured data underlying these Web pages had to build custom extractors to convert plain HTML into structured data. These efforts were often laborious and the scrapers were fragile and error-prone, breaking every time a site changed its layout."
                },
                {
                    "title": "Embedding Entities and Relations for Learning and Inference in Knowledge Bases",
                    "abstract": "Abstract: We consider learning representations of entities and relations in KBs using the neural-embedding approach. We show that most existing models, including NTN (Socher et al., 2013) and TransE (Bordes et al., 2013b), can be generalized under a unified learning framework, where entities are low-dimensional vectors learned from a neural network and relations are bilinear and/or linear mapping functions. Under this framework, we compare a variety of embedding models on the link prediction task. We show that a simple bilinear formulation achieves new state-of-the-art results for the task (achieving a top-10 accuracy of 73.2% vs. 54.7% by TransE on Freebase). Furthermore, we introduce a novel approach that utilizes the learned relation embeddings to mine logical rules such as \"BornInCity(a,b) and CityInCountry(b,c) => Nationality(a,c)\". We find that embeddings learned from the bilinear objective are particularly good at capturing relational semantics and that the composition of relations is characterized by matrix multiplication. More interestingly, we demonstrate that our embedding-based rule extraction approach successfully outperforms a state-of-the-art confidence-based rule mining approach in mining Horn rules that involve compositional reasoning."
                },
                {
                    "title": "Translating Embeddings for Modeling Multi-relational Data",
                    "abstract": "We consider the problem of embedding entities and relationships of multi-relational data in low-dimensional vector spaces. Our objective is to propose a canonical model which is easy to train, contains a reduced number of parameters and can scale up to very large databases. Hence, we propose TransE, a method which models relationships by interpreting them as translations operating on the low-dimensional embeddings of the entities. Despite its simplicity, this assumption proves to be powerful since extensive experiments show that TransE significantly outperforms state-of-the-art methods in link prediction on two knowledge bases. Besides, it can be successfully trained on a large scale data set with 1M entities, 25k relationships and more than 17M training samples."
                },
                {
                    "title": "Disease Ontology: a backbone for disease semantic integration",
                    "abstract": "The Disease Ontology (DO) database (http://disease-ontology.org) represents a comprehensive knowledge base of 8043 inherited, developmental and acquired human diseases (DO version 3, revision 2510). The DO web browser has been designed for speed, efficiency and robustness through the use of a graph database. Full-text contextual searching functionality using Lucene allows the querying of name, synonym, definition, DOID and cross-reference (xrefs) with complex Boolean search strings. The DO semantically integrates disease and medical vocabularies through extensive cross mapping and integration of MeSH, ICD, NCI's thesaurus, SNOMED CT and OMIM disease-specific terms and identifiers. The DO is utilized for disease annotation by major biomedical databases (e.g. Array Express, NIF, IEDB), as a standard representation of human disease in biomedical ontologies (e.g. IDO, Cell line ontology, NIFSTD ontology, Experimental Factor Ontology, Influenza Ontology), and as an ontological cross mappings resource between DO, MeSH and OMIM (e.g. GeneWiki). The DO project (http://diseaseontology.sf.net) has been incorporated into open source tools (e.g. Gene Answers, FunDO) to connect gene and disease biomedical data through the lens of human disease. The next iteration of the DO web browser will integrate DO's extended relations and logical definition representation along with these biomedical resource cross-mappings."
                },
                {
                    "title": "WordNet: A Lexical Database for English",
                    "abstract": "Because meaningful sentences are composed of meaningful words, any system that hopes to process natural languages as people do must have information about words and their meanings. This information is traditionally provided through dictionaries, and machine-readable dictionaries are now widely available. But dictionary entries evolved for the convenience of human readers, not for machines. WordNet1 provides a more effective combination of traditional lexicographic information and modern computing. WordNet is an online lexical database designed for use under program control. English nouns, verbs, adjectives, and adverbs are organized into sets of synonyms, each representing a lexicalized concept. Semantic relations link the synonym sets [4]."
                },
                {
                    "title": "Does BERT Know that the IS-A Relation Is Transitive?",
                    "abstract": "The success of a natural language processing (NLP) system on a task does not amount to fully understanding the complexity of the task, typified by many deep learning models. One such question is: can a black-box model make logically consistent predictions for transitive relations? Recent studies suggest that pre-trained BERT can capture lexico-semantic clues from words in the context. However, to what extent BERT captures the transitive nature of some lexical relations is unclear. From a probing perspective, we examine WordNet word senses and the IS-A relation, which is a transitive relation. That is, for senses A, B, and C, A is-a B and B is-a C entail A is-a C. We aim to quantify how much BERT agrees with the transitive property of IS-A relations, via a minimalist probing setting. Our investigation reveals that BERT\u2019s predictions do not fully obey the transitivity property of the IS-A relation."
                },
                {
                    "title": "Analyzing BERT\u2019s Knowledge of Hypernymy via Prompting",
                    "abstract": "The high performance of large pretrained language models (LLMs) such as BERT on NLP tasks has prompted questions about BERT\u2019s linguistic capabilities, and how they differ from humans\u2019. In this paper, we approach this question by examining BERT\u2019s knowledge of lexical semantic relations. We focus on hypernymy, the \u201cis-a\u201d relation that relates a word to a superordinate category. We use a prompting methodology to simply ask BERT what the hypernym of a given word is. We find that, in a setting where all hypernyms are guessable via prompting, BERT knows hypernyms with up to 57% accuracy. Moreover, BERT with prompting outperforms other unsupervised models for hypernym discovery even in an unconstrained scenario. However, BERT\u2019s predictions and performance on a dataset containing uncommon hyponyms and hypernyms indicate that its knowledge of hypernymy is still limited."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Improving Language Understanding by Generative Pre-Training",
                    "abstract": "Natural language understanding comprises a wide range of diverse tasks such as textual entailment, question answering, semantic similarity assessment, and document classi\ufb01cation. Although large unlabeled text corpora are abundant, labeled data for learning these speci\ufb01c tasks is scarce, making it challenging for discriminatively trained models to perform adequately. We demonstrate that large gains on these tasks can be realized by generative pre-training of a language model on a diverse corpus of unlabeled text, followed by discriminative \ufb01ne-tuning on each speci\ufb01c task. In contrast to previous approaches, we make use of task-aware input transformations during \ufb01ne-tuning to achieve effective transfer while requiring minimal changes to the model architecture. We demonstrate the effectiveness of our approach on a wide range of benchmarks for natural language understanding. Our general task-agnostic model outperforms discriminatively trained models that use architectures speci\ufb01cally crafted for each task, signi\ufb01cantly improving upon the state of the art in 9 out of the 12 tasks studied. For instance, we achieve absolute improvements of 8.9% on commonsense reasoning (Stories Cloze Test), 5.7% on question answering (RACE), and 1.5% on textual entailment (MultiNLI)."
                },
                {
                    "title": "NLTK: The Natural Language Toolkit",
                    "abstract": "The Natural Language Toolkit is a suite of program modules, data sets and tutorials supporting research and teaching in computational linguistics and natural language processing. NLTK is written in Python and distributed under the GPL open source license. Over the past year the toolkit has been rewritten, simplifying many linguistic data structures and taking advantage of recent enhancements in the Python language. This paper reports on the simpli-\ufb01ed toolkit and explains how it is used in teaching NLP."
                },
                {
                    "title": "SNOMED clinical terms: overview of the development process and project status",
                    "abstract": "Two large health care reference terminologies, SNOMED RT and Clinical Terms Version 3 , are in the process of being merged to form a comprehensive new work referred to as SNOMED Clinical Terms. The College of American Pathologists and the United Kingdom s National Health Service have entered into a collaborative agreement to develop this new work. Both organizations have extensive terminology development and maintenance experience. This paper discusses the process and status of SNOMED CT development and how the resources and expertise of both organizations are being used to develop this new terminological resource. The preliminary results of the merger process, including mapping, the merger of upper levels of each hierarchy, and attribute harmonization are also discussed."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively encode and interpret hierarchical structures in transformer-based language models to improve their performance in natural language processing tasks?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the capabilities of language models in understanding complex relationships within language, which has significant implications for various applications such as information retrieval, question answering, and knowledge representation. By addressing the limitations of current models in capturing hierarchical information, this research could lead to more sophisticated NLP systems that better mimic human understanding, ultimately influencing future research directions and practical applications in AI.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent limitations of existing transformer architectures, which struggle to represent hierarchical relationships effectively. Naive approaches, such as simple fine-tuning or classification-based methods, may fail because they do not explicitly account for the geometric properties of hierarchical data. Technical obstacles include the need for specialized loss functions that can operate within the constraints of the model's output space, as well as the complexity of designing a training regime that effectively integrates hierarchical information without compromising the model's performance on other tasks.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on implicit methods of incorporating hierarchical information, such as using classification layers or few-shot prompting, rather than explicitly encoding hierarchies. Limitations in understanding the geometric representation of hierarchies and the lack of tailored training methodologies have hindered progress. Our approach differs by introducing hyperbolic geometry and specific loss functions designed for hierarchical representation, which directly addresses these gaps and offers a novel framework for re-training language models.\n\n**[Question 5] - What are the key components of my approach and results?**  \nWe propose a methodology that involves re-training transformer encoder-based language models as Hierarchy Transformer encoders (HiTs) using hyperbolic clustering and centripetal losses. The dataset will consist of hierarchical data representations, and we will evaluate the models using metrics from Multi-hop Inference and Mixed-hop Prediction tasks. The expected outcomes include improved performance in capturing hierarchical relationships, leading to better clustering of related entities and enhanced model interpretability in NLP tasks."
            }
        },
        "author_data": {
            "5b93de95-5752-4702-848f-18e3dfc21e88": {
                "pk": "5b93de95-5752-4702-848f-18e3dfc21e88",
                "name": "Yuan He",
                "collaborators": [
                    "Yuan Yuan He",
                    "Subhrakanti Dey"
                ],
                "domain": [
                    "Number Theory",
                    "Cognitive Radio",
                    "Power Allocation",
                    "Fourier Analysis"
                ],
                "publications": [
                    {
                        "title": "On the mean values of products of Dirichlet $L$-functions at positive integers",
                        "abstract": "In this paper, we study the mean value distributions of Dirichlet $L$-functions at positive integers. We give some explicit formulas for the mean values of products of two and three Dirichlet $L$-functions at positive integers weighted by Dirichlet characters that involve the Bernoulli functions and Jordan's totient functions. The results presented here are the generalizations of various known formulas."
                    },
                    {
                        "title": "Reciprocity formulas for certain generalized Hardy-Berndt sums",
                        "abstract": "In this paper, we establish some reciprocity formulas for certain generalized Hardy-Berndt sums by using the Fourier series technique and some properties of the periodic zeta function and the Lerch zeta function. It turns out that one of Hardy's reciprocity theorems is deduced as a special case."
                    },
                    {
                        "title": "Power Allocation for Outage Minimization in Cognitive Radio Networks with Limited Feedback",
                        "abstract": "We address an optimal transmit power allocation problem that minimizes the outage probability of a secondary user (SU) who is allowed to coexist with a primary user (PU) in a narrowband spectrum sharing cognitive radio network, under a long term average transmit power constraint at the secondary transmitter (SU-TX) and an average interference power constraint at the primary receiver (PU-RX), with quantized channel state information (CSI) (including both the channels from SU-TX to SU-RX, denoted as $g_1$ and the channel from SU-TX to PU-RX, denoted as $g_0$) at the SU-TX. The optimal quantization regions in the vector channel space is shown to have a 'stepwise' structure. With this structure, the above outage minimization problem can be explicitly formulated and solved by employing the Karush-Kuhn-Tucker (KKT) necessary optimality conditions to obtain a locally optimal quantized power codebook. A low-complexity near-optimal quantized power allocation algorithm is derived for the case of large number of feedback bits. An explicit expression for the asymptotic SU outage probability at high rate quantization (as the number of feedback bits goes to infinity) is also provided, and is shown to approximate the optimal outage behavior extremely well for large number of bits of feedback via numerical simulations. Numerical results also illustrate that with 6 bits of feedback, the derived algorithms provide SU outage performance very close to that with full CSI at the SU-TX."
                    },
                    {
                        "title": "Spectrum Sharing in Cognitive Radio with Quantized Channel Information",
                        "abstract": "We consider a wideband spectrum sharing system where a secondary user can share a number of orthogonal frequency bands where each band is licensed to an individual primary user. We address the problem of optimum secondary transmit power allocation for its ergodic capacity maximization subject to an average sum (across the bands) transmit power constraint and individual average interference constraints on the primary users. The major contribution of our work lies in considering quantized channel state information (CSI)(for the vector channel space consisting of all secondary-to-secondary and secondary-to-primary channels) at the secondary transmitter. It is assumed that a band manager or a cognitive radio service provider has access to the full CSI information from the secondary and primary receivers and designs (offline) an optimal power codebook based on the statistical information (channel distributions) of the channels and feeds back the index of the codebook to the secondary transmitter for every channel realization in real-time, via a delay-free noiseless limited feedback channel. A modified Generalized Lloyds-type algorithm (GLA) is designed for deriving the optimal power codebook. An approximate quantized power allocation (AQPA) algorithm is also presented, that performs very close to its GLA based counterpart for large number of feedback bits and is significantly faster. We also present an extension of the modified GLA based quantized power codebook design algorithm for the case when the feedback channel is noisy. Numerical studies illustrate that with only 3-4 bits of feedback, the modified GLA based algorithms provide secondary ergodic capacity very close to that achieved by full CSI and with only as little as 4 bits of feedback, AQPA provides a comparable performance, thus making it an attractive choice for practical implementation."
                    }
                ]
            },
            "16cb678d-3227-4d18-9a72-09233db445c1": {
                "pk": "16cb678d-3227-4d18-9a72-09233db445c1",
                "name": "Zhangdie Yuan",
                "collaborators": [
                    "Andreas Vlachos",
                    "Songbo Hu",
                    "Anna Korhonen",
                    "Ivan Vuli\u0107",
                    "Nedjma Ousidhoum",
                    "Xiaobin Wang",
                    "Eric Chamoun",
                    "Rami Aly",
                    "Chenxi Whitehouse",
                    "Zaiqiao Meng"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Fact-Checking",
                    "Dialogue Systems",
                    "Language Models"
                ],
                "publications": [
                    {
                        "title": "Zero-Shot Fact-Checking with Semantic Triples and Knowledge Graphs",
                        "abstract": "Despite progress in automated fact-checking, most systems require a significant amount of labeled training data, which is expensive. In this paper, we propose a novel zero-shot method, which instead of operating directly on the claim and evidence sentences, decomposes them into semantic triples augmented using external knowledge graphs, and uses large language models trained for natural language inference. This allows it to generalize to adversarial datasets and domains that supervised models require specific training data for. Our empirical results show that our approach outperforms previous zero-shot approaches on FEVER, FEVER-Symmetric, FEVER 2.0, and Climate-FEVER, while being comparable or better than supervised models on the adversarial and the out-of-domain datasets."
                    },
                    {
                        "title": "Varifocal Question Generation for Fact-checking",
                        "abstract": "Fact-checking requires retrieving evidence related to a claim under investigation. The task can be formulated as question generation based on a claim, followed by question answering. However, recent question generation approaches assume that the answer is known and typically contained in a passage given as input, whereas such passages are what is being sought when verifying a claim. In this paper, we present {\\it Varifocal}, a method that generates questions based on different focal points within a given claim, i.e.\\ different spans of the claim and its metadata, such as its source and date. Our method outperforms previous work on a fact-checking question generation dataset on a wide range of automatic evaluation metrics. These results are corroborated by our manual evaluation, which indicates that our method generates more relevant and informative questions. We further demonstrate the potential of focal points in generating sets of clarification questions for product descriptions."
                    },
                    {
                        "title": "DIALIGHT: Lightweight Multilingual Development and Evaluation of Task-Oriented Dialogue Systems with Large Language Models",
                        "abstract": "We present DIALIGHT, a toolkit for developing and evaluating multilingual Task-Oriented Dialogue (ToD) systems which facilitates systematic evaluations and comparisons between ToD systems using fine-tuning of Pretrained Language Models (PLMs) and those utilising the zero-shot and in-context learning capabilities of Large Language Models (LLMs). In addition to automatic evaluation, this toolkit features (i) a secure, user-friendly web interface for fine-grained human evaluation at both local utterance level and global dialogue level, and (ii) a microservice-based backend, improving efficiency and scalability. Our evaluations reveal that while PLM fine-tuning leads to higher accuracy and coherence, LLM-based systems excel in producing diverse and likeable responses. However, we also identify significant challenges of LLMs in adherence to task-specific instructions and generating outputs in multiple languages, highlighting areas for future research. We hope this open-sourced toolkit will serve as a valuable resource for researchers aiming to develop and properly evaluate multilingual ToD systems and will lower, currently still high, entry barriers in the field."
                    },
                    {
                        "title": "PRobELM: Plausibility Ranking Evaluation for Language Models",
                        "abstract": "This paper introduces PRobELM (Plausibility Ranking Evaluation for Language Models), a benchmark designed to assess language models' ability to discern more plausible from less plausible scenarios through their parametric knowledge. While benchmarks such as TruthfulQA emphasise factual accuracy or truthfulness, and others such as COPA explore plausible scenarios without explicitly incorporating world knowledge, PRobELM seeks to bridge this gap by evaluating models' capabilities to prioritise plausible scenarios that leverage world knowledge over less plausible alternatives. This design allows us to assess the potential of language models for downstream use cases such as literature-based discovery where the focus is on identifying information that is likely but not yet known. Our benchmark is constructed from a dataset curated from Wikidata edit histories, tailored to align the temporal bounds of the training data for the evaluated models. PRobELM facilitates the evaluation of language models across multiple prompting types, including statement, text completion, and question-answering. Experiments with 10 models of various sizes and architectures on the relationship between model scales, training recency, and plausibility performance, reveal that factual accuracy does not directly correlate with plausibility performance and that up-to-date training data enhances plausibility assessment across different model architectures."
                    },
                    {
                        "title": "Can Pretrained Language Models (Yet) Reason Deductively?",
                        "abstract": "Acquiring factual knowledge with Pretrained Language Models (PLMs) has attracted increasing attention, showing promising performance in many knowledge-intensive tasks. Their good performance has led the community to believe that the models do possess a modicum of reasoning competence rather than merely memorising the knowledge. In this paper, we conduct a comprehensive evaluation of the learnable deductive (also known as explicit) reasoning capability of PLMs. Through a series of controlled experiments, we posit two main findings. (i) PLMs inadequately generalise learned logic rules and perform inconsistently against simple adversarial surface form edits. (ii) While the deductive reasoning fine-tuning of PLMs does improve their performance on reasoning over unseen knowledge facts, it results in catastrophically forgetting the previously learnt knowledge. Our main results suggest that PLMs cannot yet perform reliable deductive reasoning, demonstrating the importance of controlled examinations and probing of PLMs' reasoning abilities; we reach beyond (misleading) task performance, revealing that PLMs are still far from human-level reasoning capabilities, even for simple deductive tasks."
                    }
                ]
            },
            "cb374bad-86b0-425d-b234-f90692176667": {
                "pk": "cb374bad-86b0-425d-b234-f90692176667",
                "name": "Jiaoyan Chen",
                "collaborators": [
                    "Ian Horrocks",
                    "Yuan He",
                    "Ernesto Jimenez-Ruiz",
                    "Hang Dong",
                    "Huajun Chen",
                    "Ernesto Jim\u00e9nez-Ruiz",
                    "Charles Sutton",
                    "Yinan Liu",
                    "Junwen Luo",
                    "Freddy Lecue"
                ],
                "domain": [
                    "Ontology Alignment",
                    "Knowledge Graph",
                    "Machine Learning",
                    "Spiking Neural Networks"
                ],
                "publications": [
                    {
                        "title": "An Internal Clock Based Space-time Neural Network for Motion Speed Recognition",
                        "abstract": "In this work we present a novel internal clock based space-time neural network for motion speed recognition. The developed system has a spike train encoder, a Spiking Neural Network (SNN) with internal clocking behaviors, a pattern transformation block and a Network Dynamic Dependent Plasticity (NDDP) learning block. The core principle is that the developed SNN will automatically tune its network pattern frequency (internal clock frequency) to recognize human motions in a speed domain. We employed both cartoons and real-world videos as training benchmarks, results demonstrate that our system can not only recognize motions with considerable speed differences (e.g. run, walk, jump, wonder(think) and standstill), but also motions with subtle speed gaps such as run and fast walk. The inference accuracy can be up to 83.3% (cartoon videos) and 75% (real-world videos). Meanwhile, the system only requires six video datasets in the learning stage and with up to 42 training trials. Hardware performance estimation indicates that the training time is 0.84-4.35s and power consumption is 33.26-201mW (based on an ARM Cortex M4 processor). Therefore, our system takes unique learning advantages of the requirement of the small dataset, quick learning and low power performance, which shows great potentials for edge or scalable AI-based applications."
                    },
                    {
                        "title": "Learning from Ontology Streams with Semantic Concept Drift",
                        "abstract": "Data stream learning has been largely studied for extracting knowledge structures from continuous and rapid data records. In the semantic Web, data is interpreted in ontologies and its ordered sequence is represented as an ontology stream. Our work exploits the semantics of such streams to tackle the problem of concept drift i.e., unexpected changes in data distribution, causing most of models to be less accurate as time passes. To this end we revisited (i) semantic inference in the context of supervised stream learning, and (ii) models with semantic embeddings. The experiments show accurate prediction with data from Dublin and Beijing."
                    },
                    {
                        "title": "Canonicalizing Knowledge Base Literals",
                        "abstract": "Ontology-based knowledge bases (KBs) like DBpedia are very valuable resources, but their usefulness and usability is limited by various quality issues. One such issue is the use of string literals instead of semantically typed entities. In this paper we study the automated canonicalization of such literals, i.e., replacing the literal with an existing entity from the KB or with a new entity that is typed using classes from the KB. We propose a framework that combines both reasoning and machine learning in order to predict the relevant entities and types, and we evaluate this framework against state-of-the-art baselines for both semantic typing and entity matching."
                    },
                    {
                        "title": "Human-centric Transfer Learning Explanation via Knowledge Graph [Extended Abstract]",
                        "abstract": "Transfer learning which aims at utilizing knowledge learned from one problem (source domain) to solve another different but related problem (target domain) has attracted wide research attentions. However, the current transfer learning methods are mostly uninterpretable, especially to people without ML expertise. In this extended abstract, we brief introduce two knowledge graph (KG) based frameworks towards human understandable transfer learning explanation. The first one explains the transferability of features learned by Convolutional Neural Network (CNN) from one domain to another through pre-training and fine-tuning, while the second justifies the model of a target domain predicted by models from multiple source domains in zero-shot learning (ZSL). Both methods utilize KG and its reasoning capability to provide rich and human understandable explanations to the transfer procedure."
                    },
                    {
                        "title": "Exploring Large Language Models for Ontology Alignment",
                        "abstract": "This work investigates the applicability of recent generative Large Language Models (LLMs), such as the GPT series and Flan-T5, to ontology alignment for identifying concept equivalence mappings across ontologies. To test the zero-shot performance of Flan-T5-XXL and GPT-3.5-turbo, we leverage challenging subsets from two equivalence matching datasets of the OAEI Bio-ML track, taking into account concept labels and structural contexts. Preliminary findings suggest that LLMs have the potential to outperform existing ontology alignment systems like BERTMap, given careful framework and prompt design."
                    },
                    {
                        "title": "OWL2Vec4OA: Tailoring Knowledge Graph Embeddings for Ontology Alignment",
                        "abstract": "Ontology alignment is integral to achieving semantic interoperability as the number of available ontologies covering intersecting domains is increasing. This paper proposes OWL2Vec4OA, an extension of the ontology embedding system OWL2Vec*. While OWL2Vec* has emerged as a powerful technique for ontology embedding, it currently lacks a mechanism to tailor the embedding to the ontology alignment task. OWL2Vec4OA incorporates edge confidence values from seed mappings to guide the random walk strategy. We present the theoretical foundations, implementation details, and experimental evaluation of our proposed extension, demonstrating its potential effectiveness for ontology alignment tasks."
                    },
                    {
                        "title": "BERTMap: A BERT-based Ontology Alignment System",
                        "abstract": "Ontology alignment (a.k.a ontology matching (OM)) plays a critical role in knowledge integration. Owing to the success of machine learning in many domains, it has been applied in OM. However, the existing methods, which often adopt ad-hoc feature engineering or non-contextual word embeddings, have not yet outperformed rule-based systems especially in an unsupervised setting. In this paper, we propose a novel OM system named BERTMap which can support both unsupervised and semi-supervised settings. It first predicts mappings using a classifier based on fine-tuning the contextual embedding model BERT on text semantics corpora extracted from ontologies, and then refines the mappings through extension and repair by utilizing the ontology structure and logic. Our evaluation with three alignment tasks on biomedical ontologies demonstrates that BERTMap can often perform better than the leading OM systems LogMap and AML."
                    },
                    {
                        "title": "Ontology Enrichment from Texts: A Biomedical Dataset for Concept Discovery and Placement",
                        "abstract": "Mentions of new concepts appear regularly in texts and require automated approaches to harvest and place them into Knowledge Bases (KB), e.g., ontologies and taxonomies. Existing datasets suffer from three issues, (i) mostly assuming that a new concept is pre-discovered and cannot support out-of-KB mention discovery; (ii) only using the concept label as the input along with the KB and thus lacking the contexts of a concept label; and (iii) mostly focusing on concept placement w.r.t a taxonomy of atomic concepts, instead of complex concepts, i.e., with logical operators. To address these issues, we propose a new benchmark, adapting MedMentions dataset (PubMed abstracts) with SNOMED CT versions in 2014 and 2017 under the Diseases sub-category and the broader categories of Clinical finding, Procedure, and Pharmaceutical / biologic product. We provide usage on the evaluation with the dataset for out-of-KB mention discovery and concept placement, adapting recent Large Language Model based methods."
                    },
                    {
                        "title": "TransBox: EL++-closed Ontology Embedding",
                        "abstract": "OWL (Web Ontology Language) ontologies, which are able to represent both relational and type facts as standard knowledge graphs and complex domain knowledge in Description Logic (DL) axioms, are widely adopted in domains such as healthcare and bioinformatics. Inspired by the success of knowledge graph embeddings, embedding OWL ontologies has gained significant attention in recent years. Current methods primarily focus on learning embeddings for atomic concepts and roles, enabling the evaluation based on normalized axioms through specially designed score functions. However, they often neglect the embedding of complex concepts, making it difficult to infer with more intricate axioms. This limitation reduces their effectiveness in advanced reasoning tasks, such as Ontology Learning and ontology-mediated Query Answering. In this paper, we propose EL++-closed ontology embeddings which are able to represent any logical expressions in DL via composition. Furthermore, we develop TransBox, an effective EL++-closed ontology embedding method that can handle many-to-one, one-to-many and many-to-many relations. Our extensive experiments demonstrate that TransBox often achieves state-of-the-art performance across various real-world datasets for predicting complex axioms."
                    },
                    {
                        "title": "Dual Box Embeddings for the Description Logic EL++",
                        "abstract": "OWL ontologies, whose formal semantics are rooted in Description Logic (DL), have been widely used for knowledge representation. Similar to Knowledge Graphs (KGs), ontologies are often incomplete, and maintaining and constructing them has proved challenging. While classical deductive reasoning algorithms use the precise formal semantics of an ontology to predict missing facts, recent years have witnessed growing interest in inductive reasoning techniques that can derive probable facts from an ontology. Similar to KGs, a promising approach is to learn ontology embeddings in a latent vector space, while additionally ensuring they adhere to the semantics of the underlying DL. While a variety of approaches have been proposed, current ontology embedding methods suffer from several shortcomings, especially that they all fail to faithfully model one-to-many, many-to-one, and many-to-many relations and role inclusion axioms. To address this problem and improve ontology completion performance, we propose a novel ontology embedding method named Box$^2$EL for the DL EL++, which represents both concepts and roles as boxes (i.e., axis-aligned hyperrectangles), and models inter-concept relationships using a bumping mechanism. We theoretically prove the soundness of Box$^2$EL and conduct an extensive experimental evaluation, achieving state-of-the-art results across a variety of datasets on the tasks of subsumption prediction, role assertion prediction, and approximating deductive reasoning."
                    },
                    {
                        "title": "Learning Semantic Annotations for Tabular Data",
                        "abstract": "The usefulness of tabular data such as web tables critically depends on understanding their semantics. This study focuses on column type prediction for tables without any meta data. Unlike traditional lexical matching-based methods, we propose a deep prediction model that can fully exploit a table's contextual semantics, including table locality features learned by a Hybrid Neural Network (HNN), and inter-column semantics features learned by a knowledge base (KB) lookup and query answering algorithm.It exhibits good performance not only on individual table sets, but also when transferring from one table set to another."
                    },
                    {
                        "title": "ColNet: Embedding the Semantics of Web Tables for Column Type Prediction",
                        "abstract": "Automatically annotating column types with knowledge base (KB) concepts is a critical task to gain a basic understanding of web tables. Current methods rely on either table metadata like column name or entity correspondences of cells in the KB, and may fail to deal with growing web tables with incomplete meta information. In this paper we propose a neural network based column type annotation framework named ColNet which is able to integrate KB reasoning and lookup with machine learning and can automatically train Convolutional Neural Networks for prediction. The prediction model not only considers the contextual semantics within a cell using word representation, but also embeds the semantics of a column by learning locality features from multiple cells. The method is evaluated with DBPedia and two different web table datasets, T2Dv2 from the general Web and Limaye from Wikipedia pages, and achieves higher performance than the state-of-the-art approaches."
                    },
                    {
                        "title": "When Low Resource NLP Meets Unsupervised Language Model: Meta-pretraining Then Meta-learning for Few-shot Text Classification",
                        "abstract": "Text classification tends to be difficult when data are deficient or when it is required to adapt to unseen classes. In such challenging scenarios, recent studies have often used meta-learning to simulate the few-shot task, thus negating implicit common linguistic features across tasks. This paper addresses such problems using meta-learning and unsupervised language models. Our approach is based on the insight that having a good generalization from a few examples relies on both a generic model initialization and an effective strategy for adapting this model to newly arising tasks. We show that our approach is not only simple but also produces a state-of-the-art performance on a well-studied sentiment classification dataset. It can thus be further suggested that pretraining could be a promising solution for few-shot learning of many other NLP tasks. The code and the dataset to replicate the experiments are made available at https://github.com/zxlzr/FewShotNLP."
                    },
                    {
                        "title": "PRASEMap: A Probabilistic Reasoning and Semantic Embedding based Knowledge Graph Alignment System",
                        "abstract": "Knowledge Graph (KG) alignment aims at finding equivalent entities and relations (i.e., mappings) between two KGs. The existing approaches utilize either reasoning-based or semantic embedding-based techniques, but few studies explore their combination. In this demonstration, we present PRASEMap, an unsupervised KG alignment system that iteratively computes the Mappings with both Probabilistic Reasoning (PR) And Semantic Embedding (SE) techniques. PRASEMap can support various embedding-based KG alignment approaches as the SE module, and enables easy human computer interaction that additionally provides an option for users to feed the mapping annotations back to the system for better results. The demonstration showcases these features via a stand-alone Web application with user friendly interfaces. The demo is available at https://prasemap.qizhy.com."
                    },
                    {
                        "title": "Low-resource Personal Attribute Prediction from Conversation",
                        "abstract": "Personal knowledge bases (PKBs) are crucial for a broad range of applications such as personalized recommendation and Web-based chatbots. A critical challenge to build PKBs is extracting personal attribute knowledge from users' conversation data. Given some users of a conversational system, a personal attribute and these users' utterances, our goal is to predict the ranking of the given personal attribute values for each user. Previous studies often rely on a relative number of resources such as labeled utterances and external data, yet the attribute knowledge embedded in unlabeled utterances is underutilized and their performance of predicting some difficult personal attributes is still unsatisfactory. In addition, it is found that some text classification methods could be employed to resolve this task directly. However, they also perform not well over those difficult personal attributes. In this paper, we propose a novel framework PEARL to predict personal attributes from conversations by leveraging the abundant personal attribute knowledge from utterances under a low-resource setting in which no labeled utterances or external data are utilized. PEARL combines the biterm semantic information with the word co-occurrence information seamlessly via employing the updated prior attribute knowledge to refine the biterm topic model's Gibbs sampling process in an iterative manner. The extensive experimental results show that PEARL outperforms all the baseline methods not only on the task of personal attribute prediction from conversations over two data sets, but also on the more general weakly supervised text classification task over one data set."
                    },
                    {
                        "title": "Dividing the Ontology Alignment Task with Semantic Embeddings and Logic-based Modules",
                        "abstract": "Large ontologies still pose serious challenges to state-of-the-art ontology alignment systems. In this paper we present an approach that combines a neural embedding model and logic-based modules to accurately divide an input ontology matching task into smaller and more tractable matching (sub)tasks. We have conducted a comprehensive evaluation using the datasets of the Ontology Alignment Evaluation Initiative. The results are encouraging and suggest that the proposed method is adequate in practice and can be integrated within the workflow of systems unable to cope with very large ontologies."
                    },
                    {
                        "title": "Language Model Analysis for Ontology Subsumption Inference",
                        "abstract": "Investigating whether pre-trained language models (LMs) can function as knowledge bases (KBs) has raised wide research interests recently. However, existing works focus on simple, triple-based, relational KBs, but omit more sophisticated, logic-based, conceptualised KBs such as OWL ontologies. To investigate an LM's knowledge of ontologies, we propose OntoLAMA, a set of inference-based probing tasks and datasets from ontology subsumption axioms involving both atomic and complex concepts. We conduct extensive experiments on ontologies of different domains and scales, and our results demonstrate that LMs encode relatively less background knowledge of Subsumption Inference (SI) than traditional Natural Language Inference (NLI) but can improve on SI significantly when a small number of samples are given. We will open-source our code and datasets."
                    },
                    {
                        "title": "Reveal the Unknown: Out-of-Knowledge-Base Mention Discovery with Entity Linking",
                        "abstract": "Discovering entity mentions that are out of a Knowledge Base (KB) from texts plays a critical role in KB maintenance, but has not yet been fully explored. The current methods are mostly limited to the simple threshold-based approach and feature-based classification, and the datasets for evaluation are relatively rare. We propose BLINKout, a new BERT-based Entity Linking (EL) method which can identify mentions that do not have corresponding KB entities by matching them to a special NIL entity. To better utilize BERT, we propose new techniques including NIL entity representation and classification, with synonym enhancement. We also apply KB Pruning and Versioning strategies to automatically construct out-of-KB datasets from common in-KB EL datasets. Results on five datasets of clinical notes, biomedical publications, and Wikipedia articles in various domains show the advantages of BLINKout over existing methods to identify out-of-KB mentions for the medical ontologies, UMLS, SNOMED CT, and the general KB, WikiData."
                    }
                ]
            },
            "acd49760-8b98-4e4d-a3df-05508991b32b": {
                "pk": "acd49760-8b98-4e4d-a3df-05508991b32b",
                "name": "Ian Horrocks",
                "collaborators": [
                    "Boris Motik",
                    "Jiaoyan Chen",
                    "Stephan Tobies",
                    "Ulrike Sattler",
                    "Pan Hu",
                    "Yavor Nenov",
                    "Robert Piro",
                    "Ernesto Jimenez-Ruiz",
                    "Yuan He",
                    "Markus Kr\u00f6tzsch"
                ],
                "domain": [
                    "Description Logics",
                    "Ontology Alignment",
                    "Knowledge Representation",
                    "Reasoning Systems"
                ],
                "publications": [
                    {
                        "title": "Reasoning with Axioms: Theory and Pratice",
                        "abstract": "When reasoning in description, modal or temporal logics it is often useful to consider axioms representing universal truths in the domain of discourse. Reasoning with respect to an arbitrary set of axioms is hard, even for relatively inexpressive logics, and it is essential to deal with such axioms in an efficient manner if implemented systems are to be effective in real applications. This is particularly relevant to Description Logics, where subsumption reasoning with respect to a terminology is a fundamental problem. Two optimisation techniques that have proved to be particularly effective in dealing with terminologies are lazy unfolding and absorption. In this paper we seek to improve our theoretical understanding of these important techniques. We define a formal framework that allows the techniques to be precisely described, establish conditions under which they can be safely applied, and prove that, provided these conditions are respected, subsumption testing algorithms will still function correctly. These results are used to show that the procedures used in the FaCT system are correct and, moreover, to show how efficiency can be significantly improved, while still retaining the guarantee of correctness, by relaxing the safety conditions for absorption."
                    },
                    {
                        "title": "Practical Reasoning for Expressive Description Logics",
                        "abstract": "Description Logics (DLs) are a family of knowledge representation formalisms mainly characterised by constructors to build complex concepts and roles from atomic ones. Expressive role constructors are important in many applications, but can be computationally problematical. We present an algorithm that decides satisfiability of the DL ALC extended with transitive and inverse roles, role hierarchies, and qualifying number restrictions. Early experiments indicate that this algorithm is well-suited for implementation. Additionally, we show that ALC extended with just transitive and inverse roles is still in PSPACE. Finally, we investigate the limits of decidability for this family of DLs."
                    },
                    {
                        "title": "A Description Logic Primer",
                        "abstract": "This paper provides a self-contained first introduction to description logics (DLs). The main concepts and features are explained with examples before syntax and semantics of the DL SROIQ are defined in detail. Additional sections review light-weight DL languages, discuss the relationship to the Web Ontology Language OWL and give pointers to further reading."
                    },
                    {
                        "title": "Handling owl:sameAs via Rewriting",
                        "abstract": "Rewriting is widely used to optimise owl:sameAs reasoning in materialisation based OWL 2 RL systems. We investigate issues related to both the correctness and efficiency of rewriting, and present an algorithm that guarantees correctness, improves efficiency, and can be effectively parallelised. Our evaluation shows that our approach can reduce reasoning times on practical data sets by orders of magnitude."
                    },
                    {
                        "title": "Canonicalizing Knowledge Base Literals",
                        "abstract": "Ontology-based knowledge bases (KBs) like DBpedia are very valuable resources, but their usefulness and usability is limited by various quality issues. One such issue is the use of string literals instead of semantically typed entities. In this paper we study the automated canonicalization of such literals, i.e., replacing the literal with an existing entity from the KB or with a new entity that is typed using classes from the KB. We propose a framework that combines both reasoning and machine learning in order to predict the relevant entities and types, and we evaluate this framework against state-of-the-art baselines for both semantic typing and entity matching."
                    },
                    {
                        "title": "Exploring Large Language Models for Ontology Alignment",
                        "abstract": "This work investigates the applicability of recent generative Large Language Models (LLMs), such as the GPT series and Flan-T5, to ontology alignment for identifying concept equivalence mappings across ontologies. To test the zero-shot performance of Flan-T5-XXL and GPT-3.5-turbo, we leverage challenging subsets from two equivalence matching datasets of the OAEI Bio-ML track, taking into account concept labels and structural contexts. Preliminary findings suggest that LLMs have the potential to outperform existing ontology alignment systems like BERTMap, given careful framework and prompt design."
                    },
                    {
                        "title": "Consequence-based Reasoning for Description Logics with Disjunction, Inverse Roles, Number Restrictions, and Nominals",
                        "abstract": "We present a consequence-based calculus for concept subsumption and classification in the description logic ALCHOIQ, which extends ALC with role hierarchies, inverse roles, number restrictions, and nominals. By using standard transformations, our calculus extends to SROIQ, which covers all of OWL 2 DL except for datatypes. A key feature of our calculus is its pay-as-you-go behaviour: unlike existing algorithms, our calculus is worst-case optimal for all the well-known proper fragments of ALCHOIQ, albeit not for the full logic."
                    },
                    {
                        "title": "Learning Semantic Annotations for Tabular Data",
                        "abstract": "The usefulness of tabular data such as web tables critically depends on understanding their semantics. This study focuses on column type prediction for tables without any meta data. Unlike traditional lexical matching-based methods, we propose a deep prediction model that can fully exploit a table's contextual semantics, including table locality features learned by a Hybrid Neural Network (HNN), and inter-column semantics features learned by a knowledge base (KB) lookup and query answering algorithm.It exhibits good performance not only on individual table sets, but also when transferring from one table set to another."
                    },
                    {
                        "title": "Practical Reasoning for Very Expressive Description Logics",
                        "abstract": "Description Logics (DLs) are a family of knowledge representation formalisms mainly characterised by constructors to build complex concepts and roles from atomic ones. Expressive role constructors are important in many applications, but can be computationally problematical. We present an algorithm that decides satisfiability of the DL ALC extended with transitive and inverse roles and functional restrictions with respect to general concept inclusion axioms and role hierarchies; early experiments indicate that this algorithm is well-suited for implementation. Additionally, we show that ALC extended with just transitive and inverse roles is still in PSPACE. We investigate the limits of decidability for this family of DLs, showing that relaxing the constraints placed on the kinds of roles used in number restrictions leads to the undecidability of all inference problems. Finally, we describe a number of optimisation techniques that are crucial in obtaining implementations of the decision procedures, which, despite the worst-case complexity of the problem, exhibit good performance with real-life problems."
                    },
                    {
                        "title": "Conjunctive Query Answering for the Description Logic SHIQ",
                        "abstract": "Conjunctive queries play an important role as an expressive query language for Description Logics (DLs). Although modern DLs usually provide for transitive roles, conjunctive query answering over DL knowledge bases is only poorly understood if transitive roles are admitted in the query. In this paper, we consider unions of conjunctive queries over knowledge bases formulated in the prominent DL SHIQ and allow transitive roles in both the query and the knowledge base. We show decidability of query answering in this setting and establish two tight complexity bounds: regarding combined complexity, we prove that there is a deterministic algorithm for query answering that needs time single exponential in the size of the KB and double exponential in the size of the query, which is optimal. Regarding data complexity, we prove containment in co-NP."
                    },
                    {
                        "title": "Hypertableau Reasoning for Description Logics",
                        "abstract": "We present a novel reasoning calculus for the description logic SHOIQ^+---a knowledge representation formalism with applications in areas such as the Semantic Web. Unnecessary nondeterminism and the construction of large models are two primary sources of inefficiency in the tableau-based reasoning calculi used in state-of-the-art reasoners. In order to reduce nondeterminism, we base our calculus on hypertableau and hyperresolution calculi, which we extend with a blocking condition to ensure termination. In order to reduce the size of the constructed models, we introduce anywhere pairwise blocking. We also present an improved nominal introduction rule that ensures termination in the presence of nominals, inverse roles, and number restrictions---a combination of DL constructs that has proven notoriously difficult to handle. Our implementation shows significant performance improvements over state-of-the-art reasoners on several well-known ontologies."
                    },
                    {
                        "title": "Combining Rewriting and Incremental Materialisation Maintenance for Datalog Programs with Equality",
                        "abstract": "Materialisation precomputes all consequences of a set of facts and a datalog program so that queries can be evaluated directly (i.e., independently from the program). Rewriting optimises materialisation for datalog programs with equality by replacing all equal constants with a single representative; and incremental maintenance algorithms can efficiently update a materialisation for small changes in the input facts. Both techniques are critical to practical applicability of datalog systems; however, we are unaware of an approach that combines rewriting and incremental maintenance. In this paper we present the first such combination, and we show empirically that it can speed up updates by several orders of magnitude compared to using either rewriting or incremental maintenance in isolation."
                    },
                    {
                        "title": "Optimised Maintenance of Datalog Materialisations",
                        "abstract": "To efficiently answer queries, datalog systems often materialise all consequences of a datalog program, so the materialisation must be updated whenever the input facts change. Several solutions to the materialisation update problem have been proposed. The Delete/Rederive (DRed) and the Backward/Forward (B/F) algorithms solve this problem for general datalog, but both contain steps that evaluate rules 'backwards' by matching their heads to a fact and evaluating the partially instantiated rule bodies as queries. We show that this can be a considerable source of overhead even on very small updates. In contrast, the Counting algorithm does not evaluate the rules 'backwards', but it can handle only nonrecursive rules. We present two hybrid approaches that combine DRed and B/F with Counting so as to reduce or even eliminate 'backward' rule evaluation while still handling arbitrary datalog programs. We show empirically that our hybrid algorithms are usually significantly faster than existing approaches, sometimes by orders of magnitude."
                    },
                    {
                        "title": "Modular Materialisation of Datalog Programs",
                        "abstract": "The semina\\\"ive algorithm can materialise all consequences of arbitrary datalog rules, and it also forms the basis for incremental algorithms that update a materialisation as the input facts change. Certain (combinations of) rules, however, can be handled much more efficiently using custom algorithms. To integrate such algorithms into a general reasoning approach that can handle arbitrary rules, we propose a modular framework for materialisation computation and its maintenance. We split a datalog program into modules that can be handled using specialised algorithms, and handle the remaining rules using the semina\\\"ive algorithm. We also present two algorithms for computing the transitive and the symmetric-transitive closure of a relation that can be used within our framework. Finally, we show empirically that our framework can handle arbitrary datalog programs while outperforming existing approaches, often by orders of magnitude."
                    },
                    {
                        "title": "Datalog Materialisation in Distributed RDF Stores with Dynamic Data Exchange",
                        "abstract": "Several centralised RDF systems support datalog reasoning by precomputing and storing all logically implied triples using the wellknown seminaive algorithm. Large RDF datasets often exceed the capacity of centralised RDF systems, and a common solution is to distribute the datasets in a cluster of shared-nothing servers. While numerous distributed query answering techniques are known, distributed seminaive evaluation of arbitrary datalog rules is less understood. In fact, most distributed RDF stores either support no reasoning or can handle only limited datalog fragments. In this paper we extend the dynamic data exchange approach for distributed query answering by Potter et al. [12] to a reasoning algorithm that can handle arbitrary rules while preserving important properties such as nonrepetition of inferences. We also show empirically that our algorithm scales well to very large RDF datasets"
                    },
                    {
                        "title": "Computing CQ lower-bounds over OWL 2 through approximation to RSA",
                        "abstract": "Conjunctive query (CQ) answering over knowledge bases is an important reasoning task. However, with expressive ontology languages such as OWL, query answering is computationally very expensive. The PAGOdA system addresses this issue by using a tractable reasoner to compute lower and upper-bound approximations, falling back to a fully-fledged OWL reasoner only when these bounds don't coincide. The effectiveness of this approach critically depends on the quality of the approximations, and in this paper we explore a technique for computing closer approximations via RSA, an ontology language that subsumes all the OWL 2 profiles while still maintaining tractability. We present a novel approximation of OWL 2 ontologies into RSA, and an algorithm to compute a closer (than PAGOdA) lower bound approximation using the RSA combined approach. We have implemented these algorithms in a prototypical CQ answering system, and we present a preliminary evaluation of our system that shows significant performance improvements w.r.t. PAGOdA."
                    },
                    {
                        "title": "Introducing Nominals to the Combined Query Answering Approaches for EL",
                        "abstract": "So-called combined approaches answer a conjunctive query over a description logic ontology in three steps: first, they materialise certain consequences of the ontology and the data; second, they evaluate the query over the data; and third, they filter the result of the second phase to eliminate unsound answers. Such approaches were developed for various members of the DL-Lite and the EL families of languages, but none of them can handle ontologies containing nominals. In our work, we bridge this gap and present a combined query answering approach for ELHO---a logic that contains all features of the OWL 2 EL standard apart from transitive roles and complex role inclusions. This extension is nontrivial because nominals require equality reasoning, which introduces complexity into the first and the third step. Our empirical evaluation suggests that our technique is suitable for practical application, and so it provides a practical basis for conjunctive query answering in a large fragment of OWL 2 EL."
                    },
                    {
                        "title": "Datalog Reasoning over Compressed RDF Knowledge Bases",
                        "abstract": "Materialisation is often used in RDF systems as a preprocessing step to derive all facts implied by given RDF triples and rules. Although widely used, materialisation considers all possible rule applications and can use a lot of memory for storing the derived facts, which can hinder performance. We present a novel materialisation technique that compresses the RDF triples so that the rules can sometimes be applied to multiple facts at once, and the derived facts can be represented using structure sharing. Our technique can thus require less space, as well as skip certain rule applications. Our experiments show that our technique can be very effective: when the rules are relatively simple, our system is both faster and requires less memory than prominent state-of-the-art RDF systems."
                    },
                    {
                        "title": "BERTMap: A BERT-based Ontology Alignment System",
                        "abstract": "Ontology alignment (a.k.a ontology matching (OM)) plays a critical role in knowledge integration. Owing to the success of machine learning in many domains, it has been applied in OM. However, the existing methods, which often adopt ad-hoc feature engineering or non-contextual word embeddings, have not yet outperformed rule-based systems especially in an unsupervised setting. In this paper, we propose a novel OM system named BERTMap which can support both unsupervised and semi-supervised settings. It first predicts mappings using a classifier based on fine-tuning the contextual embedding model BERT on text semantics corpora extracted from ontologies, and then refines the mappings through extension and repair by utilizing the ontology structure and logic. Our evaluation with three alignment tasks on biomedical ontologies demonstrates that BERTMap can often perform better than the leading OM systems LogMap and AML."
                    }
                ]
            }
        }
    },
    "2410.11449": {
        "paper_data": {
            "title": "Conditional Density Estimation with Histogram Trees",
            "url": "http://arxiv.org/abs/2410.11449v1",
            "arxiv_id": "2410.11449",
            "authors": [
                "Lincen Yang",
                "Matthijs van Leeuwen"
            ],
            "abstract": "Conditional density estimation (CDE) goes beyond regression by modeling the full conditional distribution, providing a richer understanding of the data than just the conditional mean in regression. This makes CDE particularly useful in critical application domains. However, interpretable CDE methods are understudied. Current methods typically employ kernel-based approaches, using kernel functions directly for kernel density estimation or as basis functions in linear models. In contrast, despite their conceptual simplicity and visualization suitability, tree-based methods -- which are arguably more comprehensible -- have been largely overlooked for CDE tasks. Thus, we propose the Conditional Density Tree (CDTree), a fully non-parametric model consisting of a decision tree in which each leaf is formed by a histogram model. Specifically, we formalize the problem of learning a CDTree using the minimum description length (MDL) principle, which eliminates the need for tuning the hyperparameter for regularization. Next, we propose an iterative algorithm that, although greedily, searches the optimal histogram for every possible node split. Our experiments demonstrate that, in comparison to existing interpretable CDE methods, CDTrees are both more accurate (as measured by the log-loss) and more robust against irrelevant features. Further, our approach leads to smaller tree sizes than existing tree-based models, which benefits interpretability.",
            "introduction": "   1 Introduction  Conditional density estimation (CDE) is a crucial yet challenging task in modeling the associations between the features and the continuous target variable, which has received a lot of research interest since the 1970s\u00a0[44]. By modeling the full conditional distribution, CDE is useful when the datasets are multi-modal, heavily skewed, or heteroscedastic. As a result, it is widely applied in various fields, including Genome\u00a0[8], Astronomy\u00a0[3], wind power forecasting\u00a0[19], and computer network\u00a0[47].   CDE provides richer information for data understanding than regression, which only models the conditional mean. This makes CDE desirable for well-informed decision-making in critical areas, such as healthcare\u00a0[32, 53, 14], which calls for interpretability.   Despite its importance, recent CDE research has predominantly focused on black-box models, such as neural networks\u00a0[10, 55, 39, 46, 50] and tree ensembles\u00a0[13]. In contrast, intrinsically interpretable models for CDE are understudied. Specifically, decision tree-based methods are largely neglected, with CADET\u00a0[6] being the only existing method to the best of our knowledge. However, CADET\u2019s assumption of a Gaussian distribution for the target variable in each leaf node limits its ability to model complex conditional densities.   As a result, kernel-based models became the standard among \u2018shallow\u2019 models for CDE, including methods based on kernel density estimation (KDE)\u00a0[44], and linear models with the basis functions chosen as Gaussian kernels\u00a0[54, 11]. Nevertheless, kernel-based models are arguably less interpretable than decision trees: as conditions in decision trees are directly readable, they are comprehensible to humans without statistical expertise (e.g., individuals affected by data-driven decisions rather than professional data analysts).   To address these limitations, we propose the Conditional Density Tree (CDTree), a flexible tree-based CDE model in which each leaf node consists of a histogram model. For illustration, Figure\u00a01 shows the CDTree learned from a dataset about the personal medical cost with demographic features\u00a0[24]. The figure visualizes the histogram models for the conditional densities on three selected leaf nodes, along with the unconditional density. The readable rules, extracted from the tree paths, and the histogram visualizations, make the results easily comprehensible to humans without statistical expertise. For instance, the difference in medical costs between smokers and non-smokers is evident when comparing the plots. The full results with histograms on all leaf nodes, as well as further descriptions of the dataset, are provided in Appendix\u00a0E.   Learning a CDTree from data is a challenging task, as it requires simultaneously optimizing the decision tree structure and the number of bins for histograms on all leaf nodes. While often used for either task, cross-validation is too time-consuming if we need to search for the optimal number of bins for every possible node split. Thus, we formalize the learning problem using the minimum description length (MDL) principle\u00a0[16, 41], which we briefly review in Section\u00a04.1. Adopting MDL eliminates the need for tuning the hyperparameter, by cross-validation, for both regularizing the decision trees\u00a0[5] and for choosing the number of bins for histograms.   Our main contributions are as follows. First, we introduce the CDTree for the CDE task and formalize the learning problem with the MDL principle. Second, we propose an iterative algorithm that can search the optimal histogram for all possible node splits. Third, we benchmark against a wide",
            "references": [
                {
                    "title": "Optimal Sparse Regression Trees",
                    "abstract": "Regression trees are one of the oldest forms of AI models, and their predictions can be made without a calculator, which makes them broadly useful, particularly for high-stakes applications. Within the large literature on regression trees, there has been little effort towards full provable optimization, mainly due to the computational hardness of the problem. This work proposes a dynamic-programming-with-bounds approach to the construction of provably-optimal sparse regression trees. We leverage a novel lower bound based on an optimal solution to the k-Means clustering algorithm on one dimensional data. We are often able to find optimal sparse trees in seconds, even for challenging datasets that involve large numbers of samples and highly-correlated features."
                },
                {
                    "title": "Truly Unordered Probabilistic Rule Sets for Multi-class Classification",
                    "abstract": "Rule set learning has long been studied and has recently been frequently revisited due to the need for interpretable models. Still, existing methods have several shortcomings: 1) most recent methods require a binary feature matrix as input, while learning rules directly from numeric variables is understudied; 2) existing methods impose orders among rules, either explicitly or implicitly, which harms interpretability; and 3) currently no method exists for learning probabilistic rule sets for multi-class target variables (there is only one for probabilistic rule lists). We propose TURS, for Truly Unordered Rule Sets, which addresses these shortcomings. We first formalize the problem of learning truly unordered rule sets. To resolve conflicts caused by overlapping rules, i.e., instances covered by multiple rules, we propose a novel approach that exploits the probabilistic properties of our rule sets. We next develop a two-phase heuristic algorithm that learns rule sets by carefully growing rules. An important innovation is that we use a surrogate score to take the global potential of the rule set into account when learning a local rule. Finally, we empirically demonstrate that, compared to non-probabilistic and (explicitly or implicitly) ordered state-of-the-art methods, our method learns rule sets that not only have better interpretability but also better predictive performance."
                },
                {
                    "title": "Posterior Representations for Bayesian Context Trees: Sampling, Estimation and Convergence",
                    "abstract": "We revisit the Bayesian Context Trees (BCT) modelling framework for discrete time series, which was recently found to be very effective in numerous tasks including model selection, estimation and prediction. A novel representation of the induced posterior distribution on model space is derived in terms of a simple branching process, and several consequences of this are explored in theory and in practice. First, it is shown that the branching process representation leads to a simple variable-dimensional Monte Carlo sampler for the joint posterior distribution on models and parameters, which can efficiently produce independent samples. This sampler is found to be more efficient than earlier MCMC samplers for the same tasks. Then, the branching process representation is used to establish the asymptotic consistency of the BCT posterior, including the derivation of an almost-sure convergence rate. Finally, an extensive study is carried out on the performance of the induced Bayesian entropy estimator. Its utility is illustrated through both simulation experiments and real-world applications, where it is found to outperform several state-of-the-art methods."
                },
                {
                    "title": "LinCDE: Conditional Density Estimation via Lindsey's Method",
                    "abstract": "Conditional density estimation is a fundamental problem in statistics, with scientific and practical applications in biology, economics, finance and environmental studies, to name a few. In this paper, we propose a conditional density estimator based on gradient boosting and Lindsey's method (LinCDE). LinCDE admits flexible modeling of the density family and can capture distributional characteristics like modality and shape. In particular, when suitably parametrized, LinCDE will produce smooth and non-negative density estimates. Furthermore, like boosted regression trees, LinCDE does automatic feature selection. We demonstrate LinCDE's efficacy through extensive simulations and three real data examples."
                },
                {
                    "title": "Conditional Density Estimation of Service Metrics for Networked Services",
                    "abstract": "We predict the conditional distributions of service metrics, such as response time or frame rate, from infrastructure measurements in a networked environment. From such distributions, key statistics of the service metrics, including mean, variance, or quantiles can be computed, which are essential for predicting SLA conformance and enabling service assurance. We present and assess two methods for prediction: (1) mixture models with Gaussian or Lognormal kernels, whose parameters are estimated using mixture density networks, a class of neural networks, and (2) histogram models, which require the target space to be discretized. We apply these methods to a VoD service and a KV store service running on our lab testbed. A comparative evaluation shows the relative effectiveness of the methods when applied to operational data. We find that both methods allow for accurate prediction. While mixture models provide a general and elegant solution, they incur a very high overhead related to hyper-parameter search and neural network training. Histogram models, on the other hand, allow for efficient training, but require adjustment to the specific use case."
                },
                {
                    "title": "Estimating Conditional Mutual Information for Discrete-Continuous Mixtures using Multi-Dimensional Adaptive Histograms",
                    "abstract": "Estimating conditional mutual information (CMI) is an essential yet challenging step in many machine learning and data mining tasks. Estimating CMI from data that contains both discrete and continuous variables, or even discrete-continuous mixture variables, is a particularly hard problem. In this paper, we show that CMI for such mixture variables, defined based on the Radon-Nikodym derivate, can be written as a sum of entropies, just like CMI for purely discrete or continuous data. Further, we show that CMI can be consistently estimated for discrete-continuous mixture variables by learning an adaptive histogram model. In practice, we estimate such a model by iteratively discretizing the continuous data points in the mixture variables. To evaluate the performance of our estimator, we benchmark it against state-of-the-art CMI estimators as well as evaluate it in a causal discovery setting."
                },
                {
                    "title": "Interpretable Machine Learning",
                    "abstract": "Interpretable machine learning has become a popular research direction as deep neural networks (DNNs) have become more powerful and their applications more mainstream, yet DNNs remain difficult to understand. Testing with Concept Activation Vectors, TCAV, (Kim et al. 2017) is an approach to interpreting DNNs in a human-friendly way and has recently received significant attention in the machine learning community. The TCAV algorithm achieves a degree of global interpretability for DNNs through human-defined concepts as explanations. This project introduces Robust TCAV, which builds on TCAV and experimentally determines best practices for this method. The objectives for Robust TCAV are 1) Making TCAV more consistent by reducing variance in the TCAV score distribution and 2) Increasing CAV and TCAV score resistance to perturbations. A difference of means method for CAV generation was determined to be the best practice to achieve both objectives. Many areas of the TCAV process are explored including CAV visualization in low dimensions, negative class selection, and activation perturbation in the direction of a CAV. Finally, a thresholding technique is considered to remove noise in TCAV scores. This project is a step in the direction of making TCAV, an already impactful algorithm in interpretability, more reliable and useful for practitioners."
                },
                {
                    "title": "Minimum Description Length Revisited",
                    "abstract": "This is an up-to-date introduction to and overview of the Minimum Description Length (MDL) Principle, a theory of inductive inference that can be applied to general problems in statistics, machine learning and pattern recognition. While MDL was originally based on data compression ideas, this introduction can be read without any knowledge thereof. It takes into account all major developments since 2007, the last time an extensive overview was written. These include new methods for model selection and averaging and hypothesis testing, as well as the first completely general definition of MDL estimators. Incorporating these developments, MDL can be seen as a powerful extension of both penalized likelihood and Bayesian approaches, in which penalization functions and prior distributions are replaced by more general luckiness functions, average-case methodology is replaced by a more robust worst-case approach, and in which methods classically viewed as highly distinct, such as AIC versus BIC and cross-validation versus Bayes can, to a large extent, be viewed from a unified perspective."
                },
                {
                    "title": "Noise Regularization for Conditional Density Estimation",
                    "abstract": "Modelling statistical relationships beyond the conditional mean is crucial in many settings. Conditional density estimation (CDE) aims to learn the full conditional probability density from data. Though highly expressive, neural network based CDE models can suffer from severe over-fitting when trained with the maximum likelihood objective. Due to the inherent structure of such models, classical regularization approaches in the parameter space are rendered ineffective. To address this issue, we develop a model-agnostic noise regularization method for CDE that adds random perturbations to the data during training. We demonstrate that the proposed approach corresponds to a smoothness regularization and prove its asymptotic consistency. In our experiments, noise regularization significantly and consistently outperforms other regularization methods across seven data sets and three CDE models. The effectiveness of noise regularization makes neural network based CDE the preferable method over previous non- and semi-parametric approaches, even when training data is scarce."
                },
                {
                    "title": "Dirac Delta Regression: Conditional Density Estimation with Clinical Trials",
                    "abstract": "Personalized medicine seeks to identify the causal effect of treatment for a particular patient as opposed to a clinical population at large. Most investigators estimate such personalized treatment effects by regressing the outcome of a randomized clinical trial (RCT) on patient covariates. The realized value of the outcome may however lie far from the conditional expectation. We therefore introduce a method called Dirac Delta Regression (DDR) that estimates the entire conditional density from RCT data in order to visualize the probabilities across all possible treatment outcomes. DDR transforms the outcome into a set of asymptotically Dirac delta distributions and then estimates the density using non-linear regression. The algorithm can identify significant patient-specific treatment effects even when no population level effect exists. Moreover, DDR outperforms state-of-the-art algorithms in conditional density estimation on average regardless of the need for causal inference."
                },
                {
                    "title": "Optimal Sparse Decision Trees",
                    "abstract": "Decision tree algorithms have been among the most popular algorithms for interpretable (transparent) machine learning since the early 1980's. The problem that has plagued decision tree algorithms since their inception is their lack of optimality, or lack of guarantees of closeness to optimality: decision tree algorithms are often greedy or myopic, and sometimes produce unquestionably suboptimal models. Hardness of decision tree optimization is both a theoretical and practical obstacle, and even careful mathematical programming approaches have not been able to solve these problems efficiently. This work introduces the first practical algorithm for optimal decision trees for binary variables. The algorithm is a co-design of analytical bounds that reduce the search space and modern systems techniques, including data structures and a custom bit-vector library. Our experiments highlight advantages in scalability, speed, and proof of optimality."
                },
                {
                    "title": "Conditional Density Estimation with Neural Networks: Best Practices and Benchmarks",
                    "abstract": "Given a set of empirical observations, conditional density estimation aims to capture the statistical relationship between a conditional variable $\\mathbf{x}$ and a dependent variable $\\mathbf{y}$ by modeling their conditional probability $p(\\mathbf{y}|\\mathbf{x})$. The paper develops best practices for conditional density estimation for finance applications with neural networks, grounded on mathematical insights and empirical evaluations. In particular, we introduce a noise regularization and data normalization scheme, alleviating problems with over-fitting, initialization and hyper-parameter sensitivity of such estimators. We compare our proposed methodology with popular semi- and non-parametric density estimators, underpin its effectiveness in various benchmarks on simulated and Euro Stoxx 50 data and show its superior performance. Our methodology allows to obtain high-quality estimators for statistical expectations of higher moments, quantiles and non-linear return transformations, with very little assumptions about the return dynamic."
                },
                {
                    "title": "extremefit: A Package for Extreme Quantiles",
                    "abstract": "extremefit is a package to estimate the extreme quantiles and probabilities of rare events. The idea of our approach is to adjust the tail of the distribution function over a threshold with a Pareto distribution. We propose a pointwise data driven procedure to choose the threshold. To illustrate the method, we use simulated data sets and three real-world data sets included in the package."
                },
                {
                    "title": "Gaussian Process Conditional Density Estimation",
                    "abstract": "Conditional Density Estimation (CDE) models deal with estimating conditional distributions. The conditions imposed on the distribution are the inputs of the model. CDE is a challenging task as there is a fundamental trade-off between model complexity, representational capacity and overfitting. In this work, we propose to extend the model's input with latent variables and use Gaussian processes (GP) to map this augmented input onto samples from the conditional distribution. Our Bayesian approach allows for the modeling of small datasets, but we also provide the machinery for it to be applied to big data using stochastic variational inference. Our approach can be used to model densities even in sparse data regions, and allows for sharing learned structure between conditions. We illustrate the effectiveness and wide-reaching applicability of our model on a variety of real-world problems, such as spatio-temporal density estimation of taxi drop-offs, non-Gaussian noise modeling, and few-shot learning on omniglot images."
                },
                {
                    "title": "RFCDE: Random Forests for Conditional Density Estimation",
                    "abstract": "Random forests is a common non-parametric regression technique which performs well for mixed-type data and irrelevant covariates, while being robust to monotonic variable transformations. Existing random forest implementations target regression or classification. We introduce the RFCDE package for fitting random forest models optimized for nonparametric conditional density estimation, including joint densities for multiple responses. This enables analysis of conditional probability distributions which is useful for propagating uncertainty and of joint distributions that describe relationships between multiple responses and covariates. RFCDE is released under the MIT open-source license and can be accessed at this https URL . Both R and Python versions, which call a common C++ library, are available."
                },
                {
                    "title": "Quotient Normalized Maximum Likelihood Criterion for Learning Bayesian Network Structures",
                    "abstract": "We introduce an information theoretic criterion for Bayesian network structure learning which we call quotient normalized maximum likelihood (qNML). In contrast to the closely related factorized normalized maximum likelihood criterion, qNML satisfies the property of score equivalence. It is also decomposable and completely free of adjustable hyperparameters. For practical computations, we identify a remarkably accurate approximation proposed earlier by Szpankowski and Weinberger. Experiments on both simulated and real data demonstrate that the new criterion leads to parsimonious models with good predictive accuracy."
                },
                {
                    "title": "Conditional Density Estimation with Bayesian Normalising Flows",
                    "abstract": "Modeling complex conditional distributions is critical in a variety of settings. Despite a long tradition of research into conditional density estimation, current methods employ either simple parametric forms or are difficult to learn in practice. This paper employs normalising flows as a flexible likelihood model and presents an efficient method for fitting them to complex densities. These estimators must trade-off between modeling distributional complexity, functional complexity and heteroscedasticity without overfitting. We recognize these trade-offs as modeling decisions and develop a Bayesian framework for placing priors over these conditional density estimators using variational Bayesian neural networks. We evaluate this method on several small benchmark regression datasets, on some of which it obtains state of the art performance. Finally, we apply the method to two spatial density modeling tasks with over 1 million datapoints using the New York City yellow taxi dataset and the Chicago crime dataset."
                },
                {
                    "title": "Bottleneck Conditional Density Estimation",
                    "abstract": "We introduce a new framework for training deep generative models for high-dimensional conditional density estimation. The Bottleneck Conditional Density Estimator (BCDE) is a variant of the conditional variational autoencoder (CVAE) that employs layer(s) of stochastic variables as the bottleneck between the input $x$ and target $y$, where both are high-dimensional. Crucially, we propose a new hybrid training method that blends the conditional generative model with a joint generative model. Hybrid blending is the key to effective training of the BCDE, which avoids overfitting and provides a novel mechanism for leveraging unlabeled data. We show that our hybrid training procedure enables models to achieve competitive results in the MNIST quadrant prediction task in the fully-supervised setting, and sets new benchmarks in the semi-supervised regime for MNIST, SVHN, and CelebA."
                },
                {
                    "title": "\u201cWhy Should I Trust You?\u201d: Explaining the Predictions of Any Classifier",
                    "abstract": "Despite widespread adoption, machine learning models remain mostly black boxes. Understanding the reasons behind predictions is, however, quite important in assessing trust, which is fundamental if one plans to take action based on a prediction, or when choosing whether to deploy a new model. Such understanding also provides insights into the model, which can be used to transform an untrustworthy model or prediction into a trustworthy one. In this work, we propose LIME, a novel explanation technique that explains the predictions of any classifier in an interpretable and faithful manner, by learning an interpretable model locally varound the prediction. We also propose a method to explain models by presenting representative individual predictions and their explanations in a non-redundant way, framing the task as a submodular optimization problem. We demonstrate the flexibility of these methods by explaining different models for text (e.g. random forests) and image classification (e.g. neural networks). We show the utility of explanations via novel experiments, both simulated and with human subjects, on various scenarios that require trust: deciding if one should trust a prediction, choosing between models, improving an untrustworthy classifier, and identifying why a classifier should not be trusted."
                },
                {
                    "title": "Variational Inference with Normalizing Flows",
                    "abstract": "The choice of approximate posterior distribution is one of the core problems in variational inference. Most applications of variational inference employ simple families of posterior approximations in order to allow for efficient inference, focusing on mean-field or other simple structured approximations. This restriction has a significant impact on the quality of inferences made using variational methods. We introduce a new approach for specifying flexible, arbitrarily complex and scalable approximate posterior distributions. Our approximations are distributions constructed through a normalizing flow, whereby a simple initial density is transformed into a more complex one by applying a sequence of invertible transformations until a desired level of complexity is attained. We use this view of normalizing flows to develop categories of finite and infinitesimal flows and provide a unified view of approaches for constructing rich posterior approximations. We demonstrate that the theoretical advantages of having posteriors that better match the true posterior, combined with the scalability of amortized variational approaches, provides a clear improvement in performance and applicability of variational inference."
                },
                {
                    "title": "The Aitchison and Aitken Kernel Function Revisited",
                    "abstract": "Over three decades ago Aitchison and Aitken proposed a novel kernel function for estimating the density functions of underlying distributions in discrete input spaces. To the best of our knowledge, it has not been shown whether this kernel function is positive definite ({\\it i.e.}, a reproducing kernel function) on these spaces. Its positive definiteness would have enriched and enlarged its applicability domain: a positive definite kernel function has an associated Reproducing Kernel Hilbert Space, a framework on which a variety of powerful statistical and machine learning schemes can be developed.This paper aims to demonstrate that Aitchison and Aitken's kernel function is indeed positive definite on discrete metric spaces. We also touch on possible applications of the proposed theorem."
                },
                {
                    "title": "Do maximum waiting times guarantees change clinical priorities? A Conditional Density Estimation approach",
                    "abstract": "The level and distribution of patient waiting times for elective treatments is a major concern in publicly-funded health care systems. Strict targets, which have specified maximum waiting times, have been introduced in the NHS over the last decade and have been criticized for distorting existing clinical priorities in scheduling hospital treatment. We demonstrate the usefulness of Conditional Density Estimation (CDE) in the evaluation of the reform using data for Scotland for 2002 and 2007. The CDE approach allows for variation in the e.ects of patients\u00e2\u20ac\u2122 characteristics over ranges of the distribution and avoids restrictive assumptions about error distribution and functional form. We find that health providers achieved the target by reducing the waiting times for long-waiting patients at the expense of short-waiting patients. We also document a change in the prioritization between di.erent patient groups with some patient groups benefiting at the cost of others."
                },
                {
                    "title": "Using Conditional Kernel Density Estimation for Wind Power Density Forecasting",
                    "abstract": "Of the various renewable energy resources, wind power is widely recognized as one of the most promising. The management of wind farms and electricity systems can benefit greatly from the availability of estimates of the probability distribution of wind power generation. However, most research has focused on point forecasting of wind power. In this article, we develop an approach to producing density forecasts for the wind power generated at individual wind farms. Our interest is in intraday data and prediction from 1 to 72 hours ahead. We model wind power in terms of wind speed and wind direction. In this framework, there are two key uncertainties. First, there is the inherent uncertainty in wind speed and direction, and we model this using a bivariate vector autoregressive moving average-generalized autoregressive conditional heteroscedastic (VARMA-GARCH) model, with a Student t error distribution, in the Cartesian space of wind speed and direction. Second, there is the stochastic nature of the relationship of wind power to wind speed (described by the power curve), and to wind direction. We model this using conditional kernel density (CKD) estimation, which enables a nonparametric modeling of the conditional density of wind power. Using Monte Carlo simulation of the VARMA-GARCH model and CKD estimation, density forecasts of wind speed and direction are converted to wind power density forecasts. Our work is novel in several respects: previous wind power studies have not modeled a stochastic power curve; to accommodate time evolution in the power curve, we incorporate a time decay factor within the CKD method; and the CKD method is conditional on a density, rather than a single value. The new approach is evaluated using datasets from four Greek wind farms."
                },
                {
                    "title": "Scikit-learn: Machine Learning in Python",
                    "abstract": "Scikit-learn is a Python module integrating a wide range of state-of-the-art machine learning algorithms for medium-scale supervised and unsupervised problems. This package focuses on bringing machine learning to non-specialists using a general-purpose high-level language. Emphasis is put on ease of use, performance, documentation, and API consistency. It has minimal dependencies and is distributed under the simplified BSD license, encouraging its use in both academic and commercial settings. Source code, binaries, and documentation can be downloaded from http://scikit-learn.sourceforge.net."
                },
                {
                    "title": "Conditional Density Estimation via Least-Squares Density Ratio Estimation",
                    "abstract": "Estimating the conditional mean of an inputoutput relation is the goal of regression. However, regression analysis is not sufficiently informative if the conditional distribution has multi-modality, is highly asymmetric, or contains heteroscedastic noise. In such scenarios, estimating the conditional distribution itself would be more useful. In this paper, we propose a novel method of conditional density estimation. Our basic idea is to express the conditional density in terms of the ratio of unconditional densities, and the ratio is directly estimated without going through density estimation. Experiments using benchmark and robot transition datasets illustrate the usefulness of the proposed approach."
                },
                {
                    "title": "Bayesian network structure learning using factorized NML universal models",
                    "abstract": "Universal codes/models can be used for data compression and model selection by the minimum description length (MDL) principle. For many interesting model classes, such as Bayesian networks, the minimax regret optimal normalized maximum likelihood (NML) universal model is computationally very demanding. We suggest a computationally feasible alternative to NML for Bayesian networks, the factorized NML universal model, where the normalization is done locally for each variable. This can be seen as an approximate sum-product algorithm. We show that this new universal model performs extremely well in model selection, compared to the existing state-of-the-art, even for small sample sizes."
                },
                {
                    "title": "Robust Machine Learning Applied to Astronomical Data Sets. III. Probabilistic Photometric Redshifts for Galaxies and Quasars in the SDSS and GALEX",
                    "abstract": "We apply machine learning in the form of a nearest neighbor instance-based algorithm (NN) to generate full photometric redshift probability density functions (PDFs) for objects in the Fifth Data Release of the Sloan Digital Sky Survey (SDSS DR5). We use a conceptually simple but novel application of NN to generate the PDFs, perturbing the object colors by their measurement error and using the resulting instances of nearest neighbor distributions to generate numerous individual redshifts. When the redshifts are compared to existing SDSS spectroscopic data, we find that the mean value of each PDF has a dispersion between the photometric and spectroscopic redshift consistent with other machine learning techniques, being \u03c3 = 0.0207 \u00b1 0.0001 for main sample galaxies to r < 17.77 mag, \u03c3 = 0.0243 \u00b1 0.0002 for luminous red galaxies to r\u2272 19.2 mag, and \u03c3 = 0.343 \u00b1 0.005 for quasars to i < 20.3 mag. The PDFs allow the selection of subsets with improved statistics. For quasars, the improvement is dramatic: for those with a single peak in their probability distribution, the dispersion is reduced from 0.343 to \u03c3 = 0.117 \u00b1 0.010, and the photometric redshift is within 0.3 of the spectroscopic redshift for 99.3% \u00b1 0.1% of the objects. Thus, for this optical quasar sample, we can virtually eliminate \u201ccatastrophic\u201d photometric redshift estimates. In addition to the SDSS sample, we incorporate ultraviolet photometry from the Third Data Release of the Galaxy Evolution Explorer All-Sky Imaging Survey (GALEX AIS GR3) to create PDFs for objects seen in both surveys. For quasars, the increased coverage of the observed-frame UV of the SED results in significant improvement over the full SDSS sample, with \u03c3 = 0.234 \u00b1 0.010. We demonstrate that this improvement is genuine and not an artifact of the SDSS-GALEX matching process."
                },
                {
                    "title": "MDL Histogram Density Estimation",
                    "abstract": "We regard histogram density estimation as a model selection problem. Our approach is based on the information-theoretic minimum description length (MDL) principle, which can be applied for tasks such as data clustering, density estimation, image denoising and model selection in general. MDLbased model selection is formalized via the normalized maximum likelihood (NML) distribution, which has several desirable optimality properties. We show how this framework can be applied for learning generic, irregular (variable-width bin) histograms, and how to compute the NML model selection criterion efficiently. We also derive a dynamic programming algorithm for finding both the MDL-optimal bin count and the cut point locations in polynomial time. Finally, we demonstrate our approach via simulation tests."
                },
                {
                    "title": "Unsupervised multiscale color image segmentation based on MDL principle",
                    "abstract": "We present an unsupervised multiscale color image segmentation algorithm. The basic idea is to apply mean shift clustering to obtain an over-segmentation and then merge regions at multiple scales to minimize the minimum description length criterion. The performance on the Berkeley segmentation benchmark compares favorably with some existing approaches"
                },
                {
                    "title": "Elements of Information Theory",
                    "abstract": "Preface to the Second Edition. Preface to the First Edition. Acknowledgments for the Second Edition. Acknowledgments for the First Edition. 1. Introduction and Preview. 1.1 Preview of the Book. 2. Entropy, Relative Entropy, and Mutual Information. 2.1 Entropy. 2.2 Joint Entropy and Conditional Entropy. 2.3 Relative Entropy and Mutual Information. 2.4 Relationship Between Entropy and Mutual Information. 2.5 Chain Rules for Entropy, Relative Entropy, and Mutual Information. 2.6 Jensen's Inequality and Its Consequences. 2.7 Log Sum Inequality and Its Applications. 2.8 Data-Processing Inequality. 2.9 Sufficient Statistics. 2.10 Fano's Inequality. Summary. Problems. Historical Notes. 3. Asymptotic Equipartition Property. 3.1 Asymptotic Equipartition Property Theorem. 3.2 Consequences of the AEP: Data Compression. 3.3 High-Probability Sets and the Typical Set. Summary. Problems. Historical Notes. 4. Entropy Rates of a Stochastic Process. 4.1 Markov Chains. 4.2 Entropy Rate. 4.3 Example: Entropy Rate of a Random Walk on a Weighted Graph. 4.4 Second Law of Thermodynamics. 4.5 Functions of Markov Chains. Summary. Problems. Historical Notes. 5. Data Compression. 5.1 Examples of Codes. 5.2 Kraft Inequality. 5.3 Optimal Codes. 5.4 Bounds on the Optimal Code Length. 5.5 Kraft Inequality for Uniquely Decodable Codes. 5.6 Huffman Codes. 5.7 Some Comments on Huffman Codes. 5.8 Optimality of Huffman Codes. 5.9 Shannon-Fano-Elias Coding. 5.10 Competitive Optimality of the Shannon Code. 5.11 Generation of Discrete Distributions from Fair Coins. Summary. Problems. Historical Notes. 6. Gambling and Data Compression. 6.1 The Horse Race. 6.2 Gambling and Side Information. 6.3 Dependent Horse Races and Entropy Rate. 6.4 The Entropy of English. 6.5 Data Compression and Gambling. 6.6 Gambling Estimate of the Entropy of English. Summary. Problems. Historical Notes. 7. Channel Capacity. 7.1 Examples of Channel Capacity. 7.2 Symmetric Channels. 7.3 Properties of Channel Capacity. 7.4 Preview of the Channel Coding Theorem. 7.5 Definitions. 7.6 Jointly Typical Sequences. 7.7 Channel Coding Theorem. 7.8 Zero-Error Codes. 7.9 Fano's Inequality and the Converse to the Coding Theorem. 7.10 Equality in the Converse to the Channel Coding Theorem. 7.11 Hamming Codes. 7.12 Feedback Capacity. 7.13 Source-Channel Separation Theorem. Summary. Problems. Historical Notes. 8. Differential Entropy. 8.1 Definitions. 8.2 AEP for Continuous Random Variables. 8.3 Relation of Differential Entropy to Discrete Entropy. 8.4 Joint and Conditional Differential Entropy. 8.5 Relative Entropy and Mutual Information. 8.6 Properties of Differential Entropy, Relative Entropy, and Mutual Information. Summary. Problems. Historical Notes. 9. Gaussian Channel. 9.1 Gaussian Channel: Definitions. 9.2 Converse to the Coding Theorem for Gaussian Channels. 9.3 Bandlimited Channels. 9.4 Parallel Gaussian Channels. 9.5 Channels with Colored Gaussian Noise. 9.6 Gaussian Channels with Feedback. Summary. Problems. Historical Notes. 10. Rate Distortion Theory. 10.1 Quantization. 10.2 Definitions. 10.3 Calculation of the Rate Distortion Function. 10.4 Converse to the Rate Distortion Theorem. 10.5 Achievability of the Rate Distortion Function. 10.6 Strongly Typical Sequences and Rate Distortion. 10.7 Characterization of the Rate Distortion Function. 10.8 Computation of Channel Capacity and the Rate Distortion Function. Summary. Problems. Historical Notes. 11. Information Theory and Statistics. 11.1 Method of Types. 11.2 Law of Large Numbers. 11.3 Universal Source Coding. 11.4 Large Deviation Theory. 11.5 Examples of Sanov's Theorem. 11.6 Conditional Limit Theorem. 11.7 Hypothesis Testing. 11.8 Chernoff-Stein Lemma. 11.9 Chernoff Information. 11.10 Fisher Information and the Cram-er-Rao Inequality. Summary. Problems. Historical Notes. 12. Maximum Entropy. 12.1 Maximum Entropy Distributions. 12.2 Examples. 12.3 Anomalous Maximum Entropy Problem. 12.4 Spectrum Estimation. 12.5 Entropy Rates of a Gaussian Process. 12.6 Burg's Maximum Entropy Theorem. Summary. Problems. Historical Notes. 13. Universal Source Coding. 13.1 Universal Codes and Channel Capacity. 13.2 Universal Coding for Binary Sequences. 13.3 Arithmetic Coding. 13.4 Lempel-Ziv Coding. 13.5 Optimality of Lempel-Ziv Algorithms. Compression. Summary. Problems. Historical Notes. 14. Kolmogorov Complexity. 14.1 Models of Computation. 14.2 Kolmogorov Complexity: Definitions and Examples. 14.3 Kolmogorov Complexity and Entropy. 14.4 Kolmogorov Complexity of Integers. 14.5 Algorithmically Random and Incompressible Sequences. 14.6 Universal Probability. 14.7 Kolmogorov complexity. 14.9 Universal Gambling. 14.10 Occam's Razor. 14.11 Kolmogorov Complexity and Universal Probability. 14.12 Kolmogorov Sufficient Statistic. 14.13 Minimum Description Length Principle. Summary. Problems. Historical Notes. 15. Network Information Theory. 15.1 Gaussian Multiple-User Channels. 15.2 Jointly Typical Sequences. 15.3 Multiple-Access Channel. 15.4 Encoding of Correlated Sources. 15.5 Duality Between Slepian-Wolf Encoding and Multiple-Access Channels. 15.6 Broadcast Channel. 15.7 Relay Channel. 15.8 Source Coding with Side Information. 15.9 Rate Distortion with Side Information. 15.10 General Multiterminal Networks. Summary. Problems. Historical Notes. 16. Information Theory and Portfolio Theory. 16.1 The Stock Market: Some Definitions. 16.2 Kuhn-Tucker Characterization of the Log-Optimal Portfolio. 16.3 Asymptotic Optimality of the Log-Optimal Portfolio. 16.4 Side Information and the Growth Rate. 16.5 Investment in Stationary Markets. 16.6 Competitive Optimality of the Log-Optimal Portfolio. 16.7 Universal Portfolios. 16.8 Shannon-McMillan-Breiman Theorem (General AEP). Summary. Problems. Historical Notes. 17. Inequalities in Information Theory. 17.1 Basic Inequalities of Information Theory. 17.2 Differential Entropy. 17.3 Bounds on Entropy and Relative Entropy. 17.4 Inequalities for Types. 17.5 Combinatorial Bounds on Entropy. 17.6 Entropy Rates of Subsets. 17.7 Entropy and Fisher Information. 17.8 Entropy Power Inequality and Brunn-Minkowski Inequality. 17.9 Inequalities for Determinants. 17.10 Inequalities for Ratios of Determinants. Summary. Problems. Historical Notes. Bibliography. List of Symbols. Index."
                },
                {
                    "title": "Cross-Validation and the Estimation of Conditional Probability Densities",
                    "abstract": "Many practical problems, especially some connected with forecasting, require nonparametric estimation of conditional densities from mixed data. For example, given an explanatory data vector X for a prospective customer, with components that could include the customer's salary, occupation, age, sex, marital status, and address, a company might wish to estimate the density of the expenditure, Y, that could be made by that person, basing the inference on observations of (X, Y) for previous clients. Choosing appropriate smoothing parameters for this problem can be tricky, not in the least because plug-in rules take a particularly complex form in the case of mixed data. An obvious difficulty is that there exists no general formula for the optimal smoothing parameters. More insidiously, and more seriously, it can be difficult to determine which components of X are relevant to the problem of conditional inference. For example, if the jth component of X is independent of Y, then that component is irrelevant to estimating the density of Y given X, and ideally should be dropped before conducting inference. In this article we show that cross-validation overcomes these difficulties. It automatically determines which components are relevant and which are not, through assigning large smoothing parameters to the latter and consequently shrinking them toward the uniform distribution on the respective marginals. This effectively removes irrelevant components from contention, by suppressing their contribution to estimator variance; they already have very small bias, a consequence of their independence of Y. Cross-validation also yields important information about which components are relevant; the relevant components are precisely those that cross-validation has chosen to smooth in a traditional way, by assigning them smoothing parameters of conventional size. Indeed, cross-validation produces asymptotically optimal smoothing for relevant components, while eliminating irrelevant components by oversmoothing. In the problem of nonparametric estimation of a conditional density, cross-validation comes into its own as a method with no obvious peers."
                },
                {
                    "title": "Top-down induction of model trees with regression and splitting nodes",
                    "abstract": "Model trees are an extension of regression trees that associate leaves with multiple regression models. In this paper, a method for the data-driven construction of model trees is presented, namely, the stepwise model tree induction (SMOTI) method. Its main characteristic is the induction of trees with two types of nodes: regression nodes, which perform only straight-line regression, and splitting nodes, which partition the feature space. The multiple linear model associated with each leaf is then built stepwise by combining straight-line regressions reported along the path from the root to the leaf. In this way, internal regression nodes contribute to the definition of multiple models and have a \"global\" effect, while straight-line regressions at leaves have only \"local\" effects. Experimental results on artificially generated data sets show that SMOTI outperforms two model tree induction systems, M5' and RETIS, in accuracy. Results on benchmark data sets used for studies on both regression and model trees show that SMOTI performs better than RETIS in accuracy, while it is not possible to draw statistically significant conclusions on the comparison with M5'. Model trees induced by SMOTI are generally simple and easily interpretable and their analysis often reveals interesting patterns."
                },
                {
                    "title": "Extreme Value Distributions: Theory and Applications",
                    "abstract": "Univariate extreme value distributions generalized extreme value distributions multivariate extreme value distributions."
                },
                {
                    "title": "A flexible approach for estimating the effects of covariates on health expenditures.",
                    "abstract": "Our estimation strategy uses sequences of conditional probability functions, similar to those used in discrete time hazard rate analyses, to construct a discrete approximation to the density function of an outcome of interest conditional on exogenous explanatory variables. Once the conditional density function has been constructed, we can examine expectations of arbitrary functions of the outcome of interest and evaluate how these expectations vary with observed exogenous covariates. We demonstrate the features and precision of the conditional density estimation method (and compare it to other commonly used methods) through Monte Carlo experiments and an application to health expenditures using the RAND Health Insurance Experiment data. Overall, we find that the approximate conditional density estimator provides accurate and precise estimates of derivatives of expected outcomes for a wide range of types of explanatory variables."
                },
                {
                    "title": "Estimation of conditional densities and sensitivity measures in nonlinear dynamical systems",
                    "abstract": "Using locally polynomial regression, we develop nonparametric estimators for the conditional density function and its square root, and their partial derivatives. Two measures of sensitivity to initial conditions in nonlinear stochastic dynamic systems are proposed, one of which relates Fisher information with initial-value sensitivity in dynamical systems. We propose estimators for these, and show asymptotic normality for one of them. We further propose a simple method for choosing the bandwidth. The methods are illustrated by simulation of two well-known models in dynamical systems."
                },
                {
                    "title": "C4.5: Programs for Machine Learning",
                    "abstract": "From the Publisher: \nClassifier systems play a major role in machine learning and knowledge-based systems, and Ross Quinlan's work on ID3 and C4.5 is widely acknowledged to have made some of the most significant contributions to their development. This book is a complete guide to the C4.5 system as implemented in C for the UNIX environment. It contains a comprehensive guide to the system's use , the source code (about 8,800 lines), and implementation notes. The source code and sample datasets are also available on a 3.5-inch floppy diskette for a Sun workstation. \n \nC4.5 starts with large sets of cases belonging to known classes. The cases, described by any mixture of nominal and numeric properties, are scrutinized for patterns that allow the classes to be reliably discriminated. These patterns are then expressed as models, in the form of decision trees or sets of if-then rules, that can be used to classify new cases, with emphasis on making the models understandable as well as accurate. The system has been applied successfully to tasks involving tens of thousands of cases described by hundreds of properties. The book starts from simple core learning methods and shows how they can be elaborated and extended to deal with typical problems such as missing data and over hitting. Advantages and disadvantages of the C4.5 approach are discussed and illustrated with several case studies. \n \nThis book and software should be of interest to developers of classification-based intelligent systems and to students in machine learning and expert systems courses."
                },
                {
                    "title": "A UNIVERSAL PRIOR FOR INTEGERS AND ESTIMATION BY MINIMUM DESCRIPTION LENGTH",
                    "abstract": "of the number of bits required to write down the observed data, has been reformulated to extend the classical maximum likelihood principle. The principle permits estimation of the number of the parameters in statistical models in addition to their values and even of the way the parameters appear in the models; i.e., of the model structures. The principle rests on a new way to interpret and construct a universal prior distribution for the integers, which makes sense even when the parameter is an individual object. Truncated realvalued parameters are converted to integers by dividing them by their precision, and their prior is determined from the universal prior for the integers by optimizing the precision. 1. Introduction. In this paper we study estimation based upon the principle of minimizing the total number of binary digits required to rewrite the observed data, when each observation is given with some precision. Instead of attempting at an absolutely shortest description, which would be futile, we look for the optimum relative to a class of parametrically given distributions. This Minimum Description Length (MDL) principle, which we introduced in a less comprehensive form in [25], turns out to degenerate to the more familiar Maximum Likelihood (ML) principle in case the number of parameters in the models is fixed, so that the description length of the parameters themselves can be ignored. In another extreme case, where the parameters determine the data, it similarly degenerates to Jaynes's principle of maximum entropy, [14]. But the main power of the new criterion is that it permits estimates of the entire model, its parameters, their number, and even the way the parameters appear in the model; i.e., the model structure. Hence, there will be no need to supplement the estimated parameters with a separate hypothesis test to decide whether a model is adequately parameterized or, perhaps, over parameterized."
                },
                {
                    "title": "Comparison of Probability Distributions",
                    "abstract": "OFTEN, more than one probability distribution is theoretically feasible when considering statistical models for an experiment. The problem of determination of the more plausible distribution using likelihood procedures (see, for example, Sprott and Kalbfleisch, 1969) will be discussed for the simple case where all observations are made under the same response conditions. (Lindsey, 1974, will consider this problem when independent variables are present.) To do this using likelihood inference, a base statistical model must be introduced with which all other distributions under consideration may be compared. The derivation which follows yields the multinomial model as the base model. Several approaches have been suggested in the literature to the problem of determining which of a number of possible models best describes a set of data. Cox (1961, 1962) develops asymptotic Neyman-Pearson likelihood ratio tests and suggests an alternative approach involving a combination, either additive or multiplicative, of the density functions, with estimation of additional parameters. This approach is further developed by Atkinson (1970). When prior probabilities, both for each model and for the parameters within the models, are available, Lindley (1961, p. 456) gives a posterior odds ratio of the two models using Bayes's theorem. When applicable (i.e. when prior probabilities are available), this approach may be used with the methods developed below."
                },
                {
                    "title": "Breast cancer statistics, 2013",
                    "abstract": "In this article, the American Cancer Society provides an overview of female breast cancer statistics in the United States, including data on incidence, mortality, survival, and screening. Approximately 232,340 new cases of invasive breast cancer and 39,620 breast cancer deaths are expected to occur among US women in 2013. One in 8 women in the United States will develop breast cancer in her lifetime. Breast cancer incidence rates increased slightly among African American women; decreased among Hispanic women; and were stable among whites, Asian Americans/Pacific Islanders, and American Indians/Alaska Natives from 2006 to 2010. Historically, white women have had the highest breast cancer incidence rates among women aged 40 years and older; however, incidence rates are converging among white and African American women, particularly among women aged 50 years to 59 years. Incidence rates increased for estrogen receptor\u2010positive breast cancers in the youngest white women, Hispanic women aged 60 years to 69 years, and all but the oldest African American women. In contrast, estrogen receptor\u2010negative breast cancers declined among most age and racial/ethnic groups. These divergent trends may reflect etiologic heterogeneity and the differing effects of some factors, such as obesity and parity, on risk by tumor subtype. Since 1990, breast cancer death rates have dropped by 34% and this decrease was evident in all racial/ethnic groups except American Indians/Alaska Natives. Nevertheless, survival disparities persist by race/ethnicity, with African American women having the poorest breast cancer survival of any racial/ethnic group. Continued progress in the control of breast cancer will require sustained and increased efforts to provide high\u2010quality screening, diagnosis, and treatment to all segments of the population. CA Cancer J Clin 2014;64:52\u201362. \u00a9 2013 American Cancer Society, Inc."
                },
                {
                    "title": "Statsmodels: Econometric and Statistical Modeling with Python",
                    "abstract": "Statsmodels is a library for statistical and econometric analysis in Python. This paper discusses the current relationship between statistics and Python and open source more generally, outlining how the statsmodels package fills a gap in this relationship. An overview of statsmodels is provided, including a discussion of the overarching design and philosophy, what can be found in the package, and some usage examples. The paper concludes with a look at what the future holds."
                },
                {
                    "title": "Computing the Multinomial Stochastic Complexity in Sub-Linear Time",
                    "abstract": "Stochastic complexity is an objective, information-theoretic criterion for model selection. In this paper we study the stochastic complexity of multinomial variables, which forms an important building block for learning probabilistic graphical models in the discrete data setting. The fastest existing algorithms for computing the multinomial stochastic complexity have the time complexity of O(n), where n is the number of data points, but in this paper we derive sub-linear time algorithms for this task using a finite precision approach. The main idea here is that in practice we do not need exact numbers, but finite floating-point precision is sufficient for typical statistical applications of stochastic complexity. We prove that if we use only finite precision (e.g. double precision) and precomputed sufficient statistics, we can in fact do the computations in sub-linear time with respect to data size and have the overall time complexity of O( \u221a dn + L), where d is precision in digits and L is the number of values of the multinomial variable. We present two fast algorithms based on our results and discuss how these results can be exploited in the task of learning the structure of a probabilistic graphical model."
                },
                {
                    "title": "ANALYZING THE STOCHASTIC COMPLEXITY VIA TREE POLYNOMIALS",
                    "abstract": "To permit formation of a threading element as a closed-loop structure, while being capable of accommodating threading paths of different lengths, the threading element is formed as a sprocket chain which has a long return portion (1) in a return path, coupled to a driven sprocket wheel (2), and connecting portions (3, 4) of a length suitable for the respective threading paths (B, C), and selectively connectable with the return portion (1), to thereby form a closed loop with the selected connecting portion (3 or 4). Pneumatically operated pistons located on path selection switches (5, 6) push the last link of the return portion into engagement with a selected end of the connecting portion, for example by engaging a cross pin (26) in a slotted end link (22, 23; 24, 25) of the connecting portions. The system is entirely automatically controllable, and additional pneumatically operated pistons (7, 10; 8, 11; 9, 12) engage between the links of the link pairs of the respective return and connecting portions of the sprocket chain to hold the respective portions in position during switch-over of the terminal link and, after switch-over, to retain in position that one of the connecting portions which is not connected, for future use."
                },
                {
                    "title": "Catalan Numbers",
                    "abstract": "Sequences and arrays whose terms enumerate combinatorial structures have many applications in computer science. Knowledge (or estimation) of such integer-valued functions is, for example, needed in analyzing the complexity of an algorithm. Familiar examples are the polynomials in n, exponential functions (with an integer base) with exponent a polynomial in n, factorials, and the binomial coefficients. Less familiar are the Stirling numbers considered elsewhere in this book. The sequence of positive integers to be met here, called the Catalan* numbers, enumerate combinatorial structures of many different types. Those include nonnegative paths in the plane, well-formed sequences of parentheses, full binary trees, well-parenthesized products of variables, stack permutations, and triangulations of a convex polygon. After defining the Catalan numbers explicitly by formula, we will show by a combinatorial argument that they count nonnegative paths in the plane. The size of each set of structures subsequently considered is shown to be a Catalan number by establishing a one-to-one correspondence between that set and a set of structures shown earlier to be of that size."
                },
                {
                    "title": "Mixture Density Networks",
                    "abstract": "p(t | x) t x x x = 0.8 = 0.5 = 0.2 Figure 7: Plot of the conditional probability densities of the target data, for various values of x, obtained by taking vertical slices through the contours in Figure 6, for x = 0:2, x = 0:5 and x = 0:8. It is clear that the Mixture Density Network is able to capture correctly the multimodal nature of the target data density function at intermediate values of x. x Figure 8: Plot of the priors i (x) as a function of x for the 3 kernel functions from the same Mixture Density Network as was used to plot Figure 6. At both small and large values of x, where the conditional probability density of the target data is unimodal, only one of the kernels has a prior probability which diiers signiicantly from zero. At intermediate values of x, where the conditional density is tri-modal, the three kernels have comparable priors."
                },
                {
                    "title": "Learning With Continuous Classes",
                    "abstract": ": Some empirical learning tasks are concerned with predicting values rather than the more familiar categories. This paper describes a new system, m5, that constructs tree-based piecewise linear models. Four case studies are presented in which m5 is compared to other methods."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop an intrinsically interpretable tree-based model for conditional density estimation (CDE) that effectively captures complex conditional densities without relying on restrictive assumptions?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of CDE, as it addresses the need for interpretable models that can provide richer insights into data distributions, particularly in critical areas like healthcare. By improving interpretability, researchers and practitioners can make better-informed decisions based on the model outputs. This work could pave the way for future research into other interpretable models and applications across various domains, enhancing the understanding of complex data relationships and potentially leading to more effective interventions in fields such as genomics and environmental science.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in simultaneously optimizing the decision tree structure and the number of bins for histograms at each leaf node, which is a complex combinatorial problem. Naive approaches may fail because they often rely on cross-validation, which is computationally expensive and impractical for large datasets. Additionally, the need to balance model complexity with interpretability adds another layer of difficulty, as overly complex models can obscure the insights that decision trees are meant to provide.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on black-box models, neglecting the development of interpretable tree-based methods for CDE. Existing solutions, like CADET, impose restrictive assumptions (e.g., Gaussian distribution in leaf nodes) that limit their applicability to complex data. Barriers such as the lack of effective optimization techniques for tree structures and histogram binning have hindered progress. Our approach differs by utilizing the minimum description length (MDL) principle to formalize the learning problem, allowing for a more flexible and interpretable model without the need for extensive hyperparameter tuning.\n\n**[Question 5] - What are the key components of my approach and results?**  \nWe propose the Conditional Density Tree (CDTree), which utilizes a histogram model at each leaf node to represent conditional densities. The methodology involves formalizing the learning problem using the MDL principle, which streamlines the optimization process. We will use a dataset related to personal medical costs with demographic features, and our evaluation metric will focus on the interpretability and accuracy of the density estimates. Expected outcomes include a model that not only captures complex conditional distributions effectively but also provides clear, interpretable"
            }
        },
        "author_data": {
            "2e5b1cf4-6bd4-48f0-b8ba-e22a33a46cd5": {
                "pk": "2e5b1cf4-6bd4-48f0-b8ba-e22a33a46cd5",
                "name": "Lincen Yang",
                "collaborators": [
                    "Matthijs van Leeuwen",
                    "Mitra Baratchi",
                    "Alexander Marx"
                ],
                "domain": [
                    "Rule Learning",
                    "Interpretability",
                    "Machine Learning",
                    "Data Mining"
                ],
                "publications": [
                    {
                        "title": "Probabilistic Truly Unordered Rule Sets",
                        "abstract": "Rule set learning has recently been frequently revisited because of its interpretability. Existing methods have several shortcomings though. First, most existing methods impose orders among rules, either explicitly or implicitly, which makes the models less comprehensible. Second, due to the difficulty of handling conflicts caused by overlaps (i.e., instances covered by multiple rules), existing methods often do not consider probabilistic rules. Third, learning classification rules for multi-class target is understudied, as most existing methods focus on binary classification or multi-class classification via the ``one-versus-rest\" approach.   To address these shortcomings, we propose TURS, for Truly Unordered Rule Sets. To resolve conflicts caused by overlapping rules, we propose a novel model that exploits the probabilistic properties of our rule sets, with the intuition of only allowing rules to overlap if they have similar probabilistic outputs. We next formalize the problem of learning a TURS model based on the MDL principle and develop a carefully designed heuristic algorithm. We benchmark against a wide range of rule-based methods and demonstrate that our method learns rule sets that have lower model complexity and highly competitive predictive performance. In addition, we empirically show that rules in our model are empirically ``independent\" and hence truly unordered."
                    },
                    {
                        "title": "Truly Unordered Probabilistic Rule Sets for Multi-class Classification",
                        "abstract": "Rule set learning has long been studied and has recently been frequently revisited due to the need for interpretable models. Still, existing methods have several shortcomings: 1) most recent methods require a binary feature matrix as input, while learning rules directly from numeric variables is understudied; 2) existing methods impose orders among rules, either explicitly or implicitly, which harms interpretability; and 3) currently no method exists for learning probabilistic rule sets for multi-class target variables (there is only one for probabilistic rule lists).   We propose TURS, for Truly Unordered Rule Sets, which addresses these shortcomings. We first formalize the problem of learning truly unordered rule sets. To resolve conflicts caused by overlapping rules, i.e., instances covered by multiple rules, we propose a novel approach that exploits the probabilistic properties of our rule sets. We next develop a two-phase heuristic algorithm that learns rule sets by carefully growing rules. An important innovation is that we use a surrogate score to take the global potential of the rule set into account when learning a local rule.   Finally, we empirically demonstrate that, compared to non-probabilistic and (explicitly or implicitly) ordered state-of-the-art methods, our method learns rule sets that not only have better interpretability but also better predictive performance."
                    },
                    {
                        "title": "Unsupervised Discretization by Two-dimensional MDL-based Histogram",
                        "abstract": "Unsupervised discretization is a crucial step in many knowledge discovery tasks. The state-of-the-art method for one-dimensional data infers locally adaptive histograms using the minimum description length (MDL) principle, but the multi-dimensional case is far less studied: current methods consider the dimensions one at a time (if not independently), which result in discretizations based on rectangular cells of adaptive size. Unfortunately, this approach is unable to adequately characterize dependencies among dimensions and/or results in discretizations consisting of more cells (or bins) than is desirable.   To address this problem, we propose an expressive model class that allows for far more flexible partitions of two-dimensional data. We extend the state of the art for the one-dimensional case to obtain a model selection problem based on the normalized maximum likelihood, a form of refined MDL. As the flexibility of our model class comes at the cost of a vast search space, we introduce a heuristic algorithm, named PALM, which Partitions each dimension ALternately and then Merges neighboring regions, all using the MDL principle. Experiments on synthetic data show that PALM 1) accurately reveals ground truth partitions that are within the model class (i.e., the search space), given a large enough sample size; 2) approximates well a wide range of partitions outside the model class; 3) converges, in contrast to the state-of-the-art multivariate discretization method IPD. Finally, we apply our algorithm to three spatial datasets, and we demonstrate that, compared to kernel density estimation (KDE), our algorithm not only reveals more detailed density changes, but also fits unseen data better, as measured by the log-likelihood."
                    },
                    {
                        "title": "Estimating Conditional Mutual Information for Discrete-Continuous Mixtures using Multi-Dimensional Adaptive Histograms",
                        "abstract": "Estimating conditional mutual information (CMI) is an essential yet challenging step in many machine learning and data mining tasks. Estimating CMI from data that contains both discrete and continuous variables, or even discrete-continuous mixture variables, is a particularly hard problem. In this paper, we show that CMI for such mixture variables, defined based on the Radon-Nikodym derivate, can be written as a sum of entropies, just like CMI for purely discrete or continuous data. Further, we show that CMI can be consistently estimated for discrete-continuous mixture variables by learning an adaptive histogram model. In practice, we estimate such a model by iteratively discretizing the continuous data points in the mixture variables. To evaluate the performance of our estimator, we benchmark it against state-of-the-art CMI estimators as well as evaluate it in a causal discovery setting."
                    }
                ]
            },
            "f0a01ff8-d228-4261-b9c6-3f2e8ffaf2c6": {
                "pk": "f0a01ff8-d228-4261-b9c6-3f2e8ffaf2c6",
                "name": "Matthijs van Leeuwen",
                "collaborators": [
                    "Zhong Li",
                    "Lincen Yang",
                    "Thomas B\u00e4ck",
                    "Hugo M. Proen\u00e7a",
                    "Vladimir Dzyuba",
                    "Sander van Rijn",
                    "Antti Ukkonen",
                    "Micky Faas",
                    "Yuxuan Zhu",
                    "Jiayang Shi"
                ],
                "domain": [
                    "Anomaly Detection",
                    "Pattern Mining",
                    "Explainable AI",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Estimating the pattern frequency spectrum inside the browser",
                        "abstract": "We present a browser application for estimating the number of frequent patterns, in particular itemsets, as well as the pattern frequency spectrum. The pattern frequency spectrum is defined as the function that shows for every value of the frequency threshold $\\sigma$ the number of patterns that are frequent in a given dataset. Our demo implements a recent algorithm proposed by the authors for finding the spectrum. The demo is 100% JavaScript, and runs in all modern browsers. We observe that modern JavaScript engines can deliver performance that makes it viable to run non-trivial data analysis algorithms in browser applications."
                    },
                    {
                        "title": "Vouw: Geometric Pattern Mining using the MDL Principle",
                        "abstract": "We introduce geometric pattern mining, the problem of finding recurring local structure in discrete, geometric matrices. It differs from existing pattern mining problems by identifying complex spatial relations between elements, resulting in arbitrarily shaped patterns. After we formalise this new type of pattern mining, we propose an approach to selecting a set of patterns using the Minimum Description Length principle. We demonstrate the potential of our approach by introducing Vouw, a heuristic algorithm for mining exact geometric patterns. We show that Vouw delivers high-quality results with a synthetic benchmark."
                    },
                    {
                        "title": "Feature Selection for Fault Detection and Prediction based on Event Log Analysis",
                        "abstract": "Event logs are widely used for anomaly detection and prediction in complex systems. Existing log-based anomaly detection methods usually consist of four main steps: log collection, log parsing, feature extraction, and anomaly detection, wherein the feature extraction step extracts useful features for anomaly detection by counting log events. For a complex system, such as a lithography machine consisting of a large number of subsystems, its log may contain thousands of different events, resulting in abounding extracted features. However, when anomaly detection is performed at the subsystem level, analyzing all features becomes expensive and unnecessary. To mitigate this problem, we develop a feature selection method for log-based anomaly detection and prediction, largely improving the effectiveness and efficiency."
                    },
                    {
                        "title": "A Survey on Explainable Anomaly Detection",
                        "abstract": "In the past two decades, most research on anomaly detection has focused on improving the accuracy of the detection, while largely ignoring the explainability of the corresponding methods and thus leaving the explanation of outcomes to practitioners. As anomaly detection algorithms are increasingly used in safety-critical domains, providing explanations for the high-stakes decisions made in those domains has become an ethical and regulatory requirement. Therefore, this work provides a comprehensive and structured survey on state-of-the-art explainable anomaly detection techniques. We propose a taxonomy based on the main aspects that characterize each explainable anomaly detection technique, aiming to help practitioners and researchers find the explainable anomaly detection method that best suits their needs."
                    },
                    {
                        "title": "Explainable Contextual Anomaly Detection using Quantile Regression Forests",
                        "abstract": "Traditional anomaly detection methods aim to identify objects that deviate from most other objects by treating all features equally. In contrast, contextual anomaly detection methods aim to detect objects that deviate from other objects within a context of similar objects by dividing the features into contextual features and behavioral features. In this paper, we develop connections between dependency-based traditional anomaly detection methods and contextual anomaly detection methods. Based on resulting insights, we propose a novel approach to inherently interpretable contextual anomaly detection that uses Quantile Regression Forests to model dependencies between features. Extensive experiments on various synthetic and real-world datasets demonstrate that our method outperforms state-of-the-art anomaly detection methods in identifying contextual anomalies in terms of accuracy and interpretability."
                    },
                    {
                        "title": "Probabilistic Truly Unordered Rule Sets",
                        "abstract": "Rule set learning has recently been frequently revisited because of its interpretability. Existing methods have several shortcomings though. First, most existing methods impose orders among rules, either explicitly or implicitly, which makes the models less comprehensible. Second, due to the difficulty of handling conflicts caused by overlaps (i.e., instances covered by multiple rules), existing methods often do not consider probabilistic rules. Third, learning classification rules for multi-class target is understudied, as most existing methods focus on binary classification or multi-class classification via the ``one-versus-rest\" approach.   To address these shortcomings, we propose TURS, for Truly Unordered Rule Sets. To resolve conflicts caused by overlapping rules, we propose a novel model that exploits the probabilistic properties of our rule sets, with the intuition of only allowing rules to overlap if they have similar probabilistic outputs. We next formalize the problem of learning a TURS model based on the MDL principle and develop a carefully designed heuristic algorithm. We benchmark against a wide range of rule-based methods and demonstrate that our method learns rule sets that have lower model complexity and highly competitive predictive performance. In addition, we empirically show that rules in our model are empirically ``independent\" and hence truly unordered."
                    },
                    {
                        "title": "Interpretable multiclass classification by MDL-based rule lists",
                        "abstract": "Interpretable classifiers have recently witnessed an increase in attention from the data mining community because they are inherently easier to understand and explain than their more complex counterparts. Examples of interpretable classification models include decision trees, rule sets, and rule lists. Learning such models often involves optimizing hyperparameters, which typically requires substantial amounts of data and may result in relatively large models. In this paper, we consider the problem of learning compact yet accurate probabilistic rule lists for multiclass classification. Specifically, we propose a novel formalization based on probabilistic rule lists and the minimum description length (MDL) principle. This results in virtually parameter-free model selection that naturally allows to trade-off model complexity with goodness of fit, by which overfitting and the need for hyperparameter tuning are effectively avoided. Finally, we introduce the Classy algorithm, which greedily finds rule lists according to the proposed criterion. We empirically demonstrate that Classy selects small probabilistic rule lists that outperform state-of-the-art classifiers when it comes to the combination of predictive performance and interpretability. We show that Classy is insensitive to its only parameter, i.e., the candidate set, and that compression on the training set correlates with classification performance, validating our MDL-based selection criterion."
                    },
                    {
                        "title": "Learning what matters - Sampling interesting patterns",
                        "abstract": "In the field of exploratory data mining, local structure in data can be described by patterns and discovered by mining algorithms. Although many solutions have been proposed to address the redundancy problems in pattern mining, most of them either provide succinct pattern sets or take the interests of the user into account-but not both. Consequently, the analyst has to invest substantial effort in identifying those patterns that are relevant to her specific interests and goals. To address this problem, we propose a novel approach that combines pattern sampling with interactive data mining. In particular, we introduce the LetSIP algorithm, which builds upon recent advances in 1) weighted sampling in SAT and 2) learning to rank in interactive pattern mining. Specifically, it exploits user feedback to directly learn the parameters of the sampling distribution that represents the user's interests. We compare the performance of the proposed algorithm to the state-of-the-art in interactive pattern mining by emulating the interests of a user. The resulting system allows efficient and interleaved learning and sampling, thus user-specific anytime data exploration. Finally, LetSIP demonstrates favourable trade-offs concerning both quality-diversity and exploitation-exploration when compared to existing methods."
                    },
                    {
                        "title": "Truly Unordered Probabilistic Rule Sets for Multi-class Classification",
                        "abstract": "Rule set learning has long been studied and has recently been frequently revisited due to the need for interpretable models. Still, existing methods have several shortcomings: 1) most recent methods require a binary feature matrix as input, while learning rules directly from numeric variables is understudied; 2) existing methods impose orders among rules, either explicitly or implicitly, which harms interpretability; and 3) currently no method exists for learning probabilistic rule sets for multi-class target variables (there is only one for probabilistic rule lists).   We propose TURS, for Truly Unordered Rule Sets, which addresses these shortcomings. We first formalize the problem of learning truly unordered rule sets. To resolve conflicts caused by overlapping rules, i.e., instances covered by multiple rules, we propose a novel approach that exploits the probabilistic properties of our rule sets. We next develop a two-phase heuristic algorithm that learns rule sets by carefully growing rules. An important innovation is that we use a surrogate score to take the global potential of the rule set into account when learning a local rule.   Finally, we empirically demonstrate that, compared to non-probabilistic and (explicitly or implicitly) ordered state-of-the-art methods, our method learns rule sets that not only have better interpretability but also better predictive performance."
                    },
                    {
                        "title": "Graph Neural Networks based Log Anomaly Detection and Explanation",
                        "abstract": "Event logs are widely used to record the status of high-tech systems, making log anomaly detection important for monitoring those systems. Most existing log anomaly detection methods take a log event count matrix or log event sequences as input, exploiting quantitative and/or sequential relationships between log events to detect anomalies. Unfortunately, only considering quantitative or sequential relationships may result in low detection accuracy. To alleviate this problem, we propose a graph-based method for unsupervised log anomaly detection, dubbed Logs2Graphs, which first converts event logs into attributed, directed, and weighted graphs, and then leverages graph neural networks to perform graph-level anomaly detection. Specifically, we introduce One-Class Digraph Inception Convolutional Networks, abbreviated as OCDiGCN, a novel graph neural network model for detecting graph-level anomalies in a collection of attributed, directed, and weighted graphs. By coupling the graph representation and anomaly detection steps, OCDiGCN can learn a representation that is especially suited for anomaly detection, resulting in a high detection accuracy. Importantly, for each identified anomaly, we additionally provide a small subset of nodes that play a crucial role in OCDiGCN's prediction as explanations, which can offer valuable cues for subsequent root cause diagnosis. Experiments on five benchmark datasets show that Logs2Graphs performs at least on par with state-of-the-art log anomaly detection methods on simple datasets while largely outperforming state-of-the-art log anomaly detection methods on complicated datasets."
                    },
                    {
                        "title": "Estimating Conditional Mutual Information for Discrete-Continuous Mixtures using Multi-Dimensional Adaptive Histograms",
                        "abstract": "Estimating conditional mutual information (CMI) is an essential yet challenging step in many machine learning and data mining tasks. Estimating CMI from data that contains both discrete and continuous variables, or even discrete-continuous mixture variables, is a particularly hard problem. In this paper, we show that CMI for such mixture variables, defined based on the Radon-Nikodym derivate, can be written as a sum of entropies, just like CMI for purely discrete or continuous data. Further, we show that CMI can be consistently estimated for discrete-continuous mixture variables by learning an adaptive histogram model. In practice, we estimate such a model by iteratively discretizing the continuous data points in the mixture variables. To evaluate the performance of our estimator, we benchmark it against state-of-the-art CMI estimators as well as evaluate it in a causal discovery setting."
                    },
                    {
                        "title": "Flexible constrained sampling with guarantees for pattern mining",
                        "abstract": "Pattern sampling has been proposed as a potential solution to the infamous pattern explosion. Instead of enumerating all patterns that satisfy the constraints, individual patterns are sampled proportional to a given quality measure. Several sampling algorithms have been proposed, but each of them has its limitations when it comes to 1) flexibility in terms of quality measures and constraints that can be used, and/or 2) guarantees with respect to sampling accuracy. We therefore present Flexics, the first flexible pattern sampler that supports a broad class of quality measures and constraints, while providing strong guarantees regarding sampling accuracy. To achieve this, we leverage the perspective on pattern mining as a constraint satisfaction problem and build upon the latest advances in sampling solutions in SAT as well as existing pattern mining algorithms. Furthermore, the proposed algorithm is applicable to a variety of pattern languages, which allows us to introduce and tackle the novel task of sampling sets of patterns. We introduce and empirically evaluate two variants of Flexics: 1) a generic variant that addresses the well-known itemset sampling task and the novel pattern set sampling task as well as a wide range of expressive constraints within these tasks, and 2) a specialized variant that exploits existing frequent itemset techniques to achieve substantial speed-ups. Experiments show that Flexics is both accurate and efficient, making it a useful tool for pattern-based data exploration."
                    },
                    {
                        "title": "Unsupervised Discretization by Two-dimensional MDL-based Histogram",
                        "abstract": "Unsupervised discretization is a crucial step in many knowledge discovery tasks. The state-of-the-art method for one-dimensional data infers locally adaptive histograms using the minimum description length (MDL) principle, but the multi-dimensional case is far less studied: current methods consider the dimensions one at a time (if not independently), which result in discretizations based on rectangular cells of adaptive size. Unfortunately, this approach is unable to adequately characterize dependencies among dimensions and/or results in discretizations consisting of more cells (or bins) than is desirable.   To address this problem, we propose an expressive model class that allows for far more flexible partitions of two-dimensional data. We extend the state of the art for the one-dimensional case to obtain a model selection problem based on the normalized maximum likelihood, a form of refined MDL. As the flexibility of our model class comes at the cost of a vast search space, we introduce a heuristic algorithm, named PALM, which Partitions each dimension ALternately and then Merges neighboring regions, all using the MDL principle. Experiments on synthetic data show that PALM 1) accurately reveals ground truth partitions that are within the model class (i.e., the search space), given a large enough sample size; 2) approximates well a wide range of partitions outside the model class; 3) converges, in contrast to the state-of-the-art multivariate discretization method IPD. Finally, we apply our algorithm to three spatial datasets, and we demonstrate that, compared to kernel density estimation (KDE), our algorithm not only reveals more detailed density changes, but also fits unseen data better, as measured by the log-likelihood."
                    },
                    {
                        "title": "WEARDA: Recording Wearable Sensor Data for Human Activity Monitoring",
                        "abstract": "We present WEARDA, the open source WEARable sensor Data Acquisition software package. WEARDA facilitates the acquisition of human activity data with smartwatches and is primarily aimed at researchers who require transparency, full control, and access to raw sensor data. It provides functionality to simultaneously record raw data from four sensors -- tri-axis accelerometer, tri-axis gyroscope, barometer, and GPS -- which should enable researchers to, for example, estimate energy expenditure and mine movement trajectories. A Samsung smartwatch running the Tizen OS was chosen because of 1) the required functionalities of the smartwatch software API, 2) the availability of software development tools and accessible documentation, 3) having the required sensors, and 4) the requirements on case design for acceptance by the target user group. WEARDA addresses five practical challenges concerning preparation, measurement, logistics, privacy preservation, and reproducibility to ensure efficient and errorless data collection. The software package was initially created for the project \"Dementia back at the heart of the community\", and has been successfully used in that context."
                    },
                    {
                        "title": "Local Subspace-Based Outlier Detection using Global Neighbourhoods",
                        "abstract": "Outlier detection in high-dimensional data is a challenging yet important task, as it has applications in, e.g., fraud detection and quality control. State-of-the-art density-based algorithms perform well because they 1) take the local neighbourhoods of data points into account and 2) consider feature subspaces. In highly complex and high-dimensional data, however, existing methods are likely to overlook important outliers because they do not explicitly take into account that the data is often a mixture distribution of multiple components.   We therefore introduce GLOSS, an algorithm that performs local subspace outlier detection using global neighbourhoods. Experiments on synthetic data demonstrate that GLOSS more accurately detects local outliers in mixed data than its competitors. Moreover, experiments on real-world data show that our approach identifies relevant outliers overlooked by existing methods, confirming that one should keep an eye on the global perspective even when doing local outlier detection."
                    },
                    {
                        "title": "Evolving the Structure of Evolution Strategies",
                        "abstract": "Various variants of the well known Covariance Matrix Adaptation Evolution Strategy (CMA-ES) have been proposed recently, which improve the empirical performance of the original algorithm by structural modifications. However, in practice it is often unclear which variation is best suited to the specific optimization problem at hand. As one approach to tackle this issue, algorithmic mechanisms attached to CMA-ES variants are considered and extracted as functional \\emph{modules}, allowing for combinations of them. This leads to a configuration space over ES structures, which enables the exploration of algorithm structures and paves the way toward novel algorithm generation. Specifically, eleven modules are incorporated in this framework with two or three alternative configurations for each module, resulting in $4\\,608$ algorithms. A self-adaptive Genetic Algorithm (GA) is used to efficiently evolve effective ES-structures for given classes of optimization problems, outperforming any classical CMA-ES variants from literature. The proposed approach is evaluated on noiseless functions from BBOB suite. Furthermore, such an observation is again confirmed on different function groups and dimensionality, indicating the feasibility of ES configuration on real-world problem classes."
                    },
                    {
                        "title": "Finding Efficient Trade-offs in Multi-Fidelity Response Surface Modeling",
                        "abstract": "In the context of optimization approaches to engineering applications, time-consuming simulations are often utilized which can be configured to deliver solutions for various levels of accuracy, commonly referred to as different fidelity levels. It is common practice to train hierarchical surrogate models on the objective functions in order to speed-up the optimization process. These operate under the assumption that there is a correlation between the high- and low-fidelity versions of the problem that can be exploited to cheaply gain information. In the practical scenario where the computational budget has to be allocated between multiple fidelities, limited guidelines are available to help make that division. In this paper we evaluate a range of different choices for a two-fidelity setup that provide helpful intuitions about the trade-off between evaluating in high- or low-fidelity. We present a heuristic method based on subsampling from an initial Design of Experiments (DoE) to find a suitable division of the computational budget between the fidelity levels. This enables the setup of multi-fidelity optimizations which utilize the available computational budget efficiently, independent of the multi-fidelity model used."
                    },
                    {
                        "title": "Discovering outstanding subgroup lists for numeric targets using MDL",
                        "abstract": "The task of subgroup discovery (SD) is to find interpretable descriptions of subsets of a dataset that stand out with respect to a target attribute. To address the problem of mining large numbers of redundant subgroups, subgroup set discovery (SSD) has been proposed. State-of-the-art SSD methods have their limitations though, as they typically heavily rely on heuristics and/or user-chosen hyperparameters.   We propose a dispersion-aware problem formulation for subgroup set discovery that is based on the minimum description length (MDL) principle and subgroup lists. We argue that the best subgroup list is the one that best summarizes the data given the overall distribution of the target. We restrict our focus to a single numeric target variable and show that our formalization coincides with an existing quality measure when finding a single subgroup, but that-in addition-it allows to trade off subgroup quality with the complexity of the subgroup. We next propose SSD++, a heuristic algorithm for which we empirically demonstrate that it returns outstanding subgroup lists: non-redundant sets of compact subgroups that stand out by having strongly deviating means and small spread."
                    },
                    {
                        "title": "Explainable Graph Neural Networks Under Fire",
                        "abstract": "Predictions made by graph neural networks (GNNs) usually lack interpretability due to their complex computational behavior and the abstract nature of graphs. In an attempt to tackle this, many GNN explanation methods have emerged. Their goal is to explain a model's predictions and thereby obtain trust when GNN models are deployed in decision critical applications. Most GNN explanation methods work in a post-hoc manner and provide explanations in the form of a small subset of important edges and/or nodes. In this paper we demonstrate that these explanations can unfortunately not be trusted, as common GNN explanation methods turn out to be highly susceptible to adversarial perturbations. That is, even small perturbations of the original graph structure that preserve the model's predictions may yield drastically different explanations. This calls into question the trustworthiness and practical utility of post-hoc explanation methods for GNNs. To be able to attack GNN explanation models, we devise a novel attack method dubbed \\textit{GXAttack}, the first \\textit{optimization-based} adversarial white-box attack method for post-hoc GNN explanations under such settings. Due to the devastating effectiveness of our attack, we call for an adversarial evaluation of future GNN explainers to demonstrate their robustness. For reproducibility, our code is available via GitHub."
                    }
                ]
            }
        }
    },
    "2404.04286": {
        "paper_data": {
            "title": "Bias Amplification in Language Model Evolution: An Iterated Learning Perspective",
            "url": "http://arxiv.org/abs/2404.04286v2",
            "arxiv_id": "2404.04286",
            "authors": [
                "Yi Ren",
                "Shangmin Guo",
                "Linlu Qiu",
                "Bailin Wang",
                "Danica J. Sutherland"
            ],
            "abstract": "With the widespread adoption of Large Language Models (LLMs), the prevalence of iterative interactions among these models is anticipated to increase. Notably, recent advancements in multi-round self-improving methods allow LLMs to generate new examples for training subsequent models. At the same time, multi-agent LLM systems, involving automated interactions among agents, are also increasing in prominence. Thus, in both short and long terms, LLMs may actively engage in an evolutionary process. We draw parallels between the behavior of LLMs and the evolution of human culture, as the latter has been extensively studied by cognitive scientists for decades. Our approach involves leveraging Iterated Learning (IL), a Bayesian framework that elucidates how subtle biases are magnified during human cultural evolution, to explain some behaviors of LLMs. This paper outlines key characteristics of agents' behavior in the Bayesian-IL framework, including predictions that are supported by experimental verification with various LLMs. This theoretical framework could help to more effectively predict and guide the evolution of LLMs in desired directions.",
            "introduction": " Introduction Recent large language models (LLMs) have shown remarkable instruction-following ability and an increasing number of applications; it is thus reasonable to expect they are likely to become more widespread. Moreover, interactions between LLMs (either multiple models, or different generations of the same model) may also become very commonplace in the near future. In fact, many recent works consider iterative on-policy self-data-augmentation solutions to break through the bottleneck of human-generated supervisions, e.g., self-instruct (Y . Wang et al. 2022), self-refine (Madaan et al. 2023), hypothesis refinement (Qiu et al. 2024), self-distill (C. Xu et al. 2023), self-instruct (Y . Wang et al. 2022), self-reward (Weizhe et al. 2024), self-feedback (W. Xu et al. 2024), RAFT (Dong et al. 2023), ReST (Gulcehre et al. 2023), iterated DPO (Xiong, Dong, Ye, Zhong, et al. 2023), OAIF (Guo et al. 2024), SPIN (Z. Chen et al. 2024), and many more. Whether the model\u2019s knowledge is updated through in-weight or in-context mechanisms, these Background and Related Work 2.1 Iterated Learning Iterated learning (IL) is a hypothetical procedure to simulate how the tendency of specific properties of human culture or language gradually emerges and becomes dominant. It is based on studying the behaviors of a chain of intelligent agents. From the perspective of an individual agent, the process involves initially acquiring knowledge from its predecessor ( imitation ), refining its beliefs while using them to conduct tasks ( interaction with the world), and subsequently imparting its knowledge to the agents in the next generation ( transmission ). Cognitive scientists have applied this framework to explain various evolutionary phenomena of human society, including the emergence of compositionality in human language (Kirby et al. 2015), patterns in human object categorisation (T. L. Griffiths et al. 2008), and the evolution of color naming systems (Carlsson et al. 2023). The framework has also seen recent success with neural network agents, including in emergent communication (Guo et al. 2019; Ren et al. 2020), machine translation (Y . Lu et al. 2020), visual question answering (Vani et al. 2021), large vision-language models (Chenhao et al. 2024), and representation learning (Ren et al. 2023), indicative that this framework could also be useful for more general deep learning systems, like LLMs. 2.2 Related On-policy Self-data-augmentation Methods A.1 Assumptions of Bayesian-IL and practical Scenarios As typical in theoretical machine learning research, some assumptions are needed to prove Appendix D and E). The phenomena claimed by the theory still hold. \u2022The Bayesian learning assumption, i.e., Plm(h) =P(h|d)\u221dp(d|h)P0(h). Although this assumption is necessary for drawing a parallel between iterated learning and the EM algorithm and hence getting a guarantee for the amplified bias, the practical in-weights learning (IWL for short) method usually does not strictly follow this assumption, because people usually early stop the training before the model perfectly learns all dt. However, abstract with varying interpretations, such as the level of conciseness, helpfulness, or even the writer\u2019s preference for using rhyme. Consider a general Bayesian agent whose behavior can be depicted by two basic procedures: learning andsampling . Learning involves updating the agent\u2019s knowledge based on observations, while sampling is a procedure wherein the agent generates data based on its knowledge. In this context, the agent\u2019s knowledge is encapsulated by its posterior over the hypotheses, i.e., Plm(h). In Bayesian learning, we assume the agent holds a prior",
            "references": [
                {
                    "title": "Learning Dynamics of LLM Finetuning",
                    "abstract": "Learning dynamics, which describes how the learning of specific training examples influences the model's predictions on other examples, gives us a powerful tool for understanding the behavior of deep learning systems. We study the learning dynamics of large language models during different types of finetuning, by analyzing the step-wise decomposition of how influence accumulates among different potential responses. Our framework allows a uniform interpretation of many interesting observations about the training of popular algorithms for both instruction tuning and preference tuning. In particular, we propose a hypothetical explanation of why specific types of hallucination are strengthened after finetuning, e.g., the model might use phrases or facts in the response for question B to answer question A, or the model might keep repeating similar simple phrases when generating responses. We also extend our framework and highlight a unique\"squeezing effect\"to explain a previously observed phenomenon in off-policy direct preference optimization (DPO), where running DPO for too long makes even the desired outputs less likely. This framework also provides insights into where the benefits of on-policy DPO and other variants come from. The analysis not only provides a novel perspective of understanding LLM's finetuning but also inspires a simple, effective method to improve alignment performance."
                },
                {
                    "title": "Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators",
                    "abstract": "LLM-based auto-annotators have become a key component of the LLM development process due to their cost-effectiveness and scalability compared to human-based evaluation. However, these auto-annotators can introduce complex biases that are hard to remove. Even simple, known confounders such as preference for longer outputs remain in existing automated evaluation metrics. We propose a simple regression analysis approach for controlling biases in auto-evaluations. As a real case study, we focus on reducing the length bias of AlpacaEval, a fast and affordable benchmark for chat LLMs that uses LLMs to estimate response quality. Despite being highly correlated with human preferences, AlpacaEval is known to favor models that generate longer outputs. We introduce a length-controlled AlpacaEval that aims to answer the counterfactual question:\"What would the preference be if the model's and baseline's output had the same length?\". To achieve this, we first fit a generalized linear model to predict the biased output of interest (auto-annotator preferences) based on the mediators we want to control for (length difference) and other relevant features. We then obtain length-controlled preferences by predicting preferences while conditioning the GLM with a zero difference in lengths. Length-controlling not only improves the robustness of the metric to manipulations in model verbosity, we also find that it increases the Spearman correlation with LMSYS' Chatbot Arena from 0.94 to 0.98. We release the code and leaderboard at https://tatsu-lab.github.io/alpaca_eval/ ."
                },
                {
                    "title": "Iterated Learning Improves Compositionality in Large Vision-Language Models",
                    "abstract": "A fundamental characteristic common to both human vision and natural language is their compositional nature. Yet, despite the performance gains contributed by large vision and language pretraining, recent investigations find that most-if not all-our state-of-the-art vision-language models struggle at compositionality. They are unable to distinguish between images of \u201ca girl in white facing a man in black\u201d and \u201ca girl in blackfacing a man in white\u201d. More-over, prior work suggests that compositionality doesn't arise with scale: larger model sizes or training data don't help. This paper develops a new iterated training algorithm that incentivizes compositionality. We draw on decades of cognitive science research that identifies cultural transmission-the need to teach a new generation-as a necessary inductive prior that incentivizes humans to develop compositional languages. Specifically, we reframe vision-language contrastive learning as the Lewis Signaling Game between a vision agent and a language agent, and ope rationalize cultural transmission by iteratively resetting one of the agent's weights during training. After every iteration, this training paradigm induces representations that become \u201ceasier to learn\u201d, a property of compositional languages: e.g. our model trained on CC3M and CC12M improves standard CLIP by 4.7%, 4.0% respectfully in the SugarCrepe benchmark."
                },
                {
                    "title": "Direct Language Model Alignment from Online AI Feedback",
                    "abstract": "Direct alignment from preferences (DAP) methods, such as DPO, have recently emerged as efficient alternatives to reinforcement learning from human feedback (RLHF), that do not require a separate reward model. However, the preference datasets used in DAP methods are usually collected ahead of training and never updated, thus the feedback is purely offline. Moreover, responses in these datasets are often sampled from a language model distinct from the one being aligned, and since the model evolves over training, the alignment phase is inevitably off-policy. In this study, we posit that online feedback is key and improves DAP methods. Our method, online AI feedback (OAIF), uses an LLM as annotator: on each training iteration, we sample two responses from the current model and prompt the LLM annotator to choose which one is preferred, thus providing online feedback. Despite its simplicity, we demonstrate via human evaluation in several tasks that OAIF outperforms both offline DAP and RLHF methods. We further show that the feedback leveraged in OAIF is easily controllable, via instruction prompts to the LLM annotator."
                },
                {
                    "title": "Self-Rewarding Language Models",
                    "abstract": "We posit that to achieve superhuman agents, future models require superhuman feedback in order to provide an adequate training signal. Current approaches commonly train reward models from human preferences, which may then be bottlenecked by human performance level, and secondly these separate frozen reward models cannot then learn to improve during LLM training. In this work, we study Self-Rewarding Language Models, where the language model itself is used via LLM-as-a-Judge prompting to provide its own rewards during training. We show that during Iterative DPO training that not only does instruction following ability improve, but also the ability to provide high-quality rewards to itself. Fine-tuning Llama 2 70B on three iterations of our approach yields a model that outperforms many existing systems on the AlpacaEval 2.0 leaderboard, including Claude 2, Gemini Pro, and GPT-4 0613. While there is much left still to explore, this work opens the door to the possibility of models that can continually improve in both axes."
                },
                {
                    "title": "Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models",
                    "abstract": "Harnessing the power of human-annotated data through Supervised Fine-Tuning (SFT) is pivotal for advancing Large Language Models (LLMs). In this paper, we delve into the prospect of growing a strong LLM out of a weak one without the need for acquiring additional human-annotated data. We propose a new fine-tuning method called Self-Play fIne-tuNing (SPIN), which starts from a supervised fine-tuned model. At the heart of SPIN lies a self-play mechanism, where the LLM refines its capability by playing against instances of itself. More specifically, the LLM generates its own training data from its previous iterations, refining its policy by discerning these self-generated responses from those obtained from human-annotated data. Our method progressively elevates the LLM from a nascent model to a formidable one, unlocking the full potential of human-annotated demonstration data for SFT. Theoretically, we prove that the global optimum to the training objective function of our method is achieved only when the LLM policy aligns with the target data distribution. Empirically, we evaluate our method on several benchmark datasets including the HuggingFace Open LLM Leaderboard, MT-Bench, and datasets from Big-Bench. Our results show that SPIN can significantly improve the LLM's performance across a variety of benchmarks and even outperform models trained through direct preference optimization (DPO) supplemented with extra GPT-4 preference data. This sheds light on the promise of self-play, enabling the achievement of human-level performance in LLMs without the need for expert opponents. Codes are available at https://github.com/uclaml/SPIN."
                },
                {
                    "title": "Iterative Preference Learning from Human Feedback: Bridging Theory and Practice for RLHF under KL-constraint",
                    "abstract": "This paper studies the alignment process of generative models with Reinforcement Learning from Human Feedback (RLHF). We first identify the primary challenges of existing popular methods like offline PPO and offline DPO as lacking in strategical exploration of the environment. Then, to understand the mathematical principle of RLHF, we consider a standard mathematical formulation, the reverse-KL regularized contextual bandit for RLHF. Despite its widespread practical application, a rigorous theoretical analysis of this formulation remains open. We investigate its behavior in three distinct settings -- offline, online, and hybrid -- and propose efficient algorithms with finite-sample theoretical guarantees. Moving towards practical applications, our framework, with a robust approximation of the information-theoretical policy improvement oracle, naturally gives rise to several novel RLHF algorithms. This includes an iterative version of the Direct Preference Optimization (DPO) algorithm for online settings, and a multi-step rejection sampling strategy for offline scenarios. Our empirical evaluations on real-world alignment experiment of large language model demonstrate that these proposed methods significantly surpass existing strong baselines, such as DPO and Rejection Sampling Optimization (RSO), showcasing the connections between solid theoretical foundations and their potent practical implementations."
                },
                {
                    "title": "Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision",
                    "abstract": "Widely used alignment techniques, such as reinforcement learning from human feedback (RLHF), rely on the ability of humans to supervise model behavior - for example, to evaluate whether a model faithfully followed instructions or generated safe outputs. However, future superhuman models will behave in complex ways too difficult for humans to reliably evaluate; humans will only be able to weakly supervise superhuman models. We study an analogy to this problem: can weak model supervision elicit the full capabilities of a much stronger model? We test this using a range of pretrained language models in the GPT-4 family on natural language processing (NLP), chess, and reward modeling tasks. We find that when we naively finetune strong pretrained models on labels generated by a weak model, they consistently perform better than their weak supervisors, a phenomenon we call weak-to-strong generalization. However, we are still far from recovering the full capabilities of strong models with naive finetuning alone, suggesting that techniques like RLHF may scale poorly to superhuman models without further work. We find that simple methods can often significantly improve weak-to-strong generalization: for example, when finetuning GPT-4 with a GPT-2-level supervisor and an auxiliary confidence loss, we can recover close to GPT-3.5-level performance on NLP tasks. Our results suggest that it is feasible to make empirical progress today on a fundamental challenge of aligning superhuman models."
                },
                {
                    "title": "Bayes in the age of intelligent machines",
                    "abstract": "The success of methods based on artificial neural networks in creating intelligent machines seems like it might pose a challenge to explanations of human cognition in terms of Bayesian inference. We argue that this is not the case, and that in fact these systems offer new opportunities for Bayesian modeling. Specifically, we argue that Bayesian models of cognition and artificial neural networks lie at different levels of analysis and are complementary modeling approaches, together offering a way to understand human cognition that spans these levels. We also argue that the same perspective can be applied to intelligent machines, where a Bayesian approach may be uniquely valuable in understanding the behavior of large, opaque artificial neural networks that are trained on proprietary data."
                },
                {
                    "title": "Phenomenal Yet Puzzling: Testing Inductive Reasoning Capabilities of Language Models with Hypothesis Refinement",
                    "abstract": "The ability to derive underlying principles from a handful of observations and then generalize to novel situations -- known as inductive reasoning -- is central to human intelligence. Prior work suggests that language models (LMs) often fall short on inductive reasoning, despite achieving impressive success on research benchmarks. In this work, we conduct a systematic study of the inductive reasoning capabilities of LMs through iterative hypothesis refinement, a technique that more closely mirrors the human inductive process than standard input-output prompting. Iterative hypothesis refinement employs a three-step process: proposing, selecting, and refining hypotheses in the form of textual rules. By examining the intermediate rules, we observe that LMs are phenomenal hypothesis proposers (i.e., generating candidate rules), and when coupled with a (task-specific) symbolic interpreter that is able to systematically filter the proposed set of rules, this hybrid approach achieves strong results across inductive reasoning benchmarks that require inducing causal relations, language-like instructions, and symbolic concepts. However, they also behave as puzzling inductive reasoners, showing notable performance gaps between rule induction (i.e., identifying plausible rules) and rule application (i.e., applying proposed rules to instances), suggesting that LMs are proposing hypotheses without being able to actually apply the rules. Through empirical and human analyses, we further reveal several discrepancies between the inductive reasoning processes of LMs and humans, shedding light on both the potentials and limitations of using LMs in inductive reasoning tasks."
                },
                {
                    "title": "Embers of Autoregression: Understanding Large Language Models Through the Problem They are Trained to Solve",
                    "abstract": "The widespread adoption of large language models (LLMs) makes it important to recognize their strengths and limitations. We argue that in order to develop a holistic understanding of these systems we need to consider the problem that they were trained to solve: next-word prediction over Internet text. By recognizing the pressures that this task exerts we can make predictions about the strategies that LLMs will adopt, allowing us to reason about when they will succeed or fail. This approach - which we call the teleological approach - leads us to identify three factors that we hypothesize will influence LLM accuracy: the probability of the task to be performed, the probability of the target output, and the probability of the provided input. We predict that LLMs will achieve higher accuracy when these probabilities are high than when they are low - even in deterministic settings where probability should not matter. To test our predictions, we evaluate two LLMs (GPT-3.5 and GPT-4) on eleven tasks, and we find robust evidence that LLMs are influenced by probability in the ways that we have hypothesized. In many cases, the experiments reveal surprising failure modes. For instance, GPT-4's accuracy at decoding a simple cipher is 51% when the output is a high-probability word sequence but only 13% when it is low-probability. These results show that AI practitioners should be careful about using LLMs in low-probability situations. More broadly, we conclude that we should not evaluate LLMs as if they are humans but should instead treat them as a distinct type of system - one that has been shaped by its own particular set of pressures."
                },
                {
                    "title": "Reinforced Self-Training (ReST) for Language Modeling",
                    "abstract": "Reinforcement learning from human feedback (RLHF) can improve the quality of large language model's (LLM) outputs by aligning them with human preferences. We propose a simple algorithm for aligning LLMs with human preferences inspired by growing batch reinforcement learning (RL), which we call Reinforced Self-Training (ReST). Given an initial LLM policy, ReST produces a dataset by generating samples from the policy, which are then used to improve the LLM policy using offline RL algorithms. ReST is more efficient than typical online RLHF methods because the training dataset is produced offline, which allows data reuse. While ReST is a general approach applicable to all generative learning settings, we focus on its application to machine translation. Our results show that ReST can substantially improve translation quality, as measured by automated metrics and human evaluation on machine translation benchmarks in a compute and sample-efficient manner."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks",
                    "abstract": "The impressive performance of recent language models across a wide range of tasks suggests that they possess a degree of abstract reasoning skills. Are these skills general and transferable, or specialized to specific tasks seen during pretraining? To disentangle these effects, we propose an evaluation framework based on \u201ccounterfactual\u201d task variants that deviate from the default assumptions underlying standard tasks. Across a suite of 11 tasks, we observe nontrivial performance on the counterfactual variants, but nevertheless find that performance substantially and consistently degrades compared to the default conditions. This suggests that while current LMs may possess abstract task-solving skills to an extent, they often also rely on narrow, non-transferable procedures for task-solving. These results motivate a more careful interpretation of language model performance that teases apart these aspects."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment",
                    "abstract": "Generative foundation models are susceptible to implicit biases that can arise from extensive unsupervised training data. Such biases can produce suboptimal samples, skewed outcomes, and unfairness, with potentially serious consequences. Consequently, aligning these models with human ethics and preferences is an essential step toward ensuring their responsible and effective deployment in real-world applications. Prior research has primarily employed Reinforcement Learning from Human Feedback (RLHF) to address this problem, where generative models are fine-tuned with RL algorithms guided by a human-feedback-informed reward model. However, the inefficiencies and instabilities associated with RL algorithms frequently present substantial obstacles to the successful alignment, necessitating the development of a more robust and streamlined approach. To this end, we introduce a new framework, Reward rAnked FineTuning (RAFT), designed to align generative models effectively. Utilizing a reward model and a sufficient number of samples, our approach selects the high-quality samples, discarding those that exhibit undesired behavior, and subsequently enhancing the model by fine-tuning on these filtered samples. Our studies show that RAFT can effectively improve the model performance in both reward learning and other automated metrics in both large language models and diffusion models."
                },
                {
                    "title": "Self-Refine: Iterative Refinement with Self-Feedback",
                    "abstract": "Like humans, large language models (LLMs) do not always generate the best output on their first try. Motivated by how humans refine their written text, we introduce Self-Refine, an approach for improving initial outputs from LLMs through iterative feedback and refinement. The main idea is to generate an initial output using an LLMs; then, the same LLMs provides feedback for its output and uses it to refine itself, iteratively. Self-Refine does not require any supervised training data, additional training, or reinforcement learning, and instead uses a single LLM as the generator, refiner, and feedback provider. We evaluate Self-Refine across 7 diverse tasks, ranging from dialog response generation to mathematical reasoning, using state-of-the-art (GPT-3.5, ChatGPT, and GPT-4) LLMs. Across all evaluated tasks, outputs generated with Self-Refine are preferred by humans and automatic metrics over those generated with the same LLM using conventional one-step generation, improving by ~20% absolute on average in task performance. Our work demonstrates that even state-of-the-art LLMs like GPT-4 can be further improved at test time using our simple, standalone approach."
                },
                {
                    "title": "Self-Instruct: Aligning Language Models with Self-Generated Instructions",
                    "abstract": "Large \u201cinstruction-tuned\u201d language models (i.e., finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is often limited in quantity, diversity, and creativity, therefore hindering the generality of the tuned model. We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off their own generations. Our pipeline generates instructions, input, and output samples from a language model, then filters invalid or similar ones before using them to finetune the original model. Applying our method to the vanilla GPT3, we demonstrate a 33% absolute improvement over the original model on Super-NaturalInstructions, on par with the performance of InstructGPT-001, which was trained with private user data and human annotations. For further evaluation, we curate a set of expert-written instructions for novel tasks, and show through human evaluation that tuning GPT3 with Self-Instruct outperforms using existing public instruction datasets by a large margin, leaving only a 5% absolute gap behind InstructGPT-001. Self-Instruct provides an almost annotation-free method for aligning pre-trained language models with instructions, and we release our large synthetic dataset to facilitate future studies on instruction tuning."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "An Explanation of In-context Learning as Implicit Bayesian Inference",
                    "abstract": "Large language models (LMs) such as GPT-3 have the surprising ability to do in-context learning, where the model learns to do a downstream task simply by conditioning on a prompt consisting of input-output examples. The LM learns from these examples without being explicitly pretrained to learn. Thus, it is unclear what enables in-context learning. In this paper, we study how in-context learning can emerge when pretraining documents have long-range coherence. Here, the LM must infer a latent document-level concept to generate coherent next tokens during pretraining. At test time, in-context learning occurs when the LM also infers a shared latent concept between examples in a prompt. We prove when this occurs despite a distribution mismatch between prompts and pretraining data in a setting where the pretraining distribution is a mixture of HMMs. In contrast to messy large-scale datasets used to train LMs capable of in-context learning, we generate a small-scale synthetic dataset (GINC) where Transformers and LSTMs both exhibit in-context learning. Beyond the theory, experiments on GINC exhibit large-scale real-world phenomena including improved in-context performance with model scaling (despite the same pretraining loss), sensitivity to example order, and instances where zero-shot is better than few-shot in-context learning."
                },
                {
                    "title": "Iterated learning for emergent systematicity in VQA",
                    "abstract": "Although neural module networks have an architectural bias towards compositionality, they require gold standard layouts to generalize systematically in practice. When instead learning layouts and modules jointly, compositionality does not arise automatically and an explicit pressure is necessary for the emergence of layouts exhibiting the right structure. We propose to address this problem using iterated learning, a cognitive science theory of the emergence of compositional languages in nature that has primarily been applied to simple referential games in machine learning. Considering the layouts of module networks as samples from an emergent language, we use iterated learning to encourage the development of structure within this language. We show that the resulting layouts support systematic generalization in neural agents solving the more complex task of visual question-answering. Our regularized iterated learning method can outperform baselines without iterated learning on SHAPES-SyGeT (SHAPES Systematic Generalization Test), a new split of the SHAPES dataset we introduce to evaluate systematic generalization, and on CLOSURE, an extension of CLEVR also designed to test systematic generalization. We demonstrate superior performance in recovering ground-truth compositional program structure with limited supervision on both SHAPES-SyGeT and CLEVR."
                },
                {
                    "title": "ACRE: Abstract Causal REasoning Beyond Covariation",
                    "abstract": "Causal induction, i.e., identifying unobservable mechanisms that lead to the observable relations among variables, has played a pivotal role in modern scientific discovery, especially in scenarios with only sparse and limited data. Humans, even young toddlers, can induce causal relationships surprisingly well in various settings despite its notorious difficulty. However, in contrast to the commonplace trait of human cognition is the lack of a diagnostic benchmark to measure causal induction for modern Artificial Intelligence (AI) systems. Therefore, in this work, we introduce the Abstract Causal REasoning (ACRE) dataset for systematic evaluation of current vision systems in causal induction. Motivated by the stream of research on causal discovery in Blicket experiments, we query a visual reasoning system with the following four types of questions in either an independent scenario or an interventional scenario: direct, indirect, screening-off, and backward-blocking, intentionally going beyond the simple strategy of inducing causal relationships by covariation. By analyzing visual reasoning architectures on this testbed, we notice that pure neural models tend towards an associative strategy under their chance-level performance, whereas neuro-symbolic combinations struggle in backward-blocking reasoning. These deficiencies call for future research in models with a more comprehensive capability of causal induction."
                },
                {
                    "title": "Countering Language Drift with Seeded Iterated Learning",
                    "abstract": "Pretraining on human corpus and then finetuning in a simulator has become a standard pipeline for training a goal-oriented dialogue agent. Nevertheless, as soon as the agents are finetuned to maximize task completion, they suffer from the so-called language drift phenomenon: they slowly lose syntactic and semantic properties of language as they only focus on solving the task. In this paper, we propose a generic approach to counter language drift called Seeded iterated learning (SIL). We periodically refine a pretrained student agent by imitating data sampled from a newly generated teacher agent. At each time step, the teacher is created by copying the student agent, before being finetuned to maximize task completion. SIL does not require external syntactic constraint nor semantic knowledge, making it a valuable task-agnostic finetuning protocol. We evaluate SIL in a toy-setting Lewis Game, and then scale it up to the translation game with natural language. In both settings, SIL helps counter language drift as well as it improves the task completion compared to baselines."
                },
                {
                    "title": "Self-Distillation Amplifies Regularization in Hilbert Space",
                    "abstract": "Knowledge distillation introduced in the deep learning context is a method to transfer knowledge from one architecture to another. In particular, when the architectures are identical, this is called self-distillation. The idea is to feed in predictions of the trained model as new target values for retraining (and iterate this loop possibly a few times). It has been empirically observed that the self-distilled model often achieves higher accuracy on held out data. Why this happens, however, has been a mystery: the self-distillation dynamics does not receive any new information about the task and solely evolves by looping over training. To the best of our knowledge, there is no rigorous understanding of why this happens. This work provides the first theoretical analysis of self-distillation. We focus on fitting a nonlinear function to training data, where the model space is Hilbert space and fitting is subject to L2 regularization in this function space. We show that self-distillation iterations modify regularization by progressively limiting the number of basis functions that can be used to represent the solution. This implies (as we also verify empirically) that while a few rounds of self-distillation may reduce over-fitting, further rounds may lead to under-fitting and thus worse performance."
                },
                {
                    "title": "Compositional Languages Emerge in a Neural Iterated Learning Model",
                    "abstract": "The principle of compositionality, which enables natural language to represent complex concepts via a structured combination of simpler ones, allows us to convey an open-ended set of messages using a limited vocabulary. If compositionality is indeed a natural property of language, we may expect it to appear in communication protocols that are created by neural agents via grounded language learning. Inspired by the iterated learning framework, which simulates the process of language evolution, we propose an effective neural iterated learning algorithm that, when applied to interacting neural agents, facilitates the emergence of a more structured type of language. Indeed, these languages provide specific advantages to neural agents during training, which translates as a larger posterior probability, which is then incrementally amplified via the iterated learning procedure. Our experiments confirm our analysis, and also demonstrate that the emerged languages largely improve the generalization of the neural agent communication."
                },
                {
                    "title": "The Emergence of Compositional Languages for Numeric Concepts Through Iterated Learning in Neural Agents",
                    "abstract": "Since first introduced, computer simulation has been an increasingly important tool in evolutionary linguistics. Recently, with the development of deep learning techniques, research in grounded language learning has also started to focus on facilitating the emergence of compositional languages without pre-defined elementary linguistic knowledge. In this work, we explore the emergence of compositional languages for numeric concepts in multi-agent communication systems. We demonstrate that compositional language for encoding numeric concepts can emerge through iterated learning in populations of deep neural network agents. However, language properties greatly depend on the input representations given to agents. We found that compositional languages only emerge if they require less iterations to be fully learnt than other non-degenerate languages for agents on a given input representation."
                },
                {
                    "title": "When Extremists Win: Cultural Transmission Via Iterated Learning When Populations Are Heterogeneous",
                    "abstract": "How does the process of information transmission affect the cultural or linguistic products that emerge? This question is often studied experimentally and computationally via iterated learning, a procedure in which participants learn from previous participants in a chain. Iterated learning is a powerful tool because, when all participants share the same priors, the stationary distributions of the iterated learning chains reveal those priors. In many situations, however, it is unreasonable to assume that all participants share the same prior beliefs. We present four simulation studies and one experiment demonstrating that when the population of learners is heterogeneous, the behavior of an iterated learning chain can be unpredictable and is often systematically distorted by the learners with the most extreme biases. This results in group-level outcomes that reflect neither the behavior of any individuals within the population nor the overall population average. We discuss implications for the use of iterated learning as a methodological tool as well as for the processes that might have shaped cultural and linguistic evolution in the real world."
                },
                {
                    "title": "The Interactive Evolution of Human Communication Systems",
                    "abstract": "This paper compares two explanations of the process by which human communication systems evolve: iterated learning and social collaboration. It then reports an experiment testing the social collaboration account. Participants engaged in a graphical communication task either as a member of a community, where they interacted with seven different partners drawn from the same pool, or as a member of an isolated pair, where they interacted with the same partner across the same number of games. Participants' horizontal, pair-wise interactions led \"bottom up\" to the creation of an effective and efficient shared sign system in the community condition. Furthermore, the community-evolved sign systems were as effective and efficient as the local sign systems developed by isolated pairs. Finally, and as predicted by a social collaboration account, and not by an iterated learning account, interaction was critical to the creation of shared sign systems, with different isolated pairs establishing different local sign systems and different communities establishing different global sign systems."
                },
                {
                    "title": "Cumulative cultural evolution in the laboratory: An experimental approach to the origins of structure in human language",
                    "abstract": "We introduce an experimental paradigm for studying the cumulative cultural evolution of language. In doing so we provide the first experimental validation for the idea that cultural transmission can lead to the appearance of design without a designer. Our experiments involve the iterated learning of artificial languages by human participants. We show that languages transmitted culturally evolve in such a way as to maximize their own transmissibility: over time, the languages in our experiments become easier to learn and increasingly structured. Furthermore, this structure emerges purely as a consequence of the transmission of language over generations, without any intentional design on the part of individual language learners. Previous computational and mathematical models suggest that iterated learning provides an explanation for the structure of human language and link particular aspects of linguistic structure with particular constraints acting on language during its transmission. The experimental work presented here shows that the predictions of these models, and models of cultural evolution more generally, can be tested in the laboratory."
                },
                {
                    "title": "Using Category Structures to Test Iterated Learning as a Method for Identifying Inductive Biases",
                    "abstract": "Many of the problems studied in cognitive science are inductive problems, requiring people to evaluate hypotheses in the light of data. The key to solving these problems successfully is having the right inductive biases-assumptions about the world that make it possible to choose between hypotheses that are equally consistent with the observed data. This article explores a novel experimental method for identifying the biases that guide human inductive inferences. The idea behind this method is simple: This article uses the responses produced by a participant on one trial to generate the stimuli that either they or another participant will see on the next. A formal analysis of this \"iterated learning\" procedure, based on the assumption that the learners are Bayesian agents, predicts that it should reveal the inductive biases of these learners, as expressed in a prior probability distribution over hypotheses. This article presents a series of experiments using stimuli based on a well-studied set of category structures, demonstrating that iterated learning can be used to reveal the inductive biases of human learners."
                },
                {
                    "title": "Innateness and culture in the evolution of language",
                    "abstract": "Human language arises from biological evolution, individual learning, and cultural transmission, but the interaction of these three processes has not been widely studied. We set out a formal framework for analyzing cultural transmission, which allows us to investigate how innate learning biases are related to universal properties of language. We show that cultural transmission can magnify weak biases into strong linguistic universals, undermining one of the arguments for strong innate constraints on language learning. As a consequence, the strength of innate biases can be shielded from natural selection, allowing these genes to drift. Furthermore, even when there is no natural selection, cultural transmission can produce apparent adaptations. Cultural transmission thus provides an alternative to traditional nativist and adaptationist explanations for the properties of human languages."
                },
                {
                    "title": "The stochastic EM algorithm: estimation and asymptotic results",
                    "abstract": "The EM algorithm is a much used tool for maximum likelihood estimation in missing or incomplete data problems. However, calculating the conditional expectation required in the E-step of the algorithm may be infeasible, especially when this expectation is a large sum or a high-dimensional integral. Instead the expectation can be estimated by simulation. This is the common idea in the stochastic EM algorithm and the Monte Carlo EM algorithm. In this paper some asymptotic results for the Stochastic EM algorithm are given, and estimation based on this algorithm is discussed. In particular, asymptotic equivalence of certain simple estimators is shown, and a simulation experiment is carried out to investigate this equivalence in small and moderate samples. Furthermore, some implementation issues and the possibility of allowing unidentified parameters in the algorithm are discussed."
                },
                {
                    "title": "Iterated learning and communication jointly explain efficient color naming systems",
                    "abstract": "It has been argued that semantic systems re\ufb02ect pressure for ef\ufb01ciency, and a current debate concerns the cultural evolutionary process that produces this pattern. We consider ef\ufb01-ciency as instantiated in the Information Bottleneck (IB) principle, and a model of cultural evolution that combines iterated learning and communication. We show that this model, instantiated in neural networks, converges to color naming systems that are ef\ufb01cient in the IB sense and similar to human color naming systems. We also show that iterated learning alone, and communication alone, do not yield the same outcome as clearly."
                },
                {
                    "title": "Gibbs Sampling from Human Feedback: A Provable KL- constrained Framework for RLHF",
                    "abstract": "This paper studies the theoretical framework of the alignment process of generative models with Reinforcement Learning from Human Feedback (RLHF). We consider a standard mathematical formulation, the reverse-KL regularized contextual bandit for RLHF. Despite its widespread practical application, a rigorous theoretical analysis of this formulation remains open. We investigate its theoretical properties both in offline and online settings and propose efficient algorithms with finite-sample theoretical guarantees. Our work bridges the gap between theory and practice by linking our theoretical insights with existing practical alignment algorithms such as Direct Preference Optimization (DPO) and Rejection Sampling Optimization (RSO). Furthermore, these findings and connections also offer both theoretical and practical communities new tools and insights for future algorithmic design of alignment algorithms."
                },
                {
                    "title": "Iterated learning and the cultural ratchet",
                    "abstract": "Iterated Learning and the Cultural Ratchet Aaron Beppu (abeppu@berkeley.edu) Thomas L. Griffiths (tom griffiths@berkeley.edu) Department of Psychology and Cognitive Science Program University of California Berkeley, Berkeley, CA 94720 USA Abstract How does the behavior of individuals in a society influence whether knowledge accumulates over generations? We explore this question using a simple model of cultural evolution as a process of \u201citerated learning,\u201d where each agent in a sequence learns and passes on a piece of information. Using both math- ematical analyses involving rational Bayesian agents and lab- oratory experiments with human participants, we vary whether agents observe data from the environment and what kind of information they receive from the previous agent. Our mathe- matical and empirical results both suggest that merely observ- ing the behavior of other learners is not sufficient to produce cumulative cultural evolution, but that knowledge can be accu- mulated over generations when agents are able to communicate the plausibility of different hypotheses. Introduction What we as humans know about the world is largely informed by what we can learn from others. Our beliefs and behaviors are built up not only by our own interactions with the environ- ment around us, but by observations of the behavior of oth- ers, and by receiving explicit statements of belief. Through this ability to build off the knowledge and behaviors of those around us, human societies are able to accumulate informa- tion, the so-called \u201cratchet effect\u201d (Tomasello, Kruger, & Rat- ner, 1993). Large, lasting changes in our beliefs develop by accruing small, incremental changes over many generations. In this paper, we use a simple model to characterize the con- ditions necessary for this accrual, and compare the predic- tions of this model to experimental results. In particular, we suggest that these changes will accumulate if learners have access to the actual beliefs of their teachers, as opposed to merely observations of behavior informed by those beliefs. Much of the previous discussion about cumulative cultural evolution has focused on the relationship between the teacher and the learner. For example, Tomasello (1993) largely claims that the ratchet effect is a product of sophisticated imi- tative learning, emphasizing the advantages of joint attention and a theory of mind. Gergely and Csibra (2005) suggest that \u201cnatural pedagogy\u201d plays a role, focusing on how human teachers go out of their way to ease the learning process for children, providing highly tailored information and a greatly enriched learning environment. Both of these mechanisms increase the fidelity of cultural transmission, providing the opportunity for the ratchet effect occur. Rather than focusing on the specific actions or abilities of teachers and learners, we consider the kind of informa- tion passed between generations. We model the process of cultural evolution using an \u201citerated learning\u201d framework (Kirby, 2001), in which successive generations are modeled as a sequence of agents in which each agent learns from the previous agent. Iterated learning models have recently been used in simulations exploring the emergence of linguis- tic structures (Kirby, 2001) and in experiments with human participants revealing inductive biases in specific domains (Kalish, Griffiths, & Lewandowsky, 2007). We explore vari- ants of this framework where each agent learns not only from the behavior of the previous generation, but also from ob- serving the world and from theories summarizing the infer- ences of the previous generation. The consequences of each variant for the accumulation of knowledge are determined through a combination of mathematical analyses of sequences of Bayesian agents and an experiment in which cultural evo- lution is simulated in the laboratory with human participants. Our mathematical and empirical analyses produce two re- sults that help to identify the conditions under which cumu- lative cultural evolution can occur. First, we show that if the information passed from teacher to learner is limited to the learner\u2019s observations of the teacher\u2019s behavior, then cumula- tive cultural evolution will not take place. We then go on to show that cumulative cultural evolution can occur if teachers give learners information about which hypotheses they be- lieve are plausible, rather than just behaviors consistent with the most plausible hypothesis. These results suggest that a capacity for communicating such beliefs may be necessary in order for the cultural ratchet to operate. The plan of the paper is as follows. The next section intro- duces our formal framework, and summarizes our mathemati- cal results. We then outline the motivation for our experiment with human learners, and describe the procedure and results of that experiment. The paper concludes with a discussion of the implications of these results for understanding the con- ditions under which the cultural ratchet can operate, and the kinds of cognitive capacities that teachers and learners need to possess in order to satisfy these conditions. Analyzing Cultural Transmission Our first step in exploring the conditions that result in cu- mulative cultural evolution is analyzing variants of iterated learning in which the learners are rational Bayesian agents. The assumption of rationality allows us to examine how ideal learners use the information transmitted between generations, and allows us to be explicit about the biases that influence learning. Specifically, we assume that all agents have a shared set of hypotheses H, and a prior distribution over these hy- potheses p(h) describing the degree to which each h \u2208 H is believed to be true in the absence of any evidence. Starting with these prior beliefs, agents observe the world and receive information from other agents, resulting in data,"
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively leverage iterated learning frameworks to enhance the instruction-following capabilities of large language models through self-data-augmentation methods?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem could significantly advance the research community's understanding of how language models can learn and evolve over time, similar to human cultural and linguistic development. By improving the instruction-following abilities of LLMs, we can unlock new practical applications in areas such as automated content generation, personalized learning systems, and more efficient human-computer interactions. This research could pave the way for future studies on the dynamics of knowledge transfer and refinement in AI systems, ultimately leading to more robust and adaptable models.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem include the complexity of modeling the iterative learning process in a way that accurately reflects human-like knowledge transmission and refinement. Naive approaches may fail because they do not account for the nuanced interactions between multiple generations of models or the subtleties of knowledge representation and evolution. Additionally, technical obstacles such as ensuring stability in learning, managing the trade-off between exploration and exploitation, and effectively measuring the impact of self-data-augmentation methods complicate the implementation of a successful solution.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on isolated aspects of language model training or self-data-augmentation without integrating the iterated learning framework comprehensively. Limitations in understanding the dynamics of knowledge transfer and the lack of robust methodologies to simulate these processes have hindered progress. Additionally, existing solutions may not have adequately addressed the complexities of multi-agent interactions or the iterative nature of learning. Our approach aims to fill these gaps by providing a structured methodology that combines insights from cognitive science with advanced machine learning techniques.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves implementing a Bayesian iterated learning framework that utilizes self-data-augmentation techniques across multiple generations of language models. We will use a diverse dataset of instructional prompts and responses to train the models, measuring their performance through metrics such as instruction-following accuracy and knowledge retention over iterations. The expected outcomes include improved model performance in following complex instructions, enhanced adaptability to new tasks, and insights into the mechanisms of knowledge evolution in AI systems."
            }
        },
        "author_data": {
            "3447b661-4a7e-47b6-be8e-5006bdf06896": {
                "pk": "3447b661-4a7e-47b6-be8e-5006bdf06896",
                "name": "Yi Ren",
                "collaborators": [
                    "Donald Goldfarb",
                    "Hongyan Tang",
                    "Sandra Hirche",
                    "Sheng Cheng",
                    "Yang Jiao",
                    "Danica J. Sutherland",
                    "Jing Liu",
                    "Yi Chen",
                    "Lei Zhang",
                    "Tanner Merry"
                ],
                "domain": [
                    "Optimization",
                    "Deep Learning",
                    "Recommender Systems",
                    "Robotics"
                ],
                "publications": [
                    {
                        "title": "Efficient Subsampled Gauss-Newton and Natural Gradient Methods for Training Neural Networks",
                        "abstract": "We present practical Levenberg-Marquardt variants of Gauss-Newton and natural gradient methods for solving non-convex optimization problems that arise in training deep neural networks involving enormous numbers of variables and huge data sets. Our methods use subsampled Gauss-Newton or Fisher information matrices and either subsampled gradient estimates (fully stochastic) or full gradients (semi-stochastic), which, in the latter case, we prove convergent to a stationary point. By using the Sherman-Morrison-Woodbury formula with automatic differentiation (backpropagation) we show how our methods can be implemented to perform efficiently. Finally, numerical results are presented to demonstrate the effectiveness of our proposed methods."
                    },
                    {
                        "title": "Fully distributed cooperation for networked uncertain mobile manipulators",
                        "abstract": "This paper investigates a fully distributed cooperation scheme for networked mobile manipulators. To achieve cooperative task allocation in a distributed way, an adaptation-based estimation law is established for each robotic agent to estimate the desired local trajectory. In addition, wrench synthesis is analyzed in detail to lay a solid foundation for tight cooperation tasks. Together with the estimated task, a set of distributed adaptive controllers is proposed to achieve motion synchronization of the mobile manipulator ensemble over a directed graph with a spanning tree irrespective of the kinematic and dynamic uncertainties in both the mobile manipulators and the tightly grasped object. The controlled synchronization alleviates the performance degradation caused by the estimation/tracking discrepancy during the transient phase. The proposed scheme requires no persistent excitation condition and avoids the use of noisy Cartesian-space velocities. Furthermore, it is independent from the object's center of mass by employing formation-based task allocation and a task-oriented strategy. These attractive attributes facilitate the practical application of the scheme. It is theoretically proven that convergence of the cooperative task tracking error is guaranteed. Simulation results validate the efficacy and demonstrate the expected performance of the proposed scheme."
                    },
                    {
                        "title": "Tensor Normal Training for Deep Learning Models",
                        "abstract": "Despite the predominant use of first-order methods for training deep learning models, second-order methods, and in particular, natural gradient methods, remain of interest because of their potential for accelerating training through the use of curvature information. Several methods with non-diagonal preconditioning matrices, including KFAC, Shampoo, and K-BFGS, have been proposed and shown to be effective. Based on the so-called tensor normal (TN) distribution, we propose and analyze a brand new approximate natural gradient method, Tensor Normal Training (TNT), which like Shampoo, only requires knowledge of the shape of the training parameters. By approximating the probabilistically based Fisher matrix, as opposed to the empirical Fisher matrix, our method uses the block-wise covariance of the sampling based gradient as the pre-conditioning matrix. Moreover, the assumption that the sampling-based (tensor) gradient follows a TN distribution, ensures that its covariance has a Kronecker separable structure, which leads to a tractable approximation to the Fisher matrix. Consequently, TNT's memory requirements and per-iteration computational costs are only slightly higher than those for first-order methods. In our experiments, TNT exhibited superior optimization performance to state-of-the-art first-order methods, and comparable optimization performance to the state-of-the-art second-order methods KFAC and Shampoo. Moreover, TNT demonstrated its ability to generalize as well as first-order methods, while using fewer epochs."
                    },
                    {
                        "title": "Data-Driven Learning of 3-Point Correlation Functions as Microstructure Representations",
                        "abstract": "This paper considers the open challenge of identifying complete, concise, and explainable quantitative microstructure representations for disordered heterogeneous material systems. Completeness and conciseness have been achieved through existing data-driven methods, e.g., deep generative models, which, however, do not provide mathematically explainable latent representations. This study investigates representations composed of three-point correlation functions, which are a special type of spatial convolutions. We show that a variety of microstructures can be characterized by a concise subset of three-point correlations, and the identification of such subsets can be achieved by Bayesian optimization. Lastly, we show that the proposed representation can directly be used to compute material properties based on the effective medium theory."
                    },
                    {
                        "title": "Understanding Simplicity Bias towards Compositional Mappings via Learning Dynamics",
                        "abstract": "Obtaining compositional mappings is important for the model to generalize well compositionally. To better understand when and how to encourage the model to learn such mappings, we study their uniqueness through different perspectives. Specifically, we first show that the compositional mappings are the simplest bijections through the lens of coding length (i.e., an upper bound of their Kolmogorov complexity). This property explains why models having such mappings can generalize well. We further show that the simplicity bias is usually an intrinsic property of neural network training via gradient descent. That partially explains why some models spontaneously generalize well when they are trained appropriately."
                    },
                    {
                        "title": "Liquid Metal Enabled Droplet Circuits",
                        "abstract": "Conventional electrical circuits are generally rigid in their components and working styles which are not flexible and stretchable. From an alternative, liquid metal based soft electronics is offering important opportunities for innovating modern bioelectronics and electrical engineering. However, its running in wet environments such as aqueous solution, biological tissues or allied subjects still encounters many technical challenges. Here, we proposed a new conceptual electrical circuit, termed as droplet circuits, to fulfill the special needs as raised in the above mentioned areas. Such unconventional circuits are immersed in solution and composed of liquid metal droplets, conductive ions or wires such as carbon nanotubes. With specifically designed topological or directional structures/patterns, the liquid metal droplets composing the circuit can be discretely existing and disconnected from each other, while achieving the function of electron transport through conductive routes or quantum tunneling effect. The conductive wires serve as the electron transfer stations when the distance between two separate liquid metal droplets is far beyond than that quantum tunneling effects can support. The unique advantage of the current droplet circuit lies in that it allows parallel electron transport, high flexibility, self-healing, regulativity and multi-point connectivity, without needing to worry about circuit break. This would extend the category of classical electrical circuits into the newly emerging areas like realizing room temperature quantum computing, making brain-like intelligence or nerve-machine interface electronics etc. The mechanisms and potential scientific issues of the droplet circuits are interpreted. Future prospects along this direction are outlined."
                    },
                    {
                        "title": "When Shall I Be Empathetic? The Utility of Empathetic Parameter Estimation in Multi-Agent Interactions",
                        "abstract": "Human-robot interactions (HRI) can be modeled as dynamic or differential games with incomplete information, where each agent holds private reward parameters. Due to the open challenge in finding perfect Bayesian equilibria of such games, existing studies often consider approximated solutions composed of parameter estimation and motion planning steps, in order to decouple the belief and physical dynamics. In parameter estimation, current approaches often assume that the reward parameters of the robot are known by the humans. We argue that by falsely conditioning on this assumption, the robot performs non-empathetic estimation of the humans' parameters, leading to undesirable values even in the simplest interactions. We test this argument by studying a two-vehicle uncontrolled intersection case with short reaction time. Results show that when both agents are unknowingly aggressive (or non-aggressive), empathy leads to more effective parameter estimation and higher reward values, suggesting that empathy is necessary when the true parameters of agents mismatch with their common belief. The proposed estimation and planning algorithms are therefore more robust than the existing approaches, by fully acknowledging the nature of information asymmetry in HRI. Lastly, we introduce value approximation techniques for real-time execution of the proposed algorithms."
                    },
                    {
                        "title": "Example-Based Image Synthesis via Randomized Patch-Matching",
                        "abstract": "Image and texture synthesis is a challenging task that has long been drawing attention in the fields of image processing, graphics, and machine learning. This problem consists of modelling the desired type of images, either through training examples or via a parametric modeling, and then generating images that belong to the same statistical origin.   This work addresses the image synthesis task, focusing on two specific families of images -- handwritten digits and face images. This paper offers two main contributions. First, we suggest a simple and intuitive algorithm capable of generating such images in a unified way. The proposed approach taken is pyramidal, consisting of upscaling and refining the estimated image several times. For each upscaling stage, the algorithm randomly draws small patches from a patch database, and merges these to form a coherent and novel image with high visual quality. The second contribution is a general framework for the evaluation of the generation performance, which combines three aspects: the likelihood, the originality and the spread of the synthesized images. We assess the proposed synthesis scheme and show that the results are similar in nature, and yet different from the ones found in the training set, suggesting that true synthesis effect has been obtained."
                    },
                    {
                        "title": "Adaptive Questionnaires for Direct Identification of Optimal Product Design",
                        "abstract": "We consider the problem of identifying the most profitable product design from a finite set of candidates under unknown consumer preference. A standard approach to this problem follows a two-step strategy: First, estimate the preference of the consumer population, represented as a point in part-worth space, using an adaptive discrete-choice questionnaire. Second, integrate the estimated part-worth vector with engineering feasibility and cost models to determine the optimal design. In this work, we (1) demonstrate that accurate preference estimation is neither necessary nor sufficient for identifying the optimal design, (2) introduce a novel adaptive questionnaire that leverages knowledge about engineering feasibility and manufacturing costs to directly determine the optimal design, and (3) interpret product design in terms of a nonlinear segmentation of part-worth space, and use this interpretation to illuminate the intrinsic difficulty of optimal design in the presence of noisy questionnaire responses. We establish the superiority of the proposed approach using a well-documented optimal product design task. This study demonstrates how the identification of optimal product design can be accelerated by integrating marketing and manufacturing knowledge into the adaptive questionnaire."
                    },
                    {
                        "title": "Design and Analysis of Dynamic Auto Scaling Algorithm (DASA) for 5G Mobile Networks",
                        "abstract": "Network Function Virtualization (NFV) enables mobile operators to virtualize their network entities as Virtualized Network Functions (VNFs), offering fine-grained on-demand network capabilities. VNFs can be dynamically scale-in/out to meet the performance requirements for future 5G networks. However, designing an auto-scaling algorithm with low operation cost and low latency while considering the capacity of legacy network equipment is a challenge. In this paper, we propose a VNF Dynamic Auto Scaling Algorithm (DASA) considering the tradeoff between performance and operation cost. We also develop an analytical model to quantify the tradeoff and validate the analysis through extensive simulations. The system is modeled as a queueing model while legacy network equipment is considered as a reserved block of servers. The VNF instances are powered on and off according to the number of job requests. The results show that the proposed DASA can significantly reduce operation cost given the latency upper-bound. Moreover, the models provide a quick way to evaluate the cost-performance tradeoff without wide deployment, which can save cost and time."
                    },
                    {
                        "title": "Unbiased Learning to Rank with Biased Continuous Feedback",
                        "abstract": "It is a well-known challenge to learn an unbiased ranker with biased feedback. Unbiased learning-to-rank(LTR) algorithms, which are verified to model the relative relevance accurately based on noisy feedback, are appealing candidates and have already been applied in many applications with single categorical labels, such as user click signals. Nevertheless, the existing unbiased LTR methods cannot properly handle continuous feedback, which are essential for many industrial applications, such as content recommender systems.   To provide personalized high-quality recommendation results, recommender systems need model both categorical and continuous biased feedback, such as click and dwell time. Accordingly, we design a novel unbiased LTR algorithm to tackle the challenges, which innovatively models position bias in the pairwise fashion and introduces the pairwise trust bias to separate the position bias, trust bias, and user relevance explicitly and can work for both continuous and categorical feedback. Experiment results on public benchmark datasets and internal live traffic of a large-scale recommender system at Tencent News show superior results for continuous labels and also competitive performance for categorical labels of the proposed method."
                    },
                    {
                        "title": "Disentangled Counterfactual Reasoning for Unbiased Sequential Recommendation",
                        "abstract": "Sequential recommender systems have achieved state-of-the-art recommendation performance by modeling the sequential dynamics of user activities. However, in most recommendation scenarios, the popular items comprise the major part of the previous user actions. Therefore, the learned models are biased towards the popular items irrespective of the user's real interests. In this paper, we propose a structural causal model-based method to address the popularity bias issue for sequential recommendation model learning. For more generalizable modeling, we disentangle the popularity and interest representations at both the item side and user context side. Based on the disentangled representation, we identify a more effective structural causal graph for general recommendation applications. Then, we design delicate sequential models to apply the aforementioned causal graph to the sequential recommendation scenario for unbiased prediction with counterfactual reasoning. Furthermore, we conduct extensive offline experiments and online A/B tests to verify the proposed \\textbf{DCR} (Disentangled Counterfactual Reasoning) method's superior overall performance and understand the effectiveness of the various introduced components. Based on our knowledge, this is the first structural causal model specifically designed for the popularity bias correction of sequential recommendation models, which achieves significant performance gains over the existing methods."
                    },
                    {
                        "title": "Deep Mutual Learning across Task Towers for Effective Multi-Task Recommender Learning",
                        "abstract": "Recommender systems usually leverage multi-task learning methods to simultaneously optimize several objectives because of the multi-faceted user behavior data. The typical way of conducting multi-task learning is to establish appropriate parameter sharing across multiple tasks at lower layers while reserving a separate task tower for each task at upper layers. Since the task towers exert direct impact on the prediction results, we argue that the architecture of standalone task towers is sub-optimal for promoting positive knowledge sharing. Accordingly, we propose the framework of Deep Mutual Learning across task towers, which is compatible with various backbone multi-task networks. Extensive offline experiments and online AB tests are conducted to evaluate and verify the proposed approach's effectiveness."
                    }
                ]
            },
            "1e8691c2-ec9c-4bce-b266-de588865c481": {
                "pk": "1e8691c2-ec9c-4bce-b266-de588865c481",
                "name": "Shangmin Guo",
                "collaborators": [
                    "Yi Ren",
                    "Stefano V. Albrecht",
                    "Kenny Smith",
                    "Samuel Garcin",
                    "Kory Mathewson",
                    "Danica J. Sutherland",
                    "Simon Kirby",
                    "James Doran",
                    "Christopher G. Lucas",
                    "Tianlin Liu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Reinforcement Learning",
                    "Multi-Agent Systems",
                    "Compositionality"
                ],
                "publications": [
                    {
                        "title": "Emergence of Numeric Concepts in Multi-Agent Autonomous Communication",
                        "abstract": "With the rapid development of deep learning, most of current state-of-the-art techniques in natural langauge processing are based on deep learning models trained with argescaled static textual corpora. However, we human beings learn and understand in a different way. Thus, grounded language learning argues that models need to learn and understand language by the experience and perceptions obtained by interacting with enviroments, like how humans do. With the help of deep reinforcement learning techniques, there are already lots of works focusing on facilitating the emergence of communication protocols that have compositionalities like natural languages among computational agents population. Unlike these works, we, on the other hand, focus on the numeric concepts which correspond to abstractions in cognition and function words in natural language. Based on a specifically designed language game, we verify that computational agents are capable of transmitting numeric concepts during autonomous communication, and the emergent communication protocols can reflect the underlying structure of meaning space. Although their encodeing method is not compositional like natural languages from a perspective of human beings, the emergent languages can be generalised to unseen inputs and, more importantly, are easier for models to learn. Besides, iterated learning can help further improving the compositionality of the emergent languages, under the measurement of topological similarity. Furthermore, we experiment another representation method, i.e. directly encode numerals into concatenations of one-hot vectors, and find that the emergent languages would become compositional like human natural languages. Thus, we argue that there are 2 important factors for the emergence of compositional languages."
                    },
                    {
                        "title": "Inductive Bias and Language Expressivity in Emergent Communication",
                        "abstract": "Referential games and reconstruction games are the most common game types for studying emergent languages. We investigate how the type of the language game affects the emergent language in terms of: i) language compositionality and ii) transfer of an emergent language to a task different from its origin, which we refer to as language expressivity. With empirical experiments on a handcrafted symbolic dataset, we show that languages emerged from different games have different compositionality and further different expressivity."
                    },
                    {
                        "title": "lpNTK: Better Generalisation with Less Data via Sample Interaction During Learning",
                        "abstract": "Although much research has been done on proposing new models or loss functions to improve the generalisation of artificial neural networks (ANNs), less attention has been directed to the impact of the training data on generalisation. In this work, we start from approximating the interaction between samples, i.e. how learning one sample would modify the model's prediction on other samples. Through analysing the terms involved in weight updates in supervised learning, we find that labels influence the interaction between samples. Therefore, we propose the labelled pseudo Neural Tangent Kernel (lpNTK) which takes label information into consideration when measuring the interactions between samples. We first prove that lpNTK asymptotically converges to the empirical neural tangent kernel in terms of the Frobenius norm under certain assumptions. Secondly, we illustrate how lpNTK helps to understand learning phenomena identified in previous work, specifically the learning difficulty of samples and forgetting events during learning. Moreover, we also show that using lpNTK to identify and remove poisoning training samples does not hurt the generalisation performance of ANNs."
                    },
                    {
                        "title": "Better Supervisory Signals by Observing Learning Paths",
                        "abstract": "Better-supervised models might have better performance. In this paper, we first clarify what makes for good supervision for a classification problem, and then explain two existing label refining methods, label smoothing and knowledge distillation, in terms of our proposed criterion. To further answer why and how better supervision emerges, we observe the learning path, i.e., the trajectory of the model's predictions during training, for each training sample. We find that the model can spontaneously refine \"bad\" labels through a \"zig-zag\" learning path, which occurs on both toy and real datasets. Observing the learning path not only provides a new perspective for understanding knowledge distillation, overfitting, and learning dynamics, but also reveals that the supervisory signal of a teacher network can be very unstable near the best points in training on real tasks. Inspired by this, we propose a new knowledge distillation scheme, Filter-KD, which improves downstream classification performance in various settings."
                    },
                    {
                        "title": "How to prepare your task head for finetuning",
                        "abstract": "In deep learning, transferring information from a pretrained network to a downstream task by finetuning has many benefits. The choice of task head plays an important role in fine-tuning, as the pretrained and downstream tasks are usually different. Although there exist many different designs for finetuning, a full understanding of when and why these algorithms work has been elusive. We analyze how the choice of task head controls feature adaptation and hence influences the downstream performance. By decomposing the learning dynamics of adaptation, we find that the key aspect is the training accuracy and loss at the beginning of finetuning, which determines the \"energy\" available for the feature's adaptation. We identify a significant trend in the effect of changes in this initial energy on the resulting features after fine-tuning. Specifically, as the energy increases, the Euclidean and cosine distances between the resulting and original features increase, while their dot products (and the resulting features' norm) first increase and then decrease. Inspired by this, we give several practical principles that lead to better downstream performance. We analytically prove this trend in an overparamterized linear setting and verify its applicability to different experimental settings."
                    },
                    {
                        "title": "Compositional Languages Emerge in a Neural Iterated Learning Model",
                        "abstract": "The principle of compositionality, which enables natural language to represent complex concepts via a structured combination of simpler ones, allows us to convey an open-ended set of messages using a limited vocabulary. If compositionality is indeed a natural property of language, we may expect it to appear in communication protocols that are created by neural agents in language games. In this paper, we propose an effective neural iterated learning (NIL) algorithm that, when applied to interacting neural agents, facilitates the emergence of a more structured type of language. Indeed, these languages provide learning speed advantages to neural agents during training, which can be incrementally amplified via NIL. We provide a probabilistic model of NIL and an explanation of why the advantage of compositional language exist. Our experiments confirm our analysis, and also demonstrate that the emerged languages largely improve the generalizing power of the neural agent communication."
                    },
                    {
                        "title": "Expressivity of Emergent Language is a Trade-off between Contextual Complexity and Unpredictability",
                        "abstract": "Researchers are using deep learning models to explore the emergence of language in various language games, where agents interact and develop an emergent language to solve tasks. We focus on the factors that determine the expressivity of emergent languages, which reflects the amount of information about input spaces those languages are capable of encoding. We measure the expressivity of emergent languages based on the generalisation performance across different games, and demonstrate that the expressivity of emergent languages is a trade-off between the complexity and unpredictability of the context those languages emerged from. Another contribution of this work is the discovery of message type collapse, i.e. the number of unique messages is lower than that of inputs. We also show that using the contrastive loss proposed by Chen et al. (2020) can alleviate this problem."
                    },
                    {
                        "title": "Economics Arena for Large Language Models",
                        "abstract": "Large language models (LLMs) have been extensively used as the backbones for general-purpose agents, and some economics literature suggest that LLMs are capable of playing various types of economics games. Following these works, to overcome the limitation of evaluating LLMs using static benchmarks, we propose to explore competitive games as an evaluation for LLMs to incorporate multi-players and dynamicise the environment. By varying the game history revealed to LLMs-based players, we find that most of LLMs are rational in that they play strategies that can increase their payoffs, but not as rational as indicated by Nash Equilibria (NEs). Moreover, when game history are available, certain types of LLMs, such as GPT-4, can converge faster to the NE strategies, which suggests higher rationality level in comparison to other models. In the meantime, certain types of LLMs can win more often when game history are available, and we argue that the winning rate reflects the reasoning ability with respect to the strategies of other players. Throughout all our experiments, we observe that the ability to strictly follow the game rules described by natural languages also vary among the LLMs we tested. In this work, we provide an economics arena for the LLMs research community as a dynamic simulation to test the above-mentioned abilities of LLMs, i.e. rationality, strategic reasoning ability, and instruction-following capability."
                    },
                    {
                        "title": "The Emergence of Compositional Languages for Numeric Concepts Through Iterated Learning in Neural Agents",
                        "abstract": "Since first introduced, computer simulation has been an increasingly important tool in evolutionary linguistics. Recently, with the development of deep learning techniques, research in grounded language learning has also started to focus on facilitating the emergence of compositional languages without pre-defined elementary linguistic knowledge. In this work, we explore the emergence of compositional languages for numeric concepts in multi-agent communication systems. We demonstrate that compositional language for encoding numeric concepts can emerge through iterated learning in populations of deep neural network agents. However, language properties greatly depend on the input representations given to agents. We found that compositional languages only emerge if they require less iterations to be fully learnt than other non-degenerate languages for agents on a given input representation."
                    },
                    {
                        "title": "Smoothing Matters: Momentum Transformer for Domain Adaptive Semantic Segmentation",
                        "abstract": "After the great success of Vision Transformer variants (ViTs) in computer vision, it has also demonstrated great potential in domain adaptive semantic segmentation. Unfortunately, straightforwardly applying local ViTs in domain adaptive semantic segmentation does not bring in expected improvement. We find that the pitfall of local ViTs is due to the severe high-frequency components generated during both the pseudo-label construction and features alignment for target domains. These high-frequency components make the training of local ViTs very unsmooth and hurt their transferability. In this paper, we introduce a low-pass filtering mechanism, momentum network, to smooth the learning dynamics of target domain features and pseudo labels. Furthermore, we propose a dynamic of discrepancy measurement to align the distributions in the source and target domains via dynamic weights to evaluate the importance of the samples. After tackling the above issues, extensive experiments on sim2real benchmarks show that the proposed method outperforms the state-of-the-art methods. Our codes are available at https://github.com/alpc91/TransDA"
                    },
                    {
                        "title": "How the level sampling process impacts zero-shot generalisation in deep reinforcement learning",
                        "abstract": "A key limitation preventing the wider adoption of autonomous agents trained via deep reinforcement learning (RL) is their limited ability to generalise to new environments, even when these share similar characteristics with environments encountered during training. In this work, we investigate how a non-uniform sampling strategy of individual environment instances, or levels, affects the zero-shot generalisation (ZSG) ability of RL agents, considering two failure modes: overfitting and over-generalisation. As a first step, we measure the mutual information (MI) between the agent's internal representation and the set of training levels, which we find to be well-correlated to instance overfitting. In contrast to uniform sampling, adaptive sampling strategies prioritising levels based on their value loss are more effective at maintaining lower MI, which provides a novel theoretical justification for this class of techniques. We then turn our attention to unsupervised environment design (UED) methods, which adaptively generate new training levels and minimise MI more effectively than methods sampling from a fixed set. However, we find UED methods significantly shift the training distribution, resulting in over-generalisation and worse ZSG performance over the distribution of interest. To prevent both instance overfitting and over-generalisation, we introduce self-supervised environment design (SSED). SSED generates levels using a variational autoencoder, effectively reducing MI while minimising the shift with the distribution of interest, and leads to statistically significant improvements in ZSG over fixed-set level sampling strategies and UED methods."
                    },
                    {
                        "title": "DRED: Zero-Shot Transfer in Reinforcement Learning via Data-Regularised Environment Design",
                        "abstract": "Autonomous agents trained using deep reinforcement learning (RL) often lack the ability to successfully generalise to new environments, even when these environments share characteristics with the ones they have encountered during training. In this work, we investigate how the sampling of individual environment instances, or levels, affects the zero-shot generalisation (ZSG) ability of RL agents. We discover that, for deep actor-critic architectures sharing their base layers, prioritising levels according to their value loss minimises the mutual information between the agent's internal representation and the set of training levels in the generated training data. This provides a novel theoretical justification for the regularisation achieved by certain adaptive sampling strategies. We then turn our attention to unsupervised environment design (UED) methods, which assume control over level generation. We find that existing UED methods can significantly shift the training distribution, which translates to low ZSG performance. To prevent both overfitting and distributional shift, we introduce data-regularised environment design (DRED). DRED generates levels using a generative model trained to approximate the ground truth distribution of an initial set of level parameters. Through its grounding, DRED achieves significant improvements in ZSG over adaptive level sampling strategies and UED methods. Our code and experimental data are available at https://github.com/uoe-agents/dred."
                    },
                    {
                        "title": "Decoding-time Realignment of Language Models",
                        "abstract": "Aligning language models with human preferences is crucial for reducing errors and biases in these models. Alignment techniques, such as reinforcement learning from human feedback (RLHF), are typically cast as optimizing a tradeoff between human preference rewards and a proximity regularization term that encourages staying close to the unaligned model. Selecting an appropriate level of regularization is critical: insufficient regularization can lead to reduced model capabilities due to reward hacking, whereas excessive regularization hinders alignment. Traditional methods for finding the optimal regularization level require retraining multiple models with varying regularization strengths. This process, however, is resource-intensive, especially for large models. To address this challenge, we propose decoding-time realignment (DeRa), a simple method to explore and evaluate different regularization strengths in aligned models without retraining. DeRa enables control over the degree of alignment, allowing users to smoothly transition between unaligned and aligned models. It also enhances the efficiency of hyperparameter tuning by enabling the identification of effective regularization strengths using a validation dataset."
                    },
                    {
                        "title": "Direct Language Model Alignment from Online AI Feedback",
                        "abstract": "Direct alignment from preferences (DAP) methods, such as DPO, have recently emerged as efficient alternatives to reinforcement learning from human feedback (RLHF), that do not require a separate reward model. However, the preference datasets used in DAP methods are usually collected ahead of training and never updated, thus the feedback is purely offline. Moreover, responses in these datasets are often sampled from a language model distinct from the one being aligned, and since the model evolves over training, the alignment phase is inevitably off-policy. In this study, we posit that online feedback is key and improves DAP methods. Our method, online AI feedback (OAIF), uses an LLM as annotator: on each training iteration, we sample two responses from the current model and prompt the LLM annotator to choose which one is preferred, thus providing online feedback. Despite its simplicity, we demonstrate via human evaluation in several tasks that OAIF outperforms both offline DAP and RLHF methods. We further show that the feedback leveraged in OAIF is easily controllable, via instruction prompts to the LLM annotator."
                    },
                    {
                        "title": "Deep Reinforcement Learning for Multi-Agent Interaction",
                        "abstract": "The development of autonomous agents which can interact with other agents to accomplish a given task is a core area of research in artificial intelligence and machine learning. Towards this goal, the Autonomous Agents Research Group develops novel machine learning algorithms for autonomous systems control, with a specific focus on deep reinforcement learning and multi-agent reinforcement learning. Research problems include scalable learning of coordinated agent policies and inter-agent communication; reasoning about the behaviours, goals, and composition of other agents from limited observations; and sample-efficient learning based on intrinsic motivation, curriculum learning, causal inference, and representation learning. This article provides a broad overview of the ongoing research portfolio of the group and discusses open problems for future directions."
                    }
                ]
            },
            "a8e96755-a812-4828-b471-be7a42d7876e": {
                "pk": "a8e96755-a812-4828-b471-be7a42d7876e",
                "name": "Linlu Qiu",
                "collaborators": [
                    "Peter Shaw",
                    "Fei Sha",
                    "Panupong Pasupat",
                    "Yoon Kim",
                    "Hexiang Hu",
                    "Bowen Zhang",
                    "Jiangmiao Pang",
                    "Haofeng Chen",
                    "Trevor Darrell",
                    "Fisher Yu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Object Tracking",
                    "Inductive Reasoning"
                ],
                "publications": [
                    {
                        "title": "Systematic Generalization on gSCAN: What is Nearly Solved and What is Next?",
                        "abstract": "We analyze the grounded SCAN (gSCAN) benchmark, which was recently proposed to study systematic generalization for grounded language understanding. First, we study which aspects of the original benchmark can be solved by commonly used methods in multi-modal research. We find that a general-purpose Transformer-based model with cross-modal attention achieves strong performance on a majority of the gSCAN splits, surprisingly outperforming more specialized approaches from prior work. Furthermore, our analysis suggests that many of the remaining errors reveal the same fundamental challenge in systematic generalization of linguistic constructs regardless of visual context. Second, inspired by this finding, we propose challenging new tasks for gSCAN by generating data to incorporate relations between objects in the visual environment. Finally, we find that current models are surprisingly data inefficient given the narrow scope of commands in gSCAN, suggesting another challenge for future work."
                    },
                    {
                        "title": "Visually Grounded Concept Composition",
                        "abstract": "We investigate ways to compose complex concepts in texts from primitive ones while grounding them in images. We propose Concept and Relation Graph (CRG), which builds on top of constituency analysis and consists of recursively combined concepts with predicate functions. Meanwhile, we propose a concept composition neural network called Composer to leverage the CRG for visually grounded concept learning. Specifically, we learn the grounding of both primitive and all composed concepts by aligning them to images and show that learning to compose leads to more robust grounding results, measured in text-to-image matching accuracy. Notably, our model can model grounded concepts forming at both the finer-grained sentence level and the coarser-grained intermediate level (or word-level). Composer leads to pronounced improvement in matching accuracy when the evaluation data has significant compound divergence from the training data."
                    },
                    {
                        "title": "Learning to Reason via Program Generation, Emulation, and Search",
                        "abstract": "Program synthesis with language models (LMs) has unlocked a large set of reasoning abilities; code-tuned LMs have proven adept at generating programs that solve a wide variety of algorithmic symbolic manipulation tasks (e.g. word concatenation). However, not all reasoning tasks are easily expressible as code, e.g. tasks involving commonsense reasoning, moral decision-making, and sarcasm understanding. Our goal is to extend an LM's program synthesis skills to such tasks and evaluate the results via pseudo-programs, namely Python programs where some leaf function calls are left undefined. To that end, we propose, Code Generation and Emulated EXecution (CoGEX). CoGEX works by (1) training LMs to generate their own pseudo-programs, (2) teaching them to emulate their generated program's execution, including those leaf functions, allowing the LM's knowledge to fill in the execution gaps; and (3) using them to search over many programs to find an optimal one. To adapt the CoGEX model to a new task, we introduce a method for performing program search to find a single program whose pseudo-execution yields optimal performance when applied to all the instances of a given dataset. We show that our approach yields large improvements compared to standard in-context learning approaches on a battery of tasks, both algorithmic and soft reasoning. This result thus demonstrates that code synthesis can be applied to a much broader class of problems than previously considered. Our released dataset, fine-tuned models, and implementation can be found at \\url{https://github.com/nweir127/CoGEX}."
                    },
                    {
                        "title": "Quasi-Dense Similarity Learning for Multiple Object Tracking",
                        "abstract": "Similarity learning has been recognized as a crucial step for object tracking. However, existing multiple object tracking methods only use sparse ground truth matching as the training objective, while ignoring the majority of the informative regions on the images. In this paper, we present Quasi-Dense Similarity Learning, which densely samples hundreds of region proposals on a pair of images for contrastive learning. We can directly combine this similarity learning with existing detection methods to build Quasi-Dense Tracking (QDTrack) without turning to displacement regression or motion priors. We also find that the resulting distinctive feature space admits a simple nearest neighbor search at the inference time. Despite its simplicity, QDTrack outperforms all existing methods on MOT, BDD100K, Waymo, and TAO tracking benchmarks. It achieves 68.7 MOTA at 20.3 FPS on MOT17 without using external training data. Compared to methods with similar detectors, it boosts almost 10 points of MOTA and significantly decreases the number of ID switches on BDD100K and Waymo datasets. Our code and trained models are available at http://vis.xyz/pub/qdtrack."
                    },
                    {
                        "title": "Improving Compositional Generalization with Latent Structure and Data Augmentation",
                        "abstract": "Generic unstructured neural networks have been shown to struggle on out-of-distribution compositional generalization. Compositional data augmentation via example recombination has transferred some prior knowledge about compositionality to such black-box neural models for several semantic parsing tasks, but this often required task-specific engineering or provided limited gains.   We present a more powerful data recombination method using a model called Compositional Structure Learner (CSL). CSL is a generative model with a quasi-synchronous context-free grammar backbone, which we induce from the training data. We sample recombined examples from CSL and add them to the fine-tuning data of a pre-trained sequence-to-sequence model (T5). This procedure effectively transfers most of CSL's compositional bias to T5 for diagnostic tasks, and results in a model even stronger than a T5-CSL ensemble on two real world compositional generalization tasks. This results in new state-of-the-art performance for these challenging semantic parsing tasks requiring generalization to both natural language variation and novel compositions of elements."
                    },
                    {
                        "title": "Evaluating the Impact of Model Scale for Compositional Generalization in Semantic Parsing",
                        "abstract": "Despite their strong performance on many tasks, pre-trained language models have been shown to struggle on out-of-distribution compositional generalization. Meanwhile, recent work has shown considerable improvements on many NLP tasks from model scaling. Can scaling up model size also improve compositional generalization in semantic parsing? We evaluate encoder-decoder models up to 11B parameters and decoder-only models up to 540B parameters, and compare model scaling curves for three different methods for applying a pre-trained language model to a new task: fine-tuning all parameters, prompt tuning, and in-context learning. We observe that fine-tuning generally has flat or negative scaling curves on out-of-distribution compositional generalization in semantic parsing evaluations. In-context learning has positive scaling curves, but is generally outperformed by much smaller fine-tuned models. Prompt-tuning can outperform fine-tuning, suggesting further potential improvements from scaling as it exhibits a more positive scaling curve. Additionally, we identify several error trends that vary with model scale. For example, larger models are generally better at modeling the syntax of the output space, but are also more prone to certain types of overfitting. Overall, our study highlights limitations of current techniques for effectively leveraging model scale for compositional generalization, while our analysis also suggests promising directions for future work."
                    },
                    {
                        "title": "Generate-and-Retrieve: use your predictions to improve retrieval for semantic parsing",
                        "abstract": "A common recent approach to semantic parsing augments sequence-to-sequence models by retrieving and appending a set of training samples, called exemplars. The effectiveness of this recipe is limited by the ability to retrieve informative exemplars that help produce the correct parse, which is especially challenging in low-resource settings. Existing retrieval is commonly based on similarity of query and exemplar inputs. We propose GandR, a retrieval procedure that retrieves exemplars for which outputs are also similar. GandRfirst generates a preliminary prediction with input-based retrieval. Then, it retrieves exemplars with outputs similar to the preliminary prediction which are used to generate a final prediction. GandR sets the state of the art on multiple low-resource semantic parsing tasks."
                    },
                    {
                        "title": "Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks",
                        "abstract": "The impressive performance of recent language models across a wide range of tasks suggests that they possess a degree of abstract reasoning skills. Are these skills general and transferable, or specialized to specific tasks seen during pretraining? To disentangle these effects, we propose an evaluation framework based on \"counterfactual\" task variants that deviate from the default assumptions underlying standard tasks. Across a suite of 11 tasks, we observe nontrivial performance on the counterfactual variants, but nevertheless find that performance substantially and consistently degrades compared to the default conditions. This suggests that while current LMs may possess abstract task-solving skills to an extent, they often also rely on narrow, non-transferable procedures for task-solving. These results motivate a more careful interpretation of language model performance that teases apart these aspects of behavior."
                    },
                    {
                        "title": "QDTrack: Quasi-Dense Similarity Learning for Appearance-Only Multiple Object Tracking",
                        "abstract": "Similarity learning has been recognized as a crucial step for object tracking. However, existing multiple object tracking methods only use sparse ground truth matching as the training objective, while ignoring the majority of the informative regions in images. In this paper, we present Quasi-Dense Similarity Learning, which densely samples hundreds of object regions on a pair of images for contrastive learning. We combine this similarity learning with multiple existing object detectors to build Quasi-Dense Tracking (QDTrack), which does not require displacement regression or motion priors. We find that the resulting distinctive feature space admits a simple nearest neighbor search at inference time for object association. In addition, we show that our similarity learning scheme is not limited to video data, but can learn effective instance similarity even from static input, enabling a competitive tracking performance without training on videos or using tracking supervision. We conduct extensive experiments on a wide variety of popular MOT benchmarks. We find that, despite its simplicity, QDTrack rivals the performance of state-of-the-art tracking methods on all benchmarks and sets a new state-of-the-art on the large-scale BDD100K MOT benchmark, while introducing negligible computational overhead to the detector."
                    },
                    {
                        "title": "Lookback Lens: Detecting and Mitigating Contextual Hallucinations in Large Language Models Using Only Attention Maps",
                        "abstract": "When asked to summarize articles or answer questions given a passage, large language models (LLMs) can hallucinate details and respond with unsubstantiated answers that are inaccurate with respect to the input context. This paper describes a simple approach for detecting such contextual hallucinations. We hypothesize that contextual hallucinations are related to the extent to which an LLM attends to information in the provided context versus its own generations. Based on this intuition, we propose a simple hallucination detection model whose input features are given by the ratio of attention weights on the context versus newly generated tokens (for each attention head). We find that a linear classifier based on these lookback ratio features is as effective as a richer detector that utilizes the entire hidden states of an LLM or a text-based entailment model. The lookback ratio-based detector -- Lookback Lens -- is found to transfer across tasks and even models, allowing a detector that is trained on a 7B model to be applied (without retraining) to a larger 13B model. We further apply this detector to mitigate contextual hallucinations, and find that a simple classifier-guided decoding approach is able to reduce the amount of hallucination, for example by 9.6% in the XSum summarization task."
                    },
                    {
                        "title": "Phenomenal Yet Puzzling: Testing Inductive Reasoning Capabilities of Language Models with Hypothesis Refinement",
                        "abstract": "The ability to derive underlying principles from a handful of observations and then generalize to novel situations -- known as inductive reasoning -- is central to human intelligence. Prior work suggests that language models (LMs) often fall short on inductive reasoning, despite achieving impressive success on research benchmarks. In this work, we conduct a systematic study of the inductive reasoning capabilities of LMs through iterative hypothesis refinement, a technique that more closely mirrors the human inductive process than standard input-output prompting. Iterative hypothesis refinement employs a three-step process: proposing, selecting, and refining hypotheses in the form of textual rules. By examining the intermediate rules, we observe that LMs are phenomenal hypothesis proposers (i.e., generating candidate rules), and when coupled with a (task-specific) symbolic interpreter that is able to systematically filter the proposed set of rules, this hybrid approach achieves strong results across inductive reasoning benchmarks that require inducing causal relations, language-like instructions, and symbolic concepts. However, they also behave as puzzling inductive reasoners, showing notable performance gaps between rule induction (i.e., identifying plausible rules) and rule application (i.e., applying proposed rules to instances), suggesting that LMs are proposing hypotheses without being able to actually apply the rules. Through empirical and human analyses, we further reveal several discrepancies between the inductive reasoning processes of LMs and humans, shedding light on both the potentials and limitations of using LMs in inductive reasoning tasks."
                    }
                ]
            },
            "d88dc78d-c0b4-4249-a73c-487177b6d04d": {
                "pk": "d88dc78d-c0b4-4249-a73c-487177b6d04d",
                "name": "Bailin Wang",
                "collaborators": [
                    "Ivan Titov",
                    "Mirella Lapata",
                    "Wei Lu",
                    "Yoon Kim",
                    "Jacob Andreas",
                    "Ekin Aky\u00fcrek",
                    "Yu Wang",
                    "Hongxia Jin",
                    "Henry Conklin",
                    "Kenny Smith"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Semantic Parsing",
                    "Machine Learning",
                    "Meta-Learning"
                ],
                "publications": [
                    {
                        "title": "Neural Segmental Hypergraphs for Overlapping Mention Recognition",
                        "abstract": "In this work, we propose a novel segmental hypergraph representation to model overlapping entity mentions that are prevalent in many practical datasets. We show that our model built on top of such a new representation is able to capture features and interactions that cannot be captured by previous models while maintaining a low time complexity for inference. We also present a theoretical analysis to formally assess how our representation is better than alternative representations reported in the literature in terms of representational power. Coupled with neural networks for feature learning, our model achieves the state-of-the-art performance in three benchmark datasets annotated with overlapping mentions."
                    },
                    {
                        "title": "Combining Spans into Entities: A Neural Two-Stage Approach for Recognizing Discontiguous Entities",
                        "abstract": "In medical documents, it is possible that an entity of interest not only contains a discontiguous sequence of words but also overlaps with another entity. Entities of such structures are intrinsically hard to recognize due to the large space of possible entity combinations. In this work, we propose a neural two-stage approach to recognize discontiguous and overlapping entities by decomposing this problem into two subtasks: 1) it first detects all the overlapping spans that either form entities on their own or present as segments of discontiguous entities, based on the representation of segmental hypergraph, 2) next it learns to combine these segments into discontiguous entities with a classifier, which filters out other incorrect combinations of segments. Two neural components are designed for these subtasks respectively and they are learned jointly using a shared encoder for text. Our model achieves the state-of-the-art performance in a standard dataset, even in the absence of external features that previous methods used."
                    },
                    {
                        "title": "Learning Semantic Parsers from Denotations with Latent Structured Alignments and Abstract Programs",
                        "abstract": "Semantic parsing aims to map natural language utterances onto machine interpretable meaning representations, aka programs whose execution against a real-world environment produces a denotation. Weakly-supervised semantic parsers are trained on utterance-denotation pairs treating programs as latent. The task is challenging due to the large search space and spuriousness of programs which may execute to the correct answer but do not generalize to unseen examples. Our goal is to instill an inductive bias in the parser to help it distinguish between spurious and correct programs. We capitalize on the intuition that correct programs would likely respect certain structural constraints were they to be aligned to the question (e.g., program fragments are unlikely to align to overlapping text spans) and propose to model alignments as structured latent variables. In order to make the latent-alignment framework tractable, we decompose the parsing task into (1) predicting a partial \"abstract program\" and (2) refining it while modeling structured alignments with differential dynamic programming. We obtain state-of-the-art performance on the WIKITABLEQUESTIONS and WIKISQL datasets. When compared to a standard attention baseline, we observe that the proposed structured-alignment mechanism is highly beneficial."
                    },
                    {
                        "title": "Meta-Learning for Domain Generalization in Semantic Parsing",
                        "abstract": "The importance of building semantic parsers which can be applied to new domains and generate programs unseen at training has long been acknowledged, and datasets testing out-of-domain performance are becoming increasingly available. However, little or no attention has been devoted to learning algorithms or objectives which promote domain generalization, with virtually all existing approaches relying on standard supervised learning. In this work, we use a meta-learning framework which targets zero-shot domain generalization for semantic parsing. We apply a model-agnostic training algorithm that simulates zero-shot parsing by constructing virtual train and test sets from disjoint domains. The learning objective capitalizes on the intuition that gradient steps that improve source-domain performance should also improve target-domain performance, thus encouraging a parser to generalize to unseen target domains. Experimental results on the (English) Spider and Chinese Spider datasets show that the meta-learning objective significantly boosts the performance of a baseline parser."
                    },
                    {
                        "title": "Learning from Executions for Semantic Parsing",
                        "abstract": "Semantic parsing aims at translating natural language (NL) utterances onto machine-interpretable programs, which can be executed against a real-world environment. The expensive annotation of utterance-program pairs has long been acknowledged as a major bottleneck for the deployment of contemporary neural models to real-life applications. In this work, we focus on the task of semi-supervised learning where a limited amount of annotated data is available together with many unlabeled NL utterances. Based on the observation that programs which correspond to NL utterances must be always executable, we propose to encourage a parser to generate executable programs for unlabeled utterances. Due to the large search space of executable programs, conventional methods that use approximations based on beam-search such as self-training and top-k marginal likelihood training, do not perform as well. Instead, we view the problem of learning from executions from the perspective of posterior regularization and propose a set of new training objectives. Experimental results on Overnight and GeoQuery show that our new objectives outperform conventional methods, bridging the gap between semi-supervised and supervised learning."
                    },
                    {
                        "title": "Structured Reordering for Modeling Latent Alignments in Sequence Transduction",
                        "abstract": "Despite success in many domains, neural models struggle in settings where train and test examples are drawn from different distributions. In particular, in contrast to humans, conventional sequence-to-sequence (seq2seq) models fail to generalize systematically, i.e., interpret sentences representing novel combinations of concepts (e.g., text segments) seen in training. Traditional grammar formalisms excel in such settings by implicitly encoding alignments between input and output segments, but are hard to scale and maintain. Instead of engineering a grammar, we directly model segment-to-segment alignments as discrete structured latent variables within a neural seq2seq model. To efficiently explore the large space of alignments, we introduce a reorder-first align-later framework whose central component is a neural reordering module producing {\\it separable} permutations. We present an efficient dynamic programming algorithm performing exact marginal inference of separable permutations, and, thus, enabling end-to-end differentiable training of our model. The resulting seq2seq model exhibits better systematic generalization than standard models on synthetic problems and NLP tasks (i.e., semantic parsing and machine translation)."
                    },
                    {
                        "title": "In-Context Language Learning: Architectures and Algorithms",
                        "abstract": "Large-scale neural language models exhibit a remarkable capacity for in-context learning (ICL): they can infer novel functions from datasets provided as input. Most of our current understanding of when and how ICL arises comes from LMs trained on extremely simple learning problems like linear regression and associative recall. There remains a significant gap between these model problems and the \"real\" ICL exhibited by LMs trained on large text corpora, which involves not just retrieval and function approximation but free-form generation of language and other structured outputs. In this paper, we study ICL through the lens of a new family of model problems we term in context language learning (ICLL). In ICLL, LMs are presented with a set of strings from a formal language, and must generate additional strings from the same language. We focus on in-context learning of regular languages generated by random finite automata. We evaluate a diverse set of neural sequence models (including several RNNs, Transformers, and state-space model variants) on regular ICLL tasks, aiming to answer three questions: (1) Which model classes are empirically capable of ICLL? (2) What algorithmic solutions do successful models implement to perform ICLL? (3) What architectural changes can improve ICLL in less performant models? We first show that Transformers significantly outperform neural sequence models with recurrent or convolutional representations on ICLL tasks. Next, we provide evidence that their ability to do so relies on specialized \"n-gram heads\" (higher-order variants of induction heads) that compute input-conditional next-token distributions. Finally, we show that hard-wiring these heads into neural models improves performance not just on ICLL, but natural language modeling -- improving the perplexity of 340M-parameter models by up to 1.14 points (6.7%) on the SlimPajama dataset."
                    },
                    {
                        "title": "A Neural Transition-based Model for Nested Mention Recognition",
                        "abstract": "It is common that entity mentions can contain other mentions recursively. This paper introduces a scalable transition-based method to model the nested structure of mentions. We first map a sentence with nested mentions to a designated forest where each mention corresponds to a constituent of the forest. Our shift-reduce based system then learns to construct the forest structure in a bottom-up manner through an action sequence whose maximal length is guaranteed to be three times of the sentence length. Based on Stack-LSTM which is employed to efficiently and effectively represent the states of the system in a continuous space, our system is further incorporated with a character-based component to capture letter-level patterns. Our model achieves the state-of-the-art results on ACE datasets, showing its effectiveness in detecting nested mentions."
                    },
                    {
                        "title": "Meta-Learning to Compositionally Generalize",
                        "abstract": "Natural language is compositional; the meaning of a sentence is a function of the meaning of its parts. This property allows humans to create and interpret novel sentences, generalizing robustly outside their prior experience. Neural networks have been shown to struggle with this kind of generalization, in particular performing poorly on tasks designed to assess compositional generalization (i.e. where training and testing distributions differ in ways that would be trivial for a compositional strategy to resolve). Their poor performance on these tasks may in part be due to the nature of supervised learning which assumes training and testing data to be drawn from the same distribution. We implement a meta-learning augmented version of supervised learning whose objective directly optimizes for out-of-distribution generalization. We construct pairs of tasks for meta-learning by sub-sampling existing training data. Each pair of tasks is constructed to contain relevant examples, as determined by a similarity metric, in an effort to inhibit models from memorizing their input. Experimental results on the COGS and SCAN datasets show that our similarity-driven meta-learning can improve generalization performance."
                    },
                    {
                        "title": "Hierarchical Phrase-based Sequence-to-Sequence Learning",
                        "abstract": "We describe a neural transducer that maintains the flexibility of standard sequence-to-sequence (seq2seq) models while incorporating hierarchical phrases as a source of inductive bias during training and as explicit constraints during inference. Our approach trains two models: a discriminative parser based on a bracketing transduction grammar whose derivation tree hierarchically aligns source and target phrases, and a neural seq2seq model that learns to translate the aligned phrases one-by-one. We use the same seq2seq model to translate at all phrase scales, which results in two inference modes: one mode in which the parser is discarded and only the seq2seq component is used at the sequence-level, and another in which the parser is combined with the seq2seq model. Decoding in the latter mode is done with the cube-pruned CKY algorithm, which is more involved but can make use of new translation rules during inference. We formalize our model as a source-conditioned synchronous grammar and develop an efficient variational inference algorithm for training. When applied on top of both randomly initialized and pretrained seq2seq models, we find that both inference modes performs well compared to baselines on small scale machine translation benchmarks."
                    },
                    {
                        "title": "Gated Linear Attention Transformers with Hardware-Efficient Training",
                        "abstract": "Transformers with linear attention allow for efficient parallel training but can simultaneously be formulated as an RNN with 2D (matrix-valued) hidden states, thus enjoying linear-time inference complexity. However, linear attention generally underperforms ordinary softmax attention. Moreover, current implementations of linear attention lack I/O-awareness and are thus slower than highly optimized implementations of softmax attention. This work describes a hardware-efficient algorithm for linear attention that trades off memory movement against parallelizability. The resulting implementation, dubbed FLASHLINEARATTENTION, is faster than FLASHATTENTION-2 (Dao, 2023) as a standalone layer even on short sequence lengths (e.g., 1K). We then generalize this algorithm to a more expressive variant of linear attention with data-dependent gates. When used as a replacement for the standard attention layer in Transformers, the resulting gated linear attention (GLA) Transformer is found to perform competitively against the LLaMA-architecture Transformer (Touvron et al., 2023) as well recent linear-time-inference baselines such as RetNet (Sun et al., 2023a) and Mamba (Gu & Dao, 2023) on moderate-scale language modeling experiments. GLA Transformer is especially effective at length generalization, enabling a model trained on 2K to generalize to sequences longer than 20K without significant perplexity degradations. For training speed, the GLA Transformer has higher throughput than a similarly-sized Mamba model."
                    },
                    {
                        "title": "Improving Generalization in Language Model-Based Text-to-SQL Semantic Parsing: Two Simple Semantic Boundary-Based Techniques",
                        "abstract": "Compositional and domain generalization present significant challenges in semantic parsing, even for state-of-the-art semantic parsers based on pre-trained language models (LMs). In this study, we empirically investigate improving an LM's generalization in semantic parsing with two simple techniques: at the token level, we introduce a token preprocessing method to preserve the semantic boundaries of tokens produced by LM tokenizers; at the sequence level, we propose to use special tokens to mark the boundaries of components aligned between input and output. Our experimental results on two text-to-SQL semantic parsing datasets show that our token preprocessing, although simple, can substantially improve the LM performance on both types of generalization, and our component boundary marking method is particularly helpful for compositional generalization."
                    }
                ]
            },
            "24f07358-8fe8-410e-b177-bcaa08e16161": {
                "pk": "24f07358-8fe8-410e-b177-bcaa08e16161",
                "name": "Danica J. Sutherland",
                "collaborators": [
                    "Wonho Bae",
                    "Arthur Gretton",
                    "Yi Ren",
                    "Michael Arbel",
                    "Namrata Deka",
                    "Jeff Schneider",
                    "Heiko Strathmann",
                    "Miko\u0142aj Bi\u0144kowski",
                    "Mohamad Amin Mohamadi",
                    "Jing Wang"
                ],
                "domain": [
                    "Kernel Methods",
                    "Active Learning",
                    "Generative Models",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Fixing an error in Caponnetto and de Vito (2007)",
                        "abstract": "The seminal paper of Caponnetto and de Vito (2007) provides minimax-optimal rates for kernel ridge regression in a very general setting. Its proof, however, contains an error in its bound on the effective dimensionality. In this note, we explain the mistake, provide a correct bound, and show that the main theorem remains true."
                    },
                    {
                        "title": "Unbiased estimators for the variance of MMD estimators",
                        "abstract": "The maximum mean discrepancy (MMD) is a kernel-based distance between probability distributions useful in many applications (Gretton et al. 2012), bearing a simple estimator with pleasing computational and statistical properties. Being able to efficiently estimate the variance of this estimator is very helpful to various problems in two-sample testing. Towards this end, Bounliphone et al. (2016) used the theory of U-statistics to derive estimators for the variance of an MMD estimator, and differences between two such estimators. Their estimator, however, drops lower-order terms, and is unnecessarily biased. We show in this note - extending and correcting work of Sutherland et al. (2017) - that we can find a truly unbiased estimator for the actual variance of both the squared MMD estimator and the difference of two correlated squared MMD estimators, at essentially no additional computational cost."
                    },
                    {
                        "title": "On the Error of Random Fourier Features",
                        "abstract": "Kernel methods give powerful, flexible, and theoretically grounded approaches to solving many problems in machine learning. The standard approach, however, requires pairwise evaluations of a kernel function, which can lead to scalability issues for very large datasets. Rahimi and Recht (2007) suggested a popular approach to handling this problem, known as random Fourier features. The quality of this approximation, however, is not well understood. We improve the uniform error bound of that paper, as well as giving novel understandings of the embedding's variance, approximation error, and use in some machine learning methods. We also point out that surprisingly, of the two main variants of those features, the more widely used is strictly higher-variance for the Gaussian kernel and has worse bounds."
                    },
                    {
                        "title": "Understanding Simplicity Bias towards Compositional Mappings via Learning Dynamics",
                        "abstract": "Obtaining compositional mappings is important for the model to generalize well compositionally. To better understand when and how to encourage the model to learn such mappings, we study their uniqueness through different perspectives. Specifically, we first show that the compositional mappings are the simplest bijections through the lens of coding length (i.e., an upper bound of their Kolmogorov complexity). This property explains why models having such mappings can generalize well. We further show that the simplicity bias is usually an intrinsic property of neural network training via gradient descent. That partially explains why some models spontaneously generalize well when they are trained appropriately."
                    },
                    {
                        "title": "MMD-B-Fair: Learning Fair Representations with Statistical Testing",
                        "abstract": "We introduce a method, MMD-B-Fair, to learn fair representations of data via kernel two-sample testing. We find neural features of our data where a maximum mean discrepancy (MMD) test cannot distinguish between representations of different sensitive groups, while preserving information about the target attributes. Minimizing the power of an MMD test is more difficult than maximizing it (as done in previous work), because the test threshold's complex behavior cannot be simply ignored. Our method exploits the simple asymptotics of block testing schemes to efficiently find fair representations without requiring complex adversarial optimization or generative modelling schemes widely used by existing work on fair representation learning. We evaluate our approach on various datasets, showing its ability to ``hide'' information about sensitive attributes, and its effectiveness in downstream transfer tasks."
                    },
                    {
                        "title": "Exploring Active Learning in Meta-Learning: Enhancing Context Set Labeling",
                        "abstract": "Most meta-learning methods assume that the (very small) context set used to establish a new task at test time is passively provided. In some settings, however, it is feasible to actively select which points to label; the potential gain from a careful choice is substantial, but the setting requires major differences from typical active learning setups. We clarify the ways in which active meta-learning can be used to label a context set, depending on which parts of the meta-learning process use active learning. Within this framework, we propose a natural algorithm based on fitting Gaussian mixtures for selecting which points to label; though simple, the algorithm also has theoretical motivation. The proposed algorithm outperforms state-of-the-art active learning methods when used with various meta-learning algorithms across several benchmark datasets."
                    },
                    {
                        "title": "Generalized Coverage for More Robust Low-Budget Active Learning",
                        "abstract": "The ProbCover method of Yehuda et al. is a well-motivated algorithm for active learning in low-budget regimes, which attempts to \"cover\" the data distribution with balls of a given radius at selected data points. We demonstrate, however, that the performance of this algorithm is extremely sensitive to the choice of this radius hyper-parameter, and that tuning it is quite difficult, with the original heuristic frequently failing. We thus introduce (and theoretically motivate) a generalized notion of \"coverage,\" including ProbCover's objective as a special case, but also allowing smoother notions that are far more robust to hyper-parameter choice. We propose an efficient greedy method to optimize this coverage, generalizing ProbCover's algorithm; due to its close connection to kernel herding, we call it \"MaxHerding.\" The objective can also be optimized non-greedily through a variant of $k$-medoids, clarifying the relationship to other low-budget active learning methods. In comprehensive experiments, MaxHerding surpasses existing active learning methods across multiple low-budget image classification benchmarks, and does so with less computational cost than most competitive methods."
                    },
                    {
                        "title": "Learning Dynamics of LLM Finetuning",
                        "abstract": "Learning dynamics, which describes how the learning of specific training examples influences the model's predictions on other examples, gives us a powerful tool for understanding the behavior of deep learning systems. We study the learning dynamics of large language models during different types of finetuning, by analyzing the step-wise decomposition of how influence accumulates among different potential responses. Our framework allows a uniform interpretation of many interesting observations about the training of popular algorithms for both instruction tuning and preference tuning. In particular, we propose a hypothetical explanation of why specific types of hallucination are strengthened after finetuning, e.g., the model might use phrases or facts in the response for question B to answer question A, or the model might keep repeating similar simple phrases when generating responses. We also extend our framework and highlight a unique \"squeezing effect\" to explain a previously observed phenomenon in off-policy direct preference optimization (DPO), where running DPO for too long makes even the desired outputs less likely. This framework also provides insights into where the benefits of on-policy DPO and other variants come from. The analysis not only provides a novel perspective of understanding LLM's finetuning but also inspires a simple, effective method to improve alignment performance."
                    },
                    {
                        "title": "Learning deep kernels for exponential family densities",
                        "abstract": "The kernel exponential family is a rich class of distributions, which can be fit efficiently and with statistical guarantees by score matching. Being required to choose a priori a simple kernel such as the Gaussian, however, limits its practical applicability. We provide a scheme for learning a kernel parameterized by a deep network, which can find complex location-dependent local features of the data geometry. This gives a very rich class of density models, capable of fitting complex structures on moderate-dimensional problems. Compared to deep density models fit via maximum likelihood, our approach provides a complementary set of strengths and tradeoffs: in empirical studies, the former can yield higher likelihoods, whereas the latter gives better estimates of the gradient of the log density, the score, which describes the distribution's shape."
                    },
                    {
                        "title": "Kernels on Sample Sets via Nonparametric Divergence Estimates",
                        "abstract": "Most machine learning algorithms, such as classification or regression, treat the individual data point as the object of interest. Here we consider extending machine learning algorithms to operate on groups of data points. We suggest treating a group of data points as an i.i.d. sample set from an underlying feature distribution for that group. Our approach employs kernel machines with a kernel on i.i.d. sample sets of vectors. We define certain kernel functions on pairs of distributions, and then use a nonparametric estimator to consistently estimate those functions based on sample sets. The projection of the estimated Gram matrix to the cone of symmetric positive semi-definite matrices enables us to use kernel machines for classification, regression, anomaly detection, and low-dimensional embedding in the space of distributions. We present several numerical experiments both on real and simulated datasets to demonstrate the advantages of our new approach."
                    },
                    {
                        "title": "On gradient regularizers for MMD GANs",
                        "abstract": "We propose a principled method for gradient-based regularization of the critic of GAN-like models trained by adversarially optimizing the kernel of a Maximum Mean Discrepancy (MMD). We show that controlling the gradient of the critic is vital to having a sensible loss function, and devise a method to enforce exact, analytical gradient constraints at no additional cost compared to existing approximate techniques based on additive regularizers. The new loss function is provably continuous, and experiments show that it stabilizes and accelerates training, giving image generation models that outperform state-of-the art methods on $160 \\times 160$ CelebA and $64 \\times 64$ unconditional ImageNet."
                    },
                    {
                        "title": "On Uniform Convergence and Low-Norm Interpolation Learning",
                        "abstract": "We consider an underdetermined noisy linear regression model where the minimum-norm interpolating predictor is known to be consistent, and ask: can uniform convergence in a norm ball, or at least (following Nagarajan and Kolter) the subset of a norm ball that the algorithm selects on a typical input set, explain this success? We show that uniformly bounding the difference between empirical and population errors cannot show any learning in the norm ball, and cannot show consistency for any set, even one depending on the exact algorithm and distribution. But we argue we can explain the consistency of the minimal-norm interpolator with a slightly weaker, yet standard, notion: uniform convergence of zero-error predictors in a norm ball. We use this to bound the generalization error of low- (but not minimal-) norm interpolating predictors."
                    },
                    {
                        "title": "Better Supervisory Signals by Observing Learning Paths",
                        "abstract": "Better-supervised models might have better performance. In this paper, we first clarify what makes for good supervision for a classification problem, and then explain two existing label refining methods, label smoothing and knowledge distillation, in terms of our proposed criterion. To further answer why and how better supervision emerges, we observe the learning path, i.e., the trajectory of the model's predictions during training, for each training sample. We find that the model can spontaneously refine \"bad\" labels through a \"zig-zag\" learning path, which occurs on both toy and real datasets. Observing the learning path not only provides a new perspective for understanding knowledge distillation, overfitting, and learning dynamics, but also reveals that the supervisory signal of a teacher network can be very unstable near the best points in training on real tasks. Inspired by this, we propose a new knowledge distillation scheme, Filter-KD, which improves downstream classification performance in various settings."
                    },
                    {
                        "title": "A Fast, Well-Founded Approximation to the Empirical Neural Tangent Kernel",
                        "abstract": "Empirical neural tangent kernels (eNTKs) can provide a good understanding of a given network's representation: they are often far less expensive to compute and applicable more broadly than infinite width NTKs. For networks with O output units (e.g. an O-class classifier), however, the eNTK on N inputs is of size $NO \\times NO$, taking $O((NO)^2)$ memory and up to $O((NO)^3)$ computation. Most existing applications have therefore used one of a handful of approximations yielding $N \\times N$ kernel matrices, saving orders of magnitude of computation, but with limited to no justification. We prove that one such approximation, which we call \"sum of logits\", converges to the true eNTK at initialization for any network with a wide final \"readout\" layer. Our experiments demonstrate the quality of this approximation for various uses across a range of settings."
                    },
                    {
                        "title": "Learning Deep Kernels for Non-Parametric Independence Testing",
                        "abstract": "The Hilbert-Schmidt Independence Criterion (HSIC) is a powerful tool for nonparametric detection of dependence between random variables. It crucially depends, however, on the selection of reasonable kernels; commonly-used choices like the Gaussian kernel, or the kernel that yields the distance covariance, are sufficient only for amply sized samples from data distributions with relatively simple forms of dependence. We propose a scheme for selecting the kernels used in an HSIC-based independence test, based on maximizing an estimate of the asymptotic test power. We prove that maximizing this estimate indeed approximately maximizes the true power of the test, and demonstrate that our learned kernels can identify forms of structured dependence between random variables in various experiments."
                    },
                    {
                        "title": "Efficient and principled score estimation with Nystr\u00f6m kernel exponential families",
                        "abstract": "We propose a fast method with statistical guarantees for learning an exponential family density model where the natural parameter is in a reproducing kernel Hilbert space, and may be infinite-dimensional. The model is learned by fitting the derivative of the log density, the score, thus avoiding the need to compute a normalization constant. Our approach improves the computational efficiency of an earlier solution by using a low-rank, Nystr\\\"om-like solution. The new solution retains the consistency and convergence rates of the full-rank solution (exactly in Fisher distance, and nearly in other distances), with guarantees on the degree of cost and storage reduction. We evaluate the method in experiments on density estimation and in the construction of an adaptive Hamiltonian Monte Carlo sampler. Compared to an existing score learning approach using a denoising autoencoder, our estimator is empirically more data-efficient when estimating the score, runs faster, and has fewer parameters (which can be tuned in a principled and interpretable way), in addition to providing statistical guarantees."
                    },
                    {
                        "title": "Demystifying MMD GANs",
                        "abstract": "We investigate the training and performance of generative adversarial networks using the Maximum Mean Discrepancy (MMD) as critic, termed MMD GANs. As our main theoretical contribution, we clarify the situation with bias in GAN loss functions raised by recent work: we show that gradient estimators used in the optimization process for both MMD GANs and Wasserstein GANs are unbiased, but learning a discriminator based on samples leads to biased gradients for the generator parameters. We also discuss the issue of kernel choice for the MMD critic, and characterize the kernel corresponding to the energy distance used for the Cramer GAN critic. Being an integral probability metric, the MMD benefits from training strategies recently developed for Wasserstein GANs. In experiments, the MMD GAN is able to employ a smaller critic network than the Wasserstein GAN, resulting in a simpler and faster-training algorithm with matching performance. We also propose an improved measure of GAN convergence, the Kernel Inception Distance, and show how to use it to dynamically adapt learning rates during GAN training."
                    },
                    {
                        "title": "Evaluating Graph Generative Models with Contrastively Learned Features",
                        "abstract": "A wide range of models have been proposed for Graph Generative Models, necessitating effective methods to evaluate their quality. So far, most techniques use either traditional metrics based on subgraph counting, or the representations of randomly initialized Graph Neural Networks (GNNs). We propose using representations from contrastively trained GNNs, rather than random GNNs, and show this gives more reliable evaluation metrics. Neither traditional approaches nor GNN-based approaches dominate the other, however: we give examples of graphs that each approach is unable to distinguish. We demonstrate that Graph Substructure Networks (GSNs), which in a way combine both approaches, are better at distinguishing the distances between graph datasets."
                    },
                    {
                        "title": "Making Look-Ahead Active Learning Strategies Feasible with Neural Tangent Kernels",
                        "abstract": "We propose a new method for approximating active learning acquisition strategies that are based on retraining with hypothetically-labeled candidate data points. Although this is usually infeasible with deep networks, we use the neural tangent kernel to approximate the result of retraining, and prove that this approximation works asymptotically even in an active learning setup -- approximating \"look-ahead\" selection criteria with far less computation required. This also enables us to conduct sequential active learning, i.e. updating the model in a streaming regime, without needing to retrain the model with SGD after adding each new data point. Moreover, our querying strategy, which better understands how the model's predictions will change by adding new data points in comparison to the standard (\"myopic\") criteria, beats other look-ahead strategies by large margins, and achieves equal or better performance compared to state-of-the-art methods on several benchmark datasets in pool-based active learning."
                    }
                ]
            }
        }
    },
    "2309.01786": {
        "paper_data": {
            "title": "Safe and Robust Watermark Injection with a Single OoD Image",
            "url": "http://arxiv.org/abs/2309.01786v2",
            "arxiv_id": "2309.01786",
            "authors": [
                "Shuyang Yu",
                "Junyuan Hong",
                "Haobo Zhang",
                "Haotao Wang",
                "Zhangyang Wang",
                "Jiayu Zhou"
            ],
            "abstract": "Training a high-performance deep neural network requires large amounts of data and computational resources. Protecting the intellectual property (IP) and commercial ownership of a deep model is challenging yet increasingly crucial. A major stream of watermarking strategies implants verifiable backdoor triggers by poisoning training samples, but these are often unrealistic due to data privacy and safety concerns and are vulnerable to minor model changes such as fine-tuning. To overcome these challenges, we propose a safe and robust backdoor-based watermark injection technique that leverages the diverse knowledge from a single out-of-distribution (OoD) image, which serves as a secret key for IP verification. The independence of training data makes it agnostic to third-party promises of IP security. We induce robustness via random perturbation of model parameters during watermark injection to defend against common watermark removal attacks, including fine-tuning, pruning, and model extraction. Our experimental results demonstrate that the proposed watermarking approach is not only time- and sample-efficient without training data, but also robust against the watermark removal attacks above.",
            "introduction": "   1 Introduction  In the era of deep learning, training a high-performance large model requires curating a massive amount of training data from different sources, powerful computational resources, and often great efforts from human experts. For example, large language models such as GPT-3 are large models trained on private datasets, incurring a significant training cost\u00a0(Floridi & Chiriatti, 2020). The risk of illegal reproduction or duplication of such high-value DNN models is a growing concern. The recent Facebook leaked LLAMA model provides a notable example of this risk\u00a0(Hern, 2023). Therefore, it is essential to protect the intellectual property of the model and the rights of the model owners. Recently, watermarking\u00a0(Adi et\u00a0al., 2018; Darvish\u00a0Rouhani et\u00a0al., 2019; Uchida et\u00a0al., 2017; Zhang et\u00a0al., 2018; Chen et\u00a0al., 2021; Li et\u00a0al., 2021) has been introduced to protect the copyright of the DNNs. Most existing watermarking methods can be categorized into two mainstreams, including parameter-embedding\u00a0(Kuribayashi et\u00a0al., 2021; Uchida et\u00a0al., 2017; Mehta et\u00a0al., 2022) and backdoor-based\u00a0(Goldblum et\u00a0al., 2022; Li et\u00a0al., 2022) techniques. Parameter-embedding techniques require white-box access to the suspicious model, which is often unrealistic in practical detection scenarios. This paper places emphasis on backdoor-based approaches, which taint the training dataset by incorporating trigger patches into a set of images referred to as verification samples (trigger set), and modifying the labels to a designated class, forcing the model to memorize the trigger pattern during fine-tuning. Then the owner of the model can perform an intellectual property (IP) inspection by assessing the correspondence between the model\u2019s outputs on the verification samples with the trigger and the intended target labels.   Existing backdoor-based watermarking methods suffer from major challenges in safety, efficiency, and robustness. Typically injection of backdoors requires full or partial access to the original training data. When protecting models, such access can be prohibitive, mostly due to data safety and confidentiality. For example, someone trying to protect a model fine-tuned upon a foundation model and a model publisher vending models uploaded by their users. Another example is an independent IP protection department or a third party that is in charge of model protection for redistribution. Yet another scenario is federated learning\u00a0(Kone\u010dn\u1ef3 et\u00a0al., 2016), where the server does not have access to any in-distribution (ID) data, but is motivated to inject a watermark to protect the ownership of the global model. Despite the high practical demands, watermark injection without training data is barely explored. Although some existing methods tried to export or synthesize out-of-distribution (OoD) samples as triggers to insert watermark\u00a0(Wang et\u00a0al., 2022b; Zhang et\u00a0al., 2018), the original training data is still essential to maintain the utility of the model, i.e., prediction performance on clean samples. Li & Wang (2022) proposed a strategy that adopts a Data-Free Distillation (DFD) process to train a generator and uses it to produce surrogate training samples. However, training the generator is time-consuming and may take hundreds of epochs\u00a0(Fang et\u00a0al., 2019). Another critical issue with backdoor-based watermarks is their known vulnerability against minor model changes, such as fine-tuning\u00a0(Adi et\u00a0al., 2018; Uchida et\u00a0al., 2017; Garg et\u00a0al., 2020), and this vulnerability greatly limited the practical applications of backdoor-based watermarks.   To address these challenges, in this work, we propose a practical watermark strategy that is based on efficient fine-tuning, using safe",
            "references": [
                {
                    "title": "Revisiting Data-Free Knowledge Distillation with Poisoned Teachers",
                    "abstract": "Data-free knowledge distillation (KD) helps transfer knowledge from a pre-trained model (known as the teacher model) to a smaller model (known as the student model) without access to the original training data used for training the teacher model. However, the security of the synthetic or out-of-distribution (OOD) data required in data-free KD is largely unknown and under-explored. In this work, we make the first effort to uncover the security risk of data-free KD w.r.t. untrusted pre-trained models. We then propose Anti-Backdoor Data-Free KD (ABD), the first plug-in defensive method for data-free KD methods to mitigate the chance of potential backdoors being transferred. We empirically evaluate the effectiveness of our proposed ABD in diminishing transferred backdoor knowledge while maintaining compatible downstream performances as the vanilla KD. We envision this work as a milestone for alarming and mitigating the potential backdoors in data-free KD. Codes are released at https://github.com/illidanlab/ABD."
                },
                {
                    "title": "Sharpness-Aware Data Poisoning Attack",
                    "abstract": "Recent research has highlighted the vulnerability of Deep Neural Networks (DNNs) against data poisoning attacks. These attacks aim to inject poisoning samples into the models' training dataset such that the trained models have inference failures. While previous studies have executed different types of attacks, one major challenge that greatly limits their effectiveness is the uncertainty of the re-training process after the injection of poisoning samples, including the re-training initialization or algorithms. To address this challenge, we propose a novel attack method called ''Sharpness-Aware Data Poisoning Attack (SAPA)''. In particular, it leverages the concept of DNNs' loss landscape sharpness to optimize the poisoning effect on the worst re-trained model. It helps enhance the preservation of the poisoning effect, regardless of the specific retraining procedure employed. Extensive experiments demonstrate that SAPA offers a general and principled strategy that significantly enhances various types of poisoning attacks."
                },
                {
                    "title": "Free Fine-tuning: A Plug-and-Play Watermarking Scheme for Deep Neural Networks",
                    "abstract": "Watermarking has been widely adopted for protecting the intellectual property (IP) of Deep Neural Networks (DNN) to defend the unauthorized distribution. Unfortunately, studies have shown that the popular data-poisoning DNN watermarking scheme via tedious model fine-tuning on a poisoned dataset (carefully-crafted sample-label pairs) is not efficient in tackling the tasks on challenging datasets and production-level DNN model protection. To address the aforementioned limitation, in this paper, we propose a plug-and-play watermarking scheme for DNN models by injecting an independent proprietary model into the target model to serve the watermark embedding and ownership verification. In contrast to the prior studies, our proposed method by incorporating a proprietary model is free of target model fine-tuning without involving any parameters update of the target model, thus the fidelity is well preserved and scalable to challenging real tasks. Experimental results on real-world challenging datasets (e.g., ImageNet) and production-level DNN models demonstrated its effectiveness, fidelity w.r.t. the functionality preservation of the target model, robustness against popular watermark removal attacks, and the plug-and-play deployment. The source code and models are available at https://github.com/AntigoneRandy/PTYNet."
                },
                {
                    "title": "Trap and Replace: Defending Backdoor Attacks by Trapping Them into an Easy-to-Replace Subnetwork",
                    "abstract": "Deep neural networks (DNNs) are vulnerable to backdoor attacks. Previous works have shown it extremely challenging to unlearn the undesired backdoor behavior from the network, since the entire network can be affected by the backdoor samples. In this paper, we propose a brand-new backdoor defense strategy, which makes it much easier to remove the harmful influence of backdoor samples from the model. Our defense strategy, Trap and Replace, consists of two stages. In the first stage, we bait and trap the backdoors in a small and easy-to-replace subnetwork. Specifically, we add an auxiliary image reconstruction head on top of the stem network shared with a light-weighted classification head. The intuition is that the auxiliary image reconstruction task encourages the stem network to keep sufficient low-level visual features that are hard to learn but semantically correct, instead of overfitting to the easy-to-learn but semantically incorrect backdoor correlations. As a result, when trained on backdoored datasets, the backdoors are easily baited towards the unprotected classification head, since it is much more vulnerable than the shared stem, leaving the stem network hardly poisoned. In the second stage, we replace the poisoned light-weighted classification head with an untainted one, by re-training it from scratch only on a small holdout dataset with clean samples, while fixing the stem network. As a result, both the stem and the classification head in the final network are hardly affected by backdoor training samples. We evaluate our method against ten different backdoor attacks. Our method outperforms previous state-of-the-art methods by up to 20.57%, 9.80%, and 13.72% attack success rate and on-average 3.14%, 1.80%, and 1.21% clean classification accuracy on CIFAR10, GTSRB, and ImageNet-12, respectively. Code is available at https://github.com/VITA-Group/Trap-and-Replace-Backdoor-Defense."
                },
                {
                    "title": "Knowledge-Free Black-Box Watermark and Ownership Proof for Image Classification Neural Networks",
                    "abstract": "Watermarking has become a plausible candidate for ownership verification and intellectual property protection of deep neural networks. Regarding image classification neural networks, current watermarking schemes uniformly resort to backdoor triggers. However, injecting a backdoor into a neural network requires knowledge of the training dataset, which is usually unavailable in the real-world commercialization. Meanwhile, established watermarking schemes oversight the potential damage of exposed evidence during ownership verification and the watermarking algorithms themselves. Those concerns decline current watermarking schemes from industrial applications. To confront these challenges, we propose a knowledge-free black-box watermarking scheme for image classification neural networks. The image generator obtained from a data-free distillation process is leveraged to stabilize the network's performance during the backdoor injection. A delicate encoding and verification protocol is designed to ensure the scheme's security against knowledgable adversaries. We also give a pioneering analysis of the capacity of the watermarking scheme. Experiment results proved the functionality-preserving capability and security of the proposed watermarking scheme."
                },
                {
                    "title": "AIME: Watermarking AI Models by Leveraging Errors",
                    "abstract": "The recent evolution of deep neural networks (DNNs) has made running complex data analytics tasks, which range from natural language processing, object detection to autonomous cars, artificial intelligence (AI) warfare, cloud, healthcare, industrial robots, and edge devices feasible. The benefits of AI are indisputable. However, there are several concerns regarding the security of the deployed AI models, such as reverse engineering and Intellectual Property (IP) piracy. Accumulating a sufficiently large amount of data - building, training, improvement, and model deployment require immense human and computational power, making the process expensive. Therefore, it is of utmost importance to protect the model against IP infringement. We propose AIME, a novel watermarking framework that captures model inaccuracy during the training phase and converts it into the owner-specific unique signature. The watermark is embedded within the class mispredictions of the DNN model. Watermark extraction is performed when the model is queried by an owner-specific sequence of key inputs, and the signature is decoded from the sequence of model predictions. AIME works with negligible watermark embedding runtime overhead while preserving the accurate functionality of the DNN. We have performed a comprehensive evaluation of AIME, which models on MNIST, Fashion-MNIST, and CIFAR-10 dataset and corroborated its effectiveness, robustness, and performance."
                },
                {
                    "title": "Copy, Right? A Testing Framework for Copyright Protection of Deep Learning Models",
                    "abstract": "Deep learning models, especially those large-scale and high-performance ones, can be very costly to train, demanding a considerable amount of data and computational resources. As a result, deep learning models have become one of the most valuable assets in modern artificial intelligence. Unauthorized duplication or reproduction of deep learning models can lead to copyright infringement and cause huge economic losses to model owners, calling for effective copyright protection techniques. Existing protection techniques are mostly based on watermarking, which embeds an owner-specified watermark into the model. While being able to provide exact ownership verification, these techniques are 1) invasive, i.e., they need to tamper with the training process, which may affect the model utility or introduce new security risks into the model; 2) prone to adaptive attacks that attempt to remove/replace the watermark or adversarially block the retrieval of the watermark; and 3) not robust to the emerging model extraction attacks. Latest fingerprinting work on deep learning models, though being non-invasive, also falls short when facing the diverse and ever-growing attack scenarios.In this paper, we propose a novel testing framework for deep learning copyright protection: DEEPJUDGE. DEEPJUDGE quantitatively tests the similarities between two deep learning models: a victim model and a suspect model. It leverages a diverse set of testing metrics and efficient test case generation algorithms to produce a chain of supporting evidence to help determine whether a suspect model is a copy of the victim model. Advantages of DEEPJUDGE include: 1) non-invasive, as it works directly on the model and does not tamper with the training process; 2) efficient, as it only needs a small set of seed test cases and a quick scan of the two models; 3) flexible, i.e., it can easily incorporate new testing metrics or test case generation methods to obtain more confident and robust judgement; and 4) fairly robust to model extraction attacks and adaptive attacks. We verify the effectiveness of DEEPJUDGE under three typical copyright infringement scenarios, including model finetuning, pruning and extraction, via extensive experiments on both image classification and speech recognition datasets with a variety of model architectures."
                },
                {
                    "title": "You are caught stealing my winning lottery ticket! Making a lottery ticket claim its ownership",
                    "abstract": "Despite tremendous success in many application scenarios, the training and inference costs of using deep learning are also rapidly increasing over time. The lottery ticket hypothesis (LTH) emerges as a promising framework to leverage a special sparse subnetwork (i.e., winning ticket) instead of a full model for both training and inference, that can lower both costs without sacrificing the performance. The main resource bottleneck of LTH is however the extraordinary cost to find the sparse mask of the winning ticket. That makes the found winning ticket become a valuable asset to the owners, highlighting the necessity of protecting its copyright. Our setting adds a new dimension to the recently soaring interest in protecting against the intellectual property (IP) infringement of deep models and verifying their ownerships, since they take owners' massive/unique resources to develop or train. While existing methods explored encrypted weights or predictions, we investigate a unique way to leverage sparse topological information to perform lottery verification, by developing several graph-based signatures that can be embedded as credentials. By further combining trigger set-based methods, our proposal can work in both white-box and black-box verification scenarios. Through extensive experiments, we demonstrate the effectiveness of lottery verification in diverse models (ResNet-20, ResNet-18, ResNet-50) on CIFAR-10 and CIFAR-100. Specifically, our verification is shown to be robust to removal attacks such as model fine-tuning and pruning, as well as several ambiguity attacks. Our codes are available at https://github.com/VITA-Group/NO-stealing-LTH."
                },
                {
                    "title": "White-Box Watermarking Scheme for Fully-Connected Layers in Fine-Tuning Model",
                    "abstract": "For the protection of trained deep neural network(DNN) models, embedding watermarks into the weights of the DNN model have been considered. However, the amount of change in the weights is large in the conventional methods, and it is reported that the existence of hidden watermark can be detected from the analysis of weight variance. This helps attackers to modify the watermark by effectively adding noise to the weight. In this paper, we focus on the fully-connected layers of fine-tuning models and apply a quantization-based watermarking method to the weights sampled from the layers. The advantage of the proposed method is that the change caused by watermark embedding is much smaller and the distortion converges gradually without using any loss function. The validity of the proposed method was evaluated by varying the conditions during the training of DNN model. The results shows the impact of training for DNN model, effectiveness of the embedding method, and high robustness against pruning attacks."
                },
                {
                    "title": "Rethinking the Backdoor Attacks\u2019 Triggers: A Frequency Perspective",
                    "abstract": "Backdoor attacks have been considered a severe security threat to deep learning. Such attacks can make models perform abnormally on inputs with predefined triggers and still retain state-of-the-art performance on clean data. While backdoor attacks have been thoroughly investigated in the image domain from both attackers\u2019 and defenders\u2019 sides, an analysis in the frequency domain has been missing thus far.This paper first revisits existing backdoor triggers from a frequency perspective and performs a comprehensive analysis. Our results show that many current backdoor attacks exhibit severe high-frequency artifacts, which persist across different datasets and resolutions. We further demonstrate these high-frequency artifacts enable a simple way to detect existing backdoor triggers at a detection rate of 98.50% without prior knowledge of the attack details and the target model. Acknowledging previous attacks\u2019 weaknesses, we propose a practical way to create smooth backdoor triggers without high-frequency artifacts and study their detectability. We show that existing defense works can benefit by incorporating these smooth triggers into their design consideration. Moreover, we show that the detector tuned over stronger smooth triggers can generalize well to unseen weak smooth triggers. In short, our work emphasizes the importance of considering frequency analysis when designing both backdoor attacks and defenses in deep learning."
                },
                {
                    "title": "Dataset Security for Machine Learning: Data Poisoning, Backdoor Attacks, and Defenses",
                    "abstract": "As machine learning systems grow in scale, so do their training data requirements, forcing practitioners to automate and outsource the curation of training data in order to achieve state-of-the-art performance. The absence of trustworthy human supervision over the data collection process exposes organizations to security vulnerabilities; training data can be manipulated to control and degrade the downstream behaviors of learned models. The goal of this work is to systematically categorize and discuss a wide range of dataset vulnerabilities and exploits, approaches for defending against these threats, and an array of open problems in this space."
                },
                {
                    "title": "Energy-based Out-of-distribution Detection",
                    "abstract": "Determining whether inputs are out-of-distribution (OOD) is an essential building block for safely deploying machine learning models in the open world. However, previous methods relying on the softmax confidence score suffer from overconfident posterior distributions for OOD data. We propose a unified framework for OOD detection that uses an energy score. We show that energy scores better distinguish in- and out-of-distribution samples than the traditional approach using the softmax scores. Unlike softmax confidence scores, energy scores are theoretically aligned with the probability density of the inputs and are less susceptible to the overconfidence issue. Within this framework, energy can be flexibly used as a scoring function for any pre-trained neural classifier as well as a trainable cost function to shape the energy surface explicitly for OOD detection. On a CIFAR-10 pre-trained WideResNet, using the energy score reduces the average FPR (at TPR 95%) by 18.03% compared to the softmax confidence score. With energy-based training, our method outperforms the state-of-the-art on common benchmarks."
                },
                {
                    "title": "ES Attack: Model Stealing Against Deep Neural Networks Without Data Hurdles",
                    "abstract": "Deep neural networks (DNNs) have become the essential components for various commercialized machine learning services, such as Machine Learning as a Service (MLaaS). Recent studies show that machine learning services face severe privacy threats - well-trained DNNs owned by MLaaS providers can be stolen through public APIs, namely model stealing attacks. However, most existing works undervalued the impact of such attacks, where a successful attack has to acquire confidential training data or auxiliary data regarding the victim DNN. In this paper, we propose ES Attack, a novel model stealing attack without any data hurdles. By using heuristically generated synthetic data, ES Attack iteratively trains a substitute model and eventually achieves a functionally equivalent copy of the victim DNN. The experimental results reveal the severity of ES Attack: i) ES Attack successfully steals the victim model without data hurdles, and ES Attack even outperforms most existing model stealing attacks using auxiliary data in terms of model accuracy; ii) most countermeasures are ineffective in defending ES Attack; iii) ES Attack facilitates further attacks relying on the stolen model."
                },
                {
                    "title": "Can Adversarial Weight Perturbations Inject Neural Backdoors",
                    "abstract": "Adversarial machine learning has exposed several security hazards of neural models. Thus far, the concept of an \"adversarial perturbation\" has exclusively been used with reference to the input space referring to a small, imperceptible change which can cause a ML model to err. In this work we extend the idea of \"adversarial perturbations\" to the space of model weights, specifically to inject backdoors in trained DNNs, which exposes a security risk of publicly available trained models. Here, injecting a backdoor refers to obtaining a desired outcome from the model when a trigger pattern is added to the input, while retaining the original predictions on a non-triggered input. From the perspective of an adversary, we characterize these adversarial perturbations to be constrained within an \u2113\u221e norm around the original model weights. We introduce adversarial perturbations in model weights using a composite loss on the predictions of the original model and the desired trigger through projected gradient descent. Our results show that backdoors can be successfully injected with a very small average relative change in model weight values for several CV and NLP applications."
                },
                {
                    "title": "Adversarial Weight Perturbation Helps Robust Generalization",
                    "abstract": "The study on improving the robustness of deep neural networks against adversarial examples grows rapidly in recent years. Among them, adversarial training is the most promising one, which flattens the input loss landscape (loss change with respect to input) via training on adversarially perturbed examples. However, how the widely used weight loss landscape (loss change with respect to weight) performs in adversarial training is rarely explored. In this paper, we investigate the weight loss landscape from a new perspective, and identify a clear correlation between the flatness of weight loss landscape and robust generalization gap. Several well-recognized adversarial training improvements, such as early stopping, designing new objective functions, or leveraging unlabeled data, all implicitly flatten the weight loss landscape. Based on these observations, we propose a simple yet effective Adversarial Weight Perturbation (AWP) to explicitly regularize the flatness of weight loss landscape, forming a double-perturbation mechanism in the adversarial training framework that adversarially perturbs both inputs and weights. Extensive experiments demonstrate that AWP indeed brings flatter weight loss landscape and can be easily incorporated into various existing adversarial training methods to further boost their adversarial robustness."
                },
                {
                    "title": "Comparing Rewinding and Fine-tuning in Neural Network Pruning",
                    "abstract": "Many neural network pruning algorithms proceed in three steps: train the network to completion, remove unwanted structure to compress the network, and retrain the remaining structure to recover lost accuracy. The standard retraining technique, fine-tuning, trains the unpruned weights from their final trained values using a small fixed learning rate. In this paper, we compare fine-tuning to alternative retraining techniques. Weight rewinding (as proposed by Frankle et al., (2019)), rewinds unpruned weights to their values from earlier in training and retrains them from there using the original training schedule. Learning rate rewinding (which we propose) trains the unpruned weights from their final values using the same learning rate schedule as weight rewinding. Both rewinding techniques outperform fine-tuning, forming the basis of a network-agnostic pruning algorithm that matches the accuracy and compression ratios of several more network-specific state-of-the-art techniques."
                },
                {
                    "title": "Entangled Watermarks as a Defense against Model Extraction",
                    "abstract": "Machine learning involves expensive data collection and training procedures. Model owners may be concerned that valuable intellectual property can be leaked if adversaries mount model extraction attacks. Because it is difficult to defend against model extraction without sacrificing significant prediction accuracy, watermarking leverages unused model capacity to have the model overfit to outlier input-output pairs, which are not sampled from the task distribution and are only known to the defender. The defender then demonstrates knowledge of the input-output pairs to claim ownership of the model at inference. The effectiveness of watermarks remains limited because they are distinct from the task distribution and can thus be easily removed through compression or other forms of knowledge transfer. \nWe introduce Entangled Watermarking Embeddings (EWE). Our approach encourages the model to learn common features for classifying data that is sampled from the task distribution, but also data that encodes watermarks. An adversary attempting to remove watermarks that are entangled with legitimate data is also forced to sacrifice performance on legitimate data. Experiments on MNIST, Fashion-MNIST, and Google Speech Commands validate that the defender can claim model ownership with 95% confidence after less than 10 queries to the stolen copy, at a modest cost of 1% accuracy in the defended model's performance."
                },
                {
                    "title": "Data-Free Adversarial Distillation",
                    "abstract": "Knowledge Distillation (KD) has made remarkable progress in the last few years and become a popular paradigm for model compression and knowledge transfer. However, almost all existing KD algorithms are data-driven, i.e., relying on a large amount of original training data or alternative data, which is usually unavailable in real-world scenarios. In this paper, we devote ourselves to this challenging problem and propose a novel adversarial distillation mechanism to craft a compact student model without any real-world data. We introduce a model discrepancy to quantificationally measure the difference between student and teacher models and construct an optimizable upper bound. In our work, the student and the teacher jointly act the role of the discriminator to reduce this discrepancy, when a generator adversarially produces some \"hard samples\" to enlarge it. Extensive experiments demonstrate that the proposed data-free method yields comparable performance to existing data-driven methods. More strikingly, our approach can be directly extended to semantic segmentation, which is more complicated than classification, and our approach achieves state-of-the-art results. Code and pretrained models are available at this https URL."
                },
                {
                    "title": "TBT: Targeted Neural Network Attack With Bit Trojan",
                    "abstract": "Security of modern Deep Neural Networks (DNNs) is under severe scrutiny as the deployment of these models become widespread in many intelligence-based applications. Most recently, DNNs are attacked through Trojan which can effectively infect the model during the training phase and get activated only through specific input patterns (i.e, trigger) during inference. In this work, for the first time, we propose a novel Targeted Bit Trojan(TBT) method, which can insert a targeted neural Trojan into a DNN through bit-flip attack. Our algorithm efficiently generates a trigger specifically designed to locate certain vulnerable bits of DNN weights stored in main memory (i.e., DRAM). The objective is that once the attacker flips these vulnerable bits, the network still operates with normal inference accuracy with benign input. However, when the attacker activates the trigger by embedding it with any input, the network is forced to classify all inputs to a certain target class. We demonstrate that flipping only several vulnerable bits identified by our method, using available bit-flip techniques (i.e, row-hammer), can transform a fully functional DNN model into a Trojan-infected model. We perform extensive experiments of CIFAR-10, SVHN and ImageNet datasets on both VGG-16 and Resnet-18 architectures. Our proposed TBT could classify 92 of test images to a target class with as little as 84 bit-flips out of 88 million weight bits on Resnet-18 for CIFAR10 dataset."
                },
                {
                    "title": "Invisible Backdoor Attacks on Deep Neural Networks Via Steganography and Regularization",
                    "abstract": "Deep neural networks (DNNs) have been proven vulnerable to backdoor attacks, where hidden features (patterns) trained to a normal model, which is only activated by some specific input (called triggers), trick the model into producing unexpected behavior. In this article, we create covert and scattered triggers for backdoor attacks, invisible backdoors, where triggers can fool both DNN models and human inspection. We apply our invisible backdoors through two state-of-the-art methods of embedding triggers for backdoor attacks. The first approach on Badnets embeds the trigger into DNNs through steganography. The second approach of a trojan attack uses two types of additional regularization terms to generate the triggers with irregular shape and size. We use the Attack Success Rate and Functionality to measure the performance of our attacks. We introduce two novel definitions of invisibility for human perception; one is conceptualized by the Perceptual Adversarial Similarity Score (PASS) and the other is Learned Perceptual Image Patch Similarity (LPIPS). We show that the proposed invisible backdoors can be fairly effective across various DNN models as well as four datasets MNIST, CIFAR-10, CIFAR-100, and GTSRB, by measuring their attack success rates for the adversary, functionality for the normal users, and invisibility scores for the administrators. We finally argue that the proposed invisible backdoor attacks can effectively thwart the state-of-the-art trojan backdoor detection approaches."
                },
                {
                    "title": "A critical analysis of self-supervision, or what we can learn from a single image",
                    "abstract": "We look critically at popular self-supervision techniques for learning deep convolutional neural networks without manual labels. We show that three different and representative methods, BiGAN, RotNet and DeepCluster, can learn the first few layers of a convolutional network from a single image as well as using millions of images and manual labels, provided that strong data augmentation is used. However, for deeper layers the gap with manual supervision cannot be closed even if millions of unlabelled images are used for training. We conclude that: (1) the weights of the early layers of deep networks contain limited information about the statistics of natural images, that (2) such low-level statistics can be learned through self-supervision just as well as through strong supervision, and that (3) the low-level statistics can be captured via synthetic transformations instead of using a large image dataset."
                },
                {
                    "title": "BadNets: Evaluating Backdooring Attacks on Deep Neural Networks",
                    "abstract": "Deep learning-based techniques have achieved state-of-the-art performance on a wide variety of recognition and classification tasks. However, these networks are typically computationally expensive to train, requiring weeks of computation on many GPUs; as a result, many users outsource the training procedure to the cloud or rely on pre-trained models that are then fine-tuned for a specific task. In this paper, we show that the outsourced training introduces new security risks: an adversary can create a maliciously trained network (a backdoored neural network, or a BadNet) that has the state-of-the-art performance on the user\u2019s training and validation samples but behaves badly on specific attacker-chosen inputs. We first explore the properties of BadNets in a toy example, by creating a backdoored handwritten digit classifier. Next, we demonstrate backdoors in a more realistic scenario by creating a U.S. street sign classifier that identifies stop signs as speed limits when a special sticker is added to the stop sign; we then show in addition that the backdoor in our U.S. street sign detector can persist even if the network is later retrained for another task and cause a drop in an accuracy of 25% on average when the backdoor trigger is present. These results demonstrate that backdoors in neural networks are both powerful and\u2014because the behavior of neural networks is difficult to explicate\u2014stealthy. This paper provides motivation for further research into techniques for verifying and inspecting neural networks, just as we have developed tools for verifying and debugging software."
                },
                {
                    "title": "DeepSigns: An End-to-End Watermarking Framework for Ownership Protection of Deep Neural Networks",
                    "abstract": "Deep Learning (DL) models have created a paradigm shift in our ability to comprehend raw data in various important fields, ranging from intelligence warfare and healthcare to autonomous transportation and automated manufacturing. A practical concern, in the rush to adopt DL models as a service, is protecting the models against Intellectual Property (IP) infringement. DL models are commonly built by allocating substantial computational resources that process vast amounts of proprietary training data. The resulting models are therefore considered to be an IP of the model builder and need to be protected to preserve the owner's competitive advantage. We propose DeepSigns, the first end-to-end IP protection framework that enables developers to systematically insert digital watermarks in the target DL model before distributing the model. DeepSigns is encapsulated as a high-level wrapper that can be leveraged within common deep learning frameworks including TensorFlow and PyTorch. The libraries in DeepSigns work by dynamically learning the Probability Density Function (pdf) of activation maps obtained in different layers of a DL model. DeepSigns uses the low probabilistic regions within the model to gradually embed the owner's signature (watermark) during DL training while minimally affecting the overall accuracy and training overhead. DeepSigns can demonstrably withstand various removal and transformation attacks, including model pruning, model fine-tuning, and watermark overwriting. We evaluate DeepSigns performance on a wide variety of DL architectures including wide residual convolution neural networks, multi-layer perceptrons, and long short-term memory models. Our extensive evaluations corroborate DeepSigns' effectiveness and applicability. We further provide a highly-optimized accompanying API to facilitate training watermarked neural networks with a training overhead as low as 2.2%."
                },
                {
                    "title": "Knockoff Nets: Stealing Functionality of Black-Box Models",
                    "abstract": "Machine Learning (ML) models are increasingly deployed in the wild to perform a wide range of tasks. In this work, we ask to what extent can an adversary steal functionality of such ``victim'' models based solely on blackbox interactions: image in, predictions out. In contrast to prior work, we study complex victim blackbox models, and an adversary lacking knowledge of train/test data used by the model, its internals, and semantics over model outputs. We formulate model functionality stealing as a two-step approach: (i) querying a set of input images to the blackbox model to obtain predictions; and (ii) training a ``knockoff'' with queried image-prediction pairs. We make multiple remarkable observations: (a) querying random images from a different distribution than that of the blackbox training data results in a well-performing knockoff; (b) this is possible even when the knockoff is represented using a different architecture; and (c) our reinforcement learning approach additionally improves query sample efficiency in certain settings and provides performance gains. We validate model functionality stealing on a range of datasets and tasks, as well as show that a reasonable knockoff of an image analysis API could be created for as little as $30."
                },
                {
                    "title": "Rethinking the Value of Network Pruning",
                    "abstract": "Network pruning is widely used for reducing the heavy inference cost of deep models in low-resource settings. A typical pruning algorithm is a three-stage pipeline, i.e., training (a large model), pruning and fine-tuning. During pruning, according to a certain criterion, redundant weights are pruned and important weights are kept to best preserve the accuracy. In this work, we make several surprising observations which contradict common beliefs. For all state-of-the-art structured pruning algorithms we examined, fine-tuning a pruned model only gives comparable or worse performance than training that model with randomly initialized weights. For pruning algorithms which assume a predefined target network architecture, one can get rid of the full pipeline and directly train the target network from scratch. Our observations are consistent for multiple network architectures, datasets, and tasks, which imply that: 1) training a large, over-parameterized model is often not necessary to obtain an efficient final model, 2) learned \"important\" weights of the large model are typically not useful for the small pruned model, 3) the pruned architecture itself, rather than a set of inherited \"important\" weights, is more crucial to the efficiency in the final model, which suggests that in some cases pruning can be useful as an architecture search paradigm. Our results suggest the need for more careful baseline evaluations in future research on structured pruning methods. We also compare with the \"Lottery Ticket Hypothesis\" (Frankle & Carbin 2019), and find that with optimal learning rate, the \"winning ticket\" initialization as used in Frankle & Carbin (2019) does not bring improvement over random initialization."
                },
                {
                    "title": "Protecting Intellectual Property of Deep Neural Networks with Watermarking",
                    "abstract": "Deep learning technologies, which are the key components of state-of-the-art Artificial Intelligence (AI) services, have shown great success in providing human-level capabilities for a variety of tasks, such as visual analysis, speech recognition, and natural language processing and etc. Building a production-level deep learning model is a non-trivial task, which requires a large amount of training data, powerful computing resources, and human expertises. Therefore, illegitimate reproducing, distribution, and the derivation of proprietary deep learning models can lead to copyright infringement and economic harm to model creators. Therefore, it is essential to devise a technique to protect the intellectual property of deep learning models and enable external verification of the model ownership. In this paper, we generalize the \"digital watermarking'' concept from multimedia ownership verification to deep neural network (DNNs) models. We investigate three DNN-applicable watermark generation algorithms, propose a watermark implanting approach to infuse watermark into deep learning models, and design a remote verification mechanism to determine the model ownership. By extending the intrinsic generalization and memorization capabilities of deep neural networks, we enable the models to learn specially crafted watermarks at training and activate with pre-specified predictions when observing the watermark patterns at inference. We evaluate our approach with two image recognition benchmark datasets. Our framework accurately (100%) and quickly verifies the ownership of all the remotely deployed deep learning models without affecting the model accuracy for normal input data. In addition, the embedded watermarks in DNN models are robust and resilient to different counter-watermark mechanisms, such as fine-tuning, parameter pruning, and model inversion attacks."
                },
                {
                    "title": "PRADA: Protecting Against DNN Model Stealing Attacks",
                    "abstract": "Machine learning (ML) applications are increasingly prevalent. Protecting the confidentiality of ML models becomes paramount for two reasons: (a) a model can be a business advantage to its owner, and (b) an adversary may use a stolen model to find transferable adversarial examples that can evade classification by the original model. Access to the model can be restricted to be only via well-defined prediction APIs. Nevertheless, prediction APIs still provide enough information to allow an adversary to mount model extraction attacks by sending repeated queries via the prediction API. In this paper, we describe new model extraction attacks using novel approaches for generating synthetic queries, and optimizing training hyperparameters. Our attacks outperform state-of-the-art model extraction in terms of transferability of both targeted and non-targeted adversarial examples (up to +29-44 percentage points, pp), and prediction accuracy (up to +46 pp) on two datasets. We provide take-aways on how to perform effective model extraction attacks. We then propose PRADA, the first step towards generic and effective detection of DNN model extraction attacks. It analyzes the distribution of consecutive API queries and raises an alarm when this distribution deviates from benign behavior. We show that PRADA can detect all prior model extraction attacks with no false positives."
                },
                {
                    "title": "Turning Your Weakness Into a Strength: Watermarking Deep Neural Networks by Backdooring",
                    "abstract": "Deep Neural Networks have recently gained lots of success after enabling several breakthroughs in notoriously challenging problems. Training these networks is computationally expensive and requires vast amounts of training data. Selling such pre-trained models can, therefore, be a lucrative business model. Unfortunately, once the models are sold they can be easily copied and redistributed. To avoid this, a tracking mechanism to identify models as the intellectual property of a particular vendor is necessary. In this work, we present an approach for watermarking Deep Neural Networks in a black-box way. Our scheme works for general classification tasks and can easily be combined with current learning algorithms. We show experimentally that such a watermark has no noticeable impact on the primary task that the model is designed for. Moreover, we evaluate the robustness of our proposal against a multitude of practical attacks."
                },
                {
                    "title": "A Downsampled Variant of ImageNet as an Alternative to the CIFAR datasets",
                    "abstract": "The original ImageNet dataset is a popular large-scale benchmark for training Deep Neural Networks. Since the cost of performing experiments (e.g, algorithm design, architecture search, and hyperparameter tuning) on the original dataset might be prohibitive, we propose to consider a downsampled version of ImageNet. In contrast to the CIFAR datasets and earlier downsampled versions of ImageNet, our proposed ImageNet32$\\times$32 (and its variants ImageNet64$\\times$64 and ImageNet16$\\times$16) contains exactly the same number of classes and images as ImageNet, with the only difference that the images are downsampled to 32$\\times$32 pixels per image (64$\\times$64 and 16$\\times$16 pixels for the variants, respectively). Experiments on these downsampled variants are dramatically faster than on the original ImageNet and the characteristics of the downsampled datasets with respect to optimal hyperparameters appear to remain similar. The proposed datasets and scripts to reproduce our results are available at this http URL and this https URL"
                },
                {
                    "title": "Embedding Watermarks into Deep Neural Networks",
                    "abstract": "Significant progress has been made with deep neural networks recently. Sharing trained models of deep neural networks has been a very important in the rapid progress of research and development of these systems. At the same time, it is necessary to protect the rights to shared trained models. To this end, we propose to use digital watermarking technology to protect intellectual property and detect intellectual property infringement in the use of trained models. First, we formulate a new problem: embedding watermarks into deep neural networks. Second, we propose a general framework for embedding a watermark in model parameters, using a parameter regularizer. Our approach does not impair the performance of networks into which a watermark is placed because the watermark is embedded while training the host network. Finally, we perform comprehensive experiments to reveal the potential of watermarking deep neural networks as the basis of this new research effort. We show that our framework can embed a watermark during the training of a deep neural network from scratch, and during fine-tuning and distilling, without impairing its performance. The embedded watermark does not disappear even after fine-tuning or parameter pruning; the watermark remains complete even after 65% of parameters are pruned."
                },
                {
                    "title": "Federated Learning: Strategies for Improving Communication Efficiency",
                    "abstract": "Federated Learning is a machine learning setting where the goal is to train a high-quality centralized model while training data remains distributed over a large number of clients each with unreliable and relatively slow network connections. We consider learning algorithms for this setting where on each round, each client independently computes an update to the current model based on its local data, and communicates this update to a central server, where the client-side updates are aggregated to compute a new global model. The typical clients in this setting are mobile phones, and communication efficiency is of the utmost importance. In this paper, we propose two ways to reduce the uplink communication costs: structured updates, where we directly learn an update from a restricted space parametrized using a smaller number of variables, e.g. either low-rank or a random mask; and sketched updates, where we learn a full model update and then compress it using a combination of quantization, random rotations, and subsampling before sending it to the server. Experiments on both convolutional and recurrent networks show that the proposed methods can reduce the communication cost by two orders of magnitude."
                },
                {
                    "title": "Stealing Machine Learning Models via Prediction APIs",
                    "abstract": "Machine learning (ML) models may be deemed confidential due to their sensitive training data, commercial value, or use in security applications. Increasingly often, confidential ML models are being deployed with publicly accessible query interfaces. ML-as-a-service (\"predictive analytics\") systems are an example: Some allow users to train models on potentially sensitive data and charge others for access on a pay-per-query basis. \nThe tension between model confidentiality and public access motivates our investigation of model extraction attacks. In such attacks, an adversary with black-box access, but no prior knowledge of an ML model's parameters or training data, aims to duplicate the functionality of (i.e., \"steal\") the model. Unlike in classical learning theory settings, ML-as-a-service offerings may accept partial feature vectors as inputs and include confidence values with predictions. Given these practices, we show simple, efficient attacks that extract target ML models with near-perfect fidelity for popular model classes including logistic regression, neural networks, and decision trees. We demonstrate these attacks against the online services of BigML and Amazon Machine Learning. We further show that the natural countermeasure of omitting confidence values from model outputs still admits potentially harmful model extraction attacks. Our results highlight the need for careful ML model deployment and new model extraction countermeasures."
                },
                {
                    "title": "Wide Residual Networks",
                    "abstract": "Deep residual networks were shown to be able to scale up to thousands of layers and still have improving performance. However, each fraction of a percent of improved accuracy costs nearly doubling the number of layers, and so training very deep residual networks has a problem of diminishing feature reuse, which makes these networks very slow to train. To tackle these problems, in this paper we conduct a detailed experimental study on the architecture of ResNet blocks, based on which we propose a novel architecture where we decrease depth and increase width of residual networks. We call the resulting network structures wide residual networks (WRNs) and show that these are far superior over their commonly used thin and very deep counterparts. For example, we demonstrate that even a simple 16-layer-deep wide residual network outperforms in accuracy and efficiency all previous deep residual networks, including thousand-layer-deep networks, achieving new state-of-the-art results on CIFAR, SVHN, COCO, and significant improvements on ImageNet. Our code and models are available at this https URL"
                },
                {
                    "title": "Communication-Efficient Learning of Deep Networks from Decentralized Data",
                    "abstract": "Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data center and training there using conventional approaches. We advocate an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. \nWe present a practical method for the federated learning of deep networks based on iterative model averaging, and conduct an extensive empirical evaluation, considering five different model architectures and four datasets. These experiments demonstrate the approach is robust to the unbalanced and non-IID data distributions that are a defining characteristic of this setting. Communication costs are the principal constraint, and we show a reduction in required communication rounds by 10-100x as compared to synchronized stochastic gradient descent."
                },
                {
                    "title": "Practical Black-Box Attacks against Machine Learning",
                    "abstract": "Machine learning (ML) models, e.g., deep neural networks (DNNs), are vulnerable to adversarial examples: malicious inputs modified to yield erroneous model outputs, while appearing unmodified to human observers. Potential attacks include having malicious content like malware identified as legitimate or controlling vehicle behavior. Yet, all existing adversarial example attacks require knowledge of either the model internals or its training data. We introduce the first practical demonstration of an attacker controlling a remotely hosted DNN with no such knowledge. Indeed, the only capability of our black-box adversary is to observe labels given by the DNN to chosen inputs. Our attack strategy consists in training a local model to substitute for the target DNN, using inputs synthetically generated by an adversary and labeled by the target DNN. We use the local substitute to craft adversarial examples, and find that they are misclassified by the targeted DNN. To perform a real-world and properly-blinded evaluation, we attack a DNN hosted by MetaMind, an online deep learning API. We find that their DNN misclassifies 84.24% of the adversarial examples crafted with our substitute. We demonstrate the general applicability of our strategy to many ML techniques by conducting the same attack against models hosted by Amazon and Google, using logistic regression substitutes. They yield adversarial examples misclassified by Amazon and Google at rates of 96.19% and 88.94%. We also find that this black-box attack strategy is capable of evading defense strategies previously found to make adversarial example crafting harder."
                },
                {
                    "title": "Margin-based Neural Network Watermarking",
                    "abstract": "As Machine Learning as a Service (MLaaS) platforms become prevalent, deep neural network (DNN) watermarking techniques are gaining increasing attention, which enables one to verify the ownership of a target DNN model in a black-box scenario. Unfortunately, previous watermarking methods are vulnerable to functionality stealing attacks, thus allowing an adversary to falsely claim the ownership of a DNN model stolen from its original owner. In this work, we propose a novel margin-based DNN watermarking approach that is robust to the functionality stealing attacks based on model extraction and distillation. Specifically, during training, our method maximizes the margins of watermarked samples by using projected gradient ascent on them so that their predicted labels cannot change without compromising the accuracy of the model that the attacker tries to steal. We validate our method on multiple benchmarks and show that our watermarking method successfully defends against model extraction attacks, outperforming relevant baselines."
                },
                {
                    "title": "Leveraging Multi-task Learning for Umambiguous and Flexible Deep Neural Network Watermarking",
                    "abstract": "Deep neural networks are playing an important role in many real-life applications. An important prerequisite in commer-cializing deep neural networks is the identi\ufb01cation of their genuine owners. Therefore, watermarking schemes that embed the owner\u2019s identity information into the models have been proposed. However, current schemes cannot meet all the security requirements such as unambiguity and are in\ufb02ex-ible since most of them focus on classi\ufb01cation models. To meet the formal de\ufb01nitions of the security requirements and increase the applicability of deep neural network watermarking schemes, we propose a new method, MTLSign , based on multi-task learning. By treating the watermark embedding as an extra task, the security requirements are explicitly formulated and met with well-designed regularizers and components from cryptography. Experiments have demonstrated that MTLSign is \ufb02exible and robust for practical security in machine learning applications."
                },
                {
                    "title": "Extrapolating from a Single Image to a Thousand Classes using Distillation",
                    "abstract": "What can neural networks learn about the visual world from a single image? While it obviously cannot contain the multitudes of possible objects, scenes and lighting conditions that exist \u2013 within the space of all possible 256 3 \u00b7 224 \u00b7 224 224 -sized square images, it might still provide a strong prior for natural images. To analyze this hypothesis, we develop a framework for training neural networks from scratch us-ing a single image by means of knowledge distillation from a supervised pretrained teacher. With this, we \ufb01nd that the answer to the above question is: \u2018surprisingly, a lot\u2019. In quantitative terms, we \ufb01nd top-1 accuracies of 94% / 74% on CIFAR-10/100, 59 % on ImageNet and, by extending this method to audio, 84 % on SpeechCommands. In extensive analyses we disentangle the effect of augmentations, choice of source image and network architectures and also discover \u201cpanda neurons\u201d in networks that have never seen a panda. This work shows that one image can be used to extrapolate to thousands of object classes and motivates a renewed research agenda on the fundamental interplay of augmentations and images. Webpage: 1"
                },
                {
                    "title": "Trojaning Attack on Neural Networks",
                    "abstract": "\u2014With the fast spread of machine learning tech- niques, sharing and adopting public machine learning models become very popular. This gives attackers many new opportunities. In this paper, we propose a trojaning attack on neural networks. As the models are not intuitive for human to understand, the attack features stealthiness. Deploying trojaned models can cause various severe consequences including endangering human lives (in applications like autonomous driving). We \ufb01rst inverse the neural network to generate a general trojan trigger , and then retrain the model with reversed engineered training data to inject malicious behaviors to the model. The malicious behaviors are only activated by inputs stamped with the trojan trigger. In our attack, we do not need to tamper with the original training process, which usually takes weeks to months. Instead, it takes minutes to hours to apply our attack. Also, we do not require the datasets that are used to train the model. In practice, the datasets are usually not shared due to privacy or copyright concerns. We use \ufb01ve different applications to demonstrate the power of our attack, and perform a deep analysis on the possible factors that affect the attack. The results show that our attack is highly effective and ef\ufb01cient. The trojaned behaviors can be successfully triggered (with nearly 100% possibility) without affecting its test accuracy for normal input and even with better accuracy on public dataset. Also, it only takes a small amount of time to attack a complex neuron network model. In the end, we also discuss possible defense against such attacks."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively implement a backdoor-based watermarking strategy for deep neural networks that ensures intellectual property protection without requiring access to the original training data?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing concern of intellectual property theft in the era of large-scale deep learning models. By developing a robust watermarking technique that does not rely on original training data, we can enhance the security and ownership verification of models, thereby fostering trust in AI technologies. This advancement could lead to practical applications in various sectors, including model sharing, federated learning, and commercial AI services, ultimately influencing future research directions in model security and copyright protection.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to create a watermarking method that is both effective and efficient without access to the original training data. Naive approaches may fail because they often rely on the availability of training samples to ensure model utility and performance. Additionally, technical obstacles include ensuring the robustness of the watermark against model fine-tuning and other modifications, as well as maintaining the model's predictive accuracy on clean samples. The complexity of designing a watermark that can withstand minor changes while still being detectable poses a significant hurdle.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on watermarking methods that require access to the original training data, which is often impractical in real-world scenarios. Existing solutions have limitations in terms of safety, efficiency, and robustness, particularly in the context of backdoor-based techniques. Barriers such as the need for extensive computational resources and the vulnerability of watermarks to model alterations have hindered progress. Our approach differs by proposing a method that circumvents these limitations, focusing on efficient fine-tuning and safe watermark injection without the need for original data.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a backdoor-based watermarking strategy that utilizes efficient fine-tuning techniques. We will employ a dataset of verification samples with trigger patches and designated labels to create the watermark. The performance of our method will be evaluated using metrics such as watermark detection accuracy and model utility on clean samples. We expect our approach to yield a robust watermark that can withstand minor model changes while ensuring the model's predictive performance remains intact, thereby"
            }
        },
        "author_data": {
            "25326f7c-5f63-4e18-b605-457f387e9522": {
                "pk": "25326f7c-5f63-4e18-b605-457f387e9522",
                "name": "Shuyang Yu",
                "collaborators": [
                    "Jiayu Zhou",
                    "Yijiang Pang",
                    "Hoang Bao",
                    "Junyuan Hong",
                    "Yi Zeng",
                    "Fei Wang",
                    "Ruoxi Jia",
                    "A. P. Reprint",
                    "Bao Hoang"
                ],
                "domain": [
                    "Optimization",
                    "Federated Learning",
                    "Machine Learning",
                    "Hyperparameter Tuning"
                ],
                "publications": [
                    {
                        "title": "Towards Stability of Parameter-free Optimization",
                        "abstract": "Hyperparameter tuning, particularly the selection of an appropriate learning rate in adaptive gradient training methods, remains a challenge. To tackle this challenge, in this paper, we propose a novel parameter-free optimizer, \\textsc{AdamG} (Adam with the golden step size), designed to automatically adapt to diverse optimization problems without manual tuning. The core technique underlying \\textsc{AdamG} is our golden step size derived for the AdaGrad-Norm algorithm, which is expected to help AdaGrad-Norm preserve the tuning-free convergence and approximate the optimal step size in expectation w.r.t. various optimization scenarios. To better evaluate tuning-free performance, we propose a novel evaluation criterion, \\textit{reliability}, to comprehensively assess the efficacy of parameter-free optimizers in addition to classical performance criteria. Empirical results demonstrate that compared with other parameter-free baselines, \\textsc{AdamG} achieves superior performance, which is consistently on par with Adam using a manually tuned learning rate across various optimization tasks."
                    },
                    {
                        "title": "Who Leaked the Model? Tracking IP Infringers in Accountable Federated Learning",
                        "abstract": "Federated learning (FL) emerges as an effective collaborative learning framework to coordinate data and computation resources from massive and distributed clients in training. Such collaboration results in non-trivial intellectual property (IP) represented by the model parameters that should be protected and shared by the whole party rather than an individual user. Meanwhile, the distributed nature of FL endorses a malicious client the convenience to compromise IP through illegal model leakage to unauthorized third parties. To block such IP leakage, it is essential to make the IP identifiable in the shared model and locate the anonymous infringer who first leaks it. The collective challenges call for \\emph{accountable federated learning}, which requires verifiable ownership of the model and is capable of revealing the infringer's identity upon leakage. In this paper, we propose Decodable Unique Watermarking (DUW) for complying with the requirements of accountable FL. Specifically, before a global model is sent to a client in an FL round, DUW encodes a client-unique key into the model by leveraging a backdoor-based watermark injection. To identify the infringer of a leaked model, DUW examines the model and checks if the triggers can be decoded as the corresponding keys. Extensive empirical results show that DUW is highly effective and robust, achieving over $99\\%$ watermark success rate for Digits, CIFAR-10, and CIFAR-100 datasets under heterogeneous FL settings, and identifying the IP infringer with $100\\%$ accuracy even after common watermark removal attempts."
                    },
                    {
                        "title": "T OWARDS R ELIABILITY OF P ARAMETER - FREE O PTIMIZATION",
                        "abstract": "Hyperparameter tuning, particularly the selection of an appropriate learning rate in adaptive gradient training methods, remains a challenge. To tackle this challenge, in this paper, we propose a novel parameter-free optimizer, A DAM G (Adam with the golden step size), designed to automatically adapt to diverse optimization problems without manual tuning. The core technique underlying A DAM G is our golden step size derived for the AdaGrad-Norm algorithm, which is expected to help AdaGrad-Norm preserve the tuning-free convergence and approximate the optimal step size in expectation w.r"
                    }
                ]
            },
            "9e8412d0-83e5-4599-99a7-aa610fdc043e": {
                "pk": "9e8412d0-83e5-4599-99a7-aa610fdc043e",
                "name": "Junyuan Hong",
                "collaborators": [
                    "Jiayu Zhou",
                    "Zhangyang Wang",
                    "Haotao Wang",
                    "Yuyang Deng",
                    "M. Mahdavi",
                    "Shuyang Yu",
                    "Zhuangdi Zhu",
                    "Lingjuan Lyu",
                    "Yi Zeng",
                    "Haobo Zhang"
                ],
                "domain": [
                    "Federated Learning",
                    "Privacy-Preserving Machine Learning",
                    "Deep Learning",
                    "Knowledge Distillation"
                ],
                "publications": [
                    {
                        "title": "On the Generalization Ability of Unsupervised Pretraining",
                        "abstract": "Recent advances in unsupervised learning have shown that unsupervised pre-training, followed by fine-tuning, can improve model generalization. However, a rigorous understanding of how the representation function learned on an unlabeled dataset affects the generalization of the fine-tuned model is lacking. Existing theoretical research does not adequately account for the heterogeneity of the distribution and tasks in pre-training and fine-tuning stage. To bridge this gap, this paper introduces a novel theoretical framework that illuminates the critical factor influencing the transferability of knowledge acquired during unsupervised pre-training to the subsequent fine-tuning phase, ultimately affecting the generalization capabilities of the fine-tuned model on downstream tasks. We apply our theoretical framework to analyze generalization bound of two distinct scenarios: Context Encoder pre-training with deep neural networks and Masked Autoencoder pre-training with deep transformers, followed by fine-tuning on a binary classification task. Finally, inspired by our findings, we propose a novel regularization method during pre-training to further enhances the generalization of fine-tuned model. Overall, our results contribute to a better understanding of unsupervised pre-training and fine-tuning paradigm, and can shed light on the design of more effective pre-training algorithms."
                    },
                    {
                        "title": "Turning the Curse of Heterogeneity in Federated Learning into a Blessing for Out-of-Distribution Detection",
                        "abstract": "Deep neural networks have witnessed huge successes in many challenging prediction tasks and yet they often suffer from out-of-distribution (OoD) samples, misclassifying them with high con\ufb01dence. Recent advances show promising OoD detection performance for centralized training, and however, OoD detection in federated learning (FL) is largely overlooked, even though many security sensitive applications such as autonomous driving and voice recognition authorization are commonly trained using FL for data privacy concerns. The main challenge that prevents previous state-of-the-art OoD detection methods from being incorporated to FL is that they require large amount of real OoD samples. However, in real-world scenarios, such large-scale OoD training data can be costly or even infeasible to obtain, especially for resource-limited local devices. On the other hand, a notorious challenge in FL is data heterogeneity where each client collects non-identically and independently distributed (non-iid) data. We propose to take advantage of such heterogeneity and turn the curse into a blessing that facilitates OoD detection in FL. The key is that for each client, non-iid data from other clients (unseen external classes) can serve as an alternative to real OoD samples. Speci\ufb01cally, we propose a novel Federated Out-of-Distribution Synthesizer (F OSTER ), which learns a class-conditional generator to synthesize virtual external-class OoD samples, and maintains data con\ufb01dentiality and communication ef\ufb01ciency required by FL. Experimental results show that our method outperforms the state-of-the-art for"
                    },
                    {
                        "title": "FedNoisy: Federated Noisy Label Learning Benchmark",
                        "abstract": "Federated learning has gained popularity for distributed learning without aggregating sensitive data from clients. But meanwhile, the distributed and isolated nature of data isolation may be complicated by data quality, making it more vulnerable to noisy labels. Many efforts exist to defend against the negative impacts of noisy labels in centralized or federated settings. However, there is a lack of a benchmark that comprehensively considers the impact of noisy labels in a wide variety of typical FL settings. In this work, we serve the first standardized benchmark that can help researchers fully explore potential federated noisy settings. Also, we conduct comprehensive experiments to explore the characteristics of these data settings and unravel challenging scenarios on the federated noisy label learning, which may guide method development in the future. We highlight the 20 basic settings for more than 5 datasets proposed in our benchmark and standardized simulation pipeline for federated noisy label learning. We hope this benchmark can facilitate idea verification in federated learning with noisy labels. \\texttt{FedNoisy} is available at \\codeword{https://github.com/SMILELab-FL/FedNoisy}."
                    },
                    {
                        "title": "Who Leaked the Model? Tracking IP Infringers in Accountable Federated Learning",
                        "abstract": "Federated learning (FL) emerges as an effective collaborative learning framework to coordinate data and computation resources from massive and distributed clients in training. Such collaboration results in non-trivial intellectual property (IP) represented by the model parameters that should be protected and shared by the whole party rather than an individual user. Meanwhile, the distributed nature of FL endorses a malicious client the convenience to compromise IP through illegal model leakage to unauthorized third parties. To block such IP leakage, it is essential to make the IP identifiable in the shared model and locate the anonymous infringer who first leaks it. The collective challenges call for \\emph{accountable federated learning}, which requires verifiable ownership of the model and is capable of revealing the infringer's identity upon leakage. In this paper, we propose Decodable Unique Watermarking (DUW) for complying with the requirements of accountable FL. Specifically, before a global model is sent to a client in an FL round, DUW encodes a client-unique key into the model by leveraging a backdoor-based watermark injection. To identify the infringer of a leaked model, DUW examines the model and checks if the triggers can be decoded as the corresponding keys. Extensive empirical results show that DUW is highly effective and robust, achieving over $99\\%$ watermark success rate for Digits, CIFAR-10, and CIFAR-100 datasets under heterogeneous FL settings, and identifying the IP infringer with $100\\%$ accuracy even after common watermark removal attempts."
                    },
                    {
                        "title": "International Workshop on Federated Learning for Distributed Data Mining",
                        "abstract": "The past decade has witnessed wide applications of machine learning to various domains for decision-making, including crime detection, urban planning, drug discovery, and health monitoring, which benefited from surging data resources. As data collection in real-world applications is often done in different locations, being able to mine and discover knowledge from distributed data sources is an essential requirement for building powerful predictive models. However, directly uploading all data sources to an untrustworthy centralized data server for learning will lead to risks of privacy leakage. Federated Learning (FL) emerges as a decentralized learning framework that aggregates knowledge from distributed data without centralizing them, hence mitigating privacy risks. By hosting this workshop, we aim to attract a broad spectrum of audiences, including researchers and practitioners from academia and industry interested in the latest advances in FL. As an effort to advance the fundamental development of FL in data mining, this workshop will encourage ideas exchange on the trustworthiness, scalability, robustness, and broad applications of FL."
                    },
                    {
                        "title": "Understanding Deep Gradient Leakage via Inversion Influence Functions",
                        "abstract": "Deep Gradient Leakage (DGL) is a highly effective attack that recovers private training images from gradient vectors. This attack casts significant privacy challenges on distributed learning from clients with sensitive data, where clients are required to share gradients. Defending against such attacks requires but lacks an understanding of when and how privacy leakage happens, mostly because of the black-box nature of deep networks. In this paper, we propose a novel Inversion Influence Function (I2F) that establishes a closed-form connection between the recovered images and the private gradients by implicitly solving the DGL problem. Compared to directly solving DGL, I2F is scalable for analyzing deep networks, requiring only oracle access to gradients and Jacobian-vector products. We empirically demonstrate that I2F effectively approximated the DGL generally on different model architectures, datasets, modalities, attack implementations, and perturbation-based defenses. With this novel tool, we provide insights into effective gradient perturbation directions, the unfairness of privacy protection, and privacy-preferred model initialization. Our codes are provided in https://github.com/illidanlab/inversion-influence-function."
                    },
                    {
                        "title": "DP-OPT: Make Large Language Model Your Privacy-Preserving Prompt Engineer",
                        "abstract": "Large Language Models (LLMs) have emerged as dominant tools for various tasks, particularly when tailored for a specific target by prompt tuning. Nevertheless, concerns surrounding data privacy present obstacles due to the tuned prompts' dependency on sensitive private information. A practical solution is to host a local LLM and optimize a soft prompt privately using data. Yet, hosting a local model becomes problematic when model ownership is protected. Alternative methods, like sending data to the model's provider for training, intensify these privacy issues facing an untrusted provider. In this paper, we present a novel solution called Differentially-Private Offsite Prompt Tuning (DP-OPT) to address this challenge. Our approach involves tuning a discrete prompt on the client side and then applying it to the desired cloud models. We demonstrate that prompts suggested by LLMs themselves can be transferred without compromising performance significantly. To ensure that the prompts do not leak private information, we introduce the first private prompt generation mechanism, by a differentially-private (DP) ensemble of in-context learning with private demonstrations. With DP-OPT, generating privacy-preserving prompts by Vicuna-7b can yield competitive performance compared to non-private in-context learning on GPT3.5 or local private prompt tuning. Codes are available at https://github.com/VITA-Group/DP-OPT ."
                    },
                    {
                        "title": "Revisiting Data-Free Knowledge Distillation with Poisoned Teachers",
                        "abstract": "Data-free knowledge distillation (KD) helps transfer knowledge from a pre-trained model (known as the teacher model) to a smaller model (known as the student model) without access to the original training data used for training the teacher model. However, the security of the synthetic or out-of-distribution (OOD) data required in data-free KD is largely unknown and under-explored. In this work, we make the first effort to uncover the security risk of data-free KD w.r.t. untrusted pre-trained models. We then propose Anti-Backdoor Data-Free KD (ABD), the first plug-in defensive method for data-free KD methods to mitigate the chance of potential backdoors being transferred. We empirically evaluate the effectiveness of our proposed ABD in diminishing transferred backdoor knowledge while maintaining compatible downstream performances as the vanilla KD. We envision this work as a milestone for alarming and mitigating the potential backdoors in data-free KD. Codes are released at https://github.com/illidanlab/ABD."
                    },
                    {
                        "title": "MECTA: Memory-Economic Continual Test-Time Model Adaptation",
                        "abstract": "Continual Test-time Adaptation (CTA) is a promising art to secure accuracy gains in continually-changing environments. The state-of-the-art adaptations improve out-of-distribution model accuracy via computation-efficient online test-time gradient descents but meanwhile cost about times of memory versus the inference, even if only a small portion of parameters are updated. Such high memory consumption of CTA substantially impedes wide applications of advanced CTA on memory-constrained devices. In this paper, we provide a novel solution, dubbed MECTA, to drastically improve the memory efficiency of gradient-based CTA. Our profiling shows that the major memory overhead comes from the intermediate cache for back-propagation, which scales by the batch size, channel, and layer number. Therefore, we propose to reduce batch sizes, adopt an adaptive normalization layer to maintain stable and accurate predictions, and stop the back-propagation caching heuristically. On the other hand, we prune the networks to reduce the computation and memory overheads in optimization and recover the parameters afterward to avoid forgetting. The proposed MECTA is efficient and can be seamlessly plugged into state-of-the-art CTA algorithms at negligible overhead on computation and memory. On three datasets, CIFAR10, CIFAR100, and ImageNet, MECTA improves the accuracy by at least 6% with constrained memory and significantly reduces the memory costs of ResNet50 on ImageNet by at least 70% with comparable accuracy. Our codes can be accessed at https://github.com/SonyAI/MECTA ."
                    },
                    {
                        "title": "On the Hardness of Robustness Transfer: A Perspective from Rademacher Complexity over Symmetric Difference Hypothesis Space",
                        "abstract": "Recent studies demonstrated that the adversarially robust learning under $\\ell_\\infty$ attack is harder to generalize to different domains than standard domain adaptation. How to transfer robustness across different domains has been a key question in domain adaptation field. To investigate the fundamental difficulty behind adversarially robust domain adaptation (or robustness transfer), we propose to analyze a key complexity measure that controls the cross-domain generalization: the adversarial Rademacher complexity over {\\em symmetric difference hypothesis space} $\\mathcal{H} \\Delta \\mathcal{H}$. For linear models, we show that adversarial version of this complexity is always greater than the non-adversarial one, which reveals the intrinsic hardness of adversarially robust domain adaptation. We also establish upper bounds on this complexity measure. Then we extend them to the ReLU neural network class by upper bounding the adversarial Rademacher complexity in the binary classification setting. Finally, even though the robust domain adaptation is provably harder, we do find positive relation between robust learning and standard domain adaptation. We explain \\emph{how adversarial training helps domain adaptation in terms of standard risk}. We believe our results initiate the study of the generalization theory of adversarially robust domain adaptation, and could shed lights on distributed adversarially robust learning from heterogeneous sources, e.g., federated learning scenario."
                    },
                    {
                        "title": "A Privacy-Preserving Hybrid Federated Learning Framework for Financial Crime Detection",
                        "abstract": "The recent decade witnessed a surge of increase in financial crimes across the public and private sectors, with an average cost of scams of $102m to financial institutions in 2022. Developing a mechanism for battling financial crimes is an impending task that requires in-depth collaboration from multiple institutions, and yet such collaboration imposed significant technical challenges due to the privacy and security requirements of distributed financial data. For example, consider the modern payment network systems, which can generate millions of transactions per day across a large number of global institutions. Training a detection model of fraudulent transactions requires not only secured transactions but also the private account activities of those involved in each transaction from corresponding bank systems. The distributed nature of both samples and features prevents most existing learning systems from being directly adopted to handle the data mining task. In this paper, we collectively address these challenges by proposing a hybrid federated learning system that offers secure and privacy-aware learning and inference for financial crime detection. We conduct extensive empirical studies to evaluate the proposed framework's detection performance and privacy-protection capability, evaluating its robustness against common malicious attacks of collaborative learning. We release our source code at https://github.com/illidanlab/HyFL ."
                    },
                    {
                        "title": "Trap and Replace: Defending Backdoor Attacks by Trapping Them into an Easy-to-Replace Subnetwork",
                        "abstract": "Deep neural networks (DNNs) are vulnerable to backdoor attacks. Previous works have shown it extremely challenging to unlearn the undesired backdoor behavior from the network, since the entire network can be affected by the backdoor samples. In this paper, we propose a brand-new backdoor defense strategy, which makes it much easier to remove the harmful influence of backdoor samples from the model. Our defense strategy, Trap and Replace, consists of two stages. In the first stage, we bait and trap the backdoors in a small and easy-to-replace subnetwork. Specifically, we add an auxiliary image reconstruction head on top of the stem network shared with a light-weighted classification head. The intuition is that the auxiliary image reconstruction task encourages the stem network to keep sufficient low-level visual features that are hard to learn but semantically correct, instead of overfitting to the easy-to-learn but semantically incorrect backdoor correlations. As a result, when trained on backdoored datasets, the backdoors are easily baited towards the unprotected classification head, since it is much more vulnerable than the shared stem, leaving the stem network hardly poisoned. In the second stage, we replace the poisoned light-weighted classification head with an untainted one, by re-training it from scratch only on a small holdout dataset with clean samples, while fixing the stem network. As a result, both the stem and the classification head in the final network are hardly affected by backdoor training samples. We evaluate our method against ten different backdoor attacks. Our method outperforms previous state-of-the-art methods by up to 20.57%, 9.80%, and 13.72% attack success rate and on-average 3.14%, 1.80%, and 1.21% clean classification accuracy on CIFAR10, GTSRB, and ImageNet-12, respectively. Code is available at https://github.com/VITA-Group/Trap-and-Replace-Backdoor-Defense."
                    },
                    {
                        "title": "Resilient and Communication Efficient Learning for Heterogeneous Federated Systems",
                        "abstract": "The rise of Federated Learning (FL) is bringing machine learning to edge computing by utilizing data scattered across edge devices. However, the heterogeneity of edge network topologies and the uncertainty of wireless transmission are two major obstructions of FL's wide application in edge computing, leading to prohibitive convergence time and high communication cost. In this work, we propose an FL scheme to address both challenges simultaneously. Specifically, we enable edge devices to learn self-distilled neural networks that are readily prunable to arbitrary sizes, which capture the knowledge of the learning domain in a nested and progressive manner. Not only does our approach tackle system heterogeneity by serving edge devices with varying model architectures, but it also alleviates the issue of connection uncertainty by allowing transmitting part of the model parameters under faulty network connections, without wasting the contributing knowledge of the transmitted parameters. Extensive empirical studies show that under system heterogeneity and network instability, our approach demonstrates significant resilience and higher communication efficiency compared to the state-of-the-art."
                    },
                    {
                        "title": "Efficient Split-Mix Federated Learning for On-Demand and In-Situ Customization",
                        "abstract": "Federated learning (FL) provides a distributed learning framework for multiple participants to collaborate learning without sharing raw data. In many practical FL scenarios, participants have heterogeneous resources due to disparities in hardware and inference dynamics that require quickly loading models of different sizes and levels of robustness. The heterogeneity and dynamics together impose significant challenges to existing FL approaches and thus greatly limit FL's applicability. In this paper, we propose a novel Split-Mix FL strategy for heterogeneous participants that, once training is done, provides in-situ customization of model sizes and robustness. Specifically, we achieve customization by learning a set of base sub-networks of different sizes and robustness levels, which are later aggregated on-demand according to inference requirements. This split-mix strategy achieves customization with high efficiency in communication, storage, and inference. Extensive experiments demonstrate that our method provides better in-situ customization than the existing heterogeneous-architecture FL methods. Codes and pre-trained models are available: https://github.com/illidanlab/SplitMix."
                    },
                    {
                        "title": "How Robust is Your Fairness? Evaluating and Sustaining Fairness under Unseen Distribution Shifts",
                        "abstract": "Increasing concerns have been raised on deep learning fairness in recent years. Existing fairness-aware machine learning methods mainly focus on the fairness of in-distribution data. However, in real-world applications, it is common to have distribution shift between the training and test data. In this paper, we first show that the fairness achieved by existing methods can be easily broken by slight distribution shifts. To solve this problem, we propose a novel fairness learning method termed CUrvature MAtching (CUMA), which can achieve robust fairness generalizable to unseen domains with unknown distributional shifts. Specifically, CUMA enforces the model to have similar generalization ability on the majority and minority groups, by matching the loss curvature distributions of the two groups. We evaluate our method on three popular fairness datasets. Compared with existing methods, CUMA achieves superior fairness under unseen distribution shifts, without sacrificing either the overall accuracy or the in-distribution fairness."
                    },
                    {
                        "title": "Outsourcing Training without Uploading Data via Efficient Collaborative Open-Source Sampling",
                        "abstract": "As deep learning blooms with growing demand for computation and data resources, outsourcing model training to a powerful cloud server becomes an attractive alternative to training at a low-power and cost-effective end device. Traditional outsourcing requires uploading device data to the cloud server, which can be infeasible in many real-world applications due to the often sensitive nature of the collected data and the limited communication bandwidth. To tackle these challenges, we propose to leverage widely available open-source data, which is a massive dataset collected from public and heterogeneous sources (e.g., Internet images). We develop a novel strategy called Efficient Collaborative Open-source Sampling (ECOS) to construct a proximal proxy dataset from open-source data for cloud training, in lieu of client data. ECOS probes open-source data on the cloud server to sense the distribution of client data via a communication- and computation-efficient sampling process, which only communicates a few compressed public features and client scalar responses. Extensive empirical studies show that the proposed ECOS improves the quality of automated client labeling, model compression, and label outsourcing when applied in various learning scenarios."
                    },
                    {
                        "title": "Dynamic Privacy Budget Allocation Improves Data Efficiency of Differentially Private Gradient Descent",
                        "abstract": "Protecting privacy in learning while maintaining the model performance has become increasingly critical in many applications that involve sensitive data. A popular private learning framework is differentially private learning composed of many privatized gradient iterations by noising and clipping. Under the privacy constraint, it has been shown that the dynamic policies could improve the final iterate loss, namely the quality of published models. In this talk, we will introduce these dynamic techniques for learning rate, batch size, noise magnitude and gradient clipping. Also, we discuss how the dynamic policy could change the convergence bounds which further provides insight of the impact of dynamic methods."
                    },
                    {
                        "title": "Data-Free Knowledge Distillation for Heterogeneous Federated Learning",
                        "abstract": "Federated Learning (FL) is a decentralized machine-learning paradigm in which a global server iteratively aggregates the model parameters of local users without accessing their data. User heterogeneity has imposed significant challenges to FL, which can incur drifted global models that are slow to converge. Knowledge Distillation has recently emerged to tackle this issue, by refining the server model using aggregated knowledge from heterogeneous users, other than directly aggregating their model parameters. This approach, however, depends on a proxy dataset, making it impractical unless such prerequisite is satisfied. Moreover, the ensemble knowledge is not fully utilized to guide local model learning, which may in turn affect the quality of the aggregated model. In this work, we propose a data-free knowledge distillation approach to address heterogeneous FL, where the server learns a lightweight generator to ensemble user information in a data-free manner, which is then broadcasted to users, regulating local training using the learned knowledge as an inductive bias. Empirical studies powered by theoretical implications show that, our approach facilitates FL with better generalization performance using fewer communication rounds, compared with the state-of-the-art."
                    }
                ]
            },
            "b071fe73-cc5c-4493-b31b-1051cef989c4": {
                "pk": "b071fe73-cc5c-4493-b31b-1051cef989c4",
                "name": "Haobo Zhang",
                "collaborators": [
                    "Junyuan Hong",
                    "Yuyang Deng",
                    "M. Mahdavi",
                    "Jiayu Zhou"
                ],
                "domain": [
                    "Privacy Preservation",
                    "Deep Learning",
                    "Adversarial Attacks",
                    "Gradient Leakage"
                ],
                "publications": [
                    {
                        "title": "Understanding Deep Gradient Leakage via Inversion Influence Functions",
                        "abstract": "Deep Gradient Leakage (DGL) is a highly effective attack that recovers private training images from gradient vectors. This attack casts significant privacy challenges on distributed learning from clients with sensitive data, where clients are required to share gradients. Defending against such attacks requires but lacks an understanding of when and how privacy leakage happens, mostly because of the black-box nature of deep networks. In this paper, we propose a novel Inversion Influence Function (I2F) that establishes a closed-form connection between the recovered images and the private gradients by implicitly solving the DGL problem. Compared to directly solving DGL, I2F is scalable for analyzing deep networks, requiring only oracle access to gradients and Jacobian-vector products. We empirically demonstrate that I2F effectively approximated the DGL generally on different model architectures, datasets, modalities, attack implementations, and perturbation-based defenses. With this novel tool, we provide insights into effective gradient perturbation directions, the unfairness of privacy protection, and privacy-preferred model initialization. Our codes are provided in https://github.com/illidanlab/inversion-influence-function."
                    }
                ]
            },
            "bd491a31-cb5a-4719-9404-5ea6efada5ab": {
                "pk": "bd491a31-cb5a-4719-9404-5ea6efada5ab",
                "name": "Haotao Wang",
                "collaborators": [
                    "Zhangyang Wang",
                    "Junyuan Hong",
                    "Jiayu Zhou",
                    "Aston Zhang",
                    "Chaowei Xiao",
                    "Zhiding Yu",
                    "Shuai Zheng",
                    "Mu Li",
                    "Tianlong Chen",
                    "Kede Ma"
                ],
                "domain": [
                    "Federated Learning",
                    "Adversarial Training",
                    "Deep Learning",
                    "Out-of-Distribution Detection"
                ],
                "publications": [
                    {
                        "title": "Turning the Curse of Heterogeneity in Federated Learning into a Blessing for Out-of-Distribution Detection",
                        "abstract": "Deep neural networks have witnessed huge successes in many challenging prediction tasks and yet they often suffer from out-of-distribution (OoD) samples, misclassifying them with high con\ufb01dence. Recent advances show promising OoD detection performance for centralized training, and however, OoD detection in federated learning (FL) is largely overlooked, even though many security sensitive applications such as autonomous driving and voice recognition authorization are commonly trained using FL for data privacy concerns. The main challenge that prevents previous state-of-the-art OoD detection methods from being incorporated to FL is that they require large amount of real OoD samples. However, in real-world scenarios, such large-scale OoD training data can be costly or even infeasible to obtain, especially for resource-limited local devices. On the other hand, a notorious challenge in FL is data heterogeneity where each client collects non-identically and independently distributed (non-iid) data. We propose to take advantage of such heterogeneity and turn the curse into a blessing that facilitates OoD detection in FL. The key is that for each client, non-iid data from other clients (unseen external classes) can serve as an alternative to real OoD samples. Speci\ufb01cally, we propose a novel Federated Out-of-Distribution Synthesizer (F OSTER ), which learns a class-conditional generator to synthesize virtual external-class OoD samples, and maintains data con\ufb01dentiality and communication ef\ufb01ciency required by FL. Experimental results show that our method outperforms the state-of-the-art for"
                    },
                    {
                        "title": "Graph Mixture of Experts: Learning on Large-Scale Graphs with Explicit Diversity Modeling",
                        "abstract": "Graph neural networks (GNNs) have found extensive applications in learning from graph data. However, real-world graphs often possess diverse structures and comprise nodes and edges of varying types. To bolster the generalization capacity of GNNs, it has become customary to augment training graph structures through techniques like graph augmentations and large-scale pre-training on a wider array of graphs. Balancing this diversity while avoiding increased computational costs and the notorious trainability issues of GNNs is crucial. This study introduces the concept of Mixture-of-Experts (MoE) to GNNs, with the aim of augmenting their capacity to adapt to a diverse range of training graph structures, without incurring explosive computational overhead. The proposed Graph Mixture of Experts (GMoE) model empowers individual nodes in the graph to dynamically and adaptively select more general information aggregation experts. These experts are trained to capture distinct subgroups of graph structures and to incorporate information with varying hop sizes, where those with larger hop sizes specialize in gathering information over longer distances. The effectiveness of GMoE is validated through a series of experiments on a diverse set of tasks, including graph, node, and link prediction, using the OGB benchmark. Notably, it enhances ROC-AUC by $1.81\\%$ in ogbg-molhiv and by $1.40\\%$ in ogbg-molbbbp, when compared to the non-MoE baselines. Our code is publicly available at https://github.com/VITA-Group/Graph-Mixture-of-Experts."
                    },
                    {
                        "title": "Trap and Replace: Defending Backdoor Attacks by Trapping Them into an Easy-to-Replace Subnetwork",
                        "abstract": "Deep neural networks (DNNs) are vulnerable to backdoor attacks. Previous works have shown it extremely challenging to unlearn the undesired backdoor behavior from the network, since the entire network can be affected by the backdoor samples. In this paper, we propose a brand-new backdoor defense strategy, which makes it much easier to remove the harmful influence of backdoor samples from the model. Our defense strategy, Trap and Replace, consists of two stages. In the first stage, we bait and trap the backdoors in a small and easy-to-replace subnetwork. Specifically, we add an auxiliary image reconstruction head on top of the stem network shared with a light-weighted classification head. The intuition is that the auxiliary image reconstruction task encourages the stem network to keep sufficient low-level visual features that are hard to learn but semantically correct, instead of overfitting to the easy-to-learn but semantically incorrect backdoor correlations. As a result, when trained on backdoored datasets, the backdoors are easily baited towards the unprotected classification head, since it is much more vulnerable than the shared stem, leaving the stem network hardly poisoned. In the second stage, we replace the poisoned light-weighted classification head with an untainted one, by re-training it from scratch only on a small holdout dataset with clean samples, while fixing the stem network. As a result, both the stem and the classification head in the final network are hardly affected by backdoor training samples. We evaluate our method against ten different backdoor attacks. Our method outperforms previous state-of-the-art methods by up to 20.57%, 9.80%, and 13.72% attack success rate and on-average 3.14%, 1.80%, and 1.21% clean classification accuracy on CIFAR10, GTSRB, and ImageNet-12, respectively. Code is available at https://github.com/VITA-Group/Trap-and-Replace-Backdoor-Defense."
                    },
                    {
                        "title": "Removing Batch Normalization Boosts Adversarial Training",
                        "abstract": "Adversarial training (AT) defends deep neural networks against adversarial attacks. One challenge that limits its practical application is the performance degradation on clean samples. A major bottleneck identified by previous works is the widely used batch normalization (BN), which struggles to model the different statistics of clean and adversarial training samples in AT. Although the dominant approach is to extend BN to capture this mixture of distribution, we propose to completely eliminate this bottleneck by removing all BN layers in AT. Our normalizer-free robust training (NoFrost) method extends recent advances in normalizer-free networks to AT for its unexplored advantage on handling the mixture distribution challenge. We show that NoFrost achieves adversarial robustness with only a minor sacrifice on clean sample accuracy. On ImageNet with ResNet50, NoFrost achieves $74.06\\%$ clean accuracy, which drops merely $2.00\\%$ from standard training. In contrast, BN-based AT obtains $59.28\\%$ clean accuracy, suffering a significant $16.78\\%$ drop from standard training. In addition, NoFrost achieves a $23.56\\%$ adversarial robustness against PGD attack, which improves the $13.57\\%$ robustness in BN-based AT. We observe better model smoothness and larger decision margins from NoFrost, which make the models less sensitive to input perturbations and thus more robust. Moreover, when incorporating more data augmentations into NoFrost, it achieves comprehensive robustness against multiple distribution shifts. Code and pre-trained models are public at https://github.com/amazon-research/normalizer-free-robust-training."
                    },
                    {
                        "title": "AutoMARS: Searching to Compress Multi-Modality Recommendation Systems",
                        "abstract": "Web applications utilize Recommendation Systems (RS) to address the problem of consumer over-choices. Recent works have taken advantage of multi-modality or multi-view, input information (such as user interaction, images, texts, rating scores) to boost recommendation system performance compared with using single-modality information. However, the use of multi-modality input demands much higher computational cost and storage capacity. On the other hand, the real-world RS services usually have strict budgets on both time and space for a good customer experience. As a result, the model efficiency of multi-modality recommendation systems has gained increasing importance. While unfortunately, to the best of our knowledge, there is no existing study of a generic compression framework for multi-modality RS. In this paper, we investigate, for the first time, how to compress a multi-modality recommendation system with a fixed budget. Assuming that input information from different modalities are of unequal importance, a good compression algorithm should learn to automatically allocate different resource budgets to each input, based on their importance in maximally preserving recommendation efficacy. To this end, we leverage the tools of neural architecture search (NAS) and distillation and propose Auto Multi-modAlity Recommendation System (AutoMARS), a unified modality-aware model compression framework dedicated to multi-modality recommendation systems. We demonstrate the effectiveness and generality of AutoMARS by testing it on three different Amazon datasets of various sparsity. AutoMARS demonstrates superior multi-modality compression performance than previous state-of-the-art compression methods. For example on the Amazon Beauty dataset, we achieve on average a 20% higher accuracy over previous state-of-the-art methods, while enjoying 65% reduction over baselines. Codes are available at: https://github.com/VITA-Group/AutoMARS."
                    },
                    {
                        "title": "Partial and Asymmetric Contrastive Learning for Out-of-Distribution Detection in Long-Tailed Recognition",
                        "abstract": "Existing out-of-distribution (OOD) detection methods are typically benchmarked on training sets with balanced class distributions. However, in real-world applications, it is common for the training sets to have long-tailed distributions. In this work, we first demonstrate that existing OOD detection methods commonly suffer from significant performance degradation when the training set is long-tail distributed. Through analysis, we posit that this is because the models struggle to distinguish the minority tail-class in-distribution samples, from the true OOD samples, making the tail classes more prone to be falsely detected as OOD. To solve this problem, we propose Partial and Asymmetric Supervised Contrastive Learning (PASCL), which explicitly encourages the model to distinguish between tail-class in-distribution samples and OOD samples. To further boost in-distribution classification accuracy, we propose Auxiliary Branch Finetuning, which uses two separate branches of BN and classification layers for anomaly detection and in-distribution classification, respectively. The intuition is that in-distribution and OOD anomaly data have different underlying distributions. Our method outperforms previous state-of-the-art method by $1.29\\%$, $1.45\\%$, $0.69\\%$ anomaly detection false positive rate (FPR) and $3.24\\%$, $4.06\\%$, $7.89\\%$ in-distribution classification accuracy on CIFAR10-LT, CIFAR100-LT, and ImageNet-LT, respectively. Code and pre-trained models are available at https://github.com/amazon-research/long-tailed-ood-detection."
                    },
                    {
                        "title": "Efficient Split-Mix Federated Learning for On-Demand and In-Situ Customization",
                        "abstract": "Federated learning (FL) provides a distributed learning framework for multiple participants to collaborate learning without sharing raw data. In many practical FL scenarios, participants have heterogeneous resources due to disparities in hardware and inference dynamics that require quickly loading models of different sizes and levels of robustness. The heterogeneity and dynamics together impose significant challenges to existing FL approaches and thus greatly limit FL's applicability. In this paper, we propose a novel Split-Mix FL strategy for heterogeneous participants that, once training is done, provides in-situ customization of model sizes and robustness. Specifically, we achieve customization by learning a set of base sub-networks of different sizes and robustness levels, which are later aggregated on-demand according to inference requirements. This split-mix strategy achieves customization with high efficiency in communication, storage, and inference. Extensive experiments demonstrate that our method provides better in-situ customization than the existing heterogeneous-architecture FL methods. Codes and pre-trained models are available: https://github.com/illidanlab/SplitMix."
                    },
                    {
                        "title": "How Robust is Your Fairness? Evaluating and Sustaining Fairness under Unseen Distribution Shifts",
                        "abstract": "Increasing concerns have been raised on deep learning fairness in recent years. Existing fairness-aware machine learning methods mainly focus on the fairness of in-distribution data. However, in real-world applications, it is common to have distribution shift between the training and test data. In this paper, we first show that the fairness achieved by existing methods can be easily broken by slight distribution shifts. To solve this problem, we propose a novel fairness learning method termed CUrvature MAtching (CUMA), which can achieve robust fairness generalizable to unseen domains with unknown distributional shifts. Specifically, CUMA enforces the model to have similar generalization ability on the majority and minority groups, by matching the loss curvature distributions of the two groups. We evaluate our method on three popular fairness datasets. Compared with existing methods, CUMA achieves superior fairness under unseen distribution shifts, without sacrificing either the overall accuracy or the in-distribution fairness."
                    },
                    {
                        "title": "UMEC: Unified model and embedding compression for efficient recommendation systems",
                        "abstract": "in the"
                    },
                    {
                        "title": "AugMax: Adversarial Composition of Random Augmentations for Robust Training",
                        "abstract": "Data augmentation is a simple yet effective way to improve the robustness of deep neural networks (DNNs). Diversity and hardness are two complementary dimensions of data augmentation to achieve robustness. For example, AugMix explores random compositions of a diverse set of augmentations to enhance broader coverage, while adversarial training generates adversarially hard samples to spot the weakness. Motivated by this, we propose a data augmentation framework, termed AugMax, to unify the two aspects of diversity and hardness. AugMax first randomly samples multiple augmentation operators and then learns an adversarial mixture of the selected operators. Being a stronger form of data augmentation, AugMax leads to a significantly augmented input distribution which makes model training more challenging. To solve this problem, we further design a disentangled normalization module, termed DuBIN (Dual-Batch-and-Instance Normalization), that disentangles the instance-wise feature heterogeneity arising from AugMax. Experiments show that AugMax-DuBIN leads to significantly improved out-of-distribution robustness, outperforming prior arts by 3.03%, 3.49%, 1.82% and 0.71% on CIFAR10-C, CIFAR100-C, Tiny ImageNet-C and ImageNet-C. Codes and pretrained models are available: https://github.com/VITA-Group/AugMax."
                    },
                    {
                        "title": "Troubleshooting Blind Image Quality Models in the Wild",
                        "abstract": "Recently, the group maximum differentiation competition (gMAD) has been used to improve blind image quality assessment (BIQA) models, with the help of full-reference metrics. When applying this type of approach to troubleshoot \"best-performing\" BIQA models in the wild, we are faced with a practical challenge: it is highly nontrivial to obtain stronger competing models for efficient failure-spotting. Inspired by recent findings that difficult samples of deep models may be exposed through network pruning, we construct a set of \"self-competitors,\" as random ensembles of pruned versions of the target model to be improved. Diverse failures can then be efficiently identified via self-gMAD competition. Next, we fine-tune both the target and its pruned variants on the human-rated gMAD set. This allows all models to learn from their respective failures, preparing themselves for the next round of self-gMAD competition. Experimental results demonstrate that our method efficiently troubleshoots BIQA models in the wild with improved generalizability."
                    },
                    {
                        "title": "Federated Robustness Propagation: Sharing Adversarial Robustness in Federated Learning",
                        "abstract": "Federated learning (FL) emerges as a popular distributed learning schema that learns a model from a set of participating users without requiring raw data to be shared. One major challenge of FL comes from heterogeneity in users, which may have distributionally different (or non-iid ) data and varying computation resources. Just like in centralized learning, FL users also desire model robustness against malicious attackers at test time. Whereas adversarial training (AT) provides a sound solution for centralized learning, extending its usage for FL users has imposed signi\ufb01cant challenges, as many users may have very limited training data as well as tight computational budgets, to afford the data-hungry and costly AT. In this paper, we study a novel learning setting that propagates adversarial robustness from high-resource users that can afford AT, to those low-resource users that cannot afford it, during the FL process. We show that existing FL techniques cannot effectively propagate adversarial robustness among non-iid users, and propose a simple yet effective propagation approach that transfers robustness through carefully designed batch-normalization statistics. We demonstrate the rationality and effectiveness of our method through extensive experiments. Especially, the proposed method is shown to grant FL remarkable robustness even when only a small portion of users afford AT during learning. Codes will be published upon acceptance."
                    },
                    {
                        "title": "Practical Machine Learning Safety: A Survey and Primer",
                        "abstract": "The open-world deployment of Machine Learning (ML) algorithms in safety-critical applications such as autonomous vehicles needs to address a variety of ML vulnerabilities such as interpretability, verifiability, and performance limitations. Research explores different approaches to improve ML dependability by proposing new models and training techniques to reduce generalization error, achieve domain adaptation, and detect outlier examples and adversarial attacks. In this paper, we review and organize practical ML techniques that can improve the safety and dependability of ML algorithms and therefore ML-based software. Our organization maps state-of-the-art ML techniques to safety strategies in order to enhance the dependability of the ML algorithm from different aspects, and discuss research gaps as well as promising solutions"
                    },
                    {
                        "title": "Learning Model-Based Privacy Protection under Budget Constraints",
                        "abstract": "Protecting privacy in gradient-based learning has become increasingly critical as more sensitive information is being used. Many existing solutions seek to protect the sensitive gradients by constraining the overall privacy cost within a constant budget, where the protection is hand-designed and empirically calibrated to boost the utility of the resulting model. However, it remains challenging to choose the proper protection adapted for specific constraints so that the utility is maximized. To this end, we propose a novel Learning-to-Protect algorithm that automatically learns a model-based protector from a set of non-private learning tasks. The learned protector can be applied to private learning tasks to improve utility within the specific privacy budget constraint. Our empirical studies on both synthetic and real datasets demonstrate that the proposed algorithm can achieve a superior utility with a given privacy constraint and generalize well to new private datasets distributed differently as compared to the hand-designed competitors."
                    },
                    {
                        "title": "Federated Robustness Propagation: Sharing Adversarial Robustness in Heterogeneous Federated Learning",
                        "abstract": "Federated learning (FL) emerges as a popular distributed learning schema that learns a model from a set of participating users without sharing raw data. One major challenge of FL comes with heterogeneous users, who may have distributionally different (or non-iid) data and varying computation resources. As federated users would use the model for prediction, they often demand the trained model to be robust against malicious attackers at test time. Whereas adversarial training (AT) provides a sound solution for centralized learning, extending its usage for federated users has imposed significant challenges, as many users may have very limited training data and tight computational budgets, to afford the data-hungry and costly AT. In this paper, we study a novel FL strategy: propagating adversarial robustness from rich-resource users that can afford AT, to those with poor resources that cannot afford it, during federated learning. We show that existing FL techniques cannot be effectively integrated with the strategy to propagate robustness among non-iid users and propose an efficient propagation approach by the proper use of batch-normalization. We demonstrate the rationality and effectiveness of our method through extensive experiments. Especially, the proposed method is shown to grant federated models remarkable robustness even when only a small portion of users afford AT during learning. Source code will be released."
                    },
                    {
                        "title": "Taxonomy of Machine Learning Safety: A Survey and Primer",
                        "abstract": "The open-world deployment of Machine Learning (ML) algorithms in safety-critical applications such as autonomous vehicles needs to address a variety of ML vulnerabilities such as interpretability, verifiability, and performance limitations. Research explores different approaches to improve ML dependability by proposing new models and training techniques to reduce generalization error, achieve domain adaptation, and detect outlier examples and adversarial attacks. However, there is a missing connection between ongoing ML research and well-established safety principles. In this article, we present a structured and comprehensive review of ML techniques to improve the dependability of ML algorithms in uncontrolled open-world settings. From this review, we propose the Taxonomy of ML Safety that maps state-of-the-art ML techniques to key engineering safety strategies. Our taxonomy of ML safety presents a safety-oriented categorization of ML techniques to provide guidance for improving dependability of the ML design and development. The proposed taxonomy can serve as a safety checklist to aid designers in improving coverage and diversity of safety strategies employed in any given ML system."
                    },
                    {
                        "title": "AutoGAN-Distiller: Searching to Compress Generative Adversarial Networks",
                        "abstract": "The compression of Generative Adversarial Networks (GANs) has lately drawn attention, due to the increasing demand for deploying GANs into mobile devices for numerous applications such as image translation, enhancement and editing. However, compared to the substantial efforts to compressing other deep models, the research on compressing GANs (usually the generators) remains at its infancy stage. Existing GAN compression algorithms are limited to handling specific GAN architectures and losses. Inspired by the recent success of AutoML in deep compression, we introduce AutoML to GAN compression and develop an AutoGAN-Distiller (AGD) framework. Starting with a specifically designed efficient search space, AGD performs an end-to-end discovery for new efficient generators, given the target computational resource constraints. The search is guided by the original GAN model via knowledge distillation, therefore fulfilling the compression. AGD is fully automatic, standalone (i.e., needing no trained discriminators), and generically applicable to various GAN models. We evaluate AGD in two representative GAN tasks: image translation and super resolution. Without bells and whistles, AGD yields remarkably lightweight yet more competitive compressed models, that largely outperform existing alternatives. Our codes and pretrained models are available at this https URL."
                    },
                    {
                        "title": "I Am Going MAD: Maximum Discrepancy Competition for Comparing Classifiers Adaptively",
                        "abstract": "The learning of hierarchical representations for image classification has experienced an impressive series of successes due in part to the availability of large-scale labeled data for training. On the other hand, the trained classifiers have traditionally been evaluated on small and fixed sets of test images, which are deemed to be extremely sparsely distributed in the space of all natural images. It is thus questionable whether recent performance improvements on the excessively re-used test sets generalize to real-world natural images with much richer content variations. Inspired by efficient stimulus selection for testing perceptual models in psychophysical and physiological studies, we present an alternative framework for comparing image classifiers, which we name the MAximum Discrepancy (MAD) competition. Rather than comparing image classifiers using fixed test images, we adaptively sample a small test set from an arbitrarily large corpus of unlabeled images so as to maximize the discrepancies between the classifiers, measured by the distance over WordNet hierarchy. Human labeling on the resulting model-dependent image sets reveals the relative performance of the competing classifiers, and provides useful insights on potential ways to improve them. We report the MAD competition results of eleven ImageNet classifiers while noting that the framework is readily extensible and cost-effective to add future classifiers into the competition. Codes can be found at this https URL."
                    }
                ]
            },
            "d1d8e78c-1bf5-4b77-bacb-cb369dd1fc2f": {
                "pk": "d1d8e78c-1bf5-4b77-bacb-cb369dd1fc2f",
                "name": "Zhangyang Wang",
                "collaborators": [
                    "Junyuan Hong",
                    "Bo Li",
                    "Zhangheng Li",
                    "Chulin Xie",
                    "Dan Hendrycks",
                    "Dawn Song",
                    "Chenhui Zhang",
                    "Qinbin Li",
                    "Jeffrey Tan",
                    "Rachel Xin"
                ],
                "domain": [
                    "Privacy",
                    "Large Language Models",
                    "Differential Privacy",
                    "Trustworthiness"
                ],
                "publications": [
                    {
                        "title": "Shake to Leak: Fine-tuning Diffusion Models Can Amplify the Generative Privacy Risk",
                        "abstract": "While diffusion models have recently demonstrated remarkable progress in generating realistic images, privacy risks also arise: published models or APIs could generate training images and thus leak privacy-sensitive training information. In this paper, we reveal a new risk, Shake-to-Leak (S2L), that fine-tuning the pre-trained models with manipulated data can amplify the existing privacy risks. We demonstrate that S2L could occur in various standard fine-tuning strategies for diffusion models, including concept-injection methods (DreamBooth and Textual Inversion) and parameter-efficient methods (LoRA and Hypernetwork), as well as their combinations. In the worst case, S2L can amplify the state-of-the-art membership inference attack (MIA) on diffusion models by 5.4% (absolute difference) AUC and can increase extracted private samples from almost 0 samples to 16.3 samples on average per target domain. This discovery underscores that the privacy risk with diffusion models is even more severe than previously recognized. Codes are available at https://github.com/VITA-Group/Shake-to-Leak."
                    },
                    {
                        "title": "LLM-PBE: Assessing Data Privacy in Large Language Models",
                        "abstract": "Large Language Models (LLMs) have become integral to numerous domains, significantly advancing applications in data management, mining, and analysis. Their profound capabilities in processing and interpreting complex language data, however, bring to light pressing concerns regarding data privacy, especially the risk of unintentional training data leakage. Despite the critical nature of this issue, there has been no existing literature to offer a comprehensive assessment of data privacy risks in LLMs. Addressing this gap, our paper introduces LLM-PBE, a toolkit crafted specifically for the systematic evaluation of data privacy risks in LLMs. LLM-PBE is designed to analyze privacy across the entire lifecycle of LLMs, incorporating diverse attack and defense strategies, and handling various data types and metrics. Through detailed experimentation with multiple LLMs, LLM-PBE facilitates an in-depth exploration of data privacy concerns, shedding light on influential factors such as model size, data characteristics, and evolving temporal dimensions. This study not only enriches the understanding of privacy issues in LLMs but also serves as a vital resource for future research in the field. Aimed at enhancing the breadth of knowledge in this area, the findings, resources, and our full technical report are made available at https://llm-pbe.github.io/, providing an open platform for academic and practical advancements in LLM privacy assessment."
                    },
                    {
                        "title": "Decoding Compressed Trust: Scrutinizing the Trustworthiness of Efficient LLMs Under Compression",
                        "abstract": "Compressing high-capability Large Language Models (LLMs) has emerged as a favored strategy for resource-efficient inferences. While state-of-the-art (SoTA) compression methods boast impressive advancements in preserving benign task performance, the potential risks of compression in terms of safety and trustworthiness have been largely neglected. This study conducts the first, thorough evaluation of three (3) leading LLMs using five (5) SoTA compression techniques across eight (8) trustworthiness dimensions. Our experiments highlight the intricate interplay between compression and trustworthiness, revealing some interesting patterns. We find that quantization is currently a more effective approach than pruning in achieving efficiency and trustworthiness simultaneously. For instance, a 4-bit quantized model retains the trustworthiness of its original counterpart, but model pruning significantly degrades trustworthiness, even at 50% sparsity. Moreover, employing quantization within a moderate bit range could unexpectedly improve certain trustworthiness dimensions such as ethics and fairness. Conversely, extreme quantization to very low bit levels (3 bits) tends to reduce trustworthiness significantly. This increased risk cannot be uncovered by looking at benign performance alone, in turn, mandating comprehensive trustworthiness evaluation in practice. These findings culminate in practical recommendations for simultaneously achieving high utility, efficiency, and trustworthiness in LLMs. Code and models are available at https://decoding-comp-trust.github.io."
                    },
                    {
                        "title": "DP-OPT: Make Large Language Model Your Privacy-Preserving Prompt Engineer",
                        "abstract": "Large Language Models (LLMs) have emerged as dominant tools for various tasks, particularly when tailored for a specific target by prompt tuning. Nevertheless, concerns surrounding data privacy present obstacles due to the tuned prompts' dependency on sensitive private information. A practical solution is to host a local LLM and optimize a soft prompt privately using data. Yet, hosting a local model becomes problematic when model ownership is protected. Alternative methods, like sending data to the model's provider for training, intensify these privacy issues facing an untrusted provider. In this paper, we present a novel solution called Differentially-Private Offsite Prompt Tuning (DP-OPT) to address this challenge. Our approach involves tuning a discrete prompt on the client side and then applying it to the desired cloud models. We demonstrate that prompts suggested by LLMs themselves can be transferred without compromising performance significantly. To ensure that the prompts do not leak private information, we introduce the first private prompt generation mechanism, by a differentially-private (DP) ensemble of in-context learning with private demonstrations. With DP-OPT, generating privacy-preserving prompts by Vicuna-7b can yield competitive performance compared to non-private in-context learning on GPT3.5 or local private prompt tuning. Codes are available at https://github.com/VITA-Group/DP-OPT ."
                    }
                ]
            },
            "b23eaf5f-3565-4e91-b95e-baa4663bfc00": {
                "pk": "b23eaf5f-3565-4e91-b95e-baa4663bfc00",
                "name": "Jiayu Zhou",
                "collaborators": [
                    "Junyuan Hong",
                    "Zhangyang Wang",
                    "Haotao Wang",
                    "Shuyang Yu",
                    "Yuyang Deng",
                    "M. Mahdavi",
                    "Yi Zeng",
                    "Haobo Zhang",
                    "Lingjuan Lyu",
                    "S. Drew"
                ],
                "domain": [
                    "Federated Learning",
                    "Privacy Protection",
                    "Deep Learning",
                    "Knowledge Distillation"
                ],
                "publications": [
                    {
                        "title": "On the Generalization Ability of Unsupervised Pretraining",
                        "abstract": "Recent advances in unsupervised learning have shown that unsupervised pre-training, followed by fine-tuning, can improve model generalization. However, a rigorous understanding of how the representation function learned on an unlabeled dataset affects the generalization of the fine-tuned model is lacking. Existing theoretical research does not adequately account for the heterogeneity of the distribution and tasks in pre-training and fine-tuning stage. To bridge this gap, this paper introduces a novel theoretical framework that illuminates the critical factor influencing the transferability of knowledge acquired during unsupervised pre-training to the subsequent fine-tuning phase, ultimately affecting the generalization capabilities of the fine-tuned model on downstream tasks. We apply our theoretical framework to analyze generalization bound of two distinct scenarios: Context Encoder pre-training with deep neural networks and Masked Autoencoder pre-training with deep transformers, followed by fine-tuning on a binary classification task. Finally, inspired by our findings, we propose a novel regularization method during pre-training to further enhances the generalization of fine-tuned model. Overall, our results contribute to a better understanding of unsupervised pre-training and fine-tuning paradigm, and can shed light on the design of more effective pre-training algorithms."
                    },
                    {
                        "title": "Turning the Curse of Heterogeneity in Federated Learning into a Blessing for Out-of-Distribution Detection",
                        "abstract": "Deep neural networks have witnessed huge successes in many challenging prediction tasks and yet they often suffer from out-of-distribution (OoD) samples, misclassifying them with high con\ufb01dence. Recent advances show promising OoD detection performance for centralized training, and however, OoD detection in federated learning (FL) is largely overlooked, even though many security sensitive applications such as autonomous driving and voice recognition authorization are commonly trained using FL for data privacy concerns. The main challenge that prevents previous state-of-the-art OoD detection methods from being incorporated to FL is that they require large amount of real OoD samples. However, in real-world scenarios, such large-scale OoD training data can be costly or even infeasible to obtain, especially for resource-limited local devices. On the other hand, a notorious challenge in FL is data heterogeneity where each client collects non-identically and independently distributed (non-iid) data. We propose to take advantage of such heterogeneity and turn the curse into a blessing that facilitates OoD detection in FL. The key is that for each client, non-iid data from other clients (unseen external classes) can serve as an alternative to real OoD samples. Speci\ufb01cally, we propose a novel Federated Out-of-Distribution Synthesizer (F OSTER ), which learns a class-conditional generator to synthesize virtual external-class OoD samples, and maintains data con\ufb01dentiality and communication ef\ufb01ciency required by FL. Experimental results show that our method outperforms the state-of-the-art for"
                    },
                    {
                        "title": "FedNoisy: Federated Noisy Label Learning Benchmark",
                        "abstract": "Federated learning has gained popularity for distributed learning without aggregating sensitive data from clients. But meanwhile, the distributed and isolated nature of data isolation may be complicated by data quality, making it more vulnerable to noisy labels. Many efforts exist to defend against the negative impacts of noisy labels in centralized or federated settings. However, there is a lack of a benchmark that comprehensively considers the impact of noisy labels in a wide variety of typical FL settings. In this work, we serve the first standardized benchmark that can help researchers fully explore potential federated noisy settings. Also, we conduct comprehensive experiments to explore the characteristics of these data settings and unravel challenging scenarios on the federated noisy label learning, which may guide method development in the future. We highlight the 20 basic settings for more than 5 datasets proposed in our benchmark and standardized simulation pipeline for federated noisy label learning. We hope this benchmark can facilitate idea verification in federated learning with noisy labels. \\texttt{FedNoisy} is available at \\codeword{https://github.com/SMILELab-FL/FedNoisy}."
                    },
                    {
                        "title": "Who Leaked the Model? Tracking IP Infringers in Accountable Federated Learning",
                        "abstract": "Federated learning (FL) emerges as an effective collaborative learning framework to coordinate data and computation resources from massive and distributed clients in training. Such collaboration results in non-trivial intellectual property (IP) represented by the model parameters that should be protected and shared by the whole party rather than an individual user. Meanwhile, the distributed nature of FL endorses a malicious client the convenience to compromise IP through illegal model leakage to unauthorized third parties. To block such IP leakage, it is essential to make the IP identifiable in the shared model and locate the anonymous infringer who first leaks it. The collective challenges call for \\emph{accountable federated learning}, which requires verifiable ownership of the model and is capable of revealing the infringer's identity upon leakage. In this paper, we propose Decodable Unique Watermarking (DUW) for complying with the requirements of accountable FL. Specifically, before a global model is sent to a client in an FL round, DUW encodes a client-unique key into the model by leveraging a backdoor-based watermark injection. To identify the infringer of a leaked model, DUW examines the model and checks if the triggers can be decoded as the corresponding keys. Extensive empirical results show that DUW is highly effective and robust, achieving over $99\\%$ watermark success rate for Digits, CIFAR-10, and CIFAR-100 datasets under heterogeneous FL settings, and identifying the IP infringer with $100\\%$ accuracy even after common watermark removal attempts."
                    },
                    {
                        "title": "Understanding Deep Gradient Leakage via Inversion Influence Functions",
                        "abstract": "Deep Gradient Leakage (DGL) is a highly effective attack that recovers private training images from gradient vectors. This attack casts significant privacy challenges on distributed learning from clients with sensitive data, where clients are required to share gradients. Defending against such attacks requires but lacks an understanding of when and how privacy leakage happens, mostly because of the black-box nature of deep networks. In this paper, we propose a novel Inversion Influence Function (I2F) that establishes a closed-form connection between the recovered images and the private gradients by implicitly solving the DGL problem. Compared to directly solving DGL, I2F is scalable for analyzing deep networks, requiring only oracle access to gradients and Jacobian-vector products. We empirically demonstrate that I2F effectively approximated the DGL generally on different model architectures, datasets, modalities, attack implementations, and perturbation-based defenses. With this novel tool, we provide insights into effective gradient perturbation directions, the unfairness of privacy protection, and privacy-preferred model initialization. Our codes are provided in https://github.com/illidanlab/inversion-influence-function."
                    },
                    {
                        "title": "Revisiting Data-Free Knowledge Distillation with Poisoned Teachers",
                        "abstract": "Data-free knowledge distillation (KD) helps transfer knowledge from a pre-trained model (known as the teacher model) to a smaller model (known as the student model) without access to the original training data used for training the teacher model. However, the security of the synthetic or out-of-distribution (OOD) data required in data-free KD is largely unknown and under-explored. In this work, we make the first effort to uncover the security risk of data-free KD w.r.t. untrusted pre-trained models. We then propose Anti-Backdoor Data-Free KD (ABD), the first plug-in defensive method for data-free KD methods to mitigate the chance of potential backdoors being transferred. We empirically evaluate the effectiveness of our proposed ABD in diminishing transferred backdoor knowledge while maintaining compatible downstream performances as the vanilla KD. We envision this work as a milestone for alarming and mitigating the potential backdoors in data-free KD. Codes are released at https://github.com/illidanlab/ABD."
                    },
                    {
                        "title": "MECTA: Memory-Economic Continual Test-Time Model Adaptation",
                        "abstract": "Continual Test-time Adaptation (CTA) is a promising art to secure accuracy gains in continually-changing environments. The state-of-the-art adaptations improve out-of-distribution model accuracy via computation-efficient online test-time gradient descents but meanwhile cost about times of memory versus the inference, even if only a small portion of parameters are updated. Such high memory consumption of CTA substantially impedes wide applications of advanced CTA on memory-constrained devices. In this paper, we provide a novel solution, dubbed MECTA, to drastically improve the memory efficiency of gradient-based CTA. Our profiling shows that the major memory overhead comes from the intermediate cache for back-propagation, which scales by the batch size, channel, and layer number. Therefore, we propose to reduce batch sizes, adopt an adaptive normalization layer to maintain stable and accurate predictions, and stop the back-propagation caching heuristically. On the other hand, we prune the networks to reduce the computation and memory overheads in optimization and recover the parameters afterward to avoid forgetting. The proposed MECTA is efficient and can be seamlessly plugged into state-of-the-art CTA algorithms at negligible overhead on computation and memory. On three datasets, CIFAR10, CIFAR100, and ImageNet, MECTA improves the accuracy by at least 6% with constrained memory and significantly reduces the memory costs of ResNet50 on ImageNet by at least 70% with comparable accuracy. Our codes can be accessed at https://github.com/SonyAI/MECTA ."
                    },
                    {
                        "title": "A Privacy-Preserving Hybrid Federated Learning Framework for Financial Crime Detection",
                        "abstract": "The recent decade witnessed a surge of increase in financial crimes across the public and private sectors, with an average cost of scams of $102m to financial institutions in 2022. Developing a mechanism for battling financial crimes is an impending task that requires in-depth collaboration from multiple institutions, and yet such collaboration imposed significant technical challenges due to the privacy and security requirements of distributed financial data. For example, consider the modern payment network systems, which can generate millions of transactions per day across a large number of global institutions. Training a detection model of fraudulent transactions requires not only secured transactions but also the private account activities of those involved in each transaction from corresponding bank systems. The distributed nature of both samples and features prevents most existing learning systems from being directly adopted to handle the data mining task. In this paper, we collectively address these challenges by proposing a hybrid federated learning system that offers secure and privacy-aware learning and inference for financial crime detection. We conduct extensive empirical studies to evaluate the proposed framework's detection performance and privacy-protection capability, evaluating its robustness against common malicious attacks of collaborative learning. We release our source code at https://github.com/illidanlab/HyFL ."
                    },
                    {
                        "title": "Trap and Replace: Defending Backdoor Attacks by Trapping Them into an Easy-to-Replace Subnetwork",
                        "abstract": "Deep neural networks (DNNs) are vulnerable to backdoor attacks. Previous works have shown it extremely challenging to unlearn the undesired backdoor behavior from the network, since the entire network can be affected by the backdoor samples. In this paper, we propose a brand-new backdoor defense strategy, which makes it much easier to remove the harmful influence of backdoor samples from the model. Our defense strategy, Trap and Replace, consists of two stages. In the first stage, we bait and trap the backdoors in a small and easy-to-replace subnetwork. Specifically, we add an auxiliary image reconstruction head on top of the stem network shared with a light-weighted classification head. The intuition is that the auxiliary image reconstruction task encourages the stem network to keep sufficient low-level visual features that are hard to learn but semantically correct, instead of overfitting to the easy-to-learn but semantically incorrect backdoor correlations. As a result, when trained on backdoored datasets, the backdoors are easily baited towards the unprotected classification head, since it is much more vulnerable than the shared stem, leaving the stem network hardly poisoned. In the second stage, we replace the poisoned light-weighted classification head with an untainted one, by re-training it from scratch only on a small holdout dataset with clean samples, while fixing the stem network. As a result, both the stem and the classification head in the final network are hardly affected by backdoor training samples. We evaluate our method against ten different backdoor attacks. Our method outperforms previous state-of-the-art methods by up to 20.57%, 9.80%, and 13.72% attack success rate and on-average 3.14%, 1.80%, and 1.21% clean classification accuracy on CIFAR10, GTSRB, and ImageNet-12, respectively. Code is available at https://github.com/VITA-Group/Trap-and-Replace-Backdoor-Defense."
                    },
                    {
                        "title": "Resilient and Communication Efficient Learning for Heterogeneous Federated Systems",
                        "abstract": "The rise of Federated Learning (FL) is bringing machine learning to edge computing by utilizing data scattered across edge devices. However, the heterogeneity of edge network topologies and the uncertainty of wireless transmission are two major obstructions of FL's wide application in edge computing, leading to prohibitive convergence time and high communication cost. In this work, we propose an FL scheme to address both challenges simultaneously. Specifically, we enable edge devices to learn self-distilled neural networks that are readily prunable to arbitrary sizes, which capture the knowledge of the learning domain in a nested and progressive manner. Not only does our approach tackle system heterogeneity by serving edge devices with varying model architectures, but it also alleviates the issue of connection uncertainty by allowing transmitting part of the model parameters under faulty network connections, without wasting the contributing knowledge of the transmitted parameters. Extensive empirical studies show that under system heterogeneity and network instability, our approach demonstrates significant resilience and higher communication efficiency compared to the state-of-the-art."
                    },
                    {
                        "title": "Efficient Split-Mix Federated Learning for On-Demand and In-Situ Customization",
                        "abstract": "Federated learning (FL) provides a distributed learning framework for multiple participants to collaborate learning without sharing raw data. In many practical FL scenarios, participants have heterogeneous resources due to disparities in hardware and inference dynamics that require quickly loading models of different sizes and levels of robustness. The heterogeneity and dynamics together impose significant challenges to existing FL approaches and thus greatly limit FL's applicability. In this paper, we propose a novel Split-Mix FL strategy for heterogeneous participants that, once training is done, provides in-situ customization of model sizes and robustness. Specifically, we achieve customization by learning a set of base sub-networks of different sizes and robustness levels, which are later aggregated on-demand according to inference requirements. This split-mix strategy achieves customization with high efficiency in communication, storage, and inference. Extensive experiments demonstrate that our method provides better in-situ customization than the existing heterogeneous-architecture FL methods. Codes and pre-trained models are available: https://github.com/illidanlab/SplitMix."
                    },
                    {
                        "title": "How Robust is Your Fairness? Evaluating and Sustaining Fairness under Unseen Distribution Shifts",
                        "abstract": "Increasing concerns have been raised on deep learning fairness in recent years. Existing fairness-aware machine learning methods mainly focus on the fairness of in-distribution data. However, in real-world applications, it is common to have distribution shift between the training and test data. In this paper, we first show that the fairness achieved by existing methods can be easily broken by slight distribution shifts. To solve this problem, we propose a novel fairness learning method termed CUrvature MAtching (CUMA), which can achieve robust fairness generalizable to unseen domains with unknown distributional shifts. Specifically, CUMA enforces the model to have similar generalization ability on the majority and minority groups, by matching the loss curvature distributions of the two groups. We evaluate our method on three popular fairness datasets. Compared with existing methods, CUMA achieves superior fairness under unseen distribution shifts, without sacrificing either the overall accuracy or the in-distribution fairness."
                    },
                    {
                        "title": "Outsourcing Training without Uploading Data via Efficient Collaborative Open-Source Sampling",
                        "abstract": "As deep learning blooms with growing demand for computation and data resources, outsourcing model training to a powerful cloud server becomes an attractive alternative to training at a low-power and cost-effective end device. Traditional outsourcing requires uploading device data to the cloud server, which can be infeasible in many real-world applications due to the often sensitive nature of the collected data and the limited communication bandwidth. To tackle these challenges, we propose to leverage widely available open-source data, which is a massive dataset collected from public and heterogeneous sources (e.g., Internet images). We develop a novel strategy called Efficient Collaborative Open-source Sampling (ECOS) to construct a proximal proxy dataset from open-source data for cloud training, in lieu of client data. ECOS probes open-source data on the cloud server to sense the distribution of client data via a communication- and computation-efficient sampling process, which only communicates a few compressed public features and client scalar responses. Extensive empirical studies show that the proposed ECOS improves the quality of automated client labeling, model compression, and label outsourcing when applied in various learning scenarios."
                    },
                    {
                        "title": "Dynamic Privacy Budget Allocation Improves Data Efficiency of Differentially Private Gradient Descent",
                        "abstract": "Protecting privacy in learning while maintaining the model performance has become increasingly critical in many applications that involve sensitive data. A popular private learning framework is differentially private learning composed of many privatized gradient iterations by noising and clipping. Under the privacy constraint, it has been shown that the dynamic policies could improve the final iterate loss, namely the quality of published models. In this talk, we will introduce these dynamic techniques for learning rate, batch size, noise magnitude and gradient clipping. Also, we discuss how the dynamic policy could change the convergence bounds which further provides insight of the impact of dynamic methods."
                    },
                    {
                        "title": "Data-Free Knowledge Distillation for Heterogeneous Federated Learning",
                        "abstract": "Federated Learning (FL) is a decentralized machine-learning paradigm in which a global server iteratively aggregates the model parameters of local users without accessing their data. User heterogeneity has imposed significant challenges to FL, which can incur drifted global models that are slow to converge. Knowledge Distillation has recently emerged to tackle this issue, by refining the server model using aggregated knowledge from heterogeneous users, other than directly aggregating their model parameters. This approach, however, depends on a proxy dataset, making it impractical unless such prerequisite is satisfied. Moreover, the ensemble knowledge is not fully utilized to guide local model learning, which may in turn affect the quality of the aggregated model. In this work, we propose a data-free knowledge distillation approach to address heterogeneous FL, where the server learns a lightweight generator to ensemble user information in a data-free manner, which is then broadcasted to users, regulating local training using the learned knowledge as an inductive bias. Empirical studies powered by theoretical implications show that, our approach facilitates FL with better generalization performance using fewer communication rounds, compared with the state-of-the-art."
                    },
                    {
                        "title": "On Dynamic Noise Influence in Differentially Private Learning",
                        "abstract": "Protecting privacy in learning while maintaining the model performance has become increasingly critical in many applications that involve sensitive data. Private Gradient Descent (PGD) is a commonly used private learning framework, which noises gradients based on the Differential Privacy protocol. Recent studies show that dynamic privacy schedules of decreasing noise magnitudes can improve loss at the final iteration, and yet theoretical understandings of the effectiveness of such schedules and their connections to optimization algorithms remain limited. In this paper, we provide comprehensive analysis of noise influence in dynamic privacy schedules to answer these critical questions. We first present a dynamic noise schedule minimizing the utility upper bound of PGD, and show how the noise influence from each optimization step collectively impacts utility of the final model. Our study also reveals how impacts from dynamic noise influence change when momentum is used. We empirically show the connection exists for general non-convex losses, and the influence is greatly impacted by the loss curvature."
                    },
                    {
                        "title": "Federated Robustness Propagation: Sharing Adversarial Robustness in Federated Learning",
                        "abstract": "Federated learning (FL) emerges as a popular distributed learning schema that learns a model from a set of participating users without requiring raw data to be shared. One major challenge of FL comes from heterogeneity in users, which may have distributionally different (or non-iid ) data and varying computation resources. Just like in centralized learning, FL users also desire model robustness against malicious attackers at test time. Whereas adversarial training (AT) provides a sound solution for centralized learning, extending its usage for FL users has imposed signi\ufb01cant challenges, as many users may have very limited training data as well as tight computational budgets, to afford the data-hungry and costly AT. In this paper, we study a novel learning setting that propagates adversarial robustness from high-resource users that can afford AT, to those low-resource users that cannot afford it, during the FL process. We show that existing FL techniques cannot effectively propagate adversarial robustness among non-iid users, and propose a simple yet effective propagation approach that transfers robustness through carefully designed batch-normalization statistics. We demonstrate the rationality and effectiveness of our method through extensive experiments. Especially, the proposed method is shown to grant FL remarkable robustness even when only a small portion of users afford AT during learning. Codes will be published upon acceptance."
                    },
                    {
                        "title": "Learning Model-Based Privacy Protection under Budget Constraints",
                        "abstract": "Protecting privacy in gradient-based learning has become increasingly critical as more sensitive information is being used. Many existing solutions seek to protect the sensitive gradients by constraining the overall privacy cost within a constant budget, where the protection is hand-designed and empirically calibrated to boost the utility of the resulting model. However, it remains challenging to choose the proper protection adapted for specific constraints so that the utility is maximized. To this end, we propose a novel Learning-to-Protect algorithm that automatically learns a model-based protector from a set of non-private learning tasks. The learned protector can be applied to private learning tasks to improve utility within the specific privacy budget constraint. Our empirical studies on both synthetic and real datasets demonstrate that the proposed algorithm can achieve a superior utility with a given privacy constraint and generalize well to new private datasets distributed differently as compared to the hand-designed competitors."
                    }
                ]
            }
        }
    },
    "2406.07217": {
        "paper_data": {
            "title": "A Synthetic Dataset for Personal Attribute Inference",
            "url": "http://arxiv.org/abs/2406.07217v1",
            "arxiv_id": "2406.07217",
            "authors": [
                "Hanna Yukhymenko",
                "Robin Staab",
                "Mark Vero",
                "Martin Vechev"
            ],
            "abstract": "Recently, powerful Large Language Models (LLMs) have become easily accessible to hundreds of millions of users worldwide. However, their strong capabilities and vast world knowledge do not come without associated privacy risks. In this work, we focus on the emerging privacy threat LLMs pose - the ability to accurately infer personal information from online texts. Despite the growing importance of LLM-based author profiling, research in this area has been hampered by a lack of suitable public datasets, largely due to ethical and privacy concerns associated with real personal data. In this work, we take two steps to address this problem: (i) we construct a simulation framework for the popular social media platform Reddit using LLM agents seeded with synthetic personal profiles; (ii) using this framework, we generate SynthPAI, a diverse synthetic dataset of over 7800 comments manually labeled for personal attributes. We validate our dataset with a human study showing that humans barely outperform random guessing on the task of distinguishing our synthetic comments from real ones. Further, we verify that our dataset enables meaningful personal attribute inference research by showing across 18 state-of-the-art LLMs that our synthetic comments allow us to draw the same conclusions as real-world data. Together, this indicates that our dataset and pipeline provide a strong and privacy-preserving basis for future research toward understanding and mitigating the inference-based privacy threats LLMs pose.",
            "introduction": " Introduction The increasing capabilities of Large Language Models (LLMs) alongside their widespread adoption have sparked ample discussions about their potential misuse [ 1,2]. Besides issues such as toxic, unsafe, or otherwise harmful generations, potential privacy implications of LLMs have been receiving an increasing amount of attention [ 3\u20135]. Most existing LLM privacy research focuses on memoriza- tion, the issue of models repeating training data at inference time [ 3]. However, as recently shown by Staab et al. [6], and previously outlined by Weidinger et al. [7], current frontier LLMs can also be misused to accurately infer personal attributes from online texts at a scale unattainable by humans. Even despite the legal relevance of such automated personal attribute inferences (PAIs) (\u00a72), research in this area has been significantly held back by a lack of suitable openly available datasets. The PAI Data Gap This lack of data is a result of multiple factors: As PAI research, by nature, deals with personal (and sensitive) data underlying strict data protection regulations (e.g., GDPR \u00a72), making real-world datasets public is legally challenging. While prior work on PAI used real-world data [ 6,8], they each had to construct their datasets from scratch, not releasing them due to associated ethical and legal concerns. Staab et al. [6]tried to address this issue to an extent by releasing synthetic examples instead of their actual data. However, they acknowledge that these samples fall short of Preprint.arXiv:2406.07217v1  [cs.LG]  11 Jun 2024\u2026 Synthetic Pro\ufb01les  Turn-Wise Interaction  in Comment Threads  LLM-Aided Personal  Attribute Labeling  + + \u2026Age: 26    Sex: Male    Income: Low  Sex: Female    Place of Birth: NYC  1 2 3 Personal LLM Agents  Agents Add  Comments Figure 1: Overview of our personalized LLM agent-based thread simulation framework and the curation of SynthPAI. First, in step 1, we create diverse synthetic profiles and seed LLM agents with them. Then, in step 2, we let the agents interact to generate comment threads. Finally, in step 3, aided by an LLM, we label the generated comments for inferrable personal attributes. real-world texts both in setting and style and, as such, are not a viable long-term replacement for real data. Additionally, due to the considerable effort in obtaining personal attribute labels over free text, even the few partially available datasets (\u00a72) contain labels for only a very restricted set of attributes. This Work In this work, we bridge this gap by (i) introducing a novel framework simulating popular comment-thread-focused social media platforms such as Reddit using personalized LLM agents and (ii) instantiating this framework to produce a synthetic dataset, SynthPAI, of over 7800 comments with hand-curated personal attribute labels. As our pipeline does not require any real data, it is fully privacy-preserving. Our experimental evaluation in \u00a74 shows that SynthPAI is realistic, diverse, and enables PAI research that is representative of results. In Fig. 19a we present the agreement between participants on the labels of the comments. Notably we find that there is no significant difference in the agreement between participants based on the source of the comment. This indicates that even across humans the comments from SynthPAI are perceived as realistic as the real comments. In Fig. 19b we present the histogram of reviewer accuracies. We find that the majority of",
            "references": [
                {
                    "title": "Private Attribute Inference from Images with Vision-Language Models",
                    "abstract": "As large language models (LLMs) become ubiquitous in our daily tasks and digital interactions, associated privacy risks are increasingly in focus. While LLM privacy research has primarily focused on the leakage of model training data, it has recently been shown that the increase in models' capabilities has enabled LLMs to make accurate privacy-infringing inferences from previously unseen texts. With the rise of multimodal vision-language models (VLMs), capable of understanding both images and text, a pertinent question is whether such results transfer to the previously unexplored domain of benign images posted online. To investigate the risks associated with the image reasoning capabilities of newly emerging VLMs, we compile an image dataset with human-annotated labels of the image owner's personal attributes. In order to understand the additional privacy risk posed by VLMs beyond traditional human attribute recognition, our dataset consists of images where the inferable private attributes do not stem from direct depictions of humans. On this dataset, we evaluate the inferential capabilities of 7 state-of-the-art VLMs, finding that they can infer various personal attributes at up to 77.6% accuracy. Concerningly, we observe that accuracy scales with the general capabilities of the models, implying that future models can be misused as stronger adversaries, establishing an imperative for the development of adequate defenses."
                },
                {
                    "title": "Yi: Open Foundation Models by 01.AI",
                    "abstract": "We introduce the Yi model family, a series of language and multimodal models that demonstrate strong multi-dimensional capabilities. The Yi model family is based on 6B and 34B pretrained language models, then we extend them to chat models, 200K long context models, depth-upscaled models, and vision-language models. Our base models achieve strong performance on a wide range of benchmarks like MMLU, and our finetuned chat models deliver strong human preference rate on major evaluation platforms like AlpacaEval and Chatbot Arena. Building upon our scalable super-computing infrastructure and the classical transformer architecture, we attribute the performance of Yi models primarily to its data quality resulting from our data-engineering efforts. For pretraining, we construct 3.1 trillion tokens of English and Chinese corpora using a cascaded data deduplication and quality filtering pipeline. For finetuning, we polish a small scale (less than 10K) instruction dataset over multiple iterations such that every single instance has been verified directly by our machine learning engineers. For vision-language, we combine the chat language model with a vision transformer encoder and train the model to align visual representations to the semantic space of the language model. We further extend the context length to 200K through lightweight continual pretraining and demonstrate strong needle-in-a-haystack retrieval performance. We show that extending the depth of the pretrained checkpoint through continual pretraining further improves performance. We believe that given our current results, continuing to scale up model parameters using thoroughly optimized data will lead to even stronger frontier models."
                },
                {
                    "title": "Large Language Models are Advanced Anonymizers",
                    "abstract": "Recent work in privacy research on large language models has shown that they achieve near human-level performance at inferring personal data from real-world online texts. With consistently increasing model capabilities, existing text anonymization methods are currently lacking behind regulatory requirements and adversarial threats. This raises the question of how individuals can effectively protect their personal data in sharing online texts. In this work, we take two steps to answer this question: We first present a new setting for evaluating anonymizations in the face of adversarial LLMs inferences, allowing for a natural measurement of anonymization performance while remedying some of the shortcomings of previous metrics. We then present our LLM-based adversarial anonymization framework leveraging the strong inferential capabilities of LLMs to inform our anonymization procedure. In our experimental evaluation, we show on real-world and synthetic online texts how adversarial anonymization outperforms current industry-grade anonymizers both in terms of the resulting utility and privacy."
                },
                {
                    "title": "Reducing Privacy Risks in Online Self-Disclosures with Language Models",
                    "abstract": "Self-disclosure, while being common and rewarding in social media interaction, also poses privacy risks. In this paper, we take the initiative to protect the user-side privacy associated with online self-disclosure through detection and abstraction. We develop a taxonomy of 19 self-disclosure categories and curate a large corpus consisting of 4.8K annotated disclosure spans. We then fine-tune a language model for detection, achieving over 65% partial span F$_1$. We further conduct an HCI user study, with 82% of participants viewing the model positively, highlighting its real-world applicability. Motivated by the user feedback, we introduce the task of self-disclosure abstraction, which is rephrasing disclosures into less specific terms while preserving their utility, e.g.,\"Im 16F\"to\"I'm a teenage girl\". We explore various fine-tuning strategies, and our best model can generate diverse abstractions that moderately reduce privacy risks while maintaining high utility according to human evaluation. To help users in deciding which disclosures to abstract, we present a task of rating their importance for context understanding. Our fine-tuned model achieves 80% accuracy, on-par with GPT-3.5. Given safety and privacy considerations, we will only release our corpus and models to researcher who agree to the ethical guidelines outlined in Ethics Statement."
                },
                {
                    "title": "When Large Language Models contradict humans? Large Language Models' Sycophantic Behaviour",
                    "abstract": "Large Language Models have been demonstrating the ability to solve complex tasks by delivering answers that are positively evaluated by humans due in part to the intensive use of human feedback that refines responses. However, the suggestibility transmitted through human feedback increases the inclination to produce responses that correspond to the users' beliefs or misleading prompts as opposed to true facts, a behaviour known as sycophancy. This phenomenon decreases the bias, robustness, and, consequently, their reliability. In this paper, we shed light on the suggestibility of Large Language Models (LLMs) to sycophantic behaviour, demonstrating these tendencies via human-influenced prompts over different tasks. Our investigation reveals that LLMs show sycophantic tendencies when responding to queries involving subjective opinions and statements that should elicit a contrary response based on facts. In contrast, when confronted with mathematical tasks or queries that have an objective answer, these models at various scales seem not to follow the users' hints by demonstrating confidence in delivering the correct answers."
                },
                {
                    "title": "Beyond Memorization: Violating Privacy Via Inference with Large Language Models",
                    "abstract": "Current privacy research on large language models (LLMs) primarily focuses on the issue of extracting memorized training data. At the same time, models' inference capabilities have increased drastically. This raises the key question of whether current LLMs could violate individuals' privacy by inferring personal attributes from text given at inference time. In this work, we present the first comprehensive study on the capabilities of pretrained LLMs to infer personal attributes from text. We construct a dataset consisting of real Reddit profiles, and show that current LLMs can infer a wide range of personal attributes (e.g., location, income, sex), achieving up to $85\\%$ top-1 and $95\\%$ top-3 accuracy at a fraction of the cost ($100\\times$) and time ($240\\times$) required by humans. As people increasingly interact with LLM-powered chatbots across all aspects of life, we also explore the emerging threat of privacy-invasive chatbots trying to extract personal information through seemingly benign questions. Finally, we show that common mitigations, i.e., text anonymization and model alignment, are currently ineffective at protecting user privacy against LLM inference. Our findings highlight that current LLMs can infer personal data at a previously unattainable scale. In the absence of working defenses, we advocate for a broader discussion around LLM privacy implications beyond memorization, striving for a wider privacy protection."
                },
                {
                    "title": "Qwen Technical Report",
                    "abstract": "Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural language processing tasks that were previously thought to be exclusive to humans. In this work, we introduce Qwen, the first installment of our large language model series. Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts. It includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques. The base language models consistently demonstrate superior performance across a multitude of downstream tasks, and the chat models, particularly those trained using Reinforcement Learning from Human Feedback (RLHF), are highly competitive. The chat models possess advanced tool-use and planning capabilities for creating agent applications, showcasing impressive performance even when compared to bigger models on complex tasks like utilizing a code interpreter. Furthermore, we have developed coding-specialized models, Code-Qwen and Code-Qwen-Chat, as well as mathematics-focused models, Math-Qwen-Chat, which are built upon base language models. These models demonstrate significantly improved performance in comparison with open-source models, and slightly fall behind the proprietary models."
                },
                {
                    "title": "LLM-powered Data Augmentation for Enhanced Crosslingual Performance",
                    "abstract": "This paper explores the potential of leveraging Large Language Models (LLMs) for data augmentation in multilingual commonsense reasoning datasets where the available training data is extremely limited. To achieve this, we utilise several LLMs, namely Dolly-v2, StableVicuna, ChatGPT, and GPT-4, to augment three datasets: XCOPA, XWinograd, and XStoryCloze. Subsequently, we evaluate the effectiveness of fine-tuning smaller multilingual models, mBERT and XLMR, using the synthesised data. We compare the performance of training with data generated in English and target languages, as well as translated English-generated data, revealing the overall advantages of incorporating data generated by LLMs, e.g. a notable 13.4 accuracy score improvement for the best case. Furthermore, we conduct a human evaluation by asking native speakers to assess the naturalness and logical coherence of the generated examples across different languages. The results of the evaluation indicate that LLMs such as ChatGPT and GPT-4 excel at producing natural and coherent text in most languages, however, they struggle to generate meaningful text in certain languages like Tamil. We also observe that ChatGPT falls short in generating plausible alternatives compared to the original dataset, whereas examples from GPT-4 exhibit competitive logical consistency."
                },
                {
                    "title": "WizardLM: Empowering Large Language Models to Follow Complex Instructions",
                    "abstract": "Training large language models (LLMs) with open-domain instruction following data brings colossal success. However, manually creating such instruction data is very time-consuming and labor-intensive. Moreover, humans may struggle to produce high-complexity instructions. In this paper, we show an avenue for creating large amounts of instruction data with varying levels of complexity using LLM instead of humans. Starting with an initial set of instructions, we use our proposed Evol-Instruct to rewrite them step by step into more complex instructions. Then, we mix all generated instruction data to fine-tune LLaMA. We call the resulting model WizardLM. Human evaluations on a complexity-balanced test bed and Vicuna's testset show that instructions from Evol-Instruct are superior to human-created ones. By analyzing the human evaluation results of the high complexity part, we demonstrate that outputs from our WizardLM are preferred to outputs from OpenAI ChatGPT. In GPT-4 automatic evaluation, WizardLM achieves more than 90\\% capacity of ChatGPT on 17 out of 29 skills. Even though WizardLM still lags behind ChatGPT in some aspects, our findings suggest that fine-tuning with AI-evolved instructions is a promising direction for enhancing LLMs. Our code and data are public at https://github.com/nlpxucan/WizardLM"
                },
                {
                    "title": "AI, write an essay for me: A large-scale comparison of human-written versus ChatGPT-generated essays",
                    "abstract": "Background: Recently, ChatGPT and similar generative AI models have attracted hundreds of millions of users and become part of the public discourse. Many believe that such models will disrupt society and will result in a significant change in the education system and information generation in the future. So far, this belief is based on either colloquial evidence or benchmarks from the owners of the models -- both lack scientific rigour. Objective: Through a large-scale study comparing human-written versus ChatGPT-generated argumentative student essays, we systematically assess the quality of the AI-generated content. Methods: A large corpus of essays was rated using standard criteria by a large number of human experts (teachers). We augment the analysis with a consideration of the linguistic characteristics of the generated essays. Results: Our results demonstrate that ChatGPT generates essays that are rated higher for quality than human-written essays. The writing style of the AI models exhibits linguistic characteristics that are different from those of the human-written essays, e.g., it is characterized by fewer discourse and epistemic markers, but more nominalizations and greater lexical diversity. Conclusions: Our results clearly demonstrate that models like ChatGPT outperform humans in generating argumentative essays. Since the technology is readily available for anyone to use, educators must act immediately. We must re-invent homework and develop teaching concepts that utilize these AI models in the same way as math utilized the calculator: teach the general concepts first and then use AI tools to free up time for other learning objectives."
                },
                {
                    "title": "Evaluating Large Language Models in Generating Synthetic HCI Research Data: a Case Study",
                    "abstract": "Collecting data is one of the bottlenecks of Human-Computer Interaction (HCI) research. Motivated by this, we explore the potential of large language models (LLMs) in generating synthetic user research data. We use OpenAI\u2019s GPT-3 model to generate open-ended questionnaire responses about experiencing video games as art, a topic not tractable with traditional computational user models. We test whether synthetic responses can be distinguished from real responses, analyze errors of synthetic data, and investigate content similarities between synthetic and real data. We conclude that GPT-3 can, in this context, yield believable accounts of HCI experiences. Given the low cost and high speed of LLM data generation, synthetic data should be useful in ideating and piloting new experiments, although any findings must obviously always be validated with real data. The results also raise concerns: if employed by malicious users of crowdsourcing services, LLMs may make crowdsourcing of self-report data fundamentally unreliable."
                },
                {
                    "title": "Whose Opinions Do Language Models Reflect?",
                    "abstract": "Language models (LMs) are increasingly being used in open-ended contexts, where the opinions reflected by LMs in response to subjective queries can have a profound impact, both on user satisfaction, as well as shaping the views of society at large. In this work, we put forth a quantitative framework to investigate the opinions reflected by LMs -- by leveraging high-quality public opinion polls and their associated human responses. Using this framework, we create OpinionsQA, a new dataset for evaluating the alignment of LM opinions with those of 60 US demographic groups over topics ranging from abortion to automation. Across topics, we find substantial misalignment between the views reflected by current LMs and those of US demographic groups: on par with the Democrat-Republican divide on climate change. Notably, this misalignment persists even after explicitly steering the LMs towards particular demographic groups. Our analysis not only confirms prior observations about the left-leaning tendencies of some human feedback-tuned LMs, but also surfaces groups whose opinions are poorly reflected by current LMs (e.g., 65+ and widowed individuals). Our code and data are available at https://github.com/tatsu-lab/opinions_qa."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Language Models are Realistic Tabular Data Generators",
                    "abstract": "Tabular data is among the oldest and most ubiquitous forms of data. However, the generation of synthetic samples with the original data's characteristics remains a significant challenge for tabular data. While many generative models from the computer vision domain, such as variational autoencoders or generative adversarial networks, have been adapted for tabular data generation, less research has been directed towards recent transformer-based large language models (LLMs), which are also generative in nature. To this end, we propose GReaT (Generation of Realistic Tabular data), which exploits an auto-regressive generative LLM to sample synthetic and yet highly realistic tabular data. Furthermore, GReaT can model tabular data distributions by conditioning on any subset of features; the remaining features are sampled without additional overhead. We demonstrate the effectiveness of the proposed approach in a series of experiments that quantify the validity and quality of the produced data samples from multiple angles. We find that GReaT maintains state-of-the-art performance across numerous real-world and synthetic data sets with heterogeneous feature types coming in various sizes."
                },
                {
                    "title": "Quantifying Memorization Across Neural Language Models",
                    "abstract": "Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Ethical and social risks of harm from Language Models",
                    "abstract": "This paper aims to help structure the risk landscape associated with large-scale Language Models (LMs). In order to foster advances in responsible innovation, an in-depth understanding of the potential risks posed by these models is needed. A wide range of established and anticipated risks are analysed in detail, drawing on multidisciplinary expertise and literature from computer science, linguistics, and social sciences. We outline six specific risk areas: I. Discrimination, Exclusion and Toxicity, II. Information Hazards, III. Misinformation Harms, V. Malicious Uses, V. Human-Computer Interaction Harms, VI. Automation, Access, and Environmental Harms. The first area concerns the perpetuation of stereotypes, unfair discrimination, exclusionary norms, toxic language, and lower performance by social group for LMs. The second focuses on risks from private data leaks or LMs correctly inferring sensitive information. The third addresses risks arising from poor, false or misleading information including in sensitive domains, and knock-on risks such as the erosion of trust in shared information. The fourth considers risks from actors who try to use LMs to cause harm. The fifth focuses on risks specific to LLMs used to underpin conversational agents that interact with human users, including unsafe use, manipulation or deception. The sixth discusses the risk of environmental harm, job automation, and other challenges that may have a disparate effect on different social groups or communities. In total, we review 21 risks in-depth. We discuss the points of origin of different risks and point to potential mitigation approaches. Lastly, we discuss organisational responsibilities in implementing mitigations, and the role of collaboration and participation. We highlight directions for further research, particularly on expanding the toolkit for assessing and evaluating the outlined risks in LMs."
                },
                {
                    "title": "GPT3Mix: Leveraging Large-scale Language Models for Text Augmentation",
                    "abstract": "Large-scale language models such as GPT-3 are excellent few-shot learners, allowing them to be controlled via natural text prompts. Recent studies report that prompt-based direct classification eliminates the need for fine-tuning but lacks data and inference scalability. This paper proposes a novel data augmentation technique that leverages large-scale language models to generate realistic text samples from a mixture of real samples. We also propose utilizing soft-labels predicted by the language models, effectively distilling knowledge from the large-scale language models and creating textual perturbations simultaneously. We perform data augmentation experiments on diverse classification tasks and show that our method hugely outperforms existing text augmentation methods. Ablation studies and a qualitative analysis provide more insights into our approach."
                },
                {
                    "title": "Generating Datasets with Pretrained Language Models",
                    "abstract": "To obtain high-quality sentence embeddings from pretrained language models (PLMs), they must either be augmented with additional pretraining objectives or finetuned on a large set of labeled text pairs. While the latter approach typically outperforms the former, it requires great human effort to generate suitable datasets of sufficient size. In this paper, we show how PLMs can be leveraged to obtain high-quality sentence embeddings without the need for labeled data, finetuning or modifications to the pretraining objective: We utilize the generative abilities of large and high-performing PLMs to generate entire datasets of labeled text pairs from scratch, which we then use for finetuning much smaller and more efficient models. Our fully unsupervised approach outperforms strong baselines on several semantic textual similarity datasets."
                },
                {
                    "title": "Generating Synthetic Text Data to Evaluate Causal Inference Methods",
                    "abstract": "Drawing causal conclusions from observational data requires making assumptions about the true data-generating process. Causal inference research typically considers low-dimensional data, such as categorical or numerical fields in structured medical records. High-dimensional and unstructured data such as natural language complicates the evaluation of causal inference methods; such evaluations rely on synthetic datasets with known causal effects. Models for natural language generation have been widely studied and perform well empirically. However, existing methods not immediately applicable to producing synthetic datasets for causal evaluations, as they do not allow for quantifying a causal effect on the text itself. In this work, we develop a framework for adapting existing generation models to produce synthetic text datasets with known causal effects. We use this framework to perform an empirical comparison of four recently-proposed methods for estimating causal effects from text data. We release our code and synthetic datasets."
                },
                {
                    "title": "Extracting Training Data from Large Language Models",
                    "abstract": "It has become common to publish large (billion parameter) language models that have been trained on private datasets. This paper demonstrates that in such settings, an adversary can perform a training data extraction attack to recover individual training examples by querying the language model. \nWe demonstrate our attack on GPT-2, a language model trained on scrapes of the public Internet, and are able to extract hundreds of verbatim text sequences from the model's training data. These extracted examples include (public) personally identifiable information (names, phone numbers, and email addresses), IRC conversations, code, and 128-bit UUIDs. Our attack is possible even though each of the above sequences are included in just one document in the training data. \nWe comprehensively evaluate our extraction attack to understand the factors that contribute to its success. For example, we find that larger models are more vulnerable than smaller models. We conclude by drawing lessons and discussing possible safeguards for training large language models."
                },
                {
                    "title": "Data Augmentation using Pre-trained Transformer Models",
                    "abstract": "Language model based pre-trained models such as BERT have provided significant gains across different NLP tasks. In this paper, we study different types of transformer based pre-trained models such as auto-regressive models (GPT-2), auto-encoder models (BERT), and seq2seq models (BART) for conditional data augmentation. We show that prepending the class labels to text sequences provides a simple yet effective way to condition the pre-trained models for data augmentation. Additionally, on three classification benchmarks, pre-trained Seq2Seq model outperforms other data augmentation methods in a low-resource setting. Further, we explore how different pre-trained model based data augmentation differs in-terms of data diversity, and how well such methods preserve the class-label information."
                },
                {
                    "title": "A4NT: Author Attribute Anonymity by Adversarial Training of Neural Machine Translation",
                    "abstract": "Text-based analysis methods allow to reveal privacy relevant author attributes such as gender, age and identify of the text's author. Such methods can compromise the privacy of an anonymous author even when the author tries to remove privacy sensitive content. In this paper, we propose an automatic method, called Adversarial Author Attribute Anonymity Neural Translation ($A^4NT$), to combat such text-based adversaries. We combine sequence-to-sequence language models used in machine translation and generative adversarial networks to obfuscate author attributes. Unlike machine translation techniques which need paired data, our method can be trained on unpaired corpora of text containing different authors. Importantly, we propose and evaluate techniques to impose constraints on our $A^4NT$ to preserve the semantics of the input text. $A^4NT$ learns to make minimal changes to the input text to successfully fool author attribute classifiers, while aiming to maintain the meaning of the input. We show through experiments on two different datasets and three settings that our proposed method is effective in fooling the author attribute classifiers and thereby improving the anonymity of authors."
                },
                {
                    "title": "Evaluating Reading Comprehension Exercises Generated by LLMs: A Showcase of ChatGPT in Education Applications",
                    "abstract": "The recent advancement of pre-trained Large Language Models (LLMs), such as OpenAI\u2019s ChatGPT, has led to transformative changes across fields. For example, developing intelligent systems in the educational sector that leverage the linguistic capabilities of LLMs demonstrates a visible potential. Though researchers have recently explored how ChatGPT could possibly assist in student learning, few studies have applied these techniques to real-world classroom settings involving teachers and students. In this study, we implement a reading comprehension exercise generation system that provides high-quality and personalized reading materials for middle school English learners in China. Extensive evaluations of the generated reading passages and corresponding exercise questions, conducted both automatically and manually, demonstrate that the system-generated materials are suitable for students and even surpass the quality of existing human-written ones. By incorporating first-hand feedback and suggestions from experienced educators, this study serves as a meaningful pioneering application of ChatGPT, shedding light on the future design and implementation of LLM-based systems in the educational context."
                },
                {
                    "title": "Adaptive Testing and Debugging of NLP Models",
                    "abstract": "Current approaches to testing and debugging NLP models rely on highly variable human creativity and extensive labor, or only work for a very restrictive class of bugs. We present AdaTest, a process which uses large scale language models (LMs) in partnership with human feedback to automatically write unit tests highlighting bugs in a target model. Such bugs are then addressed through an iterative text-fix-retest loop, inspired by traditional software development. In experiments with expert and non-expert users and commercial / research models for 8 different tasks, AdaTest makes users 5-10x more effective at finding bugs than current approaches, and helps users effectively fix bugs without adding new bugs."
                },
                {
                    "title": "Preventing Verbatim Memorization in Language Models Gives a False Sense of Privacy",
                    "abstract": "Studying data memorization in neural language models helps us understand the risks (e.g., to privacy or copyright) associated with models regurgitating training data, and aids in the evaluation of potential countermeasures. Many prior works\u2014and some recently deployed defenses\u2014focus on \u201cverbatim memo-rization\u201d, de\ufb01ned as a model generation that exactly matches a substring from the training set. We argue that verbatim memorization def-initions are too restrictive and fail to capture more subtle forms of memorization. Speci\ufb01cally, we design and implement an ef\ufb01cient defense based on Bloom \ufb01lters that perfectly prevents all verbatim memorization. And yet, we demonstrate that this \u201cperfect\u201d \ufb01lter does not prevent the leakage of training data. Indeed, it is easily circumvented by plausible and mini-mally modi\ufb01ed \u201cstyle-transfer\u201d prompts\u2014and in some cases even the non-modi\ufb01ed original prompts\u2014to extract memorized information. For example, instructing the model to output ALL-CAPITAL texts bypasses memorization checks based on verbatim matching. We conclude by discussing potential alternative de\ufb01nitions and why de\ufb01ning memorization is a dif\ufb01cult yet crucial open question for neural language models."
                },
                {
                    "title": "Overview of the 6th Author Profiling Task at PAN 2018: Multimodal Gender Identification in Twitter",
                    "abstract": "This overview presents the framework and the results of the Author Profiling shared task at PAN 2018. The objective of this year\u2019s task is to address gender identification from a multimodal perspective, where not only texts but also images are given. For this purpose a corpus with Twitter data has been provided, covering the languages Arabic, English, and Spanish. Altogether, the approaches of 23 participants are evaluated."
                },
                {
                    "title": "Text and Image Synergy with Feature Cross Technique for Gender Identification: Notebook for PAN at CLEF 2018",
                    "abstract": "This paper describes a neural network model for the author profiling task of PAN@CLEF 2018. Traditional machine learning models have shown superior performances for the author profiling task in past PAN series. However, these models often require careful feature-engineering to improve their performance. On the other hand, neural network approaches have recently shown advanced performances in both natural language processing (NLP) and computer vision (CV) tasks. We tackle the author profiling task using neural networks for texts and images. In order to leverage the synergy of the texts and images, we propose Text Image Fusion Neural Network (TIFNN), which considers their interaction. In an in-house experiment, TIFNN achieved accuracies of 84-90% for different languages when used for gender identification."
                },
                {
                    "title": "Gender Identification in Twitter using N-grams and LSA: Notebook for PAN at CLEF 2018",
                    "abstract": "In this paper, we describe the participation of the Natural Language Processing Lab of the University of Ottawa in the author profiling shared task at PAN 2018. We present our approach to gender identification in Twitter performed on the tweet corpus provided by CLEF for the task. Our approach takes advantage of textual information solely, and consists of tweet preprocessing, feature construction, dimensionality reduction using Latent Semantic Analysis (LSA), and classification model construction. We propose a linear Support Vector Machine (SVM) classifier, with different types of word and character n-grams as features. Our model was the best-performing model in textual classification, with the accuracy of 0.8221, 0.82, and 0.809 on the English, Spanish, and Arabic datasets respectively. Considering the combination of textual and image classification, and all three datasets, our model ranked second in the task."
                },
                {
                    "title": "Overview of the 5th Author Profiling Task at PAN 2017: Gender and Language Variety Identification in Twitter",
                    "abstract": "This overview presents the framework and the results of the Author Pro\ufb01ling task at PAN 2017. The objective of this year is to address gender and language variety identi\ufb01cation. For this purpose a corpus from Twitter has been provided for four different languages: Arabic, English, Portuguese, and Spanish. Altogether, the approaches of 22 participants are evaluated."
                },
                {
                    "title": "A Spanish text corpus for the author profiling task",
                    "abstract": "Author Profiling is the task of predicting characteristics of the author of a text, such as age, gender, personality, native language, etc. This is a task of growing importance due to its potential applications in security, crime and marketing, among others. One of the main difficulties in this field is the lack of reliable text collections (corpora) to train and test automatically derived classifiers, in particular in specific languages such as Spanish. Although some recent data sets were generated for the PAN competitions, these documents have a lot of \u201cnoise\u201d that prevent researchers from obtaining more general conclusions about this task when more formal documents are used. In this context, this work proposes and describes SpanText, a data collection of formal texts in Spanish language which is, as far as we know, the first collection with these characteristics for the author profiling task. Besides, an experimental study is carried out where the difference in performance obtained with formal and informal texts is clearly established and opens interesting research lines to get a deeper understanding of the particularities that each type of documents poses to the author profiling task."
                },
                {
                    "title": "Overview of the Author Profiling Task at PAN 2013",
                    "abstract": "This overview presents the framework and results for the Author Pro- filing task at PAN 2013. We describe in detail the corpus and its characteristics, and the evaluation framework we used to measure the participants performance to solve the problem of identifying age and gender from anonymous texts. Finally, the approaches of the 21 participants and their results are described."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we create a privacy-preserving synthetic dataset that accurately simulates personal attribute inferences from comment threads on social media platforms?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the significant gap in available datasets for studying personal attribute inferences (PAIs) from online texts. By providing a realistic and diverse synthetic dataset, SynthPAI can facilitate further research into the implications of LLMs on privacy and personal data, ultimately leading to better understanding and mitigation of potential misuse. This work could advance knowledge in the fields of machine learning, data privacy, and ethical AI, while also enabling practical applications in developing safer LLMs and informing policy decisions regarding data protection.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to create a dataset that is both realistic and representative of real-world interactions while adhering to strict data protection regulations. Naive approaches may fail because they might not capture the nuanced and diverse nature of human interactions or could inadvertently expose sensitive information. Additionally, the technical complexity of simulating realistic comment threads and accurately labeling personal attributes without real data presents significant obstacles that must be overcome.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the lack of openly available datasets due to legal and ethical concerns surrounding personal data. While some studies have attempted to create datasets, they often relied on real-world data that could not be shared publicly. Existing synthetic datasets have not adequately captured the complexity and realism of actual comment threads. Our approach differs by using a novel framework that generates synthetic profiles and simulates interactions without requiring real data, thus overcoming the barriers that have previously hindered PAI research.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves creating personalized LLM agents that simulate interactions on comment-thread-focused social media platforms. We will generate a synthetic dataset, SynthPAI, consisting of over 7,800 comments with hand-curated personal attribute labels. The evaluation will focus on the realism and diversity of the dataset, using metrics such as inter-rater agreement among human reviewers to assess the perceived authenticity of the comments. We expect that SynthPAI will provide a valuable resource for PAI research, enabling more comprehensive studies while ensuring privacy preservation."
            }
        },
        "author_data": {
            "bb44608e-58c5-4f42-abf4-9fccced297b9": {
                "pk": "bb44608e-58c5-4f42-abf4-9fccced297b9",
                "name": "Hanna Yukhymenko",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "2b75ae50-6add-4dc8-8630-02769c269b36": {
                "pk": "2b75ae50-6add-4dc8-8630-02769c269b36",
                "name": "Robin Staab",
                "collaborators": [
                    "Martin Vechev",
                    "Nikola Jovanovi\u0107",
                    "Mark Vero",
                    "Mislav Balunovi\u0107",
                    "Thibaud Gloaguen",
                    "Maximilian Baader",
                    "Batuhan T\u00f6mek\u00e7e",
                    "Mark Niklas M\u00fcller",
                    "Marc Fischer",
                    "Dimitar I. Dimitrov"
                ],
                "domain": [
                    "Privacy",
                    "Large Language Models",
                    "Anonymization",
                    "Fairness"
                ],
                "publications": [
                    {
                        "title": "Large Language Models are Advanced Anonymizers",
                        "abstract": "Recent work in privacy research on large language models has shown that they achieve near human-level performance at inferring personal data from real-world online texts. With consistently increasing model capabilities, existing text anonymization methods are currently lacking behind regulatory requirements and adversarial threats. This raises the question of how individuals can effectively protect their personal data in sharing online texts. In this work, we take two steps to answer this question: We first present a new setting for evaluating anonymizations in the face of adversarial LLMs inferences, allowing for a natural measurement of anonymization performance while remedying some of the shortcomings of previous metrics. We then present our LLM-based adversarial anonymization framework leveraging the strong inferential capabilities of LLMs to inform our anonymization procedure. In our experimental evaluation, we show on real-world and synthetic online texts how adversarial anonymization outperforms current industry-grade anonymizers both in terms of the resulting utility and privacy."
                    },
                    {
                        "title": "Beyond Memorization: Violating Privacy Via Inference with Large Language Models",
                        "abstract": "Current privacy research on large language models (LLMs) primarily focuses on the issue of extracting memorized training data. At the same time, models' inference capabilities have increased drastically. This raises the key question of whether current LLMs could violate individuals' privacy by inferring personal attributes from text given at inference time. In this work, we present the first comprehensive study on the capabilities of pretrained LLMs to infer personal attributes from text. We construct a dataset consisting of real Reddit profiles, and show that current LLMs can infer a wide range of personal attributes (e.g., location, income, sex), achieving up to $85\\%$ top-1 and $95\\%$ top-3 accuracy at a fraction of the cost ($100\\times$) and time ($240\\times$) required by humans. As people increasingly interact with LLM-powered chatbots across all aspects of life, we also explore the emerging threat of privacy-invasive chatbots trying to extract personal information through seemingly benign questions. Finally, we show that common mitigations, i.e., text anonymization and model alignment, are currently ineffective at protecting user privacy against LLM inference. Our findings highlight that current LLMs can infer personal data at a previously unattainable scale. In the absence of working defenses, we advocate for a broader discussion around LLM privacy implications beyond memorization, striving for a wider privacy protection."
                    },
                    {
                        "title": "Watermark Stealing in Large Language Models",
                        "abstract": "LLM watermarking has attracted attention as a promising way to detect AI-generated content, with some works suggesting that current schemes may already be fit for deployment. In this work we dispute this claim, identifying watermark stealing (WS) as a fundamental vulnerability of these schemes. We show that querying the API of the watermarked LLM to approximately reverse-engineer a watermark enables practical spoofing attacks, as hypothesized in prior work, but also greatly boosts scrubbing attacks, which was previously unnoticed. We are the first to propose an automated WS algorithm and use it in the first comprehensive study of spoofing and scrubbing in realistic settings. We show that for under $50 an attacker can both spoof and scrub state-of-the-art schemes previously considered safe, with average success rate of over 80%. Our findings challenge common beliefs about LLM watermarking, stressing the need for more robust schemes. We make all our code and additional examples available at https://watermark-stealing.org."
                    },
                    {
                        "title": "Discovering Clues of Spoofed LM Watermarks",
                        "abstract": "LLM watermarks stand out as a promising way to attribute ownership of LLM-generated text. One threat to watermark credibility comes from spoofing attacks, where an unauthorized third party forges the watermark, enabling it to falsely attribute arbitrary texts to a particular LLM. While recent works have demonstrated that state-of-the-art schemes are in fact vulnerable to spoofing, they lack deeper qualitative analysis of the texts produced by spoofing methods. In this work, we for the first time reveal that there are observable differences between genuine and spoofed watermark texts. Namely, we show that regardless of their underlying approach, all current spoofing methods consistently leave observable artifacts in spoofed texts, indicative of watermark forgery. We build upon these findings to propose rigorous statistical tests that reliably reveal the presence of such artifacts, effectively discovering that a watermark was spoofed. Our experimental evaluation shows high test power across all current spoofing methods, providing insights into their fundamental limitations, and suggesting a way to mitigate this threat."
                    },
                    {
                        "title": "Ward: Provable RAG Dataset Inference via LLM Watermarks",
                        "abstract": "Retrieval-Augmented Generation (RAG) improves LLMs by enabling them to incorporate external data during generation. This raises concerns for data owners regarding unauthorized use of their content in RAG systems. Despite its importance, the challenge of detecting such unauthorized usage remains underexplored, with existing datasets and methodologies from adjacent fields being ill-suited for its study. In this work, we take several steps to bridge this gap. First, we formalize this problem as (black-box) RAG Dataset Inference (RAG-DI). To facilitate research on this challenge, we further introduce a novel dataset specifically designed for benchmarking RAG-DI methods under realistic conditions, and propose a set of baseline approaches. Building on this foundation, we introduce Ward, a RAG-DI method based on LLM watermarks that enables data owners to obtain rigorous statistical guarantees regarding the usage of their dataset in a RAG system. In our experimental evaluation, we show that Ward consistently outperforms all baselines across many challenging settings, achieving higher accuracy, superior query efficiency and robustness. Our work provides a foundation for future studies of RAG-DI and highlights LLM watermarks as a promising approach to this problem."
                    },
                    {
                        "title": "From Principle to Practice: Vertical Data Minimization for Machine Learning",
                        "abstract": "Aiming to train and deploy predictive models, organizations collect large amounts of detailed client data, risking the exposure of private information in the event of a breach. To mitigate this, policymakers increasingly demand compliance with the data minimization (DM) principle, restricting data collection to only that data which is relevant and necessary for the task. Despite regulatory pressure, the problem of deploying machine learning models that obey DM has so far received little attention. In this work, we address this challenge in a comprehensive manner. We propose a novel vertical DM (vDM) workflow based on data generalization, which by design ensures that no full-resolution client data is collected during training and deployment of models, benefiting client privacy by reducing the attack surface in case of a breach. We formalize and study the corresponding problem of finding generalizations that both maximize data utility and minimize empirical privacy risk, which we quantify by introducing a diverse set of policy-aligned adversarial scenarios. Finally, we propose a range of baseline vDM algorithms, as well as Privacy-aware Tree (PAT), an especially effective vDM algorithm that outperforms all baselines across several settings. We plan to release our code as a publicly available library, helping advance the standardization of DM for machine learning. Overall, we believe our work can help lay the foundation for further exploration and adoption of DM principles in real-world applications."
                    },
                    {
                        "title": "Private Attribute Inference from Images with Vision-Language Models",
                        "abstract": "As large language models (LLMs) become ubiquitous in our daily tasks and digital interactions, associated privacy risks are increasingly in focus. While LLM privacy research has primarily focused on the leakage of model training data, it has recently been shown that the increase in models' capabilities has enabled LLMs to make accurate privacy-infringing inferences from previously unseen texts. With the rise of multimodal vision-language models (VLMs), capable of understanding both images and text, a pertinent question is whether such results transfer to the previously unexplored domain of benign images posted online. To investigate the risks associated with the image reasoning capabilities of newly emerging VLMs, we compile an image dataset with human-annotated labels of the image owner's personal attributes. In order to understand the additional privacy risk posed by VLMs beyond traditional human attribute recognition, our dataset consists of images where the inferable private attributes do not stem from direct depictions of humans. On this dataset, we evaluate the inferential capabilities of 7 state-of-the-art VLMs, finding that they can infer various personal attributes at up to 77.6% accuracy. Concerningly, we observe that accuracy scales with the general capabilities of the models, implying that future models can be misused as stronger adversaries, establishing an imperative for the development of adequate defenses."
                    },
                    {
                        "title": "Black-Box Detection of Language Model Watermarks",
                        "abstract": "Watermarking has emerged as a promising way to detect LLM-generated text. To apply a watermark an LLM provider, given a secret key, augments generations with a signal that is later detectable by any party with the same key. Recent work has proposed three main families of watermarking schemes, two of which focus on the property of preserving the LLM distribution. This is motivated by it being a tractable proxy for maintaining LLM capabilities, but also by the idea that concealing a watermark deployment makes it harder for malicious actors to hide misuse by avoiding a certain LLM or attacking its watermark. Yet, despite much discourse around detectability, no prior work has investigated if any of these scheme families are detectable in a realistic black-box setting. We tackle this for the first time, developing rigorous statistical tests to detect the presence of all three most popular watermarking scheme families using only a limited number of black-box queries. We experimentally confirm the effectiveness of our methods on a range of schemes and a diverse set of open-source models. Our findings indicate that current watermarking schemes are more detectable than previously believed, and that obscuring the fact that a watermark was deployed may not be a viable way for providers to protect against adversaries. We further apply our methods to test for watermark presence behind the most popular public APIs: GPT4, Claude 3, Gemini 1.0 Pro, finding no strong evidence of a watermark at this point in time."
                    },
                    {
                        "title": "Abstract Interpretation of Fixpoint Iterators with Applications to Neural Networks",
                        "abstract": "We present a new abstract interpretation framework for the precise over-approximation of numerical fixpoint iterators. Our key observation is that unlike in standard abstract interpretation (AI), typically used to over-approximate all reachable program states, in this setting, one only needs to abstract the concrete fixpoints, i.e., the final program states. Our framework targets numerical fixpoint iterators with convergence and uniqueness guarantees in the concrete and is based on two major technical contributions: (i) theoretical insights which allow us to compute sound and precise fixpoint abstractions without using joins, and (ii) a new abstract domain, CH-Zonotope, which admits efficient propagation and inclusion checks while retaining high precision. We implement our framework in a tool called CRAFT and evaluate it on a novel fixpoint-based neural network architecture (monDEQ) that is particularly challenging to verify. Our extensive evaluation demonstrates that CRAFT exceeds the state-of-the-art performance in terms of speed (two orders of magnitude), scalability (one order of magnitude), and precision (25% higher certified accuracies)."
                    },
                    {
                        "title": "Bayesian Framework for Gradient Leakage",
                        "abstract": "Federated learning is an established method for training machine learning models without sharing training data. However, recent work has shown that it cannot guarantee data privacy as shared gradients can still leak sensitive information. To formalize the problem of gradient leakage, we propose a theoretical framework that enables, for the first time, analysis of the Bayes optimal adversary phrased as an optimization problem. We demonstrate that existing leakage attacks can be seen as approximations of this optimal adversary with different assumptions on the probability distributions of the input data and gradients. Our experiments confirm the effectiveness of the Bayes optimal adversary when it has knowledge of the underlying distribution. Further, our experimental evaluation shows that several existing heuristic defenses are not effective against stronger attacks, especially early in the training process. Thus, our findings indicate that the construction of more effective defenses and their evaluation remains an open problem."
                    },
                    {
                        "title": "Exploiting LLM Quantization",
                        "abstract": "Quantization leverages lower-precision weights to reduce the memory usage of large language models (LLMs) and is a key technique for enabling their deployment on commodity hardware. While LLM quantization's impact on utility has been extensively explored, this work for the first time studies its adverse effects from a security perspective. We reveal that widely used quantization methods can be exploited to produce a harmful quantized LLM, even though the full-precision counterpart appears benign, potentially tricking users into deploying the malicious quantized model. We demonstrate this threat using a three-staged attack framework: (i) first, we obtain a malicious LLM through fine-tuning on an adversarial task; (ii) next, we quantize the malicious model and calculate constraints that characterize all full-precision models that map to the same quantized model; (iii) finally, using projected gradient descent, we tune out the poisoned behavior from the full-precision model while ensuring that its weights satisfy the constraints computed in step (ii). This procedure results in an LLM that exhibits benign behavior in full precision but when quantized, it follows the adversarial behavior injected in step (i). We experimentally demonstrate the feasibility and severity of such an attack across three diverse scenarios: vulnerable code generation, content injection, and over-refusal attack. In practice, the adversary could host the resulting full-precision model on an LLM community hub such as Hugging Face, exposing millions of users to the threat of deploying its malicious quantized version on their devices."
                    },
                    {
                        "title": "Back to the Drawing Board for Fair Representation Learning",
                        "abstract": "The goal of Fair Representation Learning (FRL) is to mitigate biases in machine learning models by learning data representations that enable high accuracy on downstream tasks while minimizing discrimination based on sensitive attributes. The evaluation of FRL methods in many recent works primarily focuses on the tradeoff between downstream fairness and accuracy with respect to a single task that was used to approximate the utility of representations during training (proxy task). This incentivizes retaining only features relevant to the proxy task while discarding all other information. In extreme cases, this can cause the learned representations to collapse to a trivial, binary value, rendering them unusable in transfer settings. In this work, we argue that this approach is fundamentally mismatched with the original motivation of FRL, which arises from settings with many downstream tasks unknown at training time (transfer tasks). To remedy this, we propose to refocus the evaluation protocol of FRL methods primarily around the performance on transfer tasks. A key challenge when conducting such an evaluation is the lack of adequate benchmarks. We address this by formulating four criteria that a suitable evaluation procedure should fulfill. Based on these, we propose TransFair, a benchmark that satisfies these criteria, consisting of novel variations of popular FRL datasets with carefully calibrated transfer tasks. In this setting, we reevaluate state-of-the-art FRL methods, observing that they often overfit to the proxy task, which causes them to underperform on certain transfer tasks. We further highlight the importance of task-agnostic learning signals for FRL methods, as they can lead to more transferrable representations."
                    }
                ]
            },
            "dbb95783-0376-4ef3-862a-ab288ad01cf0": {
                "pk": "dbb95783-0376-4ef3-862a-ab288ad01cf0",
                "name": "Mark Vero",
                "collaborators": [
                    "Martin Vechev",
                    "Robin Staab",
                    "Mislav Balunovi\u0107",
                    "Jingxuan He",
                    "Nikola Jovanovi\u0107",
                    "Batuhan T\u00f6mek\u00e7e",
                    "Dimitar I. Dimitrov",
                    "Kazuki Egashira",
                    "Ang\u00e9line Pouget",
                    "Gabriela Krasnopolska"
                ],
                "domain": [
                    "Privacy",
                    "Synthetic Data",
                    "Machine Learning",
                    "Security"
                ],
                "publications": [
                    {
                        "title": "CuTS: Customizable Tabular Synthetic Data Generation",
                        "abstract": "Privacy, data quality, and data sharing concerns pose a key limitation for tabular data applications. While generating synthetic data resembling the original distribution addresses some of these issues, most applications would benefit from additional customization on the generated data. However, existing synthetic data approaches are limited to particular constraints, e.g., differential privacy (DP) or fairness. In this work, we introduce CuTS, the first customizable synthetic tabular data generation framework. Customization in CuTS is achieved via declarative statistical and logical expressions, supporting a wide range of requirements (e.g., DP or fairness, among others). To ensure high synthetic data quality in the presence of custom specifications, CuTS is pre-trained on the original dataset and fine-tuned on a differentiable loss automatically derived from the provided specifications using novel relaxations. We evaluate CuTS over four datasets and on numerous custom specifications, outperforming state-of-the-art specialized approaches on several tasks while being more general. In particular, at the same fairness level, we achieve 2.3% higher downstream accuracy than the state-of-the-art in fair synthetic data generation on the Adult dataset."
                    },
                    {
                        "title": "Private Attribute Inference from Images with Vision-Language Models",
                        "abstract": "As large language models (LLMs) become ubiquitous in our daily tasks and digital interactions, associated privacy risks are increasingly in focus. While LLM privacy research has primarily focused on the leakage of model training data, it has recently been shown that the increase in models' capabilities has enabled LLMs to make accurate privacy-infringing inferences from previously unseen texts. With the rise of multimodal vision-language models (VLMs), capable of understanding both images and text, a pertinent question is whether such results transfer to the previously unexplored domain of benign images posted online. To investigate the risks associated with the image reasoning capabilities of newly emerging VLMs, we compile an image dataset with human-annotated labels of the image owner's personal attributes. In order to understand the additional privacy risk posed by VLMs beyond traditional human attribute recognition, our dataset consists of images where the inferable private attributes do not stem from direct depictions of humans. On this dataset, we evaluate the inferential capabilities of 7 state-of-the-art VLMs, finding that they can infer various personal attributes at up to 77.6% accuracy. Concerningly, we observe that accuracy scales with the general capabilities of the models, implying that future models can be misused as stronger adversaries, establishing an imperative for the development of adequate defenses."
                    },
                    {
                        "title": "TabLeak: Tabular Data Leakage in Federated Learning",
                        "abstract": "While federated learning (FL) promises to preserve privacy, recent works in the image and text domains have shown that training updates leak private client data. However, most high-stakes applications of FL (e.g., in healthcare and finance) use tabular data, where the risk of data leakage has not yet been explored. A successful attack for tabular data must address two key challenges unique to the domain: (i) obtaining a solution to a high-variance mixed discrete-continuous optimization problem, and (ii) enabling human assessment of the reconstruction as unlike for image and text data, direct human inspection is not possible. In this work we address these challenges and propose TabLeak, the first comprehensive reconstruction attack on tabular data. TabLeak is based on two key contributions: (i) a method which leverages a softmax relaxation and pooled ensembling to solve the optimization problem, and (ii) an entropy-based uncertainty quantification scheme to enable human assessment. We evaluate TabLeak on four tabular datasets for both FedSGD and FedAvg training protocols, and show that it successfully breaks several settings previously deemed safe. For instance, we extract large subsets of private data at >90% accuracy even at the large batch size of 128. Our findings demonstrate that current high-stakes tabular FL is excessively vulnerable to leakage attacks."
                    },
                    {
                        "title": "Exploiting LLM Quantization",
                        "abstract": "Quantization leverages lower-precision weights to reduce the memory usage of large language models (LLMs) and is a key technique for enabling their deployment on commodity hardware. While LLM quantization's impact on utility has been extensively explored, this work for the first time studies its adverse effects from a security perspective. We reveal that widely used quantization methods can be exploited to produce a harmful quantized LLM, even though the full-precision counterpart appears benign, potentially tricking users into deploying the malicious quantized model. We demonstrate this threat using a three-staged attack framework: (i) first, we obtain a malicious LLM through fine-tuning on an adversarial task; (ii) next, we quantize the malicious model and calculate constraints that characterize all full-precision models that map to the same quantized model; (iii) finally, using projected gradient descent, we tune out the poisoned behavior from the full-precision model while ensuring that its weights satisfy the constraints computed in step (ii). This procedure results in an LLM that exhibits benign behavior in full precision but when quantized, it follows the adversarial behavior injected in step (i). We experimentally demonstrate the feasibility and severity of such an attack across three diverse scenarios: vulnerable code generation, content injection, and over-refusal attack. In practice, the adversary could host the resulting full-precision model on an LLM community hub such as Hugging Face, exposing millions of users to the threat of deploying its malicious quantized version on their devices."
                    },
                    {
                        "title": "Back to the Drawing Board for Fair Representation Learning",
                        "abstract": "The goal of Fair Representation Learning (FRL) is to mitigate biases in machine learning models by learning data representations that enable high accuracy on downstream tasks while minimizing discrimination based on sensitive attributes. The evaluation of FRL methods in many recent works primarily focuses on the tradeoff between downstream fairness and accuracy with respect to a single task that was used to approximate the utility of representations during training (proxy task). This incentivizes retaining only features relevant to the proxy task while discarding all other information. In extreme cases, this can cause the learned representations to collapse to a trivial, binary value, rendering them unusable in transfer settings. In this work, we argue that this approach is fundamentally mismatched with the original motivation of FRL, which arises from settings with many downstream tasks unknown at training time (transfer tasks). To remedy this, we propose to refocus the evaluation protocol of FRL methods primarily around the performance on transfer tasks. A key challenge when conducting such an evaluation is the lack of adequate benchmarks. We address this by formulating four criteria that a suitable evaluation procedure should fulfill. Based on these, we propose TransFair, a benchmark that satisfies these criteria, consisting of novel variations of popular FRL datasets with carefully calibrated transfer tasks. In this setting, we reevaluate state-of-the-art FRL methods, observing that they often overfit to the proxy task, which causes them to underperform on certain transfer tasks. We further highlight the importance of task-agnostic learning signals for FRL methods, as they can lead to more transferrable representations."
                    },
                    {
                        "title": "Instruction Tuning for Secure Code Generation",
                        "abstract": "Modern language models (LMs) have gained widespread acceptance in everyday and professional contexts, particularly in programming. An essential procedure enabling this adoption is instruction tuning, which substantially enhances LMs' practical utility by training them to follow user instructions and human preferences. However, existing instruction tuning schemes overlook a crucial aspect: the security of generated code. As a result, even the state-of-the-art instruction-tuned LMs frequently produce unsafe code, posing significant security risks. In this work, we introduce SafeCoder to address this gap. SafeCoder performs security-centric fine-tuning using a diverse and high-quality dataset that we collected using an automated pipeline. We integrate the security fine-tuning with standard instruction tuning, to facilitate a joint optimization of both security and utility. Despite its simplicity, we show that SafeCoder is effective across a variety of popular LMs and datasets. It is able to drastically improve security (by about 30%), while preserving utility."
                    },
                    {
                        "title": "Large Language Models are Advanced Anonymizers",
                        "abstract": "Recent work in privacy research on large language models has shown that they achieve near human-level performance at inferring personal data from real-world online texts. With consistently increasing model capabilities, existing text anonymization methods are currently lacking behind regulatory requirements and adversarial threats. This raises the question of how individuals can effectively protect their personal data in sharing online texts. In this work, we take two steps to answer this question: We first present a new setting for evaluating anonymizations in the face of adversarial LLMs inferences, allowing for a natural measurement of anonymization performance while remedying some of the shortcomings of previous metrics. We then present our LLM-based adversarial anonymization framework leveraging the strong inferential capabilities of LLMs to inform our anonymization procedure. In our experimental evaluation, we show on real-world and synthetic online texts how adversarial anonymization outperforms current industry-grade anonymizers both in terms of the resulting utility and privacy."
                    },
                    {
                        "title": "Beyond Memorization: Violating Privacy Via Inference with Large Language Models",
                        "abstract": "Current privacy research on large language models (LLMs) primarily focuses on the issue of extracting memorized training data. At the same time, models' inference capabilities have increased drastically. This raises the key question of whether current LLMs could violate individuals' privacy by inferring personal attributes from text given at inference time. In this work, we present the first comprehensive study on the capabilities of pretrained LLMs to infer personal attributes from text. We construct a dataset consisting of real Reddit profiles, and show that current LLMs can infer a wide range of personal attributes (e.g., location, income, sex), achieving up to $85\\%$ top-1 and $95\\%$ top-3 accuracy at a fraction of the cost ($100\\times$) and time ($240\\times$) required by humans. As people increasingly interact with LLM-powered chatbots across all aspects of life, we also explore the emerging threat of privacy-invasive chatbots trying to extract personal information through seemingly benign questions. Finally, we show that common mitigations, i.e., text anonymization and model alignment, are currently ineffective at protecting user privacy against LLM inference. Our findings highlight that current LLMs can infer personal data at a previously unattainable scale. In the absence of working defenses, we advocate for a broader discussion around LLM privacy implications beyond memorization, striving for a wider privacy protection."
                    },
                    {
                        "title": "COMPL-AI Framework: A Technical Interpretation and LLM Benchmarking Suite for the EU Artificial Intelligence Act",
                        "abstract": "The EU's Artificial Intelligence Act (AI Act) is a significant step towards responsible AI development, but lacks clear technical interpretation, making it difficult to assess models' compliance. This work presents COMPL-AI, a comprehensive framework consisting of (i) the first technical interpretation of the EU AI Act, translating its broad regulatory requirements into measurable technical requirements, with the focus on large language models (LLMs), and (ii) an open-source Act-centered benchmarking suite, based on thorough surveying and implementation of state-of-the-art LLM benchmarks. By evaluating 12 prominent LLMs in the context of COMPL-AI, we reveal shortcomings in existing models and benchmarks, particularly in areas like robustness, safety, diversity, and fairness. This work highlights the need for a shift in focus towards these aspects, encouraging balanced development of LLMs and more comprehensive regulation-aligned benchmarks. Simultaneously, COMPL-AI for the first time demonstrates the possibilities and difficulties of bringing the Act's obligations to a more concrete, technical level. As such, our work can serve as a useful first step towards having actionable recommendations for model providers, and contributes to ongoing efforts of the EU to enable application of the Act, such as the drafting of the GPAI Code of Practice."
                    },
                    {
                        "title": "Reducing Neural Architecture Search Spaces with Training-Free Statistics and Computational Graph Clustering",
                        "abstract": "The computational demands of neural architecture search (NAS) algorithms are usually directly proportional to the size of their target search spaces. Thus, limiting the search to high-quality subsets can greatly reduce the computational load of NAS algorithms. In this paper, we present Clustering-Based REDuction (C-BRED), a new technique to reduce the size of NAS search spaces. C-BRED reduces a NAS space by clustering the computational graphs associated with its architectures and selecting the most promising cluster using proxy statistics correlated with network accuracy. When considering the NAS-Bench-201 (NB201) data set and the CIFAR-100 task, C-BRED selects a subset with 70% average accuracy instead of the whole space's 64% average accuracy."
                    },
                    {
                        "title": "Practical Attacks against Black-box Code Completion Engines",
                        "abstract": "Modern code completion engines, powered by large language models, have demonstrated impressive capabilities to generate functionally correct code based on surrounding context. As these tools are extensively used by millions of developers, it is crucial to investigate their security implications. In this work, we present INSEC, a novel attack that directs code completion engines towards generating vulnerable code. In line with most commercial completion engines, such as GitHub Copilot, INSEC assumes only black-box query access to the targeted engine, without requiring any knowledge of the engine's internals. Our attack works by inserting a malicious attack string as a short comment in the completion input. To derive the attack string, we design a series of specialized initialization schemes and an optimization procedure for further refinement. We demonstrate the strength of INSEC not only on state-of-the-art open-source models but also on black-box commercial services such as the OpenAI API and GitHub Copilot. On a comprehensive set of security-critical test cases covering 16 CWEs across 5 programming languages, INSEC significantly increases the likelihood of the considered completion engines in generating unsafe code by >50% in absolute, while maintaining the ability in producing functionally correct code. At the same time, our attack has low resource requirements, and can be developed for a cost of well under ten USD on commodity hardware."
                    }
                ]
            },
            "30b1ddb0-e72b-4d78-99e4-7c0842542b37": {
                "pk": "30b1ddb0-e72b-4d78-99e4-7c0842542b37",
                "name": "Martin Vechev",
                "collaborators": [
                    "Maximilian Baader",
                    "Matthew Mirman",
                    "Marc Fischer",
                    "Pavol Bielik",
                    "Mislav Balunovi\u0107",
                    "Gagandeep Singh",
                    "Nikola Jovanovi\u0107",
                    "Anian Ruoss",
                    "Anouk Paradis",
                    "Benjamin Bichsel"
                ],
                "domain": [
                    "Adversarial Robustness",
                    "Neural Networks",
                    "Privacy-Preserving Machine Learning",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "A Provable Defense for Deep Residual Networks",
                        "abstract": "We present a training system, which can provably defend significantly larger neural networks than previously possible, including ResNet-34 and DenseNet-100. Our approach is based on differentiable abstract interpretation and introduces two novel concepts: (i) abstract layers for fine-tuning the precision and scalability of the abstraction, (ii) a flexible domain specific language (DSL) for describing training objectives that combine abstract and concrete losses with arbitrary specifications. Our training method is implemented in the DiffAI system."
                    },
                    {
                        "title": "Online Robustness Training for Deep Reinforcement Learning",
                        "abstract": "In deep reinforcement learning (RL), adversarial attacks can trick an agent into unwanted states and disrupt training. We propose a system called Robust Student-DQN (RS-DQN), which permits online robustness training alongside Q networks, while preserving competitive performance. We show that RS-DQN can be combined with (i) state-of-the-art adversarial training and (ii) provably robust training to obtain an agent that is resilient to strong attacks during training and evaluation."
                    },
                    {
                        "title": "Adversarial Robustness for Code",
                        "abstract": "Machine learning and deep learning in particular has been recently used to successfully address many tasks in the domain of code such as finding and fixing bugs, code completion, decompilation, type inference and many others. However, the issue of adversarial robustness of models for code has gone largely unnoticed. In this work, we explore this issue by: (i) instantiating adversarial attacks for code (a domain with discrete and highly structured inputs), (ii) showing that, similar to other domains, neural models for code are vulnerable to adversarial attacks, and (iii) combining existing and novel techniques to improve robustness while preserving high accuracy."
                    },
                    {
                        "title": "Large Language Models for Code: Security Hardening and Adversarial Testing",
                        "abstract": "Large language models (large LMs) are increasingly trained on massive codebases and used to generate code. However, LMs lack awareness of security and are found to frequently produce unsafe code. This work studies the security of LMs along two important axes: (i) security hardening, which aims to enhance LMs' reliability in generating secure code, and (ii) adversarial testing, which seeks to evaluate LMs' security at an adversarial standpoint. We address both of these by formulating a new security task called controlled code generation. The task is parametric and takes as input a binary property to guide the LM to generate secure or unsafe code, while preserving the LM's capability of generating functionally correct code. We propose a novel learning-based approach called SVEN to solve this task. SVEN leverages property-specific continuous vectors to guide program generation towards the given property, without modifying the LM's weights. Our training procedure optimizes these continuous vectors by enforcing specialized loss terms on different regions of code, using a high-quality dataset carefully curated by us. Our extensive evaluation shows that SVEN is highly effective in achieving strong security control. For instance, a state-of-the-art CodeGen LM with 2.7B parameters generates secure code for 59.1% of the time. When we employ SVEN to perform security hardening (or adversarial testing) on this LM, the ratio is significantly boosted to 92.3% (or degraded to 36.8%). Importantly, SVEN closely matches the original LMs in functional correctness."
                    },
                    {
                        "title": "Universal Approximation with Certified Networks",
                        "abstract": "Training neural networks to be certifiably robust is critical to ensure their safety against adversarial attacks. However, it is currently very difficult to train a neural network that is both accurate and certifiably robust. In this work we take a step towards addressing this challenge. We prove that for every continuous function $f$, there exists a network $n$ such that: (i) $n$ approximates $f$ arbitrarily close, and (ii) simple interval bound propagation of a region $B$ through $n$ yields a result that is arbitrarily close to the optimal output of $f$ on $B$. Our result can be seen as a Universal Approximation Theorem for interval-certified ReLU networks. To the best of our knowledge, this is the first work to prove the existence of accurate, interval-certified networks."
                    },
                    {
                        "title": "Certified Defense to Image Transformations via Randomized Smoothing",
                        "abstract": "We extend randomized smoothing to cover parameterized transformations (e.g., rotations, translations) and certify robustness in the parameter space (e.g., rotation angle). This is particularly challenging as interpolation and rounding effects mean that image transformations do not compose, in turn preventing direct certification of the perturbed image (unlike certification with $\\ell^p$ norms). We address this challenge by introducing three different kinds of defenses, each with a different guarantee (heuristic, distributional and individual) stemming from the method used to bound the interpolation error. Importantly, we show how individual certificates can be obtained via either statistical error bounds or efficient online inverse computation of the image transformation. We provide an implementation of all methods at https://github.com/eth-sri/transformation-smoothing."
                    },
                    {
                        "title": "On the Paradox of Certified Training",
                        "abstract": "Certified defenses based on convex relaxations are an established technique for training provably robust models. The key component is the choice of relaxation, varying from simple intervals to tight polyhedra. Counterintuitively, loose interval-based training often leads to higher certified robustness than what can be achieved with tighter relaxations, which is a well-known but poorly understood paradox. While recent works introduced various improvements aiming to circumvent this issue in practice, the fundamental problem of training models with high certified robustness remains unsolved. In this work, we investigate the underlying reasons behind the paradox and identify two key properties of relaxations, beyond tightness, that impact certified training dynamics: continuity and sensitivity. Our extensive experimental evaluation with a number of popular convex relaxations provides strong evidence that these factors can explain the drop in certified robustness observed for tighter relaxations. We also systematically explore modifications of existing relaxations and discover that improving unfavorable properties is challenging, as such attempts often harm other properties, revealing a complex tradeoff. Our findings represent an important first step towards understanding the intricate optimization challenges involved in certified training."
                    },
                    {
                        "title": "Automated Discovery of Adaptive Attacks on Adversarial Defenses",
                        "abstract": "Reliable evaluation of adversarial defenses is a challenging task, currently limited to an expert who manually crafts attacks that exploit the defense's inner workings or approaches based on an ensemble of fixed attacks, none of which may be effective for the specific defense at hand. Our key observation is that adaptive attacks are composed of reusable building blocks that can be formalized in a search space and used to automatically discover attacks for unknown defenses. We evaluated our approach on 24 adversarial defenses and show that it outperforms AutoAttack, the current state-of-the-art tool for reliable evaluation of adversarial defenses: our tool discovered significantly stronger attacks by producing 3.0\\%-50.8\\% additional adversarial examples for 10 models, while obtaining attacks with slightly stronger or similar strength for the remaining models."
                    },
                    {
                        "title": "Scalable Certified Segmentation via Randomized Smoothing",
                        "abstract": "We present a new certification method for image and point cloud segmentation based on randomized smoothing. The method leverages a novel scalable algorithm for prediction and certification that correctly accounts for multiple testing, necessary for ensuring statistical guarantees. The key to our approach is reliance on established multiple-testing correction mechanisms as well as the ability to abstain from classifying single pixels or points while still robustly segmenting the overall input. Our experimental evaluation on synthetic data and challenging datasets, such as Pascal Context, Cityscapes, and ShapeNet, shows that our algorithm can achieve, for the first time, competitive accuracy and certification guarantees on real-world segmentation tasks. We provide an implementation at https://github.com/eth-sri/segmentation-smoothing."
                    },
                    {
                        "title": "Learning to Infer User Interface Attributes from Images",
                        "abstract": "We explore a new domain of learning to infer user interface attributes that helps developers automate the process of user interface implementation. Concretely, given an input image created by a designer, we learn to infer its implementation which when rendered, looks visually the same as the input image. To achieve this, we take a black box rendering engine and a set of attributes it supports (e.g., colors, border radius, shadow or text properties), use it to generate a suitable synthetic training dataset, and then train specialized neural models to predict each of the attribute values. To improve pixel-level accuracy, we additionally use imitation learning to train a neural policy that refines the predicted attribute values by learning to compute the similarity of the original and rendered images in their attribute space, rather than based on the difference of pixel values. We instantiate our approach to the task of inferring Android Button attribute values and achieve 92.5% accuracy on a dataset consisting of real-world Google Play Store applications."
                    },
                    {
                        "title": "Adversarial Attacks on Probabilistic Autoregressive Forecasting Models",
                        "abstract": "We develop an effective generation of adversarial attacks on neural models that output a sequence of probability distributions rather than a sequence of single values. This setting includes the recently proposed deep probabilistic autoregressive forecasting models that estimate the probability distribution of a time series given its past and achieve state-of-the-art results in a diverse set of application domains. The key technical challenge we address is effectively differentiating through the Monte-Carlo estimation of statistics of the joint distribution of the output sequence. Additionally, we extend prior work on probabilistic forecasting to the Bayesian setting which allows conditioning on future observations, instead of only on past observations. We demonstrate that our approach can successfully generate attacks with small input perturbations in two challenging tasks where robust decision making is crucial: stock market trading and prediction of electricity consumption."
                    },
                    {
                        "title": "Robustness Certification of Generative Models",
                        "abstract": "Generative neural networks can be used to specify continuous transformations between images via latent-space interpolation. However, certifying that all images captured by the resulting path in the image manifold satisfy a given property can be very challenging. This is because this set is highly non-convex, thwarting existing scalable robustness analysis methods, which are often based on convex relaxations. We present ApproxLine, a scalable certification method that successfully verifies non-trivial specifications involving generative models and classifiers. ApproxLine can provide both sound deterministic and probabilistic guarantees, by capturing either infinite non-convex sets of neural network activation vectors or distributions over such sets. We show that ApproxLine is practically useful and can verify interesting interpolations in the networks latent space."
                    },
                    {
                        "title": "Efficient Certification of Spatial Robustness",
                        "abstract": "Recent work has exposed the vulnerability of computer vision models to vector field attacks. Due to the widespread usage of such models in safety-critical applications, it is crucial to quantify their robustness against such spatial transformations. However, existing work only provides empirical robustness quantification against vector field deformations via adversarial attacks, which lack provable guarantees. In this work, we propose novel convex relaxations, enabling us, for the first time, to provide a certificate of robustness against vector field transformations. Our relaxations are model-agnostic and can be leveraged by a wide range of neural network verifiers. Experiments on various network architectures and different datasets demonstrate the effectiveness and scalability of our method."
                    },
                    {
                        "title": "Fair Normalizing Flows",
                        "abstract": "Fair representation learning is an attractive approach that promises fairness of downstream predictors by encoding sensitive data. Unfortunately, recent work has shown that strong adversarial predictors can still exhibit unfairness by recovering sensitive attributes from these representations. In this work, we present Fair Normalizing Flows (FNF), a new approach offering more rigorous fairness guarantees for learned representations. Specifically, we consider a practical setting where we can estimate the probability density for sensitive groups. The key idea is to model the encoder as a normalizing flow trained to minimize the statistical distance between the latent representations of different groups. The main advantage of FNF is that its exact likelihood computation allows us to obtain guarantees on the maximum unfairness of any potentially adversarial downstream predictor. We experimentally demonstrate the effectiveness of FNF in enforcing various group fairness notions, as well as other attractive properties such as interpretability and transfer learning, on a variety of challenging real-world datasets."
                    },
                    {
                        "title": "The Fundamental Limits of Interval Arithmetic for Neural Networks",
                        "abstract": "Interval analysis (or interval bound propagation, IBP) is a popular technique for verifying and training provably robust deep neural networks, a fundamental challenge in the area of reliable machine learning. However, despite substantial efforts, progress on addressing this key challenge has stagnated, calling into question whether interval arithmetic is a viable path forward.   In this paper we present two fundamental results on the limitations of interval arithmetic for analyzing neural networks. Our main impossibility theorem states that for any neural network classifying just three points, there is a valid specification over these points that interval analysis can not prove. Further, in the restricted case of one-hidden-layer neural networks we show a stronger impossibility result: given any radius $\\alpha < 1$, there is a set of $O(\\alpha^{-1})$ points with robust radius $\\alpha$, separated by distance $2$, that no one-hidden-layer network can be proven to classify robustly via interval analysis."
                    },
                    {
                        "title": "Private and Reliable Neural Network Inference",
                        "abstract": "Reliable neural networks (NNs) provide important inference-time reliability guarantees such as fairness and robustness. Complementarily, privacy-preserving NN inference protects the privacy of client data. So far these two emerging areas have been largely disconnected, yet their combination will be increasingly important. In this work, we present the first system which enables privacy-preserving inference on reliable NNs. Our key idea is to design efficient fully homomorphic encryption (FHE) counterparts for the core algorithmic building blocks of randomized smoothing, a state-of-the-art technique for obtaining reliable models. The lack of required control flow in FHE makes this a demanding task, as na\\\"ive solutions lead to unacceptable runtime. We employ these building blocks to enable privacy-preserving NN inference with robustness and fairness guarantees in a system called Phoenix. Experimentally, we demonstrate that Phoenix achieves its goals without incurring prohibitive latencies. To our knowledge, this is the first work which bridges the areas of client data privacy and reliability guarantees for NNs."
                    },
                    {
                        "title": "Learning to Configure Computer Networks with Neural Algorithmic Reasoning",
                        "abstract": "We present a new method for scaling automatic configuration of computer networks. The key idea is to relax the computationally hard search problem of finding a configuration that satisfies a given specification into an approximate objective amenable to learning-based techniques. Based on this idea, we train a neural algorithmic model which learns to generate configurations likely to (fully or partially) satisfy a given specification under existing routing protocols. By relaxing the rigid satisfaction guarantees, our approach (i) enables greater flexibility: it is protocol-agnostic, enables cross-protocol reasoning, and does not depend on hardcoded rules; and (ii) finds configurations for much larger computer networks than previously possible. Our learned synthesizer is up to 490x faster than state-of-the-art SMT-based methods, while producing configurations which on average satisfy more than 93% of the provided requirements."
                    },
                    {
                        "title": "Reqomp: Space-constrained Uncomputation for Quantum Circuits",
                        "abstract": "Quantum circuits must run on quantum computers with tight limits on qubit and gate counts. To generate circuits respecting both limits, a promising opportunity is exploiting uncomputation to trade qubits for gates. We present Reqomp, a method to automatically synthesize correct and efficient uncomputation of ancillae while respecting hardware constraints. For a given circuit, Reqomp can offer a wide range of trade-offs between tightly constraining qubit count or gate count. Our evaluation demonstrates that Reqomp can significantly reduce the number of required ancilla qubits by up to 96%. On 80% of our benchmarks, the ancilla qubits required can be reduced by at least 25% while never incurring a gate count increase beyond 28%."
                    },
                    {
                        "title": "Abstraqt: Analysis of Quantum Circuits via Abstract Stabilizer Simulation",
                        "abstract": "Stabilizer simulation can efficiently simulate an important class of quantum circuits consisting exclusively of Clifford gates. However, all existing extensions of this simulation to arbitrary quantum circuits including non-Clifford gates suffer from an exponential runtime.   To address this challenge, we present a novel approach for efficient stabilizer simulation on arbitrary quantum circuits, at the cost of lost precision. Our key idea is to compress an exponential sum representation of the quantum state into a single abstract summand covering (at least) all occurring summands. This allows us to introduce an abstract stabilizer simulator that efficiently manipulates abstract summands by over-approximating the effect of circuit operations including Clifford gates, non-Clifford gates, and (internal) measurements.   We implemented our abstract simulator in a tool called Abstraqt and experimentally demonstrate that Abstraqt can establish circuit properties intractable for existing techniques."
                    },
                    {
                        "title": "CuTS: Customizable Tabular Synthetic Data Generation",
                        "abstract": "Privacy, data quality, and data sharing concerns pose a key limitation for tabular data applications. While generating synthetic data resembling the original distribution addresses some of these issues, most applications would benefit from additional customization on the generated data. However, existing synthetic data approaches are limited to particular constraints, e.g., differential privacy (DP) or fairness. In this work, we introduce CuTS, the first customizable synthetic tabular data generation framework. Customization in CuTS is achieved via declarative statistical and logical expressions, supporting a wide range of requirements (e.g., DP or fairness, among others). To ensure high synthetic data quality in the presence of custom specifications, CuTS is pre-trained on the original dataset and fine-tuned on a differentiable loss automatically derived from the provided specifications using novel relaxations. We evaluate CuTS over four datasets and on numerous custom specifications, outperforming state-of-the-art specialized approaches on several tasks while being more general. In particular, at the same fairness level, we achieve 2.3% higher downstream accuracy than the state-of-the-art in fair synthetic data generation on the Adult dataset."
                    }
                ]
            }
        }
    },
    "2405.14669": {
        "paper_data": {
            "title": "Efficiency for Free: Ideal Data Are Transportable Representations",
            "url": "http://arxiv.org/abs/2405.14669v1",
            "arxiv_id": "2405.14669",
            "authors": [
                "Peng Sun",
                "Yi Jiang",
                "Tao Lin"
            ],
            "abstract": "Data, the seminal opportunity and challenge in modern machine learning, currently constrains the scalability of representation learning and impedes the pace of model evolution. Existing paradigms tackle the issue of learning efficiency over massive datasets from the perspective of self-supervised learning and dataset distillation independently, while neglecting the untapped potential of accelerating representation learning from an intermediate standpoint. In this work, we delve into defining the ideal data properties from both optimization and generalization perspectives. We propose that model-generated representations, despite being trained on diverse tasks and architectures, converge to a shared linear space, facilitating effective linear transport between models. Furthermore, we demonstrate that these representations exhibit properties conducive to the formation of ideal data. The theoretical/empirical insights therein inspire us to propose a Representation Learning Accelerator (ReLA), which leverages a task- and architecture-agnostic, yet publicly available, free model to form a dynamic data subset and thus accelerate (self-)supervised learning. For instance, employing a CLIP ViT B/16 as a prior model for dynamic data generation, ReLA-aided BYOL can train a ResNet-50 from scratch with 50% of ImageNet-1K, yielding performance surpassing that of training on the full dataset. Additionally, employing a ResNet-18 pre-trained on CIFAR-10 can enhance ResNet-50 training on 10% of ImageNet-1K, resulting in a 7.7% increase in accuracy.",
            "introduction": "   1 Introduction  The available of massive datasets\u00a0[20, 46] and recent advances in parallel data processing\u00a0[28, 39] have facilitated the rapid evolution of large deep models, such as GPT-4\u00a0[1] and LVM\u00a0[2]. These models excel in numerous learning tasks, attributable to their impressive representation capabilities. However, the emergence of vast amounts of data within the modern deep learning paradigm raises two fundamental challenges: (i) the demand for human annotations of huge datasets consumes significant social resources\u00a0[42, 20, 40]; (ii) training large models with increasing data and model capacity suffers from intensive computational burden\u00a0[6, 16, 54].   The community has made considerable efforts to enhance learning efficiency. Self-supervised learning methods\u00a0[12, 67, 29, 10, 26, 13, 9, 3], with their superior representation learning devoid of human annotations via the self-learning paradigm, attempt to tackle the challenge (i). Concurrently, research has been conducted to mitigate data efficiency issues in challenge (ii): dataset distillation approaches\u00a0[64, 55, 11, 70, 51, 52] have successfully synthesized a small distilled dataset, on which models trained on this compact dataset can akin to one trained on the full dataset.   Table 1: Key issues in self-supervised learning methods and dataset distillation approaches.     Key issues   Solved by       Self-supervised learning methods:      (a)     inefficiency in the learning procedure compared to conventional supervised learning arises due to sub-optimal self-generating targets\u00a0[29, 61], as well as issues such as \u2018noisy mapping\u2019 discussed in Section\u00a03.3;     1 \u2004  2 \u2004  4       (b)     introducing a large volume of data marginally enhances learning performance but significantly exacerbates inefficiency due to the neural scaling law\u00a0[53, 23];     1 \u2004  4 \u2004  5     Dataset distillation approaches:      (c)     although training on the distilled dataset is efficient and effective, the distillation process of optimization-based approaches\u00a0[11, 70, 36] is computationally demanding\u00a0[18, 55], often surpassing the computational load of training on the full dataset.     4 \u2004  5       (d)     high GPU memory consumption [18] constrains the scale of model architectures during distillation, thereby limiting the scale and quality of the distilled dataset;     4 \u2004  5       (e)     a recent approach\u00a0[55] addresses issues (c) and (d) through an optimization-free distillation process. However, it necessitates a pre-trained model on the full dataset and is applicable solely to labeled datasets.     1 \u2004  4       (f)     the theoretical underpinnings of these approaches remain inadequately explored, and their efficacy lacks formal guarantees, such as generalization bounds or convergence rates for training on distilled data.     1 \u2004  3 \u2004  4        However, challenges (i) and (ii) persist and yet are far from being solved\u00a0[40, 42, 6, 16, 54], particularly the intervention of these two learning paradigms. In this manuscript, we outline the six key remaining issues in Table\u00a01, and list our five key contributions below as the first step toward bridging representation learning with data-efficient learning:    1  Revealing the optimal properties of distilled data for efficient (self-)supervised learning. Through a study on linear models, we derive an upper bound for the convergence rate of training on distilled data and demonstrate that maximizing this rate necessitates specific data properties\u2014perfect bijective mappings between the samples and targets within a dataset (c.f. Section\u00a03).     2  Identifying the inefficiency problems of (self-)supervised learning from a data-centric perspective. Specifically, we identify several problems (c.f. Section\u00a03.3) contributing to the inefficiency of (self-)supervised learning. For instance, common data augmentation techniques in modern deep learning algorithms can implicitly",
            "references": [
                {
                    "title": "CatLIP: CLIP-level Visual Recognition Accuracy with 2.7x Faster Pre-training on Web-scale Image-Text Data",
                    "abstract": "Contrastive learning has emerged as a transformative method for learning effective visual representations through the alignment of image and text embeddings. However, pairwise similarity computation in contrastive loss between image and text pairs poses computational challenges. This paper presents a novel weakly supervised pre-training of vision models on web-scale image-text data. The proposed method reframes pre-training on image-text data as a classification task. Consequently, it eliminates the need for pairwise similarity computations in contrastive loss, achieving a remarkable $2.7\\times$ acceleration in training speed compared to contrastive learning on web-scale data. Through extensive experiments spanning diverse vision tasks, including detection and segmentation, we demonstrate that the proposed method maintains high representation quality. Our source code along with pre-trained model weights and training recipes is available at \\url{https://github.com/apple/corenet}."
                },
                {
                    "title": "Elucidating the Design Space of Dataset Condensation",
                    "abstract": "Dataset condensation, a concept within data-centric learning, efficiently transfers critical attributes from an original dataset to a synthetic version, maintaining both diversity and realism. This approach significantly improves model training efficiency and is adaptable across multiple application areas. Previous methods in dataset condensation have faced challenges: some incur high computational costs which limit scalability to larger datasets (e.g., MTT, DREAM, and TESLA), while others are restricted to less optimal design spaces, which could hinder potential improvements, especially in smaller datasets (e.g., SRe2L, G-VBSM, and RDED). To address these limitations, we propose a comprehensive design framework that includes specific, effective strategies like implementing soft category-aware matching and adjusting the learning rate schedule. These strategies are grounded in empirical evidence and theoretical backing. Our resulting approach, Elucidate Dataset Condensation (EDC), establishes a benchmark for both small and large-scale dataset condensation. In our testing, EDC achieves state-of-the-art accuracy, reaching 48.6% on ImageNet-1k with a ResNet-18 model at an IPC of 10, which corresponds to a compression ratio of 0.78%. This performance exceeds those of SRe2L, G-VBSM, and RDED by margins of 27.3%, 17.2%, and 6.6%, respectively."
                },
                {
                    "title": "Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision",
                    "abstract": "Widely used alignment techniques, such as reinforcement learning from human feedback (RLHF), rely on the ability of humans to supervise model behavior - for example, to evaluate whether a model faithfully followed instructions or generated safe outputs. However, future superhuman models will behave in complex ways too difficult for humans to reliably evaluate; humans will only be able to weakly supervise superhuman models. We study an analogy to this problem: can weak model supervision elicit the full capabilities of a much stronger model? We test this using a range of pretrained language models in the GPT-4 family on natural language processing (NLP), chess, and reward modeling tasks. We find that when we naively finetune strong pretrained models on labels generated by a weak model, they consistently perform better than their weak supervisors, a phenomenon we call weak-to-strong generalization. However, we are still far from recovering the full capabilities of strong models with naive finetuning alone, suggesting that techniques like RLHF may scale poorly to superhuman models without further work. We find that simple methods can often significantly improve weak-to-strong generalization: for example, when finetuning GPT-4 with a GPT-2-level supervisor and an auxiliary confidence loss, we can recover close to GPT-3.5-level performance on NLP tasks. Our results suggest that it is feasible to make empirical progress today on a fundamental challenge of aligning superhuman models."
                },
                {
                    "title": "On the Diversity and Realism of Distilled Dataset: An Efficient Dataset Distillation Paradigm",
                    "abstract": "Contemporary machine learning, which involves training large neural networks on massive datasets, faces significant computational challenges. Dataset distillation, as a recent emerging strategy, aims to compress real-world datasets for efficient training. However, this line of research currently struggles with large-scale and high-resolution datasets, hindering its practicality and feasibility. Thus, we re-examine existing methods and identify three properties essential for real-world applications: realism, diversity, and efficiency. As a remedy, we propose RDED, a novel computationally-efficient yet effective data distillation paradigm, to enable both diversity and realism of the distilled data. Extensive empirical results over various model architectures and datasets demonstrate the advancement of RDED: we can distill a dataset to 10 images per class from full ImageNet-1K [6] within 7 minutes, achieving a notable 42% accuracy with ResNet-18 [14] on a single RTX-4090 GPU (while the SOTA only achieves 21% but requires 6 hours). Code: https://github.com/LINs-1ab/RDED."
                },
                {
                    "title": "Sequential Modeling Enables Scalable Learning for Large Vision Models",
                    "abstract": "We introduce a novel sequential modeling approach which enables learning a Large Vision Model (LVM) without making use of any linguistic data. To do this, we define a common format, \u201cvisual sentences \u201d, in which we can represent raw images and videos as well as annotated data sources such as semantic segmentations and depth reconstructions with-out needing any meta-knowledge beyond the pixels. Once this wide variety of visual data (comprising 420 billion to-kens) is represented as sequences, the model can be trained to minimize a cross-entropy loss for next token prediction. By training across various scales of model architecture and data diversity, we provide empirical evidence that our models scale effectively. Many different vision tasks can be solved by designing suitable visual prompts at test time."
                },
                {
                    "title": "Generalized Large-Scale Data Condensation via Various Backbone and Statistical Matching",
                    "abstract": "The lightweight \u201clocal-match-global\u201d matching introduced by SRe2L successfully creates a distilled dataset with comprehensive information on the full 224\u00d7224 ImageNetlk. However, this one-sided approach is limited to a particular backbone, layer, and statistics, which limits the improvement of the generalization of a distilled dataset. We suggest that sufficient and various \u201clocal-match-global\u201d matching are more precise and effective than a single one and have the ability to create a distilled dataset with richer information and better generalization ability. We call this perspective \u201cgeneralized matching\u201d and propose Generalized Various Backbone and Statistical Matching (G-VBSM) in this work, which aims to create a synthetic dataset with densities, ensuring consistency with the complete dataset across various backbones, layers, and statistics. As experimentally demonstrated, G-VBSM is the first algorithm to obtain strong performance across both small-scale and large-scale datasets. Specifically, G-VBSM achieves performances of 38.7% on CIFAR-I00, 47.6% on Tiny-ImageNet, and 31.4% on the full 224\u00d7224 ImageNet1 k, respectively11Settings: CIFAR-I00 with 128-width ConvNet under 10 images per class (lPC), Tiny-ImageNet with ResNet18 under 50 IPC, and ImageNetlk with ResNet18 under 10 IPC.. These results surpass all SOTA methods by margins of 3.9%, 6.5%, and 10.1%, respectively."
                },
                {
                    "title": "Self-Supervised Dataset Distillation for Transfer Learning",
                    "abstract": "Dataset distillation methods have achieved remarkable success in distilling a large dataset into a small set of representative samples. However, they are not designed to produce a distilled dataset that can be effectively used for facilitating self-supervised pre-training. To this end, we propose a novel problem of distilling an unlabeled dataset into a set of small synthetic samples for efficient self-supervised learning (SSL). We first prove that a gradient of synthetic samples with respect to a SSL objective in naive bilevel optimization is \\textit{biased} due to the randomness originating from data augmentations or masking. To address this issue, we propose to minimize the mean squared error (MSE) between a model's representations of the synthetic examples and their corresponding learnable target feature representations for the inner objective, which does not introduce any randomness. Our primary motivation is that the model obtained by the proposed inner optimization can mimic the \\textit{self-supervised target model}. To achieve this, we also introduce the MSE between representations of the inner model and the self-supervised target model on the original full dataset for outer optimization. Lastly, assuming that a feature extractor is fixed, we only optimize a linear head on top of the feature extractor, which allows us to reduce the computational cost and obtain a closed-form solution of the head with kernel ridge regression. We empirically validate the effectiveness of our method on various applications involving transfer learning."
                },
                {
                    "title": "Towards Lossless Dataset Distillation via Difficulty-Aligned Trajectory Matching",
                    "abstract": "The ultimate goal of Dataset Distillation is to synthesize a small synthetic dataset such that a model trained on this synthetic set will perform equally well as a model trained on the full, real dataset. Until now, no method of Dataset Distillation has reached this completely lossless goal, in part due to the fact that previous methods only remain effective when the total number of synthetic samples is extremely small. Since only so much information can be contained in such a small number of samples, it seems that to achieve truly loss dataset distillation, we must develop a distillation method that remains effective as the size of the synthetic dataset grows. In this work, we present such an algorithm and elucidate why existing methods fail to generate larger, high-quality synthetic sets. Current state-of-the-art methods rely on trajectory-matching, or optimizing the synthetic data to induce similar long-term training dynamics as the real data. We empirically find that the training stage of the trajectories we choose to match (i.e., early or late) greatly affects the effectiveness of the distilled dataset. Specifically, early trajectories (where the teacher network learns easy patterns) work well for a low-cardinality synthetic set since there are fewer examples wherein to distribute the necessary information. Conversely, late trajectories (where the teacher network learns hard patterns) provide better signals for larger synthetic sets since there are now enough samples to represent the necessary complex patterns. Based on our findings, we propose to align the difficulty of the generated patterns with the size of the synthetic dataset. In doing so, we successfully scale trajectory matching-based methods to larger synthetic datasets, achieving lossless dataset distillation for the very first time. Code and distilled datasets are available at https://gzyaftermath.github.io/DATM."
                },
                {
                    "title": "Squeeze, Recover and Relabel: Dataset Condensation at ImageNet Scale From A New Perspective",
                    "abstract": "We present a new dataset condensation framework termed Squeeze, Recover and Relabel (SRe$^2$L) that decouples the bilevel optimization of model and synthetic data during training, to handle varying scales of datasets, model architectures and image resolutions for efficient dataset condensation. The proposed method demonstrates flexibility across diverse dataset scales and exhibits multiple advantages in terms of arbitrary resolutions of synthesized images, low training cost and memory consumption with high-resolution synthesis, and the ability to scale up to arbitrary evaluation network architectures. Extensive experiments are conducted on Tiny-ImageNet and full ImageNet-1K datasets. Under 50 IPC, our approach achieves the highest 42.5% and 60.8% validation accuracy on Tiny-ImageNet and ImageNet-1K, outperforming all previous state-of-the-art methods by margins of 14.5% and 32.9%, respectively. Our approach also surpasses MTT in terms of speed by approximately 52$\\times$ (ConvNet-4) and 16$\\times$ (ResNet-18) faster with less memory consumption of 11.6$\\times$ and 6.4$\\times$ during data synthesis. Our code and condensed datasets of 50, 200 IPC with 4K recovery budget are available at https://github.com/VILA-Lab/SRe2L."
                },
                {
                    "title": "Improved Distribution Matching for Dataset Condensation",
                    "abstract": "Dataset Condensation aims to condense a large dataset into a smaller one while maintaining its ability to train a well-performing model, thus reducing the storage cost and training effort in deep learning applications. However, conventional dataset condensation methods are optimization-oriented and condense the dataset by performing gradient or parameter matching during model optimization, which is computationally intensive even on small datasets and models. In this paper, we propose a novel dataset condensation method based on distribution matching, which is more efficient and promising. Specifically, we identify two important shortcomings of naive distribution matching (i.e., imbalanced feature numbers and unvalidated embeddings for distance computation) and address them with three novel techniques (i.e., partitioning and expansion augmentation, efficient and enriched model sampling, and class-aware distribution regularization). Our simple yet effective method outperforms most previous optimization-oriented methods with much fewer computational resources, thereby scaling data condensation to larger datasets and models. Extensive experiments demonstrate the effectiveness of our method. Codes are available at https://github.com/uitrbn/IDM"
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "DREAM: Efficient Dataset Distillation by Representative Matching",
                    "abstract": "Dataset distillation aims to synthesize small datasets with little information loss from original large-scale ones for reducing storage and training costs. Recent state-of-the-art methods mainly constrain the sample synthesis process by matching synthetic images and the original ones regarding gradients, embedding distributions, or training trajectories. Although there are various matching objectives, currently the strategy for selecting original images is limited to naive random sampling. We argue that random sampling overlooks the evenness of the selected sample distribution, which may result in noisy or biased matching targets. Besides, the sample diversity is also not constrained by random sampling. These factors together lead to optimization instability in the distilling process and degrade the training efficiency. Accordingly, we propose a novel matching strategy named as Dataset distillation by REpresentAtive Matching (DREAM), where only representative original images are selected for matching. DREAM is able to be easily plugged into popular dataset distillation frameworks and reduce the distilling iterations by more than 8 times without performance drop. Given sufficient training time, DREAM further provides significant improvements and achieves state-of-the-art performances."
                },
                {
                    "title": "Dataset Distillation: A Comprehensive Review",
                    "abstract": "Recent success of deep learning is largely attributed to the sheer amount of data used for training deep neural networks. Despite the unprecedented success, the massive data, unfortunately, significantly increases the burden on storage and transmission and further gives rise to a cumbersome model training process. Besides, relying on the raw data for training per se yields concerns about privacy and copyright. To alleviate these shortcomings, dataset distillation (DD), also known as dataset condensation (DC), was introduced and has recently attracted much research attention in the community. Given an original dataset, DD aims to derive a much smaller dataset containing synthetic samples, based on which the trained models yield performance comparable with those trained on the original dataset. In this paper, we give a comprehensive review and summary of recent advances in DD and its application. We first introduce the task formally and propose an overall algorithmic framework followed by all existing DD methods. Next, we provide a systematic taxonomy of current methodologies in this area, and discuss their theoretical interconnections. We also present current challenges in DD through extensive empirical studies and envision possible directions for future works."
                },
                {
                    "title": "Accelerating Dataset Distillation via Model Augmentation",
                    "abstract": "Dataset Distillation (DD), a newly emerging field, aims at generating much smaller but efficient synthetic training datasets from large ones. Existing DD methods based on gradient matching achieve leading performance; however, they are extremely computationally intensive as they require continuously optimizing a dataset among thousands of randomly initialized models. In this paper, we assume that training the synthetic data with diverse models leads to better generalization performance. Thus we propose two model augmentation techniques, i.e. using early-stage models and parameter perturbation to learn an informative synthetic set with significantly reduced training cost. Extensive experiments demonstrate that our method achieves up to 20\u00d7 speedup and comparable performance on par with state-of-the-art methods."
                },
                {
                    "title": "Minimizing the Accumulated Trajectory Error to Improve Dataset Distillation",
                    "abstract": "Model-based deep learning has achieved astounding successes due in part to the availability of large-scale real-world data. However, processing such massive amounts of data comes at a considerable cost in terms of computations, storage, training and the search for good neural architectures. Dataset distillation has thus recently come to the fore. This paradigm involves distilling information from large real-world datasets into tiny and compact synthetic datasets such that processing the latter ideally yields similar performances as the former. State-of-the-art methods primarily rely on learning the synthetic dataset by matching the gradients obtained during training between the real and synthetic data. However, these gradient-matching methods suffer from the so-called accumulated trajectory error caused by the discrepancy between the distillation and subsequent evaluation. To mitigate the adverse impact of this accumulated trajectory error, we propose a novel approach that encourages the optimization algorithm to seek a flat trajectory. We show that the weights trained on synthetic data are robust against the accumulated errors perturbations with the regularization towards the flat trajectory. Our method, called Flat Trajectory Distillation (FTD), is shown to boost the performance of gradient-matching methods by up to 4.7% on a subset of images of the ImageNet dataset with higher resolution images. We also validate the effectiveness and generalizability of our method with datasets of different resolutions and demonstrate its applicability to neural architecture search. Code is available at. https://github.com/AngusDujw/FTD-distillation."
                },
                {
                    "title": "Scaling Up Dataset Distillation to ImageNet-1K with Constant Memory",
                    "abstract": "Dataset Distillation is a newly emerging area that aims to distill large datasets into much smaller and highly informative synthetic ones to accelerate training and reduce storage. Among various dataset distillation methods, trajectory-matching-based methods (MTT) have achieved SOTA performance in many tasks, e.g., on CIFAR-10/100. However, due to exorbitant memory consumption when unrolling optimization through SGD steps, MTT fails to scale to large-scale datasets such as ImageNet-1K. Can we scale this SOTA method to ImageNet-1K and does its effectiveness on CIFAR transfer to ImageNet-1K? To answer these questions, we first propose a procedure to exactly compute the unrolled gradient with constant memory complexity, which allows us to scale MTT to ImageNet-1K seamlessly with ~6x reduction in memory footprint. We further discover that it is challenging for MTT to handle datasets with a large number of classes, and propose a novel soft label assignment that drastically improves its convergence. The resulting algorithm sets new SOTA on ImageNet-1K: we can scale up to 50 IPCs (Image Per Class) on ImageNet-1K on a single GPU (all previous methods can only scale to 2 IPCs on ImageNet-1K), leading to the best accuracy (only 5.9% accuracy drop against full dataset training) while utilizing only 4.2% of the number of data points - an 18.2% absolute gain over prior SOTA. Our code is available at https://github.com/justincui03/tesla"
                },
                {
                    "title": "DC-BENCH: Dataset Condensation Benchmark",
                    "abstract": "Dataset Condensation is a newly emerging technique aiming at learning a tiny dataset that captures the rich information encoded in the original dataset. As the size of datasets contemporary machine learning models rely on becomes increasingly large, condensation methods become a prominent direction for accelerating network training and reducing data storage. Despite numerous methods have been proposed in this rapidly growing field, evaluating and comparing different condensation methods is non-trivial and still remains an open issue. The quality of condensed dataset are often shadowed by many critical contributing factors to the end performance, such as data augmentation and model architectures. The lack of a systematic way to evaluate and compare condensation methods not only hinders our understanding of existing techniques, but also discourages practical usage of the synthesized datasets. This work provides the first large-scale standardized benchmark on Dataset Condensation. It consists of a suite of evaluations to comprehensively reflect the generability and effectiveness of condensation methods through the lens of their generated dataset. Leveraging this benchmark, we conduct a large-scale study of current condensation methods, and report many insightful findings that open up new possibilities for future development. The benchmark library, including evaluators, baseline methods, and generated datasets, is open-sourced to facilitate future research and application."
                },
                {
                    "title": "Beyond neural scaling laws: beating power law scaling via data pruning",
                    "abstract": "Widely observed neural scaling laws, in which error falls off as a power of the training set size, model size, or both, have driven substantial performance improvements in deep learning. However, these improvements through scaling alone require considerable costs in compute and energy. Here we focus on the scaling of error with dataset size and show how in theory we can break beyond power law scaling and potentially even reduce it to exponential scaling instead if we have access to a high-quality data pruning metric that ranks the order in which training examples should be discarded to achieve any pruned dataset size. We then test this improved scaling prediction with pruned dataset size empirically, and indeed observe better than power law scaling in practice on ResNets trained on CIFAR-10, SVHN, and ImageNet. Next, given the importance of finding high-quality pruning metrics, we perform the first large-scale benchmarking study of ten different data pruning metrics on ImageNet. We find most existing high performing metrics scale poorly to ImageNet, while the best are computationally intensive and require labels for every image. We therefore developed a new simple, cheap and scalable self-supervised pruning metric that demonstrates comparable performance to the best supervised metrics. Overall, our work suggests that the discovery of good data-pruning metrics may provide a viable path forward to substantially improved neural scaling laws, thereby reducing the resource costs of modern deep learning."
                },
                {
                    "title": "Dataset Condensation via Efficient Synthetic-Data Parameterization",
                    "abstract": "The great success of machine learning with massive amounts of data comes at a price of huge computation costs and storage for training and tuning. Recent studies on dataset condensation attempt to reduce the dependence on such massive data by synthesizing a compact training dataset. However, the existing approaches have fundamental limitations in optimization due to the limited representability of synthetic datasets without considering any data regularity characteristics. To this end, we propose a novel condensation framework that generates multiple synthetic data with a limited storage budget via efficient parameterization considering data regularity. We further analyze the shortcomings of the existing gradient matching-based condensation methods and develop an effective optimization technique for improving the condensation of training data information. We propose a unified algorithm that drastically improves the quality of condensed data against the current state-of-the-art on CIFAR-10, ImageNet, and Speech Commands."
                },
                {
                    "title": "Dataset Distillation by Matching Training Trajectories",
                    "abstract": "Dataset distillation is the task of synthesizing a small dataset such that a model trained on the synthetic set will match the test accuracy of the model trained on the full dataset. In this paper, we propose a new formulation that optimizes our distilled data to guide networks to a similar state as those trained on real data across many training steps. Given a network, we train it for several iterations on our distilled data and optimize the distilled data with respect to the distance between the synthetically trained parameters and the parameters trained on real data. To efficiently obtain the initial and target network parameters for large-scale datasets, we pre-compute and store training trajectories of expert networks trained on the real dataset. Our method handily outperforms existing methods and also allows us to distill higher-resolution visual data."
                },
                {
                    "title": "Exploring the Equivalence of Siamese Self-Supervised Learning via A Unified Gradient Framework",
                    "abstract": "Self-supervised learning has shown its great potential to extract powerful visual representations without human annotations. Various works are proposed to deal with self-supervised learning from different perspectives: (1) contrastive learning methods (e.g., MoCo, SimCLR) utilize both positive and negative samples to guide the training direction; (2) asymmetric network methods (e.g., BYOL, SimSiam) get rid of negative samples via the introduction of a predictor network and the stop-gradient operation; (3) feature decorrelation methods (e.g., Barlow Twins, VICReg) instead aim to reduce the redundancy between feature dimensions. These methods appear to be quite different in the designed loss functions from various motivations. The final accuracy numbers also vary, where different networks and tricks are utilized in different works. In this work, we demonstrate that these methods can be unified into the same form. Instead of comparing their loss functions, we derive a unified formula through gradient analysis. Furthermore, we conduct fair and detailed experiments to compare their performances. It turns out that there is little gap between these methods, and the use of momentum encoder is the key factor to boost performance. From this unified framework, we propose UniGrad, a simple but effective gradient form for self-supervised learning. It does not require a memory bank or a predictor network, but can still achieve state-of-the-art performance and easily adopt other training strategies. Extensive experiments on linear evaluation and many downstream tasks also show its effectiveness. Code shall be released."
                },
                {
                    "title": "Dataset Condensation with Distribution Matching",
                    "abstract": "Computational cost of training state-of-the-art deep models in many learning problems is rapidly increasing due to more sophisticated models and larger datasets. A recent promising direction for reducing training cost is dataset condensation that aims to replace the original large training set with a significantly smaller learned synthetic set while preserving the original information. While training deep models on the small set of condensed images can be extremely fast, their synthesis remains computationally expensive due to the complex bi-level optimization and secondorder derivative computation. In this work, we propose a simple yet effective method that synthesizes condensed images by matching feature distributions of the synthetic and original training images in many sampled embedding spaces. Our method significantly reduces the synthesis cost while achieving comparable or better performance. Thanks to its efficiency, we apply our method to more realistic and larger datasets with sophisticated neural architectures and obtain a significant performance boost1. We also show promising practical benefits of our method in continual learning and neural architecture search."
                },
                {
                    "title": "Improving Contrastive Learning by Visualizing Feature Transformation",
                    "abstract": "Contrastive learning, which aims at minimizing the distance between positive pairs while maximizing that of negative ones, has been widely and successfully applied in unsupervised feature learning, where the design of positive and negative (pos/neg) pairs is one of its keys. In this paper, we attempt to devise a feature-level data manipulation, differing from data augmentation, to enhance the generic contrastive self-supervised learning. To this end, we first design a visualization scheme for pos/neg score 1 distribution, which enables us to analyze, interpret and understand the learning process. To our knowledge, this is the first attempt of its kind. More importantly, leveraging this tool, we gain some significant observations, which inspire our novel Feature Transformation proposals including the extrapolation of positives. This operation creates harder positives to boost the learning because hard positives enable the model to be more view-invariant. Besides, we propose the interpolation among negatives, which provides diversified negatives and makes the model more discriminative. It is the first attempt to deal with both challenges simultaneously. Experiment results show that our proposed Feature Transformation can improve at least 6.0% accuracy on ImageNet-100 over MoCo baseline, and about 2.0% accuracy on ImageNet-1K over the MoCoV2 baseline. Transferring to the downstream tasks successfully demonstrate our model is less task-bias. Visualization tools and codes: https://github.com/DTennant/CL-Visualizing-Feature-Transformation."
                },
                {
                    "title": "Solo-learn: A Library of Self-supervised Methods for Visual Representation Learning",
                    "abstract": "This paper presents solo-learn, a library of self-supervised methods for visual representation learning. Implemented in Python, using Pytorch and Pytorch lightning, the library fits both research and industry needs by featuring distributed training pipelines with mixed-precision, faster data loading via Nvidia DALI, online linear evaluation for better prototyping, and many additional training tricks. Our goal is to provide an easy-to-use library comprising a large amount of Self-supervised Learning (SSL) methods, that can be easily extended and fine-tuned by the community. solo-learn opens up avenues for exploiting large-budget SSL solutions on inexpensive smaller infrastructures and seeks to democratize SSL by making it accessible to all. The source code is available at https://github.com/vturrisi/solo-learn."
                },
                {
                    "title": "VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning",
                    "abstract": "Recent self-supervised methods for image representation learning are based on maximizing the agreement between embedding vectors from different views of the same image. A trivial solution is obtained when the encoder outputs constant vectors. This collapse problem is often avoided through implicit biases in the learning architecture, that often lack a clear justification or interpretation. In this paper, we introduce VICReg (Variance-Invariance-Covariance Regularization), a method that explicitly avoids the collapse problem with a simple regularization term on the variance of the embeddings along each dimension individually. VICReg combines the variance term with a decorrelation mechanism based on redundancy reduction and covariance regularization, and achieves results on par with the state of the art on several downstream tasks. In addition, we show that incorporating our new variance term into other methods helps stabilize the training and leads to performance improvements."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "Solving Inefficiency of Self-supervised Representation Learning",
                    "abstract": "Self-supervised learning (especially contrastive learning) has attracted great interest due to its huge potential in learning discriminative representations in an unsupervised manner. Despite the acknowledged successes, existing contrastive learning methods suffer from very low learning efficiency, e.g., taking about ten times more training epochs than supervised learning for comparable recognition accuracy. In this paper, we reveal two contradictory phenomena in contrastive learning that we call under-clustering and over-clustering problems, which are major obstacles to learning efficiency. Under-clustering means that the model cannot efficiently learn to discover the dissimilarity between inter-class samples when the negative sample pairs for contrastive learning are insufficient to differentiate all the actual object classes. Over-clustering implies that the model cannot efficiently learn features from excessive negative sample pairs, forcing the model to over-cluster samples of the same actual classes into different clusters. To simultaneously overcome these two problems, we propose a novel self-supervised learning framework using a truncated triplet loss. Precisely, we employ a triplet loss tending to maximize the relative distance between the positive pair and negative pairs to address the under-clustering problem; and we construct the negative pair by selecting a negative sample deputy from all negative samples to avoid the over-clustering problem, guaranteed by the Bernoulli Distribution model. We extensively evaluate our framework in several large-scale benchmarks (e.g., ImageNet, SYSU-30k, and COCO). The results demonstrate our model\u2019s superiority (e.g., the learning efficiency) over the latest state-of-the-art methods by a clear margin. See Codes1."
                },
                {
                    "title": "An Empirical Study of Training Self-Supervised Vision Transformers",
                    "abstract": "This paper does not describe a novel method. Instead, it studies a straightforward, incremental, yet must-know baseline given the recent progress in computer vision: self-supervised learning for Vision Transformers (ViT). While the training recipes for standard convolutional networks have been highly mature and robust, the recipes for ViT are yet to be built, especially in the self-supervised scenarios where training becomes more challenging. In this work, we go back to basics and investigate the effects of several fundamental components for training self-supervised ViT. We observe that instability is a major issue that degrades accuracy, and it can be hidden by apparently good results. We reveal that these results are indeed partial failure, and they can be improved when training is made more stable. We benchmark ViT results in MoCo v3 and several other self-supervised frameworks, with ablations in various aspects. We discuss the currently positive evidence as well as challenges and open questions. We hope that this work will provide useful data points and experience for future research."
                },
                {
                    "title": "Barlow Twins: Self-Supervised Learning via Redundancy Reduction",
                    "abstract": "Self-supervised learning (SSL) is rapidly closing the gap with supervised methods on large computer vision benchmarks. A successful approach to SSL is to learn embeddings which are invariant to distortions of the input sample. However, a recurring issue with this approach is the existence of trivial constant solutions. Most current methods avoid such solutions by careful implementation details. We propose an objective function that naturally avoids collapse by measuring the cross-correlation matrix between the outputs of two identical networks fed with distorted versions of a sample, and making it as close to the identity matrix as possible. This causes the embedding vectors of distorted versions of a sample to be similar, while minimizing the redundancy between the components of these vectors. The method is called Barlow Twins, owing to neuroscientist H. Barlow's redundancy-reduction principle applied to a pair of identical networks. Barlow Twins does not require large batches nor asymmetry between the network twins such as a predictor network, gradient stopping, or a moving average on the weight updates. Intriguingly it benefits from very high-dimensional output vectors. Barlow Twins outperforms previous methods on ImageNet for semi-supervised classification in the low-data regime, and is on par with current state of the art for ImageNet classification with a linear classifier head, and for transfer tasks of classification and object detection."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Understanding self-supervised Learning Dynamics without Contrastive Pairs",
                    "abstract": "While contrastive approaches of self-supervised learning (SSL) learn representations by minimizing the distance between two augmented views of the same data point (positive pairs) and maximizing views from different data points (negative pairs), recent \\emph{non-contrastive} SSL (e.g., BYOL and SimSiam) show remarkable performance {\\it without} negative pairs, with an extra learnable predictor and a stop-gradient operation. A fundamental question arises: why do these methods not collapse into trivial representations? We answer this question via a simple theoretical study and propose a novel approach, DirectPred, that \\emph{directly} sets the linear predictor based on the statistics of its inputs, without gradient training. On ImageNet, it performs comparably with more complex two-layer non-linear predictors that employ BatchNorm and outperforms a linear predictor by $2.5\\%$ in 300-epoch training (and $5\\%$ in 60-epoch). DirectPred is motivated by our theoretical study of the nonlinear learning dynamics of non-contrastive SSL in simple linear networks. Our study yields conceptual insights into how non-contrastive SSL methods learn, how they avoid representational collapse, and how multiple factors, like predictor networks, stop-gradients, exponential moving averages, and weight decay all come into play. Our simple theory recapitulates the results of real-world ablation studies in both STL-10 and ImageNet. Code is released https://github.com/facebookresearch/luckmatters/tree/master/ssl."
                },
                {
                    "title": "Exploring Simple Siamese Representation Learning",
                    "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code is made available.1"
                },
                {
                    "title": "AdCo: Adversarial Contrast for Efficient Learning of Unsupervised Representations from Self-Trained Negative Adversaries",
                    "abstract": "Contrastive learning relies on constructing a collection of negative examples that are sufficiently hard to discriminate against positive queries when their representations are self-trained. Existing contrastive learning methods either maintain a queue of negative samples over minibatches while only a small portion of them are updated in an iteration, or only use the other examples from the current minibatch as negatives. They could not closely track the change of the learned representation over iterations by updating the entire queue as a whole, or discard the useful information from the past minibatches. Alternatively, we present to directly learn a set of negative adversaries playing against the self-trained representation. Two players, the representation network and negative adversaries, are alternately updated to obtain the most challenging negative examples against which the representation of positive queries will be trained to discriminate. We further show that the negative adversaries are updated towards a weighted combination of positive queries by maximizing the adversarial contrastive loss, thereby allowing them to closely track the change of representations over time. Experiment results demonstrate the proposed Adversarial Contrastive (AdCo) model not only achieves superior performances (a top-1 accuracy of 73.2% over 200 epochs and 75.7% over 800 epochs with linear evaluation on ImageNet), but also can be pre-trained more efficiently with much shorter GPU time and fewer epochs. The source code is available at https://github.com/maple-research-lab/AdCo."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "BYOL works even without batch statistics",
                    "abstract": "Bootstrap Your Own Latent (BYOL) is a self-supervised learning approach for image representation. From an augmented view of an image, BYOL trains an online network to predict a target network representation of a different augmented view of the same image. Unlike contrastive methods, BYOL does not explicitly use a repulsion term built from negative pairs in its training objective. Yet, it avoids collapse to a trivial, constant representation. Thus, it has recently been hypothesized that batch normalization (BN) is critical to prevent collapse in BYOL. Indeed, BN flows gradients across batch elements, and could leak information about negative views in the batch, which could act as an implicit negative (contrastive) term. However, we experimentally show that replacing BN with a batch-independent normalization scheme (namely, a combination of group normalization and weight standardization) achieves performance comparable to vanilla BYOL ($73.9\\%$ vs. $74.3\\%$ top-1 accuracy under the linear evaluation protocol on ImageNet with ResNet-$50$). Our finding disproves the hypothesis that the use of batch statistics is a crucial ingredient for BYOL to learn useful representations."
                },
                {
                    "title": "Hard Negative Mixing for Contrastive Learning",
                    "abstract": "Contrastive learning has become a key component of self-supervised learning approaches for computer vision. By learning to embed two augmented versions of the same image close to each other and to push the embeddings of different images apart, one can train highly transferable visual representations. As revealed by recent studies, heavy data augmentation and large sets of negatives are both crucial in learning such representations. At the same time, data mixing strategies either at the image or the feature level improve both supervised and semi-supervised learning by synthesizing novel examples, forcing networks to learn more robust features. In this paper, we argue that an important aspect of contrastive learning, i.e., the effect of hard negatives, has so far been neglected. To get more meaningful negative samples, current top contrastive self-supervised learning approaches either substantially increase the batch sizes, or keep very large memory banks; increasing the memory size, however, leads to diminishing returns in terms of performance. We therefore start by delving deeper into a top-performing framework and show evidence that harder negatives are needed to facilitate better and faster learning. Based on these observations, and motivated by the success of data mixing, we propose hard negative mixing strategies at the feature level, that can be computed on-the-fly with a minimal computational overhead. We exhaustively ablate our approach on linear classification, object detection and instance segmentation and show that employing our hard negative mixing procedure improves the quality of visual representations learned by a state-of-the-art self-supervised learning method."
                },
                {
                    "title": "Parametric Instance Classification for Unsupervised Visual Feature Learning",
                    "abstract": "This paper presents parametric instance classification (PIC) for unsupervised visual feature learning. Unlike the state-of-the-art approaches which do instance discrimination in a dual-branch non-parametric fashion, PIC directly performs a one-branch parametric instance classification, revealing a simple framework similar to supervised classification and without the need to address the information leakage issue. We show that the simple PIC framework can be as effective as the state-of-the-art approaches, i.e. SimCLR and MoCo v2, by adapting several common component settings used in the state-of-the-art approaches. We also propose two novel techniques to further improve effectiveness and practicality of PIC: 1) a sliding-window data scheduler, instead of the previous epoch-based data scheduler, which addresses the extremely infrequent instance visiting issue in PIC and improves the effectiveness; 2) a negative sampling and weight update correction approach to reduce the training time and GPU memory consumption, which also enables application of PIC to almost unlimited training images. We hope that the PIC framework can serve as a simple baseline to facilitate future study."
                },
                {
                    "title": "Unsupervised Learning of Visual Features by Contrasting Cluster Assignments",
                    "abstract": "Unsupervised image representations have significantly reduced the gap with supervised pretraining, notably with the recent achievements of contrastive learning methods. These contrastive methods typically work online and rely on a large number of explicit pairwise feature comparisons, which is computationally challenging. In this paper, we propose an online algorithm, SwAV, that takes advantage of contrastive methods without requiring to compute pairwise comparisons. Specifically, our method simultaneously clusters the data while enforcing consistency between cluster assignments produced for different augmentations (or views) of the same image, instead of comparing features directly as in contrastive learning. Simply put, we use a swapped prediction mechanism where we predict the cluster assignment of a view from the representation of another view. Our method can be trained with large and small batches and can scale to unlimited amounts of data. Compared to previous contrastive methods, our method is more memory efficient since it does not require a large memory bank or a special momentum network. In addition, we also propose a new data augmentation strategy, multi-crop, that uses a mix of views with different resolutions in place of two full-resolution views, without increasing the memory or compute requirements much. We validate our findings by achieving 75.3% top-1 accuracy on ImageNet with ResNet-50, as well as surpassing supervised pretraining on all the considered transfer tasks."
                },
                {
                    "title": "Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning",
                    "abstract": "We introduce Bootstrap Your Own Latent (BYOL), a new approach to self-supervised image representation learning. BYOL relies on two neural networks, referred to as online and target networks, that interact and learn from each other. From an augmented view of an image, we train the online network to predict the target network representation of the same image under a different augmented view. At the same time, we update the target network with a slow-moving average of the online network. While state-of-the art methods rely on negative pairs, BYOL achieves a new state of the art without them. BYOL reaches $74.3\\%$ top-1 classification accuracy on ImageNet using a linear evaluation with a ResNet-50 architecture and $79.6\\%$ with a larger ResNet. We show that BYOL performs on par or better than the current state of the art on both transfer and semi-supervised benchmarks. Our implementation and pretrained models are given on GitHub."
                },
                {
                    "title": "Dataset Condensation with Gradient Matching",
                    "abstract": "Efficient training of deep neural networks is an increasingly important problem in the era of sophisticated architectures and large-scale datasets. This paper proposes a training set synthesis technique, called Dataset Condensation, that learns to produce a small set of informative samples for training deep neural networks from scratch in a small fraction of the required computational cost on the original data while achieving comparable results. We rigorously evaluate its performance in several computer vision benchmarks and show that it significantly outperforms the state-of-the-art methods. Finally we show promising applications of our method in continual learning and domain adaptation."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hypersphere",
                    "abstract": "Contrastive representation learning has been outstandingly successful in practice. In this work, we identify two key properties related to the contrastive loss: (1) alignment (closeness) of features from positive pairs, and (2) uniformity of the induced distribution of the (normalized) features on the hypersphere. We prove that, asymptotically, the contrastive loss optimizes these properties, and analyze their positive effects on downstream tasks. Empirically, we introduce an optimizable metric to quantify each property. Extensive experiments on standard vision and language datasets confirm the strong agreement between both metrics and downstream task performance. Remarkably, directly optimizing for these two metrics leads to representations with comparable or better performance at downstream tasks than contrastive learning. \nProject Page: this https URL \nCode: this https URL , this https URL"
                },
                {
                    "title": "Prototypical Contrastive Learning of Unsupervised Representations",
                    "abstract": "This paper presents Prototypical Contrastive Learning (PCL), an unsupervised representation learning method that addresses the fundamental limitations of instance-wise contrastive learning. PCL not only learns low-level features for the task of instance discrimination, but more importantly, it implicitly encodes semantic structures of the data into the learned embedding space. Specifically, we introduce prototypes as latent variables to help find the maximum-likelihood estimation of the network parameters in an Expectation-Maximization framework. We iteratively perform E-step as finding the distribution of prototypes via clustering and M-step as optimizing the network via contrastive learning. We propose ProtoNCE loss, a generalized version of the InfoNCE loss for contrastive learning, which encourages representations to be closer to their assigned prototypes. PCL achieves state-of-the-art results on multiple unsupervised representation learning benchmarks, with >10% accuracy improvement in low-resource transfer tasks. Code is available at this https URL."
                },
                {
                    "title": "Improved Baselines with Momentum Contrastive Learning",
                    "abstract": "Contrastive unsupervised learning has recently shown encouraging progress, e.g., in Momentum Contrast (MoCo) and SimCLR. In this note, we verify the effectiveness of two of SimCLR's design improvements by implementing them in the MoCo framework. With simple modifications to MoCo---namely, using an MLP projection head and more data augmentation---we establish stronger baselines that outperform SimCLR and do not require large training batches. We hope this will make state-of-the-art unsupervised learning research more accessible. Code will be made public."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Momentum Contrast for Unsupervised Visual Representation Learning",
                    "abstract": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a moving-averaged encoder. This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning. MoCo provides competitive results under the common linear protocol on ImageNet classification. More importantly, the representations learned by MoCo transfer well to downstream tasks. MoCo can outperform its supervised pre-training counterpart in 7 detection/segmentation tasks on PASCAL VOC, COCO, and other datasets, sometimes surpassing it by large margins. This suggests that the gap between unsupervised and supervised representation learning has been largely closed in many vision tasks."
                },
                {
                    "title": "Energy and Policy Considerations for Deep Learning in NLP",
                    "abstract": "Recent progress in hardware and methodology for training neural networks has ushered in a new generation of large networks trained on abundant data. These models have obtained notable gains in accuracy across many NLP tasks. However, these accuracy improvements depend on the availability of exceptionally large computational resources that necessitate similarly substantial energy consumption. As a result these models are costly to train and develop, both financially, due to the cost of hardware and electricity or cloud compute time, and environmentally, due to the carbon footprint required to fuel modern tensor processing hardware. In this paper we bring this issue to the attention of NLP researchers by quantifying the approximate financial and environmental costs of training a variety of recently successful neural network models for NLP. Based on these findings, we propose actionable recommendations to reduce costs and improve equity in NLP research and practice."
                },
                {
                    "title": "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks",
                    "abstract": "Convolutional Neural Networks (ConvNets) are commonly developed at a fixed resource budget, and then scaled up for better accuracy if more resources are available. In this paper, we systematically study model scaling and identify that carefully balancing network depth, width, and resolution can lead to better performance. Based on this observation, we propose a new scaling method that uniformly scales all dimensions of depth/width/resolution using a simple yet highly effective compound coefficient. We demonstrate the effectiveness of this method on scaling up MobileNets and ResNet. \nTo go even further, we use neural architecture search to design a new baseline network and scale it up to obtain a family of models, called EfficientNets, which achieve much better accuracy and efficiency than previous ConvNets. In particular, our EfficientNet-B7 achieves state-of-the-art 84.4% top-1 / 97.1% top-5 accuracy on ImageNet, while being 8.4x smaller and 6.1x faster on inference than the best existing ConvNet. Our EfficientNets also transfer well and achieve state-of-the-art accuracy on CIFAR-100 (91.7%), Flowers (98.8%), and 3 other transfer learning datasets, with an order of magnitude fewer parameters. Source code is at this https URL."
                },
                {
                    "title": "Scaling and Benchmarking Self-Supervised Visual Representation Learning",
                    "abstract": "Self-supervised learning aims to learn representations from the data itself without explicit manual supervision. Existing efforts ignore a crucial aspect of self-supervised learning - the ability to scale to large amount of data because self-supervision requires no manual labels. In this work, we revisit this principle and scale two popular self-supervised approaches to 100 million images. We show that by scaling on various axes (including data size and problem 'hardness'), one can largely match or even exceed the performance of supervised pre-training on a variety of tasks such as object detection, surface normal estimation (3D) and visual navigation using reinforcement learning. Scaling these methods also provides many interesting insights into the limitations of current self-supervised techniques and evaluations. We conclude that current self-supervised methods are not 'hard' enough to take full advantage of large scale data and do not seem to learn effective high level semantic representations. We also introduce an extensive benchmark across 9 different datasets and tasks. We believe that such a benchmark along with comparable evaluation settings is necessary to make meaningful progress. Code is at: https://github.com/facebookresearch/fair_self_supervision_benchmark."
                },
                {
                    "title": "GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism",
                    "abstract": "Scaling up deep neural network capacity has been known as an effective approach to improving model quality for several different machine learning tasks. In many cases, increasing model capacity beyond the memory limit of a single accelerator has required developing special algorithms or infrastructure. These solutions are often architecture-specific and do not transfer to other tasks. To address the need for efficient and task-independent model parallelism, we introduce GPipe, a pipeline parallelism library that allows scaling any network that can be expressed as a sequence of layers. By pipelining different sub-sequences of layers on separate accelerators, GPipe provides the flexibility of scaling a variety of different networks to gigantic sizes efficiently. Moreover, GPipe utilizes a novel batch-splitting pipelining algorithm, resulting in almost linear speedup when a model is partitioned across multiple accelerators. We demonstrate the advantages of GPipe by training large-scale neural networks on two different tasks with distinct network architectures: (i) Image Classification: We train a 557-million-parameter AmoebaNet model and attain a top-1 accuracy of 84.4% on ImageNet-2012, (ii) Multilingual Neural Machine Translation: We train a single 6-billion-parameter, 128-layer Transformer model on a corpus spanning over 100 languages and achieve better quality than all bilingual models."
                },
                {
                    "title": "Don't Use Large Mini-Batches, Use Local SGD",
                    "abstract": "Mini-batch stochastic gradient methods (SGD) are state of the art for distributed training of deep neural networks. Drastic increases in the mini-batch sizes have lead to key efficiency and scalability gains in recent years. However, progress faces a major roadblock, as models trained with large batches often do not generalize well, i.e. they do not show good accuracy on new data. As a remedy, we propose a \\emph{post-local} SGD and show that it significantly improves the generalization performance compared to large-batch training on standard benchmarks while enjoying the same efficiency (time-to-accuracy) and scalability. We further provide an extensive study of the communication efficiency vs. performance trade-offs associated with a host of \\emph{local SGD} variants."
                },
                {
                    "title": "Representation Learning with Contrastive Predictive Coding",
                    "abstract": "While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments."
                },
                {
                    "title": "MobileNetV2: Inverted Residuals and Linear Bottlenecks",
                    "abstract": "In this paper we describe a new mobile architecture, MobileNetV2, that improves the state of the art performance of mobile models on multiple tasks and benchmarks as well as across a spectrum of different model sizes. We also describe efficient ways of applying these mobile models to object detection in a novel framework we call SSDLite. Additionally, we demonstrate how to build mobile semantic segmentation models through a reduced form of DeepLabv3 which we call Mobile DeepLabv3. is based on an inverted residual structure where the shortcut connections are between the thin bottleneck layers. The intermediate expansion layer uses lightweight depthwise convolutions to filter features as a source of non-linearity. Additionally, we find that it is important to remove non-linearities in the narrow layers in order to maintain representational power. We demonstrate that this improves performance and provide an intuition that led to this design. Finally, our approach allows decoupling of the input/output domains from the expressiveness of the transformation, which provides a convenient framework for further analysis. We measure our performance on ImageNet [1] classification, COCO object detection [2], VOC image segmentation [3]. We evaluate the trade-offs between accuracy, and number of operations measured by multiply-adds (MAdd), as well as actual latency, and the number of parameters."
                },
                {
                    "title": "Visualizing the Loss Landscape of Neural Nets",
                    "abstract": "Neural network training relies on our ability to find \"good\" minimizers of highly non-convex loss functions. It is well-known that certain network architecture designs (e.g., skip connections) produce loss functions that train easier, and well-chosen training parameters (batch size, learning rate, optimizer) produce minimizers that generalize better. However, the reasons for these differences, and their effects on the underlying loss landscape, are not well understood. In this paper, we explore the structure of neural loss functions, and the effect of loss landscapes on generalization, using a range of visualization methods. First, we introduce a simple \"filter normalization\" method that helps us visualize loss function curvature and make meaningful side-by-side comparisons between loss functions. Then, using a variety of visualizations, we explore how network architecture affects the loss landscape, and how training parameters affect the shape of minimizers."
                },
                {
                    "title": "A Survey of Model Compression and Acceleration for Deep Neural Networks",
                    "abstract": "Deep neural networks (DNNs) have recently achieved great success in many visual recognition tasks. However, existing deep neural network models are computationally expensive and memory intensive, hindering their deployment in devices with low memory resources or in applications with strict latency requirements. Therefore, a natural thought is to perform model compression and acceleration in deep networks without significantly decreasing the model performance. During the past five years, tremendous progress has been made in this area. In this paper, we review the recent techniques for compacting and accelerating DNN models. In general, these techniques are divided into four categories: parameter pruning and quantization, low-rank factorization, transferred/compact convolutional filters, and knowledge distillation. Methods of parameter pruning and quantization are described first, after that the other techniques are introduced. For each category, we also provide insightful analysis about the performance, related applications, advantages, and drawbacks. Then we go through some very recent successful methods, for example, dynamic capacity networks and stochastic depths networks. After that, we survey the evaluation matrices, the main datasets used for evaluating the model performance, and recent benchmark efforts. Finally, we conclude this paper, discuss remaining the challenges and possible directions for future work."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "A Stochastic Approximation Method",
                    "abstract": "Let M(x) denote the expected value at level x of the response to a certain experiment. M(x) is assumed to be a monotone function of x but is unknown tot he experiment, and it is desire to find the solution x=0 of the equation M(x) = a, where x is a given constant. we give a method for making successive experiments at levels x1, x2,... in such a way that x, will tend to 0 in probability."
                },
                {
                    "title": "CAFE Learning to Condense Dataset by Aligning Features",
                    "abstract": "Dataset condensation aims at reducing the network training effort through condensing a cumbersome training set into a compact synthetic one. State-of-the-art approaches largely rely on learning the synthetic data by matching the gradients between the real and synthetic data batches. Despite the intuitive motivation and promising re-sults, such gradient-based methods, by nature, easily over-\ufb01t to a biased set of samples that produce dominant gradients, and thus lack a global supervision of data distribution. In this paper, we propose a novel scheme to Condense dataset by Aligning FEatures (CAFE), which explicitly attempts to preserve the real-feature distribution as well as the discriminant power of the resulting synthetic set, lending itself to strong generalization capability to various architectures. At the heart of our approach is an effective strategy to align features from the real and synthetic data across various scales, while accounting for the classi\ufb01cation of real samples. Our scheme is further backed up by a novel dynamic bi-level optimization, which adaptively adjusts parameter updates to prevent over-/under-\ufb01tting. We validate the proposed CAFE across various datasets, and demonstrate that it generally outperforms the state of the art: on the SVHN dataset, for example, the performance gain is up to 11%. Extensive experiments and analysis verify the effectiveness and necessity of proposed designs."
                },
                {
                    "title": "Tiny ImageNet Visual Recognition Challenge",
                    "abstract": "In this work, we investigate the effect of convolutional network depth, receptive field size, dropout layers, rectified activation unit type and dataset noise on its accuracy in Tiny-ImageNet Challenge settings. In order to make a thorough evaluation of the cause of the peformance improvement, we start with a basic 5 layer model with 5\u00d75 convolutional receptive fields. We keep increasing network depth or reducing receptive field size, and continue applying modern techniques, such as PReLu and dropout, to the model. Our model achieves excellent performance even compared to state-of-the-art results, with 0.444 final error rate on the test set."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Visualizing Data using t-SNE",
                    "abstract": "We present a new technique called \u201ct-SNE\u201d that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to influence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are significantly better than those produced by the other techniques on almost all of the datasets."
                },
                {
                    "title": "A Mathematical Theory of Communication",
                    "abstract": "This paper opened the new area the information theory. Before this paper, most people believed that the only way to make the error probability of transmission as small as desired is to reduce the data rate (such as a long repetition scheme). However, surprisingly this paper revealed that it does not need to reduce the data rate for achieving that much of small errors. It proved that we can get some positive data rate that has the same small error probability and also there is an upper bound of the data rate, which means we cannot achieve the data rate with any encoding scheme that has small enough error probability over the upper bound."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively bridge the gap between representation learning and data-efficient learning in self-supervised and dataset distillation approaches?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the pressing challenges of human annotation demands and computational burdens associated with large datasets and models. By improving the efficiency of self-supervised learning and dataset distillation, we can enhance model performance while reducing resource consumption. This advancement could lead to more practical applications in various fields, such as natural language processing and computer vision, ultimately fostering innovation and enabling the development of more accessible AI technologies.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities of optimizing learning procedures that rely on self-generated targets, which can be sub-optimal and lead to inefficiencies. Naive approaches may fail due to the intricate relationships between data properties and learning outcomes, as well as the high computational demands of existing distillation methods. Additionally, the need for theoretical guarantees, such as generalization bounds and convergence rates, complicates the development of effective solutions.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often overlooked the interplay between representation learning and data efficiency, leading to gaps in understanding the optimal properties of distilled data. Existing solutions have been limited by their reliance on computationally intensive optimization processes and the necessity of pre-trained models on labeled datasets. Our approach aims to address these limitations by focusing on the theoretical underpinnings of distilled data and exploring data-centric inefficiencies in self-supervised learning, which have not been adequately addressed in prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a detailed analysis of the optimal properties of distilled data for efficient learning, utilizing linear models to derive upper bounds for convergence rates. We will employ a diverse set of datasets to evaluate the performance of our approach, measuring efficiency through metrics such as convergence speed and model accuracy. The expected outcomes include a clearer understanding of the relationship between data properties and learning efficiency, as well as practical guidelines for improving self-supervised and dataset distillation methods."
            }
        },
        "author_data": {
            "2c1f0ed5-52c8-43f1-be80-f30c15af9ee2": {
                "pk": "2c1f0ed5-52c8-43f1-be80-f30c15af9ee2",
                "name": "Peng Sun",
                "collaborators": [],
                "domain": [
                    "Dynamical Systems",
                    "Ergodic Theory",
                    "Topological Entropy",
                    "Measure Theory"
                ],
                "publications": [
                    {
                        "title": "Entropy and Ergodic Measures for Toral Automorphisms",
                        "abstract": "We show that for every linear toral automorphism, especially the non-hyperbolic ones, the entropies of ergodic measures form a dense set on the interval from zero to the topological entropy."
                    },
                    {
                        "title": "Ergodic properties of N-continued fractions",
                        "abstract": "We discuss some ergodic properties of the generalized Gauss transformation $$T_N(x)=\\{\\frac{N}{x}\\}.$$ We generalize a series of results for the regular continued fractions, such as Khinchin's constant and L\\'evy's constant."
                    },
                    {
                        "title": "On the Entropy of Flows with Reparametrized Gluing Orbit Property",
                        "abstract": "We show that a flow or a semiflow with a weaker reparametrized form of gluing orbit property is either minimal or of positive topological entropy."
                    },
                    {
                        "title": "Zero-entropy dynamical systems with the gluing orbit property",
                        "abstract": "Under the assumption of the gluing orbit property, equivalent conditions to having zero topological entropy are investigated. In particular, we show that a dynamical system has the gluing orbit property and zero topological entropy if and only if it is minimal and equicontinuous."
                    },
                    {
                        "title": "Unique ergodicity for zero-entropy dynamical systems with the approximate product property",
                        "abstract": "We show that for every topological dynamical system with the approximate product property, zero topological entropy is equivalent to unique ergodicity. Equivalence of minimality is also proved under a slightly stronger condition. Moreover, we show that unique ergodicity implies the approximate product property if the system has periodic points."
                    },
                    {
                        "title": "Exponential Decay of Lebesgue Numbers",
                        "abstract": "We study the exponential rate of decay of Lebesgue numbers of open covers in topological dynamical systems. We show that topological entropy is bounded by this rate multiplied by dimension. Some corollaries and examples are discussed."
                    },
                    {
                        "title": "Exponential Decay of Expansive Constants",
                        "abstract": "A map $f$ on a compact metric space is expansive if and only if $f^n$ is expansive. We study the exponential rate of decay of the expansive constant of $f^n$. A major result is that this rate times box dimension bounds topological entropy."
                    },
                    {
                        "title": "Zero entropy invariant measures for some skew product diffeomorphisms",
                        "abstract": "In this paper we study some skew product diffeomorphisms with nonuniformly hyperbolic structure along fibers. We show that there is an invariant measure with zero entropy which has atomic conditional measures along fibers."
                    },
                    {
                        "title": "Equilibrium states of intermediate entropies",
                        "abstract": "We explore an approach to the conjecture of Katok on intermediate entropies that based on uniqueness of equilibrium states, provided the entropy function is upper semi-continuous. As an application, we prove Katok's conjecture for Ma\\~n\\'e diffeomorphisms."
                    },
                    {
                        "title": "A Generalization of Gauss-Kuzmin-L\u00e9vy Theorem",
                        "abstract": "We prove a generalized Gauss-Kuzmin-L\\'evy theorem for the $p$-numerated generalized Gauss transformation $$T_p(x)=\\{\\frac{p}{x}\\}.$$ In addition, we give an estimate for the constant that appears in the theorem."
                    },
                    {
                        "title": "A Generalization of the Gauss-Kuzmin-Wirsing constant",
                        "abstract": "We generalize the result of Wirsing on Gauss transformation to the generalized tranformation $T_p(x)=\\{\\cfrac{p}{x}\\}$ for any positive integer $p$. We give an estimate for the generalized Gauss-Kuzmin-Wirsing constant."
                    },
                    {
                        "title": "Denseness of intermediate pressures for systems with the Climenhaga-Thompson structures",
                        "abstract": "We show that systems with the structure introduced by Climenhaga and Thompson have dense intermediate pressures and dense intermediate entropies of ergodic measures. The result applies to the Ma\\~n\\'e diffeomorphisms."
                    },
                    {
                        "title": "Minimality and Gluing Orbit Property",
                        "abstract": "We show that a topological dynamical system is either minimal or have positive topological entropy. Moreover, for equicontinuous systems, we show that topological transitivity, minimality and orbit gluing property are equivalent. These facts reflect the similarity and dissimilarity of gluing orbit property with specification property."
                    },
                    {
                        "title": "Non-dense orbits of systems with approximate product property",
                        "abstract": "We show that for any topological dynamical system with approximate product property, the set of points whose forward orbits do not accumulate to any point in a large set carries full topological pressure."
                    },
                    {
                        "title": "Measures of Intermediate Entropies for Skew Product Diffeomorphisms",
                        "abstract": "In this paper we study a skew product map $F$ with a measure $\\mu$ of positive entropy. We show that if on the fibers the map are $C^{1+\\alpha}$ diffeomorphisms with nonzero Lyapunov exponents, then $F$ has ergodic measures of intermediate entropies. To construct these measures we find a set on which the return map is a skew product with horseshoes along fibers. We can control the average return time and show the maximum entropy of these measures can be arbitrarily close to $h_\\mu(F)$."
                    },
                    {
                        "title": "Ergodic measures of intermediate entropies for dynamical systems with the approximate product property",
                        "abstract": "For a dynamical system satisfying the approximate product property and asymptotically entropy expansiveness, we characterize a delicate structrue of the space of invariant measures: The ergodic measures of intermediate entropies and intermediate pressures are generic in certain subspaces. This proves a conjecture of Katok for a broad class of systems and extends a sequence of known results."
                    }
                ]
            },
            "a04e371f-7b50-4080-9e94-125b83453a4b": {
                "pk": "a04e371f-7b50-4080-9e94-125b83453a4b",
                "name": "Yi Jiang",
                "collaborators": [
                    "Hailiang Liu",
                    "Cheng Du",
                    "Shohei Shimizu",
                    "Yang Su",
                    "Zhong Wang",
                    "Yi Zhang",
                    "Xin Zhou",
                    "Xin Liang"
                ],
                "domain": [
                    "Topology",
                    "Symplectic Geometry",
                    "Causal Inference",
                    "Numerical Analysis"
                ],
                "publications": [
                    {
                        "title": "Regular Circle Actions on 2-connected 7-manifolds",
                        "abstract": "We determine the homeomorphism (resp. diffeomorphism) types of those 2-connected 7-manifolds (resp. smooth 2-connected 7-manifolds) that admit regular circle actions (resp. smooth regular circle actions)."
                    },
                    {
                        "title": "Topology of certain symplectic conifold transitions of $CP^{1}$-bundles",
                        "abstract": "In this paper, we prove the existence of certain symplectic conifold transitions on all $CP^{1}$-bundles over symplectic 4--manifolds, which generalizes Smith, Thomas and Yau's examples of symplectic conifold transitions on trivial $CP^{1}$-bundles over K\\\"ahler surfaces. Our main result is to determine the diffeomorphism types of such symplectic conifold transitions of $CP^{1}$-bundles. In particular, this implies that in the case of trivial $CP^{1}$-bundles over projective surfaces, Smith, Thomas and Yau's examples of symplectic conifold transitions are diffeomorphic to K\\\"ahler 3--folds."
                    },
                    {
                        "title": "An Invariant-region-preserving (IRP) Limiter to DG Methods for Compressible Euler Equations",
                        "abstract": "We introduce an explicit invariant-region-preserving limiter applied to DG methods for compressible Euler equations. The invariant region considered consists of positivity of density and pressure and a maximum principle of a specific entropy. The modified polynomial by the limiter preserves the cell average, lies entirely within the invariant region and does not destroy the high order of accuracy for smooth solutions. Numerical tests are presented to illustrate the properties of the limiter. In particular, the tests on Riemann problems show that the limiter helps to damp the oscillations near discontinuities."
                    },
                    {
                        "title": "Does Financial Literacy Impact Investment Participation and Retirement Planning in Japan?",
                        "abstract": "By employing causal discovery method, the Fast Causal Inference (FCI) model to analyze data from the 2022 \"Financial Literacy Survey,\" we explore the causal relationships between financial literacy and financial activities, specifically investment participation and retirement planning. Our findings indicate that increasing financial literacy may not directly boost engagement in financial investments or retirement planning in Japan, which underscores the necessity for alternative strategies to motivate financial activities among Japanese households. This research offers valuable insights for policymakers focused on improving financial well-being by advancing the use of causal discovery algorithms in understanding financial behaviors."
                    },
                    {
                        "title": "Free circle actions on $(n-1)$-connected $(2n+1)$-manifolds",
                        "abstract": "In this paper, we determine those $(n-1)$-connected $(2n+1)$-manifolds with torsion free homology that admit free circle actions up to almost diffeomorphism, provided that $n\\equiv5,7 \\mod 8$."
                    },
                    {
                        "title": "Prototype matching: children's preference for forming scientific concepts",
                        "abstract": "Inspired by a sample lesson, this paper studies and discusses children's preferences in learning scientific concepts. In a \"Dissolution\" lesson, one of the students took the demonstration experiment of \"carmine dissolves in Water\" demonstrated by the teacher as the prototype to judge whether a new phenomenon belongs to dissolution, instead of analyzing and judging the phenomenon by using the dissolution definition. Therefore, we propose a conjecture that \"prototype matching\" may be a more preferred way for children to learn concepts than thinking through inquiry experiment, analysis, deduction, etc. To this end, we conducted a targeted test on 160 fifth grade students (all of whom had learned this lesson) from a primary school in Beijing, and used goodness of fit test to statistically analyze the results. The results showed that: 1. the Chi square of the general result is 73.865, P<0.001, indicating that children did have obvious prototype preference; 2. We \"tampered\" some of the prototypes, that is, they looked like the prototypes that the teacher had told students, but they were actually wrong. However, the results showed that children still preferred these so-called \"prototypes\" (chi square is 21.823, P<0.001). Conclusion: 1. Children have an obvious preference for \"prototype matching\" in scientific concept learning, which is not only obviously deviated from the current general understanding of science education that emphasizes discovery/inquiry construction, but also points out that there may be a priority relationship among various ways of concept organization (such as definition theory, prototype theory, schema theory, etc.). 2. Children's preference for prototypes seems to be unthinking, and they will not identify the authenticity of prototypes, which is particularly noteworthy in front-line teaching."
                    },
                    {
                        "title": "An invariant-region-preserving limiter for DG schemes to isentropic Euler equations",
                        "abstract": "In this paper, we introduce an invariant-region-preserving (IRP) limiter for the p-system and the corresponding viscous p-system, both of which share the same invariant region. Rigorous analysis is presented to show that for smooth solutions the order of approximation accuracy is not destroyed by the IRP limiter, provided the cell average stays away from the boundary of the invariant region. Moreover, this limiter is explicit, and easy for computer implementation. A generic algorithm incorporating the IRP limiter is presented for high order finite volume type schemes as long as the evolved cell average of the underlying scheme stays strictly within the invariant region. For any high order discontinuous Galerkin (DG) scheme to the p-system, sufficient conditions are obtained for cell averages to stay in the invariant region. For the viscous p-system, we design both second and third order IRP DG schemes. Numerical experiments are provided to test the proven properties of the IRP limiter and the performance of IRP DG schemes."
                    },
                    {
                        "title": "Invariant-region-preserving DG methods for multi-dimensional hyperbolic conservation law systems, with an application to compressible Euler equations",
                        "abstract": "An invariant-region-preserving (IRP) limiter for multi-dimensional hyperbolic conservation law systems is introduced, as long as the system admits a global invariant region which is a convex set in the phase space. It is shown that the order of approximation accuracy is not destroyed by the IRP limiter, provided the cell average is away from the boundary of the convex set. Moreover, this limiter is explicit, and easy for computer implementation. A generic algorithm incorporating the IRP limiter is presented for high order finite volume type schemes. For arbitrarily high order discontinuous Galerkin (DG) schemes to hyperbolic conservation law systems, sufficient conditions are obtained for cell averages to remain in the invariant region provided the projected one-dimensional system shares the same invariant region as the full multi-dimensional hyperbolic system {does}. The general results are then applied to both one and two dimensional compressible Euler equations so to obtain high order IRP DG schemes. Numerical experiments are provided to validate the proven properties of the IRP limiter and the performance of IRP DG schemes for compressible Euler equations."
                    },
                    {
                        "title": "Golay Complementary Sequences of Arbitrary Length and Asymptotic Existence of Hadamard Matrices",
                        "abstract": "In this work, we construct $4$-phase Golay complementary sequence (GCS) set of cardinality $2^{3+\\lceil \\log_2 r \\rceil}$ with arbitrary sequence length $n$, where the $10^{13}$-base expansion of $n$ has $r$ nonzero digits. Specifically, the GCS octets (eight sequences) cover all the lengths no greater than $10^{13}$. Besides, based on the representation theory of signed symmetric group, we construct Hadamard matrices from some special GCS to improve their asymptotic existence: there exist Hadamard matrices of order $2^t m$ for any odd number $m$, where $t = 6\\lfloor \\frac{1}{40}\\log_{2}m\\rfloor + 10$."
                    },
                    {
                        "title": "Linkages among the Foreign Exchange, Stock, and Bond Markets in Japan and the United States",
                        "abstract": "While economic theory explains the linkages among the financial markets of different countries, empirical studies mainly verify the linkages through Granger causality, without considering latent variables or instantaneous effects. Their findings are inconsistent regarding the existence of causal linkages among financial markets, which might be attributed to differences in the focused markets, data periods, and methods applied. Our study adopts causal discovery methods including VAR-LiNGAM and LPCMCI with domain knowledge to explore the linkages among financial markets in Japan and the United States (US) for the post Covid-19 pandemic period under divergent monetary policy directions. The VAR-LiNGAM results reveal that the previous day's US market influences the following day's Japanese market for both stocks and bonds, and the bond markets of the previous day impact the following day's foreign exchange (FX) market directly and the following day's Japanese stock market indirectly. The LPCMCI results indicate the existence of potential latent confounders. Our results demonstrate that VAR-LiNGAM uniquely identifies the directed acyclic graph (DAG), and thus provides informative insight into the causal relationship when the assumptions are considered valid. Our study contributes to a better understanding of the linkages among financial markets in the analyzed data period by supporting the existence of linkages between Japan and the US for the same financial markets and among FX, stock, and bond markets, thus highlighting the importance of leveraging causal discovery methods in the financial domain."
                    },
                    {
                        "title": "Fast Adaptive Flat-histogram Ensemble for Calculating Density of States and Enhanced Sampling in Large Systems",
                        "abstract": "We presented an efficient algorithm, fast adaptive flat-histogram ensemble (FAFE), to estimate the density of states (DOS) and to enhance sampling in large systems. FAFE calculates the means of an arbitrary extensive variable $U$ in generalized ensembles to form points on the curve $\\beta_{s}(U) \\equiv \\frac{\\partial S(U)}{\\partial U}$, the derivative of the logarithmic DOS. Unlike the popular Wang-Landau-like (WLL) methods, FAFE satisfies the detailed-balance condition through out the simulation and automatically generates non-uniform $(\\beta_{i}, U_{i})$ data points to follow the real change rate of $\\beta_{s}(U)$ in different $U$ regions and in different systems. Combined with a $U-$compression transformation, FAFE reduces the required simulation steps from $O(N^{3/2})$ in WLL to $O(N^{1/2})$, where $N$ is the system size. We demonstrate the efficiency of FAFE in Lennard-Jones liquids with several $N$ values. More importantly, we show its abilities in finding and identifying different macroscopic states including meta-stable states in phase co-existing regions."
                    },
                    {
                        "title": "Nonlinear Kalman Filter Using Cramer Rao Bound",
                        "abstract": "This paper studies the optimal state estimation for a dynamic system, whose transfer function can be nonlinear and the input noise can be of arbitrary distribution. Our algorithm differs from the conventional extended Kalman filter (EKF) and the particle filter (PF) in that it estimates not only the state vector but also the Cramer-Rao bound (CRB), which serves as an accuracy indicator. Combining the state estimation, the CRB, and the incoming new measurement, the algorithm updates the state estimation according to the maximum likelihood (ML) criterion. To illustrate the effectiveness of the proposed method for autonomous driving, we apply it to estimate the position and velocity of a vehicle based on the noisy measurements of distance and Doppler offset. Simulation results show that the proposed algorithm can achieve estimation significantly more accurate than the standard EKF and the PF."
                    },
                    {
                        "title": "Constructions of Polyphase Golay Complementary Arrays",
                        "abstract": "Golay complementary matrices (GCM) have recently drawn considerable attentions owing to its potential applications in omnidirectional precoding. In this paper we generalize the GCM to multi-dimensional Golay complementary arrays (GCA) and propose new constructions of GCA pairs and GCA quads. These constructions are facilitated by introducing a set of identities over a commutative ring. We prove that a quaternary GCA pair is feasible if the product of the array sizes in all dimensions is a quaternary Golay number with an additional constraint on the factorization of the product. For the binary GCM quads, we conjecture that the feasible sizes are arbitrary, and verify for sizes within 78 $\\times$ 78 and other less densely distributed sizes. For the quaternary GCM quads, all the positive integers within 1000 can be covered for the size in one dimension."
                    },
                    {
                        "title": "Generalized Step-Chirp Sequences With Flexible Bandwidth",
                        "abstract": "Sequences with low aperiodic autocorrelation sidelobes have been extensively researched in literatures. With sufficiently low integrated sidelobe level (ISL), their power spectrums are asymptotically flat over the whole frequency domain. However, for the beam sweeping in the massive multi-input multi-output (MIMO) broadcast channels, the flat spectrum should be constrained in a passband with tunable bandwidth to achieve the flexible tradeoffs between the beamforming gain and the beam sweeping time. Motivated by this application, we construct a family of sequences termed the generalized step-chirp (GSC) sequence with a closed-form expression, where some parameters can be tuned to adjust the bandwidth flexibly. In addition to the application in beam sweeping, some GSC sequences are closely connected with Mow's unified construction of sequences with perfect periodic autocorrelations, and may have a coarser phase resolution than the Mow sequence while their ISLs are comparable."
                    }
                ]
            },
            "518acfb7-8cbc-4e4f-8f85-1120136c4135": {
                "pk": "518acfb7-8cbc-4e4f-8f85-1120136c4135",
                "name": "Tao Lin",
                "collaborators": [
                    "Yiling Chen",
                    "Xiaoming Li",
                    "Liangze Jiang",
                    "Tie Zhang",
                    "Hu Fu",
                    "Ce Li",
                    "Ruchi Guo"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Federated Learning",
                    "Bayesian Persuasion",
                    "Immersed Finite Element Method"
                ],
                "publications": [
                    {
                        "title": "Deep Learning for IoT",
                        "abstract": "Deep learning and other machine learning approaches are deployed to many systems related to Internet of Things or IoT. However, it faces challenges that adversaries can take loopholes to hack these systems through tampering history data. This paper first presents overall points of adversarial machine learning. Then, we illustrate traditional methods, such as Petri Net cannot solve this new question efficiently. To help IoT data analysis more efficient, we propose a retrieval method based on deep learning (recurrent neural network). Besides, this paper presents a research on data retrieval solution to avoid hacking by adversaries in the fields of adversary machine leaning. It further directs the new approaches in terms of how to implementing this framework in IoT settings based on adversarial deep learning."
                    },
                    {
                        "title": "How Does Independence Help Generalization? Sample Complexity of ERM on Product Distributions",
                        "abstract": "While many classical notions of learnability (e.g., PAC learnability) are distribution-free, utilizing the specific structures of an input distribution may improve learning performance. For example, a product distribution on a multi-dimensional input space has a much simpler structure than a correlated distribution. A recent paper [GHTZ21] shows that the sample complexity of a general learning problem on product distributions is polynomial in the input dimension, which is exponentially smaller than that on correlated distributions. However, the learning algorithm they use is not the standard Empirical Risk Minimization (ERM) algorithm. In this note, we characterize the sample complexity of ERM in a general learning problem on product distributions. We show that, even though product distributions are simpler than correlated distributions, ERM still needs an exponential number of samples to learn on product distributions, instead of a polynomial. This leads to the conclusion that a product distribution by itself does not make a learning problem easier -- an algorithm designed specifically for product distributions is needed."
                    },
                    {
                        "title": "On Clearing Prices in Matching Markets: A Simple Characterization without Duality",
                        "abstract": "Duality of linear programming is a standard approach to the classical weighted maximum matching problem. From an economic perspective, the dual variables can be regarded as prices of products and payoffs of buyers in a two-sided matching market. Traditional duality-based algorithms, e.g., Hungarian, essentially aims at finding a set of prices that clears the market. Under such market-clearing prices, a maximum matching is formed when buyers buy their most preferred products respectively. We study the property of market-clearing prices without the use of duality, showing that: (1) the space of market-clearing prices is convex and closed under element-wise maximum and minimum operations; (2) any market-clearing prices induce all maximum matchings."
                    },
                    {
                        "title": "Deferred Optical Photon simulation for the JUNO experiment",
                        "abstract": "The Jiangmen Underground Neutrino Observatory (JUNO) is designed to determine the neutrino mass ordering and precisely measure oscillation parameters. It is being built in South China at a depth of 700~m underground and comprises a central detector, water Cerenkov detector and top tracker. The central detector is designed to detect anti-neutrinos with an energy resolution of 3\\% at 1~MeV, using a 20 kt liquid scintillator target with 17,612 20-inch PMTs and 25,600 3-inch PMTs. The scintillator provides a light yield of approximately 10,000 photons per MeV. Monte Carlo simulation is a crucial tool for developing an understanding of detector performance, requiring the production of large samples of background processes with optical photons. Simulation of large numbers of optical photons with Geant4 is computationally challenging for both processing time and memory resources. In order to optimize resource usage, a deferred optical photon simulation workflow is proposed and implemented using Geant4 classes. The key idea is to simulate events initially without optical photons, only performing the optical photon simulation when user specified criteria are met.   In this contribution, the design and the implementation of the deferred optical photon simulation will be presented."
                    },
                    {
                        "title": "Test-Time Robust Personalization for Federated Learning",
                        "abstract": "Federated Learning (FL) is a machine learning paradigm where many clients collaboratively learn a shared global model with decentralized training data. Personalized FL additionally adapts the global model to different clients, achieving promising results on consistent local training and test distributions. However, for real-world personalized FL applications, it is crucial to go one step further: robustifying FL models under the evolving local test set during deployment, where various distribution shifts can arise. In this work, we identify the pitfalls of existing works under test-time distribution shifts and propose Federated Test-time Head Ensemble plus tuning(FedTHE+), which personalizes FL models with robustness to various test-time distribution shifts. We illustrate the advancement of FedTHE+ (and its computationally efficient variant FedTHE) over strong competitors, by training various neural architectures (CNN, ResNet, and Transformer) on CIFAR10 andImageNet with various test distributions. Along with this, we build a benchmark for assessing the performance and robustness of personalized FL methods during deployment. Code: https://github.com/LINs-lab/FedTHE."
                    },
                    {
                        "title": "A Stable Weak Galerkin Finite Element Method for Stokes Problem",
                        "abstract": "We study the weak Galerkin finite element method for Stokes problem. A new weak Galerkin finite element velocity-pressure space pair is presented which satisfies the discrete inf-sup condition. Based on this space pair, we establish a stable weak Galerkin approximation scheme without adding any stability term or penalty term. Then, we further derive the optimal error estimates for velocity and pressure approximations, respectively. Numerical experiments are provided to illustrate the theoretical analysis."
                    },
                    {
                        "title": "Learning Utilities and Equilibria in Non-Truthful Auctions",
                        "abstract": "In non-truthful auctions, agents' utility for a strategy depends on the strategies of the opponents and also the prior distribution over their private types; the set of Bayes Nash equilibria generally has an intricate dependence on the prior. Using the First Price Auction as our main demonstrating example, we show that $\\tilde O(n / \\epsilon^2)$ samples from the prior with $n$ agents suffice for an algorithm to learn the interim utilities for all monotone bidding strategies. As a consequence, this number of samples suffice for learning all approximate equilibria. We give almost matching (up to polylog factors) lower bound on the sample complexity for learning utilities. We also consider a setting where agents must pay a search cost to discover their own types. Drawing on a connection between this setting and the first price auction, discovered recently by Kleinberg et al. (2016), we show that $\\tilde O(n / \\epsilon^2)$ samples suffice for utilities and equilibria to be estimated in a near welfare-optimal descending auction in this setting. En route, we improve the sample complexity bound, recently obtained by Guo et al. (2021), for the Pandora's Box problem, which is a classical model for sequential consumer search."
                    },
                    {
                        "title": "Sample Complexity of Forecast Aggregation",
                        "abstract": "We consider a Bayesian forecast aggregation model where $n$ experts, after observing private signals about an unknown binary event, report their posterior beliefs about the event to a principal, who then aggregates the reports into a single prediction for the event. The signals of the experts and the outcome of the event follow a joint distribution that is unknown to the principal, but the principal has access to i.i.d. \"samples\" from the distribution, where each sample is a tuple of the experts' reports (not signals) and the realization of the event. Using these samples, the principal aims to find an $\\varepsilon$-approximately optimal aggregator, where optimality is measured in terms of the expected squared distance between the aggregated prediction and the realization of the event. We show that the sample complexity of this problem is at least $\\tilde \\Omega(m^{n-2} / \\varepsilon)$ for arbitrary discrete distributions, where $m$ is the size of each expert's signal space. This sample complexity grows exponentially in the number of experts $n$. But, if the experts' signals are independent conditioned on the realization of the event, then the sample complexity is significantly reduced, to $\\tilde O(1 / \\varepsilon^2)$, which does not depend on $n$. Our results can be generalized to non-binary events. The proof of our results uses a reduction from the distribution learning problem and reveals the fact that forecast aggregation is almost as difficult as distribution learning."
                    },
                    {
                        "title": "Modern Software Development for JUNO offline software",
                        "abstract": "The Jiangmen Underground Neutrino Observatory (JUNO), under construction in South China, primarily aims to determine the neutrino mass hierarchy and to precise measure the neutrino oscillation parameters. The data-taking is expected to start in 2024 and the detector plans to run for more than 20 years. The development of the JUNO offline software (JUNOSW) started in 2012, and it is quite challenging to maintain the JUNOSW for such a long time. In the last ten years, tools such as Subversion, Trac, and CMT had been adopted for software development. However, new stringent requirements came out, such as how to reduce the building time for the whole project, how to deploy offline algorithms to an online environment, and how to improve the code quality with code review and continuous integration. To meet the further requirements of software development, modern development tools are evaluated for JUNOSW, such as Git, GitLab, CMake, Docker, and Kubernetes. This contribution will present the software development system based on these modern tools for JUNOSW and the functionalities achieved: CMake macros are developed to simplify the build instructions for users; CMake generator expressions are used to control the build flags for the online and offline environments; a tool named git-junoenv is developed to help users partially checkout and build the software; a script is used to build and deploy the software on the CVMFS server; a Docker image with CVMFS client installed is created for continuous integration; a GitLab agent is set up to manage GitLab runners in Kubernetes with all the configurations in a GitLab repository."
                    },
                    {
                        "title": "Generalized Principal-Agent Problem with a Learning Agent",
                        "abstract": "Generalized principal-agent problems, including Stackelberg games, contract design, and Bayesian persuasion, are a class of economic problems where an agent best responds to a principal's committed strategy. We study repeated generalized principal-agent problems under the assumption that the principal does not have commitment power and the agent uses algorithms to learn to respond to the principal. We reduce this problem to a one-shot generalized principal-agent problem with an approximately-best-responding agent. Using this reduction, we show that: (1) if the agent uses contextual no-regret learning algorithms, then the principal can guarantee a utility that is at least the principal's optimal utility in the classic non-learning model minus the square root of the agent's regret; (2) if the agent uses contextual no-swap-regret learning algorithms, then the principal cannot obtain any utility more than the optimal utility in the non-learning model plus the agent's swap regret. But (3) if the agent uses mean-based learning algorithms (which can be no-regret but not no-swap-regret), then the principal can do significantly better than the non-learning model. These general results not only refine previous results in Stackelberg games and contract design with learning agents but also lead to new results for Bayesian persuasion with a learning agent."
                    },
                    {
                        "title": "Information Design with Unknown Prior",
                        "abstract": "Classical information design models (e.g., Bayesian persuasion and cheap talk) require players to have perfect knowledge of the prior distribution of the state of the world. Our paper studies repeated persuasion problems in which the information designer does not know the prior. The information designer learns to design signaling schemes from repeated interactions with the receiver. We design learning algorithms for the information designer to achieve no regret compared to using the optimal signaling scheme with known prior, under two models of the receiver's decision-making. (1) The first model assumes that the receiver knows the prior and can perform posterior update and best respond to signals. In this model, we design a learning algorithm for the information designer with $O(\\log T)$ regret in the general case, and another algorithm with $\\Theta(\\log \\log T)$ regret in the case where the receiver has only two actions. (2) The second model assumes that the receiver does not know the prior and employs a no-regret learning algorithm to take actions. We show that the information designer can achieve regret $O(\\sqrt{\\mathrm{rReg}(T) T})$, where $\\mathrm{rReg}(T)=o(T)$ is an upper bound on the receiver's learning regret. Our work thus provides a learning foundation for the problem of information design with unknown prior."
                    },
                    {
                        "title": "Persuading a Behavioral Agent: Approximately Best Responding and Learning",
                        "abstract": "The classic Bayesian persuasion model assumes a Bayesian and best-responding receiver. We study a relaxation of the Bayesian persuasion model where the receiver can approximately best respond to the sender's signaling scheme. We show that, under natural assumptions, (1) the sender can find a signaling scheme that guarantees itself an expected utility almost as good as its optimal utility in the classic model, no matter what approximately best-responding strategy the receiver uses; (2) on the other hand, there is no signaling scheme that gives the sender much more utility than its optimal utility in the classic model, even if the receiver uses the approximately best-responding strategy that is best for the sender. Together, (1) and (2) imply that the approximately best-responding behavior of the receiver does not affect the sender's maximal achievable utility a lot in the Bayesian persuasion problem. The proofs of both results rely on the idea of robustification of a Bayesian persuasion scheme: given a pair of the sender's signaling scheme and the receiver's strategy, we can construct another signaling scheme such that the receiver prefers to use that strategy in the new scheme more than in the original scheme, and the two schemes give the sender similar utilities. As an application of our main result (1), we show that, in a repeated Bayesian persuasion model where the receiver learns to respond to the sender by some algorithms, the sender can do almost as well as in the classic model. Interestingly, unlike (2), with a learning receiver the sender can sometimes do much better than in the classic model."
                    },
                    {
                        "title": "A Group of Immersed Finite Element Spaces For Elliptic Interface Problems",
                        "abstract": "We present a unified framework for developing and analysing immersed finite element (IFE) spaces for solving typical elliptic interface problems with interface independent meshes. This framework allows us to construct a group of new IFE spaces with either linear, or bilinear, or the rotated-Q1 polynomials. Functions in these IFE spaces are locally piecewise polynomials defined according to the sub-elements formed by the interface itself instead of its line approximation. We show that the unisolvence for these IFE spaces follows from the invertibility of the Sherman-Morrison matrix. A group of estimates and identities are established for the interface geometry and shape functions that are applicable to all of these IFE spaces. Most importantly, these fundamental preparations enable us to develop a unified multipoint Taylor expansion procedure for proving that these IFE spaces have the expected optimal approximation capability according to the involved polynomials."
                    }
                ]
            }
        }
    },
    "2402.13728": {
        "paper_data": {
            "title": "Average gradient outer product as a mechanism for deep neural collapse",
            "url": "http://arxiv.org/abs/2402.13728v5",
            "arxiv_id": "2402.13728",
            "authors": [
                "Daniel Beaglehole",
                "Peter S\u00faken\u00edk",
                "Marco Mondelli",
                "Mikhail Belkin"
            ],
            "abstract": "Deep Neural Collapse (DNC) refers to the surprisingly rigid structure of the data representations in the final layers of Deep Neural Networks (DNNs). Though the phenomenon has been measured in a variety of settings, its emergence is typically explained via data-agnostic approaches, such as the unconstrained features model. In this work, we introduce a data-dependent setting where DNC forms due to feature learning through the average gradient outer product (AGOP). The AGOP is defined with respect to a learned predictor and is equal to the uncentered covariance matrix of its input-output gradients averaged over the training dataset. The Deep Recursive Feature Machine (Deep RFM) is a method that constructs a neural network by iteratively mapping the data with the AGOP and applying an untrained random feature map. We demonstrate empirically that DNC occurs in Deep RFM across standard settings as a consequence of the projection with the AGOP matrix computed at each layer. Further, we theoretically explain DNC in Deep RFM in an asymptotic setting and as a result of kernel learning. We then provide evidence that this mechanism holds for neural networks more generally. In particular, we show that the right singular vectors and values of the weights can be responsible for the majority of within-class variability collapse for DNNs trained in the feature learning regime. As observed in recent work, this singular structure is highly correlated with that of the AGOP.",
            "introduction": "   1 Introduction  How Deep Neural Networks (DNNs) learn a transformation of the data to form a prediction and what are the properties of this transformation constitute fundamental questions in the theory of deep learning. A promising avenue to understand the hidden mechanisms of DNNs is Neural Collapse (NC) (Papyan et\u00a0al., 2020). NC is a widely observed structural property of overparametrized DNNs, occurring in the terminal phase of gradient descent training on classification tasks. This property is linked with the performance of DNNs, such as generalization and robustness (Papyan et\u00a0al., 2020; Su et\u00a0al., 2023). NC is defined by four properties that describe the geometry of feature representations of the training data in the last layer. Of these, in this paper we focus on the following two, because they are relevant in the context of deeper layers. The within-class variability collapse (NC1) states that feature vectors of the training samples within a single class collapse to the common class-mean. The orthogonality or simplex equiangular tight frame property (NC2) states that these class-means form an orthogonal or simplex equiangular tight frame. While initially observed in just the final hidden layer, these properties were measured for the intermediate layers of DNNs as well (Rangamani et\u00a0al., 2023; He and Su, 2022), indicating that NC progresses depth-wise. This has lead to the study of Deep Neural Collapse (DNC), a direct depth-wise extension of standard NC (S\u00faken\u00edk et\u00a0al., 2023).   The theoretical explanations of the formation of (D)NC have mostly relied on a data-agnostic model, known as unconstrained features model (UFM) (Mixon et\u00a0al., 2020; Lu and Steinerberger, 2022). This assumes that DNNs are infinitely expressive and, thus, optimizes for the feature vectors in the last layer directly. The deep UFM (DUFM) was then introduced to account for more layers (S\u00faken\u00edk et\u00a0al., 2023; Tirer and Bruna, 2022). The (D)UFM has been identified as a useful analogy for neural network collapse, as (D)NC emerges naturally in this setting for a variety of loss functions and training regimes (see Section\u00a02). However, (D)UFM discards the training data and most of the network completely, thus ignoring also the role of learning in DNC formation. This is a serious gap to fully understand the formation of DNC in the context of the full DNN pipeline.   In this paper, we introduce a setting where DNC forms through a learning algorithm that is highly dependent on the training data and predictor - iterated linear mapping onto the average gradient outer product (AGOP). The AGOP is the uncentered covariance matrix of the input-output gradients of a predictor averaged over its training data. Recent work has utilized this object to understand various surprising phenomena in neural networks including grokking, lottery tickets, simplicity bias, and adversarial examples (Radhakrishnan et\u00a0al., 2024a). Additional work incorporated layer-wise linear transformation with the AGOP into kernel machines, to model the deep feature learning mechanism of neural networks (Beaglehole et\u00a0al., 2023). The output of their backpropagation-free method, Deep Recursive Feature Machines (Deep RFM), is a standard neural network at i.i.d. initialization, where each random weight matrix is additionally right-multiplied by the AGOP of a kernel machine trained on the input",
            "references": [
                {
                    "title": "Neural Collapse versus Low-rank Bias: Is Deep Neural Collapse Really Optimal?",
                    "abstract": "Deep neural networks (DNNs) exhibit a surprising structure in their final layer known as neural collapse (NC), and a growing body of works has currently investigated the propagation of neural collapse to earlier layers of DNNs -- a phenomenon called deep neural collapse (DNC). However, existing theoretical results are restricted to special cases: linear models, only two layers or binary classification. In contrast, we focus on non-linear models of arbitrary depth in multi-class classification and reveal a surprising qualitative shift. As soon as we go beyond two layers or two classes, DNC stops being optimal for the deep unconstrained features model (DUFM) -- the standard theoretical framework for the analysis of collapse. The main culprit is a low-rank bias of multi-layer regularization schemes: this bias leads to optimal solutions of even lower rank than the neural collapse. We support our theoretical findings with experiments on both DUFM and real data, which show the emergence of the low-rank structure in the solution found by gradient descent."
                },
                {
                    "title": "Mechanism for feature learning in neural networks and backpropagation-free machine learning models",
                    "abstract": "Understanding how neural networks learn features, or relevant patterns in data, for prediction is necessary for their reliable use in technological and scientific applications. In this work, we presented a unifying mathematical mechanism, known as average gradient outer product (AGOP), that characterized feature learning in neural networks. We provided empirical evidence that AGOP captured features learned by various neural network architectures, including transformer-based language models, convolutional networks, multilayer perceptrons, and recurrent neural networks. Moreover, we demonstrated that AGOP, which is backpropagation-free, enabled feature learning in machine learning models, such as kernel machines, that a priori could not identify task-specific features. Overall, we established a fundamental mechanism that captured feature learning in neural networks and enabled feature learning in general machine learning models. Editor\u2019s summary How neural networks learn problem-specific features (patterns in data) and how features emerge through training remain major unsolved problems in machine learning. Understanding this mechanism provides the opportunity to design networks with improved reliability and the model transparency needed for various practical applications. Radhakrishnan et al. proposed the deep neural feature ansatz, which states that neural feature learning occurs by up-weighting the features that are most influential on model output, a process that was formulated mathematically in terms of the average gradient outer product and was supported by numerical experiments and theoretical results. The presented mechanism provides a backpropagation-free approach for feature learning in various machine learning models, including those that previously had no such capabilities. \u2014Yury Suleymanov A unified mechanism without backpropagation characterizes feature learning in neural network architectures."
                },
                {
                    "title": "Linear Recursive Feature Machines provably recover low-rank matrices",
                    "abstract": "A fundamental problem in machine learning is to understand how neural networks make accurate predictions, while seemingly bypassing the curse of dimensionality. A possible explanation is that common training algorithms for neural networks implicitly perform dimensionality reduction - a process called feature learning. Recent work posited that the effects of feature learning can be elicited from a classical statistical estimator called the average gradient outer product (AGOP). The authors proposed Recursive Feature Machines (RFMs) as an algorithm that explicitly performs feature learning by alternating between (1) reweighting the feature vectors by the AGOP and (2) learning the prediction function in the transformed space. In this work, we develop the first theoretical guarantees for how RFM performs dimensionality reduction by focusing on the class of overparametrized problems arising in sparse linear regression and low-rank matrix recovery. Specifically, we show that RFM restricted to linear models (lin-RFM) generalizes the well-studied Iteratively Reweighted Least Squares (IRLS) algorithm. Our results shed light on the connection between feature learning in neural networks and classical sparse recovery algorithms. In addition, we provide an implementation of lin-RFM that scales to matrices with millions of missing entries. Our implementation is faster than the standard IRLS algorithm as it is SVD-free. It also outperforms deep linear networks for sparse linear regression and low-rank matrix completion."
                },
                {
                    "title": "Efficient Estimation of the Central Mean Subspace via Smoothed Gradient Outer Products",
                    "abstract": "We consider the problem of sufficient dimension reduction (SDR) for multi-index models. The estimators of the central mean subspace in prior works either have slow (non-parametric) convergence rates, or rely on stringent distributional conditions (e.g., the covariate distribution $P_{\\mathbf{X}}$ being elliptical symmetric). In this paper, we show that a fast parametric convergence rate of form $C_d \\cdot n^{-1/2}$ is achievable via estimating the \\emph{expected smoothed gradient outer product}, for a general class of distribution $P_{\\mathbf{X}}$ admitting Gaussian or heavier distributions. When the link function is a polynomial with a degree of at most $r$ and $P_{\\mathbf{X}}$ is the standard Gaussian, we show that the prefactor depends on the ambient dimension $d$ as $C_d \\propto d^r$."
                },
                {
                    "title": "On the Robustness of Neural Collapse and the Neural Collapse of Robustness",
                    "abstract": "Neural Collapse refers to the curious phenomenon in the end of training of a neural network, where feature vectors and classification weights converge to a very simple geometrical arrangement (a simplex). While it has been observed empirically in various cases and has been theoretically motivated, its connection with crucial properties of neural networks, like their generalization and robustness, remains unclear. In this work, we study the stability properties of these simplices. We find that the simplex structure disappears under small adversarial attacks, and that perturbed examples\"leap\"between simplex vertices. We further analyze the geometry of networks that are optimized to be robust against adversarial perturbations of the input, and find that Neural Collapse is a pervasive phenomenon in these cases as well, with clean and perturbed representations forming aligned simplices, and giving rise to a robust simple nearest-neighbor classifier. By studying the propagation of the amount of collapse inside the network, we identify novel properties of both robust and non-robust machine learning models, and show that earlier, unlike later layers maintain reliable simplices on perturbed data."
                },
                {
                    "title": "Kernel Learning in Ridge Regression\"Automatically\"Yields Exact Low Rank Solution",
                    "abstract": "We consider kernels of the form $(x,x') \\mapsto \\phi(\\|x-x'\\|^2_\\Sigma)$ parametrized by $\\Sigma$. For such kernels, we study a variant of the kernel ridge regression problem which simultaneously optimizes the prediction function and the parameter $\\Sigma$ of the reproducing kernel Hilbert space. The eigenspace of the $\\Sigma$ learned from this kernel ridge regression problem can inform us which directions in covariate space are important for prediction. Assuming that the covariates have nonzero explanatory power for the response only through a low dimensional subspace (central mean subspace), we find that the global minimizer of the finite sample kernel learning objective is also low rank with high probability. More precisely, the rank of the minimizing $\\Sigma$ is with high probability bounded by the dimension of the central mean subspace. This phenomenon is interesting because the low rankness property is achieved without using any explicit regularization of $\\Sigma$, e.g., nuclear norm penalization. Our theory makes correspondence between the observed phenomenon and the notion of low rank set identifiability from the optimization literature. The low rankness property of the finite sample solutions exists because the population kernel learning objective grows\"sharply\"when moving away from its minimizers in any direction perpendicular to the central mean subspace."
                },
                {
                    "title": "How Far Pre-trained Models Are from Neural Collapse on the Target Dataset Informs their Transferability",
                    "abstract": "This paper focuses on model transferability estimation, i.e., assessing the performance of pre-trained models on a downstream task without performing fine-tuning. Motivated by neural collapse (NC) [25] that reveals particular feature geometry at the terminal stage of training, we consider model transferability as how far the target activations obtained by pre-trained models are from their hypothetical state in the terminal phase of the model fine-tuned on the target domain. We propose a metric that measures this proximity based on three phenomena of NC: within-class variability collapse, simplex encoded label interpolation geometry structure is formed, and the nearest center classifier becomes optimal on training data. Through experiments on 11 datasets, we confirm none of the three NC proxies are dispensable, which allows us to obtain very competitive transferability estimation accuracy with approximately 10\u00d7 wall-clock time speed up compared to state-of-the-art approaches."
                },
                {
                    "title": "Neural Collapse for Unconstrained Feature Model under Cross-entropy Loss with Imbalanced Data",
                    "abstract": "Recent years have witnessed the huge success of deep neural networks (DNNs) in various tasks of computer vision and text processing. Interestingly, these DNNs with massive number of parameters share similar structural properties on their feature representation and last-layer classifier at terminal phase of training (TPT). Specifically, if the training data are balanced (each class shares the same number of samples), it is observed that the feature vectors of samples from the same class converge to their corresponding in-class mean features and their pairwise angles are the same. This fascinating phenomenon is known as Neural Collapse (N C), first termed by Papyan, Han, and Donoho in 2019. Many recent works manage to theoretically explain this phenomenon by adopting so-called unconstrained feature model (UFM). In this paper, we study the extension of N C phenomenon to the imbalanced data under cross-entropy loss function in the context of unconstrained feature model. Our contribution is multi-fold compared with the state-of-the-art results: (a) we show that the feature vectors exhibit collapse phenomenon, i.e., the features within the same class collapse to the same mean vector; (b) the mean feature vectors no longer form an equiangular tight frame. Instead, their pairwise angles depend on the sample size; (c) we also precisely characterize the sharp threshold on which the minority collapse (the feature vectors of the minority groups collapse to one single vector) will take place; (d) finally, we argue that the effect of the imbalance in datasize diminishes as the sample size grows. Our results provide a complete picture of the N C under the cross-entropy loss for the imbalanced data. Numerical experiments confirm our theoretical analysis."
                },
                {
                    "title": "Mechanism of feature learning in convolutional neural networks",
                    "abstract": "Understanding the mechanism of how convolutional neural networks learn features from image data is a fundamental problem in machine learning and computer vision. In this work, we identify such a mechanism. We posit the Convolutional Neural Feature Ansatz, which states that covariances of filters in any convolutional layer are proportional to the average gradient outer product (AGOP) taken with respect to patches of the input to that layer. We present extensive empirical evidence for our ansatz, including identifying high correlation between covariances of filters and patch-based AGOPs for convolutional layers in standard neural architectures, such as AlexNet, VGG, and ResNets pre-trained on ImageNet. We also provide supporting theoretical evidence. We then demonstrate the generality of our result by using the patch-based AGOP to enable deep feature learning in convolutional kernel machines. We refer to the resulting algorithm as (Deep) ConvRFM and show that our algorithm recovers similar features to deep convolutional networks including the notable emergence of edge detectors. Moreover, we find that Deep ConvRFM overcomes previously identified limitations of convolutional kernels, such as their inability to adapt to local signals in images and, as a result, leads to sizable performance improvement over fixed convolutional kernels."
                },
                {
                    "title": "Neural Collapse in the Intermediate Hidden Layers of Classification Neural Networks",
                    "abstract": "Neural Collapse (NC) gives a precise description of the representations of classes in the final hidden layer of classification neural networks. This description provides insights into how these networks learn features and generalize well when trained past zero training error. However, to date, (NC) has only been studied in the final layer of these networks. In the present paper, we provide the first comprehensive empirical analysis of the emergence of (NC) in the intermediate hidden layers of these classifiers. We examine a variety of network architectures, activations, and datasets, and demonstrate that some degree of (NC) emerges in most of the intermediate hidden layers of the network, where the degree of collapse in any given layer is typically positively correlated with the depth of that layer in the neural network. Moreover, we remark that: (1) almost all of the reduction in intra-class variance in the samples occurs in the shallower layers of the networks, (2) the angular separation between class means increases consistently with hidden layer depth, and (3) simple datasets require only the shallower layers of the networks to fully learn them, whereas more difficult ones require the entire network. Ultimately, these results provide granular insights into the structural propagation of features through classification neural networks."
                },
                {
                    "title": "Catapults in SGD: spikes in the training loss and their impact on generalization through feature learning",
                    "abstract": "In this paper, we first present an explanation regarding the common occurrence of spikes in the training loss when neural networks are trained with stochastic gradient descent (SGD). We provide evidence that the spikes in the training loss of SGD are\"catapults\", an optimization phenomenon originally observed in GD with large learning rates in [Lewkowycz et al. 2020]. We empirically show that these catapults occur in a low-dimensional subspace spanned by the top eigenvectors of the tangent kernel, for both GD and SGD. Second, we posit an explanation for how catapults lead to better generalization by demonstrating that catapults promote feature learning by increasing alignment with the Average Gradient Outer Product (AGOP) of the true predictor. Furthermore, we demonstrate that a smaller batch size in SGD induces a larger number of catapults, thereby improving AGOP alignment and test performance."
                },
                {
                    "title": "Neural (Tangent Kernel) Collapse",
                    "abstract": "This work bridges two important concepts: the Neural Tangent Kernel (NTK), which captures the evolution of deep neural networks (DNNs) during training, and the Neural Collapse (NC) phenomenon, which refers to the emergence of symmetry and structure in the last-layer features of well-trained classification DNNs. We adopt the natural assumption that the empirical NTK develops a block structure aligned with the class labels, i.e., samples within the same class have stronger correlations than samples from different classes. Under this assumption, we derive the dynamics of DNNs trained with mean squared (MSE) loss and break them into interpretable phases. Moreover, we identify an invariant that captures the essence of the dynamics, and use it to prove the emergence of NC in DNNs with block-structured NTK. We provide large-scale numerical experiments on three common DNN architectures and three benchmark datasets to support our theory."
                },
                {
                    "title": "Deep Neural Collapse Is Provably Optimal for the Deep Unconstrained Features Model",
                    "abstract": "Neural collapse (NC) refers to the surprising structure of the last layer of deep neural networks in the terminal phase of gradient descent training. Recently, an increasing amount of experimental evidence has pointed to the propagation of NC to earlier layers of neural networks. However, while the NC in the last layer is well studied theoretically, much less is known about its multi-layered counterpart - deep neural collapse (DNC). In particular, existing work focuses either on linear layers or only on the last two layers at the price of an extra assumption. Our paper fills this gap by generalizing the established analytical framework for NC - the unconstrained features model - to multiple non-linear layers. Our key technical contribution is to show that, in a deep unconstrained features model, the unique global optimum for binary classification exhibits all the properties typical of DNC. This explains the existing experimental evidence of DNC. We also empirically show that (i) by optimizing deep unconstrained features models via gradient descent, the resulting solution agrees well with our theory, and (ii) trained networks recover the unconstrained features suitable for the occurrence of DNC, thus supporting the validity of this modeling principle."
                },
                {
                    "title": "ReLU soothes the NTK condition number and accelerates optimization for wide neural networks",
                    "abstract": "Rectified linear unit (ReLU), as a non-linear activation function, is well known to improve the expressivity of neural networks such that any continuous function can be approximated to arbitrary precision by a sufficiently wide neural network. In this work, we present another interesting and important feature of ReLU activation function. We show that ReLU leads to: {\\it better separation} for similar data, and {\\it better conditioning} of neural tangent kernel (NTK), which are closely related. Comparing with linear neural networks, we show that a ReLU activated wide neural network at random initialization has a larger angle separation for similar data in the feature space of model gradient, and has a smaller condition number for NTK. Note that, for a linear neural network, the data separation and NTK condition number always remain the same as in the case of a linear model. Furthermore, we show that a deeper ReLU network (i.e., with more ReLU activation operations), has a smaller NTK condition number than a shallower one. Our results imply that ReLU activation, as well as the depth of ReLU network, helps improve the gradient descent convergence rate, which is closely related to the NTK condition number."
                },
                {
                    "title": "Dynamics in Deep Classifiers Trained with the Square Loss: Normalization, Low Rank, Neural Collapse, and Generalization Bounds",
                    "abstract": "We overview several properties\u2014old and new\u2014of training overparameterized deep networks under the square loss. We first consider a model of the dynamics of gradient flow under the square loss in deep homogeneous rectified linear unit networks. We study the convergence to a solution with the absolute minimum \u03c1, which is the product of the Frobenius norms of each layer weight matrix, when normalization by Lagrange multipliers is used together with weight decay under different forms of gradient descent. A main property of the minimizers that bound their expected error for a specific network architecture is \u03c1. In particular, we derive novel norm-based bounds for convolutional layers that are orders of magnitude better than classical bounds for dense networks. Next, we prove that quasi-interpolating solutions obtained by stochastic gradient descent in the presence of weight decay have a bias toward low-rank weight matrices, which should improve generalization. The same analysis predicts the existence of an inherent stochastic gradient descent noise for deep networks. In both cases, we verify our predictions experimentally. We then predict neural collapse and its properties without any specific assumption\u2014unlike other published proofs. Our analysis supports the idea that the advantage of deep networks relative to other classifiers is greater for problems that are appropriate for sparse deep architectures such as convolutional neural networks. The reason is that compositionally sparse target functions can be approximated well by \u201csparse\u201d deep networks without incurring in the curse of dimensionality."
                },
                {
                    "title": "Neural Collapse in Deep Linear Networks: From Balanced to Imbalanced Data",
                    "abstract": "Modern deep neural networks have achieved impressive performance on tasks from image classification to natural language processing. Surprisingly, these complex systems with massive amounts of parameters exhibit the same structural properties in their last-layer features and classifiers across canonical datasets when training until convergence. In particular, it has been observed that the last-layer features collapse to their class-means, and those class-means are the vertices of a simplex Equiangular Tight Frame (ETF). This phenomenon is known as Neural Collapse (NC). Recent papers have theoretically shown that NC emerges in the global minimizers of training problems with the simplified\"unconstrained feature model\". In this context, we take a step further and prove the NC occurrences in deep linear networks for the popular mean squared error (MSE) and cross entropy (CE) losses, showing that global solutions exhibit NC properties across the linear layers. Furthermore, we extend our study to imbalanced data for MSE loss and present the first geometric analysis of NC under bias-free setting. Our results demonstrate the convergence of the last-layer features and classifiers to a geometry consisting of orthogonal vectors, whose lengths depend on the amount of data in their corresponding classes. Finally, we empirically validate our theoretical analyses on synthetic and practical network architectures with both balanced and imbalanced scenarios."
                },
                {
                    "title": "A law of data separation in deep learning",
                    "abstract": "Significance The practice of deep learning has long been shrouded in mystery, leading many to believe that the inner workings of these black-box models are chaotic during training. In this paper, we challenge this belief by presenting a simple and approximate law that deep neural networks follow when processing data in the intermediate layers. This empirical law is observed in a class of modern network architectures for vision tasks, and its emergence is shown to bring important benefits for the trained models. The significance of this law is that it allows for a perspective that provides useful insights into the practice of deep learning."
                },
                {
                    "title": "Perturbation Analysis of Neural Collapse",
                    "abstract": "Training deep neural networks for classification often includes minimizing the training loss beyond the zero training error point. In this phase of training, a\"neural collapse\"behavior has been observed: the variability of features (outputs of the penultimate layer) of within-class samples decreases and the mean features of different classes approach a certain tight frame structure. Recent works analyze this behavior via idealized unconstrained features models where all the minimizers exhibit exact collapse. However, with practical networks and datasets, the features typically do not reach exact collapse, e.g., because deep layers cannot arbitrarily modify intermediate features that are far from being collapsed. In this paper, we propose a richer model that can capture this phenomenon by forcing the features to stay in the vicinity of a predefined features matrix (e.g., intermediate features). We explore the model in the small vicinity case via perturbation analysis and establish results that cannot be obtained by the previously studied models. For example, we prove reduction in the within-class variability of the optimized features compared to the predefined input features (via analyzing gradient flow on the\"central-path\"with minimal assumptions), analyze the minimizers in the near-collapse regime, and provide insights on the effect of regularization hyperparameters on the closeness to collapse. We support our theory with experiments in practical deep learning settings."
                },
                {
                    "title": "The Asymmetric Maximum Margin Bias of Quasi-Homogeneous Neural Networks",
                    "abstract": "In this work, we explore the maximum-margin bias of quasi-homogeneous neural networks trained with gradient flow on an exponential loss and past a point of separability. We introduce the class of quasi-homogeneous models, which is expressive enough to describe nearly all neural networks with homogeneous activations, even those with biases, residual connections, and normalization layers, while structured enough to enable geometric analysis of its gradient dynamics. Using this analysis, we generalize the existing results of maximum-margin bias for homogeneous networks to this richer class of models. We find that gradient flow implicitly favors a subset of the parameters, unlike in the case of a homogeneous model where all parameters are treated equally. We demonstrate through simple examples how this strong favoritism toward minimizing an asymmetric norm can degrade the robustness of quasi-homogeneous models. On the other hand, we conjecture that this norm-minimization discards, when possible, unnecessary higher-order parameters, reducing the model to a sparser parameterization. Lastly, by applying our theorem to sufficiently expressive neural networks with normalization layers, we reveal a universal mechanism behind the empirical phenomenon of Neural Collapse."
                },
                {
                    "title": "Linking Neural Collapse and L2 Normalization with Improved Out-of-Distribution Detection in Deep Neural Networks",
                    "abstract": "We propose a simple modification to standard ResNet architectures--L2 normalization over feature space--that substantially improves out-of-distribution (OoD) performance on the previously proposed Deep Deterministic Uncertainty (DDU) benchmark. We show that this change also induces early Neural Collapse (NC), an effect linked to better OoD performance. Our method achieves comparable or superior OoD detection scores and classification accuracy in a small fraction of the training time of the benchmark. Additionally, it substantially improves worst case OoD performance over multiple, randomly initialized models. Though we do not suggest that NC is the sole mechanism or a comprehensive explanation for OoD behaviour in deep neural networks (DNN), we believe NC's simple mathematical and geometric structure can provide a framework for analysis of this complex phenomenon in future work."
                },
                {
                    "title": "Imbalance Trouble: Revisiting Neural-Collapse Geometry",
                    "abstract": "Neural Collapse refers to the remarkable structural properties characterizing the geometry of class embeddings and classifier weights, found by deep nets when trained beyond zero training error. However, this characterization only holds for balanced data. Here we thus ask whether it can be made invariant to class imbalances. Towards this end, we adopt the unconstrained-features model (UFM), a recent theoretical model for studying neural collapse, and introduce Simplex-Encoded-Labels Interpolation (SELI) as an invariant characterization of the neural collapse phenomenon. Specifically, we prove for the UFM with cross-entropy loss and vanishing regularization that, irrespective of class imbalances, the embeddings and classifiers always interpolate a simplex-encoded label matrix and that their individual geometries are determined by the SVD factors of this same label matrix. We then present extensive experiments on synthetic and real datasets that confirm convergence to the SELI geometry. However, we caution that convergence worsens with increasing imbalances. We theoretically support this finding by showing that unlike the balanced case, when minorities are present, ridge-regularization plays a critical role in tweaking the geometry. This defines new questions and motivates further investigations into the impact of class imbalances on the rates at which first-order methods converge to their asymptotically preferred solutions."
                },
                {
                    "title": "Neural Collapse: A Review on Modelling Principles and Generalization",
                    "abstract": "Deep classifier neural networks enter the terminal phase of training (TPT) when training error reaches zero and tend to exhibit intriguing Neural Collapse (NC) properties. Neural collapse essentially represents a state at which the within-class variability of final hidden layer outputs is infinitesimally small and their class means form a simplex equiangular tight frame. This simplifies the last layer behaviour to that of a nearest-class center decision rule. Despite the simplicity of this state, the dynamics and implications of reaching it are yet to be fully understood. In this work, we review the principles which aid in modelling neural collapse, followed by the implications of this state on generalization and transfer learning capabilities of neural networks. Finally, we conclude by discussing potential avenues and directions for future research."
                },
                {
                    "title": "On the Optimization Landscape of Neural Collapse under MSE Loss: Global Optimality with Unconstrained Features",
                    "abstract": "When training deep neural networks for classification tasks, an intriguing empirical phenomenon has been widely observed in the last-layer classifiers and features, where (i) the class means and the last-layer classifiers all collapse to the vertices of a Simplex Equiangular Tight Frame (ETF) up to scaling, and (ii) cross-example within-class variability of last-layer activations collapses to zero. This phenomenon is called Neural Collapse (NC), which seems to take place regardless of the choice of loss functions. In this work, we justify NC under the mean squared error (MSE) loss, where recent empirical evidence shows that it performs comparably or even better than the de-facto cross-entropy loss. Under a simplified unconstrained feature model, we provide the first global landscape analysis for vanilla nonconvex MSE loss and show that the (only!) global minimizers are neural collapse solutions, while all other critical points are strict saddles whose Hessian exhibit negative curvature directions. Furthermore, we justify the usage of rescaled MSE loss by probing the optimization landscape around the NC solutions, showing that the landscape can be improved by tuning the rescaling hyperparameters. Finally, our theoretical findings are experimentally verified on practical network architectures."
                },
                {
                    "title": "Limitations of Neural Collapse for Understanding Generalization in Deep Learning",
                    "abstract": "The recent work of Papyan, Han,&Donoho (2020) presented an intriguing\"Neural Collapse\"phenomenon, showing a structural property of interpolating classifiers in the late stage of training. This opened a rich area of exploration studying this phenomenon. Our motivation is to study the upper limits of this research program: How far will understanding Neural Collapse take us in understanding deep learning? First, we investigate its role in generalization. We refine the Neural Collapse conjecture into two separate conjectures: collapse on the train set (an optimization property) and collapse on the test distribution (a generalization property). We find that while Neural Collapse often occurs on the train set, it does not occur on the test set. We thus conclude that Neural Collapse is primarily an optimization phenomenon, with as-yet-unclear connections to generalization. Second, we investigate the role of Neural Collapse in feature learning. We show simple, realistic experiments where training longer leads to worse last-layer features, as measured by transfer-performance on a downstream task. This suggests that neural collapse is not always desirable for representation learning, as previously claimed. Finally, we give preliminary evidence of a\"cascading collapse\"phenomenon, wherein some form of Neural Collapse occurs not only for the last layer, but in earlier layers as well. We hope our work encourages the community to continue the rich line of Neural Collapse research, while also considering its inherent limitations."
                },
                {
                    "title": "Extended Unconstrained Features Model for Exploring Deep Neural Collapse",
                    "abstract": "The modern strategy for training deep neural networks for classification tasks includes optimizing the network's weights even after the training error vanishes to further push the training loss toward zero. Recently, a phenomenon termed\"neural collapse\"(NC) has been empirically observed in this training procedure. Specifically, it has been shown that the learned features (the output of the penultimate layer) of within-class samples converge to their mean, and the means of different classes exhibit a certain tight frame structure, which is also aligned with the last layer's weights. Recent papers have shown that minimizers with this structure emerge when optimizing a simplified\"unconstrained features model\"(UFM) with a regularized cross-entropy loss. In this paper, we further analyze and extend the UFM. First, we study the UFM for the regularized MSE loss, and show that the minimizers' features can have a more delicate structure than in the cross-entropy case. This affects also the structure of the weights. Then, we extend the UFM by adding another layer of weights as well as ReLU nonlinearity to the model and generalize our previous results. Finally, we empirically demonstrate the usefulness of our nonlinear extended UFM in modeling the NC phenomenon that occurs with practical networks."
                },
                {
                    "title": "An Unconstrained Layer-Peeled Perspective on Neural Collapse",
                    "abstract": "Neural collapse is a highly symmetric geometric pattern of neural networks that emerges during the terminal phase of training, with profound implications on the generalization performance and robustness of the trained networks. To understand how the last-layer features and classifiers exhibit this recently discovered implicit bias, in this paper, we introduce a surrogate model called the unconstrained layer-peeled model (ULPM). We prove that gradient flow on this model converges to critical points of a minimum-norm separation problem exhibiting neural collapse in its global minimizer. Moreover, we show that the ULPM with the cross-entropy loss has a benign global landscape for its loss function, which allows us to prove that all the critical points are strict saddle points except the global minimizers that exhibit the neural collapse phenomenon. Empirically, we show that our results also hold during the training of neural networks in real-world tasks when explicit regularization or weight decay is not used."
                },
                {
                    "title": "Neural Collapse Under MSE Loss: Proximity to and Dynamics on the Central Path",
                    "abstract": "The recently discovered Neural Collapse (NC) phenomenon occurs pervasively in today's deep net training paradigm of driving cross-entropy (CE) loss towards zero. During NC, last-layer features collapse to their class-means, both classifiers and class-means collapse to the same Simplex Equiangular Tight Frame, and classifier behavior collapses to the nearest-class-mean decision rule. Recent works demonstrated that deep nets trained with mean squared error (MSE) loss perform comparably to those trained with CE. As a preliminary, we empirically establish that NC emerges in such MSE-trained deep nets as well through experiments on three canonical networks and five benchmark datasets. We provide, in a Google Colab notebook, PyTorch code for reproducing MSE-NC and CE-NC: at https://colab.research.google.com/github/neuralcollapse/neuralcollapse/blob/main/neuralcollapse.ipynb. The analytically-tractable MSE loss offers more mathematical opportunities than the hard-to-analyze CE loss, inspiring us to leverage MSE loss towards the theoretical investigation of NC. We develop three main contributions: (I) We show a new decomposition of the MSE loss into (A) terms directly interpretable through the lens of NC and which assume the last-layer classifier is exactly the least-squares classifier; and (B) a term capturing the deviation from this least-squares classifier. (II) We exhibit experiments on canonical datasets and networks demonstrating that term-(B) is negligible during training. This motivates us to introduce a new theoretical construct: the central path, where the linear classifier stays MSE-optimal for feature activations throughout the dynamics. (III) By studying renormalized gradient flow along the central path, we derive exact dynamics that predict NC."
                },
                {
                    "title": "Exploring deep neural networks via layer-peeled model: Minority collapse in imbalanced training",
                    "abstract": "Significance The remarkable development of deep learning over the past decade relies heavily on sophisticated heuristics and tricks. To better exploit its potential in the coming decade, perhaps a rigorous framework for reasoning about deep learning is needed, which, however, is not easy to build due to the intricate details of neural networks. For near-term purposes, a practical alternative is to develop a mathematically tractable surrogate model, yet maintaining many characteristics of neural networks. This paper proposes a model of this kind that we term the Layer-Peeled Model. The effectiveness of this model is evidenced by, among others, its ability to reproduce a known empirical pattern and to predict a hitherto-unknown phenomenon when training deep-learning models on imbalanced datasets. In this paper, we introduce the Layer-Peeled Model, a nonconvex, yet analytically tractable, optimization program, in a quest to better understand deep neural networks that are trained for a sufficiently long time. As the name suggests, this model is derived by isolating the topmost layer from the remainder of the neural network, followed by imposing certain constraints separately on the two parts of the network. We demonstrate that the Layer-Peeled Model, albeit simple, inherits many characteristics of well-trained neural networks, thereby offering an effective tool for explaining and predicting common empirical patterns of deep-learning training. First, when working on class-balanced datasets, we prove that any solution to this model forms a simplex equiangular tight frame, which, in part, explains the recently discovered phenomenon of neural collapse [V. Papyan, X. Y. Han, D. L. Donoho, Proc. Natl. Acad. Sci. U.S.A. 117, 24652\u201324663 (2020)]. More importantly, when moving to the imbalanced case, our analysis of the Layer-Peeled Model reveals a hitherto-unknown phenomenon that we term Minority Collapse, which fundamentally limits the performance of deep-learning models on the minority classes. In addition, we use the Layer-Peeled Model to gain insights into how to mitigate Minority Collapse. Interestingly, this phenomenon is first predicted by the Layer-Peeled Model before being confirmed by our computational experiments."
                },
                {
                    "title": "Explicit regularization and implicit bias in deep network classifiers trained with the square loss",
                    "abstract": "Deep ReLU networks trained with the square loss have been observed to perform well in classification tasks. We provide here a theoretical justification based on analysis of the associated gradient flow. We show that convergence to a solution with the absolute minimum norm is expected when normalization techniques such as Batch Normalization (BN) or Weight Normalization (WN) are used together with Weight Decay (WD). The main property of the minimizers that bounds their expected error is the norm: we prove that among all the close-to-interpolating solutions, the ones associated with smaller Frobenius norms of the unnormalized weight matrices have better margin and better bounds on the expected classification error. With BN but in the absence of WD, the dynamical system is singular. Implicit dynamical regularization \u2013 that is zero-initial conditions biasing the dynamics towards high margin solutions \u2013 is also possible in the no-BN and no-WD case. The theory yields several predictions, including the role of BN and weight decay, aspects of Papyan, Han and Donoho\u2019s Neural Collapse and the constraints induced by BN on the network weights."
                },
                {
                    "title": "Towards Resolving the Implicit Bias of Gradient Descent for Matrix Factorization: Greedy Low-Rank Learning",
                    "abstract": "Matrix factorization is a simple and natural test-bed to investigate the implicit regularization of gradient descent. Gunasekar et al. (2018) conjectured that Gradient Flow with infinitesimal initialization converges to the solution that minimizes the nuclear norm, but a series of recent papers argued that the language of norm minimization is not sufficient to give a full characterization for the implicit regularization. In this work, we provide theoretical and empirical evidence that for depth-2 matrix factorization, gradient flow with infinitesimal initialization is mathematically equivalent to a simple heuristic rank minimization algorithm, Greedy Low-Rank Learning, under some reasonable assumptions. This generalizes the rank minimization view from previous works to a much broader setting and enables us to construct counter-examples to refute the conjecture from Gunasekar et al. (2018). We also extend the results to the case where depth $\\ge 3$, and we show that the benefit of being deeper is that the above convergence has a much weaker dependence over initialization magnitude so that this rank minimization is more likely to take effect for initialization with practical scale."
                },
                {
                    "title": "On the emergence of simplex symmetry in the final and penultimate layers of neural network classifiers",
                    "abstract": "A recent numerical study observed that neural network classi\ufb01ers enjoy a large degree of symmetry in the penultimate layer. Namely, if h ( x ) = Af ( x ) + b where A is a linear map and f is the output of the penultimate layer of the network (after activation), then all data points x i, 1 , . . . , x i,N i in a class C i are mapped to a single point y i by f and the points y i are located at the vertices of a regular k \u2212 1 -dimensional standard simplex in a high-dimensional Euclidean space. We explain this observation analytically in toy models for highly expressive deep neural networks. In complementary examples, we demonstrate rigorously that even the \ufb01nal output of the classi\ufb01er h is not uniform over data samples from a class C i if h is a shallow network (or if the deeper layers do not bring the data samples into a convenient geometric con\ufb01guration)."
                },
                {
                    "title": "Universality Laws for High-Dimensional Learning With Random Features",
                    "abstract": "We prove a universality theorem for learning with random features. Our result shows that, in terms of training and generalization errors, a random feature model with a nonlinear activation function is asymptotically equivalent to a surrogate linear Gaussian model with a matching covariance matrix. This settles a so-called Gaussian equivalence conjecture based on which several recent papers develop their results. Our method for proving the universality theorem builds on the classical Lindeberg approach. Major ingredients of the proof include a leave-one-out analysis for the optimization problem associated with the training process and a central limit theorem, obtained via Stein\u2019s method, for weakly correlated random variables."
                },
                {
                    "title": "Prevalence of neural collapse during the terminal phase of deep learning training",
                    "abstract": "Significance Modern deep neural networks for image classification have achieved superhuman performance. Yet, the complex details of trained networks have forced most practitioners and researchers to regard them as black boxes with little that could be understood. This paper considers in detail a now-standard training methodology: driving the cross-entropy loss to zero, continuing long after the classification error is already zero. Applying this methodology to an authoritative collection of standard deepnets and datasets, we observe the emergence of a simple and highly symmetric geometry of the deepnet features and of the deepnet classifier, and we document important benefits that the geometry conveys\u2014thereby helping us understand an important component of the modern deep learning training paradigm. Modern practice for training classification deepnets involves a terminal phase of training (TPT), which begins at the epoch where training error first vanishes. During TPT, the training error stays effectively zero, while training loss is pushed toward zero. Direct measurements of TPT, for three prototypical deepnet architectures and across seven canonical classification datasets, expose a pervasive inductive bias we call neural collapse (NC), involving four deeply interconnected phenomena. (NC1) Cross-example within-class variability of last-layer training activations collapses to zero, as the individual activations themselves collapse to their class means. (NC2) The class means collapse to the vertices of a simplex equiangular tight frame (ETF). (NC3) Up to rescaling, the last-layer classifiers collapse to the class means or in other words, to the simplex ETF (i.e., to a self-dual configuration). (NC4) For a given activation, the classifier\u2019s decision collapses to simply choosing whichever class has the closest train class mean (i.e., the nearest class center [NCC] decision rule). The symmetric and very simple geometry induced by the TPT confers important benefits, including better generalization performance, better robustness, and better interpretability."
                },
                {
                    "title": "The Neural Tangent Kernel in High Dimensions: Triple Descent and a Multi-Scale Theory of Generalization",
                    "abstract": "Modern deep learning models employ considerably more parameters than required to fit the training data. Whereas conventional statistical wisdom suggests such models should drastically overfit, in practice these models generalize remarkably well. An emerging paradigm for describing this unexpected behavior is in terms of a \\emph{double descent} curve, in which increasing a model's capacity causes its test error to first decrease, then increase to a maximum near the interpolation threshold, and then decrease again in the overparameterized regime. Recent efforts to explain this phenomenon theoretically have focused on simple settings, such as linear regression or kernel regression with unstructured random features, which we argue are too coarse to reveal important nuances of actual neural networks. We provide a precise high-dimensional asymptotic analysis of generalization under kernel regression with the Neural Tangent Kernel, which characterizes the behavior of wide neural networks optimized with gradient descent. Our results reveal that the test error has non-monotonic behavior deep in the overparameterized regime and can even exhibit additional peaks and descents when the number of parameters scales quadratically with the dataset size."
                },
                {
                    "title": "A Random Matrix Perspective on Mixtures of Nonlinearities for Deep Learning",
                    "abstract": "One of the distinguishing characteristics of modern deep learning systems is that they typically employ neural network architectures that utilize enormous numbers of parameters, often in the millions and sometimes even in the billions. While this paradigm has inspired significant research on the properties of large networks, relatively little work has been devoted to the fact that these networks are often used to model large complex datasets, which may themselves contain millions or even billions of constraints. In this work, we focus on this high-dimensional regime in which both the dataset size and the number of features tend to infinity. We analyze the performance of a simple regression model trained on the random features $F=f(WX+B)$ for a random weight matrix $W$ and random bias vector $B$, obtaining an exact formula for the asymptotic training error on a noisy autoencoding task. The role of the bias can be understood as parameterizing a distribution over activation functions, and our analysis directly generalizes to such distributions, even those not expressible with a traditional additive bias. Intriguingly, we find that a mixture of nonlinearities can outperform the best single nonlinearity on the noisy autoecndoing task, suggesting that mixtures of nonlinearities might be useful for approximate kernel methods or neural network architecture design."
                },
                {
                    "title": "Beyond worst-case analysis",
                    "abstract": "The need for deeply understanding when algorithms work (or not) has never been greater."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "Convex Learning of Multiple Tasks and their Structure",
                    "abstract": "Reducing the amount of human supervision is a key problem in machine learning and a natural approach is that of exploiting the relations (structure) among different tasks. This is the idea at the core of multi-task learning. In this context a fundamental question is how to incorporate the tasks structure in the learning problem. We tackle this question by studying a general computational framework that allows to encode apriori knowledge of the tasks structure in the form of a convex penalty; in this setting a variety of previously proposed methods can be recovered as special cases, including linear and non-linear approaches. Within this framework, we show that tasks and their structure can be efficiently learned considering a convex optimization problem that can be approached by means of block coordinate methods such as alternating minimization and for which we prove convergence to the global minimum."
                },
                {
                    "title": "A Consistent Estimator of the Expected Gradient Outerproduct",
                    "abstract": "In high-dimensional classification or regression problems, the expected gradient outerproduct (EGOP) of the unknown regression function f, namely Ex (\u2207 f(X) \u00b7 \u2207f(X)\u22a4), is known to recover those directions v \u2208 \u211dd most relevant to predicting the output Y. \n \nHowever, just as in gradient estimation, optimal estimators of the EGOP can be expensive in practice. We show that a simple rough estimator, much cheaper in practice, suffices to obtain significant improvements on real-world nonparametric classification and regression tasks. Furthermore, we prove that, despite its simplicity, this rough estimator remains statistically consistent under mild conditions."
                },
                {
                    "title": "The spectrum of kernel random matrices",
                    "abstract": "We place ourselves in the setting of high-dimensional statistical inference where the number of variables p in a dataset of interest is of the same order of magnitude as the number of observations n. We consider the spectrum of certain kernel random matrices, in particular n x n matrices whose (i, j)th entry is f (X' i X j /p) or f (\u2225X i - X j \u2225 2 /p) where p is the dimension of the data, and X i are independent data vectors. Here f is assumed to be a locally smooth function. The study is motivated by questions arising in statistics and computer science where these matrices are used to perform, among other things, nonlinear versions of principal component analysis. Surprisingly, we show that in high-dimensions, and for the models we analyze, the problem becomes essentially linear\u2014which is at odds with heuristics sometimes used to justify the usage of these methods. The analysis also highlights certain peculiarities of models widely studied in random matrix theory and raises some questions about their relevance as tools to model high-dimensional data encountered in practice."
                },
                {
                    "title": "Kernel Methods for Deep Learning",
                    "abstract": "We introduce a new family of positive-definite kernel functions that mimic the computation in large, multilayer neural nets. These kernel functions can be used in shallow architectures, such as support vector machines (SVMs), or in deep kernel-based architectures that we call multilayer kernel machines (MKMs). We evaluate SVMs and MKMs with these kernel functions on problems designed to illustrate the advantages of deep architectures. On several problems, we obtain better results than previous, leading benchmarks from both SVMs with Gaussian kernels as well as deep belief nets."
                },
                {
                    "title": "Kernels for Multi--task Learning",
                    "abstract": "This paper provides a foundation for multi-task learning using reproducing kernel Hilbert spaces of vector-valued functions. In this setting, the kernel is a matrix-valued function. Some explicit examples will be described which go beyond our earlier results in [7]. In particular, we characterize classes of matrix- valued kernels which are linear and are of the dot product or the translation invariant type. We discuss how these kernels can be used to model relations between the tasks and present linear multi-task learning algorithms. Finally, we present a novel proof of the representer theorem for a minimizer of a regularization functional which is based on the notion of minimal norm interpolation."
                },
                {
                    "title": "Smoothed analysis of algorithms: why the simplex algorithm usually takes polynomial time",
                    "abstract": "We introduce the smoothed analysis of algorithms, which is a hybrid of the worst-case and average-case analysis of algorithms. Essentially, we study the performance of algorithms under small random perturbations of their inputs. We show that the shadow-vertex simplex algorithm has polynomial smoothed complexity."
                },
                {
                    "title": "Gradient descent induces alignment between weights and the empirical NTK for deep non-linear networks",
                    "abstract": "Understanding the mechanisms through which neural networks extract statistics from input-label pairs is one of the most important unsolved problems in supervised learning. Prior works have identified that the gram matrices of the weights in trained neural networks of general architectures are proportional to the average gradient outer product of the model, in a statement known as the Neural Feature Ansatz (NFA). However, the reason these quantities become correlated during training is poorly understood. In this work, we explain the emergence of this correlation. We identify that the NFA is equivalent to alignment between the left singular structure of the weight matrices and a significant component of the empirical neural tangent kernels associated with those weights. We establish that the NFA introduced in prior works is driven by a centered NFA that isolates this alignment. We show that the speed of NFA development can be predicted analytically at early training times in terms of simple statistics of the inputs and labels. Finally, we introduce a simple intervention to increase NFA correlation at any given layer, which dramatically improves the quality of features learned."
                },
                {
                    "title": "Feature learning in deep classifiers through Intermediate Neural Collapse",
                    "abstract": "In this paper, we conduct an empirical study of the feature learning process in deep classifiers. Recent research has identified a training phenomenon called Neural Collapse (NC), in which the top-layer feature embeddings of samples from the same class tend to concentrate around their means, and the top layer\u2019s weights align with those features. Our study aims to investigate if these properties extend to intermediate layers. We empirically study the evolution of the covariance and mean of representations across different layers and show that as we move deeper into a trained neural network, the within-class covariance decreases relative to the between-class covariance. Additionally, we find that in the top layers, where the between-class covariance is dominant, the sub-space spanned by the class means aligns with the subspace spanned by the most significant singular vector components of the weight matrix in the corresponding layer. Finally, we discuss the relationship between NC and Associative Memo-ries (Willshaw et al., 1969)."
                },
                {
                    "title": "Feature learning in neural networks and kernel machines that recursively learn features",
                    "abstract": "Neural networks have achieved impressive results on many technological and scienti\ufb01c tasks. Yet, their empirical successes have outpaced our fundamental understanding of their structure and function. Identifying mechanisms driving the successes of neural networks can provide principled approaches for improving neural network performance and developing simple and e\ufb00ective alternatives. In this work, we isolate a key mechanism driving feature learning in fully connected neural networks by connecting neural feature learning to a statistical estimator known as average gradient outer product. We subsequently leverage this mechanism to design Recursive Feature Machines (RFMs), which are kernel machines that learn features. We show that RFMs (1) accurately capture features learned by deep fully connected neural networks, and (2) outperform a broad spectrum of models including neural networks on tabular data. Furthermore, we show how RFMs shed light on recently observed deep learning phenomena including grokking, lottery tickets, simplicity biases, and spurious features. We provide a Python implementation to make our method easily accessible. 4"
                },
                {
                    "title": "Linear Convergence Analysis of Neural Collapse with Unconstrained Features",
                    "abstract": "In this work, we study the recently discovered neural collapse (NC) phenomenon, which is prevalent in training over-parameterized deep neural networks for classification tasks. Existing work has shown that any optimal solution of the trained problem for classification tasks is an NC solution and has a benign landscape under the unconstrained feature model. However, these results do not provide an answer to the question of how quickly gradient descent can find an NC solution. To answer this question, we prove an error bound property of the trained problem, which refers to the inequality that bounds the distance of a point to the optimal solution set by the norm of its gradient, under the unconstrained feature model. Using this error bound, we show linear convergence of gradient descent for finding an NC solution."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does the learning algorithm based on iterated linear mapping onto the average gradient outer product (AGOP) contribute to the formation of Deep Neural Collapse (DNC) in deep neural networks?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing our understanding of the mechanisms behind deep learning, particularly in how DNC influences the performance of deep neural networks. By elucidating the relationship between the learning algorithm and DNC, this research could lead to improved model architectures and training strategies, enhancing generalization and robustness in practical applications. Furthermore, it could inspire future research into the geometric properties of feature representations, potentially leading to novel insights in both theoretical and applied machine learning.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem stem from the complexity of deep neural networks and the intricate interplay between their architecture, training data, and learning algorithms. Naive approaches may fail because they often overlook the role of the training data in shaping the feature representations and the dynamics of DNC formation. Additionally, the theoretical frameworks currently in use, such as the deep unconstrained features model (DUFM), do not adequately account for the learning process, leading to a gap in understanding how DNC emerges in practice. Overcoming these obstacles requires a nuanced approach that integrates both theoretical and empirical insights.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on data-agnostic models that do not consider the influence of training data on DNC formation. This limitation has hindered a comprehensive understanding of how DNC develops in the context of the full deep learning pipeline. Additionally, existing models like DUFM have not adequately captured the learning dynamics involved in DNC. Our approach differs by introducing a learning algorithm that explicitly incorporates the AGOP, allowing for a more accurate representation of how DNC forms through the training process, thus addressing the gaps left by prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves utilizing the average gradient outer product (AGOP) as a foundation for understanding DNC formation through a learning algorithm based on iterated linear mapping. We will analyze the feature representations in deep neural networks using a dataset relevant to classification tasks, employing metrics that assess the geometric properties of the learned features, such as within-class variability and orthogonality. The expected outcomes include a clearer"
            }
        },
        "author_data": {
            "9a2e5c26-d542-4294-a25e-f74e5bde27a7": {
                "pk": "9a2e5c26-d542-4294-a25e-f74e5bde27a7",
                "name": "Daniel Beaglehole",
                "collaborators": [
                    "Mikhail Belkin",
                    "Parthe Pandit",
                    "Adityanarayanan Radhakrishnan",
                    "Thomas Valles",
                    "Alexandr Andoni",
                    "Max Hopkins",
                    "Daniel Kane",
                    "Sihan Liu",
                    "Shachar Lovett",
                    "Ioannis Mitliagkas"
                ],
                "domain": [
                    "Game Theory",
                    "Machine Learning",
                    "Neural Networks",
                    "Feature Learning"
                ],
                "publications": [
                    {
                        "title": "Fast, optimal, and dynamic electoral campaign budgeting by a generalized Colonel Blotto game",
                        "abstract": "The Colonel Blotto game is a deeply studied theoretical model for competitive allocation environments including elections, advertising, and ecology. However, the original formulation of Colonel Blotto has had few practical implications due to the lack of fast algorithms to compute its optimal strategies and the limited applicability of its winner-take-all reward distribution. We demonstrate that the Colonel Blotto game can be a practical model for competitive allocation environments by implementing the multiplicative weights update algorithm from Beaglehole et al. (2023). In particular, using that this algorithm allows for arbitrary winning-rules, we study strategies for a more realistic model of political campaigning we term Electoral Colonel Blotto. Contrary to existing theory and the implemented allocation strategies from U.S. presidential elections, we find that the optimal response to Democratic and Republican strategies in the 2008 and 2020 presidential elections was to focus allocations on a subset of states and sacrifice winning probability on others. We also found that campaigners should compete for undecided voters even in states where the opponent has significantly many more decided voters."
                    },
                    {
                        "title": "On the Inconsistency of Kernel Ridgeless Regression in Fixed Dimensions",
                        "abstract": "``Benign overfitting'', the ability of certain algorithms to interpolate noisy training data and yet perform well out-of-sample, has been a topic of considerable recent interest. We show, using a fixed design setup, that an important class of predictors, kernel machines with translation-invariant kernels, does not exhibit benign overfitting in fixed dimensions. In particular, the estimated predictor does not converge to the ground truth with increasing sample size, for any non-zero regression function and any (even adaptive) bandwidth selection. To prove these results, we give exact expressions for the generalization error, and its decomposition in terms of an approximation error and an estimation error that elicits a trade-off based on the selection of the kernel bandwidth. Our results apply to commonly used translation-invariant kernels such as Gaussian, Laplace, and Cauchy."
                    },
                    {
                        "title": "Mechanism of feature learning in convolutional neural networks",
                        "abstract": "Understanding the mechanism of how convolutional neural networks learn features from image data is a fundamental problem in machine learning and computer vision. In this work, we identify such a mechanism. We posit the Convolutional Neural Feature Ansatz, which states that covariances of filters in any convolutional layer are proportional to the average gradient outer product (AGOP) taken with respect to patches of the input to that layer. We present extensive empirical evidence for our ansatz, including identifying high correlation between covariances of filters and patch-based AGOPs for convolutional layers in standard neural architectures, such as AlexNet, VGG, and ResNets pre-trained on ImageNet. We also provide supporting theoretical evidence. We then demonstrate the generality of our result by using the patch-based AGOP to enable deep feature learning in convolutional kernel machines. We refer to the resulting algorithm as (Deep) ConvRFM and show that our algorithm recovers similar features to deep convolutional networks including the notable emergence of edge detectors. Moreover, we find that Deep ConvRFM overcomes previously identified limitations of convolutional kernels, such as their inability to adapt to local signals in images and, as a result, leads to sizable performance improvement over fixed convolutional kernels."
                    },
                    {
                        "title": "Learning to Hash Robustly, Guaranteed",
                        "abstract": "The indexing algorithms for the high-dimensional nearest neighbor search (NNS) with the best worst-case guarantees are based on the randomized Locality Sensitive Hashing (LSH), and its derivatives. In practice, many heuristic approaches exist to \"learn\" the best indexing method in order to speed-up NNS, crucially adapting to the structure of the given dataset.   Oftentimes, these heuristics outperform the LSH-based algorithms on real datasets, but, almost always, come at the cost of losing the guarantees of either correctness or robust performance on adversarial queries, or apply to datasets with an assumed extra structure/model. In this paper, we design an NNS algorithm for the Hamming space that has worst-case guarantees essentially matching that of theoretical algorithms, while optimizing the hashing to the structure of the dataset (think instance-optimal algorithms) for performance on the minimum-performing query. We evaluate the algorithm's ability to optimize for a given dataset both theoretically and practically. On the theoretical side, we exhibit a natural setting (dataset model) where our algorithm is much better than the standard theoretical one. On the practical side, we run experiments that show that our algorithm has a 1.8x and 2.1x better recall on the worst-performing queries to the MNIST and ImageNet datasets."
                    },
                    {
                        "title": "Sampling Equilibria: Fast No-Regret Learning in Structured Games",
                        "abstract": "Learning and equilibrium computation in games are fundamental problems across computer science and economics, with applications ranging from politics to machine learning. Much of the work in this area revolves around a simple algorithm termed \\emph{randomized weighted majority} (RWM), also known as \"Hedge\" or \"Multiplicative Weights Update,\" which is well known to achieve statistically optimal rates in adversarial settings (Littlestone and Warmuth '94, Freund and Schapire '99). Unfortunately, RWM comes with an inherent computational barrier: it requires maintaining and sampling from a distribution over all possible actions. In typical settings of interest the action space is exponentially large, seemingly rendering RWM useless in practice.   In this work, we refute this notion for a broad variety of \\emph{structured} games, showing it is possible to efficiently (approximately) sample the action space in RWM in \\emph{polylogarithmic} time. This gives the first efficient no-regret algorithms for problems such as the \\emph{(discrete) Colonel Blotto game}, \\emph{matroid congestion}, \\emph{matroid security}, and basic \\emph{dueling games}. As an immediate corollary, we give a polylogarithmic time meta-algorithm to compute approximate Nash Equilibria for these games that is exponentially faster than prior methods in several important settings. Further, our algorithm is the first to efficiently compute equilibria for more involved variants of these games with general sums, more than two players, and, for Colonel Blotto, multiple resource types."
                    },
                    {
                        "title": "Feature learning as alignment: a structural property of gradient descent in non-linear neural networks",
                        "abstract": "Understanding the mechanisms through which neural networks extract statistics from input-label pairs through feature learning is one of the most important unsolved problems in supervised learning. Prior works demonstrated that the gram matrices of the weights (the neural feature matrices, NFM) and the average gradient outer products (AGOP) become correlated during training, in a statement known as the neural feature ansatz (NFA). Through the NFA, the authors introduce mapping with the AGOP as a general mechanism for neural feature learning. However, these works do not provide a theoretical explanation for this correlation or its origins. In this work, we further clarify the nature of this correlation, and explain its emergence. We show that this correlation is equivalent to alignment between the left singular structure of the weight matrices and the newly defined pre-activation tangent features at each layer. We further establish that the alignment is driven by the interaction of weight changes induced by SGD with the pre-activation features, and analyze the resulting dynamics analytically at early times in terms of simple statistics of the inputs and labels. Finally, motivated by the observation that the NFA is driven by this centered correlation, we introduce a simple optimization rule that dramatically increases the NFA correlations at any given layer and improves the quality of features learned."
                    },
                    {
                        "title": "Mechanism of feature learning in deep fully connected networks and kernel machines that recursively learn features",
                        "abstract": "In recent years neural networks have achieved impressive results on many technological and scientific tasks. Yet, the mechanism through which these models automatically select features, or patterns in data, for prediction remains unclear. Identifying such a mechanism is key to advancing performance and interpretability of neural networks and promoting reliable adoption of these models in scientific applications. In this paper, we identify and characterize the mechanism through which deep fully connected neural networks learn features. We posit the Deep Neural Feature Ansatz, which states that neural feature learning occurs by implementing the average gradient outer product to up-weight features strongly related to model output. Our ansatz sheds light on various deep learning phenomena including emergence of spurious features and simplicity biases and how pruning networks can increase performance, the \"lottery ticket hypothesis.\" Moreover, the mechanism identified in our work leads to a backpropagation-free method for feature learning with any machine learning model. To demonstrate the effectiveness of this feature learning mechanism, we use it to enable feature learning in classical, non-feature learning models known as kernel machines and show that the resulting models, which we refer to as Recursive Feature Machines, achieve state-of-the-art performance on tabular data."
                    },
                    {
                        "title": "Emergence in non-neural models: grokking modular arithmetic via average gradient outer product",
                        "abstract": "Neural networks trained to solve modular arithmetic tasks exhibit grokking, a phenomenon where the test accuracy starts improving long after the model achieves 100% training accuracy in the training process. It is often taken as an example of \"emergence\", where model ability manifests sharply through a phase transition. In this work, we show that the phenomenon of grokking is not specific to neural networks nor to gradient descent-based optimization. Specifically, we show that this phenomenon occurs when learning modular arithmetic with Recursive Feature Machines (RFM), an iterative algorithm that uses the Average Gradient Outer Product (AGOP) to enable task-specific feature learning with general machine learning models. When used in conjunction with kernel machines, iterating RFM results in a fast transition from random, near zero, test accuracy to perfect test accuracy. This transition cannot be predicted from the training loss, which is identically zero, nor from the test loss, which remains constant in initial iterations. Instead, as we show, the transition is completely determined by feature learning: RFM gradually learns block-circulant features to solve modular arithmetic. Paralleling the results for RFM, we show that neural networks that solve modular arithmetic also learn block-circulant features. Furthermore, we present theoretical evidence that RFM uses such block-circulant features to implement the Fourier Multiplication Algorithm, which prior work posited as the generalizing solution neural networks learn on these tasks. Our results demonstrate that emergence can result purely from learning task-relevant features and is not specific to neural architectures nor gradient descent-based optimization methods. Furthermore, our work provides more evidence for AGOP as a key mechanism for feature learning in neural networks."
                    }
                ]
            },
            "f32b9739-742a-45e7-9899-0b0e96cb1c03": {
                "pk": "f32b9739-742a-45e7-9899-0b0e96cb1c03",
                "name": "Peter S\u00faken\u00edk",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "a166913d-9e97-41fa-b4c0-b42c4b79589d": {
                "pk": "a166913d-9e97-41fa-b4c0-b42c4b79589d",
                "name": "Marco Mondelli",
                "collaborators": [
                    "R\u00fcdiger Urbanke",
                    "S. Hamed Hassani",
                    "Simone Bombari",
                    "Andrea Montanari",
                    "Ramji Venkataramanan",
                    "Quynh Nguyen",
                    "Shrinivas Kudekar",
                    "Eren \u015ea\u015fo\u011flu",
                    "Seyyed Ali Hashemi",
                    "John Cioffi"
                ],
                "domain": [
                    "Coding Theory",
                    "Neural Networks",
                    "Statistical Learning",
                    "Information Theory"
                ],
                "publications": [
                    {
                        "title": "Reed-Muller Codes Achieve Capacity on the Binary Erasure Channel under MAP Decoding",
                        "abstract": "We show that Reed-Muller codes achieve capacity under maximum a posteriori bit decoding for transmission over the binary erasure channel for all rates $0 < R < 1$. The proof is generic and applies to other codes with sufficient amount of symmetry as well. The main idea is to combine the following observations: (i) monotone functions experience a sharp threshold behavior, (ii) the extrinsic information transfer (EXIT) functions are monotone, (iii) Reed--Muller codes are 2-transitive and thus the EXIT functions associated with their codeword bits are all equal, and (iv) therefore the Area Theorem for the average EXIT functions implies that RM codes' threshold is at channel capacity."
                    },
                    {
                        "title": "On the Connection Between Learning Two-Layers Neural Networks and Tensor Decomposition",
                        "abstract": "We establish connections between the problem of learning a two-layer neural network and tensor decomposition. We consider a model with feature vectors $\\boldsymbol x \\in \\mathbb R^d$, $r$ hidden units with weights $\\{\\boldsymbol w_i\\}_{1\\le i \\le r}$ and output $y\\in \\mathbb R$, i.e., $y=\\sum_{i=1}^r \\sigma( \\boldsymbol w_i^{\\mathsf T}\\boldsymbol x)$, with activation functions given by low-degree polynomials. In particular, if $\\sigma(x) = a_0+a_1x+a_3x^3$, we prove that no polynomial-time learning algorithm can outperform the trivial predictor that assigns to each example the response variable $\\mathbb E(y)$, when $d^{3/2}\\ll r\\ll d^2$. Our conclusion holds for a `natural data distribution', namely standard Gaussian feature vectors $\\boldsymbol x$, and output distributed according to a two-layer neural network with random isotropic weights, and under a certain complexity-theoretic assumption on tensor decomposition. Roughly speaking, we assume that no polynomial-time algorithm can substantially outperform current methods for tensor decomposition based on the sum-of-squares hierarchy.   We also prove generalizations of this statement for higher degree polynomial activations, and non-random weight vectors. Remarkably, several existing algorithms for learning two-layer networks with rigorous guarantees are based on tensor decomposition. Our results support the idea that this is indeed the core computational difficulty in learning such networks, under the stated generative model for the data. As a side result, we show that under this model learning the network requires accurate learning of its weights, a property that does not hold in a more general setting."
                    },
                    {
                        "title": "Approximate Message Passing with Spectral Initialization for Generalized Linear Models",
                        "abstract": "We consider the problem of estimating a signal from measurements obtained via a generalized linear model. We focus on estimators based on approximate message passing (AMP), a family of iterative algorithms with many appealing features: the performance of AMP in the high-dimensional limit can be succinctly characterized under suitable model assumptions; AMP can also be tailored to the empirical distribution of the signal entries, and for a wide class of estimation problems, AMP is conjectured to be optimal among all polynomial-time algorithms.   However, a major issue of AMP is that in many models (such as phase retrieval), it requires an initialization correlated with the ground-truth signal and independent from the measurement matrix. Assuming that such an initialization is available is typically not realistic. In this paper, we solve this problem by proposing an AMP algorithm initialized with a spectral estimator. With such an initialization, the standard AMP analysis fails since the spectral estimator depends in a complicated way on the design matrix. Our main contribution is a rigorous characterization of the performance of AMP with spectral initialization in the high-dimensional limit. The key technical idea is to define and analyze a two-phase artificial AMP algorithm that first produces the spectral estimator, and then closely approximates the iterates of the true AMP. We also provide numerical results that demonstrate the validity of the proposed approach."
                    },
                    {
                        "title": "PCA Initialization for Approximate Message Passing in Rotationally Invariant Models",
                        "abstract": "We study the problem of estimating a rank-$1$ signal in the presence of rotationally invariant noise-a class of perturbations more general than Gaussian noise. Principal Component Analysis (PCA) provides a natural estimator, and sharp results on its performance have been obtained in the high-dimensional regime. Recently, an Approximate Message Passing (AMP) algorithm has been proposed as an alternative estimator with the potential to improve the accuracy of PCA. However, the existing analysis of AMP requires an initialization that is both correlated with the signal and independent of the noise, which is often unrealistic in practice. In this work, we combine the two methods, and propose to initialize AMP with PCA. Our main result is a rigorous asymptotic characterization of the performance of this estimator. Both the AMP algorithm and its analysis differ from those previously derived in the Gaussian setting: at every iteration, our AMP algorithm requires a specific term to account for PCA initialization, while in the Gaussian case, PCA initialization affects only the first iteration of AMP. The proof is based on a two-phase artificial AMP that first approximates the PCA estimator and then mimics the true AMP. Our numerical simulations show an excellent agreement between AMP results and theoretical predictions, and suggest an interesting open direction on achieving Bayes-optimal performance."
                    },
                    {
                        "title": "Fundamental Limits of Weak Recovery with Applications to Phase Retrieval",
                        "abstract": "In phase retrieval we want to recover an unknown signal $\\boldsymbol x\\in\\mathbb C^d$ from $n$ quadratic measurements of the form $y_i = |\\langle{\\boldsymbol a}_i,{\\boldsymbol x}\\rangle|^2+w_i$ where $\\boldsymbol a_i\\in \\mathbb C^d$ are known sensing vectors and $w_i$ is measurement noise. We ask the following weak recovery question: what is the minimum number of measurements $n$ needed to produce an estimator $\\hat{\\boldsymbol x}(\\boldsymbol y)$ that is positively correlated with the signal $\\boldsymbol x$? We consider the case of Gaussian vectors $\\boldsymbol a_i$. We prove that - in the high-dimensional limit - a sharp phase transition takes place, and we locate the threshold in the regime of vanishingly small noise. For $n\\le d-o(d)$ no estimator can do significantly better than random and achieve a strictly positive correlation. For $n\\ge d+o(d)$ a simple spectral estimator achieves a positive correlation. Surprisingly, numerical simulations with the same spectral estimator demonstrate promising performance with realistic sensing matrices. Spectral methods are used to initialize non-convex optimization algorithms in phase retrieval, and our approach can boost the performance in this setting as well.   Our impossibility result is based on classical information-theory arguments. The spectral algorithm computes the leading eigenvector of a weighted empirical covariance matrix. We obtain a sharp characterization of the spectral properties of this random matrix using tools from free probability and generalizing a recent result by Lu and Li. Both the upper and lower bound generalize beyond phase retrieval to measurements $y_i$ produced according to a generalized linear model. As a byproduct of our analysis, we compare the threshold of the proposed spectral method with that of a message passing algorithm."
                    },
                    {
                        "title": "Sublinear Latency for Simplified Successive Cancellation Decoding of Polar Codes",
                        "abstract": "This work analyzes the latency of the simplified successive cancellation (SSC) decoding scheme for polar codes proposed by Alamdar-Yazdi and Kschischang. It is shown that, unlike conventional successive cancellation decoding, where latency is linear in the block length, the latency of SSC decoding is sublinear. More specifically, the latency of SSC decoding is $O(N^{1-1/\\mu})$, where $N$ is the block length and $\\mu$ is the scaling exponent of the channel, which captures the speed of convergence of the rate to capacity. Numerical results demonstrate the tightness of the bound and show that most of the latency reduction arises from the parallel decoding of subcodes of rate $0$ or $1$."
                    },
                    {
                        "title": "Global Convergence of Deep Networks with One Wide Layer Followed by Pyramidal Topology",
                        "abstract": "Recent works have shown that gradient descent can find a global minimum for over-parameterized neural networks where the widths of all the hidden layers scale polynomially with $N$ ($N$ being the number of training samples). In this paper, we prove that, for deep networks, a single layer of width $N$ following the input layer suffices to ensure a similar guarantee. In particular, all the remaining layers are allowed to have constant widths, and form a pyramidal topology. We show an application of our result to the widely used LeCun's initialization and obtain an over-parameterization requirement for the single wide layer of order $N^2.$"
                    },
                    {
                        "title": "Polar Coded Computing: The Role of the Scaling Exponent",
                        "abstract": "We consider the problem of coded distributed computing using polar codes. The average execution time of a coded computing system is related to the error probability for transmission over the binary erasure channel in recent work by Soleymani, Jamali and Mahdavifar, where the performance of binary linear codes is investigated. In this paper, we focus on polar codes and unveil a connection between the average execution time and the scaling exponent $\\mu$ of the family of codes. The scaling exponent has emerged as a central object in the finite-length characterization of polar codes, and it captures the speed of convergence to capacity. In particular, we show that (i) the gap between the normalized average execution time of polar codes and that of optimal MDS codes is $O(n^{-1/\\mu})$, and (ii) this upper bound can be improved to roughly $O(n^{-1/2})$ by considering polar codes with large kernels. We conjecture that these bounds could be improved to $O(n^{-2/\\mu})$ and $O(n^{-1})$, respectively, and provide a heuristic argument as well as numerical evidence supporting this view."
                    },
                    {
                        "title": "Matrix Denoising with Doubly Heteroscedastic Noise: Fundamental Limits and Optimal Spectral Methods",
                        "abstract": "We study the matrix denoising problem of estimating the singular vectors of a rank-$1$ signal corrupted by noise with both column and row correlations. Existing works are either unable to pinpoint the exact asymptotic estimation error or, when they do so, the resulting approaches (e.g., based on whitening or singular value shrinkage) remain vastly suboptimal. On top of this, most of the literature has focused on the special case of estimating the left singular vector of the signal when the noise only possesses row correlation (one-sided heteroscedasticity). In contrast, our work establishes the information-theoretic and algorithmic limits of matrix denoising with doubly heteroscedastic noise. We characterize the exact asymptotic minimum mean square error, and design a novel spectral estimator with rigorous optimality guarantees: under a technical condition, it attains positive correlation with the signals whenever information-theoretically possible and, for one-sided heteroscedasticity, it also achieves the Bayes-optimal error. Numerical experiments demonstrate the significant advantage of our theoretically principled method with the state of the art. The proofs draw connections with statistical physics and approximate message passing, departing drastically from standard random matrix theory techniques."
                    },
                    {
                        "title": "Privacy for Free in the Over-Parameterized Regime",
                        "abstract": "Differentially private gradient descent (DP-GD) is a popular algorithm to train deep learning models with provable guarantees on the privacy of the training data. In the last decade, the problem of understanding its performance cost with respect to standard GD has received remarkable attention from the research community, which formally derived upper bounds on the excess population risk $R_{P}$ in different learning settings. However, existing bounds typically degrade with over-parameterization, i.e., as the number of parameters $p$ gets larger than the number of training samples $n$ -- a regime which is ubiquitous in current deep-learning practice. As a result, the lack of theoretical insights leaves practitioners without clear guidance, leading some to reduce the effective number of trainable parameters to improve performance, while others use larger models to achieve better results through scale. In this work, we show that in the popular random features model with quadratic loss, for any sufficiently large $p$, privacy can be obtained for free, i.e., $\\left|R_{P} \\right| = o(1)$, not only when the privacy parameter $\\varepsilon$ has constant order, but also in the strongly private setting $\\varepsilon = o(1)$. This challenges the common wisdom that over-parameterization inherently hinders performance in private learning."
                    },
                    {
                        "title": "Scaling Exponent of List Decoders with Applications to Polar Codes",
                        "abstract": "Motivated by the significant performance gains which polar codes experience under successive cancellation list decoding, their scaling exponent is studied as a function of the list size. In particular, the error probability is fixed and the trade-off between block length and back-off from capacity is analyzed. A lower bound is provided on the error probability under $\\rm MAP$ decoding with list size $L$ for any binary-input memoryless output-symmetric channel and for any class of linear codes such that their minimum distance is unbounded as the block length grows large. Then, it is shown that under $\\rm MAP$ decoding, although the introduction of a list can significantly improve the involved constants, the scaling exponent itself, i.e., the speed at which capacity is approached, stays unaffected for any finite list size. In particular, this result applies to polar codes, since their minimum distance tends to infinity as the block length increases. A similar result is proved for genie-aided successive cancellation decoding when transmission takes place over the binary erasure channel, namely, the scaling exponent remains constant for any fixed number of helps from the genie. Note that since genie-aided successive cancellation decoding might be strictly worse than successive cancellation list decoding, the problem of establishing the scaling exponent of the latter remains open."
                    },
                    {
                        "title": "From Polar to Reed-Muller Codes: a Technique to Improve the Finite-Length Performance",
                        "abstract": "We explore the relationship between polar and RM codes and we describe a coding scheme which improves upon the performance of the standard polar code at practical block lengths. Our starting point is the experimental observation that RM codes have a smaller error probability than polar codes under MAP decoding. This motivates us to introduce a family of codes that \"interpolates\" between RM and polar codes, call this family ${\\mathcal C}_{\\rm inter} = \\{C_{\\alpha} : \\alpha \\in [0, 1]\\}$, where $C_{\\alpha} \\big |_{\\alpha = 1}$ is the original polar code, and $C_{\\alpha} \\big |_{\\alpha = 0}$ is an RM code. Based on numerical observations, we remark that the error probability under MAP decoding is an increasing function of $\\alpha$. MAP decoding has in general exponential complexity, but empirically the performance of polar codes at finite block lengths is boosted by moving along the family ${\\mathcal C}_{\\rm inter}$ even under low-complexity decoding schemes such as, for instance, belief propagation or successive cancellation list decoder. We demonstrate the performance gain via numerical simulations for transmission over the erasure channel as well as the Gaussian channel."
                    },
                    {
                        "title": "Tight Bounds on the Smallest Eigenvalue of the Neural Tangent Kernel for Deep ReLU Networks",
                        "abstract": "A recent line of work has analyzed the theoretical properties of deep neural networks via the Neural Tangent Kernel (NTK). In particular, the smallest eigenvalue of the NTK has been related to the memorization capacity, the global convergence of gradient descent algorithms and the generalization of deep nets. However, existing results either provide bounds in the two-layer setting or assume that the spectrum of the NTK matrices is bounded away from 0 for multi-layer networks. In this paper, we provide tight bounds on the smallest eigenvalue of NTK matrices for deep ReLU nets, both in the limiting case of infinite widths and for finite widths. In the finite-width setting, the network architectures we consider are fairly general: we require the existence of a wide layer with roughly order of $N$ neurons, $N$ being the number of data samples; and the scaling of the remaining layer widths is arbitrary (up to logarithmic factors). To obtain our results, we analyze various quantities of independent interest: we give lower bounds on the smallest singular value of hidden feature matrices, and upper bounds on the Lipschitz constant of input-output feature maps."
                    },
                    {
                        "title": "Wide Neural Networks Trained with Weight Decay Provably Exhibit Neural Collapse",
                        "abstract": "Deep neural networks (DNNs) at convergence consistently represent the training data in the last layer via a highly symmetric geometric structure referred to as neural collapse. This empirical evidence has spurred a line of theoretical research aimed at proving the emergence of neural collapse, mostly focusing on the unconstrained features model. Here, the features of the penultimate layer are free variables, which makes the model data-agnostic and, hence, puts into question its ability to capture DNN training. Our work addresses the issue, moving away from unconstrained features and studying DNNs that end with at least two linear layers. We first prove generic guarantees on neural collapse that assume (i) low training error and balancedness of the linear layers (for within-class variability collapse), and (ii) bounded conditioning of the features before the linear part (for orthogonality of class-means, as well as their alignment with weight matrices). We then show that such assumptions hold for gradient descent training with weight decay: (i) for networks with a wide first layer, we prove low training error and balancedness, and (ii) for solutions that are either nearly optimal or stable under large learning rates, we additionally prove the bounded conditioning. Taken together, our results are the first to show neural collapse in the end-to-end training of DNNs."
                    },
                    {
                        "title": "Capacity-Achieving Rate-Compatible Polar Codes for General Channels",
                        "abstract": "We present a rate-compatible polar coding scheme that achieves the capacity of any family of channels. Our solution generalizes the previous results [1], [2] that provide capacity-achieving rate-compatible polar codes for a degraded family of channels. The motivation for our extension comes from the fact that in many practical scenarios, e.g., MIMO systems and non-Gaussian interference, the channels cannot be ordered by degradation. The main technical contribution of this paper consists in removing the degradation condition. To do so, we exploit the ideas coming from the construction of universal polar codes.   Our scheme possesses the usual attractive features of polar codes: low complexity code construction, encoding, and decoding; super-polynomial scaling of the error probability with the block length; and absence of error floors. On the negative side, the scaling of the gap to capacity with the block length is slower than in standard polar codes, and we prove an upper bound on the scaling exponent."
                    },
                    {
                        "title": "Landscape Connectivity and Dropout Stability of SGD Solutions for Over-parameterized Neural Networks",
                        "abstract": "The optimization of multilayer neural networks typically leads to a solution with zero training error, yet the landscape can exhibit spurious local minima and the minima can be disconnected. In this paper, we shed light on this phenomenon: we show that the combination of stochastic gradient descent (SGD) and over-parameterization makes the landscape of multilayer neural networks approximately connected and thus more favorable to optimization. More specifically, we prove that SGD solutions are connected via a piecewise linear path, and the increase in loss along this path vanishes as the number of neurons grows large. This result is a consequence of the fact that the parameters found by SGD are increasingly dropout stable as the network becomes wider. We show that, if we remove part of the neurons (and suitably rescale the remaining ones), the change in loss is independent of the total number of neurons, and it depends only on how many neurons are left. Our results exhibit a mild dependence on the input dimension: they are dimension-free for two-layer networks and depend linearly on the dimension for multilayer networks. We validate our theoretical findings with numerical experiments for different architectures and classification tasks."
                    },
                    {
                        "title": "How Spurious Features Are Memorized: Precise Analysis for Random and NTK Features",
                        "abstract": "Deep learning models are known to overfit and memorize spurious features in the training dataset. While numerous empirical studies have aimed at understanding this phenomenon, a rigorous theoretical framework to quantify it is still missing. In this paper, we consider spurious features that are uncorrelated with the learning task, and we provide a precise characterization of how they are memorized via two separate terms: (i) the stability of the model with respect to individual training samples, and (ii) the feature alignment between the spurious feature and the full sample. While the first term is well established in learning theory and it is connected to the generalization error in classical work, the second one is, to the best of our knowledge, novel. Our key technical result gives a precise characterization of the feature alignment for the two prototypical settings of random features (RF) and neural tangent kernel (NTK) regression. We prove that the memorization of spurious features weakens as the generalization capability increases and, through the analysis of the feature alignment, we unveil the role of the model and of its activation function. Numerical experiments show the predictive power of our theory on standard datasets (MNIST, CIFAR-10)."
                    },
                    {
                        "title": "Towards Understanding the Word Sensitivity of Attention Layers: A Study via Random Features",
                        "abstract": "Understanding the reasons behind the exceptional success of transformers requires a better analysis of why attention layers are suitable for NLP tasks. In particular, such tasks require predictive models to capture contextual meaning which often depends on one or few words, even if the sentence is long. Our work studies this key property, dubbed word sensitivity (WS), in the prototypical setting of random features. We show that attention layers enjoy high WS, namely, there exists a vector in the space of embeddings that largely perturbs the random attention features map. The argument critically exploits the role of the softmax in the attention layer, highlighting its benefit compared to other activations (e.g., ReLU). In contrast, the WS of standard random features is of order $1/\\sqrt{n}$, $n$ being the number of words in the textual sample, and thus it decays with the length of the context. We then translate these results on the word sensitivity into generalization bounds: due to their low WS, random features provably cannot learn to distinguish between two sentences that differ only in a single word; in contrast, due to their high WS, random attention features have higher generalization capabilities. We validate our theoretical results with experimental evidence over the BERT-Base word embeddings of the imdb review dataset."
                    },
                    {
                        "title": "Achieving Marton's Region for Broadcast Channels Using Polar Codes",
                        "abstract": "This paper presents polar coding schemes for the 2-user discrete memoryless broadcast channel (DM-BC) which achieve Marton's region with both common and private messages. This is the best achievable rate region known to date, and it is tight for all classes of 2-user DM-BCs whose capacity regions are known. To accomplish this task, we first construct polar codes for both the superposition as well as the binning strategy. By combining these two schemes, we obtain Marton's region with private messages only. Finally, we show how to handle the case of common information. The proposed coding schemes possess the usual advantages of polar codes, i.e., they have low encoding and decoding complexity and a super-polynomial decay rate of the error probability.   We follow the lead of Goela, Abbe, and Gastpar, who recently introduced polar codes emulating the superposition and binning schemes. In order to align the polar indices, for both schemes, their solution involves some degradedness constraints that are assumed to hold between the auxiliary random variables and the channel outputs. To remove these constraints, we consider the transmission of $k$ blocks and employ a chaining construction that guarantees the proper alignment of the polarized indices. The techniques described in this work are quite general, and they can be adopted to many other multi-terminal scenarios whenever there polar indices need to be aligned."
                    },
                    {
                        "title": "How to Achieve the Capacity of Asymmetric Channels",
                        "abstract": "We survey coding techniques that enable reliable transmission at rates that approach the capacity of an arbitrary discrete memoryless channel. In particular, we take the point of view of modern coding theory and discuss how recent advances in coding for symmetric channels help provide more efficient solutions for the asymmetric case. We consider, in more detail, three basic coding paradigms.   The first one is Gallager's scheme that consists of concatenating a linear code with a non-linear mapping so that the input distribution can be appropriately shaped. We explicitly show that both polar codes and spatially coupled codes can be employed in this scenario. Furthermore, we derive a scaling law between the gap to capacity, the cardinality of the input and output alphabets, and the required size of the mapper.   The second one is an integrated scheme in which the code is used both for source coding, in order to create codewords distributed according to the capacity-achieving input distribution, and for channel coding, in order to provide error protection. Such a technique has been recently introduced by Honda and Yamamoto in the context of polar codes, and we show how to apply it also to the design of sparse graph codes.   The third paradigm is based on an idea of B\\\"ocherer and Mathar, and separates the two tasks of source coding and channel coding by a chaining construction that binds together several codewords. We present conditions for the source code and the channel code, and we describe how to combine any source code with any channel code that fulfill those conditions, in order to provide capacity-achieving schemes for asymmetric channels. In particular, we show that polar codes, spatially coupled codes, and homophonic codes are suitable as basic building blocks of the proposed coding strategy."
                    }
                ]
            },
            "e60f775d-ab59-46b7-919e-417465016f8a": {
                "pk": "e60f775d-ab59-46b7-919e-417465016f8a",
                "name": "Mikhail Belkin",
                "collaborators": [
                    "Chaoyue Liu",
                    "Kaushik Sinha",
                    "Like Hui",
                    "Siyuan Ma",
                    "Jihun Hamm",
                    "Alexander Rakhlin",
                    "Alexandre B. Tsybakov",
                    "John Goldsmith",
                    "Xueyuan Zhou",
                    "Nikhil Ghosh"
                ],
                "domain": [
                    "Machine Learning",
                    "Deep Learning",
                    "Kernel Methods",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Fit without fear: remarkable mathematical phenomena of deep learning through the prism of interpolation",
                        "abstract": "In the past decade the mathematical theory of machine learning has lagged far behind the triumphs of deep neural networks on practical challenges. However, the gap between theory and practice is gradually starting to close. In this paper I will attempt to assemble some pieces of the remarkable and still incomplete mathematical mosaic emerging from the efforts to understand the foundations of deep learning. The two key themes will be interpolation, and its sibling, over-parameterization. Interpolation corresponds to fitting data, even noisy data, exactly. Over-parameterization enables interpolation and provides flexibility to select a right interpolating model.   As we will see, just as a physical prism separates colors mixed within a ray of light, the figurative prism of interpolation helps to disentangle generalization and optimization properties within the complex picture of modern Machine Learning. This article is written with belief and hope that clearer understanding of these issues brings us a step closer toward a general theory of deep learning and machine learning."
                    },
                    {
                        "title": "Approximation beats concentration? An approximation view on inference with smooth radial kernels",
                        "abstract": "Positive definite kernels and their associated Reproducing Kernel Hilbert Spaces provide a mathematically compelling and practically competitive framework for learning from data.   In this paper we take the approximation theory point of view to explore various aspects of smooth kernels related to their inferential properties. We analyze eigenvalue decay of kernels operators and matrices, properties of eigenfunctions/eigenvectors and \"Fourier\" coefficients of functions in the kernel space restricted to a discrete set of data points. We also investigate the fitting capacity of kernels, giving explicit bounds on the fat shattering dimension of the balls in Reproducing Kernel Hilbert spaces. Interestingly, the same properties that make kernels very effective approximators for functions in their \"native\" kernel space, also limit their capacity to represent arbitrary functions. We discuss various implications, including those for gradient descent type methods.   It is important to note that most of our bounds are measure independent. Moreover, at least in moderate dimension, the bounds for eigenvalues are much tighter than the bounds which can be obtained from the usual matrix concentration results. For example, we see that the eigenvalues of kernel matrices show nearly exponential decay with constants depending only on the kernel and the domain. We call this \"approximation beats concentration\" phenomenon as even when the data are sampled from a probability distribution, some of their aspects are better understood in terms of approximation theory."
                    },
                    {
                        "title": "Probabilistic Zero-shot Classification with Semantic Rankings",
                        "abstract": "In this paper we propose a non-metric ranking-based representation of semantic similarity that allows natural aggregation of semantic information from multiple heterogeneous sources. We apply the ranking-based representation to zero-shot learning problems, and present deterministic and probabilistic zero-shot classifiers which can be built from pre-trained classifiers without retraining. We demonstrate their the advantages on two large real-world image datasets. In particular, we show that aggregating different sources of semantic information, including crowd-sourcing, leads to more accurate classification."
                    },
                    {
                        "title": "Parametrized Accelerated Methods Free of Condition Number",
                        "abstract": "Analyses of accelerated (momentum-based) gradient descent usually assume bounded condition number to obtain exponential convergence rates. However, in many real problems, e.g., kernel methods or deep neural networks, the condition number, even locally, can be unbounded, unknown or mis-estimated. This poses problems in both implementing and analyzing accelerated algorithms. In this paper, we address this issue by proposing parametrized accelerated methods by considering the condition number as a free parameter. We provide spectral-level analysis for several important accelerated algorithms, obtain explicit expressions and improve worst case convergence rates. Moreover, we show that those algorithm converge exponentially even when the condition number is unknown or mis-estimated."
                    },
                    {
                        "title": "Does data interpolation contradict statistical optimality?",
                        "abstract": "We show that learning methods interpolating the training data can achieve optimal rates for the problems of nonparametric regression and prediction with square loss."
                    },
                    {
                        "title": "Using eigenvectors of the bigram graph to infer morpheme identity",
                        "abstract": "This paper describes the results of some experiments exploring statistical methods to infer syntactic behavior of words and morphemes from a raw corpus in an unsupervised fashion. It shares certain points in common with Brown et al (1992) and work that has grown out of that: it employs statistical techniques to analyze syntactic behavior based on what words occur adjacent to a given word. However, we use an eigenvector decomposition of a nearest-neighbor graph to produce a two-dimensional rendering of the words of a corpus in which words of the same syntactic category tend to form neighborhoods. We exploit this technique for extending the value of automatic learning of morphology. In particular, we look at the suffixes derived from a corpus by unsupervised learning of morphology, and we ask which of these suffixes have a consistent syntactic function (e.g., in English, -tion is primarily a mark of nouns, but -s marks both noun plurals and 3rd person present on verbs), and we determine that this method works well for this task."
                    },
                    {
                        "title": "Polynomial Learning of Distribution Families",
                        "abstract": "The question of polynomial learnability of probability distributions, particularly Gaussian mixture distributions, has recently received significant attention in theoretical computer science and machine learning. However, despite major progress, the general question of polynomial learnability of Gaussian mixture distributions still remained open. The current work resolves the question of polynomial learnability for Gaussian mixtures in high dimension with an arbitrary fixed number of components. The result on learning Gaussian mixtures relies on an analysis of distributions belonging to what we call \"polynomial families\" in low dimension. These families are characterized by their moments being polynomial in parameters and include almost all common probability distributions as well as their mixtures and products. Using tools from real algebraic geometry, we show that parameters of any distribution belonging to such a family can be learned in polynomial time and using a polynomial number of sample points. The result on learning polynomial families is quite general and is of independent interest. To estimate parameters of a Gaussian mixture distribution in high dimensions, we provide a deterministic algorithm for dimensionality reduction. This allows us to reduce learning a high-dimensional mixture to a polynomial number of parameter estimations in low dimension. Combining this reduction with the results on polynomial families yields our result on learning arbitrary Gaussian mixtures in high dimensions."
                    },
                    {
                        "title": "Behavior of Graph Laplacians on Manifolds with Boundary",
                        "abstract": "In manifold learning, algorithms based on graph Laplacians constructed from data have received considerable attention both in practical applications and theoretical analysis. In particular, the convergence of graph Laplacians obtained from sampled data to certain continuous operators has become an active research topic recently. Most of the existing work has been done under the assumption that the data is sampled from a manifold without boundary or that the functions of interests are evaluated at a point away from the boundary. However, the question of boundary behavior is of considerable practical and theoretical interest. In this paper we provide an analysis of the behavior of graph Laplacians at a point near or on the boundary, discuss their convergence rates and their implications and provide some numerical results. It turns out that while points near the boundary occupy only a small part of the total volume of a manifold, the behavior of graph Laplacian there has different scaling properties from its behavior elsewhere on the manifold, with global effects on the whole manifold, an observation with potentially important implications for the general problem of learning on manifolds."
                    },
                    {
                        "title": "A Universal Trade-off Between the Model Size, Test Loss, and Training Loss of Linear Predictors",
                        "abstract": "In this work we establish an algorithm and distribution independent non-asymptotic trade-off between the model size, excess test loss, and training loss of linear predictors. Specifically, we show that models that perform well on the test data (have low excess loss) are either \"classical\" -- have training loss close to the noise level, or are \"modern\" -- have a much larger number of parameters compared to the minimum needed to fit the training data exactly.   We also provide a more precise asymptotic analysis when the limiting spectral distribution of the whitened features is Marchenko-Pastur. Remarkably, while the Marchenko-Pastur analysis is far more precise near the interpolation peak, where the number of parameters is just enough to fit the training data, it coincides exactly with the distribution independent bound as the level of overparametrization increases."
                    },
                    {
                        "title": "Cut your Losses with Squentropy",
                        "abstract": "Nearly all practical neural models for classification are trained using cross-entropy loss. Yet this ubiquitous choice is supported by little theoretical or empirical evidence. Recent work (Hui & Belkin, 2020) suggests that training using the (rescaled) square loss is often superior in terms of the classification accuracy. In this paper we propose the \"squentropy\" loss, which is the sum of two terms: the cross-entropy loss and the average square loss over the incorrect classes. We provide an extensive set of experiments on multi-class classification problems showing that the squentropy loss outperforms both the pure cross entropy and rescaled square losses in terms of the classification accuracy. We also demonstrate that it provides significantly better model calibration than either of these alternative losses and, furthermore, has less variance with respect to the random initialization. Additionally, in contrast to the square loss, squentropy loss can typically be trained using exactly the same optimization parameters, including the learning rate, as the standard cross-entropy loss, making it a true \"plug-and-play\" replacement. Finally, unlike the rescaled square loss, multiclass squentropy contains no parameters that need to be adjusted."
                    },
                    {
                        "title": "Learning Gaussian Mixtures with Arbitrary Separation",
                        "abstract": "In this paper we present a method for learning the parameters of a mixture of $k$ identical spherical Gaussians in $n$-dimensional space with an arbitrarily small separation between the components. Our algorithm is polynomial in all parameters other than $k$. The algorithm is based on an appropriate grid search over the space of parameters. The theoretical analysis of the algorithm hinges on a reduction of the problem to 1 dimension and showing that two 1-dimensional mixtures whose densities are close in the $L^2$ norm must have similar means and mixing coefficients. To produce such a lower bound for the $L^2$ norm in terms of the distances between the corresponding means, we analyze the behavior of the Fourier transform of a mixture of Gaussians in 1 dimension around the origin, which turns out to be closely related to the properties of the Vandermonde matrix obtained from the component means. Analysis of this matrix together with basic function approximation results allows us to provide a lower bound for the norm of the mixture in the Fourier domain.   In recent years much research has been aimed at understanding the computational aspects of learning parameters of Gaussians mixture distributions in high dimension. To the best of our knowledge all existing work on learning parameters of Gaussian mixtures assumes minimum separation between components of the mixture which is an increasing function of either the dimension of the space $n$ or the number of components $k$. In our paper we prove the first result showing that parameters of a $n$-dimensional Gaussian mixture model with arbitrarily small component separation can be learned in time polynomial in $n$."
                    },
                    {
                        "title": "Laplacian Support Vector Machines Trained in the Primal",
                        "abstract": "In the last few years, due to the growing ubiquity of unlabeled data, much effort has been spent by the machine learning community to develop better understanding and improve the quality of classifiers exploiting unlabeled data. Following the manifold regularization approach, Laplacian Support Vector Machines (LapSVMs) have shown the state of the art performance in semi--supervised classification. In this paper we present two strategies to solve the primal LapSVM problem, in order to overcome some issues of the original dual formulation. Whereas training a LapSVM in the dual requires two steps, using the primal form allows us to collapse training to a single step. Moreover, the computational complexity of the training algorithm is reduced from O(n^3) to O(n^2) using preconditioned conjugate gradient, where n is the combined number of labeled and unlabeled examples. We speed up training by using an early stopping strategy based on the prediction on unlabeled data or, if available, on labeled validation examples. This allows the algorithm to quickly compute approximate solutions with roughly the same classification accuracy as the optimal ones, considerably reducing the training time. Due to its simplicity, training LapSVM in the primal can be the starting point for additional enhancements of the original LapSVM formulation, such as those for dealing with large datasets. We present an extensive experimental evaluation on real world data showing the benefits of the proposed approach."
                    },
                    {
                        "title": "Accelerating SGD with momentum for over-parameterized learning",
                        "abstract": "Nesterov SGD is widely used for training modern neural networks and other machine learning models. Yet, its advantages over SGD have not been theoretically clarified. Indeed, as we show in our paper, both theoretically and empirically, Nesterov SGD with any parameter selection does not in general provide acceleration over ordinary SGD. Furthermore, Nesterov SGD may diverge for step sizes that ensure convergence of ordinary SGD. This is in contrast to the classical results in the deterministic scenario, where the same step size ensures accelerated convergence of the Nesterov's method over optimal gradient descent.   To address the non-acceleration issue, we introduce a compensation term to Nesterov SGD. The resulting algorithm, which we call MaSS, converges for same step sizes as SGD. We prove that MaSS obtains an accelerated convergence rates over SGD for any mini-batch size in the linear setting. For full batch, the convergence rate of MaSS matches the well-known accelerated rate of the Nesterov's method.   We also analyze the practically important question of the dependence of the convergence rate and optimal hyper-parameters on the mini-batch size, demonstrating three distinct regimes: linear scaling, diminishing returns and saturation.   Experimental evaluation of MaSS for several standard architectures of deep networks, including ResNet and convolutional networks, shows improved performance over SGD, Nesterov SGD and Adam."
                    },
                    {
                        "title": "Evaluation of Neural Architectures Trained with Square Loss vs Cross-Entropy in Classification Tasks",
                        "abstract": "Modern neural architectures for classification tasks are trained using the cross-entropy loss, which is widely believed to be empirically superior to the square loss. In this work we provide evidence indicating that this belief may not be well-founded. We explore several major neural architectures and a range of standard benchmark datasets for NLP, automatic speech recognition (ASR) and computer vision tasks to show that these architectures, with the same hyper-parameter settings as reported in the literature, perform comparably or better when trained with the square loss, even after equalizing computational resources. Indeed, we observe that the square loss produces better results in the dominant majority of NLP and ASR experiments. Cross-entropy appears to have a slight edge on computer vision tasks.   We argue that there is little compelling empirical or theoretical evidence indicating a clear-cut advantage to the cross-entropy loss. Indeed, in our experiments, performance on nearly all non-vision tasks can be improved, sometimes significantly, by switching to the square loss. Furthermore, training with square loss appears to be less sensitive to the randomness in initialization. We posit that training using the square loss for classification needs to be a part of best practices of modern deep learning on equal footing with cross-entropy."
                    },
                    {
                        "title": "Diving into the shallows: a computational perspective on large-scale shallow learning",
                        "abstract": "In this paper we first identify a basic limitation in gradient descent-based optimization methods when used in conjunctions with smooth kernels. An analysis based on the spectral properties of the kernel demonstrates that only a vanishingly small portion of the function space is reachable after a polynomial number of gradient descent iterations. This lack of approximating power drastically limits gradient descent for a fixed computational budget leading to serious over-regularization/underfitting. The issue is purely algorithmic, persisting even in the limit of infinite data.   To address this shortcoming in practice, we introduce EigenPro iteration, based on a preconditioning scheme using a small number of approximately computed eigenvectors. It can also be viewed as learning a new kernel optimized for gradient descent. It turns out that injecting this small (computationally inexpensive and SGD-compatible) amount of approximate second-order information leads to major improvements in convergence. For large data, this translates into significant performance boost over the standard kernel methods. In particular, we are able to consistently match or improve the state-of-the-art results recently reported in the literature with a small fraction of their computational budget.   Finally, we feel that these results show a need for a broader computational perspective on modern large-scale learning to complement more traditional statistical and convergence analyses. In particular, many phenomena of large-scale high-dimensional inference are best understood in terms of optimization on infinite dimensional Hilbert spaces, where standard algorithms can sometimes have properties at odds with finite-dimensional intuition. A systematic analysis concentrating on the approximation power of such algorithms within a budget of computation may lead to progress both in theory and practice."
                    },
                    {
                        "title": "Kernel machines that adapt to GPUs for effective large batch training",
                        "abstract": "Modern machine learning models are typically trained using Stochastic Gradient Descent (SGD) on massively parallel computing resources such as GPUs. Increasing mini-batch size is a simple and direct way to utilize the parallel computing capacity. For small batch an increase in batch size results in the proportional reduction in the training time, a phenomenon known as linear scaling. However, increasing batch size beyond a certain value leads to no further improvement in training time. In this paper we develop the first analytical framework that extends linear scaling to match the parallel computing capacity of a resource. The framework is designed for a class of classical kernel machines. It automatically modifies a standard kernel machine to output a mathematically equivalent prediction function, yet allowing for extended linear scaling, i.e., higher effective parallelization and faster training time on given hardware.   The resulting algorithms are accurate, principled and very fast. For example, using a single Titan Xp GPU, training on ImageNet with $1.3\\times 10^6$ data points and $1000$ labels takes under an hour, while smaller datasets, such as MNIST, take seconds. As the parameters are chosen analytically, based on the theoretical bounds, little tuning beyond selecting the kernel and the kernel parameter is needed, further facilitating the practical use of these methods."
                    },
                    {
                        "title": "Local Quadratic Convergence of Stochastic Gradient Descent with Adaptive Step Size",
                        "abstract": "Establishing a fast rate of convergence for optimization methods is crucial to their applicability in practice. With the increasing popularity of deep learning over the past decade, stochastic gradient descent and its adaptive variants (e.g. Adagrad, Adam, etc.) have become prominent methods of choice for machine learning practitioners. While a large number of works have demonstrated that these first order optimization methods can achieve sub-linear or linear convergence, we establish local quadratic convergence for stochastic gradient descent with adaptive step size for problems such as matrix inversion."
                    },
                    {
                        "title": "Transition to Linearity of General Neural Networks with Directed Acyclic Graph Architecture",
                        "abstract": "In this paper we show that feedforward neural networks corresponding to arbitrary directed acyclic graphs undergo transition to linearity as their \"width\" approaches infinity. The width of these general networks is characterized by the minimum in-degree of their neurons, except for the input and first layers. Our results identify the mathematical structure underlying transition to linearity and generalize a number of recent works aimed at characterizing transition to linearity or constancy of the Neural Tangent Kernel for standard architectures."
                    },
                    {
                        "title": "Inverse Density as an Inverse Problem: The Fredholm Equation Approach",
                        "abstract": "In this paper we address the problem of estimating the ratio $\\frac{q}{p}$ where $p$ is a density function and $q$ is another density, or, more generally an arbitrary function. Knowing or approximating this ratio is needed in various problems of inference and integration, in particular, when one needs to average a function with respect to one probability distribution, given a sample from another. It is often referred as {\\it importance sampling} in statistical inference and is also closely related to the problem of {\\it covariate shift} in transfer learning as well as to various MCMC methods. It may also be useful for separating the underlying geometry of a space, say a manifold, from the density function defined on it.   Our approach is based on reformulating the problem of estimating $\\frac{q}{p}$ as an inverse problem in terms of an integral operator corresponding to a kernel, and thus reducing it to an integral equation, known as the Fredholm problem of the first kind. This formulation, combined with the techniques of regularization and kernel methods, leads to a principled kernel-based framework for constructing algorithms and for analyzing them theoretically.   The resulting family of algorithms (FIRE, for Fredholm Inverse Regularized Estimator) is flexible, simple and easy to implement.   We provide detailed theoretical analysis including concentration bounds and convergence rates for the Gaussian kernel in the case of densities defined on $\\R^d$, compact domains in $\\R^d$ and smooth $d$-dimensional sub-manifolds of the Euclidean space.   We also show experimental results including applications to classification and semi-supervised learning within the covariate shift framework and demonstrate some encouraging experimental comparisons. We also show how the parameters of our algorithms can be chosen in a completely unsupervised manner."
                    }
                ]
            }
        }
    },
    "2406.03417": {
        "paper_data": {
            "title": "CoFie: Learning Compact Neural Surface Representations with Coordinate Fields",
            "url": "http://arxiv.org/abs/2406.03417v1",
            "arxiv_id": "2406.03417",
            "authors": [
                "Hanwen Jiang",
                "Haitao Yang",
                "Georgios Pavlakos",
                "Qixing Huang"
            ],
            "abstract": "This paper introduces CoFie, a novel local geometry-aware neural surface representation. CoFie is motivated by the theoretical analysis of local SDFs with quadratic approximation. We find that local shapes are highly compressive in an aligned coordinate frame defined by the normal and tangent directions of local shapes. Accordingly, we introduce Coordinate Field, which is a composition of coordinate frames of all local shapes. The Coordinate Field is optimizable and is used to transform the local shapes from the world coordinate frame to the aligned shape coordinate frame. It largely reduces the complexity of local shapes and benefits the learning of MLP-based implicit representations. Moreover, we introduce quadratic layers into the MLP to enhance expressiveness concerning local shape geometry. CoFie is a generalizable surface representation. It is trained on a curated set of 3D shapes and works on novel shape instances during testing. When using the same amount of parameters with prior works, CoFie reduces the shape error by 48% and 56% on novel instances of both training and unseen shape categories. Moreover, CoFie demonstrates comparable performance to prior works when using only 70% fewer parameters.",
            "introduction": "   1 Introduction  In the realm of geometry modeling, neural implicit shape representations have become a powerful tool\u00a0[28, 7, 4, 12, 35, 33, 2]. These representations typically use latent codes to represent shapes and employ multilayer perceptions (MLPs) to decode their Signed Distance Functions (SDFs). Early works in this field use a single latent code to represent an entire shape\u00a0[28]. Nevertheless, the decoded SDFs usually lack geometry details. To improve the shape modeling quality, recent approaches have introduced local-based designs\u00a0[4, 23, 36]. By decomposing an entire shape into many local surfaces, the shape modeling task becomes effortless \u2013 local surfaces are in simpler geometry which are easier to represent. Despite the progress, the local-aware design significantly increases the number of parameters, as each local surface is represented by one or even multiple latent codes. Thus, proposing a neural surface representation that is both accurate and compact is necessary.   To achieve this goal, we argue it is important to understand the properties of local surfaces. Following prior works\u00a0[26, 42, 9, 38], we approximate the local geometry with quadratic patches\u00a0[8] and perform analysis. Results show the feasibility of fitting the geometry of a specific category of quadratic patches. In detail, the quadratic patches are aligned with the coordinate system defined by the normal, principal directions, and principal curvatures of quadratic patch\u00a0[8, 26]. However, when the quadratic patches are not aligned \u2013 they are freely transformed with random rotations and translations in 3D, mimicking real local surfaces \u2013 the optimization will be easily trapped into local minima. This analysis reveals the difficulty of jointly recovering transformation information and geometry of local patches.   Figure 1: CoFie is a local geometry-aware shape representation. (Left) CoFie divides a shape into non-overlapping local patches, where each local patch is represented by an MLP-based Signed Distance Function. (Right) CoFie introduces Coordinate Field, which attaches a coordinate frame to each local patch. It transforms local patches from the world coordinate system to an aligned coordinate system, reducing shape complexity.   Based on the analysis, we propose CoFie, a novel local geometry-aware neural surface representation. The key insight of CoFie is decomposing the transformation information of local shapes from its geometry. As shown in Fig.\u00a01, we associate each local surface with a learnable coordinate frame, which forms a Coordinate Field. We use the Coordinate Field to transform all local surfaces into an aligned coordinate system, reducing their spatial complexity. Thus, the geometry space of local surfaces becomes more compact, where the MLP-based neural SDFs are easier to learn.   An important design aspect is how to represent the Coordinate Field. Departing from the implicit-based representations, we use an explicit representation. Specifically, the coordinate frame of each local surface is parameterized by a rotation and a translation, forming a 6 Degree-of-Freedom pose. Moreover, we initialize the rotation using the estimated normal, principal direction, and principal curvature of a local surface. This design makes CoFie local geometry-aware and facilitates the learning of Coordinate Fields.   To better represent local surfaces\u2019 geometry, we introduce quadratic layers to the MLP. Prior works typically employ ReLU-based MLP with shallow layers and limited hidden size\u00a0[28, 4]. Thus, the MLP is",
            "references": [
                {
                    "title": "LRM: Large Reconstruction Model for Single Image to 3D",
                    "abstract": "We propose the first Large Reconstruction Model (LRM) that predicts the 3D model of an object from a single input image within just 5 seconds. In contrast to many previous methods that are trained on small-scale datasets such as ShapeNet in a category-specific fashion, LRM adopts a highly scalable transformer-based architecture with 500 million learnable parameters to directly predict a neural radiance field (NeRF) from the input image. We train our model in an end-to-end manner on massive multi-view data containing around 1 million objects, including both synthetic renderings from Objaverse and real captures from MVImgNet. This combination of a high-capacity model and large-scale training data empowers our model to be highly generalizable and produce high-quality 3D reconstructions from various testing inputs, including real-world in-the-wild captures and images created by generative models. Video demos and interactable 3D meshes can be found on our LRM project webpage: https://yiconghong.me/LRM."
                },
                {
                    "title": "LEAP: Liberate Sparse-view 3D Modeling from Camera Poses",
                    "abstract": "Are camera poses necessary for multi-view 3D modeling? Existing approaches predominantly assume access to accurate camera poses. While this assumption might hold for dense views, accurately estimating camera poses for sparse views is often elusive. Our analysis reveals that noisy estimated poses lead to degraded performance for existing sparse-view 3D modeling methods. To address this issue, we present LEAP, a novel pose-free approach, therefore challenging the prevailing notion that camera poses are indispensable. LEAP discards pose-based operations and learns geometric knowledge from data. LEAP is equipped with a neural volume, which is shared across scenes and is parameterized to encode geometry and texture priors. For each incoming scene, we update the neural volume by aggregating 2D image features in a feature-similarity-driven manner. The updated neural volume is decoded into the radiance field, enabling novel view synthesis from any viewpoint. On both object-centric and scene-level datasets, we show that LEAP significantly outperforms prior methods when they employ predicted poses from state-of-the-art pose estimators. Notably, LEAP performs on par with prior approaches that use ground-truth poses while running $400\\times$ faster than PixelNeRF. We show LEAP generalizes to novel object categories and scenes, and learns knowledge closely resembles epipolar geometry. Project page: https://hwjiang1510.github.io/LEAP/"
                },
                {
                    "title": "Neural-Singular-Hessian: Implicit Neural Representation of Unoriented Point Clouds by Enforcing Singular Hessian",
                    "abstract": "Neural implicit representation is a promising approach for reconstructing surfaces from point clouds. Existing methods combine various regularization terms, such as the Eikonal and Laplacian energy terms, to enforce the learned neural function to possess the properties of a Signed Distance Function (SDF). However, inferring the actual topology and geometry of the underlying surface from poor-quality unoriented point clouds remains challenging. In accordance with Differential Geometry, the Hessian of the SDF is singular for points within the differential thin-shell space surrounding the surface. Our approach enforces the Hessian of the neural implicit function to have a zero determinant for points near the surface. This technique aligns the gradients for a near-surface point and its on-surface projection point, producing a rough but faithful shape within just a few iterations. By annealing the weight of the singular-Hessian term, our approach ultimately produces a high-fidelity reconstruction result. Extensive experimental results demonstrate that our approach effectively suppresses ghost geometry and recovers details from unoriented point clouds with better expressiveness than existing fitting-based methods."
                },
                {
                    "title": "3DShape2VecSet: A 3D Shape Representation for Neural Fields and Generative Diffusion Models",
                    "abstract": "We introduce 3DShape2VecSet, a novel shape representation for neural fields designed for generative diffusion models. Our shape representation can encode 3D shapes given as surface models or point clouds, and represents them as neural fields. The concept of neural fields has previously been combined with a global latent vector, a regular grid of latent vectors, or an irregular grid of latent vectors. Our new representation encodes neural fields on top of a set of vectors. We draw from multiple concepts, such as the radial basis function representation, and the cross attention and self-attention function, to design a learnable representation that is especially suitable for processing with transformers. Our results show improved performance in 3D shape encoding and 3D shape generative modeling tasks. We demonstrate a wide variety of generative applications: unconditioned generation, category-conditioned generation, text-conditioned generation, point-cloud completion, and image-conditioned generation. Code: https://1zb.github.io/3DShape2VecSet/."
                },
                {
                    "title": "Few-View Object Reconstruction with Unknown Categories and Camera Poses",
                    "abstract": "While object reconstruction has made great strides in recent years, current methods typically require densely captured images and/or known camera poses, and generalize poorly to novel object categories. To step toward object reconstruction in the wild, this work explores reconstructing general real-world objects from a few images without known camera poses or object categories. The crux of our work is solving two fundamental 3D vision problems \u2014 shape reconstruction and pose estimation \u2014 in a unified approach. Our approach captures the synergies of these two problems: reliable camera pose estimation gives rise to accurate shape reconstruction, and the accurate reconstruction, in turn, induces robust correlations between different views and facilitates pose estimation. Our method FORGE predicts 3D features from each view and leverages them in conjunction with the input images to establish cross-view correlations for estimating relative camera poses. The 3D features are then transformed by the estimated poses into a shared space and are fused into a neural radiance field. The reconstruction results are rendered by volume rendering techniques, enabling us to train the model without 3D shape groundtruth. Our experiments on both real and synthetic datasets, as well as in the wild images, show that FORGE reliably reconstructs objects from five views. Our pose estimation method outperforms existing ones by a large margin. The reconstruction results under predicted poses are comparable to the ones using ground-truth poses. And the performance on novel testing categories matches the results on categories seen during training."
                },
                {
                    "title": "Dual octree graph networks for learning adaptive volumetric shape representations",
                    "abstract": "We present an adaptive deep representation of volumetric fields of 3D shapes and an efficient approach to learn this deep representation for high-quality 3D shape reconstruction and auto-encoding. Our method encodes the volumetric field of a 3D shape with an adaptive feature volume organized by an octree and applies a compact multilayer perceptron network for mapping the features to the field value at each 3D position. An encoder-decoder network is designed to learn the adaptive feature volume based on the graph convolutions over the dual graph of octree nodes. The core of our network is a new graph convolution operator defined over a regular grid of features fused from irregular neighboring octree nodes at different levels, which not only reduces the computational and memory cost of the convolutions over irregular neighboring octree nodes, but also improves the performance of feature learning. Our method effectively encodes shape details, enables fast 3D shape reconstruction, and exhibits good generality for modeling 3D shapes out of training categories. We evaluate our method on a set of reconstruction tasks of 3D shapes and scenes and validate its superiority over other existing approaches. Our code, data, and trained models are available at https://wang-ps.github.io/dualocnn."
                },
                {
                    "title": "FIRe: Fast Inverse Rendering using Directional and Signed Distance Functions",
                    "abstract": "Neural 3D implicit representations learn priors that are useful for diverse applications, such as single- or multiple-view 3D reconstruction. A major downside of existing approaches while rendering an image is that they require evaluating the network multiple times per camera ray so that the high computational time forms a bottleneck for downstream applications. We address this problem by introducing a novel neural scene representation that we call the directional distance function (DDF). To this end, we learn a signed distance function (SDF) along with our DDF model to represent a class of shapes. Specifically, our DDF is defined on the unit sphere and predicts the distance to the surface along any given direction. Therefore, our DDF allows rendering images with just a single network evaluation per camera ray. Based on our DDF, we present a novel fast algorithm (FIRe) to reconstruct 3D shapes given a posed depth map. We evaluate our proposed method on 3D reconstruction from single-view depth images, where we empirically show that our algorithm reconstructs 3D shapes more accurately and it is more than 15 times faster (per iteration) than competing methods."
                },
                {
                    "title": "Efficient Geometry-aware 3D Generative Adversarial Networks",
                    "abstract": "Unsupervised generation of high-quality multi-view-consistent images and 3D shapes using only collections of single-view 2D photographs has been a long-standing challenge. Existing 3D GANs are either compute intensive or make approximations that are not 3D-consistent; the former limits quality and resolution of the generated images and the latter adversely affects multi-view consistency and shape quality. In this work, we improve the computational efficiency and image quality of 3D GANs without overly relying on these approximations. We introduce an expressive hybrid explicit implicit network architecture that, together with other design choices, synthesizes not only high-resolution multi-view-consistent images in real time but also produces high-quality 3D geometry. By decoupling feature generation and neural rendering, our framework is able to leverage state-of-the-art 2D CNN generators, such as StyleGAN2, and inherit their efficiency and expressiveness. We demonstrate state-of-the-art 3D-aware synthesis with FFHQ and AFHQ Cats, among other experiments."
                },
                {
                    "title": "Representing 3D Shapes with Probabilistic Directed Distance Fields",
                    "abstract": "Differentiable rendering is an essential operation in modern vision, allowing inverse graphics approaches to 3D understanding to be utilized in modern machine learning frameworks. Explicit shape representations (voxels, point clouds, or meshes), while relatively easily rendered, often suffer from limited geometric fidelity or topological con-straints. On the other hand, implicit representations (occu-pancy, distance, or radiance fields) preserve greater fidelity, but suffer from complex or inefficient rendering processes, limiting scalability. In this work, we endeavour to address both shortcomings with a novel shape representation that allows fast differentiable rendering within an implicit ar-chitecture. Building on implicit distance representations, we define Directed Distance Fields (DDFs), which map an oriented point (position and direction) to surface visibility and depth. Such a field can render a depth map with a single forward pass per pixel, enable differential surface geometry extraction (e.g., surface normals and curvatures) via network derivatives, be easily composed, and permit extraction of classical unsigned distance fields. Using probabilistic DDFs (PDDFs), we show how to model inherent discontinuities in the underlying field. Finally, we apply our method to fitting single shapes, unpaired 3D-aware generative image modelling, and single-image 3D reconstruction tasks, showcasing strong performance with simple architectural components via the versatility of our representation."
                },
                {
                    "title": "Deep Marching Tetrahedra: a Hybrid Representation for High-Resolution 3D Shape Synthesis",
                    "abstract": "We introduce DMTet, a deep 3D conditional generative model that can synthesize high-resolution 3D shapes using simple user guides such as coarse voxels. It marries the merits of implicit and explicit 3D representations by leveraging a novel hybrid 3D representation. Compared to the current implicit approaches, which are trained to regress the signed distance values, DMTet directly optimizes for the reconstructed surface, which enables us to synthesize finer geometric details with fewer artifacts. Unlike deep 3D generative models that directly generate explicit representations such as meshes, our model can synthesize shapes with arbitrary topology. The core of DMTet includes a deformable tetrahedral grid that encodes a discretized signed distance function and a differentiable marching tetrahedra layer that converts the implicit signed distance representation to the explicit surface mesh representation. This combination allows joint optimization of the surface geometry and topology as well as generation of the hierarchy of subdivisions using reconstruction and adversarial losses defined explicitly on the surface mesh. Our approach significantly outperforms existing work on conditional shape synthesis from coarse voxel inputs, trained on a dataset of complex 3D animal shapes. Project page: https://nv-tlabs.github.io/DMTet/."
                },
                {
                    "title": "OctField: Hierarchical Implicit Functions for 3D Modeling",
                    "abstract": "Recent advances in localized implicit functions have enabled neural implicit representation to be scalable to large scenes. However, the regular subdivision of 3D space employed by these approaches fails to take into account the sparsity of the surface occupancy and the varying granularities of geometric details. As a result, its memory footprint grows cubically with the input volume, leading to a prohibitive computational cost even at a moderately dense decomposition. In this work, we present a learnable hierarchical implicit representation for 3D surfaces, coded OctField, that allows high-precision encoding of intricate surfaces with low memory and computational budget. The key to our approach is an adaptive decomposition of 3D scenes that only distributes local implicit functions around the surface of interest. We achieve this goal by introducing a hierarchical octree structure to adaptively subdivide the 3D space according to the surface occupancy and the richness of part geometry. As octree is discrete and non-differentiable, we further propose a novel hierarchical network that models the subdivision of octree cells as a probabilistic process and recursively encodes and decodes both octree structure and surface geometry in a differentiable manner. We demonstrate the value of OctField for a range of shape modeling and reconstruction tasks, showing superiority over alternative approaches."
                },
                {
                    "title": "Mixture of volumetric primitives for efficient neural rendering",
                    "abstract": "Real-time rendering and animation of humans is a core function in games, movies, and telepresence applications. Existing methods have a number of drawbacks we aim to address with our work. Triangle meshes have difficulty modeling thin structures like hair, volumetric representations like Neural Volumes are too low-resolution given a reasonable memory budget, and high-resolution implicit representations like Neural Radiance Fields are too slow for use in real-time applications. We present Mixture of Volumetric Primitives (MVP), a representation for rendering dynamic 3D content that combines the completeness of volumetric representations with the efficiency of primitive-based rendering, e.g., point-based or mesh-based methods. Our approach achieves this by leveraging spatially shared computation with a convolutional architecture and by minimizing computation in empty regions of space with volumetric primitives that can move to cover only occupied regions. Our parameterization supports the integration of correspondence and tracking constraints, while being robust to areas where classical tracking fails, such as around thin or translucent structures and areas with large topological variability. MVP is a hybrid that generalizes both volumetric and primitive-based representations. Through a series of extensive experiments we demonstrate that it inherits the strengths of each, while avoiding many of their limitations. We also compare our approach to several state-of-the-art methods and demonstrate that MVP produces superior results in terms of quality and runtime performance."
                },
                {
                    "title": "Neural Geometric Level of Detail: Real-time Rendering with Implicit 3D Shapes",
                    "abstract": "Neural signed distance functions (SDFs) are emerging as an effective representation for 3D shapes. State-of-the-art methods typically encode the SDF with a large, fixed-size neural network to approximate complex shapes with implicit surfaces. Rendering with these large networks is, however, computationally expensive since it requires many forward passes through the network for every pixel, making these representations impractical for real-time graphics. We introduce an efficient neural representation that, for the first time, enables real-time rendering of high-fidelity neural SDFs, while achieving state-of-the-art geometry reconstruction quality. We represent implicit surfaces using an octree-based feature volume which adaptively fits shapes with multiple discrete levels of detail (LODs), and enables continuous LOD with SDF interpolation. We further develop an efficient algorithm to directly render our novel neural SDF representation in real-time by querying only the necessary LODs with sparse octree traversal. We show that our representation is 2\u20133 orders of magnitude more efficient in terms of rendering speed compared to previous works. Furthermore, it produces state-of-the-art reconstruction quality for complex shapes under both 3D geometric and 2D image-space metrics."
                },
                {
                    "title": "Deep Implicit Templates for 3D Shape Representation",
                    "abstract": "Deep implicit functions (DIFs), as a kind of 3D shape representation, are becoming more and more popular in the 3D vision community due to their compactness and strong representation power. However, unlike polygon mesh-based templates, it remains a challenge to reason dense correspondences or other semantic relationships across shapes represented by DIFs, which limits its applications in texture transfer, shape analysis and so on. To overcome this limitation and also make DIFs more interpretable, we propose Deep Implicit Templates, a new 3D shape representation that supports explicit correspondence reasoning in deep implicit representations. Our key idea is to formulate DIFs as conditional deformations of a template implicit function. To this end, we propose Spatial Warping LSTM, which de-composes the conditional spatial transformation into multiple point-wise transformations and guarantees generalization capability. Moreover, the training loss is carefully designed in order to achieve high reconstruction accuracy while learning a plausible template with accurate correspondences in an unsupervised manner. Experiments show that our method can not only learn a common implicit tem-plate for a collection of shapes, but also establish dense correspondences across all the shapes simultaneously with-out any supervision."
                },
                {
                    "title": "Constructing Multilayer Perceptrons as Piecewise Low-Order Polynomial Approximators: A Signal Processing Approach",
                    "abstract": "The construction of a multilayer perceptron (MLP) as a piecewise low-order polynomial approximator using a signal processing approach is presented in this work. The constructed MLP contains one input, one intermediate and one output layers. Its construction includes the specification of neuron numbers and all filter weights. Through the construction, a one-to-one correspondence between the approximation of an MLP and that of a piecewise low-order polynomial is established. Comparison between piecewise polynomial and MLP approximations is made. Since the approximation capability of piecewise low-order polynomials is well understood, our findings shed light on the universal approximation capability of an MLP."
                },
                {
                    "title": "DSG-Net: Learning Disentangled Structure and Geometry for 3D Shape Generation",
                    "abstract": "3D shape generation is a fundamental operation in computer graphics. While significant progress has been made, especially with recent deep generative models, it remains a challenge to synthesize high-quality shapes with rich geometric details and complex structures, in a controllable manner. To tackle this, we introduce DSG-Net, a deep neural network that learns a disentangled structured & geometric mesh representation for 3D shapes, where two key aspects of shapes, geometry and structure, are encoded in a synergistic manner to ensure plausibility of the generated shapes, while also being disentangled as much as possible. This supports a range of novel shape generation applications with disentangled control, such as interpolation of structure (geometry) while keeping geometry (structure) unchanged. To achieve this, we simultaneously learn structure and geometry through variational autoencoders (VAEs) in a hierarchical manner for both, with bijective mappings at each level. In this manner, we effectively encode geometry and structure in separate latent spaces, while ensuring their compatibility: the structure is used to guide the geometry and vice versa. At the leaf level, the part geometry is represented using a conditional part VAE, to encode high-quality geometric details, guided by the structure context as the condition. Our method not only supports controllable generation applications, but also produces high-quality synthesized shapes, outperforming state-of-the-art methods."
                },
                {
                    "title": "Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains",
                    "abstract": "We show that passing input points through a simple Fourier feature mapping enables a multilayer perceptron (MLP) to learn high-frequency functions in low-dimensional problem domains. These results shed light on recent advances in computer vision and graphics that achieve state-of-the-art results by using MLPs to represent complex 3D objects and scenes. Using tools from the neural tangent kernel (NTK) literature, we show that a standard MLP fails to learn high frequencies both in theory and in practice. To overcome this spectral bias, we use a Fourier feature mapping to transform the effective NTK into a stationary kernel with a tunable bandwidth. We suggest an approach for selecting problem-specific Fourier features that greatly improves the performance of MLPs for low-dimensional regression tasks relevant to the computer vision and graphics communities."
                },
                {
                    "title": "Implicit Neural Representations with Periodic Activation Functions",
                    "abstract": "Implicitly defined, continuous, differentiable signal representations parameterized by neural networks have emerged as a powerful paradigm, offering many possible benefits over conventional representations. However, current network architectures for such implicit neural representations are incapable of modeling signals with fine detail, and fail to represent a signal's spatial and temporal derivatives, despite the fact that these are essential to many physical signals defined implicitly as the solution to partial differential equations. We propose to leverage periodic activation functions for implicit neural representations and demonstrate that these networks, dubbed sinusoidal representation networks or Sirens, are ideally suited for representing complex natural signals and their derivatives. We analyze Siren activation statistics to propose a principled initialization scheme and demonstrate the representation of images, wavefields, video, sound, and their derivatives. Further, we show how Sirens can be leveraged to solve challenging boundary value problems, such as particular Eikonal equations (yielding signed distance functions), the Poisson equation, and the Helmholtz and wave equations. Lastly, we combine Sirens with hypernetworks to learn priors over the space of Siren functions."
                },
                {
                    "title": "SAL++: Sign Agnostic Learning with Derivatives",
                    "abstract": "Learning 3D geometry directly from raw data, such as point clouds, triangle soups, or un-oriented meshes is still a challenging task that feeds many downstream computer vision and graphics applications. In this paper, we introduce SAL++: a method for learning implicit neural representations of shapes directly from such raw data. We build upon the recent sign agnostic learning (SAL) approach and generalize it to include derivative data in a sign agnostic manner. In more detail, given the unsigned distance function to the input raw data, we suggest a novel sign agnostic regression loss, incorporating both pointwise values and gradients of the unsigned distance function. Optimizing this loss leads to a signed implicit function solution, the zero level set of which is a high quality, valid manifold approximation to the input 3D data. We demonstrate the efficacy of SAL++ shape space learning from two challenging datasets: ShapeNet that contains inconsistent orientation and non-manifold meshes, and D-Faust that contains raw 3D scans (triangle soups). On both these datasets, we present state-of-the-art results."
                },
                {
                    "title": "Local Implicit Grid Representations for 3D Scenes",
                    "abstract": "Shape priors learned from data are commonly used to reconstruct 3D objects from partial or noisy data. Yet no such shape priors are available for indoor scenes, since typical 3D autoencoders cannot handle their scale, complexity, or diversity. In this paper, we introduce Local Implicit Grid Representations, a new 3D shape representation designed for scalability and generality. The motivating idea is that most 3D surfaces share geometric details at some scale -- i.e., at a scale smaller than an entire object and larger than a small patch. We train an autoencoder to learn an embedding of local crops of 3D shapes at that size. Then, we use the decoder as a component in a shape optimization that solves for a set of latent codes on a regular grid of overlapping crops such that an interpolation of the decoded local shapes matches a partial or noisy observation. We demonstrate the value of this proposed approach for 3D surface reconstruction from sparse point observations, showing significantly better results than alternative approaches."
                },
                {
                    "title": "Implicit Geometric Regularization for Learning Shapes",
                    "abstract": "Representing shapes as level sets of neural networks has been recently proved to be useful for different shape analysis and reconstruction tasks. So far, such representations were computed using either: (i) pre-computed implicit shape representations; or (ii) loss functions explicitly defined over the neural level sets. In this paper we offer a new paradigm for computing high fidelity implicit neural representations directly from raw data (i.e., point clouds, with or without normal information). We observe that a rather simple loss function, encouraging the neural network to vanish on the input point cloud and to have a unit norm gradient, possesses an implicit geometric regularization property that favors smooth and natural zero level set surfaces, avoiding bad zero-loss solutions. We provide a theoretical analysis of this property for the linear case, and show that, in practice, our method leads to state of the art implicit neural representations with higher level-of-details and fidelity compared to previous methods."
                },
                {
                    "title": "SAL: Sign Agnostic Learning of Shapes From Raw Data",
                    "abstract": "Recently, neural networks have been used as implicit representations for surface reconstruction, modelling, learning, and generation. So far, training neural networks to be implicit representations of surfaces required training data sampled from a ground-truth signed implicit functions such as signed distance or occupancy functions, which are notoriously hard to compute. In this paper we introduce Sign Agnostic Learning (SAL), a deep learning approach for learning implicit shape representations directly from raw, unsigned geometric data, such as point clouds and triangle soups. We have tested SAL on the challenging problem of surface reconstruction from an un-oriented point cloud, as well as end-to-end human shape space learning directly from raw scans dataset, and achieved state of the art reconstructions compared to current approaches. We believe SAL opens the door to many geometric deep learning applications with real-world data, alleviating the usual painstaking, often manual pre-process."
                },
                {
                    "title": "SDM-NET: Deep Generative Network for Structured Deformable Mesh",
                    "abstract": "We introduce SDM-NET, a deep generative neural network which produces structured deformable meshes. Specifically, the network is trained to generate a spatial arrangement of closed, deformable mesh parts, which respect the global part structure of a shape collection, e.g., chairs, airplanes, etc. Our key observation is that while the overall structure of a 3D shape can be complex, the shape can usually be decomposed into a set of parts, each homeomorphic to a box, and the finer-scale geometry of the part can be recovered by deforming the box. The architecture of SDM-NET is that of a two-level variational autoencoder (VAE). At the part level, a PartVAE learns a deformable model of part geometries. At the structural level, we train a Structured Parts VAE (SP-VAE), which jointly learns the part structure of a shape collection and the part geometries, ensuring a coherence between global shape structure and surface details. Through extensive experiments and comparisons with the state-of-the-art deep generative models of shapes, we demonstrate the superiority of SDM-NET in generating meshes with visual quality, flexible topology, and meaningful structures, which benefit shape interpolation and other subsequently modeling tasks."
                },
                {
                    "title": "DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation",
                    "abstract": "Computer graphics, 3D computer vision and robotics communities have produced multiple approaches to representing 3D geometry for rendering and reconstruction. These provide trade-offs across fidelity, efficiency and compression capabilities. In this work, we introduce DeepSDF, a learned continuous Signed Distance Function (SDF) representation of a class of shapes that enables high quality shape representation, interpolation and completion from partial and noisy 3D input data. DeepSDF, like its classical counterpart, represents a shape's surface by a continuous volumetric field: the magnitude of a point in the field represents the distance to the surface boundary and the sign indicates whether the region is inside (-) or outside (+) of the shape, hence our representation implicitly encodes a shape's boundary as the zero-level-set of the learned function while explicitly representing the classification of space as being part of the shapes interior or not. While classical SDF's both in analytical or discretized voxel form typically represent the surface of a single shape, DeepSDF can represent an entire class of shapes. Furthermore, we show state-of-the-art performance for learned 3D shape representation and completion while reducing the model size by an order of magnitude compared with previous work."
                },
                {
                    "title": "Learning Implicit Fields for Generative Shape Modeling",
                    "abstract": "We advocate the use of implicit fields for learning generative models of shapes and introduce an implicit field decoder, called IM-NET, for shape generation, aimed at improving the visual quality of the generated shapes. An implicit field assigns a value to each point in 3D space, so that a shape can be extracted as an iso-surface. IM-NET is trained to perform this assignment by means of a binary classifier. Specifically, it takes a point coordinate, along with a feature vector encoding a shape, and outputs a value which indicates whether the point is outside the shape or not. By replacing conventional decoders by our implicit decoder for representation learning (via IM-AE) and shape generation (via IM-GAN), we demonstrate superior results for tasks such as generative shape modeling, interpolation, and single-view 3D reconstruction, particularly in terms of visual quality. Code and supplementary material are available at https://github.com/czq142857/implicit-decoder."
                },
                {
                    "title": "Practical 3D frame field generation",
                    "abstract": "Given a tetrahedral mesh, the algorithm described in this article produces a smooth 3D frame field, i.e. a set of three orthogonal directions associated with each vertex of the input mesh. The field varies smoothly inside the volume, and matches the normals of the volume boundary. Such a 3D frame field is a key component for some hexahedral meshing algorithms, where it is used to steer the placement of the generated elements. We improve the state-of-the art in terms of quality, efficiency and reproducibility. Our main contribution is a non-trivial extension in 3D of the existing least-squares approach used for optimizing a 2D frame field. Our algorithm is inspired by the method proposed by Huang et al. [2011], improved with an initialization that directly enforces boundary conditions. Our initialization alone is a fast and easy way to generate frames fields that are suitable for remeshing applications. For better robustness and quality, the field can be further optimized using nonlinear optimization as in Li et al [2012]. We make the remark that sampling the field on vertices instead of tetrahedra significantly improves both performance and quality."
                },
                {
                    "title": "Directional Field Synthesis, Design, and Processing",
                    "abstract": "Direction fields and vector fields play an increasingly important role in computer graphics and geometry processing. The synthesis of directional fields on surfaces, or other spatial domains, is a fundamental step in numerous applications, such as mesh generation, deformation, texture mapping, and many more. The wide range of applications resulted in definitions for many types of directional fields: from vector and tensor fields, over line and cross fields, to frame and vector\u2010set fields. Depending on the application at hand, researchers have used various notions of objectives and constraints to synthesize such fields. These notions are defined in terms of fairness, feature alignment, symmetry, or field topology, to mention just a few. To facilitate these objectives, various representations, discretizations, and optimization strategies have been developed. These choices come with varying strengths and weaknesses. This report provides a systematic overview of directional field synthesis for graphics applications, the challenges it poses, and the methods developed in recent years to address these challenges."
                },
                {
                    "title": "Instant field-aligned meshes",
                    "abstract": "We present a novel approach to remesh a surface into an isotropic triangular or quad-dominant mesh using a unified local smoothing operator that optimizes both the edge orientations and vertex positions in the output mesh. Our algorithm produces meshes with high isotropy while naturally aligning and snapping edges to sharp features. The method is simple to implement and parallelize, and it can process a variety of input surface representations, such as point clouds, range scans and triangle meshes. Our full pipeline executes instantly (less than a second) on meshes with hundreds of thousands of faces, enabling new types of interactive workflows. Since our algorithm avoids any global optimization, and its key steps scale linearly with input size, we are able to process extremely large meshes and point clouds, with sizes exceeding several hundred million elements. To demonstrate the robustness and effectiveness of our method, we apply it to hundreds of models of varying complexity and provide our cross-platform reference implementation in the supplemental material."
                },
                {
                    "title": "From Shading to Local Shape",
                    "abstract": "We develop a framework for extracting a concise representation of the shape information available from diffuse shading in a small image patch. This produces a mid-level scene descriptor, comprised of local shape distributions that are inferred separately at every image patch across multiple scales. The framework is based on a quadratic representation of local shape that, in the absence of noise, has guarantees on recovering accurate local shape and lighting. And when noise is present, the inferred local shape distributions provide useful shape information without over-committing to any particular image explanation. These local shape distributions naturally encode the fact that some smooth diffuse regions are more informative than others, and they enable efficient and robust reconstruction of object-scale shape. Experimental results show that this approach to surface reconstruction compares well against the state-of-art on both synthetic images and captured photographs."
                },
                {
                    "title": "Globally optimal direction fields",
                    "abstract": "We present a method for constructing smooth n-direction fields (line fields, cross fields, etc.) on surfaces that is an order of magnitude faster than state-of-the-art methods, while still producing fields of equal or better quality. Fields produced by the method are globally optimal in the sense that they minimize a simple, well-defined quadratic smoothness energy over all possible configurations of singularities (number, location, and index). The method is fully automatic and can optionally produce fields aligned with a given guidance field such as principal curvature directions. Computationally the smoothest field is found via a sparse eigenvalue problem involving a matrix similar to the cotan-Laplacian. When a guidance field is present, finding the optimal field amounts to solving a single linear system."
                },
                {
                    "title": "Periodic global parameterization",
                    "abstract": "We present a new globally smooth parameterization method for the triangulated surfaces of arbitrary topology. Given two orthogonal piecewise linear vector fields defined over the input mesh (typically the estimated principal curvature directions), our method computes two piecewise linear periodic functions, aligned with the input vector fields, by minimizing an objective function. The bivariate function they define is a smooth parameterization almost everywhere on the surface, except in the vicinity of singular vertices, edges, and triangles, where the derivatives of the parameterization vanish. We extract a quadrilateral chart layout from the parameterization function and propose an automatic procedure to detect the singularities, and fix them by splitting and reparameterizing the containing charts. Our method can construct both quasiconformal (angle preserving) and quasi-isometric (angle and area preserving) parameterizations. The more restrictive class of quasi-isometric parameterizations is constructed at the expense of introducing more singularities. The constructed parameterizations can be used for a variety of geometry processing applications. Since we can align the parameterization with the principal curvature directions, our result is particularly suitable for surface fitting and remeshing."
                },
                {
                    "title": "Provably good moving least squares",
                    "abstract": "We analyze a moving least squares algorithm for reconstructing a surface from point cloud data. Our algorithm defines an implicit function I whose zero set U is the reconstructed surface. We prove that I is a good approximation to the signed distance function of the sampled surface F and that U is geometrically close to and homeomorphic to F. Our proof requires sampling conditions similar to \u03b5-sampling, used in Delaunay reconstruction algorithms."
                },
                {
                    "title": "Multi-level partition of unity implicits",
                    "abstract": "We present a new shape representation, the multi-level partition of unity implicit surface, that allows us to construct surface models from very large sets of points. There are three key ingredients to our approach: 1) piecewise quadratic functions that capture the local shape of the surface, 2) weighting functions (the partitions of unity) that blend together these local shape functions, and 3) an octree subdivision method that adapts to variations in the complexity of the local shape.Our approach gives us considerable flexibility in the choice of local shape functions, and in particular we can accurately represent sharp features such as edges and corners by selecting appropriate shape functions. An error-controlled subdivision leads to an adaptive approximation whose time and memory consumption depends on the required accuracy. Due to the separation of local approximation and local blending, the representation is not global and can be created and evaluated rapidly. Because our surfaces are described using implicit functions, operations such as shape blending, offsets, deformations and CSG are simple to perform."
                },
                {
                    "title": "Three-Dimensional Shape Analysis Using Local Shape Descriptors",
                    "abstract": "The three-dimensional shape analysis problem is a very demanding test of shape analysis algorithms. Previous approaches to the problem have employed global features such as moments and Fourier descriptors. Global features lack the capacity for solving the partial shape recognition problem, in which only part of the unknown shape is available. Previous approaches to local shape analysis have employed structural (syntactic) methods, but these methods have so far failed to solve the three-dimensional problem. This paper describes a hybrid structural/statistical local shape analysis algorithm which is applied to the three-dimensional problem."
                },
                {
                    "title": "Mesh R-CNN",
                    "abstract": "Rapid advances in 2D perception have led to systems that accurately detect objects in real-world images. However, these systems make predictions in 2D, ignoring the 3D structure of the world. Concurrently, advances in 3D shape prediction have mostly focused on synthetic benchmarks and isolated objects. We unify advances in these two areas. We propose a system that detects objects in real-world images and produces a triangle mesh giving the full 3D shape of each detected object. Our system, called Mesh R-CNN, augments Mask R-CNN with a mesh prediction branch that outputs meshes with varying topological structure by \ufb01rst predicting coarse voxel representations which are converted to meshes and re\ufb01ned with a graph convolution network operating over the mesh\u2019s vertices and edges. We validate our mesh prediction branch on ShapeNet, where we out-perform prior work on single-image shape prediction. We then deploy our full Mesh R-CNN system on Pix3D, where we jointly detect objects and predict their 3D shapes."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.GR"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a compact and accurate neural surface representation that effectively captures the geometry of local surfaces while overcoming the challenges of transformation alignment in 3D space?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of geometry modeling, particularly in applications such as computer graphics, robotics, and virtual reality, where accurate shape representation is essential. A successful approach could lead to more efficient algorithms for shape reconstruction and manipulation, enabling researchers to create more detailed and realistic models. Furthermore, it could inspire future research into other forms of shape representation and contribute to the development of more sophisticated machine learning techniques that leverage local geometry.\n\n**[Question 3] - Why is it hard?**  \nThe complexity of this problem arises from the need to jointly recover transformation information and the geometry of local patches, which can easily lead to optimization challenges such as getting trapped in local minima. Naive approaches that treat local surfaces uniformly may fail to capture the intricacies of their geometry, especially when the patches are freely transformed in 3D space. Additionally, the increased number of parameters associated with local surface representations complicates the learning process, making it difficult to achieve both accuracy and compactness.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on using single latent codes for entire shapes or local-based designs that significantly increase parameter counts without addressing the alignment of local surfaces. The lack of a method to effectively separate transformation information from geometry has been a barrier to progress. Existing solutions often rely on implicit representations that do not adequately capture the complexities of local geometry. Our approach, CoFie, differs by introducing a learnable Coordinate Field that explicitly represents the transformation of local surfaces, allowing for better alignment and compact representation.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, CoFie, involves decomposing shapes into non-overlapping local patches, each represented by an MLP-based Signed Distance Function (SDF). We introduce a Coordinate Field that provides a learnable coordinate frame for each local surface, transforming them into an aligned coordinate system to reduce spatial complexity. The representation of the Coordinate Field is parameterized by a 6 Degree-of-Freedom pose, initialized using geometric properties of the local surface. We will evaluate our approach using standard shape datasets and metrics such as reconstruction accuracy and parameter efficiency. Expected outcomes include improved shape"
            }
        },
        "author_data": {
            "d337404c-4e96-456a-abe2-2b014d956895": {
                "pk": "d337404c-4e96-456a-abe2-2b014d956895",
                "name": "Hanwen Jiang",
                "collaborators": [
                    "Zhenyu Jiang",
                    "Qixing Huang",
                    "Yuke Zhu",
                    "Kristen Grauman",
                    "Shaowei Liu",
                    "Xiaolong Wang",
                    "Yue Zhao",
                    "Santhosh Kumar Ramakrishnan",
                    "Georgios Pavlakos",
                    "Jiashun Wang"
                ],
                "domain": [
                    "Computer Vision",
                    "3D Reconstruction",
                    "Self-Supervised Learning",
                    "Robotics"
                ],
                "publications": [
                    {
                        "title": "Doduo: Learning Dense Visual Correspondence from Unsupervised Semantic-Aware Flow",
                        "abstract": "Dense visual correspondence plays a vital role in robotic perception. This work focuses on establishing the dense correspondence between a pair of images that captures dynamic scenes undergoing substantial transformations. We introduce Doduo to learn general dense visual correspondence from in-the-wild images and videos without ground truth supervision. Given a pair of images, it estimates the dense flow field encoding the displacement of each pixel in one image to its corresponding pixel in the other image. Doduo uses flow-based warping to acquire supervisory signals for the training. Incorporating semantic priors with self-supervised flow training, Doduo produces accurate dense correspondence robust to the dynamic changes of the scenes. Trained on an in-the-wild video dataset, Doduo illustrates superior performance on point-level correspondence estimation over existing self-supervised correspondence learning baselines. We also apply Doduo to articulation estimation and zero-shot goal-conditioned manipulation, underlining its practical applications in robotics. Code and additional visualizations are available at https://ut-austin-rpl.github.io/Doduo"
                    },
                    {
                        "title": "LEAP: Liberate Sparse-view 3D Modeling from Camera Poses",
                        "abstract": "Are camera poses necessary for multi-view 3D modeling? Existing approaches predominantly assume access to accurate camera poses. While this assumption might hold for dense views, accurately estimating camera poses for sparse views is often elusive. Our analysis reveals that noisy estimated poses lead to degraded performance for existing sparse-view 3D modeling methods. To address this issue, we present LEAP, a novel pose-free approach, therefore challenging the prevailing notion that camera poses are indispensable. LEAP discards pose-based operations and learns geometric knowledge from data. LEAP is equipped with a neural volume, which is shared across scenes and is parameterized to encode geometry and texture priors. For each incoming scene, we update the neural volume by aggregating 2D image features in a feature-similarity-driven manner. The updated neural volume is decoded into the radiance field, enabling novel view synthesis from any viewpoint. On both object-centric and scene-level datasets, we show that LEAP significantly outperforms prior methods when they employ predicted poses from state-of-the-art pose estimators. Notably, LEAP performs on par with prior approaches that use ground-truth poses while running $400\\times$ faster than PixelNeRF. We show LEAP generalizes to novel object categories and scenes, and learns knowledge closely resembles epipolar geometry. Project page: https://hwjiang1510.github.io/LEAP/"
                    },
                    {
                        "title": "Single-Stage Visual Query Localization in Egocentric Videos",
                        "abstract": "Visual Query Localization on long-form egocentric videos requires spatio-temporal search and localization of visually specified objects and is vital to build episodic memory systems. Prior work develops complex multi-stage pipelines that leverage well-established object detection and tracking methods to perform VQL. However, each stage is independently trained and the complexity of the pipeline results in slow inference speeds. We propose VQLoC, a novel single-stage VQL framework that is end-to-end trainable. Our key idea is to first build a holistic understanding of the query-video relationship and then perform spatio-temporal localization in a single shot manner. Specifically, we establish the query-video relationship by jointly considering query-to-frame correspondences between the query and each video frame and frame-to-frame correspondences between nearby video frames. Our experiments demonstrate that our approach outperforms prior VQL methods by 20% accuracy while obtaining a 10x improvement in inference speed. VQLoC is also the top entry on the Ego4D VQ2D challenge leaderboard. Project page: https://hwjiang1510.github.io/VQLoC/"
                    },
                    {
                        "title": "Real3D: Scaling Up Large Reconstruction Models with Real-World Images",
                        "abstract": "The default strategy for training single-view Large Reconstruction Models (LRMs) follows the fully supervised route using large-scale datasets of synthetic 3D assets or multi-view captures. Although these resources simplify the training procedure, they are hard to scale up beyond the existing datasets and they are not necessarily representative of the real distribution of object shapes. To address these limitations, in this paper, we introduce Real3D, the first LRM system that can be trained using single-view real-world images. Real3D introduces a novel self-training framework that can benefit from both the existing synthetic data and diverse single-view real images. We propose two unsupervised losses that allow us to supervise LRMs at the pixel- and semantic-level, even for training examples without ground-truth 3D or novel views. To further improve performance and scale up the image data, we develop an automatic data curation approach to collect high-quality examples from in-the-wild images. Our experiments show that Real3D consistently outperforms prior work in four diverse evaluation settings that include real and synthetic data, as well as both in-domain and out-of-domain shapes. Code and model can be found here: https://hwjiang1510.github.io/Real3D/"
                    },
                    {
                        "title": "Few-View Object Reconstruction with Unknown Categories and Camera Poses",
                        "abstract": "While object reconstruction has made great strides in recent years, current methods typically require densely captured images and/or known camera poses, and generalize poorly to novel object categories. To step toward object reconstruction in the wild, this work explores reconstructing general real-world objects from a few images without known camera poses or object categories. The crux of our work is solving two fundamental 3D vision problems -- shape reconstruction and pose estimation -- in a unified approach. Our approach captures the synergies of these two problems: reliable camera pose estimation gives rise to accurate shape reconstruction, and the accurate reconstruction, in turn, induces robust correspondence between different views and facilitates pose estimation. Our method FORGE predicts 3D features from each view and leverages them in conjunction with the input images to establish cross-view correspondence for estimating relative camera poses. The 3D features are then transformed by the estimated poses into a shared space and are fused into a neural radiance field. The reconstruction results are rendered by volume rendering techniques, enabling us to train the model without 3D shape ground-truth. Our experiments show that FORGE reliably reconstructs objects from five views. Our pose estimation method outperforms existing ones by a large margin. The reconstruction results under predicted poses are comparable to the ones using ground-truth poses. The performance on novel testing categories matches the results on categories seen during training. Project page: https://ut-austin-rpl.github.io/FORGE/"
                    },
                    {
                        "title": "Hand-Object Contact Consistency Reasoning for Human Grasps Generation",
                        "abstract": "While predicting robot grasps with parallel jaw grippers have been well studied and widely applied in robot manipulation tasks, the study on natural human grasp generation with a multi-finger hand remains a very challenging problem. In this paper, we propose to generate human grasps given a 3D object in the world. Our key observation is that it is crucial to model the consistency between the hand contact points and object contact regions. That is, we encourage the prior hand contact points to be close to the object surface and the object common contact regions to be touched by the hand at the same time. Based on the hand-object contact consistency, we design novel objectives in training the human grasp generation model and also a new self-supervised task which allows the grasp generation network to be adjusted even during test time. Our experiments show significant improvement in human grasp generation over state-of-the-art approaches by a large margin. More interestingly, by optimizing the model during test time with the self-supervised task, it helps achieve larger gain on unseen and out-of-domain objects. Project page: https://hwjiang1510.github.io/GraspTTA/"
                    },
                    {
                        "title": "OmniGlue: Generalizable Feature Matching with Foundation Model Guidance",
                        "abstract": "The image matching field has been witnessing a continuous emergence of novel learnable feature matching techniques, with ever-improving performance on conventional benchmarks. However, our investigation shows that despite these gains, their potential for real-world applications is restricted by their limited generalization capabilities to novel image domains. In this paper, we introduce OmniGlue, the first learnable image matcher that is designed with generalization as a core principle. OmniGlue leverages broad knowledge from a vision foundation model to guide the feature matching process, boosting generalization to domains not seen at training time. Additionally, we propose a novel keypoint position-guided attention mechanism which disentangles spatial and appearance information, leading to enhanced matching descriptors. We perform comprehensive experiments on a suite of $7$ datasets with varied image domains, including scene-level, object-centric and aerial images. OmniGlue's novel components lead to relative gains on unseen domains of $20.9\\%$ with respect to a directly comparable reference model, while also outperforming the recent LightGlue method by $9.5\\%$ relatively.Code and model can be found at https://hwjiang1510.github.io/OmniGlue"
                    },
                    {
                        "title": "Semi-Supervised 3D Hand-Object Poses Estimation with Interactions in Time",
                        "abstract": "Estimating 3D hand and object pose from a single image is an extremely challenging problem: hands and objects are often self-occluded during interactions, and the 3D annotations are scarce as even humans cannot directly label the ground-truths from a single image perfectly. To tackle these challenges, we propose a unified framework for estimating the 3D hand and object poses with semi-supervised learning. We build a joint learning framework where we perform explicit contextual reasoning between hand and object representations by a Transformer. Going beyond limited 3D annotations in a single image, we leverage the spatial-temporal consistency in large-scale hand-object videos as a constraint for generating pseudo labels in semi-supervised learning. Our method not only improves hand pose estimation in challenging real-world dataset, but also substantially improve the object pose which has fewer ground-truths per instance. By training with large-scale diverse videos, our model also generalizes better across multiple out-of-domain datasets. Project page and code: https://stevenlsw.github.io/Semi-Hand-Object"
                    }
                ]
            },
            "20df65cf-245b-4e0f-a62e-f9b342d9bafc": {
                "pk": "20df65cf-245b-4e0f-a62e-f9b342d9bafc",
                "name": "Haitao Yang",
                "collaborators": [
                    "Yajing Xu",
                    "Mingfei Cheng",
                    "Si Li",
                    "Zaiwei Zhang",
                    "Bo Sun",
                    "Qixing Huang"
                ],
                "domain": [
                    "Deep Learning",
                    "Computer Vision",
                    "3D Object Detection",
                    "Generative Adversarial Networks"
                ],
                "publications": [
                    {
                        "title": "Cyclone intensity estimate with context-aware cyclegan",
                        "abstract": "Deep learning approaches to cyclone intensity estimationhave recently shown promising results. However, sufferingfrom the extreme scarcity of cyclone data on specific in-tensity, most existing deep learning methods fail to achievesatisfactory performance on cyclone intensity estimation,especially on classes with few instances. To avoid the degra-dation of recognition performance caused by scarce samples,we propose a context-aware CycleGAN which learns the la-tent evolution features from adjacent cyclone intensity andsynthesizes CNN features of classes lacking samples fromunpaired source classes. Specifically, our approach synthe-sizes features conditioned on the learned evolution features,while the extra information is not required. Experimentalresults of several evaluation methods show the effectivenessof our approach, even can predicting unseen classes."
                    },
                    {
                        "title": "H3DNet: 3D Object Detection Using Hybrid Geometric Primitives",
                        "abstract": "We introduce H3DNet, which takes a colorless 3D point cloud as input and outputs a collection of oriented object bounding boxes (or BB) and their semantic labels. The critical idea of H3DNet is to predict a hybrid set of geometric primitives, i.e., BB centers, BB face centers, and BB edge centers. We show how to convert the predicted geometric primitives into object proposals by defining a distance function between an object and the geometric primitives. This distance function enables continuous optimization of object proposals, and its local minimums provide high-fidelity object proposals. H3DNet then utilizes a matching and refinement module to classify object proposals into detected objects and fine-tune the geometric parameters of the detected objects. The hybrid set of geometric primitives not only provides more accurate signals for object detection than using a single type of geometric primitives, but it also provides an overcomplete set of constraints on the resulting 3D layout. Therefore, H3DNet can tolerate outliers in predicted geometric primitives. Our model achieves state-of-the-art 3D detection results on two large datasets with real 3D scans, ScanNet and SUN RGB-D."
                    }
                ]
            },
            "1d81930b-aa6d-470f-9939-1b38c602885f": {
                "pk": "1d81930b-aa6d-470f-9939-1b38c602885f",
                "name": "Georgios Pavlakos",
                "collaborators": [
                    "Kostas Daniilidis",
                    "Angjoo Kanazawa",
                    "Jitendra Malik",
                    "Xiaowei Zhou",
                    "Nikos Kolotouros",
                    "Konstantinos G. Derpanis",
                    "Jathushan Rajasegaran",
                    "Ethan Weber",
                    "Matthew Tancik",
                    "Luyang Zhu"
                ],
                "domain": [
                    "3D Human Pose Estimation",
                    "Computer Vision",
                    "Deep Learning",
                    "Human-Computer Interaction"
                ],
                "publications": [
                    {
                        "title": "TexturePose: Supervising Human Mesh Estimation with Texture Consistency",
                        "abstract": "This work addresses the problem of model-based human pose estimation. Recent approaches have made significant progress towards regressing the parameters of parametric human body models directly from images. Because of the absence of images with 3D shape ground truth, relevant approaches rely on 2D annotations or sophisticated architecture designs. In this work, we advocate that there are more cues we can leverage, which are available for free in natural images, i.e., without getting more annotations, or modifying the network architecture. We propose a natural form of supervision, that capitalizes on the appearance constancy of a person among different frames (or viewpoints). This seemingly insignificant and often overlooked cue goes a long way for model-based pose estimation. The parametric model we employ allows us to compute a texture map for each frame. Assuming that the texture of the person does not change dramatically between frames, we can apply a novel texture consistency loss, which enforces that each point in the texture map has the same texture value across all frames. Since the texture is transferred in this common texture map space, no camera motion computation is necessary, or even an assumption of smoothness among frames. This makes our proposed supervision applicable in a variety of settings, ranging from monocular video, to multi-view images. We benchmark our approach against strong baselines that require the same or even more annotations that we do and we consistently outperform them. Simultaneously, we achieve state-of-the-art results among model-based pose estimation approaches in different benchmarks. The project website with videos, results, and code can be found at https://seas.upenn.edu/~pavlakos/projects/texturepose."
                    },
                    {
                        "title": "Ordinal Depth Supervision for 3D Human Pose Estimation",
                        "abstract": "Our ability to train end-to-end systems for 3D human pose estimation from single images is currently constrained by the limited availability of 3D annotations for natural images. Most datasets are captured using Motion Capture (MoCap) systems in a studio setting and it is difficult to reach the variability of 2D human pose datasets, like MPII or LSP. To alleviate the need for accurate 3D ground truth, we propose to use a weaker supervision signal provided by the ordinal depths of human joints. This information can be acquired by human annotators for a wide range of images and poses. We showcase the effectiveness and flexibility of training Convolutional Networks (ConvNets) with these ordinal relations in different settings, always achieving competitive performance with ConvNets trained with accurate 3D joint coordinates. Additionally, to demonstrate the potential of the approach, we augment the popular LSP and MPII datasets with ordinal depth annotations. This extension allows us to present quantitative and qualitative evaluation in non-studio conditions. Simultaneously, these ordinal annotations can be easily incorporated in the training procedure of typical ConvNets for 3D human pose. Through this inclusion we achieve new state-of-the-art performance for the relevant benchmarks and validate the effectiveness of ordinal depth supervision for 3D human pose."
                    },
                    {
                        "title": "The One Where They Reconstructed 3D Humans and Environments in TV Shows",
                        "abstract": "TV shows depict a wide variety of human behaviors and have been studied extensively for their potential to be a rich source of data for many applications. However, the majority of the existing work focuses on 2D recognition tasks. In this paper, we make the observation that there is a certain persistence in TV shows, i.e., repetition of the environments and the humans, which makes possible the 3D reconstruction of this content. Building on this insight, we propose an automatic approach that operates on an entire season of a TV show and aggregates information in 3D; we build a 3D model of the environment, compute camera information, static 3D scene structure and body scale information. Then, we demonstrate how this information acts as rich 3D context that can guide and improve the recovery of 3D human pose and position in these environments. Moreover, we show that reasoning about humans and their environment in 3D enables a broad range of downstream applications: re-identification, gaze estimation, cinematography and image editing. We apply our approach on environments from seven iconic TV shows and perform an extensive evaluation of the proposed system."
                    },
                    {
                        "title": "Coarse-to-Fine Volumetric Prediction for Single-Image 3D Human Pose",
                        "abstract": "This paper addresses the challenge of 3D human pose estimation from a single color image. Despite the general success of the end-to-end learning paradigm, top performing approaches employ a two-step solution consisting of a Convolutional Network (ConvNet) for 2D joint localization and a subsequent optimization step to recover 3D pose. In this paper, we identify the representation of 3D pose as a critical issue with current ConvNet approaches and make two important contributions towards validating the value of end-to-end learning for this task. First, we propose a fine discretization of the 3D space around the subject and train a ConvNet to predict per voxel likelihoods for each joint. This creates a natural representation for 3D pose and greatly improves performance over the direct regression of joint coordinates. Second, to further improve upon initial estimates, we employ a coarse-to-fine prediction scheme. This step addresses the large dimensionality increase and enables iterative refinement and repeated processing of the image features. The proposed approach outperforms all state-of-the-art methods on standard benchmarks achieving a relative error reduction greater than 30% on average. Additionally, we investigate using our volumetric representation in a related architecture which is suboptimal compared to our end-to-end approach, but is of practical interest, since it enables training when no image with corresponding 3D groundtruth is available, and allows us to present compelling results for in-the-wild images."
                    },
                    {
                        "title": "Learning to Estimate 3D Human Pose and Shape from a Single Color Image",
                        "abstract": "This work addresses the problem of estimating the full body 3D human pose and shape from a single color image. This is a task where iterative optimization-based solutions have typically prevailed, while Convolutional Networks (ConvNets) have suffered because of the lack of training data and their low resolution 3D predictions. Our work aims to bridge this gap and proposes an efficient and effective direct prediction method based on ConvNets. Central part to our approach is the incorporation of a parametric statistical body shape model (SMPL) within our end-to-end framework. This allows us to get very detailed 3D mesh results, while requiring estimation only of a small number of parameters, making it friendly for direct network prediction. Interestingly, we demonstrate that these parameters can be predicted reliably only from 2D keypoints and masks. These are typical outputs of generic 2D human analysis ConvNets, allowing us to relax the massive requirement that images with 3D shape ground truth are available for training. Simultaneously, by maintaining differentiability, at training time we generate the 3D mesh from the estimated parameters and optimize explicitly for the surface using a 3D per-vertex loss. Finally, a differentiable renderer is employed to project the 3D mesh to the image, which enables further refinement of the network, by optimizing for the consistency of the projection with 2D annotations (i.e., 2D keypoints or masks). The proposed approach outperforms previous baselines on this task and offers an attractive solution for direct prediction of 3D shape from a single color image."
                    },
                    {
                        "title": "Human Mesh Recovery from Multiple Shots",
                        "abstract": "Videos from edited media like movies are a useful, yet under-explored source of information. The rich variety of appearance and interactions between humans depicted over a large temporal context in these films could be a valuable source of data. However, the richness of data comes at the expense of fundamental challenges such as abrupt shot changes and close up shots of actors with heavy truncation, which limits the applicability of existing human 3D understanding methods. In this paper, we address these limitations with an insight that while shot changes of the same scene incur a discontinuity between frames, the 3D structure of the scene still changes smoothly. This allows us to handle frames before and after the shot change as multi-view signal that provide strong cues to recover the 3D state of the actors. We propose a multi-shot optimization framework, which leads to improved 3D reconstruction and mining of long sequences with pseudo ground truth 3D human mesh. We show that the resulting data is beneficial in the training of various human mesh recovery models: for single image, we achieve improved robustness; for video we propose a pure transformer-based temporal encoder, which can naturally handle missing observations due to shot changes in the input frames. We demonstrate the importance of the insight and proposed models through extensive experiments. The tools we develop open the door to processing and analyzing in 3D content from a large library of edited media, which could be helpful for many downstream applications. Project page: https://geopavlakos.github.io/multishot"
                    },
                    {
                        "title": "Probabilistic Modeling for Human Mesh Recovery",
                        "abstract": "This paper focuses on the problem of 3D human reconstruction from 2D evidence. Although this is an inherently ambiguous problem, the majority of recent works avoid the uncertainty modeling and typically regress a single estimate for a given input. In contrast to that, in this work, we propose to embrace the reconstruction ambiguity and we recast the problem as learning a mapping from the input to a distribution of plausible 3D poses. Our approach is based on the normalizing flows model and offers a series of advantages. For conventional applications, where a single 3D estimate is required, our formulation allows for efficient mode computation. Using the mode leads to performance that is comparable with the state of the art among deterministic unimodal regression models. Simultaneously, since we have access to the likelihood of each sample, we demonstrate that our model is useful in a series of downstream tasks, where we leverage the probabilistic nature of the prediction as a tool for more accurate estimation. These tasks include reconstruction from multiple uncalibrated views, as well as human model fitting, where our model acts as a powerful image-based prior for mesh recovery. Our results validate the importance of probabilistic modeling, and indicate state-of-the-art performance across a variety of settings. Code and models are available at: https://www.seas.upenn.edu/~nkolot/projects/prohmr."
                    },
                    {
                        "title": "Harvesting Multiple Views for Marker-less 3D Human Pose Annotations",
                        "abstract": "Recent advances with Convolutional Networks (ConvNets) have shifted the bottleneck for many computer vision tasks to annotated data collection. In this paper, we present a geometry-driven approach to automatically collect annotations for human pose prediction tasks. Starting from a generic ConvNet for 2D human pose, and assuming a multi-view setup, we describe an automatic way to collect accurate 3D human pose annotations. We capitalize on constraints offered by the 3D geometry of the camera setup and the 3D structure of the human body to probabilistically combine per view 2D ConvNet predictions into a globally optimal 3D pose. This 3D pose is used as the basis for harvesting annotations. The benefit of the annotations produced automatically with our approach is demonstrated in two challenging settings: (i) fine-tuning a generic ConvNet-based 2D pose predictor to capture the discriminative aspects of a subject's appearance (i.e.,\"personalization\"), and (ii) training a ConvNet from scratch for single view 3D human pose prediction without leveraging 3D pose groundtruth. The proposed multi-view pose estimator achieves state-of-the-art results on standard benchmarks, demonstrating the effectiveness of our method in exploiting the available multi-view information."
                    },
                    {
                        "title": "6-DoF Object Pose from Semantic Keypoints",
                        "abstract": "This paper presents a novel approach to estimating the continuous six degree of freedom (6-DoF) pose (3D translation and rotation) of an object from a single RGB image. The approach combines semantic keypoints predicted by a convolutional network (convnet) with a deformable shape model. Unlike prior work, we are agnostic to whether the object is textured or textureless, as the convnet learns the optimal representation from the available training image data. Furthermore, the approach can be applied to instance- and class-based pose recovery. Empirically, we show that the proposed approach can accurately recover the 6-DoF object pose for both instance- and class-based scenarios with a cluttered background. For class-based object pose estimation, state-of-the-art accuracy is shown on the large-scale PASCAL3D+ dataset."
                    },
                    {
                        "title": "Reconstructing Hands in 3D with Transformers",
                        "abstract": "We present an approach that can reconstruct hands in 3D from monocular input. Our approach for Hand Mesh Recovery, HaMeR, follows a fully transformer-based architecture and can analyze hands with significantly increased accuracy and robustness compared to previous work. The key to HaMeR's success lies in scaling up both the data used for training and the capacity of the deep network for hand reconstruction. For training data, we combine multiple datasets that contain 2D or 3D hand annotations. For the deep model, we use a large scale Vision Transformer architecture. Our final model consistently outperforms the previous baselines on popular 3D hand pose benchmarks. To further evaluate the effect of our design in non-controlled settings, we annotate existing in-the-wild datasets with 2D hand keypoint annotations. On this newly collected dataset of annotations, HInt, we demonstrate significant improvements over existing baselines. We make our code, data and models available on the project website: https://geopavlakos.github.io/hamer/."
                    },
                    {
                        "title": "Independent Sign Language Recognition with 3D Body, Hands, and Face Reconstruction",
                        "abstract": "Independent Sign Language Recognition is a complex visual recognition problem that combines several challenging tasks of Computer Vision due to the necessity to exploit and fuse information from hand gestures, body features and facial expressions. While many state-of-the-art works have managed to deeply elaborate on these features independently, to the best of our knowledge, no work has adequately combined all three information channels to efficiently recognize Sign Language. In this work, we employ SMPL-X, a contemporary parametric model that enables joint extraction of 3D body shape, face and hands information from a single image. We use this holistic 3D reconstruction for SLR, demonstrating that it leads to higher accuracy than recognition from raw RGB images and their optical flow fed into the state-of-the-art I3D-type network for 3D action recognition and from 2D Openpose skeletons fed into a Recurrent Neural Network. Finally, a set of experiments on the body, face and hand features showed that neglecting any of these, significantly reduces the classification accuracy, proving the importance of jointly modeling body shape, facial expression and hand pose for Sign Language Recognition."
                    },
                    {
                        "title": "Tracking People with 3D Representations",
                        "abstract": "We present a novel approach for tracking multiple people in video. Unlike past approaches which employ 2D representations, we focus on using 3D representations of people, located in three-dimensional space. To this end, we develop a method, Human Mesh and Appearance Recovery (HMAR) which in addition to extracting the 3D geometry of the person as a SMPL mesh, also extracts appearance as a texture map on the triangles of the mesh. This serves as a 3D representation for appearance that is robust to viewpoint and pose changes. Given a video clip, we first detect bounding boxes corresponding to people, and for each one, we extract 3D appearance, pose, and location information using HMAR. These embedding vectors are then sent to a transformer, which performs spatio-temporal aggregation of the representations over the duration of the sequence. The similarity of the resulting representations is used to solve for associations that assigns each person to a tracklet. We evaluate our approach on the Posetrack, MuPoTs and AVA datasets. We find that 3D representations are more effective than 2D representations for tracking in these settings, and we obtain state-of-the-art performance. Code and results are available at: https://brjathu.github.io/T3DP."
                    },
                    {
                        "title": "Tracking People by Predicting 3D Appearance, Location & Pose",
                        "abstract": "In this paper, we present an approach for tracking people in monocular videos, by predicting their future 3D representations. To achieve this, we first lift people to 3D from a single frame in a robust way. This lifting includes information about the 3D pose of the person, his or her location in the 3D space, and the 3D appearance. As we track a person, we collect 3D observations over time in a tracklet representation. Given the 3D nature of our observations, we build temporal models for each one of the previous attributes. We use these models to predict the future state of the tracklet, including 3D location, 3D appearance, and 3D pose. For a future frame, we compute the similarity between the predicted state of a tracklet and the single frame observations in a probabilistic manner. Association is solved with simple Hungarian matching, and the matches are used to update the respective tracklets. We evaluate our approach on various benchmarks and report state-of-the-art results."
                    },
                    {
                        "title": "Decoupling Human and Camera Motion from Videos in the Wild",
                        "abstract": "We propose a method to reconstruct global human trajectories from videos in the wild. Our optimization method decouples the camera and human motion, which allows us to place people in the same world coordinate frame. Most existing methods do not model the camera motion; methods that rely on the background pixels to infer 3D human motion usually require a full scene reconstruction, which is often not possible for in-the-wild videos. However, even when existing SLAM systems cannot recover accurate scene reconstructions, the background pixel motion still provides enough signal to constrain the camera motion. We show that relative camera estimates along with data-driven human motion priors can resolve the scene scale ambiguity and recover global human trajectories. Our method robustly recovers the global 3D trajectories of people in challenging in-the-wild videos, such as PoseTrack. We quantify our improvement over existing methods on 3D human dataset Egobody. We further demonstrate that our recovered camera scale allows us to reason about motion of multiple people in a shared coordinate frame, which improves performance of downstream tracking in PoseTrack. Code and video results can be found at https://vye16.github.io/slahmr."
                    },
                    {
                        "title": "Evaluating Zero-Shot GPT-4V Performance on 3D Visual Question Answering Benchmarks",
                        "abstract": "As interest in \"reformulating\" the 3D Visual Question Answering (VQA) problem in the context of foundation models grows, it is imperative to assess how these new paradigms influence existing closed-vocabulary datasets. In this case study, we evaluate the zero-shot performance of foundational models (GPT-4 Vision and GPT-4) on well-established 3D VQA benchmarks, namely 3D-VQA and ScanQA. We provide an investigation to contextualize the performance of GPT-based agents relative to traditional modeling approaches. We find that GPT-based agents without any fine-tuning perform on par with the closed vocabulary approaches. Our findings corroborate recent results that \"blind\" models establish a surprisingly strong baseline in closed-vocabulary settings. We demonstrate that agents benefit significantly from scene-specific vocabulary via in-context textual grounding. By presenting a preliminary comparison with previous baselines, we hope to inform the community's ongoing efforts to refine multi-modal 3D benchmarks."
                    },
                    {
                        "title": "Real3D: Scaling Up Large Reconstruction Models with Real-World Images",
                        "abstract": "The default strategy for training single-view Large Reconstruction Models (LRMs) follows the fully supervised route using large-scale datasets of synthetic 3D assets or multi-view captures. Although these resources simplify the training procedure, they are hard to scale up beyond the existing datasets and they are not necessarily representative of the real distribution of object shapes. To address these limitations, in this paper, we introduce Real3D, the first LRM system that can be trained using single-view real-world images. Real3D introduces a novel self-training framework that can benefit from both the existing synthetic data and diverse single-view real images. We propose two unsupervised losses that allow us to supervise LRMs at the pixel- and semantic-level, even for training examples without ground-truth 3D or novel views. To further improve performance and scale up the image data, we develop an automatic data curation approach to collect high-quality examples from in-the-wild images. Our experiments show that Real3D consistently outperforms prior work in four diverse evaluation settings that include real and synthetic data, as well as both in-domain and out-of-domain shapes. Code and model can be found here: https://hwjiang1510.github.io/Real3D/"
                    },
                    {
                        "title": "Convolutional Mesh Regression for Single-Image Human Shape Reconstruction",
                        "abstract": "This paper addresses the problem of 3D human pose and shape estimation from a single image. Previous approaches consider a parametric model of the human body, SMPL, and attempt to regress the model parameters that give rise to a mesh consistent with image evidence. This parameter regression has been a very challenging task, with model-based approaches underperforming compared to nonparametric solutions in terms of pose estimation. In our work, we propose to relax this heavy reliance on the model's parameter space. We still retain the topology of the SMPL template mesh, but instead of predicting model parameters, we directly regress the 3D location of the mesh vertices. This is a heavy task for a typical network, but our key insight is that the regression becomes significantly easier using a Graph-CNN. This architecture allows us to explicitly encode the template mesh structure within the network and leverage the spatial locality the mesh has to offer. Image-based features are attached to the mesh vertices and the Graph-CNN is responsible to process them on the mesh structure, while the regression target for each vertex is its 3D location. Having recovered the complete 3D geometry of the mesh, if we still require a specific model parametrization, this can be reliably regressed from the vertices locations. We demonstrate the flexibility and the effectiveness of our proposed graph-based mesh regression by attaching different types of features on the mesh vertices. In all cases, we outperform the comparable baselines relying on model parameter regression, while we also achieve state-of-the-art results among model-based pose estimation approaches."
                    },
                    {
                        "title": "Learning Articulated Shape with Keypoint Pseudo-labels from Web Images",
                        "abstract": "This paper shows that it is possible to learn models for monocular 3D reconstruction of articulated objects (e.g., horses, cows, sheep), using as few as 50-150 images labeled with 2D keypoints. Our proposed approach involves training category-specific keypoint estimators, generating 2D keypoint pseudo-labels on unlabeled web images, and using both the labeled and self-labeled sets to train 3D reconstruction models. It is based on two key insights: (1) 2D keypoint estimation networks trained on as few as 50-150 images of a given object category generalize well and generate reliable pseudo-labels; (2) a data selection mechanism can automatically create a \"curated\" subset of the unlabeled web images that can be used for training -- we evaluate four data selection methods. Coupling these two insights enables us to train models that effectively utilize web images, resulting in improved 3D reconstruction performance for several articulated object categories beyond the fully-supervised baseline. Our approach can quickly bootstrap a model and requires only a few images labeled with 2D keypoints. This requirement can be easily satisfied for any new object category. To showcase the practicality of our approach for predicting the 3D shape of arbitrary object categories, we annotate 2D keypoints on giraffe and bear images from COCO -- the annotation process takes less than 1 minute per image."
                    },
                    {
                        "title": "Reconstructing Hand-Held Objects in 3D",
                        "abstract": "Objects manipulated by the hand (i.e., manipulanda) are particularly challenging to reconstruct from in-the-wild RGB images or videos. Not only does the hand occlude much of the object, but also the object is often only visible in a small number of image pixels. At the same time, two strong anchors emerge in this setting: (1) estimated 3D hands help disambiguate the location and scale of the object, and (2) the set of manipulanda is small relative to all possible objects. With these insights in mind, we present a scalable paradigm for handheld object reconstruction that builds on recent breakthroughs in large language/vision models and 3D object datasets. Our model, MCC-Hand-Object (MCC-HO), jointly reconstructs hand and object geometry given a single RGB image and inferred 3D hand as inputs. Subsequently, we use GPT-4(V) to retrieve a 3D object model that matches the object in the image and rigidly align the model to the network-inferred geometry; we call this alignment Retrieval-Augmented Reconstruction (RAR). Experiments demonstrate that MCC-HO achieves state-of-the-art performance on lab and Internet datasets, and we show how RAR can be used to automatically obtain 3D labels for in-the-wild images of hand-object interactions."
                    }
                ]
            },
            "2fa970a2-d5c8-4649-966c-423ad99549e4": {
                "pk": "2fa970a2-d5c8-4649-966c-423ad99549e4",
                "name": "Qixing Huang",
                "collaborators": [
                    "Chen Song",
                    "Bo Sun",
                    "Zhenpei Yang",
                    "Leonidas Guibas",
                    "Siming Yan",
                    "Zaiwei Zhang",
                    "Chandrajit Bajaj",
                    "Jiaxin Lu",
                    "Hanwen Jiang",
                    "Xiangru Huang"
                ],
                "domain": [
                    "3D Vision",
                    "Semantic Segmentation",
                    "Object Detection",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Instance-aware 3D Semantic Segmentation powered by Shape Generators and Classifiers",
                        "abstract": "Existing 3D semantic segmentation methods rely on point-wise or voxel-wise feature descriptors to output segmentation predictions. However, these descriptors are often supervised at point or voxel level, leading to segmentation models that can behave poorly at instance-level. In this paper, we proposed a novel instance-aware approach for 3D semantic segmentation. Our method combines several geometry processing tasks supervised at instance-level to promote the consistency of the learned feature representation. Specifically, our methods use shape generators and shape classifiers to perform shape reconstruction and classification tasks for each shape instance. This enforces the feature representation to faithfully encode both structural and local shape information, with an awareness of shape instances. In the experiments, our method significantly outperform existing approaches in 3D semantic segmentation on several public benchmarks, such as Waymo Open Dataset, SemanticKITTI and ScanNetV2."
                    },
                    {
                        "title": "3D Object Classification via Spherical Projections",
                        "abstract": "In this paper, we introduce a new method for classifying 3D objects. Our main idea is to project a 3D object onto a spherical domain centered around its barycenter and develop neural network to classify the spherical projection. We introduce two complementary projections. The first captures depth variations of a 3D object, and the second captures contour-information viewed from different angles. Spherical projections combine key advantages of two main-stream 3D classification methods: image-based and 3D-based. Specifically, spherical projections are locally planar, allowing us to use massive image datasets (e.g, ImageNet) for pre-training. Also spherical projections are similar to voxel-based methods, as they encode complete information of a 3D object in a single neural network capturing dependencies across different views. Our novel network design can fully utilize these advantages. Experimental results on ModelNet40 and ShapeNetCore show that our method is superior to prior methods."
                    },
                    {
                        "title": "HybridPose: 6D Object Pose Estimation under Hybrid Representations",
                        "abstract": "We introduce HybridPose, a novel 6D object pose estimation approach. HybridPose utilizes a hybrid intermediate representation to express different geometric information in the input image, including keypoints, edge vectors, and symmetry correspondences. Compared to a unitary representation, our hybrid representation allows pose regression to exploit more and diverse features when one type of predicted representation is inaccurate (e.g., because of occlusion). Different intermediate representations used by HybridPose can all be predicted by the same simple neural network, and outliers in predicted intermediate representations are filtered by a robust regression module. Compared to state-of-the-art pose estimation approaches, HybridPose is comparable in running time and accuracy. For example, on Occlusion Linemod dataset, our method achieves a prediction speed of 30 fps with a mean ADD(-S) accuracy of 47.5%, representing a state-of-the-art performance. The implementation of HybridPose is available at https://github.com/chensong1995/HybridPose."
                    },
                    {
                        "title": "Distributable Consistent Multi-Object Matching",
                        "abstract": "In this paper we propose an optimization-based framework to multiple object matching. The framework takes maps computed between pairs of objects as input, and outputs maps that are consistent among all pairs of objects. The central idea of our approach is to divide the input object collection into overlapping sub-collections and enforce map consistency among each sub-collection. This leads to a distributed formulation, which is scalable to large-scale datasets. We also present an equivalence condition between this decoupled scheme and the original scheme. Experiments on both synthetic and real-world datasets show that our framework is competitive against state-of-the-art multi-object matching techniques."
                    },
                    {
                        "title": "Extreme Relative Pose Network under Hybrid Representations",
                        "abstract": "In this paper, we introduce a novel RGB-D based relative pose estimation approach that is suitable for small-overlapping or non-overlapping scans and can output multiple relative poses. Our method performs scene completion and matches the completed scans. However, instead of using a fixed representation for completion, the key idea is to utilize hybrid representations that combine 360-image, 2D image-based layout, and planar patches. This approach offers adaptively feature representations for relative pose estimation. Besides, we introduce a global-2-local matching procedure, which utilizes initial relative poses obtained during the global phase to detect and then integrate geometric relations for pose refinement. Experimental results justify the potential of this approach across a wide range of benchmark datasets. For example, on ScanNet, the rotation translation errors of the top-1/top-5 predictions of our approach are 28.6/0.90m and 16.8/0.76m, respectively. Our approach also considerably boosts the performance of multi-scan reconstruction in few-view reconstruction settings."
                    },
                    {
                        "title": "Joint alignment of multiple protein-protein interaction networks via convex optimization",
                        "abstract": "Motivation: High-throughput experimental techniques have been producing more and more protein-protein interaction (PPI) data. PPI network alignment greatly benefits the understanding of evolutionary relationship among species, helps identify conserved sub-networks and provides extra information for functional annotations. Although a few methods have been developed for multiple PPI network alignment, the alignment quality is still far away from perfect and thus, new network alignment methods are needed. Result: In this paper, we present a novel method, denoted as ConvexAlign, for joint alignment of multiple PPI networks by convex optimization of a scoring function composed of sequence similarity, topological score and interaction conservation score. In contrast to existing methods that generate multiple alignments in a greedy or progressive manner, our convex method optimizes alignments globally and enforces consistency among all pairwise alignments, resulting in much better alignment quality. Tested on both synthetic and real data, our experimental results show that ConvexAlign outperforms several popular methods in producing functionally coherent alignments. ConvexAlign even has a larger advantage over the others in aligning real PPI networks. ConvexAlign also finds a few conserved complexes among 5 species which cannot be detected by the other methods."
                    },
                    {
                        "title": "StarMap for Category-Agnostic Keypoint and Viewpoint Estimation",
                        "abstract": "Semantic keypoints provide concise abstractions for a variety of visual understanding tasks. Existing methods define semantic keypoints separately for each category with a fixed number of semantic labels in fixed indices. As a result, this keypoint representation is in-feasible when objects have a varying number of parts, e.g. chairs with varying number of legs. We propose a category-agnostic keypoint representation, which combines a multi-peak heatmap (StarMap) for all the keypoints and their corresponding features as 3D locations in the canonical viewpoint (CanViewFeature) defined for each instance. Our intuition is that the 3D locations of the keypoints in canonical object views contain rich semantic and compositional information. Using our flexible representation, we demonstrate competitive performance in keypoint detection and localization compared to category-specific state-of-the-art methods. Moreover, we show that when augmented with an additional depth channel (DepthMap) to lift the 2D keypoints to 3D, our representation can achieve state-of-the-art results in viewpoint estimation. Finally, we show that our category-agnostic keypoint representation can be generalized to novel categories."
                    },
                    {
                        "title": "H3DNet: 3D Object Detection Using Hybrid Geometric Primitives",
                        "abstract": "We introduce H3DNet, which takes a colorless 3D point cloud as input and outputs a collection of oriented object bounding boxes (or BB) and their semantic labels. The critical idea of H3DNet is to predict a hybrid set of geometric primitives, i.e., BB centers, BB face centers, and BB edge centers. We show how to convert the predicted geometric primitives into object proposals by defining a distance function between an object and the geometric primitives. This distance function enables continuous optimization of object proposals, and its local minimums provide high-fidelity object proposals. H3DNet then utilizes a matching and refinement module to classify object proposals into detected objects and fine-tune the geometric parameters of the detected objects. The hybrid set of geometric primitives not only provides more accurate signals for object detection than using a single type of geometric primitives, but it also provides an overcomplete set of constraints on the resulting 3D layout. Therefore, H3DNet can tolerate outliers in predicted geometric primitives. Our model achieves state-of-the-art 3D detection results on two large datasets with real 3D scans, ScanNet and SUN RGB-D."
                    },
                    {
                        "title": "MVS2D: Efficient Multi-view Stereo via Attention-Driven 2D Convolutions",
                        "abstract": "Deep learning has made significant impacts on multi-view stereo systems. State-of-the-art approaches typically involve building a cost volume, followed by multiple 3D convolution operations to recover the input image's pixel-wise depth. While such end-to-end learning of plane-sweeping stereo advances public benchmarks' accuracy, they are typically very slow to compute. We present \\ouralg, a highly efficient multi-view stereo algorithm that seamlessly integrates multi-view constraints into single-view networks via an attention mechanism. Since \\ouralg only builds on 2D convolutions, it is at least $2\\times$ faster than all the notable counterparts. Moreover, our algorithm produces precise depth estimations and 3D reconstructions, achieving state-of-the-art results on challenging benchmarks ScanNet, SUN3D, RGBD, and the classical DTU dataset. our algorithm also out-performs all other algorithms in the setting of inexact camera poses. Our code is released at \\url{https://github.com/zhenpeiyang/MVS2D}"
                    },
                    {
                        "title": "E-CIR: Event-Enhanced Continuous Intensity Recovery",
                        "abstract": "A camera begins to sense light the moment we press the shutter button. During the exposure interval, relative motion between the scene and the camera causes motion blur, a common undesirable visual artifact. This paper presents E-CIR, which converts a blurry image into a sharp video represented as a parametric function from time to intensity. E-CIR leverages events as an auxiliary input. We discuss how to exploit the temporal event structure to construct the parametric bases. We demonstrate how to train a deep learning model to predict the function coefficients. To improve the appearance consistency, we further introduce a refinement module to propagate visual features among consecutive frames. Compared to state-of-the-art event-enhanced deblurring approaches, E-CIR generates smoother and more realistic results. The implementation of E-CIR is available at https://github.com/chensong1995/E-CIR."
                    },
                    {
                        "title": "HM3D-ABO: A Photo-realistic Dataset for Object-centric Multi-view 3D Reconstruction",
                        "abstract": "Reconstructing 3D objects is an important computer vision task that has wide application in AR/VR. Deep learning algorithm developed for this task usually relies on an unrealistic synthetic dataset, such as ShapeNet and Things3D. On the other hand, existing real-captured object-centric datasets usually do not have enough annotation to enable supervised training or reliable evaluation. In this technical report, we present a photo-realistic object-centric dataset HM3D-ABO. It is constructed by composing realistic indoor scene and realistic object. For each configuration, we provide multi-view RGB observations, a water-tight mesh model for the object, ground truth depth map and object mask. The proposed dataset could also be useful for tasks such as camera pose estimation and novel-view synthesis. The dataset generation code is released at https://github.com/zhenpeiyang/HM3D-ABO."
                    },
                    {
                        "title": "DeblurSR: Event-Based Motion Deblurring Under the Spiking Representation",
                        "abstract": "We present DeblurSR, a novel motion deblurring approach that converts a blurry image into a sharp video. DeblurSR utilizes event data to compensate for motion ambiguities and exploits the spiking representation to parameterize the sharp output video as a mapping from time to intensity. Our key contribution, the Spiking Representation (SR), is inspired by the neuromorphic principles determining how biological neurons communicate with each other in living organisms. We discuss why the spikes can represent sharp edges and how the spiking parameters are interpreted from the neuromorphic perspective. DeblurSR has higher output quality and requires fewer computing resources than state-of-the-art event-based motion deblurring methods. We additionally show that our approach easily extends to video super-resolution when combined with recent advances in implicit neural representation. The implementation and animated visualization of DeblurSR are available at https://github.com/chensong1995/DeblurSR."
                    },
                    {
                        "title": "Jigsaw: Learning to Assemble Multiple Fractured Objects",
                        "abstract": "Automated assembly of 3D fractures is essential in orthopedics, archaeology, and our daily life. This paper presents Jigsaw, a novel framework for assembling physically broken 3D objects from multiple pieces. Our approach leverages hierarchical features of global and local geometry to match and align the fracture surfaces. Our framework consists of four components: (1) front-end point feature extractor with attention layers, (2) surface segmentation to separate fracture and original parts, (3) multi-parts matching to find correspondences among fracture surface points, and (4) robust global alignment to recover the global poses of the pieces. We show how to jointly learn segmentation and matching and seamlessly integrate feature matching and rigidity constraints. We evaluate Jigsaw on the Breaking Bad dataset and achieve superior performance compared to state-of-the-art methods. Our method also generalizes well to diverse fracture modes, objects, and unseen instances. To the best of our knowledge, this is the first learning-based method designed specifically for 3D fracture assembly over multiple pieces. Our code is available at https://jiaxin-lu.github.io/Jigsaw/."
                    },
                    {
                        "title": "Real3D: Scaling Up Large Reconstruction Models with Real-World Images",
                        "abstract": "The default strategy for training single-view Large Reconstruction Models (LRMs) follows the fully supervised route using large-scale datasets of synthetic 3D assets or multi-view captures. Although these resources simplify the training procedure, they are hard to scale up beyond the existing datasets and they are not necessarily representative of the real distribution of object shapes. To address these limitations, in this paper, we introduce Real3D, the first LRM system that can be trained using single-view real-world images. Real3D introduces a novel self-training framework that can benefit from both the existing synthetic data and diverse single-view real images. We propose two unsupervised losses that allow us to supervise LRMs at the pixel- and semantic-level, even for training examples without ground-truth 3D or novel views. To further improve performance and scale up the image data, we develop an automatic data curation approach to collect high-quality examples from in-the-wild images. Our experiments show that Real3D consistently outperforms prior work in four diverse evaluation settings that include real and synthetic data, as well as both in-domain and out-of-domain shapes. Code and model can be found here: https://hwjiang1510.github.io/Real3D/"
                    },
                    {
                        "title": "PPLNs: Parametric Piecewise Linear Networks for Event-Based Temporal Modeling and Beyond",
                        "abstract": "We present Parametric Piecewise Linear Networks (PPLNs) for temporal vision inference. Motivated by the neuromorphic principles that regulate biological neural behaviors, PPLNs are ideal for processing data captured by event cameras, which are built to simulate neural activities in the human retina. We discuss how to represent the membrane potential of an artificial neuron by a parametric piecewise linear function with learnable coefficients. This design echoes the idea of building deep models from learnable parametric functions recently popularized by Kolmogorov-Arnold Networks (KANs). Experiments demonstrate the state-of-the-art performance of PPLNs in event-based and image-based vision applications, including steering prediction, human pose estimation, and motion deblurring. The source code of our implementation is available at https://github.com/chensong1995/PPLN."
                    },
                    {
                        "title": "Jigsaw++: Imagining Complete Shape Priors for Object Reassembly",
                        "abstract": "The automatic assembly problem has attracted increasing interest due to its complex challenges that involve 3D representation. This paper introduces Jigsaw++, a novel generative method designed to tackle the multifaceted challenges of reconstruction for the reassembly problem. Existing approach focusing primarily on piecewise information for both part and fracture assembly, often overlooking the integration of complete object prior. Jigsaw++ distinguishes itself by learning a category-agnostic shape prior of complete objects. It employs the proposed \"retargeting\" strategy that effectively leverages the output of any existing assembly method to generate complete shape reconstructions. This capability allows it to function orthogonally to the current methods. Through extensive evaluations on Breaking Bad dataset and PartNet, Jigsaw++ has demonstrated its effectiveness, reducing reconstruction errors and enhancing the precision of shape reconstruction, which sets a new direction for future reassembly model developments."
                    },
                    {
                        "title": "Scalable Semidefinite Relaxation for Maximum A Posterior Estimation",
                        "abstract": "Maximum a posteriori (MAP) inference over discrete Markov random fields is a fundamental task spanning a wide spectrum of real-world applications, which is known to be NP-hard for general graphs. In this paper, we propose a novel semidefinite relaxation formulation (referred to as SDR) to estimate the MAP assignment. Algorithmically, we develop an accelerated variant of the alternating direction method of multipliers (referred to as SDPAD-LR) that can effectively exploit the special structure of the new relaxation. Encouragingly, the proposed procedure allows solving SDR for large-scale problems, e.g., problems on a grid graph comprising hundreds of thousands of variables with multiple states per node. Compared with prior SDP solvers, SDPAD-LR is capable of attaining comparable accuracy while exhibiting remarkably improved scalability, in contrast to the commonly held belief that semidefinite relaxation can only been applied on small-scale MRF problems. We have evaluated the performance of SDR on various benchmark datasets including OPENGM2 and PIC in terms of both the quality of the solutions and computation time. Experimental results demonstrate that for a broad class of problems, SDPAD-LR outperforms state-of-the-art algorithms in producing better MAP assignment in an efficient manner."
                    },
                    {
                        "title": "Multi-View Representation is What You Need for Point-Cloud Pre-Training",
                        "abstract": "A promising direction for pre-training 3D point clouds is to leverage the massive amount of data in 2D, whereas the domain gap between 2D and 3D creates a fundamental challenge. This paper proposes a novel approach to point-cloud pre-training that learns 3D representations by leveraging pre-trained 2D networks. Different from the popular practice of predicting 2D features first and then obtaining 3D features through dimensionality lifting, our approach directly uses a 3D network for feature extraction. We train the 3D feature extraction network with the help of the novel 2D knowledge transfer loss, which enforces the 2D projections of the 3D feature to be consistent with the output of pre-trained 2D networks. To prevent the feature from discarding 3D signals, we introduce the multi-view consistency loss that additionally encourages the projected 2D feature representations to capture pixel-wise correspondences across different views. Such correspondences induce 3D geometry and effectively retain 3D features in the projected 2D features. Experimental results demonstrate that our pre-trained model can be successfully transferred to various downstream tasks, including 3D shape classification, part segmentation, 3D object detection, and semantic segmentation, achieving state-of-the-art performance."
                    },
                    {
                        "title": "LEAP: Liberate Sparse-view 3D Modeling from Camera Poses",
                        "abstract": "Are camera poses necessary for multi-view 3D modeling? Existing approaches predominantly assume access to accurate camera poses. While this assumption might hold for dense views, accurately estimating camera poses for sparse views is often elusive. Our analysis reveals that noisy estimated poses lead to degraded performance for existing sparse-view 3D modeling methods. To address this issue, we present LEAP, a novel pose-free approach, therefore challenging the prevailing notion that camera poses are indispensable. LEAP discards pose-based operations and learns geometric knowledge from data. LEAP is equipped with a neural volume, which is shared across scenes and is parameterized to encode geometry and texture priors. For each incoming scene, we update the neural volume by aggregating 2D image features in a feature-similarity-driven manner. The updated neural volume is decoded into the radiance field, enabling novel view synthesis from any viewpoint. On both object-centric and scene-level datasets, we show that LEAP significantly outperforms prior methods when they employ predicted poses from state-of-the-art pose estimators. Notably, LEAP performs on par with prior approaches that use ground-truth poses while running $400\\times$ faster than PixelNeRF. We show LEAP generalizes to novel object categories and scenes, and learns knowledge closely resembles epipolar geometry. Project page: https://hwjiang1510.github.io/LEAP/"
                    }
                ]
            }
        }
    },
    "2310.18297": {
        "paper_data": {
            "title": "Image Clustering Conditioned on Text Criteria",
            "url": "http://arxiv.org/abs/2310.18297v4",
            "arxiv_id": "2310.18297",
            "authors": [
                "Sehyun Kwon",
                "Jaeseung Park",
                "Minkyu Kim",
                "Jaewoong Cho",
                "Ernest K. Ryu",
                "Kangwook Lee"
            ],
            "abstract": "Classical clustering methods do not provide users with direct control of the clustering results, and the clustering results may not be consistent with the relevant criterion that a user has in mind. In this work, we present a new methodology for performing image clustering based on user-specified text criteria by leveraging modern vision-language models and large language models. We call our method Image Clustering Conditioned on Text Criteria (IC|TC), and it represents a different paradigm of image clustering. IC|TC requires a minimal and practical degree of human intervention and grants the user significant control over the clustering results in return. Our experiments show that IC|TC can effectively cluster images with various criteria, such as human action, physical location, or the person's mood, while significantly outperforming baselines.",
            "introduction": "   1 Introduction  Image clustering has been studied as a prototypical unsupervised learning task, and it has been used to organize large volumes of visual data (Platt et\u00a0al., 2003), to reduce the cost of labeling an unlabeled image dataset (Russell et\u00a0al., 2008; Schmarje et\u00a0al., 2022), and to enhance image retrieval systems (Wu et\u00a0al., 2000; J\u00e9gou and Chum, 2012). Modern deep image clustering methods are often evaluated against pre-defined class labels of datasets viewed as the ground truth.    In practice, however, a user may have a criterion in mind for how to cluster or organize a set of images. The user may even want multiple clustering results of the same dataset based on different criteria. (See Figure\u00a01.) But, classical clustering methods offer no direct mechanism for the user to control the clustering criterion; the clustering criteria for existing methods are likely determined by the inductive biases of the neural networks and the loss function, data augmentations, and feature extractors used within the method. This necessitates a new paradigm in image clustering, enabling diverse outcomes from a single dataset based on user-specified criteria and revolutionizing the conventional, implicitly dictated clustering processes.   Recently, foundation models have received significant recent interest due to their ability to understand and follow human instructions at an unprecedented level. Large language models (LLMs) (Brown et\u00a0al., 2020; Chowdhery et\u00a0al., 2022; Touvron et\u00a0al., 2023a; b; Chiang et\u00a0al., 2023; OpenAI, 2023; Adams et\u00a0al., 2023) perform remarkably well on a wide range of natural language tasks such as understanding, summarizing, and reasoning in zero- or few-shot settings. Vision-language models (VLMs)\u00a0(Alayrac et\u00a0al., 2022; Liu et\u00a0al., 2023; Awadalla et\u00a0al., 2023; Dai et\u00a0al., 2023; Li et\u00a0al., 2023a; Zhu et\u00a0al., 2023; Gong et\u00a0al., 2023) interpret natural language instructions in visual contexts and produce responses that seemingly exhibit in-depth image analyses and complex reasoning.   In this work, we present a new methodology based on foundation models for performing image clustering based on user-specified criteria provided in natural language text. We call our method Image Clustering Conditioned on Text Criteria (IC||||TC), and it represents a different paradigm of image clustering: the user directs the method with the relevant clustering criterion, the same dataset can be clustered with multiple different criteria, and if the clustering results are not satisfactory, the user can edit the text criterion to iteratively refine the clustering results. IC||||TC requires a minimal and practical degree of human intervention and grants the user significant control over the clustering results in return, and we argue that this makes IC||||TC more practical and powerful compared to the classical purely unsupervised clustering methods.                   (a)  Sample images from the clustering results on the Stanford 40 Action dataset. Each result is obtained using a different text criterion: Action, Location, and Mood.              (b)  Sample images from the clustering results on the PPMI dataset using the text criterion Instrument with different cluster numbers K=2\ud835\udc3e2K=2italic_K = 2 and 7777.       Figure 1: Sample images from clustering results of IC||||TC. The method finds clusters consistent with the user-specified text criterion. Furthermore, IC||||TC provides cluster names (texts above each image cluster) along with the clusters, enhancing the interpretability of clustering results.      1.1 Contribution  Our main contributions are the",
            "references": [
                {
                    "title": "ChatGPT-Powered Hierarchical Comparisons for Image Classification",
                    "abstract": "The zero-shot open-vocabulary challenge in image classification is tackled by pretrained vision-language models like CLIP, which benefit from incorporating class-specific knowledge from large language models (LLMs) like ChatGPT. However, biases in CLIP lead to similar descriptions for distinct but related classes, prompting our novel image classification framework via hierarchical comparisons: using LLMs to recursively group classes into hierarchies and classifying images by comparing image-text embeddings at each hierarchy level, resulting in an intuitive, effective, and explainable approach."
                },
                {
                    "title": "Image Clustering with External Guidance",
                    "abstract": "The core of clustering is incorporating prior knowledge to construct supervision signals. From classic k-means based on data compactness to recent contrastive clustering guided by self-supervision, the evolution of clustering methods intrinsically corresponds to the progression of supervision signals. At present, substantial efforts have been devoted to mining internal supervision signals from data. Nevertheless, the abundant external knowledge such as semantic descriptions, which naturally conduces to clustering, is regrettably overlooked. In this work, we propose leveraging external knowledge as a new supervision signal to guide clustering, even though it seems irrelevant to the given data. To implement and validate our idea, we design an externally guided clustering method (Text-Aided Clustering, TAC), which leverages the textual semantics of WordNet to facilitate image clustering. Specifically, TAC first selects and retrieves WordNet nouns that best distinguish images to enhance the feature discriminability. Then, to improve image clustering performance, TAC collaborates text and image modalities by mutually distilling cross-modal neighborhood information. Experiments demonstrate that TAC achieves state-of-the-art performance on five widely used and three more challenging image clustering benchmarks, including the full ImageNet-1K dataset."
                },
                {
                    "title": "Learning Spatial-context-aware Global Visual Feature Representation for Instance Image Retrieval",
                    "abstract": "In instance image retrieval, considering local spatial information within an image has proven effective to boost retrieval performance, as demonstrated by local visual descriptor based geometric verification. Nevertheless, it will be highly valuable to make ordinary global image representations spatial-context-aware because global representation based image retrieval is appealing thanks to its algorithmic simplicity, low memory cost, and being friendly to sophisticated data structures. To this end, we propose a novel feature learning framework for instance image retrieval, which embeds local spatial context information into the learned global feature representations. Specifically, in parallel to the visual feature branch in a CNN backbone, we design a spatial context branch that consists of two modules called online token learning and distance encoding. For each local descriptor learned in CNN, the former module is used to indicate the types of its surrounding descriptors, while their spatial distribution information is captured by the latter module. After that, the visual feature branch and the spatial context branch are fused to produce a single global feature representation per image. As experimentally demonstrated, with the spatial-context-aware characteristic, we can well improve the performance of global representation based image retrieval while maintaining all of its appealing properties. Our code is available at https://github.com/Zy-Zhang/SpCa."
                },
                {
                    "title": "From Sparse to Dense: GPT-4 Summarization with Chain of Density Prompting",
                    "abstract": "Selecting the \u201cright\u201d amount of information to include in a summary is a difficult task. A good summary should be detailed and entity-centric without being overly dense and hard to follow. To better understand this tradeoff, we solicit increasingly dense GPT-4 summaries with what we refer to as a \u201cChain of Density\u201d (CoD) prompt. Specifically, GPT-4 generates an initial entity-sparse summary before iteratively incorporating missing salient entities without increasing the length. Summaries generated by CoD are more abstractive, exhibit more fusion, and have less of a lead bias than GPT-4 summaries generated by a vanilla prompt. We conduct a human preference study on 100 CNN DailyMail articles and find that humans prefer GPT-4 summaries that are more dense than those generated by a vanilla prompt and almost as dense as human written summaries. Qualitative analysis supports the notion that there exists a tradeoff between informativeness and readability. 500 annotated CoD summaries, as well as an extra 5,000 unannotated summaries, are freely available on HuggingFace (https://huggingface.co/datasets/griffin/chain_of_density)."
                },
                {
                    "title": "FACET: Fairness in Computer Vision Evaluation Benchmark",
                    "abstract": "Computer vision models have known performance disparities across attributes such as gender and skin tone. This means during tasks such as classification and detection, model performance differs for certain classes based on the demographics of the people in the image. These disparities have been shown to exist, but until now there has not been a unified approach to measure these differences for common use-cases of computer vision models. We present a new benchmark named FACET (FAirness in Computer Vision EvaluaTion), a large, publicly available evaluation set of 32k images for some of the most common vision tasks - image classification, object detection and segmentation. For every image in FACET, we hired expert reviewers to manually annotate person-related attributes such as perceived skin tone and hair type, manually draw bounding boxes and label fine-grained person-related classes such as disk jockey or guitarist. In addition, we use FACET to benchmark state-of-the-art vision models and present a deeper understanding of potential performance disparities and challenges across sensitive demographic attributes. With the exhaustive annotations collected, we probe models using single demographics attributes as well as multiple attributes using an intersectional approach (e.g. hair color and perceived skin tone). Our results show that classification, detection, segmentation, and visual grounding models exhibit performance disparities across demographic attributes and intersections of attributes. These harms suggest that not all people represented in datasets receive fair and equitable treatment in these vision tasks. We hope current and future results using our benchmark will contribute to fairer, more robust vision models. FACET is available publicly at https://facet.metademolab.com."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Large Language Models Enable Few-Shot Clustering",
                    "abstract": "Unlike traditional unsupervised clustering, semi-supervised clustering allows users to provide meaningful structure to the data, which helps the clustering algorithm to match the user\u2019s intent. Existing approaches to semi-supervised clustering require a significant amount of feedback from an expert to improve the clusters. In this paper, we ask whether a large language model (LLM) can amplify an expert\u2019s guidance to enable query-efficient, few-shot semi-supervised text clustering. We show that LLMs are surprisingly effective at improving clustering. We explore three stages where LLMs can be incorporated into clustering: before clustering (improving input features), during clustering (by providing constraints to the clusterer), and after clustering (using LLMs post-correction). We find that incorporating LLMs in the first two stages routinely provides significant improvements in cluster quality, and that LLMs enable a user to make trade-offs between cost and accuracy to produce desired clusters. We release our code and LLM prompts for the public to use.1"
                },
                {
                    "title": "PointClustering: Unsupervised Point Cloud Pre-training using Transformation Invariance in Clustering",
                    "abstract": "Feature invariance under different data transformations, i.e., transformation invariance, can be regarded as a type of self-supervision for representation learning. In this paper, we present PointClustering, a new unsupervised representation learning scheme that leverages transformation invariance for point cloud pre-training. PointClustering formulates the pretext task as deep clustering and employs transformation invariance as an inductive bias, following the philosophy that common point cloud transformation will not change the geometric properties and semantics. Technically, PointClustering iteratively optimizes the feature clusters and backbone, and delves into the transformation invariance as learning regularization from two perspectives: point level and instance level. Point-level invariance learning maintains local geometric properties through gathering point features of one instance across transformations, while instance-level invariance learning further measures clusters over the entire dataset to explore semantics of instances. Our PointClustering is architecture-agnostic and readily applicable to MLP-based, CNN-based and Transformer-based backbones. We empirically demonstrate that the models pre-learnt on the ScanNet dataset by PointClustering provide superior performances on six benchmark-s, across downstream tasks of classification and segmentation. More remarkably, PoinrClustering achieves an accuracy of 94.5% on ModelNet40 with Transformer backbone. Source code is available at https://github.com/FuchenUSTC/PointClustering."
                },
                {
                    "title": "Asymmetric Feature Fusion for Image Retrieval",
                    "abstract": "In asymmetric retrieval systems, models with different capacities are deployed on platforms with different computational and storage resources. Despite the great progress, existing approaches still suffer from a dilemma between retrieval efficiency and asymmetric accuracy due to the limited capacity of the lightweight query model. In this work, we propose an Asymmetric Feature Fusion (AFF) paradigm, which advances existing asymmetric retrieval systems by considering the complementarity among different features just at the gallery side. Specifically, it first embeds each gallery image into various features, e.g., local features and global features. Then, a dynamic mixer is introduced to aggregate these features into compact embedding for efficient search. On the query side, only a single lightweight model is deployed for feature extraction. The query model and dynamic mixer are jointly trained by sharing a momentum-updated classifier. Notably, the proposed paradigm boosts the accuracy of asymmetric retrieval without introducing any extra overhead to the query side. Exhaustive experiments on various landmark retrieval datasets demonstrate the superiority of our paradigm."
                },
                {
                    "title": "Revisiting Self-Similarity: Structural Embedding for Image Retrieval",
                    "abstract": "Despite advances in global image representation, existing image retrieval approaches rarely consider geometric structure during the global retrieval stage. In this work, we revisit the conventional self-similarity descriptor from a convolutional perspective, to encode both the visual and structural cues of the image to global image representation. Our proposed network, named Structural Embedding Network (SENet), captures the internal structure of the images and gradually compresses them into dense self-similarity descriptors while learning diverse structures from various images. These self-similarity descriptors and original image features are fused and then pooled into global embedding, so that global embedding can represent both geometric and visual cues of the image. Along with this novel structural embedding, our proposed network sets new state-of-the-art performances on several image retrieval benchmarks, convincing its robustness to look-alike distractors. The code and models are available: https://github.com/sungonce/SENet."
                },
                {
                    "title": "InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning",
                    "abstract": "Large-scale pre-training and instruction tuning have been successful at creating general-purpose language models with broad competence. However, building general-purpose vision-language models is challenging due to the rich input distributions and task diversity resulting from the additional visual input. Although vision-language pretraining has been widely studied, vision-language instruction tuning remains under-explored. In this paper, we conduct a systematic and comprehensive study on vision-language instruction tuning based on the pretrained BLIP-2 models. We gather 26 publicly available datasets, covering a wide variety of tasks and capabilities, and transform them into instruction tuning format. Additionally, we introduce an instruction-aware Query Transformer, which extracts informative features tailored to the given instruction. Trained on 13 held-in datasets, InstructBLIP attains state-of-the-art zero-shot performance across all 13 held-out datasets, substantially outperforming BLIP-2 and larger Flamingo models. Our models also lead to state-of-the-art performance when finetuned on individual downstream tasks (e.g., 90.7% accuracy on ScienceQA questions with image contexts). Furthermore, we qualitatively demonstrate the advantages of InstructBLIP over concurrent multimodal models. All InstructBLIP models are open-sourced at https://github.com/salesforce/LAVIS/tree/main/projects/instructblip."
                },
                {
                    "title": "MultiModal-GPT: A Vision and Language Model for Dialogue with Humans",
                    "abstract": "We present a vision and language model named MultiModal-GPT to conduct multi-round dialogue with humans. MultiModal-GPT can follow various instructions from humans, such as generating a detailed caption, counting the number of interested objects, and answering general questions from users. MultiModal-GPT is parameter-efficiently fine-tuned from OpenFlamingo, with Low-rank Adapter (LoRA) added both in the cross-attention part and the self-attention part of the language model. We first construct instruction templates with vision and language data for multi-modality instruction tuning to make the model understand and follow human instructions. We find the quality of training data is vital for the dialogue performance, where few data containing short answers can lead the model to respond shortly to any instructions. To further enhance the ability to chat with humans of the MultiModal-GPT, we utilize language-only instruction-following data to train the MultiModal-GPT jointly. The joint training of language-only and visual-language instructions with the \\emph{same} instruction template effectively improves dialogue performance. Various demos show the ability of continuous dialogue of MultiModal-GPT with humans. Code, dataset, and demo are at https://github.com/open-mmlab/Multimodal-GPT"
                },
                {
                    "title": "Otter: A Multi-Modal Model with In-Context Instruction Tuning",
                    "abstract": "Large language models (LLMs) have demonstrated significant universal capabilities as few/zero-shot learners in various tasks due to their pre-training on vast amounts of text data, as exemplified by GPT-3, which boosted to InstrctGPT and ChatGPT, effectively following natural language instructions to accomplish real-world tasks. In this paper, we propose to introduce instruction tuning into multi-modal models, motivated by the Flamingo model's upstream interleaved format pretraining dataset. We adopt a similar approach to construct our MultI-Modal In-Context Instruction Tuning (MIMIC-IT) dataset. We then introduce Otter, a multi-modal model based on OpenFlamingo (open-sourced version of DeepMind's Flamingo), trained on MIMIC-IT and showcasing improved instruction-following ability and in-context learning. We also optimize OpenFlamingo's implementation for researchers, democratizing the required training resources from 1$\\times$ A100 GPU to 4$\\times$ RTX-3090 GPUs, and integrate both OpenFlamingo and Otter into Huggingface Transformers for more researchers to incorporate the models into their customized training and inference pipelines."
                },
                {
                    "title": "Rotation and Translation Invariant Representation Learning with Implicit Neural Representations",
                    "abstract": "In many computer vision applications, images are acquired with arbitrary or random rotations and translations, and in such setups, it is desirable to obtain semantic representations disentangled from the image orientation. Examples of such applications include semiconductor wafer defect inspection, plankton microscope images, and inference on single-particle cryo-electron microscopy (cryo-EM) micro-graphs. In this work, we propose Invariant Representation Learning with Implicit Neural Representation (IRL-INR), which uses an implicit neural representation (INR) with a hypernetwork to obtain semantic representations disentangled from the orientation of the image. We show that IRL-INR can effectively learn disentangled semantic representations on more complex images compared to those considered in prior works and show that these semantic representations synergize well with SCAN to produce state-of-the-art unsupervised clustering results."
                },
                {
                    "title": "MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models",
                    "abstract": "The recent GPT-4 has demonstrated extraordinary multi-modal abilities, such as directly generating websites from handwritten text and identifying humorous elements within images. These features are rarely observed in previous vision-language models. However, the technical details behind GPT-4 continue to remain undisclosed. We believe that the enhanced multi-modal generation capabilities of GPT-4 stem from the utilization of sophisticated large language models (LLM). To examine this phenomenon, we present MiniGPT-4, which aligns a frozen visual encoder with a frozen advanced LLM, Vicuna, using one projection layer. Our work, for the first time, uncovers that properly aligning the visual features with an advanced large language model can possess numerous advanced multi-modal abilities demonstrated by GPT-4, such as detailed image description generation and website creation from hand-drawn drafts. Furthermore, we also observe other emerging capabilities in MiniGPT-4, including writing stories and poems inspired by given images, teaching users how to cook based on food photos, and so on. In our experiment, we found that the model trained on short image caption pairs could produce unnatural language outputs (e.g., repetition and fragmentation). To address this problem, we curate a detailed image description dataset in the second stage to finetune the model, which consequently improves the model's generation reliability and overall usability. Our code, pre-trained model, and collected dataset are available at https://minigpt-4.github.io/."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "DivClust: Controlling Diversity in Deep Clustering",
                    "abstract": "Clustering has been a major research topic in the field of machine learning, one to which Deep Learning has recently been applied with significant success. However, an aspect of clustering that is not addressed by existing deep clustering methods, is that of efficiently producing multiple, diverse partitionings for a given dataset. This is particularly important, as a diverse set of base clusterings are necessary for consensus clustering, which has been found to produce better and more robust results than relying on a single clustering. To address this gap, we propose Div-Clust, a diversity controlling loss that can be incorporated into existing deep clustering frameworks to produce multiple clusterings with the desired degree of diversity. We conduct experiments with multiple datasets and deep clustering frameworks and show that: a) our method effectively controls diversity across frameworks and datasets with very small additional computational cost, b) the sets of clusterings learned by DivClust include solutions that significantly outperform single-clustering baselines, and c) using an off-the-shelf consensus clustering algorithm, DivClust produces consensus clustering solutions that consistently outperform single-clustering baselines, effectively improving the performance of the base deep clustering framework. Code is available at https://github.com/ManiadisG/DivClust."
                },
                {
                    "title": "HiCLIP: Contrastive Language-Image Pretraining with Hierarchy-aware Attention",
                    "abstract": "The success of large-scale contrastive vision-language pretraining (CLIP) has benefited both visual recognition and multimodal content understanding. The concise design brings CLIP the advantage in inference efficiency against other vision-language models with heavier cross-attention fusion layers, making it a popular choice for a wide spectrum of downstream tasks. However, CLIP does not explicitly capture the hierarchical nature of high-level and fine-grained semantics conveyed in images and texts, which is arguably critical to vision-language understanding and reasoning. To this end, we equip both the visual and language branches in CLIP with hierarchy-aware attentions, namely Hierarchy-aware CLIP (HiCLIP), to progressively discover semantic hierarchies layer-by-layer from both images and texts in an unsupervised manner. As a result, such hierarchical aggregation significantly improves the cross-modal alignment. To demonstrate the advantages of HiCLIP, we conduct qualitative analysis on its unsupervised hierarchy induction during inference, as well as extensive quantitative experiments on both visual recognition and vision-language downstream tasks."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "Fashion Image Retrieval with Text Feedback by Additive Attention Compositional Learning",
                    "abstract": "Effective fashion image retrieval with text feedback stands to impact a range of real-world applications, such as e-commerce. Given a source image and text feedback that describes the desired modifications to that image, the goal is to retrieve the target images that resemble the source yet satisfy the given modifications by composing a multi-modal (image-text) query. We propose a novel solution to this problem, Additive Attention Compositional Learning (AACL), that uses a multi-modal transformer-based architecture and effectively models the image-text contexts. Specifically, we propose a novel image-text composition module based on additive attention that can be seamlessly plugged into deep neural networks. We also introduce a new challenging benchmark derived from the Shopping100k dataset. AACL is evaluated on three large-scale datasets (FashionIQ, Fashion200k, and Shopping100k), each with strong baselines. Extensive experiments show that AACL achieves new state-of-the-art results on all three datasets."
                },
                {
                    "title": "Visual Classification via Description from Large Language Models",
                    "abstract": "Vision-language models (VLMs) such as CLIP have shown promising performance on a variety of recognition tasks using the standard zero-shot classification procedure -- computing similarity between the query image and the embedded words for each category. By only using the category name, they neglect to make use of the rich context of additional information that language affords. The procedure gives no intermediate understanding of why a category is chosen, and furthermore provides no mechanism for adjusting the criteria used towards this decision. We present an alternative framework for classification with VLMs, which we call classification by description. We ask VLMs to check for descriptive features rather than broad categories: to find a tiger, look for its stripes; its claws; and more. By basing decisions on these descriptors, we can provide additional cues that encourage using the features we want to be used. In the process, we can get a clear idea of what features the model uses to construct its decision; it gains some level of inherent explainability. We query large language models (e.g., GPT-3) for these descriptors to obtain them in a scalable way. Extensive experiments show our framework has numerous advantages past interpretability. We show improvements in accuracy on ImageNet across distribution shifts; demonstrate the ability to adapt VLMs to recognize concepts unseen during training; and illustrate how descriptors can be edited to effectively mitigate bias compared to the baseline."
                },
                {
                    "title": "Deep Fair Clustering via Maximizing and Minimizing Mutual Information: Theory, Algorithm and Metric",
                    "abstract": "Fair clustering aims to divide data into distinct clusters while preventing sensitive attributes (e.g., gender, race, RNA sequencing technique) from dominating the clustering. Although a number of works have been conducted and achieved huge success recently, most of them are heuristical, and there lacks a unified theory for algorithm design. In this work, we fill this blank by developing a mutual information theory for deep fair clustering and accordingly designing a novel algorithm, dubbed FCMI. In brief, through maximizing and minimizing mutual information, FCMI is designed to achieve four characteristics highly expected by deep fair clustering, i.e., compact, balanced, and fair clusters, as well as informative features. Besides the contributions to theory and algorithm, another contribution of this work is proposing a novel fair clustering metric built upon information theory as well. Unlike existing evaluation metrics, our metric measures the clustering quality and fairness as a whole instead of separate manner. To verify the effectiveness of the proposed FCMI, we conduct experiments on six benchmarks including a single-cell RNA-seq atlas compared with 11 state-of-the-art methods in terms of five metrics. The code could be accessed from https://pengxi.me."
                },
                {
                    "title": "Semantic-enhanced Image Clustering",
                    "abstract": "Image clustering is an important and open challenging task in computer vision. Although many methods have been proposed to solve the image clustering task, they only explore images and uncover clusters according to the image features, thus being unable to distinguish visually similar but semantically different images. In this paper, we propose to investigate the task of image clustering with the help of visual-language pre-training model. Different from the zero-shot setting, in which the class names are known, we only know the number of clusters in this setting. Therefore, how to map images to a proper semantic space and how to cluster images from both image and semantic spaces are two key problems. To solve the above problems, we propose a novel image clustering method guided by the visual-language pre-training model CLIP, named Semantic-Enhanced Image Clustering (SIC). In this new method, we propose a method to map the given images to a proper semantic space first and efficient methods to generate pseudo-labels according to the relationships between images and semantics. Finally, we propose to perform clustering with consistency learning in both image space and semantic space, in a self-supervised learning fashion. The theoretical result of convergence analysis shows that our proposed method can converge at a sublinear speed. Theoretical analysis of expectation risk also shows that we can reduce the expectation risk by improving neighborhood consistency, increasing prediction confidence, or reducing neighborhood imbalance. Experimental results on five benchmark datasets clearly show the superiority of our new method."
                },
                {
                    "title": "LIFT: Language-Interfaced Fine-Tuning for Non-Language Machine Learning Tasks",
                    "abstract": "Fine-tuning pretrained language models (LMs) without making any architectural changes has become a norm for learning various language downstream tasks. However, for non-language downstream tasks, a common practice is to employ task-specific designs for input, output layers, and loss functions. For instance, it is possible to fine-tune an LM into an MNIST classifier by replacing the word embedding layer with an image patch embedding layer, the word token output layer with a 10-way output layer, and the word prediction loss with a 10-way classification loss, respectively. A natural question arises: Can LM fine-tuning solve non-language downstream tasks without changing the model architecture or loss function? To answer this, we propose Language-Interfaced Fine-Tuning (LIFT) and study its efficacy and limitations by conducting an extensive empirical study on a suite of non-language classification and regression tasks. LIFT does not make any changes to the model architecture or loss function, and it solely relies on the natural language interface, enabling\"no-code machine learning with LMs.\"We find that LIFT performs comparably well across a wide range of low-dimensional classification and regression tasks, matching the performances of the best baselines in many cases, especially for the classification tasks. We also report experimental results on the fundamental properties of LIFT, including inductive bias, robustness, and sample complexity. We also analyze the effect of pretraining on LIFT and a few properties/techniques specific to LIFT, e.g., context-aware learning via appropriate prompting, calibrated predictions, data generation, and two-stage fine-tuning. Our code is available at https://github.com/UW-Madison-Lee-Lab/LanguageInterfacedFineTuning."
                },
                {
                    "title": "GLIPv2: Unifying Localization and Vision-Language Understanding",
                    "abstract": "We present GLIPv2, a grounded VL understanding model, that serves both localization tasks (e.g., object detection, instance segmentation) and Vision-Language (VL) understanding tasks (e.g., VQA, image captioning). GLIPv2 elegantly unifies localization pre-training and Vision-Language Pre-training (VLP) with three pre-training tasks: phrase grounding as a VL reformulation of the detection task, region-word contrastive learning as a novel region-word level contrastive learning task, and the masked language modeling. This unification not only simplifies the previous multi-stage VLP procedure but also achieves mutual benefits between localization and understanding tasks. Experimental results show that a single GLIPv2 model (all model weights are shared) achieves near SoTA performance on various localization and understanding tasks. The model also shows (1) strong zero-shot and few-shot adaption performance on open-vocabulary object detection tasks and (2) superior grounding capability on VL understanding tasks. Code will be released at https://github.com/microsoft/GLIP."
                },
                {
                    "title": "Conditioned and composed image retrieval combining and partially fine-tuning CLIP-based features",
                    "abstract": "In this paper, we present an approach for conditioned and composed image retrieval based on CLIP features. In this extension of content-based image retrieval (CBIR) an image is combined with a text that provides information regarding user intentions, and is relevant for application domains like e-commerce. The proposed method is based on an initial training stage where a simple combination of visual and textual features is used, to fine-tune the CLIP text encoder. Then in a second training stage we learn a more complex combiner network that merges visual and textual features. Contrastive learning is used in both stages. The proposed approach obtains state-of-the-art performance for conditioned CBIR on the FashionIQ dataset and for composed CBIR on the more recent CIRR dataset."
                },
                {
                    "title": "Flamingo: a Visual Language Model for Few-Shot Learning",
                    "abstract": "Building models that can be rapidly adapted to novel tasks using only a handful of annotated examples is an open challenge for multimodal machine learning research. We introduce Flamingo, a family of Visual Language Models (VLM) with this ability. We propose key architectural innovations to: (i) bridge powerful pretrained vision-only and language-only models, (ii) handle sequences of arbitrarily interleaved visual and textual data, and (iii) seamlessly ingest images or videos as inputs. Thanks to their flexibility, Flamingo models can be trained on large-scale multimodal web corpora containing arbitrarily interleaved text and images, which is key to endow them with in-context few-shot learning capabilities. We perform a thorough evaluation of our models, exploring and measuring their ability to rapidly adapt to a variety of image and video tasks. These include open-ended tasks such as visual question-answering, where the model is prompted with a question which it has to answer; captioning tasks, which evaluate the ability to describe a scene or an event; and close-ended tasks such as multiple-choice visual question-answering. For tasks lying anywhere on this spectrum, a single Flamingo model can achieve a new state of the art with few-shot learning, simply by prompting the model with task-specific examples. On numerous benchmarks, Flamingo outperforms models fine-tuned on thousands of times more task-specific data."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "GLaM: Efficient Scaling of Language Models with Mixture-of-Experts",
                    "abstract": "Scaling language models with more data, compute and parameters has driven significant progress in natural language processing. For example, thanks to scaling, GPT-3 was able to achieve strong results on in-context learning tasks. However, training these large dense models requires significant amounts of computing resources. In this paper, we propose and develop a family of language models named GLaM (Generalist Language Model), which uses a sparsely activated mixture-of-experts architecture to scale the model capacity while also incurring substantially less training cost compared to dense variants. The largest GLaM has 1.2 trillion parameters, which is approximately 7x larger than GPT-3. It consumes only 1/3 of the energy used to train GPT-3 and requires half of the computation flops for inference, while still achieving better overall zero-shot and one-shot performance across 29 NLP tasks."
                },
                {
                    "title": "Grounded Language-Image Pre-training",
                    "abstract": "This paper presents a grounded language-image pretraining (GLIP) model for learning object-level, language-aware, and semantic-rich visual representations. GLIP unifies object detection and phrase grounding for pre-training. The unification brings two benefits: 1) it allows GLIP to learn from both detection and grounding data to improve both tasks and bootstrap a good grounding model; 2) GLIP can leverage massive image-text pairs by generating grounding boxes in a self-training fashion, making the learned representations semantic-rich. In our experiments, we pre-train GLIP on 27M grounding data, including 3M human-annotated and 24M web-crawled image-text pairs. The learned representations demonstrate strong zero-shot and few-shot transferability to various object-level recognition tasks. 1) When directly evaluated on COCO and LVIS (without seeing any images in COCO during pre-training), GLIP achieves 49.8 AP and 26.9 AP, respectively, surpassing many supervised baselines.11Supervised baselines on COCO object detection: Faster-RCNN w/ ResNet50 (40.2) or ResNet101 (42.0), and DyHead w/ Swin-Tiny (49.7). 2) After fine-tuned on COCO, GLIP achieves 60.8 AP on val and 61.5 AP on test-dev, surpassing prior SoTA. 3) When transferred to 13 downstream object detection tasks, a 1-shot GLIP rivals with a fully-supervised Dynamic Head. Code will be released at https://github.com/microsoft/GLIP."
                },
                {
                    "title": "ClipCap: CLIP Prefix for Image Captioning",
                    "abstract": "Image captioning is a fundamental task in vision-language understanding, where the model predicts a textual informative caption to a given input image. In this paper, we present a simple approach to address this task. We use CLIP encoding as a prefix to the caption, by employing a simple mapping network, and then fine-tunes a language model to generate the image captions. The recently proposed CLIP model contains rich semantic features which were trained with textual context, making it best for vision-language perception. Our key idea is that together with a pre-trained language model (GPT2), we obtain a wide understanding of both visual and textual data. Hence, our approach only requires rather quick training to produce a competent captioning model. Without additional annotations or pre-training, it efficiently generates meaningful captions for large-scale and diverse datasets. Surprisingly, our method works well even when only the mapping network is trained, while both CLIP and the language model remain frozen, allowing a lighter architecture with less trainable parameters. Through quantitative evaluation, we demonstrate our model achieves comparable results to state-of-the-art methods on the challenging Conceptual Captions and nocaps datasets, while it is simpler, faster, and lighter. Our code is available in https://github.com/rmokady/CLIP_prefix_caption."
                },
                {
                    "title": "Multitask Prompted Training Enables Zero-Shot Task Generalization",
                    "abstract": "Large language models have recently been shown to attain reasonable zero-shot generalization on a diverse set of tasks (Brown et al., 2020). It has been hypothesized that this is a consequence of implicit multitask learning in language models' pretraining (Radford et al., 2019). Can zero-shot generalization instead be directly induced by explicit multitask learning? To test this question at scale, we develop a system for easily mapping any natural language tasks into a human-readable prompted form. We convert a large set of supervised datasets, each with multiple prompts with diverse wording. These prompted datasets allow for benchmarking the ability of a model to perform completely held-out tasks. We fine-tune a pretrained encoder-decoder model (Raffel et al., 2020; Lester et al., 2021) on this multitask mixture covering a wide variety of tasks. The model attains strong zero-shot performance on several standard datasets, often outperforming models up to 16x its size. Further, our approach attains strong performance on a subset of tasks from the BIG-bench benchmark, outperforming models up to 6x its size. All trained models are available at https://github.com/bigscience-workshop/t-zero and all prompts are available at https://github.com/bigscience-workshop/promptsource."
                },
                {
                    "title": "Finetuned Language Models Are Zero-Shot Learners",
                    "abstract": "This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning."
                },
                {
                    "title": "Image Retrieval on Real-life Images with Pre-trained Vision-and-Language Models",
                    "abstract": "We extend the task of composed image retrieval, where an input query consists of an image and short textual description of how to modify the image. Existing methods have only been applied to non-complex images within narrow domains, such as fashion products, thereby limiting the scope of study on in-depth visual reasoning in rich image and language contexts. To address this issue, we collect the Compose Image Retrieval on Real-life images (CIRR) dataset, which consists of over 36,000 pairs of crowd-sourced, open-domain images with human-generated modifying text. To extend current methods to the open-domain, we propose CIRPLANT, a transformer based model that leverages rich pre-trained vision-and-language (V&L) knowledge for modifying visual features conditioned on natural language. Retrieval is then done by nearest neighbor lookup on the modified features. We demonstrate that with a relatively simple architecture, CIRPLANT outperforms existing methods on open-domain images, while matching state-of-the-art accuracy on the existing narrow datasets, such as fashion. Together with the release of CIRR, we believe this work will inspire further research on composed image retrieval. Our dataset, code and pre-trained models are available at https://cuberick-orion.github.io/CIRR/."
                },
                {
                    "title": "You Never Cluster Alone",
                    "abstract": "Recent advances in self-supervised learning with instance-level contrastive objectives facilitate unsupervised clustering. However, a standalone datum is not perceiving the context of the holistic cluster, and may undergo sub-optimal assignment. In this paper, we extend the mainstream contrastive learning paradigm to a cluster-level scheme, where all the data subjected to the same cluster contribute to a unified representation that encodes the context of each data group. Contrastive learning with this representation then rewards the assignment of each datum. To implement this vision, we propose twin-contrast clustering (TCC). We define a set of categorical variables as clustering assignment confidence, which links the instance-level learning track with the cluster-level one. On one hand, with the corresponding assignment variables being the weight, a weighted aggregation along the data points implements the set representation of a cluster. We further propose heuristic cluster augmentation equivalents to enable cluster-level contrastive learning. On the other hand, we derive the evidence lower-bound of the instance-level contrastive objective with the assignments. By reparametrizing the assignment variables, TCC is trained end-to-end, requiring no alternating steps. Extensive experiments show that TCC outperforms the state-of-the-art on challenging benchmarks."
                },
                {
                    "title": "Graph Contrastive Clustering",
                    "abstract": "Recently, some contrastive learning methods have been proposed to simultaneously learn representations and clustering assignments, achieving significant improvements. However, these methods do not take the category information and clustering objective into consideration, thus the learned representations are not optimal for clustering and the performance might be limited. Towards this issue, we first propose a novel graph contrastive learning framework, and then apply it to the clustering task, resulting in the Graph Constrastive Clustering (GCC) method. Different from basic contrastive clustering that only assumes an image and its augmentation should share similar representation and clustering assignments, we lift the instance-level consistency to the cluster-level consistency with the assumption that samples in one cluster and their augmentations should all be similar. Specifically, on the one hand, we propose the graph Laplacian based contrastive loss to learn more discriminative and clustering-friendly features. On the other hand, we propose a novel graph-based contrastive learning strategy to learn more compact clustering assignments. Both of them incorporate the latent category information to reduce the intra-cluster variance as well as increase the inter-cluster variance. Experiments on six commonly used datasets demonstrate the superiority of our proposed approach over the state-of-the-art methods. 1"
                },
                {
                    "title": "PiCIE: Unsupervised Semantic Segmentation using Invariance and Equivariance in Clustering",
                    "abstract": "We present a new framework for semantic segmentation without annotations via clustering. Off-the-shelf clustering methods are limited to curated, single-label, and objectcentric images yet real-world data are dominantly uncurated, multi-label, and scene-centric. We extend clustering from images to pixels and assign separate cluster membership to different instances within each image. However, solely relying on pixel-wise feature similarity fails to learn high-level semantic concepts and overfits to lowlevel visual cues. We propose a method to incorporate geometric consistency as an inductive bias to learn invariance and equivariance for photometric and geometric variations. With our novel learning objective, our framework can learn high-level semantic concepts. Our method, PiCIE (Pixel-level feature Clustering using Invariance and Equivariance), is the first method capable of segmenting both things and stuff categories without any hyperparameter tuning or task-specific pre-processing. Our method largely outperforms existing baselines on COCO [31] and Cityscapes [8] with +17.5 Acc. and +4.5 mIoU. We show that PiCIE gives a better initialization for standard supervised training. The code is available at https://github.com/janghyuncho/PiCIE."
                },
                {
                    "title": "SPICE: Semantic Pseudo-Labeling for Image Clustering",
                    "abstract": "The similarity among samples and the discrepancy among clusters are two crucial aspects of image clustering. However, current deep clustering methods suffer from inaccurate estimation of either feature similarity or semantic discrepancy. In this paper, we present a Semantic Pseudo-labeling-based Image ClustEring (SPICE) framework, which divides the clustering network into a feature model for measuring the instance-level similarity and a clustering head for identifying the cluster-level discrepancy. We design two semantics-aware pseudo-labeling algorithms, prototype pseudo-labeling and reliable pseudo-labeling, which enable accurate and reliable self-supervision over clustering. Without using any ground-truth label, we optimize the clustering network in three stages: 1) train the feature model through contrastive learning to measure the instance similarity; 2) train the clustering head with the prototype pseudo-labeling algorithm to identify cluster semantics; and 3) jointly train the feature model and clustering head with the reliable pseudo-labeling algorithm to improve the clustering performance. Extensive experimental results demonstrate that SPICE achieves significant improvements (~10%) over existing methods and establishes the new state-of-the-art clustering results on six balanced benchmark datasets in terms of three popular metrics. Importantly, SPICE significantly reduces the gap between unsupervised and fully-supervised classification; e.g. there is only 2% (91.8% vs 93.8%) accuracy difference on CIFAR-10. Our code is made publicly available at https://github.com/niuchuangnn/SPICE."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision",
                    "abstract": "Pre-trained representations are becoming crucial for many NLP and perception tasks. While representation learning in NLP has transitioned to training on raw text without human annotations, visual and vision-language representations still rely heavily on curated training datasets that are expensive or require expert knowledge. For vision applications, representations are mostly learned using datasets with explicit class labels such as ImageNet or OpenImages. For vision-language, popular datasets like Conceptual Captions, MSCOCO, or CLIP all involve a non-trivial data collection (and cleaning) process. This costly curation process limits the size of datasets and hence hinders the scaling of trained models. In this paper, we leverage a noisy dataset of over one billion image alt-text pairs, obtained without expensive filtering or post-processing steps in the Conceptual Captions dataset. A simple dual-encoder architecture learns to align visual and language representations of the image and text pairs using a contrastive loss. We show that the scale of our corpus can make up for its noise and leads to state-of-the-art representations even with such a simple learning scheme. Our visual representation achieves strong performance when transferred to classification tasks such as ImageNet and VTAB. The aligned visual and language representations enables zero-shot image classification and also set new state-of-the-art results on Flickr30K and MSCOCO image-text retrieval benchmarks, even when compared with more sophisticated cross-attention models. The representations also enable cross-modality search with complex text and text + image queries."
                },
                {
                    "title": "Improving Unsupervised Image Clustering With Robust Learning",
                    "abstract": "Unsupervised image clustering methods often introduce alternative objectives to indirectly train the model and are subject to faulty predictions and overconfident results. To overcome these challenges, the current research proposes an innovative model RUC that is inspired by robust learning. RUC\u2019s novelty is at utilizing pseudo-labels of existing image clustering models as a noisy dataset that may include misclassified samples. Its retraining process can revise misaligned knowledge and alleviate the overconfidence problem in predictions. The model\u2019s flexible structure makes it possible to be used as an add-on module to other clustering methods and helps them achieve better performance on multiple datasets. Extensive experiments show that the proposed model can adjust the model confidence with better calibration and gain additional robustness against adversarial noise."
                },
                {
                    "title": "Deep Fair Clustering for Visual Learning",
                    "abstract": "Fair clustering aims to hide sensitive attributes during data partition by balancing the distribution of protected subgroups in each cluster. Existing work attempts to address this problem by reducing it to a classical balanced clustering with a constraint on the proportion of protected subgroups of the input space. However, the input space may limit the clustering performance, and so far only low-dimensional datasets have been considered. In light of these limitations, in this paper, we propose Deep Fair Clustering (DFC) to learn fair and clustering-favorable representations for clustering simultaneously. Our approach could effectively filter out sensitive attributes from representations, and also lead to representations that are amenable for the following cluster analysis. Theoretically, we show that our fairness constraint in DFC will not incur much loss in terms of several clustering metrics. Empirically, we provide extensive experimental demonstrations on four visual datasets to corroborate the superior performance of the proposed approach over existing fair clustering and deep clustering methods on both cluster validity and fairness criterion."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Self-Supervised Learning of Pretext-Invariant Representations",
                    "abstract": "The goal of self-supervised learning from images is to construct image representations that are semantically meaningful via pretext tasks that do not require semantic annotations. Many pretext tasks lead to representations that are covariant with image transformations. We argue that, instead, semantic representations ought to be invariant under such transformations. Specifically, we develop Pretext-Invariant Representation Learning (PIRL, pronounced as `pearl') that learns invariant representations based on pretext tasks. We use PIRL with a commonly used pretext task that involves solving jigsaw puzzles. We find that PIRL substantially improves the semantic quality of the learned image representations. Our approach sets a new state-of-the-art in self-supervised learning from images on several popular benchmarks for self-supervised learning. Despite being unsupervised, PIRL outperforms supervised pre-training in learning image representations for object detection. Altogether, our results demonstrate the potential of self-supervised representations with good invariance properties."
                },
                {
                    "title": "Local Features and Visual Words Emerge in Activations",
                    "abstract": "We propose a novel method of deep spatial matching (DSM) for image retrieval. Initial ranking is based on image descriptors extracted from convolutional neural network activations by global pooling, as in recent state-of-the-art work. However, the same sparse 3D activation tensor is also approximated by a collection of local features. These local features are then robustly matched to approximate the optimal alignment of the tensors. This happens without any network modification, additional layers or training. No local feature detection happens on the original image. No local feature descriptors and no visual vocabulary are needed throughout the whole process. We experimentally show that the proposed method achieves the state-of-the-art performance on standard benchmarks across different network architectures and different global pooling methods. The highest gain in performance is achieved when diffusion on the nearest-neighbor graph of global descriptors is initiated from spatially verified images."
                },
                {
                    "title": "D2-Net: A Trainable CNN for Joint Description and Detection of Local Features",
                    "abstract": "In this work we address the problem of finding reliable pixel-level correspondences under difficult imaging conditions. We propose an approach where a single convolutional neural network plays a dual role: It is simultaneously a dense feature descriptor and a feature detector. By postponing the detection to a later stage, the obtained keypoints are more stable than their traditional counterparts based on early detection of low-level structures. We show that this model can be trained using pixel correspondences extracted from readily available large-scale SfM reconstructions, without any further annotations. The proposed method obtains state-of-the-art performance on both the difficult Aachen Day-Night localization dataset and the InLoc indoor localization benchmark, as well as competitive performance on other benchmarks for image matching and 3D reconstruction."
                },
                {
                    "title": "Composing Text and Image for Image Retrieval - an Empirical Odyssey",
                    "abstract": "In this paper, we study the task of image retrieval, where the input query is specified in the form of an image plus some text that describes desired modifications to the input image. For example, we may present an image of the Eiffel tower, and ask the system to find images which are visually similar, but are modified in small ways, such as being taken at nighttime instead of during the day. o tackle this task, we embed the query (reference image plus modification text) and the target (images). The encoding function of the image text query learns a representation, such that the similarity with the target image representation is high iff it is a ``positive match''. We propose a new way to combine image and text through residual connection, that is designed for this retrieval task. We show this outperforms existing approaches on 3 different datasets, namely Fashion-200k, MIT-States and a new synthetic dataset we create based on CLEVR. We also show that our approach can be used to perform image classification with compositionally novel labels, and we outperform previous methods on MIT-States on this task."
                },
                {
                    "title": "Invariant Information Clustering for Unsupervised Image Classification and Segmentation",
                    "abstract": "We present a novel clustering objective that learns a neural network classifier from scratch, given only unlabelled data samples. The model discovers clusters that accurately match semantic classes, achieving state-of-the-art results in eight unsupervised clustering benchmarks spanning image classification and segmentation. These include STL10, an unsupervised variant of ImageNet, and CIFAR10, where we significantly beat the accuracy of our closest competitors by 6.6 and 9.5 absolute percentage points respectively. The method is not specialised to computer vision and operates on any paired dataset samples; in our experiments we use random transforms to obtain a pair from each image. The trained network directly outputs semantic labels, rather than high dimensional representations that need external processing to be usable for semantic clustering. The objective is simply to maximise mutual information between the class assignments of each pair. It is easy to implement and rigorously grounded in information theory, meaning we effortlessly avoid degenerate solutions that other clustering methods are susceptible to. In addition to the fully unsupervised mode, we also test two semi-supervised settings. The first achieves 88.8% accuracy on STL10 classification, setting a new global state-of-the-art over all existing methods (whether supervised, semi-supervised or unsupervised). The second shows robustness to 90% reductions in label coverage, of relevance to applications that wish to make use of small amounts of labels. github.com/xu-ji/IIC"
                },
                {
                    "title": "Local Descriptors Optimized for Average Precision",
                    "abstract": "Extraction of local feature descriptors is a vital stage in the solution pipelines for numerous computer vision tasks. Learning-based approaches improve performance in certain tasks, but still cannot replace handcrafted features in general. In this paper, we improve the learning of local feature descriptors by optimizing the performance of descriptor matching, which is a common stage that follows descriptor extraction in local feature based pipelines, and can be formulated as nearest neighbor retrieval. Specifically, we directly optimize a ranking-based retrieval performance metric, Average Precision, using deep neural networks. This general-purpose solution can also be viewed as a listwise learning to rank approach, which is advantageous compared to recent local ranking approaches. On standard benchmarks, descriptors learned with our formulation achieve state-of-the-art results in patch verification, patch retrieval, and image matching,"
                },
                {
                    "title": "SuperPoint: Self-Supervised Interest Point Detection and Description",
                    "abstract": "This paper presents a self-supervised framework for training interest point detectors and descriptors suitable for a large number of multiple-view geometry problems in computer vision. As opposed to patch-based neural networks, our fully-convolutional model operates on full-sized images and jointly computes pixel-level interest point locations and associated descriptors in one forward pass. We introduce Homographic Adaptation, a multi-scale, multi-homography approach for boosting interest point detection repeatability and performing cross-domain adaptation (e.g., synthetic-to-real). Our model, when trained on the MS-COCO generic image dataset using Homographic Adaptation, is able to repeatedly detect a much richer set of interest points than the initial pre-adapted deep model and any other traditional corner detector. The final system gives rise to state-of-the-art homography estimation results on HPatches when compared to LIFT, SIFT and ORB."
                },
                {
                    "title": "Large-Scale Image Retrieval with Attentive Deep Local Features",
                    "abstract": "We propose an attentive local feature descriptor suitable for large-scale image retrieval, referred to as DELE (DEep Local Feature). The new feature is based on convolutional neural networks, which are trained only with image-level annotations on a landmark image dataset. To identify semantically useful local features for image retrieval, we also propose an attention mechanism for key point selection, which shares most network layers with the descriptor. This frame-work can be used for image retrieval as a drop-in replacement for other keypoint detectors and descriptors, enabling more accurate feature matching and geometric verification. Our system produces reliable confidence scores to reject false positives\u2013in particular, it is robust against queries that have no correct match in the database. To evaluate the proposed descriptor, we introduce a new large-scale dataset, referred to as Google-Landmarks dataset, which involves challenges in both database and query such as background clutter, partial occlusion, multiple landmarks, objects in variable scales, etc. We show that DELE outperforms the state-of-the-art global and local descriptors in the large-scale setting by significant margins."
                },
                {
                    "title": "Particular object retrieval with integral max-pooling of CNN activations",
                    "abstract": "Recently, image representation built upon Convolutional Neural Network (CNN) has been shown to provide effective descriptors for image search, outperforming pre-CNN features as short-vector representations. Yet such models are not compatible with geometry-aware re-ranking methods and still outperformed, on some particular object retrieval benchmarks, by traditional image search systems relying on precise descriptor matching, geometric re-ranking, or query expansion. This work revisits both retrieval stages, namely initial search and re-ranking, by employing the same primitive information derived from the CNN. We build compact feature vectors that encode several image regions without the need to feed multiple inputs to the network. Furthermore, we extend integral images to handle max-pooling on convolutional layer activations, allowing us to efficiently localize matching objects. The resulting bounding box is finally used for image re-ranking. As a result, this paper significantly improves existing CNN-based recognition pipeline: We report for the first time results competing with traditional methods on the challenging Oxford5k and Paris6k datasets."
                },
                {
                    "title": "Learning Deep Features for Scene Recognition using Places Database",
                    "abstract": "Scene recognition is one of the hallmark tasks of computer vision, allowing definition of a context for object recognition. Whereas the tremendous recent progress in object recognition tasks is due to the availability of large datasets like ImageNet and the rise of Convolutional Neural Networks (CNNs) for learning high-level features, performance at scene recognition has not attained the same level of success. This may be because current deep features trained from ImageNet are not competitive enough for such tasks. Here, we introduce a new scene-centric database called Places with over 7 million labeled pictures of scenes. We propose new methods to compare the density and diversity of image datasets and show that Places is as dense as other scene datasets and has more diversity. Using CNN, we learn deep features for scene recognition tasks, and establish new state-of-the-art results on several scene-centric datasets. A visualization of the CNN layers' responses allows us to show differences in the internal representations of object-centric and scene-centric networks."
                },
                {
                    "title": "Clustering: probably approximately useless?",
                    "abstract": "Clustering never seems to live up to the hype. To paraphrase the popular saying, clustering looks good in theory, yet often fails to deliver in practice. Why? You would think that something so simple and elegant as finding groups of similar items in data would be incredibly useful. Yet often it isn't. The problem is that clustering rarely finds the groups you want, or expected, or that are most useful for the task at hand. There are so many good ways to cluster a dataset that the odds of coming up with the clustering that is best for what you are doing now are small. How do we fix this and make clustering more useful in practice? How do we make clustering do what you want, while still giving it the freedom to \"do its own thing\" and surprise us?"
                },
                {
                    "title": "Human action recognition by learning bases of action attributes and parts",
                    "abstract": "In this work, we propose to use attributes and parts for recognizing human actions in still images. We define action attributes as the verbs that describe the properties of human actions, while the parts of actions are objects and poselets that are closely related to the actions. We jointly model the attributes and parts by learning a set of sparse bases that are shown to carry much semantic meaning. Then, the attributes and parts of an action image can be reconstructed from sparse coefficients with respect to the learned bases. This dual sparsity provides theoretical guarantee of our bases learning and feature reconstruction approach. On the PASCAL action dataset and a new \u201cStanford 40 Actions\u201d dataset, we show that our method extracts meaningful high-order interactions between attributes and parts in human actions while achieving state-of-the-art classification performance."
                },
                {
                    "title": "Grouplet: A structured image representation for recognizing human and object interactions",
                    "abstract": "Psychologists have proposed that many human-object interaction activities form unique classes of scenes. Recognizing these scenes is important for many social functions. To enable a computer to do this is however a challenging task. Take people-playing-musical-instrument (PPMI) as an example; to distinguish a person playing violin from a person just holding a violin requires subtle distinction of characteristic image features and feature arrangements that differentiate these two scenes. Most of the existing image representation methods are either too coarse (e.g. BoW) or too sparse (e.g. constellation models) for performing this task. In this paper, we propose a new image feature representation called \u201cgrouplet\u201d. The grouplet captures the structured information of an image by encoding a number of discriminative visual features and their spatial configurations. Using a dataset of 7 different PPMI activities, we show that grouplets are more effective in classifying and detecting human-object interactions than other state-of-the-art methods. In particular, our method can make a robust distinction between humans playing the instruments and humans co-occurring with the instruments without playing."
                },
                {
                    "title": "Locality-constrained Linear Coding for image classification",
                    "abstract": "The traditional SPM approach based on bag-of-features (BoF) requires nonlinear classifiers to achieve good image classification performance. This paper presents a simple but effective coding scheme called Locality-constrained Linear Coding (LLC) in place of the VQ coding in traditional SPM. LLC utilizes the locality constraints to project each descriptor into its local-coordinate system, and the projected coordinates are integrated by max pooling to generate the final representation. With linear classifier, the proposed approach performs remarkably better than the traditional nonlinear SPM, achieving state-of-the-art performance on several benchmarks. Compared with the sparse coding strategy [22], the objective function used by LLC has an analytical solution. In addition, the paper proposes a fast approximated LLC method by first performing a K-nearest-neighbor search and then solving a constrained least square fitting problem, bearing computational complexity of O(M + K2). Hence even with very large codebooks, our system can still process multiple frames per second. This efficiency significantly adds to the practical values of LLC for real applications."
                },
                {
                    "title": "Clustering: Science or Art?",
                    "abstract": "We examine whether the quality of different clustering algorithms can be compared by a general, scientifically sound procedure which is independent of particular clustering algorithms. We argue that the major obstacle is the difficulty in evaluating a clustering algorithm without taking into account the context: why does the user cluster his data in the first place, and what does he want to do with the clustering afterwards? We argue that clustering should not be treated as an application-independent mathematical problem, but should always be studied in the context of its end-use. Different techniques to evaluate clustering algorithms have to be developed for different uses of clustering. To simplify this procedure we argue that it will be useful to build a \"taxonomy of clustering problems\" to identify clustering applications which can be treated in a unified way and that such an effort will be more fruitful than attempting the impossible--developing \"optimal\" domain-independent clustering algorithms or even classifying clustering algorithms in terms of how they work."
                },
                {
                    "title": "Non-redundant Multi-view Clustering via Orthogonalization",
                    "abstract": "Typical clustering algorithms output a single clustering of the data. However, in real world applications, data can often be interpreted in many different ways; data can have different groupings that are reasonable and interesting from different perspectives. This is especially true for high-dimensional data, where different feature subspaces may reveal different structures of the data. Why commit to one clustering solution while all these alternative clustering views might be interesting to the user. In this paper, we propose a new clustering paradigm for explorative data analysis: find all non-redundant clustering views of the data, where data points of one cluster can belong to different clusters in other views. We present a framework to solve this problem and suggest two approaches within this framework: (1) orthogonal clustering, and (2) clustering in orthogonal subspaces. In essence, both approaches find alternative ways to partition the data by projecting it to a space that is orthogonal to our current solution. The first approach seeks orthogonality in the cluster space, while the second approach seeks orthogonality in the feature space. We test our framework on both synthetic and high-dimensional benchmark data sets, and the results show that indeed our approaches were able to discover varied solutions that are interesting and meaningful."
                },
                {
                    "title": "LLAMA: automatic hypertext generation utilizing language models",
                    "abstract": "Manual hypertext construction is labour intensive and prone to error. Robust systems capable of automatic hypertext generation (AHG) could be of direct benefit to those individuals responsible for hypertext authoring. In this paper we propose a novel technique for the autonomous creation of hypertext which is dependent upon language models. This work is strongly influenced by those algorithms which process the hyperlinked structure of a corpus in an attempt to find authoritative sources. The algorithm was evaluated by experimental comparison with human hypertext authors, and we found that both approaches produced broadly similar results."
                },
                {
                    "title": "Meta Clustering",
                    "abstract": "Clustering is ill-defined. Unlike supervised learning where labels lead to crisp performance criteria such as accuracy and squared error, clustering quality depends on how the clusters will be used. Devising clustering criteria that capture what users need is difficult. Most clustering algorithms search for optimal clusterings based on a pre-specified clustering criterion. Our approach differs. We search for many alternate clusterings of the data, and then allow users to select the clustering(s) that best fit their needs. Meta clustering first finds a variety of clusterings and then clusters this diverse set of clusterings so that users must only examine a small number of qualitatively different clusterings. We present methods for automatically generating a diverse set of alternate clusterings, as well as methods for grouping clusterings into meta clusters. We evaluate meta clustering on four test problems and two case studies. Surprisingly, clusterings that would be of most interest to users often are not very compact clusterings."
                },
                {
                    "title": "PhotoTOC: automatic clustering for browsing personal photographs",
                    "abstract": "This paper presents photo table of contents (PhotoTOC), a system that helps users find digital photographs in their own collection of photographs. PhotoTOC is a browsing user interface that uses an overview+detail design. The detail view is a temporally ordered list of all of the user's photographs. The overview of the user's collection is automatically generated by an image clustering algorithm, which clusters on the creation time and the color of the photographs. PhotoTOC was tested on users' own photographs against three other browsers. Searching for images with PhotoTOC was subjectively rated easier than all of the other browsers. This result shows that automatic organization of personal photographs facilitates efficient and satisfying search."
                },
                {
                    "title": "No free lunch theorems for optimization",
                    "abstract": "A framework is developed to explore the connection between effective optimization algorithms and the problems they are solving. A number of \"no free lunch\" (NFL) theorems are presented which establish that for any algorithm, any elevated performance over one class of problems is offset by performance over another class. These theorems result in a geometric interpretation of what it means for an algorithm to be well suited to an optimization problem. Applications of the NFL theorems to information-theoretic aspects of optimization and benchmark measures of performance are also presented. Other issues addressed include time-varying optimization problems and a priori \"head-to-head\" minimax distinctions between optimization algorithms, distinctions that result despite the NFL theorems' enforcing of a type of uniformity over all algorithms."
                },
                {
                    "title": "R2D2: Reliable and Repeatable Detector and Descriptor",
                    "abstract": "Interest point detection and local feature description are fundamental steps in many computer vision applications. Classical approaches are based on a detect-then-describe paradigm where separate handcrafted methods are used to first identify repeatable keypoints and then represent them with a local descriptor. Neural networks trained with metric learning losses have recently caught up with these techniques, focusing on learning repeatable saliency maps for keypoint detection or learning descriptors at the detected keypoint locations. In this work, we argue that repeatable regions are not necessarily discriminative and can therefore lead to select suboptimal keypoints. Furthermore, we claim that descriptors should be learned only in regions for which matching can be performed with high confidence. We thus propose to jointly learn keypoint detection and description together with a predictor of the local descriptor discriminativeness. This allows to avoid ambiguous areas, thus leading to reliable keypoint detection and description. Our detection-and-description approach simultaneously outputs sparse, repeatable and reliable keypoints that outperforms state-of-the-art detectors and descriptors on the HPatches dataset and on the recent Aachen Day-Night localization benchmark."
                },
                {
                    "title": "Learning local feature descriptors with triplets and shallow convolutional neural networks",
                    "abstract": "It has recently been demonstrated that local feature descriptors based on convolutional neural networks (CNN) can significantly improve the matching performance. Previous work on learning such descriptors has focused on exploiting pairs of positive and negative patches to learn discriminative CNN representations. In this work, we propose to utilize triplets of training samples, together with in-triplet mining of hard negatives. We show that our method achieves state of the art results, without the computational overhead typically associated with mining of negatives and with lower complexity of the network architecture. We compare our approach to recently introduced convolutional local feature descriptors, and demonstrate the advantages of the proposed methods in terms of performance and speed. We also examine different loss functions associated with triplets."
                },
                {
                    "title": "Neural Codes for Image Retrieval",
                    "abstract": "This seminar report focuses on using convolutional neural networks for image retrieval. Firstly, we give a thorough discussion of several state-of-the-art techniques in image retrieval by considering the associated subproblems: image description, descriptor compression, nearest-neighbor search and query expansion. We discuss both the aggregation of local descriptors using clustering and metric learning techniques as well as global descriptors. Subsequently, we briefly introduce the basic concepts of deep convolutional neural networks, focusing on the architecture proposed by Krizhevsky et al. [KSH12]. We discuss different types of layers commonly used in recent architectures, for example convolutional layers, non-linearity and rectification layers, pooling layers as well as local contrast normalization layers. Finally, we shortly review supervised training techniques based on stochastic gradient descent and regularization techniques such as dropout and weight decay. Finally, following Babenko et al. [BSCL14], we discuss the use of feature activations in intermediate layers as image representation for image retrieval. After presenting experiments and comparing convolutional neural networks for image retrieval with other state-of-the-art techniques, we conclude by motivating the combined use of deep architectures and hand-crafted image representations for accurate and efficient image retrieval."
                },
                {
                    "title": "An Impossibility Theorem for Clustering",
                    "abstract": "Although the study of clustering is centered around an intuitively compelling goal, it has been very difficult to develop a unified framework for reasoning about it at a technical level, and profoundly diverse approaches to clustering abound in the research community. Here we suggest a formal perspective on the difficulty in finding such a unification, in the form of an impossibility theorem: for a set of three simple properties, we show that there is no clustering function satisfying all three. Relaxations of these properties expose some of the interesting (and unavoidable) trade-offs at work in well-studied clustering techniques such as single-linkage, sum-of-pairs, k-means, and k-median."
                },
                {
                    "title": "Discriminant-EM algorithm with application to image retrieval",
                    "abstract": "In many vision applications, the practice of supervised learning faces several difficulties, one of which is that insufficient labeled training data result in poor generalization. In image retrieval, we have very few labeled images from query and relevance feedback so that it is hard to automatically weight image features and select similarity metrics for image classification. This paper investigates the possibility of including an unlabeled data set to make up the insufficiency of labeled data. Different from most current research in image retrieval, the proposed approach tries to cast image retrieval as a transductive learning problem, in which the generalization of an image classifier is only defined on a set of images such as the given image database. Formulating this transductive problem in a probabilistic framework the proposed algorithm, Discriminant EM (D-EM) not only estimates the parameters of a generative model but also finds a linear transformation to relax the assumption of probabilistic structure of data distributions as well as select good features automatically. Our experiments show that D-EM has a satisfactory performance in image retrieval applications. D-EM algorithm has the potential to many other applications."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a methodology for image clustering that allows users to specify diverse clustering criteria in natural language, enabling multiple clustering outcomes from the same dataset?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is significant for the research community as it addresses the limitations of traditional unsupervised clustering methods, which do not allow for user-defined criteria. By enabling user-directed clustering, this approach could lead to more relevant and interpretable clustering results, enhancing applications in image retrieval, organization, and analysis. This could also inspire future research into more interactive and user-centric machine learning models, ultimately advancing knowledge in both unsupervised learning and human-computer interaction.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the complexity of accurately interpreting natural language instructions in the context of image data, as well as the need to design a flexible clustering algorithm that can adapt to various user-specified criteria. Naive approaches may fail due to the inherent ambiguity in natural language and the difficulty in mapping these instructions to effective clustering strategies. Additionally, technical obstacles such as ensuring the robustness of the clustering results across different criteria and maintaining computational efficiency must be addressed.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on unsupervised clustering methods that rely on fixed inductive biases and do not incorporate user input. The lack of integration between natural language processing and image clustering has been a significant barrier. Existing solutions have not provided a mechanism for users to influence clustering outcomes directly. Our approach differs by leveraging foundation models to interpret user-specified criteria, allowing for a more dynamic and user-driven clustering process that enhances flexibility and interpretability.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Image Clustering Conditioned on Text Criteria (IC||||TC), utilizes foundation models to interpret natural language instructions for clustering images. We will use datasets such as the Stanford 40 Action dataset and the PPMI dataset, evaluating the clustering results based on user-defined text criteria. The metric for success will include the relevance and interpretability of the clusters, as well as user satisfaction with the clustering outcomes. We expect that IC||||TC will produce diverse and meaningful clusters that align with user specifications, allowing for iterative refinement of results based on user feedback."
            }
        },
        "author_data": {
            "d2c8b439-565d-45d7-bc81-8ddea818077c": {
                "pk": "d2c8b439-565d-45d7-bc81-8ddea818077c",
                "name": "Sehyun Kwon",
                "collaborators": [
                    "Ernest K. Ryu",
                    "Jaeyeon Kim",
                    "Joo Young Choi",
                    "Jongho Park",
                    "Jaewoong Cho",
                    "Jason D. Lee",
                    "Albert No",
                    "Taeho Yoon"
                ],
                "domain": [
                    "In-context Learning",
                    "Generative Adversarial Networks",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Task Diversity Shortens the ICL Plateau",
                        "abstract": "In-context learning (ICL) describes a language model's ability to generate outputs based on a set of input demonstrations and a subsequent query. To understand this remarkable capability, researchers have studied simplified, stylized models. These studies have consistently observed long loss plateaus, during which models exhibit minimal improvement, followed by a sudden, rapid surge of learning. In this work, we reveal that training on multiple diverse ICL tasks simultaneously shortens the loss plateaus, making each task easier to learn. This finding is surprising as it contradicts the natural intuition that the combined complexity of multiple ICL tasks would lengthen the learning process, not shorten it. Our result suggests that the recent success in large-scale training of language models may be attributed not only to the richness of the data at scale but also to the easier optimization (training) induced by the diversity of natural language training data."
                    },
                    {
                        "title": "WGAN with an Infinitely Wide Generator Has No Spurious Stationary Points",
                        "abstract": "Generative adversarial networks (GAN) are a widely used class of deep generative models, but their minimax training dynamics are not understood very well. In this work, we show that GANs with a 2-layer infinite-width generator and a 2-layer finite-width discriminator trained with stochastic gradient ascent-descent have no spurious stationary points. We then show that when the width of the generator is finite but wide, there are no spurious stationary points within a ball whose radius becomes arbitrarily large (to cover the entire parameter space) as the width goes to infinity."
                    }
                ]
            },
            "d6251fec-e565-455f-a2be-54d361d805b7": {
                "pk": "d6251fec-e565-455f-a2be-54d361d805b7",
                "name": "Jaeseung Park",
                "collaborators": [
                    "Jongho Park",
                    "Zheyang Xiong",
                    "Nayoung Lee",
                    "Jaewoong Cho",
                    "Samet Oymak",
                    "Kangwook Lee",
                    "Dimitris Papailiopoulos"
                ],
                "domain": [
                    "State-Space Models",
                    "Language Modeling",
                    "Hybrid Architectures",
                    "In-Context Learning"
                ],
                "publications": [
                    {
                        "title": "Can Mamba Learn How to Learn? A Comparative Study on In-Context Learning Tasks",
                        "abstract": "State-space models (SSMs), such as Mamba (Gu&Dao, 2023), have been proposed as alternatives to Transformer networks in language modeling, by incorporating gating, convolutions, and input-dependent token selection to mitigate the quadratic cost of multi-head attention. Although SSMs exhibit competitive performance, their in-context learning (ICL) capabilities, a remarkable emergent property of modern language models that enables task execution without parameter optimization, remain underexplored compared to Transformers. In this study, we evaluate the ICL performance of SSMs, focusing on Mamba, against Transformer models across various tasks. Our results show that SSMs perform comparably to Transformers in standard regression ICL tasks, while outperforming them in tasks like sparse parity learning. However, SSMs fall short in tasks involving non-standard retrieval functionality. To address these limitations, we introduce a hybrid model, MambaFormer, that combines Mamba with attention blocks, surpassing individual models in tasks where they struggle independently. Our findings suggest that hybrid architectures offer promising avenues for enhancing ICL in language models."
                    }
                ]
            },
            "3fe8b240-4402-4c56-90cf-f53b12f4a3ab": {
                "pk": "3fe8b240-4402-4c56-90cf-f53b12f4a3ab",
                "name": "Minkyu Kim",
                "collaborators": [
                    "Jinwoo Shin",
                    "Sangwoo Mo",
                    "Kyungmin Lee",
                    "Dongmin Park",
                    "Sebin Kim",
                    "Taehong Moon",
                    "Kangwook Lee",
                    "Jaewoong Cho",
                    "Younghyun Kim",
                    "Jaeho Lee"
                ],
                "domain": [
                    "Computer Vision",
                    "Natural Language Processing",
                    "Robustness",
                    "Semi-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Rare-to-Frequent: Unlocking Compositional Generation Power of Diffusion Models on Rare Concepts with LLM Guidance",
                        "abstract": "State-of-the-art text-to-image (T2I) diffusion models often struggle to generate rare compositions of concepts, e.g., objects with unusual attributes. In this paper, we show that the compositional generation power of diffusion models on such rare concepts can be significantly enhanced by the Large Language Model (LLM) guidance. We start with empirical and theoretical analysis, demonstrating that exposing frequent concepts relevant to the target rare concepts during the diffusion sampling process yields more accurate concept composition. Based on this, we propose a training-free approach, R2F, that plans and executes the overall rare-to-frequent concept guidance throughout the diffusion inference by leveraging the abundant semantic knowledge in LLMs. Our framework is flexible across any pre-trained diffusion models and LLMs, and can be seamlessly integrated with the region-guided diffusion approaches. Extensive experiments on three datasets, including our newly proposed benchmark, RareBench, containing various prompts with rare compositions of concepts, R2F significantly surpasses existing models including SD3.0 and FLUX by up to 28.1%p in T2I alignment. Code is available at https://github.com/krafton-ai/Rare2Frequent."
                    },
                    {
                        "title": "Discovering and Mitigating Visual Biases Through Keyword Explanation",
                        "abstract": "Addressing biases in computer vision models is crucial for real-world AI deployments. However, mitigating visual biases is challenging due to their unexplainable nature, often identified indirectly through visualization or sample statistics, which necessitates additional human supervision for interpretation. To tackle this issue, we propose the Bias-to-Text (B2T) framework, which interprets visual biases as keywords. Specifically, we extract common keywords from the captions of mispredicted images to identify potential biases in the model. We then validate these keywords by measuring their similarity to the mispredicted images using a vision-language scoring model. The keyword explanation form of visual bias offers several advantages, such as a clear group naming for bias discovery and a natural extension for debiasing using these group names. Our experiments demonstrate that B2T can identify known biases, such as gender bias in CelebA, background bias in Waterbirds, and distribution shifts in ImageNet-R/C. Additionally, B2T uncovers novel biases in larger datasets, such as Dollar Street and ImageNet. For example, we discovered a contextual bias between \u201cbee\u201d and \u201cflower\u201d in ImageNet. We also highlight various applications of B2T keywords, including debiased training, CLIP prompting, and model comparison. 11Code: https://github.com/alinlab/b2t"
                    },
                    {
                        "title": "S-CLIP: Semi-supervised Vision-Language Learning using Few Specialist Captions",
                        "abstract": "Vision-language models, such as contrastive language-image pre-training (CLIP), have demonstrated impressive results in natural image domains. However, these models often struggle when applied to specialized domains like remote sensing, and adapting to such domains is challenging due to the limited number of image-text pairs available for training. To address this, we propose S-CLIP, a semi-supervised learning method for training CLIP that utilizes additional unpaired images. S-CLIP employs two pseudo-labeling strategies specifically designed for contrastive learning and the language modality. The caption-level pseudo-label is given by a combination of captions of paired images, obtained by solving an optimal transport problem between unpaired and paired images. The keyword-level pseudo-label is given by a keyword in the caption of the nearest paired image, trained through partial label learning that assumes a candidate set of labels for supervision instead of the exact one. By combining these objectives, S-CLIP significantly enhances the training of CLIP using only a few image-text pairs, as demonstrated in various specialist domains, including remote sensing, fashion, scientific figures, and comics. For instance, S-CLIP improves CLIP by 10% for zero-shot classification and 4% for image-text retrieval on the remote sensing benchmark, matching the performance of supervised CLIP while using three times fewer image-text pairs."
                    },
                    {
                        "title": "SmoothMix: Training Confidence-calibrated Smoothed Classifiers for Certified Robustness",
                        "abstract": "Randomized smoothing is currently a state-of-the-art method to construct a certifiably robust classifier from neural networks against $\\ell_2$-adversarial perturbations. Under the paradigm, the robustness of a classifier is aligned with the prediction confidence, i.e., the higher confidence from a smoothed classifier implies the better robustness. This motivates us to rethink the fundamental trade-off between accuracy and robustness in terms of calibrating confidences of a smoothed classifier. In this paper, we propose a simple training scheme, coined SmoothMix, to control the robustness of smoothed classifiers via self-mixup: it trains on convex combinations of samples along the direction of adversarial perturbation for each input. The proposed procedure effectively identifies over-confident, near off-class samples as a cause of limited robustness in case of smoothed classifiers, and offers an intuitive way to adaptively set a new decision boundary between these samples for better robustness. Our experimental results demonstrate that the proposed method can significantly improve the certified $\\ell_2$-robustness of smoothed classifiers compared to existing state-of-the-art robust training methods."
                    }
                ]
            },
            "88983c5b-955f-46c5-a6ca-3077a18585f0": {
                "pk": "88983c5b-955f-46c5-a6ca-3077a18585f0",
                "name": "Jaewoong Cho",
                "collaborators": [
                    "Jongho Park",
                    "Jaeyeon Kim",
                    "Sehyun Kwon",
                    "Joo Young Choi",
                    "Jason D. Lee",
                    "Ernest K. Ryu",
                    "Jaeseung Park",
                    "Zheyang Xiong",
                    "Nayoung Lee",
                    "Samet Oymak"
                ],
                "domain": [
                    "In-context Learning",
                    "Language Models",
                    "State-space Models",
                    "Hybrid Architectures"
                ],
                "publications": [
                    {
                        "title": "Task Diversity Shortens the ICL Plateau",
                        "abstract": "In-context learning (ICL) describes a language model's ability to generate outputs based on a set of input demonstrations and a subsequent query. To understand this remarkable capability, researchers have studied simplified, stylized models. These studies have consistently observed long loss plateaus, during which models exhibit minimal improvement, followed by a sudden, rapid surge of learning. In this work, we reveal that training on multiple diverse ICL tasks simultaneously shortens the loss plateaus, making each task easier to learn. This finding is surprising as it contradicts the natural intuition that the combined complexity of multiple ICL tasks would lengthen the learning process, not shorten it. Our result suggests that the recent success in large-scale training of language models may be attributed not only to the richness of the data at scale but also to the easier optimization (training) induced by the diversity of natural language training data."
                    },
                    {
                        "title": "Can Mamba Learn How to Learn? A Comparative Study on In-Context Learning Tasks",
                        "abstract": "State-space models (SSMs), such as Mamba (Gu&Dao, 2023), have been proposed as alternatives to Transformer networks in language modeling, by incorporating gating, convolutions, and input-dependent token selection to mitigate the quadratic cost of multi-head attention. Although SSMs exhibit competitive performance, their in-context learning (ICL) capabilities, a remarkable emergent property of modern language models that enables task execution without parameter optimization, remain underexplored compared to Transformers. In this study, we evaluate the ICL performance of SSMs, focusing on Mamba, against Transformer models across various tasks. Our results show that SSMs perform comparably to Transformers in standard regression ICL tasks, while outperforming them in tasks like sparse parity learning. However, SSMs fall short in tasks involving non-standard retrieval functionality. To address these limitations, we introduce a hybrid model, MambaFormer, that combines Mamba with attention blocks, surpassing individual models in tasks where they struggle independently. Our findings suggest that hybrid architectures offer promising avenues for enhancing ICL in language models."
                    }
                ]
            },
            "b5fef4e6-25ff-4797-82da-ff64df737b74": {
                "pk": "b5fef4e6-25ff-4797-82da-ff64df737b74",
                "name": "Ernest K. Ryu",
                "collaborators": [
                    "Jason D. Lee",
                    "Joo Young Choi",
                    "Jaewoong Cho",
                    "Uijeong Jang",
                    "Jaeyeon Kim",
                    "Sehyun Kwon",
                    "Jongho Park",
                    "Jaesung R. Park",
                    "Inkyu Park",
                    "Albert No"
                ],
                "domain": [
                    "Machine Learning",
                    "Natural Language Processing",
                    "Diffusion Models",
                    "Parameter-Efficient Fine-Tuning"
                ],
                "publications": [
                    {
                        "title": "LoRA Training in the NTK Regime has No Spurious Local Minima",
                        "abstract": "Low-rank adaptation (LoRA) has become the standard approach for parameter-efficient fine-tuning of large language models (LLM), but our theoretical understanding of LoRA has been limited. In this work, we theoretically analyze LoRA fine-tuning in the neural tangent kernel (NTK) regime with $N$ data points, showing: (i) full fine-tuning (without LoRA) admits a low-rank solution of rank $r\\lesssim \\sqrt{N}$; (ii) using LoRA with rank $r\\gtrsim \\sqrt{N}$ eliminates spurious local minima, allowing gradient descent to find the low-rank solutions; (iii) the low-rank solution found using LoRA generalizes well."
                    },
                    {
                        "title": "Task Diversity Shortens the ICL Plateau",
                        "abstract": "In-context learning (ICL) describes a language model's ability to generate outputs based on a set of input demonstrations and a subsequent query. To understand this remarkable capability, researchers have studied simplified, stylized models. These studies have consistently observed long loss plateaus, during which models exhibit minimal improvement, followed by a sudden, rapid surge of learning. In this work, we reveal that training on multiple diverse ICL tasks simultaneously shortens the loss plateaus, making each task easier to learn. This finding is surprising as it contradicts the natural intuition that the combined complexity of multiple ICL tasks would lengthen the learning process, not shorten it. Our result suggests that the recent success in large-scale training of language models may be attributed not only to the richness of the data at scale but also to the easier optimization (training) induced by the diversity of natural language training data."
                    },
                    {
                        "title": "Simple Drop-in LoRA Conditioning on Attention Layers Will Improve Your Diffusion Model",
                        "abstract": "Current state-of-the-art diffusion models employ U-Net architectures containing convolutional and (qkv) self-attention layers. The U-Net processes images while being conditioned on the time embedding input for each sampling step and the class or caption embedding input corresponding to the desired conditional generation. Such conditioning involves scale-and-shift operations to the convolutional layers but does not directly affect the attention layers. While these standard architectural choices are certainly effective, not conditioning the attention layers feels arbitrary and potentially suboptimal. In this work, we show that simply adding LoRA conditioning to the attention layers without changing or tuning the other parts of the U-Net architecture improves the image generation quality. For example, a drop-in addition of LoRA conditioning to EDM diffusion model yields FID scores of 1.91/1.75 for unconditional and class-conditional CIFAR-10 generation, improving upon the baseline of 1.97/1.79."
                    }
                ]
            },
            "1a110182-8fcf-40ee-a290-cb71742a28bb": {
                "pk": "1a110182-8fcf-40ee-a290-cb71742a28bb",
                "name": "Kangwook Lee",
                "collaborators": [
                    "Zheyang Xiong",
                    "Dimitris Papailiopoulos",
                    "Vasileios Papageorgiou",
                    "Liu Yang",
                    "Samet Oymak",
                    "Ziyang Cai",
                    "John Cooper",
                    "Albert Ge",
                    "Zack Sifakis",
                    "Angeliki Giannou"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "In-Context Learning",
                    "Hybrid Models"
                ],
                "publications": [
                    {
                        "title": "Everything Everywhere All at Once: LLMs can In-Context Learn Multiple Tasks in Superposition",
                        "abstract": "Large Language Models (LLMs) have demonstrated remarkable in-context learning (ICL) capabilities. In this study, we explore a surprising phenomenon related to ICL: LLMs can perform multiple, computationally distinct ICL tasks simultaneously, during a single inference call, a capability we term\"task superposition\". We provide empirical evidence of this phenomenon across various LLM families and scales and show that this phenomenon emerges even if we train the model to in-context learn one task at a time. We offer theoretical explanations that this capability is well within the expressive power of transformers. We also explore how LLMs internally compose task vectors during superposition. Furthermore, we show that larger models can solve more ICL tasks in parallel, and better calibrate their output distribution. Our findings offer insights into the latent capabilities of LLMs, further substantiate the perspective of\"LLMs as superposition of simulators\", and raise questions about the mechanisms enabling simultaneous task execution."
                    },
                    {
                        "title": "Can Mamba Learn How to Learn? A Comparative Study on In-Context Learning Tasks",
                        "abstract": "State-space models (SSMs), such as Mamba (Gu&Dao, 2023), have been proposed as alternatives to Transformer networks in language modeling, by incorporating gating, convolutions, and input-dependent token selection to mitigate the quadratic cost of multi-head attention. Although SSMs exhibit competitive performance, their in-context learning (ICL) capabilities, a remarkable emergent property of modern language models that enables task execution without parameter optimization, remain underexplored compared to Transformers. In this study, we evaluate the ICL performance of SSMs, focusing on Mamba, against Transformer models across various tasks. Our results show that SSMs perform comparably to Transformers in standard regression ICL tasks, while outperforming them in tasks like sparse parity learning. However, SSMs fall short in tasks involving non-standard retrieval functionality. To address these limitations, we introduce a hybrid model, MambaFormer, that combines Mamba with attention blocks, surpassing individual models in tasks where they struggle independently. Our findings suggest that hybrid architectures offer promising avenues for enhancing ICL in language models."
                    },
                    {
                        "title": "From Artificial Needles to Real Haystacks: Improving Retrieval Capabilities in LLMs by Finetuning on Synthetic Data",
                        "abstract": "Recent studies have shown that Large Language Models (LLMs) struggle to accurately retrieve information and maintain reasoning capabilities when processing long-context inputs. To address these limitations, we propose a finetuning approach utilizing a carefully designed synthetic dataset comprising numerical key-value retrieval tasks. Our experiments on models like GPT-3.5 Turbo and Mistral 7B demonstrate that finetuning LLMs on this dataset significantly improves LLMs' information retrieval and reasoning capabilities in longer-context settings. We present an analysis of the finetuned models, illustrating the transfer of skills from synthetic to real task evaluations (e.g., $10.5\\%$ improvement on $20$ documents MDQA at position $10$ for GPT-3.5 Turbo). We also find that finetuned LLMs' performance on general benchmarks remains almost constant while LLMs finetuned on other baseline long-context augmentation data can encourage hallucination (e.g., on TriviaQA, Mistral 7B finetuned on our synthetic data cause no performance drop while other baseline data can cause a drop that ranges from $2.33\\%$ to $6.19\\%$). Our study highlights the potential of finetuning on synthetic data for improving the performance of LLMs on longer-context tasks."
                    },
                    {
                        "title": "Looped Transformers are Better at Learning Learning Algorithms",
                        "abstract": "Transformers have demonstrated effectiveness in in-context solving data-fitting problems from various (latent) models, as reported by Garg et al. However, the absence of an inherent iterative structure in the transformer architecture presents a challenge in emulating the iterative algorithms, which are commonly employed in traditional machine learning methods. To address this, we propose the utilization of looped transformer architecture and its associated training methodology, with the aim of incorporating iterative characteristics into the transformer architectures. Experimental results suggest that the looped transformer achieves performance comparable to the standard transformer in solving various data-fitting problems, while utilizing less than 10% of the parameter count."
                    }
                ]
            }
        }
    },
    "2408.07307": {
        "paper_data": {
            "title": "Nonlocal Attention Operator: Materializing Hidden Knowledge Towards Interpretable Physics Discovery",
            "url": "http://arxiv.org/abs/2408.07307v1",
            "arxiv_id": "2408.07307",
            "authors": [
                "Yue Yu",
                "Ning Liu",
                "Fei Lu",
                "Tian Gao",
                "Siavash Jafarzadeh",
                "Stewart Silling"
            ],
            "abstract": "Despite the recent popularity of attention-based neural architectures in core AI fields like natural language processing (NLP) and computer vision (CV), their potential in modeling complex physical systems remains under-explored. Learning problems in physical systems are often characterized as discovering operators that map between function spaces based on a few instances of function pairs. This task frequently presents a severely ill-posed PDE inverse problem. In this work, we propose a novel neural operator architecture based on the attention mechanism, which we coin Nonlocal Attention Operator (NAO), and explore its capability towards developing a foundation physical model. In particular, we show that the attention mechanism is equivalent to a double integral operator that enables nonlocal interactions among spatial tokens, with a data-dependent kernel characterizing the inverse mapping from data to the hidden parameter field of the underlying operator. As such, the attention mechanism extracts global prior information from training data generated by multiple systems, and suggests the exploratory space in the form of a nonlinear kernel map. Consequently, NAO can address ill-posedness and rank deficiency in inverse PDE problems by encoding regularization and achieving generalizability. We empirically demonstrate the advantages of NAO over baseline neural models in terms of generalizability to unseen data resolutions and system states. Our work not only suggests a novel neural operator architecture for learning interpretable foundation models of physical systems, but also offers a new perspective towards understanding the attention mechanism.",
            "introduction": "   1 Introduction  The interpretability of machine learning (ML) models has become increasingly important from the security and robustness standpoints (Rudin et\u00a0al., 2022; Molnar, 2020). This is particularly true in physics modeling problems that can affect human lives, where not only the accuracy but also the transparency of data-driven models are essential in making decisions (Coorey et\u00a0al., 2022; Ferrari and Willcox, 2024). Nevertheless, it remains challenging to discover the underlying physical system and the governing mechanism from data. Taking the material modeling task for instance, given that only the deformation field is observable, the goal of discovering the underlying material parameter field and mechanism presents an ill-posed unsupervised learning task. That means, even if an ML model can serve as a good surrogate to predict the corresponding loading field from a given deformation field, its inference of the material parameters can still drastically deteriorate.   Figure 1: Illustration of NAO\u2019s architecture.   To discover an interpretable mechanism for physical systems, a major challenge is to infer the governing laws of these systems that are often high- or infinite-dimensional, from data that are comprised of discrete measurements of continuous functions. Therefore, a data-driven surrogate model needs to learn not only the mapping between input and output function pairs, but also the mapping from given function pairs to the hidden state. From the PDE-based modeling standpoint, learning a surrogate model corresponds to a forward problem, whereas inferring the underlying mechanism is an inverse problem. The latter is generally an enduring ill-posed problem, especially when the measurements are scarce. Unfortunately, such an ill-posedness issue may become even more severe in neural network models, due to the inherent bias of neural network approximations (Xu et\u00a0al., 2019). To tackle this challenge, many deep learning methods have recently been proposed as inverse PDE solvers (Fan and Ying, 2023; Molinaro et\u00a0al., 2023; Jiang et\u00a0al., 2022; Chen et\u00a0al., 2023). The central idea is to incorporate prior information into the learning scheme, in the form of governing PDEs (Yang et\u00a0al., 2021; Li et\u00a0al., 2021), regularizers (Dittmer et\u00a0al., 2020; Obmann et\u00a0al., 2020; Ding et\u00a0al., 2022; Chen et\u00a0al., 2023), or additional operator structures (Uhlmann, 2009; Lai et\u00a0al., 2019; Yilmaz, 2001). However, such prior information is often either unavailable or problem-specific in complex systems. As a result, these methods can only solve the inverse problem for a particular system, and one has to start from scratch when the system varies (e.g., when the material of the specimen undergoes degradation in a material modeling task).   In this work, we propose Nonlocal Attention Operator (NAO), a novel attention-based neural operator architecture to simultaneously solve both forward and inverse modeling problems. Neural operators (NOs) (Li et\u00a0al., 2020a, c) learn mappings between infinite-dimensional function spaces in the form of integral operators, hence they provide promising tools for the discovery of continuum physical laws by manifesting the mapping between spatial and/or spatiotemporal data; see You et\u00a0al. (2022); Liu et\u00a0al. (2024a, b, 2023); Ong et\u00a0al. (2022); Cao (2021); Lu et\u00a0al. (2019, 2021); Goswami et\u00a0al. (2022); Gupta et\u00a0al. (2021) and references therein. However, most NOs focus on providing an efficient surrogate for the underlying physical system as a forward",
            "references": [
                {
                    "title": "Deep Neural Operator Enabled Digital Twin Modeling for Additive Manufacturing",
                    "abstract": "A digital twin (DT), with the components of a physics-based model, a data-driven model, and a machine learning (ML) enabled efficient surrogate, behaves as a virtual twin of the real-world physical process. In terms of Laser Powder Bed Fusion (L-PBF) based additive manufacturing (AM), a DT can predict the current and future states of the melt pool and the resulting defects corresponding to the input laser parameters, evolve itself by assimilating in-situ sensor data, and optimize the laser parameters to mitigate defect formation. In this paper, we present a deep neural operator enabled computational framework of the DT for closed-loop feedback control of the L-PBF process. This is accomplished by building a high-fidelity computational model to accurately represent the melt pool states, an efficient surrogate model to approximate the melt pool solution field, followed by an physics-based procedure to extract information from the computed melt pool simulation that can further be correlated to the defect quantities of interest (e.g., surface roughness). In particular, we leverage the data generated from the high-fidelity physics-based model and train a series of Fourier neural operator (FNO) based ML models to effectively learn the relation between the input laser parameters and the corresponding full temperature field of the melt pool. Subsequently, a set of physics-informed variables such as the melt pool dimensions and the peak temperature can be extracted to compute the resulting defects. An optimization algorithm is then exercised to control laser input and minimize defects. On the other hand, the constructed DT can also evolve with the physical twin via offline finetuning and online material calibration. Finally, a probabilistic framework is adopted for uncertainty quantification. The developed DT is envisioned to guide the AM process and facilitate high-quality manufacturing."
                },
                {
                    "title": "Towards a Foundation Model for Partial Differential Equations: Multi-Operator Learning and Extrapolation",
                    "abstract": "Foundation models, such as large language models, have demonstrated success in addressing various language and image processing tasks. In this work, we introduce a multi-modal foundation model for scientific problems, named PROSE-PDE. Our model, designed for bi-modality to bi-modality learning, is a multi-operator learning approach which can predict future states of spatiotemporal systems while concurrently learning the underlying governing equations of the physical system. Specifically, we focus on multi-operator learning by training distinct one-dimensional time-dependent nonlinear constant coefficient partial differential equations, with potential applications to many physical applications including physics, geology, and biology. More importantly, we provide three extrapolation studies to demonstrate that PROSE-PDE can generalize physical features through the robust training of multiple operators and that the proposed model can extrapolate to predict PDE solutions whose models or data were unseen during the training. Furthermore, we show through systematic numerical experiments that the utilization of the symbolic modality in our model effectively resolves the well-posedness problems with training multiple operators and thus enhances our model's predictive capabilities."
                },
                {
                    "title": "MODNO: Multi Operator Learning With Distributed Neural Operators",
                    "abstract": "The study of operator learning involves the utilization of neural networks to approximate operators. Traditionally, the focus has been on single-operator learning (SOL). However, recent advances have rapidly expanded this to include the approximation of multiple operators using foundation models equipped with millions or billions of trainable parameters, leading to the research of multi-operator learning (MOL). In this paper, we present a novel distributed training approach aimed at enabling a single neural operator with significantly fewer parameters to effectively tackle multi-operator learning challenges, all without incurring additional average costs. Our method is applicable to various neural operators, such as Deep Operator Neural Networks (DON). The core idea is to independently learn the output basis functions for each operator using its dedicated data, while simultaneously centralizing the learning of the input function encoding shared by all operators using the entire dataset. Through a systematic study of five numerical examples, we compare the accuracy and cost of training a single neural operator for each operator independently versus training a MOL model using our proposed method. Our results demonstrate enhanced efficiency and satisfactory accuracy. Moreover, our approach illustrates that some operators with limited data can be more effectively constructed with the aid of data from analogous operators through MOL learning. This highlights another MOL's potential to bolster operator learning."
                },
                {
                    "title": "PDEformer: Towards a Foundation Model for One-Dimensional Partial Differential Equations",
                    "abstract": "This paper introduces PDEformer, a neural solver for partial differential equations (PDEs) capable of simultaneously addressing various types of PDEs. We propose to represent the PDE in the form of a computational graph, facilitating the seamless integration of both symbolic and numerical information inherent in a PDE. A graph Transformer and an implicit neural representation (INR) are employed to generate mesh-free predicted solutions. Following pretraining on data exhibiting a certain level of diversity, our model achieves zero-shot accuracies on benchmark datasets that is comparable to those of specifically trained expert models. Additionally, PDEformer demonstrates promising results in the inverse problem of PDE coefficient recovery."
                },
                {
                    "title": "PDE Generalization of In-Context Operator Networks: A Study on 1D Scalar Nonlinear Conservation Laws",
                    "abstract": "Can we build a single large model for a wide range of PDE-related scientific learning tasks? Can this model generalize to new PDEs, even of new forms, without any fine-tuning? In-context operator learning and the corresponding model In-Context Operator Networks (ICON) represent an initial exploration of these questions. The capability of ICON regarding the first question has been demonstrated previously. In this paper, we present a detailed methodology for solving PDE problems with ICON, and show how a single ICON model can make forward and reverse predictions for different equations with different strides, provided with appropriately designed data prompts. We show the positive evidence to the second question, i.e., ICON can generalize well to some PDEs with new forms without any fine-tuning. This is exemplified through a study on 1D scalar nonlinear conservation laws, a family of PDEs with temporal evolution. We also show how to broaden the range of problems that an ICON model can address, by transforming functions and equations to ICON's capability scope. We believe that the progress in this paper is a significant step towards the goal of training a foundation model for PDE-related tasks under the in-context operator learning framework."
                },
                {
                    "title": "Peridynamic Neural Operators: A Data-Driven Nonlocal Constitutive Model for Complex Material Responses",
                    "abstract": "Neural operators, which can act as implicit solution operators of hidden governing equations, have recently become popular tools for learning the responses of complex real-world physical systems. Nevertheless, most neural operator applications have thus far been data-driven and neglect the intrinsic preservation of fundamental physical laws in data. In this work, we introduce a novel integral neural operator architecture called the Peridynamic Neural Operator (PNO) that learns a nonlocal constitutive law from data. This neural operator provides a forward model in the form of state-based peridynamics, with objectivity and momentum balance laws automatically guaranteed. As applications, we demonstrate the expressivity and efficacy of our model in learning complex material behaviors from both synthetic and experimental data sets. We show that, owing to its ability to capture complex responses, our learned neural operator achieves improved accuracy and efficiency compared to baseline models that use predefined constitutive laws. Moreover, by preserving the essential physical laws within the neural network architecture, the PNO is robust in treating noisy data. The method shows generalizability to different domain configurations, external loadings, and discretizations."
                },
                {
                    "title": "Harnessing the Power of Neural Operators with Automatically Encoded Conservation Laws",
                    "abstract": "Neural operators (NOs) have emerged as effective tools for modeling complex physical systems in scientific machine learning. In NOs, a central characteristic is to learn the governing physical laws directly from data. In contrast to other machine learning applications, partial knowledge is often known a priori about the physical system at hand whereby quantities such as mass, energy and momentum are exactly conserved. Currently, NOs have to learn these conservation laws from data and can only approximately satisfy them due to finite training data and random noise. In this work, we introduce conservation law-encoded neural operators (clawNOs), a suite of NOs that endow inference with automatic satisfaction of such conservation laws. ClawNOs are built with a divergence-free prediction of the solution field, with which the continuity equation is automatically guaranteed. As a consequence, clawNOs are compliant with the most fundamental and ubiquitous conservation laws essential for correct physical consistency. As demonstrations, we consider a wide variety of scientific applications ranging from constitutive modeling of material deformation, incompressible fluid dynamics, to atmospheric simulation. ClawNOs significantly outperform the state-of-the-art NOs in learning efficacy, especially in small-data regimes."
                },
                {
                    "title": "Let Data Talk: Data-regularized Operator Learning Theory for Inverse Problems",
                    "abstract": "Regularization plays a pivotal role in integrating prior information into inverse problems. While many deep learning methods have been proposed to solve inverse problems, determining where to apply regularization remains a crucial consideration. Typical methods regularize neural networks via architecture, wherein neural network functions parametrize the parameter of interest or the regularization term. We introduce a novel approach, denoted as the\"data-regularized operator learning\"(DaROL) method, designed to address PDE inverse problems. The DaROL method trains a neural network on data, regularized through common techniques such as Tikhonov variational methods and Bayesian inference. The DaROL method offers flexibility across different frameworks, faster inverse problem-solving, and a simpler structure that separates regularization and neural network training. We demonstrate that training a neural network on the regularized data is equivalent to supervised learning for a regularized inverse map. Furthermore, we provide sufficient conditions for the smoothness of such a regularized inverse map and estimate the learning error in terms of neural network size and the number of training samples."
                },
                {
                    "title": "Domain Agnostic Fourier Neural Operators",
                    "abstract": "Fourier neural operators (FNOs) can learn highly nonlinear mappings between function spaces, and have recently become a popular tool for learning responses of complex physical systems. However, to achieve good accuracy and efficiency, FNOs rely on the Fast Fourier transform (FFT), which is restricted to modeling problems on rectangular domains. To lift such a restriction and permit FFT on irregular geometries as well as topology changes, we introduce domain agnostic Fourier neural operator (DAFNO), a novel neural operator architecture for learning surrogates with irregular geometries and evolving domains. The key idea is to incorporate a smoothed characteristic function in the integral layer architecture of FNOs, and leverage FFT to achieve rapid computations, in such a way that the geometric information is explicitly encoded in the architecture. In our empirical evaluation, DAFNO has achieved state-of-the-art accuracy as compared to baseline neural operator models on two benchmark datasets of material modeling and airfoil simulation. To further demonstrate the capability and generalizability of DAFNO in handling complex domains with topology changes, we consider a brittle material fracture evolution problem. With only one training crack simulation sample, DAFNO has achieved generalizability to unseen loading scenarios and substantially different crack patterns from the trained scenario. Our code and data accompanying this paper are available at https://github.com/ningliu-iga/DAFNO."
                },
                {
                    "title": "Multiscale Attention via Wavelet Neural Operators for Vision Transformers",
                    "abstract": "Transformers have achieved widespread success in computer vision. At their heart, there is a Self-Attention (SA) mechanism, an inductive bias that associates each token in the input with every other token through a weighted basis. The standard SA mechanism has quadratic complexity with the sequence length, which impedes its utility to long sequences appearing in high resolution vision. Recently, inspired by operator learning for PDEs, Adaptive Fourier Neural Operators (AFNO) were introduced for high resolution attention based on global convolution that is efficiently implemented via FFT. However, the AFNO global filtering cannot well represent small and moderate scale structures that commonly appear in natural images. To leverage the coarse-to-fine scale structures we introduce a Multiscale Wavelet Attention (MWA) by leveraging wavelet neural operators which incurs linear complexity in the sequence size. We replace the attention in ViT with MWA and our experiments with CIFAR and Tiny-ImageNet classification demonstrate significant improvement over alternative Fourier-based attentions such as AFNO and Global Filter Network (GFN)."
                },
                {
                    "title": "Super-Resolution Neural Operator",
                    "abstract": "We propose Super-resolution Neural Operator (SRNO), a deep operator learning framework that can resolve high-resolution (HR) images at arbitrary scales from the low-resolution (LR) counterparts. Treating the LR-HR image pairs as continuous functions approximated with different grid sizes, SRNO learns the mapping between the corresponding function spaces. From the perspective of approximation theory, SRNO first embeds the LR input into a higher-dimensional latent representation space, trying to capture sufficient basis functions, and then iteratively approximates the implicit image function with a kernel integral mechanism, followed by a final dimensionality reduction step to generate the RGB representation at the target coordinates. The key characteristics distinguishing SRNO from prior continuous SR works are: 1) the kernel integral in each layer is efficiently implemented via the Galerkin-type attention, which possesses non-local properties in the spatial domain and therefore benefits the grid-free continuum; and 2) the multilayer attention architecture allows for the dynamic latent basis update, which is crucial for SR problems to \u201challucinate\u201d high-frequency information from the LR image. Experiments show that SRNO outperforms existing continuous SR methods in terms of both accuracy and running time. Our code is at https://github.com/2y7c3/Super-Resolution-Neural-Operator."
                },
                {
                    "title": "GNOT: A General Neural Operator Transformer for Operator Learning",
                    "abstract": "Learning partial differential equations' (PDEs) solution operators is an essential problem in machine learning. However, there are several challenges for learning operators in practical applications like the irregular mesh, multiple input functions, and complexity of the PDEs' solution. To address these challenges, we propose a general neural operator transformer (GNOT), a scalable and effective transformer-based framework for learning operators. By designing a novel heterogeneous normalized attention layer, our model is highly flexible to handle multiple input functions and irregular meshes. Besides, we introduce a geometric gating mechanism which could be viewed as a soft domain decomposition to solve the multi-scale problems. The large model capacity of the transformer architecture grants our model the possibility to scale to large datasets and practical problems. We conduct extensive experiments on multiple challenging datasets from different domains and achieve a remarkable improvement compared with alternative methods. Our code and data are publicly available at \\url{https://github.com/thu-ml/GNOT}."
                },
                {
                    "title": "Neural Inverse Operators for Solving PDE Inverse Problems",
                    "abstract": "A large class of inverse problems for PDEs are only well-defined as mappings from operators to functions. Existing operator learning frameworks map functions to functions and need to be modified to learn inverse maps from data. We propose a novel architecture termed Neural Inverse Operators (NIOs) to solve these PDE inverse problems. Motivated by the underlying mathematical structure, NIO is based on a suitable composition of DeepONets and FNOs to approximate mappings from operators to functions. A variety of experiments are presented to demonstrate that NIOs significantly outperform baselines and solve PDE inverse problems robustly, accurately and are several orders of magnitude faster than existing direct and PDE-constrained optimization methods."
                },
                {
                    "title": "INO: Invariant Neural Operators for Learning Complex Physical Systems with Momentum Conservation",
                    "abstract": "Neural operators, which emerge as implicit solution operators of hidden governing equations, have recently become popular tools for learning responses of complex real-world physical systems. Nevertheless, the majority of neural operator applications has thus far been data-driven, which neglects the intrinsic preservation of fundamental physical laws in data. In this paper, we introduce a novel integral neural operator architecture, to learn physical models with fundamental conservation laws automatically guaranteed. In particular, by replacing the frame-dependent position information with its invariant counterpart in the kernel space, the proposed neural operator is by design translation- and rotation-invariant, and consequently abides by the conservation laws of linear and angular momentums. As applications, we demonstrate the expressivity and efficacy of our model in learning complex material behaviors from both synthetic and experimental datasets, and show that, by automatically satisfying these essential physical laws, our learned neural operator is not only generalizable in handling translated and rotated datasets, but also achieves state-of-the-art accuracy and efficiency as compared to baseline neural operator models."
                },
                {
                    "title": "Reinforced Inverse Scattering",
                    "abstract": "Inverse wave scattering aims at determining the properties of an object using data on how the object scatters incoming waves. In order to collect information, sensors are put in different locations to send and receive waves from each other. The choice of sensor positions and incident wave frequencies determines the reconstruction quality of scatterer properties. This paper introduces reinforcement learning to develop precision imaging that decides sensor positions and wave frequencies adaptive to different scatterers in an intelligent way, thus obtaining a signi\ufb01cant improvement in reconstruction quality with limited imaging resources. Extensive numerical results will be provided to demonstrate the superiority of the proposed method over existing methods."
                },
                {
                    "title": "Transformer for Partial Differential Equations' Operator Learning",
                    "abstract": "Data-driven learning of partial differential equations' solution operators has recently emerged as a promising paradigm for approximating the underlying solutions. The solution operators are usually parameterized by deep learning models that are built upon problem-specific inductive biases. An example is a convolutional or a graph neural network that exploits the local grid structure where functions' values are sampled. The attention mechanism, on the other hand, provides a flexible way to implicitly exploit the patterns within inputs, and furthermore, relationship between arbitrary query locations and inputs. In this work, we present an attention-based framework for data-driven operator learning, which we term Operator Transformer (OFormer). Our framework is built upon self-attention, cross-attention, and a set of point-wise multilayer perceptrons (MLPs), and thus it makes few assumptions on the sampling pattern of the input function or query locations. We show that the proposed framework is competitive on standard benchmark problems and can flexibly be adapted to randomly sampled input."
                },
                {
                    "title": "Nonparametric Learning of Kernels in Nonlocal Operators",
                    "abstract": "Nonlocal operators with integral kernels have become a popular tool for designing solution maps between function spaces, due to their efficiency in representing long-range dependence and the attractive feature of being resolution-invariant. In this work, we provide a rigorous identifiability analysis and convergence study for learning kernels in nonlocal operators. It is found that kernel estimation is an ill-posed or even ill-defined inverse problem, leading to divergent estimators in the presence of modeling errors or measurement noises. To resolve this issue, we propose a nonparametric regression algorithm with a novel data-adaptive RKHS Tikhonov regularization method based on the function space of identifiability. The method yields a noisy-robust convergent estimator of the kernel as the data resolution refines, on both synthetic and real-world datasets. In particular, the method successfully learns a homogenized model for stress wave propagation in a heterogeneous solid, revealing the unknown governing laws from real-world data at the microscale. Our regularization method outperforms baseline methods in robustness, generalizability, and accuracy."
                },
                {
                    "title": "IAE-Net: Integral Autoencoders for Discretization-Invariant Learning",
                    "abstract": "Discretization invariant learning aims at learning in the in\ufb01nite-dimensional function spaces with the capacity to process heterogeneous discrete representations of functions as inputs and/or outputs of a learning model. This paper proposes a novel deep learning framework based on integral autoencoders (IAE-Net) for discretization invariant learning. The basic building block of IAE-Net consists of an encoder and a decoder as integral transforms with data-driven kernels, and a fully connected neural network between the encoder and decoder. This basic building block is applied in parallel in a wide multi-channel structure, which is repeatedly composed to form a deep and densely connected neural network with skip connections as IAE-Net. IAE-Net is trained with randomized data augmentation that generates training data with heterogeneous structures to facilitate the performance of discretization invariant learning. The proposed IAE-Net is tested with various applications in predictive data science, solving forward and inverse problems in scienti\ufb01c computing, and signal/image processing. Compared with alternatives in the literature, IAE-Net achieves state-of-the-art performance in existing applications and creates a wide range of new applications where existing methods fail."
                },
                {
                    "title": "Data adaptive RKHS Tikhonov regularization for learning kernels in operators",
                    "abstract": "We present DARTR: a Data Adaptive RKHS Tikhonov Regularization method for the linear inverse problem of nonparametric learning of function parameters in operators. A key ingredient is a system intrinsic data-adaptive (SIDA) RKHS, whose norm restricts the learning to take place in the function space of identifiability. DARTR utilizes this norm and selects the regularization parameter by the L-curve method. We illustrate its performance in examples including integral operators, nonlinear operators and nonlocal operators with discrete synthetic data. Numerical results show that DARTR leads to an accurate estimator robust to both numerical error due to discrete data and noise in data, and the estimator converges at a consistent rate as the data mesh refines under different levels of noises, outperforming two baseline regularizers using $l^2$ and $L^2$ norms."
                },
                {
                    "title": "Coupling Deep Learning with Full Waveform Inversion",
                    "abstract": "Full waveform inversion (FWI) aims at reconstructing unknown physical coefficients in wave equations using the wave field data generated from multiple incoming sources. In this work, we propose an offline-online computational strategy for coupling classical least-squares based computational inversion with modern deep learning based approaches for FWI to achieve advantages that can not be achieved with only one of the components. In a nutshell, we develop an offline learning strategy to construct a robust approximation to the inverse operator and utilize it to design a new objective function for the online inversion with new datasets. We demonstrate through numerical simulations that our coupling strategy improves the computational efficiency of FWI with reliable offline training on moderate computational resources (in terms of both the size of the training dataset and the computational cost needed)."
                },
                {
                    "title": "Adaptive Fourier Neural Operators: Efficient Token Mixers for Transformers",
                    "abstract": "Vision transformers have delivered tremendous success in representation learning. This is primarily due to effective token mixing through self attention. However, this scales quadratically with the number of pixels, which becomes infeasible for high-resolution inputs. To cope with this challenge, we propose Adaptive Fourier Neural Operator (AFNO) as an efficient token mixer that learns to mix in the Fourier domain. AFNO is based on a principled foundation of operator learning which allows us to frame token mixing as a continuous global convolution without any dependence on the input resolution. This principle was previously used to design FNO, which solves global convolution efficiently in the Fourier domain and has shown promise in learning challenging PDEs. To handle challenges in visual representation learning such as discontinuities in images and high resolution inputs, we propose principled architectural modifications to FNO which results in memory and computational efficiency. This includes imposing a block-diagonal structure on the channel mixing weights, adaptively sharing weights across tokens, and sparsifying the frequency modes via soft-thresholding and shrinkage. The resulting model is highly parallel with a quasi-linear complexity and has linear memory in the sequence size. AFNO outperforms self-attention mechanisms for few-shot segmentation in terms of both efficiency and accuracy. For Cityscapes segmentation with the Segformer-B3 backbone, AFNO can handle a sequence size of 65k and outperforms other efficient self-attention mechanisms."
                },
                {
                    "title": "Physics-Informed Neural Operator for Learning Partial Differential Equations",
                    "abstract": "In this paper, we propose physics-informed neural operators (PINO) that combine training data and physics constraints to learn the solution operator of a given family of parametric Partial Differential Equations (PDE). PINO is the first hybrid approach incorporating data and PDE constraints at different resolutions to learn the operator. Specifically, in PINO, we combine coarse-resolution training data with PDE constraints imposed at a higher resolution. The resulting PINO model can accurately approximate the ground-truth solution operator for many popular PDE families and shows no degradation in accuracy even under zero-shot super-resolution, i.e., being able to predict beyond the resolution of training data. PINO uses the Fourier neural operator (FNO) framework that is guaranteed to be a universal approximator for any continuous operator and discretization convergent in the limit of mesh refinement. By adding PDE constraints to FNO at a higher resolution, we obtain a high-fidelity reconstruction of the ground-truth operator. Moreover, PINO succeeds in settings where no training data is available and only PDE constraints are imposed, while previous approaches, such as the Physics-Informed Neural Network (PINN), fail due to optimization challenges, e.g., in multi-scale dynamic systems such as Kolmogorov flows."
                },
                {
                    "title": "Multiwavelet-based Operator Learning for Differential Equations",
                    "abstract": "The solution of a partial differential equation can be obtained by computing the inverse operator map between the input and the solution space. Towards this end, we introduce a \\textit{multiwavelet-based neural operator learning scheme} that compresses the associated operator's kernel using fine-grained wavelets. By explicitly embedding the inverse multiwavelet filters, we learn the projection of the kernel onto fixed multiwavelet polynomial bases. The projected kernel is trained at multiple scales derived from using repeated computation of multiwavelet transform. This allows learning the complex dependencies at various scales and results in a resolution-independent scheme. Compare to the prior works, we exploit the fundamental properties of the operator's kernel which enable numerically efficient representation. We perform experiments on the Korteweg-de Vries (KdV) equation, Burgers' equation, Darcy Flow, and Navier-Stokes equation. Compared with the existing neural operator approaches, our model shows significantly higher accuracy and achieves state-of-the-art in a range of datasets. For the time-varying equations, the proposed method exhibits a ($2X-10X$) improvement ($0.0018$ ($0.0033$) relative $L2$ error for Burgers' (KdV) equation). By learning the mappings between function spaces, the proposed method has the ability to find the solution of a high-resolution input after learning from lower-resolution data."
                },
                {
                    "title": "Global Filter Networks for Image Classification",
                    "abstract": "Recent advances in self-attention and pure multi-layer perceptrons (MLP) models for vision have shown great potential in achieving promising performance with fewer inductive biases. These models are generally based on learning interaction among spatial locations from raw data. The complexity of self-attention and MLP grows quadratically as the image size increases, which makes these models hard to scale up when high-resolution features are required. In this paper, we present the Global Filter Network (GFNet), a conceptually simple yet computationally efficient architecture, that learns long-term spatial dependencies in the frequency domain with log-linear complexity. Our architecture replaces the self-attention layer in vision transformers with three key operations: a 2D discrete Fourier transform, an element-wise multiplication between frequency-domain features and learnable global filters, and a 2D inverse Fourier transform. We exhibit favorable accuracy/complexity trade-offs of our models on both ImageNet and downstream tasks. Our results demonstrate that GFNet can be a very competitive alternative to transformer-style models and CNNs in efficiency, generalization ability and robustness. Code is available at https://github.com/raoyongming/GFNet"
                },
                {
                    "title": "Choose a Transformer: Fourier or Galerkin",
                    "abstract": "In this paper, we apply the self-attention from the state-of-the-art Transformer in Attention Is All You Need for the first time to a data-driven operator learning problem related to partial differential equations. An effort is put together to explain the heuristics of, and to improve the efficacy of the attention mechanism. By employing the operator approximation theory in Hilbert spaces, it is demonstrated for the first time that the softmax normalization in the scaled dot-product attention is sufficient but not necessary. Without softmax, the approximation capacity of a linearized Transformer variant can be proved to be comparable to a Petrov-Galerkin projection layer-wise, and the estimate is independent with respect to the sequence length. A new layer normalization scheme mimicking the Petrov-Galerkin projection is proposed to allow a scaling to propagate through attention layers, which helps the model achieve remarkable accuracy in operator learning tasks with unnormalized data. Finally, we present three operator learning experiments, including the viscid Burgers' equation, an interface Darcy flow, and an inverse interface coefficient identification problem. The newly proposed simple attention-based operator learner, Galerkin Transformer, shows significant improvements in both training cost and evaluation accuracy over its softmax-normalized counterparts."
                },
                {
                    "title": "Pay Attention to MLPs",
                    "abstract": "Transformers have become one of the most important architectural innovations in deep learning and have enabled many breakthroughs over the past few years. Here we propose a simple network architecture, gMLP, based on MLPs with gating, and show that it can perform as well as Transformers in key language and vision applications. Our comparisons show that self-attention is not critical for Vision Transformers, as gMLP can achieve the same accuracy. For BERT, our model achieves parity with Transformers on pretraining perplexity and is better on some downstream NLP tasks. On finetuning tasks where gMLP performs worse, making the gMLP model substantially larger can close the gap with Transformers. In general, our experiments show that gMLP can scale as well as Transformers over increased data and compute."
                },
                {
                    "title": "FNet: Mixing Tokens with Fourier Transforms",
                    "abstract": "We show that Transformer encoder architectures can be sped up, with limited accuracy costs, by replacing the self-attention sublayers with simple linear transformations that \u201cmix\u201d input tokens. Most surprisingly, we find that replacing the self-attention sublayer in a Transformer encoder with a standard, unparameterized Fourier Transform achieves 92-97% of the accuracy of BERT counterparts on the GLUE benchmark, but trains 80% faster on GPUs and 70% faster on TPUs at standard 512 input lengths. At longer input lengths, our FNet model is significantly faster: when compared to the \u201cefficient Transformers\u201d on the Long Range Arena benchmark, FNet matches the accuracy of the most accurate models, while outpacing the fastest models across all sequence lengths on GPUs (and across relatively shorter lengths on TPUs). Finally, FNet has a light memory footprint and is particularly efficient at smaller model sizes; for a fixed speed and accuracy budget, small FNet models outperform Transformer counterparts."
                },
                {
                    "title": "ResMLP: Feedforward Networks for Image Classification With Data-Efficient Training",
                    "abstract": "We present ResMLP, an architecture built entirely upon multi-layer perceptrons for image classification. It is a simple residual network that alternates (i) a linear layer in which image patches interact, independently and identically across channels, and (ii) a two-layer feed-forward network in which channels interact independently per patch. When trained with a modern training strategy using heavy data-augmentation and optionally distillation, it attains surprisingly good accuracy/complexity trade-offs on ImageNet. We also train ResMLP models in a self-supervised setup, to further remove priors from employing a labelled dataset. Finally, by adapting our model to machine translation we achieve surprisingly good results. We share pre-trained models and our code based on the Timm library."
                },
                {
                    "title": "MLP-Mixer: An all-MLP Architecture for Vision",
                    "abstract": "Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e.\"mixing\"the per-location features), and one with MLPs applied across patches (i.e.\"mixing\"spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers."
                },
                {
                    "title": "Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges",
                    "abstract": "Interpretability in machine learning (ML) is crucial for high stakes decisions and troubleshooting. In this work, we provide fundamental principles for interpretable ML, and dispel common misunderstandings that dilute the importance of this crucial topic. We also identify 10 technical challenge areas in interpretable machine learning and provide history and background on each problem. Some of these problems are classically important, and some are recent problems that have arisen in the last few years. These problems are: (1) Optimizing sparse logical models such as decision trees; (2) Optimization of scoring systems; (3) Placing constraints into generalized additive models to encourage sparsity and better interpretability; (4) Modern case-based reasoning, including neural networks and matching for causal inference; (5) Complete supervised disentanglement of neural networks; (6) Complete or even partial unsupervised disentanglement of neural networks; (7) Dimensionality reduction for data visualization; (8) Machine learning models that can incorporate physics and other generative or causal constraints; (9) Characterization of the\"Rashomon set\"of good models; and (10) Interpretable reinforcement learning. This survey is suitable as a starting point for statisticians and computer scientists interested in working in interpretable machine learning."
                },
                {
                    "title": "Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention",
                    "abstract": "Transformers achieve remarkable performance in several tasks but due to their quadratic complexity, with respect to the input's length, they are prohibitively slow for very long sequences. To address this limitation, we express the self-attention as a linear dot-product of kernel feature maps and make use of the associativity property of matrix products to reduce the complexity from $\\mathcal{O}\\left(N^2\\right)$ to $\\mathcal{O}\\left(N\\right)$, where $N$ is the sequence length. We show that this formulation permits an iterative implementation that dramatically accelerates autoregressive transformers and reveals their relationship to recurrent neural networks. Our linear transformers achieve similar performance to vanilla transformers and they are up to 4000x faster on autoregressive prediction of very long sequences."
                },
                {
                    "title": "Multipole Graph Neural Operator for Parametric Partial Differential Equations",
                    "abstract": "One of the main challenges in using deep learning-based methods for simulating physical systems and solving partial differential equations (PDEs) is formulating physics-based data in the desired structure for neural networks. Graph neural networks (GNNs) have gained popularity in this area since graphs offer a natural way of modeling particle interactions and provide a clear way of discretizing the continuum models. However, the graphs constructed for approximating such tasks usually ignore long-range interactions due to unfavorable scaling of the computational complexity with respect to the number of nodes. The errors due to these approximations scale with the discretization of the system, thereby not allowing for generalization under mesh-refinement. Inspired by the classical multipole methods, we propose a novel multi-level graph neural network framework that captures interaction at all ranges with only linear complexity. Our multi-level formulation is equivalent to recursively adding inducing points to the kernel matrix, unifying GNNs with multi-resolution matrix factorization of the kernel. Experiments confirm our multi-graph network learns discretization-invariant solution operators to PDEs and can be evaluated in linear time."
                },
                {
                    "title": "Linformer: Self-Attention with Linear Complexity",
                    "abstract": "Large transformer models have shown extraordinary success in achieving state-of-the-art results in many natural language processing applications. However, training and deploying these models can be prohibitively costly for long sequences, as the standard self-attention mechanism of the Transformer uses $O(n^2)$ time and space with respect to sequence length. In this paper, we demonstrate that the self-attention mechanism can be approximated by a low-rank matrix. We further exploit this finding to propose a new self-attention mechanism, which reduces the overall self-attention complexity from $O(n^2)$ to $O(n)$ in both time and space. The resulting linear transformer, the \\textit{Linformer}, performs on par with standard Transformer models, while being much more memory- and time-efficient."
                },
                {
                    "title": "Neural Operator: Graph Kernel Network for Partial Differential Equations",
                    "abstract": "The classical development of neural networks has been primarily for mappings between a finite-dimensional Euclidean space and a set of classes, or between two finite-dimensional Euclidean spaces. The purpose of this work is to generalize neural networks so that they can learn mappings between infinite-dimensional spaces (operators). The key innovation in our work is that a single set of network parameters, within a carefully designed network architecture, may be used to describe mappings between infinite-dimensional spaces and between different finite-dimensional approximations of those spaces. We formulate approximation of the infinite-dimensional mapping by composing nonlinear activation functions and a class of integral operators. The kernel integration is computed by message passing on graph networks. This approach has substantial practical consequences which we will illustrate in the context of mappings between input data to partial differential equations (PDEs) and their solutions. In this context, such learned networks can generalize among different approximation methods for the PDE (such as finite difference or finite element methods) and among approximations corresponding to different underlying levels of resolution and discretization. Experiments confirm that the proposed graph kernel network does have the desired properties and show competitive performance compared to the state of the art solvers."
                },
                {
                    "title": "Deep synthesis regularization of inverse problems",
                    "abstract": "Recently, a large number of efficient deep learning methods for solving inverse problems have been developed and show outstanding numerical performance. For these deep learning methods, however, a solid theoretical foundation in the form of reconstruction guarantees is missing. In contrast, for classical reconstruction methods, such as convex variational and frame-based regularization, theoretical convergence and convergence rate results are well established. In this paper, we introduce deep synthesis regularization (DESYRE) using neural networks as nonlinear synthesis operator bridging the gap between these two worlds. The proposed method allows to exploit the deep learning benefits of being well adjustable to available training data and on the other hand comes with a solid mathematical foundation. We present a complete convergence analysis with convergence rates for the proposed deep synthesis regularization. We present a strategy for constructing a synthesis network as part of an analysis-synthesis sequence together with an appropriate training strategy. Numerical results show the plausibility of our approach."
                },
                {
                    "title": "Interpretable Machine Learning",
                    "abstract": "Interpretable machine learning has become a popular research direction as deep neural networks (DNNs) have become more powerful and their applications more mainstream, yet DNNs remain difficult to understand. Testing with Concept Activation Vectors, TCAV, (Kim et al. 2017) is an approach to interpreting DNNs in a human-friendly way and has recently received significant attention in the machine learning community. The TCAV algorithm achieves a degree of global interpretability for DNNs through human-defined concepts as explanations. This project introduces Robust TCAV, which builds on TCAV and experimentally determines best practices for this method. The objectives for Robust TCAV are 1) Making TCAV more consistent by reducing variance in the TCAV score distribution and 2) Increasing CAV and TCAV score resistance to perturbations. A difference of means method for CAV generation was determined to be the best practice to achieve both objectives. Many areas of the TCAV process are explored including CAV visualization in low dimensions, negative class selection, and activation perturbation in the direction of a CAV. Finally, a thresholding technique is considered to remove noise in TCAV scores. This project is a step in the direction of making TCAV, an already impactful algorithm in interpretability, more reliable and useful for practitioners."
                },
                {
                    "title": "Axial Attention in Multidimensional Transformers",
                    "abstract": "We propose Axial Transformers, a self-attention-based autoregressive model for images and other data organized as high dimensional tensors. Existing autoregressive models either suffer from excessively large computational resource requirements for high dimensional data, or make compromises in terms of distribution expressiveness or ease of implementation in order to decrease resource requirements. Our architecture, by contrast, maintains both full expressiveness over joint distributions over data and ease of implementation with standard deep learning frameworks, while requiring reasonable memory and computation and achieving state-of-the-art results on standard generative modeling benchmarks. Our models are based on axial attention, a simple generalization of self-attention that naturally aligns with the multiple dimensions of the tensors in both the encoding and the decoding settings. Notably the proposed structure of the layers allows for the vast majority of the context to be computed in parallel during decoding without introducing any independence assumptions. This semi-parallel structure goes a long way to making decoding from even a very large Axial Transformer broadly applicable. We demonstrate state-of-the-art results for the Axial Transformer on the ImageNet-32 and ImageNet-64 image benchmarks as well as on the BAIR Robotic Pushing video benchmark. We open source the implementation of Axial Transformers."
                },
                {
                    "title": "Ab-Initio Solution of the Many-Electron Schr\u00f6dinger Equation with Deep Neural Networks",
                    "abstract": "Given access to accurate solutions of the many-electron Schr\\\"odinger equation, nearly all chemistry could be derived from first principles. Exact wavefunctions of interesting chemical systems are out of reach because they are NP-hard to compute in general, but approximations can be found using polynomially-scaling algorithms. The key challenge for many of these algorithms is the choice of wavefunction approximation, or Ansatz, which must trade off between efficiency and accuracy. Neural networks have shown impressive power as accurate practical function approximators and promise as a compact wavefunction Ansatz for spin systems, but problems in electronic structure require wavefunctions that obey Fermi-Dirac statistics. Here we introduce a novel deep learning architecture, the Fermionic Neural Network, as a powerful wavefunction Ansatz for many-electron systems. The Fermionic Neural Network is able to achieve accuracy beyond other variational quantum Monte Carlo Ans\\\"atze on a variety of atoms and small molecules. Using no data other than atomic positions and charges, we predict the dissociation curves of the nitrogen molecule and hydrogen chain, two challenging strongly-correlated systems, to significantly higher accuracy than the coupled cluster method, widely considered the most accurate scalable method for quantum chemistry at equilibrium geometry. This demonstrates that deep neural networks can improve the accuracy of variational quantum Monte Carlo to the point where it outperforms other ab-initio quantum chemistry methods, opening the possibility of accurate direct optimisation of wavefunctions for previously intractable molecules and solids."
                },
                {
                    "title": "Multiscale Modeling, Homogenization and Nonlocal Effects: Mathematical and Computational Issues",
                    "abstract": "In this work, we review the connection between the subjects of homogenization and nonlocal modeling and discuss the relevant computational issues. By further exploring this connection, we hope to promote the cross fertilization of ideas from the different research fronts. We illustrate how homogenization may help characterizing the nature and the form of nonlocal interactions hypothesized in nonlocal models. We also offer some perspective on how studies of nonlocality may help the development of more effective numerical methods for homogenization."
                },
                {
                    "title": "Transformer Dissection: An Unified Understanding for Transformer\u2019s Attention via the Lens of Kernel",
                    "abstract": "Transformer is a powerful architecture that achieves superior performance on various sequence learning tasks, including neural machine translation, language understanding, and sequence prediction. At the core of the Transformer is the attention mechanism, which concurrently processes all inputs in the streams. In this paper, we present a new formulation of attention via the lens of the kernel. To be more precise, we realize that the attention can be seen as applying kernel smoother over the inputs with the kernel scores being the similarities between inputs. This new formulation gives us a better way to understand individual components of the Transformer\u2019s attention, such as the better way to integrate the positional embedding. Another important advantage of our kernel-based formulation is that it paves the way to a larger space of composing Transformer\u2019s attention. As an example, we propose a new variant of Transformer\u2019s attention which models the input as a product of symmetric kernels. This approach achieves competitive performance to the current state of the art model with less computation. In our experiments, we empirically study different kernel construction strategies on two widely used tasks: neural machine translation and sequence prediction."
                },
                {
                    "title": "Generating Long Sequences with Sparse Transformers",
                    "abstract": "Transformers are powerful sequence models, but require time and memory that grows quadratically with the sequence length. In this paper we introduce sparse factorizations of the attention matrix which reduce this to $O(n \\sqrt{n})$. We also introduce a) a variation on architecture and initialization to train deeper networks, b) the recomputation of attention matrices to save memory, and c) fast attention kernels for training. We call networks with these changes Sparse Transformers, and show they can model sequences tens of thousands of timesteps long using hundreds of layers. We use the same architecture to model images, audio, and text from raw bytes, setting a new state of the art for density modeling of Enwik8, CIFAR-10, and ImageNet-64. We generate unconditional samples that demonstrate global coherence and great diversity, and show it is possible in principle to use self-attention to model sequences of length one million or more."
                },
                {
                    "title": "Machine learning and the physical sciences",
                    "abstract": "Machine learning (ML) encompasses a broad range of algorithms and modeling tools used for a vast array of data processing tasks, which has entered most scientific disciplines in recent years. This article reviews in a selective way the recent research on the interface between machine learning and the physical sciences. This includes conceptual developments in ML motivated by physical insights, applications of machine learning techniques to several domains in physics, and cross fertilization between the two fields. After giving a basic notion of machine learning methods and principles, examples are described of how statistical physics is used to understand methods in ML. This review then describes applications of ML methods in particle physics and cosmology, quantum many-body physics, quantum computing, and chemical and material physics. Research and development into novel computing architectures aimed at accelerating ML are also highlighted. Each of the sections describe recent successes as well as domain-specific methodology and challenges."
                },
                {
                    "title": "Frequency Principle: Fourier Analysis Sheds Light on Deep Neural Networks",
                    "abstract": "We study the training process of Deep Neural Networks (DNNs) from the Fourier analysis perspective. We demonstrate a very universal Frequency Principle (F-Principle) --- DNNs often fit target functions from low to high frequencies --- on high-dimensional benchmark datasets such as MNIST/CIFAR10 and deep neural networks such as VGG16. This F-Principle of DNNs is opposite to the behavior of most conventional iterative numerical schemes (e.g., Jacobi method), which exhibit faster convergence for higher frequencies for various scientific computing problems. With a simple theory, we illustrate that this F-Principle results from the regularity of the commonly used activation functions. The F-Principle implies an implicit bias that DNNs tend to fit training data by a low-frequency function. This understanding provides an explanation of good generalization of DNNs on most real datasets and bad generalization of DNNs on parity function or randomized dataset."
                },
                {
                    "title": "Inverse Problems for the Stationary Transport Equation in the Diffusion Scaling",
                    "abstract": "We consider the inverse problem of reconstructing the optical parameters of the radiative transfer equation (RTE) from boundary measurements in the diffusion limit. In the diffusive regime (the Knudsen number $\\mathsf{Kn}\\ll 1$), the forward problem for the stationary RTE is well approximated by an elliptic equation. However, the connection between the inverse problem for the RTE and the inverse problem for the elliptic equation has not been fully developed. This problem is particularly interesting because the former one is mildly ill-posed , with a Lipschitz type stability estimate, while the latter is well known to be severely ill-posed with a logarithmic type stability estimate. In this paper, we derive stability estimates for the inverse problem for RTE and examine its dependence on $\\mathsf{Kn}$. We show that the stability is Lipschitz in all regimes, but the coefficient deteriorates as $e^{\\frac{1}{\\mathsf{Kn}}}$, making the inverse problem of RTE severely ill-posed when $\\mathsf{Kn}$ is small. In this way we connect the two inverse problems. Numerical results agree with the analysis of worsening stability as the Knudsen number gets smaller."
                },
                {
                    "title": "Deep Potential Molecular Dynamics: a scalable model with the accuracy of quantum mechanics",
                    "abstract": "We introduce a scheme for molecular simulations, the deep potential molecular dynamics (DPMD) method, based on a many-body potential and interatomic forces generated by a carefully crafted deep neural network trained with ab\u00a0initio data. The neural network model preserves all the natural symmetries in the problem. It is first-principles based in the sense that there are no ad\u00a0hoc components aside from the network model. We show that the proposed scheme provides an efficient and accurate protocol in a variety of systems, including bulk materials and molecules. In all these cases, DPMD gives results that are essentially indistinguishable from the original data, at a cost that scales linearly with system size."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Electrical impedance tomography and Calder\u00f3n's problem",
                    "abstract": "We survey mathematical developments in the inverse method of electrical impedance tomography which consists in determining the electrical properties of a medium by making voltage and current measurements at the boundary of the medium. In the mathematical literature, this is also known as Calder\u00f3n's problem from Calder\u00f3n's pioneer contribution (Calder\u00f3n 1980 Seminar on Numerical Analysis and its Applications to Continuum Physics (R\u00edo de Janeiro, 1980) p 65 (Soc. Brasil. Mat.)). We concentrate this review around the topic of complex geometrical optics solutions that have led to many advances in the field. In the last section, we review some counterexamples to Calder\u00f3n's problems that have attracted a lot of interest because of connections with cloaking and invisibility."
                },
                {
                    "title": "Autoprogressive training of neural network constitutive models",
                    "abstract": "A new method, termed autoprogressive training, for training neural networks to learn complex stress\u2013strain behaviour of materials using global load\u2013deflection response measured in a structural test is described. The richness of the constitutive information that is generally implicitly contained in the results of structural tests may in many cases make it possible to train a neural network material model from only a small number of such tests, thus overcoming one of the perceived limitations of a neural network approach to modelling of material behaviour; namely, that a voluminous amount of material test data is required. The method uses the partially-trained neural network in a central way in an iterative non-linear finite element analysis of the test specimen in order to extract approximate, but gradually improving, stress\u2013strain information with which to train the neural network.\r\n\r\nAn example is presented in which a simple neural network constitutive model of a T300/976 graphite/epoxy unidirectional lamina is trained, using the load\u2013deflection response recorded during a destructive compressive test of a [(\u00b145)6]S laminated structural plate containing an open hole. The results of a subsequent forward analysis are also presented, in which the trained material model is used to simulate the response of a compressively loaded [(\u00b130)6]S structural laminate containing an open hole. Avenues for further improvement of the neural network model are also suggested.\r\n\r\nThe proposed autoprogressive algorithm appears to have wide application in the general area of Non-Destructive Evaluation (NDE) and damage detection. Most NDE experiments can be viewed as structural tests and the proposed methodology can be used to determine certain damage indices, similar to the way in which constitutive models are determined. \u00a9 1998 John Wiley & Sons, Ltd."
                },
                {
                    "title": "Knowledge-Based Modeling of Material Behavior with Neural Networks",
                    "abstract": "To date, material modeling has involved the development of mathematical models of material behavior derived from human observation of, and reasoning with, experimental data. An alternative, discussed in this paper, is to use a computation and knowledge representation paradigm, called neural networks, developed by researchers in connectionism (a subfield of artificial intelligence) to model material behavior. The main benefits in using a neural\u2010network approach are that all behavior can be represented within a unified environment of a neural network and that the network is built directly from experimental data using the self\u2010organizing capabilities of the neural network, i.e., the network is presented with the experimental data and \u201clearns\u201d the relationships between stresses and strains. Such a modeling strategy has important implications for modeling the behavior of modern, complex materials, such as composites. In this paper, the behaviors of concrete in the state of plane stress under monotonic biaxial ..."
                },
                {
                    "title": "HT-Net: Hierarchical Transformer based Operator Learning Model for Multiscale PDEs",
                    "abstract": ". Complex nonlinear interplays of multiple scales give rise to many interesting physical phenomena and pose major di\ufb03culties for the computer simulation of multiscale PDE models in areas such as reservoir simulation, high frequency scattering and turbulence modeling. In this paper, we introduce a hierarchical transformer (HT) scheme to e\ufb03ciently learn the solution operator for multiscale PDEs. We construct a hierarchical architecture with scale adaptive interaction range, such that the features can be computed in a nested manner and with a controllable linear cost. Self-attentions over a hierarchy of levels can be used to encode and decode the multiscale solution space over all scale ranges. In addition, we adopt an empirical H 1 loss function to counteract the spectral bias of the neural network approximation for multiscale functions. In the numerical experiments, we demonstrate the superior performance of the HT scheme compared with state-of-the-art (SOTA) methods for representative multiscale problems."
                },
                {
                    "title": "Physics-Informed Neural Operators",
                    "abstract": ". Standard neural networks can approximate general nonlinear operators, represented either explicitly by a combination of mathematical operators, e.g. , in an advection-di\ufb00usion-reaction partial di\ufb00erential equation, or simply as a black box, e.g. , a system-of-systems. The \ufb01rst neural operator was the Deep Operator Network (DeepONet), proposed in 2019 based on rigorous approximation theory. Since then, a few other less general operators have been published, e.g., based on graph neural networks or Fourier transforms. For black box systems, training of neural operators is data-driven only but if the governing equations are known they can be incorporated into the loss function during training to develop physics-informed neural operators. Neural operators can be used as surrogates in design problems, uncertainty quanti\ufb01cation, autonomous systems, and almost in any application requiring real-time inference. More-over, independently pre-trained DeepONets can be used as components of a complex multi-physics system by coupling them together with relatively light training. Here, we present a review of DeepONet, the Fourier neural operator, and the graph neural operator, as well as appropriate extensions with feature expansions, and highlight their usefulness in diverse applications in computational mechanics, including porous media, \ufb02uid mechanics, and solid mechanics. schematic of a polymorphic DeepONet and its resemblance to a human neuron. An operator network is made up of two deep"
                }
            ],
            "categories": [
                "cs.LG",
                "math.AP"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively infer the underlying physical parameters and governing mechanisms of complex systems from limited and discrete observational data in material modeling tasks?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning, particularly in applications where understanding the underlying physical laws is essential for safety and reliability, such as in engineering and materials science. By developing interpretable models that can accurately infer physical parameters, we can enhance the robustness and transparency of data-driven decision-making processes. This research could lead to significant advancements in predictive modeling, enabling more effective designs and optimizations in various industries, ultimately impacting future research directions by fostering the integration of machine learning with physics-based modeling.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the ill-posed nature of the inverse problem, where limited and discrete measurements of continuous functions can lead to multiple plausible solutions for the underlying parameters. Naive approaches may fail because they do not account for the high-dimensionality of the governing laws or the inherent biases in neural network approximations. Additionally, the scarcity of data exacerbates the ill-posedness, making it difficult to accurately infer the hidden states. Overcoming these technical and theoretical obstacles requires sophisticated methodologies that can effectively integrate prior knowledge and handle the complexities of the underlying physical systems.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on specific systems, limiting the applicability of existing solutions. Many methods rely on prior information, such as governing PDEs or regularizers, which may not be available or applicable across different scenarios. This has created a barrier to generalizing solutions for varying systems, such as those experiencing material degradation. Our approach, utilizing the Nonlocal Attention Operator (NAO), aims to address these limitations by providing a flexible framework that can simultaneously tackle both forward and inverse modeling problems, thus improving upon prior work by enabling broader applicability and adaptability.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the Nonlocal Attention Operator (NAO), which leverages attention-based neural operator architecture to learn mappings between infinite-dimensional function spaces. We will utilize a dataset comprising deformation fields and corresponding loading fields in material modeling tasks. The performance will be evaluated using metrics that assess both the accuracy of parameter inference and the interpretability of the learned models. We expect that our approach will yield a"
            }
        },
        "author_data": {
            "40900f79-c0ce-4ecd-aa05-80ecf2783061": {
                "pk": "40900f79-c0ce-4ecd-aa05-80ecf2783061",
                "name": "Yue Yu",
                "collaborators": [],
                "domain": [
                    "Computational Physics",
                    "Finite Element Method",
                    "Quantum Mechanics",
                    "Numerical Analysis"
                ],
                "publications": [
                    {
                        "title": "varFEM: variational formulation based programming for finite element methods in Matlab",
                        "abstract": "This paper summarizes the development of varFEM, which provides a realization of the programming style in FreeFEM by using the Matlab language."
                    },
                    {
                        "title": "Tripled composite fermion liquid in generalized Pfaffian of v=5/2 and non-abelian anyonic quasiholes with k=3",
                        "abstract": "This paper has been withdrawn by the author, due to an important sign error in Eqn. 2."
                    },
                    {
                        "title": "Implementation of Polygonal Mesh Refinement in MATLAB",
                        "abstract": "We present a simple and efficient MATLAB implementation of the local refinement for polygonal meshes. The purpose of this implementation is primarily educational, especially on the teaching of adaptive virtual element methods."
                    },
                    {
                        "title": "Programming of linear virtual element methods in three dimensions",
                        "abstract": "We present a simple and efficient MATLAB implementation of the linear virtual element method for the three dimensional Poisson equation. The purpose of this software is primarily educational, to demonstrate how the key components of the method can be translated into code."
                    },
                    {
                        "title": "A Systematic Improvement for Calculation to Conductivity in Anomalous Propagation of Surface Acoustic Wave at $\u03bd={1/2}$",
                        "abstract": "We report a systematic improvement to calculate the conductivity which associates to the anomaly of the propagation of surface acoustic waves at $\\nu={1/2}$ above a two-dimensional electron gas. We try to resolve the discrepancy between theoretical and experimental values for the magnitude of $\\sigma_{xx}(q)$ by considering the contribution to the response functions from the self-interaction among the Chern-Simons gauge fluctuations."
                    },
                    {
                        "title": "Linearizations via Dirac delta function",
                        "abstract": "This note is to show that the position-space embedding in \\cite{ESP2021embedding} in the position and occupation bases can be obtained by considering the dynamics of Dirac delta function $$\\delta(\\mathbf{x}- \\mathbf{z}(t)) = \\delta(x_1-z_1(t))\\cdots \\delta(x_d-z_d(t)),$$ where $\\mathbf{z}(t)\\in \\mathbb{R}^d$ is the solution of a nonlinear dynamical system and $\\mathbf{x}\\in \\mathbb{R}^d$ is a variable in the position space."
                    },
                    {
                        "title": "A Microscopic Model of Edge States of Fractional Quantum Hall Liquid: From Composite Fermions to Calogero-Sutherland Model",
                        "abstract": "Based on the composite fermion approach, we derive a microscopic theory describing the low-lying edge excitations in the fractional quantum Hall liquid with $\\nu=\\frac{\\nu^*}{\\tilde\\phi\\nu^*+1}$. For $\\nu^*>0$, it is found that the composite fermion model reduces to an SU$(\\nu^*)$ Calogero-Sutherland model in the one-dimensional limit, whereas it is not exact soluble for $\\nu^*<0$. However, the ground states in both cases can be found and the low-lying excitations can be shown the chiral Luttinger liquid behaviors since a gap exists between the right- and left-moving sectors in each branch of the azimuthal excitations."
                    },
                    {
                        "title": "Comments on `Rashba precession in quantum wire with interaction'",
                        "abstract": "In a recent Rapid Communication (Phys. Rev. B {\\bf 63}, 121210(R) (2001)), Ha\\\"usler showed that the interaction between electrons in quantum wires may enhance the persistent spin current arising from Rashba spin-orbital coupling. In this Comments, we would like to point out that this 'enhancement' comes from a misunderstanding to the boosting persistent current in the Luttinger liquid theory. A correct calculation will not give such an enhancement of the persistent spin current. Meanwhile, we provide a Luttinger liquid theory with Rashba spin-orbital interaction by bosonization, which may show how the Rashba precession is in a Luttinger liquid."
                    },
                    {
                        "title": "Time-evolution stability of order parameters and phase diagrams of bosons on optical lattice",
                        "abstract": "Stemming from the Heisenberg equations of motion, we study the time-evolution stability of the order parameters for the cold atoms on optical lattices. The requirement of this stability of the order parameters endows the phase diagram with a fruitful structure in the superfluid phase. For the one-component Bose-Hubbard model, we see that this stability of order parameter leads to a physically receivable phase diagram. For two-component bosons, we show that the molecules are preformed in the atomic superfluid and then condenses into a molecular superfluid phase at a critical repulsive inter-species interaction, which resembles the pre-pairing mechanism in high $T_c$ superconductor of Cu-O cuprates."
                    },
                    {
                        "title": "Calogero-Sutherland-Lieb-Liniger gas in one-dimensional cold atoms",
                        "abstract": "We study an array of cigar-like Bose atom condensates confined in a cylinder and examine the competition between the dipole-dipole and the short range interactions. The system is effectively reduced to a one-dimensional boson one with a contact and inverse square interactions. We call this system the Calogero-Sutherland-Lieb-Liniger gas. The universal properties of the ground state are analyzed by the renormalization group theory. By using the bosonization techniques to the excluson gas, we calculate the non-universal exponent depending on the microscopic parameters. This exponent may be experimentally measurable."
                    },
                    {
                        "title": "mVEM: A MATLAB Software Package for the Virtual Element Methods",
                        "abstract": "This paper summarizes the development of mVEM, a MATLAB software package containing efficient and easy-following codes for various virtual element methods (VEMs) published in the literature. We explain in detail the numerical implementation of the mixed VEMs for the Darcy problem and the three-dimensional linear VEMs for the Poisson equation. For other model problems, we present the construction of the discrete methods and only provide the implementation of the elliptic projection matrices. Some mesh related functions are also given in the package, including the mesh generation and refinement in two or three dimensions. mVEM is free and open source software."
                    },
                    {
                        "title": "Unified Implementation of Finite Element Methods involving Jumps and Averages in Matlab",
                        "abstract": "We provide unified implementations of the finite element methods involving jumps and averages in Maltab by combing the use of the software package varFEM, including the adaptive finite element methods for the Poisson equation and the $C^0$ interior penalty methods for the biharmonic equation. The design ideas can be extended to other Galerkin-based methods, for example, the discontinuous Galerkin methods and the virtual element methods."
                    },
                    {
                        "title": "From Composite Fermions to Calogero-Sutherland Model: Edge of Fractional Quantum Hall Liquid and the Dimension Reduction",
                        "abstract": "We derive a microscopic model describing the low-lying edge excitations in the fractional quantum Hall liquid with $\\nu=\\frac{\\nu^*} {\\tilde\\phi\\nu^*+1}$. For $\\nu^*>0$, it is found that the composite fermion model reduces to an SU$(\\nu^*)$ Calogero-Sutherland model in a dimension reduction, whereas it is not exact soluble for $\\nu^*<0$. However, the ground states in both cases can be found and the low-lying excitations can be shown the chiral Luttinger liquid behaviors. On the other hand, we shows that the finite temperature behavior of $G-T$ curve will deviate from the prediction of the chiral Luttinger liquid. We also point out that the suppression of the `spin' degrees of freedom agrees with very recent experiments by Chang et al. The two-boson model of Lee and Wen is described microscopically."
                    },
                    {
                        "title": "Incompressible excitonic superfluid of ultracold Bose atoms in an optical lattice: a new superfluid phase in the one-component Bose-Hubbard model",
                        "abstract": "We predict that a new superfluid phase, the incompressible excitonic superfluid (IESF), in the phase diagram of ultracold Bose atoms in $d>1$ dimensional optical lattices, which is caused by the spontaneous breaking of the symmetry of translation of the lattice. Within mean field theory, the critical temperature of the phase transition from this IESF to the normal fluid (NF) is calculated and the triple-critical point of the three phases is determined. We also investigate both configuration and gauge field fluctuations and show the IESF state is stable against these fluctuations. We expect this IESF phase can be experimentally observed by loading cold Bose atoms into a two-dimensional lattice where the atom filling fraction deviates slightly from exact commensurations. The signatures distinguishing this IESF from the common atom superfluid (ASF) are that (i) the critical temperature of the IEST/NF transition is independent of interaction, unlike the ASF/NF transition; (ii) the IESF is incompressible while the ASF is compressible."
                    },
                    {
                        "title": "Short-range coherence of a lattice Bose atom gas in the Mott insulating phase",
                        "abstract": "We study the short-range coherence of ultracold lattice Bose gases in the Mott insulating phase. We calculate the visibility of the interference pattern and the results agree quantitatively with the recent experimental measurement [Phys. Rev. Lett. 95, 050404 (2005)]. The visibility deviation from the inversely linear dependence on the bare on-site interaction U_0 is explained both in smaller and larger U_0. For a smaller U_0, it comes from a second order correction. For a larger U_0, except the breakdown of adiabaticity as analyzed by Gerbier et al, there might be another source to cause this deviation, which is the diversity between $U_0$ determined by the single atom Wannier function and the effective on site interaction U_eff for a multi-occupation per site."
                    },
                    {
                        "title": "Calogero-Sutherland gas of ultracold Bose atoms",
                        "abstract": "We show that the Calogero-Sutherland (C-S) gas, a famous exact soluble one-dimensional system with an inverse square long range interaction, can be realized by dimension reduction in a cold Bose atom system with a dipole-dipole interaction. Depending on the orientation of the dipoles, the effective interaction is either attractive or repulsive. The low-lying effective theory may be a Luttinger liquid when the exclusion statistics parameter $\\lambda$ may be well-defined. We hope that the C-S gas can be realized experimentally and the Luttinger liquid character can be observed."
                    },
                    {
                        "title": "Dynamics of edge Majorana fermions in $\u03bd=\\frac{5}2$ fractional quantum Hall effects",
                        "abstract": "Commencing with the composite fermion description of the $\\nu=5/2$ fractional quantum Hall effect, we study the dynamics of the edge neutral Majorana fermions. We confirm that these neutral modes are chiral and show that a conventional p-wave pairing interaction between CFs does not contribute to the dynamics of the edge neutral fermions. We find an important bilinear coupling between the charged and neutral modes. We show that owing to this coupling, the dispersion of the neutral modes is linear and their velocities are proportional to the wave vector of the charged mode. This dynamic origin of the motion of the edge Majorana fermions was never expected before."
                    },
                    {
                        "title": "Quaternate generalization of Pfaffian state at $\u03bd=5/2$",
                        "abstract": "We consider a quaternately generalized Pfaffian QGPf$(\\frac{1}{J(z_i,z_j,z_k,z_l)})[J(z_1,...,z_N)]^2$ in which the square of Vandermonde determinant, $[J(z_1,...,z_N)]^2$, implies the upmost Landau level is half filled. This wave function is the unique highest density zero energy state of a special short range interacting Hamiltonian. One can think this quaternate composite fermion liquid as a competing ground state of Moore-Read (MR) Pfaffian state at $\\nu=5/2$. The degeneracy of the quasihole excitations above the QGPf is higher than that of Moore-Read even Read-Rezayi quasiholes. The QGPf is related to a unitary conformal field theory with $Z_2\\times Z_2$ parafermions in coset space $SU(3)_2/U(1)^2$ . Because of the level-rank duality between $SU(3)_2$ and $SU(2)_3$ in conformal field theory, these quasiholes above this QGPf state obeying non-abelian anyonic statistics are expected to support the universal quantum computation at $\\nu=5/2$ as Read-Rezayi quasiholes at $\\nu=13/5$. The edge states of QGPf are very different from those of the Pfaffian's."
                    },
                    {
                        "title": "Seiberg-Witten monopoles: Weyl metal coupled to chiral magnets",
                        "abstract": "We study a Weyl (semi)metal which couples to local magnets. In the continuum limit, the Hamiltonian of the system matches the Chern-Simons-Maxwell-Dirac functional and then the ground state is governed by generalized Seiberg-Witten (SW) or Freund equations in terms of the sign of Dzyaloshinskii-Moriya coupling. The ground states determined by the Freund equations may either be monopolar Weyl semimetal accompanied by the ferromagnetic magnets (MWFM) or SW monopoles which consist of spheric Weyl fermions coupled to chiral magnets, depending on the strength of the Kondo coupling. In the latter phase, the topological ground state is characterized by SW invariants and with a Weyl surface on which the Weyl metal is of an exotic dispersion $\\propto \\sqrt k$. There are also the metastable SW monopole solutions carrying an opposite SW invariant for the SW equations while the ground state in this case is the MWFM state."
                    },
                    {
                        "title": "A lowest-order locking-free nonconforming virtual element method based on the reduced integration technique for linear elasticity problems",
                        "abstract": "We develop a lowest-order nonconforming virtual element method for planar linear elasticity, which can be viewed as an extension of the idea in Falk (1991) to the virtual element method (VEM), with the family of polygonal meshes satisfying a very general geometric assumption. The method is shown to be uniformly convergent for the nearly incompressible case with optimal rates of convergence. The crucial step is to establish the discrete Korn's inequality, yielding the coercivity of the discrete bilinear form. We also provide a unified locking-free scheme both for the conforming and nonconforming VEMs in the lowest order case. Numerical results validate the feasibility and effectiveness of the proposed numerical algorithms."
                    }
                ]
            },
            "d612dbf5-59a9-4567-9c87-712d1c815c5e": {
                "pk": "d612dbf5-59a9-4567-9c87-712d1c815c5e",
                "name": "Ning Liu",
                "collaborators": [
                    "Zhanchun Tu",
                    "Z. C. Tu"
                ],
                "domain": [
                    "Bose-Einstein Condensates",
                    "Quantum Physics",
                    "Particle Physics",
                    "Numerical Simulation"
                ],
                "publications": [
                    {
                        "title": "Weak and strong coupling polarons in binary Bose-Einstein condensates",
                        "abstract": "The Bose polaron is a quasiparticle that arises from the interaction between impurities and Bogoliubov excitation in Bose-Einstein condensates, analogous to the polaron formed by electrons and phonons in solid-state physics. In this paper, we investigate the effect of phase separation on weakly coupled and strongly coupled Bose polarons. Our findings reveal that phase separation induces a remarkable alteration in the properties of weakly coupled Bose polarons. However, in the case of strong coupling, phase separation cannot destroy the highly self-trapping state of Bose polarons."
                    },
                    {
                        "title": "QCD corrections to the production of $t\\bar{t}\u03b3$ at the ILC",
                        "abstract": "A precise calculation of the top quark pair production associated with a hard photon is essential for testing the electroweak property of the top quark in the Standard Model (SM). We investigate the one-loop QCD corrections to the process $e^{+}e^{-} \\to t\\bar{t}\\gamma$ at the International Linear Collider (ILC), and find that the $K$-factor can be as large as 1.238 (1.105, 1.060) for a center-of-mass energy $\\sqrt{s}=500 (800, 1500)$ GeV. The transverse momentum distributions of the top quark and photon are respectively shown at leading order (LO) and next-to-leading order(NLO). Due to the asymmetric rapidity distribution of the top (anti-top) quark, we also study the top quark forward-backward asymmetry ($A^{t}_{FB}$) in $t\\bar{t}\\gamma$ production at NLO, which is found to be 45.82 (55.25, 55.89)% for $\\sqrt{s}=500 (800, 1500)$ GeV."
                    },
                    {
                        "title": "Elastic characteristics for naturally-fractured reservoirs using an integrated LSM-DFN scheme with rough contact deformation",
                        "abstract": "Crack micro-geometries and tribological properties pose an important impact on the elastic characteristics of fractured rocks. Numerical simulation as a promising way for this issue still faces some challenges. With the rapid development of computers and computational techniques, discrete-based numerical approaches with desirable properties have been increasingly developed, but few attempts to consider the particle surface roughness in a lattice type model. For this purpose, an integrated numerical scheme accounting rough contact deformation is developed by coupling modified LSM and DFN modeling for predicting the effective mechanical properties of a realistic outcrop. Smooth joint logic is introduced to consider contact and slip behaviors at fracture surfaces and a modified contact relation to estimating the normal force-displacement from rough contact deformation. Improved constitutive laws are developed and employed for rock matrix and rough fracture surface and implemented in the modified LSM. Complex fracture networks presented by DFNs are automatically extracted based on the gradient Hough transform algorithm. This developed framework is validated by classic equivalent medium theories. It shows the model could be used to emulate naturally-fractured media and to quantitatively investigate the effects of fracture attributes and micro-scale surface roughness on the compression mechanism."
                    },
                    {
                        "title": "Dynamics of Momentum Distribution and Structure Factor in a Weakly Interacting Bose Gas with a Periodical Modulation",
                        "abstract": "The momentum distribution and dynamical structure factor in a weakly interacting Bose gas with a time-dependent periodic modulation in terms of the Bogoliubov treatment are investigated. The evolution equation related to the Bogoliubov weights happens to be a solvable Mathieu equation when the coupling strength is periodically modulated. An exact relation between the time derivatives of momentum distribution and dynamical structure factor is derived, which indicates that the single-particle property strongly related to the two-body property in the evolutions of Bose-Einstein condensates. It is found that the momentum distribution and dynamical structure factor cannot display periodical behavior. For stable dynamics, some particular peaks in the curves of momentum distribution and dynamical structure factor appear synchronously, which is consistent with the derivative relation."
                    },
                    {
                        "title": "Polarons in Binary Bose-Einstein Condensates",
                        "abstract": "Bose polarons are quasiparticles formed through the interaction between impurities and Bose-Einstein condensates. In this paper, we derive an effective Fr\\\"{o}hlich Hamiltonian using the generalized Bogoliubov transformation. The effective Fr\\\"{o}hlich Hamiltonian encompasses two types of effective interactions: impurity-density (ID) coupling and impurity-spin (IS) coupling. Furthermore, we employ the Lee-Low-Pines variational approach to investigate the relevant properties of Bose polarons induced by the ID and IS coupling. These properties include the ground state energy, effective mass, and average number of virtual phonons. Our findings reveal that the contribution resulting from IS couplings to the ground energy decreases to zero near the miscible-immiscible boundary. Additionally, the increase of the IS coupling induces a greater number of virtual phonons, impeding the movement of impurities and leading to a significant increase in the effective mass of Bose polarons."
                    }
                ]
            },
            "4292d9c8-f10a-4042-bd10-189ffe4111e2": {
                "pk": "4292d9c8-f10a-4042-bd10-189ffe4111e2",
                "name": "Fei Lu",
                "collaborators": [
                    "Quanjun Lang",
                    "Yaozhong Hu",
                    "David Nualart",
                    "Alexandre J. Chorin",
                    "Zhongyang Li",
                    "Zehong Zhang",
                    "Meng Fang",
                    "Xiangpeng Wan",
                    "Fei Xing",
                    "Kai Zou"
                ],
                "domain": [
                    "Stochastic Processes",
                    "Machine Learning",
                    "Nonlinear Dynamics",
                    "Bayesian Inference"
                ],
                "publications": [
                    {
                        "title": "Data-driven model reduction for stochastic Burgers equations",
                        "abstract": "We present a class of efficient parametric closure models for 1D stochastic Burgers equations. Casting it as statistical learning of the flow map, we derive the parametric form by representing the unresolved high wavenumber Fourier modes as functionals of the resolved variables' trajectory. The reduced models are nonlinear autoregression (NAR) time series models, with coefficients estimated from data by least squares. The NAR models can accurately reproduce the energy spectrum, the invariant densities, and the autocorrelations. Taking advantage of the simplicity of the NAR models, we investigate maximal and optimal space-time reduction. Reduction in space dimension is unlimited, and NAR models with two Fourier modes can perform well. The NAR model's stability limits time reduction, with a maximal time step smaller than that of the K-mode Galerkin system. We report a potential criterion for optimal space-time reduction: the NAR models achieve minimal relative error in the energy spectrum at the time step where the K-mode Galerkin system's mean CFL number agrees with the full model's."
                    },
                    {
                        "title": "Identifiability of interaction kernels in mean-field equations of interacting particles",
                        "abstract": "This study examines the identifiability of interaction kernels in mean-field equations of interacting particles or agents, an area of growing interest across various scientific and engineering fields. The main focus is identifying data-dependent function spaces where a quadratic loss functional possesses a unique minimizer. We consider two data-adaptive $L^2$ spaces: one weighted by a data-adaptive measure and the other using the Lebesgue measure. In each $L^2$ space, we show that the function space of identifiability is the closure of the RKHS associated with the integral operator of inversion.   Alongside prior research, our study completes a full characterization of identifiability in interacting particle systems with either finite or infinite particles, highlighting critical differences between these two settings. Moreover, the identifiability analysis has important implications for computational practice. It shows that the inverse problem is ill-posed, necessitating regularization. Our numerical demonstrations show that the weighted $L^2$ space is preferable over the unweighted $L^2$ space, as it yields more accurate regularized estimators."
                    },
                    {
                        "title": "Small noise analysis for Tikhonov and RKHS regularizations",
                        "abstract": "Regularization plays a pivotal role in ill-posed machine learning and inverse problems. However, the fundamental comparative analysis of various regularization norms remains open. We establish a small noise analysis framework to assess the effects of norms in Tikhonov and RKHS regularizations, in the context of ill-posed linear inverse problems with Gaussian noise. This framework studies the convergence rates of regularized estimators in the small noise limit and reveals the potential instability of the conventional L2-regularizer. We solve such instability by proposing an innovative class of adaptive fractional RKHS regularizers, which covers the L2 Tikhonov and RKHS regularizations by adjusting the fractional smoothness parameter. A surprising insight is that over-smoothing via these fractional RKHSs consistently yields optimal convergence rates, but the optimal hyper-parameter may decay too fast to be selected in practice."
                    },
                    {
                        "title": "On the coercivity condition in the learning of interacting particle systems",
                        "abstract": "In the learning of systems of interacting particles or agents, coercivity condition ensures identifiability of the interaction functions, providing the foundation of learning by nonparametric regression. The coercivity condition is equivalent to the strictly positive definiteness of an integral kernel arising in the learning. We show that for a class of interaction functions such that the system is ergodic, the integral kernel is strictly positive definite, and hence the coercivity condition holds true."
                    },
                    {
                        "title": "Cluster Prediction for Opinion Dynamics from Partial Observations",
                        "abstract": "We present a Bayesian approach to predict the clustering of opinions for a system of interacting agents from partial observations. The Bayesian formulation overcomes the unobservability of the system and quantifies the uncertainty in the prediction. We characterize the clustering by the posterior of the clusters' sizes and centers, and we represent the posterior by samples. To overcome the challenge in sampling the high-dimensional posterior, we introduce an auxiliary implicit sampling (AIS) algorithm using two-step observations. Numerical results show that the AIS algorithm leads to accurate predictions of the sizes and centers for the leading clusters, in both cases of noiseless and noisy observations. In particular, the centers are predicted with high success rates, but the sizes exhibit a considerable uncertainty that is sensitive to observation noise and the observation ratio."
                    },
                    {
                        "title": "Learning interaction kernels in mean-field equations of 1st-order systems of interacting particles",
                        "abstract": "We introduce a nonparametric algorithm to learn interaction kernels of mean-field equations for 1st-order systems of interacting particles. The data consist of discrete space-time observations of the solution. By least squares with regularization, the algorithm learns the kernel on data-adaptive hypothesis spaces efficiently. A key ingredient is a probabilistic error functional derived from the likelihood of the mean-field equation's diffusion process. The estimator converges, in a reproducing kernel Hilbert space and an L2 space under an identifiability condition, at a rate optimal in the sense that it equals the numerical integrator's order. We demonstrate our algorithm on three typical examples: the opinion dynamics with a piecewise linear kernel, the granular media model with a quadratic kernel, and the aggregation-diffusion with a repulsive-attractive kernel."
                    },
                    {
                        "title": "Non-degeneracy of some Sobolev Pseudo-norms of fractional Brownian motion",
                        "abstract": "Applying an upper bound estimate for small $L^{2}$ ball probability for fractional Brownian motion (fBm), we prove the non-degeneracy of some Sobolev pseudo-norms of fBm."
                    },
                    {
                        "title": "MathOdyssey: Benchmarking Mathematical Problem-Solving Skills in Large Language Models Using Odyssey Math Data",
                        "abstract": "Large language models (LLMs) have significantly advanced natural language understanding and demonstrated strong problem-solving abilities. Despite these successes, most LLMs still struggle with solving mathematical problems due to the intricate reasoning required. This paper investigates the mathematical problem-solving capabilities of LLMs using the newly developed \"MathOdyssey\" dataset. The dataset includes diverse mathematical problems at high school and university levels, created by experts from notable institutions to rigorously test LLMs in advanced problem-solving scenarios and cover a wider range of subject areas. By providing the MathOdyssey dataset as a resource to the AI community, we aim to contribute to the understanding and improvement of AI capabilities in complex mathematical problem-solving. We conduct benchmarking on open-source models, such as Llama-3 and DBRX-Instruct, and closed-source models from the GPT series and Gemini models. Our results indicate that while LLMs perform well on routine and moderately difficult tasks, they face significant challenges with Olympiad-level problems and complex university-level questions. Our analysis shows a narrowing performance gap between open-source and closed-source models, yet substantial challenges remain, particularly with the most demanding problems. This study highlights the ongoing need for research to enhance the mathematical reasoning of LLMs. The dataset, results, and code are publicly available."
                    },
                    {
                        "title": "H\u00f6lder Continuity of the Solution for a Class of Nonlinear SPDE Arising from One Dimensional Superprocesses",
                        "abstract": "The H\\\"older continuity of the solution to a nonlinear stochastic partial differential equation arising from one dimensional super process is obtained. It is proved that the H\\\"older exponent in time variable is as close as to 1/4, improving the result of 1/10 in a recent paper by Li et al [3]. The method is to use the Malliavin calculus. The H\\\"older continuity in spatial variable x of exponent 1/2 is also obtained by using this new approach. This H\\\"older continuity result is sharp since the corresponding linear heat equation has the same H\\\"older continuity."
                    },
                    {
                        "title": "Convergence of densities of some functionals of Gaussian processes",
                        "abstract": "The aim of this paper is to establish the uniform convergence of the densities of a sequence of random variables, which are functionals of an underlying Gaussian process, to a normal density. Precise estimates for the uniform distance are derived by using the techniques of Malliavin calculus, combined with Stein's method for normal approximation. We need to assume some non-degeneracy conditions. First, the study is focused on random variables in a fixed Wiener chaos, and later, the results are extended to the uniform convergence of the derivatives of the densities and to the case of random vectors in some fixed chaos, which are uniformly non-degenerate in the sense of Malliavin calculus. Explicit upper bounds for the uniform norm are obtained for random variables in the second Wiener chaos, and an application to the convergence of densities of the least square estimator for the drift parameter in Ornstein-Uhlenbeck processes is discussed."
                    },
                    {
                        "title": "A discrete approach to stochastic parametrization and dimensional reduction in nonlinear dynamics",
                        "abstract": "Many physical systems are described by nonlinear differential equations that are too complicated to solve in full. A natural way to proceed is to divide the variables into those that are of direct interest and those that are not, formulate solvable approximate equations for the variables of greater interest, and use data and statistical methods to account for the impact of the other variables. In the present paper the problem is considered in a fully discrete-time setting, which simplifies both the analysis of the data and the numerical algorithms. The resulting time series are identified by a NARMAX (nonlinear autoregression moving average with exogenous input) representation familiar from engineering practice. The connections with the Mori-Zwanzig formalism of statistical physics are discussed, as well as an application to the Lorenz 96 system."
                    },
                    {
                        "title": "Data-based stochastic model reduction for the Kuramoto--Sivashinsky equation",
                        "abstract": "The problem of constructing data-based, predictive, reduced models for the Kuramoto-Sivashinsky equation is considered, under circumstances where one has observation data only for a small subset of the dynamical variables. Accurate prediction is achieved by developing a discrete-time stochastic reduced system, based on a NARMAX (Nonlinear Autoregressive Moving Average with eXogenous input) representation. The practical issue, with the NARMAX representation as with any other, is to identify an efficient structure, i.e., one with a small number of terms and coefficients. This is accomplished here by estimating coefficients for an approximate inertial form. The broader significance of the results is discussed."
                    },
                    {
                        "title": "Domain Adaptive Monocular Depth Estimation With Semantic Information",
                        "abstract": "The advent of deep learning has brought an impressive advance to monocular depth estimation, e.g., supervised monocular depth estimation has been thoroughly investigated. However, the large amount of the RGB-to-depth dataset may not be always available since collecting accurate depth ground truth according to the RGB image is a time-consuming and expensive task. Although the network can be trained on an alternative dataset to overcome the dataset scale problem, the trained model is hard to generalize to the target domain due to the domain discrepancy. Adversarial domain alignment has demonstrated its efficacy to mitigate the domain shift on simple image classification tasks in previous works. However, traditional approaches hardly handle the conditional alignment as they solely consider the feature map of the network. In this paper, we propose an adversarial training model that leverages semantic information to narrow the domain gap. Based on the experiments conducted on the datasets for the monocular depth estimation task including KITTI and Cityscapes, the proposed compact model achieves state-of-the-art performance comparable to complex latest models and shows favorable results on boundaries and objects at far distances."
                    },
                    {
                        "title": "Unsupervised learning of observation functions in state-space models by nonparametric moment methods",
                        "abstract": "We investigate the unsupervised learning of non-invertible observation functions in nonlinear state-space models. Assuming abundant data of the observation process along with the distribution of the state process, we introduce a nonparametric generalized moment method to estimate the observation function via constrained regression. The major challenge comes from the non-invertibility of the observation function and the lack of data pairs between the state and observation. We address the fundamental issue of identifiability from quadratic loss functionals and show that the function space of identifiability is the closure of a RKHS that is intrinsic to the state process. Numerical results show that the first two moments and temporal correlations, along with upper and lower bounds, can identify functions ranging from piecewise polynomials to smooth functions, leading to convergent estimators. The limitations of this method, such as non-identifiability due to symmetry and stationarity, are also discussed."
                    }
                ]
            },
            "0cb2181b-3cc9-42c7-b53c-5a17468e6023": {
                "pk": "0cb2181b-3cc9-42c7-b53c-5a17468e6023",
                "name": "Tian Gao",
                "collaborators": [
                    "Fei Yen",
                    "Jinglai Li",
                    "Yongsheng Liu",
                    "Adam Berlie",
                    "Qichuan Yang",
                    "Zhiqiang He",
                    "Kai Zhao",
                    "Clemens Rosenbaum",
                    "Tim Klinger",
                    "Jie Chen"
                ],
                "domain": [
                    "Optimization",
                    "Quantum Phenomena",
                    "Localization",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "A Derivative-Free Trust-Region Algorithm for Reliability-Based Optimization",
                        "abstract": "In this note, we present a derivative-free trust-region (TR) algorithm for reliability based optimization (RBO) problems. The proposed algorithm consists of solving a set of subproblems, in which simple surrogate models of the reliability constraints are constructed and used in solving the subproblems. Taking advantage of the special structure of the RBO problems, we employ a sample reweighting method to evaluate the failure probabilities, which constructs the surrogate for the reliability constraints by performing only a single full reliability evaluation in each iteration. With numerical experiments, we illustrate that the proposed algorithm is competitive against existing methods."
                    },
                    {
                        "title": "Dielectric Anomaly in Ice near 20 K; Evidence of Macroscopic Quantum Phenomena",
                        "abstract": "H2O is one of the most important substances needed in sustaining life; but yet not much is known about its ground state. Here, a previously unidentified anomaly is identified in the form of a minimum in the imaginary part of the dielectric constant with respect to temperature near 20 K while the real part remains monotonic. Isothermal dispersion and absorption measurements show coinciding results. For the case of heavy ice (D2O), no anomaly was identified confirming an apparent isotope effect. Concerted quantum tunneling of protons is believed to be the main cause behind the reported anomaly. Our findings identify another system that exhibits macroscopic quantum phenomena of which rarely occur in nature."
                    },
                    {
                        "title": "Possible magnetism based on orbital motion of protons in ice",
                        "abstract": "A peak anomaly is observed in the magnetic susceptibility as a function of temperature in solid H2O near Tp=60 K. At external magnetic fields below 2 kOe, Tp becomes positive in the temperature range between 45 and 66 K. The magnetic field dependence of the susceptibility in the same temperature range exhibits an inverted ferromagnetic hysteretic loop superimposed on top of the diamagnetic signature of ice at fields below 600 Oe. We suggest that a fraction of protons that are capable of undergoing correlated tunneling in a hexagonal path without disrupting the stoichiometry of the lattice create an induced magnetic field opposite to the induced magnetic field created by the electrons upon application of an external field which counters the overall diamagnetism of the material."
                    },
                    {
                        "title": "A Time Localization System in Smart Home Using Hierarchical Structure and Dynamic Frequency",
                        "abstract": "Both GPS and WiFi based localization have been exploited in recent years, yet most researches focus on localizing at home without environment context. Besides, the near home or workplace area is complex and has little attention in smart home or IOT. Therefore, after exploring the realistic route in and out of building, we conducted a time localization system (TLS) based on off-the-shelf smart phones with WiFi identification. TLS can identify the received signal strength indication (RSSI) of home and construct radio map of users' time route without site survey. As designed to service the smart devices in home, TLS applies the time interval as the distance of positions and as the variables of WiFi environment to mark time points. Experimental results with real users show that TLS as a service system for timeline localization achieves a median accuracy of 70 seconds and is more robust compared with nearest neighbor localization approach."
                    },
                    {
                        "title": "e-QRAQ: A Multi-turn Reasoning Dataset and Simulator with Explanations",
                        "abstract": "In this paper we present a new dataset and user simulator e-QRAQ (explainable Query, Reason, and Answer Question) which tests an Agent's ability to read an ambiguous text; ask questions until it can answer a challenge question; and explain the reasoning behind its questions and answer. The User simulator provides the Agent with a short, ambiguous story and a challenge question about the story. The story is ambiguous because some of the entities have been replaced by variables. At each turn the Agent may ask for the value of a variable or try to answer the challenge question. In response the User simulator provides a natural language explanation of why the Agent's query or answer was useful in narrowing down the set of possible answers, or not. To demonstrate one potential application of the e-QRAQ dataset, we train a new neural architecture based on End-to-End Memory Networks to successfully generate both predictions and partial explanations of its current understanding of the problem. We observe a strong correlation between the quality of the prediction and explanation."
                    },
                    {
                        "title": "A Sequential Set Generation Method for Predicting Set-Valued Outputs",
                        "abstract": "Consider a general machine learning setting where the output is a set of labels or sequences. This output set is unordered and its size varies with the input. Whereas multi-label classification methods seem a natural first resort, they are not readily applicable to set-valued outputs because of the growth rate of the output space; and because conventional sequence generation doesn't reflect sets' order-free nature. In this paper, we propose a unified framework--sequential set generation (SSG)--that can handle output sets of labels and sequences. SSG is a meta-algorithm that leverages any probabilistic learning method for label or sequence prediction, but employs a proper regularization such that a new label or sequence is generated repeatedly until the full set is produced. Though SSG is sequential in nature, it does not penalize the ordering of the appearance of the set elements and can be applied to a variety of set output problems, such as a set of classification labels or sequences. We perform experiments with both benchmark and synthetic data sets and demonstrate SSG's strong performance over baseline methods."
                    }
                ]
            },
            "d0c95634-8b9d-446c-9ca1-634bebb0fac2": {
                "pk": "d0c95634-8b9d-446c-9ca1-634bebb0fac2",
                "name": "Siavash Jafarzadeh",
                "collaborators": [
                    "Yue Yu",
                    "Florin Bobaru",
                    "Ning Liu",
                    "Farzaneh Mousavi",
                    "Adam Larios",
                    "Stewart Silling",
                    "Michael Hillman",
                    "Zhongqiang Zhang",
                    "Brian Y. Lattimer",
                    "Shuna Ni"
                ],
                "domain": [
                    "Computational Mechanics",
                    "Neural Operators",
                    "Machine Learning",
                    "Peridynamics"
                ],
                "publications": [
                    {
                        "title": "An Ultra-high-speed Reproducing Kernel Particle Method",
                        "abstract": "In this work, the fast-convolving reproducing kernel particle method (FC-RKPM) is introduced. This method is hundreds to millions of times faster than the traditional RKPM for 3D meshfree simulations. In this approach, the meshfree discretizations with RK approximation are expressed in terms of convolution sums. Fast Fourier transform (FFT) is then used to efficiently compute the convolutions. Certain modifications to the domain and shape functions are considered to maintain generality for complex geometries and arbitrary boundary conditions. The new method does not need to identify, store, and loop over the neighbors which is one of the bottleneck of the traditional meshfree methods. As a result, the run-times and memory allocations are independent of the number of neighbors and the shape function support size. As a model problem, the method is laid out for a Galerkin weak form of the Poisson problem with the RK approximation, and is verified in 1D, 2D, and 3D. Tables with run-times and allocated memory are presented to compare the performance of FC-RKPM with the traditional method in 3D. The performance is studied for various node numbers, support size, and approximation degree. All the implementation details and the roadmap for software development are also provided. Application of the new method to nonlinear and explicit problems are briefly discussed as well."
                    },
                    {
                        "title": "An ordinary state-based peridynamic elastoplastic 2D model consistent with J2 plasticity",
                        "abstract": "We present a new ordinary state-based peridynamic model (OSB-PD) in 2D consistent with J2 plasticity using a novel decomposition for force and extension states. A new strategy for testing the consistency of an OSB-PD formulation for elastoplasticity is introduced. In contrast with other similar models, the new elasto-plastic OSB-PD model is objective and works for large rotations. We formulate two rate-independent yield functions equivalent to J2 plasticity with associated flow rules, based on force states and on the strain energy density. An efficient return mapping algorithm for peridynamic formulations in plasticity is proposed. We verify the model against results obtained with Abaqus for the classical formulation using two examples, one featuring plastic deformations under large rotations. The evolution of the plastic zone computed with the new model matches that from the classical model found with Abaqus."
                    },
                    {
                        "title": "A general and fast convolution-based method for peridynamics: applications to elasticity and brittle fracture",
                        "abstract": "We introduce a general and fast convolution-based method (FCBM) for peridynamics (PD). Expressing the PD integrals in terms of convolutions and computing them by fast Fourier transform (FFT), we reduce the computational complexity of PD models from O(N^2) to O(Nlog_2 N), with N being the total number of discretization nodes. Initial neighbor identification and storing neighbor information is not required, and, as a consequence, memory allocation scales with O(N) instead of O(N^2), common for existing methods. The method is applicable to bounded domains with arbitrary shapes and boundary conditions via an embedded constraint (EC) approach. We explain the FCBM-EC formulation for certain bond-based and state-based, linear and nonlinear PD models of elasticity and dynamic brittle fracture, as applications. We solve a 3D elastostatic problem and show that the FCBM reduces the computational time from days to hours and from years to days, compared with the original meshfree discretization for PD models. Large-scale computations of PD models are feasible with the new method, and we demonstrate its versatility by simulating, with ease, the difficult problem of multiple crack branching in a brittle plate."
                    },
                    {
                        "title": "Efficient solutions for nonlocal diffusion problems via boundary-adapted spectral methods",
                        "abstract": "We introduce an efficient boundary-adapted spectral method for peridynamic diffusion problems with arbitrary boundary conditions. The spectral approach transforms the convolution integral in the peridynamic formulation into a multiplication in the Fourier space, resulting in computations that scale as O(NlogN). The limitation of regular spectral methods to periodic problems is eliminated using the volume penalization method. We show that arbitrary boundary conditions or volume constraints can be enforced in this way to achieve high levels of accuracy. To test the performance of our approach we compare the computational results with analytical solutions of the nonlocal problem. The performance is tested with convergence studies in terms of nodal discretization and the size of the penalization parameter in problems with Dirichlet and Neumann boundary conditions."
                    },
                    {
                        "title": "Domain Agnostic Fourier Neural Operators",
                        "abstract": "Fourier neural operators (FNOs) can learn highly nonlinear mappings between function spaces, and have recently become a popular tool for learning responses of complex physical systems. However, to achieve good accuracy and efficiency, FNOs rely on the Fast Fourier transform (FFT), which is restricted to modeling problems on rectangular domains. To lift such a restriction and permit FFT on irregular geometries as well as topology changes, we introduce domain agnostic Fourier neural operator (DAFNO), a novel neural operator architecture for learning surrogates with irregular geometries and evolving domains. The key idea is to incorporate a smoothed characteristic function in the integral layer architecture of FNOs, and leverage FFT to achieve rapid computations, in such a way that the geometric information is explicitly encoded in the architecture. In our empirical evaluation, DAFNO has achieved state-of-the-art accuracy as compared to baseline neural operator models on two benchmark datasets of material modeling and airfoil simulation. To further demonstrate the capability and generalizability of DAFNO in handling complex domains with topology changes, we consider a brittle material fracture evolution problem. With only one training crack simulation sample, DAFNO has achieved generalizability to unseen loading scenarios and substantially different crack patterns from the trained scenario. Our code and data accompanying this paper are available at https://github.com/ningliu-iga/DAFNO."
                    },
                    {
                        "title": "Peridynamic Neural Operators: A Data-Driven Nonlocal Constitutive Model for Complex Material Responses",
                        "abstract": "Neural operators, which can act as implicit solution operators of hidden governing equations, have recently become popular tools for learning the responses of complex real-world physical systems. Nevertheless, most neural operator applications have thus far been data-driven and neglect the intrinsic preservation of fundamental physical laws in data. In this work, we introduce a novel integral neural operator architecture called the Peridynamic Neural Operator (PNO) that learns a nonlocal constitutive law from data. This neural operator provides a forward model in the form of state-based peridynamics, with objectivity and momentum balance laws automatically guaranteed. As applications, we demonstrate the expressivity and efficacy of our model in learning complex material behaviors from both synthetic and experimental data sets. We show that, owing to its ability to capture complex responses, our learned neural operator achieves improved accuracy and efficiency compared to baseline models that use predefined constitutive laws. Moreover, by preserving the essential physical laws within the neural network architecture, the PNO is robust in treating noisy data. The method shows generalizability to different domain configurations, external loadings, and discretizations."
                    },
                    {
                        "title": "Large language models, physics-based modeling, experimental measurements: the trinity of data-scarce learning of polymer properties",
                        "abstract": "Large language models (LLMs) bear promise as a fast and accurate material modeling paradigm for evaluation, analysis, and design. Their vast number of trainable parameters necessitates a wealth of data to achieve accuracy and mitigate overfitting. However, experimental measurements are often limited and costly to obtain in sufficient quantities for finetuning. To this end, we present a physics-based training pipeline that tackles the pathology of data scarcity. The core enabler is a physics-based modeling framework that generates a multitude of synthetic data to align the LLM to a physically consistent initial state before finetuning. Our framework features a two-phase training strategy: (1) utilizing the large-in-amount while less accurate synthetic data for supervised pretraining, and (2) finetuning the phase-1 model with limited experimental data. We empirically demonstrate that supervised pretraining is vital to obtaining accurate finetuned LLMs, via the lens of learning polymer flammability metrics where cone calorimeter data is sparse."
                    },
                    {
                        "title": "Heterogeneous Peridynamic Neural Operators: Discover Biotissue Constitutive Law and Microstructure From Digital Image Correlation Measurements",
                        "abstract": "Human tissues are highly organized structures with collagen fiber arrangements varying from point to point. Anisotropy of the tissue arises from the natural orientation of the fibers, resulting in location-dependent anisotropy. Heterogeneity also plays an important role in tissue function. It is therefore critical to discover and understand the distribution of fiber orientations from experimental mechanical measurements such as digital image correlation (DIC) data. To this end, we introduce the Heterogeneous Peridynamic Neural Operator (HeteroPNO) approach for data-driven constitutive modeling of heterogeneous anisotropic materials. Our goal is to learn a nonlocal constitutive law together with the material microstructure, in the form of a heterogeneous fiber orientation field, from load-displacement field measurements. We propose a two-phase learning approach. Firstly, we learn a homogeneous constitutive law in the form of a neural network-based kernel function and a nonlocal bond force, to capture complex homogeneous material responses from data. Then, in the second phase we reinitialize the learnt bond force and the kernel function, and training them together with a fiber orientation field for each material point. Owing to the state-based peridynamic skeleton, our HeteroPNO-learned material models are objective and have the balance of linear and angular momentum guaranteed. Moreover, the effects from heterogeneity and nonlinear constitutive relationship are captured by the kernel function and the bond force respectively, enabling physical interpretability. As a result, our HeteroPNO architecture can learn a constitutive model for a biological tissue with anisotropic heterogeneous response undergoing large deformation regime. Moreover, the framework is capable to provide displacement and stress field predictions for new and unseen loading instances."
                    }
                ]
            },
            "21498bae-2b91-4d23-a831-fa5ea56ed752": {
                "pk": "21498bae-2b91-4d23-a831-fa5ea56ed752",
                "name": "Stewart Silling",
                "collaborators": [
                    "Yue Yu",
                    "Marta D'Elia",
                    "Huaiqian You",
                    "Robert Lipton",
                    "Richard Lehoucq",
                    "Siavash Jafarzadeh",
                    "Ning Liu",
                    "Zhongqiang Zhang",
                    "Tian Gao",
                    "Yiming Fan"
                ],
                "domain": [
                    "Machine Learning",
                    "Continuum Mechanics",
                    "Fracture Mechanics",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "Complex fracture nucleation and evolution with nonlocal elastodynamics",
                        "abstract": "A mechanical model is introduced for predicting the initiation and evolution of complex fracture patterns without the need for a damage variable or law. The model, a continuum variant of Newton's second law, uses integral rather than partial differential operators where the region of integration is over finite domain. The force interaction is derived from a novel nonconvex strain energy density function, resulting in a nonmonotonic material model. The resulting equation of motion is proved to be mathematically well-posed. The model has the capacity to simulate nucleation and growth of multiple, mutually interacting dynamic fractures. In the limit of zero region of integration, the model reproduces the classic Griffith model of brittle fracture. The simplicity of the formulation avoids the need for supplemental kinetic relations that dictate crack growth or the need for an explicit damage evolution law."
                    },
                    {
                        "title": "Data-driven learning of nonlocal models: from high-fidelity simulations to constitutive laws",
                        "abstract": "We show that machine learning can improve the accuracy of simulations of stress waves in one-dimensional composite materials. We propose a data-driven technique to learn nonlocal constitutive laws for stress wave propagation models. The method is an optimization-based technique in which the nonlocal kernel function is approximated via Bernstein polynomials. The kernel, including both its functional form and parameters, is derived so that when used in a nonlocal solver, it generates solutions that closely match high-fidelity data. The optimal kernel therefore acts as a homogenized nonlocal continuum model that accurately reproduces wave motion in a smaller-scale, more detailed model that can include multiple materials. We apply this technique to wave propagation within a heterogeneous bar with a periodic microstructure. Several one-dimensional numerical tests illustrate the accuracy of our algorithm. The optimal kernel is demonstrated to reproduce high-fidelity data for a composite material in applications that are substantially different from the problems used as training data."
                    },
                    {
                        "title": "A data-driven peridynamic continuum model for upscaling molecular dynamics",
                        "abstract": "Nonlocal models, including peridynamics, often use integral operators that embed lengthscales in their definition. However, the integrands in these operators are difficult to define from the data that are typically available for a given physical system, such as laboratory mechanical property tests. In contrast, molecular dynamics (MD) does not require these integrands, but it suffers from computational limitations in the length and time scales it can address. To combine the strengths of both methods and to obtain a coarse-grained, homogenized continuum model that efficiently and accurately captures materials' behavior, we propose a learning framework to extract, from MD data, an optimal Linear Peridynamic Solid (LPS) model as a surrogate for MD displacements. To maximize the accuracy of the learnt model we allow the peridynamic influence function to be partially negative, while preserving the well-posedness of the resulting model. To achieve this, we provide sufficient well-posedness conditions for discretized LPS models with sign-changing influence functions and develop a constrained optimization algorithm that minimizes the equation residual while enforcing such solvability conditions. This framework guarantees that the resulting model is mathematically well-posed, physically consistent, and that it generalizes well to settings that are different from the ones used during training. We illustrate the efficacy of the proposed approach with several numerical tests for single layer graphene. Our two-dimensional tests show the robustness of the proposed algorithm on validation data sets that include thermal noise, different domain shapes and external loadings, and discretizations substantially different from the ones used for training."
                    },
                    {
                        "title": "Peridynamic Neural Operators: A Data-Driven Nonlocal Constitutive Model for Complex Material Responses",
                        "abstract": "Neural operators, which can act as implicit solution operators of hidden governing equations, have recently become popular tools for learning the responses of complex real-world physical systems. Nevertheless, most neural operator applications have thus far been data-driven and neglect the intrinsic preservation of fundamental physical laws in data. In this work, we introduce a novel integral neural operator architecture called the Peridynamic Neural Operator (PNO) that learns a nonlocal constitutive law from data. This neural operator provides a forward model in the form of state-based peridynamics, with objectivity and momentum balance laws automatically guaranteed. As applications, we demonstrate the expressivity and efficacy of our model in learning complex material behaviors from both synthetic and experimental data sets. We show that, owing to its ability to capture complex responses, our learned neural operator achieves improved accuracy and efficiency compared to baseline models that use predefined constitutive laws. Moreover, by preserving the essential physical laws within the neural network architecture, the PNO is robust in treating noisy data. The method shows generalizability to different domain configurations, external loadings, and discretizations."
                    },
                    {
                        "title": "Nonlocal Kernel Network (NKN): a Stable and Resolution-Independent Deep Neural Network",
                        "abstract": "Neural operators have recently become popular tools for designing solution maps between function spaces in the form of neural networks. Differently from classical scientific machine learning approaches that learn parameters of a known partial differential equation (PDE) for a single instance of the input parameters at a fixed resolution, neural operators approximate the solution map of a family of PDEs. Despite their success, the uses of neural operators are so far restricted to relatively shallow neural networks and confined to learning hidden governing laws. In this work, we propose a novel nonlocal neural operator, which we refer to as nonlocal kernel network (NKN), that is resolution independent, characterized by deep neural networks, and capable of handling a variety of tasks such as learning governing equations and classifying images. Our NKN stems from the interpretation of the neural network as a discrete nonlocal diffusion reaction equation that, in the limit of infinite layers, is equivalent to a parabolic nonlocal equation, whose stability is analyzed via nonlocal vector calculus. The resemblance with integral forms of neural operators allows NKNs to capture long-range dependencies in the feature space, while the continuous treatment of node-to-node interactions makes NKNs resolution independent. The resemblance with neural ODEs, reinterpreted in a nonlocal sense, and the stable network dynamics between layers allow for generalization of NKN's optimal parameters from shallow to deep networks. This fact enables the use of shallow-to-deep initialization techniques. Our tests show that NKNs outperform baseline methods in both learning governing equations and image classification tasks and generalize well to different resolutions and depths."
                    },
                    {
                        "title": "Bayesian Nonlocal Operator Regression (BNOR): A Data-Driven Learning Framework of Nonlocal Models with Uncertainty Quantification",
                        "abstract": "We consider the problem of modeling heterogeneous materials where micro-scale dynamics and interactions affect global behavior. In the presence of heterogeneities in material microstructure it is often impractical, if not impossible, to provide quantitative characterization of material response. The goal of this work is to develop a Bayesian framework for uncertainty quantification (UQ) in material response prediction when using nonlocal models. Our approach combines the nonlocal operator regression (NOR) technique and Bayesian inference. Specifically, we use a Markov chain Monte Carlo (MCMC) method to sample the posterior probability distribution on parameters involved in the nonlocal constitutive law, and associated modeling discrepancies relative to higher fidelity computations. As an application, we consider the propagation of stress waves through a one-dimensional heterogeneous bar with randomly generated microstructure. Several numerical tests illustrate the construction, enabling UQ in nonlocal model predictions. Although nonlocal models have become popular means for homogenization, their statistical calibration with respect to high-fidelity models has not been presented before. This work is a first step towards statistical characterization of nonlocal model discrepancy in the context of homogenization."
                    },
                    {
                        "title": "Towards a unified nonlocal, peridynamics framework for the coarse-graining of molecular dynamics data with fractures",
                        "abstract": "Molecular dynamics (MD) has served as a powerful tool for designing materials with reduced reliance on laboratory testing. However, the use of MD directly to treat the deformation and failure of materials at the mesoscale is still largely beyond reach. Herein, we propose a learning framework to extract a peridynamic model as a mesoscale continuum surrogate from MD simulated material fracture datasets. Firstly, we develop a novel coarse-graining method, to automatically handle the material fracture and its corresponding discontinuities in MD displacement dataset. Inspired by the Weighted Essentially Non-Oscillatory scheme, the key idea lies at an adaptive procedure to automatically choose the locally smoothest stencil, then reconstruct the coarse-grained material displacement field as piecewise smooth solutions containing discontinuities. Then, based on the coarse-grained MD data, a two-phase optimization-based learning approach is proposed to infer the optimal peridynamics model with damage criterion. In the first phase, we identify the optimal nonlocal kernel function from datasets without material damage, to capture the material stiffness properties. Then, in the second phase, the material damage criterion is learnt as a smoothed step function from the data with fractures. As a result, a peridynamics surrogate is obtained. Our peridynamics surrogate model can be employed in further prediction tasks with different grid resolutions from training, and hence allows for substantial reductions in computational cost compared with MD. We illustrate the efficacy of the proposed approach with several numerical tests for single layer graphene. Our tests show that the proposed data-driven model is robust and generalizable: it is capable in modeling the initialization and growth of fractures under discretization and loading settings that are different from the ones used during training."
                    }
                ]
            }
        }
    },
    "2405.15682": {
        "paper_data": {
            "title": "The Road Less Scheduled",
            "url": "http://arxiv.org/abs/2405.15682v3",
            "arxiv_id": "2405.15682",
            "authors": [
                "Aaron Defazio",
                "Xingyu Alice Yang",
                "Harsh Mehta",
                "Konstantin Mishchenko",
                "Ahmed Khaled",
                "Ashok Cutkosky"
            ],
            "abstract": "Existing learning rate schedules that do not require specification of the optimization stopping step T are greatly out-performed by learning rate schedules that depend on T. We propose an approach that avoids the need for this stopping time by eschewing the use of schedules entirely, while exhibiting state-of-the-art performance compared to schedules across a wide family of problems ranging from convex problems to large-scale deep learning problems. Our Schedule-Free approach introduces no additional hyper-parameters over standard optimizers with momentum. Our method is a direct consequence of a new theory we develop that unifies scheduling and iterate averaging. An open source implementation of our method is available (https://github.com/facebookresearch/schedule_free).",
            "introduction": "   1 Introduction  The theory of optimization, as applied in machine learning, has been successful at providing precise, prescriptive results for many problems. However, even in the simplest setting of stochastic gradient descent (SGD) applied to convex Lipschitz functions, there are glaring gaps between what our current theory prescribes and the methods used in practice.   Consider the stochastic gradient descent (SGD) step with step size \u03b3>0\ud835\udefe0\\gamma>0italic_\u03b3 > 0, zt+1=zt\u2212\u03b3\u2062gtsubscript\ud835\udc67\ud835\udc611subscript\ud835\udc67\ud835\udc61\ud835\udefesubscript\ud835\udc54\ud835\udc61z_{t+1}=z_{t}-\\gamma g_{t}italic_z start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT = italic_z start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT - italic_\u03b3 italic_g start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT where gtsubscript\ud835\udc54\ud835\udc61g_{t}italic_g start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT is the stochastic (sub-)gradient at step t\ud835\udc61titalic_t (formally defined in Section\u00a01.2) of a convex Lipschitz function f\ud835\udc53fitalic_f. Although standard practice for many classes of problems, classical convergence theory suggests that the expected loss of this z\ud835\udc67zitalic_z sequence is suboptimal, and that the Polyak-Ruppert (PR) average x\ud835\udc65xitalic_x of the sequence should be returned instead (Polyak,, 1990; Ruppert,, 1988):    zt+1subscript\ud835\udc67\ud835\udc611\\displaystyle z_{t+1}italic_z start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT =zt\u2212\u03b3\u2062gtabsentsubscript\ud835\udc67\ud835\udc61\ud835\udefesubscript\ud835\udc54\ud835\udc61\\displaystyle=z_{t}-\\gamma g_{t}= italic_z start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT - italic_\u03b3 italic_g start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT     xt+1subscript\ud835\udc65\ud835\udc611\\displaystyle x_{t+1}italic_x start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT =(1\u2212ct+1)\u2062xt+ct+1\u2062zt+1,absent1subscript\ud835\udc50\ud835\udc611subscript\ud835\udc65\ud835\udc61subscript\ud835\udc50\ud835\udc611subscript\ud835\udc67\ud835\udc611\\displaystyle=\\left(1-c_{t+1}\\right)x_{t}+c_{t+1}z_{t+1},= ( 1 - italic_c start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT ) italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT + italic_c start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT italic_z start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT ,    where using ct+1=1/(t+1)subscript\ud835\udc50\ud835\udc6111\ud835\udc611c_{t+1}=1/(t+1)italic_c start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT = 1 / ( italic_t + 1 ) results in xt=1T\u2062\u2211t=1Tztsubscript\ud835\udc65\ud835\udc611\ud835\udc47superscriptsubscript\ud835\udc611\ud835\udc47subscript\ud835\udc67\ud835\udc61x_{t}=\\frac{1}{T}\\sum_{t=1}^{T}z_{t}italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT = divide start_ARG 1 end_ARG start_ARG italic_T end_ARG \u2211 start_POSTSUBSCRIPT italic_t = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_T end_POSTSUPERSCRIPT italic_z start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT. Despite their theoretical optimality, PR averages give much worse results in practice than using the last-iterate of SGD (Figure\u00a02) \u2014 a folk-law result in the field of optimization, and a large theory-practice gap that is often attributed to the mismatch between this simplified problem class and the complexity of problems addressed in practice.   Recently, Zamani and Glineur, (2023) and Defazio et\u00a0al., (2023) showed that the exact worst-case optimal rates can be achieved via carefully chosen learning rate sequences (also known as schedules) alone, without the use of averaging. This result suggests that schedules have, in some sense, the same role to play as PR averaging in optimization. However, schedules have a critical disadvantage: they require setting the optimization stopping time T\ud835\udc47Titalic_T in advance.   Motivated by the theory-practice gap for Polyak-Ruppert averaging, we ask the following question:  Do there exist iterate averaging approaches that match the empirical performance of learning rate schedules, without sacrificing theoretical guarantees?  By developing a new link between averaging and learning rate sequences, we introduce a new approach to averaging that maintains the worst-case convergence rate theory of PR averaging, while matching and often exceeding the performance of schedule-based approaches \u2013 firmly answering this question in the affirmative.    1.1 Summary of Results    \u2022  Our approach does not require the stopping time T\ud835\udc47Titalic_T to be known or set in advance. It closely tracks the Pareto frontier of loss versus training time during a single training run (Figure\u00a01), while requiring no additional hyper-parameters over the base SGD or Adam optimizer.    \u2022  Our approach uses an alternative form of momentum. This form has appealing theoretical properties: it is worst case optimal for any choice of the momentum parameter in the convex Lipschitz",
            "references": [
                {
                    "title": "When, Why and How Much? Adaptive Learning Rate Scheduling by Refinement",
                    "abstract": "Learning rate schedules used in practice bear little resemblance to those recommended by theory. We close much of this theory/practice gap, and as a consequence are able to derive new problem-adaptive learning rate schedules. Our key technical contribution is a refined analysis of learning rate schedules for a wide class of optimization algorithms (including SGD). In contrast to most prior works that study the convergence of the average iterate, we study the last iterate, which is what most people use in practice. When considering only worst-case analysis, our theory predicts that the best choice is the linear decay schedule: a popular choice in practice that sets the stepsize proportionally to $1 - t/T$, where $t$ is the current iteration and $T$ is the total number of steps. To go beyond this worst-case analysis, we use the observed gradient norms to derive schedules refined for any particular task. These refined schedules exhibit learning rate warm-up and rapid learning rate annealing near the end of training. Ours is the first systematic approach to automatically yield both of these properties. We perform the most comprehensive evaluation of learning rate schedules to date, evaluating across 10 diverse deep learning problems, a series of LLMs, and a suite of logistic regression problems. We validate that overall, the linear-decay schedule matches or outperforms all commonly used default schedules including cosine annealing, and that our schedule refinement method gives further improvements."
                },
                {
                    "title": "Exact convergence rate of the last iterate in subgradient methods",
                    "abstract": "We study the convergence of the last iterate in subgradient methods applied to the minimization of a nonsmooth convex function with bounded subgradients. We first introduce a proof technique that generalizes the standard analysis of subgradient methods. It is based on tracking the distance between the current iterate and a different reference point at each iteration. Using this technique, we obtain the exact worst-case convergence rate for the objective accuracy of the last iterate of the projected subgradient method with either constant step sizes or constant step lengths. Tightness is shown with a worst-case instance matching the established convergence rate. We also derive the value of the optimal constant step size when performing $N$ iterations, for which we find that the last iterate accuracy is smaller than $B R \\sqrt{1+\\log(N)/4}/{\\sqrt{N+1}}$ %$\\frac{B R \\log N}{\\sqrt{N+1}}$ , where $B$ is a bound on the subgradient norm and $R$ is a bound on the distance between the initial iterate and a minimizer. Finally, we introduce a new optimal subgradient method that achieves the best possible last-iterate accuracy after a given number $N$ of iterations. Its convergence rate ${B R}/{\\sqrt{N+1}}$ matches exactly the lower bound on the performance of any black-box method on the considered problem class. We also show that there is no universal sequence of step sizes that simultaneously achieves this optimal rate at each iteration, meaning that the dependence of the step size sequence in $N$ is unavoidable."
                },
                {
                    "title": "Benchmarking Neural Network Training Algorithms",
                    "abstract": "Training algorithms, broadly construed, are an essential part of every deep learning pipeline. Training algorithm improvements that speed up training across a wide variety of workloads (e.g., better update rules, tuning protocols, learning rate schedules, or data selection schemes) could save time, save computational resources, and lead to better, more accurate, models. Unfortunately, as a community, we are currently unable to reliably identify training algorithm improvements, or even determine the state-of-the-art training algorithm. In this work, using concrete experiments, we argue that real progress in speeding up training requires new benchmarks that resolve three basic challenges faced by empirical comparisons of training algorithms: (1) how to decide when training is complete and precisely measure training time, (2) how to handle the sensitivity of measurements to exact workload details, and (3) how to fairly compare algorithms that require hyperparameter tuning. In order to address these challenges, we introduce a new, competitive, time-to-result benchmark using multiple workloads running on fixed hardware, the AlgoPerf: Training Algorithms benchmark. Our benchmark includes a set of workload variants that make it possible to detect benchmark submissions that are more robust to workload changes than current widely-used methods. Finally, we evaluate baseline submissions constructed using various optimizers that represent current practice, as well as other optimizers that have recently received attention in the literature. These baseline results collectively demonstrate the feasibility of our benchmark, show that non-trivial gaps between methods exist, and set a provisional state-of-the-art for future benchmark submissions to try and surpass."
                },
                {
                    "title": "Early Weight Averaging meets High Learning Rates for LLM Pre-training",
                    "abstract": "Training Large Language Models (LLMs) incurs significant cost; hence, any strategy that accelerates model convergence is helpful. In this paper, we investigate the ability of a simple idea checkpoint averaging along the trajectory of a training run to improve both convergence and generalization quite early on during training. Here we show that models trained with high learning rates observe higher gains due to checkpoint averaging. Furthermore, these gains are amplified when checkpoints are sampled with considerable spacing in training steps. Our training recipe outperforms conventional training and popular checkpoint averaging baselines such as exponential moving average (EMA) and stochastic moving average (SWA). We evaluate our training recipe by pre-training LLMs, where high learning rates are inherently preferred due to extremely large batch sizes. Specifically, we pre-trained nanoGPT-2 models of varying sizes, small (125M), medium (335M), and large (770M)on the OpenWebText dataset, comprised of 9B tokens. Additionally, we present results for publicly available Pythia LLMs, ranging from 1B to 12B, which were trained on the PILE-deduped dataset containing 207B tokens."
                },
                {
                    "title": "Scaling Vision Transformers to 22 Billion Parameters",
                    "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                },
                {
                    "title": "Learning-Rate-Free Learning by D-Adaptation",
                    "abstract": "D-Adaptation is an approach to automatically setting the learning rate which asymptotically achieves the optimal rate of convergence for minimizing convex Lipschitz functions, with no back-tracking or line searches, and no additional function value or gradient evaluations per step. Our approach is the first hyper-parameter free method for this class without additional multiplicative log factors in the convergence rate. We present extensive experiments for SGD and Adam variants of our method, where the method automatically matches hand-tuned learning rates across more than a dozen diverse machine learning problems, including large-scale vision and language problems. An open-source implementation is available."
                },
                {
                    "title": "Stop Wasting My Time! Saving Days of ImageNet and BERT Training with Latest Weight Averaging",
                    "abstract": "Training vision or language models on large datasets can take days, if not weeks. We show that averaging the weights of the k latest checkpoints, each collected at the end of an epoch, can speed up the training progression in terms of loss and accuracy by dozens of epochs, corresponding to time savings up to ~68 and ~30 GPU hours when training a ResNet50 on ImageNet and RoBERTa-Base model on WikiText-103, respectively. We also provide the code and model checkpoint trajectory to reproduce the results and facilitate research on reusing historical weights for faster convergence."
                },
                {
                    "title": "Fast Benchmarking of Accuracy vs. Training Time with Cyclic Learning Rates",
                    "abstract": "Benchmarking the tradeoff between neural network accuracy and training time is computationally expensive. Here we show how a multiplicative cyclic learning rate schedule can be used to construct a tradeoff curve in a single training run. We generate cyclic tradeoff curves for combinations of training methods such as Blurpool, Channels Last, Label Smoothing and MixUp, and high-light how these cyclic tradeoff curves can be used to evaluate the effects of algorithmic choices on network training ef\ufb01ciency."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "Rethinking \"Batch\" in BatchNorm",
                    "abstract": "BatchNorm is a critical building block in modern convolutional neural networks. Its unique property of operating on \"batches\" instead of individual samples introduces significantly different behaviors from most other operations in deep learning. As a result, it leads to many hidden caveats that can negatively impact model's performance in subtle ways. This paper thoroughly reviews such problems in visual recognition tasks, and shows that a key to address them is to rethink different choices in the concept of \"batch\" in BatchNorm. By presenting these caveats and their mitigations, we hope this review can help researchers use BatchNorm more effectively."
                },
                {
                    "title": "Adaptivity without Compromise: A Momentumized, Adaptive, Dual Averaged Gradient Method for Stochastic Optimization",
                    "abstract": "We introduce MADGRAD, a novel optimization method in the family of AdaGrad adaptive gradient methods. MADGRAD shows excellent performance on deep learning optimization problems from multiple fields, including classification and image-to-image tasks in vision, and recurrent and bidirectionally-masked models in natural language processing. For each of these tasks, MADGRAD matches or outperforms both SGD and ADAM in test set performance, even on problems for which adaptive methods normally perform poorly."
                },
                {
                    "title": "Momentum via Primal Averaging: Theoretical Insights and Learning Rate Schedules for Non-Convex Optimization",
                    "abstract": "Momentum methods are now used pervasively within the machine learning community for training non-convex models such as deep neural networks. Empirically, they out perform traditional stochastic gradient descent (SGD) approaches. In this work we develop a Lyapunov analysis of SGD with momentum (SGD+M), by utilizing a equivalent rewriting of the method known as the stochastic primal averaging (SPA) form. This analysis is much tighter than previous theory in the non-convex case, and due to this we are able to give precise insights into when SGD+M may out-perform SGD, and what hyper-parameter schedules will work and why."
                },
                {
                    "title": "The Power of Factorial Powers: New Parameter settings for (Stochastic) Optimization",
                    "abstract": "The convergence rates for convex and non-convex optimization methods depend on the choice of a host of constants, including step sizes, Lyapunov function constants and momentum constants. In this work we propose the use of factorial powers as a flexible tool for defining constants that appear in convergence proofs. We list a number of remarkable properties that these sequences enjoy, and show how they can be applied to convergence proofs to simplify or improve the convergence rates of the momentum method, accelerated gradient and the stochastic variance reduced method (SVRG)."
                },
                {
                    "title": "Conformer: Convolution-augmented Transformer for Speech Recognition",
                    "abstract": "Recently Transformer and Convolution neural network (CNN) based models have shown promising results in Automatic Speech Recognition (ASR), outperforming Recurrent neural networks (RNNs). Transformer models are good at capturing content-based global interactions, while CNNs exploit local features effectively. In this work, we achieve the best of both worlds by studying how to combine convolution neural networks and transformers to model both local and global dependencies of an audio sequence in a parameter-efficient way. To this regard, we propose the convolution-augmented transformer for speech recognition, named Conformer. Conformer significantly outperforms the previous Transformer and CNN based models achieving state-of-the-art accuracies. On the widely used LibriSpeech benchmark, our model achieves WER of 2.1%/4.3% without using a language model and 1.9%/3.9% with an external language model on test/testother. We also observe competitive performance of 2.7%/6.3% with a small model of only 10M parameters."
                },
                {
                    "title": "Open Graph Benchmark: Datasets for Machine Learning on Graphs",
                    "abstract": "We present the Open Graph Benchmark (OGB), a diverse set of challenging and realistic benchmark datasets to facilitate scalable, robust, and reproducible graph machine learning (ML) research. OGB datasets are large-scale (up to 100+ million nodes and 1+ billion edges), encompass multiple important graph ML tasks, and cover a diverse range of domains, ranging from social and information networks to biological networks, molecular graphs, source code ASTs, and knowledge graphs. For each dataset, we provide a unified evaluation protocol using meaningful application-specific data splits and evaluation metrics. In addition to building the datasets, we also perform extensive benchmark experiments for each dataset. Our experiments suggest that OGB datasets present significant challenges of scalability to large-scale graphs and out-of-distribution generalization under realistic data splits, indicating fruitful opportunities for future research. Finally, OGB provides an automated end-to-end graph ML pipeline that simplifies and standardizes the process of graph data loading, experimental setup, and model evaluation. OGB will be regularly updated and welcomes inputs from the community. OGB datasets as well as data loaders, evaluation scripts, baseline code, and leaderboards are publicly available at this https URL ."
                },
                {
                    "title": "Primal Averaging: A New Gradient Evaluation Step to Attain the Optimal Individual Convergence",
                    "abstract": "Many well-known first-order gradient methods have been extended to cope with large-scale composite problems, which often arise as a regularized empirical risk minimization in machine learning. However, their optimal convergence is attained only in terms of the weighted average of past iterative solutions. How to make the individual convergence of stochastic gradient descent (SGD) optimal, especially for strongly convex problems has now become a challenging problem in the machine learning community. On the other hand, Nesterov\u2019s recent weighted averaging strategy succeeds in achieving the optimal individual convergence of dual averaging (DA) but it fails in the basic mirror descent (MD). In this paper, a new primal averaging (PA) gradient operation step is presented, in which the gradient evaluation is imposed on the weighted average of all past iterative solutions. We prove that simply modifying the gradient operation step in MD by PA strategy suffices to recover the optimal individual rate for general convex problems. Along this line, the optimal individual rate of convergence for strongly convex problems can also be achieved by imposing the strong convexity on the gradient operation step. Furthermore, we extend PA-MD to solve regularized nonsmooth learning problems in the stochastic setting, which reveals that PA strategy is a simple yet effective extra step toward the optimal individual convergence of SGD. Several real experiments on sparse learning and SVM problems verify the correctness of our theoretical analysis."
                },
                {
                    "title": "A Modern Introduction to Online Learning",
                    "abstract": "In this monograph, I introduce the basic concepts of Online Learning through a modern view of Online Convex Optimization. Here, online learning refers to the framework of regret minimization under worst-case assumptions. I present first-order and second-order algorithms for online learning with convex losses, in Euclidean and non-Euclidean settings. All the algorithms are clearly presented as instantiation of Online Mirror Descent or Follow-The-Regularized-Leader and their variants. Particular attention is given to the issue of tuning the parameters of the algorithms and learning in unbounded domains, through adaptive and parameter-free online learning algorithms. Non-convex losses are dealt through convex surrogate losses and through randomization. The bandit setting is also briefly discussed, touching on the problem of adversarial and stochastic multi-armed bandits. These notes do not require prior knowledge of convex analysis and all the required mathematical tools are rigorously explained. Moreover, all the proofs have been carefully chosen to be as simple and as short as possible."
                },
                {
                    "title": "UniXGrad: A Universal, Adaptive Algorithm with Optimal Guarantees for Constrained Optimization",
                    "abstract": "We propose a novel adaptive, accelerated algorithm for the stochastic constrained convex optimization setting.Our method, which is inspired by the Mirror-Prox method, \\emph{simultaneously} achieves the optimal rates for smooth/non-smooth problems with either deterministic/stochastic first-order oracles. This is done without any prior knowledge of the smoothness nor the noise properties of the problem. To the best of our knowledge, this is the first adaptive, unified algorithm that achieves the optimal rates in the constrained setting. We demonstrate the practical performance of our framework through extensive numerical experiments."
                },
                {
                    "title": "Lookahead Optimizer: k steps forward, 1 step back",
                    "abstract": "The vast majority of successful deep neural networks are trained using variants of stochastic gradient descent (SGD) algorithms. Recent attempts to improve SGD can be broadly categorized into two approaches: (1) adaptive learning rate schemes, such as AdaGrad and Adam, and (2) accelerated schemes, such as heavy-ball and Nesterov momentum. In this paper, we propose a new optimization algorithm, Lookahead, that is orthogonal to these previous approaches and iteratively updates two sets of weights. Intuitively, the algorithm chooses a search direction by looking ahead at the sequence of fast weights generated by another optimizer. We show that Lookahead improves the learning stability and lowers the variance of its inner optimizer with negligible computation and memory cost. We empirically demonstrate Lookahead can significantly improve the performance of SGD and Adam, even with their default hyperparameter settings on ImageNet, CIFAR-10/100, neural machine translation, and Penn Treebank."
                },
                {
                    "title": "Deep Learning Recommendation Model for Personalization and Recommendation Systems",
                    "abstract": "With the advent of deep learning, neural network-based recommendation models have emerged as an important tool for tackling personalization and recommendation tasks. These networks differ significantly from other deep learning networks due to their need to handle categorical features and are not well studied or understood. In this paper, we develop a state-of-the-art deep learning recommendation model (DLRM) and provide its implementation in both PyTorch and Caffe2 frameworks. In addition, we design a specialized parallelization scheme utilizing model parallelism on the embedding tables to mitigate memory constraints while exploiting data parallelism to scale-out compute from the fully-connected layers. We compare DLRM against existing recommendation models and characterize its performance on the Big Basin AI platform, demonstrating its usefulness as a benchmark for future algorithmic experimentation and system co-design."
                },
                {
                    "title": "fastMRI: An Open Dataset and Benchmarks for Accelerated MRI",
                    "abstract": "Accelerating Magnetic Resonance Imaging (MRI) by taking fewer measurements has the potential to reduce medical costs, minimize stress to patients and make MRI possible in applications where it is currently prohibitively slow or expensive. We introduce the fastMRI dataset, a large-scale collection of both raw MR measurements and clinical MR images, that can be used for training and evaluation of machine-learning approaches to MR image reconstruction. By introducing standardized evaluation criteria and a freely-accessible dataset, our goal is to help the community make rapid advances in the state of the art for MR image reconstruction. We also provide a self-contained introduction to MRI for machine learning researchers with no medical imaging background."
                },
                {
                    "title": "Relational inductive biases, deep learning, and graph networks",
                    "abstract": "Artificial intelligence (AI) has undergone a renaissance recently, making major progress in key domains such as vision, language, control, and decision-making. This has been due, in part, to cheap data and cheap compute resources, which have fit the natural strengths of deep learning. However, many defining characteristics of human intelligence, which developed under much different pressures, remain out of reach for current approaches. In particular, generalizing beyond one's experiences--a hallmark of human intelligence from infancy--remains a formidable challenge for modern AI. \nThe following is part position paper, part review, and part unification. We argue that combinatorial generalization must be a top priority for AI to achieve human-like abilities, and that structured representations and computations are key to realizing this objective. Just as biology uses nature and nurture cooperatively, we reject the false choice between \"hand-engineering\" and \"end-to-end\" learning, and instead advocate for an approach which benefits from their complementary strengths. We explore how using relational inductive biases within deep learning architectures can facilitate learning about entities, relations, and rules for composing them. We present a new building block for the AI toolkit with a strong relational inductive bias--the graph network--which generalizes and extends various approaches for neural networks that operate on graphs, and provides a straightforward interface for manipulating structured knowledge and producing structured behaviors. We discuss how graph networks can support relational reasoning and combinatorial generalization, laying the foundation for more sophisticated, interpretable, and flexible patterns of reasoning. As a companion to this paper, we have released an open-source software library for building graph networks, with demonstrations of how to use them in practice."
                },
                {
                    "title": "Averaging Weights Leads to Wider Optima and Better Generalization",
                    "abstract": "Deep neural networks are typically trained by optimizing a loss function with an SGD variant, in conjunction with a decaying learning rate, until convergence. We show that simple averaging of multiple points along the trajectory of SGD, with a cyclical or constant learning rate, leads to better generalization than conventional training. We also show that this Stochastic Weight Averaging (SWA) procedure finds much flatter solutions than SGD, and approximates the recent Fast Geometric Ensembling (FGE) approach with a single model. Using SWA we achieve notable improvement in test accuracy over conventional SGD training on a range of state-of-the-art residual networks, PyramidNets, DenseNets, and Shake-Shake networks on CIFAR-10, CIFAR-100, and ImageNet. In short, SWA is extremely easy to implement, improves generalization, and has almost no computational overhead."
                },
                {
                    "title": "On the Convergence of Adam and Beyond",
                    "abstract": "Several recently proposed stochastic optimization methods that have been successfully used in training deep networks such as RMSProp, Adam, Adadelta, Nadam are based on using gradient updates scaled by square roots of exponential moving averages of squared past gradients. In many applications, e.g. learning with large output spaces, it has been empirically observed that these algorithms fail to converge to an optimal solution (or a critical point in nonconvex settings). We show that one cause for such failures is the exponential moving average used in the algorithms. We provide an explicit example of a simple convex optimization setting where Adam does not converge to the optimal solution, and describe the precise problems with the previous analysis of Adam algorithm. Our analysis suggests that the convergence issues can be fixed by endowing such algorithms with `long-term memory' of past gradients, and propose new variants of the Adam algorithm which not only fix the convergence issues but often also lead to improved empirical performance."
                },
                {
                    "title": "A Modular Analysis of Adaptive (Non-)Convex Optimization: Optimism, Composite Objectives, and Variational Bounds",
                    "abstract": "Recently, much work has been done on extending the scope of online learning and incremental stochastic optimization algorithms. In this paper we contribute to this effort in two ways: First, based on a new regret decomposition and a generalization of Bregman divergences, we provide a self-contained, modular analysis of the two workhorses of online learning: (general) adaptive versions of Mirror Descent (MD) and the Follow-the-Regularized-Leader (FTRL) algorithms. The analysis is done with extra care so as not to introduce assumptions not needed in the proofs and allows to combine, in a straightforward way, different algorithmic ideas (e.g., adaptivity, optimism, implicit updates) and learning settings (e.g., strongly convex or composite objectives). This way we are able to reprove, extend and refine a large body of the literature, while keeping the proofs concise. The second contribution is a byproduct of this careful analysis: We present algorithms with improved variational bounds for smooth, composite objectives, including a new family of optimistic MD algorithms with only one projection step per round. Furthermore, we provide a simple extension of adaptive regret bounds to practically relevant non-convex problem settings with essentially no extra effort."
                },
                {
                    "title": "Findings of the 2017 Conference on Machine Translation (WMT17)",
                    "abstract": "This paper presents the results of the WMT17 shared tasks, which included \nthree machine translation (MT) tasks (news, biomedical, and multimodal), two evaluation tasks (metrics and run-time estimation of MT quality), an automatic post-editing task, a neural MT training task, and a bandit learning task."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Densely Connected Convolutional Networks",
                    "abstract": "Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections&#x2014;one between each layer and its subsequent layer&#x2014;our network has L(L+1)/2 direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters. We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet). DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less memory and computation to achieve high performance. Code and pre-trained models are available at https://github.com/liuzhuang13/DenseNet."
                },
                {
                    "title": "Introduction to Online Convex Optimization",
                    "abstract": "This monograph portrays optimization as a process. In many practical applications the environment is so complex that it is infeasible to lay out a comprehensive theoretical model and use classical algorithmic theory and mathematical optimization. It is necessary as well as beneficial to take a robust approach, by applying an optimization method that learns as one goes along, learning from experience as more aspects of the problem are observed. This view of optimization as a process has become prominent in varied fields and has led to some spectacular success in modeling and systems that are now part of our daily lives."
                },
                {
                    "title": "Sequence-to-Sequence Learning as Beam-Search Optimization",
                    "abstract": "Sequence-to-Sequence (seq2seq) modeling has rapidly become an important general-purpose NLP tool that has proven effective for many text-generation and sequence-labeling tasks. Seq2seq builds on deep neural language modeling and inherits its remarkable accuracy in estimating local, next-word distributions. In this work, we introduce a model and beam-search training scheme, based on the work of Daume III and Marcu (2005), that extends seq2seq to learn global sequence scores. This structured approach avoids classical biases associated with local training and unifies the training loss with the test-time usage, while preserving the proven model architecture of seq2seq and its efficient training approach. We show that our system outperforms a highly-optimized attention-based seq2seq system and other baselines on three different sequence to sequence tasks: word ordering, parsing, and machine translation."
                },
                {
                    "title": "Wide Residual Networks",
                    "abstract": "Deep residual networks were shown to be able to scale up to thousands of layers and still have improving performance. However, each fraction of a percent of improved accuracy costs nearly doubling the number of layers, and so training very deep residual networks has a problem of diminishing feature reuse, which makes these networks very slow to train. To tackle these problems, in this paper we conduct a detailed experimental study on the architecture of ResNet blocks, based on which we propose a novel architecture where we decrease depth and increase width of residual networks. We call the resulting network structures wide residual networks (WRNs) and show that these are far superior over their commonly used thin and very deep counterparts. For example, we demonstrate that even a simple 16-layer-deep wide residual network outperforms in accuracy and efficiency all previous deep residual networks, including thousand-layer-deep networks, achieving new state-of-the-art results on CIFAR, SVHN, COCO, and significant improvements on ImageNet. Our code and models are available at this https URL"
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Rethinking the Inception Architecture for Computer Vision",
                    "abstract": "Convolutional networks are at the core of most state of-the-art computer vision solutions for a wide variety of tasks. Since 2014 very deep convolutional networks started to become mainstream, yielding substantial gains in various benchmarks. Although increased model size and computational cost tend to translate to immediate quality gains for most tasks (as long as enough labeled data is provided for training), computational efficiency and low parameter count are still enabling factors for various use cases such as mobile vision and big-data scenarios. Here we are exploring ways to scale up networks in ways that aim at utilizing the added computation as efficiently as possible by suitably factorized convolutions and aggressive regularization. We benchmark our methods on the ILSVRC 2012 classification challenge validation set demonstrate substantial gains over the state of the art: 21:2% top-1 and 5:6% top-5 error for single frame evaluation using a network with a computational cost of 5 billion multiply-adds per inference and with using less than 25 million parameters. With an ensemble of 4 models and multi-crop evaluation, we report 3:5% top-5 error and 17:3% top-1 error on the validation set and 3:6% top-5 error on the official test set."
                },
                {
                    "title": "Librispeech: An ASR corpus based on public domain audio books",
                    "abstract": "This paper introduces a new corpus of read English speech, suitable for training and evaluating speech recognition systems. The LibriSpeech corpus is derived from audiobooks that are part of the LibriVox project, and contains 1000 hours of speech sampled at 16 kHz. We have made the corpus freely available for download, along with separately prepared language-model training data and pre-built language models. We show that acoustic models trained on LibriSpeech give lower error rate on the Wall Street Journal (WSJ) test sets than models trained on WSJ itself. We are also releasing Kaldi scripts that make it easy to build these systems."
                },
                {
                    "title": "Adam: A Method for Stochastic Optimization",
                    "abstract": "We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
                },
                {
                    "title": "On the importance of initialization and momentum in deep learning",
                    "abstract": "Deep and recurrent neural networks (DNNs and RNNs respectively) are powerful models that were considered to be almost impossible to train using stochastic gradient descent with momentum. In this paper, we show that when stochastic gradient descent with momentum uses a well-designed random initialization and a particular type of slowly increasing schedule for the momentum parameter, it can train both DNNs and RNNs (on datasets with long-term dependencies) to levels of performance that were previously achievable only with Hessian-Free optimization. We find that both the initialization and the momentum are crucial since poorly initialized networks cannot be trained with momentum and well-initialized networks perform markedly worse when the momentum is absent or poorly tuned. \n \nOur success training these models suggests that previous attempts to train deep and recurrent neural networks from random initializations have likely failed due to poor initialization schemes. Furthermore, carefully tuned momentum methods suffice for dealing with the curvature issues in deep and recurrent network training objectives without the need for sophisticated second-order methods."
                },
                {
                    "title": "Non-strongly-convex smooth stochastic approximation with convergence rate O(1/n)",
                    "abstract": "We consider the stochastic approximation problem where a convex function has to be minimized, given only the knowledge of unbiased estimates of its gradients at certain points, a framework which includes machine learning methods based on the minimization of the empirical risk. We focus on problems without strong convexity, for which all previously known algorithms achieve a convergence rate for function values of O(1/\u221an) after n iterations. We consider and analyze two algorithms that achieve a rate of O(1/n) for classical supervised learning problems. For least-squares regression, we show that averaged stochastic gradient descent with constant step-size achieves the desired rate. For logistic regression, this is achieved by a simple novel stochastic gradient algorithm that (a) constructs successive local quadratic approximations of the loss functions, while (b) preserving the same running-time complexity as stochastic gradient descent. For these algorithms, we provide a non-asymptotic analysis of the generalization error (in expectation, and also in high probability for least-squares), and run extensive experiments showing that they often outperform existing approaches."
                },
                {
                    "title": "A simpler approach to obtaining an O(1/t) convergence rate for the projected stochastic subgradient method",
                    "abstract": "In this note, we present a new averaging technique for the projected stochastic subgradient method. By using a weighted average with a weight of t+1 for each iterate w_t at iteration t, we obtain the convergence rate of O(1/t) with both an easy proof and an easy implementation. The new scheme is compared empirically to existing techniques, with similar performance behavior."
                },
                {
                    "title": "Stochastic Gradient Descent for Non-smooth Optimization: Convergence Results and Optimal Averaging Schemes",
                    "abstract": "Stochastic Gradient Descent (SGD) is one of the simplest and most popular stochastic optimization methods. While it has already been theoretically studied for decades, the classical analysis usually required non-trivial smoothness assumptions, which do not apply to many modern applications of SGD with non-smooth objective functions such as support vector machines. In this paper, we investigate the performance of SGD without such smoothness assumptions, as well as a running average scheme to convert the SGD iterates to a solution with optimal optimization accuracy. In this framework, we prove that after T rounds, the suboptimality of the last SGD iterate scales as O(log(T)/\u221aT) for non-smooth convex objective functions, and O(log(T)/T) in the non-smooth strongly convex case. To the best of our knowledge, these are the first bounds of this kind, and almost match the minimax-optimal rates obtainable by appropriate averaging schemes. We also propose a new and simple averaging scheme, which not only attains optimal rates, but can also be easily computed on-the-fly (in contrast, the suffix averaging scheme proposed in Rakhlin et al. (2011) is not as simple to implement). Finally, we provide some experimental illustrations."
                },
                {
                    "title": "Online Learning with Predictable Sequences",
                    "abstract": "We present methods for online linear optimization that take advantage of benign (as opposed to worst-case) sequences. Specically if the sequence encountered by the learner is described well by a known \\predictable process\", the algorithms presented enjoy tighter bounds as compared to the typical worst case bounds. Additionally, the methods achieve the usual worst-case regret bounds if the sequence is not benign. Our approach can be seen as a way of adding prior knowledge about the sequence within the paradigm of online learning. The setting is shown to encompass partial and side information. Variance and path-length bounds [11, 9] can be seen as particular examples of online learning with simple predictable sequences. We further extend our methods and results to include competing with a set of possible predictable processes (models), that is \\learning\" the predictable process itself concurrently with using it to obtain better regret guarantees. We show that such model selection is possible under various assumptions on the available feedback. Our results suggest a promising direction of further research with potential applications to stock market and time series prediction."
                },
                {
                    "title": "Making Gradient Descent Optimal for Strongly Convex Stochastic Optimization",
                    "abstract": "Stochastic gradient descent (SGD) is a simple and popular method to solve stochastic optimization problems which arise in machine learning. For strongly convex problems, its convergence rate was known to be O(log(T)/T), by running SGD for T iterations and returning the average point. However, recent results showed that using a different algorithm, one can get an optimal O(1/T) rate. This might lead one to believe that standard SGD is suboptimal, and maybe should even be replaced as a method of choice. In this paper, we investigate the optimality of SGD in a stochastic setting. We show that for smooth problems, the algorithm attains the optimal O(1/T) rate. However, for non-smooth problems, the convergence rate with averaging might really be \u03a9(log(T)/T), and this is not just an artifact of the analysis. On the flip side, we show that a simple modification of the averaging step suffices to recover the O(1/T) rate, and no other change of the algorithm is necessary. We also present experimental results which support our findings, and point out open problems."
                },
                {
                    "title": "Adaptive Subgradient Methods for Online Learning and Stochastic Optimization",
                    "abstract": "We present a new family of subgradient methods that dynamically incorporate knowledge of the geometry of the data observed in earlier iterations to perform more informative gradient-based learning. Metaphorically, the adaptation allows us to find needles in haystacks in the form of very predictive but rarely seen features. Our paradigm stems from recent advances in stochastic optimization and online learning which employ proximal functions to control the gradient steps of the algorithm. We describe and analyze an apparatus for adaptively modifying the proximal function, which significantly simplifies setting a learning rate and results in regret guarantees that are provably as good as the best proximal function that can be chosen in hindsight. We give several efficient algorithms for empirical risk minimization problems with common and important regularization functions and domain constraints. We experimentally study our theoretical analysis and show that adaptive subgradient methods outperform state-of-the-art, yet non-adaptive, subgradient algorithms."
                },
                {
                    "title": "Online Convex Programming and Generalized Infinitesimal Gradient Ascent",
                    "abstract": "Convex programming involves a convex set F \u2286 Rn and a convex cost function c : F \u2192 R. The goal of convex programming is to find a point in F which minimizes c. In online convex programming, the convex set is known in advance, but in each step of some repeated optimization problem, one must select a point in F before seeing the cost function for that step. This can be used to model factory production, farm production, and many other industrial optimization problems where one is unaware of the value of the items produced until they have already been constructed. We introduce an algorithm for this domain. We also apply this algorithm to repeated games, and show that it is really a generalization of infinitesimal gradient ascent, and the results here imply that generalized infinitesimal gradient ascent (GIGA) is universally consistent."
                },
                {
                    "title": "On the generalization ability of on-line learning algorithms",
                    "abstract": "In this paper, it is shown how to extract a hypothesis with small risk from the ensemble of hypotheses generated by an arbitrary on-line learning algorithm run on an independent and identically distributed (i.i.d.) sample of data. Using a simple large deviation argument, we prove tight data-dependent bounds for the risk of this hypothesis in terms of an easily computable statistic M/sub n/ associated with the on-line performance of the ensemble. Via sharp pointwise bounds on M/sub n/, we then obtain risk tail bounds for kernel perceptron algorithms in terms of the spectrum of the empirical kernel matrix. These bounds reveal that the linear hypotheses found via our approach achieve optimal tradeoffs between hinge loss and margin size over the class of all linear functions, an issue that was left open by previous results. A distinctive feature of our approach is that the key tools for our analysis come from the model of prediction of individual sequences; i.e., a model making no probabilistic assumptions on the source generating the data. In fact, these tools turn out to be so powerful that we only need very elementary statistical facts to obtain our final risk bounds."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Report on the 11th IWSLT evaluation campaign",
                    "abstract": "The paper overviews the 11th evaluation campaign organized by the IWSLT workshop. The 2014 evaluation offered multiple tracks on lecture transcription and translation based on the TED Talks corpus. In particular, this year IWSLT included three automatic speech recognition tracks, on English, German and Italian, \ufb01ve speech translation tracks, from English to French, English to German, German to English, English to Italian, and Italian to English, and \ufb01ve text translation track, also from English to French, English to German, German to English, English to Italian, and Italian to English. In addition to the of\ufb01cial tracks, speech and text translation optional tracks were offered, globally involving 12 other languages: Arabic, Spanish, Portuguese (B), Hebrew, Chinese, Polish, Persian, Slovenian, Turkish, Dutch, Romanian, Russian. Overall, 21 teams participated in the evaluation, for a total of 76 primary runs submitted. Participants were also asked to submit runs on the 2013 test set (progress test set), in order to measure the progress of systems with respect to the previous year. All runs were evaluated with objective metrics, and submissions for two of the of\ufb01cial text translation tracks were also evaluated with human post-editing."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "math.OC",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nDo there exist iterate averaging approaches that match the empirical performance of learning rate schedules, without sacrificing theoretical guarantees?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for bridging the theory-practice gap in optimization methods used in machine learning. By developing an averaging approach that aligns with the empirical performance of learning rate schedules, we can enhance the effectiveness of optimization algorithms like SGD and Adam. This advancement could lead to more efficient training processes, improved model performance, and a deeper understanding of the dynamics between learning rates and averaging techniques. Ultimately, it could influence future research directions in optimization theory and practical applications in various machine learning tasks.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in reconciling the theoretical optimality of Polyak-Ruppert (PR) averaging with its suboptimal empirical performance compared to last-iterate methods. Naive approaches may fail because they do not account for the complexities of real-world problems, which often deviate from the simplified assumptions of classical convergence theory. Additionally, developing a method that does not require prior knowledge of the stopping time T introduces further complexity, as it necessitates a dynamic adaptation of the averaging process to maintain optimal performance.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either learning rate schedules or averaging techniques in isolation, often overlooking the potential synergies between the two. The lack of a comprehensive framework that integrates these concepts has hindered progress. Additionally, existing solutions have been limited by their reliance on predetermined stopping times, which restricts their applicability in practice. Our approach differs by establishing a new link between averaging and learning rate sequences, allowing for a more flexible and effective optimization strategy.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a new averaging technique that maintains the worst-case convergence rate of PR averaging while matching or exceeding the performance of learning rate schedules. We will utilize a dataset relevant to convex Lipschitz functions and evaluate our approach using standard metrics for convergence and empirical performance. The expected outcomes include improved optimization results without the need for setting a stopping time in advance, as well as a demonstration of the theoretical properties of our alternative momentum form, which is optimal for any choice of the momentum parameter."
            }
        },
        "author_data": {
            "3aed3d1b-f960-4520-a431-eefd83e165a9": {
                "pk": "3aed3d1b-f960-4520-a431-eefd83e165a9",
                "name": "Aaron Defazio",
                "collaborators": [
                    "L\u00e9on Bottou",
                    "Samy Jelassi",
                    "Konstantin Mishchenko",
                    "Thore Graepel",
                    "Robert M. Gower",
                    "Baoyu Zhou",
                    "Lin Xiao",
                    "Tiberio Caetano",
                    "Mark Tygert",
                    "Ashok Cutkosky"
                ],
                "domain": [
                    "Optimization",
                    "Deep Learning",
                    "Machine Learning",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "A Simple Practical Accelerated Method for Finite Sums",
                        "abstract": "We describe a novel optimization method for finite sums (such as empirical risk minimization problems) building on the recently introduced SAGA method. Our method achieves an accelerated convergence rate on strongly convex smooth problems. Our method has only one parameter (a step size), and is radically simpler than other accelerated methods for finite sums. Additionally it can be applied when the terms are non-smooth, yielding a method applicable in many areas where operator splitting methods would traditionally be applied."
                    },
                    {
                        "title": "Offset Sampling Improves Deep Learning based Accelerated MRI Reconstructions by Exploiting Symmetry",
                        "abstract": "Deep learning approaches to accelerated MRI take a matrix of sampled Fourier-space lines as input and produce a spatial image as output. In this work we show that by careful choice of the offset used in the sampling procedure, the symmetries in k-space can be better exploited, producing higher quality reconstructions than given by standard equally-spaced samples or randomized samples motivated by compressed sensing."
                    },
                    {
                        "title": "On the Curved Geometry of Accelerated Optimization",
                        "abstract": "In this work we propose a differential geometric motivation for Nesterov's accelerated gradient method (AGM) for strongly-convex problems. By considering the optimization procedure as occurring on a Riemannian manifold with a natural structure, The AGM method can be seen as the proximal point method applied in this curved space. This viewpoint can also be extended to the continuous time case, where the accelerated gradient method arises from the natural block-implicit Euler discretization of an ODE on the manifold. We provide an analysis of the convergence rate of this ODE for quadratic objectives."
                    },
                    {
                        "title": "Momentum via Primal Averaging: Theoretical Insights and Learning Rate Schedules for Non-Convex Optimization",
                        "abstract": "Momentum methods are now used pervasively within the machine learning community for training non-convex models such as deep neural networks. Empirically, they out perform traditional stochastic gradient descent (SGD) approaches. In this work we develop a Lyapunov analysis of SGD with momentum (SGD+M), by utilizing a equivalent rewriting of the method known as the stochastic primal averaging (SPA) form. This analysis is much tighter than previous theory in the non-convex case, and due to this we are able to give precise insights into when SGD+M may out-perform SGD, and what hyper-parameter schedules will work and why."
                    },
                    {
                        "title": "New Optimisation Methods for Machine Learning",
                        "abstract": "A thesis submitted for the degree of Doctor of Philosophy of The Australian National University.   In this work we introduce several new optimisation methods for problems in machine learning. Our algorithms broadly fall into two categories: optimisation of finite sums and of graph structured objectives. The finite sum problem is simply the minimisation of objective functions that are naturally expressed as a summation over a large number of terms, where each term has a similar or identical weight. Such objectives most often appear in machine learning in the empirical risk minimisation framework in the non-online learning setting. The second category, that of graph structured objectives, consists of objectives that result from applying maximum likelihood to Markov random field models. Unlike the finite sum case, all the non-linearity is contained within a partition function term, which does not readily decompose into a summation.   For the finite sum problem, we introduce the Finito and SAGA algorithms, as well as variants of each.   For graph-structured problems, we take three complementary approaches. We look at learning the parameters for a fixed structure, learning the structure independently, and learning both simultaneously. Specifically, for the combined approach, we introduce a new method for encouraging graph structures with the \"scale-free\" property. For the structure learning problem, we establish SHORTCUT, a O(n^{2.5}) expected time approximate structure learning method for Gaussian graphical models. For problems where the structure is known but the parameters unknown, we introduce an approximate maximum likelihood learning algorithm that is capable of learning a useful subclass of Gaussian graphical models."
                    },
                    {
                        "title": "On the Ineffectiveness of Variance Reduced Optimization for Deep Learning",
                        "abstract": "The application of stochastic variance reduction to optimization has shown remarkable recent theoretical and practical success. The applicability of these techniques to the hard non-convex optimization problems encountered during training of modern deep neural networks is an open problem. We show that naive application of the SVRG technique and related approaches fail, and explore why."
                    },
                    {
                        "title": "Controlling Covariate Shift using Balanced Normalization of Weights",
                        "abstract": "We introduce a new normalization technique that exhibits the fast convergence properties of batch normalization using a transformation of layer weights instead of layer outputs. The proposed technique keeps the contribution of positive and negative weights to the layer output balanced. We validate our method on a set of standard benchmarks including CIFAR-10/100, SVHN and ILSVRC 2012 ImageNet."
                    },
                    {
                        "title": "A Comparison of learning algorithms on the Arcade Learning Environment",
                        "abstract": "Reinforcement learning agents have traditionally been evaluated on small toy problems. With advances in computing power and the advent of the Arcade Learning Environment, it is now possible to evaluate algorithms on diverse and difficult problems within a consistent framework. We discuss some challenges posed by the arcade learning environment which do not manifest in simpler environments. We then provide a comparison of model-free, linear learning algorithms on this challenging problem set."
                    },
                    {
                        "title": "Adaptivity without Compromise: A Momentumized, Adaptive, Dual Averaged Gradient Method for Stochastic Optimization",
                        "abstract": "We introduce MADGRAD, a novel optimization method in the family of AdaGrad adaptive gradient methods. MADGRAD shows excellent performance on deep learning optimization problems from multiple fields, including classification and image-to-image tasks in vision, and recurrent and bidirectionally-masked models in natural language processing. For each of these tasks, MADGRAD matches or outperforms both SGD and ADAM in test set performance, even on problems for which adaptive methods normally perform poorly."
                    },
                    {
                        "title": "The Power of Factorial Powers: New Parameter settings for (Stochastic) Optimization",
                        "abstract": "The convergence rates for convex and non-convex optimization methods depend on the choice of a host of constants, including step sizes, Lyapunov function constants and momentum constants. In this work we propose the use of factorial powers as a flexible tool for defining constants that appear in convergence proofs. We list a number of remarkable properties that these sequences enjoy, and show how they can be applied to convergence proofs to simplify or improve the convergence rates of the momentum method, accelerated gradient and the stochastic variance reduced method (SVRG)."
                    },
                    {
                        "title": "Grad-GradaGrad? A Non-Monotone Adaptive Stochastic Gradient Method",
                        "abstract": "The classical AdaGrad method adapts the learning rate by dividing by the square root of a sum of squared gradients. Because this sum on the denominator is increasing, the method can only decrease step sizes over time, and requires a learning rate scaling hyper-parameter to be carefully tuned. To overcome this restriction, we introduce GradaGrad, a method in the same family that naturally grows or shrinks the learning rate based on a different accumulation in the denominator, one that can both increase and decrease. We show that it obeys a similar convergence rate as AdaGrad and demonstrate its non-monotone adaptation capability with experiments."
                    },
                    {
                        "title": "Beyond Folklore: A Scaling Calculus for the Design and Initialization of ReLU Networks",
                        "abstract": "We propose a system for calculating a \"scaling constant\" for layers and weights of neural networks. We relate this scaling constant to two important quantities that relate to the optimizability of neural networks, and argue that a network that is \"preconditioned\" via scaling, in the sense that all weights have the same scaling constant, will be easier to train. This scaling calculus results in a number of consequences, among them the fact that the geometric mean of the fan-in and fan-out, rather than the fan-in, fan-out, or arithmetic mean, should be used for the initialization of the variance of weights in a neural network. Our system allows for the off-line design & engineering of ReLU neural networks, potentially replacing blind experimentation."
                    },
                    {
                        "title": "Learning-Rate-Free Learning by D-Adaptation",
                        "abstract": "D-Adaptation is an approach to automatically setting the learning rate which asymptotically achieves the optimal rate of convergence for minimizing convex Lipschitz functions, with no back-tracking or line searches, and no additional function value or gradient evaluations per step. Our approach is the first hyper-parameter free method for this class without additional multiplicative log factors in the convergence rate. We present extensive experiments for SGD and Adam variants of our method, where the method automatically matches hand-tuned learning rates across more than a dozen diverse machine learning problems, including large-scale vision and language problems.   An open-source implementation is available."
                    },
                    {
                        "title": "A Graphical Model Formulation of Collaborative Filtering Neighbourhood Methods with Fast Maximum Entropy Training",
                        "abstract": "Item neighbourhood methods for collaborative filtering learn a weighted graph over the set of items, where each item is connected to those it is most similar to. The prediction of a user's rating on an item is then given by that rating of neighbouring items, weighted by their similarity. This paper presents a new neighbourhood approach which we call item fields, whereby an undirected graphical model is formed over the item graph. The resulting prediction rule is a simple generalization of the classical approaches, which takes into account non-local information in the graph, allowing its best results to be obtained when using drastically fewer edges than other neighbourhood approaches. A fast approximate maximum entropy training method based on the Bethe approximation is presented, which uses a simple gradient ascent procedure. When using precomputed sufficient statistics on the Movielens datasets, our method is faster than maximum likelihood approaches by two orders of magnitude."
                    },
                    {
                        "title": "Methods of interpreting error estimates for grayscale image reconstructions",
                        "abstract": "One representation of possible errors in a grayscale image reconstruction is as another grayscale image estimating potentially worrisome differences between the reconstruction and the actual \"ground-truth\" reality. Visualizations and summary statistics can aid in the interpretation of such a representation of error estimates. Visualizations include suitable colorizations of the reconstruction, as well as the obvious \"correction\" of the reconstruction by subtracting off the error estimates. The canonical summary statistic would be the root-mean-square of the error estimates. Numerical examples involving cranial magnetic-resonance imaging clarify the relative merits of the various methods in the context of compressed sensing. Unfortunately, the colorizations appear likely to be too distracting for actual clinical practice, and the root-mean-square gets swamped by background noise in the error estimates. Fortunately, straightforward displays of the error estimates and of the \"corrected\" reconstruction are illuminating, and the root-mean-square improves greatly after mild blurring of the error estimates; the blurring is barely perceptible to the human eye yet smooths away background noise that would otherwise overwhelm the root-mean-square."
                    },
                    {
                        "title": "Dual Averaging is Surprisingly Effective for Deep Learning Optimization",
                        "abstract": "First-order stochastic optimization methods are currently the most widely used class of methods for training deep neural networks. However, the choice of the optimizer has become an ad-hoc rule that can significantly affect the performance. For instance, SGD with momentum (SGD+M) is typically used in computer vision (CV) and Adam is used for training transformer models for Natural Language Processing (NLP). Using the wrong method can lead to significant performance degradation. Inspired by the dual averaging algorithm, we propose Modernized Dual Averaging (MDA), an optimizer that is able to perform as well as SGD+M in CV and as Adam in NLP. Our method is not adaptive and is significantly simpler than Adam. We show that MDA induces a decaying uncentered $L_2$-regularization compared to vanilla SGD+M and hypothesize that this may explain why it works on NLP problems where SGD+M fails."
                    },
                    {
                        "title": "Prodigy: An Expeditiously Adaptive Parameter-Free Learner",
                        "abstract": "We consider the problem of estimating the learning rate in adaptive methods, such as AdaGrad and Adam. We propose Prodigy, an algorithm that provably estimates the distance to the solution $D$, which is needed to set the learning rate optimally. At its core, Prodigy is a modification of the D-Adaptation method for learning-rate-free learning. It improves upon the convergence rate of D-Adaptation by a factor of $O(\\sqrt{\\log(D/d_0)})$, where $d_0$ is the initial estimate of $D$. We test Prodigy on 12 common logistic-regression benchmark datasets, VGG11 and ResNet-50 training on CIFAR10, ViT training on Imagenet, LSTM training on IWSLT14, DLRM training on Criteo dataset, VarNet on Knee MRI dataset, as well as RoBERTa and GPT transformer training on BookWiki. Our experimental results show that our approach consistently outperforms D-Adaptation and reaches test accuracy values close to that of hand-tuned Adam."
                    },
                    {
                        "title": "Mechanic: A Learning Rate Tuner",
                        "abstract": "We introduce a technique for tuning the learning rate scale factor of any base optimization algorithm and schedule automatically, which we call \\textsc{mechanic}. Our method provides a practical realization of recent theoretical reductions for accomplishing a similar goal in online convex optimization. We rigorously evaluate \\textsc{mechanic} on a range of large scale deep learning tasks with varying batch sizes, schedules, and base optimization algorithms. These experiments demonstrate that depending on the problem, \\textsc{mechanic} either comes very close to, matches or even improves upon manual tuning of learning rates."
                    }
                ]
            },
            "c6ee3194-c47b-4b42-9f5b-56d0e854576b": {
                "pk": "c6ee3194-c47b-4b42-9f5b-56d0e854576b",
                "name": "Xingyu Alice Yang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "5f40cc02-2911-4d63-9320-8920baa6a7c7": {
                "pk": "5f40cc02-2911-4d63-9320-8920baa6a7c7",
                "name": "Harsh Mehta",
                "collaborators": [
                    "Ashok Cutkosky",
                    "Suvodip Mukherjee",
                    "Behnam Neyshabur",
                    "Ankit Gupta",
                    "Jeffery Lagarias",
                    "Aaron Defazio",
                    "Steffen Rendle",
                    "Walid Krichene",
                    "Li Zhang",
                    "Abhradeep Thakurta"
                ],
                "domain": [
                    "Machine Learning",
                    "Optimization",
                    "Graph Theory",
                    "Astrophysics"
                ],
                "publications": [
                    {
                        "title": "The L1 norm of the generalized de la Vallee Poussin kernel",
                        "abstract": "Charles de la Vall'ee Poussin defined two different kernels that bear his name. This paper considers the one are a linear combinations of two Fej'er kernels, which are known as the delayed means. We show that the $L^1$ norms are constant in families of delayed means, and determine the exact value"
                    },
                    {
                        "title": "Counting extensions of number fields with Frobenius Galois group",
                        "abstract": "Let $G$ be a Frobenius group with an abelian Frobenius kernel $F$ and let $k$ be a finite extension of $\\mathbb{Q}$. We obtain an upper bound for the number of degree $|F|$ algebraic extensions $K/k$ with Galois group $G$ with the norm of the discriminant $\\mathcal{N}_{k/\\mathbb{Q}}(d_{K/k})$ bounded above by $X$. We extend this method for any group $G$ that has an abelian normal subgroup. If $G$ has an abelian normal subgroup, then we obtain upper bounds for the number of degree $|G|$ extensions $N/k$ with Galois group $G$ with bounded norm of the discriminant. Malle made a conjecture about what the order of magnitude of this quantity should be as the degree of the extension $d$ and underlying Galois group $G$ vary. We show that under the $\\ell$-torsion conjecture, the upper bounds we achieve for certain pairs $d$ and $G$ agree with the prediction of Malle. Unconditionally we show that the upper bound for the number of degree 6 extensions with Galois group $A_4$ also satisfies Malle's weak conjecture."
                    },
                    {
                        "title": "Products of Farey Fractions",
                        "abstract": "The {Farey fractions} $F_n$ of order $n$ consist of all fractions $\\frac{h}{k}$ in lowest terms lying in the closed unit interval and having denominator at most $n$. This paper considers the products $F_n$ of all nonzero Farey fractions of order $n$. It studies their growth measured by $\\log(F_n)$ and their divisibility properties by powers of a fixed prime, given by $ord_p(F_n)$, as a function of $n$. The growth of $\\log(F_n)$ is related to the Riemann hypothesis. This paper theoretically and empirically studies the functions $ord_p(F_n)$ and formulates several unproved properties (P1)-(P4) they may have. It presents evidence raising the possibility that the Riemann hypothesis may also be encoded in $ord_p(F_n)$ for a single prime $p$. This encoding makes use of a relation of these products to the products $G_n$ of all reduced and unreduced Farey fractions of order $n$, which are connected by M\\\"obius inversion. It involves new arithmetic functions which mix the M\\\"obius function with functions of radix expansions to a fixed prime base $p$."
                    },
                    {
                        "title": "SpectrAx: Spectral Search of Axion-Like Particles Using Multi-Band Observations of Galaxy Clusters from SKA, SO, CMB-S4 and eROSITA",
                        "abstract": "The existence of axions or Axion-Like Particles (ALPs) has been predicted by various Beyond Standard Model (BSM) theories, and the proposed photon-ALP interaction is one of the ways to probe them. Such an interaction will lead to photon-ALP resonant conversion in galaxy clusters, resulting in a polarized spectral distortion in the CMB along the cluster line of sight. The estimation of this signal from galaxy clusters requires an estimation of the electron density and magnetic field in galaxy clusters. We have developed a new Bayesian framework \\texttt{SpectrAx} that can use observations from different electromagnetic bands such as radio, CMB, optical, and X-ray to infer the astrophysical properties of a galaxy cluster, such as cluster its redshift, electron density and magnetic field, along with the BSM physics such as ALPs. By using the simulated data for upcoming CMB surveys such as Simons Observatory (SO) and CMB-S4 in combination with Square Kilometer Array (SKA) and extended ROentgen Survey with an Imaging Telescope Array (eROSITA) we demonstrate the capability in accurately inferring the ALPs coupling strength along with the radial profile of electron density and magnetic field from galaxy clusters. The application of this framework to the data from future surveys by combining SKA+SO+eROSITA and SKA+CMB-S4+eROSITA will make it possible for the first time to explore both astrophysics and BSM physics from low-redshift galaxy clusters using a multi-band approach."
                    },
                    {
                        "title": "A power spectrum approach to the search for Axion-like Particles from resolved galaxy clusters using CMB as a backlight",
                        "abstract": "Axions or ALPs are hypothetical particles predicted by BSM theories, which make one of the dark matter candidates. These particles can convert into photons and vice-versa in the presence of magnetic field, with a probability decided by its coupling strength $\\mathrm{g_{a\\gamma}}$. One of the ways to detect these particles is using the CMB as a backlight. As the CMB photons pass through a galaxy cluster, they can get converted into ALPs in the mass range $10^{-15}$ eV to $10^{-11}$ eV through resonant conversion in the presence of cluster magnetic fields. This leads to a polarized spectral distortion ($\\alpha$-distortion) in the CMB as the photon polarization parallel to the magnetic field in the galaxy cluster is involved in the conversion. The fluctuations in the magnetic field and electron density in a galaxy cluster lead to spatially varying $\\alpha$-distortion around the cluster, with a power spectrum that is different from the lensed CMB polarization power spectrum for the standard model of cosmology. By measuring the difference in the polarization power spectrum around a galaxy cluster from the all-sky signal, one can find new $\\alpha$-distortion in the sky. For galaxy clusters resolvable in multiple EM bands, one can measure the coupling strength $\\mathrm{g_{a\\gamma}}$ from the ALP power spectrum. Using multi-frequency techniques like ILC to clean the foregrounds, we show that the new power spectrum-based approach of the resolved galaxy clusters from upcoming CMB experiments such as Simons Observatory and CMB-S4 can detect (or put constraints) on the ALP-photon coupling strength of $\\mathrm{g_{a\\gamma} < 5.2 \\times 10^{-12} \\, GeV^{-1}}$ and $\\mathrm{g_{a\\gamma} < 3.6 \\times 10^{-12} \\, GeV^{-1}}$ at 95\\% C.I. respectively for ALPs of masses $10^{-13}$ eV or for smaller $\\mathrm{g_{a\\gamma}}$ for lighter ALP masses (Abridged)."
                    },
                    {
                        "title": "A Diffused Background from Axion-like Particles in the Microwave Sky",
                        "abstract": "The nature of dark matter is an unsolved cosmological problem and axions are one of the weakly interacting cold dark matter candidates. Axions or ALPs (Axion-like particles) are pseudo-scalar bosons predicted by beyond-standard model theories. The weak coupling of ALPs with photons leads to the conversion of CMB photons to ALPs in the presence of a transverse magnetic field. If they have the same mass as the effective mass of a photon in a plasma, the resonant conversion would cause a polarized spectral distortion leading to temperature fluctuations with the distortion spectrum. The probability of resonant conversion depends on the properties of the cluster such as the magnetic field, electron density, and its redshift. We show that this kind of conversion can happen in numerous unresolved galaxy clusters up to high redshifts, which will lead to a diffused polarised anisotropy signal in the microwave sky. The spectrum of the signal and its shape in the angular scale will be different from the lensed CMB polarization signal. This new polarised distortion spectrum will be correlated with the distribution of clusters in the universe and hence, with the large-scale structure. The spectrum can then be probed using its spectral and spatial variation with respect to the CMB and various foregrounds. An SNR of $\\sim$ 4.36 and $\\sim$ 93.87 are possible in the CMB-S4 145 GHz band and CMB-HD 150 GHz band respectively for a photon-ALPs coupling strength of $\\mathrm{g_{a \\gamma} = 10^{-12} \\, GeV^{-1}}$ using galaxy clusters beyond redshift z $= 1$. The same signal would lead to additional RMS fluctuations of $\\sim \\mathrm{7.5 \\times 10^{-2} \\, \\mu K}$ at 145 GHz. In the absence of any signal, future CMB experiments such as Simons Observatory (SO), CMB-S4, and CMB-HD can put constraints on coupling strength better than current bounds from particle physics experiment CERN Axion Solar Telescope (CAST)."
                    },
                    {
                        "title": "Momentum Improves Normalized SGD",
                        "abstract": "We provide an improved analysis of normalized SGD showing that adding momentum provably removes the need for large batch sizes on non-convex objectives. Then, we consider the case of objectives with bounded second derivative and show that in this case a small tweak to the momentum formula allows normalized SGD with momentum to find an $\\epsilon$-critical point in $O(1/\\epsilon^{3.5})$ iterations, matching the best-known rates without accruing any logarithmic factors or dependence on dimension. We also provide an adaptive method that automatically improves convergence rates when the variance in the gradients is small. Finally, we show that our method is effective when employed on popular large scale tasks such as ResNet-50 and BERT pretraining, matching the performance of the disparate methods used to get state-of-the-art results on both tasks."
                    },
                    {
                        "title": "High-probability Bounds for Non-Convex Stochastic Optimization with Heavy Tails",
                        "abstract": "We consider non-convex stochastic optimization using first-order algorithms for which the gradient estimates may have heavy tails. We show that a combination of gradient clipping, momentum, and normalized gradient descent yields convergence to critical points in high-probability with best-known rates for smooth losses when the gradients only have bounded $\\mathfrak{p}$th moments for some $\\mathfrak{p}\\in(1,2]$. We then consider the case of second-order smooth losses, which to our knowledge have not been studied in this setting, and again obtain high-probability bounds for any $\\mathfrak{p}$. Moreover, our results hold for arbitrary smooth norms, in contrast to the typical SGD analysis which requires a Hilbert space norm. Further, we show that after a suitable \"burn-in\" period, the objective value will monotonically decrease for every iteration until a critical point is identified, which provides intuition behind the popular practice of learning rate \"warm-up\" and also yields a last-iterate guarantee."
                    },
                    {
                        "title": "Mechanic: A Learning Rate Tuner",
                        "abstract": "We introduce a technique for tuning the learning rate scale factor of any base optimization algorithm and schedule automatically, which we call \\textsc{mechanic}. Our method provides a practical realization of recent theoretical reductions for accomplishing a similar goal in online convex optimization. We rigorously evaluate \\textsc{mechanic} on a range of large scale deep learning tasks with varying batch sizes, schedules, and base optimization algorithms. These experiments demonstrate that depending on the problem, \\textsc{mechanic} either comes very close to, matches or even improves upon manual tuning of learning rates."
                    },
                    {
                        "title": "Extreme Memorization via Scale of Initialization",
                        "abstract": "We construct an experimental setup in which changing the scale of initialization strongly impacts the implicit regularization induced by SGD, interpolating from good generalization performance to completely memorizing the training set while making little progress on the test set. Moreover, we find that the extent and manner in which generalization ability is affected depends on the activation and loss function used, with $\\sin$ activation demonstrating extreme memorization. In the case of the homogeneous ReLU activation, we show that this behavior can be attributed to the loss function. Our empirical investigation reveals that increasing the scale of initialization correlates with misalignment of representations and gradients across examples in the same class. This insight allows us to devise an alignment measure over gradients and representations which can capture this phenomenon. We demonstrate that our alignment measure correlates with generalization of deep models trained on image classification tasks."
                    },
                    {
                        "title": "ALX: Large Scale Matrix Factorization on TPUs",
                        "abstract": "We present ALX, an open-source library for distributed matrix factorization using Alternating Least Squares, written in JAX. Our design allows for efficient use of the TPU architecture and scales well to matrix factorization problems of O(B) rows/columns by scaling the number of available TPU cores. In order to spur future research on large scale matrix factorization methods and to illustrate the scalability properties of our own implementation, we also built a real world web link prediction dataset called WebGraph. This dataset can be easily modeled as a matrix factorization problem. We created several variants of this dataset based on locality and sparsity properties of sub-graphs. The largest variant of WebGraph has around 365M nodes and training a single epoch finishes in about 20 minutes with 256 TPU cores. We include speed and performance numbers of ALX on all variants of WebGraph. Both the framework code and the dataset is open-sourced."
                    },
                    {
                        "title": "Long Range Language Modeling via Gated State Spaces",
                        "abstract": "State space models have shown to be effective at modeling long range dependencies, specially on sequence classification tasks. In this work we focus on autoregressive sequence modeling over English books, Github source code and ArXiv mathematics articles. Based on recent developments around the effectiveness of gated activation functions, we propose a new layer named Gated State Space (GSS) and show that it trains significantly faster than the diagonal version of S4 (i.e. DSS) on TPUs, is fairly competitive with several well-tuned Transformer-based baselines and exhibits zero-shot generalization to longer inputs while being straightforward to implement. Finally, we show that leveraging self-attention to model local dependencies improves the performance of GSS even further."
                    },
                    {
                        "title": "Large Scale Transfer Learning for Differentially Private Image Classification",
                        "abstract": "Differential Privacy (DP) provides a formal framework for training machine learning models with individual example level privacy. In the field of deep learning, Differentially Private Stochastic Gradient Descent (DP-SGD) has emerged as a popular private training algorithm. Unfortunately, the computational cost of training large-scale models with DP-SGD is substantially higher than non-private training. This is further exacerbated by the fact that increasing the number of parameters leads to larger degradation in utility with DP. In this work, we zoom in on the ImageNet dataset and demonstrate that, similar to the non-private case, pre-training over-parameterized models on a large public dataset can lead to substantial gains when the model is finetuned privately. Moreover, by systematically comparing private and non-private models across a range of large batch sizes, we find that similar to non-private setting, choice of optimizer can further improve performance substantially with DP. By using LAMB optimizer with DP-SGD we saw improvement of up to 20$\\%$ points (absolute). Finally, we show that finetuning just the last layer for a \\emph{single step} in the full batch setting, combined with extremely small-scale (near-zero) initialization leads to both SOTA results of 81.7 $\\%$ under a wide privacy budget range of $\\epsilon \\in [4, 10]$ and $\\delta$ = $10^{-6}$ while minimizing the computational overhead substantially."
                    },
                    {
                        "title": "Products of binomial coefficients and unreduced Farey fractions",
                        "abstract": "This paper studies the product $\\bar{G}_n$ of the binomial coefficients in the n-th row of Pascal's triangle, which equals the reciprocal of the product of all the reduced and unreduced Farey fractions of order n. It studies its size as a real number, measured by its logarithm $log(\\bar{G}_n)$, and its prime factorization, measured by the order of divisibility by a fixed prime p, each viewed as a function of n. It derives three formulas for its prime power divisibility, $ord_p(\\bar{G}_n)$, two of which relate it to base p radix expansions of n, and which display different facets of its behavior. These formulas are used to determine the maximal growth rate of each $ord_p(\\bar{G}_n)$ and structure of the fluctuations of these functions. It also defines analogous functions for all integer bases $b$ replacing prime bases. A final topic relates the factorizations of $\\bar{G}_n$ to Chebyshev-type prime-counting estimates and the prime number theorem."
                    },
                    {
                        "title": "Convexifying Transformers: Improving optimization and understanding of transformer networks",
                        "abstract": "Understanding the fundamental mechanism behind the success of transformer networks is still an open problem in the deep learning literature. Although their remarkable performance has been mostly attributed to the self-attention mechanism, the literature still lacks a solid analysis of these networks and interpretation of the functions learned by them. To this end, we study the training problem of attention/transformer networks and introduce a novel convex analytic approach to improve the understanding and optimization of these networks. Particularly, we first introduce a convex alternative to the self-attention mechanism and reformulate the regularized training problem of transformer networks with our alternative convex attention. Then, we cast the reformulation as a convex optimization problem that is interpretable and easier to optimize. Moreover, as a byproduct of our convex analysis, we reveal an implicit regularization mechanism, which promotes sparsity across tokens. Therefore, we not only improve the optimization of attention/transformer networks but also provide a solid theoretical understanding of the functions learned by them. We also demonstrate the effectiveness of our theory through several numerical experiments."
                    },
                    {
                        "title": "Optimal Stochastic Non-smooth Non-convex Optimization through Online-to-Non-convex Conversion",
                        "abstract": "We present new algorithms for optimizing non-smooth, non-convex stochastic objectives based on a novel analysis technique. This improves the current best-known complexity for finding a $(\\delta,\\epsilon)$-stationary point from $O(\\epsilon^{-4}\\delta^{-1})$ stochastic gradient queries to $O(\\epsilon^{-3}\\delta^{-1})$, which we also show to be optimal. Our primary technique is a reduction from non-smooth non-convex optimization to online learning, after which our results follow from standard regret bounds in online learning. For deterministic and second-order smooth objectives, applying more advanced optimistic online learning techniques enables a new complexity of $O(\\epsilon^{-1.5}\\delta^{-0.5})$. Our techniques also recover all optimal or best-known results for finding $\\epsilon$ stationary points of smooth or second-order smooth objectives in both stochastic and deterministic settings."
                    },
                    {
                        "title": "Simplifying and Understanding State Space Models with Diagonal Linear RNNs",
                        "abstract": "Sequence models based on linear state spaces (SSMs) have recently emerged as a promising choice of architecture for modeling long range dependencies across various modalities. However, they invariably rely on discretization of a continuous state space, which complicates their presentation and understanding. In this work, we dispose of the discretization step, and propose a model based on vanilla Diagonal Linear RNNs ($\\mathrm{DLR}$). We empirically show that, despite being conceptually much simpler, $\\mathrm{DLR}$ is as performant as previously-proposed SSMs on a variety of tasks and benchmarks including Long Range Arena and raw speech classification. Moreover, we characterize the expressivity of SSMs (including $\\mathrm{DLR}$) and attention-based models via a suite of $13$ synthetic sequence-to-sequence tasks involving interactions over tens of thousands of tokens, ranging from simple operations, such as shifting an input sequence, to detecting co-dependent visual features over long spatial ranges in flattened images. We find that while SSMs report near-perfect performance on tasks that can be modeled via $\\textit{few}$ convolutional kernels, they struggle on tasks requiring $\\textit{many}$ such kernels and especially when the desired sequence manipulation is $\\textit{context-dependent}$. Despite these limitations, $\\mathrm{DLR}$ reaches high performance on two higher-order reasoning tasks $\\mathrm{ListOpsSubTrees}$ and $\\mathrm{PathfinderSegmentation}\\text{-}\\mathrm{256}$ with input lengths $8K$ and $65K$ respectively, and gives encouraging performance on $\\mathrm{PathfinderSegmentation}\\text{-}\\mathrm{512}$ with input length $262K$ for which attention is not a viable choice."
                    },
                    {
                        "title": "Retouchdown: Adding Touchdown to StreetLearn as a Shareable Resource for Language Grounding Tasks in Street View",
                        "abstract": "The Touchdown dataset (Chen et al., 2019) provides instructions by human annotators for navigation through New York City streets and for resolving spatial descriptions at a given location. To enable the wider research community to work effectively with the Touchdown tasks, we are publicly releasing the 29k raw Street View panoramas needed for Touchdown. We follow the process used for the StreetLearn data release (Mirowski et al., 2019) to check panoramas for personally identifiable information and blur them as necessary. These have been added to the StreetLearn dataset and can be obtained via the same process as used previously for StreetLearn. We also provide a reference implementation for both of the Touchdown tasks: vision and language navigation (VLN) and spatial description resolution (SDR). We compare our model results to those given in Chen et al. (2019) and show that the panoramas we have added to StreetLearn fully support both Touchdown tasks and can be used effectively for further research and comparison."
                    }
                ]
            },
            "d6cb0adc-8eee-4d57-a803-357e6c61a840": {
                "pk": "d6cb0adc-8eee-4d57-a803-357e6c61a840",
                "name": "Konstantin Mishchenko",
                "collaborators": [
                    "Peter Richt\u00e1rik",
                    "Yura Malitsky",
                    "Ahmed Khaled",
                    "Dmitry Kovalev",
                    "Aaron Defazio",
                    "Mallory Montgomery",
                    "Federico Vaggi",
                    "Rustem Islamov",
                    "Eduard Gorbunov",
                    "Samuel Horv\u00e1th"
                ],
                "domain": [
                    "Optimization",
                    "Machine Learning",
                    "Federated Learning",
                    "Stochastic Methods"
                ],
                "publications": [
                    {
                        "title": "Sinkhorn Algorithm as a Special Case of Stochastic Mirror Descent",
                        "abstract": "We present a new perspective on the celebrated Sinkhorn algorithm by showing that is a special case of incremental/stochastic mirror descent. In order to see this, one should simply plug Kullback-Leibler divergence in both mirror map and the objective function. Since the problem has unbounded domain, the objective function is neither smooth nor it has bounded gradients. However, one can still approach the problem using the notion of relative smoothness, obtaining that the stochastic objective is 1-relative smooth. The discovered equivalence allows us to propose 1) new methods for optimal transport, 2) an extension of Sinkhorn algorithm beyond two constraints."
                    },
                    {
                        "title": "Regularized Newton Method with Global $O(1/k^2)$ Convergence",
                        "abstract": "We present a Newton-type method that converges fast from any initialization and for arbitrary convex objectives with Lipschitz Hessians. We achieve this by merging the ideas of cubic regularization with a certain adaptive Levenberg--Marquardt penalty. In particular, we show that the iterates given by $x^{k+1}=x^k - \\bigl(\\nabla^2 f(x^k) + \\sqrt{H\\|\\nabla f(x^k)\\|} \\mathbf{I}\\bigr)^{-1}\\nabla f(x^k)$, where $H>0$ is a constant, converge globally with a $\\mathcal{O}(\\frac{1}{k^2})$ rate. Our method is the first variant of Newton's method that has both cheap iterations and provably fast global convergence. Moreover, we prove that locally our method converges superlinearly when the objective is strongly convex. To boost the method's performance, we present a line search procedure that does not need prior knowledge of $H$ and is provably efficient."
                    },
                    {
                        "title": "On Seven Fundamental Optimization Challenges in Machine Learning",
                        "abstract": "Many recent successes of machine learning went hand in hand with advances in optimization. The exchange of ideas between these fields has worked both ways, with machine learning building on standard optimization procedures such as gradient descent, as well as with new directions in the optimization theory stemming from machine learning applications. In this thesis, we discuss new developments in optimization inspired by the needs and practice of machine learning, federated learning, and data science. In particular, we consider seven key challenges of mathematical optimization and develop a solution to each.   Our first contribution is the resolution of a key open problem in Federated Learning: we establish the first theoretical guarantees for the famous Local SGD algorithm in the heterogeneous data regime. As the second challenge, we close the gap between the upper and lower bounds for the theory of two algorithms known as Random Reshuffling (RR) and Shuffle-Once that are widely used in practice, and set as the default data selection strategies for SGD in modern machine learning software. Our third contribution can be seen as a combination of our new theory for proximal RR and Local SGD yielding a new algorithm, which we call FedRR. Unlike Local SGD, FedRR can provably beat gradient descent in communication complexity in the heterogeneous data regime. The fourth challenge is related to the class of adaptive methods. In particular, we present the first parameter-free stepsize rule for gradient descent that provably works for any locally smooth convex objective. The fifth challenge is the development of an algorithm for distributed optimization with quantized updates that preserves linear convergence of gradient descent. Finally, in our sixth and seventh challenges, we develop new VR mechanisms applicable to the non-smooth setting based on proximal operators."
                    },
                    {
                        "title": "Prodigy: An Expeditiously Adaptive Parameter-Free Learner",
                        "abstract": "We consider the problem of estimating the learning rate in adaptive methods, such as AdaGrad and Adam. We propose Prodigy, an algorithm that provably estimates the distance to the solution $D$, which is needed to set the learning rate optimally. At its core, Prodigy is a modification of the D-Adaptation method for learning-rate-free learning. It improves upon the convergence rate of D-Adaptation by a factor of $O(\\sqrt{\\log(D/d_0)})$, where $d_0$ is the initial estimate of $D$. We test Prodigy on 12 common logistic-regression benchmark datasets, VGG11 and ResNet-50 training on CIFAR10, ViT training on Imagenet, LSTM training on IWSLT14, DLRM training on Criteo dataset, VarNet on Knee MRI dataset, as well as RoBERTa and GPT transformer training on BookWiki. Our experimental results show that our approach consistently outperforms D-Adaptation and reaches test accuracy values close to that of hand-tuned Adam."
                    },
                    {
                        "title": "Adaptive Proximal Gradient Method for Convex Optimization",
                        "abstract": "In this paper, we explore two fundamental first-order algorithms in convex optimization, namely, gradient descent (GD) and proximal gradient method (ProxGD). Our focus is on making these algorithms entirely adaptive by leveraging local curvature information of smooth functions. We propose adaptive versions of GD and ProxGD that are based on observed gradient differences and, thus, have no added computational costs. Moreover, we prove convergence of our methods assuming only local Lipschitzness of the gradient. In addition, the proposed versions allow for even larger stepsizes than those initially suggested in [MM20]."
                    },
                    {
                        "title": "A Stochastic Penalty Model for Convex and Nonconvex Optimization with Big Constraints",
                        "abstract": "The last decade witnessed a rise in the importance of supervised learning applications involving {\\em big data} and {\\em big models}. Big data refers to situations where the amounts of training data available and needed causes difficulties in the training phase of the pipeline. Big model refers to situations where large dimensional and over-parameterized models are needed for the application at hand. Both of these phenomena lead to a dramatic increase in research activity aimed at taming the issues via the design of new sophisticated optimization algorithms. In this paper we turn attention to the {\\em big constraints} scenario and argue that elaborate machine learning systems of the future will necessarily need to account for a large number of real-world constraints, which will need to be incorporated in the training process. This line of work is largely unexplored, and provides ample opportunities for future work and applications. To handle the {\\em big constraints} regime, we propose a {\\em stochastic penalty} formulation which {\\em reduces the problem to the well understood big data regime}. Our formulation has many interesting properties which relate it to the original problem in various ways, with mathematical guarantees. We give a number of results specialized to nonconvex loss functions, smooth convex functions, strongly convex functions and convex constraints. We show through experiments that our approach can beat competing approaches by several orders of magnitude when a medium accuracy solution is required."
                    },
                    {
                        "title": "A Stochastic Decoupling Method for Minimizing the Sum of Smooth and Non-Smooth Functions",
                        "abstract": "We consider the problem of minimizing the sum of three convex functions: i) a smooth function $f$ in the form of an expectation or a finite average, ii) a non-smooth function $g$ in the form of a finite average of proximable functions $g_j$, and iii) a proximable regularizer $R$. We design a variance-reduced method which is able to progressively learn the proximal operator of $g$ via the computation of the proximal operator of a single randomly selected function $g_j$ in each iteration only. Our method can provably and efficiently accommodate many strategies for the estimation of the gradient of $f$, including via standard and variance-reduced stochastic estimation, effectively decoupling the smooth part of the problem from the non-smooth part. We prove a number of iteration complexity results, including a general ${\\cal O}(1/t)$ rate, ${\\cal O}(1/t^2)$ rate in the case of strongly convex smooth $f$, and several linear rates in special cases, including accelerated linear rate. For example, our method achieves a linear rate for the problem of minimizing a strongly convex function $f$ subject to linear constraints under no assumption on the constraints beyond consistency. When combined with SGD or SAGA estimators for the gradient of $f$, this leads to a very efficient method for empirical risk minimization. Our method generalizes several existing algorithms, including forward-backward splitting, Douglas-Rachford splitting, proximal SGD, proximal SAGA, SDCA, randomized Kaczmarz and Point-SAGA. However, our method leads to many new specific methods in special cases; for instance, we obtain the first randomized variant of the Dykstra's method for projection onto the intersection of closed convex sets."
                    },
                    {
                        "title": "Adaptive Gradient Descent without Descent",
                        "abstract": "We present a strikingly simple proof that two rules are sufficient to automate gradient descent: 1) don't increase the stepsize too fast and 2) don't overstep the local curvature. No need for functional values, no line search, no information about the function except for the gradients. By following these rules, you get a method adaptive to the local geometry, with convergence guarantees depending only on the smoothness in a neighborhood of a solution. Given that the problem is convex, our method converges even if the global smoothness constant is infinity. As an illustration, it can minimize arbitrary continuously twice-differentiable convex function. We examine its performance on a range of convex and nonconvex problems, including logistic regression and matrix factorization."
                    },
                    {
                        "title": "Learning-Rate-Free Learning by D-Adaptation",
                        "abstract": "D-Adaptation is an approach to automatically setting the learning rate which asymptotically achieves the optimal rate of convergence for minimizing convex Lipschitz functions, with no back-tracking or line searches, and no additional function value or gradient evaluations per step. Our approach is the first hyper-parameter free method for this class without additional multiplicative log factors in the convergence rate. We present extensive experiments for SGD and Adam variants of our method, where the method automatically matches hand-tuned learning rates across more than a dozen diverse machine learning problems, including large-scale vision and language problems.   An open-source implementation is available."
                    },
                    {
                        "title": "A Self-supervised Approach to Hierarchical Forecasting with Applications to Groupwise Synthetic Controls",
                        "abstract": "When forecasting time series with a hierarchical structure, the existing state of the art is to forecast each time series independently, and, in a post-treatment step, to reconcile the time series in a way that respects the hierarchy (Hyndman et al., 2011; Wickramasuriya et al., 2018). We propose a new loss function that can be incorporated into any maximum likelihood objective with hierarchical data, resulting in reconciled estimates with confidence intervals that correctly account for additional uncertainty due to imperfect reconciliation. We evaluate our method using a non-linear model and synthetic data on a counterfactual forecasting problem, where we have access to the ground truth and contemporaneous covariates, and show that we largely improve over the existing state-of-the-art method."
                    },
                    {
                        "title": "Proximal and Federated Random Reshuffling",
                        "abstract": "Random Reshuffling (RR), also known as Stochastic Gradient Descent (SGD) without replacement, is a popular and theoretically grounded method for finite-sum minimization. We propose two new algorithms: Proximal and Federated Random Reshuffing (ProxRR and FedRR). The first algorithm, ProxRR, solves composite convex finite-sum minimization problems in which the objective is the sum of a (potentially non-smooth) convex regularizer and an average of $n$ smooth objectives. We obtain the second algorithm, FedRR, as a special case of ProxRR applied to a reformulation of distributed problems with either homogeneous or heterogeneous data. We study the algorithms' convergence properties with constant and decreasing stepsizes, and show that they have considerable advantages over Proximal and Local SGD. In particular, our methods have superior complexities and ProxRR evaluates the proximal operator once per epoch only. When the proximal operator is expensive to compute, this small difference makes ProxRR up to $n$ times faster than algorithms that evaluate the proximal operator in every iteration. We give examples of practical optimization tasks where the proximal operator is difficult to compute and ProxRR has a clear advantage. Finally, we corroborate our results with experiments on real data sets."
                    },
                    {
                        "title": "Stochastic Newton and Cubic Newton Methods with Simple Local Linear-Quadratic Rates",
                        "abstract": "We present two new remarkably simple stochastic second-order methods for minimizing the average of a very large number of sufficiently smooth and strongly convex functions. The first is a stochastic variant of Newton's method (SN), and the second is a stochastic variant of cubically regularized Newton's method (SCN). We establish local linear-quadratic convergence results. Unlike existing stochastic variants of second order methods, which require the evaluation of a large number of gradients and/or Hessians in each iteration to guarantee convergence, our methods do not have this shortcoming. For instance, the simplest variants of our methods in each iteration need to compute the gradient and Hessian of a {\\em single} randomly selected function only. In contrast to most existing stochastic Newton and quasi-Newton methods, our approach guarantees local convergence faster than with first-order oracle and adapts to the problem's curvature. Interestingly, our method is not unbiased, so our theory provides new intuition for designing new stochastic methods."
                    },
                    {
                        "title": "Partially Personalized Federated Learning: Breaking the Curse of Data Heterogeneity",
                        "abstract": "We present a partially personalized formulation of Federated Learning (FL) that strikes a balance between the flexibility of personalization and cooperativeness of global training. In our framework, we split the variables into global parameters, which are shared across all clients, and individual local parameters, which are kept private. We prove that under the right split of parameters, it is possible to find global parameters that allow each client to fit their data perfectly, and refer to the obtained problem as overpersonalized. For instance, the shared global parameters can be used to learn good data representations, whereas the personalized layers are fine-tuned for a specific client. Moreover, we present a simple algorithm for the partially personalized formulation that offers significant benefits to all clients. In particular, it breaks the curse of data heterogeneity in several settings, such as training with local steps, asynchronous training, and Byzantine-robust training."
                    },
                    {
                        "title": "IntSGD: Adaptive Floatless Compression of Stochastic Gradients",
                        "abstract": "We propose a family of adaptive integer compression operators for distributed Stochastic Gradient Descent (SGD) that do not communicate a single float. This is achieved by multiplying floating-point vectors with a number known to every device and then rounding to integers. In contrast to the prior work on integer compression for SwitchML by Sapio et al. (2021), our IntSGD method is provably convergent and computationally cheaper as it estimates the scaling of vectors adaptively. Our theory shows that the iteration complexity of IntSGD matches that of SGD up to constant factors for both convex and non-convex, smooth and non-smooth functions, with and without overparameterization. Moreover, our algorithm can also be tailored for the popular all-reduce primitive and shows promising empirical performance."
                    },
                    {
                        "title": "Revisiting Stochastic Extragradient",
                        "abstract": "We fix a fundamental issue in the stochastic extragradient method by providing a new sampling strategy that is motivated by approximating implicit updates. Since the existing stochastic extragradient algorithm, called Mirror-Prox, of (Juditsky et al., 2011) diverges on a simple bilinear problem when the domain is not bounded, we prove guarantees for solving variational inequality that go beyond existing settings. Furthermore, we illustrate numerically that the proposed variant converges faster than many other methods on bilinear saddle-point problems. We also discuss how extragradient can be applied to training Generative Adversarial Networks (GANs) and how it compares to other methods. Our experiments on GANs demonstrate that the introduced approach may make the training faster in terms of data passes, while its higher iteration complexity makes the advantage smaller."
                    },
                    {
                        "title": "Random Reshuffling: Simple Analysis with Vast Improvements",
                        "abstract": "Random Reshuffling (RR) is an algorithm for minimizing finite-sum functions that utilizes iterative gradient descent steps in conjunction with data reshuffling. Often contrasted with its sibling Stochastic Gradient Descent (SGD), RR is usually faster in practice and enjoys significant popularity in convex and non-convex optimization. The convergence rate of RR has attracted substantial attention recently and, for strongly convex and smooth functions, it was shown to converge faster than SGD if 1) the stepsize is small, 2) the gradients are bounded, and 3) the number of epochs is large. We remove these 3 assumptions, improve the dependence on the condition number from $\\kappa^2$ to $\\kappa$ (resp. from $\\kappa$ to $\\sqrt{\\kappa}$) and, in addition, show that RR has a different type of variance. We argue through theory and experiments that the new variance type gives an additional justification of the superior performance of RR. To go beyond strong convexity, we present several results for non-strongly convex and non-convex objectives. We show that in all cases, our theory improves upon existing literature. Finally, we prove fast convergence of the Shuffle-Once (SO) algorithm, which shuffles the data only once, at the beginning of the optimization process. Our theory for strongly-convex objectives tightly matches the known lower bounds for both RR and SO and substantiates the common practical heuristic of shuffling once or only a few times. As a byproduct of our analysis, we also get new results for the Incremental Gradient algorithm (IG), which does not shuffle the data at all."
                    },
                    {
                        "title": "SEGA: Variance Reduction via Gradient Sketching",
                        "abstract": "We propose a randomized first order optimization method--SEGA (SkEtched GrAdient method)-- which progressively throughout its iterations builds a variance-reduced estimate of the gradient from random linear measurements (sketches) of the gradient obtained from an oracle. In each iteration, SEGA updates the current estimate of the gradient through a sketch-and-project operation using the information provided by the latest sketch, and this is subsequently used to compute an unbiased estimate of the true gradient through a random relaxation procedure. This unbiased estimate is then used to perform a gradient step. Unlike standard subspace descent methods, such as coordinate descent, SEGA can be used for optimization problems with a non-separable proximal term. We provide a general convergence analysis and prove linear convergence for strongly convex objectives. In the special case of coordinate sketches, SEGA can be enhanced with various techniques such as importance sampling, minibatching and acceleration, and its rate is up to a small constant factor identical to the best-known rate of coordinate descent."
                    },
                    {
                        "title": "MISO is Making a Comeback With Better Proofs and Rates",
                        "abstract": "MISO, also known as Finito, was one of the first stochastic variance reduced methods discovered, yet its popularity is fairly low. Its initial analysis was significantly limited by the so-called Big Data assumption. Although the assumption was lifted in subsequent work using negative momentum, this introduced a new parameter and required knowledge of strong convexity and smoothness constants, which is rarely possible in practice. We rehabilitate the method by introducing a new variant that needs only smoothness constant and does not have any extra parameters. Furthermore, when removing the strong convexity constant from the stepsize, we present a new analysis of the method, which no longer uses the assumption that every component is strongly convex. This allows us to also obtain so far unknown nonconvex convergence of MISO. To make the proposed method efficient in practice, we derive minibatching bounds with arbitrary uniform sampling that lead to linear speedup when the expected minibatch size is in a certain range. Our numerical experiments show that MISO is a serious competitor to SAGA and SVRG and sometimes outperforms them on real datasets."
                    },
                    {
                        "title": "A Distributed Flexible Delay-tolerant Proximal Gradient Algorithm",
                        "abstract": "We develop and analyze an asynchronous algorithm for distributed convex optimization when the objective writes a sum of smooth functions, local to each worker, and a non-smooth function. Unlike many existing methods, our distributed algorithm is adjustable to various levels of communication cost, delays, machines computational power, and functions smoothness. A unique feature is that the stepsizes do not depend on communication delays nor number of machines, which is highly desirable for scalability. We prove that the algorithm converges linearly in the strongly convex case, and provide guarantees of convergence for the non-strongly convex case. The obtained rates are the same as the vanilla proximal gradient algorithm over some introduced epoch sequence that subsumes the delays of the system. We provide numerical results on large-scale machine learning problems to demonstrate the merits of the proposed method."
                    },
                    {
                        "title": "First Analysis of Local GD on Heterogeneous Data",
                        "abstract": "We provide the first convergence analysis of local gradient descent for minimizing the average of smooth and convex but otherwise arbitrary functions. Problems of this form and local gradient descent as a solution method are of importance in federated learning, where each function is based on private data stored by a user on a mobile device, and the data of different users can be arbitrarily heterogeneous. We show that in a low accuracy regime, the method has the same communication complexity as gradient descent."
                    }
                ]
            },
            "56487216-65c6-4946-8470-ac4f3833c8a2": {
                "pk": "56487216-65c6-4946-8470-ac4f3833c8a2",
                "name": "Ahmed Khaled",
                "collaborators": [
                    "Peter Richt\u00e1rik",
                    "Chi Jin",
                    "Konstantin Mishchenko",
                    "Ahmed Khaled Zaher"
                ],
                "domain": [
                    "Optimization",
                    "Federated Learning",
                    "Machine Learning",
                    "Data Analysis"
                ],
                "publications": [
                    {
                        "title": "Gradient Descent with Compressed Iterates",
                        "abstract": "We propose and analyze a new type of stochastic first order method: gradient descent with compressed iterates (GDCI). GDCI in each iteration first compresses the current iterate using a lossy randomized compression technique, and subsequently takes a gradient step. This method is a distillation of a key ingredient in the current practice of federated learning, where a model needs to be compressed by a mobile device before it is sent back to a server for aggregation. Our analysis provides a step towards closing the gap between the theory and practice of federated learning, and opens the possibility for many extensions."
                    },
                    {
                        "title": "Better Theory for SGD in the Nonconvex World",
                        "abstract": "Large-scale nonconvex optimization problems are ubiquitous in modern machine learning, and among practitioners interested in solving them, Stochastic Gradient Descent (SGD) reigns supreme. We revisit the analysis of SGD in the nonconvex setting and propose a new variant of the recently introduced expected smoothness assumption which governs the behaviour of the second moment of the stochastic gradient. We show that our assumption is both more general and more reasonable than assumptions made in all prior work. Moreover, our results yield the optimal $\\mathcal{O}(\\varepsilon^{-4})$ rate for finding a stationary point of nonconvex smooth functions, and recover the optimal $\\mathcal{O}(\\varepsilon^{-1})$ rate for finding a global solution if the Polyak-{\\L}ojasiewicz condition is satisfied. We compare against convergence rates under convexity and prove a theorem on the convergence of SGD under Quadratic Functional Growth and convexity, which might be of independent interest. Moreover, we perform our analysis in a framework which allows for a detailed study of the effects of a wide array of sampling strategies and minibatch sizes for finite-sum optimization problems. We corroborate our theoretical results with experiments on real and synthetic data."
                    },
                    {
                        "title": "Faster federated optimization under second-order similarity",
                        "abstract": "Federated learning (FL) is a subfield of machine learning where multiple clients try to collaboratively learn a model over a network under communication constraints. We consider finite-sum federated optimization under a second-order function similarity condition and strong convexity, and propose two new algorithms: SVRP and Catalyzed SVRP. This second-order similarity condition has grown popular recently, and is satisfied in many applications including distributed statistical learning and differentially private empirical risk minimization. The first algorithm, SVRP, combines approximate stochastic proximal point evaluations, client sampling, and variance reduction. We show that SVRP is communication efficient and achieves superior performance to many existing algorithms when function similarity is high enough. Our second algorithm, Catalyzed SVRP, is a Catalyst-accelerated variant of SVRP that achieves even better performance and uniformly improves upon existing algorithms for federated optimization under second-order similarity and strong convexity. In the course of analyzing these algorithms, we provide a new analysis of the Stochastic Proximal Point Method (SPPM) that might be of independent interest. Our analysis of SPPM is simple, allows for approximate proximal point evaluations, does not require any smoothness assumptions, and shows a clear benefit in communication complexity over ordinary distributed stochastic gradient descent."
                    },
                    {
                        "title": "Parameterized Algorithms for Scalable Interprocedural Data-flow Analysis",
                        "abstract": "Data-flow analysis is a general technique used to compute information of interest at different points of a program and is considered to be a cornerstone of static analysis. In this thesis, we consider interprocedural data-flow analysis as formalized by the standard IFDS framework, which can express many widely-used static analyses such as reaching definitions, live variables, and null-pointer. We focus on the well-studied on-demand setting in which queries arrive one-by-one in a stream and each query should be answered as fast as possible. While the classical IFDS algorithm provides a polynomial-time solution to this problem, it is not scalable in practice. Specifically, it either requires a quadratic-time preprocessing phase or takes linear time per query, both of which are untenable for modern huge codebases with hundreds of thousands of lines. Previous works have already shown that parameterizing the problem by the treewidth of the program's control-flow graph is promising and can lead to significant gains in efficiency. Unfortunately, these results were only applicable to the limited special case of same-context queries.   In this work, we obtain significant speedups for the general case of on-demand IFDS with queries that are not necessarily same-context. This is achieved by exploiting a new graph sparsity parameter, namely the treedepth of the program's call graph. Our approach is the first to exploit the sparsity of control-flow graphs and call graphs at the same time and parameterize by both treewidth and treedepth. We obtain an algorithm with a linear preprocessing phase that can answer each query in constant time with respect to the input size. Finally, we show experimental results demonstrating that our approach significantly outperforms the classical IFDS and its on-demand variant."
                    },
                    {
                        "title": "Tuning-Free Stochastic Optimization",
                        "abstract": "Large-scale machine learning problems make the cost of hyperparameter tuning ever more prohibitive. This creates a need for algorithms that can tune themselves on-the-fly. We formalize the notion of \"tuning-free\" algorithms that can match the performance of optimally-tuned optimization algorithms up to polylogarithmic factors given only loose hints on the relevant problem parameters. We consider in particular algorithms that can match optimally-tuned Stochastic Gradient Descent (SGD). When the domain of optimization is bounded, we show tuning-free matching of SGD is possible and achieved by several existing algorithms. We prove that for the task of minimizing a convex and smooth or Lipschitz function over an unbounded domain, tuning-free optimization is impossible. We discuss conditions under which tuning-free optimization is possible even over unbounded domains. In particular, we show that the recently proposed DoG and DoWG algorithms are tuning-free when the noise distribution is sufficiently well-behaved. For the task of finding a stationary point of a smooth and potentially nonconvex function, we give a variant of SGD that matches the best-known high-probability convergence rate for tuned SGD at only an additional polylogarithmic cost. However, we also give an impossibility result that shows no algorithm can hope to match the optimal expected convergence rate for tuned SGD with high probability."
                    },
                    {
                        "title": "First Analysis of Local GD on Heterogeneous Data",
                        "abstract": "We provide the first convergence analysis of local gradient descent for minimizing the average of smooth and convex but otherwise arbitrary functions. Problems of this form and local gradient descent as a solution method are of importance in federated learning, where each function is based on private data stored by a user on a mobile device, and the data of different users can be arbitrarily heterogeneous. We show that in a low accuracy regime, the method has the same communication complexity as gradient descent."
                    },
                    {
                        "title": "Tighter Theory for Local SGD on Identical and Heterogeneous Data",
                        "abstract": "We provide a new analysis of local SGD, removing unnecessary assumptions and elaborating on the difference between two data regimes: identical and heterogeneous. In both cases, we improve the existing theory and provide values of the optimal stepsize and optimal number of local iterations. Our bounds are based on a new notion of variance that is specific to local SGD methods with different data. The tightness of our results is guaranteed by recovering known statements when we plug $H=1$, where $H$ is the number of local steps. The empirical evidence further validates the severe impact of data heterogeneity on the performance of local SGD."
                    },
                    {
                        "title": "DoWG Unleashed: An Efficient Universal Parameter-Free Gradient Descent Method",
                        "abstract": "This paper proposes a new easy-to-implement parameter-free gradient-based optimizer: DoWG (Distance over Weighted Gradients). We prove that DoWG is efficient -- matching the convergence rate of optimally tuned gradient descent in convex optimization up to a logarithmic factor without tuning any parameters, and universal -- automatically adapting to both smooth and nonsmooth problems. While popular algorithms following the AdaGrad framework compute a running average of the squared gradients to use for normalization, DoWG maintains a new distance-based weighted version of the running average, which is crucial to achieve the desired properties. To complement our theory, we also show empirically that DoWG trains at the edge of stability, and validate its effectiveness on practical machine learning tasks."
                    }
                ]
            },
            "0b480649-8c1a-4064-8f1e-b2eb3c76d2d5": {
                "pk": "0b480649-8c1a-4064-8f1e-b2eb3c76d2d5",
                "name": "Ashok Cutkosky",
                "collaborators": [
                    "Kwabena Boahen",
                    "Francesco Orabona",
                    "Harsh Mehta",
                    "Hoang Tran",
                    "Tamas Sarlos",
                    "Zakaria Mhammedi",
                    "Robert Busa-Fekete",
                    "Jiujia Zhang",
                    "Qinzi Zhang",
                    "Kwangjun Ahn"
                ],
                "domain": [
                    "Online Learning",
                    "Stochastic Optimization",
                    "Convex Analysis",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Associated Primes of the Square of the Alexander Dual of Hypergraphs",
                        "abstract": "The purpose of this paper is to provide methods for determining the associated primes of the square of the Alexander dual of the edge ideal for an m-hypergraph H. We prove a general method for detecting associated primes of the square of the Alexander dual of the edge ideal based on combinatorial conditions on the m-hypergraph. Also, we demonstrate a more efficient combinatorial criterion for detecting the non-existence of non-minimal associated primes. In investigating 3-hypergraphs, we prove a surprising extension of the previously discovered results for 2-hypergraphs (simple graphs). For 2-hypergraphs, associated primes of the square of the Alexander dual of the edge ideal are either of height 2 or of odd height greater than 2. However, we prove that in the 3-hypergraph case, there is no such restriction - or indeed any restriction - on the heights of the associated primes. Further, we generalize this result to any dimension greater than 3. Specifically, given any integers m, q, and n with 3\\leq m\\leq q\\leq n, we construct a m-hypergraph of size n with an associated prime of height q. We further prove that it is possible to construct connected m-hypergraphs under the same conditions."
                    },
                    {
                        "title": "Combining Online Learning Guarantees",
                        "abstract": "We show how to take any two parameter-free online learning algorithms with different regret guarantees and obtain a single algorithm whose regret is the minimum of the two base algorithms. Our method is embarrassingly simple: just add the iterates. This trick can generate efficient algorithms that adapt to many norms simultaneously, as well as providing diagonal-style algorithms that still maintain dimension-free guarantees. We then proceed to show how a variant on this idea yields a black-box procedure for generating optimistic online learning algorithms. This yields the first optimistic regret guarantees in the unconstrained setting and generically increases adaptivity. Further, our optimistic algorithms are guaranteed to do no worse than their non-optimistic counterparts regardless of the quality of the optimistic estimates provided to the algorithm."
                    },
                    {
                        "title": "Artificial Constraints and Lipschitz Hints for Unconstrained Online Learning",
                        "abstract": "We provide algorithms that guarantee regret $R_T(u)\\le \\tilde O(G\\|u\\|^3 + G(\\|u\\|+1)\\sqrt{T})$ or $R_T(u)\\le \\tilde O(G\\|u\\|^3T^{1/3} + GT^{1/3}+ G\\|u\\|\\sqrt{T})$ for online convex optimization with $G$-Lipschitz losses for any comparison point $u$ without prior knowledge of either $G$ or $\\|u\\|$. Previous algorithms dispense with the $O(\\|u\\|^3)$ term at the expense of knowledge of one or both of these parameters, while a lower bound shows that some additional penalty term over $G\\|u\\|\\sqrt{T}$ is necessary. Previous penalties were exponential while our bounds are polynomial in all quantities. Further, given a known bound $\\|u\\|\\le D$, our same techniques allow us to design algorithms that adapt optimally to the unknown value of $\\|u\\|$ without requiring knowledge of $G$."
                    },
                    {
                        "title": "Anytime Online-to-Batch Conversions, Optimism, and Acceleration",
                        "abstract": "A standard way to obtain convergence guarantees in stochastic convex optimization is to run an online learning algorithm and then output the average of its iterates: the actual iterates of the online learning algorithm do not come with individual guarantees. We close this gap by introducing a black-box modification to any online learning algorithm whose iterates converge to the optimum in stochastic scenarios. We then consider the case of smooth losses, and show that combining our approach with optimistic online learning algorithms immediately yields a fast convergence rate of $O(L/T^{3/2}+\\sigma/\\sqrt{T})$ on $L$-smooth problems with $\\sigma^2$ variance in the gradients. Finally, we provide a reduction that converts any adaptive online algorithm into one that obtains the optimal accelerated rate of $\\tilde O(L/T^2 + \\sigma/\\sqrt{T})$, while still maintaining $\\tilde O(1/\\sqrt{T})$ convergence in the non-smooth setting. Importantly, our algorithms adapt to $L$ and $\\sigma$ automatically: they do not need to know either to obtain these rates."
                    },
                    {
                        "title": "Adaptive Online Learning with Varying Norms",
                        "abstract": "Given any increasing sequence of norms $\\|\\cdot\\|_0,\\dots,\\|\\cdot\\|_{T-1}$, we provide an online convex optimization algorithm that outputs points $w_t$ in some domain $W$ in response to convex losses $\\ell_t:W\\to \\mathbb{R}$ that guarantees regret $R_T(u)=\\sum_{t=1}^T \\ell_t(w_t)-\\ell_t(u)\\le \\tilde O\\left(\\|u\\|_{T-1}\\sqrt{\\sum_{t=1}^T \\|g_t\\|_{t-1,\\star}^2}\\right)$ where $g_t$ is a subgradient of $\\ell_t$ at $w_t$. Our method does not require tuning to the value of $u$ and allows for arbitrary convex $W$. We apply this result to obtain new \"full-matrix\"-style regret bounds. Along the way, we provide a new examination of the full-matrix AdaGrad algorithm, suggesting a better learning rate value that improves significantly upon prior analysis. We use our new techniques to tune AdaGrad on-the-fly, realizing our improved bound in a concrete algorithm."
                    },
                    {
                        "title": "Online Convex Optimization with Unconstrained Domains and Losses",
                        "abstract": "We propose an online convex optimization algorithm (RescaledExp) that achieves optimal regret in the unconstrained setting without prior knowledge of any bounds on the loss functions. We prove a lower bound showing an exponential separation between the regret of existing algorithms that require a known bound on the loss functions and any algorithm that does not require such knowledge. RescaledExp matches this lower bound asymptotically in the number of iterations. RescaledExp is naturally hyperparameter-free and we demonstrate empirically that it matches prior optimization algorithms that require hyperparameter optimization."
                    },
                    {
                        "title": "Black-Box Reductions for Parameter-free Online Learning in Banach Spaces",
                        "abstract": "We introduce several new black-box reductions that significantly improve the design of adaptive and parameter-free online learning algorithms by simplifying analysis, improving regret guarantees, and sometimes even improving runtime. We reduce parameter-free online learning to online exp-concave optimization, we reduce optimization in a Banach space to one-dimensional optimization, and we reduce optimization over a constrained domain to unconstrained optimization. All of our reductions run as fast as online gradient descent. We use our new techniques to improve upon the previously best regret bounds for parameter-free learning, and do so for arbitrary norms."
                    },
                    {
                        "title": "Matrix-Free Preconditioning in Online Learning",
                        "abstract": "We provide an online convex optimization algorithm with regret that interpolates between the regret of an algorithm using an optimal preconditioning matrix and one using a diagonal preconditioning matrix. Our regret bound is never worse than that obtained by diagonal preconditioning, and in certain setting even surpasses that of algorithms with full-matrix preconditioning. Importantly, our algorithm runs in the same time and space complexity as online gradient descent. Along the way we incorporate new techniques that mildly streamline and improve logarithmic factors in prior regret analyses. We conclude by benchmarking our algorithm on synthetic data and deep learning tasks."
                    },
                    {
                        "title": "Fully Unconstrained Online Learning",
                        "abstract": "We provide an online learning algorithm that obtains regret $G\\|w_\\star\\|\\sqrt{T\\log(\\|w_\\star\\|G\\sqrt{T})} + \\|w_\\star\\|^2 + G^2$ on $G$-Lipschitz convex losses for any comparison point $w_\\star$ without knowing either $G$ or $\\|w_\\star\\|$. Importantly, this matches the optimal bound $G\\|w_\\star\\|\\sqrt{T}$ available with such knowledge (up to logarithmic factors), unless either $\\|w_\\star\\|$ or $G$ is so large that even $G\\|w_\\star\\|\\sqrt{T}$ is roughly linear in $T$. Thus, it matches the optimal bound in all cases in which one can achieve sublinear regret, which arguably most \"interesting\" scenarios."
                    },
                    {
                        "title": "Momentum Improves Normalized SGD",
                        "abstract": "We provide an improved analysis of normalized SGD showing that adding momentum provably removes the need for large batch sizes on non-convex objectives. Then, we consider the case of objectives with bounded second derivative and show that in this case a small tweak to the momentum formula allows normalized SGD with momentum to find an $\\epsilon$-critical point in $O(1/\\epsilon^{3.5})$ iterations, matching the best-known rates without accruing any logarithmic factors or dependence on dimension. We also provide an adaptive method that automatically improves convergence rates when the variance in the gradients is small. Finally, we show that our method is effective when employed on popular large scale tasks such as ResNet-50 and BERT pretraining, matching the performance of the disparate methods used to get state-of-the-art results on both tasks."
                    },
                    {
                        "title": "Online Learning Without Prior Information",
                        "abstract": "The vast majority of optimization and online learning algorithms today require some prior information about the data (often in the form of bounds on gradients or on the optimal parameter value). When this information is not available, these algorithms require laborious manual tuning of various hyperparameters, motivating the search for algorithms that can adapt to the data with no prior information. We describe a frontier of new lower bounds on the performance of such algorithms, reflecting a tradeoff between a term that depends on the optimal parameter value and a term that depends on the gradients' rate of growth. Further, we construct a family of algorithms whose performance matches any desired point on this frontier, which no previous algorithm reaches."
                    },
                    {
                        "title": "Distributed Stochastic Optimization via Adaptive SGD",
                        "abstract": "Stochastic convex optimization algorithms are the most popular way to train machine learning models on large-scale data. Scaling up the training process of these models is crucial, but the most popular algorithm, Stochastic Gradient Descent (SGD), is a serial method that is surprisingly hard to parallelize. In this paper, we propose an efficient distributed stochastic optimization method by combining adaptivity with variance reduction techniques. Our analysis yields a linear speedup in the number of machines, constant memory footprint, and only a logarithmic number of communication rounds. Critically, our approach is a black-box reduction that parallelizes any serial online learning algorithm, streamlining prior analysis and allowing us to leverage the significant progress that has been made in designing adaptive algorithms. In particular, we achieve optimal convergence rates without any prior knowledge of smoothness parameters, yielding a more robust algorithm that reduces the need for hyperparameter tuning. We implement our algorithm in the Spark distributed framework and exhibit dramatic performance gains on large-scale logistic regression problems."
                    },
                    {
                        "title": "High-probability Bounds for Non-Convex Stochastic Optimization with Heavy Tails",
                        "abstract": "We consider non-convex stochastic optimization using first-order algorithms for which the gradient estimates may have heavy tails. We show that a combination of gradient clipping, momentum, and normalized gradient descent yields convergence to critical points in high-probability with best-known rates for smooth losses when the gradients only have bounded $\\mathfrak{p}$th moments for some $\\mathfrak{p}\\in(1,2]$. We then consider the case of second-order smooth losses, which to our knowledge have not been studied in this setting, and again obtain high-probability bounds for any $\\mathfrak{p}$. Moreover, our results hold for arbitrary smooth norms, in contrast to the typical SGD analysis which requires a Hilbert space norm. Further, we show that after a suitable \"burn-in\" period, the objective value will monotonically decrease for every iteration until a critical point is identified, which provides intuition behind the popular practice of learning rate \"warm-up\" and also yields a last-iterate guarantee."
                    },
                    {
                        "title": "Better SGD using Second-order Momentum",
                        "abstract": "We develop a new algorithm for non-convex stochastic optimization that finds an $\\epsilon$-critical point in the optimal $O(\\epsilon^{-3})$ stochastic gradient and Hessian-vector product computations. Our algorithm uses Hessian-vector products to \"correct\" a bias term in the momentum of SGD with momentum. This leads to better gradient estimates in a manner analogous to variance reduction methods. In contrast to prior work, we do not require excessively large batch sizes, and are able to provide an adaptive algorithm whose convergence rate automatically improves with decreasing variance in the gradient estimates. We validate our results on a variety of large-scale deep learning architectures and benchmarks tasks."
                    },
                    {
                        "title": "Momentum Aggregation for Private Non-convex ERM",
                        "abstract": "We introduce new algorithms and convergence guarantees for privacy-preserving non-convex Empirical Risk Minimization (ERM) on smooth $d$-dimensional objectives. We develop an improved sensitivity analysis of stochastic gradient descent on smooth objectives that exploits the recurrence of examples in different epochs. By combining this new approach with recent analysis of momentum with private aggregation techniques, we provide an $(\\epsilon,\\delta)$-differential private algorithm that finds a gradient of norm $\\tilde O\\left(\\frac{d^{1/3}}{(\\epsilon N)^{2/3}}\\right)$ in $O\\left(\\frac{N^{7/3}\\epsilon^{4/3}}{d^{2/3}}\\right)$ gradient evaluations, improving the previous best gradient bound of $\\tilde O\\left(\\frac{d^{1/4}}{\\sqrt{\\epsilon N}}\\right)$."
                    },
                    {
                        "title": "Parameter-free Regret in High Probability with Heavy Tails",
                        "abstract": "We present new algorithms for online convex optimization over unbounded domains that obtain parameter-free regret in high-probability given access only to potentially heavy-tailed subgradient estimates. Previous work in unbounded domains considers only in-expectation results for sub-exponential subgradients. Unlike in the bounded domain case, we cannot rely on straight-forward martingale concentration due to exponentially large iterates produced by the algorithm. We develop new regularization techniques to overcome these problems. Overall, with probability at most $\\delta$, for all comparators $\\mathbf{u}$ our algorithm achieves regret $\\tilde{O}(\\| \\mathbf{u} \\| T^{1/\\mathfrak{p}} \\log (1/\\delta))$ for subgradients with bounded $\\mathfrak{p}^{th}$ moments for some $\\mathfrak{p} \\in (1, 2]$."
                    },
                    {
                        "title": "Random Scaling and Momentum for Non-smooth Non-convex Optimization",
                        "abstract": "Training neural networks requires optimizing a loss function that may be highly irregular, and in particular neither convex nor smooth. Popular training algorithms are based on stochastic gradient descent with momentum (SGDM), for which classical analysis applies only if the loss is either convex or smooth. We show that a very small modification to SGDM closes this gap: simply scale the update at each time point by an exponentially distributed random scalar. The resulting algorithm achieves optimal convergence guarantees. Intriguingly, this result is not derived by a specific analysis of SGDM: instead, it falls naturally out of a more general framework for converting online convex optimization algorithms to non-convex optimization algorithms."
                    },
                    {
                        "title": "Adam with model exponential moving average is effective for nonconvex optimization",
                        "abstract": "In this work, we offer a theoretical analysis of two modern optimization techniques for training large and complex models: (i) adaptive optimization algorithms, such as Adam, and (ii) the model exponential moving average (EMA). Specifically, we demonstrate that a clipped version of Adam with model EMA achieves the optimal convergence rates in various nonconvex optimization settings, both smooth and nonsmooth. Moreover, when the scale varies significantly across different coordinates, we demonstrate that the coordinate-wise adaptivity of Adam is provably advantageous. Notably, unlike previous analyses of Adam, our analysis crucially relies on its core elements -- momentum and discounting factors -- as well as model EMA, motivating their wide applications in practice."
                    },
                    {
                        "title": "Upper Confidence Bounds for Combining Stochastic Bandits",
                        "abstract": "We provide a simple method to combine stochastic bandit algorithms. Our approach is based on a \"meta-UCB\" procedure that treats each of $N$ individual bandit algorithms as arms in a higher-level $N$-armed bandit problem that we solve with a variant of the classic UCB algorithm. Our final regret depends only on the regret of the base algorithm with the best regret in hindsight. This approach provides an easy and intuitive alternative strategy to the CORRAL algorithm for adversarial bandits, without requiring the stability conditions imposed by CORRAL on the base algorithms. Our results match lower bounds in several settings, and we provide empirical validation of our algorithm on misspecified linear bandit and model selection problems."
                    },
                    {
                        "title": "Momentum-Based Variance Reduction in Non-Convex SGD",
                        "abstract": "Variance reduction has emerged in recent years as a strong competitor to stochastic gradient descent in non-convex problems, providing the first algorithms to improve upon the converge rate of stochastic gradient descent for finding first-order critical points. However, variance reduction techniques typically require carefully tuned learning rates and willingness to use excessively large \"mega-batches\" in order to achieve their improved results. We present a new algorithm, STORM, that does not require any batches and makes use of adaptive learning rates, enabling simpler implementation and less hyperparameter tuning. Our technique for removing the batches uses a variant of momentum to achieve variance reduction in non-convex optimization. On smooth losses $F$, STORM finds a point $\\boldsymbol{x}$ with $\\mathbb{E}[\\|\\nabla F(\\boldsymbol{x})\\|]\\le O(1/\\sqrt{T}+\\sigma^{1/3}/T^{1/3})$ in $T$ iterations with $\\sigma^2$ variance in the gradients, matching the optimal rate but without requiring knowledge of $\\sigma$."
                    }
                ]
            }
        }
    },
    "2309.05660": {
        "paper_data": {
            "title": "Hypothesis Search: Inductive Reasoning with Language Models",
            "url": "http://arxiv.org/abs/2309.05660v2",
            "arxiv_id": "2309.05660",
            "authors": [
                "Ruocheng Wang",
                "Eric Zelikman",
                "Gabriel Poesia",
                "Yewen Pu",
                "Nick Haber",
                "Noah D. Goodman"
            ],
            "abstract": "Inductive reasoning is a core problem-solving capacity: humans can identify underlying principles from a few examples, which robustly generalize to novel scenarios. Recent work evaluates large language models (LLMs) on inductive reasoning tasks by directly prompting them yielding \"in context learning.\" This works well for straightforward inductive tasks but performs poorly on complex tasks such as the Abstraction and Reasoning Corpus (ARC). In this work, we propose to improve the inductive reasoning ability of LLMs by generating explicit hypotheses at multiple levels of abstraction: we prompt the LLM to propose multiple abstract hypotheses about the problem, in natural language, then implement the natural language hypotheses as concrete Python programs. These programs can be verified by running on observed examples and generalized to novel inputs. To reduce the hypothesis search space, we explore steps to filter the set of hypotheses to implement: we either ask the LLM to summarize them into a smaller set of hypotheses or ask human annotators to select a subset. We verify our pipeline's effectiveness on the ARC visual inductive reasoning benchmark, its variant 1D-ARC, string transformation dataset SyGuS, and list transformation dataset List Functions. On a random 100-problem subset of ARC, our automated pipeline using LLM summaries achieves 30% accuracy, outperforming the direct prompting baseline (accuracy of 17%). With the minimal human input of selecting from LLM-generated candidates, performance is boosted to 33%. Our ablations show that both abstract hypothesis generation and concrete program representations benefit LLMs on inductive reasoning tasks.",
            "introduction": "ABSTRACT Inductive reasoning is a core problem-solving capacity: humans can identify under- lying principles from a few examples, which robustly generalize to novel scenarios. Recent work evaluates large language models (LLMs) on inductive reasoning tasks by directly prompting them yielding \u201cin context learning.\u201d This works well for straightforward inductive tasks but performs poorly on complex tasks such as the Abstraction and Reasoning Corpus (ARC). In this work, we propose to improve the inductive reasoning ability of LLMs by generating explicit hypotheses at multiple levels of abstraction: we prompt the LLM to propose multipleabstract reasoning tasks. In Proceedings of the Annual Meeting of the Cognitive Science Society , 2021. Nate Kushman. Generating computer programs from natural language descriptions . PhD thesis, Massachusetts Institute of Technology, 2015. Andrew K Lampinen, Ishita Dasgupta, Stephanie CY Chan, Kory Matthewson, Michael Henry Tessler, Antonia Creswell, James L McClelland, Jane X Wang, and Felix Hill. Can language models learn from explanations in context? arXiv preprint arXiv:2204.02329 , 2022. Vu Le, Daniel Perelman, Oleksandr Polozov, Mohammad Raza, Abhishek Udupa, and Sumit Gulwani. Interactive program synthesis. arXiv preprint arXiv:1703.03539 , 2017. Chin-Yew Lin. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out , pp. 74\u201381, 2004. 11Published as a conference paper at ICLR 2024 Jiayuan Mao, Chuang Gan, Pushmeet Kohli, Joshua B Tenenbaum, and Jiajun Wu. The neuro- symbolic concept learner: Interpreting scenes, words, and sentences from natural supervision. In ICLR , 2019. Ana Marasovi \u00b4c, Iz Beltagy, Doug Downey, and Matthew E Peters. Few-shot self-rationalization with natural language prompts. arXiv preprint arXiv:2111.08284 , 2021. Suvir Mirchandani, Fei Xia, Pete Florence, Brian Ichter, Danny Driess, Montserrat Gonzalez Arenas, Kanishka Rao, Dorsa Sadigh, and Andy Zeng. Large language models as general pattern machines. arXiv preprint arXiv:2307.04721 , 2023. Maxwell Nye, Anders Johan Andreassen, Guy Gur-Ari, Henryk Michalewski, Jacob Austin, David Bieber, David Dohan, Aitor Lewkowycz, Maarten Bosma, David Luan, et al. Show your work: Scratchpads for intermediate computation with language models. arXiv preprint arXiv:2112.00114 , 2021. Augustus Odena, Kensen Shi, David Bieber, Rishabh Singh, Charles Sutton, and Hanjun Dai. Bustle: Bottom-up program synthesis through learning-guided exploration. arXiv preprint arXiv:2007.14381 , 2020. Nancy Pang. Deep learning for code repair. University of British Columbia , 2018. Charles S Peirce. Questions concerning certain faculties claimed for man. The Journal of Speculative Philosophy , 2(2):103\u2013114, 1868. Md Mostafizer Rahman, Yutaka Watanobe, and Keita Nakamura. A bidirectional lstm language model for code evaluation and repair. Symmetry , 13(2):247, 2021. Nazneen Fatema Rajani, Bryan McCann, Caiming Xiong, and Richard Socher. Explain yourself! leveraging language models for commonsense reasoning. ACL, 2019. Joshua S Rule, Joshua B Tenenbaum, and Steven T Piantadosi. The child as hacker. Trends in cognitive sciences , 24(11):900\u2013915, 2020. Peter Schulam, Roni Rosenfeld, and Premkumar Devanbu. Building statistical language models of code. In 2013 1st International Workshop on Data Analysis Patterns in Software Engineering (DAPSE) , pp. 1\u20133. IEEE, 2013. Eric Schulte, Stephanie Forrest, and Westley Weimer. Automated program repair through the evolution of assembly code. In Proceedings of the 25th IEEE/ACM International Conference on Automated Software Engineering , pp. 313\u2013316, 2010. Gopal Sharma, Rishabh Goyal, Difan Liu, Evangelos Kalogerakis, and Subhransu Maji. Csgnet: Neural shape parser for constructive solid geometry. 2018 ieee. In CVF Conference on Computer Vision and Pattern Recognition , pp. 5515\u20135523, 2017. Kensen Shi, Hanjun Dai, Kevin Ellis, and Charles Sutton. Crossbeam: Learning to search in bottom-up program synthesis. In ICLR",
            "references": [
                {
                    "title": "Large Language Models as General Pattern Machines",
                    "abstract": "We observe that pre-trained large language models (LLMs) are capable of autoregressively completing complex token sequences -- from arbitrary ones procedurally generated by probabilistic context-free grammars (PCFG), to more rich spatial patterns found in the Abstraction and Reasoning Corpus (ARC), a general AI benchmark, prompted in the style of ASCII art. Surprisingly, pattern completion proficiency can be partially retained even when the sequences are expressed using tokens randomly sampled from the vocabulary. These results suggest that without any additional training, LLMs can serve as general sequence modelers, driven by in-context learning. In this work, we investigate how these zero-shot capabilities may be applied to problems in robotics -- from extrapolating sequences of numbers that represent states over time to complete simple motions, to least-to-most prompting of reward-conditioned trajectories that can discover and represent closed-loop policies (e.g., a stabilizing controller for CartPole). While difficult to deploy today for real systems due to latency, context size limitations, and compute costs, the approach of using LLMs to drive low-level control may provide an exciting glimpse into how the patterns among words could be transferred to actions."
                },
                {
                    "title": "LLMs and the Abstraction and Reasoning Corpus: Successes, Failures, and the Importance of Object-based Representations",
                    "abstract": "Can a Large Language Model (LLM) solve simple abstract reasoning problems? We explore this broad question through a systematic analysis of GPT on the Abstraction and Reasoning Corpus (ARC), a representative benchmark of abstract reasoning ability from limited examples in which solutions require some\"core knowledge\"of concepts such as objects, goal states, counting, and basic geometry. GPT-4 solves only 13/50 of the most straightforward ARC tasks when using textual encodings for their two-dimensional input-output grids. Our failure analysis reveals that GPT-4's capacity to identify objects and reason about them is significantly influenced by the sequential nature of the text that represents an object within a text encoding of a task. To test this hypothesis, we design a new benchmark, the 1D-ARC, which consists of one-dimensional (array-like) tasks that are more conducive to GPT-based reasoning, and where it indeed performs better than on the (2D) ARC. To alleviate this issue, we propose an object-based representation that is obtained through an external tool, resulting in nearly doubling the performance on solved ARC tasks and near-perfect scores on the easier 1D-ARC. Although the state-of-the-art GPT-4 is unable to\"reason\"perfectly within non-language domains such as the 1D-ARC or a simple ARC subset, our study reveals that the use of object-based representations can significantly improve its reasoning ability. Visualizations, GPT logs, and data are available at https://khalil-research.github.io/LLM4ARC."
                },
                {
                    "title": "Teaching Large Language Models to Self-Debug",
                    "abstract": "Large language models (LLMs) have achieved impressive performance on code generation. However, for complex programming tasks, generating the correct solution in one go becomes challenging, thus some prior works have designed program repair approaches to improve code generation performance. In this work, we propose Self-Debugging, which teaches a large language model to debug its predicted program via few-shot demonstrations. In particular, we demonstrate that Self-Debugging can teach the large language model to perform rubber duck debugging; i.e., without any human feedback on the code correctness or error messages, the model is able to identify its mistakes by investigating the execution results and explaining the generated code in natural language. Self-Debugging achieves the state-of-the-art performance on several code generation benchmarks, including the Spider dataset for text-to-SQL generation, TransCoder for C++-to-Python translation, and MBPP for text-to-Python generation. On the Spider benchmark where there are no unit tests to verify the correctness of predictions, Self-Debugging with code explanation consistently improves the baseline by 2-3%, and improves the prediction accuracy on problems of the hardest level by 9%. On TransCoder and MBPP where unit tests are available, Self-Debugging improves the baseline accuracy by up to 12%. Meanwhile, by leveraging feedback messages and reusing failed predictions, Self-Debugging notably improves sample efficiency, and can match or outperform baseline models that generate more than 10x candidate programs."
                },
                {
                    "title": "ViperGPT: Visual Inference via Python Execution for Reasoning",
                    "abstract": "Answering visual queries is a complex task that requires both visual processing and reasoning. End-to-end models, the dominant approach for this task, do not explicitly differentiate between the two, limiting interpretability and generalization. Learning modular programs presents a promising alternative, but has proven challenging due to the difficulty of learning both the programs and modules simultaneously. We introduce ${\\color{green}{\\text{ViperGPT}}}$, a framework that leverages code-generation models to compose vision-and-language models into subroutines to produce a result for any query. ${\\color{green}{\\text{ViperGPT}}}$ utilizes a provided API to access the available modules, and composes them by generating Python code that is later executed. This simple approach requires no further training, and achieves state-of-the-art results across various complex visual tasks."
                },
                {
                    "title": "PaLM-E: An Embodied Multimodal Language Model",
                    "abstract": "Large language models excel at a wide range of complex tasks. However, enabling general inference in the real world, e.g., for robotics problems, raises the challenge of grounding. We propose embodied language models to directly incorporate real-world continuous sensor modalities into language models and thereby establish the link between words and percepts. Input to our embodied language model are multi-modal sentences that interleave visual, continuous state estimation, and textual input encodings. We train these encodings end-to-end, in conjunction with a pre-trained large language model, for multiple embodied tasks including sequential robotic manipulation planning, visual question answering, and captioning. Our evaluations show that PaLM-E, a single large embodied multimodal model, can address a variety of embodied reasoning tasks, from a variety of observation modalities, on multiple embodiments, and further, exhibits positive transfer: the model benefits from diverse joint training across internet-scale language, vision, and visual-language domains. Our largest model, PaLM-E-562B with 562B parameters, in addition to being trained on robotics tasks, is a visual-language generalist with state-of-the-art performance on OK-VQA, and retains generalist language capabilities with increasing scale."
                },
                {
                    "title": "Parsel\ud83e\udd86: Algorithmic Reasoning with Language Models by Composing Decompositions",
                    "abstract": "Despite recent success in large language model (LLM) reasoning, LLMs struggle with hierarchical multi-step reasoning tasks like generating complex programs. For these tasks, humans often start with a high-level algorithmic design and implement each part gradually. We introduce Parsel, a framework enabling automatic implementation and validation of complex algorithms with code LLMs. With Parsel, we automatically decompose algorithmic tasks into hierarchical natural language function descriptions and then search over combinations of possible function implementations using tests. We show that Parsel can be used across domains requiring hierarchical reasoning, including program synthesis and robotic planning. We find that, using Parsel, LLMs solve more competition-level problems in the APPS dataset, resulting in pass rates over 75\\% higher than prior results from directly sampling AlphaCode and Codex, while often using a smaller sample budget. Moreover, with automatically generated tests, we find that Parsel can improve the state-of-the-art pass@1 performance on HumanEval from 67\\% to 85\\%. We also find that LLM-generated robotic plans using Parsel are more than twice as likely to be considered accurate than directly generated plans. Lastly, we explore how Parsel addresses LLM limitations and discuss how Parsel may be useful for human programmers. We release our code at https://github.com/ezelikman/parsel"
                },
                {
                    "title": "Coder Reviewer Reranking for Code Generation",
                    "abstract": "Sampling diverse programs from a code language model and reranking with model likelihood is a popular method for code generation but it is prone to preferring degenerate solutions. Inspired by collaborative programming, we propose Coder-Reviewer reranking. We augment Coder language models from past work, which generate programs given language instructions, with Reviewer models, which evaluate the likelihood of the instruction given the generated programs. We perform an extensive study across six datasets with eight models from three model families. Experimental results show that Coder-Reviewer reranking leads to consistent and significant improvement (up to 17% absolute accuracy gain) over reranking with the Coder model only. When combined with executability filtering, Coder-Reviewer reranking can often outperform the minimum Bayes risk method. Coder-Reviewer reranking is easy to implement by prompting, can generalize to different programming languages, and works well with off-the-shelf hyperparameters."
                },
                {
                    "title": "Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks",
                    "abstract": "Recently, there has been significant progress in teaching language models to perform step-by-step reasoning to solve complex numerical reasoning tasks. Chain-of-thoughts prompting (CoT) is by far the state-of-art method for these tasks. CoT uses language models to perform both reasoning and computation in the multi-step `thought' process. To disentangle computation from reasoning, we propose `Program of Thoughts' (PoT), which uses language models (mainly Codex) to express the reasoning process as a program. The computation is relegated to an external computer, which executes the generated programs to derive the answer. We evaluate PoT on five math word problem datasets (GSM, AQuA, SVAMP, TabMWP, MultiArith) and three financial-QA datasets (FinQA, ConvFinQA, TATQA) for both few-shot and zero-shot setups. Under both few-shot and zero-shot settings, PoT can show an average performance gain over CoT by around 12\\% across all the evaluated datasets. By combining PoT with self-consistency decoding, we can achieve SoTA performance on all math problem datasets and near-SoTA performance on financial datasets. All of our data and code are released in Github https://github.com/wenhuchen/Program-of-Thoughts"
                },
                {
                    "title": "PAL: Program-aided Language Models",
                    "abstract": "Large language models (LLMs) have recently demonstrated an impressive ability to perform arithmetic and symbolic reasoning tasks, when provided with a few examples at test time (\"few-shot prompting\"). Much of this success can be attributed to prompting methods such as\"chain-of-thought'', which employ LLMs for both understanding the problem description by decomposing it into steps, as well as solving each step of the problem. While LLMs seem to be adept at this sort of step-by-step decomposition, LLMs often make logical and arithmetic mistakes in the solution part, even when the problem is decomposed correctly. In this paper, we present Program-Aided Language models (PAL): a novel approach that uses the LLM to read natural language problems and generate programs as the intermediate reasoning steps, but offloads the solution step to a runtime such as a Python interpreter. With PAL, decomposing the natural language problem into runnable steps remains the only learning task for the LLM, while solving is delegated to the interpreter. We demonstrate this synergy between a neural LLM and a symbolic interpreter across 13 mathematical, symbolic, and algorithmic reasoning tasks from BIG-Bench Hard and other benchmarks. In all these natural language reasoning tasks, generating code using an LLM and reasoning using a Python interpreter leads to more accurate results than much larger models. For example, PAL using Codex achieves state-of-the-art few-shot accuracy on the GSM8K benchmark of math word problems, surpassing PaLM-540B which uses chain-of-thought by absolute 15% top-1. Our code and data are publicly available at http://reasonwithpal.com/ ."
                },
                {
                    "title": "Visual Programming: Compositional visual reasoning without training",
                    "abstract": "We present Visprog, a neuro-symbolic approach to solving complex and compositional visual tasks given natural language instructions. Visprog avoids the need for any task-specific training. Instead, it uses the incontext learning ability of large language models to generate python-like modular programs, which are then executed to get both the solution and a comprehensive and interpretable rationale. Each line of the generated program may invoke one of several off-the-shelf computer vision models, image processing subroutines, or python functions to produce intermediate outputs that may be consumed by subsequent parts of the program. We demonstrate the flexibility of VIsPROG on 4 diverse tasks - compositional visual question answering, zero-shot reasoning on image pairs, factual knowledge object tagging, and language-guided image editing. We believe neuro-symbolic approaches like Visprog are an exciting avenue to easily and effectively expand the scope of AI systems to serve the long tail of complex tasks that people may wish to perform."
                },
                {
                    "title": "Teaching Algorithmic Reasoning via In-context Learning",
                    "abstract": "Large language models (LLMs) have shown increasing in-context learning capabilities through scaling up model and data size. Despite this progress, LLMs are still unable to solve algorithmic reasoning problems. While providing a rationale with the final answer has led to further improvements in multi-step reasoning problems, Anil et al. 2022 showed that even simple algorithmic reasoning tasks such as parity are far from solved. In this work, we identify and study four key stages for successfully teaching algorithmic reasoning to LLMs: (1) formulating algorithms as skills, (2) teaching multiple skills simultaneously (skill accumulation), (3) teaching how to combine skills (skill composition) and (4) teaching how to use skills as tools. We show that it is possible to teach algorithmic reasoning to LLMs via in-context learning, which we refer to as algorithmic prompting. We evaluate our approach on a variety of arithmetic and quantitative reasoning tasks, and demonstrate significant boosts in performance over existing prompting techniques. In particular, for long parity, addition, multiplication and subtraction, we achieve an error reduction of approximately 10x, 9x, 5x and 2x respectively compared to the best available baselines."
                },
                {
                    "title": "Graphs, Constraints, and Search for the Abstraction and Reasoning Corpus",
                    "abstract": "The Abstraction and Reasoning Corpus (ARC) aims at benchmarking the performance of general artificial intelligence algorithms. The ARC's focus on broad generalization and few-shot learning has made it difficult to solve using pure machine learning. A more promising approach has been to perform program synthesis within an appropriately designed Domain Specific Language (DSL). However, these too have seen limited success. We propose Abstract Reasoning with Graph Abstractions (ARGA), a new object-centric framework that first represents images using graphs and then performs a search for a correct program in a DSL that is based on the abstracted graph space. The complexity of this combinatorial search is tamed through the use of constraint acquisition, state hashing, and Tabu search. An extensive set of experiments demonstrates the promise of ARGA in tackling some of the complicated object-centric tasks of the ARC rather efficiently, producing programs that are correct and easy to understand."
                },
                {
                    "title": "Language models show human-like content effects on reasoning",
                    "abstract": "Reasoning is a key ability for an intelligent system. Large language models (LMs) achieve above-chance performance on abstract reasoning tasks, but exhibit many imperfections. However, human abstract reasoning is also imperfect. For example, human reasoning is affected by our real-world knowledge and beliefs, and shows notable\"content effects\"; humans reason more reliably when the semantic content of a problem supports the correct logical inferences. These content-entangled reasoning patterns play a central role in debates about the fundamental nature of human intelligence. Here, we investigate whether language models $\\unicode{x2014}$ whose prior expectations capture some aspects of human knowledge $\\unicode{x2014}$ similarly mix content into their answers to logical problems. We explored this question across three logical reasoning tasks: natural language inference, judging the logical validity of syllogisms, and the Wason selection task. We evaluate state of the art large language models, as well as humans, and find that the language models reflect many of the same patterns observed in humans across these tasks $\\unicode{x2014}$ like humans, models answer more accurately when the semantic content of a task supports the logical inferences. These parallels are reflected both in answer patterns, and in lower-level features like the relationship between model answer distributions and human response times. Our findings have implications for understanding both these cognitive effects in humans, and the factors that contribute to language model performance."
                },
                {
                    "title": "Can language models learn from explanations in context?",
                    "abstract": "Language Models (LMs) can perform new tasks by adapting to a few in-context examples. For humans, explanations that connect examples to task principles can improve learning. We therefore investigate whether explanations of few-shot examples can help LMs. We annotate questions from 40 challenging tasks with answer explanations, and various matched control explanations. We evaluate how different types of explanations, instructions, and controls affect zero- and few-shot performance. We analyze these results using statistical multilevel modeling techniques that account for the nested dependencies among conditions, tasks, prompts, and models. We find that explanations can improve performance -- even without tuning. Furthermore, explanations hand-tuned for performance on a small validation set offer substantially larger benefits, and building a prompt by selecting examples and explanations together substantially improves performance over selecting examples alone. Finally, even untuned explanations outperform carefully matched controls, suggesting that the benefits are due to the link between an example and its explanation, rather than lower-level features. However, only large models benefit. In summary, explanations can support the in-context learning of large LMs on challenging tasks."
                },
                {
                    "title": "STaR: Bootstrapping Reasoning With Reasoning",
                    "abstract": "Generating step-by-step\"chain-of-thought\"rationales improves language model performance on complex reasoning tasks like mathematics or commonsense question-answering. However, inducing language model rationale generation currently requires either constructing massive rationale datasets or sacrificing accuracy by using only few-shot inference. We propose a technique to iteratively leverage a small number of rationale examples and a large dataset without rationales, to bootstrap the ability to perform successively more complex reasoning. This technique, the\"Self-Taught Reasoner\"(STaR), relies on a simple loop: generate rationales to answer many questions, prompted with a few rationale examples; if the generated answers are wrong, try again to generate a rationale given the correct answer; fine-tune on all the rationales that ultimately yielded correct answers; repeat. We show that STaR significantly improves performance on multiple datasets compared to a model fine-tuned to directly predict final answers, and performs comparably to fine-tuning a 30$\\times$ larger state-of-the-art language model on CommensenseQA. Thus, STaR lets a model improve itself by learning from its own generated reasoning."
                },
                {
                    "title": "CrossBeam: Learning to Search in Bottom-Up Program Synthesis",
                    "abstract": "Many approaches to program synthesis perform a search within an enormous space of programs to find one that satisfies a given specification. Prior works have used neural models to guide combinatorial search algorithms, but such approaches still explore a huge portion of the search space and quickly become intractable as the size of the desired program increases. To tame the search space blowup, we propose training a neural model to learn a hands-on search policy for bottom-up synthesis, instead of relying on a combinatorial search algorithm. Our approach, called CrossBeam, uses the neural model to choose how to combine previously-explored programs into new programs, taking into account the search history and partial program executions. Motivated by work in structured prediction on learning to search, CrossBeam is trained on-policy using data extracted from its own bottom-up searches on training tasks. We evaluate CrossBeam in two very different domains, string manipulation and logic programming. We observe that CrossBeam learns to search efficiently, exploring much smaller portions of the program space compared to the state-of-the-art."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Jigsaw: Large Language Models meet Program Synthesis",
                    "abstract": "Large pre-trained language models such as GPT-3 [10], Codex [11], and Coogle's language model [7] are now capable of generating code from natural language specifications of programmer intent. We view these developments with a mixture of optimism and caution. On the optimistic side, such large language models have the potential to improve productivity by providing an automated AI pair programmer for every programmer in the world. On the cautionary side, since these large language models do not understand program semantics, they offer no guarantees about quality of the suggested code. In this paper, we present an approach to augment these large language models with post-processing steps based on program analysis and synthesis techniques, that understand the syntax and semantics of programs. Further, we show that such techniques can make use of user feedback and improve with usage. We present our experiences from building and evaluating such a tool Jigsaw, targeted at synthesizing code for using Python Pandas API using multi-modal inputs. Our experience suggests that as these large language models evolve for synthesizing code from intent, Jigsaw has an important role to play in improving the accuracy of the systems."
                },
                {
                    "title": "Show Your Work: Scratchpads for Intermediate Computation with Language Models",
                    "abstract": "Large pre-trained language models perform remarkably well on tasks that can be done\"in one pass\", such as generating realistic text or synthesizing computer programs. However, they struggle with tasks that require unbounded multi-step computation, such as adding integers or executing programs. Surprisingly, we find that these same models are able to perform complex multi-step computations -- even in the few-shot regime -- when asked to perform the operation\"step by step\", showing the results of intermediate computations. In particular, we train transformers to perform multi-step computations by asking them to emit intermediate computation steps into a\"scratchpad\". On a series of increasingly complex tasks ranging from long addition to the execution of arbitrary programs, we show that scratchpads dramatically improve the ability of language models to perform multi-step computations."
                },
                {
                    "title": "Few-Shot Self-Rationalization with Natural Language Prompts",
                    "abstract": "Self-rationalization models that predict task labels and generate free-text elaborations for their predictions could enable more intuitive interaction with NLP systems. These models are, however, currently trained with a large amount of human-written free-text explanations for each task which hinders their broader usage. We propose to study a more realistic setting of self-rationalization using few training examples. We present FEB -- a standardized collection of four existing English-language datasets and associated metrics. We identify the right prompting approach by extensively exploring natural language prompts on FEB. Then, by using this prompt and scaling the model size, we demonstrate that making progress on few-shot self-rationalization is possible. We show there is still ample room for improvement in this task: the average plausibility of generated explanations assessed by human annotators is at most 51% (with GPT-3), while plausibility of human explanations is 76%. We hope that FEB and our proposed approach will spur the community to take on the few-shot self-rationalization challenge."
                },
                {
                    "title": "Program Synthesis with Large Language Models",
                    "abstract": "This paper explores the limits of the current generation of large language models for program synthesis in general purpose programming languages. We evaluate a collection of such models (with between 244M and 137B parameters) on two new benchmarks, MBPP and MathQA-Python, in both the few-shot and fine-tuning regimes. Our benchmarks are designed to measure the ability of these models to synthesize short Python programs from natural language descriptions. The Mostly Basic Programming Problems (MBPP) dataset contains 974 programming tasks, designed to be solvable by entry-level programmers. The MathQA-Python dataset, a Python version of the MathQA benchmark, contains 23914 problems that evaluate the ability of the models to synthesize code from more complex text. On both datasets, we find that synthesis performance scales log-linearly with model size. Our largest models, even without finetuning on a code dataset, can synthesize solutions to 59.6 percent of the problems from MBPP using few-shot learning with a well-designed prompt. Fine-tuning on a held-out portion of the dataset improves performance by about 10 percentage points across most model sizes. On the MathQA-Python dataset, the largest fine-tuned model achieves 83.8 percent accuracy. Going further, we study the model's ability to engage in dialog about code, incorporating human feedback to improve its solutions. We find that natural language feedback from a human halves the error rate compared to the model's initial prediction. Additionally, we conduct an error analysis to shed light on where these models fall short and what types of programs are most difficult to generate. Finally, we explore the semantic grounding of these models by fine-tuning them to predict the results of program execution. We find that even our best models are generally unable to predict the output of a program given a specific input."
                },
                {
                    "title": "Evaluating Large Language Models Trained on Code",
                    "abstract": "We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics."
                },
                {
                    "title": "Communicating Natural Programs to Humans and Machines",
                    "abstract": "The Abstraction and Reasoning Corpus (ARC) is a set of procedural tasks that tests an agent's ability to flexibly solve novel problems. While most ARC tasks are easy for humans, they are challenging for state-of-the-art AI. What makes building intelligent systems that can generalize to novel situations such as ARC difficult? We posit that the answer might be found by studying the difference of \\emph{language}: While humans readily generate and interpret instructions in a general language, computer systems are shackled to a narrow domain-specific language that they can precisely execute. We present LARC, the \\textit{Language-complete ARC}: a collection of natural language descriptions by a group of human participants who instruct each other on how to solve ARC tasks using language alone, which contains successful instructions for 88\\% of the ARC tasks. We analyze the collected instructions as `natural programs', finding that while they resemble computer programs, they are distinct in two ways: First, they contain a wide range of primitives; Second, they frequently leverage communicative strategies beyond directly executable codes. We demonstrate that these two distinctions prevent current program synthesis techniques from leveraging LARC to its full potential, and give concrete suggestions on how to build the next-generation program synthesizers."
                },
                {
                    "title": "Fast and flexible: Human program induction in abstract reasoning tasks",
                    "abstract": "The Abstraction and Reasoning Corpus (ARC) is a challenging program induction dataset that was recently proposed by Chollet (2019). Here, we report the first set of results collected from a behavioral study of humans solving a subset of tasks from ARC (40 out of 1000). Although this subset of tasks contains considerable variation, our results showed that humans were able to infer the underlying program and generate the correct test output for a novel test input example, with an average of 80% of tasks solved per participant, and with 65% of tasks being solved by more than 80% of participants. Additionally, we find interesting patterns of behavioral consistency and variability within the action sequences during the generation process, the natural language descriptions to describe the transformations for each task, and the errors people made. Our findings suggest that people can quickly and reliably determine the relevant features and properties of a task to compose a correct solution. Future modeling work could incorporate these findings, potentially by connecting the natural language descriptions we collected here to the underlying semantics of ARC."
                },
                {
                    "title": "BUSTLE: Bottom-up program-Synthesis Through Learning-guided Exploration",
                    "abstract": "Program synthesis is challenging largely because of the difficulty of search in a large space of programs. Human programmers routinely tackle the task of writing complex programs by writing sub-programs and then analysing their intermediate results to compose them in appropriate ways. Motivated by this intuition, we present a new synthesis approach that leverages learning to guide a bottom-up search over programs. In particular, we train a model to prioritize compositions of intermediate values during search conditioned on a given set of input-output examples. This is a powerful combination because of several emergent properties: First, in bottom-up search, intermediate programs can be executed, providing semantic information to the neural network. Second, given the concrete values from those executions, we can exploit rich features based on recent work on property signatures. Finally, bottom-up search allows the system substantial flexibility in what order to generate the solution, allowing the synthesizer to build up a program from multiple smaller sub-programs. Overall, our empirical evaluation finds that the combination of learning and bottom-up search is remarkably effective, even with simple supervised learning approaches. We demonstrate the effectiveness of our technique on a new data set for synthesis of string transformation programs."
                },
                {
                    "title": "DreamCoder: growing generalizable, interpretable knowledge with wake\u2013sleep Bayesian program learning",
                    "abstract": "Expert problem-solving is driven by powerful languages for thinking about problems and their solutions. Acquiring expertise means learning these languages\u2014systems of concepts, alongside the skills to use them. We present DreamCoder, a system that learns to solve problems by writing programs. It builds expertise by creating domain-specific programming languages for expressing domain concepts, together with neural networks to guide the search for programs within these languages. A \u2018wake\u2013sleep\u2019 learning algorithm alternately extends the language with new symbolic abstractions and trains the neural network on imagined and replayed problems. DreamCoder solves both classic inductive programming tasks and creative tasks such as drawing pictures and building scenes. It rediscovers the basics of modern functional programming, vector algebra and classical physics, including Newton\u2019s and Coulomb\u2019s laws. Concepts are built compositionally from those learned earlier, yielding multilayered symbolic representations that are interpretable and transferrable to new tasks, while still growing scalably and flexibly with experience. This article is part of a discussion meeting issue \u2018Cognitive artificial intelligence\u2019."
                },
                {
                    "title": "Unsupervised Commonsense Question Answering with Self-Talk",
                    "abstract": "Natural language understanding involves reading between the lines with implicit background knowledge. Current systems either rely on pre-trained language models as the sole implicit source of world knowledge, or resort to external knowledge bases (KBs) to incorporate additional relevant knowledge. We propose an unsupervised framework based on \\emph{self-talk} as a novel alternative to multiple-choice commonsense tasks. Inspired by inquiry-based discovery learning (Bruner, 1961), our approach inquires language models with a number of information seeking questions such as \"$\\textit{what is the definition of ...}$\" to discover additional background knowledge. Empirical results demonstrate that the self-talk procedure substantially improves the performance of zero-shot language model baselines on four out of six commonsense benchmarks, and competes with models that obtain knowledge from external KBs. While our approach improves performance on several benchmarks, the self-talk induced knowledge even when leading to correct answers is not always seen as useful by human judges, raising interesting questions about the inner-workings of pre-trained language models for commonsense reasoning."
                },
                {
                    "title": "Write, Execute, Assess: Program Synthesis with a REPL",
                    "abstract": "We present a neural program synthesis approach integrating components which write, execute, and assess code to navigate the search space of possible programs. We equip the search process with an interpreter or a read-eval-print-loop (REPL), which immediately executes partially written programs, exposing their semantics. The REPL addresses a basic challenge of program synthesis: tiny changes in syntax can lead to huge changes in semantics. We train a pair of models, a policy that proposes the new piece of code to write, and a value function that assesses the prospects of the code written so-far. At test time we can combine these models with a Sequential Monte Carlo algorithm. We apply our approach to two domains: synthesizing text editing programs and inferring 2D and 3D graphics programs."
                },
                {
                    "title": "Explain Yourself! Leveraging Language Models for Commonsense Reasoning",
                    "abstract": "Deep learning models perform poorly on tasks that require commonsense reasoning, which often necessitates some form of world-knowledge or reasoning over information not immediately present in the input. We collect human explanations for commonsense reasoning in the form of natural language sequences and highlighted annotations in a new dataset called Common Sense Explanations (CoS-E). We use CoS-E to train language models to automatically generate explanations that can be used during training and inference in a novel Commonsense Auto-Generated Explanation (CAGE) framework. CAGE improves the state-of-the-art by 10% on the challenging CommonsenseQA task. We further study commonsense reasoning in DNNs using both human and auto-generated explanations including transfer to out-of-domain tasks. Empirical results indicate that we can effectively leverage language models for commonsense reasoning."
                },
                {
                    "title": "The Neuro-Symbolic Concept Learner: Interpreting Scenes, Words, and Sentences From Natural Supervision",
                    "abstract": "We propose the Neuro-Symbolic Concept Learner (NS-CL), a model that learns visual concepts, words, and semantic parsing of sentences without explicit supervision on any of them; instead, our model learns by simply looking at images and reading paired questions and answers. Our model builds an object-based scene representation and translates sentences into executable, symbolic programs. To bridge the learning of two modules, we use a neuro-symbolic reasoning module that executes these programs on the latent scene representation. Analogical to human concept learning, the perception module learns visual concepts based on the language description of the object being referred to. Meanwhile, the learned visual concepts facilitate learning new words and parsing new sentences. We use curriculum learning to guide the searching over the large compositional space of images and language. Extensive experiments demonstrate the accuracy and efficiency of our model on learning visual concepts, word representations, and semantic parsing of sentences. Further, our method allows easy generalization to new object attributes, compositions, language concepts, scenes and questions, and even new program domains. It also empowers applications including visual question answering and bidirectional image-text retrieval."
                },
                {
                    "title": "Neural Program Repair by Jointly Learning to Localize and Repair",
                    "abstract": "Due to its potential to improve programmer productivity and software quality, automated program repair has been an active topic of research. Newer techniques harness neural networks to learn directly from examples of buggy programs and their fixes. In this work, we consider a recently identified class of bugs called variable-misuse bugs. The state-of-the-art solution for variable misuse enumerates potential fixes for all possible bug locations in a program, before selecting the best prediction. We show that it is beneficial to train a model that jointly and directly localizes and repairs variable-misuse bugs. We present multi-headed pointer networks for this purpose, with one head each for localization and repair. The experimental results show that the joint model significantly outperforms an enumerative solution that uses a pointer based model for repair alone."
                },
                {
                    "title": "CSGNet: Neural Shape Parser for Constructive Solid Geometry",
                    "abstract": "We present a neural architecture that takes as input a 2D or 3D shape and outputs a program that generates the shape. The instructions in our program are based on constructive solid geometry principles, i.e., a set of boolean operations on shape primitives defined recursively. Bottom-up techniques for this shape parsing task rely on primitive detection and are inherently slow since the search space over possible primitive combinations is large. In contrast, our model uses a recurrent neural network that parses the input shape in a top-down manner, which is significantly faster and yields a compact and easy-to-interpret sequence of modeling instructions. Our model is also more effective as a shape detector compared to existing state-of-the-art detection techniques. We finally demonstrate that our network can be trained on novel datasets without ground-truth program annotations through policy gradient techniques."
                },
                {
                    "title": "Learning with Latent Language",
                    "abstract": "The named concepts and compositional operators present in natural language provide a rich source of information about the abstractions humans use to navigate the world. Can this linguistic background knowledge improve the generality and efficiency of learned classifiers and control policies? This paper aims to show that using the space of natural language strings as a parameter space is an effective way to capture natural task structure. In a pretraining phase, we learn a language interpretation model that transforms inputs (e.g. images) into outputs (e.g. labels) given natural language descriptions. To learn a new concept (e.g. a classifier), we search directly in the space of descriptions to minimize the interpreter\u2019s loss on training examples. Crucially, our models do not require language data to learn these concepts: language is used only in pretraining to impose structure on subsequent learning. Results on image classification, text editing, and reinforcement learning show that, in all settings, models with a linguistic parameterization outperform those without."
                },
                {
                    "title": "Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning",
                    "abstract": "Relational databases store a significant amount of the worlds data. However, accessing this data currently requires users to understand a query language such as SQL. We propose Seq2SQL, a deep neural network for translating natural language questions to corresponding SQL queries. Our model uses rewards from in the loop query execution over the database to learn a policy to generate the query, which contains unordered parts that are less suitable for optimization via cross entropy loss. Moreover, Seq2SQL leverages the structure of SQL to prune the space of generated queries and significantly simplify the generation problem. In addition to the model, we release WikiSQL, a dataset of 80654 hand-annotated examples of questions and SQL queries distributed across 24241 tables fromWikipedia that is an order of magnitude larger than comparable datasets. By applying policy based reinforcement learning with a query execution environment to WikiSQL, Seq2SQL outperforms a state-of-the-art semantic parser, improving execution accuracy from 35.9% to 59.4% and logical form accuracy from 23.4% to 48.3%."
                },
                {
                    "title": "From Language to Programs: Bridging Reinforcement Learning and Maximum Marginal Likelihood",
                    "abstract": "Our goal is to learn a semantic parser that maps natural language utterances into executable programs when only indirect supervision is available: examples are labeled with the correct execution result, but not the program itself. Consequently, we must search the space of programs for those that output the correct result, while not being misled by spurious programs: incorrect programs that coincidentally output the correct result. We connect two common learning paradigms, reinforcement learning (RL) and maximum marginal likelihood (MML), and then present a new learning algorithm that combines the strengths of both. The new algorithm guards against spurious programs by combining the systematic search traditionally employed in MML with the randomized exploration of RL, and by updating parameters such that probability is spread more evenly across consistent programs. We apply our learning algorithm to a new neural semantic parser and show significant gains over existing state-of-the-art results on a recent context-dependent semantic parsing task."
                },
                {
                    "title": "RobustFill: Neural Program Learning under Noisy I/O",
                    "abstract": "The problem of automatically generating a computer program from some specification has been studied since the early days of AI. Recently, two competing approaches for automatic program learning have received significant attention: (1) neural program synthesis, where a neural network is conditioned on input/output (I/O) examples and learns to generate a program, and (2) neural program induction, where a neural network generates new outputs directly using a latent program representation. Here, for the first time, we directly compare both approaches on a large-scale, real-world learning task and we additionally contrast to rule-based program synthesis, which uses hand-crafted semantics to guide the program generation. Our neural models use a modified attention RNN to allow encoding of variable-sized sets of I/O pairs, which achieve 92% accuracy on a real-world test set, compared to the 34% accuracy of the previous best neural synthesis approach. The synthesis model also outperforms a comparable induction model on this task, but we more importantly demonstrate that the strength of each approach is highly dependent on the evaluation metric and end-user application. Finally, we show that we can train our neural models to remain very robust to the type of noise expected in real-world data (e.g., typos), while a highly-engineered rule-based system fails entirely."
                },
                {
                    "title": "Interactive Program Synthesis",
                    "abstract": "Program synthesis from incomplete specifications (e.g. input-output examples) has gained popularity and found real-world applications, primarily due to its ease-of-use. Since this technology is often used in an interactive setting, efficiency and correctness are often the key user expectations from a system based on such technologies. Ensuring efficiency is challenging since the highly combinatorial nature of program synthesis algorithms does not fit in a 1-2 second response expectation of a user-facing system. Meeting correctness expectations is also difficult, given that the specifications provided are incomplete, and that the users of such systems are typically non-programmers. \nIn this paper, we describe how interactivity can be leveraged to develop efficient synthesis algorithms, as well as to decrease the cognitive burden that a user endures trying to ensure that the system produces the desired program. We build a formal model of user interaction along three dimensions: incremental algorithm, step-based problem formulation, and feedback-based intent refinement. We then illustrate the effectiveness of each of these forms of interactivity with respect to synthesis performance and correctness on a set of real-world case studies."
                },
                {
                    "title": "Learning to Compose Neural Networks for Question Answering",
                    "abstract": "We describe a question answering model that applies to both images and structured knowledge bases. The model uses natural language strings to automatically assemble neural networks from a collection of composable modules. Parameters for these modules are learned jointly with network-assembly parameters via reinforcement learning, with only (world, question, answer) triples as supervision. Our approach, which we term a dynamic neural model network, achieves state-of-the-art results on benchmark datasets in both visual and structured domains."
                },
                {
                    "title": "Language to Logical Form with Neural Attention",
                    "abstract": "Semantic parsing aims at mapping natural language to machine interpretable meaning representations. Traditional approaches rely on high-quality lexicons, manually-built templates, and linguistic features which are either domain- or representation-specific. In this paper we present a general method based on an attention-enhanced encoder-decoder model. We encode input utterances into vector representations, and generate their logical forms by conditioning the output sequences or trees on the encoding vectors. Experimental results on four datasets show that our approach performs competitively without using hand-engineered features and is easy to adapt across domains and meaning representations."
                },
                {
                    "title": "Program Synthesis Using Natural Language",
                    "abstract": "Interacting with computers is a ubiquitous activity for millions of people. Repetitive or specialized tasks often require creation of small, often one-off, programs. End-users struggle with learning and using the myriad of domain-specific languages (DSLs) to effectively accomplish these tasks. We present a general framework for constructing program synthesizers that take natural language (NL) inputs and produce expressions in a target DSL. The framework takes as input a DSL definition and training data consisting of NL/DSL pairs. From these it constructs a synthesizer by learning optimal weights and classifiers (using NLP features) that rank the outputs of a keyword-programming based translation. We applied our framework to three domains: repetitive text editing, an intelligent tutoring system, and flight information queries. On 1200+ English descriptions, the respective synthesizers rank the desired program as the top-1 and top-3 for 80% and 90% descriptions respectively."
                },
                {
                    "title": "One and Done? Optimal Decisions From Very Few Samples",
                    "abstract": "In many learning or inference tasks human behavior approximates that of a Bayesian ideal observer, suggesting that, at some level, cognition can be described as Bayesian inference. However, a number of findings have highlighted an intriguing mismatch between human behavior and standard assumptions about optimality: People often appear to make decisions based on just one or a few samples from the appropriate posterior probability distribution, rather than using the full distribution. Although sampling-based approximations are a common way to implement Bayesian inference, the very limited numbers of samples often used by humans seem insufficient to approximate the required probability distributions very accurately. Here, we consider this discrepancy in the broader framework of statistical decision theory, and ask: If people are making decisions based on samples--but as samples are costly--how many samples should people use to optimize their total expected or worst-case reward over a large number of decisions? We find that under reasonable assumptions about the time costs of sampling, making many quick but locally suboptimal decisions based on very few samples may be the globally optimal strategy over long periods. These results help to reconcile a large body of work showing sampling-based or probability matching behavior with the hypothesis that human cognition can be understood in Bayesian terms, and they suggest promising future directions for studies of resource-constrained cognition."
                },
                {
                    "title": "Building Statistical Language Models of code",
                    "abstract": "We present the Source Code Statistical Language Model data analysis pattern. Statistical language models have been an enabling tool for a wide array of important language technologies. Speech recognition, machine translation, and document summarization (to name a few) all rely on statistical language models to assign probability estimates to natural language utterances or sentences. In this data analysis pattern, we describe the process of building n-gram language models over software source files. We hope that by introducing the empirical software engineering community to best practices that have been established over the years in research for natural languages, statistical language models can become a tool that SE researchers are able to use to explore new research directions."
                },
                {
                    "title": "How to Grow a Mind: Statistics, Structure, and Abstraction",
                    "abstract": "In coming to understand the world\u2014in learning concepts, acquiring language, and grasping causal relations\u2014our minds make inferences that appear to go far beyond the data available. How do we do it? This review describes recent approaches to reverse-engineering human learning and cognitive development and, in parallel, engineering more humanlike machine learning systems. Computational models that perform probabilistic inference over hierarchies of flexibly structured representations can address some of the deepest questions about the nature and origins of human thought: How does abstract knowledge guide learning and reasoning from sparse data? What forms does our knowledge take, across different domains and tasks? And how is that abstract knowledge itself acquired?"
                },
                {
                    "title": "Automated program repair through the evolution of assembly code",
                    "abstract": "A method is described for automatically repairing legacy software at the assembly code level using evolutionary computation. The technique is demonstrated on Java byte code and x86 assembly programs, showing how to find program variations that correct defects while retaining desired behavior. Test cases are used to demonstrate the defect and define required functionality. The paper explores advantages of assembly-level repair over earlier work at the source code level - the ability to repair programs written in many different languages; and the ability to repair bugs that were previously intractable. The paper reports experimental results showing reasonable performance of assembly language repair even on non-trivial programs"
                },
                {
                    "title": "Code template inference using language models",
                    "abstract": "This paper investigates the use of a natural language processing technique that automatically detects project-specific code templates (i.e., frequently used code blocks), which can be made available to software developers within an integrated development environment. During software development, programmers often and in some cases unknowingly rewrite the same code block that represents some functionality. These frequently used code blocks can inform the existence and possible use of code templates. Many existing code editors support code templates, but programmers are expected to manually define these templates and subsequently add them as templates in the editor. Furthermore, the support of editors to provide templates based on the editing context is still limited. The use of n-gram language models within the context of software development is described and evaluated to overcome these restrictions. The technique can search for project-specific code templates and present these templates to the programmer based on the current editing context."
                },
                {
                    "title": "A Rational Analysis of Rule-Based Concept Learning",
                    "abstract": "This article proposes a new model of human concept learning that provides a rational analysis of learning feature-based concepts. This model is built upon Bayesian inference for a grammatically structured hypothesis space-a concept language of logical rules. This article compares the model predictions to human generalization judgments in several well-known category learning experiments, and finds good agreement for both average and individual participant generalizations. This article further investigates judgments for a broad set of 7-feature concepts-a more natural setting in several ways-and again finds that the model explains human performance."
                },
                {
                    "title": "ROUGE: A Package for Automatic Evaluation of Summaries",
                    "abstract": "ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST."
                },
                {
                    "title": "Analogical Reasoning and Computer Programming",
                    "abstract": "Investigations of the cognitive demands of programming can inform teaching and validate claims that important cognitive skills are inherent in programming. Given reports of experts' use of analogical problem solving in programming, the study reported here related analogical reasoning to Logo programming mastery among high school students. Correlational analyses related pretests of analogical reasoning to posttests of programming mastery. As predicted, a significant correlation was found between analogical reasoning and the ability to write subprocedures which can be reused for several different programs. This sophisticated programming skill requires recognition of structural similarities among distinct programming tasks. A final, general discussion considers analogical reasoning skill as a cognitive demand and consequence of programming."
                },
                {
                    "title": "Using Large Language Models to Simulate Multiple Humans",
                    "abstract": "We propose a method for using a large language model, such as GPT-3, to simulate responses of different humans in a given context. We test our method by attempting to reproduce well-established economic, psycholinguistic, and social experiments. The method requires prompt templates for each experiment. Simulations are run by varying the (hypothetical) subject details, such as name, and analyzing the text generated by the language model. To validate our methodology, we use GPT-3 to simulate the Ultimatum Game , garden path sentences , risk aversion , and the Milgram Shock experiments. In order to address concerns of exposure to these studies in training data, we also evaluate simulations on novel variants of these studies. We show that it is possible to simulate responses of different people and that their responses are consistent with prior human studies from the literature. Across all studies, the distributions generated by larger language models better align with prior experimental results, suggesting a trend that future language models may be used for even more faithful simulations of human responses. Our use of a language model for simulation is contrasted with anthropomorphic views of a language model as having its own behavior."
                },
                {
                    "title": "A Bidirectional LSTM Language Model for Code Evaluation and Repair",
                    "abstract": "Programming is a vital skill in computer science and engineering-related disciplines. However, developing source code is an error-prone task. Logical errors in code are particularly hard to identify for both students and professionals, and a single error is unexpected to end-users. At present, conventional compilers have difficulty identifying many of the errors (especially logical errors) that can occur in code. To mitigate this problem, we propose a language model for evaluating source codes using a bidirectional long short-term memory (BiLSTM) neural network. We trained the BiLSTM model with a large number of source codes with tuning various hyperparameters. We then used the model to evaluate incorrect code and assessed the model\u2019s performance in three principal areas: source code error detection, suggestions for incorrect code repair, and erroneous code classification. Experimental results showed that the proposed BiLSTM model achieved 50.88% correctness in identifying errors and providing suggestions. Moreover, the model achieved an F-score of approximately 97%, outperforming other state-of-the-art models (recurrent neural networks (RNNs) and long short-term memory (LSTM))."
                },
                {
                    "title": "Questions Concerning Certain Faculties Claimed for Man",
                    "abstract": "Throughout this paper, the term intuition will be taken as signifying a cognition not determined by a previous cognition of the same object, and therefore so determined by something out of the consciousness.1 Let me request the reader to note this. Intuition here will be nearly the same as \"premiss not itself a conclusion\"; the only difference being that premisses and conclusions are judgments, whereas an intuition may, as far as its definition states, be any kind of cognition whatever. But just as a conclusion (good or bad) is determined in the mind of the reasoner by its premiss, so cognitions not judgments may be determined by previous cognitions; and a (W2.194) cognition not so determined, and therefore determined directly by the transcendental object, is to be termed an intuition."
                },
                {
                    "title": "Generating computer programs from natural language descriptions",
                    "abstract": "This thesis addresses the problem of learning to\ntranslate natural language into preexisting programming languages\nsupported by widely-deployed computer systems. Generating programs\nfor existing computer systems enables us to take advantage of two\nimportant capabilities of these systems: computing the semantic\nequivalence between programs, and executing the programs to obtain\na result. We present probabilistic models and inference algorithms\nwhich integrate these capabilities into the learning process. We\nuse these to build systems that learn to generate programs from\nnatural language in three different computing domains: text\nprocessing, solving math problems, and performing robotic tasks in\na virtual world. In all cases the resulting systems provide\nsignificant performance gains over strong baselines which do not\nexploit the underlying system capabilities to help interpret the\ntext."
                },
                {
                    "title": "Learning a theory of causality.",
                    "abstract": "The very early appearance of abstract knowledge is often taken as evidence for innateness. We explore the relative learning speeds of abstract and specific knowledge within a Bayesian framework and the role for innate structure. We focus on knowledge about causality, seen as a domain-general intuitive theory, and ask whether this knowledge can be learned from co-occurrence of events. We begin by phrasing the causal Bayes nets theory of causality and a range of alternatives in a logical language for relational theories. This allows us to explore simultaneous inductive learning of an abstract theory of causality and a causal model for each of several causal systems. We find that the correct theory of causality can be learned relatively quickly, often becoming available before specific causal theories have been learned--an effect we term the blessing of abstraction. We then explore the effect of providing a variety of auxiliary evidence and find that a collection of simple perceptual input analyzers can help to bootstrap abstract knowledge. Together, these results suggest that the most efficient route to causal knowledge may be to build in not an abstract notion of causality but a powerful inductive learning mechanism and a variety of perceptual supports. While these results are purely computational, they have implications for cognitive development, which we explore in the conclusion."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we enhance the inductive reasoning capabilities of large language models (LLMs) to perform better on complex reasoning tasks, such as those found in the Abstraction and Reasoning Corpus (ARC)?\n\n**[Question 2] - Why is it interesting and important?**  \nImproving the inductive reasoning abilities of LLMs has significant implications for the research community, as it could lead to advancements in artificial intelligence that more closely mimic human cognitive processes. This could open new avenues for research in cognitive science, machine learning, and natural language processing. By addressing this problem, we could develop models that not only perform better on complex reasoning tasks but also contribute to practical applications in areas such as automated reasoning, decision-making systems, and human-computer interaction.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of inductive reasoning tasks, which often require multi-level abstraction and the ability to generalize from limited examples. Naive approaches may fail because they do not account for the nuanced relationships between concepts or the need for hierarchical reasoning. Technical obstacles include the limitations of current LLM architectures in handling abstract reasoning and the difficulty in generating and evaluating hypotheses at multiple levels of abstraction.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on straightforward inductive tasks, often overlooking the complexities involved in more challenging scenarios like those presented in the ARC. Existing solutions may lack the necessary framework for multi-level abstraction or fail to generate explicit hypotheses effectively. Our approach differs by explicitly prompting LLMs to generate multiple abstract reasoning tasks, thereby addressing the limitations of prior work and providing a structured method for enhancing reasoning capabilities.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves prompting LLMs to generate explicit hypotheses at multiple levels of abstraction when faced with complex reasoning tasks. We will utilize the Abstraction and Reasoning Corpus (ARC) as our primary dataset and evaluate performance using metrics such as accuracy and generalization ability. The expected outcomes include improved performance of LLMs on complex reasoning tasks, demonstrating their enhanced inductive reasoning capabilities and providing insights into the underlying mechanisms of reasoning in language models."
            }
        },
        "author_data": {
            "66980ea8-94bf-4520-afcf-3cd0746f0dc8": {
                "pk": "66980ea8-94bf-4520-afcf-3cd0746f0dc8",
                "name": "Ruocheng Wang",
                "collaborators": [
                    "Yunzhi Zhang",
                    "Jiayuan Mao",
                    "Jiajun Wu",
                    "Pei Xu",
                    "Albert Wu",
                    "Sirui Chen",
                    "Clemens Eppner",
                    "C. K. Liu",
                    "Chin-Yi Cheng",
                    "S. Gershman"
                ],
                "domain": [
                    "Robotics",
                    "Machine Learning",
                    "Computer Vision",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Synchronize Dual Hands for Physics-Based Dexterous Guitar Playing",
                        "abstract": "We present a novel approach to synthesize dexterous motions for physically simulated hands in tasks that require coordination between the control of two hands with high temporal precision. Instead of directly learning a joint policy to control two hands, our approach performs bimanual control through cooperative learning where each hand is treated as an individual agent. The individual policies for each hand are first trained separately, and then synchronized through latent space manipulation in a centralized environment to serve as a joint policy for two-hand control. By doing so, we avoid directly performing policy learning in the joint state-action space of two hands with higher dimensions, greatly improving the overall training efficiency. We demonstrate the effectiveness of our proposed approach in the challenging guitar-playing task. The virtual guitarist trained by our approach can synthesize motions from unstructured reference data of general guitar-playing practice motions, and accurately play diverse rhythms with complex chord pressing and string picking patterns based on the input guitar tabs that do not exist in the references. Along with this paper, we provide the motion capture data that we collected as the reference for policy training. Code is available at: https://pei-xu.github.io/guitar."
                    },
                    {
                        "title": "One-Shot Transfer of Long-Horizon Extrinsic Manipulation Through Contact Retargeting",
                        "abstract": "Extrinsic manipulation, the use of environment contacts to achieve manipulation objectives, enables strategies that are otherwise impossible with a parallel jaw gripper. However, orchestrating a long-horizon sequence of contact interactions between the robot, object, and environment is notoriously challenging due to the scene diversity, large action space, and difficult contact dynamics. We observe that most extrinsic manipulation are combinations of short-horizon primitives, each of which depend strongly on initializing from a desirable contact configuration to succeed. Therefore, we propose to generalize one extrinsic manipulation trajectory to diverse objects and environments by retargeting contact requirements. We prepare a single library of robust short-horizon, goal-conditioned primitive policies, and design a framework to compose state constraints stemming from contacts specifications of each primitive. Given a test scene and a single demo prescribing the primitive sequence, our method enforces the state constraints on the test scene and find intermediate goal states using inverse kinematics. The goals are then tracked by the primitive policies. Using a 7+1 DoF robotic arm-gripper system, we achieved an overall success rate of 80.5% on hardware over 4 long-horizon extrinsic manipulation tasks, each with up to 4 primitives. Our experiments cover 10 objects and 6 environment configurations. We further show empirically that our method admits a wide range of demonstrations, and that contact retargeting is indeed the key to successfully combining primitives for long-horizon extrinsic manipulation. Code and additional details are available at stanford-tml.github.io/extrinsic-manipulation."
                    },
                    {
                        "title": "IKEA-Manual: Seeing Shape Assembly Step by Step",
                        "abstract": "Human-designed visual manuals are crucial components in shape assembly activities. They provide step-by-step guidance on how we should move and connect different parts in a convenient and physically-realizable way. While there has been an ongoing effort in building agents that perform assembly tasks, the information in human-design manuals has been largely overlooked. We identify that this is due to 1) a lack of realistic 3D assembly objects that have paired manuals and 2) the difficulty of extracting structured information from purely image-based manuals. Motivated by this observation, we present IKEA-Manual, a dataset consisting of 102 IKEA objects paired with assembly manuals. We provide fine-grained annotations on the IKEA objects and assembly manuals, including decomposed assembly parts, assembly plans, manual segmentation, and 2D-3D correspondence between 3D parts and visual manuals. We illustrate the broad application of our dataset on four tasks related to shape assembly: assembly plan generation, part segmentation, pose estimation, and 3D part assembly."
                    },
                    {
                        "title": "Translating a Visual LEGO Manual to a Machine-Executable Plan",
                        "abstract": "We study the problem of translating an image-based, step-by-step assembly manual created by human designers into machine-interpretable instructions. We formulate this problem as a sequential prediction task: at each step, our model reads the manual, locates the components to be added to the current shape, and infers their 3D poses. This task poses the challenge of establishing a 2D-3D correspondence between the manual image and the real 3D object, and 3D pose estimation for unseen 3D objects, since a new component to be added in a step can be an object built from previous steps. To address these two challenges, we present a novel learning-based framework, the Manual-to-Executable-Plan Network (MEPNet), which reconstructs the assembly steps from a sequence of manual images. The key idea is to integrate neural 2D keypoint detection modules and 2D-3D projection algorithms for high-precision prediction and strong generalization to unseen components. The MEPNet outperforms existing methods on three newly collected LEGO manual datasets and a Minecraft house dataset."
                    },
                    {
                        "title": "Language-Mediated, Object-Centric Representation Learning",
                        "abstract": "We present Language-mediated, Object-centric Representation Learning (LORL), a paradigm for learning disentangled, object-centric scene representations from vision and language. LORL builds upon recent advances in unsupervised object discovery and segmentation, notably MONet and Slot Attention. While these algorithms learn an object-centric representation just by reconstructing the input image, LORL enables them to further learn to associate the learned representations to concepts, i.e., words for object categories, properties, and spatial relationships, from language input. These object-centric concepts derived from language facilitate the learning of object-centric representations. LORL can be integrated with various unsupervised object discovery algorithms that are language-agnostic. Experiments show that the integration of LORL consistently improves the performance of unsupervised object discovery methods on two datasets via the help of language. We also show that concepts learned by LORL, in conjunction with object discovery methods, aid downstream tasks such as referring expression comprehension."
                    },
                    {
                        "title": "Legal Summarization for Multi-role Debate Dialogue via Controversy Focus Mining and Multi-task Learning",
                        "abstract": "Multi-role court debate is a critical component in a civil trial where parties from different camps (plaintiff, defendant, witness, judge, etc.) actively involved. Unlike other types of dialogue, court debate can be lengthy, and important information, with respect to the controversy focus(es), often hides within the redundant and colloquial dialogue data. Summarizing court debate can be a novel but significant task to assist judge to effectively make the legal decision for the target trial. In this work, we propose an innovative end-to-end model to address this problem. Unlike prior summarization efforts, the proposed model projects the multi-role debate into the controversy focus space, which enables high-quality essential utterance(s) extraction in terms of legal knowledge and judicial factors. An extensive set of experiments with a large civil trial dataset shows that the proposed model can provide more accurate and readable summarization against several alternatives in the multi-role court debate scene."
                    }
                ]
            },
            "d7294ac4-7843-49c1-b8e1-beb4a12126ac": {
                "pk": "d7294ac4-7843-49c1-b8e1-beb4a12126ac",
                "name": "Eric Zelikman",
                "collaborators": [
                    "Noah D. Goodman",
                    "Percy Liang",
                    "Qian Huang",
                    "Kanishk Gandhi",
                    "Yuhuai Wu",
                    "Nick Haber",
                    "Esin Durmus",
                    "Shibani Santurkar",
                    "Faisal Ladhak",
                    "Tatsunori Hashimoto"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Evaluation Metrics",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "PERSONA: A Reproducible Testbed for Pluralistic Alignment",
                        "abstract": "The rapid advancement of language models (LMs) necessitates robust alignment with diverse user values. However, current preference optimization approaches often fail to capture the plurality of user opinions, instead reinforcing majority viewpoints and marginalizing minority perspectives. We introduce PERSONA, a reproducible test bed designed to evaluate and improve pluralistic alignment of LMs. We procedurally generate diverse user profiles from US census data, resulting in 1,586 synthetic personas with varied demographic and idiosyncratic attributes. We then generate a large-scale evaluation dataset containing 3,868 prompts and 317,200 feedback pairs obtained from our synthetic personas. Leveraging this dataset, we systematically evaluate LM capabilities in role-playing diverse users, verified through human judges, and the establishment of both a benchmark, PERSONA Bench, for pluralistic alignment approaches as well as an extensive dataset to create new and future benchmarks. The full dataset and benchmarks are available here: https://www.synthlabs.ai/research/persona."
                    },
                    {
                        "title": "When Benchmarks are Targets: Revealing the Sensitivity of Large Language Model Leaderboards",
                        "abstract": "Large Language Model (LLM) leaderboards based on benchmark rankings are regularly used to guide practitioners in model selection. Often, the published leaderboard rankings are taken at face value - we show this is a (potentially costly) mistake. Under existing leaderboards, the relative performance of LLMs is highly sensitive to (often minute) details. We show that for popular multiple-choice question benchmarks (e.g., MMLU), minor perturbations to the benchmark, such as changing the order of choices or the method of answer selection, result in changes in rankings up to 8 positions. We explain this phenomenon by conducting systematic experiments over three broad categories of benchmark perturbations and identifying the sources of this behavior. Our analysis results in several best-practice recommendations, including the advantage of a hybrid scoring method for answer selection. Our study highlights the dangers of relying on simple benchmark evaluations and charts the path for more robust evaluation schemes on the existing benchmarks. The code for this paper is available at https://github.com/National-Center-for-AI-Saudi-Arabia/lm-evaluation-harness."
                    },
                    {
                        "title": "Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking",
                        "abstract": "When writing and talking, people sometimes pause to think. Although reasoning-focused works have often framed reasoning as a method of answering questions or completing agentic tasks, reasoning is implicit in almost all written text. For example, this applies to the steps not stated between the lines of a proof or to the theory of mind underlying a conversation. In the Self-Taught Reasoner (STaR, Zelikman et al. 2022), useful thinking is learned by inferring rationales from few-shot examples in question-answering and learning from those that lead to a correct answer. This is a highly constrained setting -- ideally, a language model could instead learn to infer unstated rationales in arbitrary text. We present Quiet-STaR, a generalization of STaR in which LMs learn to generate rationales at each token to explain future text, improving their predictions. We address key challenges, including 1) the computational cost of generating continuations, 2) the fact that the LM does not initially know how to generate or use internal thoughts, and 3) the need to predict beyond individual next tokens. To resolve these, we propose a tokenwise parallel sampling algorithm, using learnable tokens indicating a thought's start and end, and an extended teacher-forcing technique. Encouragingly, generated rationales disproportionately help model difficult-to-predict tokens and improve the LM's ability to directly answer difficult questions. In particular, after continued pretraining of an LM on a corpus of internet text with Quiet-STaR, we find zero-shot improvements on GSM8K (5.9%$\\rightarrow$10.9%) and CommonsenseQA (36.3%$\\rightarrow$47.2%) and observe a perplexity improvement of difficult tokens in natural text. Crucially, these improvements require no fine-tuning on these tasks. Quiet-STaR marks a step towards LMs that can learn to reason in a more general and scalable way."
                    },
                    {
                        "title": "Self-Supervised Alignment with Mutual Information: Learning to Follow Principles without Preference Labels",
                        "abstract": "When prompting a language model (LM), users often expect the model to adhere to a set of behavioral principles across diverse tasks, such as producing insightful content while avoiding harmful or biased language. Instilling such principles (i.e., a constitution) into a model is resource-intensive, technically challenging, and generally requires human preference labels or examples. We introduce SAMI, an iterative algorithm that finetunes a pretrained language model (without requiring preference labels or demonstrations) to increase the conditional mutual information between constitutions and self-generated responses given queries from a dataset. On single-turn dialogue and summarization, a SAMI-trained mistral-7b outperforms the initial pretrained model, with win rates between 66% and 77%. Strikingly, it also surpasses an instruction-finetuned baseline (mistral-7b-instruct) with win rates between 55% and 57% on single-turn dialogue. SAMI requires a model that writes the principles. To avoid dependence on strong models for writing principles, we align a strong pretrained model (mixtral-8x7b) using constitutions written by a weak instruction-finetuned model (mistral-7b-instruct), achieving a 65% win rate on summarization. Finally, we investigate whether SAMI generalizes to diverse summarization principles (e.g.,\"summaries should be scientific\") and scales to stronger models (llama3-70b), finding that it achieves win rates of up to 68% for learned and 67% for held-out principles compared to the base model. Our results show that a pretrained LM can learn to follow constitutions without using preference labels, demonstrations, or human oversight."
                    },
                    {
                        "title": "Generating and Evaluating Tests for K-12 Students with Language Model Simulations: A Case Study on Sentence Reading Efficiency",
                        "abstract": "Developing an educational test can be expensive and time-consuming, as each item must be written by experts and then evaluated by collecting hundreds of student responses. Moreover, many tests require multiple distinct sets of questions administered throughout the school year to closely monitor students' progress, known as parallel tests. In this study, we focus on tests of silent sentence reading efficiency, used to assess students' reading ability over time. To generate high-quality parallel tests, we propose to fine-tune large language models (LLMs) to simulate how previous students would have responded to unseen items. With these simulated responses, we can estimate each item's difficulty and ambiguity. We first use GPT-4 to generate new test items following a list of expert-developed rules and then apply a fine-tuned LLM to filter the items based on criteria from psychological measurements. We also propose an optimal-transport-inspired technique for generating parallel tests and show the generated tests closely correspond to the original test's difficulty and reliability based on crowdworker responses. Our evaluation of a generated test with 234 students from grades 2 to 8 produces test scores highly correlated (r=0.93) to those of a standard test form written by human experts and evaluated across thousands of K-12 students."
                    },
                    {
                        "title": "Self-Taught Optimizer (STOP): Recursively Self-Improving Code Generation",
                        "abstract": "Several recent advances in AI systems solve problems by providing a\"scaffolding\"program that structures multiple calls to language models (LMs) to generate better outputs. A scaffolding program is written in a programming language such as Python. In this work, we use a language-model-infused scaffolding program to improve itself. We start with a seed\"improver\"that improves an input program according to a given utility function by querying an LM several times and returning the best solution. We then run this seed improver to improve itself. Across a small set of downstream tasks, the resulting improved improver generates programs with significantly better performance than its seed improver. A variety of self-improvement strategies are proposed by the language model, including beam search, genetic algorithms, and simulated annealing. Since the language models themselves are not altered, this is not full recursive self-improvement. Nonetheless, it demonstrates that a modern language model, GPT-4 in our experiments, is capable of writing code that can call itself to improve itself. We consider concerns around the development of self-improving technologies and evaluate the frequency with which the generated code bypasses a sandbox."
                    },
                    {
                        "title": "ContextRef: Evaluating Referenceless Metrics For Image Description Generation",
                        "abstract": "Referenceless metrics (e.g., CLIPScore) use pretrained vision--language models to assess image descriptions directly without costly ground-truth reference texts. Such methods can facilitate rapid progress, but only if they truly align with human preference judgments. In this paper, we introduce ContextRef, a benchmark for assessing referenceless metrics for such alignment. ContextRef has two components: human ratings along a variety of established quality dimensions, and ten diverse robustness checks designed to uncover fundamental weaknesses. A crucial aspect of ContextRef is that images and descriptions are presented in context, reflecting prior work showing that context is important for description quality. Using ContextRef, we assess a variety of pretrained models, scoring functions, and techniques for incorporating context. None of the methods is successful with ContextRef, but we show that careful fine-tuning yields substantial improvements. ContextRef remains a challenging benchmark though, in large part due to the challenge of context dependence."
                    },
                    {
                        "title": "SkyGPT: Probabilistic Short-term Solar Forecasting Using Synthetic Sky Videos from Physics-constrained VideoGPT",
                        "abstract": "In recent years, deep learning-based solar forecasting using all-sky images has emerged as a promising approach for alleviating uncertainty in PV power generation. However, the stochastic nature of cloud movement remains a major challenge for accurate and reliable solar forecasting. With the recent advances in generative artificial intelligence, the synthesis of visually plausible yet diversified sky videos has potential for aiding in forecasts. In this study, we introduce \\emph{SkyGPT}, a physics-informed stochastic video prediction model that is able to generate multiple possible future images of the sky with diverse cloud motion patterns, by using past sky image sequences as input. Extensive experiments and comparison with benchmark video prediction models demonstrate the effectiveness of the proposed model in capturing cloud dynamics and generating future sky images with high realism and diversity. Furthermore, we feed the generated future sky images from the video prediction models for 15-minute-ahead probabilistic solar forecasting for a 30-kW roof-top PV system, and compare it with an end-to-end deep learning baseline model SUNSET and a smart persistence model. Better PV output prediction reliability and sharpness is observed by using the predicted sky images generated with SkyGPT compared with other benchmark models, achieving a continuous ranked probability score (CRPS) of 2.81 (13\\% better than SUNSET and 23\\% better than smart persistence) and a Winkler score of 26.70 for the test set. Although an arbitrary number of futures can be generated from a historical sky image sequence, the results suggest that 10 future scenarios is a good choice that balances probabilistic solar forecasting performance and computational cost."
                    },
                    {
                        "title": "Lexinvariant Language Models",
                        "abstract": "Token embeddings, a mapping from discrete lexical symbols to continuous vectors, are at the heart of any language model (LM). However, lexical symbol meanings can also be determined and even redefined by their structural role in a long context. In this paper, we ask: is it possible for a language model to be performant without \\emph{any} fixed token embeddings? Such a language model would have to rely entirely on the co-occurence and repetition of tokens in the context rather than the \\textit{a priori} identity of any token. To answer this, we study \\textit{lexinvariant}language models that are invariant to lexical symbols and therefore do not need fixed token embeddings in practice. First, we prove that we can construct a lexinvariant LM to converge to the true language model at a uniform rate that is polynomial in terms of the context length, with a constant factor that is sublinear in the vocabulary size. Second, to build a lexinvariant LM, we simply encode tokens using random Gaussian vectors, such that each token maps to the same representation within each sequence but different representations across sequences. Empirically, we demonstrate that it can indeed attain perplexity comparable to that of a standard language model, given a sufficiently long context. We further explore two properties of the lexinvariant language models: First, given text generated from a substitution cipher of English, it implicitly implements Bayesian in-context deciphering and infers the mapping to the underlying real tokens with high accuracy. Second, it has on average 4X better accuracy over synthetic in-context reasoning tasks. Finally, we discuss regularizing standard language models towards lexinvariance and potential practical applications."
                    },
                    {
                        "title": "Certified Reasoning with Language Models",
                        "abstract": "Language models often achieve higher accuracy when reasoning step-by-step in complex tasks. However, their reasoning can be unsound, inconsistent, or rely on undesirable prior assumptions. To tackle these issues, we introduce a class of tools for language models called guides that use state and incremental constraints to guide generation. A guide can be invoked by the model to constrain its own generation to a set of valid statements given by the tool. In turn, the model's choices can change the guide's state. We show how a general system for logical reasoning can be used as a guide, which we call LogicGuide. Given a reasoning problem in natural language, a model can formalize its assumptions for LogicGuide and then guarantee that its reasoning steps are sound. In experiments with the PrOntoQA and ProofWriter reasoning datasets, LogicGuide significantly improves the performance of GPT-3, GPT-3.5 Turbo and LLaMA (accuracy gains up to 35%). LogicGuide also drastically reduces content effects: the interference of prior and current assumptions that both humans and language models have been shown to suffer from. Finally, we explore bootstrapping LLaMA 13B from its own reasoning and find that LogicGuide is critical: by training only on certified self-generated reasoning, LLaMA can self-improve, avoiding learning from its own hallucinations."
                    },
                    {
                        "title": "Holistic Evaluation of Language Models",
                        "abstract": "Language models (LMs) like GPT\u20103, PaLM, and ChatGPT are the foundation for almost all major language technologies, but their capabilities, limitations, and risks are not well understood. We present Holistic Evaluation of Language Models (HELM) to improve the transparency of LMs. LMs can serve many purposes and their behavior should satisfy many desiderata. To navigate the vast space of potential scenarios and metrics, we taxonomize the space and select representative subsets. We evaluate models on 16 core scenarios and 7 metrics, exposing important trade\u2010offs. We supplement our core evaluation with seven targeted evaluations to deeply analyze specific aspects (including world knowledge, reasoning, regurgitation of copyrighted content, and generation of disinformation). We benchmark 30 LMs, from OpenAI, Microsoft, Google, Meta, Cohere, AI21 Labs, and others. Prior to HELM, models were evaluated on just 17.9% of the core HELM scenarios, with some prominent models not sharing a single scenario in common. We improve this to 96.0%: all 30 models are now benchmarked under the same standardized conditions. Our evaluation surfaces 25 top\u2010level findings. For full transparency, we release all raw model prompts and completions publicly. HELM is a living benchmark for the community, continuously updated with new scenarios, metrics, and models https://crfm.stanford.edu/helm/latest/."
                    },
                    {
                        "title": "Just One Byte (per gradient): A Note on Low-Bandwidth Decentralized Language Model Finetuning Using Shared Randomness",
                        "abstract": "Language model training in distributed settings is limited by the communication cost of gradient exchanges. In this short note, we extend recent work from Malladi et al. (2023), using shared randomness to perform distributed fine-tuning with low bandwidth. The method is a natural decentralized extension of memory-efficient Simultaneous Perturbation Stochastic Approximation (SPSA). Each iteration, each machine seeds a Random Number Generator (RNG) to perform local reproducible perturbations on model weights and calculate and exchange scalar projected gradients, which are then used to update each model. By using a (machine, sample) identifier as the random seed, each model can regenerate one another's perturbations. As machines only exchange single-byte projected gradients, this is highly communication efficient. There are also potential privacy benefits, as projected gradients may be calculated on different training data, and models never access the other's data. Our approach not only drastically reduces communication bandwidth requirements but also accommodates dynamic addition or removal of machines during the training process and retains the memory-efficient and inference-only advantages of recent work. We perform proof-of-concept experiments to demonstrate the potential usefulness of this method, building off of rich literature on distributed optimization and memory-efficient training."
                    },
                    {
                        "title": "Certified Deductive Reasoning with Language Models",
                        "abstract": "Language models often achieve higher accuracy when reasoning step-by-step in complex tasks. However, even when arriving at a correct final answer, their rationales are often logically unsound or inconsistent. This is a major issue when reliable reasoning traces are needed, such when fine-tuning on model-generated reasoning for self-improvement. To tackle these issues, we introduce a class of tools for language models called \\emph{guides}, that use state and incremental constraints to guide generation. A guide can be invoked by the model to constrain its own generation to a set of valid statements given by the tool. In turn, the model's choices can change the guide's state. We show how a general system for logical reasoning can be used as a guide, which we call \\textsc{LogicGuide}. Given a reasoning problem in natural language, a model can formalize its assumptions for \\textsc{LogicGuide} and guarantee that its step-by-step reasoning is sound. In experiments on PrOntoQA, ProofWriter and Syllogism Validity datasets, \\textsc{LogicGuide} significantly improves the performance of GPT-3, GPT-3.5 Turbo and LLaMA (accuracy gains up to 35\\%), while drastically reducing \\emph{content effects} -- the interference between unwanted prior assumptions and reasoning, which humans and language models suffer from. We then explore bootstrapping GPT-3.5 Turbo and LLaMA using their own reasoning traces. We find that LogicGuide is critical: by training only on certified self-generated reasoning, models can self-improve, avoiding learning from their own hallucinations. Moreover, bootstrapped models enjoy significant boosts on ReClor, a challenging real-world reasoning dataset, even when not relying on formalization at inference time."
                    },
                    {
                        "title": "Parsel: A Unified Natural Language Framework for Algorithmic Reasoning",
                        "abstract": "Despite recent success in large language model (LLM) reasoning, LLMs still struggle with hierarchical multi-step reasoning like generating complex programs. In these cases, humans often start with a high-level algorithmic design and implement each part gradually. We introduce Parsel 2 , a framework enabling automatic implementation and validation of complex algorithms with code LLMs, based on hierarchical function descriptions in natural language. Parsel can be used across domains requiring hierarchical reasoning, e.g. code synthesis, theorem proving, and robotic planning. We demonstrate Parsel\u2019s capabilities by using it to generate complex programs that cannot currently be automatically implemented from one description and backtranslating Python programs in the APPS dataset. Beyond modeling capabilities, Parsel allows problem-solving with high-level algorithmic designs, bene\ufb01ting both students and professional programmers."
                    },
                    {
                        "title": "STaR: Bootstrapping Reasoning With Reasoning",
                        "abstract": "Generating step-by-step\"chain-of-thought\"rationales improves language model performance on complex reasoning tasks like mathematics or commonsense question-answering. However, inducing language model rationale generation currently requires either constructing massive rationale datasets or sacrificing accuracy by using only few-shot inference. We propose a technique to iteratively leverage a small number of rationale examples and a large dataset without rationales, to bootstrap the ability to perform successively more complex reasoning. This technique, the\"Self-Taught Reasoner\"(STaR), relies on a simple loop: generate rationales to answer many questions, prompted with a few rationale examples; if the generated answers are wrong, try again to generate a rationale given the correct answer; fine-tune on all the rationales that ultimately yielded correct answers; repeat. We show that STaR significantly improves performance on multiple datasets compared to a model fine-tuned to directly predict final answers, and performs comparably to fine-tuning a 30$\\times$ larger state-of-the-art language model on CommensenseQA. Thus, STaR lets a model improve itself by learning from its own generated reasoning."
                    },
                    {
                        "title": "Holistic Evaluation of Language Models",
                        "abstract": "Language models (LMs) are becoming the foundation for almost all major language technologies, but their capabilities, limitations, and risks are not well understood. We present Holistic Evaluation of Language Models (HELM) to improve the transparency of language models. First, we taxonomize the vast space of potential scenarios (i.e. use cases) and metrics (i.e. desiderata) that are of interest for LMs. Then we select a broad subset based on coverage and feasibility, noting what's missing or underrepresented (e.g. question answering for neglected English dialects, metrics for trustworthiness). Second, we adopt a multi-metric approach: We measure 7 metrics (accuracy, calibration, robustness, fairness, bias, toxicity, and efficiency) for each of 16 core scenarios when possible (87.5% of the time). This ensures metrics beyond accuracy don't fall to the wayside, and that trade-offs are clearly exposed. We also perform 7 targeted evaluations, based on 26 targeted scenarios, to analyze specific aspects (e.g. reasoning, disinformation). Third, we conduct a large-scale evaluation of 30 prominent language models (spanning open, limited-access, and closed models) on all 42 scenarios, 21 of which were not previously used in mainstream LM evaluation. Prior to HELM, models on average were evaluated on just 17.9% of the core HELM scenarios, with some prominent models not sharing a single scenario in common. We improve this to 96.0%: now all 30 models have been densely benchmarked on the same core scenarios and metrics under standardized conditions. Our evaluation surfaces 25 top-level findings. For full transparency, we release all raw model prompts and completions publicly for further analysis, as well as a general modular toolkit. We intend for HELM to be a living benchmark for the community, continuously updated with new scenarios, metrics, and models."
                    },
                    {
                        "title": "Context Matters for Image Descriptions for Accessibility: Challenges for Referenceless Evaluation Metrics",
                        "abstract": "Few images on the Web receive alt-text descriptions that would make them accessible to blind and low vision (BLV) users. Image-based NLG systems have progressed to the point where they can begin to address this persistent societal problem, but these systems will not be fully successful unless we evaluate them on metrics that guide their development correctly. Here, we argue against current referenceless metrics \u2013 those that don\u2019t rely on human-generated ground-truth descriptions \u2013 on the grounds that they do not align with the needs of BLV users. The fundamental shortcoming of these metrics is that they do not take context into account, whereas contextual information is highly valued by BLV users. To substantiate these claims, we present a study with BLV participants who rated descriptions along a variety of dimensions. An in-depth analysis reveals that the lack of context-awareness makes current referenceless metrics inadequate for advancing image accessibility. As a proof-of-concept, we provide a contextual version of the referenceless metric CLIPScore which begins to address the disconnect to the BLV data."
                    }
                ]
            },
            "5847d3fb-6622-4a0f-a1c8-fa4ac5f45a9f": {
                "pk": "5847d3fb-6622-4a0f-a1c8-fa4ac5f45a9f",
                "name": "Gabriel Poesia",
                "collaborators": [
                    "Noah D. Goodman",
                    "E. Zelikman",
                    "Nick Haber",
                    "Zhening Li",
                    "Armando Solar-Lezama",
                    "Kanishk Gandhi",
                    "Qian Huang",
                    "Shubhra Mishra",
                    "Belinda Mo",
                    "M. Doumbouya"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Reinforcement Learning",
                    "Large Language Models",
                    "Mathematical Reasoning"
                ],
                "publications": [
                    {
                        "title": "MathCAMPS: Fine-grained Synthesis of Mathematical Problems From Human Curricula",
                        "abstract": "Mathematical problem solving is an important skill for Large Language Models (LLMs), both as an important capability and a proxy for a range of reasoning abilities. Existing benchmarks probe a diverse set of skills, but they yield aggregate accuracy metrics, obscuring specific abilities or weaknesses. Furthermore, they are difficult to extend with new problems, risking data contamination over time. To address these challenges, we propose MathCAMPS: a method to synthesize high-quality mathematical problems at scale, grounded on 44 fine-grained\"standards\"from the Mathematics Common Core (CC) Standard for K-8 grades. We encode each standard in a formal grammar, allowing us to sample diverse symbolic problems and their answers. We then use LLMs to realize the symbolic problems into word problems. We propose a cycle-consistency method for validating problem faithfulness. Finally, we derive follow-up questions from symbolic structures and convert them into follow-up word problems - a novel task of mathematical dialogue that probes for robustness in understanding. Experiments on 23 LLMs show surprising failures even in the strongest models (in particular when asked simple follow-up questions). Moreover, we evaluate training checkpoints of Pythia 12B on MathCAMPS, allowing us to analyze when particular mathematical skills develop during its training. Our framework enables the community to reproduce and extend our pipeline for a fraction of the typical cost of building new high-quality datasets."
                    },
                    {
                        "title": "h4rm3l: A Dynamic Benchmark of Composable Jailbreak Attacks for LLM Safety Assessment",
                        "abstract": "The safety of Large Language Models (LLMs) remains a critical concern due to a lack of adequate benchmarks for systematically evaluating their ability to resist generating harmful content. Previous efforts towards automated red teaming involve static or templated sets of illicit requests and adversarial prompts which have limited utility given jailbreak attacks' evolving and composable nature. We propose a novel dynamic benchmark of composable jailbreak attacks to move beyond static datasets and taxonomies of attacks and harms. Our approach consists of three components collectively called h4rm3l: (1) a domain-specific language that formally expresses jailbreak attacks as compositions of parameterized prompt transformation primitives, (2) bandit-based few-shot program synthesis algorithms that generate novel attacks optimized to penetrate the safety filters of a target black box LLM, and (3) open-source automated red-teaming software employing the previous two components. We use h4rm3l to generate a dataset of 2656 successful novel jailbreak attacks targeting 6 state-of-the-art (SOTA) open-source and proprietary LLMs. Several of our synthesized attacks are more effective than previously reported ones, with Attack Success Rates exceeding 90% on SOTA closed language models such as claude-3-haiku and GPT4-o. By generating datasets of jailbreak attacks in a unified formal representation, h4rm3l enables reproducible benchmarking and automated red-teaming, contributes to understanding LLM safety limitations, and supports the development of robust defenses in an increasingly LLM-integrated world. Warning: This paper and related research artifacts contain offensive and potentially disturbing prompts and model-generated content."
                    },
                    {
                        "title": "Learning Formal Mathematics From Intrinsic Motivation",
                        "abstract": "How did humanity coax mathematics from the aether? We explore the Platonic view that mathematics can be discovered from its axioms - a game of conjecture and proof. We describe Minimo (Mathematics from Intrinsic Motivation): an agent that jointly learns to pose challenging problems for itself (conjecturing) and solve them (theorem proving). Given a mathematical domain axiomatized in dependent type theory, we first combine methods for constrained decoding and type-directed synthesis to sample valid conjectures from a language model. Our method guarantees well-formed conjectures by construction, even as we start with a randomly initialized model. We use the same model to represent a policy and value function for guiding proof search. Our agent targets generating hard but provable conjectures - a moving target, since its own theorem proving ability also improves as it trains. We propose novel methods for hindsight relabeling on proof search trees to significantly improve the agent's sample efficiency in both tasks. Experiments on 3 axiomatic domains (propositional logic, arithmetic and group theory) demonstrate that our agent can bootstrap from only the axioms, self-improving in generating true and challenging conjectures and in finding proofs."
                    },
                    {
                        "title": "When Do Skills Help Reinforcement Learning? A Theoretical Analysis of Temporal Abstractions",
                        "abstract": "Skills are temporal abstractions that are intended to improve reinforcement learning (RL) performance through hierarchical RL. Despite our intuition about the properties of an environment that make skills useful, a precise characterization has been absent. We provide the first such characterization, focusing on the utility of deterministic skills in deterministic sparse-reward environments with finite action spaces. We show theoretically and empirically that RL performance gain from skills is worse in environments where solutions to states are less compressible. Additional theoretical results suggest that skills benefit exploration more than they benefit learning from existing experience, and that using unexpressive skills such as macroactions may worsen RL performance. We hope our findings can guide research on automatic skill discovery and help RL practitioners better decide when and how to use skills."
                    },
                    {
                        "title": "Generalizing BERT through Multi-Task Learning",
                        "abstract": "Bidirectional Encoder Representations from Transformers or BERT is a model that uses word representations to perform tasks such as sentiment analysis, paraphrase detection, and semantic textual similarity. In this paper, a basic BERT model is implemented as a baseline model. Then, in an attempt to improve BERT\u2019s ability to generalize well across the three downstream tasks listed above, this paper implements a multi-task method of training BERT"
                    },
                    {
                        "title": "Pre-training BERT: Swapped Subject Phrase Detection",
                        "abstract": "BERT is a transformer-based model that generates contextual word representations (Devlin et al., 2019). These representations can then be fine-tuned and used for downstream tasks such as sentiment analysis, paraphrase detection, and semantic textual similarity detection. In this project, we examine the success of a novel pre-training task aimed at teaching the model topic-level features, where we randomly (with 50% probability) swap subject phrases in two unrelated sentences and have the BERT model predict whether or not they were swapped. Through our experiments, we find that the swapped sentence phrase detection task is more beneficial for downstream task performance than additional pre-training on the masked language model objective using in-task data, and that pre-training on a combination of the two tasks leads to even higher performance and a more generalized model."
                    },
                    {
                        "title": "Solving Math Word Problems by Combining Language Models With Symbolic Solvers",
                        "abstract": "Automatically generating high-quality step-by-step solutions to math word problems has many applications in education. Recently, combining large language models (LLMs) with external tools to perform complex reasoning and calculation has emerged as a promising direction for solving math word problems, but prior approaches such as Program-Aided Language model (PAL) are biased towards simple procedural problems and less effective for problems that require declarative reasoning. We propose an approach that combines an LLM that can incrementally formalize word problems as a set of variables and equations with an external symbolic solver that can solve the equations. Our approach achieves comparable accuracy to the original PAL on the GSM8K benchmark of math word problems and outperforms PAL by an absolute 20% on ALGEBRA, a new dataset of more challenging word problems extracted from Algebra textbooks. Our work highlights the benefits of using declarative and incremental representations when interfacing with an external tool for solving complex math word problems. Our data and prompts are publicly available at https://github.com/joyheyueya/declarative-math-word-problem."
                    },
                    {
                        "title": "Enhanced generalizable minBERT model for multiple downstream tasks",
                        "abstract": "This project aims to 1) build a BERT model that could work for Sentiment Analysis and extend the model to multiple downstream tasks 2) explore and analyze performance impact from different improvement approaches. For part 1, we\u2019ve built a BERT model that could work for three downstream tasks and got 0.348 average score as baseline. For part 2, we explored a few different approaches and finally got 0.663 average score for test Set"
                    },
                    {
                        "title": "Multitask BERT: Exploration and Extension",
                        "abstract": "This project will be exploring the BERT model and possible extensions. The BERT model has been an important innovation in natural language processing, and performs extremely well on single-task evaluation. The goal of this project is to extend its performance to multitask scenarios, specifically on sentiment analysis, paraphrase detection, and semantic textual similarity. This project explored four ways of extending BERT: including additional dropout layers with layer normalization to reduce overfitting, including RELU layers for performance improvement , including cosine similarity when completing tasks which focus on similarity, and multitask finetuning by training all three tasks at the same time and combining their loss functions. These extensions concretely improved performance of the model over baselines. This project also utilized parameter search to locate ideal hyperparameters for the improved model, which similarly boosted performance on the three tasks of relevance. The main findings were that including additional dropout layers significantly reduced overfitting, cosine similarity was effective in increasing comparison-based scores, and multitask finetuning and RELU layers had smaller, but definite improvements. Increasing batch sizes notably improved efficiency with slight impact on scores, increasing dropout rates decreased overfitting but hurt performance, and increasing the learning rate had harmful effects on the optimal model\u2019s performance."
                    },
                    {
                        "title": "Can Visual Scratchpads With Diagrammatic Abstractions Augment LLM Reasoning?",
                        "abstract": "When humans reason about complex text-based questions, we leverage diagrammatic abstractions drawn on a visual scratchpad. In this paper, we introduce and explore the capabilities of Visual-Scratchpad, a method that augments a large language foundation model (LLM) with diagrammatic execution and readout. We enable the LLM to generate drawing commands and to readout abstractions from the resulting picture. The visual readout operation uses a visual foundation model , optionally finetuned with expert iteration. Here, we show that although Visual-Scratchpad outperforms an inference-only LLM, it surprisingly yields worse performance compared to a single finetuned LLM. Through experiments, we propose that this gap is due to the failure mode of vision foundation models in understanding abstractions in diagrams"
                    },
                    {
                        "title": "Certified Reasoning with Language Models",
                        "abstract": "Language models often achieve higher accuracy when reasoning step-by-step in complex tasks. However, their reasoning can be unsound, inconsistent, or rely on undesirable prior assumptions. To tackle these issues, we introduce a class of tools for language models called guides that use state and incremental constraints to guide generation. A guide can be invoked by the model to constrain its own generation to a set of valid statements given by the tool. In turn, the model's choices can change the guide's state. We show how a general system for logical reasoning can be used as a guide, which we call LogicGuide. Given a reasoning problem in natural language, a model can formalize its assumptions for LogicGuide and then guarantee that its reasoning steps are sound. In experiments with the PrOntoQA and ProofWriter reasoning datasets, LogicGuide significantly improves the performance of GPT-3, GPT-3.5 Turbo and LLaMA (accuracy gains up to 35%). LogicGuide also drastically reduces content effects: the interference of prior and current assumptions that both humans and language models have been shown to suffer from. Finally, we explore bootstrapping LLaMA 13B from its own reasoning and find that LogicGuide is critical: by training only on certified self-generated reasoning, LLaMA can self-improve, avoiding learning from its own hallucinations."
                    },
                    {
                        "title": "Certified Deductive Reasoning with Language Models",
                        "abstract": "Language models often achieve higher accuracy when reasoning step-by-step in complex tasks. However, even when arriving at a correct final answer, their rationales are often logically unsound or inconsistent. This is a major issue when reliable reasoning traces are needed, such when fine-tuning on model-generated reasoning for self-improvement. To tackle these issues, we introduce a class of tools for language models called \\emph{guides}, that use state and incremental constraints to guide generation. A guide can be invoked by the model to constrain its own generation to a set of valid statements given by the tool. In turn, the model's choices can change the guide's state. We show how a general system for logical reasoning can be used as a guide, which we call \\textsc{LogicGuide}. Given a reasoning problem in natural language, a model can formalize its assumptions for \\textsc{LogicGuide} and guarantee that its step-by-step reasoning is sound. In experiments on PrOntoQA, ProofWriter and Syllogism Validity datasets, \\textsc{LogicGuide} significantly improves the performance of GPT-3, GPT-3.5 Turbo and LLaMA (accuracy gains up to 35\\%), while drastically reducing \\emph{content effects} -- the interference between unwanted prior assumptions and reasoning, which humans and language models suffer from. We then explore bootstrapping GPT-3.5 Turbo and LLaMA using their own reasoning traces. We find that LogicGuide is critical: by training only on certified self-generated reasoning, models can self-improve, avoiding learning from their own hallucinations. Moreover, bootstrapped models enjoy significant boosts on ReClor, a challenging real-world reasoning dataset, even when not relying on formalization at inference time."
                    },
                    {
                        "title": "Parsel: A Unified Natural Language Framework for Algorithmic Reasoning",
                        "abstract": "Despite recent success in large language model (LLM) reasoning, LLMs still struggle with hierarchical multi-step reasoning like generating complex programs. In these cases, humans often start with a high-level algorithmic design and implement each part gradually. We introduce Parsel 2 , a framework enabling automatic implementation and validation of complex algorithms with code LLMs, based on hierarchical function descriptions in natural language. Parsel can be used across domains requiring hierarchical reasoning, e.g. code synthesis, theorem proving, and robotic planning. We demonstrate Parsel\u2019s capabilities by using it to generate complex programs that cannot currently be automatically implemented from one description and backtranslating Python programs in the APPS dataset. Beyond modeling capabilities, Parsel allows problem-solving with high-level algorithmic designs, bene\ufb01ting both students and professional programmers."
                    },
                    {
                        "title": "Emoji Prediction with Transformer Models",
                        "abstract": "Text messaging has become a dominant form of communication, and emojis play a crucial role in conveying emotions for many people, which makes accurate emoji prediction for messages a useful task to explore. This project aims to predict emojis for texts in supervised setting and generalize to new emojis in zero-shot setting. We formulate supervised emoji prediction as a multi-class classification problem and fine-tune pretrained BERT and GPT-2 for this task. Further, we adapt instruction tuning that is originally proposed in the FLAN paper to GPT-2 to generalize to zero-shot setting. We find that the fine-tuned models achieve decent results for the supervised task, but instruction tuning fails to help GPT-2 generate accurate predictions for the zero-shot task due to the small scale of the dataset and model compared to those used for FLAN."
                    },
                    {
                        "title": "LEMMA: Bootstrapping High-Level Mathematical Reasoning with Learned Symbolic Abstractions",
                        "abstract": "Humans tame the complexity of mathematical reasoning by developing hierarchies of abstractions. With proper abstractions, solutions to hard problems can be expressed concisely, thus making them more likely to be found. In this paper, we propose Learning Mathematical Abstractions (LEMMA): an algorithm that implements this idea for reinforcement learning agents in mathematical domains. LEMMA augments Expert Iteration with an abstraction step, where solutions found so far are revisited and rewritten in terms of new higher-level actions, which then become available to solve new problems. We evaluate LEMMA on two mathematical reasoning tasks--equation solving and fraction simplification--in a step-by-step fashion. In these two domains, LEMMA improves the ability of an existing agent, both solving more problems and generalizing more effectively to harder problems than those seen during training."
                    },
                    {
                        "title": "Left to the Reader: Abstracting Solutions in Mathematical Reasoning",
                        "abstract": "Formal mathematical reasoning is unique in its precision: any valid conclusion can be justified by a sequence of base axioms. But human-written proofs or solutions rarely operate at that level. Instead, obvious steps are skipped to provide a simple, lucid argument. This is especially important in an educational setting, where too many details in an example solution, or too few, can confuse a student. What are the key steps for humans in a given formal solution? We investigate several computational hypotheses in the context of equation solving. Specifi- cally, we take a reinforcement learning agent that solves equations using low-level axioms, and propose a series of methods for abstracting its solutions by selecting key steps. We consider methods based on the semantic distance between subse- quent steps, based on the steps with the highest uncertainty for the agent, and based on transitions between latent \u201chigh-level skills\u201d learned from a large number of agent-produced solutions. In a human evaluation we find that skill-base simplifi- cations were judged most useful. These results suggest new directions for understanding human mathematical reasoning."
                    },
                    {
                        "title": "Peano: learning formal mathematical reasoning",
                        "abstract": "General mathematical reasoning is computationally undecidable, but humans routinely solve new problems. Moreover, discoveries developed over centuries are taught to subsequent generations quickly. What structure enables this, and how might that inform automated mathematical reasoning? We posit that central to both puzzles is the structure of procedural abstractions underlying mathematics. We explore this idea in a case study on five sections of beginning algebra on the Khan Academy platform. To define a computational foundation, we introduce Peano, a theorem-proving environment where the set of valid actions at any point is finite. We use Peano to formalize introductory algebra problems and axioms, obtaining well-defined search problems. We observe existing reinforcement learning methods for symbolic reasoning to be insufficient to solve harder problems. Adding the ability to induce reusable abstractions (\u2018tactics\u2019) from its own solutions allows an agent to make steady progress, solving all problems. Furthermore, these abstractions induce an order to the problems, seen at random during training. The recovered order has significant agreement with the expert-designed Khan Academy curriculum, and second-generation agents trained on the recovered curriculum learn significantly faster. These results illustrate the synergistic role of abstractions and curricula in the cultural transmission of mathematics. This article is part of a discussion meeting issue \u2018Cognitive artificial intelligence\u2019."
                    },
                    {
                        "title": "Synchromesh: Reliable code generation from pre-trained language models",
                        "abstract": "Large pre-trained language models have been used to generate code,providing a flexible interface for synthesizing programs from natural language specifications. However, they often violate syntactic and semantic rules of their output language, limiting their practical usability. In this paper, we propose Synchromesh: a framework for substantially improving the reliability of pre-trained models for code generation. Synchromesh comprises two components. First, it retrieves few-shot examples from a training bank using Target Similarity Tuning (TST), a novel method for semantic example selection. TST learns to recognize utterances that describe similar target programs despite differences in surface natural language features. Then, Synchromesh feeds the examples to a pre-trained language model and samples programs using Constrained Semantic Decoding (CSD): a general framework for constraining the output to a set of valid programs in the target language. CSD leverages constraints on partial outputs to sample complete correct programs, and needs neither re-training nor fine-tuning of the language model. We evaluate our methods by synthesizing code from natural language descriptions using GPT-3 and Codex in three real-world languages: SQL queries, Vega-Lite visualizations and SMCalFlow programs. These domains showcase rich constraints that CSD is able to enforce, including syntax, scope, typing rules, and contextual logic. We observe substantial complementary gains from CSD and TST in prediction accuracy and in effectively preventing run-time errors."
                    },
                    {
                        "title": "Parsel\ud83e\udd86: Algorithmic Reasoning with Language Models by Composing Decompositions",
                        "abstract": "Despite recent success in large language model (LLM) reasoning, LLMs struggle with hierarchical multi-step reasoning tasks like generating complex programs. For these tasks, humans often start with a high-level algorithmic design and implement each part gradually. We introduce Parsel, a framework enabling automatic implementation and validation of complex algorithms with code LLMs. With Parsel, we automatically decompose algorithmic tasks into hierarchical natural language function descriptions and then search over combinations of possible function implementations using tests. We show that Parsel can be used across domains requiring hierarchical reasoning, including program synthesis and robotic planning. We find that, using Parsel, LLMs solve more competition-level problems in the APPS dataset, resulting in pass rates over 75\\% higher than prior results from directly sampling AlphaCode and Codex, while often using a smaller sample budget. Moreover, with automatically generated tests, we find that Parsel can improve the state-of-the-art pass@1 performance on HumanEval from 67\\% to 85\\%. We also find that LLM-generated robotic plans using Parsel are more than twice as likely to be considered accurate than directly generated plans. Lastly, we explore how Parsel addresses LLM limitations and discuss how Parsel may be useful for human programmers. We release our code at https://github.com/ezelikman/parsel"
                    }
                ]
            },
            "26cae974-de2b-47eb-a4a4-072d56cee7c0": {
                "pk": "26cae974-de2b-47eb-a4a4-072d56cee7c0",
                "name": "Yewen Pu",
                "collaborators": [
                    "Rui Zhang",
                    "Hanna Hajishirzi",
                    "Yutong Wu",
                    "Di Huang",
                    "Wenxuan Shi",
                    "Wei Wang",
                    "Lingzhe Gao",
                    "Shihao Liu",
                    "Ziyuan Nan",
                    "Kaizhao Yuan"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Code Generation",
                    "Differentiable Physics"
                ],
                "publications": [
                    {
                        "title": "InverseCoder: Unleashing the Power of Instruction-Tuned Code LLMs with Inverse-Instruct",
                        "abstract": "Recent advancements in open-source code large language models (LLMs) have demonstrated remarkable coding abilities by fine-tuning on the data generated from powerful closed-source LLMs such as GPT-3.5 and GPT-4 for instruction tuning. This paper explores how to further improve an instruction-tuned code LLM by generating data from itself rather than querying closed-source LLMs. Our key observation is the misalignment between the translation of formal and informal languages: translating formal language (i.e., code) to informal language (i.e., natural language) is more straightforward than the reverse. Based on this observation, we propose INVERSE-INSTRUCT, which summarizes instructions from code snippets instead of the reverse. Specifically, given an instruction tuning corpus for code and the resulting instruction-tuned code LLM, we ask the code LLM to generate additional high-quality instructions for the original corpus through code summarization and self-evaluation. Then, we fine-tune the base LLM on the combination of the original corpus and the self-generated one, which yields a stronger instruction-tuned LLM. We present a series of code LLMs named InverseCoder, which surpasses the performance of the original code LLMs on a wide range of benchmarks, including Python text-to-code generation, multilingual coding, and data-science code generation."
                    },
                    {
                        "title": "DiffVL: Scaling Up Soft Body Manipulation using Vision-Language Driven Differentiable Physics",
                        "abstract": "Combining gradient-based trajectory optimization with differentiable physics simulation is an efficient technique for solving soft-body manipulation problems. Using a well-crafted optimization objective, the solver can quickly converge onto a valid trajectory. However, writing the appropriate objective functions requires expert knowledge, making it difficult to collect a large set of naturalistic problems from non-expert users. We introduce DiffVL, a method that enables non-expert users to communicate soft-body manipulation tasks -- a combination of vision and natural language, given in multiple stages -- that can be readily leveraged by a differential physics solver. We have developed GUI tools that enable non-expert users to specify 100 tasks inspired by real-life soft-body manipulations from online videos, which we'll make public. We leverage large language models to translate task descriptions into machine-interpretable optimization objectives. The optimization objectives can help differentiable physics solvers to solve these long-horizon multistage tasks that are challenging for previous baselines."
                    },
                    {
                        "title": "D IV K NOW QA: Assessing the Reasoning Ability of LLMs via Open-Domain Question Answering over Knowledge Base and Text",
                        "abstract": "Large Language Models (LLMs) have exhib-001 ited impressive generation capabilities, but 002 they suffer from hallucinations when solely"
                    }
                ]
            },
            "16cac62a-adfa-42b3-90be-deda11af0e70": {
                "pk": "16cac62a-adfa-42b3-90be-deda11af0e70",
                "name": "Nick Haber",
                "collaborators": [
                    "E. Zelikman",
                    "Noah D. Goodman",
                    "Gabriel Poesia",
                    "Hanna Hajishirzi",
                    "David Broman",
                    "Georges Harik",
                    "Yijia Shao",
                    "Varuna Jayasiri",
                    "Wanjing Anya Ma",
                    "Jasmine E. Tran"
                ],
                "domain": [
                    "Mathematics",
                    "Language Models",
                    "Reasoning",
                    "Educational Assessment"
                ],
                "publications": [
                    {
                        "title": "Learning Formal Mathematics From Intrinsic Motivation",
                        "abstract": "How did humanity coax mathematics from the aether? We explore the Platonic view that mathematics can be discovered from its axioms - a game of conjecture and proof. We describe Minimo (Mathematics from Intrinsic Motivation): an agent that jointly learns to pose challenging problems for itself (conjecturing) and solve them (theorem proving). Given a mathematical domain axiomatized in dependent type theory, we first combine methods for constrained decoding and type-directed synthesis to sample valid conjectures from a language model. Our method guarantees well-formed conjectures by construction, even as we start with a randomly initialized model. We use the same model to represent a policy and value function for guiding proof search. Our agent targets generating hard but provable conjectures - a moving target, since its own theorem proving ability also improves as it trains. We propose novel methods for hindsight relabeling on proof search trees to significantly improve the agent's sample efficiency in both tasks. Experiments on 3 axiomatic domains (propositional logic, arithmetic and group theory) demonstrate that our agent can bootstrap from only the axioms, self-improving in generating true and challenging conjectures and in finding proofs."
                    },
                    {
                        "title": "Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking",
                        "abstract": "When writing and talking, people sometimes pause to think. Although reasoning-focused works have often framed reasoning as a method of answering questions or completing agentic tasks, reasoning is implicit in almost all written text. For example, this applies to the steps not stated between the lines of a proof or to the theory of mind underlying a conversation. In the Self-Taught Reasoner (STaR, Zelikman et al. 2022), useful thinking is learned by inferring rationales from few-shot examples in question-answering and learning from those that lead to a correct answer. This is a highly constrained setting -- ideally, a language model could instead learn to infer unstated rationales in arbitrary text. We present Quiet-STaR, a generalization of STaR in which LMs learn to generate rationales at each token to explain future text, improving their predictions. We address key challenges, including 1) the computational cost of generating continuations, 2) the fact that the LM does not initially know how to generate or use internal thoughts, and 3) the need to predict beyond individual next tokens. To resolve these, we propose a tokenwise parallel sampling algorithm, using learnable tokens indicating a thought's start and end, and an extended teacher-forcing technique. Encouragingly, generated rationales disproportionately help model difficult-to-predict tokens and improve the LM's ability to directly answer difficult questions. In particular, after continued pretraining of an LM on a corpus of internet text with Quiet-STaR, we find zero-shot improvements on GSM8K (5.9%$\\rightarrow$10.9%) and CommonsenseQA (36.3%$\\rightarrow$47.2%) and observe a perplexity improvement of difficult tokens in natural text. Crucially, these improvements require no fine-tuning on these tasks. Quiet-STaR marks a step towards LMs that can learn to reason in a more general and scalable way."
                    },
                    {
                        "title": "Generating and Evaluating Tests for K-12 Students with Language Model Simulations: A Case Study on Sentence Reading Efficiency",
                        "abstract": "Developing an educational test can be expensive and time-consuming, as each item must be written by experts and then evaluated by collecting hundreds of student responses. Moreover, many tests require multiple distinct sets of questions administered throughout the school year to closely monitor students' progress, known as parallel tests. In this study, we focus on tests of silent sentence reading efficiency, used to assess students' reading ability over time. To generate high-quality parallel tests, we propose to fine-tune large language models (LLMs) to simulate how previous students would have responded to unseen items. With these simulated responses, we can estimate each item's difficulty and ambiguity. We first use GPT-4 to generate new test items following a list of expert-developed rules and then apply a fine-tuned LLM to filter the items based on criteria from psychological measurements. We also propose an optimal-transport-inspired technique for generating parallel tests and show the generated tests closely correspond to the original test's difficulty and reliability based on crowdworker responses. Our evaluation of a generated test with 234 students from grades 2 to 8 produces test scores highly correlated (r=0.93) to those of a standard test form written by human experts and evaluated across thousands of K-12 students."
                    },
                    {
                        "title": "ContextRef: Evaluating Referenceless Metrics For Image Description Generation",
                        "abstract": "Referenceless metrics (e.g., CLIPScore) use pretrained vision--language models to assess image descriptions directly without costly ground-truth reference texts. Such methods can facilitate rapid progress, but only if they truly align with human preference judgments. In this paper, we introduce ContextRef, a benchmark for assessing referenceless metrics for such alignment. ContextRef has two components: human ratings along a variety of established quality dimensions, and ten diverse robustness checks designed to uncover fundamental weaknesses. A crucial aspect of ContextRef is that images and descriptions are presented in context, reflecting prior work showing that context is important for description quality. Using ContextRef, we assess a variety of pretrained models, scoring functions, and techniques for incorporating context. None of the methods is successful with ContextRef, but we show that careful fine-tuning yields substantial improvements. ContextRef remains a challenging benchmark though, in large part due to the challenge of context dependence."
                    },
                    {
                        "title": "D IV K NOW QA: Assessing the Reasoning Ability of LLMs via Open-Domain Question Answering over Knowledge Base and Text",
                        "abstract": "Large Language Models (LLMs) have exhib-001 ited impressive generation capabilities, but 002 they suffer from hallucinations when solely"
                    }
                ]
            },
            "38b64e79-7ee3-43aa-92c2-84c65da7b610": {
                "pk": "38b64e79-7ee3-43aa-92c2-84c65da7b610",
                "name": "Noah D. Goodman",
                "collaborators": [
                    "Jesse Mu",
                    "Dilip Arumugam",
                    "Gabriel Poesia",
                    "Kanishk Gandhi",
                    "E. Zelikman",
                    "Zhengxuan Wu",
                    "Atticus Geiger",
                    "Christopher Potts",
                    "Mark K. Ho",
                    "Benjamin Van Roy"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Cognitive Science",
                    "Explainable AI"
                ],
                "publications": [
                    {
                        "title": "From Word Models to World Models: Translating from Natural Language to the Probabilistic Language of Thought",
                        "abstract": "How does language inform our downstream thinking? In particular, how do humans make meaning from language--and how can we leverage a theory of linguistic meaning to build machines that think in more human-like ways? In this paper, we propose rational meaning construction, a computational framework for language-informed thinking that combines neural language models with probabilistic models for rational inference. We frame linguistic meaning as a context-sensitive mapping from natural language into a probabilistic language of thought (PLoT)--a general-purpose symbolic substrate for generative world modeling. Our architecture integrates two computational tools that have not previously come together: we model thinking with probabilistic programs, an expressive representation for commonsense reasoning; and we model meaning construction with large language models (LLMs), which support broad-coverage translation from natural language utterances to code expressions in a probabilistic programming language. We illustrate our framework through examples covering four core domains from cognitive science: probabilistic reasoning, logical and relational reasoning, visual and physical reasoning, and social reasoning. In each, we show that LLMs can generate context-sensitive translations that capture pragmatically-appropriate linguistic meanings, while Bayesian inference with the generated programs supports coherent and robust commonsense reasoning. We extend our framework to integrate cognitively-motivated symbolic modules (physics simulators, graphics engines, and planning algorithms) to provide a unified commonsense thinking interface from language. Finally, we explore how language can drive the construction of world models themselves. We hope this work will provide a roadmap towards cognitive models and AI systems that synthesize the insights of both modern and classical computational perspectives."
                    },
                    {
                        "title": "Characterizing tradeoffs between teaching via language and demonstrations in multi-agent systems",
                        "abstract": "Humans teach others about the world through language and demonstration. When might one of these modalities be more effective than the other? In this work, we study the factors that modulate the effectiveness of language vs. demonstration using multi-agent systems to model human communication. Specifically, we train neural network agents to teach via language or demonstration in a grounded communication task, manipulating 1) the inherent difficulty of the task and 2) the competence of the teacher. We find that teaching by demonstration is more effective in the simplest settings, but language is more effective as task difficulty increases, due to its ability to generalize more effectively to unseen scenarios. Overall, these results provide converging evidence for a tradeoff between language and demonstration as teaching modalities in humans, and make the novel predictions that demonstration may be optimal for easy tasks, while language enables generalization in more challenging settings."
                    },
                    {
                        "title": "Interpretability at Scale: Identifying Causal Mechanisms in Alpaca",
                        "abstract": "Obtaining human-interpretable explanations of large, general-purpose language models is an urgent goal for AI safety. However, it is just as important that our interpretability methods are faithful to the causal dynamics underlying model behavior and able to robustly generalize to unseen inputs. Distributed Alignment Search (DAS) is a powerful gradient descent method grounded in a theory of causal abstraction that has uncovered perfect alignments between interpretable symbolic algorithms and small deep learning models fine-tuned for specific tasks. In the present paper, we scale DAS significantly by replacing the remaining brute-force search steps with learned parameters -- an approach we call Boundless DAS. This enables us to efficiently search for interpretable causal structure in large language models while they follow instructions. We apply Boundless DAS to the Alpaca model (7B parameters), which, off the shelf, solves a simple numerical reasoning problem. With Boundless DAS, we discover that Alpaca does this by implementing a causal model with two interpretable boolean variables. Furthermore, we find that the alignment of neural representations with these variables is robust to changes in inputs and instructions. These findings mark a first step toward faithfully understanding the inner-workings of our ever-growing and most widely deployed language models. Our tool is extensible to larger LLMs and is released publicly at `https://github.com/stanfordnlp/pyvene`."
                    },
                    {
                        "title": "Bayesian Reinforcement Learning With Limited Cognitive Load",
                        "abstract": "Abstract All biological and artificial agents must act given limits on their ability to acquire and process information. As such, a general theory of adaptive behavior should be able to account for the complex interactions between an agent\u2019s learning history, decisions, and capacity constraints. Recent work in computer science has begun to clarify the principles that shape these dynamics by bridging ideas from reinforcement learning, Bayesian decision-making, and rate-distortion theory. This body of work provides an account of capacity-limited Bayesian reinforcement learning, a unifying normative framework for modeling the effect of processing constraints on learning and action selection. Here, we provide an accessible review of recent algorithms and theoretical results in this setting, paying special attention to how these ideas can be applied to studying questions in the cognitive and behavioral sciences."
                    },
                    {
                        "title": "Solving Math Word Problems by Combining Language Models With Symbolic Solvers",
                        "abstract": "Automatically generating high-quality step-by-step solutions to math word problems has many applications in education. Recently, combining large language models (LLMs) with external tools to perform complex reasoning and calculation has emerged as a promising direction for solving math word problems, but prior approaches such as Program-Aided Language model (PAL) are biased towards simple procedural problems and less effective for problems that require declarative reasoning. We propose an approach that combines an LLM that can incrementally formalize word problems as a set of variables and equations with an external symbolic solver that can solve the equations. Our approach achieves comparable accuracy to the original PAL on the GSM8K benchmark of math word problems and outperforms PAL by an absolute 20% on ALGEBRA, a new dataset of more challenging word problems extracted from Algebra textbooks. Our work highlights the benefits of using declarative and incremental representations when interfacing with an external tool for solving complex math word problems. Our data and prompts are publicly available at https://github.com/joyheyueya/declarative-math-word-problem."
                    },
                    {
                        "title": "Why think step-by-step? Reasoning emerges from the locality of experience",
                        "abstract": "Humans have a powerful and mysterious capacity to reason. By working through a series of purely mental steps, we can make inferences we would not be capable of making directly -- despite the fact that we get no additional data from the world. Similarly, when large language models generate a series of intermediate steps (a chain of thought) before answering a question, they often produce better answers than they otherwise would. We investigate why and how chain-of-thought reasoning is useful in language models, testing the hypothesis that reasoning is effective when training data consists of local clusters of variables that influence each other strongly. These training conditions enable the chaining of accurate local inferences in order to estimate relationships between variables that were not seen together in training. We prove that there will exist a\"reasoning gap\", where reasoning through intermediate variables improves inference, for the simple case of an autoregressive density estimator trained on local samples from a chain-structured probabilistic model. We then test our hypothesis empirically in more complex models, training an autoregressive language model on samples from Bayes nets but only including a subset of variables in each sample. We test language models' ability to match conditional probabilities with and without intermediate reasoning steps, finding that intermediate steps are only helpful when the training data is locally structured with respect to dependencies between variables and that the combination of locally-structured observations and reasoning is much more data-efficient than training on all variables. Our results illustrate how the effectiveness of reasoning step by step is rooted in the local statistical structure of the training data."
                    },
                    {
                        "title": "Overinformative Question Answering by Humans and Machines",
                        "abstract": "When faced with a polar question, speakers often provide overinformative answers going beyond a simple\"yes\"or\"no\". But what principles guide the selection of additional information? In this paper, we provide experimental evidence from two studies suggesting that overinformativeness in human answering is driven by considerations of relevance to the questioner's goals which they flexibly adjust given the functional context in which the question is uttered. We take these human results as a strong benchmark for investigating question-answering performance in state-of-the-art neural language models, conducting an extensive evaluation on items from human experiments. We find that most models fail to adjust their answering behavior in a human-like way and tend to include irrelevant information. We show that GPT-3 is highly sensitive to the form of the prompt and only achieves human-like answer patterns when guided by an example and cognitively-motivated explanation."
                    },
                    {
                        "title": "Learning to Compress Prompts with Gist Tokens",
                        "abstract": "Prompting is the primary way to utilize the multitask capabilities of language models (LMs), but prompts occupy valuable space in the input context window, and repeatedly encoding the same prompt is computationally inefficient. Finetuning and distillation methods allow for specialization of LMs without prompting, but require retraining the model for each task. To avoid this trade-off entirely, we present gisting, which trains an LM to compress prompts into smaller sets of\"gist\"tokens which can be cached and reused for compute efficiency. Gist models can be trained with no additional cost over standard instruction finetuning by simply modifying Transformer attention masks to encourage prompt compression. On decoder (LLaMA-7B) and encoder-decoder (FLAN-T5-XXL) LMs, gisting enables up to 26x compression of prompts, resulting in up to 40% FLOPs reductions, 4.2% wall time speedups, and storage savings, all with minimal loss in output quality."
                    },
                    {
                        "title": "Finding Alignments Between Interpretable Causal Variables and Distributed Neural Representations",
                        "abstract": "Causal abstraction is a promising theoretical framework for explainable artificial intelligence that defines when an interpretable high-level causal model is a faithful simplification of a low-level deep learning system. However, existing causal abstraction methods have two major limitations: they require a brute-force search over alignments between the high-level model and the low-level one, and they presuppose that variables in the high-level model will align with disjoint sets of neurons in the low-level one. In this paper, we present distributed alignment search (DAS), which overcomes these limitations. In DAS, we find the alignment between high-level and low-level models using gradient descent rather than conducting a brute-force search, and we allow individual neurons to play multiple distinct roles by analyzing representations in non-standard bases-distributed representations. Our experiments show that DAS can discover internal structure that prior approaches miss. Overall, DAS removes previous obstacles to conducting causal abstraction analyses and allows us to find conceptual structure in trained neural nets."
                    },
                    {
                        "title": "Certified Reasoning with Language Models",
                        "abstract": "Language models often achieve higher accuracy when reasoning step-by-step in complex tasks. However, their reasoning can be unsound, inconsistent, or rely on undesirable prior assumptions. To tackle these issues, we introduce a class of tools for language models called guides that use state and incremental constraints to guide generation. A guide can be invoked by the model to constrain its own generation to a set of valid statements given by the tool. In turn, the model's choices can change the guide's state. We show how a general system for logical reasoning can be used as a guide, which we call LogicGuide. Given a reasoning problem in natural language, a model can formalize its assumptions for LogicGuide and then guarantee that its reasoning steps are sound. In experiments with the PrOntoQA and ProofWriter reasoning datasets, LogicGuide significantly improves the performance of GPT-3, GPT-3.5 Turbo and LLaMA (accuracy gains up to 35%). LogicGuide also drastically reduces content effects: the interference of prior and current assumptions that both humans and language models have been shown to suffer from. Finally, we explore bootstrapping LLaMA 13B from its own reasoning and find that LogicGuide is critical: by training only on certified self-generated reasoning, LLaMA can self-improve, avoiding learning from its own hallucinations."
                    },
                    {
                        "title": "Just One Byte (per gradient): A Note on Low-Bandwidth Decentralized Language Model Finetuning Using Shared Randomness",
                        "abstract": "Language model training in distributed settings is limited by the communication cost of gradient exchanges. In this short note, we extend recent work from Malladi et al. (2023), using shared randomness to perform distributed fine-tuning with low bandwidth. The method is a natural decentralized extension of memory-efficient Simultaneous Perturbation Stochastic Approximation (SPSA). Each iteration, each machine seeds a Random Number Generator (RNG) to perform local reproducible perturbations on model weights and calculate and exchange scalar projected gradients, which are then used to update each model. By using a (machine, sample) identifier as the random seed, each model can regenerate one another's perturbations. As machines only exchange single-byte projected gradients, this is highly communication efficient. There are also potential privacy benefits, as projected gradients may be calculated on different training data, and models never access the other's data. Our approach not only drastically reduces communication bandwidth requirements but also accommodates dynamic addition or removal of machines during the training process and retains the memory-efficient and inference-only advantages of recent work. We perform proof-of-concept experiments to demonstrate the potential usefulness of this method, building off of rich literature on distributed optimization and memory-efficient training."
                    },
                    {
                        "title": "Strategic Reasoning with Language Models",
                        "abstract": "Strategic reasoning enables agents to cooperate, communicate, and compete with other agents in diverse situations. Existing approaches to solving strategic games rely on extensive training, yielding strategies that do not generalize to new scenarios or games without retraining. Large Language Models (LLMs), with their ability to comprehend and generate complex, context-rich language, could prove powerful as tools for strategic gameplay. This paper introduces an approach that uses pretrained LLMs with few-shot chain-of-thought examples to enable strategic reasoning for AI agents. Our approach uses systematically generated demonstrations of reasoning about states, values, and beliefs to prompt the model. Using extensive variations of simple matrix games, we show that strategies that are derived based on systematically generated prompts generalize almost perfectly to new game structures, alternate objectives, and hidden information. Additionally, we demonstrate our approach can lead to human-like negotiation strategies in realistic scenarios without any extra training or fine-tuning. Our results highlight the ability of LLMs, guided by systematic reasoning demonstrations, to adapt and excel in diverse strategic scenarios."
                    },
                    {
                        "title": "CLEVRER-Humans: Describing Physical and Causal Events the Human Way",
                        "abstract": "Building machines that can reason about physical events and their causal relationships is crucial for flexible interaction with the physical world. However, most existing physical and causal reasoning benchmarks are exclusively based on synthetically generated events and synthetic natural language descriptions of causal relationships. This design brings up two issues. First, there is a lack of diversity in both event types and natural language descriptions; second, causal relationships based on manually-defined heuristics are different from human judgments. To address both shortcomings, we present the CLEVRER-Humans benchmark, a video reasoning dataset for causal judgment of physical events with human labels. We employ two techniques to improve data collection efficiency: first, a novel iterative event cloze task to elicit a new representation of events in videos, which we term Causal Event Graphs (CEGs); second, a data augmentation technique based on neural language generative models. We convert the collected CEGs into questions and answers to be consistent with prior work. Finally, we study a collection of baseline approaches for CLEVRER-Humans question-answering, highlighting the great challenges set forth by our benchmark."
                    },
                    {
                        "title": "Understanding Social Reasoning in Language Models with Language Models",
                        "abstract": "As Large Language Models (LLMs) become increasingly integrated into our everyday lives, understanding their ability to comprehend human mental states becomes critical for ensuring effective interactions. However, despite the recent attempts to assess the Theory-of-Mind (ToM) reasoning capabilities of LLMs, the degree to which these models can align with human ToM remains a nuanced topic of exploration. This is primarily due to two distinct challenges: (1) the presence of inconsistent results from previous evaluations, and (2) concerns surrounding the validity of existing evaluation methodologies. To address these challenges, we present a novel framework for procedurally generating evaluations with LLMs by populating causal templates. Using our framework, we create a new social reasoning benchmark (BigToM) for LLMs which consists of 25 controls and 5,000 model-written evaluations. We find that human participants rate the quality of our benchmark higher than previous crowd-sourced evaluations and comparable to expert-written evaluations. Using BigToM, we evaluate the social reasoning capabilities of a variety of LLMs and compare model performances with human performance. Our results suggest that GPT4 has ToM capabilities that mirror human inference patterns, though less reliable, while other LLMs struggle."
                    },
                    {
                        "title": "On Rate-Distortion Theory in Capacity-Limited Cognition & Reinforcement Learning",
                        "abstract": "Throughout the cognitive-science literature, there is widespread agreement that decision-making agents operating in the real world do so under limited information-processing capabilities and without access to unbounded cognitive or computational resources. Prior work has drawn inspiration from this fact and leveraged an information-theoretic model of such behaviors or policies as communication channels operating under a bounded rate constraint. Meanwhile, a parallel line of work also capitalizes on the same principles from rate-distortion theory to formalize capacity-limited decision making through the notion of a learning target, which facilitates Bayesian regret bounds for provably-efficient learning algorithms. In this paper, we aim to elucidate this latter perspective by presenting a brief survey of these information-theoretic models of capacity-limited decision making in biological and artificial agents."
                    },
                    {
                        "title": "I N THE ZONE: M EASURING DIFFICULTY AND PROGRESSION IN CURRICULUM GENERATION",
                        "abstract": "A common strategy in curriculum generation for reinforcement learning is to train a teacher network to generate tasks that enable student learning. But, what kind of tasks enables this? One answer is tasks belonging to a student\u2019s zone of proximal development (ZPD), a concept from developmental psychology. These are tasks that are not too easy and not too hard for the student. Albeit intuitive, ZPD is not well understood computationally. We propose ZONE , a novel computational framework that operationalizes ZPD. It formalizes ZPD through the language of Bayesian probability theory, revealing that tasks should be selected by difficulty (the student\u2019s probability of task success) and learning progression (the degree of change in the student\u2019s model parameters). ZONE instantiates two techniques that enforce the teacher to pick tasks within the student\u2019s ZPD. One is R EJECT , which rejects tasks outside of a difficulty scope"
                    },
                    {
                        "title": "Parsel: A Unified Natural Language Framework for Algorithmic Reasoning",
                        "abstract": "Despite recent success in large language model (LLM) reasoning, LLMs still struggle with hierarchical multi-step reasoning like generating complex programs. In these cases, humans often start with a high-level algorithmic design and implement each part gradually. We introduce Parsel 2 , a framework enabling automatic implementation and validation of complex algorithms with code LLMs, based on hierarchical function descriptions in natural language. Parsel can be used across domains requiring hierarchical reasoning, e.g. code synthesis, theorem proving, and robotic planning. We demonstrate Parsel\u2019s capabilities by using it to generate complex programs that cannot currently be automatically implemented from one description and backtranslating Python programs in the APPS dataset. Beyond modeling capabilities, Parsel allows problem-solving with high-level algorithmic designs, bene\ufb01ting both students and professional programmers."
                    }
                ]
            }
        }
    },
    "2305.19521": {
        "paper_data": {
            "title": "Incremental Randomized Smoothing Certification",
            "url": "http://arxiv.org/abs/2305.19521v2",
            "arxiv_id": "2305.19521",
            "authors": [
                "Shubham Ugare",
                "Tarun Suresh",
                "Debangshu Banerjee",
                "Gagandeep Singh",
                "Sasa Misailovic"
            ],
            "abstract": "Randomized smoothing-based certification is an effective approach for obtaining robustness certificates of deep neural networks (DNNs) against adversarial attacks. This method constructs a smoothed DNN model and certifies its robustness through statistical sampling, but it is computationally expensive, especially when certifying with a large number of samples. Furthermore, when the smoothed model is modified (e.g., quantized or pruned), certification guarantees may not hold for the modified DNN, and recertifying from scratch can be prohibitively expensive.   We present the first approach for incremental robustness certification for randomized smoothing, IRS. We show how to reuse the certification guarantees for the original smoothed model to certify an approximated model with very few samples. IRS significantly reduces the computational cost of certifying modified DNNs while maintaining strong robustness guarantees. We experimentally demonstrate the effectiveness of our approach, showing up to 3x certification speedup over the certification that applies randomized smoothing of the approximate model from scratch.",
            "introduction": "   1 Introduction  Ensuring the robustness of deep neural networks (DNNs) to input perturbations is gaining increased attention from both users and regulators in various application domains\u00a0Bojarski et\u00a0al. (2016); Amato et\u00a0al. (2013); Julian et\u00a0al. (2018); ISO . Out of many techniques for obtaining robustness certificaties, statistical methods currently offer the greatest scalability. Randomized smoothing (RS) is a popular statistical certification method by constructing a smoothed model g\ud835\udc54gitalic_g from a base network f\ud835\udc53fitalic_f under noise\u00a0Cohen et\u00a0al. (2019). To certify the model g\ud835\udc54gitalic_g on an input, RS certification checks if the estimated lower bound on the probability of the top class is greater than the upper bound on the probability of the runner-up class (with high confidence). RS certification computes the certified accuracy metric of the DNN on the set of test inputs as a proxy for the DNN robustness. However, despite its effectiveness, RS-based certification can be computationally expensive as it requires DNN inference on a large number of corruptions per input.   The high cost of certification complicates the DNN deployment process, which has become increasingly iterative: the networks are often modified post-training to improve their execution time and/or accuracy. Especially, deploying DNNs on real-world systems with bounded computing resources (e.g., edge devices or GPUs with limited memory), has led to various techniques for approximating DNNs. Common approximation techniques include quantization \u2013 reducing the numerical precision of weights\u00a0(Fiesler et\u00a0al., 1990; Balzer et\u00a0al., 1991), and pruning \u2013 removing weights that have minimal impact on accuracy\u00a0(Janowsky, 1989; Reed, 1993).    Common to all of these approximations is that the network behavior (e.g., the classifications) remains the same on most inputs, its architecture does not change, and many weights are only slightly changed. When a user seeks to select a robust and accurate DNN from these possible approximations, RS needs to be performed to compute the robustness of all candidate networks.  For instance, in the context of approximation tuning, there are multiple choices for approximation where different quantization or pruning strategies are applied at different layers. Tools such as Chen et\u00a0al. (2018b; a); Sharif et\u00a0al. (2019); Zhao et\u00a0al. (2023) use approximations iteratively and test the network at each step. To ensure DNN robustness when using such tools, one would need to check certified accuracy, computed using RS on test data in each step. However, performing RS to compute certified accuracy from scratch can take hours as shown in our experiments even for a single network (with only 500 test images).  Therefore, a major encumbrance of almost all existing RS-based certification practices in the above setting, is that the expensive certification needs to be re-run from scratch for each approximate network. Overcoming this main limitation requires addressing the following fundamental problem:     How can we leverage the information generated while certifying a given network to speed up the certification of similar networks?    This Work.  We present the first incremental RS-based certification framework called Incremental Randomized Smoothing (IRS) to answer this question. The primary objective of our work is to improve the sample complexity of the certification process of similar networks on a predefined test set. Improved sample complexity results in overall speedup in certification, and it reduces the energy requirement and memory footprint of the",
            "references": [
                {
                    "title": "Artificial Neural Networks in Medical Diagnosis",
                    "abstract": "Artificial neural networks are finding many uses in the medical diagnosis application. The goal of this paper is to evaluate artificial neural network in disease diagnosis. Two cases are studied. The first one is acute nephritis disease; data is the disease symptoms. The second is the heart disease; data is on cardiac Single Proton Emission Computed Tomography (SPECT) images. Each patient classified into two categories: infected and non-infected. Classification is an important tool in medical diagnosis decision support. Feed-forward back propagation neural network is used as a classifier to distinguish between infected or non-infected person in both cases. The results of applying the artificial neural networks methodology to acute nephritis diagnosis based upon selected symptoms show abilities of the network to learn the patterns corresponding to symptoms of the person. In this study, the data were obtained from UCI machine learning repository in order to diagnosed diseases. The data is separated into inputs and targets. The targets for the neural network will be identified with 1's as infected and will be identified with 0's as non-infected. In the diagnosis of acute nephritis disease; the percent correctly classified in the simulation sample by the feed-forward back propagation network is 99 percent while in the diagnosis of heart disease; the percent correctly classified in the simulation sample by the feed-forward back propagation network is 95 percent."
                },
                {
                    "title": "Synthesizing Precise Static Analyzers for Automatic Differentiation",
                    "abstract": "We present Pasado, a technique for synthesizing precise static analyzers for Automatic Differentiation. Our technique allows one to automatically construct a static analyzer specialized for the Chain Rule, Product Rule, and Quotient Rule computations for Automatic Differentiation in a way that abstracts all of the nonlinear operations of each respective rule simultaneously. By directly synthesizing an abstract transformer for the composite expressions of these 3 most common rules of AD, we are able to obtain significant precision improvement compared to prior works which compose standard abstract transformers together suboptimally. We prove our synthesized static analyzers sound and additionally demonstrate the generality of our approach by instantiating these AD static analyzers with different nonlinear functions, different abstract domains (both intervals and zonotopes) and both forward-mode and reverse-mode AD. We evaluate Pasado on multiple case studies, namely soundly computing bounds on a neural network\u2019s local Lipschitz constant, soundly bounding the sensitivities of financial models, certifying monotonicity, and lastly, bounding sensitivities of the solutions of differential equations from climate science and chemistry for verified ranges of initial conditions and parameters. The local Lipschitz constants computed by Pasado on our largest CNN are up to 2750\u00d7 more precise compared to the existing state-of-the-art zonotope analysis. The bounds obtained on the sensitivities of the climate, chemical, and financial differential equation solutions are between 1.31 \u2212 2.81\u00d7 more precise (on average) compared to a state-of-the-art zonotope analysis."
                },
                {
                    "title": "Incremental Verification of Neural Networks",
                    "abstract": "Complete verification of deep neural networks (DNNs) can exactly determine whether the DNN satisfies a desired trustworthy property (e.g., robustness, fairness) on an infinite set of inputs or not. Despite the tremendous progress to improve the scalability of complete verifiers over the years on individual DNNs, they are inherently inefficient when a deployed DNN is updated to improve its inference speed or accuracy. The inefficiency is because the expensive verifier needs to be run from scratch on the updated DNN. To improve efficiency, we propose a new, general framework for incremental and complete DNN verification based on the design of novel theory, data structure, and algorithms. Our contributions implemented in a tool named IVAN yield an overall geometric mean speedup of 2.4x for verifying challenging MNIST and CIFAR10 classifiers and a geometric mean speedup of 3.8x for the ACAS-XU classifiers over the state-of-the-art baselines."
                },
                {
                    "title": "Collective Robustness Certificates: Exploiting Interdependence in Graph Neural Networks",
                    "abstract": "In tasks like node classification, image segmentation, and named-entity recognition we have a classifier that simultaneously outputs multiple predictions (a vector of labels) based on a single input, i.e. a single graph, image, or document respectively. Existing adversarial robustness certificates consider each prediction independently and are thus overly pessimistic for such tasks. They implicitly assume that an adversary can use different perturbed inputs to attack different predictions, ignoring the fact that we have a single shared input. We propose the first collective robustness certificate which computes the number of predictions that are simultaneously guaranteed to remain stable under perturbation, i.e. cannot be attacked. We focus on Graph Neural Networks and leverage their locality property - perturbations only affect the predictions in a close neighborhood - to fuse multiple single-node certificates into a drastically stronger collective certificate. For example, on the Citeseer dataset our collective certificate for node classification increases the average number of certifiable feature perturbations from $7$ to $351$."
                },
                {
                    "title": "A general construction for abstract interpretation of higher-order automatic differentiation",
                    "abstract": "We present a novel, general construction to abstractly interpret higher-order automatic differentiation (AD). Our construction allows one to instantiate an abstract interpreter for computing derivatives up to a chosen order. Furthermore, since our construction reduces the problem of abstractly reasoning about derivatives to abstractly reasoning about real-valued straight-line programs, it can be instantiated with almost any numerical abstract domain, both relational and non-relational. We formally establish the soundness of this construction. We implement our technique by instantiating our construction with both the non-relational interval domain and the relational zonotope domain to compute both first and higher-order derivatives. In the latter case, we are the first to apply a relational domain to automatic differentiation for abstracting higher-order derivatives, and hence we are also the first abstract interpretation work to track correlations across not only different variables, but different orders of derivatives. We evaluate these instantiations on multiple case studies, namely robustly explaining a neural network and more precisely computing a neural network\u2019s Lipschitz constant. For robust interpretation, first and second derivatives computed via zonotope AD are up to 4.76\u00d7 and 6.98\u00d7 more precise, respectively, compared to interval AD. For Lipschitz certification, we obtain bounds that are up to 11,850\u00d7 more precise with zonotopes, compared to the state-of-the-art interval-based tool."
                },
                {
                    "title": "Sound Randomized Smoothing in Floating-Point Arithmetics",
                    "abstract": "Randomized smoothing is sound when using infinite precision. However, we show that randomized smoothing is no longer sound for limited floating-point precision. We present a simple example where randomized smoothing certifies a radius of $1.26$ around a point, even though there is an adversarial example in the distance $0.8$ and extend this example further to provide false certificates for CIFAR10. We discuss the implicit assumptions of randomized smoothing and show that they do not apply to generic image classification models whose smoothed versions are commonly certified. In order to overcome this problem, we propose a sound approach to randomized smoothing when using floating-point precision with essentially equal speed and matching the certificates of the standard, unsound practice for standard classifiers tested so far. Our only assumption is that we have access to a fair coin."
                },
                {
                    "title": "Proof transfer for fast certification of multiple approximate neural networks",
                    "abstract": "Developers of machine learning applications often apply post-training neural network optimizations, such as quantization and pruning, that approximate a neural network to speed up inference and reduce energy consumption, while maintaining high accuracy and robustness. Despite a recent surge in techniques for the robustness verification of neural networks, a major limitation of almost all state-of-the-art approaches is that the verification needs to be run from scratch every time the network is even slightly modified. Running precise end-to-end verification from scratch for every new network is expensive and impractical in many scenarios that use or compare multiple approximate network versions, and the robustness of all the networks needs to be verified efficiently. We present FANC, the first general technique for transferring proofs between a given network and its multiple approximate versions without compromising verifier precision. To reuse the proofs obtained when verifying the original network, FANC generates a set of templates \u2013 connected symbolic shapes at intermediate layers of the original network \u2013 that capture the proof of the property to be verified. We present novel algorithms for generating and transforming templates that generalize to a broad range of approximate networks and reduce the verification cost. We present a comprehensive evaluation demonstrating the effectiveness of our approach. We consider a diverse set of networks obtained by applying popular approximation techniques such as quantization and pruning on fully-connected and convolutional architectures and verify their robustness against different adversarial attacks such as adversarial patches, L0, rotation and brightening. Our results indicate that FANC can significantly speed up verification with state-of-the-art verifier, DeepZ by up to 4.1x."
                },
                {
                    "title": "Statheros: Compiler for Efficient Low-Precision Probabilistic Programming",
                    "abstract": "As Edge and IoT computing devices process noisy data or make decisions in uncertain environments, they require frameworks for inexpensive, yet accurate probabilistic inference. Probabilistic programming has emerged as a powerful way for developers to write high-level programs, while abstracting away the implementation details of inference. However, the existing algorithms are slow and often assumed to require precise calculations. We present Statheros, the first compiler for low-level, fixed-point approximation of probabilistic programming. Statheros compiles programs to fixed-point inference procedures and is able to determine the optimal fixed-point type to use. We evaluate Statheros on 13 benchmarks and three embedded platforms. The results show that Statheros-generated code is 11. 5x (Arduino), 3. 8x (PocketBeagle), and 2. 2x (Raspberry Pi) faster than single-precision floating-point computation, with minimal accuracy loss."
                },
                {
                    "title": "Integer-arithmetic-only Certified Robustness for Quantized Neural Networks",
                    "abstract": "Adversarial data examples have drawn significant attention from the machine learning and security communities. A line of work on tackling adversarial examples is certified robustness via randomized smoothing that can provide a theoretical robustness guarantee. However, such a mechanism usually uses floating-point arithmetic for calculations in inference and requires large memory footprints and daunting computational costs. These defensive models cannot run efficiently on edge devices nor be deployed on integer-only logical units such as Turing Tensor Cores or integer-only ARM processors. To overcome these challenges, we propose an integer randomized smoothing approach with quantization to convert any classifier into a new smoothed classifier, which uses integer-only arithmetic for certified robustness against adversarial perturbations. We prove a tight robustness guarantee under \u21132-norm for the proposed approach. We show our approach can obtain a comparable accuracy and 4\u00d7 \u223c 5\u00d7 speedup over floating-point arithmetic certified robust methods on general-purpose CPUs and mobile devices on two distinct datasets (CIFAR-10 and Caltech-101)."
                },
                {
                    "title": "Scalable Certified Segmentation via Randomized Smoothing",
                    "abstract": "We present a new certification method for image and point cloud segmentation based on randomized smoothing. The method leverages a novel scalable algorithm for prediction and certification that correctly accounts for multiple testing, necessary for ensuring statistical guarantees. The key to our approach is reliance on established multiple-testing correction mechanisms as well as the ability to abstain from classifying single pixels or points while still robustly segmenting the overall input. Our experimental evaluation on synthetic data and challenging datasets, such as Pascal Context, Cityscapes, and ShapeNet, shows that our algorithm can achieve, for the first time, competitive accuracy and certification guarantees on real-world segmentation tasks. We provide an implementation at https://github.com/eth-sri/segmentation-smoothing."
                },
                {
                    "title": "Boosting Randomized Smoothing with Variance Reduced Classifiers",
                    "abstract": "Randomized Smoothing (RS) is a promising method for obtaining robustness certificates by evaluating a base model under noise. In this work, we: (i) theoretically motivate why ensembles are a particularly suitable choice as base models for RS, and (ii) empirically confirm this choice, obtaining state-of-the-art results in multiple settings. The key insight of our work is that the reduced variance of ensembles over the perturbations introduced in RS leads to significantly more consistent classifications for a given input. This, in turn, leads to substantially increased certifiable radii for samples close to the decision boundary. Additionally, we introduce key optimizations which enable an up to 55-fold decrease in sample complexity of RS for predetermined radii, thus drastically reducing its computational overhead. Experimentally, we show that ensembles of only 3 to 10 classifiers consistently improve on their strongest constituting model with respect to their average certified radius (ACR) by 5% to 21% on both CIFAR10 and ImageNet, achieving a new state-of-the-art ACR of 0.86 and 1.11, respectively. We release all code and models required to reproduce our results at https://github.com/eth-sri/smoothing-ensembles."
                },
                {
                    "title": "Demanded abstract interpretation",
                    "abstract": "We consider the problem of making expressive static analyzers interactive. Formal static analysis is seeing increasingly widespread adoption as a tool for verification and bug-finding, but even with powerful cloud infrastructure it can take minutes or hours to get batch analysis results after a code change. While existing techniques offer some demand-driven or incremental aspects for certain classes of analysis, the fundamental challenge we tackle is doing both for arbitrary abstract interpreters. Our technique, demanded abstract interpretation, lifts program syntax and analysis state to a dynamically evolving graph structure, in which program edits, client-issued queries, and evaluation of abstract semantics are all treated uniformly. The key difficulty addressed by our approach is the application of general incremental computation techniques to the complex, cyclic dependency structure induced by abstract interpretation of loops with widening operators. We prove that desirable abstract interpretation meta-properties, including soundness and termination, are preserved in our approach, and that demanded analysis results are equal to those computed by a batch abstract interpretation. Experimental results suggest promise for a prototype demanded abstract interpretation framework: by combining incremental and demand-driven techniques, our framework consistently delivers analysis results at interactive speeds, answering 95% of queries within 1.2 seconds."
                },
                {
                    "title": "PointGuard: Provably Robust 3D Point Cloud Classification",
                    "abstract": "3D point cloud classification has many safety-critical applications such as autonomous driving and robotic grasping. However, several studies showed that it is vulnerable to adversarial attacks. In particular, an attacker can make a classifier predict an incorrect label for a 3D point cloud via carefully modifying, adding, and/or deleting a small number of its points. Randomized smoothing is state-of-the-art technique to build certifiably robust 2D image classifiers. However, when applied to 3D point cloud classification, randomized smoothing can only certify robustness against adversarially modified points.In this work, we propose PointGuard, the first defense that has provable robustness guarantees against adversarially modified, added, and/or deleted points. Specifically, given a 3D point cloud and an arbitrary point cloud classifier, our PointGuard first creates multiple subsampled point clouds, each of which contains a random subset of the points in the original point cloud; then our PointGuard predicts the label of the original point cloud as the majority vote among the labels of the subsampled point clouds predicted by the point cloud classifier. Our first major theoretical contribution is that we show PointGuard provably predicts the same label for a 3D point cloud when the number of adversarially modified, added, and/or deleted points is bounded. Our second major theoretical contribution is that we prove the tightness of our derived bound when no assumptions on the point cloud classifier are made. Moreover, we design an efficient algorithm to compute our certified robustness guarantees. We also empirically evaluate PointGuard on ModelNet40 and ScanNet benchmark datasets."
                },
                {
                    "title": "Higher-Order Certification for Randomized Smoothing",
                    "abstract": "Randomized smoothing is a recently proposed defense against adversarial attacks that has achieved SOTA provable robustness against $\\ell_2$ perturbations. A number of publications have extended the guarantees to other metrics, such as $\\ell_1$ or $\\ell_\\infty$, by using different smoothing measures. Although the current framework has been shown to yield near-optimal $\\ell_p$ radii, the total safety region certified by the current framework can be arbitrarily small compared to the optimal. In this work, we propose a framework to improve the certified safety region for these smoothed classifiers without changing the underlying smoothing scheme. The theoretical contributions are as follows: 1) We generalize the certification for randomized smoothing by reformulating certified radius calculation as a nested optimization problem over a class of functions. 2) We provide a method to calculate the certified safety region using $0^{th}$-order and $1^{st}$-order information for Gaussian-smoothed classifiers. We also provide a framework that generalizes the calculation for certification using higher-order information. 3) We design efficient, high-confidence estimators for the relevant statistics of the first-order information. Combining the theoretical contribution 2) and 3) allows us to certify safety region that are significantly larger than the ones provided by the current methods. On CIFAR10 and Imagenet datasets, the new regions certified by our approach achieve significant improvements on general $\\ell_1$ certified radii and on the $\\ell_2$ certified radii for color-space attacks ($\\ell_2$ restricted to 1 channel) while also achieving smaller improvements on the general $\\ell_2$ certified radii. Our framework can also provide a way to circumvent the current impossibility results on achieving higher magnitude of certified radii without requiring the use of data-dependent smoothing techniques."
                },
                {
                    "title": "Certifying Confidence via Randomized Smoothing",
                    "abstract": "Randomized smoothing has been shown to provide good certified-robustness guarantees for high-dimensional classification problems. It uses the probabilities of predicting the top two most-likely classes around an input point under a smoothing distribution to generate a certified radius for a classifier's prediction. However, most smoothing methods do not give us any information about the \\emph{confidence} with which the underlying classifier (e.g., deep neural network) makes a prediction. In this work, we propose a method to generate certified radii for the prediction confidence of the smoothed classifier. We consider two notions for quantifying confidence: average prediction score of a class and the margin by which the average prediction score of one class exceeds that of another. We modify the Neyman-Pearson lemma (a key theorem in randomized smoothing) to design a procedure for computing the certified radius where the confidence is guaranteed to stay above a certain threshold. Our experimental results on CIFAR-10 and ImageNet datasets show that using information about the distribution of the confidence scores allows us to achieve a significantly better certified radius than ignoring it. Thus, we demonstrate that extra information about the base classifier at the input point can help improve certified guarantees for the smoothed classifier."
                },
                {
                    "title": "Certified Robustness of Graph Classification against Topology Attack with Randomized Smoothing",
                    "abstract": "Graph classification has practical applications in diverse fields. Recent studies show that graph-based machine learning models are especially vulnerable to adversarial perturbations due to the non i.i. d nature of graph data. By adding or deleting a small number of edges in the graph, adversaries could greatly change the graph label predicted by a graph classification model. In this work, we propose to build a smoothed graph classification model with certified robustness guarantee. We have proven that the resulting graph classification model would output the same prediction for a graph under l0 bounded adversarial perturbation. We also evaluate the effectiveness of our approach under graph convolutional network (GCN) based multi-class graph classification model."
                },
                {
                    "title": "Efficient Robustness Certificates for Discrete Data: Sparsity-Aware Randomized Smoothing for Graphs, Images and More",
                    "abstract": "Existing techniques for certifying the robustness of models for discrete data either work only for a small class of models or are general at the expense of efficiency or tightness. Moreover, they do not account for sparsity in the input which, as our findings show, is often essential for obtaining non-trivial guarantees. We propose a model-agnostic certificate based on the randomized smoothing framework which subsumes earlier work and is tight, efficient, and sparsity-aware. Its computational complexity does not depend on the number of discrete categories or the dimension of the input (e.g. the graph size), making it highly scalable. We show the effectiveness of our approach on a wide variety of models, datasets, and tasks -- specifically highlighting its use for Graph Neural Networks. So far, obtaining provable guarantees for GNNs has been difficult due to the discrete and non-i.i.d. nature of graph data. Our method can certify any GNN and handles perturbations to both the graph structure and the node attributes."
                },
                {
                    "title": "Certified Robustness of Graph Neural Networks against Adversarial Structural Perturbation",
                    "abstract": "Graph neural networks (GNNs) have recently gained much attention for node and graph classification tasks on graph-structured data. However, multiple recent works showed that an attacker can easily make GNNs predict incorrectly via perturbing the graph structure, i.e., adding or deleting edges in the graph. We aim to defend against such attacks via developing certifiably robust GNNs. Specifically, we prove the first certified robustness guarantee of any GNN for both node and graph classifications against structural perturbation. Moreover, we show that our certified robustness guarantee is tight. Our results are based on a recently proposed technique called randomized smoothing, which we extend to graph data. We also empirically evaluate our method for both node and graph classifications on multiple GNNs and multiple benchmark datasets. For instance, on the Cora dataset, Graph Convolutional Network with our randomized smoothing can achieve a certified accuracy of 0.49 when the attacker can arbitrarily add/delete at most 15 edges in the graph."
                },
                {
                    "title": "Detection as Regression: Certified Object Detection by Median Smoothing",
                    "abstract": "Despite the vulnerability of object detectors to adversarial attacks, very few defenses are known to date. While adversarial training can improve the empirical robustness of image classifiers, a direct extension to object detection is very expensive. This work is motivated by recent progress on certified classification by randomized smoothing. We start by presenting a reduction from object detection to a regression problem. Then, to enable certified regression, where standard mean smoothing fails, we propose median smoothing, which is of independent interest. We obtain the first model-agnostic, training-free, and certified defense for object detection against $\\ell_2$-bounded attacks."
                },
                {
                    "title": "A Framework for robustness Certification of Smoothed Classifiers using F-Divergences",
                    "abstract": "Formal verification techniques that compute provable guarantees on properties of machine learning models, like robustness to norm-bounded adversarial perturbations, have yielded impressive results. Although most techniques developed so far requires knowledge of the architecture of the machine learning model and remains hard to scale to complex prediction pipelines, the method of randomized smoothing has been shown to overcome many of these obstacles. By requiring only black-box access to the underlying model, randomized smoothing scales to large architectures and is agnostic to the internals of the network. However, past work on randomized smoothing has focused on restricted classes of smoothing measures or perturbations (like Gaussian or discrete) and has only been able to prove robustness with respect to simple norm bounds. In this paper we introduce a general framework for proving robustness properties of smoothed machine learning models in the black-box setting. Specifically, we extend randomized smoothing procedures to handle arbitrary smoothing measures and prove robustness of the smoothed classifier by using $f$-divergences. Our methodology achieves state-of-the-art}certified robustness on MNIST, CIFAR-10 and ImageNet and also audio classification task, Librispeech, with respect to several classes of adversarial perturbations."
                },
                {
                    "title": "TSS: Transformation-Specific Smoothing for Robustness Certification",
                    "abstract": "As machine learning (ML) systems become pervasive, safeguarding their security is critical. However, recently it has been demonstrated that motivated adversaries are able to mislead ML systems by perturbing test data using semantic transformations. While there exists a rich body of research providing provable robustness guarantees for ML models against Lp bounded adversarial perturbations, guarantees against semantic perturbations remain largely underexplored. In this paper, we provide TSS-a unified framework for certifying ML robustness against general adversarial semantic transformations. First, depending on the properties of each transformation, we divide common transformations into two categories, namely resolvable (e.g., Gaussian blur) and differentially resolvable (e.g., rotation) transformations. For the former, we propose transformation-specific randomized smoothing strategies and obtain strong robustness certification. The latter category covers transformations that involve interpolation errors, and we propose a novel approach based on stratified sampling to certify the robustness. Our framework TSS leverages these certification strategies and combines with consistency-enhanced training to provide rigorous certification of robustness. We conduct extensive experiments on over ten types of challenging semantic transformations and show that TSS significantly outperforms the state of the art. Moreover, to the best of our knowledge, TSS is the first approach that achieves nontrivial certified robustness on the large-scale ImageNet dataset. For instance, our framework achieves 30.4% certified robust accuracy against rotation attack (within \u00b130\u00b0) on ImageNet. Moreover, to consider a broader range of transformations, we show TSS is also robust against adaptive attacks and unforeseen image corruptions such as CIFAR-10-C and ImageNet-C."
                },
                {
                    "title": "Certified Defense to Image Transformations via Randomized Smoothing",
                    "abstract": "We extend randomized smoothing to cover parameterized transformations (e.g., rotations, translations) and certify robustness in the parameter space (e.g., rotation angle). This is particularly challenging as interpolation and rounding effects mean that image transformations do not compose, in turn preventing direct certification of the perturbed image (unlike certification with $\\ell^p$ norms). We address this challenge by introducing three different defenses, each with a different guarantee (heuristic, distributional and individual) stemming from the method used to bound the interpolation error. Importantly, in the individual case, we show how to efficiently compute the inverse of an image transformation, enabling us to provide individual guarantees in the online setting. We provide an implementation of all methods at https://github.com/eth-sri/transformation-smoothing."
                },
                {
                    "title": "(De)Randomized Smoothing for Certifiable Defense against Patch Attacks",
                    "abstract": "Patch adversarial attacks on images, in which the attacker can distort pixels within a region of bounded size, are an important threat model since they provide a quantitative model for physical adversarial attacks. In this paper, we introduce a certifiable defense against patch attacks that guarantees for a given image and patch attack size, no patch adversarial examples exist. Our method is related to the broad class of randomized smoothing robustness schemes which provide high-confidence probabilistic robustness certificates. By exploiting the fact that patch attacks are more constrained than general sparse attacks, we derive meaningfully large robustness certificates. Additionally, the algorithm we propose is de-randomized, providing deterministic certificates. To the best of our knowledge, there exists only one prior method for certifiable defense against patch attacks, which relies on interval bound propagation. While this sole existing method performs well on MNIST, it has several limitations: it requires computationally expensive training, does not scale to ImageNet, and performs poorly on CIFAR-10. In contrast, our proposed method effectively addresses all of these issues: our classifier can be trained quickly, achieves high clean and certified robust accuracy on CIFAR-10, and provides certificates at the ImageNet scale. For example, for a 5*5 patch attack on CIFAR-10, our method achieves up to around 57.8% certified accuracy (with a classifier around 83.9% clean accuracy), compared to at most 30.3% certified accuracy for the existing method (with a classifier with around 47.8% clean accuracy), effectively establishing a new state-of-the-art. Code is available at this https URL."
                },
                {
                    "title": "HYDRA: Pruning Adversarially Robust Neural Networks",
                    "abstract": "In safety-critical but computationally resource-constrained applications, deep learning faces two key challenges: lack of robustness against adversarial attacks and large neural network size (often millions of parameters). While the research community has extensively explored the use of robust training and network pruning independently to address one of these challenges, only a few recent works have studied them jointly. However, these works inherit a heuristic pruning strategy that was developed for benign training, which performs poorly when integrated with robust training techniques, including adversarial training and verifiable robust training. To overcome this challenge, we propose to make pruning techniques aware of the robust training objective and let the training objective guide the search for which connections to prune. We realize this insight by formulating the pruning objective as an empirical risk minimization problem which is solved efficiently using SGD. We demonstrate that our approach, titled HYDRA, achieves compressed networks with state-of-the-art benign and robust accuracy, simultaneously. We demonstrate the success of our approach across CIFAR-10, SVHN, and ImageNet dataset with four robust training techniques: iterative adversarial training, randomized smoothing, MixTrain, and CROWN-IBP. We also demonstrate the existence of highly robust sub-networks within non-robust networks. Our code and compressed networks are publicly available at \\url{this https URL}."
                },
                {
                    "title": "Black-Box Certification with Randomized Smoothing: A Functional Optimization Based Framework",
                    "abstract": "Randomized classifiers have been shown to provide a promising approach for achieving certified robustness against adversarial attacks in deep learning. However, most existing methods only leverage Gaussian smoothing noise and only work for $\\ell_2$ perturbation. We propose a general framework of adversarial certification with non-Gaussian noise and for more general types of attacks, from a unified functional optimization perspective. Our new framework allows us to identify a key trade-off between accuracy and robustness via designing smoothing distributions, helping to design new families of non-Gaussian smoothing distributions that work more efficiently for different $\\ell_p$ settings, including $\\ell_1$, $\\ell_2$ and $\\ell_\\infty$ attacks. Our proposed methods achieve better certification results than previous works and provide a new perspective on randomized smoothing certification."
                },
                {
                    "title": "Randomized Smoothing of All Shapes and Sizes",
                    "abstract": "Randomized smoothing is the current state-of-the-art defense with provable robustness against $\\ell_2$ adversarial attacks. Many works have devised new randomized smoothing schemes for other metrics, such as $\\ell_1$ or $\\ell_\\infty$; however, substantial effort was needed to derive such new guarantees. This begs the question: can we find a general theory for randomized smoothing? \nWe propose a novel framework for devising and analyzing randomized smoothing schemes, and validate its effectiveness in practice. Our theoretical contributions are: (1) we show that for an appropriate notion of \"optimal\", the optimal smoothing distributions for any \"nice\" norms have level sets given by the norm's *Wulff Crystal*; (2) we propose two novel and complementary methods for deriving provably robust radii for any smoothing distribution; and, (3) we show fundamental limits to current randomized smoothing techniques via the theory of *Banach space cotypes*. By combining (1) and (2), we significantly improve the state-of-the-art certified accuracy in $\\ell_1$ on standard datasets. Meanwhile, we show using (3) that with only label statistics under random input perturbations, randomized smoothing cannot achieve nontrivial certified accuracy against perturbations of $\\ell_p$-norm $\\Omega(\\min(1, d^{\\frac{1}{p} - \\frac{1}{2}}))$, when the input dimension $d$ is large. We provide code in this http URL."
                },
                {
                    "title": "Certified Robustness to Label-Flipping Attacks via Randomized Smoothing",
                    "abstract": "Machine learning algorithms are known to be susceptible to data poisoning attacks, where an adversary manipulates the training data to degrade performance of the resulting classifier. In this work, we present a unifying view of randomized smoothing over arbitrary functions, and we leverage this novel characterization to propose a new strategy for building classifiers that are pointwise-certifiably robust to general data poisoning attacks. As a specific instantiation, we utilize our framework to build linear classifiers that are robust to a strong variant of label flipping, where each test example is targeted independently. In other words, for each test point, our classifier includes a certification that its prediction would be the same had some number of training labels been changed adversarially. Randomized smoothing has previously been used to guarantee---with high probability---test-time robustness to adversarial manipulation of the input to a classifier; we derive a variant which provides a deterministic, analytical bound, sidestepping the probabilistic certificates that traditionally result from the sampling subprocedure. Further, we obtain these certified bounds with minimal additional runtime complexity over standard classification and no assumptions on the train or test distributions. We generalize our results to the multi-class case, providing the first multi-class classification algorithm that is certifiably robust to label-flipping attacks."
                },
                {
                    "title": "Certified Robustness for Top-k Predictions against Adversarial Perturbations via Randomized Smoothing",
                    "abstract": "It is well-known that classifiers are vulnerable to adversarial perturbations. To defend against adversarial perturbations, various certified robustness results have been derived. However, existing certified robustnesses are limited to top-1 predictions. In many real-world applications, top-$k$ predictions are more relevant. In this work, we aim to derive certified robustness for top-$k$ predictions. In particular, our certified robustness is based on randomized smoothing, which turns any classifier to a new classifier via adding noise to an input example. We adopt randomized smoothing because it is scalable to large-scale neural networks and applicable to any classifier. We derive a tight robustness in $\\ell_2$ norm for top-$k$ predictions when using randomized smoothing with Gaussian noise. We find that generalizing the certified robustness from top-1 to top-$k$ predictions faces significant technical challenges. We also empirically evaluate our method on CIFAR10 and ImageNet. For example, our method can obtain an ImageNet classifier with a certified top-5 accuracy of 62.8\\% when the $\\ell_2$-norms of the adversarial perturbations are less than 0.5 (=127/255). Our code is publicly available at: \\url{https://github.com/jjy1994/Certify_Topk}."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "ApproxHPVM: a portable compiler IR for accuracy-aware optimizations",
                    "abstract": "We propose ApproxHPVM, a compiler IR and system designed to enable accuracy-aware performance and energy tuning on heterogeneous systems with multiple compute units and approximation methods. ApproxHPVM automatically translates end-to-end application-level quality metrics into accuracy requirements for individual operations. ApproxHPVM uses a hardware-agnostic accuracy-tuning phase to do this translation that provides greater portability across heterogeneous hardware platforms and enables future capabilities like accuracy-aware dynamic scheduling and design space exploration. ApproxHPVM incorporates three main components: (a) a compiler IR with hardware-agnostic approximation metrics, (b) a hardware-agnostic accuracy-tuning phase to identify error-tolerant computations, and (c) an accuracy-aware hardware scheduler that maps error-tolerant computations to approximate hardware components. As ApproxHPVM does not incorporate any hardware-specific knowledge as part of the IR, it can serve as a portable virtual ISA that can be shipped to all kinds of hardware platforms. We evaluate our framework on nine benchmarks from the deep learning domain and five image processing benchmarks. Our results show that our framework can offload chunks of approximable computations to special-purpose accelerators that provide significant gains in performance and energy, while staying within user-specified application-level quality metrics with high probability. Across the 14 benchmarks, we observe from 1-9x performance speedups and 1.1-11.3x energy reduction for very small reductions in accuracy."
                },
                {
                    "title": "Branch and Bound for Piecewise Linear Neural Network Verification",
                    "abstract": "The success of Deep Learning and its potential use in many safety-critical applications has motivated research on formal verification of Neural Network (NN) models. In this context, verification involves proving or disproving that an NN model satisfies certain input-output properties. Despite the reputation of learned NN models as black boxes, and the theoretical hardness of proving useful properties about them, researchers have been successful in verifying some classes of models by exploiting their piecewise linear structure and taking insights from formal methods such as Satisifiability Modulo Theory. However, these methods are still far from scaling to realistic neural networks. To facilitate progress on this crucial area, we exploit the Mixed Integer Linear Programming (MIP) formulation of verification to propose a family of algorithms based on Branch-and-Bound (BaB). We show that our family contains previous verification methods as special cases. With the help of the BaB framework, we make three key contributions. Firstly, we identify new methods that combine the strengths of multiple existing approaches, accomplishing significant performance improvements over previous state of the art. Secondly, we introduce an effective branching strategy on ReLU non-linearities. This branching strategy allows us to efficiently and successfully deal with high input dimensional problems with convolutional network architecture, on which previous methods fail frequently. Finally, we propose comprehensive test data sets and benchmarks which includes a collection of previously released testcases. We use the data sets to conduct a thorough experimental comparison of existing and new algorithms and to provide an inclusive analysis of the factors impacting the hardness of verification problems."
                },
                {
                    "title": "Tight Certificates of Adversarial Robustness for Randomly Smoothed Classifiers",
                    "abstract": "Strong theoretical guarantees of robustness can be given for ensembles of classifiers generated by input randomization. Specifically, an $\\ell_2$ bounded adversary cannot alter the ensemble prediction generated by an additive isotropic Gaussian noise, where the radius for the adversary depends on both the variance of the distribution as well as the ensemble margin at the point of interest. We build on and considerably expand this work across broad classes of distributions. In particular, we offer adversarial robustness guarantees and associated algorithms for the discrete case where the adversary is $\\ell_0$ bounded. Moreover, we exemplify how the guarantees can be tightened with specific assumptions about the function class of the classifier such as a decision tree. We empirically illustrate these results with and without functional restrictions across image and molecule datasets."
                },
                {
                    "title": "Certified Adversarial Robustness via Randomized Smoothing",
                    "abstract": "We show how to turn any classifier that classifies well under Gaussian noise into a new classifier that is certifiably robust to adversarial perturbations under the $\\ell_2$ norm. This \"randomized smoothing\" technique has been proposed recently in the literature, but existing guarantees are loose. We prove a tight robustness guarantee in $\\ell_2$ norm for smoothing with Gaussian noise. We use randomized smoothing to obtain an ImageNet classifier with e.g. a certified top-1 accuracy of 49% under adversarial perturbations with $\\ell_2$ norm less than 0.5 (=127/255). No certified defense has been shown feasible on ImageNet except for smoothing. On smaller-scale datasets where competing approaches to certified $\\ell_2$ robustness are viable, smoothing delivers higher certified accuracies. Our strong empirical results suggest that randomized smoothing is a promising direction for future research into adversarially robust classification. Code and models are available at this http URL."
                },
                {
                    "title": "Deep Neural Network Compression for Aircraft Collision Avoidance Systems",
                    "abstract": "One approach to designing decision making logic for an aircraft collision avoidance system frames the problem as a Markov decision process and optimizes the system using dynamic programming. The resulting collision avoidance strategy can be represented as a numeric table. This methodology has been used in the development of the Airborne Collision Avoidance System X (ACAS X) family of collision avoidance systems for manned and unmanned aircraft, but the high dimensionality of the state space leads to very large tables. To improve storage efficiency, a deep neural network is used to approximate the table. With the use of an asymmetric loss function and a gradient descent algorithm, the parameters for this network can be trained to provide accurate estimates of table values while preserving the relative preferences of the possible advisories for each state. By training multiple networks to represent subtables, the network also decreases the required runtime for computing the collision avoidance advisory. Simulation studies show that the network improves the safety and efficiency of the collision avoidance system. Because only the network parameters need to be stored, the required storage space is reduced by a factor of 1000, enabling the collision avoidance system to operate using current avionics systems."
                },
                {
                    "title": "Continuous Reasoning: Scaling the impact of formal methods",
                    "abstract": "This paper describes work in continuous reasoning, where formal reasoning about a (changing) codebase is done in a fashion which mirrors the iterative, continuous model of software development that is increasingly practiced in industry. We suggest that advances in continuous reasoning will allow formal reasoning to scale to more programs, and more programmers. The paper describes the rationale for continuous reasoning, outlines some success cases from within industry, and proposes directions for work by the scientific community."
                },
                {
                    "title": "Learning to Optimize Tensor Programs",
                    "abstract": "We introduce a learning-based framework to optimize tensor programs for deep learning workloads. Efficient implementations of tensor operators, such as matrix multiplication and high dimensional convolution, are key enablers of effective deep learning systems. However, existing systems rely on manually optimized libraries such as cuDNN where only a narrow range of server class GPUs are well-supported. The reliance on hardware-specific operator libraries limits the applicability of high-level graph optimizations and incurs significant engineering costs when deploying to new hardware targets. We use learning to remove this engineering burden. We learn domain-specific statistical cost models to guide the search of tensor operator implementations over billions of possible program variants. We further accelerate the search by effective model transfer across workloads. Experimental results show that our framework delivers performance competitive with state-of-the-art hand-tuned libraries for low-power CPU, mobile GPU, and server-class GPU."
                },
                {
                    "title": "The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks",
                    "abstract": "Neural network pruning techniques can reduce the parameter counts of trained networks by over 90%, decreasing storage requirements and improving computational performance of inference without compromising accuracy. However, contemporary experience is that the sparse architectures produced by pruning are difficult to train from the start, which would similarly improve training performance. \nWe find that a standard pruning technique naturally uncovers subnetworks whose initializations made them capable of training effectively. Based on these results, we articulate the \"lottery ticket hypothesis:\" dense, randomly-initialized, feed-forward networks contain subnetworks (\"winning tickets\") that - when trained in isolation - reach test accuracy comparable to the original network in a similar number of iterations. The winning tickets we find have won the initialization lottery: their connections have initial weights that make training particularly effective. \nWe present an algorithm to identify winning tickets and a series of experiments that support the lottery ticket hypothesis and the importance of these fortuitous initializations. We consistently find winning tickets that are less than 10-20% of the size of several fully-connected and convolutional feed-forward architectures for MNIST and CIFAR10. Above this size, the winning tickets that we find learn faster than the original network and reach higher test accuracy."
                },
                {
                    "title": "Evaluating Robustness of Neural Networks with Mixed Integer Programming",
                    "abstract": "Neural networks have demonstrated considerable\nsuccess on a wide variety of real-world problems. However, neural\nnetworks can be fooled by adversarial examples \u00a0\u2013 slightly\nperturbed inputs that are misclassified with high confidence.\nVerification of networks enables us to gauge their vulnerability to\nsuch adversarial examples. We formulate verification of\npiecewise-linear neural networks as a mixed integer program. Our\nverifier finds minimum adversarial distortions two to three orders\nof magnitude more quickly than the state-of-the-art. We achieve\nthis via tight formulations for non-linearities, as well as a novel\npresolve algorithm that makes full use of all information\navailable. The computational speedup enables us to verify\nproperties on convolutional networks with an order of magnitude\nmore ReLUs than had been previously verified by any complete\nverifier, and we determine for the first time the exact adversarial\naccuracy of an MNIST classifier to perturbations with bounded\nl[infinity] norm e = 0:1. On this network, we find an adversarial\nexample for 4.38% of samples, and a certificate of robustness for\nthe remainder. Across a variety of robust training procedures, we\nare able to certify more samples than the state-of-the-art and find\nmore adversarial examples than a strong first-order attack for\nevery network."
                },
                {
                    "title": "Incremental Network Quantization: Towards Lossless CNNs with Low-Precision Weights",
                    "abstract": "This paper presents incremental network quantization (INQ), a novel method, targeting to efficiently convert any pre-trained full-precision convolutional neural network (CNN) model into a low-precision version whose weights are constrained to be either powers of two or zero. Unlike existing methods which are struggled in noticeable accuracy loss, our INQ has the potential to resolve this issue, as benefiting from two innovations. On one hand, we introduce three interdependent operations, namely weight partition, group-wise quantization and re-training. A well-proven measure is employed to divide the weights in each layer of a pre-trained CNN model into two disjoint groups. The weights in the first group are responsible to form a low-precision base, thus they are quantized by a variable-length encoding method. The weights in the other group are responsible to compensate for the accuracy loss from the quantization, thus they are the ones to be re-trained. On the other hand, these three operations are repeated on the latest re-trained group in an iterative manner until all the weights are converted into low-precision ones, acting as an incremental network quantization and accuracy enhancement procedure. Extensive experiments on the ImageNet classification task using almost all known deep CNN architectures including AlexNet, VGG-16, GoogleNet and ResNets well testify the efficacy of the proposed method. Specifically, at 5-bit quantization, our models have improved accuracy than the 32-bit floating-point references. Taking ResNet-18 as an example, we further show that our quantized models with 4-bit, 3-bit and 2-bit ternary weights have improved or very similar accuracy against its 32-bit floating-point baseline. Besides, impressive results with the combination of network pruning and INQ are also reported. The code is available at this https URL."
                },
                {
                    "title": "End to End Learning for Self-Driving Cars",
                    "abstract": "We trained a convolutional neural network (CNN) to map raw pixels from a single front-facing camera directly to steering commands. This end-to-end approach proved surprisingly powerful. With minimum training data from humans the system learns to drive in traffic on local roads with or without lane markings and on highways. It also operates in areas with unclear visual guidance such as in parking lots and on unpaved roads. \nThe system automatically learns internal representations of the necessary processing steps such as detecting useful road features with only the human steering angle as the training signal. We never explicitly trained it to detect, for example, the outline of roads. \nCompared to explicit decomposition of the problem, such as lane marking detection, path planning, and control, our end-to-end system optimizes all processing steps simultaneously. We argue that this will eventually lead to better performance and smaller systems. Better performance will result because the internal components self-optimize to maximize overall system performance, instead of optimizing human-selected intermediate criteria, e.g., lane detection. Such criteria understandably are selected for ease of human interpretation which doesn't automatically guarantee maximum system performance. Smaller networks are possible because the system learns to solve the problem with the minimal number of processing steps. \nWe used an NVIDIA DevBox and Torch 7 for training and an NVIDIA DRIVE(TM) PX self-driving car computer also running Torch 7 for determining where to drive. The system operates at 30 frames per second (FPS)."
                },
                {
                    "title": "Precision reuse for efficient regression verification",
                    "abstract": "Continuous testing during development is a well-established technique for software-quality assurance. Continuous model checking from revision to revision is not yet established as a standard practice, because the enormous resource consumption makes its application impractical. Model checkers compute a large number of verification facts that are necessary for verifying if a given specification holds. We have identified a category of such intermediate results that are easy to store and efficient to reuse: abstraction precisions. The precision of an abstract domain specifies the level of abstraction that the analysis works on. Precisions are thus a precious result of the verification effort and it is a waste of resources to throw them away after each verification run. In particular, precisions are reasonably small and thus easy to store; they are easy to process and have a large impact on resource consumption. We experimentally show the impact of precision reuse on industrial verification problems created from 62 Linux kernel device drivers with 1119 revisions."
                },
                {
                    "title": "An incremental verification framework for component-based software systems",
                    "abstract": "We present a tool-supported framework for the efficient reverification of component-based software systems after changes such as additions, removals or modifications of components. The incremental verification engine at the core of our INcremental VErification STrategy (INVEST) framework uses high-level algebraic representations of component-based systems to identify and execute the minimal set of component-wise reverification steps after a system change. The generality of the INVEST engine allows its integration with existing assume-guarantee verification paradigms. We illustrate this integration for an existing technique for the assume-guarantee verification of probabilistic systems. The resulting instance of the INVEST framework can reverify probabilistic safety properties of a cloud-deployed software system in a fraction of the time required by compositional assume-guarantee verification alone."
                },
                {
                    "title": "The cost of using exact confidence intervals for a binomial proportion",
                    "abstract": "When computing a confidence interval for a binomial proportion p one must choose between using an exact interval, which has a coverage probability of at least 1 a for all values of p, and a shorter approximate interval, which may have lower coverage for some p but that on average has coverage equal to 1 a. We investigate the cost of using the exact one and two-sided Clopper-Pearson confidence intervals rat her than shorter approximate intervals, first in terms of increased expected length and then in terms of the increase in sample size required to obtain a desired expected length. Using asymptotic expansions, we also give a closed-form formula for determining the sample size for the exact Clopper-Pearson methods. For two-sided intervals, our investigation reveals an interesting connection between the frequentist Clopper-Pearson interval and Bayesian intervals based on noninformative priors."
                },
                {
                    "title": "Green: reducing, reusing and recycling constraints in program analysis",
                    "abstract": "The analysis of constraints plays an important role in many aspects of software engineering, for example constraint satisfiability checking is central to symbolic execution. However, the norm is to recompute results in each analysis. We propose a different approach where every call to the solver is wrapped in a check to see if the result is not already available. While many tools use some form of results caching, the novelty of our approach is the persistence of results across runs, across programs being analyzed, across different analyses and even across physical location. Achieving such reuse requires that constraints be distilled into their essential parts and represented in a canonical form.\n In this paper, we describe the key ideas of our approach and its implementation, the Green solver interface, which reduces constraints to a simple form, allows for reuse of constraint solutions within an analysis run, and allows for recycling constraint solutions produced in one analysis run for use in other analysis runs. We describe how we integrated Green into two existing symbolic execution tools and demonstrate the reuse we achieve in the different settings."
                },
                {
                    "title": "Regression model checking",
                    "abstract": "Model checking is a promising technique for verifying program behavior and is increasingly finding usage in industry. To date, however, researchers have primarily considered model checking of single versions of programs. It is well understood that model checking can be very expensive for large, complex programs. Thus, simply reapplying model checking techniques on subsequent versions of programs as they evolve, in the limited time that is typically available for validating new releases, presents challenges. To address these challenges, we have developed a new technique for regression model checking (RMC), that applies model checking incrementally to new versions of systems. We report results of an empirical study examining the effectiveness of our technique; our results show that it is significantly faster than traditional model checking."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Approximate is Better than \u201cExact\u201d for Interval Estimation of Binomial Proportions",
                    "abstract": "Abstract For interval estimation of a proportion, coverage probabilities tend to be too large for \u201cexact\u201d confidence intervals based on inverting the binomial test and too small for the interval based on inverting the Wald large-sample normal test (i.e., sample proportion \u00b1 z-score \u00d7 estimated standard error). Wilson's suggestion of inverting the related score test with null rather than estimated standard error yields coverage probabilities close to nominal confidence levels, even for very small sample sizes. The 95% score interval has similar behavior as the adjusted Wald interval obtained after adding two \u201csuccesses\u201d and two \u201cfailures\u201d to the sample. In elementary courses, with the score and adjusted Wald methods it is unnecessary to provide students with awkward sample size guidelines."
                },
                {
                    "title": "Pruning algorithms-a survey",
                    "abstract": "A rule of thumb for obtaining good generalization in systems trained by examples is that one should use the smallest system that will fit the data. Unfortunately, it usually is not obvious what size is best; a system that is too small will not be able to learn the data while one that is just big enough may learn very slowly and be very sensitive to initial conditions and learning parameters. This paper is a survey of neural network pruning algorithms. The approach taken by the methods described here is to train a network that is larger than necessary and then remove the parts that are not needed."
                },
                {
                    "title": "Weight discretization paradigm for optical neural networks",
                    "abstract": "Neural networks are a primary candidate architecture for optical computing. One of the major problems in using neural networks for optical computers is that the information holders: the interconnection strengths (or weights) are normally real valued (continuous), whereas optics (light) is only capable of representing a few distinguishable intensity levels (discrete). In this paper a weight discretization paradigm is presented for back(ward error) propagation neural networks which can work with a very limited number of discretization levels. The number of interconnections in a (fully connected) neural network grows quadratically with the number of neurons of the network. Optics can handle a large number of interconnections because of the fact that light beams do not interfere with each other. A vast amount of light beams can therefore be used per unit of area. However the number of different values one can represent in a light beam is very limited. A flexible, portable (machine independent) neural network software package which is capable of weight discretization, is presented. The development of the software and some experiments have been done on personal computers. The major part of the testing, which requires a lot of computation, has been done using a CRAY X-MP/24 super computer."
                },
                {
                    "title": "Pruning versus clipping in neural networks.",
                    "abstract": "Dans un reseau neuronal, le nombre d'interconnexions est reduit par elimination des liaisons les plus faibles. Les performances sont ensuite ameliorees par application du theoreme d'apprentissage"
                },
                {
                    "title": "ApproxCaliper: A Programmable Framework for Application-aware Neural Network Optimization",
                    "abstract": "To deploy compute-intensive neural networks on resource-constrained edge systems, developers use model optimization techniques that reduce model size and computational cost. Existing optimization tools are application-agnostic \u2013 they optimize model parameters solely in view of the neural network accuracy \u2013 and can thus miss optimization opportunities. We propose ApproxCaliper , the first programmable framework for application-aware neural network optimization . By incorporating application-specific goals, ApproxCaliper facilitates more aggressive optimization of the neural networks compared to application-agnostic techniques. We perform experiments on five different neural networks used in two real-world robotics systems: a commercial agriculture robot and a simulation of an autonomous electric cart. Compared to Learning Rate Rewinding (LRR), a state-of-the-art structured pruning tool used in an application agnostic setting, ApproxCaliper achieves 5.3 \u00d7 higher speedup and 2.9 \u00d7 lower GPU resource utilization, and 36 \u00d7 and 6.1 \u00d7 additional model size reduction for the two evaluated benchmarks, respectively."
                },
                {
                    "title": "Beta-CROWN: Efficient Bound Propagation with Per-neuron Split Constraints for Complete and Incomplete Neural Network Verification",
                    "abstract": "Recent works in neural network veri\ufb01cation show that cheap incomplete veri\ufb01ers such as CROWN, based upon bound propagations, can effectively be used in Branch-and-Bound (BaB) methods to accelerate complete veri\ufb01cation, achieving signi\ufb01cant speedups compared to expensive linear programming (LP) based techniques. However, they cannot fully handle the per-neuron split constraints introduced by BaB like LP veri\ufb01ers do, leading to looser bounds and hurting their veri-\ufb01cation ef\ufb01ciency. In this work, we develop \u03b2 - CROWN, a new bound propagation based method that can fully encode per-neuron splits via opti-mizable parameters \u03b2 . When the optimizable parameters are jointly optimized in intermediate layers, \u03b2 -CROWN has the potential of producing better bounds than typical LP veri\ufb01ers with neuron split constraints, while being ef\ufb01ciently parallelizable on GPUs. Applied to the complete veri\ufb01cation setting, \u03b2 -CROWN is close to three orders of magnitude faster than LP-based BaB methods for robustness veri\ufb01cation, and also over twice faster than state-of-the-art GPU-based complete veri-\ufb01ers with similar timeout rates. By terminating BaB early, our method can also be used for incomplete veri\ufb01cation. Compared to the state-of-the-art semide\ufb01nite-programming (SDP) based veri-\ufb01er"
                },
                {
                    "title": "Online Verification of Deep Neural Networks under Domain or Weight Shift",
                    "abstract": "Although neural networks are widely used, it remains challenging to formally verify the safety and robustness of neural networks in real-world applications. Existing methods are designed to verify the network before use, which is limited to relatively simple speci\ufb01cations and \ufb01xed networks. These methods are not ready to be applied to real-world problems with complex and/or dynamically changing speci\ufb01cations and networks. To effectively handle dynamically changing speci\ufb01cations and networks, the veri\ufb01cation needs to be performed online when these changes take place. However, it is still challenging to run existing veri\ufb01cation algorithms online. Our key insight is that we can leverage the temporal dependencies of these changes to accelerate the veri\ufb01cation process, e.g., by warm starting new online veri\ufb01cation using previous veri\ufb01ed results. This paper establishes a novel framework for scalable online veri\ufb01cation to solve real-world veri\ufb01cation problems with dynamically changing speci\ufb01cations and/or networks, known as domain shift and weight shift respectively. We propose three types of techniques (branch management, perturbation tolerance analysis, and incremental computation) to accelerate the online veri\ufb01cation of deep neural networks. Experiment results show that our online veri\ufb01cation algorithm is up to two orders of magnitude faster than existing veri\ufb01cation algorithms, and thus can scale to real-world applications."
                },
                {
                    "title": "Hidden Cost of Randomized Smoothing",
                    "abstract": "The fragility of modern machine learning models has drawn a considerable amount of attention from both academia and the public. While immense interests were in either crafting adversarial attacks as a way to measure the robustness of neural networks or devising worst-case analytical robustness veri\ufb01cation with guarantees, few methods could enjoy both scalability and robustness guarantees at the same time. As an alternative to these attempts, randomized smoothing adopts a di\ufb00erent prediction rule that enables statistical robustness arguments which easily scale to large networks. However, in this paper, we point out the side e\ufb00ects of current randomized smoothing work\ufb02ows. Speci\ufb01cally, we articulate and prove two major points: 1) the decision boundaries of smoothed clas-si\ufb01ers will shrink, resulting in disparity in class-wise accuracy; 2) applying noise augmentation in the training process does not necessarily resolve the shrinking issue due to the inconsistent learning objectives."
                },
                {
                    "title": "TVM: An Automated End-to-End Optimizing Compiler for Deep Learning",
                    "abstract": "There is an increasing need to bring machine learning to a wide diversity of hardware devices. Current frameworks rely on vendor-speci\ufb01c operator libraries and optimize for a narrow range of server-class GPUs. Deploying workloads to new platforms such as mobile phones, embedded devices, and accelerators (e.g., FPGAs, ASICs) requires signi\ufb01cant manual effort. We propose TVM, a compiler that exposes graph-level and operator-level optimizations to provide performance portability to deep learning workloads across diverse hardware back-ends. TVM solves optimization challenges speci\ufb01c to deep learning such as high-level operator fusion, mapping to arbitrary hardware primitives, and memory latency hiding. TVM also offers automated optimization of low-level programs to hardware characteristics by employing a novel learning-based cost modeling method for rapid exploration of code optimizations. Experimental results demonstrate that TVM delivers performance across hardware back-ends that are competitive with state-of-the-art hand tuned libraries for low-power CPU, mobile GPU, and server-class GPUs. We also demonstrate TVM\u2019s ability to target new accelerator back-ends by targeting an FPGA-based generic deep learning accelerator. The system is open sourced and in production use inside several major companies."
                },
                {
                    "title": "Toward Continuous Verification of DNNs",
                    "abstract": "Deep neural network (DNN) verification is used to assess whether a DNN meets a desired trust-worthy property (e.g., robustness, fairness) on an infinite input set. While there have been significant advancements in scalable verifiers for individual DNNs, they suffer from inefficiency when developers update a deployed DNN (e.g. through quantization, pruning, or fine-tuning) to enhance its inference speed or accuracy. The verification is inefficient because the developers need to re-run the computationally expensive verifier from scratch on the updated DNN. To address this issue, we propose an incremental DNN verification framework, leveraging novel theory, data structures, and algorithms. Our previous works Ugare et al. (2023a) and Ugare et al. (2022) focused on incremental verification, and in this paper, we provide a summary of these advancements and explore the potential of incremental verification in enabling real-world DNN verification systems."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CR",
                "cs.PL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we leverage the information generated while certifying a given network to speed up the certification of similar networks using Incremental Randomized Smoothing (IRS)?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the significant challenge of computational expense in robustness certification of deep neural networks (DNNs). By improving the efficiency of the certification process, this research could lead to more practical applications of DNNs in real-world scenarios, particularly in resource-constrained environments like edge devices. The findings could influence future research by encouraging the development of more scalable certification methods, ultimately advancing knowledge in DNN robustness and enabling broader adoption of certified models in critical applications such as autonomous driving and healthcare.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent computational complexity of Randomized Smoothing (RS) certification, which requires extensive DNN inference on numerous corruptions per input. Naive approaches may fail because they do not account for the shared characteristics among similar networks, leading to redundant computations. Additionally, the need to maintain high confidence in certified accuracy while optimizing for speed introduces technical obstacles, such as ensuring that the incremental approach does not compromise the robustness guarantees provided by RS.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on standalone RS certification without considering the potential for incremental improvements across similar networks. Limitations in existing solutions include a lack of methodologies that utilize previously certified information to enhance efficiency. Barriers such as the absence of frameworks that systematically address the certification of approximated networks have prevented progress. Our approach differs by introducing the Incremental Randomized Smoothing (IRS) framework, which explicitly aims to reduce the computational burden by reusing certification data, thus improving upon prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the Incremental Randomized Smoothing (IRS) framework, which will utilize a predefined test set to improve the sample complexity of the certification process for similar networks. We will evaluate the performance of IRS using various DNN architectures and approximation techniques, measuring the speedup in certification time and the reduction in energy requirements and memory footprint. The expected outcomes include a significant decrease in the time required for certification while maintaining robust accuracy guarantees, thereby facilitating the deployment of certified DNNs in practical applications"
            }
        },
        "author_data": {
            "27d00bf6-ee64-4822-a329-02a5e70535d7": {
                "pk": "27d00bf6-ee64-4822-a329-02a5e70535d7",
                "name": "Shubham Ugare",
                "collaborators": [
                    "Gagandeep Singh",
                    "Sasa Misailovic",
                    "Tarun Suresh",
                    "Debangshu Banerjee",
                    "Jacob S. Laurel",
                    "Rem Yang",
                    "Hangoo Kang",
                    "Rohan Gumaste",
                    "Robert Nagel",
                    "Atharva Sehgal"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Network Verification",
                    "Probabilistic Programming",
                    "Code Generation"
                ],
                "publications": [
                    {
                        "title": "Is Watermarking LLM-Generated Code Robust?",
                        "abstract": "We present the first study of the robustness of existing watermarking techniques on Python code generated by large language models. Although existing works showed that watermarking can be robust for natural language, we show that it is easy to remove these watermarks on code by semantic-preserving transformations."
                    },
                    {
                        "title": "SynCode: LLM Generation with Grammar Augmentation",
                        "abstract": "LLMs are widely used in complex AI applications. These applications underscore the need for LLM outputs to adhere to a specific format, for their integration with other components in the systems. Typically the format rules e.g., for data serialization formats such as JSON, YAML, or Code in Programming Language are expressed as context-free grammar (CFG). Due to the hallucinations and unreliability of LLMs, instructing LLMs to adhere to specified syntax becomes an increasingly important challenge. We present SynCode, a novel framework for efficient and general syntactical decoding with LLMs, to address this challenge. SynCode ensures soundness and completeness with respect to the CFG of a formal language, effectively retaining valid tokens while filtering out invalid ones. SynCode uses an offline-constructed, efficient lookup table, the DFA mask store, derived from the DFA of the language's grammar for efficient generation. SynCode seamlessly integrates with any language defined by CFG, as evidenced by experiments focusing on generating JSON, Python, and Go outputs. Our experiments evaluating the effectiveness of SynCode for JSON generation demonstrate that SynCode eliminates all syntax errors and significantly outperforms state-of-the-art baselines. Furthermore, our results underscore how SynCode significantly reduces 96.07% of syntax errors in generated Python and Go code, showcasing its substantial impact on enhancing syntactical precision in LLM generation. Our code is available at https://github.com/uiuc-focal-lab/syncode"
                    },
                    {
                        "title": "IterGen: Iterative Structured LLM Generation",
                        "abstract": "Large Language Models (LLMs) are widely used for tasks such as natural language and code generation. Still, their outputs often suffer from issues like privacy violations, and semantically inaccurate code generation. Current libraries for LLM generation rely on left-to-right decoding without systematic support for backtracking, limiting the ability to correct or refine outputs mid-generation. To address this issue, we introduce IterGen, an intuitive framework for iterative, grammar-guided LLM generation that enables users to move both forward and backward within the generated output based on grammar symbols. By leveraging a symbol-to-position mapping, IterGen ensures efficient and structured generation while allowing for corrections during the process. We demonstrate IterGen's effectiveness in two important applications: reducing privacy leakage in LLM outputs and improving the accuracy of LLM-generated SQL queries. Our code is available at https://github.com/uiuc-arc/itergen"
                    },
                    {
                        "title": "Incremental Verification of Neural Networks",
                        "abstract": "Complete verification of deep neural networks (DNNs) can exactly determine whether the DNN satisfies a desired trustworthy property (e.g., robustness, fairness) on an infinite set of inputs or not. Despite the tremendous progress to improve the scalability of complete verifiers over the years on individual DNNs, they are inherently inefficient when a deployed DNN is updated to improve its inference speed or accuracy. The inefficiency is because the expensive verifier needs to be run from scratch on the updated DNN. To improve efficiency, we propose a new, general framework for incremental and complete DNN verification based on the design of novel theory, data structure, and algorithms. Our contributions implemented in a tool named IVAN yield an overall geometric mean speedup of 2.4x for verifying challenging MNIST and CIFAR10 classifiers and a geometric mean speedup of 3.8x for the ACAS-XU classifiers over the state-of-the-art baselines."
                    },
                    {
                        "title": "A general construction for abstract interpretation of higher-order automatic differentiation",
                        "abstract": "We present a novel, general construction to abstractly interpret higher-order automatic differentiation (AD). Our construction allows one to instantiate an abstract interpreter for computing derivatives up to a chosen order. Furthermore, since our construction reduces the problem of abstractly reasoning about derivatives to abstractly reasoning about real-valued straight-line programs, it can be instantiated with almost any numerical abstract domain, both relational and non-relational. We formally establish the soundness of this construction. We implement our technique by instantiating our construction with both the non-relational interval domain and the relational zonotope domain to compute both first and higher-order derivatives. In the latter case, we are the first to apply a relational domain to automatic differentiation for abstracting higher-order derivatives, and hence we are also the first abstract interpretation work to track correlations across not only different variables, but different orders of derivatives. We evaluate these instantiations on multiple case studies, namely robustly explaining a neural network and more precisely computing a neural network\u2019s Lipschitz constant. For robust interpretation, first and second derivatives computed via zonotope AD are up to 4.76\u00d7 and 6.98\u00d7 more precise, respectively, compared to interval AD. For Lipschitz certification, we obtain bounds that are up to 11,850\u00d7 more precise with zonotopes, compared to the state-of-the-art interval-based tool."
                    },
                    {
                        "title": "Proof transfer for fast certification of multiple approximate neural networks",
                        "abstract": "Developers of machine learning applications often apply post-training neural network optimizations, such as quantization and pruning, that approximate a neural network to speed up inference and reduce energy consumption, while maintaining high accuracy and robustness. Despite a recent surge in techniques for the robustness verification of neural networks, a major limitation of almost all state-of-the-art approaches is that the verification needs to be run from scratch every time the network is even slightly modified. Running precise end-to-end verification from scratch for every new network is expensive and impractical in many scenarios that use or compare multiple approximate network versions, and the robustness of all the networks needs to be verified efficiently. We present FANC, the first general technique for transferring proofs between a given network and its multiple approximate versions without compromising verifier precision. To reuse the proofs obtained when verifying the original network, FANC generates a set of templates \u2013 connected symbolic shapes at intermediate layers of the original network \u2013 that capture the proof of the property to be verified. We present novel algorithms for generating and transforming templates that generalize to a broad range of approximate networks and reduce the verification cost. We present a comprehensive evaluation demonstrating the effectiveness of our approach. We consider a diverse set of networks obtained by applying popular approximation techniques such as quantization and pruning on fully-connected and convolutional architectures and verify their robustness against different adversarial attacks such as adversarial patches, L0, rotation and brightening. Our results indicate that FANC can significantly speed up verification with state-of-the-art verifier, DeepZ by up to 4.1x."
                    },
                    {
                        "title": "Statheros: Compiler for Efficient Low-Precision Probabilistic Programming",
                        "abstract": "As Edge and IoT computing devices process noisy data or make decisions in uncertain environments, they require frameworks for inexpensive, yet accurate probabilistic inference. Probabilistic programming has emerged as a powerful way for developers to write high-level programs, while abstracting away the implementation details of inference. However, the existing algorithms are slow and often assumed to require precise calculations. We present Statheros, the first compiler for low-level, fixed-point approximation of probabilistic programming. Statheros compiles programs to fixed-point inference procedures and is able to determine the optimal fixed-point type to use. We evaluate Statheros on 13 benchmarks and three embedded platforms. The results show that Statheros-generated code is 11. 5x (Arduino), 3. 8x (PocketBeagle), and 2. 2x (Raspberry Pi) faster than single-precision floating-point computation, with minimal accuracy loss."
                    },
                    {
                        "title": "Toward Continuous Verification of DNNs",
                        "abstract": "Deep neural network (DNN) verification is used to assess whether a DNN meets a desired trust-worthy property (e.g., robustness, fairness) on an infinite input set. While there have been significant advancements in scalable verifiers for individual DNNs, they suffer from inefficiency when developers update a deployed DNN (e.g. through quantization, pruning, or fine-tuning) to enhance its inference speed or accuracy. The verification is inefficient because the developers need to re-run the computationally expensive verifier from scratch on the updated DNN. To address this issue, we propose an incremental DNN verification framework, leveraging novel theory, data structures, and algorithms. Our previous works Ugare et al. (2023a) and Ugare et al. (2022) focused on incremental verification, and in this paper, we provide a summary of these advancements and explore the potential of incremental verification in enabling real-world DNN verification systems."
                    }
                ]
            },
            "6f89b4ba-3f13-4a27-88b3-549a8904e02e": {
                "pk": "6f89b4ba-3f13-4a27-88b3-549a8904e02e",
                "name": "Tarun Suresh",
                "collaborators": [
                    "Gagandeep Singh",
                    "Shubham Ugare",
                    "Sasa Misailovic",
                    "Rohan Gumaste",
                    "Hangoo Kang",
                    "Yinglun Xu",
                    "David Zhu",
                    "Ruirui Li",
                    "Zhengyang Wang",
                    "Haoming Jiang"
                ],
                "domain": [
                    "Large Language Models",
                    "Code Generation",
                    "Reinforcement Learning",
                    "DNN Verification"
                ],
                "publications": [
                    {
                        "title": "Is Watermarking LLM-Generated Code Robust?",
                        "abstract": "We present the first study of the robustness of existing watermarking techniques on Python code generated by large language models. Although existing works showed that watermarking can be robust for natural language, we show that it is easy to remove these watermarks on code by semantic-preserving transformations."
                    },
                    {
                        "title": "SynCode: LLM Generation with Grammar Augmentation",
                        "abstract": "LLMs are widely used in complex AI applications. These applications underscore the need for LLM outputs to adhere to a specific format, for their integration with other components in the systems. Typically the format rules e.g., for data serialization formats such as JSON, YAML, or Code in Programming Language are expressed as context-free grammar (CFG). Due to the hallucinations and unreliability of LLMs, instructing LLMs to adhere to specified syntax becomes an increasingly important challenge. We present SynCode, a novel framework for efficient and general syntactical decoding with LLMs, to address this challenge. SynCode ensures soundness and completeness with respect to the CFG of a formal language, effectively retaining valid tokens while filtering out invalid ones. SynCode uses an offline-constructed, efficient lookup table, the DFA mask store, derived from the DFA of the language's grammar for efficient generation. SynCode seamlessly integrates with any language defined by CFG, as evidenced by experiments focusing on generating JSON, Python, and Go outputs. Our experiments evaluating the effectiveness of SynCode for JSON generation demonstrate that SynCode eliminates all syntax errors and significantly outperforms state-of-the-art baselines. Furthermore, our results underscore how SynCode significantly reduces 96.07% of syntax errors in generated Python and Go code, showcasing its substantial impact on enhancing syntactical precision in LLM generation. Our code is available at https://github.com/uiuc-focal-lab/syncode"
                    },
                    {
                        "title": "IterGen: Iterative Structured LLM Generation",
                        "abstract": "Large Language Models (LLMs) are widely used for tasks such as natural language and code generation. Still, their outputs often suffer from issues like privacy violations, and semantically inaccurate code generation. Current libraries for LLM generation rely on left-to-right decoding without systematic support for backtracking, limiting the ability to correct or refine outputs mid-generation. To address this issue, we introduce IterGen, an intuitive framework for iterative, grammar-guided LLM generation that enables users to move both forward and backward within the generated output based on grammar symbols. By leveraging a symbol-to-position mapping, IterGen ensures efficient and structured generation while allowing for corrections during the process. We demonstrate IterGen's effectiveness in two important applications: reducing privacy leakage in LLM outputs and improving the accuracy of LLM-generated SQL queries. Our code is available at https://github.com/uiuc-arc/itergen"
                    },
                    {
                        "title": "Two-Step Offline Preference-Based Reinforcement Learning with Constrained Actions",
                        "abstract": "Preference-based reinforcement learning (PBRL) in the offline setting has succeeded greatly in industrial applications such as chatbots. A two-step learning framework where one applies a reinforcement learning step after a reward modeling step has been widely adopted for the problem. However, such a method faces challenges from the risk of reward hacking and the complexity of reinforcement learning. To overcome the challenge, our insight is that both challenges come from the state-actions not supported in the dataset. Such state-actions are unreliable and increase the complexity of the reinforcement learning problem at the second step. Based on the insight, we develop a novel two-step learning method called PRC: preference-based reinforcement learning with constrained actions. The high-level idea is to limit the reinforcement learning agent to optimize over a constrained action space that excludes the out-of-distribution state-actions. We empirically verify that our method has high learning efficiency on various datasets in robotic control environments."
                    },
                    {
                        "title": "Toward Continuous Verification of DNNs",
                        "abstract": "Deep neural network (DNN) verification is used to assess whether a DNN meets a desired trust-worthy property (e.g., robustness, fairness) on an infinite input set. While there have been significant advancements in scalable verifiers for individual DNNs, they suffer from inefficiency when developers update a deployed DNN (e.g. through quantization, pruning, or fine-tuning) to enhance its inference speed or accuracy. The verification is inefficient because the developers need to re-run the computationally expensive verifier from scratch on the updated DNN. To address this issue, we propose an incremental DNN verification framework, leveraging novel theory, data structures, and algorithms. Our previous works Ugare et al. (2023a) and Ugare et al. (2022) focused on incremental verification, and in this paper, we provide a summary of these advancements and explore the potential of incremental verification in enabling real-world DNN verification systems."
                    }
                ]
            },
            "af28b79f-f66a-4080-a695-9b2716256415": {
                "pk": "af28b79f-f66a-4080-a695-9b2716256415",
                "name": "Debangshu Banerjee",
                "collaborators": [
                    "Gagandeep Singh",
                    "Aditya Gopalan",
                    "Nilotpal Biswas",
                    "S. Bhattacharya",
                    "Changming Xu",
                    "Avaljot Singh",
                    "Shubham Ugare",
                    "Sasa Misailovic",
                    "Kavita Wagh",
                    "Avishek Ghosh"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Network Verification",
                    "Reinforcement Learning",
                    "Adversarial Robustness"
                ],
                "publications": [
                    {
                        "title": "Relational DNN Verification With Cross Executional Bound Refinement",
                        "abstract": "We focus on verifying relational properties defined over deep neural networks (DNNs) such as robustness against universal adversarial perturbations (UAP), certified worst-case hamming distance for binary string classifications, etc. Precise verification of these properties requires reasoning about multiple executions of the same DNN. However, most of the existing works in DNN verification only handle properties defined over single executions and as a result, are imprecise for relational properties. Though few recent works for relational DNN verification, capture linear dependencies between the inputs of multiple executions, they do not leverage dependencies between the outputs of hidden layers producing imprecise results. We develop a scalable relational verifier RACoon that utilizes cross-execution dependencies at all layers of the DNN gaining substantial precision over SOTA baselines on a wide range of datasets, networks, and relational properties."
                    },
                    {
                        "title": "Input-Relational Verification of Deep Neural Networks",
                        "abstract": "We consider the verification of input-relational properties defined over deep neural networks (DNNs) such as robustness against universal adversarial perturbations, monotonicity, etc. Precise verification of these properties requires reasoning about multiple executions of the same DNN. We introduce a novel concept of difference tracking to compute the difference between the outputs of two executions of the same DNN at all layers. We design a new abstract domain, DiffPoly for efficient difference tracking that can scale large DNNs. DiffPoly is equipped with custom abstract transformers for common activation functions (ReLU, Tanh, Sigmoid, etc.) and affine layers and can create precise linear cross-execution constraints. We implement an input-relational verifier for DNNs called RaVeN which uses DiffPoly and linear program formulations to handle a wide range of input-relational properties. Our experimental results on challenging benchmarks show that by leveraging precise linear constraints defined over multiple executions of the DNN, RaVeN gains substantial precision over baselines on a wide range of datasets, networks, and input-relational properties."
                    },
                    {
                        "title": "Markov Chain Concentration with an Application in Reinforcement Learning",
                        "abstract": "Given $X_1,\\cdot ,X_N$ random variables whose joint distribution is given as $\\mu$ we will use the Martingale Method to show any Lipshitz Function $f$ over these random variables is subgaussian. The Variance parameter however can have a simple expression under certain conditions. For example under the assumption that the random variables follow a Markov Chain and that the function is Lipschitz under a Weighted Hamming Metric. We shall conclude with certain well known techniques from concentration of suprema of random processes with applications in Reinforcement Learning"
                    },
                    {
                        "title": "Interpreting Robustness Proofs of Deep Neural Networks",
                        "abstract": "In recent years numerous methods have been developed to formally verify the robustness of deep neural networks (DNNs). Though the proposed techniques are effective in providing mathematical guarantees about the DNNs behavior, it is not clear whether the proofs generated by these methods are human-interpretable. In this paper, we bridge this gap by developing new concepts, algorithms, and representations to generate human understandable interpretations of the proofs. Leveraging the proposed method, we show that the robustness proofs of standard DNNs rely on spurious input features, while the proofs of DNNs trained to be provably robust filter out even the semantically meaningful features. The proofs for the DNNs combining adversarial and provably robust training are the most effective at selectively filtering out spurious features as well as relying on human-understandable input features."
                    },
                    {
                        "title": "On the Minimax Regret for Linear Bandits in a wide variety of Action Spaces",
                        "abstract": "As noted in the works of Lattimore and Szepesv\u00b4ari [2020], it has been mentioned that it is an open problem to characterize the minimax regret of linear bandits in a wide variety of action spaces. In this article we present an optimal regret lower bound for a wide class of convex action spaces."
                    },
                    {
                        "title": "Incremental Verification of Neural Networks",
                        "abstract": "Complete verification of deep neural networks (DNNs) can exactly determine whether the DNN satisfies a desired trustworthy property (e.g., robustness, fairness) on an infinite set of inputs or not. Despite the tremendous progress to improve the scalability of complete verifiers over the years on individual DNNs, they are inherently inefficient when a deployed DNN is updated to improve its inference speed or accuracy. The inefficiency is because the expensive verifier needs to be run from scratch on the updated DNN. To improve efficiency, we propose a new, general framework for incremental and complete DNN verification based on the design of novel theory, data structure, and algorithms. Our contributions implemented in a tool named IVAN yield an overall geometric mean speedup of 2.4x for verifying challenging MNIST and CIFAR10 classifiers and a geometric mean speedup of 3.8x for the ACAS-XU classifiers over the state-of-the-art baselines."
                    },
                    {
                        "title": "Bad Values but Good Behavior: Learning Highly Misspecified Bandits and MDPs",
                        "abstract": "Parametric, feature-based reward models are employed by a variety of algorithms in decision-making settings such as bandits and Markov decision processes (MDPs). The typical assumption under which the algorithms are analysed is realizability, i.e., that the true values of actions are perfectly explained by some parametric model in the class. We are, however, interested in the situation where the true values are (significantly) misspecified with respect to the model class. For parameterized bandits, contextual bandits and MDPs, we identify structural conditions, depending on the problem instance and model class, under which basic algorithms such as $\\epsilon$-greedy, LinUCB and fitted Q-learning provably learn optimal policies under even highly misspecified models. This is in contrast to existing worst-case results for, say misspecified bandits, which show regret bounds that scale linearly with time, and shows that there can be a nontrivially large set of bandit instances that are robust to misspecification."
                    },
                    {
                        "title": "Critic Algorithms using Cooperative Networks",
                        "abstract": "An algorithm is proposed for policy evaluation in Markov Decision Processes which gives good empirical results with respect to convergence rates. The algorithm tracks the Projected Bellman Error and is implemented as a true gradient based algorithm. In this respect this algorithm differs from TD($\\lambda$) class of algorithms. This algorithm tracks the Projected Bellman Algorithm and is therefore different from the class of residual algorithms. Further the convergence of this algorithm is empirically much faster than GTD2 class of algorithms which aim at tracking the Projected Bellman Error. We implemented proposed algorithm in DQN and DDPG framework and found that our algorithm achieves comparable results in both of these experiments"
                    },
                    {
                        "title": "Exploration in Linear Bandits with Rich Action Sets and its Implications for Inference",
                        "abstract": "We present a non-asymptotic lower bound on the eigenspectrum of the design matrix generated by any linear bandit algorithm with sub-linear regret when the action set has well-behaved curvature. Specifically, we show that the minimum eigenvalue of the expected design matrix grows as $\\Omega(\\sqrt{n})$ whenever the expected cumulative regret of the algorithm is $O(\\sqrt{n})$, where $n$ is the learning horizon, and the action-space has a constant Hessian around the optimal arm. This shows that such action-spaces force a polynomial lower bound rather than a logarithmic lower bound, as shown by \\cite{lattimore2017end}, in discrete (i.e., well-separated) action spaces. Furthermore, while the previous result is shown to hold only in the asymptotic regime (as $n \\to \\infty$), our result for these\"locally rich\"action spaces is any-time. Additionally, under a mild technical assumption, we obtain a similar lower bound on the minimum eigen value holding with high probability. We apply our result to two practical scenarios -- \\emph{model selection} and \\emph{clustering} in linear bandits. For model selection, we show that an epoch-based linear bandit algorithm adapts to the true model complexity at a rate exponential in the number of epochs, by virtue of our novel spectral bound. For clustering, we consider a multi agent framework where we show, by leveraging the spectral result, that no forced exploration is necessary -- the agents can run a linear bandit algorithm and estimate their underlying parameters at once, and hence incur a low regret."
                    },
                    {
                        "title": "Minimising the duration of a system-controlled virtual reality tour",
                        "abstract": "The duration of a virtual reality tour can be a crucial factor in affecting user experience. Lengthy exposure to a virtual environment can cause severe health issues similar to motion sickness, and wearing the Head-Mounted Display for a long duration can make the user fatigued. Minimising the duration of VR exposure can always help in addressing these issues. In this work, we have proposed an approach to minimise the duration of a system-controlled virtual tour by optimising the path connecting all the sites of the virtual environment. To optimise the duration, we theoretically compute the optimal time and the path to cover all the places of a virtual environment by reducing the problem to the famous Vehicle Routing Problem (VRP). Following our approach, we have also created a VR tour for the largest river island in the world, Majuli (Assam, India)."
                    },
                    {
                        "title": "Natural walking speed prediction in Virtual Reality while using target selection-based locomotion",
                        "abstract": "Travelling speed plays an essential role in the overall user experience while navigating inside a virtual environment. Researchers have used various travelling speed that matches the user speed profile in order to give a natural walking experience. However, predicting a user\u2019s instantaneous walking speed can be challenging when there is no continuous input from the user. Target selection-based techniques are those where the user selects the target to reach there automatically. These techniques also lack naturalness due to their low interaction fidelity. In this work, we have proposed a mathematical model that can dynamically compute the instantaneous natural walking speed while moving from one point to another in a virtual environment. We formulated our model with the help of user studies."
                    },
                    {
                        "title": "Vertex guarding for dynamic orthogonal art galleries",
                        "abstract": "We devise an algorithm for surveying a dynamic orthogonal polygonal domain by placing one guard at each vertex in a subset of its vertices, i.e., whenever an orthogonal polygonal domain [Formula: see text] is modified to result in another orthogonal polygonal domain [Formula: see text], our algorithm updates the set of vertex guards surveying [Formula: see text] so that the updated guard set surveys [Formula: see text]. Our algorithm modifies the guard placement in [Formula: see text] amortized time, while ensuring the updated orthogonal polygonal domain with [Formula: see text] holes and [Formula: see text] vertices is guarded using at most [Formula: see text] vertex guards. For the special case of the initial orthogonal polygon being hole-free and each update resulting in a hole-free orthogonal polygon, our guard update algorithm takes [Formula: see text] worst-case time. Here, [Formula: see text] and [Formula: see text] are the number of vertices of the orthogonal polygon before and after the update, respectively; and, [Formula: see text] is the sum of [Formula: see text] and the number of updates to a few structures maintained by our algorithm. Further, by giving a construction, we show it suffices for the algorithm to consider only the case in which the parity of the number of reflex vertices of both [Formula: see text] and [Formula: see text] are equal."
                    },
                    {
                        "title": "HEX and Neurodynamic Programming",
                        "abstract": "Hex is a complex game with a high branching factor. For the first time Hex is being attempted to be solved without the use of game tree structures and associated methods of pruning. We also are abstaining from any heuristic information about Virtual Connections or Semi Virtual Connections which were previously used in all previous known computer versions of the game. The H-search algorithm which was the basis of finding such connections and had been used with success in previous Hex playing agents has been forgone. Instead what we use is reinforcement learning through self play and approximations through neural networks to by pass the problem of high branching factor and maintaining large tables for state-action evaluations. Our code is based primarily on NeuroHex. The inspiration is drawn from the recent success of AlphaGo Zero."
                    }
                ]
            },
            "79d3f1ae-c8c4-4475-a331-6e89b0ac0bfc": {
                "pk": "79d3f1ae-c8c4-4475-a331-6e89b0ac0bfc",
                "name": "Gagandeep Singh",
                "collaborators": [
                    "Martin T. Vechev",
                    "Markus P\u00fcschel",
                    "Jacob S. Laurel",
                    "Rem Yang",
                    "Sasa Misailovic",
                    "Christoph M\u00fcller",
                    "O. Mutlu",
                    "Shubham Ugare",
                    "Anian Ruoss",
                    "Mislav Balunovic"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Federated Learning",
                    "Formal Verification"
                ],
                "publications": [
                    {
                        "title": "Evaluating Machine LearningWorkloads on Memory-Centric Computing Systems",
                        "abstract": "Training machine learning (ML) algorithms is a computationally intensive process, which is frequently memory-bound due to repeatedly accessing large training datasets. As a result, processor-centric systems (CPU, GPU) waste large amounts of energy and execution cycles due to the data movement between memory units and processing units. Memory-centric computing systems, i.e., systems with processing-in-memory (PIM) capabilities, can alleviate this data movement bottleneck. Our goal is to understand the potential of general-purpose PIM architectures to accelerate ML training. To do so, we (1) implement several classic ML algorithms (namely, linear regression, logistic regression, decision tree, K-Means clustering) on a real-world generalpurpose PIM architecture, (2) evaluate and characterize them in terms of accuracy, performance and scaling, and (3) compare to their counterpart state-of-the-art implementations on CPU and GPU. Our evaluation on a real memory-centric computing system with more than 2500 PIM cores shows that PIM greatly accelerates memorybound ML workloads, when the necessary operations and datatypes are natively supported by PIM hardware. For example, our PIM implementation of decision tree is 27\u00d7 faster than the CPU implementation on an 8-core Intel Xeon, and 1.34\u00d7 faster than the GPU implementation on an NVIDIA A100. Our PIM implementation of K-Means clustering is 2.8\u00d7 and 3.2\u00d7 faster than CPU and GPU implementations, respectively. We provide several key observations, takeaways, and recommendations for users of ML workloads, programmers of PIM architectures, and hardware designers and architects of future memory-centric computing systems. We open-source all our code and datasets at https://github.com/CMU-SA FARI/pim-ml."
                    },
                    {
                        "title": "FedCompass: Efficient Cross-Silo Federated Learning on Heterogeneous Client Devices using a Computing Power Aware Scheduler",
                        "abstract": "Cross-silo federated learning offers a promising solution to collaboratively train robust and generalized AI models without compromising the privacy of local datasets, e.g., healthcare, financial, as well as scientific projects that lack a centralized data facility. Nonetheless, because of the disparity of computing resources among different clients (i.e., device heterogeneity), synchronous federated learning algorithms suffer from degraded efficiency when waiting for straggler clients. Similarly, asynchronous federated learning algorithms experience degradation in the convergence rate and final model accuracy on non-identically and independently distributed (non-IID) heterogeneous datasets due to stale local models and client drift. To address these limitations in cross-silo federated learning with heterogeneous clients and data, we propose FedCompass, an innovative semi-asynchronous federated learning algorithm with a computing power-aware scheduler on the server side, which adaptively assigns varying amounts of training tasks to different clients using the knowledge of the computing power of individual clients. FedCompass ensures that multiple locally trained models from clients are received almost simultaneously as a group for aggregation, effectively reducing the staleness of local models. At the same time, the overall training process remains asynchronous, eliminating prolonged waiting periods from straggler clients. Using diverse non-IID heterogeneous distributed datasets, we demonstrate that FedCompass achieves faster convergence and higher accuracy than other asynchronous algorithms while remaining more efficient than synchronous algorithms when performing federated learning on heterogeneous clients. The source code for FedCompass is available at https://github.com/APPFL/FedCompass."
                    },
                    {
                        "title": "RawHash: enabling fast and accurate real-time analysis of raw nanopore signals for large genomes",
                        "abstract": "Nanopore sequencers generate electrical raw signals in real-time while sequencing long genomic strands. These raw signals can be analyzed as they are generated, providing an opportunity for real-time genome analysis. An important feature of nanopore sequencing, Read Until, can eject strands from sequencers without fully sequencing them, which provides opportunities to computationally reduce the sequencing time and cost. However, existing works utilizing Read Until either 1) require powerful computational resources that may not be available for portable sequencers or 2) lack scalability for large genomes, rendering them inaccurate or ineffective. We propose RawHash, the first mechanism that can accurately and efficiently perform real-time analysis of nanopore raw signals for large genomes using a hash-based similarity search. To enable this, RawHash ensures the signals corresponding to the same DNA content lead to the same hash value, regardless of the slight variations in these signals. RawHash achieves an accurate hash-based similarity search via an effective quantization of the raw signals such that signals corresponding to the same DNA content have the same quantized value and, subsequently, the same hash value. We evaluate RawHash on three applications: 1) read mapping, 2) relative abundance estimation, and 3) contamination analysis. Our evaluations show that RawHash is the only tool that can provide high accuracy and high throughput for analyzing large genomes in real-time. When compared to the state-of-the-art techniques, UNCALLED and Sigmap, RawHash provides 1) 25.8\u00d7 and 3.4\u00d7 better average throughput and 2) significantly better accuracy for large genomes, respectively. Source code is available at https://github.com/CMU-SAFARI/RawHash."
                    },
                    {
                        "title": "Scalable verification of GNN-based job schedulers",
                        "abstract": "Recently, Graph Neural Networks (GNNs) have been applied for scheduling jobs over clusters, achieving better performance than hand-crafted heuristics. Despite their impressive performance, concerns remain over whether these GNN-based job schedulers meet users\u2019 expectations about other important properties, such as strategy-proofness, sharing incentive, and stability. In this work, we consider formal verification of GNN-based job schedulers. We address several domain-specific challenges such as networks that are deeper and specifications that are richer than those encountered when verifying image and NLP classifiers. We develop vegas, the first general framework for verifying both single-step and multi-step properties of these schedulers based on carefully designed algorithms that combine abstractions, refinements, solvers, and proof transfer. Our experimental results show that vegas achieves significant speed-up when verifying important properties of a state-of-the-art GNN-based scheduler compared to previous methods."
                    },
                    {
                        "title": "A dual number abstraction for static analysis of Clarke Jacobians",
                        "abstract": "We present a novel abstraction for bounding the Clarke Jacobian of a Lipschitz continuous, but not necessarily differentiable function over a local input region. To do so, we leverage a novel abstract domain built upon dual numbers, adapted to soundly over-approximate all first derivatives needed to compute the Clarke Jacobian. We formally prove that our novel forward-mode dual interval evaluation produces a sound, interval domain-based over-approximation of the true Clarke Jacobian for a given input region. Due to the generality of our formalism, we can compute and analyze interval Clarke Jacobians for a broader class of functions than previous works supported \u2013 specifically, arbitrary compositions of neural networks with Lipschitz, but non-differentiable perturbations. We implement our technique in a tool called DeepJ and evaluate it on multiple deep neural networks and non-differentiable input perturbations to showcase both the generality and scalability of our analysis. Concretely, we can obtain interval Clarke Jacobians to analyze Lipschitz robustness and local optimization landscapes of both fully-connected and convolutional neural networks for rotational, contrast variation, and haze perturbations, as well as their compositions."
                    },
                    {
                        "title": "Provable Defense Against Geometric Transformations",
                        "abstract": "Geometric image transformations that arise in the real world, such as scaling and rotation, have been shown to easily deceive deep neural networks (DNNs). Hence, training DNNs to be certifiably robust to these perturbations is critical. However, no prior work has been able to incorporate the objective of deterministic certified robustness against geometric transformations into the training procedure, as existing verifiers are exceedingly slow. To address these challenges, we propose the first provable defense for deterministic certified geometric robustness. Our framework leverages a novel GPU-optimized verifier that can certify images between 60$\\times$ to 42,600$\\times$ faster than existing geometric robustness verifiers, and thus unlike existing works, is fast enough for use in training. Across multiple datasets, our results show that networks trained via our framework consistently achieve state-of-the-art deterministic certified geometric robustness and clean accuracy. Furthermore, for the first time, we verify the geometric robustness of a neural network for the challenging, real-world setting of autonomous driving."
                    },
                    {
                        "title": "A general construction for abstract interpretation of higher-order automatic differentiation",
                        "abstract": "We present a novel, general construction to abstractly interpret higher-order automatic differentiation (AD). Our construction allows one to instantiate an abstract interpreter for computing derivatives up to a chosen order. Furthermore, since our construction reduces the problem of abstractly reasoning about derivatives to abstractly reasoning about real-valued straight-line programs, it can be instantiated with almost any numerical abstract domain, both relational and non-relational. We formally establish the soundness of this construction. We implement our technique by instantiating our construction with both the non-relational interval domain and the relational zonotope domain to compute both first and higher-order derivatives. In the latter case, we are the first to apply a relational domain to automatic differentiation for abstracting higher-order derivatives, and hence we are also the first abstract interpretation work to track correlations across not only different variables, but different orders of derivatives. We evaluate these instantiations on multiple case studies, namely robustly explaining a neural network and more precisely computing a neural network\u2019s Lipschitz constant. For robust interpretation, first and second derivatives computed via zonotope AD are up to 4.76\u00d7 and 6.98\u00d7 more precise, respectively, compared to interval AD. For Lipschitz certification, we obtain bounds that are up to 11,850\u00d7 more precise with zonotopes, compared to the state-of-the-art interval-based tool."
                    },
                    {
                        "title": "Proof transfer for fast certification of multiple approximate neural networks",
                        "abstract": "Developers of machine learning applications often apply post-training neural network optimizations, such as quantization and pruning, that approximate a neural network to speed up inference and reduce energy consumption, while maintaining high accuracy and robustness. Despite a recent surge in techniques for the robustness verification of neural networks, a major limitation of almost all state-of-the-art approaches is that the verification needs to be run from scratch every time the network is even slightly modified. Running precise end-to-end verification from scratch for every new network is expensive and impractical in many scenarios that use or compare multiple approximate network versions, and the robustness of all the networks needs to be verified efficiently. We present FANC, the first general technique for transferring proofs between a given network and its multiple approximate versions without compromising verifier precision. To reuse the proofs obtained when verifying the original network, FANC generates a set of templates \u2013 connected symbolic shapes at intermediate layers of the original network \u2013 that capture the proof of the property to be verified. We present novel algorithms for generating and transforming templates that generalize to a broad range of approximate networks and reduce the verification cost. We present a comprehensive evaluation demonstrating the effectiveness of our approach. We consider a diverse set of networks obtained by applying popular approximation techniques such as quantization and pruning on fully-connected and convolutional architectures and verify their robustness against different adversarial attacks such as adversarial patches, L0, rotation and brightening. Our results indicate that FANC can significantly speed up verification with state-of-the-art verifier, DeepZ by up to 4.1x."
                    },
                    {
                        "title": "Efficient Reward Poisoning Attacks on Online Deep Reinforcement Learning",
                        "abstract": "We study reward poisoning attacks on online deep reinforcement learning (DRL), where the attacker is oblivious to the learning algorithm used by the agent and the dynamics of the environment. We demonstrate the intrinsic vulnerability of state-of-the-art DRL algorithms by designing a general, black-box reward poisoning framework called adversarial MDP attacks. We instantiate our framework to construct two new attacks which only corrupt the rewards for a small fraction of the total training timesteps and make the agent learn a low-performing policy. We provide a theoretical analysis of the efficiency of our attack and perform an extensive empirical evaluation. Our results show that our attacks efficiently poison agents learning in several popular classical control and MuJoCo environments with a variety of state-of-the-art DRL algorithms, such as DQN, PPO, SAC, etc."
                    },
                    {
                        "title": "Probabilistic Trust Intervals for Out of Distribution Detection",
                        "abstract": "Building neural network classifiers with an ability to distinguish between in and out-of distribution inputs is an important step towards faithful deep learning systems. Some of the successful approaches for this, resort to architectural novelties, such as ensembles, with increased complexities in terms of the number of parameters and training procedures. Whereas some other approaches make use of surrogate samples, which are easy to create and work as proxies for actual out-of-distribution (OOD) samples, to train the networks for OOD detection. In this paper, we propose a very simple approach for enhancing the ability of a pretrained network to detect OOD inputs without even altering the original parameter values. We define a probabilistic trust interval for each weight parameter of the network and optimize its size according to the in-distribution (ID) inputs. It allows the network to sample additional weight values along with the original values at the time of inference and use the observed disagreement among the corresponding outputs for OOD detection. In order to capture the disagreement effectively, we also propose a measure and establish its suitability using empirical evidence. Our approach outperforms the existing state-of-the-art methods on various OOD datasets by considerable margins without using any real or surrogate OOD samples. We also analyze the performance of our approach on adversarial and corrupted inputs such as CIFAR-10-C and demonstrate its ability to clearly distinguish such inputs as well. By using fundamental theorem of calculus on neural networks, we explain why our technique doesn't need to observe OOD samples during training to achieve results better than the previous works."
                    },
                    {
                        "title": "Robustness Certification for Point Cloud Models",
                        "abstract": "The use of deep 3D point cloud models in safety-critical applications, such as autonomous driving, dictates the need to certify the robustness of these models to real-world trans-formations. This is technically challenging, as it requires a scalable verifier tailored to point cloud models that handles a wide range of semantic 3D transformations. In this work, we address this challenge and introduce 3DCertify, the first verifier able to certify the robustness of point cloud models. 3DCertify is based on two key insights: (i) a generic relaxation based on first-order Taylor approximations, applicable to any differentiable transformation, and (ii) a precise relaxation for global feature pooling, which is more complex than pointwise activations (e.g., ReLU or sigmoid) but commonly employed in point cloud models. We demonstrate the effectiveness of 3DCertify by performing an extensive evaluation on a wide range of 3D transformations (e.g., rotation, twisting) for both classification and part segmentation tasks. For example, we can certify robustness against rotations by \u00b160\u00b0 for 95.7% of point clouds, and our max pool relaxation increases certification by up to 15.6%."
                    },
                    {
                        "title": "PRIMA: general and precise neural network certification via scalable convex hull approximations",
                        "abstract": "Formal verification of neural networks is critical for their safe adoption in real-world applications. However, designing a precise and scalable verifier which can handle different activation functions, realistic network architectures and relevant specifications remains an open and difficult challenge. In this paper, we take a major step forward in addressing this challenge and present a new verification framework, called PRIMA. PRIMA is both (i) general: it handles any non-linear activation function, and (ii) precise: it computes precise convex abstractions involving multiple neurons via novel convex hull approximation algorithms that leverage concepts from computational geometry. The algorithms have polynomial complexity, yield fewer constraints, and minimize precision loss. We evaluate the effectiveness of PRIMA on a variety of challenging tasks from prior work. Our results show that PRIMA is significantly more precise than the state-of-the-art, verifying robustness to input perturbations for up to 20%, 30%, and 34% more images than existing work on ReLU-, Sigmoid-, and Tanh-based networks, respectively. Further, PRIMA enables, for the first time, the precise verification of a realistic neural network for autonomous driving within a few minutes."
                    },
                    {
                        "title": "Precise Multi-Neuron Abstractions for Neural Network Certification",
                        "abstract": "Formal verification of neural networks is critical for their safe adoption in real-world applications. However, designing a verifier which can handle realistic networks in a precise manner remains an open and difficult challenge. In this paper, we take a major step in addressing this challenge and present a new framework, called PRIMA, that computes precise convex approximations of arbitrary non-linear activations. PRIMA is based on novel approximation algorithms that compute the convex hull of polytopes, leveraging concepts from computational geometry. The algorithms have polynomial complexity, yield fewer constraints, and minimize precision loss. We evaluate the effectiveness of PRIMA on challenging neural networks with ReLU, Sigmoid, and Tanh activations. Our results show that PRIMA is significantly more precise than the state-of-the-art, verifying robustness for up to 16%, 30%, and 34% more images than prior work on ReLU-, Sigmoid-, and Tanh-based networks, respectively."
                    },
                    {
                        "title": "ERAN User Manual",
                        "abstract": "Neural networks are becoming more and more important. They are applied in many real-world applications, such as speech recognition [Hinton et al. 2012], medical diagnostic [Al-Shayea 2011], and autonomous driving [Bojarski et al. 2016] among other fields. Thus it is critical that neural networks behave reliably. This has led to the need to verify the safety of these networks. A number of efforts have been made in this direction, from less scalable but complete verifiers based on SMT solving [Katz et al. 2017, Ehlers 2017, Bunel et al. 2017], mixed-integer linear programming [Tjeng and Tedrake 2017], or Lipschitz optimization [Ruan et al. 2018] to more scalable but incomplete verifiers based on abstract interpretation [Gehr et al. 2018, Singh et al. 2018, Singh et al. 2019a], duality [Dvijotham et al. 2018, Kolter and Wong 2017], semidefinite programming [Raghunathan et al. 2018, Dvijotham et al. 2019] or linear relaxations [Weng et al. 2018, Zhang et al. 2018, Boopathy et al. 2019]. All of the above works require specific network formats, which limits usability. The Secure, Reliable, and Intelligent Systems Lab (SRI) at ETH aims to provide all of its state-of-the-art research on neural network verification in one tool with the ETH Robustness Analyzer for Neural Networks (ERAN). The goal of this work is to extend ERAN for handling more diverse network formats and architectures and add testing support to increase the trust in the verification process. This thesis describes the newest version of the ERAN, which offers complete and incomplete verification methods for feedforward, convolutional and residual neural networks. ERAN can analyze ONNX and TensorFlow models directly, making it accessible and easy to use."
                    },
                    {
                        "title": "Neural Network Robustness Verification on GPUs",
                        "abstract": "Certifying the robustness of neural networks against adversarial attacks is critical to their reliable adoption in real-world systems including autonomous driving and medical diagnosis. Unfortunately, state-of-the-art verifiers either do not scale to larger networks or are too imprecise to prove robustness, which limits their practical adoption. In this work, we introduce GPUPoly, a scalable verifier that can prove the robustness of significantly larger deep neural networks than possible with prior work. The key insight behind GPUPoly is the design of custom, sound polyhedra algorithms for neural network verification on a GPU. Our algorithms leverage the available GPU parallelism and the inherent sparsity of the underlying neural network verification task. GPUPoly scales to very large networks: for example, it can prove the robustness of a 1M neuron, 34-layer deep residual network in about 1 minute. We believe GPUPoly is a promising step towards the practical verification of large real-world networks."
                    },
                    {
                        "title": "Provably Robust Adversarial Examples",
                        "abstract": "We introduce the concept of provably robust adversarial examples for deep neural networks - connected input regions constructed from standard adversarial examples which are guaranteed to be robust to a set of real-world perturbations (such as changes in pixel intensity and geometric transformations). We present a novel method called PARADE for generating these regions in a scalable manner which works by iteratively refining the region initially obtained via sampling until a refined region is certified to be adversarial with existing state-of-the-art verifiers. At each step, a novel optimization procedure is applied to maximize the region's volume under the constraint that the convex relaxation of the network behavior with respect to the region implies a chosen bound on the certification objective. Our experimental evaluation shows the effectiveness of PARADE: it successfully finds large provably robust regions including ones containing $\\approx 10^{573}$ adversarial examples for pixel intensity and $\\approx 10^{599}$ for geometric perturbations. The provability enables our robust examples to be significantly more effective against state-of-the-art defenses based on randomized smoothing than the individual attacks used to construct the regions."
                    },
                    {
                        "title": "Scaling Polyhedral Neural Network Verification on GPUs",
                        "abstract": "Certifying the robustness of neural networks against adversarial attacks is essential to their reliable adoption in safety-critical systems such as autonomous driving and medical diagnosis. Unfortunately, state-of-the-art verifiers either do not scale to bigger networks or are too imprecise to prove robustness, limiting their practical adoption. In this work, we introduce GPUPoly, a scalable verifier that can prove the robustness of significantly larger deep neural networks than previously possible. The key technical insight behind GPUPoly is the design of custom, sound polyhedra algorithms for neural network verification on a GPU. Our algorithms leverage the available GPU parallelism and inherent sparsity of the underlying verification task. GPUPoly scales to large networks: for example, it can prove the robustness of a 1M neuron, 34-layer deep residual network in approximately 34.5 ms. We believe GPUPoly is a promising step towards practical verification of real-world neural networks."
                    },
                    {
                        "title": "Adversarial Attacks on Probabilistic Autoregressive Forecasting Models",
                        "abstract": "We develop an effective generation of adversarial attacks on neural models that output a sequence of probability distributions rather than a sequence of single values. This setting includes the recently proposed deep probabilistic autoregressive forecasting models that estimate the probability distribution of a time series given its past and achieve state-of-the-art results in a diverse set of application domains. The key technical challenge we address is effectively differentiating through the Monte-Carlo estimation of statistics of the joint distribution of the output sequence. Additionally, we extend prior work on probabilistic forecasting to the Bayesian setting which allows conditioning on future observations, instead of only on past observations. We demonstrate that our approach can successfully generate attacks with small input perturbations in two challenging tasks where robust decision making is crucial: stock market trading and prediction of electricity consumption."
                    },
                    {
                        "title": "Learning fast and precise numerical analysis",
                        "abstract": "Numerical abstract domains are a key component of modern static analyzers. Despite recent advances, precise analysis with highly expressive domains remains too costly for many real-world programs. To address this challenge, we introduce a new data-driven method, called LAIT, that produces a faster and more scalable numerical analysis without significant loss of precision. Our approach is based on the key insight that sequences of abstract elements produced by the analyzer contain redundancy which can be exploited to increase performance without compromising precision significantly. Concretely, we present an iterative learning algorithm that learns a neural policy that identifies and removes redundant constraints at various points in the sequence. We believe that our method is generic and can be applied to various numerical domains. We instantiate LAIT for the widely used Polyhedra and Octagon domains. Our evaluation of LAIT on a range of real-world applications with both domains shows that while the approach is designed to be generic, it is orders of magnitude faster on the most costly benchmarks than a state-of-the-art numerical library while maintaining close-to-original analysis precision. Further, LAIT outperforms hand-crafted heuristics and a domain-specific learning approach in terms of both precision and speed."
                    }
                ]
            },
            "811cf2eb-f4b5-47ff-87c9-4edfcd9f0817": {
                "pk": "811cf2eb-f4b5-47ff-87c9-4edfcd9f0817",
                "name": "Sasa Misailovic",
                "collaborators": [
                    "Gagandeep Singh",
                    "Shubham Ugare",
                    "Jacob S. Laurel",
                    "Tarun Suresh",
                    "Yifan Zhao",
                    "Hashim Sharif",
                    "Vikram S. Adve",
                    "Keyur Joshi",
                    "Rem Yang",
                    "Shanto Rahman"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Code Generation",
                    "Robustness"
                ],
                "publications": [
                    {
                        "title": "Is Watermarking LLM-Generated Code Robust?",
                        "abstract": "We present the first study of the robustness of existing watermarking techniques on Python code generated by large language models. Although existing works showed that watermarking can be robust for natural language, we show that it is easy to remove these watermarks on code by semantic-preserving transformations."
                    },
                    {
                        "title": "Quantizing Large-Language Models for Predicting Flaky Tests",
                        "abstract": "A major challenge in regression testing practice is the presence of flaky tests, which non-deterministically pass or fail when run on the same code. Previous research identified multiple categories of flaky tests. Prior research has also de-veloped techniques for automatically detecting which tests are flaky or categorizing flaky tests, but these techniques generally involve repeatedly rerunning tests in various ways, making them costly to use. Although several recent approaches have utilized large-language models (LLMs) to predict which tests are flaky or predict flaky-test categories without needing to rerun tests, they are costly to use due to relying on a large neural network to perform feature extraction and prediction. We propose FlakyQ to improve the effectiveness of LLM-based flaky-test prediction by quantizing LLM's weights. The quantized LLM can extract features from test code more efficiently. To make up for loss in prediction performance due to quantization, we further train a traditional ML classifier (e.g., a random forest) to learn from the quantized LLM-extracted features and do the same prediction. The final model has similar prediction performance while running faster than the non-quantized LLM. Our evaluation finds that FlakyQ classifiers consistently improves prediction time over the non-quantized LLM classifier, saving 25.4% in prediction time over all tests, along with a 48.4 % reduction in memory usage. Furthermore, prediction performance is equal or better than the non-quantized LLM classifier."
                    },
                    {
                        "title": "SynCode: LLM Generation with Grammar Augmentation",
                        "abstract": "LLMs are widely used in complex AI applications. These applications underscore the need for LLM outputs to adhere to a specific format, for their integration with other components in the systems. Typically the format rules e.g., for data serialization formats such as JSON, YAML, or Code in Programming Language are expressed as context-free grammar (CFG). Due to the hallucinations and unreliability of LLMs, instructing LLMs to adhere to specified syntax becomes an increasingly important challenge. We present SynCode, a novel framework for efficient and general syntactical decoding with LLMs, to address this challenge. SynCode ensures soundness and completeness with respect to the CFG of a formal language, effectively retaining valid tokens while filtering out invalid ones. SynCode uses an offline-constructed, efficient lookup table, the DFA mask store, derived from the DFA of the language's grammar for efficient generation. SynCode seamlessly integrates with any language defined by CFG, as evidenced by experiments focusing on generating JSON, Python, and Go outputs. Our experiments evaluating the effectiveness of SynCode for JSON generation demonstrate that SynCode eliminates all syntax errors and significantly outperforms state-of-the-art baselines. Furthermore, our results underscore how SynCode significantly reduces 96.07% of syntax errors in generated Python and Go code, showcasing its substantial impact on enhancing syntactical precision in LLM generation. Our code is available at https://github.com/uiuc-focal-lab/syncode"
                    },
                    {
                        "title": "Felix: Optimizing Tensor Programs with Gradient Descent",
                        "abstract": "Obtaining high-performance implementations of tensor programs such as deep neural networks on a wide range of hardware remains a challenging task. Search-based tensor program optimizers can automatically find high-performance programs on a given hardware platform, but the search process in existing tools suffer from low efficiency, requiring hours or days of time to discover good programs due to the size of the search space. We present Felix, a novel gradient-based compiler optimization framework for tensor-based programs. Felix creates a differentiable space of tensor programs that is amenable to search by gradient descent. Felix applies continuous relaxation on the space of programs and creates differentiable estimator of program latency, allowing efficient search of program candidates using gradient descent, in contrast to conventional approaches that search over a non-differentiable objective function over a discrete search space. We perform an extensive evaluation on six deep neural networks for vision and natural language processing tasks on three GPU-based platforms. Our experiments show that Felix surpasses the performance of off-the-shelf inference frameworks - PyTorch, Tensorflow, and TensorRT - within 7 minutes of search time on average. Felix also finds optimized programs significantly faster than TVM Ansor, a state-of-the-art search-based optimizer for tensor programs."
                    },
                    {
                        "title": "IterGen: Iterative Structured LLM Generation",
                        "abstract": "Large Language Models (LLMs) are widely used for tasks such as natural language and code generation. Still, their outputs often suffer from issues like privacy violations, and semantically inaccurate code generation. Current libraries for LLM generation rely on left-to-right decoding without systematic support for backtracking, limiting the ability to correct or refine outputs mid-generation. To address this issue, we introduce IterGen, an intuitive framework for iterative, grammar-guided LLM generation that enables users to move both forward and backward within the generated output based on grammar symbols. By leveraging a symbol-to-position mapping, IterGen ensures efficient and structured generation while allowing for corrections during the process. We demonstrate IterGen's effectiveness in two important applications: reducing privacy leakage in LLM outputs and improving the accuracy of LLM-generated SQL queries. Our code is available at https://github.com/uiuc-arc/itergen"
                    },
                    {
                        "title": "FastFlip: Compositional Error Injection Analysis",
                        "abstract": "Instruction-level error injection analyses aim to find instructions where errors often lead to unacceptable outcomes like Silent Data Corruptions (SDCs). These analyses require significant time, which is especially problematic if developers wish to regularly analyze software that evolves over time. We present FastFlip, a combination of empirical error injection and symbolic SDC propagation analyses that enables fast, compositional error injection analysis of evolving programs. FastFlip calculates how SDCs propagate across program sections and correctly accounts for unexpected side effects that can occur due to errors. Using FastFlip, we analyze five benchmarks, plus two modified versions of each benchmark. FastFlip speeds up the analysis of incrementally modified programs by $3.2\\times$ (geomean). FastFlip selects a set of instructions to protect against SDCs that minimizes the runtime cost of protection while protecting against a developer-specified target fraction of all SDC-causing errors."
                    },
                    {
                        "title": "Mobiprox: Supporting Dynamic Approximate Computing on Mobiles",
                        "abstract": "Runtime-tunable context-dependent network compression would make mobile deep learning (DL) adaptable to often varying resource availability, input \u201cdifficulty,\u201d or user needs. The existing compression techniques significantly reduce the memory, processing, and energy tax of DL, yet, the resulting models tend to be permanently impaired, sacrificing the inference power for reduced resource usage. The existing tunable compression approaches, on the other hand, require expensive retraining, do not support arbitrary strategies for adapting the compression and do not provide mobile-ready implementations. In this article, we present Mobiprox, a framework enabling mobile DL with flexible precision. Mobiprox implements tunable approximations of tensor operations and enables runtime-adaptable approximation of individual network layers. A profiler and a tuner included with Mobiprox identify the most promising neural network approximation configurations leading to the desired inference quality with the minimal use of resources. Furthermore, we develop control strategies that depending on contextual factors, such as the input data difficulty, dynamically adjust the approximation levels across a mobile DL model\u2019s layers. We implement Mobiprox in Android OS and through experiments in diverse mobile domains, including human activity recognition and spoken keyword detection, demonstrate that it can save up to 15% system-wide energy with a minimal impact on the inference accuracy."
                    },
                    {
                        "title": "ASTRA: Understanding the practical impact of robustness for probabilistic programs",
                        "abstract": "We present the \ufb01rst systematic study of effectiveness of robustness transformations on a diverse set of 24 probabilistic programs representing generalized linear models, mixture models, and time-series models. We evaluate \ufb01ve robustness transformations from literature on each model. We quantify and present insights on (1) the improvement of the posterior prediction accuracy and (2) the execution time overhead of the robusti\ufb01ed programs, in the presence of three input noise models. To automate the evaluation of various robustness transformations, we developed ASTRA \u2013 a novel framework for quantifying the robustness of probabilistic programs and exploring the trade-offs between robustness and execution time. Our experimental results indicate that the existing transformations are often suitable only for speci\ufb01c noise models, can signi\ufb01cantly increase execution time, and have non-trivial interaction with the inference algorithms."
                    },
                    {
                        "title": "Synthesizing Precise Static Analyzers for Automatic Differentiation",
                        "abstract": "We present Pasado, a technique for synthesizing precise static analyzers for Automatic Differentiation. Our technique allows one to automatically construct a static analyzer specialized for the Chain Rule, Product Rule, and Quotient Rule computations for Automatic Differentiation in a way that abstracts all of the nonlinear operations of each respective rule simultaneously. By directly synthesizing an abstract transformer for the composite expressions of these 3 most common rules of AD, we are able to obtain significant precision improvement compared to prior works which compose standard abstract transformers together suboptimally. We prove our synthesized static analyzers sound and additionally demonstrate the generality of our approach by instantiating these AD static analyzers with different nonlinear functions, different abstract domains (both intervals and zonotopes) and both forward-mode and reverse-mode AD. We evaluate Pasado on multiple case studies, namely soundly computing bounds on a neural network\u2019s local Lipschitz constant, soundly bounding the sensitivities of financial models, certifying monotonicity, and lastly, bounding sensitivities of the solutions of differential equations from climate science and chemistry for verified ranges of initial conditions and parameters. The local Lipschitz constants computed by Pasado on our largest CNN are up to 2750\u00d7 more precise compared to the existing state-of-the-art zonotope analysis. The bounds obtained on the sensitivities of the climate, chemical, and financial differential equation solutions are between 1.31 \u2212 2.81\u00d7 more precise (on average) compared to a state-of-the-art zonotope analysis."
                    },
                    {
                        "title": "ApproxCaliper: A Programmable Framework for Application-aware Neural Network Optimization",
                        "abstract": "To deploy compute-intensive neural networks on resource-constrained edge systems, developers use model optimization techniques that reduce model size and computational cost. Existing optimization tools are application-agnostic \u2013 they optimize model parameters solely in view of the neural network accuracy \u2013 and can thus miss optimization opportunities. We propose ApproxCaliper , the first programmable framework for application-aware neural network optimization . By incorporating application-specific goals, ApproxCaliper facilitates more aggressive optimization of the neural networks compared to application-agnostic techniques. We perform experiments on five different neural networks used in two real-world robotics systems: a commercial agriculture robot and a simulation of an autonomous electric cart. Compared to Learning Rate Rewinding (LRR), a state-of-the-art structured pruning tool used in an application agnostic setting, ApproxCaliper achieves 5.3 \u00d7 higher speedup and 2.9 \u00d7 lower GPU resource utilization, and 36 \u00d7 and 6.1 \u00d7 additional model size reduction for the two evaluated benchmarks, respectively."
                    },
                    {
                        "title": "ViX: Analysis-driven Compiler for Efficient Low-Precision Variational Inference",
                        "abstract": "As large quantities of stochastic data are processed onboard tiny edge devices, these systems must constantly make decisions under uncertainty. This challenge necessitates principled embedded compiler support for time- and energy-efficient probabilistic inference. However, compiling probabilistic inference to run on the edge is significantly understudied, and the existing research is limited to computationally expensive MCMC algorithms. Hence, these works cannot leverage faster variational inference algorithms which can better scale to larger data sizes that are representative of realistic workloads in the edge setting. However, naively writing code for differentiable inference on resource-constrained edge devices is challenging due to the need for expensive floating point computations. Even when using reduced precision, a developer still faces the challenge of choosing the right quantization scheme, as gradients can be notoriously unstable in the face of low-precision. To address these challenges, we propose ViX which is the first compiler for low-precision probabilistic programming with variational inference. ViX generates optimized variational inference code in reduced precision by automatically exploiting Bayesian domain knowledge and analytical mathematical properties to ensure that low-precision gradients can still be effectively used. ViX can scale inference to much larger data-sets than previous compilers for resource-constrained probabilistic programming while attaining both high accuracy and significant speedup. Our evaluation of ViX across 7 benchmarks shows that ViX-generated code is up to 8.15\u00d7 faster than performing the same variational inference in 32-bit floating point and also up to 22.67\u00d7 faster than performing the variational inference in 64-bit double precision, all with minimal accuracy loss. Further, on a subset of our benchmarks, ViX can scale inference to data sizes between 16\u201380\u00d7 larger than the existing state-of-the-art tool Statheros."
                    },
                    {
                        "title": "Incremental Verification of Neural Networks",
                        "abstract": "Complete verification of deep neural networks (DNNs) can exactly determine whether the DNN satisfies a desired trustworthy property (e.g., robustness, fairness) on an infinite set of inputs or not. Despite the tremendous progress to improve the scalability of complete verifiers over the years on individual DNNs, they are inherently inefficient when a deployed DNN is updated to improve its inference speed or accuracy. The inefficiency is because the expensive verifier needs to be run from scratch on the updated DNN. To improve efficiency, we propose a new, general framework for incremental and complete DNN verification based on the design of novel theory, data structure, and algorithms. Our contributions implemented in a tool named IVAN yield an overall geometric mean speedup of 2.4x for verifying challenging MNIST and CIFAR10 classifiers and a geometric mean speedup of 3.8x for the ACAS-XU classifiers over the state-of-the-art baselines."
                    },
                    {
                        "title": "GAS: Generating Fast and Accurate Surrogate Models for Autonomous Vehicle Systems",
                        "abstract": "Modern autonomous vehicle systems use complex perception and control components. These components can rapidly change during development of such systems, requiring constant re-testing. Unfortunately, high-fidelity simulations of these complex systems for evaluating vehicle safety are costly. The complexity also hinders the creation of less computationally intensive surrogate models. We present GAS, the first approach for creating surrogate models of complete (perception, control, and dynamics) autonomous vehicle systems containing complex perception and/or control components. GAS's two-stage approach first replaces complex perception components with a perception model. Then, GAS constructs a polynomial surrogate model of the complete vehicle system using Generalized Polynomial Chaos (GPC). We demonstrate the use of these surrogate models in two applications. First, we estimate the probability that the vehicle will enter an unsafe state over time. Second, we perform global sensitivity analysis of the vehicle system with respect to its state in a previous time step. GAS's approach also allows for reuse of the perception model when vehicle control and dynamics characteristics are altered during vehicle development, saving significant time. We consider five scenarios concerning crop management vehicles that must not crash into adjacent crops, self driving cars that must stay within their lane, and unmanned aircraft that must avoid collision. Each of the systems in these scenarios contain a complex perception or control component. Using GAS, we generate surrogate models for these systems, and evaluate the generated models in the applications described above. GAS's surrogate models provide an average speedup of $3.7\\times$ for safe state probability estimation (minimum $2.1\\times$) and $1.4\\times$ for sensitivity analysis (minimum $1.3\\times$), while still maintaining high accuracy."
                    },
                    {
                        "title": "A dual number abstraction for static analysis of Clarke Jacobians",
                        "abstract": "We present a novel abstraction for bounding the Clarke Jacobian of a Lipschitz continuous, but not necessarily differentiable function over a local input region. To do so, we leverage a novel abstract domain built upon dual numbers, adapted to soundly over-approximate all first derivatives needed to compute the Clarke Jacobian. We formally prove that our novel forward-mode dual interval evaluation produces a sound, interval domain-based over-approximation of the true Clarke Jacobian for a given input region. Due to the generality of our formalism, we can compute and analyze interval Clarke Jacobians for a broader class of functions than previous works supported \u2013 specifically, arbitrary compositions of neural networks with Lipschitz, but non-differentiable perturbations. We implement our technique in a tool called DeepJ and evaluate it on multiple deep neural networks and non-differentiable input perturbations to showcase both the generality and scalability of our analysis. Concretely, we can obtain interval Clarke Jacobians to analyze Lipschitz robustness and local optimization landscapes of both fully-connected and convolutional neural networks for rotational, contrast variation, and haze perturbations, as well as their compositions."
                    },
                    {
                        "title": "Provable Defense Against Geometric Transformations",
                        "abstract": "Geometric image transformations that arise in the real world, such as scaling and rotation, have been shown to easily deceive deep neural networks (DNNs). Hence, training DNNs to be certifiably robust to these perturbations is critical. However, no prior work has been able to incorporate the objective of deterministic certified robustness against geometric transformations into the training procedure, as existing verifiers are exceedingly slow. To address these challenges, we propose the first provable defense for deterministic certified geometric robustness. Our framework leverages a novel GPU-optimized verifier that can certify images between 60$\\times$ to 42,600$\\times$ faster than existing geometric robustness verifiers, and thus unlike existing works, is fast enough for use in training. Across multiple datasets, our results show that networks trained via our framework consistently achieve state-of-the-art deterministic certified geometric robustness and clean accuracy. Furthermore, for the first time, we verify the geometric robustness of a neural network for the challenging, real-world setting of autonomous driving."
                    }
                ]
            }
        }
    },
    "2402.07876": {
        "paper_data": {
            "title": "Policy Improvement using Language Feedback Models",
            "url": "http://arxiv.org/abs/2402.07876v6",
            "arxiv_id": "2402.07876",
            "authors": [
                "Victor Zhong",
                "Dipendra Misra",
                "Xingdi Yuan",
                "Marc-Alexandre C\u00f4t\u00e9"
            ],
            "abstract": "We introduce Language Feedback Models (LFMs) that identify desirable behaviour - actions that help achieve tasks specified in the instruction - for imitation learning in instruction following. To train LFMs, we obtain feedback from Large Language Models (LLMs) on visual trajectories verbalized to language descriptions. First, by using LFMs to identify desirable behaviour to imitate, we improve in task-completion rate over strong behavioural cloning baselines on three distinct language grounding environments (Touchdown, ScienceWorld, and ALFWorld). Second, LFMs outperform using LLMs as experts to directly predict actions, when controlling for the number of LLM output tokens. Third, LFMs generalize to unseen environments, improving task-completion rate by 3.5-12.0% through one round of adaptation. Finally, LFM can be modified to provide human-interpretable feedback without performance loss, allowing human verification of desirable behaviour for imitation learning.",
            "introduction": "   1 Introduction     (a)  Learning a small and cost-effective\u00a0Language Feedback Model\u00a0from LLM feedback. We roll out an initial policy, then prompt an LLM to provide feedback on what actions the policy took during the rollout were productive in achieving the task outlined in the instruction. We then use this data to train a feedback model that predicts whether an action is productive given the instruction.       (b)  Policy improvement by imitating desirable behaviour identified by a learned feedback model. Given the instruction, we roll out a base policy, then identify productive actions that help achieve tasks specified in the instruction using the trained feedback model. Finally, we update the base policy by imitating productive actions.       (c)  Example of desirable behaviour identified in ALFWorld, a kitchen instruction following benchmark.     Figure 1:  Given an environment and instructions to follow, we assume a verbalization procedure that converts observations to language descriptions. Policy improvement using\u00a0Language Feedback Model\u00a0involves (a) training a feedback model, then (b) using it to identify desirable behaviour for policy improvement via imitation learning. The feedback model is yellow, other models purple, and generated intermediate data\u00a0green. An example of\u00a0LFM-identified behaviour is shown in (c).    Sample-efficiency and generalizability are two primary challenges in learning instruction following agents in grounded environments\u00a0[26, 23, 2]. First, we want an agent that is sample-efficient: it learns from few demonstrations of how to act according to instructions. Second, we want an agent that is generalizable: it should act successfully in novel environments according to new instructions after training. Reinforcement learning (RL;\u00a0Sutton and Barto [40]) and imitation learning (IL;\u00a0Schaal [32],\u00a0Abbeel and Ng [1]) are two techniques for learning agents for instruction following in grounded environments. These techniques often require large numbers of trials and errors or expensive-to-obtain expert demonstrations. Recent work show that pretrained large language models (LLMs) exhibit sample-efficient learning through prompting and in-context learning for textual\u00a0[10] and grounded problems such as robotic control\u00a0[2]. However, for instruction following in grounded problems, current methods rely on LLMs on-line during inference, which is impractical and expensive.   We develop a sample-efficient and cost-effective technique that uses LLMs to train\u00a0Language Feedback Models (LFMs) for policy improvement in instruction following. Figure\u00a01\u00a0illustrates policy improvement using\u00a0LFMs. Consider the task of interacting with objects in a kitchen to follow instructions shown in Figure\u00a01(c). First, in Figure\u00a01(a), given a grounded environment and a base policy (i.e.\u00a0a behaviour cloned policy), we roll out the base policy to collect a small set of trajectories for different instructions. Next, we verbalize observations in the trajectory by describing scenes in language. For each instruction and verbalized trajectory pair, we query an LLM to provide feedback identifying which behaviour in the trajectory is productive to solving the task identified in the instruction (i.e.\u00a0answer yes or no). For instance, given an instruction \u201cput a clean slice of lettuce in the refridgerator\u201d, GPT-4\u00a0[29]\u00a0is able to deduce that key milestones are 1) find the lettuce, 2) slice it 3) wash it in the sink, and 4) put it in the fridge. Consequently, such an LLM is able to identify when an agent is exhibiting\u00a0desirable behaviour conducive to solving tasks outlined in the instruction, for instance by taking the lettuce to the sink, versus undesirable behaviour, for instance by cooking the lettuce. We define desirable behaviour as productive",
            "references": [
                {
                    "title": "OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments",
                    "abstract": "Autonomous agents that accomplish complex computer tasks with minimal human interventions have the potential to transform human-computer interaction, significantly enhancing accessibility and productivity. However, existing benchmarks either lack an interactive environment or are limited to environments specific to certain applications or domains, failing to reflect the diverse and complex nature of real-world computer use, thereby limiting the scope of tasks and agent scalability. To address this issue, we introduce OSWorld, the first-of-its-kind scalable, real computer environment for multimodal agents, supporting task setup, execution-based evaluation, and interactive learning across various operating systems such as Ubuntu, Windows, and macOS. OSWorld can serve as a unified, integrated computer environment for assessing open-ended computer tasks that involve arbitrary applications. Building upon OSWorld, we create a benchmark of 369 computer tasks involving real web and desktop apps in open domains, OS file I/O, and workflows spanning multiple applications. Each task example is derived from real-world computer use cases and includes a detailed initial state setup configuration and a custom execution-based evaluation script for reliable, reproducible evaluation. Extensive evaluation of state-of-the-art LLM/VLM-based agents on OSWorld reveals significant deficiencies in their ability to serve as computer assistants. While humans can accomplish over 72.36% of the tasks, the best model achieves only 12.24% success, primarily struggling with GUI grounding and operational knowledge. Comprehensive analysis using OSWorld provides valuable insights for developing multimodal generalist agents that were not possible with previous benchmarks. Our code, environment, baseline models, and data are publicly available at https://os-world.github.io."
                },
                {
                    "title": "Motif: Intrinsic Motivation from Artificial Intelligence Feedback",
                    "abstract": "Exploring rich environments and evaluating one's actions without prior knowledge is immensely challenging. In this paper, we propose Motif, a general method to interface such prior knowledge from a Large Language Model (LLM) with an agent. Motif is based on the idea of grounding LLMs for decision-making without requiring them to interact with the environment: it elicits preferences from an LLM over pairs of captions to construct an intrinsic reward, which is then used to train agents with reinforcement learning. We evaluate Motif's performance and behavior on the challenging, open-ended and procedurally-generated NetHack game. Surprisingly, by only learning to maximize its intrinsic reward, Motif achieves a higher game score than an algorithm directly trained to maximize the score itself. When combining Motif's intrinsic reward with the environment reward, our method significantly outperforms existing approaches and makes progress on tasks where no advancements have ever been made without demonstrations. Finally, we show that Motif mostly generates intuitive human-aligned behaviors which can be steered easily through prompt modifications, while scaling well with the LLM size and the amount of information given in the prompt."
                },
                {
                    "title": "VELMA: Verbalization Embodiment of LLM Agents for Vision and Language Navigation in Street View",
                    "abstract": "Incremental decision making in real-world environments is one of the most challenging tasks in embodied artificial intelligence. One particularly demanding scenario is Vision and Language Navigation (VLN) which requires visual and natural language understanding as well as spatial and temporal reasoning capabilities. The embodied agent needs to ground its understanding of navigation instructions in observations of a real-world environment like Street View. Despite the impressive results of LLMs in other research areas, it is an ongoing problem of how to best connect them with an interactive visual environment. In this work, we propose VELMA, an embodied LLM agent that uses a verbalization of the trajectory and of visual environment observations as contextual prompt for the next action. Visual information is verbalized by a pipeline that extracts landmarks from the human written navigation instructions and uses CLIP to determine their visibility in the current panorama view. We show that VELMA is able to successfully follow navigation instructions in Street View with only two in-context examples. We further finetune the LLM agent on a few thousand examples and achieve around 25% relative improvement in task completion over the previous state-of-the-art for two datasets."
                },
                {
                    "title": "Mind2Web: Towards a Generalist Agent for the Web",
                    "abstract": "We introduce Mind2Web, the first dataset for developing and evaluating generalist agents for the web that can follow language instructions to complete complex tasks on any website. Existing datasets for web agents either use simulated websites or only cover a limited set of websites and tasks, thus not suitable for generalist web agents. With over 2,000 open-ended tasks collected from 137 websites spanning 31 domains and crowdsourced action sequences for the tasks, Mind2Web provides three necessary ingredients for building generalist web agents: 1) diverse domains, websites, and tasks, 2) use of real-world websites instead of simulated and simplified ones, and 3) a broad spectrum of user interaction patterns. Based on Mind2Web, we conduct an initial exploration of using large language models (LLMs) for building generalist web agents. While the raw HTML of real-world websites are often too large to be fed to LLMs, we show that first filtering it with a small LM significantly improves the effectiveness and efficiency of LLMs. Our solution demonstrates a decent level of performance, even on websites or entire domains the model has never seen before, but there is still a substantial room to improve towards truly generalizable agents. We open-source our dataset, model implementation, and trained models (https://osu-nlp-group.github.io/Mind2Web) to facilitate further research on building a generalist agent for the web."
                },
                {
                    "title": "Augmenting Autotelic Agents with Large Language Models",
                    "abstract": "Humans learn to master open-ended repertoires of skills by imagining and practicing their own goals. This autotelic learning process, literally the pursuit of self-generated (auto) goals (telos), becomes more and more open-ended as the goals become more diverse, abstract and creative. The resulting exploration of the space of possible skills is supported by an inter-individual exploration: goal representations are culturally evolved and transmitted across individuals, in particular using language. Current artificial agents mostly rely on predefined goal representations corresponding to goal spaces that are either bounded (e.g. list of instructions), or unbounded (e.g. the space of possible visual inputs) but are rarely endowed with the ability to reshape their goal representations, to form new abstractions or to imagine creative goals. In this paper, we introduce a language model augmented autotelic agent (LMA3) that leverages a pretrained language model (LM) to support the representation, generation and learning of diverse, abstract, human-relevant goals. The LM is used as an imperfect model of human cultural transmission; an attempt to capture aspects of humans' common-sense, intuitive physics and overall interests. Specifically, it supports three key components of the autotelic architecture: 1)~a relabeler that describes the goals achieved in the agent's trajectories, 2)~a goal generator that suggests new high-level goals along with their decomposition into subgoals the agent already masters, and 3)~reward functions for each of these goals. Without relying on any hand-coded goal representations, reward functions or curriculum, we show that LMA3 agents learn to master a large diversity of skills in a task-agnostic text-based environment."
                },
                {
                    "title": "Reflexion: language agents with verbal reinforcement learning",
                    "abstract": "Large language models (LLMs) have been increasingly used to interact with external environments (e.g., games, compilers, APIs) as goal-driven agents. However, it remains challenging for these language agents to quickly and efficiently learn from trial-and-error as traditional reinforcement learning methods require extensive training samples and expensive model fine-tuning. We propose Reflexion, a novel framework to reinforce language agents not by updating weights, but instead through linguistic feedback. Concretely, Reflexion agents verbally reflect on task feedback signals, then maintain their own reflective text in an episodic memory buffer to induce better decision-making in subsequent trials. Reflexion is flexible enough to incorporate various types (scalar values or free-form language) and sources (external or internally simulated) of feedback signals, and obtains significant improvements over a baseline agent across diverse tasks (sequential decision-making, coding, language reasoning). For example, Reflexion achieves a 91% pass@1 accuracy on the HumanEval coding benchmark, surpassing the previous state-of-the-art GPT-4 that achieves 80%. We also conduct ablation and analysis studies using different feedback signals, feedback incorporation methods, and agent types, and provide insights into how they affect performance."
                },
                {
                    "title": "Guiding Pretraining in Reinforcement Learning with Large Language Models",
                    "abstract": "Reinforcement learning algorithms typically struggle in the absence of a dense, well-shaped reward function. Intrinsically motivated exploration methods address this limitation by rewarding agents for visiting novel states or transitions, but these methods offer limited benefits in large environments where most discovered novelty is irrelevant for downstream tasks. We describe a method that uses background knowledge from text corpora to shape exploration. This method, called ELLM (Exploring with LLMs) rewards an agent for achieving goals suggested by a language model prompted with a description of the agent's current state. By leveraging large-scale language model pretraining, ELLM guides agents toward human-meaningful and plausibly useful behaviors without requiring a human in the loop. We evaluate ELLM in the Crafter game environment and the Housekeep robotic simulator, showing that ELLM-trained agents have better coverage of common-sense behaviors during pretraining and usually match or improve performance on a range of downstream tasks. Code available at https://github.com/yuqingd/ellm."
                },
                {
                    "title": "ReAct: Synergizing Reasoning and Acting in Language Models",
                    "abstract": "While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources, such as knowledge bases or environments, to gather additional information. We apply our approach, named ReAct, to a diverse set of language and decision making tasks and demonstrate its effectiveness over state-of-the-art baselines, as well as improved human interpretability and trustworthiness over methods without reasoning or acting components. Concretely, on question answering (HotpotQA) and fact verification (Fever), ReAct overcomes issues of hallucination and error propagation prevalent in chain-of-thought reasoning by interacting with a simple Wikipedia API, and generates human-like task-solving trajectories that are more interpretable than baselines without reasoning traces. On two interactive decision making benchmarks (ALFWorld and WebShop), ReAct outperforms imitation and reinforcement learning methods by an absolute success rate of 34% and 10% respectively, while being prompted with only one or two in-context examples. Project site with code: https://react-lm.github.io"
                },
                {
                    "title": "Inner Monologue: Embodied Reasoning through Planning with Language Models",
                    "abstract": "Recent works have shown how the reasoning capabilities of Large Language Models (LLMs) can be applied to domains beyond natural language processing, such as planning and interaction for robots. These embodied problems require an agent to understand many semantic aspects of the world: the repertoire of skills available, how these skills influence the world, and how changes to the world map back to the language. LLMs planning in embodied environments need to consider not just what skills to do, but also how and when to do them - answers that change over time in response to the agent's own choices. In this work, we investigate to what extent LLMs used in such embodied contexts can reason over sources of feedback provided through natural language, without any additional training. We propose that by leveraging environment feedback, LLMs are able to form an inner monologue that allows them to more richly process and plan in robotic control scenarios. We investigate a variety of sources of feedback, such as success detection, scene description, and human interaction. We find that closed-loop language feedback significantly improves high-level instruction completion on three domains, including simulated and real table top rearrangement tasks and long-horizon mobile manipulation tasks in a kitchen environment in the real world."
                },
                {
                    "title": "Do As I Can, Not As I Say: Grounding Language in Robotic Affordances",
                    "abstract": "Large language models can encode a wealth of semantic knowledge about the world. Such knowledge could be extremely useful to robots aiming to act upon high-level, temporally extended instructions expressed in natural language. However, a significant weakness of language models is that they lack real-world experience, which makes it difficult to leverage them for decision making within a given embodiment. For example, asking a language model to describe how to clean a spill might result in a reasonable narrative, but it may not be applicable to a particular agent, such as a robot, that needs to perform this task in a particular environment. We propose to provide real-world grounding by means of pretrained skills, which are used to constrain the model to propose natural language actions that are both feasible and contextually appropriate. The robot can act as the language model's\"hands and eyes,\"while the language model supplies high-level semantic knowledge about the task. We show how low-level skills can be combined with large language models so that the language model provides high-level knowledge about the procedures for performing complex and temporally-extended instructions, while value functions associated with these skills provide the grounding necessary to connect this knowledge to a particular physical environment. We evaluate our method on a number of real-world robotic tasks, where we show the need for real-world grounding and that this approach is capable of completing long-horizon, abstract, natural language instructions on a mobile manipulator. The project's website and the video can be found at https://say-can.github.io/."
                },
                {
                    "title": "Analyzing Generalization of Vision and Language Navigation to Unseen Outdoor Areas",
                    "abstract": "Vision and language navigation (VLN) is a challenging visually-grounded language understanding task. Given a natural language navigation instruction, a visual agent interacts with a graph-based environment equipped with panorama images and tries to follow the described route. Most prior work has been conducted in indoor scenarios where best results were obtained for navigation on routes that are similar to the training routes, with sharp drops in performance when testing on unseen environments. We focus on VLN in outdoor scenarios and find that in contrast to indoor VLN, most of the gain in outdoor VLN on unseen data is due to features like junction type embedding or heading delta that are specific to the respective environment graph, while image information plays a very minor role in generalizing VLN to unseen outdoor areas. These findings show a bias to specifics of graph representations of urban environments, demanding that VLN tasks grow in scale and diversity of geographical environments."
                },
                {
                    "title": "ScienceWorld: Is your Agent Smarter than a 5th Grader?",
                    "abstract": "We present ScienceWorld, a benchmark to test agents\u2019 scientific reasoning abilities in a new interactive text environment at the level of a standard elementary school science curriculum. Despite the transformer-based progress seen in question-answering and scientific text processing, we find that current models cannot reason about or explain learned science concepts in novel contexts. For instance, models can easily answer what the conductivity of a known material is but struggle when asked how they would conduct an experiment in a grounded environment to find the conductivity of an unknown material. This begs the question of whether current models are simply retrieving answers by way of seeing a large number of similar examples or if they have learned to reason about concepts in a reusable manner. We hypothesize that agents need to be grounded in interactive environments to achieve such reasoning capabilities. Our experiments provide empirical evidence supporting this hypothesis \u2013 showing that a 1.5 million parameter agent trained interactively for 100k steps outperforms a 11 billion parameter model statically trained for scientific question-answering and reasoning from millions of expert demonstrations."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "SILG: The Multi-environment Symbolic Interactive Language Grounding Benchmark",
                    "abstract": "Existing work in language grounding typically study single environments. How do we build unified models that apply across multiple environments? We propose the multi-environment Symbolic Interactive Language Grounding benchmark (SILG), which unifies a collection of diverse grounded language learning environments under a common interface. SILG consists of grid-world environments that require generalization to new dynamics, entities, and partially observed worlds (RTFM, Messenger, NetHack), as well as symbolic counterparts of visual worlds that require interpreting rich natural language with respect to complex scenes (ALFWorld, Touchdown). Together, these environments provide diverse grounding challenges in richness of observation space, action space, language specification, and plan complexity. In addition, we propose the first shared model architecture for RL on these environments, and evaluate recent advances such as egocentric local convolution, recurrent state-tracking, entity-centric attention, and pretrained LM using SILG. Our shared architecture achieves comparable performance to environment-specific architectures. Moreover, we find that many recent modelling advances do not result in significant gains on environments other than the one they were designed for. This highlights the need for a multi-environment benchmark. Finally, the best models significantly underperform humans on SILG, which suggests ample room for future work. We hope SILG enables the community to quickly identify new methodologies for language grounding that generalize to a diverse set of environments and their associated challenges."
                },
                {
                    "title": "CLIPort: What and Where Pathways for Robotic Manipulation",
                    "abstract": "How can we imbue robots with the ability to manipulate objects precisely but also to reason about them in terms of abstract concepts? Recent works in manipulation have shown that end-to-end networks can learn dexterous skills that require precise spatial reasoning, but these methods often fail to generalize to new goals or quickly learn transferable concepts across tasks. In parallel, there has been great progress in learning generalizable semantic representations for vision and language by training on large-scale internet data, however these representations lack the spatial understanding necessary for fine-grained manipulation. To this end, we propose a framework that combines the best of both worlds: a two-stream architecture with semantic and spatial pathways for vision-based manipulation. Specifically, we present CLIPort, a language-conditioned imitation-learning agent that combines the broad semantic understanding (what) of CLIP [1] with the spatial precision (where) of Transporter [2]. Our end-to-end framework is capable of solving a variety of language-specified tabletop tasks from packing unseen objects to folding cloths, all without any explicit representations of object poses, instance segmentations, memory, symbolic states, or syntactic structures. Experiments in simulated and real-world settings show that our approach is data efficient in few-shot settings and generalizes effectively to seen and unseen semantic concepts. We even learn one multi-task policy for 10 simulated and 9 real-world tasks that is better or comparable to single-task policies."
                },
                {
                    "title": "Language Models are Few-Shot Butlers",
                    "abstract": "Pretrained language models demonstrate strong performance in most NLP tasks when fine-tuned on small task-specific datasets. Hence, these autoregressive models constitute ideal agents to operate in text-based environments where language understanding and generative capabilities are essential. Nonetheless, collecting expert demonstrations in such environments is a time-consuming endeavour. We introduce a two-stage procedure to learn from a small set of demonstrations and further improve by interacting with an environment. We show that language models fine-tuned with only 1.2% of the expert demonstrations and a simple reinforcement learning algorithm achieve a 51% absolute improvement in success rate over existing methods in the ALFWorld environment."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "ALFWorld: Aligning Text and Embodied Environments for Interactive Learning",
                    "abstract": "Given a simple request (e.g., Put a washed apple in the kitchen fridge), humans can reason in purely abstract terms by imagining action sequences and scoring their likelihood of success, prototypicality, and efficiency, all without moving a muscle. Once we see the kitchen in question, we can update our abstract plans to fit the scene. Embodied agents require the same abilities, but existing work does not yet provide the infrastructure necessary for both reasoning abstractly and executing concretely. We address this limitation by introducing ALFWorld, a simulator that enables agents to learn abstract, text-based policies in TextWorld (Cote et al., 2018) and then execute goals from the ALFRED benchmark (Shridhar et al., 2020) in a rich visual environment. ALFWorld enables the creation of a new BUTLER agent whose abstract knowledge, learned in TextWorld, corresponds directly to concrete, visually grounded actions. In turn, as we demonstrate empirically, this fosters better agent generalization than training only in the visually grounded environment. BUTLER's simple, modular design factors the problem to allow researchers to focus on models for improving every piece of the pipeline (language understanding, planning, navigation, visual scene understanding, and so forth)."
                },
                {
                    "title": "Learning to summarize from human feedback",
                    "abstract": "As language models become more powerful, training and evaluation are increasingly bottlenecked by the data and metrics used for a particular task. For example, summarization models are often trained to predict human reference summaries and evaluated using ROUGE, but both of these metrics are rough proxies for what we really care about---summary quality. In this work, we show that it is possible to significantly improve summary quality by training a model to optimize for human preferences. We collect a large, high-quality dataset of human comparisons between summaries, train a model to predict the human-preferred summary, and use that model as a reward function to fine-tune a summarization policy using reinforcement learning. We apply our method to a version of the TL;DR dataset of Reddit posts and find that our models significantly outperform both human reference summaries and much larger models fine-tuned with supervised learning alone. Our models also transfer to CNN/DM news articles, producing summaries nearly as good as the human reference without any news-specific fine-tuning. We conduct extensive analyses to understand our human feedback dataset and fine-tuned models We establish that our reward model generalizes to new datasets, and that optimizing our reward model results in better summaries than optimizing ROUGE according to humans. We hope the evidence from our paper motivates machine learning researchers to pay closer attention to how their training loss affects the model behavior they actually want."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "ALFRED: A Benchmark for Interpreting Grounded Instructions for Everyday Tasks",
                    "abstract": "We present ALFRED (Action Learning From Realistic Environments and Directives), a benchmark for learning a mapping from natural language instructions and egocentric vision to sequences of actions for household tasks. ALFRED includes long, compositional tasks with non-reversible state changes to shrink the gap between research benchmarks and real-world applications. ALFRED consists of expert demonstrations in interactive visual environments for 25k natural language directives. These directives contain both high-level goals like \u201cRinse off a mug and place it in the coffee maker.\u201d and low-level language instructions like \u201cWalk to the coffee maker on the right.\u201d ALFRED tasks are more complex in terms of sequence length, action space, and language than existing vision- and-language task datasets. We show that a baseline model based on recent embodied vision-and-language tasks performs poorly on ALFRED, suggesting that there is significant room for developing innovative grounded visual language understanding models with this benchmark."
                },
                {
                    "title": "Learning to Map Natural Language Instructions to Physical Quadcopter Control using Simulated Flight",
                    "abstract": "We propose a joint simulation and real-world learning framework for mapping navigation instructions and raw first-person observations to continuous control. Our model estimates the need for environment exploration, predicts the likelihood of visiting environment positions during execution, and controls the agent to both explore and visit high-likelihood positions. We introduce Supervised Reinforcement Asynchronous Learning (SuReAL). Learning uses both simulation and real environments without requiring autonomous flight in the physical environment during training, and combines supervised learning for predicting positions to visit and reinforcement learning for continuous control. We evaluate our approach on a natural language instruction-following task with a physical quadcopter, and demonstrate effective execution and exploration behavior."
                },
                {
                    "title": "On Warm-Starting Neural Network Training",
                    "abstract": "In many real-world deployments of machine learning systems, data arrive piecemeal. These learning scenarios may be passive, where data arrive incrementally due to structural properties of the problem (e.g., daily financial data) or active, where samples are selected according to a measure of their quality (e.g., experimental design). In both of these cases, we are building a sequence of models that incorporate an increasing amount of data. We would like each of these models in the sequence to be performant and take advantage of all the data that are available to that point. Conventional intuition suggests that when solving a sequence of related optimization problems of this form, it should be possible to initialize using the solution of the previous iterate---to \"warm start\" the optimization rather than initialize from scratch---and see reductions in wall-clock time. However, in practice this warm-starting seems to yield poorer generalization performance than models that have fresh random initializations, even though the final training losses are similar. While it appears that some hyperparameter settings allow a practitioner to close this generalization gap, they seem to only do so in regimes that damage the wall-clock gains of the warm start. Nevertheless, it is highly desirable to be able to warm-start neural network training, as it would dramatically reduce the resource usage associated with the construction of performant deep learning systems. In this work, we take a closer look at this empirical phenomenon and try to understand when and how it occurs. We also provide a surprisingly simple trick that overcomes this pathology in several important situations, and present experiments that elucidate some of its properties."
                },
                {
                    "title": "RTFM: Generalising to Novel Environment Dynamics via Reading",
                    "abstract": "Obtaining policies that can generalise to new environments in reinforcement learning is challenging. In this work, we demonstrate that language understanding via a reading policy learner is a promising vehicle for generalisation to new environments. We propose a grounded policy learning problem, Read to Fight Monsters (RTFM), in which the agent must jointly reason over a language goal, relevant dynamics described in a document, and environment observations. We procedurally generate environment dynamics and corresponding language descriptions of the dynamics, such that agents must read to understand new environment dynamics instead of memorising any particular information. In addition, we propose txt2$\\pi$, a model that captures three-way interactions between the goal, document, and observations. On RTFM, txt2$\\pi$ generalises to new environments with dynamics not seen during training via reading. Furthermore, our model outperforms baselines such as FiLM and language-conditioned CNNs on RTFM. Through curriculum learning, txt2$\\pi$ produces policies that excel on complex RTFM tasks requiring several reasoning and coreference steps."
                },
                {
                    "title": "Fine-Tuning Language Models from Human Preferences",
                    "abstract": "Reward learning enables the application of reinforcement learning (RL) to tasks where reward is defined by human judgment, building a model of reward by asking humans questions. Most work on reward learning has used simulated environments, but complex information about values is often expressed in natural language, and we believe reward learning for language is a key to making RL practical and safe for real-world tasks. In this paper, we build on advances in generative pretraining of language models to apply reward learning to four natural language tasks: continuing text with positive sentiment or physically descriptive language, and summarization tasks on the TL;DR and CNN/Daily Mail datasets. For stylistic continuation we achieve good results with only 5,000 comparisons evaluated by humans. For summarization, models trained with 60,000 comparisons copy whole sentences from the input but skip irrelevant preamble; this leads to reasonable ROUGE scores and very good performance according to our human labelers, but may be exploiting the fact that labelers rely on simple heuristics."
                },
                {
                    "title": "TOUCHDOWN: Natural Language Navigation and Spatial Reasoning in Visual Street Environments",
                    "abstract": "We study the problem of jointly reasoning about language and vision through a navigation and spatial reasoning task. We introduce the Touchdown task and dataset, where an agent must first follow navigation instructions in a Street View environment to a goal position, and then guess a location in its observed environment described in natural language to find a hidden object. The data contains 9326 examples of English instructions and spatial descriptions paired with demonstrations. We perform qualitative linguistic analysis, and show that the data displays a rich use of spatial reasoning. Empirical analysis shows the data presents an open challenge to existing methods."
                },
                {
                    "title": "Behavioral Cloning from Observation",
                    "abstract": "Humans often learn how to perform tasks via imitation: they observe others perform a task, and then very quickly infer the appropriate actions to take based on their observations. While extending this paradigm to autonomous agents is a well-studied problem in general, there are two particular aspects that have largely been overlooked: (1) that the learning is done from observation only (i.e., without explicit action information), and (2) that the learning is typically done very quickly. In this work, we propose a two-phase, autonomous imitation learning technique called behavioral cloning from observation (BCO), that aims to provide improved performance with respect to both of these aspects. First, we allow the agent to acquire experience in a self-supervised fashion. This experience is used to develop a model which is then utilized to learn a particular task by observing an expert perform that task without the knowledge of the specific actions taken. We experimentally compare BCO to imitation learning methods, including the state-of-the-art, generative adversarial imitation learning (GAIL) technique, and we show comparable task performance in several different simulation domains while exhibiting increased learning speed after expert trajectories become available."
                },
                {
                    "title": "Unified Pragmatic Models for Generating and Following Instructions",
                    "abstract": "We show that explicit pragmatic inference aids in correctly generating and following natural language instructions for complex, sequential tasks. Our pragmatics-enabled models reason about why speakers produce certain instructions, and about how listeners will react upon hearing them. Like previous pragmatic models, we use learned base listener and speaker models to build a pragmatic speaker that uses the base listener to simulate the interpretation of candidate descriptions, and a pragmatic listener that reasons counterfactually about alternative descriptions. We extend these models to tasks with sequential structure. Evaluation of language generation and interpretation shows that pragmatic inference improves state-of-the-art listener models (at correctly interpreting human instructions) and speaker models (at producing instructions correctly interpreted by humans) in diverse settings."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Alignment-Based Compositional Semantics for Instruction Following",
                    "abstract": "This paper describes an alignment-based model for interpreting natural language instructions in context. We approach instruction following as a search over plans, scoring sequences of actions conditioned on structured observations of text and the environment. By explicitly modeling both the low-level compositional structure of individual actions and the high-level structure of full plans, we are able to learn both grounded representations of sentence meaning and pragmatic constraints on interpretation. To demonstrate the model\u2019s flexibility, we apply it to a diverse set of benchmark tasks. On every task, we outperform strong task-specific baselines, and achieve several new state-of-the-art results."
                },
                {
                    "title": "Scheduled Sampling for Sequence Prediction with Recurrent Neural Networks",
                    "abstract": "Recurrent Neural Networks can be trained to produce sequences of tokens given some input, as exemplified by recent results in machine translation and image captioning. The current approach to training them consists of maximizing the likelihood of each token in the sequence given the current (recurrent) state and the previous token. At inference, the unknown previous token is then replaced by a token generated by the model itself. This discrepancy between training and inference can yield errors that can accumulate quickly along the generated sequence. We propose a curriculum learning strategy to gently change the training process from a fully guided scheme using the true previous token, towards a less guided scheme which mostly uses the generated token instead. Experiments on several sequence prediction tasks show that this approach yields significant improvements. Moreover, it was used succesfully in our winning entry to the MSCOCO image captioning challenge, 2015."
                },
                {
                    "title": "Learning to Interpret Natural Language Navigation Instructions from Observations",
                    "abstract": "\n \n The ability to understand natural-language instructions is critical to building intelligent agents that interact with humans. We present a system that learns to transform natural-language navigation instructions into executable formal plans. Given no prior linguistic knowledge, the system learns by simply observing how humans follow navigation instructions. The system is evaluated in three complex virtual indoor environments with numerous objects and landmarks. A previously collected realistic corpus of complex English navigation instructions for these environments is used for training and testing data. By using a learned lexicon to refine inferred plans and a supervised learner to induce a semantic parser, the system is able to automatically learnto correctly interpret a reasonable fraction of the complex instructions in this corpus.\n \n"
                },
                {
                    "title": "A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning",
                    "abstract": "Sequential prediction problems such as imitation learning, where future observations depend on previous predictions (actions), violate the common i.i.d. assumptions made in statistical learning. This leads to poor performance in theory and often in practice. Some recent approaches provide stronger guarantees in this setting, but remain somewhat unsatisfactory as they train either non-stationary or stochastic policies and require a large number of iterations. In this paper, we propose a new iterative algorithm, which trains a stationary deterministic policy, that can be seen as a no regret algorithm in an online learning setting. We show that any such no regret algorithm, combined with additional reduction assumptions, must find a policy with good performance under the distribution of observations it induces in such sequential settings. We demonstrate that this new approach outperforms previous approaches on two challenging imitation learning problems and a benchmark sequence labeling problem."
                },
                {
                    "title": "Toward understanding natural language directions",
                    "abstract": "Speaking using unconstrained natural language is an intuitive and flexible way for humans to interact with robots. Understanding this kind of linguistic input is challenging because diverse words and phrases must be mapped into structures that the robot can understand, and elements in those structures must be grounded in an uncertain environment. We present a system that follows natural language directions by extracting a sequence of spatial description clauses from the linguistic input and then infers the most probable path through the environment given only information about the environmental geometry and detected visible objects. We use a probabilistic graphical model that factors into three key components. The first component grounds landmark phrases such as \u201cthe computers\u201d in the perceptual frame of the robot by exploiting co-occurrence statistics from a database of tagged images such as Flickr. Second, a spatial reasoning component judges how well spatial relations such as \u201cpast the computers\u201d describe a path. Finally, verb phrases such as \u201cturn right\u201d are modeled according to the amount of change in orientation in the path. Our system follows 60% of the directions in our corpus to within 15 meters of the true destination, significantly outperforming other approaches."
                },
                {
                    "title": "Walk the Talk: Connecting Language, Knowledge, and Action in Route Instructions",
                    "abstract": "Following verbal route instructions requires knowledge of language, space, action and perception. We present MARCO, an agent that follows free-form, natural language route instructions by representing and executing a sequence of compound action specifications that model which actions to take under which conditions. MARCO infers implicit actions from knowledge of both linguistic conditional phrases and from spatial action and local configurations. Thus, MARCO performs explicit actions, implicit actions necessary to achieve the stated conditions, and exploratory actions to learn about the world. \n \nWe gathered a corpus of 786 route instructions from six people in three large-scale virtual indoor environments. Thirtysix other people followed these instructions and rated them for quality. These human participants finished at the intended destination on 69% of the trials. MARCO followed the same instructions in the same environments, with a success rate of 61%. We measured the efficacy of action inference with MARCO variants lacking action inference: executing only explicit actions, MARCO succeeded on just 28% of the trials. For this task, inferring implicit actions is essential to follow poor instructions, but is also crucial for many highly-rated route instructions."
                },
                {
                    "title": "Apprenticeship learning via inverse reinforcement learning",
                    "abstract": "We consider learning in a Markov decision process where we are not explicitly given a reward function, but where instead we can observe an expert demonstrating the task that we want to learn to perform. This setting is useful in applications (such as the task of driving) where it may be difficult to write down an explicit reward function specifying exactly how different desiderata should be traded off. We think of the expert as trying to maximize a reward function that is expressible as a linear combination of known features, and give an algorithm for learning the task demonstrated by the expert. Our algorithm is based on using \"inverse reinforcement learning\" to try to recover the unknown reward function. We show that our algorithm terminates in a small number of iterations, and that even though we may never recover the expert's reward function, the policy output by the algorithm will attain performance close to that of the expert, where here performance is measured with respect to the expert's unknown reward function."
                },
                {
                    "title": "Text2Reward: Automated Dense Reward Function Generation for Reinforcement Learning",
                    "abstract": "Designing reward functions is a longstanding challenge in reinforcement learning (RL); it requires specialized knowledge or domain data, leading to high costs for development. To address this, we introduce T EXT 2R EWARD , a data-free framework that automates the generation of dense reward functions based on large language models (LLMs). Given a goal described in natural language, T EXT 2R EWARD generates dense reward functions as an executable program grounded in a compact representation of the environment. Unlike inverse RL and recent work that uses LLMs to write sparse reward codes, T EXT 2R EWARD produces interpretable, free-form dense reward codes that cover a wide range of tasks, utilize existing packages, and allow iterative refinement with human feedback. We evaluate T EXT 2R EWARD on two robotic manipulation benchmarks (M ANI S KILL 2, M ETA W ORLD ) and two locomotion environments of M U J O C O . On 13 of the 17 manipulation tasks, policies trained with generated reward codes achieve similar or better task success rates and convergence speed than expert-written reward codes. For locomotion tasks, our method learns six novel locomotion behaviors with a success rate exceeding 94%. Furthermore, we show that the policies trained in the simulator with our method can be deployed in the real world. Finally, T EXT 2R EWARD further improves the policies by refining their reward functions with human feedback. Video results are available at https://text-to-reward.github.io."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Natural Language Navigation and Spatial Reasoning in Visual Street Environments",
                    "abstract": "We study the problem of jointly reasoning about language and vision through a navigation and spatial reasoning task. We introduce the TOUCHDOWN task and dataset, where an agent must first follow navigation instructions in a real-life visual urban environment to a goal position, and then identify in the observed image a location described in natural language to find a hidden object. The data contains 9,326 examples of English instructions and spatial descriptions paired with demonstrations. We perform qualitative linguistic analysis, and show that the data displays richer use of spatial reasoning compared to related resources. Empirical analysis shows the data presents an open challenge to existing methods."
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                },
                {
                    "title": "A Framework for Behavioural Cloning",
                    "abstract": "This paper describes recent experiments in automatically constructing reactive agents. The method used is behavioural cloning, where the logged data from skilled, human operators are input to an induction program which outputs a control strategy for a complex control task. Initial studies were able to successfully construct such behavioural clones, but suered from several drawbacks, namely, that the clones were brittle and dicult to understand. Current research is aimed at solving these problems by learning in a framework where there is a separation between an agent\u2019s goals and its knowledge of how to achieve them."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively learn a small and cost-effective Language Feedback Model (LFM) from large language model (LLM) feedback to improve policy performance in instruction-following tasks within grounded environments?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem has significant implications for the research community as it addresses the challenge of sample efficiency and generalizability in instruction-following agents. By developing a method that leverages LLMs to provide feedback for policy improvement, we can reduce the reliance on extensive trial-and-error learning and expensive expert demonstrations. This advancement could lead to more practical applications in robotics and AI, enabling agents to learn effectively from fewer examples and adapt to new environments and instructions. Ultimately, this research could pave the way for more intelligent and autonomous systems capable of performing complex tasks in real-world scenarios.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need for sample efficiency and generalizability in learning agents. Traditional reinforcement learning (RL) and imitation learning (IL) methods often require large amounts of data, which can be impractical and costly to obtain. Naive approaches may fail because they do not effectively utilize the feedback from LLMs to identify productive actions, leading to suboptimal policy updates. Additionally, the complexities of accurately verbalizing observations and interpreting LLM feedback in a grounded environment introduce technical and practical obstacles that must be addressed to ensure the effectiveness of the learned feedback model.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on RL and IL techniques that require extensive data collection, which limits their applicability in real-world scenarios. Existing solutions often rely on LLMs during inference, making them impractical due to high computational costs. The gap in research lies in the integration of LLM feedback into a structured learning framework that can efficiently improve policies without the need for continuous online LLM usage. Our approach differs by proposing a method that trains a feedback model offline using LLM feedback, allowing for more efficient policy improvement and reducing the dependency on real-time LLM interactions.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the following key components: (1) Collecting a small set of trajectories from a base policy in a grounded environment, (2) Verbalizing observations from these trajectories, (3) Querying"
            }
        },
        "author_data": {
            "e836d4cc-fa5d-42a6-8b69-e116dab515f0": {
                "pk": "e836d4cc-fa5d-42a6-8b69-e116dab515f0",
                "name": "Victor Zhong",
                "collaborators": [
                    "Luke Zettlemoyer",
                    "Caiming Xiong",
                    "Richard Socher",
                    "Tim Rockt\u00e4schel",
                    "Edward Grefenstette",
                    "Machel Reid",
                    "Sewon Min",
                    "Jesse Mu",
                    "Hannaneh Hajishirzi",
                    "Austin W. Hanjie"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Question Answering",
                    "Reinforcement Learning",
                    "Dialogue Systems"
                ],
                "publications": [
                    {
                        "title": "E3: Entailment-driven Extracting and Editing for Conversational Machine Reading",
                        "abstract": "Conversational machine reading systems help users answer high-level questions (e.g. determine if they qualify for particular government benefits) when they do not know the exact rules by which the determination is made(e.g. whether they need certain income levels or veteran status). The key challenge is that these rules are only provided in the form of a procedural text (e.g. guidelines from government website) which the system must read to figure out what to ask the user. We present a new conversational machine reading model that jointly extracts a set of decision rules from the procedural text while reasoning about which are entailed by the conversational history and which still need to be edited to create questions for the user. On the recently introduced ShARC conversational machine reading dataset, our Entailment-driven Extract and Edit network (E3) achieves a new state-of-the-art, outperforming existing systems as well as a new BERT-based baseline. In addition, by explicitly highlighting which information still needs to be gathered, E3 provides a more explainable alternative to prior work. We release source code for our models and experiments at https://github.com/vzhong/e3."
                    },
                    {
                        "title": "LEWIS: Levenshtein Editing for Unsupervised Text Style Transfer",
                        "abstract": "Many types of text style transfer can be achieved with only small, precise edits (e.g. sentiment transfer from I had a terrible time... to I had a great time...). We propose a coarse-to-fine editor for style transfer that transforms text using Levenshtein edit operations (e.g. insert, replace, delete). Unlike prior single-span edit methods, our method concurrently edits multiple spans in the source text. To train without parallel style text pairs (e.g. pairs of +/- sentiment statements), we propose an unsupervised data synthesis procedure. We first convert text to style-agnostic templates using style classifier attention (e.g. I had a SLOT time...), then fill in slots in these templates using fine-tuned pretrained language models. Our method outperforms existing generation and editing style transfer methods on sentiment (Yelp, Amazon) and politeness (Polite) transfer. In particular, multi-span editing achieves higher performance and more diverse output than single-span editing. Moreover, compared to previous methods on unsupervised data synthesis, our method results in higher quality parallel style pairs and improves model performance."
                    },
                    {
                        "title": "DCN+: Mixed Objective and Deep Residual Coattention for Question Answering",
                        "abstract": "Traditional models for question answering optimize using cross entropy loss, which encourages exact answers at the cost of penalizing nearby or overlapping answers that are sometimes equally accurate. We propose a mixed objective that combines cross entropy loss with self-critical policy learning. The objective uses rewards derived from word overlap to solve the misalignment between evaluation metric and optimization objective. In addition to the mixed objective, we improve dynamic coattention networks (DCN) with a deep residual coattention encoder that is inspired by recent work in deep self-attention and residual networks. Our proposals improve model performance across question types and input lengths, especially for long questions that requires the ability to capture long-term dependencies. On the Stanford Question Answering Dataset, our model achieves state-of-the-art results with 75.1% exact match accuracy and 83.1% F1, while the ensemble obtains 78.9% exact match accuracy and 86.0% F1."
                    },
                    {
                        "title": "Global-Locally Self-Attentive Dialogue State Tracker",
                        "abstract": "Dialogue state tracking, which estimates user goals and requests given the dialogue context, is an essential part of task-oriented dialogue systems. In this paper, we propose the Global-Locally Self-Attentive Dialogue State Tracker (GLAD), which learns representations of the user utterance and previous system actions with global-local modules. Our model uses global modules to share parameters between estimators for different types (called slots) of dialogue states, and uses local modules to learn slot-specific features. We show that this significantly improves tracking of rare states and achieves state-of-the-art performance on the WoZ and DSTC2 state tracking tasks. GLAD obtains 88.1% joint goal accuracy and 97.1% request accuracy on WoZ, outperforming prior work by 3.7% and 5.5%. On DSTC2, our model obtains 74.5% joint goal accuracy and 97.5% request accuracy, outperforming prior work by 1.1% and 1.0%."
                    },
                    {
                        "title": "Dynamic Coattention Networks For Question Answering",
                        "abstract": "Several deep learning models have been proposed for question answering. However, due to their single-pass nature, they have no way to recover from local maxima corresponding to incorrect answers. To address this problem, we introduce the Dynamic Coattention Network (DCN) for question answering. The DCN first fuses co-dependent representations of the question and the document in order to focus on relevant parts of both. Then a dynamic pointing decoder iterates over potential answer spans. This iterative procedure enables the model to recover from initial local maxima corresponding to incorrect answers. On the Stanford question answering dataset, a single DCN model improves the previous state of the art from 71.0% F1 to 75.9%, while a DCN ensemble obtains 80.4% F1."
                    },
                    {
                        "title": "RTFM: Generalising to Novel Environment Dynamics via Reading",
                        "abstract": "Obtaining policies that can generalise to new environments in reinforcement learning is challenging. In this work, we demonstrate that language understanding via a reading policy learner is a promising vehicle for generalisation to new environments. We propose a grounded policy learning problem, Read to Fight Monsters (RTFM), in which the agent must jointly reason over a language goal, relevant dynamics described in a document, and environment observations. We procedurally generate environment dynamics and corresponding language descriptions of the dynamics, such that agents must read to understand new environment dynamics instead of memorising any particular information. In addition, we propose txt2$\\pi$, a model that captures three-way interactions between the goal, document, and observations. On RTFM, txt2$\\pi$ generalises to new environments with dynamics not seen during training via reading. Furthermore, our model outperforms baselines such as FiLM and language-conditioned CNNs on RTFM. Through curriculum learning, txt2$\\pi$ produces policies that excel on complex RTFM tasks requiring several reasoning and coreference steps."
                    },
                    {
                        "title": "Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning",
                        "abstract": "A significant amount of the world's knowledge is stored in relational databases. However, the ability for users to retrieve facts from a database is limited due to a lack of understanding of query languages such as SQL. We propose Seq2SQL, a deep neural network for translating natural language questions to corresponding SQL queries. Our model leverages the structure of SQL queries to significantly reduce the output space of generated queries. Moreover, we use rewards from in-the-loop query execution over the database to learn a policy to generate unordered parts of the query, which we show are less suitable for optimization via cross entropy loss. In addition, we will publish WikiSQL, a dataset of 80654 hand-annotated examples of questions and SQL queries distributed across 24241 tables from Wikipedia. This dataset is required to train our model and is an order of magnitude larger than comparable datasets. By applying policy-based reinforcement learning with a query execution environment to WikiSQL, our model Seq2SQL outperforms attentional sequence to sequence models, improving execution accuracy from 35.9% to 59.4% and logical form accuracy from 23.4% to 48.3%."
                    },
                    {
                        "title": "Efficient and Robust Question Answering from Minimal Context over Documents",
                        "abstract": "Neural models for question answering (QA) over documents have achieved significant performance improvements. Although effective, these models do not scale to large corpora due to their complex modeling of interactions between the document and the question. Moreover, recent work has shown that such models are sensitive to adversarial inputs. In this paper, we study the minimal context required to answer the question, and find that most questions in existing datasets can be answered with a small set of sentences. Inspired by this observation, we propose a simple sentence selector to select the minimal set of sentences to feed into the QA model. Our overall system achieves significant reductions in training (up to 15 times) and inference times (up to 13 times), with accuracy comparable to or better than the state-of-the-art on SQuAD, NewsQA, TriviaQA and SQuAD-Open. Furthermore, our experimental results and analyses show that our approach is more robust to adversarial inputs."
                    },
                    {
                        "title": "Multi-hop Reading Comprehension through Question Decomposition and Rescoring",
                        "abstract": "Multi-hop Reading Comprehension (RC) requires reasoning and aggregation across several paragraphs. We propose a system for multi-hop RC that decomposes a compositional question into simpler sub-questions that can be answered by off-the-shelf single-hop RC models. Since annotations for such decomposition are expensive, we recast sub-question generation as a span prediction problem and show that our method, trained using only 400 labeled examples, generates sub-questions that are as effective as human-authored sub-questions. We also introduce a new global rescoring approach that considers each decomposition (i.e. the sub-questions and their answers) to select the best final answer, greatly improving overall performance. Our experiments on HotpotQA show that this approach achieves the state-of-the-art results, while providing explainable evidence for its decision making in the form of sub-questions."
                    },
                    {
                        "title": "Grounding Language to Entities and Dynamics for Generalization in Reinforcement Learning",
                        "abstract": "We investigate the use of natural language to drive the generalization of control policies and introduce the new multi-task environment Messenger with free-form text manuals describing the environment dynamics. Unlike previous work, Messenger does not assume prior knowledge connecting text and state observations $-$ the control policy must simultaneously ground the game manual to entity symbols and dynamics in the environment. We develop a new model, EMMA (Entity Mapper with Multi-modal Attention) which uses an entity-conditioned attention module that allows for selective focus over relevant descriptions in the manual for each entity in the environment. EMMA is end-to-end differentiable and learns a latent grounding of entities and dynamics from text to observations using only environment rewards. EMMA achieves successful zero-shot generalization to unseen games with new dynamics, obtaining a 40% higher win rate compared to multiple baselines. However, win rate on the hardest stage of Messenger remains low (10%), demonstrating the need for additional work in this direction."
                    },
                    {
                        "title": "M2D2: A Massively Multi-domain Language Modeling Dataset",
                        "abstract": "We present M2D2, a fine-grained, massively multi-domain corpus for studying domain adaptation in language models (LMs). M2D2 consists of 8.5B tokens and spans 145 domains extracted from Wikipedia and Semantic Scholar. Using ontologies derived from Wikipedia and ArXiv categories, we organize the domains in each data source into 22 groups. This two-level hierarchy enables the study of relationships between domains and their effects on in- and out-of-domain performance after adaptation. We also present a number of insights into the nature of effective domain adaptation in LMs, as examples of the new types of studies M2D2 enables. To improve in-domain performance, we show the benefits of adapting the LM along a domain hierarchy; adapting to smaller amounts of fine-grained domain-specific data can lead to larger in-domain performance gains than larger amounts of weakly relevant data. We further demonstrate a trade-off between in-domain specialization and out-of-domain generalization within and across ontologies, as well as a strong correlation between out-of-domain performance and lexical overlap between domains."
                    },
                    {
                        "title": "Coarse-grain Fine-grain Coattention Network for Multi-evidence Question Answering",
                        "abstract": "End-to-end neural models have made significant progress in question answering, however recent studies show that these models implicitly assume that the answer and evidence appear close together in a single document. In this work, we propose the Coarse-grain Fine-grain Coattention Network (CFC), a new question answering model that combines information from evidence across multiple documents. The CFC consists of a coarse-grain module that interprets documents with respect to the query then finds a relevant answer, and a fine-grain module which scores each candidate answer by comparing its occurrences across all of the documents with the query. We design these modules using hierarchies of coattention and self-attention, which learn to emphasize different parts of the input. On the Qangaroo WikiHop multi-evidence question answering task, the CFC obtains a new state-of-the-art result of 70.6% on the blind test set, outperforming the previous best by 3% accuracy despite not using pretrained contextual encoders."
                    },
                    {
                        "title": "Grounded Adaptation for Zero-shot Executable Semantic Parsing",
                        "abstract": "We propose Grounded Adaptation for Zero-shot Executable Semantic Parsing (GAZP) to adapt an existing semantic parser to new environments (e.g. new database schemas). GAZP combines a forward semantic parser with a backward utterance generator to synthesize data (e.g. utterances and SQL queries) in the new environment, then selects cycle-consistent examples to adapt the parser. Unlike data-augmentation, which typically synthesizes unverified examples in the training environment, GAZP synthesizes examples in the new environment whose input-output consistency are verified. On the Spider, Sparc, and CoSQL zero-shot semantic parsing tasks, GAZP improves logical form and execution accuracy of the baseline parser. Our analyses show that GAZP outperforms data-augmentation in the training environment, performance increases with the amount of GAZP-synthesized data, and cycle-consistency is central to successful adaptation."
                    },
                    {
                        "title": "Improving Policy Learning via Language Dynamics Distillation",
                        "abstract": "Recent work has shown that augmenting environments with language descriptions improves policy learning. However, for environments with complex language abstractions, learning how to ground language to observations is difficult due to sparse, delayed rewards. We propose Language Dynamics Distillation (LDD), which pretrains a model to predict environment dynamics given demonstrations with language descriptions, and then fine-tunes these language-aware pretrained representations via reinforcement learning (RL). In this way, the model is trained to both maximize expected reward and retain knowledge about how language relates to environment dynamics. On SILG, a benchmark of five tasks with language descriptions that evaluate distinct generalization challenges on unseen environments (NetHack, ALFWorld, RTFM, Messenger, and Touchdown), LDD outperforms tabula-rasa RL, VAE pretraining, and methods that learn from unlabeled demonstrations in inverse RL and reward shaping with pretrained experts. In our analyses, we show that language descriptions in demonstrations improve sample-efficiency and generalization across environments, and that dynamics modelling with expert demonstrations is more effective than with non-experts."
                    },
                    {
                        "title": "RoMQA: A Benchmark for Robust, Multi-evidence, Multi-answer Question Answering",
                        "abstract": "We introduce RoMQA, the first benchmark for robust, multi-evidence, multi-answer question answering (QA). RoMQA contains clusters of questions that are derived from related constraints mined from the Wikidata knowledge graph. RoMQA evaluates robustness of QA models to varying constraints by measuring worst-case performance within each question cluster. Compared to prior QA datasets, RoMQA has more human-written questions that require reasoning over more evidence text and have, on average, many more correct answers. In addition, human annotators rate RoMQA questions as more natural or likely to be asked by people. We evaluate state-of-the-art large language models in zero-shot, few-shot, and fine-tuning settings, and find that RoMQA is challenging: zero-shot and few-shot models perform similarly to naive baselines, while supervised retrieval methods perform well below gold evidence upper bounds. Moreover, existing models are not robust to variations in question constraints, but can be made more robust by tuning on clusters of related questions. Our results show that RoMQA is a challenging benchmark for large language models, and provides a quantifiable test to build more robust QA methods."
                    },
                    {
                        "title": "Improving Intrinsic Exploration with Language Abstractions",
                        "abstract": "Reinforcement learning (RL) agents are particularly hard to train when rewards are sparse. One common solution is to use intrinsic rewards to encourage agents to explore their environment. However, recent intrinsic exploration methods often use state-based novelty measures which reward low-level exploration and may not scale to domains requiring more abstract skills. Instead, we explore natural language as a general medium for highlighting relevant abstractions in an environment. Unlike previous work, we evaluate whether language can improve over existing exploration methods by directly extending (and comparing to) competitive intrinsic exploration baselines: AMIGo (Campero et al., 2021) and NovelD (Zhang et al., 2021). These language-based variants outperform their non-linguistic forms by 47-85% across 13 challenging tasks from the MiniGrid and MiniHack environment suites."
                    }
                ]
            },
            "c0890c4b-cc4e-46a2-9cec-9b65533e8dab": {
                "pk": "c0890c4b-cc4e-46a2-9cec-9b65533e8dab",
                "name": "Dipendra Misra",
                "collaborators": [
                    "Kavosh Asadi",
                    "John Langford",
                    "Andrew Bennett",
                    "Michael L. Littman",
                    "Yoav Artzi",
                    "Akshay Krishnamurthy",
                    "Jordan T. Ash",
                    "Robert E. Schapire",
                    "Evan Cater",
                    "Valts Blukis"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Neural Networks",
                    "Natural Language Processing",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "Have you tried Neural Topic Models? Comparative Analysis of Neural and Non-Neural Topic Models with Application to COVID-19 Twitter Data",
                        "abstract": "Topic models are widely used in studying social phenomena. We conduct a comparative study examining state-of-the-art neural versus non-neural topic models, performing a rigorous quantitative and qualitative assessment on a dataset of tweets about the COVID-19 pandemic. Our results show that not only do neural topic models outperform their classical counterparts on standard evaluation metrics, but they also produce more coherent topics, which are of great benefit when studying complex social problems. We also propose a novel regularization term for neural topic models, which is designed to address the well-documented problem of mode collapse, and demonstrate its effectiveness."
                    },
                    {
                        "title": "The Truth is in There: Improving Reasoning in Language Models with Layer-Selective Rank Reduction",
                        "abstract": "Transformer-based Large Language Models (LLMs) have become a fixture in modern machine learning. Correspondingly, significant resources are allocated towards research that aims to further advance this technology, typically resulting in models of increasing size that are trained on increasing amounts of data. This work, however, demonstrates the surprising result that it is often possible to significantly improve the performance of LLMs by selectively removing higher-order components of their weight matrices. This simple intervention, which we call LAyer-SElective Rank reduction (LASER), can be done on a model after training has completed, and requires no additional parameters or data. We show extensive experiments demonstrating the generality of this finding across language models and datasets, and provide in-depth analyses offering insights into both when LASER is effective and the mechanism by which it operates."
                    },
                    {
                        "title": "Lipschitz Continuity in Model-based Reinforcement Learning",
                        "abstract": "We examine the impact of learning Lipschitz continuous models in the context of model-based reinforcement learning. We provide a novel bound on multi-step prediction error of Lipschitz models where we quantify the error using the Wasserstein metric. We go on to prove an error bound for the value-function estimate arising from Lipschitz models and show that the estimated value function is itself Lipschitz. We conclude with empirical results that show the benefits of controlling the Lipschitz constant of neural-network models."
                    },
                    {
                        "title": "Provable Interactive Learning with Hindsight Instruction Feedback",
                        "abstract": "We study interactive learning in a setting where the agent has to generate a response (e.g., an action or trajectory) given a context and an instruction. In contrast, to typical approaches that train the system using reward or expert supervision on response, we study learning with hindsight instruction where a teacher provides an instruction that is most suitable for the agent's generated response. This hindsight labeling of instruction is often easier to provide than providing expert supervision of the optimal response which may require expert knowledge or can be impractical to elicit. We initiate the theoretical analysis of interactive learning with hindsight labeling. We first provide a lower bound showing that in general, the regret of any algorithm must scale with the size of the agent's response space. We then study a specialized setting where the underlying instruction-response distribution can be decomposed as a low-rank matrix. We introduce an algorithm called LORIL for this setting and show that its regret scales as $\\sqrt{T}$ where $T$ is the number of rounds and depends on the intrinsic rank but does not depend on the size of the agent's response space. We provide experiments in two domains showing that LORIL outperforms baselines even when the low-rank assumption is violated."
                    },
                    {
                        "title": "Mapping Instructions and Visual Observations to Actions with Reinforcement Learning",
                        "abstract": "We propose to directly map raw visual observations and text input to actions for instruction execution. While existing approaches assume access to structured environment representations or use a pipeline of separately trained models, we learn a single model to jointly reason about linguistic and visual input. We use reinforcement learning in a contextual bandit setting to train a neural network agent. To guide the agent's exploration, we use reward shaping with different forms of supervision. Our approach does not require intermediate representations, planning procedures, or training different models. We evaluate in a simulated environment, and show significant improvements over supervised learning and common reinforcement learning variants."
                    },
                    {
                        "title": "Provable Safe Reinforcement Learning with Binary Feedback",
                        "abstract": "Safety is a crucial necessity in many applications of reinforcement learning (RL), whether robotic, automotive, or medical. Many existing approaches to safe RL rely on receiving numeric safety feedback, but in many cases this feedback can only take binary values; that is, whether an action in a given state is safe or unsafe. This is particularly true when feedback comes from human experts. We therefore consider the problem of provable safe RL when given access to an offline oracle providing binary feedback on the safety of state, action pairs. We provide a novel meta algorithm, SABRE, which can be applied to any MDP setting given access to a blackbox PAC RL algorithm for that setting. SABRE applies concepts from active learning to reinforcement learning to provably control the number of queries to the safety oracle. SABRE works by iteratively exploring the state space to find regions where the agent is currently uncertain about safety. Our main theoretical results shows that, under appropriate technical assumptions, SABRE never takes unsafe actions during training, and is guaranteed to return a near-optimal safe policy with high probability. We provide a discussion of how our meta-algorithm may be applied to various settings studied in both theoretical and empirical frameworks."
                    },
                    {
                        "title": "Kinematic State Abstraction and Provably Efficient Rich-Observation Reinforcement Learning",
                        "abstract": "We present an algorithm, HOMER, for exploration and reinforcement learning in rich observation environments that are summarizable by an unknown latent state space. The algorithm interleaves representation learning to identify a new notion of kinematic state abstraction with strategic exploration to reach new states using the learned abstraction. The algorithm provably explores the environment with sample complexity scaling polynomially in the number of latent states and the time horizon, and, crucially, with no dependence on the size of the observation space, which could be infinitely large. This exploration guarantee further enables sample-efficient global policy optimization for any reward function. On the computational side, we show that the algorithm can be implemented efficiently whenever certain supervised learning problems are tractable. Empirically, we evaluate HOMER on a challenging exploration problem, where we show that the algorithm is exponentially more sample efficient than standard reinforcement learning baselines."
                    },
                    {
                        "title": "Investigating the Role of Negatives in Contrastive Representation Learning",
                        "abstract": "Noise contrastive learning is a popular technique for unsupervised representation learning. In this approach, a representation is obtained via reduction to supervised learning, where given a notion of semantic similarity, the learner tries to distinguish a similar (positive) example from a collection of random (negative) examples. The success of modern contrastive learning pipelines relies on many parameters such as the choice of data augmentation, the number of negative examples, and the batch size; however, there is limited understanding as to how these parameters interact and affect downstream performance. We focus on disambiguating the role of one of these parameters: the number of negative examples. Theoretically, we show the existence of a collision-coverage trade-off suggesting that the optimal number of negative examples should scale with the number of underlying concepts in the data. Empirically, we scrutinize the role of the number of negatives in both NLP and vision tasks. In the NLP task, we find that the results broadly agree with our theory, while our vision experiments are murkier with performance sometimes even being insensitive to the number of negatives. We discuss plausible explanations for this behavior and suggest future directions to better align theory and practice."
                    },
                    {
                        "title": "Towards a Simple Approach to Multi-step Model-based Reinforcement Learning",
                        "abstract": "When environmental interaction is expensive, model-based reinforcement learning offers a solution by planning ahead and avoiding costly mistakes. Model-based agents typically learn a single-step transition model. In this paper, we propose a multi-step model that predicts the outcome of an action sequence with variable length. We show that this model is easy to learn, and that the model can make policy-conditional predictions. We report preliminary results that show a clear advantage for the multi-step model compared to its one-step counterpart."
                    },
                    {
                        "title": "Mapping Navigation Instructions to Continuous Control Actions with Position-Visitation Prediction",
                        "abstract": "We propose an approach for mapping natural language instructions and raw observations to continuous control of a quadcopter drone. Our model predicts interpretable position-visitation distributions indicating where the agent should go during execution and where it should stop, and uses the predicted distributions to select the actions to execute. This two-step model decomposition allows for simple and efficient training using a combination of supervised learning and imitation learning. We evaluate our approach with a realistic drone simulator, and demonstrate absolute task-completion accuracy improvements of 16.85% over two state-of-the-art instruction-following methods."
                    },
                    {
                        "title": "Interactive Learning from Activity Description",
                        "abstract": "We present a novel interactive learning protocol that enables training request-fulfilling agents by verbally describing their activities. Unlike imitation learning (IL), our protocol allows the teaching agent to provide feedback in a language that is most appropriate for them. Compared with reward in reinforcement learning (RL), the description feedback is richer and allows for improved sample complexity. We develop a probabilistic framework and an algorithm that practically implements our protocol. Empirical results in two challenging request-fulfilling problems demonstrate the strengths of our approach: compared with RL baselines, it is more sample-efficient; compared with IL baselines, it achieves competitive success rates without requiring the teaching agent to be able to demonstrate the desired behavior using the learning agent's actions. Apart from empirical evaluation, we also provide theoretical guarantees for our algorithm under certain assumptions about the teacher and the environment."
                    },
                    {
                        "title": "Equivalence Between Wasserstein and Value-Aware Loss for Model-based Reinforcement Learning",
                        "abstract": "Learning a generative model is a key component of model-based reinforcement learning. Though learning a good model in the tabular setting is a simple task, learning a useful model in the approximate setting is challenging. In this context, an important question is the loss function used for model learning as varying the loss function can have a remarkable impact on effectiveness of planning. Recently Farahmand et al. (2017) proposed a value-aware model learning (VAML) objective that captures the structure of value function during model learning. Using tools from Asadi et al. (2018), we show that minimizing the VAML objective is in fact equivalent to minimizing the Wasserstein metric. This equivalence improves our understanding of value-aware models, and also creates a theoretical foundation for applications of Wasserstein in model-based reinforcement~learning."
                    },
                    {
                        "title": "Combating the Compounding-Error Problem with a Multi-step Model",
                        "abstract": "Model-based reinforcement learning is an appealing framework for creating agents that learn, plan, and act in sequential environments. Model-based algorithms typically involve learning a transition model that takes a state and an action and outputs the next state---a one-step model. This model can be composed with itself to enable predicting multiple steps into the future, but one-step prediction errors can get magnified, leading to unacceptable inaccuracy. This compounding-error problem plagues planning and undermines model-based reinforcement learning. In this paper, we address the compounding-error problem by introducing a multi-step model that directly outputs the outcome of executing a sequence of actions. Novel theoretical and empirical results indicate that the multi-step model is more conducive to efficient value-function estimation, and it yields better action selection compared to the one-step model. These results make a strong case for using multi-step models in the context of model-based reinforcement learning."
                    },
                    {
                        "title": "Mapping Instructions to Actions in 3D Environments with Visual Goal Prediction",
                        "abstract": "We propose to decompose instruction execution to goal prediction and action generation. We design a model that maps raw visual observations to goals using LINGUNET, a language-conditioned image generation network, and then generates the actions required to complete them. Our model is trained from demonstration only without external resources. To evaluate our approach, we introduce two benchmarks for instruction following: LANI, a navigation task; and CHAI, where an agent executes household instructions. Our evaluation demonstrates the advantages of our model decomposition, and illustrates the challenges posed by our new benchmarks."
                    },
                    {
                        "title": "Provable RL with Exogenous Distractors via Multistep Inverse Dynamics",
                        "abstract": "Many real-world applications of reinforcement learning (RL) require the agent to deal with high-dimensional observations such as those generated from a megapixel camera. Prior work has addressed such problems with representation learning, through which the agent can provably extract endogenous, latent state information from raw observations and subsequently plan efficiently. However, such approaches can fail in the presence of temporally correlated noise in the observations, a phenomenon that is common in practice. We initiate the formal study of latent state discovery in the presence of such exogenous noise sources by proposing a new model, the Exogenous Block MDP (EX-BMDP), for rich observation RL. We start by establishing several negative results, by highlighting failure cases of prior representation learning based approaches. Then, we introduce the Predictive Path Elimination (PPE) algorithm, that learns a generalization of inverse dynamics and is provably sample and computationally efficient in EX-BMDPs when the endogenous state dynamics are near deterministic. The sample complexity of PPE depends polynomially on the size of the latent endogenous state space while not directly depending on the size of the observation space, nor the exogenous state space. We provide experiments on challenging exploration problems which show that our approach works empirically."
                    },
                    {
                        "title": "Provably Sample-Efficient RL with Side Information about Latent Dynamics",
                        "abstract": "We study reinforcement learning (RL) in settings where observations are high-dimensional, but where an RL agent has access to abstract knowledge about the structure of the state space, as is the case, for example, when a robot is tasked to go to a specific room in a building using observations from its own camera, while having access to the floor plan. We formalize this setting as transfer reinforcement learning from an abstract simulator, which we assume is deterministic (such as a simple model of moving around the floor plan), but which is only required to capture the target domain's latent-state dynamics approximately up to unknown (bounded) perturbations (to account for environment stochasticity). Crucially, we assume no prior knowledge about the structure of observations in the target domain except that they can be used to identify the latent states (but the decoding map is unknown). Under these assumptions, we present an algorithm, called TASID, that learns a robust policy in the target domain, with sample complexity that is polynomial in the horizon, and independent of the number of states, which is not possible without access to some prior knowledge. In synthetic experiments, we verify various properties of our algorithm and show that it empirically outperforms transfer RL algorithms that require access to \"full simulators\" (i.e., those that also simulate observations)."
                    },
                    {
                        "title": "Survival Instinct in Offline Reinforcement Learning",
                        "abstract": "We present a novel observation about the behavior of offline reinforcement learning (RL) algorithms: on many benchmark datasets, offline RL can produce well-performing and safe policies even when trained with \"wrong\" reward labels, such as those that are zero everywhere or are negatives of the true rewards. This phenomenon cannot be easily explained by offline RL's return maximization objective. Moreover, it gives offline RL a degree of robustness that is uncharacteristic of its online RL counterparts, which are known to be sensitive to reward design. We demonstrate that this surprising robustness property is attributable to an interplay between the notion of pessimism in offline RL algorithms and certain implicit biases in common data collection practices. As we prove in this work, pessimism endows the agent with a \"survival instinct\", i.e., an incentive to stay within the data support in the long term, while the limited and biased data coverage further constrains the set of survival policies. Formally, given a reward class -- which may not even contain the true reward -- we identify conditions on the training data distribution that enable offline RL to learn a near-optimal and safe policy from any reward within the class. We argue that the survival instinct should be taken into account when interpreting results from existing offline RL benchmarks and when creating future ones. Our empirical and theoretical results suggest a new paradigm for RL, whereby an agent is nudged to learn a desirable behavior with imperfect reward but purposely biased data coverage."
                    },
                    {
                        "title": "Towards Principled Representation Learning from Videos for Reinforcement Learning",
                        "abstract": "We study pre-training representations for decision-making using video data, which is abundantly available for tasks such as game agents and software testing. Even though significant empirical advances have been made on this problem, a theoretical understanding remains absent. We initiate the theoretical investigation into principled approaches for representation learning and focus on learning the latent state representations of the underlying MDP using video data. We study two types of settings: one where there is iid noise in the observation, and a more challenging setting where there is also the presence of exogenous noise, which is non-iid noise that is temporally correlated, such as the motion of people or cars in the background. We study three commonly used approaches: autoencoding, temporal contrastive learning, and forward modeling. We prove upper bounds for temporal contrastive learning and forward modeling in the presence of only iid noise. We show that these approaches can learn the latent state and use it to do efficient downstream RL with polynomial sample complexity. When exogenous noise is also present, we establish a lower bound result showing that the sample complexity of learning from video data can be exponentially worse than learning from action-labeled trajectory data. This partially explains why reinforcement learning with video pre-training is hard. We evaluate these representational learning methods in two visual domains, yielding results that are consistent with our theoretical findings."
                    },
                    {
                        "title": "Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning",
                        "abstract": "Imitation learning (IL) aims to mimic the behavior of an expert in a sequential decision making task by learning from demonstrations, and has been widely applied to robotics, autonomous driving, and autoregressive text generation. The simplest approach to IL, behavior cloning (BC), is thought to incur sample complexity with unfavorable quadratic dependence on the problem horizon, motivating a variety of different online algorithms that attain improved linear horizon dependence under stronger assumptions on the data and the learner's access to the expert.   We revisit the apparent gap between offline and online IL from a learning-theoretic perspective, with a focus on general policy classes up to and including deep neural networks. Through a new analysis of behavior cloning with the logarithmic loss, we show that it is possible to achieve horizon-independent sample complexity in offline IL whenever (i) the range of the cumulative payoffs is controlled, and (ii) an appropriate notion of supervised learning complexity for the policy class is controlled. Specializing our results to deterministic, stationary policies, we show that the gap between offline and online IL is not fundamental: (i) it is possible to achieve linear dependence on horizon in offline IL under dense rewards (matching what was previously only known to be achievable in online IL); and (ii) without further assumptions on the policy class, online IL cannot improve over offline IL with the logarithmic loss, even in benign MDPs. We complement our theoretical results with experiments on standard RL tasks and autoregressive language generation to validate the practical relevance of our findings."
                    }
                ]
            },
            "75af0bad-490e-4e8f-8554-04f94dc0ff12": {
                "pk": "75af0bad-490e-4e8f-8554-04f94dc0ff12",
                "name": "Xingdi Yuan",
                "collaborators": [
                    "Adam Trischler",
                    "Tong Wang",
                    "Rui Meng",
                    "Daqing He",
                    "Marc-Alexandre C\u00f4t\u00e9",
                    "Kaheer Suleman",
                    "Tsendsuren Munkhdalai",
                    "Khushboo Thaker",
                    "Peter Brusilovsky",
                    "Jie Fu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Reading Comprehension",
                    "Knowledge Graphs",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Interactive Machine Comprehension with Dynamic Knowledge Graphs",
                        "abstract": "Interactive machine reading comprehension (iMRC) is machine comprehension tasks where knowledge sources are partially observable. An agent must interact with an environment sequentially to gather necessary knowledge in order to answer a question. We hypothesize that graph representations are good inductive biases, which can serve as an agent's memory mechanism in iMRC tasks. We explore four different categories of graphs that can capture text information at various levels. We describe methods that dynamically build and update these graphs during information gathering, as well as neural models to encode graph representations in RL agents. Extensive experiments on iSQuAD suggest that graph representations can result in significant performance improvements for RL agents."
                    },
                    {
                        "title": "A Joint Model for Question Answering and Question Generation",
                        "abstract": "We propose a generative machine comprehension model that learns jointly to ask and answer questions based on documents. The proposed model uses a sequence-to-sequence framework that encodes the document and generates a question (answer) given an answer (question). Significant improvement in model performance is observed empirically on the SQuAD corpus, confirming our hypothesis that the model benefits from jointly learning to perform both tasks. We believe the joint model's novelty offers a new perspective on machine comprehension beyond architectural engineering, and serves as a first step towards autonomous information seeking."
                    },
                    {
                        "title": "Natural Language Comprehension with the EpiReader",
                        "abstract": "We present the EpiReader, a novel model for machine comprehension of text. Machine comprehension of unstructured, real-world text is a major research goal for natural language processing. Current tests of machine comprehension pose questions whose answers can be inferred from some supporting text, and evaluate a model's response to the questions. The EpiReader is an end-to-end neural model comprising two components: the first component proposes a small set of candidate answers after comparing a question to its supporting text, and the second component formulates hypotheses using the proposed candidates and the question, then reranks the hypotheses based on their estimated concordance with the supporting text. We present experiments demonstrating that the EpiReader sets a new state-of-the-art on the CNN and Children's Book Test machine comprehension benchmarks, outperforming previous neural models by a significant margin."
                    },
                    {
                        "title": "Rapid Adaptation with Conditionally Shifted Neurons",
                        "abstract": "We describe a mechanism by which artificial neural networks can learn rapid adaptation - the ability to adapt on the fly, with little data, to new tasks - that we call conditionally shifted neurons. We apply this mechanism in the framework of metalearning, where the aim is to replicate some of the flexibility of human learning in machines. Conditionally shifted neurons modify their activation values with task-specific shifts retrieved from a memory module, which is populated rapidly based on limited task experience. On metalearning benchmarks from the vision and language domains, models augmented with conditionally shifted neurons achieve state-of-the-art results."
                    },
                    {
                        "title": "Building Dynamic Knowledge Graphs from Text using Machine Reading Comprehension",
                        "abstract": "We propose a neural machine-reading model that constructs dynamic knowledge graphs from procedural text. It builds these graphs recurrently for each step of the described procedure, and uses them to track the evolving states of participant entities. We harness and extend a recently proposed machine reading comprehension (MRC) model to query for entity states, since these states are generally communicated in spans of text and MRC models perform well in extracting entity-centric spans. The explicit, structured, and evolving knowledge graph representations that our model constructs can be used in downstream question answering tasks to improve machine comprehension of text, as we demonstrate empirically. On two comprehension tasks from the recently proposed PROPARA dataset (Dalvi et al., 2018), our model achieves state-of-the-art results. We further show that our model is competitive on the RECIPES dataset (Kiddon et al., 2015), suggesting it may be generally applicable. We present some evidence that the model's knowledge graphs help it to impose commonsense constraints on its predictions."
                    },
                    {
                        "title": "Interactive Fiction Games: A Colossal Adventure",
                        "abstract": "A hallmark of human intelligence is the ability to understand and communicate with language. Interactive Fiction games are fully text-based simulation environments where a player issues text commands to effect change in the environment and progress through the story. We argue that IF games are an excellent testbed for studying language-based autonomous agents. In particular, IF games combine challenges of combinatorial action spaces, language understanding, and commonsense reasoning. To facilitate rapid development of language-based agents, we introduce Jericho, a learning environment for man-made IF games and conduct a comprehensive study of text-agents across a rich set of games, highlighting directions in which agents can improve."
                    },
                    {
                        "title": "General-to-Specific Transfer Labeling for Domain Adaptable Keyphrase Generation",
                        "abstract": "Training keyphrase generation (KPG) models require a large amount of annotated data, which can be prohibitively expensive and often limited to specific domains. In this study, we first demonstrate that large distribution shifts among different domains severely hinder the transferability of KPG models. We then propose a three-stage pipeline, which gradually guides KPG models' learning focus from general syntactical features to domain-related semantics, in a data-efficient manner. With Domain-general Phrase pre-training, we pre-train Sequence-to-Sequence models with generic phrase annotations that are widely available on the web, which enables the models to generate phrases in a wide range of domains. The resulting model is then applied in the Transfer Labeling stage to produce domain-specific pseudo keyphrases, which help adapt models to a new domain. Finally, we fine-tune the model with limited data with true labels to fully adapt it to the target domain. Our experiment results show that the proposed process can produce good-quality keyphrases in new domains and achieve consistent improvements after adaptation with limited in-domain annotated data. All code and datasets are available at https://github.com/memray/OpenNMT-kpg-release."
                    },
                    {
                        "title": "One Size Does Not Fit All: Generating and Evaluating Variable Number of Keyphrases",
                        "abstract": "Different texts shall by nature correspond to different number of keyphrases. This desideratum is largely missing from existing neural keyphrase generation models. In this study, we address this problem from both modeling and evaluation perspectives.   We first propose a recurrent generative model that generates multiple keyphrases as delimiter-separated sequences. Generation diversity is further enhanced with two novel techniques by manipulating decoder hidden states. In contrast to previous approaches, our model is capable of generating diverse keyphrases and controlling number of outputs.   We further propose two evaluation metrics tailored towards the variable-number generation. We also introduce a new dataset StackEx that expands beyond the only existing genre (i.e., academic writing) in keyphrase generation tasks. With both previous and new evaluation metrics, our model outperforms strong baselines on all datasets."
                    },
                    {
                        "title": "Interactive Machine Comprehension with Information Seeking Agents",
                        "abstract": "Existing machine reading comprehension (MRC) models do not scale effectively to real-world applications like web-level information retrieval and question answering (QA). We argue that this stems from the nature of MRC datasets: most of these are static environments wherein the supporting documents and all necessary information are fully observed. In this paper, we propose a simple method that reframes existing MRC datasets as interactive, partially observable environments. Specifically, we \"occlude\" the majority of a document's text and add context-sensitive commands that reveal \"glimpses\" of the hidden text to a model. We repurpose SQuAD and NewsQA as an initial case study, and then show how the interactive corpora can be used to train a model that seeks relevant information through sequential decision making. We believe that this setting can contribute in scaling models to web-level QA scenarios."
                    },
                    {
                        "title": "Does Order Matter? An Empirical Study on Generating Multiple Keyphrases as a Sequence",
                        "abstract": "Recently, concatenating multiple keyphrases as a target sequence has been proposed as a new learning paradigm for keyphrase generation. Existing studies concatenate target keyphrases in different orders but no study has examined the effects of ordering on models' behavior. In this paper, we propose several orderings for concatenation and inspect the important factors for training a successful keyphrase generation model. By running comprehensive comparisons, we observe one preferable ordering and summarize a number of empirical findings and challenges, which can shed light on future research on this line of work."
                    },
                    {
                        "title": "An Empirical Study on Neural Keyphrase Generation",
                        "abstract": "Recent years have seen a flourishing of neural keyphrase generation (KPG) works, including the release of several large-scale datasets and a host of new models to tackle them. Model performance on KPG tasks has increased significantly with evolving deep learning research. However, there lacks a comprehensive comparison among different model designs, and a thorough investigation on related factors that may affect a KPG system's generalization performance. In this empirical study, we aim to fill this gap by providing extensive experimental results and analyzing the most crucial factors impacting the generalizability of KPG models. We hope this study can help clarify some of the uncertainties surrounding the KPG task and facilitate future research on this topic."
                    },
                    {
                        "title": "It Takes Two to Tango: Navigating Conceptualizations of NLP Tasks and Measurements of Performance",
                        "abstract": "Progress in NLP is increasingly measured through benchmarks; hence, contextualizing progress requires understanding when and why practitioners may disagree about the validity of benchmarks. We develop a taxonomy of disagreement, drawing on tools from measurement modeling, and distinguish between two types of disagreement: 1) how tasks are conceptualized and 2) how measurements of model performance are operationalized. To provide evidence for our taxonomy, we conduct a meta-analysis of relevant literature to understand how NLP tasks are conceptualized, as well as a survey of practitioners about their impressions of different factors that affect benchmark validity. Our meta-analysis and survey across eight tasks, ranging from coreference resolution to question answering, uncover that tasks are generally not clearly and consistently conceptualized and benchmarks suffer from operationalization disagreements. These findings support our proposed taxonomy of disagreement. Finally, based on our taxonomy, we present a framework for constructing benchmarks and documenting their limitations."
                    },
                    {
                        "title": "Building Dynamic Knowledge Graphs from Text-based Games",
                        "abstract": "We are interested in learning how to update Knowledge Graphs (KG) from text. In this preliminary work, we propose a novel Sequence-to-Sequence (Seq2Seq) architecture to generate elementary KG operations. Furthermore, we introduce a new dataset for KG extraction built upon text-based game transitions (over 300k data points). We conduct experiments and discuss the results."
                    },
                    {
                        "title": "NewsQA: A Machine Comprehension Dataset",
                        "abstract": "We present NewsQA, a challenging machine comprehension dataset of over 100,000 human-generated question-answer pairs. Crowdworkers supply questions and answers based on a set of over 10,000 news articles from CNN, with answers consisting of spans of text from the corresponding articles. We collect this dataset through a four-stage process designed to solicit exploratory questions that require reasoning. A thorough analysis confirms that NewsQA demands abilities beyond simple word matching and recognizing textual entailment. We measure human performance on the dataset and compare it to several strong neural models. The performance gap between humans and machines (0.198 in F1) indicates that significant progress can be made on NewsQA through future research. The dataset is freely available at https://datasets.maluuba.com/NewsQA."
                    },
                    {
                        "title": "Neural Models for Key Phrase Detection and Question Generation",
                        "abstract": "We propose a two-stage neural model to tackle question generation from documents. First, our model estimates the probability that word sequences in a document are ones that a human would pick when selecting candidate answers by training a neural key-phrase extractor on the answers in a question-answering corpus. Predicted key phrases then act as target answers and condition a sequence-to-sequence question-generation model with a copy mechanism. Empirically, our key-phrase extraction model significantly outperforms an entity-tagging baseline and existing rule-based approaches. We further demonstrate that our question generation system formulates fluent, answerable questions from key phrases. This two-stage system could be used to augment or generate reading comprehension datasets, which may be leveraged to improve machine reading systems or in educational settings."
                    },
                    {
                        "title": "Towards Solving Text-based Games by Producing Adaptive Action Spaces",
                        "abstract": "To solve a text-based game, an agent needs to formulate valid text commands for a given context and find the ones that lead to success. Recent attempts at solving text-based games with deep reinforcement learning have focused on the latter, i.e., learning to act optimally when valid actions are known in advance. In this work, we propose to tackle the first task and train a model that generates the set of all valid commands for a given context. We try three generative models on a dataset generated with Textworld. The best model can generate valid commands which were unseen at training and achieve high $F_1$ score on the test set."
                    },
                    {
                        "title": "Bringing Structure into Summaries: a Faceted Summarization Dataset for Long Scientific Documents",
                        "abstract": "Faceted summarization provides briefings of a document from different perspectives. Readers can quickly comprehend the main points of a long document with the help of a structured outline. However, little research has been conducted on this subject, partially due to the lack of large-scale faceted summarization datasets. In this study, we present FacetSum, a faceted summarization benchmark built on Emerald journal articles, covering a diverse range of domains. Different from traditional document-summary pairs, FacetSum provides multiple summaries, each targeted at specific sections of a long document, including the purpose, method, findings, and value. Analyses and empirical results on our dataset reveal the importance of bringing structure into summaries. We believe FacetSum will spur further advances in summarization research and foster the development of NLP systems that can leverage the structured information in both long texts and summaries."
                    },
                    {
                        "title": "Supporting Qualitative Analysis with Large Language Models: Combining Codebook with GPT-3 for Deductive Coding",
                        "abstract": "Qualitative analysis of textual contents unpacks rich and valuable information by assigning labels to the data. However, this process is often labor-intensive, particularly when working with large datasets. While recent AI-based tools demonstrate utility, researchers may not have readily available AI resources and expertise, let alone be challenged by the limited generalizability of those task-specific models. In this study, we explored the use of large language models (LLMs) in supporting deductive coding, a major category of qualitative analysis where researchers use pre-determined codebooks to label the data into a fixed set of codes. Instead of training task-specific models, a pre-trained LLM could be used directly for various tasks without fine-tuning through prompt learning. Using a curiosity-driven questions coding task as a case study, we found, by combining GPT-3 with expert-drafted codebooks, our proposed approach achieved fair to substantial agreements with expert-coded results. We lay out challenges and opportunities in using LLMs to support qualitative coding and beyond."
                    },
                    {
                        "title": "Think Before You Act: Decision Transformers with Working Memory",
                        "abstract": "Decision Transformer-based decision-making agents have shown the ability to generalize across multiple tasks. However, their performance relies on massive data and computation. We argue that this inefficiency stems from the forgetting phenomenon, in which a model memorizes its behaviors in parameters throughout training. As a result, training on a new task may deteriorate the model's performance on previous tasks. In contrast to LLMs' implicit memory mechanism, the human brain utilizes distributed memory storage, which helps manage and organize multiple skills efficiently, mitigating the forgetting phenomenon. Inspired by this, we propose a working memory module to store, blend, and retrieve information for different downstream tasks. Evaluation results show that the proposed method improves training efficiency and generalization in Atari games and Meta-World object manipulation tasks. Moreover, we demonstrate that memory fine-tuning further enhances the adaptability of the proposed architecture."
                    }
                ]
            },
            "180c6670-2a81-4421-8184-8697033dfc3f": {
                "pk": "180c6670-2a81-4421-8184-8697033dfc3f",
                "name": "Marc-Alexandre C\u00f4t\u00e9",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2401.12332": {
        "paper_data": {
            "title": "A Precise Characterization of SGD Stability Using Loss Surface Geometry",
            "url": "http://arxiv.org/abs/2401.12332v1",
            "arxiv_id": "2401.12332",
            "authors": [
                "Gregory Dexter",
                "Borja Ocejo",
                "Sathiya Keerthi",
                "Aman Gupta",
                "Ayan Acharya",
                "Rajiv Khanna"
            ],
            "abstract": "Stochastic Gradient Descent (SGD) stands as a cornerstone optimization algorithm with proven real-world empirical successes but relatively limited theoretical understanding. Recent research has illuminated a key factor contributing to its practical efficacy: the implicit regularization it instigates. Several studies have investigated the linear stability property of SGD in the vicinity of a stationary point as a predictive proxy for sharpness and generalization error in overparameterized neural networks (Wu et al., 2022; Jastrzebski et al., 2019; Cohen et al., 2021). In this paper, we delve deeper into the relationship between linear stability and sharpness. More specifically, we meticulously delineate the necessary and sufficient conditions for linear stability, contingent on hyperparameters of SGD and the sharpness at the optimum. Towards this end, we introduce a novel coherence measure of the loss Hessian that encapsulates pertinent geometric properties of the loss function that are relevant to the linear stability of SGD. It enables us to provide a simplified sufficient condition for identifying linear instability at an optimum. Notably, compared to previous works, our analysis relies on significantly milder assumptions and is applicable for a broader class of loss functions than known before, encompassing not only mean-squared error but also cross-entropy loss.",
            "introduction": "   1 Introduction  Stochastic Gradient Descent (SGD) is a fundamental optimization algorithm widely used in practice. In addition to its computational efficiency, there is irrefutable evidence of its superior generalization performance even on non-convex functions, including neural networks\u00a0(Bottou, 1991). For large over-parameterized neural networks, the number of points to fit is often much less than the number of free parameters in the model. In this case, there is often a high-dimensional manifold of model weights that can perfectly fit the data\u00a0(Cooper, 2021); hence, focusing solely on the ability of an optimizer to minimize the loss function ignores a central part of training such networks. The primary goal in a model is not to achieve high performance on a training data set but rather to achieve strong generalization performance on previously unseen data. Although we currently lack a comprehensive theoretical explanation for the empirical success of SGD in these models, a promising hypothesis suggests that SGD naturally applies a form of implicit regularization (Zhang et\u00a0al., 2017; Neyshabur et\u00a0al., 2015) when multiple optima are present\u00a0(Keskar et\u00a0al., 2017; Liu et\u00a0al., 2020). This phenomenon guides the iterative process towards more favorable optima purely through algorithmic choices.   In order to measure this distinguishing favorability between more and less desirable optima, prior work has proposed the concept of sharpness at a minimum as an indicator of the generalization performance of the trained model. Lower sharpness is often indicative of better generalization performance\u00a0(Hochreiter & Schmidhuber, 1997). There is a wealth of empirical work exploring the relationship between sharpness and generalization performance, particularly in networks trained with SGD, e.g., (Jiang et\u00a0al., 2019; Jastrzebski et\u00a0al., 2019; Andriushchenko et\u00a0al., 2023; Wu et\u00a0al., 2017; Chaudhari et\u00a0al., 2017; Izmailov et\u00a0al., 2018). Furthermore, these ideas have led to new optimizers which deliberately reduce sharpness and are observed to attain improved empirical performance (Behdin et\u00a0al., 2023; Foret et\u00a0al., 2020). Although the connection between sharpness and generalization performance isn\u2019t precise or completely understood, the partial achievements of this theory has inspired several works, including ours, to investigate how SGD implicitly tends to converge to flatter optima.   Sharpness has been defined in several ways in prior literature, but most commonly, the sharpness of a trained neural network at a minimum is the maximum eigenvalue of the Hessian of the loss with respect to weights. Intuitively, one can see that if \ud835\udc30*superscript\ud835\udc30\\mathbf{w}^{*}bold_w start_POSTSUPERSCRIPT * end_POSTSUPERSCRIPT is a stationary point of a smooth function f\u2062(\ud835\udc30)\ud835\udc53\ud835\udc30f(\\mathbf{w})italic_f ( bold_w ) with Hessian \ud835\udc07\u2062(\ud835\udc30)\ud835\udc07\ud835\udc30\\mathbf{H}(\\mathbf{w})bold_H ( bold_w ), then for perturbation \ud835\udc2f\ud835\udc2f\\mathbf{v}bold_v such that \u2016\ud835\udc2f\u20162=\u03f5subscriptnorm\ud835\udc2f2italic-\u03f5\\|\\mathbf{v}\\|_{2}=\\epsilon\u2225 bold_v \u2225 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT = italic_\u03f5, f\u2062(\ud835\udc30*+\ud835\udc2f)<f\u2062(\ud835\udc30*)+O\u2062(\u03f52)\u22c5\u03bb1\u2062(\ud835\udc07\u2062(\ud835\udc30*))\ud835\udc53superscript\ud835\udc30\ud835\udc2f\ud835\udc53superscript\ud835\udc30\u22c5\ud835\udc42superscriptitalic-\u03f52subscript\ud835\udf061\ud835\udc07superscript\ud835\udc30f(\\mathbf{w}^{*}+\\mathbf{v})<f(\\mathbf{w}^{*})+O(\\epsilon^{2})\\cdot\\lambda_{1}% (\\mathbf{H}(\\mathbf{w}^{*}))italic_f ( bold_w start_POSTSUPERSCRIPT * end_POSTSUPERSCRIPT + bold_v ) < italic_f ( bold_w start_POSTSUPERSCRIPT * end_POSTSUPERSCRIPT ) + italic_O ( italic_\u03f5 start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ) \u22c5 italic_\u03bb start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( bold_H ( bold_w start_POSTSUPERSCRIPT * end_POSTSUPERSCRIPT ) ), where \u03bb1\u2062(\u22c5)subscript\ud835\udf061\u22c5\\lambda_{1}(\\cdot)italic_\u03bb start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( \u22c5 ) denotes the maximum eigenvalue. This relation follows from the Taylor expansion of f\u2062(\u22c5)\ud835\udc53\u22c5f(\\cdot)italic_f ( \u22c5 ) around \ud835\udc30*superscript\ud835\udc30\\mathbf{w}^{*}bold_w start_POSTSUPERSCRIPT * end_POSTSUPERSCRIPT, and we see that the sharpness at \ud835\udc30*superscript\ud835\udc30\\mathbf{w}^{*}bold_w start_POSTSUPERSCRIPT * end_POSTSUPERSCRIPT, i.e., \u03bb1\u2062(\ud835\udc07\u2062(\ud835\udc30*))subscript\ud835\udf061\ud835\udc07superscript\ud835\udc30\\lambda_{1}(\\mathbf{H}(\\mathbf{w}^{*}))italic_\u03bb start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( bold_H ( bold_w start_POSTSUPERSCRIPT * end_POSTSUPERSCRIPT ) ), determines how",
            "references": [
                {
                    "title": "The Implicit Regularization of Dynamical Stability in Stochastic Gradient Descent",
                    "abstract": "In this paper, we study the implicit regularization of stochastic gradient descent (SGD) through the lens of {\\em dynamical stability} (Wu et al., 2018). We start by revising existing stability analyses of SGD, showing how the Frobenius norm and trace of Hessian relate to different notions of stability. Notably, if a global minimum is linearly stable for SGD, then the trace of Hessian must be less than or equal to $2/\\eta$, where $\\eta$ denotes the learning rate. By contrast, for gradient descent (GD), the stability imposes a similar constraint but only on the largest eigenvalue of Hessian. We then turn to analyze the generalization properties of these stable minima, focusing specifically on two-layer ReLU networks and diagonal linear networks. Notably, we establish the {\\em equivalence} between these metrics of sharpness and certain parameter norms for the two models, which allows us to show that the stable minima of SGD provably generalize well. By contrast, the stability-induced regularization of GD is provably too weak to ensure satisfactory generalization. This discrepancy provides an explanation of why SGD often generalizes better than GD. Note that the learning rate (LR) plays a pivotal role in the strength of stability-induced regularization. As the LR increases, the regularization effect becomes more pronounced, elucidating why SGD with a larger LR consistently demonstrates superior generalization capabilities. Additionally, numerical experiments are provided to support our theoretical findings."
                },
                {
                    "title": "Type-II Saddles and Probabilistic Stability of Stochastic Gradient Descent",
                    "abstract": "Characterizing and understanding the dynamics of stochastic gradient descent (SGD) around saddle points remains an open problem. We first show that saddle points in neural networks can be divided into two types, among which the Type-II saddles are especially difficult to escape from because the gradient noise vanishes at the saddle. The dynamics of SGD around these saddles are thus to leading order described by a random matrix product process, and it is thus natural to study the dynamics of SGD around these saddles using the notion of probabilistic stability and the related Lyapunov exponent. Theoretically, we link the study of SGD dynamics to well-known concepts in ergodic theory, which we leverage to show that saddle points can be either attractive or repulsive for SGD, and its dynamics can be classified into four different phases, depending on the signal-to-noise ratio in the gradient close to the saddle."
                },
                {
                    "title": "mSAM: Micro-Batch-Averaged Sharpness-Aware Minimization",
                    "abstract": "Modern deep learning models are over-parameterized, where different optima can result in widely varying generalization performance. The Sharpness-Aware Minimization (SAM) technique modifies the fundamental loss function that steers gradient descent methods toward flatter minima, which are believed to exhibit enhanced generalization prowess. Our study delves into a specific variant of SAM known as micro-batch SAM (mSAM). This variation involves aggregating updates derived from adversarial perturbations across multiple shards (micro-batches) of a mini-batch during training. We extend a recently developed and well-studied general framework for flatness analysis to theoretically show that SAM achieves flatter minima than SGD, and mSAM achieves even flatter minima than SAM. We provide a thorough empirical evaluation of various image classification and natural language processing tasks to substantiate this theoretical advancement. We also show that contrary to previous work, mSAM can be implemented in a flexible and parallelizable manner without significantly increasing computational costs. Our implementation of mSAM yields superior generalization performance across a wide range of tasks compared to SAM, further supporting our theoretical framework."
                },
                {
                    "title": "SAM operates far from home: eigenvalue regularization as a dynamical phenomenon",
                    "abstract": "The Sharpness Aware Minimization (SAM) optimization algorithm has been shown to control large eigenvalues of the loss Hessian and provide generalization benefits in a variety of settings. The original motivation for SAM was a modified loss function which penalized sharp minima; subsequent analyses have also focused on the behavior near minima. However, our work reveals that SAM provides a strong regularization of the eigenvalues throughout the learning trajectory. We show that in a simplified setting, SAM dynamically induces a stabilization related to the edge of stability (EOS) phenomenon observed in large learning rate gradient descent. Our theory predicts the largest eigenvalue as a function of the learning rate and SAM radius parameters. Finally, we show that practical models can also exhibit this EOS stabilization, and that understanding SAM must account for these dynamics far away from any minima."
                },
                {
                    "title": "A modern look at the relationship between sharpness and generalization",
                    "abstract": "Sharpness of minima is a promising quantity that can correlate with generalization in deep networks and, when optimized during training, can improve generalization. However, standard sharpness is not invariant under reparametrizations of neural networks, and, to fix this, reparametrization-invariant sharpness definitions have been proposed, most prominently adaptive sharpness (Kwon et al., 2021). But does it really capture generalization in modern practical settings? We comprehensively explore this question in a detailed study of various definitions of adaptive sharpness in settings ranging from training from scratch on ImageNet and CIFAR-10 to fine-tuning CLIP on ImageNet and BERT on MNLI. We focus mostly on transformers for which little is known in terms of sharpness despite their widespread usage. Overall, we observe that sharpness does not correlate well with generalization but rather with some training parameters like the learning rate that can be positively or negatively correlated with generalization depending on the setup. Interestingly, in multiple cases, we observe a consistent negative correlation of sharpness with out-of-distribution error implying that sharper minima can generalize better. Finally, we illustrate on a simple model that the right sharpness measure is highly data-dependent, and that we do not understand well this aspect for realistic data distributions. The code of our experiments is available at https://github.com/tml-epfl/sharpness-vs-generalization."
                },
                {
                    "title": "How Does Sharpness-Aware Minimization Minimize Sharpness?",
                    "abstract": "Sharpness-Aware Minimization (SAM) is a highly effective regularization technique for improving the generalization of deep neural networks for various settings. However, the underlying working of SAM remains elusive because of various intriguing approximations in the theoretical characterizations. SAM intends to penalize a notion of sharpness of the model but implements a computationally efficient variant; moreover, a third notion of sharpness was used for proving generalization guarantees. The subtle differences in these notions of sharpness can indeed lead to significantly different empirical results. This paper rigorously nails down the exact sharpness notion that SAM regularizes and clarifies the underlying mechanism. We also show that the two steps of approximations in the original motivation of SAM individually lead to inaccurate local conclusions, but their combination accidentally reveals the correct effect, when full-batch gradients are applied. Furthermore, we also prove that the stochastic version of SAM in fact regularizes another notion of sharpness, which is most likely to be the preferred notion for practical performance. The key mechanism behind this intriguing phenomenon is the implicit alignment between the gradient and the top eigenvector of Hessian when running SAM."
                },
                {
                    "title": "Rethinking Sharpness-Aware Minimization as Variational Inference",
                    "abstract": "Sharpness-aware minimization (SAM) aims to improve the generalisation of gradient-based learning by seeking out \ufb02at minima. In this work, we establish connections between SAM and Mean-Field Variational Inference (MFVI) of neural network parameters. We show that both these methods have interpretations as opti-mizing notions of \ufb02atness, and when using the reparametrisation trick, they both boil down to calculating the gradient at a perturbed version of the current mean parameter. This thinking motivates our study of algorithms that combine or in-terpolate between SAM and MFVI. We evaluate the proposed variational algorithms on several benchmark datasets, and compare their performance to variants of SAM. Taking a broader perspective, our work suggests that SAM-like updates can be used as a drop-in replacement for the reparametrisation trick."
                },
                {
                    "title": "The Dynamics of Sharpness-Aware Minimization: Bouncing Across Ravines and Drifting Towards Wide Minima",
                    "abstract": "We consider Sharpness-Aware Minimization (SAM), a gradient-based optimization method for deep networks that has exhibited performance improvements on image and language prediction problems. We show that when SAM is applied with a convex quadratic objective, for most random initializations it converges to a cycle that oscillates between either side of the minimum in the direction with the largest curvature, and we provide bounds on the rate of convergence. In the non-quadratic case, we show that such oscillations effectively perform gradient descent, with a smaller step-size, on the spectral norm of the Hessian. In such cases, SAM's update may be regarded as a third derivative -- the derivative of the Hessian in the leading eigenvector direction -- that encourages drift toward wider minima."
                },
                {
                    "title": "Adaptive Gradient Methods at the Edge of Stability",
                    "abstract": "Very little is known about the training dynamics of adaptive gradient methods like Adam in deep learning. In this paper, we shed light on the behavior of these algorithms in the full-batch and sufficiently large batch settings. Specifically, we empirically demonstrate that during full-batch training, the maximum eigenvalue of the preconditioned Hessian typically equilibrates at a certain numerical value -- the stability threshold of a gradient descent algorithm. For Adam with step size $\\eta$ and $\\beta_1 = 0.9$, this stability threshold is $38/\\eta$. Similar effects occur during minibatch training, especially as the batch size grows. Yet, even though adaptive methods train at the ``Adaptive Edge of Stability'' (AEoS), their behavior in this regime differs in a significant way from that of non-adaptive methods at the EoS. Whereas non-adaptive algorithms at the EoS are blocked from entering high-curvature regions of the loss landscape, adaptive gradient methods at the AEoS can keep advancing into high-curvature regions, while adapting the preconditioner to compensate. Our findings can serve as a foundation for the community's future understanding of adaptive gradient methods in deep learning."
                },
                {
                    "title": "The alignment property of SGD noise and how it helps select flat minima: A stability analysis",
                    "abstract": "The phenomenon that stochastic gradient descent (SGD) favors flat minima has played a critical role in understanding the implicit regularization of SGD. In this paper, we provide an explanation of this striking phenomenon by relating the particular noise structure of SGD to its \\emph{linear stability} (Wu et al., 2018). Specifically, we consider training over-parameterized models with square loss. We prove that if a global minimum $\\theta^*$ is linearly stable for SGD, then it must satisfy $\\|H(\\theta^*)\\|_F\\leq O(\\sqrt{B}/\\eta)$, where $\\|H(\\theta^*)\\|_F, B,\\eta$ denote the Frobenius norm of Hessian at $\\theta^*$, batch size, and learning rate, respectively. Otherwise, SGD will escape from that minimum \\emph{exponentially} fast. Hence, for minima accessible to SGD, the sharpness -- as measured by the Frobenius norm of the Hessian -- is bounded \\emph{independently} of the model size and sample size. The key to obtaining these results is exploiting the particular structure of SGD noise: The noise concentrates in sharp directions of local landscape and the magnitude is proportional to loss value. This alignment property of SGD noise provably holds for linear networks and random feature models (RFMs), and is empirically verified for nonlinear networks. Moreover, the validity and practical relevance of our theoretical findings are also justified by extensive experiments on CIFAR-10 dataset."
                },
                {
                    "title": "Surrogate Gap Minimization Improves Sharpness-Aware Training",
                    "abstract": "The recently proposed Sharpness-Aware Minimization (SAM) improves generalization by minimizing a \\textit{perturbed loss} defined as the maximum loss within a neighborhood in the parameter space. However, we show that both sharp and flat minima can have a low perturbed loss, implying that SAM does not always prefer flat minima. Instead, we define a \\textit{surrogate gap}, a measure equivalent to the dominant eigenvalue of Hessian at a local minimum when the radius of the neighborhood (to derive the perturbed loss) is small. The surrogate gap is easy to compute and feasible for direct minimization during training. Based on the above observations, we propose Surrogate \\textbf{G}ap Guided \\textbf{S}harpness-\\textbf{A}ware \\textbf{M}inimization (GSAM), a novel improvement over SAM with negligible computation overhead. Conceptually, GSAM consists of two steps: 1) a gradient descent like SAM to minimize the perturbed loss, and 2) an \\textit{ascent} step in the \\textit{orthogonal} direction (after gradient decomposition) to minimize the surrogate gap and yet not affect the perturbed loss. GSAM seeks a region with both small loss (by step 1) and low sharpness (by step 2), giving rise to a model with high generalization capabilities. Theoretically, we show the convergence of GSAM and provably better generalization than SAM. Empirically, GSAM consistently improves generalization (e.g., +3.2\\% over SAM and +5.4\\% over AdamW on ImageNet top-1 accuracy for ViT-B/32). Code is released at \\url{ https://sites.google.com/view/gsam-iclr22/home}."
                },
                {
                    "title": "Towards Efficient and Scalable Sharpness-Aware Minimization",
                    "abstract": "Recently, Sharpness-Aware Minimization (SAM), which connects the geometry of the loss landscape and generalization, has demonstrated a significant performance boost on training large-scale models such as vision transformers. However, the update rule of SAM requires two sequential (non-parallelizable) gradient computations at each step, which can double the computational overhead. In this paper, we propose a novel algorithm LookSAM - that only periodically calculates the inner gradient ascent, to significantly reduce the additional training cost of SAM. The empirical results illustrate that LookSAM achieves similar accuracy gains to SAM while being tremendously faster - it enjoys comparable computational complexity with first-order optimizers such as SGD or Adam. To further evaluate the performance and scalability of LookSAM, we incorporate a layer-wise modification and perform experiments in the large-batch training scenario, which is more prone to converge to sharp local minima. Equipped with the proposed algorithms, we are the first to successfully scale up the batch size when training Vision Transformers (ViTs). With a 64k batch size, we are able to train ViTs from scratch in minutes while maintaining competitive performance. The code is available here: https://github.com/yong-6/LookSAM"
                },
                {
                    "title": "On the Power-Law Hessian Spectrums in Deep Learning",
                    "abstract": "It is well-known that the Hessian of deep loss landscape matters to optimization, generalization, and even robustness of deep learning. Recent works empirically discovered that the Hessian spectrum in deep learning has a two-component structure that consists of a small number of large eigenvalues and a large number of nearly-zero eigenvalues. However, the theoretical mechanism or the mathematical behind the Hessian spectrum is still largely under-explored. To the best of our knowledge, we are the first to demonstrate that the Hessian spectrums of well-trained deep neural networks exhibit simple power-law structures. Inspired by the statistical physical theories and the spectral analysis of natural proteins, we provide a maximum-entropy theoretical interpretation for explaining why the power-law structure exist and suggest a spectral parallel between protein evolution and training of deep neural networks. By conducing extensive experiments, we further use the power-law spectral framework as a useful tool to explore multiple novel behaviors of deep learning."
                },
                {
                    "title": "On Linear Stability of SGD and Input-Smoothness of Neural Networks",
                    "abstract": "The multiplicative structure of parameters and input data in the first layer of neural networks is explored to build connection between the landscape of the loss function with respect to parameters and the landscape of the model function with respect to input data. By this connection, it is shown that flat minima regularize the gradient of the model function, which explains the good generalization performance of flat minima. Then, we go beyond the flatness and consider high-order moments of the gradient noise, and show that Stochastic Gradient Descent (SGD) tends to impose constraints on these moments by a linear stability analysis of SGD around global minima. Together with the multiplicative structure, we identify the Sobolev regularization effect of SGD, i.e. SGD regularizes the Sobolev seminorms of the model function with respect to the input data. Finally, bounds for generalization error and adversarial robustness are provided for solutions found by SGD under assumptions of the data distribution."
                },
                {
                    "title": "Gradient Descent on Neural Networks Typically Occurs at the Edge of Stability",
                    "abstract": "We empirically demonstrate that full-batch gradient descent on neural network training objectives typically operates in a regime we call the Edge of Stability. In this regime, the maximum eigenvalue of the training loss Hessian hovers just above the numerical value $2 / \\text{(step size)}$, and the training loss behaves non-monotonically over short timescales, yet consistently decreases over long timescales. Since this behavior is inconsistent with several widespread presumptions in the field of optimization, our findings raise questions as to whether these presumptions are relevant to neural network training. We hope that our findings will inspire future efforts aimed at rigorously understanding optimization at the Edge of Stability. Code is available at https://github.com/locuslab/edge-of-stability."
                },
                {
                    "title": "ASAM: Adaptive Sharpness-Aware Minimization for Scale-Invariant Learning of Deep Neural Networks",
                    "abstract": "Recently, learning algorithms motivated from sharpness of loss surface as an effective measure of generalization gap have shown state-of-the-art performances. Nevertheless, sharpness defined in a rigid region with a fixed radius, has a drawback in sensitivity to parameter re-scaling which leaves the loss unaffected, leading to weakening of the connection between sharpness and generalization gap. In this paper, we introduce the concept of adaptive sharpness which is scale-invariant and propose the corresponding generalization bound. We suggest a novel learning method, adaptive sharpness-aware minimization (ASAM), utilizing the proposed generalization bound. Experimental results in various benchmark datasets show that ASAM contributes to significant improvement of model generalization performance."
                },
                {
                    "title": "Sharpness-Aware Minimization for Efficiently Improving Generalization",
                    "abstract": "In today's heavily overparameterized models, the value of the training loss provides few guarantees on model generalization ability. Indeed, optimizing only the training loss value, as is commonly done, can easily lead to suboptimal model quality. Motivated by the connection between geometry of the loss landscape and generalization---including a generalization bound that we prove here---we introduce a novel, effective procedure for instead simultaneously minimizing loss value and loss sharpness. In particular, our procedure, Sharpness-Aware Minimization (SAM), seeks parameters that lie in neighborhoods having uniformly low loss; this formulation results in a min-max optimization problem on which gradient descent can be performed efficiently. We present empirical results showing that SAM improves model generalization across a variety of benchmark datasets (e.g., CIFAR-{10, 100}, ImageNet, finetuning tasks) and models, yielding novel state-of-the-art performance for several. Additionally, we find that SAM natively provides robustness to label noise on par with that provided by state-of-the-art procedures that specifically target learning with noisy labels."
                },
                {
                    "title": "The Break-Even Point on Optimization Trajectories of Deep Neural Networks",
                    "abstract": "The early phase of training of deep neural networks is critical for their final performance. In this work, we study how the hyperparameters of stochastic gradient descent (SGD) used in the early phase of training affect the rest of the optimization trajectory. We argue for the existence of the \"break-even\" point on this trajectory, beyond which the curvature of the loss surface and noise in the gradient are implicitly regularized by SGD. In particular, we demonstrate on multiple classification tasks that using a large learning rate in the initial phase of training reduces the variance of the gradient, and improves the conditioning of the covariance of gradients. These effects are beneficial from the optimization perspective and become visible after the break-even point. Complementing prior work, we also show that using a low learning rate results in bad conditioning of the loss surface even for a neural network with batch normalization layers. In short, our work shows that key properties of the loss surface are strongly influenced by SGD in the early phase of training. We argue that studying the impact of the identified effects on generalization is a promising future direction."
                },
                {
                    "title": "A Diffusion Theory For Deep Learning Dynamics: Stochastic Gradient Descent Exponentially Favors Flat Minima",
                    "abstract": "Stochastic Gradient Descent (SGD) and its variants are mainstream methods for training deep networks in practice. SGD is known to find a flat minimum that often generalizes well. However, it is mathematically unclear how deep learning can select a flat minimum among so many minima. To answer the question quantitatively, we develop a density diffusion theory (DDT) to reveal how minima selection quantitatively depends on the minima sharpness and the hyperparameters. To the best of our knowledge, we are the first to theoretically and empirically prove that, benefited from the Hessian-dependent covariance of stochastic gradient noise, SGD favors flat minima exponentially more than sharp minima, while Gradient Descent (GD) with injected white noise favors flat minima only polynomially more than sharp minima. We also reveal that either a small learning rate or large-batch training requires exponentially many iterations to escape from minima in terms of the ratio of the batch size and learning rate. Thus, large-batch training cannot search flat minima efficiently in a realistic computational time."
                },
                {
                    "title": "Fantastic Generalization Measures and Where to Find Them",
                    "abstract": "Generalization of deep networks has been of great interest in recent years, resulting in a number of theoretically and empirically motivated complexity measures. However, most papers proposing such measures study only a small set of models, leaving open the question of whether the conclusion drawn from those experiments would remain valid in other settings. We present the first large scale study of generalization in deep networks. We investigate more then 40 complexity measures taken from both theoretical bounds and empirical studies. We train over 10,000 convolutional networks by systematically varying commonly used hyperparameters. Hoping to uncover potentially causal relationships between each measure and generalization, we analyze carefully controlled experiments and show surprising failures of some measures as well as promising measures for further research."
                },
                {
                    "title": "Bad Global Minima Exist and SGD Can Reach Them",
                    "abstract": "Several recent works have aimed to explain why severely overparameterized models, generalize well when trained by Stochastic Gradient Descent (SGD). The emergent consensus explanation has two parts: the first is that there are \"no bad local minima\", while the second is that SGD performs implicit regularization by having a bias towards low complexity models. We revisit both of these ideas in the context of image classification with common deep neural network architectures. Our first finding is that there exist bad global minima, i.e., models that fit the training set perfectly, yet have poor generalization. Our second finding is that given only unlabeled training data, we can easily construct initializations that will cause SGD to quickly converge to such bad global minima. For example, on CIFAR, CINIC10, and (Restricted) ImageNet, this can be achieved by starting SGD at a model derived by fitting random labels on the training data: while subsequent SGD training (with the correct labels) will reach zero training error, the resulting model will exhibit a test accuracy degradation of up to 40% compared to training from a random initialization. Finally, we show that regularization seems to provide SGD with an escape route: once heuristics such as data augmentation are used, starting from a complex model (adversarial initialization) has no effect on the test accuracy."
                },
                {
                    "title": "A Convergence Theory for Deep Learning via Over-Parameterization",
                    "abstract": "Deep neural networks (DNNs) have demonstrated dominating performance in many fields; since AlexNet, networks used in practice are going wider and deeper. On the theoretical side, a long line of works has been focusing on training neural networks with one hidden layer. The theory of multi-layer networks remains largely unsettled. \nIn this work, we prove why stochastic gradient descent (SGD) can find $\\textit{global minima}$ on the training objective of DNNs in $\\textit{polynomial time}$. We only make two assumptions: the inputs are non-degenerate and the network is over-parameterized. The latter means the network width is sufficiently large: $\\textit{polynomial}$ in $L$, the number of layers and in $n$, the number of samples. \nOur key technique is to derive that, in a sufficiently large neighborhood of the random initialization, the optimization landscape is almost-convex and semi-smooth even with ReLU activations. This implies an equivalence between over-parameterized neural networks and neural tangent kernel (NTK) in the finite (and polynomial) width setting. \nAs concrete examples, starting from randomly initialized weights, we prove that SGD can attain 100% training accuracy in classification tasks, or minimize regression loss in linear convergence speed, with running time polynomial in $n,L$. Our theory applies to the widely-used but non-smooth ReLU activation, and to any smooth and possibly non-convex loss functions. In terms of network architectures, our theory at least applies to fully-connected neural networks, convolutional neural networks (CNN), and residual neural networks (ResNet)."
                },
                {
                    "title": "Averaging Weights Leads to Wider Optima and Better Generalization",
                    "abstract": "Deep neural networks are typically trained by optimizing a loss function with an SGD variant, in conjunction with a decaying learning rate, until convergence. We show that simple averaging of multiple points along the trajectory of SGD, with a cyclical or constant learning rate, leads to better generalization than conventional training. We also show that this Stochastic Weight Averaging (SWA) procedure finds much flatter solutions than SGD, and approximates the recent Fast Geometric Ensembling (FGE) approach with a single model. Using SWA we achieve notable improvement in test accuracy over conventional SGD training on a range of state-of-the-art residual networks, PyramidNets, DenseNets, and Shake-Shake networks on CIFAR-10, CIFAR-100, and ImageNet. In short, SWA is extremely easy to implement, improves generalization, and has almost no computational overhead."
                },
                {
                    "title": "Towards Understanding Generalization of Deep Learning: Perspective of Loss Landscapes",
                    "abstract": "It is widely observed that deep learning models with learned parameters generalize well, even with much more model parameters than the number of training samples. We systematically investigate the underlying reasons why deep neural networks often generalize well, and reveal the difference between the minima (with the same training error) that generalize well and those they don't. We show that it is the characteristics the landscape of the loss function that explains the good generalization capability. For the landscape of loss function for deep networks, the volume of basin of attraction of good minima dominates over that of poor minima, which guarantees optimization methods with random initialization to converge to good minima. We theoretically justify our findings through analyzing 2-layer neural networks; and show that the low-complexity solutions have a small norm of Hessian matrix with respect to model parameters. For deeper networks, extensive numerical evidence helps to support our arguments."
                },
                {
                    "title": "Entropy-SGD: biasing gradient descent into wide valleys",
                    "abstract": "This paper proposes a new optimization algorithm called Entropy-SGD for training deep neural networks that is motivated by the local geometry of the energy landscape. Local extrema with low generalization error have a large proportion of almost-zero eigenvalues in the Hessian with very few positive or negative eigenvalues. We leverage upon this observation to construct a local-entropy-based objective function that favors well-generalizable solutions lying in large flat regions of the energy landscape, while avoiding poorly-generalizable solutions located in the sharp valleys. Conceptually, our algorithm resembles two nested loops of SGD where we use Langevin dynamics in the inner loop to compute the gradient of the local entropy before each update of the weights. We show that the new objective has a smoother energy landscape and show improved generalization over SGD using uniform stability, under certain assumptions. Our experiments on convolutional and recurrent networks demonstrate that Entropy-SGD compares favorably to state-of-the-art techniques in terms of generalization error and training time."
                },
                {
                    "title": "Understanding deep learning requires rethinking generalization",
                    "abstract": "Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family, or to the regularization techniques used during training. \nThrough extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization, and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. \nWe interpret our experimental findings by comparison with traditional models."
                },
                {
                    "title": "On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima",
                    "abstract": "The stochastic gradient descent (SGD) method and its variants are algorithms of choice for many Deep Learning tasks. These methods operate in a small-batch regime wherein a fraction of the training data, say $32$-$512$ data points, is sampled to compute an approximation to the gradient. It has been observed in practice that when using a larger batch there is a degradation in the quality of the model, as measured by its ability to generalize. We investigate the cause for this generalization drop in the large-batch regime and present numerical evidence that supports the view that large-batch methods tend to converge to sharp minimizers of the training and testing functions - and as is well known, sharp minima lead to poorer generalization. In contrast, small-batch methods consistently converge to flat minimizers, and our experiments support a commonly held view that this is due to the inherent noise in the gradient estimation. We discuss several strategies to attempt to help large-batch methods eliminate this generalization gap."
                },
                {
                    "title": "Rethinking the Inception Architecture for Computer Vision",
                    "abstract": "Convolutional networks are at the core of most state of-the-art computer vision solutions for a wide variety of tasks. Since 2014 very deep convolutional networks started to become mainstream, yielding substantial gains in various benchmarks. Although increased model size and computational cost tend to translate to immediate quality gains for most tasks (as long as enough labeled data is provided for training), computational efficiency and low parameter count are still enabling factors for various use cases such as mobile vision and big-data scenarios. Here we are exploring ways to scale up networks in ways that aim at utilizing the added computation as efficiently as possible by suitably factorized convolutions and aggressive regularization. We benchmark our methods on the ILSVRC 2012 classification challenge validation set demonstrate substantial gains over the state of the art: 21:2% top-1 and 5:6% top-5 error for single frame evaluation using a network with a computational cost of 5 billion multiply-adds per inference and with using less than 25 million parameters. With an ensemble of 4 models and multi-crop evaluation, we report 3:5% top-5 error and 17:3% top-1 error on the validation set and 3:6% top-5 error on the official test set."
                },
                {
                    "title": "Stochastic Modified Equations and Adaptive Stochastic Gradient Algorithms",
                    "abstract": "We develop the method of stochastic modified equations (SME), in which stochastic gradient algorithms are approximated in the weak sense by continuous-time stochastic differential equations. We exploit the continuous formulation together with optimal control theory to derive novel adaptive hyper-parameter adjustment policies. Our algorithms have competitive performance with the added benefit of being robust to varying models and datasets. This provides a general methodology for the analysis and design of stochastic gradient algorithms."
                },
                {
                    "title": "In Search of the Real Inductive Bias: On the Role of Implicit Regularization in Deep Learning",
                    "abstract": "We present experiments demonstrating that some other form of capacity control, different from network size, plays a central role in learning multilayer feed-forward networks. We argue, partially through analogy to matrix factorization, that this is an inductive bias that can help shed light on deep learning."
                },
                {
                    "title": "The conference paper",
                    "abstract": "Purpose - A satire on conference paper presentations. Design/methodology/approach - A poetic critique of conference presenters and their occasional failure to engage with their audience. Findings - What may be fascinating and of assumed and obvious importance to a research presenter may be quite obscure to their audience, who nonetheless participate in the ritual of conference attendance and polite \u201clistening\u201d. Research limitations/implications - The reader may discover even more meaning in this poem than its original author envisaged. Originality/value - A poignant reminder of the perils of many conference presentations."
                },
                {
                    "title": "How SGD Selects the Global Minima in Over-parameterized Learning : A Dynamical Stability Perspective",
                    "abstract": "The question of which global minima are accessible by a stochastic gradient decent (SGD) algorithm with specific learning rate and batch size is studied from the perspective of dynamical stability. The concept of non-uniformity is introduced, which, together with sharpness, characterizes the stability property of a global minimum and hence the accessibility of a particular SGD algorithm to that global minimum. In particular, this analysis shows that learning rate and batch size play different roles in minima selection. Extensive empirical results seem to correlate well with the theoretical findings and provide further support to these claims."
                },
                {
                    "title": "Flat Minima",
                    "abstract": "We present a new algorithm for finding low-complexity neural networks with high generalization capability. The algorithm searches for a flat minimum of the error function. A flat minimum is a large connected region in weight space where the error remains approximately constant. An MDL-based, Bayesian argument suggests that flat minima correspond to simple networks and low expected overfitting. The argument is based on a Gibbs algorithm variant and a novel way of splitting generalization error into underfitting and overfitting error. Unlike many previous approaches, ours does not require gaussian assumptions and does not depend on a good weight prior. Instead we have a prior over input output functions, thus taking into account net architecture and training set. Although our algorithm requires the computation of second-order derivatives, it has backpropagation's order of complexity. Automatically, it effectively prunes units, weights, and input lines. Various experiments with feedforward and recurrent nets are described. In an application to stock market prediction, flat minimum search outperforms conventional backprop, weight decay, and optimal brain surgeon/optimal brain damage."
                },
                {
                    "title": "Stochastic Gradient Learning in Neural Networks",
                    "abstract": "Many connectionist learning algorithms consists of minimizing a cost of the form C(w) = E(J(z,w)) = J(z,w)dP(z) where dP is an unknown probability distribution that characterizes the problem to learn, and J, the loss function, defines the learning system itself. This popular statistical formulation has led to many theoretical results. The minimization of such a cost may be achieved with a stochastic gradient descent algorithm, e.g.: wt+1 = wt \u2212 \u025bt\u2207wJ(z,wt) With some restrictions on J and C, this algorithm converges, even if J is non differentiable on a set of measure 0. Links with simulated annealing are depicted."
                }
            ],
            "categories": [
                "cs.LG",
                "math.OC"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow does Stochastic Gradient Descent (SGD) implicitly guide the optimization process towards flatter optima in neural networks, and what is the relationship between sharpness at a minimum and generalization performance?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it could provide a deeper theoretical understanding of why SGD performs well in practice, particularly in the context of over-parameterized models. By elucidating the connection between sharpness and generalization, this research could lead to the development of more effective optimization algorithms that enhance model performance on unseen data. Furthermore, it could inspire new methodologies for training neural networks, ultimately advancing knowledge in machine learning and leading to practical applications in various domains, such as computer vision and natural language processing.\n\n### [Question 3] - Why is it hard?\nThe challenges in addressing this problem stem from the complex nature of high-dimensional optimization landscapes in neural networks, where multiple local minima exist. Naive approaches may fail because they often overlook the nuanced relationship between sharpness and generalization, focusing solely on loss minimization. Additionally, the lack of a comprehensive theoretical framework to explain SGD's behavior complicates the analysis. Technical obstacles include accurately measuring sharpness and establishing a clear causal link between sharpness and generalization performance, which requires sophisticated mathematical tools and empirical validation.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on empirical observations rather than providing a theoretical foundation for the relationship between sharpness and generalization. Existing solutions often lack a unified approach to defining and measuring sharpness, leading to inconsistencies in findings. Barriers such as the complexity of high-dimensional optimization and the difficulty in isolating the effects of SGD's implicit regularization have hindered progress. Our approach aims to bridge these gaps by offering a clearer definition of sharpness and a systematic investigation of its implications for generalization, thereby improving upon prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves analyzing the sharpness of minima found by SGD in neural networks by calculating the maximum eigenvalue of the Hessian of the loss function with respect to the model weights. We will utilize a diverse set of datasets, including benchmark image classification tasks, to evaluate the generalization performance of models trained with SGD. The primary metric for assessment will be the sharpness at the minima, correlated with the models' performance on unseen data. We expect"
            }
        },
        "author_data": {
            "dacaca0d-66ce-4b27-af3d-1923b09bd62d": {
                "pk": "dacaca0d-66ce-4b27-af3d-1923b09bd62d",
                "name": "Gregory Dexter",
                "collaborators": [
                    "Kayhan Behdin",
                    "Qingquan Song",
                    "Aman Gupta",
                    "S. Keerthi",
                    "A. Acharya",
                    "Borja Ocejo",
                    "Rajiv Khanna",
                    "D. Durfee",
                    "R. Mazumder"
                ],
                "domain": [
                    "Deep Learning",
                    "Optimization",
                    "Natural Language Processing",
                    "Image Classification"
                ],
                "publications": [
                    {
                        "title": "mSAM: Micro-Batch-Averaged Sharpness-Aware Minimization",
                        "abstract": "Modern deep learning models are over-parameterized, where different optima can result in widely varying generalization performance. The Sharpness-Aware Minimization (SAM) technique modifies the fundamental loss function that steers gradient descent methods toward flatter minima, which are believed to exhibit enhanced generalization prowess. Our study delves into a specific variant of SAM known as micro-batch SAM (mSAM). This variation involves aggregating updates derived from adversarial perturbations across multiple shards (micro-batches) of a mini-batch during training. We extend a recently developed and well-studied general framework for flatness analysis to theoretically show that SAM achieves flatter minima than SGD, and mSAM achieves even flatter minima than SAM. We provide a thorough empirical evaluation of various image classification and natural language processing tasks to substantiate this theoretical advancement. We also show that contrary to previous work, mSAM can be implemented in a flexible and parallelizable manner without significantly increasing computational costs. Our implementation of mSAM yields superior generalization performance across a wide range of tasks compared to SAM, further supporting our theoretical framework."
                    }
                ]
            },
            "3e337508-9f47-4994-8c7f-a87371dd6fcb": {
                "pk": "3e337508-9f47-4994-8c7f-a87371dd6fcb",
                "name": "Borja Ocejo",
                "collaborators": [
                    "Aman Gupta",
                    "A. Acharya",
                    "Kinjal Basu",
                    "Siyuan Gao",
                    "Ankan Saha",
                    "Keerthi Selvaraj",
                    "Parag Agrawal",
                    "S. Keerthi",
                    "R. Mazumder",
                    "Rahul Mazumdar"
                ],
                "domain": [
                    "Recommender Systems",
                    "Linear Programming",
                    "Optimization",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Promoting Inactive Members in Edge-Building Marketplace",
                        "abstract": "Social networks are platforms where content creators and consumers share and consume content. The edge recommendation system, which determines who a member should connect with, significantly impacts the reach and engagement of the audience on such networks. This paper emphasizes improving the experience of inactive members (IMs) who do not have a large connection network by recommending better connections. To that end, we propose a multi-objective linear optimization framework and solve it using accelerated gradient descent. We report our findings regarding key business metrics related to user engagement on LinkedIn, a professional network with over 850 million members."
                    },
                    {
                        "title": "Practical Design of Performant Recommender Systems using Large-scale Linear Programming-based Global Inference",
                        "abstract": "Several key problems in web-scale recommender systems, such as optimal matching and allocation, can be formulated as large-scale linear programs (LPs) [4, 1]. These LPs take predictions from ML models such as probabilities of click, like, etc. as inputs and optimize recommendations made to users. In recent years, there has been an explosion in the research and development of large-scale recommender systems, but effective optimization of business objectives using the output of those systems remains a challenge. Although LPs can help optimize such business objectives, and algorithms for solving LPs have existed since the 1950s [5, 8], generic LP solvers cannot handle the scale of these problems. At LinkedIn, we have developed algorithms that can solve LPs of various forms with trillions of variables in a Spark-based library called \"DuaLip\" [7], a novel distributed solver that solves a perturbation of the LP problem at scale via gradient-based algorithms on the smooth dual of the perturbed LP. DuaLip has been deployed in production at LinkedIn and powers several very large-scale recommender systems. DuaLip is open-sourced and extensible in terms of features and algorithms. In this first-of-its-kind tutorial, we will motivate the application of LPs to improve recommender systems, cover the theory of key LP algorithms [8, 6], and introduce DuaLip (https://github.com/linkedin/DuaLip), a highly performant Spark-based library that solves extreme-scale LPs for a large variety of recommender system problems. We will describe practical successes of large-scale LP in the industry [3, 2, 9] followed by a hands-on exercise to run DuaLip."
                    },
                    {
                        "title": "Optimizing for Member Value in an Edge Building Marketplace",
                        "abstract": "Social networks are prosperous marketplaces where creators and consumers congregate to share and consume various content. In general, products that rank content for distribution (such as newsfeeds, stories, and notifications) and are related to edge recommendations (such as connect to members, follow celebrities or groups or hashtags) optimize the experience of active users. Typically, such users generate ample interaction data amenable to accurate model training and prediction. In contrast, we prioritize enhancing the experience of inactive members (IMs) who do not have a rich connection network. We formulate strategies for recommending superior edges to help members grow their connection network. Adapting the recommendations provides enormous value to the IMs and can significantly influence their future behaviour and engagement with the ecosystem. To that end, we propose a general and scalable multi-objective optimization (MOO) framework to provide more value to IMs as invitation recipients on LinkedIn, a professional network with over 900M members. To deal with the enormous scale, we formulate the problem as a massive constrained linear optimization involving billions of variables and millions of constraints and efficiently solve it using accelerated gradient descent,making this the largest deployment of LP-based recommender systems worldwide. Furthermore, the proposed MOO paradigm can solve the general problem of matching different types of entities in an m-sided marketplace. Finally, we discuss the challenges and benefits of implementing and ramping our method in production at scale at LinkedIn and report our findings about the core business metrics related to users' engagement and network health."
                    },
                    {
                        "title": "mSAM: Micro-Batch-Averaged Sharpness-Aware Minimization",
                        "abstract": "Modern deep learning models are over-parameterized, where different optima can result in widely varying generalization performance. The Sharpness-Aware Minimization (SAM) technique modifies the fundamental loss function that steers gradient descent methods toward flatter minima, which are believed to exhibit enhanced generalization prowess. Our study delves into a specific variant of SAM known as micro-batch SAM (mSAM). This variation involves aggregating updates derived from adversarial perturbations across multiple shards (micro-batches) of a mini-batch during training. We extend a recently developed and well-studied general framework for flatness analysis to theoretically show that SAM achieves flatter minima than SGD, and mSAM achieves even flatter minima than SAM. We provide a thorough empirical evaluation of various image classification and natural language processing tasks to substantiate this theoretical advancement. We also show that contrary to previous work, mSAM can be implemented in a flexible and parallelizable manner without significantly increasing computational costs. Our implementation of mSAM yields superior generalization performance across a wide range of tasks compared to SAM, further supporting our theoretical framework."
                    }
                ]
            },
            "d57827bd-cef4-4413-a1f4-e154b5a1f6c6": {
                "pk": "d57827bd-cef4-4413-a1f4-e154b5a1f6c6",
                "name": "Sathiya Keerthi",
                "collaborators": [
                    "Aman Gupta",
                    "Yunbo Ouyang",
                    "Fedor Borisyuk",
                    "Qingquan Song",
                    "Siyu Zhu",
                    "Ruofan Wang",
                    "Prakruthi Prabhakar",
                    "Gaurav Srivastava",
                    "Tianqi Wang",
                    "Zeinab S. Jalali"
                ],
                "domain": [
                    "Recommender Systems",
                    "Deep Learning",
                    "Model Compression",
                    "Meta-Learning"
                ],
                "publications": [
                    {
                        "title": "LiMAML: Personalization of Deep Recommender Models via Meta Learning",
                        "abstract": "In the realm of recommender systems, the ubiquitous adoption of deep neural networks has emerged as a dominant paradigm for modeling diverse business objectives. As user bases continue to expand, the necessity of personalization and frequent model updates have assumed paramount significance to ensure the delivery of relevant and refreshed experiences to a diverse array of members. In this work, we introduce an innovative meta-learning solution tailored to the personalization of models for individual members and other entities, coupled with the frequent updates based on the latest user interaction signals. Specifically, we leverage the Model-Agnostic Meta Learning (MAML) algorithm to adapt per-task sub-networks using recent user interaction data. Given the near infeasibility of productionizing original MAML-based models in online recommendation systems, we propose an efficient strategy to operationalize meta-learned sub-networks in production, which involves transforming them into fixed-sized vectors, termed meta embeddings, thereby enabling the seamless deployment of models with hundreds of billions of parameters for online serving. Through extensive experimentation on production data drawn from various applications at LinkedIn, we demonstrate that the proposed solution consistently outperforms the baseline models of those applications, including strong baselines such as using wide-and-deep ID based personalization approach. Our approach has enabled the deployment of a range of highly personalized AI models across diverse LinkedIn applications, leading to substantial improvements in business metrics as well as refreshed experience for our members."
                    },
                    {
                        "title": "LiRank: Industrial Large Scale Ranking Models at LinkedIn",
                        "abstract": "We present LiRank, a large-scale ranking framework at LinkedIn that brings to production state-of-the-art modeling architectures and optimization methods. We unveil several modeling improvements, including Residual DCN, which adds attention and residual connections to the famous DCNv2 architecture. We share insights into combining and tuning SOTA architectures to create a unified model, including Dense Gating, Transformers and Residual DCN. We also propose novel techniques for calibration and describe how we productionalized deep learning based explore/exploit methods. To enable effective, production-grade serving of large ranking models, we detail how to train and compress models using quantization and vocabulary compression. We provide details about the deployment setup for large-scale use cases of Feed ranking, Jobs Recommendations, and Ads click-through rate (CTR) prediction. We summarize our learnings from various A/B tests by elucidating the most effective technical approaches. These ideas have contributed to relative metrics improvements across the board at LinkedIn: +0.5% member sessions in the Feed, +1.76% qualified job applications for Jobs search and recommendations, and +4.3% for Ads CTR. We hope this work can provide practical insights and solutions for practitioners interested in leveraging large-scale deep ranking systems."
                    },
                    {
                        "title": "QuantEase: Optimization-based Quantization for Language Models - An Efficient and Intuitive Algorithm",
                        "abstract": "With the rising popularity of Large Language Models (LLMs), there has been an increasing interest in compression techniques that enable their efficient deployment. This study focuses on the Post-Training Quantization (PTQ) of LLMs. Drawing from recent advances, our work introduces QuantEase, a layer-wise quantization framework where individual layers undergo separate quantization. The problem is framed as a discrete-structured non-convex optimization, prompting the development of algorithms rooted in Coordinate Descent (CD) techniques. These CD-based methods provide high-quality solutions to the complex non-convex layer-wise quantization problems. Notably, our CD-based approach features straightforward updates, relying solely on matrix and vector operations, circumventing the need for matrix inversion or decomposition. We also explore an outlier-aware variant of our approach, allowing for retaining significant weights (outliers) with complete precision. Our proposal attains state-of-the-art performance in terms of perplexity and zero-shot accuracy in empirical evaluations across various LLMs and datasets, with relative improvements up to 15% over methods such as GPTQ. Leveraging careful linear algebra optimizations, QuantEase can quantize models like Falcon-180B on a single NVIDIA A100 GPU in $\\sim$3 hours. Particularly noteworthy is our outlier-aware algorithm's capability to achieve near or sub-3-bit quantization of LLMs with an acceptable drop in accuracy, obviating the need for non-uniform quantization or grouping techniques, improving upon methods such as SpQR by up to two times in terms of perplexity."
                    },
                    {
                        "title": "Automatic web-scale information extraction",
                        "abstract": "In this demonstration, we showcase the technologies that we are building at Yahoo! for Web-scale Information Extraction. Given any new Website, containing semi-structured information about a pre-specified set of schemas, we show how to populate objects in the corresponding schema by automatically extracting information from the Website."
                    },
                    {
                        "title": "A fast procedure for computing the distance between complex objects in three-dimensional space",
                        "abstract": "An algorithm for computing the Euclidean distance between a pair of convex sets in R/sup m/ is described. Extensive numerical experience with a broad family of polytopes in R/sup 3/ shows that the computational cost is approximately linear in the total number of vertices specifying the two polytopes. The algorithm has special features which makes its application in a variety of robotics problems attractive. These features are discussed and an example of collision detection is given. >"
                    },
                    {
                        "title": "Adam vs. SGD: Closing the generalization gap on image classi\ufb01cation",
                        "abstract": "Adam is an adaptive deep neural network training optimizer that has been widely used across a variety of applications. However, on image classi\ufb01cation problems, its generalization performance is signi\ufb01cantly worse than stochastic gradient descent (SGD). By tuning several inner hyperparameters of Adam, it is possible to lift the performance of Adam and close this gap; but this makes the use of Adam computationally expensive. In this paper, we use a new training approach based on layer-wise weight normalization (LAWN) to solidly improve Adam\u2019s performance and close the gap with SGD. LAWN also helps reduce the impact of batch size on Adam\u2019s performance. With speed in tact and performance vastly improved, the Adam-LAWN combination becomes an attractive optimizer for use in image classi\ufb01cation."
                    }
                ]
            },
            "6056472e-12bb-4953-abd4-90291ed66158": {
                "pk": "6056472e-12bb-4953-abd4-90291ed66158",
                "name": "Aman Gupta",
                "collaborators": [
                    "Kinjal Basu",
                    "Borja Ocejo",
                    "Keerthi Selvaraj",
                    "A. Acharya",
                    "Siyuan Gao",
                    "Ankan Saha",
                    "Parag Agrawal",
                    "S. Keerthi",
                    "Miao Cheng",
                    "R. Mazumder"
                ],
                "domain": [
                    "Recommender Systems",
                    "Linear Programming",
                    "Optimization",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Promoting Inactive Members in Edge-Building Marketplace",
                        "abstract": "Social networks are platforms where content creators and consumers share and consume content. The edge recommendation system, which determines who a member should connect with, significantly impacts the reach and engagement of the audience on such networks. This paper emphasizes improving the experience of inactive members (IMs) who do not have a large connection network by recommending better connections. To that end, we propose a multi-objective linear optimization framework and solve it using accelerated gradient descent. We report our findings regarding key business metrics related to user engagement on LinkedIn, a professional network with over 850 million members."
                    },
                    {
                        "title": "Practical Design of Performant Recommender Systems using Large-scale Linear Programming-based Global Inference",
                        "abstract": "Several key problems in web-scale recommender systems, such as optimal matching and allocation, can be formulated as large-scale linear programs (LPs) [4, 1]. These LPs take predictions from ML models such as probabilities of click, like, etc. as inputs and optimize recommendations made to users. In recent years, there has been an explosion in the research and development of large-scale recommender systems, but effective optimization of business objectives using the output of those systems remains a challenge. Although LPs can help optimize such business objectives, and algorithms for solving LPs have existed since the 1950s [5, 8], generic LP solvers cannot handle the scale of these problems. At LinkedIn, we have developed algorithms that can solve LPs of various forms with trillions of variables in a Spark-based library called \"DuaLip\" [7], a novel distributed solver that solves a perturbation of the LP problem at scale via gradient-based algorithms on the smooth dual of the perturbed LP. DuaLip has been deployed in production at LinkedIn and powers several very large-scale recommender systems. DuaLip is open-sourced and extensible in terms of features and algorithms. In this first-of-its-kind tutorial, we will motivate the application of LPs to improve recommender systems, cover the theory of key LP algorithms [8, 6], and introduce DuaLip (https://github.com/linkedin/DuaLip), a highly performant Spark-based library that solves extreme-scale LPs for a large variety of recommender system problems. We will describe practical successes of large-scale LP in the industry [3, 2, 9] followed by a hands-on exercise to run DuaLip."
                    },
                    {
                        "title": "Optimizing for Member Value in an Edge Building Marketplace",
                        "abstract": "Social networks are prosperous marketplaces where creators and consumers congregate to share and consume various content. In general, products that rank content for distribution (such as newsfeeds, stories, and notifications) and are related to edge recommendations (such as connect to members, follow celebrities or groups or hashtags) optimize the experience of active users. Typically, such users generate ample interaction data amenable to accurate model training and prediction. In contrast, we prioritize enhancing the experience of inactive members (IMs) who do not have a rich connection network. We formulate strategies for recommending superior edges to help members grow their connection network. Adapting the recommendations provides enormous value to the IMs and can significantly influence their future behaviour and engagement with the ecosystem. To that end, we propose a general and scalable multi-objective optimization (MOO) framework to provide more value to IMs as invitation recipients on LinkedIn, a professional network with over 900M members. To deal with the enormous scale, we formulate the problem as a massive constrained linear optimization involving billions of variables and millions of constraints and efficiently solve it using accelerated gradient descent,making this the largest deployment of LP-based recommender systems worldwide. Furthermore, the proposed MOO paradigm can solve the general problem of matching different types of entities in an m-sided marketplace. Finally, we discuss the challenges and benefits of implementing and ramping our method in production at scale at LinkedIn and report our findings about the core business metrics related to users' engagement and network health."
                    },
                    {
                        "title": "QuantEase: Optimization-based Quantization for Language Models - An Efficient and Intuitive Algorithm",
                        "abstract": "With the rising popularity of Large Language Models (LLMs), there has been an increasing interest in compression techniques that enable their efficient deployment. This study focuses on the Post-Training Quantization (PTQ) of LLMs. Drawing from recent advances, our work introduces QuantEase, a layer-wise quantization framework where individual layers undergo separate quantization. The problem is framed as a discrete-structured non-convex optimization, prompting the development of algorithms rooted in Coordinate Descent (CD) techniques. These CD-based methods provide high-quality solutions to the complex non-convex layer-wise quantization problems. Notably, our CD-based approach features straightforward updates, relying solely on matrix and vector operations, circumventing the need for matrix inversion or decomposition. We also explore an outlier-aware variant of our approach, allowing for retaining significant weights (outliers) with complete precision. Our proposal attains state-of-the-art performance in terms of perplexity and zero-shot accuracy in empirical evaluations across various LLMs and datasets, with relative improvements up to 15% over methods such as GPTQ. Leveraging careful linear algebra optimizations, QuantEase can quantize models like Falcon-180B on a single NVIDIA A100 GPU in $\\sim$3 hours. Particularly noteworthy is our outlier-aware algorithm's capability to achieve near or sub-3-bit quantization of LLMs with an acceptable drop in accuracy, obviating the need for non-uniform quantization or grouping techniques, improving upon methods such as SpQR by up to two times in terms of perplexity."
                    },
                    {
                        "title": "A Light-speed Linear Program Solver for Personalized Recommendation with Diversity Constraints",
                        "abstract": "We study a structured linear program (LP) that emerges in the need of ranking candidates or items in personalized recommender systems. Since the candidate set is only known in real time, the LP also needs to be formed and solved in real time. Latency and user experience are major considerations, requiring the LP to be solved within just a few milliseconds. Although typical instances of the problem are not very large in size, this stringent time limit appears to be beyond the capability of most existing (commercial) LP solvers, which can take $20$ milliseconds or more to find a solution. Thus, reliable methods that address the real-world complication of latency become necessary. In this paper, we propose a fast specialized LP solver for a structured problem with diversity constraints. Our method solves the dual problem, making use of the piece-wise affine structure of the dual objective function, with an additional screening technique that helps reduce the dimensionality of the problem as the algorithm progresses. Experiments reveal that our method can solve the problem within roughly 1 millisecond, yielding a 20x improvement in speed over efficient off-the-shelf LP solvers. This speed-up can help improve the quality of recommendations without affecting user experience, highlighting how optimization can provide solid orthogonal value to machine-learned recommender systems."
                    },
                    {
                        "title": "Total Coloring Algorithm for Graphs",
                        "abstract": "A total coloring of a graph is an assignment of colors to all the elements of graph such that no two adjacent or incident elements receive the same color. Behzad and Vizing conjectured that for any graph \ufffd\ufffd the following inequality holds: \u0394(\ufffd\ufffd ) + 1 \u2264 \ufffd\ufffd \u2032\u2032 (\ufffd\ufffd ) \u2264 \u0394(\ufffd\ufffd ) + 2, where \u0394(\ufffd\ufffd ) is the maximum degree of \ufffd\ufffd . Total coloring algorithm is a NP-hard algorithm. In this paper, we give the total coloring algorithm for any graph and we discuss the complexity of the algorithm."
                    },
                    {
                        "title": "2016 Ieee International Conference on Big Data (big Data) Optimizing Callout in Unified Ad Markets",
                        "abstract": "In the past, online ad networks owned their supply of impressions for ads with publishers and matched it to their own demand (advertisers). However, in the past few years, with programmatic ad selling becoming common place, these ad networks are now appealing to demand from multitude of demand side platforms (DSPs). The general approach is to partition the demand into pools (based on type of ads or the demand party). An inevitable challenge is the impedance mismatch between the marketplace and the demand pools. Thus, a central problem is to figure out how to handle each supply request and selectively send requests to demand pool so as to optimize the performance of entire supply. This is the Callout Problem. We present a large scale data analysis and control system that (a) continually learns changing traffic patterns, capacities, bids, revenue etc, and (b) picks the best slice of traffic to send to each demand pool, subject to their capacity constraints. The optimization is based on a greedy solution to the underlying knapsack problem which easily adapts as capacities change over time. We have implemented and deployed this solution for the callout problem in one of the largest mobile ad marketplaces(InMobi) and has been operational for several months. In this paper, we will describe the scale of the problem, our solution and our observations from operational experience with it. We believe a well-engineered solution to the Callout problem is essential for many ad networks in the online ad world."
                    }
                ]
            },
            "9886e444-7312-457a-a395-778e444fcddd": {
                "pk": "9886e444-7312-457a-a395-778e444fcddd",
                "name": "Ayan Acharya",
                "collaborators": [
                    "Rahul Mazumder",
                    "Aman Gupta",
                    "Siyuan Gao",
                    "Ankan Saha",
                    "Borja Ocejo",
                    "Kinjal Basu",
                    "Keerthi Selvaraj",
                    "Parag Agrawal",
                    "Kayhan Behdin",
                    "Qingquan Song"
                ],
                "domain": [
                    "Recommender Systems",
                    "Large Language Models",
                    "Optimization",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Optimizing for Member Value in an Edge Building Marketplace",
                        "abstract": "Social networks are prosperous marketplaces where creators and consumers congregate to share and consume various content. In general, products that rank content for distribution (such as newsfeeds, stories, and notifications) and are related to edge recommendations (such as connect to members, follow celebrities or groups or hashtags) optimize the experience of active users. Typically, such users generate ample interaction data amenable to accurate model training and prediction. In contrast, we prioritize enhancing the experience of inactive members (IMs) who do not have a rich connection network. We formulate strategies for recommending superior edges to help members grow their connection network. Adapting the recommendations provides enormous value to the IMs and can significantly influence their future behaviour and engagement with the ecosystem. To that end, we propose a general and scalable multi-objective optimization (MOO) framework to provide more value to IMs as invitation recipients on LinkedIn, a professional network with over 900M members. To deal with the enormous scale, we formulate the problem as a massive constrained linear optimization involving billions of variables and millions of constraints and efficiently solve it using accelerated gradient descent,making this the largest deployment of LP-based recommender systems worldwide. Furthermore, the proposed MOO paradigm can solve the general problem of matching different types of entities in an m-sided marketplace. Finally, we discuss the challenges and benefits of implementing and ramping our method in production at scale at LinkedIn and report our findings about the core business metrics related to users' engagement and network health."
                    },
                    {
                        "title": "QuantEase: Optimization-based Quantization for Language Models - An Efficient and Intuitive Algorithm",
                        "abstract": "With the rising popularity of Large Language Models (LLMs), there has been an increasing interest in compression techniques that enable their efficient deployment. This study focuses on the Post-Training Quantization (PTQ) of LLMs. Drawing from recent advances, our work introduces QuantEase, a layer-wise quantization framework where individual layers undergo separate quantization. The problem is framed as a discrete-structured non-convex optimization, prompting the development of algorithms rooted in Coordinate Descent (CD) techniques. These CD-based methods provide high-quality solutions to the complex non-convex layer-wise quantization problems. Notably, our CD-based approach features straightforward updates, relying solely on matrix and vector operations, circumventing the need for matrix inversion or decomposition. We also explore an outlier-aware variant of our approach, allowing for retaining significant weights (outliers) with complete precision. Our proposal attains state-of-the-art performance in terms of perplexity and zero-shot accuracy in empirical evaluations across various LLMs and datasets, with relative improvements up to 15% over methods such as GPTQ. Leveraging careful linear algebra optimizations, QuantEase can quantize models like Falcon-180B on a single NVIDIA A100 GPU in $\\sim$3 hours. Particularly noteworthy is our outlier-aware algorithm's capability to achieve near or sub-3-bit quantization of LLMs with an acceptable drop in accuracy, obviating the need for non-uniform quantization or grouping techniques, improving upon methods such as SpQR by up to two times in terms of perplexity."
                    }
                ]
            },
            "0faa115c-020e-4574-b4f5-68827553ed79": {
                "pk": "0faa115c-020e-4574-b4f5-68827553ed79",
                "name": "Rajiv Khanna",
                "collaborators": [
                    "Gregory Dexter",
                    "Liam Hodgkinson",
                    "Umut Simsekli",
                    "Michael W. Mahoney",
                    "Vipul Gupta",
                    "Avishek Ghosh",
                    "Michal Derezinski",
                    "K. Ramchandran",
                    "Young In Kim",
                    "Pratiksha Agrawal"
                ],
                "domain": [
                    "Optimization",
                    "Machine Learning",
                    "Generalization",
                    "Distributed Systems"
                ],
                "publications": [
                    {
                        "title": "On Memorization and Privacy risks of Sharpness Aware Minimization",
                        "abstract": "In many recent works, there is an increased focus on designing algorithms that seek flatter optima for neural network loss optimization as there is empirical evidence that it leads to better generalization performance in many datasets. In this work, we dissect these performance gains through the lens of data memorization in overparameterized models. We define a new metric that helps us identify which data points specifically do algorithms seeking flatter optima do better when compared to vanilla SGD. We find that the generalization gains achieved by Sharpness Aware Minimization (SAM) are particularly pronounced for atypical data points, which necessitate memorization. This insight helps us unearth higher privacy risks associated with SAM, which we verify through exhaustive empirical evaluations. Finally, we propose mitigation strategies to achieve a more desirable accuracy vs privacy tradeoff."
                    },
                    {
                        "title": "mSAM: Micro-Batch-Averaged Sharpness-Aware Minimization",
                        "abstract": "Modern deep learning models are over-parameterized, where different optima can result in widely varying generalization performance. The Sharpness-Aware Minimization (SAM) technique modifies the fundamental loss function that steers gradient descent methods toward flatter minima, which are believed to exhibit enhanced generalization prowess. Our study delves into a specific variant of SAM known as micro-batch SAM (mSAM). This variation involves aggregating updates derived from adversarial perturbations across multiple shards (micro-batches) of a mini-batch during training. We extend a recently developed and well-studied general framework for flatness analysis to theoretically show that SAM achieves flatter minima than SGD, and mSAM achieves even flatter minima than SAM. We provide a thorough empirical evaluation of various image classification and natural language processing tasks to substantiate this theoretical advancement. We also show that contrary to previous work, mSAM can be implemented in a flexible and parallelizable manner without significantly increasing computational costs. Our implementation of mSAM yields superior generalization performance across a wide range of tasks compared to SAM, further supporting our theoretical framework."
                    },
                    {
                        "title": "Feature Space Sketching for Logistic Regression",
                        "abstract": "We present novel bounds for coreset construction, feature selection, and dimensionality reduction for logistic regression. All three approaches can be thought of as sketching the logistic regression inputs. On the coreset construction front, we resolve open problems from prior work and present novel bounds for the complexity of coreset construction methods. On the feature selection and dimensionality reduction front, we initiate the study of forward error bounds for logistic regression. Our bounds are tight up to constant factors and our forward error bounds can be extended to Generalized Linear Models."
                    },
                    {
                        "title": "Generalization Guarantees via Algorithm-dependent Rademacher Complexity",
                        "abstract": "Algorithm- and data-dependent generalization bounds are required to explain the generalization behavior of modern machine learning algorithms. In this context, there exists information theoretic generalization bounds that involve (various forms of) mutual information, as well as bounds based on hypothesis set stability. We propose a conceptually related, but technically distinct complexity measure to control generalization error, which is the empirical Rademacher complexity of an algorithm- and data-dependent hypothesis class. Combining standard properties of Rademacher complexity with the convenient structure of this class, we are able to (i) obtain novel bounds based on the finite fractal dimension, which (a) extend previous fractal dimension-type bounds from continuous to finite hypothesis classes, and (b) avoid a mutual information term that was required in prior work; (ii) we greatly simplify the proof of a recent dimension-independent generalization bound for stochastic gradient descent; and (iii) we easily recover results for VC classes and compression schemes, similar to approaches based on conditional mutual information."
                    },
                    {
                        "title": "Generalization Bounds using Lower Tail Exponents in Stochastic Optimizers",
                        "abstract": "Despite the ubiquitous use of stochastic optimization algorithms in machine learning, the precise impact of these algorithms and their dynamics on generalization performance in realis-tic non-convex settings is still poorly understood. While recent work has revealed connections between generalization and heavy-tailed behavior in stochastic optimization, this work mainly relied on continuous-time approximations; and a rigorous treatment for the original discrete-time iterations is yet to be performed. To bridge this gap, we present novel bounds linking generalization to the lower tail exponent of the transition kernel associated with the optimizer around a local minimum, in both discrete- and continuous-time settings. To achieve this, we \ufb01rst prove a data-and algorithm-dependent generalization bound in terms of the celebrated Fernique\u2013Talagrand functional applied to the trajectory of the optimizer. Then, we specialize this result by exploiting the Markovian structure of stochastic optimizers, and derive bounds in terms of their (data-dependent) transition kernels. We support our theory with empirical results from a variety of neural networks, showing correlations between generalization error and lower tail exponents."
                    },
                    {
                        "title": "LocalNewton: Reducing Communication Bottleneck for Distributed Learning",
                        "abstract": "To address the communication bottleneck problem in distributed optimization within a master-worker framework, we propose LocalNewton, a distributed second-order algorithm with local averaging. In LocalNewton, the worker machines update their model in every iteration by finding a suitable second-order descent direction using only the data and model stored in their own local memory. We let the workers run multiple such iterations locally and communicate the models to the master node only once every few (say L) iterations. LocalNewton is highly practical since it requires only one hyperparameter, the number L of local iterations. We use novel matrix concentration-based techniques to obtain theoretical guarantees for LocalNewton, and we validate them with detailed empirical evaluation. To enhance practicability, we devise an adaptive scheme to choose L, and we show that this reduces the number of local iterations in worker machines between two model synchronizations as the training proceeds, successively refining the model quality at the master. Via extensive experiments using several real-world datasets with AWS Lambda workers and an AWS EC2 master, we show that LocalNewton requires fewer than 60% of the communication rounds (between master and workers) and less than 40% of the end-to-end running time, compared to state-of-the-art algorithms, to reach the same training~loss."
                    },
                    {
                        "title": "LocalNewton: Reducing communication rounds for distributed learning",
                        "abstract": "To address the communication bottleneck problem in distributed optimization within a master-worker framework, we propose LocalNewton, a distributed second-order algorithm with local averaging . In LocalNewton, the worker machines update their model in every iteration by \ufb01nding a suitable second-order descent direction using only the data and model stored in their own local memory. We let the workers run multiple such iterations locally and communicate the models to the master node only once every few (say L ) iterations. Lo-calNewton is highly practical since it requires only one hyperparameter, the number L of local iterations. We use novel matrix concentration based techniques to obtain theoretical guarantees for Lo-calNewton, and we validate them with detailed empirical evaluation. To enhance practicability, we devise an adaptive scheme to choose L , and we show that this reduces the number of local iterations in worker machines between two model syn-chronizations as the training proceeds, successively re\ufb01ning the model quality at the master. Via extensive experiments using several real-world datasets with AWS Lambda workers and an AWS EC2 master, we show that LocalNewton requires fewer than 60% of the communication rounds (between master and workers) and less than 40% of the end-to-end running time, compared to state-of-the-art algorithms, to reach the same training loss."
                    },
                    {
                        "title": "Estimating rates of rare events with multiple hierarchies through scalable log-linear models",
                        "abstract": "We consider the problem of estimating rates of rare events for high dimensional, multivariate categorical data where several dimensions are hierarchical. Such problems are routine in several data mining applications including computational advertising, our main focus in this paper. We propose LMMH, a novel log-linear modeling method that scales to massive data applications with billions of training records and several million potential predictors in a map-reduce framework. Our method exploits correlations in aggregates observed at multiple resolutions when working with multiple hierarchies; stable estimates at coarser resolution provide informative prior information to improve estimates at finer resolutions. Other than prediction accuracy and scalability, our method has an inbuilt variable screening procedure based on a \"spike and slab prior\" that provides parsimony by removing non-informative predictors without hurting predictive accuracy. We perform large scale experiments on data from real computational advertising applications and illustrate our approach on datasets with several billion records and hundreds of millions of predictors. Extensive comparisons with other benchmark methods show significant improvements in prediction accuracy."
                    }
                ]
            }
        }
    },
    "2406.04325": {
        "paper_data": {
            "title": "ShareGPT4Video: Improving Video Understanding and Generation with Better Captions",
            "url": "http://arxiv.org/abs/2406.04325v1",
            "arxiv_id": "2406.04325",
            "authors": [
                "Lin Chen",
                "Xilin Wei",
                "Jinsong Li",
                "Xiaoyi Dong",
                "Pan Zhang",
                "Yuhang Zang",
                "Zehui Chen",
                "Haodong Duan",
                "Bin Lin",
                "Zhenyu Tang",
                "Li Yuan",
                "Yu Qiao",
                "Dahua Lin",
                "Feng Zhao",
                "Jiaqi Wang"
            ],
            "abstract": "We present the ShareGPT4Video series, aiming to facilitate the video understanding of large video-language models (LVLMs) and the video generation of text-to-video models (T2VMs) via dense and precise captions. The series comprises: 1) ShareGPT4Video, 40K GPT4V annotated dense captions of videos with various lengths and sources, developed through carefully designed data filtering and annotating strategy. 2) ShareCaptioner-Video, an efficient and capable captioning model for arbitrary videos, with 4.8M high-quality aesthetic videos annotated by it. 3) ShareGPT4Video-8B, a simple yet superb LVLM that reached SOTA performance on three advancing video benchmarks. To achieve this, taking aside the non-scalable costly human annotators, we find using GPT4V to caption video with a naive multi-frame or frame-concatenation input strategy leads to less detailed and sometimes temporal-confused results. We argue the challenge of designing a high-quality video captioning strategy lies in three aspects: 1) Inter-frame precise temporal change understanding. 2) Intra-frame detailed content description. 3) Frame-number scalability for arbitrary-length videos. To this end, we meticulously designed a differential video captioning strategy, which is stable, scalable, and efficient for generating captions for videos with arbitrary resolution, aspect ratios, and length. Based on it, we construct ShareGPT4Video, which contains 40K high-quality videos spanning a wide range of categories, and the resulting captions encompass rich world knowledge, object attributes, camera movements, and crucially, detailed and precise temporal descriptions of events. Based on ShareGPT4Video, we further develop ShareCaptioner-Video, a superior captioner capable of efficiently generating high-quality captions for arbitrary videos...",
            "introduction": "   1 Introduction  Recent advancements in multi-modal learning, driven by large language models, have led to progress in image-text dialogue[35, 10, 16, 56, 63, 13, 2, 65] and text-to-image generation tasks[3, 52, 6, 53, 70, 8]. This has inspired a shift towards video understanding[34, 1, 32, 69, 42, 40] and generation tasks[54, 20, 60, 57, 4, 39], allowing for user interaction across video and language modalities. Thus, the detailed and high-fidelity video captions, which bridge the aforementioned modalities, is instrumental in propelling the advancements within the field.   Despite the rich semantic and temporal content of videos, they are often paired with brief captions in existing data. These short descriptions limit the detailed video understanding and the controllability of the video generation. While the importance of detailed captions is recognized in image-text dialogue[10, 7, 55] and text-to-image generation tasks[8, 22], similar efforts are lacking in video understanding and generation.   However, creating large-scale, high-quality video captions is a challenging task. Detailed captioning for long videos is non-trivial and time-consuming even for humans, hindering large-scale annotation. Current open-source LVLMs lack this capability, and closed-source APIs do not yet support video inputs.   On the other hand, if we roughly degrade the input from video to multiple frames, even GPT4V struggles to describe the video with satisfied quality. For example, an intuitive idea is to provide multiple frames with timestamps to the GPT4V and generate the caption, while we find that GPT4V is unstable and sometimes misunderstands the temporal relation between the frames, and its performance further degrades with the increasing of video frame number. Other solutions such as concatenating all the frames into a large image are nonhelpful to the temporal problem, and the caption loses details as the frame number increases. We also showcase these problems in Figure\u00a011-12   We posit that the challenge of devising an effective video captioning strategy is rooted in three fundamental aspects: 1) Inter-frame precise temporal change understanding: The temporal dimension distinguishes videos from images. An imprecise temporal description can significantly diminish the quality of the video caption and lead to confusion in the training models. 2) Intra-frame detailed content description: Detailed descriptions\u00a0[10] are crucial for aligning modalities between image and text, which are also important for video-text alignment. 3) Frame-number scalability for arbitrary-length videos: Videos encountered in the wild can vary greatly in length. An ideal captioning strategy should be resilient to this variability and generate appropriate captions for videos of any length.   To this end, we present the Differential Sliding-Window Captioning strategy (DiffSW), which is stable, scalable, and efficient for generating captions for arbitrary videos. The central concept of DiffSW is translating the all-frames-to-caption task into a differential description task. Specifically, we generate a detailed caption for the first frame and apply a sliding window of length two to the subsequent frames in chronological order. The powerful image multi-modal model, GPT4V [46], is tasked with identifying the changes between frames based on three inputs: the previous frame, its differential caption, and the current frame. This encompasses alterations in camera movement, object movement, character actions, and scene transitions. Upon acquiring all differential captions, these are input into GPT4 [45] to",
            "references": [
                {
                    "title": "Q-Instruct: Improving Low-Level Visual Abilities for Multi-Modality Foundation Models",
                    "abstract": "Multi-modality large language models (MLLMs), as represented by GPT-4V, have introduced a paradigm shift for visual perception and understanding tasks, that a variety of abilities can be achieved within one foundation model. While current MLLMs demonstrate primary low-level visual abilities from the identification of low-level visual attributes (e.g., clarity, brightness) to the evaluation on image quality, there's still an imperative to further improve the accuracy of MLLMs to substantially alleviate human burdens. To address this, we collect the first dataset consisting of human natural language feedback on low-level vision. Each feedback offers a comprehensive description of an image's low-level visual attributes, culminating in an overall quality assessment. The constructed Q-Pathway dataset includes 58K detailed human feedbacks on 18,973 multi-sourced images with diverse low-level appearance. To ensure MLLMs can adeptly handle diverse queries, we further propose a GPT-participated transformation to convert these feedbacks into a rich set of 200K instruction-response pairs, termed Q-Instruct. Experimental results indicate that the Q-Instruct consistently elevates various low-level visual capabilities across multiple base models. We anticipate that our datasets can pave the way for a future that foundation models can assist humans on low-level visual tasks."
                },
                {
                    "title": "InternLM-XComposer2-4KHD: A Pioneering Large Vision-Language Model Handling Resolutions from 336 Pixels to 4K HD",
                    "abstract": "The Large Vision-Language Model (LVLM) field has seen significant advancements, yet its progression has been hindered by challenges in comprehending fine-grained visual content due to limited resolution. Recent efforts have aimed to enhance the high-resolution understanding capabilities of LVLMs, yet they remain capped at approximately 1500 x 1500 pixels and constrained to a relatively narrow resolution range. This paper represents InternLM-XComposer2-4KHD, a groundbreaking exploration into elevating LVLM resolution capabilities up to 4K HD (3840 x 1600) and beyond. Concurrently, considering the ultra-high resolution may not be necessary in all scenarios, it supports a wide range of diverse resolutions from 336 pixels to 4K standard, significantly broadening its scope of applicability. Specifically, this research advances the patch division paradigm by introducing a novel extension: dynamic resolution with automatic patch configuration. It maintains the training image aspect ratios while automatically varying patch counts and configuring layouts based on a pre-trained Vision Transformer (ViT) (336 x 336), leading to dynamic training resolution from 336 pixels to 4K standard. Our research demonstrates that scaling training resolution up to 4K HD leads to consistent performance enhancements without hitting the ceiling of potential improvements. InternLM-XComposer2-4KHD shows superb capability that matches or even surpasses GPT-4V and Gemini Pro in 10 of the 16 benchmarks. The InternLM-XComposer2-4KHD model series with 7B parameters are publicly available at https://github.com/InternLM/InternLM-XComposer."
                },
                {
                    "title": "MiniGPT4-Video: Advancing Multimodal LLMs for Video Understanding with Interleaved Visual-Textual Tokens",
                    "abstract": "This paper introduces MiniGPT4-Video, a multimodal Large Language Model (LLM) designed specifically for video understanding. The model is capable of processing both temporal visual and textual data, making it adept at understanding the complexities of videos. Building upon the success of MiniGPT-v2, which excelled in translating visual features into the LLM space for single images and achieved impressive results on various image-text benchmarks, this paper extends the model's capabilities to process a sequence of frames, enabling it to comprehend videos. MiniGPT4-video does not only consider visual content but also incorporates textual conversations, allowing the model to effectively answer queries involving both visual and text components. The proposed model outperforms existing state-of-the-art methods, registering gains of 4.22%, 1.13%, 20.82%, and 13.1% on the MSVD, MSRVTT, TGIF, and TVQA benchmarks respectively. Our models and code have been made publicly available here https://vision-cair.github.io/MiniGPT4-video/"
                },
                {
                    "title": "Are We on the Right Way for Evaluating Large Vision-Language Models?",
                    "abstract": "Large vision-language models (LVLMs) have recently achieved rapid progress, sparking numerous studies to evaluate their multi-modal capabilities. However, we dig into current evaluation works and identify two primary issues: 1) Visual content is unnecessary for many samples. The answers can be directly inferred from the questions and options, or the world knowledge embedded in LLMs. This phenomenon is prevalent across current benchmarks. For instance, GeminiPro achieves 42.9% on the MMMU benchmark without any visual input, and outperforms the random choice baseline across six benchmarks over 24% on average. 2) Unintentional data leakage exists in LLM and LVLM training. LLM and LVLM could still answer some visual-necessary questions without visual content, indicating the memorizing of these samples within large-scale training data. For example, Sphinx-X-MoE gets 43.6% on MMMU without accessing images, surpassing its LLM backbone with 17.9%. Both problems lead to misjudgments of actual multi-modal gains and potentially misguide the study of LVLM. To this end, we present MMStar, an elite vision-indispensable multi-modal benchmark comprising 1,500 samples meticulously selected by humans. MMStar benchmarks 6 core capabilities and 18 detailed axes, aiming to evaluate LVLMs' multi-modal capacities with carefully balanced and purified samples. These samples are first roughly selected from current benchmarks with an automated pipeline, human review is then involved to ensure each curated sample exhibits visual dependency, minimal data leakage, and requires advanced multi-modal capabilities. Moreover, two metrics are developed to measure data leakage and actual performance gain in multi-modal training. We evaluate 16 leading LVLMs on MMStar to assess their multi-modal capabilities, and on 7 benchmarks with the proposed metrics to investigate their data leakage and actual multi-modal gain."
                },
                {
                    "title": "An Image Grid Can Be Worth a Video: Zero-shot Video Question Answering Using a VLM",
                    "abstract": "Stimulated by the sophisticated reasoning capabilities of recent Large Language Models (LLMs), a variety of strategies for bridging video modality have been devised. A prominent strategy involves Video Language Models (VideoLMs), which train a learnable interface with video data to connect advanced vision encoders with LLMs. Recently, an alternative strategy has surfaced, employing readily available foundation models, such as VideoLMs and LLMs, across multiple stages for modality bridging. In this study, we introduce a simple yet novel strategy where only a single Vision Language Model (VLM) is utilized. Our starting point is the plain insight that a video comprises a series of images, or frames, interwoven with temporal information. The essence of video comprehension lies in adeptly managing the temporal aspects along with the spatial details of each frame. Initially, we transform a video into a single composite image by arranging multiple frames in a grid layout. The resulting single image is termed as an image grid. This format, while maintaining the appearance of a solitary image, effectively retains temporal information within the grid structure. Therefore, the image grid approach enables direct application of a single high-performance VLM without necessitating any video-data training. Our extensive experimental analysis across ten zero-shot video question answering benchmarks, including five open-ended and five multiple-choice benchmarks, reveals that the proposed Image Grid Vision Language Model (IG-VLM) surpasses the existing methods in nine out of ten benchmarks."
                },
                {
                    "title": "Yi: Open Foundation Models by 01.AI",
                    "abstract": "We introduce the Yi model family, a series of language and multimodal models that demonstrate strong multi-dimensional capabilities. The Yi model family is based on 6B and 34B pretrained language models, then we extend them to chat models, 200K long context models, depth-upscaled models, and vision-language models. Our base models achieve strong performance on a wide range of benchmarks like MMLU, and our finetuned chat models deliver strong human preference rate on major evaluation platforms like AlpacaEval and Chatbot Arena. Building upon our scalable super-computing infrastructure and the classical transformer architecture, we attribute the performance of Yi models primarily to its data quality resulting from our data-engineering efforts. For pretraining, we construct 3.1 trillion tokens of English and Chinese corpora using a cascaded data deduplication and quality filtering pipeline. For finetuning, we polish a small scale (less than 10K) instruction dataset over multiple iterations such that every single instance has been verified directly by our machine learning engineers. For vision-language, we combine the chat language model with a vision transformer encoder and train the model to align visual representations to the semantic space of the language model. We further extend the context length to 200K through lightweight continual pretraining and demonstrate strong needle-in-a-haystack retrieval performance. We show that extending the depth of the pretrained checkpoint through continual pretraining further improves performance. We believe that given our current results, continuing to scale up model parameters using thoroughly optimized data will lead to even stronger frontier models."
                },
                {
                    "title": "TempCompass: Do Video LLMs Really Understand Videos?",
                    "abstract": "Recently, there is a surge in interest surrounding video large language models (Video LLMs). However, existing benchmarks fail to provide a comprehensive feedback on the temporal perception ability of Video LLMs. On the one hand, most of them are unable to distinguish between different temporal aspects (e.g., speed, direction) and thus cannot reflect the nuanced performance on these specific aspects. On the other hand, they are limited in the diversity of task formats (e.g., only multi-choice QA), which hinders the understanding of how temporal perception performance may vary across different types of tasks. Motivated by these two problems, we propose the \\textbf{TempCompass} benchmark, which introduces a diversity of temporal aspects and task formats. To collect high-quality test data, we devise two novel strategies: (1) In video collection, we construct conflicting videos that share the same static content but differ in a specific temporal aspect, which prevents Video LLMs from leveraging single-frame bias or language priors. (2) To collect the task instructions, we propose a paradigm where humans first annotate meta-information for a video and then an LLM generates the instruction. We also design an LLM-based approach to automatically and accurately evaluate the responses from Video LLMs. Based on TempCompass, we comprehensively evaluate 8 state-of-the-art (SOTA) Video LLMs and 3 Image LLMs, and reveal the discerning fact that these models exhibit notably poor temporal perception ability. Our data will be available at https://github.com/llyx97/TempCompass."
                },
                {
                    "title": "Panda-70M: Captioning 70M Videos with Multiple Cross-Modality Teachers",
                    "abstract": "The quality of the data and annotation upper-bounds the quality of a downstream model. While there exist large text corpora and image-text pairs, high-quality video-text data is much harder to collect. First of all, manual labeling is more time-consuming, as it requires an annotator to watch an entire video. Second, videos have a temporal dimension, consisting of several scenes stacked together, and showing multiple actions. Accordingly, to establish a video dataset with high-quality captions, we propose an automatic approach leveraging multimodal inputs, such as textual video description, subtitles, and individual video frames. Specifically, we curate 3.8M high-resolution videos from the publicly available HD-VILA-100M dataset. We then split them into semantically consistent video clips, and apply multiple cross-modality teacher models to obtain captions for each video. Next, we finetune a retrieval model on a small subset where the best caption of each video is manually selected and then employ the model in the whole dataset to select the best caption as the annotation. In this way, we get 70M videos paired with high-quality text captions. We dub the dataset as Panda-70M. We show the value of the proposed dataset on three downstream tasks: video captioning, video and text retrieval, and text-driven video generation. The models trained on the proposed data score substantially better on the majority of metrics across all the tasks."
                },
                {
                    "title": "Sora: A Review on Background, Technology, Limitations, and Opportunities of Large Vision Models",
                    "abstract": "Sora is a text-to-video generative AI model, released by OpenAI in February 2024. The model is trained to generate videos of realistic or imaginative scenes from text instructions and show potential in simulating the physical world. Based on public technical reports and reverse engineering, this paper presents a comprehensive review of the model's background, related technologies, applications, remaining challenges, and future directions of text-to-video AI models. We first trace Sora's development and investigate the underlying technologies used to build this\"world simulator\". Then, we describe in detail the applications and potential impact of Sora in multiple industries ranging from film-making and education to marketing. We discuss the main challenges and limitations that need to be addressed to widely deploy Sora, such as ensuring safe and unbiased video generation. Lastly, we discuss the future development of Sora and video generation models in general, and how advancements in the field could enable new ways of human-AI interaction, boosting productivity and creativity of video generation."
                },
                {
                    "title": "TinyLLaVA: A Framework of Small-scale Large Multimodal Models",
                    "abstract": "We present the TinyLLaVA framework that provides a unified perspective in designing and analyzing the small-scale Large Multimodal Models (LMMs). We empirically study the effects of different vision encoders, connection modules, language models, training data and training recipes. Our extensive experiments showed that better quality of data combined with better training recipes, smaller LMMs can consistently achieve on-par performances compared to bigger LMMs. Under our framework, we train a family of small-scale LMMs. Our best model, TinyLLaVA-3.1B, achieves better overall performance against existing 7B models such as LLaVA-1.5 and Qwen-VL. We hope our findings can serve as baselines for future research in terms of data scaling, training setups and model selections. Our model weights and codes will be made public."
                },
                {
                    "title": "ALLaVA: Harnessing GPT4V-Synthesized Data for Lite Vision-Language Models",
                    "abstract": "Large vision-language models (LVLMs) have shown premise in a broad range of vision-language tasks with their strong reasoning and generalization capabilities. However, they require considerable computational resources for training and deployment. This study aims to bridge the performance gap between traditional-scale LVLMs and resource-friendly lite versions by adopting high-quality training data. To this end, we propose a comprehensive pipeline for generating a synthetic dataset. The key idea is to leverage strong proprietary models to generate (i) fine-grained image annotations for vision-language alignment and (ii) complex reasoning visual question-answering pairs for visual instruction fine-tuning, yielding 1.3M samples in total. We train a series of lite VLMs on the synthetic dataset and experimental results demonstrate the effectiveness of the proposed scheme, where they achieve competitive performance on 17 benchmarks among 4B LVLMs, and even perform on par with 7B/13B-scale models on various benchmarks. This work highlights the feasibility of adopting high-quality data in crafting more efficient LVLMs. We name our dataset \\textit{ALLaVA}, and open-source it to research community for developing better resource-efficient LVLMs for wider usage."
                },
                {
                    "title": "InternLM-XComposer2: Mastering Free-form Text-Image Composition and Comprehension in Vision-Language Large Model",
                    "abstract": "We introduce InternLM-XComposer2, a cutting-edge vision-language model excelling in free-form text-image composition and comprehension. This model goes beyond conventional vision-language understanding, adeptly crafting interleaved text-image content from diverse inputs like outlines, detailed textual specifications, and reference images, enabling highly customizable content creation. InternLM-XComposer2 proposes a Partial LoRA (PLoRA) approach that applies additional LoRA parameters exclusively to image tokens to preserve the integrity of pre-trained language knowledge, striking a balance between precise vision understanding and text composition with literary talent. Experimental results demonstrate the superiority of InternLM-XComposer2 based on InternLM2-7B in producing high-quality long-text multi-modal content and its exceptional vision-language understanding performance across various benchmarks, where it not only significantly outperforms existing multimodal models but also matches or even surpasses GPT-4V and Gemini Pro in certain assessments. This highlights its remarkable proficiency in the realm of multimodal understanding. The InternLM-XComposer2 model series with 7B parameters are publicly available at https://github.com/InternLM/InternLM-XComposer."
                },
                {
                    "title": "Latte: Latent Diffusion Transformer for Video Generation",
                    "abstract": "We propose a novel Latent Diffusion Transformer, namely Latte, for video generation. Latte first extracts spatio-temporal tokens from input videos and then adopts a series of Transformer blocks to model video distribution in the latent space. In order to model a substantial number of tokens extracted from videos, four efficient variants are introduced from the perspective of decomposing the spatial and temporal dimensions of input videos. To improve the quality of generated videos, we determine the best practices of Latte through rigorous experimental analysis, including video clip patch embedding, model variants, timestep-class information injection, temporal positional embedding, and learning strategies. Our comprehensive evaluation demonstrates that Latte achieves state-of-the-art performance across four standard video generation datasets, i.e., FaceForensics, SkyTimelapse, UCF101, and Taichi-HD. In addition, we extend Latte to text-to-video generation (T2V) task, where Latte achieves comparable results compared to recent T2V models. We strongly believe that Latte provides valuable insights for future research on incorporating Transformers into diffusion models for video generation."
                },
                {
                    "title": "Intern VL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks",
                    "abstract": "The exponential growth of large language models (LLMs) has opened up numerous possibilities for multi-modal AGI systems. However, the progress in vision and vision-language foundation models, which are also critical elements of multi-modal AGI, has not kept pace with LLMs. In this work, we design a large-scale vision-language foun-dation model (Intern VL), which scales up the vision foun-dation model to 6 billion parameters and progressively aligns it with the LLM, using web-scale image-text data from various sources. This model can be broadly applied to and achieve state-of-the-art performance on 32 generic visual-linguistic benchmarks including visual perception tasks such as image-level or pixel-level recognition, vision-language tasks such as zero-shot image/video classification, zero-shot image/video-text retrieval, and link with LLMs to create multi-modal dialogue systems. It has powerful visual capabilities and can be a good alternative to the ViT-22B. We hope that our research could contribute to the development of multi-modal large models."
                },
                {
                    "title": "LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models",
                    "abstract": "In this work, we present a novel method to tackle the token generation challenge in Vision Language Models (VLMs) for video and image understanding, called LLaMA-VID. Current VLMs, while proficient in tasks like image captioning and visual question answering, face computational burdens when processing long videos due to the excessive visual tokens. LLaMA-VID addresses this issue by representing each frame with two distinct tokens, namely context token and content token. The context token encodes the overall image context based on user input, whereas the content token encapsulates visual cues in each frame. This dual-token strategy significantly reduces the overload of long videos while preserving critical information. Generally, LLaMA-VID empowers existing frameworks to support hour-long videos and pushes their upper limit with an extra context token. It is proved to surpass previous methods on most of video- or image-based benchmarks. Code is available https://github.com/dvlab-research/LLaMA-VID}{https://github.com/dvlab-research/LLaMA-VID"
                },
                {
                    "title": "MVBench: A Comprehensive Multi-modal Video Understanding Benchmark",
                    "abstract": "With the rapid development of Multi-modal Large language Models (MLLMs), a number of diagnostic bench-marks have recently emerged to evaluate the comprehension capabilities of these models. However, most bench-marks predominantly assess spatial understanding in the static image tasks, while overlooking temporal understanding in the dynamic video tasks. To alleviate this issue, we introduce a comprehensive Multi-modal Video understanding Benchmark, namely MVBench, which covers 20 chal-lenging video tasks that cannot be effectively solved with a single frame. Specifically, we first introduce a novel static-to-dynamic method to define these temporal-related tasks. By transforming various static tasks into dynamic ones, we enable the systematic generation of video tasks that require a broad spectrum of temporal skills, ranging from perception to cognition. Then, guided by the task definition, we au-tomatically convert public video annotations into multiple-choice QA to evaluate each task. On one hand, such a distinct paradigm allows us to build MVBench efficiently, without much manual intervention. On the other hand, it guarantees evaluation fairness with ground-truth video an-notations, avoiding the biased scoring of LLMs. More-over, we further develop a robust video MLLM baseline, i.e., VideoChat2, by progressive multi-modal training with di-verse instruction-tuning data. The extensive results on our MVBench reveal that, the existing MLLMs are far from sat-isfactory in temporal understanding, while our VideoChat2 largely surpasses these leading models by over 15% on MVBench. All models and data are available at https://github.com/OpenGVLab/Ask-Anything."
                },
                {
                    "title": "Video-Bench: A Comprehensive Benchmark and Toolkit for Evaluating Video-based Large Language Models",
                    "abstract": "Video-based large language models (Video-LLMs) have been recently introduced, targeting both fundamental improvements in perception and comprehension, and a diverse range of user inquiries. In pursuit of the ultimate goal of achieving artificial general intelligence, a truly intelligent Video-LLM model should not only see and understand the surroundings, but also possess human-level commonsense, and make well-informed decisions for the users. To guide the development of such a model, the establishment of a robust and comprehensive evaluation system becomes crucial. To this end, this paper proposes \\textit{Video-Bench}, a new comprehensive benchmark along with a toolkit specifically designed for evaluating Video-LLMs. The benchmark comprises 10 meticulously crafted tasks, evaluating the capabilities of Video-LLMs across three distinct levels: Video-exclusive Understanding, Prior Knowledge-based Question-Answering, and Comprehension and Decision-making. In addition, we introduce an automatic toolkit tailored to process model outputs for various tasks, facilitating the calculation of metrics and generating convenient final scores. We evaluate 8 representative Video-LLMs using \\textit{Video-Bench}. The findings reveal that current Video-LLMs still fall considerably short of achieving human-like comprehension and analysis of real-world videos, offering valuable insights for future research directions. The benchmark and toolkit are available at: \\url{https://github.com/PKU-YuanGroup/Video-Bench}."
                },
                {
                    "title": "Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets",
                    "abstract": "We present Stable Video Diffusion - a latent video diffusion model for high-resolution, state-of-the-art text-to-video and image-to-video generation. Recently, latent diffusion models trained for 2D image synthesis have been turned into generative video models by inserting temporal layers and finetuning them on small, high-quality video datasets. However, training methods in the literature vary widely, and the field has yet to agree on a unified strategy for curating video data. In this paper, we identify and evaluate three different stages for successful training of video LDMs: text-to-image pretraining, video pretraining, and high-quality video finetuning. Furthermore, we demonstrate the necessity of a well-curated pretraining dataset for generating high-quality videos and present a systematic curation process to train a strong base model, including captioning and filtering strategies. We then explore the impact of finetuning our base model on high-quality data and train a text-to-video model that is competitive with closed-source video generation. We also show that our base model provides a powerful motion representation for downstream tasks such as image-to-video generation and adaptability to camera motion-specific LoRA modules. Finally, we demonstrate that our model provides a strong multi-view 3D-prior and can serve as a base to finetune a multi-view diffusion model that jointly generates multiple views of objects in a feedforward fashion, outperforming image-based methods at a fraction of their compute budget. We release code and model weights at https://github.com/Stability-AI/generative-models ."
                },
                {
                    "title": "ShareGPT4V: Improving Large Multi-Modal Models with Better Captions",
                    "abstract": "In the realm of large multi-modal models (LMMs), efficient modality alignment is crucial yet often constrained by the scarcity of high-quality image-text data. To address this bottleneck, we introduce the ShareGPT4V dataset, a pioneering large-scale resource featuring 1.2 million highly descriptive captions, which surpasses existing datasets in diversity and information content, covering world knowledge, object properties, spatial relationships, and aesthetic evaluations. Specifically, ShareGPT4V originates from a curated 100K high-quality captions collected from advanced GPT4-Vision and has been expanded to 1.2M with a superb caption model trained on this subset. ShareGPT4V first demonstrates its effectiveness for the Supervised Fine-Tuning (SFT) phase, by substituting an equivalent quantity of detailed captions in existing SFT datasets with a subset of our high-quality captions, significantly enhancing the LMMs like LLaVA-7B, LLaVA-1.5-13B, and Qwen-VL-Chat-7B on the MME and MMBench benchmarks, with respective gains of 222.8/22.0/22.3 and 2.7/1.3/1.5. We further incorporate ShareGPT4V data into both the pre-training and SFT phases, obtaining ShareGPT4V-7B, a superior LMM based on a simple architecture that has remarkable performance across a majority of the multi-modal benchmarks. This project is available at https://ShareGPT4V.github.io to serve as a pivotal resource for advancing the LMMs community."
                },
                {
                    "title": "Make Pixels Dance: High-Dynamic Video Generation",
                    "abstract": "Creating high-dynamic videos such as motion-rich actions and sophisticated visual effects poses a significant challenge in the field of artificial intelligence. Unfortu-nately, current state-of-the-art video generation methods, primarily focusing on text-to- video generation, tend to pro-duce video clips with minimal motions despite maintaining high fidelity. We argue that relying solely on text in-structions is insufficient and suboptimal for video generation. In this paper, we introduce PixelDance, a novel approach based on diffusion models that incorporates image instructions for both the first and last frames in conjunction with text instructions for video generation. Comprehensive experimental results demonstrate that PixelDance trained with public data exhibits significantly better proficiency in synthesizing videos with complex scenes and intricate motions, setting a new standard for video generation."
                },
                {
                    "title": "Video-LLaVA: Learning United Visual Representation by Alignment Before Projection",
                    "abstract": "The Large Vision-Language Model (LVLM) has enhanced the performance of various downstream tasks in visual-language understanding. Most existing approaches encode images and videos into separate feature spaces, which are then fed as inputs to large language models. However, due to the lack of unified tokenization for images and videos, namely misalignment before projection, it becomes challenging for a Large Language Model (LLM) to learn multi-modal interactions from several poor projection layers. In this work, we unify visual representation into the language feature space to advance the foundational LLM towards a unified LVLM. As a result, we establish a simple but robust LVLM baseline, Video-LLaVA, which learns from a mixed dataset of images and videos, mutually enhancing each other. Video-LLaVA achieves superior performances on a broad range of 9 image benchmarks across 5 image question-answering datasets and 4 image benchmark toolkits. Additionally, our Video-LLaVA also outperforms Video-ChatGPT by 5.8%, 9.9%, 18.6%, and 10.1% on MSRVTT, MSVD, TGIF, and ActivityNet, respectively. Notably, extensive experiments demonstrate that Video-LLaVA mutually benefits images and videos within a unified visual representation, outperforming models designed specifically for images or videos. We aim for this work to provide modest insights into the multi-modal inputs for the LLM. Code address: \\href{https://github.com/PKU-YuanGroup/Video-LLaVA}"
                },
                {
                    "title": "Chat-UniVi: Unified Visual Representation Empowers Large Language Models with Image and Video Understanding",
                    "abstract": "Large language models have demonstrated impressive universal capabilities across a wide range of open-ended tasks and have extended their utility to encompass multi-modal conversations. However, existing methods encounter challenges in effectively handling both image and video understanding, particularly with limited visual tokens. In this work, we introduce Chat-UniVi, a Unified Vision-language model capable of comprehending and engaging in conver-sations involving images and videos through a unified visual representation. Specifically, we employ a set of dynamic visual tokens to uniformly represent images and videos. This representation framework empowers the model to ef-ficiently utilize a limited number of visual tokens to simul-taneously capture the spatial details necessary for images and the comprehensive temporal relationship required for videos. Moreover, we leverage a multi-scale representation, enabling the model to perceive both high-level seman-tic concepts and low-level visual details. Notably, Chat-UniVi is trained on a mixed dataset containing both images and videos, allowing direct application to tasks involving both mediums without requiring any modifications. Exten-sive experimental results demonstrate that Chat- UniVi con-sistently outperforms even existing methods exclusively de-signed for either images or videos. Code is available at https://github.com/PKu-Yuan Group/Chat-UniVi."
                },
                {
                    "title": "To See is to Believe: Prompting GPT-4V for Better Visual Instruction Tuning",
                    "abstract": "Existing visual instruction tuning methods typically prompt large language models with textual descriptions to generate instruction-following data. Despite the promising performance achieved, these descriptions are derived from image annotations, which are oftentimes coarse-grained. Furthermore, the instructions might even contradict the visual content without observing the entire visual context. To address this challenge, we introduce a fine-grained visual instruction dataset, LVIS-Instruct4V, which contains 220K visually aligned and context-aware instructions produced by prompting the powerful GPT-4V with images from LVIS. Through experimental validation and case studies, we demonstrate that high-quality visual instructional data could improve the performance of LLaVA-1.5, a state-of-the-art large multimodal model, across a wide spectrum of benchmarks by clear margins. Notably, by simply replacing the LLaVA-Instruct with our LVIS-Instruct4V, we achieve better results than LLaVA on most challenging LMM benchmarks, e.g., LLaVA$^w$ (76.7 vs. 70.7) and MM-Vet (40.2 vs. 35.4). We release our data and model at https://github.com/X2FD/LVIS-INSTRUCT4V."
                },
                {
                    "title": "Monkey: Image Resolution and Text Label are Important Things for Large Multi-Modal Models",
                    "abstract": "Large Multimodal Models (LMMs) have shown promise in vision-language tasks but struggle with high-resolution input and detailed scene understanding. Addressing these challenges, we introduce Monkey to enhance LMM capabilities. Firstly, Monkey processes input images by dividing them into uniform patches, each matching the size (e.g., 448\u00d7448) used in the original training of the well-trained vision encoder. Equipped with individual adapter for each patch, Monkey can handle higher resolutions up to 1344\u00d7896 pixels, enabling the detailed capture of complex visual information. Secondly, it employs a multi-level description generation method, enriching the context for scene-object associations. This two-part strategy ensures more effective learning from generated data: the higher resolution allows for a more detailed capture of visuals, which in turn enhances the effectiveness of comprehensive descriptions. Extensive ablative results validate the effectiveness of our designs. Additionally, experiments on 18 datasets further demonstrate that Monkey surpasses existing LMMs in many tasks like Image Captioning and various Visual Question Answering formats. Specially, in qualitative tests focused on dense text question answering, Monkey has exhibited encouraging results compared with GPT4V. Code is available at https://github.com/Yuliang-Liu/Monkey."
                },
                {
                    "title": "CogVLM: Visual Expert for Pretrained Language Models",
                    "abstract": "We introduce CogVLM, a powerful open-source visual language foundation model. Different from the popular shallow alignment method which maps image features into the input space of language model, CogVLM bridges the gap between the frozen pretrained language model and image encoder by a trainable visual expert module in the attention and FFN layers. As a result, CogVLM enables deep fusion of vision language features without sacrificing any performance on NLP tasks. CogVLM-17B achieves state-of-the-art performance on 10 classic cross-modal benchmarks, including NoCaps, Flicker30k captioning, RefCOCO, RefCOCO+, RefCOCOg, Visual7W, GQA, ScienceQA, VizWiz VQA and TDIUC, and ranks the 2nd on VQAv2, OKVQA, TextVQA, COCO captioning, etc., surpassing or matching PaLI-X 55B. Codes and checkpoints are available at https://github.com/THUDM/CogVLM."
                },
                {
                    "title": "Improved Baselines with Visual Instruction Tuning",
                    "abstract": "Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this paper, we present the first systematic study to investigate the design choices of LMMs in a controlled setting under the LLaVA framework. We show that the fully-connected vision-language connector in LLaVA is surprisingly power-ful and data-efficient. With simple modifications to LLa VA, namely, using CLIP- ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~ 1 day on a single 8-AI00 node. Furthermore, we present some early exploration of open problems in LMMs, including scaling to higher resolution inputs, compositional capabilities, and model hallucination, etc. We hope this makes state-of-the-art LMM research more accessible. Code and model will be publicly available."
                },
                {
                    "title": "LAVIE: High-Quality Video Generation with Cascaded Latent Diffusion Models",
                    "abstract": "This work aims to learn a high-quality text-to-video (T2V) generative model by leveraging a pre-trained text-to-image (T2I) model as a basis. It is a highly desirable yet challenging task to simultaneously a) accomplish the synthesis of visually realistic and temporally coherent videos while b) preserving the strong creative generation nature of the pre-trained T2I model. To this end, we propose LaVie, an integrated video generation framework that operates on cascaded video latent diffusion models, comprising a base T2V model, a temporal interpolation model, and a video super-resolution model. Our key insights are two-fold: 1) We reveal that the incorporation of simple temporal self-attentions, coupled with rotary positional encoding, adequately captures the temporal correlations inherent in video data. 2) Additionally, we validate that the process of joint image-video fine-tuning plays a pivotal role in producing high-quality and creative outcomes. To enhance the performance of LaVie, we contribute a comprehensive and diverse video dataset named Vimeo25M, consisting of 25 million text-video pairs that prioritize quality, diversity, and aesthetic appeal. Extensive experiments demonstrate that LaVie achieves state-of-the-art performance both quantitatively and qualitatively. Furthermore, we showcase the versatility of pre-trained LaVie models in various long video generation and personalized video synthesis applications."
                },
                {
                    "title": "Q-Bench: A Benchmark for General-Purpose Foundation Models on Low-level Vision",
                    "abstract": "The rapid evolution of Multi-modality Large Language Models (MLLMs) has catalyzed a shift in computer vision from specialized models to general-purpose foundation models. Nevertheless, there is still an inadequacy in assessing the abilities of MLLMs on low-level visual perception and understanding. To address this gap, we present Q-Bench, a holistic benchmark crafted to systematically evaluate potential abilities of MLLMs on three realms: low-level visual perception, low-level visual description, and overall visual quality assessment. a) To evaluate the low-level perception ability, we construct the LLVisionQA dataset, consisting of 2,990 diverse-sourced images, each equipped with a human-asked question focusing on its low-level attributes. We then measure the correctness of MLLMs on answering these questions. b) To examine the description ability of MLLMs on low-level information, we propose the LLDescribe dataset consisting of long expert-labelled golden low-level text descriptions on 499 images, and a GPT-involved comparison pipeline between outputs of MLLMs and the golden descriptions. c) Besides these two tasks, we further measure their visual quality assessment ability to align with human opinion scores. Specifically, we design a softmax-based strategy that enables MLLMs to predict quantifiable quality scores, and evaluate them on various existing image quality assessment (IQA) datasets. Our evaluation across the three abilities confirms that MLLMs possess preliminary low-level visual skills. However, these skills are still unstable and relatively imprecise, indicating the need for specific enhancements on MLLMs towards these abilities. We hope that our benchmark can encourage the research community to delve deeper to discover and enhance these untapped potentials of MLLMs. Project Page: https://q-future.github.io/Q-Bench."
                },
                {
                    "title": "Valley: Video Assistant with Large Language model Enhanced abilitY",
                    "abstract": "Large language models (LLMs), with their remarkable conversational capabilities, have demonstrated impressive performance across various applications and have emerged as formidable AI assistants. In view of this, it raises an intuitive question: Can we harness the power of LLMs to build multimodal AI assistants for visual applications? Recently, several multi-modal models have been developed for this purpose. They typically pre-train an adaptation module to align the semantics of the vision encoder and language model, followed by fine-tuning on instruction-following data. However, despite the success of this pipeline in image and language understanding, its effectiveness in joint video and language understanding has not been widely explored. In this paper, we aim to develop a novel multi-modal foundation model capable of comprehending video, image, and language within a general framework. To achieve this goal, we introduce Valley, a Video Assistant with Large Language model Enhanced abilitY. The Valley consists of a LLM, a temporal modeling module, a visual encoder, and a simple projection module designed to bridge visual and textual modes. To empower Valley with video comprehension and instruction-following capabilities, we construct a video instruction dataset and adopt a two-stage tuning procedure to train it. Specifically, we employ ChatGPT to facilitate the construction of task-oriented conversation data encompassing various tasks, including multi-shot captions, long video descriptions, action recognition, causal relationship inference, etc. Subsequently, we adopt a pre-training-then-instructions-tuned pipeline to align visual and textual modalities and improve the instruction-following capability of Valley. Qualitative experiments demonstrate that Valley has the potential to function as a highly effective video assistant that can make complex video understanding scenarios easy."
                },
                {
                    "title": "Video-ChatGPT: Towards Detailed Video Understanding via Large Vision and Language Models",
                    "abstract": "Conversation agents fueled by Large Language Models (LLMs) are providing a new way to interact with visual data. While there have been initial attempts for image-based conversation models, this work addresses the under-explored field of \\emph{video-based conversation} by introducing Video-ChatGPT. It is a multimodal model that merges a video-adapted visual encoder with an LLM. The resulting model is capable of understanding and generating detailed conversations about videos. We introduce a new dataset of 100,000 video-instruction pairs used to train Video-ChatGPT acquired via manual and semi-automated pipeline that is easily scalable and robust to label noise. We also develop a quantitative evaluation framework for video-based dialogue models to objectively analyze the strengths and weaknesses of video-based dialogue models. Code: https://github.com/mbzuai-oryx/Video-ChatGPT."
                },
                {
                    "title": "Video-LLaMA: An Instruction-tuned Audio-Visual Language Model for Video Understanding",
                    "abstract": "We present Video-LLaMA a multi-modal framework that empowers Large Language Models (LLMs) with the capability of understanding both visual and auditory content in the video. Video-LLaMA bootstraps cross-modal training from the frozen pre-trained visual and audio encoders and the frozen LLMs. Unlike previous works that complement LLMs to process the visual or audio signals only, Video-LLaMA enables video comprehension by tackling two challenges: (1) capturing the temporal changes in visual scenes, (2) integrating audio-visual signals. To counter the first challenge, we propose a Video Q-former to assemble a pre-trained image encoder into our video encoder and introduce a video-to-text generation task to learn video-language correspondence. For the second challenge, we leverage ImageBind, a universal embedding model aligning multiple modalities, as the pre-trained audio encoder and introduce an Audio Q-former on top of ImageBind to learn reasonable auditory query embeddings for the LLM module. To align the output of both visual and audio encoders with LLM's embedding space, we first train Video-LLaMA on massive video/image-caption pairs and then tune our model with visual-instruction datasets of moderate amount but higher quality. We found Video-LLaMA shows the ability to perceive and comprehend video content and generate meaningful responses grounded in the visual and auditory information presented in the videos."
                },
                {
                    "title": "VideoChat: Chat-Centric Video Understanding",
                    "abstract": "In this paper, we initiate an attempt of developing an end-to-end chat-centric video understanding system, coined as VideoChat. It integrates video foundation models and large language models via a learnable neural interface, excelling in spatiotemporal reasoning, event localization, and causal relationship inference. To instructively tune this system, we build a video-centric instruction dataset, composed of thousands of videos associated with detailed descriptions and conversations. This dataset emphasizes spatiotemporal reasoning and captures causal relationships, providing a valuable asset for training our chat-centric video understanding system. Preliminary qualitative experiments demonstrate the potential of our system across a broad spectrum of video applications, which could serve as a simple prototype system for future research on chat-centric video understanding. Access our code and data at https://github.com/OpenGVLab/Ask-Anything"
                },
                {
                    "title": "Otter: A Multi-Modal Model with In-Context Instruction Tuning",
                    "abstract": "Large language models (LLMs) have demonstrated significant universal capabilities as few/zero-shot learners in various tasks due to their pre-training on vast amounts of text data, as exemplified by GPT-3, which boosted to InstrctGPT and ChatGPT, effectively following natural language instructions to accomplish real-world tasks. In this paper, we propose to introduce instruction tuning into multi-modal models, motivated by the Flamingo model's upstream interleaved format pretraining dataset. We adopt a similar approach to construct our MultI-Modal In-Context Instruction Tuning (MIMIC-IT) dataset. We then introduce Otter, a multi-modal model based on OpenFlamingo (open-sourced version of DeepMind's Flamingo), trained on MIMIC-IT and showcasing improved instruction-following ability and in-context learning. We also optimize OpenFlamingo's implementation for researchers, democratizing the required training resources from 1$\\times$ A100 GPU to 4$\\times$ RTX-3090 GPUs, and integrate both OpenFlamingo and Otter into Huggingface Transformers for more researchers to incorporate the models into their customized training and inference pipelines."
                },
                {
                    "title": "mPLUG-Owl: Modularization Empowers Large Language Models with Multimodality",
                    "abstract": "Large language models (LLMs) have demonstrated impressive zero-shot abilities on a variety of open-ended tasks, while recent research has also explored the use of LLMs for multi-modal generation. In this study, we introduce mPLUG-Owl, a novel training paradigm that equips LLMs with multi-modal abilities through modularized learning of foundation LLM, a visual knowledge module, and a visual abstractor module. This approach can support multiple modalities and facilitate diverse unimodal and multimodal abilities through modality collaboration. The training paradigm of mPLUG-Owl involves a two-stage method for aligning image and text, which learns visual knowledge with the assistance of LLM while maintaining and even improving the generation abilities of LLM. In the first stage, the visual knowledge module and abstractor module are trained with a frozen LLM module to align the image and text. In the second stage, language-only and multi-modal supervised datasets are used to jointly fine-tune a low-rank adaption (LoRA) module on LLM and the abstractor module by freezing the visual knowledge module. We carefully build a visually-related instruction evaluation set OwlEval. Experimental results show that our model outperforms existing multi-modal models, demonstrating mPLUG-Owl's impressive instruction and visual understanding ability, multi-turn conversation ability, and knowledge reasoning ability. Besides, we observe some unexpected and exciting abilities such as multi-image correlation and scene text understanding, which makes it possible to leverage it for harder real scenarios, such as vision-only document comprehension. Our code, pre-trained model, instruction-tuned models, and evaluation set are available at https://github.com/X-PLUG/mPLUG-Owl. The online demo is available at https://www.modelscope.cn/studios/damo/mPLUG-Owl."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "Video ChatCaptioner: Towards Enriched Spatiotemporal Descriptions",
                    "abstract": "Video captioning aims to convey dynamic scenes from videos using natural language, facilitating the understanding of spatiotemporal information within our environment. Although there have been recent advances, generating detailed and enriched video descriptions continues to be a substantial challenge. In this work, we introduce Video ChatCaptioner, an innovative approach for creating more comprehensive spatiotemporal video descriptions. Our method employs a ChatGPT model as a controller, specifically designed to select frames for posing video content-driven questions. Subsequently, a robust algorithm is utilized to answer these visual queries. This question-answer framework effectively uncovers intricate video details and shows promise as a method for enhancing video content. Following multiple conversational rounds, ChatGPT can summarize enriched video content based on previous conversations. We qualitatively demonstrate that our Video ChatCaptioner can generate captions containing more visual details about the videos. The code is publicly available at https://github.com/Vision-CAIR/ChatCaptioner"
                },
                {
                    "title": "LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention",
                    "abstract": "We present LLaMA-Adapter, a lightweight adaption method to efficiently fine-tune LLaMA into an instruction-following model. Using 52K self-instruct demonstrations, LLaMA-Adapter only introduces 1.2M learnable parameters upon the frozen LLaMA 7B model, and costs less than one hour for fine-tuning on 8 A100 GPUs. Specifically, we adopt a set of learnable adaption prompts, and prepend them to the word tokens at higher transformer layers. Then, a zero-initialized attention mechanism with zero gating is proposed, which adaptively injects the new instructional cues into LLaMA, while effectively preserves its pre-trained knowledge. With our efficient training, LLaMA-Adapter can generate high-quality responses, comparable to Alpaca with fully fine-tuned 7B parameters. Besides language commands, our approach can be simply extended to multi-modal instructions for learning image-conditioned LLaMA model, which achieves superior reasoning performance on ScienceQA and COCO Caption benchmarks. Furthermore, we also evaluate the zero-initialized attention mechanism for fine-tuning other pre-trained models (ViT, RoBERTa) on traditional vision and language tasks, demonstrating the superior generalization capacity of our approach. Code is released at https://github.com/OpenGVLab/LLaMA-Adapter."
                },
                {
                    "title": "Adding Conditional Control to Text-to-Image Diffusion Models",
                    "abstract": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "Muse: Text-To-Image Generation via Masked Generative Transformers",
                    "abstract": "We present Muse, a text-to-image Transformer model that achieves state-of-the-art image generation performance while being significantly more efficient than diffusion or autoregressive models. Muse is trained on a masked modeling task in discrete token space: given the text embedding extracted from a pre-trained large language model (LLM), Muse is trained to predict randomly masked image tokens. Compared to pixel-space diffusion models, such as Imagen and DALL-E 2, Muse is significantly more efficient due to the use of discrete tokens and requiring fewer sampling iterations; compared to autoregressive models, such as Parti, Muse is more efficient due to the use of parallel decoding. The use of a pre-trained LLM enables fine-grained language understanding, translating to high-fidelity image generation and the understanding of visual concepts such as objects, their spatial relationships, pose, cardinality etc. Our 900M parameter model achieves a new SOTA on CC3M, with an FID score of 6.06. The Muse 3B parameter model achieves an FID of 7.88 on zero-shot COCO evaluation, along with a CLIP score of 0.32. Muse also directly enables a number of image editing applications without the need to fine-tune or invert the model: inpainting, outpainting, and mask-free editing. More results are available at https://muse-model.github.io"
                },
                {
                    "title": "Tune-A-Video: One-Shot Tuning of Image Diffusion Models for Text-to-Video Generation",
                    "abstract": "To replicate the success of text-to-image (T2I) generation, recent works employ large-scale video datasets to train a text-to-video (T2V) generator. Despite their promising results, such paradigm is computationally expensive. In this work, we propose a new T2V generation setting\u2014One-Shot Video Tuning, where only one text-video pair is presented. Our model is built on state-of-the-art T2I diffusion models pre-trained on massive image data. We make two key observations: 1) T2I models can generate still images that represent verb terms; 2) extending T2I models to generate multiple images concurrently exhibits surprisingly good content consistency. To further learn continuous motion, we introduce Tune-A-Video, which involves a tailored spatio-temporal attention mechanism and an efficient one-shot tuning strategy. At inference, we employ DDIM inversion to provide structure guidance for sampling. Extensive qualitative and numerical experiments demonstrate the remarkable ability of our method across various applications."
                },
                {
                    "title": "Scalable Diffusion Models with Transformers",
                    "abstract": "We explore a new class of diffusion models based on the transformer architecture. We train latent diffusion models of images, replacing the commonly-used U-Net backbone with a transformer that operates on latent patches. We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops. We find that DiTs with higher Gflops\u2014through increased transformer depth/width or increased number of input tokens\u2014consistently have lower FID. In addition to possessing good scalability properties, our largest DiT-XL/2 models outperform all prior diffusion models on the class-conditional ImageNet 512\u00d7512 and 256\u00d7256 benchmarks, achieving a state-of-the-art FID of 2.27 on the latter."
                },
                {
                    "title": "MagicVideo: Efficient Video Generation With Latent Diffusion Models",
                    "abstract": "We present an efficient text-to-video generation framework based on latent diffusion models, termed MagicVideo. MagicVideo can generate smooth video clips that are concordant with the given text descriptions. Due to a novel and efficient 3D U-Net design and modeling video distributions in a low-dimensional space, MagicVideo can synthesize video clips with 256x256 spatial resolution on a single GPU card, which takes around 64x fewer computations than the Video Diffusion Models (VDM) in terms of FLOPs. In specific, unlike existing works that directly train video models in the RGB space, we use a pre-trained VAE to map video clips into a low-dimensional latent space and learn the distribution of videos' latent codes via a diffusion model. Besides, we introduce two new designs to adapt the U-Net denoiser trained on image tasks to video data: a frame-wise lightweight adaptor for the image-to-video distribution adjustment and a directed temporal attention module to capture temporal dependencies across frames. Thus, we can exploit the informative weights of convolution operators from a text-to-image model for accelerating video training. To ameliorate the pixel dithering in the generated videos, we also propose a novel VideoVAE auto-encoder for better RGB reconstruction. We conduct extensive experiments and demonstrate that MagicVideo can generate high-quality video clips with either realistic or imaginary content. Refer to \\url{https://magicvideo.github.io/#} for more examples."
                },
                {
                    "title": "Imagen Video: High Definition Video Generation with Diffusion Models",
                    "abstract": "We present Imagen Video, a text-conditional video generation system based on a cascade of video diffusion models. Given a text prompt, Imagen Video generates high definition videos using a base video generation model and a sequence of interleaved spatial and temporal video super-resolution models. We describe how we scale up the system as a high definition text-to-video model including design decisions such as the choice of fully-convolutional temporal and spatial super-resolution models at certain resolutions, and the choice of the v-parameterization of diffusion models. In addition, we confirm and transfer findings from previous work on diffusion-based image generation to the video generation setting. Finally, we apply progressive distillation to our video models with classifier-free guidance for fast, high quality sampling. We find Imagen Video not only capable of generating videos of high fidelity, but also having a high degree of controllability and world knowledge, including the ability to generate diverse videos and text animations in various artistic styles and with 3D object understanding. See https://imagen.research.google/video/ for samples."
                },
                {
                    "title": "Make-A-Video: Text-to-Video Generation without Text-Video Data",
                    "abstract": "We propose Make-A-Video -- an approach for directly translating the tremendous recent progress in Text-to-Image (T2I) generation to Text-to-Video (T2V). Our intuition is simple: learn what the world looks like and how it is described from paired text-image data, and learn how the world moves from unsupervised video footage. Make-A-Video has three advantages: (1) it accelerates training of the T2V model (it does not need to learn visual and multimodal representations from scratch), (2) it does not require paired text-video data, and (3) the generated videos inherit the vastness (diversity in aesthetic, fantastical depictions, etc.) of today's image generation models. We design a simple yet effective way to build on T2I models with novel and effective spatial-temporal modules. First, we decompose the full temporal U-Net and attention tensors and approximate them in space and time. Second, we design a spatial temporal pipeline to generate high resolution and frame rate videos with a video decoder, interpolation model and two super resolution models that can enable various applications besides T2V. In all aspects, spatial and temporal resolution, faithfulness to text, and quality, Make-A-Video sets the new state-of-the-art in text-to-video generation, as determined by both qualitative and quantitative measures."
                },
                {
                    "title": "DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation",
                    "abstract": "Large text-to-image models achieved a remarkable leap in the evolution of AI, enabling high-quality and diverse synthesis of images from a given text prompt. However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts. In this work, we present a new approach for \u201cpersonalization\u201d of text-to-image diffusion models. Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject. Once the subject is embedded in the output domain of the model, the unique identifier can be used to synthesize novel photorealistic images of the subject contextualized in different scenes. By leveraging the semantic prior embedded in the model with a new autogenous class-specific prior preservation loss, our technique enables synthesizing the subject in diverse scenes, poses, views and lighting conditions that do not appear in the reference images. We apply our technique to several previously-unassailable tasks, including subject recontextualization, text-guided view synthesis, and artistic rendering, all while preserving the subject's key features. We also provide a new dataset and evaluation protocol for this new task of subject-driven generation. Project page: https://dreambooth.github.io/"
                },
                {
                    "title": "CogVideo: Large-scale Pretraining for Text-to-Video Generation via Transformers",
                    "abstract": "Large-scale pretrained transformers have created milestones in text (GPT-3) and text-to-image (DALL-E and CogView) generation. Its application to video generation is still facing many challenges: The potential huge computation cost makes the training from scratch unaffordable; The scarcity and weak relevance of text-video datasets hinder the model understanding complex movement semantics. In this work, we present 9B-parameter transformer CogVideo, trained by inheriting a pretrained text-to-image model, CogView2. We also propose multi-frame-rate hierarchical training strategy to better align text and video clips. As (probably) the first open-source large-scale pretrained text-to-video model, CogVideo outperforms all publicly available models at a large margin in machine and human evaluations."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Ego4D: Around the World in 3,000 Hours of Egocentric Video",
                    "abstract": "We introduce Ego4D, a massive-scale egocentric video dataset and benchmark suite. It offers 3,670 hours of dailylife activity video spanning hundreds of scenarios (household, outdoor, workplace, leisure, etc.) captured by 931 unique camera wearers from 74 worldwide locations and 9 different countries. The approach to collection is designed to uphold rigorous privacy and ethics standards, with consenting participants and robust de-identification procedures where relevant. Ego4D dramatically expands the volume of diverse egocentric video footage publicly available to the research community. Portions of the video are accompanied by audio, 3D meshes of the environment, eye gaze, stereo, and/or synchronized videos from multiple egocentric cameras at the same event. Furthermore, we present a host of new benchmark challenges centered around understanding the first-person visual experience in the past (querying an episodic memory), present (analyzing hand-object manipulation, audio-visual conversation, and social interactions), and future (forecasting activities). By publicly sharing this massive annotated dataset and benchmark suite, we aim to push the frontier of first-person perception. Project page: https://ego4d-data.org/"
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "NExT-QA: Next Phase of Question-Answering to Explaining Temporal Actions",
                    "abstract": "We introduce NExT-QA, a rigorously designed video question answering (VideoQA) benchmark to advance video understanding from describing to explaining the temporal actions. Based on the dataset, we set up multi-choice and open-ended QA tasks targeting causal action reasoning, temporal action reasoning, and common scene comprehension. Through extensive analysis of baselines and established VideoQA techniques, we find that top-performing methods excel at shallow scene descriptions but are weak in causal and temporal action reasoning. Furthermore, the models that are effective on multi-choice QA, when adapted to open-ended QA, still struggle in generalizing the answers. This raises doubt on the ability of these models to reason and highlights possibilities for improvement. With detailed results for different question types and heuristic observations for future works, we hope NExT-QA will guide the next generation of VQA research to go beyond superficial description towards a deeper understanding of videos. (The dataset and related resources are available at https://github.com/doc-doc/NExT-QA.git)."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "CLEVRER: CoLlision Events for Video REpresentation and Reasoning",
                    "abstract": "The ability to reason about temporal and causal events from videos lies at the core of human intelligence. Most video reasoning benchmarks, however, focus on pattern recognition from complex visual and language input, instead of on causal structure. We study the complementary problem, exploring the temporal and causal structures behind videos of objects with simple visual appearance. To this end, we introduce the CoLlision Events for Video REpresentation and Reasoning (CLEVRER), a diagnostic video dataset for systematic evaluation of computational models on a wide range of reasoning tasks. Motivated by the theory of human casual judgment, CLEVRER includes four types of questions: descriptive (e.g., \"what color\"), explanatory (\"what is responsible for\"), predictive (\"what will happen next\"), and counterfactual (\"what if\"). We evaluate various state-of-the-art models for visual reasoning on our benchmark. While these models thrive on the perception-based task (descriptive), they perform poorly on the causal tasks (explanatory, predictive and counterfactual), suggesting that a principled approach for causal reasoning should incorporate the capability of both perceiving complex visual and language inputs, and understanding the underlying dynamics and causal relations. We also study an oracle model that explicitly combines these components via symbolic representations."
                },
                {
                    "title": "ActivityNet-QA: A Dataset for Understanding Complex Web Videos via Question Answering",
                    "abstract": "Recent developments in modeling language and vision have been successfully applied to image question answering. It is both crucial and natural to extend this research direction to the video domain for video question answering (VideoQA). Compared to the image domain where large scale and fully annotated benchmark datasets exists, VideoQA datasets are limited to small scale and are automatically generated, etc. These limitations restrict their applicability in practice. Here we introduce ActivityNet-QA, a fully annotated and large scale VideoQA dataset. The dataset consists of 58,000 QA pairs on 5,800 complex web videos derived from the popular ActivityNet dataset. We present a statistical analysis of our ActivityNet-QA dataset and conduct extensive experiments on it by comparing existing VideoQA baselines. Moreover, we explore various video representation strategies to improve VideoQA performance, especially for long videos."
                },
                {
                    "title": "BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning",
                    "abstract": "Datasets drive vision progress, yet existing driving datasets are impoverished in terms of visual content and supported tasks to study multitask learning for autonomous driving. Researchers are usually constrained to study a small set of problems on one dataset, while real-world computer vision applications require performing tasks of various complexities. We construct BDD100K, the largest driving video dataset with 100K videos and 10 tasks to evaluate the exciting progress of image recognition algorithms on autonomous driving. The dataset possesses geographic, environmental, and weather diversity, which is useful for training models that are less likely to be surprised by new conditions. Based on this diverse dataset, we build a benchmark for heterogeneous multitask learning and study how to solve the tasks together. Our experiments show that special training strategies are needed for existing models to perform such heterogeneous tasks. BDD100K opens the door for future studies in this important venue."
                },
                {
                    "title": "Video Question Answering via Gradually Refined Attention over Appearance and Motion",
                    "abstract": "Recently image question answering (ImageQA) has gained lots of attention in the research community. However, as its natural extension, video question answering (VideoQA) is less explored. Although both tasks look similar, VideoQA is more challenging mainly because of the complexity and diversity of videos. As such, simply extending the ImageQA methods to videos is insufficient and suboptimal. Particularly, working with the video needs to model its inherent temporal structure and analyze the diverse information it contains. In this paper, we consider exploiting the appearance and motion information resided in the video with a novel attention mechanism. More specifically, we propose an end-to-end model which gradually refines its attention over the appearance and motion features of the video using the question as guidance. The question is processed word by word until the model generates the final optimized attention. The weighted representation of the video, as well as other contextual information, are used to generate the answer. Extensive experiments show the advantages of our model compared to other baseline models. We also demonstrate the effectiveness of our model by analyzing the refined attention weights during the question answering procedure."
                },
                {
                    "title": "TGIF: A New Dataset and Benchmark on Animated GIF Description",
                    "abstract": "With the recent popularity of animated GIFs on social media, there is need for ways to index them with rich meta-data. To advance research on animated GIF understanding, we collected a new dataset, Tumblr GIF (TGIF), with 100K animated GIFs from Tumblr and 120K natural language descriptions obtained via crowdsourcing. The motivation for this work is to develop a testbed for image sequence description systems, where the task is to generate natural language descriptions for animated GIFs or video clips. To ensure a high quality dataset, we developed a series of novel quality controls to validate free-form text input from crowd-workers. We show that there is unambiguous association between visual content and natural language descriptions in our dataset, making it an ideal benchmark for the visual content captioning task. We perform extensive statistical analyses to compare our dataset to existing image and video description datasets. Next, we provide baseline results on the animated GIF description task, using three representative techniques: nearest neighbor, statistical machine translation, and recurrent neural networks. Finally, we show that models fine-tuned from our animated GIF description dataset can be helpful for automatic movie description."
                },
                {
                    "title": "Efficient key-frame extraction and video analysis",
                    "abstract": "Content-based video indexing and retrieval has its foundations in the analyses of the prime video temporal structures. Consequently, technologies for video segmentation and key-frame extraction have become crucial for the development of advanced digital video systems. Conventional algorithms for video partitioning and key-frame extraction are mainly implemented autonomously. By focusing the analysis on compressed video features, this paper introduces a real-time algorithm for scene change detection and key-frame extraction that generates frame difference metrics by analysing the statistics of the macro-block features extracted from an MPEG compressed stream. The key-frame extraction method is implemented using difference metrics in curve simplification by means of a discrete contour evolution algorithm. This approach resulted in a fast and robust algorithm. Results of computer simulations are reported."
                },
                {
                    "title": "Adaptive key frame extraction using unsupervised clustering",
                    "abstract": "Key frame extraction has been recognized as one of the important research issues in video information retrieval. Although progress has been made in key frame extraction, the existing approaches are either computationally expensive or ineffective in capturing salient visual content. We first discuss the importance of key frame selection; and then review and evaluate the existing approaches. To overcome the shortcomings of the existing approaches, we introduce a new algorithm for key frame extraction based on unsupervised clustering. The proposed algorithm is both computationally simple and able to adapt to the visual content. The efficiency and effectiveness are validated by large amount of real-world videos."
                },
                {
                    "title": "Qwen-VL: A Frontier Large Vision-Language Model with Versatile Abilities",
                    "abstract": "We introduce the Qwen-VL series, a set of large-scale vision-language models designed to perceive and understand both text and images. Comprising Qwen-VL and Qwen-VL-Chat, these models exhibit remarkable performance in tasks like image captioning, question answering, visual localization, and flexible interaction. The evaluation covers a wide range of tasks including zero-shot captioning, visual or document visual question answering, and grounding. We demonstrate the Qwen-VL outperforms existing Large Vision Language Models (LVLMs). We present their architecture, training, capabilities, and performance, highlighting their contributions to advancing multimodal artificial intelligence. Code, demo and models are available at https://github.com/QwenLM/Qwen-VL ."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Improving Image Generation with Better Captions",
                    "abstract": "We show that prompt following abilities of text-to-image models can be substantially improved by training on highly descriptive generated image captions. Existing text-to-image models struggle to follow detailed image descriptions and often ignore words or confuse the meaning of prompts. We hypothesize that this issue stems from noisy and inaccurate image captions in the training dataset. We address this by training a bespoke image captioner and use it to recaption the training dataset. We then train several text-to-image models and find that training on these synthetic captions reliably improves prompt following ability. Finally, we use these findings to build DALL-E 3: a new text-to-image generation system, and benchmark its performance on an evaluation designed to measure prompt following, coherence, and aesthetics, finding that it compares favorably to competitors. We publish samples and code for these evaluations so that future research can continue optimizing this important aspect of text-to-image systems."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively generate detailed and high-fidelity captions for arbitrary-length videos that accurately reflect both inter-frame temporal changes and intra-frame content descriptions?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of generating detailed video captions is crucial for advancing multi-modal learning, particularly in video understanding and generation tasks. High-quality captions can enhance the performance of models in image-text dialogue and text-to-image generation, leading to improved user interaction across video and language modalities. This research could pave the way for more sophisticated applications in areas such as video summarization, content creation, and accessibility, ultimately enriching the research community's understanding of video data and its potential uses.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in generating detailed video captions stem from three main aspects: 1) Understanding precise temporal changes between frames is complex, as inaccuracies can lead to poor caption quality. 2) Providing detailed content descriptions for each frame is essential for effective video-text alignment, yet difficult to achieve consistently. 3) The variability in video lengths necessitates a scalable approach that can handle arbitrary-length videos, which is not straightforward. Naive methods, such as treating video frames as static images or concatenating frames, fail to capture the dynamic nature of video content and often result in loss of detail and coherence.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on short captions for videos, neglecting the need for detailed descriptions. Existing solutions have been limited by the lack of effective large-scale video captioning models and the challenges of human annotation for long videos. Additionally, current open-source models do not support video inputs, and closed-source APIs have not yet addressed this gap. Our approach, the Differential Sliding-Window Captioning strategy (DiffSW), improves upon prior work by translating the captioning task into a differential description task, allowing for a more nuanced understanding of temporal changes and detailed content.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, DiffSW, involves generating a detailed caption for the first frame of a video and then applying a sliding window of length two to subsequent frames. The model, GPT4V, will identify changes between frames based on three inputs: the previous frame, its differential caption, and the current frame. This approach addresses the challenges of inter-frame temporal understanding, intra"
            }
        },
        "author_data": {
            "6f14b29e-7417-4aa3-b375-f45dcb8352b8": {
                "pk": "6f14b29e-7417-4aa3-b375-f45dcb8352b8",
                "name": "Lin Chen",
                "collaborators": [
                    "Qingtao Chen",
                    "Yutong Jiang",
                    "Lin Zhang",
                    "Kaifeng Bu"
                ],
                "domain": [
                    "Quantum Physics",
                    "Algebraic Geometry",
                    "Representation Theory",
                    "Mathematical Physics"
                ],
                "publications": [
                    {
                        "title": "Symmetrized cut-join equation of Marino-Vafa formula",
                        "abstract": "In this note, we symmetrized the cut-join equation from the proof of Marino-Vafa formula by applying a transcendental changing of variable. One can derive more recursion formulas of Hodge integrals out of this polynomial equations. We also give some applications."
                    },
                    {
                        "title": "Bouchard-Klemm-Marino-Pasquetti Conjecture for $\\mathbb{C}^3$",
                        "abstract": "In this paper, we give a proof of the Bouchard-Klemm-Marino-Pasquetti conjecture for a framed vertex, by using the symmetrized Cut-Join Equation developed in a previous paper."
                    },
                    {
                        "title": "Length of separable states and symmetrical informationally complete (SIC) POVM",
                        "abstract": "This short note reviews the notion and fundamental properties of SIC-POVM and its connection with the length of separable states. We also review the t-design."
                    },
                    {
                        "title": "Framed Deformation of Galois Representation",
                        "abstract": "We studied framed deformations of two dimensional Galois representation of which the residue representation restrict to decomposition groups are scalars, and established a modular lifting theorem for certain cases. We then proved a family version of the result, and used it to determine the structure of deformation rings over characteristic zero fields. These can be applied to the study of exceptional zero of p-adic L-function."
                    },
                    {
                        "title": "Deligne--Lusztig duality on the moduli stack of bundles",
                        "abstract": "Let $Bun_G(X)$ be the moduli stack of $G$-torsors on a smooth projective curve $X$ for a reductive group $G$. We prove a conjecture made by Drinfeld-Wang and Gaitsgory on the Deligne-Lusztig duality for D-modules on $Bun_G(X)$. This conjecture relates Drinfeld-Gaitsgory's pseudo-identity functors to the enhanced Eisenstein series and geometric constant term functors on $DMod(Bun_G(X))$. We also prove a \"second adjointness\" result for these enhanced functors."
                    },
                    {
                        "title": "Nearby cycles on Drinfeld-Gaitsgory-Vinberg Interpolation Grassmannian and long intertwining functor",
                        "abstract": "Let $G$ be a reductive group and $U,U^-$ be the unipotent radicals of a pair of opposite parabolic subgroups $P,P^-$. We prove that the DG-categories of $U(\\!(t)\\!)$-equivariant and $U^-(\\!(t)\\!)$-equivariant D-modules on the affine Grassmannian $Gr_G$ are canonically dual to each other. We show that the unit object witnessing this duality is given by nearby cycles on the Drinfeld-Gaitsgory-Vinberg interpolation Grassmannian defined in arXiv:1805.07721. We study various properties of the mentioned nearby cycles, in particular compare them with the nearby cycles studied in arXiv:1411.4206 and arXiv:1607.00586. We also generalize our results to the Beilinson-Drinfeld Grassmannian $Gr_{G,X^I}$ and to the affine flag variety $Fl_G$."
                    },
                    {
                        "title": "Orthogonal Quantum Group Invariants of Links",
                        "abstract": "We study the Chern-Simons partition function of orthogonal quantum group invariants, and propose a new orthogonal Labastida-Mari\\~{n}o-Ooguri-Vafa conjecture as well as degree conjecture for free energy associated to the orthogonal Chern-Simons partition function. We prove the degree conjecture and some interesting cases of orthogonal LMOV conjecture. In particular, We provide a formula of colored Kauffman polynomials for torus knots and links, and applied this formula to verify certain case of the conjecture at roots of unity except $1$. We also derive formulas of Lickorish-Millett type for Kauffman polynomials and relate all these to the orthogonal LMOV conjecture."
                    },
                    {
                        "title": "Nonlocality via multiqubit orthogonal product bases",
                        "abstract": "We investigate the quantum nonlocality via the discrimination on two, three and four-qubit orthogonal product bases (OPBs). We show that every two-qubit, and some three and four-qubit OPBs can be locally distinguished. It turns out that the remaining three and four-qubit OPBs cannot be locally distinguished, and thus they show the quantum nonlocality without entanglement. We also distinguish them by merging some systems using quantum teleportation with assisted Bell states."
                    },
                    {
                        "title": "Fidelity between a bipartite state and another one undergoing local unitary dynamics",
                        "abstract": "The fidelity and local unitary transformation are two widely useful notions in quantum physics. We study two constrained optimization problems in terms of the maximal and minimal fidelity between two bipartite quantum states undergoing local unitary dynamics. The problems are related to the geometric measure of entanglement and the distillability problem. We show that the problems can be reduced to semi-definite programming optimization problems. We give close-form formulae of the fidelity when the two states are both pure states, or a pure product state and the Werner state. We explain from the point of view of local unitary actions that why the entanglement in Werner states is hard to accessible. For general mixed states, we give upper and lower bounds of the fidelity using tools such as affine fidelity, channels and relative entropy from information theory. We also investigate the power of local unitaries and quantification for the commutativity of quantum states, and the equivalence of the two optimization problems."
                    }
                ]
            },
            "84e3aaa6-c3b3-4198-9c41-bc2528e81aca": {
                "pk": "84e3aaa6-c3b3-4198-9c41-bc2528e81aca",
                "name": "Xilin Wei",
                "collaborators": [
                    "Beichen Zhang",
                    "Kun Zhou",
                    "Wayne Xin Zhao",
                    "Jing Sha",
                    "Shijin Wang",
                    "Ji-Rong Wen"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Reasoning",
                    "Dataset Development"
                ],
                "publications": [
                    {
                        "title": "Evaluating and Improving Tool-Augmented Computation-Intensive Math Reasoning",
                        "abstract": "Chain-of-thought prompting~(CoT) and tool augmentation have been validated in recent work as effective practices for improving large language models~(LLMs) to perform step-by-step reasoning on complex math-related tasks. However, most existing math reasoning datasets may be not able to fully evaluate and analyze the ability of LLMs in manipulating tools and performing reasoning, as they may only require very few invocations of tools or miss annotations for evaluating intermediate reasoning steps. To address the issue, we construct \\textbf{CARP}, a new Chinese dataset consisting of 4,886 computation-intensive algebra problems with formulated annotations on intermediate steps. In CARP, we test four LLMs with CoT prompting, and find that they are all prone to make mistakes at the early steps of the solution, leading to wrong answers. Based on this finding, we propose a new approach that can deliberate the reasoning steps with tool interfaces, namely \\textbf{DELI}. In DELI, we first initialize a step-by-step solution based on retrieved exemplars, then iterate two deliberation procedures that check and refine the intermediate steps of the generated solution, from the perspectives of tool manipulation and natural language reasoning, until obtaining converged solutions or reaching the maximum turn. Experimental results on CARP and six other datasets show that the proposed DELI mostly outperforms competitive baselines, and can further boost the performance of existing CoT methods. Our data and code are available in \\url{https://github.com/RUCAIBox/CARP}."
                    }
                ]
            },
            "9636c67e-8a17-4bcb-8545-a55dc1a98fc4": {
                "pk": "9636c67e-8a17-4bcb-8545-a55dc1a98fc4",
                "name": "Jinsong Li",
                "collaborators": [
                    "Haichou Li",
                    "Jinsong Liu",
                    "Hongyu Wang"
                ],
                "domain": [
                    "Complex Analysis",
                    "Carleson Measures",
                    "Pseudoconvex Domains",
                    "Function Spaces"
                ],
                "publications": [
                    {
                        "title": "Carleson measures on convex domains",
                        "abstract": "Following M.Abate and A.Saracco's work on strongly pseudoconvex domains in $\\mathbb{C}^n$, we characterize Carleson measures of $A^2(D)$ in bounded convex domains with smooth boundary of finite type. We also give examples of Carleson measures with uniformly discrete (with respect to the Kobayashi distance) sequences."
                    }
                ]
            },
            "4a805eee-f82d-4abf-a3db-a9b6db411604": {
                "pk": "4a805eee-f82d-4abf-a3db-a9b6db411604",
                "name": "Xiaoyi Dong",
                "collaborators": [
                    "Xuebing Wu",
                    "Weiming Zhang",
                    "Nenghai Yu",
                    "Michael M. De Robertis"
                ],
                "domain": [
                    "Astrophysics",
                    "Quasars",
                    "Black Hole",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Herschel observed Stripe 82 quasars and their host galaxies: connections between the AGN activity and the host galaxy star formation",
                        "abstract": "In this work, we present a study of 207 quasars selected from the Sloan Digital Sky Survey quasar catalogs and the Herschel Stripe 82 survey. Quasars within this sample are high luminosity quasars with a mean bolometric luminosity of $10^{46.4}$ erg s$^{-1}$. The redshift range of this sample is within $z<4$, with a mean value of $1.5\\pm0.78$. Because we only selected quasars that have been detected in all three Herschel-SPIRE bands, the quasar sample is complete yet highly biased. Based on the multi-wavelength photometric observation data, we conducted a spectral energy distribution (SED) fitting through UV to FIR. Parameters such as active galactic nucleus (AGN) luminosity, FIR luminosity, stellar mass, as well as many other AGN and galaxy properties are deduced from the SED fitting results. The mean star formation rate (SFR) of the sample is 419 $M_{\\odot}$ yr$^{-1}$ and the mean gas mass is $\\sim 10^{11.3}$ $M_{\\odot}$. All these results point to an IR luminous quasar system. Comparing with star formation main sequence (MS) galaxies, at least 80 out of 207 quasars are hosted by starburst galaxies. It supports the statement that luminous AGNs are more likely to be associated with major mergers. The SFR increases with the redshift up to $z=2$. It is correlated with the AGN bolometric luminosity, where $L_{\\rm FIR} \\propto L_{\\rm Bol}^{0.46\\pm0.03}$. The AGN bolometric luminosity is also correlated with the host galaxy mass and gas mass. Yet the correlation between $L_{\\rm FIR}$ and $L_{\\rm Bol}$ has higher significant level, implies that the link between AGN accretion and the SFR is more primal. The $M_{\\rm BH}/M_{\\ast}$ ratio of our sample is 0.02, higher than the value 0.005 in the local Universe. It might indicate an evolutionary trend of the $M_{\\rm BH} - M_{\\ast}$ scaling relation."
                    },
                    {
                        "title": "CAAD 2018: Powerful None-Access Black-Box Attack Based on Adversarial Transformation Network",
                        "abstract": "In this paper, we propose an improvement of Adversarial Transformation Networks(ATN) to generate adversarial examples, which can fool white-box models and black-box models with a state of the art performance and won the 2rd place in the non-target task in CAAD 2018."
                    },
                    {
                        "title": "Low-luminosity Active Galaxies and their Central Black Holes",
                        "abstract": "Central black hole masses for 118 spiral galaxies representing morphological stages S0/a through Sc and taken from the large spectroscopic survey of Ho, Filippenko & Sargent (1997) are derived using 2MASS Ks data. Black hole (BH) masses are found using a calibrated black-hole - Ks bulge luminosity relation, while bulge luminosities are measured using GALFIT, a two-dimensional bulge/disk decomposition routine.   The BH masses are correlated against a variety of nuclear and host-galaxy properties. Nuclear properties such as line width and line ratios show a very high degree of correlation with BH mass. The excellent correlation with line-width supports the view that the emission-line gas is in virial equilibrium with either the BH or bulge potential. The very good emission-line ratio correlations may indicate a change in ionizing continuum shape with BH mass in the sense that more massive BHs generate harder spectra.   Apart from the inclination-corrected rotational velocity, no excellent correlations are found between BH mass and host-galaxy properties.   Significant differences are found between the distributions of BH masses in early-, mid- and later-type spiral galaxies in the sense that early-type galaxies have preferentially larger central BHs. The line-width distributions show a marked difference among the subsamples in the sense that earlier-type galaxies have larger line widths. There are also clear differences in line ratios between subsamples likely related to the level of ionization in the gas. Finally, a Ks-band Simien & de Vaucouleurs diagram shows excellent agreement with the original B-band relation."
                    }
                ]
            },
            "c4f7fa8d-78a1-40c5-b9fb-77402b77e897": {
                "pk": "c4f7fa8d-78a1-40c5-b9fb-77402b77e897",
                "name": "Pan Zhang",
                "collaborators": [
                    "Feng Pan",
                    "Sujie Li",
                    "Pengfei Zhou",
                    "Liang Zhang",
                    "Keyang Chen",
                    "Changpeng Pan",
                    "Zhenghan Shen",
                    "Hanyan Cao",
                    "Yijia Wang"
                ],
                "domain": [
                    "Statistical Physics",
                    "Quantum Computing",
                    "Machine Learning",
                    "Graph Theory"
                ],
                "publications": [
                    {
                        "title": "Inference of kinetic Ising model on sparse graphs",
                        "abstract": "Based on dynamical cavity method, we propose an approach to the inference of kinetic Ising model, which asks to reconstruct couplings and external fields from given time-dependent output of original system. Our approach gives an exact result on tree graphs and a good approximation on sparse graphs, it can be seen as an extension of Belief Propagation inference of static Ising model to kinetic Ising model. While existing mean field methods to the kinetic Ising inference e.g., na\\\" ive mean-field, TAP equation and simply mean-field, use approximations which calculate magnetizations and correlations at time $t$ from statistics of data at time $t-1$, dynamical cavity method can use statistics of data at times earlier than $t-1$ to capture more correlations at different time steps. Extensive numerical experiments show that our inference method is superior to existing mean-field approaches on diluted networks."
                    },
                    {
                        "title": "Non-backtracking operator for Ising model and its application in attractor neural networks",
                        "abstract": "The non-backtracking operator was recently shown to give a redemption for spectral clustering in sparse graphs. In this paper we consider non-backtracking operator for Ising model on a general graph with a general coupling distribution by linearizing Belief Propagation algorithm at paramagnetic fixed-point. The spectrum of the operator is studied, the sharp edge of bulk and possible real eigenvalues outside the bulk are computed analytically as a function of couplings and temperature. We show the applications of the operator in attractor neural networks. At thermodynamic limit, our result recovers the phase boundaries of Hopfield model obtained by replica method. On single instances of Hopfield model, its eigenvectors can be used to retrieve all patterns simultaneously. We also give an example on how to control the neural networks, i.e. making network more sparse while keeping patterns stable, using the non-backtracking operator and matrix perturbation theory."
                    },
                    {
                        "title": "Robust Spectral Detection of Global Structures in the Data by Learning a Regularization",
                        "abstract": "Spectral methods are popular in detecting global structures in the given data that can be represented as a matrix. However when the data matrix is sparse or noisy, classic spectral methods usually fail to work, due to localization of eigenvectors (or singular vectors) induced by the sparsity or noise. In this work, we propose a general method to solve the localization problem by learning a regularization matrix from the localized eigenvectors. Using matrix perturbation analysis, we demonstrate that the learned regularizations suppress down the eigenvalues associated with localized eigenvectors and enable us to recover the informative eigenvectors representing the global structure. We show applications of our method in several inference problems: community detection in networks, clustering from pairwise similarities, rank estimation and matrix completion problems. Using extensive experiments, we illustrate that our method solves the localization problem and works down to the theoretical detectability limits in different kinds of synthetic data. This is in contrast with existing spectral algorithms based on data matrix, non-backtracking matrix, Laplacians and those with rank-one regularizations, which perform poorly in the sparse case with noise."
                    },
                    {
                        "title": "Spectral estimation of the percolation transition in clustered networks",
                        "abstract": "There have been several spectral bounds for the percolation transition in networks, using spectrum of matrices associated with the network such as the adjacency matrix and the non-backtracking matrix. However they are far from being tight when the network is sparse and displays clustering or transitivity, which is represented by existence of short loops e.g. triangles. In this work, for the bond percolation, we first propose a message passing algorithm for calculating size of percolating clusters considering effects of triangles, then relate the percolation transition to the leading eigenvalue of a matrix that we name the triangle-non-backtracking matrix, by analyzing stability of the message passing equations. We establish that our method gives a tighter lower-bound to the bond percolation transition than previous spectral bounds, and it becomes exact for an infinite network with no loops longer than 3. We evaluate numerically our methods on synthetic and real-world networks, and discuss further generalizations of our approach to include higher-order sub-structures."
                    },
                    {
                        "title": "Evaluating accuracy of community detection using the relative normalized mutual information",
                        "abstract": "The Normalized Mutual Information (NMI) has been widely used to evaluate the accuracy of community detection algorithms. However in this article we show that the NMI is seriously affected by systematic errors due to finite size of networks, and may give a wrong estimate of performance of algorithms in some cases. We give a simple theory to the finite-size effect of NMI and test our theory numerically. Then we propose a new metric for the accuracy of community detection, namely the relative Normalized Mutual Information (rNMI), which considers statistical significance of the NMI by comparing it with the expected NMI of random partitions. Our numerical experiments show that the rNMI overcomes the finite-size effect of the NMI."
                    },
                    {
                        "title": "Gradient Flows of Higher Order Yang-Mills-Higgs Functionals",
                        "abstract": "In this paper, we define a family of functionals generalizing the Yang-Mills-Higgs functional on a closed Riemannian manifold. Then we prove the short time existence of the corresponding gradient flow by a gauge fixing technique. The lack of maximal principle for the higher order operator brings us a lot of inconvenience during the estimates for the Higgs field. We observe that the $L^2$-bound of the Higgs field is enough for energy estimates in $4$ dimension, and we show that, provided the order of derivatives, appearing in the higher order Yang-Mills-Higgs functionals, is strictly greater than 1, solutions to the gradient flow do not hit any finite time singularities. As for the Yang-Mills-Higgs $k$-functional with Higgs self-interaction, we show that, provided $\\dim(M)<2(k+1)$, the associated gradient flow admits long time existence with smooth initial data. The proof depends on local $L^2$-derivative estimates, energy estimates and blow-up analysis."
                    },
                    {
                        "title": "Simulating the Sycamore quantum supremacy circuits",
                        "abstract": "We propose a general tensor network method for simulating quantum circuits. The method is massively more efficient in computing a large number of correlated bitstring amplitudes and probabilities than existing methods. As an application, we study the sampling problem of Google's Sycamore circuits, which are believed to be beyond the reach of classical supercomputers and have been used to demonstrate quantum supremacy. Using our method, employing a small computational cluster containing 60 graphical processing units (GPUs), we have generated one million correlated bitstrings with some entries fixed, from the Sycamore circuit with 53 qubits and 20 cycles, with linear cross-entropy benchmark (XEB) fidelity equals 0.739, which is much higher than those in Google's quantum supremacy experiments."
                    },
                    {
                        "title": "Minimality of a Kind of Pseudo-Umbilical Totally Real Submanifolds in Non-Flat Complex Space Forms",
                        "abstract": "In this paper, by studying the position of umbilical normal vectors in the normal bundle, we prove that pseudo-umbilical totally real submanifolds with flat normal connection in non-flat complex space forms must be minimal."
                    },
                    {
                        "title": "Solving the sampling problem of the Sycamore quantum circuits",
                        "abstract": "We study the problem of generating independent samples from the output distribution of Google's Sycamore quantum circuits with a target fidelity, which is believed to be beyond the reach of classical supercomputers and has been used to demonstrate quantum supremacy. We propose a new method to classically solve this problem by contracting the corresponding tensor network just once, and is massively more efficient than existing methods in obtaining a large number of uncorrelated samples with a target fidelity. For the Sycamore quantum supremacy circuit with $53$ qubits and $20$ cycles, we have generated one million uncorrelated bitstrings $\\{\\mathbf s\\}$ which are sampled from a distribution $\\hat P(\\mathbf s)=|\\hat \\psi(\\mathbf s)|^2$, where the approximate state $\\hat \\psi$ has fidelity $F\\approx 0.0037$. The whole computation has cost about $15$ hours on a computational cluster with $512$ GPUs. The obtained one million samples, the contraction code and contraction order is made public. If our algorithm could be implemented with high efficiency on a modern supercomputer with ExaFLOPS performance, we estimate that ideally, the simulation would cost a few dozens of seconds, which is faster than Google's quantum hardware."
                    },
                    {
                        "title": "The Limit of the Yang-Mills-Higgs Flow for twisted Higgs pairs",
                        "abstract": "In this paper, we consider the Yang-Mills-Higgs flow for twisted Higgs pairs over K\\\"ahler manifolds. We prove that this flow converges to a reflexive twisted Higgs sheaf outside a closed subset of codimension $4$, and the limiting twisted Higgs sheaf is isomorphic to the double dual of the graded twisted Higgs sheaves associated to the Harder-Narasimhan--eshadri filtration of the initial twisted Higgs bundle."
                    },
                    {
                        "title": "qecGPT: decoding Quantum Error-correcting Codes with Generative Pre-trained Transformers",
                        "abstract": "We propose a general framework for decoding quantum error-correcting codes with generative modeling. The model utilizes autoregressive neural networks, specifically Transformers, to learn the joint probability of logical operators and syndromes. This training is in an unsupervised way, without the need for labeled training data, and is thus referred to as pre-training. After the pre-training, the model can efficiently compute the likelihood of logical operators for any given syndrome, using maximum likelihood decoding. It can directly generate the most-likely logical operators with computational complexity $\\mathcal O(2k)$ in the number of logical qubits $k$, which is significantly better than the conventional maximum likelihood decoding algorithms that require $\\mathcal O(4^k)$ computation. Based on the pre-trained model, we further propose refinement to achieve more accurately the likelihood of logical operators for a given syndrome by directly sampling the stabilizer operators. We perform numerical experiments on stabilizer codes with small code distances, using both depolarizing error models and error models with correlated noise. The results show that our approach provides significantly better decoding accuracy than the minimum weight perfect matching and belief-propagation-based algorithms. Our framework is general and can be applied to any error model and quantum codes with different topologies such as surface codes and quantum LDPC codes. Furthermore, it leverages the parallelization capabilities of GPUs, enabling simultaneous decoding of a large number of syndromes. Our approach sheds light on the efficient and accurate decoding of quantum error-correcting codes using generative artificial intelligence and modern computational power."
                    },
                    {
                        "title": "Boltzmann machines as two-dimensional tensor networks",
                        "abstract": "Restricted Boltzmann machines (RBM) and deep Boltzmann machines (DBM) are important models in machine learning, and recently found numerous applications in quantum many-body physics. We show that there are fundamental connections between them and tensor networks. In particular, we demonstrate that any RBM and DBM can be exactly represented as a two-dimensional tensor network. This representation gives an understanding of the expressive power of RBM and DBM using entanglement structures of the tensor networks, also provides an efficient tensor network contraction algorithm for the computing partition function of RBM and DBM. Using numerical experiments, we demonstrate that the proposed algorithm is much more accurate than the state-of-the-art machine learning methods in estimating the partition function of restricted Boltzmann machines and deep Boltzmann machines, and have potential applications in training deep Boltzmann machines for general machine learning tasks."
                    },
                    {
                        "title": "Contracting Arbitrary Tensor Networks: General Approximate Algorithm and Applications in Graphical Models and Quantum Circuit Simulations",
                        "abstract": "We present a general method for approximately contracting tensor networks with an arbitrary connectivity. This enables us to release the computational power of tensor networks to wide use in inference and learning problems defined on general graphs. We show applications of our algorithm in graphical models, specifically on estimating free energy of spin glasses defined on various of graphs, where our method largely outperforms existing algorithms including the mean-field methods and the recently proposed neural-network-based methods. We further apply our method to the simulation of random quantum circuits, and demonstrate that, with a trade off of negligible truncation errors, our method is able to simulate large quantum circuits that are out of reach of the state-of-the-art simulation methods."
                    }
                ]
            },
            "2310c461-d6d0-4572-8146-610083d1cba2": {
                "pk": "2310c461-d6d0-4572-8146-610083d1cba2",
                "name": "Yuhang Zang",
                "collaborators": [
                    "Chen Change Loy",
                    "Chen Huang",
                    "Kaiyang Zhou",
                    "Pan Zhang",
                    "Xiaoyi Dong",
                    "Jiaqi Wang",
                    "Yuhang Cao",
                    "Dahua Lin",
                    "Wei Li",
                    "Hanlin Goh"
                ],
                "domain": [
                    "Computer Vision",
                    "Long-Tailed Learning",
                    "Multimodal Learning",
                    "Object Detection"
                ],
                "publications": [
                    {
                        "title": "FASA: Feature Augmentation and Sampling Adaptation for Long-Tailed Instance Segmentation",
                        "abstract": "Recent methods for long-tailed instance segmentation still struggle on rare object classes with few training data. We propose a simple yet effective method, Feature Augmentation and Sampling Adaptation (FASA), that addresses the data scarcity issue by augmenting the feature space especially for rare classes. Both the Feature Augmentation (FA) and feature sampling components are adaptive to the actual training status -- FA is informed by the feature mean and variance of observed real samples from past iterations, and we sample the generated virtual features in a loss-adapted manner to avoid over-fitting. FASA does not require any elaborate loss design, and removes the need for inter-class transfer learning that often involves large cost and manually-defined head/tail class groups. We show FASA is a fast, generic method that can be easily plugged into standard or long-tailed segmentation frameworks, with consistent performance gains and little added cost. FASA is also applicable to other tasks like long-tailed classification with state-of-the-art performance."
                    },
                    {
                        "title": "Overcoming the Pitfalls of Vision-Language Model Finetuning for OOD Generalization",
                        "abstract": "Existing vision-language models exhibit strong generalization on a variety of visual domains and tasks. However, such models mainly perform zero-shot recognition in a closed-set manner, and thus struggle to handle open-domain visual concepts by design. There are recent finetuning methods, such as prompt learning, that not only study the discrimination between in-distribution (ID) and out-of-distribution (OOD) samples, but also show some improvements in both ID and OOD accuracies. In this paper, we first demonstrate that vision-language models, after long enough finetuning but without proper regularization, tend to overfit the known classes in the given dataset, with degraded performance on unknown classes. Then we propose a novel approach OGEN to address this pitfall, with the main focus on improving the OOD GENeralization of finetuned models. Specifically, a class-conditional feature generator is introduced to synthesize OOD features using just the class name of any unknown class. Such synthesized features will provide useful knowledge about unknowns and help regularize the decision boundary between ID and OOD data when optimized jointly. Equally important is our adaptive self-distillation mechanism to regularize our feature generation model during joint optimization, i.e., adaptively transferring knowledge between model states to further prevent overfitting. Experiments validate that our method yields convincing gains in OOD generalization performance in different settings. Code: https://github.com/apple/ml-ogen."
                    },
                    {
                        "title": "Semi-Supervised and Long-Tailed Object Detection with CascadeMatch",
                        "abstract": "This paper focuses on long-tailed object detection in the semi-supervised learning setting, which poses realistic challenges, but has rarely been studied in the literature. We propose a novel pseudo-labeling-based detector called CascadeMatch. Our detector features a cascade network architecture, which has multi-stage detection heads with progressive confidence thresholds. To avoid manually tuning the thresholds, we design a new adaptive pseudo-label mining mechanism to automatically identify suitable values from data. To mitigate confirmation bias, where a model is negatively reinforced by incorrect pseudo-labels produced by itself, each detection head is trained by the ensemble pseudo-labels of all detection heads. Experiments on two long-tailed datasets, i.e., LVIS and COCO-LT, demonstrate that CascadeMatch surpasses existing state-of-the-art semi-supervised approaches -- across a wide range of detection architectures -- in handling long-tailed object detection. For instance, CascadeMatch outperforms Unbiased Teacher by 1.9 AP Fix on LVIS when using a ResNet50-based Cascade R-CNN structure, and by 1.7 AP Fix when using Sparse R-CNN with a Transformer encoder. We also show that CascadeMatch can even handle the challenging sparsely annotated object detection problem."
                    },
                    {
                        "title": "BroadWay: Boost Your Text-to-Video Generation Model in a Training-free Way",
                        "abstract": "The text-to-video (T2V) generation models, offering convenient visual creation, have recently garnered increasing attention. Despite their substantial potential, the generated videos may present artifacts, including structural implausibility, temporal inconsistency, and a lack of motion, often resulting in near-static video. In this work, we have identified a correlation between the disparity of temporal attention maps across different blocks and the occurrence of temporal inconsistencies. Additionally, we have observed that the energy contained within the temporal attention maps is directly related to the magnitude of motion amplitude in the generated videos. Based on these observations, we present BroadWay, a training-free method to improve the quality of text-to-video generation without introducing additional parameters, augmenting memory or sampling time. Specifically, BroadWay is composed of two principal components: 1) Temporal Self-Guidance improves the structural plausibility and temporal consistency of generated videos by reducing the disparity between the temporal attention maps across various decoder blocks. 2) Fourier-based Motion Enhancement enhances the magnitude and richness of motion by amplifying the energy of the map. Extensive experiments demonstrate that BroadWay significantly improves the quality of text-to-video generation with negligible additional cost."
                    },
                    {
                        "title": "MIA-DPO: Multi-Image Augmented Direct Preference Optimization For Large Vision-Language Models",
                        "abstract": "Visual preference alignment involves training Large Vision-Language Models (LVLMs) to predict human preferences between visual inputs. This is typically achieved by using labeled datasets of chosen/rejected pairs and employing optimization algorithms like direct preference optimization (DPO). Existing visual alignment methods, primarily designed for single-image scenarios, struggle to effectively handle the complexity of multi-image tasks due to the scarcity of diverse training data and the high cost of annotating chosen/rejected pairs. We present Multi-Image Augmented Direct Preference Optimization (MIA-DPO), a visual preference alignment approach that effectively handles multi-image inputs. MIA-DPO mitigates the scarcity of diverse multi-image training data by extending single-image data with unrelated images arranged in grid collages or pic-in-pic formats, significantly reducing the costs associated with multi-image data annotations. Our observation reveals that attention values of LVLMs vary considerably across different images. We use attention values to identify and filter out rejected responses the model may have mistakenly focused on. Our attention-aware selection for constructing the chosen/rejected pairs without relying on (i) human annotation, (ii) extra data, and (iii) external models or APIs. MIA-DPO is compatible with various architectures and outperforms existing methods on five multi-image benchmarks, achieving an average performance boost of 3.0% on LLaVA-v1.5 and 4.3% on the recent InternLM-XC2.5. Moreover, MIA-DPO has a minimal effect on the model's ability to understand single images."
                    },
                    {
                        "title": "Unified Vision and Language Prompt Learning",
                        "abstract": "Prompt tuning, a parameter- and data-efficient transfer learning paradigm that tunes only a small number of parameters in a model's input space, has become a trend in the vision community since the emergence of large vision-language models like CLIP. We present a systematic study on two representative prompt tuning methods, namely text prompt tuning and visual prompt tuning. A major finding is that none of the unimodal prompt tuning methods performs consistently well: text prompt tuning fails on data with high intra-class visual variances while visual prompt tuning cannot handle low inter-class variances. To combine the best from both worlds, we propose a simple approach called Unified Prompt Tuning (UPT), which essentially learns a tiny neural network to jointly optimize prompts across different modalities. Extensive experiments on over 11 vision datasets show that UPT achieves a better trade-off than the unimodal counterparts on few-shot learning benchmarks, as well as on domain generalization benchmarks. Code and models will be released to facilitate future research."
                    },
                    {
                        "title": "Contextual Object Detection with Multimodal Large Language Models",
                        "abstract": "Recent Multimodal Large Language Models (MLLMs) are remarkable in vision-language tasks, such as image captioning and question answering, but lack the essential perception ability, i.e., object detection. In this work, we address this limitation by introducing a novel research problem of contextual object detection -- understanding visible objects within different human-AI interactive contexts. Three representative scenarios are investigated, including the language cloze test, visual captioning, and question answering. Moreover, we present ContextDET, a unified multimodal model that is capable of end-to-end differentiable modeling of visual-language contexts, so as to locate, identify, and associate visual objects with language inputs for human-AI interaction. Our ContextDET involves three key submodels: (i) a visual encoder for extracting visual representations, (ii) a pre-trained LLM for multimodal context decoding, and (iii) a visual decoder for predicting bounding boxes given contextual object words. The new generate-then-detect framework enables us to detect object words within human vocabulary. Extensive experiments show the advantages of ContextDET on our proposed CODE benchmark, open-vocabulary detection, and referring image segmentation. Github: https://github.com/yuhangzang/ContextDET."
                    },
                    {
                        "title": "Deciphering Cross-Modal Alignment in Large Vision-Language Models with Modality Integration Rate",
                        "abstract": "We present the Modality Integration Rate (MIR), an effective, robust, and generalized metric to indicate the multi-modal pre-training quality of Large Vision Language Models (LVLMs). Large-scale pre-training plays a critical role in building capable LVLMs, while evaluating its training quality without the costly supervised fine-tuning stage is under-explored. Loss, perplexity, and in-context evaluation results are commonly used pre-training metrics for Large Language Models (LLMs), while we observed that these metrics are less indicative when aligning a well-trained LLM with a new modality. Due to the lack of proper metrics, the research of LVLMs in the critical pre-training stage is hindered greatly, including the training data choice, efficient module design, etc. In this paper, we propose evaluating the pre-training quality from the inter-modal distribution distance perspective and present MIR, the Modality Integration Rate, which is 1) \\textbf{Effective} to represent the pre-training quality and show a positive relation with the benchmark performance after supervised fine-tuning. 2) \\textbf{Robust} toward different training/evaluation data. 3) \\textbf{Generalize} across training configurations and architecture choices. We conduct a series of pre-training experiments to explore the effectiveness of MIR and observe satisfactory results that MIR is indicative about training data selection, training strategy schedule, and model architecture design to get better pre-training results. We hope MIR could be a helpful metric for building capable LVLMs and inspire the following research about modality alignment in different areas. Our code is at: https://github.com/shikiw/Modality-Integration-Rate."
                    },
                    {
                        "title": "SAM2Long: Enhancing SAM 2 for Long Video Segmentation with a Training-Free Memory Tree",
                        "abstract": "The Segment Anything Model 2 (SAM 2) has emerged as a powerful foundation model for object segmentation in both images and videos, paving the way for various downstream video applications. The crucial design of SAM 2 for video segmentation is its memory module, which prompts object-aware memories from previous frames for current frame prediction. However, its greedy-selection memory design suffers from the \"error accumulation\" problem, where an errored or missed mask will cascade and influence the segmentation of the subsequent frames, which limits the performance of SAM 2 toward complex long-term videos. To this end, we introduce SAM2Long, an improved training-free video object segmentation strategy, which considers the segmentation uncertainty within each frame and chooses the video-level optimal results from multiple segmentation pathways in a constrained tree search manner. In practice, we maintain a fixed number of segmentation pathways throughout the video. For each frame, multiple masks are proposed based on the existing pathways, creating various candidate branches. We then select the same fixed number of branches with higher cumulative scores as the new pathways for the next frame. After processing the final frame, the pathway with the highest cumulative score is chosen as the final segmentation result. Benefiting from its heuristic search design, SAM2Long is robust toward occlusions and object reappearances, and can effectively segment and track objects for complex long-term videos. Notably, SAM2Long achieves an average improvement of 3.0 points across all 24 head-to-head comparisons, with gains of up to 5.3 points in J&F on long-term video object segmentation benchmarks such as SA-V and LVOS. The code is released at https://github.com/Mark12Ding/SAM2Long."
                    },
                    {
                        "title": "Long-CLIP: Unlocking the Long-Text Capability of CLIP",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) has been the cornerstone for zero-shot classification, text-image retrieval, and text-image generation by aligning image and text modalities. Despite its widespread adoption, a significant limitation of CLIP lies in the inadequate length of text input. The length of the text token is restricted to 77, and an empirical study shows the actual effective length is even less than 20. This prevents CLIP from handling detailed descriptions, limiting its applications for image retrieval and text-to-image generation with extensive prerequisites. To this end, we propose Long-CLIP as a plug-and-play alternative to CLIP that supports long-text input, retains or even surpasses its zero-shot generalizability, and aligns the CLIP latent space, making it readily replace CLIP without any further adaptation in downstream frameworks. Nevertheless, achieving this goal is far from straightforward, as simplistic fine-tuning can result in a significant degradation of CLIP's performance. Moreover, substituting the text encoder with a language model supporting longer contexts necessitates pretraining with vast amounts of data, incurring significant expenses. Accordingly, Long-CLIP introduces an efficient fine-tuning solution on CLIP with two novel strategies designed to maintain the original capabilities, including (1) a knowledge-preserved stretching of positional embedding and (2) a primary component matching of CLIP features. With leveraging just one million extra long text-image pairs, Long-CLIP has shown the superiority to CLIP for about 20% in long caption text-image retrieval and 6% in traditional text-image retrieval tasks, e.g., COCO and Flickr30k. Furthermore, Long-CLIP offers enhanced capabilities for generating images from detailed text descriptions by replacing CLIP in a plug-and-play manner."
                    },
                    {
                        "title": "Open-Vocabulary DETR with Conditional Matching",
                        "abstract": "Open-vocabulary object detection, which is concerned with the problem of detecting novel objects guided by natural language, has gained increasing attention from the community. Ideally, we would like to extend an open-vocabulary detector such that it can produce bounding box predictions based on user inputs in form of either natural language or exemplar image. This offers great flexibility and user experience for human-computer interaction. To this end, we propose a novel open-vocabulary detector based on DETR -- hence the name OV-DETR -- which, once trained, can detect any object given its class name or an exemplar image. The biggest challenge of turning DETR into an open-vocabulary detector is that it is impossible to calculate the classification cost matrix of novel classes without access to their labeled images. To overcome this challenge, we formulate the learning objective as a binary matching one between input queries (class name or exemplar image) and the corresponding objects, which learns useful correspondence to generalize to unseen queries during testing. For training, we choose to condition the Transformer decoder on the input embeddings obtained from a pre-trained vision-language model like CLIP, in order to enable matching for both text and image queries. With extensive experiments on LVIS and COCO datasets, we demonstrate that our OV-DETR -- the first end-to-end Transformer-based open-vocabulary detector -- achieves non-trivial improvements over current state of the arts."
                    },
                    {
                        "title": "Scene Text Detection with Supervised Pyramid Context Network",
                        "abstract": "Scene text detection methods based on deep learning have achieved remarkable results over the past years. However, due to the high diversity and complexity of natural scenes, previous state-of-the-art text detection methods may still produce a considerable amount of false positives, when applied to images captured in real-world environments. To tackle this issue, mainly inspired by Mask R-CNN, we propose in this paper an effective model for scene text detection, which is based on Feature Pyramid Network (FPN) and instance segmentation. We propose a supervised pyramid context network (SPCNET) to precisely locate text regions while suppressing false positives. Benefited from the guidance of semantic information and sharing FPN, SPCNET obtains significantly enhanced performance while introducing marginal extra computation. Experiments on standard datasets demonstrate that our SPCNET clearly outperforms start-of-the-art methods. Specifically, it achieves an F-measure of 92.1% on ICDAR2013, 87.2% on ICDAR2015, 74.1% on ICDAR2017 MLT and 82.9% on Total-Text."
                    },
                    {
                        "title": "KPNet: Towards Minimal Face Detector",
                        "abstract": "The small receptive field and capacity of minimal neural networks limit their performance when using them to be the backbone of detectors. In this work, we find that the appearance feature of a generic face is discriminative enough for a tiny and shallow neural network to verify from the background. And the essential barriers behind us are 1) the vague definition of the face bounding box and 2) tricky design of anchor-boxes or receptive field. Unlike most top-down methods for joint face detection and alignment, the proposed KPNet detects small facial keypoints instead of the whole face by in a bottom-up manner. It first predicts the facial landmarks from a low-resolution image via the well-designed fine-grained scale approximation and scale adaptive soft-argmax operator. Finally, the precise face bounding boxes, no matter how we define it, can be inferred from the keypoints. Without any complex head architecture or meticulous network designing, the KPNet achieves state-of-the-art accuracy on generic face detection and alignment benchmarks with only $\\sim1M$ parameters, which runs at 1000fps on GPU and is easy to perform real-time on most modern front-end chips."
                    },
                    {
                        "title": "On-Device Domain Generalization",
                        "abstract": "We present a systematic study of domain generalization (DG) for tiny neural networks. This problem is critical to on-device machine learning applications but has been overlooked in the literature where research has been merely focused on large models. Tiny neural networks have much fewer parameters and lower complexity and therefore should not be trained the same way as their large counterparts for DG applications. By conducting extensive experiments, we find that knowledge distillation (KD), a well-known technique for model compression, is much better for tackling the on-device DG problem than conventional DG methods. Another interesting observation is that the teacher-student gap on out-of-distribution data is bigger than that on in-distribution data, which highlights the capacity mismatch issue as well as the shortcoming of KD. We further propose a method called out-of-distribution knowledge distillation (OKD) where the idea is to teach the student how the teacher handles out-of-distribution data synthesized via disruptive data augmentation. Without adding any extra parameter to the model -- hence keeping the deployment cost unchanged -- OKD significantly improves DG performance for tiny neural networks in a variety of on-device DG scenarios for image and speech applications. We also contribute a scalable approach for synthesizing visual domain shifts, along with a new suite of DG datasets to complement existing testbeds."
                    }
                ]
            },
            "b6816d52-ff06-4e3e-aaf0-2650678d1a86": {
                "pk": "b6816d52-ff06-4e3e-aaf0-2650678d1a86",
                "name": "Zehui Chen",
                "collaborators": [
                    "Feng Zhao",
                    "Zhenyu Li",
                    "Qiaofei Li",
                    "Shiquan Zhang",
                    "Liangji Fang",
                    "Qinhong Jiang",
                    "Lara Dolecek",
                    "Xianming Liu",
                    "Junjun Jiang",
                    "Qiuchen Wang"
                ],
                "domain": [
                    "3D Object Detection",
                    "Deep Learning",
                    "Computer Vision",
                    "Memory Systems"
                ],
                "publications": [
                    {
                        "title": "Write and Read Channel Models for 1S1R Crossbar Resistive Memory with High Line Resistance",
                        "abstract": "Crossbar resistive memory with 1 Selector 1 Resistor (1S1R) structure is attractive for low-cost and high-density nonvolatile memory applications. As technology scales down to the single-nm regime, the increasing resistivity of wordline/bitline becomes a limiting factor to device reliability. This paper presents write/read communication channels while considering the line resistance and device variabilities by statistically relating the degraded write/read margins and the channel parameters. Binary asymmetric channel (BAC) models are proposed for the write/read operations. Simulations based on these models suggest that the bit-error rate of devices are highly non-uniform across the memory array. These models provide quantitative tools for evaluating the trade-offs between memory reliability and design parameters, such as array size, technology nodes, and aspect ratio, and also for designing coding-theoretic solutions that would be most effective for crossbar memory."
                    },
                    {
                        "title": "Channel Models and Coding Solutions for 1S1R Crossbar Resistive Memory with High Line Resistance",
                        "abstract": "Crossbar resistive memory with the 1 Selector 1 Resistor (1S1R) structure is attractive for nonvolatile, high-density, and low-latency storage-class memory applications. As technology scales down to the single-nm regime, the increasing resistivity of wordline/bitline becomes a limiting factor to device reliability. This paper presents write/read communication channels while considering the line resistance and device variabilities by statistically relating the degraded write/read margins and the channel parameters. Binary asymmetric channel (BAC) models are proposed for the write/read operations. Simulations based on these models suggest that the bit-error rate of devices are highly non-uniform across the memory array. These models provide quantitative tools for evaluating the trade-offs between memory reliability and design parameters, such as array size, technology nodes, and aspect ratio, and also for designing coding-theoretic solutions that would be most effective for crossbar memory. Method for optimizing the read threshold is proposed to reduce the raw bit-error rate (RBER). We propose two schemes for efficient channel coding based on Bose-Chaudhuri-Hocquenghem (BCH) codes. An interleaved coding scheme is proposed to mitigate the non-uniformity of reliability and a location dependent coding framework is proposed to leverage this non-uniformity. Both of our proposed coding schemes effectively reduce the undetected bit-error rate (UBER)."
                    },
                    {
                        "title": "Towards Fine-grained Large Object Segmentation 1st Place Solution to 3D AI Challenge 2020 -- Instance Segmentation Track",
                        "abstract": "This technical report introduces our solutions of Team 'FineGrainedSeg' for Instance Segmentation track in 3D AI Challenge 2020. In order to handle extremely large objects in 3D-FUTURE, we adopt PointRend as our basic framework, which outputs more fine-grained masks compared to HTC and SOLOv2. Our final submission is an ensemble of 5 PointRend models, which achieves the 1st place on both validation and test leaderboards. The code is available at https://github.com/zehuichen123/3DFuture_ins_seg."
                    },
                    {
                        "title": "DepthFormer: Exploiting Long-Range Correlation and Local Information for Accurate Monocular Depth Estimation",
                        "abstract": "This paper aims to address the problem of supervised monocular depth estimation. We start with a meticulous pilot study to demonstrate that the long-range correlation is essential for accurate depth estimation. Therefore, we propose to leverage the Transformer to model this global context with an effective attention mechanism. We also adopt an additional convolution branch to preserve the local information as the Transformer lacks the spatial inductive bias in modeling such contents. However, independent branches lead to a shortage of connections between features. To bridge this gap, we design a hierarchical aggregation and heterogeneous interaction module to enhance the Transformer features via element-wise interaction and model the affinity between the Transformer and the CNN features in a set-to-set translation manner. Due to the unbearable memory cost caused by global attention on high-resolution feature maps, we introduce the deformable scheme to reduce the complexity. Extensive experiments on the KITTI, NYU, and SUN RGB-D datasets demonstrate that our proposed model, termed DepthFormer, surpasses state-of-the-art monocular depth estimation methods with prominent margins. Notably, it achieves the most competitive result on the highly competitive KITTI depth estimation benchmark. Our codes and models are available at https://github.com/zhyever/Monocular-Depth-Estimation-Toolbox."
                    },
                    {
                        "title": "1st Place Solutions of Waymo Open Dataset Challenge 2020 -- 2D Object Detection Track",
                        "abstract": "In this technical report, we present our solutions of Waymo Open Dataset (WOD) Challenge 2020 - 2D Object Track. We adopt FPN as our basic framework. Cascade RCNN, stacked PAFPN Neck and Double-Head are used for performance improvements. In order to handle the small object detection problem in WOD, we use very large image scales for both training and testing. Using our methods, our team RW-TSDet achieved the 1st place in the 2D Object Detection Track."
                    },
                    {
                        "title": "LiteDepth: Digging into Fast and Accurate Depth Estimation on Mobile Devices",
                        "abstract": "Monocular depth estimation is an essential task in the computer vision community. While tremendous successful methods have obtained excellent results, most of them are computationally expensive and not applicable for real-time on-device inference. In this paper, we aim to address more practical applications of monocular depth estimation, where the solution should consider not only the precision but also the inference time on mobile devices. To this end, we first develop an end-to-end learning-based model with a tiny weight size (1.4MB) and a short inference time (27FPS on Raspberry Pi 4). Then, we propose a simple yet effective data augmentation strategy, called R2 crop, to boost the model performance. Moreover, we observe that the simple lightweight model trained with only one single loss term will suffer from performance bottleneck. To alleviate this issue, we adopt multiple loss terms to provide sufficient constraints during the training stage. Furthermore, with a simple dynamic re-weight strategy, we can avoid the time-consuming hyper-parameter choice of loss terms. Finally, we adopt the structure-aware distillation to further improve the model performance. Notably, our solution named LiteDepth ranks 2nd in the MAI&AIM2022 Monocular Depth Estimation Challenge}, with a si-RMSE of 0.311, an RMSE of 3.79, and the inference time is 37$ms$ tested on the Raspberry Pi 4. Notably, we provide the fastest solution to the challenge. Codes and models will be released at \\url{https://github.com/zhyever/LiteDepth}."
                    },
                    {
                        "title": "A Vanilla Multi-Task Framework for Dense Visual Prediction Solution to 1st VCL Challenge -- Multi-Task Robustness Track",
                        "abstract": "In this report, we present our solution to the multi-task robustness track of the 1st Visual Continual Learning (VCL) Challenge at ICCV 2023 Workshop. We propose a vanilla framework named UniNet that seamlessly combines various visual perception algorithms into a multi-task model. Specifically, we choose DETR3D, Mask2Former, and BinsFormer for 3D object detection, instance segmentation, and depth estimation tasks, respectively. The final submission is a single model with InternImage-L backbone, and achieves a 49.6 overall score (29.5 Det mAP, 80.3 mTPS, 46.4 Seg mAP, and 7.93 silog) on SHIFT validation set. Besides, we provide some interesting observations in our experiments which may facilitate the development of multi-task learning in dense visual prediction."
                    },
                    {
                        "title": "Point-DETR3D: Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection",
                        "abstract": "Training high-accuracy 3D detectors necessitates massive labeled 3D annotations with 7 degree-of-freedom, which is laborious and time-consuming. Therefore, the form of point annotations is proposed to offer significant prospects for practical applications in 3D detection, which is not only more accessible and less expensive but also provides strong spatial information for object localization. In this paper, we empirically discover that it is non-trivial to merely adapt Point-DETR to its 3D form, encountering two main bottlenecks: 1) it fails to encode strong 3D prior into the model, and 2) it generates low-quality pseudo labels in distant regions due to the extreme sparsity of LiDAR points. To overcome these challenges, we introduce Point-DETR3D, a teacher-student framework for weakly semi-supervised 3D detection, designed to fully capitalize on point-wise supervision within a constrained instance-wise annotation budget.Different from Point-DETR which encodes 3D positional information solely through a point encoder, we propose an explicit positional query initialization strategy to enhance the positional prior. Considering the low quality of pseudo labels at distant regions produced by the teacher model, we enhance the detector's perception by incorporating dense imagery data through a novel Cross-Modal Deformable RoI Fusion (D-RoI).Moreover, an innovative point-guided self-supervised learning technique is proposed to allow for fully exploiting point priors, even in student models.Extensive experiments on representative nuScenes dataset demonstrate our Point-DETR3D obtains significant improvements compared to previous works. Notably, with only 5% of labeled data, Point-DETR3D achieves over 90% performance of its fully supervised counterpart."
                    },
                    {
                        "title": "MindSearch: Mimicking Human Minds Elicits Deep AI Searcher",
                        "abstract": "Information seeking and integration is a complex cognitive task that consumes enormous time and effort. Inspired by the remarkable progress of Large Language Models, recent works attempt to solve this task by combining LLMs and search engines. However, these methods still obtain unsatisfying performance due to three challenges: (1) complex requests often cannot be accurately and completely retrieved by the search engine once (2) corresponding information to be integrated is spread over multiple web pages along with massive noise, and (3) a large number of web pages with long contents may quickly exceed the maximum context length of LLMs. Inspired by the cognitive process when humans solve these problems, we introduce MindSearch to mimic the human minds in web information seeking and integration, which can be instantiated by a simple yet effective LLM-based multi-agent framework. The WebPlanner models the human mind of multi-step information seeking as a dynamic graph construction process: it decomposes the user query into atomic sub-questions as nodes in the graph and progressively extends the graph based on the search result from WebSearcher. Tasked with each sub-question, WebSearcher performs hierarchical information retrieval with search engines and collects valuable information for WebPlanner. The multi-agent design of MindSearch enables the whole framework to seek and integrate information parallelly from larger-scale (e.g., more than 300) web pages in 3 minutes, which is worth 3 hours of human effort. MindSearch demonstrates significant improvement in the response quality in terms of depth and breadth, on both close-set and open-set QA problems. Besides, responses from MindSearch based on InternLM2.5-7B are preferable by humans to ChatGPT-Web and Perplexity.ai applications, which implies that MindSearch can already deliver a competitive solution to the proprietary AI search engine."
                    },
                    {
                        "title": "Disentangle Your Dense Object Detector",
                        "abstract": "Deep learning-based dense object detectors have achieved great success in the past few years and have been applied to numerous multimedia applications such as video understanding. However, the current training pipeline for dense detectors is compromised to lots of conjunctions that may not hold. In this paper, we investigate three such important conjunctions: 1) only samples assigned as positive in classification head are used to train the regression head; 2) classification and regression share the same input feature and computational fields defined by the parallel head architecture; and 3) samples distributed in different feature pyramid layers are treated equally when computing the loss. We first carry out a series of pilot experiments to show disentangling such conjunctions can lead to persistent performance improvement. Then, based on these findings, we propose Disentangled Dense Object Detector (DDOD), in which simple and effective disentanglement mechanisms are designed and integrated into the current state-of-the-art dense object detectors. Extensive experiments on MS COCO benchmark show that our approach can lead to 2.0 mAP, 2.4 mAP and 2.2 mAP absolute improvements on RetinaNet, FCOS, and ATSS baselines with negligible extra overhead. Notably, our best model reaches 55.0 mAP on the COCO test-dev set and 93.5 AP on the hard subset of WIDER FACE, achieving new state-of-the-art performance on these two competitive benchmarks. Code is available at https://github.com/zehuichen123/DDOD."
                    },
                    {
                        "title": "AutoAlignV2: Deformable Feature Aggregation for Dynamic Multi-Modal 3D Object Detection",
                        "abstract": "Point clouds and RGB images are two general perceptional sources in autonomous driving. The former can provide accurate localization of objects, and the latter is denser and richer in semantic information. Recently, AutoAlign presents a learnable paradigm in combining these two modalities for 3D object detection. However, it suffers from high computational cost introduced by the global-wise attention. To solve the problem, we propose Cross-Domain DeformCAFA module in this work. It attends to sparse learnable sampling points for cross-modal relational modeling, which enhances the tolerance to calibration error and greatly speeds up the feature aggregation across different modalities. To overcome the complex GT-AUG under multi-modal settings, we design a simple yet effective cross-modal augmentation strategy on convex combination of image patches given their depth information. Moreover, by carrying out a novel image-level dropout training scheme, our model is able to infer in a dynamic manner. To this end, we propose AutoAlignV2, a faster and stronger multi-modal 3D detection framework, built on top of AutoAlign. Extensive experiments on nuScenes benchmark demonstrate the effectiveness and efficiency of AutoAlignV2. Notably, our best model reaches 72.4 NDS on nuScenes test leaderboard, achieving new state-of-the-art results among all published multi-modal 3D object detectors. Code will be available at https://github.com/zehuichen123/AutoAlignV2."
                    },
                    {
                        "title": "BEVDistill: Cross-Modal BEV Distillation for Multi-View 3D Object Detection",
                        "abstract": "3D object detection from multiple image views is a fundamental and challenging task for visual scene understanding. Owing to its low cost and high efficiency, multi-view 3D object detection has demonstrated promising application prospects. However, accurately detecting objects through perspective views is extremely difficult due to the lack of depth information. Current approaches tend to adopt heavy backbones for image encoders, making them inapplicable for real-world deployment. Different from the images, LiDAR points are superior in providing spatial cues, resulting in highly precise localization. In this paper, we explore the incorporation of LiDAR-based detectors for multi-view 3D object detection. Instead of directly training a depth prediction network, we unify the image and LiDAR features in the Bird-Eye-View (BEV) space and adaptively transfer knowledge across non-homogenous representations in a teacher-student paradigm. To this end, we propose \\textbf{BEVDistill}, a cross-modal BEV knowledge distillation (KD) framework for multi-view 3D object detection. Extensive experiments demonstrate that the proposed method outperforms current KD approaches on a highly-competitive baseline, BEVFormer, without introducing any extra cost in the inference phase. Notably, our best model achieves 59.4 NDS on the nuScenes test leaderboard, achieving new state-of-the-art in comparison with various image-based detectors. Code will be available at https://github.com/zehuichen123/BEVDistill."
                    },
                    {
                        "title": "Graph-DETR3D: Rethinking Overlapping Regions for Multi-View 3D Object Detection",
                        "abstract": "3D object detection from multiple image views is a fundamental and challenging task for visual scene understanding. Due to its low cost and high efficiency, multi-view 3D object detection has demonstrated promising application prospects. However, accurately detecting objects through perspective views in the 3D space is extremely difficult due to the lack of depth information. Recently, DETR3D introduces a novel 3D-2D query paradigm in aggregating multi-view images for 3D object detection and achieves state-of-the-art performance. In this paper, with intensive pilot experiments, we quantify the objects located at different regions and find that the \"truncated instances\" (i.e., at the border regions of each image) are the main bottleneck hindering the performance of DETR3D. Although it merges multiple features from two adjacent views in the overlapping regions, DETR3D still suffers from insufficient feature aggregation, thus missing the chance to fully boost the detection performance. In an effort to tackle the problem, we propose Graph-DETR3D to automatically aggregate multi-view imagery information through graph structure learning (GSL). It constructs a dynamic 3D graph between each object query and 2D feature maps to enhance the object representations, especially at the border regions. Besides, Graph-DETR3D benefits from a novel depth-invariant multi-scale training strategy, which maintains the visual depth consistency by simultaneously scaling the image size and the object depth. Extensive experiments on the nuScenes dataset demonstrate the effectiveness and efficiency of our Graph-DETR3D. Notably, our best model achieves 49.5 NDS on the nuScenes test leaderboard, achieving new state-of-the-art in comparison with various published image-view 3D object detectors."
                    }
                ]
            },
            "d9a6accc-aaaf-4665-a5d4-5b33020d585f": {
                "pk": "d9a6accc-aaaf-4665-a5d4-5b33020d585f",
                "name": "Haodong Duan",
                "collaborators": [
                    "Dahua Lin",
                    "Kai Chen",
                    "Jiaqi Wang",
                    "Yue Zhao",
                    "Yuanjun Xiong",
                    "Wentao Liu",
                    "Songyang Zhang",
                    "Xinyu Fang",
                    "Nanxuan Zhao",
                    "Bingzhe Wu"
                ],
                "domain": [
                    "Skeleton-based Action Recognition",
                    "Graph Neural Network",
                    "Self-Supervised Learning",
                    "Video Understanding"
                ],
                "publications": [
                    {
                        "title": "PYSKL: Towards Good Practices for Skeleton Action Recognition",
                        "abstract": "We present PYSKL: an open-source toolbox for skeleton-based action recognition based on PyTorch. The toolbox supports a wide variety of skeleton action recognition algorithms, including approaches based on GCN and CNN. In contrast to existing open-source skeleton action recognition projects that include only one or two algorithms, PYSKL implements six different algorithms under a unified framework with both the latest and original good practices to ease the comparison of efficacy and efficiency. We also provide an original GCN-based skeleton action recognition model named ST-GCN++, which achieves competitive recognition performance without any complicated attention schemes, serving as a strong baseline. Meanwhile, PYSKL supports the training and testing of nine skeleton-based action recognition benchmarks and achieves state-of-the-art recognition performance on eight of them. To facilitate future research on skeleton action recognition, we also provide a large number of trained models and detailed benchmark results to give some insights. PYSKL is released at https://github.com/kennymckormick/pyskl and is actively maintained. We will update this report when we add new features or benchmarks. The current version corresponds to PYSKL v0.2."
                    },
                    {
                        "title": "DG-STGCN: Dynamic Spatial-Temporal Modeling for Skeleton-based Action Recognition",
                        "abstract": "Graph convolution networks (GCN) have been widely used in skeleton-based action recognition. We note that existing GCN-based approaches primarily rely on prescribed graphical structures (ie., a manually defined topology of skeleton joints), which limits their flexibility to capture complicated correlations between joints. To move beyond this limitation, we propose a new framework for skeleton-based action recognition, namely Dynamic Group Spatio-Temporal GCN (DG-STGCN). It consists of two modules, DG-GCN and DG-TCN, respectively, for spatial and temporal modeling. In particular, DG-GCN uses learned affinity matrices to capture dynamic graphical structures instead of relying on a prescribed one, while DG-TCN performs group-wise temporal convolutions with varying receptive fields and incorporates a dynamic joint-skeleton fusion module for adaptive multi-level temporal modeling. On a wide range of benchmarks, including NTURGB+D, Kinetics-Skeleton, BABEL, and Toyota SmartHome, DG-STGCN consistently outperforms state-of-the-art methods, often by a notable margin."
                    },
                    {
                        "title": "TransRank: Self-supervised Video Representation Learning via Ranking-based Transformation Recognition",
                        "abstract": "Recognizing transformation types applied to a video clip (RecogTrans) is a long-established paradigm for self-supervised video representation learning, which achieves much inferior performance compared to instance discrimination approaches (InstDisc) in recent works. However, based on a thorough comparison of representative RecogTrans and InstDisc methods, we observe the great potential of RecogTrans on both semantic-related and temporal-related downstream tasks. Based on hard-label classification, existing RecogTrans approaches suffer from noisy supervision signals in pre-training. To mitigate this problem, we developed TransRank, a unified framework for recognizing Transformations in a Ranking formulation. TransRank provides accurate supervision signals by recognizing transformations relatively, consistently outperforming the classification-based formulation. Meanwhile, the unified framework can be instantiated with an arbitrary set of temporal or spatial transformations, demonstrating good generality. With a ranking-based formulation and several empirical practices, we achieve competitive performance on video retrieval and action recognition. Under the same setting, TransRank surpasses the previous state-of-the-art method by 6.4% on UCF101 and 8.3% on HMDB51 for action recognition (Top1 Acc); improves video retrieval on UCF101 by 20.4% (R@1). The promising results validate that RecogTrans is still a worth exploring paradigm for video self-supervised learning. Codes will be released at https://github.com/kennymckormick/TransRank."
                    },
                    {
                        "title": "SRPGAN: Perceptual Generative Adversarial Network for Single Image Super Resolution",
                        "abstract": "Single image super resolution (SISR) is to reconstruct a high resolution image from a single low resolution image. The SISR task has been a very attractive research topic over the last two decades. In recent years, convolutional neural network (CNN) based models have achieved great performance on SISR task. Despite the breakthroughs achieved by using CNN models, there are still some problems remaining unsolved, such as how to recover high frequency details of high resolution images. Previous CNN based models always use a pixel wise loss, such as l2 loss. Although the high resolution images constructed by these models have high peak signal-to-noise ratio (PSNR), they often tend to be blurry and lack high-frequency details, especially at a large scaling factor. In this paper, we build a super resolution perceptual generative adversarial network (SRPGAN) framework for SISR tasks. In the framework, we propose a robust perceptual loss based on the discriminator of the built SRPGAN model. We use the Charbonnier loss function to build the content loss and combine it with the proposed perceptual loss and the adversarial loss. Compared with other state-of-the-art methods, our method has demonstrated great ability to construct images with sharp edges and rich details. We also evaluate our method on different benchmarks and compare it with previous CNN based methods. The results show that our method can achieve much higher structural similarity index (SSIM) scores on most of the benchmarks than the previous state-of-art methods."
                    },
                    {
                        "title": "Revisiting Skeleton-based Action Recognition",
                        "abstract": "Human skeleton, as a compact representation of human action, has received increasing attention in recent years. Many skeleton-based action recognition methods adopt graph convolutional networks (GCN) to extract features on top of human skeletons. Despite the positive results shown in previous works, GCN-based methods are subject to limitations in robustness, interoperability, and scalability. In this work, we propose PoseC3D, a new approach to skeleton-based action recognition, which relies on a 3D heatmap stack instead of a graph sequence as the base representation of human skeletons. Compared to GCN-based methods, PoseC3D is more effective in learning spatiotemporal features, more robust against pose estimation noises, and generalizes better in cross-dataset settings. Also, PoseC3D can handle multiple-person scenarios without additional computation cost, and its features can be easily integrated with other modalities at early fusion stages, which provides a great design space to further boost the performance. On four challenging datasets, PoseC3D consistently obtains superior performance, when used alone on skeletons and in combination with the RGB modality."
                    },
                    {
                        "title": "Mitigating Representation Bias in Action Recognition: Algorithms and Benchmarks",
                        "abstract": "Deep learning models have achieved excellent recognition results on large-scale video benchmarks. However, they perform poorly when applied to videos with rare scenes or objects, primarily due to the bias of existing video datasets. We tackle this problem from two different angles: algorithm and dataset. From the perspective of algorithms, we propose Spatial-aware Multi-Aspect Debiasing (SMAD), which incorporates both explicit debiasing with multi-aspect adversarial training and implicit debiasing with the spatial actionness reweighting module, to learn a more generic representation invariant to non-action aspects. To neutralize the intrinsic dataset bias, we propose OmniDebias to leverage web data for joint training selectively, which can achieve higher performance with far fewer web data. To verify the effectiveness, we establish evaluation protocols and perform extensive experiments on both re-distributed splits of existing datasets and a new evaluation dataset focusing on the action with rare scenes. We also show that the debiased representation can generalize better when transferred to other datasets and tasks."
                    },
                    {
                        "title": "Omni-sourced Webly-supervised Learning for Video Recognition",
                        "abstract": "We introduce OmniSource, a novel framework for leveraging web data to train video recognition models. OmniSource overcomes the barriers between data formats, such as images, short videos, and long untrimmed videos for webly-supervised learning. First, data samples with multiple formats, curated by task-specific data collection and automatically filtered by a teacher model, are transformed into a unified form. Then a joint-training strategy is proposed to deal with the domain gaps between multiple data sources and formats in webly-supervised learning. Several good practices, including data balancing, resampling, and cross-dataset mixup are adopted in joint training. Experiments show that by utilizing data from multiple sources and formats, OmniSource is more data-efficient in training. With only 3.5M images and 800K minutes videos crawled from the internet without human labeling (less than 2% of prior works), our models learned with OmniSource improve Top-1 accuracy of 2D- and 3D-ConvNet baseline models by 3.0% and 3.9%, respectively, on the Kinetics-400 benchmark. With OmniSource, we establish new records with different pretraining strategies for video recognition. Our best models achieve 80.4%, 80.5%, and 83.6 Top-1 accuracies on the Kinetics-400 benchmark respectively for training-from-scratch, ImageNet pre-training and IG-65M pre-training."
                    },
                    {
                        "title": "Ada-LEval: Evaluating long-context LLMs with length-adaptable benchmarks",
                        "abstract": "Recently, the large language model (LLM) community has shown increasing interest in enhancing LLMs' capability to handle extremely long documents. As various long-text techniques and model architectures emerge, the precise and detailed evaluation of models' long-text capabilities has become increasingly important. Existing long-text evaluation benchmarks, such as L-Eval and LongBench, construct long-text test sets based on open-source datasets, focusing mainly on QA and summarization tasks. These datasets include test samples of varying lengths (from 2k to 32k+) entangled together, making it challenging to assess model capabilities across different length ranges. Moreover, they do not cover the ultralong settings (100k+ tokens) that the latest LLMs claim to achieve. In this paper, we introduce Ada-LEval, a length-adaptable benchmark for evaluating the long-context understanding of LLMs. Ada-LEval includes two challenging subsets, TSort and BestAnswer, which enable a more reliable evaluation of LLMs' long context capabilities. These benchmarks support intricate manipulation of the length of test cases, and can easily produce text samples up to 128k tokens. We evaluate 4 state-of-the-art closed-source API models and 6 open-source models with Ada-LEval. The evaluation results demonstrate the limitations of current LLMs, especially in ultra-long-context settings. Our code is available at https://github.com/open-compass/Ada-LEval."
                    },
                    {
                        "title": "OCSampler: Compressing Videos to One Clip with Single-step Sampling",
                        "abstract": "In this paper, we propose a framework named OCSampler to explore a compact yet effective video representation with one short clip for efficient video recognition. Recent works prefer to formulate frame sampling as a sequential decision task by selecting frames one by one according to their importance, while we present a new paradigm of learning instance-specific video condensation policies to select informative frames for representing the entire video only in a single step. Our basic motivation is that the efficient video recognition task lies in processing a whole sequence at once rather than picking up frames sequentially. Accordingly, these policies are derived from a light-weighted skim network together with a simple yet effective policy network within one step. Moreover, we extend the proposed method with a frame number budget, enabling the framework to produce correct predictions in high confidence with as few frames as possible. Experiments on four benchmarks, i.e., ActivityNet, Mini-Kinetics, FCVID, Mini-Sports1M, demonstrate the effectiveness of our OCSampler over previous methods in terms of accuracy, theoretical computational expense, actual inference speed. We also evaluate its generalization power across different classifiers, sampled frames, and search spaces. Especially, we achieve 76.9% mAP and 21.7 GFLOPs on ActivityNet with an impressive throughput: 123.9 Videos/s on a single TITAN Xp GPU."
                    },
                    {
                        "title": "Self-supervised Action Representation Learning from Partial Spatio-Temporal Skeleton Sequences",
                        "abstract": "Self-supervised learning has demonstrated remarkable capability in representation learning for skeleton-based action recognition. Existing methods mainly focus on applying global data augmentation to generate different views of the skeleton sequence for contrastive learning. However, due to the rich action clues in the skeleton sequences, existing methods may only take a global perspective to learn to discriminate different skeletons without thoroughly leveraging the local relationship between different skeleton joints and video frames, which is essential for real-world applications. In this work, we propose a Partial Spatio-Temporal Learning (PSTL) framework to exploit the local relationship from a partial skeleton sequences built by a unique spatio-temporal masking strategy. Specifically, we construct a negative-sample-free triplet steam structure that is composed of an anchor stream without any masking, a spatial masking stream with Central Spatial Masking (CSM), and a temporal masking stream with Motion Attention Temporal Masking (MATM). The feature cross-correlation matrix is measured between the anchor stream and the other two masking streams, respectively. (1) Central Spatial Masking discards selected joints from the feature calculation process, where the joints with a higher degree of centrality have a higher possibility of being selected. (2) Motion Attention Temporal Masking leverages the motion of action and remove frames that move faster with a higher possibility. Our method achieves SOTA performance on NTU-60, NTU-120 and PKU-MMD under various downstream tasks. A practical evaluation is performed where some skeleton joints are lost in downstream tasks. In contrast to previous methods that suffer from large performance drops, our PSTL can still achieve remarkable results, validating the robustness of our method. Code: https://github.com/YujieOuO/PSTL.git."
                    },
                    {
                        "title": "TRB: A Novel Triplet Representation for Understanding 2D Human Body",
                        "abstract": "Human pose and shape are two important components of 2D human body. However, how to efficiently represent both of them in images is still an open question. In this paper, we propose the Triplet Representation for Body (TRB) -- a compact 2D human body representation, with skeleton keypoints capturing human pose information and contour keypoints containing human shape information. TRB not only preserves the flexibility of skeleton keypoint representation, but also contains rich pose and human shape information. Therefore, it promises broader application areas, such as human shape editing and conditional image generation. We further introduce the challenging problem of TRB estimation, where joint learning of human pose and shape is required. We construct several large-scale TRB estimation datasets, based on popular 2D pose datasets: LSP, MPII, COCO. To effectively solve TRB estimation, we propose a two-branch network (TRB-net) with three novel techniques, namely X-structure (Xs), Directional Convolution (DC) and Pairwise Mapping (PM), to enforce multi-level message passing for joint feature learning. We evaluate our proposed TRB-net and several leading approaches on our proposed TRB datasets, and demonstrate the superiority of our method through extensive evaluations."
                    },
                    {
                        "title": "SkeleTR: Towrads Skeleton-based Action Recognition in the Wild",
                        "abstract": "We present SkeleTR, a new framework for skeleton-based action recognition. In contrast to prior work, which focuses mainly on controlled environments, we target more general scenarios that typically involve a variable number of people and various forms of interaction between people. SkeleTR works with a two-stage paradigm. It first models the intra-person skeleton dynamics for each skeleton sequence with graph convolutions, and then uses stacked Transformer encoders to capture person interactions that are important for action recognition in general scenarios. To mitigate the negative impact of inaccurate skeleton associations, SkeleTR takes relative short skeleton sequences as input and increases the number of sequences. As a unified solution, SkeleTR can be directly applied to multiple skeleton-based action tasks, including video-level action classification, instance-level action detection, and group-level activity recognition. It also enables transfer learning and joint training across different action tasks and datasets, which result in performance improvement. When evaluated on various skeleton-based action recognition benchmarks, SkeleTR achieves the state-of-the-art performance."
                    },
                    {
                        "title": "ProSA: Assessing and Understanding the Prompt Sensitivity of LLMs",
                        "abstract": "Large language models (LLMs) have demonstrated impressive capabilities across various tasks, but their performance is highly sensitive to the prompts utilized. This variability poses challenges for accurate assessment and user satisfaction. Current research frequently overlooks instance-level prompt variations and their implications on subjective evaluations. To address these shortcomings, we introduce ProSA, a framework designed to evaluate and comprehend prompt sensitivity in LLMs. ProSA incorporates a novel sensitivity metric, PromptSensiScore, and leverages decoding confidence to elucidate underlying mechanisms. Our extensive study, spanning multiple tasks, uncovers that prompt sensitivity fluctuates across datasets and models, with larger models exhibiting enhanced robustness. We observe that few-shot examples can alleviate this sensitivity issue, and subjective evaluations are also susceptible to prompt sensitivities, particularly in complex, reasoning-oriented tasks. Furthermore, our findings indicate that higher model confidence correlates with increased prompt robustness. We believe this work will serve as a helpful tool in studying prompt sensitivity of LLMs. The project is released at: https://github.com/open-compass/ProSA ."
                    },
                    {
                        "title": "MMBench-Video: A Long-Form Multi-Shot Benchmark for Holistic Video Understanding",
                        "abstract": "The advent of large vision-language models (LVLMs) has spurred research into their applications in multi-modal contexts, particularly in video understanding. Traditional VideoQA benchmarks, despite providing quantitative metrics, often fail to encompass the full spectrum of video content and inadequately assess models' temporal comprehension. To address these limitations, we introduce MMBench-Video, a quantitative benchmark designed to rigorously evaluate LVLMs' proficiency in video understanding. MMBench-Video incorporates lengthy videos from YouTube and employs free-form questions, mirroring practical use cases. The benchmark is meticulously crafted to probe the models' temporal reasoning skills, with all questions human-annotated according to a carefully constructed ability taxonomy. We employ GPT-4 for automated assessment, demonstrating superior accuracy and robustness over earlier LLM-based evaluations. Utilizing MMBench-Video, we have conducted comprehensive evaluations that include both proprietary and open-source LVLMs for images and videos. MMBench-Video stands as a valuable resource for the research community, facilitating improved evaluation of LVLMs and catalyzing progress in the field of video understanding. The evalutation code of MMBench-Video will be integrated into VLMEvalKit: https://github.com/open-compass/VLMEvalKit."
                    }
                ]
            },
            "85ec0db9-d177-40f7-9010-0366305b541e": {
                "pk": "85ec0db9-d177-40f7-9010-0366305b541e",
                "name": "Bin Lin",
                "collaborators": [
                    "Gregorio Robles",
                    "Dejun Qian",
                    "Yang Ye",
                    "Bin Zhu",
                    "Jiaxi Cui",
                    "Munan Ning",
                    "Peng Jin",
                    "Li Yuan",
                    "H. Gordon Berry",
                    "Tomohiro Shibata"
                ],
                "domain": [
                    "Software Engineering",
                    "Natural Language Processing",
                    "Multi-modal Learning",
                    "Symbolic Execution"
                ],
                "publications": [
                    {
                        "title": "Investigating the Impact of Vocabulary Difficulty and Code Naturalness on Program Comprehension",
                        "abstract": "Context: Developers spend most of their time comprehending source code during software development. Automatically assessing how readable and understandable source code is can provide various benefits in different tasks, such as task triaging and code reviews. While several studies have proposed approaches to predict software readability and understandability, most of them only focus on local characteristics of source code. Besides, the performance of understandability prediction is far from satisfactory.   Objective: In this study, we aim to assess readability and understandability from the perspective of language acquisition. More specifically, we would like to investigate whether code readability and understandability are correlated with the naturalness and vocabulary difficulty of source code.   Method: To assess code naturalness, we adopted the cross-entropy metric, while we use a manually crafted list of code elements with their assigned advancement levels to assess the vocabulary difficulty. We will conduct a statistical analysis to understand their correlations and analyze whether code naturalness and vocabulary difficulty can be used to improve the performance of code readability and understandability prediction methods. The study will be conducted on existing datasets."
                    },
                    {
                        "title": "Regression Testing of Virtual Prototypes Using Symbolic Execution",
                        "abstract": "Recently virtual platforms and virtual prototyping techniques have been widely applied for accelerating software development in electronics companies. It has been proved that these techniques can greatly shorten time-to-market and improve product quality. One challenge is how to test and validate a virtual prototype. In this paper, we present how to conduct regression testing of virtual prototypes in different versions using symbolic execution. Suppose we have old and new versions of a virtual prototype, we first apply symbolic execution to the new version and collect all path constraints. Then the collected path constraints are used for guiding the symbolic execution of the old version. For each path explored, we compare the device states between two versions to check if they behave the same. We have applied this approach to a widely-used virtual prototype and detected numerous differences. The experimental results show that our approach is useful and efficient."
                    },
                    {
                        "title": "Video-LLaVA: Learning United Visual Representation by Alignment Before Projection",
                        "abstract": "The Large Vision-Language Model (LVLM) has enhanced the performance of various downstream tasks in visual-language understanding. Most existing approaches encode images and videos into separate feature spaces, which are then fed as inputs to large language models. However, due to the lack of unified tokenization for images and videos, namely misalignment before projection, it becomes challenging for a Large Language Model (LLM) to learn multi-modal interactions from several poor projection layers. In this work, we unify visual representation into the language feature space to advance the foundational LLM towards a unified LVLM. As a result, we establish a simple but robust LVLM baseline, Video-LLaVA, which learns from a mixed dataset of images and videos, mutually enhancing each other. Video-LLaVA achieves superior performances on a broad range of 9 image benchmarks across 5 image question-answering datasets and 4 image benchmark toolkits. Additionally, our Video-LLaVA also outperforms Video-ChatGPT by 5.8%, 9.9%, 18.6%, and 10.1% on MSRVTT, MSVD, TGIF, and ActivityNet, respectively. Notably, extensive experiments demonstrate that Video-LLaVA mutually benefits images and videos within a unified visual representation, outperforming models designed specifically for images or videos. We aim for this work to provide modest insights into the multi-modal inputs for the LLM. Code address: \\href{https://github.com/PKU-YuanGroup/Video-LLaVA}"
                    },
                    {
                        "title": "Beam-gas Spectroscopy of Sextet Transitions in O3+, F4+ and Ne5+",
                        "abstract": "We present VUV observations of transitions between doubly excited sextet states in O3+, F4+ and Ne5+. Spectra were produced by collisions of an oxygen, uorine and neon beam with a nitrogen gas jet target. Prepared beam-gas experiment yields new and explicit information on doubly core-excited ions. Some observed lines were assigned to the 1s2s2p3 6S-1s2p33s, 3d 6P electric-dipole transitions in O3+, F4+ and Ne5+. Three lines have been reassigned. Present data are the first explicit measurements on transitions between sextet states in boronlike ions by beam-gas spectroscopy."
                    }
                ]
            },
            "2183be5a-8360-4a11-a46b-80ca7376c868": {
                "pk": "2183be5a-8360-4a11-a46b-80ca7376c868",
                "name": "Zhenyu Tang",
                "collaborators": [
                    "Dinesh Manocha",
                    "Anton Ratnarajah",
                    "Hsien-Yu Meng",
                    "Rohith Aralikatti",
                    "Hui-Ling Zhen",
                    "Nicolas Morales",
                    "Shaoting Zhang",
                    "Xiaosong Wang",
                    "John D. Kanu",
                    "Kevin Hogan"
                ],
                "domain": [
                    "Acoustic Modeling",
                    "Speech Recognition",
                    "Deep Learning",
                    "Geometric Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Scene-aware Far-field Automatic Speech Recognition",
                        "abstract": "We propose a novel method for generating scene-aware training data for far-field automatic speech recognition. We use a deep learning-based estimator to non-intrusively compute the sub-band reverberation time of an environment from its speech samples. We model the acoustic characteristics of a scene with its reverberation time and represent it using a multivariate Gaussian distribution. We use this distribution to select acoustic impulse responses from a large real-world dataset for augmenting speech data. The speech recognition system trained on our scene-aware data consistently outperforms the system trained using many more random acoustic impulse responses on the REVERB and the AMI far-field benchmarks. In practice, we obtain 2.64% absolute improvement in word error rate compared with using training data of the same size with uniformly distributed reverberation times."
                    },
                    {
                        "title": "Anti-Chaos Control via Nonlinear Schr\u00f6dinger Equations for the secured optical communication",
                        "abstract": "Coupled nonlinear Schr\\\"odinger equations, governing the propagation of envelopes of electromagnetic waves in birefringent optical fibers, are studied in this paper for their potential applications in the secured optical communication. Periodicity and integrability of the CNLS equations are obtained via the phase-plane analysis. With the time-delay and perturbations introduced, CNLS equations are chaotified and a chaotic system is proposed. Numerical and analytical methods are conducted on such system: (I) Phase projections are given and the final chaotic states can be observed. (II) Power spectra and the largest Lyapunov exponents are calculated to corroborate that those motions are indeed chaotic."
                    },
                    {
                        "title": "Receiver Placement for Speech Enhancement using Sound Propagation Optimization",
                        "abstract": "A common problem in acoustic design is the placement of speakers or receivers for public address systems, telecommunications, and home smart speakers or digital personal assistants. We present a novel algorithm to automatically place a speaker or receiver in a room to improve the intelligibility of spoken phrases in a design. Our technique uses a sound propagation optimization formulation to maximize the Speech Transmission Index (STI) by computing an optimal location of the sound receiver. We use an efficient and accurate hybrid sound propagation technique on complex 3D models to compute the Room Impulse Responses (RIR) and evaluate their impact on the STI. The overall algorithm computes a globally optimal position of the receiver that reduces the effects of reverberation and noise over many source positions. We evaluate our algorithm on various indoor 3D models, all showing significant improvement in STI, based on accurate sound propagation."
                    },
                    {
                        "title": "Chaotification via Higher-order Nonlinear Schr\u00f6dinger Equations for Secured Communication",
                        "abstract": "Higher-order nonlinear Schr\\\"odinger(HNLS) equation which can be used to describe the propagation of short light pulses in the optical fibers, is studied in this paper. Using the phase plane analysis, HNLS equation is reduced into the equivalent dynamical system, the periodicity of such system is obtained with the phase projections and power spectra given. By means of the time-delay feedback method, with the original dynamical system rewritten, we construct a single-input single-output system, and propose a chaotic system based on the chaotification of HNLS. Numerical studies have been conducted on such system. Chaotic motions with different time delays are displayed. Power spectra of such chaotic motions are calculated. Lyapunov exponents are given to corroborate that those motions are indeed chaotic."
                    },
                    {
                        "title": "Low-frequency Compensated Synthetic Impulse Responses for Improved Far-field Speech Recognition",
                        "abstract": "We propose a method for generating low-frequency compensated synthetic impulse responses that improve the performance of far-field speech recognition systems trained on artificially augmented datasets. We design linear-phase filters that adapt the simulated impulse responses to equalization distributions corresponding to real-world captured impulse responses. Our filtered synthetic impulse responses are then used to augment clean speech data from LibriSpeech dataset [1]. We evaluate the performance of our method on the real-world LibriSpeech test set. In practice, our low-frequency compensated synthetic dataset can reduce the word-error-rate by up to 8.8% for far-field speech recognition."
                    },
                    {
                        "title": "Learning Acoustic Scattering Fields for Dynamic Interactive Sound Propagation",
                        "abstract": "We present a novel hybrid sound propagation algorithm for interactive applications. Our approach is designed for dynamic scenes and uses a neural network-based learned scattered field representation along with ray tracing to generate specular, diffuse, diffraction, and occlusion effects efficiently. We use geometric deep learning to approximate the acoustic scattering field using spherical harmonics. We use a large 3D dataset for training, and compare its accuracy with the ground truth generated using an accurate wave-based solver. The additional overhead of computing the learned scattered field at runtime is small and we demonstrate its interactive performance by generating plausible sound effects in dynamic scenes with diffraction and occlusion effects. We demonstrate the perceptual benefits of our approach based on an audio-visual user study."
                    },
                    {
                        "title": "IR-GAN: Room Impulse Response Generator for Far-field Speech Recognition",
                        "abstract": "We present a Generative Adversarial Network (GAN) based room impulse response generator (IR-GAN) for generating realistic synthetic room impulse responses (RIRs). IR-GAN extracts acoustic parameters from captured real-world RIRs and uses these parameters to generate new synthetic RIRs. We use these generated synthetic RIRs to improve far-field automatic speech recognition in new environments that are different from the ones used in training datasets. In particular, we augment the far-field speech training set by convolving our synthesized RIRs with a clean LibriSpeech dataset. We evaluate the quality of our synthetic RIRs on the real-world LibriSpeech test set created using real-world RIRs from the BUT ReverbDB and AIR datasets. Our IR-GAN reports up to an 8.95% lower error rate than Geometric Acoustic Simulator (GAS) in far-field speech recognition benchmarks. We further improve the performance when we combine our synthetic RIRs with synthetic impulse responses generated using GAS. This combination can reduce the word error rate by up to 14.3% in far-field speech recognition benchmarks."
                    },
                    {
                        "title": "GWA: A Large High-Quality Acoustic Dataset for Audio Processing",
                        "abstract": "We present the Geometric-Wave Acoustic (GWA) dataset, a large-scale audio dataset of about 2 million synthetic room impulse responses (IRs) and their corresponding detailed geometric and simulation configurations. Our dataset samples acoustic environments from over 6.8K high-quality diverse and professionally designed houses represented as semantically labeled 3D meshes. We also present a novel real-world acoustic materials assignment scheme based on semantic matching that uses a sentence transformer model. We compute high-quality impulse responses corresponding to accurate low-frequency and high-frequency wave effects by automatically calibrating geometric acoustic ray-tracing with a finite-difference time-domain wave solver. We demonstrate the higher accuracy of our IRs by comparing with recorded IRs from complex real-world environments. Moreover, we highlight the benefits of GWA on audio deep learning tasks such as automated speech recognition, speech enhancement, and speech separation. This dataset is the first data with accurate wave acoustic simulations in complex scenes. Codes and data are available at https://gamma.umd.edu/pro/sound/gwa."
                    },
                    {
                        "title": "Exploring Data Redundancy in Real-world Image Classification through Data Selection",
                        "abstract": "Deep learning models often require large amounts of data for training, leading to increased costs. It is particularly challenging in medical imaging, i.e., gathering distributed data for centralized training, and meanwhile, obtaining quality labels remains a tedious job. Many methods have been proposed to address this issue in various training paradigms, e.g., continual learning, active learning, and federated learning, which indeed demonstrate certain forms of the data valuation process. However, existing methods are either overly intuitive or limited to common clean/toy datasets in the experiments. In this work, we present two data valuation metrics based on Synaptic Intelligence and gradient norms, respectively, to study the redundancy in real-world image data. Novel online and offline data selection algorithms are then proposed via clustering and grouping based on the examined data values. Our online approach effectively evaluates data utilizing layerwise model parameter updates and gradients in each epoch and can accelerate model training with fewer epochs and a subset (e.g., 19%-59%) of data while maintaining equivalent levels of accuracy in a variety of datasets. It also extends to the offline coreset construction, producing subsets of only 18%-30% of the original. The codes for the proposed adaptive data selection and coreset computation are available (https://github.com/ZhenyuTANG2023/data_selection)."
                    },
                    {
                        "title": "Point-based Acoustic Scattering for Interactive Sound Propagation via Surface Encoding",
                        "abstract": "We present a novel geometric deep learning method to compute the acoustic scattering properties of geometric objects. Our learning algorithm uses a point cloud representation of objects to compute the scattering properties and integrates them with ray tracing for interactive sound propagation in dynamic scenes. We use discrete Laplacian-based surface encoders and approximate the neighborhood of each point using a shared multi-layer perceptron. We show that our formulation is permutation invariant and present a neural network that computes the scattering function using spherical harmonics. Our approach can handle objects with arbitrary topologies and deforming models, and takes less than 1ms per object on a commodity GPU. We have analyzed the accuracy and perform validation on thousands of unseen 3D objects and highlight the benefits over other point-based geometric deep learning methods. To the best of our knowledge, this is the first real-time learning algorithm that can approximate the acoustic scattering properties of arbitrary objects with high accuracy."
                    },
                    {
                        "title": "Improving Reverberant Speech Separation with Multi-stage Training and Curriculum Learning",
                        "abstract": "We present a novel approach that improves the performance of reverberant speech separation. Our approach is based on an accurate geometric acoustic simulator (GAS) which generates realistic room impulse responses (RIRs) by modeling both specular and diffuse reflections. We also propose three training methods - pre-training, multi-stage training and curriculum learning that significantly improve separation quality in the presence of reverberation. We also demonstrate that mixing the synthetic RIRs with a small number of real RIRs during training enhances separation performance. We evaluate our approach on reverberant mixtures generated from real, recorded data (in several different room configurations) from the VOiCES dataset. Our novel approach (curriculum learning+pre-training+multi-stage training) results in a significant relative improvement over prior techniques based on image source method (ISM)."
                    },
                    {
                        "title": "TS-RIR: Translated synthetic room impulse responses for speech augmentation",
                        "abstract": "We present a method for improving the quality of synthetic room impulse responses for far-field speech recognition. We bridge the gap between the fidelity of synthetic room impulse responses (RIRs) and the real room impulse responses using our novel, TS-RIRGAN architecture. Given a synthetic RIR in the form of raw audio, we use TS-RIRGAN to translate it into a real RIR. We also perform real-world sub-band room equalization on the translated synthetic RIR. Our overall approach improves the quality of synthetic RIRs by compensating low-frequency wave effects, similar to those in real RIRs. We evaluate the performance of improved synthetic RIRs on a far-field speech dataset augmented by convolving the LibriSpeech clean speech dataset [1] with RIRs and adding background noise. We show that far-field speech augmented using our improved synthetic RIRs reduces the word error rate by up to 19.9% in Kaldi far-field automatic speech recognition benchmark [2]."
                    },
                    {
                        "title": "Synthetic Wave-Geometric Impulse Responses for Improved Speech Dereverberation",
                        "abstract": "We present a novel approach to improve the performance of learning-based speech dereverberation using accurate synthetic datasets. Our approach is designed to recover the reverb-free signal from a reverberant speech signal. We show that accurately simulating the low-frequency components of Room Impulse Responses (RIRs) is important to achieving good dereverberation. We use the GWA dataset that consists of synthetic RIRs generated in a hybrid fashion: an accurate wave-based solver is used to simulate the lower frequencies and geometric ray tracing methods simulate the higher frequencies. We demonstrate that speech dereverberation models trained on hybrid synthetic RIRs outperform models trained on RIRs generated by prior geometric ray tracing methods on four real-world RIR datasets."
                    },
                    {
                        "title": "Regression and Classification for Direction-of-Arrival Estimation with Convolutional Recurrent Neural Networks",
                        "abstract": "We present a novel learning-based approach to estimate the direction-of-arrival (DOA) of a sound source using a convolutional recurrent neural network (CRNN) trained via regression on synthetic data and Cartesian labels. We also describe an improved method to generate synthetic data to train the neural network using state-of-the-art sound propagation algorithms that model specular as well as diffuse reflections of sound. We compare our model against three other CRNNs trained using different formulations of the same problem: classification on categorical labels, and regression on spherical coordinate labels. In practice, our model achieves up to 43% decrease in angular error over prior methods. The use of diffuse reflection results in 34% and 41% reduction in angular prediction errors on LOCATA and SOFA datasets, respectively, over prior methods based on image-source methods. Our method results in an additional 3% error reduction over prior schemes that use classification based networks, and we use 36% fewer network parameters."
                    },
                    {
                        "title": "Improving Reverberant Speech Training Using Diffuse Acoustic Simulation",
                        "abstract": "We present an efficient and realistic geometric acoustic simulation approach for generating and augmenting training data in speech-related machine learning tasks. Our physically-based acoustic simulation method is capable of modeling occlusion, specular and diffuse reflections of sound in complicated acoustic environments, whereas the classical image method can only model specular reflections in simple room settings. We show that by using our synthetic training data, the same neural networks gain significant performance improvement on real test sets in far-field speech recognition by 1.58% and keyword spotting by 21%, without fine-tuning using real impulse responses."
                    }
                ]
            },
            "768f317a-c02a-4cb6-b9ac-506a4ddb826b": {
                "pk": "768f317a-c02a-4cb6-b9ac-506a4ddb826b",
                "name": "Li Yuan",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "9a0a8c52-fb4a-431a-b20a-2b4ce13c53fa": {
                "pk": "9a0a8c52-fb4a-431a-b20a-2b4ce13c53fa",
                "name": "Yu Qiao",
                "collaborators": [
                    "Jia Zhao",
                    "Zhaoru Shang",
                    "Catarina Carvalho",
                    "Daniel Wiechmann",
                    "Elma Kerz",
                    "Honglei Lang",
                    "Yanbin Yin",
                    "Benzhuo Lu",
                    "Meng Wang",
                    "Xuefeng Yin"
                ],
                "domain": [
                    "Mathematical Physics",
                    "Thermodynamics",
                    "Machine Learning",
                    "Algebraic Structures"
                ],
                "publications": [
                    {
                        "title": "Analytical Solution of the Degeneracy Pressure of Two-Dimensional Quantum Gas",
                        "abstract": "The exact equations of state of two-dimensional Bose gas (Equation 14) and Fermi gas (Equation 24) are derived, to calculate the degeneracy pressure. They are consistent with the second law of thermodynamics in the entire phase space. Under relevant conditions, they converge to the conventional equations."
                    },
                    {
                        "title": "Fredholm Groupoids and Layer Potentials on Conical Domains",
                        "abstract": "We show that layer potential groupoids for conical domains constructed in an earlier paper (Carvalho-Qiao, Central European J. Math., 2013) are Fredholm groupoids, which enables us to deal with many analysis problems on singular spaces in a unified treatment. As an application, we obtain Fredholm criteria for operators on layer potential groupoids."
                    },
                    {
                        "title": "Cohomology and relative Rota-Baxter-Nijenhuis structures on LieYRep pairs",
                        "abstract": "A LieYRep pair consists of a Lie-Yamaguti algebra and its representation. In this paper, we establish the cohomology theory of LieYRep pairs and characterize their linear deformations by the second cohomology group. Then we introduce the notion of relative Rota-Baxter-Nijenhuis structures on LieYRep pairs, investigate their properties, and prove that a relative Rota-Baxter-Nijenhuis structure gives rise to a pair of compatible relative Rota-Baxter operators under a certain condition. Finally, we show the equivalence between $r$-matrix-Nijenhuis structures and Rota-Baxter-Nijenhuis structures on Lie-Yamaguti algebras."
                    },
                    {
                        "title": "On Lie Bialgebroid Crossed Modules",
                        "abstract": "We study Lie bialgebroid crossed modules which are pairs of Lie algebroid crossed modules in duality that canonically give rise to Lie bialgebroids. A one-one correspondence between such Lie bialgebroid crossed modules and co-quadratic Manin triples $(K,P,Q)$ is established, where $K$ is a co-quadratic Lie algebroid and $(P,Q)$ is a pair of transverse Dirac structures in $K$."
                    },
                    {
                        "title": "Improvements in continuum modeling for biomolecular systems",
                        "abstract": "Modeling of biomolecular systems plays an essential role in understanding biological processes, such as ionic flow across channels, protein modification or interaction, and cell signaling. The continuum model described by the Poisson-Boltzmann (PB)/Poisson-Nernst-Planck (PNP) equations has made great contributions towards simulation of these processes. However, the model has shortcomings in its commonly used form and cannot capture (or cannot accurately capture) some important physical properties of biological systems. Considerable efforts have been made to improve the continuum model to account for discrete particle interactions and to make progress in numerical methods to provide accurate and efficient simulation. This review will summarize recent main improvements in continuum modeling for biomolecular systems, with focus on the size-modified models, the coupling of the classical density functional theory and PNP equations, the coupling of polar and nonpolar interactions, and numerical progress."
                    },
                    {
                        "title": "Producing Useful Work in a Cycle by Absorbing Heat from a Single Thermal Reservoir: An Investigation on a Locally Nonchaotic Energy Barrier",
                        "abstract": "In the current research, we investigate the concept of spontaneously nonequilibrium dimension (SND), and show that a SND-based system can break the second law of thermodynamics. The main characteristic of the SND is the inherent nonequilibrium particle crossing ratio. A locally nonchaotic energy barrier is employed to form the model system. On the one hand, when the barrier width is much smaller than the mean free path of the particles, the system cannot reach thermodynamic equilibrium; on the other hand, the nonequilibrium particle distribution allows for production of useful work in a cycle by absorbing heat from a single thermal reservoir. Such system performance is demonstrated by a Monte Carlo simulation. It should be attributed to the unbalanced cross-influence of the thermally correlated thermodynamic forces, incompatible with the conventional framework of statistical mechanics. No Maxwell's demon is involved. Similar effects may be achieved by a number of variants, e.g., when the barrier is switchable or there are distributed nonchaotic traps."
                    },
                    {
                        "title": "Cohomology and deformations of Relative Rota-Baxter operators on Lie-Yamaguti algebras",
                        "abstract": "In this paper, we establish the cohomology of relative Rota-Baxter operators on Lie-Yamaguti algebras via the Yamaguti cohomology. Then we use this type of cohomology to characterize deformations of relative Rota-Baxter operators on Lie-Yamaguti algebras. We show that if two linear or formal deformations of a relative Rota-Baxter operator are equivalent, then their infinitesimals are in the same cohomology class in the first cohomology group. Moreover, an order $n$ deformation of a relative Rota-Baxter operator can be extended to an order $n+1$ deformation if and only if the obstruction class in the second cohomology group is trivial."
                    },
                    {
                        "title": "The classical Lie-Yamaguti Yang-Baxter equation and Lie-Yamaguti bialgebras",
                        "abstract": "In this paper, we develop the bialgebra theory for Lie-Yamaguti algebras. For this purpose, we exploit two types of compatibility conditions: local cocycle condition and double construction. We define the classical Yang-Baxter equation in Lie-Yamaguti algebras and show that a solution to the classical Yang-Baxter equation corresponds to a relative Rota-Baxter operator with respect to the coadjoint representation. Furthermore, we generalize some results by Bai in [1] and Semonov-Tian-Shansky in [19] to the context of Lie-Yamaguti algebras. Then we introduce the notion of matched pairs of Lie-Yamaguti algebras, which leads us to the concept of double construction Lie-Yamaguti bialgebras following the Manin triple approach to Lie bialgebras. We prove that matched pairs, Manin triples of Lie-Yamaguti algebras, and double construction Lie-Yamaguti bialgebras are equivalent. Finally, we clarify that a local cocycle condition is a special case of a double construction for Lie-Yamaguti bialgebras."
                    },
                    {
                        "title": "On the second law of thermodynamics: Spontaneous cold-to-hot heat transfer in a nonchaotic medium",
                        "abstract": "It has long been known that, fundamentally different from a large body of rarefied gas, when a Knudsen gas is immersed in a thermal bath, it may never reach thermal equilibrium. The root cause is nonchaoticity: as the particle-particle collisions are sparse, the particle trajectories tend to be independent of each other. Usually, this counterintuitive phenomenon is studied through kinetic theory and is not considered a thermodynamic problem. In current research, we show that if incorporated in a compound setup, such an intrinsically nonequilibrium behavior has nontrivial consequences and cannot circumvent thermodynamics: cold-to-hot heat transfer may happen spontaneously, either continuously (with an energy barrier) or cyclically (with time-dependent entropy barriers). It allows for production of useful work by absorbing heat from a single thermal reservoir without any other effect. As the system obeys the first law of thermodynamics, the refrigeration statement and the heat-engine statement of the second law of thermodynamics cannot be applied. The basic principle of maximum entropy is adhered to."
                    },
                    {
                        "title": "Experiment on intrinsically nonequilibrium distribution of large ions in charged small nanopores",
                        "abstract": "Recent theoretical research on locally nonchaotic gravitational energy barrier led to an interesting finding: beyond the boundaries of Boltzmann's H-theorem, there may be macroscopic systems with nontrivial energy properties. The fundamental mechanism is rooted in the intrinsically nonequilibrium steady state. In the current investigation, we experimentally verify this concept, with the weak gravitational force being changed to the strong Coulomb force. The tests are performed on capacitive cells with the same nanoporous electrodes and various electrolyte concentrations. The key characteristic is that the nanopore size is only slightly larger than the ion size, less than twice the ion size. The confinement effect of the nanopore walls plays the role of the spontaneously nonequilibrium dimension (SND). At first glance, the capacitive cells exhibit \"normal\" charge curves. However, their steady-state ion distribution significantly differs from thermodynamic equilibrium. The measured potential difference is nearly an order of magnitude larger than the prediction of equilibrium thermodynamics. Such phenomena are in line with the molecular dynamics simulations reported in the open literature."
                    },
                    {
                        "title": "Alzheimer's Disease Detection from Spontaneous Speech through Combining Linguistic Complexity and (Dis)Fluency Features with Pretrained Language Models",
                        "abstract": "In this paper, we combined linguistic complexity and (dis)fluency features with pretrained language models for the task of Alzheimer's disease detection of the 2021 ADReSSo (Alzheimer's Dementia Recognition through Spontaneous Speech) challenge. An accuracy of 83.1% was achieved on the test set, which amounts to an improvement of 4.23% over the baseline model. Our best-performing model that integrated component models using a stacking ensemble technique performed equally well on cross-validation and test data, indicating that it is robust against overfitting."
                    },
                    {
                        "title": "MANTIS at #SMM4H 2023: Leveraging Hybrid and Ensemble Models for Detection of Social Anxiety Disorder on Reddit",
                        "abstract": "This paper presents our system employed for the Social Media Mining for Health 2023 Shared Task 4: Binary classification of English Reddit posts self-reporting a social anxiety disorder diagnosis. We systematically investigate and contrast the efficacy of hybrid and ensemble models that harness specialized medical domain-adapted transformers in conjunction with BiLSTM neural networks. The evaluation results outline that our best performing model obtained 89.31% F1 on the validation set and 83.76% F1 on the test set."
                    },
                    {
                        "title": "Knowledge-based Fully Convolutional Network and Its Application in Segmentation of Lung CT Images",
                        "abstract": "A variety of deep neural networks have been applied in medical image segmentation and achieve good performance. Unlike natural images, medical images of the same imaging modality are characterized by the same pattern, which indicates that same normal organs or tissues locate at similar positions in the images. Thus, in this paper we try to incorporate the prior knowledge of medical images into the structure of neural networks such that the prior knowledge can be utilized for accurate segmentation. Based on this idea, we propose a novel deep network called knowledge-based fully convolutional network (KFCN) for medical image segmentation. The segmentation function and corresponding error is analyzed. We show the existence of an asymptotically stable region for KFCN which traditional FCN doesn't possess. Experiments validate our knowledge assumption about the incorporation of prior knowledge into the convolution kernels of KFCN and show that KFCN can achieve a reasonable segmentation and a satisfactory accuracy."
                    },
                    {
                        "title": "Layer potentials C*-algebras of domains with conical points",
                        "abstract": "To a domain with conical points \\Omega, we associate a natural C*-algebra that is motivated by the study of boundary value problems on \\Omega, especially using the method of layer potentials. In two dimensions, we allow \\Omega to be a domain with ramified cracks. We construct an explicit groupoid associated to the boundary of \\Omega and use the theory of pseudodifferential operators on groupoids and its representations to obtain our layer potentials C*-algebra. We study its structure, compute the associated K-groups, and prove Fredholm conditions for the natural pseudodifferential operators affiliated to this C*-algebra."
                    },
                    {
                        "title": "On the second law of thermodynamics: An ideal-gas flow spontaneously induced by a locally nonchaotic energy barrier",
                        "abstract": "People are well aware that, inherently, certain small-scale nonchaotic particle movements are not governed by thermodynamics. Usually, such phenomena are studied by kinetic theory and their energy properties are considered \"trivial\". In current research, we show that, beyond the boundary of the second law of thermodynamics where Boltzmann's H-theorem does not apply, there are also large-sized systems of nontrivial energy properties: when the system is isolated, entropy can decrease; from a single thermal reservoir, the system can absorb heat and produce useful work without any other effect. The key concept is local nonchaoticity, demonstrated by using a narrow energy barrier. The barrier width is much less than the nominal particle mean free path, so that inside the barrier, the particle-particle collisions are sparse and the particle trajectories tend to be locally nonchaotic. Across the barrier, the steady-state particle flux ratio is intrinsically in a non-Boltzmann form. With a step-ramp structure, the nonequilibrium effect spreads to the entire system, and a global flow is generated spontaneously from the random thermal motion. The deviation from thermodynamic equilibrium is steady and significant, and compatible with the basic principle of maximum entropy. These theoretical and numerical analyses may shed light on the fundamentals of thermodynamics and statistical mechanics."
                    },
                    {
                        "title": "DeepWriter: A Multi-Stream Deep CNN for Text-independent Writer Identification",
                        "abstract": "Text-independent writer identification is challenging due to the huge variation of written contents and the ambiguous written styles of different writers. This paper proposes DeepWriter, a deep multi-stream CNN to learn deep powerful representation for recognizing writers. DeepWriter takes local handwritten patches as input and is trained with softmax classification loss. The main contributions are: 1) we design and optimize multi-stream structure for writer identification task; 2) we introduce data augmentation learning to enhance the performance of DeepWriter; 3) we introduce a patch scanning strategy to handle text image with different lengths. In addition, we find that different languages such as English and Chinese may share common features for writer identification, and joint training can yield better performance. Experimental results on IAM and HWDB datasets show that our models achieve high identification accuracy: 99.01% on 301 writers and 97.03% on 657 writers with one English sentence input, 93.85% on 300 writers with one Chinese character input, which outperform previous methods with a large margin. Moreover, our models obtain accuracy of 98.01% on 301 writers with only 4 English alphabets as input."
                    },
                    {
                        "title": "Content Rating Classification for Fan Fiction",
                        "abstract": "Content ratings can enable audiences to determine the suitability of various media products. With the recent advent of fan fiction, the critical issue of fan fiction content ratings has emerged. Whether fan fiction content ratings are done voluntarily or required by regulation, there is the need to automate the content rating classification. The problem is to take fan fiction text and determine the appropriate content rating. Methods for other domains, such as online books, have been attempted though none have been applied to fan fiction. We propose natural language processing techniques, including traditional and deep learning methods, to automatically determine the content rating. We show that these methods produce poor accuracy results for multi-classification. We then demonstrate that treating the problem as a binary classification problem produces better accuracy. Finally, we believe and provide some evidence that the current approach of self-annotating has led to incorrect labels limiting classification results."
                    }
                ]
            },
            "d5bd0b89-696a-4f2d-98d4-cc3741a691bc": {
                "pk": "d5bd0b89-696a-4f2d-98d4-cc3741a691bc",
                "name": "Dahua Lin",
                "collaborators": [
                    "Bo Dai",
                    "Yuanjun Xiong",
                    "Zhirong Wu",
                    "Sijie Yan",
                    "Sanja Fidler",
                    "Xiaoou Tang",
                    "Junjie Yan",
                    "Jianqiao Wangni",
                    "Zhiyu Min",
                    "Zhizhong Li"
                ],
                "domain": [
                    "Computer Vision",
                    "Image Captioning",
                    "Deep Learning",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Contrastive Learning for Image Captioning",
                        "abstract": "Image captioning, a popular topic in computer vision, has achieved substantial progress in recent years. However, the distinctiveness of natural descriptions is often overlooked in previous work. It is closely related to the quality of captions, as distinctive captions are more likely to describe images with their unique aspects. In this work, we propose a new learning method, Contrastive Learning (CL), for image captioning. Specifically, via two constraints formulated on top of a reference model, the proposed method can encourage distinctiveness, while maintaining the overall quality of the generated captions. We tested our method on two challenging datasets, where it improves the baseline model by significant margins. We also showed in our studies that the proposed method is generic and can be used for models with various structures."
                    },
                    {
                        "title": "Learning Sparse Visual Representations with Leaky Capped Norm Regularizers",
                        "abstract": "Sparsity inducing regularization is an important part for learning over-complete visual representations. Despite the popularity of $\\ell_1$ regularization, in this paper, we investigate the usage of non-convex regularizations in this problem. Our contribution consists of three parts. First, we propose the leaky capped norm regularization (LCNR), which allows model weights below a certain threshold to be regularized more strongly as opposed to those above, therefore imposes strong sparsity and only introduces controllable estimation bias. We propose a majorization-minimization algorithm to optimize the joint objective function. Second, our study over monocular 3D shape recovery and neural networks with LCNR outperforms $\\ell_1$ and other non-convex regularizations, achieving state-of-the-art performance and faster convergence. Third, we prove a theoretical global convergence speed on the 3D recovery problem. To the best of our knowledge, this is the first convergence analysis of the 3D recovery problem."
                    },
                    {
                        "title": "Probabilistic Ensemble of Collaborative Filters",
                        "abstract": "Collaborative filtering is an important technique for recommendation. Whereas it has been repeatedly shown to be effective in previous work, its performance remains unsatisfactory in many real-world applications, especially those where the items or users are highly diverse. In this paper, we explore an ensemble-based framework to enhance the capability of a recommender in handling diverse data. Specifically, we formulate a probabilistic model which integrates the items, the users, as well as the associations between them into a generative process. On top of this formulation, we further derive a progressive algorithm to construct an ensemble of collaborative filters. In each iteration, a new filter is derived from re-weighted entries and incorporated into the ensemble. It is noteworthy that while the algorithmic procedure of our algorithm is apparently similar to boosting, it is derived from an essentially different formulation and thus differs in several key technical aspects. We tested the proposed method on three large datasets, and observed substantial improvement over the state of the art, including L2Boost, an effective method based on boosting."
                    },
                    {
                        "title": "Integrating Specialized Classifiers Based on Continuous Time Markov Chain",
                        "abstract": "Specialized classifiers, namely those dedicated to a subset of classes, are often adopted in real-world recognition systems. However, integrating such classifiers is nontrivial. Existing methods, e.g. weighted average, usually implicitly assume that all constituents of an ensemble cover the same set of classes. Such methods can produce misleading predictions when used to combine specialized classifiers. This work explores a novel approach. Instead of combining predictions from individual classifiers directly, it first decomposes the predictions into sets of pairwise preferences, treating them as transition channels between classes, and thereon constructs a continuous-time Markov chain, and use the equilibrium distribution of this chain as the final prediction. This way allows us to form a coherent picture over all specialized predictions. On large public datasets, the proposed method obtains considerable improvement compared to mainstream ensemble methods, especially when the classifier coverage is highly unbalanced."
                    },
                    {
                        "title": "Scalable Estimation of Dirichlet Process Mixture Models on Distributed Data",
                        "abstract": "We consider the estimation of Dirichlet Process Mixture Models (DPMMs) in distributed environments, where data are distributed across multiple computing nodes. A key advantage of Bayesian nonparametric models such as DPMMs is that they allow new components to be introduced on the fly as needed. This, however, posts an important challenge to distributed estimation -- how to handle new components efficiently and consistently. To tackle this problem, we propose a new estimation method, which allows new components to be created locally in individual computing nodes. Components corresponding to the same cluster will be identified and merged via a probabilistic consolidation scheme. In this way, we can maintain the consistency of estimation with very low communication cost. Experiments on large real-world data sets show that the proposed method can achieve high scalability in distributed and asynchronous environments without compromising the mixing performance."
                    },
                    {
                        "title": "Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition",
                        "abstract": "Dynamics of human body skeletons convey significant information for human action recognition. Conventional approaches for modeling skeletons usually rely on hand-crafted parts or traversal rules, thus resulting in limited expressive power and difficulties of generalization. In this work, we propose a novel model of dynamic skeletons called Spatial-Temporal Graph Convolutional Networks (ST-GCN), which moves beyond the limitations of previous methods by automatically learning both the spatial and temporal patterns from data. This formulation not only leads to greater expressive power but also stronger generalization capability. On two large datasets, Kinetics and NTU-RGBD, it achieves substantial improvements over mainstream methods."
                    },
                    {
                        "title": "Motion Guided 3D Pose Estimation from Videos",
                        "abstract": "We propose a new loss function, called motion loss, for the problem of monocular 3D Human pose estimation from 2D pose. In computing motion loss, a simple yet effective representation for keypoint motion, called pairwise motion encoding, is introduced. We design a new graph convolutional network architecture, U-shaped GCN (UGCN). It captures both short-term and long-term motion information to fully leverage the additional supervision from the motion loss. We experiment training UGCN with the motion loss on two large scale benchmarks: Human3.6M and MPI-INF-3DHP. Our model surpasses other state-of-the-art models by a large margin. It also demonstrates strong capacity in producing smooth 3D sequences and recovering keypoint motion."
                    },
                    {
                        "title": "MINI: Mining Implicit Novel Instances for Few-Shot Object Detection",
                        "abstract": "Learning from a few training samples is a desirable ability of an object detector, inspiring the explorations of Few-Shot Object Detection (FSOD). Most existing approaches employ a pretrain-transfer paradigm. The model is first pre-trained on base classes with abundant data and then transferred to novel classes with a few annotated samples. Despite the substantial progress, the FSOD performance is still far behind satisfactory. During pre-training, due to the co-occurrence between base and novel classes, the model is learned to treat the co-occurred novel classes as backgrounds. During transferring, given scarce samples of novel classes, the model suffers from learning discriminative features to distinguish novel instances from backgrounds and base classes. To overcome the obstacles, we propose a novel framework, Mining Implicit Novel Instances (MINI), to mine the implicit novel instances as auxiliary training samples, which widely exist in abundant base data but are not annotated. MINI comprises an offline mining mechanism and an online mining mechanism. The offline mining mechanism leverages a self-supervised discriminative model to collaboratively mine implicit novel instances with a trained FSOD network. Taking the mined novel instances as auxiliary training samples, the online mining mechanism takes a teacher-student framework to simultaneously update the FSOD network and the mined implicit novel instances on the fly. Extensive experiments on PASCAL VOC and MS-COCO datasets show MINI achieves new state-of-the-art performance on any shot and split. The significant performance improvements demonstrate the superiority of our method."
                    },
                    {
                        "title": "Generating Multi-Sentence Lingual Descriptions of Indoor Scenes",
                        "abstract": "This paper proposes a novel framework for generating lingual descriptions of indoor scenes. Whereas substantial efforts have been made to tackle this problem, previous approaches focusing primarily on generating a single sentence for each image, which is not sufficient for describing complex scenes. We attempt to go beyond this, by generating coherent descriptions with multiple sentences. Our approach is distinguished from conventional ones in several aspects: (1) a 3D visual parsing system that jointly infers objects, attributes, and relations; (2) a generative grammar learned automatically from training text; and (3) a text generation algorithm that takes into account the coherence among sentences. Experiments on the augmented NYU-v2 dataset show that our framework can generate natural descriptions with substantially higher ROGUE scores compared to those produced by the baseline."
                    },
                    {
                        "title": "Adjustable Bounded Rectifiers: Towards Deep Binary Representations",
                        "abstract": "Binary representation is desirable for its memory efficiency, computation speed and robustness. In this paper, we propose adjustable bounded rectifiers to learn binary representations for deep neural networks. While hard constraining representations across layers to be binary makes training unreasonably difficult, we softly encourage activations to diverge from real values to binary by approximating step functions. Our final representation is completely binary. We test our approach on MNIST, CIFAR10, and ILSVRC2012 dataset, and systematically study the training dynamics of the binarization process. Our approach can binarize the last layer representation without loss of performance and binarize all the layers with reasonably small degradations. The memory space that it saves may allow more sophisticated models to be deployed, thus compensating the loss. To the best of our knowledge, this is the first work to report results on current deep network architectures using complete binary middle representations. Given the learned representations, we find that the firing or inhibition of a binary neuron is usually associated with a meaningful interpretation across different classes. This suggests that the semantic structure of a neural network may be manifested through a guided binarization process."
                    },
                    {
                        "title": "Deep Markov Random Field for Image Modeling",
                        "abstract": "Markov Random Fields (MRFs), a formulation widely used in generative image modeling, have long been plagued by the lack of expressive power. This issue is primarily due to the fact that conventional MRFs formulations tend to use simplistic factors to capture local patterns. In this paper, we move beyond such limitations, and propose a novel MRF model that uses fully-connected neurons to express the complex interactions among pixels. Through theoretical analysis, we reveal an inherent connection between this model and recurrent neural networks, and thereon derive an approximated feed-forward network that couples multiple RNNs along opposite directions. This formulation combines the expressive power of deep neural networks and the cyclic dependency structure of MRF in a unified model, bringing the modeling capability to a new level. The feed-forward approximation also allows it to be efficiently learned from data. Experimental results on a variety of low-level vision tasks show notable improvement over state-of-the-arts."
                    },
                    {
                        "title": "Detecting Visual Relationships with Deep Relational Networks",
                        "abstract": "Relationships among objects play a crucial role in image understanding. Despite the great success of deep learning techniques in recognizing individual objects, reasoning about the relationships among objects remains a challenging task. Previous methods often treat this as a classification problem, considering each type of relationship (e.g. \"ride\") or each distinct visual phrase (e.g. \"person-ride-horse\") as a category. Such approaches are faced with significant difficulties caused by the high diversity of visual appearance for each kind of relationships or the large number of distinct visual phrases. We propose an integrated framework to tackle this problem. At the heart of this framework is the Deep Relational Network, a novel formulation designed specifically for exploiting the statistical dependencies between objects and their relationships. On two large datasets, the proposed method achieves substantial improvement over state-of-the-art."
                    },
                    {
                        "title": "Peephole: Predicting Network Performance Before Training",
                        "abstract": "The quest for performant networks has been a significant force that drives the advancements of deep learning in recent years. While rewarding, improving network design has never been an easy journey. The large design space combined with the tremendous cost required for network training poses a major obstacle to this endeavor. In this work, we propose a new approach to this problem, namely, predicting the performance of a network before training, based on its architecture. Specifically, we develop a unified way to encode individual layers into vectors and bring them together to form an integrated description via LSTM. Taking advantage of the recurrent network's strong expressive power, this method can reliably predict the performances of various network architectures. Our empirical studies showed that it not only achieved accurate predictions but also produced consistent rankings across datasets -- a key desideratum in performance prediction."
                    },
                    {
                        "title": "Accelerated Training for Massive Classification via Dynamic Class Selection",
                        "abstract": "Massive classification, a classification task defined over a vast number of classes (hundreds of thousands or even millions), has become an essential part of many real-world systems, such as face recognition. Existing methods, including the deep networks that achieved remarkable success in recent years, were mostly devised for problems with a moderate number of classes. They would meet with substantial difficulties, e.g. excessive memory demand and computational cost, when applied to massive problems. We present a new method to tackle this problem. This method can efficiently and accurately identify a small number of \"active classes\" for each mini-batch, based on a set of dynamic class hierarchies constructed on the fly. We also develop an adaptive allocation scheme thereon, which leads to a better tradeoff between performance and cost. On several large-scale benchmarks, our method significantly reduces the training cost and memory demand, while maintaining competitive performance."
                    },
                    {
                        "title": "Unsupervised Feature Learning via Non-Parametric Instance-level Discrimination",
                        "abstract": "Neural net classifiers trained on data with annotated class labels can also capture apparent visual similarity among categories without being directed to do so. We study whether this observation can be extended beyond the conventional domain of supervised learning: Can we learn a good feature representation that captures apparent similarity among instances, instead of classes, by merely asking the feature to be discriminative of individual instances? We formulate this intuition as a non-parametric classification problem at the instance-level, and use noise-contrastive estimation to tackle the computational challenges imposed by the large number of instance classes. Our experimental results demonstrate that, under unsupervised learning settings, our method surpasses the state-of-the-art on ImageNet classification by a large margin. Our method is also remarkable for consistently improving test performance with more training data and better network architectures. By fine-tuning the learned feature, we further obtain competitive results for semi-supervised learning and object detection tasks. Our non-parametric model is highly compact: With 128 features per image, our method requires only 600MB storage for a million images, enabling fast nearest neighbour retrieval at the run time."
                    },
                    {
                        "title": "A Neural Compositional Paradigm for Image Captioning",
                        "abstract": "Mainstream captioning models often follow a sequential structure to generate captions, leading to issues such as introduction of irrelevant semantics, lack of diversity in the generated captions, and inadequate generalization performance. In this paper, we present an alternative paradigm for image captioning, which factorizes the captioning procedure into two stages: (1) extracting an explicit semantic representation from the given image; and (2) constructing the caption based on a recursive compositional procedure in a bottom-up manner. Compared to conventional ones, our paradigm better preserves the semantic content through an explicit factorization of semantics and syntax. By using the compositional generation procedure, caption construction follows a recursive structure, which naturally fits the properties of human language. Moreover, the proposed compositional procedure requires less data to train, generalizes better, and yields more diverse captions."
                    },
                    {
                        "title": "Recursive Visual Sound Separation Using Minus-Plus Net",
                        "abstract": "Sounds provide rich semantics, complementary to visual data, for many tasks. However, in practice, sounds from multiple sources are often mixed together. In this paper we propose a novel framework, referred to as MinusPlus Network (MP-Net), for the task of visual sound separation. MP-Net separates sounds recursively in the order of average energy, removing the separated sound from the mixture at the end of each prediction, until the mixture becomes empty or contains only noise. In this way, MP-Net could be applied to sound mixtures with arbitrary numbers and types of sounds. Moreover, while MP-Net keeps removing sounds with large energy from the mixture, sounds with small energy could emerge and become clearer, so that the separation is more accurate. Compared to previous methods, MP-Net obtains state-of-the-art results on two large scale datasets, across mixtures with different types and numbers of sounds."
                    },
                    {
                        "title": "Low-Latency Video Semantic Segmentation",
                        "abstract": "Recent years have seen remarkable progress in semantic segmentation. Yet, it remains a challenging task to apply segmentation techniques to video-based applications. Specifically, the high throughput of video streams, the sheer cost of running fully convolutional networks, together with the low-latency requirements in many real-world applications, e.g. autonomous driving, present a significant challenge to the design of the video segmentation framework. To tackle this combined challenge, we develop a framework for video semantic segmentation, which incorporates two novel components: (1) a feature propagation module that adaptively fuses features over time via spatially variant convolution, thus reducing the cost of per-frame computation; and (2) an adaptive scheduler that dynamically allocate computation based on accuracy prediction. Both components work together to ensure low latency while maintaining high segmentation quality. On both Cityscapes and CamVid, the proposed framework obtained competitive performance compared to the state of the art, while substantially reducing the latency, from 360 ms to 119 ms."
                    },
                    {
                        "title": "Unifying Identification and Context Learning for Person Recognition",
                        "abstract": "Despite the great success of face recognition techniques, recognizing persons under unconstrained settings remains challenging. Issues like profile views, unfavorable lighting, and occlusions can cause substantial difficulties. Previous works have attempted to tackle this problem by exploiting the context, e.g. clothes and social relations. While showing promising improvement, they are usually limited in two important aspects, relying on simple heuristics to combine different cues and separating the construction of context from people identities. In this work, we aim to move beyond such limitations and propose a new framework to leverage context for person recognition. In particular, we propose a Region Attention Network, which is learned to adaptively combine visual cues with instance-dependent weights. We also develop a unified formulation, where the social contexts are learned along with the reasoning of people identities. These models substantially improve the robustness when working with the complex contextual relations in unconstrained environments. On two large datasets, PIPA and Cast In Movies (CIM), a new dataset proposed in this work, our method consistently achieves state-of-the-art performance under multiple evaluation policies."
                    },
                    {
                        "title": "Rethinking the Form of Latent States in Image Captioning",
                        "abstract": "RNNs and their variants have been widely adopted for image captioning. In RNNs, the production of a caption is driven by a sequence of latent states. Existing captioning models usually represent latent states as vectors, taking this practice for granted. We rethink this choice and study an alternative formulation, namely using two-dimensional maps to encode latent states. This is motivated by the curiosity about a question: how the spatial structures in the latent states affect the resultant captions? Our study on MSCOCO and Flickr30k leads to two significant observations. First, the formulation with 2D states is generally more effective in captioning, consistently achieving higher performance with comparable parameter sizes. Second, 2D states preserve spatial locality. Taking advantage of this, we visually reveal the internal dynamics in the process of caption generation, as well as the connections between input visual domain and output linguistic domain."
                    }
                ]
            },
            "45d3dab9-03ed-4da3-9e0b-6254b4d0840b": {
                "pk": "45d3dab9-03ed-4da3-9e0b-6254b4d0840b",
                "name": "Feng Zhao",
                "collaborators": [],
                "domain": [
                    "Statistical Modeling",
                    "Protein Folding",
                    "Machine Learning",
                    "Conditional Random Fields"
                ],
                "publications": [
                    {
                        "title": "Moments of the multivariate Beta distribution",
                        "abstract": "In this paper, we extend Beta distribution to 2 by 2 matrix and give the analytical formula for its moments. Our analytical formula can be used to analyze the asymptotic behavior of Beta distribution for 2 by 2 matrix."
                    },
                    {
                        "title": "Structural Learning for Template-free Protein Folding",
                        "abstract": "The thesis is aimed to solve the template-free protein folding problem by tackling two important components: efficient sampling in vast conformation space, and design of knowledge-based potentials with high accuracy. We have proposed the first-order and second-order CRF-Sampler to sample structures from the continuous local dihedral angles space by modeling the lower and higher order conditional dependency between neighboring dihedral angles given the primary sequence information. A framework combining the Conditional Random Fields and the energy function is introduced to guide the local conformation sampling using long range constraints with the energy function.   The relationship between the sequence profile and the local dihedral angle distribution is nonlinear. Hence we proposed the CNF-Folder to model this complex relationship by applying a novel machine learning model Conditional Neural Fields which utilizes the structural graphical model with the neural network. CRF-Samplers and CNF-Folder perform very well in CASP8 and CASP9.   Further, a novel pairwise distance statistical potential (EPAD) is designed to capture the dependency of the energy profile on the positions of the interacting amino acids as well as the types of those amino acids, opposing the common assumption that this energy profile depends only on the types of amino acids. EPAD has also been successfully applied in the CASP 10 Free Modeling experiment with CNF-Folder, especially outstanding on some uncommon structured targets."
                    }
                ]
            },
            "23ffba67-fd50-4fa1-8f7c-95593346779e": {
                "pk": "23ffba67-fd50-4fa1-8f7c-95593346779e",
                "name": "Jiaqi Wang",
                "collaborators": [
                    "Xi Li",
                    "Chen Wu",
                    "Shuzhen Yang",
                    "Daqing Yang",
                    "Dongyan Fu",
                    "Yubing Dong",
                    "Hang Xie",
                    "Yihong Wu",
                    "Yong Ao",
                    "Wenming Zou"
                ],
                "domain": [
                    "Federated Learning",
                    "Privacy",
                    "Machine Learning",
                    "Graph Theory"
                ],
                "publications": [
                    {
                        "title": "An In-depth Review of Privacy Concerns Raised by the COVID-19 Pandemic",
                        "abstract": "COVID-19 has hugely changed our lives, work, and interactions with people. With more and more online activities, people are easily exposed to privacy threats. In this paper, we explore how users self-disclose on social media and privacy concerns raised from these behaviors. Based on recent news, techniques, and research, we indicate three increasing privacy threats caused by the COVID-19 pandemic. After that, we provide a systematic analysis of potential privacy issues related to the COVID pandemic. Furthermore, we propose a series of research directions about online user self-disclosure and privacy issues for future work as well as possible solutions."
                    },
                    {
                        "title": "Stochastic maximum principle for recursive optimal control problems with varying terminal time",
                        "abstract": "This paper introduces a new recursive stochastic optimal control problem driven by a forward-backward stochastic differential equations (FBSDEs), where the ter?minal time varies according to the constraints of the state of the forward equation. This new optimal control problem can be used to describe the investment portfolio problems with the varying investment period. Based on novel \\r{ho}-moving variational and adjoint equations, we establish the stochastic maximum principle for this optimal control problem including the classical optimal control problem as a particular case. Furthermore, we propose an example to verify our main results."
                    },
                    {
                        "title": "Position Paper: Assessing Robustness, Privacy, and Fairness in Federated Learning Integrated with Foundation Models",
                        "abstract": "Federated Learning (FL), while a breakthrough in decentralized machine learning, contends with significant challenges such as limited data availability and the variability of computational resources, which can stifle the performance and scalability of the models. The integration of Foundation Models (FMs) into FL presents a compelling solution to these issues, with the potential to enhance data richness and reduce computational demands through pre-training and data augmentation. However, this incorporation introduces novel issues in terms of robustness, privacy, and fairness, which have not been sufficiently addressed in the existing research. We make a preliminary investigation into this field by systematically evaluating the implications of FM-FL integration across these dimensions. We analyze the trade-offs involved, uncover the threats and issues introduced by this integration, and propose a set of criteria and strategies for navigating these challenges. Furthermore, we identify potential research directions for advancing this field, laying a foundation for future development in creating reliable, secure, and equitable FL systems."
                    },
                    {
                        "title": "Queue layouts and nonrepetitive colouring of planar graphs and powers of trees",
                        "abstract": "Dujmovi\\'{c}, Joret, Micek, Morin, Ueckerdt and Wood recently in [Planar graphs have bounded queue-number, Journal of the ACM, Volume 67, Issue 4, Article No.: 22, August 2020] showed some attractive graph product structure theorems for planar graphs. By using the product structure, they proved that planar graphs have bounded queue-number $48$; in [Planar graphs have bounded nonrepetitive chromatic number, Advances in Combinatorics, 5, 11 pp, 2020], the authors proved that planar graphs have bounded nonrepetitive chromatic number $768$.   In this paper, still by using some product structure theorem, we improve the upper bound of queue-number of planar graphs to $27$ and the non-repetitive chromatic number to $320$. We also study powers of trees. We show a graph product structure theorem of the $k$-th power $T^k$ of tree $T$, then use it giving an upper bound of the nonrepetitive~chromatic~number of $T^k$. We also give an asymptotically tight upper bound of the queue-number of $T^k$."
                    },
                    {
                        "title": "A systematical study of the nucleon form factors with the pion cloud effect",
                        "abstract": "The electromagnetic and gravitational form factors of the nucleon are studied simultaneously using a covariant quark-diquark approach, and the pion cloud effect on the form factors is explicitly discussed. In this study, the electromagnetic form factors are first calculated to determine the parameters of our approach. Then, the gravitational form factors of the nucleon are evaluated with the same parameters and the cloud effect is addressed. The mechanical properties, including mass radii, energy densities, and spin distributions, are shown and discussed."
                    },
                    {
                        "title": "Unveiling Backdoor Risks Brought by Foundation Models in Heterogeneous Federated Learning",
                        "abstract": "The foundation models (FMs) have been used to generate synthetic public datasets for the heterogeneous federated learning (HFL) problem where each client uses a unique model architecture. However, the vulnerabilities of integrating FMs, especially against backdoor attacks, are not well-explored in the HFL contexts. In this paper, we introduce a novel backdoor attack mechanism for HFL that circumvents the need for client compromise or ongoing participation in the FL process. This method plants and transfers the backdoor through a generated synthetic public dataset, which could help evade existing backdoor defenses in FL by presenting normal client behaviors. Empirical experiments across different HFL configurations and benchmark datasets demonstrate the effectiveness of our attack compared to traditional client-based attacks. Our findings reveal significant security risks in developing robust FM-assisted HFL systems. This research contributes to enhancing the safety and integrity of FL systems, highlighting the need for advanced security measures in the era of FMs."
                    },
                    {
                        "title": "Vulnerabilities of Foundation Model Integrated Federated Learning Under Adversarial Threats",
                        "abstract": "Federated Learning (FL) addresses critical issues in machine learning related to data privacy and security, yet suffering from data insufficiency and imbalance under certain circumstances. The emergence of foundation models (FMs) offers potential solutions to the limitations of existing FL frameworks, e.g., by generating synthetic data for model initialization. However, due to the inherent safety concerns of FMs, integrating FMs into FL could introduce new risks, which remains largely unexplored. To address this gap, we conduct the first investigation on the vulnerability of FM integrated FL (FM-FL) under adversarial threats. Based on a unified framework of FM-FL, we introduce a novel attack strategy that exploits safety issues of FM to compromise FL client models. Through extensive experiments with well-known models and benchmark datasets in both image and text domains, we reveal the high susceptibility of the FM-FL to this new threat under various FL configurations. Furthermore, we find that existing FL defense strategies offer limited protection against this novel attack approach. This research highlights the critical need for enhanced security measures in FL in the era of FMs."
                    },
                    {
                        "title": "Angular Position Sensor Based on Anisotropic Magnetoresistive and Anomalous Nernst Effect",
                        "abstract": "Magnetic position sensors find extensive applications in various industrial sectors and consumer products. However, measuring angles in the full range of 0{\\deg} to 360{\\deg} in a wide field range using a single magnetic sensor remains a challenge. Here, we propose a magnetic position sensor based on a single Wheatstone bridge structure made from a single ferromagnetic layer. By measuring the anisotropic magnetoresistance (AMR) signal from the bridge and two sets of anomalous Nernst effect (ANE) signals from the transverse ports on two perpendicular Wheatstone bridge arms concurrently, we show that it is possible to achieve 0{\\deg} to 360{\\deg} angle detection using a single bridge sensor. The combined use of AMR and ANE signals allows to achieve a mean angle error in the range of 0.51{\\deg} to 1.05{\\deg} within a field range of 100 Oe to 10,000 Oe."
                    },
                    {
                        "title": "On the existence and regularity of vector solutions for quasilinear systems with linear coupling",
                        "abstract": "We study a class of linearly coupled system of quasilinear equations. Under some assumptions on the nonlinear terms, we establish some results about the existence and regularity of vector solutions for the p-Laplacian systems by using variational methods. In particular, we get two pairs of nontrivial solutions. We also study their different asymptotic behavior of solutions as the coupling parameter tends to zero."
                    },
                    {
                        "title": "Hierarchical Attention Generative Adversarial Networks for Cross-domain Sentiment Classification",
                        "abstract": "Cross-domain sentiment classification (CDSC) is an importance task in domain adaptation and sentiment classification. Due to the domain discrepancy, a sentiment classifier trained on source domain data may not works well on target domain data. In recent years, many researchers have used deep neural network models for cross-domain sentiment classification task, many of which use Gradient Reversal Layer (GRL) to design an adversarial network structure to train a domain-shared sentiment classifier. Different from those methods, we proposed Hierarchical Attention Generative Adversarial Networks (HAGAN) which alternately trains a generator and a discriminator in order to produce a document representation which is sentiment-distinguishable but domain-indistinguishable. Besides, the HAGAN model applies Bidirectional Gated Recurrent Unit (Bi-GRU) to encode the contextual information of a word and a sentence into the document representation. In addition, the HAGAN model use hierarchical attention mechanism to optimize the document representation and automatically capture the pivots and non-pivots. The experiments on Amazon review dataset show the effectiveness of HAGAN."
                    }
                ]
            }
        }
    },
    "2310.13061": {
        "paper_data": {
            "title": "To grok or not to grok: Disentangling generalization and memorization on corrupted algorithmic datasets",
            "url": "http://arxiv.org/abs/2310.13061v2",
            "arxiv_id": "2310.13061",
            "authors": [
                "Darshil Doshi",
                "Aritra Das",
                "Tianyu He",
                "Andrey Gromov"
            ],
            "abstract": "Robust generalization is a major challenge in deep learning, particularly when the number of trainable parameters is very large. In general, it is very difficult to know if the network has memorized a particular set of examples or understood the underlying rule (or both). Motivated by this challenge, we study an interpretable model where generalizing representations are understood analytically, and are easily distinguishable from the memorizing ones. Namely, we consider multi-layer perceptron (MLP) and Transformer architectures trained on modular arithmetic tasks, where ($\\xi \\cdot 100\\%$) of labels are corrupted (\\emph{i.e.} some results of the modular operations in the training set are incorrect). We show that (i) it is possible for the network to memorize the corrupted labels \\emph{and} achieve $100\\%$ generalization at the same time; (ii) the memorizing neurons can be identified and pruned, lowering the accuracy on corrupted data and improving the accuracy on uncorrupted data; (iii) regularization methods such as weight decay, dropout and BatchNorm force the network to ignore the corrupted data during optimization, and achieve $100\\%$ accuracy on the uncorrupted dataset; and (iv) the effect of these regularization methods is (``mechanistically'') interpretable: weight decay and dropout force all the neurons to learn generalizing representations, while BatchNorm de-amplifies the output of memorizing neurons and amplifies the output of the generalizing ones. Finally, we show that in the presence of regularization, the training dynamics involves two consecutive stages: first, the network undergoes \\emph{grokking} dynamics reaching high train \\emph{and} test accuracy; second, it unlearns the memorizing representations, where the train accuracy suddenly jumps from $100\\%$ to $100 (1-\\xi)\\%$.",
            "introduction": "   1 Introduction  The astounding progress of deep learning in the last decade has been facilitated by massive, high-quality datasets. Annotated real-world datasets inevitably contain noisy labels, due to biases of annotation schemes (Paolacci et\u00a0al., 2010; Cothey, 2004) or inherent ambiguity (Beyer et\u00a0al., 2020). A key challenge in training large models is to prevent overfitting the noisy data and attain robust generalization performance. On the other hand, in large models, it is possible for memorization and generalization to coexist (Zhang et\u00a0al., 2017; 2021). By and large, the tussle between memorization and generalization, especially in the presence of label corruption, remains poorly understood.   In generative language models the problem of memorization is even more nuanced. On the one hand, some factual knowledge is critical for the language models to produce accurate information. On the other hand, verbatim memorization of the training data is generally unwanted due to privacy concerns (Carlini et\u00a0al., 2021). The full scope of the conditions that affect memorization in large language models (LLMs) is still under investigation (Carlini et\u00a0al., 2022; Tirumala et\u00a0al., 2022). In particular, larger models as well as repeated data (Hernandez et\u00a0al., 2022) favor memorization.   In this work, we take an approach to disentangling generalization and memorization motivated by theoretical physics. Namely, we focus on simple, analytically tractable, (\u201cmechanistically\u201d) interpretable models trained on algorithmic datasets. More concretely, we study two-layer Multilayer Perceptrons (MLP) trained on modular arithmetic tasks. In this case, the network is trained to learn a rule, such as z=(m+n)%\u2062p\u2261m+n\u2062mod\u2062p\ud835\udc67percent\ud835\udc5a\ud835\udc5b\ud835\udc5d\ud835\udc5a\ud835\udc5bmod\ud835\udc5dz=(m+n)\\%p\\equiv m+n\\,\\,\\textrm{mod}\\,\\,pitalic_z = ( italic_m + italic_n ) % italic_p \u2261 italic_m + italic_n mod italic_p from a number of examples. These tasks are phrased as classification problems. We then force the network to memorize \u03be\u22c5100%\u22c5\ud835\udf09percent100\\xi\\cdot 100\\%italic_\u03be \u22c5 100 % of the examples by corrupting the labels, i.e. injecting data points where (m+n)%\u2062ppercent\ud835\udc5a\ud835\udc5b\ud835\udc5d(m+n)\\%p( italic_m + italic_n ) % italic_p is mapped to z\u2032\u2260zsuperscript\ud835\udc67\u2032\ud835\udc67z^{\\prime}\\neq zitalic_z start_POSTSUPERSCRIPT \u2032 end_POSTSUPERSCRIPT \u2260 italic_z. In this setting, the generalizing representations are understood qualitatively Nanda et\u00a0al. (2023); Liu et\u00a0al. (2022) as well as quantitatively, thanks to the availability of an analytic solution for the representations described in Gromov (2023). We empirically show that sufficiently large networks can memorize the corrupted examples while achieving nearly perfect generalization on the test set. The neurons responsible for memorization can be identified using inverse participation ratio (IPR) and pruned away, leading to perfect accuracy on the un-corrupted dataset. We further show that regularization dramatically affects the network\u2019s ability to memorize the corrupted examples. Weight decay, Dropout and BatchNorm all prevent memorization, but in different ways: Weight decay and Dropout eliminate the memorizing neurons by converting them into generalizing ones, while BatchNorm de-amplifies the signal coming from the memorizing neurons without eliminating them.    1.1 Preliminaries  Figure 1: Training curves in various phases. All plots are made for networks trained with data-fraction \u03b1=0.5\ud835\udefc0.5\\alpha=0.5italic_\u03b1 = 0.5 and corruption-fraction \u03be=0.35\ud835\udf090.35\\xi=0.35italic_\u03be = 0.35. (a) No regularization: Coexistence of generalization and memorization \u2013 both train and test accuracies are high. (b)(c) Adding weight decay: The network generalizes on the test data but does not memorize the corrupted training data, resulting in a negative generalization gap! Remarkably, the network predicts the \u201ctrue\u201d",
            "references": [
                {
                    "title": "The Clock and the Pizza: Two Stories in Mechanistic Explanation of Neural Networks",
                    "abstract": "Do neural networks, trained on well-understood algorithmic tasks, reliably rediscover known algorithms for solving those tasks? Several recent studies, on tasks ranging from group arithmetic to in-context linear regression, have suggested that the answer is yes. Using modular addition as a prototypical problem, we show that algorithm discovery in neural networks is sometimes more complex. Small changes to model hyperparameters and initializations can induce the discovery of qualitatively different algorithms from a fixed training set, and even parallel implementations of multiple such algorithms. Some networks trained to perform modular addition implement a familiar Clock algorithm; others implement a previously undescribed, less intuitive, but comprehensible procedure which we term the Pizza algorithm, or a variety of even more complex procedures. Our results show that even simple learning problems can admit a surprising diversity of solutions, motivating the development of new tools for characterizing the behavior of neural networks across their algorithmic phase space."
                },
                {
                    "title": "Predicting Grokking Long Before it Happens: A look into the loss landscape of models which grok",
                    "abstract": "This paper focuses on predicting the occurrence of grokking in neural networks, a phenomenon in which perfect generalization emerges long after signs of overfitting or memorization are observed. It has been reported that grokking can only be observed with certain hyper-parameters. This makes it critical to identify the parameters that lead to grokking. However, since grokking occurs after a large number of epochs, searching for the hyper-parameters that lead to it is time-consuming. In this paper, we propose a low-cost method to predict grokking without training for a large number of epochs. In essence, by studying the learning curve of the first few epochs, we show that one can predict whether grokking will occur later on. Specifically, if certain oscillations occur in the early epochs, one can expect grokking to occur if the model is trained for a much longer period of time. We propose using the spectral signature of a learning curve derived by applying the Fourier transform to quantify the amplitude of low-frequency components to detect the presence of such oscillations. We also present additional experiments aimed at explaining the cause of these oscillations and characterizing the loss landscape."
                },
                {
                    "title": "A Tale of Two Circuits: Grokking as Competition of Sparse and Dense Subnetworks",
                    "abstract": "Grokking is a phenomenon where a model trained on an algorithmic task first overfits but, then, after a large amount of additional training, undergoes a phase transition to generalize perfectly. We empirically study the internal structure of networks undergoing grokking on the sparse parity task, and find that the grokking phase transition corresponds to the emergence of a sparse subnetwork that dominates model predictions. On an optimization level, we find that this subnetwork arises when a small subset of neurons undergoes rapid norm growth, whereas the other neurons in the network decay slowly in norm. Thus, we suggest that the grokking phase transition can be understood to emerge from competition of two largely distinct subnetworks: a dense one that dominates before the transition and generalizes poorly, and a sparse one that dominates afterwards."
                },
                {
                    "title": "Progress measures for grokking via mechanistic interpretability",
                    "abstract": "Neural networks often exhibit emergent behavior, where qualitatively new capabilities arise from scaling up the amount of parameters, training data, or training steps. One approach to understanding emergence is to find continuous \\textit{progress measures} that underlie the seemingly discontinuous qualitative changes. We argue that progress measures can be found via mechanistic interpretability: reverse-engineering learned behaviors into their individual components. As a case study, we investigate the recently-discovered phenomenon of ``grokking'' exhibited by small transformers trained on modular addition tasks. We fully reverse engineer the algorithm learned by these networks, which uses discrete Fourier transforms and trigonometric identities to convert addition to rotation about a circle. We confirm the algorithm by analyzing the activations and weights and by performing ablations in Fourier space. Based on this understanding, we define progress measures that allow us to study the dynamics of training and split training into three continuous phases: memorization, circuit formation, and cleanup. Our results show that grokking, rather than being a sudden shift, arises from the gradual amplification of structured mechanisms encoded in the weights, followed by the later removal of memorizing components."
                },
                {
                    "title": "Grokking phase transitions in learning local rules with gradient descent",
                    "abstract": "We discuss two solvable grokking (generalisation beyond overfitting) models in a rule learning scenario. We show that grokking is a phase transition and find exact analytic expressions for the critical exponents, grokking probability, and grokking time distribution. Further, we introduce a tensor-network map that connects the proposed grokking setup with the standard (perceptron) statistical learning theory and show that grokking is a consequence of the locality of the teacher model. As an example, we analyse the cellular automata learning task, numerically determine the critical exponent and the grokking time distributions and compare them with the prediction of the proposed grokking model. Finally, we numerically analyse the connection between structure formation and grokking."
                },
                {
                    "title": "Omnigrok: Grokking Beyond Algorithmic Data",
                    "abstract": "Grokking, the unusual phenomenon for algorithmic datasets where generalization happens long after overfitting the training data, has remained elusive. We aim to understand grokking by analyzing the loss landscapes of neural networks, identifying the mismatch between training and test losses as the cause for grokking. We refer to this as the\"LU mechanism\"because training and test losses (against model weight norm) typically resemble\"L\"and\"U\", respectively. This simple mechanism can nicely explain many aspects of grokking: data size dependence, weight decay dependence, the emergence of representations, etc. Guided by the intuitive picture, we are able to induce grokking on tasks involving images, language and molecules. In the reverse direction, we are able to eliminate grokking for algorithmic datasets. We attribute the dramatic nature of grokking for algorithmic datasets to representation learning."
                },
                {
                    "title": "Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limit",
                    "abstract": "There is mounting evidence of emergent phenomena in the capabilities of deep learning methods as we scale up datasets, model sizes, and training times. While there are some accounts of how these resources modulate statistical capacity, far less is known about their effect on the computational problem of model training. This work conducts such an exploration through the lens of learning a $k$-sparse parity of $n$ bits, a canonical discrete search problem which is statistically easy but computationally hard. Empirically, we find that a variety of neural networks successfully learn sparse parities, with discontinuous phase transitions in the training curves. On small instances, learning abruptly occurs at approximately $n^{O(k)}$ iterations; this nearly matches SQ lower bounds, despite the apparent lack of a sparse prior. Our theoretical analysis shows that these observations are not explained by a Langevin-like mechanism, whereby SGD\"stumbles in the dark\"until it finds the hidden set of features (a natural algorithm which also runs in $n^{O(k)}$ time). Instead, we show that SGD gradually amplifies the sparse solution via a Fourier gap in the population gradient, making continual progress that is invisible to loss and error metrics."
                },
                {
                    "title": "Review\u2013A Survey of Learning from Noisy Labels",
                    "abstract": "Deep Learning has achieved remarkable successes in many industry applications and scientific research fields. One essential reason is that deep models can learn rich information from large-scale training datasets through supervised learning. It has been well accepted that the robust deep models heavily rely on the quality of data labels. However, current large-scale datasets mostly involve noisy labels, which are caused by sensor errors, human mistakes, or inaccuracy of search engines, and may severely degrade the performance of deep models. In this survey, we summaries existing works on noisy label learning into two main categories, Loss Correction and Sample Selection, and present their methodologies, commonly used experimental setups, datasets, and the state-of-the-art results. Finally, we discuss a promising research direction that might be valuable for the future study."
                },
                {
                    "title": "Memorization Without Overfitting: Analyzing the Training Dynamics of Large Language Models",
                    "abstract": "Despite their wide adoption, the underlying training and memorization dynamics of very large language models is not well understood. We empirically study exact memorization in causal and masked language modeling, across model sizes and throughout the training process. We measure the effects of dataset size, learning rate, and model size on memorization, finding that larger language models memorize training data faster across all settings. Surprisingly, we show that larger models can memorize a larger portion of the data before over-fitting and tend to forget less throughout the training process. We also analyze the memorization dynamics of different parts of speech and find that models memorize nouns and numbers first; we hypothesize and provide empirical evidence that nouns and numbers act as a unique identifier for memorizing individual training examples. Together, these findings present another piece of the broader puzzle of trying to understand what actually improves as models get bigger."
                },
                {
                    "title": "Scaling Laws and Interpretability of Learning from Repeated Data",
                    "abstract": "Recent large language models have been trained on vast datasets, but also often on repeated data, either intentionally for the purpose of upweighting higher quality data, or unintentionally because data deduplication is not perfect and the model is exposed to repeated data at the sentence, paragraph, or document level. Some works have reported substantial negative performance effects of this repeated data. In this paper we attempt to study repeated data systematically and to understand its effects mechanistically. To do this, we train a family of models where most of the data is unique but a small fraction of it is repeated many times. We find a strong double descent phenomenon, in which repeated data can lead test loss to increase midway through training. A predictable range of repetition frequency leads to surprisingly severe degradation in performance. For instance, performance of an 800M parameter model can be degraded to that of a 2x smaller model (400M params) by repeating 0.1% of the data 100 times, despite the other 90% of the training tokens remaining unique. We suspect there is a range in the middle where the data can be memorized and doing so consumes a large fraction of the model's capacity, and this may be where the peak of degradation occurs. Finally, we connect these observations to recent mechanistic interpretability work - attempting to reverse engineer the detailed computations performed by the model - by showing that data repetition disproportionately damages copying and internal structures associated with generalization, such as induction heads, providing a possible mechanism for the shift from generalization to memorization. Taken together, these results provide a hypothesis for why repeating a relatively small fraction of data in large language models could lead to disproportionately large harms to performance."
                },
                {
                    "title": "Towards Understanding Grokking: An Effective Theory of Representation Learning",
                    "abstract": "We aim to understand grokking, a phenomenon where models generalize long after overfitting their training set. We present both a microscopic analysis anchored by an effective theory and a macroscopic analysis of phase diagrams describing learning performance across hyperparameters. We find that generalization originates from structured representations whose training dynamics and dependence on training set size can be predicted by our effective theory in a toy setting. We observe empirically the presence of four learning phases: comprehension, grokking, memorization, and confusion. We find representation learning to occur only in a\"Goldilocks zone\"(including comprehension and grokking) between memorization and confusion. We find on transformers the grokking phase stays closer to the memorization phase (compared to the comprehension phase), leading to delayed generalization. The Goldilocks phase is reminiscent of\"intelligence from starvation\"in Darwinian evolution, where resource limitations drive discovery of more efficient solutions. This study not only provides intuitive explanations of the origin of grokking, but also highlights the usefulness of physics-inspired tools, e.g., effective theories and phase diagrams, for understanding deep learning."
                },
                {
                    "title": "Quantifying Memorization Across Neural Language Models",
                    "abstract": "Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations."
                },
                {
                    "title": "Deep Learning on a Data Diet: Finding Important Examples Early in Training",
                    "abstract": "Recent success in deep learning has partially been driven by training increasingly overparametrized networks on ever larger datasets. It is therefore natural to ask: how much of the data is superfluous, which examples are important for generalization, and how do we find them? In this work, we make the striking observation that, in standard vision datasets, simple scores averaged over several weight initializations can be used to identify important examples very early in training. We propose two such scores -- the Gradient Normed (GraNd) and the Error L2-Norm (EL2N) scores -- and demonstrate their efficacy on a range of architectures and datasets by pruning significant fractions of training data without sacrificing test accuracy. In fact, using EL2N scores calculated a few epochs into training, we can prune half of the CIFAR10 training set while slightly improving test accuracy. Furthermore, for a given dataset, EL2N scores from one architecture or hyperparameter configuration generalize to other configurations. Compared to recent work that prunes data by discarding examples that are rarely forgotten over the course of training, our scores use only local information early in training. We also use our scores to detect noisy examples and study training dynamics through the lens of important examples -- we investigate how the data distribution shapes the loss surface and identify subspaces of the model's data representation that are relatively stable over training."
                },
                {
                    "title": "Understanding deep learning (still) requires rethinking generalization",
                    "abstract": "Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small gap between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family or to the regularization techniques used during training. Through extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. We interpret our experimental findings by comparison with traditional models. We supplement this republication with a new section at the end summarizing recent progresses in the field since the original version of this paper."
                },
                {
                    "title": "Extracting Training Data from Large Language Models",
                    "abstract": "It has become common to publish large (billion parameter) language models that have been trained on private datasets. This paper demonstrates that in such settings, an adversary can perform a training data extraction attack to recover individual training examples by querying the language model. \nWe demonstrate our attack on GPT-2, a language model trained on scrapes of the public Internet, and are able to extract hundreds of verbatim text sequences from the model's training data. These extracted examples include (public) personally identifiable information (names, phone numbers, and email addresses), IRC conversations, code, and 128-bit UUIDs. Our attack is possible even though each of the above sequences are included in just one document in the training data. \nWe comprehensively evaluate our extraction attack to understand the factors that contribute to its success. For example, we find that larger models are more vulnerable than smaller models. We conclude by drawing lessons and discussing possible safeguards for training large language models."
                },
                {
                    "title": "Learning From Noisy Labels With Deep Neural Networks: A Survey",
                    "abstract": "Deep learning has achieved remarkable success in numerous domains with help from large amounts of big data. However, the quality of data labels is a concern because of the lack of high-quality labels in many real-world scenarios. As noisy labels severely degrade the generalization performance of deep neural networks, learning from noisy labels (robust training) is becoming an important task in modern deep learning applications. In this survey, we first describe the problem of learning with label noise from a supervised learning perspective. Next, we provide a comprehensive review of 62 state-of-the-art robust training methods, all of which are categorized into five groups according to their methodological difference, followed by a systematic comparison of six properties used to evaluate their superiority. Subsequently, we perform an in-depth analysis of noise rate estimation and summarize the typically used evaluation methodology, including public noisy datasets and evaluation metrics. Finally, we present several promising research directions that can serve as a guideline for future studies."
                },
                {
                    "title": "Are we done with ImageNet?",
                    "abstract": "Yes, and no. We ask whether recent progress on the ImageNet classification benchmark continues to represent meaningful generalization, or whether the community has started to overfit to the idiosyncrasies of its labeling procedure. We therefore develop a significantly more robust procedure for collecting human annotations of the ImageNet validation set. Using these new labels, we reassess the accuracy of recently proposed ImageNet classifiers, and find their gains to be substantially smaller than those reported on the original labels. Furthermore, we find the original ImageNet labels to no longer be the best predictors of this independently-collected set, indicating that their usefulness in evaluating vision models may be nearing an end. Nevertheless, we find our annotation procedure to have largely remedied the errors in the original labels, reinforcing ImageNet as a powerful benchmark for future research in visual recognition."
                },
                {
                    "title": "Masking: A New Perspective of Noisy Supervision",
                    "abstract": "It is important to learn various types of classifiers given training data with noisy labels. Noisy labels, in the most popular noise model hitherto, are corrupted from ground-truth labels by an unknown noise transition matrix. Thus, by estimating this matrix, classifiers can escape from overfitting those noisy labels. However, such estimation is practically difficult, due to either the indirect nature of two-step approaches, or not big enough data to afford end-to-end approaches. In this paper, we propose a human-assisted approach called Masking that conveys human cognition of invalid class transitions and naturally speculates the structure of the noise transition matrix. To this end, we derive a structure-aware probabilistic model incorporating a structure prior, and solve the challenges from structure extraction and structure alignment. Thanks to Masking, we only estimate unmasked noise transition probabilities and the burden of estimation is tremendously reduced. We conduct extensive experiments on CIFAR-10 and CIFAR-100 with three noise structures as well as the industrial-level Clothing1M with agnostic noise structure, and the results show that Masking can improve the robustness of classifiers significantly."
                },
                {
                    "title": "Not All Samples Are Created Equal: Deep Learning with Importance Sampling",
                    "abstract": "Deep neural network training spends most of the computation on examples that are properly handled, and could be ignored. We propose to mitigate this phenomenon with a principled importance sampling scheme that focuses computation on \"informative\" examples, and reduces the variance of the stochastic gradients during training. Our contribution is twofold: first, we derive a tractable upper bound to the per-sample gradient norm, and second we derive an estimator of the variance reduction achieved with importance sampling, which enables us to switch it on when it will result in an actual speedup. The resulting scheme can be used by changing a few lines of code in a standard SGD procedure, and we demonstrate experimentally, on image classification, CNN fine-tuning, and RNN training, that for a fixed wall-clock time budget, it provides a reduction of the train losses of up to an order of magnitude and a relative improvement of test errors between 5% and 17%."
                },
                {
                    "title": "On the importance of single directions for generalization",
                    "abstract": "Despite their ability to memorize large datasets, deep neural networks often achieve good generalization performance. However, the differences between the learned solutions of networks which generalize and those which do not remain unclear. Additionally, the tuning properties of single directions (defined as the activation of a single unit or some linear combination of units in response to some input) have been highlighted, but their importance has not been evaluated. Here, we connect these lines of inquiry to demonstrate that a network's reliance on single directions is a good predictor of its generalization performance, across networks trained on datasets with different fractions of corrupted labels, across ensembles of networks trained on datasets with unmodified labels, across different hyperparameters, and over the course of training. While dropout only regularizes this quantity up to a point, batch normalization implicitly discourages single direction reliance, in part by decreasing the class selectivity of individual units. Finally, we find that class selectivity is a poor predictor of task importance, suggesting not only that networks which generalize well minimize their dependence on individual units by reducing their selectivity, but also that individually selective units may not be necessary for strong network performance."
                },
                {
                    "title": "Deep Learning From Noisy Image Labels With Quality Embedding",
                    "abstract": "There is an emerging trend to leverage noisy image datasets in many visual recognition tasks. However, the label noise among datasets severely degenerates the performance of deep learning approaches. Recently, one mainstream is to introduce the latent label to handle label noise, which has shown promising improvement in the network designs. Nevertheless, the mismatch between latent labels and noisy labels still affects the predictions in such methods. To address this issue, we propose a probabilistic model, which explicitly introduces an extra variable to represent the trustworthiness of noisy labels, termed as the quality variable. Our key idea is to identify the mismatch between the latent and noisy labels by embedding the quality variables into different subspaces, which effectively minimizes the influence of label noise. At the same time, reliable labels are still able to be applied for training. To instantiate the model, we further propose a contrastive-additive noise network (CAN), which consists of two important layers: 1) the contrastive layer that estimates the quality variable in the embedding space to reduce the influence of noisy labels and 2) the additive layer that aggregates the prior prediction and noisy labels as the posterior to train the classifier. Moreover, to tackle the challenges in optimization, we deduce an SGD algorithm with the reparameterization tricks, which makes our method scalable to big data. We validate the proposed method on a range of noisy image datasets. Comprehensive results have demonstrated that CAN outperforms the state-of-the-art deep learning approaches."
                },
                {
                    "title": "Exploring Generalization in Deep Learning",
                    "abstract": "With a goal of understanding what drives generalization in deep networks, we consider several recently suggested explanations, including norm-based control, sharpness and robustness. We study how these measures can ensure generalization, highlighting the importance of scale normalization, and making a connection between sharpness and PAC-Bayes theory. We then investigate how well the measures explain different observed phenomena."
                },
                {
                    "title": "Robust Loss Functions under Label Noise for Deep Neural Networks",
                    "abstract": "\n \n In many applications of classifier learning, training data suffers from label noise. Deep networks are learned using huge training data where the problem of noisy labels is particularly relevant. The current techniques proposed for learning deep networks under label noise focus on modifying the network architecture and on algorithms for estimating true labels from noisy labels. An alternate approach would be to look for loss functions that are inherently noise-tolerant. For binary classification there exist theoretical results on loss functions that are robust to label noise. In this paper, we provide some sufficient conditions on a loss function so that risk minimization under that loss function would be inherently tolerant to label noise for multiclass classification problems. These results generalize the existing results on noise-tolerant loss functions for binary classification. We study some of the widely used loss functions in deep networks and show that the loss function based on mean absolute value of error is inherently robust to label noise. Thus standard back propagation is enough to learn the true classifier even under label noise. Through experiments, we illustrate the robustness of risk minimization with such loss functions for learning neural networks.\n \n"
                },
                {
                    "title": "Understanding deep learning requires rethinking generalization",
                    "abstract": "Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family, or to the regularization techniques used during training. \nThrough extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization, and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. \nWe interpret our experimental findings by comparison with traditional models."
                },
                {
                    "title": "Learning from massive noisy labeled data for image classification",
                    "abstract": "Large-scale supervised datasets are crucial to train convolutional neural networks (CNNs) for various computer vision problems. However, obtaining a massive amount of well-labeled data is usually very expensive and time consuming. In this paper, we introduce a general framework to train CNNs with only a limited number of clean labels and millions of easily obtained noisy labels. We model the relationships between images, class labels and label noises with a probabilistic graphical model and further integrate it into an end-to-end deep learning system. To demonstrate the effectiveness of our approach, we collect a large-scale real-world clothing classification dataset with both noisy and clean labels. Experiments on this dataset indicate that our approach can better correct the noisy labels and improves the performance of trained CNNs."
                },
                {
                    "title": "Norm-Based Capacity Control in Neural Networks",
                    "abstract": "We investigate the capacity, convexity and characterization of a general family of norm-constrained feed-forward networks."
                },
                {
                    "title": "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift",
                    "abstract": "Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.82% top-5 test error, exceeding the accuracy of human raters."
                },
                {
                    "title": "Dropout Training as Adaptive Regularization",
                    "abstract": "Dropout and other feature noising schemes control overfitting by artificially corrupting the training data. For generalized linear models, dropout performs a form of adaptive regularization. Using this viewpoint, we show that the dropout regularizer is first-order equivalent to an L2 regularizer applied after scaling the features by an estimate of the inverse diagonal Fisher information matrix. We also establish a connection to AdaGrad, an online learning algorithm, and find that a close relative of AdaGrad operates by repeatedly solving linear dropout-regularized problems. By casting dropout as regularization, we develop a natural semi-supervised algorithm that uses unlabeled data to create a better adaptive regularizer. We apply this idea to document classification tasks, and show that it consistently boosts the performance of dropout training, improving on state-of-the-art results on the IMDB reviews dataset."
                },
                {
                    "title": "Improving neural networks by preventing co-adaptation of feature detectors",
                    "abstract": "When a large feedforward neural network is trained on a small training set, it typically performs poorly on held-out test data. This \"overfitting\" is greatly reduced by randomly omitting half of the feature detectors on each training case. This prevents complex co-adaptations in which a feature detector is only helpful in the context of several other specific feature detectors. Instead, each neuron learns to detect a feature that is generally helpful for producing the correct answer given the combinatorially large variety of internal contexts in which it must operate. Random \"dropout\" gives big improvements on many benchmark tasks and sets new records for speech and object recognition."
                },
                {
                    "title": "Web-crawling reliability",
                    "abstract": "In this article, I investigate the reliability, in the social science sense, of collecting informetric data about the World Wide Web by Web crawling. The investigation includes a critical examination of the practice of Web crawling and contrasts the results of content crawling with the results of link crawling. It is shown that Web crawling by search engines is intentionally biased and selective. I also report the results of a large-scale experimental simulation of Web crawling that illustrates the effects of different crawling policies on data collection. It is concluded that the reliability of Web crawling as a data collection technique is improved by fuller reporting of relevant crawling policies."
                },
                {
                    "title": "The Sample Complexity of Pattern Classification with Neural Networks: The Size of the Weights is More Important than the Size of the Network",
                    "abstract": "Sample complexity results from computational learning theory, when applied to neural network learning for pattern classification problems, suggest that for good generalization performance the number of training examples should grow at least linearly with the number of adjustable parameters in the network. Results in this paper show that if a large neural network is used for a pattern classification problem and the learning algorithm finds a network with small weights that has small squared error on the training patterns, then the generalization performance depends on the size of the weights rather than the number of weights. For example, consider a two-layer feedforward network of sigmoid units, in which the sum of the magnitudes of the weights associated with each unit is bounded by A and the input dimension is n. We show that the misclassification probability is no more than a certain error estimate (that is related to squared error on the training set) plus A/sup 3/ /spl radic/((log n)/m) (ignoring log A and log m factors), where m is the number of training patterns. This may explain the generalization performance of neural networks, particularly when the number of training examples is considerably smaller than the number of weights. It also supports heuristics (such as weight decay and early stopping) that attempt to keep the weights small during training. The proof techniques appear to be useful for the analysis of other pattern classifiers: when the input domain is a totally bounded metric space, we use the same approach to give upper bounds on misclassification probability for classifiers with decision boundaries that are far from the training examples."
                },
                {
                    "title": "Dropout: a simple way to prevent neural networks from overfitting",
                    "abstract": "Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units (along with their connections) from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different \"thinned\" networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights. This significantly reduces overfitting and gives major improvements over other regularization methods. We show that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets."
                }
            ],
            "categories": [
                "cs.LG",
                "cond-mat.dis-nn",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively disentangle generalization and memorization in large neural networks when training on datasets with noisy labels?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the fundamental challenge of overfitting to noisy data, which is prevalent in real-world applications. By understanding the dynamics between memorization and generalization, we can improve the robustness of machine learning models, leading to better performance in practical applications such as natural language processing and computer vision. This research could pave the way for more reliable models that can generalize well while minimizing privacy concerns associated with data memorization.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complex interplay between memorization and generalization, especially in the presence of label corruption. Naive approaches may fail because they do not account for the nuanced behavior of large models, which can simultaneously memorize and generalize. Technical obstacles include identifying the specific neurons responsible for memorization and understanding how different regularization techniques affect this balance. Theoretical complexities arise from the need to analyze the model's behavior in a mathematically rigorous way, particularly when dealing with corrupted labels.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often overlooked the intricate relationship between memorization and generalization, focusing instead on either aspect in isolation. Limitations in existing solutions include a lack of analytical frameworks to study these phenomena and insufficient empirical evidence to support theoretical claims. Barriers such as the difficulty in isolating the effects of different regularization techniques have also hindered progress. Our approach differs by utilizing simple, interpretable models and algorithmic datasets, allowing for a clearer understanding of the mechanisms at play.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves training two-layer Multilayer Perceptrons (MLPs) on modular arithmetic tasks with corrupted labels. We will analyze the network's performance using metrics such as accuracy on both corrupted and uncorrupted datasets, as well as the inverse participation ratio (IPR) to identify memorizing neurons. The expected outcomes include demonstrating that large networks can achieve high generalization despite memorizing corrupted examples and that regularization techniques can be effectively employed to enhance generalization while mitigating memorization."
            }
        },
        "author_data": {
            "2c232ed2-ff8d-4869-b005-d13440de3c92": {
                "pk": "2c232ed2-ff8d-4869-b005-d13440de3c92",
                "name": "Darshil Doshi",
                "collaborators": [
                    "Tianyu He",
                    "Andrey Gromov",
                    "A. Gromov",
                    "Aritra Das"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Theoretical Physics",
                    "Gaussian Processes"
                ],
                "publications": [
                    {
                        "title": "Grokking Modular Polynomials",
                        "abstract": "Neural networks readily learn a subset of the modular arithmetic tasks, while failing to generalize on the rest. This limitation remains unmoved by the choice of architecture and training strategies. On the other hand, an analytical solution for the weights of Multi-layer Perceptron (MLP) networks that generalize on the modular addition task is known in the literature. In this work, we (i) extend the class of analytical solutions to include modular multiplication as well as modular addition with many terms. Additionally, we show that real networks trained on these datasets learn similar solutions upon generalization (grokking). (ii) We combine these\"expert\"solutions to construct networks that generalize on arbitrary modular polynomials. (iii) We hypothesize a classification of modular polynomials into learnable and non-learnable via neural networks training; and provide experimental evidence supporting our claims."
                    },
                    {
                        "title": "Learning to grok: Emergence of in-context learning and skill composition in modular arithmetic tasks",
                        "abstract": "Large language models can solve tasks that were not present in the training set. This capability is believed to be due to in-context learning and skill composition. In this work, we study the emergence of in-context learning and skill composition in a collection of modular arithmetic tasks. Specifically, we consider a finite collection of linear modular functions $z = a \\, x + b \\, y \\;\\mathrm{mod}\\; p$ labeled by the vector $(a, b) \\in \\mathbb{Z}_p^2$. We use some of these tasks for pre-training and the rest for out-of-distribution testing. We empirically show that a GPT-style transformer exhibits a transition from in-distribution to out-of-distribution generalization as the number of pre-training tasks increases. We find that the smallest model capable of out-of-distribution generalization requires two transformer blocks, while for deeper models, the out-of-distribution generalization phase is \\emph{transient}, necessitating early stopping. Finally, we perform an interpretability study of the pre-trained models, revealing the highly structured representations in both phases; and discuss the learnt algorithm."
                    },
                    {
                        "title": "AutoInit: Automatic Initialization via Jacobian Tuning",
                        "abstract": "Good initialization is essential for training Deep Neural Networks (DNNs). Oftentimes such initialization is found through a trial and error approach, which has to be applied anew every time an architecture is substantially modified, or inherited from smaller size networks leading to sub-optimal initialization. In this work we introduce a new and cheap algorithm, that allows one to find a good initialization automatically, for general feed-forward DNNs. The algorithm utilizes the Jacobian between adjacent network blocks to tune the network hyperparameters to criticality. We solve the dynamics of the algorithm for fully connected networks with ReLU and derive conditions for its convergence. We then extend the discussion to more general architectures with BatchNorm and residual connections. Finally, we apply our method to ResMLP and VGG architectures, where the automatic one-shot initialization found by our method shows good performance on vision tasks."
                    },
                    {
                        "title": "Critical Initialization of Wide and Deep Neural Networks using Partial Jacobians: General Theory and Applications",
                        "abstract": "Deep neural networks are notorious for defying theoretical treatment. However, when the number of parameters in each layer tends to infinity, the network function is a Gaussian process (GP) and quantitatively predictive description is possible. Gaussian approximation allows one to formulate criteria for selecting hyperparameters, such as variances of weights and biases, as well as the learning rate. These criteria rely on the notion of criticality defined for deep neural networks. In this work we describe a new practical way to diagnose criticality. We introduce \\emph{partial Jacobians} of a network, defined as derivatives of preactivations in layer $l$ with respect to preactivations in layer $l_0\\leq l$. We derive recurrence relations for the norms of partial Jacobians and utilize these relations to analyze criticality of deep fully connected neural networks with LayerNorm and/or residual connections. We derive and implement a simple and cheap numerical test that allows one to select optimal initialization for a broad class of deep neural networks; containing fully connected, convolutional and normalization layers. Using these tools we show quantitatively that proper stacking of the LayerNorm (applied to preactivations) and residual connections leads to an architecture that is critical for any initialization. Finally, we apply our methods to analyze ResNet and MLP-Mixer architectures; demonstrating the everywhere-critical regime."
                    },
                    {
                        "title": "Critical initialization of wide and deep neural networks through partial Jacobians: general theory and applications to LayerNorm",
                        "abstract": "Deep neural networks are notorious for defying theoretical treatment. However, when the number of parameters in each layer tends to in\ufb01nity the network function is a Gaussian process (GP) and quantitatively predictive description is possible. Gaussian approximation allows to formulate criteria for selecting hyperparameters, such as variances of weights and biases, as well as the learning rate. These criteria rely on the notion of criticality de\ufb01ned for deep neural networks. In this work we describe a new way to diagnose (both theoretically and empirically) this criticality. To that end, we introduce partial Jacobians of a network, de\ufb01ned as derivatives of preactivations in layer l with respect to preactivations in layer l 0 < l . These quantities are particularly useful when the network architecture involves many different layers. We discuss various properties of the partial Jacobians such as their scaling with depth and relation to the neural tangent kernel (NTK). We derive the recurrence relations for the partial Jacobians and utilize them to analyze criticality of deep MLP networks with (and without) LayerNorm. We \ufb01nd that the normalization layer changes the optimal values of hyperparameters and critical exponents. We argue that LayerNorm is more stable when applied to preactivations, rather than activations due to larger correlation depth."
                    },
                    {
                        "title": "Vortices and Fractons",
                        "abstract": "We discuss a simple and experimentally available realization of fracton physics. We note that superfluid vortices form a Hamiltonian system that conserves total dipole moment and trace of the quadrupole moment of vorticity; thereby establishing a relation to a traceless scalar charge theory in two spatial dimensions. Next we consider the limit where the number of vortices is large and show that emergent vortex hydrodynamics also conserves these moments. Finally, we show the motion of vortices and of fractons on curved surfaces agree, thereby opening a route to experimental study of the interplay between fracton physics and curved space. Our conclusions also apply to charged particles in strong magnetic field."
                    }
                ]
            },
            "ce4d3d15-a8b4-483c-9630-4ea8956de6f4": {
                "pk": "ce4d3d15-a8b4-483c-9630-4ea8956de6f4",
                "name": "Aritra Das",
                "collaborators": [
                    "Darshil Doshi",
                    "Tianyu He",
                    "Andrey Gromov"
                ],
                "domain": [
                    "Neural Networks",
                    "In-Context Learning",
                    "Modular Arithmetic",
                    "Generalization"
                ],
                "publications": [
                    {
                        "title": "Grokking Modular Polynomials",
                        "abstract": "Neural networks readily learn a subset of the modular arithmetic tasks, while failing to generalize on the rest. This limitation remains unmoved by the choice of architecture and training strategies. On the other hand, an analytical solution for the weights of Multi-layer Perceptron (MLP) networks that generalize on the modular addition task is known in the literature. In this work, we (i) extend the class of analytical solutions to include modular multiplication as well as modular addition with many terms. Additionally, we show that real networks trained on these datasets learn similar solutions upon generalization (grokking). (ii) We combine these\"expert\"solutions to construct networks that generalize on arbitrary modular polynomials. (iii) We hypothesize a classification of modular polynomials into learnable and non-learnable via neural networks training; and provide experimental evidence supporting our claims."
                    },
                    {
                        "title": "Learning to grok: Emergence of in-context learning and skill composition in modular arithmetic tasks",
                        "abstract": "Large language models can solve tasks that were not present in the training set. This capability is believed to be due to in-context learning and skill composition. In this work, we study the emergence of in-context learning and skill composition in a collection of modular arithmetic tasks. Specifically, we consider a finite collection of linear modular functions $z = a \\, x + b \\, y \\;\\mathrm{mod}\\; p$ labeled by the vector $(a, b) \\in \\mathbb{Z}_p^2$. We use some of these tasks for pre-training and the rest for out-of-distribution testing. We empirically show that a GPT-style transformer exhibits a transition from in-distribution to out-of-distribution generalization as the number of pre-training tasks increases. We find that the smallest model capable of out-of-distribution generalization requires two transformer blocks, while for deeper models, the out-of-distribution generalization phase is \\emph{transient}, necessitating early stopping. Finally, we perform an interpretability study of the pre-trained models, revealing the highly structured representations in both phases; and discuss the learnt algorithm."
                    }
                ]
            },
            "b82515fa-bb25-4ac3-9eaa-07522db9a2b4": {
                "pk": "b82515fa-bb25-4ac3-9eaa-07522db9a2b4",
                "name": "Tianyu He",
                "collaborators": [
                    "Darshil Doshi",
                    "Andrey Gromov",
                    "Aritra Das"
                ],
                "domain": [
                    "Neural Networks",
                    "In-Context Learning",
                    "Gaussian Processes",
                    "Modular Arithmetic"
                ],
                "publications": [
                    {
                        "title": "Grokking Modular Polynomials",
                        "abstract": "Neural networks readily learn a subset of the modular arithmetic tasks, while failing to generalize on the rest. This limitation remains unmoved by the choice of architecture and training strategies. On the other hand, an analytical solution for the weights of Multi-layer Perceptron (MLP) networks that generalize on the modular addition task is known in the literature. In this work, we (i) extend the class of analytical solutions to include modular multiplication as well as modular addition with many terms. Additionally, we show that real networks trained on these datasets learn similar solutions upon generalization (grokking). (ii) We combine these\"expert\"solutions to construct networks that generalize on arbitrary modular polynomials. (iii) We hypothesize a classification of modular polynomials into learnable and non-learnable via neural networks training; and provide experimental evidence supporting our claims."
                    },
                    {
                        "title": "Learning to grok: Emergence of in-context learning and skill composition in modular arithmetic tasks",
                        "abstract": "Large language models can solve tasks that were not present in the training set. This capability is believed to be due to in-context learning and skill composition. In this work, we study the emergence of in-context learning and skill composition in a collection of modular arithmetic tasks. Specifically, we consider a finite collection of linear modular functions $z = a \\, x + b \\, y \\;\\mathrm{mod}\\; p$ labeled by the vector $(a, b) \\in \\mathbb{Z}_p^2$. We use some of these tasks for pre-training and the rest for out-of-distribution testing. We empirically show that a GPT-style transformer exhibits a transition from in-distribution to out-of-distribution generalization as the number of pre-training tasks increases. We find that the smallest model capable of out-of-distribution generalization requires two transformer blocks, while for deeper models, the out-of-distribution generalization phase is \\emph{transient}, necessitating early stopping. Finally, we perform an interpretability study of the pre-trained models, revealing the highly structured representations in both phases; and discuss the learnt algorithm."
                    },
                    {
                        "title": "Critical initialization of wide and deep neural networks through partial Jacobians: general theory and applications to LayerNorm",
                        "abstract": "Deep neural networks are notorious for defying theoretical treatment. However, when the number of parameters in each layer tends to in\ufb01nity the network function is a Gaussian process (GP) and quantitatively predictive description is possible. Gaussian approximation allows to formulate criteria for selecting hyperparameters, such as variances of weights and biases, as well as the learning rate. These criteria rely on the notion of criticality de\ufb01ned for deep neural networks. In this work we describe a new way to diagnose (both theoretically and empirically) this criticality. To that end, we introduce partial Jacobians of a network, de\ufb01ned as derivatives of preactivations in layer l with respect to preactivations in layer l 0 < l . These quantities are particularly useful when the network architecture involves many different layers. We discuss various properties of the partial Jacobians such as their scaling with depth and relation to the neural tangent kernel (NTK). We derive the recurrence relations for the partial Jacobians and utilize them to analyze criticality of deep MLP networks with (and without) LayerNorm. We \ufb01nd that the normalization layer changes the optimal values of hyperparameters and critical exponents. We argue that LayerNorm is more stable when applied to preactivations, rather than activations due to larger correlation depth."
                    }
                ]
            },
            "ffa02457-c1aa-48cc-aad6-264a4e5e14b3": {
                "pk": "ffa02457-c1aa-48cc-aad6-264a4e5e14b3",
                "name": "Andrey Gromov",
                "collaborators": [
                    "Darshil Doshi",
                    "Tianyu He",
                    "Aritra Das"
                ],
                "domain": [
                    "Neural Networks",
                    "In-context Learning",
                    "Gaussian Processes",
                    "Modular Arithmetic"
                ],
                "publications": [
                    {
                        "title": "Grokking Modular Polynomials",
                        "abstract": "Neural networks readily learn a subset of the modular arithmetic tasks, while failing to generalize on the rest. This limitation remains unmoved by the choice of architecture and training strategies. On the other hand, an analytical solution for the weights of Multi-layer Perceptron (MLP) networks that generalize on the modular addition task is known in the literature. In this work, we (i) extend the class of analytical solutions to include modular multiplication as well as modular addition with many terms. Additionally, we show that real networks trained on these datasets learn similar solutions upon generalization (grokking). (ii) We combine these\"expert\"solutions to construct networks that generalize on arbitrary modular polynomials. (iii) We hypothesize a classification of modular polynomials into learnable and non-learnable via neural networks training; and provide experimental evidence supporting our claims."
                    },
                    {
                        "title": "Learning to grok: Emergence of in-context learning and skill composition in modular arithmetic tasks",
                        "abstract": "Large language models can solve tasks that were not present in the training set. This capability is believed to be due to in-context learning and skill composition. In this work, we study the emergence of in-context learning and skill composition in a collection of modular arithmetic tasks. Specifically, we consider a finite collection of linear modular functions $z = a \\, x + b \\, y \\;\\mathrm{mod}\\; p$ labeled by the vector $(a, b) \\in \\mathbb{Z}_p^2$. We use some of these tasks for pre-training and the rest for out-of-distribution testing. We empirically show that a GPT-style transformer exhibits a transition from in-distribution to out-of-distribution generalization as the number of pre-training tasks increases. We find that the smallest model capable of out-of-distribution generalization requires two transformer blocks, while for deeper models, the out-of-distribution generalization phase is \\emph{transient}, necessitating early stopping. Finally, we perform an interpretability study of the pre-trained models, revealing the highly structured representations in both phases; and discuss the learnt algorithm."
                    },
                    {
                        "title": "Critical initialization of wide and deep neural networks through partial Jacobians: general theory and applications to LayerNorm",
                        "abstract": "Deep neural networks are notorious for defying theoretical treatment. However, when the number of parameters in each layer tends to in\ufb01nity the network function is a Gaussian process (GP) and quantitatively predictive description is possible. Gaussian approximation allows to formulate criteria for selecting hyperparameters, such as variances of weights and biases, as well as the learning rate. These criteria rely on the notion of criticality de\ufb01ned for deep neural networks. In this work we describe a new way to diagnose (both theoretically and empirically) this criticality. To that end, we introduce partial Jacobians of a network, de\ufb01ned as derivatives of preactivations in layer l with respect to preactivations in layer l 0 < l . These quantities are particularly useful when the network architecture involves many different layers. We discuss various properties of the partial Jacobians such as their scaling with depth and relation to the neural tangent kernel (NTK). We derive the recurrence relations for the partial Jacobians and utilize them to analyze criticality of deep MLP networks with (and without) LayerNorm. We \ufb01nd that the normalization layer changes the optimal values of hyperparameters and critical exponents. We argue that LayerNorm is more stable when applied to preactivations, rather than activations due to larger correlation depth."
                    }
                ]
            }
        }
    },
    "2402.18815": {
        "paper_data": {
            "title": "How do Large Language Models Handle Multilingualism?",
            "url": "http://arxiv.org/abs/2402.18815v2",
            "arxiv_id": "2402.18815",
            "authors": [
                "Yiran Zhao",
                "Wenxuan Zhang",
                "Guizhen Chen",
                "Kenji Kawaguchi",
                "Lidong Bing"
            ],
            "abstract": "Large language models (LLMs) have demonstrated impressive capabilities across diverse languages. This study explores how LLMs handle multilingualism. Based on observed language ratio shifts among layers and the relationships between network structures and certain capabilities, we hypothesize the LLM's multilingual workflow ($\\texttt{MWork}$): LLMs initially understand the query, converting multilingual inputs into English for task-solving. In the intermediate layers, they employ English for thinking and incorporate multilingual knowledge with self-attention and feed-forward structures, respectively. In the final layers, LLMs generate responses aligned with the original language of the query. To verify $\\texttt{MWork}$, we introduce Parallel Language-specific Neuron Detection ($\\texttt{PLND}$) to identify activated neurons for inputs in different languages without any labeled data. Using $\\texttt{PLND}$, we validate $\\texttt{MWork}$ through extensive experiments involving the deactivation of language-specific neurons across various layers and structures. Moreover, $\\texttt{MWork}$ allows fine-tuning of language-specific neurons with a small dataset, enhancing multilingual abilities in a specific language without compromising others. This approach results in an average improvement of $3.6\\%$ for high-resource languages and $2.3\\%$ for low-resource languages across all tasks with just $400$ documents.",
            "introduction": "   1 Introduction  Recent advancements in large language models (LLMs) \u00a0(OpenAI, 2023; Touvron et\u00a0al., 2023; Team et\u00a0al., 2023) have dramatically transformed the field of natural language processing (NLP). Thanks to the extensive pretraining on massive corpora mixed with different languages, these models demonstrate remarkable capabilities in understanding and generating text across multiple languages\u00a0(Huang et\u00a0al., 2023; Zhu et\u00a0al., 2023; Zhang et\u00a0al., 2023a; Zhao et\u00a0al., 2024a). Despite these advancements, the intricate mechanism of their multilingual processing behavior remains largely unclear, which leads to an important research question: How do large language models handle multilingualism?   To understand the working mechanism of LLMs, existing studies mainly focus on the relationship between model architectures and certain capabilities, with some investigating reasoning abilities with self-attention layers\u00a0(Hou et\u00a0al., 2023; Stolfo et\u00a0al., 2023; Friedman et\u00a0al., 2023), and others interpreting feed-forward layers as key-value memories for storing factual knowledge\u00a0(Geva et\u00a0al., 2021; Dai et\u00a0al., 2022a; Meng et\u00a0al., 2022). However, these works solely center on English and neglect the multilingual features of LLMs in their interpretations.      (a) Vicuna-13b-v1.5     (b) BLOOMZ-7b1    Figure 1: Ratio of English and non-English tokens among layers given non-English queries.   To gain an initial understanding of the multilingual mechanism of LLMs, we test LLMs with various non-English queries and decode the hidden embeddings of each layer to tokens within the LLM\u2019s vocabulary. Subsequently, we classify these decoded tokens into either English or non-English, and analyze the ratio. Figure\u00a01 illustrates the ratio of English and non-English tokens for each layer of two LLMs. We observe that non-English queries initially generate non-English embeddings as expected. However, as queries progress through the middle layers, the representations become English-centric. In the final layers, there is a reversion to predominantly non-English embeddings, matching the non-English queries.   Figure 2: Our hypothesized multilingual workflow MWork.   Motivated by the observed transformation above, we hypothesize a three-stage multilingual workflow: understanding, task-solving, and generating. This involves understanding the original non-English queries and interpreting them in English, solving tasks in English, and reverting outputs back to the original language. Furthermore, building upon previous studies that link self-attention structures to reasoning and feed-forward structures to factual knowledge storage (Hou et\u00a0al., 2023; Geva et\u00a0al., 2021), we further decouple the task-solving stage into reasoning with self-attention structures and extracting multilingual knowledge with feed-forward structures. Therefore, our hypothesized Multilingual Workflow (MWork) illustrated in Figure 2 outlines the three operational stages of LLMs in processing multilingual queries: Initially, LLMs understand queries by converting diverse linguistic features into a unified representation. In the task-solving phase, LLMs think in English and incorporate multilingual knowledge to obtain factual content, using self-attention and feed-forward structures, respectively. Finally, models generate responses in the original language as the original query.   To verify the proposed MWork workflow, we extract language-specific parameters and selectively deactivate them within different structures, thereby assessing the functionality of corresponding structures and validating our hypothesis. To identify the parameters to be activated, we develop a novel approach called Parallel Language-specific Neuron Detection (PLND). Unlike existing methods that rely on fine-tuning(Frankle and Carbin, 2018; Aghajanyan et\u00a0al., 2021; Zhang et\u00a0al., 2023b), labeled data (Tang et\u00a0al., 2024; Liu et\u00a0al., 2024), or parallel corpora (Libovick\u1ef3 et\u00a0al., 2020; Tanti et\u00a0al., 2021; Zhang et\u00a0al., 2024) to detect activated parameters, PLND measures the",
            "references": [
                {
                    "title": "AdaMergeX: Cross-Lingual Transfer with Large Language Models via Adaptive Adapter Merging",
                    "abstract": "As an effective alternative to the direct fine-tuning on target tasks in specific languages, cross-lingual transfer addresses the challenges of limited training data by decoupling ''task ability'' and ''language ability'' by fine-tuning on the target task in the source language and another selected task in the target language, respectively. However, they fail to fully separate the task ability from the source language or the language ability from the chosen task. In this paper, we acknowledge the mutual reliance between task ability and language ability and direct our attention toward the gap between the target language and the source language on tasks. As the gap removes the impact of tasks, we assume that it remains consistent across tasks. Based on this assumption, we propose a new cross-lingual transfer method called $\\texttt{AdaMergeX}$ that utilizes adaptive adapter merging. By introducing a reference task, we can determine that the divergence of adapters fine-tuned on the reference task in both languages follows the same distribution as the divergence of adapters fine-tuned on the target task in both languages. Hence, we can obtain target adapters by combining the other three adapters. Furthermore, we propose a structure-adaptive adapter merging method. Our empirical results demonstrate that our approach yields new and effective cross-lingual transfer, outperforming existing methods across all settings."
                },
                {
                    "title": "Language-Specific Neurons: The Key to Multilingual Capabilities in Large Language Models",
                    "abstract": "Large language models (LLMs) demonstrate remarkable multilingual capabilities without being pre-trained on specially curated multilingual parallel corpora. It remains a challenging problem to explain the underlying mechanisms by which LLMs process multilingual texts. In this paper, we delve into the composition of Transformer architectures in LLMs to pinpoint language-specific regions. Specially, we propose a novel detection method, language activation probability entropy (LAPE), to identify language-specific neurons within LLMs. Based on LAPE, we conduct comprehensive experiments on several representative LLMs, such as LLaMA-2, BLOOM, and Mistral. Our findings indicate that LLMs' proficiency in processing a particular language is predominantly due to a small subset of neurons, primarily situated in the models' top and bottom layers. Furthermore, we showcase the feasibility to\"steer\"the output language of LLMs by selectively activating or deactivating language-specific neurons. Our research provides important evidence to the understanding and exploration of the multilingual capabilities of LLMs."
                },
                {
                    "title": "LLaMA Beyond English: An Empirical Study on Language Capability Transfer",
                    "abstract": "In recent times, substantial advancements have been witnessed in large language models (LLMs), exemplified by ChatGPT, showcasing remarkable proficiency across a range of complex tasks. However, many mainstream LLMs (e.g. LLaMA) are pretrained on English-dominant corpus, which limits their performance in other non-English languages. In this paper, we focus on how to effectively transfer the capabilities of language generation and following instructions to a non-English language. To answer this question, we conduct an extensive empirical investigation based on LLaMA, accumulating over 1440 GPU hours. We analyze the impact of key factors such as vocabulary extension, further pretraining, and instruction tuning on transfer. To accurately assess the model's level of knowledge, we employ four widely used standardized testing benchmarks: C-Eval, MMLU, AGI-Eval, and GAOKAO-Bench. Furthermore, a comprehensive evaluation of the model's response quality is conducted, considering aspects such as accuracy, fluency, informativeness, logical coherence, and harmlessness, based on LLM-Eval, a benchmarks consisting instruction tasks from 17 diverse categories. Our evaluation results demonstrate that comparable performance to state-of-the-art transfer models can be achieved with less than 1% of the pretraining data, both in terms of knowledge alignment and response quality. Furthermore, the experimental outcomes across the thirteen low-resource languages also exhibit similar trends. We anticipate that the conclusions revealed by the experiments will aid the community in developing non-English LLMs."
                },
                {
                    "title": "The Truth is in There: Improving Reasoning in Language Models with Layer-Selective Rank Reduction",
                    "abstract": "Transformer-based Large Language Models (LLMs) have become a fixture in modern machine learning. Correspondingly, significant resources are allocated towards research that aims to further advance this technology, typically resulting in models of increasing size that are trained on increasing amounts of data. This work, however, demonstrates the surprising result that it is often possible to significantly improve the performance of LLMs by selectively removing higher-order components of their weight matrices. This simple intervention, which we call LAyer-SElective Rank reduction (LASER), can be done on a model after training has completed, and requires no additional parameters or data. We show extensive experiments demonstrating the generality of this finding across language models and datasets, and provide in-depth analyses offering insights into both when LASER is effective and the mechanism by which it operates."
                },
                {
                    "title": "Interpretability Illusions in the Generalization of Simplified Models",
                    "abstract": "A common method to study deep learning systems is to use simplified model representations--for example, using singular value decomposition to visualize the model's hidden states in a lower dimensional space. This approach assumes that the results of these simplifications are faithful to the original model. Here, we illustrate an important caveat to this assumption: even if the simplified representations can accurately approximate the full model on the training set, they may fail to accurately capture the model's behavior out of distribution. We illustrate this by training Transformer models on controlled datasets with systematic generalization splits, including the Dyck balanced-parenthesis languages and a code completion task. We simplify these models using tools like dimensionality reduction and clustering, and then explicitly test how these simplified proxies match the behavior of the original model. We find consistent generalization gaps: cases in which the simplified proxies are more faithful to the original model on the in-distribution evaluations and less faithful on various tests of systematic generalization. This includes cases where the original model generalizes systematically but the simplified proxies fail, and cases where the simplified proxies generalize better. Together, our results raise questions about the extent to which mechanistic interpretations derived using tools like SVD can reliably predict what a model will do in novel situations."
                },
                {
                    "title": "SeaLLMs - Large Language Models for Southeast Asia",
                    "abstract": "Despite the remarkable achievements of large language models (LLMs) in various tasks, there remains a linguistic bias that favors high-resource languages, such as English, often at the expense of low-resource and regional languages. To address this imbalance, we introduce SeaLLMs, an innovative series of language models that specifically focuses on Southeast Asian (SEA) languages. SeaLLMs are built upon the Llama-2 model and further advanced through continued pre-training with an extended vocabulary, specialized instruction and alignment tuning to better capture the intricacies of regional languages. This allows them to respect and reflect local cultural norms, customs, stylistic preferences, and legal considerations. Our comprehensive evaluation demonstrates that SeaLLM-13b models exhibit superior performance across a wide spectrum of linguistic tasks and assistant-style instruction-following capabilities relative to comparable open-source models. Moreover, they outperform ChatGPT-3.5 in non-Latin languages, such as Thai, Khmer, Lao, and Burmese, by large margins while remaining lightweight and cost-effective to operate."
                },
                {
                    "title": "Towards a Mechanistic Interpretation of Multi-Step Reasoning Capabilities of Language Models",
                    "abstract": "Recent work has shown that language models (LMs) have strong multi-step (i.e., procedural) reasoning capabilities. However, it is unclear whether LMs perform these tasks by cheating with answers memorized from pretraining corpus, or, via a multi-step reasoning mechanism. In this paper, we try to answer this question by exploring a mechanistic interpretation of LMs for multi-step reasoning tasks. Concretely, we hypothesize that the LM implicitly embeds a reasoning tree resembling the correct reasoning process within it. We test this hypothesis by introducing a new probing approach (called MechanisticProbe) that recovers the reasoning tree from the model's attention patterns. We use our probe to analyze two LMs: GPT-2 on a synthetic task (k-th smallest element), and LLaMA on two simple language-based reasoning tasks (ProofWriter&AI2 Reasoning Challenge). We show that MechanisticProbe is able to detect the information of the reasoning tree from the model's attentions for most examples, suggesting that the LM indeed is going through a process of multi-step reasoning within its architecture in many cases."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Enhancing Cross-lingual Transfer via Phonemic Transcription Integration",
                    "abstract": "Previous cross-lingual transfer methods are restricted to orthographic representation learning via textual scripts. This limitation hampers cross-lingual transfer and is biased towards languages sharing similar well-known scripts. To alleviate the gap between languages from different writing scripts, we propose PhoneXL, a framework incorporating phonemic transcriptions as an additional linguistic modality beyond the traditional orthographic transcriptions for cross-lingual transfer. Particularly, we propose unsupervised alignment objectives to capture (1) local one-to-one alignment between the two different modalities, (2) alignment via multi-modality contexts to leverage information from additional modalities, and (3) alignment via multilingual contexts where additional bilingual dictionaries are incorporated. We also release the first phonemic-orthographic alignment dataset on two token-level tasks (Named Entity Recognition and Part-of-Speech Tagging) among the understudied but interconnected Chinese-Japanese-Korean-Vietnamese (CJKV) languages. Our pilot study reveals phonemic transcription provides essential information beyond the orthography to enhance cross-lingual transfer and bridge the gap among CJKV languages, leading to consistent improvements on cross-lingual token-level tasks over orthographic-based multilingual PLMs."
                },
                {
                    "title": "M3Exam: A Multilingual, Multimodal, Multilevel Benchmark for Examining Large Language Models",
                    "abstract": "Despite the existence of various benchmarks for evaluating natural language processing models, we argue that human exams are a more suitable means of evaluating general intelligence for large language models (LLMs), as they inherently demand a much wider range of abilities such as language understanding, domain knowledge, and problem-solving skills. To this end, we introduce M3Exam, a novel benchmark sourced from real and official human exam questions for evaluating LLMs in a multilingual, multimodal, and multilevel context. M3Exam exhibits three unique characteristics: (1) multilingualism, encompassing questions from multiple countries that require strong multilingual proficiency and cultural knowledge; (2) multimodality, accounting for the multimodal nature of many exam questions to test the model's multimodal understanding capability; and (3) multilevel structure, featuring exams from three critical educational periods to comprehensively assess a model's proficiency at different levels. In total, M3Exam contains 12,317 questions in 9 diverse languages with three educational levels, where about 23\\% of the questions require processing images for successful solving. We assess the performance of top-performing LLMs on M3Exam and find that current models, including GPT-4, still struggle with multilingual text, particularly in low-resource and non-Latin script languages. Multimodal LLMs also perform poorly with complex multimodal questions. We believe that M3Exam can be a valuable resource for comprehensively evaluating LLMs by examining their multilingual and multimodal abilities and tracking their development. Data and evaluation code is available at \\url{https://github.com/DAMO-NLP-SG/M3Exam}."
                },
                {
                    "title": "Inference-Time Intervention: Eliciting Truthful Answers from a Language Model",
                    "abstract": "We introduce Inference-Time Intervention (ITI), a technique designed to enhance the\"truthfulness\"of large language models (LLMs). ITI operates by shifting model activations during inference, following a set of directions across a limited number of attention heads. This intervention significantly improves the performance of LLaMA models on the TruthfulQA benchmark. On an instruction-finetuned LLaMA called Alpaca, ITI improves its truthfulness from 32.5% to 65.1%. We identify a tradeoff between truthfulness and helpfulness and demonstrate how to balance it by tuning the intervention strength. ITI is minimally invasive and computationally inexpensive. Moreover, the technique is data efficient: while approaches like RLHF require extensive annotations, ITI locates truthful directions using only few hundred examples. Our findings suggest that LLMs may have an internal representation of the likelihood of something being true, even as they produce falsehoods on the surface."
                },
                {
                    "title": "Machine-Created Universal Language for Cross-lingual Transfer",
                    "abstract": "There are two primary approaches to addressing cross-lingual transfer: multilingual pre-training, which implicitly aligns the hidden representations of various languages, and translate-test, which explicitly translates different languages into an intermediate language, such as English. Translate-test offers better interpretability compared to multilingual pre-training. However, it has lower performance than multilingual pre-training and struggles with word-level tasks due to translation altering word order. As a result, we propose a new Machine-created Universal Language (MUL) as an alternative intermediate language. MUL comprises a set of discrete symbols forming a universal vocabulary and a natural language to MUL translator for converting multiple natural languages to MUL. MUL unifies shared concepts from various languages into a single universal word, enhancing cross-language transfer. Additionally, MUL retains language-specific words and word order, allowing the model to be easily applied to word-level tasks. Our experiments demonstrate that translating into MUL yields improved performance compared to multilingual pre-training, and our analysis indicates that MUL possesses strong interpretability. The code is at: https://github.com/microsoft/Unicoder/tree/master/MCUL."
                },
                {
                    "title": "Enhancing Cross-lingual Natural Language Inference by Soft Prompting with Multilingual Verbalizer",
                    "abstract": "Cross-lingual natural language inference is a fundamental problem in cross-lingual language understanding. Many recent works have used prompt learning to address the lack of annotated parallel corpora in XNLI. However, these methods adopt discrete prompting by simply translating the templates to the target language and need external expert knowledge to design the templates. Besides, discrete prompts of human-designed template words are not trainable vectors and can not be migrated to target languages in the inference stage flexibly. In this paper, we propose a novel Soft prompt learning framework with the Multilingual Verbalizer (SoftMV) for XNLI. SoftMV first constructs cloze-style question with soft prompts for the input sample. Then we leverage bilingual dictionaries to generate an augmented multilingual question for the original question. SoftMV adopts a multilingual verbalizer to align the representations of original and augmented multilingual questions into the same semantic space with consistency regularization. Experimental results on XNLI demonstrate that SoftMV can achieve state-of-the-art performance and significantly outperform the previous methods under the few-shot and full-shot cross-lingual transfer settings."
                },
                {
                    "title": "Not All Languages Are Created Equal in LLMs: Improving Multilingual Capability by Cross-Lingual-Thought Prompting",
                    "abstract": "Large language models (LLMs) demonstrate impressive multilingual capability, but their performance varies substantially across different languages. In this work, we introduce a simple yet effective method, called cross-lingual-thought prompting (XLT), to systematically improve the multilingual capability of LLMs. Specifically, XLT is a generic template prompt that stimulates cross-lingual and logical reasoning skills to enhance task performance across languages. We conduct comprehensive evaluations on 7 typical benchmarks related to reasoning, understanding, and generation tasks, covering both high-resource and low-resource languages. Experimental results show that XLT not only remarkably enhances the performance of various multilingual tasks but also significantly reduces the gap between the average performance and the best performance of each task in different languages. Notably, XLT brings over 10 points of average improvement in arithmetic reasoning and open-domain question-answering tasks."
                },
                {
                    "title": "Multilingual LLMs are Better Cross-lingual In-context Learners with Alignment",
                    "abstract": "In-context learning (ICL) unfolds as large language models become capable of inferring test labels conditioned on a few labeled samples without any gradient update. ICL-enabled large language models provide a promising step forward toward bypassing recurrent annotation costs in a low-resource setting. Yet, only a handful of past studies have explored ICL in a cross-lingual setting, in which the need for transferring label-knowledge from a high-resource language to a low-resource one is immensely crucial. To bridge the gap, we provide the first in-depth analysis of ICL for cross-lingual text classification. We find that the prevalent mode of selecting random input-label pairs to construct the prompt-context is severely limited in the case of cross-lingual ICL, primarily due to the lack of alignment in the input as well as the output spaces. To mitigate this, we propose a novel prompt construction strategy \u2014 Cross-lingual In-context Source Target Alignment (X-InSTA). With an injected coherence in the semantics of the input examples and a task-based alignment across the source and target languages, X-InSTA is able to outperform random prompt selection by a large margin across three different tasks using 44 different cross-lingual pairs."
                },
                {
                    "title": "Efficient and Effective Text Encoding for Chinese LLaMA and Alpaca",
                    "abstract": "Large Language Models (LLMs), such as ChatGPT and GPT-4, have dramatically transformed natural language processing research and shown promising strides towards Artificial General Intelligence (AGI). Nonetheless, the high costs associated with training and deploying LLMs present substantial obstacles to transparent, accessible academic research. While several large language models, such as LLaMA, have been open-sourced by the community, these predominantly focus on English corpora, limiting their usefulness for other languages. In this paper, we propose a method to augment LLaMA with capabilities for understanding and generating Chinese text and its ability to follow instructions. We achieve this by extending LLaMA's existing vocabulary with an additional 20,000 Chinese tokens, thereby improving its encoding efficiency and semantic understanding of Chinese. We further incorporate secondary pre-training using Chinese data and fine-tune the model with Chinese instruction datasets, significantly enhancing the model's ability to comprehend and execute instructions. Our experimental results indicate that the newly proposed model markedly enhances the original LLaMA's proficiency in understanding and generating Chinese content. Additionally, the results on the C-Eval dataset yield competitive performance among the models with several times the size of ours. We have made our pre-trained models, training scripts, and other resources available through GitHub, fostering open research for our community. Chinese LLaMA series: \\url{https://github.com/ymcui/Chinese-LLaMA-Alpaca} and Chinese Llama-2 series: \\url{https://github.com/ymcui/Chinese-LLaMA-Alpaca-2}"
                },
                {
                    "title": "Language Models are Multilingual Chain-of-Thought Reasoners",
                    "abstract": "We evaluate the reasoning abilities of large language models in multilingual settings. We introduce the Multilingual Grade School Math (MGSM) benchmark, by manually translating 250 grade-school math problems from the GSM8K dataset (Cobbe et al., 2021) into ten typologically diverse languages. We find that the ability to solve MGSM problems via chain-of-thought prompting emerges with increasing model scale, and that models have strikingly strong multilingual reasoning abilities, even in underrepresented languages such as Bengali and Swahili. Finally, we show that the multilingual reasoning abilities of language models extend to other tasks such as commonsense reasoning and word-in-context semantic judgment. The MGSM benchmark is publicly available at https://github.com/google-research/url-nlp."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space",
                    "abstract": "Transformer-based language models (LMs) are at the core of modern NLP, but their internal prediction construction process is opaque and largely not understood. In this work, we make a substantial step towards unveiling this underlying prediction process, by reverse-engineering the operation of the feed-forward network (FFN) layers, one of the building blocks of transformer models. We view the token representation as a changing distribution over the vocabulary, and the output from each FFN layer as an additive update to that distribution. Then, we analyze the FFN updates in the vocabulary space, showing that each update can be decomposed to sub-updates corresponding to single FFN parameter vectors, each promoting concepts that are often human-interpretable. We then leverage these findings for controlling LM predictions, where we reduce the toxicity of GPT2 by almost 50%, and for improving computation efficiency with a simple early exit rule, saving 20% of computation on average."
                },
                {
                    "title": "Cross-Lingual Ability of Multilingual Masked Language Models: A Study of Language Structure",
                    "abstract": "Multilingual pre-trained language models, such as mBERT and XLM-R, have shown impressive cross-lingual ability. Surprisingly, both of them use multilingual masked language model (MLM) without any cross-lingual supervision or aligned data. Despite the encouraging results, we still lack a clear understanding of why cross-lingual ability could emerge from multilingual MLM. In our work, we argue that cross-language ability comes from the commonality between languages. Specifically, we study three language properties: constituent order, composition and word co-occurrence. First, we create an artificial language by modifying property in source language. Then we study the contribution of modified property through the change of cross-language transfer results on target language. We conduct experiments on six languages and two cross-lingual NLP tasks (textual entailment, sentence retrieval). Our main conclusion is that the contribution of constituent order and word co-occurrence is limited, while the composition is more crucial to the success of cross-linguistic transfer."
                },
                {
                    "title": "Locating and Editing Factual Associations in GPT",
                    "abstract": "We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's factual predictions. This reveals a distinct set of steps in middle-layer feed-forward modules that mediate factual predictions while processing subject tokens. To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update specific factual associations using Rank-One Model Editing (ROME). We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another. Our results confirm an important role for mid-layer feed-forward modules in storing factual associations and suggest that direct manipulation of computational mechanisms may be a feasible approach for model editing. The code, dataset, visualizations, and an interactive demo notebook are available at https://rome.baulab.info/"
                },
                {
                    "title": "On the Language-specificity of Multilingual BERT and the Impact of Fine-tuning",
                    "abstract": "Recent work has shown evidence that the knowledge acquired by multilingual BERT (mBERT) has two components: a language-specific and a language-neutral one. This paper analyses the relationship between them, in the context of fine-tuning on two tasks \u2013 POS tagging and natural language inference \u2013 which require the model to bring to bear different degrees of language-specific knowledge. Visualisations reveal that mBERT loses the ability to cluster representations by language after fine-tuning, a result that is supported by evidence from language identification experiments. However, further experiments on \u2018unlearning\u2019 language-specific representations using gradient reversal and iterative adversarial learning are shown not to add further improvement to the language-independent component over and above the effect of fine-tuning. The results presented here suggest that the process of fine-tuning causes a reorganisation of the model\u2019s limited representational capacity, enhancing language-independent representations at the expense of language-specific ones."
                },
                {
                    "title": "A Primer on Pretrained Multilingual Language Models",
                    "abstract": "Multilingual Language Models (\\MLLMs) such as mBERT, XLM, XLM-R, \\textit{etc.} have emerged as a viable option for bringing the power of pretraining to a large number of languages. Given their success in zero-shot transfer learning, there has emerged a large body of work in (i) building bigger \\MLLMs~covering a large number of languages (ii) creating exhaustive benchmarks covering a wider variety of tasks and languages for evaluating \\MLLMs~ (iii) analysing the performance of \\MLLMs~on monolingual, zero-shot cross-lingual and bilingual tasks (iv) understanding the universal language patterns (if any) learnt by \\MLLMs~ and (v) augmenting the (often) limited capacity of \\MLLMs~ to improve their performance on seen or even unseen languages. In this survey, we review the existing literature covering the above broad areas of research pertaining to \\MLLMs. Based on our survey, we recommend some promising directions of future research."
                },
                {
                    "title": "XL-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages",
                    "abstract": "Contemporary works on abstractive text summarization have focused primarily on high-resource languages like English, mostly due to the limited availability of datasets for low/mid-resource ones. In this work, we present XL-Sum, a comprehensive and diverse dataset comprising 1 million professionally annotated article-summary pairs from BBC, extracted using a set of carefully designed heuristics. The dataset covers 44 languages ranging from low to high-resource, for many of which no public dataset is currently available. XL-Sum is highly abstractive, concise, and of high quality, as indicated by human and intrinsic evaluation. We fine-tune mT5, a state-of-the-art pretrained multilingual model, with XL-Sum and experiment on multilingual and low-resource summarization tasks. XL-Sum induces competitive results compared to the ones obtained using similar monolingual datasets: we show higher than 11 ROUGE-2 scores on 10 languages we benchmark on, with some of them exceeding 15, as obtained by multilingual training. Additionally, training on low-resource languages individually also provides competitive performance. To the best of our knowledge, XL-Sum is the largest abstractive summarization dataset in terms of the number of samples collected from a single source and the number of languages covered. We are releasing our dataset and models to encourage future research on multilingual abstractive summarization. The resources can be found at \\url{https://github.com/csebuetnlp/xl-sum}."
                },
                {
                    "title": "Common Sense Beyond English: Evaluating and Improving Multilingual Language Models for Commonsense Reasoning",
                    "abstract": "Commonsense reasoning research has so far been limited to English. We aim to evaluate and improve popular multilingual language models (ML-LMs) to help advance commonsense reasoning (CSR) beyond English. We collect the Mickey corpus, consisting of 561k sentences in 11 different languages, which can be used for analyzing and improving ML-LMs. We propose Mickey Probe, a language-general probing task for fairly evaluating the common sense of popular ML-LMs across different languages. In addition, we also create two new datasets, X-CSQA and X-CODAH, by translating their English versions to 14 other languages, so that we can evaluate popular ML-LMs for cross-lingual commonsense reasoning. To improve the performance beyond English, we propose a simple yet effective method \u2014 multilingual contrastive pretraining (MCP). It significantly enhances sentence representations, yielding a large performance gain on both benchmarks (e.g., +2.7% accuracy for X-CSQA over XLM-R_L)."
                },
                {
                    "title": "Transformer Feed-Forward Layers Are Key-Value Memories",
                    "abstract": "Feed-forward layers constitute two-thirds of a transformer model\u2019s parameters, yet their role in the network remains under-explored. We show that feed-forward layers in transformer-based language models operate as key-value memories, where each key correlates with textual patterns in the training examples, and each value induces a distribution over the output vocabulary. Our experiments show that the learned patterns are human-interpretable, and that lower layers tend to capture shallow patterns, while upper layers learn more semantic ones. The values complement the keys\u2019 input patterns by inducing output distributions that concentrate probability mass on tokens likely to appear immediately after each pattern, particularly in the upper layers. Finally, we demonstrate that the output of a feed-forward layer is a composition of its memories, which is subsequently refined throughout the model\u2019s layers via residual connections to produce the final output distribution."
                },
                {
                    "title": "Intrinsic Dimensionality Explains the Effectiveness of Language Model Fine-Tuning",
                    "abstract": "Although pretrained language models can be fine-tuned to produce state-of-the-art results for a very wide range of language understanding tasks, the dynamics of this process are not well understood, especially in the low data regime. Why can we use relatively vanilla gradient descent algorithms (e.g., without strong regularization) to tune a model with hundreds of millions of parameters on datasets with only hundreds or thousands of labeled examples? In this paper, we argue that analyzing fine-tuning through the lens of intrinsic dimension provides us with empirical and theoretical intuitions to explain this remarkable phenomenon. We empirically show that common pre-trained models have a very low intrinsic dimension; in other words, there exists a low dimension reparameterization that is as effective for fine-tuning as the full parameter space. For example, by optimizing only 200 trainable parameters randomly projected back into the full space, we can tune a RoBERTa model to achieve 90% of the full parameter performance levels on MRPC. Furthermore, we empirically show that pre-training implicitly minimizes intrinsic dimension and, perhaps surprisingly, larger models tend to have lower intrinsic dimension after a fixed number of pre-training updates, at least in part explaining their extreme effectiveness. Lastly, we connect intrinsic dimensionality with low dimensional task representations and compression based generalization bounds to provide intrinsic-dimension-based generalization bounds that are independent of the full parameter count."
                },
                {
                    "title": "Language ID in the Wild: Unexpected Challenges on the Path to a Thousand-Language Web Text Corpus",
                    "abstract": "Large text corpora are increasingly important for a wide variety of Natural Language Processing (NLP) tasks, and automatic language identification (LangID) is a core technology needed to collect such datasets in a multilingual context. LangID is largely treated as solved in the literature, with models reported that achieve over 90% average F1 on as many as 1,366 languages. We train LangID models on up to 1,629 languages with comparable quality on held-out test sets, but find that human-judged LangID accuracy for web-crawl text corpora created using these models is only around 5% for many lower-resource languages, suggesting a need for more robust evaluation. Further analysis revealed a variety of error modes, arising from domain mismatch, class imbalance, language similarity, and insufficiently expressive models. We propose two classes of techniques to mitigate these errors: wordlist-based tunable-precision filters (for which we release curated lists in about 500 languages) and transformer-based semi-supervised LangID models, which increase median dataset precision from 5.5% to 71.2%. These techniques enable us to create an initial data set covering 100K or more relatively clean sentences in each of 500+ languages, paving the way towards a 1,000-language web text corpus."
                },
                {
                    "title": "MAD-X: An Adapter-based Framework for Multi-task Cross-lingual Transfer",
                    "abstract": "The main goal behind state-of-the-art pretrained multilingual models such as multilingual BERT and XLM-R is enabling and bootstrapping NLP applications in low-resource languages through zero-shot or few-shot cross-lingual transfer. However, due to limited model capacity, their transfer performance is the weakest exactly on such low-resource languages and languages unseen during pretraining. We propose MAD-X, an adapter-based framework that enables high portability and parameter-efficient transfer to arbitrary tasks and languages by learning modular language and task representations. In addition, we introduce a novel invertible adapter architecture and a strong baseline method for adapting a pretrained multilingual model to a new language. MAD-X outperforms the state of the art in cross-lingual transfer across a representative set of typologically diverse languages on named entity recognition and achieves competitive results on question answering."
                },
                {
                    "title": "On the Language Neutrality of Pre-trained Multilingual Representations",
                    "abstract": "Multilingual contextual embeddings, such as multilingual BERT and XLM-RoBERTa, have proved useful for many multi-lingual tasks. Previous work probed the cross-linguality of the representations indirectly using zero-shot transfer learning on morphological and syntactic tasks. We instead investigate the language-neutrality of multilingual contextual embeddings directly and with respect to lexical semantics. Our results show that contextual embeddings are more language-neutral and, in general, more informative than aligned static word-type embeddings, which are explicitly trained for language neutrality. Contextual embeddings are still only moderately language-neutral by default, so we propose two simple methods for achieving stronger language neutrality: first, by unsupervised centering of the representation for each language and second, by fitting an explicit projection on small parallel data. Besides, we show how to reach state-of-the-art accuracy on language identification and match the performance of statistical methods for word alignment of parallel sentences without using parallel data."
                },
                {
                    "title": "Cross-Lingual Ability of Multilingual BERT: An Empirical Study",
                    "abstract": "Recent work has exhibited the surprising cross-lingual abilities of multilingual BERT (M-BERT) -- surprising since it is trained without any cross-lingual objective and with no aligned data. In this work, we provide a comprehensive study of the contribution of different components in M-BERT to its cross-lingual ability. We study the impact of linguistic properties of the languages, the architecture of the model, and the learning objectives. The experimental study is done in the context of three typologically different languages -- Spanish, Hindi, and Russian -- and using two conceptually different NLP tasks, textual entailment and named entity recognition. Among our key conclusions is the fact that the lexical overlap between languages plays a negligible role in the cross-lingual success, while the depth of the network is an integral part of it. All our models and implementations can be found on our project page: this http URL ."
                },
                {
                    "title": "On the Cross-lingual Transferability of Monolingual Representations",
                    "abstract": "State-of-the-art unsupervised multilingual models (e.g., multilingual BERT) have been shown to generalize in a zero-shot cross-lingual setting. This generalization ability has been attributed to the use of a shared subword vocabulary and joint training across multiple languages giving rise to deep multilingual abstractions. We evaluate this hypothesis by designing an alternative approach that transfers a monolingual model to new languages at the lexical level. More concretely, we first train a transformer-based masked language model on one language, and transfer it to a new language by learning a new embedding matrix with the same masked language modeling objective, freezing parameters of all other layers. This approach does not rely on a shared vocabulary or joint training. However, we show that it is competitive with multilingual BERT on standard cross-lingual classification benchmarks and on a new Cross-lingual Question Answering Dataset (XQuAD). Our results contradict common beliefs of the basis of the generalization ability of multilingual models and suggest that deep monolingual models learn some abstractions that generalize across languages. We also release XQuAD as a more comprehensive cross-lingual benchmark, which comprises 240 paragraphs and 1190 question-answer pairs from SQuAD v1.1 translated into ten languages by professional translators."
                },
                {
                    "title": "Designing and Interpreting Probes with Control Tasks",
                    "abstract": "Probes, supervised models trained to predict properties (like parts-of-speech) from representations (like ELMo), have achieved high accuracy on a range of linguistic tasks. But does this mean that the representations encode linguistic structure or just that the probe has learned the linguistic task? In this paper, we propose control tasks, which associate word types with random outputs, to complement linguistic tasks. By construction, these tasks can only be learned by the probe itself. So a good probe, (one that reflects the representation), should be selective, achieving high linguistic task accuracy and low control task accuracy. The selectivity of a probe puts linguistic task accuracy in context with the probe\u2019s capacity to memorize from word types. We construct control tasks for English part-of-speech tagging and dependency edge prediction, and show that popular probes on ELMo representations are not selective. We also find that dropout, commonly used to control probe complexity, is ineffective for improving selectivity of MLPs, but that other forms of regularization are effective. Finally, we find that while probes on the first layer of ELMo yield slightly better part-of-speech tagging accuracy than the second, probes on the second layer are substantially more selective, which raises the question of which layer better represents parts-of-speech."
                },
                {
                    "title": "A Multiscale Visualization of Attention in the Transformer Model",
                    "abstract": "The Transformer is a sequence model that forgoes traditional recurrent architectures in favor of a fully attention-based approach. Besides improving performance, an advantage of using attention is that it can also help to interpret a model by showing how the model assigns weight to different input elements. However, the multi-layer, multi-head attention mechanism in the Transformer model can be difficult to decipher. To make the model more accessible, we introduce an open-source tool that visualizes attention at multiple scales, each of which provides a unique perspective on the attention mechanism. We demonstrate the tool on BERT and OpenAI GPT-2 and present three example use cases: detecting model bias, locating relevant attention heads, and linking neurons to model behavior."
                },
                {
                    "title": "Choosing Transfer Languages for Cross-Lingual Learning",
                    "abstract": "Cross-lingual transfer, where a high-resource transfer language is used to improve the accuracy of a low-resource task language, is now an invaluable tool for improving performance of natural language processing (NLP) on low-resource languages. However, given a particular task language, it is not clear which language to transfer from, and the standard strategy is to select languages based on ad hoc criteria, usually the intuition of the experimenter. Since a large number of features contribute to the success of cross-lingual transfer (including phylogenetic similarity, typological properties, lexical overlap, or size of available data), even the most enlightened experimenter rarely considers all these factors for the particular task at hand. In this paper, we consider this task of automatically selecting optimal transfer languages as a ranking problem, and build models that consider the aforementioned features to perform this prediction. In experiments on representative NLP tasks, we demonstrate that our model predicts good transfer languages much better than ad hoc baselines considering single features in isolation, and glean insights on what features are most informative for each different NLP tasks, which may inform future ad hoc selection even without use of our method."
                },
                {
                    "title": "The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks",
                    "abstract": "Neural network pruning techniques can reduce the parameter counts of trained networks by over 90%, decreasing storage requirements and improving computational performance of inference without compromising accuracy. However, contemporary experience is that the sparse architectures produced by pruning are difficult to train from the start, which would similarly improve training performance. \nWe find that a standard pruning technique naturally uncovers subnetworks whose initializations made them capable of training effectively. Based on these results, we articulate the \"lottery ticket hypothesis:\" dense, randomly-initialized, feed-forward networks contain subnetworks (\"winning tickets\") that - when trained in isolation - reach test accuracy comparable to the original network in a similar number of iterations. The winning tickets we find have won the initialization lottery: their connections have initial weights that make training particularly effective. \nWe present an algorithm to identify winning tickets and a series of experiments that support the lottery ticket hypothesis and the importance of these fortuitous initializations. We consistently find winning tickets that are less than 10-20% of the size of several fully-connected and convolutional feed-forward architectures for MNIST and CIFAR10. Above this size, the winning tickets that we find learn faster than the original network and reach higher test accuracy."
                },
                {
                    "title": "Cross-Lingual Transfer Learning for POS Tagging without Cross-Lingual Resources",
                    "abstract": "Training a POS tagging model with crosslingual transfer learning usually requires linguistic knowledge and resources about the relation between the source language and the target language. In this paper, we introduce a cross-lingual transfer learning model for POS tagging without ancillary resources such as parallel corpora. The proposed cross-lingual model utilizes a common BLSTM that enables knowledge transfer from other languages, and private BLSTMs for language-specific representations. The cross-lingual model is trained with language-adversarial training and bidirectional language modeling as auxiliary objectives to better represent language-general information while not losing the information about a specific target language. Evaluating on POS datasets from 14 languages in the Universal Dependencies corpus, we show that the proposed transfer learning model improves the POS tagging performance of the target languages without exploiting any linguistic knowledge between the source language and the target language."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow do large language models handle multilingualism?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding how large language models (LLMs) process multilingual inputs is crucial for advancing the field of natural language processing (NLP). By addressing this question, we can uncover the underlying mechanisms that enable LLMs to understand and generate text in multiple languages, which has significant implications for improving model performance across diverse linguistic contexts. This research could lead to more effective multilingual applications, enhance cross-lingual transfer learning, and inform the design of future models that are better equipped to handle the complexities of multilingual communication.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge in solving this problem lies in the intricate and often opaque nature of LLMs' internal mechanisms. Naive approaches may fail because they do not account for the dynamic transformations that occur within the model's layers as it processes multilingual queries. The complexities include understanding how different layers interact with language-specific features, the need for a comprehensive analysis of embeddings across multiple languages, and the difficulty in isolating the contributions of various model components (like self-attention and feed-forward layers) to the multilingual processing workflow. Additionally, the lack of interpretability in LLMs makes it hard to ascertain how they manage language transitions.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on monolingual capabilities or specific language pairs, neglecting the broader multilingual features of LLMs. Existing studies often lack a holistic view of how LLMs transition between languages, and many rely on methods that do not effectively capture the nuances of multilingual processing. Barriers include the absence of robust methodologies for analyzing language-specific parameters and the reliance on fine-tuning or labeled data, which may not be applicable in a multilingual context. Our approach, which introduces the Parallel Language-specific Neuron Detection (PLND) method, aims to overcome these limitations by providing a novel way to identify and assess language-specific parameters without the constraints of previous methods.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a three-stage multilingual workflow (MWork) consisting of understanding, task-solving, and generating. We will test LLMs with various non-English queries and decode the hidden embeddings of each layer to analyze the ratio of English and non-English tokens. The dataset will include diverse multilingual queries"
            }
        },
        "author_data": {
            "e2465f08-4db3-4064-b7e8-f2003907b002": {
                "pk": "e2465f08-4db3-4064-b7e8-f2003907b002",
                "name": "Yiran Zhao",
                "collaborators": [
                    "Tiangang Cui",
                    "Wenxuan Zhang",
                    "Huiming Wang",
                    "Kenji Kawaguchi",
                    "Lidong Bing"
                ],
                "domain": [
                    "Machine Learning",
                    "Bayesian Inference",
                    "Cross-lingual Transfer",
                    "Tensor Decomposition"
                ],
                "publications": [
                    {
                        "title": "Tensor-train methods for sequential state and parameter learning in state-space models",
                        "abstract": "We consider sequential state and parameter learning in state-space models with intractable state transition and observation processes. By exploiting low-rank tensor train (TT) decompositions, we propose new sequential learning methods for joint parameter and state estimation under the Bayesian framework. Our key innovation is the introduction of scalable function approximation tools such as TT for recursively learning the sequentially updated posterior distributions. The function approximation perspective of our methods offers tractable error analysis and potentially alleviates the particle degeneracy faced by many particle-based methods. In addition to the new insights into the algorithmic design, our methods complement conventional particle-based methods. Our TT-based approximations naturally define conditional Knothe--Rosenblatt (KR) rearrangements that lead to parameter estimation, filtering, smoothing and path estimation accompanying our sequential learning algorithms, which open the door to removing potential approximation bias. We also explore several preconditioning techniques based on either linear or nonlinear KR rearrangements to enhance the approximation power of TT for practical problems. We demonstrate the efficacy and efficiency of our proposed methods on several state-space models, in which our methods achieve state-of-the-art estimation accuracy and computational performance."
                    },
                    {
                        "title": "AdaMergeX: Cross-Lingual Transfer with Large Language Models via Adaptive Adapter Merging",
                        "abstract": "As an effective alternative to the direct fine-tuning on target tasks in specific languages, cross-lingual transfer addresses the challenges of limited training data by decoupling ''task ability'' and ''language ability'' by fine-tuning on the target task in the source language and another selected task in the target language, respectively. However, they fail to fully separate the task ability from the source language or the language ability from the chosen task. In this paper, we acknowledge the mutual reliance between task ability and language ability and direct our attention toward the gap between the target language and the source language on tasks. As the gap removes the impact of tasks, we assume that it remains consistent across tasks. Based on this assumption, we propose a new cross-lingual transfer method called $\\texttt{AdaMergeX}$ that utilizes adaptive adapter merging. By introducing a reference task, we can determine that the divergence of adapters fine-tuned on the reference task in both languages follows the same distribution as the divergence of adapters fine-tuned on the target task in both languages. Hence, we can obtain target adapters by combining the other three adapters. Furthermore, we propose a structure-adaptive adapter merging method. Our empirical results demonstrate that our approach yields new and effective cross-lingual transfer, outperforming existing methods across all settings."
                    }
                ]
            },
            "d58e2802-b28c-41ae-9e14-a2090255f8f4": {
                "pk": "d58e2802-b28c-41ae-9e14-a2090255f8f4",
                "name": "Wenxuan Zhang",
                "collaborators": [
                    "Wai Lam",
                    "Yang Deng",
                    "Paul Janson",
                    "Mohamed Elhoseiny",
                    "Lidong Bing",
                    "Chang Gao",
                    "Jing Ma",
                    "Rahaf Aljundi",
                    "Xin Li",
                    "Kai Yi"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Question Answering",
                    "Information Extraction",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Answer Ranking for Product-Related Questions via Multiple Semantic Relations Modeling",
                        "abstract": "Many E-commerce sites now offer product-specific question answering platforms for users to communicate with each other by posting and answering questions during online shopping. However, the multiple answers provided by ordinary users usually vary diversely in their qualities and thus need to be appropriately ranked for each question to improve user satisfaction. It can be observed that product reviews usually provide useful information for a given question, and thus can assist the ranking process. In this paper, we investigate the answer ranking problem for product-related questions, with the relevant reviews treated as auxiliary information that can be exploited for facilitating the ranking. We propose an answer ranking model named MUSE which carefully models multiple semantic relations among the question, answers, and relevant reviews. Specifically, MUSE constructs a multi-semantic relation graph with the question, each answer, and each review snippet as nodes. Then a customized graph convolutional neural network is designed for explicitly modeling the semantic relevance between the question and answers, the content consistency among answers, and the textual entailment between answers and reviews. Extensive experiments on real-world E-commerce datasets across three product categories show that our proposed model achieves superior performance on the concerned answer ranking task."
                    },
                    {
                        "title": "Review-guided Helpful Answer Identification in E-commerce",
                        "abstract": "Product-specific community question answering platforms can greatly help address the concerns of potential customers. However, the user-provided answers on such platforms often vary a lot in their qualities. Helpfulness votes from the community can indicate the overall quality of the answer, but they are often missing. Accurately predicting the helpfulness of an answer to a given question and thus identifying helpful answers is becoming a demanding need. Since the helpfulness of an answer depends on multiple perspectives instead of only topical relevance investigated in typical QA tasks, common answer selection algorithms are insufficient for tackling this task. In this paper, we propose the Review-guided Answer Helpfulness Prediction (RAHP) model that not only considers the interactions between QA pairs but also investigates the opinion coherence between the answer and crowds' opinions reflected in the reviews, which is another important factor to identify helpful answers. Moreover, we tackle the task of determining opinion coherence as a language inference problem and explore the utilization of pre-training strategy to transfer the textual inference knowledge obtained from a specifically designed trained network. Extensive experiments conducted on real-world data across seven product categories show that our proposed model achieves superior performance on the prediction task."
                    },
                    {
                        "title": "Overcoming Generic Knowledge Loss with Selective Parameter Update",
                        "abstract": "Foundation models encompass an extensive knowledge base and offer remarkable transferability. However, this knowledge becomes outdated or insufficient over time. The challenge lies in continuously updating foundation models to accommodate novel information while retaining their original capabilities. Leveraging the fact that foundation models have initial knowledge on various tasks and domains, we propose a novel approach that, instead of updating all parameters equally, localizes the updates to a sparse set of parameters relevant to the task being learned. We strike a balance between efficiency and new task performance, while maintaining the transferability and generalizability of foundation models. We extensively evaluate our method on foundational vision-language models with a diverse spectrum of continual learning tasks. Our method achieves improvements on the accuracy of the newly learned tasks up to 7% while preserving the pretraining knowledge with a negligible decrease of 0.9% on a representative control set accuracy."
                    },
                    {
                        "title": "Exploiting BERT for End-to-End Aspect-based Sentiment Analysis",
                        "abstract": "In this paper, we investigate the modeling power of contextualized embeddings from pre-trained language models, e.g. BERT, on the E2E-ABSA task. Specifically, we build a series of simple yet insightful neural baselines to deal with E2E-ABSA. The experimental results show that even with a simple linear classification layer, our BERT-based architecture can outperform state-of-the-art works. Besides, we also standardize the comparative study by consistently utilizing a hold-out validation dataset for model selection, which is largely ignored by previous works. Therefore, our work can serve as a BERT-based benchmark for E2E-ABSA."
                    },
                    {
                        "title": "UniGDD: A Unified Generative Framework for Goal-Oriented Document-Grounded Dialogue",
                        "abstract": "The goal-oriented document-grounded dialogue aims at responding to the user query based on the dialogue context and supporting document. Existing studies tackle this problem by decomposing it into two sub-tasks: knowledge identification and response generation. However, such pipeline methods would unavoidably suffer from the error propagation issue. This paper proposes to unify these two sub-tasks via sequentially generating the grounding knowledge and the response. We further develop a prompt-connected multi-task learning strategy to model the characteristics and connections of different tasks and introduce linear temperature scheduling to reduce the negative effect of irrelevant document information. Experimental results demonstrate the effectiveness of our framework."
                    },
                    {
                        "title": "Multi-hop Inference for Question-driven Summarization",
                        "abstract": "Question-driven summarization has been recently studied as an effective approach to summarizing the source document to produce concise but informative answers for non-factoid questions. In this work, we propose a novel question-driven abstractive summarization method, Multi-hop Selective Generator (MSG), to incorporate multi-hop reasoning into question-driven summarization and, meanwhile, provide justifications for the generated summaries. Specifically, we jointly model the relevance to the question and the interrelation among different sentences via a human-like multi-hop inference module, which captures important sentences for justifying the summarized answer. A gated selective pointer generator network with a multi-view coverage mechanism is designed to integrate diverse information from different perspectives. Experimental results show that the proposed method consistently outperforms state-of-the-art methods on two non-factoid QA datasets, namely WikiHow and PubMedQA."
                    },
                    {
                        "title": "Learning to Rank Question Answer Pairs with Bilateral Contrastive Data Augmentation",
                        "abstract": "In this work, we propose a novel and easy-to-apply data augmentation strategy, namely Bilateral Generation (BiG), with a contrastive training objective for improving the performance of ranking question answer pairs with existing labeled data. In specific, we synthesize pseudo-positive QA pairs in contrast to the original negative QA pairs with two pre-trained generation models, one for question generation, the other for answer generation, which are fine-tuned on the limited positive QA pairs from the original dataset. With the augmented dataset, we design a contrastive training objective for learning to rank question answer pairs. Experimental results on three benchmark datasets show that our method significantly improves the performance of ranking models by making full use of existing labeled data and can be easily applied to different ranking models."
                    },
                    {
                        "title": "Domain-Aware Continual Zero-Shot Learning",
                        "abstract": "Modern visual systems have a wide range of potential applications in vision tasks for natural science research, such as aiding in species discovery, monitoring animals in the wild, and so on. However, real-world vision tasks may experience changes in environmental conditions, leading to shifts in how captured images are presented. To address this issue, we introduce Domain-Aware Continual Zero-Shot Learning (DACZSL), a task to recognize images of unseen categories in continuously changing domains. Accordingly, we propose a Domain-Invariant Network (DIN) to learn factorized features for shifting domains and improved textual representation for unseen classes. DIN continually learns a global shared network for domain-invariant and task-invariant features, and per-task private networks for task-specific features. Furthermore, we enhance the dual network with class-wise learnable prompts to improve class-level text representation, thereby improving zero-shot prediction of future unseen classes. To evaluate DACZSL, we introduce two benchmarks, DomainNet-CZSL and iWildCam-CZSL. Our results show that DIN significantly outperforms existing baselines by over 5% in harmonic accuracy and over 1% in backward transfer and achieves a new SoTA."
                    },
                    {
                        "title": "Product Question Answering in E-Commerce: A Survey",
                        "abstract": "Product question answering (PQA), aiming to automatically provide instant responses to customer's questions in E-Commerce platforms, has drawn increasing attention in recent years. Compared with typical QA problems, PQA exhibits unique challenges such as the subjectivity and reliability of user-generated contents in E-commerce platforms. Therefore, various problem settings and novel methods have been proposed to capture these special characteristics. In this paper, we aim to systematically review existing research efforts on PQA. Specifically, we categorize PQA studies into four problem settings in terms of the form of provided answers. We analyze the pros and cons, as well as present existing datasets and evaluation protocols for each setting. We further summarize the most significant challenges that characterize PQA from general QA applications and discuss their corresponding solutions. Finally, we conclude this paper by providing the prospect on several future directions."
                    },
                    {
                        "title": "Easy-to-Hard Learning for Information Extraction",
                        "abstract": "Information extraction (IE) systems aim to automatically extract structured information, such as named entities, relations between entities, and events, from unstructured texts. While most existing work addresses a particular IE task, universally modeling various IE tasks with one model has achieved great success recently. Despite their success, they employ a one-stage learning strategy, i.e., directly learning to extract the target structure given the input text, which contradicts the human learning process. In this paper, we propose a unified easy-to-hard learning framework consisting of three stages, i.e., the easy stage, the hard stage, and the main stage, for IE by mimicking the human learning process. By breaking down the learning process into multiple stages, our framework facilitates the model to acquire general IE task knowledge and improve its generalization ability. Extensive experiments across four IE tasks demonstrate the effectiveness of our framework. We achieve new state-of-the-art results on 13 out of 17 datasets. Our code is available at \\url{https://github.com/DAMO-NLP-SG/IE-E2H}."
                    }
                ]
            },
            "89df31c7-94d1-49de-b209-8f827c67d964": {
                "pk": "89df31c7-94d1-49de-b209-8f827c67d964",
                "name": "Guizhen Chen",
                "collaborators": [
                    "Lidong Bing",
                    "Luu Anh Tuan",
                    "Yew Ken Chia",
                    "Soujanya Poria",
                    "Wenxuan Zhang",
                    "Anh Tuan Luu",
                    "Chang Gao",
                    "Wai Lam",
                    "Liying Cheng",
                    "Hui Chen"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Data Augmentation",
                    "Instruction Tuning"
                ],
                "publications": [
                    {
                        "title": "JsonTuning: Towards Generalizable, Robust, and Controllable Instruction Tuning",
                        "abstract": "Instruction tuning has become an essential process for optimizing the performance of large language models (LLMs). However, current text-to-text instruction tuning methods, referred to as TextTuning, exhibit significant limitations in terms of generalization, robustness, and controllability, primarily due to the absence of explicit task structures. In this paper, we introduce JsonTuning, a novel structure-to-structure approach for instruction tuning. By utilizing the versatile and structured format of JSON to represent tasks, JsonTuning enhances generalization by enabling the model to comprehend essential task elements and their interrelations, improves robustness by reducing ambiguity, and increases controllability by providing explicit control over the output. We conduct a comprehensive comparative analysis between JsonTuning and TextTuning using various language models and evaluation benchmarks. Our experimental results demonstrate that JsonTuning consistently outperforms TextTuning across a range of applications, showing marked improvements in performance, robustness, and controllability. By addressing the inherent limitations of TextTuning, JsonTuning reveals significant potential for developing more effective and reliable LLMs capable of managing diverse scenarios."
                    },
                    {
                        "title": "Exploring the Potential of Large Language Models in Computational Argumentation",
                        "abstract": "Computational argumentation has become an essential tool in various domains, including law, public policy, and artificial intelligence. It is an emerging research field in natural language processing that attracts increasing attention. Research on computational argumentation mainly involves two types of tasks: argument mining and argument generation. As large language models (LLMs) have demonstrated impressive capabilities in understanding context and generating natural language, it is worthwhile to evaluate the performance of LLMs on diverse computational argumentation tasks. This work aims to embark on an assessment of LLMs, such as ChatGPT, Flan models, and LLaMA2 models, in both zero-shot and few-shot settings. We organize existing tasks into six main categories and standardize the format of fourteen openly available datasets. In addition, we present a new benchmark dataset on counter speech generation that aims to holistically evaluate the end-to-end performance of LLMs on argument mining and argument generation. Extensive experiments show that LLMs exhibit commendable performance across most of the datasets, demonstrating their capabilities in the field of argumentation. Our analysis offers valuable suggestions for evaluating computational argumentation and its integration with LLMs in future research endeavors."
                    },
                    {
                        "title": "Contrastive Chain-of-Thought Prompting",
                        "abstract": "Despite the success of chain of thought in enhancing language model reasoning, the underlying process remains less well understood. Although logically sound reasoning appears inherently crucial for chain of thought, prior studies surprisingly reveal minimal impact when using invalid demonstrations instead. Furthermore, the conventional chain of thought does not inform language models on what mistakes to avoid, which potentially leads to more errors. Hence, inspired by how humans can learn from both positive and negative examples, we propose contrastive chain of thought to enhance language model reasoning. Compared to the conventional chain of thought, our approach provides both valid and invalid reasoning demonstrations, to guide the model to reason step-by-step while reducing reasoning mistakes. To improve generalization, we introduce an automatic method to construct contrastive demonstrations. Our experiments on reasoning benchmarks demonstrate that contrastive chain of thought can serve as a general enhancement of chain-of-thought prompting."
                    },
                    {
                        "title": "Domain-Expanded ASTE: Rethinking Generalization in Aspect Sentiment Triplet Extraction",
                        "abstract": "Aspect Sentiment Triplet Extraction (ASTE) is a subtask of Aspect-Based Sentiment Analysis (ABSA) that considers each opinion term, their expressed sentiment, and the corresponding aspect targets. However, existing methods are limited to the in-domain setting with two domains. Hence, we propose a domain-expanded benchmark to address the in-domain, out-of-domain and cross-domain settings. We support the new benchmark by annotating more than 4000 data samples for two new domains based on hotel and cosmetics reviews. Our analysis of five existing methods shows that while there is a significant gap between in-domain and out-of-domain performance, generative methods have a strong potential for domain generalization. Our datasets, code implementation and models are available at https://github.com/DAMO-NLP-SG/domain-expanded-aste ."
                    },
                    {
                        "title": "Reasoning Paths Optimization: Learning to Reason and Explore From Diverse Paths",
                        "abstract": "Advanced models such as OpenAI o1 exhibit impressive problem-solving capabilities through step-by-step reasoning. However, they may still falter on more complex problems, making errors that disrupt their reasoning paths. We attribute this to the expansive solution space, where each step has the risk of diverging into mistakes. To enhance language model reasoning, we introduce a specialized training framework called Reasoning Paths Optimization (RPO), which enables learning to reason and explore from diverse paths. Our approach encourages favorable branches at each reasoning step while penalizing unfavorable ones, enhancing the model's overall problem-solving performance. Reasoning Paths Optimization does not rely on large-scale human-annotated rationales or outputs from closed-source models, making it scalable and data-efficient. We focus on multi-step reasoning tasks, such as math word problems and science-based exam questions. The experiments demonstrate that our framework significantly enhances the reasoning performance of large language models, with up to 3.1% and 4.3% improvement on GSM8K and MMLU (STEM) respectively. Our data and code can be found at https://reasoning-paths.github.io."
                    },
                    {
                        "title": "Zero-Shot Text Classification via Self-Supervised Tuning",
                        "abstract": "Existing solutions to zero-shot text classification either conduct prompting with pre-trained language models, which is sensitive to the choices of templates, or rely on large-scale annotated data of relevant tasks for meta-tuning. In this work, we propose a new paradigm based on self-supervised learning to solve zero-shot text classification tasks by tuning the language models with unlabeled data, called self-supervised tuning. By exploring the inherent structure of free texts, we propose a new learning objective called first sentence prediction to bridge the gap between unlabeled data and text classification tasks. After tuning the model to learn to predict the first sentence in a paragraph based on the rest, the model is able to conduct zero-shot inference on unseen tasks such as topic classification and sentiment analysis. Experimental results show that our model outperforms the state-of-the-art baselines on 7 out of 10 tasks. Moreover, the analysis reveals that our model is less sensitive to the prompt design. Our code and pre-trained models are publicly available at https://github.com/DAMO-NLP-SG/SSTuning ."
                    },
                    {
                        "title": "Data Augmentation using Large Language Models: Data Perspectives, Learning Paradigms and Challenges",
                        "abstract": "In the rapidly evolving field of large language models (LLMs), data augmentation (DA) has emerged as a pivotal technique for enhancing model performance by diversifying training examples without the need for additional data collection. This survey explores the transformative impact of LLMs on DA, particularly addressing the unique challenges and opportunities they present in the context of natural language processing (NLP) and beyond. From both data and learning perspectives, we examine various strategies that utilize LLMs for data augmentation, including a novel exploration of learning paradigms where LLM-generated data is used for diverse forms of further training. Additionally, this paper highlights the primary open challenges faced in this domain, ranging from controllable data augmentation to multi-modal data augmentation. This survey highlights a paradigm shift introduced by LLMs in DA, and aims to serve as a comprehensive guide for researchers and practitioners."
                    }
                ]
            },
            "6ed90478-d3b4-4733-adad-62811f62a379": {
                "pk": "6ed90478-d3b4-4733-adad-62811f62a379",
                "name": "Kenji Kawaguchi",
                "collaborators": [
                    "Leslie Pack Kaelbling",
                    "Jiaoyang Huang",
                    "Yoshua Bengio",
                    "Haihao Lu",
                    "Zhun Deng",
                    "Vikas Verma",
                    "Jascha Sohl-Dickstein",
                    "Mauricio Araya",
                    "Qingyun Sun",
                    "Tom\u00e1s Lozano-P\u00e9rez"
                ],
                "domain": [
                    "Deep Learning",
                    "Non-convex Optimization",
                    "Reinforcement Learning",
                    "Bayesian Methods"
                ],
                "publications": [
                    {
                        "title": "Bounded Optimal Exploration in MDP",
                        "abstract": "Within the framework of probably approximately correct Markov decision processes (PAC-MDP), much theoretical work has focused on methods to attain near optimality after a relatively long period of learning and exploration. However, practical concerns require the attainment of satisfactory behavior within a short period of time. In this paper, we relax the PAC-MDP conditions to reconcile theoretically driven exploration methods and practical needs. We propose simple algorithms for discrete and continuous state spaces, and illustrate the benefits of our proposed relaxation via theoretical analyses and numerical examples. Our algorithms also maintain anytime error bounds and average loss bounds. Our approach accommodates both Bayesian and non-Bayesian methods."
                    },
                    {
                        "title": "On the Theory of Implicit Deep Learning: Global Convergence with Implicit Layers",
                        "abstract": "A deep equilibrium model uses implicit layers, which are implicitly defined through an equilibrium point of an infinite sequence of computation. It avoids any explicit computation of the infinite sequence by finding an equilibrium point directly via root-finding and by computing gradients via implicit differentiation. In this paper, we analyze the gradient dynamics of deep equilibrium models with nonlinearity only on weight matrices and non-convex objective functions of weights for regression and classification. Despite non-convexity, convergence to global optimum at a linear rate is guaranteed without any assumption on the width of the models, allowing the width to be smaller than the output dimension and the number of data points. Moreover, we prove a relation between the gradient dynamics of the deep implicit layer and the dynamics of trust region Newton method of a shallow explicit layer. This mathematically proven relation along with our numerical observation suggests the importance of understanding implicit bias of implicit layers and an open problem on the topic. Our proofs deal with implicit layers, weight tying and nonlinearity on weights, and differ from those in the related literature."
                    },
                    {
                        "title": "Deep Learning without Poor Local Minima",
                        "abstract": "In this paper, we prove a conjecture published in 1989 and also partially address an open problem announced at the Conference on Learning Theory (COLT) 2015. With no unrealistic assumption, we first prove the following statements for the squared loss function of deep linear neural networks with any depth and any widths: 1) the function is non-convex and non-concave, 2) every local minimum is a global minimum, 3) every critical point that is not a global minimum is a saddle point, and 4) there exist \"bad\" saddle points (where the Hessian has no negative eigenvalue) for the deeper networks (with more than three layers), whereas there is no bad saddle point for the shallow networks (with three layers). Moreover, for deep nonlinear neural networks, we prove the same four statements via a reduction to a deep linear model under the independence assumption adopted from recent work. As a result, we present an instance, for which we can answer the following question: how difficult is it to directly train a deep model in theory? It is more difficult than the classical machine learning models (because of the non-convexity), but not too difficult (because of the nonexistence of poor local minima). Furthermore, the mathematically proven existence of bad saddle points for deeper models would suggest a possible open problem. We note that even though we have advanced the theoretical foundations of deep learning and non-convex optimization, there is still a gap between theory and practice."
                    },
                    {
                        "title": "Depth with Nonlinearity Creates No Bad Local Minima in ResNets",
                        "abstract": "In this paper, we prove that depth with nonlinearity creates no bad local minima in a type of arbitrarily deep ResNets with arbitrary nonlinear activation functions, in the sense that the values of all local minima are no worse than the global minimum value of corresponding classical machine-learning models, and are guaranteed to further improve via residual representations. As a result, this paper provides an affirmative answer to an open question stated in a paper in the conference on Neural Information Processing Systems 2018. This paper advances the optimization theory of deep learning only for ResNets and not for other network architectures."
                    },
                    {
                        "title": "Ordered SGD: A New Stochastic Optimization Framework for Empirical Risk Minimization",
                        "abstract": "We propose a new stochastic optimization framework for empirical risk minimization problems such as those that arise in machine learning. The traditional approaches, such as (mini-batch) stochastic gradient descent (SGD), utilize an unbiased gradient estimator of the empirical average loss. In contrast, we develop a computationally efficient method to construct a gradient estimator that is purposely biased toward those observations with higher current losses. On the theory side, we show that the proposed method minimizes a new ordered modification of the empirical average loss, and is guaranteed to converge at a sublinear rate to a global optimum for convex loss and to a critical point for weakly convex (non-convex) loss. Furthermore, we prove a new generalization bound for the proposed algorithm. On the empirical side, the numerical experiments show that our proposed method consistently improves the test errors compared with the standard mini-batch SGD in various models including SVM, logistic regression, and deep learning problems."
                    },
                    {
                        "title": "Gradient Descent Finds Global Minima for Generalizable Deep Neural Networks of Practical Sizes",
                        "abstract": "In this paper, we theoretically prove that gradient descent can find a global minimum of non-convex optimization of all layers for nonlinear deep neural networks of sizes commonly encountered in practice. The theory developed in this paper only requires the practical degrees of over-parameterization unlike previous theories. Our theory only requires the number of trainable parameters to increase linearly as the number of training samples increases. This allows the size of the deep neural networks to be consistent with practice and to be several orders of magnitude smaller than that required by the previous theories. Moreover, we prove that the linear increase of the size of the network is the optimal rate and that it cannot be improved, except by a logarithmic factor. Furthermore, deep neural networks with the trainability guarantee are shown to generalize well to unseen test samples with a natural dataset but not a random dataset."
                    },
                    {
                        "title": "Eliminating all bad Local Minima from Loss Landscapes without even adding an Extra Unit",
                        "abstract": "Recent work has noted that all bad local minima can be removed from neural network loss landscapes, by adding a single unit with a particular parameterization. We show that the core technique from these papers can be used to remove all bad local minima from any loss landscape, so long as the global minimum has a loss of zero. This procedure does not require the addition of auxiliary units, or even that the loss be associated with a neural network. The method of action involves all bad local minima being converted into bad (non-local) minima at infinity in terms of auxiliary parameters."
                    },
                    {
                        "title": "Depth Creates No Bad Local Minima",
                        "abstract": "In deep learning, \\textit{depth}, as well as \\textit{nonlinearity}, create non-convex loss surfaces. Then, does depth alone create bad local minima? In this paper, we prove that without nonlinearity, depth alone does not create bad local minima, although it induces non-convex loss surface. Using this insight, we greatly simplify a recently proposed proof to show that all of the local minima of feedforward deep linear neural networks are global minima. Our theoretical results generalize previous results with fewer assumptions, and this analysis provides a method to show similar results beyond square loss in deep linear models."
                    },
                    {
                        "title": "A Greedy Approximation of Bayesian Reinforcement Learning with Probably Optimistic Transition Model",
                        "abstract": "Bayesian Reinforcement Learning (RL) is capable of not only incorporating domain knowledge, but also solving the exploration-exploitation dilemma in a natural way. As Bayesian RL is intractable except for special cases, previous work has proposed several approximation methods. However, these methods are usually too sensitive to parameter values, and finding an acceptable parameter setting is practically impossible in many applications. In this paper, we propose a new algorithm that greedily approximates Bayesian RL to achieve robustness in parameter space. We show that for a desired learning behavior, our proposed algorithm has a polynomial sample complexity that is lower than those of existing algorithms. We also demonstrate that the proposed algorithm naturally outperforms other existing algorithms when the prior distributions are not significantly misleading. On the other hand, the proposed algorithm cannot handle greatly misspecified priors as well as the other algorithms can. This is a natural consequence of the fact that the proposed algorithm is greedier than the other algorithms. Accordingly, we discuss a way to select an appropriate algorithm for different tasks based on the algorithms' greediness. We also introduce a new way of simplifying Bayesian planning, based on which future work would be able to derive new algorithms."
                    },
                    {
                        "title": "A Recipe for Global Convergence Guarantee in Deep Neural Networks",
                        "abstract": "Existing global convergence guarantees of (stochastic) gradient descent do not apply to practical deep networks in the practical regime of deep learning beyond the neural tangent kernel (NTK) regime. This paper proposes an algorithm, which is ensured to have global convergence guarantees in the practical regime beyond the NTK regime, under a verifiable condition called the expressivity condition. The expressivity condition is defined to be both data-dependent and architecture-dependent, which is the key property that makes our results applicable for practical settings beyond the NTK regime. On the one hand, the expressivity condition is theoretically proven to hold data-independently for fully-connected deep neural networks with narrow hidden layers and a single wide layer. On the other hand, the expressivity condition is numerically shown to hold data-dependently for deep (convolutional) ResNet with batch normalization with various standard image datasets. We also show that the proposed algorithm has generalization performances comparable with those of the heuristic algorithm, with the same hyper-parameters and total number of iterations. Therefore, the proposed algorithm can be viewed as a step towards providing theoretical guarantees for deep learning in the practical regime."
                    },
                    {
                        "title": "Bayesian Optimization with Exponential Convergence",
                        "abstract": "This paper presents a Bayesian optimization method with exponential convergence without the need of auxiliary optimization and without the delta-cover sampling. Most Bayesian optimization methods require auxiliary optimization: an additional non-convex global optimization problem, which can be time-consuming and hard to implement in practice. Also, the existing Bayesian optimization method with exponential convergence requires access to the delta-cover sampling, which was considered to be impractical. Our approach eliminates both requirements and achieves an exponential convergence rate."
                    },
                    {
                        "title": "Global Continuous Optimization with Error Bound and Fast Convergence",
                        "abstract": "This paper considers global optimization with a black-box unknown objective function that can be non-convex and non-differentiable. Such a difficult optimization problem arises in many real-world applications, such as parameter tuning in machine learning, engineering design problem, and planning with a complex physics simulator. This paper proposes a new global optimization algorithm, called Locally Oriented Global Optimization (LOGO), to aim for both fast convergence in practice and finite-time error bound in theory. The advantage and usage of the new algorithm are illustrated via theoretical analysis and an experiment conducted with 11 benchmark test functions. Further, we modify the LOGO algorithm to specifically solve a planning problem via policy search with continuous state/action space and long time horizon while maintaining its finite-time error bound. We apply the proposed planning method to accident management of a nuclear power plant. The result of the application study demonstrates the practical utility of our method."
                    },
                    {
                        "title": "Effect of Depth and Width on Local Minima in Deep Learning",
                        "abstract": "In this paper, we analyze the effects of depth and width on the quality of local minima, without strong over-parameterization and simplification assumptions in the literature. Without any simplification assumption, for deep nonlinear neural networks with the squared loss, we theoretically show that the quality of local minima tends to improve towards the global minimum value as depth and width increase. Furthermore, with a locally-induced structure on deep nonlinear neural networks, the values of local minima of neural networks are theoretically proven to be no worse than the globally optimal values of corresponding classical machine learning models. We empirically support our theoretical observation with a synthetic dataset as well as MNIST, CIFAR-10 and SVHN datasets. When compared to previous studies with strong over-parameterization assumptions, the results in this paper do not require over-parameterization, and instead show the gradual effects of over-parameterization as consequences of general results."
                    },
                    {
                        "title": "Generalization in Deep Learning",
                        "abstract": "This paper provides theoretical insights into why and how deep learning can generalize well, despite its large capacity, complexity, possible algorithmic instability, nonrobustness, and sharp minima, responding to an open question in the literature. We also discuss approaches to provide non-vacuous generalization guarantees for deep learning. Based on theoretical observations, we propose new open problems and discuss the limitations of our results."
                    },
                    {
                        "title": "Elimination of All Bad Local Minima in Deep Learning",
                        "abstract": "In this paper, we theoretically prove that adding one special neuron per output unit eliminates all suboptimal local minima of any deep neural network, for multi-class classification, binary classification, and regression with an arbitrary loss function, under practical assumptions. At every local minimum of any deep neural network with these added neurons, the set of parameters of the original neural network (without added neurons) is guaranteed to be a global minimum of the original neural network. The effects of the added neurons are proven to automatically vanish at every local minimum. Moreover, we provide a novel theoretical characterization of a failure mode of eliminating suboptimal local minima via an additional theorem and several examples. This paper also introduces a novel proof technique based on the perturbable gradient basis (PGB) necessary condition of local minima, which provides new insight into the elimination of local minima and is applicable to analyze various models and transformations of objective functions beyond the elimination of local minima."
                    },
                    {
                        "title": "Robustness Implies Generalization via Data-Dependent Generalization Bounds",
                        "abstract": "This paper proves that robustness implies generalization via data-dependent generalization bounds. As a result, robustness and generalization are shown to be connected closely in a data-dependent manner. Our bounds improve previous bounds in two directions, to solve an open problem that has seen little development since 2010. The first is to reduce the dependence on the covering number. The second is to remove the dependence on the hypothesis space. We present several examples, including ones for lasso and deep learning, in which our bounds are provably preferable. The experiments on real-world data and theoretical models demonstrate near-exponential improvements in various situations. To achieve these improvements, we do not require additional assumptions on the unknown distribution; instead, we only incorporate an observable and computable property of the training samples. A key technical innovation is an improved concentration bound for multinomial random variables that is of independent interest beyond robustness and generalization."
                    },
                    {
                        "title": "Every Local Minimum Value is the Global Minimum Value of Induced Model in Non-convex Machine Learning",
                        "abstract": "For nonconvex optimization in machine learning, this article proves that every local minimum achieves the globally optimal value of the perturbable gradient basis model at any differentiable point. As a result, nonconvex machine learning is theoretically as supported as convex machine learning with a handcrafted basis in terms of the loss at differentiable local minima, except in the case when a preference is given to the handcrafted basis over the perturbable gradient basis. The proofs of these results are derived under mild assumptions. Accordingly, the proven results are directly applicable to many machine learning models, including practical deep neural networks, without any modification of practical methods. Furthermore, as special cases of our general results, this article improves or complements several state-of-the-art theoretical results on deep neural networks, deep residual networks, and overparameterized deep neural networks with a unified proof technique and novel geometric insights. A special case of our results also contributes to the theoretical foundation of representation learning."
                    },
                    {
                        "title": "Deep Semi-Random Features for Nonlinear Function Approximation",
                        "abstract": "We propose semi-random features for nonlinear function approximation. The flexibility of semi-random feature lies between the fully adjustable units in deep learning and the random features used in kernel methods. For one hidden layer models with semi-random features, we prove with no unrealistic assumptions that the model classes contain an arbitrarily good function as the width increases (universality), and despite non-convexity, we can find such a good function (optimization theory) that generalizes to unseen new data (generalization bound). For deep models, with no unrealistic assumptions, we prove universal approximation ability, a lower bound on approximation error, a partial optimization guarantee, and a generalization bound. Depending on the problems, the generalization bound of deep semi-random features can be exponentially better than the known bounds of deep ReLU nets; our generalization error bound can be independent of the depth, the number of trainable weights as well as the input dimensionality. In experiments, we show that semi-random features can match the performance of neural networks by using slightly more units, and it outperforms random features by using significantly fewer units. Moreover, we introduce a new implicit ensemble method by using semi-random features."
                    },
                    {
                        "title": "How Does Information Bottleneck Help Deep Learning?",
                        "abstract": "Numerous deep learning algorithms have been inspired by and understood via the notion of information bottleneck, where unnecessary information is (often implicitly) minimized while task-relevant information is maximized. However, a rigorous argument for justifying why it is desirable to control information bottlenecks has been elusive. In this paper, we provide the first rigorous learning theory for justifying the benefit of information bottleneck in deep learning by mathematically relating information bottleneck to generalization errors. Our theory proves that controlling information bottleneck is one way to control generalization errors in deep learning, although it is not the only or necessary way. We investigate the merit of our new mathematical findings with experiments across a range of architectures and learning settings. In many cases, generalization errors are shown to correlate with the degree of information bottleneck: i.e., the amount of the unnecessary information at hidden layers. This paper provides a theoretical foundation for current and future methods through the lens of information bottleneck. Our new generalization bounds scale with the degree of information bottleneck, unlike the previous bounds that scale with the number of parameters, VC dimension, Rademacher complexity, stability or robustness. Our code is publicly available at: https://github.com/xu-ji/information-bottleneck"
                    },
                    {
                        "title": "Generalization in Machine Learning via Analytical Learning Theory",
                        "abstract": "This paper introduces a novel measure-theoretic theory for machine learning that does not require statistical assumptions. Based on this theory, a new regularization method in deep learning is derived and shown to outperform previous methods in CIFAR-10, CIFAR-100, and SVHN. Moreover, the proposed theory provides a theoretical basis for a family of practically successful regularization methods in deep learning. We discuss several consequences of our results on one-shot learning, representation learning, deep learning, and curriculum learning. Unlike statistical learning theory, the proposed learning theory analyzes each problem instance individually via measure theory, rather than a set of problem instances via statistics. As a result, it provides different types of results and insights when compared to statistical learning theory."
                    }
                ]
            },
            "e894cc30-1574-4a19-87f6-e2dd902a0cf9": {
                "pk": "e894cc30-1574-4a19-87f6-e2dd902a0cf9",
                "name": "Lidong Bing",
                "collaborators": [
                    "Wai Lam",
                    "Xin Li",
                    "Piji Li",
                    "Wenxuan Zhang",
                    "Zihao Wang",
                    "Lu Xu",
                    "Qingyu Tan",
                    "Hwee Tou Ng",
                    "Haitian Sun",
                    "William W. Cohen"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Information Extraction",
                    "Summarization"
                ],
                "publications": [
                    {
                        "title": "Semi-Supervised Learning with Declaratively Specified Entropy Constraints",
                        "abstract": "We propose a technique for declaratively specifying strategies for semi-supervised learning (SSL). The proposed method can be used to specify ensembles of semi-supervised learning, as well as agreement constraints and entropic regularization constraints between these learners, and can be used to model both well-known heuristics such as co-training and novel domain-specific heuristics. In addition to representing individual SSL heuristics, we show that multiple heuristics can also be automatically combined using Bayesian optimization methods. We show consistent improvements on a suite of well-studied SSL benchmarks, including a new state-of-the-art result on a difficult relation extraction task."
                    },
                    {
                        "title": "Exploiting BERT for End-to-End Aspect-based Sentiment Analysis",
                        "abstract": "In this paper, we investigate the modeling power of contextualized embeddings from pre-trained language models, e.g. BERT, on the E2E-ABSA task. Specifically, we build a series of simple yet insightful neural baselines to deal with E2E-ABSA. The experimental results show that even with a simple linear classification layer, our BERT-based architecture can outperform state-of-the-art works. Besides, we also standardize the comparative study by consistently utilizing a hold-out validation dataset for model selection, which is largely ignored by previous works. Therefore, our work can serve as a BERT-based benchmark for E2E-ABSA."
                    },
                    {
                        "title": "Deep Recurrent Generative Decoder for Abstractive Text Summarization",
                        "abstract": "We propose a new framework for abstractive text summarization based on a sequence-to-sequence oriented encoder-decoder model equipped with a deep recurrent generative decoder (DRGN).   Latent structure information implied in the target summaries is learned based on a recurrent latent random model for improving the summarization quality.   Neural variational inference is employed to address the intractable posterior inference for the recurrent latent variables.   Abstractive summaries are generated based on both the generative latent variables and the discriminative deterministic states.   Extensive experiments on some benchmark datasets in different languages show that DRGN achieves improvements over the state-of-the-art methods."
                    },
                    {
                        "title": "Reader-Aware Multi-Document Summarization: An Enhanced Model and The First Dataset",
                        "abstract": "We investigate the problem of reader-aware multi-document summarization (RA-MDS) and introduce a new dataset for this problem. To tackle RA-MDS, we extend a variational auto-encodes (VAEs) based MDS framework by jointly considering news documents and reader comments. To conduct evaluation for summarization performance, we prepare a new dataset. We describe the methods for data collection, aspect annotation, and summary writing as well as scrutinizing by experts. Experimental results show that reader comments can improve the summarization performance, which also demonstrates the usefulness of the proposed dataset. The annotated dataset for RA-MDS is available online."
                    },
                    {
                        "title": "A Unified Model for Opinion Target Extraction and Target Sentiment Prediction",
                        "abstract": "Target-based sentiment analysis involves opinion target extraction and target sentiment classification. However, most of the existing works usually studied one of these two sub-tasks alone, which hinders their practical use. This paper aims to solve the complete task of target-based sentiment analysis in an end-to-end fashion, and presents a novel unified model which applies a unified tagging scheme. Our framework involves two stacked recurrent neural networks: The upper one predicts the unified tags to produce the final output results of the primary target-based sentiment analysis; The lower one performs an auxiliary target boundary prediction aiming at guiding the upper network to improve the performance of the primary task. To explore the inter-task dependency, we propose to explicitly model the constrained transitions from target boundaries to target sentiment polarities. We also propose to maintain the sentiment consistency within an opinion target via a gate mechanism which models the relation between the features for the current word and the previous word. We conduct extensive experiments on three benchmark datasets and our framework achieves consistently superior results."
                    },
                    {
                        "title": "Persona-Aware Tips Generation",
                        "abstract": "Tips, as a compacted and concise form of reviews, were paid less attention by researchers. In this paper, we investigate the task of tips generation by considering the `persona' information which captures the intrinsic language style of the users or the different characteristics of the product items. In order to exploit the persona information, we propose a framework based on adversarial variational auto-encoders (aVAE) for persona modeling from the historical tips and reviews of users and items. The latent variables from aVAE are regarded as persona embeddings. Besides representing persona using the latent embeddings, we design a persona memory for storing the persona related words for users and items. Pointer Network is used to retrieve persona wordings from the memory when generating tips. Moreover, the persona embeddings are used as latent factors by a rating prediction component to predict the sentiment of a user over an item. Finally, the persona embeddings and the sentiment information are incorporated into a recurrent neural networks based tips generation component. Extensive experimental results are reported and discussed to elaborate the peculiarities of our framework."
                    },
                    {
                        "title": "Actor-Critic based Training Framework for Abstractive Summarization",
                        "abstract": "We present a training framework for neural abstractive summarization based on actor-critic approaches from reinforcement learning. In the traditional neural network based methods, the objective is only to maximize the likelihood of the predicted summaries, no other assessment constraints are considered, which may generate low-quality summaries or even incorrect sentences. To alleviate this problem, we employ an actor-critic framework to enhance the training procedure. For the actor, we employ the typical attention based sequence-to-sequence (seq2seq) framework as the policy network for summary generation. For the critic, we combine the maximum likelihood estimator with a well designed global summary quality estimator which is a neural network based binary classifier aiming to make the generated summaries indistinguishable from the human-written ones. Policy gradient method is used to conduct the parameter learning. An alternating training strategy is proposed to conduct the joint training of the actor and critic models. Extensive experiments on some benchmark datasets in different languages show that our framework achieves improvements over the state-of-the-art methods."
                    },
                    {
                        "title": "Transformation Networks for Target-Oriented Sentiment Classification",
                        "abstract": "Target-oriented sentiment classification aims at classifying sentiment polarities over individual opinion targets in a sentence. RNN with attention seems a good fit for the characteristics of this task, and indeed it achieves the state-of-the-art performance. After re-examining the drawbacks of attention mechanism and the obstacles that block CNN to perform well in this classification task, we propose a new model to overcome these issues. Instead of attention, our model employs a CNN layer to extract salient features from the transformed word representations originated from a bi-directional RNN layer. Between the two layers, we propose a component to generate target-specific representations of words in the sentence, meanwhile incorporate a mechanism for preserving the original contextual information from the RNN layer. Experiments show that our model achieves a new state-of-the-art performance on a few benchmarks."
                    },
                    {
                        "title": "Learning Span-Level Interactions for Aspect Sentiment Triplet Extraction",
                        "abstract": "Aspect Sentiment Triplet Extraction (ASTE) is the most recent subtask of ABSA which outputs triplets of an aspect target, its associated sentiment, and the corresponding opinion term. Recent models perform the triplet extraction in an end-to-end manner but heavily rely on the interactions between each target word and opinion word. Thereby, they cannot perform well on targets and opinions which contain multiple words. Our proposed span-level approach explicitly considers the interaction between the whole spans of targets and opinions when predicting their sentiment relation. Thus, it can make predictions with the semantics of whole spans, ensuring better sentiment consistency. To ease the high computational cost caused by span enumeration, we propose a dual-channel span pruning strategy by incorporating supervision from the Aspect Term Extraction (ATE) and Opinion Term Extraction (OTE) tasks. This strategy not only improves computational efficiency but also distinguishes the opinion and target spans more properly. Our framework simultaneously achieves strong performance for the ASTE as well as ATE and OTE tasks. In particular, our analysis shows that our span-level approach achieves more significant improvements over the baselines on triplets with multi-word targets or opinions."
                    },
                    {
                        "title": "Enhancing Cross-lingual Prompting with Dual Prompt Augmentation",
                        "abstract": "Prompting shows promising results in few-shot scenarios. However, its strength for multilingual/cross-lingual problems has not been fully exploited. Zhao and Sch\\\"utze (2021) made initial explorations in this direction by presenting that cross-lingual prompting outperforms cross-lingual finetuning. In this paper, we conduct an empirical exploration on the effect of each component in cross-lingual prompting and derive language-agnostic Universal Prompting, which helps alleviate the discrepancies between source-language training and target-language inference. Based on this, we propose DPA, a dual prompt augmentation framework, aiming at relieving the data scarcity issue in few-shot cross-lingual prompting. Notably, for XNLI, our method achieves 46.54% with only 16 English training examples per class, significantly better than 34.99% of finetuning. Our code is available at https://github.com/DAMO-NLP-SG/DPA."
                    },
                    {
                        "title": "Towards Integration of Discriminability and Robustness for Document-Level Relation Extraction",
                        "abstract": "Document-level relation extraction (DocRE) predicts relations for entity pairs that rely on long-range context-dependent reasoning in a document. As a typical multi-label classification problem, DocRE faces the challenge of effectively distinguishing a small set of positive relations from the majority of negative ones. This challenge becomes even more difficult to overcome when there exists a significant number of annotation errors in the dataset. In this work, we aim to achieve better integration of both the discriminability and robustness for the DocRE problem. Specifically, we first design an effective loss function to endow high discriminability to both probabilistic outputs and internal representations. We innovatively customize entropy minimization and supervised contrastive learning for the challenging multi-label and long-tailed learning problems. To ameliorate the impact of label errors, we equipped our method with a novel negative label sampling strategy to strengthen the model robustness. In addition, we introduce two new data regimes to mimic more realistic scenarios with annotation errors and evaluate our sampling strategy. Experimental results verify the effectiveness of each component and show that our method achieves new state-of-the-art results on the DocRED dataset, its recently cleaned version, Re-DocRED, and the proposed data regimes."
                    },
                    {
                        "title": "Easy-to-Hard Learning for Information Extraction",
                        "abstract": "Information extraction (IE) systems aim to automatically extract structured information, such as named entities, relations between entities, and events, from unstructured texts. While most existing work addresses a particular IE task, universally modeling various IE tasks with one model has achieved great success recently. Despite their success, they employ a one-stage learning strategy, i.e., directly learning to extract the target structure given the input text, which contradicts the human learning process. In this paper, we propose a unified easy-to-hard learning framework consisting of three stages, i.e., the easy stage, the hard stage, and the main stage, for IE by mimicking the human learning process. By breaking down the learning process into multiple stages, our framework facilitates the model to acquire general IE task knowledge and improve its generalization ability. Extensive experiments across four IE tasks demonstrate the effectiveness of our framework. We achieve new state-of-the-art results on 13 out of 17 datasets. Our code is available at \\url{https://github.com/DAMO-NLP-SG/IE-E2H}."
                    },
                    {
                        "title": "Better Sampling of Negatives for Distantly Supervised Named Entity Recognition",
                        "abstract": "Distantly supervised named entity recognition (DS-NER) has been proposed to exploit the automatically labeled training data instead of human annotations. The distantly annotated datasets are often noisy and contain a considerable number of false negatives. The recent approach uses a weighted sampling approach to select a subset of negative samples for training. However, it requires a good classifier to assign weights to the negative samples. In this paper, we propose a simple and straightforward approach for selecting the top negative samples that have high similarities with all the positive samples for training. Our method achieves consistent performance improvements on four distantly supervised NER datasets. Our analysis also shows that it is critical to differentiate the true negatives from the false negatives."
                    },
                    {
                        "title": "mPMR: A Multilingual Pre-trained Machine Reader at Scale",
                        "abstract": "We present multilingual Pre-trained Machine Reader (mPMR), a novel method for multilingual machine reading comprehension (MRC)-style pre-training. mPMR aims to guide multilingual pre-trained language models (mPLMs) to perform natural language understanding (NLU) including both sequence classification and span extraction in multiple languages. To achieve cross-lingual generalization when only source-language fine-tuning data is available, existing mPLMs solely transfer NLU capability from a source language to target languages. In contrast, mPMR allows the direct inheritance of multilingual NLU capability from the MRC-style pre-training to downstream tasks. Therefore, mPMR acquires better NLU capability for target languages. mPMR also provides a unified solver for tackling cross-lingual span extraction and sequence classification, thereby enabling the extraction of rationales to explain the sentence-pair classification process."
                    },
                    {
                        "title": "Is GPT-4 a Good Data Analyst?",
                        "abstract": "As large language models (LLMs) have demonstrated their powerful capabilities in plenty of domains and tasks, including context understanding, code generation, language generation, data storytelling, etc., many data analysts may raise concerns if their jobs will be replaced by artificial intelligence (AI). This controversial topic has drawn great attention in public. However, we are still at a stage of divergent opinions without any definitive conclusion. Motivated by this, we raise the research question of \"is GPT-4 a good data analyst?\" in this work and aim to answer it by conducting head-to-head comparative studies. In detail, we regard GPT-4 as a data analyst to perform end-to-end data analysis with databases from a wide range of domains. We propose a framework to tackle the problems by carefully designing the prompts for GPT-4 to conduct experiments. We also design several task-specific evaluation metrics to systematically compare the performance between several professional human data analysts and GPT-4. Experimental results show that GPT-4 can achieve comparable performance to humans. We also provide in-depth discussions about our results to shed light on further studies before reaching the conclusion that GPT-4 can replace data analysts."
                    },
                    {
                        "title": "Towards Benchmarking and Improving the Temporal Reasoning Capability of Large Language Models",
                        "abstract": "Reasoning about time is of fundamental importance. Many facts are time-dependent. For example, athletes change teams from time to time, and different government officials are elected periodically. Previous time-dependent question answering (QA) datasets tend to be biased in either their coverage of time spans or question types. In this paper, we introduce a comprehensive probing dataset \\tempreason to evaluate the temporal reasoning capability of large language models. Our dataset includes questions of three temporal reasoning levels. In addition, we also propose a novel learning framework to improve the temporal reasoning capability of large language models, based on temporal span extraction and time-sensitive reinforcement learning. We conducted experiments in closed book QA, open book QA, and reasoning QA settings and demonstrated the effectiveness of our approach. Our code and data are released on https://github.com/DAMO-NLP-SG/TempReason."
                    },
                    {
                        "title": "Once Upon a $\\textit{Time}$ in $\\textit{Graph}$: Relative-Time Pretraining for Complex Temporal Reasoning",
                        "abstract": "Our physical world is constantly evolving over time, rendering challenges for pre-trained language models to understand and reason over the temporal contexts of texts. Existing work focuses on strengthening the direct association between a piece of text and its time-stamp. However, the knowledge-time association is usually insufficient for the downstream tasks that require reasoning over temporal dependencies between knowledge. In this work, we make use of the underlying nature of time, all temporally-scoped sentences are strung together through a one-dimensional time axis, and suggest creating a graph structure based on the relative placements of events along the time axis. Inspired by the graph view, we propose RemeMo ($\\underline{Re}$lative Ti$\\underline{me}$ $\\underline{Mo}$deling), which explicitly connects all temporally-scoped facts by modeling the time relations between any two sentences. Experimental results show that RemeMo outperforms the baseline T5 on multiple temporal question answering datasets under various settings. Further analysis suggests that RemeMo is especially good at modeling long-range complex temporal dependencies. We release our code and pre-trained checkpoints at $\\href{https://github.com/DAMO-NLP-SG/RemeMo}{\\text{this url}}$."
                    },
                    {
                        "title": "Towards Robust Temporal Reasoning of Large Language Models via a Multi-Hop QA Dataset and Pseudo-Instruction Tuning",
                        "abstract": "Knowledge in the real world is being updated constantly. However, it is costly to frequently update large language models (LLMs). Therefore, it is crucial for LLMs to understand the concept of temporal knowledge. However, prior works on temporal question answering (TQA) did not emphasize multi-answer and multi-hop types of temporal reasoning. In this paper, we propose a complex temporal question-answering dataset Complex-TR that focuses on multi-answer and multi-hop temporal reasoning. Besides, we also propose a novel data augmentation strategy to improve the complex temporal reasoning capability and robustness of LLMs. We conducted experiments on multiple temporal QA datasets. Experimental results show that our method is able to improve LLMs' performance on temporal QA benchmarks by significant margins. Our code and data are released at: https://github.com/nusnlp/complex-tr."
                    },
                    {
                        "title": "Salience Estimation with Multi-Attention Learning for Abstractive Text Summarization",
                        "abstract": "Attention mechanism plays a dominant role in the sequence generation models and has been used to improve the performance of machine translation and abstractive text summarization. Different from neural machine translation, in the task of text summarization, salience estimation for words, phrases or sentences is a critical component, since the output summary is a distillation of the input text. Although the typical attention mechanism can conduct text fragment selection from the input text conditioned on the decoder states, there is still a gap to conduct direct and effective salience detection. To bring back direct salience estimation for summarization with neural networks, we propose a Multi-Attention Learning framework which contains two new attention learning components for salience estimation: supervised attention learning and unsupervised attention learning. We regard the attention weights as the salience information, which means that the semantic units with large attention value will be more important. The context information obtained based on the estimated salience is incorporated with the typical attention mechanism in the decoder to conduct summary generation. Extensive experiments on some benchmark datasets in different languages demonstrate the effectiveness of the proposed framework for the task of abstractive summarization."
                    },
                    {
                        "title": "Video-LLaMA: An Instruction-tuned Audio-Visual Language Model for Video Understanding",
                        "abstract": "We present Video-LLaMA a multi-modal framework that empowers Large Language Models (LLMs) with the capability of understanding both visual and auditory content in the video. Video-LLaMA bootstraps cross-modal training from the frozen pre-trained visual and audio encoders and the frozen LLMs. Unlike previous works that complement LLMs to process the visual or audio signals only, Video-LLaMA enables video comprehension by tackling two challenges: (1) capturing the temporal changes in visual scenes, (2) integrating audio-visual signals. To counter the first challenge, we propose a Video Q-former to assemble a pre-trained image encoder into our video encoder and introduce a video-to-text generation task to learn video-language correspondence. For the second challenge, we leverage ImageBind, a universal embedding model aligning multiple modalities, as the pre-trained audio encoder and introduce an Audio Q-former on top of ImageBind to learn reasonable auditory query embeddings for the LLM module. To align the output of both visual and audio encoders with LLM's embedding space, we first train Video-LLaMA on massive video/image-caption pairs and then tune our model with visual-instruction datasets of moderate amount but higher quality. We found Video-LLaMA shows the ability to perceive and comprehend video content and generate meaningful responses grounded in the visual and auditory information presented in the videos."
                    }
                ]
            }
        }
    },
    "2406.05658": {
        "paper_data": {
            "title": "Visual Prompt Tuning in Null Space for Continual Learning",
            "url": "http://arxiv.org/abs/2406.05658v3",
            "arxiv_id": "2406.05658",
            "authors": [
                "Yue Lu",
                "Shizhou Zhang",
                "De Cheng",
                "Yinghui Xing",
                "Nannan Wang",
                "Peng Wang",
                "Yanning Zhang"
            ],
            "abstract": "Existing prompt-tuning methods have demonstrated impressive performances in continual learning (CL), by selecting and updating relevant prompts in the vision-transformer models. On the contrary, this paper aims to learn each task by tuning the prompts in the direction orthogonal to the subspace spanned by previous tasks' features, so as to ensure no interference on tasks that have been learned to overcome catastrophic forgetting in CL. However, different from the orthogonal projection in the traditional CNN architecture, the prompt gradient orthogonal projection in the ViT architecture shows completely different and greater challenges, i.e., 1) the high-order and non-linear self-attention operation; 2) the drift of prompt distribution brought by the LayerNorm in the transformer block. Theoretically, we have finally deduced two consistency conditions to achieve the prompt gradient orthogonal projection, which provide a theoretical guarantee of eliminating interference on previously learned knowledge via the self-attention mechanism in visual prompt tuning. In practice, an effective null-space-based approximation solution has been proposed to implement the prompt gradient orthogonal projection. Extensive experimental results demonstrate the effectiveness of anti-forgetting on four class-incremental benchmarks with diverse pre-trained baseline models, and our approach achieves superior performances to state-of-the-art methods. Our code is available at https://github.com/zugexiaodui/VPTinNSforCL.",
            "introduction": "   1 Introduction  Continual learning (CL) is crucial for AI models to adapt to the ever-changing environment by learning sequentially arrived data, where the catastrophic forgetting is the key challenge\u00a0[21, 28]. Recently, prompt tuning-based continual learning methods\u00a0[40, 32, 34, 43, 10, 22, 38, 45, 20, 12, 18] have been attracting increasing attention due to their impressive performances in the CL field. Existing prompt tuning-based works tackle the downstream continual learning problem by selecting and updating relevant prompts, which is encoded with full task-specific knowledge while exploiting the general knowledge of the pre-trained ViTs\u00a0[40, 39].   On the contrary, this paper aims to learn each task by tuning the prompts in the direction orthogonal to the subspace spanned by previous tasks\u2019 features, so as to ensure no interference with tasks that have been learned to overcome catastrophic forgetting in CL. It is worth noting that forgetting can be theoretically resolved by gradient orthogonal projection methods\u00a0[42, 31, 36, 44], which have been extensively explored especially when adapting CNN models. Nevertheless, it remains a huge gap to introduce the orthogonal projection-based methods of CNNs to visual prompt tuning due to the following challenges: 1) the high-order and non-linear self-attention operation; 2) the drift of prompt distribution brought by the LayerNorm in the transformer block. For the linear operation in convolution or fully-connected layers, the output features of old tasks can remain unchanged by updating the weights in the orthogonal subspace of previous input features. While for self-attention, three linear transformations are employed on input tokens, followed by high-order and non-linear operations for the self-attention interaction of tokens. It makes the relationship between the update of prompts and the output image tokens much more complex, far exceeding mere linearity.   In this work, we theoretically deduced two consistency conditions to achieve the prompt gradient orthogonal projection, which provide a theoretical guarantee of eliminating interference on previously learned knowledge via the self-attention mechanism in visual prompt tuning. To be concrete, we firstly take the full self-attention and LayerNorm into consideration and derive a strict condition for eliminating the interference through a comprehensive analysis of the forward propagation of the ViT layer. Then we further propose to convert the condition of self-attention into its two sufficient conditions, which enables us to address the challenge of high order and nonlinearity. Thirdly, we propose a constraint of invariant prompt distribution that removes the obstacle to the final simplification of the conditions brought by the LayerNorm. The consistency conditions reveal that if the prompt update can be orthogonal to (1) the normalized previous input image tokens projected with the second-order qkv-transformation matrices of the pre-trained model, and (2) the activated attention map generated by image queries and prompt keys, the interference in visual prompt tuning can be eliminated theoretically.   In practice, based on the proposed consistency conditions, an effective null-space-based approximation solution\u00a0[36] has been proposed to implement the prompt gradient orthogonal projection, while the invariant prompt distribution constraint is implemented by incorporating a loss function which penalizes the drifting of prompt distribution over sequential tasks. We validate our Null-Space Projection for Prompts (NSP2) approach on extensive class-incremental benchmarks: 10- and 20-split CIFAR-100, 10-split ImageNet-R\u00a0[39] and 10-split DomainNet\u00a0[38], with the sequential",
            "references": [
                {
                    "title": "InfLoRA: Interference-Free Low-Rank Adaptation for Continual Learning",
                    "abstract": "Continual learning requires the model to learn multiple tasks sequentially. In continual learning, the model should possess the ability to maintain its performance on old tasks (stability) and the ability to adapt to new tasks continuously (plasticity). Recently, parameter-efficient finetuning (PEFT), which involves freezing a pretrained model and injecting a small number of learnable parameters to adapt to downstream tasks, has gained increasing popularity in continual learning. Although existing continual learning methods based on PEFT have demonstrated superior performance compared to those not based on PEFT, most of them do not consider how to eliminate the interference of the new task on the old tasks, which inhibits the model from making a good tradeoff between stability and plasticity. In this work, we propose a new PEFT method, called interference-free low-rank adaptation (InfLoRA), for continual learning. InfLoRA injects a small number of parameters to reparameterize the pretrained weights and shows that fine-tuning these injected parameters is equivalent to fine-tuning the pretrained weights within a subspace. Furthermore, InfLoRA designs this subspace to eliminate the interference of the new task on the old tasks, making a good tradeoff between stability and plasticity. Experimental results show that InfLoRA outperforms existing state-of-the-art continual learning methods on multiple datasets. Code is available at https://github.com/liangyanshuo/InfLoRA."
                },
                {
                    "title": "Evolving Parameterized Prompt Memory for Continual Learning",
                    "abstract": "Recent studies have demonstrated the potency of leveraging prompts in Transformers for continual learning (CL). Nevertheless, employing a discrete key-prompt bottleneck can lead to selection mismatches and inappropriate prompt associations during testing. Furthermore, this approach hinders adaptive prompting due to the lack of shareability among nearly identical instances at more granular level. To address these challenges, we introduce the Evolving Parameterized Prompt Memory (EvoPrompt), a novel method involving adaptive and continuous prompting attached to pre-trained Vision Transformer (ViT), conditioned on specific instance. We formulate a continuous prompt function as a neural bottleneck and encode the collection of prompts on network weights. We establish a paired prompt memory system consisting of a stable reference and a flexible working prompt memory. Inspired by linear mode connectivity, we progressively fuse the working prompt memory and reference prompt memory during inter-task periods, resulting in continually evolved prompt memory. This fusion involves aligning functionally equivalent prompts using optimal transport and aggregating them in parameter space with an adjustable bias based on prompt node attribution. Additionally, to enhance backward compatibility, we propose compositional classifier initialization, which leverages prior prototypes from pre-trained models to guide the initialization of new classifiers in a subspace-aware manner. Comprehensive experiments validate that our approach achieves state-of-the-art performance in both class and domain incremental learning scenarios."
                },
                {
                    "title": "Expandable Subspace Ensemble for Pre-Trained Model-Based Class-Incremental Learning",
                    "abstract": "Class-Incremental Learning (CIL) requires a learning system to continually learn new classes without forgetting. Despite the strong performance of Pre-Trained Models (PTMs) in CIL, a critical issue persists: learning new classes often results in the overwriting of old ones. Excessive modification of the network causes forgetting, while minimal adjustments lead to an inadequate fit for new classes. As a result, it is desired to figure out a way of efficient model updating without harming former knowledge. In this paper, we propose ExpAndable Subspace Ensem-ble (EASE) for PTM-based CIL. To enable model updating without conflict, we train a distinct lightweight adapter module for each new task, aiming to create task-specific subspaces. These adapters span a high-dimensional feature space, enabling joint decision-making across multiple sub-spaces. As data evolves, the expanding subspaces render the old class classifiers incompatible with new-stage spaces. Correspondingly, we design a semantic-guided prototype complement strategy that synthesizes old classes' new features without using any old class instance. Extensive experiments on seven benchmark datasets verify EASE's state-of-the-art performance. Code is available at: https://github.com/sun-hailong/CVPR24-Ease"
                },
                {
                    "title": "Consistent Prompting for Rehearsal-Free Continual Learning",
                    "abstract": "Continual learning empowers models to adapt autonomously to the ever-changing environment or data streams without forgetting old knowledge. Prompt-based approaches are built on frozen pre-trained models to learn the task-specific prompts and classifiers efficiently. Existing prompt-based methods are inconsistent between training and testing, limiting their effectiveness. Two types of inconsistency are revealed. Test predictions are made from all classifiers while training only focuses on the current task classifier without holistic alignment, leading to Classifier inconsistency. Prompt inconsistency indicates that the prompt selected during testing may not correspond to the one associated with this task during training. In this paper, we propose a novel prompt-based method, Consistent Prompting (CPrompt), for more aligned training and testing. Specifically, all existing classifiers are exposed to prompt training, resulting in classifier consistency learning. In addition, prompt consistency learning is proposed to enhance prediction robustness and boost prompt selection accuracy. Our Consistent Prompting surpasses its prompt-based counterparts and achieves state-of-the-art performance on multiple continual learning benchmarks. Detailed analysis shows that improvements come from more consistent training and testing. Our code is available at https://github.com/Zhanxin-Gao/CPrompt."
                },
                {
                    "title": "OVOR: OnePrompt with Virtual Outlier Regularization for Rehearsal-Free Class-Incremental Learning",
                    "abstract": "Recent works have shown that by using large pre-trained models along with learnable prompts, rehearsal-free methods for class-incremental learning (CIL) settings can achieve superior performance to prominent rehearsal-based ones. Rehearsal-free CIL methods struggle with distinguishing classes from different tasks, as those are not trained together. In this work we propose a regularization method based on virtual outliers to tighten decision boundaries of the classifier, such that confusion of classes among different tasks is mitigated. Recent prompt-based methods often require a pool of task-specific prompts, in order to prevent overwriting knowledge of previous tasks with that of the new task, leading to extra computation in querying and composing an appropriate prompt from the pool. This additional cost can be eliminated, without sacrificing accuracy, as we reveal in the paper. We illustrate that a simplified prompt-based method can achieve results comparable to previous state-of-the-art (SOTA) methods equipped with a prompt pool, using much less learnable parameters and lower inference cost. Our regularization method has demonstrated its compatibility with different prompt-based methods, boosting those previous SOTA rehearsal-free CIL methods' accuracy on the ImageNet-R and CIFAR-100 benchmarks. Our source code is available at https://github.com/jpmorganchase/ovor."
                },
                {
                    "title": "Hierarchical Decomposition of Prompt-Based Continual Learning: Rethinking Obscured Sub-optimality",
                    "abstract": "Prompt-based continual learning is an emerging direction in leveraging pre-trained knowledge for downstream continual learning, and has almost reached the performance pinnacle under supervised pre-training. However, our empirical research reveals that the current strategies fall short of their full potential under the more realistic self-supervised pre-training, which is essential for handling vast quantities of unlabeled data in practice. This is largely due to the difficulty of task-specific knowledge being incorporated into instructed representations via prompt parameters and predicted by uninstructed representations at test time. To overcome the exposed sub-optimality, we conduct a theoretical analysis of the continual learning objective in the context of pre-training, and decompose it into hierarchical components: within-task prediction, task-identity inference, and task-adaptive prediction. Following these empirical and theoretical insights, we propose Hierarchical Decomposition (HiDe-)Prompt, an innovative approach that explicitly optimizes the hierarchical components with an ensemble of task-specific prompts and statistics of both uninstructed and instructed representations, further with the coordination of a contrastive regularization strategy. Our extensive experiments demonstrate the superior performance of HiDe-Prompt and its robustness to pre-training paradigms in continual learning (e.g., up to 15.01% and 9.61% lead on Split CIFAR-100 and Split ImageNet-R, respectively). Our code is available at \\url{https://github.com/thu-ml/HiDe-Prompt}."
                },
                {
                    "title": "Generating Instance-level Prompts for Rehearsal-free Continual Learning",
                    "abstract": "We introduce Domain-Adaptive Prompt (DAP), a novel method for continual learning using Vision Transformers (ViT). Prompt-based continual learning has recently gained attention due to its rehearsal-free nature. Currently, the prompt pool, which is suggested by prompt-based continual learning, is key to effectively exploiting the frozen pretrained ViT backbone in a sequence of tasks. However, we observe that the use of a prompt pool creates a domain scalability problem between pre-training and continual learning. This problem arises due to the inherent encoding of group-level instructions within the prompt pool. To address this problem, we propose DAP, a pool-free approach that generates a suitable prompt in an instance-level manner at inference time. We optimize an adaptive prompt generator that creates instance-specific fine-grained instructions required for each input, enabling enhanced model plasticity and reduced forgetting. Our experiments on seven datasets with varying degrees of domain similarity to ImageNet demonstrate the superiority of DAP over state-of-the-art prompt-based methods. Code is publicly available at https://github.com/naver-ai/dap-cl."
                },
                {
                    "title": "Introducing Language Guidance in Prompt-based Continual Learning",
                    "abstract": "Continual Learning aims to learn a single model on a sequence of tasks without having access to data from previous tasks. The biggest challenge in the domain still remains catastrophic forgetting: a loss in performance on seen classes of earlier tasks. Some existing methods rely on an expensive replay buffer to store a chunk of data from previous tasks. This, while promising, becomes expensive when the number of tasks becomes large or data can not be stored for privacy reasons. As an alternative, prompt-based methods have been proposed that store the task information in a learnable prompt pool. This prompt pool instructs a frozen image encoder on how to solve each task. While the model faces a disjoint set of classes in each task in this setting, we argue that these classes can be encoded to the same embedding space of a pre-trained language encoder. In this work, we propose Language Guidance for Prompt-based Continual Learning (LGCL) as a plug-in for prompt-based methods. LGCL is model agnostic and introduces language guidance at the task level in the prompt pool and at the class level on the output feature of the vision encoder. We show with extensive experimentation that LGCL consistently improves the performance of prompt-based continual learning methods to set a new state-of-the art. LGCL achieves these performance improvements without needing any additional learnable parameters."
                },
                {
                    "title": "When Prompt-based Incremental Learning Does Not Meet Strong Pretraining",
                    "abstract": "Incremental learning aims to overcome catastrophic forgetting when learning deep networks from sequential tasks. With impressive learning efficiency and performance, prompt-based methods adopt a fixed backbone to sequential tasks by learning task-specific prompts. However, existing prompt-based methods heavily rely on strong pretraining (typically trained on ImageNet-21k), and we find that their models could be trapped if the potential gap between the pretraining task and unknown future tasks is large. In this work, we develop a learnable Adaptive Prompt Generator (APG). The key is to unify the prompt retrieval and prompt learning processes into a learnable prompt generator. Hence, the whole prompting process can be optimized to reduce the negative effects of the gap between tasks effectively. To make our APG avoid learning ineffective knowledge, we maintain a knowledge pool to regularize APG with the feature distribution of each class. Extensive experiments show that our method significantly outperforms advanced methods in exemplar-free incremental learning without (strong) pretraining. Besides, under strong pretraining, our method also has comparable performance to existing prompt-based models, showing that our method can still benefit from pretraining. Codes can be found at https://github.com/TOM-tym/APG"
                },
                {
                    "title": "On the Effectiveness of LayerNorm Tuning for Continual Learning in Vision Transformers",
                    "abstract": "State-of-the-art rehearsal-free continual learning methods exploit the peculiarities of Vision Transformers to learn task-specific prompts, drastically reducing catastrophic forgetting. However, there is a tradeoff between the number of learned parameters and the performance, making such models computationally expensive. In this work, we aim to reduce this cost while maintaining competitive performance. We achieve this by revisiting and extending a simple transfer learning idea: learning task-specific normalization layers. Specifically, we tune the scale and bias parameters of LayerNorm for each continual learning task, selecting them at inference time based on the similarity between task-specific keys and the output of the pre-trained model. To make the classifier robust to incorrect selection of parameters during inference, we introduce a two-stage training procedure, where we first optimize the task-specific parameters and then train the classifier with the same selection procedure of the inference time. Experiments on ImageNet-R and CIFAR-100 show that our method achieves results that are either superior or on par with the state of the art while being computationally cheaper.1"
                },
                {
                    "title": "RanPAC: Random Projections and Pre-trained Models for Continual Learning",
                    "abstract": "Continual learning (CL) aims to incrementally learn different tasks (such as classification) in a non-stationary data stream without forgetting old ones. Most CL works focus on tackling catastrophic forgetting under a learning-from-scratch paradigm. However, with the increasing prominence of foundation models, pre-trained models equipped with informative representations have become available for various downstream requirements. Several CL methods based on pre-trained models have been explored, either utilizing pre-extracted features directly (which makes bridging distribution gaps challenging) or incorporating adaptors (which may be subject to forgetting). In this paper, we propose a concise and effective approach for CL with pre-trained models. Given that forgetting occurs during parameter updating, we contemplate an alternative approach that exploits training-free random projectors and class-prototype accumulation, which thus bypasses the issue. Specifically, we inject a frozen Random Projection layer with nonlinear activation between the pre-trained model's feature representations and output head, which captures interactions between features with expanded dimensionality, providing enhanced linear separability for class-prototype-based CL. We also demonstrate the importance of decorrelating the class-prototypes to reduce the distribution disparity when using pre-trained representations. These techniques prove to be effective and circumvent the problem of forgetting for both class- and domain-incremental continual learning. Compared to previous methods applied to pre-trained ViT-B/16 models, we reduce final error rates by between 20% and 62% on seven class-incremental benchmarks, despite not using any rehearsal memory. We conclude that the full potential of pre-trained models for simple, effective, and fast CL has not hitherto been fully tapped. Code is at github.com/RanPAC/RanPAC."
                },
                {
                    "title": "Rethinking Gradient Projection Continual Learning: Stability/Plasticity Feature Space Decoupling",
                    "abstract": "Continual learning aims to incrementally learn novel classes over time, while not forgetting the learned knowledge. Recent studies have found that learning would not forget if the updated gradient is orthogonal to the feature space. However, previous approaches require the gradient to be fully orthogonal to the whole feature space, leading to poor plasticity, as the feasible gradient direction becomes narrow when the tasks continually come, i.e., feature space is unlimitedly expanded. In this paper, we propose a space decoupling (SD) algorithm to decouple the feature space into a pair of complementary subspaces, i.e., the stability space $\\mathcal{I}$ and the plasticity space $\\mathcal{R}. \\mathcal{I}$ is established by conducting space intersection between the historic and current feature space, and thus $\\mathcal{I}$ contains more task-shared bases. $\\mathcal{R}$ is constructed by seeking the orthogonal complementary subspace of $T$ and thus $\\mathcal{R}$ mainly contains task-specific bases. By putting distinguishing constraints on $\\mathcal{R}$ and $\\mathcal{I}$, our method achieves a better balance between stability and plasticity. Extensive experiments are conducted by applying SD to gradient projection baselines, and show SD is model-agnostic and achieves SOTA results on publicly available datasets."
                },
                {
                    "title": "AttriCLIP: A Non-Incremental Learner for Incremental Knowledge Learning",
                    "abstract": "Continual learning aims to enable a model to incrementally learn knowledge from sequentially arrived data. Previous works adopt the conventional classification architecture, which consists of a feature extractor and a classifier. The feature extractor is shared across sequentially arrived tasks or classes, but one specific group of weights of the classifier corresponding to one new class should be incrementally expanded. Consequently, the parameters of a continual learner gradually increase. Moreover, as the classifier contains all historical arrived classes, a certain size of the memory is usually required to store rehearsal data to mitigate classifier bias and catastrophic forgetting. In this paper, we propose a non-incremental learner, named AttriCLIP, to incrementally extract knowledge of new classes or tasks. Specifically, AttriCLIP is built upon the pre-trained visual-language model CLIP. Its image encoder and text encoder are fixed to extract features from both images and text. Text consists of a category name and a fixed number of learnable parameters which are selected from our designed attribute word bank and serve as attributes. As we compute the visual and textual similarity for classification, AttriCLIP is a non-incremental learner. The attribute prompts, which encode the common knowledge useful for classification, can effectively mitigate the catastrophic forgetting and avoid constructing a replay memory. We evaluate our AttriCLIP and compare it with CLIP-based and previous state-of-the-art continual learning methods in realistic settings with domain-shift and long-sequence learning. The results show that our method performs favorably against previous state-of-the-arts. The implementation code will be available at https://gitee.com/mindspore/models/tree/master/research/cv/AttriCLIP."
                },
                {
                    "title": "A Unified Continual Learning Framework with General Parameter-Efficient Tuning",
                    "abstract": "The \"pre-training \u2192 downstream adaptation\" presents both new opportunities and challenges for Continual Learning (CL). Although the recent state-of-the-art in CL is achieved through Parameter-Efficient-Tuning (PET) adaptation paradigm, only prompt has been explored, limiting its application to Transformers only. In this paper, we position prompting as one instantiation of PET, and propose a unified CL framework with general PET, dubbed as Learning-Accumulation-Ensemble (LAE). PET, e.g., using Adapter, LoRA, or Prefix, can adapt a pre-trained model to downstream tasks with fewer parameters and resources. Given a PET method, our LAE framework incorporates it for CL with three novel designs. 1) Learning: the pre-trained model adapts to the new task by tuning an online PET module, along with our adaptation speed calibration to align different PET modules, 2) Accumulation: the task-specific knowledge learned by the online PET module is accumulated into an offline PET module through momentum update, 3) Ensemble: During inference, we respectively construct two experts with online/offline PET modules (which are favored by the novel/historical tasks) for prediction ensemble. We show that LAE is compatible with a battery of PET methods and gains strong CL capability. For example, LAE with Adaptor PET surpasses the prior state-of-the-art by 1.3% and 3.6% in last-incremental accuracy on CIFAR100 and ImageNet-R datasets, respectively. Code is available at https://github.com/gqk/LAE."
                },
                {
                    "title": "Steering Prototypes with Prompt-tuning for Rehearsal-free Continual Learning",
                    "abstract": "In the context of continual learning, prototypes\u2014as representative class embeddings\u2014offer advantages in memory conservation and the mitigation of catastrophic forgetting. However, challenges related to semantic drift and prototype interference persist. In this study, we introduce the Contrastive Prototypical Prompt (CPP) approach. Through task-specific prompt-tuning, underpinned by a contrastive learning objective, we effectively address both aforementioned challenges. Our evaluations on four challenging class-incremental benchmarks reveal that CPP achieves a significant 4% to 6% improvement over state-of-the-art methods. Importantly, CPP operates without a rehearsal buffer and narrows the performance divergence between continual and offline joint-learning, suggesting an innovative scheme for Transformer-based continual learning systems1."
                },
                {
                    "title": "SLCA: Slow Learner with Classifier Alignment for Continual Learning on a Pre-trained Model",
                    "abstract": "The goal of continual learning is to improve the performance of recognition models in learning sequentially arrived data. Although most existing works are established on the premise of learning from scratch, growing efforts have been devoted to incorporating the benefits of pre-training. However, how to adaptively exploit the pre-trained knowledge for each incremental task while maintaining its generalizability remains an open question. In this work, we present an extensive analysis for continual learning on a pre-trained model (CLPM), and attribute the key challenge to a progressive overfitting problem. Observing that selectively reducing the learning rate can almost resolve this issue in the representation layer, we propose a simple but extremely effective approach named Slow Learner with Classifier Alignment (SLCA), which further improves the classification layer by modeling the class-wise distributions and aligning the classification layers in a post-hoc fashion. Across a variety of scenarios, our proposal provides substantial improvements for CLPM (e.g., up to 49.76%, 50.05%, 44.69% and 40.16% on Split CIFAR-100, Split ImageNet-R, Split CUB-200 and Split Cars-196, respectively), and thus outperforms state-of-the-art approaches by a large margin. Based on such a strong baseline, critical factors and promising directions are analyzed in-depth to facilitate subsequent research. Code has been made available at: https://github.com/GengDavid/SLCA."
                },
                {
                    "title": "\u00c0-la-carte Prompt Tuning (APT): Combining Distinct Data Via Composable Prompting",
                    "abstract": "We introduce \u00c0-la-carte Prompt Tuning (APT), a transformer-based scheme to tune prompts on distinct data so that they can be arbitrarily composed at inference time. The individual prompts can be trained in isolation, possibly on different devices, at different times, and on different distributions or domains. Furthermore each prompt only contains information about the subset of data it was exposed to during training. During inference, models can be assembled based on arbitrary selections of data sources, which we call \u00e0-la-carte learning. \u00c0-la-carte learning enables constructing bespoke models specific to each user's individual access rights and preferences. We can add or remove information from the model by simply adding or removing the corresponding prompts without retraining from scratch. We demonstrate that \u00e0-la-carte built models achieve accuracy within 5% of models trained on the union of the respective sources, with comparable cost in terms of training and inference time. For the continual learning benchmarks Split CIFAR- 100 and CORe50, we achieve state-of-the-art performance."
                },
                {
                    "title": "Reproducible Scaling Laws for Contrastive Language-Image Learning",
                    "abstract": "Scaling up neural networks has led to remarkable performance across a wide range of tasks. Moreover, performance often follows reliable scaling laws as a function of training set size, model size, and compute, which offers valuable guidance as large-scale experiments are becoming increasingly expensive. However, previous work on scaling laws has primarily used private data & models or focused on uni-modal language or vision learning. To address these limitations, we investigate scaling laws for contrastive language-image pre-training (CLIP) with the public LAION dataset and the open-source OpenCLIP repository. Our large-scale experiments involve models trained on up to two billion image-text pairs and identify power law scaling for multiple downstream tasks including zero-shot classification, retrieval, linear probing, and end-to-end fine-tuning. We find that the training distribution plays a key role in scaling laws as the OpenAI and OpenCLIP models exhibit different scaling behavior despite identical model architectures and similar training recipes. We open-source our evaluation workflow and all models, including the largest public CLIP models, to ensure reproducibility and make scaling laws research more accessible. Source code and instructions to reproduce this study is available at https://github.eom/LAION-AI/sealing-laws-openelip."
                },
                {
                    "title": "Isolation and Impartial Aggregation: A Paradigm of Incremental Learning without Interference",
                    "abstract": "This paper focuses on the prevalent stage interference and stage performance imbalance of incremental learning. To avoid obvious stage learning bottlenecks, we propose a new incremental learning framework, which leverages a series of stage-isolated classifiers to perform the learning task at each stage, without interference from others. To be concrete, to aggregate multiple stage classifiers as a uniform one impartially, we first introduce a temperature-controlled energy metric for indicating the confidence score levels of the stage classifiers. We then propose an anchor-based energy self-normalization strategy to ensure the stage classifiers work at the same energy level. Finally, we design a voting-based inference augmentation strategy for robust inference. The proposed method is rehearsal-free and can work for almost all incremental learning scenarios. We evaluate the proposed method on four large datasets. Extensive results demonstrate the superiority of the proposed method in setting up new state-of-the-art overall performance. Code is available at https://github.com/iamwangyabin/ESN."
                },
                {
                    "title": "CODA-Prompt: COntinual Decomposed Attention-Based Prompting for Rehearsal-Free Continual Learning",
                    "abstract": "Computer vision models suffer from a phenomenon known as catastrophic forgetting when learning novel concepts from continuously shifting training data. Typical solutions for this continual learning problem require extensive rehearsal of previously seen data, which increases memory costs and may violate data privacy. Recently, the emergence of large-scale pre-trained vision transformer models has enabled prompting approaches as an alternative to data-rehearsal. These approaches rely on a key-query mechanism to generate prompts and have been found to be highly resistant to catastrophic forgetting in the well-established rehearsal-free continual learning setting. However, the key mechanism of these methods is not trained end-to-end with the task sequence. Our experiments show that this leads to a reduction in their plasticity, hence sacrificing new task accuracy, and inability to benefit from expanded parameter capacity. We instead propose to learn a set of prompt components which are assembled with input-conditioned weights to produce input-conditioned prompts, resulting in a novel attention-based end-to-end key-query scheme. Our experiments show that we outperform the current SOTA method DualPrompt on established benchmarks by as much as 4.5% in average final accuracy. We also outperform the state of art by as much as 4.4% accuracy on a continual learning benchmark which contains both class-incremental and domain-incremental task shifts, corresponding to many practical settings. Our code is available at https://github.com/GT-RIPL/CODA-Prompt"
                },
                {
                    "title": "S-Prompts Learning with Pre-trained Transformers: An Occam's Razor for Domain Incremental Learning",
                    "abstract": "State-of-the-art deep neural networks are still struggling to address the catastrophic forgetting problem in continual learning. In this paper, we propose one simple paradigm (named as S-Prompting) and two concrete approaches to highly reduce the forgetting degree in one of the most typical continual learning scenarios, i.e., domain increment learning (DIL). The key idea of the paradigm is to learn prompts independently across domains with pre-trained transformers, avoiding the use of exemplars that commonly appear in conventional methods. This results in a win-win game where the prompting can achieve the best for each domain. The independent prompting across domains only requests one single cross-entropy loss for training and one simple K-NN operation as a domain identifier for inference. The learning paradigm derives an image prompt learning approach and a novel language-image prompt learning approach. Owning an excellent scalability (0.03% parameter increase per domain), the best of our approaches achieves a remarkable relative improvement (an average of about 30%) over the best of the state-of-the-art exemplar-free methods for three standard DIL tasks, and even surpasses the best of them relatively by about 6% in average when they use exemplars. Source code is available at \\url{https://github.com/iamwangyabin/S-Prompts}."
                },
                {
                    "title": "Balancing Stability and Plasticity through Advanced Null Space in Continual Learning",
                    "abstract": "Continual learning is a learning paradigm that learns tasks sequentially with resources constraints, in which the key challenge is stability-plasticity dilemma, i.e., it is uneasy to simultaneously have the stability to prevent catastrophic forgetting of old tasks and the plasticity to learn new tasks well. In this paper, we propose a new continual learning approach, Advanced Null Space (AdNS), to balance the stability and plasticity without storing any old data of previous tasks. Specifically, to obtain better stability, AdNS makes use of low-rank approximation to obtain a novel null space and projects the gradient onto the null space to prevent the interference on the past tasks. To control the generation of the null space, we introduce a non-uniform constraint strength to further reduce forgetting. Furthermore, we present a simple but effective method, intra-task distillation, to improve the performance of the current task. Finally, we theoretically find that null space plays a key role in plasticity and stability, respectively. Experimental results show that the proposed method can achieve better performance compared to state-of-the-art continual learning approaches."
                },
                {
                    "title": "DualPrompt: Complementary Prompting for Rehearsal-free Continual Learning",
                    "abstract": "Continual learning aims to enable a single model to learn a sequence of tasks without catastrophic forgetting. Top-performing methods usually require a rehearsal buffer to store past pristine examples for experience replay, which, however, limits their practical value due to privacy and memory constraints. In this work, we present a simple yet effective framework, DualPrompt, which learns a tiny set of parameters, called prompts, to properly instruct a pre-trained model to learn tasks arriving sequentially without buffering past examples. DualPrompt presents a novel approach to attach complementary prompts to the pre-trained backbone, and then formulates the objective as learning task-invariant and task-specific\"instructions\". With extensive experimental validation, DualPrompt consistently sets state-of-the-art performance under the challenging class-incremental setting. In particular, DualPrompt outperforms recent advanced continual learning methods with relatively large buffer sizes. We also introduce a more challenging benchmark, Split ImageNet-R, to help generalize rehearsal-free continual learning research. Source code is available at https://github.com/google-research/l2p."
                },
                {
                    "title": "Visual Prompt Tuning",
                    "abstract": "The current modus operandi in adapting pre-trained models involves updating all the backbone parameters, ie, full fine-tuning. This paper introduces Visual Prompt Tuning (VPT) as an efficient and effective alternative to full fine-tuning for large-scale Transformer models in vision. Taking inspiration from recent advances in efficiently tuning large language models, VPT introduces only a small amount (less than 1% of model parameters) of trainable parameters in the input space while keeping the model backbone frozen. Via extensive experiments on a wide variety of downstream recognition tasks, we show that VPT achieves significant performance gains compared to other parameter efficient tuning protocols. Most importantly, VPT even outperforms full fine-tuning in many cases across model capacities and training data scales, while reducing per-task storage cost."
                },
                {
                    "title": "Learning to Prompt for Continual Learning",
                    "abstract": "The mainstream paradigm behind continual learning has been to adapt the model parameters to non-stationary data distributions, where catastrophic forgetting is the central challenge. Typical methods rely on a rehearsal buffer or known task identity at test time to retrieve learned knowl-edge and address forgetting, while this work presents a new paradigm for continual learning that aims to train a more succinct memory system without accessing task identity at test time. Our method learns to dynamically prompt (L2P) a pre-trained model to learn tasks sequen-tially under different task transitions. In our proposed framework, prompts are small learnable parameters, which are maintained in a memory space. The objective is to optimize prompts to instruct the model prediction and ex-plicitly manage task-invariant and task-specific knowledge while maintaining model plasticity. We conduct comprehen-sive experiments under popular image classification bench-marks with different challenging continual learning set-tings, where L2P consistently outperforms prior state-of-the-art methods. Surprisingly, L2P achieves competitive results against rehearsal-based methods even without a re-hearsal buffer and is directly applicable to challenging task-agnostic continual learning. Source code is available at https://github.com/google-research/12p."
                },
                {
                    "title": "Flattening Sharpness for Dynamic Gradient Projection Memory Benefits Continual Learning",
                    "abstract": "The backpropagation networks are notably susceptible to catastrophic forgetting, where networks tend to forget previously learned skills upon learning new ones. To address such the 'sensitivity-stability' dilemma, most previous efforts have been contributed to minimizing the empirical risk with different parameter regularization terms and episodic memory, but rarely exploring the usages of the weight loss landscape. In this paper, we investigate the relationship between the weight loss landscape and sensitivity-stability in the continual learning scenario, based on which, we propose a novel method, Flattening Sharpness for Dynamic Gradient Projection Memory (FS-DGPM). In particular, we introduce a soft weight to represent the importance of each basis representing past tasks in GPM, which can be adaptively learned during the learning process, so that less important bases can be dynamically released to improve the sensitivity of new skill learning. We further introduce Flattening Sharpness (FS) to reduce the generalization gap by explicitly regulating the flatness of the weight loss landscape of all seen tasks. As demonstrated empirically, our proposed method consistently outperforms baselines with the superior ability to learn new skills while alleviating forgetting effectively."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "ImageNet-21K Pretraining for the Masses",
                    "abstract": "ImageNet-1K serves as the primary dataset for pretraining deep learning models for computer vision tasks. ImageNet-21K dataset, which is bigger and more diverse, is used less frequently for pretraining, mainly due to its complexity, low accessibility, and underestimation of its added value. This paper aims to close this gap, and make high-quality efficient pretraining on ImageNet-21K available for everyone. Via a dedicated preprocessing stage, utilization of WordNet hierarchical structure, and a novel training scheme called semantic softmax, we show that various models significantly benefit from ImageNet-21K pretraining on numerous datasets and tasks, including small mobile-oriented models. We also show that we outperform previous ImageNet-21K pretraining schemes for prominent new models like ViT and Mixer. Our proposed pretraining pipeline is efficient, accessible, and leads to SoTA reproducible results, from a publicly available dataset. The training code and pretrained models are available at: https://github.com/Alibaba-MIIL/ImageNet21K"
                },
                {
                    "title": "Gradient Projection Memory for Continual Learning",
                    "abstract": "The ability to learn continually without forgetting the past tasks is a desired attribute for artificial learning systems. Existing approaches to enable such learning in artificial neural networks usually rely on network growth, importance based weight update or replay of old data from the memory. In contrast, we propose a novel approach where a neural network learns new tasks by taking gradient steps in the orthogonal direction to the gradient subspaces deemed important for the past tasks. We find the bases of these subspaces by analyzing network representations (activations) after learning each task with Singular Value Decomposition (SVD) in a single shot manner and store them in the memory as Gradient Projection Memory (GPM). With qualitative and quantitative analyses, we show that such orthogonal gradient descent induces minimum to no interference with the past tasks, thereby mitigates forgetting. We evaluate our algorithm on diverse image classification datasets with short and long sequences of tasks and report better or on-par performance compared to the state-of-the-art approaches."
                },
                {
                    "title": "Training Networks in Null Space of Feature Covariance for Continual Learning",
                    "abstract": "In the setting of continual learning, a network is trained on a sequence of tasks, and suffers from catastrophic forgetting. To balance plasticity and stability of network in continual learning, in this paper, we propose a novel network training algorithm Adam-NSCL which sequentially optimizes network parameters in the null space of all previous tasks. We first propose two mathematical conditions respectively for achieving network stability and plasticity in continual learning. Based on them, the network training for sequential tasks without forgetting can be simply achieved by projecting the candidate parameter update into the approximate null space of all previous tasks in the network training process, where the candidate parameter update can be generated by Adam. The approximate null space can be derived by applying singular value decomposition to the un-centered covariance matrix of all input features of previous tasks for each linear layer. For efficiency, the uncentered covariance matrix can be incrementally computed after learning each task. We also empirically verify the rationality of the approximate null space at each linear layer. We apply our approach to training networks for continual learning on benchmark datasets of CIFAR-100 and TinyImageNet, and the results suggest that the proposed approach outperforms or matches the state-ot-the-art continual learning approaches."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Continual Learning in Low-rank Orthogonal Subspaces",
                    "abstract": "In continual learning (CL), a learner is faced with a sequence of tasks, arriving one after the other, and the goal is to remember all the tasks once the continual learning experience is finished. The prior art in CL uses episodic memory, parameter regularization or extensible network structures to reduce interference among tasks, but in the end, all the approaches learn different tasks in a joint vector space. We believe this invariably leads to interference among different tasks. We propose to learn tasks in different (low-rank) vector subspaces that are kept orthogonal to each other in order to minimize interference. Further, to keep the gradients of different tasks coming from these subspaces orthogonal to each other, we learn isometric mappings by posing network training as an optimization problem over the Stiefel manifold. To the best of our understanding, we report, for the first time, strong results over experience-replay baseline with and without memory on standard classification benchmarks in continual learning. The code is made publicly available."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "AutoAugment: Learning Augmentation Strategies From Data",
                    "abstract": "Data augmentation is an effective technique for improving the accuracy of modern image classifiers. However, current data augmentation implementations are manually designed. In this paper, we describe a simple procedure called AutoAugment to automatically search for improved data augmentation policies. In our implementation, we have designed a search space where a policy consists of many sub-policies, one of which is randomly chosen for each image in each mini-batch. A sub-policy consists of two operations, each operation being an image processing function such as translation, rotation, or shearing, and the probabilities and magnitudes with which the functions are applied. We use a search algorithm to find the best policy such that the neural network yields the highest validation accuracy on a target dataset. Our method achieves state-of-the-art accuracy on CIFAR-10, CIFAR-100, SVHN, and ImageNet (without additional data). On ImageNet, we attain a Top-1 accuracy of 83.5% which is 0.4% better than the previous record of 83.1%. On CIFAR-10, we achieve an error rate of 1.5%, which is 0.6% better than the previous state-of-the-art. Augmentation policies we find are transferable between datasets. The policy learned on ImageNet transfers well to achieve significant improvements on other datasets, such as Oxford Flowers, Caltech-101, Oxford-IIT Pets, FGVC Aircraft, and Stanford Cars."
                },
                {
                    "title": "Moment Matching for Multi-Source Domain Adaptation",
                    "abstract": "Conventional unsupervised domain adaptation (UDA) assumes that training data are sampled from a single domain. This neglects the more practical scenario where training data are collected from multiple sources, requiring multi-source domain adaptation. We make three major contributions towards addressing this problem. First, we collect and annotate by far the largest UDA dataset, called DomainNet, which contains six domains and about 0.6 million images distributed among 345 categories, addressing the gap in data availability for multi-source UDA research. Second, we propose a new deep learning approach, Moment Matching for Multi-Source Domain Adaptation (M3SDA), which aims to transfer knowledge learned from multiple labeled source domains to an unlabeled target domain by dynamically aligning moments of their feature distributions. Third, we provide new theoretical insights specifically for moment matching approaches in both single and multiple source domain adaptation. Extensive experiments are conducted to demonstrate the power of our new dataset in benchmarking state-of-the-art multi-source domain adaptation methods, as well as the advantage of our proposed model. Dataset and Code are available at http://ai.bu.edu/M3SDA/"
                },
                {
                    "title": "Layer Normalization",
                    "abstract": "Training state-of-the-art, deep neural networks is computationally expensive. One way to reduce the training time is to normalize the activities of the neurons. A recently introduced technique called batch normalization uses the distribution of the summed input to a neuron over a mini-batch of training cases to compute a mean and variance which are then used to normalize the summed input to that neuron on each training case. This significantly reduces the training time in feed-forward neural networks. However, the effect of batch normalization is dependent on the mini-batch size and it is not obvious how to apply it to recurrent neural networks. In this paper, we transpose batch normalization into layer normalization by computing the mean and variance used for normalization from all of the summed inputs to the neurons in a layer on a single training case. Like batch normalization, we also give each neuron its own adaptive bias and gain which are applied after the normalization but before the non-linearity. Unlike batch normalization, layer normalization performs exactly the same computation at training and test times. It is also straightforward to apply to recurrent neural networks by computing the normalization statistics separately at each time step. Layer normalization is very effective at stabilizing the hidden state dynamics in recurrent networks. Empirically, we show that layer normalization can substantially reduce the training time compared with previously published techniques."
                },
                {
                    "title": "High Performance Convolutional Neural Networks for Document Processing",
                    "abstract": "Convolutional neural networks (CNNs) are well known for producing state-of-the-art recognizers for document processing [1]. However, they can be difficult to implement and are usually slower than traditional multi-layer perceptrons (MLPs). We present three novel approaches to speeding up CNNs: a) unrolling convolution, b) using BLAS (basic linear algebra subroutines), and c) using GPUs (graphic processing units). Unrolled convolution converts the processing in each convolutional layer (both forward-propagation and back-propagation) into a matrix-matrix product. The matrix-matrix product representation of CNNs makes their implementation as easy as MLPs. BLAS is used to efficiently compute matrix products on the CPU. We also present a pixel shader based GPU implementation of CNNs. Results on character recognition problems indicate that unrolled convolution with BLAS produces a dramatic 2.4X\u22123.0X speedup. The GPU implementation is even faster and produces a 3.1X\u22124.1X speedup."
                },
                {
                    "title": "Connectionist models of recognition memory: constraints imposed by learning and forgetting functions.",
                    "abstract": "Multilayer connectionist models of memory based on the encoder model using the backpropagation learning rule are evaluated. The models are applied to standard recognition memory procedures in which items are studied sequentially and then tested for retention. Sequential learning in these models leads to 2 major problems. First, well-learned information is forgotten rapidly as new information is learned. Second, discrimination between studied items and new items either decreases or is nonmonotonic as a function of learning. To address these problems, manipulations of the network within the multilayer model and several variants of the multilayer model were examined, including a model with prelearned memory and a context model, but none solved the problems. The problems discussed provide limitations on connectionist models applied to human memory and in tasks where information to be learned is not all available during learning."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively implement prompt gradient orthogonal projection in visual prompt tuning to mitigate catastrophic forgetting in continual learning?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of continual learning (CL) in machine learning, as it addresses the significant challenge of catastrophic forgetting when models are trained on sequential data. By developing methods that allow AI models to retain previously learned knowledge while adapting to new tasks, this research could lead to more robust and flexible AI systems. The implications extend to various practical applications, such as autonomous systems, personalized AI, and any domain where models must continuously learn from new data without losing prior knowledge. This work could inspire future research to explore more sophisticated approaches to CL, potentially leading to breakthroughs in how AI systems learn and adapt over time.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe problem is complex due to several challenges inherent in the self-attention mechanism of transformers, which involves high-order and non-linear operations that complicate the relationship between prompt updates and output features. Naive approaches may fail because they do not account for the intricate interactions within the self-attention layers, leading to interference with previously learned tasks. Additionally, the presence of LayerNorm introduces further complications by altering the distribution of prompts, making it difficult to maintain orthogonality with respect to past tasks. Overcoming these technical and theoretical obstacles requires a deep understanding of the underlying mechanisms of visual prompt tuning and the development of innovative solutions to ensure effective gradient orthogonal projection.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on linear operations in CNNs, which do not translate directly to the complexities of self-attention in transformers. The lack of attention to the non-linear interactions and the effects of LayerNorm has left a gap in the literature regarding orthogonal projection methods in visual prompt tuning. Existing solutions have not adequately addressed the need for maintaining orthogonality in a high-dimensional, non-linear space, which is essential for preventing catastrophic forgetting. This work differs by providing a theoretical framework that establishes consistency conditions for prompt updates, thus offering a novel approach that integrates insights from both the self-attention mechanism and the constraints imposed by LayerNorm.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves deriving two consistency conditions for achieving prompt gradient orthogonal projection in visual prompt tuning."
            }
        },
        "author_data": {
            "2cfcb3de-ffb5-4356-a58f-eedf71a5ce28": {
                "pk": "2cfcb3de-ffb5-4356-a58f-eedf71a5ce28",
                "name": "Yue Lu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "ce7907dc-abf4-441d-8926-f8633fd84471": {
                "pk": "ce7907dc-abf4-441d-8926-f8633fd84471",
                "name": "Shizhou Zhang",
                "collaborators": [
                    "Yanning Zhang",
                    "Peng Wang",
                    "Guoqiang Liang",
                    "Yinghui Xing",
                    "De Cheng",
                    "Bingliang Jiao",
                    "Yifei Yang",
                    "Lingyan Ran",
                    "Congqi Cao",
                    "Xin Zhang"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Image Processing",
                    "Anomaly Detection"
                ],
                "publications": [
                    {
                        "title": "DMAT: A Dynamic Mask-Aware Transformer for Human De-occlusion",
                        "abstract": "Human de-occlusion, which aims to infer the appearance of invisible human parts from an occluded image, has great value in many human-related tasks, such as person re-id, and intention inference. To address this task, this paper proposes a dynamic mask-aware transformer (DMAT), which dynamically augments information from human regions and weakens that from occlusion. First, to enhance token representation, we design an expanded convolution head with enlarged kernels, which captures more local valid context and mitigates the influence of surrounding occlusion. To concentrate on the visible human parts, we propose a novel dynamic multi-head human-mask guided attention mechanism through integrating multiple masks, which can prevent the de-occluded regions from assimilating to the background. Besides, a region upsampling strategy is utilized to alleviate the impact of occlusion on interpolated images. During model learning, an amodal loss is developed to further emphasize the recovery effect of human regions, which also refines the model's convergence. Extensive experiments on the AHP dataset demonstrate its superior performance compared to recent state-of-the-art methods."
                    },
                    {
                        "title": "Adaptive Graph Convolutional Networks for Weakly Supervised Anomaly Detection in Videos",
                        "abstract": "For weakly supervised anomaly detection, most existing work is limited to the problem of inadequate video representation due to the inability of modeling long-term contextual information. To solve this, we propose a novel weakly supervised adaptive graph convolutional network (WAGCN) to model the complex contextual relationship among video segments. By which, we fully consider the influence of other video segments on the current one when generating the anomaly probability score for each segment. Firstly, we combine the temporal consistency as well as feature similarity of video segments to construct a global graph, which makes full use of the association information among spatial-temporal features of anomalous events in videos. Secondly, we propose a graph learning layer in order to break the limitation of setting topology manually, which can extract graph adjacency matrix based on data adaptively and effectively. Extensive experiments on two public datasets (i.e., UCF-Crime dataset and ShanghaiTech dataset) demonstrate the effectiveness of our approach which achieves state-of-the-art performance."
                    },
                    {
                        "title": "CrossDiff: Exploring Self-Supervised Representation of Pansharpening via Cross-Predictive Diffusion Model",
                        "abstract": "Fusion of a panchromatic (PAN) image and corresponding multispectral (MS) image is also known as pansharpening, which aims to combine abundant spatial details of PAN and spectral information of MS. Due to the absence of high-resolution MS images, available deep-learning-based methods usually follow the paradigm of training at reduced resolution and testing at both reduced and full resolution. When taking original MS and PAN images as inputs, they always obtain sub-optimal results due to the scale variation. In this paper, we propose to explore the self-supervised representation of pansharpening by designing a cross-predictive diffusion model, named CrossDiff. It has two-stage training. In the first stage, we introduce a cross-predictive pretext task to pre-train the UNet structure based on conditional DDPM, while in the second stage, the encoders of the UNets are frozen to directly extract spatial and spectral features from PAN and MS, and only the fusion head is trained to adapt for pansharpening task. Extensive experiments show the effectiveness and superiority of the proposed model compared with state-of-the-art supervised and unsupervised methods. Besides, the cross-sensor experiments also verify the generalization ability of proposed self-supervised representation learners for other satellite's datasets. We will release our code for reproducibility."
                    },
                    {
                        "title": "Unsupervised Video Summarization with a Convolutional Attentive Adversarial Network",
                        "abstract": "With the explosive growth of video data, video summarization, which attempts to seek the minimum subset of frames while still conveying the main story, has become one of the hottest topics. Nowadays, substantial achievements have been made by supervised learning techniques, especially after the emergence of deep learning. However, it is extremely expensive and difficult to collect human annotation for large-scale video datasets. To address this problem, we propose a convolutional attentive adversarial network (CAAN), whose key idea is to build a deep summarizer in an unsupervised way. Upon the generative adversarial network, our overall framework consists of a generator and a discriminator. The former predicts importance scores for all frames of a video while the latter tries to distinguish the score-weighted frame features from original frame features. Specifically, the generator employs a fully convolutional sequence network to extract global representation of a video, and an attention-based network to output normalized importance scores. To learn the parameters, our objective function is composed of three loss functions, which can guide the frame-level importance score prediction collaboratively. To validate this proposed method, we have conducted extensive experiments on two public benchmarks SumMe and TVSum. The results show the superiority of our proposed method against other state-of-the-art unsupervised approaches. Our method even outperforms some published supervised approaches."
                    },
                    {
                        "title": "Attend to the Difference: Cross-Modality Person Re-identification via Contrastive Correlation",
                        "abstract": "The problem of cross-modality person re-identification has been receiving increasing attention recently, due to its practical significance. Motivated by the fact that human usually attend to the difference when they compare two similar objects, we propose a dual-path cross-modality feature learning framework which preserves intrinsic spatial strictures and attends to the difference of input cross-modality image pairs. Our framework is composed by two main components: a Dual-path Spatial-structure-preserving Common Space Network (DSCSN) and a Contrastive Correlation Network (CCN). The former embeds cross-modality images into a common 3D tensor space without losing spatial structures, while the latter extracts contrastive features by dynamically comparing input image pairs. Note that the representations generated for the input RGB and Infrared images are mutually dependant to each other. We conduct extensive experiments on two public available RGB-IR ReID datasets, SYSU-MM01 and RegDB, and our proposed method outperforms state-of-the-art algorithms by a large margin with both full and simplified evaluation modes."
                    },
                    {
                        "title": "Weakly Supervised Video Anomaly Detection Based on Cross-Batch Clustering Guidance",
                        "abstract": "Weakly supervised video anomaly detection (WSVAD) is a challenging task since only video-level labels are available for training. In previous studies, the discriminative power of the learned features is not strong enough, and the data imbalance resulting from the mini-batch training strategy is ignored. To address these two issues, we propose a novel WSVAD method based on cross-batch clustering guidance. To enhance the discriminative power of features, we propose a batch clustering based loss to encourage a clustering branch to generate distinct normal and abnormal clusters based on a batch of data. Meanwhile, we design a cross-batch learning strategy by introducing clustering results from previous mini-batches to reduce the impact of data imbalance. In addition, we propose to generate more accurate segment-level anomaly scores based on batch clustering guidance further improving the performance of WSVAD. Extensive experiments on two public datasets demonstrate the effectiveness of our approach."
                    },
                    {
                        "title": "Person Re-identification in Aerial Imagery",
                        "abstract": "Nowadays, with the rapid development of consumer Unmanned Aerial Vehicles (UAVs), visual surveillance by utilizing the UAV platform has been very attractive. Most of the research works for UAV captured visual data are mainly focused on the tasks of object detection and tracking. However, limited attention has been paid to the task of person Re-identification (ReID) which has been widely studied in ordinary surveillance cameras with fixed emplacements. In this paper, to facilitate the research of person ReID in aerial imagery, we collect a large scale airborne person ReID dataset named as Person ReID for Aerial Imagery (PRAI-1581), which consists of 39,461 images of 1581 person identities. The images of the dataset are shot by two DJI consumer UAVs flying at an altitude ranging from 20 to 60 meters above the ground, which covers most of the real UAV surveillance scenarios. In addition, we propose to utilize subspace pooling of convolution feature maps to represent the input person images. Our method can learn a discriminative and compact feature representation for ReID in aerial imagery and can be trained in an end-to-end fashion efficiently. We conduct extensive experiments on the proposed dataset and the experimental results demonstrate that re-identify persons in aerial imagery is a challenging problem, where our method performs favorably against state of the arts. Our dataset can be accessed via \\url{https://github.com/stormyoung/PRAI-1581}."
                    },
                    {
                        "title": "MS-DETR: Multispectral Pedestrian Detection Transformer with Loosely Coupled Fusion and Modality-Balanced Optimization",
                        "abstract": "Multispectral pedestrian detection is an important task for many around-the-clock applications, since the visible and thermal modalities can provide complementary information especially under low light conditions. Most of the available multispectral pedestrian detectors are based on non-end-to-end detectors, while in this paper, we propose MultiSpectral pedestrian DEtection TRansformer (MS-DETR), an end-to-end multispectral pedestrian detector, which extends DETR into the field of multi-modal detection. MS-DETR consists of two modality-specific backbones and Transformer encoders, followed by a multi-modal Transformer decoder, and the visible and thermal features are fused in the multi-modal Transformer decoder. To well resist the misalignment between multi-modal images, we design a loosely coupled fusion strategy by sparsely sampling some keypoints from multi-modal features independently and fusing them with adaptively learned attention weights. Moreover, based on the insight that not only different modalities, but also different pedestrian instances tend to have different confidence scores to final detection, we further propose an instance-aware modality-balanced optimization strategy, which preserves visible and thermal decoder branches and aligns their predicted slots through an instance-wise dynamic loss. Our end-to-end MS-DETR shows superior performance on the challenging KAIST, CVC-14 and LLVIP benchmark datasets. The source code is available at https://github.com/YinghuiXing/MS-DETR ."
                    },
                    {
                        "title": "Vehicle Re-identification in Aerial Imagery: Dataset and Approach",
                        "abstract": "In this work, we construct a large-scale dataset for vehicle re-identification (ReID), which contains 137k images of 13k vehicle instances captured by UAV-mounted cameras. To our knowledge, it is the largest UAV-based vehicle ReID dataset. To increase intra-class variation, each vehicle is captured by at least two UAVs at different locations, with diverse view-angles and flight-altitudes. We manually label a variety of vehicle attributes, including vehicle type, color, skylight, bumper, spare tire and luggage rack. Furthermore, for each vehicle image, the annotator is also required to mark the discriminative parts that helps them to distinguish this particular vehicle from others. Besides the dataset, we also design a specific vehicle ReID algorithm to make full use of the rich annotation information. It is capable of explicitly detecting discriminative parts for each specific vehicle and significantly outperforms the evaluated baselines and state-of-the-art vehicle ReID approaches."
                    },
                    {
                        "title": "Pre-train, Adapt and Detect: Multi-Task Adapter Tuning for Camouflaged Object Detection",
                        "abstract": "Camouflaged object detection (COD), aiming to segment camouflaged objects which exhibit similar patterns with the background, is a challenging task. Most existing works are dedicated to establishing specialized modules to identify camouflaged objects with complete and fine details, while the boundary can not be well located for the lack of object-related semantics. In this paper, we propose a novel ``pre-train, adapt and detect\" paradigm to detect camouflaged objects. By introducing a large pre-trained model, abundant knowledge learned from massive multi-modal data can be directly transferred to COD. A lightweight parallel adapter is inserted to adjust the features suitable for the downstream COD task. Extensive experiments on four challenging benchmark datasets demonstrate that our method outperforms existing state-of-the-art COD models by large margins. Moreover, we design a multi-task learning scheme for tuning the adapter to exploit the shareable knowledge across different semantic classes. Comprehensive experimental results showed that the generalization ability of our model can be substantially improved with multi-task adapter initialization on source tasks and multi-task adaptation on target tasks."
                    },
                    {
                        "title": "Dual Modality Prompt Tuning for Vision-Language Pre-Trained Model",
                        "abstract": "With the emergence of large pre-trained vison-language model like CLIP, transferable representations can be adapted to a wide range of downstream tasks via prompt tuning. Prompt tuning tries to probe the beneficial information for downstream tasks from the general knowledge stored in the pre-trained model. A recently proposed method named Context Optimization (CoOp) introduces a set of learnable vectors as text prompt from the language side. However, tuning the text prompt alone can only adjust the synthesized \"classifier\", while the computed visual features of the image encoder can not be affected , thus leading to sub-optimal solutions. In this paper, we propose a novel Dual-modality Prompt Tuning (DPT) paradigm through learning text and visual prompts simultaneously. To make the final image feature concentrate more on the target visual concept, a Class-Aware Visual Prompt Tuning (CAVPT) scheme is further proposed in our DPT, where the class-aware visual prompt is generated dynamically by performing the cross attention between text prompts features and image patch token embeddings to encode both the downstream task-related information and visual instance information. Extensive experimental results on 11 datasets demonstrate the effectiveness and generalization ability of the proposed method. Our code is available in https://github.com/fanrena/DPT."
                    },
                    {
                        "title": "Ground-to-Aerial Person Search: Benchmark Dataset and Approach",
                        "abstract": "In this work, we construct a large-scale dataset for Ground-to-Aerial Person Search, named G2APS, which contains 31,770 images of 260,559 annotated bounding boxes for 2,644 identities appearing in both of the UAVs and ground surveillance cameras. To our knowledge, this is the first dataset for cross-platform intelligent surveillance applications, where the UAVs could work as a powerful complement for the ground surveillance cameras. To more realistically simulate the actual cross-platform Ground-to-Aerial surveillance scenarios, the surveillance cameras are fixed about 2 meters above the ground, while the UAVs capture videos of persons at different location, with a variety of view-angles, flight attitudes and flight modes. Therefore, the dataset has the following unique characteristics: 1) drastic view-angle changes between query and gallery person images from cross-platform cameras; 2) diverse resolutions, poses and views of the person images under 9 rich real-world scenarios. On basis of the G2APS benchmark dataset, we demonstrate detailed analysis about current two-step and end-to-end person search methods, and further propose a simple yet effective knowledge distillation scheme on the head of the ReID network, which achieves state-of-the-art performances on both of the G2APS and the previous two public person search datasets, i.e., PRW and CUHK-SYSU. The dataset and source code available on \\url{https://github.com/yqc123456/HKD_for_person_search}."
                    },
                    {
                        "title": "Efficient Bilateral Cross-Modality Cluster Matching for Unsupervised Visible-Infrared Person ReID",
                        "abstract": "Unsupervised visible-infrared person re-identification (USL-VI-ReID) aims to match pedestrian images of the same identity from different modalities without annotations. Existing works mainly focus on alleviating the modality gap by aligning instance-level features of the unlabeled samples. However, the relationships between cross-modality clusters are not well explored. To this end, we propose a novel bilateral cluster matching-based learning framework to reduce the modality gap by matching cross-modality clusters. Specifically, we design a Many-to-many Bilateral Cross-Modality Cluster Matching (MBCCM) algorithm through optimizing the maximum matching problem in a bipartite graph. Then, the matched pairwise clusters utilize shared visible and infrared pseudo-labels during the model training. Under such a supervisory signal, a Modality-Specific and Modality-Agnostic (MSMA) contrastive learning framework is proposed to align features jointly at a cluster-level. Meanwhile, the cross-modality Consistency Constraint (CC) is proposed to explicitly reduce the large modality discrepancy. Extensive experiments on the public SYSU-MM01 and RegDB datasets demonstrate the effectiveness of the proposed method, surpassing state-of-the-art approaches by a large margin of 8.76% mAP on average."
                    },
                    {
                        "title": "Dual-Modal Prompting for Sketch-Based Image Retrieval",
                        "abstract": "Sketch-based image retrieval (SBIR) associates hand-drawn sketches with their corresponding realistic images. In this study, we aim to tackle two major challenges of this task simultaneously: i) zero-shot, dealing with unseen categories, and ii) fine-grained, referring to intra-category instance-level retrieval. Our key innovation lies in the realization that solely addressing this cross-category and fine-grained recognition task from the generalization perspective may be inadequate since the knowledge accumulated from limited seen categories might not be fully valuable or transferable to unseen target categories. Inspired by this, in this work, we propose a dual-modal prompting CLIP (DP-CLIP) network, in which an adaptive prompting strategy is designed. Specifically, to facilitate the adaptation of our DP-CLIP toward unpredictable target categories, we employ a set of images within the target category and the textual category label to respectively construct a set of category-adaptive prompt tokens and channel scales. By integrating the generated guidance, DP-CLIP could gain valuable category-centric insights, efficiently adapting to novel categories and capturing unique discriminative clues for effective retrieval within each target category. With these designs, our DP-CLIP outperforms the state-of-the-art fine-grained zero-shot SBIR method by 7.3% in Acc.@1 on the Sketchy dataset. Meanwhile, in the other two category-level zero-shot SBIR benchmarks, our method also achieves promising performance."
                    },
                    {
                        "title": "Cross-Platform Video Person ReID: A New Benchmark Dataset and Adaptation Approach",
                        "abstract": "In this paper, we construct a large-scale benchmark dataset for Ground-to-Aerial Video-based person Re-Identification, named G2A-VReID, which comprises 185,907 images and 5,576 tracklets, featuring 2,788 distinct identities. To our knowledge, this is the first dataset for video ReID under Ground-to-Aerial scenarios. G2A-VReID dataset has the following characteristics: 1) Drastic view changes; 2) Large number of annotated identities; 3) Rich outdoor scenarios; 4) Huge difference in resolution. Additionally, we propose a new benchmark approach for cross-platform ReID by transforming the cross-platform visual alignment problem into visual-semantic alignment through vision-language model (i.e., CLIP) and applying a parameter-efficient Video Set-Level-Adapter module to adapt image-based foundation model to video ReID tasks, termed VSLA-CLIP. Besides, to further reduce the great discrepancy across the platforms, we also devise the platform-bridge prompts for efficient visual feature alignment. Extensive experiments demonstrate the superiority of the proposed method on all existing video ReID datasets and our proposed G2A-VReID dataset."
                    },
                    {
                        "title": "Text-based Person Search in Full Images via Semantic-Driven Proposal Generation",
                        "abstract": "Finding target persons in full scene images with a query of text description has important practical applications in intelligent video surveillance.However, different from the real-world scenarios where the bounding boxes are not available, existing text-based person retrieval methods mainly focus on the cross modal matching between the query text descriptions and the gallery of cropped pedestrian images. To close the gap, we study the problem of text-based person search in full images by proposing a new end-to-end learning framework which jointly optimize the pedestrian detection, identification and visual-semantic feature embedding tasks. To take full advantage of the query text, the semantic features are leveraged to instruct the Region Proposal Network to pay more attention to the text-described proposals. Besides, a cross-scale visual-semantic embedding mechanism is utilized to improve the performance. To validate the proposed method, we collect and annotate two large-scale benchmark datasets based on the widely adopted image-based person search datasets CUHK-SYSU and PRW. Comprehensive experiments are conducted on the two datasets and compared with the baseline methods, our method achieves the state-of-the-art performance."
                    },
                    {
                        "title": "Generalizable Person Re-Identification via Viewpoint Alignment and Fusion",
                        "abstract": "In the current person Re-identification (ReID) methods, most domain generalization works focus on dealing with style differences between domains while largely ignoring unpredictable camera view change, which we identify as another major factor leading to a poor generalization of ReID methods. To tackle the viewpoint change, this work proposes to use a 3D dense pose estimation model and a texture mapping module to map the pedestrian images to canonical view images. Due to the imperfection of the texture mapping module, the canonical view images may lose the discriminative detail clues from the original images, and thus directly using them for ReID will inevitably result in poor performance. To handle this issue, we propose to fuse the original image and canonical view image via a transformer-based module. The key insight of this design is that the cross-attention mechanism in the transformer could be an ideal solution to align the discriminative texture clues from the original image with the canonical view image, which could compensate for the low-quality texture information of the canonical view image. Through extensive experiments, we show that our method can lead to superior performance over the existing approaches in various evaluation settings."
                    },
                    {
                        "title": "Frequency-Guided Spatial Adaptation for Camouflaged Object Detection",
                        "abstract": "Camouflaged object detection (COD) aims to segment camouflaged objects which exhibit very similar patterns with the surrounding environment. Recent research works have shown that enhancing the feature representation via the frequency information can greatly alleviate the ambiguity problem between the foreground objects and the background.With the emergence of vision foundation models, like InternImage, Segment Anything Model etc, adapting the pretrained model on COD tasks with a lightweight adapter module shows a novel and promising research direction. Existing adapter modules mainly care about the feature adaptation in the spatial domain. In this paper, we propose a novel frequency-guided spatial adaptation method for COD task. Specifically, we transform the input features of the adapter into frequency domain. By grouping and interacting with frequency components located within non overlapping circles in the spectrogram, different frequency components are dynamically enhanced or weakened, making the intensity of image details and contour features adaptively adjusted. At the same time, the features that are conducive to distinguishing object and background are highlighted, indirectly implying the position and shape of camouflaged object. We conduct extensive experiments on four widely adopted benchmark datasets and the proposed method outperforms 26 state-of-the-art methods with large margins. Code will be released."
                    }
                ]
            },
            "fa2c2e4c-dd56-4329-a5b5-34ce8143c260": {
                "pk": "fa2c2e4c-dd56-4329-a5b5-34ce8143c260",
                "name": "De Cheng",
                "collaborators": [
                    "Nannan Wang",
                    "Xinbo Gao",
                    "Junwei Han",
                    "Dingwen Zhang",
                    "Yubin Wang",
                    "Xinyang Jiang",
                    "Dongsheng Li",
                    "Cairong Zhao",
                    "Peiliang Huang",
                    "Jingyu Zhou"
                ],
                "domain": [
                    "Zero-shot Learning",
                    "Unsupervised Learning",
                    "Video Analysis",
                    "Incremental Learning"
                ],
                "publications": [
                    {
                        "title": "Robust Region Feature Synthesizer for Zero-Shot Object Detection",
                        "abstract": "Zero-shot object detection aims at incorporating class semantic vectors to realize the detection of (both seen and) unseen classes given an unconstrained test image. In this study, we reveal the core challenges in this research area: how to synthesize robust region features (for unseen objects) that are as intra-class diverse and inter-class separable as the real samples, so that strong unseen object detectors can be trained upon them. To address these challenges, we build a novel zero-shot object detection framework that contains an Intra-class Semantic Diverging component and an Inter-class Structure Preserving component. The former is used to realize the one-to-more mapping to obtain diverse visual features from each class semantic vector, preventing miss-classifying the real unseen objects as image backgrounds. While the latter is used to avoid the synthesized features too scattered to mix up the inter-class and foreground-background relationship. To demonstrate the effectiveness of the proposed approach, comprehensive experiments on PASCAL VOC, COCO, and DIOR datasets are conducted. Notably, our approach achieves the new state-of-the-art performance on PASCAL VOC and COCO and it is the first study to carry out zero-shot object detection in remote sensing imagery."
                    },
                    {
                        "title": "Hybrid Dynamic Contrast and Probability Distillation for Unsupervised Person Re-Id",
                        "abstract": "Unsupervised person re-identification (Re-Id) has attracted increasing attention due to its practical application in the read-world video surveillance system. The traditional unsupervised Re-Id are mostly based on the method alternating between clustering and fine-tuning with the classification or metric learning objectives on the grouped clusters. However, since person Re-Id is an open-set problem, the clustering based methods often leave out lots of outlier instances or group the instances into the wrong clusters, thus they can not make full use of the training samples as a whole. To solve these problems, we present the hybrid dynamic cluster contrast and probability distillation algorithm. It formulates the unsupervised Re-Id problem into an unified local-to-global dynamic contrastive learning and self-supervised probability distillation framework. Specifically, the proposed method can make the utmost of the self-supervised signals of all the clustered and un-clustered instances, from both the instances' self-contrastive level and the probability distillation respective, in the memory-based non-parametric manner. Besides, the proposed hybrid local-to-global contrastive learning can take full advantage of the informative and valuable training examples for effective and robust training. Extensive experiment results show that the proposed method achieves superior performances to state-of-the-art methods, under both the purely unsupervised and unsupervised domain adaptation experiment settings."
                    },
                    {
                        "title": "Exploring Homogeneous and Heterogeneous Consistent Label Associations for Unsupervised Visible-Infrared Person ReID",
                        "abstract": "Unsupervised visible-infrared person re-identification (USL-VI-ReID) aims to retrieve pedestrian images of the same identity from different modalities without annotations. While prior work focuses on establishing cross-modality pseudo-label associations to bridge the modality-gap, they ignore maintaining the instance-level homogeneous and heterogeneous consistency in pseudo-label space, resulting in coarse associations. In response, we introduce a Modality-Unified Label Transfer (MULT) module that simultaneously accounts for both homogeneous and heterogeneous fine-grained instance-level structures, yielding high-quality cross-modality label associations. It models both homogeneous and heterogeneous affinities, leveraging them to define the inconsistency for the pseudo-labels and then minimize it, leading to pseudo-labels that maintain alignment across modalities and consistency within intra-modality structures. Additionally, a straightforward plug-and-play Online Cross-memory Label Refinement (OCLR) module is proposed to further mitigate the impact of noisy pseudo-labels while simultaneously aligning different modalities, coupled with a Modality-Invariant Representation Learning (MIRL) framework. Experiments demonstrate that our proposed method outperforms existing USL-VI-ReID methods, highlighting the superiority of our MULT in comparison to other cross-modality association methods. The code will be available."
                    },
                    {
                        "title": "Neighbour Consistency Guided Pseudo-Label Refinement for Unsupervised Person Re-Identification",
                        "abstract": "Unsupervised person re-identification (ReID) aims at learning discriminative identity features for person retrieval without any annotations. Recent advances accomplish this task by leveraging clustering-based pseudo labels, but these pseudo labels are inevitably noisy which deteriorate model performance. In this paper, we propose a Neighbour Consistency guided Pseudo Label Refinement (NCPLR) framework, which can be regarded as a transductive form of label propagation under the assumption that the prediction of each example should be similar to its nearest neighbours'. Specifically, the refined label for each training instance can be obtained by the original clustering result and a weighted ensemble of its neighbours' predictions, with weights determined according to their similarities in the feature space. In addition, we consider the clustering-based unsupervised person ReID as a label-noise learning problem. Then, we proposed an explicit neighbour consistency regularization to reduce model susceptibility to over-fitting while improving the training stability. The NCPLR method is simple yet effective, and can be seamlessly integrated into existing clustering-based unsupervised algorithms. Extensive experimental results on five ReID datasets demonstrate the effectiveness of the proposed method, and showing superior performance to state-of-the-art methods by a large margin."
                    },
                    {
                        "title": "ActPrompt: In-Domain Feature Adaptation via Action Cues for Video Temporal Grounding",
                        "abstract": "Video temporal grounding is an emerging topic aiming to identify specific clips within videos. In addition to pre-trained video models, contemporary methods utilize pre-trained vision-language models (VLM) to capture detailed characteristics of diverse scenes and objects from video frames. However, as pre-trained on images, VLM may struggle to distinguish action-sensitive patterns from static objects, making it necessary to adapt them to specific data domains for effective feature representation over temporal grounding. We address two primary challenges to achieve this goal. Specifically, to mitigate high adaptation costs, we propose an efficient preliminary in-domain fine-tuning paradigm for feature adaptation, where downstream-adaptive features are learned through several pretext tasks. Furthermore, to integrate action-sensitive information into VLM, we introduce Action-Cue-Injected Temporal Prompt Learning (ActPrompt), which injects action cues into the image encoder of VLM for better discovering action-sensitive patterns. Extensive experiments demonstrate that ActPrompt is an off-the-shelf training framework that can be effectively applied to various SOTA methods, resulting in notable improvements. The complete code used in this study is provided in the supplementary materials."
                    },
                    {
                        "title": "Learning Hierarchical Prompt with Structured Linguistic Knowledge for Vision-Language Models",
                        "abstract": "Prompt learning has become a prevalent strategy for adapting vision-language foundation models to downstream tasks. As large language models (LLMs) have emerged, recent studies have explored the use of category-related descriptions as input to enhance prompt effectiveness. Nevertheless, conventional descriptions fall short of structured information that effectively represents the interconnections among entities or attributes linked to a particular category. To address this limitation and prioritize harnessing structured knowledge, this paper advocates for leveraging LLMs to build a graph for each description to model the entities and attributes describing the category, as well as their correlations. Preexisting prompt tuning methods exhibit inadequacies in managing this structured knowledge. Consequently, we propose a novel approach called Hierarchical Prompt Tuning (HPT), which enables simultaneous modeling of both structured and conventional linguistic knowledge. Specifically, we introduce a relationship-guided attention module to capture pair-wise associations among entities and attributes for low-level prompt learning. In addition, by incorporating high-level and global-level prompts modeling overall semantics, the proposed hierarchical structure forges cross-level interlinks and empowers the model to handle more complex and long-term relationships. Extensive experiments demonstrate that our HPT shows strong effectiveness and generalizes much better than existing SOTA methods. Our code is available at https://github.com/Vill-Lab/2024-AAAI-HPT."
                    },
                    {
                        "title": "Center-Sensitive Kernel Optimization for Efficient On-Device Incremental Learning",
                        "abstract": "To facilitate the evolution of edge intelligence in ever-changing environments, we study on-device incremental learning constrained in limited computation resource in this paper. Current on-device training methods just focus on efficient training without considering the catastrophic forgetting, preventing the model getting stronger when continually exploring the world. To solve this problem, a direct solution is to involve the existing incremental learning mechanisms into the on-device training framework. Unfortunately, such a manner cannot work well as those mechanisms usually introduce large additional computational cost to the network optimization process, which would inevitably exceed the memory capacity of the edge devices. To address this issue, this paper makes an early effort to propose a simple but effective edge-friendly incremental learning framework. Based on an empirical study on the knowledge intensity of the kernel elements of the neural network, we find that the center kernel is the key for maximizing the knowledge intensity for learning new data, while freezing the other kernel elements would get a good balance on the model's capacity for overcoming catastrophic forgetting. Upon this finding, we further design a center-sensitive kernel optimization framework to largely alleviate the cost of the gradient computation and back-propagation. Besides, a dynamic channel element selection strategy is also proposed to facilitate a sparse orthogonal gradient projection for further reducing the optimization complexity, upon the knowledge explored from the new task data. Extensive experiments validate our method is efficient and effective, e.g., our method achieves average accuracy boost of 38.08% with even less memory and approximate computation compared to existing on-device training methods, indicating its significant potential for on-device incremental learning."
                    }
                ]
            },
            "1cf61d82-4b2c-426c-92d8-385850e4ca63": {
                "pk": "1cf61d82-4b2c-426c-92d8-385850e4ca63",
                "name": "Yinghui Xing",
                "collaborators": [
                    "Yanning Zhang",
                    "Shizhou Zhang",
                    "Guoqiang Liang",
                    "Lingyan Ran",
                    "De Cheng",
                    "Peng Wang",
                    "Yan Zhang",
                    "Xiuwei Zhang",
                    "Song Wang",
                    "Dexuan Kong"
                ],
                "domain": [
                    "Image Processing",
                    "Deep Learning",
                    "Remote Sensing",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Pansharpening via Frequency-Aware Fusion Network with Explicit Similarity Constraints",
                        "abstract": "The process of fusing a high spatial resolution (HR) panchromatic (PAN) image and a low spatial resolution (LR) multispectral (MS) image to obtain an HRMS image is known as pansharpening. With the development of convolutional neural networks, the performance of pansharpening methods has been improved, however, the blurry effects and the spectral distortion still exist in their fusion results due to the insufficiency in details learning and the frequency mismatch between MSand PAN. Therefore, the improvement of spatial details at the premise of reducing spectral distortion is still a challenge. In this paper, we propose a frequency-aware fusion network (FAFNet) together with a novel high-frequency feature similarity loss to address above mentioned problems. FAFNet is mainly composed of two kinds of blocks, where the frequency aware blocks aim to extract features in the frequency domain with the help of discrete wavelet transform (DWT) layers, and the frequency fusion blocks reconstruct and transform the features from frequency domain to spatial domain with the assistance of inverse DWT (IDWT) layers. Finally, the fusion results are obtained through a convolutional block. In order to learn the correspondence, we also propose a high-frequency feature similarity loss to constrain the HF features derived from PAN and MS branches, so that HF features of PAN can reasonably be used to supplement that of MS. Experimental results on three datasets at both reduced- and full-resolution demonstrate the superiority of the proposed method compared with several state-of-the-art pansharpening models."
                    },
                    {
                        "title": "PC-GANs: Progressive Compensation Generative Adversarial Networks for Pan-sharpening",
                        "abstract": "The fusion of multispectral and panchromatic images is always dubbed pansharpening. Most of the available deep learning-based pan-sharpening methods sharpen the multispectral images through a one-step scheme, which strongly depends on the reconstruction ability of the network. However, remote sensing images always have large variations, as a result, these one-step methods are vulnerable to the error accumulation and thus incapable of preserving spatial details as well as the spectral information. In this paper, we propose a novel two-step model for pan-sharpening that sharpens the MS image through the progressive compensation of the spatial and spectral information. Firstly, a deep multiscale guided generative adversarial network is used to preliminarily enhance the spatial resolution of the MS image. Starting from the pre-sharpened MS image in the coarse domain, our approach then progressively refines the spatial and spectral residuals over a couple of generative adversarial networks (GANs) that have reverse architectures. The whole model is composed of triple GANs, and based on the specific architecture, a joint compensation loss function is designed to enable the triple GANs to be trained simultaneously. Moreover, the spatial-spectral residual compensation structure proposed in this paper can be extended to other pan-sharpening methods to further enhance their fusion results. Extensive experiments are performed on different datasets and the results demonstrate the effectiveness and efficiency of our proposed method."
                    },
                    {
                        "title": "CrossDiff: Exploring Self-Supervised Representation of Pansharpening via Cross-Predictive Diffusion Model",
                        "abstract": "Fusion of a panchromatic (PAN) image and corresponding multispectral (MS) image is also known as pansharpening, which aims to combine abundant spatial details of PAN and spectral information of MS. Due to the absence of high-resolution MS images, available deep-learning-based methods usually follow the paradigm of training at reduced resolution and testing at both reduced and full resolution. When taking original MS and PAN images as inputs, they always obtain sub-optimal results due to the scale variation. In this paper, we propose to explore the self-supervised representation of pansharpening by designing a cross-predictive diffusion model, named CrossDiff. It has two-stage training. In the first stage, we introduce a cross-predictive pretext task to pre-train the UNet structure based on conditional DDPM, while in the second stage, the encoders of the UNets are frozen to directly extract spatial and spectral features from PAN and MS, and only the fusion head is trained to adapt for pansharpening task. Extensive experiments show the effectiveness and superiority of the proposed model compared with state-of-the-art supervised and unsupervised methods. Besides, the cross-sensor experiments also verify the generalization ability of proposed self-supervised representation learners for other satellite's datasets. We will release our code for reproducibility."
                    },
                    {
                        "title": "DDF: A Novel Dual-Domain Image Fusion Strategy for Remote Sensing Image Semantic Segmentation with Unsupervised Domain Adaptation",
                        "abstract": "Semantic segmentation of remote sensing images is a challenging and hot issue due to the large amount of unlabeled data. Unsupervised domain adaptation (UDA) has proven to be advantageous in incorporating unclassified information from the target domain. However, independently fine-tuning UDA models on the source and target domains has a limited effect on the outcome. This paper proposes a hybrid training strategy as well as a novel dual-domain image fusion strategy that effectively utilizes the original image, transformation image, and intermediate domain information. Moreover, to enhance the precision of pseudo-labels, we present a pseudo-label region-specific weight strategy. The efficacy of our approach is substantiated by extensive benchmark experiments and ablation studies conducted on the ISPRS Vaihingen and Potsdam datasets."
                    },
                    {
                        "title": "Pre-train, Adapt and Detect: Multi-Task Adapter Tuning for Camouflaged Object Detection",
                        "abstract": "Camouflaged object detection (COD), aiming to segment camouflaged objects which exhibit similar patterns with the background, is a challenging task. Most existing works are dedicated to establishing specialized modules to identify camouflaged objects with complete and fine details, while the boundary can not be well located for the lack of object-related semantics. In this paper, we propose a novel ``pre-train, adapt and detect\" paradigm to detect camouflaged objects. By introducing a large pre-trained model, abundant knowledge learned from massive multi-modal data can be directly transferred to COD. A lightweight parallel adapter is inserted to adjust the features suitable for the downstream COD task. Extensive experiments on four challenging benchmark datasets demonstrate that our method outperforms existing state-of-the-art COD models by large margins. Moreover, we design a multi-task learning scheme for tuning the adapter to exploit the shareable knowledge across different semantic classes. Comprehensive experimental results showed that the generalization ability of our model can be substantially improved with multi-task adapter initialization on source tasks and multi-task adaptation on target tasks."
                    },
                    {
                        "title": "MS-DETR: Multispectral Pedestrian Detection Transformer with Loosely Coupled Fusion and Modality-Balanced Optimization",
                        "abstract": "Multispectral pedestrian detection is an important task for many around-the-clock applications, since the visible and thermal modalities can provide complementary information especially under low light conditions. Most of the available multispectral pedestrian detectors are based on non-end-to-end detectors, while in this paper, we propose MultiSpectral pedestrian DEtection TRansformer (MS-DETR), an end-to-end multispectral pedestrian detector, which extends DETR into the field of multi-modal detection. MS-DETR consists of two modality-specific backbones and Transformer encoders, followed by a multi-modal Transformer decoder, and the visible and thermal features are fused in the multi-modal Transformer decoder. To well resist the misalignment between multi-modal images, we design a loosely coupled fusion strategy by sparsely sampling some keypoints from multi-modal features independently and fusing them with adaptively learned attention weights. Moreover, based on the insight that not only different modalities, but also different pedestrian instances tend to have different confidence scores to final detection, we further propose an instance-aware modality-balanced optimization strategy, which preserves visible and thermal decoder branches and aligns their predicted slots through an instance-wise dynamic loss. Our end-to-end MS-DETR shows superior performance on the challenging KAIST, CVC-14 and LLVIP benchmark datasets. The source code is available at https://github.com/YinghuiXing/MS-DETR ."
                    },
                    {
                        "title": "Dual Modality Prompt Tuning for Vision-Language Pre-Trained Model",
                        "abstract": "With the emergence of large pre-trained vison-language model like CLIP, transferable representations can be adapted to a wide range of downstream tasks via prompt tuning. Prompt tuning tries to probe the beneficial information for downstream tasks from the general knowledge stored in the pre-trained model. A recently proposed method named Context Optimization (CoOp) introduces a set of learnable vectors as text prompt from the language side. However, tuning the text prompt alone can only adjust the synthesized \"classifier\", while the computed visual features of the image encoder can not be affected , thus leading to sub-optimal solutions. In this paper, we propose a novel Dual-modality Prompt Tuning (DPT) paradigm through learning text and visual prompts simultaneously. To make the final image feature concentrate more on the target visual concept, a Class-Aware Visual Prompt Tuning (CAVPT) scheme is further proposed in our DPT, where the class-aware visual prompt is generated dynamically by performing the cross attention between text prompts features and image patch token embeddings to encode both the downstream task-related information and visual instance information. Extensive experimental results on 11 datasets demonstrate the effectiveness and generalization ability of the proposed method. Our code is available in https://github.com/fanrena/DPT."
                    },
                    {
                        "title": "Ground-to-Aerial Person Search: Benchmark Dataset and Approach",
                        "abstract": "In this work, we construct a large-scale dataset for Ground-to-Aerial Person Search, named G2APS, which contains 31,770 images of 260,559 annotated bounding boxes for 2,644 identities appearing in both of the UAVs and ground surveillance cameras. To our knowledge, this is the first dataset for cross-platform intelligent surveillance applications, where the UAVs could work as a powerful complement for the ground surveillance cameras. To more realistically simulate the actual cross-platform Ground-to-Aerial surveillance scenarios, the surveillance cameras are fixed about 2 meters above the ground, while the UAVs capture videos of persons at different location, with a variety of view-angles, flight attitudes and flight modes. Therefore, the dataset has the following unique characteristics: 1) drastic view-angle changes between query and gallery person images from cross-platform cameras; 2) diverse resolutions, poses and views of the person images under 9 rich real-world scenarios. On basis of the G2APS benchmark dataset, we demonstrate detailed analysis about current two-step and end-to-end person search methods, and further propose a simple yet effective knowledge distillation scheme on the head of the ReID network, which achieves state-of-the-art performances on both of the G2APS and the previous two public person search datasets, i.e., PRW and CUHK-SYSU. The dataset and source code available on \\url{https://github.com/yqc123456/HKD_for_person_search}."
                    },
                    {
                        "title": "Cross-Platform Video Person ReID: A New Benchmark Dataset and Adaptation Approach",
                        "abstract": "In this paper, we construct a large-scale benchmark dataset for Ground-to-Aerial Video-based person Re-Identification, named G2A-VReID, which comprises 185,907 images and 5,576 tracklets, featuring 2,788 distinct identities. To our knowledge, this is the first dataset for video ReID under Ground-to-Aerial scenarios. G2A-VReID dataset has the following characteristics: 1) Drastic view changes; 2) Large number of annotated identities; 3) Rich outdoor scenarios; 4) Huge difference in resolution. Additionally, we propose a new benchmark approach for cross-platform ReID by transforming the cross-platform visual alignment problem into visual-semantic alignment through vision-language model (i.e., CLIP) and applying a parameter-efficient Video Set-Level-Adapter module to adapt image-based foundation model to video ReID tasks, termed VSLA-CLIP. Besides, to further reduce the great discrepancy across the platforms, we also devise the platform-bridge prompts for efficient visual feature alignment. Extensive experiments demonstrate the superiority of the proposed method on all existing video ReID datasets and our proposed G2A-VReID dataset."
                    },
                    {
                        "title": "Text-based Person Search in Full Images via Semantic-Driven Proposal Generation",
                        "abstract": "Finding target persons in full scene images with a query of text description has important practical applications in intelligent video surveillance.However, different from the real-world scenarios where the bounding boxes are not available, existing text-based person retrieval methods mainly focus on the cross modal matching between the query text descriptions and the gallery of cropped pedestrian images. To close the gap, we study the problem of text-based person search in full images by proposing a new end-to-end learning framework which jointly optimize the pedestrian detection, identification and visual-semantic feature embedding tasks. To take full advantage of the query text, the semantic features are leveraged to instruct the Region Proposal Network to pay more attention to the text-described proposals. Besides, a cross-scale visual-semantic embedding mechanism is utilized to improve the performance. To validate the proposed method, we collect and annotate two large-scale benchmark datasets based on the widely adopted image-based person search datasets CUHK-SYSU and PRW. Comprehensive experiments are conducted on the two datasets and compared with the baseline methods, our method achieves the state-of-the-art performance."
                    },
                    {
                        "title": "Frequency-Guided Spatial Adaptation for Camouflaged Object Detection",
                        "abstract": "Camouflaged object detection (COD) aims to segment camouflaged objects which exhibit very similar patterns with the surrounding environment. Recent research works have shown that enhancing the feature representation via the frequency information can greatly alleviate the ambiguity problem between the foreground objects and the background.With the emergence of vision foundation models, like InternImage, Segment Anything Model etc, adapting the pretrained model on COD tasks with a lightweight adapter module shows a novel and promising research direction. Existing adapter modules mainly care about the feature adaptation in the spatial domain. In this paper, we propose a novel frequency-guided spatial adaptation method for COD task. Specifically, we transform the input features of the adapter into frequency domain. By grouping and interacting with frequency components located within non overlapping circles in the spectrogram, different frequency components are dynamically enhanced or weakened, making the intensity of image details and contour features adaptively adjusted. At the same time, the features that are conducive to distinguishing object and background are highlighted, indirectly implying the position and shape of camouflaged object. We conduct extensive experiments on four widely adopted benchmark datasets and the proposed method outperforms 26 state-of-the-art methods with large margins. Code will be released."
                    }
                ]
            },
            "91530277-e3ad-419d-b6bc-9715476f1433": {
                "pk": "91530277-e3ad-419d-b6bc-9715476f1433",
                "name": "Nannan Wang",
                "collaborators": [
                    "Xinbo Gao",
                    "Mingrui Zhu",
                    "Xiaoyu Wang",
                    "Jie Li",
                    "De Cheng",
                    "Chunlei Peng",
                    "Yun Yi",
                    "Lixin Xu",
                    "Dawei Zhou",
                    "Bo Han"
                ],
                "domain": [
                    "Computer Vision",
                    "Face Recognition",
                    "Deep Learning",
                    "Adversarial Learning"
                ],
                "publications": [
                    {
                        "title": "Strong Gravitational Lensing and Its Cosmic Constraints",
                        "abstract": "In this paper, we propose a new method to use the strong lensing data sets to constrain a cosmological model. By taking the ratio $\\mathcal{D}^{obs}_{ij}=\\theta_{\\mathrm{E_{\\mathrm{i}}}}\\sigma_{\\mathrm{0_{\\mathrm{j}}}}^2/\\theta_{\\mathrm{E_{\\mathrm{j}}}}\\sigma_{\\mathrm{0_{\\mathrm{i}}}}^2$ as cosmic observations, one can {\\it completely} eliminate the uncertainty caused by the relation $\\sigma_{\\mathrm{SIS}}=f_{\\mathrm{E}}\\sigma_0$ which characterizes the relation between the stellar velocity dispersion $\\sigma_0$ and the velocity dispersion $\\sigma_{SIS}$. Via our method, a relative tight constraint to the cosmological model space can be obtained, for the spatially flat $\\Lambda$CDM model as an example $\\Omega_m=0.143_{- 0.143-0.143-0.143}^{+ 0.000769+0.143+0.489}$ in $3\\sigma$ regions. And by using this method, one can also probe the nature of dark energy and the spatial curvature of our Universe."
                    },
                    {
                        "title": "Random Sampling for Fast Face Sketch Synthesis",
                        "abstract": "Exemplar-based face sketch synthesis plays an important role in both digital entertainment and law enforcement. It generally consists of two parts: neighbor selection and reconstruction weight representation. The most time-consuming or main computation complexity for exemplar-based face sketch synthesis methods lies in the neighbor selection process. State-of-the-art face sketch synthesis methods perform neighbor selection online in a data-driven manner by $K$ nearest neighbor ($K$-NN) searching. Actually, the online search increases the time consuming for synthesis. Moreover, since these methods need to traverse the whole training dataset for neighbor selection, the computational complexity increases with the scale of the training database and hence these methods have limited scalability. In this paper, we proposed a simple but effective offline random sampling in place of online $K$-NN search to improve the synthesis efficiency. Extensive experiments on public face sketch databases demonstrate the superiority of the proposed method in comparison to state-of-the-art methods, in terms of both synthesis quality and time consumption. The proposed method could be extended to other heterogeneous face image transformation problems such as face hallucination. We release the source codes of our proposed methods and the evaluation metrics for future study online: http://www.ihitworld.com/RSLCR.html."
                    },
                    {
                        "title": "Graphical Representation for Heterogeneous Face Recognition",
                        "abstract": "Heterogeneous face recognition (HFR) refers to matching face images acquired from different sources (i.e., different sensors or different wavelengths) for identification. HFR plays an important role in both biometrics research and industry. In spite of promising progresses achieved in recent years, HFR is still a challenging problem due to the difficulty to represent two heterogeneous images in a homogeneous manner. Existing HFR methods either represent an image ignoring the spatial information, or rely on a transformation procedure which complicates the recognition task. Considering these problems, we propose a novel graphical representation based HFR method (G-HFR) in this paper. Markov networks are employed to represent heterogeneous image patches separately, which takes the spatial compatibility between neighboring image patches into consideration. A coupled representation similarity metric (CRSM) is designed to measure the similarity between obtained graphical representations. Extensive experiments conducted on multiple HFR scenarios (viewed sketch, forensic sketch, near infrared image, and thermal infrared image) show that the proposed method outperforms state-of-the-art methods."
                    },
                    {
                        "title": "Sparse Graphical Representation based Discriminant Analysis for Heterogeneous Face Recognition",
                        "abstract": "Face images captured in heterogeneous environments, e.g., sketches generated by the artists or composite-generation software, photos taken by common cameras and infrared images captured by corresponding infrared imaging devices, usually subject to large texture (i.e., style) differences. This results in heavily degraded performance of conventional face recognition methods in comparison with the performance on images captured in homogeneous environments. In this paper, we propose a novel sparse graphical representation based discriminant analysis (SGR-DA) approach to address aforementioned face recognition in heterogeneous scenarios. An adaptive sparse graphical representation scheme is designed to represent heterogeneous face images, where a Markov networks model is constructed to generate adaptive sparse vectors. To handle the complex facial structure and further improve the discriminability, a spatial partition-based discriminant analysis framework is presented to refine the adaptive sparse vectors for face matching. We conducted experiments on six commonly used heterogeneous face datasets and experimental results illustrate that our proposed SGR-DA approach achieves superior performance in comparison with state-of-the-art methods."
                    },
                    {
                        "title": "Modeling Adversarial Noise for Adversarial Training",
                        "abstract": "Deep neural networks have been demonstrated to be vulnerable to adversarial noise, promoting the development of defense against adversarial attacks. Motivated by the fact that adversarial noise contains well-generalizing features and that the relationship between adversarial data and natural data can help infer natural data and make reliable predictions, in this paper, we study to model adversarial noise by learning the transition relationship between adversarial labels (i.e. the flipped labels used to generate adversarial data) and natural labels (i.e. the ground truth labels of the natural data). Specifically, we introduce an instance-dependent transition matrix to relate adversarial labels and natural labels, which can be seamlessly embedded with the target model (enabling us to model stronger adaptive adversarial noise). Empirical evaluations demonstrate that our method could effectively improve adversarial accuracy."
                    },
                    {
                        "title": "Human-Inspired Facial Sketch Synthesis with Dynamic Adaptation",
                        "abstract": "Facial sketch synthesis (FSS) aims to generate a vivid sketch portrait from a given facial photo. Existing FSS methods merely rely on 2D representations of facial semantic or appearance. However, professional human artists usually use outlines or shadings to covey 3D geometry. Thus facial 3D geometry (e.g. depth map) is extremely important for FSS. Besides, different artists may use diverse drawing techniques and create multiple styles of sketches; but the style is globally consistent in a sketch. Inspired by such observations, in this paper, we propose a novel Human-Inspired Dynamic Adaptation (HIDA) method. Specially, we propose to dynamically modulate neuron activations based on a joint consideration of both facial 3D geometry and 2D appearance, as well as globally consistent style control. Besides, we use deformable convolutions at coarse-scales to align deep features, for generating abstract and distinct outlines. Experiments show that HIDA can generate high-quality sketches in multiple styles, and significantly outperforms previous methods, over a large range of challenging faces. Besides, HIDA allows precise style control of the synthesized sketch, and generalizes well to natural scenes and other artistic styles. Our code and results have been released online at: https://github.com/AiArt-HDU/HIDA."
                    },
                    {
                        "title": "Exploring Homogeneous and Heterogeneous Consistent Label Associations for Unsupervised Visible-Infrared Person ReID",
                        "abstract": "Unsupervised visible-infrared person re-identification (USL-VI-ReID) aims to retrieve pedestrian images of the same identity from different modalities without annotations. While prior work focuses on establishing cross-modality pseudo-label associations to bridge the modality-gap, they ignore maintaining the instance-level homogeneous and heterogeneous consistency in pseudo-label space, resulting in coarse associations. In response, we introduce a Modality-Unified Label Transfer (MULT) module that simultaneously accounts for both homogeneous and heterogeneous fine-grained instance-level structures, yielding high-quality cross-modality label associations. It models both homogeneous and heterogeneous affinities, leveraging them to define the inconsistency for the pseudo-labels and then minimize it, leading to pseudo-labels that maintain alignment across modalities and consistency within intra-modality structures. Additionally, a straightforward plug-and-play Online Cross-memory Label Refinement (OCLR) module is proposed to further mitigate the impact of noisy pseudo-labels while simultaneously aligning different modalities, coupled with a Modality-Invariant Representation Learning (MIRL) framework. Experiments demonstrate that our proposed method outperforms existing USL-VI-ReID methods, highlighting the superiority of our MULT in comparison to other cross-modality association methods. The code will be available."
                    },
                    {
                        "title": "Facial Feature Point Detection: A Comprehensive Survey",
                        "abstract": "This paper presents a comprehensive survey of facial feature point detection with the assistance of abundant manually labeled images. Facial feature point detection favors many applications such as face recognition, animation, tracking, hallucination, expression analysis and 3D face modeling. Existing methods can be categorized into the following four groups: constrained local model (CLM)-based, active appearance model (AAM)-based, regression-based, and other methods. CLM-based methods consist of a shape model and a number of local experts, each of which is utilized to detect a facial feature point. AAM-based methods fit a shape model to an image by minimizing texture synthesis errors. Regression-based methods directly learn a mapping function from facial image appearance to facial feature points. Besides the above three major categories of methods, there are also minor categories of methods which we classify into other methods: graphical model-based methods, joint face alignment methods, independent facial feature point detectors, and deep learning-based methods. Though significant progress has been made, facial feature point detection is limited in its success by wild and real-world conditions: variations across poses, expressions, illuminations, and occlusions. A comparative illustration and analysis of representative methods provide us a holistic understanding and deep insight into facial feature point detection, which also motivates us to explore promising future directions."
                    },
                    {
                        "title": "Hybrid Dynamic Contrast and Probability Distillation for Unsupervised Person Re-Id",
                        "abstract": "Unsupervised person re-identification (Re-Id) has attracted increasing attention due to its practical application in the read-world video surveillance system. The traditional unsupervised Re-Id are mostly based on the method alternating between clustering and fine-tuning with the classification or metric learning objectives on the grouped clusters. However, since person Re-Id is an open-set problem, the clustering based methods often leave out lots of outlier instances or group the instances into the wrong clusters, thus they can not make full use of the training samples as a whole. To solve these problems, we present the hybrid dynamic cluster contrast and probability distillation algorithm. It formulates the unsupervised Re-Id problem into an unified local-to-global dynamic contrastive learning and self-supervised probability distillation framework. Specifically, the proposed method can make the utmost of the self-supervised signals of all the clustered and un-clustered instances, from both the instances' self-contrastive level and the probability distillation respective, in the memory-based non-parametric manner. Besides, the proposed hybrid local-to-global contrastive learning can take full advantage of the informative and valuable training examples for effective and robust training. Extensive experiment results show that the proposed method achieves superior performances to state-of-the-art methods, under both the purely unsupervised and unsupervised domain adaptation experiment settings."
                    },
                    {
                        "title": "Adapt and Align to Improve Zero-Shot Sketch-Based Image Retrieval",
                        "abstract": "Zero-shot sketch-based image retrieval (ZS-SBIR) is challenging due to the cross-domain nature of sketches and photos, as well as the semantic gap between seen and unseen image distributions. Previous methods fine-tune pre-trained models with various side information and learning strategies to learn a compact feature space that is shared between the sketch and photo domains and bridges seen and unseen classes. However, these efforts are inadequate in adapting domains and transferring knowledge from seen to unseen classes. In this paper, we present an effective ``Adapt and Align'' approach to address the key challenges. Specifically, we insert simple and lightweight domain adapters to learn new abstract concepts of the sketch domain and improve cross-domain representation capabilities. Inspired by recent advances in image-text foundation models (e.g., CLIP) on zero-shot scenarios, we explicitly align the learned image embedding with a more semantic text embedding to achieve the desired knowledge transfer from seen to unseen classes. Extensive experiments on three benchmark datasets and two popular backbones demonstrate the superiority of our method in terms of retrieval accuracy and flexibility."
                    },
                    {
                        "title": "HFORD: High-Fidelity and Occlusion-Robust De-identification for Face Privacy Protection",
                        "abstract": "With the popularity of smart devices and the development of computer vision technology, concerns about face privacy protection are growing. The face de-identification technique is a practical way to solve the identity protection problem. The existing facial de-identification methods have revealed several problems, including the impact on the realism of anonymized results when faced with occlusions and the inability to maintain identity-irrelevant details in anonymized results. We present a High-Fidelity and Occlusion-Robust De-identification (HFORD) method to deal with these issues. This approach can disentangle identities and attributes while preserving image-specific details such as background, facial features (e.g., wrinkles), and lighting, even in occluded scenes. To disentangle the latent codes in the GAN inversion space, we introduce an Identity Disentanglement Module (IDM). This module selects the latent codes that are closely related to the identity. It further separates the latent codes into identity-related codes and attribute-related codes, enabling the network to preserve attributes while only modifying the identity. To ensure the preservation of image details and enhance the network's robustness to occlusions, we propose an Attribute Retention Module (ARM). This module adaptively preserves identity-irrelevant details and facial occlusions and blends them into the generated results in a modulated manner. Extensive experiments show that our method has higher quality, better detail fidelity, and stronger occlusion robustness than other face de-identification methods."
                    },
                    {
                        "title": "Semi-parametric Makeup Transfer via Semantic-aware Correspondence",
                        "abstract": "The large discrepancy between the source non-makeup image and the reference makeup image is one of the key challenges in makeup transfer. Conventional approaches for makeup transfer either learn disentangled representation or perform pixel-wise correspondence in a parametric way between two images. We argue that non-parametric techniques have a high potential for addressing the pose, expression, and occlusion discrepancies. To this end, this paper proposes a \\textbf{S}emi-\\textbf{p}arametric \\textbf{M}akeup \\textbf{T}ransfer (SpMT) method, which combines the reciprocal strengths of non-parametric and parametric mechanisms. The non-parametric component is a novel \\textbf{S}emantic-\\textbf{a}ware \\textbf{C}orrespondence (SaC) module that explicitly reconstructs content representation with makeup representation under the strong constraint of component semantics. The reconstructed representation is desired to preserve the spatial and identity information of the source image while \"wearing\" the makeup of the reference image. The output image is synthesized via a parametric decoder that draws on the reconstructed representation. Extensive experiments demonstrate the superiority of our method in terms of visual quality, robustness, and flexibility. Code and pre-trained model are available at \\url{https://github.com/AnonymScholar/SpMT."
                    },
                    {
                        "title": "Neighbour Consistency Guided Pseudo-Label Refinement for Unsupervised Person Re-Identification",
                        "abstract": "Unsupervised person re-identification (ReID) aims at learning discriminative identity features for person retrieval without any annotations. Recent advances accomplish this task by leveraging clustering-based pseudo labels, but these pseudo labels are inevitably noisy which deteriorate model performance. In this paper, we propose a Neighbour Consistency guided Pseudo Label Refinement (NCPLR) framework, which can be regarded as a transductive form of label propagation under the assumption that the prediction of each example should be similar to its nearest neighbours'. Specifically, the refined label for each training instance can be obtained by the original clustering result and a weighted ensemble of its neighbours' predictions, with weights determined according to their similarities in the feature space. In addition, we consider the clustering-based unsupervised person ReID as a label-noise learning problem. Then, we proposed an explicit neighbour consistency regularization to reduce model susceptibility to over-fitting while improving the training stability. The NCPLR method is simple yet effective, and can be seamlessly integrated into existing clustering-based unsupervised algorithms. Extensive experimental results on five ReID datasets demonstrate the effectiveness of the proposed method, and showing superior performance to state-of-the-art methods by a large margin."
                    },
                    {
                        "title": "All-to-key Attention for Arbitrary Style Transfer",
                        "abstract": "Attention-based arbitrary style transfer studies have shown promising performance in synthesizing vivid local style details. They typically use the all-to-all attention mechanism -- each position of content features is fully matched to all positions of style features. However, all-to-all attention tends to generate distorted style patterns and has quadratic complexity, limiting the effectiveness and efficiency of arbitrary style transfer. In this paper, we propose a novel all-to-key attention mechanism -- each position of content features is matched to stable key positions of style features -- that is more in line with the characteristics of style transfer. Specifically, it integrates two newly proposed attention forms: distributed and progressive attention. Distributed attention assigns attention to key style representations that depict the style distribution of local regions; Progressive attention pays attention from coarse-grained regions to fine-grained key positions. The resultant module, dubbed StyA2K, shows extraordinary performance in preserving the semantic structure and rendering consistent style patterns. Qualitative and quantitative comparisons with state-of-the-art methods demonstrate the superior performance of our approach."
                    },
                    {
                        "title": "NAR-Former: Neural Architecture Representation Learning towards Holistic Attributes Prediction",
                        "abstract": "With the wide and deep adoption of deep learning models in real applications, there is an increasing need to model and learn the representations of the neural networks themselves. These models can be used to estimate attributes of different neural network architectures such as the accuracy and latency, without running the actual training or inference tasks. In this paper, we propose a neural architecture representation model that can be used to estimate these attributes holistically. Specifically, we first propose a simple and effective tokenizer to encode both the operation and topology information of a neural network into a single sequence. Then, we design a multi-stage fusion transformer to build a compact vector representation from the converted sequence. For efficient model training, we further propose an information flow consistency augmentation and correspondingly design an architecture consistency loss, which brings more benefits with less augmentation samples compared with previous random augmentation strategies. Experiment results on NAS-Bench-101, NAS-Bench-201, DARTS search space and NNLQP show that our proposed framework can be used to predict the aforementioned latency and accuracy attributes of both cell architectures and whole deep neural networks, and achieves promising performance. Code is available at https://github.com/yuny220/NAR-Former."
                    },
                    {
                        "title": "Few-shot Face Image Translation via GAN Prior Distillation",
                        "abstract": "Face image translation has made notable progress in recent years. However, when training on limited data, the performance of existing approaches significantly declines. Although some studies have attempted to tackle this problem, they either failed to achieve the few-shot setting (less than 10) or can only get suboptimal results. In this paper, we propose GAN Prior Distillation (GPD) to enable effective few-shot face image translation. GPD contains two models: a teacher network with GAN Prior and a student network that fulfills end-to-end translation. Specifically, we adapt the teacher network trained on large-scale data in the source domain to the target domain with only a few samples, where it can learn the target domain's knowledge. Then, we can achieve few-shot augmentation by generating source domain and target domain images simultaneously with the same latent codes. We propose an anchor-based knowledge distillation module that can fully use the difference between the training and the augmented data to distill the knowledge of the teacher network into the student network. The trained student network achieves excellent generalization performance with the absorption of additional knowledge. Qualitative and quantitative experiments demonstrate that our method achieves superior results than state-of-the-art approaches in a few-shot setting."
                    }
                ]
            },
            "7f64d92a-f841-4abb-b292-4b78646b091a": {
                "pk": "7f64d92a-f841-4abb-b292-4b78646b091a",
                "name": "Peng Wang",
                "collaborators": [],
                "domain": [
                    "Geometry",
                    "Willmore Surfaces",
                    "Ontology Matching",
                    "Crowd Behavior"
                ],
                "publications": [
                    {
                        "title": "On the Willmore functional of 2-tori in some product Riemannian manifolds",
                        "abstract": "We discuss the minimum of Willmore functional of torus in a Riemannian manifold $N$, especially for the case that $N$ is a product manifold. We show that when $N=S^2\\times S^1$, the minimum of $W(T^2)$ is 0, and when $N=R^2\\times S^1$, there exists no torus having least Willmore functional. When $N=H^2(-c)\\times S^1$, and $x=\\gamma\\times S^1$, the minimum of $W(x)$ is $2\\pi^2\\sqrt{c}$."
                    },
                    {
                        "title": "Virialization in Dark Energy Cosmology",
                        "abstract": "We discuss the issue of energy nonconservation in the virialzation process of spherical collapse model with homogeneous dark energy. We propose an approximation scheme to find the virialization radius. By comparing various schemes and estimating the parameter characterizing the ratio of dark energy to dark matter at the turn-around time, we conclude that the problem of energy nonconservation may have sizable effects in fitting models to observations."
                    },
                    {
                        "title": "Horizon Entropy in Modified Gravity",
                        "abstract": "We present an observation about the proposal that four-dimensional modification of general relativity may explain the observed cosmic acceleration today. Assuming that the thermodynamical nature of gravity theory continues to hold in modified gravity theories, we derive the modified horizon entropy formula from the modified Friedmann equation. We argue that our results imply that there are conceptual problems in some models of four-dimensional modification of general relativity."
                    },
                    {
                        "title": "Blaschke's problem for timelike surfaces in pseudo-Riemannian space forms",
                        "abstract": "We show that isothermic surfaces and S-Willmore surfaces are also the solutions to the corresponding Blaschke's problem for both spacelike and timelike surfaces in pseudo-Riemannian space forms. For timelike surfaces both Willmore and isothermic, we obtain a description by minimal surfaces similar to the classical results of Thomsen."
                    },
                    {
                        "title": "Willmore surfaces in spheres via loop groups II: a coarse classification of Willmore two-spheres by potentials",
                        "abstract": "Applying the DPW version of the theory developed by Burstall and Guest for harmonic maps of finite uniton type, we derive a coarse classification of Willmore two-spheres in $S^{n+2}$ in terms of the normalized potential of their (harmonic) conformal Gauss maps. Moreover, for the case of $S^6$, some geometric properties of the corresponding Willmore two-spheres are discussed. The classical classification of Willmore two-spheres in $S^4$ are also derived as a corollary."
                    },
                    {
                        "title": "Willmore surfaces in spheres via loop groups IV: on totally isotropic Willmore two-spheres in $S^6$",
                        "abstract": "Totally isotropic surfaces in $S^6$ are not necessarily Willmore surfaces. Therefore it is the first goal of this paper to derive a geometric characterization of totally isotropic Willmore two-spheres in $S^6$. This will naturally yield to a description of such surfaces in terms of the loop group language. Moreover, applying the loop group method, we also obtain an algorithm to construct totally isotropic Willmore two-spheres in $S^6$. As an application, a new, totally isotropic Willmore two-sphere which is not S-Willmore (i.e., has no dual surface) in $S^6$ is constructed, illustrating the theory of this paper."
                    },
                    {
                        "title": "Construction of Willmore two-spheres via harmonic maps into $SO^+(1,n+3)/(SO^+(1,1)\\times SO(n+2))$",
                        "abstract": "This paper aims to provide a description of totally isotropic Willmore two-spheres and their adjoint transforms. We first recall the isotropic harmonic maps which are introduced by H\\'elein, Xia-Shen and Ma for the study of Willmore surfaces. Then we derive a description of the normalized potential (some Lie algebra valued meromorphic 1-forms) of totally isotropic Willmore two-spheres in terms of the isotropic harmonic maps. In particular, the corresponding isotropic harmonic maps are of finite uniton type. The proof also contains a concrete way to construct examples of totally isotropic Willmore two-spheres and their adjoint transforms. As illustrations, two kinds of examples are obtained this way."
                    },
                    {
                        "title": "Willmore surfaces in spheres via loop groups III: on minimal surfaces in space forms",
                        "abstract": "The family of Willmore immersions from a Riemann surface into $S^{n+2}$ can be divided naturally into the subfamily of Willmore surfaces conformally equivalent to a minimal surface in $\\R^{n+2}$ and those which are not conformally equivalent to a minimal surface in $\\R^{n+2}$. On the level of their conformal Gauss maps into $Gr_{1,3}(\\R^{1,n+3})=SO^+(1,n+3)/SO^+(1,3)\\times SO(n)$ these two classes of Willmore immersions into $S^{n+2}$ correspond to conformally harmonic maps for which every image point, considered as a 4-dimensional Lorentzian subspace of $\\R^{1,n+3}$, contains a fixed lightlike vector or where it does not contain such a \"constant lightlike vector\". Using the loop group formalism for the construction of Willmore immersions we characterize in this paper precisely those normalized potentials which correspond to conformally harmonic maps containing a lightlike vector. Since the special form of these potentials can easily be avoided, we also precisely characterize those potentials which produce Willmore immersions into $ S^{n+2}$ which are not conformal to a minimal surface in $\\R^{n+2}$. It turns out that our proof also works analogously for minimal immersions into the other space forms."
                    },
                    {
                        "title": "A characterization of the Ejiri torus in $S^{5}$",
                        "abstract": "Ejiri's torus in $S^5$ is the first example of Willmore surface which is not conformally equivalent to any minimal surface in any space forms. Li and Vrancken classified all Willmore surfaces of tensor product in $S^{n}$ by reducing them into elastic curves in $S^3$, and the Ejiri torus appeared as a special example. In this paper, we first prove that among all Willmore tori of tensor product, the Willmore functional of the Ejiri torus in $S^5$ attains the minimum $2\\pi^2\\sqrt{3}$. Then we show that all Willmore tori of tensor product are unstable when the co-dimension is big enough. We also show that the Ejiri torus is unstable in $S^5$. Moreover, similar to Li and Vrancken, we classify all constrained Willmore surfaces of tensor product by reducing them with elastic curves in $S^3$. All constrained Willmore tori obtained this way are also shown to be unstable when the co-dimension is big enough.   We conjecture that a Willmore torus having Willmore functional between $2\\pi^2$ and $2\\pi^2\\sqrt{3}$ is either the Clifford torus, or the Ejiri torus."
                    },
                    {
                        "title": "Generalized polar transforms of spacelike isothermic surfaces",
                        "abstract": "In this paper, we generalize the polar transforms of spacelike isothermic surfaces in $Q^4_1$ to n-dimensional pseudo-Riemannian space forms $Q^n_r$. We show that there exist $c-$polar spacelike isothermic surfaces derived from a spacelike isothermic surface in $Q^n_r$, which are into $S^{n+1}_r(c)$, $H^{n+1}_{r-1}(c)$ or $Q^n_r$ depending on $c>0,<0,$ or $=0$. The $c-$polar isothermic surfaces can be characterized as generalized $H-$surfaces with null minimal sections. We also prove that if both the original surface and its $c-$polar surface are closed immersion, then they have the same Willmore functional. As examples, we discuss some product surfaces and compute the $c-$polar transforms of them. In the end, we derive the permutability theorems for $c-$polar transforms and Darboux transform and spectral transform of isothermic surfaces."
                    },
                    {
                        "title": "On Willmore surfaces in S^n of flat normal bundle",
                        "abstract": "We discuss several kinds of Willmore surfaces of flat normal bundle in this paper. First we show that every S-Willmore surface with flat normal bundle in $S^n$ must locate in some $S^3\\subset S^n$, from which we characterize Clifford torus as the only non-equatorial homogeneous minimal surface in $S^n$ with flat normal bundle, which improve a result of K. Yang. Then we derived that every Willmore two sphere with flat normal bundle in $S^n$ is conformal to a minimal surface with embedded planer ends in $\\mathbb{R}^3$. We also point out that for a class of Willmore tori, they have flat normal bundle if and only if they locate in some $S^3$. In the end, we show that a Willmore surface with flat normal bundle must locate in some $S^6$"
                    },
                    {
                        "title": "Matching Weak Informative Ontologies",
                        "abstract": "Most existing ontology matching methods utilize the literal information to discover alignments. However, some literal information in ontologies may be opaque and some ontologies may not have sufficient literal information. In this paper, these ontologies are named as weak informative ontologies (WIOs) and it is challenging for existing methods to matching WIOs. On one hand, string-based and linguistic-based matching methods cannot work well for WIOs. On the other hand, some matching methods use external resources to improve their performance, but collecting and processing external resources is still time-consuming. To address this issue, this paper proposes a practical method for matching WIOs by employing the ontology structure information to discover alignments. First, the semantic subgraphs are extracted from the ontology graph to capture the precise meanings of ontology elements. Then, a new similarity propagation model is designed for matching WIOs. Meanwhile, in order to avoid meaningless propagation, the similarity propagation is constrained by semantic subgraphs and other conditions. Consequently, the similarity propagation model ensures a balance between efficiency and quality during matching. Finally, the similarity propagation model uses a few credible alignments as seeds to find more alignments, and some useful strategies are adopted to improve the performance. This matching method for WIOs has been implemented in the ontology matching system Lily. Experimental results on public OAEI benchmark datasets demonstrate that Lily significantly outperforms most of the state-of-the-art works in both WIO matching tasks and general ontology matching tasks. In particular, Lily increases the recall by a large margin, while it still obtains high precision of matching results."
                    },
                    {
                        "title": "Understanding Social-Force Model in Psychological Principles of Collective Behavior",
                        "abstract": "To well understand crowd behavior, microscopic models have been developed in recent decades, in which an individual's behavioral/psychological status can be modeled and simulated. A well-known model is the social-force model innovated by physical scientists (Helbing and Molnar, 1995; Helbing, Farkas and Vicsek, 2000; Helbing et al., 2002). This model has been widely accepted and mainly used in simulation of crowd evacuation in the past decade. A problem, however, is that the testing results of the model were not explained in consistency with the psychological findings, resulting in misunderstanding of the model by psychologists. This paper will bridge the gap between psychological studies and physical explanation about this model. We reinterpret this physics-based model from a psychological perspective, clarifying that the model is consistent with psychological theories on stress, including time-related stress and interpersonal stress. Based on the conception of stress, we renew the model at both micro-and-macro level, referring to multi-agent simulation in a microscopic sense and fluid-based analysis in a macroscopic sense. The cognition and behavior of individual agents are critically modeled as response to environmental stimuli. Existing simulation results such as faster-is-slower effect will be reinterpreted by Yerkes-Dodson law, and herding and grouping effect as well as oscillation phenomenon are further discussed for pedestrian crowd. In brief the social-force model exhibits a bridge between the physics laws and psychological principles regarding crowd motion, and this paper will renew and reinterpret the model on the foundation of psychological studies."
                    }
                ]
            },
            "4436b2fa-0c1a-4477-98d4-00de34c6e7f1": {
                "pk": "4436b2fa-0c1a-4477-98d4-00de34c6e7f1",
                "name": "Yanning Zhang",
                "collaborators": [
                    "Congqi Cao",
                    "Aiqing Fang",
                    "Xinbo Zhao",
                    "Jiaqi Yang",
                    "Peng Wang",
                    "Yue Lu",
                    "Haichao Zhang",
                    "David Wipf",
                    "Lijun Zhang",
                    "Wei Suo"
                ],
                "domain": [
                    "Computer Vision",
                    "Image Processing",
                    "Deep Learning",
                    "Anomaly Detection"
                ],
                "publications": [
                    {
                        "title": "Learning to Compare Relation: Semantic Alignment for Few-Shot Learning",
                        "abstract": "Few-shot learning is a fundamental and challenging problem since it requires recognizing novel categories from only a few examples. The objects for recognition have multiple variants and can locate anywhere in images. Directly comparing query images with example images can not handle content misalignment. The representation and metric for comparison are critical but challenging to learn due to the scarcity and wide variation of the samples in few-shot learning. In this paper, we present a novel semantic alignment model to compare relations, which is robust to content misalignment. We propose to add two key ingredients to existing few-shot learning frameworks for better feature and metric learning ability. First, we introduce a semantic alignment loss to align the relation statistics of the features from samples that belong to the same category. And second, local and global mutual information maximization is introduced, allowing for representations that contain locally-consistent and intra-class shared information across structural locations in an image. Thirdly, we introduce a principled approach to weigh multiple loss functions by considering the homoscedastic uncertainty of each stream. We conduct extensive experiments on several few-shot learning datasets. Experimental results show that the proposed method is capable of comparing relations with semantic alignment strategies, and achieves state-of-the-art performance."
                    },
                    {
                        "title": "Image Super-Resolution via Sparse Bayesian Modeling of Natural Images",
                        "abstract": "Image super-resolution (SR) is one of the long-standing and active topics in image processing community. A large body of works for image super resolution formulate the problem with Bayesian modeling techniques and then obtain its Maximum-A-Posteriori (MAP) solution, which actually boils down to a regularized regression task over separable regularization term. Although straightforward, this approach cannot exploit the full potential offered by the probabilistic modeling, as only the posterior mode is sought. Also, the separable property of the regularization term can not capture any correlations between the sparse coefficients, which sacrifices much on its modeling accuracy. We propose a Bayesian image SR algorithm via sparse modeling of natural images. The sparsity property of the latent high resolution image is exploited by introducing latent variables into the high-order Markov Random Field (MRF) which capture the content adaptive variance by pixel-wise adaptation. The high-resolution image is estimated via Empirical Bayesian estimation scheme, which is substantially faster than our previous approach based on Markov Chain Monte Carlo sampling [1]. It is shown that the actual cost function for the proposed approach actually incorporates a non-factorial regularization term over the sparse coefficients. Experimental results indicate that the proposed method can generate competitive or better results than \\emph{state-of-the-art} SR algorithms."
                    },
                    {
                        "title": "A Plug-and-Play Method for Rare Human-Object Interactions Detection by Bridging Domain Gap",
                        "abstract": "Human-object interactions (HOI) detection aims at capturing human-object pairs in images and corresponding actions. It is an important step toward high-level visual reasoning and scene understanding. However, due to the natural bias from the real world, existing methods mostly struggle with rare human-object pairs and lead to sub-optimal results. Recently, with the development of the generative model, a straightforward approach is to construct a more balanced dataset based on a group of supplementary samples. Unfortunately, there is a significant domain gap between the generated data and the original data, and simply merging the generated images into the original dataset cannot significantly boost the performance. To alleviate the above problem, we present a novel model-agnostic framework called \\textbf{C}ontext-\\textbf{E}nhanced \\textbf{F}eature \\textbf{A}lignment (CEFA) module, which can effectively align the generated data with the original data at the feature level and bridge the domain gap. Specifically, CEFA consists of a feature alignment module and a context enhancement module. On one hand, considering the crucial role of human-object pairs information in HOI tasks, the feature alignment module aligns the human-object pairs by aggregating instance information. On the other hand, to mitigate the issue of losing important context information caused by the traditional discriminator-style alignment method, we employ a context-enhanced image reconstruction module to improve the model's learning ability of contextual cues. Extensive experiments have shown that our method can serve as a plug-and-play module to improve the detection performance of HOI models on rare categories\\footnote{https://github.com/LijunZhang01/CEFA}."
                    },
                    {
                        "title": "Inter-sphere consistency-based method for camera-projector pair calibration",
                        "abstract": "We construct constraints from consistency between estimated parameters from different spheres, termed inter-sphere consistency. It facilitates more flexible calibration using only two spheres, which has been considered a challenging and not well addressed ill-posed problem."
                    },
                    {
                        "title": "A Pulmonary Nodule Detection Model Based on Progressive Resolution and Hierarchical Saliency",
                        "abstract": "Detection of pulmonary nodules on chest CT is an essential step in the early diagnosis of lung cancer, which is critical for best patient care. Although a number of computer-aided nodule detection methods have been published in the literature, these methods still have two major drawbacks: missing out true nodules during the detection of nodule candidates and less-accurate identification of nodules from non-nodule. In this paper, we propose an automated pulmonary nodule detection algorithm that jointly combines progressive resolution and hierarchical saliency. Specifically, we design a 3D progressive resolution-based densely dilated FCN, namely the progressive resolution network (PRN), to detect nodule candidates inside the lung, and construct a densely dilated 3D CNN with hierarchical saliency, namely the hierarchical saliency network (HSN), to simultaneously identify genuine nodules from those candidates and estimate the diameters of nodules. We evaluated our algorithm on the benchmark LUng Nodule Analysis 2016 (LUNA16) dataset and achieved a state-of-the-art detection score. Our results suggest that the proposed algorithm can effectively detect pulmonary nodules on chest CT and accurately estimate their diameters."
                    },
                    {
                        "title": "Non-linear and Selective Fusion of Cross-Modal Images",
                        "abstract": "The human visual perception system has strong robustness in image fusion. This robustness is based on human visual perception system's characteristics of feature selection and non-linear fusion of different features. In order to simulate the human visual perception mechanism in image fusion tasks, we propose a multi-source image fusion framework that combines illuminance factors and attention mechanisms. The framework effectively combines traditional image features and modern deep learning features. First, we perform multi-scale decomposition of multi-source images. Then, the visual saliency map and the deep feature map are combined with the illuminance fusion factor to perform high-low frequency nonlinear fusion. Secondly, the characteristics of high and low frequency fusion are selected through the channel attention network to obtain the final fusion map. By simulating the nonlinear characteristics and selection characteristics of the human visual perception system in image fusion, the fused image is more in line with the human visual perception mechanism. Finally, we validate our fusion framework on public datasets of infrared and visible images, medical images and multi-focus images. The experimental results demonstrate the superiority of our fusion framework over state-of-arts in visual quality, objective fusion metrics and robustness."
                    },
                    {
                        "title": "IllumiNet: Transferring Illumination from Planar Surfaces to Virtual Objects in Augmented Reality",
                        "abstract": "This paper presents an illumination estimation method for virtual objects in real environment by learning. While previous works tackled this problem by reconstructing high dynamic range (HDR) environment maps or the corresponding spherical harmonics, we do not seek to recover the lighting environment of the entire scene. Given a single RGB image, our method directly infers the relit virtual object by transferring the illumination features extracted from planar surfaces in the scene to the desired geometries. Compared to previous works, our approach is more robust as it works in both indoor and outdoor environments with spatially-varying illumination. Experiments and evaluation results show that our approach outperforms the state-of-the-art quantitatively and qualitatively, achieving realistic augmented experience."
                    },
                    {
                        "title": "Dynamic Image Restoration and Fusion Based on Dynamic Degradation",
                        "abstract": "The deep-learning-based image restoration and fusion methods have achieved remarkable results. However, the existing restoration and fusion methods paid little research attention to the robustness problem caused by dynamic degradation. In this paper, we propose a novel dynamic image restoration and fusion neural network, termed as DDRF-Net, which is capable of solving two problems, i.e., static restoration and fusion, dynamic degradation. In order to solve the static fusion problem of existing methods, dynamic convolution is introduced to learn dynamic restoration and fusion weights. In addition, a dynamic degradation kernel is proposed to improve the robustness of image restoration and fusion. Our network framework can effectively combine image degradation with image fusion tasks, provide more detailed information for image fusion tasks through image restoration loss, and optimize image restoration tasks through image fusion loss. Therefore, the stumbling blocks of deep learning in image fusion, e.g., static fusion weight and specifically designed network architecture, are greatly mitigated. Extensive experiments show that our method is more superior compared with the state-of-the-art methods."
                    },
                    {
                        "title": "Cross-Modal Image Fusion Theory Guided by Subjective Visual Attention",
                        "abstract": "The human visual perception system has very strong robustness and contextual awareness in a variety of image processing tasks. This robustness and the perception ability of contextual awareness is closely related to the characteristics of multi-task auxiliary learning and subjective attention of the human visual perception system. In order to improve the robustness and contextual awareness of image fusion tasks, we proposed a multi-task auxiliary learning image fusion theory guided by subjective attention. The image fusion theory effectively unifies the subjective task intention and prior knowledge of human brain. In order to achieve our proposed image fusion theory, we first analyze the mechanism of multi-task auxiliary learning, build a multi-task auxiliary learning network. Secondly, based on the human visual attention perception mechanism, we introduce the human visual attention network guided by subjective tasks on the basis of the multi-task auxiliary learning network. The subjective intention is introduced by the subjective attention task model, so that the network can fuse images according to the subjective intention. Finally, in order to verify the superiority of our image fusion theory, we carried out experiments on the combined vision system image data set, and the infrared and visible image data set for experimental verification. The experimental results demonstrate the superiority of our fusion theory over state-of-arts in contextual awareness and robustness."
                    },
                    {
                        "title": "Learnable Locality-Sensitive Hashing for Video Anomaly Detection",
                        "abstract": "Video anomaly detection (VAD) mainly refers to identifying anomalous events that have not occurred in the training set where only normal samples are available. Existing works usually formulate VAD as a reconstruction or prediction problem. However, the adaptability and scalability of these methods are limited. In this paper, we propose a novel distance-based VAD method to take advantage of all the available normal data efficiently and flexibly. In our method, the smaller the distance between a testing sample and normal samples, the higher the probability that the testing sample is normal. Specifically, we propose to use locality-sensitive hashing (LSH) to map samples whose similarity exceeds a certain threshold into the same bucket in advance. In this manner, the complexity of near neighbor search is cut down significantly. To make the samples that are semantically similar get closer and samples not similar get further apart, we propose a novel learnable version of LSH that embeds LSH into a neural network and optimizes the hash functions with contrastive learning strategy. The proposed method is robust to data imbalance and can handle the large intra-class variations in normal data flexibly. Besides, it has a good ability of scalability. Extensive experiments demonstrate the superiority of our method, which achieves new state-of-the-art results on VAD benchmarks."
                    },
                    {
                        "title": "Context Recovery and Knowledge Retrieval: A Novel Two-Stream Framework for Video Anomaly Detection",
                        "abstract": "Video anomaly detection aims to find the events in a video that do not conform to the expected behavior. The prevalent methods mainly detect anomalies by snippet reconstruction or future frame prediction error. However, the error is highly dependent on the local context of the current snippet and lacks the understanding of normality. To address this issue, we propose to detect anomalous events not only by the local context, but also according to the consistency between the testing event and the knowledge about normality from the training data. Concretely, we propose a novel two-stream framework based on context recovery and knowledge retrieval, where the two streams can complement each other. For the context recovery stream, we propose a spatiotemporal U-Net which can fully utilize the motion information to predict the future frame. Furthermore, we propose a maximum local error mechanism to alleviate the problem of large recovery errors caused by complex foreground objects. For the knowledge retrieval stream, we propose an improved learnable locality-sensitive hashing, which optimizes hash functions via a Siamese network and a mutual difference loss. The knowledge about normality is encoded and stored in hash tables, and the distance between the testing event and the knowledge representation is used to reveal the probability of anomaly. Finally, we fuse the anomaly scores from the two streams to detect anomalies. Extensive experiments demonstrate the effectiveness and complementarity of the two streams, whereby the proposed two-stream framework achieves state-of-the-art performance on four datasets."
                    },
                    {
                        "title": "Generating Content for HDR Deghosting from Frequency View",
                        "abstract": "Recovering ghost-free High Dynamic Range (HDR) images from multiple Low Dynamic Range (LDR) images becomes challenging when the LDR images exhibit saturation and significant motion. Recent Diffusion Models (DMs) have been introduced in HDR imaging field, demonstrating promising performance, particularly in achieving visually perceptible results compared to previous DNN-based methods. However, DMs require extensive iterations with large models to estimate entire images, resulting in inefficiency that hinders their practical application. To address this challenge, we propose the Low-Frequency aware Diffusion (LF-Diff) model for ghost-free HDR imaging. The key idea of LF-Diff is implementing the DMs in a highly compacted latent space and integrating it into a regression-based model to enhance the details of reconstructed images. Specifically, as low-frequency information is closely related to human visual perception we propose to utilize DMs to create compact low-frequency priors for the reconstruction process. In addition, to take full advantage of the above low-frequency priors, the Dynamic HDR Reconstruction Network (DHRNet) is carried out in a regression-based manner to obtain final HDR images. Extensive experiments conducted on synthetic and real-world benchmark datasets demonstrate that our LF-Diff performs favorably against several state-of-the-art methods and is 10$\\times$ faster than previous DM-based methods."
                    },
                    {
                        "title": "3D Registration with Maximal Cliques",
                        "abstract": "As a fundamental problem in computer vision, 3D point cloud registration (PCR) aims to seek the optimal pose to align a point cloud pair. In this paper, we present a 3D registration method with maximal cliques (MAC). The key insight is to loosen the previous maximum clique constraint, and mine more local consensus information in a graph for accurate pose hypotheses generation: 1) A compatibility graph is constructed to render the affinity relationship between initial correspondences. 2) We search for maximal cliques in the graph, each of which represents a consensus set. We perform node-guided clique selection then, where each node corresponds to the maximal clique with the greatest graph weight. 3) Transformation hypotheses are computed for the selected cliques by the SVD algorithm and the best hypothesis is used to perform registration. Extensive experiments on U3M, 3DMatch, 3DLoMatch and KITTI demonstrate that MAC effectively increases registration accuracy, outperforms various state-of-the-art methods and boosts the performance of deep-learned methods. MAC combined with deep-learned methods achieves state-of-the-art registration recall of 95.7% / 78.9% on 3DMatch / 3DLoMatch."
                    },
                    {
                        "title": "Adaptive Graph Convolutional Networks for Weakly Supervised Anomaly Detection in Videos",
                        "abstract": "For weakly supervised anomaly detection, most existing work is limited to the problem of inadequate video representation due to the inability of modeling long-term contextual information. To solve this, we propose a novel weakly supervised adaptive graph convolutional network (WAGCN) to model the complex contextual relationship among video segments. By which, we fully consider the influence of other video segments on the current one when generating the anomaly probability score for each segment. Firstly, we combine the temporal consistency as well as feature similarity of video segments to construct a global graph, which makes full use of the association information among spatial-temporal features of anomalous events in videos. Secondly, we propose a graph learning layer in order to break the limitation of setting topology manually, which can extract graph adjacency matrix based on data adaptively and effectively. Extensive experiments on two public datasets (i.e., UCF-Crime dataset and ShanghaiTech dataset) demonstrate the effectiveness of our approach which achieves state-of-the-art performance."
                    },
                    {
                        "title": "Age-minimal Multicast by Graph Attention Reinforcement Learning",
                        "abstract": "Age of Information (AoI) is an emerging metric used to assess the timeliness of information, gaining research interest in real-time multicast applications such as video streaming and metaverse platforms. In this paper, we consider a dynamic multicast network with energy constraints, where our objective is to minimize the expected time-average AoI through energy-constrained multicast routing and scheduling. The inherent complexity of the problem, given the NP-hardness and intertwined scheduling and routing decisions, makes existing approaches inapplicable. To address these challenges, we decompose the original problem into two subtasks, each amenable to reinforcement learning (RL) methods. Subsequently, we propose an innovative framework based on graph attention networks (GATs) to effectively capture graph information with superior generalization capabilities. To validate our framework, we conduct experiments on three datasets including a real-world dataset called AS-733, and show that our proposed scheme reduces the average weighted AoI by 62.9% and reduces the energy consumption by at most 72.5% compared to baselines."
                    }
                ]
            }
        }
    },
    "2310.17191": {
        "paper_data": {
            "title": "How do Language Models Bind Entities in Context?",
            "url": "http://arxiv.org/abs/2310.17191v2",
            "arxiv_id": "2310.17191",
            "authors": [
                "Jiahai Feng",
                "Jacob Steinhardt"
            ],
            "abstract": "To correctly use in-context information, language models (LMs) must bind entities to their attributes. For example, given a context describing a \"green square\" and a \"blue circle\", LMs must bind the shapes to their respective colors. We analyze LM representations and identify the binding ID mechanism: a general mechanism for solving the binding problem, which we observe in every sufficiently large model from the Pythia and LLaMA families. Using causal interventions, we show that LMs' internal activations represent binding information by attaching binding ID vectors to corresponding entities and attributes. We further show that binding ID vectors form a continuous subspace, in which distances between binding ID vectors reflect their discernability. Overall, our results uncover interpretable strategies in LMs for representing symbolic knowledge in-context, providing a step towards understanding general in-context reasoning in large-scale LMs.",
            "introduction": "   1 Introduction  Modern language models (LMs) excel at many reasoning benchmarks, suggesting that they can perform general purpose reasoning across many domains. However, the mechanisms that underlie LM reasoning remain largely unknown (R\u00e4uker et\u00a0al., 2023). The deployment of LMs in society has led to calls to better understand these mechanisms (Hendrycks et\u00a0al., 2021), so as to know why they work and when they fail (Mu & Andreas, 2020; Hernandez et\u00a0al., 2021; Vig et\u00a0al., 2020b).   In this work, we seek to understand binding, a foundational skill that underlies many compositional reasoning capabilities (Fodor & Pylyshyn, 1988) such as entity tracking (Kim & Schuster, 2023). How humans solve binding, i.e. recognize features of an object as bound to that object and not to others, is a fundamental problem in psychology\u00a0(Treisman, 1996). Here, we study binding in LMs.   Binding arises any time the LM has to reason about two or more objects of the same kind. For example, consider the following passage involving two people and two countries:       Context: Alice lives in the capital city of France. Bob lives in the capital city of Thailand.   Question: Which city does Bob live in?   (1)    In this example the LM has to represent the associations lives(Alice, Paris) and lives(Bob, Bangkok). We call this the binding problem\u2014for the predicate lives, Alice is bound to Paris and Bob to Bangkok. Since predicates are bound in-context, binding must occur in the activations, rather than in the weights as with factual recall (Meng et\u00a0al., 2022). This raises the question: how do LMs represent binding information in the context such that they can be later recalled?   Overall, our key technical contribution is the identification of a robust general mechanism in LMs for solving the binding problem. The mechanism relies on binding IDs, which are abstract concepts that LMs use internally to mark variables in the same predicate apart from variables in other predicates (Fig. 1). Using causal mediation analysis we empirically verify two key properties of the binding ID mechanism (section\u00a03): factorizability and position independence.   Turning to the structure of binding IDs, we find that binding IDs are represented as vectors which are bound to variables by simple addition (section\u00a04) in the activation space. Further, we show that binding IDs occupy a subspace, in the sense that linear combinations of binding IDs are still valid binding IDs, even though random vectors are not.   Lastly, we find that binding IDs are ubiquitous and transferable (section\u00a05). They are used by every sufficiently large model in the LLaMA (Touvron et\u00a0al., 2023) and Pythia (Biderman et\u00a0al., 2023) families, and their fidelity increases with scale. They are used for a variety of synthetic binding tasks with different surface forms, and binding vectors from one task transfer to other tasks. Finally, we qualify our findings by showing that despite their ubiquity, binding IDs are not universal: we exhibit a question-answering task where an alternate mechanism, \u201cdirect binding\u201d, is used instead (appendix\u00a0E). We release code and datasets here: https://github.com/jiahai-feng/binding-iclr   Figure 1: The Binding ID mechanism. The LM learns abstract binding IDs (drawn as triangles or squares) which distinguish between entity-attribute pairs. Binding functions \u0393Esubscript\u0393\ud835\udc38\\Gamma_{E}roman_\u0393 start_POSTSUBSCRIPT italic_E end_POSTSUBSCRIPT and \u0393Asubscript\u0393\ud835\udc34\\Gamma_{A}roman_\u0393",
            "references": [
                {
                    "title": "Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking",
                    "abstract": "Fine-tuning on generalized tasks such as instruction following, code generation, and mathematics has been shown to enhance language models' performance on a range of tasks. Nevertheless, explanations of how such fine-tuning influences the internal computations in these models remain elusive. We study how fine-tuning affects the internal mechanisms implemented in language models. As a case study, we explore the property of entity tracking, a crucial facet of language comprehension, where models fine-tuned on mathematics have substantial performance gains. We identify the mechanism that enables entity tracking and show that (i) in both the original model and its fine-tuned versions primarily the same circuit implements entity tracking. In fact, the entity tracking circuit of the original model on the fine-tuned versions performs better than the full original model. (ii) The circuits of all the models implement roughly the same functionality: Entity tracking is performed by tracking the position of the correct entity in both the original model and its fine-tuned versions. (iii) Performance boost in the fine-tuned models is primarily attributed to its improved ability to handle the augmented positional information. To uncover these findings, we employ: Patch Patching, DCM, which automatically detects model components responsible for specific semantics, and CMAP, a new approach for patching activations across models to reveal improved mechanisms. Our findings suggest that fine-tuning enhances, rather than fundamentally alters, the mechanistic operation of the model."
                },
                {
                    "title": "Linearity of Relation Decoding in Transformer Language Models",
                    "abstract": "Much of the knowledge encoded in transformer language models (LMs) may be expressed in terms of relations: relations between words and their synonyms, entities and their attributes, etc. We show that, for a subset of relations, this computation is well-approximated by a single linear transformation on the subject representation. Linear relation representations may be obtained by constructing a first-order approximation to the LM from a single prompt, and they exist for a variety of factual, commonsense, and linguistic relations. However, we also identify many cases in which LM predictions capture relational knowledge accurately, but this knowledge is not linearly encoded in their representations. Our results thus reveal a simple, interpretable, but heterogeneously deployed knowledge representation strategy in transformer LMs."
                },
                {
                    "title": "How Far Can Camels Go? Exploring the State of Instruction Tuning on Open Resources",
                    "abstract": "In this work we explore recent advances in instruction-tuning language models on a range of open instruction-following datasets. Despite recent claims that open models can be on par with state-of-the-art proprietary models, these claims are often accompanied by limited evaluation, making it difficult to compare models across the board and determine the utility of various resources. We provide a large set of instruction-tuned models from 6.7B to 65B parameters in size, trained on 12 instruction datasets ranging from manually curated (e.g., OpenAssistant) to synthetic and distilled (e.g., Alpaca) and systematically evaluate them on their factual knowledge, reasoning, multilinguality, coding, and open-ended instruction following abilities through a collection of automatic, model-based, and human-based metrics. We further introduce T\\\"ulu, our best performing instruction-tuned model suite finetuned on a combination of high-quality open resources. Our experiments show that different instruction-tuning datasets can uncover or enhance specific skills, while no single dataset (or combination) provides the best performance across all evaluations. Interestingly, we find that model and human preference-based evaluations fail to reflect differences in model capabilities exposed by benchmark-based evaluations, suggesting the need for the type of systemic evaluation performed in this work. Our evaluations show that the best model in any given evaluation reaches on average 87% of ChatGPT performance, and 73% of GPT-4 performance, suggesting that further investment in building better base models and instruction-tuning data is required to close the gap. We release our instruction-tuned models, including a fully finetuned 65B T\\\"ulu, along with our code, data, and evaluation framework at https://github.com/allenai/open-instruct to facilitate future research."
                },
                {
                    "title": "Interpretability at Scale: Identifying Causal Mechanisms in Alpaca",
                    "abstract": "Obtaining human-interpretable explanations of large, general-purpose language models is an urgent goal for AI safety. However, it is just as important that our interpretability methods are faithful to the causal dynamics underlying model behavior and able to robustly generalize to unseen inputs. Distributed Alignment Search (DAS) is a powerful gradient descent method grounded in a theory of causal abstraction that has uncovered perfect alignments between interpretable symbolic algorithms and small deep learning models fine-tuned for specific tasks. In the present paper, we scale DAS significantly by replacing the remaining brute-force search steps with learned parameters -- an approach we call Boundless DAS. This enables us to efficiently search for interpretable causal structure in large language models while they follow instructions. We apply Boundless DAS to the Alpaca model (7B parameters), which, off the shelf, solves a simple numerical reasoning problem. With Boundless DAS, we discover that Alpaca does this by implementing a causal model with two interpretable boolean variables. Furthermore, we find that the alignment of neural representations with these variables is robust to changes in inputs and instructions. These findings mark a first step toward faithfully understanding the inner-workings of our ever-growing and most widely deployed language models. Our tool is extensible to larger LLMs and is released publicly at `https://github.com/stanfordnlp/pyvene`."
                },
                {
                    "title": "Entity Tracking in Language Models",
                    "abstract": "Keeping track of how states of entities change as a text or dialog unfolds is a key prerequisite to discourse understanding. Yet, there have been few systematic investigations into the ability of large language models (LLMs) to track discourse entities. In this work, we present a task probing to what extent a language model can infer the final state of an entity given an English description of the initial state and a series of state-changing operations. We use this task to first investigate whether Flan-T5, GPT-3 and GPT-3.5 can track the state of entities, and find that only GPT-3.5 models, which have been pretrained on large amounts of code, exhibit this ability. We then investigate whether smaller models pretrained primarily on text can learn to track entities, through finetuning T5 on several training/evaluation splits. While performance degrades for more complex splits, we find that even when evaluated on a different set of entities from training or longer operation sequences, a finetuned model can perform non-trivial entity tracking. Taken together, these results suggest that language models can learn to track entities but pretraining on text corpora alone does not make this capacity surface."
                },
                {
                    "title": "Dissecting Recall of Factual Associations in Auto-Regressive Language Models",
                    "abstract": "Transformer-based language models (LMs) are known to capture factual knowledge in their parameters. While previous work looked into where factual associations are stored, only little is known about how they are retrieved internally during inference. We investigate this question through the lens of information flow. Given a subject-relation query, we study how the model aggregates information about the subject and relation to predict the correct attribute. With interventions on attention edges, we first identify two critical points where information propagates to the prediction: one from the relation positions followed by another from the subject positions. Next, by analyzing the information at these points, we unveil a three-step internal mechanism for attribute extraction. First, the representation at the last-subject position goes through an enrichment process, driven by the early MLP sublayers, to encode many subject-related attributes. Second, information from the relation propagates to the prediction. Third, the prediction representation\"queries\"the enriched subject to extract the attribute. Perhaps surprisingly, this extraction is typically done via attention heads, which often encode subject-attribute mappings in their parameters. Overall, our findings introduce a comprehensive view of how factual associations are stored and extracted internally in LMs, facilitating future research on knowledge localization and editing."
                },
                {
                    "title": "Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling",
                    "abstract": "How do large language models (LLMs) develop and evolve over the course of training? How do these patterns change as models scale? To answer these questions, we introduce \\textit{Pythia}, a suite of 16 LLMs all trained on public data seen in the exact same order and ranging in size from 70M to 12B parameters. We provide public access to 154 checkpoints for each one of the 16 models, alongside tools to download and reconstruct their exact training dataloaders for further study. We intend \\textit{Pythia} to facilitate research in many areas, and we present several case studies including novel results in memorization, term frequency effects on few-shot performance, and reducing gender bias. We demonstrate that this highly controlled setup can be used to yield novel insights toward LLMs and their training dynamics. Trained models, analysis code, training code, and training data can be found at \\url{https://github.com/EleutherAI/pythia}."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Progress measures for grokking via mechanistic interpretability",
                    "abstract": "Neural networks often exhibit emergent behavior, where qualitatively new capabilities arise from scaling up the amount of parameters, training data, or training steps. One approach to understanding emergence is to find continuous \\textit{progress measures} that underlie the seemingly discontinuous qualitative changes. We argue that progress measures can be found via mechanistic interpretability: reverse-engineering learned behaviors into their individual components. As a case study, we investigate the recently-discovered phenomenon of ``grokking'' exhibited by small transformers trained on modular addition tasks. We fully reverse engineer the algorithm learned by these networks, which uses discrete Fourier transforms and trigonometric identities to convert addition to rotation about a circle. We confirm the algorithm by analyzing the activations and weights and by performing ablations in Fourier space. Based on this understanding, we define progress measures that allow us to study the dynamics of training and split training into three continuous phases: memorization, circuit formation, and cleanup. Our results show that grokking, rather than being a sudden shift, arises from the gradual amplification of structured mechanisms encoded in the weights, followed by the later removal of memorizing components."
                },
                {
                    "title": "Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small",
                    "abstract": "Research in mechanistic interpretability seeks to explain behaviors of machine learning models in terms of their internal components. However, most previous work either focuses on simple behaviors in small models, or describes complicated behaviors in larger models with broad strokes. In this work, we bridge this gap by presenting an explanation for how GPT-2 small performs a natural language task called indirect object identification (IOI). Our explanation encompasses 26 attention heads grouped into 7 main classes, which we discovered using a combination of interpretability approaches relying on causal interventions. To our knowledge, this investigation is the largest end-to-end attempt at reverse-engineering a natural behavior\"in the wild\"in a language model. We evaluate the reliability of our explanation using three quantitative criteria--faithfulness, completeness and minimality. Though these criteria support our explanation, they also point to remaining gaps in our understanding. Our work provides evidence that a mechanistic understanding of large ML models is feasible, opening opportunities to scale our understanding to both larger models and more complex tasks."
                },
                {
                    "title": "Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task",
                    "abstract": "Language models show a surprising range of capabilities, but the source of their apparent competence is unclear. Do these networks just memorize a collection of surface statistics, or do they rely on internal representations of the process that generates the sequences they see? We investigate this question by applying a variant of the GPT model to the task of predicting legal moves in a simple board game, Othello. Although the network has no a priori knowledge of the game or its rules, we uncover evidence of an emergent nonlinear internal representation of the board state. Interventional experiments indicate this representation can be used to control the output of the network and create\"latent saliency maps\"that can help explain predictions in human terms."
                },
                {
                    "title": "Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networks",
                    "abstract": "The last decade of machine learning has seen drastic increases in scale and capabilities. Deep neural networks (DNNs) are increasingly being deployed in the real world. However, they are difficult to analyze, raising concerns about using them without a rigorous understanding of how they function. Effective tools for interpreting them will be important for building more trustworthy AI by helping to identify problems, fix bugs, and improve basic understanding. In particular, \u201cinner\u201d interpretability techniques, which focus on explaining the internal components of DNNs, are well-suited for developing a mechanistic understanding, guiding manual modifications, and reverse engineering solutions. Much recent work has focused on DNN interpretability, and rapid progress has thus far made a thorough systematization of methods difficult. In this survey, we review over 300 works with a focus on inner interpretability tools. We introduce a taxonomy that classifies methods by what part of the network they help to explain (weights, neurons, subnetworks, or latent representations) and whether they are implemented during (intrinsic) or after (post hoc) training. To our knowledge, we are also the first to survey a number of connections between interpretability research and work in adversarial robustness, continual learning, modularity, network compression, and studying the human visual system. We discuss key challenges and argue that the status quo in interpretability research is largely unproductive. Finally, we highlight the importance of future work that emphasizes diagnostics, debugging, adversaries, and benchmarking in order to make interpretability tools more useful to engineers in practical applications."
                },
                {
                    "title": "Locating and Editing Factual Associations in GPT",
                    "abstract": "We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's factual predictions. This reveals a distinct set of steps in middle-layer feed-forward modules that mediate factual predictions while processing subject tokens. To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update specific factual associations using Rank-One Model Editing (ROME). We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another. Our results confirm an important role for mid-layer feed-forward modules in storing factual associations and suggest that direct manipulation of computational mechanisms may be a feasible approach for model editing. The code, dataset, visualizations, and an interactive demo notebook are available at https://rome.baulab.info/"
                },
                {
                    "title": "Natural Language Descriptions of Deep Visual Features",
                    "abstract": "Some neurons in deep networks specialize in recognizing highly specific perceptual, structural, or semantic features of inputs. In computer vision, techniques exist for identifying neurons that respond to individual concept categories like colors, textures, and object classes. But these techniques are limited in scope, labeling only a small subset of neurons and behaviors in any network. Is a richer characterization of neuron-level computation possible? We introduce a procedure (called MILAN, for mutual-information-guided linguistic annotation of neurons) that automatically labels neurons with open-ended, compositional, natural language descriptions. Given a neuron, MILAN generates a description by searching for a natural language string that maximizes pointwise mutual information with the image regions in which the neuron is active. MILAN produces fine-grained descriptions that capture categorical, relational, and logical structure in learned features. These descriptions obtain high agreement with human-generated feature descriptions across a diverse set of model architectures and tasks, and can aid in understanding and controlling learned models. We highlight three applications of natural language neuron descriptions. First, we use MILAN for analysis, characterizing the distribution and importance of neurons selective for attribute, category, and relational information in vision models. Second, we use MILAN for auditing, surfacing neurons sensitive to human faces in datasets designed to obscure them. Finally, we use MILAN for editing, improving robustness in an image classifier by deleting neurons sensitive to text features spuriously correlated with class labels."
                },
                {
                    "title": "Unsolved Problems in ML Safety",
                    "abstract": "Machine learning (ML) systems are rapidly increasing in size, are acquiring new capabilities, and are increasingly deployed in high-stakes settings. As with other powerful technologies, safety for ML should be a leading research priority. In response to emerging safety challenges in ML, such as those introduced by recent large-scale models, we provide a new roadmap for ML Safety and refine the technical problems that the field needs to address. We present four problems ready for research, namely withstanding hazards (\"Robustness\"), identifying hazards (\"Monitoring\"), reducing inherent model hazards (\"Alignment\"), and reducing systemic hazards (\"Systemic Safety\"). Throughout, we clarify each problem's motivation and provide concrete research directions."
                },
                {
                    "title": "Causal Abstractions of Neural Networks",
                    "abstract": "Structural analysis methods (e.g., probing and feature attribution) are increasingly important tools for neural network analysis. We propose a new structural analysis method grounded in a formal theory of causal abstraction that provides rich characterizations of model-internal representations and their roles in input/output behavior. In this method, neural representations are aligned with variables in interpretable causal models, and then interchange interventions are used to experimentally verify that the neural representations have the causal properties of their aligned variables. We apply this method in a case study to analyze neural models trained on Multiply Quantified Natural Language Inference (MQNLI) corpus, a highly complex NLI dataset that was constructed with a tree-structured natural logic causal model. We discover that a BERT-based model with state-of-the-art performance successfully realizes parts of the natural logic model's causal structure, whereas a simpler baseline model fails to show any such structure, demonstrating that BERT representations encode the compositional structure of MQNLI."
                },
                {
                    "title": "Implicit Representations of Meaning in Neural Language Models",
                    "abstract": "Does the effectiveness of neural language models derive entirely from accurate modeling of surface word co-occurrence statistics, or do these models represent and reason about the world they describe? In BART and T5 transformer language models, we identify contextual word representations that function as *models of entities and situations* as they evolve throughout a discourse. These neural representations have functional similarities to linguistic models of dynamic semantics: they support a linear readout of each entity\u2019s current properties and relations, and can be manipulated with predictable effects on language generation. Our results indicate that prediction in pretrained neural language models is supported, at least in part, by dynamic representations of meaning and implicit simulation of entity state, and that this behavior can be learned with only text as training data."
                },
                {
                    "title": "Knowledge Neurons in Pretrained Transformers",
                    "abstract": "Large-scale pretrained language models are surprisingly good at recalling factual knowledge presented in the training corpus. In this paper, we present preliminary studies on how factual knowledge is stored in pretrained Transformers by introducing the concept of knowledge neurons. Specifically, we examine the fill-in-the-blank cloze task for BERT. Given a relational fact, we propose a knowledge attribution method to identify the neurons that express the fact. We find that the activation of such knowledge neurons is positively correlated to the expression of their corresponding facts. In our case studies, we attempt to leverage knowledge neurons to edit (such as update, and erase) specific factual knowledge without fine-tuning. Our results shed light on understanding the storage of knowledge within pretrained Transformers."
                },
                {
                    "title": "Calibrate Before Use: Improving Few-Shot Performance of Language Models",
                    "abstract": "GPT-3 can perform numerous tasks when provided a natural language prompt that contains a few training examples. We show that this type of few-shot learning can be unstable: the choice of prompt format, training examples, and even the order of the training examples can cause accuracy to vary from near chance to near state-of-the-art. We demonstrate that this instability arises from the bias of language models towards predicting certain answers, e.g., those that are placed near the end of the prompt or are common in the pre-training data. To mitigate this, we first estimate the model's bias towards each answer by asking for its prediction when given the training prompt and a content-free test input such as \"N/A\". We then fit calibration parameters that cause the prediction for this input to be uniform across answers. On a diverse set of tasks, this contextual calibration procedure substantially improves GPT-3 and GPT-2's average accuracy (up to 30.0% absolute) and reduces variance across different choices of the prompt."
                },
                {
                    "title": "Transformer Feed-Forward Layers Are Key-Value Memories",
                    "abstract": "Feed-forward layers constitute two-thirds of a transformer model\u2019s parameters, yet their role in the network remains under-explored. We show that feed-forward layers in transformer-based language models operate as key-value memories, where each key correlates with textual patterns in the training examples, and each value induces a distribution over the output vocabulary. Our experiments show that the learned patterns are human-interpretable, and that lower layers tend to capture shallow patterns, while upper layers learn more semantic ones. The values complement the keys\u2019 input patterns by inducing output distributions that concentrate probability mass on tokens likely to appear immediately after each pattern, particularly in the upper layers. Finally, we demonstrate that the output of a feed-forward layer is a composition of its memories, which is subsequently refined throughout the model\u2019s layers via residual connections to produce the final output distribution."
                },
                {
                    "title": "Compositional Explanations of Neurons",
                    "abstract": "We describe a procedure for explaining neurons in deep representations by identifying compositional logical concepts that closely approximate neuron behavior. Compared to prior work that uses atomic labels as explanations, analyzing neurons compositionally allows us to more precisely and expressively characterize their behavior. We use this procedure to answer several questions on interpretability in models for vision and natural language processing. First, we examine the kinds of abstractions learned by neurons. In image classification, we find that many neurons learn highly abstract but semantically coherent visual concepts, while other polysemantic neurons detect multiple unrelated features; in natural language inference (NLI), neurons learn shallow lexical heuristics from dataset biases. Second, we see whether compositional explanations give us insight into model performance: vision neurons that detect human-interpretable concepts are positively correlated with task performance, while NLI neurons that fire for shallow heuristics are negatively correlated with task performance. Finally, we show how compositional explanations provide an accessible way for end users to produce simple \"copy-paste\" adversarial examples that change model behavior in predictable ways."
                },
                {
                    "title": "A Primer in BERTology: What We Know About How BERT Works",
                    "abstract": "Abstract Transformer-based models have pushed state of the art in many areas of NLP, but our understanding of what is behind their success is still limited. This paper is the first survey of over 150 studies of the popular BERT model. We review the current state of knowledge about how BERT works, what kind of information it learns and how it is represented, common modifications to its training objectives and architecture, the overparameterization issue, and approaches to compression. We then outline directions for future research."
                },
                {
                    "title": "Designing and Interpreting Probes with Control Tasks",
                    "abstract": "Probes, supervised models trained to predict properties (like parts-of-speech) from representations (like ELMo), have achieved high accuracy on a range of linguistic tasks. But does this mean that the representations encode linguistic structure or just that the probe has learned the linguistic task? In this paper, we propose control tasks, which associate word types with random outputs, to complement linguistic tasks. By construction, these tasks can only be learned by the probe itself. So a good probe, (one that reflects the representation), should be selective, achieving high linguistic task accuracy and low control task accuracy. The selectivity of a probe puts linguistic task accuracy in context with the probe\u2019s capacity to memorize from word types. We construct control tasks for English part-of-speech tagging and dependency edge prediction, and show that popular probes on ELMo representations are not selective. We also find that dropout, commonly used to control probe complexity, is ineffective for improving selectivity of MLPs, but that other forms of regularization are effective. Finally, we find that while probes on the first layer of ELMo yield slightly better part-of-speech tagging accuracy than the second, probes on the second layer are substantially more selective, which raises the question of which layer better represents parts-of-speech."
                },
                {
                    "title": "BERT Rediscovers the Classical NLP Pipeline",
                    "abstract": "Pre-trained text encoders have rapidly advanced the state of the art on many NLP tasks. We focus on one such model, BERT, and aim to quantify where linguistic information is captured within the network. We find that the model represents the steps of the traditional NLP pipeline in an interpretable and localizable way, and that the regions responsible for each step appear in the expected sequence: POS tagging, parsing, NER, semantic roles, then coreference. Qualitative analysis reveals that the model can and often does adjust this pipeline dynamically, revising lower-level decisions on the basis of disambiguating information from higher-level representations."
                },
                {
                    "title": "Bias in Bios: A Case Study of Semantic Representation Bias in a High-Stakes Setting",
                    "abstract": "We present a large-scale study of gender bias in occupation classification, a task where the use of machine learning may lead to negative outcomes on peoples' lives. We analyze the potential allocation harms that can result from semantic representation bias. To do so, we study the impact on occupation classification of including explicit gender indicators---such as first names and pronouns---in different semantic representations of online biographies. Additionally, we quantify the bias that remains when these indicators are \"scrubbed,\" and describe proxy behavior that occurs in the absence of explicit gender indicators. As we demonstrate, differences in true positive rates between genders are correlated with existing gender imbalances in occupations, which may compound these imbalances."
                },
                {
                    "title": "Analysis Methods in Neural Language Processing: A Survey",
                    "abstract": "The field of natural language processing has seen impressive progress in recent years, with neural network models replacing many of the traditional systems. A plethora of new models have been proposed, many of which are thought to be opaque compared to their feature-rich counterparts. This has led researchers to analyze, interpret, and evaluate neural networks in novel and more fine-grained ways. In this survey paper, we review analysis methods in neural language processing, categorize them according to prominent research trends, highlight existing limitations, and point to potential directions for future work."
                },
                {
                    "title": "Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank",
                    "abstract": "Semantic word spaces have been very useful but cannot express the meaning of longer phrases in a principled way. Further progress towards understanding compositionality in tasks such as sentiment detection requires richer supervised training and evaluation resources and more powerful models of composition. To remedy this, we introduce a Sentiment Treebank. It includes fine grained sentiment labels for 215,154 phrases in the parse trees of 11,855 sentences and presents new challenges for sentiment compositionality. To address them, we introduce the Recursive Neural Tensor Network. When trained on the new treebank, this model outperforms all previous methods on several metrics. It pushes the state of the art in single sentence positive/negative classification from 80% up to 85.4%. The accuracy of predicting fine-grained sentiment labels for all phrases reaches 80.7%, an improvement of 9.7% over bag of features baselines. Lastly, it is the only model that can accurately capture the effects of negation and its scope at various tree levels for both positive and negative phrases."
                },
                {
                    "title": "Linguistic Regularities in Continuous Space Word Representations",
                    "abstract": "Continuous space language models have recently demonstrated outstanding results across a variety of tasks. In this paper, we examine the vector-space word representations that are implicitly learned by the input-layer weights. We find that these representations are surprisingly good at capturing syntactic and semantic regularities in language, and that each relationship is characterized by a relation-specific vector offset. This allows vector-oriented reasoning based on the offsets between words. For example, the male/female relationship is automatically learned, and with the induced vector representations, \u201cKing Man + Woman\u201d results in a vector very close to \u201cQueen.\u201d We demonstrate that the word vectors capture syntactic regularities by means of syntactic analogy questions (provided with this paper), and are able to correctly answer almost 40% of the questions. We demonstrate that the word vectors capture semantic regularities by using the vector offset method to answer SemEval-2012 Task 2 questions. Remarkably, this method outperforms the best previous systems."
                },
                {
                    "title": "Measuring and Manipulating Knowledge Representations in Language Models",
                    "abstract": "Neural language models (LMs) represent facts about the world described by text. Sometimes these facts derive from training data (in most LMs, a representation of the word banana encodes the fact that bananas are fruits). Sometimes facts derive from input text itself (a representation of the sentence I poured out the bottle encodes the fact that the bottle became empty). Tools for inspecting and modifying LM fact representations would be useful almost everywhere LMs are used: making it possible to update them when the world changes, to localize and remove sources of bias, and to identify errors in generated text. We describe REMEDI , an approach for querying and modifying factual knowledge in LMs. REMEDI learns a map from textual queries to fact encodings in an LM\u2019s internal representation system. These encodings can be used as knowledge editors : by adding them to LM hidden representations, we can modify downstream generation to be consistent with new facts. REMEDI encodings can also be used as model probes : by comparing them to LM representations, we can ascertain what properties LMs attribute to mentioned entities, and predict when they will generate outputs that con\ufb02ict with background knowledge or input text. REMEDI thus links work on probing, prompting, and model editing, and offers steps toward general tools for \ufb01ne-grained inspection and control of knowledge in LMs."
                },
                {
                    "title": "Mapping Language Models to Grounded Conceptual Spaces",
                    "abstract": "A fundamental criticism of text-only language models (LMs) is their lack of grounding \u2014that is, the ability to tie a word for which they have learned a representation to its referent in the non-linguistic world. However, despite this limitation, large pre-trained LMs have been shown to have a remarkable grasp of the conceptual structure of language, as demonstrated by their ability to answer questions, generate \ufb02uent text, or make inferences about entities, objects, and properties that they have never physically observed. In this work we investigate the extent to which the rich conceptual structure that LMs learn indeed re\ufb02ects the conceptual structure of the non-linguistic world\u2014which is something that LMs have never observed. We do this by testing whether the LMs can learn to map an entire conceptual domain (e.g., direction or colour) onto a grounded world representation given only a small number of examples. For example, we show a model what the word \u201cleft\u201d means using a textual depiction of a grid world, and assess how well it can generalise to related concepts, for example, the word \u201cright\u201d , in a similar grid world. We investigate a range of generative language models of varying sizes (including GPT-2 and GPT-3), and see that although the smaller models struggle to perform this mapping, the largest model can not only learn to ground the concepts that it is explicitly taught, but appears to generalise to several instances of unseen concepts as well. Our results suggest an alternative means of building grounded language models: rather than learning grounded representations \u201cfrom scratch\u201d, it is possible that large text-only models learn a suf\ufb01ciently rich conceptual structure that could allow them to be grounded in a data-ef\ufb01cient way."
                },
                {
                    "title": "Investigating Gender Bias in Language Models Using Causal Mediation Analysis",
                    "abstract": "Many interpretation methods for neural models in natural language processing investigate how information is encoded inside hidden representations. However, these methods can only measure whether the information exists, not whether it is actually used by the model. We propose a methodology grounded in the theory of causal mediation analysis for interpreting which parts of a model are causally implicated in its behavior. The approach enables us to analyze the mechanisms that facilitate the \ufb02ow of information from input to output through various model components, known as mediators. As a case study, we apply this methodology to analyzing gender bias in pre-trained Transformer language models. We study the role of individual neurons and attention heads in mediating gender bias across three datasets designed to gauge a model\u2019s sensitivity to gender bias. Our mediation analysis reveals that gender bias effects are concentrated in speci\ufb01c components of the model that may exhibit highly specialized behavior."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow do language models represent and utilize binding information in the context of compositional reasoning?\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding the binding mechanisms in language models is crucial for advancing the field of machine learning, particularly in natural language processing. By elucidating how LMs handle binding, we can improve their reasoning capabilities, leading to more reliable and interpretable models. This research could pave the way for enhanced applications in various domains, such as AI-driven dialogue systems, automated reasoning, and cognitive modeling, ultimately influencing future research directions in both theoretical and practical aspects of AI.\n\n**[Question 3] - Why is it hard?**  \nThe binding problem is complex due to the need for language models to manage multiple entities and their relationships simultaneously. Naive approaches may fail because they do not account for the contextual nature of binding, where associations must be dynamically represented in the model's activations rather than relying solely on static weights. Additionally, the challenge lies in empirically verifying the mechanisms of binding, as it requires sophisticated analysis techniques and a deep understanding of the model's internal representations.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often overlooked the specific mechanisms underlying binding in language models, focusing instead on broader performance metrics. Limitations in empirical methodologies and a lack of targeted investigations into the internal workings of LMs have hindered progress. Our approach differs by employing causal mediation analysis to identify and verify the properties of binding IDs, providing a clearer understanding of how these mechanisms operate within LMs, which has not been thoroughly explored in prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves analyzing the binding ID mechanism in language models through causal mediation analysis to verify its properties, such as factorizability and position independence. We will utilize datasets from the LLaMA and Pythia families to investigate the representation of binding IDs as vectors in the activation space. The expected outcomes include a comprehensive understanding of how binding IDs function, their transferability across tasks, and the identification of scenarios where alternative mechanisms may be employed, ultimately contributing to the broader understanding of reasoning in language models."
            }
        },
        "author_data": {
            "f0f51e3c-4e11-41b6-8a96-3bbd09ec5efd": {
                "pk": "f0f51e3c-4e11-41b6-8a96-3bbd09ec5efd",
                "name": "Jiahai Feng",
                "collaborators": [
                    "Stuart Russell",
                    "Jacob Steinhardt"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Interpretability",
                    "Bias Detection"
                ],
                "publications": [
                    {
                        "title": "Monitoring Latent World States in Language Models with Propositional Probes",
                        "abstract": "Language models are susceptible to bias, sycophancy, backdoors, and other tendencies that lead to unfaithful responses to the input context. Interpreting internal states of language models could help monitor and correct unfaithful behavior. We hypothesize that language models represent their input contexts in a latent world model, and seek to extract this latent world state from the activations. We do so with 'propositional probes', which compositionally probe tokens for lexical information and bind them into logical propositions representing the world state. For example, given the input context ''Greg is a nurse. Laura is a physicist.'', we decode the propositions ''WorksAs(Greg, nurse)'' and ''WorksAs(Laura, physicist)'' from the model's activations. Key to this is identifying a 'binding subspace' in which bound tokens have high similarity (''Greg'' and ''nurse'') but unbound ones do not (''Greg'' and ''physicist''). We validate propositional probes in a closed-world setting with finitely many predicates and properties. Despite being trained on simple templated contexts, propositional probes generalize to contexts rewritten as short stories and translated to Spanish. Moreover, we find that in three settings where language models respond unfaithfully to the input context -- prompt injections, backdoor attacks, and gender bias -- the decoded propositions remain faithful. This suggests that language models often encode a faithful world model but decode it unfaithfully, which motivates the search for better interpretability tools for monitoring LMs."
                    }
                ]
            },
            "05af9677-56e6-4013-9b0d-2f195f8cfd46": {
                "pk": "05af9677-56e6-4013-9b0d-2f195f8cfd46",
                "name": "Jacob Steinhardt",
                "collaborators": [
                    "Jiahai Feng",
                    "Stuart Russell"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Interpretability",
                    "Bias Detection"
                ],
                "publications": [
                    {
                        "title": "Monitoring Latent World States in Language Models with Propositional Probes",
                        "abstract": "Language models are susceptible to bias, sycophancy, backdoors, and other tendencies that lead to unfaithful responses to the input context. Interpreting internal states of language models could help monitor and correct unfaithful behavior. We hypothesize that language models represent their input contexts in a latent world model, and seek to extract this latent world state from the activations. We do so with 'propositional probes', which compositionally probe tokens for lexical information and bind them into logical propositions representing the world state. For example, given the input context ''Greg is a nurse. Laura is a physicist.'', we decode the propositions ''WorksAs(Greg, nurse)'' and ''WorksAs(Laura, physicist)'' from the model's activations. Key to this is identifying a 'binding subspace' in which bound tokens have high similarity (''Greg'' and ''nurse'') but unbound ones do not (''Greg'' and ''physicist''). We validate propositional probes in a closed-world setting with finitely many predicates and properties. Despite being trained on simple templated contexts, propositional probes generalize to contexts rewritten as short stories and translated to Spanish. Moreover, we find that in three settings where language models respond unfaithfully to the input context -- prompt injections, backdoor attacks, and gender bias -- the decoded propositions remain faithful. This suggests that language models often encode a faithful world model but decode it unfaithfully, which motivates the search for better interpretability tools for monitoring LMs."
                    }
                ]
            }
        }
    },
    "2404.16767": {
        "paper_data": {
            "title": "REBEL: Reinforcement Learning via Regressing Relative Rewards",
            "url": "http://arxiv.org/abs/2404.16767v3",
            "arxiv_id": "2404.16767",
            "authors": [
                "Zhaolin Gao",
                "Jonathan D. Chang",
                "Wenhao Zhan",
                "Owen Oertell",
                "Gokul Swamy",
                "Kiant\u00e9 Brantley",
                "Thorsten Joachims",
                "J. Andrew Bagnell",
                "Jason D. Lee",
                "Wen Sun"
            ],
            "abstract": "While originally developed for continuous control problems, Proximal Policy Optimization (PPO) has emerged as the work-horse of a variety of reinforcement learning (RL) applications, including the fine-tuning of generative models. Unfortunately, PPO requires multiple heuristics to enable stable convergence (e.g. value networks, clipping), and is notorious for its sensitivity to the precise implementation of these components. In response, we take a step back and ask what a minimalist RL algorithm for the era of generative models would look like. We propose REBEL, an algorithm that cleanly reduces the problem of policy optimization to regressing the relative reward between two completions to a prompt in terms of the policy, enabling strikingly lightweight implementation. In theory, we prove that fundamental RL algorithms like Natural Policy Gradient can be seen as variants of REBEL, which allows us to match the strongest known theoretical guarantees in terms of convergence and sample complexity in the RL literature. REBEL can also cleanly incorporate offline data and be extended to handle the intransitive preferences we frequently see in practice. Empirically, we find that REBEL provides a unified approach to language modeling and image generation with stronger or similar performance as PPO and DPO, all while being simpler to implement and more computationally efficient than PPO. When fine-tuning Llama-3-8B-Instruct, REBEL achieves strong performance in AlpacaEval 2.0, MT-Bench, and Open LLM Leaderboard.",
            "introduction": "   1 Introduction  The generality of the reinforcement learning (RL) paradigm is striking: from continuous control problems (Kalashnikov et\u00a0al., 2018) to, more recently, the fine-tuning of generative models (Stiennon et\u00a0al., 2022; Ouyang et\u00a0al., 2022), RL has enabled concrete progress across a variety of decision-making tasks. Specifically, when it comes to fine-tuning generative models, Proximal Policy Optimization (PPO, Schulman et\u00a0al. (2017)) has emerged as the de-facto RL algorithm of choice, from language models (LLMs) (Ziegler et\u00a0al., 2020; Stiennon et\u00a0al., 2022; Ouyang et\u00a0al., 2022; Touvron et\u00a0al., 2023) to generative image models (Black et\u00a0al., 2023; Fan et\u00a0al., 2024; Oertell et\u00a0al., 2024).   If we take a step back however, it is odd that we are using an algorithm designed for optimizing two-layer networks for continuous control tasks from scratch, even though we are now fine-tuning generative models with billions of parameters. In the continuous control setting, the randomly initialized neural networks and the possible stochasticity in the dynamics necessitate variance reduction through a learned value function as a baseline (Schulman et\u00a0al., 2015), while clipping updates is important to limit distribution shift from iteration to iteration (Kakade and Langford, 2002). This means that when applied to generative model fine-tuning, we need to store four models in memory simultaneously (the policy, the reference policy, the critic, and the reward model), each with billions of parameters. Furthermore, we often add a KL regularization to the base model for fine-tuning, making explicit clipping unnecessary nor advisable, as pointed out by Ahmadian et\u00a0al. (2024). Even outside of the generative modeling context, PPO is notorious for the wide range of performances measured, with differences being attributed to seemingly inconsequential implementation details (Henderson et\u00a0al., 2019; Engstrom et\u00a0al., 2020). This begs the question: are there simpler algorithms that better scale to modern RL applications?   Figure 1: We present REBEL: a simple and scalable RL algorithm that performs policy optimization via iteratively regressing the difference in rewards in terms of the policy, allowing us to eliminate much of the complexity (e.g. value functions, clipping) of algorithms like PPO (Schulman et\u00a0al., 2017). We apply REBEL to problems in both image generation and language modeling and find that despite its conceptual and implementation-level simplicity, REBEL is able to match or sometimes outperform the performance of PPO while out-performing purely offline techniques like DPO (Rafailov et\u00a0al., 2023). REBEL also achieves strong performance on common benchmarks such as AlpacaEval when fine-tuning a Llama-3-8B model.   Our answer is REBEL: an algorithm that reduces the problem of RL to solving a sequence of squared loss regression problems on iteratively collected datasets. Each regression problem directly uses the policy to predict the difference in rewards. This allows us to eliminate the complexity of using value functions, avoids heuristics like clipping, and scales easily to problems in both language modeling and image generation. Our key insight is that a regressor that can predict the difference in rewards between trajectories in a dataset implicitly captures an improved policy.   Rather than being a mere heuristic, REBEL comes with strong guarantees in theory and can be seen as a strict generalization of classical techniques (e.g., NPG) in reinforcement learning. Furthermore, REBEL",
            "references": [
                {
                    "title": "Interpretable Preferences via Multi-Objective Reward Modeling and Mixture-of-Experts",
                    "abstract": "Reinforcement learning from human feedback (RLHF) has emerged as the primary method for aligning large language models (LLMs) with human preferences. The RLHF process typically starts by training a reward model (RM) using human preference data. Conventional RMs are trained on pairwise responses to the same user request, with relative ratings indicating which response humans prefer. The trained RM serves as a proxy for human preferences. However, due to the black-box nature of RMs, their outputs lack interpretability, as humans cannot intuitively understand why an RM thinks a response is good or not. As RMs act as human preference proxies, we believe they should be human-interpretable to ensure that their internal decision processes are consistent with human preferences and to prevent reward hacking in LLM alignment. To build RMs with interpretable preferences, we propose a two-stage approach: i) train an Absolute-Rating Multi-Objective Reward Model (ArmoRM) with multi-dimensional absolute-rating data, each dimension corresponding to a human-interpretable objective (e.g., honesty, verbosity, safety); ii) employ a Mixture-of-Experts (MoE) strategy with a gating network that automatically selects the most suitable reward objectives based on the context. We efficiently trained an ArmoRM with Llama-3 8B and a gating network consisting of a shallow MLP on top of the ArmoRM. Our trained model, ArmoRM-Llama3-8B, obtains state-of-the-art performance on RewardBench, a benchmark evaluating RMs for language modeling. Notably, the performance of our model surpasses the LLM-as-a-judge method with GPT-4 judges by a margin, and approaches the performance of the much larger Nemotron-4 340B reward model."
                },
                {
                    "title": "From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipeline",
                    "abstract": "The rapid evolution of Large Language Models (LLMs) has outpaced the development of model evaluation, highlighting the need for continuous curation of new, challenging benchmarks. However, manual curation of high-quality, human-aligned benchmarks is expensive and time-consuming. To address this, we introduce BenchBuilder, an automated pipeline that leverages LLMs to curate high-quality, open-ended prompts from large, crowd-sourced datasets, enabling continuous benchmark updates without human in the loop. We apply BenchBuilder to datasets such as Chatbot Arena and WildChat-1M, extracting challenging prompts and utilizing LLM-as-a-Judge for automatic model evaluation. To validate benchmark quality, we propose new metrics to measure a benchmark's alignment with human preferences and ability to separate models. We release Arena-Hard-Auto, a benchmark consisting 500 challenging prompts curated by BenchBuilder. Arena-Hard-Auto provides 3x higher separation of model performances compared to MT-Bench and achieves 98.6% correlation with human preference rankings, all at a cost of $20. Our work sets a new framework for the scalable curation of automated benchmarks from extensive data."
                },
                {
                    "title": "Self-Play Preference Optimization for Language Model Alignment",
                    "abstract": "Standard reinforcement learning from human feedback (RLHF) approaches relying on parametric models like the Bradley-Terry model fall short in capturing the intransitivity and irrationality in human preferences. Recent advancements suggest that directly working with preference probabilities can yield a more accurate reflection of human preferences, enabling more flexible and accurate language model alignment. In this paper, we propose a self-play-based method for language model alignment, which treats the problem as a constant-sum two-player game aimed at identifying the Nash equilibrium policy. Our approach, dubbed Self-Play Preference Optimization (SPPO), utilizes iterative policy updates to provably approximate the Nash equilibrium. Additionally, we propose a new SPPO objective which is both strongly motivated by theory and is simple and effective in practice. In our experiments, using only 60k prompts (without responses) from the UltraFeedback dataset and without any prompt augmentation, by leveraging a pre-trained preference model PairRM with only 0.4B parameters, SPPO can obtain a model from fine-tuning Mistral-7B-Instruct-v0.2 that achieves the state-of-the-art length-controlled win-rate of 28.53% against GPT-4-Turbo on AlpacaEval 2.0. It also outperforms the (iterative) DPO and IPO on MT-Bench, Arena-Hard, and the Open LLM Leaderboard. Starting from a stronger base model Llama-3-8B-Instruct, we are able to achieve a length-controlled win rate of 38.77%. Notably, the strong performance of SPPO is achieved without additional external supervision (e.g., responses, preferences, etc.) from GPT-4 or other stronger language models. Codes are available at https://github.com/uclaml/SPPO."
                },
                {
                    "title": "Preference Fine-Tuning of LLMs Should Leverage Suboptimal, On-Policy Data",
                    "abstract": "Learning from preference labels plays a crucial role in fine-tuning large language models. There are several distinct approaches for preference fine-tuning, including supervised learning, on-policy reinforcement learning (RL), and contrastive learning. Different methods come with different implementation tradeoffs and performance differences, and existing empirical findings present different conclusions, for instance, some results show that online RL is quite important to attain good fine-tuning results, while others find (offline) contrastive or even purely supervised methods sufficient. This raises a natural question: what kind of approaches are important for fine-tuning with preference data and why? In this paper, we answer this question by performing a rigorous analysis of a number of fine-tuning techniques on didactic and full-scale LLM problems. Our main finding is that, in general, approaches that use on-policy sampling or attempt to push down the likelihood on certain responses (i.e., employ a\"negative gradient\") outperform offline and maximum likelihood objectives. We conceptualize our insights and unify methods that use on-policy sampling or negative gradient under a notion of mode-seeking objectives for categorical distributions. Mode-seeking objectives are able to alter probability mass on specific bins of a categorical distribution at a fast rate compared to maximum likelihood, allowing them to relocate masses across bins more effectively. Our analysis prescribes actionable insights for preference fine-tuning of LLMs and informs how data should be collected for maximal improvement."
                },
                {
                    "title": "Dataset Reset Policy Optimization for RLHF",
                    "abstract": "Reinforcement Learning (RL) from Human Preference-based feedback is a popular paradigm for fine-tuning generative models, which has produced impressive models such as GPT-4 and Claude3 Opus. This framework often consists of two steps: learning a reward model from an offline preference dataset followed by running online RL to optimize the learned reward model. In this work, leveraging the idea of reset, we propose a new RLHF algorithm with provable guarantees. Motivated by the fact that offline preference dataset provides informative states (i.e., data that is preferred by the labelers), our new algorithm, Dataset Reset Policy Optimization (DR-PO), integrates the existing offline preference dataset into the online policy training procedure via dataset reset: it directly resets the policy optimizer to the states in the offline dataset, instead of always starting from the initial state distribution. In theory, we show that DR-PO learns to perform at least as good as any policy that is covered by the offline dataset under general function approximation with finite sample complexity. In experiments, we demonstrate that on both the TL;DR summarization and the Anthropic Helpful Harmful (HH) dataset, the generation from DR-PO is better than that from Proximal Policy Optimization (PPO) and Direction Preference Optimization (DPO), under the metric of GPT4 win-rate. Code for this work can be found at https://github.com/Cornell-RL/drpo."
                },
                {
                    "title": "Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators",
                    "abstract": "LLM-based auto-annotators have become a key component of the LLM development process due to their cost-effectiveness and scalability compared to human-based evaluation. However, these auto-annotators can introduce complex biases that are hard to remove. Even simple, known confounders such as preference for longer outputs remain in existing automated evaluation metrics. We propose a simple regression analysis approach for controlling biases in auto-evaluations. As a real case study, we focus on reducing the length bias of AlpacaEval, a fast and affordable benchmark for chat LLMs that uses LLMs to estimate response quality. Despite being highly correlated with human preferences, AlpacaEval is known to favor models that generate longer outputs. We introduce a length-controlled AlpacaEval that aims to answer the counterfactual question:\"What would the preference be if the model's and baseline's output had the same length?\". To achieve this, we first fit a generalized linear model to predict the biased output of interest (auto-annotator preferences) based on the mediators we want to control for (length difference) and other relevant features. We then obtain length-controlled preferences by predicting preferences while conditioning the GLM with a zero difference in lengths. Length-controlling not only improves the robustness of the metric to manipulations in model verbosity, we also find that it increases the Spearman correlation with LMSYS' Chatbot Arena from 0.94 to 0.98. We release the code and leaderboard at https://tatsu-lab.github.io/alpaca_eval/ ."
                },
                {
                    "title": "Direct Nash Optimization: Teaching Language Models to Self-Improve with General Preferences",
                    "abstract": "This paper studies post-training large language models (LLMs) using preference feedback from a powerful oracle to help a model iteratively improve over itself. The typical approach for post-training LLMs involves Reinforcement Learning from Human Feedback (RLHF), which traditionally separates reward learning and subsequent policy optimization. However, such a reward maximization approach is limited by the nature of\"point-wise\"rewards (such as Bradley-Terry model), which fails to express complex intransitive or cyclic preference relations. While advances on RLHF show reward learning and policy optimization can be merged into a single contrastive objective for stability, they yet still remain tethered to the reward maximization framework. Recently, a new wave of research sidesteps the reward maximization presumptions in favor of directly optimizing over\"pair-wise\"or general preferences. In this paper, we introduce Direct Nash Optimization (DNO), a provable and scalable algorithm that marries the simplicity and stability of contrastive learning with theoretical generality from optimizing general preferences. Because DNO is a batched on-policy algorithm using a regression-based objective, its implementation is straightforward and efficient. Moreover, DNO enjoys monotonic improvement across iterations that help it improve even over a strong teacher (such as GPT-4). In our experiments, a resulting 7B parameter Orca-2.5 model aligned by DNO achieves the state-of-the-art win-rate against GPT-4-Turbo of 33% on AlpacaEval 2.0 (even after controlling for response length), an absolute gain of 26% (7% to 33%) over the initializing model. It outperforms models with far more parameters, including Mistral Large, Self-Rewarding LM (70B parameters), and older versions of GPT-4."
                },
                {
                    "title": "RL for Consistency Models: Faster Reward Guided Text-to-Image Generation",
                    "abstract": "Reinforcement learning (RL) has improved guided image generation with diffusion models by directly optimizing rewards that capture image quality, aesthetics, and instruction following capabilities. However, the resulting generative policies inherit the same iterative sampling process of diffusion models that causes slow generation. To overcome this limitation, consistency models proposed learning a new class of generative models that directly map noise to data, resulting in a model that can generate an image in as few as one sampling iteration. In this work, to optimize text-to-image generative models for task specific rewards and enable fast training and inference, we propose a framework for fine-tuning consistency models via RL. Our framework, called Reinforcement Learning for Consistency Model (RLCM), frames the iterative inference process of a consistency model as an RL procedure. Comparing to RL finetuned diffusion models, RLCM trains significantly faster, improves the quality of the generation measured under the reward objectives, and speeds up the inference procedure by generating high quality images with as few as two inference steps. Experimentally, we show that RLCM can adapt text-to-image consistency models to objectives that are challenging to express with prompting, such as image compressibility, and those derived from human feedback, such as aesthetic quality. Our code is available at https://rlcm.owenoertell.com."
                },
                {
                    "title": "The N+ Implementation Details of RLHF with PPO: A Case Study on TL;DR Summarization",
                    "abstract": "This work is the first to openly reproduce the Reinforcement Learning from Human Feedback (RLHF) scaling behaviors reported in OpenAI's seminal TL;DR summarization work. We create an RLHF pipeline from scratch, enumerate over 20 key implementation details, and share key insights during the reproduction. Our RLHF-trained Pythia models demonstrate significant gains in response quality that scale with model size, with our 2.8B, 6.9B models outperforming OpenAI's released 1.3B checkpoint. We publicly release the trained model checkpoints and code to facilitate further research and accelerate progress in the field (\\url{https://github.com/vwxyzjn/summarize_from_feedback_details})."
                },
                {
                    "title": "RewardBench: Evaluating Reward Models for Language Modeling",
                    "abstract": "Reward models (RMs) are at the crux of successfully using RLHF to align pretrained models to human preferences, yet there has been relatively little study that focuses on evaluation of those models. Evaluating reward models presents an opportunity to understand the opaque technologies used for alignment of language models and which values are embedded in them. Resources for reward model training and understanding are sparse in the nascent open-source community around them. To enhance scientific understanding of reward models, we present RewardBench, a benchmark dataset and code-base for evaluation. The RewardBench dataset is a collection of prompt-chosen-rejected trios spanning chat, reasoning, and safety, to benchmark how reward models perform on challenging, structured and out-of-distribution queries. We create specific comparison datasets for RMs that have subtle, but verifiable reasons (e.g. bugs, incorrect facts) why one answer should be preferred to another. On the RewardBench leaderboard, we evaluate reward models trained with a variety of methods, such as the direct MLE training of classifiers and the implicit reward modeling of Direct Preference Optimization (DPO). We present many findings on propensity for refusals, reasoning limitations, and instruction following shortcomings of various reward models towards a better understanding of the RLHF process."
                },
                {
                    "title": "Human Alignment of Large Language Models through Online Preference Optimisation",
                    "abstract": "Ensuring alignment of language models' outputs with human preferences is critical to guarantee a useful, safe, and pleasant user experience. Thus, human alignment has been extensively studied recently and several methods such as Reinforcement Learning from Human Feedback (RLHF), Direct Policy Optimisation (DPO) and Sequence Likelihood Calibration (SLiC) have emerged. In this paper, our contribution is two-fold. First, we show the equivalence between two recent alignment methods, namely Identity Policy Optimisation (IPO) and Nash Mirror Descent (Nash-MD). Second, we introduce a generalisation of IPO, named IPO-MD, that leverages the regularised sampling approach proposed by Nash-MD. This equivalence may seem surprising at first sight, since IPO is an offline method whereas Nash-MD is an online method using a preference model. However, this equivalence can be proven when we consider the online version of IPO, that is when both generations are sampled by the online policy and annotated by a trained preference model. Optimising the IPO loss with such a stream of data becomes then equivalent to finding the Nash equilibrium of the preference model through self-play. Building on this equivalence, we introduce the IPO-MD algorithm that generates data with a mixture policy (between the online and reference policy) similarly as the general Nash-MD algorithm. We compare online-IPO and IPO-MD to different online versions of existing losses on preference data such as DPO and SLiC on a summarisation task."
                },
                {
                    "title": "Stop Regressing: Training Value Functions via Classification for Scalable Deep RL",
                    "abstract": "Value functions are a central component of deep reinforcement learning (RL). These functions, parameterized by neural networks, are trained using a mean squared error regression objective to match bootstrapped target values. However, scaling value-based RL methods that use regression to large networks, such as high-capacity Transformers, has proven challenging. This difficulty is in stark contrast to supervised learning: by leveraging a cross-entropy classification loss, supervised methods have scaled reliably to massive networks. Observing this discrepancy, in this paper, we investigate whether the scalability of deep RL can also be improved simply by using classification in place of regression for training value functions. We demonstrate that value functions trained with categorical cross-entropy significantly improves performance and scalability in a variety of domains. These include: single-task RL on Atari 2600 games with SoftMoEs, multi-task RL on Atari with large-scale ResNets, robotic manipulation with Q-transformers, playing Chess without search, and a language-agent Wordle task with high-capacity Transformers, achieving state-of-the-art results on these domains. Through careful analysis, we show that the benefits of categorical cross-entropy primarily stem from its ability to mitigate issues inherent to value-based RL, such as noisy targets and non-stationarity. Overall, we argue that a simple shift to training value functions with categorical cross-entropy can yield substantial improvements in the scalability of deep RL at little-to-no cost."
                },
                {
                    "title": "Don't Forget Your Reward Values: Language Model Alignment via Value-based Calibration",
                    "abstract": "While Reinforcement Learning from Human Feedback (RLHF) significantly enhances the generation quality of Large Language Models (LLMs), recent studies have raised concerns regarding the complexity and instability associated with the Proximal Policy Optimization (PPO) algorithm, proposing a series of order-based calibration methods as viable alternatives. This paper delves further into current order-based methods, examining their inefficiencies in utilizing reward values and addressing misalignment issues. Building upon these findings, we propose a novel \\textbf{V}alue-based \\textbf{C}ali\\textbf{B}ration (VCB) method to better align LLMs with human preferences. Experimental results demonstrate that VCB surpasses existing alignment methods on AI assistant and summarization datasets, providing impressive generalizability, robustness, and stability in diverse settings."
                },
                {
                    "title": "Hybrid Inverse Reinforcement Learning",
                    "abstract": "The inverse reinforcement learning approach to imitation learning is a double-edged sword. On the one hand, it can enable learning from a smaller number of expert demonstrations with more robustness to error compounding than behavioral cloning approaches. On the other hand, it requires that the learner repeatedly solve a computationally expensive reinforcement learning (RL) problem. Often, much of this computation is wasted searching over policies very dissimilar to the expert's. In this work, we propose using hybrid RL -- training on a mixture of online and expert data -- to curtail unnecessary exploration. Intuitively, the expert data focuses the learner on good states during training, which reduces the amount of exploration required to compute a strong policy. Notably, such an approach doesn't need the ability to reset the learner to arbitrary states in the environment, a requirement of prior work in efficient inverse RL. More formally, we derive a reduction from inverse RL to expert-competitive RL (rather than globally optimal RL) that allows us to dramatically reduce interaction during the inner policy search loop while maintaining the benefits of the IRL approach. This allows us to derive both model-free and model-based hybrid inverse RL algorithms with strong policy performance guarantees. Empirically, we find that our approaches are significantly more sample efficient than standard inverse RL and several other baselines on a suite of continuous control tasks."
                },
                {
                    "title": "More Benefits of Being Distributional: Second-Order Bounds for Reinforcement Learning",
                    "abstract": "In this paper, we prove that Distributional Reinforcement Learning (DistRL), which learns the return distribution, can obtain second-order bounds in both online and offline RL in general settings with function approximation. Second-order bounds are instance-dependent bounds that scale with the variance of return, which we prove are tighter than the previously known small-loss bounds of distributional RL. To the best of our knowledge, our results are the first second-order bounds for low-rank MDPs and for offline RL. When specializing to contextual bandits (one-step RL problem), we show that a distributional learning based optimism algorithm achieves a second-order worst-case regret bound, and a second-order gap dependent bound, simultaneously. We also empirically demonstrate the benefit of DistRL in contextual bandits on real-world datasets. We highlight that our analysis with DistRL is relatively simple, follows the general framework of optimism in the face of uncertainty and does not require weighted regression. Our results suggest that DistRL is a promising framework for obtaining second-order bounds in general RL settings, thus further reinforcing the benefits of DistRL."
                },
                {
                    "title": "Direct Language Model Alignment from Online AI Feedback",
                    "abstract": "Direct alignment from preferences (DAP) methods, such as DPO, have recently emerged as efficient alternatives to reinforcement learning from human feedback (RLHF), that do not require a separate reward model. However, the preference datasets used in DAP methods are usually collected ahead of training and never updated, thus the feedback is purely offline. Moreover, responses in these datasets are often sampled from a language model distinct from the one being aligned, and since the model evolves over training, the alignment phase is inevitably off-policy. In this study, we posit that online feedback is key and improves DAP methods. Our method, online AI feedback (OAIF), uses an LLM as annotator: on each training iteration, we sample two responses from the current model and prompt the LLM annotator to choose which one is preferred, thus providing online feedback. Despite its simplicity, we demonstrate via human evaluation in several tasks that OAIF outperforms both offline DAP and RLHF methods. We further show that the feedback leveraged in OAIF is easily controllable, via instruction prompts to the LLM annotator."
                },
                {
                    "title": "A Minimaximalist Approach to Reinforcement Learning from Human Feedback",
                    "abstract": "We present Self-Play Preference Optimization (SPO), an algorithm for reinforcement learning from human feedback. Our approach is minimalist in that it does not require training a reward model nor unstable adversarial training and is therefore rather simple to implement. Our approach is maximalist in that it provably handles non-Markovian, intransitive, and stochastic preferences while being robust to the compounding errors that plague offline approaches to sequential prediction. To achieve the preceding qualities, we build upon the concept of a Minimax Winner (MW), a notion of preference aggregation from the social choice theory literature that frames learning from preferences as a zero-sum game between two policies. By leveraging the symmetry of this game, we prove that rather than using the traditional technique of dueling two policies to compute the MW, we can simply have a single agent play against itself while maintaining strong convergence guarantees. Practically, this corresponds to sampling multiple trajectories from a policy, asking a preference or teacher model to compare them, and then using the proportion of wins as the reward for a particular trajectory. We demonstrate that on a suite of continuous control tasks, we are able to learn significantly more efficiently than reward-model based approaches while maintaining robustness to the intransitive and stochastic preferences that frequently occur in practice when aggregating human judgments."
                },
                {
                    "title": "Iterative Preference Learning from Human Feedback: Bridging Theory and Practice for RLHF under KL-constraint",
                    "abstract": "This paper studies the alignment process of generative models with Reinforcement Learning from Human Feedback (RLHF). We first identify the primary challenges of existing popular methods like offline PPO and offline DPO as lacking in strategical exploration of the environment. Then, to understand the mathematical principle of RLHF, we consider a standard mathematical formulation, the reverse-KL regularized contextual bandit for RLHF. Despite its widespread practical application, a rigorous theoretical analysis of this formulation remains open. We investigate its behavior in three distinct settings -- offline, online, and hybrid -- and propose efficient algorithms with finite-sample theoretical guarantees. Moving towards practical applications, our framework, with a robust approximation of the information-theoretical policy improvement oracle, naturally gives rise to several novel RLHF algorithms. This includes an iterative version of the Direct Preference Optimization (DPO) algorithm for online settings, and a multi-step rejection sampling strategy for offline scenarios. Our empirical evaluations on real-world alignment experiment of large language model demonstrate that these proposed methods significantly surpass existing strong baselines, such as DPO and Rejection Sampling Optimization (RSO), showcasing the connections between solid theoretical foundations and their potent practical implementations."
                },
                {
                    "title": "Nash Learning from Human Feedback",
                    "abstract": "Reinforcement learning from human feedback (RLHF) has emerged as the main paradigm for aligning large language models (LLMs) with human preferences. Typically, RLHF involves the initial step of learning a reward model from human feedback, often expressed as preferences between pairs of text generations produced by a pre-trained LLM. Subsequently, the LLM's policy is fine-tuned by optimizing it to maximize the reward model through a reinforcement learning algorithm. However, an inherent limitation of current reward models is their inability to fully represent the richness of human preferences and their dependency on the sampling distribution. In this study, we introduce an alternative pipeline for the fine-tuning of LLMs using pairwise human feedback. Our approach entails the initial learning of a preference model, which is conditioned on two inputs given a prompt, followed by the pursuit of a policy that consistently generates responses preferred over those generated by any competing policy, thus defining the Nash equilibrium of this preference model. We term this approach Nash learning from human feedback (NLHF). In the context of a tabular policy representation, we present a novel algorithmic solution, Nash-MD, founded on the principles of mirror descent. This algorithm produces a sequence of policies, with the last iteration converging to the regularized Nash equilibrium. Additionally, we explore parametric representations of policies and introduce gradient descent algorithms for deep-learning architectures. To demonstrate the effectiveness of our approach, we present experimental results involving the fine-tuning of a LLM for a text summarization task. We believe NLHF offers a compelling avenue for preference learning and policy optimization with the potential of advancing the field of aligning LLMs with human preferences."
                },
                {
                    "title": "Offline Data Enhanced On-Policy Policy Gradient with Provable Guarantees",
                    "abstract": "Hybrid RL is the setting where an RL agent has access to both offline data and online data by interacting with the real-world environment. In this work, we propose a new hybrid RL algorithm that combines an on-policy actor-critic method with offline data. On-policy methods such as policy gradient and natural policy gradient (NPG) have shown to be more robust to model misspecification, though sometimes it may not be as sample efficient as methods that rely on off-policy learning. On the other hand, offline methods that depend on off-policy training often require strong assumptions in theory and are less stable to train in practice. Our new approach integrates a procedure of off-policy training on the offline data into an on-policy NPG framework. We show that our approach, in theory, can obtain a best-of-both-worlds type of result -- it achieves the state-of-art theoretical guarantees of offline RL when offline RL-specific assumptions hold, while at the same time maintaining the theoretical guarantees of on-policy NPG regardless of the offline RL assumptions' validity. Experimentally, in challenging rich-observation environments, we show that our approach outperforms a state-of-the-art hybrid RL baseline which only relies on off-policy policy optimization, demonstrating the empirical benefit of combining on-policy and off-policy learning. Our code is publicly available at https://github.com/YifeiZhou02/HNPG."
                },
                {
                    "title": "A General Theoretical Paradigm to Understand Learning from Human Preferences",
                    "abstract": "The prevalent deployment of learning from human preferences through reinforcement learning (RLHF) relies on two important approximations: the first assumes that pairwise preferences can be substituted with pointwise rewards. The second assumes that a reward model trained on these pointwise rewards can generalize from collected data to out-of-distribution data sampled by the policy. Recently, Direct Preference Optimisation (DPO) has been proposed as an approach that bypasses the second approximation and learn directly a policy from collected data without the reward modelling stage. However, this method still heavily relies on the first approximation. In this paper we try to gain a deeper theoretical understanding of these practical algorithms. In particular we derive a new general objective called $\\Psi$PO for learning from human preferences that is expressed in terms of pairwise preferences and therefore bypasses both approximations. This new general objective allows us to perform an in-depth analysis of the behavior of RLHF and DPO (as special cases of $\\Psi$PO) and to identify their potential pitfalls. We then consider another special case for $\\Psi$PO by setting $\\Psi$ simply to Identity, for which we can derive an efficient optimisation procedure, prove performance guarantees and demonstrate its empirical superiority to DPO on some illustrative examples."
                },
                {
                    "title": "Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inference",
                    "abstract": "Latent Diffusion models (LDMs) have achieved remarkable results in synthesizing high-resolution images. However, the iterative sampling process is computationally intensive and leads to slow generation. Inspired by Consistency Models (song et al.), we propose Latent Consistency Models (LCMs), enabling swift inference with minimal steps on any pre-trained LDMs, including Stable Diffusion (rombach et al). Viewing the guided reverse diffusion process as solving an augmented probability flow ODE (PF-ODE), LCMs are designed to directly predict the solution of such ODE in latent space, mitigating the need for numerous iterations and allowing rapid, high-fidelity sampling. Efficiently distilled from pre-trained classifier-free guided diffusion models, a high-quality 768 x 768 2~4-step LCM takes only 32 A100 GPU hours for training. Furthermore, we introduce Latent Consistency Fine-tuning (LCF), a novel method that is tailored for fine-tuning LCMs on customized image datasets. Evaluation on the LAION-5B-Aesthetics dataset demonstrates that LCMs achieve state-of-the-art text-to-image generation performance with few-step inference. Project Page: https://latent-consistency-models.github.io/"
                },
                {
                    "title": "Beyond One-Preference-Fits-All Alignment: Multi-Objective Direct Preference Optimization",
                    "abstract": "A single language model, even when aligned with labelers through reinforcement learning from human feedback (RLHF), may not suit all human preferences. Recent approaches therefore prefer customization, gathering multi-dimensional feedback, and creating distinct reward models for each dimension. Different language models are then optimized for various preferences using multi-objective RLHF (MORLHF) with varying reward weights. However, RL fine-tuning is unstable and resource-heavy, especially with diverse and usually conflicting objectives. In this paper, we present Multi-Objective Direct Preference Optimization (MODPO), an RL-free extension of Direct Preference Optimization (DPO) for multiple alignment objectives. Essentially, MODPO folds language modeling directly into reward modeling, training language models as implicit collective reward models that combine all objectives with specific weights. MODPO theoretically yields the same optimal solutions as MORLHF but is practically more stable and efficient. Empirical results in safety alignment and long-form question answering show that MODPO matches or outperforms existing methods, producing a Pareto front of language models catering to diverse preferences with three times less computational resources compared to MORLHF. Code is available at https://github.com/ZHZisZZ/modpo."
                },
                {
                    "title": "OpenChat: Advancing Open-source Language Models with Mixed-Quality Data",
                    "abstract": "Nowadays, open-source large language models like LLaMA have emerged. Recent developments have incorporated supervised fine-tuning (SFT) and reinforcement learning fine-tuning (RLFT) to align these models with human goals. However, SFT methods treat all training data with mixed quality equally, while RLFT methods require high-quality pairwise or ranking-based preference data. In this study, we present a novel framework, named OpenChat, to advance open-source language models with mixed-quality data. Specifically, we consider the general SFT training data, consisting of a small amount of expert data mixed with a large proportion of sub-optimal data, without any preference labels. We propose the C(onditioned)-RLFT, which regards different data sources as coarse-grained reward labels and learns a class-conditioned policy to leverage complementary data quality information. Interestingly, the optimal policy in C-RLFT can be easily solved through single-stage, RL-free supervised learning, which is lightweight and avoids costly human preference labeling. Through extensive experiments on three standard benchmarks, our openchat-13b fine-tuned with C-RLFT achieves the highest average performance among all 13b open-source language models. Moreover, we use AGIEval to validate the model generalization performance, in which only openchat-13b surpasses the base model. Finally, we conduct a series of analyses to shed light on the effectiveness and robustness of OpenChat. Our code, data, and models are publicly available at https://github.com/imoneoi/openchat and https://huggingface.co/openchat."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Deep Reinforcement Learning",
                    "abstract": "Deep Reinforcement Learning (DRL) is a powerful technique for learning policies for complex decision-making tasks. In this paper, we provide an overview of DRL, including its basic components, key algorithms and techniques, and applications in areas s.a. robotics, game playing, and autonomous driving. We also discuss some of the challenges and limitations of DRL, s.a. sample inefficiency and safety concerns, and we identify some of the promising directions for future research in DRL, s.a. meta-learning, hierarchical reinforcement learning, and combining DRL with formal techniques. In the second part of the paper, we discuss several important applications of DRL, including transfer learning, multi-agent reinforcement learning, and explainable reinforcement learning. We also explore the combination of DRL with formal techniques, a promising area of research for ensuring the safety and reliability of DRL applications. Finally, we identify some of the limitations and open issues in DRL, including sample efficiency, safety, and scalability concerns. To help practitioners effectively apply DRL in their work, we provide recommendations for starting with simple problems, choosing appropriate algorithms and architectures, paying attention to safety and ethics, collaborating with experts, and staying up to date with the latest research in the field. We conclude by highlighting the potential impact of DRL in a wide range of applications and emphasizing the need for careful consideration of the ethical and societal implications of DRL."
                },
                {
                    "title": "Is RLHF More Difficult than Standard RL?",
                    "abstract": "Reinforcement learning from Human Feedback (RLHF) learns from preference signals, while standard Reinforcement Learning (RL) directly learns from reward signals. Preferences arguably contain less information than rewards, which makes preference-based RL seemingly more difficult. This paper theoretically proves that, for a wide range of preference models, we can solve preference-based RL directly using existing algorithms and techniques for reward-based RL, with small or no extra costs. Specifically, (1) for preferences that are drawn from reward-based probabilistic models, we reduce the problem to robust reward-based RL that can tolerate small errors in rewards; (2) for general arbitrary preferences where the objective is to find the von Neumann winner, we reduce the problem to multiagent reward-based RL which finds Nash equilibria for factored Markov games under a restricted set of policies. The latter case can be further reduce to adversarial MDP when preferences only depend on the final state. We instantiate all reward-based RL subroutines by concrete provable algorithms, and apply our theory to a large class of models including tabular MDPs and MDPs with generic function approximation. We further provide guarantees when K-wise comparisons are available."
                },
                {
                    "title": "Learning to Generate Better Than Your LLM",
                    "abstract": "Reinforcement learning (RL) has emerged as a powerful paradigm for fine-tuning Large Language Models (LLMs) for text generation. In particular, recent LLMs such as ChatGPT and GPT-4 can engage in fluent conversations with users after finetuning with RL. Capitalizing on key properties of text generation, we seek to investigate RL algorithms beyond general purpose algorithms like Proximal Policy Optimization (PPO). In particular, we extend RL algorithms to allow them to interact with a dynamic black-box guide LLM and propose RL with guided feedback (RLGF), a suite of RL algorithms for LLM fine-tuning. We provide two ways for the guide LLM to interact with the LLM to be optimized for maximizing rewards. The guide LLM can generate text which serves as additional starting states for the RL optimization procedure. The guide LLM can also be used to complete the partial sentences generated by the LLM that is being optimized, treating the guide LLM as an expert to imitate and surpass eventually. We experiment on the IMDB positive sentiment, CommonGen, and TL;DR summarization tasks. We show that our RL algorithms achieve higher performance than supervised learning (SL) and the RL baseline PPO, demonstrating the benefit of interaction with the guide LLM. On both CommonGen and TL;DR, we not only outperform our SL baselines but also improve upon PPO across a variety of metrics beyond the one we optimized for. Our code can be found at https://github.com/Cornell-RL/tril."
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "Fine-Tuning Language Models with Advantage-Induced Policy Alignment",
                    "abstract": "Reinforcement learning from human feedback (RLHF) has emerged as a reliable approach to aligning large language models (LLMs) to human preferences. Among the plethora of RLHF techniques, proximal policy optimization (PPO) is of the most widely used methods. Despite its popularity, however, PPO may suffer from mode collapse, instability, and poor sample efficiency. We show that these issues can be alleviated by a novel algorithm that we refer to as Advantage-Induced Policy Alignment (APA), which leverages a squared error loss function based on the estimated advantages. We demonstrate empirically that APA consistently outperforms PPO in language tasks by a large margin, when a separate reward model is employed as the evaluator. In addition, compared with PPO, APA offers a more stable form of control over the deviation from the model's initial policy, ensuring that the model improves its performance without collapsing to deterministic output. In addition to empirical results, we also provide a theoretical justification supporting the design of our loss function."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "The Benefits of Being Distributional: Small-Loss Bounds for Reinforcement Learning",
                    "abstract": "While distributional reinforcement learning (DistRL) has been empirically effective, the question of when and why it is better than vanilla, non-distributional RL has remained unanswered. This paper explains the benefits of DistRL through the lens of small-loss bounds, which are instance-dependent bounds that scale with optimal achievable cost. Particularly, our bounds converge much faster than those from non-distributional approaches if the optimal cost is small. As warmup, we propose a distributional contextual bandit (DistCB) algorithm, which we show enjoys small-loss regret bounds and empirically outperforms the state-of-the-art on three real-world tasks. In online RL, we propose a DistRL algorithm that constructs confidence sets using maximum likelihood estimation. We prove that our algorithm enjoys novel small-loss PAC bounds in low-rank MDPs. As part of our analysis, we introduce the $\\ell_1$ distributional eluder dimension which may be of independent interest. Then, in offline RL, we show that pessimistic DistRL enjoys small-loss PAC bounds that are novel to the offline setting and are more robust to bad single-policy coverage."
                },
                {
                    "title": "Coherent Soft Imitation Learning",
                    "abstract": "Imitation learning methods seek to learn from an expert either through behavioral cloning (BC) of the policy or inverse reinforcement learning (IRL) of the reward. Such methods enable agents to learn complex tasks from humans that are difficult to capture with hand-designed reward functions. Choosing BC or IRL for imitation depends on the quality and state-action coverage of the demonstrations, as well as additional access to the Markov decision process. Hybrid strategies that combine BC and IRL are not common, as initial policy optimization against inaccurate rewards diminishes the benefit of pretraining the policy with BC. This work derives an imitation method that captures the strengths of both BC and IRL. In the entropy-regularized ('soft') reinforcement learning setting, we show that the behaviour-cloned policy can be used as both a shaped reward and a critic hypothesis space by inverting the regularized policy update. This coherency facilitates fine-tuning cloned policies using the reward estimate and additional interactions with the environment. This approach conveniently achieves imitation learning through initial behaviour cloning, followed by refinement via RL with online or offline data sources. The simplicity of the approach enables graceful scaling to high-dimensional and vision-based tasks, with stable learning and minimal hyperparameter tuning, in contrast to adversarial approaches. For the open-source implementation and simulation results, see https://joemwatson.github.io/csil/."
                },
                {
                    "title": "Training Diffusion Models with Reinforcement Learning",
                    "abstract": "Diffusion models are a class of flexible generative models trained with an approximation to the log-likelihood objective. However, most use cases of diffusion models are not concerned with likelihoods, but instead with downstream objectives such as human-perceived image quality or drug effectiveness. In this paper, we investigate reinforcement learning methods for directly optimizing diffusion models for such objectives. We describe how posing denoising as a multi-step decision-making problem enables a class of policy gradient algorithms, which we refer to as denoising diffusion policy optimization (DDPO), that are more effective than alternative reward-weighted likelihood approaches. Empirically, DDPO is able to adapt text-to-image diffusion models to objectives that are difficult to express via prompting, such as image compressibility, and those derived from human feedback, such as aesthetic quality. Finally, we show that DDPO can improve prompt-image alignment using feedback from a vision-language model without the need for additional data collection or human annotation. The project's website can be found at http://rl-diffusion.github.io ."
                },
                {
                    "title": "AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedback",
                    "abstract": "Large language models (LLMs) such as ChatGPT have seen widespread adoption due to their strong instruction-following abilities. Developing these LLMs involves a complex yet poorly understood workflow requiring training with human feedback. Replicating and understanding this instruction-following requires tackling three major challenges: the high cost of data collection, the lack of trustworthy evaluation, and the absence of reference method implementations. We address these challenges with AlpacaFarm, a simulator that enables research and development for learning from feedback at a low cost. First, we design LLM prompts to simulate human feedback that are 50x cheaper than crowdworkers and display high agreement with humans. Second, we propose an automatic evaluation and validate it against human instructions obtained on real-world interactions. Third, we contribute reference implementations for several methods (PPO, DPO, best-of-n, expert iteration, and more) that learn from pairwise feedback. Finally, as an end-to-end validation of AlpacaFarm, we train and evaluate eleven models on 10k pairs of real human feedback and show that rankings of models trained in AlpacaFarm match rankings of models trained on human data. As a demonstration of the research possible in AlpacaFarm, we find that methods that use a reward model can substantially improve over supervised fine-tuning and that our reference PPO implementation leads to a +10% improvement in win-rate against Davinci003. We release all components of AlpacaFarm at https://github.com/tatsu-lab/alpaca_farm."
                },
                {
                    "title": "Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling",
                    "abstract": "How do large language models (LLMs) develop and evolve over the course of training? How do these patterns change as models scale? To answer these questions, we introduce \\textit{Pythia}, a suite of 16 LLMs all trained on public data seen in the exact same order and ranging in size from 70M to 12B parameters. We provide public access to 154 checkpoints for each one of the 16 models, alongside tools to download and reconstruct their exact training dataloaders for further study. We intend \\textit{Pythia} to facilitate research in many areas, and we present several case studies including novel results in memorization, term frequency effects on few-shot performance, and reducing gender bias. We demonstrate that this highly controlled setup can be used to yield novel insights toward LLMs and their training dynamics. Trained models, analysis code, training code, and training data can be found at \\url{https://github.com/EleutherAI/pythia}."
                },
                {
                    "title": "Inverse Reinforcement Learning without Reinforcement Learning",
                    "abstract": "Inverse Reinforcement Learning (IRL) is a powerful set of techniques for imitation learning that aims to learn a reward function that rationalizes expert demonstrations. Unfortunately, traditional IRL methods suffer from a computational weakness: they require repeatedly solving a hard reinforcement learning (RL) problem as a subroutine. This is counter-intuitive from the viewpoint of reductions: we have reduced the easier problem of imitation learning to repeatedly solving the harder problem of RL. Another thread of work has proved that access to the side-information of the distribution of states where a strong policy spends time can dramatically reduce the sample and computational complexities of solving an RL problem. In this work, we demonstrate for the first time a more informed imitation learning reduction where we utilize the state distribution of the expert to alleviate the global exploration component of the RL subroutine, providing an exponential speedup in theory. In practice, we find that we are able to significantly speed up the prior art on continuous control tasks."
                },
                {
                    "title": "Consistency Models",
                    "abstract": "Diffusion models have significantly advanced the fields of image, audio, and video generation, but they depend on an iterative sampling process that causes slow generation. To overcome this limitation, we propose consistency models, a new family of models that generate high quality samples by directly mapping noise to data. They support fast one-step generation by design, while still allowing multistep sampling to trade compute for sample quality. They also support zero-shot data editing, such as image inpainting, colorization, and super-resolution, without requiring explicit training on these tasks. Consistency models can be trained either by distilling pre-trained diffusion models, or as standalone generative models altogether. Through extensive experiments, we demonstrate that they outperform existing distillation techniques for diffusion models in one- and few-step sampling, achieving the new state-of-the-art FID of 3.55 on CIFAR-10 and 6.20 on ImageNet 64x64 for one-step generation. When trained in isolation, consistency models become a new family of generative models that can outperform existing one-step, non-adversarial generative models on standard benchmarks such as CIFAR-10, ImageNet 64x64 and LSUN 256x256."
                },
                {
                    "title": "Aligning Text-to-Image Models using Human Feedback",
                    "abstract": "Deep generative models have shown impressive results in text-to-image synthesis. However, current text-to-image models often generate images that are inadequately aligned with text prompts. We propose a fine-tuning method for aligning such models using human feedback, comprising three stages. First, we collect human feedback assessing model output alignment from a set of diverse text prompts. We then use the human-labeled image-text dataset to train a reward function that predicts human feedback. Lastly, the text-to-image model is fine-tuned by maximizing reward-weighted likelihood to improve image-text alignment. Our method generates objects with specified colors, counts and backgrounds more accurately than the pre-trained model. We also analyze several design choices and find that careful investigations on such design choices are important in balancing the alignment-fidelity tradeoffs. Our results demonstrate the potential for learning from human feedback to significantly improve text-to-image models."
                },
                {
                    "title": "Efficient Online Reinforcement Learning with Offline Data",
                    "abstract": "Sample efficiency and exploration remain major challenges in online reinforcement learning (RL). A powerful approach that can be applied to address these issues is the inclusion of offline data, such as prior trajectories from a human expert or a sub-optimal exploration policy. Previous methods have relied on extensive modifications and additional complexity to ensure the effective use of this data. Instead, we ask: can we simply apply existing off-policy methods to leverage offline data when learning online? In this work, we demonstrate that the answer is yes; however, a set of minimal but important changes to existing off-policy RL algorithms are required to achieve reliable performance. We extensively ablate these design choices, demonstrating the key factors that most affect performance, and arrive at a set of recommendations that practitioners can readily apply, whether their data comprise a small number of expert demonstrations or large volumes of sub-optimal trajectories. We see that correct application of these simple recommendations can provide a $\\mathbf{2.5\\times}$ improvement over existing approaches across a diverse set of competitive benchmarks, with no additional computational overhead. We have released our code at https://github.com/ikostrikov/rlpd."
                },
                {
                    "title": "Principled Reinforcement Learning with Human Feedback from Pairwise or K-wise Comparisons",
                    "abstract": "We provide a theoretical framework for Reinforcement Learning with Human Feedback (RLHF). Our analysis shows that when the true reward function is linear, the widely used maximum likelihood estimator (MLE) converges under both the Bradley-Terry-Luce (BTL) model and the Plackett-Luce (PL) model. However, we show that when training a policy based on the learned reward model, MLE fails while a pessimistic MLE provides policies with improved performance under certain coverage assumptions. Additionally, we demonstrate that under the PL model, the true MLE and an alternative MLE that splits the $K$-wise comparison into pairwise comparisons both converge. Moreover, the true MLE is asymptotically more efficient. Our results validate the empirical success of existing RLHF algorithms in InstructGPT and provide new insights for algorithm design. Furthermore, our results unify the problem of RLHF and max-entropy Inverse Reinforcement Learning (IRL), and provide the first sample complexity bound for max-entropy IRL."
                },
                {
                    "title": "Hybrid RL: Using Both Offline and Online Data Can Make RL Efficient",
                    "abstract": "We consider a hybrid reinforcement learning setting (Hybrid RL), in which an agent has access to an offline dataset and the ability to collect experience via real-world online interaction. The framework mitigates the challenges that arise in both pure offline and online RL settings, allowing for the design of simple and highly effective algorithms, in both theory and practice. We demonstrate these advantages by adapting the classical Q learning/iteration algorithm to the hybrid setting, which we call Hybrid Q-Learning or Hy-Q. In our theoretical results, we prove that the algorithm is both computationally and statistically efficient whenever the offline dataset supports a high-quality policy and the environment has bounded bilinear rank. Notably, we require no assumptions on the coverage provided by the initial distribution, in contrast with guarantees for policy gradient/iteration methods. In our experimental results, we show that Hy-Q with neural network function approximation outperforms state-of-the-art online, offline, and hybrid RL baselines on challenging benchmarks, including Montezuma's Revenge."
                },
                {
                    "title": "Is Reinforcement Learning (Not) for Natural Language Processing?: Benchmarks, Baselines, and Building Blocks for Natural Language Policy Optimization",
                    "abstract": "We tackle the problem of aligning pre-trained large language models (LMs) with human preferences. If we view text generation as a sequential decision-making problem, reinforcement learning (RL) appears to be a natural conceptual framework. However, using RL for LM-based generation faces empirical challenges, including training instability due to the combinatorial action space, as well as a lack of open-source libraries and benchmarks customized for LM alignment. Thus, a question rises in the research community: is RL a practical paradigm for NLP? To help answer this, we first introduce an open-source modular library, RL4LMs (Reinforcement Learning for Language Models), for optimizing language generators with RL. The library consists of on-policy RL algorithms that can be used to train any encoder or encoder-decoder LM in the HuggingFace library (Wolf et al. 2020) with an arbitrary reward function. Next, we present the GRUE (General Reinforced-language Understanding Evaluation) benchmark, a set of 6 language generation tasks which are supervised not by target strings, but by reward functions which capture automated measures of human preference. GRUE is the first leaderboard-style evaluation of RL algorithms for NLP tasks. Finally, we introduce an easy-to-use, performant RL algorithm, NLPO (Natural Language Policy Optimization) that learns to effectively reduce the combinatorial action space in language generation. We show 1) that RL techniques are generally better than supervised methods at aligning LMs to human preferences; and 2) that NLPO exhibits greater stability and performance than previous policy gradient methods (e.g., PPO (Schulman et al. 2017)), based on both automatic and human evaluations."
                },
                {
                    "title": "Provably Efficient Offline Multi-agent Reinforcement Learning via Strategy-wise Bonus",
                    "abstract": "This paper considers offline multi-agent reinforcement learning. We propose the strategy-wise concentration principle which directly builds a confidence interval for the joint strategy, in contrast to the point-wise concentration principle that builds a confidence interval for each point in the joint action space. For two-player zero-sum Markov games, by exploiting the convexity of the strategy-wise bonus, we propose a computationally efficient algorithm whose sample complexity enjoys a better dependency on the number of actions than the prior methods based on the point-wise bonus. Furthermore, for offline multi-agent general-sum Markov games, based on the strategy-wise bonus and a novel surrogate function, we give the first algorithm whose sample complexity only scales $\\sum_{i=1}^mA_i$ where $A_i$ is the action size of the $i$-th player and $m$ is the number of players. In sharp contrast, the sample complexity of methods based on the point-wise bonus would scale with the size of the joint action space $\\Pi_{i=1}^m A_i$ due to the curse of multiagents. Lastly, all of our algorithms can naturally take a pre-specified strategy class $\\Pi$ as input and output a strategy that is close to the best strategy in $\\Pi$. In this setting, the sample complexity only scales with $\\log |\\Pi|$ instead of $\\sum_{i=1}^mA_i$."
                },
                {
                    "title": "Nearly Minimax Optimal Offline Reinforcement Learning with Linear Function Approximation: Single-Agent MDP and Markov Game",
                    "abstract": "Offline reinforcement learning (RL) aims at learning an optimal strategy using a pre-collected dataset without further interactions with the environment. While various algorithms have been proposed for offline RL in the previous literature, the minimax optimality has only been (nearly) established for tabular Markov decision processes (MDPs). In this paper, we focus on offline RL with linear function approximation and propose a new pessimism-based algorithm for offline linear MDP. At the core of our algorithm is the uncertainty decomposition via a reference function, which is new in the literature of offline RL under linear function approximation. Theoretical analysis demonstrates that our algorithm can match the performance lower bound up to logarithmic factors. We also extend our techniques to the two-player zero-sum Markov games (MGs), and establish a new performance lower bound for MGs, which tightens the existing result, and verifies the nearly minimax optimality of the proposed algorithm. To the best of our knowledge, these are the first computationally efficient and nearly minimax optimal algorithms for offline single-agent MDPs and MGs with linear function approximation."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Pessimistic Minimax Value Iteration: Provably Efficient Equilibrium Learning from Offline Datasets",
                    "abstract": "We study episodic two-player zero-sum Markov games (MGs) in the offline setting, where the goal is to find an approximate Nash equilibrium (NE) policy pair based on a dataset collected a priori. When the dataset does not have uniform coverage over all policy pairs, finding an approximate NE involves challenges in three aspects: (i) distributional shift between the behavior policy and the optimal policy, (ii) function approximation to handle large state space, and (iii) minimax optimization for equilibrium solving. We propose a pessimism-based algorithm, dubbed as pessimistic minimax value iteration (PMVI), which overcomes the distributional shift by constructing pessimistic estimates of the value functions for both players and outputs a policy pair by solving NEs based on the two value functions. Furthermore, we establish a data-dependent upper bound on the suboptimality which recovers a sublinear rate without the assumption on uniform coverage of the dataset. We also prove an information-theoretical lower bound, which suggests that the data-dependent term in the upper bound is intrinsic. Our theoretical results also highlight a notion of\"relative uncertainty\", which characterizes the necessary and sufficient condition for achieving sample efficiency in offline MGs. To the best of our knowledge, we provide the first nearly minimax optimal result for offline MGs with function approximation."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "WebGPT: Browser-assisted question-answering with human feedback",
                    "abstract": "We fine-tune GPT-3 to answer long-form questions using a text-based web-browsing environment, which allows the model to search and navigate the web. By setting up the task so that it can be performed by humans, we are able to train models on the task using imitation learning, and then optimize answer quality with human feedback. To make human evaluation of factual accuracy easier, models must collect references while browsing in support of their answers. We train and evaluate our models on ELI5, a dataset of questions asked by Reddit users. Our best model is obtained by fine-tuning GPT-3 using behavior cloning, and then performing rejection sampling against a reward model trained to predict human preferences. This model's answers are preferred by humans 56% of the time to those of our human demonstrators, and 69% of the time to the highest-voted answer from Reddit."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "TruthfulQA: Measuring How Models Mimic Human Falsehoods",
                    "abstract": "We propose a benchmark to measure whether a language model is truthful in generating answers to questions. The benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics. We crafted questions that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. We tested GPT-3, GPT-Neo/J, GPT-2 and a T5-based model. The best model was truthful on 58% of questions, while human performance was 94%. Models generated many false answers that mimic popular misconceptions and have the potential to deceive humans. The largest models were generally the least truthful. This contrasts with other NLP tasks, where performance improves with model size. However, this result is expected if false answers are learned from the training distribution. We suggest that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web."
                },
                {
                    "title": "Deep Reinforcement Learning at the Edge of the Statistical Precipice",
                    "abstract": "Deep reinforcement learning (RL) algorithms are predominantly evaluated by comparing their relative performance on a large suite of tasks. Most published results on deep RL benchmarks compare point estimates of aggregate performance such as mean and median scores across tasks, ignoring the statistical uncertainty implied by the use of a finite number of training runs. Beginning with the Arcade Learning Environment (ALE), the shift towards computationally-demanding benchmarks has led to the practice of evaluating only a small number of runs per task, exacerbating the statistical uncertainty in point estimates. In this paper, we argue that reliable evaluation in the few run deep RL regime cannot ignore the uncertainty in results without running the risk of slowing down progress in the field. We illustrate this point using a case study on the Atari 100k benchmark, where we find substantial discrepancies between conclusions drawn from point estimates alone versus a more thorough statistical analysis. With the aim of increasing the field's confidence in reported results with a handful of runs, we advocate for reporting interval estimates of aggregate performance and propose performance profiles to account for the variability in results, as well as present more robust and efficient aggregate metrics, such as interquartile mean scores, to achieve small uncertainty in results. Using such statistical tools, we scrutinize performance evaluations of existing algorithms on other widely used RL benchmarks including the ALE, Procgen, and the DeepMind Control Suite, again revealing discrepancies in prior comparisons. Our findings call for a change in how we evaluate performance in deep RL, for which we present a more rigorous evaluation methodology, accompanied with an open-source library rliable, to prevent unreliable results from stagnating the field."
                },
                {
                    "title": "Efficient First-Order Contextual Bandits: Prediction, Allocation, and Triangular Discrimination",
                    "abstract": "A recurring theme in statistical learning, online learning, and beyond is that faster convergence rates are possible for problems with low noise, often quantified by the performance of the best hypothesis; such results are known as first-order or small-loss guarantees. While first-order guarantees are relatively well understood in statistical and online learning, adapting to low noise in contextual bandits (and more broadly, decision making) presents major algorithmic challenges. In a COLT 2017 open problem, Agarwal, Krishnamurthy, Langford, Luo, and Schapire asked whether first-order guarantees are even possible for contextual bandits and -- if so -- whether they can be attained by efficient algorithms. We give a resolution to this question by providing an optimal and efficient reduction from contextual bandits to online regression with the logarithmic (or, cross-entropy) loss. Our algorithm is simple and practical, readily accommodates rich function classes, and requires no distributional assumptions beyond realizability. In a large-scale empirical evaluation, we find that our approach typically outperforms comparable non-first-order methods. On the technical side, we show that the logarithmic loss and an information-theoretic quantity called the triangular discrimination play a fundamental role in obtaining first-order guarantees, and we combine this observation with new refinements to the regression oracle reduction framework of Foster and Rakhlin. The use of triangular discrimination yields novel results even for the classical statistical learning model, and we anticipate that it will find broader use."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Of Moments and Matching: A Game-Theoretic Framework for Closing the Imitation Gap",
                    "abstract": "We provide a unifying view of a large family of previous imitation learning algorithms through the lens of moment matching. At its core, our classification scheme is based on whether the learner attempts to match (1) reward or (2) action-value moments of the expert's behavior, with each option leading to differing algorithmic approaches. By considering adversarially chosen divergences between learner and expert behavior, we are able to derive bounds on policy performance that apply for all algorithms in each of these classes, the first to our knowledge. We also introduce the notion of moment recoverability, implicit in many previous analyses of imitation learning, which allows us to cleanly delineate how well each algorithmic family is able to mitigate compounding errors. We derive three novel algorithm templates (AdVIL, AdRIL, and DAeQuIL) with strong guarantees, simple implementation, and competitive empirical performance."
                },
                {
                    "title": "\"Learning to be a Learner\"",
                    "abstract": "Learning to be a learner\u201d is how Tessa, a grades 6-12 teacher librarian, said she would describe the idea of self-regulated learning to her students. Teacher librarians are in a unique position within schools to truly serve as lifelong learning coaches for students, focused on process and skills instead of content. Further, their reach extends across levels in the school, touching every single student through their teaching and the development and maintenance of the library collection. This paper presents findings from a study investigating how US teacher librarians apply metacognitive strategies in their teaching. An unexpected theme emerged from the interviews as participants described lifelong learning skills they strive to impart to students. These skills included cognitive, personal, and interpersonal skills as defined by UNICEF (2003) and exemplified by the theme for the 2013 International Association of School Librarians\u2019 Conference."
                },
                {
                    "title": "VarGrad: A Low-Variance Gradient Estimator for Variational Inference",
                    "abstract": "We analyse the properties of an unbiased gradient estimator of the ELBO for variational inference, based on the score function method with leave-one-out control variates. We show that this gradient estimator can be obtained using a new loss, defined as the variance of the log-ratio between the exact posterior and the variational approximation, which we call the $\\textit{log-variance loss}$. Under certain conditions, the gradient of the log-variance loss equals the gradient of the (negative) ELBO. We show theoretically that this gradient estimator, which we call $\\textit{VarGrad}$ due to its connection to the log-variance loss, exhibits lower variance than the score function method in certain settings, and that the leave-one-out control variate coefficients are close to the optimal ones. We empirically demonstrate that VarGrad offers a favourable variance versus computation trade-off compared to other state-of-the-art estimators on a discrete VAE."
                },
                {
                    "title": "Revisiting Design Choices in Proximal Policy Optimization",
                    "abstract": "Proximal Policy Optimization (PPO) is a popular deep policy gradient algorithm. In standard implementations, PPO regularizes policy updates with clipped probability ratios, and parameterizes policies with either continuous Gaussian distributions or discrete Softmax distributions. These design choices are widely accepted, and motivated by empirical performance comparisons on MuJoCo and Atari benchmarks. \nWe revisit these practices outside the regime of current benchmarks, and expose three failure modes of standard PPO. We explain why standard design choices are problematic in these cases, and show that alternative choices of surrogate objectives and policy parameterizations can prevent the failure modes. We hope that our work serves as a reminder that many algorithmic design choices in reinforcement learning are tied to specific simulation environments. We should not implicitly accept these choices as a standard part of a more general algorithm."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "Learning to summarize from human feedback",
                    "abstract": "As language models become more powerful, training and evaluation are increasingly bottlenecked by the data and metrics used for a particular task. For example, summarization models are often trained to predict human reference summaries and evaluated using ROUGE, but both of these metrics are rough proxies for what we really care about---summary quality. In this work, we show that it is possible to significantly improve summary quality by training a model to optimize for human preferences. We collect a large, high-quality dataset of human comparisons between summaries, train a model to predict the human-preferred summary, and use that model as a reward function to fine-tune a summarization policy using reinforcement learning. We apply our method to a version of the TL;DR dataset of Reddit posts and find that our models significantly outperform both human reference summaries and much larger models fine-tuned with supervised learning alone. Our models also transfer to CNN/DM news articles, producing summaries nearly as good as the human reference without any news-specific fine-tuning. We conduct extensive analyses to understand our human feedback dataset and fine-tuned models We establish that our reward model generalizes to new datasets, and that optimizing our reward model results in better summaries than optimizing ROUGE according to humans. We hope the evidence from our paper motivates machine learning researchers to pay closer attention to how their training loss affects the model behavior they actually want."
                },
                {
                    "title": "PC-PG: Policy Cover Directed Exploration for Provable Policy Gradient Learning",
                    "abstract": "Direct policy gradient methods for reinforcement learning are a successful approach for a variety of reasons: they are model free, they directly optimize the performance metric of interest, and they allow for richly parameterized policies. Their primary drawback is that, by being local in nature, they fail to adequately explore the environment. In contrast, while model-based approaches and Q-learning directly handle exploration through the use of optimism, their ability to handle model misspecification and function approximation is far less evident. This work introduces the the Policy Cover-Policy Gradient (PC-PG) algorithm, which provably balances the exploration vs. exploitation tradeoff using an ensemble of learned policies (the policy cover). PC-PG enjoys polynomial sample complexity and run time for both tabular MDPs and, more generally, linear MDPs in an infinite dimensional RKHS. Furthermore, PC-PG also has strong guarantees under model misspecification that go beyond the standard worst case $\\ell_{\\infty}$ assumptions; this includes approximation guarantees for state aggregation under an average case error assumption, along with guarantees under a more general assumption where the approximation error under distribution shift is controlled. We complement the theory with empirical evaluation across a variety of domains in both reward-free and reward-driven settings."
                },
                {
                    "title": "Implementation Matters in Deep Policy Gradients: A Case Study on PPO and TRPO",
                    "abstract": "We study the roots of algorithmic progress in deep policy gradient algorithms through a case study on two popular algorithms: Proximal Policy Optimization (PPO) and Trust Region Policy Optimization (TRPO). Specifically, we investigate the consequences of \"code-level optimizations:\" algorithm augmentations found only in implementations or described as auxiliary details to the core algorithm. Seemingly of secondary importance, such optimizations turn out to have a major impact on agent behavior. Our results show that they (a) are responsible for most of PPO's gain in cumulative reward over TRPO, and (b) fundamentally change how RL methods function. These insights show the difficulty and importance of attributing performance gains in deep reinforcement learning. Code for reproducing our results is available at this https URL ."
                },
                {
                    "title": "Advantage-Weighted Regression: Simple and Scalable Off-Policy Reinforcement Learning",
                    "abstract": "In this paper, we aim to develop a simple and scalable reinforcement learning algorithm that uses standard supervised learning methods as subroutines. Our goal is an algorithm that utilizes only simple and convergent maximum likelihood loss functions, while also being able to leverage off-policy data. Our proposed approach, which we refer to as advantage-weighted regression (AWR), consists of two standard supervised learning steps: one to regress onto target values for a value function, and another to regress onto weighted target actions for the policy. The method is simple and general, can accommodate continuous and discrete actions, and can be implemented in just a few lines of code on top of standard supervised learning methods. We provide a theoretical motivation for AWR and analyze its properties when incorporating off-policy data from experience replay. We evaluate AWR on a suite of standard OpenAI Gym benchmark tasks, and show that it achieves competitive performance compared to a number of well-established state-of-the-art RL algorithms. AWR is also able to acquire more effective policies than most off-policy algorithms when learning from purely static datasets with no additional environmental interactions. Furthermore, we demonstrate our algorithm on challenging continuous control tasks with highly complex simulated characters."
                },
                {
                    "title": "Fine-Tuning Language Models from Human Preferences",
                    "abstract": "Reward learning enables the application of reinforcement learning (RL) to tasks where reward is defined by human judgment, building a model of reward by asking humans questions. Most work on reward learning has used simulated environments, but complex information about values is often expressed in natural language, and we believe reward learning for language is a key to making RL practical and safe for real-world tasks. In this paper, we build on advances in generative pretraining of language models to apply reward learning to four natural language tasks: continuing text with positive sentiment or physically descriptive language, and summarization tasks on the TL;DR and CNN/Daily Mail datasets. For stylistic continuation we achieve good results with only 5,000 comparisons evaluated by humans. For summarization, models trained with 60,000 comparisons copy whole sentences from the input but skip irrelevant preamble; this leads to reasonable ROUGE scores and very good performance according to our human labelers, but may be exploiting the fact that labelers rely on simple heuristics."
                },
                {
                    "title": "Adaptive Trust Region Policy Optimization: Global Convergence and Faster Rates for Regularized MDPs",
                    "abstract": "Trust region policy optimization (TRPO) is a popular and empirically successful policy search algorithm in Reinforcement Learning (RL) in which a surrogate problem, that restricts consecutive policies to be \u2018close\u2019 to one another, is iteratively solved. Nevertheless, TRPO has been considered a heuristic algorithm inspired by Conservative Policy Iteration (CPI). We show that the adaptive scaling mechanism used in TRPO is in fact the natural \u201cRL version\u201d of traditional trust-region methods from convex analysis. We first analyze TRPO in the planning setting, in which we have access to the model and the entire state space. Then, we consider sample-based TRPO and establish \u00d5(1/\u221aN) convergence rate to the global optimum. Importantly, the adaptive scaling mechanism allows us to analyze TRPO in regularized MDPs for which we prove fast rates of \u00d5(1/N), much like results in convex optimization. This is the first result in RL of better rates when regularizing the instantaneous cost or reward."
                },
                {
                    "title": "On the Theory of Policy Gradient Methods: Optimality, Approximation, and Distribution Shift",
                    "abstract": "Policy gradient methods are among the most effective methods in challenging reinforcement learning problems with large state and/or action spaces. However, little is known about even their most basic theoretical convergence properties, including: if and how fast they converge to a globally optimal solution or how they cope with approximation error due to using a restricted class of parametric policies. This work provides provable characterizations of the computational, approximation, and sample size properties of policy gradient methods in the context of discounted Markov Decision Processes (MDPs). We focus on both: \"tabular\" policy parameterizations, where the optimal policy is contained in the class and where we show global convergence to the optimal policy; and parametric policy classes (considering both log-linear and neural policy classes), which may not contain the optimal policy and where we provide agnostic learning results. One central contribution of this work is in providing approximation guarantees that are average case -- which avoid explicit worst-case dependencies on the size of state space -- by making a formal connection to supervised learning under distribution shift. This characterization shows an important interplay between estimation error, approximation error, and exploration (as characterized through a precisely defined condition number)."
                },
                {
                    "title": "HellaSwag: Can a Machine Really Finish Your Sentence?",
                    "abstract": "Recent work by Zellers et al. (2018) introduced a new task of commonsense natural language inference: given an event description such as \u201cA woman sits at a piano,\u201d a machine must select the most likely followup: \u201cShe sets her fingers on the keys.\u201d With the introduction of BERT, near human-level performance was reached. Does this mean that machines can perform human level commonsense inference? In this paper, we show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans (>95% accuracy), state-of-the-art models struggle (<48%). We achieve this via Adversarial Filtering (AF), a data collection paradigm wherein a series of discriminators iteratively select an adversarial set of machine-generated wrong answers. AF proves to be surprisingly robust. The key insight is to scale up the length and complexity of the dataset examples towards a critical \u2018Goldilocks\u2019 zone wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models. Our construction of HellaSwag, and its resulting difficulty, sheds light on the inner workings of deep pretrained models. More broadly, it suggests a new path forward for NLP research, in which benchmarks co-evolve with the evolving state-of-the-art in an adversarial way, so as to present ever-harder challenges."
                },
                {
                    "title": "QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation",
                    "abstract": "In this paper, we study the problem of learning vision-based dynamic manipulation skills using a scalable reinforcement learning approach. We study this problem in the context of grasping, a longstanding challenge in robotic manipulation. In contrast to static learning behaviors that choose a grasp point and then execute the desired grasp, our method enables closed-loop vision-based control, whereby the robot continuously updates its grasp strategy based on the most recent observations to optimize long-horizon grasp success. To that end, we introduce QT-Opt, a scalable self-supervised vision-based reinforcement learning framework that can leverage over 580k real-world grasp attempts to train a deep neural network Q-function with over 1.2M parameters to perform closed-loop, real-world grasping that generalizes to 96% grasp success on unseen objects. Aside from attaining a very high success rate, our method exhibits behaviors that are quite distinct from more standard grasping systems: using only RGB vision-based perception from an over-the-shoulder camera, our method automatically learns regrasping strategies, probes objects to find the most effective grasps, learns to reposition objects and perform other non-prehensile pre-grasp manipulations, and responds dynamically to disturbances and perturbations."
                },
                {
                    "title": "Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge",
                    "abstract": "We present a new question set, text corpus, and baselines assembled to encourage AI research in advanced question answering. Together, these constitute the AI2 Reasoning Challenge (ARC), which requires far more powerful knowledge and reasoning than previous challenges such as SQuAD or SNLI. The ARC question set is partitioned into a Challenge Set and an Easy Set, where the Challenge Set contains only questions answered incorrectly by both a retrieval-based algorithm and a word co-occurence algorithm. The dataset contains only natural, grade-school science questions (authored for human tests), and is the largest public-domain set of this kind (7,787 questions). We test several baselines on the Challenge Set, including leading neural models from the SQuAD and SNLI tasks, and find that none are able to significantly outperform a random baseline, reflecting the difficult nature of this task. We are also releasing the ARC Corpus, a corpus of 14M science sentences relevant to the task, and implementations of the three neural baseline models tested. Can your model perform better? We pose ARC as a challenge to the community."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Deep Reinforcement Learning that Matters",
                    "abstract": "\n \n In recent years, significant progress has been made in solving challenging problems across various domains using deep reinforcement learning (RL). Reproducing existing work and accurately judging the improvements offered by novel methods is vital to sustaining this progress. Unfortunately, reproducing results for state-of-the-art deep RL methods is seldom straightforward. In particular, non-determinism in standard benchmark environments, combined with variance intrinsic to the methods, can make reported results tough to interpret. Without significance metrics and tighter standardization of experimental reporting, it is difficult to determine whether improvements over the prior state-of-the-art are meaningful. In this paper, we investigate challenges posed by reproducibility, proper experimental techniques, and reporting procedures. We illustrate the variability in reported metrics and results when comparing against common baselines and suggest guidelines to make future results in deep RL more reproducible. We aim to spur discussion about how to ensure continued progress in the field by minimizing wasted effort stemming from results that are non-reproducible and easily misinterpreted.\n \n"
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "Deep Reinforcement Learning from Human Preferences",
                    "abstract": "For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems. In this work, we explore goals defined in terms of (non-expert) human preferences between pairs of trajectory segments. We show that this approach can effectively solve complex RL tasks without access to the reward function, including Atari games and simulated robot locomotion, while providing feedback on less than one percent of our agent's interactions with the environment. This reduces the cost of human oversight far enough that it can be practically applied to state-of-the-art RL systems. To demonstrate the flexibility of our approach, we show that we can successfully train complex novel behaviors with about an hour of human time. These behaviors and environments are considerably more complex than any that have been previously learned from human feedback."
                },
                {
                    "title": "Asynchronous Methods for Deep Reinforcement Learning",
                    "abstract": "We propose a conceptually simple and lightweight framework for deep reinforcement learning that uses asynchronous gradient descent for optimization of deep neural network controllers. We present asynchronous variants of four standard reinforcement learning algorithms and show that parallel actor-learners have a stabilizing effect on training allowing all four methods to successfully train neural network controllers. The best performing method, an asynchronous variant of actor-critic, surpasses the current state-of-the-art on the Atari domain while training for half the time on a single multi-core CPU instead of a GPU. Furthermore, we show that asynchronous actor-critic succeeds on a wide variety of continuous motor control problems as well as on a new task of navigating random 3D mazes using a visual input."
                },
                {
                    "title": "High-Dimensional Continuous Control Using Generalized Advantage Estimation",
                    "abstract": "Policy gradient methods are an appealing approach in reinforcement learning because they directly optimize the cumulative reward and can straightforwardly be used with nonlinear function approximators such as neural networks. The two main challenges are the large number of samples typically required, and the difficulty of obtaining stable and steady improvement despite the nonstationarity of the incoming data. We address the first challenge by using value functions to substantially reduce the variance of policy gradient estimates at the cost of some bias, with an exponentially-weighted estimator of the advantage function that is analogous to TD(lambda). We address the second challenge by using trust region optimization procedure for both the policy and the value function, which are represented by neural networks. \nOur approach yields strong empirical results on highly challenging 3D locomotion tasks, learning running gaits for bipedal and quadrupedal simulated robots, and learning a policy for getting the biped to stand up from starting out lying on the ground. In contrast to a body of prior work that uses hand-crafted policy representations, our neural network policies map directly from raw kinematics to joint torques. Our algorithm is fully model-free, and the amount of simulated experience required for the learning tasks on 3D bipeds corresponds to 1-2 weeks of real time."
                },
                {
                    "title": "Contextual Dueling Bandits",
                    "abstract": "We consider the problem of learning to choose actions using contextual information when provided with limited feedback in the form of relative pairwise comparisons. We study this problem in the dueling-bandits framework of Yue et al. (2009), which we extend to incorporate context. Roughly, the learner's goal is to find the best policy, or way of behaving, in some space of policies, although \"best\" is not always so clearly defined. Here, we propose a new and natural solution concept, rooted in game theory, called a von Neumann winner, a randomized policy that beats or ties every other policy. We show that this notion overcomes important limitations of existing solutions, particularly the Condorcet winner which has typically been used in the past, but which requires strong and often unrealistic assumptions. We then present three efficient algorithms for online learning in our setting, and for approximating a von Neumann winner from batch-like data. The first of these algorithms achieves particularly low regret, even when data is adversarial, although its time and space requirements are linear in the size of the policy space. The other two algorithms require time and space only logarithmic in the size of the policy space when provided access to an oracle for solving classification problems on the space."
                },
                {
                    "title": "Trust Region Policy Optimization",
                    "abstract": "In this article, we describe a method for optimizing control policies, with guaranteed monotonic improvement. By making several approximations to the theoretically-justified scheme, we develop a practical algorithm, called Trust Region Policy Optimization (TRPO). This algorithm is effective for optimizing large nonlinear policies such as neural networks. Our experiments demonstrate its robust performance on a wide variety of tasks: learning simulated robotic swimming, hopping, and walking gaits; and playing Atari games using images of the screen as input. Despite its approximations that deviate from the theory, TRPO tends to give monotonic improvement, with little tuning of hyperparameters."
                },
                {
                    "title": "Reinforcement and Imitation Learning via Interactive No-Regret Learning",
                    "abstract": "Recent work has demonstrated that problems-- particularly imitation learning and structured prediction-- where a learner's predictions influence the input-distribution it is tested on can be naturally addressed by an interactive approach and analyzed using no-regret online learning. These approaches to imitation learning, however, neither require nor benefit from information about the cost of actions. We extend existing results in two directions: first, we develop an interactive imitation learning approach that leverages cost information; second, we extend the technique to address reinforcement learning. The results provide theoretical support to the commonly observed successes of online approximate policy iteration. Our approach suggests a broad new family of algorithms and provides a unifying view of existing techniques for imitation and reinforcement learning."
                },
                {
                    "title": "A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning",
                    "abstract": "Sequential prediction problems such as imitation learning, where future observations depend on previous predictions (actions), violate the common i.i.d. assumptions made in statistical learning. This leads to poor performance in theory and often in practice. Some recent approaches provide stronger guarantees in this setting, but remain somewhat unsatisfactory as they train either non-stationary or stochastic policies and require a large number of iterations. In this paper, we propose a new iterative algorithm, which trains a stationary deterministic policy, that can be seen as a no regret algorithm in an online learning setting. We show that any such no regret algorithm, combined with additional reduction assumptions, must find a policy with good performance under the distribution of observations it induces in such sequential settings. We demonstrate that this new approach outperforms previous approaches on two challenging imitation learning problems and a benchmark sequence labeling problem."
                },
                {
                    "title": "Maximum Entropy Inverse Reinforcement Learning",
                    "abstract": "Recent research has shown the benefit of framing problems of imitation learning as solutions to Markov Decision Problems. This approach reduces learning to the problem of recovering a utility function that makes the behavior induced by a near-optimal policy closely mimic demonstrated behavior. In this work, we develop a probabilistic approach based on the principle of maximum entropy. Our approach provides a well-defined, globally normalized distribution over decision sequences, while providing the same performance guarantees as existing methods. \n \nWe develop our technique in the context of modeling real-world navigation and driving behaviors where collected data is inherently noisy and imperfect. Our probabilistic approach enables modeling of route preferences as well as a powerful new approach to inferring destinations and routes based on partial trajectories."
                },
                {
                    "title": "Reinforcement learning by reward-weighted regression for operational space control",
                    "abstract": "Many robot control problems of practical importance, including operational space control, can be reformulated as immediate reward reinforcement learning problems. However, few of the known optimization or reinforcement learning algorithms can be used in online learning control for robots, as they are either prohibitively slow, do not scale to interesting domains of complex robots, or require trying out policies generated by random search, which are infeasible for a physical system. Using a generalization of the EM-base reinforcement learning framework suggested by Dayan & Hinton, we reduce the problem of learning with immediate rewards to a reward-weighted regression problem with an adaptive, integrated reward transformation for faster convergence. The resulting algorithm is efficient, learns smoothly without dangerous jumps in solution space, and works well in applications of complex high degree-of-freedom robots."
                },
                {
                    "title": "Game theory, maximum entropy, minimum discrepancy and robust Bayesian decision theory",
                    "abstract": "We describe and develop a close relationship between two problems that have customarily been regarded as distinct: that of maximizing entropy, and that of minimizing worst-case expected loss. Using a formulation grounded in the equilibrium theory of zero-sum games between Decision Maker and Nature, these two problems are shown to be dual to each other, the solution to each providing that to the other. Although Topsoe described this connection for the Shannon entropy over 20 years ago, it does not appear to be widely known even in that important special case. We here generalize this theory to apply to arbitrary decision problems and loss functions. We indicate how an appropriate generalized definition of entropy can be associated with such a problem, and we show that, subject to certain regularity conditions, the above-mentioned duality continues to apply in this extended context. This simultaneously provides a possible rationale for maximizing entropy and a tool for finding robust Bayes acts. We also describe the essential identity between the problem of maximizing entropy and that of minimizing a related discrepancy or divergence between distributions. This leads to an extension, to arbitrary discrepancies, of a well-known minimax theorem for the case of Kullback-Leibler divergence (the redundancy-capacity theorem of information theory). For the important case of families of distributions having certain mean values specified, we develop simple sufficient conditions and methods for identifying the desired solutions. We use this theory to introduce a new concept of generalized exponential family linked to the specific decision problem under consideration, and we demonstrate that this shares many of the properties of standard exponential families. Finally, we show that the existence of an equilibrium in our game can be rephrased in terms of a Pythagorean property of the related divergence, thus generalizing previously announced results for Kullback-Leibler and Bregman divergences."
                },
                {
                    "title": "Policy Search by Dynamic Programming",
                    "abstract": "We consider the policy search approach to reinforcement learning. We show that if a \"baseline distribution\" is given (indicating roughly how often we expect a good policy to visit each state), then we can derive a policy search algorithm that terminates in a finite number of steps, and for which we can provide non-trivial performance guarantees. We also demonstrate this algorithm on several grid-world POMDPs, a planar biped walking robot, and a double-pole balancing problem."
                },
                {
                    "title": "Error Bounds for Approximate Policy Iteration",
                    "abstract": "In Dynamic Programming, convergence of algorithms such as Value Iteration or Policy Iteration results -in discounted problems- from a contraction property of the back-up operator, guaranteeing convergence to its fixed-point. When approximation is considered, known results in Approximate Policy Iteration provide bounds on the closeness to optimality of the approximate value function obtained by successive policy improvement steps as a function of the maximum norm of value determination errors during policy evaluation steps. Unfortunately, such results have limited practical range since most function approximators (such as linear regression) select the best fit in a given class of parameterized functions by minimizing some (weighted) quadratic norm. \n \nIn this paper, we provide error bounds for Approximate Policy Iteration using quadratic norms, and illustrate those results in the case of feature-based linear function approximation."
                },
                {
                    "title": "Covariant policy search",
                    "abstract": "We investigate the problem of non-covariant behavior of policy gradient reinforcement learning algorithms. The policy gradient approach is amenable to analysis by information geometric methods. This leads us to propose a natural metric on controller parameterization that results from considering the manifold of probability distributions over paths induced by a stochastic controller. Investigation of this approach leads to a covariant gradient ascent rule. Interesting properties of this rule are discussed, including its relation with actor-critic style reinforcement learning algorithms. The algorithms discussed here are computationally quite efficient and on some interesting problems lead to dramatic performance improvement over noncovariant rules."
                },
                {
                    "title": "A Natural Policy Gradient",
                    "abstract": "We provide a natural gradient method that represents the steepest descent direction based on the underlying structure of the parameter space. Although gradient methods cannot make large changes in the values of the parameters, we show that the natural gradient is moving toward choosing a greedy optimal action rather than just a better action. These greedy optimal actions are those that would be chosen under one improvement step of policy iteration with approximate, compatible value functions, as defined by Sutton et al. [9]. We then show drastic performance improvements in simple MDPs and in the more challenging MDP of Tetris."
                },
                {
                    "title": "Policy Gradient Methods for Reinforcement Learning with Function Approximation",
                    "abstract": "Function approximation is essential to reinforcement learning, but the standard approach of approximating a value function and determining a policy from it has so far proven theoretically intractable. In this paper we explore an alternative approach in which the policy is explicitly represented by its own function approximator, independent of the value function, and is updated according to the gradient of expected reward with respect to the policy parameters. Williams's REINFORCE method and actor-critic methods are examples of this approach. Our main new result is to show that the gradient can be written in a form suitable for estimation from experience aided by an approximate action-value or advantage function. Using this result, we prove for the first time that a version of policy iteration with arbitrary differentiable function approximation is convergent to a locally optimal policy."
                },
                {
                    "title": "Probabilistic Social Choice Based on Simple Voting Comparisons",
                    "abstract": "A social choice procedure is developed for selecting an alternative from a finite set on the basis of paired-comparison voting. Ballot data are used to construct a lottery on the alternatives that is socially as preferred as every other lottery. The constructed lottery is then used to select a winner. An axiomatization of social preferences among lotteries that justifies the procedure is included. The procedure will always select a consensus majority alternative when one exists, and it will never select an alternative that is Pareto dominated by another alternative."
                },
                {
                    "title": "On a Class of Equilibrium Conditions for Majority Rule",
                    "abstract": "The various conditions for non-intransitivity of majority rule formulated over the past decade have been concerned with choices over arbitrary, usually finite, sets of discrete alternatives. In many economic and other social choice problems, however, the possible choices constitute a point set in some appropriately defined multi-dimensional commodity or policy space. It is shown that in problems of this kind, when voter preferences can be represented by quasi-concave, differentiable utility functions, the various equilibrium conditions for majority rule are incompatible with even a very modest degree of heterogeneity of tastes, and for most purposes are probably not significantly less restrictive than the extreme condition of complete unanimity of individual preferences."
                },
                {
                    "title": "On Defining Areas of Voter Choice: Professor Tullock on Stable Voting",
                    "abstract": "I. Degree of voter agreement, 478. \u2014 II. The conditions of noncircular voting, 481. \u2014 III. Size of the decision set, 482. \u2014 IV. Democratic processes, 485. \u2014 V. Conclusions, 486. \u2014 Mathematical note, 487."
                },
                {
                    "title": "A Theoretical Analysis of Nash Learning from Human Feedback under General KL-Regularized Preference",
                    "abstract": "Reinforcement Learning from Human Feedback (RLHF) learns from the preference signal provided by a probabilistic preference model, which takes a prompt and two responses as input, and produces a score indicating the preference of one response against another. So far, the most popular RLHF paradigm is reward-based, which starts with an initial step of reward modeling, and the constructed reward is then used to provide a reward signal for the subsequent reward optimization stage. However, the existence of a reward function is a strong assumption and the reward-based RLHF is limited in expressivity and cannot capture the real-world complicated human preference. In this work, we provide theoretical insights for a recently proposed learning paradigm, Nash learning from human feedback (NLHF), which considered a general preference model and formulated the alignment process as a game between two competitive LLMs. The learning objective is to \ufb01nd a policy that consistently generates responses preferred over any competing policy while staying close to the initial model. The objective is de\ufb01ned as the Nash equilibrium (NE) of the KL-regularized preference model. We aim to make the \ufb01rst attempt to study the theoretical learnability of the KL-regularized NLHF by considering both of\ufb02ine and online settings. For the of\ufb02ine learning from a pre-collected dataset, we propose algorithms that are ef\ufb01cient under suitable coverage conditions of the dataset. For batch online learning from iterative interactions with a preference oracle, our proposed algorithm enjoys a \ufb01nite sample guarantee under the structural condition of the underlying preference model. Our results connect the new NLHF paradigm with traditional RL theory, and validate the potential of reward-model-free learning under general preference."
                },
                {
                    "title": "UltraFeedback: Boosting Language Models with High-quality Feedback",
                    "abstract": "Reinforcement learning from human feedback (RLHF) has become a pivot technique in aligning large language models (LLMs) with human preferences. In RLHF practice, preference data plays a crucial role in bridging human proclivity and LLMs. However, the scarcity of diverse, naturalistic datasets of human preferences on LLM outputs at scale poses a great challenge to RLHF as well as feedback learning research within the open-source community. Current preference datasets, either proprietary or limited in size and prompt variety, result in limited RLHF adoption in open-source models and hinder further exploration. In this study, we propose ULTRAFEEDBACK, a large-scale, high-quality, and diversified preference dataset designed to overcome these limitations and foster RLHF development. To create ULTRAFEEDBACK, we compile a diverse array of instructions and models from multiple sources to produce comparative data. We meticulously devise annotation instructions and employ GPT-4 to offer detailed feedback in both numerical and textual forms. ULTRAFEEDBACK establishes a reproducible and expandable preference data construction pipeline, serving as a solid foundation for future RLHF and feedback learning research. Utilizing ULTRAFEEDBACK, we train various models to demonstrate its effectiveness, including the reward model UltraRM, chat language model UltraLM-13B-PPO, and critique model UltraCM. Experimental results indicate that our models outperform existing open-source models, achieving top performance across multiple benchmarks. Our data and models are available at https://github.com/thunlp/UltraFeedback."
                },
                {
                    "title": "When are Offline Two-Player Zero-Sum Markov Games Solvable?",
                    "abstract": "We study what dataset assumption permits solving offline two-player zero-sum Markov games. In stark contrast to the offline single-agent Markov decision process, we show that the single strategy concentration assumption is insufficient for learning the Nash equilibrium (NE) strategy in offline two-player zero-sum Markov games. On the other hand, we propose a new assumption named unilateral concentration and design a pessimism-type algorithm that is provably efficient under this assumption. In addition, we show that the unilateral concentration assumption is necessary for learning an NE strategy. Furthermore, our algorithm can achieve minimax sample complexity without any modification for two widely studied settings: dataset with uniform concentration assumption and turn-based Markov games. Our work serves as an important initial step towards understanding offline multi-agent reinforcement learning."
                },
                {
                    "title": "An Adversarial Winograd Schema Challenge at Scale",
                    "abstract": "The Winograd Schema Challenge (WSC), proposed by Levesque et al. (2011) as an alternative to the Turing Test, was originally designed as a pronoun resolution problem that cannot be solved based on statistical patterns in large text corpora. However, recent studies suggest that current WSC datasets, even when composed carefully by experts, are still prone to such biases that statistical methods can exploit. We introduce WINOGRANDE, a new collection of WSC problems that are adversarially constructed to be robust against spurious statistical biases. While the original WSC dataset provided only 273 instances, WINOGRANDE includes 43,985 instances, half of which are determined as adversarial. Key to our approach is a novel adversarial filtering algorithm AFLITE for systematic bias reduction, combined with a careful crowdsourcing design. Despite the significant increase in training data, the performance of existing stateof-the-art methods remains modest (61.6%) and contrasts with high human performance (90.8%) for the binary questions. In addition, WINOGRANDE allows us to use transfer learning for achieving new state-of-the-art results on the original WSC and related datasets. Finally, we discuss how biases lead to overestimating the true capabilities of machine commonsense."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CL",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nAre there simpler algorithms that better scale to modern reinforcement learning applications than existing methods like Proximal Policy Optimization (PPO)?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it could lead to more efficient and scalable reinforcement learning algorithms, which are essential for fine-tuning large generative models. A simpler approach could reduce computational overhead and complexity, making RL more accessible and applicable across various domains. This advancement could inspire future research to explore novel algorithms that prioritize simplicity and scalability, potentially leading to breakthroughs in both theoretical understanding and practical applications in AI.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of existing algorithms like PPO, which require managing multiple models and intricate mechanisms such as value functions and clipping. Naive approaches may fail because they do not adequately address the need for variance reduction and stability in policy updates, especially when dealing with high-dimensional generative models. Additionally, the need to balance performance with simplicity poses a significant technical obstacle, as simplifying the algorithm could lead to performance degradation if not done carefully.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on refining existing algorithms like PPO without questioning their foundational assumptions or exploring alternative approaches. Barriers include a lack of theoretical frameworks that support simpler methods and the entrenched belief in the efficacy of complex algorithms for high-dimensional tasks. The proposed approach, REBEL, differs by reducing the problem to a series of regression tasks, which simplifies the optimization process and eliminates the need for value functions and clipping, thus addressing the limitations of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves the REBEL algorithm, which optimizes policy by iteratively regressing the difference in rewards based on the policy itself. The approach will utilize datasets from both image generation and language modeling tasks, with performance metrics including comparison against PPO and offline techniques like DPO. Expected outcomes include matching or surpassing the performance of PPO while demonstrating a significant reduction in complexity, thereby validating the effectiveness of REBEL as a scalable and simpler alternative for reinforcement learning applications."
            }
        },
        "author_data": {
            "b09869e0-88dd-48f4-936f-47f09db6e4a5": {
                "pk": "b09869e0-88dd-48f4-936f-47f09db6e4a5",
                "name": "Zhaolin Gao",
                "collaborators": [
                    "Kiant\u00e9 Brantley",
                    "Thorsten Joachims",
                    "Joyce Zhou",
                    "Yijia Dai",
                    "Wanyu Lin",
                    "Baochun Li",
                    "Wenhao Zhan",
                    "Jonathan D. Chang",
                    "Gokul Swamy",
                    "Jason D. Lee"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Semi-Supervised Learning",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Reviewer2: Optimizing Review Generation Through Prompt Generation",
                        "abstract": "Recent developments in LLMs offer new opportunities for assisting authors in improving their work. In this paper, we envision a use case where authors can receive LLM-generated reviews that uncover weak points in the current draft. While initial methods for automated review generation already exist, these methods tend to produce reviews that lack detail, and they do not cover the range of opinions that human reviewers produce. To address this shortcoming, we propose an efficient two-stage review generation framework called Reviewer2. Unlike prior work, this approach explicitly models the distribution of possible aspects that the review may address. We show that this leads to more detailed reviews that better cover the range of aspects that human reviewers identify in the draft. As part of the research, we generate a large-scale review dataset of 27k papers and 99k reviews that we annotate with aspect prompts, which we make available as a resource for future research."
                    },
                    {
                        "title": "End-to-end Training for Recommendation with Language-based User Profiles",
                        "abstract": "Many online platforms maintain user profiles for personalization. Unfortunately, these profiles are typically not interpretable or easily modifiable by the user. To remedy this shortcoming, we explore natural language-based user profiles, as they promise enhanced transparency and scrutability of recommender systems. While existing work has shown that language-based profiles from standard LLMs can be effective, such generalist LLMs are unlikely to be optimal for this task. In this paper, we introduce LangPTune, the first end-to-end learning method for training LLMs to produce language-based user profiles that optimize recommendation effectiveness. Through comprehensive evaluations of LangPTune across various training configurations and benchmarks, we demonstrate that our approach significantly outperforms existing profile-based methods. In addition, it approaches performance levels comparable to state-of-the-art, less transparent recommender systems, providing a robust and interpretable alternative to conventional systems. Finally, we validate the relative interpretability of these language-based user profiles through user studies involving crowdworkers and GPT-4-based evaluations. Implementation of LangPTune can be found at https://github.com/ZhaolinGao/LangPTune."
                    },
                    {
                        "title": "Shoestring: Graph-Based Semi-Supervised Learning with Severely Limited Labeled Data",
                        "abstract": "Graph-based semi-supervised learning has been shown to be one of the most effective approaches for classification tasks from a wide range of domains, such as image classification and text classification, as they can exploit the connectivity patterns between labeled and unlabeled samples to improve learning performance. In this work, we advance this effective learning paradigm towards a scenario where labeled data are severely limited. More specifically, we address the problem of graph-based semi-supervised learning in the presence of severely limited labeled samples, and propose a new framework, called {\\em Shoestring}, that improves the learning performance through semantic transfer from these very few labeled samples to large numbers of unlabeled samples.   In particular, our framework learns a metric space in which classification can be performed by computing the similarity to centroid embedding of each class. {\\em Shoestring} is trained in an end-to-end fashion to learn to leverage the semantic knowledge of limited labeled samples as well as their connectivity patterns with large numbers of unlabeled samples simultaneously. By combining {\\em Shoestring} with graph convolutional networks, label propagation and their recent label-efficient variations (IGCN and GLP), we are able to achieve state-of-the-art node classification performance in the presence of very few labeled samples. In addition, we demonstrate the effectiveness of our framework on image classification tasks in the few-shot learning regime, with significant gains on miniImageNet ($2.57\\%\\sim3.59\\%$) and tieredImageNet ($1.05\\%\\sim2.70\\%$)."
                    },
                    {
                        "title": "Regressing the Relative Future: Efficient Policy Optimization for Multi-turn RLHF",
                        "abstract": "Large Language Models (LLMs) have achieved remarkable success at tasks like summarization that involve a single turn of interaction. However, they can still struggle with multi-turn tasks like dialogue that require long-term planning. Previous works on multi-turn dialogue extend single-turn reinforcement learning from human feedback (RLHF) methods to the multi-turn setting by treating all prior dialogue turns as a long context. Such approaches suffer from covariate shift: the conversations in the training set have previous turns generated by some reference policy, which means that low training error may not necessarily correspond to good performance when the learner is actually in the conversation loop. In response, we introduce REgressing the RELative FUture (REFUEL), an efficient policy optimization approach designed to address multi-turn RLHF in LLMs. REFUEL employs a single model to estimate $Q$-values and trains on self-generated data, addressing the covariate shift issue. REFUEL frames the multi-turn RLHF problem as a sequence of regression tasks on iteratively collected datasets, enabling ease of implementation. Theoretically, we prove that REFUEL can match the performance of any policy covered by the training set. Empirically, we evaluate our algorithm by using Llama-3.1-70B-it to simulate a user in conversation with our model. REFUEL consistently outperforms state-of-the-art methods such as DPO and REBEL across various settings. Furthermore, despite having only 8 billion parameters, Llama-3-8B-it fine-tuned with REFUEL outperforms Llama-3.1-70B-it on long multi-turn dialogues. Implementation of REFUEL can be found at https://github.com/ZhaolinGao/REFUEL/, and models trained by REFUEL can be found at https://huggingface.co/Cornell-AGI."
                    }
                ]
            },
            "108b2e21-0164-4e5b-b8b3-c754a246ed0b": {
                "pk": "108b2e21-0164-4e5b-b8b3-c754a246ed0b",
                "name": "Jonathan D. Chang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "67524918-84af-486a-8a68-fd1f38933b93": {
                "pk": "67524918-84af-486a-8a68-fd1f38933b93",
                "name": "Wenhao Zhan",
                "collaborators": [
                    "Jason D. Lee",
                    "Masatoshi Uehara",
                    "Wen Sun",
                    "Haoyue Tang",
                    "Jintao Wang",
                    "Zhuoran Yang",
                    "Baihe Huang",
                    "Audrey Huang",
                    "Nan Jiang",
                    "Gen Li"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Privacy-Preserving Data",
                    "Markov Decision Processes",
                    "Decentralized Learning"
                ],
                "publications": [
                    {
                        "title": "Strong Privacy and Utility Guarantee: Over-the-Air Statistical Estimation",
                        "abstract": "We consider the privacy problem of statistical estimation from distributed data, where users communicate to a central processor over a Gaussian multiple-access channel(MAC). To avoid the inevitable sacrifice of data utility for privacy in digital transmission schemes, we devise an over-the-air estimation strategy which utilizes the additive nature of MAC channel. Using the mutual information between the channel outputs and users' data as the metric, we obtain the privacy bounds for our scheme and validate that it can guarantee strong privacy without incurring larger estimation error. Further, to increase the robustness of our methods, we adjust our primary schemes by adding Gaussian noises locally and derive the corresponding minimax mean squared error under conditional mutual information constraints. Comparing the performance of our methods to the digital ones, we show that the minimax error decreases by $O(\\frac{1}{n})$ in general, which suggests the advantages of over-the-air estimation for preserving data privacy and utility."
                    },
                    {
                        "title": "Delay Optimal Cross-Layer Scheduling Over Markov Channels with Power Constraint",
                        "abstract": "We consider a scenario where a power constrained transmitter delivers randomly arriving packets to the destination over Markov time-varying channel and adapts different transmission power to each channel state in order to guarantee successful transmission. To minimize the expected average transmission delay of each packet, we formulate the problem into a constrained Markov decision process (CMDP). We reveal the queue-length threshold structure of the optimal policy, i.e., the transmitter sends packets if and only if the queue length surpasses a threshold and obtain the optimal cross-layer scheduling strategy through linear programming (LP). Numerical results validate the performance of the proposed strategy and illustrate a delay-power tradeoff in such scenario."
                    },
                    {
                        "title": "Decentralized Optimistic Hyperpolicy Mirror Descent: Provably No-Regret Learning in Markov Games",
                        "abstract": "We study decentralized policy learning in Markov games where we control a single agent to play with nonstationary and possibly adversarial opponents. Our goal is to develop a no-regret online learning algorithm that (i) takes actions based on the local information observed by the agent and (ii) is able to find the best policy in hindsight. For such a problem, the nonstationary state transitions due to the varying opponent pose a significant challenge. In light of a recent hardness result \\citep{liu2022learning}, we focus on the setting where the opponent's previous policies are revealed to the agent for decision making. With such an information structure, we propose a new algorithm, \\underline{D}ecentralized \\underline{O}ptimistic hype\\underline{R}policy m\\underline{I}rror de\\underline{S}cent (DORIS), which achieves $\\sqrt{K}$-regret in the context of general function approximation, where $K$ is the number of episodes. Moreover, when all the agents adopt DORIS, we prove that their mixture policy constitutes an approximate coarse correlated equilibrium. In particular, DORIS maintains a \\textit{hyperpolicy} which is a distribution over the policy space. The hyperpolicy is updated via mirror descent, where the update direction is obtained by an optimistic variant of least-squares policy evaluation. Furthermore, to illustrate the power of our method, we apply DORIS to constrained and vector-valued MDPs, which can be formulated as zero-sum Markov games with a fictitious opponent."
                    },
                    {
                        "title": "Provable Reward-Agnostic Preference-Based Reinforcement Learning",
                        "abstract": "Preference-based Reinforcement Learning (PbRL) is a paradigm in which an RL agent learns to optimize a task using pair-wise preference-based feedback over trajectories, rather than explicit reward signals. While PbRL has demonstrated practical success in fine-tuning language models, existing theoretical work focuses on regret minimization and fails to capture most of the practical frameworks. In this study, we fill in such a gap between theoretical PbRL and practical algorithms by proposing a theoretical reward-agnostic PbRL framework where exploratory trajectories that enable accurate learning of hidden reward functions are acquired before collecting any human feedback. Theoretical analysis demonstrates that our algorithm requires less human feedback for learning the optimal policy under preference-based models with linear parameterization and unknown transitions, compared to the existing theoretical literature. Specifically, our framework can incorporate linear and low-rank MDPs with efficient sample complexity. Additionally, we investigate reward-agnostic RL with action-based comparison feedback and introduce an efficient querying algorithm tailored to this scenario."
                    },
                    {
                        "title": "PAC Reinforcement Learning for Predictive State Representations",
                        "abstract": "In this paper we study online Reinforcement Learning (RL) in partially observable dynamical systems. We focus on the Predictive State Representations (PSRs) model, which is an expressive model that captures other well-known models such as Partially Observable Markov Decision Processes (POMDP). PSR represents the states using a set of predictions of future observations and is defined entirely using observable quantities. We develop a novel model-based algorithm for PSRs that can learn a near optimal policy in sample complexity scaling polynomially with respect to all the relevant parameters of the systems. Our algorithm naturally works with function approximation to extend to systems with potentially large state and observation spaces. We show that given a realizable model class, the sample complexity of learning the near optimal policy only scales polynomially with respect to the statistical complexity of the model class, without any explicit polynomial dependence on the size of the state and observation spaces. Notably, our work is the first work that shows polynomial sample complexities to compete with the globally optimal policy in PSRs. Finally, we demonstrate how our general theorem can be directly used to derive sample complexity bounds for special models including $m$-step weakly revealing and $m$-step decodable tabular POMDPs, POMDPs with low-rank latent transition, and POMDPs with linear emission and latent transition."
                    },
                    {
                        "title": "Offline Reinforcement Learning with Realizability and Single-policy Concentrability",
                        "abstract": "Sample-efficiency guarantees for offline reinforcement learning (RL) often rely on strong assumptions on both the function classes (e.g., Bellman-completeness) and the data coverage (e.g., all-policy concentrability). Despite the recent efforts on relaxing these assumptions, existing works are only able to relax one of the two factors, leaving the strong assumption on the other factor intact. As an important open problem, can we achieve sample-efficient offline RL with weak assumptions on both factors?   In this paper we answer the question in the positive. We analyze a simple algorithm based on the primal-dual formulation of MDPs, where the dual variables (discounted occupancy) are modeled using a density-ratio function against offline data. With proper regularization, we show that the algorithm enjoys polynomial sample complexity, under only realizability and single-policy concentrability. We also provide alternative analyses based on different assumptions to shed light on the nature of primal-dual algorithms for offline RL."
                    },
                    {
                        "title": "Reward-agnostic Fine-tuning: Provable Statistical Benefits of Hybrid Reinforcement Learning",
                        "abstract": "This paper studies tabular reinforcement learning (RL) in the hybrid setting, which assumes access to both an offline dataset and online interactions with the unknown environment. A central question boils down to how to efficiently utilize online data collection to strengthen and complement the offline dataset and enable effective policy fine-tuning. Leveraging recent advances in reward-agnostic exploration and model-based offline RL, we design a three-stage hybrid RL algorithm that beats the best of both worlds -- pure offline RL and pure online RL -- in terms of sample complexities. The proposed algorithm does not require any reward information during data collection. Our theory is developed based on a new notion called single-policy partial concentrability, which captures the trade-off between distribution mismatch and miscoverage and guides the interplay between offline and online data."
                    }
                ]
            },
            "5811938c-dc35-4cfb-8bc1-20e2e95edbd7": {
                "pk": "5811938c-dc35-4cfb-8bc1-20e2e95edbd7",
                "name": "Owen Oertell",
                "collaborators": [
                    "Wen Sun",
                    "Jonathan D. Chang",
                    "Kiant\u00e9 Brantley",
                    "Kaiwen Wang",
                    "Alekh Agarwal",
                    "Nathan Kallus",
                    "Yiyi Zhang",
                    "Wenhao Zhan",
                    "Dipendra Misra",
                    "Jason D. Lee"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Generative Models",
                    "Human Feedback",
                    "Distributional Learning"
                ],
                "publications": [
                    {
                        "title": "More Benefits of Being Distributional: Second-Order Bounds for Reinforcement Learning",
                        "abstract": "In this paper, we prove that Distributional Reinforcement Learning (DistRL), which learns the return distribution, can obtain second-order bounds in both online and offline RL in general settings with function approximation. Second-order bounds are instance-dependent bounds that scale with the variance of return, which we prove are tighter than the previously known small-loss bounds of distributional RL. To the best of our knowledge, our results are the first second-order bounds for low-rank MDPs and for offline RL. When specializing to contextual bandits (one-step RL problem), we show that a distributional learning based optimism algorithm achieves a second-order worst-case regret bound, and a second-order gap dependent bound, simultaneously. We also empirically demonstrate the benefit of DistRL in contextual bandits on real-world datasets. We highlight that our analysis with DistRL is relatively simple, follows the general framework of optimism in the face of uncertainty and does not require weighted regression. Our results suggest that DistRL is a promising framework for obtaining second-order bounds in general RL settings, thus further reinforcing the benefits of DistRL."
                    },
                    {
                        "title": "RL for Consistency Models: Faster Reward Guided Text-to-Image Generation",
                        "abstract": "Reinforcement learning (RL) has improved guided image generation with diffusion models by directly optimizing rewards that capture image quality, aesthetics, and instruction following capabilities. However, the resulting generative policies inherit the same iterative sampling process of diffusion models that causes slow generation. To overcome this limitation, consistency models proposed learning a new class of generative models that directly map noise to data, resulting in a model that can generate an image in as few as one sampling iteration. In this work, to optimize text-to-image generative models for task specific rewards and enable fast training and inference, we propose a framework for fine-tuning consistency models via RL. Our framework, called Reinforcement Learning for Consistency Model (RLCM), frames the iterative inference process of a consistency model as an RL procedure. Comparing to RL finetuned diffusion models, RLCM trains significantly faster, improves the quality of the generation measured under the reward objectives, and speeds up the inference procedure by generating high quality images with as few as two inference steps. Experimentally, we show that RLCM can adapt text-to-image consistency models to objectives that are challenging to express with prompting, such as image compressibility, and those derived from human feedback, such as aesthetic quality. Our code is available at https://rlcm.owenoertell.com."
                    },
                    {
                        "title": "Dataset Reset Policy Optimization for RLHF",
                        "abstract": "Reinforcement Learning (RL) from Human Preference-based feedback is a popular paradigm for fine-tuning generative models, which has produced impressive models such as GPT-4 and Claude3 Opus. This framework often consists of two steps: learning a reward model from an offline preference dataset followed by running online RL to optimize the learned reward model. In this work, leveraging the idea of reset, we propose a new RLHF algorithm with provable guarantees. Motivated by the fact that offline preference dataset provides informative states (i.e., data that is preferred by the labelers), our new algorithm, Dataset Reset Policy Optimization (DR-PO), integrates the existing offline preference dataset into the online policy training procedure via dataset reset: it directly resets the policy optimizer to the states in the offline dataset, instead of always starting from the initial state distribution. In theory, we show that DR-PO learns to perform at least as good as any policy that is covered by the offline dataset under general function approximation with finite sample complexity. In experiments, we demonstrate that on both the TL;DR summarization and the Anthropic Helpful Harmful (HH) dataset, the generation from DR-PO is better than that from Proximal Policy Optimization (PPO) and Direction Preference Optimization (DPO), under the metric of GPT4 win-rate. Code for this work can be found at https://github.com/Cornell-RL/drpo."
                    }
                ]
            },
            "2870ebff-c7a4-4374-a688-5af13eecfd9e": {
                "pk": "2870ebff-c7a4-4374-a688-5af13eecfd9e",
                "name": "Gokul Swamy",
                "collaborators": [
                    "Zhiwei Steven Wu",
                    "Sanjiban Choudhury",
                    "J. Andrew Bagnell",
                    "Wen Sun",
                    "Kiant\u00e9 Brantley",
                    "Patrick Chao",
                    "Alexander Li",
                    "Jens Schulz",
                    "Rohan Choudhury",
                    "Dylan Hadfield-Menell"
                ],
                "domain": [
                    "Imitation Learning",
                    "Reinforcement Learning",
                    "Human-Robot Interaction",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Generative Models for Pose Transfer",
                        "abstract": "We investigate nearest neighbor and generative models for transferring pose between persons. We take in a video of one person performing a sequence of actions and attempt to generate a video of another person performing the same actions. Our generative model (pix2pix) outperforms k-NN at both generating corresponding frames and generalizing outside the demonstrated action set. Our most salient contribution is determining a pipeline (pose detection, face detection, k-NN based pairing) that is effective at perform-ing the desired task. We also detail several iterative improvements and failure modes."
                    },
                    {
                        "title": "On the Utility of Model Learning in HRI",
                        "abstract": "Fundamental to robotics is the debate between model-based and model-free learning: should the robot build an explicit model of the world, or learn a policy directly? In the context of HRI, part of the world to be modeled is the human. One option is for the robot to treat the human as a black box and learn a policy for how they act directly. But it can also model the human as an agent, and rely on a \"theory of mind\" to guide or bias the learning (grey box). We contribute a characterization of the performance of these methods for an autonomous driving task under the optimistic case of having an ideal theory of mind, as well as under different scenarios in which the assumptions behind the robot's theory of mind for the human are wrong, as they inevitably will be in practice."
                    },
                    {
                        "title": "A Critique of Strictly Batch Imitation Learning",
                        "abstract": "Recent work by Jarrett et al. attempts to frame the problem of offline imitation learning (IL) as one of learning a joint energy-based model, with the hope of out-performing standard behavioral cloning. We suggest that notational issues obscure how the psuedo-state visitation distribution the authors propose to optimize might be disconnected from the policy's $\\textit{true}$ state visitation distribution. We further construct natural examples where the parameter coupling advocated by Jarrett et al. leads to inconsistent estimates of the expert's policy, unlike behavioral cloning."
                    },
                    {
                        "title": "Game-Theoretic Algorithms for Conditional Moment Matching",
                        "abstract": "A variety of problems in econometrics and machine learning, including instrumental variable regression and Bellman residual minimization, can be formulated as satisfying a set of conditional moment restrictions (CMR). We derive a general, game-theoretic strategy for satisfying CMR that scales to nonlinear problems, is amenable to gradient-based optimization, and is able to account for finite sample uncertainty. We recover the approaches of Dikkala et al. and Dai et al. as special cases of our general framework before detailing various extensions and how to efficiently solve the game defined by CMR."
                    },
                    {
                        "title": "Diffusing States and Matching Scores: A New Framework for Imitation Learning",
                        "abstract": "Adversarial Imitation Learning is traditionally framed as a two-player zero-sum game between a learner and an adversarially chosen cost function, and can therefore be thought of as the sequential generalization of a Generative Adversarial Network (GAN). A prominent example of this framework is Generative Adversarial Imitation Learning (GAIL). However, in recent years, diffusion models have emerged as a non-adversarial alternative to GANs that merely require training a score function via regression, yet produce generations of a higher quality. In response, we investigate how to lift insights from diffusion modeling to the sequential setting. We propose diffusing states and performing score-matching along diffused states to measure the discrepancy between the expert's and learner's states. Thus, our approach only requires training score functions to predict noises via standard regression, making it significantly easier and more stable to train than adversarial methods. Theoretically, we prove first- and second-order instance-dependent bounds with linear scaling in the horizon, proving that our approach avoids the compounding errors that stymie offline approaches to imitation learning. Empirically, we show our approach outperforms GAN-style imitation learning baselines across various continuous control problems, including complex tasks like controlling humanoids to walk, sit, and crawl."
                    },
                    {
                        "title": "Scaled Autonomy: Enabling Human Operators to Control Robot Fleets",
                        "abstract": "Autonomous robots often encounter challenging situations where their control policies fail and an expert human operator must briefly intervene, e.g., through teleoperation. In settings where multiple robots act in separate environments, a single human operator can manage a fleet of robots by identifying and teleoperating one robot at any given time. The key challenge is that users have limited attention: as the number of robots increases, users lose the ability to decide which robot requires teleoperation the most. Our goal is to automate this decision, thereby enabling users to supervise more robots than their attention would normally allow for. Our insight is that we can model the user's choice of which robot to control as an approximately optimal decision that maximizes the user's utility function. We learn a model of the user's preferences from observations of the user's choices in easy settings with a few robots, and use it in challenging settings with more robots to automatically identify which robot the user would most likely choose to control, if they were able to evaluate the states of all robots at all times. We run simulation experiments and a user study with twelve participants that show our method can be used to assist users in performing a simulated navigation task. We also run a hardware demonstration that illustrates how our method can be applied to a real-world mobile robot navigation task."
                    },
                    {
                        "title": "Multi-Agent Imitation Learning: Value is Easy, Regret is Hard",
                        "abstract": "We study a multi-agent imitation learning (MAIL) problem where we take the perspective of a learner attempting to coordinate a group of agents based on demonstrations of an expert doing so. Most prior work in MAIL essentially reduces the problem to matching the behavior of the expert within the support of the demonstrations. While doing so is sufficient to drive the value gap between the learner and the expert to zero under the assumption that agents are non-strategic, it does not guarantee robustness to deviations by strategic agents. Intuitively, this is because strategic deviations can depend on a counterfactual quantity: the coordinator's recommendations outside of the state distribution their recommendations induce. In response, we initiate the study of an alternative objective for MAIL in Markov Games we term the regret gap that explicitly accounts for potential deviations by agents in the group. We first perform an in-depth exploration of the relationship between the value and regret gaps. First, we show that while the value gap can be efficiently minimized via a direct extension of single-agent IL algorithms, even value equivalence can lead to an arbitrarily large regret gap. This implies that achieving regret equivalence is harder than achieving value equivalence in MAIL. We then provide a pair of efficient reductions to no-regret online convex optimization that are capable of minimizing the regret gap (a) under a coverage assumption on the expert (MALICE) or (b) with access to a queryable expert (BLADES)."
                    },
                    {
                        "title": "Of Moments and Matching: A Game-Theoretic Framework for Closing the Imitation Gap",
                        "abstract": "We provide a unifying view of a large family of previous imitation learning algorithms through the lens of moment matching. At its core, our classification scheme is based on whether the learner attempts to match (1) reward or (2) action-value moments of the expert's behavior, with each option leading to differing algorithmic approaches. By considering adversarially chosen divergences between learner and expert behavior, we are able to derive bounds on policy performance that apply for all algorithms in each of these classes, the first to our knowledge. We also introduce the notion of moment recoverability, implicit in many previous analyses of imitation learning, which allows us to cleanly delineate how well each algorithmic family is able to mitigate compounding errors. We derive three novel algorithm templates (AdVIL, AdRIL, and DAeQuIL) with strong guarantees, simple implementation, and competitive empirical performance."
                    },
                    {
                        "title": "Causal Imitation Learning under Temporally Correlated Noise",
                        "abstract": "We develop algorithms for imitation learning from policy data that was corrupted by temporally correlated noise in expert actions. When noise affects multiple timesteps of recorded data, it can manifest as spurious correlations between states and actions that a learner might latch on to, leading to poor policy performance. To break up these spurious correlations, we apply modern variants of the instrumental variable regression (IVR) technique of econometrics, enabling us to recover the underlying policy without requiring access to an interactive expert. In particular, we present two techniques, one of a generative-modeling flavor (DoubIL) that can utilize access to a simulator, and one of a game-theoretic flavor (ResiduIL) that can be run entirely offline. We find both of our algorithms compare favorably to behavioral cloning on simulated control tasks."
                    },
                    {
                        "title": "Inverse Reinforcement Learning without Reinforcement Learning",
                        "abstract": "Inverse Reinforcement Learning (IRL) is a powerful set of techniques for imitation learning that aims to learn a reward function that rationalizes expert demonstrations. Unfortunately, traditional IRL methods suffer from a computational weakness: they require repeatedly solving a hard reinforcement learning (RL) problem as a subroutine. This is counter-intuitive from the viewpoint of reductions: we have reduced the easier problem of imitation learning to repeatedly solving the harder problem of RL. Another thread of work has proved that access to the side-information of the distribution of states where a strong policy spends time can dramatically reduce the sample and computational complexities of solving an RL problem. In this work, we demonstrate for the first time a more informed imitation learning reduction where we utilize the state distribution of the expert to alleviate the global exploration component of the RL subroutine, providing an exponential speedup in theory. In practice, we find that we are able to significantly speed up the prior art on continuous control tasks."
                    },
                    {
                        "title": "Learning Shared Safety Constraints from Multi-task Demonstrations",
                        "abstract": "Regardless of the particular task we want them to perform in an environment, there are often shared safety constraints we want our agents to respect. For example, regardless of whether it is making a sandwich or clearing the table, a kitchen robot should not break a plate. Manually specifying such a constraint can be both time-consuming and error-prone. We show how to learn constraints from expert demonstrations of safe task completion by extending inverse reinforcement learning (IRL) techniques to the space of constraints. Intuitively, we learn constraints that forbid highly rewarding behavior that the expert could have taken but chose not to. Unfortunately, the constraint learning problem is rather ill-posed and typically leads to overly conservative constraints that forbid all behavior that the expert did not take. We counter this by leveraging diverse demonstrations that naturally occur in multi-task settings to learn a tighter set of constraints. We validate our method with simulation experiments on high-dimensional continuous control tasks."
                    },
                    {
                        "title": "Sequence Model Imitation Learning with Unobserved Contexts",
                        "abstract": "We consider imitation learning problems where the learner's ability to mimic the expert increases throughout the course of an episode as more information is revealed. One example of this is when the expert has access to privileged information: while the learner might not be able to accurately reproduce expert behavior early on in an episode, by considering the entire history of states and actions, they might be able to eventually identify the hidden context and act as the expert would. We prove that on-policy imitation learning algorithms (with or without access to a queryable expert) are better equipped to handle these sorts of asymptotically realizable problems than off-policy methods. This is because on-policy algorithms provably learn to recover from their initially suboptimal actions, while off-policy methods treat their suboptimal past actions as though they came from the expert. This often manifests as a latching behavior: a naive repetition of past actions. We conduct experiments in a toy bandit domain that show that there exist sharp phase transitions of whether off-policy approaches are able to match expert performance asymptotically, in contrast to the uniformly good performance of on-policy approaches. We demonstrate that on several continuous control tasks, on-policy approaches are able to use history to identify the context while off-policy approaches actually perform worse when given access to history."
                    },
                    {
                        "title": "A Minimaximalist Approach to Reinforcement Learning from Human Feedback",
                        "abstract": "We present Self-Play Preference Optimization (SPO), an algorithm for reinforcement learning from human feedback. Our approach is minimalist in that it does not require training a reward model nor unstable adversarial training and is therefore rather simple to implement. Our approach is maximalist in that it provably handles non-Markovian, intransitive, and stochastic preferences while being robust to the compounding errors that plague offline approaches to sequential prediction. To achieve the preceding qualities, we build upon the concept of a Minimax Winner (MW), a notion of preference aggregation from the social choice theory literature that frames learning from preferences as a zero-sum game between two policies. By leveraging the symmetry of this game, we prove that rather than using the traditional technique of dueling two policies to compute the MW, we can simply have a single agent play against itself while maintaining strong convergence guarantees. Practically, this corresponds to sampling multiple trajectories from a policy, asking a preference or teacher model to compare them, and then using the proportion of wins as the reward for a particular trajectory. We demonstrate that on a suite of continuous control tasks, we are able to learn significantly more efficiently than reward-model based approaches while maintaining robustness to the intransitive and stochastic preferences that frequently occur in practice when aggregating human judgments."
                    },
                    {
                        "title": "Hybrid Inverse Reinforcement Learning",
                        "abstract": "The inverse reinforcement learning approach to imitation learning is a double-edged sword. On the one hand, it can enable learning from a smaller number of expert demonstrations with more robustness to error compounding than behavioral cloning approaches. On the other hand, it requires that the learner repeatedly solve a computationally expensive reinforcement learning (RL) problem. Often, much of this computation is wasted searching over policies very dissimilar to the expert's. In this work, we propose using hybrid RL -- training on a mixture of online and expert data -- to curtail unnecessary exploration. Intuitively, the expert data focuses the learner on good states during training, which reduces the amount of exploration required to compute a strong policy. Notably, such an approach doesn't need the ability to reset the learner to arbitrary states in the environment, a requirement of prior work in efficient inverse RL. More formally, we derive a reduction from inverse RL to expert-competitive RL (rather than globally optimal RL) that allows us to dramatically reduce interaction during the inner policy search loop while maintaining the benefits of the IRL approach. This allows us to derive both model-free and model-based hybrid inverse RL algorithms with strong policy performance guarantees. Empirically, we find that our approaches are significantly more sample efficient than standard inverse RL and several other baselines on a suite of continuous control tasks."
                    },
                    {
                        "title": "The Importance of Online Data: Understanding Preference Fine-tuning via Coverage",
                        "abstract": "Learning from human preference data has emerged as the dominant paradigm for fine-tuning large language models (LLMs). The two most common families of techniques -- online reinforcement learning (RL) such as Proximal Policy Optimization (PPO) and offline contrastive methods such as Direct Preference Optimization (DPO) -- were positioned as equivalent in prior work due to the fact that both have to start from the same offline preference dataset. To further expand our theoretical understanding of the similarities and differences between online and offline techniques for preference fine-tuning, we conduct a rigorous analysis through the lens of dataset coverage, a concept that captures how the training data covers the test distribution and is widely used in RL. We prove that a global coverage condition is both necessary and sufficient for offline contrastive methods to converge to the optimal policy, but a weaker partial coverage condition suffices for online RL methods. This separation provides one explanation of why online RL methods can perform better than offline methods, especially when the offline preference data is not diverse enough. Finally, motivated by our preceding theoretical observations, we derive a hybrid preference optimization (HyPO) algorithm that uses offline data for contrastive-based preference optimization and online data for KL regularization. Theoretically and empirically, we demonstrate that HyPO is more performant than its pure offline counterpart DPO, while still preserving its computation and memory efficiency."
                    },
                    {
                        "title": "EvIL: Evolution Strategies for Generalisable Imitation Learning",
                        "abstract": "Often times in imitation learning (IL), the environment we collect expert demonstrations in and the environment we want to deploy our learned policy in aren't exactly the same (e.g. demonstrations collected in simulation but deployment in the real world). Compared to policy-centric approaches to IL like behavioural cloning, reward-centric approaches like inverse reinforcement learning (IRL) often better replicate expert behaviour in new environments. This transfer is usually performed by optimising the recovered reward under the dynamics of the target environment. However, (a) we find that modern deep IL algorithms frequently recover rewards which induce policies far weaker than the expert, even in the same environment the demonstrations were collected in. Furthermore, (b) these rewards are often quite poorly shaped, necessitating extensive environment interaction to optimise effectively. We provide simple and scalable fixes to both of these concerns. For (a), we find that reward model ensembles combined with a slightly different training objective significantly improves re-training and transfer performance. For (b), we propose a novel evolution-strategies based method EvIL to optimise for a reward-shaping term that speeds up re-training in the target environment, closing a gap left open by the classical theory of IRL. On a suite of continuous control tasks, we are able to re-train policies in target (and source) environments more interaction-efficiently than prior work."
                    },
                    {
                        "title": "Minimax Optimal Online Imitation Learning via Replay Estimation",
                        "abstract": "Online imitation learning is the problem of how best to mimic expert demonstrations, given access to the environment or an accurate simulator. Prior work has shown that in the infinite sample regime, exact moment matching achieves value equivalence to the expert policy. However, in the finite sample regime, even if one has no optimization error, empirical variance can lead to a performance gap that scales with $H^2 / N$ for behavioral cloning and $H / \\sqrt{N}$ for online moment matching, where $H$ is the horizon and $N$ is the size of the expert dataset. We introduce the technique of replay estimation to reduce this empirical variance: by repeatedly executing cached expert actions in a stochastic simulator, we compute a smoother expert visitation distribution estimate to match. In the presence of general function approximation, we prove a meta theorem reducing the performance gap of our approach to the parameter estimation error for offline classification (i.e. learning the expert policy). In the tabular setting or with linear function approximation, our meta theorem shows that the performance gap incurred by our approach achieves the optimal $\\widetilde{O} \\left( \\min({H^{3/2}} / {N}, {H} / {\\sqrt{N}} \\right)$ dependency, under significantly weaker assumptions compared to prior work. We implement multiple instantiations of our approach on several continuous control tasks and find that we are able to significantly improve policy performance across a variety of dataset sizes."
                    },
                    {
                        "title": "The Virtues of Pessimism in Inverse Reinforcement Learning",
                        "abstract": "Inverse Reinforcement Learning (IRL) is a powerful framework for learning complex behaviors from expert demonstrations. However, it traditionally requires repeatedly solving a computationally expensive reinforcement learning (RL) problem in its inner loop. It is desirable to reduce the exploration burden by leveraging expert demonstrations in the inner-loop RL. As an example, recent work resets the learner to expert states in order to inform the learner of high-reward expert states. However, such an approach is infeasible in the real world. In this work, we consider an alternative approach to speeding up the RL subroutine in IRL: \\emph{pessimism}, i.e., staying close to the expert's data distribution, instantiated via the use of offline RL algorithms. We formalize a connection between offline RL and IRL, enabling us to use an arbitrary offline RL algorithm to improve the sample efficiency of IRL. We validate our theory experimentally by demonstrating a strong correlation between the efficacy of an offline RL algorithm and how well it works as part of an IRL procedure. By using a strong offline RL algorithm as part of an IRL procedure, we are able to find policies that match expert performance significantly more efficiently than the prior art."
                    },
                    {
                        "title": "Regressing the Relative Future: Efficient Policy Optimization for Multi-turn RLHF",
                        "abstract": "Large Language Models (LLMs) have achieved remarkable success at tasks like summarization that involve a single turn of interaction. However, they can still struggle with multi-turn tasks like dialogue that require long-term planning. Previous works on multi-turn dialogue extend single-turn reinforcement learning from human feedback (RLHF) methods to the multi-turn setting by treating all prior dialogue turns as a long context. Such approaches suffer from covariate shift: the conversations in the training set have previous turns generated by some reference policy, which means that low training error may not necessarily correspond to good performance when the learner is actually in the conversation loop. In response, we introduce REgressing the RELative FUture (REFUEL), an efficient policy optimization approach designed to address multi-turn RLHF in LLMs. REFUEL employs a single model to estimate $Q$-values and trains on self-generated data, addressing the covariate shift issue. REFUEL frames the multi-turn RLHF problem as a sequence of regression tasks on iteratively collected datasets, enabling ease of implementation. Theoretically, we prove that REFUEL can match the performance of any policy covered by the training set. Empirically, we evaluate our algorithm by using Llama-3.1-70B-it to simulate a user in conversation with our model. REFUEL consistently outperforms state-of-the-art methods such as DPO and REBEL across various settings. Furthermore, despite having only 8 billion parameters, Llama-3-8B-it fine-tuned with REFUEL outperforms Llama-3.1-70B-it on long multi-turn dialogues. Implementation of REFUEL can be found at https://github.com/ZhaolinGao/REFUEL/, and models trained by REFUEL can be found at https://huggingface.co/Cornell-AGI."
                    }
                ]
            },
            "3f63d232-8cff-4d4d-a173-952dbff42866": {
                "pk": "3f63d232-8cff-4d4d-a173-952dbff42866",
                "name": "Kiant\u00e9 Brantley",
                "collaborators": [
                    "Hal Daum\u00e9 III",
                    "Amr Sharaf",
                    "Zhichong Fang",
                    "Sarah Dean",
                    "Thorsten Joachims",
                    "Anne Wu",
                    "Yoav Artzi",
                    "Soroush Mehri",
                    "Geoffrey J. Gordon",
                    "Sean Welleck"
                ],
                "domain": [
                    "Imitation Learning",
                    "Reinforcement Learning",
                    "Active Learning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Active Imitation Learning with Noisy Guidance",
                        "abstract": "Imitation learning algorithms provide state-of-the-art results on many structured prediction tasks by learning near-optimal search policies. Such algorithms assume training-time access to an expert that can provide the optimal action at any queried state; unfortunately, the number of such queries is often prohibitive, frequently rendering these approaches impractical. To combat this query complexity, we consider an active learning setting in which the learning algorithm has additional access to a much cheaper noisy heuristic that provides noisy guidance. Our algorithm, LEAQI, learns a difference classifier that predicts when the expert is likely to disagree with the heuristic, and queries the expert only when necessary. We apply LEAQI to three sequence labeling tasks, demonstrating significantly fewer queries to the expert and comparable (or better) accuracies over a passive approach."
                    },
                    {
                        "title": "Ranking with Long-Term Constraints",
                        "abstract": "The feedback that users provide through their choices (e.g., clicks, purchases) is one of the most common types of data readily available for training search and recommendation algorithms. However, myopically training systems based on choice data may only improve short-term engagement, but not the long-term sustainability of the platform and the long-term benefits to its users, content providers, and other stakeholders. In this paper, we thus develop a new framework in which decision makers (e.g., platform operators, regulators, users) can express long-term goals for the behavior of the platform (e.g., fairness, revenue distribution, legal requirements). These goals take the form of exposure or impact targets that go well beyond individual sessions, and we provide new control-based algorithms to achieve these goals. In particular, the controllers are designed to achieve the stated long-term goals with minimum impact on short-term engagement. Beyond the principled theoretical derivation of the controllers, we evaluate the algorithms on both synthetic and real-world data. While all controllers perform well, we find that they provide interesting trade-offs in efficiency, robustness, and the ability to plan ahead."
                    },
                    {
                        "title": "A Surprising Failure? Multimodal LLMs and the NLVR Challenge",
                        "abstract": "This study evaluates three state-of-the-art MLLMs -- GPT-4V, Gemini Pro, and the open-source model IDEFICS -- on the compositional natural language vision reasoning task NLVR. Given a human-written sentence paired with a synthetic image, this task requires the model to determine the truth value of the sentence with respect to the image. Despite the strong performance demonstrated by these models, we observe they perform poorly on NLVR, which was constructed to require compositional and spatial reasoning, and to be robust for semantic and systematic biases."
                    },
                    {
                        "title": "Successor Feature Sets: Generalizing Successor Representations Across Policies",
                        "abstract": "Successor-style representations have many advantages for reinforcement learning: for example, they can help an agent generalize from past experience to new goals, and they have been proposed as explanations of behavioral and neural data from human and animal learners. They also form a natural bridge between model-based and model-free RL methods: like the former they make predictions about future experiences, and like the latter they allow efficient prediction of total discounted rewards. However, successor-style representations are not optimized to generalize across policies: typically, we maintain a limited-length list of policies, and share information among them by representation learning or GPI. Successor-style representations also typically make no provision for gathering information or reasoning about latent variables. To address these limitations, we bring together ideas from predictive state representations, belief space value iteration, successor features, and convex analysis: we develop a new, general successor-style representation, together with a Bellman equation that connects multiple sources of information within this representation, including different latent states, policies, and reward functions. The new representation is highly expressive: for example, it lets us efficiently read off an optimal policy for a new reward function, or a policy that imitates a new demonstration. For this paper, we focus on exact computation of the new representation in small, known environments, since even this restricted setting offers plenty of interesting questions. Our implementation does not scale to large, unknown environments -- nor would we expect it to, since it generalizes POMDP value iteration, which is difficult to scale. However, we believe that future work will allow us to extend our ideas to approximate reasoning in large, unknown environments."
                    },
                    {
                        "title": "Non-Monotonic Sequential Text Generation",
                        "abstract": "Standard sequential generation methods assume a pre-specified generation order, such as text generation methods which generate words from left to right. In this work, we propose a framework for training models of text generation that operate in non-monotonic orders; the model directly learns good orders, without any additional annotation. Our framework operates by generating a word at an arbitrary position, and then recursively generating words to its left and then words to its right, yielding a binary tree. Learning is framed as imitation learning, including a coaching method which moves from imitating an oracle to reinforcing the policy's own preferences. Experimental results demonstrate that using the proposed method, it is possible to learn policies which generate text without pre-specifying a generation order, while achieving competitive performance with conventional left-to-right generation."
                    },
                    {
                        "title": "Constrained episodic reinforcement learning in concave-convex and knapsack settings",
                        "abstract": "We propose an algorithm for tabular episodic reinforcement learning with constraints. We provide a modular analysis with strong theoretical guarantees for settings with concave rewards and convex constraints, and for settings with hard constraints (knapsacks). Most of the previous work in constrained reinforcement learning is limited to linear constraints, and the remaining work focuses on either the feasibility question or settings with a single episode. Our experiments demonstrate that the proposed algorithm significantly outperforms these approaches in existing constrained episodic environments."
                    }
                ]
            },
            "7b9350b4-15ee-47fd-91a1-635d92ea2973": {
                "pk": "7b9350b4-15ee-47fd-91a1-635d92ea2973",
                "name": "Thorsten Joachims",
                "collaborators": [
                    "Filip Radlinski",
                    "Pannaga Shivaswamy",
                    "Ashudeep Singh",
                    "Yuta Saito",
                    "Lequn Wang",
                    "Carmel Domshlak",
                    "Adith Swaminathan",
                    "Joyce Zhou",
                    "Karthik Raman",
                    "Richa Rastogi"
                ],
                "domain": [
                    "Information Retrieval",
                    "Machine Learning",
                    "Fairness in AI",
                    "Recommendation Systems"
                ],
                "publications": [
                    {
                        "title": "Evaluating the Robustness of Learning from Implicit Feedback",
                        "abstract": "This paper evaluates the robustness of learning from implicit feedback in web search. In particular, we create a model of user behavior by drawing upon user studies in laboratory and real-world settings. The model is used to understand the effect of user behavior on the performance of a learning algorithm for ranked retrieval. We explore a wide range of possible user behaviors and find that learning from implicit feedback can be surprisingly robust. This complements previous results that demonstrated our algorithm's effectiveness in a real-world search engine application."
                    },
                    {
                        "title": "Unstructuring User Preferences: Efficient Non-Parametric Utility Revelation",
                        "abstract": "Tackling the problem of ordinal preference revelation and reasoning, we propose a novel methodology for generating an ordinal utility function from a set of qualitative preference statements. To the best of our knowledge, our proposal constitutes the first nonparametric solution for this problem that is both efficient and semantically sound. Our initial experiments provide strong evidence for practical effectiveness of our approach."
                    },
                    {
                        "title": "Query Chains: Learning to Rank from Implicit Feedback",
                        "abstract": "This paper presents a novel approach for using clickthrough data to learn ranked retrieval functions for web search results. We observe that users searching the web often perform a sequence, or chain, of queries with a similar information need. Using query chains, we generate new types of preference judgments from search engine logs, thus taking advantage of user intelligence in reformulating queries. To validate our method we perform a controlled user study comparing generated preference judgments to explicit relevance judgments. We also implemented a real-world search engine to test our approach, using a modified ranking SVM to learn an improved ranking function from preference data. Our results demonstrate significant improvements in the ranking given by the search engine. The learned rankings outperform both a static ranking function, as well as one trained without considering query chains."
                    },
                    {
                        "title": "Minimally Invasive Randomization for Collecting Unbiased Preferences from Clickthrough Logs",
                        "abstract": "Clickthrough data is a particularly inexpensive and plentiful resource to obtain implicit relevance feedback for improving and personalizing search engines. However, it is well known that the probability of a user clicking on a result is strongly biased toward documents presented higher in the result set irrespective of relevance. We introduce a simple method to modify the presentation of search results that provably gives relevance judgments that are unaffected by presentation bias under reasonable assumptions. We validate this property of the training data in interactive real world experiments. Finally, we show that using these unbiased relevance judgments learning methods can be guaranteed to converge to an ideal ranking given sufficient data."
                    },
                    {
                        "title": "Online Structured Prediction via Coactive Learning",
                        "abstract": "We propose Coactive Learning as a model of interaction between a learning system and a human user, where both have the common goal of providing results of maximum utility to the user. At each step, the system (e.g. search engine) receives a context (e.g. query) and predicts an object (e.g. ranking). The user responds by correcting the system if necessary, providing a slightly improved -- but not necessarily optimal -- object as feedback. We argue that such feedback can often be inferred from observable user behavior, for example, from clicks in web-search. Evaluating predictions by their cardinal utility to the user, we propose efficient learning algorithms that have ${\\cal O}(\\frac{1}{\\sqrt{T}})$ average regret, even though the learning algorithm never observes cardinal utility values as in conventional online learning. We demonstrate the applicability of our model and learning algorithms on a movie recommendation task, as well as ranking for web-search."
                    },
                    {
                        "title": "Counterfactual Risk Minimization: Learning from Logged Bandit Feedback",
                        "abstract": "We develop a learning principle and an efficient algorithm for batch learning from logged bandit feedback. This learning setting is ubiquitous in online systems (e.g., ad placement, web search, recommendation), where an algorithm makes a prediction (e.g., ad ranking) for a given input (e.g., query) and observes bandit feedback (e.g., user clicks on presented ads). We first address the counterfactual nature of the learning problem through propensity scoring. Next, we prove generalization error bounds that account for the variance of the propensity-weighted empirical risk estimator. These constructive bounds give rise to the Counterfactual Risk Minimization (CRM) principle. We show how CRM can be used to derive a new learning method -- called Policy Optimizer for Exponential Models (POEM) -- for learning stochastic linear rules for structured output prediction. We present a decomposition of the POEM objective that enables efficient stochastic gradient optimization. POEM is evaluated on several multi-label classification problems showing substantially improved robustness and generalization performance compared to the state-of-the-art."
                    },
                    {
                        "title": "Policy Learning for Fairness in Ranking",
                        "abstract": "Conventional Learning-to-Rank (LTR) methods optimize the utility of the rankings to the users, but they are oblivious to their impact on the ranked items. However, there has been a growing understanding that the latter is important to consider for a wide range of ranking applications (e.g. online marketplaces, job placement, admissions). To address this need, we propose a general LTR framework that can optimize a wide range of utility metrics (e.g. NDCG) while satisfying fairness of exposure constraints with respect to the items. This framework expands the class of learnable ranking functions to stochastic ranking policies, which provides a language for rigorously expressing fairness specifications. Furthermore, we provide a new LTR algorithm called Fair-PG-Rank for directly searching the space of fair ranking policies via a policy-gradient approach. Beyond the theoretical evidence in deriving the framework and the algorithm, we provide empirical results on simulated and real-world datasets verifying the effectiveness of the approach in individual and group-fairness settings."
                    },
                    {
                        "title": "Fairness of Exposure in Rankings",
                        "abstract": "Rankings are ubiquitous in the online world today. As we have transitioned from finding books in libraries to ranking products, jobs, job applicants, opinions and potential romantic partners, there is a substantial precedent that ranking systems have a responsibility not only to their users but also to the items being ranked. To address these often conflicting responsibilities, we propose a conceptual and computational framework that allows the formulation of fairness constraints on rankings in terms of exposure allocation. As part of this framework, we develop efficient algorithms for finding rankings that maximize the utility for the user while provably satisfying a specifiable notion of fairness. Since fairness goals can be application specific, we show how a broad range of fairness constraints can be implemented using our framework, including forms of demographic parity, disparate treatment, and disparate impact constraints. We illustrate the effect of these constraints by providing empirical results on two ranking problems."
                    },
                    {
                        "title": "Off-Policy Evaluation for Large Action Spaces via Embeddings",
                        "abstract": "Off-policy evaluation (OPE) in contextual bandits has seen rapid adoption in real-world systems, since it enables offline evaluation of new policies using only historic log data. Unfortunately, when the number of actions is large, existing OPE estimators -- most of which are based on inverse propensity score weighting -- degrade severely and can suffer from extreme bias and variance. This foils the use of OPE in many applications from recommender systems to language models. To overcome this issue, we propose a new OPE estimator that leverages marginalized importance weights when action embeddings provide structure in the action space. We characterize the bias, variance, and mean squared error of the proposed estimator and analyze the conditions under which the action embedding provides statistical benefits over conventional estimators. In addition to the theoretical analysis, we find that the empirical performance improvement can be substantial, enabling reliable OPE even when existing estimators collapse due to a large number of actions."
                    },
                    {
                        "title": "GPT as a Baseline for Recommendation Explanation Texts",
                        "abstract": "In this work, we establish a baseline potential for how modern model-generated text explanations of movie recommendations may help users, and explore what different components of these text explanations that users like or dislike, especially in contrast to existing human movie reviews. We found that participants gave no significantly different rankings between movies, nor did they give significantly different individual quality scores to reviews of movies that they had never seen before. However, participants did mark reviews as significantly better when they were movies they had seen before. We also explore specific aspects of movie review texts that participants marked as important for each quality. Overall, we establish that modern LLMs are a promising source of recommendation explanations, and we intend on further exploring personalizable text explanations in the future."
                    },
                    {
                        "title": "Methods for Ordinal Peer Grading",
                        "abstract": "MOOCs have the potential to revolutionize higher education with their wide outreach and accessibility, but they require instructors to come up with scalable alternates to traditional student evaluation. Peer grading -- having students assess each other -- is a promising approach to tackling the problem of evaluation at scale, since the number of \"graders\" naturally scales with the number of students. However, students are not trained in grading, which means that one cannot expect the same level of grading skills as in traditional settings. Drawing on broad evidence that ordinal feedback is easier to provide and more reliable than cardinal feedback, it is therefore desirable to allow peer graders to make ordinal statements (e.g. \"project X is better than project Y\") and not require them to make cardinal statements (e.g. \"project X is a B-\"). Thus, in this paper we study the problem of automatically inferring student grades from ordinal peer feedback, as opposed to existing methods that require cardinal peer feedback. We formulate the ordinal peer grading problem as a type of rank aggregation problem, and explore several probabilistic models under which to estimate student grades and grader reliability. We study the applicability of these methods using peer grading data collected from a real class -- with instructor and TA grades as a baseline -- and demonstrate the efficacy of ordinal feedback techniques in comparison to existing cardinal peer grading methods. Finally, we compare these peer-grading techniques to traditional evaluation techniques."
                    },
                    {
                        "title": "User Fairness, Item Fairness, and Diversity for Rankings in Two-Sided Markets",
                        "abstract": "Ranking items by their probability of relevance has long been the goal of conventional ranking systems. While this maximizes traditional criteria of ranking performance, there is a growing understanding that it is an oversimplification in online platforms that serve not only a diverse user population, but also the producers of the items. In particular, ranking algorithms are expected to be fair in how they serve all groups of users -- not just the majority group -- and they also need to be fair in how they divide exposure among the items. These fairness considerations can partially be met by adding diversity to the rankings, as done in several recent works. However, we show in this paper that user fairness, item fairness and diversity are fundamentally different concepts. In particular, we find that algorithms that consider only one of the three desiderata can fail to satisfy and even harm the other two. To overcome this shortcoming, we present the first ranking algorithm that explicitly enforces all three desiderata. The algorithm optimizes user and item fairness as a convex optimization problem which can be solved optimally. From its solution, a ranking policy can be derived via a novel Birkhoff-von Neumann decomposition algorithm that optimizes diversity. Beyond the theoretical analysis, we investigate empirically on a new benchmark dataset how effectively the proposed ranking algorithm can control user fairness, item fairness and diversity, as well as the trade-offs between them."
                    },
                    {
                        "title": "Uncertainty Quantification for Fairness in Two-Stage Recommender Systems",
                        "abstract": "Many large-scale recommender systems consist of two stages. The first stage efficiently screens the complete pool of items for a small subset of promising candidates, from which the second-stage model curates the final recommendations. In this paper, we investigate how to ensure group fairness to the items in this two-stage architecture. In particular, we find that existing first-stage recommenders might select an irrecoverably unfair set of candidates such that there is no hope for the second-stage recommender to deliver fair recommendations. To this end, motivated by recent advances in uncertainty quantification, we propose two threshold-policy selection rules that can provide distribution-free and finite-sample guarantees on fairness in first-stage recommenders. More concretely, given any relevance model of queries and items and a point-wise lower confidence bound on the expected number of relevant items for each threshold-policy, the two rules find near-optimal sets of candidates that contain enough relevant items in expectation from each group of items. To instantiate the rules, we demonstrate how to derive such confidence bounds from potentially partial and biased user feedback data, which are abundant in many large-scale recommender systems. In addition, we provide both finite-sample and asymptotic analyses of how close the two threshold selection rules are to the optimal thresholds. Beyond this theoretical analysis, we show empirically that these two rules can consistently select enough relevant items from each group while minimizing the size of the candidate sets for a wide range of settings."
                    },
                    {
                        "title": "Fairness in Ranking under Disparate Uncertainty",
                        "abstract": "Ranking is a ubiquitous method for focusing the attention of human evaluators on a manageable subset of options. Its use as part of human decision-making processes ranges from surfacing potentially relevant products on an e-commerce site to prioritizing college applications for human review. While ranking can make human evaluation more effective by focusing attention on the most promising options, we argue that it can introduce unfairness if the uncertainty of the underlying relevance model differs between groups of options. Unfortunately, such disparity in uncertainty appears widespread, often to the detriment of minority groups for which relevance estimates can have higher uncertainty due to a lack of data or appropriate features. To address this fairness issue, we propose Equal-Opportunity Ranking (EOR) as a new fairness criterion for ranking and show that it corresponds to a group-wise fair lottery among the relevant options even in the presence of disparate uncertainty. EOR optimizes for an even cost burden on all groups, unlike the conventional Probability Ranking Principle, and is fundamentally different from existing notions of fairness in rankings, such as demographic parity and proportional Rooney rule constraints that are motivated by proportional representation relative to group size. To make EOR ranking practical, we present an efficient algorithm for computing it in time $O(n \\log(n))$ and prove its close approximation guarantee to the globally optimal solution. In a comprehensive empirical evaluation on synthetic data, a US Census dataset, and a real-world audit of Amazon search queries, we find that the algorithm reliably guarantees EOR fairness while providing effective rankings."
                    },
                    {
                        "title": "Fair Ranking as Fair Division: Impact-Based Individual Fairness in Ranking",
                        "abstract": "Rankings have become the primary interface in two-sided online markets. Many have noted that the rankings not only affect the satisfaction of the users (e.g., customers, listeners, employers, travelers), but that the position in the ranking allocates exposure -- and thus economic opportunity -- to the ranked items (e.g., articles, products, songs, job seekers, restaurants, hotels). This has raised questions of fairness to the items, and most existing works have addressed fairness by explicitly linking item exposure to item relevance. However, we argue that any particular choice of such a link function may be difficult to defend, and we show that the resulting rankings can still be unfair. To avoid these shortcomings, we develop a new axiomatic approach that is rooted in principles of fair division. This not only avoids the need to choose a link function, but also more meaningfully quantifies the impact on the items beyond exposure. Our axioms of envy-freeness and dominance over uniform ranking postulate that for a fair ranking policy every item should prefer their own rank allocation over that of any other item, and that no item should be actively disadvantaged by the rankings. To compute ranking policies that are fair according to these axioms, we propose a new ranking objective related to the Nash Social Welfare. We show that the solution has guarantees regarding its envy-freeness, its dominance over uniform rankings for every item, and its Pareto optimality. In contrast, we show that conventional exposure-based fairness can produce large amounts of envy and have a highly disparate impact on the items. Beyond these theoretical results, we illustrate empirically how our framework controls the trade-off between impact-based individual item fairness and user utility."
                    },
                    {
                        "title": "Recommending Related Papers Based on Digital Library Access Records",
                        "abstract": "An important goal for digital libraries is to enable researchers to more easily explore related work. While citation data is often used as an indicator of relatedness, in this paper we demonstrate that digital access records (e.g. http-server logs) can be used as indicators as well. In particular, we show that measures based on co-access provide better coverage than co-citation, that they are available much sooner, and that they are more accurate for recent papers."
                    },
                    {
                        "title": "Large-Margin Learning of Submodular Summarization Methods",
                        "abstract": "In this paper, we present a supervised learning approach to training submodular scoring functions for extractive multi-document summarization. By taking a structured predicition approach, we provide a large-margin method that directly optimizes a convex relaxation of the desired performance measure. The learning method applies to all submodular summarization methods, and we demonstrate its effectiveness for both pairwise as well as coverage-based scoring functions on multiple datasets. Compared to state-of-the-art functions that were tuned manually, our method significantly improves performance and enables high-fidelity models with numbers of parameters well beyond what could reasonbly be tuned by hand."
                    }
                ]
            },
            "cc5d2d19-dcb1-4b6b-b4e6-17ac0d0f7014": {
                "pk": "cc5d2d19-dcb1-4b6b-b4e6-17ac0d0f7014",
                "name": "J. Andrew Bagnell",
                "collaborators": [
                    "Wen Sun",
                    "Geoffrey J. Gordon",
                    "Byron Boots",
                    "Kevin Waugh",
                    "Stephane Ross",
                    "Anirudh Vemula",
                    "Alexander Grubb",
                    "Martial Hebert",
                    "Gokul Swamy",
                    "Sanjiban Choudhury"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Imitation Learning",
                    "Game Theory",
                    "Structured Prediction"
                ],
                "publications": [
                    {
                        "title": "Deeply AggreVaTeD: Differentiable Imitation Learning for Sequential Prediction",
                        "abstract": "Researchers have demonstrated state-of-the-art performance in sequential decision making problems (e.g., robotics control, sequential prediction) with deep neural network models. One often has access to near-optimal oracles that achieve good performance on the task during training. We demonstrate that AggreVaTeD --- a policy gradient extension of the Imitation Learning (IL) approach of (Ross & Bagnell, 2014) --- can leverage such an oracle to achieve faster and better solutions with less training data than a less-informed Reinforcement Learning (RL) technique. Using both feedforward and recurrent neural network predictors, we present stochastic gradient procedures on a sequential prediction task, dependency-parsing from raw image data, as well as on various high dimensional robotics control problems. We also provide a comprehensive theoretical study of IL that demonstrates we can expect up to exponentially lower sample complexity for learning with AggreVaTeD than with RL algorithms, which backs our empirical findings. Our results and theory indicate that the proposed approach can achieve superior performance with respect to the oracle when the demonstrator is sub-optimal."
                    },
                    {
                        "title": "A Unified View of Large-scale Zero-sum Equilibrium Computation",
                        "abstract": "The task of computing approximate Nash equilibria in large zero-sum extensive-form games has received a tremendous amount of attention due mainly to the Annual Computer Poker Competition. Immediately after its inception, two competing and seemingly different approaches emerged---one an application of no-regret online learning, the other a sophisticated gradient method applied to a convex-concave saddle-point formulation. Since then, both approaches have grown in relative isolation with advancements on one side not effecting the other. In this paper, we rectify this by dissecting and, in a sense, unify the two views."
                    },
                    {
                        "title": "SpeedMachines: Anytime Structured Prediction",
                        "abstract": "Structured prediction plays a central role in machine learning applications from computational biology to computer vision. These models require significantly more computation than unstructured models, and, in many applications, algorithms may need to make predictions within a computational budget or in an anytime fashion. In this work we propose an anytime technique for learning structured prediction that, at training time, incorporates both structural elements and feature computation trade-offs that affect test-time inference. We apply our technique to the challenging problem of scene understanding in computer vision and demonstrate efficient and anytime predictions that gradually improve towards state-of-the-art classification performance as the allotted time increases."
                    },
                    {
                        "title": "A Critique of Strictly Batch Imitation Learning",
                        "abstract": "Recent work by Jarrett et al. attempts to frame the problem of offline imitation learning (IL) as one of learning a joint energy-based model, with the hope of out-performing standard behavioral cloning. We suggest that notational issues obscure how the psuedo-state visitation distribution the authors propose to optimize might be disconnected from the policy's $\\textit{true}$ state visitation distribution. We further construct natural examples where the parameter coupling advocated by Jarrett et al. leads to inconsistent estimates of the expert's policy, unlike behavioral cloning."
                    },
                    {
                        "title": "Game-Theoretic Algorithms for Conditional Moment Matching",
                        "abstract": "A variety of problems in econometrics and machine learning, including instrumental variable regression and Bellman residual minimization, can be formulated as satisfying a set of conditional moment restrictions (CMR). We derive a general, game-theoretic strategy for satisfying CMR that scales to nonlinear problems, is amenable to gradient-based optimization, and is able to account for finite sample uncertainty. We recover the approaches of Dikkala et al. and Dai et al. as special cases of our general framework before detailing various extensions and how to efficiently solve the game defined by CMR."
                    },
                    {
                        "title": "Agnostic System Identification for Model-Based Reinforcement Learning",
                        "abstract": "A fundamental problem in control is to learn a model of a system from observations that is useful for controller synthesis. To provide good performance guarantees, existing methods must assume that the real system is in the class of models considered during learning. We present an iterative method with strong guarantees even in the agnostic case where the system is not in the class. In particular, we show that any no-regret online learning algorithm can be used to obtain a near-optimal policy, provided some model achieves low training error and access to a good exploration distribution. Our approach applies to both discrete and continuous domains. We demonstrate its efficacy and scalability on a challenging helicopter domain from the literature."
                    },
                    {
                        "title": "Reinforcement and Imitation Learning via Interactive No-Regret Learning",
                        "abstract": "Recent work has demonstrated that problems-- particularly imitation learning and structured prediction-- where a learner's predictions influence the input-distribution it is tested on can be naturally addressed by an interactive approach and analyzed using no-regret online learning. These approaches to imitation learning, however, neither require nor benefit from information about the cost of actions. We extend existing results in two directions: first, we develop an interactive imitation learning approach that leverages cost information; second, we extend the technique to address reinforcement learning. The results provide theoretical support to the commonly observed successes of online approximate policy iteration. Our approach suggests a broad new family of algorithms and provides a unifying view of existing techniques for imitation and reinforcement learning."
                    },
                    {
                        "title": "TRON: A Fast Solver for Trajectory Optimization with Non-Smooth Cost Functions",
                        "abstract": "Trajectory optimization is an important tool for control and planning of complex, underactuated robots, and has shown impressive results in real world robotic tasks. However, in applications where the cost function to be optimized is non-smooth, modern trajectory optimization methods have extremely slow convergence. In this work, we present TRON, an iterative solver that can be used for efficient trajectory optimization in applications with non-smooth cost functions that are composed of smooth components. TRON achieves this by exploiting the structure of the objective to adaptively smooth the cost function, resulting in a sequence of objectives that can be efficiently optimized. TRON is provably guaranteed to converge to the global optimum of the non-smooth convex cost function when the dynamics are linear, and to a stationary point when the dynamics are nonlinear. Empirically, we show that TRON has faster convergence and lower final costs when compared to other trajectory optimization methods on a range of simulated tasks including collision-free motion planning for a mobile robot, sparse optimal control for surgical needle, and a satellite rendezvous problem."
                    },
                    {
                        "title": "A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning",
                        "abstract": "Sequential prediction problems such as imitation learning, where future observations depend on previous predictions (actions), violate the common i.i.d. assumptions made in statistical learning. This leads to poor performance in theory and often in practice. Some recent approaches provide stronger guarantees in this setting, but remain somewhat unsatisfactory as they train either non-stationary or stochastic policies and require a large number of iterations. In this paper, we propose a new iterative algorithm, which trains a stationary deterministic policy, that can be seen as a no regret algorithm in an online learning setting. We show that any such no regret algorithm, combined with additional reduction assumptions, must find a policy with good performance under the distribution of observations it induces in such sequential settings. We demonstrate that this new approach outperforms previous approaches on two challenging imitation learning problems and a benchmark sequence labeling problem."
                    },
                    {
                        "title": "Dual Policy Iteration",
                        "abstract": "Recently, a novel class of Approximate Policy Iteration (API) algorithms have demonstrated impressive practical performance (e.g., ExIt from [2], AlphaGo-Zero from [27]). This new family of algorithms maintains, and alternately optimizes, two policies: a fast, reactive policy (e.g., a deep neural network) deployed at test time, and a slow, non-reactive policy (e.g., Tree Search), that can plan multiple steps ahead. The reactive policy is updated under supervision from the non-reactive policy, while the non-reactive policy is improved with guidance from the reactive policy. In this work we study this Dual Policy Iteration (DPI) strategy in an alternating optimization framework and provide a convergence analysis that extends existing API theory. We also develop a special instance of this framework which reduces the update of non-reactive policies to model-based optimal control using learned local models, and provides a theoretically sound way of unifying model-free and model-based RL approaches with unknown dynamics. We demonstrate the efficacy of our approach on various continuous control Markov Decision Processes."
                    },
                    {
                        "title": "Generalized Boosting Algorithms for Convex Optimization",
                        "abstract": "Boosting is a popular way to derive powerful learners from simpler hypothesis classes. Following previous work (Mason et al., 1999; Friedman, 2000) on general boosting frameworks, we analyze gradient-based descent algorithms for boosting with respect to any convex objective and introduce a new measure of weak learner performance into this setting which generalizes existing work. We present the weak to strong learning guarantees for the existing gradient boosting work for strongly-smooth, strongly-convex objectives under this new measure of performance, and also demonstrate that this work fails for non-smooth objectives. To address this issue, we present new algorithms which extend this boosting approach to arbitrary convex loss functions and give corresponding weak to strong convergence results. In addition, we demonstrate experimental results that support our analysis and demonstrate the need for the new algorithms we present."
                    },
                    {
                        "title": "Truncated Horizon Policy Search: Combining Reinforcement Learning & Imitation Learning",
                        "abstract": "In this paper, we propose to combine imitation and reinforcement learning via the idea of reward shaping using an oracle. We study the effectiveness of the near-optimal cost-to-go oracle on the planning horizon and demonstrate that the cost-to-go oracle shortens the learner's planning horizon as function of its accuracy: a globally optimal oracle can shorten the planning horizon to one, leading to a one-step greedy Markov Decision Process which is much easier to optimize, while an oracle that is far away from the optimality requires planning over a longer horizon to achieve near-optimal performance. Hence our new insight bridges the gap and interpolates between imitation learning and reinforcement learning. Motivated by the above mentioned insights, we propose Truncated HORizon Policy Search (THOR), a method that focuses on searching for policies that maximize the total reshaped reward over a finite planning horizon when the oracle is sub-optimal. We experimentally demonstrate that a gradient-based implementation of THOR can achieve superior performance compared to RL baselines and IL baselines even when the oracle is sub-optimal."
                    },
                    {
                        "title": "Contrasting Exploration in Parameter and Action Space: A Zeroth-Order Optimization Perspective",
                        "abstract": "Black-box optimizers that explore in parameter space have often been shown to outperform more sophisticated action space exploration methods developed specifically for the reinforcement learning problem. We examine these black-box methods closely to identify situations in which they are worse than action space exploration methods and those in which they are superior. Through simple theoretical analyses, we prove that complexity of exploration in parameter space depends on the dimensionality of parameter space, while complexity of exploration in action space depends on both the dimensionality of action space and horizon length. This is also demonstrated empirically by comparing simple exploration methods on several model problems, including Contextual Bandit, Linear Regression and Reinforcement Learning in continuous control."
                    },
                    {
                        "title": "Exploration in Action Space",
                        "abstract": "Parameter space exploration methods with black-box optimization have recently been shown to outperform state-of-the-art approaches in continuous control reinforcement learning domains. In this paper, we examine reasons why these methods work better and the situations in which they are worse than traditional action space exploration methods. Through a simple theoretical analysis, we show that when the parametric complexity required to solve the reinforcement learning problem is greater than the product of action space dimensionality and horizon length, exploration in action space is preferred. This is also shown empirically by comparing simple exploration methods on several toy problems."
                    },
                    {
                        "title": "Learning Transferable Policies for Monocular Reactive MAV Control",
                        "abstract": "The ability to transfer knowledge gained in previous tasks into new contexts is one of the most important mechanisms of human learning. Despite this, adapting autonomous behavior to be reused in partially similar settings is still an open problem in current robotics research. In this paper, we take a small step in this direction and propose a generic framework for learning transferable motion policies. Our goal is to solve a learning problem in a target domain by utilizing the training data in a different but related source domain. We present this in the context of an autonomous MAV flight using monocular reactive control, and demonstrate the efficacy of our proposed approach through extensive real-world flight experiments in outdoor cluttered environments."
                    },
                    {
                        "title": "Hybrid Reinforcement Learning from Offline Observation Alone",
                        "abstract": "We consider the hybrid reinforcement learning setting where the agent has access to both offline data and online interactive access. While Reinforcement Learning (RL) research typically assumes offline data contains complete action, reward and transition information, datasets with only state information (also known as observation-only datasets) are more general, abundant and practical. This motivates our study of the hybrid RL with observation-only offline dataset framework. While the task of competing with the best policy \"covered\" by the offline data can be solved if a reset model of the environment is provided (i.e., one that can be reset to any state), we show evidence of hardness when only given the weaker trace model (i.e., one can only reset to the initial states and must produce full traces through the environment), without further assumption of admissibility of the offline data. Under the admissibility assumptions -- that the offline data could actually be produced by the policy class we consider -- we propose the first algorithm in the trace model setting that provably matches the performance of algorithms that leverage a reset model. We also perform proof-of-concept experiments that suggest the effectiveness of our algorithm in practice."
                    },
                    {
                        "title": "Computational Rationalization: The Inverse Equilibrium Problem",
                        "abstract": "Modeling the purposeful behavior of imperfect agents from a small number of observations is a challenging task. When restricted to the single-agent decision-theoretic setting, inverse optimal control techniques assume that observed behavior is an approximately optimal solution to an unknown decision problem. These techniques learn a utility function that explains the example behavior and can then be used to accurately predict or imitate future behavior in similar observed or unobserved situations.   In this work, we consider similar tasks in competitive and cooperative multi-agent domains. Here, unlike single-agent settings, a player cannot myopically maximize its reward --- it must speculate on how the other agents may act to influence the game's outcome. Employing the game-theoretic notion of regret and the principle of maximum entropy, we introduce a technique for predicting and generalizing behavior, as well as recovering a reward function in these domains."
                    },
                    {
                        "title": "Computational Rationalization: The Inverse Equilibrium Problem",
                        "abstract": "Modeling the purposeful behavior of imperfect agents from a small number of observations is a challenging task. When restricted to the single-agent decision-theoretic setting, inverse optimal control techniques assume that observed behavior is an approximately optimal solution to an unknown decision problem. These techniques learn a utility function that explains the example behavior and can then be used to accurately predict or imitate future behavior in similar observed or unobserved situations.   In this work, we consider similar tasks in competitive and cooperative multi-agent domains. Here, unlike single-agent settings, a player cannot myopically maximize its reward; it must speculate on how the other agents may act to influence the game's outcome. Employing the game-theoretic notion of regret and the principle of maximum entropy, we introduce a technique for predicting and generalizing behavior."
                    }
                ]
            },
            "6d18f8c6-0371-45e6-89a0-9fd5221cd296": {
                "pk": "6d18f8c6-0371-45e6-89a0-9fd5221cd296",
                "name": "Jason D. Lee",
                "collaborators": [],
                "domain": [
                    "Statistics",
                    "Selective Inference",
                    "Graphical Models",
                    "High-Dimensional Learning"
                ],
                "publications": [
                    {
                        "title": "Selective Inference and Learning Mixed Graphical Models",
                        "abstract": "This thesis studies two problems in modern statistics. First, we study selective inference, or inference for hypothesis that are chosen after looking at the data. The motiving application is inference for regression coefficients selected by the lasso. We present the Condition-on-Selection method that allows for valid selective inference, and study its application to the lasso, and several other selection algorithms.   In the second part, we consider the problem of learning the structure of a pairwise graphical model over continuous and discrete variables. We present a new pairwise model for graphical models with both continuous and discrete variables that is amenable to structure learning. In previous work, authors have considered structure learning of Gaussian graphical models and structure learning of discrete models. Our approach is a natural generalization of these two lines of work to the mixed case. The penalization scheme involves a novel symmetric use of the group-lasso norm and follows naturally from a particular parametrization of the model. We provide conditions under which our estimator is model selection consistent in the high-dimensional regime."
                    }
                ]
            },
            "4b080911-e6dc-4e82-8e24-49e4f7f516a7": {
                "pk": "4b080911-e6dc-4e82-8e24-49e4f7f516a7",
                "name": "Wen Sun",
                "collaborators": [],
                "domain": [
                    "Stochastic Processes",
                    "Large Deviations",
                    "Coagulation-Fragmentation",
                    "Random Graphs"
                ],
                "publications": [
                    {
                        "title": "On the Condensation and fluctuations in reversible coagulation-fragmentation models",
                        "abstract": "We study the condensation phenomenon for the invariant measures of the mean-field model of reversible coagulation-fragmentation processes conditioned to a supercritical density of particles. It is shown that when the parameters of the associated balance equation satisfy a subexponential tail condition, there is one single giant particle that corresponds to the missing mass in the macroscopic limit. We also show that in this case, the rest of the particles are asymptotically \\emph{i.i.d.} according to the normalized equilibrium state of the limit hydrodynamic differential equation. Conditions for the normal fluctuations and the $\\alpha$-stable fluctuations around the condensed mass are given. We obtain the large deviation principle for the empirical measure of the masses of the particles at equilibrium as well."
                    },
                    {
                        "title": "A conditional compound Poisson process approach to the sparse Erd\u0151s-R\u00e9nyi random graphs: moderate deviations",
                        "abstract": "We construct a compound Poisson process conditioned on its random summation that represents the sizes of the connected components in the sparse Erd\\H{o}s-R\\'enyi random graph $G(n,c/n)$. This new representation depicts a connection between the phase transition in the sparse random graph and the condensation transition in the zero-range model. Under this framework, we can derive moderate deviation principles for the maximun component, total number of connected components and empirical measure of the sizes in the non-critical regimes. Large deviation results are discussed."
                    },
                    {
                        "title": "A Functional Central Limit Theorem for the Becker-D\u00f6ring model",
                        "abstract": "We investigate the fluctuations of the stochastic Becker-D\\\"oring model of polymerization when the initial size of the system converges to infinity. A functional central limit problem is proved for the vector of the number of polymers of a given size. It is shown that the stochastic process associated to fluctuations is converging to the strong solution of an infinite dimensional stochastic differential equation (SDE) in a Hilbert space.   We also prove that, at equilibrium, the solution of this SDE is a Gaussian process. The proofs are based on a specific representation of the evolution equations, the introduction of a convenient Hilbert space and several technical estimates to control the fluctuations, especially of the first coordinate which interacts with all components of the infinite dimensional vector representing the state of the process."
                    },
                    {
                        "title": "Pathwise Large deviations for the pure jump $k$-nary interacting particle systems",
                        "abstract": "A pathwise large deviation result is proved for the pure jump models of $k$-nary interacting particle system introduced by Kolokoltsov that generalize classical Boltzmann's collision model, Smoluchovski's coagulation model and many others. The upper bound is obtained by following the standard methods of using a process \"perturbed\" by a regular function. To show the lower bound, we propose a family of orthogonal martingale measures and prove a coupling for the general perturbations. The rate function is studied based on the idea of L\\'eonard with a simplification by considering the conjugation of integral functionals on a subspace of $L^{\\infty}$. General \"gelling\" solutions in the domain of the rate function are also discussed."
                    }
                ]
            }
        }
    },
    "2307.16230": {
        "paper_data": {
            "title": "An Unforgeable Publicly Verifiable Watermark for Large Language Models",
            "url": "http://arxiv.org/abs/2307.16230v7",
            "arxiv_id": "2307.16230",
            "authors": [
                "Aiwei Liu",
                "Leyi Pan",
                "Xuming Hu",
                "Shu'ang Li",
                "Lijie Wen",
                "Irwin King",
                "Philip S. Yu"
            ],
            "abstract": "Recently, text watermarking algorithms for large language models (LLMs) have been proposed to mitigate the potential harms of text generated by LLMs, including fake news and copyright issues. However, current watermark detection algorithms require the secret key used in the watermark generation process, making them susceptible to security breaches and counterfeiting during public detection. To address this limitation, we propose an unforgeable publicly verifiable watermark algorithm named UPV that uses two different neural networks for watermark generation and detection, instead of using the same key at both stages. Meanwhile, the token embedding parameters are shared between the generation and detection networks, which makes the detection network achieve a high accuracy very efficiently. Experiments demonstrate that our algorithm attains high detection accuracy and computational efficiency through neural networks. Subsequent analysis confirms the high complexity involved in forging the watermark from the detection network. Our code is available at \\href{https://github.com/THU-BPM/unforgeable_watermark}{https://github.com/THU-BPM/unforgeable\\_watermark}. Additionally, our algorithm could also be accessed through MarkLLM \\citep{pan2024markllm} \\footnote{https://github.com/THU-BPM/MarkLLM}.",
            "introduction": "   1 Introduction  With the development of current large language models (LLMs), many LLMs, like GPT-4 (OpenAI, 2023) and Claude, could rapidly generate human-like texts. This has led to numerous risks, such as the generation of a vast amount of false information on the Internet (Pan et\u00a0al., 2023), and the infringement of copyrights of creative works (Chen et\u00a0al., 2023). Therefore, texts generated by LLMs need to be detected and tagged.   At present, some watermarking algorithms for LLM have proved successful in making machine-generated texts detectable by adding implicit features during the text generation process that are difficult for humans to discover but easily detected by the specially designed method (Christ et\u00a0al., 2023; Kirchenbauer et\u00a0al., 2023). The current watermark algorithms for large models utilize a shared key during the generation and detection of watermarks. They work well when the detection access is restricted to the watermark owner only. However, in many situations, when third-party watermark detection is required, the exposure of the shared key would enable others to forge the watermark. Therefore, preventing the watermark forge in the public detection setting, is of great importance.   In this work, we propose UPV, the first unforgeable publicly verifiable watermarking algorithm for LLMs. Our approach adopts the commonly used watermark schema, which embeds small watermark signals into LLM\u2019s logits during generation Kirchenbauer et\u00a0al. (2023); Zhao et\u00a0al. (2023). The main difference is that our algorithms could detect the watermark without the generation key. To conceal the details of the watermark in the detection process, we propose using two separate neural networks for watermark generation and detection, rather than relying on the same secret key for both stages. The unforgeability of our algorithm stems from a computational asymmetry: constructing a watermark generation network from a watermark detection network is notably more complex than the reverse process (Section 4.5). To enhance the accuracy of the detection network while alleviating the complexity of the training process, we have shared the token embedding parameters between the detector and the generator. This provides some prior information to the detector while maintaining unforgeable. Specifically, our watermark generator takes a window of w\ud835\udc64witalic_w tokens as input and predicts the watermark signals of the last token. The text detector directly takes all tokens from the text as input and outputs whether the complete text contains the watermark.   Figure 1: The left segment outlines the token logits generation process of the watermarked language model. Initially, the origin model generates token logits; these are then refined by the watermark generator to increase the probability of watermarked tokens (denoted in green). Operating within a predefined token window, this generator network (center) determines the final token\u2019s watermark label. The watermark detector (right) evaluates the entire text to ascertain watermark presence.   In our experiments, the watermark detection algorithm exhibits a detection performance of nearly 99% F1 score, almost identical to the key-based watermark algorithm, which is our theoretical upper bound. Notably, the computational burden of our watermark generation and detection network is minimal, especially when compared to large language models, due to its significantly lower parameter count. Furthermore, we have demonstrated the difficulty of cracking the watermarking",
            "references": [
                {
                    "title": "A Survey of Text Watermarking in the Era of Large Language Models",
                    "abstract": "Text watermarking algorithms are crucial for protecting the copyright of textual content. Historically, their capabilities and application scenarios were limited. However, recent advancements in large language models (LLMs) have revolutionized these techniques. LLMs not only enhance text watermarking algorithms with their advanced abilities but also create a need for employing these algorithms to protect their own copyrights or prevent potential misuse. This paper conducts a comprehensive survey of the current state of text watermarking technology, covering four main aspects: (1) an overview and comparison of different text watermarking techniques; (2) evaluation methods for text watermarking algorithms, including their detectability, impact on text or LLM quality, robustness under target or untargeted attacks; (3) potential application scenarios for text watermarking technology; (4) current challenges and future directions for text watermarking. This survey aims to provide researchers with a thorough understanding of text watermarking technology in the era of LLM, thereby promoting its further advancement."
                },
                {
                    "title": "AI Alignment: A Comprehensive Survey",
                    "abstract": "AI alignment aims to make AI systems behave in line with human intentions and values. As AI systems grow more capable, so do risks from misalignment. To provide a comprehensive and up-to-date overview of the alignment field, in this survey, we delve into the core concepts, methodology, and practice of alignment. First, we identify four principles as the key objectives of AI alignment: Robustness, Interpretability, Controllability, and Ethicality (RICE). Guided by these four principles, we outline the landscape of current alignment research and decompose them into two key components: forward alignment and backward alignment. The former aims to make AI systems aligned via alignment training, while the latter aims to gain evidence about the systems' alignment and govern them appropriately to avoid exacerbating misalignment risks. On forward alignment, we discuss techniques for learning from feedback and learning under distribution shift. On backward alignment, we discuss assurance techniques and governance practices. We also release and continually update the website (www.alignmentsurvey.com) which features tutorials, collections of papers, blog posts, and other resources."
                },
                {
                    "title": "A Semantic Invariant Robust Watermark for Large Language Models",
                    "abstract": "Watermark algorithms for large language models (LLMs) have achieved extremely high accuracy in detecting text generated by LLMs. Such algorithms typically involve adding extra watermark logits to the LLM's logits at each generation step. However, prior algorithms face a trade-off between attack robustness and security robustness. This is because the watermark logits for a token are determined by a certain number of preceding tokens; a small number leads to low security robustness, while a large number results in insufficient attack robustness. In this work, we propose a semantic invariant watermarking method for LLMs that provides both attack robustness and security robustness. The watermark logits in our work are determined by the semantics of all preceding tokens. Specifically, we utilize another embedding LLM to generate semantic embeddings for all preceding tokens, and then these semantic embeddings are transformed into the watermark logits through our trained watermark model. Subsequent analyses and experiments demonstrated the attack robustness of our method in semantically invariant settings: synonym substitution and text paraphrasing settings. Finally, we also show that our watermark possesses adequate security robustness. Our code and data are available at \\href{https://github.com/THU-BPM/Robust_Watermark}{https://github.com/THU-BPM/Robust\\_Watermark}. Additionally, our algorithm could also be accessed through MarkLLM \\citep{pan2024markllm} \\footnote{https://github.com/THU-BPM/MarkLLM}."
                },
                {
                    "title": "Do Large Language Models Know about Facts?",
                    "abstract": "Large language models (LLMs) have recently driven striking performance improvements across a range of natural language processing tasks. The factual knowledge acquired during pretraining and instruction tuning can be useful in various downstream tasks, such as question answering, and language generation. Unlike conventional Knowledge Bases (KBs) that explicitly store factual knowledge, LLMs implicitly store facts in their parameters. Content generated by the LLMs can often exhibit inaccuracies or deviations from the truth, due to facts that can be incorrectly induced or become obsolete over time. To this end, we aim to comprehensively evaluate the extent and scope of factual knowledge within LLMs by designing the benchmark Pinocchio. Pinocchio contains 20K diverse factual questions that span different sources, timelines, domains, regions, and languages. Furthermore, we investigate whether LLMs are able to compose multiple facts, update factual knowledge temporally, reason over multiple pieces of facts, identify subtle factual differences, and resist adversarial examples. Extensive experiments on different sizes and types of LLMs show that existing LLMs still lack factual knowledge and suffer from various spurious correlations. We believe this is a critical bottleneck for realizing trustworthy artificial intelligence. The dataset Pinocchio and our codes will be publicly available."
                },
                {
                    "title": "Robust Distortion-free Watermarks for Language Models",
                    "abstract": "We propose a methodology for planting watermarks in text from an autoregressive language model that are robust to perturbations without changing the distribution over text up to a certain maximum generation budget. We generate watermarked text by mapping a sequence of random numbers -- which we compute using a randomized watermark key -- to a sample from the language model. To detect watermarked text, any party who knows the key can align the text to the random number sequence. We instantiate our watermark methodology with two sampling schemes: inverse transform sampling and exponential minimum sampling. We apply these watermarks to three language models -- OPT-1.3B, LLaMA-7B and Alpaca-7B -- to experimentally validate their statistical power and robustness to various paraphrasing attacks. Notably, for both the OPT-1.3B and LLaMA-7B models, we find we can reliably detect watermarked text ($p \\leq 0.01$) from $35$ tokens even after corrupting between $40$-$50\\%$ of the tokens via random edits (i.e., substitutions, insertions or deletions). For the Alpaca-7B model, we conduct a case study on the feasibility of watermarking responses to typical user instructions. Due to the lower entropy of the responses, detection is more difficult: around $25\\%$ of the responses -- whose median length is around $100$ tokens -- are detectable with $p \\leq 0.01$, and the watermark is also less robust to certain automated paraphrasing attacks we implement."
                },
                {
                    "title": "RADAR: Robust AI-Text Detection via Adversarial Learning",
                    "abstract": "Recent advances in large language models (LLMs) and the intensifying popularity of ChatGPT-like applications have blurred the boundary of high-quality text generation between humans and machines. However, in addition to the anticipated revolutionary changes to our technology and society, the difficulty of distinguishing LLM-generated texts (AI-text) from human-generated texts poses new challenges of misuse and fairness, such as fake content generation, plagiarism, and false accusations of innocent writers. While existing works show that current AI-text detectors are not robust to LLM-based paraphrasing, this paper aims to bridge this gap by proposing a new framework called RADAR, which jointly trains a robust AI-text detector via adversarial learning. RADAR is based on adversarial training of a paraphraser and a detector. The paraphraser's goal is to generate realistic content to evade AI-text detection. RADAR uses the feedback from the detector to update the paraphraser, and vice versa. Evaluated with 8 different LLMs (Pythia, Dolly 2.0, Palmyra, Camel, GPT-J, Dolly 1.0, LLaMA, and Vicuna) across 4 datasets, experimental results show that RADAR significantly outperforms existing AI-text detection methods, especially when paraphrasing is in place. We also identify the strong transferability of RADAR from instruction-tuned LLMs to other LLMs, and evaluate the improved capability of RADAR via GPT-3.5-Turbo."
                },
                {
                    "title": "Provable Robust Watermarking for AI-Generated Text",
                    "abstract": "We study the problem of watermarking large language models (LLMs) generated text -- one of the most promising approaches for addressing the safety challenges of LLM usage. In this paper, we propose a rigorous theoretical framework to quantify the effectiveness and robustness of LLM watermarks. We propose a robust and high-quality watermark method, Unigram-Watermark, by extending an existing approach with a simplified fixed grouping strategy. We prove that our watermark method enjoys guaranteed generation quality, correctness in watermark detection, and is robust against text editing and paraphrasing. Experiments on three varying LLMs and two datasets verify that our Unigram-Watermark achieves superior detection accuracy and comparable generation quality in perplexity, thus promoting the responsible use of LLMs. Code is available at https://github.com/XuandongZhao/Unigram-Watermark."
                },
                {
                    "title": "Undetectable Watermarks for Language Models",
                    "abstract": "Recent advances in the capabilities of large language models such as GPT-4 have spurred increasing concern about our ability to detect AI-generated text. Prior works have suggested methods of embedding watermarks in model outputs, by noticeably altering the output distribution. We ask: Is it possible to introduce a watermark without incurring any detectable change to the output distribution? To this end we introduce a cryptographically-inspired notion of undetectable watermarks for language models. That is, watermarks can be detected only with the knowledge of a secret key; without the secret key, it is computationally intractable to distinguish watermarked outputs from those of the original model. In particular, it is impossible for a user to observe any degradation in the quality of the text. Crucially, watermarks should remain undetectable even when the user is allowed to adaptively query the model with arbitrarily chosen prompts. We construct undetectable watermarks based on the existence of one-way functions, a standard assumption in cryptography."
                },
                {
                    "title": "Who Wrote this Code? Watermarking for Code Generation",
                    "abstract": "Since the remarkable generation performance of large language models raised ethical and legal concerns, approaches to detect machine-generated text by embedding watermarks are being developed. However, we discover that the existing works fail to function appropriately in code generation tasks due to the task's nature of having low entropy. Extending a logit-modifying watermark method, we propose Selective WatErmarking via Entropy Thresholding (SWEET), which enhances detection ability and mitigates code quality degeneration by removing low-entropy segments at generating and detecting watermarks. Our experiments show that SWEET significantly improves code quality preservation while outperforming all baselines, including post-hoc detection methods, in detecting machine-generated code text. Our code is available in https://github.com/hongcheki/sweet-watermark."
                },
                {
                    "title": "On the Risk of Misinformation Pollution with Large Language Models",
                    "abstract": "In this paper, we comprehensively investigate the potential misuse of modern Large Language Models (LLMs) for generating credible-sounding misinformation and its subsequent impact on information-intensive applications, particularly Open-Domain Question Answering (ODQA) systems. We establish a threat model and simulate potential misuse scenarios, both unintentional and intentional, to assess the extent to which LLMs can be utilized to produce misinformation. Our study reveals that LLMs can act as effective misinformation generators, leading to a significant degradation in the performance of ODQA systems. To mitigate the harm caused by LLM-generated misinformation, we explore three defense strategies: prompting, misinformation detection, and majority voting. While initial results show promising trends for these defensive strategies, much more work needs to be done to address the challenge of misinformation pollution. Our work highlights the need for further research and interdisciplinary collaboration to address LLM-generated misinformation and to promote responsible use of LLMs."
                },
                {
                    "title": "DetectLLM: Leveraging Log Rank Information for Zero-Shot Detection of Machine-Generated Text",
                    "abstract": "With the rapid progress of large language models (LLMs) and the huge amount of text they generated, it becomes more and more impractical to manually distinguish whether a text is machine-generated. Given the growing use of LLMs in social media and education, it prompts us to develop methods to detect machine-generated text, preventing malicious usage such as plagiarism, misinformation, and propaganda. Previous work has studied several zero-shot methods, which require no training data. These methods achieve good performance, but there is still a lot of room for improvement. In this paper, we introduce two novel zero-shot methods for detecting machine-generated text by leveraging the log rank information. One is called DetectLLM-LRR, which is fast and efficient, and the other is called DetectLLM-NPR, which is more accurate, but slower due to the need for perturbations. Our experiments on three datasets and seven language models show that our proposed methods improve over the state of the art by 3.9 and 1.75 AUROC points absolute. Moreover, DetectLLM-NPR needs fewer perturbations than previous work to achieve the same level of performance, which makes it more practical for real-world use. We also investigate the efficiency--performance trade-off based on users preference on these two measures and we provide intuition for using them in practice effectively. We release the data and the code of both methods in https://github.com/mbzuai-nlp/DetectLLM"
                },
                {
                    "title": "G3Detector: General GPT-Generated Text Detector",
                    "abstract": "The burgeoning progress in the field of Large Language Models (LLMs) heralds significant benefits due to their unparalleled capacities. However, it is critical to acknowledge the potential misuse of these models, which could give rise to a spectrum of social and ethical dilemmas. Despite numerous preceding efforts centered around distinguishing synthetic text, most existing detection systems fail to identify data synthesized by the latest LLMs, such as ChatGPT and GPT-4. In response to this challenge, we introduce an unpretentious yet potent detection approach proficient in identifying synthetic text across a wide array of fields. Moreover, our detector demonstrates outstanding performance uniformly across various model architectures and decoding strategies. It also possesses the capability to identify text generated utilizing a potent detection-evasion technique. Our comprehensive research underlines our commitment to boosting the robustness and efficiency of machine-generated text detection mechanisms, particularly in the context of swiftly progressing and increasingly adaptive AI technologies."
                },
                {
                    "title": "Smaller Language Models are Better Black-box Machine-Generated Text Detectors",
                    "abstract": "With the advent of fluent generative language models that can produce convincing utterances very similar to those written by humans, distinguishing whether a piece of text is machine-generated or human-written becomes more challenging and more important, as such models could be used to spread misinformation, fake news, fake reviews and to mimic certain authors and figures. To this end, there have been a slew of methods proposed to detect machine-generated text. Most of these methods need access to the logits of the target model or need the ability to sample from the target. One such black-box detection method relies on the observation that generated text is locally optimal under the likelihood function of the generator, while human-written text is not. We find that overall, smaller and partially-trained models are better universal text detectors: they can more precisely detect text generated from both small and larger models. Interestingly, we find that whether the detector and generator were trained on the same data is not critically important to the detection success. For instance the OPT-125M model has an AUC of 0.81 in detecting ChatGPT generations, whereas a larger model from the GPT family, GPTJ-6B, has AUC of 0.45."
                },
                {
                    "title": "Robust Multi-bit Natural Language Watermarking through Invariant Features",
                    "abstract": "Recent years have witnessed a proliferation of valuable original natural language contents found in subscription-based media outlets, web novel platforms, and outputs of large language models. However, these contents are susceptible to illegal piracy and potential misuse without proper security measures. This calls for a secure watermarking system to guarantee copyright protection through leakage tracing or ownership identification. To effectively combat piracy and protect copyrights, a multi-bit watermarking framework should be able to embed adequate bits of information and extract the watermarks in a robust manner despite possible corruption. In this work, we explore ways to advance both payload and robustness by following a well-known proposition from image watermarking and identify features in natural language that are invariant to minor corruption. Through a systematic analysis of the possible sources of errors, we further propose a corruption-resistant infill model. Our full method improves upon the previous work on robustness by +16.8% point on average on four datasets, three corruption types, and two corruption ratios"
                },
                {
                    "title": "On the Possibilities of AI-Generated Text Detection",
                    "abstract": "Our work addresses the critical issue of distinguishing text generated by Large Language Models (LLMs) from human-produced text, a task essential for numerous applications. Despite ongoing debate about the feasibility of such differentiation, we present evidence supporting its consistent achievability, except when human and machine text distributions are indistinguishable across their entire support. Drawing from information theory, we argue that as machine-generated text approximates human-like quality, the sample size needed for detection increases. We establish precise sample complexity bounds for detecting AI-generated text, laying groundwork for future research aimed at developing advanced, multi-sample detectors. Our empirical evaluations across multiple datasets (Xsum, Squad, IMDb, and Kaggle FakeNews) confirm the viability of enhanced detection methods. We test various state-of-the-art text generators, including GPT-2, GPT-3.5-Turbo, Llama, Llama-2-13B-Chat-HF, and Llama-2-70B-Chat-HF, against detectors, including oBERTa-Large/Base-Detector, GPTZero. Our findings align with OpenAI's empirical data related to sequence length, marking the first theoretical substantiation for these observations."
                },
                {
                    "title": "Can AI-Generated Text be Reliably Detected?",
                    "abstract": "The unregulated use of LLMs can potentially lead to malicious consequences such as plagiarism, generating fake news, spamming, etc. Therefore, reliable detection of AI-generated text can be critical to ensure the responsible use of LLMs. Recent works attempt to tackle this problem either using certain model signatures present in the generated text outputs or by applying watermarking techniques that imprint specific patterns onto them. In this paper, we show that these detectors are not reliable in practical scenarios. In particular, we develop a recursive paraphrasing attack to apply on AI text, which can break a whole range of detectors, including the ones using the watermarking schemes as well as neural network-based detectors, zero-shot classifiers, and retrieval-based detectors. Our experiments include passages around 300 tokens in length, showing the sensitivity of the detectors even in the case of relatively long passages. We also observe that our recursive paraphrasing only degrades text quality slightly, measured via human studies, and metrics such as perplexity scores and accuracy on text benchmarks. Additionally, we show that even LLMs protected by watermarking schemes can be vulnerable against spoofing attacks aimed to mislead detectors to classify human-written text as AI-generated, potentially causing reputational damages to the developers. In particular, we show that an adversary can infer hidden AI text signatures of the LLM outputs without having white-box access to the detection method. Finally, we provide a theoretical connection between the AUROC of the best possible detector and the Total Variation distance between human and AI text distributions that can be used to study the fundamental hardness of the reliable detection problem for advanced language models. Our code is publicly available at https://github.com/vinusankars/Reliability-of-AI-text-detectors."
                },
                {
                    "title": "A comprehensive evaluation of ChatGPT's zero-shot Text-to-SQL capability",
                    "abstract": "This paper presents the first comprehensive analysis of ChatGPT's Text-to-SQL ability. Given the recent emergence of large-scale conversational language model ChatGPT and its impressive capabilities in both conversational abilities and code generation, we sought to evaluate its Text-to-SQL performance. We conducted experiments on 12 benchmark datasets with different languages, settings, or scenarios, and the results demonstrate that ChatGPT has strong text-to-SQL abilities. Although there is still a gap from the current state-of-the-art (SOTA) model performance, considering that the experiment was conducted in a zero-shot scenario, ChatGPT's performance is still impressive. Notably, in the ADVETA (RPL) scenario, the zero-shot ChatGPT even outperforms the SOTA model that requires fine-tuning on the Spider dataset by 4.1\\%, demonstrating its potential for use in practical applications. To support further research in related fields, we have made the data generated by ChatGPT publicly available at https://github.com/THU-BPM/chatgpt-sql."
                },
                {
                    "title": "A Pathway Towards Responsible AI Generated Content",
                    "abstract": "AI Generated Content (AIGC) has received tremendous attention within the past few years, with content ranging from image, text, to audio, video, etc. Meanwhile, AIGC has become a double-edged sword and recently received much criticism regarding its responsible usage. In this article, we focus on three main concerns that may hinder the healthy development and deployment of AIGC in practice, including risks from privacy; bias, toxicity, misinformation; and intellectual property (IP). By documenting known and potential risks, as well as any possible misuse scenarios of AIGC, the aim is to sound the alarm of potential risks and misuse, help society to eliminate obstacles, and promote the more ethical and secure deployment of AIGC."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "A Watermark for Large Language Models",
                    "abstract": "Potential harms of large language models can be mitigated by watermarking model output, i.e., embedding signals into generated text that are invisible to humans but algorithmically detectable from a short span of tokens. We propose a watermarking framework for proprietary language models. The watermark can be embedded with negligible impact on text quality, and can be detected using an efficient open-source algorithm without access to the language model API or parameters. The watermark works by selecting a randomized set of\"green\"tokens before a word is generated, and then softly promoting use of green tokens during sampling. We propose a statistical test for detecting the watermark with interpretable p-values, and derive an information-theoretic framework for analyzing the sensitivity of the watermark. We test the watermark using a multi-billion parameter model from the Open Pretrained Transformer (OPT) family, and discuss robustness and security."
                },
                {
                    "title": "Character-level White-Box Adversarial Attacks against Transformers via Attachable Subwords Substitution",
                    "abstract": "We propose the first character-level white-box adversarial attack method against transformer models. The intuition of our method comes from the observation that words are split into subtokens before being fed into the transformer models and the substitution between two close subtokens has a similar effect with the character modification. Our method mainly contains three steps. First, a gradient-based method is adopted to find the most vulnerable words in the sentence. Then we split the selected words into subtokens to replace the origin tokenization result from the transformer tokenizer. Finally, we utilize an adversarial loss to guide the substitution of attachable subtokens in which the Gumbel-softmax trick is introduced to ensure gradient propagation.Meanwhile, we introduce the visual and length constraint in the optimization process to achieve minimum character modifications.Extensive experiments on both sentence-level and token-level tasks demonstrate that our method could outperform the previous attack methods in terms of success rate and edit distance. Furthermore, human evaluation verifies our adversarial examples could preserve their origin labels."
                },
                {
                    "title": "No Language Left Behind: Scaling Human-Centered Machine Translation",
                    "abstract": "Driven by the goal of eradicating language barriers on a global scale, machine translation has solidified itself as a key focus of artificial intelligence research today. However, such efforts have coalesced around a small subset of languages, leaving behind the vast majority of mostly low-resource languages. What does it take to break the 200 language barrier while ensuring safe, high quality results, all while keeping ethical considerations in mind? In No Language Left Behind, we took on this challenge by first contextualizing the need for low-resource language translation support through exploratory interviews with native speakers. Then, we created datasets and models aimed at narrowing the performance gap between low and high-resource languages. More specifically, we developed a conditional compute model based on Sparsely Gated Mixture of Experts that is trained on data obtained with novel and effective data mining techniques tailored for low-resource languages. We propose multiple architectural and training improvements to counteract overfitting while training on thousands of tasks. Critically, we evaluated the performance of over 40,000 different translation directions using a human-translated benchmark, Flores-200, and combined human evaluation with a novel toxicity benchmark covering all languages in Flores-200 to assess translation safety. Our model achieves an improvement of 44% BLEU relative to the previous state-of-the-art, laying important groundwork towards realizing a universal translation system. Finally, we open source all contributions described in this work, accessible at https://github.com/facebookresearch/fairseq/tree/nllb."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "Adversarial Watermarking Transformer: Towards Tracing Text Provenance with Data Hiding",
                    "abstract": "Recent advances in natural language generation have introduced powerful language models with high-quality output text. However, this raises concerns about the potential misuse of such models for malicious purposes. In this paper, we study natural language watermarking as a defense to help better mark and trace the provenance of text. We introduce the Adversarial Watermarking Transformer (AWT) with a jointly trained encoder-decoder and adversarial training that, given an input text and a binary message, generates an output text that is unobtrusively encoded with the given message. We further study different training and inference strategies to achieve minimal changes to the semantics and correctness of the input text.AWT is the first end-to-end model to hide data in text by automatically learning -without ground truth- word substitutions along with their locations in order to encode the message. We empirically show that our model is effective in largely preserving text utility and decoding the watermark while hiding its presence against adversaries. Additionally, we demonstrate that our method is robust against a range of attacks."
                },
                {
                    "title": "BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension",
                    "abstract": "We present BART, a denoising autoencoder for pretraining sequence-to-sequence models. BART is trained by (1) corrupting text with an arbitrary noising function, and (2) learning a model to reconstruct the original text. It uses a standard Tranformer-based neural machine translation architecture which, despite its simplicity, can be seen as generalizing BERT (due to the bidirectional encoder), GPT (with the left-to-right decoder), and other recent pretraining schemes. We evaluate a number of noising approaches, finding the best performance by both randomly shuffling the order of sentences and using a novel in-filling scheme, where spans of text are replaced with a single mask token. BART is particularly effective when fine tuned for text generation but also works well for comprehension tasks. It matches the performance of RoBERTa on GLUE and SQuAD, and achieves new state-of-the-art results on a range of abstractive dialogue, question answering, and summarization tasks, with gains of up to 3.5 ROUGE. BART also provides a 1.1 BLEU increase over a back-translation system for machine translation, with only target language pretraining. We also replicate other pretraining schemes within the BART framework, to understand their effect on end-task performance."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "RoBERTa: A Robustly Optimized BERT Pretraining Approach",
                    "abstract": "Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code."
                },
                {
                    "title": "Results of the WMT14 Metrics Shared Task",
                    "abstract": "This paper presents the results of the WMT14 Metrics Shared Task. We asked participants of this task to score the outputs of the MT systems involved in WMT14 Shared Translation Task. We collected scores of 23 metrics from 12 research groups. In addition to that we computed scores of 6 standard metrics (BLEU, NIST, WER, PER, TER and CDER) as baselines. The collected scores were evaluated in terms of system level correlation (how well each metric\u2019s scores correlate with WMT14 official manual ranking of systems) and in terms of segment level correlation (how often a metric agrees with humans in comparing two translations of a particular sentence)."
                },
                {
                    "title": "Long Short-Term Memory",
                    "abstract": "Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, LSTM can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units. Multiplicative gate units learn to open and close access to the constant error flow. LSTM is local in space and time; its computational complexity per time step and weight is O. 1. Our experiments with artificial data involve local, distributed, real-valued, and noisy pattern representations. In comparisons with real-time recurrent learning, back propagation through time, recurrent cascade correlation, Elman nets, and neural sequence chunking, LSTM leads to many more successful runs, and learns much faster. LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.CL",
                "68T50",
                "I.2.7"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop an unforgeable publicly verifiable watermarking algorithm for large language models (LLMs) that allows for the detection of machine-generated texts without exposing a shared key?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing concerns surrounding the misuse of LLMs, including the proliferation of false information and copyright infringement. By creating a robust watermarking system, we can enhance the integrity of generated content, thereby fostering trust in AI-generated texts. This advancement could lead to new standards in content verification, influencing future research in AI ethics, copyright law, and content authenticity. Additionally, practical applications could emerge in various sectors, such as journalism, academia, and digital media, where the authenticity of information is paramount.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in creating a watermarking system that is both effective and secure in a public detection setting. Naive approaches may fail because they often rely on shared keys, which can be compromised, allowing for watermark forgery. The technical complexities include ensuring that the watermark is imperceptible to human readers while remaining detectable by the algorithm. Theoretical obstacles involve the need for a computationally asymmetric design, where the generation of a watermark is significantly more complex than its detection, which complicates the development of a secure and efficient system.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on watermarking algorithms that depend on shared keys, which limits their applicability in public detection scenarios. The lack of a method that allows for secure detection without exposing the key has been a significant barrier. Additionally, existing solutions often do not address the computational asymmetry required for unforgeability. Our approach differs by utilizing two separate neural networks for watermark generation and detection, thereby enhancing security and efficiency while maintaining the integrity of the watermarking process.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of a watermark generator and detector using two distinct neural networks. The generator refines token logits to embed watermark signals, while the detector evaluates the entire text for watermark presence. We will use a dataset of machine-generated texts and evaluate our algorithm's performance using the F1 score as the primary metric. The expected outcome is a watermark detection performance nearing 99% F1 score,"
            }
        },
        "author_data": {
            "50b00d46-ff26-4bf2-9f32-60a3a0df3c61": {
                "pk": "50b00d46-ff26-4bf2-9f32-60a3a0df3c61",
                "name": "Aiwei Liu",
                "collaborators": [
                    "Xuming Hu",
                    "Lijie Wen",
                    "Philip S. Yu",
                    "Irwin King",
                    "Leyi Pan",
                    "Zitian Gao",
                    "Yijian Lu",
                    "Haoping Bai",
                    "Zhiyun Lu",
                    "Xiang Kong",
                    "Simon Wang",
                    "Jiulong Shan",
                    "Jingjing Li",
                    "Zhiwei He",
                    "Binglin Zhou",
                    "Yifei Zhang",
                    "Dianzhi Yu",
                    "Chenwei Zhang",
                    "Shiao Meng",
                    "Zhonghao Li",
                    "Kening Zheng",
                    "Sirui Huang",
                    "Hui Xiong",
                    "Yanchao Sun",
                    "Albin Madappally Jose",
                    "Xiaojiang Liu",
                    "Meng Cao",
                    "Zexuan Qiu",
                    "Zijing Ou",
                    "Bin Wu",
                    "Yutong Li",
                    "Lu Chen",
                    "Kai Yu",
                    "Guanyu Zhou",
                    "Yibo Yan",
                    "Xin Zou",
                    "Kun Wang",
                    "Xuandong Zhao",
                    "Shuliang Liu",
                    "Lingzhe Zhang",
                    "Tong Jia",
                    "Mengxi Jia",
                    "Yifan Wu",
                    "Yong Yang",
                    "Zhonghai Wu",
                    "Ying Li",
                    "Yichen Di",
                    "Sheng Guan",
                    "Yiming Liu",
                    "Liancheng Fang",
                    "Xinni Zhang",
                    "Yankai Chen",
                    "Zhaochen Hong",
                    "Hong-ping Hao",
                    "Xing Wang",
                    "Zhaopeng Tu",
                    "Zhuosheng Zhang",
                    "Rui Wang",
                    "Fukun Ma",
                    "Yawen Yang",
                    "Shuang Li",
                    "Mengsi Cao",
                    "Boye Niu",
                    "Xuzheng He",
                    "Haotian Xu",
                    "Hongzhang Liu",
                    "Zeqi Tan",
                    "Xin Zhang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Information Retrieval",
                    "Watermarking"
                ],
                "publications": [
                    {
                        "title": "Refiner: Restructure Retrieval Content Efficiently to Advance Question-Answering Capabilities",
                        "abstract": "Large Language Models (LLMs) are limited by their parametric knowledge, leading to hallucinations in knowledge-extensive tasks. To address this, Retrieval-Augmented Generation (RAG) incorporates external document chunks to expand LLM knowledge. Furthermore, compressing information from document chunks through extraction or summarization can improve LLM performance. Nonetheless, LLMs still struggle to notice and utilize scattered key information, a problem known as the\"lost-in-the-middle\"syndrome. Therefore, we typically need to restructure the content for LLM to recognize the key information. We propose $\\textit{Refiner}$, an end-to-end extract-and-restructure paradigm that operates in the post-retrieval process of RAG. $\\textit{Refiner}$ leverages a single decoder-only LLM to adaptively extract query-relevant contents verbatim along with the necessary context, and section them based on their interconnectedness, thereby highlights information distinction, and aligns downstream LLMs with the original context effectively. Experiments show that a trained $\\textit{Refiner}$ (with 7B parameters) exhibits significant gain to downstream LLM in improving answer accuracy, and outperforms other state-of-the-art advanced RAG and concurrent compressing approaches in various single-hop and multi-hop QA tasks. Notably, $\\textit{Refiner}$ achieves a 80.5% tokens reduction and a 1.6-7.0% improvement margin in multi-hop tasks compared to the next best solution. $\\textit{Refiner}$ is a plug-and-play solution that can be seamlessly integrated with RAG systems, facilitating its application across diverse open-source frameworks."
                    },
                    {
                        "title": "TIS-DPO: Token-level Importance Sampling for Direct Preference Optimization With Estimated Weights",
                        "abstract": "Direct Preference Optimization (DPO) has been widely adopted for preference alignment of Large Language Models (LLMs) due to its simplicity and effectiveness. However, DPO is derived as a bandit problem in which the whole response is treated as a single arm, ignoring the importance differences between tokens, which may affect optimization efficiency and make it difficult to achieve optimal results. In this work, we propose that the optimal data for DPO has equal expected rewards for each token in winning and losing responses, as there is no difference in token importance. However, since the optimal dataset is unavailable in practice, we propose using the original dataset for importance sampling to achieve unbiased optimization. Accordingly, we propose a token-level importance sampling DPO objective named TIS-DPO that assigns importance weights to each token based on its reward. Inspired by previous works, we estimate the token importance weights using the difference in prediction probabilities from a pair of contrastive LLMs. We explore three methods to construct these contrastive LLMs: (1) guiding the original LLM with contrastive prompts, (2) training two separate LLMs using winning and losing responses, and (3) performing forward and reverse DPO training with winning and losing responses. Experiments show that TIS-DPO significantly outperforms various baseline methods on harmlessness and helpfulness alignment and summarization tasks. We also visualize the estimated weights, demonstrating their ability to identify key token positions."
                    },
                    {
                        "title": "Entropy-Based Decoding for Retrieval-Augmented Large Language Models",
                        "abstract": "Augmenting Large Language Models (LLMs) with retrieved external knowledge has proven effective for improving the factual accuracy of generated responses. Despite their success, retrieval-augmented LLMs still face the distractibility issue, where the generated responses are negatively influenced by noise from both external and internal knowledge sources. In this paper, we introduce a novel, training-free decoding method guided by entropy considerations to mitigate this issue. Our approach utilizes entropy-based document-parallel ensemble decoding to prioritize low-entropy distributions from retrieved documents, thereby enhancing the extraction of relevant information of context. Additionally, it incorporates a contrastive decoding mechanism that contrasts the obtained low-entropy ensemble distribution with the high-entropy distribution derived from the model's internal knowledge across layers, which ensures a greater emphasis on reliable external information. Extensive experiments on open-domain question answering datasets demonstrate the superiority of our method."
                    },
                    {
                        "title": "ChatCite: LLM Agent with Human Workflow Guidance for Comparative Literature Summary",
                        "abstract": "The literature review is an indispensable step in the research process. It provides the benefit of comprehending the research problem and understanding the current research situation while conducting a comparative analysis of prior works. However, literature summary is challenging and time consuming. The previous LLM-based studies on literature review mainly focused on the complete process, including literature retrieval, screening, and summarization. However, for the summarization step, simple CoT method often lacks the ability to provide extensive comparative summary. In this work, we firstly focus on the independent literature summarization step and introduce ChatCite, an LLM agent with human workflow guidance for comparative literature summary. This agent, by mimicking the human workflow, first extracts key elements from relevant literature and then generates summaries using a Reflective Incremental Mechanism. In order to better evaluate the quality of the generated summaries, we devised a LLM-based automatic evaluation metric, G-Score, in refer to the human evaluation criteria. The ChatCite agent outperformed other models in various dimensions in the experiments. The literature summaries generated by ChatCite can also be directly used for drafting literature reviews."
                    },
                    {
                        "title": "Mitigating Modality Prior-Induced Hallucinations in Multimodal Large Language Models via Deciphering Attention Causality",
                        "abstract": "Multimodal Large Language Models (MLLMs) have emerged as a central focus in both industry and academia, but often suffer from biases introduced by visual and language priors, which can lead to multimodal hallucination. These biases arise from the visual encoder and the Large Language Model (LLM) backbone, affecting the attention mechanism responsible for aligning multimodal inputs. Existing decoding-based mitigation methods focus on statistical correlations and overlook the causal relationships between attention mechanisms and model output, limiting their effectiveness in addressing these biases. To tackle this issue, we propose a causal inference framework termed CausalMM that applies structural causal modeling to MLLMs, treating modality priors as a confounder between attention mechanisms and output. Specifically, by employing backdoor adjustment and counterfactual reasoning at both the visual and language attention levels, our method mitigates the negative effects of modality priors and enhances the alignment of MLLM's inputs and outputs, with a maximum score improvement of 65.3% on 6 VLind-Bench indicators and 164 points on MME Benchmark compared to conventional methods. Extensive experiments validate the effectiveness of our approach while being a plug-and-play solution. Our code is available at: https://github.com/The-Martyr/CausalMM"
                    },
                    {
                        "title": "MarkLLM: An Open-Source Toolkit for LLM Watermarking",
                        "abstract": "LLM watermarking, which embeds imperceptible yet algorithmically detectable signals in model outputs to identify LLM-generated text, has become crucial in mitigating the potential misuse of large language models. However, the abundance of LLM watermarking algorithms, their intricate mechanisms, and the complex evaluation procedures and perspectives pose challenges for researchers and the community to easily experiment with, understand, and assess the latest advancements. To address these issues, we introduce MarkLLM, an open-source toolkit for LLM watermarking. MarkLLM offers a unified and extensible framework for implementing LLM watermarking algorithms, while providing user-friendly interfaces to ensure ease of access. Furthermore, it enhances understanding by supporting automatic visualization of the underlying mechanisms of these algorithms. For evaluation, MarkLLM offers a comprehensive suite of 12 tools spanning three perspectives, along with two types of automated evaluation pipelines. Through MarkLLM, we aim to support researchers while improving the comprehension and involvement of the general public in LLM watermarking technology, fostering consensus and driving further advancements in research and application. Our code is available at https://github.com/THU-BPM/MarkLLM."
                    },
                    {
                        "title": "A Survey of AIOps for Failure Management in the Era of Large Language Models",
                        "abstract": "As software systems grow increasingly intricate, Artificial Intelligence for IT Operations (AIOps) methods have been widely used in software system failure management to ensure the high availability and reliability of large-scale distributed software systems. However, these methods still face several challenges, such as lack of cross-platform generality and cross-task flexibility. Fortunately, recent advancements in large language models (LLMs) can significantly address these challenges, and many approaches have already been proposed to explore this field. However, there is currently no comprehensive survey that discusses the differences between LLM-based AIOps and traditional AIOps methods. Therefore, this paper presents a comprehensive survey of AIOps technology for failure management in the LLM era. It includes a detailed definition of AIOps tasks for failure management, the data sources for AIOps, and the LLM-based approaches adopted for AIOps. Additionally, this survey explores the AIOps subtasks, the specific LLM-based approaches suitable for different AIOps subtasks, and the challenges and future directions of the domain, aiming to further its development and application."
                    },
                    {
                        "title": "WaterSeeker: Pioneering Efficient Detection of Watermarked Segments in Large Documents",
                        "abstract": "Watermarking algorithms for large language models (LLMs) have attained high accuracy in detecting LLM-generated text. However, existing methods primarily focus on distinguishing fully watermarked text from non-watermarked text, overlooking real-world scenarios where LLMs generate only small sections within large documents. In this scenario, balancing time complexity and detection performance poses significant challenges. This paper presents WaterSeeker, a novel approach to efficiently detect and locate watermarked segments amid extensive natural text. It first applies an efficient anomaly extraction method to preliminarily locate suspicious watermarked regions. Following this, it conducts a local traversal and performs full-text detection for more precise verification. Theoretical analysis and experimental results demonstrate that WaterSeeker achieves a superior balance between detection accuracy and computational efficiency. Moreover, WaterSeeker's localization ability supports the development of interpretable AI detection systems. This work pioneers a new direction in watermarked segment detection, facilitating more reliable AI-generated content identification.Our code is available at https://github.com/THU-BPM/WaterSeeker."
                    },
                    {
                        "title": "Can Watermarked LLMs be Identified by Users via Crafted Prompts?",
                        "abstract": "Text watermarking for Large Language Models (LLMs) has made significant progress in detecting LLM outputs and preventing misuse. Current watermarking techniques offer high detectability, minimal impact on text quality, and robustness to text editing. However, current researches lack investigation into the imperceptibility of watermarking techniques in LLM services. This is crucial as LLM providers may not want to disclose the presence of watermarks in real-world scenarios, as it could reduce user willingness to use the service and make watermarks more vulnerable to attacks. This work is the first to investigate the imperceptibility of watermarked LLMs. We design an identification algorithm called Water-Probe that detects watermarks through well-designed prompts to the LLM. Our key motivation is that current watermarked LLMs expose consistent biases under the same watermark key, resulting in similar differences across prompts under different watermark keys. Experiments show that almost all mainstream watermarking algorithms are easily identified with our well-designed prompts, while Water-Probe demonstrates a minimal false positive rate for non-watermarked LLMs. Finally, we propose that the key to enhancing the imperceptibility of watermarked LLMs is to increase the randomness of watermark key selection. Based on this, we introduce the Water-Bag strategy, which significantly improves watermark imperceptibility by merging multiple watermark keys."
                    },
                    {
                        "title": "Recent Advances of Multimodal Continual Learning: A Comprehensive Survey",
                        "abstract": "Continual learning (CL) aims to empower machine learning models to learn continually from new data, while building upon previously acquired knowledge without forgetting. As machine learning models have evolved from small to large pre-trained architectures, and from supporting unimodal to multimodal data, multimodal continual learning (MMCL) methods have recently emerged. The primary challenge of MMCL is that it goes beyond a simple stacking of unimodal CL methods, as such straightforward approaches often yield unsatisfactory performance. In this work, we present the first comprehensive survey on MMCL. We provide essential background knowledge and MMCL settings, as well as a structured taxonomy of MMCL methods. We categorize existing MMCL methods into four categories, i.e., regularization-based, architecture-based, replay-based, and prompt-based methods, explaining their methodologies and highlighting their key innovations. Additionally, to prompt further research in this field, we summarize open MMCL datasets and benchmarks, and discuss several promising future directions for investigation and development. We have also created a GitHub repository for indexing relevant MMCL papers and open resources available at https://github.com/LucyDYu/Awesome-Multimodal-Continual-Learning."
                    },
                    {
                        "title": "Reading Broadly to Open Your Mind: Improving Open Relation Extraction With Search Documents Under Self-Supervisions",
                        "abstract": "Open relation extraction is the task of extracting open-domain relation facts from natural language sentences. Existing works either utilize distant-supervised annotations to train a supervised classifier over pre-defined relations, or adopt unsupervised methods with additional dependency on external assumptions. However, these works can only obtain information signals from limited existing knowledge bases or datasets. In this work, we propose a self-supervised framework named Web-SelfORE, which exploits self-supervised signals by requiring a large pretrained language model to extensively read real-world relevant documents from the web, and obtain contextualized relational features by mixing contextualized representations of entities from different documents. We perform adaptive clustering on contextualized relational features and bootstrap the self-supervised signals by improving contextualized features in relation classification. We additionally compare the effectiveness of self-supervisions brought by different document sources, and introduce relevance and redundancy evaluation metrics to obtain higher-quality self-supervisions. Experimental results on four public datasets show the effectiveness and robustness of Web-SelfORE on open-domain relation extraction task when comparing with competitive baselines."
                    },
                    {
                        "title": "Can Watermarks Survive Translation? On the Cross-lingual Consistency of Text Watermark for Large Language Models",
                        "abstract": "Text watermarking technology aims to tag and identify content produced by large language models (LLMs) to prevent misuse. In this study, we introduce the concept of cross-lingual consistency in text watermarking, which assesses the ability of text watermarks to maintain their effectiveness after being translated into other languages. Preliminary empirical results from two LLMs and three watermarking methods reveal that current text watermarking technologies lack consistency when texts are translated into various languages. Based on this observation, we propose a Cross-lingual Watermark Removal Attack (CWRA) to bypass watermarking by first obtaining a response from an LLM in a pivot language, which is then translated into the target language. CWRA can effectively remove watermarks, decreasing the AUCs to a random-guessing level without performance loss. Furthermore, we analyze two key factors that contribute to the cross-lingual consistency in text watermarking and propose X-SIR as a defense method against CWRA. Code: https://github.com/zwhe99/X-SIR."
                    },
                    {
                        "title": "On the Robustness of Document-Level Relation Extraction Models to Entity Name Variations",
                        "abstract": "Driven by the demand for cross-sentence and large-scale relation extraction, document-level relation extraction (DocRE) has attracted increasing research interest. Despite the continuous improvement in performance, we find that existing DocRE models which initially perform well may make more mistakes when merely changing the entity names in the document, hindering the generalization to novel entity names. To this end, we systematically investigate the robustness of DocRE models to entity name variations in this work. We first propose a principled pipeline to generate entity-renamed documents by replacing the original entity names with names from Wikidata. By applying the pipeline to DocRED and Re-DocRED datasets, we construct two novel benchmarks named Env-DocRED and Env-Re-DocRED for robustness evaluation. Experimental results show that both three representative DocRE models and two in-context learned large language models consistently lack sufficient robustness to entity name variations, particularly on cross-sentence relation instances and documents with more entities. Finally, we propose an entity variation robust training method which not only improves the robustness of DocRE models but also enhances their understanding and reasoning capabilities. We further verify that the basic idea of this method can be effectively transferred to in-context learning for DocRE as well."
                    },
                    {
                        "title": "Direct Large Language Model Alignment Through Self-Rewarding Contrastive Prompt Distillation",
                        "abstract": "Aligning large language models (LLMs) with human expectations without human-annotated preference data is an important problem. In this paper, we propose a method to evaluate the response preference by using the output probabilities of response pairs under contrastive prompt pairs, which could achieve better performance on LLaMA2-7B and LLaMA2-13B compared to RLAIF. Based on this, we propose an automatic alignment method, Direct Large Model Alignment (DLMA). First, we use contrastive prompt pairs to automatically generate preference data. Then, we continue to evaluate the generated preference data using contrastive prompt pairs and calculate a self-rewarding score. Finally, we use the DPO algorithm to effectively align LLMs by combining this self-rewarding score. In the experimental stage, our DLMA method could surpass the \\texttt{RLHF} method without relying on human-annotated preference data."
                    },
                    {
                        "title": "Interpretable Contrastive Monte Carlo Tree Search Reasoning",
                        "abstract": "We propose SC-MCTS*: a novel Monte Carlo Tree Search (MCTS) reasoning algorithm for Large Language Models (LLMs), significantly improves both reasoning accuracy and speed. Our motivation comes from: 1. Previous MCTS LLM reasoning works often overlooked its biggest drawback--slower speed compared to CoT; 2. Previous research mainly used MCTS as a tool for LLM reasoning on various tasks with limited quantitative analysis or ablation studies of its components from reasoning interpretability perspective. 3. The reward model is the most crucial component in MCTS, however previous work has rarely conducted in-depth study or improvement of MCTS's reward models. Thus, we conducted extensive ablation studies and quantitative analysis on components of MCTS, revealing the impact of each component on the MCTS reasoning performance of LLMs. Building on this, (i) we designed a highly interpretable reward model based on the principle of contrastive decoding and (ii) achieved an average speed improvement of 51.9% per node using speculative decoding. Additionally, (iii) we improved UCT node selection strategy and backpropagation used in previous works, resulting in significant performance improvement. We outperformed o1-mini by an average of 17.4% on the Blocksworld multi-step reasoning dataset using Llama-3.1-70B with SC-MCTS*. Our code is available at \\url{https://github.com/zitian-gao/SC-MCTS}."
                    },
                    {
                        "title": "An Entropy-based Text Watermarking Detection Method",
                        "abstract": "Text watermarking algorithms for large language models (LLMs) can effectively identify machine-generated texts by embedding and detecting hidden features in the text. Although the current text watermarking algorithms perform well in most high-entropy scenarios, its performance in low-entropy scenarios still needs to be improved. In this work, we opine that the influence of token entropy should be fully considered in the watermark detection process, $i.e.$, the weight of each token during watermark detection should be customized according to its entropy, rather than setting the weights of all tokens to the same value as in previous methods. Specifically, we propose \\textbf{E}ntropy-based Text \\textbf{W}atermarking \\textbf{D}etection (\\textbf{EWD}) that gives higher-entropy tokens higher influence weights during watermark detection, so as to better reflect the degree of watermarking. Furthermore, the proposed detection process is training-free and fully automated. From the experiments, we demonstrate that our EWD can achieve better detection performance in low-entropy scenarios, and our method is also general and can be applied to texts with different entropy distributions. Our code and data is available\\footnote{\\url{https://github.com/luyijian3/EWD}}. Additionally, our algorithm could be accessed through MarkLLM \\cite{pan2024markllm}\\footnote{\\url{https://github.com/THU-BPM/MarkLLM}}."
                    },
                    {
                        "title": "GDA: Generative Data Augmentation Techniques for Relation Extraction Tasks",
                        "abstract": "Relation extraction (RE) tasks show promising performance in extracting relations from two entities mentioned in sentences, given sufficient annotations available during training. Such annotations would be labor-intensive to obtain in practice. Existing work adopts data augmentation techniques to generate pseudo-annotated sentences beyond limited annotations. These techniques neither preserve the semantic consistency of the original sentences when rule-based augmentations are adopted, nor preserve the syntax structure of sentences when expressing relations using seq2seq models, resulting in less diverse augmentations. In this work, we propose a dedicated augmentation technique for relational texts, named GDA, which uses two complementary modules to preserve both semantic consistency and syntax structures. We adopt a generative formulation and design a multi-tasking solution to achieve synergies. Furthermore, GDA adopts entity hints as the prior knowledge of the generative model to augment diverse sentences. Experimental results in three datasets under a low-resource setting showed that GDA could bring {\\em 2.0\\%} F1 improvements compared with no augmentation technique. Source code and data are available."
                    }
                ]
            },
            "97515290-a420-4c28-92cf-a84f5b9f59f3": {
                "pk": "97515290-a420-4c28-92cf-a84f5b9f59f3",
                "name": "Leyi Pan",
                "collaborators": [
                    "Aiwei Liu",
                    "Lijie Wen",
                    "Xuming Hu",
                    "Irwin King",
                    "Philip S. Yu",
                    "Yijian Lu",
                    "Zitian Gao",
                    "Zhiwei He",
                    "Xuandong Zhao",
                    "Binglin Zhou",
                    "Shuliang Liu",
                    "Yichen Di",
                    "Sheng Guan",
                    "Yiming Liu",
                    "Yifei Zhang",
                    "Liancheng Fang",
                    "Jingjing Li",
                    "Shiao Meng",
                    "Shuang Li"
                ],
                "domain": [
                    "LLM Watermarking",
                    "Text Detection",
                    "Neural Networks",
                    "AI Safety"
                ],
                "publications": [
                    {
                        "title": "MarkLLM: An Open-Source Toolkit for LLM Watermarking",
                        "abstract": "LLM watermarking, which embeds imperceptible yet algorithmically detectable signals in model outputs to identify LLM-generated text, has become crucial in mitigating the potential misuse of large language models. However, the abundance of LLM watermarking algorithms, their intricate mechanisms, and the complex evaluation procedures and perspectives pose challenges for researchers and the community to easily experiment with, understand, and assess the latest advancements. To address these issues, we introduce MarkLLM, an open-source toolkit for LLM watermarking. MarkLLM offers a unified and extensible framework for implementing LLM watermarking algorithms, while providing user-friendly interfaces to ensure ease of access. Furthermore, it enhances understanding by supporting automatic visualization of the underlying mechanisms of these algorithms. For evaluation, MarkLLM offers a comprehensive suite of 12 tools spanning three perspectives, along with two types of automated evaluation pipelines. Through MarkLLM, we aim to support researchers while improving the comprehension and involvement of the general public in LLM watermarking technology, fostering consensus and driving further advancements in research and application. Our code is available at https://github.com/THU-BPM/MarkLLM."
                    },
                    {
                        "title": "WaterSeeker: Pioneering Efficient Detection of Watermarked Segments in Large Documents",
                        "abstract": "Watermarking algorithms for large language models (LLMs) have attained high accuracy in detecting LLM-generated text. However, existing methods primarily focus on distinguishing fully watermarked text from non-watermarked text, overlooking real-world scenarios where LLMs generate only small sections within large documents. In this scenario, balancing time complexity and detection performance poses significant challenges. This paper presents WaterSeeker, a novel approach to efficiently detect and locate watermarked segments amid extensive natural text. It first applies an efficient anomaly extraction method to preliminarily locate suspicious watermarked regions. Following this, it conducts a local traversal and performs full-text detection for more precise verification. Theoretical analysis and experimental results demonstrate that WaterSeeker achieves a superior balance between detection accuracy and computational efficiency. Moreover, WaterSeeker's localization ability supports the development of interpretable AI detection systems. This work pioneers a new direction in watermarked segment detection, facilitating more reliable AI-generated content identification.Our code is available at https://github.com/THU-BPM/WaterSeeker."
                    },
                    {
                        "title": "Can Watermarked LLMs be Identified by Users via Crafted Prompts?",
                        "abstract": "Text watermarking for Large Language Models (LLMs) has made significant progress in detecting LLM outputs and preventing misuse. Current watermarking techniques offer high detectability, minimal impact on text quality, and robustness to text editing. However, current researches lack investigation into the imperceptibility of watermarking techniques in LLM services. This is crucial as LLM providers may not want to disclose the presence of watermarks in real-world scenarios, as it could reduce user willingness to use the service and make watermarks more vulnerable to attacks. This work is the first to investigate the imperceptibility of watermarked LLMs. We design an identification algorithm called Water-Probe that detects watermarks through well-designed prompts to the LLM. Our key motivation is that current watermarked LLMs expose consistent biases under the same watermark key, resulting in similar differences across prompts under different watermark keys. Experiments show that almost all mainstream watermarking algorithms are easily identified with our well-designed prompts, while Water-Probe demonstrates a minimal false positive rate for non-watermarked LLMs. Finally, we propose that the key to enhancing the imperceptibility of watermarked LLMs is to increase the randomness of watermark key selection. Based on this, we introduce the Water-Bag strategy, which significantly improves watermark imperceptibility by merging multiple watermark keys."
                    },
                    {
                        "title": "A Survey of Text Watermarking in the Era of Large Language Models",
                        "abstract": "Text watermarking algorithms are crucial for protecting the copyright of textual content. Historically, their capabilities and application scenarios were limited. However, recent advancements in large language models (LLMs) have revolutionized these techniques. LLMs not only enhance text watermarking algorithms with their advanced abilities but also create a need for employing these algorithms to protect their own copyrights or prevent potential misuse. This paper conducts a comprehensive survey of the current state of text watermarking technology, covering four main aspects: (1) an overview and comparison of different text watermarking techniques; (2) evaluation methods for text watermarking algorithms, including their detectability, impact on text or LLM quality, robustness under target or untargeted attacks; (3) potential application scenarios for text watermarking technology; (4) current challenges and future directions for text watermarking. This survey aims to provide researchers with a thorough understanding of text watermarking technology in the era of LLM, thereby promoting its further advancement."
                    },
                    {
                        "title": "A Semantic Invariant Robust Watermark for Large Language Models",
                        "abstract": "Watermark algorithms for large language models (LLMs) have achieved extremely high accuracy in detecting text generated by LLMs. Such algorithms typically involve adding extra watermark logits to the LLM's logits at each generation step. However, prior algorithms face a trade-off between attack robustness and security robustness. This is because the watermark logits for a token are determined by a certain number of preceding tokens; a small number leads to low security robustness, while a large number results in insufficient attack robustness. In this work, we propose a semantic invariant watermarking method for LLMs that provides both attack robustness and security robustness. The watermark logits in our work are determined by the semantics of all preceding tokens. Specifically, we utilize another embedding LLM to generate semantic embeddings for all preceding tokens, and then these semantic embeddings are transformed into the watermark logits through our trained watermark model. Subsequent analyses and experiments demonstrated the attack robustness of our method in semantically invariant settings: synonym substitution and text paraphrasing settings. Finally, we also show that our watermark possesses adequate security robustness. Our code and data are available at \\href{https://github.com/THU-BPM/Robust_Watermark}{https://github.com/THU-BPM/Robust\\_Watermark}. Additionally, our algorithm could also be accessed through MarkLLM \\citep{pan2024markllm} \\footnote{https://github.com/THU-BPM/MarkLLM}."
                    },
                    {
                        "title": "A Private Watermark for Large Language Models",
                        "abstract": "Recently, text watermarking algorithms for large language models (LLMs) have been mitigating the potential harms of text generated by the LLMs, including fake news and copyright issues. However, the watermark detection of current text algorithms requires the key from the generation process, making them susceptible to breaches and counterfeiting. In this work, we propose the first private watermarking algorithm, which extends the current text watermarking algorithms by using two different neural networks respectively for watermark generation and detection, rather than using the same key at both stages. Meanwhile, part of the parameters of the watermark generation and detection networks are shared, which makes the detection network achieve a high accuracy very efficiently. Experiments show that our algorithm ensures high detection accuracy with minimal impact on generation and detection speed, due to the small parameter size of both networks. Additionally, our subsequent analysis demonstrates the difficulty of reverting the watermark generation rules from the detection network."
                    }
                ]
            },
            "ccab4d6e-b14c-474d-8e23-864622651949": {
                "pk": "ccab4d6e-b14c-474d-8e23-864622651949",
                "name": "Xuming Hu",
                "collaborators": [
                    "Philip S. Yu",
                    "Lijie Wen",
                    "Aiwei Liu",
                    "Yinghui Li",
                    "Yangning Li",
                    "Junzhe Chen",
                    "Zhijiang Guo",
                    "Hai-Tao Zheng",
                    "Shiao Meng",
                    "Leyi Pan",
                    "Zitian Gao",
                    "Irwin King",
                    "Xiaochuan Li",
                    "Xiaoguang Li",
                    "Yasheng Wang",
                    "Qun Liu",
                    "Shirong Ma",
                    "Zhaochen Hong",
                    "Wenhao Jiang",
                    "Zhonghao Li",
                    "Kening Zheng",
                    "Sirui Huang",
                    "Hui Xiong",
                    "Zhiwei He",
                    "Xuandong Zhao",
                    "Yijian Lu",
                    "Binglin Zhou",
                    "Shuliang Liu",
                    "Hkustgz",
                    "Lingzhe Zhang",
                    "Tong Jia",
                    "Mengxi Jia",
                    "Yifan Wu",
                    "Yong Yang",
                    "Zhonghai Wu",
                    "Ying Li",
                    "Sheng Guan",
                    "Yiming Liu",
                    "Yifei Zhang",
                    "Liancheng Fang",
                    "Qingyu Zhou",
                    "Yuanzhen Luo",
                    "Chenwei Zhang",
                    "Qingsong Lv",
                    "Tianyu Yu",
                    "Shulin Huang",
                    "Tingwei Lu",
                    "Hui Wang",
                    "Yong Jiang",
                    "Zhichao Lin",
                    "Xiaobin Wang",
                    "Pengjun Xie",
                    "Shang Qin",
                    "Jingheng Ye",
                    "Libo Qin",
                    "Shuodi Liu",
                    "Shiyu Huang",
                    "Weijuan Tu",
                    "Zhaofeng He",
                    "Fukun Ma",
                    "Yawen Yang",
                    "Shuang Li",
                    "Boye Niu",
                    "Xuzheng He",
                    "Haotian Xu",
                    "Hongzhang Liu",
                    "Guan-Huei Wu",
                    "Zhiyang Teng",
                    "I. King"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Information Retrieval",
                    "Adversarial Learning"
                ],
                "publications": [
                    {
                        "title": "Refiner: Restructure Retrieval Content Efficiently to Advance Question-Answering Capabilities",
                        "abstract": "Large Language Models (LLMs) are limited by their parametric knowledge, leading to hallucinations in knowledge-extensive tasks. To address this, Retrieval-Augmented Generation (RAG) incorporates external document chunks to expand LLM knowledge. Furthermore, compressing information from document chunks through extraction or summarization can improve LLM performance. Nonetheless, LLMs still struggle to notice and utilize scattered key information, a problem known as the\"lost-in-the-middle\"syndrome. Therefore, we typically need to restructure the content for LLM to recognize the key information. We propose $\\textit{Refiner}$, an end-to-end extract-and-restructure paradigm that operates in the post-retrieval process of RAG. $\\textit{Refiner}$ leverages a single decoder-only LLM to adaptively extract query-relevant contents verbatim along with the necessary context, and section them based on their interconnectedness, thereby highlights information distinction, and aligns downstream LLMs with the original context effectively. Experiments show that a trained $\\textit{Refiner}$ (with 7B parameters) exhibits significant gain to downstream LLM in improving answer accuracy, and outperforms other state-of-the-art advanced RAG and concurrent compressing approaches in various single-hop and multi-hop QA tasks. Notably, $\\textit{Refiner}$ achieves a 80.5% tokens reduction and a 1.6-7.0% improvement margin in multi-hop tasks compared to the next best solution. $\\textit{Refiner}$ is a plug-and-play solution that can be seamlessly integrated with RAG systems, facilitating its application across diverse open-source frameworks."
                    },
                    {
                        "title": "MarkLLM: An Open-Source Toolkit for LLM Watermarking",
                        "abstract": "LLM watermarking, which embeds imperceptible yet algorithmically detectable signals in model outputs to identify LLM-generated text, has become crucial in mitigating the potential misuse of large language models. However, the abundance of LLM watermarking algorithms, their intricate mechanisms, and the complex evaluation procedures and perspectives pose challenges for researchers and the community to easily experiment with, understand, and assess the latest advancements. To address these issues, we introduce MarkLLM, an open-source toolkit for LLM watermarking. MarkLLM offers a unified and extensible framework for implementing LLM watermarking algorithms, while providing user-friendly interfaces to ensure ease of access. Furthermore, it enhances understanding by supporting automatic visualization of the underlying mechanisms of these algorithms. For evaluation, MarkLLM offers a comprehensive suite of 12 tools spanning three perspectives, along with two types of automated evaluation pipelines. Through MarkLLM, we aim to support researchers while improving the comprehension and involvement of the general public in LLM watermarking technology, fostering consensus and driving further advancements in research and application. Our code is available at https://github.com/THU-BPM/MarkLLM."
                    },
                    {
                        "title": "Evaluating Robustness of Generative Search Engine on Adversarial Factoid Questions",
                        "abstract": "Generative search engines have the potential to transform how people seek information on-line, but generated responses from existing large language models (LLMs)-backed generative search engines may not always be accurate. Nonetheless, retrieval-augmented generation exacerbates safety concerns, since adversaries may successfully evade the entire sys-tem by subtly manipulating the most vulnerable part of a claim. To this end, we pro-pose evaluating the robustness of generative search engines in the realistic and high-risk setting, where adversaries have only black-box system access and seek to deceive the model into returning incorrect responses. Through a comprehensive human evaluation of various generative search engines, such as Bing Chat, PerplexityAI, and YouChat across diverse queries, we demonstrate the effectiveness of adversarial factual questions in inducing incorrect responses. Moreover, retrieval-augmented generation exhibits a higher susceptibility to factual errors compared to LLMs without retrieval. These findings highlight the potential security risks of these systems and emphasize the need for rigorous evaluation before deployment. Our constructed dataset and codes are available at: https://github.com/HKUSTGZ-NLP/Adversarial-Attack."
                    },
                    {
                        "title": "A Survey of AIOps for Failure Management in the Era of Large Language Models",
                        "abstract": "As software systems grow increasingly intricate, Artificial Intelligence for IT Operations (AIOps) methods have been widely used in software system failure management to ensure the high availability and reliability of large-scale distributed software systems. However, these methods still face several challenges, such as lack of cross-platform generality and cross-task flexibility. Fortunately, recent advancements in large language models (LLMs) can significantly address these challenges, and many approaches have already been proposed to explore this field. However, there is currently no comprehensive survey that discusses the differences between LLM-based AIOps and traditional AIOps methods. Therefore, this paper presents a comprehensive survey of AIOps technology for failure management in the LLM era. It includes a detailed definition of AIOps tasks for failure management, the data sources for AIOps, and the LLM-based approaches adopted for AIOps. Additionally, this survey explores the AIOps subtasks, the specific LLM-based approaches suitable for different AIOps subtasks, and the challenges and future directions of the domain, aiming to further its development and application."
                    },
                    {
                        "title": "Can Watermarked LLMs be Identified by Users via Crafted Prompts?",
                        "abstract": "Text watermarking for Large Language Models (LLMs) has made significant progress in detecting LLM outputs and preventing misuse. Current watermarking techniques offer high detectability, minimal impact on text quality, and robustness to text editing. However, current researches lack investigation into the imperceptibility of watermarking techniques in LLM services. This is crucial as LLM providers may not want to disclose the presence of watermarks in real-world scenarios, as it could reduce user willingness to use the service and make watermarks more vulnerable to attacks. This work is the first to investigate the imperceptibility of watermarked LLMs. We design an identification algorithm called Water-Probe that detects watermarks through well-designed prompts to the LLM. Our key motivation is that current watermarked LLMs expose consistent biases under the same watermark key, resulting in similar differences across prompts under different watermark keys. Experiments show that almost all mainstream watermarking algorithms are easily identified with our well-designed prompts, while Water-Probe demonstrates a minimal false positive rate for non-watermarked LLMs. Finally, we propose that the key to enhancing the imperceptibility of watermarked LLMs is to increase the randomness of watermark key selection. Based on this, we introduce the Water-Bag strategy, which significantly improves watermark imperceptibility by merging multiple watermark keys."
                    },
                    {
                        "title": "When LLMs Meet Cunning Texts: A Fallacy Understanding Benchmark for Large Language Models",
                        "abstract": "Recently, Large Language Models (LLMs) make remarkable evolutions in language understanding and generation. Following this, various benchmarks for measuring all kinds of capabilities of LLMs have sprung up. In this paper, we challenge the reasoning and understanding abilities of LLMs by proposing a FaLlacy Understanding Benchmark (FLUB) containing cunning texts that are easy for humans to understand but difficult for models to grasp. Specifically, the cunning texts that FLUB focuses on mainly consist of the tricky, humorous, and misleading texts collected from the real internet environment. And we design three tasks with increasing difficulty in the FLUB benchmark to evaluate the fallacy understanding ability of LLMs. Based on FLUB, we investigate the performance of multiple representative and advanced LLMs, reflecting our FLUB is challenging and worthy of more future study. Interesting discoveries and valuable insights are achieved in our extensive experiments and detailed analyses. We hope that our benchmark can encourage the community to improve LLMs' ability to understand fallacies. Our data and codes are available at https://github.com/THUKElab/FLUB."
                    },
                    {
                        "title": "Evaluating Robustness of Generative Search Engine on Adversarial Factual Questions",
                        "abstract": "Generative search engines have the potential to transform how people seek information online, but generated responses from existing large language models (LLMs)-backed generative search engines may not always be accurate. Nonetheless, retrieval-augmented generation exacerbates safety concerns, since adversaries may successfully evade the entire system by subtly manipulating the most vulnerable part of a claim. To this end, we propose evaluating the robustness of generative search engines in the realistic and high-risk setting, where adversaries have only black-box system access and seek to deceive the model into returning incorrect responses. Through a comprehensive human evaluation of various generative search engines, such as Bing Chat, PerplexityAI, and YouChat across diverse queries, we demonstrate the effectiveness of adversarial factual questions in inducing incorrect responses. Moreover, retrieval-augmented generation exhibits a higher susceptibility to factual errors compared to LLMs without retrieval. These findings highlight the potential security risks of these systems and emphasize the need for rigorous evaluation before deployment."
                    },
                    {
                        "title": "Reading Broadly to Open Your Mind: Improving Open Relation Extraction With Search Documents Under Self-Supervisions",
                        "abstract": "Open relation extraction is the task of extracting open-domain relation facts from natural language sentences. Existing works either utilize distant-supervised annotations to train a supervised classifier over pre-defined relations, or adopt unsupervised methods with additional dependency on external assumptions. However, these works can only obtain information signals from limited existing knowledge bases or datasets. In this work, we propose a self-supervised framework named Web-SelfORE, which exploits self-supervised signals by requiring a large pretrained language model to extensively read real-world relevant documents from the web, and obtain contextualized relational features by mixing contextualized representations of entities from different documents. We perform adaptive clustering on contextualized relational features and bootstrap the self-supervised signals by improving contextualized features in relation classification. We additionally compare the effectiveness of self-supervisions brought by different document sources, and introduce relevance and redundancy evaluation metrics to obtain higher-quality self-supervisions. Experimental results on four public datasets show the effectiveness and robustness of Web-SelfORE on open-domain relation extraction task when comparing with competitive baselines."
                    },
                    {
                        "title": "UltraWiki: Ultra-fine-grained Entity Set Expansion with Negative Seed Entities",
                        "abstract": "Entity Set Expansion (ESE) aims to identify new entities belonging to the same semantic class as a given set of seed entities. Traditional methods primarily relied on positive seed entities to represent a target semantic class, which poses challenge for the representation of ultra-fine-grained semantic classes. Ultra-fine-grained semantic classes are defined based on fine-grained semantic classes with more specific attribute constraints. Describing it with positive seed entities alone cause two issues: (i) Ambiguity among ultra-fine-grained semantic classes. (ii) Inability to define\"unwanted\"semantic. Due to these inherent shortcomings, previous methods struggle to address the ultra-fine-grained ESE (Ultra-ESE). To solve this issue, we first introduce negative seed entities in the inputs, which belong to the same fine-grained semantic class as the positive seed entities but differ in certain attributes. Negative seed entities eliminate the semantic ambiguity by contrast between positive and negative attributes. Meanwhile, it provide a straightforward way to express\"unwanted\". To assess model performance in Ultra-ESE, we constructed UltraWiki, the first large-scale dataset tailored for Ultra-ESE. UltraWiki encompasses 236 ultra-fine-grained semantic classes, where each query of them is represented with 3-5 positive and negative seed entities. A retrieval-based framework RetExpan and a generation-based framework GenExpan are proposed to comprehensively assess the efficacy of large language models from two different paradigms in Ultra-ESE. Moreover, we devised three strategies to enhance models' comprehension of ultra-fine-grained entities semantics: contrastive learning, retrieval augmentation, and chain-of-thought reasoning. Extensive experiments confirm the effectiveness of our proposed strategies and also reveal that there remains a large space for improvement in Ultra-ESE."
                    },
                    {
                        "title": "Three Heads Are Better than One: Improving Cross-Domain NER with Progressive Decomposed Network",
                        "abstract": "Cross-domain named entity recognition (NER) tasks encourage NER models to transfer knowledge from data-rich source domains to sparsely labeled target domains. Previous works adopt the paradigms of pre-training on the source domain followed by fine-tuning on the target domain. However, these works ignore that general labeled NER source domain data can be easily retrieved in the real world, and soliciting more source domains could bring more benefits. Unfortunately, previous paradigms cannot efficiently transfer knowledge from multiple source domains. In this work, to transfer multiple source domains' knowledge, we decouple the NER task into the pipeline tasks of mention detection and entity typing, where the mention detection unifies the training object across domains, thus providing the entity typing with higher-quality entity mentions. Additionally, we request multiple general source domain models to suggest the potential named entities for sentences in the target domain explicitly, and transfer their knowledge to the target domain models through the knowledge progressive networks implicitly. Furthermore, we propose two methods to analyze in which source domain knowledge transfer occurs, thus helping us judge which source domain brings the greatest benefit. In our experiment, we develop a Chinese cross-domain NER dataset. Our model improved the F1 score by an average of 12.50% across 8 Chinese and English datasets compared to models without source domain data."
                    },
                    {
                        "title": "Rethinking the Roles of Large Language Models in Chinese Grammatical Error Correction",
                        "abstract": "Recently, Large Language Models (LLMs) have been widely studied by researchers for their roles in various downstream NLP tasks. As a fundamental task in the NLP field, Chinese Grammatical Error Correction (CGEC) aims to correct all potential grammatical errors in the input sentences. Previous studies have shown that LLMs' performance as correctors on CGEC remains unsatisfactory due to its challenging task focus. To promote the CGEC field to better adapt to the era of LLMs, we rethink the roles of LLMs in the CGEC task so that they can be better utilized and explored in CGEC. Considering the rich grammatical knowledge stored in LLMs and their powerful semantic understanding capabilities, we utilize LLMs as explainers to provide explanation information for the CGEC small models during error correction to enhance performance. We also use LLMs as evaluators to bring more reasonable CGEC evaluations, thus alleviating the troubles caused by the subjectivity of the CGEC task. In particular, our work is also an active exploration of how LLMs and small models better collaborate in downstream tasks. Extensive experiments and detailed analyses on widely used datasets verify the effectiveness of our thinking intuition and the proposed methods."
                    },
                    {
                        "title": "LLMArena: Assessing Capabilities of Large Language Models in Dynamic Multi-Agent Environments",
                        "abstract": "Recent advancements in large language models (LLMs) have revealed their potential for achieving autonomous agents possessing human-level intelligence. However, existing benchmarks for evaluating LLM Agents either use static datasets, potentially leading to data leakage or focus only on single-agent scenarios, overlooking the complexities of multi-agent interactions. There is a lack of a benchmark that evaluates the diverse capabilities of LLM agents in multi-agent, dynamic environments. To this end, we introduce LLMArena, a novel and easily extensible framework for evaluating the diverse capabilities of LLM in multi-agent dynamic environments. LLMArena encompasses seven distinct gaming environments, employing Trueskill scoring to assess crucial abilities in LLM agents, including spatial reasoning, strategic planning, numerical reasoning, risk assessment, communication, opponent modeling, and team collaboration. We conduct an extensive experiment and human evaluation among different sizes and types of LLMs, showing that LLMs still have a significant journey ahead in their development towards becoming fully autonomous agents, especially in opponent modeling and team collaboration. We hope LLMArena could guide future research towards enhancing these capabilities in LLMs, ultimately leading to more sophisticated and practical applications in dynamic, multi-agent settings. The code and data will be available."
                    },
                    {
                        "title": "On the Robustness of Document-Level Relation Extraction Models to Entity Name Variations",
                        "abstract": "Driven by the demand for cross-sentence and large-scale relation extraction, document-level relation extraction (DocRE) has attracted increasing research interest. Despite the continuous improvement in performance, we find that existing DocRE models which initially perform well may make more mistakes when merely changing the entity names in the document, hindering the generalization to novel entity names. To this end, we systematically investigate the robustness of DocRE models to entity name variations in this work. We first propose a principled pipeline to generate entity-renamed documents by replacing the original entity names with names from Wikidata. By applying the pipeline to DocRED and Re-DocRED datasets, we construct two novel benchmarks named Env-DocRED and Env-Re-DocRED for robustness evaluation. Experimental results show that both three representative DocRE models and two in-context learned large language models consistently lack sufficient robustness to entity name variations, particularly on cross-sentence relation instances and documents with more entities. Finally, we propose an entity variation robust training method which not only improves the robustness of DocRE models but also enhances their understanding and reasoning capabilities. We further verify that the basic idea of this method can be effectively transferred to in-context learning for DocRE as well."
                    },
                    {
                        "title": "Interpretable Contrastive Monte Carlo Tree Search Reasoning",
                        "abstract": "We propose SC-MCTS*: a novel Monte Carlo Tree Search (MCTS) reasoning algorithm for Large Language Models (LLMs), significantly improves both reasoning accuracy and speed. Our motivation comes from: 1. Previous MCTS LLM reasoning works often overlooked its biggest drawback--slower speed compared to CoT; 2. Previous research mainly used MCTS as a tool for LLM reasoning on various tasks with limited quantitative analysis or ablation studies of its components from reasoning interpretability perspective. 3. The reward model is the most crucial component in MCTS, however previous work has rarely conducted in-depth study or improvement of MCTS's reward models. Thus, we conducted extensive ablation studies and quantitative analysis on components of MCTS, revealing the impact of each component on the MCTS reasoning performance of LLMs. Building on this, (i) we designed a highly interpretable reward model based on the principle of contrastive decoding and (ii) achieved an average speed improvement of 51.9% per node using speculative decoding. Additionally, (iii) we improved UCT node selection strategy and backpropagation used in previous works, resulting in significant performance improvement. We outperformed o1-mini by an average of 17.4% on the Blocksworld multi-step reasoning dataset using Llama-3.1-70B with SC-MCTS*. Our code is available at \\url{https://github.com/zitian-gao/SC-MCTS}."
                    },
                    {
                        "title": "SelfLRE: Self-refining Representation Learning for Low-resource Relation Extraction",
                        "abstract": "Low-resource relation extraction (LRE) aims to extract potential relations from limited labeled corpus to handle the problem of scarcity of human annotations. Previous works mainly consist of two categories of methods: (1) Self-training methods, which improve themselves through the models' predictions, thus suffering from confirmation bias when the predictions are wrong. (2) Self-ensembling methods, which learn task-agnostic representations, therefore, generally do not work well for specific tasks. In our work, we propose a novel LRE architecture named SelfLRE, which leverages two complementary modules, one module uses self-training to obtain pseudo-labels for unlabeled data, and the other module uses self-ensembling learning to obtain the task-agnostic representations, and leverages the existing pseudo-labels to refine the better task-specific representations on unlabeled data. The two models are jointly trained through multi-task learning to iteratively improve the effect of LRE task. Experiments on three public datasets show that SelfLRE achieves 1.81% performance gain over the SOTA baseline. Source code is available at: https://github.com/THU-BPM/SelfLRE."
                    },
                    {
                        "title": "Give Me More Details: Improving Fact-Checking with Latent Retrieval",
                        "abstract": "Evidence plays a crucial role in automated fact-checking. When verifying real-world claims, existing fact-checking systems either assume the evidence sentences are given or use the search snippets returned by the search engine. Such methods ignore the challenges of collecting evidence and may not provide sufficient information to verify real-world claims. Aiming at building a better fact-checking system, we propose to incorporate full text from source documents as evidence and introduce two enriched datasets. The first one is a multilingual dataset, while the second one is monolingual (English). We further develop a latent variable model to jointly extract evidence sentences from documents and perform claim verification. Experiments indicate that including source documents can provide sufficient contextual clues even when gold evidence sentences are not annotated. The proposed system is able to achieve significant improvements upon best-reported models under different settings."
                    },
                    {
                        "title": "Multimodal Relation Extraction with Cross-Modal Retrieval and Synthesis",
                        "abstract": "Multimodal relation extraction (MRE) is the task of identifying the semantic relationships between two entities based on the context of the sentence image pair. Existing retrieval-augmented approaches mainly focused on modeling the retrieved textual knowledge, but this may not be able to accurately identify complex relations. To improve the prediction, this research proposes to retrieve textual and visual evidence based on the object, sentence, and whole image. We further develop a novel approach to synthesize the object-level, image-level, and sentence-level information for better reasoning between the same and different modalities. Extensive experiments and analyses show that the proposed method is able to effectively select and compare evidence across modalities and significantly outperforms state-of-the-art models."
                    }
                ]
            },
            "97f0dfc1-2bf5-44f4-b912-ac50d603cde3": {
                "pk": "97f0dfc1-2bf5-44f4-b912-ac50d603cde3",
                "name": "Shu'ang Li",
                "collaborators": [
                    "Lijie Wen",
                    "Xuming Hu",
                    "Aiwei Liu",
                    "Fukun Ma",
                    "Yawen Yang",
                    "Philip S. Yu",
                    "Li Lin",
                    "Shiao Meng",
                    "Xiaojuan Zhang",
                    "Jun Fu",
                    "Jiahua Wang",
                    "W. Liu",
                    "Leyi Pan",
                    "Irwin King",
                    "Dali Chen",
                    "Yixin Cao",
                    "Lifu Huang",
                    "L. Wen",
                    "Jianmin Wang",
                    "Shuliang Liu",
                    "Chenwei Zhang",
                    "Honghai Yu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Multi-modal Learning",
                    "Relation Extraction",
                    "Contrastive Learning"
                ],
                "publications": [
                    {
                        "title": "FSMR: A Feature Swapping Multi-modal Reasoning Approach with Joint Textual and Visual Clues",
                        "abstract": "Multi-modal reasoning plays a vital role in bridging the gap between textual and visual information, enabling a deeper understanding of the context. This paper presents the Feature Swapping Multi-modal Reasoning (FSMR) model, designed to enhance multi-modal reasoning through feature swapping. FSMR leverages a pre-trained visual-language model as an encoder, accommodating both text and image inputs for effective feature representation from both modalities. It introduces a unique feature swapping module, enabling the exchange of features between identified objects in images and corresponding vocabulary words in text, thereby enhancing the model's comprehension of the interplay between images and text. To further bolster its multi-modal alignment capabilities, FSMR incorporates a multi-modal cross-attention mechanism, facilitating the joint modeling of textual and visual information. During training, we employ image-text matching and cross-entropy losses to ensure semantic consistency between visual and language elements. Extensive experiments on the PMR dataset demonstrate FSMR's superiority over state-of-the-art baseline models across various performance metrics."
                    },
                    {
                        "title": "On the Robustness of Document-Level Relation Extraction Models to Entity Name Variations",
                        "abstract": "Driven by the demand for cross-sentence and large-scale relation extraction, document-level relation extraction (DocRE) has attracted increasing research interest. Despite the continuous improvement in performance, we find that existing DocRE models which initially perform well may make more mistakes when merely changing the entity names in the document, hindering the generalization to novel entity names. To this end, we systematically investigate the robustness of DocRE models to entity name variations in this work. We first propose a principled pipeline to generate entity-renamed documents by replacing the original entity names with names from Wikidata. By applying the pipeline to DocRED and Re-DocRED datasets, we construct two novel benchmarks named Env-DocRED and Env-Re-DocRED for robustness evaluation. Experimental results show that both three representative DocRE models and two in-context learned large language models consistently lack sufficient robustness to entity name variations, particularly on cross-sentence relation instances and documents with more entities. Finally, we propose an entity variation robust training method which not only improves the robustness of DocRE models but also enhances their understanding and reasoning capabilities. We further verify that the basic idea of this method can be effectively transferred to in-context learning for DocRE as well."
                    },
                    {
                        "title": "Contrastive Disentangled Learning on Graph for Node Classification",
                        "abstract": "Contrastive learning methods have attracted considerable attention due to their remarkable success in analyzing graph-structured data. Inspired by the success of contrastive learning, we propose a novel framework for contrastive disentangled learning on graphs, employing a disentangled graph encoder and two carefully crafted self-supervision signals. Specifically, we introduce a disentangled graph encoder to enforce the framework to distinguish various latent factors corresponding to underlying semantic information and learn the disentangled node embeddings. Moreover, to overcome the heavy reliance on labels, we design two self-supervision signals, namely node specificity and channel independence, which capture informative knowledge without the need for labeled data, thereby guiding the automatic disentanglement of nodes. Finally, we perform node classification tasks on three citation networks by using the disentangled node embeddings, and the relevant analysis is provided. Experimental results validate the effectiveness of the proposed framework compared with various baselines."
                    },
                    {
                        "title": "Exploring the Compositional Generalization in Context Dependent Text-to-SQL Parsing",
                        "abstract": "In the context-dependent Text-to-SQL task, the generated SQL statements are refined iteratively based on the user input utterance from each interaction. The input text from each interaction can be viewed as component modifications to the previous SQL statements, which could be further extracted as the modification patterns. Since these modification patterns could also be combined with other SQL statements, the models are supposed to have the compositional generalization to these novel combinations. This work is the first exploration of compositional generalization in context-dependent Text-to-SQL scenarios. To facilitate related studies, we constructed two challenging benchmarks named \\textsc{CoSQL-CG} and \\textsc{SParC-CG} by recombining the modification patterns and existing SQL statements. The following experiments show that all current models struggle on our proposed benchmarks. Furthermore, we found that better aligning the previous SQL statements with the input utterance could give models better compositional generalization ability. Based on these observations, we propose a method named \\texttt{p-align} to improve the compositional generalization of Text-to-SQL models. Further experiments validate the effectiveness of our method. Source code and data are available."
                    },
                    {
                        "title": "Enhancing Cross-lingual Natural Language Inference by Soft Prompting with Multilingual Verbalizer",
                        "abstract": "Cross-lingual natural language inference is a fundamental problem in cross-lingual language understanding. Many recent works have used prompt learning to address the lack of annotated parallel corpora in XNLI. However, these methods adopt discrete prompting by simply translating the templates to the target language and need external expert knowledge to design the templates. Besides, discrete prompts of human-designed template words are not trainable vectors and can not be migrated to target languages in the inference stage flexibly. In this paper, we propose a novel Soft prompt learning framework with the Multilingual Verbalizer (SoftMV) for XNLI. SoftMV first constructs cloze-style question with soft prompts for the input sample. Then we leverage bilingual dictionaries to generate an augmented multilingual question for the original question. SoftMV adopts a multilingual verbalizer to align the representations of original and augmented multilingual questions into the same semantic space with consistency regularization. Experimental results on XNLI demonstrate that SoftMV can achieve state-of-the-art performance and significantly outperform the previous methods under the few-shot and full-shot cross-lingual transfer settings."
                    },
                    {
                        "title": "AMR-based Network for Aspect-based Sentiment Analysis",
                        "abstract": "Aspect-based sentiment analysis (ABSA) is a fine-grained sentiment classification task. Many recent works have used dependency trees to extract the relation between aspects and contexts and have achieved significant improvements. However, further improvement is limited due to the potential mismatch between the dependency tree as a syntactic structure and the sentiment classification as a semantic task. To alleviate this gap, we replace the syntactic dependency tree with the semantic structure named Abstract Meaning Representation (AMR) and propose a model called AMR-based Path Aggregation Relational Network (APARN) to take full advantage of semantic structures. In particular, we design the path aggregator and the relation-enhanced self-attention mechanism that complement each other. The path aggregator extracts semantic features from AMRs under the guidance of sentence information, while the relation-enhanced self-attention mechanism in turn improves sentence features with refined semantic information. Experimental results on four public datasets demonstrate 1.13% average F1 improvement of APARN in ABSA when compared with state-of-the-art baselines."
                    },
                    {
                        "title": "Gaussian Prior Reinforcement Learning for Nested Named Entity Recognition",
                        "abstract": "Named Entity Recognition (NER) is a well and widely studied task in natural language processing. Recently, the nested NER has attracted more attention since its practicality and difficulty. Existing works for nested NER ignore the recognition order and boundary position relation of nested entities. To address these issues, we propose a novel seq2seq model named GPRL, which formulates the nested NER task as an entity triplet sequence generation process. GPRL adopts the reinforcement learning method to generate entity triplets de-coupling the entity order in gold labels and expects to learn a reasonable recognition order of entities via trial and error. Based on statistics of boundary distance for nested entities, GPRL designs a Gaussian prior to represent the boundary distance distribution between nested entities and adjust the out-put probability distribution of nested boundary tokens. Experiments on three nested NER datasets demonstrate that GPRL outperforms previous nested NER models."
                    },
                    {
                        "title": "A Private Watermark for Large Language Models",
                        "abstract": "Recently, text watermarking algorithms for large language models (LLMs) have been mitigating the potential harms of text generated by the LLMs, including fake news and copyright issues. However, the watermark detection of current text algorithms requires the key from the generation process, making them susceptible to breaches and counterfeiting. In this work, we propose the first private watermarking algorithm, which extends the current text watermarking algorithms by using two different neural networks respectively for watermark generation and detection, rather than using the same key at both stages. Meanwhile, part of the parameters of the watermark generation and detection networks are shared, which makes the detection network achieve a high accuracy very efficiently. Experiments show that our algorithm ensures high detection accuracy with minimal impact on generation and detection speed, due to the small parameter size of both networks. Additionally, our subsequent analysis demonstrates the difficulty of reverting the watermark generation rules from the detection network."
                    },
                    {
                        "title": "RAPL: A Relation-Aware Prototype Learning Approach for Few-Shot Document-Level Relation Extraction",
                        "abstract": "How to identify semantic relations among entities in a document when only a few labeled documents are available? Few-shot document-level relation extraction (FSDLRE) is crucial for addressing the pervasive data scarcity problem in real-world scenarios. Metric-based meta-learning is an effective framework widely adopted for FSDLRE, which constructs class prototypes for classification. However, existing works often struggle to obtain class prototypes with accurate relational semantics: 1) To build prototype for a target relation type, they aggregate the representations of all entity pairs holding that relation, while these entity pairs may also hold other relations, thus disturbing the prototype. 2) They use a set of generic NOTA (none-of-the-above) prototypes across all tasks, neglecting that the NOTA semantics differs in tasks with different target relation types. In this paper, we propose a relation-aware prototype learning method for FSDLRE to strengthen the relational semantics of prototype representations. By judiciously leveraging the relation descriptions and realistic NOTA instances as guidance, our method effectively refines the relation prototypes and generates task-specific NOTA prototypes. Extensive experiments demonstrate that our method outperforms state-of-the-art approaches by average 2.61% $F_1$ across various settings of two FSDLRE benchmarks."
                    },
                    {
                        "title": "Variational Disentangled Graph Auto-Encoders for Link Prediction",
                        "abstract": "With the explosion of graph-structured data, link prediction has emerged as an increasingly important task. Embedding methods for link prediction utilize neural networks to generate node embeddings, which are subsequently employed to predict links between nodes. However, the existing embedding methods typically take a holistic strategy to learn node embeddings and ignore the entanglement of latent factors. As a result, entangled embeddings fail to effectively capture the underlying information and are vulnerable to irrelevant information, leading to unconvincing and uninterpretable link prediction results. To address these challenges, this paper proposes a novel framework with two variants, the disentangled graph auto-encoder (DGAE) and the variational disentangled graph auto-encoder (VDGAE). Our work provides a pioneering effort to apply the disentanglement strategy to link prediction. The proposed framework infers the latent factors that cause edges in the graph and disentangles the representation into multiple channels corresponding to unique latent factors, which contributes to improving the performance of link prediction. To further encourage the embeddings to capture mutually exclusive latent factors, we introduce mutual information regularization to enhance the independence among different channels. Extensive experiments on various real-world benchmarks demonstrate that our proposed methods achieve state-of-the-art results compared to a variety of strong baselines on link prediction tasks. Qualitative analysis on the synthetic dataset also illustrates that the proposed methods can capture distinct latent factors that cause links, providing empirical evidence that our models are able to explain the results of link prediction to some extent. All code will be made publicly available upon publication of the paper."
                    },
                    {
                        "title": "What Makes the Story Forward?: Inferring Commonsense Explanations as Prompts for Future Event Generation",
                        "abstract": "Prediction over event sequences is critical for many real-world applications in Information Retrieval and Natural Language Processing. Future Event Generation (FEG) is a challenging task in event sequence prediction because it requires not only fluent text generation but also commonsense reasoning to maintain the logical coherence of the entire event story. In this paper, we propose a novel explainable FEG framework, Coep. It highlights and integrates two types of event knowledge, sequential knowledge of direct event-event relations and inferential knowledge that reflects the intermediate character psychology between events, such as intents, causes, reactions, which intrinsically pushes the story forward. To alleviate the knowledge forgetting issue, we design two modules, IM and GM, for each type of knowledge, which are combined via prompt tuning. First, IM focuses on understanding inferential knowledge to generate commonsense explanations and provide a soft prompt vector for GM. We also design a contrastive discriminator for better generalization ability. Second, GM generates future events by modeling direct sequential knowledge with the guidance of IM. Automatic and human evaluation demonstrate that our approach can generate more coherent, specific, and logical future events."
                    },
                    {
                        "title": "HiURE: Hierarchical Exemplar Contrastive Learning for Unsupervised Relation Extraction",
                        "abstract": "Unsupervised relation extraction aims to extract the relationship between entities from natural language sentences without prior information on relational scope or distribution. Existing works either utilize self-supervised schemes to refine relational feature signals by iteratively leveraging adaptive clustering and classification that provoke gradual drift problems, or adopt instance-wise contrastive learning which unreasonably pushes apart those sentence pairs that are semantically similar. To overcome these defects, we propose a novel contrastive learning framework named HiURE, which has the capability to derive hierarchical signals from relational feature space using cross hierarchy attention and effectively optimize relation representation of sentences under exemplar-wise contrastive learning. Experimental results on two public datasets demonstrate the advanced effectiveness and robustness of HiURE on unsupervised relation extraction when compared with state-of-the-art models."
                    },
                    {
                        "title": "Character-level White-Box Adversarial Attacks against Transformers via Attachable Subwords Substitution",
                        "abstract": "We propose the first character-level white-box adversarial attack method against transformer models. The intuition of our method comes from the observation that words are split into subtokens before being fed into the transformer models and the substitution between two close subtokens has a similar effect with the character modification. Our method mainly contains three steps. First, a gradient-based method is adopted to find the most vulnerable words in the sentence. Then we split the selected words into subtokens to replace the origin tokenization result from the transformer tokenizer. Finally, we utilize an adversarial loss to guide the substitution of attachable subtokens in which the Gumbel-softmax trick is introduced to ensure gradient propagation.Meanwhile, we introduce the visual and length constraint in the optimization process to achieve minimum character modifications.Extensive experiments on both sentence-level and token-level tasks demonstrate that our method could outperform the previous attack methods in terms of success rate and edit distance. Furthermore, human evaluation verifies our adversarial examples could preserve their origin labels."
                    },
                    {
                        "title": "Pair-Level Supervised Contrastive Learning for Natural Language Inference",
                        "abstract": "Natural language inference (NLI) is an increasingly important task for natural language understanding, which requires one to infer the relationship between the sentence pair (premise and hypothesis). Many recent works have used contrastive learning by incorporating the relationship of the sentence pair from NLI datasets to learn sentence representation. However, these methods only focus on comparisons with sentence-level representations. In this paper, we propose a Pair-level Supervised Contrastive Learning approach (PairSCL). We adopt a cross attention module to learn the joint representations of the sentence pairs. A contrastive learning objective is designed to distinguish the varied classes of sentence pairs by pulling those in one class together and pushing apart the pairs in other classes. We evaluate PairSCL on two public datasets of NLI where the accuracy of PairSCL outperforms other methods by 2.1% on average. Furthermore, our method outperforms the previous state-of-the-art method on seven transfer tasks of text classification.1"
                    },
                    {
                        "title": "A Multi-Level Supervised Contrastive Learning Framework for Low-Resource Natural Language Inference",
                        "abstract": "Natural Language Inference (NLI) is a growingly essential task in natural language understanding, which requires inferring the relationship between the sentence pairs (premise and hypothesis). Recently, low-resource natural language inference has gained increasing attention, due to significant savings in manual annotation costs and a better fit with real-world scenarios. Existing works fail to characterize discriminative representations between different classes with limited training data, which may cause faults in label prediction. Here we propose a multi-level supervised contrastive learning framework named MultiSCL for low-resource natural language inference. MultiSCL leverages a sentence-level and pair-level contrastive learning objective to discriminate between different classes of sentence pairs by bringing those in one class together and pushing away those in different classes. MultiSCL adopts a data augmentation module that generates different views for input samples to better learn the latent representation. The pair-level representation is obtained from a cross attention module. We conduct extensive experiments on two public NLI datasets in low-resource settings, and the accuracy of MultiSCL exceeds other models by 1.8%, 3.1% and 4.1% on SNLI, MNLI and Sick with 5 instances per label respectively. Moreover, our method outperforms the previous state-of-the-art method on cross-domain tasks of text classification."
                    }
                ]
            },
            "f583cb06-6864-40c5-83e3-fcd838367746": {
                "pk": "f583cb06-6864-40c5-83e3-fcd838367746",
                "name": "Lijie Wen",
                "collaborators": [
                    "Xuming Hu",
                    "Philip S. Yu",
                    "Aiwei Liu",
                    "Junzhe Chen",
                    "Zhijiang Guo",
                    "Shiao Meng",
                    "Shuang Li",
                    "Leyi Pan",
                    "Irwin King",
                    "Xiaochuan Li",
                    "Fukun Ma",
                    "Yawen Yang",
                    "Yijian Lu",
                    "Yinghui Li",
                    "Yangning Li",
                    "Xiaoguang Li",
                    "Yasheng Wang",
                    "Qun Liu",
                    "Zhaochen Hong",
                    "Zhiwei He",
                    "Zitian Gao",
                    "Xuandong Zhao",
                    "Binglin Zhou",
                    "Shuliang Liu",
                    "Hkustgz",
                    "Chenwei Zhang",
                    "Jiahua Wang",
                    "Shuodi Liu",
                    "Shiyu Huang",
                    "Weijuan Tu",
                    "Zhaofeng He",
                    "Guan-Huei Wu",
                    "Jingjing Li",
                    "W. Liu",
                    "Yufei Guo"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Information Retrieval",
                    "Multi-modal Learning"
                ],
                "publications": [
                    {
                        "title": "MarkLLM: An Open-Source Toolkit for LLM Watermarking",
                        "abstract": "LLM watermarking, which embeds imperceptible yet algorithmically detectable signals in model outputs to identify LLM-generated text, has become crucial in mitigating the potential misuse of large language models. However, the abundance of LLM watermarking algorithms, their intricate mechanisms, and the complex evaluation procedures and perspectives pose challenges for researchers and the community to easily experiment with, understand, and assess the latest advancements. To address these issues, we introduce MarkLLM, an open-source toolkit for LLM watermarking. MarkLLM offers a unified and extensible framework for implementing LLM watermarking algorithms, while providing user-friendly interfaces to ensure ease of access. Furthermore, it enhances understanding by supporting automatic visualization of the underlying mechanisms of these algorithms. For evaluation, MarkLLM offers a comprehensive suite of 12 tools spanning three perspectives, along with two types of automated evaluation pipelines. Through MarkLLM, we aim to support researchers while improving the comprehension and involvement of the general public in LLM watermarking technology, fostering consensus and driving further advancements in research and application. Our code is available at https://github.com/THU-BPM/MarkLLM."
                    },
                    {
                        "title": "Evaluating Robustness of Generative Search Engine on Adversarial Factoid Questions",
                        "abstract": "Generative search engines have the potential to transform how people seek information on-line, but generated responses from existing large language models (LLMs)-backed generative search engines may not always be accurate. Nonetheless, retrieval-augmented generation exacerbates safety concerns, since adversaries may successfully evade the entire sys-tem by subtly manipulating the most vulnerable part of a claim. To this end, we pro-pose evaluating the robustness of generative search engines in the realistic and high-risk setting, where adversaries have only black-box system access and seek to deceive the model into returning incorrect responses. Through a comprehensive human evaluation of various generative search engines, such as Bing Chat, PerplexityAI, and YouChat across diverse queries, we demonstrate the effectiveness of adversarial factual questions in inducing incorrect responses. Moreover, retrieval-augmented generation exhibits a higher susceptibility to factual errors compared to LLMs without retrieval. These findings highlight the potential security risks of these systems and emphasize the need for rigorous evaluation before deployment. Our constructed dataset and codes are available at: https://github.com/HKUSTGZ-NLP/Adversarial-Attack."
                    },
                    {
                        "title": "Evaluating Robustness of Generative Search Engine on Adversarial Factual Questions",
                        "abstract": "Generative search engines have the potential to transform how people seek information online, but generated responses from existing large language models (LLMs)-backed generative search engines may not always be accurate. Nonetheless, retrieval-augmented generation exacerbates safety concerns, since adversaries may successfully evade the entire system by subtly manipulating the most vulnerable part of a claim. To this end, we propose evaluating the robustness of generative search engines in the realistic and high-risk setting, where adversaries have only black-box system access and seek to deceive the model into returning incorrect responses. Through a comprehensive human evaluation of various generative search engines, such as Bing Chat, PerplexityAI, and YouChat across diverse queries, we demonstrate the effectiveness of adversarial factual questions in inducing incorrect responses. Moreover, retrieval-augmented generation exhibits a higher susceptibility to factual errors compared to LLMs without retrieval. These findings highlight the potential security risks of these systems and emphasize the need for rigorous evaluation before deployment."
                    },
                    {
                        "title": "Reading Broadly to Open Your Mind: Improving Open Relation Extraction With Search Documents Under Self-Supervisions",
                        "abstract": "Open relation extraction is the task of extracting open-domain relation facts from natural language sentences. Existing works either utilize distant-supervised annotations to train a supervised classifier over pre-defined relations, or adopt unsupervised methods with additional dependency on external assumptions. However, these works can only obtain information signals from limited existing knowledge bases or datasets. In this work, we propose a self-supervised framework named Web-SelfORE, which exploits self-supervised signals by requiring a large pretrained language model to extensively read real-world relevant documents from the web, and obtain contextualized relational features by mixing contextualized representations of entities from different documents. We perform adaptive clustering on contextualized relational features and bootstrap the self-supervised signals by improving contextualized features in relation classification. We additionally compare the effectiveness of self-supervisions brought by different document sources, and introduce relevance and redundancy evaluation metrics to obtain higher-quality self-supervisions. Experimental results on four public datasets show the effectiveness and robustness of Web-SelfORE on open-domain relation extraction task when comparing with competitive baselines."
                    },
                    {
                        "title": "FSMR: A Feature Swapping Multi-modal Reasoning Approach with Joint Textual and Visual Clues",
                        "abstract": "Multi-modal reasoning plays a vital role in bridging the gap between textual and visual information, enabling a deeper understanding of the context. This paper presents the Feature Swapping Multi-modal Reasoning (FSMR) model, designed to enhance multi-modal reasoning through feature swapping. FSMR leverages a pre-trained visual-language model as an encoder, accommodating both text and image inputs for effective feature representation from both modalities. It introduces a unique feature swapping module, enabling the exchange of features between identified objects in images and corresponding vocabulary words in text, thereby enhancing the model's comprehension of the interplay between images and text. To further bolster its multi-modal alignment capabilities, FSMR incorporates a multi-modal cross-attention mechanism, facilitating the joint modeling of textual and visual information. During training, we employ image-text matching and cross-entropy losses to ensure semantic consistency between visual and language elements. Extensive experiments on the PMR dataset demonstrate FSMR's superiority over state-of-the-art baseline models across various performance metrics."
                    },
                    {
                        "title": "LLMArena: Assessing Capabilities of Large Language Models in Dynamic Multi-Agent Environments",
                        "abstract": "Recent advancements in large language models (LLMs) have revealed their potential for achieving autonomous agents possessing human-level intelligence. However, existing benchmarks for evaluating LLM Agents either use static datasets, potentially leading to data leakage or focus only on single-agent scenarios, overlooking the complexities of multi-agent interactions. There is a lack of a benchmark that evaluates the diverse capabilities of LLM agents in multi-agent, dynamic environments. To this end, we introduce LLMArena, a novel and easily extensible framework for evaluating the diverse capabilities of LLM in multi-agent dynamic environments. LLMArena encompasses seven distinct gaming environments, employing Trueskill scoring to assess crucial abilities in LLM agents, including spatial reasoning, strategic planning, numerical reasoning, risk assessment, communication, opponent modeling, and team collaboration. We conduct an extensive experiment and human evaluation among different sizes and types of LLMs, showing that LLMs still have a significant journey ahead in their development towards becoming fully autonomous agents, especially in opponent modeling and team collaboration. We hope LLMArena could guide future research towards enhancing these capabilities in LLMs, ultimately leading to more sophisticated and practical applications in dynamic, multi-agent settings. The code and data will be available."
                    },
                    {
                        "title": "On the Robustness of Document-Level Relation Extraction Models to Entity Name Variations",
                        "abstract": "Driven by the demand for cross-sentence and large-scale relation extraction, document-level relation extraction (DocRE) has attracted increasing research interest. Despite the continuous improvement in performance, we find that existing DocRE models which initially perform well may make more mistakes when merely changing the entity names in the document, hindering the generalization to novel entity names. To this end, we systematically investigate the robustness of DocRE models to entity name variations in this work. We first propose a principled pipeline to generate entity-renamed documents by replacing the original entity names with names from Wikidata. By applying the pipeline to DocRED and Re-DocRED datasets, we construct two novel benchmarks named Env-DocRED and Env-Re-DocRED for robustness evaluation. Experimental results show that both three representative DocRE models and two in-context learned large language models consistently lack sufficient robustness to entity name variations, particularly on cross-sentence relation instances and documents with more entities. Finally, we propose an entity variation robust training method which not only improves the robustness of DocRE models but also enhances their understanding and reasoning capabilities. We further verify that the basic idea of this method can be effectively transferred to in-context learning for DocRE as well."
                    },
                    {
                        "title": "SelfLRE: Self-refining Representation Learning for Low-resource Relation Extraction",
                        "abstract": "Low-resource relation extraction (LRE) aims to extract potential relations from limited labeled corpus to handle the problem of scarcity of human annotations. Previous works mainly consist of two categories of methods: (1) Self-training methods, which improve themselves through the models' predictions, thus suffering from confirmation bias when the predictions are wrong. (2) Self-ensembling methods, which learn task-agnostic representations, therefore, generally do not work well for specific tasks. In our work, we propose a novel LRE architecture named SelfLRE, which leverages two complementary modules, one module uses self-training to obtain pseudo-labels for unlabeled data, and the other module uses self-ensembling learning to obtain the task-agnostic representations, and leverages the existing pseudo-labels to refine the better task-specific representations on unlabeled data. The two models are jointly trained through multi-task learning to iteratively improve the effect of LRE task. Experiments on three public datasets show that SelfLRE achieves 1.81% performance gain over the SOTA baseline. Source code is available at: https://github.com/THU-BPM/SelfLRE."
                    },
                    {
                        "title": "Give Me More Details: Improving Fact-Checking with Latent Retrieval",
                        "abstract": "Evidence plays a crucial role in automated fact-checking. When verifying real-world claims, existing fact-checking systems either assume the evidence sentences are given or use the search snippets returned by the search engine. Such methods ignore the challenges of collecting evidence and may not provide sufficient information to verify real-world claims. Aiming at building a better fact-checking system, we propose to incorporate full text from source documents as evidence and introduce two enriched datasets. The first one is a multilingual dataset, while the second one is monolingual (English). We further develop a latent variable model to jointly extract evidence sentences from documents and perform claim verification. Experiments indicate that including source documents can provide sufficient contextual clues even when gold evidence sentences are not annotated. The proposed system is able to achieve significant improvements upon best-reported models under different settings."
                    },
                    {
                        "title": "A Survey of Text Watermarking in the Era of Large Language Models",
                        "abstract": "Text watermarking algorithms are crucial for protecting the copyright of textual content. Historically, their capabilities and application scenarios were limited. However, recent advancements in large language models (LLMs) have revolutionized these techniques. LLMs not only enhance text watermarking algorithms with their advanced abilities but also create a need for employing these algorithms to protect their own copyrights or prevent potential misuse. This paper conducts a comprehensive survey of the current state of text watermarking technology, covering four main aspects: (1) an overview and comparison of different text watermarking techniques; (2) evaluation methods for text watermarking algorithms, including their detectability, impact on text or LLM quality, robustness under target or untargeted attacks; (3) potential application scenarios for text watermarking technology; (4) current challenges and future directions for text watermarking. This survey aims to provide researchers with a thorough understanding of text watermarking technology in the era of LLM, thereby promoting its further advancement."
                    },
                    {
                        "title": "Exploring the Compositional Generalization in Context Dependent Text-to-SQL Parsing",
                        "abstract": "In the context-dependent Text-to-SQL task, the generated SQL statements are refined iteratively based on the user input utterance from each interaction. The input text from each interaction can be viewed as component modifications to the previous SQL statements, which could be further extracted as the modification patterns. Since these modification patterns could also be combined with other SQL statements, the models are supposed to have the compositional generalization to these novel combinations. This work is the first exploration of compositional generalization in context-dependent Text-to-SQL scenarios. To facilitate related studies, we constructed two challenging benchmarks named \\textsc{CoSQL-CG} and \\textsc{SParC-CG} by recombining the modification patterns and existing SQL statements. The following experiments show that all current models struggle on our proposed benchmarks. Furthermore, we found that better aligning the previous SQL statements with the input utterance could give models better compositional generalization ability. Based on these observations, we propose a method named \\texttt{p-align} to improve the compositional generalization of Text-to-SQL models. Further experiments validate the effectiveness of our method. Source code and data are available."
                    },
                    {
                        "title": "A comprehensive evaluation of ChatGPT's zero-shot Text-to-SQL capability",
                        "abstract": "This paper presents the first comprehensive analysis of ChatGPT's Text-to-SQL ability. Given the recent emergence of large-scale conversational language model ChatGPT and its impressive capabilities in both conversational abilities and code generation, we sought to evaluate its Text-to-SQL performance. We conducted experiments on 12 benchmark datasets with different languages, settings, or scenarios, and the results demonstrate that ChatGPT has strong text-to-SQL abilities. Although there is still a gap from the current state-of-the-art (SOTA) model performance, considering that the experiment was conducted in a zero-shot scenario, ChatGPT's performance is still impressive. Notably, in the ADVETA (RPL) scenario, the zero-shot ChatGPT even outperforms the SOTA model that requires fine-tuning on the Spider dataset by 4.1\\%, demonstrating its potential for use in practical applications. To support further research in related fields, we have made the data generated by ChatGPT publicly available at https://github.com/THU-BPM/chatgpt-sql."
                    },
                    {
                        "title": "Prompt Me Up: Unleashing the Power of Alignments for Multimodal Entity and Relation Extraction",
                        "abstract": "How can we better extract entities and relations from text? Using multimodal extraction with images and text obtains more signals for entities and relations, and aligns them through graphs or hierarchical fusion, aiding in extraction. Despite attempts at various fusions, previous works have overlooked many unlabeled image-caption pairs, such as NewsCLIPing. This paper proposes innovative pre-training objectives for entity-object and relation-image alignment, extracting objects from images and aligning them with entity and relation prompts for soft pseudo-labels. These labels are used as self-supervised signals for pre-training, enhancing the ability to extract entities and relations. Experiments on three datasets show an average 3.41% F1 improvement over prior SOTA. Additionally, our method is orthogonal to previous multimodal fusions, and using it on prior SOTA fusions further improves 5.47% F1."
                    },
                    {
                        "title": "MR2: A Benchmark for Multimodal Retrieval-Augmented Rumor Detection in Social Media",
                        "abstract": "As social media platforms are evolving from text-based forums into multi-modal environments, the nature of misinformation in social media is also transforming accordingly. Misinformation spreaders have recently targeted contextual connections between the modalities e.g., text and image. However, existing datasets for rumor detection mainly focus on a single modality i.e., text. To bridge this gap, we construct MR2, a multimodal multilingual retrieval-augmented dataset for rumor detection. The dataset covers rumors with images and texts, and provides evidence from both modalities that are retrieved from the Internet. Further, we develop established baselines and conduct a detailed analysis of the systems evaluated on the dataset. Extensive experiments show that MR2 will provide a challenging testbed for developing rumor detection systems designed to retrieve and reason over social media posts. Source code and data are available at: https://github.com/THU-BPM/MR2."
                    },
                    {
                        "title": "Enhancing Cross-lingual Natural Language Inference by Soft Prompting with Multilingual Verbalizer",
                        "abstract": "Cross-lingual natural language inference is a fundamental problem in cross-lingual language understanding. Many recent works have used prompt learning to address the lack of annotated parallel corpora in XNLI. However, these methods adopt discrete prompting by simply translating the templates to the target language and need external expert knowledge to design the templates. Besides, discrete prompts of human-designed template words are not trainable vectors and can not be migrated to target languages in the inference stage flexibly. In this paper, we propose a novel Soft prompt learning framework with the Multilingual Verbalizer (SoftMV) for XNLI. SoftMV first constructs cloze-style question with soft prompts for the input sample. Then we leverage bilingual dictionaries to generate an augmented multilingual question for the original question. SoftMV adopts a multilingual verbalizer to align the representations of original and augmented multilingual questions into the same semantic space with consistency regularization. Experimental results on XNLI demonstrate that SoftMV can achieve state-of-the-art performance and significantly outperform the previous methods under the few-shot and full-shot cross-lingual transfer settings."
                    },
                    {
                        "title": "A Semantic Invariant Robust Watermark for Large Language Models",
                        "abstract": "Watermark algorithms for large language models (LLMs) have achieved extremely high accuracy in detecting text generated by LLMs. Such algorithms typically involve adding extra watermark logits to the LLM's logits at each generation step. However, prior algorithms face a trade-off between attack robustness and security robustness. This is because the watermark logits for a token are determined by a certain number of preceding tokens; a small number leads to low security robustness, while a large number results in insufficient attack robustness. In this work, we propose a semantic invariant watermarking method for LLMs that provides both attack robustness and security robustness. The watermark logits in our work are determined by the semantics of all preceding tokens. Specifically, we utilize another embedding LLM to generate semantic embeddings for all preceding tokens, and then these semantic embeddings are transformed into the watermark logits through our trained watermark model. Subsequent analyses and experiments demonstrated the attack robustness of our method in semantically invariant settings: synonym substitution and text paraphrasing settings. Finally, we also show that our watermark possesses adequate security robustness. Our code and data are available at \\href{https://github.com/THU-BPM/Robust_Watermark}{https://github.com/THU-BPM/Robust\\_Watermark}. Additionally, our algorithm could also be accessed through MarkLLM \\citep{pan2024markllm} \\footnote{https://github.com/THU-BPM/MarkLLM}."
                    },
                    {
                        "title": "Do Large Language Models Know about Facts?",
                        "abstract": "Large language models (LLMs) have recently driven striking performance improvements across a range of natural language processing tasks. The factual knowledge acquired during pretraining and instruction tuning can be useful in various downstream tasks, such as question answering, and language generation. Unlike conventional Knowledge Bases (KBs) that explicitly store factual knowledge, LLMs implicitly store facts in their parameters. Content generated by the LLMs can often exhibit inaccuracies or deviations from the truth, due to facts that can be incorrectly induced or become obsolete over time. To this end, we aim to comprehensively evaluate the extent and scope of factual knowledge within LLMs by designing the benchmark Pinocchio. Pinocchio contains 20K diverse factual questions that span different sources, timelines, domains, regions, and languages. Furthermore, we investigate whether LLMs are able to compose multiple facts, update factual knowledge temporally, reason over multiple pieces of facts, identify subtle factual differences, and resist adversarial examples. Extensive experiments on different sizes and types of LLMs show that existing LLMs still lack factual knowledge and suffer from various spurious correlations. We believe this is a critical bottleneck for realizing trustworthy artificial intelligence. The dataset Pinocchio and our codes will be publicly available."
                    },
                    {
                        "title": "Read it Twice: Towards Faithfully Interpretable Fact Verification by Revisiting Evidence",
                        "abstract": "Real-world fact verification task aims to verify the factuality of a claim by retrieving evidence from the source document. The quality of the retrieved evidence plays an important role in claim verification. Ideally, the retrieved evidence should be faithful (reflecting the model's decision-making process in claim verification) and plausible (convincing to humans), and can improve the accuracy of verification task. Although existing approaches leverage the similarity measure of semantic or surface form between claims and documents to retrieve evidence, they all rely on certain heuristics that prevent them from satisfying all three requirements. In light of this, we propose a fact verification model named ReRead to retrieve evidence and verify claim that: (1) Train the evidence retriever to obtain interpretable evidence (i.e., faithfulness and plausibility criteria); (2) Train the claim verifier to revisit the evidence retrieved by the optimized evidence retriever to improve the accuracy. The proposed system is able to achieve significant improvements upon best-reported models under different settings."
                    }
                ]
            },
            "eddf8fdc-3057-478b-ae30-d56b76c80ac0": {
                "pk": "eddf8fdc-3057-478b-ae30-d56b76c80ac0",
                "name": "Irwin King",
                "collaborators": [
                    "Yankai Chen",
                    "Yifei Zhang",
                    "Menglin Yang",
                    "Ziqiao Meng",
                    "Peilin Zhao",
                    "Yang Yu",
                    "Zixing Song",
                    "Xuming Hu",
                    "Chenwei Zhang",
                    "Philip S. Yu",
                    "Min Zhou",
                    "Chenshuo Ma",
                    "Aiwei Liu",
                    "Zeqi Tan",
                    "Xin Zhang",
                    "Quoc-Tuan Truong",
                    "Xin Shen",
                    "Ming Wang",
                    "Jin Li",
                    "Jim Chan",
                    "Yueen Ma",
                    "Dafeng Chi",
                    "Jingjing Li",
                    "Yuzheng Zhuang",
                    "Jianye Hao",
                    "Zhaochen Hong",
                    "Xinyu Fu",
                    "Zhihang Hu",
                    "Qinze Yu",
                    "Yucheng Guo",
                    "Taifeng Wang",
                    "Xin Gao",
                    "Le Song",
                    "Yu Li",
                    "Bo Xiong",
                    "H. Xiong",
                    "Dun Zeng",
                    "Jinglong Luo",
                    "Zenglin Xu",
                    "Ming Jin",
                    "Huan Yee Koh",
                    "Qingsong Wen",
                    "Daniele Zambon",
                    "C. Alippi",
                    "G. I. Webb",
                    "Shirui Pan",
                    "Rex Ying",
                    "Conghao Xiong",
                    "Hao Chen",
                    "Joseph J. Y. Sung"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Drug Discovery",
                    "Natural Language Processing",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "Doubly Stochastic Graph-based Non-autoregressive Reaction Prediction",
                        "abstract": "Organic reaction prediction is a critical task in drug discovery. Recently, researchers have achieved non-autoregressive reaction prediction by modeling the redistribution of electrons, resulting in state-of-the-art top-1 accuracy, and enabling parallel sampling. However, the current non-autoregressive decoder does not satisfy two essential rules of electron redistribution modeling simultaneously: the electron-counting rule and the symmetry rule. This violation of the physical constraints of chemical reactions impairs model performance. In this work, we propose a new framework called ReactionSink that combines two doubly stochastic self-attention mappings to obtain electron redistribution predictions that follow both constraints. We further extend our solution to a general multi-head attention mechanism with augmented constraints. To achieve this, we apply Sinkhorn's algorithm to iteratively update self-attention mappings, which imposes doubly conservative constraints as additional informative priors on electron redistribution modeling. We theoretically demonstrate that our ReactionSink can simultaneously satisfy both rules, which the current decoder mechanism cannot do. Empirical results show that our approach consistently improves the predictive performance of non-autoregressive models and does not bring an unbearable additional computational cost."
                    },
                    {
                        "title": "WSFE: Wasserstein Sub-graph Feature Encoder for Effective User Segmentation in Collaborative Filtering",
                        "abstract": "Maximizing the user-item engagement based on vectorized embeddings is a standard procedure of recent recommender models. Despite the superior performance for item recommendations, these methods however implicitly deprioritize the modeling of user-wise similarity in the embedding space; consequently, identifying similar users is underperforming, and additional processing schemes are usually required otherwise. To avoid thorough model re-training, we propose WSFE, a model-agnostic and training-free representation encoder, to be flexibly employed on the fly for effective user segmentation. Underpinned by the optimal transport theory, the encoded representations from WSFE present a matched user-wise similarity/distance measurement between the realistic and embedding space. We incorporate WSFE into six state-of-the-art recommender models and conduct extensive experiments on six real-world datasets. The empirical analyses well demonstrate the superiority and generality of WSFE to fuel multiple downstream tasks with diverse underlying targets in recommendation."
                    },
                    {
                        "title": "GDA: Generative Data Augmentation Techniques for Relation Extraction Tasks",
                        "abstract": "Relation extraction (RE) tasks show promising performance in extracting relations from two entities mentioned in sentences, given sufficient annotations available during training. Such annotations would be labor-intensive to obtain in practice. Existing work adopts data augmentation techniques to generate pseudo-annotated sentences beyond limited annotations. These techniques neither preserve the semantic consistency of the original sentences when rule-based augmentations are adopted, nor preserve the syntax structure of sentences when expressing relations using seq2seq models, resulting in less diverse augmentations. In this work, we propose a dedicated augmentation technique for relational texts, named GDA, which uses two complementary modules to preserve both semantic consistency and syntax structures. We adopt a generative formulation and design a multi-tasking solution to achieve synergies. Furthermore, GDA adopts entity hints as the prior knowledge of the generative model to augment diverse sentences. Experimental results in three datasets under a low-resource setting showed that GDA could bring {\\em 2.0\\%} F1 improvements compared with no augmentation technique. Source code and data are available."
                    },
                    {
                        "title": "Topological Representation Learning for E-commerce Shopping Behaviors",
                        "abstract": "Learning compact representation from customer shopping behaviors is at the core of web-scale E-commerce recommender systems. At Amazon, we put great efforts into learning embedding of customer engagements in order to fuel multiple downstream tasks for better recommendation services. In this work, we define the notion of shopping trajectory that consists of customer interactions at the categorical level of products, then construct an end-to-end model namely C-STAR which is capable of learning rich embedding for representing the variable-length customer trajectory. C-STAR explicitly captures the trajectory distribution similarity and trajectory topological semantics , providing a coarse-to-fine trajectory representation learning paradigm both structurally and semantically. We evaluate the model on Amazon proprietary data as well as four public datasets, where the learned embeddings have shown to be effective for customer-centric tasks including customer segmentation and shopping trajectory completion ."
                    },
                    {
                        "title": "Contrastive Cross-scale Graph Knowledge Synergy",
                        "abstract": "Graph representation learning via Contrastive Learning (GCL) has drawn considerable attention recently. Efforts are mainly focused on gathering more global information via contrasting on a single high-level graph view, which, however, underestimates the inherent complex and hierarchical properties in many real-world networks, leading to sub-optimal embeddings. To incorporate these properties of a complex graph, we propose Cross-Scale Contrastive Graph Knowledge Synergy (CGKS), a generic feature learning framework, to advance graph contrastive learning with enhanced generalization ability and the awareness of latent anatomies. Specifically, to maintain the hierarchical information, we create a so-call graph pyramid (GP) consisting of coarse-grained graph views. Each graph view is obtained via the careful design topology-aware graph coarsening layer that extends the Laplacian Eigenmaps with negative sampling. To promote cross-scale information sharing and knowledge interactions among GP, we propose a novel joint optimization formula that contains a pairwise contrastive loss between any two coarse-grained graph views. This synergy loss not only promotes knowledge sharing that yields informative representations, but also stabilizes the training process. Experiments on various downstream tasks demonstrate the substantial improvements of the proposed method over its counterparts."
                    },
                    {
                        "title": "VOLTA: Improving Generative Diversity by Variational Mutual Information Maximizing Autoencoder",
                        "abstract": "The natural language generation domain has witnessed great success thanks to Transformer models. Although they have achieved state-of-the-art generative quality, they often neglect generative diversity. Prior attempts to tackle this issue suffer from either low model capacity or over-complicated architectures. Some recent methods employ the VAE framework to enhance diversity, but their latent variables fully depend on the input context, restricting exploration of the latent space. In this paper, we introduce VOLTA, a framework that elevates generative diversity by bridging Transformer with VAE via a more effective cross-attention-based connection, departing from conventional embedding concatenation or summation. Additionally, we propose integrating InfoGAN-style latent codes to enable input-independent variability, further diversifying the generation. Moreover, our framework accommodates discrete inputs alongside its existing support for continuous inputs. We perform comprehensive experiments with two types of Transformers on six datasets from three different NLG tasks to show that our approach can significantly improve generative diversity while maintaining generative quality."
                    },
                    {
                        "title": "Think Rationally about What You See: Continuous Rationale Extraction for Relation Extraction",
                        "abstract": "Relation extraction (RE) aims to extract potential relations according to the context of two entities, thus, deriving rational contexts from sentences plays an important role. Previous works either focus on how to leverage the entity information (e.g., entity types, entity verbalization) to inference relations, but ignore context-focused content, or use counterfactual thinking to remove the model's bias of potential relations in entities, but the relation reasoning process will still be hindered by irrelevant content. Therefore, how to preserve relevant content and remove noisy segments from sentences is a crucial task. In addition, retained content needs to be fluent enough to maintain semantic coherence and interpretability. In this work, we propose a novel rationale extraction framework named RE2, which leverages two continuity and sparsity factors to obtain relevant and coherent rationales from sentences. To solve the problem that the gold rationales are not labeled, RE2 applies an optimizable binary mask to each token in the sentence, and adjust the rationales that need to be selected according to the relation label. Experiments on four datasets show that RE2 surpasses baselines."
                    },
                    {
                        "title": "FedHGN: A Federated Framework for Heterogeneous Graph Neural Networks",
                        "abstract": "Heterogeneous graph neural networks (HGNNs) can learn from typed and relational graph data more effectively than conventional GNNs. With larger parameter spaces, HGNNs may require more training data, which is often scarce in real-world applications due to privacy regulations (e.g., GDPR). Federated graph learning (FGL) enables multiple clients to train a GNN collaboratively without sharing their local data. However, existing FGL methods mainly focus on homogeneous GNNs or knowledge graph embeddings; few have considered heterogeneous graphs and HGNNs. In federated heterogeneous graph learning, clients may have private graph schemas. Conventional FL/FGL methods attempting to define a global HGNN model would violate schema privacy. To address these challenges, we propose FedHGN, a novel and general FGL framework for HGNNs. FedHGN adopts schema-weight decoupling to enable schema-agnostic knowledge sharing and employs coefficients alignment to stabilize the training process and improve HGNN performance. With better privacy preservation, FedHGN consistently outperforms local training and conventional FL methods on three widely adopted heterogeneous graph datasets with varying client numbers. The code is available at https://github.com/cynricfu/FedHGN."
                    },
                    {
                        "title": "Drug Synergistic Combinations Predictions via Large-Scale Pre-Training and Graph Structure Learning",
                        "abstract": "Drug combination therapy is a well-established strategy for disease treatment with better effectiveness and less safety degradation. However, identifying novel drug combinations through wet-lab experiments is resource intensive due to the vast combinatorial search space. Recently, computational approaches, specifically deep learning models have emerged as an efficient way to discover synergistic combinations. While previous methods reported fair performance, their models usually do not take advantage of multi-modal data and they are unable to handle new drugs or cell lines. In this study, we collected data from various datasets covering various drug-related aspects. Then, we take advantage of large-scale pre-training models to generate informative representations and features for drugs, proteins, and diseases. Based on that, a message-passing graph is built on top to propagate information together with graph structure learning flexibility. This is first introduced in the biological networks and enables us to generate pseudo-relations in the graph. Our framework achieves state-of-the-art results in comparison with other deep learning-based methods on synergistic prediction benchmark datasets. We are also capable of inferencing new drug combination data in a test on an independent set released by AstraZeneca, where 10% of improvement over previous methods is observed. In addition, we're robust against unseen drugs and surpass almost 15% AU ROC compared to the second-best model. We believe our framework contributes to both the future wet-lab discovery of novel drugs and the building of promising guidance for precise combination medicine."
                    },
                    {
                        "title": "Hyperbolic Graph Neural Networks: A Tutorial on Methods and Applications",
                        "abstract": "Graph Neural Networks (GNNs) generalize conventional neural networks to graph-structured data and have received considerable attention owing to their impressive performance. In spite of the notable successes, the performance of Euclidean models is inherently bounded and limited by the representation ability of Euclidean geometry, especially when it comes to datasets with highly non-Euclidean latent anatomy. Recently, hyperbolic spaces have emerged as a promising alternative for processing graph data with tree-like structure or power-law distribution and a surge of works on either methods or novel applications have been seen. Unlike Euclidean space, which expands polynomially, hyperbolic space grows exponentially with its radius, making it more suitable for modeling complex real-world data. Hence, it gains natural advantages in abstracting tree-like graphs with a hierarchical organization or power-law distribution. To support the burgeoning interest in Hyperbolic Graph Neural Networks (HGNNs), the primary goal of this tutorial is to give a systematical review of the methods, applications, and challenges in this fast-growing and vibrant area, with the express purpose of being accessible to all audiences.More specifically, we will first give a brief introduction to graph neural networks as well as some preliminary of Riemannian manifold and hyperbolic geometry. We then will comprehensively revisit the technical details of the developed HGNNs, by unifying them into a general framework and summarizing the variants of each component. Besides, we will introduce applications deployed in a variety of fields. Finally, we will discuss several challenges and present the potential solutions to address them, including some initial attempts of our own, which potentially paves the path for the further flourishing of the research community."
                    },
                    {
                        "title": "A Survey of Trustworthy Federated Learning with Perspectives on Security, Robustness and Privacy",
                        "abstract": "Trustworthy artificial intelligence (AI) technology has revolutionized daily life and greatly benefited human society. Among various AI technologies, Federated Learning (FL) stands out as a promising solution for diverse real-world scenarios, ranging from risk evaluation systems in finance to cutting-edge technologies like drug discovery in life sciences. However, challenges around data isolation and privacy threaten the trustworthiness of FL systems. Adversarial attacks against data privacy, learning algorithm stability, and system confidentiality are particularly concerning in the context of distributed training in federated learning. Therefore, it is crucial to develop FL in a trustworthy manner, with a focus on robustness and privacy. In this survey, we propose a comprehensive roadmap for developing trustworthy FL systems and summarize existing efforts from two key aspects: robustness and privacy. We outline the threats that pose vulnerabilities to trustworthy federated learning across different stages of development, including data processing, model training, and deployment. To guide the selection of the most appropriate defense methods, we discuss specific technical solutions for realizing each aspect of Trustworthy FL (TFL). Our approach differs from previous work that primarily discusses TFL from a legal perspective or presents FL from a high-level, non-technical viewpoint."
                    },
                    {
                        "title": "A Unified View of Deep Learning for Reaction and Retrosynthesis Prediction: Current Status and Future Challenges",
                        "abstract": "Reaction and retrosynthesis prediction are two fundamental tasks in computational chemistry. In recent years, these two tasks have attracted great attentions from both machine learning and drug discovery communities. Various deep learning approaches have been proposed to tackle these two problems and achieved initial success. In this survey, we conduct a comprehensive investigation on advanced deep learning-based reaction and retrosynthesis prediction models. We first summarize the design mechanism, strengths and weaknesses of the state-of-the-art approaches. Then we further discuss limitations of current solutions and open challenges in the problem itself. Last but not the least, we present some promising directions to facilitate future research. To our best knowledge, this paper is the first comprehensive and systematic survey on unified understanding of reaction and retrosynthesis prediction."
                    },
                    {
                        "title": "A Survey on Graph Neural Networks for Time Series: Forecasting, Classification, Imputation, and Anomaly Detection",
                        "abstract": "Time series are the primary data type used to record dynamic system measurements and generated in great volume by both physical sensors and online processes (virtual sensors). Time series analytics is therefore crucial to unlocking the wealth of information implicit in available data. With the recent advancements in graph neural networks (GNNs), there has been a surge in GNN-based approaches for time series analysis. These approaches can explicitly model inter-temporal and inter-variable relationships, which traditional and other deep neural network-based methods struggle to do. In this survey, we provide a comprehensive review of graph neural networks for time series analysis (GNN4TS), encompassing four fundamental dimensions: forecasting, classification, anomaly detection, and imputation. Our aim is to guide designers and practitioners to understand, build applications, and advance research of GNN4TS. At first, we provide a comprehensive task-oriented taxonomy of GNN4TS. Then, we present and discuss representative research works and introduce mainstream applications of GNN4TS. A comprehensive discussion of potential future research directions completes the survey. This survey, for the first time, brings together a vast array of knowledge on GNN-based time series research, highlighting foundations, practical applications, and opportunities of graph neural networks for time series analysis."
                    },
                    {
                        "title": "Hyperbolic Representation Learning: Revisiting and Advancing",
                        "abstract": "The non-Euclidean geometry of hyperbolic spaces has recently garnered considerable attention in the realm of representation learning. Current endeavors in hyperbolic representation largely presuppose that the underlying hierarchies can be automatically inferred and preserved through the adaptive optimization process. This assumption, however, is questionable and requires further validation. In this work, we first introduce a position-tracking mechanism to scrutinize existing prevalent \\hlms, revealing that the learned representations are sub-optimal and unsatisfactory. To address this, we propose a simple yet effective method, hyperbolic informed embedding (HIE), by incorporating cost-free hierarchical information deduced from the hyperbolic distance of the node to origin (i.e., induced hyperbolic norm) to advance existing \\hlms. The proposed method HIE is both task-agnostic and model-agnostic, enabling its seamless integration with a broad spectrum of models and tasks. Extensive experiments across various models and different tasks demonstrate the versatility and adaptability of the proposed method. Remarkably, our method achieves a remarkable improvement of up to 21.4\\% compared to the competing baselines."
                    },
                    {
                        "title": "Diagnose Like a Pathologist: Transformer-Enabled Hierarchical Attention-Guided Multiple Instance Learning for Whole Slide Image Classification",
                        "abstract": "Multiple Instance Learning (MIL) and transformers are increasingly popular in histopathology Whole Slide Image (WSI) classification. However, unlike human pathologists who selectively observe specific regions of histopathology tissues under different magnifications, most methods do not incorporate multiple resolutions of the WSIs, hierarchically and attentively, thereby leading to a loss of focus on the WSIs and information from other resolutions. To resolve this issue, we propose a Hierarchical Attention-Guided Multiple Instance Learning framework to fully exploit the WSIs. This framework can dynamically and attentively discover the discriminative regions across multiple resolutions of the WSIs. Within this framework, an Integrated Attention Transformer is proposed to further enhance the performance of the transformer and obtain a more holistic WSI (bag) representation. This transformer consists of multiple Integrated Attention Modules, which is the combination of a transformer layer and an aggregation module that produces a bag representation based on every instance representation in that bag. The experimental results show that our method achieved state-of-the-art performances on multiple datasets, including Camelyon16, TCGA-RCC, TCGA-NSCLC, and an in-house IMGC dataset. The code is available at https://github.com/BearCleverProud/HAG-MIL."
                    }
                ]
            },
            "dc70f7b4-d9fe-48ee-a723-e68577808a8e": {
                "pk": "dc70f7b4-d9fe-48ee-a723-e68577808a8e",
                "name": "Philip S. Yu",
                "collaborators": [
                    "Xuming Hu",
                    "Lijie Wen",
                    "Aiwei Liu",
                    "Zhijiang Guo",
                    "Junzhe Chen",
                    "Irwin King",
                    "Zhiwei Liu",
                    "Shiao Meng",
                    "Chenwei Zhang",
                    "Ziwei Fan",
                    "Shuang Li",
                    "Yawen Yang",
                    "Fukun Ma",
                    "Zhaochen Hong",
                    "Guan-Huei Wu",
                    "Zhiyang Teng",
                    "I. King",
                    "Zeqi Tan",
                    "Xin Zhang",
                    "Hao Peng",
                    "Leyi Pan",
                    "Yijian Lu",
                    "Jingjing Li",
                    "Lili Zhang",
                    "Ning Yang",
                    "Yanchao Sun",
                    "Shelby Heinecke",
                    "Jianguo Zhang",
                    "Haiquan Wang",
                    "Caiming Xiong",
                    "Yu Wang",
                    "Liangwei Yang",
                    "Yu-Chu Chang",
                    "Xu Wang",
                    "Jindong Wang",
                    "Yuanyi Wu",
                    "Kaijie Zhu",
                    "Hao Chen",
                    "Linyi Yang",
                    "Xiaoyuan Yi",
                    "Cunxiang Wang",
                    "Yidong Wang",
                    "Weirong Ye",
                    "Yue Zhang",
                    "Yi Chang",
                    "Qian Yang",
                    "Xingxu Xie",
                    "Xiaochuan Li",
                    "Yufei Guo",
                    "Yue Huang",
                    "Qihui Zhang",
                    "Lichao Sun"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Relation Extraction",
                    "Multimodal Learning",
                    "Adversarial Training"
                ],
                "publications": [
                    {
                        "title": "SelfLRE: Self-refining Representation Learning for Low-resource Relation Extraction",
                        "abstract": "Low-resource relation extraction (LRE) aims to extract potential relations from limited labeled corpus to handle the problem of scarcity of human annotations. Previous works mainly consist of two categories of methods: (1) Self-training methods, which improve themselves through the models' predictions, thus suffering from confirmation bias when the predictions are wrong. (2) Self-ensembling methods, which learn task-agnostic representations, therefore, generally do not work well for specific tasks. In our work, we propose a novel LRE architecture named SelfLRE, which leverages two complementary modules, one module uses self-training to obtain pseudo-labels for unlabeled data, and the other module uses self-ensembling learning to obtain the task-agnostic representations, and leverages the existing pseudo-labels to refine the better task-specific representations on unlabeled data. The two models are jointly trained through multi-task learning to iteratively improve the effect of LRE task. Experiments on three public datasets show that SelfLRE achieves 1.81% performance gain over the SOTA baseline. Source code is available at: https://github.com/THU-BPM/SelfLRE."
                    },
                    {
                        "title": "Give Me More Details: Improving Fact-Checking with Latent Retrieval",
                        "abstract": "Evidence plays a crucial role in automated fact-checking. When verifying real-world claims, existing fact-checking systems either assume the evidence sentences are given or use the search snippets returned by the search engine. Such methods ignore the challenges of collecting evidence and may not provide sufficient information to verify real-world claims. Aiming at building a better fact-checking system, we propose to incorporate full text from source documents as evidence and introduce two enriched datasets. The first one is a multilingual dataset, while the second one is monolingual (English). We further develop a latent variable model to jointly extract evidence sentences from documents and perform claim verification. Experiments indicate that including source documents can provide sufficient contextual clues even when gold evidence sentences are not annotated. The proposed system is able to achieve significant improvements upon best-reported models under different settings."
                    },
                    {
                        "title": "Multimodal Relation Extraction with Cross-Modal Retrieval and Synthesis",
                        "abstract": "Multimodal relation extraction (MRE) is the task of identifying the semantic relationships between two entities based on the context of the sentence image pair. Existing retrieval-augmented approaches mainly focused on modeling the retrieved textual knowledge, but this may not be able to accurately identify complex relations. To improve the prediction, this research proposes to retrieve textual and visual evidence based on the object, sentence, and whole image. We further develop a novel approach to synthesize the object-level, image-level, and sentence-level information for better reasoning between the same and different modalities. Extensive experiments and analyses show that the proposed method is able to effectively select and compare evidence across modalities and significantly outperforms state-of-the-art models."
                    },
                    {
                        "title": "GDA: Generative Data Augmentation Techniques for Relation Extraction Tasks",
                        "abstract": "Relation extraction (RE) tasks show promising performance in extracting relations from two entities mentioned in sentences, given sufficient annotations available during training. Such annotations would be labor-intensive to obtain in practice. Existing work adopts data augmentation techniques to generate pseudo-annotated sentences beyond limited annotations. These techniques neither preserve the semantic consistency of the original sentences when rule-based augmentations are adopted, nor preserve the syntax structure of sentences when expressing relations using seq2seq models, resulting in less diverse augmentations. In this work, we propose a dedicated augmentation technique for relational texts, named GDA, which uses two complementary modules to preserve both semantic consistency and syntax structures. We adopt a generative formulation and design a multi-tasking solution to achieve synergies. Furthermore, GDA adopts entity hints as the prior knowledge of the generative model to augment diverse sentences. Experimental results in three datasets under a low-resource setting showed that GDA could bring {\\em 2.0\\%} F1 improvements compared with no augmentation technique. Source code and data are available."
                    },
                    {
                        "title": "Sequential Recommendation with Controllable Diversification: Representation Degeneration and Diversity",
                        "abstract": "Sequential recommendation (SR) models the dynamic user preferences and generates the next-item prediction as the affinity between the sequence and items, in a joint latent space with low dimensions (i.e., the sequence and item embedding space). Both sequence and item representations suffer from the representation degeneration issue due to the user/item long-tail distributions, where tail users/ items are indistinguishably distributed as a narrow cone in the latent space. We argue that the representation degeneration issue is the root cause of insufficient recommendation diversity in existing SR methods, impairing the user potential exploration and further worsening the echo chamber issue. In this work, we first disclose the connection between the representation degeneration and recommendation diversity, in which severer representation degeneration indicates lower recommendation diversity. We then propose a novel Singular sPectrum sMoothing regularization for Recommendation (SPMRec), which acts as a controllable surrogate to alleviate the degeneration and achieve the balance between recommendation diversity and performance. The proposed smoothing regularization alleviates the degeneration by maximizing the area under the singular value curve, which is also the diversity surrogate. We conduct experiments on four benchmark datasets to demonstrate the superiority of SPMRec, and show that the proposed singular spectrum smoothing can control the balance of recommendation performance and diversity simultaneously."
                    },
                    {
                        "title": "A Survey of Text Watermarking in the Era of Large Language Models",
                        "abstract": "Text watermarking algorithms are crucial for protecting the copyright of textual content. Historically, their capabilities and application scenarios were limited. However, recent advancements in large language models (LLMs) have revolutionized these techniques. LLMs not only enhance text watermarking algorithms with their advanced abilities but also create a need for employing these algorithms to protect their own copyrights or prevent potential misuse. This paper conducts a comprehensive survey of the current state of text watermarking technology, covering four main aspects: (1) an overview and comparison of different text watermarking techniques; (2) evaluation methods for text watermarking algorithms, including their detectability, impact on text or LLM quality, robustness under target or untargeted attacks; (3) potential application scenarios for text watermarking technology; (4) current challenges and future directions for text watermarking. This survey aims to provide researchers with a thorough understanding of text watermarking technology in the era of LLM, thereby promoting its further advancement."
                    },
                    {
                        "title": "Provable Unrestricted Adversarial Training without Compromise with Generalizability",
                        "abstract": "Adversarial training (AT) is widely considered as the most promising strategy to defend against adversarial attacks and has drawn increasing interest from researchers. However, the existing AT methods still suffer from two challenges. First, they are unable to handle unrestricted adversarial examples (UAEs), which are built from scratch, as opposed to restricted adversarial examples (RAEs), which are created by adding perturbations bound by an lp norm to observed examples. Second, the existing AT methods often achieve adversarial robustness at the expense of standard generalizability (i.e., the accuracy on natural examples) because they make a tradeoff between them. To overcome these challenges, we propose a unique viewpoint that understands UAEs as imperceptibly perturbed unobserved examples. Also, we find that the tradeoff results from the separation of the distributions of adversarial examples and natural examples. Based on these ideas, we propose a novel AT approach called Provable Unrestricted Adversarial Training (PUAT), which can provide a target classifier with comprehensive adversarial robustness against both UAE and RAE, and simultaneously improve its standard generalizability. Particularly, PUAT utilizes partially labeled data to achieve effective UAE generation by accurately capturing the natural data distribution through a novel augmented triple-GAN. At the same time, PUAT extends the traditional AT by introducing the supervised loss of the target classifier into the adversarial loss and achieves the alignment between the UAE distribution, the natural data distribution, and the distribution learned by the classifier, with the collaboration of the augmented triple-GAN. Finally, the solid theoretical analysis and extensive experiments conducted on widely-used benchmarks demonstrate the superiority of PUAT."
                    },
                    {
                        "title": "Zero-shot Item-based Recommendation via Multi-task Product Knowledge Graph Pre-Training",
                        "abstract": "Existing recommender systems face difficulties with zero-shot items, i.e. items that have no historical interactions with users during the training stage. Though recent works extract universal item representation via pre-trained language models (PLMs), they ignore the crucial item relationships. This paper presents a novel paradigm for the Zero-Shot Item-based Recommendation (ZSIR) task, which pre-trains a model on product knowledge graph (PKG) to refine the item features from PLMs. We identify three challenges for pre-training PKG, which are multi-type relations in PKG, semantic divergence between item generic information and relations and domain discrepancy from PKG to downstream ZSIR task. We address the challenges by proposing four pre-training tasks and novel task-oriented adaptation (ToA) layers. Moreover, this paper discusses how to fine-tune the model on new recommendation task such that the ToA layers are adapted to ZSIR task. Comprehensive experiments on 18 markets dataset are conducted to verify the effectiveness of the proposed MPKG model."
                    },
                    {
                        "title": "Conditional Denoising Diffusion for Sequential Recommendation",
                        "abstract": "Generative models have attracted significant interest due to their ability to handle uncertainty by learning the inherent data distributions. However, two prominent generative models, namely Generative Adversarial Networks (GANs) and Variational AutoEncoders (VAEs), exhibit challenges that impede achieving optimal performance in sequential recommendation tasks. Specifically, GANs suffer from unstable optimization, while VAEs are prone to posterior collapse and over-smoothed generations. The sparse and noisy nature of sequential recommendation further exacerbates these issues. In response to these limitations, we present a conditional denoising diffusion model, which includes a sequence encoder, a cross-attentive denoising decoder, and a step-wise diffuser. This approach streamlines the optimization and generation process by dividing it into easier and tractable steps in a conditional autoregressive manner. Furthermore, we introduce a novel optimization schema that incorporates both cross-divergence loss and contrastive loss. This novel training schema enables the model to generate high-quality sequence/item representations and meanwhile precluding collapse. We conducted comprehensive experiments on four benchmark datasets, and the superior performance achieved by our model attests to its efficacy."
                    },
                    {
                        "title": "A comprehensive evaluation of ChatGPT's zero-shot Text-to-SQL capability",
                        "abstract": "This paper presents the first comprehensive analysis of ChatGPT's Text-to-SQL ability. Given the recent emergence of large-scale conversational language model ChatGPT and its impressive capabilities in both conversational abilities and code generation, we sought to evaluate its Text-to-SQL performance. We conducted experiments on 12 benchmark datasets with different languages, settings, or scenarios, and the results demonstrate that ChatGPT has strong text-to-SQL abilities. Although there is still a gap from the current state-of-the-art (SOTA) model performance, considering that the experiment was conducted in a zero-shot scenario, ChatGPT's performance is still impressive. Notably, in the ADVETA (RPL) scenario, the zero-shot ChatGPT even outperforms the SOTA model that requires fine-tuning on the Spider dataset by 4.1\\%, demonstrating its potential for use in practical applications. To support further research in related fields, we have made the data generated by ChatGPT publicly available at https://github.com/THU-BPM/chatgpt-sql."
                    },
                    {
                        "title": "Prompt Me Up: Unleashing the Power of Alignments for Multimodal Entity and Relation Extraction",
                        "abstract": "How can we better extract entities and relations from text? Using multimodal extraction with images and text obtains more signals for entities and relations, and aligns them through graphs or hierarchical fusion, aiding in extraction. Despite attempts at various fusions, previous works have overlooked many unlabeled image-caption pairs, such as NewsCLIPing. This paper proposes innovative pre-training objectives for entity-object and relation-image alignment, extracting objects from images and aligning them with entity and relation prompts for soft pseudo-labels. These labels are used as self-supervised signals for pre-training, enhancing the ability to extract entities and relations. Experiments on three datasets show an average 3.41% F1 improvement over prior SOTA. Additionally, our method is orthogonal to previous multimodal fusions, and using it on prior SOTA fusions further improves 5.47% F1."
                    },
                    {
                        "title": "MR2: A Benchmark for Multimodal Retrieval-Augmented Rumor Detection in Social Media",
                        "abstract": "As social media platforms are evolving from text-based forums into multi-modal environments, the nature of misinformation in social media is also transforming accordingly. Misinformation spreaders have recently targeted contextual connections between the modalities e.g., text and image. However, existing datasets for rumor detection mainly focus on a single modality i.e., text. To bridge this gap, we construct MR2, a multimodal multilingual retrieval-augmented dataset for rumor detection. The dataset covers rumors with images and texts, and provides evidence from both modalities that are retrieved from the Internet. Further, we develop established baselines and conduct a detailed analysis of the systems evaluated on the dataset. Extensive experiments show that MR2 will provide a challenging testbed for developing rumor detection systems designed to retrieve and reason over social media posts. Source code and data are available at: https://github.com/THU-BPM/MR2."
                    },
                    {
                        "title": "Enhancing Cross-lingual Natural Language Inference by Soft Prompting with Multilingual Verbalizer",
                        "abstract": "Cross-lingual natural language inference is a fundamental problem in cross-lingual language understanding. Many recent works have used prompt learning to address the lack of annotated parallel corpora in XNLI. However, these methods adopt discrete prompting by simply translating the templates to the target language and need external expert knowledge to design the templates. Besides, discrete prompts of human-designed template words are not trainable vectors and can not be migrated to target languages in the inference stage flexibly. In this paper, we propose a novel Soft prompt learning framework with the Multilingual Verbalizer (SoftMV) for XNLI. SoftMV first constructs cloze-style question with soft prompts for the input sample. Then we leverage bilingual dictionaries to generate an augmented multilingual question for the original question. SoftMV adopts a multilingual verbalizer to align the representations of original and augmented multilingual questions into the same semantic space with consistency regularization. Experimental results on XNLI demonstrate that SoftMV can achieve state-of-the-art performance and significantly outperform the previous methods under the few-shot and full-shot cross-lingual transfer settings."
                    },
                    {
                        "title": "Think Rationally about What You See: Continuous Rationale Extraction for Relation Extraction",
                        "abstract": "Relation extraction (RE) aims to extract potential relations according to the context of two entities, thus, deriving rational contexts from sentences plays an important role. Previous works either focus on how to leverage the entity information (e.g., entity types, entity verbalization) to inference relations, but ignore context-focused content, or use counterfactual thinking to remove the model's bias of potential relations in entities, but the relation reasoning process will still be hindered by irrelevant content. Therefore, how to preserve relevant content and remove noisy segments from sentences is a crucial task. In addition, retained content needs to be fluent enough to maintain semantic coherence and interpretability. In this work, we propose a novel rationale extraction framework named RE2, which leverages two continuity and sparsity factors to obtain relevant and coherent rationales from sentences. To solve the problem that the gold rationales are not labeled, RE2 applies an optimizable binary mask to each token in the sentence, and adjust the rationales that need to be selected according to the relation label. Experiments on four datasets show that RE2 surpasses baselines."
                    },
                    {
                        "title": "A Survey on Evaluation of Large Language Models",
                        "abstract": "Large language models (LLMs) are gaining increasing popularity in both academia and industry, owing to their unprecedented performance in various applications. As LLMs continue to play a vital role in both research and daily use, their evaluation becomes increasingly critical, not only at the task level, but also at the society level for better understanding of their potential risks. Over the past years, significant efforts have been made to examine LLMs from various perspectives. This paper presents a comprehensive review of these evaluation methods for LLMs, focusing on three key dimensions: what to evaluate, where to evaluate, and how to evaluate. Firstly, we provide an overview from the perspective of evaluation tasks, encompassing general natural language processing tasks, reasoning, medical usage, ethics, education, natural and social sciences, agent applications, and other areas. Secondly, we answer the \u2018where\u2019 and \u2018how\u2019 questions by diving into the evaluation methods and benchmarks, which serve as crucial components in assessing the performance of LLMs. Then, we summarize the success and failure cases of LLMs in different tasks. Finally, we shed light on several future challenges that lie ahead in LLMs evaluation. Our aim is to offer invaluable insights to researchers in the realm of LLMs evaluation, thereby aiding the development of more proficient LLMs. Our key point is that evaluation should be treated as an essential discipline to better assist the development of LLMs. We consistently maintain the related open-source materials at: https://github.com/MLGroupJLU/LLM-eval-survey"
                    },
                    {
                        "title": "Do Large Language Models Know about Facts?",
                        "abstract": "Large language models (LLMs) have recently driven striking performance improvements across a range of natural language processing tasks. The factual knowledge acquired during pretraining and instruction tuning can be useful in various downstream tasks, such as question answering, and language generation. Unlike conventional Knowledge Bases (KBs) that explicitly store factual knowledge, LLMs implicitly store facts in their parameters. Content generated by the LLMs can often exhibit inaccuracies or deviations from the truth, due to facts that can be incorrectly induced or become obsolete over time. To this end, we aim to comprehensively evaluate the extent and scope of factual knowledge within LLMs by designing the benchmark Pinocchio. Pinocchio contains 20K diverse factual questions that span different sources, timelines, domains, regions, and languages. Furthermore, we investigate whether LLMs are able to compose multiple facts, update factual knowledge temporally, reason over multiple pieces of facts, identify subtle factual differences, and resist adversarial examples. Extensive experiments on different sizes and types of LLMs show that existing LLMs still lack factual knowledge and suffer from various spurious correlations. We believe this is a critical bottleneck for realizing trustworthy artificial intelligence. The dataset Pinocchio and our codes will be publicly available."
                    },
                    {
                        "title": "TrustGPT: A Benchmark for Trustworthy and Responsible Large Language Models",
                        "abstract": "Large Language Models (LLMs) such as ChatGPT, have gained significant attention due to their impressive natural language processing capabilities. It is crucial to prioritize human-centered principles when utilizing these models. Safeguarding the ethical and moral compliance of LLMs is of utmost importance. However, individual ethical issues have not been well studied on the latest LLMs. Therefore, this study aims to address these gaps by introducing a new benchmark -- TrustGPT. TrustGPT provides a comprehensive evaluation of LLMs in three crucial areas: toxicity, bias, and value-alignment. Initially, TrustGPT examines toxicity in language models by employing toxic prompt templates derived from social norms. It then quantifies the extent of bias in models by measuring quantifiable toxicity values across different groups. Lastly, TrustGPT assesses the value of conversation generation models from both active value-alignment and passive value-alignment tasks. Through the implementation of TrustGPT, this research aims to enhance our understanding of the performance of conversation generation models and promote the development of language models that are more ethical and socially responsible."
                    },
                    {
                        "title": "Read it Twice: Towards Faithfully Interpretable Fact Verification by Revisiting Evidence",
                        "abstract": "Real-world fact verification task aims to verify the factuality of a claim by retrieving evidence from the source document. The quality of the retrieved evidence plays an important role in claim verification. Ideally, the retrieved evidence should be faithful (reflecting the model's decision-making process in claim verification) and plausible (convincing to humans), and can improve the accuracy of verification task. Although existing approaches leverage the similarity measure of semantic or surface form between claims and documents to retrieve evidence, they all rely on certain heuristics that prevent them from satisfying all three requirements. In light of this, we propose a fact verification model named ReRead to retrieve evidence and verify claim that: (1) Train the evidence retriever to obtain interpretable evidence (i.e., faithfulness and plausibility criteria); (2) Train the claim verifier to revisit the evidence retrieved by the optimized evidence retriever to improve the accuracy. The proposed system is able to achieve significant improvements upon best-reported models under different settings."
                    },
                    {
                        "title": "AMR-based Network for Aspect-based Sentiment Analysis",
                        "abstract": "Aspect-based sentiment analysis (ABSA) is a fine-grained sentiment classification task. Many recent works have used dependency trees to extract the relation between aspects and contexts and have achieved significant improvements. However, further improvement is limited due to the potential mismatch between the dependency tree as a syntactic structure and the sentiment classification as a semantic task. To alleviate this gap, we replace the syntactic dependency tree with the semantic structure named Abstract Meaning Representation (AMR) and propose a model called AMR-based Path Aggregation Relational Network (APARN) to take full advantage of semantic structures. In particular, we design the path aggregator and the relation-enhanced self-attention mechanism that complement each other. The path aggregator extracts semantic features from AMRs under the guidance of sentence information, while the relation-enhanced self-attention mechanism in turn improves sentence features with refined semantic information. Experimental results on four public datasets demonstrate 1.13% average F1 improvement of APARN in ABSA when compared with state-of-the-art baselines."
                    }
                ]
            }
        }
    },
    "2408.03572": {
        "paper_data": {
            "title": "2D-OOB: Attributing Data Contribution through Joint Valuation Framework",
            "url": "http://arxiv.org/abs/2408.03572v1",
            "arxiv_id": "2408.03572",
            "authors": [
                "Yifan Sun",
                "Jingyan Shen",
                "Yongchan Kwon"
            ],
            "abstract": "Data valuation has emerged as a powerful framework to quantify the contribution of each datum to the training of a particular machine learning model. However, it is crucial to recognize that the quality of various cells within a single data point can vary greatly in practice. For example, even in the case of an abnormal data point, not all cells are necessarily noisy. The single scalar valuation assigned by existing methods blurs the distinction between noisy and clean cells of a data point, thereby compromising the interpretability of the valuation. In this paper, we propose 2D-OOB, an out-of-bag estimation framework for jointly determining helpful (or detrimental) samples, as well as the particular cells that drive them. Our comprehensive experiments demonstrate that 2D-OOB achieves state-of-the-art performance across multiple use cases, while being exponentially faster. 2D-OOB excels in detecting and rectifying fine-grained outliers at the cell level, as well as localizing backdoor triggers in data poisoning attacks.",
            "introduction": "   1 Introduction  From customer behavior prediction and medical image analysis to autonomous driving and policy making, machine learning (ML) systems process ever increasing amounts of data. In such data-rich regimes, a fraction of the samples is often noisy, incorrect annotations are likely to occur, and uniform data quality standards become difficult to enforce. To address these challenges, data valuation emerges as a research field receiving increasing attention, focusing on properly assessing the contribution of each datum to ML training [12]. These methods have proven useful in identifying low-quality samples that can be detrimental to model performance, as well as selecting subsets of data that are representative of enhanced model performance [23, 48, 27]. Furthermore, they are widely applicable in data marketplace for fair revenue allocation and incentive design [51, 45, 40].   Nevertheless, existing data valuation methods assign a scalar score to each datum, thereby failing to account for the varied roles of individual cells. This leaves the valuation rationale unclear and can be unsatisfactory and sub-optimal in various practical scenarios. Firstly, whenever a score is assigned to a data point by a particular data valuation method, it is crucial to understand the underlying justifications to ensure transparency and reliability, especially in high-stakes decision making [39]. Secondly, it is important to recognize the fact that even if a data point is of low quality, it is rarely the case that all the cells within this data point are noisy [37, 26, 43]. The absence of detailed insights into how individual cells contribute to ML training inevitably leads to discarding entire data points. This can result in substantial data waste, particularly when only a few cells are noisy and data acquisition is expensive. Finally, in data markets, different cells within a data point may originate from different data sellers [3, 10]. Consequently, a singular valuation for the entire point fails to offer equitable compensation to all contributing parties.   Figure 1: Comparison of data valuation and joint valuation. (a) Data valuation evaluates the quality of individual data points, whereas (b) joint valuation evaluates the quality of individual cells. Both panels illustrate the same hypothetical dataset, and the darker colors overlaid represent the higher quality or importance. Joint valuation provides a finer level of attributions than data valuation and aims to describe how features affect data values. As panel (b) illustrates, the joint valuation framework can identify outlier cells highlighted with blue boxes (i.e., \u221211-1- 1 in \"Income\" and 100100100100 in \"Education\") and provide quantitative interpretations of data values.    Our contributions  In this paper, we propose 2D-OOB, a powerful and efficient joint valuation framework that can attribute a data point\u2019s value to its individual features. 2D-OOB quantifies the importance of each cell in a dataset, as illustrated in Figure 1, providing interpretable insights into which cells are associated with influential data points. Our method is computationally efficient as well as theoretically supported by its connections with Data-OOB [23]. Moreover, our extensive empirical experiments demonstrate the practical effectiveness of 2D-OOB in various use cases. 2D-OOB accurately identifies cell outliers and pinpoints which cells to fix to improve model performance. 2D-OOB enables inspection of data poisoning attacks by precisely",
            "references": [
                {
                    "title": "Incentives in Private Collaborative Machine Learning",
                    "abstract": "Collaborative machine learning involves training models on data from multiple parties but must incentivize their participation. Existing data valuation methods fairly value and reward each party based on shared data or model parameters but neglect the privacy risks involved. To address this, we introduce differential privacy (DP) as an incentive. Each party can select its required DP guarantee and perturb its sufficient statistic (SS) accordingly. The mediator values the perturbed SS by the Bayesian surprise it elicits about the model parameters. As our valuation function enforces a privacy-valuation trade-off, parties are deterred from selecting excessive DP guarantees that reduce the utility of the grand coalition's model. Finally, the mediator rewards each party with different posterior samples of the model parameters. Such rewards still satisfy existing incentives like fairness but additionally preserve DP and a high similarity to the grand coalition's posterior. We empirically demonstrate the effectiveness and practicality of our approach on synthetic and real-world datasets."
                },
                {
                    "title": "Efficient Data Shapley for Weighted Nearest Neighbor Algorithms",
                    "abstract": "This work aims to address an open problem in data valuation literature concerning the efficient computation of Data Shapley for weighted $K$ nearest neighbor algorithm (WKNN-Shapley). By considering the accuracy of hard-label KNN with discretized weights as the utility function, we reframe the computation of WKNN-Shapley into a counting problem and introduce a quadratic-time algorithm, presenting a notable improvement from $O(N^K)$, the best result from existing literature. We develop a deterministic approximation algorithm that further improves computational efficiency while maintaining the key fairness properties of the Shapley value. Through extensive experiments, we demonstrate WKNN-Shapley's computational efficiency and its superior performance in discerning data quality compared to its unweighted counterpart."
                },
                {
                    "title": "2D-Shapley: A Framework for Fragmented Data Valuation",
                    "abstract": "Data valuation -- quantifying the contribution of individual data sources to certain predictive behaviors of a model -- is of great importance to enhancing the transparency of machine learning and designing incentive systems for data sharing. Existing work has focused on evaluating data sources with the shared feature or sample space. How to valuate fragmented data sources of which each only contains partial features and samples remains an open question. We start by presenting a method to calculate the counterfactual of removing a fragment from the aggregated data matrix. Based on the counterfactual calculation, we further propose 2D-Shapley, a theoretical framework for fragmented data valuation that uniquely satisfies some appealing axioms in the fragmented data context. 2D-Shapley empowers a range of new use cases, such as selecting useful data fragments, providing interpretation for sample-wise data values, and fine-grained data issue diagnosis."
                },
                {
                    "title": "OpenDataVal: a Unified Benchmark for Data Valuation",
                    "abstract": "Assessing the quality and impact of individual data points is critical for improving model performance and mitigating undesirable biases within the training dataset. Several data valuation algorithms have been proposed to quantify data quality, however, there lacks a systemic and standardized benchmarking system for data valuation. In this paper, we introduce OpenDataVal, an easy-to-use and unified benchmark framework that empowers researchers and practitioners to apply and compare various data valuation algorithms. OpenDataVal provides an integrated environment that includes (i) a diverse collection of image, natural language, and tabular datasets, (ii) implementations of eleven different state-of-the-art data valuation algorithms, and (iii) a prediction model API that can import any models in scikit-learn. Furthermore, we propose four downstream machine learning tasks for evaluating the quality of data values. We perform benchmarking analysis using OpenDataVal, quantifying and comparing the efficacy of state-of-the-art data valuation approaches. We find that no single algorithm performs uniformly best across all tasks, and an appropriate algorithm should be employed for a user's downstream task. OpenDataVal is publicly available at https://opendataval.github.io with comprehensive documentation. Furthermore, we provide a leaderboard where researchers can evaluate the effectiveness of their own data valuation algorithms."
                },
                {
                    "title": "Addressing Budget Allocation and Revenue Allocation in Data Market Environments Using an Adaptive Sampling Algorithm",
                    "abstract": "High-quality machine learning models are dependent on access to high-quality training data. When the data are not already available, it is tedious and costly to obtain them. Data markets help with identifying valuable training data: model consumers pay to train a model, the market uses that budget to identify data and train the model (the budget allocation problem), and finally the market compensates data providers according to their data contribution (revenue allocation problem). For example, a bank could pay the data market to access data from other financial institutions to train a fraud detection model. Compensating data contributors requires understanding data's contribution to the model; recent efforts to solve this revenue allocation problem based on the Shapley value are inefficient to lead to practical data markets. In this paper, we introduce a new algorithm to solve budget allocation and revenue allocation problems simultaneously in linear time. The new algorithm employs an adaptive sampling process that selects data from those providers who are contributing the most to the model. Better data means that the algorithm accesses those providers more often, and more frequent accesses corresponds to higher compensation. Furthermore, the algorithm can be deployed in both centralized and federated scenarios, boosting its applicability. We provide theoretical guarantees for the algorithm that show the budget is used efficiently and the properties of revenue allocation are similar to Shapley's. Finally, we conduct an empirical evaluation to show the performance of the algorithm in practical scenarios and when compared to other baselines. Overall, we believe that the new algorithm paves the way for the implementation of practical data markets."
                },
                {
                    "title": "Efficient Sampling Approaches to Shapley Value Approximation",
                    "abstract": "Shapley value provides a unique way to fairly assess each player's contribution in a coalition and has enjoyed many applications. However, the exact computation of Shapley value is #P-hard due to the combinatoric nature of Shapley value. Many existing applications of Shapley value are based on Monte-Carlo approximation, which requires a large number of samples and the assessment of utility on many coalitions to reach high quality approximation, and thus is still far from being efficient. Can we achieve an efficient approximation of Shapley value by smartly obtaining samples? In this paper, we treat the sampling approach to Shapley value approximation as a stratified sampling problem. Our main technical contributions are a novel stratification design and two sample allocation methods based on Neyman allocation and empirical Bernstein bound, respectively. Experimental results on several real data sets and synthetic data sets demonstrate the effectiveness and efficiency of our novel stratification design and sampling approaches."
                },
                {
                    "title": "LAVA: Data Valuation without Pre-Specified Learning Algorithms",
                    "abstract": "Traditionally, data valuation (DV) is posed as a problem of equitably splitting the validation performance of a learning algorithm among the training data. As a result, the calculated data values depend on many design choices of the underlying learning algorithm. However, this dependence is undesirable for many DV use cases, such as setting priorities over different data sources in a data acquisition process and informing pricing mechanisms in a data marketplace. In these scenarios, data needs to be valued before the actual analysis and the choice of the learning algorithm is still undetermined then. Another side-effect of the dependence is that to assess the value of individual points, one needs to re-run the learning algorithm with and without a point, which incurs a large computation burden. This work leapfrogs over the current limits of data valuation methods by introducing a new framework that can value training data in a way that is oblivious to the downstream learning algorithm. Our main results are as follows. (1) We develop a proxy for the validation performance associated with a training set based on a non-conventional class-wise Wasserstein distance between training and validation sets. We show that the distance characterizes the upper bound of the validation performance for any given model under certain Lipschitz conditions. (2) We develop a novel method to value individual data based on the sensitivity analysis of the class-wise Wasserstein distance. Importantly, these values can be directly obtained for free from the output of off-the-shelf optimization solvers when computing the distance. (3) We evaluate our new data valuation framework over various use cases related to detecting low-quality data and show that, surprisingly, the learning-agnostic feature of our framework enables a significant improvement over SOTA performance while being orders of magnitude faster."
                },
                {
                    "title": "Data-OOB: Out-of-bag Estimate as a Simple and Efficient Data Value",
                    "abstract": "Data valuation is a powerful framework for providing statistical insights into which data are beneficial or detrimental to model training. Many Shapley-based data valuation methods have shown promising results in various downstream tasks, however, they are well known to be computationally challenging as it requires training a large number of models. As a result, it has been recognized as infeasible to apply to large datasets. To address this issue, we propose Data-OOB, a new data valuation method for a bagging model that utilizes the out-of-bag estimate. The proposed method is computationally efficient and can scale to millions of data by reusing trained weak learners. Specifically, Data-OOB takes less than 2.25 hours on a single CPU processor when there are $10^6$ samples to evaluate and the input dimension is 100. Furthermore, Data-OOB has solid theoretical interpretations in that it identifies the same important data point as the infinitesimal jackknife influence function when two different points are compared. We conduct comprehensive experiments using 12 classification datasets, each with thousands of sample sizes. We demonstrate that the proposed method significantly outperforms existing state-of-the-art data valuation methods in identifying mislabeled data and finding a set of helpful (or harmful) data points, highlighting the potential for applying data values in real-world applications."
                },
                {
                    "title": "WeightedSHAP: analyzing and improving Shapley based feature attributions",
                    "abstract": "Shapley value is a popular approach for measuring the influence of individual features. While Shapley feature attribution is built upon desiderata from game theory, some of its constraints may be less natural in certain machine learning settings, leading to unintuitive model interpretation. In particular, the Shapley value uses the same weight for all marginal contributions -- i.e. it gives the same importance when a large number of other features are given versus when a small number of other features are given. This property can be problematic if larger feature sets are more or less informative than smaller feature sets. Our work performs a rigorous analysis of the potential limitations of Shapley feature attribution. We identify simple settings where the Shapley value is mathematically suboptimal by assigning larger attributions for less influential features. Motivated by this observation, we propose WeightedSHAP, which generalizes the Shapley value and learns which marginal contributions to focus directly from data. On several real-world datasets, we demonstrate that the influential features identified by WeightedSHAP are better able to recapitulate the model's predictions compared to the features identified by the Shapley value."
                },
                {
                    "title": "Data Valuation in Machine Learning: \"Ingredients\", Strategies, and Open Challenges",
                    "abstract": "Data valuation in machine learning (ML) is an emerging research area that studies the worth of data in ML. Data valuation is used in collaborative ML to determine a fair compensation for every data owner and in interpretable ML to identify the most responsible, noisy, or misleading training examples. This paper presents a comprehensive technical survey that provides a new formal study of data valuation in ML through its \u201cingredients\u201d and the corresponding properties, grounds the discussion of common desiderata satisfied by existing data valuation strategies on our proposed ingredients, and identifies open research challenges for designing new ingredients, data valuation strategies, and cost reduction techniques."
                },
                {
                    "title": "VOS: Learning What You Don't Know by Virtual Outlier Synthesis",
                    "abstract": "Out-of-distribution (OOD) detection has received much attention lately due to its importance in the safe deployment of neural networks. One of the key challenges is that models lack supervision signals from unknown data, and as a result, can produce overconfident predictions on OOD data. Previous approaches rely on real outlier datasets for model regularization, which can be costly and sometimes infeasible to obtain in practice. In this paper, we present VOS, a novel framework for OOD detection by adaptively synthesizing virtual outliers that can meaningfully regularize the model's decision boundary during training. Specifically, VOS samples virtual outliers from the low-likelihood region of the class-conditional distribution estimated in the feature space. Alongside, we introduce a novel unknown-aware training objective, which contrastively shapes the uncertainty space between the ID data and synthesized outlier data. VOS achieves competitive performance on both object detection and image classification models, reducing the FPR95 by up to 9.36% compared to the previous best method on object detectors. Code is available at https://github.com/deeplearning-wisc/vos."
                },
                {
                    "title": "Beta Shapley: a Unified and Noise-reduced Data Valuation Framework for Machine Learning",
                    "abstract": "Data Shapley has recently been proposed as a principled framework to quantify the contribution of individual datum in machine learning. It can effectively identify helpful or harmful data points for a learning algorithm. In this paper, we propose Beta Shapley, which is a substantial generalization of Data Shapley. Beta Shapley arises naturally by relaxing the efficiency axiom of the Shapley value, which is not critical for machine learning settings. Beta Shapley unifies several popular data valuation methods and includes data Shapley as a special case. Moreover, we prove that Beta Shapley has several desirable statistical properties and propose efficient algorithms to estimate it. We demonstrate that Beta Shapley outperforms state-of-the-art data valuation methods on several downstream ML tasks such as: 1) detecting mislabeled training data; 2) learning with subsamples; and 3) identifying points whose addition or removal have the largest positive or negative impact on the model."
                },
                {
                    "title": "Robust variable selection under cellwise contamination",
                    "abstract": "Cellwise outliers are widespread in real world data analysis. Traditional robust methods may fail when applied to datasets under such contamination. We introduce a variable selection procedure, that uses the Gaussian rank estimator to obtain an initial empirical covariance matrix among the response and potential predictors. We re-parameterize the classical linear regression model design matrix and the response vector such that we are able to take advantage of these robustly estimated components before applying the adaptive Lasso to obtain consistent variable selection results. The procedure is robust to cellwise outliers in low and high-dimensional settings. Empirical results show good performance compared with recently proposed robust techniques, particularly in the challenging environment when contamination rates are high but the magnitude of outliers is moderate."
                },
                {
                    "title": "TrojanZoo: Towards Unified, Holistic, and Practical Evaluation of Neural Backdoors",
                    "abstract": "Neural backdoors represent one primary threat to the security of deep learning systems. The intensive research has produced a plethora of backdoor attacks/defenses, resulting in a constant arms race. However, due to the lack of evaluation benchmarks, many critical questions remain under-explored: (i) what are the strengths and limitations of different attacks/defenses? (ii) what are the best practices to operate them? and (iii) how can the existing attacks/defenses be further improved? To bridge this gap, we design and implement TROJAN-ZOO, the first open-source platform for evaluating neural backdoor attacks/defenses in a unified, holistic, and practical manner. Thus far, focusing on the computer vision domain, it has incorporated 8 representative attacks, 14 state-of-the-art defenses, 6 attack performance metrics, 10 defense utility metrics, as well as rich tools for in-depth analysis of the attack-defense interactions. Leveraging TROJANZOO, we conduct a systematic study on the existing attacks/defenses, unveiling their complex design spectrum: both manifest intricate trade-offs among multiple desiderata (e.g., the effectiveness, evasiveness, and transferability of attacks). We further explore improving the existing attacks/defenses, leading to a number of interesting findings: (i) one-pixel triggers often suffice; (ii) training from scratch often outperforms perturbing benign models to craft trojan models; (iii) optimizing triggers and trojan models jointly greatly improves both attack effectiveness and evasiveness; (iv) individual defenses can often be evaded by adaptive attacks; and (v) exploiting model interpretability significantly improves defense robustness. We envision that TROJANZOO will serve as a valuable platform to facilitate future research on neural backdoors."
                },
                {
                    "title": "Improving KernelSHAP: Practical Shapley Value Estimation via Linear Regression",
                    "abstract": "The Shapley value solution concept from cooperative game theory has become popular for interpreting ML models, but efficiently estimating Shapley values remains challenging, particularly in the model-agnostic setting. We revisit the idea of estimating Shapley values via linear regression to understand and improve upon this approach. By analyzing KernelSHAP alongside a newly proposed unbiased estimator, we develop techniques to detect its convergence and calculate uncertainty estimates. We also find that that the original version incurs a negligible increase in bias in exchange for a significant reduction in variance, and we propose a variance reduction technique that further accelerates the convergence of both estimators. Finally, we develop a version of KernelSHAP for stochastic cooperative games that yields fast new estimators for two global explanation methods."
                },
                {
                    "title": "Understanding Global Feature Contributions With Additive Importance Measures",
                    "abstract": "Understanding the inner workings of complex machine learning models is a long-standing problem and most recent research has focused on local interpretability. To assess the role of individual input features in a global sense, we explore the perspective of defining feature importance through the predictive power associated with each feature. We introduce two notions of predictive power (model-based and universal) and formalize this approach with a framework of additive importance measures, which unifies numerous methods in the literature. We then propose SAGE, a model-agnostic method that quantifies predictive power while accounting for feature interactions. Our experiments show that SAGE can be calculated efficiently and that it assigns more accurate importance values than other methods."
                },
                {
                    "title": "What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation",
                    "abstract": "Deep learning algorithms are well-known to have a propensity for fitting the training data very well and often fit even outliers and mislabeled data points. Such fitting requires memorization of training data labels, a phenomenon that has attracted significant research interest but has not been given a compelling explanation so far. A recent work of Feldman (2019) proposes a theoretical explanation for this phenomenon based on a combination of two insights. First, natural image and data distributions are (informally) known to be long-tailed, that is have a significant fraction of rare and atypical examples. Second, in a simple theoretical model such memorization is necessary for achieving close-to-optimal generalization error when the data distribution is long-tailed. However, no direct empirical evidence for this explanation or even an approach for obtaining such evidence were given. \nIn this work we design experiments to test the key ideas in this theory. The experiments require estimation of the influence of each training example on the accuracy at each test example as well as memorization values of training examples. Estimating these quantities directly is computationally prohibitive but we show that closely-related subsampled influence and memorization values can be estimated much more efficiently. Our experiments demonstrate the significant benefits of memorization for generalization on several standard benchmarks. They also provide quantitative and visually compelling evidence for the theory put forth in (Feldman, 2019)."
                },
                {
                    "title": "Efficient computation and analysis of distributional Shapley values",
                    "abstract": "Distributional data Shapley value (DShapley) has been recently proposed as a principled framework to quantify the contribution of individual datum in machine learning. DShapley develops the foundational game theory concept of Shapley values into a statistical framework and can be applied to identify data points that are useful (or harmful) to a learning algorithm. Estimating DShapley is computationally expensive, however, and this can be a major challenge to using it in practice. Moreover, there has been little mathematical analyses of how this value depends on data characteristics. In this paper, we derive the first analytic expressions for DShapley for the canonical problems of linear regression and non-parametric density estimation. These analytic forms provide new algorithms to compute DShapley that are several orders of magnitude faster than previous state-of-the-art. Furthermore, our formulas are directly interpretable and provide quantitative insights into how the value varies for different types of data. We demonstrate the efficacy of our DShapley approach on multiple real and synthetic datasets."
                },
                {
                    "title": "Preserving Causal Constraints in Counterfactual Explanations for Machine Learning Classifiers",
                    "abstract": "To construct interpretable explanations that are consistent with the original ML model, counterfactual examples---showing how the model's output changes with small perturbations to the input---have been proposed. This paper extends the work in counterfactual explanations by addressing the challenge of feasibility of such examples. For explanations of ML models in critical domains such as healthcare and finance, counterfactual examples are useful for an end-user only to the extent that perturbation of feature inputs is feasible in the real world. We formulate the problem of feasibility as preserving causal relationships among input features and present a method that uses (partial) structural causal models to generate actionable counterfactuals. When feasibility constraints cannot be easily expressed, we consider an alternative mechanism where people can label generated CF examples on feasibility: whether it is feasible to intervene and realize the candidate CF example from the original input. To learn from this labelled feasibility data, we propose a modified variational auto encoder loss for generating CF examples that optimizes for feasibility as people interact with its output. Our experiments on Bayesian networks and the widely used ''Adult-Income'' dataset show that our proposed methods can generate counterfactual explanations that better satisfy feasibility constraints than existing methods.. Code repository can be accessed here: \\textit{this https URL}"
                },
                {
                    "title": "OpenML-Python: an extensible Python API for OpenML",
                    "abstract": "OpenML is an online platform for open science collaboration in machine learning, used to share datasets and results of machine learning experiments. In this paper we introduce \\emph{OpenML-Python}, a client API for Python, opening up the OpenML platform for a wide range of Python-based tools. It provides easy access to all datasets, tasks and experiments on OpenML from within Python. It also provides functionality to conduct machine learning experiments, upload the results to OpenML, and reproduce results which are stored on OpenML. Furthermore, it comes with a scikit-learn plugin and a plugin mechanism to easily integrate other machine learning libraries written in Python into the OpenML ecosystem. Source code and documentation is available at this https URL."
                },
                {
                    "title": "Data Valuation using Reinforcement Learning",
                    "abstract": "Quantifying the value of data is a fundamental problem in machine learning. Data valuation has multiple important use cases: (1) building insights about the learning task, (2) domain adaptation, (3) corrupted sample discovery, and (4) robust learning. To adaptively learn data values jointly with the target task predictor model, we propose a meta learning framework which we name Data Valuation using Reinforcement Learning (DVRL). We employ a data value estimator (modeled by a deep neural network) to learn how likely each datum is used in training of the predictor model. We train the data value estimator using a reinforcement signal of the reward obtained on a small validation set that reflects performance on the target task. We demonstrate that DVRL yields superior data value estimates compared to alternative methods across different types of datasets and in a diverse set of application scenarios. The corrupted sample discovery performance of DVRL is close to optimal in many regimes (i.e. as if the noisy samples were known apriori), and for domain adaptation and robust learning DVRL significantly outperforms state-of-the-art by 14.6% and 10.8%, respectively."
                },
                {
                    "title": "FACE: Feasible and Actionable Counterfactual Explanations",
                    "abstract": "Work in Counterfactual Explanations tends to focus on the principle of \"the closest possible world\" that identifies small changes leading to the desired outcome. In this paper we argue that while this approach might initially seem intuitively appealing it exhibits shortcomings not addressed in the current literature. First, a counterfactual example generated by the state-of-the-art systems is not necessarily representative of the underlying data distribution, and may therefore prescribe unachievable goals (e.g., an unsuccessful life insurance applicant with severe disability may be advised to do more sports). Secondly, the counterfactuals may not be based on a \"feasible path\" between the current state of the subject and the suggested one, making actionable recourse infeasible (e.g., low-skilled unsuccessful mortgage applicants may be told to double their salary, which may be hard without first increasing their skill level). These two shortcomings may render counterfactual explanations impractical and sometimes outright offensive. To address these two major flaws, first of all, we propose a new line of Counterfactual Explanations research aimed at providing actionable and feasible paths to transform a selected instance into one that meets a certain goal. Secondly, we propose FACE: an algorithmically sound way of uncovering these \"feasible paths\" based on the shortest path distances defined via density-weighted metrics. Our approach generates counterfactuals that are coherent with the underlying data distribution and supported by the \"feasible paths\" of change, which are achievable and can be tailored to the problem at hand."
                },
                {
                    "title": "Towards Realistic Individual Recourse and Actionable Explanations in Black-Box Decision Making Systems",
                    "abstract": "Machine learning based decision making systems are increasingly affecting humans. An individual can suffer an undesirable outcome under such decision making systems (e.g. denied credit) irrespective of whether the decision is fair or accurate. Individual recourse pertains to the problem of providing an actionable set of changes a person can undertake in order to improve their outcome. We propose a recourse algorithm that models the underlying data distribution or manifold. We then provide a mechanism to generate the smallest set of changes that will improve an individual's outcome. This mechanism can be easily used to provide recourse for any differentiable machine learning based decision making system. Further, the resulting algorithm is shown to be applicable to both supervised classification and causal decision making systems. Our work attempts to fill gaps in existing fairness literature that have primarily focused on discovering and/or algorithmically enforcing fairness constraints on decision making systems. This work also provides an alternative approach to generating counterfactual explanations."
                },
                {
                    "title": "Efficient Task-Specific Data Valuation for Nearest Neighbor Algorithms",
                    "abstract": "\n Given a data set\n D\n containing millions of data points and a data consumer who is willing to pay for $\n X\n to train a machine learning (ML) model over\n D\n ,\n how should we distribute this $X to each data point to reflect its \"value\"?\n In this paper, we define the \"relative value of data\" via the Shapley value, as it uniquely possesses properties with appealing real-world interpretations, such as fairness, rationality and decentralizability. For general, bounded utility functions, the Shapley value is known to be challenging to compute: to get Shapley values for all\n N\n data points, it requires\n O\n (2\n \n N\n \n ) model evaluations for exact computation and\n O\n (\n N\n log\n N\n ) for (\n \u03f5\n , \u03b4)-approximation.\n \n \n In this paper, we focus on one popular family of ML models relying on\n K\n -nearest neighbors (\n K\n NN). The most surprising result is that for unweighted\n K\n NN classifiers and regressors, the Shapley value of all\n N\n data points can be computed,\n exactly\n , in\n O\n (\n N\n log\n N\n ) time - an exponential improvement on computational complexity! Moreover, for (\n \u03f5\n , \u03b4)-approximation, we are able to develop an algorithm based on Locality Sensitive Hashing (LSH) with only sublinear complexity\n O\n (\n N\n \n h\n (\n \u03f5\n ,\n K\n )\n \n log\n N\n ) when\n \u03f5\n is not too small and\n K\n is not too large. We empirically evaluate our algorithms on up to 10 million data points and even our\n exact\n algorithm is up to three orders of magnitude faster than the baseline approximation algorithm. The LSH-based approximation algorithm can accelerate the value calculation process even further.\n \n \n We then extend our algorithm to other scenarios such as (1) weighed\n K\n NN classifiers, (2) different data points are clustered by different\n data curators\n , and (3) there are\n data analysts\n providing computation who also requires proper valuation.\n Some\n of these extensions, although also being improved exponentially, are less practical for exact computation (e.g.,\n O\n (\n \n N\n K\n \n ) complexity for weigthed\n K\n NN). We thus propose an Monte Carlo approximation algorithm, which is\n O\n (\n N\n (log\n N\n )\n 2\n /(log\n K\n )\n 2\n ) times more efficient than the baseline approximation algorithm.\n"
                },
                {
                    "title": "Explaining machine learning classifiers through diverse counterfactual explanations",
                    "abstract": "Post-hoc explanations of machine learning models are crucial for people to understand and act on algorithmic predictions. An intriguing class of explanations is through counterfactuals, hypothetical examples that show people how to obtain a different prediction. We posit that effective counterfactual explanations should satisfy two properties: feasibility of the counterfactual actions given user context and constraints, and diversity among the counterfactuals presented. To this end, we propose a framework for generating and evaluating a diverse set of counterfactual explanations based on determinantal point processes. To evaluate the actionability of counterfactuals, we provide metrics that enable comparison of counterfactual-based methods to other local explanation methods. We further address necessary tradeoffs and point to causal implications in optimizing for counterfactuals. Our experiments on four real-world datasets show that our framework can generate a set of counterfactuals that are diverse and well approximate local decision boundaries, outperforming prior approaches to generating diverse counterfactuals. We provide an implementation of the framework at https://github.com/microsoft/DiCE."
                },
                {
                    "title": "Data Shapley: Equitable Valuation of Data for Machine Learning",
                    "abstract": "As data becomes the fuel driving technological and economic growth, a fundamental challenge is how to quantify the value of data in algorithmic predictions and decisions. For example, in healthcare and consumer markets, it has been suggested that individuals should be compensated for the data that they generate, but it is not clear what is an equitable valuation for individual data. In this work, we develop a principled framework to address data valuation in the context of supervised machine learning. Given a learning algorithm trained on $n$ data points to produce a predictor, we propose data Shapley as a metric to quantify the value of each training datum to the predictor performance. Data Shapley value uniquely satisfies several natural properties of equitable data valuation. We develop Monte Carlo and gradient-based methods to efficiently estimate data Shapley values in practical settings where complex learning algorithms, including neural networks, are trained on large datasets. In addition to being equitable, extensive experiments across biomedical, image and synthetic data demonstrate that data Shapley has several other benefits: 1) it is more powerful than the popular leave-one-out or leverage score in providing insight on what data is more valuable for a given learning task; 2) low Shapley value data effectively capture outliers and corruptions; 3) high Shapley value data inform what type of new data to acquire to improve the predictor."
                },
                {
                    "title": "Adversarial Robustness Toolbox v1.0.0",
                    "abstract": "Adversarial Robustness Toolbox (ART) is a Python library supporting developers and researchers in defending Machine Learning models (Deep Neural Networks, Gradient Boosted Decision Trees, Support Vector Machines, Random Forests, Logistic Regression, Gaussian Processes, Decision Trees, Scikit-learn Pipelines, etc.) against adversarial threats and helps making AI systems more secure and trustworthy. Machine Learning models are vulnerable to adversarial examples, which are inputs (images, texts, tabular data, etc.) deliberately modified to produce a desired response by the Machine Learning model. ART provides the tools to build and deploy defences and test them with adversarial attacks. Defending Machine Learning models involves certifying and verifying model robustness and model hardening with approaches such as pre-processing inputs, augmenting training data with adversarial samples, and leveraging runtime detection methods to flag any inputs that might have been modified by an adversary. The attacks implemented in ART allow creating adversarial attacks against Machine Learning models which is required to test defenses with state-of-the-art threat models. Supported Machine Learning Libraries include TensorFlow (v1 and v2), Keras, PyTorch, MXNet, Scikit-learn, XGBoost, LightGBM, CatBoost, and GPy. The source code of ART is released with MIT license at https://github.com/IBM/adversarial-robustness-toolbox. The release includes code examples, notebooks with tutorials and documentation (this http URL)."
                },
                {
                    "title": "Consistent Individualized Feature Attribution for Tree Ensembles",
                    "abstract": "Interpreting predictions from tree ensemble methods such as gradient boosting machines and random forests is important, yet feature attribution for trees is often heuristic and not individualized for each prediction. Here we show that popular feature attribution methods are inconsistent, meaning they can lower a feature's assigned importance when the true impact of that feature actually increases. This is a fundamental problem that casts doubt on any comparison between features. To address it we turn to recent applications of game theory and develop fast exact tree solutions for SHAP (SHapley Additive exPlanation) values, which are the unique consistent and locally accurate attribution values. We then extend SHAP values to interaction effects and define SHAP interaction values. We propose a rich visualization of individualized feature attributions that improves over classic attribution summaries and partial dependence plots, and a unique \"supervised\" clustering (clustering based on feature attributions). We demonstrate better agreement with human intuition through a user study, exponential improvements in run time, improved clustering performance, and better identification of influential features. An implementation of our algorithm has also been merged into XGBoost and LightGBM, see this http URL for details."
                },
                {
                    "title": "Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning",
                    "abstract": "Deep learning models have achieved high performance on many tasks, and thus have been applied to many security-critical scenarios. For example, deep learning-based face recognition systems have been used to authenticate users to access many security-sensitive applications like payment apps. Such usages of deep learning systems provide the adversaries with sufficient incentives to perform attacks against these systems for their adversarial purposes. In this work, we consider a new type of attacks, called backdoor attacks, where the attacker's goal is to create a backdoor into a learning-based authentication system, so that he can easily circumvent the system by leveraging the backdoor. Specifically, the adversary aims at creating backdoor instances, so that the victim learning system will be misled to classify the backdoor instances as a target label specified by the adversary. In particular, we study backdoor poisoning attacks, which achieve backdoor attacks using poisoning strategies. Different from all existing work, our studied poisoning strategies can apply under a very weak threat model: (1) the adversary has no knowledge of the model and the training set used by the victim system; (2) the attacker is allowed to inject only a small amount of poisoning samples; (3) the backdoor key is hard to notice even by human beings to achieve stealthiness. We conduct evaluation to demonstrate that a backdoor adversary can inject only around 50 poisoning samples, while achieving an attack success rate of above 90%. We are also the first work to show that a data poisoning attack can create physically implementable backdoors without touching the training process. Our work demonstrates that backdoor poisoning attacks pose real threats to a learning system, and thus highlights the importance of further investigation and proposing defense strategies against them."
                },
                {
                    "title": "Towards better understanding of gradient-based attribution methods for Deep Neural Networks",
                    "abstract": "Understanding the flow of information in Deep Neural Networks (DNNs) is a challenging problem that has gain increasing attention over the last few years. While several methods have been proposed to explain network predictions, there have been only a few attempts to compare them from a theoretical perspective. What is more, no exhaustive empirical comparison has been performed in the past. In this work, we analyze four gradient-based attribution methods and formally prove conditions of equivalence and approximation between them. By reformulating two of these methods, we construct a unified framework which enables a direct comparison, as well as an easier implementation. Finally, we propose a novel evaluation metric, called Sensitivity-n and test the gradient-based attribution methods alongside with a simple perturbation-based attribution method on several datasets in the domains of image and text classification, using various network architectures."
                },
                {
                    "title": "Counterfactual Explanations Without Opening the Black Box: Automated Decisions and the GDPR",
                    "abstract": "There has been much discussion of the right to explanation in the EU General Data Protection Regulation, and its existence, merits, and disadvantages. Implementing a right to explanation that opens the black box of algorithmic decision-making faces major legal and technical barriers. Explaining the functionality of complex algorithmic decision-making systems and their rationale in specific cases is a technically challenging problem. Some explanations may offer little meaningful information to data subjects, raising questions around their value. Explanations of automated decisions need not hinge on the general public understanding how algorithmic systems function. Even though such interpretability is of great importance and should be pursued, explanations can, in principle, be offered without opening the black box. Looking at explanations as a means to help a data subject act rather than merely understand, one could gauge the scope and content of explanations according to the specific goal or action they are intended to support. From the perspective of individuals affected by automated decision-making, we propose three aims for explanations: (1) to inform and help the individual understand why a particular decision was reached, (2) to provide grounds to contest the decision if the outcome is undesired, and (3) to understand what would need to change in order to receive a desired result in the future, based on the current decision-making model. We assess how each of these goals finds support in the GDPR. We suggest data controllers should offer a particular type of explanation, unconditional counterfactual explanations, to support these three aims. These counterfactual explanations describe the smallest change to the world that can be made to obtain a desirable outcome, or to arrive at the closest possible world, without needing to explain the internal logic of the system."
                },
                {
                    "title": "BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain",
                    "abstract": "Deep learning-based techniques have achieved state-of-the-art performance on a wide variety of recognition and classification tasks. However, these networks are typically computationally expensive to train, requiring weeks of computation on many GPUs; as a result, many users outsource the training procedure to the cloud or rely on pre-trained models that are then fine-tuned for a specific task. In this paper we show that outsourced training introduces new security risks: an adversary can create a maliciously trained network (a backdoored neural network, or a \\emph{BadNet}) that has state-of-the-art performance on the user's training and validation samples, but behaves badly on specific attacker-chosen inputs. We first explore the properties of BadNets in a toy example, by creating a backdoored handwritten digit classifier. Next, we demonstrate backdoors in a more realistic scenario by creating a U.S. street sign classifier that identifies stop signs as speed limits when a special sticker is added to the stop sign; we then show in addition that the backdoor in our US street sign detector can persist even if the network is later retrained for another task and cause a drop in accuracy of {25}\\% on average when the backdoor trigger is present. These results demonstrate that backdoors in neural networks are both powerful and---because the behavior of neural networks is difficult to explicate---stealthy. This work provides motivation for further research into techniques for verifying and inspecting neural networks, just as we have developed tools for verifying and debugging software."
                },
                {
                    "title": "A Unified Approach to Interpreting Model Predictions",
                    "abstract": "Understanding why a model makes a certain prediction can be as crucial as the prediction's accuracy in many applications. However, the highest accuracy for large modern datasets is often achieved by complex models that even experts struggle to interpret, such as ensemble or deep learning models, creating a tension between accuracy and interpretability. In response, various methods have recently been proposed to help users interpret the predictions of complex models, but it is often unclear how these methods are related and when one method is preferable over another. To address this problem, we present a unified framework for interpreting predictions, SHAP (SHapley Additive exPlanations). SHAP assigns each feature an importance value for a particular prediction. Its novel components include: (1) the identification of a new class of additive feature importance measures, and (2) theoretical results showing there is a unique solution in this class with a set of desirable properties. The new class unifies six existing methods, notable because several recent methods in the class lack the proposed desirable properties. Based on insights from this unification, we present new methods that show improved computational performance and/or better consistency with human intuition than previous approaches."
                },
                {
                    "title": "Understanding Black-box Predictions via Influence Functions",
                    "abstract": "How can we explain the predictions of a black-box model? In this paper, we use influence functions \u2014 a classic technique from robust statistics \u2014 to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. We show that even on non-convex and non-differentiable models where the theory breaks down, approximations to influence functions can still provide valuable information. On linear models and convolutional neural networks, we demonstrate that influence functions are useful for multiple purposes: understanding model behavior, debugging models, detecting dataset errors, and even creating visually-indistinguishable training-set attacks."
                },
                {
                    "title": "Detecting Deviating Data Cells",
                    "abstract": "ABSTRACT A multivariate dataset consists of n cases in d dimensions, and is often stored in an n by d data matrix. It is well-known that real data may contain outliers. Depending on the situation, outliers may be (a) undesirable errors, which can adversely affect the data analysis, or (b) valuable nuggets of unexpected information. In statistics and data analysis, the word outlier usually refers to a row of the data matrix, and the methods to detect such outliers only work when at least half the rows are clean. But often many rows have a few contaminated cell values, which may not be visible by looking at each variable (column) separately. We propose the first method to detect deviating data cells in a multivariate sample which takes the correlations between the variables into account. It has no restriction on the number of clean rows, and can deal with high dimensions. Other advantages are that it provides predicted values of the outlying cells, while imputing missing values at the same time. We illustrate the method on several real datasets, where it uncovers more structure than found by purely columnwise methods or purely rowwise methods. The proposed method can help to diagnose why a certain row is outlying, for example, in process control. It also serves as an initial step for estimating multivariate location and scatter matrices."
                },
                {
                    "title": "On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation",
                    "abstract": "Understanding and interpreting classification decisions of automated image classification systems is of high value in many applications, as it allows to verify the reasoning of the system and provides additional information to the human expert. Although machine learning methods are solving very successfully a plethora of tasks, they have in most cases the disadvantage of acting as a black box, not providing any information about what made them arrive at a particular decision. This work proposes a general solution to the problem of understanding classification decisions by pixel-wise decomposition of nonlinear classifiers. We introduce a methodology that allows to visualize the contributions of single pixels to predictions for kernel-based classifiers over Bag of Words features and for multilayered neural networks. These pixel contributions can be visualized as heatmaps and are provided to a human expert who can intuitively not only verify the validity of the classification decision, but also focus further analysis on regions of potential interest. We evaluate our method for classifiers trained on PASCAL VOC 2009 images, synthetic image data containing geometric shapes, the MNIST handwritten digits data set and for the pre-trained ImageNet model available as part of the Caffe open source package."
                },
                {
                    "title": "Striving for Simplicity: The All Convolutional Net",
                    "abstract": "Most modern convolutional neural networks (CNNs) used for object recognition are built using the same principles: Alternating convolution and max-pooling layers followed by a small number of fully connected layers. We re-evaluate the state of the art for object recognition from small images with convolutional networks, questioning the necessity of different components in the pipeline. We find that max-pooling can simply be replaced by a convolutional layer with increased stride without loss in accuracy on several image recognition benchmarks. Following this finding -- and building on other recent work for finding simple network structures -- we propose a new architecture that consists solely of convolutional layers and yields competitive or state of the art performance on several object recognition datasets (CIFAR-10, CIFAR-100, ImageNet). To analyze the network we introduce a new variant of the \"deconvolution approach\" for visualizing features learned by CNNs, which can be applied to a broader range of network structures than existing approaches."
                },
                {
                    "title": "Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps",
                    "abstract": "This paper addresses the visualisation of image classification models, learnt using deep Convolutional Networks (ConvNets). We consider two visualisation techniques, based on computing the gradient of the class score with respect to the input image. The first one generates an image, which maximises the class score [Erhan et al., 2009], thus visualising the notion of the class, captured by a ConvNet. The second technique computes a class saliency map, specific to a given image and class. We show that such maps can be employed for weakly supervised object segmentation using classification ConvNets. Finally, we establish the connection between the gradient-based ConvNet visualisation methods and deconvolutional networks [Zeiler et al., 2013]."
                },
                {
                    "title": "A Value for n-person Games",
                    "abstract": "Introduction At the foundation of the theory of games is the assumption that the players of a game can evaluate, in their utility scales, every \u201cprospect\u201d that might arise as a result of a play. In attempting to apply the theory to any field, one would normally expect to be permitted to include, in the class of \u201cprospects,\u201d the prospect of having to play a game. The possibility of evaluating games is therefore of critical importance. So long as the theory is unable to assign values to the games typically found in application, only relatively simple situations\u2014where games do not depend on other games\u2014will be susceptible to analysis and solution. In the finite theory of von Neumann and Morgenstern difficulty in evaluation persists for the \u201cessential\u201d games, and for only those. In this note we deduce a value for the \u201cessential\u201d case and examine a number of its elementary properties. We proceed from a set of three axioms, having simple intuitive interpretations, which suffice to determine the value uniquely. Our present work, though mathematically self-contained, is founded conceptually on the von Neumann\u2014Morgenstern theory up to their introduction of characteristic functions. We thereby inherit certain important underlying assumptions: (a) that utility is objective and transferable; (b) that games are cooperative affairs; (c) that games, granting (a) and (b), are adequately represented by their characteristic functions."
                },
                {
                    "title": "DAVINZ: Data Valuation using Deep Neural Networks at Initialization",
                    "abstract": "Recent years have witnessed a surge of interest in developing trustworthy methods to evaluate the value of data in many real-world applications (e.g., collaborative machine learning, data mar-ketplaces). Existing data valuation methods typically valuate data using the generalization performance of converged machine learning models after their long-term model training, hence making data valuation on large complex deep neural networks (DNNs) unaffordable. To this end, we theoretically derive a domain-aware generalization bound to estimate the generalization performance of DNNs without model training. We then exploit this theoretically derived generalization bound to develop a novel training-free data valuation method named data valuation at initialization (D A V INZ ) on DNNs, which consistently achieves remarkable effectiveness and ef\ufb01ciency in practice. Moreover, our training-free D A V INZ , surprisingly, can even theoretically and empirically enjoy the desirable properties that training-based data valuation methods usually attain, thus making it more trustworthy in practice."
                },
                {
                    "title": "Data Banzhaf: A Data Valuation Framework with Maximal Robustness to Learning Stochasticity",
                    "abstract": ","
                },
                {
                    "title": "Trojaning Attack on Neural Networks",
                    "abstract": "\u2014With the fast spread of machine learning tech- niques, sharing and adopting public machine learning models become very popular. This gives attackers many new opportunities. In this paper, we propose a trojaning attack on neural networks. As the models are not intuitive for human to understand, the attack features stealthiness. Deploying trojaned models can cause various severe consequences including endangering human lives (in applications like autonomous driving). We \ufb01rst inverse the neural network to generate a general trojan trigger , and then retrain the model with reversed engineered training data to inject malicious behaviors to the model. The malicious behaviors are only activated by inputs stamped with the trojan trigger. In our attack, we do not need to tamper with the original training process, which usually takes weeks to months. Instead, it takes minutes to hours to apply our attack. Also, we do not require the datasets that are used to train the model. In practice, the datasets are usually not shared due to privacy or copyright concerns. We use \ufb01ve different applications to demonstrate the power of our attack, and perform a deep analysis on the possible factors that affect the attack. The results show that our attack is highly effective and ef\ufb01cient. The trojaned behaviors can be successfully triggered (with nearly 100% possibility) without affecting its test accuracy for normal input and even with better accuracy on public dataset. Also, it only takes a small amount of time to attack a complex neuron network model. In the end, we also discuss possible defense against such attacks."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively assess the contribution of individual cells within data points to machine learning training, rather than relying on scalar scores for entire data points?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for enhancing the transparency and reliability of machine learning models, especially in high-stakes applications where data quality is paramount. By providing a detailed understanding of how individual cells contribute to model performance, we can reduce data waste, improve data acquisition efficiency, and ensure fair compensation in data marketplaces. This advancement could lead to more robust machine learning systems and foster trust in automated decision-making processes, ultimately influencing future research directions in data valuation and model interpretability.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of accurately attributing the value of a data point to its individual cells, as existing methods typically assign a single score to entire data points. Naive approaches may fail because they overlook the nuanced contributions of individual features, leading to potential misinterpretations of data quality. Additionally, technical obstacles include the need for computational efficiency and theoretical grounding to support the proposed methodologies, as well as the practical difficulty of identifying and addressing noisy cells without discarding valuable data.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on scalar data valuation methods that do not account for the individual contributions of cells, leading to a lack of granularity in data assessment. Barriers include the complexity of developing a framework that can efficiently evaluate individual cells while maintaining interpretability and computational feasibility. Our approach, 2D-OOB, differs from prior work by providing a joint valuation framework that quantifies the importance of each cell, thus addressing the limitations of existing methods and offering a more nuanced understanding of data quality.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, 2D-OOB, utilizes a joint valuation framework to assess the importance of individual cells within data points. We will employ a diverse set of datasets to evaluate the effectiveness of our method, using metrics such as accuracy in identifying cell outliers and improvements in model performance. The expected outcomes include a clearer understanding of which cells contribute most significantly to data quality, the ability to pinpoint specific cells that require correction, and enhanced insights into data poisoning attacks, ultimately leading to more effective and interpretable machine learning models."
            }
        },
        "author_data": {
            "018553fa-01e1-45a5-8f84-47050ad57d14": {
                "pk": "018553fa-01e1-45a5-8f84-47050ad57d14",
                "name": "Yifan Sun",
                "collaborators": [
                    "Francis Bach",
                    "Mingshang Hu",
                    "Zhengyang Sun",
                    "Xihong Wu",
                    "Yuzhao Zhang"
                ],
                "domain": [
                    "Graph Theory",
                    "Self-Supervised Learning",
                    "Optimization",
                    "Multi-Task Learning"
                ],
                "publications": [
                    {
                        "title": "Explicit positive solutions to $G$-heat equations and the application to $G$-capacities",
                        "abstract": "In this paper, we study a class of explicit positive solutions to $G$-heat equations by solving second order nonlinear ordinary differential equations. Based on the positive solutions, we give the sharp order of $G$-capacity $c_{\\sigma}\\left( \\left[ -\\varepsilon,\\varepsilon \\right] \\right) $ when $\\varepsilon \\rightarrow0$."
                    },
                    {
                        "title": "Target observation of complex networks",
                        "abstract": "How to observe the state of a network from a limited number of measurements has become an important issue in complex networks, engineering, communication, epidemiology, etc. Under some scenarios, it is neither unfeasible nor unnecessary to observe the entire network. Therefore, we investigate the target observation of a network in this paper. We propose a target minimal dominating set problem corresponding to target observation, which is a natural generalization of classical minimal dominating set problem. Three algorithms are proposed to approximate the minimum set of occupied nodes sufficient for target observation. Extensive numerical results on computer-generated random networks and real-world networks demonstrate that the proposed algorithms offer superior performance in identification of a target minimal dominating set."
                    },
                    {
                        "title": "Embodied Self-supervised Learning by Coordinated Sampling and Training",
                        "abstract": "Self-supervised learning can significantly improve the performance of downstream tasks, however, the dimensions of learned representations normally lack explicit physical meanings. In this work, we propose a novel self-supervised approach to solve inverse problems by employing the corresponding physical forward process so that the learned representations can have explicit physical meanings. The proposed approach works in an analysis-by-synthesis manner to learn an inference network by iteratively sampling and training. At the sampling step, given observed data, the inference network is used to approximate the intractable posterior, from which we sample input parameters and feed them to a physical process to generate data in the observational space; At the training step, the same network is optimized with the sampled paired data. We prove the feasibility of the proposed method by tackling the acoustic-to-articulatory inversion problem to infer articulatory information from speech. Given an articulatory synthesizer, an inference model can be trained completely from scratch with random initialization. Our experiments demonstrate that the proposed method can converge steadily and the network learns to control the articulatory synthesizer to speak like a human. We also demonstrate that trained models can generalize well to unseen speakers or even new languages, and performance can be further improved through self-adaptation."
                    },
                    {
                        "title": "Safe Screening for the Generalized Conditional Gradient Method",
                        "abstract": "The conditional gradient method (CGM) has been widely used for fast sparse approximation, having a low per iteration computational cost for structured sparse regularizers. We explore the sparsity acquiring properties of a generalized CGM (gCGM), where the constraint is replaced by a penalty function based on a gauge penalty; this can be done without significantly increasing the per-iteration computation, and applies to general notions of sparsity. Without assuming bounded iterates, we show $O(1/t)$ convergence of the function values and gap of gCGM. We couple this with a safe screening rule, and show that at a rate $O(1/(t\\delta^2))$, the screened support matches the support at the solution, where $\\delta \\geq 0$ measures how close the problem is to being degenerate. In our experiments, we show that the gCGM for these modified penalties have similar feature selection properties as common penalties, but with potentially more stability over the choice of hyperparameter."
                    },
                    {
                        "title": "Screening for a Reweighted Penalized Conditional Gradient Method",
                        "abstract": "The conditional gradient method (CGM) is widely used in large-scale sparse convex optimization, having a low per iteration computational cost for structured sparse regularizers and a greedy approach to collecting nonzeros. We explore the sparsity acquiring properties of a general penalized CGM (P-CGM) for convex regularizers and a reweighted penalized CGM (RP-CGM) for nonconvex regularizers, replacing the usual convex constraints with gauge-inspired penalties. This generalization does not increase the per-iteration complexity noticeably. Without assuming bounded iterates or using line search, we show $O(1/t)$ convergence of the gap of each subproblem, which measures distance to a stationary point. We couple this with a screening rule which is safe in the convex case, converging to the true support at a rate $O(1/(\\delta^2))$ where $\\delta \\geq 0$ measures how close the problem is to degeneracy. In the nonconvex case the screening rule converges to the true support in a finite number of iterations, but is not necessarily safe in the intermediate iterates. In our experiments, we verify the consistency of the method and adjust the aggressiveness of the screening rule by tuning the concavity of the regularizer."
                    },
                    {
                        "title": "Semisoft Task Clustering for Multi-Task Learning",
                        "abstract": "Multi-task learning (MTL) aims to improve the performance of multiple related prediction tasks by leveraging useful information from them. Due to their flexibility and ability to reduce unknown coefficients substantially, the task-clustering-based MTL approaches have attracted considerable attention. Motivated by the idea of semisoft clustering of data, we propose a semisoft task clustering approach, which can simultaneously reveal the task cluster structure for both pure and mixed tasks as well as select the relevant features. The main assumption behind our approach is that each cluster has some pure tasks, and each mixed task can be represented by a linear combination of pure tasks in different clusters. To solve the resulting non-convex constrained optimization problem, we design an efficient three-step algorithm. The experimental results based on synthetic and real-world datasets validate the effectiveness and efficiency of the proposed approach. Finally, we extend the proposed approach to a robust task clustering problem."
                    }
                ]
            },
            "68cf27fb-bc5f-4d82-8ea1-59f0011b0543": {
                "pk": "68cf27fb-bc5f-4d82-8ea1-59f0011b0543",
                "name": "Jingyan Shen",
                "collaborators": [
                    "Yifan Sun",
                    "Yizi Zhang",
                    "Xiaoxue Xiong",
                    "Yongchan Kwon",
                    "Yunfan Jiang",
                    "Zhaohui Huang",
                    "Minsheng Hao",
                    "Xuegong Zhang",
                    "Yining Shi",
                    "Yunlong Wang"
                ],
                "domain": [
                    "Time Series Analysis",
                    "Deep Learning",
                    "Object Detection",
                    "Health Informatics"
                ],
                "publications": [
                    {
                        "title": "TimeInf: Time Series Data Contribution via Influence Functions",
                        "abstract": "Evaluating the contribution of individual data points to a model's prediction is critical for interpreting model predictions and improving model performance. Existing data contribution methods have been applied to various data types, including tabular data, images, and texts; however, their primary focus has been on i.i.d. settings. Despite the pressing need for principled approaches tailored to time series datasets, the problem of estimating data contribution in such settings remains unexplored, possibly due to challenges associated with handling inherent temporal dependencies. This paper introduces TimeInf, a data contribution estimation method for time-series datasets. TimeInf uses influence functions to attribute model predictions to individual time points while preserving temporal structures. Our extensive empirical results demonstrate that TimeInf outperforms state-of-the-art methods in identifying harmful anomalies and helpful time points for forecasting. Additionally, TimeInf offers intuitive and interpretable attributions of data values, allowing us to easily distinguish diverse anomaly patterns through visualizations."
                    },
                    {
                        "title": "MMA-RNN: A Multi-level Multi-task Attention-based Recurrent Neural Network for Discrimination and Localization of Atrial Fibrillation",
                        "abstract": "The automatic detection of atrial fibrillation based on electrocardiograph (ECG) signals has received wide attention both clinically and practically. It is challenging to process ECG signals with cyclical pattern, varying length and unstable quality due to noise and distortion. Besides, there has been insufficient research on separating persistent atrial fibrillation from paroxysmal atrial fibrillation, and little discussion on locating the onsets and end points of AF episodes. It is even more arduous to perform well on these two distinct but interrelated tasks, while avoiding the mistakes inherent from stage-by-stage approaches. This paper proposes the Multi-level Multi-task Attention-based Recurrent Neural Network for three-class discrimination on patients and localization of the exact timing of AF episodes. Our model captures three-level sequential features based on a hierarchical architecture utilizing Bidirectional Long and Short-Term Memory Network (Bi-LSTM) and attention layers, and accomplishes the two tasks simultaneously with a multi-head classifier. The model is designed as an end-to-end framework to enhance information interaction and reduce error accumulation. Finally, we conduct experiments on CPSC 2021 dataset and the result demonstrates the superior performance of our method, indicating the potential application of MMA-RNN to wearable mobile devices for routine AF monitoring and early diagnosis."
                    },
                    {
                        "title": "SRCN3D: Sparse R-CNN 3D for Compact Convolutional Multi-View 3D Object Detection and Tracking",
                        "abstract": "Detection and tracking of moving objects is an essential component in environmental perception for autonomous driving. In the flourishing field of multi-view 3D camera-based detectors, different transformer-based pipelines are designed to learn queries in 3D space from 2D feature maps of perspective views, but the dominant dense BEV query mechanism is computationally inefficient. This paper proposes Sparse R-CNN 3D (SRCN3D), a novel two-stage fully-sparse detector that incorporates sparse queries, sparse attention with box-wise sampling, and sparse prediction. SRCN3D adopts a cascade structure with the twin-track update of both a fixed number of query boxes and latent query features. Our novel sparse feature sampling module only utilizes local 2D region of interest (RoI) features calculated by the projection of 3D query boxes for further box refinement, leading to a fully-convolutional and deployment-friendly pipeline. For multi-object tracking, motion features, query features and RoI features are comprehensively utilized in multi-hypotheses data association. Extensive experiments on nuScenes dataset demonstrate that SRCN3D achieves competitive performance in both 3D object detection and multi-object tracking tasks, while also exhibiting superior efficiency compared to transformer-based methods. Code and models are available at https://github.com/synsin0/SRCN3D."
                    }
                ]
            },
            "71582240-45b1-42ba-9fcf-f063dd8ec9bd": {
                "pk": "71582240-45b1-42ba-9fcf-f063dd8ec9bd",
                "name": "Yongchan Kwon",
                "collaborators": [
                    "James Zou",
                    "Weixin Liang",
                    "Antonio Ginart",
                    "Seongok Ryu",
                    "Woo Youn Kim",
                    "Wonyoung Kim",
                    "Myunghee Cho Paik",
                    "Zhaonan Qu",
                    "Manuel A. Rivas",
                    "Eric Wu"
                ],
                "domain": [
                    "Data Valuation",
                    "Machine Learning",
                    "Causal Inference",
                    "Bayesian Methods"
                ],
                "publications": [
                    {
                        "title": "Beta Shapley: a Unified and Noise-reduced Data Valuation Framework for Machine Learning",
                        "abstract": "Data Shapley has recently been proposed as a principled framework to quantify the contribution of individual datum in machine learning. It can effectively identify helpful or harmful data points for a learning algorithm. In this paper, we propose Beta Shapley, which is a substantial generalization of Data Shapley. Beta Shapley arises naturally by relaxing the efficiency axiom of the Shapley value, which is not critical for machine learning settings. Beta Shapley unifies several popular data valuation methods and includes data Shapley as a special case. Moreover, we prove that Beta Shapley has several desirable statistical properties and propose efficient algorithms to estimate it. We demonstrate that Beta Shapley outperforms state-of-the-art data valuation methods on several downstream ML tasks such as: 1) detecting mislabeled training data; 2) learning with subsamples; and 3) identifying points whose addition or removal have the largest positive or negative impact on the model."
                    },
                    {
                        "title": "WeightedSHAP: analyzing and improving Shapley based feature attributions",
                        "abstract": "Shapley value is a popular approach for measuring the influence of individual features. While Shapley feature attribution is built upon desiderata from game theory, some of its constraints may be less natural in certain machine learning settings, leading to unintuitive model interpretation. In particular, the Shapley value uses the same weight for all marginal contributions -- i.e. it gives the same importance when a large number of other features are given versus when a small number of other features are given. This property can be problematic if larger feature sets are more or less informative than smaller feature sets. Our work performs a rigorous analysis of the potential limitations of Shapley feature attribution. We identify simple settings where the Shapley value is mathematically suboptimal by assigning larger attributions for less influential features. Motivated by this observation, we propose WeightedSHAP, which generalizes the Shapley value and learns which marginal contributions to focus directly from data. On several real-world datasets, we demonstrate that the influential features identified by WeightedSHAP are better able to recapitulate the model's predictions compared to the features identified by the Shapley value."
                    },
                    {
                        "title": "Data-OOB: Out-of-bag Estimate as a Simple and Efficient Data Value",
                        "abstract": "Data valuation is a powerful framework for providing statistical insights into which data are beneficial or detrimental to model training. Many Shapley-based data valuation methods have shown promising results in various downstream tasks, however, they are well known to be computationally challenging as it requires training a large number of models. As a result, it has been recognized as infeasible to apply to large datasets. To address this issue, we propose Data-OOB, a new data valuation method for a bagging model that utilizes the out-of-bag estimate. The proposed method is computationally efficient and can scale to millions of data by reusing trained weak learners. Specifically, Data-OOB takes less than 2.25 hours on a single CPU processor when there are $10^6$ samples to evaluate and the input dimension is 100. Furthermore, Data-OOB has solid theoretical interpretations in that it identifies the same important data point as the infinitesimal jackknife influence function when two different points are compared. We conduct comprehensive experiments using 12 classification datasets, each with thousands of sample sizes. We demonstrate that the proposed method significantly outperforms existing state-of-the-art data valuation methods in identifying mislabeled data and finding a set of helpful (or harmful) data points, highlighting the potential for applying data values in real-world applications."
                    },
                    {
                        "title": "Distributionally Robust Instrumental Variables Estimation",
                        "abstract": "Instrumental variables (IV) estimation is a fundamental method in econometrics and statistics for estimating causal effects in the presence of unobserved confounding. However, challenges such as untestable model assumptions and poor finite sample properties have undermined its reliability in practice. Viewing common issues in IV estimation as distributional uncertainties, we propose DRIVE, a distributionally robust framework of the classical IV estimation method. When the ambiguity set is based on a Wasserstein distance, DRIVE minimizes a square root ridge regularized variant of the two stage least squares (TSLS) objective. We develop a novel asymptotic theory for this regularized regression estimator based on the square root ridge, showing that it achieves consistency without requiring the regularization parameter to vanish. This result follows from a fundamental property of the square root ridge, which we call ``delayed shrinkage''. This novel property, which also holds for a class of generalized method of moments (GMM) estimators, ensures that the estimator is robust to distributional uncertainties that persist in large samples. We further derive the asymptotic distribution of Wasserstein DRIVE and propose data-driven procedures to select the regularization parameter based on theoretical results. Simulation studies confirm the superior finite sample performance of Wasserstein DRIVE. Thanks to its regularization and robustness properties, Wasserstein DRIVE could be preferable in practice, particularly when the practitioner is uncertain about model assumptions or distributional shifts in data."
                    },
                    {
                        "title": "Efficient computation and analysis of distributional Shapley values",
                        "abstract": "Distributional data Shapley value (DShapley) has recently been proposed as a principled framework to quantify the contribution of individual datum in machine learning. DShapley develops the foundational game theory concept of Shapley values into a statistical framework and can be applied to identify data points that are useful (or harmful) to a learning algorithm. Estimating DShapley is computationally expensive, however, and this can be a major challenge to using it in practice. Moreover, there has been little mathematical analyses of how this value depends on data characteristics. In this paper, we derive the first analytic expressions for DShapley for the canonical problems of linear regression, binary classification, and non-parametric density estimation. These analytic forms provide new algorithms to estimate DShapley that are several orders of magnitude faster than previous state-of-the-art methods. Furthermore, our formulas are directly interpretable and provide quantitative insights into how the value varies for different types of data. We demonstrate the practical efficacy of our approach on multiple real and synthetic datasets."
                    },
                    {
                        "title": "Competition over data: how does data purchase affect users?",
                        "abstract": "As machine learning (ML) is deployed by many competing service providers, the underlying ML predictors also compete against each other, and it is increasingly important to understand the impacts and biases from such competition. In this paper, we study what happens when the competing predictors can acquire additional labeled data to improve their prediction quality. We introduce a new environment that allows ML predictors to use active learning algorithms to purchase labeled data within their budgets while competing against each other to attract users. Our environment models a critical aspect of data acquisition in competing systems which has not been well-studied before. We found that the overall performance of an ML predictor improves when predictors can purchase additional labeled data. Surprisingly, however, the quality that users experience -- i.e. the accuracy of the predictor selected by each user -- can decrease even as the individual predictors get better. We show that this phenomenon naturally arises due to a trade-off whereby competition pushes each predictor to specialize in a subset of the population while data purchase has the effect of making predictors more uniform. We support our findings with both experiments and theories."
                    },
                    {
                        "title": "DataInf: Efficiently Estimating Data Influence in LoRA-tuned LLMs and Diffusion Models",
                        "abstract": "Quantifying the impact of training data points is crucial for understanding the outputs of machine learning models and for improving the transparency of the AI pipeline. The influence function is a principled and popular data attribution method, but its computational cost often makes it challenging to use. This issue becomes more pronounced in the setting of large language models and text-to-image models. In this work, we propose DataInf, an efficient influence approximation method that is practical for large-scale generative AI models. Leveraging an easy-to-compute closed-form expression, DataInf outperforms existing influence computation algorithms in terms of computational and memory efficiency. Our theoretical analysis shows that DataInf is particularly well-suited for parameter-efficient fine-tuning techniques such as LoRA. Through systematic empirical evaluations, we show that DataInf accurately approximates influence scores and is orders of magnitude faster than existing methods. In applications to RoBERTa-large, Llama-2-13B-chat, and stable-diffusion-v1.5 models, DataInf effectively identifies the most influential fine-tuning examples better than other approximate influence scores. Moreover, it can help to identify which data points are mislabeled."
                    },
                    {
                        "title": "Uncertainty quantification of molecular property prediction using Bayesian neural network models",
                        "abstract": "In chemistry, deep neural network models have been increasingly utilized in a variety of applications such as molecular property predictions, novel molecule designs, and planning chemical reactions. Despite the rapid increase in the use of state-of-the-art models and algorithms, deep neural network models often produce poor predictions in real applications because model performance is highly dependent on the quality of training data. In the field of molecular analysis, data are mostly obtained from either complicated chemical experiments or approximate mathematical equations, and then quality of data may be questioned.In this paper, we quantify uncertainties of prediction using Bayesian neural networks in molecular property predictions. We estimate both model-driven and data-driven uncertainties, demonstrating the usefulness of uncertainty quantification as both a quality checker and a confidence indicator with the three experiments. Our results manifest that uncertainty quantification is necessary for more reliable molecular applications and Bayesian neural network models can be a practical approach."
                    },
                    {
                        "title": "TimeInf: Time Series Data Contribution via Influence Functions",
                        "abstract": "Evaluating the contribution of individual data points to a model's prediction is critical for interpreting model predictions and improving model performance. Existing data contribution methods have been applied to various data types, including tabular data, images, and texts; however, their primary focus has been on i.i.d. settings. Despite the pressing need for principled approaches tailored to time series datasets, the problem of estimating data contribution in such settings remains unexplored, possibly due to challenges associated with handling inherent temporal dependencies. This paper introduces TimeInf, a data contribution estimation method for time-series datasets. TimeInf uses influence functions to attribute model predictions to individual time points while preserving temporal structures. Our extensive empirical results demonstrate that TimeInf outperforms state-of-the-art methods in identifying harmful anomalies and helpful time points for forecasting. Additionally, TimeInf offers intuitive and interpretable attributions of data values, allowing us to easily distinguish diverse anomaly patterns through visualizations."
                    },
                    {
                        "title": "Group Shapley Value and Counterfactual Simulations in a Structural Model",
                        "abstract": "We propose a variant of the Shapley value, the group Shapley value, to interpret counterfactual simulations in structural economic models by quantifying the importance of different components. Our framework compares two sets of parameters, partitioned into multiple groups, and applying group Shapley value decomposition yields unique additive contributions to the changes between these sets. The relative contributions sum to one, enabling us to generate an importance table that is as easily interpretable as a regression table. The group Shapley value can be characterized as the solution to a constrained weighted least squares problem. Using this property, we develop robust decomposition methods to address scenarios where inputs for the group Shapley value are missing. We first apply our methodology to a simple Roy model and then illustrate its usefulness by revisiting two published papers."
                    },
                    {
                        "title": "Competing AI: How does competition feedback affect machine learning?",
                        "abstract": "This papers studies how competition affects machine learning (ML) predictors. As ML becomes more ubiquitous, it is often deployed by companies to compete over customers. For example, digital platforms like Yelp use ML to predict user preference and make recommendations. A service that is more often queried by users, perhaps because it more accurately anticipates user preferences, is also more likely to obtain additional user data (e.g. in the form of a Yelp review). Thus, competing predictors cause feedback loops whereby a predictor's performance impacts what training data it receives and biases its predictions over time. We introduce a flexible model of competing ML predictors that enables both rapid experimentation and theoretical tractability. We show with empirical and mathematical analysis that competition causes predictors to specialize for specific sub-populations at the cost of worse performance over the general population. We further analyze the impact of predictor specialization on the overall prediction quality experienced by users. We show that having too few or too many competing predictors in a market can hurt the overall prediction quality. Our theory is complemented by experiments on several real datasets using popular learning algorithms, such as neural networks and nearest neighbor methods."
                    },
                    {
                        "title": "Principled analytic classifier for positive-unlabeled learning via weighted integral probability metric",
                        "abstract": "We consider the problem of learning a binary classifier from only positive and unlabeled observations (called PU learning). Recent studies in PU learning have shown superior performance theoretically and empirically. However, most existing algorithms may not be suitable for large-scale datasets because they face repeated computations of a large Gram matrix or require massive hyperparameter optimization. In this paper, we propose a computationally efficient and theoretically grounded PU learning algorithm. The proposed PU learning algorithm produces a closed-form classifier when the hypothesis space is a closed ball in reproducing kernel Hilbert space. In addition, we establish upper bounds of the estimation error and the excess risk. The obtained estimation error bound is sharper than existing results and the derived excess risk bound has an explicit form, which vanishes as sample sizes increase. Finally, we conduct extensive numerical experiments using both synthetic and real datasets, demonstrating improved accuracy, scalability, and robustness of the proposed algorithm."
                    },
                    {
                        "title": "OpenDataVal: a Unified Benchmark for Data Valuation",
                        "abstract": "Assessing the quality and impact of individual data points is critical for improving model performance and mitigating undesirable biases within the training dataset. Several data valuation algorithms have been proposed to quantify data quality, however, there lacks a systemic and standardized benchmarking system for data valuation. In this paper, we introduce OpenDataVal, an easy-to-use and unified benchmark framework that empowers researchers and practitioners to apply and compare various data valuation algorithms. OpenDataVal provides an integrated environment that includes (i) a diverse collection of image, natural language, and tabular datasets, (ii) implementations of eleven different state-of-the-art data valuation algorithms, and (iii) a prediction model API that can import any models in scikit-learn. Furthermore, we propose four downstream machine learning tasks for evaluating the quality of data values. We perform benchmarking analysis using OpenDataVal, quantifying and comparing the efficacy of state-of-the-art data valuation approaches. We find that no single algorithm performs uniformly best across all tasks, and an appropriate algorithm should be employed for a user's downstream task. OpenDataVal is publicly available at https://opendataval.github.io with comprehensive documentation. Furthermore, we provide a leaderboard where researchers can evaluate the effectiveness of their own data valuation algorithms."
                    },
                    {
                        "title": "Uncertainty quantification of molecular property prediction with Bayesian neural networks",
                        "abstract": "Deep neural networks have outperformed existing machine learning models in various molecular applications. In practical applications, it is still difficult to make confident decisions because of the uncertainty in predictions arisen from insufficient quality and quantity of training data. Here, we show that Bayesian neural networks are useful to quantify the uncertainty of molecular property prediction with three numerical experiments. In particular, it enables us to decompose the predictive variance into the model- and data-driven uncertainties, which helps to elucidate the source of errors. In the logP predictions, we show that data noise affected the data-driven uncertainties more significantly than the model-driven ones. Based on this analysis, we were able to find unexpected errors in the Harvard Clean Energy Project dataset. Lastly, we show that the confidence of prediction is closely related to the predictive uncertainty by performing on bio-activity and toxicity classification problems."
                    },
                    {
                        "title": "Mind the Gap: Understanding the Modality Gap in Multi-modal Contrastive Representation Learning",
                        "abstract": "We present modality gap, an intriguing geometric phenomenon of the representation space of multi-modal models. Specifically, we show that different data modalities (e.g. images and text) are embedded at arm's length in their shared representation in multi-modal models such as CLIP. Our systematic analysis demonstrates that this gap is caused by a combination of model initialization and contrastive learning optimization. In model initialization, we show empirically and theoretically that the representation of a common deep neural network is restricted to a narrow cone. As a consequence, in a multi-modal model with two encoders, the representations of the two modalities are clearly apart when the model is initialized. During optimization, contrastive learning keeps the different modalities separate by a certain distance, which is influenced by the temperature parameter in the loss function. Our experiments further demonstrate that varying the modality gap distance has a significant impact in improving the model's downstream zero-shot classification performance and fairness. Our code and data are available at https://modalitygap.readthedocs.io/"
                    },
                    {
                        "title": "Accuracy on the Curve: On the Nonlinear Correlation of ML Performance Between Data Subpopulations",
                        "abstract": "Understanding the performance of machine learning (ML) models across diverse data distributions is critically important for reliable applications. Despite recent empirical studies positing a near-perfect linear correlation between in-distribution (ID) and out-of-distribution (OOD) accuracies, we empirically demonstrate that this correlation is more nuanced under subpopulation shifts. Through rigorous experimentation and analysis across a variety of datasets, models, and training epochs, we demonstrate that OOD performance often has a nonlinear correlation with ID performance in subpopulation shifts. Our findings, which contrast previous studies that have posited a linear correlation in model performance during distribution shifts, reveal a \"moon shape\" correlation (parabolic uptrend curve) between the test performance on the majority subpopulation and the minority subpopulation. This non-trivial nonlinear correlation holds across model architectures, hyperparameters, training durations, and the imbalance between subpopulations. Furthermore, we found that the nonlinearity of this \"moon shape\" is causally influenced by the degree of spurious correlations in the training data. Our controlled experiments show that stronger spurious correlation in the training data creates more nonlinear performance correlation. We provide complementary experimental and theoretical analyses for this phenomenon, and discuss its implications for ML reliability and fairness. Our work highlights the importance of understanding the nonlinear effects of model improvement on performance in different subpopulations, and has the potential to inform the development of more equitable and responsible machine learning models."
                    },
                    {
                        "title": "Truthful Dataset Valuation by Pointwise Mutual Information",
                        "abstract": "A common way to evaluate a dataset in ML involves training a model on this dataset and assessing the model's performance on a test set. However, this approach has two issues: (1) it may incentivize undesirable data manipulation in data marketplaces, as the self-interested data providers seek to modify the dataset to maximize their evaluation scores; (2) it may select datasets that overfit to potentially small test sets. We propose a new data valuation method that provably guarantees the following: data providers always maximize their expected score by truthfully reporting their observed data. Any manipulation of the data, including but not limited to data duplication, adding random data, data removal, or re-weighting data from different groups, cannot increase their expected score. Our method, following the paradigm of proper scoring rules, measures the pointwise mutual information (PMI) of the test dataset and the evaluated dataset. However, computing the PMI of two datasets is challenging. We introduce a novel PMI measuring method that greatly improves tractability within Bayesian machine learning contexts. This is accomplished through a new characterization of PMI that relies solely on the posterior probabilities of the model parameter at an arbitrarily selected value. Finally, we support our theoretical results with simulations and further test the effectiveness of our data valuation method in identifying the top datasets among multiple data providers. Interestingly, our method outperforms the standard approach of selecting datasets based on the trained model's test performance, suggesting that our truthful valuation score can also be more robust to overfitting."
                    },
                    {
                        "title": "Principled learning method for Wasserstein distributionally robust optimization with local perturbations",
                        "abstract": "Wasserstein distributionally robust optimization (WDRO) attempts to learn a model that minimizes the local worst-case risk in the vicinity of the empirical data distribution defined by Wasserstein ball. While WDRO has received attention as a promising tool for inference since its introduction, its theoretical understanding has not been fully matured. Gao et al. (2017) proposed a minimizer based on a tractable approximation of the local worst-case risk, but without showing risk consistency. In this paper, we propose a minimizer based on a novel approximation theorem and provide the corresponding risk consistency results. Furthermore, we develop WDRO inference for locally perturbed data that include the Mixup (Zhang et al., 2017) as a special case. We show that our approximation and risk consistency results naturally extend to the cases when data are locally perturbed. Numerical experiments demonstrate robustness of the proposed method using image classification datasets. Our results show that the proposed method achieves significantly higher accuracy than baseline models on noisy datasets."
                    },
                    {
                        "title": "Rethinking Data Shapley for Data Selection Tasks: Misleads and Merits",
                        "abstract": "Data Shapley provides a principled approach to data valuation and plays a crucial role in data-centric machine learning (ML) research. Data selection is considered a standard application of Data Shapley. However, its data selection performance has shown to be inconsistent across settings in the literature. This study aims to deepen our understanding of this phenomenon. We introduce a hypothesis testing framework and show that Data Shapley's performance can be no better than random selection without specific constraints on utility functions. We identify a class of utility functions, monotonically transformed modular functions, within which Data Shapley optimally selects data. Based on this insight, we propose a heuristic for predicting Data Shapley's effectiveness in data selection tasks. Our experiments corroborate these findings, adding new insights into when Data Shapley may or may not succeed."
                    }
                ]
            }
        }
    },
    "2406.14408": {
        "paper_data": {
            "title": "FVEL: Interactive Formal Verification Environment with Large Language Models via Theorem Proving",
            "url": "http://arxiv.org/abs/2406.14408v2",
            "arxiv_id": "2406.14408",
            "authors": [
                "Xiaohan Lin",
                "Qingxing Cao",
                "Yinya Huang",
                "Haiming Wang",
                "Jianqiao Lu",
                "Zhengying Liu",
                "Linqi Song",
                "Xiaodan Liang"
            ],
            "abstract": "Formal verification (FV) has witnessed growing significance with current emerging program synthesis by the evolving large language models (LLMs). However, current formal verification mainly resorts to symbolic verifiers or hand-craft rules, resulting in limitations for extensive and flexible verification. On the other hand, formal languages for automated theorem proving, such as Isabelle, as another line of rigorous verification, are maintained with comprehensive rules and theorems. In this paper, we propose FVEL, an interactive Formal Verification Environment with LLMs. Specifically, FVEL transforms a given code to be verified into Isabelle, and then conducts verification via neural automated theorem proving with an LLM. The joined paradigm leverages the rigorous yet abundant formulated and organized rules in Isabelle and is also convenient for introducing and adjusting cutting-edge LLMs. To achieve this goal, we extract a large-scale FVELER3. The FVELER dataset includes code dependencies and verification processes that are formulated in Isabelle, containing 758 theories, 29,125 lemmas, and 200,646 proof steps in total with in-depth dependencies. We benchmark FVELER in the FVEL environment by first fine-tuning LLMs with FVELER and then evaluating them on Code2Inv and SV-COMP. The results show that FVEL with FVELER fine-tuned Llama3- 8B solves 17.39% (69 -> 81) more problems, and Mistral-7B 12% (75 -> 84) more problems in SV-COMP. And the proportion of proof errors is reduced. Project page: https://fveler.github.io/.",
            "introduction": "   1 Introduction  Formal verification (FV), or automated program verification [35, 2] checks if a code meets a specific demand and is correct to implement. As the code synthesis ability of current models [28, 25, 9] evolves rapidly, there is a growing demand for automated verification of diverse and abundant synthesis programs. However, current formal verification mainly resorts to symbolic verifiers [11, 6, 22] or hand-craft rules [39]. However, symbolic verification can not leverage the advanced reasoning ability of current large language models (LLMs), while hand-craft rules with limited execution on specific code cases have restricted abilities to general verification.   On the other hand, automated theorem proving (ATP) [42, 1, 14] is a line of work on rigorous verification with formal languages (e.g., Isabelle [30], Lean [5]) and interactive proof environments (e.g., PISA [16], LeanDojo [40]). Such formal languages and toolkits maintain corresponding libraries with a large number of human-written and checked theorems and rules, which are provided as pre-training materials for many large language models [28, 36, 15]. The ATP formulation and rules have strong expressiveness and, therefore have a great potential for describing formal verification problems and requests. As a result, the verification can be implemented under a rigorous, step-wise, and interactive ATP environment. Moreover, the pre-trained formal reasoning capabilities within LLMs and their potential to solve formal verification problems are underexplored.   To take one step towards this goal, this paper proposes FVEL, a new formal verification environment interacting with LLMs via automated theorem proving processes. Figure\u00a01 demonstrates an overview of FVEL. Specifically, the FVEL\u00a0environment takes as input a code to be verified, converts the code into Isabelle formulation, and generates a lemma in Isabelle followed by a whole proof to the lemma. FVEL\u00a0then outputs the proof result (succeed or failed being proved) as an indication of the code verification result. FVEL\u00a0interacts with an LLM by initially providing the converted Isabelle formulation to the LLM and then receiving the derived lemma on the code specification. The interaction is then continued by the LLM generating proof states and the FVEL\u00a0environment providing feedback via prover information in the PISA environment [16], such as cheating keywords sorry or opps and other error messages. As a result, a user provides her code to be verified to FVEL, and then she will receive the verification result and intermediate proving information. Note that we follow previous works [6, 39] to investigate FVEL\u00a0on C code verification in this paper. We remain the extension of FVEL\u00a0to support more program languages as a near future work.   To implement the FVEL\u00a0environment, we extract and cleanse a large-scale FVELer\u00a0dataset with deep dependencies, which can be applied as both a fine-tuning resource and evaluation benchmark. The FVELer\u00a0dataset has two main components: C code dependencies formulated by Isabelle theories, and Isabelle lemmas with their step-wise proof states. FVELer\u00a0then includes 758 theories with 29,125 lemmas and 200,646 proof steps. The dataset is then randomly split according to lemmas, resulting in training/validation/test/test-hard sets. The test-hard set data have dependencies that are challenging to find. Statistical analysis shows that FVELer\u00a0data comprehensively covers diverse dependency depths and has a remarkable number of data with very deep dependencies. For example, over 50% of lemmas have a depth greater than 78,",
            "references": [
                {
                    "title": "A Survey on Deep Learning for Theorem Proving",
                    "abstract": "Theorem proving is a fundamental aspect of mathematics, spanning from informal reasoning in natural language to rigorous derivations in formal systems. In recent years, the advancement of deep learning, especially the emergence of large language models, has sparked a notable surge of research exploring these techniques to enhance the process of theorem proving. This paper presents a comprehensive survey of deep learning for theorem proving by offering (i) a thorough review of existing approaches across various tasks such as autoformalization, premise selection, proofstep generation, and proof search; (ii) an extensive summary of curated datasets and strategies for synthetic data generation; (iii) a detailed analysis of evaluation metrics and the performance of state-of-the-art methods; and (iv) a critical discussion on the persistent challenges and the promising avenues for future exploration. Our survey aims to serve as a foundational reference for deep learning approaches in theorem proving, inspiring and catalyzing further research endeavors in this rapidly growing field. A curated list of papers is available at https://github.com/zhaoyu-li/DL4TP."
                },
                {
                    "title": "LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models",
                    "abstract": "Efficient fine-tuning is vital for adapting large language models (LLMs) to downstream tasks. However, it requires non-trivial efforts to implement these methods on different models. We present LlamaFactory, a unified framework that integrates a suite of cutting-edge efficient training methods. It provides a solution for flexibly customizing the fine-tuning of 100+ LLMs without the need for coding through the built-in web UI LlamaBoard. We empirically validate the efficiency and effectiveness of our framework on language modeling and text generation tasks. It has been released at https://github.com/hiyouga/LLaMA-Factory and received over 25,000 stars and 3,000 forks."
                },
                {
                    "title": "MUSTARD: Mastering Uniform Synthesis of Theorem and Proof Data",
                    "abstract": "Recent large language models (LLMs) have witnessed significant advancement in various tasks, including mathematical reasoning and theorem proving. As these two tasks require strict and formal multi-step inference, they are appealing domains for exploring the reasoning ability of LLMs but still face important challenges. Previous studies such as Chain-of-Thought (CoT) have revealed the effectiveness of intermediate steps guidance. However, such step-wise annotation requires heavy labor, leading to insufficient training steps for current benchmarks. To fill this gap, this work introduces MUSTARD, a data generation framework that masters uniform synthesis of theorem and proof data of high quality and diversity. MUSTARD synthesizes data in three stages: (1) It samples a few mathematical concept seeds as the problem category. (2) Then, it prompts a generative language model with the sampled concepts to obtain both the problems and their step-wise formal solutions. (3) Lastly, the framework utilizes a proof assistant (e.g., Lean Prover) to filter the valid proofs. With the proposed MUSTARD, we present a theorem-and-proof benchmark MUSTARDSAUCE with 5,866 valid data points. Each data point contains an informal statement, an informal proof, and a translated formal proof that passes the prover validation. We perform extensive analysis and demonstrate that MUSTARD generates validated high-quality step-by-step data. We further apply the MUSTARDSAUCE for fine-tuning smaller language models. The fine-tuned Llama 2-7B achieves a 15.41% average relative performance gain in automated theorem proving, and 8.18% in math word problems. Codes and data are available at https://github.com/Eleanor-H/MUSTARD."
                },
                {
                    "title": "DeepSeek-Coder: When the Large Language Model Meets Programming - The Rise of Code Intelligence",
                    "abstract": "The rapid development of large language models has revolutionized code intelligence in software development. However, the predominance of closed-source models has restricted extensive research and development. To address this, we introduce the DeepSeek-Coder series, a range of open-source code models with sizes from 1.3B to 33B, trained from scratch on 2 trillion tokens. These models are pre-trained on a high-quality project-level code corpus and employ a fill-in-the-blank task with a 16K window to enhance code generation and infilling. Our extensive evaluations demonstrate that DeepSeek-Coder not only achieves state-of-the-art performance among open-source code models across multiple benchmarks but also surpasses existing closed-source models like Codex and GPT-3.5. Furthermore, DeepSeek-Coder models are under a permissive license that allows for both research and unrestricted commercial use."
                },
                {
                    "title": "Finding Inductive Loop Invariants using Large Language Models",
                    "abstract": "Loop invariants are fundamental to reasoning about programs with loops. They establish properties about a given loop's behavior. When they additionally are inductive, they become useful for the task of formal verification that seeks to establish strong mathematical guarantees about program's runtime behavior. The inductiveness ensures that the invariants can be checked locally without consulting the entire program, thus are indispensable artifacts in a formal proof of correctness. Finding inductive loop invariants is an undecidable problem, and despite a long history of research towards practical solutions, it remains far from a solved problem. This paper investigates the capabilities of the Large Language Models (LLMs) in offering a new solution towards this old, yet important problem. To that end, we first curate a dataset of verification problems on programs with loops. Next, we design a prompt for exploiting LLMs, obtaining inductive loop invariants, that are checked for correctness using sound symbolic tools. Finally, we explore the effectiveness of using an efficient combination of a symbolic tool and an LLM on our dataset and compare it against a purely symbolic baseline. Our results demonstrate that LLMs can help improve the state-of-the-art in automated program verification."
                },
                {
                    "title": "Lemur: Integrating Large Language Models in Automated Program Verification",
                    "abstract": "The demonstrated code-understanding capability of LLMs raises the question of whether they can be used for automated program verification, a task that demands high-level abstract reasoning about program properties that is challenging for verification tools. We propose a general methodology to combine the power of LLMs and automated reasoners for automated program verification. We formally describe this methodology as a set of transition rules and prove its soundness. We instantiate the calculus as a sound automated verification procedure and demonstrate practical improvements on a set of synthetic and competition benchmarks."
                },
                {
                    "title": "LEGO-Prover: Neural Theorem Proving with Growing Libraries",
                    "abstract": "Despite the success of large language models (LLMs), the task of theorem proving still remains one of the hardest reasoning tasks that is far from being fully solved. Prior methods using language models have demonstrated promising results, but they still struggle to prove even middle school level theorems. One common limitation of these methods is that they assume a fixed theorem library during the whole theorem proving process. However, as we all know, creating new useful theorems or even new theories is not only helpful but crucial and necessary for advancing mathematics and proving harder and deeper results. In this work, we present LEGO-Prover, which employs a growing skill library containing verified lemmas as skills to augment the capability of LLMs used in theorem proving. By constructing the proof modularly, LEGO-Prover enables LLMs to utilize existing skills retrieved from the library and to create new skills during the proving process. These skills are further evolved (by prompting an LLM) to enrich the library on another scale. Modular and reusable skills are constantly added to the library to enable tackling increasingly intricate mathematical problems. Moreover, the learned library further bridges the gap between human proofs and formal proofs by making it easier to impute missing steps. LEGO-Prover advances the state-of-the-art pass rate on miniF2F-valid (48.0% to 57.0%) and miniF2F-test (45.5% to 47.1%). During the proving process, LEGO-Prover also manages to generate over 20,000 skills (theorems/lemmas) and adds them to the growing library. Our ablation study indicates that these newly added skills are indeed helpful for proving theorems, resulting in an improvement from a success rate of 47.1% to 50.4%. We also release our code and all the generated skills."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Loop Invariant Inference through SMT Solving Enhanced Reinforcement Learning",
                    "abstract": "Inferring loop invariants is one of the most challenging problems in program verification. It is highly desired to incorporate machine learning when inferring. This paper presents a Reinforcement Learning (RL) pruning framework to infer loop invariants over a general nonlinear hypothesis space. The key idea is to synergize the RL-based pruning and SMT solving to generate candidate invariants efficiently. To address the sparse reward problem in learning, we design a novel two-dimensional reward mechanism that enables the RL pruner to recognize the capability boundary of SMT solvers and learn the pruning heuristics in a few rounds. We have implemented our approach with Z3 SMT solver in the tool called LIPuS and conducted extensive experiments over the linear and nonlinear benchmarks. Experiment results show that LIPuS can solve the most cases compared to the state-of-the-art loop invariant inference tools such as Code2Inv, ICE-DT, GSpacer, SymInfer, ImplCheck, and Eldarica. Especially, LIPuS outperforms them significantly on nonlinear benchmarks."
                },
                {
                    "title": "LeanDojo: Theorem Proving with Retrieval-Augmented Language Models",
                    "abstract": "Large language models (LLMs) have shown promise in proving formal theorems using proof assistants such as Lean. However, existing methods are difficult to reproduce or build on, due to private code, data, and large compute requirements. This has created substantial barriers to research on machine learning methods for theorem proving. This paper removes these barriers by introducing LeanDojo: an open-source Lean playground consisting of toolkits, data, models, and benchmarks. LeanDojo extracts data from Lean and enables interaction with the proof environment programmatically. It contains fine-grained annotations of premises in proofs, providing valuable data for premise selection: a key bottleneck in theorem proving. Using this data, we develop ReProver (Retrieval-Augmented Prover): an LLM-based prover augmented with retrieval for selecting premises from a vast math library. It is inexpensive and needs only one GPU week of training. Our retriever leverages LeanDojo's program analysis capability to identify accessible premises and hard negative examples, which makes retrieval much more effective. Furthermore, we construct a new benchmark consisting of 98,734 theorems and proofs extracted from Lean's math library. It features challenging data split requiring the prover to generalize to theorems relying on novel premises that are never used in training. We use this benchmark for training and evaluation, and experimental results demonstrate the effectiveness of ReProver over non-retrieval baselines and GPT-4. We thus provide the first set of open-source LLM-based theorem provers without any proprietary datasets and release it under a permissive MIT license to facilitate further research."
                },
                {
                    "title": "WizardCoder: Empowering Code Large Language Models with Evol-Instruct",
                    "abstract": "Code Large Language Models (Code LLMs), such as StarCoder, have demonstrated exceptional performance in code-related tasks. However, most existing models are solely pre-trained on extensive raw code data without instruction fine-tuning. In this paper, we introduce WizardCoder, which empowers Code LLMs with complex instruction fine-tuning, by adapting the Evol-Instruct method to the domain of code. Through comprehensive experiments on four prominent code generation benchmarks, namely HumanEval, HumanEval+, MBPP, and DS-1000, we unveil the exceptional capabilities of our model. It surpasses all other open-source Code LLMs by a substantial margin. Moreover, our model even outperforms the largest closed LLMs, Anthropic's Claude and Google's Bard, on HumanEval and HumanEval+. Our code, model weights, and data are public at https://github.com/nlpxucan/WizardLM"
                },
                {
                    "title": "Thor: Wielding Hammers to Integrate Language Models and Automated Theorem Provers",
                    "abstract": "In theorem proving, the task of selecting useful premises from a large library to unlock the proof of a given conjecture is crucially important. This presents a challenge for all theorem provers, especially the ones based on language models, due to their relative inability to reason over huge volumes of premises in text form. This paper introduces Thor, a framework integrating language models and automated theorem provers to overcome this difficulty. In Thor, a class of methods called hammers that leverage the power of automated theorem provers are used for premise selection, while all other tasks are designated to language models. Thor increases a language model's success rate on the PISA dataset from $39\\%$ to $57\\%$, while solving $8.2\\%$ of problems neither language models nor automated theorem provers are able to solve on their own. Furthermore, with a significantly smaller computational budget, Thor can achieve a success rate on the MiniF2F dataset that is on par with the best existing methods. Thor can be instantiated for the majority of popular interactive theorem provers via a straightforward protocol we provide."
                },
                {
                    "title": "MiniF2F: a cross-system benchmark for formal Olympiad-level mathematics",
                    "abstract": "We present miniF2F, a dataset of formal Olympiad-level mathematics problems statements intended to provide a unified cross-system benchmark for neural theorem proving. The miniF2F benchmark currently targets Metamath, Lean, Isabelle (partially) and HOL Light (partially) and consists of 488 problem statements drawn from the AIME, AMC, and the International Mathematical Olympiad (IMO), as well as material from high-school and undergraduate mathematics courses. We report baseline results using GPT-f, a neural theorem prover based on GPT-3 and provide an analysis of its performance. We intend for miniF2F to be a community-driven effort and hope that our benchmark will help spur advances in neural theorem proving."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Proof Artifact Co-training for Theorem Proving with Language Models",
                    "abstract": "Labeled data for imitation learning of theorem proving in large libraries of formalized mathematics is scarce as such libraries require years of concentrated effort by human specialists to be built. This is particularly challenging when applying large Transformer language models to tactic prediction, because the scaling of performance with respect to model size is quickly disrupted in the data-scarce, easily-overfitted regime. We propose PACT ({\\bf P}roof {\\bf A}rtifact {\\bf C}o-{\\bf T}raining), a general methodology for extracting abundant self-supervised data from kernel-level proof terms for co-training alongside the usual tactic prediction objective. We apply this methodology to Lean, an interactive proof assistant which hosts some of the most sophisticated formalized mathematics to date. We instrument Lean with a neural theorem prover driven by a Transformer language model and show that PACT improves theorem proving success rate on a held-out suite of test theorems from 32\\% to 48\\%."
                },
                {
                    "title": "Generative Language Modeling for Automated Theorem Proving",
                    "abstract": "We explore the application of transformer-based language models to automated theorem proving. This work is motivated by the possibility that a major limitation of automated theorem provers compared to humans -- the generation of original mathematical terms -- might be addressable via generation from language models. We present an automated prover and proof assistant, GPT-f, for the Metamath formalization language, and analyze its performance. GPT-f found new short proofs that were accepted into the main Metamath library, which is to our knowledge, the first time a deep-learning based system has contributed proofs that were adopted by a formal mathematics community."
                },
                {
                    "title": "The lean mathematical library",
                    "abstract": "This paper describes mathlib, a community-driven effort to build a unified library of mathematics formalized in the Lean proof assistant. Among proof assistant libraries, it is distinguished by its dependently typed foundations, focus on classical mathematics, extensive hierarchy of structures, use of large- and small-scale automation, and distributed organization. We explain the architecture and design decisions of the library and the social organization that has led to its development."
                },
                {
                    "title": "CLN2INV: Learning Loop Invariants with Continuous Logic Networks",
                    "abstract": "Program verification offers a framework for ensuring program correctness and therefore systematically eliminating different classes of bugs. Inferring loop invariants is one of the main challenges behind automated verification of real-world programs which often contain many loops. In this paper, we present Continuous Logic Network (CLN), a novel neural architecture for automatically learning loop invariants directly from program execution traces. Unlike existing neural networks, CLNs can learn precise and explicit representations of formulas in Satisfiability Modulo Theories (SMT) for loop invariants from program execution traces. We develop a new sound and complete semantic mapping for assigning SMT formulas to continuous truth values that allows CLNs to be trained efficiently. We use CLNs to implement a new inference system for loop invariants, CLN2INV, that significantly outperforms existing approaches on the popular Code2Inv dataset. CLN2INV is the first tool to solve all 124 theoretically solvable problems in the Code2Inv dataset. Moreover, CLN2INV takes only 1.1 seconds on average for each problem, which is 40 times faster than existing approaches. We further demonstrate that CLN2INV can even learn 12 significantly more complex loop invariants than the ones required for the Code2Inv dataset."
                },
                {
                    "title": "ESBMC 5.0: An Industrial-Strength C Model Checker",
                    "abstract": "ESBMC is a mature, permissively licensed open-source context-bounded model checker for the verification of single- and multithreaded C programs. It can verify both predefined safety properties (e.g., bounds check, pointer safety, overflow) and user-defined program assertions automatically. ESBMC provides C++ and Python APIs to access internal data structures, allowing inspection and extension at any stage of the verification process. We discuss improvements over previous versions of ESBMC, including the description of new front- and back-ends, IEEE floating-point support, and an improved k-induction algorithm. A demonstration is available at https://www.youtube.com/watch?v=YcJjXHlN1v8."
                },
                {
                    "title": "Datasheets for datasets",
                    "abstract": "Documentation to facilitate communication between dataset creators and consumers."
                },
                {
                    "title": "Learning Invariants using Decision Trees",
                    "abstract": "The problem of inferring an inductive invariant for verifying program safety can be formulated in terms of binary classification. This is a standard problem in machine learning: given a sample of good and bad points, one is asked to find a classifier that generalizes from the sample and separates the two sets. Here, the good points are the reachable states of the program, and the bad points are those that reach a safety property violation. Thus, a learned classifier is a candidate invariant. In this paper, we propose a new algorithm that uses decision trees to learn candidate invariants in the form of arbitrary Boolean combinations of numerical inequalities. We have used our algorithm to verify C programs taken from the literature. The algorithm is able to infer safe invariants for a range of challenging benchmarks and compares favorably to other ML-based invariant inference techniques. In particular, it scales well to large sample sets."
                },
                {
                    "title": "seL4: formal verification of an OS kernel",
                    "abstract": "Complete formal verification is the only known way to guarantee that a system is free of programming errors.\n We present our experience in performing the formal, machine-checked verification of the seL4 microkernel from an abstract specification down to its C implementation. We assume correctness of compiler, assembly code, and hardware, and we used a unique design approach that fuses formal and operating systems techniques. To our knowledge, this is the first formal proof of functional correctness of a complete, general-purpose operating-system kernel. Functional correctness means here that the implementation always strictly follows our high-level abstract specification of kernel behaviour. This encompasses traditional design and implementation safety properties such as the kernel will never crash, and it will never perform an unsafe operation. It also proves much more: we can predict precisely how the kernel will behave in every possible situation.\n seL4, a third-generation microkernel of L4 provenance, comprises 8,700 lines of C code and 600 lines of assembler. Its performance is comparable to other high-performance L4 kernels."
                },
                {
                    "title": "E - a brainiac theorem prover",
                    "abstract": "We describe the superposition-based theorem prover E. E is a sound and complete prover for clausal first order logic with equality. Important properties of the prover include strong redundancy elimination criteria, the DISCOUNT loop proof procedure, a very flexible interface for specifying search control heuristics, and an efficient inference engine. We also discuss strength and weaknesses of the system."
                },
                {
                    "title": "DT-Solver: Automated Theorem Proving with Dynamic-Tree Sampling Guided by Proof-level Value Function",
                    "abstract": "Recent advances in neural theorem-proving resort to large language models and tree searches. When proving a theorem, a language model advises single-step actions based on the current proving state and the tree search finds a sequence of correct steps using actions given by the language model. However, prior works often conduct constant computation efforts for each proving state while ignoring that the hard states often need more exploration than easy states. Moreover, they evaluate and guide the proof search solely depending on the current proof state instead of considering the whole proof trajectory as human reasoning does. Here, to accommodate general theorems, we propose a novel Dynamic-Tree Driven Theorem Solver (DT-Solver) by guiding the search procedure with state confidence and proof-level values. Specifically, DT-Solver introduces a dynamic-tree Monte-Carlo search algorithm, which dynamically allocates computing budgets for different state confidences, guided by a new proof-level value function to discover proof states that require substantial exploration.Experiments on two popular theorem-proving datasets, PISA and Mathlib, show significant performance gains by our DT-Solver over the state-of-the-art approaches, with a 6.65% improvement on average in terms of success rate. And especially under low computing resource settings (11.03% improvement on average)."
                },
                {
                    "title": "Learning Loop Invariants for Program Verification",
                    "abstract": "A fundamental problem in program verification concerns inferring loop invariants. The problem is undecidable and even practical instances are challenging. Inspired by how human experts construct loop invariants, we propose a reasoning framework Code2Inv that constructs the solution by multi-step decision making and querying an external program graph memory block. By training with reinforcement learning, Code2Inv captures rich program features and avoids the need for ground truth solutions as supervision. Compared to previous learning tasks in domains with graph-structured data, it addresses unique challenges, such as a binary objective function and an extremely sparse reward that is given by an automated theorem prover only after the complete loop invariant is proposed. We evaluate Code2Inv on a suite of 133 benchmark problems and compare it to three state-of-the-art systems. It solves 106 problems compared to 73 by a stochastic search-based system, 77 by a heuristic search-based system, and 100 by a decision tree learning-based system. Moreover, the strategy learned can be generalized to new programs: compared to solving new instances from scratch, the pre-trained agent is more sample efficient in finding solutions."
                }
            ],
            "categories": [
                "cs.AI",
                "cs.CL",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we leverage large language models (LLMs) to enhance the formal verification process of code through automated theorem proving (ATP)?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it bridges the gap between advanced reasoning capabilities of LLMs and the rigorous requirements of formal verification. By integrating LLMs with ATP, we can improve the efficiency and effectiveness of code verification, leading to more reliable software systems. This advancement could inspire future research into hybrid models that combine machine learning with formal methods, potentially leading to practical applications in critical software systems where correctness is paramount.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in this problem stem from the complexity of formal verification itself, which requires precise reasoning and adherence to formal languages. Naive approaches may fail due to the intricate nature of code dependencies and the need for step-wise proof generation. Additionally, the integration of LLMs with ATP environments poses technical obstacles, such as ensuring that the LLM can generate valid lemmas and proof states that align with the formal requirements of the verification process. The need for high expressiveness in the formal languages and the ability to handle deep dependencies further complicates the task.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on symbolic verification or hand-crafted rules, which lack the adaptability and reasoning power of LLMs. The limitations of existing solutions include their inability to generalize across diverse code cases and the reliance on human-written theorems that may not cover all scenarios. Barriers such as the lack of a suitable dataset for training LLMs in this context and the absence of a framework that effectively combines LLMs with ATP have hindered progress. Our approach differs by proposing the FVEL environment, which systematically integrates LLMs into the formal verification process, utilizing a specially curated dataset (FVELer) that addresses these gaps.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves the development of the FVEL environment, which takes C code as input, converts it into Isabelle formulation, and generates lemmas followed by proofs. The interaction with the LLM includes providing the Isabelle formulation and receiving generated lemmas and proof states, with feedback from the ATP environment. The FVELer dataset, comprising 758 theories, 29,125 lemmas,"
            }
        },
        "author_data": {
            "85b28bc2-38ec-499c-a82f-810c928d6c13": {
                "pk": "85b28bc2-38ec-499c-a82f-810c928d6c13",
                "name": "Xiaohan Lin",
                "collaborators": [
                    "Yuanyuan Mi",
                    "Xiaolong Zou",
                    "Zilong Ji",
                    "Tiejun Huang",
                    "Si Wu",
                    "Qingxing Cao",
                    "Yinya Huang",
                    "Zhicheng Yang",
                    "Zhengying Liu",
                    "Zhenguo Li"
                ],
                "domain": [
                    "Spatiotemporal Processing",
                    "Automated Theorem Proving",
                    "Multiscale Molecular Modeling",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Spatiotemporal Information Processing with a Reservoir Decision-making Network",
                        "abstract": "Spatiotemporal information processing is fundamental to brain functions. The present study investigates a canonic neural network model for spatiotemporal pattern recognition. Specifically, the model consists of two modules, a reservoir subnetwork and a decision-making subnetwork. The former projects complex spatiotemporal patterns into spatially separated neural representations, and the latter reads out these neural representations via integrating information over time; the two modules are combined together via supervised-learning using known examples. We elucidate the working mechanism of the model and demonstrate its feasibility for discriminating complex spatiotemporal patterns. Our model reproduces the phenomenon of recognizing looming patterns in the neural system, and can learn to discriminate gait with very few training examples. We hope this study gives us insight into understanding how spatiotemporal information is processed in the brain and helps us to develop brain-inspired application algorithms."
                    },
                    {
                        "title": "ATG: Benchmarking Automated Theorem Generation for Generative Language Models",
                        "abstract": "Humans can develop new theorems to explore broader and more complex mathematical results. While current generative language models (LMs) have achieved significant improvement in automatically proving theorems, their ability to generate new or reusable theorems is still under-explored. Without the new theorems, current LMs struggle to prove harder theorems that are distant from the given hypotheses with the exponentially growing search space. Therefore, this paper proposes an Automated Theorem Generation (ATG) benchmark that evaluates whether an agent can automatically generate valuable (and possibly brand new) theorems that are applicable for downstream theorem proving as reusable knowledge. Specifically, we construct the ATG benchmark by splitting the Metamath library into three sets: axioms, library, and problem based on their proving depth. We conduct extensive experiments to investigate whether current LMs can generate theorems in the library and benefit the problem theorems proving. The results demonstrate that high-quality ATG data facilitates models' performances on downstream ATP. However, there is still room for current LMs to develop better ATG and generate more advanced and human-like theorems. We hope the new ATG challenge can shed some light on advanced complex theorem proving."
                    },
                    {
                        "title": "Invertible Coarse Graining with Physics-Informed Generative Artificial Intelligence",
                        "abstract": "Multiscale molecular modeling is widely applied in scientific research of molecular properties over large time and length scales. Two specific challenges are commonly present in multiscale modeling, provided that information between the coarse and fine representations of molecules needs to be properly exchanged: One is to construct coarse grained models by passing information from the fine to coarse levels; the other is to restore finer molecular details given coarse grained configurations. Although these two problems are commonly addressed independently, in this work, we present a theory connecting them, and develop a methodology called Cycle Coarse Graining (CCG) to solve both problems in a unified manner. In CCG, reconstruction can be achieved via a tractable deep generative model, allowing retrieval of fine details from coarse-grained simulations. The reconstruction in turn delivers better coarse-grained models which are informed of the fine-grained physics, and enables calculation of the free energies in a rare-event-free manner. CCG thus provides a systematic way for multiscale molecular modeling, where the finer details of coarse-grained simulations can be efficiently retrieved, and the coarse-grained models can be improved consistently."
                    }
                ]
            },
            "42d88c9e-f01c-4388-91df-0a2568c06339": {
                "pk": "42d88c9e-f01c-4388-91df-0a2568c06339",
                "name": "Qingxing Cao",
                "collaborators": [
                    "Xiaodan Liang",
                    "Liang Lin",
                    "Yinya Huang",
                    "Zhengying Liu",
                    "Haiming Wang",
                    "Bailin Li",
                    "Guanbin Li",
                    "Keze Wang",
                    "Zhenguo Li",
                    "Jing Xiong"
                ],
                "domain": [
                    "Visual Question Answering",
                    "Explainable AI",
                    "Natural Language Processing",
                    "Automated Theorem Proving"
                ],
                "publications": [
                    {
                        "title": "Explainable High-order Visual Question Reasoning: A New Benchmark and Knowledge-routed Network",
                        "abstract": "Explanation and high-order reasoning capabilities are crucial for real-world visual question answering with diverse levels of inference complexity (e.g., what is the dog that is near the girl playing with?) and important for users to understand and diagnose the trustworthiness of the system. Current VQA benchmarks on natural images with only an accuracy metric end up pushing the models to exploit the dataset biases and cannot provide any interpretable justification, which severally hinders advances in high-level question answering. In this work, we propose a new HVQR benchmark for evaluating explainable and high-order visual question reasoning ability with three distinguishable merits: 1) the questions often contain one or two relationship triplets, which requires the model to have the ability of multistep reasoning to predict plausible answers; 2) we provide an explicit evaluation on a multistep reasoning process that is constructed with image scene graphs and commonsense knowledge bases; and 3) each relationship triplet in a large-scale knowledge base only appears once among all questions, which poses challenges for existing networks that often attempt to overfit the knowledge base that already appears in the training set and enforces the models to handle unseen questions and knowledge fact usage. We also propose a new knowledge-routed modular network (KM-net) that incorporates the multistep reasoning process over a large knowledge base into visual question reasoning. An extensive dataset analysis and comparisons with existing models on the HVQR benchmark show that our benchmark provides explainable evaluations, comprehensive reasoning requirements and realistic challenges of VQA systems, as well as our KM-net's superiority in terms of accuracy and explanation ability."
                    },
                    {
                        "title": "Recognizing Focal Liver Lesions in Contrast-Enhanced Ultrasound with Discriminatively Trained Spatio-Temporal Model",
                        "abstract": "The aim of this study is to provide an automatic computational framework to assist clinicians in diagnosing Focal Liver Lesions (FLLs) in Contrast-Enhancement Ultrasound (CEUS). We represent FLLs in a CEUS video clip as an ensemble of Region-of-Interests (ROIs), whose locations are modeled as latent variables in a discriminative model. Different types of FLLs are characterized by both spatial and temporal enhancement patterns of the ROIs. The model is learned by iteratively inferring the optimal ROI locations and optimizing the model parameters. To efficiently search the optimal spatial and temporal locations of the ROIs, we propose a data-driven inference algorithm by combining effective spatial and temporal pruning. The experiments show that our method achieves promising results on the largest dataset in the literature (to the best of our knowledge), which we have made publicly available."
                    },
                    {
                        "title": "Attention-Aware Face Hallucination via Deep Reinforcement Learning",
                        "abstract": "Face hallucination is a domain-specific super-resolution problem with the goal to generate high-resolution (HR) faces from low-resolution (LR) input images. In contrast to existing methods that often learn a single patch-to-patch mapping from LR to HR images and are regardless of the contextual interdependency between patches, we propose a novel Attention-aware Face Hallucination (Attention-FH) framework which resorts to deep reinforcement learning for sequentially discovering attended patches and then performing the facial part enhancement by fully exploiting the global interdependency of the image. Specifically, in each time step, the recurrent policy network is proposed to dynamically specify a new attended region by incorporating what happened in the past. The state (i.e., face hallucination result for the whole image) can thus be exploited and updated by the local enhancement network on the selected region. The Attention-FH approach jointly learns the recurrent policy network and local enhancement network through maximizing the long-term reward that reflects the hallucination performance over the whole image. Therefore, our proposed Attention-FH is capable of adaptively personalizing an optimal searching path for each face image according to its own characteristic. Extensive experiments show our approach significantly surpasses the state-of-the-arts on in-the-wild faces with large pose and illumination variations."
                    },
                    {
                        "title": "Interpretable Visual Question Answering by Reasoning on Dependency Trees",
                        "abstract": "Collaborative reasoning for understanding image-question pairs is a very critical but underexplored topic in interpretable visual question answering systems. Although very recent studies have attempted to use explicit compositional processes to assemble multiple subtasks embedded in questions, their models heavily rely on annotations or handcrafted rules to obtain valid reasoning processes, which leads to either heavy workloads or poor performance on compositional reasoning. In this paper, to better align image and language domains in diverse and unrestricted cases, we propose a novel neural network model that performs global reasoning on a dependency tree parsed from the question; thus, our model is called a parse-tree-guided reasoning network (PTGRN). This network consists of three collaborative modules: i) an attention module that exploits the local visual evidence of each word parsed from the question, ii) a gated residual composition module that composes the previously mined evidence, and iii) a parse-tree-guided propagation module that passes the mined evidence along the parse tree. Thus, PTGRN is capable of building an interpretable visual question answering (VQA) system that gradually derives image cues following question-driven parse-tree reasoning. Experiments on relational datasets demonstrate the superiority of PTGRN over current state-of-the-art VQA methods, and the visualization results highlight the explainable capability of our reasoning system."
                    },
                    {
                        "title": "Visual Question Reasoning on General Dependency Tree",
                        "abstract": "The collaborative reasoning for understanding each image-question pair is very critical but under-explored for an interpretable Visual Question Answering (VQA) system. Although very recent works also tried the explicit compositional processes to assemble multiple sub-tasks embedded in the questions, their models heavily rely on the annotations or hand-crafted rules to obtain valid reasoning layout, leading to either heavy labor or poor performance on composition reasoning. In this paper, to enable global context reasoning for better aligning image and language domains in diverse and unrestricted cases, we propose a novel reasoning network called Adversarial Composition Modular Network (ACMN). This network comprises of two collaborative modules: i) an adversarial attention module to exploit the local visual evidence for each word parsed from the question; ii) a residual composition module to compose the previously mined evidence. Given a dependency parse tree for each question, the adversarial attention module progressively discovers salient regions of one word by densely combining regions of child word nodes in an adversarial manner. Then residual composition module merges the hidden representations of an arbitrary number of children through sum pooling and residual connection. Our ACMN is thus capable of building an interpretable VQA system that gradually dives the image cues following a question-driven reasoning route and makes global reasoning by incorporating the learned knowledge of all attention modules in a principled manner. Experiments on relational datasets demonstrate the superiority of our ACMN and visualization results show the explainable capability of our reasoning system."
                    },
                    {
                        "title": "Face Hallucination by Attentive Sequence Optimization with Reinforcement Learning",
                        "abstract": "Face hallucination is a domain-specific super-resolution problem that aims to generate a high-resolution (HR) face image from a low-resolution~(LR) input. In contrast to the existing patch-wise super-resolution models that divide a face image into regular patches and independently apply LR to HR mapping to each patch, we implement deep reinforcement learning and develop a novel attention-aware face hallucination (Attention-FH) framework, which recurrently learns to attend a sequence of patches and performs facial part enhancement by fully exploiting the global interdependency of the image. Specifically, our proposed framework incorporates two components: a recurrent policy network for dynamically specifying a new attended region at each time step based on the status of the super-resolved image and the past attended region sequence, and a local enhancement network for selected patch hallucination and global state updating. The Attention-FH model jointly learns the recurrent policy network and local enhancement network through maximizing a long-term reward that reflects the hallucination result with respect to the whole HR image. Extensive experiments demonstrate that our Attention-FH significantly outperforms the state-of-the-art methods on in-the-wild face images with large pose and illumination variations."
                    },
                    {
                        "title": "Linguistically Driven Graph Capsule Network for Visual Question Reasoning",
                        "abstract": "Recently, studies of visual question answering have explored various architectures of end-to-end networks and achieved promising results on both natural and synthetic datasets, which require explicitly compositional reasoning. However, it has been argued that these black-box approaches lack interpretability of results, and thus cannot perform well on generalization tasks due to overfitting the dataset bias. In this work, we aim to combine the benefits of both sides and overcome their limitations to achieve an end-to-end interpretable structural reasoning for general images without the requirement of layout annotations. Inspired by the property of a capsule network that can carve a tree structure inside a regular convolutional neural network (CNN), we propose a hierarchical compositional reasoning model called the \"Linguistically driven Graph Capsule Network\", where the compositional process is guided by the linguistic parse tree. Specifically, we bind each capsule in the lowest layer to bridge the linguistic embedding of a single word in the original question with visual evidence and then route them to the same capsule if they are siblings in the parse tree. This compositional process is achieved by performing inference on a linguistically driven conditional random field (CRF) and is performed across multiple graph capsule layers, which results in a compositional reasoning process inside a CNN. Experiments on the CLEVR dataset, CLEVR compositional generation test, and FigureQA dataset demonstrate the effectiveness and composition generalization ability of our end-to-end model."
                    },
                    {
                        "title": "REM-Net: Recursive Erasure Memory Network for Commonsense Evidence Refinement",
                        "abstract": "When answering a question, people often draw upon their rich world knowledge in addition to the particular context. While recent works retrieve supporting facts/evidence from commonsense knowledge bases to supply additional information to each question, there is still ample opportunity to advance it on the quality of the evidence. It is crucial since the quality of the evidence is the key to answering commonsense questions, and even determines the upper bound on the QA systems performance. In this paper, we propose a recursive erasure memory network (REM-Net) to cope with the quality improvement of evidence. To address this, REM-Net is equipped with a module to refine the evidence by recursively erasing the low-quality evidence that does not explain the question answering. Besides, instead of retrieving evidence from existing knowledge bases, REM-Net leverages a pre-trained generative model to generate candidate evidence customized for the question. We conduct experiments on two commonsense question answering datasets, WIQA and CosmosQA. The results demonstrate the performance of REM-Net and show that the refined evidence is explainable."
                    },
                    {
                        "title": "ATG: Benchmarking Automated Theorem Generation for Generative Language Models",
                        "abstract": "Humans can develop new theorems to explore broader and more complex mathematical results. While current generative language models (LMs) have achieved significant improvement in automatically proving theorems, their ability to generate new or reusable theorems is still under-explored. Without the new theorems, current LMs struggle to prove harder theorems that are distant from the given hypotheses with the exponentially growing search space. Therefore, this paper proposes an Automated Theorem Generation (ATG) benchmark that evaluates whether an agent can automatically generate valuable (and possibly brand new) theorems that are applicable for downstream theorem proving as reusable knowledge. Specifically, we construct the ATG benchmark by splitting the Metamath library into three sets: axioms, library, and problem based on their proving depth. We conduct extensive experiments to investigate whether current LMs can generate theorems in the library and benefit the problem theorems proving. The results demonstrate that high-quality ATG data facilitates models' performances on downstream ATP. However, there is still room for current LMs to develop better ATG and generate more advanced and human-like theorems. We hope the new ATG challenge can shed some light on advanced complex theorem proving."
                    },
                    {
                        "title": "Knowledge-Routed Visual Question Reasoning: Challenges for Deep Representation Embedding",
                        "abstract": "Though beneficial for encouraging the Visual Question Answering (VQA) models to discover the underlying knowledge by exploiting the input-output correlation beyond image and text contexts, the existing knowledge VQA datasets are mostly annotated in a crowdsource way, e.g., collecting questions and external reasons from different users via the internet. In addition to the challenge of knowledge reasoning, how to deal with the annotator bias also remains unsolved, which often leads to superficial over-fitted correlations between questions and answers. To address this issue, we propose a novel dataset named Knowledge-Routed Visual Question Reasoning for VQA model evaluation. Considering that a desirable VQA model should correctly perceive the image context, understand the question, and incorporate its learned knowledge, our proposed dataset aims to cutoff the shortcut learning exploited by the current deep embedding models and push the research boundary of the knowledge-based visual question reasoning. Specifically, we generate the question-answer pair based on both the Visual Genome scene graph and an external knowledge base with controlled programs to disentangle the knowledge from other biases. The programs can select one or two triplets from the scene graph or knowledge base to push multi-step reasoning, avoid answer ambiguity, and balanced the answer distribution. In contrast to the existing VQA datasets, we further imply the following two major constraints on the programs to incorporate knowledge reasoning: i) multiple knowledge triplets can be related to the question, but only one knowledge relates to the image object. This can enforce the VQA model to correctly perceive the image instead of guessing the knowledge based on the given question solely; ii) all questions are based on different knowledge, but the candidate answers are the same for both the training and test sets."
                    },
                    {
                        "title": "MUSTARD: Mastering Uniform Synthesis of Theorem and Proof Data",
                        "abstract": "Recent large language models (LLMs) have witnessed significant advancement in various tasks, including mathematical reasoning and theorem proving. As these two tasks require strict and formal multi-step inference, they are appealing domains for exploring the reasoning ability of LLMs but still face important challenges. Previous studies such as Chain-of-Thought (CoT) have revealed the effectiveness of intermediate steps guidance. However, such step-wise annotation requires heavy labor, leading to insufficient training steps for current benchmarks. To fill this gap, this work introduces MUSTARD, a data generation framework that masters uniform synthesis of theorem and proof data of high quality and diversity. MUSTARD synthesizes data in three stages: (1) It samples a few mathematical concept seeds as the problem category. (2) Then, it prompts a generative language model with the sampled concepts to obtain both the problems and their step-wise formal solutions. (3) Lastly, the framework utilizes a proof assistant (e.g., Lean Prover) to filter the valid proofs. With the proposed MUSTARD, we present a theorem-and-proof benchmark MUSTARDSAUCE with 5,866 valid data points. Each data point contains an informal statement, an informal proof, and a translated formal proof that passes the prover validation. We perform extensive analysis and demonstrate that MUSTARD generates validated high-quality step-by-step data. We further apply the MUSTARDSAUCE for fine-tuning smaller language models. The fine-tuned Llama 2-7B achieves a 15.41% average relative performance gain in automated theorem proving, and 8.18% in math word problems. Codes and data are available at https://github.com/Eleanor-H/MUSTARD."
                    },
                    {
                        "title": "TRIGO: Benchmarking Formal Mathematical Proof Reduction for Generative Language Models",
                        "abstract": "Automated theorem proving (ATP) has become an appealing domain for exploring the reasoning ability of the recent successful generative language models. However, current ATP benchmarks mainly focus on symbolic inference, but rarely involve the understanding of complex number combination reasoning. In this work, we propose TRIGO, an ATP benchmark that not only requires a model to reduce a trigonometric expression with step-by-step proofs but also evaluates a generative LM's reasoning ability on formulas and its capability to manipulate, group, and factor number terms. We gather trigonometric expressions and their reduced forms from the web, annotate the simplification process manually, and translate it into the Lean formal language system. We then automatically generate additional examples from the annotated samples to expand the dataset. Furthermore, we develop an automatic generator based on Lean-Gym to create dataset splits of varying difficulties and distributions in order to thoroughly analyze the model's generalization ability. Our extensive experiments show our proposed TRIGO poses a new challenge for advanced generative LM's including GPT-4 which is pre-trained on a considerable amount of open-source formal theorem-proving language data, and provide a new tool to study the generative LM's ability on both formal and mathematical reasoning."
                    },
                    {
                        "title": "LEGO-Prover: Neural Theorem Proving with Growing Libraries",
                        "abstract": "Despite the success of large language models (LLMs), the task of theorem proving still remains one of the hardest reasoning tasks that is far from being fully solved. Prior methods using language models have demonstrated promising results, but they still struggle to prove even middle school level theorems. One common limitation of these methods is that they assume a fixed theorem library during the whole theorem proving process. However, as we all know, creating new useful theorems or even new theories is not only helpful but crucial and necessary for advancing mathematics and proving harder and deeper results. In this work, we present LEGO-Prover, which employs a growing skill library containing verified lemmas as skills to augment the capability of LLMs used in theorem proving. By constructing the proof modularly, LEGO-Prover enables LLMs to utilize existing skills retrieved from the library and to create new skills during the proving process. These skills are further evolved (by prompting an LLM) to enrich the library on another scale. Modular and reusable skills are constantly added to the library to enable tackling increasingly intricate mathematical problems. Moreover, the learned library further bridges the gap between human proofs and formal proofs by making it easier to impute missing steps. LEGO-Prover advances the state-of-the-art pass rate on miniF2F-valid (48.0% to 57.0%) and miniF2F-test (45.5% to 47.1%). During the proving process, LEGO-Prover also manages to generate over 20,000 skills (theorems/lemmas) and adds them to the growing library. Our ablation study indicates that these newly added skills are indeed helpful for proving theorems, resulting in an improvement from a success rate of 47.1% to 50.4%. We also release our code and all the generated skills."
                    },
                    {
                        "title": "DQ-LoRe: Dual Queries with Low Rank Approximation Re-ranking for In-Context Learning",
                        "abstract": "Recent advances in natural language processing, primarily propelled by Large Language Models (LLMs), have showcased their remarkable capabilities grounded in in-context learning. A promising avenue for guiding LLMs in intricate reasoning tasks involves the utilization of intermediate reasoning steps within the Chain-of-Thought (CoT) paradigm. Nevertheless, the central challenge lies in the effective selection of exemplars for facilitating in-context learning. In this study, we introduce a framework that leverages Dual Queries and Low-rank approximation Re-ranking (DQ-LoRe) to automatically select exemplars for in-context learning. Dual Queries first query LLM to obtain LLM-generated knowledge such as CoT, then query the retriever to obtain the final exemplars via both question and the knowledge. Moreover, for the second query, LoRe employs dimensionality reduction techniques to refine exemplar selection, ensuring close alignment with the input question's knowledge. Through extensive experiments, we demonstrate that DQ-LoRe significantly outperforms prior state-of-the-art methods in the automatic selection of exemplars for GPT-4, enhancing performance from 92.5% to 94.2%. Our comprehensive analysis further reveals that DQ-LoRe consistently outperforms retrieval-based approaches in terms of both performance and adaptability, especially in scenarios characterized by distribution shifts. DQ-LoRe pushes the boundary of in-context learning and opens up new avenues for addressing complex reasoning challenges. Our code is released at https://github.com/AI4fun/DQ-LoRe}{https://github.com/AI4fun/DQ-LoRe."
                    }
                ]
            },
            "bccd13ac-b175-4176-9273-dd26815cd907": {
                "pk": "bccd13ac-b175-4176-9273-dd26815cd907",
                "name": "Yinya Huang",
                "collaborators": [
                    "Xiaodan Liang",
                    "Zhengying Liu",
                    "Zhicheng Yang",
                    "Qingxing Cao",
                    "Jing Xiong",
                    "Zhijiang Guo",
                    "Linqi Song",
                    "Haiming Wang",
                    "Zhenguo Li",
                    "Jianqiao Lu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Logical Reasoning",
                    "Automated Theorem Proving",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "DAGN: Discourse-Aware Graph Network for Logical Reasoning",
                        "abstract": "Recent QA with logical reasoning questions requires passage-level relations among the sentences. However, current approaches still focus on sentence-level relations interacting among tokens. In this work, we explore aggregating passage-level clues for solving logical reasoning QA by using discourse-based information. We propose a discourse-aware graph network (DAGN) that reasons relying on the discourse structure of the texts. The model encodes discourse information as a graph with elementary discourse units (EDUs) and discourse relations, and learns the discourse-aware features via a graph network for downstream QA tasks. Experiments are conducted on two logical reasoning QA datasets, ReClor and LogiQA, and our proposed DAGN achieves competitive results. The source code is available at https://github.com/Eleanor-H/DAGN."
                    },
                    {
                        "title": "REM-Net: Recursive Erasure Memory Network for Commonsense Evidence Refinement",
                        "abstract": "When answering a question, people often draw upon their rich world knowledge in addition to the particular context. While recent works retrieve supporting facts/evidence from commonsense knowledge bases to supply additional information to each question, there is still ample opportunity to advance it on the quality of the evidence. It is crucial since the quality of the evidence is the key to answering commonsense questions, and even determines the upper bound on the QA systems performance. In this paper, we propose a recursive erasure memory network (REM-Net) to cope with the quality improvement of evidence. To address this, REM-Net is equipped with a module to refine the evidence by recursively erasing the low-quality evidence that does not explain the question answering. Besides, instead of retrieving evidence from existing knowledge bases, REM-Net leverages a pre-trained generative model to generate candidate evidence customized for the question. We conduct experiments on two commonsense question answering datasets, WIQA and CosmosQA. The results demonstrate the performance of REM-Net and show that the refined evidence is explainable."
                    },
                    {
                        "title": "Discourse-Aware Graph Networks for Textual Logical Reasoning",
                        "abstract": "Textual logical reasoning, especially question-answering (QA) tasks with logical reasoning, requires awareness of particular logical structures. The passage-level logical relations represent entailment or contradiction between propositional units (e.g., a concluding sentence). However, such structures are unexplored as current QA systems focus on entity-based relations. In this work, we propose logic structural-constraint modeling to solve the logical reasoning QA and introduce discourse-aware graph networks (DAGNs). The networks first construct logic graphs leveraging in-line discourse connectives and generic logic theories, then learn logic representations by end-to-end evolving the logic relations with an edge-reasoning mechanism and updating the graph features. This pipeline is applied to a general encoder, whose fundamental features are joined with the high-level logic features for answer prediction. Experiments on three textual logical reasoning datasets demonstrate the reasonability of the logical structures built in DAGNs and the effectiveness of the learned logic features. Moreover, zero-shot transfer results show the features' generality to unseen logical texts."
                    },
                    {
                        "title": "MetaLogic: Logical Reasoning Explanations with Fine-Grained Structure",
                        "abstract": "In this paper, we propose a comprehensive benchmark to investigate models' logical reasoning capabilities in complex real-life scenarios. Current explanation datasets often employ synthetic data with simple reasoning structures. Therefore, it cannot express more complex reasoning processes, such as the rebuttal to a reasoning step and the degree of certainty of the evidence. To this end, we propose a comprehensive logical reasoning explanation form. Based on the multi-hop chain of reasoning, the explanation form includes three main components: (1) The condition of rebuttal that the reasoning node can be challenged; (2) Logical formulae that uncover the internal texture of reasoning nodes; (3) Reasoning strength indicated by degrees of certainty. The fine-grained structure conforms to the real logical reasoning scenario, better fitting the human cognitive process but, simultaneously, is more challenging for the current models. We evaluate the current best models' performance on this new explanation form. The experimental results show that generating reasoning graphs remains a challenging task for current models, even with the help of giant pre-trained language models."
                    },
                    {
                        "title": "ATG: Benchmarking Automated Theorem Generation for Generative Language Models",
                        "abstract": "Humans can develop new theorems to explore broader and more complex mathematical results. While current generative language models (LMs) have achieved significant improvement in automatically proving theorems, their ability to generate new or reusable theorems is still under-explored. Without the new theorems, current LMs struggle to prove harder theorems that are distant from the given hypotheses with the exponentially growing search space. Therefore, this paper proposes an Automated Theorem Generation (ATG) benchmark that evaluates whether an agent can automatically generate valuable (and possibly brand new) theorems that are applicable for downstream theorem proving as reusable knowledge. Specifically, we construct the ATG benchmark by splitting the Metamath library into three sets: axioms, library, and problem based on their proving depth. We conduct extensive experiments to investigate whether current LMs can generate theorems in the library and benefit the problem theorems proving. The results demonstrate that high-quality ATG data facilitates models' performances on downstream ATP. However, there is still room for current LMs to develop better ATG and generate more advanced and human-like theorems. We hope the new ATG challenge can shed some light on advanced complex theorem proving."
                    },
                    {
                        "title": "FormalAlign: Automated Alignment Evaluation for Autoformalization",
                        "abstract": "Autoformalization aims to convert informal mathematical proofs into machine-verifiable formats, bridging the gap between natural and formal languages. However, ensuring semantic alignment between the informal and formalized statements remains challenging. Existing approaches heavily rely on manual verification, hindering scalability. To address this, we introduce \\textsc{FormalAlign}, the first automated framework designed for evaluating the alignment between natural and formal languages in autoformalization. \\textsc{FormalAlign} trains on both the autoformalization sequence generation task and the representational alignment between input and output, employing a dual loss that combines a pair of mutually enhancing autoformalization and alignment tasks. Evaluated across four benchmarks augmented by our proposed misalignment strategies, \\textsc{FormalAlign} demonstrates superior performance. In our experiments, \\textsc{FormalAlign} outperforms GPT-4, achieving an Alignment-Selection Score 11.58\\% higher on \\forml-Basic (99.21\\% vs. 88.91\\%) and 3.19\\% higher on MiniF2F-Valid (66.39\\% vs. 64.34\\%). This effective alignment evaluation significantly reduces the need for manual verification. Both the dataset and code can be accessed via~\\url{https://github.com/rookie-joe/FormalAlign}."
                    },
                    {
                        "title": "AlignedCoT: Prompting Large Language Models via Native-Speaking Demonstrations",
                        "abstract": "Large Language Models prompting, such as using in-context demonstrations, is a mainstream technique for invoking LLMs to perform high-performance and solid complex reasoning (e.g., mathematical reasoning, commonsense reasoning), and has the potential for further human-machine collaborative scientific findings. However, current LLMs are delicate and elusive in prompt words and styles. And there is an unseen gap between LLM understanding and human-written prompts. This paper introduces Alignedcot, an LLM-acquainted prompting technique that includes proficient ``native-speaking'' in in-context learning for the LLMs. Specifically, it achieves consistent and correct step-wise prompts in zero-shot scenarios by progressively probing, refining, and formatting the LLM chain of thoughts so that free from handcrafted few-shot demonstrations while maintaining the prompt quality. We conduct experiments on mathematical reasoning and commonsense reasoning. We find that LLMs with Alignedcot perform significantly superior to them with human-crafted demonstrations. We further apply Alignedcot for rewriting the GSM8K training set, resulting in a GSM8K-Align dataset. We observe its benefits for retrieval augmented generation. The code and data can be found at https://github.com/yangzhch6/AlignedCoT."
                    },
                    {
                        "title": "CLOMO: Counterfactual Logical Modification with Large Language Models",
                        "abstract": "In this study, we delve into the realm of counterfactual reasoning capabilities of large language models (LLMs). Our primary objective is to cultivate the counterfactual thought processes within LLMs and rigorously assess these processes for their validity. Specifically, we introduce a novel task, Counterfactual Logical Modification (CLOMO), and a high-quality human-annotated benchmark. In this task, LLMs must adeptly alter a given argumentative text to uphold a predetermined logical relationship. To effectively evaluate a generation model's counterfactual capabilities, we propose an innovative evaluation metric, the decomposed Self-Evaluation Score (SES) to directly evaluate the natural language output of LLMs instead of modeling the task as a multiple-choice problem. Analysis shows that the proposed automatic metric aligns well with human preference. Our experimental results show that while LLMs demonstrate a notable capacity for logical counterfactual thinking, there remains a discernible gap between their current abilities and human performance. Code and data are available at https://github.com/Eleanor-H/CLOMO."
                    },
                    {
                        "title": "MUSTARD: Mastering Uniform Synthesis of Theorem and Proof Data",
                        "abstract": "Recent large language models (LLMs) have witnessed significant advancement in various tasks, including mathematical reasoning and theorem proving. As these two tasks require strict and formal multi-step inference, they are appealing domains for exploring the reasoning ability of LLMs but still face important challenges. Previous studies such as Chain-of-Thought (CoT) have revealed the effectiveness of intermediate steps guidance. However, such step-wise annotation requires heavy labor, leading to insufficient training steps for current benchmarks. To fill this gap, this work introduces MUSTARD, a data generation framework that masters uniform synthesis of theorem and proof data of high quality and diversity. MUSTARD synthesizes data in three stages: (1) It samples a few mathematical concept seeds as the problem category. (2) Then, it prompts a generative language model with the sampled concepts to obtain both the problems and their step-wise formal solutions. (3) Lastly, the framework utilizes a proof assistant (e.g., Lean Prover) to filter the valid proofs. With the proposed MUSTARD, we present a theorem-and-proof benchmark MUSTARDSAUCE with 5,866 valid data points. Each data point contains an informal statement, an informal proof, and a translated formal proof that passes the prover validation. We perform extensive analysis and demonstrate that MUSTARD generates validated high-quality step-by-step data. We further apply the MUSTARDSAUCE for fine-tuning smaller language models. The fine-tuned Llama 2-7B achieves a 15.41% average relative performance gain in automated theorem proving, and 8.18% in math word problems. Codes and data are available at https://github.com/Eleanor-H/MUSTARD."
                    },
                    {
                        "title": "Integrating Large Language Models into Recommendation via Mutual Augmentation and Adaptive Aggregation",
                        "abstract": "Conventional recommendation methods have achieved notable advancements by harnessing collaborative or sequential information from user behavior. Recently, large language models (LLMs) have gained prominence for their capabilities in understanding and reasoning over textual semantics, and have found utility in various domains, including recommendation. Conventional recommendation methods and LLMs each have their strengths and weaknesses. While conventional methods excel at mining collaborative information and modeling sequential behavior, they struggle with data sparsity and the long-tail problem. LLMs, on the other hand, are proficient at utilizing rich textual contexts but face challenges in mining collaborative or sequential information. Despite their individual successes, there is a significant gap in leveraging their combined potential to enhance recommendation performance.   In this paper, we introduce a general and model-agnostic framework known as \\textbf{L}arge \\textbf{la}nguage model with \\textbf{m}utual augmentation and \\textbf{a}daptive aggregation for \\textbf{Rec}ommendation (\\textbf{Llama4Rec}). Llama4Rec synergistically combines conventional and LLM-based recommendation models. Llama4Rec proposes data augmentation and prompt augmentation strategies tailored to enhance the conventional model and LLM respectively. An adaptive aggregation module is adopted to combine the predictions of both kinds of models to refine the final recommendation results. Empirical studies on three real-world datasets validate the superiority of Llama4Rec, demonstrating its consistent outperformance of baseline methods and significant improvements in recommendation performance."
                    },
                    {
                        "title": "AutoCV: Empowering Reasoning with Automated Process Labeling via Confidence Variation",
                        "abstract": "In this work, we propose a novel method named \\textbf{Auto}mated Process Labeling via \\textbf{C}onfidence \\textbf{V}ariation (\\textbf{\\textsc{AutoCV}}) to enhance the reasoning capabilities of large language models (LLMs) by automatically annotating the reasoning steps. Our approach begins by training a verification model on the correctness of final answers, enabling it to generate automatic process annotations. This verification model assigns a confidence score to each reasoning step, indicating the probability of arriving at the correct final answer from that point onward. We detect relative changes in the verification's confidence scores across reasoning steps to automatically annotate the reasoning process. This alleviates the need for numerous manual annotations or the high computational costs associated with model-induced annotation approaches. We experimentally validate that the confidence variations learned by the verification model trained on the final answer correctness can effectively identify errors in the reasoning steps. Subsequently, we demonstrate that the process annotations generated by \\textsc{AutoCV} can improve the accuracy of the verification model in selecting the correct answer from multiple outputs generated by LLMs. Notably, we achieve substantial improvements across five datasets in mathematics and commonsense reasoning. The source code of \\textsc{AutoCV} is available at \\url{https://github.com/rookie-joe/AUTOCV}."
                    },
                    {
                        "title": "OptiBench Meets ReSocratic: Measure and Improve LLMs for Optimization Modeling",
                        "abstract": "Large language models (LLMs) have exhibited their problem-solving abilities in mathematical reasoning. Solving realistic optimization (OPT) problems in application scenarios requires advanced and applied mathematics ability. However, current OPT benchmarks that merely solve linear programming are far from complex realistic situations. In this work, we propose OptiBench, a benchmark for End-to-end optimization problem-solving with human-readable inputs and outputs. OptiBench contains rich optimization problems, including linear and nonlinear programming with or without tabular data, which can comprehensively evaluate LLMs' solving ability. In our benchmark, LLMs are required to call a code solver to provide precise numerical answers. Furthermore, to alleviate the data scarcity for optimization problems, and to bridge the gap between open-source LLMs on a small scale (e.g., Llama-3-8b) and closed-source LLMs (e.g., GPT-4), we further propose a data synthesis method namely ReSocratic. Unlike general data synthesis methods that proceed from questions to answers, \\ReSocratic first incrementally synthesizes formatted optimization demonstration with mathematical formulations step by step and then back-translates the generated demonstrations into questions. Based on this, we synthesize the ReSocratic-29k dataset. We further conduct supervised fine-tuning with ReSocratic-29k on multiple open-source models. Experimental results show that ReSocratic-29k significantly improves the performance of open-source models."
                    },
                    {
                        "title": "TRIGO: Benchmarking Formal Mathematical Proof Reduction for Generative Language Models",
                        "abstract": "Automated theorem proving (ATP) has become an appealing domain for exploring the reasoning ability of the recent successful generative language models. However, current ATP benchmarks mainly focus on symbolic inference, but rarely involve the understanding of complex number combination reasoning. In this work, we propose TRIGO, an ATP benchmark that not only requires a model to reduce a trigonometric expression with step-by-step proofs but also evaluates a generative LM's reasoning ability on formulas and its capability to manipulate, group, and factor number terms. We gather trigonometric expressions and their reduced forms from the web, annotate the simplification process manually, and translate it into the Lean formal language system. We then automatically generate additional examples from the annotated samples to expand the dataset. Furthermore, we develop an automatic generator based on Lean-Gym to create dataset splits of varying difficulties and distributions in order to thoroughly analyze the model's generalization ability. Our extensive experiments show our proposed TRIGO poses a new challenge for advanced generative LM's including GPT-4 which is pre-trained on a considerable amount of open-source formal theorem-proving language data, and provide a new tool to study the generative LM's ability on both formal and mathematical reasoning."
                    },
                    {
                        "title": "RecRanker: Instruction Tuning Large Language Model as Ranker for Top-k Recommendation",
                        "abstract": "Large Language Models (LLMs) have demonstrated remarkable capabilities and have been extensively deployed across various domains, including recommender systems. Prior research has employed specialized \\textit{prompts} to leverage the in-context learning capabilities of LLMs for recommendation purposes. More recent studies have utilized instruction tuning techniques to align LLMs with human preferences, promising more effective recommendations. However, existing methods suffer from several limitations. The full potential of LLMs is not fully elicited due to low-quality tuning data and the overlooked integration of conventional recommender signals. Furthermore, LLMs may generate inconsistent responses for different ranking tasks in the recommendation, potentially leading to unreliable results.   In this paper, we introduce \\textbf{RecRanker}, tailored for instruction tuning LLMs to serve as the \\textbf{Ranker} for top-\\textit{k} \\textbf{Rec}ommendations. Specifically, we introduce an adaptive sampling module for sampling high-quality, representative, and diverse training data. To enhance the prompt, we introduce a position shifting strategy to mitigate position bias and augment the prompt with auxiliary information from conventional recommendation models, thereby enriching the contextual understanding of the LLM. Subsequently, we utilize the sampled data to assemble an instruction-tuning dataset with the augmented prompts comprising three distinct ranking tasks: pointwise, pairwise, and listwise rankings. We further propose a hybrid ranking method to enhance the model performance by ensembling these ranking tasks. Our empirical evaluations demonstrate the effectiveness of our proposed RecRanker in both direct and sequential recommendation scenarios."
                    },
                    {
                        "title": "Proving Theorems Recursively",
                        "abstract": "Recent advances in automated theorem proving leverages language models to explore expanded search spaces by step-by-step proof generation. However, such approaches are usually based on short-sighted heuristics (e.g., log probability or value function scores) that potentially lead to suboptimal or even distracting subgoals, preventing us from finding longer proofs. To address this challenge, we propose POETRY (PrOvE Theorems RecursivelY), which proves theorems in a recursive, level-by-level manner in the Isabelle theorem prover. Unlike previous step-by-step methods, POETRY searches for a verifiable sketch of the proof at each level and focuses on solving the current level's theorem or conjecture. Detailed proofs of intermediate conjectures within the sketch are temporarily replaced by a placeholder tactic called sorry, deferring their proofs to subsequent levels. This approach allows the theorem to be tackled incrementally by outlining the overall theorem at the first level and then solving the intermediate conjectures at deeper levels. Experiments are conducted on the miniF2F and PISA datasets and significant performance gains are observed in our POETRY approach over state-of-the-art methods. POETRY on miniF2F achieves an average proving success rate improvement of 5.1%. Moreover, we observe a substantial increase in the maximum proof length found by POETRY, from 10 to 26."
                    },
                    {
                        "title": "Process-Driven Autoformalization in Lean 4",
                        "abstract": "Autoformalization, the conversion of natural language mathematics into formal languages, offers significant potential for advancing mathematical reasoning. However, existing efforts are limited to formal languages with substantial online corpora and struggle to keep pace with rapidly evolving languages like Lean 4. To bridge this gap, we propose a new benchmark \\textbf{Form}alization for \\textbf{L}ean~\\textbf{4} (\\textbf{\\name}) designed to evaluate the autoformalization capabilities of large language models (LLMs). This benchmark encompasses a comprehensive assessment of questions, answers, formal statements, and proofs. Additionally, we introduce a \\textbf{P}rocess-\\textbf{S}upervised \\textbf{V}erifier (\\textbf{PSV}) model that leverages the precise feedback from Lean 4 compilers to enhance autoformalization. Our experiments demonstrate that the PSV method improves autoformalization, enabling higher accuracy using less filtered training data. Furthermore, when fine-tuned with data containing detailed process information, PSV can leverage the data more effectively, leading to more significant improvements in autoformalization for Lean 4. Our dataset and code are available at \\url{https://github.com/rookie-joe/PDA}."
                    },
                    {
                        "title": "LEGO-Prover: Neural Theorem Proving with Growing Libraries",
                        "abstract": "Despite the success of large language models (LLMs), the task of theorem proving still remains one of the hardest reasoning tasks that is far from being fully solved. Prior methods using language models have demonstrated promising results, but they still struggle to prove even middle school level theorems. One common limitation of these methods is that they assume a fixed theorem library during the whole theorem proving process. However, as we all know, creating new useful theorems or even new theories is not only helpful but crucial and necessary for advancing mathematics and proving harder and deeper results. In this work, we present LEGO-Prover, which employs a growing skill library containing verified lemmas as skills to augment the capability of LLMs used in theorem proving. By constructing the proof modularly, LEGO-Prover enables LLMs to utilize existing skills retrieved from the library and to create new skills during the proving process. These skills are further evolved (by prompting an LLM) to enrich the library on another scale. Modular and reusable skills are constantly added to the library to enable tackling increasingly intricate mathematical problems. Moreover, the learned library further bridges the gap between human proofs and formal proofs by making it easier to impute missing steps. LEGO-Prover advances the state-of-the-art pass rate on miniF2F-valid (48.0% to 57.0%) and miniF2F-test (45.5% to 47.1%). During the proving process, LEGO-Prover also manages to generate over 20,000 skills (theorems/lemmas) and adds them to the growing library. Our ablation study indicates that these newly added skills are indeed helpful for proving theorems, resulting in an improvement from a success rate of 47.1% to 50.4%. We also release our code and all the generated skills."
                    }
                ]
            },
            "9ea31c7f-e779-41f6-b2e8-3122b458ece5": {
                "pk": "9ea31c7f-e779-41f6-b2e8-3122b458ece5",
                "name": "Haiming Wang",
                "collaborators": [
                    "Cheng-Xiang Wang",
                    "Jie Huang",
                    "Jun Wang",
                    "Xiqi Gao",
                    "Xiaohu You",
                    "Yang Hao",
                    "Zhikun Zhang",
                    "Tianhao Wang",
                    "Shibo He",
                    "Michael Backes"
                ],
                "domain": [
                    "Terahertz Communication",
                    "6G Wireless Systems",
                    "MIMO",
                    "Differential Privacy"
                ],
                "publications": [
                    {
                        "title": "A Novel 3D Space-Time-Frequency Non-Stationary Channel Model for 6G THz Indoor Communication Systems",
                        "abstract": "Terahertz (THz) communication is now being considered as one of possible technologies for the sixth generation (6G) communication systems. In this paper, a novel three-dimensional (3D) space-time-frequency non-stationary massive multiple-input multiple-output (MIMO) channel model for 6G THz indoor communication systems is proposed. In this geometry-based stochastic model (GBSM), the initialization and evolution of parameters in time, space, and frequency domains are developed to generate the complete channel transfer function (CTF). Based on the proposed model, the correlation functions including time auto-correlation function (ACF), spatial crosscorrelation function (CCF), and frequency correlation function (FCF) are investigated. The results show that the statistical properties of the simulation model match well with those of the theoretical model. The stationary intervals at different frequencies are simulated. The non-stationarity in time, space, and frequency domains is verified by theoretical derivations and simulations."
                    },
                    {
                        "title": "A General 3D Space-Time-Frequency Non-Stationary THz Channel Model for 6G Ultra-Massive MIMO Wireless Communication Systems",
                        "abstract": "In this paper, a novel three-dimensional (3D) space-time-frequency (STF) non-stationary geometry-based stochastic model (GBSM) is proposed for the sixth generation (6G) terahertz (THz) wireless communication systems. The proposed THz channel model is very general having the capability to capture different channel characteristics in multiple THz application scenarios such as indoor scenarios, device-to-device (D2D) communications, ultra-massive multiple-input multiple-output (MIMO) communications, and long traveling paths of users. Also, the generality of the proposed channel model is demonstrated by the fact that it can easily be reduced to different simplified channel models to fit specific scenarios by properly adjusting model parameters. The proposed general channel model takes into consideration the non-stationarities in space, time, and frequency domains caused by ultra-massive MIMO, long traveling paths, and large bandwidths of THz communications, respectively. Statistical properties of the proposed general THz channel model are investigated. The accuracy and generality of the proposed channel model are verified by comparing the simulation results of the relative angle spread and root mean square (RMS) delay spread with corresponding channel measurements."
                    },
                    {
                        "title": "6G Oriented Wireless Communication Channel Characteristics Analysis and Modeling",
                        "abstract": "Based on the vision on the 6G wireless communication network, i.e., global coverage, all spectrums and all applications, we comprehensively survey 6G related wireless channel measurements, channel characteristics, and channel models for all frequency bands and all scenarios. Millimeter wave (mmWave), terahertz (THz), optical band, satellite, unmanned aerial vehicle (UAV), maritime, underwater acoustic, high-speed train (HST), vehicle-to-vehicle (V2V), massive/ ultra-massive multiple-input multiple-output (MIMO), orbital angular momentum (OAM), and industry Internet of things (IoT) communication channels were particularly investigated. The related 6G channel measurement and modeling results were also given. Finally, future research challenges on 6G channel measurements and modeling were pointed out."
                    },
                    {
                        "title": "PrivTrace: Differentially Private Trajectory Synthesis by Adaptive Markov Model",
                        "abstract": "Publishing trajectory data (individual's movement information) is very useful, but it also raises privacy concerns. To handle the privacy concern, in this paper, we apply differential privacy, the standard technique for data privacy, together with Markov chain model, to generate synthetic trajectories. We notice that existing studies all use Markov chain model and thus propose a framework to analyze the usage of the Markov chain model in this problem. Based on the analysis, we come up with an effective algorithm PrivTrace that uses the first-order and second-order Markov model adaptively. We evaluate PrivTrace and existing methods on synthetic and real-world datasets to demonstrate the superiority of our method."
                    },
                    {
                        "title": "Over-the-Air Federated Learning in Cell-Free MIMO with Long-term Power Constraint",
                        "abstract": "Wireless networks supporting artificial intelligence have gained significant attention, with Over-the-Air Federated Learning emerging as a key application due to its unique transmission and distributed computing characteristics. This paper derives error bounds for Over-the-Air Federated Learning in a Cell-free MIMO system and formulates an optimization problem to minimize optimality gap via joint optimization of power control and beamforming. We introduce the MOP-LOFPC algorithm, which employs Lyapunov optimization to decouple long-term constraints across rounds while requiring only causal channel state information. Experimental results demonstrate that MOP-LOFPC achieves a better and more flexible trade-off between the model's training loss and adherence to long-term power constraints compared to existing baselines."
                    },
                    {
                        "title": "A Novel 3D Non-Stationary GBSM for 6G THz Ultra-Massive MIMO Wireless Systems",
                        "abstract": "Terahertz (THz) communication is now being considered as one of possible technologies for the sixth generation (6G) wireless communication systems. In this paper, a novel three-dimensional (3D) space-time-frequency non-stationary theoretical channel model is first proposed for 6G THz wireless communication systems employing ultra-massive multiple-input multiple-output (MIMO) technologies with long traveling paths. Considering frequency-dependent diffuse scattering, which is a special property of THz channels different from millimeter wave (mmWave) channels, the relative angles and delays of rays within one cluster will evolve in the frequency domain. Then, a corresponding simulation model is proposed with discrete angles calculated using the method of equal area (MEA). The statistical properties of the proposed theoretical and simulation models are derived and compared, showing good agreements. The accuracy and flexibility of the proposed simulation model are demonstrated by comparing the simulation results of the relative angle spread and root mean square (RMS) delay spread with corresponding measurements."
                    }
                ]
            },
            "f8e6ddd9-e6f4-4341-bf35-22bbd09df054": {
                "pk": "f8e6ddd9-e6f4-4341-bf35-22bbd09df054",
                "name": "Jianqiao Lu",
                "collaborators": [
                    "Zhijiang Guo",
                    "Yingjia Wan",
                    "Yinya Huang",
                    "Wenyong Huang",
                    "Jing Xiong",
                    "Zhengying Liu",
                    "Xingshan Zeng",
                    "Zeyu Cao",
                    "Jianbo Dai",
                    "Wanjun Zhong"
                ],
                "domain": [
                    "Online Algorithms",
                    "Autoformalization",
                    "Speech Processing",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "Online Matching Meets Sampling Without Replacement",
                        "abstract": "Sampling without replacement is a natural online rounding strategy for converting fractional bipartite matching into an integral one. In Online Bipartite Matching, we can use the Balance algorithm to fractionally match each online vertex, and then sample an unmatched offline neighbor with probability proportional to the fractional matching. In Online Stochastic Matching, we can take the solution to a linear program relaxation as a reference, and then match each online vertex to an unmatched offline neighbor with probability proportional to the fractional matching of the online vertex's type. On the one hand, we find empirical evidence that online matching algorithms based on sampling without replacement outperform existing algorithms. On the other hand, the literature offers little theoretical understanding of the power of sampling without replacement in online matching problems.   This paper fills the gap in the literature by giving the first non-trivial competitive analyses of sampling without replacement for online matching problems. In Online Stochastic Matching, we develop a potential function analysis framework to show that sampling without replacement is at least $0.707$-competitive. The new analysis framework further allows us to derandomize the algorithm to obtain the first polynomial-time deterministic algorithm that breaks the $1-\\frac{1}{e}$ barrier. In Online Bipartite Matching, we show that sampling without replacement provides provable online correlated selection guarantees when the selection probabilities correspond to the fractional matching chosen by the Balance algorithm. As a result, we prove that sampling without replacement is at least $0.513$-competitive for Online Bipartite Matching."
                    },
                    {
                        "title": "FormalAlign: Automated Alignment Evaluation for Autoformalization",
                        "abstract": "Autoformalization aims to convert informal mathematical proofs into machine-verifiable formats, bridging the gap between natural and formal languages. However, ensuring semantic alignment between the informal and formalized statements remains challenging. Existing approaches heavily rely on manual verification, hindering scalability. To address this, we introduce \\textsc{FormalAlign}, the first automated framework designed for evaluating the alignment between natural and formal languages in autoformalization. \\textsc{FormalAlign} trains on both the autoformalization sequence generation task and the representational alignment between input and output, employing a dual loss that combines a pair of mutually enhancing autoformalization and alignment tasks. Evaluated across four benchmarks augmented by our proposed misalignment strategies, \\textsc{FormalAlign} demonstrates superior performance. In our experiments, \\textsc{FormalAlign} outperforms GPT-4, achieving an Alignment-Selection Score 11.58\\% higher on \\forml-Basic (99.21\\% vs. 88.91\\%) and 3.19\\% higher on MiniF2F-Valid (66.39\\% vs. 64.34\\%). This effective alignment evaluation significantly reduces the need for manual verification. Both the dataset and code can be accessed via~\\url{https://github.com/rookie-joe/FormalAlign}."
                    },
                    {
                        "title": "Improving End-to-End Speech Processing by Efficient Text Data Utilization with Latent Synthesis",
                        "abstract": "Training a high performance end-to-end speech (E2E) processing model requires an enormous amount of labeled speech data, especially in the era of data-centric artificial intelligence. However, labeled speech data are usually scarcer and more expensive for collection, compared to textual data. We propose Latent Synthesis (LaSyn), an efficient textual data utilization framework for E2E speech processing models. We train a latent synthesizer to convert textual data into an intermediate latent representation of a pre-trained speech model. These pseudo acoustic representations of textual data augment acoustic data for model training. We evaluate LaSyn on low-resource automatic speech recognition (ASR) and spoken language understanding (SLU) tasks. For ASR, LaSyn improves an E2E baseline trained on LibriSpeech train-clean-100, with relative word error rate reductions over 22.3% on different test sets. For SLU, LaSyn improves our E2E baseline by absolute 4.1% for intent classification accuracy and 3.8% for slot filling SLU-F1 on SLURP, and absolute 4.49% and 2.25% for exact match (EM) and EM-Tree accuracies on STOP respectively. With fewer parameters, the results of LaSyn are competitive to published state-of-the-art works. The results demonstrate the quality of the augmented training data."
                    },
                    {
                        "title": "Scaling Laws for Mixed quantization in Large Language Models",
                        "abstract": "Post-training quantization of Large Language Models (LLMs) has proven effective in reducing the computational requirements for running inference on these models. In this study, we focus on a straightforward question: When aiming for a specific accuracy or perplexity target for low-precision quantization, how many high-precision numbers or calculations are required to preserve as we scale LLMs to larger sizes? We first introduce a critical metric named the quantization ratio, which compares the number of parameters quantized to low-precision arithmetic against the total parameter count. Through extensive and carefully controlled experiments across different model families, arithmetic types, and quantization granularities (e.g. layer-wise, matmul-wise), we identify two central phenomenons. 1) The larger the models, the better they can preserve performance with an increased quantization ratio, as measured by perplexity in pre-training tasks or accuracy in downstream tasks. 2) The finer the granularity of mixed-precision quantization (e.g., matmul-wise), the more the model can increase the quantization ratio. We believe these observed phenomena offer valuable insights for future AI hardware design and the development of advanced Efficient AI algorithms."
                    },
                    {
                        "title": "MHPP: Exploring the Capabilities and Limitations of Language Models Beyond Basic Code Generation",
                        "abstract": "Recent advancements in large language models (LLMs) have greatly improved code generation, specifically at the function level. For instance, GPT-4 has achieved an 88.4% pass rate on HumanEval. However, this draws into question the adequacy of existing benchmarks in thoroughly assessing function-level code generation capabilities. Our study analyzed two common benchmarks, HumanEval and MBPP, and found that these might not thoroughly evaluate LLMs' code generation capacities due to limitations in quality, difficulty, and granularity. To resolve this, we introduce the Mostly Hard Python Problems (MHPP) dataset, consisting of 140 unique human-curated problems. By focusing on the combination of natural language and code reasoning, MHPP gauges LLMs' abilities to comprehend specifications and restrictions, engage in multi-step reasoning, and apply coding knowledge effectively. Initial evaluations of 22 LLMs using MHPP showed many high-performing models on HumanEval failed to achieve similar success on MHPP. Moreover, MHPP highlighted various previously undiscovered limitations within various LLMs, leading us to believe that it could pave the way for a better understanding of LLMs' capabilities and limitations. Dataset and code are available at https://github.com/SparksofAGI/MHPP."
                    },
                    {
                        "title": "AutoCV: Empowering Reasoning with Automated Process Labeling via Confidence Variation",
                        "abstract": "In this work, we propose a novel method named \\textbf{Auto}mated Process Labeling via \\textbf{C}onfidence \\textbf{V}ariation (\\textbf{\\textsc{AutoCV}}) to enhance the reasoning capabilities of large language models (LLMs) by automatically annotating the reasoning steps. Our approach begins by training a verification model on the correctness of final answers, enabling it to generate automatic process annotations. This verification model assigns a confidence score to each reasoning step, indicating the probability of arriving at the correct final answer from that point onward. We detect relative changes in the verification's confidence scores across reasoning steps to automatically annotate the reasoning process. This alleviates the need for numerous manual annotations or the high computational costs associated with model-induced annotation approaches. We experimentally validate that the confidence variations learned by the verification model trained on the final answer correctness can effectively identify errors in the reasoning steps. Subsequently, we demonstrate that the process annotations generated by \\textsc{AutoCV} can improve the accuracy of the verification model in selecting the correct answer from multiple outputs generated by LLMs. Notably, we achieve substantial improvements across five datasets in mathematics and commonsense reasoning. The source code of \\textsc{AutoCV} is available at \\url{https://github.com/rookie-joe/AUTOCV}."
                    },
                    {
                        "title": "SELF: Self-Evolution with Language Feedback",
                        "abstract": "Large Language Models (LLMs) have demonstrated remarkable versatility across various domains. To further advance LLMs, we propose 'SELF' (Self-Evolution with Language Feedback), a novel approach that enables LLMs to self-improve through self-reflection, akin to human learning processes. SELF initiates with a meta-skill learning process that equips the LLMs with capabilities for self-feedback and self-refinement. Subsequently, the model undergoes an iterative process of self-evolution. In each iteration, it utilizes an unlabeled dataset of instructions to generate initial responses. These responses are enhanced through self-feedback and self-refinement. The model is then fine-tuned using this enhanced data. The model undergoes progressive improvement through this iterative self-evolution process. Moreover, the SELF framework enables the model to apply self-refinement during inference, which further improves response quality. Our experiments in mathematics and general tasks demonstrate that SELF can enhance the capabilities of LLMs without human intervention. The SELF framework indicates a promising direction for the autonomous evolution of LLMs, transitioning them from passive information receivers to active participants in their development."
                    },
                    {
                        "title": "YODA: Teacher-Student Progressive Learning for Language Models",
                        "abstract": "Although large language models (LLMs) have demonstrated adeptness in a range of tasks, they still lag behind human learning efficiency. This disparity is often linked to the inherent human capacity to learn from basic examples, gradually generalize and handle more complex problems, and refine their skills with continuous feedback. Inspired by this, this paper introduces YODA, a novel teacher-student progressive learning framework that emulates the teacher-student education process to improve the efficacy of model fine-tuning. The framework operates on an interactive \\textit{basic-generalized-harder} loop. The teacher agent provides tailored feedback on the student's answers, and systematically organizes the education process. This process unfolds by teaching the student basic examples, reinforcing understanding through generalized questions, and then enhancing learning by posing questions with progressively enhanced complexity. With the teacher's guidance, the student learns to iteratively refine its answer with feedback, and forms a robust and comprehensive understanding of the posed questions. The systematic procedural data, which reflects the progressive learning process of humans, is then utilized for model training. Taking math reasoning as a testbed, experiments show that training LLaMA2 with data from YODA improves SFT with significant performance gain (+17.01\\% on GSM8K and +9.98\\% on MATH). In addition, we find that training with curriculum learning further improves learning robustness."
                    },
                    {
                        "title": "Process-Driven Autoformalization in Lean 4",
                        "abstract": "Autoformalization, the conversion of natural language mathematics into formal languages, offers significant potential for advancing mathematical reasoning. However, existing efforts are limited to formal languages with substantial online corpora and struggle to keep pace with rapidly evolving languages like Lean 4. To bridge this gap, we propose a new benchmark \\textbf{Form}alization for \\textbf{L}ean~\\textbf{4} (\\textbf{\\name}) designed to evaluate the autoformalization capabilities of large language models (LLMs). This benchmark encompasses a comprehensive assessment of questions, answers, formal statements, and proofs. Additionally, we introduce a \\textbf{P}rocess-\\textbf{S}upervised \\textbf{V}erifier (\\textbf{PSV}) model that leverages the precise feedback from Lean 4 compilers to enhance autoformalization. Our experiments demonstrate that the PSV method improves autoformalization, enabling higher accuracy using less filtered training data. Furthermore, when fine-tuned with data containing detailed process information, PSV can leverage the data more effectively, leading to more significant improvements in autoformalization for Lean 4. Our dataset and code are available at \\url{https://github.com/rookie-joe/PDA}."
                    }
                ]
            },
            "a1e799d3-3534-466f-940b-9c55899dca94": {
                "pk": "a1e799d3-3534-466f-940b-9c55899dca94",
                "name": "Zhengying Liu",
                "collaborators": [
                    "Zhenguo Li",
                    "Yinya Huang",
                    "Haiming Wang",
                    "Chuanyang Zheng",
                    "Xiaodan Liang",
                    "Huajian Xin",
                    "Lin Li",
                    "Enze Xie",
                    "Qingxing Cao",
                    "Jing Xiong"
                ],
                "domain": [
                    "Automated Theorem Proving",
                    "Large Language Models",
                    "Machine Learning",
                    "AutoML"
                ],
                "publications": [
                    {
                        "title": "Learning to Prove Trigonometric Identities",
                        "abstract": "Automatic theorem proving with deep learning methods has attracted attentions recently. In this paper, we construct an automatic proof system for trigonometric identities. We define the normalized form of trigonometric identities, design a set of rules for the proof and put forward a method which can generate theoretically infinite trigonometric identities. Our goal is not only to complete the proof, but to complete the proof in as few steps as possible. For this reason, we design a model to learn proof data generated by random BFS (rBFS), and it is proved theoretically and experimentally that the model can outperform rBFS after a simple imitation learning. After further improvement through reinforcement learning, we get AutoTrig, which can give proof steps for identities in almost as short steps as BFS (theoretically shortest method), with a time cost of only one-thousandth. In addition, AutoTrig also beats Sympy, Matlab and human in the synthetic dataset, and performs well in many generalization tasks."
                    },
                    {
                        "title": "AgEBO-Tabular: Joint Neural Architecture and Hyperparameter Search with Autotuned Data-Parallel Training for Tabular Data",
                        "abstract": "Developing high-performing predictive models for large tabular data sets is a challenging task. The state-of-the-art methods are based on expert-developed model ensembles from different supervised learning methods. Recently, automated machine learning (AutoML) is emerging as a promising approach to automate predictive model development. Neural architecture search (NAS) is an AutoML approach that generates and evaluates multiple neural network architectures concurrently and improves the accuracy of the generated models iteratively. A key issue in NAS, particularly for large data sets, is the large computation time required to evaluate each generated architecture. While data-parallel training is a promising approach that can address this issue, its use within NAS is difficult. For different data sets, the data-parallel training settings such as the number of parallel processes, learning rate, and batch size need to be adapted to achieve high accuracy and reduction in training time. To that end, we have developed AgEBO-Tabular, an approach to combine aging evolution (AgE), a parallel NAS method that searches over neural architecture space, and an asynchronous Bayesian optimization method for tuning the hyperparameters of the data-parallel training simultaneously. We demonstrate the efficacy of the proposed method to generate high-performing neural network models for large tabular benchmark data sets. Furthermore, we demonstrate that the automatically discovered neural network models using our method outperform the state-of-the-art AutoML ensemble models in inference speed by two orders of magnitude while reaching similar accuracy values."
                    },
                    {
                        "title": "Progressive-Hint Prompting Improves Reasoning in Large Language Models",
                        "abstract": "The performance of Large Language Models (LLMs) in reasoning tasks depends heavily on prompt design, with Chain-of-Thought (CoT) and self-consistency being critical methods that enhance this ability. However, these methods do not fully exploit the answers generated by the LLM to guide subsequent responses. This paper proposes a new prompting method, named Progressive-Hint Prompting (PHP), that enables automatic multiple interactions between users and LLMs by using previously generated answers as hints to progressively guide toward the correct answers. PHP is orthogonal to CoT and self-consistency, making it easy to combine with state-of-the-art techniques to further improve performance. We conducted extensive and comprehensive experiments on seven benchmarks. The results show that PHP significantly improves accuracy while remaining highly efficient. For instance, with text-davinci-003, we observed a 4.2% improvement on GSM8K with greedy decoding compared to Complex CoT, and a 46.17% reduction in sample paths with self-consistency. With GPT-4 and PHP, we achieve state-of-the-art performances on SVAMP (89.1% -> 91.9%), GSM8K (92% -> 95.5%), AQuA (76.4% -> 79.9%) and MATH (50.3% -> 53.9%)."
                    },
                    {
                        "title": "ATG: Benchmarking Automated Theorem Generation for Generative Language Models",
                        "abstract": "Humans can develop new theorems to explore broader and more complex mathematical results. While current generative language models (LMs) have achieved significant improvement in automatically proving theorems, their ability to generate new or reusable theorems is still under-explored. Without the new theorems, current LMs struggle to prove harder theorems that are distant from the given hypotheses with the exponentially growing search space. Therefore, this paper proposes an Automated Theorem Generation (ATG) benchmark that evaluates whether an agent can automatically generate valuable (and possibly brand new) theorems that are applicable for downstream theorem proving as reusable knowledge. Specifically, we construct the ATG benchmark by splitting the Metamath library into three sets: axioms, library, and problem based on their proving depth. We conduct extensive experiments to investigate whether current LMs can generate theorems in the library and benefit the problem theorems proving. The results demonstrate that high-quality ATG data facilitates models' performances on downstream ATP. However, there is still room for current LMs to develop better ATG and generate more advanced and human-like theorems. We hope the new ATG challenge can shed some light on advanced complex theorem proving."
                    },
                    {
                        "title": "FormalAlign: Automated Alignment Evaluation for Autoformalization",
                        "abstract": "Autoformalization aims to convert informal mathematical proofs into machine-verifiable formats, bridging the gap between natural and formal languages. However, ensuring semantic alignment between the informal and formalized statements remains challenging. Existing approaches heavily rely on manual verification, hindering scalability. To address this, we introduce \\textsc{FormalAlign}, the first automated framework designed for evaluating the alignment between natural and formal languages in autoformalization. \\textsc{FormalAlign} trains on both the autoformalization sequence generation task and the representational alignment between input and output, employing a dual loss that combines a pair of mutually enhancing autoformalization and alignment tasks. Evaluated across four benchmarks augmented by our proposed misalignment strategies, \\textsc{FormalAlign} demonstrates superior performance. In our experiments, \\textsc{FormalAlign} outperforms GPT-4, achieving an Alignment-Selection Score 11.58\\% higher on \\forml-Basic (99.21\\% vs. 88.91\\%) and 3.19\\% higher on MiniF2F-Valid (66.39\\% vs. 64.34\\%). This effective alignment evaluation significantly reduces the need for manual verification. Both the dataset and code can be accessed via~\\url{https://github.com/rookie-joe/FormalAlign}."
                    },
                    {
                        "title": "Large Language Models as Automated Aligners for benchmarking Vision-Language Models",
                        "abstract": "With the advancements in Large Language Models (LLMs), Vision-Language Models (VLMs) have reached a new level of sophistication, showing notable competence in executing intricate cognition and reasoning tasks. However, existing evaluation benchmarks, primarily relying on rigid, hand-crafted datasets to measure task-specific performance, face significant limitations in assessing the alignment of these increasingly anthropomorphic models with human intelligence. In this work, we address the limitations via Auto-Bench, which delves into exploring LLMs as proficient aligners, measuring the alignment between VLMs and human intelligence and value through automatic data curation and assessment. Specifically, for data curation, Auto-Bench utilizes LLMs (e.g., GPT-4) to automatically generate a vast set of question-answer-reasoning triplets via prompting on visual symbolic representations (e.g., captions, object locations, instance relationships, and etc.). The curated data closely matches human intent, owing to the extensive world knowledge embedded in LLMs. Through this pipeline, a total of 28.5K human-verified and 3,504K unfiltered question-answer-reasoning triplets have been curated, covering 4 primary abilities and 16 sub-abilities. We subsequently engage LLMs like GPT-3.5 to serve as judges, implementing the quantitative and qualitative automated assessments to facilitate a comprehensive evaluation of VLMs. Our validation results reveal that LLMs are proficient in both evaluation data curation and model assessment, achieving an average agreement rate of 85%. We envision Auto-Bench as a flexible, scalable, and comprehensive benchmark for evaluating the evolving sophisticated VLMs."
                    },
                    {
                        "title": "LEAP nets for power grid perturbations",
                        "abstract": "We propose a novel neural network embedding approach to model power transmission grids, in which high voltage lines are disconnected and reconnected with one-another from time to time, either accidentally or willfully. We call our architeture LEAP net, for Latent Encoding of Atypical Perturbation. Our method implements a form of transfer learning, permitting to train on a few source domains, then generalize to new target domains, without learning on any example of that domain. We evaluate the viability of this technique to rapidly assess cu-rative actions that human operators take in emergency situations, using real historical data, from the French high voltage power grid."
                    },
                    {
                        "title": "Advances in MetaDL: AAAI 2021 challenge and workshop",
                        "abstract": "To stimulate advances in metalearning using deep learning techniques (MetaDL), we organized in 2021 a challenge and an associated workshop. This paper presents the design of the challenge and its results, and summarizes presentations made at the workshop. The challenge focused on few-shot learning classification tasks of small images. Participants' code submissions were run in a uniform manner, under tight computational constraints. This put pressure on solution designs to use existing architecture backbones and/or pre-trained networks. Winning methods featured various classifiers trained on top of the second last layer of popular CNN backbones, fined-tuned on the meta-training data (not necessarily in an episodic manner), then trained on the labeled support and tested on the unlabeled query sets of the meta-test data."
                    },
                    {
                        "title": "Forward-Backward Reasoning in Large Language Models for Mathematical Verification",
                        "abstract": "Self-Consistency samples diverse reasoning chains with answers and chooses the final answer by majority voting. It is based on forward reasoning and cannot further improve performance by sampling more reasoning chains when saturated. To further boost performance, we introduce backward reasoning to verify candidate answers. Specifically, for mathematical tasks, we mask a number in the question and ask the LLM to answer a backward question created by a simple template, i.e., to predict the masked number when a candidate answer is provided. Instead of using forward or backward reasoning alone, we propose FOBAR to combine FOrward and BAckward Reasoning for verification. Extensive experiments on six standard mathematical data sets and three LLMs show that FOBAR achieves state-of-the-art performance. In particular, FOBAR outperforms Self-Consistency, which uses forward reasoning alone, demonstrating that combining forward and forward reasoning is better. In addition, FOBAR performs better than existing verification methods, showing the effectiveness of the simple template used in backward reasoning and the proposed combination. Extensions to non-mathematical problems are also discussed and validated empirically."
                    },
                    {
                        "title": "Lyra: Orchestrating Dual Correction in Automated Theorem Proving",
                        "abstract": "Large Language Models (LLMs) present an intriguing avenue for exploration in the field of formal theorem proving. Nevertheless, their full potential, particularly concerning the mitigation of hallucinations and refinement through prover error messages, remains an area that has yet to be thoroughly investigated. To enhance the effectiveness of LLMs in the field, we introduce the Lyra, a new framework that employs two distinct correction mechanisms: Tool Correction (TC) and Conjecture Correction (CC). To implement Tool Correction in the post-processing of formal proofs, we leverage prior knowledge to utilize predefined prover tools (e.g., Sledgehammer) for guiding the replacement of incorrect tools. Tool Correction significantly contributes to mitigating hallucinations, thereby improving the overall accuracy of the proof. In addition, we introduce Conjecture Correction, an error feedback mechanism designed to interact with prover to refine formal proof conjectures with prover error messages. Compared to the previous refinement framework, the proposed Conjecture Correction refines generation with instruction but does not collect paired (generation, error & refinement) prompts. Our method has achieved state-of-the-art (SOTA) performance on both miniF2F validation (48.0% -> 55.3%) and test (45.5% -> 51.2%). We also present 3 IMO problems solved by Lyra. We believe Tool Correction (post-process for hallucination mitigation) and Conjecture Correction (subgoal adjustment from interaction with environment) could provide a promising avenue for future research in this field."
                    },
                    {
                        "title": "MUSTARD: Mastering Uniform Synthesis of Theorem and Proof Data",
                        "abstract": "Recent large language models (LLMs) have witnessed significant advancement in various tasks, including mathematical reasoning and theorem proving. As these two tasks require strict and formal multi-step inference, they are appealing domains for exploring the reasoning ability of LLMs but still face important challenges. Previous studies such as Chain-of-Thought (CoT) have revealed the effectiveness of intermediate steps guidance. However, such step-wise annotation requires heavy labor, leading to insufficient training steps for current benchmarks. To fill this gap, this work introduces MUSTARD, a data generation framework that masters uniform synthesis of theorem and proof data of high quality and diversity. MUSTARD synthesizes data in three stages: (1) It samples a few mathematical concept seeds as the problem category. (2) Then, it prompts a generative language model with the sampled concepts to obtain both the problems and their step-wise formal solutions. (3) Lastly, the framework utilizes a proof assistant (e.g., Lean Prover) to filter the valid proofs. With the proposed MUSTARD, we present a theorem-and-proof benchmark MUSTARDSAUCE with 5,866 valid data points. Each data point contains an informal statement, an informal proof, and a translated formal proof that passes the prover validation. We perform extensive analysis and demonstrate that MUSTARD generates validated high-quality step-by-step data. We further apply the MUSTARDSAUCE for fine-tuning smaller language models. The fine-tuned Llama 2-7B achieves a 15.41% average relative performance gain in automated theorem proving, and 8.18% in math word problems. Codes and data are available at https://github.com/Eleanor-H/MUSTARD."
                    },
                    {
                        "title": "FIMO: A Challenge Formal Dataset for Automated Theorem Proving",
                        "abstract": "We present FIMO, an innovative dataset comprising formal mathematical problem statements sourced from the International Mathematical Olympiad (IMO) Shortlisted Problems. Designed to facilitate advanced automated theorem proving at the IMO level, FIMO is currently tailored for the Lean formal language. It comprises 149 formal problem statements, accompanied by both informal problem descriptions and their corresponding LaTeX-based informal proofs. Through initial experiments involving GPT-4, our findings underscore the existing limitations in current methodologies, indicating a substantial journey ahead before achieving satisfactory IMO-level automated theorem proving outcomes."
                    },
                    {
                        "title": "MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Models",
                        "abstract": "Large language models (LLMs) have pushed the limits of natural language understanding and exhibited excellent problem-solving ability. Despite the great success, most existing open-source LLMs (e.g., LLaMA-2) are still far away from satisfactory for solving mathematical problem due to the complex reasoning procedures. To bridge this gap, we propose MetaMath, a fine-tuned language model that specializes in mathematical reasoning. Specifically, we start by bootstrapping mathematical questions by rewriting the question from multiple perspectives without extra knowledge, which results in a new dataset called MetaMathQA. Then we fine-tune the LLaMA-2 models on MetaMathQA. Experimental results on two popular benchmarks (i.e., GSM8K and MATH) for mathematical reasoning demonstrate that MetaMath outperforms a suite of open-source LLMs by a significant margin. Our MetaMath-7B model achieves 66.4% on GSM8K and 19.4% on MATH, exceeding the state-of-the-art models of the same size by 11.5% and 8.7%. Particularly, MetaMath-70B achieves an accuracy of 82.3% on GSM8K, slightly better than GPT-3.5-Turbo. We release all the MetaMathQA dataset, the MetaMath models with different model sizes and the training code for public use."
                    },
                    {
                        "title": "TRIGO: Benchmarking Formal Mathematical Proof Reduction for Generative Language Models",
                        "abstract": "Automated theorem proving (ATP) has become an appealing domain for exploring the reasoning ability of the recent successful generative language models. However, current ATP benchmarks mainly focus on symbolic inference, but rarely involve the understanding of complex number combination reasoning. In this work, we propose TRIGO, an ATP benchmark that not only requires a model to reduce a trigonometric expression with step-by-step proofs but also evaluates a generative LM's reasoning ability on formulas and its capability to manipulate, group, and factor number terms. We gather trigonometric expressions and their reduced forms from the web, annotate the simplification process manually, and translate it into the Lean formal language system. We then automatically generate additional examples from the annotated samples to expand the dataset. Furthermore, we develop an automatic generator based on Lean-Gym to create dataset splits of varying difficulties and distributions in order to thoroughly analyze the model's generalization ability. Our extensive experiments show our proposed TRIGO poses a new challenge for advanced generative LM's including GPT-4 which is pre-trained on a considerable amount of open-source formal theorem-proving language data, and provide a new tool to study the generative LM's ability on both formal and mathematical reasoning."
                    },
                    {
                        "title": "Gaining Wisdom from Setbacks: Aligning Large Language Models via Mistake Analysis",
                        "abstract": "The rapid development of large language models (LLMs) has not only provided numerous opportunities but also presented significant challenges. This becomes particularly evident when LLMs inadvertently generate harmful or toxic content, either unintentionally or because of intentional inducement. Existing alignment methods usually direct LLMs toward the favorable outcomes by utilizing human-annotated, flawless instruction-response pairs. Conversely, this study proposes a novel alignment technique based on mistake analysis, which deliberately exposes LLMs to erroneous content to learn the reasons for mistakes and how to avoid them. In this case, mistakes are repurposed into valuable data for alignment, effectively helping to avoid the production of erroneous responses. Without external models or human annotations, our method leverages a model's intrinsic ability to discern undesirable mistakes and improves the safety of its generated responses. Experimental results reveal that our method outperforms existing alignment approaches in enhancing model safety while maintaining the overall utility."
                    },
                    {
                        "title": "Process-Driven Autoformalization in Lean 4",
                        "abstract": "Autoformalization, the conversion of natural language mathematics into formal languages, offers significant potential for advancing mathematical reasoning. However, existing efforts are limited to formal languages with substantial online corpora and struggle to keep pace with rapidly evolving languages like Lean 4. To bridge this gap, we propose a new benchmark \\textbf{Form}alization for \\textbf{L}ean~\\textbf{4} (\\textbf{\\name}) designed to evaluate the autoformalization capabilities of large language models (LLMs). This benchmark encompasses a comprehensive assessment of questions, answers, formal statements, and proofs. Additionally, we introduce a \\textbf{P}rocess-\\textbf{S}upervised \\textbf{V}erifier (\\textbf{PSV}) model that leverages the precise feedback from Lean 4 compilers to enhance autoformalization. Our experiments demonstrate that the PSV method improves autoformalization, enabling higher accuracy using less filtered training data. Furthermore, when fine-tuned with data containing detailed process information, PSV can leverage the data more effectively, leading to more significant improvements in autoformalization for Lean 4. Our dataset and code are available at \\url{https://github.com/rookie-joe/PDA}."
                    },
                    {
                        "title": "Proving Theorems Recursively",
                        "abstract": "Recent advances in automated theorem proving leverages language models to explore expanded search spaces by step-by-step proof generation. However, such approaches are usually based on short-sighted heuristics (e.g., log probability or value function scores) that potentially lead to suboptimal or even distracting subgoals, preventing us from finding longer proofs. To address this challenge, we propose POETRY (PrOvE Theorems RecursivelY), which proves theorems in a recursive, level-by-level manner in the Isabelle theorem prover. Unlike previous step-by-step methods, POETRY searches for a verifiable sketch of the proof at each level and focuses on solving the current level's theorem or conjecture. Detailed proofs of intermediate conjectures within the sketch are temporarily replaced by a placeholder tactic called sorry, deferring their proofs to subsequent levels. This approach allows the theorem to be tackled incrementally by outlining the overall theorem at the first level and then solving the intermediate conjectures at deeper levels. Experiments are conducted on the miniF2F and PISA datasets and significant performance gains are observed in our POETRY approach over state-of-the-art methods. POETRY on miniF2F achieves an average proving success rate improvement of 5.1%. Moreover, we observe a substantial increase in the maximum proof length found by POETRY, from 10 to 26."
                    },
                    {
                        "title": "LEGO-Prover: Neural Theorem Proving with Growing Libraries",
                        "abstract": "Despite the success of large language models (LLMs), the task of theorem proving still remains one of the hardest reasoning tasks that is far from being fully solved. Prior methods using language models have demonstrated promising results, but they still struggle to prove even middle school level theorems. One common limitation of these methods is that they assume a fixed theorem library during the whole theorem proving process. However, as we all know, creating new useful theorems or even new theories is not only helpful but crucial and necessary for advancing mathematics and proving harder and deeper results. In this work, we present LEGO-Prover, which employs a growing skill library containing verified lemmas as skills to augment the capability of LLMs used in theorem proving. By constructing the proof modularly, LEGO-Prover enables LLMs to utilize existing skills retrieved from the library and to create new skills during the proving process. These skills are further evolved (by prompting an LLM) to enrich the library on another scale. Modular and reusable skills are constantly added to the library to enable tackling increasingly intricate mathematical problems. Moreover, the learned library further bridges the gap between human proofs and formal proofs by making it easier to impute missing steps. LEGO-Prover advances the state-of-the-art pass rate on miniF2F-valid (48.0% to 57.0%) and miniF2F-test (45.5% to 47.1%). During the proving process, LEGO-Prover also manages to generate over 20,000 skills (theorems/lemmas) and adds them to the growing library. Our ablation study indicates that these newly added skills are indeed helpful for proving theorems, resulting in an improvement from a success rate of 47.1% to 50.4%. We also release our code and all the generated skills."
                    }
                ]
            },
            "77254400-4380-4ee7-95fb-a13e57d2f39f": {
                "pk": "77254400-4380-4ee7-95fb-a13e57d2f39f",
                "name": "Linqi Song",
                "collaborators": [
                    "Christina Fragouli",
                    "Tan Li",
                    "Mohammed Karmoose",
                    "Martina Cardone",
                    "Jie Xu",
                    "Haoyue Tang",
                    "Jintao Wang",
                    "Jian Song",
                    "Erkao Bao",
                    "Devavrat Shah"
                ],
                "domain": [
                    "Active Learning",
                    "Multi-Armed Bandit",
                    "Index Coding",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "Stream-based Online Active Learning in a Contextual Multi-Armed Bandit Framework",
                        "abstract": "We study the stream-based online active learning in a contextual multi-armed bandit framework. In this framework, the reward depends on both the arm and the context. In a stream-based active learning setting, obtaining the ground truth of the reward is costly, and the conventional contextual multi-armed bandit algorithm fails to achieve a sublinear regret due to this cost. Hence, the algorithm needs to determine whether or not to request the ground truth of the reward at current time slot. In our framework, we consider a stream-based active learning setting in which a query request for the ground truth is sent to the annotator, together with some prior information of the ground truth. Depending on the accuracy of the prior information, the query cost varies. Our algorithm mainly carries out two operations: the refinement of the context and arm spaces and the selection of actions. In our algorithm, the partitions of the context space and the arm space are maintained for a certain time slots, and then become finer as more information about the rewards accumulates. We use a strategic way to select the arms and to request the ground truth of the reward, aiming to maximize the total reward. We analytically show that the regret is sublinear and in the same order with that of the conventional contextual multi-armed bandit algorithms, where no query cost"
                    },
                    {
                        "title": "Content-type coding",
                        "abstract": "This paper is motivated by the observation that, in many cases, we do not need to serve specific messages, but rather, any message within a content-type. Content-type traffic pervades a host of applications today, ranging from search engines and recommender networks to newsfeeds and advertisement networks. The paper asks a novel question: if there are benefits in designing network and channel codes specifically tailored to content-type requests. It provides three examples of content-type formulations to argue that, indeed in some cases we can have significant such benefits."
                    },
                    {
                        "title": "Equivariant neural networks and equivarification",
                        "abstract": "We provide a process to modify a neural network to an equivariant one, which we call equivarification. As an illustration, we build an equivariant neural network for image classification by equivarifying a convolutional neural network."
                    },
                    {
                        "title": "A Deterministic Algorithm for Pliable Index Coding",
                        "abstract": "Pliable index coding considers a server with m messages, and n clients where each has as side information a subset of the messages. We seek to minimize the number of transmissions the server should make, so that each client receives (any) one message she does not already have. Previous work has shown that the server can achieve this using O(\\log^2(n)) transmissions and needs at least \\Omega(log(n)) transmissions in the worst case, but finding a code of optimal length is NP-hard. In this paper, we propose a deterministic algorithm that we prove achieves this upper bound, that is, in an order almost as the worst-case optimal code length. We also establish a connection between the pliable index coding problem and the minrank problem over a family of mixed matrices."
                    },
                    {
                        "title": "Making recommendations bandwidth aware",
                        "abstract": "This paper asks how much we can gain in terms of bandwidth and user satisfaction, if recommender systems became bandwidth aware and took into account not only the user preferences, but also the fact that they may need to serve these users under bandwidth constraints, as is the case over wireless networks. We formulate this as a new problem in the context of index coding: we relax the index coding requirements to capture scenarios where each client has preferences associated with messages. The client is satisfied to receive any message she does not already have, with a satisfaction proportional to her preference for that message. We consistently find, over a number of scenarios we sample, that although the optimization problems are in general NP-hard, significant bandwidth savings are possible even when restricted to polynomial time algorithms."
                    },
                    {
                        "title": "Dynamic Edge Caching with Popularity Drifting",
                        "abstract": "Caching at the network edge devices such as wireless caching stations (WCS) is a key technology in the 5G network. The spatial-temporal diversity of content popularity requires different content to be cached in different WCSs and periodically updated to adapt to temporal changes. In this paper, we study how the popularity drifting speed affects the number of required broadcast transmissions by the MBS and then design coded transmission schemes by leveraging the broadcast advantage under the index coding framework. The key idea is that files already cached in WCSs, which although may be currently unpopular, can serve as side information to facilitate coded broadcast transmission for cache updating. Our algorithm extends existing index coding-based schemes from a single-request scenario to a multiple-request scenario via a \"dynamic coloring\" approach. Simulation results indicate that a significant bandwidth saving can be achieved by adopting our scheme."
                    },
                    {
                        "title": "A Polynomial-Time Algorithm for Pliable Index Coding",
                        "abstract": "In pliable index coding, we consider a server with $m$ messages and $n$ clients where each client has as side information a subset of the messages. We seek to minimize the number of broadcast transmissions, so that each client can recover any one unknown message she does not already have. Previous work has shown that the pliable index coding problem is NP-hard and requires at most $\\mathcal{O}(\\log^2(n))$ broadcast transmissions, which indicates exponential savings over the conventional index coding that requires in the worst case $\\mathcal{O}(n)$ transmissions. In this work, building on a decoding criterion that we propose, we first design a deterministic polynomial-time algorithm that can realize the exponential benefits, by achieving, in the worst case, a performance upper bounded by $\\mathcal{O}(\\log^2(n))$ broadcast transmissions. We extend our algorithm to the $t$-requests case, where each client requires $t$ unknown messages that she does not have, and show that our algorithm requires at most $\\mathcal{O}(t\\log(n)+\\log^2(n))$ broadcast transmissions. We construct lower bound instances that require at least $\\Omega(\\log(n))$ transmissions for linear pliable index coding and at least $\\Omega(t+\\log(n))$ transmissions for the $t$-requests case, indicating that both our upper and lower bounds are polynomials of $\\log(n)$ and differ within a factor of $\\mathcal{O}(\\log(n))$. Finally, we provide a probabilistic analysis and show that the required number of transmissions is almost surely $\\Theta(\\log(n))$, as compared to $\\Theta(n/\\log(n))$ for index coding. Our numerical experiments show that our algorithm outperforms existing algorithms for pliable index coding by up to $50\\%$ less transmissions."
                    },
                    {
                        "title": "A Contextual Bandit Approach for Stream-Based Active Learning",
                        "abstract": "Contextual bandit algorithms -- a class of multi-armed bandit algorithms that exploit the contextual information -- have been shown to be effective in solving sequential decision making problems under uncertainty. A common assumption adopted in the literature is that the realized (ground truth) reward by taking the selected action is observed by the learner at no cost, which, however, is not realistic in many practical scenarios. When observing the ground truth reward is costly, a key challenge for the learner is how to judiciously acquire the ground truth by assessing the benefits and costs in order to balance learning efficiency and learning cost. From the information theoretic perspective, a perhaps even more interesting question is how much efficiency might be lost due to this cost. In this paper, we design a novel contextual bandit-based learning algorithm and endow it with the active learning capability. The key feature of our algorithm is that in addition to sending a query to an annotator for the ground truth, prior information about the ground truth learned by the learner is sent together, thereby reducing the query cost. We prove that by carefully choosing the algorithm parameters, the learning regret of the proposed algorithm achieves the same order as that of conventional contextual bandit algorithms in cost-free scenarios, implying that, surprisingly, cost due to acquiring the ground truth does not increase the learning regret in the long-run. Our analysis shows that prior information about the ground truth plays a critical role in improving the system performance in scenarios where active learning is necessary."
                    },
                    {
                        "title": "Privacy-Preserving Communication-Efficient Federated Multi-Armed Bandits",
                        "abstract": "Communication bottleneck and data privacy are two critical concerns in federated multi-armed bandit (MAB) problems, such as situations in decision-making and recommendations of connected vehicles via wireless. In this paper, we design the privacy-preserving communication-efficient algorithm in such problems and study the interactions among privacy, communication and learning performance in terms of the regret. To be specific, we design privacy-preserving learning algorithms and communication protocols and derive the learning regret when networked private agents are performing online bandit learning in a master-worker, a decentralized and a hybrid structure. Our bandit learning algorithms are based on epoch-wise sub-optimal arm eliminations at each agent and agents exchange learning knowledge with the server/each other at the end of each epoch. Furthermore, we adopt the differential privacy (DP) approach to protect the data privacy at each agent when exchanging information; and we curtail communication costs by making less frequent communications with fewer agents participation. By analyzing the regret of our proposed algorithmic framework in the master-worker, decentralized and hybrid structures, we theoretically show tradeoffs between regret and communication costs/privacy. Finally, we empirically show these trade-offs which are consistent with our theoretical analysis."
                    },
                    {
                        "title": "Regret vs. Bandwidth Trade-off for Recommendation Systems",
                        "abstract": "We consider recommendation systems that need to operate under wireless bandwidth constraints, measured as number of broadcast transmissions, and demonstrate a (tight for some instances) tradeoff between regret and bandwidth for two scenarios: the case of multi-armed bandit with context, and the case where there is a latent structure in the message space that we can exploit to reduce the learning phase."
                    },
                    {
                        "title": "A Pliable Index Coding Approach to Data Shuffling",
                        "abstract": "A promising research area that has recently emerged, is on how to use index coding to improve the communication efficiency in distributed computing systems, especially for data shuffling in iterative computations. In this paper, we posit that pliable index coding can offer a more efficient framework for data shuffling, as it can better leverage the many possible shuffling choices to reduce the number of transmissions. We theoretically analyze pliable index coding under data shuffling constraints, and design a hierarchical data-shuffling scheme that uses pliable coding as a component. We find benefits up to $O(ns/m)$ over index coding, where $ns/m$ is the average number of workers caching a message, and $m$, $n$, and $s$ are the numbers of messages, workers, and cache size, respectively."
                    },
                    {
                        "title": "The Benefit of Being Flexible in Distributed Computation",
                        "abstract": "In wireless distributed computing, networked nodes perform intermediate computations over data placed in their memory and exchange these intermediate values to calculate function values. In this paper we consider an asymmetric setting where each node has access to a random subset of the data, i.e., we cannot control the data placement. The paper makes a simple point: we can realize significant benefits if we are allowed to be \"flexible\", and decide which node computes which function, in our system. We make this argument in the case where each function depends on only two of the data messages, as is the case in similarity searches. We establish a percolation in the behavior of the system, where, depending on the amount of observed data, by being flexible, we may need no communication at all."
                    },
                    {
                        "title": "Federated Recommendation System via Differential Privacy",
                        "abstract": "In this paper, we are interested in what we term the federated private bandits framework, that combines differential privacy with multi-agent bandit learning. We explore how differential privacy based Upper Confidence Bound (UCB) methods can be applied to multi-agent environments, and in particular to federated learning environments both in `master-worker' and `fully decentralized' settings. We provide a theoretical analysis on the privacy and regret performance of the proposed methods and explore the tradeoffs between these two."
                    },
                    {
                        "title": "CSAGN: Conversational Structure Aware Graph Network for Conversational Semantic Role Labeling",
                        "abstract": "Conversational semantic role labeling (CSRL) is believed to be a crucial step towards dialogue understanding. However, it remains a major challenge for existing CSRL parser to handle conversational structural information. In this paper, we present a simple and effective architecture for CSRL which aims to address this problem. Our model is based on a conversational structure-aware graph network which explicitly encodes the speaker dependent information. We also propose a multi-task learning method to further improve the model. Experimental results on benchmark datasets show that our model with our proposed training objectives significantly outperforms previous baselines."
                    },
                    {
                        "title": "Minimizing Age of Information with Power Constraints: Multi-user Opportunistic Scheduling in Multi-State Time-Varying Channels",
                        "abstract": "This work is motivated by the need of collecting fresh data from power-constrained sensors in the industrial Internet of Things (IIoT) network. A recently proposed metric, the Age of Information (AoI) is adopted to measure data freshness from the perspective of the central controller in the IIoT network. We wonder what is the minimum average AoI the network can achieve and how to design scheduling algorithms to approach it. To answer these questions when the channel states of the network are Markov time-varying and scheduling decisions are restricted to bandwidth constraint, we first decouple the multi-sensor scheduling problem into a single-sensor constrained Markov decision process (CMDP) through relaxation of the hard bandwidth constraint. Next we exploit the threshold structure of the optimal policy for the decoupled single sensor CMDP and obtain the optimum solution through linear programming (LP). Finally, an asymptotically optimal truncated policy that can satisfy the hard bandwidth constraint is built upon the optimal solution to each of the decoupled single-sensor. Our investigation shows that to obtain a small AoI performance: (1) The scheduler exploits good channels to schedule sensors supported by limited power; (2) Sensors equipped with enough transmission power are updated in a timely manner such that the bandwidth constraint can be satisfied."
                    },
                    {
                        "title": "Scheduling to Minimize Age of Information in Multi-State Time-Varying Networks with Power Constraints",
                        "abstract": "In this paper, we study how to collect fresh data in time-varying networks with power constrained users. We measure data freshness from the perspective of the central controller by using the metric Age of Information, namely the time elapsed since the generation time-stamp of the freshest information. We wonder what is the minimum AoI performance the network can achieve and how to design scheduling algorithms to approach it. To answer these questions when scheduling decisions are restricted to bandwidth constraint, we first decouple the multi-user scheduling problem into a single user constrained Markov decision process (CMDP) through relaxation of the hard bandwidth constraint. Next we exploit the threshold structure of the optimal policy for the decoupled single user CMDP and obtain the optimum solution through linear programming (LP). Finally, an asymptotic optimal truncated policy that can satisfy the hard bandwidth constraint is built upon the optimal solution to each of the decoupled single-user sub-problem. The performance is verified through simulations. Our investigation shows that to obtain a small AoI performance, the scheduler exploits good channels to schedule users supported by limited power. Users equipped with enough transmission power are updated in a timely manner such that the bandwidth constraint can be satisfied."
                    },
                    {
                        "title": "Private Broadcasting: an Index Coding Approach",
                        "abstract": "Using a broadcast channel to transmit clients' data requests may impose privacy risks. In this paper, we address such privacy concerns in the index coding framework. We show how a malicious client can infer some information about the requests and side information of other clients by learning the encoding matrix used by the server. We propose an information-theoretic metric to measure the level of privacy and show how encoding matrices can be designed to achieve specific privacy guarantees. We then consider a special scenario for which we design a transmission scheme and derive the achieved levels of privacy in closed-form. We also derive upper bounds and we compare them to the levels of privacy achieved by our scheme, highlighting that an inherent trade-off exists between protecting privacy of the request and of the side information of the clients."
                    },
                    {
                        "title": "Privacy in Index Coding: Improved Bounds and Coding Schemes",
                        "abstract": "It was recently observed in [1], that in index coding, learning the coding matrix used by the server can pose privacy concerns: curious clients can extract information about the requests and side information of other clients. One approach to mitigate such concerns is the use of $k$-limited-access schemes [1], that restrict each client to learn only part of the index coding matrix, and in particular, at most $k$ rows. These schemes transform a linear index coding matrix of rank $T$ to an alternate one, such that each client needs to learn at most $k$ of the coding matrix rows to decode its requested message. This paper analyzes $k$-limited-access schemes. First, a worst-case scenario, where the total number of clients $n$ is $2^T-1$ is studied. For this case, a novel construction of the coding matrix is provided and shown to be order-optimal in the number of transmissions. Then, the case of a general $n$ is considered and two different schemes are designed and analytically and numerically assessed in their performance. It is shown that these schemes perform better than the one designed for the case $n=2^T-1$."
                    },
                    {
                        "title": "Distributed Thompson Sampling",
                        "abstract": "We study a cooperative multi-agent multi-armed bandits with M agents and K arms. The goal of the agents is to minimized the cumulative regret. We adapt a traditional Thompson Sampling algoirthm under the distributed setting. However, with agent's ability to communicate, we note that communication may further reduce the upper bound of the regret for a distributed Thompson Sampling approach. To further improve the performance of distributed Thompson Sampling, we propose a distributed Elimination based Thompson Sampling algorithm that allow the agents to learn collaboratively. We analyse the algorithm under Bernoulli reward and derived a problem dependent upper bound on the cumulative regret."
                    },
                    {
                        "title": "Preserving Privacy while Broadcasting: $k$-Limited-Access Schemes",
                        "abstract": "Index coding employs coding across clients within the same broadcast domain. This typically assumes that all clients learn the coding matrix so that they can decode and retrieve their requested data. However, learning the coding matrix can pose privacy concerns: it may enable clients to infer information about the requests and side information of other clients [1]. In this paper, we formalize the intuition that the achieved privacy can increase by decreasing the number of rows of the coding matrix that a client learns. Based on this, we propose the use of $k$-limited-access schemes: given an index coding scheme that employs $T$ transmissions, we create a $k$-limited-access scheme with $T_k\\geq T$ transmissions, and with the property that each client learns at most $k$ rows of the coding matrix to decode its message. We derive upper and lower bounds on $T_k$ for all values of $k$, and develop deterministic designs for these schemes for which $T_k$ has an order-optimal exponent for some regimes."
                    }
                ]
            },
            "32c12263-ef4e-40dc-afc6-698a08031299": {
                "pk": "32c12263-ef4e-40dc-afc6-698a08031299",
                "name": "Xiaodan Liang",
                "collaborators": [
                    "Liang Lin",
                    "Qingxing Cao",
                    "Eric Xing",
                    "Xiaojun Chang",
                    "Fengda Zhu",
                    "Ke Gong",
                    "Siyi Hu",
                    "Luona Yang",
                    "Tairui Wang",
                    "Jinghui Qin"
                ],
                "domain": [
                    "Computer Vision",
                    "Machine Learning",
                    "Reinforcement Learning",
                    "Semantic Segmentation"
                ],
                "publications": [
                    {
                        "title": "Complex Background Subtraction by Pursuing Dynamic Spatio-Temporal Models",
                        "abstract": "Although it has been widely discussed in video surveillance, background subtraction is still an open problem in the context of complex scenarios, e.g., dynamic backgrounds, illumination variations, and indistinct foreground objects. To address these challenges, we propose an effective background subtraction method by learning and maintaining an array of dynamic texture models within the spatio-temporal representations. At any location of the scene, we extract a sequence of regular video bricks, i.e. video volumes spanning over both spatial and temporal domain. The background modeling is thus posed as pursuing subspaces within the video bricks while adapting the scene variations. For each sequence of video bricks, we pursue the subspace by employing the ARMA (Auto Regressive Moving Average) Model that jointly characterizes the appearance consistency and temporal coherence of the observations. During online processing, we incrementally update the subspaces to cope with disturbances from foreground objects and scene changes. In the experiments, we validate the proposed method in several complex scenarios, and show superior performances over other state-of-the-art approaches of background subtraction. The empirical studies of parameter setting and component analysis are presented as well."
                    },
                    {
                        "title": "Recognizing Focal Liver Lesions in Contrast-Enhanced Ultrasound with Discriminatively Trained Spatio-Temporal Model",
                        "abstract": "The aim of this study is to provide an automatic computational framework to assist clinicians in diagnosing Focal Liver Lesions (FLLs) in Contrast-Enhancement Ultrasound (CEUS). We represent FLLs in a CEUS video clip as an ensemble of Region-of-Interests (ROIs), whose locations are modeled as latent variables in a discriminative model. Different types of FLLs are characterized by both spatial and temporal enhancement patterns of the ROIs. The model is learned by iteratively inferring the optimal ROI locations and optimizing the model parameters. To efficiently search the optimal spatial and temporal locations of the ROIs, we propose a data-driven inference algorithm by combining effective spatial and temporal pruning. The experiments show that our method achieves promising results on the largest dataset in the literature (to the best of our knowledge), which we have made publicly available."
                    },
                    {
                        "title": "Learning Latent Spatio-Temporal Compositional Model for Human Action Recognition",
                        "abstract": "Action recognition is an important problem in multimedia understanding. This paper addresses this problem by building an expressive compositional action model. We model one action instance in the video with an ensemble of spatio-temporal compositions: a number of discrete temporal anchor frames, each of which is further decomposed to a layout of deformable parts. In this way, our model can identify a Spatio-Temporal And-Or Graph (STAOG) to represent the latent structure of actions e.g. triple jumping, swinging and high jumping. The STAOG model comprises four layers: (i) a batch of leaf-nodes in bottom for detecting various action parts within video patches; (ii) the or-nodes over bottom, i.e. switch variables to activate their children leaf-nodes for structural variability; (iii) the and-nodes within an anchor frame for verifying spatial composition; and (iv) the root-node at top for aggregating scores over temporal anchor frames. Moreover, the contextual interactions are defined between leaf-nodes in both spatial and temporal domains. For model training, we develop a novel weakly supervised learning algorithm which iteratively determines the structural configuration (e.g. the production of leaf-nodes associated with the or-nodes) along with the optimization of multi-layer parameters. By fully exploiting spatio-temporal compositions and interactions, our approach handles well large intra-class action variance (e.g. different views, individual appearances, spatio-temporal structures). The experimental results on the challenging databases demonstrate superior performance of our approach over other competing methods."
                    },
                    {
                        "title": "Generative Semantic Manipulation with Contrasting GAN",
                        "abstract": "Generative Adversarial Networks (GANs) have recently achieved significant improvement on paired/unpaired image-to-image translation, such as photo$\\rightarrow$ sketch and artist painting style transfer. However, existing models can only be capable of transferring the low-level information (e.g. color or texture changes), but fail to edit high-level semantic meanings (e.g., geometric structure or content) of objects. On the other hand, while some researches can synthesize compelling real-world images given a class label or caption, they cannot condition on arbitrary shapes or structures, which largely limits their application scenarios and interpretive capability of model results. In this work, we focus on a more challenging semantic manipulation task, which aims to modify the semantic meaning of an object while preserving its own characteristics (e.g. viewpoints and shapes), such as cow$\\rightarrow$sheep, motor$\\rightarrow$ bicycle, cat$\\rightarrow$dog. To tackle such large semantic changes, we introduce a contrasting GAN (contrast-GAN) with a novel adversarial contrasting objective. Instead of directly making the synthesized samples close to target data as previous GANs did, our adversarial contrasting objective optimizes over the distance comparisons between samples, that is, enforcing the manipulated data be semantically closer to the real data with target category than the input data. Equipped with the new contrasting objective, a novel mask-conditional contrast-GAN architecture is proposed to enable disentangle image background with object semantic changes. Experiments on several semantic manipulation tasks on ImageNet and MSCOCO dataset show considerable performance gain by our contrast-GAN over other conditional GANs. Quantitative results further demonstrate the superiority of our model on generating manipulated results with high visual fidelity and reasonable object semantics."
                    },
                    {
                        "title": "Dynamic-structured Semantic Propagation Network",
                        "abstract": "Semantic concept hierarchy is still under-explored for semantic segmentation due to the inefficiency and complicated optimization of incorporating structural inference into dense prediction. This lack of modeling semantic correlations also makes prior works must tune highly-specified models for each task due to the label discrepancy across datasets. It severely limits the generalization capability of segmentation models for open set concept vocabulary and annotation utilization. In this paper, we propose a Dynamic-Structured Semantic Propagation Network (DSSPN) that builds a semantic neuron graph by explicitly incorporating the semantic concept hierarchy into network construction. Each neuron represents the instantiated module for recognizing a specific type of entity such as a super-class (e.g. food) or a specific concept (e.g. pizza). During training, DSSPN performs the dynamic-structured neuron computation graph by only activating a sub-graph of neurons for each image in a principled way. A dense semantic-enhanced neural block is proposed to propagate the learned knowledge of all ancestor neurons into each fine-grained child neuron for feature evolving. Another merit of such semantic explainable structure is the ability of learning a unified model concurrently on diverse datasets by selectively activating different neuron sub-graphs for each annotation at each step. Extensive experiments on four public semantic segmentation datasets (i.e. ADE20K, COCO-Stuff, Cityscape and Mapillary) demonstrate the superiority of our DSSPN over state-of-the-art segmentation models. Moreoever, we demonstrate a universal segmentation model that is jointly trained on diverse datasets can surpass the performance of the common fine-tuning scheme for exploiting multiple domain knowledge."
                    },
                    {
                        "title": "Explainable High-order Visual Question Reasoning: A New Benchmark and Knowledge-routed Network",
                        "abstract": "Explanation and high-order reasoning capabilities are crucial for real-world visual question answering with diverse levels of inference complexity (e.g., what is the dog that is near the girl playing with?) and important for users to understand and diagnose the trustworthiness of the system. Current VQA benchmarks on natural images with only an accuracy metric end up pushing the models to exploit the dataset biases and cannot provide any interpretable justification, which severally hinders advances in high-level question answering. In this work, we propose a new HVQR benchmark for evaluating explainable and high-order visual question reasoning ability with three distinguishable merits: 1) the questions often contain one or two relationship triplets, which requires the model to have the ability of multistep reasoning to predict plausible answers; 2) we provide an explicit evaluation on a multistep reasoning process that is constructed with image scene graphs and commonsense knowledge bases; and 3) each relationship triplet in a large-scale knowledge base only appears once among all questions, which poses challenges for existing networks that often attempt to overfit the knowledge base that already appears in the training set and enforces the models to handle unseen questions and knowledge fact usage. We also propose a new knowledge-routed modular network (KM-net) that incorporates the multistep reasoning process over a large knowledge base into visual question reasoning. An extensive dataset analysis and comparisons with existing models on the HVQR benchmark show that our benchmark provides explainable evaluations, comprehensive reasoning requirements and realistic challenges of VQA systems, as well as our KM-net's superiority in terms of accuracy and explanation ability."
                    },
                    {
                        "title": "Adaptive Temporal Encoding Network for Video Instance-level Human Parsing",
                        "abstract": "Beyond the existing single-person and multiple-person human parsing tasks in static images, this paper makes the first attempt to investigate a more realistic video instance-level human parsing that simultaneously segments out each person instance and parses each instance into more fine-grained parts (e.g., head, leg, dress). We introduce a novel Adaptive Temporal Encoding Network (ATEN) that alternatively performs temporal encoding among key frames and flow-guided feature propagation from other consecutive frames between two key frames. Specifically, ATEN first incorporates a Parsing-RCNN to produce the instance-level parsing result for each key frame, which integrates both the global human parsing and instance-level human segmentation into a unified model. To balance between accuracy and efficiency, the flow-guided feature propagation is used to directly parse consecutive frames according to their identified temporal consistency with key frames. On the other hand, ATEN leverages the convolution gated recurrent units (convGRU) to exploit temporal changes over a series of key frames, which are further used to facilitate the frame-level instance-level parsing. By alternatively performing direct feature propagation between consistent frames and temporal encoding network among key frames, our ATEN achieves a good balance between frame-level accuracy and time efficiency, which is a common crucial problem in video object segmentation research. To demonstrate the superiority of our ATEN, extensive experiments are conducted on the most popular video segmentation benchmark (DAVIS) and a newly collected Video Instance-level Parsing (VIP) dataset, which is the first video instance-level human parsing dataset comprised of 404 sequences and over 20k frames with instance-level and pixel-wise annotations."
                    },
                    {
                        "title": "Modern Augmented Reality: Applications, Trends, and Future Directions",
                        "abstract": "Augmented reality (AR) is one of the relatively old, yet trending areas in the intersection of computer vision and computer graphics with numerous applications in several areas, from gaming and entertainment, to education and healthcare. Although it has been around for nearly fifty years, it has seen a lot of interest by the research community in the recent years, mainly because of the huge success of deep learning models for various computer vision and AR applications, which made creating new generations of AR technologies possible. This work tries to provide an overview of modern augmented reality, from both application-level and technical perspective. We first give an overview of main AR applications, grouped into more than ten categories. We then give an overview of around 100 recent promising machine learning based works developed for AR systems, such as deep learning works for AR shopping (clothing, makeup), AR based image filters (such as Snapchat's lenses), AR animations, and more. In the end we discuss about some of the current challenges in AR domain, and the future directions in this area."
                    },
                    {
                        "title": "Policy Diagnosis via Measuring Role Diversity in Cooperative Multi-agent RL",
                        "abstract": "Cooperative multi-agent reinforcement learning (MARL) is making rapid progress for solving tasks in a grid world and real-world scenarios, in which agents are given different attributes and goals, resulting in different behavior through the whole multi-agent task. In this study, we quantify the agent's behavior difference and build its relationship with the policy performance via {\\bf Role Diversity}, a metric to measure the characteristics of MARL tasks. We define role diversity from three perspectives: action-based, trajectory-based, and contribution-based to fully measure a multi-agent task. Through theoretical analysis, we find that the error bound in MARL can be decomposed into three parts that have a strong relation to the role diversity. The decomposed factors can significantly impact policy optimization on three popular directions including parameter sharing, communication mechanism, and credit assignment. The main experimental platforms are based on {\\bf Multiagent Particle Environment (MPE)} and {\\bf The StarCraft Multi-Agent Challenge (SMAC). Extensive experiments} clearly show that role diversity can serve as a robust measurement for the characteristics of a multi-agent cooperation task and help diagnose whether the policy fits the current multi-agent system for a better policy performance."
                    },
                    {
                        "title": "Real-to-Virtual Domain Unification for End-to-End Autonomous Driving",
                        "abstract": "In the spectrum of vision-based autonomous driving, vanilla end-to-end models are not interpretable and suboptimal in performance, while mediated perception models require additional intermediate representations such as segmentation masks or detection bounding boxes, whose annotation can be prohibitively expensive as we move to a larger scale. More critically, all prior works fail to deal with the notorious domain shift if we were to merge data collected from different sources, which greatly hinders the model generalization ability. In this work, we address the above limitations by taking advantage of virtual data collected from driving simulators, and present DU-drive, an unsupervised real-to-virtual domain unification framework for end-to-end autonomous driving. It first transforms real driving data to its less complex counterpart in the virtual domain and then predicts vehicle control commands from the generated virtual image. Our framework has three unique advantages: 1) it maps driving data collected from a variety of source distributions into a unified domain, effectively eliminating domain shift; 2) the learned virtual representation is simpler than the input real image and closer in form to the \"minimum sufficient statistic\" for the prediction task, which relieves the burden of the compression phase while optimizing the information bottleneck tradeoff and leads to superior prediction performance; 3) it takes advantage of annotated virtual data which is unlimited and free to obtain. Extensive experiments on two public driving datasets and two driving simulators demonstrate the performance superiority and interpretive capability of DU-drive."
                    },
                    {
                        "title": "Dynamic Knowledge Routing Network For Target-Guided Open-Domain Conversation",
                        "abstract": "Target-guided open-domain conversation aims to proactively and naturally guide a dialogue agent or human to achieve specific goals, topics or keywords during open-ended conversations. Existing methods mainly rely on single-turn datadriven learning and simple target-guided strategy without considering semantic or factual knowledge relations among candidate topics/keywords. This results in poor transition smoothness and low success rate. In this work, we adopt a structured approach that controls the intended content of system responses by introducing coarse-grained keywords, attains smooth conversation transition through turn-level supervised learning and knowledge relations between candidate keywords, and drives an conversation towards an specified target with discourse-level guiding strategy. Specially, we propose a novel dynamic knowledge routing network (DKRN) which considers semantic knowledge relations among candidate keywords for accurate next topic prediction of next discourse. With the help of more accurate keyword prediction, our keyword-augmented response retrieval module can achieve better retrieval performance and more meaningful conversations. Besides, we also propose a novel dual discourse-level target-guided strategy to guide conversations to reach their goals smoothly with higher success rate. Furthermore, to push the research boundary of target-guided open-domain conversation to match real-world scenarios better, we introduce a new large-scale Chinese target-guided open-domain conversation dataset (more than 900K conversations) crawled from Sina Weibo. Quantitative and human evaluations show our method can produce meaningful and effective target-guided conversations, significantly improving over other state-of-the-art methods by more than 20% in success rate and more than 0.6 in average smoothness score."
                    },
                    {
                        "title": "CIRL: Controllable Imitative Reinforcement Learning for Vision-based Self-driving",
                        "abstract": "Autonomous urban driving navigation with complex multi-agent dynamics is under-explored due to the difficulty of learning an optimal driving policy. The traditional modular pipeline heavily relies on hand-designed rules and the pre-processing perception system while the supervised learning-based models are limited by the accessibility of extensive human experience. We present a general and principled Controllable Imitative Reinforcement Learning (CIRL) approach which successfully makes the driving agent achieve higher success rates based on only vision inputs in a high-fidelity car simulator. To alleviate the low exploration efficiency for large continuous action space that often prohibits the use of classical RL on challenging real tasks, our CIRL explores over a reasonably constrained action space guided by encoded experiences that imitate human demonstrations, building upon Deep Deterministic Policy Gradient (DDPG). Moreover, we propose to specialize adaptive policies and steering-angle reward designs for different control signals (i.e. follow, straight, turn right, turn left) based on the shared representations to improve the model capability in tackling with diverse cases. Extensive experiments on CARLA driving benchmark demonstrate that CIRL substantially outperforms all previous methods in terms of the percentage of successfully completed episodes on a variety of goal-directed driving tasks. We also show its superior generalization capability in unseen environments. To our knowledge, this is the first successful case of the learned driving policy through reinforcement learning in the high-fidelity simulator, which performs better-than supervised imitation learning."
                    },
                    {
                        "title": "Vision-Language Navigation with Self-Supervised Auxiliary Reasoning Tasks",
                        "abstract": "Vision-Language Navigation (VLN) is a task where agents learn to navigate following natural language instructions. The key to this task is to perceive both the visual scene and natural language sequentially. Conventional approaches exploit the vision and language features in cross-modal grounding. However, the VLN task remains challenging, since previous works have neglected the rich semantic information contained in the environment (such as implicit navigation graphs or sub-trajectory semantics). In this paper, we introduce Auxiliary Reasoning Navigation (AuxRN), a framework with four self-supervised auxiliary reasoning tasks to take advantage of the additional training signals derived from the semantic information. The auxiliary tasks have four reasoning objectives: explaining the previous actions, estimating the navigation progress, predicting the next orientation, and evaluating the trajectory consistency. As a result, these additional training signals help the agent to acquire knowledge of semantic representations in order to reason about its activity and build a thorough perception of the environment. Our experiments indicate that auxiliary reasoning tasks improve both the performance of the main task and the model generalizability by a large margin. Empirically, we demonstrate that an agent trained with self-supervised auxiliary reasoning tasks substantially outperforms the previous state-of-the-art method, being the best existing approach on the standard benchmark."
                    },
                    {
                        "title": "UPDeT: Universal Multi-agent Reinforcement Learning via Policy Decoupling with Transformers",
                        "abstract": "Recent advances in multi-agent reinforcement learning have been largely limited in training one model from scratch for every new task. The limitation is due to the restricted model architecture related to fixed input and output dimensions. This hinders the experience accumulation and transfer of the learned agent over tasks with diverse levels of difficulty (e.g. 3 vs 3 or 5 vs 6 multi-agent games). In this paper, we make the first attempt to explore a universal multi-agent reinforcement learning pipeline, designing one single architecture to fit tasks with the requirement of different observation and action configurations. Unlike previous RNN-based models, we utilize a transformer-based model to generate a flexible policy by decoupling the policy distribution from the intertwined input observation with an importance weight measured by the merits of the self-attention mechanism. Compared to a standard transformer block, the proposed model, named as Universal Policy Decoupling Transformer (UPDeT), further relaxes the action restriction and makes the multi-agent task's decision process more explainable. UPDeT is general enough to be plugged into any multi-agent reinforcement learning pipeline and equip them with strong generalization abilities that enables the handling of multiple tasks at a time. Extensive experiments on large-scale SMAC multi-agent competitive games demonstrate that the proposed UPDeT-based multi-agent reinforcement learning achieves significant results relative to state-of-the-art approaches, demonstrating advantageous transfer capability in terms of both performance and training speed (10 times faster)."
                    },
                    {
                        "title": "Look into Person: Joint Body Parsing & Pose Estimation Network and A New Benchmark",
                        "abstract": "Human parsing and pose estimation have recently received considerable interest due to their substantial application potentials. However, the existing datasets have limited numbers of images and annotations and lack a variety of human appearances and coverage of challenging cases in unconstrained environments. In this paper, we introduce a new benchmark named \"Look into Person (LIP)\" that provides a significant advancement in terms of scalability, diversity, and difficulty, which are crucial for future developments in human-centric analysis. This comprehensive dataset contains over 50,000 elaborately annotated images with 19 semantic part labels and 16 body joints, which are captured from a broad range of viewpoints, occlusions, and background complexities. Using these rich annotations, we perform detailed analyses of the leading human parsing and pose estimation approaches, thereby obtaining insights into the successes and failures of these methods. To further explore and take advantage of the semantic correlation of these two tasks, we propose a novel joint human parsing and pose estimation network to explore efficient context modeling, which can simultaneously predict parsing and pose with extremely high quality. Furthermore, we simplify the network to solve human parsing by exploring a novel self-supervised structure-sensitive learning approach, which imposes human pose structures into the parsing results without resorting to extra supervision. The dataset, code and models are available at http://www.sysu-hcp.net/lip/."
                    },
                    {
                        "title": "Linguistically Driven Graph Capsule Network for Visual Question Reasoning",
                        "abstract": "Recently, studies of visual question answering have explored various architectures of end-to-end networks and achieved promising results on both natural and synthetic datasets, which require explicitly compositional reasoning. However, it has been argued that these black-box approaches lack interpretability of results, and thus cannot perform well on generalization tasks due to overfitting the dataset bias. In this work, we aim to combine the benefits of both sides and overcome their limitations to achieve an end-to-end interpretable structural reasoning for general images without the requirement of layout annotations. Inspired by the property of a capsule network that can carve a tree structure inside a regular convolutional neural network (CNN), we propose a hierarchical compositional reasoning model called the \"Linguistically driven Graph Capsule Network\", where the compositional process is guided by the linguistic parse tree. Specifically, we bind each capsule in the lowest layer to bridge the linguistic embedding of a single word in the original question with visual evidence and then route them to the same capsule if they are siblings in the parse tree. This compositional process is achieved by performing inference on a linguistically driven conditional random field (CRF) and is performed across multiple graph capsule layers, which results in a compositional reasoning process inside a CNN. Experiments on the CLEVR dataset, CLEVR compositional generation test, and FigureQA dataset demonstrate the effectiveness and composition generalization ability of our end-to-end model."
                    },
                    {
                        "title": "Visual-Language Navigation Pretraining via Prompt-based Environmental Self-exploration",
                        "abstract": "Vision-language navigation (VLN) is a challenging task due to its large searching space in the environment. To address this problem, previous works have proposed some methods of fine-tuning a large model that pretrained on large-scale datasets. However, the conventional fine-tuning methods require extra human-labeled navigation data and lack self-exploration capabilities in environments, which hinders their generalization of unseen scenes. To improve the ability of fast cross-domain adaptation, we propose Prompt-based Environmental Self-exploration (ProbES), which can self-explore the environments by sampling trajectories and automatically generates structured instructions via a large-scale cross-modal pretrained model (CLIP). Our method fully utilizes the knowledge learned from CLIP to build an in-domain dataset by self-exploration without human labeling. Unlike the conventional approach of fine-tuning, we introduce prompt-based learning to achieve fast adaptation for language embeddings, which substantially improves the learning efficiency by leveraging prior knowledge. By automatically synthesizing trajectory-instruction pairs in any environment without human supervision and efficient prompt-based learning, our model can adapt to diverse vision-language navigation tasks, including VLN and REVERIE. Both qualitative and quantitative results show that our ProbES significantly improves the generalization ability of the navigation model."
                    },
                    {
                        "title": "Geometric Scene Parsing with Hierarchical LSTM",
                        "abstract": "This paper addresses the problem of geometric scene parsing, i.e. simultaneously labeling geometric surfaces (e.g. sky, ground and vertical plane) and determining the interaction relations (e.g. layering, supporting, siding and affinity) between main regions. This problem is more challenging than the traditional semantic scene labeling, as recovering geometric structures necessarily requires the rich and diverse contextual information. To achieve these goals, we propose a novel recurrent neural network model, named Hierarchical Long Short-Term Memory (H-LSTM). It contains two coupled sub-networks: the Pixel LSTM (P-LSTM) and the Multi-scale Super-pixel LSTM (MS-LSTM) for handling the surface labeling and relation prediction, respectively. The two sub-networks provide complementary information to each other to exploit hierarchical scene contexts, and they are jointly optimized for boosting the performance. Our extensive experiments show that our model is capable of parsing scene geometric structures and outperforming several state-of-the-art methods by large margins. In addition, we show promising 3D reconstruction results from the still images based on the geometric parsing."
                    },
                    {
                        "title": "GRADE: Automatic Graph-Enhanced Coherence Metric for Evaluating Open-Domain Dialogue Systems",
                        "abstract": "Automatically evaluating dialogue coherence is a challenging but high-demand ability for developing high-quality open-domain dialogue systems. However, current evaluation metrics consider only surface features or utterance-level semantics, without explicitly considering the fine-grained topic transition dynamics of dialogue flows. Here, we first consider that the graph structure constituted with topics in a dialogue can accurately depict the underlying communication logic, which is a more natural way to produce persuasive metrics. Capitalized on the topic-level dialogue graph, we propose a new evaluation metric GRADE, which stands for Graph-enhanced Representations for Automatic Dialogue Evaluation. Specifically, GRADE incorporates both coarse-grained utterance-level contextualized representations and fine-grained topic-level graph representations to evaluate dialogue coherence. The graph representations are obtained by reasoning over topic-level dialogue graphs enhanced with the evidence from a commonsense graph, including k-hop neighboring representations and hop-attention weights. Experimental results show that our GRADE significantly outperforms other state-of-the-art metrics on measuring diverse dialogue models in terms of the Pearson and Spearman correlations with human judgements. Besides, we release a new large-scale human evaluation benchmark to facilitate future research on automatic metrics."
                    }
                ]
            }
        }
    },
    "2410.03813": {
        "paper_data": {
            "title": "SOI: Scaling Down Computational Complexity by Estimating Partial States of the Model",
            "url": "http://arxiv.org/abs/2410.03813v1",
            "arxiv_id": "2410.03813",
            "authors": [
                "Grzegorz Stefa\u0144ski",
                "Pawe\u0142 Daniluk",
                "Artur Szumaczuk",
                "Jakub Tkaczuk"
            ],
            "abstract": "Consumer electronics used to follow the miniaturization trend described by Moore's Law. Despite increased processing power in Microcontroller Units (MCUs), MCUs used in the smallest appliances are still not capable of running even moderately big, state-of-the-art artificial neural networks (ANNs) especially in time-sensitive scenarios. In this work, we present a novel method called Scattered Online Inference (SOI) that aims to reduce the computational complexity of ANNs. SOI leverages the continuity and seasonality of time-series data and model predictions, enabling extrapolation for processing speed improvements, particularly in deeper layers. By applying compression, SOI generates more general inner partial states of ANN, allowing skipping full model recalculation at each inference.",
            "introduction": "   1 Introduction  Moore\u2019s Law (Moore, 1998) predicts that the number of transistors on a microchip will double every 2 years. As the number of transistors increases, the computational capacity of systems-on-chip (SoCs) also grows. Accompanied by the further miniaturization of components, the increased complexity of SoCs allows for the production of more compact appliances with equal or higher capabilities.   Despite the continuous development of more advanced hardware, the smallest appliances still cannot fully benefit from increasingly popular neural-based solutions, as they are not able to run them efficiently. Examples of such appliances include wireless in-ear headphones, smart watches, and AR glasses. Furthermore, the size of state-of-the-art neural networks is growing at much faster rate than that described by Moore\u2019s Law (Xu et\u00a0al., 2018). Adding to the challenge, according to researchers (Leiserson et\u00a0al., 2020), Moore\u2019s Law is starting to decelerate due to the physical constraints of semiconductors, and the \u201croom at the bottom\u201d is depleting.   Currently available neural network technologies enable machines to outperform humans in numerous applications in terms of measured performance (Nguyen et\u00a0al., 2020; Bakhtin et\u00a0al., 2022; Schrittwieser et\u00a0al., 2020). However, despite this achievement, the same models fall significantly short when it comes to energy efficiency compared to humans. The human brain consumes a mere 20 Watts of power (Laughlin et\u00a0al., 1998; Sengupta & Stemmler, 2014; Balasubramanian, 2021) and according to Xu et\u00a0al. (2018) it is estimated to be over five orders of magnitude more energy efficient than modern neural networks.   This disparity in energy efficiency can be attributed to the common pursuit of the highest model quality in the literature, showcasing the full capabilities of the developed technology, often at the expense of efficiency. This behavior is justifiable due to the ease of comparing different solutions using well-defined metrics that are independent of the hardware and software. Conversely, estimating a model\u2019s energy efficiency is a more complex task, influenced by various factors including the running environment.   However, this trend within our community may prove restrictive, particularly for applications like real-time systems. These applications naturally demand optimal performance alongside high-efficiency solutions, rendering most current Deep Neural Networks (DNNs) impractical for such tasks. Furthermore, due to the substantial discrepancy between assumptions for high-efficiency on-device solutions and high-performing monolithic models, compressing these models may not be a viable means of applying state-of-the-art DNNs to real-time tasks.   The importance of neural system efficiency is also increasingly significant from ecological and economic standpoints (Lacoste et\u00a0al., 2019; Patterson et\u00a0al., 2021).    1.1 Related Works  The Short-Term Memory Convolution (STMC) (Stefa\u0144ski et\u00a0al., 2023) was devised to enhance the efficiency of Convolutional Neural Networks (CNNs) inference by reducing computation requirements at each step via eliminating the need to recompute prior states. The authors achieved a notable 5.68-fold reduction in inference time and a 30% decrease in memory consumption compared to the Incremental Inference method (Romaniuk et\u00a0al., 2020). STMC enables the conversion of a standard CNN model, which typically requires an input of size at least as large as its receptive field, into a model capable of processing a single input frame at a time, akin to Long-Short Term Memory networks (LSTM). The SOI method is built",
            "references": [
                {
                    "title": "Short-Term Memory Convolutions",
                    "abstract": "The real-time processing of time series signals is a critical issue for many real-life applications. The idea of real-time processing is especially important in audio domain as the human perception of sound is sensitive to any kind of disturbance in perceived signals, especially the lag between auditory and visual modalities. The rise of deep learning (DL) models complicated the landscape of signal processing. Although they often have superior quality compared to standard DSP methods, this advantage is diminished by higher latency. In this work we propose novel method for minimization of inference time latency and memory consumption, called Short-Term Memory Convolution (STMC) and its transposed counterpart. The main advantage of STMC is the low latency comparable to long short-term memory (LSTM) networks. Furthermore, the training of STMC-based models is faster and more stable as the method is based solely on convolutional neural networks (CNNs). In this study we demonstrate an application of this solution to a U-Net model for a speech separation task and GhostNet model in acoustic scene classification (ASC) task. In case of speech separation we achieved a 5-fold reduction in inference time and a 2-fold reduction in latency without affecting the output quality. The inference time for ASC task was up to 4 times faster while preserving the original accuracy."
                },
                {
                    "title": "Human-level play in the game of Diplomacy by combining language models with strategic reasoning",
                    "abstract": "Despite much progress in training artificial intelligence (AI) systems to imitate human language, building agents that use language to communicate intentionally with humans in interactive environments remains a major challenge. We introduce Cicero, the first AI agent to achieve human-level performance in Diplomacy, a strategy game involving both cooperation and competition that emphasizes natural language negotiation and tactical coordination between seven players. Cicero integrates a language model with planning and reinforcement learning algorithms by inferring players\u2019 beliefs and intentions from its conversations and generating dialogue in pursuit of its plans. Across 40 games of an anonymous online Diplomacy league, Cicero achieved more than double the average score of the human players and ranked in the top 10% of participants who played more than one game. Description AI masters Diplomacy The game Diplomacy has been a major challenge for artificial intelligence (AI). Unlike other competitive games that AI has recently mastered, such as chess, Go, and poker, Diplomacy cannot be solved purely through self-play; it requires the development of an agent to understand other players\u2019 motivations and perspectives and to use natural language to negotiate complex shared plans. The Meta Fundamental AI Research Diplomacy Team (FAIR) et al. developed an agent that is able to play the full natural language form of the game and demonstrates performance well above the human average in an online Diplomacy league. The present work has far-reaching implications for the development of cooperative AI and language models for communication with people, even when interactions involve a mixture of aligned and competing interests. \u2014YS Artificial intelligence demonstrates human-level performance in the strategic board game Diplomacy."
                },
                {
                    "title": "Quantization",
                    "abstract": "The basic goal of quantization for probability distribution is to reduce the number of values, which is typically uncountable, describing a probability distribution to some finite set and thus to make an approximation of a continuous probability distribution by a discrete distribution. It has broad application in signal processing and data compression. In this paper, first we define the uniform distributions on different curves such as a line segment, a circle, and the boundary of an equilateral triangle. Then, we give the exact formulas to determine the optimal sets of n-means and the nth quantization errors for different values of n with respect to the uniform distributions defined on the curves. In each case, we further calculate the quantization dimension and show that it is equal to the dimension of the object; and the quantization coefficient exists as a finite positive number. This supports the well-known result of Bucklew and Wise (1982), which says that for a Borel probability measure P with non-vanishing absolutely continuous part the quantization coefficient exists as a finite positive number."
                },
                {
                    "title": "ResNet10: A lightweight residual network for remote sensing image classification",
                    "abstract": "Remote sensing image classification is challenging due to its complex background and various types. Currently, the commonly used lightweight neural network still has the problem of excessive calculation and parameter amount, which makes cheap mobile devices with low computing power unable to quickly complete remote sensing image classification tasks. In response to this problem, we propose a lightweight residual neural network-ResNet10 that is more suitable for use on cheap mobile devices with low computing power. Experiments based on the UCM Land-Use public data set show that the classification accuracy of ResNet10 can reach 96.2%, the size of the model is one-tenth that of ResNetl8, and it can classify 105 remote sensing images per second. ResNet10 makes it possible for remote sensing image classification to be used on mobile devices."
                },
                {
                    "title": "Carbon Emissions and Large Neural Network Training",
                    "abstract": "The computation demand for machine learning (ML) has grown rapidly recently, which comes with a number of costs. Estimating the energy cost helps measure its environmental impact and finding greener strategies, yet it is challenging without detailed information. We calculate the energy use and carbon footprint of several recent large models-T5, Meena, GShard, Switch Transformer, and GPT-3-and refine earlier estimates for the neural architecture search that found Evolved Transformer. We highlight the following opportunities to improve energy efficiency and CO2 equivalent emissions (CO2e): Large but sparsely activated DNNs can consume<1/10th the energy of large, dense DNNs without sacrificing accuracy despite using as many or even more parameters. Geographic location matters for ML workload scheduling since the fraction of carbon-free energy and resulting CO2e vary ~5X-10X, even within the same country and the same organization. We are now optimizing where and when large models are trained. Specific datacenter infrastructure matters, as Cloud datacenters can be ~1.4-2X more energy efficient than typical datacenters, and the ML-oriented accelerators inside them can be ~2-5X more effective than off-the-shelf systems. Remarkably, the choice of DNN, datacenter, and processor can reduce the carbon footprint up to ~100-1000X. These large factors also make retroactive estimates of energy cost difficult. To avoid miscalculations, we believe ML papers requiring large computational resources should make energy consumption and CO2e explicit when practical. We are working to be more transparent about energy use and CO2e in our future research. To help reduce the carbon footprint of ML, we believe energy usage and CO2e should be a key metric in evaluating models, and we are collaborating with MLPerf developers to include energy usage during training and inference in this industry standard benchmark."
                },
                {
                    "title": "Super-Human Performance in Online Low-latency Recognition of Conversational Speech",
                    "abstract": "Achieving super-human performance in recognizing human speech has been a goal for several decades, as researchers have worked on increasingly challenging tasks. In the 1990's it was discovered, that conversational speech between two humans turns out to be considerably more difficult than read speech as hesitations, disfluencies, false starts and sloppy articulation complicate acoustic processing and require robust handling of acoustic, lexical and language context, jointly. Early attempts with statistical models could only reach error rates over 50% and far from human performance (WER of around 5.5%). Neural hybrid models and recent attention-based encoder-decoder models have considerably improved performance as such contexts can now be learned in an integral fashion. However, processing such contexts requires an entire utterance presentation and thus introduces unwanted delays before a recognition result can be output. In this paper, we address performance as well as latency. We present results for a system that can achieve super-human performance (at a WER of 5.0%, over the Switchboard conversational benchmark) at a word based latency of only 1 second behind a speaker's speech. The system uses multiple attention-based encoder-decoder networks integrated within a novel low latency incremental inference approach."
                },
                {
                    "title": "Efficient Low-Latency Speech Enhancement with Mobile Audio Streaming Networks",
                    "abstract": "We propose Mobile Audio Streaming Networks (MASnet) for efficient low-latency speech enhancement, which is particularly suitable for mobile devices and other applications where computational capacity is a limitation. MASnet processes linear-scale spectrograms, transforming successive noisy frames into complex-valued ratio masks which are then applied to the respective noisy frames. MASnet can operate in a low-latency incremental inference mode which matches the complexity of layer-by-layer batch mode. Compared to a similar fully-convolutional architecture, MASnet incorporates depthwise and pointwise convolutions for a large reduction in fused multiply-accumulate operations per second (FMA/s), at the cost of some reduction in SNR."
                },
                {
                    "title": "There\u2019s plenty of room at the Top: What will drive computer performance after Moore\u2019s law?",
                    "abstract": "From bottom to top The doubling of the number of transistors on a chip every 2 years, a seemly inevitable trend that has been called Moore's law, has contributed immensely to improvements in computer performance. However, silicon-based transistors cannot get much smaller than they are today, and other approaches should be explored to keep performance growing. Leiserson et al. review recent examples and argue that the most promising place to look is at the top of the computing stack, where improvements in software, algorithms, and hardware architecture can bring the much-needed boost. Science, this issue p. eaam9744 BACKGROUND Improvements in computing power can claim a large share of the credit for many of the things that we take for granted in our modern lives: cellphones that are more powerful than room-sized computers from 25 years ago, internet access for nearly half the world, and drug discoveries enabled by powerful supercomputers. Society has come to rely on computers whose performance increases exponentially over time. Much of the improvement in computer performance comes from decades of miniaturization of computer components, a trend that was foreseen by the Nobel Prize\u2013winning physicist Richard Feynman in his 1959 address, \u201cThere\u2019s Plenty of Room at the Bottom,\u201d to the American Physical Society. In 1975, Intel founder Gordon Moore predicted the regularity of this miniaturization trend, now called Moore\u2019s law, which, until recently, doubled the number of transistors on computer chips every 2 years. Unfortunately, semiconductor miniaturization is running out of steam as a viable way to grow computer performance\u2014there isn\u2019t much more room at the \u201cBottom.\u201d If growth in computing power stalls, practically all industries will face challenges to their productivity. Nevertheless, opportunities for growth in computing performance will still be available, especially at the \u201cTop\u201d of the computing-technology stack: software, algorithms, and hardware architecture. ADVANCES Software can be made more efficient by performance engineering: restructuring software to make it run faster. Performance engineering can remove inefficiencies in programs, known as software bloat, arising from traditional software-development strategies that aim to minimize an application\u2019s development time rather than the time it takes to run. Performance engineering can also tailor software to the hardware on which it runs, for example, to take advantage of parallel processors and vector units. Algorithms offer more-efficient ways to solve problems. Indeed, since the late 1970s, the time to solve the maximum-flow problem improved nearly as much from algorithmic advances as from hardware speedups. But progress on a given algorithmic problem occurs unevenly and sporadically and must ultimately face diminishing returns. As such, we see the biggest benefits coming from algorithms for new problem domains (e.g., machine learning) and from developing new theoretical machine models that better reflect emerging hardware. Hardware architectures can be streamlined\u2014for instance, through processor simplification, where a complex processing core is replaced with a simpler core that requires fewer transistors. The freed-up transistor budget can then be redeployed in other ways\u2014for example, by increasing the number of processor cores running in parallel, which can lead to large efficiency gains for problems that can exploit parallelism. Another form of streamlining is domain specialization, where hardware is customized for a particular application domain. This type of specialization jettisons processor functionality that is not needed for the domain. It can also allow more customization to the specific characteristics of the domain, for instance, by decreasing floating-point precision for machine-learning applications. In the post-Moore era, performance improvements from software, algorithms, and hardware architecture will increasingly require concurrent changes across other levels of the stack. These changes will be easier to implement, from engineering-management and economic points of view, if they occur within big system components: reusable software with typically more than a million lines of code or hardware of comparable complexity. When a single organization or company controls a big component, modularity can be more easily reengineered to obtain performance gains. Moreover, costs and benefits can be pooled so that important but costly changes in one part of the big component can be justified by benefits elsewhere in the same component. OUTLOOK As miniaturization wanes, the silicon-fabrication improvements at the Bottom will no longer provide the predictable, broad-based gains in computer performance that society has enjoyed for more than 50 years. Software performance engineering, development of algorithms, and hardware streamlining at the Top can continue to make computer applications faster in the post-Moore era. Unlike the historical gains at the Bottom, however, gains at the Top will be opportunistic, uneven, and sporadic. Moreover, they will be subject to diminishing returns as specific computations become better explored. Performance gains after Moore\u2019s law ends. In the post-Moore era, improvements in computing power will increasingly come from technologies at the \u201cTop\u201d of the computing stack, not from those at the \u201cBottom\u201d, reversing the historical trend. CREDIT: N. CARY/SCIENCE The miniaturization of semiconductor transistors has driven the growth in computer performance for more than 50 years. As miniaturization approaches its limits, bringing an end to Moore\u2019s law, performance gains will need to come from software, algorithms, and hardware. We refer to these technologies as the \u201cTop\u201d of the computing stack to distinguish them from the traditional technologies at the \u201cBottom\u201d: semiconductor physics and silicon-fabrication technology. In the post-Moore era, the Top will provide substantial performance gains, but these gains will be opportunistic, uneven, and sporadic, and they will suffer from the law of diminishing returns. Big system components offer a promising context for tackling the challenges of working at the Top."
                },
                {
                    "title": "The INTERSPEECH 2020 Deep Noise Suppression Challenge: Datasets, Subjective Testing Framework, and Challenge Results",
                    "abstract": "The INTERSPEECH 2020 Deep Noise Suppression (DNS) Challenge is intended to promote collaborative research in real-time single-channel Speech Enhancement aimed to maximize the subjective (perceptual) quality of the enhanced speech. A typical approach to evaluate the noise suppression methods is to use objective metrics on the test set obtained by splitting the original dataset. While the performance is good on the synthetic test set, often the model performance degrades significantly on real recordings. Also, most of the conventional objective metrics do not correlate well with subjective tests and lab subjective tests are not scalable for a large test set. In this challenge, we open-sourced a large clean speech and noise corpus for training the noise suppression models and a representative test set to real-world scenarios consisting of both synthetic and real recordings. We also open-sourced an online subjective test framework based on ITU-T P.808 for researchers to reliably test their developments. We evaluated the results using P.808 on a blind test set. The results and the key learnings from the challenge are discussed. The datasets and scripts can be found here for quick access this https URL."
                },
                {
                    "title": "GhostNet: More Features From Cheap Operations",
                    "abstract": "Deploying convolutional neural networks (CNNs) on embedded devices is difficult due to the limited memory and computation resources. The redundancy in feature maps is an important characteristic of those successful CNNs, but has rarely been investigated in neural architecture design. This paper proposes a novel Ghost module to generate more feature maps from cheap operations. Based on a set of intrinsic feature maps, we apply a series of linear transformations with cheap cost to generate many ghost feature maps that could fully reveal information underlying intrinsic features. The proposed Ghost module can be taken as a plug-and-play component to upgrade existing convolutional neural networks. Ghost bottlenecks are designed to stack Ghost modules, and then the lightweight GhostNet can be easily established. Experiments conducted on benchmarks demonstrate that the proposed Ghost module is an impressive alternative of convolution layers in baseline models, and our GhostNet can achieve higher recognition performance (e.g. 75.7% top-1 accuracy) than MobileNetV3 with similar computational cost on the ImageNet ILSVRC-2012 classification dataset. Code is available at https://github.com/huawei-noah/ghostnet."
                },
                {
                    "title": "Quantifying the Carbon Emissions of Machine Learning",
                    "abstract": "From an environmental standpoint, there are a few crucial aspects of training a neural network that have a major impact on the quantity of carbon that it emits. These factors include: the location of the server used for training and the energy grid that it uses, the length of the training procedure, and even the make and model of hardware on which the training takes place. In order to approximate these emissions, we present our Machine Learning Emissions Calculator, a tool for our community to better understand the environmental impact of training ML models. We accompany this tool with an explanation of the factors cited above, as well as concrete actions that individual practitioners and organizations can take to mitigate their carbon emissions."
                },
                {
                    "title": "Neural Speed Reading with Structural-Jump-LSTM",
                    "abstract": "Recurrent neural networks (RNNs) can model natural language by sequentially 'reading' input tokens and outputting a distributed representation of each token. Due to the sequential nature of RNNs, inference time is linearly dependent on the input length, and all inputs are read regardless of their importance. Efforts to speed up this inference, known as 'neural speed reading', either ignore or skim over part of the input. We present Structural-Jump-LSTM: the first neural speed reading model to both skip and jump text during inference. The model consists of a standard LSTM and two agents: one capable of skipping single words when reading, and one capable of exploiting punctuation structure (sub-sentence separators (,:), sentence end symbols (.!?), or end of text markers) to jump ahead after reading a word. A comprehensive experimental evaluation of our model against all five state-of-the-art neural reading models shows that Structural-Jump-LSTM achieves the best overall floating point operations (FLOP) reduction (hence is faster), while keeping the same accuracy or even improving it compared to a vanilla LSTM that reads the whole text."
                },
                {
                    "title": "Neural Speed Reading via Skim-RNN",
                    "abstract": "Inspired by the principles of speed reading, we introduce Skim-RNN, a recurrent neural network (RNN) that dynamically decides to update only a small fraction of the hidden state for relatively unimportant input tokens. Skim-RNN gives computational advantage over an RNN that always updates the entire hidden state. Skim-RNN uses the same input and output interfaces as a standard RNN and can be easily used instead of RNNs in existing models. In our experiments, we show that Skim-RNN can achieve significantly reduced computational cost without losing accuracy compared to standard RNNs across five different natural language tasks. In addition, we demonstrate that the trade-off between accuracy and speed of Skim-RNN can be dynamically controlled during inference time in a stable manner. Our analysis also shows that Skim-RNN running on a single CPU offers lower latency compared to standard RNNs on GPUs."
                },
                {
                    "title": "Skip RNN: Learning to Skip State Updates in Recurrent Neural Networks",
                    "abstract": "Recurrent Neural Networks (RNNs) continue to show outstanding performance in sequence modeling tasks. However, training RNNs on long sequences often face challenges like slow inference, vanishing gradients and difficulty in capturing long term dependencies. In backpropagation through time settings, these issues are tightly coupled with the large, sequential computational graph resulting from unfolding the RNN in time. We introduce the Skip RNN model which extends existing RNN models by learning to skip state updates and shortens the effective size of the computational graph. This model can also be encouraged to perform fewer state updates through a budget constraint. We evaluate the proposed model on various tasks and show how it can reduce the number of required RNN updates while preserving, and sometimes even improving, the performance of the baseline RNN models. Source code is publicly available at this https URL ."
                },
                {
                    "title": "Learning to Skim Text",
                    "abstract": "Recurrent Neural Networks are showing much promise in many sub-areas of natural language processing, ranging from document classification to machine translation to automatic question answering. Despite their promise, many recurrent models have to read the whole text word by word, making it slow to handle long documents. For example, it is difficult to use a recurrent network to read a book and answer questions about it. In this paper, we present an approach of reading text while skipping irrelevant information if needed. The underlying model is a recurrent network that learns how far to jump after reading a few words of the input text. We employ a standard policy gradient method to train the model to make discrete jumping decisions. In our benchmarks on four different tasks, including number prediction, sentiment analysis, news article classification and automatic Q&A, our proposed model, a modified LSTM with jumping, is up to 6 times faster than the standard sequential LSTM, while maintaining the same or even better accuracy."
                },
                {
                    "title": "Multi-Scale Dense Networks for Resource Efficient Image Classification",
                    "abstract": "In this paper we investigate image classification with computational resource limits at test time. Two such settings are: 1. anytime classification, where the network's prediction for a test example is progressively updated, facilitating the output of a prediction at any time; and 2. budgeted batch classification, where a fixed amount of computation is available to classify a set of examples that can be spent unevenly across \"easier\" and \"harder\" inputs. In contrast to most prior work, such as the popular Viola and Jones algorithm, our approach is based on convolutional neural networks. We train multiple classifiers with varying resource demands, which we adaptively apply during test time. To maximally re-use computation between the classifiers, we incorporate them as early-exits into a single deep convolutional neural network and inter-connect them with dense connectivity. To facilitate high quality classification early on, we use a two-dimensional multi-scale network architecture that maintains coarse and fine level features all-throughout the network. Experiments on three image-classification tasks demonstrate that our framework substantially improves the existing state-of-the-art in both settings."
                },
                {
                    "title": "Adaptive Neural Networks for Efficient Inference",
                    "abstract": "We present an approach to adaptively utilize deep neural networks in order to reduce the evaluation time on new examples without loss of accuracy. Rather than attempting to redesign or approximate existing networks, we propose two schemes that adaptively utilize networks. We first pose an adaptive network evaluation scheme, where we learn a system to adaptively choose the components of a deep network to be evaluated for each example. By allowing examples correctly classified using early layers of the system to exit, we avoid the computational time associated with full evaluation of the network. We extend this to learn a network selection system that adaptively selects the network to be evaluated for each example. We show that computational time can be dramatically reduced by exploiting the fact that many examples can be correctly classified using relatively efficient networks and that complex, computationally costly networks are only necessary for a small fraction of examples. We pose a global objective for learning an adaptive early exit or network selection policy and solve it by reducing the policy learning problem to a layer-by-layer weighted binary classification problem. Empirically, these approaches yield dramatic reductions in computational cost, with up to a 2.8x speedup on state-of-the-art networks from the ImageNet image recognition challenge with minimal (<1%) loss of top5 accuracy."
                },
                {
                    "title": "BranchyNet: Fast inference via early exiting from deep neural networks",
                    "abstract": "Deep neural networks are state of the art methods for many learning tasks due to their ability to extract increasingly better features at each network layer. However, the improved performance of additional layers in a deep network comes at the cost of added latency and energy usage in feedforward inference. As networks continue to get deeper and larger, these costs become more prohibitive for real-time and energy-sensitive applications. To address this issue, we present BranchyNet, a novel deep network architecture that is augmented with additional side branch classifiers. The architecture allows prediction results for a large portion of test samples to exit the network early via these branches when samples can already be inferred with high confidence. BranchyNet exploits the observation that features learned at an early layer of a network may often be sufficient for the classification of many data points. For more difficult samples, which are expected less frequently, BranchyNet will use further or all network layers to provide the best likelihood of correct prediction. We study the BranchyNet architecture using several well-known networks (LeNet, AlexNet, ResNet) and datasets (MNIST, CIFAR10) and show that it can both improve accuracy and significantly reduce the inference time of the network."
                },
                {
                    "title": "Variable Computation in Recurrent Neural Networks",
                    "abstract": "Recurrent neural networks (RNNs) have been used extensively and with increasing success to model various types of sequential data. Much of this progress has been achieved through devising recurrent units and architectures with the flexibility to capture complex statistics in the data, such as long range dependency or localized attention phenomena. However, while many sequential data (such as video, speech or language) can have highly variable information flow, most recurrent models still consume input features at a constant rate and perform a constant number of computations per time step, which can be detrimental to both speed and model capacity. In this paper, we explore a modification to existing recurrent units which allows them to learn to vary the amount of computation they perform at each step, without prior knowledge of the sequence's time structure. We show experimentally that not only do our models require fewer operations, they also lead to better performance overall on evaluation tasks."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Learning both Weights and Connections for Efficient Neural Network",
                    "abstract": "Neural networks are both computationally intensive and memory intensive, making them difficult to deploy on embedded systems. Also, conventional networks fix the architecture before training starts; as a result, training cannot improve the architecture. To address these limitations, we describe a method to reduce the storage and computation required by neural networks by an order of magnitude without affecting their accuracy by learning only the important connections. Our method prunes redundant connections using a three-step method. First, we train the network to learn which connections are important. Next, we prune the unimportant connections. Finally, we retrain the network to fine tune the weights of the remaining connections. On the ImageNet dataset, our method reduced the number of parameters of AlexNet by a factor of 9x, from 61 million to 6.7 million, without incurring accuracy loss. Similar experiments with VGG-16 found that the number of parameters can be reduced by 13x, from 138 million to 10.3 million, again with no loss of accuracy."
                },
                {
                    "title": "Power Consumption During Neuronal Computation",
                    "abstract": "Maintaining the ability of the nervous system to perceive, remember, process, and react to the outside world requires a continuous energy supply. Yet the overall power consumption is remarkably low, which has inspired engineers to mimic nervous systems in designing artificial cochlea, retinal implants, and brain-computer interfaces (BCIs) to improve the quality of life in patients. Such neuromorphic devices are both energy efficient and increasingly able to emulate many functions of the human nervous system. We examine the energy constraints of neuronal signaling within biology, review the quantitative tradeoff between energy use and information processing, and ask whether the biophysics and design of nerve cells minimizes energy consumption."
                },
                {
                    "title": "HMDB: A large video database for human motion recognition",
                    "abstract": "With nearly one billion online videos viewed everyday, an emerging new frontier in computer vision research is recognition and search in video. While much effort has been devoted to the collection and annotation of large scalable static image datasets containing thousands of image categories, human action datasets lag far behind. Current action recognition databases contain on the order of ten different action categories collected under fairly controlled conditions. State-of-the-art performance on these datasets is now near ceiling and thus there is a need for the design and creation of new benchmarks. To address this issue we collected the largest action video database to-date with 51 action categories, which in total contain around 7,000 manually annotated clips extracted from a variety of sources ranging from digitized movies to YouTube. We use this database to evaluate the performance of two representative computer vision systems for action recognition and explore the robustness of these methods under various conditions such as camera motion, viewpoint, video quality and occlusion."
                },
                {
                    "title": "Cramming More Components Onto Integrated Circuits",
                    "abstract": "The future of integrated electronics is the future of electronics itself. The advantages of integration will bring about a proliferation of electronics, pushing this science into many new areas. Integrated circuits will lead to such wonders as home computers\u2014or at least terminals connected to a central computer\u2014automatic controls for automobiles, and personal portable communications equipment. The electronic wristwatch needs only a display to be feasible today. But the biggest potential lies in the production of large systems. In telephone communications, integrated circuits in digital filters will separate channels on multiplex equipment. Integrated circuits will also switch telephone circuits and perform data processing. Computers will be more powerful, and will be organized in completely different ways. For example, memories built of integrated electronics may be distributed throughout the machine instead of being concentrated in a central unit. In addition, the improved reliability made possible by integrated circuits will allow the construction of larger processing units. Machines similar to those in existence today will be built at lower costs and with faster turnaround."
                },
                {
                    "title": "Optimal Brain Damage",
                    "abstract": "We have used information-theoretic ideas to derive a class of practical and nearly optimal schemes for adapting the size of a neural network. By removing unimportant weights from a network, several improvements can be expected: better generalization, fewer training examples required, and improved speed of learning and/or classification. The basic idea is to use second-derivative information to make a tradeoff between network complexity and training set error. Experiments confirm the usefulness of the methods on a real-world application."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.SD",
                "eess.AS"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop energy-efficient neural network models that maintain high performance for real-time applications on compact devices, given the limitations of current hardware and the increasing complexity of state-of-the-art neural networks?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing need for efficient machine learning models that can operate on resource-constrained devices, such as wearables and IoT devices. By advancing energy efficiency in neural networks, we can enable a wider range of applications, particularly in real-time systems where performance and efficiency are paramount. This research could lead to breakthroughs in sustainable AI, reducing the ecological footprint of machine learning technologies, and fostering innovation in areas like smart devices, healthcare, and autonomous systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent trade-off between model performance and energy efficiency. Current approaches often prioritize accuracy and complexity, leading to models that are computationally intensive and unsuitable for low-power environments. Naive methods, such as simple model compression or pruning, may fail to achieve the necessary efficiency without sacrificing performance. Additionally, the variability in running environments and the need for real-time processing introduce significant technical and practical obstacles that complicate the development of efficient neural networks.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on maximizing model accuracy without adequately addressing energy efficiency, leading to a gap in solutions suitable for real-time applications on compact devices. Existing methods often overlook the unique requirements of low-power environments, and there has been a lack of comprehensive frameworks that integrate efficiency metrics into the model development process. Our approach aims to bridge this gap by introducing novel methodologies that prioritize energy efficiency while maintaining high performance, thus improving upon prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of a new neural network architecture that incorporates energy-efficient design principles, leveraging techniques such as Short-Term Memory Convolution (STMC) to reduce computational requirements. We will utilize benchmark datasets relevant to real-time applications and evaluate our models using metrics that assess both performance and energy consumption. The expected outcomes include a significant reduction in inference time and memory usage while achieving competitive accuracy, ultimately demonstrating the feasibility of deploying high-performance neural networks on compact devices."
            }
        },
        "author_data": {
            "29057f61-ff65-48f6-9cee-1eb80e8c56c0": {
                "pk": "29057f61-ff65-48f6-9cee-1eb80e8c56c0",
                "name": "Grzegorz Stefa\u0144ski",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "edd4e2eb-a231-49ca-9635-b2b145237707": {
                "pk": "edd4e2eb-a231-49ca-9635-b2b145237707",
                "name": "Pawe\u0142 Daniluk",
                "collaborators": [
                    "Grzegorz Stefa\u0144ski",
                    "Krzysztof Arendt",
                    "Bart\u0142omiej Jasik",
                    "Artur Szumaczuk"
                ],
                "domain": [
                    "Deep Learning",
                    "Signal Processing",
                    "Audio Processing",
                    "Convolutional Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Short-Term Memory Convolutions",
                        "abstract": "The real-time processing of time series signals is a critical issue for many real-life applications. The idea of real-time processing is especially important in audio domain as the human perception of sound is sensitive to any kind of disturbance in perceived signals, especially the lag between auditory and visual modalities. The rise of deep learning (DL) models complicated the landscape of signal processing. Although they often have superior quality compared to standard DSP methods, this advantage is diminished by higher latency. In this work we propose novel method for minimization of inference time latency and memory consumption, called Short-Term Memory Convolution (STMC) and its transposed counterpart. The main advantage of STMC is the low latency comparable to long short-term memory (LSTM) networks. Furthermore, the training of STMC-based models is faster and more stable as the method is based solely on convolutional neural networks (CNNs). In this study we demonstrate an application of this solution to a U-Net model for a speech separation task and GhostNet model in acoustic scene classification (ASC) task. In case of speech separation we achieved a 5-fold reduction in inference time and a 2-fold reduction in latency without affecting the output quality. The inference time for ASC task was up to 4 times faster while preserving the original accuracy."
                    }
                ]
            },
            "05458b5a-1c22-48a5-aa17-5a99d1ea483b": {
                "pk": "05458b5a-1c22-48a5-aa17-5a99d1ea483b",
                "name": "Artur Szumaczuk",
                "collaborators": [
                    "Grzegorz Stefa\u0144ski",
                    "Krzysztof Arendt",
                    "Pawe\u0142 Daniluk",
                    "Bart\u0142omiej Jasik"
                ],
                "domain": [
                    "Deep Learning",
                    "Signal Processing",
                    "Audio Processing",
                    "Convolutional Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Short-Term Memory Convolutions",
                        "abstract": "The real-time processing of time series signals is a critical issue for many real-life applications. The idea of real-time processing is especially important in audio domain as the human perception of sound is sensitive to any kind of disturbance in perceived signals, especially the lag between auditory and visual modalities. The rise of deep learning (DL) models complicated the landscape of signal processing. Although they often have superior quality compared to standard DSP methods, this advantage is diminished by higher latency. In this work we propose novel method for minimization of inference time latency and memory consumption, called Short-Term Memory Convolution (STMC) and its transposed counterpart. The main advantage of STMC is the low latency comparable to long short-term memory (LSTM) networks. Furthermore, the training of STMC-based models is faster and more stable as the method is based solely on convolutional neural networks (CNNs). In this study we demonstrate an application of this solution to a U-Net model for a speech separation task and GhostNet model in acoustic scene classification (ASC) task. In case of speech separation we achieved a 5-fold reduction in inference time and a 2-fold reduction in latency without affecting the output quality. The inference time for ASC task was up to 4 times faster while preserving the original accuracy."
                    }
                ]
            },
            "63e4341e-c9cc-443c-9d01-dad332857cdd": {
                "pk": "63e4341e-c9cc-443c-9d01-dad332857cdd",
                "name": "Jakub Tkaczuk",
                "collaborators": [
                    "S\u0142awomir Kapka"
                ],
                "domain": [
                    "Sound Event Localization",
                    "Deep Learning",
                    "Audio Processing"
                ],
                "publications": [
                    {
                        "title": "CoLoC: Conditioned Localizer and Classifier for Sound Event Localization and Detection",
                        "abstract": "In this article, we describe Conditioned Localizer and Classifier (CoLoC) which is a novel solution for Sound Event Localization and Detection (SELD). The solution constitutes of two stages: the localization is done first and is followed by classification conditioned by the output of the localizer. In order to resolve the problem of the unknown number of sources we incorporate the idea borrowed from Sequential Set Generation (SSG). Models from both stages are SELDnet-like CRNNs, but with single outputs. Conducted reasoning shows that such two single-output models are fit for SELD task. We show that our solution improves on the baseline system in most metrics on the STARSS22 Dataset."
                    }
                ]
            }
        }
    },
    "2406.09358": {
        "paper_data": {
            "title": "Understanding Hallucinations in Diffusion Models through Mode Interpolation",
            "url": "http://arxiv.org/abs/2406.09358v2",
            "arxiv_id": "2406.09358",
            "authors": [
                "Sumukh K Aithal",
                "Pratyush Maini",
                "Zachary C. Lipton",
                "J. Zico Kolter"
            ],
            "abstract": "Colloquially speaking, image generation models based upon diffusion processes are frequently said to exhibit \"hallucinations,\" samples that could never occur in the training data. But where do such hallucinations come from? In this paper, we study a particular failure mode in diffusion models, which we term mode interpolation. Specifically, we find that diffusion models smoothly \"interpolate\" between nearby data modes in the training set, to generate samples that are completely outside the support of the original training distribution; this phenomenon leads diffusion models to generate artifacts that never existed in real data (i.e., hallucinations). We systematically study the reasons for, and the manifestation of this phenomenon. Through experiments on 1D and 2D Gaussians, we show how a discontinuous loss landscape in the diffusion model's decoder leads to a region where any smooth approximation will cause such hallucinations. Through experiments on artificial datasets with various shapes, we show how hallucination leads to the generation of combinations of shapes that never existed. Finally, we show that diffusion models in fact know when they go out of support and hallucinate. This is captured by the high variance in the trajectory of the generated sample towards the final few backward sampling process. Using a simple metric to capture this variance, we can remove over 95% of hallucinations at generation time while retaining 96% of in-support samples. We conclude our exploration by showing the implications of such hallucination (and its removal) on the collapse (and stabilization) of recursive training on synthetic data with experiments on MNIST and 2D Gaussians dataset. We release our code at https://github.com/locuslab/diffusion-model-hallucination.",
            "introduction": "   1 Introduction  The high quality and diversity of images generated by diffusion models\u00a0[37, 14] have made them the de facto standard generative models across various tasks including video generation [6], image inpainting [23], image super-resolution [10], data augmentation [43], and others. As a result of their uptake, large volumes of synthetic data are rapidly proliferating on the internet. The next generation of generative models will likely be exposed to many machine-generated instances during their training, making it crucial to understand ways in which diffusion models fail to model the true underlying data distribution. Like other generative model families, much research has been done to understand the failure modes of diffusion models as well. Past works have identified, and attempted to explain and remedy failures such as, training instabilities\u00a0[16], memorization \u00a0[7, 38] and inaccurate modeling of objects such as hands and legs\u00a0[27, 22, 4].   In this work, we formalize and study a particular failure mode of diffusion models that we call hallucination\u2014a phenomenon where diffusion models generate samples that lie completely out of the support of the training distribution of the model. As a contemporary example, hallucinations manifest in large generative models like StableDiffusion\u00a0[32] in the form of hands with extra (or missing) fingers or limbs. We begin our investigation with a surprising observation that an unconditional diffusion model trained on a distribution of simple shapes, generates images with combinations of shapes (or artifacts) that never existed in the original training distribution (Figure\u00a01). While extensive research on generative models has focused on the phenomenon of \u2018mode collapse\u2019\u00a0[47], which leads to a loss of diversity in the sampled distribution, such studies often overlook the complex nature of real data which typically comprise multiple distinct modes on a complex data manifold, and the effects of their mutual interactions are thus neglected. In our work, we explain hallucinations by introducing a novel phenomenon we term \u2018mode interpolation\u2019 that considers this mutual interaction.   To understand the cause of these hallucinations and their relationship to mode interpolation, we construct simplified 1-d and 2-d mixture of Gaussian setups and train diffusion models on them (\u00a7\u00a04). We observe that when the true data distribution occurs in disjoint modes, diffusion models are unable to model a true approximation of the underlying distribution. This is because there exist \u2018step functions\u2019 between different modes, but the score function learned by the DDPM is a smooth approximation of the same, leading to interpolation between the nearest modes, even when these interpolated values are entirely absent from the training data. Moreover, we observation that hallucinated samples usually have very high variance towards the end of their trajectory during the reverse diffusion process. Based on this observation, we use the trajectory variance during sampling as a metric to detect hallucinations (\u00a7\u00a05), and show that diffusion models usually \u2018know\u2019 when they hallucinate, allowing detection with sensitivity and specificity >0.92absent0.92>0.92> 0.92 in our experiments.   We explore mode interpolation as a potential explanation for the common failure of large-scale generative models, to accurately generate human hands. To demonstrate this concretely, we trained a diffusion model on a dataset of high-quality hand images and observed that it generated hands",
            "references": [
                {
                    "title": "Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data",
                    "abstract": "The proliferation of generative models, combined with pretraining on web-scale data, raises a timely question: what happens when these models are trained on their own generated outputs? Recent investigations into model-data feedback loops proposed that such loops would lead to a phenomenon termed model collapse, under which performance progressively degrades with each model-data feedback iteration until fitted models become useless. However, those studies largely assumed that new data replace old data over time, where an arguably more realistic assumption is that data accumulate over time. In this paper, we ask: what effect does accumulating data have on model collapse? We empirically study this question by pretraining sequences of language models on text corpora. We confirm that replacing the original real data by each generation's synthetic data does indeed tend towards model collapse, then demonstrate that accumulating the successive generations of synthetic data alongside the original real data avoids model collapse; these results hold across a range of model sizes, architectures, and hyperparameters. We obtain similar results for deep generative models on other types of real data: diffusion models for molecule conformation generation and variational autoencoders for image generation. To understand why accumulating data can avoid model collapse, we use an analytically tractable framework introduced by prior work in which a sequence of linear models are fit to the previous models' outputs. Previous work used this framework to show that if data are replaced, the test error increases with the number of model-fitting iterations; we extend this argument to prove that if data instead accumulate, the test error has a finite upper bound independent of the number of iterations, meaning model collapse no longer occurs."
                },
                {
                    "title": "HanDiffuser: Text-to-Image Generation with Realistic Hand Appearances",
                    "abstract": "Text-to-image generative models can generate high-quality humans, but realism is lost when generating hands. Common artifacts include irregular hand poses, shapes, incorrect numbers of fingers, and physically implausible finger orientations. To generate images with realistic hands, we propose a novel diffusion-based architecture called HanDiffuser that achieves realism by injecting hand embeddings in the generative process. HanDiffuser consists of two components: a Text-to-Hand-Params diffusion model to generate SMPL-Body and MANO-Hand parameters from input text prompts, and a Text-Guided Hand-Params-to-Image diffusion model to synthesize images by conditioning on the prompts and hand parameters generated by the previous component. We incorporate multiple aspects of hand representation, including 3D shapes and joint-level finger positions, orientations and articulations, for robust learning and reliable performance during inference. We conduct extensive quantitative and qualitative experiments and perform user studies to demonstrate the efficacy of our method in generating images with high-quality hands. Project page: https://supreethn.github.io/research/handiffuser/index.html"
                },
                {
                    "title": "Towards Theoretical Understandings of Self-Consuming Generative Models",
                    "abstract": "This paper tackles the emerging challenge of training generative models within a self-consuming loop, wherein successive generations of models are recursively trained on mixtures of real and synthetic data from previous generations. We construct a theoretical framework to rigorously evaluate how this training procedure impacts the data distributions learned by future models, including parametric and non-parametric models. Specifically, we derive bounds on the total variation (TV) distance between the synthetic data distributions produced by future models and the original real data distribution under various mixed training scenarios for diffusion models with a one-hidden-layer neural network score function. Our analysis demonstrates that this distance can be effectively controlled under the condition that mixed training dataset sizes or proportions of real data are large enough. Interestingly, we further unveil a phase transition induced by expanding synthetic data amounts, proving theoretically that while the TV distance exhibits an initial ascent, it declines beyond a threshold point. Finally, we present results for kernel density estimation, delivering nuanced insights such as the impact of mixed data training on error propagation."
                },
                {
                    "title": "A Tale of Tails: Model Collapse as a Change of Scaling Laws",
                    "abstract": "As AI model size grows, neural scaling laws have become a crucial tool to predict the improvements of large models when increasing capacity and the size of original (human or natural) training data. Yet, the widespread use of popular models means that the ecosystem of online data and text will co-evolve to progressively contain increased amounts of synthesized data. In this paper we ask: How will the scaling laws change in the inevitable regime where synthetic data makes its way into the training corpus? Will future models, still improve, or be doomed to degenerate up to total (model) collapse? We develop a theoretical framework of model collapse through the lens of scaling laws. We discover a wide range of decay phenomena, analyzing loss of scaling, shifted scaling with number of generations, the ''un-learning\"of skills, and grokking when mixing human and synthesized data. Our theory is validated by large-scale experiments with a transformer on an arithmetic task and text generation using the large language model Llama2."
                },
                {
                    "title": "Large Language Models Suffer From Their Own Output: An Analysis of the Self-Consuming Training Loop",
                    "abstract": "Large Language Models (LLM) are already widely used to generate content for a variety of online platforms. As we are not able to safely distinguish LLM-generated content from human-produced content, LLM-generated content is used to train the next generation of LLMs, giving rise to a self-consuming training loop. From the image generation domain we know that such a self-consuming training loop reduces both quality and diversity of images finally ending in a model collapse. However, it is unclear whether this alarming effect can also be observed for LLMs. Therefore, we present the first study investigating the self-consuming training loop for LLMs. Further, we propose a novel method based on logic expressions that allows us to unambiguously verify the correctness of LLM-generated content, which is difficult for natural language text. We find that the self-consuming training loop produces correct outputs, however, the output declines in its diversity depending on the proportion of the used generated data. Fresh data can slow down this decline, but not stop it. Given these concerning results, we encourage researchers to study methods to negate this process."
                },
                {
                    "title": "ScaleLong: Towards More Stable Training of Diffusion Model via Scaling Network Long Skip Connection",
                    "abstract": "In diffusion models, UNet is the most popular network backbone, since its long skip connects (LSCs) to connect distant network blocks can aggregate long-distant information and alleviate vanishing gradient. Unfortunately, UNet often suffers from unstable training in diffusion models which can be alleviated by scaling its LSC coefficients smaller. However, theoretical understandings of the instability of UNet in diffusion models and also the performance improvement of LSC scaling remain absent yet. To solve this issue, we theoretically show that the coefficients of LSCs in UNet have big effects on the stableness of the forward and backward propagation and robustness of UNet. Specifically, the hidden feature and gradient of UNet at any layer can oscillate and their oscillation ranges are actually large which explains the instability of UNet training. Moreover, UNet is also provably sensitive to perturbed input, and predicts an output distant from the desired output, yielding oscillatory loss and thus oscillatory gradient. Besides, we also observe the theoretical benefits of the LSC coefficient scaling of UNet in the stableness of hidden features and gradient and also robustness. Finally, inspired by our theory, we propose an effective coefficient scaling framework ScaleLong that scales the coefficients of LSC in UNet and better improves the training stability of UNet. Experimental results on four famous datasets show that our methods are superior to stabilize training and yield about 1.5x training acceleration on different diffusion models with UNet or UViT backbones. Code: https://github.com/sail-sg/ScaleLong"
                },
                {
                    "title": "On the Stability of Iterative Retraining of Generative Models on their own Data",
                    "abstract": "Deep generative models have made tremendous progress in modeling complex data, often exhibiting generation quality that surpasses a typical human's ability to discern the authenticity of samples. Undeniably, a key driver of this success is enabled by the massive amounts of web-scale data consumed by these models. Due to these models' striking performance and ease of availability, the web will inevitably be increasingly populated with synthetic content. Such a fact directly implies that future iterations of generative models will be trained on both clean and artificially generated data from past models. In this paper, we develop a framework to rigorously study the impact of training generative models on mixed datasets -- from classical training on real data to self-consuming generative models trained on purely synthetic data. We first prove the stability of iterative training under the condition that the initial generative models approximate the data distribution well enough and the proportion of clean training data (w.r.t. synthetic data) is large enough. We empirically validate our theory on both synthetic and natural images by iteratively training normalizing flows and state-of-the-art diffusion models on CIFAR10 and FFHQ."
                },
                {
                    "title": "Cognitive Mirage: A Review of Hallucinations in Large Language Models",
                    "abstract": "As large language models continue to develop in the field of AI, text generation systems are susceptible to a worrisome phenomenon known as hallucination. In this study, we summarize recent compelling insights into hallucinations in LLMs. We present a novel taxonomy of hallucinations from various text generation tasks, thus provide theoretical insights, detection methods and improvement approaches. Based on this, future research directions are proposed. Our contribution are threefold: (1) We provide a detailed and complete taxonomy for hallucinations appearing in text generation tasks; (2) We provide theoretical analyses of hallucinations in LLMs and provide existing detection and improvement methods; (3) We propose several research directions that can be developed in the future. As hallucinations garner significant attention from the community, we will maintain updates on relevant research progress."
                },
                {
                    "title": "Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Models",
                    "abstract": "While large language models (LLMs) have demonstrated remarkable capabilities across a range of downstream tasks, a significant concern revolves around their propensity to exhibit hallucinations: LLMs occasionally generate content that diverges from the user input, contradicts previously generated context, or misaligns with established world knowledge. This phenomenon poses a substantial challenge to the reliability of LLMs in real-world scenarios. In this paper, we survey recent efforts on the detection, explanation, and mitigation of hallucination, with an emphasis on the unique challenges posed by LLMs. We present taxonomies of the LLM hallucination phenomena and evaluation benchmarks, analyze existing approaches aiming at mitigating LLM hallucination, and discuss potential directions for future research."
                },
                {
                    "title": "Self-Consuming Generative Models Go MAD",
                    "abstract": "Seismic advances in generative AI algorithms have led to the temptation to use AI-synthesized data to train next-generation models. Repeating this process creates autophagous (\u201cself-consuming\u201d) loops whose properties are poorly understood. We conduct a thorough analysis using state-of-the-art generative image models of three autophagous loop families that differ in how they incorporate fixed or fresh real training data and whether previous generations' samples have been biased to trade off data quality versus diversity. Our primary conclusion across all scenarios is that without enough fresh real data in each generation of an autophagous loop, future generative models are doomed to have their quality (precision) or diversity (recall) progressively decrease. We term this condition Model Autophagy Disorder (MAD) and show that appreciable MADness arises in just a few generations."
                },
                {
                    "title": "Textbooks Are All You Need",
                    "abstract": "We introduce phi-1, a new large language model for code, with significantly smaller size than competing models: phi-1 is a Transformer-based model with 1.3B parameters, trained for 4 days on 8 A100s, using a selection of ``textbook quality\"data from the web (6B tokens) and synthetically generated textbooks and exercises with GPT-3.5 (1B tokens). Despite this small scale, phi-1 attains pass@1 accuracy 50.6% on HumanEval and 55.5% on MBPP. It also displays surprising emergent properties compared to phi-1-base, our model before our finetuning stage on a dataset of coding exercises, and phi-1-small, a smaller model with 350M parameters trained with the same pipeline as phi-1 that still achieves 45% on HumanEval."
                },
                {
                    "title": "Towards Understanding the Interplay of Generative Artificial Intelligence and the Internet",
                    "abstract": "The rapid adoption of generative Artificial Intelligence (AI) tools that can generate realistic images or text, such as DALL-E, MidJourney, or ChatGPT, have put the societal impacts of these technologies at the center of public debate. These tools are possible due to the massive amount of data (text and images) that is publicly available through the Internet. At the same time, these generative AI tools become content creators that are already contributing to the data that is available to train future models. Therefore, future versions of generative AI tools will be trained with a mix of human-created and AI-generated content, causing a potential feedback loop between generative AI and public data repositories. This interaction raises many questions: how will future versions of generative AI tools behave when trained on a mixture of real and AI generated data? Will they evolve and improve with the new data sets or on the contrary will they degrade? Will evolution introduce biases or reduce diversity in subsequent generations of generative AI tools? What are the societal implications of the possible degradation of these models? Can we mitigate the effects of this feedback loop? In this document, we explore the effect of this interaction and report some initial results using simple diffusion models trained with various image datasets. Our results show that the quality and diversity of the generated images can degrade over time suggesting that incorporating AI-created data can have undesired effects on future versions of generative models."
                },
                {
                    "title": "The Curse of Recursion: Training on Generated Data Makes Models Forget",
                    "abstract": "Stable Diffusion revolutionised image creation from descriptive text. GPT-2, GPT-3(.5) and GPT-4 demonstrated astonishing performance across a variety of language tasks. ChatGPT introduced such language models to the general public. It is now clear that large language models (LLMs) are here to stay, and will bring about drastic change in the whole ecosystem of online text and images. In this paper we consider what the future might hold. What will happen to GPT-{n} once LLMs contribute much of the language found online? We find that use of model-generated content in training causes irreversible defects in the resulting models, where tails of the original content distribution disappear. We refer to this effect as Model Collapse and show that it can occur in Variational Autoencoders, Gaussian Mixture Models and LLMs. We build theoretical intuition behind the phenomenon and portray its ubiquity amongst all learned generative models. We demonstrate that it has to be taken seriously if we are to sustain the benefits of training from large-scale data scraped from the web. Indeed, the value of data collected about genuine human interactions with systems will be increasingly valuable in the presence of content generated by LLMs in data crawled from the Internet."
                },
                {
                    "title": "Generating Images of Rare Concepts Using Pre-trained Diffusion Models",
                    "abstract": "Text-to-image diffusion models can synthesize high quality images, but they have various limitations. Here we highlight a common failure mode of these models, namely, generating uncommon concepts and structured concepts like hand palms. We show that their limitation is partly due to the long-tail nature of their training data: web-crawled data sets are strongly unbalanced, causing models to under-represent concepts from the tail of the distribution. We characterize the effect of unbalanced training data on text-to-image models and offer a remedy. We show that rare concepts can be correctly generated by carefully selecting suitable generation seeds in the noise space, using a small reference set of images, a technique that we call SeedSelect. SeedSelect does not require retraining or finetuning the diffusion model. We assess the faithfulness, quality and diversity of SeedSelect in creating rare objects and generating complex formations like hand images, and find it consistently achieves superior performance. We further show the advantage of SeedSelect in semantic data augmentation. Generating semantically appropriate images can successfully improve performance in few-shot recognition benchmarks, for classes from the head and from the tail of the training data of diffusion models."
                },
                {
                    "title": "Implicit Diffusion Models for Continuous Super-Resolution",
                    "abstract": "Image super-resolution (SR) has attracted increasing attention due to its widespread applications. However, current SR methods generally suffer from over-smoothing and artifacts, and most work only with fixed magnifications. This paper introduces an Implicit Diffusion Model (IDM) for high-fidelity continuous image super-resolution. IDM integrates an implicit neural representation and a denoising diffusion model in a unified end-to-end framework, where the implicit neural representation is adopted in the decoding process to learn continuous-resolution representation. Furthermore, we design a scale-adaptive conditioning mechanism that consists of a low-resolution (LR) conditioning network and a scaling factor. The scaling factor regulates the resolution and accordingly modulates the proportion of the LR information and generated features in the final output, which enables the model to accommodate the continuous-resolution requirement. Extensive experiments validate the effectiveness of our IDM and demonstrate its superior performance over prior arts. The source code will be available at https://github.com/Ree1s/IDM."
                },
                {
                    "title": "Qualitative Failures of Image Generation Models and Their Application in Detecting Deepfakes",
                    "abstract": "The ability of image and video generation models to create photorealistic images has reached unprecedented heights, making it difficult to distinguish between real and fake images in many cases. However, despite this progress, a gap remains between the quality of generated images and those found in the real world. To address this, we have reviewed a vast body of literature from both academic publications and social media to identify qualitative shortcomings in image generation models, which we have classified into five categories. By understanding these failures, we can identify areas where these models need improvement, as well as develop strategies for detecting deep fakes. The prevalence of deep fakes in today's society is a serious concern, and our findings can help mitigate their negative impact."
                },
                {
                    "title": "Combining Generative Artificial Intelligence (AI) and the Internet: Heading towards Evolution or Degradation?",
                    "abstract": "In the span of a few months, generative Artificial Intelligence (AI) tools that can generate realistic images or text have taken the Internet by storm, making them one of the technologies with fastest adoption ever. Some of these generative AI tools such as DALL-E, MidJourney, or ChatGPT have gained wide public notoriety. Interestingly, these tools are possible because of the massive amount of data (text and images) available on the Internet. The tools are trained on massive data sets that are scraped from Internet sites. And now, these generative AI tools are creating massive amounts of new data that are being fed into the Internet. Therefore, future versions of generative AI tools will be trained with Internet data that is a mix of original and AI-generated data. As time goes on, a mixture of original data and data generated by different versions of AI tools will populate the Internet. This raises a few intriguing questions: how will future versions of generative AI tools behave when trained on a mixture of real and AI generated data? Will they evolve with the new data sets or degenerate? Will evolution introduce biases in subsequent generations of generative AI tools? In this document, we explore these questions and report some very initial simulation results using a simple image-generation AI tool. These results suggest that the quality of the generated images degrades as more AI-generated data is used for training thus suggesting that generative AI may degenerate. Although these results are preliminary and cannot be generalised without further study, they serve to illustrate the potential issues of the interaction between generative AI and the Internet."
                },
                {
                    "title": "Effective Data Augmentation With Diffusion Models",
                    "abstract": "Data augmentation is one of the most prevalent tools in deep learning, underpinning many recent advances, including those from classification, generative models, and representation learning. The standard approach to data augmentation combines simple transformations like rotations and flips to generate new images from existing ones. However, these new images lack diversity along key semantic axes present in the data. Current augmentations cannot alter the high-level semantic attributes, such as animal species present in a scene, to enhance the diversity of data. We address the lack of diversity in data augmentation with image-to-image transformations parameterized by pre-trained text-to-image diffusion models. Our method edits images to change their semantics using an off-the-shelf diffusion model, and generalizes to novel visual concepts from a few labelled examples. We evaluate our approach on few-shot image classification tasks, and on a real-world weed recognition task, and observe an improvement in accuracy in tested domains."
                },
                {
                    "title": "In-Context Retrieval-Augmented Language Models",
                    "abstract": "Abstract Retrieval-Augmented Language Modeling (RALM) methods, which condition a language model (LM) on relevant documents from a grounding corpus during generation, were shown to significantly improve language modeling performance. In addition, they can mitigate the problem of factually inaccurate text generation and provide natural source attribution mechanism. Existing RALM approaches focus on modifying the LM architecture in order to facilitate the incorporation of external information, significantly complicating deployment. This paper considers a simple alternative, which we dub In-Context RALM: leaving the LM architecture unchanged and prepending grounding documents to the input, without any further training of the LM. We show that In-Context RALM that builds on off-the-shelf general purpose retrievers provides surprisingly large LM gains across model sizes and diverse corpora. We also demonstrate that the document retrieval and ranking mechanism can be specialized to the RALM setting to further boost performance. We conclude that In-Context RALM has considerable potential to increase the prevalence of LM grounding, particularly in settings where a pretrained LM must be used without modification or even via API access.1"
                },
                {
                    "title": "Extracting Training Data from Diffusion Models",
                    "abstract": "Image diffusion models such as DALL-E 2, Imagen, and Stable Diffusion have attracted significant attention due to their ability to generate high-quality synthetic images. In this work, we show that diffusion models memorize individual images from their training data and emit them at generation time. With a generate-and-filter pipeline, we extract over a thousand training examples from state-of-the-art models, ranging from photographs of individual people to trademarked company logos. We also train hundreds of diffusion models in various settings to analyze how different modeling and data decisions affect privacy. Overall, our results show that diffusion models are much less private than prior generative models such as GANs, and that mitigating these vulnerabilities may require new advances in privacy-preserving training."
                },
                {
                    "title": "Input Perturbation Reduces Exposure Bias in Diffusion Models",
                    "abstract": "Denoising Diffusion Probabilistic Models have shown an impressive generation quality, although their long sampling chain leads to high computational costs. In this paper, we observe that a long sampling chain also leads to an error accumulation phenomenon, which is similar to the exposure bias problem in autoregressive text generation. Specifically, we note that there is a discrepancy between training and testing, since the former is conditioned on the ground truth samples, while the latter is conditioned on the previously generated results. To alleviate this problem, we propose a very simple but effective training regularization, consisting in perturbing the ground truth samples to simulate the inference time prediction errors. We empirically show that, without affecting the recall and precision, the proposed input perturbation leads to a significant improvement in the sample quality while reducing both the training and the inference times. For instance, on CelebA 64$\\times$64, we achieve a new state-of-the-art FID score of 1.27, while saving 37.5% of the training time. The code is publicly available at https://github.com/forever208/DDPM-IP"
                },
                {
                    "title": "Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Models",
                    "abstract": "Cutting-edge diffusion models produce images with high quality and customizability, enabling them to be used for commercial art and graphic design purposes. But do diffusion models create unique works of art, or are they replicating content directly from their training sets? In this work, we study image retrieval frameworks that enable us to compare generated images with training samples and detect when content has been replicated. Applying our frameworks to diffusion models trained on multiple datasets including Oxford flowers, Celeb-A, ImageNet, and LAION, we discuss how factors such as training set size impact rates of content replication. We also identify cases where diffusion models, including the popular Stable Diffusion model, blatantly copy from their training data. Project page: https://somepago.github.io/diffrep.html"
                },
                {
                    "title": "Understanding Diffusion Models: A Unified Perspective",
                    "abstract": "Diffusion models have shown incredible capabilities as generative models; indeed, they power the current state-of-the-art models on text-conditioned image generation such as Imagen and DALL-E 2. In this work we review, demystify, and unify the understanding of diffusion models across both variational and score-based perspectives. We first derive Variational Diffusion Models (VDM) as a special case of a Markovian Hierarchical Variational Autoencoder, where three key assumptions enable tractable computation and scalable optimization of the ELBO. We then prove that optimizing a VDM boils down to learning a neural network to predict one of three potential objectives: the original source input from any arbitrary noisification of it, the original source noise from any arbitrarily noisified input, or the score function of a noisified input at any arbitrary noise level. We then dive deeper into what it means to learn the score function, and connect the variational perspective of a diffusion model explicitly with the Score-based Generative Modeling perspective through Tweedie's Formula. Lastly, we cover how to learn a conditional distribution using diffusion models via guidance."
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "Understanding DDPM Latent Codes Through Optimal Transport",
                    "abstract": "Diffusion models have recently outperformed alternative approaches to model the distribution of natural images, such as GANs. Such diffusion models allow for deterministic sampling via the probability flow ODE, giving rise to a latent space and an encoder map. While having important practical applications, such as estimation of the likelihood, the theoretical properties of this map are not yet fully understood. In the present work, we partially address this question for the popular case of the VP SDE (DDPM) approach. We show that, perhaps surprisingly, the DDPM encoder map coincides with the optimal transport map for common distributions; we support this claim theoretically and by extensive numerical experiments."
                },
                {
                    "title": "RePaint: Inpainting using Denoising Diffusion Probabilistic Models",
                    "abstract": "Free-form inpainting is the task of adding new content to an image in the regions specified by an arbitrary binary mask. Most existing approaches train for a certain distribution of masks, which limits their generalization capabilities to unseen mask types. Furthermore, training with pixel-wise and perceptual losses often leads to simple textural extensions towards the missing areas instead of semantically meaningful generation. In this work, we propose RePaint: A Denoising Diffusion Probabilistic Model (DDPM) based inpainting approach that is applicable to even extreme masks. We employ a pretrained unconditional DDPM as the generative prior. To condition the generation process, we only alter the reverse diffusion iterations by sampling the unmasked regions using the given image infor-mation. Since this technique does not modify or condition the original DDPM network itself, the model produces high-quality and diverse output images for any inpainting form. We validate our method for both faces and general-purpose image inpainting using standard and extreme masks. Re-Paint outperforms state-of-the-art Autoregressive, and GAN approaches for at least five out of six mask distributions. Github Repository: git.io/RePaint"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Variational Diffusion Models",
                    "abstract": "Diffusion-based generative models have demonstrated a capacity for perceptually impressive synthesis, but can they also be great likelihood-based models? We answer this in the affirmative, and introduce a family of diffusion-based generative models that obtain state-of-the-art likelihoods on standard image density estimation benchmarks. Unlike other diffusion-based models, our method allows for efficient optimization of the noise schedule jointly with the rest of the model. We show that the variational lower bound (VLB) simplifies to a remarkably short expression in terms of the signal-to-noise ratio of the diffused data, thereby improving our theoretical understanding of this model class. Using this insight, we prove an equivalence between several models proposed in the literature. In addition, we show that the continuous-time VLB is invariant to the noise schedule, except for the signal-to-noise ratio at its endpoints. This enables us to learn a noise schedule that minimizes the variance of the resulting VLB estimator, leading to faster optimization. Combining these advances with architectural improvements, we obtain state-of-the-art likelihoods on image density estimation benchmarks, outperforming autoregressive models that have dominated these benchmarks for many years, with often significantly faster optimization. In addition, we show how to use the model as part of a bits-back compression scheme, and demonstrate lossless compression rates close to the theoretical optimum. Code is available at https://github.com/google-research/vdm ."
                },
                {
                    "title": "Improved Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) are a class of generative models which have recently been shown to produce excellent samples. We show that with a few simple modifications, DDPMs can also achieve competitive log-likelihoods while maintaining high sample quality. Additionally, we find that learning variances of the reverse diffusion process allows sampling with an order of magnitude fewer forward passes with a negligible difference in sample quality, which is important for the practical deployment of these models. We additionally use precision and recall to compare how well DDPMs and GANs cover the target distribution. Finally, we show that the sample quality and likelihood of these models scale smoothly with model capacity and training compute, making them easily scalable. We release our code at https://github.com/openai/improved-diffusion"
                },
                {
                    "title": "Maximum Likelihood Training of Score-Based Diffusion Models",
                    "abstract": "Score-based diffusion models synthesize samples by reversing a stochastic process that diffuses data to noise, and are trained by minimizing a weighted combination of score matching losses. The log-likelihood of score-based diffusion models can be tractably computed through a connection to continuous normalizing flows, but log-likelihood is not directly optimized by the weighted combination of score matching losses. We show that for a specific weighting scheme, the objective upper bounds the negative log-likelihood, thus enabling approximate maximum likelihood training of score-based diffusion models. We empirically observe that maximum likelihood training consistently improves the likelihood of score-based diffusion models across multiple datasets, stochastic processes, and model architectures. Our best models achieve negative log-likelihoods of 2.83 and 3.76 bits/dim on CIFAR-10 and ImageNet 32x32 without any data augmentation, on a par with state-of-the-art autoregressive models on these tasks."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                    "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                },
                {
                    "title": "On the convergence and mode collapse of GAN",
                    "abstract": "Generative adversarial network (GAN) is a powerful generative model. However, it suffers from several problems, such as convergence instability and mode collapse. To overcome these drawbacks, this paper presents a novel architecture of GAN, which consists of one generator and two different discriminators. With the fact that GAN is the analogy of a minimax game, the proposed architecture is as follows. The generator (G) aims to produce realistic-looking samples to fool both of two discriminators. The first discriminator (D1) rewards high scores for samples from the data distribution, while the second one (D2) favors samples from the generator conversely. Specifically, the ResBlock and minibatch discrimination (MD) architectures are adopted in D1 to improve the diversity of the samples. The leaky rectified linear unit (Leaky ReLU) and batch normalization (BN) are replaced by the scaled exponential linear unit (SELU) in D2 to alleviate the convergence problem. A new loss function that minimizes the KL divergence is designed to better optimize the model. Extensive experiments on CIFAR-10/100 datasets demonstrate that the proposed method can effectively solve the problems of convergence and mode collapse."
                },
                {
                    "title": "GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium",
                    "abstract": "Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the \"Frechet Inception Distance\" (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "Gradient-based learning applied to document recognition",
                    "abstract": "Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow do diffusion models generate hallucinated samples that lie outside the support of the training distribution, and what is the role of mode interpolation in this phenomenon?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding the hallucination phenomenon in diffusion models is crucial for the research community as it addresses a significant failure mode that can lead to the generation of unrealistic or nonsensical outputs. By solving this problem, we can improve the reliability and accuracy of generative models, which are increasingly used in various applications such as image synthesis, video generation, and data augmentation. This research could pave the way for future advancements in generative modeling techniques, enhancing their practical applications in fields like computer vision, art generation, and virtual reality, where realistic outputs are essential.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing hallucinations in diffusion models stem from the complex nature of real data distributions, which often consist of multiple distinct modes. Naive approaches may fail because they do not account for the interactions between these modes, leading to incorrect assumptions about the underlying data distribution. Additionally, the smooth approximation of the score function learned by diffusion models can result in unintended interpolation between modes, generating samples that do not exist in the training data. Overcoming these technical obstacles requires a deep understanding of the model's behavior and the development of new metrics to detect hallucinations effectively.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on issues like mode collapse and training instabilities, often neglecting the intricate interactions between distinct modes in complex data distributions. This oversight has created a gap in understanding how diffusion models can generate hallucinated outputs. Barriers to solving this problem include a lack of formalization of the hallucination phenomenon and insufficient exploration of mode interpolation. Our approach differs by explicitly investigating mode interpolation as a key factor in hallucinations, providing a novel perspective that has not been thoroughly examined in prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves constructing simplified 1-dimensional and 2-dimensional mixtures of Gaussian distributions to study the behavior of diffusion models. We will train these models on the constructed datasets and analyze the generated samples to identify hallucinations. The key metric for our analysis will be the trajectory variance during the reverse diffusion process, which we hypothesize can effectively detect hallucinations. We expect to demonstrate that diffusion"
            }
        },
        "author_data": {
            "8e51f4b5-1a3a-4039-833f-1afcf9178955": {
                "pk": "8e51f4b5-1a3a-4039-833f-1afcf9178955",
                "name": "Sumukh K Aithal",
                "collaborators": [
                    "Harsh Rangwani",
                    "R. Venkatesh Babu",
                    "Arihant Jain",
                    "Mayank Mishra"
                ],
                "domain": [
                    "Domain Adaptation",
                    "Class Imbalance",
                    "Neural Networks",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "S$^3$VAADA: Submodular Subset Selection for Virtual Adversarial Active Domain Adaptation",
                        "abstract": "Unsupervised domain adaptation (DA) methods have focused on achieving maximal performance through aligning features from source and target domains without using labeled data in the target domain. Whereas, in the real-world scenario's it might be feasible to get labels for a small proportion of target data. In these scenarios, it is important to select maximally-informative samples to label and find an effective way to combine them with the existing knowledge from source data. Towards achieving this, we propose S$^3$VAADA which i) introduces a novel submodular criterion to select a maximally informative subset to label and ii) enhances a cluster-based DA procedure through novel improvements to effectively utilize all the available data for improving generalization on target. Our approach consistently outperforms the competing state-of-the-art approaches on datasets with varying degrees of domain shifts."
                    },
                    {
                        "title": "A Closer Look at Smoothness in Domain Adversarial Training",
                        "abstract": "Domain adversarial training has been ubiquitous for achieving invariant representations and is used widely for various domain adaptation tasks. In recent times, methods converging to smooth optima have shown improved generalization for supervised learning tasks like classification. In this work, we analyze the effect of smoothness enhancing formulations on domain adversarial training, the objective of which is a combination of task loss (eg. classification, regression, etc.) and adversarial terms. We find that converging to a smooth minima with respect to (w.r.t.) task loss stabilizes the adversarial training leading to better performance on target domain. In contrast to task loss, our analysis shows that converging to smooth minima w.r.t. adversarial loss leads to sub-optimal generalization on the target domain. Based on the analysis, we introduce the Smooth Domain Adversarial Training (SDAT) procedure, which effectively enhances the performance of existing domain adversarial methods for both classification and object detection tasks. Our analysis also provides insight into the extensive usage of SGD over Adam in the community for domain adversarial training."
                    },
                    {
                        "title": "Escaping Saddle Points for Effective Generalization on Class-Imbalanced Data",
                        "abstract": "Real-world datasets exhibit imbalances of varying types and degrees. Several techniques based on re-weighting and margin adjustment of loss are often used to enhance the performance of neural networks, particularly on minority classes. In this work, we analyze the class-imbalanced learning problem by examining the loss landscape of neural networks trained with re-weighting and margin-based techniques. Specifically, we examine the spectral density of Hessian of class-wise loss, through which we observe that the network weights converge to a saddle point in the loss landscapes of minority classes. Following this observation, we also find that optimization methods designed to escape from saddle points can be effectively used to improve generalization on minority classes. We further theoretically and empirically demonstrate that Sharpness-Aware Minimization (SAM), a recent technique that encourages convergence to a flat minima, can be effectively used to escape saddle points for minority classes. Using SAM results in a 6.2\\% increase in accuracy on the minority classes over the state-of-the-art Vector Scaling Loss, leading to an overall average increase of 4\\% across imbalanced datasets. The code is available at: https://github.com/val-iisc/Saddle-LongTail."
                    }
                ]
            },
            "76dd2187-699a-4736-82e7-edbefcd5a43b": {
                "pk": "76dd2187-699a-4736-82e7-edbefcd5a43b",
                "name": "Pratyush Maini",
                "collaborators": [
                    "J. Zico Kolter",
                    "Zachary C. Lipton",
                    "Nicolas Papernot",
                    "Danish Pruthi",
                    "Sachin Goyal",
                    "Aditi Raghunathan",
                    "Keshav Kolluru",
                    "Mausam",
                    "Eric Wong",
                    "Mohammad Yaghini"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Adversarial Machine Learning",
                    "Model Extraction",
                    "Data Curation"
                ],
                "publications": [
                    {
                        "title": "Why and when should you pool? Analyzing Pooling in Recurrent Architectures",
                        "abstract": "Pooling-based recurrent neural architectures consistently outperform their counterparts without pooling. However, the reasons for their enhanced performance are largely unexamined. In this work, we examine three commonly used pooling techniques (mean-pooling, max-pooling, and attention), and propose max-attention, a novel variant that effectively captures interactions among predictive tokens in a sentence. We find that pooling-based architectures substantially differ from their non-pooling equivalents in their learning ability and positional biases--which elucidate their performance benefits. By analyzing the gradient propagation, we discover that pooling facilitates better gradient flow compared to BiLSTMs. Further, we expose how BiLSTMs are positionally biased towards tokens in the beginning and the end of a sequence. Pooling alleviates such biases. Consequently, we identify settings where pooling offers large benefits: (i) in low resource scenarios, and (ii) when important words lie towards the middle of the sentence. Among the pooling techniques studied, max-attention is the most effective, resulting in significant performance gains on several text classification tasks."
                    },
                    {
                        "title": "Adversarial Robustness Against the Union of Multiple Perturbation Models",
                        "abstract": "Owing to the susceptibility of deep learning systems to adversarial attacks, there has been a great deal of work in developing (both empirically and certifiably) robust classifiers. While most work has defended against a single type of attack, recent work has looked at defending against multiple perturbation models using simple aggregations of multiple attacks. However, these methods can be difficult to tune, and can easily result in imbalanced degrees of robustness to individual perturbation models, resulting in a sub-optimal worst-case loss over the union. In this work, we develop a natural generalization of the standard PGD-based procedure to incorporate multiple perturbation models into a single attack, by taking the worst-case over all steepest descent directions. This approach has the advantage of directly converging upon a trade-off between different perturbation models which minimizes the worst-case performance over the union. With this approach, we are able to train standard architectures which are simultaneously robust against $\\ell_\\infty$, $\\ell_2$, and $\\ell_1$ attacks, outperforming past approaches on the MNIST and CIFAR10 datasets and achieving adversarial accuracy of 47.0% against the union of ($\\ell_\\infty$, $\\ell_2$, $\\ell_1$) perturbations with radius = (0.03, 0.5, 12) on the latter, improving upon previous approaches which achieve 40.6% accuracy."
                    },
                    {
                        "title": "Dataset Inference: Ownership Resolution in Machine Learning",
                        "abstract": "With increasingly more data and computation involved in their training, machine learning models constitute valuable intellectual property. This has spurred interest in model stealing, which is made more practical by advances in learning with partial, little, or no supervision. Existing defenses focus on inserting unique watermarks in a model's decision surface, but this is insufficient: the watermarks are not sampled from the training distribution and thus are not always preserved during model stealing. In this paper, we make the key observation that knowledge contained in the stolen model's training set is what is common to all stolen copies. The adversary's goal, irrespective of the attack employed, is always to extract this knowledge or its by-products. This gives the original model's owner a strong advantage over the adversary: model owners have access to the original training data. We thus introduce $dataset$ $inference$, the process of identifying whether a suspected model copy has private knowledge from the original model's dataset, as a defense against model stealing. We develop an approach for dataset inference that combines statistical testing with the ability to estimate the distance of multiple data points to the decision boundary. Our experiments on CIFAR10, SVHN, CIFAR100 and ImageNet show that model owners can claim with confidence greater than 99% that their model (or dataset as a matter of fact) was stolen, despite only exposing 50 of the stolen model's training points. Dataset inference defends against state-of-the-art attacks even when the adversary is adaptive. Unlike prior work, it does not require retraining or overfitting the defended model."
                    },
                    {
                        "title": "LLM Dataset Inference: Did you train on my dataset?",
                        "abstract": "The proliferation of large language models (LLMs) in the real world has come with a rise in copyright cases against companies for training their models on unlicensed data from the internet. Recent works have presented methods to identify if individual text sequences were members of the model's training data, known as membership inference attacks (MIAs). We demonstrate that the apparent success of these MIAs is confounded by selecting non-members (text sequences not used for training) belonging to a different distribution from the members (e.g., temporally shifted recent Wikipedia articles compared with ones used to train the model). This distribution shift makes membership inference appear successful. However, most MIA methods perform no better than random guessing when discriminating between members and non-members from the same distribution (e.g., in this case, the same period of time). Even when MIAs work, we find that different MIAs succeed at inferring membership of samples from different distributions. Instead, we propose a new dataset inference method to accurately identify the datasets used to train large language models. This paradigm sits realistically in the modern-day copyright landscape, where authors claim that an LLM is trained over multiple documents (such as a book) written by them, rather than one particular paragraph. While dataset inference shares many of the challenges of membership inference, we solve it by selectively combining the MIAs that provide positive signal for a given distribution, and aggregating them to perform a statistical test on a given dataset. Our approach successfully distinguishes the train and test sets of different subsets of the Pile with statistically significant p-values < 0.1, without any false positives."
                    },
                    {
                        "title": "Data-Free Model Extraction",
                        "abstract": "Current model extraction attacks assume that the adversary has access to a surrogate dataset with characteristics similar to the proprietary data used to train the victim model. This requirement precludes the use of existing model extraction techniques on valuable models, such as those trained on rare or hard to acquire datasets. In contrast, we propose data-free model extraction methods that do not require a surrogate dataset. Our approach adapts techniques from the area of data-free knowledge transfer for model extraction. As part of our study, we identify that the choice of loss is critical to ensuring that the extracted model is an accurate replica of the victim model. Furthermore, we address difficulties arising from the adversary's limited access to the victim model in a black-box setting. For example, we recover the model's logits from its probability predictions to approximate gradients. We find that the proposed data-free model extraction approach achieves high-accuracy with reasonable query complexity -- 0.99x and 0.92x the victim model accuracy on SVHN and CIFAR-10 datasets given 2M and 20M queries respectively."
                    },
                    {
                        "title": "Characterizing Datapoints via Second-Split Forgetting",
                        "abstract": "Researchers investigating example hardness have increasingly focused on the dynamics by which neural networks learn and forget examples throughout training. Popular metrics derived from these dynamics include (i) the epoch at which examples are first correctly classified; (ii) the number of times their predictions flip during training; and (iii) whether their prediction flips if they are held out. However, these metrics do not distinguish among examples that are hard for distinct reasons, such as membership in a rare subpopulation, being mislabeled, or belonging to a complex subpopulation. In this paper, we propose $second$-$split$ $forgetting$ $time$ (SSFT), a complementary metric that tracks the epoch (if any) after which an original training example is forgotten as the network is fine-tuned on a randomly held out partition of the data. Across multiple benchmark datasets and modalities, we demonstrate that $mislabeled$ examples are forgotten quickly, and seemingly $rare$ examples are forgotten comparatively slowly. By contrast, metrics only considering the first split learning dynamics struggle to differentiate the two. At large learning rates, SSFT tends to be robust across architectures, optimizers, and random seeds. From a practical standpoint, the SSFT can (i) help to identify mislabeled samples, the removal of which improves generalization; and (ii) provide insights about failure modes. Through theoretical analysis addressing overparameterized linear models, we provide insights into how the observed phenomena may arise. Code for reproducing our experiments can be found here: https://github.com/pratyushmaini/ssft"
                    },
                    {
                        "title": "Rephrasing the Web: A Recipe for Compute and Data-Efficient Language Modeling",
                        "abstract": "Large language models are trained on massive scrapes of the web, which are often unstructured, noisy, and poorly phrased. Current scaling laws show that learning from such data requires an abundance of both compute and data, which grows with the size of the model being trained. This is infeasible both because of the large compute costs and duration associated with pre-training, and the impending scarcity of high-quality data on the web. In this work, we propose Web Rephrase Augmented Pre-training ($\\textbf{WRAP}$) that uses an off-the-shelf instruction-tuned model prompted to paraphrase documents on the web in specific styles such as \"like Wikipedia\" or in \"question-answer format\" to jointly pre-train LLMs on real and synthetic rephrases. First, we show that using WRAP on the C4 dataset, which is naturally noisy, speeds up pre-training by $\\sim3x$. At the same pre-training compute budget, it improves perplexity by more than 10% on average across different subsets of the Pile, and improves zero-shot question answer accuracy across 13 tasks by more than 2%. Second, we investigate the impact of the re-phrasing style on the performance of the model, offering insights into how the composition of the training data can impact the performance of LLMs in OOD settings. Our gains are attributed to the fact that re-phrased synthetic data has higher utility than just real data because it (i) incorporates style diversity that closely reflects downstream evaluation style, and (ii) has higher 'quality' than web-scraped data."
                    },
                    {
                        "title": "T-MARS: Improving Visual Representations by Circumventing Text Feature Learning",
                        "abstract": "Large web-sourced multimodal datasets have powered a slew of new methods for learning general-purpose visual representations, advancing the state of the art in computer vision and revolutionizing zero- and few-shot recognition. One crucial decision facing practitioners is how, if at all, to curate these ever-larger datasets. For example, the creators of the LAION-5B dataset chose to retain only image-caption pairs whose CLIP similarity score exceeded a designated threshold. In this paper, we propose a new state-of-the-art data filtering approach motivated by our observation that nearly 40% of LAION's images contain text that overlaps significantly with the caption. Intuitively, such data could be wasteful as it incentivizes models to perform optical character recognition rather than learning visual features. However, naively removing all such data could also be wasteful, as it throws away images that contain visual features (in addition to overlapping text). Our simple and scalable approach, T-MARS (Text Masking and Re-Scoring), filters out only those pairs where the text dominates the remaining visual features -- by first masking out the text and then filtering out those with a low CLIP similarity score of the masked image. Experimentally, T-MARS outperforms the top-ranked method on the \"medium scale\" of DataComp (a data filtering benchmark) by a margin of 6.5% on ImageNet and 4.7% on VTAB. Additionally, our systematic evaluation on various data pool sizes from 2M to 64M shows that the accuracy gains enjoyed by T-MARS linearly increase as data and compute are scaled exponentially. Code is available at https://github.com/locuslab/T-MARS."
                    },
                    {
                        "title": "Model-tuning Via Prompts Makes NLP Models Adversarially Robust",
                        "abstract": "In recent years, NLP practitioners have converged on the following practice: (i) import an off-the-shelf pretrained (masked) language model; (ii) append a multilayer perceptron atop the CLS token's hidden representation (with randomly initialized weights); and (iii) fine-tune the entire model on a downstream task (MLP-FT). This procedure has produced massive gains on standard NLP benchmarks, but these models remain brittle, even to mild adversarial perturbations. In this work, we demonstrate surprising gains in adversarial robustness enjoyed by Model-tuning Via Prompts (MVP), an alternative method of adapting to downstream tasks. Rather than appending an MLP head to make output prediction, MVP appends a prompt template to the input, and makes prediction via text infilling/completion. Across 5 NLP datasets, 4 adversarial attacks, and 3 different models, MVP improves performance against adversarial substitutions by an average of 8% over standard methods and even outperforms adversarial training-based state-of-art defenses by 3.5%. By combining MVP with adversarial training, we achieve further improvements in adversarial robustness while maintaining performance on unperturbed examples. Finally, we conduct ablations to investigate the mechanism underlying these gains. Notably, we find that the main causes of vulnerability of MLP-FT can be attributed to the misalignment between pre-training and fine-tuning tasks, and the randomly initialized MLP parameters."
                    },
                    {
                        "title": "TOFU: A Task of Fictitious Unlearning for LLMs",
                        "abstract": "Large language models trained on massive corpora of data from the web can memorize and reproduce sensitive or private data raising both legal and ethical concerns. Unlearning, or tuning models to forget information present in their training data, provides us with a way to protect private data after training. Although several methods exist for such unlearning, it is unclear to what extent they result in models equivalent to those where the data to be forgotten was never learned in the first place. To address this challenge, we present TOFU, a Task of Fictitious Unlearning, as a benchmark aimed at helping deepen our understanding of unlearning. We offer a dataset of 200 diverse synthetic author profiles, each consisting of 20 question-answer pairs, and a subset of these profiles called the forget set that serves as the target for unlearning. We compile a suite of metrics that work together to provide a holistic picture of unlearning efficacy. Finally, we provide a set of baseline results from existing unlearning algorithms. Importantly, none of the baselines we consider show effective unlearning motivating continued efforts to develop approaches for unlearning that effectively tune models so that they truly behave as if they were never trained on the forget data at all."
                    },
                    {
                        "title": "Scaling Laws for Data Filtering -- Data Curation cannot be Compute Agnostic",
                        "abstract": "Vision-language models (VLMs) are trained for thousands of GPU hours on carefully curated web datasets. In recent times, data curation has gained prominence with several works developing strategies to retain 'high-quality' subsets of 'raw' scraped data. For instance, the LAION public dataset retained only 10% of the total crawled data. However, these strategies are typically developed agnostic of the available compute for training. In this paper, we first demonstrate that making filtering decisions independent of training compute is often suboptimal: the limited high-quality data rapidly loses its utility when repeated, eventually requiring the inclusion of 'unseen' but 'lower-quality' data. To address this quality-quantity tradeoff ($\\texttt{QQT}$), we introduce neural scaling laws that account for the non-homogeneous nature of web data, an angle ignored in existing literature. Our scaling laws (i) characterize the $\\textit{differing}$ 'utility' of various quality subsets of web data; (ii) account for how utility diminishes for a data point at its 'nth' repetition; and (iii) formulate the mutual interaction of various data pools when combined, enabling the estimation of model performance on a combination of multiple data pools without ever jointly training on them. Our key message is that data curation $\\textit{cannot}$ be agnostic of the total compute that a model will be trained for. Our scaling laws allow us to curate the best possible pool for achieving top performance on Datacomp at various compute budgets, carving out a pareto-frontier for data curation. Code is available at https://github.com/locuslab/scaling_laws_data_filtering."
                    }
                ]
            },
            "6671aa96-c6c5-464b-83a2-a213f66f5681": {
                "pk": "6671aa96-c6c5-464b-83a2-a213f66f5681",
                "name": "Zachary C. Lipton",
                "collaborators": [
                    "Julian McAuley",
                    "Charles Elkan",
                    "David C. Kale",
                    "Zachary Novack",
                    "Saurabh Garg",
                    "Subarna Tripathi",
                    "Randall C. Wetzel",
                    "Sharad Vikram",
                    "John Alberg",
                    "John Berkowitz"
                ],
                "domain": [
                    "Machine Learning",
                    "Interpretability",
                    "Generative Models",
                    "Active Learning"
                ],
                "publications": [
                    {
                        "title": "Stuck in a What? Adventures in Weight Space",
                        "abstract": "Deep learning researchers commonly suggest that converged models are stuck in local minima. More recently, some researchers observed that under reasonable assumptions, the vast majority of critical points are saddle points, not true minima. Both descriptions suggest that weights converge around a point in weight space, be it a local optima or merely a critical point. However, it's possible that neither interpretation is accurate. As neural networks are typically over-complete, it's easy to show the existence of vast continuous regions through weight space with equal loss. In this paper, we build on recent work empirically characterizing the error surfaces of neural networks. We analyze training paths through weight space, presenting evidence that apparent convergence of loss does not correspond to weights arriving at critical points, but instead to large movements through flat regions of weight space. While it's trivial to show that neural network error surfaces are globally non-convex, we show that error surfaces are also locally non-convex, even after breaking symmetry with a random initialization and also after partial training."
                    },
                    {
                        "title": "The Mythos of Model Interpretability",
                        "abstract": "Supervised machine learning models boast remarkable predictive capabilities. But can you trust your model? Will it work in deployment? What else can it tell you about the world? We want models to be not only good, but interpretable. And yet the task of interpretation appears underspecified. Papers provide diverse and sometimes non-overlapping motivations for interpretability, and offer myriad notions of what attributes render models interpretable. Despite this ambiguity, many papers proclaim interpretability axiomatically, absent further explanation. In this paper, we seek to refine the discourse on interpretability. First, we examine the motivations underlying interest in interpretability, finding them to be diverse and occasionally discordant. Then, we address model properties and techniques thought to confer interpretability, identifying transparency to humans and post-hoc explanations as competing notions. Throughout, we discuss the feasibility and desirability of different notions, and question the oft-made assertions that linear models are interpretable and that deep neural networks are not."
                    },
                    {
                        "title": "The Doctor Just Won't Accept That!",
                        "abstract": "Calls to arms to build interpretable models express a well-founded discomfort with machine learning. Should a software agent that does not even know what a loan is decide who qualifies for one? Indeed, we ought to be cautious about injecting machine learning (or anything else, for that matter) into applications where there may be a significant risk of causing social harm. However, claims that stakeholders \"just won't accept that!\" do not provide a sufficient foundation for a proposed field of study. For the field of interpretable machine learning to advance, we must ask the following questions: What precisely won't various stakeholders accept? What do they want? Are these desiderata reasonable? Are they feasible? In order to answer these questions, we'll have to give real-world problems and their respective stakeholders greater consideration."
                    },
                    {
                        "title": "Precise Recovery of Latent Vectors from Generative Adversarial Networks",
                        "abstract": "Generative adversarial networks (GANs) transform latent vectors into visually plausible images. It is generally thought that the original GAN formulation gives no out-of-the-box method to reverse the mapping, projecting images back into latent space. We introduce a simple, gradient-based technique called stochastic clipping. In experiments, for images generated by the GAN, we precisely recover their latent vector pre-images 100% of the time. Additional experiments demonstrate that this method is robust to noise. Finally, we show that even for unseen images, our method appears to recover unique encodings."
                    },
                    {
                        "title": "Phenotyping of Clinical Time Series with LSTM Recurrent Neural Networks",
                        "abstract": "We present a novel application of LSTM recurrent neural networks to multilabel classification of diagnoses given variable-length time series of clinical measurements. Our method outperforms a strong baseline on a variety of metrics."
                    },
                    {
                        "title": "CHiLS: Zero-Shot Image Classification with Hierarchical Label Sets",
                        "abstract": "Open vocabulary models (e.g. CLIP) have shown strong performance on zero-shot classification through their ability generate embeddings for each class based on their (natural language) names. Prior work has focused on improving the accuracy of these models through prompt engineering or by incorporating a small amount of labeled downstream data (via finetuning). However, there has been little focus on improving the richness of the class names themselves, which can pose issues when class labels are coarsely-defined and are uninformative. We propose Classification with Hierarchical Label Sets (or CHiLS), an alternative strategy for zero-shot classification specifically designed for datasets with implicit semantic hierarchies. CHiLS proceeds in three steps: (i) for each class, produce a set of subclasses, using either existing label hierarchies or by querying GPT-3; (ii) perform the standard zero-shot CLIP procedure as though these subclasses were the labels of interest; (iii) map the predicted subclass back to its parent to produce the final prediction. Across numerous datasets with underlying hierarchical structure, CHiLS leads to improved accuracy in situations both with and without ground-truth hierarchical information. CHiLS is simple to implement within existing zero-shot pipelines and requires no additional training cost. Code is available at: https://github.com/acmi-lab/CHILS."
                    },
                    {
                        "title": "Generative Concatenative Nets Jointly Learn to Write and Classify Reviews",
                        "abstract": "A recommender system's basic task is to estimate how users will respond to unseen items. This is typically modeled in terms of how a user might rate a product, but here we aim to extend such approaches to model how a user would write about the product. To do so, we design a character-level Recurrent Neural Network (RNN) that generates personalized product reviews. The network convincingly learns styles and opinions of nearly 1000 distinct authors, using a large corpus of reviews from BeerAdvocate.com. It also tailors reviews to describe specific items, categories, and star ratings. Using a simple input replication strategy, the Generative Concatenative Network (GCN) preserves the signal of static auxiliary inputs across wide sequence intervals. Without any additional training, the generative model can classify reviews, identifying the author of the review, the product category, and the sentiment (rating), with remarkable accuracy. Our evaluation shows the GCN captures complex dynamics in text, such as the effect of negation, misspellings, slang, and large vocabularies gracefully absent any machinery explicitly dedicated to the purpose."
                    },
                    {
                        "title": "Efficient Elastic Net Regularization for Sparse Linear Models",
                        "abstract": "This paper presents an algorithm for efficient training of sparse linear models with elastic net regularization. Extending previous work on delayed updates, the new algorithm applies stochastic gradient updates to non-zero features only, bringing weights current as needed with closed-form updates. Closed-form delayed updates for the $\\ell_1$, $\\ell_{\\infty}$, and rarely used $\\ell_2$ regularizers have been described previously. This paper provides closed-form updates for the popular squared norm $\\ell^2_2$ and elastic net regularizers.   We provide dynamic programming algorithms that perform each delayed update in constant time. The new $\\ell^2_2$ and elastic net methods handle both fixed and varying learning rates, and both standard {stochastic gradient descent} (SGD) and {forward backward splitting (FoBoS)}. Experimental results show that on a bag-of-words dataset with $260,941$ features, but only $88$ nonzero features on average per training example, the dynamic programming method trains a logistic regression classifier with elastic net regularization over $2000$ times faster than otherwise."
                    },
                    {
                        "title": "Improving Factor-Based Quantitative Investing by Forecasting Company Fundamentals",
                        "abstract": "On a periodic basis, publicly traded companies are required to report fundamentals: financial data such as revenue, operating income, debt, among others. These data points provide some insight into the financial health of a company. Academic research has identified some factors, i.e. computed features of the reported data, that are known through retrospective analysis to outperform the market average. Two popular factors are the book value normalized by market capitalization (book-to-market) and the operating income normalized by the enterprise value (EBIT/EV). In this paper: we first show through simulation that if we could (clairvoyantly) select stocks using factors calculated on future fundamentals (via oracle), then our portfolios would far outperform a standard factor approach. Motivated by this analysis, we train deep neural networks to forecast future fundamentals based on a trailing 5-years window. Quantitative analysis demonstrates a significant improvement in MSE over a naive strategy. Moreover, in retrospective analysis using an industry-grade stock portfolio simulator (backtester), we show an improvement in compounded annual return to 17.1% (MLP) vs 14.4% for a standard factor model."
                    },
                    {
                        "title": "A Critical Review of Recurrent Neural Networks for Sequence Learning",
                        "abstract": "Countless learning tasks require dealing with sequential data. Image captioning, speech synthesis, and music generation all require that a model produce outputs that are sequences. In other domains, such as time series prediction, video analysis, and musical information retrieval, a model must learn from inputs that are sequences. Interactive tasks, such as translating natural language, engaging in dialogue, and controlling a robot, often demand both capabilities. Recurrent neural networks (RNNs) are connectionist models that capture the dynamics of sequences via cycles in the network of nodes. Unlike standard feedforward neural networks, recurrent networks retain a state that can represent information from an arbitrarily long context window. Although recurrent neural networks have traditionally been difficult to train, and often contain millions of parameters, recent advances in network architectures, optimization techniques, and parallel computation have enabled successful large-scale learning with them. In recent years, systems based on long short-term memory (LSTM) and bidirectional (BRNN) architectures have demonstrated ground-breaking performance on tasks as varied as image captioning, language translation, and handwriting recognition. In this survey, we review and synthesize the research that over the past three decades first yielded and then made practical these powerful learning models. When appropriate, we reconcile conflicting notation and nomenclature. Our goal is to provide a self-contained explication of the state of the art together with a historical perspective and references to primary research."
                    },
                    {
                        "title": "Does mitigating ML's impact disparity require treatment disparity?",
                        "abstract": "Following related work in law and policy, two notions of disparity have come to shape the study of fairness in algorithmic decision-making. Algorithms exhibit treatment disparity if they formally treat members of protected subgroups differently; algorithms exhibit impact disparity when outcomes differ across subgroups, even if the correlation arises unintentionally. Naturally, we can achieve impact parity through purposeful treatment disparity. In one thread of technical work, papers aim to reconcile the two forms of parity proposing disparate learning processes (DLPs). Here, the learning algorithm can see group membership during training but produce a classifier that is group-blind at test time. In this paper, we show theoretically that: (i) When other features correlate to group membership, DLPs will (indirectly) implement treatment disparity, undermining the policy desiderata they are designed to address; (ii) When group membership is partly revealed by other features, DLPs induce within-class discrimination; and (iii) In general, DLPs provide a suboptimal trade-off between accuracy and impact parity. Based on our technical analysis, we argue that transparent treatment disparity is preferable to occluded methods for achieving impact parity. Experimental results on several real-world datasets highlight the practical consequences of applying DLPs vs. per-group thresholds."
                    },
                    {
                        "title": "Troubling Trends in Machine Learning Scholarship",
                        "abstract": "Collectively, machine learning (ML) researchers are engaged in the creation and dissemination of knowledge about data-driven algorithms. In a given paper, researchers might aspire to any subset of the following goals, among others: to theoretically characterize what is learnable, to obtain understanding through empirically rigorous experiments, or to build a working system that has high predictive accuracy. While determining which knowledge warrants inquiry may be subjective, once the topic is fixed, papers are most valuable to the community when they act in service of the reader, creating foundational knowledge and communicating as clearly as possible.   Recent progress in machine learning comes despite frequent departures from these ideals. In this paper, we focus on the following four patterns that appear to us to be trending in ML scholarship: (i) failure to distinguish between explanation and speculation; (ii) failure to identify the sources of empirical gains, e.g., emphasizing unnecessary modifications to neural architectures when gains actually stem from hyper-parameter tuning; (iii) mathiness: the use of mathematics that obfuscates or impresses rather than clarifies, e.g., by confusing technical and non-technical concepts; and (iv) misuse of language, e.g., by choosing terms of art with colloquial connotations or by overloading established technical terms.   While the causes behind these patterns are uncertain, possibilities include the rapid expansion of the community, the consequent thinness of the reviewer pool, and the often-misaligned incentives between scholarship and short-term measures of success (e.g., bibliometrics, attention, and entrepreneurial opportunity). While each pattern offers a corresponding remedy (don't do it), we also discuss some speculative suggestions for how the community might combat these trends."
                    },
                    {
                        "title": "Modeling Missing Data in Clinical Time Series with RNNs",
                        "abstract": "We demonstrate a simple strategy to cope with missing data in sequential inputs, addressing the task of multilabel classification of diagnoses given clinical time series. Collected from the pediatric intensive care unit (PICU) at Children's Hospital Los Angeles, our data consists of multivariate time series of observations. The measurements are irregularly spaced, leading to missingness patterns in temporally discretized sequences. While these artifacts are typically handled by imputation, we achieve superior predictive performance by treating the artifacts as features. Unlike linear models, recurrent neural networks can realize this improvement using only simple binary indicators of missingness. For linear models, we show an alternative strategy to capture this signal. Training models on missingness patterns only, we show that for some diseases, what tests are run can be as predictive as the results themselves."
                    },
                    {
                        "title": "Detecting and Correcting for Label Shift with Black Box Predictors",
                        "abstract": "Faced with distribution shift between training and test set, we wish to detect and quantify the shift, and to correct our classifiers without test set labels. Motivated by medical diagnosis, where diseases (targets) cause symptoms (observations), we focus on label shift, where the label marginal $p(y)$ changes but the conditional $p(x| y)$ does not. We propose Black Box Shift Estimation (BBSE) to estimate the test distribution $p(y)$. BBSE exploits arbitrary black box predictors to reduce dimensionality prior to shift correction. While better predictors give tighter estimates, BBSE works even when predictors are biased, inaccurate, or uncalibrated, so long as their confusion matrices are invertible. We prove BBSE's consistency, bound its error, and introduce a statistical test that uses BBSE to detect shift. We also leverage BBSE to correct classifiers. Experiments demonstrate accurate estimates and improved prediction, even on high-dimensional datasets of natural images."
                    },
                    {
                        "title": "Entity Projection via Machine Translation for Cross-Lingual NER",
                        "abstract": "Although over 100 languages are supported by strong off-the-shelf machine translation systems, only a subset of them possess large annotated corpora for named entity recognition. Motivated by this fact, we leverage machine translation to improve annotation-projection approaches to cross-lingual named entity recognition. We propose a system that improves over prior entity-projection methods by: (a) leveraging machine translation systems twice: first for translating sentences and subsequently for translating entities; (b) matching entities based on orthographic and phonetic similarity; and (c) identifying matches based on distributional statistics derived from the dataset. Our approach improves upon current state-of-the-art methods for cross-lingual named entity recognition on 5 diverse languages by an average of 4.1 points. Further, our method achieves state-of-the-art F_1 scores for Armenian, outperforming even a monolingual model trained on Armenian source data."
                    },
                    {
                        "title": "Algorithmic Fairness from a Non-ideal Perspective",
                        "abstract": "Inspired by recent breakthroughs in predictive modeling, practitioners in both industry and government have turned to machine learning with hopes of operationalizing predictions to drive automated decisions. Unfortunately, many social desiderata concerning consequential decisions, such as justice or fairness, have no natural formulation within a purely predictive framework. In efforts to mitigate these problems, researchers have proposed a variety of metrics for quantifying deviations from various statistical parities that we might expect to observe in a fair world and offered a variety of algorithms in attempts to satisfy subsets of these parities or to trade off the degree to which they are satisfied against utility. In this paper, we connect this approach to \\emph{fair machine learning} to the literature on ideal and non-ideal methodological approaches in political philosophy. The ideal approach requires positing the principles according to which a just world would operate. In the most straightforward application of ideal theory, one supports a proposed policy by arguing that it closes a discrepancy between the real and the perfectly just world. However, by failing to account for the mechanisms by which our non-ideal world arose, the responsibilities of various decision-makers, and the impacts of proposed policies, naive applications of ideal thinking can lead to misguided interventions. In this paper, we demonstrate a connection between the fair machine learning literature and the ideal approach in political philosophy, and argue that the increasingly apparent shortcomings of proposed fair machine learning algorithms reflect broader troubles faced by the ideal approach. We conclude with a critical discussion of the harms of misguided solutions, a reinterpretation of impossibility results, and directions for future research."
                    },
                    {
                        "title": "Disentangling the Mechanisms Behind Implicit Regularization in SGD",
                        "abstract": "A number of competing hypotheses have been proposed to explain why small-batch Stochastic Gradient Descent (SGD)leads to improved generalization over the full-batch regime, with recent work crediting the implicit regularization of various quantities throughout training. However, to date, empirical evidence assessing the explanatory power of these hypotheses is lacking. In this paper, we conduct an extensive empirical evaluation, focusing on the ability of various theorized mechanisms to close the small-to-large batch generalization gap. Additionally, we characterize how the quantities that SGD has been claimed to (implicitly) regularize change over the course of training. By using micro-batches, i.e. disjoint smaller subsets of each mini-batch, we empirically show that explicitly penalizing the gradient norm or the Fisher Information Matrix trace, averaged over micro-batches, in the large-batch regime recovers small-batch SGD generalization, whereas Jacobian-based regularizations fail to do so. This generalization performance is shown to often be correlated with how well the regularized model's gradient norms resemble those of small-batch SGD. We additionally show that this behavior breaks down as the micro-batch size approaches the batch size. Finally, we note that in this line of inquiry, positive experimental findings on CIFAR10 are often reversed on other datasets like CIFAR100, highlighting the need to test hypotheses on a wider collection of datasets."
                    },
                    {
                        "title": "Differential Privacy and Machine Learning: a Survey and Review",
                        "abstract": "The objective of machine learning is to extract useful information from data, while privacy is preserved by concealing information. Thus it seems hard to reconcile these competing interests. However, they frequently must be balanced when mining sensitive data. For example, medical research represents an important application where it is necessary both to extract useful information and protect patient privacy. One way to resolve the conflict is to extract general characteristics of whole populations without disclosing the private information of individuals.   In this paper, we consider differential privacy, one of the most popular and powerful definitions of privacy. We explore the interplay between machine learning and differential privacy, namely privacy-preserving machine learning algorithms and learning-based data release mechanisms. We also describe some theoretical results that address what can be learned differentially privately and upper bounds of loss functions for differentially private algorithms.   Finally, we present some open questions, including how to incorporate public data, how to deal with missing data in private datasets, and whether, as the number of observed samples grows arbitrarily large, differentially private machine learning algorithms can be achieved at no cost to utility as compared to corresponding non-differentially private algorithms."
                    },
                    {
                        "title": "How Much Reading Does Reading Comprehension Require? A Critical Investigation of Popular Benchmarks",
                        "abstract": "Many recent papers address reading comprehension, where examples consist of (question, passage, answer) tuples. Presumably, a model must combine information from both questions and passages to predict corresponding answers. However, despite intense interest in the topic, with hundreds of published papers vying for leaderboard dominance, basic questions about the difficulty of many popular benchmarks remain unanswered. In this paper, we establish sensible baselines for the bAbI, SQuAD, CBT, CNN, and Who-did-What datasets, finding that question- and passage-only models often perform surprisingly well. On $14$ out of $20$ bAbI tasks, passage-only models achieve greater than $50\\%$ accuracy, sometimes matching the full model. Interestingly, while CBT provides $20$-sentence stories only the last is needed for comparably accurate prediction. By comparison, SQuAD and CNN appear better-constructed."
                    },
                    {
                        "title": "Deep Bayesian Active Learning for Natural Language Processing: Results of a Large-Scale Empirical Study",
                        "abstract": "Several recent papers investigate Active Learning (AL) for mitigating the data dependence of deep learning for natural language processing. However, the applicability of AL to real-world problems remains an open question. While in supervised learning, practitioners can try many different methods, evaluating each against a validation set before selecting a model, AL affords no such luxury. Over the course of one AL run, an agent annotates its dataset exhausting its labeling budget. Thus, given a new task, an active learner has no opportunity to compare models and acquisition functions. This paper provides a large scale empirical study of deep active learning, addressing multiple tasks and, for each, multiple datasets, multiple models, and a full suite of acquisition functions. We find that across all settings, Bayesian active learning by disagreement, using uncertainty estimates provided either by Dropout or Bayes-by Backprop significantly improves over i.i.d. baselines and usually outperforms classic uncertainty sampling."
                    }
                ]
            },
            "f8cda715-0123-44cf-bb42-f248ca0e4a6a": {
                "pk": "f8cda715-0123-44cf-bb42-f248ca0e4a6a",
                "name": "J. Zico Kolter",
                "collaborators": [
                    "Matt Wytock",
                    "Zhili Feng",
                    "Vaishnavh Nagarajan",
                    "Asher Trockman",
                    "Alnur Ali",
                    "Ryan J. Tibshirani",
                    "Brandon Amos",
                    "Vladlen Koltun",
                    "Joseph Szurley",
                    "Michal Moshkovitz"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Time Series Analysis",
                    "Neural Networks",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "Perceptual Based Adversarial Audio Attacks",
                        "abstract": "Recent work has shown the possibility of adversarial attacks on automatic speechrecognition (ASR) systems. However, in the vast majority of work in this area, theattacks have been executed only in the digital space, or have involved short phrasesand static room settings. In this paper, we demonstrate a physically realizableaudio adversarial attack. We base our approach specifically on a psychoacoustic-property-based loss function, and automated generation of room impulse responses, to create adversarial attacks that are robust when played over a speaker in multiple environments. We show that such attacks are possible even while being virtually imperceptible to listeners."
                    },
                    {
                        "title": "Probabilistic Segmentation via Total Variation Regularization",
                        "abstract": "We present a convex approach to probabilistic segmentation and modeling of time series data. Our approach builds upon recent advances in multivariate total variation regularization, and seeks to learn a separate set of parameters for the distribution over the observations at each time point, but with an additional penalty that encourages the parameters to remain constant over time. We propose efficient optimization methods for solving the resulting (large) optimization problems, and a two-stage procedure for estimating recurring clusters under such models, based upon kernel density estimation. Finally, we show on a number of real-world segmentation tasks, the resulting methods often perform as well or better than existing latent variable models, while being substantially easier to train."
                    },
                    {
                        "title": "On the Neural Tangent Kernel of Equilibrium Models",
                        "abstract": "This work studies the neural tangent kernel (NTK) of the deep equilibrium (DEQ) model, a practical ``infinite-depth'' architecture which directly computes the infinite-depth limit of a weight-tied network via root-finding. Even though the NTK of a fully-connected neural network can be stochastic if its width and depth both tend to infinity simultaneously, we show that contrarily a DEQ model still enjoys a deterministic NTK despite its width and depth going to infinity at the same time under mild conditions. Moreover, this deterministic NTK can be found efficiently via root-finding."
                    },
                    {
                        "title": "An Axiomatic Approach to Model-Agnostic Concept Explanations",
                        "abstract": "Concept explanation is a popular approach for examining how human-interpretable concepts impact the predictions of a model. However, most existing methods for concept explanations are tailored to specific models. To address this issue, this paper focuses on model-agnostic measures. Specifically, we propose an approach to concept explanations that satisfy three natural axioms: linearity, recursivity, and similarity. We then establish connections with previous concept explanation methods, offering insight into their varying semantic meanings. Experimentally, we demonstrate the utility of the new method by applying it in different scenarios: for model selection, optimizer selection, and model improvement using a kind of prompt editing for zero-shot vision language models."
                    },
                    {
                        "title": "Generalization in Deep Networks: The Role of Distance from Initialization",
                        "abstract": "Why does training deep neural networks using stochastic gradient descent (SGD) result in a generalization error that does not worsen with the number of parameters in the network? To answer this question, we advocate a notion of effective model capacity that is dependent on {\\em a given random initialization of the network} and not just the training algorithm and the data distribution. We provide empirical evidences that demonstrate that the model capacity of SGD-trained deep networks is in fact restricted through implicit regularization of {\\em the $\\ell_2$ distance from the initialization}. We also provide theoretical arguments that further highlight the need for initialization-dependent notions of model capacity. We leave as open questions how and why distance from initialization is regularized, and whether it is sufficient to explain generalization."
                    },
                    {
                        "title": "Learning Stable Deep Dynamics Models",
                        "abstract": "Deep networks are commonly used to model dynamical systems, predicting how the state of a system will evolve over time (either autonomously or in response to control inputs). Despite the predictive power of these systems, it has been difficult to make formal claims about the basic properties of the learned systems. In this paper, we propose an approach for learning dynamical systems that are guaranteed to be stable over the entire state space. The approach works by jointly learning a dynamics model and Lyapunov function that guarantees non-expansiveness of the dynamics under the learned Lyapunov function. We show that such learning systems are able to model simple dynamical systems and can be combined with additional deep generative models to learn complex dynamics, such as video textures, in a fully end-to-end fashion."
                    },
                    {
                        "title": "Monotone operator equilibrium networks",
                        "abstract": "Implicit-depth models such as Deep Equilibrium Networks have recently been shown to match or exceed the performance of traditional deep networks while being much more memory efficient. However, these models suffer from unstable convergence to a solution and lack guarantees that a solution exists. On the other hand, Neural ODEs, another class of implicit-depth models, do guarantee existence of a unique solution but perform poorly compared with traditional networks. In this paper, we develop a new class of implicit-depth model based on the theory of monotone operators, the Monotone Operator Equilibrium Network (monDEQ). We show the close connection between finding the equilibrium point of an implicit network and solving a form of monotone operator splitting problem, which admits efficient solvers with guaranteed, stable convergence. We then develop a parameterization of the network which ensures that all operators remain monotone, which guarantees the existence of a unique equilibrium point. Finally, we show how to instantiate several versions of these models, and implement the resulting iterative solvers, for structured linear operators such as multi-scale convolutions. The resulting models vastly outperform the Neural ODE-based models while also being more computationally efficient. Code is available at http://github.com/locuslab/monotone_op_net."
                    },
                    {
                        "title": "Gradient descent GAN optimization is locally stable",
                        "abstract": "Despite the growing prominence of generative adversarial networks (GANs), optimization in GANs is still a poorly understood topic. In this paper, we analyze the \"gradient descent\" form of GAN optimization i.e., the natural setting where we simultaneously take small gradient steps in both generator and discriminator parameters. We show that even though GAN optimization does not correspond to a convex-concave game (even for simple parameterizations), under proper conditions, equilibrium points of this optimization procedure are still \\emph{locally asymptotically stable} for the traditional GAN formulation. On the other hand, we show that the recently proposed Wasserstein GAN can have non-convergent limit cycles near equilibrium. Motivated by this stability analysis, we propose an additional regularization term for gradient descent GAN updates, which \\emph{is} able to guarantee local stability for both the WGAN and the traditional GAN, and also shows practical promise in speeding up convergence and addressing mode collapse."
                    },
                    {
                        "title": "Dynamic Modeling and Equilibria in Fair Decision Making",
                        "abstract": "Recent studies on fairness in automated decision making systems have both investigated the potential future impact of these decisions on the population at large, and emphasized that imposing ''typical'' fairness constraints such as demographic parity or equality of opportunity does not guarantee a benefit to disadvantaged groups. However, these previous studies have focused on either simple one-step cost/benefit criteria, or on discrete underlying state spaces. In this work, we first propose a natural continuous representation of population state, governed by the Beta distribution, using a loan granting setting as a running example. Next, we apply a model of population dynamics under lending decisions, and show that when conditional payback probabilities are estimated correctly 1) ``optimal'' behavior by lenders can lead to ''Matthew Effect'' bifurcations (i.e., ''the rich get richer and the poor get poorer''), but that 2) many common fairness constraints on the allowable policies cause groups to converge to the same equilibrium point. Last, we contrast our results in the case of misspecified conditional probability estimates with prior work, and show that for this model, different levels of group misestimation guarantees that even fair policies lead to bifurcations. We illustrate some of the modeling conclusions on real data from credit scoring."
                    },
                    {
                        "title": "Orthogonalizing Convolutional Layers with the Cayley Transform",
                        "abstract": "Recent work has highlighted several advantages of enforcing orthogonality in the weight layers of deep networks, such as maintaining the stability of activations, preserving gradient norms, and enhancing adversarial robustness by enforcing low Lipschitz constants. Although numerous methods exist for enforcing the orthogonality of fully-connected layers, those for convolutional layers are more heuristic in nature, often focusing on penalty methods or limited classes of convolutions. In this work, we propose and evaluate an alternative approach to directly parameterize convolutional layers that are constrained to be orthogonal. Specifically, we propose to apply the Cayley transform to a skew-symmetric convolution in the Fourier domain, so that the inverse convolution needed by the Cayley transform can be computed efficiently. We compare our method to previous Lipschitz-constrained and orthogonal convolutional layers and show that it indeed preserves orthogonality to a high degree even for large convolutions. Applied to the problem of certified adversarial robustness, we show that networks incorporating the layer outperform existing deterministic methods for certified defense against $\\ell_2$-norm-bounded adversaries, while scaling to larger architectures than previously investigated. Code is available at https://github.com/locuslab/orthogonal-convolutions."
                    },
                    {
                        "title": "Mimetic Initialization of Self-Attention Layers",
                        "abstract": "It is notoriously difficult to train Transformers on small datasets; typically, large pre-trained models are instead used as the starting point. We explore the weights of such pre-trained Transformers (particularly for vision) to attempt to find reasons for this discrepancy. Surprisingly, we find that simply initializing the weights of self-attention layers so that they \"look\" more like their pre-trained counterparts allows us to train vanilla Transformers faster and to higher final accuracies, particularly on vision tasks such as CIFAR-10 and ImageNet classification, where we see gains in accuracy of over 5% and 4%, respectively. Our initialization scheme is closed form, learning-free, and very simple: we set the product of the query and key weights to be approximately the identity, and the product of the value and projection weights to approximately the negative identity. As this mimics the patterns we saw in pre-trained Transformers, we call the technique \"mimetic initialization\"."
                    },
                    {
                        "title": "TorchDEQ: A Library for Deep Equilibrium Models",
                        "abstract": "Deep Equilibrium (DEQ) Models, an emerging class of implicit models that maps inputs to fixed points of neural networks, are of growing interest in the deep learning community. However, training and applying DEQ models is currently done in an ad-hoc fashion, with various techniques spread across the literature. In this work, we systematically revisit DEQs and present TorchDEQ, an out-of-the-box PyTorch-based library that allows users to define, train, and infer using DEQs over multiple domains with minimal code and best practices. Using TorchDEQ, we build a ``DEQ Zoo'' that supports six published implicit models across different domains. By developing a joint framework that incorporates the best practices across all models, we have substantially improved the performance, training stability, and efficiency of DEQs on ten datasets across all six projects in the DEQ Zoo. TorchDEQ and DEQ Zoo are released as \\href{https://github.com/locuslab/torchdeq}{open source}."
                    },
                    {
                        "title": "AcceleratedLiNGAM: Learning Causal DAGs at the speed of GPUs",
                        "abstract": "Existing causal discovery methods based on combinatorial optimization or search are slow, prohibiting their application on large-scale datasets. In response, more recent methods attempt to address this limitation by formulating causal discovery as structure learning with continuous optimization but such approaches thus far provide no statistical guarantees. In this paper, we show that by efficiently parallelizing existing causal discovery methods, we can in fact scale them to thousands of dimensions, making them practical for substantially larger-scale problems. In particular, we parallelize the LiNGAM method, which is quadratic in the number of variables, obtaining up to a 32-fold speed-up on benchmark datasets when compared with existing sequential implementations. Specifically, we focus on the causal ordering subprocedure in DirectLiNGAM and implement GPU kernels to accelerate it. This allows us to apply DirectLiNGAM to causal inference on large-scale gene expression data with genetic interventions yielding competitive results compared with specialized continuous optimization methods, and Var-LiNGAM for causal discovery on U.S. stock data."
                    },
                    {
                        "title": "The Multiple Quantile Graphical Model",
                        "abstract": "We introduce the Multiple Quantile Graphical Model (MQGM), which extends the neighborhood selection approach of Meinshausen and Buhlmann for learning sparse graphical models. The latter is defined by the basic subproblem of modeling the conditional mean of one variable as a sparse function of all others. Our approach models a set of conditional quantiles of one variable as a sparse function of all others, and hence offers a much richer, more expressive class of conditional distribution estimates. We establish that, under suitable regularity conditions, the MQGM identifies the exact conditional independencies with probability tending to one as the problem size grows, even outside of the usual homoskedastic Gaussian data model. We develop an efficient algorithm for fitting the MQGM using the alternating direction method of multipliers. We also describe a strategy for sampling from the joint distribution that underlies the MQGM estimate. Lastly, we present detailed experiments that demonstrate the flexibility and effectiveness of the MQGM in modeling hetereoskedastic non-Gaussian data."
                    },
                    {
                        "title": "A Continuous-Time View of Early Stopping for Least Squares",
                        "abstract": "We study the statistical properties of the iterates generated by gradient descent, applied to the fundamental problem of least squares regression. We take a continuous-time view, i.e., consider infinitesimal step sizes in gradient descent, in which case the iterates form a trajectory called gradient flow. Our primary focus is to compare the risk of gradient flow to that of ridge regression. Under the calibration $t=1/\\lambda$---where $t$ is the time parameter in gradient flow, and $\\lambda$ the tuning parameter in ridge regression---we prove that the risk of gradient flow is no less than 1.69 times that of ridge, along the entire path (for all $t \\geq 0$). This holds in finite samples with very weak assumptions on the data model (in particular, with no assumptions on the features $X$). We prove that the same relative risk bound holds for prediction risk, in an average sense over the underlying signal $\\beta_0$. Finally, we examine limiting risk expressions (under standard Marchenko-Pastur asymptotics), and give supporting numerical experiments."
                    },
                    {
                        "title": "A Fast Algorithm for Sparse Controller Design",
                        "abstract": "We consider the task of designing sparse control laws for large-scale systems by directly minimizing an infinite horizon quadratic cost with an $\\ell_1$ penalty on the feedback controller gains. Our focus is on an improved algorithm that allows us to scale to large systems (i.e. those where sparsity is most useful) with convergence times that are several orders of magnitude faster than existing algorithms. In particular, we develop an efficient proximal Newton method which minimizes per-iteration cost with a coordinate descent active set approach and fast numerical solutions to the Lyapunov equations. Experimentally we demonstrate the appeal of this approach on synthetic examples and real power networks significantly larger than those previously considered in the literature."
                    },
                    {
                        "title": "Contextually Supervised Source Separation with Application to Energy Disaggregation",
                        "abstract": "We propose a new framework for single-channel source separation that lies between the fully supervised and unsupervised setting. Instead of supervision, we provide input features for each source signal and use convex methods to estimate the correlations between these features and the unobserved signal decomposition. We analyze the case of $\\ell_2$ loss theoretically and show that recovery of the signal components depends only on cross-correlation between features for different signals, not on correlations between features for the same signal. Contextually supervised source separation is a natural fit for domains with large amounts of data but no explicit supervision; our motivating application is energy disaggregation of hourly smart meter data (the separation of whole-home power signals into different energy uses). Here we apply contextual supervision to disaggregate the energy usage of thousands homes over four years, a significantly larger scale than previously published efforts, and demonstrate on synthetic data that our method outperforms the unsupervised approach."
                    },
                    {
                        "title": "Input Convex Neural Networks",
                        "abstract": "This paper presents the input convex neural network architecture. These are scalar-valued (potentially deep) neural networks with constraints on the network parameters such that the output of the network is a convex function of (some of) the inputs. The networks allow for efficient inference via optimization over some inputs to the network given others, and can be applied to settings including structured prediction, data imputation, reinforcement learning, and others. In this paper we lay the basic groundwork for these models, proposing methods for inference, optimization and learning, and analyze their representational power. We show that many existing neural network architectures can be made input-convex with a minor modification, and develop specialized optimization algorithms tailored to this setting. Finally, we highlight the performance of the methods on multi-label prediction, image completion, and reinforcement learning problems, where we show improvement over the existing state of the art in many cases."
                    },
                    {
                        "title": "Trellis Networks for Sequence Modeling",
                        "abstract": "We present trellis networks, a new architecture for sequence modeling. On the one hand, a trellis network is a temporal convolutional network with special structure, characterized by weight tying across depth and direct injection of the input into deep layers. On the other hand, we show that truncated recurrent networks are equivalent to trellis networks with special sparsity structure in their weight matrices. Thus trellis networks with general weight matrices generalize truncated recurrent networks. We leverage these connections to design high-performing trellis networks that absorb structural and algorithmic elements from both recurrent and convolutional models. Experiments demonstrate that trellis networks outperform the current state of the art methods on a variety of challenging benchmarks, including word-level language modeling and character-level language modeling tasks, and stress tests designed to evaluate long-term memory retention. The code is available at https://github.com/locuslab/trellisnet ."
                    },
                    {
                        "title": "The Limited Multi-Label Projection Layer",
                        "abstract": "We propose the Limited Multi-Label (LML) projection layer as a new primitive operation for end-to-end learning systems. The LML layer provides a probabilistic way of modeling multi-label predictions limited to having exactly k labels. We derive efficient forward and backward passes for this layer and show how the layer can be used to optimize the top-k recall for multi-label tasks with incomplete label information. We evaluate LML layers on top-k CIFAR-100 classification and scene graph generation. We show that LML layers add a negligible amount of computational overhead, strictly improve the model's representational capacity, and improve accuracy. We also revisit the truncated top-k entropy method as a competitive baseline for top-k classification."
                    }
                ]
            }
        }
    },
    "2405.18765": {
        "paper_data": {
            "title": "Large Brain Model for Learning Generic Representations with Tremendous EEG Data in BCI",
            "url": "http://arxiv.org/abs/2405.18765v1",
            "arxiv_id": "2405.18765",
            "authors": [
                "Wei-Bang Jiang",
                "Li-Ming Zhao",
                "Bao-Liang Lu"
            ],
            "abstract": "The current electroencephalogram (EEG) based deep learning models are typically designed for specific datasets and applications in brain-computer interaction (BCI), limiting the scale of the models and thus diminishing their perceptual capabilities and generalizability. Recently, Large Language Models (LLMs) have achieved unprecedented success in text processing, prompting us to explore the capabilities of Large EEG Models (LEMs). We hope that LEMs can break through the limitations of different task types of EEG datasets, and obtain universal perceptual capabilities of EEG signals through unsupervised pre-training. Then the models can be fine-tuned for different downstream tasks. However, compared to text data, the volume of EEG datasets is generally small and the format varies widely. For example, there can be mismatched numbers of electrodes, unequal length data samples, varied task designs, and low signal-to-noise ratio. To overcome these challenges, we propose a unified foundation model for EEG called Large Brain Model (LaBraM). LaBraM enables cross-dataset learning by segmenting the EEG signals into EEG channel patches. Vector-quantized neural spectrum prediction is used to train a semantically rich neural tokenizer that encodes continuous raw EEG channel patches into compact neural codes. We then pre-train neural Transformers by predicting the original neural codes for the masked EEG channel patches. The LaBraMs were pre-trained on about 2,500 hours of various types of EEG signals from around 20 datasets and validated on multiple different types of downstream tasks. Experiments on abnormal detection, event type classification, emotion recognition, and gait prediction show that our LaBraM outperforms all compared SOTA methods in their respective fields. Our code is available at https://github.com/935963004/LaBraM.",
            "introduction": "   1 Introduction  Electroencephalography (EEG) is a method to record an electrogram of the spontaneous electrical activity of the brain. It is typically non-invasive, with the EEG electrodes placed along the scalp using the international 10\u201320 system. EEG signals can be formulated as a matrix of real numbers X\u2208\u211dC\u00d7T\ud835\udc4bsuperscript\u211d\ud835\udc36\ud835\udc47X\\in\\mathbb{R}^{C\\times T}italic_X \u2208 blackboard_R start_POSTSUPERSCRIPT italic_C \u00d7 italic_T end_POSTSUPERSCRIPT, where C\ud835\udc36Citalic_C is the number of EEG electrodes (channels) that may vary depending on the acquisition equipment used, and T\ud835\udc47Titalic_T represents the total number of samples, which is related to the collection time and sampling rate. As highly objective physiological signals, EEG has demonstrated remarkable potential in seizure epilepsy classification (Boonyakitanont et\u00a0al., 2020), acute stress detection (Sharma et\u00a0al., 2022), sleep stage classification (Aboalayon et\u00a0al., 2016), motor imagery recognition (Amin et\u00a0al., 2019), abnormal identification (Roy et\u00a0al., 2019), emotion analysis (Suhaimi et\u00a0al., 2020), and auditory attention detection (Biesmans et\u00a0al., 2016).   Numerous deep learning models have been proposed to address the aforementioned tasks in their respective fields. Some works apply convolutional neural networks (CNN) across and within raw EEG channels to encode spatial and temporal features (Lawhern et\u00a0al., 2018), while others preprocess the data using short-time Fourier transform (STFT) and employ Graph Neural Network (GNN) on the resulting spectrograms to obtain semantic features of brain area links (Song et\u00a0al., 2018). Researchers also segment the signal and use a CNN segment encoder with a downstream sequence model such as recurrent neural networks (RNN) to capture temporal dynamics (Xu et\u00a0al., 2020). These models primarily focus on EEG samples that adhere to specific task formats, mainly because the equipment used to collect EEG differs between datasets, which introduces mismatched channels and variable lengths. Meanwhile, EEG data collection is quite expensive, which makes it challenging to build large EEG datasets specifically designed for a particular task. To prevent overfitting, the parameters of these models need to be regulated, which in turn hampers the model\u2019s ability to learn EEG expressions and limits its generalizability. Consequently, we discovered that current EEG models are typically proprietary and lack the capacity to perform cross-task learning.   Recently, we have been impressed by the capabilities of LLMs (Ouyang et\u00a0al., 2022; Wei et\u00a0al., 2022). Specifically, Transformer-based models have demonstrated promising results in natural language processing tasks, which highlights the potential of self-supervised pre-training as a means for harnessing large-scale data. These masked language modeling tasks involve randomly masking some proportion of tokens within a text and then recovering the masked tokens based on the Transformer encoding results of the corrupted text. Motivated by these methods, we propose to apply reconstruction ideas to pre-train neural Transformers. However, it is a daunting task to directly apply LLM-style pre-training to EEG data. The challenges are summarized as follows:   1) Lack of sufficient EEG data. The acquisition of EEG data is significantly challenging compared to natural language and image data. Moreover, the annotation of EEG data usually requires a lot of effort on the part of experts in the corresponding field, thus leading to the fact that only small labeled datasets exist for specific tasks in BCI, where EEG signals are often collected from a small number",
            "references": [
                {
                    "title": "Multimodal Adaptive Emotion Transformer with Flexible Modality Inputs on A Novel Dataset with Continuous Labels",
                    "abstract": "Emotion recognition from physiological signals is a topic of widespread interest, and researchers continue to develop novel techniques for perceiving emotions. However, the emergence of deep learning has highlighted the need for high-quality emotional datasets to accurately decode human emotions. In this study, we present a novel multimodal emotion dataset that incorporates electroencephalography (EEG) and eye movement signals to systematically explore human emotions. Seven basic emotions (happy, sad, fear, disgust, surprise, anger, and neutral) are elicited by a large number of 80 videos and fully investigated with continuous labels that indicate the intensity of the corresponding emotions. Additionally, we propose a novel Multimodal Adaptive Emotion Transformer (MAET), that can flexibly process both unimodal and multimodal inputs. Adversarial training is utilized in MAET to mitigate subject discrepancy, which enhances domain generalization. Our extensive experiments, encompassing both subject-dependent and cross-subject conditions, demonstrate MAET's superior performance in handling various inputs. The filtering of data for high emotional evocation using continuous labels proved to be effective in the experiments. Furthermore, the complementary properties between EEG and eye movements are observed. Our code is available at https://github.com/935963004/MAET."
                },
                {
                    "title": "Aggregating Intrinsic Information to Enhance BCI Performance through Federated Learning",
                    "abstract": "Insufficient data is a long-standing challenge for Brain-Computer Interface (BCI) to build a high-performance deep learning model. Though numerous research groups and institutes collect a multitude of EEG datasets for the same BCI task, sharing EEG data from multiple sites is still challenging due to the heterogeneity of devices. The significance of this challenge cannot be overstated, given the critical role of data diversity in fostering model robustness. However, existing works rarely discuss this issue, predominantly centering their attention on model training within a single dataset, often in the context of inter-subject or inter-session settings. In this work, we propose a hierarchical personalized Federated Learning EEG decoding (FLEEG) framework to surmount this challenge. This innovative framework heralds a new learning paradigm for BCI, enabling datasets with disparate data formats to collaborate in the model training process. Each client is assigned a specific dataset and trains a hierarchical personalized model to manage diverse data formats and facilitate information exchange. Meanwhile, the server coordinates the training procedure to harness knowledge gleaned from all datasets, thus elevating overall performance. The framework has been evaluated in Motor Imagery (MI) classification with nine EEG datasets collected by different devices but implementing the same MI task. Results demonstrate that the proposed framework can boost classification performance up to 8.4% by enabling knowledge sharing between multiple datasets, especially for smaller datasets. Visualization results also indicate that the proposed framework can empower the local models to put a stable focus on task-related areas, yielding better performance. To the best of our knowledge, this is the first end-to-end solution to address this important challenge."
                },
                {
                    "title": "EEG decoding for datasets with heterogenous electrode configurations using transfer learning graph neural networks",
                    "abstract": "Objective. Brain-machine interfacing (BMI) has greatly benefited from adopting machine learning methods for feature learning that require extensive data for training, which are often unavailable from a single dataset. Yet, it is difficult to combine data across labs or even data within the same lab collected over the years due to the variation in recording equipment and electrode layouts resulting in shifts in data distribution, changes in data dimensionality, and altered identity of data dimensions. Our objective is to overcome this limitation and learn from many different and diverse datasets across labs with different experimental protocols. Approach. To tackle the domain adaptation problem, we developed a novel machine learning framework combining graph neural networks (GNNs) and transfer learning methodologies for non-invasive motor imagery (MI) EEG decoding, as an example of BMI. Empirically, we focus on the challenges of learning from EEG data with different electrode layouts and varying numbers of electrodes. We utilize three MI EEG databases collected using very different numbers of EEG sensors (from 22 channels to 64) and layouts (from custom layouts to 10\u201320). Main results. Our model achieved the highest accuracy with lower standard deviations on the testing datasets. This indicates that the GNN-based transfer learning framework can effectively aggregate knowledge from multiple datasets with different electrode layouts, leading to improved generalization in subject-independent MI EEG classification. Significance. The findings of this study have important implications for brain-computer-interface research, as they highlight a promising method for overcoming the limitations posed by non-unified experimental setups. By enabling the integration of diverse datasets with varying electrode layouts, our proposed approach can help advance the development and application of BMI technologies."
                },
                {
                    "title": "Generalizable Movement Intention Recognition with Multiple Heterogeneous EEG Datasets",
                    "abstract": "Human movement intention recognition is important for human-robot interaction. Existing work based on motor imagery electroencephalogram (EEG) provides a non-invasive and portable solution for intention detection. However, the data-driven methods may suffer from the limited scale and diversity of the training datasets, which result in poor generalization performance on new test subjects. It is practically difficult to directly aggregate data from multiple datasets for training, since they often employ different channels and collected data suffers from significant domain shifts caused by different devices, experiment setup, etc. On the other hand, the inter-subject heterogeneity is also substantial due to individual differences in EEG representations. In this work, we developed two networks to learn from both the shared and the complete channels across datasets, handling inter-subject and inter-dataset heterogeneity respectively. Based on both networks, we further developed an online knowledge co-distillation framework to collaboratively learn from both networks, achieving coherent performance boosts. Experimental results have shown that our proposed method can effectively aggregate knowledge from multiple datasets, demonstrating better generalization in the context of cross-subject validation."
                },
                {
                    "title": "Development of Expert-Level Classification of Seizures and Rhythmic and Periodic Patterns During EEG Interpretation",
                    "abstract": "Background and Objectives Seizures (SZs) and other SZ-like patterns of brain activity can harm the brain and contribute to in-hospital death, particularly when prolonged. However, experts qualified to interpret EEG data are scarce. Prior attempts to automate this task have been limited by small or inadequately labeled samples and have not convincingly demonstrated generalizable expert-level performance. There exists a critical unmet need for an automated method to classify SZs and other SZ-like events with expert-level reliability. This study was conducted to develop and validate a computer algorithm that matches the reliability and accuracy of experts in identifying SZs and SZ-like events, known as \u201cictal-interictal-injury continuum\u201d (IIIC) patterns on EEG, including SZs, lateralized and generalized periodic discharges (LPD, GPD), and lateralized and generalized rhythmic delta activity (LRDA, GRDA), and in differentiating these patterns from non-IIIC patterns. Methods We used 6,095 scalp EEGs from 2,711 patients with and without IIIC events to train a deep neural network, SPaRCNet, to perform IIIC event classification. Independent training and test data sets were generated from 50,697 EEG segments, independently annotated by 20 fellowship-trained neurophysiologists. We assessed whether SPaRCNet performs at or above the sensitivity, specificity, precision, and calibration of fellowship-trained neurophysiologists for identifying IIIC events. Statistical performance was assessed by the calibration index and by the percentage of experts whose operating points were below the model's receiver operating characteristic curves (ROCs) and precision recall curves (PRCs) for the 6 pattern classes. Results SPaRCNet matches or exceeds most experts in classifying IIIC events based on both calibration and discrimination metrics. For SZ, LPD, GPD, LRDA, GRDA, and \u201cother\u201d classes, SPaRCNet exceeds the following percentages of 20 experts\u2014ROC: 45%, 20%, 50%, 75%, 55%, and 40%; PRC: 50%, 35%, 50%, 90%, 70%, and 45%; and calibration: 95%, 100%, 95%, 100%, 100%, and 80%, respectively. Discussion SPaRCNet is the first algorithm to match expert performance in detecting SZs and other SZ-like events in a representative sample of EEGs. With further development, SPaRCNet may thus be a valuable tool for an expedited review of EEGs. Classification of Evidence This study provides Class II evidence that among patients with epilepsy or critical illness undergoing EEG monitoring, SPaRCNet can differentiate (IIIC) patterns from non-IIIC events and expert neurophysiologists."
                },
                {
                    "title": "BEiT v2: Masked Image Modeling with Vector-Quantized Visual Tokenizers",
                    "abstract": "Masked image modeling (MIM) has demonstrated impressive results in selfsupervised representation learning by recovering corrupted image patches. However, most existing studies operate on low-level image pixels, which hinders the exploitation of high-level semantics for representation models. In this work, we propose to use a semantic-rich visual tokenizer as the reconstruction target for masked prediction, providing a systematic way to promote MIM from pixel-level to semantic-level. Specifically, we propose vector-quantized knowledge distillation to train the tokenizer, which discretizes a continuous semantic space to compact codes. We then pretrain vision Transformers by predicting the original visual tokens for the masked image patches. Furthermore, we introduce a patch aggregation strategy which associates discrete image patches to enhance global semantic representation. Experiments on image classification and semantic segmentation show that BEIT V2 outperforms all compared MIM methods. On ImageNet-1K (224 size), the base-size BEIT V2 achieves 85.5% top-1 accuracy for fine-tuning and 80.1% top-1 accuracy for linear probing. The large-size BEIT V2 obtains 87.3% top-1 accuracy for ImageNet-1K (224 size) fine-tuning, and 56.7% mIoU on ADE20K for semantic segmentation. The code and pretrained models are available at https://aka.ms/beitv2."
                },
                {
                    "title": "Multimodal Emotion Recognition in Response to Oil Paintings",
                    "abstract": "Most previous affective studies use facial expression pictures, music or movie clips as emotional stimuli, which are either too simplified without contexts or too dynamic for emotion annotations. In this work, we evaluate the effectiveness of oil paintings as stimuli. We develop an emotion stimuli dataset with 114 oil paintings selected from subject ratings to evoke three emotional states (i.e., negative, neutral and positive), and acquire both EEG and eye tracking data from 20 subjects while watching the oil paintings. Furthermore, we propose a novel affective model for multimodal emotion recognition by 1) extracting informative features of EEG signals from both the time domain and the frequency domain, 2) exploring topological information embedded in EEG channels with graph neural networks (GNNs), and 3) combining EEG and eye tracking data with a deep autoencoder neural network. From the exper-iments, our model obtains an averaged classification accuracy of 94.72 % \u00b1 1.47 %, which demonstrates the feasibility of using oil paintings as emotion elicitation material."
                },
                {
                    "title": "Transformer Convolutional Neural Networks for Automated Artifact Detection in Scalp EEG",
                    "abstract": "It is well known that electroencephalograms (EEGs) often contain artifacts due to muscle activity, eye blinks, and various other causes. Detecting such artifacts is an essential first step toward a correct interpretation of EEGs. Although much effort has been devoted to semi-automated and automated artifact detection in EEG, the problem of artifact detection remains challenging. In this paper, we propose a convolutional neural network (CNN) enhanced by transformers using belief matching (BM) loss for automated detection of five types of artifacts: chewing, electrode pop, eye movement, muscle, and shiver. Specifically, we apply these five detectors at individual EEG channels to distinguish artifacts from background EEG. Next, for each of these five types of artifacts, we combine the output of these channel-wise detectors to detect artifacts in multi-channel EEG segments. These segment-level classifiers can detect specific artifacts with a balanced accuracy (BAC) of 0.947, 0.735, 0.826, 0.857, and 0.655 for chewing, electrode pop, eye movement, muscle, and shiver artifacts, respectively. Finally, we combine the outputs of the five segment-level detectors to perform a combined binary classification (any artifact vs. background). The resulting detector achieves a sensitivity (SEN) of 42.0%, 32.0%, and 13.3%, at a specificity (SPE) of 95%, 97%, and 99%, respectively. This artifact detection module can reject artifact segments while only removing a small fraction of the background EEG, leading to a cleaner EEG for further analysis."
                },
                {
                    "title": "Emergent Abilities of Large Language Models",
                    "abstract": "Scaling up language models has been shown to predictably improve performance and sample efficiency on a wide range of downstream tasks. This paper instead discusses an unpredictable phenomenon that we refer to as emergent abilities of large language models. We consider an ability to be emergent if it is not present in smaller models but is present in larger models. Thus, emergent abilities cannot be predicted simply by extrapolating the performance of smaller models. The existence of such emergence implies that additional scaling could further expand the range of capabilities of language models."
                },
                {
                    "title": "Neuro-BERT: Rethinking Masked Autoencoding for Self-Supervised Neurological Pretraining.",
                    "abstract": "Deep learning associated with neurological signals is poised to drive major advancements in diverse fields such as medical diagnostics, neurorehabilitation, and brain-computer interfaces. The challenge in harnessing the full potential of these signals lies in the dependency on extensive, high-quality annotated data, which is often scarce and expensive to acquire, requiring specialized infrastructure and domain expertise. To address the appetite for data in deep learning, we present Neuro-BERT, a self-supervised pre-training framework of neurological signals based on masked autoencoding in the Fourier domain. The intuition behind our approach is simple: frequency and phase distribution of neurological signals can reveal intricate neurological activities. We propose a novel pre-training task dubbed Fourier Inversion Prediction (FIP), which randomly masks out a portion of the input signal and then predicts the missing information using the Fourier inversion theorem. Pre-trained models can be potentially used for various downstream tasks such as sleep stage classification and gesture recognition. Unlike contrastive-based methods, which strongly rely on carefully hand-crafted augmentations and siamese structure, our approach works reasonably well with a simple transformer encoder with no augmentation requirements. By evaluating our method on several benchmark datasets, we show that Neuro-BERT improves downstream neurological-related tasks by a large margin."
                },
                {
                    "title": "Identifying similarities and differences in emotion recognition with EEG and eye movements among Chinese, German, and French People",
                    "abstract": "Objective. Cultures have essential influences on emotions. However, most studies on cultural influences on emotions are in the areas of psychology and neuroscience, while the existing affective models are mostly built with data from the same culture. In this paper, we identify the similarities and differences among Chinese, German, and French individuals in emotion recognition with electroencephalogram (EEG) and eye movements from an affective computing perspective. Approach. Three experimental settings were designed: intraculture subject dependent, intraculture subject independent, and cross-culture subject independent. EEG and eye movements are acquired simultaneously from Chinese, German, and French subjects while watching positive, neutral, and negative movie clips. The affective models for Chinese, German, and French subjects are constructed by using machine learning algorithms. A systematic analysis is performed from four aspects: affective model performance, neural patterns, complementary information from different modalities, and cross-cultural emotion recognition. Main results. From emotion recognition accuracies, we find that EEG and eye movements can adapt to Chinese, German, and French cultural diversities and that a cultural in-group advantage phenomenon does exist in emotion recognition with EEG. From the topomaps of EEG, we find that the \u03b3 and \u03b2 bands exhibit decreasing activities for Chinese, while for German and French, \u03b8 and \u03b1 bands exhibit increasing activities. From confusion matrices and attentional weights, we find that EEG and eye movements have complementary characteristics. From a cross-cultural emotion recognition perspective, we observe that German and French people share more similarities in topographical patterns and attentional weight distributions than Chinese people while the data from Chinese are a good fit for test data but not suitable for training data for the other two cultures. Significance. Our experimental results provide concrete evidence of the in-group advantage phenomenon, cultural influences on emotion recognition, and different neural patterns among Chinese, German, and French individuals."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Discriminating Surprise and Anger from EEG and Eye Movements with a Graph Network",
                    "abstract": "Emotion recognition based on EEG and eye movement signals has been studied extensively due to the reliability and stability of signals. The separability of four basic emotions, happy, sad, disgust and fear, has been systematically studied in the existing work. However, there is less research on the emotions of anger and surprise since they are more difficult to be elicited in lab settings. This paper investigates the discrimination ability of EEG and eye movement signals for surprise and anger. To this end, we design a stimulus paradigm that can effectively elicit surprise and anger. We propose a novel Graph Convolutional Network with Channel Attention (GCNCA) to classify three emotions, anger, surprise and neutrality. Experimental results indicate that: a) the proposed GCNCA model has an excellent classification accuracy of 86.47% using EEG and 84.22% using eye movement signals, which are better than other baseline methods; b) EEG and eye movements have a good ability to discriminate surprise and anger, while EEG performs better than eye movements; c) the high-frequency bands of EEG are more distinguishable on classifying surprise and anger than the low-frequency bands; d) there are some differences in neural patterns between surprise and anger, meanwhile critical channels and channel connections of EEG are found."
                },
                {
                    "title": "Recent Advances in the TUH EEG Corpus: Improving the Interrater Agreement for Artifacts and Epileptiform Events",
                    "abstract": "The Temple University Hospital EEG Corpus (TUEG) [1] is the largest publicly available EEG corpus of its type and currently has over 5,000 subscribers (we currently average 35 new subscribers a week). Several valuable subsets of this corpus have been developed including the Temple University Hospital EEG Seizure Corpus (TUSZ) [2] and the Temple University Hospital EEG Artifact Corpus (TUAR) [3]. TUSZ contains manually annotated seizure events and has been widely used to develop seizure detection and prediction technology [4]. TUAR contains manually annotated artifacts and has been used to improve machine learning performance on seizure detection tasks [5]. In this poster, we will discuss recent improvements made to both corpora that are creating opportunities to improve machine learning performance."
                },
                {
                    "title": "SimMIM: a Simple Framework for Masked Image Modeling",
                    "abstract": "This paper presents SimMIM, a simple framework for masked image modeling. We have simplified recently proposed relevant approaches, without the need for special designs, such as block-wise masking and tokenization via discrete VAE or clustering. To investigate what makes a masked image modeling task learn good representations, we systematically study the major components in our framework, and find that the simple designs of each component have revealed very strong representation learning performance: 1) random masking of the input image with a moderately large masked patch size (e.g., 32) makes a powerful pre-text task; 2) predicting RGB values of raw pixels by direct regression performs no worse than the patch classification approaches with complex designs; 3) the prediction head can be as light as a linear layer, with no worse performance than heavier ones. Using ViT-B, our approach achieves 83.8% top-1 fine-tuning accuracy on ImageNet-1K by pre-training also on this dataset, surpassing previous best approach by +0.6%. When applied to a larger model with about 650 million parameters, SwinV2-H, it achieves 87.1% top-1 accuracy on ImageNet-1K using only ImageNet-1K data. We also leverage this approach to address the data-hungry issue faced by large-scale model training, that a 3B model (Swin V2-G) is successfully trained to achieve state-of-the-art accuracy on four representative vision benchmarks using 40\u00d7 less labelled data than that in previous practice (JFT-3B). The code is available at https://github.com/microsoft/SimMIM."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "BEiT: BERT Pre-Training of Image Transformers",
                    "abstract": "We introduce a self-supervised vision representation model BEiT, which stands for Bidirectional Encoder representation from Image Transformers. Following BERT developed in the natural language processing area, we propose a masked image modeling task to pretrain vision Transformers. Specifically, each image has two views in our pre-training, i.e, image patches (such as 16x16 pixels), and visual tokens (i.e., discrete tokens). We first\"tokenize\"the original image into visual tokens. Then we randomly mask some image patches and fed them into the backbone Transformer. The pre-training objective is to recover the original visual tokens based on the corrupted image patches. After pre-training BEiT, we directly fine-tune the model parameters on downstream tasks by appending task layers upon the pretrained encoder. Experimental results on image classification and semantic segmentation show that our model achieves competitive results with previous pre-training methods. For example, base-size BEiT achieves 83.2% top-1 accuracy on ImageNet-1K, significantly outperforming from-scratch DeiT training (81.8%) with the same setup. Moreover, large-size BEiT obtains 86.3% only using ImageNet-1K, even outperforming ViT-L with supervised pre-training on ImageNet-22K (85.2%). The code and pretrained models are available at https://aka.ms/beit."
                },
                {
                    "title": "Transformer-based Spatial-Temporal Feature Learning for EEG Decoding",
                    "abstract": "At present, people usually use some methods based on convolutional neural networks (CNNs) for Electroencephalograph (EEG) decoding. However, CNNs have limitations in perceiving global dependencies, which is not adequate for common EEG paradigms with a strong overall relationship. Regarding this issue, we propose a novel EEG decoding method that mainly relies on the attention mechanism. The EEG data is firstly preprocessed and spatially filtered. And then, we apply attention transforming on the feature-channel dimension so that the model can enhance more relevant spatial features. The most crucial step is to slice the data in the time dimension for attention transforming, and finally obtain a highly distinguishable representation. At this time, global averaging pooling and a simple fully-connected layer are used to classify different categories of EEG data. Experiments on two public datasets indicate that the strategy of attention transforming effectively utilizes spatial and temporal features. And we have reached the level of the state-of-the-art in multi-classification of EEG, with fewer parameters. As far as we know, it is the first time that a detailed and complete method based on the transformer idea has been proposed in this field. It has good potential to promote the practicality of brain-computer interface (BCI). The source code can be found at: \\textit{https://github.com/anranknight/EEG-Transformer}."
                },
                {
                    "title": "Discrimination of Decision Confidence Levels from EEG Signals",
                    "abstract": "To explore the capability of utilizing electroencephalograms (EEGs) for the measurement of human decision confidence levels, this paper develops a new visual perceptual decision confidence experiment. In this experiment, a visual perceptual decision-making task is performed by 14 participants, and their EEG data are recorded. The problem of measuring decision confidence levels is considered to be a pattern classification task, and two pattern classifiers are trained with differential entropy (DE), power spectral density (PSD), differential asymmetry (DASM), rational asymmetry (RASM), and asymmetry (ASM) features extracted from multichannel EEG data. We compare the performance of these features and find that the DE feature performs better than the others for measuring levels of decision confidence. The experimental results indicate that EEG signals offer good capability for measuring human decision confidence levels. The best performance of our proposed method in measuring five levels of decision confidence reaches an accuracy of 49.14 % and F1-score of 45.07 %, and for the extreme levels of decision confidence, the recognition accuracy reaches 91.28 %, with an average F1-score of 88.92 %. Topographic maps are also used to depict the neural patterns of EEG signals, suggesting that the posterior parietal cortex and occipital cortex might be sensitive brain areas for indicating decision confidence."
                },
                {
                    "title": "Comparing Recognition Performance and Robustness of Multimodal Deep Learning Models for Multimodal Emotion Recognition",
                    "abstract": "Multimodal signals are powerful for emotion recognition since they can represent emotions comprehensively. In this article, we compare the recognition performance and robustness of two multimodal emotion recognition models: 1) deep canonical correlation analysis (DCCA) and 2) bimodal deep autoencoder (BDAE). The contributions of this article are threefold: 1) we propose two methods for extending the original DCCA model for multimodal fusion: a) weighted sum fusion and b) attention-based fusion; 2) we systemically compare the performance of DCCA, BDAE, and traditional approaches on five multimodal data sets; and 3) we investigate the robustness of DCCA, BDAE, and traditional approaches on SEED-V and DREAMER data sets under two conditions: 1) adding noises to multimodal features and 2) replacing electroencephalography features with noises. Our experimental results demonstrate that DCCA achieves state-of-the-art recognition results on all five data sets: 1) 94.6% on the SEED data set; 2) 87.5% on the SEED-IV data set; 3) 84.3% and 85.6% on the DEAP data set; 4) 85.3% on the SEED-V data set; and 5) 89.0%, 90.6%, and 90.7% on the DREAMER data set. Meanwhile, DCCA has greater robustness when adding various amounts of noises to the SEED-V and DREAMER data sets. By visualizing features before and after DCCA transformation on the SEED-V data set, we find that the transformed features are more homogeneous and discriminative across emotions."
                },
                {
                    "title": "BENDR: Using Transformers and a Contrastive Self-Supervised Learning Task to Learn From Massive Amounts of EEG Data",
                    "abstract": "Deep neural networks (DNNs) used for brain\u2013computer interface (BCI) classification are commonly expected to learn general features when trained across a variety of contexts, such that these features could be fine-tuned to specific contexts. While some success is found in such an approach, we suggest that this interpretation is limited and an alternative would better leverage the newly (publicly) available massive electroencephalography (EEG) datasets. We consider how to adapt techniques and architectures used for language modeling (LM) that appear capable of ingesting awesome amounts of data toward the development of encephalography modeling with DNNs in the same vein. We specifically adapt an approach effectively used for automatic speech recognition, which similarly (to LMs) uses a self-supervised training objective to learn compressed representations of raw data signals. After adaptation to EEG, we find that a single pre-trained model is capable of modeling completely novel raw EEG sequences recorded with differing hardware, and different subjects performing different tasks. Furthermore, both the internal representations of this model and the entire architecture can be fine-tuned to a variety of downstream BCI and EEG classification tasks, outperforming prior work in more task-specific (sleep stage classification) self-supervision."
                },
                {
                    "title": "A One-Dimensional CNN-LSTM Model for Epileptic Seizure Recognition Using EEG Signal Analysis",
                    "abstract": "Frequent epileptic seizures cause damage to the human brain, resulting in memory impairment, mental decline, and so on. Therefore, it is important to detect epileptic seizures and provide medical treatment in a timely manner. Currently, medical experts recognize epileptic seizure activity through the visual inspection of electroencephalographic (EEG) signal recordings of patients based on their experience, which takes much time and effort. In view of this, this paper proposes a one-dimensional convolutional neural network-long short-term memory (1D CNN-LSTM) model for automatic recognition of epileptic seizures through EEG signal analysis. Firstly, the raw EEG signal data are pre-processed and normalized. Then, a 1D convolutional neural network (CNN) is designed to effectively extract the features of the normalized EEG sequence data. In addition, the extracted features are then processed by the LSTM layers in order to further extract the temporal features. After that, the output features are fed into several fully connected layers for final epileptic seizure recognition. The performance of the proposed 1D CNN-LSTM model is verified on the public UCI epileptic seizure recognition data set. Experiments results show that the proposed method achieves high recognition accuracies of 99.39% and 82.00% on the binary and five-class epileptic seizure recognition tasks, respectively. Comparing results with traditional machine learning methods including k-nearest neighbors, support vector machines, and decision trees, other deep learning methods including standard deep neural network and CNN further verify the superiority of the proposed method."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "EEG-Based Emotion Recognition: A State-of-the-Art Review of Current Trends and Opportunities",
                    "abstract": "Emotions are fundamental for human beings and play an important role in human cognition. Emotion is commonly associated with logical decision making, perception, human interaction, and to a certain extent, human intelligence itself. With the growing interest of the research community towards establishing some meaningful \u201cemotional\u201d interactions between humans and computers, the need for reliable and deployable solutions for the identification of human emotional states is required. Recent developments in using electroencephalography (EEG) for emotion recognition have garnered strong interest from the research community as the latest developments in consumer-grade wearable EEG solutions can provide a cheap, portable, and simple solution for identifying emotions. Since the last comprehensive review was conducted back from the years 2009 to 2016, this paper will update on the current progress of emotion recognition using EEG signals from 2016 to 2019. The focus on this state-of-the-art review focuses on the elements of emotion stimuli type and presentation approach, study size, EEG hardware, machine learning classifiers, and classification approach. From this state-of-the-art review, we suggest several future research opportunities including proposing a different approach in presenting the stimuli in the form of virtual reality (VR). To this end, an additional section devoted specifically to reviewing only VR studies within this research domain is presented as the motivation for this proposed new approach using VR as the stimuli presentation device. This review paper is intended to be useful for the research community working on emotion recognition using EEG signals as well as for those who are venturing into this field of research."
                },
                {
                    "title": "Uncovering the structure of clinical EEG signals with self-supervised learning",
                    "abstract": "Objective. Supervised learning paradigms are often limited by the amount of labeled data that is available. This phenomenon is particularly problematic in clinically-relevant data, such as electroencephalography (EEG), where labeling can be costly in terms of specialized expertise and human processing time. Consequently, deep learning architectures designed to learn on EEG data have yielded relatively shallow models and performances at best similar to those of traditional feature-based approaches. However, in most situations, unlabeled data is available in abundance. By extracting information from this unlabeled data, it might be possible to reach competitive performance with deep neural networks despite limited access to labels. Approach. We investigated self-supervised learning (SSL), a promising technique for discovering structure in unlabeled data, to learn representations of EEG signals. Specifically, we explored two tasks based on temporal context prediction as well as contrastive predictive coding on two clinically-relevant problems: EEG-based sleep staging and pathology detection. We conducted experiments on two large public datasets with thousands of recordings and performed baseline comparisons with purely supervised and hand-engineered approaches. Main results. Linear classifiers trained on SSL-learned features consistently outperformed purely supervised deep neural networks in low-labeled data regimes while reaching competitive performance when all labels were available. Additionally, the embeddings learned with each method revealed clear latent structures related to physiological and clinical phenomena, such as age effects. Significance. We demonstrate the benefit of SSL approaches on EEG data. Our results suggest that self-supervision may pave the way to a wider use of deep learning models on EEG data."
                },
                {
                    "title": "EEG Synchronization Analysis for Seizure Prediction: A Study on Data of Noninvasive Recordings",
                    "abstract": "Objective: Epilepsy is a neurological disorder arising from anomalies of the electrical activity in the brain, affecting ~65 million individuals worldwide. Prediction methods, typically based on machine learning methods, require a large amount of data for training, in order to correctly classify seizures with small false alarm rates. Methods: In this work, we present a new database containing EEG scalp signals of 14 epileptic patients acquired at the Unit of Neurology and Neurophysiology of the University of Siena, Italy. Furthermore, a patient-specific seizure prediction method, based on the detection of synchronization patterns in the EEG, is proposed and tested on the data of the database. The use of noninvasive EEG data aims to explore the possibility of developing a noninvasive monitoring/control device for the prediction of seizures. The prediction method employs synchronization measures computed over all channel pairs and a computationally inexpensive threshold-based classification approach. Results and conclusions: The experimental analysis, performed by inspection and by the proposed threshold-based classifier on all the patients of the database, shows that the features extracted by the synchronization measures are able to detect preictal and ictal states and allow the prediction of the seizures few minutes before the seizure onsets."
                },
                {
                    "title": "Generative Pretraining From Pixels",
                    "abstract": "Inspired by progress in unsupervised representation learning for natural language, we examine whether similar models can learn useful representations for images. We train a sequence Transformer to auto-regressively predict pixels, without incorporating knowledge of the 2D input structure. Despite training on low-resolution ImageNet without labels, we \ufb01nd that a GPT-2 scale model learns strong image representations as measured by linear probing, \ufb01ne-tuning, and low-data classi\ufb01cation. On CIFAR-10, we achieve 96.3% accuracy with a linear probe, outperforming a supervised Wide ResNet, and 99.0% accuracy with full \ufb01ne-tuning, matching the top supervised pre-trained models. An even larger model trained on a mix-ture of ImageNet and web images is competitive with self-supervised benchmarks on ImageNet, achieving 72.0% top-1 accuracy on a linear probe of our features."
                },
                {
                    "title": "Emotion Recognition under Sleep Deprivation Using a Multimodal Residual LSTM Network",
                    "abstract": "Emotion recognition under sleep deprivation is instructive for the study of mental disorders such as major depressive disorder. Previous studies on emotion recognition under sleep deprivation have been mainly based on psychological research techniques. In this paper, we introduce a multilayer weight-sharing multimodal residual LSTM network for emotion recognition under sleep deprivation. The advantage of our proposed method is that it allows for the combination of three different features: the electroencephalography (EEG) single-channel differential entropy (DE) features, EEG functional strength features with topological correlation connectivity, and eye movement features. The experiments under the conditions of sleep deprivation, sleep recovery and baseline are designed and conducted. The experimental results demonstrate that the proposed method significantly enhances the performance compared with the simple concatenation of the features of different modalities, and the best mean accuracies of 86.86% and 82.03% are achieved for four emotions (happiness, sadness, fear, and neutral) in subject-dependent and cross-subject emotion recognition tasks under 30 hours of sleep deprivation, respectively. The classification accuracy of the happiness emotion is obviously impaired under sleep deprivation, indicating that sleep deprivation impairs the stimulation of the happiness emotion, and one night of sleep recovery can reactivate the elicitation of the happiness emotion to the baseline level. Furthermore, we study the brain neural patterns of the four emotional states. The prefrontal area becomes less activated for the happiness emotion and sadness emotion in the gamma band under sleep deprivation, while the neural pattern of the fear emotion is highly robust with respect to sleep deprivation."
                },
                {
                    "title": "EEG Emotion Recognition Using Dynamical Graph Convolutional Neural Networks",
                    "abstract": "In this paper, a multichannel EEG emotion recognition method based on a novel dynamical graph convolutional neural networks (DGCNN) is proposed. The basic idea of the proposed EEG emotion recognition method is to use a graph to model the multichannel EEG features and then perform EEG emotion classification based on this model. Different from the traditional graph convolutional neural networks (GCNN) methods, the proposed DGCNN method can dynamically learn the intrinsic relationship between different electroencephalogram (EEG) channels, represented by an adjacency matrix, via training a neural network so as to benefit for more discriminative EEG feature extraction. Then, the learned adjacency matrix is used to learn more discriminative features for improving the EEG emotion recognition. We conduct extensive experiments on the SJTU emotion EEG dataset (SEED) and DREAMER dataset. The experimental results demonstrate that the proposed method achieves better recognition performance than the state-of-the-art methods, in which the average recognition accuracy of 90.4 percent is achieved for subject dependent experiment while 79.95 percent for subject independent cross-validation one on the SEED database, and the average accuracies of 86.23, 84.54 and 85.02 percent are respectively obtained for valence, arousal and dominance classifications on the DREAMER database."
                },
                {
                    "title": "wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations",
                    "abstract": "We show for the first time that learning powerful representations from speech audio alone followed by fine-tuning on transcribed speech can outperform the best semi-supervised methods while being conceptually simpler. wav2vec 2.0 masks the speech input in the latent space and solves a contrastive task defined over a quantization of the latent representations which are jointly learned. Experiments using all labeled data of Librispeech achieve 1.8/3.3 WER on the clean/other test sets. When lowering the amount of labeled data to one hour, wav2vec 2.0 outperforms the previous state of the art on the 100 hour subset while using 100 times less labeled data. Using just ten minutes of labeled data and pre-training on 53k hours of unlabeled data still achieves 4.8/8.2 WER. This demonstrates the feasibility of speech recognition with limited amounts of labeled data."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Prediction of Reaction Time and Vigilance Variability From Spatio-Spectral Features of Resting-State EEG in a Long Sustained Attention Task",
                    "abstract": "Resting-state brain networks represent the intrinsic state of the brain during the majority of cognitive and sensorimotor tasks. However, no study has yet presented concise predictors of task-induced vigilance variability from spectro-spatial features of the resting-state electroencephalograms (EEG). In this study, ten healthy volunteers have participated in fixed-sequence, varying-duration sessions of sustained attention to response task (SART) for over 100 minutes. A novel and adaptive cumulative vigilance scoring (CVS) scheme is proposed based on tonic performance and response time. Multiple linear regression (MLR) using feature relevance analysis has shown that average CVS, average response time, and variabilities of these scores can be predicted (p < 0.05) from the resting-state band-power ratios of EEG signals. Cross-validated neural networks also captured different associations for narrow-band beta and wide-band gamma and differences between the high- and low-attention networks in temporal regions. The proposed framework and these first findings on stable and significant attention predictors from the power ratios of resting-state EEG can be useful in brain-computer interfacing and vigilance monitoring applications."
                },
                {
                    "title": "Brain Invaders calibration-less P300-based BCI with modulation of flash duration Dataset (bi2015a)",
                    "abstract": "We describe the experimental procedures for an experiment dataset that we have made publicly available at https://doi.org/10.5281/zenodo.3266930 in mat and csv formats. This dataset contains electroencephalographic (EEG) recordings of 50 subjects playing to a visual P300 Brain-Computer Interface (BCI) videogame named Brain Invaders. The interface uses the oddball paradigm on a grid of 36 symbols (1 Target, 35 Non-Target) that are flashed pseudo-randomly to elicit the P300 response. EEG data were recorded using 32 active wet electrodes with three conditions: flash duration 50ms, 80ms or 110ms. The experiment took place at GIPSA-lab, Grenoble, France, in 2015. Python code for manipulating the data is available at https://github.com/plcrodrigues/py.BI.EEG.2015a-GIPSA. The full ID of this dataset bi2015a."
                },
                {
                    "title": "EmotionMeter: A Multimodal Framework for Recognizing Human Emotions",
                    "abstract": "In this paper, we present a multimodal emotion recognition framework called EmotionMeter that combines brain waves and eye movements. To increase the feasibility and wearability of EmotionMeter in real-world applications, we design a six-electrode placement above the ears to collect electroencephalography (EEG) signals. We combine EEG and eye movements for integrating the internal cognitive states and external subconscious behaviors of users to improve the recognition accuracy of EmotionMeter. The experimental results demonstrate that modality fusion with multimodal deep neural networks can significantly enhance the performance compared with a single modality, and the best mean accuracy of 85.11% is achieved for four emotions (happy, sad, fear, and neutral). We explore the complementary characteristics of EEG and eye movements for their representational capacities and identify that EEG has the advantage of classifying happy emotion, whereas eye movements outperform EEG in recognizing fear emotion. To investigate the stability of EmotionMeter over time, each subject performs the experiments three times on different days. EmotionMeter obtains a mean recognition accuracy of 72.39% across sessions with the six-electrode EEG and eye movement features. These experimental results demonstrate the effectiveness of EmotionMeter within and between sessions."
                },
                {
                    "title": "The Temple University Hospital Seizure Detection Corpus",
                    "abstract": "We introduce the TUH EEG Seizure Corpus (TUSZ), which is the largest open source corpus of its type, and represents an accurate characterization of clinical conditions. In this paper, we describe the techniques used to develop TUSZ, evaluate their effectiveness, and present some descriptive statistics on the resulting corpus."
                },
                {
                    "title": "Big data resources for EEGs: Enabling deep learning research",
                    "abstract": "The Temple University Hospital (TUH) electroencephalography (EEG) Corpus is the world's largest open source EEG corpus of its kind [1]. This corpus consists of over 25,000 EEG studies and over 14,000 patients, and includes a neurologist's interpretation of the test, a brief medical history of the patient, and demographic information about the patients such as gender and age. This database represents the efforts of the Department of Neurology and the Neural Engineering Data Consortium to support the use of EEG data in machine learning research. The data was collected in normal clinical settings and hence includes many non-epileptic features such as muscle and movement artifacts, and a variety of channel configurations that cannot be found in currently available, more sanitized datasets. This is the first dataset of its kind to contain a sufficient amount of EEG data to support the application of state of the art deep learning algorithms. The most recent release of this corpus is vl.0.0 which includes 13,550 patients, 23,218 EEG sessions with reports and 61,634 EEG files."
                },
                {
                    "title": "Electroencephalographic slowing: A primary source of error in automatic seizure detection",
                    "abstract": "Although a seizure event represents a major deviation from a baseline electroencephalographic signal, there are features of seizure morphology that can be seen in non-epileptic portions of the record. A transient decrease in frequency, referred to as slowing, is a generally abnormal but not necessarily epileptic EEG variant. Seizure termination is often associated with a period of slowing between the period of peak amplitude and frequency of the seizure and the return to baseline. In annotation of seizure events in the TUH EEG Seizure Corpus, independent slowing events were identified as a major source of false alarm error. Preliminary results demonstrated the difficulty in automatic differentiation between seizure events and independent slowing events. The TUH EEG Slowing database, a subset of the TUH EEG Corpus, was created, and is introduced here, to aid in the development of a seizure detection tool that can differentiate between slowing at the end of a seizure and an independent nonseizure slowing event. The corpus contains 100 10-second samples each of background, slowing, and seizure events. Preliminary experiments show that 77% sensitivity can be achieved in seizure detection using models trained on all three sample types compared to 43% sensitivity with only seizure and background samples."
                },
                {
                    "title": "Neural Discrete Representation Learning",
                    "abstract": "Learning useful representations without supervision remains a key challenge in machine learning. In this paper, we propose a simple yet powerful generative model that learns such discrete representations. Our model, the Vector Quantised-Variational AutoEncoder (VQ-VAE), differs from VAEs in two key ways: the encoder network outputs discrete, rather than continuous, codes; and the prior is learnt rather than static. In order to learn a discrete latent representation, we incorporate ideas from vector quantisation (VQ). Using the VQ method allows the model to circumvent issues of \"posterior collapse\" -- where the latents are ignored when they are paired with a powerful autoregressive decoder -- typically observed in the VAE framework. Pairing these representations with an autoregressive prior, the model can generate high quality images, videos, and speech as well as doing high quality speaker conversion and unsupervised learning of phonemes, providing further evidence of the utility of the learnt representations."
                },
                {
                    "title": "The Effect of Electroencephalogram (EEG) Reference Choice on Information-Theoretic Measures of the Complexity and Integration of EEG Signals",
                    "abstract": "Converging evidence suggests that human cognition and behavior emerge from functional brain networks interacting on local and global scales. We investigated two information-theoretic measures of functional brain segregation and integration\u2014interaction complexity CI(X), and integration I(X)\u2014as applied to electroencephalographic (EEG) signals and how these measures are affected by choice of EEG reference. CI(X) is a statistical measure of the system entropy accounted for by interactions among its elements, whereas I(X) indexes the overall deviation from statistical independence of the individual elements of a system. We recorded 72 channels of scalp EEG from human participants who sat in a wakeful resting state (interleaved counterbalanced eyes-open and eyes-closed blocks). CI(X) and I(X) of the EEG signals were computed using four different EEG references: linked-mastoids (LM) reference, average (AVG) reference, a Laplacian (LAP) \u201creference-free\u201d transformation, and an infinity (INF) reference estimated via the Reference Electrode Standardization Technique (REST). Fourier-based power spectral density (PSD), a standard measure of resting state activity, was computed for comparison and as a check of data integrity and quality. We also performed dipole source modeling in order to assess the accuracy of neural source CI(X) and I(X) estimates obtained from scalp-level EEG signals. CI(X) was largest for the LAP transformation, smallest for the LM reference, and at intermediate values for the AVG and INF references. I(X) was smallest for the LAP transformation, largest for the LM reference, and at intermediate values for the AVG and INF references. Furthermore, across all references, CI(X) and I(X) reliably distinguished between resting-state conditions (larger values for eyes-open vs. eyes-closed). These findings occurred in the context of the overall expected pattern of resting state PSD. Dipole modeling showed that simulated scalp EEG-level CI(X) and I(X) reflected changes in underlying neural source dependencies, but only for higher levels of integration and with highest accuracy for the LAP transformation. Our observations suggest that the Laplacian-transformation should be preferred for the computation of scalp-level CI(X) and I(X) due to its positive impact on EEG signal quality and statistics, reduction of volume-conduction, and the higher accuracy this provides when estimating scalp-level EEG complexity and integration."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Auditory-Inspired Speech Envelope Extraction Methods for Improved EEG-Based Auditory Attention Detection in a Cocktail Party Scenario",
                    "abstract": "This paper considers the auditory attention detection (AAD) paradigm, where the goal is to determine which of two simultaneous speakers a person is attending to. The paradigm relies on recordings of the listener\u2019s brain activity, e.g., from electroencephalography (EEG). To perform AAD, decoded EEG signals are typically correlated with the temporal envelopes of the speech signals of the separate speakers. In this paper, we study how the inclusion of various degrees of auditory modelling in this speech envelope extraction process affects the AAD performance, where the best performance is found for an auditory-inspired linear filter bank followed by power law compression. These two modelling stages are computationally cheap, which is important for implementation in wearable devices, such as future neuro-steered auditory prostheses. We also introduce a more natural way to combine recordings (over trials and subjects) to train the decoder, which reduces the dependence of the algorithm on regularization parameters. Finally, we investigate the simultaneous design of the EEG decoder and the audio subband envelope recombination weights vector using either a norm-constrained least squares or a canonical correlation analysis, but conclude that this increases computational complexity without improving AAD performance."
                },
                {
                    "title": "EEGNet: a compact convolutional neural network for EEG-based brain\u2013computer interfaces",
                    "abstract": "Objective. Brain\u2013computer interfaces (BCI) enable direct communication with a computer, using neural activity as the control signal. This neural signal is generally chosen from a variety of well-studied electroencephalogram (EEG) signals. For a given BCI paradigm, feature extractors and classifiers are tailored to the distinct characteristics of its expected EEG control signal, limiting its application to that specific signal. Convolutional neural networks (CNNs), which have been used in computer vision and speech recognition to perform automatic feature extraction and classification, have successfully been applied to EEG-based BCIs; however, they have mainly been applied to single BCI paradigms and thus it remains unclear how these architectures generalize to other paradigms. Here, we ask if we can design a single CNN architecture to accurately classify EEG signals from different BCI paradigms, while simultaneously being as compact as possible. Approach. In this work we introduce EEGNet, a compact convolutional neural network for EEG-based BCIs. We introduce the use of depthwise and separable convolutions to construct an EEG-specific model which encapsulates well-known EEG feature extraction concepts for BCI. We compare EEGNet, both for within-subject and cross-subject classification, to current state-of-the-art approaches across four BCI paradigms: P300 visual-evoked potentials, error-related negativity responses (ERN), movement-related cortical potentials (MRCP), and sensory motor rhythms (SMR). Main results. We show that EEGNet generalizes across paradigms better than, and achieves comparably high performance to, the reference algorithms when only limited training data is available across all tested paradigms. In addition, we demonstrate three different approaches to visualize the contents of a trained EEGNet model to enable interpretation of the learned features. Significance. Our results suggest that EEGNet is robust enough to learn a wide variety of interpretable features over a range of BCI tasks. Our models can be found at: https://github.com/vlawhern/arl-eegmodels."
                },
                {
                    "title": "Sleep Stage Classification Using EEG Signal Analysis: A Comprehensive Survey and New Investigation",
                    "abstract": "Sleep specialists often conduct manual sleep stage scoring by visually inspecting the patient\u2019s neurophysiological signals collected at sleep labs. This is, generally, a very difficult, tedious and time-consuming task. The limitations of manual sleep stage scoring have escalated the demand for developing Automatic Sleep Stage Classification (ASSC) systems. Sleep stage classification refers to identifying the various stages of sleep and is a critical step in an effort to assist physicians in the diagnosis and treatment of related sleep disorders. The aim of this paper is to survey the progress and challenges in various existing Electroencephalogram (EEG) signal-based methods used for sleep stage identification at each phase; including pre-processing, feature extraction and classification; in an attempt to find the research gaps and possibly introduce a reasonable solution. Many of the prior and current related studies use multiple EEG channels, and are based on 30 s or 20 s epoch lengths which affect the feasibility and speed of ASSC for real-time applications. Thus, in this paper, we also present a novel and efficient technique that can be implemented in an embedded hardware device to identify sleep stages using new statistical features applied to 10 s epochs of single-channel EEG signals. In this study, the PhysioNet Sleep European Data Format (EDF) Database was used. The proposed methodology achieves an average classification sensitivity, specificity and accuracy of 89.06%, 98.61% and 93.13%, respectively, when the decision tree classifier is applied. Finally, our new method is compared with those in recently published studies, which reiterates the high classification accuracy performance."
                },
                {
                    "title": "Layer Normalization",
                    "abstract": "Training state-of-the-art, deep neural networks is computationally expensive. One way to reduce the training time is to normalize the activities of the neurons. A recently introduced technique called batch normalization uses the distribution of the summed input to a neuron over a mini-batch of training cases to compute a mean and variance which are then used to normalize the summed input to that neuron on each training case. This significantly reduces the training time in feed-forward neural networks. However, the effect of batch normalization is dependent on the mini-batch size and it is not obvious how to apply it to recurrent neural networks. In this paper, we transpose batch normalization into layer normalization by computing the mean and variance used for normalization from all of the summed inputs to the neurons in a layer on a single training case. Like batch normalization, we also give each neuron its own adaptive bias and gain which are applied after the normalization but before the non-linearity. Unlike batch normalization, layer normalization performs exactly the same computation at training and test times. It is also straightforward to apply to recurrent neural networks by computing the normalization statistics separately at each time step. Layer normalization is very effective at stabilizing the hidden state dynamics in recurrent networks. Empirically, we show that layer normalization can substantially reduce the training time compared with previously published techniques."
                },
                {
                    "title": "Gaussian Error Linear Units (GELUs)",
                    "abstract": "We propose the Gaussian Error Linear Unit (GELU), a high-performing neural network activation function. The GELU activation function is $x\\Phi(x)$, where $\\Phi(x)$ the standard Gaussian cumulative distribution function. The GELU nonlinearity weights inputs by their value, rather than gates inputs by their sign as in ReLUs ($x\\mathbf{1}_{x>0}$). We perform an empirical evaluation of the GELU nonlinearity against the ReLU and ELU activations and find performance improvements across all considered computer vision, natural language processing, and speech tasks."
                },
                {
                    "title": "The Temple University Hospital EEG Data Corpus",
                    "abstract": "The electroencephalogram (EEG) is an excellent tool for probing neural function, both in clinical and research environments, due to its low cost, non-invasive nature, and pervasiveness. In the clinic, the EEG is the standard test for diagnosing and characterizing epilepsy and stroke, as well as a host of other trauma and pathology related conditions (Tatum et al., 2007; Yamada and Meng, 2009). In research laboratories, EEG is used to study neural responses to external stimuli, motor planning and execution, and brain-computer interfaces (Lebedev and Nicolelis, 2006; Wang et al., 2013). While human interpretation is still the gold standard for EEG analysis in the clinic, a host of software tools exist to facilitate the process or to make predictive analyses such as seizure prediction. \n \nRecently, a confluence of events has underscored the need for robust EEG tools. First, there has been a renewed push via the White House BRAIN initiative to understand neural function and disease (Weiss, 2013). Secondly, there is an increased awareness on brain injury owing to both the influx of injured warfighters and numerous high-profile athletes found to have chronic brain damage (McKee et al., 2009; Stern et al., 2011). And thirdly, a wave of consumer grade scalp sensors has entered the market, allowing end users to monitor sleep, arousal, and mood (Liao et al., 2012). \n \nIn all these applications, there is a need for robust signal processing tools to analyze the EEG data. Historically, EEG signal processing tools have been devised using either ad hoc heuristic methods, or by training pattern recognition engines on small data sets (Gotman, 1982). These methods have yielded limited results, owing mostly to the fact that brain signals (and EEG in particular) are characterized by great variability, which can only be properly interpreted by building statistical models using massive amounts of data (Alotaiby et al., 2014; Ramgopal et al., 2014). Unfortunately, despite EEG being perhaps the most pervasive modality for acquiring brain signals, there is a severe lack of data in the public domain. For example, the \u201cEEG Motor Movement/Imagery Dataset\u201d (http://www.physionet.org/pn4/eegmmidb/) contains ~1500 recordings of 1 or 2 min duration apiece from 109 subjects (Goldberger et al., 2000; Schalk et al., 2004). The CHB-MIT database contains data from 22 subjects, mostly pediatric (Shoeb, 2009). A database from Karunya University contains 175 16-channel EEGs of duration 10 s (Selvaraj et al., 2014). One of the most extensive databases for supporting epilepsy research is the European Epilepsy Database (http://epilepsy-database.eu/), which contains 250 datasets from 30 unique patients, but sells for \u20ac3000. Other databases, such as ieee.org, contain a wealth of data from more invasive modalities such as electrocorticogram, but little or no EEG. \n \nThis lack of publically available data is ironic considering that hundreds of thousands of EEGs are administered annually in clinical settings around the world. Relatively little of this data is publicly available to the research community in a form that is useful to machine learning research. Massive amounts of EEG data would allow the use of state-of-the-art machine learning algorithms to discover new diagnostics and validate clinical practice. Furthermore, it is desirable that such data be collected in clinical settings, as opposed to tightly controlled research environments, since \u201cclinical-grade\u201d data is inherently more variable with respect to parameters such as electrode location, clinical environment, equipment, and noise. Capturing this variability is critical to the development of robust, high performance technology that has real-world impact. \n \nIn this work, we describe a new corpus, the TUH-EEG Corpus, which is an ongoing data collection effort that has recently released 14 years of clinical EEG data collected at Temple University Hospital. The records have been curated, organized, and paired with textual clinician reports that describe the patients and scans. The corpus is publicly available from the Neural Engineering Data Consortium (www.nedcdata.org) (Picone and Obeid, 2016)."
                },
                {
                    "title": "Investigating Critical Frequency Bands and Channels for EEG-Based Emotion Recognition with Deep Neural Networks",
                    "abstract": "To investigate critical frequency bands and channels, this paper introduces deep belief networks (DBNs) to constructing EEG-based emotion recognition models for three emotions: positive, neutral and negative. We develop an EEG dataset acquired from 15 subjects. Each subject performs the experiments twice at the interval of a few days. DBNs are trained with differential entropy features extracted from multichannel EEG data. We examine the weights of the trained DBNs and investigate the critical frequency bands and channels. Four different profiles of 4, 6, 9, and 12 channels are selected. The recognition accuracies of these four profiles are relatively stable with the best accuracy of 86.65%, which is even better than that of the original 62 channels. The critical frequency bands and channels determined by using the weights of trained DBNs are consistent with the existing observations. In addition, our experiment results show that neural signatures associated with different emotions do exist and they share commonality across sessions and individuals. We compare the performance of deep models with shallow models. The average accuracies of DBN, SVM, LR, and KNN are 86.08%, 83.99%, 82.70%, and 72.60%, respectively."
                },
                {
                    "title": "BCI2000: a general-purpose brain-computer interface (BCI) system",
                    "abstract": "Many laboratories have begun to develop brain-computer interface (BCI) systems that provide communication and control capabilities to people with severe motor disabilities. Further progress and realization of practical applications depends on systematic evaluations and comparisons of different brain signals, recording methods, processing algorithms, output formats, and operating protocols. However, the typical BCI system is designed specifically for one particular BCI method and is, therefore, not suited to the systematic studies that are essential for continued progress. In response to this problem, we have developed a documented general-purpose BCI research and development platform called BCI2000. BCI2000 can incorporate alone or in combination any brain signals, signal processing methods, output devices, and operating protocols. This report is intended to describe to investigators, biomedical engineers, and computer scientists the concepts that the BCI2000 system is based upon and gives examples of successful BCI implementations using this system. To date, we have used BCI2000 to create BCI systems for a variety of brain signals, processing methods, and applications. The data show that these systems function well in online operation and that BCI2000 satisfies the stringent real-time requirements of BCI systems. By substantially reducing labor and cost, BCI2000 facilitates the implementation of different BCI systems and other psychophysiological experiments. It is available with full documentation and free of charge for research or educational purposes and is currently being used in a variety of studies by many research groups."
                },
                {
                    "title": "Learning Topology-Agnostic EEG Representations with Geometry-Aware Modeling",
                    "abstract": "Large-scale pre-training has shown great potential to enhance models on down-stream tasks in vision and language. Developing similar techniques for scalp electroencephalogram (EEG) is suitable since unlabelled data is plentiful. Meanwhile, various sampling channel selections and inherent structural and spatial information bring challenges and avenues to improve existing pre-training strategies further. In order to break boundaries between different EEG resources and facilitate cross-dataset EEG pre-training, we propose to map all kinds of channel selections to a unified topology. We further introduce MMM , a pre-training framework with M ulti-dimensional position encoding, M ulti-level channel hierarchy, M ulti-stage pre-training strategy built on the unified topology to obtain topology-agnostic representations. Experiments demonstrate that our approach yields impressive improvements over previous state-of-the-art techniques on emotional recognition benchmark datasets. 1"
                },
                {
                    "title": "BIOT: Biosignal Transformer for Cross-data Learning in the Wild",
                    "abstract": "Biological signals, such as electroencephalograms (EEG), play a crucial role in numerous clinical applications, exhibiting diverse data formats and quality profiles. Current deep learning models for biosignals (based on CNN, RNN, and Transformers) are typically specialized for specific datasets and clinical settings, limiting their broader applicability. This paper explores the development of a flexible biosignal encoder architecture that can enable pre-training on multiple datasets and fine-tuned on downstream biosignal tasks with different formats. To overcome the unique challenges associated with biosignals of various formats, such as mismatched channels, variable sample lengths, and prevalent missing values, we propose Biosignal Transformer ( BIOT ). The proposed BIOT model can enable cross-data learning with mismatched channels, variable lengths, and missing values by tokenizing different biosignals into unified \"sentences\" structure. Specifically, we tokenize each channel separately into fixed-length segments containing local signal features and then re-arrange the segments to form a long \"sentence\". Channel embeddings and relative position embeddings are added to each segment (viewed as \"token\") to preserve spatio-temporal features. The BIOT model is versatile and applicable to various biosignal learning settings across different datasets, including joint pre-training for larger models. Comprehensive evaluations on EEG, electrocardiogram (ECG), and human activity sensory signals demonstrate that BIOT outperforms robust baselines in common settings and facilitates learning across multiple datasets with different formats. Using CHB-MIT seizure detection task as an example, our vanilla BIOT model shows 3% improvement over baselines in balanced accuracy, and the pre-trained BIOT models (optimized from other data sources) can further bring up to 4% improvements. Our repository is public at https://github"
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Improving Language Understanding by Generative Pre-Training",
                    "abstract": "Natural language understanding comprises a wide range of diverse tasks such as textual entailment, question answering, semantic similarity assessment, and document classi\ufb01cation. Although large unlabeled text corpora are abundant, labeled data for learning these speci\ufb01c tasks is scarce, making it challenging for discriminatively trained models to perform adequately. We demonstrate that large gains on these tasks can be realized by generative pre-training of a language model on a diverse corpus of unlabeled text, followed by discriminative \ufb01ne-tuning on each speci\ufb01c task. In contrast to previous approaches, we make use of task-aware input transformations during \ufb01ne-tuning to achieve effective transfer while requiring minimal changes to the model architecture. We demonstrate the effectiveness of our approach on a wide range of benchmarks for natural language understanding. Our general task-agnostic model outperforms discriminatively trained models that use architectures speci\ufb01cally crafted for each task, signi\ufb01cantly improving upon the state of the art in 9 out of the 12 tasks studied. For instance, we achieve absolute improvements of 8.9% on commonsense reasoning (Stories Cloze Test), 5.7% on question answering (RACE), and 1.5% on textual entailment (MultiNLI)."
                },
                {
                    "title": "Descriptor : A mobile brain-body imaging dataset recorded during treadmill walking with a brain-computer interface",
                    "abstract": "We present a mobile brain-body imaging (MoBI) dataset acquired during treadmill walking in a braincomputer interface (BCI) task. The data were collected from eight healthy subjects, each having three identical trials. Each trial consisted of three conditions: standing, treadmill walking, and treadmill walking with a closed-loop BCI. During the BCI condition, subjects used their brain activity to control a virtual avatar on a screen to walk in real-time. Robust procedures were designed to record lower limb joint angles (bilateral hip, knee, and ankle) using goniometers synchronized with 60-channel scalp electroencephalography (EEG). Additionally, electrooculogram (EOG), EEG electrodes impedance, and digitized EEG channel locations were acquired to aid artifact removal and EEG dipole-source localization. This dataset is unique in that it is the first published MoBI dataset recorded during walking. It is useful in addressing several important open research questions, such as how EEG is coupled with gait cycle during closed-loop BCI, how BCI influences neural activity during walking, and how a BCI decoder may be optimized."
                },
                {
                    "title": "Objective and Subjective Evaluation of Online Error Correction during P300-Based Spelling",
                    "abstract": "Error potentials (ErrP) are alterations of EEG traces following the subject's perception of erroneous feedbacks. They provide a way to recognize misinterpreted commands in brain-computer interfaces (BCI). However, this has been evaluated online in only a couple of studies and mostly with very few subjects. In this study, we implemented a P300-based BCI, including not only online error detection but also, for the first time, automatic correction. We evaluated it in 16 healthy volunteers. Whenever an error was detected, a new decision was made based on the second best guess of a probabilistic classifier. At the group level, correction did neither improve nor deteriorate spelling accuracy. However, automatic correction yielded a higher bit rate than a respelling strategy. Furthermore, the fine examination of interindividual differences in the efficiency of error correction and spelling clearly distinguished between two groups who differed according to individual specificity in ErrP detection. The high specificity group had larger evoked responses and made fewer errors which were corrected more efficiently, yielding a 4% improvement in spelling accuracy and a higher bit rate. Altogether, our results suggest that the more the subject is engaged into the task, the more useful and well accepted the automatic error correction."
                },
                {
                    "title": "Emotion Detection in the Loop from Brain Signals and Facial Images",
                    "abstract": "In this project, we intended to develop techniques for multimodal emotion detection, one modality being brain signals via fNIRS, the second modality being face video and the third modality being the scalp EEG signals. EEG and\nfNIRS provided us with an \u201cinternal\u201d look at the emotion generation processes, while video sequence gave us an \u201cexternal\u201d look on the \u201csame\u201d phenomenon.\nFusions of fNIRS with video and of EEG with fNIRS were considered. Fusion of all three modalities was not considered due to the extensive noise on the EEG signals caused by facial muscle movements, which are required for emotion detection from video sequences.\nBesides the techniques mentioned above, peripheral signals, namely, respiration, cardiac rate, and galvanic skin resistance were also measured from the subjects during \u201cfNIRS + EEG\u201d recordings. These signals provided us with extra information about the emotional state of the subjects.\nThe critical point in the success of this project was to be able to build a \u201cgood\u201d database. Good data acquisition means synchronous data and requires the definition of some specific experimental protocols for emotions elicitation. Thus, we devoted much of our time to data acquisition throughout the workshop, which resulted in a large enough database for\nmaking the first analyses. Results presented in this report should be considered as preliminary. However, they are promising enough to extend the scope of the research."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively apply self-supervised pre-training techniques from large language models to enhance the generalizability and performance of deep learning models for various EEG-based tasks, given the challenges of limited data and variability in EEG signal acquisition?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem could significantly advance the research community's understanding of cross-task learning in EEG analysis, enabling the development of more robust and versatile models that can generalize across different EEG tasks. This could lead to practical applications in clinical settings, such as improved diagnostics for epilepsy, stress detection, and other neurological conditions. By leveraging self-supervised learning, we could also reduce the reliance on large labeled datasets, making EEG research more accessible and efficient.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem include the scarcity of labeled EEG data, which complicates the training of deep learning models. Naive approaches may fail due to the high variability in EEG signal acquisition methods, leading to mismatched channels and lengths across datasets. Additionally, the complexity of EEG signals requires sophisticated models that can capture both spatial and temporal features, which is difficult to achieve without overfitting. The need for expert annotation further limits the availability of training data, making it hard to implement effective self-supervised learning strategies.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on specific EEG tasks without considering the potential for cross-task learning, leading to models that are often proprietary and limited in their applicability. The lack of large, standardized EEG datasets has hindered the development of generalizable models. Additionally, existing approaches have not effectively integrated self-supervised learning techniques from natural language processing into EEG analysis. Our approach aims to bridge this gap by applying reconstruction ideas from large language models to EEG data, thus improving upon prior work by enabling a more flexible and scalable learning framework.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves pre-training neural Transformers using reconstruction techniques adapted for EEG data. We will utilize a diverse set of publicly available EEG datasets to create a robust training framework. The evaluation metrics will include classification accuracy and generalization performance across multiple EEG tasks. We expect that our approach will yield models capable of effectively learning from limited data, demonstrating improved performance in various EEG applications, and facilitating cross-task learning that has been previously unattain"
            }
        },
        "author_data": {
            "5be4d13b-603f-4108-868f-940de7d155b4": {
                "pk": "5be4d13b-603f-4108-868f-940de7d155b4",
                "name": "Wei-Bang Jiang",
                "collaborators": [
                    "Bao-Liang Lu",
                    "Wei-Long Zheng",
                    "Hui Zheng",
                    "Haiteng Wang",
                    "Zhongtao Chen",
                    "Li He",
                    "Pei-Yang Lin",
                    "Peng-Hu Wei",
                    "Guo-Guang Zhao",
                    "Yunzhe Liu"
                ],
                "domain": [
                    "Brain-Computer Interface",
                    "Emotion Recognition",
                    "Electroencephalography",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "Du-IN-v2: Unleashing the Power of Vector Quantization for Decoding Cognitive States from Intracranial Neural Signals",
                        "abstract": "While invasive brain-computer interfaces have shown promise for high-performance speech decoding under medical use, the potential of intracranial stereoElectroEn-cephaloGraphy (sEEG), which causes less damage to patients, remains underex-plored. With the rapid progress in representation learning, leveraging abundant pure recordings to further enhance speech decoding becomes increasingly attractive. However, some popular methods pre-train temporal models based on brain-level tokens, overlooking the brain\u2019s desynchronization nature; others pre-train spatial-temporal models based on channel-level tokens, yet fail to evaluate them on more challenging tasks, e.g., speech decoding, which demands intricate processing in specific brain regions. To tackle these issues, we introduce a general pre-training framework for speech decoding \u2013 Du-IN-v2, which can extract contextual embeddings based on region-level tokens through discrete codex-guided mask modeling. To further push its limits, we propose Decoupling Product Quantization (DPQ), where different codexes are designed to extract different parts of brain dynamics. Our model achieves SOTA performance on both the 61-word classification task and the 49-syllable sequence classification task, surpassing all baselines. Model comparison and ablation studies reveal that our design choices, including (i) temporal modeling based on region-level tokens by utilizing 1D depthwise convolution to fuse channels in vSMC and STG regions and (ii) self-supervision by discrete decoupling codex-guided mask modeling, significantly contribute to these performances. Collectively, our approach, inspired by neuroscience findings, capitalizing on region-level representations from specific brain regions, is suitable for invasive brain modeling. It marks a promising neuro-inspired AI approach in BCI."
                    },
                    {
                        "title": "Functional Emotion Transformer for EEG-Assisted Cross-Modal Emotion Recognition",
                        "abstract": "Multimodal emotion recognition based on electroencephalography (EEG) and eye movements has attracted increasing attention due to their high performance and complementary properties. However, there are two challenges that hinder its practical applications: the inconvenient EEG data collection and high-cost data annotation. In contrast, eye movements are convenient to obtain and process in real scenarios. To combine high performance of EEG and easy setups of eye tracking, we propose a novel EEG-assisted Contrastive Learning Framework with a Functional Emotion Transformer (ECO-FET) for cross-modal emotion recognition. ECO-FET leverages both the functional brain connectivity and the spectral-spatial-temporal domain of EEG signals simultaneously, which dramatically benefit the learning of eye movements. The whole process consists of three phases: pre-training, test, and fine-tuning. ECO-FET exploits the complementary information provided by multiple modalities during pre-training in order to improve the performance of unimodal models. In the pre-training phase, unlabeled EEG and eye movement data are fed into the model to contrastively learn the emotional latent representations between the two modalities, while in the test phase, eye movements and few labeled EEG samples are used to predict different emotions. Experimental results on three public datasets demonstrate that ECO-FET surpasses the state-of-the-art dramatically."
                    },
                    {
                        "title": "CEMOAE: A Dynamic Autoencoder with Masked Channel Modeling for Robust EEG-Based Emotion Recognition",
                        "abstract": "Emotion recognition through electroencephalography (EEG) has been an area of active research, but the inherent sensitivity of EEG signals to noise and artifacts poses significant challenges, especially in real-world settings. These complications often necessitate the removal of corrupted channels, making it crucial to develop robust models capable of maintaining performance even when few channels are available. To address this, we propose the Corrupted EMOtion AutoEncoder (CEMOAE), an innovative approach that leverages masked channel modeling to maintain robust performance, achieved through three components: masked autoencoder pretraining for robust representation learning, random masked auxiliary task for implicit modeling of channel corruption, and masked auto-repair to explicitly narrow the data distribution gap between high-quality and corrupted EEG signals. Specifically, we first pretrain a masked autoencoder with the dynamic masking strategy for feature extractor initialization and channel recovery. During the finetuning stage, we mask EEG data using the auxiliary task to mimic real-world EEG corruption. We then employ the pretrained autoencoder to repair these signals and finetune the feature extractor for emotion recognition. Experiments on the SEED dataset demonstrate that CEMOAE achieves SOTA performance for emotion recognition under the random channel corruption simulation, validating the effectiveness of the proposed techniques."
                    },
                    {
                        "title": "Du-IN: Discrete units-guided mask modeling for decoding speech from Intracranial Neural signals",
                        "abstract": "Invasive brain-computer interfaces with Electrocorticography (ECoG) have shown promise for high-performance speech decoding in medical applications, but less damaging methods like intracranial stereo-electroencephalography (sEEG) remain underexplored. With rapid advances in representation learning, leveraging abundant recordings to enhance speech decoding is increasingly attractive. However, popular methods often pre-train temporal models based on brain-level tokens, overlooking that brain activities in different regions are highly desynchronized during tasks. Alternatively, they pre-train spatial-temporal models based on channel-level tokens but fail to evaluate them on challenging tasks like speech decoding, which requires intricate processing in specific language-related areas. To address this issue, we collected a well-annotated Chinese word-reading sEEG dataset targeting language-related brain networks from 12 subjects. Using this benchmark, we developed the Du-IN model, which extracts contextual embeddings based on region-level tokens through discrete codex-guided mask modeling. Our model achieves state-of-the-art performance on the 61-word classification task, surpassing all baselines. Model comparisons and ablation studies reveal that our design choices, including (i) temporal modeling based on region-level tokens by utilizing 1D depthwise convolution to fuse channels in the lateral sensorimotor cortex (vSMC) and superior temporal gyrus (STG) and (ii) self-supervision through discrete codex-guided mask modeling, significantly contribute to this performance. Overall, our approach -- inspired by neuroscience findings and capitalizing on region-level representations from specific brain regions -- is suitable for invasive brain modeling and represents a promising neuro-inspired AI approach in brain-computer interfaces."
                    },
                    {
                        "title": "NeuroLM: A Universal Multi-task Foundation Model for Bridging the Gap between Language and EEG Signals",
                        "abstract": "Recent advancements for large-scale pre-training with neural signals such as electroencephalogram (EEG) have shown promising results, significantly boosting the development of brain-computer interfaces (BCIs) and healthcare. However, these pre-trained models often require full fine-tuning on each downstream task to achieve substantial improvements, limiting their versatility and usability, and leading to considerable resource wastage. To tackle these challenges, we propose NeuroLM, the first multi-task foundation model that leverages the capabilities of Large Language Models (LLMs) by regarding EEG signals as a foreign language, endowing the model with multi-task learning and inference capabilities. Our approach begins with learning a text-aligned neural tokenizer through vector-quantized temporal-frequency prediction, which encodes EEG signals into discrete neural tokens. These EEG tokens, generated by the frozen vector-quantized (VQ) encoder, are then fed into an LLM that learns causal EEG information via multi-channel autoregression. Consequently, NeuroLM can understand both EEG and language modalities. Finally, multi-task instruction tuning adapts NeuroLM to various downstream tasks. We are the first to demonstrate that, by specific incorporation with LLMs, NeuroLM unifies diverse EEG tasks within a single model through instruction tuning. The largest variant NeuroLM-XL has record-breaking 1.7B parameters for EEG signal processing, and is pre-trained on a large-scale corpus comprising approximately 25,000-hour EEG data. When evaluated on six diverse downstream datasets, NeuroLM showcases the huge potential of this multi-task learning paradigm."
                    },
                    {
                        "title": "Multimodal Adaptive Emotion Transformer with Flexible Modality Inputs on A Novel Dataset with Continuous Labels",
                        "abstract": "Emotion recognition from physiological signals is a topic of widespread interest, and researchers continue to develop novel techniques for perceiving emotions. However, the emergence of deep learning has highlighted the need for high-quality emotional datasets to accurately decode human emotions. In this study, we present a novel multimodal emotion dataset that incorporates electroencephalography (EEG) and eye movement signals to systematically explore human emotions. Seven basic emotions (happy, sad, fear, disgust, surprise, anger, and neutral) are elicited by a large number of 80 videos and fully investigated with continuous labels that indicate the intensity of the corresponding emotions. Additionally, we propose a novel Multimodal Adaptive Emotion Transformer (MAET), that can flexibly process both unimodal and multimodal inputs. Adversarial training is utilized in MAET to mitigate subject discrepancy, which enhances domain generalization. Our extensive experiments, encompassing both subject-dependent and cross-subject conditions, demonstrate MAET's superior performance in handling various inputs. The filtering of data for high emotional evocation using continuous labels proved to be effective in the experiments. Furthermore, the complementary properties between EEG and eye movements are observed. Our code is available at https://github.com/935963004/MAET."
                    },
                    {
                        "title": "Identifying sex differences in EEG-based emotion recognition using graph convolutional network with attention mechanism",
                        "abstract": "Objective. Sex differences in emotions have been widely perceived via self-reports, peripheral physiological signals and brain imaging techniques. However, how sex differences are reflected in the electroencephalography (EEG) neural patterns of emotions remains unresolved. In this paper, we detect sex differences in emotional EEG patterns, investigate the consistency of such differences in various emotion datasets across cultures, and study how sex as a factor affects the performance of EEG-based emotion recognition models. Approach. We thoroughly assess sex differences in emotional EEG patterns on five public datasets, including SEED, SEED-IV, SEED-V, DEAP and DREAMER, systematically examine the sex-specific EEG patterns for happy, sad, fearful, disgusted and neutral emotions, and implement deep learning models for sex-specific emotion recognition. Main results. (1) Sex differences exist in various emotion types and both Western and Eastern cultures; (2) The emotion patterns of females are more stable than those of males, and the patterns of happiness from females are in sharp contrast with the patterns of sadness, fear and disgust, while the energy levels are more balanced for males; (3) The key features for emotion recognition are mainly located at the frontal and temporal sites for females and distributed more evenly over the whole brain for males, and (4) the same-sex emotion recognition models outperform the corresponding cross-sex models. Significance. These findings extend efforts to characterize sex differences in emotional brain activation, provide new physiological evidence for sex-specific emotion processing, and reinforce the message that sex differences should be carefully considered in affective research and precision medicine."
                    },
                    {
                        "title": "Elastic Graph Transformer Networks for EEG-Based Emotion Recognition",
                        "abstract": "Electroencephalogram (EEG) has been applied in emotion recognition due to excellent temporal resolution with less competitive spatial resolution. This leads to the consequence that the majority of EEG-based emotion recognition models emphasize on exploiting temporal features while ignoring the efficient information provided by spatial resolution. To extract more informative representations, we propose an elastic Graph Transformer network for emotion recognition (EmoGT) inspired by the advantages of Transformer in time-series analysis and the superior performance of graph convolutional networks in topological analysis. Moreover, it is able to be flexibly expanded to cope with multimodal inputs by employing specially designed structures. Experimental results on 3 public datasets demonstrate that our models outperform the state-of-the-art results by 3% on average in both single and multimodal cases, indicating the effectiveness of utilizing temporal and spatial information simultaneously."
                    },
                    {
                        "title": "Increasing the Stability of EEG-based Emotion Recognition with a Variant of Neural Processes",
                        "abstract": "Electroencephalography (EEG) signals provide an incredible promising access to decode human emotions in affective computing. Many approaches have been applied to building EEG-based affective models and much endeavor is made to improve the performance of affective models, where test data has similar quality as training data. However, due to the strong sensitivity of EEG to external factors such as body movement and electromagnetic interference, EEG signals usually have a lot of noise and subjects have to remain as motionless as possible in a quiet environment, which is difficult to be satisfied in real applications and severely influences the user experience. To deal with this problem, a common way is to drop those channels with intensive noise. However, this results in the loss of critical information and neither existing machine learning (ML) nor deep learning (DL) approaches can handle this situation well, especially when too many channels are missed. In this paper, we propose a robust variant of the neural processes model and evaluate the stability of our model under various circumstances to simulate random data corruption in real applications. We conduct two categories of experiments by controlling the number and the places of missing channels separately and compare with classical ML and DL models. The results demonstrate that the performance of our model significantly outperforms the existing models, even using only five channels. We also explore the critical brain regions in the current EEG electrode distribution. Final performance manifests our model has extreme stability in dealing with intractable situations and sheds light on the widespread usage of portable EEG-based affective computing."
                    },
                    {
                        "title": "Emotion Transformer Fusion: Complementary Representation Properties of EEG and Eye Movements on Recognizing Anger and Surprise",
                        "abstract": "Emotion recognition plays an important role in di-agnosing and treating many mental disorders as well as affective computing. Among six basic emotions, anger and surprise are relatively hard to be elicited in lab settings, and the complementary representation properties of encephalography (EEG) and eye movement signals on recognizing anger and surprise emotions remain unknown. Although the transformer architecture has the ability of parallelism which avoids many sequential operations as recurrent and convolutional layers, the knowledge of its performance and effectiveness on multimodal emotion recognition from EEG and eye movement signals is limited. To tackle these issues, we elaborately design the experiment and stimuli materials to effectively elicit surprise, anger, and neutral emotions, and propose an Emotion Transformer Fusion (ETF) model based on pure attention mechanism. Results of extensive experiments with multiple models on our dataset indicate that the complementary information of EEG and eye movements significantly improves the performance of discriminating anger, surprise and neutral emotions. Meanwhile, our proposed architecture outperforms baseline models with higher parallelism, which proves the capability of Transformer based architecture on multimodal emotion recognition with EEG and eye movement signals."
                    },
                    {
                        "title": "Discriminating Surprise and Anger from EEG and Eye Movements with a Graph Network",
                        "abstract": "Emotion recognition based on EEG and eye movement signals has been studied extensively due to the reliability and stability of signals. The separability of four basic emotions, happy, sad, disgust and fear, has been systematically studied in the existing work. However, there is less research on the emotions of anger and surprise since they are more difficult to be elicited in lab settings. This paper investigates the discrimination ability of EEG and eye movement signals for surprise and anger. To this end, we design a stimulus paradigm that can effectively elicit surprise and anger. We propose a novel Graph Convolutional Network with Channel Attention (GCNCA) to classify three emotions, anger, surprise and neutrality. Experimental results indicate that: a) the proposed GCNCA model has an excellent classification accuracy of 86.47% using EEG and 84.22% using eye movement signals, which are better than other baseline methods; b) EEG and eye movements have a good ability to discriminate surprise and anger, while EEG performs better than eye movements; c) the high-frequency bands of EEG are more distinguishable on classifying surprise and anger than the low-frequency bands; d) there are some differences in neural patterns between surprise and anger, meanwhile critical channels and channel connections of EEG are found."
                    },
                    {
                        "title": "Network for End-to-End Robust EEG-Based Emotion Recognition",
                        "abstract": ". Emotion recognition based on electroencephalography (EEG) is attracting more and more interest in a\ufb00ective computing. Previous studies have predominantly relied on manually extracted features from EEG signals. It remains largely unexplored in the utilization of raw EEG signals, which contain more temporal information but present a signi\ufb01cant challenge due to their abundance of redundant data and susceptibility to contamination from other physiological signals, such as electrooculography (EOG) and electromyography (EMG). To cope with the high dimensionality and noise interference in end-to-end EEG-based emotion recognition tasks, we introduce a Two-Stream Spectral-Temporal Denoising Network (TS-STDN) which takes into account the spectral and temporal aspects of EEG signals. Moreover, two U-net modules are adopted to reconstruct clean EEG signals in both spectral and temporal domains while extracting discriminative features from noisy data for classifying emotions. Extensive experiments are conducted on two public datasets, SEED and SEED-IV, with the original EEG signals and the noisy EEG signals contaminated by EMG signals. Compared to the baselines, our TS-STDN model exhibits a notable improvement in accuracy, demonstrating an increase of 6% and 8% on the clean data and 11% and 10% on the noisy data, which shows the robustness of the model."
                    }
                ]
            },
            "94b85f6e-2e01-4f51-8d2e-e147ea8a1b30": {
                "pk": "94b85f6e-2e01-4f51-8d2e-e147ea8a1b30",
                "name": "Li-Ming Zhao",
                "collaborators": [
                    "Bao-Liang Lu",
                    "Wei-Long Zheng",
                    "Rui Li",
                    "Ziyi Li",
                    "Cheng Fei",
                    "Xue Yan",
                    "Tian-Fang Ma",
                    "Luyu Liu",
                    "Dan Peng",
                    "Yong Lu"
                ],
                "domain": [
                    "Emotion Recognition",
                    "EEG",
                    "Multimodal Learning",
                    "Affective Computing"
                ],
                "publications": [
                    {
                        "title": "Temporal-Spatial Prediction: Pre-Training on Diverse Datasets for EEG Classification",
                        "abstract": "Electroencephalogram (EEG) classification tasks have received increasing attention because its high application value. Meanwhile, the great success of general pre-training models in language processing areas inspires us to excavate the potential of an EEG pre-trained model. This model is expected to adapt to diverse downstream tasks. However, current studies either ignore the temporal or spatial domain in EEG signals, or only use single datasets in pre-training. The proposed Temporal-Spatial Prediction (TSP) model effectively solve these issues. Specifically, the output of the TSP encoder serves as the input of two tasks: spatial prediction, i.e., masked autoencoder, and temporal prediction, i.e, contrastive predictive coding. In addition, in order to provide more diverse information and thus benefit the downstream fine-tuning, we pre-train TSP on six large EEG datasets with four different numbers of channels. Results on three public downstream datasets SEED, SEED-IV, TUEV demonstrate that TSP achieves the state-of-the-art performance on different EEG classification tasks. In addition, according to the ablation experiments, TSP performs better than the single-domain method, i.e. Temporal Prediction (TP) model and Spatial Prediction (SP) model."
                    },
                    {
                        "title": "Detecting Major Depression Disorder with Multiview Eye Movement Features in a Novel Oil Painting Paradigm",
                        "abstract": "Major Depressive Disorder (MDD) is a debilitating condition marked by persistent low mood, reduced interest, cognitive impairments, and vegetative neurological symptoms such as sleep and appetite disturbances. In this paper, we collected eye movement signals from 40 patients diagnosed with MDD and 40 healthy controls to study the relation between eye movements and cognitive processes for depression detection. The eye movement data were captured during a novel emotional cognition task using oil paintings. Subsequently, the data were transformed into multiview eye movement features, including heatmaps, trajectories, and statistical vectors. Rigorous statistical analyses were then conducted on these features to identify significant patterns and correlations between eye movements and depressive symptoms. A multiview invariant & specific eye movement model (MISEYE) was proposed to fuse different eye movement features. The proposed achieved an accuracy rate of 79.88% in depression detection. This performance surpassed not only the outcomes of single-mode approaches and combinations of any two features but also outperformed other fusion methodologies. These findings not only shed light on the intricate relationship between eye movement patterns and MDD but also underscore the potential of eye-tracking technology in psychiatric research."
                    },
                    {
                        "title": "Tagging Continuous Labels for EEG-based Emotion Classification",
                        "abstract": "EEG-based emotion classification has long been a critical task in the field of affective brain-computer interface (aBCI). The majority of leading researches construct supervised learning models based on labeled datasets. Several datasets have been released, including different kinds of emotions while utilizing various forms of stimulus materials. However, they adopt discrete labeling methods, in which the EEG data collected during the same stimulus material are given a same label. These methods neglect the fact that emotion changes continuously, and mislabeled data possibly exist. The imprecision of discrete labels may hinder the progress of emotion classification in concerned works. Therefore, we develop an efficient system in this paper to support continuous labeling by giving each sample a unique label, and construct a continuously labeled EEG emotion dataset. Using our dataset with continuous labels, we demonstrate the superiority of continuous labeling in emotion classification through experiments on several classification models. We further utilize the continuous labels to identify the EEG features under induced and non-induced emotions in both our dataset and a public dataset. Our experimental results reveal the learnability and generality of the relation between the EEG features and their continuous labels."
                    },
                    {
                        "title": "EEG-Eye Movements Cross-Modal Decision Confidence Measurement with Generative Adversarial Networks",
                        "abstract": "Decision confidence is an individual's feeling of correctness or optimization when making a decision. Various physiological signals, including electroencephalography (EEG) and eye movements have been studied extensively in measuring levels of decision confidence in humans. While multimodal fusion generally performs better than single-modal approaches, it requires data from different modalities at a greater cost. In particular, collection of EEG data is more complicated and time consuming while eye movement signals are much easier to acquire. To tackle this problem, we propose a cross-modal method based on generative adversarial learning. In our method, the intrinsic relationship between eye movement and EEG features in a high-level feature space can be learned in the training phase, and then we can obtain multimodal information during the test phase when only eye movements are available as inputs. Experimental results on the SEED-VPDC dataset demonstrate that our proposed method outperforms single-modal methods trained and tested only on eye movement signals with an improvement of approximately 5.43% in accuracy, and maintains competitive performance in comparison with multimodal methods. Our cross-modal approach requires only eye movements as inputs and reduces reliance on EEG data, making the decision confidence measurement more applicable and practicable."
                    },
                    {
                        "title": "A Cross-modality Deep Learning Method for Measuring Decision Confidence from Eye Movement Signals",
                        "abstract": "Electroencephalography (EEG) signals can effectively measure the level of human decision confidence. However, it is difficult to acquire EEG signals in practice due to the ex-pensive cost and complex operation, while eye movement signals are much easier to acquire and process. To tackle this problem, we propose a cross-modality deep learning method based on deep canoncial correlation analysis (CDCCA) to transform each modality separately and coordinate different modalities into a hyperspace by using specific canonical correlation analysis constraints. In our proposed method, only eye movement signals are used as inputs in the test phase and the knowledge from EEG signals is learned in the training stage. Experimental results on two human decision confidence datasets demonstrate that our proposed method achieves advanced performance compared with the existing single-modal approaches trained and tested on eye movement signals and maintains a competitive accuracy in comparison with multimodal models."
                    },
                    {
                        "title": "Plug-and-Play Domain Adaptation for Cross-Subject EEG-based Emotion Recognition",
                        "abstract": "Human emotion decoding in affective brain-computer interfaces suffers a major setback due to the inter-subject variability of electroencephalography (EEG) signals. Existing approaches usually require amassing extensive EEG data of each new subject, which is prohibitively time-consuming along with poor user experience. To tackle this issue, we divide EEG representations into private components specific to each subject and shared emotional components that are universal to all subjects. According to this representation partition, we propose a plug-and-play domain adaptation method for dealing with the inter-subject variability. In the training phase, subject-invariant emotional representations and private components of source subjects are separately captured by a shared encoder and private encoders. Furthermore, we build one emotion classifier on the shared partition and subjects' individual classifiers on the combination of these two partitions. In the calibration phase, the model only requires few unlabeled EEG data from incoming target subjects to model their private components. Therefore, besides the shared emotion classifier, we have another pipeline to use the knowledge of source subjects through the similarity of private components. In the test phase, we integrate predictions of the shared emotion classifier with those of individual classifiers ensemble after modulation by similarity weights. Experimental results on the SEED dataset show that our model greatly shortens the calibration time within a minute while maintaining the recognition accuracy, all of which make emotion decoding more generalizable and practicable."
                    },
                    {
                        "title": "Simplifying Multimodal Emotion Recognition with Single Eye Movement Modality",
                        "abstract": "Multimodal emotion recognition has long been a popular topic in affective computing since it significantly enhances the performance compared with that of a single modality. Among all, the combination of electroencephalography (EEG) and eye movement signals is one of the most attractive practices due to their complementarity and objectivity. However, the high cost and inconvenience of EEG signal acquisition severely hamper the popularization of multimodal emotion recognition in practical scenarios, while eye movement signals are much easier to acquire. To increase the feasibility and the generalization ability of emotion decoding without compromising the performance, we propose a generative adversarial network-based framework. In our model, a single modality of eye movements is used as input and it is capable of mapping the information onto multimodal features. Experimental results on SEED series datasets with different emotion categories demonstrate that our model with multimodal features generated by the single eye movement modality maintains competitive accuracies compared to those with multimodality input and drastically outperforms those single-modal emotion classifiers. This illustrates that the model has the potential to reduce the dependence on multimodalities without sacrificing performance which makes emotion recognition more applicable and practicable."
                    },
                    {
                        "title": "A Novel Experiment Setting for Cross-subject Emotion Recognition",
                        "abstract": "Recently, cross-subject emotion recognition attracts widespread attention. The current emotional experiments mainly use video clips of different emotions as stimulus materials, but the videos watched by different subjects are the same, which may introduce the same noise pattern in the collected data. However, the traditional experiment settings for cross-subject emotion recognition models couldn\u2019t eliminate the impact of same video clips on recognition results, which may lead to a bias on classification. In this paper, we propose a novel experiment setting for cross-subject emotion recognition. We evaluate different experiment settings on four public emotion datasets, DEAP, SEED, SEED-IV and SEED-V. The experimental results demonstrate the deficiencies of the traditional experiment settings and the advantages of our proposed experiment setting."
                    },
                    {
                        "title": "Discriminating Surprise and Anger from EEG and Eye Movements with a Graph Network",
                        "abstract": "Emotion recognition based on EEG and eye movement signals has been studied extensively due to the reliability and stability of signals. The separability of four basic emotions, happy, sad, disgust and fear, has been systematically studied in the existing work. However, there is less research on the emotions of anger and surprise since they are more difficult to be elicited in lab settings. This paper investigates the discrimination ability of EEG and eye movement signals for surprise and anger. To this end, we design a stimulus paradigm that can effectively elicit surprise and anger. We propose a novel Graph Convolutional Network with Channel Attention (GCNCA) to classify three emotions, anger, surprise and neutrality. Experimental results indicate that: a) the proposed GCNCA model has an excellent classification accuracy of 86.47% using EEG and 84.22% using eye movement signals, which are better than other baseline methods; b) EEG and eye movements have a good ability to discriminate surprise and anger, while EEG performs better than eye movements; c) the high-frequency bands of EEG are more distinguishable on classifying surprise and anger than the low-frequency bands; d) there are some differences in neural patterns between surprise and anger, meanwhile critical channels and channel connections of EEG are found."
                    },
                    {
                        "title": "Classification of Five Emotions from EEG and Eye Movement Signals: Complementary Representation Properties",
                        "abstract": "Recently, various multimodal approaches to enhancing the performance of affective models have been developed. In this paper, we investigate the complementary representation properties of EEG and eye movement signals on classification for five human emotions: happy, sad, fear, disgust, and neutral. We compare the performance of single modality and two different modality fusion approaches. The results indicate that EEG is superior to eye movements in classifying happy, sad and disgust emotions, whereas eye movements outperform EEG in recognizing fear and neutral emotions. Compared with eye movements, EEG has the advantage of classifying the five emotions, with the mean accuracies of 69.50% and 59.81%, respectively. Due to the complementary representation properties, the modality fusion with bimodal deep auto-encoder significantly improves the classification accuracy to 79.71%. Furthermore, we study the neural patterns of five emotion states and the recognition performance of different eye movement features. The results reveal that five emotions have distinguishable neural patterns and pupil diameter has a relatively high discrimination ability than the other eye movement features."
                    },
                    {
                        "title": "Multimodal Emotion Recognition from Eye Image, Eye Movement and EEG Using Deep Neural Networks",
                        "abstract": "In consideration of the complexity of recording electroencephalography(EEG), some researchers are trying to find new features of emotion recognition. In order to investigate the potential of eye tracking glasses for multimodal emotion recognition, we collect and use eye images to classify five emotions along with eye movements and EEG. We compare four combinations of the three different types of data and two kinds of fusion methods, feature level fusion and Bimodal Deep AutoEncoder (BDAE). According to the three-modality fusion features generated by BDAE, the best mean accuracy of 79.63% is achieved. By analyzing the confusion matrices, we find that the three modalities can provide complementary information for recognizing five emotions. Meanwhile, the experimental results indicate that the classifiers with eye image and eye movement fusion features can achieve a comparable classification accuracy of 71.99%."
                    },
                    {
                        "title": "Generating Multimodal Features for Emotion Classification from Eye Movement Signals",
                        "abstract": ". In recent years, multimodal signals such as EEG and eye movements have been widely used and achieved a better performance than single modality in emotion classi\ufb01cation. However, inconvenience in EEG signals collecting procedure still remains a key problem in practical applications. In this paper, we propose a novel method for generating multimodal features from eye movement signals. As a result, we could simplify classi\ufb01cation procedure and use EEG and eye movement signals in training stage and eye movement signals in test stage. We evaluate the e\ufb00ectiveness of our proposed method on classi\ufb01cation of \ufb01ve human emotions, which include disgust, fear, sad, neutral and happy. The experimental results indicate that our proposed method could really generate reliable multimodal representations and has a nearly comparable performance (72.80%), with multimodal models (79.70%), while it also outperforms using single eye movement signals (59.66%) and EEG signals (68.58%)."
                    }
                ]
            },
            "147bb94d-77ac-40d0-9101-817e6d0bbe72": {
                "pk": "147bb94d-77ac-40d0-9101-817e6d0bbe72",
                "name": "Bao-Liang Lu",
                "collaborators": [
                    "Wei-Long Zheng",
                    "Dan Peng",
                    "Wei-Bang Jiang",
                    "Ziyi Li",
                    "Li-Ming Zhao",
                    "Luyu Liu",
                    "Yong Lu",
                    "Yun Luo",
                    "Wei Liu",
                    "Xuan-Hao Liu"
                ],
                "domain": [
                    "EEG",
                    "Emotion Recognition",
                    "Deep Learning",
                    "Multimodal Analysis"
                ],
                "publications": [
                    {
                        "title": "Temporal-Spatial Prediction: Pre-Training on Diverse Datasets for EEG Classification",
                        "abstract": "Electroencephalogram (EEG) classification tasks have received increasing attention because its high application value. Meanwhile, the great success of general pre-training models in language processing areas inspires us to excavate the potential of an EEG pre-trained model. This model is expected to adapt to diverse downstream tasks. However, current studies either ignore the temporal or spatial domain in EEG signals, or only use single datasets in pre-training. The proposed Temporal-Spatial Prediction (TSP) model effectively solve these issues. Specifically, the output of the TSP encoder serves as the input of two tasks: spatial prediction, i.e., masked autoencoder, and temporal prediction, i.e, contrastive predictive coding. In addition, in order to provide more diverse information and thus benefit the downstream fine-tuning, we pre-train TSP on six large EEG datasets with four different numbers of channels. Results on three public downstream datasets SEED, SEED-IV, TUEV demonstrate that TSP achieves the state-of-the-art performance on different EEG classification tasks. In addition, according to the ablation experiments, TSP performs better than the single-domain method, i.e. Temporal Prediction (TP) model and Spatial Prediction (SP) model."
                    },
                    {
                        "title": "Focused State Recognition Using EEG with Eye Movement-Assisted Annotation",
                        "abstract": "With the rapid advancement in machine learning, the recognition and analysis of brain activity based on EEG and eye movement signals have attained a high level of sophistication. Utilizing deep learning models for learning EEG and eye movement features proves effective in classifying brain activities. A focused state indicates intense concentration on a task or thought. Distinguishing focused and unfocused states can be achieved through eye movement behaviors, reflecting variations in brain activities. By calculating binocular focusing point disparity in eye movement signals and integrating relevant EEG features, we propose an annotation method for focused states. The resulting comprehensive dataset, derived from raw data processed through a bio-acquisition device, includes both EEG features and focused labels annotated by eye movements. Extensive training and testing on several deep learning models, particularly the Transformer, yielded a 90.16% accuracy on the subject-dependent experiments. The validity of this approach was demonstrated, with cross-subject experiments, key frequency band and brain region analyses confirming its generalizability and providing physiological explanations."
                    },
                    {
                        "title": "Functional Emotion Transformer for EEG-Assisted Cross-Modal Emotion Recognition",
                        "abstract": "Multimodal emotion recognition based on electroencephalography (EEG) and eye movements has attracted increasing attention due to their high performance and complementary properties. However, there are two challenges that hinder its practical applications: the inconvenient EEG data collection and high-cost data annotation. In contrast, eye movements are convenient to obtain and process in real scenarios. To combine high performance of EEG and easy setups of eye tracking, we propose a novel EEG-assisted Contrastive Learning Framework with a Functional Emotion Transformer (ECO-FET) for cross-modal emotion recognition. ECO-FET leverages both the functional brain connectivity and the spectral-spatial-temporal domain of EEG signals simultaneously, which dramatically benefit the learning of eye movements. The whole process consists of three phases: pre-training, test, and fine-tuning. ECO-FET exploits the complementary information provided by multiple modalities during pre-training in order to improve the performance of unimodal models. In the pre-training phase, unlabeled EEG and eye movement data are fed into the model to contrastively learn the emotional latent representations between the two modalities, while in the test phase, eye movements and few labeled EEG samples are used to predict different emotions. Experimental results on three public datasets demonstrate that ECO-FET surpasses the state-of-the-art dramatically."
                    },
                    {
                        "title": "CEMOAE: A Dynamic Autoencoder with Masked Channel Modeling for Robust EEG-Based Emotion Recognition",
                        "abstract": "Emotion recognition through electroencephalography (EEG) has been an area of active research, but the inherent sensitivity of EEG signals to noise and artifacts poses significant challenges, especially in real-world settings. These complications often necessitate the removal of corrupted channels, making it crucial to develop robust models capable of maintaining performance even when few channels are available. To address this, we propose the Corrupted EMOtion AutoEncoder (CEMOAE), an innovative approach that leverages masked channel modeling to maintain robust performance, achieved through three components: masked autoencoder pretraining for robust representation learning, random masked auxiliary task for implicit modeling of channel corruption, and masked auto-repair to explicitly narrow the data distribution gap between high-quality and corrupted EEG signals. Specifically, we first pretrain a masked autoencoder with the dynamic masking strategy for feature extractor initialization and channel recovery. During the finetuning stage, we mask EEG data using the auxiliary task to mimic real-world EEG corruption. We then employ the pretrained autoencoder to repair these signals and finetune the feature extractor for emotion recognition. Experiments on the SEED dataset demonstrate that CEMOAE achieves SOTA performance for emotion recognition under the random channel corruption simulation, validating the effectiveness of the proposed techniques."
                    },
                    {
                        "title": "Detecting Major Depression Disorder with Multiview Eye Movement Features in a Novel Oil Painting Paradigm",
                        "abstract": "Major Depressive Disorder (MDD) is a debilitating condition marked by persistent low mood, reduced interest, cognitive impairments, and vegetative neurological symptoms such as sleep and appetite disturbances. In this paper, we collected eye movement signals from 40 patients diagnosed with MDD and 40 healthy controls to study the relation between eye movements and cognitive processes for depression detection. The eye movement data were captured during a novel emotional cognition task using oil paintings. Subsequently, the data were transformed into multiview eye movement features, including heatmaps, trajectories, and statistical vectors. Rigorous statistical analyses were then conducted on these features to identify significant patterns and correlations between eye movements and depressive symptoms. A multiview invariant & specific eye movement model (MISEYE) was proposed to fuse different eye movement features. The proposed achieved an accuracy rate of 79.88% in depression detection. This performance surpassed not only the outcomes of single-mode approaches and combinations of any two features but also outperformed other fusion methodologies. These findings not only shed light on the intricate relationship between eye movement patterns and MDD but also underscore the potential of eye-tracking technology in psychiatric research."
                    },
                    {
                        "title": "A cross-scenario and cross-subject domain adaptation method for driving fatigue detection",
                        "abstract": "Objective. The scarcity of electroencephalogram (EEG) data, coupled with individual and scenario variations, leads to considerable challenges in real-world EEG-based driver fatigue detection. We propose a domain adaptation method that utilizes EEG data collected from a laboratory to supplement real-world EEG data and constructs a cross-scenario and cross-subject driver fatigue detection model for real-world scenarios. Approach. First, we collect EEG data from subjects participating in a driving experiment conducted in both laboratory and real-world scenarios. To address the issue of data scarcity, we build a real-world fatigued driving detection model by integrating the real-world data with the laboratory data. Then, we propose a method named cross-scenario and cross-subject domain adaptation (CS2DA), which aims to eliminate the domain shift problem caused by individual variances and scenario differences. Adversarial learning is adopted to extract the common features observed across different subjects within the same scenario. The multikernel maximum mean discrepancy (MK-MMD) method is applied to further minimize scenario differences. Additionally, we propose a conditional MK-MMD constraint to better utilize label information. Finally, we use seven rules to fuse the predicted labels. Main results. We evaluate the CS2DA method through extensive experiments conducted on the two EEG datasets created in this work: the SEED-VLA and the SEED-VRW datasets. Different domain adaptation methods are used to construct a real-world fatigued driving detection model using data from laboratory and real-world scenarios, as well as a combination of both. Our findings show that the proposed CS2DA method outperforms the existing traditional and adversarial learning-based domain adaptation approaches. We also find that combining data from both laboratory and real-world scenarios improves the performance of the model. Significance. This study contributes two EEG-based fatigue driving datasets and demonstrates that the proposed CS2DA method can effectively enhance the performance of a real-world fatigued driving detection model."
                    },
                    {
                        "title": "Professor X: Manipulating EEG BCI with Invisible and Robust Backdoor Attack",
                        "abstract": "While electroencephalogram (EEG) based brain-computer interface (BCI) has been widely used for medical diagnosis, health care, and device control, the safety of EEG BCI has long been neglected. In this paper, we propose Professor X, an invisible and robust\"mind-controller\"that can arbitrarily manipulate the outputs of EEG BCI through backdoor attack, to alert the EEG community of the potential hazard. However, existing EEG attacks mainly focus on single-target class attacks, and they either require engaging the training stage of the target BCI, or fail to maintain high stealthiness. Addressing these limitations, Professor X exploits a three-stage clean label poisoning attack: 1) selecting one trigger for each class; 2) learning optimal injecting EEG electrodes and frequencies strategy with reinforcement learning for each trigger; 3) generating poisoned samples by injecting the corresponding trigger's frequencies into poisoned data for each class by linearly interpolating the spectral amplitude of both data according to previously learned strategies. Experiments on datasets of three common EEG tasks demonstrate the effectiveness and robustness of Professor X, which also easily bypasses existing backdoor defenses."
                    },
                    {
                        "title": "Multimodal Multi-View Spectral-Spatial-Temporal Masked Autoencoder for Self-Supervised Emotion Recognition",
                        "abstract": "Emotion recognition is a primary and complex task in emotional intelligence. Due to the complexity of human emotions, utilizing multimodal fusion methods can enhance the performance by leveraging the complementary properties of different modalities. In this paper, we propose a Multimodal Multi-view Spectral-Spatial-Temporal Masked Autoencoder (Multimodal MV-SSTMA) with self-supervised learning to investigate multimodal emotion recognition based on electroencephalogram (EEG) and eye movement signals. Our experimental process comprises three stages: 1) In the pre-training stage, we employ MV-SSTMA to train feature extractors for EEG and eye movement signals; 2) In the fine-tuning stage, the labeled data are input to the feature extractors to fuse and fine-tune the features; 3) In the testing stage, our model is applied to recognize emotions with test data to calculate the accuracies of different methods. Our experimental results demonstrate that the multimodal fusion model outperforms the unimodal model on both SEED-IV and SEED-V datasets. In addition, the proposed model can still effectively recognize emotions with various ratios of missing data. These results underscore the efficiency of multimodal self-supervised learning and data fusion in emotion recognition."
                    },
                    {
                        "title": "Deep Depression Detection with Resting-State and Cognitive-Task EEG",
                        "abstract": "Depression is a common mental disorder that negatively affects physical health and personal, social and occupational functioning. Currently, accurate and objective diagnosis of depression remains challenging, and electroencephalography (EEG) provides promising clinical practice or home use due to considerable performance and low cost. This work investigates the capabilities of deep neural networks with EEG-based neural patterns from both resting states and cognitive tasks for depression detection. We collect EEG signals from 33 depressed patients and 40 healthy controls using wearable dry electrodes and build Attentive Simple Graph Convolutional network and Transformer neural network for objective depression detection. Four experiment stages, including two resting states and two cognitive tasks, are designed to characterize the alteration of relevant neural patterns in the depressed patients, in terms of decreased energy and impaired performance in sustained attention and response inhibition. The Transformer model achieves an AUC of 0.94 on the Continuous Performance Test-Identical Pairs version (sensitivity: 0.87, specificity: 0.91) and the Stroop Color Word Test (sensitivity: 0.93, specificity: 0.88), and an AUC of 0.89 on the two resting states (sensitivity: 0.85 and 0.87, specificity: 0.88 and 0.90, respectively), indicating the potential of EEG-based neural patterns in identifying depression. These findings provide new insights into the research of depression mechanisms and EEG-based depression biomarkers."
                    },
                    {
                        "title": "Tagging Continuous Labels for EEG-based Emotion Classification",
                        "abstract": "EEG-based emotion classification has long been a critical task in the field of affective brain-computer interface (aBCI). The majority of leading researches construct supervised learning models based on labeled datasets. Several datasets have been released, including different kinds of emotions while utilizing various forms of stimulus materials. However, they adopt discrete labeling methods, in which the EEG data collected during the same stimulus material are given a same label. These methods neglect the fact that emotion changes continuously, and mislabeled data possibly exist. The imprecision of discrete labels may hinder the progress of emotion classification in concerned works. Therefore, we develop an efficient system in this paper to support continuous labeling by giving each sample a unique label, and construct a continuously labeled EEG emotion dataset. Using our dataset with continuous labels, we demonstrate the superiority of continuous labeling in emotion classification through experiments on several classification models. We further utilize the continuous labels to identify the EEG features under induced and non-induced emotions in both our dataset and a public dataset. Our experimental results reveal the learnability and generality of the relation between the EEG features and their continuous labels."
                    },
                    {
                        "title": "Multimodal Adaptive Emotion Transformer with Flexible Modality Inputs on A Novel Dataset with Continuous Labels",
                        "abstract": "Emotion recognition from physiological signals is a topic of widespread interest, and researchers continue to develop novel techniques for perceiving emotions. However, the emergence of deep learning has highlighted the need for high-quality emotional datasets to accurately decode human emotions. In this study, we present a novel multimodal emotion dataset that incorporates electroencephalography (EEG) and eye movement signals to systematically explore human emotions. Seven basic emotions (happy, sad, fear, disgust, surprise, anger, and neutral) are elicited by a large number of 80 videos and fully investigated with continuous labels that indicate the intensity of the corresponding emotions. Additionally, we propose a novel Multimodal Adaptive Emotion Transformer (MAET), that can flexibly process both unimodal and multimodal inputs. Adversarial training is utilized in MAET to mitigate subject discrepancy, which enhances domain generalization. Our extensive experiments, encompassing both subject-dependent and cross-subject conditions, demonstrate MAET's superior performance in handling various inputs. The filtering of data for high emotional evocation using continuous labels proved to be effective in the experiments. Furthermore, the complementary properties between EEG and eye movements are observed. Our code is available at https://github.com/935963004/MAET."
                    },
                    {
                        "title": "Identifying sex differences in EEG-based emotion recognition using graph convolutional network with attention mechanism",
                        "abstract": "Objective. Sex differences in emotions have been widely perceived via self-reports, peripheral physiological signals and brain imaging techniques. However, how sex differences are reflected in the electroencephalography (EEG) neural patterns of emotions remains unresolved. In this paper, we detect sex differences in emotional EEG patterns, investigate the consistency of such differences in various emotion datasets across cultures, and study how sex as a factor affects the performance of EEG-based emotion recognition models. Approach. We thoroughly assess sex differences in emotional EEG patterns on five public datasets, including SEED, SEED-IV, SEED-V, DEAP and DREAMER, systematically examine the sex-specific EEG patterns for happy, sad, fearful, disgusted and neutral emotions, and implement deep learning models for sex-specific emotion recognition. Main results. (1) Sex differences exist in various emotion types and both Western and Eastern cultures; (2) The emotion patterns of females are more stable than those of males, and the patterns of happiness from females are in sharp contrast with the patterns of sadness, fear and disgust, while the energy levels are more balanced for males; (3) The key features for emotion recognition are mainly located at the frontal and temporal sites for females and distributed more evenly over the whole brain for males, and (4) the same-sex emotion recognition models outperform the corresponding cross-sex models. Significance. These findings extend efforts to characterize sex differences in emotional brain activation, provide new physiological evidence for sex-specific emotion processing, and reinforce the message that sex differences should be carefully considered in affective research and precision medicine."
                    },
                    {
                        "title": "Investigating Emotion EEG Patterns for Depression Detection with Attentive Simple Graph Convolutional Network",
                        "abstract": "Depression severely limits daily functioning, diminishes quality of life and possibly leads to self-harm and suicide. Noninvasive electroencephalography (EEG) has been shown effective as biomarkers for objective depression diagnose and treatment response prediction, and dry EEG electrodes further extend its availability for clinical use. Even though many efforts have been made to identify depression biomarkers, searching reliable EEG biomarkers for depression detection remains challenging. This work presents a systematic investigation of capabilities of emotion EEG patterns for depression detection using a dry EEG electrode system. We design an emotion elicitation paradigm with happy, neutral and sad emotions and collect EEG signals during film watching from 33 depressed patients and 40 healthy controls. The mean activation levels at frontal and temporal sites in the alpha, beta and gamma bands of the depressed group are different to those of the healthy group, indicating the impacts of depressive symptoms on the emotion experiences. To leverage the topology information among EEG channels for emotion recognition and depression detection, an Attentive Simple Graph Convolutional network is built. The deep depression-health classifier achieves a sensitivity of 81.93% and a specificity of 91.69% on the happy emotions, suggesting the promising use of the emotion neural patterns for distinguishing the depressed patients from the healthy controls."
                    },
                    {
                        "title": "Objective Depression Detection Using EEG and Eye Movement Signals Induced by Oil Paintings",
                        "abstract": "Depression is a mental disorder characterized by persistent sadness and loss of interest, which has become one of the leading causes of disability worldwide. There are currently no objective diagnostic standards for depression in clinical practice. Previous studies have shown that depression causes both brain abnormalities and behavioral disorders. In this study, both electroencephalography (EEG) and eye movement signals were used to objectively detect depression. By presenting 40 carefully selected oil paintings\u201420 positive and 20 negative\u2014as stimuli, we were able to successfully evoke emotions in 48 depressed patients (DPs) and 40 healthy controls (HCs) from three centers. We then used Transformer, a deep learning model, to conduct emotion recognition and depression detection. The experimental results demonstrate that: a) Transformer achieves the best accuracies of 89.21% and 92.19% in emotion recognition and depression detection, respectively; b) The HC group has higher accuracies than the DP group in emotion recognition for both subject-dependent and subject-independent experiments; c) The neural pattern differences do exist between DPs and HCs, and we find the consistent asymmetry of the neural patterns in DPs; d) For depression detection, using single oil painting achieves the best accuracies, and using negative oil paintings has higher accuracies than using positive oil paintings. These findings suggest that EEG and eye movement signals induced by oil paintings can be used to objectively identify depression."
                    },
                    {
                        "title": "Transformer-Based Domain Adaptation for Multi-Modal Emotion Recognition in Response to Game Animation Videos",
                        "abstract": "Emotion recognition researches necessitate the strategic selection of stimuli to evoke targeted emotions for robust physiological analyses. This study pioneers the use of Genshin Impact game animation videos to induce positive and neutral emotional states. Notably, it introduces a Transformer-based feature extractor, enhancing Domain-Adversarial Neural Networks (DANN) to advance domain adaptation capabilities. Leveraging the inherent advantages of the Transformer architecture, including parallel processing and handling intricate time-series Electroencephalogram (EEG) and eye movement data, this innovation is reinforced by a discriminator with gradient reversal layers, harmonizing source and target domain distributions. Empirical results demonstrate the effectiveness of the Transformer-based DANN model in cross-subject multimodal emotion recognition, achieving a remarkable prediction accuracy of 83.38% across 59 subjects."
                    },
                    {
                        "title": "Network for End-to-End Robust EEG-Based Emotion Recognition",
                        "abstract": ". Emotion recognition based on electroencephalography (EEG) is attracting more and more interest in a\ufb00ective computing. Previous studies have predominantly relied on manually extracted features from EEG signals. It remains largely unexplored in the utilization of raw EEG signals, which contain more temporal information but present a signi\ufb01cant challenge due to their abundance of redundant data and susceptibility to contamination from other physiological signals, such as electrooculography (EOG) and electromyography (EMG). To cope with the high dimensionality and noise interference in end-to-end EEG-based emotion recognition tasks, we introduce a Two-Stream Spectral-Temporal Denoising Network (TS-STDN) which takes into account the spectral and temporal aspects of EEG signals. Moreover, two U-net modules are adopted to reconstruct clean EEG signals in both spectral and temporal domains while extracting discriminative features from noisy data for classifying emotions. Extensive experiments are conducted on two public datasets, SEED and SEED-IV, with the original EEG signals and the noisy EEG signals contaminated by EMG signals. Compared to the baselines, our TS-STDN model exhibits a notable improvement in accuracy, demonstrating an increase of 6% and 8% on the clean data and 11% and 10% on the noisy data, which shows the robustness of the model."
                    }
                ]
            }
        }
    },
    "2310.05842": {
        "paper_data": {
            "title": "Robust Angular Synchronization via Directed Graph Neural Networks",
            "url": "http://arxiv.org/abs/2310.05842v2",
            "arxiv_id": "2310.05842",
            "authors": [
                "Yixuan He",
                "Gesine Reinert",
                "David Wipf",
                "Mihai Cucuringu"
            ],
            "abstract": "The angular synchronization problem aims to accurately estimate (up to a constant additive phase) a set of unknown angles $\\theta_1, \\dots, \\theta_n\\in[0, 2\\pi)$ from $m$ noisy measurements of their offsets $\\theta_i-\\theta_j \\;\\mbox{mod} \\; 2\\pi.$ Applications include, for example, sensor network localization, phase retrieval, and distributed clock synchronization. An extension of the problem to the heterogeneous setting (dubbed $k$-synchronization) is to estimate $k$ groups of angles simultaneously, given noisy observations (with unknown group assignment) from each group. Existing methods for angular synchronization usually perform poorly in high-noise regimes, which are common in applications. In this paper, we leverage neural networks for the angular synchronization problem, and its heterogeneous extension, by proposing GNNSync, a theoretically-grounded end-to-end trainable framework using directed graph neural networks. In addition, new loss functions are devised to encode synchronization objectives. Experimental results on extensive data sets demonstrate that GNNSync attains competitive, and often superior, performance against a comprehensive set of baselines for the angular synchronization problem and its extension, validating the robustness of GNNSync even at high noise levels.",
            "introduction": "ABSTRACT The angular synchronization problem aims to accurately estimate (up to a constant additive phase) a set of unknown angles \u03b81, . . . , \u03b8 nPr0,2\u03c0qfrom mnoisy mea- surements of their offsets \u03b8i\u00b4\u03b8jmod2\u03c0.Applications include, for example, sensor network localization, phase retrieval, and distributed clock synchronization. An extension of the problem to the heterogeneous setting (dubbed k-synchronization) is to estimate kgroups of angles simultaneously, given noisy observations (with unknown group assignment) from each group. Existingmethods fall behind the trivial baseline. To show the effect of a linear combination of LcycleandLupset, we empirically test Lcycle`\u03c4Lupset, with\u03c4varying from 0 to 0.9; see Fig. 33 (the others are omitted but could lead to the sameresults for \u03b7\u201c0.9as in general allexperiments also show that as the problem becomes harder (e.g. as the noise level increases and the network becomes sparser), a smaller coefficient of Lupset(even zero) is preferred, which indicates that Lcycleplays a more essential role in the more challenging scenarios. To assess the effect of fine-tuning (via projected gradient steps) over the baselines, we apply the same number of projected gradient descent steps as GNNSync to the comparative baselines and report the performance in Figures 34 and 35. We observe that even when applying these fine-tuning steps, the baselines are usually beaten by our end-to-end trainable GNNSync pipeline. 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.02.53.03.5MSE k1EROp5N360stylegamma (a) ERO 1pp\u201c0.05q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.02.5MSE k1EROp5N360stylemulti normal1 (b) ERO 2pp\u201c0.05q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.02.53.0MSE k1EROp5N360stylemulti normal0 (c) ERO 3pp\u201c0.05q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.02.5MSE k1EROp5N360styleblock normal6 (d) ERO 4pp\u201c0.05q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.02.53.03.5MSE k1EROp10N360stylegamma (e) ERO 1pp\u201c0.1q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.0MSE k1EROp10N360stylemulti normal1 (f) ERO 2pp\u201c0.1q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.02.5MSE k1EROp10N360stylemulti normal0 (g) ERO 3pp\u201c0.1q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.02.5MSE k1EROp10N360styleblock normal6 (h) ERO 4pp\u201c0.1q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.02.53.0MSE k1EROp15N360stylegamma (i) ERO 1pp\u201c0.15q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.0MSE k1EROp15N360stylemulti normal1 (j) ERO 2pp\u201c0.15q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.02.5MSE k1EROp15N360stylemulti normal0 (k) ERO 3pp\u201c0.15q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 \u03b70.00.51.01.52.0MSE k1EROp15N360styleblock normal6 (l) ERO 4pp\u201c0.15q Figure 28: MSE performance comparison on GNNSync variants on angular synchronization ( k\u201c1) for ERO models. pis the network density and \u03b7is the noise level. Error bars indicate one standard deviation. 41Published as a conference paper at ICLR 2024 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b71.001.251.501.752.002.252.502.753.00MSE k2EROp5N360stylegamma (a) ERO 1pp\u201c0.05q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b70.51.01.52.02.5MSE k2EROp5N360stylemulti normal1 (b) ERO 2pp\u201c0.05q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b70.751.001.251.501.752.002.252.50MSE k2EROp5N360stylemulti normal0 (c) ERO 3pp\u201c0.05q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b70.51.01.52.02.5MSE k2EROp5N360styleblock normal6 (d) ERO 4pp\u201c0.05q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b70.51.01.52.02.5MSE k2EROp10N360stylegamma (e) ERO 1pp\u201c0.1q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b70.51.01.52.0MSE k2EROp10N360stylemulti normal1 (f) ERO 2pp\u201c0.1q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b70.60.81.01.21.41.61.82.02.2MSE k2EROp10N360stylemulti normal0 (g) ERO 3pp\u201c0.1q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b70.500.751.001.251.501.752.00MSE k2EROp10N360styleblock normal6 (h) ERO 4pp\u201c0.1q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b70.51.01.52.02.5MSE k2EROp15N360stylegamma (i) ERO 1pp\u201c0.15q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b70.000.250.500.751.001.251.501.752.00MSE k2EROp15N360stylemulti normal1 (j) ERO 2pp\u201c0.15q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b70.60.81.01.21.41.61.82.02.2MSE k2EROp15N360stylemulti normal0 (k) ERO 3pp\u201c0.15q 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 \u03b70.500.751.001.251.501.752.00MSE k2EROp15N360styleblock normal6 (l) ERO 4pp\u201c0.15q Figure 29:",
            "references": [
                {
                    "title": "Unrolled Algorithms for Group Synchronization",
                    "abstract": "The group synchronization problem involves estimating a collection of group elements from noisy measurements of their pairwise ratios. This task is a key component in many computational problems, including the molecular reconstruction problem in single-particle cryo-electron microscopy (cryo-EM). The standard methods to estimate the group elements are based on iteratively applying linear and non-linear operators, and are not necessarily optimal. Motivated by the structural similarity to deep neural networks, we adopt the concept of algorithm unrolling, where training data is used to optimize the algorithm. We design unrolled algorithms for several group synchronization instances, including synchronization over the group of 3-D rotations: the synchronization problem in cryo-EM. We also apply a similar approach to the multi-reference alignment problem. We show by numerical experiments that the unrolling strategy outperforms existing synchronization algorithms in a wide variety of scenarios."
                },
                {
                    "title": "Robust Group Synchronization via Quadratic Programming",
                    "abstract": "We propose a novel quadratic programming formulation for estimating the corruption levels in group synchronization, and use these estimates to solve this problem. Our objective function exploits the cycle consistency of the group and we thus refer to our method as detection and estimation of structural consistency (DESC). This general framework can be extended to other algebraic and geometric structures. Our formulation has the following advantages: it can tolerate corruption as high as the information-theoretic bound, it does not require a good initialization for the estimates of group elements, it has a simple interpretation, and under some mild conditions the global minimum of our objective function exactly recovers the corruption levels. We demonstrate the competitive accuracy of our approach on both synthetic and real data experiments of rotation averaging."
                },
                {
                    "title": "GNNRank: Learning Global Rankings from Pairwise Comparisons via Directed Graph Neural Networks",
                    "abstract": "Recovering global rankings from pairwise comparisons has wide applications from time synchronization to sports team ranking. Pairwise comparisons corresponding to matches in a competition can be construed as edges in a directed graph (digraph), whose nodes represent e.g. competitors with an unknown rank. In this paper, we introduce neural networks into the ranking recovery problem by proposing the so-called GNNRank, a trainable GNN-based framework with digraph embedding. Moreover, new objectives are devised to encode ranking upsets/violations. The framework involves a ranking score estimation approach, and adds an inductive bias by unfolding the Fiedler vector computation of the graph constructed from a learnable similarity matrix. Experimental results on extensive data sets show that our methods attain competitive and often superior performance against baselines, as well as showing promising transfer ability. Codes and preprocessed data are at: \\url{https://github.com/SherylHYX/GNNRank}."
                },
                {
                    "title": "DIGRAC: Digraph Clustering Based on Flow Imbalance",
                    "abstract": "Node clustering is a powerful tool in the analysis of networks. We introduce a graph neural network framework, named DIGRAC, to obtain node embeddings for directed networks in a self-supervised manner, including a novel probabilistic imbalance loss, which can be used for network clustering. Here, we propose \\textit{directed flow imbalance} measures, which are tightly related to directionality, to reveal clusters in the network even when there is no density difference between clusters. In contrast to standard approaches in the literature, in this paper, directionality is not treated as a nuisance, but rather contains the main signal. DIGRAC optimizes directed flow imbalance for clustering without requiring label supervision, unlike existing graph neural network methods, and can naturally incorporate node features, unlike existing spectral methods. Extensive experimental results on synthetic data, in the form of directed stochastic block models, and real-world data at different scales, demonstrate that our method, based on flow imbalance, attains state-of-the-art results on directed graph clustering when compared against 10 state-of-the-art methods from the literature, for a wide range of noise and sparsity levels, graph structures, and topologies, and even outperforms supervised methods."
                },
                {
                    "title": "An extension of the angular synchronization problem to the heterogeneous setting",
                    "abstract": "<p style='text-indent:20px;'>Given an undirected measurement graph <inline-formula><tex-math id=\"M1\">\\begin{document}$ G = ([n], E) $\\end{document}</tex-math></inline-formula>, the classical angular synchronization problem consists of recovering unknown angles <inline-formula><tex-math id=\"M2\">\\begin{document}$ \\theta_1, \\dots, \\theta_n $\\end{document}</tex-math></inline-formula> from a collection of noisy pairwise measurements of the form <inline-formula><tex-math id=\"M3\">\\begin{document}$ (\\theta_i - \\theta_j) \\mod 2\\pi $\\end{document}</tex-math></inline-formula>, for each <inline-formula><tex-math id=\"M4\">\\begin{document}$ \\{i, j\\} \\in E $\\end{document}</tex-math></inline-formula>. This problem arises in a variety of applications, including computer vision, time synchronization of distributed networks, and ranking from preference relationships. In this paper, we consider a generalization to the setting where there exist <inline-formula><tex-math id=\"M5\">\\begin{document}$ k $\\end{document}</tex-math></inline-formula> unknown groups of angles <inline-formula><tex-math id=\"M6\">\\begin{document}$ \\theta_{l, 1}, \\dots, \\theta_{l, n} $\\end{document}</tex-math></inline-formula>, for <inline-formula><tex-math id=\"M7\">\\begin{document}$ l = 1, \\dots, k $\\end{document}</tex-math></inline-formula>. For each <inline-formula><tex-math id=\"M8\">\\begin{document}$ {\\left\\{{{i, j}}\\right\\}} \\in E $\\end{document}</tex-math></inline-formula>, we are given noisy pairwise measurements of the form <inline-formula><tex-math id=\"M9\">\\begin{document}$ \\theta_{\\ell, i} - \\theta_{\\ell, j} $\\end{document}</tex-math></inline-formula> for an <i>unknown</i> <inline-formula><tex-math id=\"M10\">\\begin{document}$ \\ell \\in \\{1, 2, \\ldots, k\\} $\\end{document}</tex-math></inline-formula>. This can be thought of as a natural extension of the angular synchronization problem to the heterogeneous setting of multiple groups of angles, where the measurement graph has an unknown edge-disjoint decomposition <inline-formula><tex-math id=\"M11\">\\begin{document}$ G = G_1 \\cup G_2 \\ldots \\cup G_k $\\end{document}</tex-math></inline-formula>, where the <inline-formula><tex-math id=\"M12\">\\begin{document}$ G_i $\\end{document}</tex-math></inline-formula>'s denote the subgraphs of edges corresponding to each group. We propose a probabilistic generative model for this problem, along with a spectral algorithm for which we provide a detailed theoretical analysis in terms of robustness against both sampling sparsity and noise. The theoretical findings are complemented by a comprehensive set of numerical experiments, showcasing the efficacy of our algorithm under various parameter regimes. Finally, we consider an application of bi-synchronization to the graph realization problem, and provide along the way an iterative graph disentangling procedure that uncovers the subgraphs <inline-formula><tex-math id=\"M13\">\\begin{document}$ G_i $\\end{document}</tex-math></inline-formula>, <inline-formula><tex-math id=\"M14\">\\begin{document}$ i = 1, \\ldots, k $\\end{document}</tex-math></inline-formula> which is of independent interest, as it is shown to improve the final recovery accuracy across all the experiments considered.</p>"
                },
                {
                    "title": "A Unified Approach to Synchronization Problems over Subgroups of the Orthogonal Group",
                    "abstract": "Given a group $\\mathcal{G}$, the problem of synchronization over the group $\\mathcal{G}$ is a constrained estimation problem where a collection of group elements $G^*_1, \\dots, G^*_n \\in \\mathcal{G}$ are estimated based on noisy observations of pairwise ratios $G^*_i {G^*_j}^{-1}$ for an incomplete set of index pairs $(i,j)$. This problem has gained much attention recently and finds lots of applications due to its appearance in a wide range of scientific and engineering areas. In this paper, we consider the class of synchronization problems over a closed subgroup of the orthogonal group, which covers many instances of group synchronization problems that arise in practice. Our contributions are threefold. First, we propose a unified approach to solve this class of group synchronization problems, which consists of a suitable initialization and an iterative refinement procedure via the generalized power method. Second, we derive a master theorem on the performance guarantee of the proposed approach. Under certain conditions on the subgroup, the measurement model, the noise model and the initialization, the estimation error of the iterates of our approach decreases geometrically. As our third contribution, we study concrete examples of the subgroup (including the orthogonal group, the special orthogonal group, the permutation group and the cyclic group), the measurement model, the noise model and the initialization. The validity of the related conditions in the master theorem are proved for these specific examples. Numerical experiments are also presented. Experiment results show that our approach outperforms existing approaches in terms of computational speed, scalability and estimation error."
                },
                {
                    "title": "Graph Neural Networks: Architectures, Stability, and Transferability",
                    "abstract": "Graph neural networks (GNNs) are information processing architectures for signals supported on graphs. They are presented here as generalizations of convolutional neural networks (CNNs) in which individual layers contain banks of graph convolutional filters instead of banks of classical convolutional filters. Otherwise, GNNs operate as CNNs. Filters are composed of pointwise nonlinearities and stacked in layers. It is shown that GNN architectures exhibit equivariance to permutation and stability to graph deformations. These properties help explain the good performance of GNNs that can be observed empirically. It is also shown that if graphs converge to a limit object, a graphon, GNNs converge to a corresponding limit object, a graphon neural network. This convergence justifies the transferability of GNNs across networks with different numbers of nodes. Concepts are illustrated by the application of GNNs to recommendation systems, decentralized collaborative control, and wireless communication networks."
                },
                {
                    "title": "Message Passing Least Squares Framework and its Application to Rotation Synchronization",
                    "abstract": "We propose an efficient algorithm for solving group synchronization under high levels of corruption and noise, while we focus on rotation synchronization. We first describe our recent theoretically guaranteed message passing algorithm that estimates the corruption levels of the measured group ratios. We then propose a novel reweighted least squares method to estimate the group elements, where the weights are initialized and iteratively updated using the estimated corruption levels. We demonstrate the superior performance of our algorithm over state-of-the-art methods for rotation synchronization using both synthetic and real data."
                },
                {
                    "title": "Understanding Notions of Stationarity in Nonsmooth Optimization: A Guided Tour of Various Constructions of Subdifferential for Nonsmooth Functions",
                    "abstract": "Many contemporary applications in signal processing and machine learning give rise to structured nonconvex nonsmooth optimization problems that can often be tackled by simple iterative methods quite effectively. One of the keys to understanding such a phenomenon-and, in fact, a very difficult conundrum even for experts-lies in the study of \"stationary points\" of the problem in question. Unlike smooth optimization, for which the definition of a stationary point is rather standard, there are myriad definitions of stationarity in nonsmooth optimization. In this article, we provide an introduction to different stationarity concepts for several important classes of nonconvex nonsmooth functions, discuss the geometric interpretations of these concepts, and further clarify their relationships. We then demonstrate the relevance of these constructions in some representative applications and indicate how they could affect the performance of iterative methods for addressing these applications."
                },
                {
                    "title": "Well-conditioned ptychographic imaging via lost subspace completion",
                    "abstract": "Ptychography, a special case of the phase retrieval problem, is a popular method in modern imaging. Its measurements are based on the shifts of a locally supported window function. In general, direct recovery of an object from such measurements is known to be an ill-posed problem. Although for some windows the conditioning can be controlled, for a number of important cases it is not possible, for instance for Gaussian windows. In this paper we develop a subspace completion algorithm, which enables stable reconstruction for a much wider choice of windows, including Gaussian windows. The combination with a regularization technique leads to improved conditioning and better noise robustness."
                },
                {
                    "title": "Multi-Frequency Phase Synchronization",
                    "abstract": "We propose a novel formulation for phase synchronization -- the statistical problem of jointly estimating alignment angles from noisy pairwise comparisons -- as a nonconvex optimization problem that enforces consistency among the pairwise comparisons in multiple frequency channels. Inspired by harmonic retrieval in signal processing, we develop a simple yet efficient two-stage algorithm that leverages the multi-frequency information. We demonstrate in theory and practice that the proposed algorithm significantly outperforms state-of-the-art phase synchronization algorithms, at a mild computational costs incurred by using the extra frequency channels. We also extend our algorithmic framework to general synchronization problems over compact Lie groups."
                },
                {
                    "title": "Solving Jigsaw Puzzles By the Graph Connection Laplacian",
                    "abstract": "We propose a novel mathematical framework to address the problem of automatically solving large jigsaw puzzles. This problem assumes a large image which is cut into equal square pieces that are arbitrarily rotated and shifted and asks to recover the original image given the transformed pieces. The main contribution of this work is a theoretically-guaranteed method for recovering the unknown orientations of the puzzle pieces by using the graph connection Laplacian associated with the puzzle. Iterative application of this method and other methods for recovering the unknown shifts result in a solution for the large jigsaw puzzle problem. This solution is not greedy, unlike many other solutions. Numerical experiments demonstrate the competitive performance of the proposed method."
                },
                {
                    "title": "Translation Synchronization via Truncated Least Squares",
                    "abstract": "In this paper, we introduce a robust algorithm, TranSync, for the 1D translation synchronization problem, in which the aim is to recover the global coordinates of a set of nodes from noisy measurements of relative coordinates along an observation graph. The basic idea of TranSync is to apply truncated least squares, where the solution at each step is used to gradually prune out noisy measurements. We analyze TranSync under both deterministic and randomized noisy models, demonstrating its robustness and stability. Experimental results on synthetic and real datasets show that TranSync is superior to state-of-the-art convex formulations in terms of both efficiency and accuracy."
                },
                {
                    "title": "Nonconvex Phase Synchronization",
                    "abstract": "We estimate $n$ phases (angles) from noisy pairwise relative phase measurements. The task is modeled as a nonconvex least-squares optimization problem. It was recently shown that this problem can be solved in polynomial time via convex relaxation, under some conditions on the noise. In this paper, under similar but more restrictive conditions, we show that a modified version of the power method converges to the global optimum. This is simpler and (empirically) faster than convex approaches. Empirically, they both succeed in the same regime. Further analysis shows that, in the same noise regime as previously studied, second-order necessary optimality conditions for this quadratically constrained quadratic program are also sufficient, despite nonconvexity."
                },
                {
                    "title": "Sync-Rank: Robust Ranking, Constrained Ranking and Rank Aggregation via Eigenvector and SDP Synchronization",
                    "abstract": "We consider the classical problem of establishing a statistical ranking of a set of n items given a set of inconsistent and incomplete pairwise comparisons between such items. Instantiations of this problem occur in numerous applications in data analysis (e.g., ranking teams in sports data), computer vision, and machine learning. We formulate the above problem of ranking with incomplete noisy information as an instance of the group synchronization problem over the group SO(2) of planar rotations, whose usefulness has been demonstrated in numerous applications in recent years in computer vision and graphics, sensor network localization and structural biology. Its least squares solution can be approximated by either a spectral or a semidefinite programming (SDP) relaxation, followed by a rounding procedure. We perform extensive numerical simulations on both synthetic and real-world data sets, which show that our proposed method compares favorably to other ranking methods from the recent literature. Existing theoretical guarantees on the group synchronization problem imply lower bounds on the largest amount of noise permissible in the data while still achieving an approximate recovery of the ground truth ranking. We propose a similar synchronization-based algorithm for the rank-aggregation problem, which integrates in a globally consistent ranking many pairwise rank-offsets or partial rankings, given by different rating systems on the same set of items, an approach which yields significantly more accurate results than other aggregation methods, including Rank-Centrality, a recent state-of-the-art algorithm. Furthermore, we discuss the problem of semi-supervised ranking when there is available information on the ground truth rank of a subset of players, and propose an algorithm based on SDP which is able to recover the ranking of the remaining players, subject to such hard constraints. Finally, synchronization-based ranking, combined with a spectral technique for the densest subgraph problem, makes it possible to extract partial rankings that other methods are not able to find, in other words, to identify the rank of a small subset of players whose pairwise rank comparisons are less noisy than the rest of the data. We discuss a number of related open questions which we defer for future investigation."
                },
                {
                    "title": "SerialRank: Spectral Ranking using Seriation",
                    "abstract": "We describe a seriation algorithm for ranking a set of n items given pairwise comparisons between these items. Intuitively, the algorithm assigns similar rankings to items that compare similarly with all others. It does so by constructing a similarity matrix from pairwise comparisons, using seriation methods to reorder this matrix and construct a ranking. We first show that this spectral seriation algorithm recovers the true ranking when all pairwise comparisons are observed and consistent with a total order. We then show that ranking reconstruction is still exact even when some pairwise comparisons are corrupted or missing, and that seriation based spectral ranking is more robust to noise than other scoring methods. An additional benefit of the seriation formulation is that it allows us to solve semi-supervised ranking problems. Experiments on both synthetic and real datasets demonstrate that seriation based spectral ranking achieves competitive and in some cases superior performance compared to classical ranking methods."
                },
                {
                    "title": "A useful variant of the Davis--Kahan theorem for statisticians",
                    "abstract": "The Davis\u2013Kahan theorem is used in the analysis of many statistical procedures to bound the distance between subspaces spanned by population eigenvectors and their sample versions. It relies on an eigenvalue separation condition between certain population and sample eigenvalues. We present a variant of this result that depends only on a population eigenvalue separation condition, making it more natural and convenient for direct application in statistical contexts, and provide an improvement in many cases to the usual bound in the statistical literature. We also give an extension to situations where the matrices under study may be asymmetric or even non-square, and where interest is in the distance between subspaces spanned by corresponding singular vectors."
                },
                {
                    "title": "Sensor network localization by eigenvector synchronization over the euclidean group",
                    "abstract": "We present a new approach to localization of sensors from noisy measurements of a subset of their Euclidean distances. Our algorithm starts by finding, embedding, and aligning uniquely realizable subsets of neighboring sensors called patches. In the noise-free case, each patch agrees with its global positioning up to an unknown rigid motion of translation, rotation, and possibly reflection. The reflections and rotations are estimated using the recently developed eigenvector synchronization algorithm, while the translations are estimated by solving an overdetermined linear system. The algorithm is scalable as the number of nodes increases and can be implemented in a distributed fashion. Extensive numerical experiments show that it compares favorably to other existing algorithms in terms of robustness to noise, sparse connectivity, and running time. While our approach is applicable to higher dimensions, in the current article, we focus on the two-dimensional case."
                },
                {
                    "title": "Eigenvector Synchronization, Graph Rigidity and the Molecule Problem",
                    "abstract": "The graph realization problem has received a great deal of attention in recent years, due to its importance in applications such as wireless sensor networks and structural biology. In this paper, we extend the previous work and propose the 3D-As-Synchronized-As-Possible (3D-ASAP) algorithm, for the graph realization problem in \u211d3, given a sparse and noisy set of distance measurements. 3D-ASAP is a divide and conquer, non-incremental and non-iterative algorithm, which integrates local distance information into a global structure determination. Our approach starts with identifying, for every node, a subgraph of its 1-hop neighborhood graph, which can be accurately embedded in its own coordinate system. In the noise-free case, the computed coordinates of the sensors in each patch must agree with their global positioning up to some unknown rigid motion, that is, up to translation, rotation and possibly reflection. In other words, to every patch, there corresponds an element of the Euclidean group, Euc(3), of rigid transformations in \u211d3, and the goal was to estimate the group elements that will properly align all the patches in a globally consistent way. Furthermore, 3D-ASAP successfully incorporates information specific to the molecule problem in structural biology, in particular information on known substructures and their orientation. In addition, we also propose 3D-spectral-partitioning (SP)-ASAP, a faster version of 3D-ASAP, which uses a spectral partitioning algorithm as a pre-processing step for dividing the initial graph into smaller subgraphs. Our extensive numerical simulations show that 3D-ASAP and 3D-SP-ASAP are very robust to high levels of noise in the measured distances and to sparse connectivity in the measurement graph, and compare favorably with similar state-of-the-art localization algorithms."
                },
                {
                    "title": "Three-Dimensional Structure Determination from Common Lines in Cryo-EM by Eigenvectors and Semidefinite Programming",
                    "abstract": "The cryo-electron microscopy reconstruction problem is to find the three-dimensional (3D) structure of a macromolecule given noisy samples of its two-dimensional projection images at unknown random directions. Present algorithms for finding an initial 3D structure model are based on the \"angular reconstitution\" method in which a coordinate system is established from three projections, and the orientation of the particle giving rise to each image is deduced from common lines among the images. However, a reliable detection of common lines is difficult due to the low signal-to-noise ratio of the images. In this paper we describe two algorithms for finding the unknown imaging directions of all projections by minimizing global self-consistency errors. In the first algorithm, the minimizer is obtained by computing the three largest eigenvectors of a specially designed symmetric matrix derived from the common lines, while the second algorithm is based on semidefinite programming (SDP). Compared with existing algorithms, the advantages of our algorithms are five-fold: first, they accurately estimate all orientations at very low common-line detection rates; second, they are extremely fast, as they involve only the computation of a few top eigenvectors or a sparse SDP; third, they are nonsequential and use the information in all common lines at once; fourth, they are amenable to a rigorous mathematical analysis using spectral analysis and random matrix theory; and finally, the algorithms are optimal in the sense that they reach the information theoretic Shannon bound up to a constant for an idealized probabilistic model."
                },
                {
                    "title": "Angular embedding: From jarring intensity differences to perceived luminance",
                    "abstract": "Our goal is to turn an intensity image into its perceived luminance without parsing it into depths, surfaces, or scene illuminations. We start with jarring intensity differences at two scales mixed according to edges, identified by a pixel-centric edge detector. We propose angular embedding as a more robust, efficient, and versatile alternative to LS, LLE, and NCUTS for obtaining a global brightness ordering from local differences. Our model explains a variety of brightness illusions with a single algorithm. Brightness of a pixel can be understood locally as its intensity deviating in the gradient direction and globally as finding its rank relative to others, particularly the lightest and darkest ones."
                },
                {
                    "title": "Exploring Network Structure, Dynamics, and Function using NetworkX",
                    "abstract": "NetworkX is a Python language package for exploration and analysis of networks and network algorithms. The core package provides data structures for representing many types of networks, or graphs, including simple graphs, directed graphs, and graphs with parallel edges and self-loops. The nodes in NetworkX graphs can be any (hashable) Python object and edges can contain arbitrary data; this flexibility makes NetworkX ideal for representing networks found in many different scientific fields. In addition to the basic data structures many graph algorithms are implemented for calculating network properties and structure measures: shortest paths, betweenness centrality, clustering, and degree distribution and many more. NetworkX can read and write various graph formats for easy exchange with existing data, and provides generators for many classic graphs and popular graph models, such as the Erdos-Renyi, Small World, and Barabasi-Albert models. The ease-of-use and flexibility of the Python programming language together with connection to the SciPy tools make NetworkX a powerful tool for scientific computations. We discuss some of our recent work studying synchronization of coupled oscillators to demonstrate how NetworkX enables research in the field of computational networks."
                },
                {
                    "title": "Distributed Clock Synchronization over Wireless Networks: Algorithms and Analysis",
                    "abstract": "We analyze the spatial smoothing algorithm of Solis, Borkar and Kumar (2005) for clock synchronization over multi-hop wireless networks. In particular, for a model of a random wireless network we show that with high probability the error variance is O(1) as the number of nodes in the network increases. This provides support for the feasibility of time-based computing n large wireless networks. We also provide bounds on the settling time of a distributed algorithm"
                },
                {
                    "title": "Object Tracking by Scattered Directional Sensors",
                    "abstract": "We address the problem of how to track objects moving at constant velocity using only directional sensors in a wireless sensor network. We model the field of vision of each directional sensor as a line, with the measured data being the times at which sensors detect objects crossing their lines. The network is initially deployed by scattering sensors, and the locations and directions in which the sensor point are also unknown a priori, in addition to the trajectories of the objects. The estimation problem involves the solution of a highly non-convex optimization problem. However we develop a three phase algorithm to solve the problem. It first chooses a coordinate basis adapted to the motions of the first two objects. Then it localizes the sensors with respect to that basis, and refines it as new objects arrive. Finally it transforms the basis if the GPS positions of six of the deployed nodes are known."
                },
                {
                    "title": "Wireless sensor networks for habitat monitoring",
                    "abstract": "We provide an in-depth study of applying wireless sensor networks to real-world habitat monitoring. A set of system design requirements are developed that cover the hardware design of the nodes, the design of the sensor network, and the capabilities for remote data access and management. A system architecture is proposed to address these requirements for habitat monitoring in general, and an instance of the architecture for monitoring seabird nesting environment and behavior is presented. The currently deployed network consists of 32 nodes on a small island off the coast of Maine streaming useful live data onto the web. The application-driven design exercise serves to identify important areas of further work in data sampling, communications, network retasking, and health monitoring."
                },
                {
                    "title": "Relative Perturbation Theory: II. Eigenspace and Singular Subspace Variations",
                    "abstract": "The classical perturbation theory for Hermitian matrix eigenvalue and singular value problems provides bounds on invariant subspace variations that are proportional to the reciprocals of absolute gaps between subsets of spectra or subsets of singular values. These bounds may be bad news for invariant subspaces corresponding to clustered eigenvalues or clustered singular values of much smaller magnitudes than the norms of matrices under considerations. In this paper, we consider how eigenspaces of a Hermitian matrix A change when it is perturbed to $\\widetilde A=D^*AD$ and how singular spaces of a (nonsquare) matrix B change when it is perturbed to $\\widetilde B=D_1^*BD_2$, where D, D1, and D2 are nonsingular. It is proved that under these kinds of perturbations, the changes of invariant subspaces are proportional to the reciprocals of relative gaps between subsets of spectra or subsets of singular values. The classical Davis--Kahan $\\sin\\theta$ theorems and Wedin $\\sin\\theta$ theorems are extended."
                },
                {
                    "title": "The Rotation of Eigenvectors by a Perturbation. III",
                    "abstract": "When a Hermitian linear operator is slightly perturbed, by how much can its invariant subspaces change? Given some approximations to a cluster of neighboring eigenvalues and to the corresponding eigenvectors of a real symmetric matrix, and given an estimate for the gap that separates the cluster from all other eigenvalues, how much can the subspace spanned by the eigenvectors differ from the subspace spanned by our approximations? These questions are closely related; both are investigated here. The difference between the two subspaces is characterized in terms of certain angles through which one subspace must be rotated in order most directly to reach the other. These angles unify the treatment of natural geometric, operator-theoretic and error-analytic questions concerning those subspaces. Sharp bounds upon trigonometric functions of these angles are obtained from the gap and from bounds upon either the perturbation (1st question) or a computable residual (2nd question). An example is included."
                },
                {
                    "title": "PyTorch Geometric Signed Directed: A Survey and Software on Graph Neural Networks for Signed and Directed Graphs",
                    "abstract": "Signed networks are ubiquitous in many real-world applications (e.g., social networks encoding trust/distrust relationships, correlation networks arising from time series data). While many signed networks are directed, there is a lack of survey papers and software packages on graph neural networks (GNNs) specially designed for directed networks. In this paper, we present PyTorch Geometric Signed Directed, a survey and software on GNNs for signed and directed networks. We review typical tasks, loss functions and evaluation metrics in the analysis of signed and directed networks, discuss data used in related experiments, and provide an overview of methods proposed. The deep learning framework consists of easy-to-use GNN models, synthetic and real-world data, as well as task-specific evaluation metrics and loss functions for signed and directed networks. The software is presented in a modular fashion, so that signed and directed networks can also be treated separately. As an extension library for PyTorch Geometric, our proposed software is maintained with open-source re-leases, detailed documentation, continuous integration, unit tests and code coverage checks. Our code is publicly available at https: //github.com/SherylHYX/pytorch_geometric_signed_directed. \u2022 Software and its engineering \u2192 Designing software ; \u2022 Computing methodologies \u2192 Machine learning ."
                },
                {
                    "title": "Digraph Inception Convolutional Networks",
                    "abstract": "Graph Convolutional Networks (GCNs) have shown promising results in modeling graph-structured data. However, they have dif\ufb01culty with processing digraphs because of two reasons: 1) transforming directed to undirected graph to guarantee the symmetry of graph Laplacian is not reasonable since it not only misleads message passing scheme to aggregate incorrect weights but also deprives the unique characteristics of digraph structure; 2) due to the \ufb01xed receptive \ufb01eld in each layer, GCNs fail to obtain multi-scale features that can boost their performance. In this paper, we theoretically extend spectral-based graph convolution to digraphs and derive a simpli\ufb01ed form using personalized PageRank. Speci\ufb01cally, we present the Digraph Inception Convolutional Networks (DiGCN) which utilizes digraph convolution and k th -order proximity to achieve larger receptive \ufb01elds and learn multi-scale features in digraphs. We empirically show that DiGCN can encode more structural information from digraphs than GCNs and help achieve better performance when generalized to other models. Moreover, experiments on various benchmarks demonstrate its superiority against the state-of-the-art methods."
                },
                {
                    "title": "Angular Embedding: A Robust Quadratic Criterion",
                    "abstract": "Given the size and confidence of pairwise local orderings, angular embedding (AE) finds a global ordering with a near-global optimal eigensolution. As a quadratic criterion in the complex domain, AE is remarkably robust to outliers, unlike its real domain counterpart LS, the least squares embedding. Our comparative study of LS and AE reveals that AE's robustness is due not to the particular choice of the criterion, but to the choice of representation in the complex domain. When the embedding is encoded in the angular space, we not only have a nonconvex error function that delivers robustness, but also have a Hermitian graph Laplacian that completely determines the optimum and delivers efficiency. The high quality of embedding by AE in the presence of outliers can hardly be matched by LS, its corresponding L1 norm formulation, or their bounded versions. These results suggest that the key to overcoming outliers lies not with additionally imposing constraints on the embedding solution, but with adaptively penalizing inconsistency between measurements themselves. AE thus significantly advances statistical ranking methods by removing the impact of outliers directly without explicit inconsistency characterization, and advances spectral clustering methods by covering the entire size-confidence measurement space and providing an ordered cluster organization."
                },
                {
                    "title": "MEASUREMENTS: IMPROVED ROBUSTNESS VIA EIGENVECTOR-BASED ANGULAR SYNCHRONIZATION",
                    "abstract": ". We improve a phase retrieval approach that uses correlation-based measurements with compactly supported measurement masks [27]. The improved algorithm admits deterministic measurement constructions together with a robust, fast recovery algorithm that consists of solving a system of linear equations in a lifted space, followed by \ufb01nding an eigenvector (e.g., via an inverse power iteration). Theoretical reconstruction error guarantees from [27] are improved as a result for the new and more robust reconstruction approach proposed herein. Numerical experiments demonstrate robustness and computational e\ufb03ciency that outperforms competing approaches on large problems. Finally, we show that this approach also trivially extends to phase retrieval problems based on windowed Fourier measurements."
                }
            ],
            "categories": [
                "cs.LG",
                "math.OC",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively solve the angular synchronization problem in heterogeneous settings (k-synchronization) to accurately estimate multiple groups of unknown angles from noisy measurements?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the angular synchronization problem has significant implications for various fields, including sensor network localization, phase retrieval, and distributed clock synchronization. Addressing this problem can lead to advancements in the accuracy and efficiency of these applications, fostering further research in distributed systems and improving real-world implementations. By enhancing our understanding of synchronization in noisy environments, this research could pave the way for innovative solutions in communication networks, robotics, and IoT systems, ultimately contributing to the development of more reliable and robust technologies.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe angular synchronization problem is challenging due to the inherent noise in measurements and the complexity of accurately estimating angles without prior knowledge of group assignments. Naive approaches may fail because they do not account for the noise and the potential overlap between groups, leading to incorrect estimations. Additionally, the problem's non-convex nature complicates optimization, making it difficult to find global solutions. Technical obstacles include the need for efficient algorithms that can handle varying noise levels and network sparsity, as well as the requirement for robust methods that can generalize across different scenarios.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on simpler versions of the synchronization problem or has not adequately addressed the complexities introduced by heterogeneous settings. Limitations in existing methods include their inability to effectively manage noise and group assignment uncertainties, leading to suboptimal performance compared to trivial baselines. Barriers such as a lack of end-to-end trainable frameworks and insufficient empirical testing in challenging scenarios have hindered progress. Our approach differs by introducing a novel end-to-end trainable pipeline (GNNSync) that integrates advanced optimization techniques and empirical validation, aiming to outperform existing methods.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the GNNSync pipeline, which utilizes a combination of Lcycle and Lupset loss functions to optimize angle estimation. We will employ synthetic datasets with varying noise levels and network densities to evaluate performance. The primary metric for assessment will be the Mean Squared Error (MSE) of angle estimations. We expect our approach to yield significantly improved accuracy in angle estimation compared to"
            }
        },
        "author_data": {
            "2ac50356-af1b-4186-9741-c51ad757df5d": {
                "pk": "2ac50356-af1b-4186-9741-c51ad757df5d",
                "name": "Yixuan He",
                "collaborators": [
                    "G. Reinert",
                    "Mihai Cucuringu",
                    "Junchi Yan",
                    "Xitong Zhang",
                    "Quan Gan",
                    "Junjie Huang",
                    "Benedek Rozemberczki",
                    "Michael Perlmutter",
                    "Delu Zeng",
                    "Haoyu Geng",
                    "Chao Chen",
                    "Gang Zeng",
                    "Zhaobing Han",
                    "Hua Chai",
                    "Xuhong Wang",
                    "Sirui Chen",
                    "Minjie Wang",
                    "Yupu Yang",
                    "D. Wipf",
                    "Nathan Brugnone",
                    "M. Hirn",
                    "Songchao Wang",
                    "Rohitash Chandra",
                    "P. Scherer",
                    "G. Panagopoulos",
                    "M. Astefanoaei",
                    "Oliver Kiss",
                    "Ferenc B\u00e9res",
                    "Nicolas Collignon",
                    "Rik Sarkar",
                    "Tianyi Hu",
                    "Li Liu",
                    "Zhi-hong Chen",
                    "Jiabin Huang",
                    "Jie Chen",
                    "J. Paisley",
                    "Kazuya Uesato",
                    "H. Yamana"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Data Clustering",
                    "Temporal Graphs"
                ],
                "publications": [
                    {
                        "title": "Non-canonical Word Order and Its Discourse Functions in English Fairy Tales",
                        "abstract": ": What is noticeable is that in English, there exist a variety of expressions to convey a given proposition. Why would the English language provide the speaker with so many syntactic options for saying what amounts to the same thing? Drawing lessons from previous studies, this research found that the main reason of the phenomenon is that multiple choices allow the speaker to mark the information status of different constituents that make up the sentence. By altering the world order of the proposition, the speaker could provide the hearer with an inference of which part of the information is given and which part is new, so that the hearer could understand and follow the speaker more easily. Among so many different syntactic ways of expressing a given proposition, non-canonical constructions have caught researchers\u2019 eyes for years due to its function of increasing discourse coherence while communicating and its significance in meeting the immediate communication needs. Therefore, this paper specializes in three typical non-canonical utterances, locative inversion, there insertion and left dislocation, with instances selected from a series of fairy tales to further explore how they achieve coherent function and help the hearer or the reader follow the story easily."
                    },
                    {
                        "title": "Pyramid Graph Neural Network: A Graph Sampling and Filtering Approach for Multi-scale Disentangled Representations",
                        "abstract": "Spectral methods for graph neural networks (GNNs) have achieved great success. Despite their success, many works have shown that existing approaches are mainly focused on low-frequency information which may not be pertinent to the task at hand. Recent efforts have been made to design new graph filters for wider frequency profiles, but it remains an open problem how to learn multi-scale disentangled node embeddings in the graph Fourier domain. In this paper, we propose a graph (signal) sampling and filtering framework, entitled Pyramid Graph Neural Network (PyGNN), which follows the Downsampling-Filtering-Upsampling-Decoding scheme. To be specific, we develop an \u03c9-bandlimited downsampling approach to split input graph into subgraphs for the reduction of high-frequency components, then perform spectral graph filters on subgraphs to achieve node embeddings with different frequency bands, and propose a Laplacian smoothing-based upsampling approach to extrapolate the node embedding on subgraphs to the full set of vertices on the original graph. In the end, we add frequency-aware gated units to decode node embeddings of different frequencies for downstream tasks. Results on both homophilic and heterophilic graph datasets show its superiority over state-of-the-art methods."
                    },
                    {
                        "title": "CEP3: Community Event Prediction with Neural Point Process on Graph",
                        "abstract": "Many real world applications can be formulated as event forecasting on Continuous Time Dynamic Graphs (CTDGs) where the occurrence of a timed event between two entities is represented as an edge along with its occurrence timestamp in the graphs.However, most previous works approach the problem in compromised settings, either formulating it as a link prediction task on the graph given the event time or a time prediction problem given which event will happen next. In this paper, we propose a novel model combining Graph Neural Networks and Marked Temporal Point Process (MTPP) that jointly forecasts multiple link events and their timestamps on communities over a CTDG. Moreover, to scale our model to large graphs, we factorize the jointly event prediction problem into three easier conditional probability modeling problems.To evaluate the effectiveness of our model and the rationale behind such a decomposition, we establish a set of benchmarks and evaluation metrics for this event forecasting task. Our experiments demonstrate the superior performance of our model in terms of both model accuracy and training efficiency."
                    },
                    {
                        "title": "GNNRank: Learning Global Rankings from Pairwise Comparisons via Directed Graph Neural Networks",
                        "abstract": "Recovering global rankings from pairwise comparisons has wide applications from time synchronization to sports team ranking. Pairwise comparisons corresponding to matches in a competition can be construed as edges in a directed graph (digraph), whose nodes represent e.g. competitors with an unknown rank. In this paper, we introduce neural networks into the ranking recovery problem by proposing the so-called GNNRank, a trainable GNN-based framework with digraph embedding. Moreover, new objectives are devised to encode ranking upsets/violations. The framework involves a ranking score estimation approach, and adds an inductive bias by unfolding the Fiedler vector computation of the graph constructed from a learnable similarity matrix. Experimental results on extensive data sets show that our methods attain competitive and often superior performance against baselines, as well as showing promising transfer ability. Codes and preprocessed data are at: \\url{https://github.com/SherylHYX/GNNRank}."
                    },
                    {
                        "title": "PyTorch Geometric Signed Directed: A Survey and Software on Graph Neural Networks for Signed and Directed Graphs",
                        "abstract": "Signed networks are ubiquitous in many real-world applications (e.g., social networks encoding trust/distrust relationships, correlation networks arising from time series data). While many signed networks are directed, there is a lack of survey papers and software packages on graph neural networks (GNNs) specially designed for directed networks. In this paper, we present PyTorch Geometric Signed Directed, a survey and software on GNNs for signed and directed networks. We review typical tasks, loss functions and evaluation metrics in the analysis of signed and directed networks, discuss data used in related experiments, and provide an overview of methods proposed. The deep learning framework consists of easy-to-use GNN models, synthetic and real-world data, as well as task-specific evaluation metrics and loss functions for signed and directed networks. The software is presented in a modular fashion, so that signed and directed networks can also be treated separately. As an extension library for PyTorch Geometric, our proposed software is maintained with open-source re-leases, detailed documentation, continuous integration, unit tests and code coverage checks. Our code is publicly available at https: //github.com/SherylHYX/pytorch_geometric_signed_directed. \u2022 Software and its engineering \u2192 Designing software ; \u2022 Computing methodologies \u2192 Machine learning ."
                    },
                    {
                        "title": "PyTorch Geometric Signed Directed: A Software Package on Graph Neural Networks for Signed and Directed Graphs",
                        "abstract": "Networks are ubiquitous in many real-world applications (e.g., social networks encoding trust/distrust relationships, correlation networks arising from time series data). While many networks are signed or directed, or both, there is a lack of unified software packages on graph neural networks (GNNs) specially designed for signed and directed networks. In this paper, we present PyTorch Geometric Signed Directed (PyGSD), a software package which fills this gap. Along the way, we evaluate the implemented methods with experiments with a view to providing insights into which method to choose for a given task. The deep learning framework consists of easy-to-use GNN models, synthetic and real-world data, as well as task-specific evaluation metrics and loss functions for signed and directed networks. As an extension library for PyG, our proposed software is maintained with open-source releases, detailed documentation, continuous integration, unit tests and code coverage checks. The GitHub repository of the library is https://github.com/SherylHYX/pytorch_geometric_signed_directed."
                    },
                    {
                        "title": "MSGNN: A Spectral Graph Neural Network Based on a Novel Magnetic Signed Laplacian",
                        "abstract": "Signed and directed networks are ubiquitous in real-world applications. However, there has been relatively little work proposing spectral graph neural networks (GNNs) for such networks. Here we introduce a signed directed Laplacian matrix, which we call the magnetic signed Laplacian, as a natural generalization of both the signed Laplacian on signed graphs and the magnetic Laplacian on directed graphs. We then use this matrix to construct a novel efficient spectral GNN architecture and conduct extensive experiments on both node clustering and link prediction tasks. In these experiments, we consider tasks related to signed information, tasks related to directional information, and tasks related to both signed and directional information. We demonstrate that our proposed spectral GNN is effective for incorporating both signed and directional information, and attains leading performance on a wide range of data sets. Additionally, we provide a novel synthetic network model, which we refer to as the Signed Directed Stochastic Block Model, and a number of novel real-world data sets based on lead-lag relationships in financial time series."
                    },
                    {
                        "title": "GNNs for Node Clustering in Signed and Directed Networks",
                        "abstract": "With an increasing number of applications where data can be represented as graphs, graph neural networks are a useful tool to apply deep learning to graph data. In particular, node clustering is an important problem in network analysis. Signed and directed networks are important types of networks that are linked to many real-world problems; their asymmetry provides a challenge for many clustering methods. We propose two graph neural network models for node clustering in signed networks and directed networks, respectively. The methods are end-to-end in combining embedding generation and clustering without an intermediate step. Experimental results on a synthetic signed stochastic block model, a polarized version of it, and real-world data at different scales, demonstrate that our proposed methods can achieve comparable or better results than state-of-the-art node clustering methods, for a wide range of noise and sparsity levels. The introduced models complement existing well-performing methods through the possibility of including exogenous information, in the form of node-level features or labels."
                    },
                    {
                        "title": "MagNet: A Neural Network for Directed Graphs",
                        "abstract": "The prevalence of graph-based data has spurred the rapid development of graph neural networks (GNNs) and related machine learning algorithms. Yet, despite the many datasets naturally modeled as directed graphs, including citation, website, and traffic networks, the vast majority of this research focuses on undirected graphs. In this paper, we propose MagNet, a GNN for directed graphs based on a complex Hermitian matrix known as the magnetic Laplacian. This matrix encodes undirected geometric structure in the magnitude of its entries and directional information in their phase. A \"charge\" parameter attunes spectral information to variation among directed cycles. We apply our network to a variety of directed graph node classification and link prediction tasks showing that MagNet performs well on all tasks and that its performance exceeds all other methods on a majority of such tasks. The underlying principles of MagNet are such that it can be adapted to other GNN architectures."
                    },
                    {
                        "title": "DIGRAC: Digraph Clustering Based on Flow Imbalance",
                        "abstract": "Node clustering is a powerful tool in the analysis of networks. We introduce a graph neural network framework, named DIGRAC, to obtain node embeddings for directed networks in a self-supervised manner, including a novel probabilistic imbalance loss, which can be used for network clustering. Here, we propose \\textit{directed flow imbalance} measures, which are tightly related to directionality, to reveal clusters in the network even when there is no density difference between clusters. In contrast to standard approaches in the literature, in this paper, directionality is not treated as a nuisance, but rather contains the main signal. DIGRAC optimizes directed flow imbalance for clustering without requiring label supervision, unlike existing graph neural network methods, and can naturally incorporate node features, unlike existing spectral methods. Extensive experimental results on synthetic data, in the form of directed stochastic block models, and real-world data at different scales, demonstrate that our method, based on flow imbalance, attains state-of-the-art results on directed graph clustering when compared against 10 state-of-the-art methods from the literature, for a wide range of noise and sparsity levels, graph structures, and topologies, and even outperforms supervised methods."
                    },
                    {
                        "title": "SSSNET: Semi-Supervised Signed Network Clustering",
                        "abstract": "Node embeddings are a powerful tool in the analysis of networks; yet, their full potential for the important task of node clustering has not been fully exploited. In particular, most state-of-the-art methods generating node embeddings of signed networks focus on link sign prediction, and those that pertain to node clustering are usually not graph neural network (GNN) methods. Here, we introduce a novel probabilistic balanced normalized cut loss for training nodes in a GNN framework for semi-supervised signed network clustering, called SSSNET. The method is end-to-end in combining embedding generation and clustering without an intermediate step; it has node clustering as main focus, with an emphasis on polarization effects arising in networks. The main novelty of our approach is a new take on the role of social balance theory for signed network embeddings. The standard heuristic for justifying the criteria for the embeddings hinges on the assumption that\"an enemy's enemy is a friend\". Here, instead, a neutral stance is assumed on whether or not the enemy of an enemy is a friend. Experimental results on various data sets, including a synthetic signed stochastic block model, a polarized version of it, and real-world data at different scales, demonstrate that SSSNET can achieve comparable or better results than state-of-the-art spectral clustering methods, for a wide range of noise and sparsity levels. SSSNET complements existing methods through the possibility of including exogenous information, in the form of node-level features or labels."
                    },
                    {
                        "title": "Bayesian neural networks for stock price forecasting before and during COVID-19 pandemic",
                        "abstract": "Recently, there has been much attention in the use of machine learning methods, particularly deep learning for stock price prediction. A major limitation of conventional deep learning is uncertainty quantification in predictions which affect investor confidence. Bayesian neural networks feature Bayesian inference for providing inference (training) of model parameters that provides a rigorous methodology for uncertainty quantification in predictions. Markov Chain Monte Carlo (MCMC) sampling methods have been prominent in implementing inference of Bayesian neural networks; however certain limitations existed due to a large number of parameters and the need for better computational resources. Recently, there has been much progress in the area of Bayesian neural networks given the use of Langevin gradients with parallel tempering MCMC that can be implemented in a parallel computing environment. The COVID-19 pandemic had a drastic impact in the world economy and stock markets given different levels of lockdowns due to rise and fall of daily infections. It is important to investigate the performance of related forecasting models during the COVID-19 pandemic given the volatility in stock markets. In this paper, we use novel Bayesian neural networks for multi-step-ahead stock price forecasting before and during COVID-19. We also investigate if the pre-COVID-19 datasets are useful of modelling stock price forecasting during COVID-19. Our results indicate due to high volatility in the stock-price during COVID-19, it is more challenging to provide forecasting. However, we found that Bayesian neural networks could provide reasonable predictions with uncertainty quantification despite high market volatility during the first peak of the COVID-19 pandemic."
                    },
                    {
                        "title": "PyTorch Geometric Temporal: Spatiotemporal Signal Processing with Neural Machine Learning Models",
                        "abstract": "We present PyTorch Geometric Temporal, a deep learning framework combining state-of-the-art machine learning algorithms for neural spatiotemporal signal processing. The main goal of the library is to make temporal geometric deep learning available for researchers and machine learning practitioners in a unified easy-to-use framework. PyTorch Geometric Temporal was created with foundations on existing libraries in the PyTorch eco-system, streamlined neural network layer definitions, temporal snapshot generators for batching, and integrated benchmark datasets. These features are illustrated with a tutorial-like case study. Experiments demonstrate the predictive performance of the models implemented in the library on real-world problems such as epidemiological forecasting, ride-hail demand prediction, and web traffic management. Our sensitivity analysis of runtime shows that the framework can potentially operate on web-scale datasets with rich temporal features and spatial structure."
                    },
                    {
                        "title": "DIGRAC: Digraph Clustering with Flow Imbalance",
                        "abstract": "Node clustering is a powerful tool in the analysis of networks. Here, we introduce a graph neural network framework with a novel scalable Directed Mixed Path Aggregation (DIMPA) scheme to obtain node embeddings for directed networks in a self-supervised manner, including a novel probabilistic imbalance loss. The method is end-to-end in combining embedding generation and clustering without an intermediate step. In contrast to standard approaches in the literature, in this paper, directionality is not treated as a nuisance, but rather contains the main signal. In particular, we leverage the recently introduced cut \ufb02ow imbalance measure, which is tightly related to directionality; cut \ufb02ow imbalance is optimized without resorting to spectral methods or cluster labels. Experimental results on synthetic data, in the form of directed stochastic block models and real-world data at different scales, demonstrate that our method attains state-of-the-art results on directed graph clustering, for a wide range of noise and sparsity levels and graph structures."
                    },
                    {
                        "title": "Scan-Flood Fill(SCAFF): An Efficient Automatic Precise Region Filling Algorithm for Complicated Regions",
                        "abstract": "Recently, instant level labeling for supervised machine learning requires a considerable number of filled masks. In this paper, we propose an efficient automatic region filling algorithm for complicated regions. Distinguishing between adjacent connected regions, the Main Filling Process scans through all pixels and fills all the pixels except boundary ones with either exterior or interior label color. In this way, we succeed in classifying all the pixels inside the region except boundary ones in the given image to form two groups: a background group and a mask group. We then set all exterior label pixels to background color, and interior label pixels to mask color. With this algorithm, we are able to generate output masks precisely and efficiently even for complicated regions as long as boundary pixels are given. Experimental results show that the proposed algorithm can generate precise masks that allow for various machine learning tasks such as supervised training. This algorithm can effectively handle multiple regions, complicated 'holes' and regions whose boundaries touch the image border. By testing the algorithm on both toy and practical images, we show that the performance of Scan-flood Fill(SCAFF) has achieved favorable results."
                    },
                    {
                        "title": "Ro-SOS: Metric Expression Network (MEnet) for Robust Salient Object Segmentation",
                        "abstract": "Although deep CNNs have brought significant improvement to image saliency detection, most CNN based models are sensitive to distortion such as compression and noise. In this paper, we propose an end-to-end generic salient object segmentation model called Metric Expression Network (MEnet) to deal with saliency detection with the tolerance of distortion. Within MEnet, a new topological metric space is constructed, whose implicit metric is determined by the deep network. As a result, we manage to group all the pixels in the observed image semantically within this latent space into two regions: a salient region and a non-salient region. With this architecture, all feature extractions are carried out at the pixel level, enabling fine granularity of output boundaries of the salient objects. What's more, we try to give a general analysis for the noise robustness of the network in the sense of Lipschitz and Jacobian literature. Experiments demonstrate that robust salient maps facilitating object segmentation can be generated by the proposed metric. Tests on several public benchmarks show that MEnet has achieved desirable performance. Furthermore, by direct computation and measuring the robustness, the proposed method outperforms previous CNN-based methods on distorted inputs."
                    },
                    {
                        "title": "Comparison of Community Detection Methods for Facebook Ego Network",
                        "abstract": "People tend to build and maintain their friendship relying on SNS nowadays. Thus, the problem of how to organize the social network accurately and automatically has become important to be considered. As this problem draws a lot of attention from both users and companies, there already existed many SNS providing such functions, such as Twitter list, Facebook lists, etc. However, the functions are laborious, means it requires users to manually organize and update their friends\u2019 circles or lists every time they add new friends. Instead of using the functions, we may adopt community clustering methods for SNS. In this research, we applied three different common clustering methods to SNAP Facebook dataset and compared the accuracy of these three methods to confirm the adoptability for organizing social network accurately and automatically. Our result shows that LDA performs the best because both of its complexity compared to k-means and of free from setting the number of communities compared to HDP. Keyword: Facebook, Data Clustering, Data Mining"
                    }
                ]
            },
            "8f433a25-9cb6-4c71-824e-fb8d726dc45a": {
                "pk": "8f433a25-9cb6-4c71-824e-fb8d726dc45a",
                "name": "Gesine Reinert",
                "collaborators": [
                    "Mihai Cucuringu",
                    "Wenkai Xu",
                    "Yixuan He",
                    "Andrew Elliott",
                    "Stratis Limnios",
                    "Yutong Lu",
                    "Xitong Zhang",
                    "Junjie Huang",
                    "James B. Wilsenach",
                    "Charlotte M. Deane",
                    "Katie Warnaby",
                    "Gholamali Aminian",
                    "Lukasz Szpruch",
                    "Samuel N. Cohen",
                    "Ruikang Ouyang",
                    "Jase Clarkson",
                    "Tadas Temvcinas",
                    "Vidit Nanda",
                    "Praveen Selvaraj",
                    "C. Maple",
                    "Anastasia Mantziou",
                    "Victor Meirinhos",
                    "Quan Gan",
                    "D. Wipf",
                    "Junchi Yan",
                    "Javier Pardo-Diaz",
                    "Mariano Beguerisse-D\u00edaz",
                    "P. Poole",
                    "C. Deane",
                    "Marie Ernst",
                    "Yvik Swan",
                    "Benedek Rozemberczki"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Generative Modeling",
                    "Time Series Analysis",
                    "Statistical Learning"
                ],
                "publications": [
                    {
                        "title": "SteinGen: Generating Fidelitous and Diverse Graph Samples",
                        "abstract": "Generating graphs that preserve characteristic structures while promoting sample diversity can be challenging, especially when the number of graph observations is small. Here, we tackle the problem of graph generation from only one observed graph. The classical approach of graph generation from parametric models relies on the estimation of parameters, which can be inconsistent or expensive to compute due to intractable normalisation constants. Generative modelling based on machine learning techniques to generate high-quality graph samples avoids parameter estimation but usually requires abundant training samples. Our proposed generating procedure, SteinGen, which is phrased in the setting of graphs as realisations of exponential random graph models, combines ideas from Stein's method and MCMC by employing Markovian dynamics which are based on a Stein operator for the target model. SteinGen uses the Glauber dynamics associated with an estimated Stein operator to generate a sample, and re-estimates the Stein operator from the sample after every sampling step. We show that on a class of exponential random graph models this novel\"estimation and re-estimation\"generation strategy yields high distributional similarity (high fidelity) to the original data, combined with high sample diversity."
                    },
                    {
                        "title": "Graph models of brain state in deep anaesthesia reveal sink state dynamics of reduced spatiotemporal complexity and integration",
                        "abstract": "Deep anaesthetisation is an important surgical and explorative tool in the study of consciousness. Much work has been done to connect the deeply anaesthetised condition with decreased functional complexity. However, anaesthesia-induced unconsciousness is also a dynamic condition in which functional activity and complexity may fluctuate, being perturbed by internal or external (e.g. noxious) stimuli. We use fMRI data from a cohort undergoing ultra-slow propofol induction and a dynamic graph modelling framework which characterises changes in functional activity and connectivity as changes in brain state. We examine the group-level dynamics of brain state activity, complexity, integration and modularization in deep anaesthesia and wakefulness and propose novel measures for temporal complexity of brain state dynamics. We find that deep anaesthesia states are less temporally complex and less functionally-related to one another than wakeful states. Anaesthesia dynamics are dominated by a handful of sink states that act as low-complexity attractors to which subjects repeatedly return. Our analysis suggests that dynamic functional organisation in anaesthesia can be characterised by temporally stable, modular communities that change little in time, stratifying regional activity and functionally isolating cortical from both subcortical, and in particular, thalamic brain areas. Using data from a large human structural tractography atlas, we show that functional modularisation in deep anaesthesia more closely resembles normative structural organisation than does functional organisation in wakefulness. We find evidence suggesting that brain state trajectories and in particular, the dominance of low complexity attractor-like sink states in anaesthesia, appear to depend on subject-specific age and anaesthesia susceptibility factors."
                    },
                    {
                        "title": "Generalization Error of Graph Neural Networks in the Mean-field Regime",
                        "abstract": "This work provides a theoretical framework for assessing the generalization error of graph neural networks in the over-parameterized regime, where the number of parameters surpasses the quantity of data points. We explore two widely utilized types of graph neural networks: graph convolutional neural networks and message passing graph neural networks. Prior to this study, existing bounds on the generalization error in the over-parametrized regime were uninformative, limiting our understanding of over-parameterized network performance. Our novel approach involves deriving upper bounds within the mean-field regime for evaluating the generalization error of these graph neural networks. We establish upper bounds with a convergence rate of $O(1/n)$, where $n$ is the number of graph samples. These upper bounds offer a theoretical assurance of the networks' performance on unseen data in the challenging over-parameterized regime and overall contribute to our understanding of their performance."
                    },
                    {
                        "title": "L2G2G: a Scalable Local-to-Global Network Embedding with Graph Autoencoders",
                        "abstract": "For analysing real-world networks, graph representation learning is a popular tool. These methods, such as a graph autoencoder (GAE), typically rely on low-dimensional representations, also called embeddings, which are obtained through minimising a loss function; these embeddings are used with a decoder for downstream tasks such as node classification and edge prediction. While GAEs tend to be fairly accurate, they suffer from scalability issues. For improved speed, a Local2Global approach, which combines graph patch embeddings based on eigenvector synchronisation, was shown to be fast and achieve good accuracy. Here we propose L2G2G, a Local2Global method which improves GAE accuracy without sacrificing scalability. This improvement is achieved by dynamically synchronising the latent node representations, while training the GAEs. It also benefits from the decoder computing an only local patch loss. Hence, aligning the local embeddings in each epoch utilises more information from the graph than a single post-training alignment does, while maintaining scalability. We illustrate on synthetic benchmarks, as well as real-world examples, that L2G2G achieves higher accuracy than the standard Local2Global approach and scales efficiently on the larger data sets. We find that for large and dense networks, it even outperforms the slow, but assumed more accurate, GAEs."
                    },
                    {
                        "title": "Split Conformal Prediction under Data Contamination",
                        "abstract": "Conformal prediction is a non-parametric technique for constructing prediction intervals or sets from arbitrary predictive models under the assumption that the data is exchangeable. It is popular as it comes with theoretical guarantees on the marginal coverage of the prediction sets and the split conformal prediction variant has a very low computational cost compared to model training. We study the robustness of split conformal prediction in a data contamination setting, where we assume a small fraction of the calibration scores are drawn from a different distribution than the bulk. We quantify the impact of the corrupted data on the coverage and efficiency of the constructed sets when evaluated on\"clean\"test points, and verify our results with numerical experiments. Moreover, we propose an adjustment in the classification setting which we call Contamination Robust Conformal Prediction, and verify the efficacy of our approach using both synthetic and real datasets."
                    },
                    {
                        "title": "Goodness-of-fit via Count Statistics in Dense Random Simplicial Complexes",
                        "abstract": "A key object of study in stochastic topology is a random simplicial complex. In this work we study a multi-parameter random simplicial complex model, where the probability of including a $k$-simplex, given the lower dimensional structure, is fixed. This leads to a conditionally independent probabilistic structure. This model includes the Erd\\H{o}s-R\\'enyi random graph, the random clique complex as well as the Linial-Meshulam complex as special cases. The model is studied from both probabilistic and statistical points of view. We prove multivariate central limit theorems with bounds and known limiting covariance structure for the subcomplex counts and the number of critical simplices under a lexicographical acyclic partial matching. We use the CLTs to develop a goodness-of-fit test for this random model and evaluate its empirical performance. In order for the test to be applicable in practice, we also prove that the MLE estimators are asymptotically unbiased, consistent, uncorrelated and normally distributed."
                    },
                    {
                        "title": "SaGess: Sampling Graph Denoising Diffusion Model for Scalable Graph Generation",
                        "abstract": "Over recent years, denoising diffusion generative models have come to be considered as state-of-the-art methods for synthetic data generation, especially in the case of generating images. These approaches have also proved successful in other applications such as tabular and graph data generation. However, due to computational complexity, to this date, the application of these techniques to graph data has been restricted to small graphs, such as those used in molecular modeling. In this paper, we propose SaGess, a discrete denoising diffusion approach, which is able to generate large real-world networks by augmenting a diffusion model (DiGress) with a generalized divide-and-conquer framework. The algorithm is capable of generating larger graphs by sampling a covering of subgraphs of the initial graph in order to train DiGress. SaGess then constructs a synthetic graph using the subgraphs that have been generated by DiGress. We evaluate the quality of the synthetic data sets against several competitor methods by comparing graph statistics between the original and synthetic samples, as well as evaluating the utility of the synthetic data set produced by using it to train a task-driven model, namely link prediction. In our experiments, SaGess, outperforms most of the one-shot state-of-the-art graph generating methods by a significant factor, both on the graph metrics and on the link prediction task."
                    },
                    {
                        "title": "The GNAR-edge model: a network autoregressive model for networks with time-varying edge weights",
                        "abstract": "  In economic and financial applications, there is often the need for analysing multivariate time series, comprising of time series for a range of quantities. In some applications, such complex systems can be associated with some underlying network describing pairwise relationships among the quantities. Accounting for the underlying network structure for the analysis of this type of multivariate time series is required for assessing estimation error and can be particularly informative for forecasting. Our work is motivated by a dataset consisting of time series of industry-to-industry transactions. In this example, pairwise relationships between Standard Industrial Classification (SIC) codes can be represented using a network, with SIC codes as nodes and pairwise transactions between SIC codes as edges, while the observed time series of the amounts of the transactions for each pair of SIC codes can be regarded as time-varying weights on the edges. Inspired by Knight et al. (2020, J. Stat. Softw., 96, 1\u201336), we introduce the GNAR-edge model which allows modelling of multiple time series utilizing the network structure, assuming that each edge weight depends not only on its past values, but also on past values of its neighbouring edges, for a range of neighbourhood stages. The method is validated through simulations. Results from the implementation of the GNAR-edge model on the real industry-to-industry data show good fitting and predictive performance of the model. The predictive performance is improved when sparsifying the network using a lead\u2013lag analysis and thresholding edges according to a lead\u2013lag score."
                    },
                    {
                        "title": "Co-Trading Networks for Modeling Dynamic Interdependency Structures and Estimating High-Dimensional Covariances in US Equity Markets",
                        "abstract": "The time proximity of trades across stocks reveals interesting topological structures of the equity market in the United States. In this article, we investigate how such concurrent cross-stock trading behaviors, which we denote as co-trading, shape the market structures and affect stock price co-movements. By leveraging a co-trading-based pairwise similarity measure, we propose a novel method to construct dynamic networks of stocks. Our empirical studies employ high-frequency limit order book data from 2017-01-03 to 2019-12-09. By applying spectral clustering on co-trading networks, we uncover economically meaningful clusters of stocks. Beyond the static Global Industry Classification Standard (GICS) sectors, our data-driven clusters capture the time evolution of the dependency among stocks. Furthermore, we demonstrate statistically significant positive relations between low-latency co-trading and return covariance. With the aid of co-trading networks, we develop a robust estimator for high-dimensional covariance matrix, which yields superior economic value on portfolio allocation. The mean-variance portfolios based on our covariance estimates achieve both lower volatility and higher Sharpe ratios than standard benchmarks."
                    },
                    {
                        "title": "A Kernelised Stein Statistic for Assessing Implicit Generative Models",
                        "abstract": "Synthetic data generation has become a key ingredient for training machine learning procedures, addressing tasks such as data augmentation, analysing privacy-sensitive data, or visualising representative samples. Assessing the quality of such synthetic data generators hence has to be addressed. As (deep) generative models for synthetic data often do not admit explicit probability distributions, classical statistical procedures for assessing model goodness-of-fit may not be applicable. In this paper, we propose a principled procedure to assess the quality of a synthetic data generator. The procedure is a kernelised Stein discrepancy (KSD)-type test which is based on a non-parametric Stein operator for the synthetic data generator of interest. This operator is estimated from samples which are obtained from the synthetic data generator and hence can be applied even when the model is only implicit. In contrast to classical testing, the sample size from the synthetic data generator can be as large as desired, while the size of the observed data, which the generator aims to emulate is fixed. Experimental results on synthetic distributions and trained generative models on synthetic and real datasets illustrate that the method shows improved power performance compared to existing approaches."
                    },
                    {
                        "title": "Trade Co-occurrence, Trade Flow Decomposition, and Conditional Order Imbalance in Equity Markets",
                        "abstract": "The time proximity of high-frequency trades can contain a salient signal. In this paper, we propose a method to classify every trade, based on its proximity with other trades in the market within a short period of time, into five types. By means of a suitably defined normalized order imbalance associated to each type of trade, which we denote as conditional order imbalance (COI), we investigate the price impact of the decomposed trade flows. Our empirical findings indicate strong positive correlations between contemporaneous returns and COIs. In terms of predictability, we document that associations with future returns are positive for COIs of trades which are isolated from trades of stocks other than themselves, and negative otherwise. Furthermore, trading strategies which we develop using COIs achieve conspicuous returns and Sharpe ratios, in an extensive experimental setup on a universe of 457 stocks using daily data for a period of four years."
                    },
                    {
                        "title": "GNNRank: Learning Global Rankings from Pairwise Comparisons via Directed Graph Neural Networks",
                        "abstract": "Recovering global rankings from pairwise comparisons has wide applications from time synchronization to sports team ranking. Pairwise comparisons corresponding to matches in a competition can be construed as edges in a directed graph (digraph), whose nodes represent e.g. competitors with an unknown rank. In this paper, we introduce neural networks into the ranking recovery problem by proposing the so-called GNNRank, a trainable GNN-based framework with digraph embedding. Moreover, new objectives are devised to encode ranking upsets/violations. The framework involves a ranking score estimation approach, and adds an inductive bias by unfolding the Fiedler vector computation of the graph constructed from a learnable similarity matrix. Experimental results on extensive data sets show that our methods attain competitive and often superior performance against baselines, as well as showing promising transfer ability. Codes and preprocessed data are at: \\url{https://github.com/SherylHYX/GNNRank}."
                    },
                    {
                        "title": "Extracting Information from Gene Coexpression Networks of Rhizobium leguminosarum",
                        "abstract": "Nitrogen uptake in legumes is facilitated by bacteria such as Rhizobium leguminosarum. For this bacterium, gene expression data are available, but functional gene annotation is less well developed than for other model organisms. More annotations could lead to a better understanding of the pathways for growth, plant colonization, and nitrogen fixation in R. leguminosarum. In this study, we present a pipeline that combines novel scores from gene coexpression network analysis in a principled way to identify the genes that are associated with certain growth conditions or highly coexpressed with a predefined set of genes of interest. This association may lead to putative functional annotation or to a prioritized list of genes for further study."
                    },
                    {
                        "title": "AgraSSt: Approximate Graph Stein Statistics for Interpretable Assessment of Implicit Graph Generators",
                        "abstract": "We propose and analyse a novel statistical procedure, coined AgraSSt, to assess the quality of graph generators that may not be available in explicit form. In particular, AgraSSt can be used to determine whether a learnt graph generating process is capable of generating graphs that resemble a given input graph. Inspired by Stein operators for random graphs, the key idea of AgraSSt is the construction of a kernel discrepancy based on an operator obtained from the graph generator. AgraSSt can provide interpretable criticisms for a graph generator training procedure and help identify reliable sample batches for downstream tasks. Using Stein`s method we give theoretical guarantees for a broad class of random graph models. We provide empirical results on both synthetic input graphs with known graph generation procedures, and real-world input graphs that the state-of-the-art (deep) generative models for graphs are trained on."
                    },
                    {
                        "title": "PyTorch Geometric Signed Directed: A Survey and Software on Graph Neural Networks for Signed and Directed Graphs",
                        "abstract": "Signed networks are ubiquitous in many real-world applications (e.g., social networks encoding trust/distrust relationships, correlation networks arising from time series data). While many signed networks are directed, there is a lack of survey papers and software packages on graph neural networks (GNNs) specially designed for directed networks. In this paper, we present PyTorch Geometric Signed Directed, a survey and software on GNNs for signed and directed networks. We review typical tasks, loss functions and evaluation metrics in the analysis of signed and directed networks, discuss data used in related experiments, and provide an overview of methods proposed. The deep learning framework consists of easy-to-use GNN models, synthetic and real-world data, as well as task-specific evaluation metrics and loss functions for signed and directed networks. The software is presented in a modular fashion, so that signed and directed networks can also be treated separately. As an extension library for PyTorch Geometric, our proposed software is maintained with open-source re-leases, detailed documentation, continuous integration, unit tests and code coverage checks. Our code is publicly available at https: //github.com/SherylHYX/pytorch_geometric_signed_directed. \u2022 Software and its engineering \u2192 Designing software ; \u2022 Computing methodologies \u2192 Machine learning ."
                    },
                    {
                        "title": "On Papathanasiou\u2019s covariance expansions",
                        "abstract": ". In this paper we provide a probabilistic representation of Lagrange\u2019s identity which we use to obtain Papathanasiou-type variance expansions of arbitrary order. Our expansions lead to generalized sequences of weights which depend on an arbitrarily chosen sequence of (non-decreasing) test functions. The expansions hold for univariate target distribution under weak assumptions, in particular they hold for continuous and lattice distributions alike. The weights are studied under di\ufb00erent sets of assumptions either on the test functions or on the underlying distributions. Many concrete illustrations for standard probability distributions are provided (including Pearson, Ord, Laplace, Rayleigh, Cauchy, and Levy distributions)."
                    },
                    {
                        "title": "PyTorch Geometric Signed Directed: A Software Package on Graph Neural Networks for Signed and Directed Graphs",
                        "abstract": "Networks are ubiquitous in many real-world applications (e.g., social networks encoding trust/distrust relationships, correlation networks arising from time series data). While many networks are signed or directed, or both, there is a lack of unified software packages on graph neural networks (GNNs) specially designed for signed and directed networks. In this paper, we present PyTorch Geometric Signed Directed (PyGSD), a software package which fills this gap. Along the way, we evaluate the implemented methods with experiments with a view to providing insights into which method to choose for a given task. The deep learning framework consists of easy-to-use GNN models, synthetic and real-world data, as well as task-specific evaluation metrics and loss functions for signed and directed networks. As an extension library for PyG, our proposed software is maintained with open-source releases, detailed documentation, continuous integration, unit tests and code coverage checks. The GitHub repository of the library is https://github.com/SherylHYX/pytorch_geometric_signed_directed."
                    }
                ]
            },
            "68920b67-4e2b-4901-9bd3-6e4f21bc78b1": {
                "pk": "68920b67-4e2b-4901-9bd3-6e4f21bc78b1",
                "name": "David Wipf",
                "collaborators": [
                    "Minjie Wang",
                    "Tong He",
                    "Zhenkun Cai",
                    "Amber Yijia Zheng",
                    "Yixuan Qiu",
                    "Xiangkun Hu",
                    "Quan Gan",
                    "Ning Li",
                    "Jianheng Tang",
                    "Yanlin Zhang",
                    "Zizhao Zhang",
                    "Zunyao Mao",
                    "Yakun Song",
                    "Yanbo Wang",
                    "Jiahang Li",
                    "Han Zhang",
                    "Guang Yang",
                    "Xiao Qin",
                    "Chuan Lei",
                    "Mu-Nan Zhang",
                    "Weinan Zhang",
                    "Christos Faloutsos",
                    "Zheng Zhang",
                    "Haitian Jiang",
                    "Renjie Liu",
                    "Xiao Yan"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Bilevel Optimization",
                    "Machine Learning",
                    "Relational Databases"
                ],
                "publications": [
                    {
                        "title": "BloomGML: Graph Machine Learning through the Lens of Bilevel Optimization",
                        "abstract": "Bilevel optimization refers to scenarios whereby the optimal solution of a lower-level energy function serves as input features to an upper-level objective of interest. These optimal features typically depend on tunable parameters of the lower-level energy in such a way that the entire bilevel pipeline can be trained end-to-end. Although not generally presented as such, this paper demonstrates how a variety of graph learning techniques can be recast as special cases of bilevel optimization or simplifications thereof. In brief, building on prior work we first derive a more flexible class of energy functions that, when paired with various descent steps (e.g., gradient descent, proximal methods, momentum, etc.), form graph neural network (GNN) message-passing layers; critically, we also carefully unpack where any residual approximation error lies with respect to the underlying constituent message-passing functions. We then probe several simplifications of this framework to derive close connections with non-GNN-based graph learning approaches, including knowledge graph embeddings, various forms of label propagation, and efficient graph-regularized MLP models. And finally, we present supporting empirical results that demonstrate the versatility of the proposed bilevel lens, which we refer to as BloomGML, referencing that BiLevel Optimization Offers More Graph Machine Learning. Our code is available at https://github.com/amberyzheng/BloomGML. Let graph ML bloom."
                    },
                    {
                        "title": "New Desiderata for Direct Preference Optimization",
                        "abstract": "Large language models in the past have typically relied on some form of reinforcement learning with human feedback (RLHF) to better align model responses with human preferences. However, because of oft-observed instabilities when implementing these RLHF pipelines, various reparameterization techniques have recently been introduced to sidestep the need for separately learning an RL reward model. Instead, directly fine-tuning for human preferences is achieved via the minimization of a single closed-form training objective, a process originally referred to as direct preference optimization (DPO) and followed by several notable descendants. Although effective in certain real-world settings, we introduce new evaluation criteria that serve to highlight unresolved shortcomings in the ability of existing DPO methods to interpolate between a pre-trained reference model and empirical measures of human preferences, as well as unavoidable trade-offs in how low- and high-quality responses are regularized and constraints are handled. Our insights then motivate an alternative DPO-like loss that provably mitigates these limitations. Empirical results serve to corroborate notable aspects of our analyses."
                    },
                    {
                        "title": "4DBInfer: A 4D Benchmarking Toolbox for Graph-Centric Predictive Modeling on Relational DBs",
                        "abstract": "Although RDBs store vast amounts of rich, informative data spread across interconnected tables, the progress of predictive machine learning models as applied to such tasks arguably falls well behind advances in other domains such as computer vision or natural language processing. This deficit stems, at least in part, from the lack of established/public RDB benchmarks as needed for training and evaluation purposes. As a result, related model development thus far often defaults to tabular approaches trained on ubiquitous single-table benchmarks, or on the relational side, graph-based alternatives such as GNNs applied to a completely different set of graph datasets devoid of tabular characteristics. To more precisely target RDBs lying at the nexus of these two complementary regimes, we explore a broad class of baseline models predicated on: (i) converting multi-table datasets into graphs using various strategies equipped with efficient subsampling, while preserving tabular characteristics; and (ii) trainable models with well-matched inductive biases that output predictions based on these input subgraphs. Then, to address the dearth of suitable public benchmarks and reduce siloed comparisons, we assemble a diverse collection of (i) large-scale RDB datasets and (ii) coincident predictive tasks. From a delivery standpoint, we operationalize the above four dimensions (4D) of exploration within a unified, scalable open-source toolbox called 4DBInfer. We conclude by presenting evaluations using 4DBInfer, the results of which highlight the importance of considering each such dimension in the design of RDB predictive models, as well as the limitations of more naive approaches such as simply joining adjacent tables. Our source code is released at https://github.com/awslabs/multi-table-benchmark ."
                    },
                    {
                        "title": "MuseGNN: Interpretable and Convergent Graph Neural Network Layers at Scale",
                        "abstract": "Among the many variants of graph neural network (GNN) architectures capable of modeling data with cross-instance relations, an important subclass involves layers designed such that the forward pass iteratively reduces a graph-regularized energy function of interest. In this way, node embeddings produced at the output layer dually serve as both predictive features for solving downstream tasks (e.g., node classification) and energy function minimizers that inherit desirable inductive biases and interpretability. However, scaling GNN architectures constructed in this way remains challenging, in part because the convergence of the forward pass may involve models with considerable depth. To tackle this limitation, we propose a sampling-based energy function and scalable GNN layers that iteratively reduce it, guided by convergence guarantees in certain settings. We also instantiate a full GNN architecture based on these designs, and the model achieves competitive accuracy and scalability when applied to the largest publicly-available node classification benchmark exceeding 1TB in size."
                    }
                ]
            },
            "3a0054f0-9f11-4735-b1d4-39d077f1f006": {
                "pk": "3a0054f0-9f11-4735-b1d4-39d077f1f006",
                "name": "Mihai Cucuringu",
                "collaborators": [
                    "G. Reinert",
                    "Yixuan He",
                    "Andrew Elliott",
                    "Stratis Limnios",
                    "Yutong Lu",
                    "Xitong Zhang",
                    "Junjie Huang",
                    "Ruikang Ouyang",
                    "Jase Clarkson",
                    "Wenkai Xu",
                    "Daniel Cunha Oliveira",
                    "Xi Lin",
                    "Andre Fujita",
                    "Praveen Selvaraj",
                    "C. Maple",
                    "Anastasia Mantziou",
                    "Victor Meirinhos",
                    "Quan Gan",
                    "D. Wipf",
                    "Junchi Yan",
                    "Benedek Rozemberczki",
                    "Michael Perlmutter",
                    "J. Clarkson",
                    "Songchao Wang"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Time Series Analysis",
                    "Generative Models",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "L2G2G: a Scalable Local-to-Global Network Embedding with Graph Autoencoders",
                        "abstract": "For analysing real-world networks, graph representation learning is a popular tool. These methods, such as a graph autoencoder (GAE), typically rely on low-dimensional representations, also called embeddings, which are obtained through minimising a loss function; these embeddings are used with a decoder for downstream tasks such as node classification and edge prediction. While GAEs tend to be fairly accurate, they suffer from scalability issues. For improved speed, a Local2Global approach, which combines graph patch embeddings based on eigenvector synchronisation, was shown to be fast and achieve good accuracy. Here we propose L2G2G, a Local2Global method which improves GAE accuracy without sacrificing scalability. This improvement is achieved by dynamically synchronising the latent node representations, while training the GAEs. It also benefits from the decoder computing an only local patch loss. Hence, aligning the local embeddings in each epoch utilises more information from the graph than a single post-training alignment does, while maintaining scalability. We illustrate on synthetic benchmarks, as well as real-world examples, that L2G2G achieves higher accuracy than the standard Local2Global approach and scales efficiently on the larger data sets. We find that for large and dense networks, it even outperforms the slow, but assumed more accurate, GAEs."
                    },
                    {
                        "title": "Split Conformal Prediction under Data Contamination",
                        "abstract": "Conformal prediction is a non-parametric technique for constructing prediction intervals or sets from arbitrary predictive models under the assumption that the data is exchangeable. It is popular as it comes with theoretical guarantees on the marginal coverage of the prediction sets and the split conformal prediction variant has a very low computational cost compared to model training. We study the robustness of split conformal prediction in a data contamination setting, where we assume a small fraction of the calibration scores are drawn from a different distribution than the bulk. We quantify the impact of the corrupted data on the coverage and efficiency of the constructed sets when evaluated on\"clean\"test points, and verify our results with numerical experiments. Moreover, we propose an adjustment in the classification setting which we call Contamination Robust Conformal Prediction, and verify the efficacy of our approach using both synthetic and real datasets."
                    },
                    {
                        "title": "Causality-Inspired Models for Financial Time Series Forecasting",
                        "abstract": "We introduce a novel framework to financial time series forecasting that leverages causality-inspired models to balance the trade-off between invariance to distributional changes and minimization of prediction errors. To the best of our knowledge, this is the first study to conduct a comprehensive comparative analysis among state-of-the-art causal discovery algorithms, benchmarked against non-causal feature selection techniques, in the application of forecasting asset returns. Empirical evaluations demonstrate the efficacy of our approach in yielding stable and accurate predictions, outperforming baseline models, particularly in tumultuous market conditions."
                    },
                    {
                        "title": "SaGess: Sampling Graph Denoising Diffusion Model for Scalable Graph Generation",
                        "abstract": "Over recent years, denoising diffusion generative models have come to be considered as state-of-the-art methods for synthetic data generation, especially in the case of generating images. These approaches have also proved successful in other applications such as tabular and graph data generation. However, due to computational complexity, to this date, the application of these techniques to graph data has been restricted to small graphs, such as those used in molecular modeling. In this paper, we propose SaGess, a discrete denoising diffusion approach, which is able to generate large real-world networks by augmenting a diffusion model (DiGress) with a generalized divide-and-conquer framework. The algorithm is capable of generating larger graphs by sampling a covering of subgraphs of the initial graph in order to train DiGress. SaGess then constructs a synthetic graph using the subgraphs that have been generated by DiGress. We evaluate the quality of the synthetic data sets against several competitor methods by comparing graph statistics between the original and synthetic samples, as well as evaluating the utility of the synthetic data set produced by using it to train a task-driven model, namely link prediction. In our experiments, SaGess, outperforms most of the one-shot state-of-the-art graph generating methods by a significant factor, both on the graph metrics and on the link prediction task."
                    },
                    {
                        "title": "The GNAR-edge model: a network autoregressive model for networks with time-varying edge weights",
                        "abstract": "  In economic and financial applications, there is often the need for analysing multivariate time series, comprising of time series for a range of quantities. In some applications, such complex systems can be associated with some underlying network describing pairwise relationships among the quantities. Accounting for the underlying network structure for the analysis of this type of multivariate time series is required for assessing estimation error and can be particularly informative for forecasting. Our work is motivated by a dataset consisting of time series of industry-to-industry transactions. In this example, pairwise relationships between Standard Industrial Classification (SIC) codes can be represented using a network, with SIC codes as nodes and pairwise transactions between SIC codes as edges, while the observed time series of the amounts of the transactions for each pair of SIC codes can be regarded as time-varying weights on the edges. Inspired by Knight et al. (2020, J. Stat. Softw., 96, 1\u201336), we introduce the GNAR-edge model which allows modelling of multiple time series utilizing the network structure, assuming that each edge weight depends not only on its past values, but also on past values of its neighbouring edges, for a range of neighbourhood stages. The method is validated through simulations. Results from the implementation of the GNAR-edge model on the real industry-to-industry data show good fitting and predictive performance of the model. The predictive performance is improved when sparsifying the network using a lead\u2013lag analysis and thresholding edges according to a lead\u2013lag score."
                    },
                    {
                        "title": "Co-Trading Networks for Modeling Dynamic Interdependency Structures and Estimating High-Dimensional Covariances in US Equity Markets",
                        "abstract": "The time proximity of trades across stocks reveals interesting topological structures of the equity market in the United States. In this article, we investigate how such concurrent cross-stock trading behaviors, which we denote as co-trading, shape the market structures and affect stock price co-movements. By leveraging a co-trading-based pairwise similarity measure, we propose a novel method to construct dynamic networks of stocks. Our empirical studies employ high-frequency limit order book data from 2017-01-03 to 2019-12-09. By applying spectral clustering on co-trading networks, we uncover economically meaningful clusters of stocks. Beyond the static Global Industry Classification Standard (GICS) sectors, our data-driven clusters capture the time evolution of the dependency among stocks. Furthermore, we demonstrate statistically significant positive relations between low-latency co-trading and return covariance. With the aid of co-trading networks, we develop a robust estimator for high-dimensional covariance matrix, which yields superior economic value on portfolio allocation. The mean-variance portfolios based on our covariance estimates achieve both lower volatility and higher Sharpe ratios than standard benchmarks."
                    },
                    {
                        "title": "Trade Co-occurrence, Trade Flow Decomposition, and Conditional Order Imbalance in Equity Markets",
                        "abstract": "The time proximity of high-frequency trades can contain a salient signal. In this paper, we propose a method to classify every trade, based on its proximity with other trades in the market within a short period of time, into five types. By means of a suitably defined normalized order imbalance associated to each type of trade, which we denote as conditional order imbalance (COI), we investigate the price impact of the decomposed trade flows. Our empirical findings indicate strong positive correlations between contemporaneous returns and COIs. In terms of predictability, we document that associations with future returns are positive for COIs of trades which are isolated from trades of stocks other than themselves, and negative otherwise. Furthermore, trading strategies which we develop using COIs achieve conspicuous returns and Sharpe ratios, in an extensive experimental setup on a universe of 457 stocks using daily data for a period of four years."
                    },
                    {
                        "title": "GNNRank: Learning Global Rankings from Pairwise Comparisons via Directed Graph Neural Networks",
                        "abstract": "Recovering global rankings from pairwise comparisons has wide applications from time synchronization to sports team ranking. Pairwise comparisons corresponding to matches in a competition can be construed as edges in a directed graph (digraph), whose nodes represent e.g. competitors with an unknown rank. In this paper, we introduce neural networks into the ranking recovery problem by proposing the so-called GNNRank, a trainable GNN-based framework with digraph embedding. Moreover, new objectives are devised to encode ranking upsets/violations. The framework involves a ranking score estimation approach, and adds an inductive bias by unfolding the Fiedler vector computation of the graph constructed from a learnable similarity matrix. Experimental results on extensive data sets show that our methods attain competitive and often superior performance against baselines, as well as showing promising transfer ability. Codes and preprocessed data are at: \\url{https://github.com/SherylHYX/GNNRank}."
                    },
                    {
                        "title": "PyTorch Geometric Signed Directed: A Survey and Software on Graph Neural Networks for Signed and Directed Graphs",
                        "abstract": "Signed networks are ubiquitous in many real-world applications (e.g., social networks encoding trust/distrust relationships, correlation networks arising from time series data). While many signed networks are directed, there is a lack of survey papers and software packages on graph neural networks (GNNs) specially designed for directed networks. In this paper, we present PyTorch Geometric Signed Directed, a survey and software on GNNs for signed and directed networks. We review typical tasks, loss functions and evaluation metrics in the analysis of signed and directed networks, discuss data used in related experiments, and provide an overview of methods proposed. The deep learning framework consists of easy-to-use GNN models, synthetic and real-world data, as well as task-specific evaluation metrics and loss functions for signed and directed networks. The software is presented in a modular fashion, so that signed and directed networks can also be treated separately. As an extension library for PyTorch Geometric, our proposed software is maintained with open-source re-leases, detailed documentation, continuous integration, unit tests and code coverage checks. Our code is publicly available at https: //github.com/SherylHYX/pytorch_geometric_signed_directed. \u2022 Software and its engineering \u2192 Designing software ; \u2022 Computing methodologies \u2192 Machine learning ."
                    },
                    {
                        "title": "PyTorch Geometric Signed Directed: A Software Package on Graph Neural Networks for Signed and Directed Graphs",
                        "abstract": "Networks are ubiquitous in many real-world applications (e.g., social networks encoding trust/distrust relationships, correlation networks arising from time series data). While many networks are signed or directed, or both, there is a lack of unified software packages on graph neural networks (GNNs) specially designed for signed and directed networks. In this paper, we present PyTorch Geometric Signed Directed (PyGSD), a software package which fills this gap. Along the way, we evaluate the implemented methods with experiments with a view to providing insights into which method to choose for a given task. The deep learning framework consists of easy-to-use GNN models, synthetic and real-world data, as well as task-specific evaluation metrics and loss functions for signed and directed networks. As an extension library for PyG, our proposed software is maintained with open-source releases, detailed documentation, continuous integration, unit tests and code coverage checks. The GitHub repository of the library is https://github.com/SherylHYX/pytorch_geometric_signed_directed."
                    },
                    {
                        "title": "MSGNN: A Spectral Graph Neural Network Based on a Novel Magnetic Signed Laplacian",
                        "abstract": "Signed and directed networks are ubiquitous in real-world applications. However, there has been relatively little work proposing spectral graph neural networks (GNNs) for such networks. Here we introduce a signed directed Laplacian matrix, which we call the magnetic signed Laplacian, as a natural generalization of both the signed Laplacian on signed graphs and the magnetic Laplacian on directed graphs. We then use this matrix to construct a novel efficient spectral GNN architecture and conduct extensive experiments on both node clustering and link prediction tasks. In these experiments, we consider tasks related to signed information, tasks related to directional information, and tasks related to both signed and directional information. We demonstrate that our proposed spectral GNN is effective for incorporating both signed and directional information, and attains leading performance on a wide range of data sets. Additionally, we provide a novel synthetic network model, which we refer to as the Signed Directed Stochastic Block Model, and a number of novel real-world data sets based on lead-lag relationships in financial time series."
                    },
                    {
                        "title": "Random Walk based Conditional Generative Model for Temporal Networks with Attributes",
                        "abstract": "We propose a novel method for graph time series generation with node and edge attributes. As graph representations for complex data become increasingly popular, we encounter many time series of graphs with temporal and attribute dependencies, such as communication networks, daily bike rentals or bank transactions. However, the analysis of such graphs can be impeded by privacy or data protection issues, calling for synthetic network time series which serve as surrogate of the observed time series. There are many methods to construct networks, for example transformer-based graph models such as T AG G EN , but it is harder to create data which are faithful to the observed dependencies, including those of attributes across time. Moreover, tabular data generation methods, such as GANs, fail to emulate complex graph structures. To circumvent these limitations, we introduce CTW ALK which combines and leverages the strengths of both methods, by coupling T AG G EN that learns the distribution of temporal random walks over the input data, and a conditional tabular GAN (CTGAN) that captures the time dependence of the features. CTW ALK is able to mimic edge weight distributions, node labels, and temporal dependencies of the data."
                    },
                    {
                        "title": "DAMNETS: A Deep Autoregressive Model for Generating Markovian Network Time Series",
                        "abstract": "Generative models for network time series (also known as dynamic graphs) have tremendous potential in fields such as epidemiology, biology and economics, where complex graph-based dynamics are core objects of study. Designing flexible and scalable generative models is a very challenging task due to the high dimensionality of the data, as well as the need to represent temporal dependencies and marginal network structure. Here we introduce DAMNETS, a scalable deep generative model for network time series. DAMNETS outperforms competing methods on all of our measures of sample quality, over both real and synthetic data sets."
                    },
                    {
                        "title": "DIGRAC: Digraph Clustering Based on Flow Imbalance",
                        "abstract": "Node clustering is a powerful tool in the analysis of networks. We introduce a graph neural network framework, named DIGRAC, to obtain node embeddings for directed networks in a self-supervised manner, including a novel probabilistic imbalance loss, which can be used for network clustering. Here, we propose \\textit{directed flow imbalance} measures, which are tightly related to directionality, to reveal clusters in the network even when there is no density difference between clusters. In contrast to standard approaches in the literature, in this paper, directionality is not treated as a nuisance, but rather contains the main signal. DIGRAC optimizes directed flow imbalance for clustering without requiring label supervision, unlike existing graph neural network methods, and can naturally incorporate node features, unlike existing spectral methods. Extensive experimental results on synthetic data, in the form of directed stochastic block models, and real-world data at different scales, demonstrate that our method, based on flow imbalance, attains state-of-the-art results on directed graph clustering when compared against 10 state-of-the-art methods from the literature, for a wide range of noise and sparsity levels, graph structures, and topologies, and even outperforms supervised methods."
                    },
                    {
                        "title": "SSSNET: Semi-Supervised Signed Network Clustering",
                        "abstract": "Node embeddings are a powerful tool in the analysis of networks; yet, their full potential for the important task of node clustering has not been fully exploited. In particular, most state-of-the-art methods generating node embeddings of signed networks focus on link sign prediction, and those that pertain to node clustering are usually not graph neural network (GNN) methods. Here, we introduce a novel probabilistic balanced normalized cut loss for training nodes in a GNN framework for semi-supervised signed network clustering, called SSSNET. The method is end-to-end in combining embedding generation and clustering without an intermediate step; it has node clustering as main focus, with an emphasis on polarization effects arising in networks. The main novelty of our approach is a new take on the role of social balance theory for signed network embeddings. The standard heuristic for justifying the criteria for the embeddings hinges on the assumption that\"an enemy's enemy is a friend\". Here, instead, a neutral stance is assumed on whether or not the enemy of an enemy is a friend. Experimental results on various data sets, including a synthetic signed stochastic block model, a polarized version of it, and real-world data at different scales, demonstrate that SSSNET can achieve comparable or better results than state-of-the-art spectral clustering methods, for a wide range of noise and sparsity levels. SSSNET complements existing methods through the possibility of including exogenous information, in the form of node-level features or labels."
                    },
                    {
                        "title": "DIGRAC: Digraph Clustering with Flow Imbalance",
                        "abstract": "Node clustering is a powerful tool in the analysis of networks. Here, we introduce a graph neural network framework with a novel scalable Directed Mixed Path Aggregation (DIMPA) scheme to obtain node embeddings for directed networks in a self-supervised manner, including a novel probabilistic imbalance loss. The method is end-to-end in combining embedding generation and clustering without an intermediate step. In contrast to standard approaches in the literature, in this paper, directionality is not treated as a nuisance, but rather contains the main signal. In particular, we leverage the recently introduced cut \ufb02ow imbalance measure, which is tightly related to directionality; cut \ufb02ow imbalance is optimized without resorting to spectral methods or cluster labels. Experimental results on synthetic data, in the form of directed stochastic block models and real-world data at different scales, demonstrate that our method attains state-of-the-art results on directed graph clustering, for a wide range of noise and sparsity levels and graph structures."
                    }
                ]
            }
        }
    },
    "2303.04209": {
        "paper_data": {
            "title": "Causal Dependence Plots",
            "url": "http://arxiv.org/abs/2303.04209v2",
            "arxiv_id": "2303.04209",
            "authors": [
                "Joshua R. Loftus",
                "Lucius E. J. Bynum",
                "Sakina Hansen"
            ],
            "abstract": "Explaining artificial intelligence or machine learning models is increasingly important. To use such data-driven systems wisely we must understand how they interact with the world, including how they depend causally on data inputs. In this work we develop Causal Dependence Plots (CDPs) to visualize how one variable--an outcome--depends on changes in another variable--a predictor--$\\textit{along with any consequent causal changes in other predictor variables}$. Crucially, CDPs differ from standard methods based on holding other predictors constant or assuming they are independent. CDPs make use of an auxiliary causal model because causal conclusions require causal assumptions. With simulations and real data experiments, we show CDPs can be combined in a modular way with methods for causal learning or sensitivity analysis. Since people often think causally about input-output dependence, CDPs can be powerful tools in the xAI or interpretable machine learning toolkit and contribute to applications like scientific machine learning and algorithmic fairness.",
            "introduction": "   1 Introduction  This paper develops Causal Dependence Plots (CDPs) to visualize causal relationships between predictor variables and an outcome variable. The idea is general, but we are motivated by explaining or interpreting AI or machine learning models (Guidotti et\u00a0al., 2018; Carvalho et\u00a0al., 2019; Gunning et\u00a0al., 2021; Molnar, 2022). For simplicity we consider supervised learning where a set of features is used to predict an outcome, i.e. regression or classification. We also focus on the model-agnostic or \"black-box\" explanation setting, where the interpreter can query the model but not access its internal structure. Interpretation methods in this setting are more broadly applicable for distributed research, but are also functionally limited to observing how the model responds to variation in the inputs. Our general approach has extensions beyond this initial application.   Visualizations and simple explanations that focus on one input variable at a time can be powerful tools for human understanding. Two popular visualization methods, the Partial Dependence Plot (PDP) of Friedman (2001) and Individual Conditional Expectation (ICE) plot from Goldstein et\u00a0al. (2015), show how model output depends on one input variable. However, just as with the interpretation of linear regression model coefficients, the relationships revealed by focusing on one predictor at a time can be misleading. When varying one input variable, we must make some choice about what values to use for the other inputs. PDPs treat other predictors as independent of the one being plotted, so they can correctly capture the model\u2019s dependence on each variable if predictors are independent and the model is additive (Hastie and Tibshirani, 1986). In general, when explaining the black-box\u2019s dependence on one input, our choice of how to handle other model inputs may break or respect the existing statistical or causal dependencies between predictors. This leads us to the following:   Problem statement. If there are causal relationships between predictors in the real world, but our visualization, interpretation, or explanation method does not respect them, then the resulting model explanation may be irrelevant or misleading for real world purposes Moraffah et\u00a0al. (2020); Shin (2021). For example, such explanations could lead to incorrect decisions about regulating or aligning algorithmic systems, sub-optimal allocations of resources based on model predictions, a breakdown between human feedback and reinforcement learning systems, or other forms of harm. In the context of scientific machine learning\u2014where explanations can be used to generate hypotheses for follow-up investigation\u2014a flawed interpretation may direct us toward spurious hypotheses. For these reasons, we may care about the causal validity of model explanations.   Our proposal.  At a high level, we use an auxiliary Explanatory Causal Model (ECM) to interpret or explain a given machine learning model. For each input predictor that we wish to explain, we use the ECM to determine how other inputs vary when that predictor is manipulated, rather than treating them as independent or fixed. We call the resulting plots Causal Dependence Plots or CDPs.   Causal models can be designed based on the desired explanation, specified with prior domain knowledge, and/or potentially learned and estimated from data.    Motivating example.  Before turning to the full details, we illustrate the idea with a simple example. Consider a mediation model",
            "references": [
                {
                    "title": "Counterfactuals for the Future",
                    "abstract": "Counterfactuals are often described as 'retrospective,' focusing on hypothetical alternatives to a realized past. This description relates to an often implicit assumption about the structure and stability of exogenous variables in the system being modeled --- an assumption that is reasonable in many settings where counterfactuals are used. In this work, we consider cases where we might reasonably make a different assumption about exogenous variables; namely, that the exogenous noise terms of each unit do exhibit some unit-specific structure and/or stability. This leads us to a different use of counterfactuals --- a forward-looking rather than retrospective counterfactual. We introduce \"counterfactual treatment choice,\" a type of treatment choice problem that motivates using forward-looking counterfactuals. We then explore how mismatches between interventional versus forward-looking counterfactual approaches to treatment choice, consistent with different assumptions about exogenous noise, can lead to counterintuitive results."
                },
                {
                    "title": "DoWhy-GCM: An extension of DoWhy for causal inference in graphical causal models",
                    "abstract": "We present DoWhy-GCM, an extension of the DoWhy Python library, which leverages graphical causal models. Unlike existing causality libraries, which mainly focus on effect estimation, DoWhy-GCM addresses diverse causal queries, such as identifying the root causes of outliers and distributional changes, attributing causal influences to the data generating process of each node, or diagnosis of causal structures. With DoWhy-GCM, users typically specify cause-effect relations via a causal graph, fit causal mechanisms, and pose causal queries -- all with just a few lines of code. The general documentation is available at https://www.pywhy.org/dowhy and the DoWhy-GCM specific code at https://github.com/py-why/dowhy/tree/main/dowhy/gcm."
                },
                {
                    "title": "Conformal Sensitivity Analysis for Individual Treatment Effects",
                    "abstract": "Abstract Estimating an individual treatment effect (ITE) is essential to personalized decision making. However, existing methods for estimating the ITE often rely on unconfoundedness, an assumption that is fundamentally untestable with observed data. To assess the robustness of individual-level causal conclusion with unconfoundedness, this article proposes a method for sensitivity analysis of the ITE, a way to estimate a range of the ITE under unobserved confounding. The method we develop quantifies unmeasured confounding through a marginal sensitivity model, and adapts the framework of conformal inference to estimate an ITE interval at a given confounding strength. In particular, we formulate this sensitivity analysis as a conformal inference problem under distribution shift, and we extend existing methods of covariate-shifted conformal inference to this more general setting. The resulting predictive interval has guaranteed nominal coverage of the ITE and provides this coverage with distribution-free and nonasymptotic guarantees. We evaluate the method on synthetic data and illustrate its application in an observational study. Supplementary materials for this article are available online."
                },
                {
                    "title": "Sensitivity analysis of individual treatment effects: A robust conformal inference approach",
                    "abstract": "Significance The individual treatment effect (ITE) describes the difference between an individual\u2019s outcome when receiving a treatment versus not. This difference may vary across individuals conditional on their characteristics. In observational studies, inference for ITEs can be invalid if one ignores unmeasured confounding factors that simultaneously influence the treatment assignment and the outcomes. We propose a framework to quantitatively understand the robustness of causal conclusions on ITEs against such potential confounding factors. This yields prediction bands, which come with rigorous uncertainty quantification tools. These tools apply regardless of the machine learning model employed to learn the treatment effect, however complicated, and regardless of the sample size."
                },
                {
                    "title": "DARPA\u2019s Explainable AI (XAI) program: A retrospective",
                    "abstract": "DARPA formulated the Explainable Artificial Intelligence (XAI) program\nin 2015 with the goal to enable end users to better understand, trust,\nand effectively manage artificially intelligent systems. In 2017, the\nfour-year XAI research program began. Now, as XAI comes to an end in\n2021, it is time to reflect on what succeeded, what failed, and what was\nlearned. This article summarizes the goals, organization, and research\nprogress of the XAI Program."
                },
                {
                    "title": "Disaggregated Interventions to Reduce Inequality",
                    "abstract": "A significant body of research in the data sciences considers unfair discrimination against social categories such as race or gender that could occur or be amplified as a result of algorithmic decisions. Simultaneously, real-world disparities continue to exist, even before algorithmic decisions are made. In this work, we draw on insights from the social sciences brought into the realm of causal modeling and constrained optimization, and develop a novel algorithmic framework for tackling pre-existing real-world disparities. The purpose of our framework, which we call the \u201cimpact remediation framework,\u201d is to measure real-world disparities and discover the optimal intervention policies that could help improve equity or access to opportunity for those who are underserved with respect to an outcome of interest. We develop a disaggregated approach to tackling pre-existing disparities that relaxes the typical set of assumptions required for the use of social categories in structural causal models. Our approach flexibly incorporates counterfactuals and is compatible with various ontological assumptions about the nature of social categories. We demonstrate impact remediation with a hypothetical case study and compare our disaggregated approach to an existing state-of-the-art approach, comparing its structure and resulting policy recommendations. In contrast to most work on optimal policy learning, we explore disparity reduction itself as an objective, explicitly focusing the power of algorithms on reducing inequality."
                },
                {
                    "title": "DoWhy: An End-to-End Library for Causal Inference",
                    "abstract": "In addition to efficient statistical estimators of a treatment's effect, successful application of causal inference requires specifying assumptions about the mechanisms underlying observed data and testing whether they are valid, and to what extent. However, most libraries for causal inference focus only on the task of providing powerful statistical estimators. We describe DoWhy, an open-source Python library that is built with causal assumptions as its first-class citizens, based on the formal framework of causal graphs to specify and test causal assumptions. DoWhy presents an API for the four steps common to any causal analysis---1) modeling the data using a causal graph and structural assumptions, 2) identifying whether the desired effect is estimable under the causal model, 3) estimating the effect using statistical estimators, and finally 4) refuting the obtained estimate through robustness checks and sensitivity analyses. In particular, DoWhy implements a number of robustness checks including placebo tests, bootstrap tests, and tests for unoberved confounding. DoWhy is an extensible library that supports interoperability with other implementations, such as EconML and CausalML for the the estimation step. The library is available at this https URL"
                },
                {
                    "title": "Survey on Causal-based Machine Learning Fairness Notions",
                    "abstract": "Addressing the problem of fairness is crucial to safely use machine learning algorithms to support decisions with a critical impact on people's lives such as job hiring, child maltreatment, disease diagnosis, loan granting, etc. Several notions of fairness have been defined and examined in the past decade, such as, statistical parity and equalized odds. The most recent fairness notions, however, are causal-based and reflect the now widely accepted idea that using causality is necessary to appropriately address the problem of fairness. This paper examines an exhaustive list of causal-based fairness notions, in particular their applicability in real-world scenarios. As the majority of causal-based fairness notions are defined in terms of non-observable quantities (e.g. interventions and counterfactuals), their applicability depends heavily on the identifiability of those quantities from observational data. In this paper, we compile the most relevant identifiability criteria for the problem of fairness from the extensive literature on identifiability theory. These criteria are then used to decide about the applicability of causal-based fairness notions in concrete discrimination scenarios."
                },
                {
                    "title": "A survey of algorithmic recourse: definitions, formulations, solutions, and prospects",
                    "abstract": "Machine learning is increasingly used to inform decision-making in sensitive situations where decisions have consequential effects on individuals' lives. In these settings, in addition to requiring models to be accurate and robust, socially relevant values such as fairness, privacy, accountability, and explainability play an important role for the adoption and impact of said technologies. In this work, we focus on algorithmic recourse, which is concerned with providing explanations and recommendations to individuals who are unfavourably treated by automated decision-making systems. We first perform an extensive literature review, and align the efforts of many authors by presenting unified definitions, formulations, and solutions to recourse. Then, we provide an overview of the prospective research directions towards which the community may engage, challenging existing assumptions and making explicit connections to other ethical challenges such as security, privacy, and fairness."
                },
                {
                    "title": "Conformal inference of counterfactuals and individual treatment effects",
                    "abstract": "Evaluating treatment effect heterogeneity widely informs treatment decision making. At the moment, much emphasis is placed on the estimation of the conditional average treatment effect via flexible machine learning algorithms. While these methods enjoy some theoretical appeal in terms of consistency and convergence rates, they generally perform poorly in terms of uncertainty quantification. This is troubling since assessing risk is crucial for reliable decision\u2010making in sensitive and uncertain environments. In this work, we propose a conformal inference\u2010based approach that can produce reliable interval estimates for counterfactuals and individual treatment effects under the potential outcome framework. For completely randomized or stratified randomized experiments with perfect compliance, the intervals have guaranteed average coverage in finite samples regardless of the unknown data generating mechanism. For randomized experiments with ignorable compliance and general observational studies obeying the strong ignorability assumption, the intervals satisfy a doubly robust property which states the following: the average coverage is approximately controlled if either the propensity score or the conditional quantiles of potential outcomes can be estimated accurately. Numerical studies on both synthetic and real data sets empirically demonstrate that existing methods suffer from a significant coverage deficit even in simple models. In contrast, our methods achieve the desired coverage with reasonably short intervals."
                },
                {
                    "title": "Explaining The Behavior Of Black-Box Prediction Algorithms With Causal Learning",
                    "abstract": "We propose to explain the behavior of black-box prediction methods (e.g., deep neural networks trained on image pixel data) using causal graphical models. Specifically, we explore learning the structure of a causal graph where the nodes represent prediction outcomes along with a set of macro-level \"interpretable\" features, while allowing for arbitrary unmeasured confounding among these variables. The resulting graph may indicate which of the interpretable features, if any, are possible causes of the prediction outcome and which may be merely associated with prediction outcomes due to confounding. The approach is motivated by a counterfactual theory of causal explanation wherein good explanations point to factors which are \"difference-makers\" in an interventionist sense. The resulting analysis may be useful in algorithm auditing and evaluation, by identifying features which make a causal difference to the algorithm's output."
                },
                {
                    "title": "Causal Interpretability for Machine Learning - Problems, Methods and Evaluation",
                    "abstract": "Machine learning models have had discernible achievements in a myriad of applications. However, most of these models are black-boxes, and it is obscure how the decisions are made by them. This makes the models unreliable and untrustworthy. To provide insights into the decision making processes of these models, a variety of traditional interpretable models have been proposed. Moreover, to generate more humanfriendly explanations, recent work on interpretability tries to answer questions related to causality such as \"Why does this model makes such decisions?\" or \"Was it a specific feature that caused the decision made by the model?\". In this work, models that aim to answer causal questions are referred to as causal interpretable models. The existing surveys have covered concepts and methodologies of traditional interpretability. In this work, we present a comprehensive survey on causal interpretable models from the aspects of the problems and methods. In addition, this survey provides in-depth insights into the existing evaluation metrics for measuring interpretability, which can help practitioners understand for what scenarios each evaluation metric is suitable."
                },
                {
                    "title": "FACE: Feasible and Actionable Counterfactual Explanations",
                    "abstract": "Work in Counterfactual Explanations tends to focus on the principle of \"the closest possible world\" that identifies small changes leading to the desired outcome. In this paper we argue that while this approach might initially seem intuitively appealing it exhibits shortcomings not addressed in the current literature. First, a counterfactual example generated by the state-of-the-art systems is not necessarily representative of the underlying data distribution, and may therefore prescribe unachievable goals (e.g., an unsuccessful life insurance applicant with severe disability may be advised to do more sports). Secondly, the counterfactuals may not be based on a \"feasible path\" between the current state of the subject and the suggested one, making actionable recourse infeasible (e.g., low-skilled unsuccessful mortgage applicants may be told to double their salary, which may be hard without first increasing their skill level). These two shortcomings may render counterfactual explanations impractical and sometimes outright offensive. To address these two major flaws, first of all, we propose a new line of Counterfactual Explanations research aimed at providing actionable and feasible paths to transform a selected instance into one that meets a certain goal. Secondly, we propose FACE: an algorithmically sound way of uncovering these \"feasible paths\" based on the shortest path distances defined via density-weighted metrics. Our approach generates counterfactuals that are coherent with the underlying data distribution and supported by the \"feasible paths\" of change, which are achievable and can be tailored to the problem at hand."
                },
                {
                    "title": "Machine Learning Interpretability: A Survey on Methods and Metrics",
                    "abstract": "Machine learning systems are becoming increasingly ubiquitous. These systems\u2019s adoption has been expanding, accelerating the shift towards a more algorithmic society, meaning that algorithmically informed decisions have greater potential for significant social impact. However, most of these accurate decision support systems remain complex black boxes, meaning their internal logic and inner workings are hidden to the user and even experts cannot fully understand the rationale behind their predictions. Moreover, new regulations and highly regulated domains have made the audit and verifiability of decisions mandatory, increasing the demand for the ability to question, understand, and trust machine learning systems, for which interpretability is indispensable. The research community has recognized this interpretability problem and focused on developing both interpretable models and explanation methods over the past few years. However, the emergence of these methods shows there is no consensus on how to assess the explanation quality. Which are the most suitable metrics to assess the quality of an explanation? The aim of this article is to provide a review of the current state of the research field on machine learning interpretability while focusing on the societal impact and on the developed methods and metrics. Furthermore, a complete literature review is presented in order to identify future directions of work on this field."
                },
                {
                    "title": "Causal Interpretations of Black-Box Models",
                    "abstract": "Abstract The fields of machine learning and causal inference have developed many concepts, tools, and theory that are potentially useful for each other. Through exploring the possibility of extracting causal interpretations from black-box machine-trained models, we briefly review the languages and concepts in causal inference that may be interesting to machine learning researchers. We start with the curious observation that Friedman\u2019s partial dependence plot has exactly the same formula as Pearl\u2019s back-door adjustment and discuss three requirements to make causal interpretations: a model with good predictive performance, some domain knowledge in the form of a causal diagram and suitable visualization tools. We provide several illustrative examples and find some interesting and potentially causal relations using visualization tools for black-box models."
                },
                {
                    "title": "Actionable Recourse in Linear Classification",
                    "abstract": "Classification models are often used to make decisions that affect humans: whether to approve a loan application, extend a job offer, or provide insurance. In such applications, individuals should have the ability to change the decision of the model. When a person is denied a loan by a credit scoring model, for example, they should be able to change the input variables of the model in a way that will guarantee approval. Otherwise, this person will be denied the loan so long as the model is deployed, and -- more importantly --will lack agency over a decision that affects their livelihood. In this paper, we propose to evaluate a linear classification model in terms of recourse, which we define as the ability of a person to change the decision of the model through actionable input variables (e.g., income vs. age or marital status). We present an integer programming toolkit to: (i) measure the feasibility and difficulty of recourse in a target population; and (ii) generate a list of actionable changes for a person to obtain a desired outcome. We discuss how our tools can inform different stakeholders by using them to audit recourse for credit scoring models built with real-world datasets. Our results illustrate how recourse can be significantly affected by common modeling practices, and motivate the need to evaluate recourse in algorithmic decision-making."
                },
                {
                    "title": "Modernizing the Bradford Hill criteria for assessing causal relationships in observational data",
                    "abstract": "Abstract Perhaps no other topic in risk analysis is more difficult, more controversial, or more important to risk management policy analysts and decision-makers than how to draw valid, correctly qualified causal conclusions from observational data. Statistical methods can readily quantify associations between observed variables using measures such as relative risk (RR) ratios, odds ratios (OR), slope coefficients for exposure or treatment variables in regression models, and quantities derived from these measures. Textbooks of epidemiology explain how to calculate population attributable fractions, attributable risks, burden-of-disease estimates, and probabilities of causation from relative risk (RR) ratios. Despite their suggestive names, these association-based measures have no necessary connection to causation if the associations on which they are based arise from bias, confounding, p-hacking, coincident historical trends, or other noncausal sources. But policy analysts and decision makers need something more: trustworthy predictions \u2013 and, later, evaluations \u2013 of the changes in outcomes caused by changes in policy variables. This concept of manipulative causation differs from the more familiar concepts of associational and attributive causation most widely used in epidemiology. Drawing on modern literature on causal discovery and inference principles and algorithms for drawing limited but useful causal conclusions from observational data, we propose seven criteria for assessing consistency of data with a manipulative causal exposure\u2013response relationship \u2013 mutual information, directed dependence, internal and external consistency, coherent causal explanation of biological plausibility, causal mediation confirmation, and refutation of non-causal explanations \u2013 and discuss to what extent it is now possible to automate discovery of manipulative causal dependencies and quantification of causal effects from observational data. We compare our proposed principles for causal discovery and inference to the traditional Bradford Hill considerations from 1965. Understanding how old and new principles are related can clarify and enrich both."
                },
                {
                    "title": "The Book of Why: The New Science of Cause and Effect",
                    "abstract": "This week on the Science podcast, Judea Pearl and Dana Mackenzie discuss strategies for causal thinking and describe its implications for artificial intelligence."
                },
                {
                    "title": "Causal Reasoning for Algorithmic Fairness",
                    "abstract": "In this work, we argue for the importance of causal reasoning in creating fair algorithms for decision making. We give a review of existing approaches to fairness, describe work in causality necessary for the understanding of causal approaches, argue why causality is necessary for any approach that wishes to be fair, and give a detailed analysis of the many recent approaches to causality-based fairness."
                },
                {
                    "title": "FASK with Interventional Knowledge Recovers Edges from the Sachs Model",
                    "abstract": "We report a procedure that, in one step from continuous data with minimal preparation, recovers the graph found by Sachs et al. \\cite{sachs2005causal}, with only a few edges different. The algorithm, Fast Adjacency Skewness (FASK), relies on a mixture of linear reasoning and reasoning from the skewness of variables; the Sachs data is a good candidate for this procedure since the skewness of the variables is quite pronounced. We review the ground truth model from Sachs et al. as well as some of the fluctuations seen in the protein abundances in the system, give the Sachs model and the FASK model, and perform a detailed comparison. Some variation in hyper-parameters is explored, though the main result uses values at or near the defaults learned from work modeling fMRI data."
                },
                {
                    "title": "Fairness in Decision-Making - The Causal Explanation Formula",
                    "abstract": "\n \n AI plays an increasingly prominent role in society since decisions that were once made by humans are now delegated to automated systems. These systems are currently in charge of deciding bank loans, criminals' incarceration, and the hiring of new employees, and it's not difficult to envision that they will in the future underpin most of the decisions in society. Despite the high complexity entailed by this task, there is still not much understanding of basic properties of such systems. For instance, we currently cannot detect (neither explain nor correct) whether an AI system can be deemed fair (i.e., is abiding by the decision-constraints agreed by society) or it is reinforcing biases and perpetuating a preceding prejudicial practice. Issues of discrimination have been discussed extensively in political and legal circles, but there exists still not much understanding of the formal conditions that a system must meet to be deemed fair. In this paper, we use the language of structural causality (Pearl, 2000) to fill in this gap. We start by introducing three new fine-grained measures of transmission of change from stimulus to effect, which we called counterfactual direct (Ctf-DE), indirect (Ctf-IE), and spurious (Ctf-SE) effects. We then derive what we call the causal explanation formula, which allows the AI designer to quantitatively evaluate fairness and explain the total observed disparity of decisions through different discriminatory mechanisms. We apply these measures to various discrimination analysis tasks and run extensive simulations, including detection, evaluation, and optimization of decision-making under fairness constraints. We conclude studying the trade-off between different types of fairness criteria (outcome and procedural), and provide a quantitative approach to policy implementation and the design of fair AI systems.\n \n"
                },
                {
                    "title": "Path-Specific Counterfactual Fairness",
                    "abstract": "We consider the problem of learning fair decision systems from data in which a sensitive attribute might affect the decision along both fair and unfair pathways. We introduce a counterfactual approach to disregard effects along unfair pathways that does not incur in the same loss of individual-specific information as previous approaches. Our method corrects observations adversely affected by the sensitive attribute, and uses these to form a decision. We leverage recent developments in deep learning and approximate inference to develop a VAE-type method that is widely applicable to complex nonlinear models."
                },
                {
                    "title": "A Survey of Methods for Explaining Black Box Models",
                    "abstract": "In recent years, many accurate decision support systems have been constructed as black boxes, that is as systems that hide their internal logic to the user. This lack of explanation constitutes both a practical and an ethical issue. The literature reports many approaches aimed at overcoming this crucial weakness, sometimes at the cost of sacrificing accuracy for interpretability. The applications in which black box decision systems can be used are various, and each approach is typically developed to provide a solution for a specific problem and, as a consequence, it explicitly or implicitly delineates its own definition of interpretability and explanation. The aim of this article is to provide a classification of the main problems addressed in the literature with respect to the notion of explanation and the type of black box system. Given a problem definition, a black box type, and a desired explanation, this survey should help the researcher to find the proposals more useful for his own work. The proposed classification of approaches to open black box models should also be useful for putting the many research open questions in perspective."
                },
                {
                    "title": "An Exact and Robust Conformal Inference Method for Counterfactual and Synthetic Controls",
                    "abstract": "Abstract We introduce new inference procedures for counterfactual and synthetic control methods for policy evaluation. We recast the causal inference problem as a counterfactual prediction and a structural breaks testing problem. This allows us to exploit insights from conformal prediction and structural breaks testing to develop permutation inference procedures that accommodate modern high-dimensional estimators, are valid under weak and easy-to-verify conditions, and are provably robust against misspecification. Our methods work in conjunction with many different approaches for predicting counterfactual mean outcomes in the absence of the policy intervention. Examples include synthetic controls, difference-in-differences, factor and matrix completion models, and (fused) time series panel data models. Our approach demonstrates an excellent small-sample performance in simulations and is taken to a data application where we re-evaluate the consequences of decriminalizing indoor prostitution. Open-source software for implementing our conformal inference methods is available."
                },
                {
                    "title": "When Worlds Collide: Integrating Different Counterfactual Assumptions in Fairness",
                    "abstract": "Machine learning is now being used to make crucial decisions about people's lives. For nearly all of these decisions there is a risk that individuals of a certain race, gender, sexual orientation, or any other subpopulation are unfairly discriminated against. Our recent method has demonstrated how to use techniques from counterfactual inference to make predictions fair across different subpopulations. This method requires that one provides the causal model that generated the data at hand. In general, validating all causal implications of the model is not possible without further assumptions. Hence, it is desirable to integrate competing causal models to provide counterfactually fair decisions, regardless of which causal \"world\" is the correct one. In this paper, we show how it is possible to make predictions that are approximately fair with respect to multiple possible causal models at once, thus mitigating the problem of exact causal specification. We frame the goal of learning a fair classifier as an optimization problem with fairness constraints entailed by competing causal explanations. We show how this optimization problem can be efficiently solved using gradient-based methods. We demonstrate the flexibility of our model on two real-world fair classification problems. We show that our model can seamlessly balance fairness in multiple worlds with prediction accuracy."
                },
                {
                    "title": "Estimation of interventional effects of features on prediction",
                    "abstract": "The interpretability of prediction mechanisms with respect to the underlying prediction problem is often unclear. While several studies have focused on developing prediction models with meaningful parameters, the causal relationships between the predictors and the actual prediction have not been considered. Here, we connect the underlying causal structure of a data generation process and the causal structure of a prediction mechanism. To achieve this, we propose a framework that identifies the feature with the greatest causal influence on the prediction and estimates the necessary causal intervention of a feature such that a desired prediction is obtained. The general concept of the framework has no restrictions regarding data linearity; however, we focus on an implementation for linear data here. The framework applicability is evaluated using artificial data and demonstrated using real-world data."
                },
                {
                    "title": "Avoiding Discrimination through Causal Reasoning",
                    "abstract": "Recent work on fairness in machine learning has focused on various statistical discrimination criteria and how they trade off. Most of these criteria are observational: They depend only on the joint distribution of predictor, protected attribute, features, and outcome. While convenient to work with, observational criteria have severe inherent limitations that prevent them from resolving matters of fairness conclusively. \nGoing beyond observational criteria, we frame the problem of discrimination based on protected attributes in the language of causal reasoning. This viewpoint shifts attention from \"What is the right fairness criterion?\" to \"What do we want to assume about the causal data generating process?\" Through the lens of causality, we make several contributions. First, we crisply articulate why and when observational criteria fail, thus formalizing what was before a matter of opinion. Second, our approach exposes previously ignored subtleties and why they are fundamental to the problem. Finally, we put forward natural causal non-discrimination criteria and develop algorithms that satisfy them."
                },
                {
                    "title": "Fair Inference on Outcomes",
                    "abstract": "In this paper, we consider the problem of fair statistical inference involving outcome variables. Examples include classification and regression problems, and estimating treatment effects in randomized trials or observational data. The issue of fairness arises in such problems where some covariates or treatments are \"sensitive,\" in the sense of having potential of creating discrimination. In this paper, we argue that the presence of discrimination can be formalized in a sensible way as the presence of an effect of a sensitive covariate on the outcome along certain causal pathways, a view which generalizes (Pearl 2009). A fair outcome model can then be learned by solving a constrained optimization problem. We discuss a number of complications that arise in classical statistical inference due to this view and provide workarounds based on recent work in causal and semi-parametric inference."
                },
                {
                    "title": "Counterfactual Fairness",
                    "abstract": "Machine learning can impact people with legal or ethical consequences when it is used to automate decisions in areas such as insurance, lending, hiring, and predictive policing. In many of these scenarios, previous decisions have been made that are unfairly biased against certain subpopulations, for example those of a particular race, gender, or sexual orientation. Since this past data may be biased, machine learning predictors must account for this to avoid perpetuating or creating discriminatory practices. In this paper, we develop a framework for modeling fairness using tools from causal inference. Our definition of counterfactual fairness captures the intuition that a decision is fair towards an individual if it the same in (a) the actual world and (b) a counterfactual world where the individual belonged to a different demographic group. We demonstrate our framework on a real-world problem of fair prediction of success in law school."
                },
                {
                    "title": "Peeking Inside the Black Box: Visualizing Statistical Learning With Plots of Individual Conditional Expectation",
                    "abstract": "This article presents individual conditional expectation (ICE) plots, a tool for visualizing the model estimated by any supervised learning algorithm. Classical partial dependence plots (PDPs) help visualize the average partial relationship between the predicted response and one or more features. In the presence of substantial interaction effects, the partial response relationship can be heterogeneous. Thus, an average curve, such as the PDP, can obfuscate the complexity of the modeled relationship. Accordingly, ICE plots refine the PDP by graphing the functional relationship between the predicted response and the feature for individual observations. Specifically, ICE plots highlight the variation in the fitted values across the range of a covariate, suggesting where and to what extent heterogeneities might exist. In addition to providing a plotting suite for exploratory analysis, we include a visual test for additive structure in the data-generating model. Through simulated examples and real datasets, we demonstrate how ICE plots can shed light on estimated models in ways PDPs cannot. Procedures outlined are available in the R package ICEbox."
                },
                {
                    "title": "On causal and anticausal learning",
                    "abstract": "We consider the problem of function estimation in the case where an underlying causal model can be inferred. This has implications for popular scenarios such as covariate shift, concept drift, transfer learning and semi-supervised learning. We argue that causal knowledge may facilitate some approaches for a given problem, and rule out others. In particular, we formulate a hypothesis for when semi-supervised learning can help, and corroborate it with empirical results."
                },
                {
                    "title": "Fast stable direct fitting and smoothness selection for generalized additive models",
                    "abstract": "Summary.\u2002 Existing computationally efficient methods for penalized likelihood generalized additive model fitting employ iterative smoothness selection on working linear models (or working mixed models). Such schemes fail to converge for a non\u2010negligible proportion of models, with failure being particularly frequent in the presence of concurvity. If smoothness selection is performed by optimizing \u2018whole model\u2019 criteria these problems disappear, but until now attempts to do this have employed finite\u2010difference\u2010based optimization schemes which are computationally inefficient and can suffer from false convergence. The paper develops the first computationally efficient method for direct generalized additive model smoothness selection. It is highly stable, but by careful structuring achieves a computational efficiency that leads, in simulations, to lower mean computation times than the schemes that are based on working model smoothness selection. The method also offers a reliable way of fitting generalized additive mixed models."
                },
                {
                    "title": "Causal Protein-Signaling Networks Derived from Multiparameter Single-Cell Data",
                    "abstract": "Machine learning was applied for the automated derivation of causal influences in cellular signaling networks. This derivation relied on the simultaneous measurement of multiple phosphorylated protein and phospholipid components in thousands of individual primary human immune system cells. Perturbing these cells with molecular interventions drove the ordering of connections between pathway components, wherein Bayesian network computational methods automatically elucidated most of the traditionally reported signaling relationships and predicted novel interpathway network causalities, which we verified experimentally. Reconstruction of network models from physiologically relevant primary single cells might be applied to understanding native-state tissue signaling biology, complex drug actions, and dysfunctional signaling in diseased cells."
                },
                {
                    "title": "Causation, Prediction, and Search",
                    "abstract": "The writing is not uniformly polished and is scattered with long, awkward sentences that require some effort to unravel. I wonder if this is the result of infelicitous translation from the original German version (Wellek 1994). There are also numerous small typographical errors. More careful editing could have solved these problems before publication. There are no exercises, and so I would hesitate to use the book as a text (although it should be noted that this is not one of the author\u2019s stated aims). Although Testing Statistical Hypotheses of Equivalence has some weaknesses, it is a useful reference for those interested in the question of equivalence testing, particularly in biological applications."
                },
                {
                    "title": "Greedy function approximation: A gradient boosting machine.",
                    "abstract": "Function estimation/approximation is viewed from the perspective of numerical optimization in function space, rather than parameter space. A connection is made between stagewise additive expansions and steepest-descent minimization. A general gradient descent boosting paradigm is developed for additive expansions based on any fitting criterion. Specific algorithms are presented for least-squares, least absolute deviation, and Huber-M loss functions for regression, and multiclass logistic likelihood for classification. Special enhancements are derived for the particular case where the individual additive components are regression trees, and tools for interpreting such TreeBoost models are presented. Gradient boosting of regression trees produces competitive, highly robust, interpretable procedures for both regression and classification, especially appropriate for mining less than clean data. Connections between this approach and the boosting methods of Freund and Shapire and Friedman, Hastie and Tibshirani are discussed."
                },
                {
                    "title": "Nuclear feature extraction for breast tumor diagnosis",
                    "abstract": "Interactive image processing techniques, along with a linear-programming-based inductive classifier, have been used to create a highly accurate system for diagnosis of breast tumors. A small fraction of a fine needle aspirate slide is selected and digitized. With an interactive interface, the user initializes active contour models, known as snakes, near the boundaries of a set of cell nuclei. The customized snakes are deformed to the exact shape of the nuclei. This allows for precise, automated analysis of nuclear size, shape and texture. Ten such features are computed for each nucleus, and the mean value, largest (or 'worst') value and standard error of each feature are found over the range of isolated cells. After 569 images were analyzed in this fashion, different combinations of features were tested to find those which best separate benign from malignant samples. Ten-fold cross-validation accuracy of 97% was achieved using a single separating plane on three of the thirty features: mean texture, worst area and worst smoothness. This represents an improvement over the best diagnostic results in the medical literature. The system is currently in use at the University of Wisconsin Hospitals. The same feature set has also been utilized in the much more difficult task of predicting distant recurrence of malignancy in patients, resulting in an accuracy of 86%."
                },
                {
                    "title": "Causal Intersectionality and Fair Ranking",
                    "abstract": "In this paper we propose a causal modeling approach to intersectional fairness, and a flexible, task-specific method for computing intersectionally fair rankings. Rankings are used in many contexts, ranging from Web search to college admissions"
                },
                {
                    "title": "A Survey of Contrastive and Counterfactual Explanation Generation Methods for Explainable Artificial Intelligence",
                    "abstract": "A number of algorithms in the field of artificial intelligence offer poorly interpretable decisions. To disclose the reasoning behind such algorithms, their output can be explained by means of so-called evidence-based (or factual) explanations. Alternatively, contrastive and counterfactual explanations justify why the output of the algorithms is not any different and how it could be changed, respectively. It is of crucial importance to bridge the gap between theoretical approaches to contrastive and counterfactual explanation and the corresponding computational frameworks. In this work we conduct a systematic literature review which provides readers with a thorough and reproducible analysis of the interdisciplinary research field under study. We first examine theoretical foundations of contrastive and counterfactual accounts of explanation. Then, we report the state-of-the-art computational frameworks for contrastive and counterfactual explanation generation. In addition, we analyze how grounded such frameworks are on the insights from the inspected theoretical approaches. As a result, we highlight a variety of properties of the approaches under study and reveal a number of shortcomings thereof. Moreover, we define a taxonomy regarding both theoretical and practical approaches to contrastive and counterfactual explanation."
                },
                {
                    "title": "Causal Discovery Toolbox: Uncovering causal relationships in Python",
                    "abstract": "This paper presents a new open source Python framework for causal discovery from observational data and domain background knowledge, aimed at causal graph and causal mechanism modeling. The Cdt package implements an end-to-end approach, recovering the direct dependencies (the skeleton of the causal graph) and the causal relationships between variables. It includes algorithms from the \u2018 Bnlearn \u2019 (Scutari, 2018) and \u2018 Pcalg \u2019 (Kalisch et al., 2018) packages, together with algorithms for pairwise causal discovery such as ANM (Hoyer et al., 2009). Cdt is available under the MIT License at https://github.com/FenTechSolutions/CausalDiscoveryToolbox ."
                },
                {
                    "title": "Comment: Graphical Models, Causality and Intervention",
                    "abstract": "I am grateful for the opportunity to respond to these two excellent papers. Although graphical models are intuitively compelling for conceptualizing statistical associations, the scientific community generally views such models with hesitancy and suspicion. The two papers before us demonstrate the use of graphs specifically, directed acyclic graphs (DAGs) -as a mathematical tool of great versatility and thus promise to make graphical languages more common in statistical analysis. In fact, I find my own views in such close agreement with those of the authors that any attempt on my part to comment directly on their work would amount to sheer repetition. Instead, as the editor suggested, I would like to provide a personal perspective on current and future developments in the areas of graphical and causal modeling. A complementary account of the evolution of belief networks is given in Pearl (1993a). I will focus on the connection between graphical models and the notion of causality in statistical analysis. This connection has been treated very cautiously in the papers before us. In Lauritzen and Spiegelhalter (1988), the graphs were called \"causal networks,\" for which the authors were criticized; they have agreed to refrain from using the word \"causal.\" In the current paper, Spiegelhalter et al. deemphasize the causal interpretation of the arcs in favor of the \"irrelevance\" interpretation. I think this retreat is regrettable for two reasons: first, causal associations are the primary source of judgments about irrelevance, and, second, rejecting the causal interpretation of arcs prevents us from using graphical models for making legitimate predictions about the effect of actions. Such predictions are indispensable in applications such as treatment management and policy analysis. I would like to supplement the discussion with an ac'count of how causal models and graphical models are related. It is generally accepted that, because they provide information about the dynamics of the system under study, causal models, regardless of how they are discovered or tested, are more useful than associational models. In other words, whereas the joint distribution"
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively visualize and interpret causal relationships between predictor variables and an outcome variable in machine learning models, particularly in the context of model-agnostic or \"black-box\" explanations?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of existing interpretation methods that often overlook causal dependencies between predictors. By developing Causal Dependence Plots (CDPs), we can provide more accurate and relevant explanations of model behavior, which can lead to better decision-making in algorithmic systems, improved resource allocation, and more reliable scientific hypotheses. This advancement in understanding causal relationships can significantly influence future research in interpretability, transparency, and trust in AI systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of accurately capturing and visualizing the interdependencies among predictor variables. Naive approaches, such as using Partial Dependence Plots (PDPs) or Individual Conditional Expectation (ICE) plots, fail because they treat predictors as independent, which can lead to misleading interpretations. The technical obstacles include the need for a robust Explanatory Causal Model (ECM) that can account for these dependencies, as well as the difficulty in specifying or learning such models from data. Theoretical challenges also arise in ensuring that the causal relationships are correctly represented and interpreted.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on model-agnostic interpretation methods that do not adequately consider the causal relationships between predictors. Limitations in existing solutions stem from a lack of integration of causal modeling into the interpretation process and the assumption of independence among predictors. Barriers include insufficient methodologies for capturing complex interactions and the absence of frameworks that allow for the incorporation of prior domain knowledge. Our approach differs by explicitly using an ECM to inform the interpretation process, thereby respecting the causal dependencies among predictors.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of Causal Dependence Plots (CDPs) that utilize an Explanatory Causal Model (ECM) to visualize the relationships between predictor variables and an outcome variable. We will apply this methodology to a relevant dataset, ensuring that the ECM is informed by domain knowledge and potentially learned from the data. The key metrics for evaluation will include the accuracy and relevance of the explanations provided by the CD"
            }
        },
        "author_data": {
            "5f213e90-c959-4d6e-84b0-1ccfad138694": {
                "pk": "5f213e90-c959-4d6e-84b0-1ccfad138694",
                "name": "Joshua R. Loftus",
                "collaborators": [
                    "Julia Stoyanovich",
                    "Jonathan E. Taylor",
                    "Chris Russell",
                    "Matt J. Kusner",
                    "Ricardo Silva",
                    "Lucius E. J. Bynum",
                    "Ke Yang",
                    "Xiaoying Tian"
                ],
                "domain": [
                    "Causal Inference",
                    "Fairness in Machine Learning",
                    "Statistical Modeling",
                    "Selective Inference"
                ],
                "publications": [
                    {
                        "title": "Selective inference after cross-validation",
                        "abstract": "This paper describes a method for performing inference on models chosen by cross-validation. When the test error being minimized in cross-validation is a residual sum of squares it can be written as a quadratic form. This allows us to apply the inference framework in Loftus et al. (2015) for models determined by quadratic constraints to the model that minimizes CV test error. Our only requirement on the model training pro- cedure is that its selection events are regions satisfying linear or quadratic constraints. This includes both Lasso and forward stepwise, which serve as our main examples throughout. We do not require knowledge of the error variance $\\sigma^2$. The procedures described here are computationally intensive methods of selecting models adaptively and performing inference for the selected model. Implementations are available in an R package."
                    },
                    {
                        "title": "Selective inference in regression models with groups of variables",
                        "abstract": "We provide a general mathematical framework for selective inference with supervised model selection procedures characterized by quadratic forms in the outcome variable. Forward stepwise with groups of variables is an important special case as it allows models with categorical variables or factors. Models can be chosen by AIC, BIC, or a fixed number of steps. We provide an exact significance test for each group of variables in the selected model based on an appropriately truncated $\\chi$ or $F$ distribution for the cases of known and unknown $\\sigma^2$ respectively. An efficient software implementation is available as a package in the R statistical programming language."
                    },
                    {
                        "title": "A significance test for forward stepwise model selection",
                        "abstract": "We apply the methods developed by Lockhart et al. (2013) and Taylor et al. (2013) on significance tests for penalized regression to forward stepwise model selection. A general framework for selection procedures described by quadratic inequalities includes a variant of forward stepwise with grouped variables, allowing us to handle categorical variables and factor models. We provide an algorithm to compute a new statistic with an exact null distribution conditional on the outcome of the model selection procedure. This new statistic, which we denote $T\\chi$, has a truncated $\\chi$ distribution under the global null. We apply this test in forward stepwise iteratively on the residual after each step. The resulting method has the computational strengths of stepwise selection and addresses the problem of invalid test statistics due to model selection. We illustrate the flexibility of this method by applying it to several specialized applications of forward stepwise including a hierarchical interactions model and a recently described additive model that adaptively chooses between linear and nonlinear effects for each variable."
                    },
                    {
                        "title": "Causal intersectionality for fair ranking",
                        "abstract": "In this paper we propose a causal modeling approach to intersectional fairness, and a flexible, task-specific method for computing intersectionally fair rankings. Rankings are used in many contexts, ranging from Web search results to college admissions, but causal inference for fair rankings has received limited attention. Additionally, the growing literature on causal fairness has directed little attention to intersectionality. By bringing these issues together in a formal causal framework we make the application of intersectionality in fair machine learning explicit, connected to important real world effects and domain knowledge, and transparent about technical limitations. We experimentally evaluate our approach on real and synthetic datasets, exploring its behaviour under different structural assumptions."
                    },
                    {
                        "title": "Causal Reasoning for Algorithmic Fairness",
                        "abstract": "In this work, we argue for the importance of causal reasoning in creating fair algorithms for decision making. We give a review of existing approaches to fairness, describe work in causality necessary for the understanding of causal approaches, argue why causality is necessary for any approach that wishes to be fair, and give a detailed analysis of the many recent approaches to causality-based fairness."
                    },
                    {
                        "title": "Selective inference with unknown variance via the square-root LASSO",
                        "abstract": "There has been much recent work on inference after model selection when the noise level is known, however, $\\sigma$ is rarely known in practice and its estimation is difficult in high-dimensional settings. In this work we propose using the square-root LASSO (also known as the scaled LASSO) to perform selective inference for the coefficients and the noise level simultaneously. The square-root LASSO has the property that choosing a reasonable tuning parameter is scale-free, namely it does not depend on the noise level in the data. We provide valid p-values and confidence intervals for the coefficients after selection, and estimates for model specific variance. Our estimates perform better than other estimates of $\\sigma^2$ in simulation."
                    },
                    {
                        "title": "Counterfactuals for the Future",
                        "abstract": "Counterfactuals are often described as 'retrospective,' focusing on hypothetical alternatives to a realized past. This description relates to an often implicit assumption about the structure and stability of exogenous variables in the system being modeled -- an assumption that is reasonable in many settings where counterfactuals are used. In this work, we consider cases where we might reasonably make a different assumption about exogenous variables, namely, that the exogenous noise terms of each unit do exhibit some unit-specific structure and/or stability. This leads us to a different use of counterfactuals -- a 'forward-looking' rather than 'retrospective' counterfactual. We introduce \"counterfactual treatment choice,\" a type of treatment choice problem that motivates using forward-looking counterfactuals. We then explore how mismatches between interventional versus forward-looking counterfactual approaches to treatment choice, consistent with different assumptions about exogenous noise, can lead to counterintuitive results."
                    },
                    {
                        "title": "A New Paradigm for Counterfactual Reasoning in Fairness and Recourse",
                        "abstract": "Counterfactuals and counterfactual reasoning underpin numerous techniques for auditing and understanding artificial intelligence (AI) systems. The traditional paradigm for counterfactual reasoning in this literature is the interventional counterfactual, where hypothetical interventions are imagined and simulated. For this reason, the starting point for causal reasoning about legal protections and demographic data in AI is an imagined intervention on a legally-protected characteristic, such as ethnicity, race, gender, disability, age, etc. We ask, for example, what would have happened had your race been different? An inherent limitation of this paradigm is that some demographic interventions -- like interventions on race -- may not translate into the formalisms of interventional counterfactuals. In this work, we explore a new paradigm based instead on the backtracking counterfactual, where rather than imagine hypothetical interventions on legally-protected characteristics, we imagine alternate initial conditions while holding these characteristics fixed. We ask instead, what would explain a counterfactual outcome for you as you actually are or could be? This alternate framework allows us to address many of the same social concerns, but to do so while asking fundamentally different questions that do not rely on demographic interventions."
                    },
                    {
                        "title": "Disaggregated Interventions to Reduce Inequality",
                        "abstract": "A significant body of research in the data sciences considers unfair discrimination against social categories such as race or gender that could occur or be amplified as a result of algorithmic decisions. Simultaneously, real-world disparities continue to exist, even before algorithmic decisions are made. In this work, we draw on insights from the social sciences brought into the realm of causal modeling and constrained optimization, and develop a novel algorithmic framework for tackling pre-existing real-world disparities. The purpose of our framework, which we call the \"impact remediation framework,\" is to measure real-world disparities and discover the optimal intervention policies that could help improve equity or access to opportunity for those who are underserved with respect to an outcome of interest. We develop a disaggregated approach to tackling pre-existing disparities that relaxes the typical set of assumptions required for the use of social categories in structural causal models. Our approach flexibly incorporates counterfactuals and is compatible with various ontological assumptions about the nature of social categories. We demonstrate impact remediation with a hypothetical case study and compare our disaggregated approach to an existing state-of-the-art approach, comparing its structure and resulting policy recommendations. In contrast to most work on optimal policy learning, we explore disparity reduction itself as an objective, explicitly focusing the power of algorithms on reducing inequality."
                    },
                    {
                        "title": "Causal Interventions for Fairness",
                        "abstract": "Most approaches in algorithmic fairness constrain machine learning methods so the resulting predictions satisfy one of several intuitive notions of fairness. While this may help private companies comply with non-discrimination laws or avoid negative publicity, we believe it is often too little, too late. By the time the training data is collected, individuals in disadvantaged groups have already suffered from discrimination and lost opportunities due to factors out of their control. In the present work we focus instead on interventions such as a new public policy, and in particular, how to maximize their positive effects while improving the fairness of the overall system. We use causal methods to model the effects of interventions, allowing for potential interference--each individual's outcome may depend on who else receives the intervention. We demonstrate this with an example of allocating a budget of teaching resources using a dataset of schools in New York City."
                    },
                    {
                        "title": "Counterfactual Fairness",
                        "abstract": "Machine learning can impact people with legal or ethical consequences when it is used to automate decisions in areas such as insurance, lending, hiring, and predictive policing. In many of these scenarios, previous decisions have been made that are unfairly biased against certain subpopulations, for example those of a particular race, gender, or sexual orientation. Since this past data may be biased, machine learning predictors must account for this to avoid perpetuating or creating discriminatory practices. In this paper, we develop a framework for modeling fairness using tools from causal inference. Our definition of counterfactual fairness captures the intuition that a decision is fair towards an individual if it is the same in (a) the actual world and (b) a counterfactual world where the individual belonged to a different demographic group. We demonstrate our framework on a real-world problem of fair prediction of success in law school."
                    }
                ]
            },
            "5a3a3b80-4f9c-416b-b364-b5cb4f99cfb5": {
                "pk": "5a3a3b80-4f9c-416b-b364-b5cb4f99cfb5",
                "name": "Lucius E. J. Bynum",
                "collaborators": [
                    "Joshua R. Loftus",
                    "Julia Stoyanovich"
                ],
                "domain": [
                    "Causal Inference",
                    "Counterfactual Reasoning",
                    "Fairness in AI",
                    "Algorithmic Equity"
                ],
                "publications": [
                    {
                        "title": "Counterfactuals for the Future",
                        "abstract": "Counterfactuals are often described as 'retrospective,' focusing on hypothetical alternatives to a realized past. This description relates to an often implicit assumption about the structure and stability of exogenous variables in the system being modeled -- an assumption that is reasonable in many settings where counterfactuals are used. In this work, we consider cases where we might reasonably make a different assumption about exogenous variables, namely, that the exogenous noise terms of each unit do exhibit some unit-specific structure and/or stability. This leads us to a different use of counterfactuals -- a 'forward-looking' rather than 'retrospective' counterfactual. We introduce \"counterfactual treatment choice,\" a type of treatment choice problem that motivates using forward-looking counterfactuals. We then explore how mismatches between interventional versus forward-looking counterfactual approaches to treatment choice, consistent with different assumptions about exogenous noise, can lead to counterintuitive results."
                    },
                    {
                        "title": "A New Paradigm for Counterfactual Reasoning in Fairness and Recourse",
                        "abstract": "Counterfactuals and counterfactual reasoning underpin numerous techniques for auditing and understanding artificial intelligence (AI) systems. The traditional paradigm for counterfactual reasoning in this literature is the interventional counterfactual, where hypothetical interventions are imagined and simulated. For this reason, the starting point for causal reasoning about legal protections and demographic data in AI is an imagined intervention on a legally-protected characteristic, such as ethnicity, race, gender, disability, age, etc. We ask, for example, what would have happened had your race been different? An inherent limitation of this paradigm is that some demographic interventions -- like interventions on race -- may not translate into the formalisms of interventional counterfactuals. In this work, we explore a new paradigm based instead on the backtracking counterfactual, where rather than imagine hypothetical interventions on legally-protected characteristics, we imagine alternate initial conditions while holding these characteristics fixed. We ask instead, what would explain a counterfactual outcome for you as you actually are or could be? This alternate framework allows us to address many of the same social concerns, but to do so while asking fundamentally different questions that do not rely on demographic interventions."
                    },
                    {
                        "title": "Disaggregated Interventions to Reduce Inequality",
                        "abstract": "A significant body of research in the data sciences considers unfair discrimination against social categories such as race or gender that could occur or be amplified as a result of algorithmic decisions. Simultaneously, real-world disparities continue to exist, even before algorithmic decisions are made. In this work, we draw on insights from the social sciences brought into the realm of causal modeling and constrained optimization, and develop a novel algorithmic framework for tackling pre-existing real-world disparities. The purpose of our framework, which we call the \"impact remediation framework,\" is to measure real-world disparities and discover the optimal intervention policies that could help improve equity or access to opportunity for those who are underserved with respect to an outcome of interest. We develop a disaggregated approach to tackling pre-existing disparities that relaxes the typical set of assumptions required for the use of social categories in structural causal models. Our approach flexibly incorporates counterfactuals and is compatible with various ontological assumptions about the nature of social categories. We demonstrate impact remediation with a hypothetical case study and compare our disaggregated approach to an existing state-of-the-art approach, comparing its structure and resulting policy recommendations. In contrast to most work on optimal policy learning, we explore disparity reduction itself as an objective, explicitly focusing the power of algorithms on reducing inequality."
                    }
                ]
            },
            "044964e0-6436-4d67-b1e7-83d8a5b622f3": {
                "pk": "044964e0-6436-4d67-b1e7-83d8a5b622f3",
                "name": "Sakina Hansen",
                "collaborators": [
                    "Christopher J. Conselice",
                    "Amelia Fraser-McKelvie",
                    "Leonardo Ferreira"
                ],
                "domain": [
                    "Machine Learning",
                    "Astronomy",
                    "Deep Learning",
                    "Galaxy Kinematics"
                ],
                "publications": [
                    {
                        "title": "Retrieving Internal Kinematics of Galaxies with Deep Learning using Single-Band Optical Images",
                        "abstract": "Using deep machine learning we show that the internal velocities of galaxies can be retrieved from optical images trained using 4596 systems observed with the SDSS-MaNGA survey. Using only $i$-band images we show that the velocity dispersions and the rotational velocities of galaxies can be measured to an accuracy of 29 km~$\\rm{s}^{-1}$ and 69 km~$\\rm{s}^{-1}$ respectively, close to the resolution limit of the spectroscopic data. This shows that galaxy structures in the optical holds important information concerning the internal properties of galaxies and that the internal kinematics of galaxies are quantitatively reflected in their stellar light distributions beyond a simple rotational vs. dispersion distinction."
                    }
                ]
            }
        }
    },
    "2402.10877": {
        "paper_data": {
            "title": "Robust agents learn causal world models",
            "url": "http://arxiv.org/abs/2402.10877v7",
            "arxiv_id": "2402.10877",
            "authors": [
                "Jonathan Richens",
                "Tom Everitt"
            ],
            "abstract": "It has long been hypothesised that causal reasoning plays a fundamental role in robust and general intelligence. However, it is not known if agents must learn causal models in order to generalise to new domains, or if other inductive biases are sufficient. We answer this question, showing that any agent capable of satisfying a regret bound under a large set of distributional shifts must have learned an approximate causal model of the data generating process, which converges to the true causal model for optimal agents. We discuss the implications of this result for several research areas including transfer learning and causal inference.",
            "introduction": "   1 Introduction  What capabilities are necessary for general intelligence (Legg & Hutter, 2007)? One candidate is causal reasoning, which plays a foundational role in human cognition (Gopnik et\u00a0al., 2007; Sloman & Lagnado, 2015). It has even been argued that human-level AI is impossible without causal reasoning (Pearl, 2018). However, recent years have seen the development of agents that do not explicitly learn or reason on causal models, but nonetheless are capable of adapting to a wide range of environments and tasks (Reed et\u00a0al., 2022; Team et\u00a0al., 2023; Brown et\u00a0al., 2020).   This raises the question, do agents have to learn causal models in order to adapt to new domains, or are other inductive biases sufficient? To answer this question, we have to be careful not to assume that agents use causal assumptions a priori. For example, transportability theory determines what causal knowledge is necessary for transfer learning when all assumptions on the data generating process (inductive biases) can be expressed as constraints on causal structure (Bareinboim & Pearl, 2016). However, deep learning algorithms can exploit a much larger set of inductive biases (Neyshabur et\u00a0al., 2014; Battaglia et\u00a0al., 2018; Rahaman et\u00a0al., 2019; Goyal & Bengio, 2022) which in many real-world settings may be sufficient to identify low regret policies without requiring causal knowledge.   The main result of this paper is to answer this question by showing that,  Any agent capable of adapting to a sufficiently large set of distributional shifts must have learned a causal model of the data generating process.    Here, adapting to a distributional shift means learning a policy that satisfies a regret bound following an intervention on the data generating process\u2014for example, changing the distribution of features or latent variables. It is known that a causal model of the data generating process can be used to identify regret-bounded policies following a distributional shift (sufficiency), with more accurate models allowing lower regret policies to be found. We prove the converse (necessity)\u2014given regret-bounded policies for a large set of distributional shifts, we can learn an approximate causal model of the data generating process, with the approximation becoming exact for optimal policies. Hence, learning a causal model of the data generating process is necessary for robust adaptation.   This has consequences for a number of fields and questions. For one, it implies that causal identification laws also constrain domain adaptation. For example, we show that adapting to covariate and label shifts is only possible if the causal relations between features and labels can be identified from the training data\u2014a non-trivial causal discovery problem. This provides further theoretical justification for causal representation learning (Sch\u00f6lkopf et\u00a0al., 2021), showing that learning causal representations is necessary for achieving strong robustness guarantees. Our result also implies that we can learn causal models from adaptive agents. We demonstrate this by solving a causal discovery task on synthetic data by observing the policy of a regret-bounded agent under distributional shifts. More speculatively, our results suggest that causal models could play a role in emergent capabilities. Agents trained to minimise a loss function across many domains are incentivized to learn a causal world model, which could in turn enable them to solve a much larger set of decision tasks they were not explicitly trained on.   Outline of",
            "references": [
                {
                    "title": "The Reasons that Agents Act: Intention and Instrumental Goals",
                    "abstract": "Intention is an important and challenging concept in AI. It is important because it underlies many other concepts we care about, such as agency, manipulation, legal responsibility, and blame. However, ascribing intent to AI systems is contentious, and there is no universally accepted theory of intention applicable to AI agents. We operationalise the intention with which an agent acts, relating to the reasons it chooses its decision. We introduce a formal definition of intention in structural causal influence models, grounded in the philosophy literature on intent and applicable to real-world machine learning systems. Through a number of examples and results, we show that our definition captures the intuitive notion of intent and satisfies desiderata set-out by past work. In addition, we show how our definition relates to past concepts, including actual causality, and the notion of instrumental goals, which is a core idea in the literature on safe AI agents. Finally, we demonstrate how our definition can be used to infer the intentions of reinforcement learning agents and language models from their behaviour."
                },
                {
                    "title": "Honesty Is the Best Policy: Defining and Mitigating AI Deception",
                    "abstract": "Deceptive agents are a challenge for the safety, trustworthiness, and cooperation of AI systems. We focus on the problem that agents might deceive in order to achieve their goals (for instance, in our experiments with language models, the goal of being evaluated as truthful). There are a number of existing definitions of deception in the literature on game theory and symbolic AI, but there is no overarching theory of deception for learning agents in games. We introduce a formal definition of deception in structural causal games, grounded in the philosophy literature, and applicable to real-world machine learning systems. Several examples and results illustrate that our formal definition aligns with the philosophical and commonsense meaning of deception. Our main technical result is to provide graphical criteria for deception. We show, experimentally, that these results can be used to mitigate deception in reinforcement learning agents and language models."
                },
                {
                    "title": "Language Models Represent Space and Time",
                    "abstract": "The capabilities of large language models (LLMs) have sparked debate over whether such systems just learn an enormous collection of superficial statistics or a set of more coherent and grounded representations that reflect the real world. We find evidence for the latter by analyzing the learned representations of three spatial datasets (world, US, NYC places) and three temporal datasets (historical figures, artworks, news headlines) in the Llama-2 family of models. We discover that LLMs learn linear representations of space and time across multiple scales. These representations are robust to prompting variations and unified across different entity types (e.g. cities and landmarks). In addition, we identify individual\"space neurons\"and\"time neurons\"that reliably encode spatial and temporal coordinates. While further investigation is needed, our results suggest modern LLMs learn rich spatiotemporal representations of the real world and possess basic ingredients of a world model."
                },
                {
                    "title": "Causal Parrots: Large Language Models May Talk Causality But Are Not Causal",
                    "abstract": "Some argue scale is all what is needed to achieve AI, covering even causal models. We make it clear that large language models (LLMs) cannot be causal and give reason onto why sometimes we might feel otherwise. To this end, we define and exemplify a new subgroup of Structural Causal Model (SCM) that we call meta SCM which encode causal facts about other SCM within their variables. We conjecture that in the cases where LLM succeed in doing causal inference, underlying was a respective meta SCM that exposed correlations between causal facts in natural language on whose data the LLM was ultimately trained. If our hypothesis holds true, then this would imply that LLMs are like parrots in that they simply recite the causal knowledge embedded in the data. Our empirical analysis provides favoring evidence that current LLMs are even weak `causal parrots.'"
                },
                {
                    "title": "Causal Reasoning and Large Language Models: Opening a New Frontier for Causality",
                    "abstract": "The causal capabilities of large language models (LLMs) are a matter of significant debate, with critical implications for the use of LLMs in societally impactful domains such as medicine, science, law, and policy. We conduct a\"behavorial\"study of LLMs to benchmark their capability in generating causal arguments. Across a wide range of tasks, we find that LLMs can generate text corresponding to correct causal arguments with high probability, surpassing the best-performing existing methods. Algorithms based on GPT-3.5 and 4 outperform existing algorithms on a pairwise causal discovery task (97%, 13 points gain), counterfactual reasoning task (92%, 20 points gain) and event causality (86% accuracy in determining necessary and sufficient causes in vignettes). We perform robustness checks across tasks and show that the capabilities cannot be explained by dataset memorization alone, especially since LLMs generalize to novel datasets that were created after the training cutoff date. That said, LLMs exhibit unpredictable failure modes, and we discuss the kinds of errors that may be improved and what are the fundamental limits of LLM-based answers. Overall, by operating on the text metadata, LLMs bring capabilities so far understood to be restricted to humans, such as using collected knowledge to generate causal graphs or identifying background causal context from natural language. As a result, LLMs may be used by human domain experts to save effort in setting up a causal analysis, one of the biggest impediments to the widespread adoption of causal methods. Given that LLMs ignore the actual data, our results also point to a fruitful research direction of developing algorithms that combine LLMs with existing causal techniques. Code and datasets are available at https://github.com/py-why/pywhy-llm."
                },
                {
                    "title": "Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task",
                    "abstract": "Language models show a surprising range of capabilities, but the source of their apparent competence is unclear. Do these networks just memorize a collection of surface statistics, or do they rely on internal representations of the process that generates the sequences they see? We investigate this question by applying a variant of the GPT model to the task of predicting legal moves in a simple board game, Othello. Although the network has no a priori knowledge of the game or its rules, we uncover evidence of an emergent nonlinear internal representation of the board state. Interventional experiments indicate this representation can be used to control the output of the network and create\"latent saliency maps\"that can help explain predictions in human terms."
                },
                {
                    "title": "Goal Misgeneralization: Why Correct Specifications Aren't Enough For Correct Goals",
                    "abstract": "The field of AI alignment is concerned with AI systems that pursue unintended goals. One commonly studied mechanism by which an unintended goal might arise is specification gaming, in which the designer-provided specification is flawed in a way that the designers did not foresee. However, an AI system may pursue an undesired goal even when the specification is correct, in the case of goal misgeneralization. Goal misgeneralization is a specific form of robustness failure for learning algorithms in which the learned program competently pursues an undesired goal that leads to good performance in training situations but bad performance in novel test situations. We demonstrate that goal misgeneralization can occur in practical systems by providing several examples in deep learning systems across a variety of domains. Extrapolating forward to more capable systems, we provide hypotheticals that illustrate how goal misgeneralization could lead to catastrophic risk. We suggest several research directions that could reduce the risk of goal misgeneralization for future systems."
                },
                {
                    "title": "Discovering Agents",
                    "abstract": "Causal models of agents have been used to analyse the safety aspects of machine learning systems. But identifying agents is non-trivial -- often the causal model is just assumed by the modeler without much justification -- and modelling failures can lead to mistakes in the safety analysis. This paper proposes the first formal causal definition of agents -- roughly that agents are systems that would adapt their policy if their actions influenced the world in a different way. From this we derive the first causal discovery algorithm for discovering agents from empirical data, and give algorithms for translating between causal models and game-theoretic influence diagrams. We demonstrate our approach by resolving some previous confusions caused by incorrect causal modelling of agents."
                },
                {
                    "title": "A Generalist Agent",
                    "abstract": "Inspired by progress in large-scale language modeling, we apply a similar approach towards building a single generalist agent beyond the realm of text outputs. The agent, which we refer to as Gato, works as a multi-modal, multi-task, multi-embodiment generalist policy. The same network with the same weights can play Atari, caption images, chat, stack blocks with a real robot arm and much more, deciding based on its context whether to output text, joint torques, button presses, or other tokens. In this report we describe the model and the data, and document the current capabilities of Gato."
                },
                {
                    "title": "Counterfactual harm",
                    "abstract": "To act safely and ethically in the real world, agents must be able to reason about harm and avoid harmful actions. However, to date there is no statistical method for measuring harm and factoring it into algorithmic decisions. In this paper we propose the first formal definition of harm and benefit using causal models. We show that any factual definition of harm must violate basic intuitions in certain scenarios, and show that standard machine learning algorithms that cannot perform counterfactual reasoning are guaranteed to pursue harmful policies following distributional shifts. We use our definition of harm to devise a framework for harm-averse decision making using counterfactual objective functions. We demonstrate this framework on the problem of identifying optimal drug doses using a dose-response model learned from randomized control trial data. We find that the standard method of selecting doses using treatment effects results in unnecessarily harmful doses, while our counterfactual approach allows us to identify doses that are significantly less harmful without sacrificing efficacy."
                },
                {
                    "title": "On Pearl\u2019s Hierarchy and the Foundations of Causal Inference",
                    "abstract": "Cause-and-effect relationships play a central role in how we perceive and make sense of the world around us, how we act upon it, and ultimately, how we under\u00ad stand ourselves. Almost two decades ago, computer scientist Judea Pearl made a breakthrough in understanding causality by discovering and systematically study\u00ad ing the \u201cLadder of Causation,\u201d a framework that highlights the distinct roles of seeing, doing, and imagining. In honor of this landmark discovery, we name this the Pearl Causal Hierarchy (PCH). In this chapter, we develop a novel and comprehensive treatment of the PCH through two complementary lenses: one logical-probabilistic and another inferential-graphical. Following Pearl\u2019s own pre\u00ad sentation of the hierarchy, we begin by showing how the PCH organically emerges from a well-specified collection of causal mechanisms (a structural causal model, or SCM). We then turn to the logical lens. Our first result, the Causal Hierarchy Theorem (CHT), demonstrates that the three layers of the hierarchy almost always separate in a measure-theoretic sense. Roughly speaking, the CHT says that data at one layer virtually always underdetermines information at higher layers. As in most practical settings the scientist does not have access to the precise form of the underlying causal mechanisms\u2014only to data generated by them with respect to some of the PCH\u2019s layers\u2014this motivates us to study inferences within the PCH through the graphical lens. Specifically, we explore a set of methods known as causal inference that enable inferences bridging the PCH\u2019s layers given a partial"
                },
                {
                    "title": "Selection, Ignorability and Challenges With Causal Fairness",
                    "abstract": "In this paper we look at popular fairness methods that use causal counterfactuals. These methods capture the intuitive notion that a prediction is fair if it coincides with the prediction that would have been made if someone\u2019s race, gender or religion were counterfactually di\ufb00erent. In order to achieve this, we must have causal models that are able to capture what someone would be like if we were to counterfactually change these traits. However, we argue that any model that can do this must lie outside the particularly well behaved class that is commonly considered in the fairness literature. This is because in fairness settings, models in this class entail a particularly strong causal assumption, normally only seen in a randomised controlled trial. We argue that in general this is unlikely to hold. Furthermore, we show in many cases it can be explicitly rejected due to the fact that samples are selected from a wider population. We show this creates di\ufb03culties for counterfactual fairness as well as for the application of more general causal fairness methods."
                },
                {
                    "title": "Current structure predictors are not learning the physics of protein folding",
                    "abstract": "Abstract Summary Motivation. Predicting the native state of a protein has long been considered a gateway problem for understanding protein folding. Recent advances in structural modeling driven by deep learning have achieved unprecedented success at predicting a protein\u2019s crystal structure, but it is not clear if these models are learning the physics of how proteins dynamically fold into their equilibrium structure or are just accurate knowledge-based predictors of the final state. Results. In this work, we compare the pathways generated by state-of-the-art protein structure prediction methods to experimental data about protein folding pathways. The methods considered were AlphaFold 2, RoseTTAFold, trRosetta, RaptorX, DMPfold, EVfold, SAINT2 and Rosetta. We find evidence that their simulated dynamics capture some information about the folding pathway, but their predictive ability is worse than a trivial classifier using sequence-agnostic features like chain length. The folding trajectories produced are also uncorrelated with experimental observables such as intermediate structures and the folding rate constant. These results suggest that recent advances in structure prediction do not yet provide an enhanced understanding of protein folding. Availability. The data underlying this article are available in GitHub at https://github.com/oxpig/structure-vs-folding/ Supplementary information Supplementary data are available at Bioinformatics online."
                },
                {
                    "title": "Promises and Challenges of Causality for Ethical Machine Learning",
                    "abstract": "Recent years have witnessed an increasing interest in causal reasoning for designing fair decision-making systems due to its compatibility with legal frameworks, interpretability for human stakeholders, and robustness to spurious correlations, among other factors. The recent attention, however, has been accompanied with great skepticism due to practical and epistemological challenges with applying existing causal fairness approaches. Motivated by the long-standing empirical work on causality in econometrics, social sciences, and biomedi- cal sciences, in this paper we lay out the conditions for appropriate application of causal fairness under the \u201cpotential outcomes framework.\u201d To this end, we propose a shift from postulating interventions on immutable social categories to their perceptions and highlight two key aspects of interven- tions that are largely overlooked in the causal fairness literature: timing and nature of manipulations. We argue that such conceptualization is key in evaluating the validity of causal assumptions and conducting sound causal analysis includ- ing avoiding post-treatment bias. Further, we illustrate how causality can address the limitations of existing fairness met- rics that depend upon statistical correlations. Speci\ufb01cally, we introduce causal variants of common statistical fairness no- tions, and we make a novel observation that under the causal framework there is no fundamental disagreement between different fairness criteria. Finally, extensive experiments on synthetic and real-world datasets including a case study on police stop and search decisions demonstrate the ef\ufb01cacy of our framework in evaluating and mitigating unfairness."
                },
                {
                    "title": "A Survey of Zero-shot Generalisation in Deep Reinforcement Learning",
                    "abstract": "The study of zero-shot generalisation (ZSG) in deep Reinforcement Learning (RL) aims to produce RL algorithms whose policies generalise well to novel unseen situations at deployment time, avoiding overfitting to their training environments. Tackling this is vital if we are to deploy reinforcement learning algorithms in real world scenarios, where the environment will be diverse, dynamic and unpredictable. This survey is an overview of this nascent field. We rely on a unifying formalism and terminology for discussing different ZSG problems, building upon previous works. We go on to categorise existing benchmarks for ZSG, as well as current methods for tackling these problems. Finally, we provide a critical discussion of the current state of the field, including recommendations for future work. Among other conclusions, we argue that taking a purely procedural content generation approach to benchmark design is not conducive to progress in ZSG, we suggest fast online adaptation and tackling RL-specific problems as some areas for future work on methods for ZSG, and we recommend building benchmarks in underexplored problem settings such as offline RL ZSG and reward-function variation."
                },
                {
                    "title": "Acquisition of chess knowledge in AlphaZero",
                    "abstract": "Significance Seventy years ago, Alan Turing conjectured that a chess-playing machine could be built that would self-learn and continuously profit from its own experience. The AlphaZero system\u2014a neural network-powered reinforcement learner\u2014passed this milestone. In this paper, we ask the following questions. How did it do it? What did it learn from its experience, and how did it encode it? Did it learn anything like a human understanding of chess, in spite of having never seen a human game? Remarkably, we find many strong correspondences between human concepts and AlphaZero\u2019s representations that emerge during training, even though none of these concepts were initially present in the network."
                },
                {
                    "title": "Can Language Models Encode Perceptual Structure Without Grounding? A Case Study in Color",
                    "abstract": "Pretrained language models have been shown to encode relational information, such as the relations between entities or concepts in knowledge-bases \u2014 (Paris, Capital, France). However, simple relations of this type can often be recovered heuristically and the extent to which models implicitly reflect topological structure that is grounded in world, such as perceptual structure, is unknown. To explore this question, we conduct a thorough case study on color. Namely, we employ a dataset of monolexemic color terms and color chips represented in CIELAB, a color space with a perceptually meaningful distance metric. Using two methods of evaluating the structural alignment of colors in this space with text-derived color term representations, we find significant correspondence. Analyzing the differences in alignment across the color spectrum, we find that warmer colors are, on average, better aligned to the perceptual color space than cooler ones, suggesting an intriguing connection to findings from recent work on efficient communication in color naming. Further analysis suggests that differences in alignment are, in part, mediated by collocationality and differences in syntactic usage, posing questions as to the relationship between color perception and usage and context."
                },
                {
                    "title": "Towards Out-Of-Distribution Generalization: A Survey",
                    "abstract": "Traditional machine learning paradigms are based on the assumption that both training and test data follow the same statistical pattern, which is mathematically referred to as Independent and Identically Distributed ($i.i.d.$). However, in real-world applications, this $i.i.d.$ assumption often fails to hold due to unforeseen distributional shifts, leading to considerable degradation in model performance upon deployment. This observed discrepancy indicates the significance of investigating the Out-of-Distribution (OOD) generalization problem. OOD generalization is an emerging topic of machine learning research that focuses on complex scenarios wherein the distributions of the test data differ from those of the training data. This paper represents the first comprehensive, systematic review of OOD generalization, encompassing a spectrum of aspects from problem definition, methodological development, and evaluation procedures, to the implications and future directions of the field. Our discussion begins with a precise, formal characterization of the OOD generalization problem. Following that, we categorize existing methodologies into three segments: unsupervised representation learning, supervised model learning, and optimization, according to their positions within the overarching learning process. We provide an in-depth discussion on representative methodologies for each category, further elucidating the theoretical links between them. Subsequently, we outline the prevailing benchmark datasets employed in OOD generalization studies. To conclude, we overview the existing body of work in this domain and suggest potential avenues for future research on OOD generalization. A summary of the OOD generalization methodologies surveyed in this paper can be accessed at http://out-of-distribution-generalization.com."
                },
                {
                    "title": "A Topological Perspective on Causal Inference",
                    "abstract": "This paper presents a topological learning-theoretic perspective on causal inference by introducing a series of topologies defined on general spaces of structural causal models (SCMs). As an illustration of the framework we prove a topological causal hierarchy theorem, showing that substantive assumption-free causal inference is possible only in a meager set of SCMs. Thanks to a known correspondence between open sets in the weak topology and statistically verifiable hypotheses, our results show that inductive assumptions sufficient to license valid causal inferences are statistically unverifiable in principle. Similar to no-free-lunch theorems for statistical inference, the present results clarify the inevitability of substantial assumptions for causal inference. An additional benefit of our topological approach is that it easily accommodates SCMs with infinitely many variables. We finally suggest that the framework may be helpful for the positive project of exploring and assessing alternative causal-inductive assumptions."
                },
                {
                    "title": "Goal Misgeneralization in Deep Reinforcement Learning",
                    "abstract": "We study goal misgeneralization, a type of out-of-distribution generalization failure in reinforcement learning (RL). Goal misgeneralization failures occur when an RL agent retains its capabilities out-of-distribution yet pursues the wrong goal. For instance, an agent might continue to competently avoid obstacles, but navigate to the wrong place. In contrast, previous works have typically focused on capability generalization failures, where an agent fails to do anything sensible at test time. We formalize this distinction between capability and goal generalization, provide the first empirical demonstrations of goal misgeneralization, and present a partial characterization of its causes."
                },
                {
                    "title": "Toward Causal Representation Learning",
                    "abstract": "The two fields of machine learning and graphical causality arose and are developed separately. However, there is, now, cross-pollination and increasing interest in both fields to benefit from the advances of the other. In this article, we review fundamental concepts of causal inference and relate them to crucial open problems of machine learning, including transfer and generalization, thereby assaying how causality can contribute to modern machine learning research. This also applies in the opposite direction: we note that most work in causality starts from the premise that the causal variables are given. A central problem for AI and causality is, thus, causal representation learning, that is, the discovery of high-level causal variables from low-level observations. Finally, we delineate some implications of causality for machine learning and propose key research areas at the intersection of both communities."
                },
                {
                    "title": "Maximum Entropy RL (Provably) Solves Some Robust RL Problems",
                    "abstract": "Many potential applications of reinforcement learning (RL) require guarantees that the agent will perform well in the face of disturbances to the dynamics or reward function. In this paper, we prove theoretically that standard maximum entropy RL is robust to some disturbances in the dynamics and the reward function. While this capability of MaxEnt RL has been observed empirically in prior work, to the best of our knowledge our work provides the first rigorous proof and theoretical characterization of the MaxEnt RL robust set. While a number of prior robust RL algorithms have been designed to handle similar disturbances to the reward function or dynamics, these methods typically require adding additional moving parts and hyperparameters on top of a base RL algorithm. In contrast, our theoretical results suggest that MaxEnt RL by itself is robust to certain disturbances, without requiring any additional modifications. While this does not imply that MaxEnt RL is the best available robust RL method, MaxEnt RL does possess a striking simplicity and appealing formal guarantees."
                },
                {
                    "title": "Domain Generalization: A Survey",
                    "abstract": "Generalization to out-of-distribution (OOD) data is a capability natural to humans yet challenging for machines to reproduce. This is because most learning algorithms strongly rely on the i.i.d. assumption on source/target data, which is often violated in practice due to domain shift. Domain generalization (DG) aims to achieve OOD generalization by using only source data for model learning. Over the last ten years, research in DG has made great progress, leading to a broad spectrum of methodologies, e.g., those based on domain alignment, meta-learning, data augmentation, or ensemble learning, to name a few; DG has also been studied in various application areas including computer vision, speech recognition, natural language processing, medical imaging, and reinforcement learning. In this paper, for the first time a comprehensive literature review in DG is provided to summarize the developments over the past decade. Specifically, we first cover the background by formally defining DG and relating it to other relevant fields like domain adaptation and transfer learning. Then, we conduct a thorough review into existing methods and theories. Finally, we conclude this survey with insights and discussions on future research directions."
                },
                {
                    "title": "D\u2019ya Like DAGs? A Survey on Structure Learning and Causal Discovery",
                    "abstract": "Causal reasoning is a crucial part of science and human intelligence. In order to discover causal relationships from data, we need structure discovery methods. We provide a review of background theory and a survey of methods for structure discovery. We primarily focus on modern, continuous optimization methods, and provide reference to further resources such as benchmark datasets and software packages. Finally, we discuss the assumptive leap required to take us from structure to causality."
                },
                {
                    "title": "Generalizing to Unseen Domains: A Survey on Domain Generalization",
                    "abstract": "Machine learning systems generally assume that the training and testing distributions are the same. To this end, a key requirement is to develop models that can generalize to unseen distributions. Domain generalization (DG), i.e., out-of-distribution generalization, has attracted increasing interests in recent years. Domain generalization deals with a challenging setting where one or several different but related domain(s) are given, and the goal is to learn a model that can generalize to an unseen test domain. Great progress has been made in the area of domain generalization for years. This paper presents the first review of recent advances in this area. First, we provide a formal definition of domain generalization and discuss several related fields. We then thoroughly review the theories related to domain generalization and carefully analyze the theory behind generalization. We categorize recent algorithms into three classes: data manipulation, representation learning, and learning strategy, and present several popular algorithms in detail for each category. Third, we introduce the commonly used datasets, applications, and our open-sourced codebase for fair evaluation. Finally, we summarize existing literature and present some potential research topics for the future."
                },
                {
                    "title": "Agent Incentives: A Causal Perspective",
                    "abstract": "We present a framework for analysing agent incentives using causal influence diagrams. We establish that a well-known criterion for value of information is complete. We propose a new graphical criterion for value of control, establishing its soundness and completeness. We also introduce two new concepts for incentive analysis: response incentives indicate which changes in the environment affect an optimal decision, while instrumental control incentives establish whether an agent can influence its utility via a variable X. For both new concepts, we provide sound and complete graphical criteria. We show by example how these results can help with evaluating the safety and fairness of an AI system"
                },
                {
                    "title": "Inductive biases for deep learning of higher-level cognition",
                    "abstract": "A fascinating hypothesis is that human and animal intelligence could be explained by a few principles (rather than an encyclopaedic list of heuristics). If that hypothesis was correct, we could more easily both understand our own intelligence and build intelligent machines. Just like in physics, the principles themselves would not be sufficient to predict the behaviour of complex systems like brains, and substantial computation might be needed to simulate human-like intelligence. This hypothesis would suggest that studying the kind of inductive biases that humans and animals exploit could help both clarify these principles and provide inspiration for AI research and neuroscience theories. Deep learning already exploits several key inductive biases, and this work considers a larger list, focusing on those which concern mostly higher-level and sequential conscious processing. The objective of clarifying these particular principles is that they could potentially help us build AI systems benefiting from humans\u2019 abilities in terms of flexible out-of-distribution and systematic generalization, which is currently an area where a large gap exists between state-of-the-art machine learning and human intelligence."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "A Calculus for Stochastic Interventions: Causal Effect Identification and Surrogate Experiments",
                    "abstract": "Some of the most prominent results in causal inference have been developed in the context of atomic interventions, following the semantics of the do-operator and the inferential power of the do-calculus. In practice, many real-world settings require more complex types of interventions that cannot be represented by a simple atomic intervention. In this paper, we investigate a general class of interventions that covers some non-trivial types of policies (conditional and stochastic), which goes beyond the atomic class. Our goal is to develop general understanding and formal machinery to be able to reason about the effects of those policies, similar to the robust treatment developed to handle the atomic case. Specifically, in this paper, we introduce a new set of inference rules (akin to do-calculus) that can be used to derive claims about general interventions, which we call \u03c3-calculus. We develop a systematic and efficient procedure for finding estimands of the effect of general policies as a function of the available observational and experimental distributions. We then prove that our algorithm and \u03c3-calculus are both sound for the tasks of identification (Pearl, 1995) and z-identification (Bareinboim and Pearl, 2012) under this class of interventions."
                },
                {
                    "title": "A Comprehensive Survey on Transfer Learning",
                    "abstract": "Transfer learning aims at improving the performance of target learners on target domains by transferring the knowledge contained in different but related source domains. In this way, the dependence on a large number of target-domain data can be reduced for constructing target learners. Due to the wide application prospects, transfer learning has become a popular and promising area in machine learning. Although there are already some valuable and impressive surveys on transfer learning, these surveys introduce approaches in a relatively isolated way and lack the recent advances in transfer learning. Due to the rapid expansion of the transfer learning area, it is both necessary and challenging to comprehensively review the relevant studies. This survey attempts to connect and systematize the existing transfer learning research studies, as well as to summarize and interpret the mechanisms and the strategies of transfer learning in a comprehensive way, which may help readers have a better understanding of the current research status and ideas. Unlike previous surveys, this survey article reviews more than 40 representative transfer learning approaches, especially homogeneous transfer learning approaches, from the perspectives of data and model. The applications of transfer learning are also briefly introduced. In order to show the performance of different transfer learning models, over 20 representative transfer learning models are used for experiments. The models are performed on three different data sets, that is, Amazon Reviews, Reuters-21578, and Office-31, and the experimental results demonstrate the importance of selecting appropriate transfer learning models for different applications in practice."
                },
                {
                    "title": "Invariant Risk Minimization",
                    "abstract": "We introduce Invariant Risk Minimization (IRM), a learning paradigm to estimate invariant correlations across multiple training distributions. To achieve this goal, IRM learns a data representation such that the optimal classifier, on top of that data representation, matches for all training distributions. Through theory and experiments, we show how the invariances learned by IRM relate to the causal structures governing the data and enable out-of-distribution generalization."
                },
                {
                    "title": "Benchmarking Neural Network Robustness to Common Corruptions and Perturbations",
                    "abstract": "In this paper we establish rigorous benchmarks for image classifier robustness. Our first benchmark, ImageNet-C, standardizes and expands the corruption robustness topic, while showing which classifiers are preferable in safety-critical applications. Then we propose a new dataset called ImageNet-P which enables researchers to benchmark a classifier's robustness to common perturbations. Unlike recent robustness research, this benchmark evaluates performance on common corruptions and perturbations not worst-case adversarial perturbations. We find that there are negligible changes in relative corruption robustness from AlexNet classifiers to ResNet classifiers. Afterward we discover ways to enhance corruption and perturbation robustness. We even find that a bypassed adversarial defense provides substantial common perturbation robustness. Together our benchmarks may aid future work toward networks that robustly generalize."
                },
                {
                    "title": "The seven tools of causal inference, with reflections on machine learning",
                    "abstract": "The kind of causal inference seen in natural human thought can be \"algorithmitized\" to help produce human-level machine intelligence."
                },
                {
                    "title": "A Survey of Unsupervised Deep Domain Adaptation",
                    "abstract": "Deep learning has produced state-of-the-art results for a variety of tasks. While such approaches for supervised learning have performed well, they assume that training and testing data are drawn from the same distribution, which may not always be the case. As a complement to this challenge, single-source unsupervised domain adaptation can handle situations where a network is trained on labeled data from a source domain and unlabeled data from a related but different target domain with the goal of performing well at test-time on the target domain. Many single-source and typically homogeneous unsupervised deep domain adaptation approaches have thus been developed, combining the powerful, hierarchical representations from deep learning with domain adaptation to reduce reliance on potentially costly target data labels. This survey will compare these approaches by examining alternative methods, the unique and common elements, results, and theoretical insights. We follow this with a look at application areas and open research directions."
                },
                {
                    "title": "On the Spectral Bias of Neural Networks",
                    "abstract": "Neural networks are known to be a class of highly expressive functions able to fit even random input-output mappings with $100\\%$ accuracy. In this work, we present properties of neural networks that complement this aspect of expressivity. By using tools from Fourier analysis, we show that deep ReLU networks are biased towards low frequency functions, meaning that they cannot have local fluctuations without affecting their global behavior. Intuitively, this property is in line with the observation that over-parameterized networks find simple patterns that generalize across data samples. We also investigate how the shape of the data manifold affects expressivity by showing evidence that learning high frequencies gets \\emph{easier} with increasing manifold complexity, and present a theoretical understanding of this behavior. Finally, we study the robustness of the frequency components with respect to parameter perturbation, to develop the intuition that the parameters must be finely tuned to express high frequency functions."
                },
                {
                    "title": "Relational inductive biases, deep learning, and graph networks",
                    "abstract": "Artificial intelligence (AI) has undergone a renaissance recently, making major progress in key domains such as vision, language, control, and decision-making. This has been due, in part, to cheap data and cheap compute resources, which have fit the natural strengths of deep learning. However, many defining characteristics of human intelligence, which developed under much different pressures, remain out of reach for current approaches. In particular, generalizing beyond one's experiences--a hallmark of human intelligence from infancy--remains a formidable challenge for modern AI. \nThe following is part position paper, part review, and part unification. We argue that combinatorial generalization must be a top priority for AI to achieve human-like abilities, and that structured representations and computations are key to realizing this objective. Just as biology uses nature and nurture cooperatively, we reject the false choice between \"hand-engineering\" and \"end-to-end\" learning, and instead advocate for an approach which benefits from their complementary strengths. We explore how using relational inductive biases within deep learning architectures can facilitate learning about entities, relations, and rules for composing them. We present a new building block for the AI toolkit with a strong relational inductive bias--the graph network--which generalizes and extends various approaches for neural networks that operate on graphs, and provides a straightforward interface for manipulating structured knowledge and producing structured behaviors. We discuss how graph networks can support relational reasoning and combinatorial generalization, laying the foundation for more sophisticated, interpretable, and flexible patterns of reasoning. As a companion to this paper, we have released an open-source software library for building graph networks, with demonstrations of how to use them in practice."
                },
                {
                    "title": "CAUSALITY FROM A DISTRIBUTIONAL ROBUSTNESS POINT OF VIEW",
                    "abstract": "We show some connections between distributional robust and causal inference. Distributional robustness is here understood as optimizing the predictive accuracy for a whole class of distributions instead of just a single target distribution. The class of distributions considered is often a ball around the empirical distribution, using a Wasserstein or Kullback-Leibler distance. Causal inference can be seen as a special case of distributional robustness, where the class of distributions is generated by a causal model. Different choices for these classes are discussed, which correspond to different assumptions about the type of interventions and presence or absence of latent confounding variables."
                },
                {
                    "title": "Why do deep convolutional networks generalize so poorly to small image transformations?",
                    "abstract": "Convolutional Neural Networks (CNNs) are commonly assumed to be invariant to small image transformations: either because of the convolutional architecture or because they were trained using data augmentation. Recently, several authors have shown that this is not the case: small translations or rescalings of the input image can drastically change the network's prediction. In this paper, we quantify this phenomena and ask why neither the convolutional architecture nor data augmentation are sufficient to achieve the desired invariance. Specifically, we show that the convolutional architecture does not give invariance since architectures ignore the classical sampling theorem, and data augmentation does not give invariance because the CNNs learn to be invariant to transformations only for images that are very similar to typical images from the training set. We discuss two possible solutions to this problem: (1) antialiasing the intermediate representations and (2) increasing data augmentation and show that they provide only a partial solution at best. Taken together, our results indicate that the problem of insuring invariance to small image transformations in neural networks while preserving high accuracy remains unsolved."
                },
                {
                    "title": "The Book of Why: The New Science of Cause and Effect",
                    "abstract": "This week on the Science podcast, Judea Pearl and Dana Mackenzie discuss strategies for causal thinking and describe its implications for artificial intelligence."
                },
                {
                    "title": "Towards Formal Definitions of Blameworthiness, Intention, and Moral Responsibility",
                    "abstract": "\n \n We provide formal definitions of degree of blameworthiness and intention relative to an epistemic state (a probability over causal models and a utility function on outcomes). These, together with a definition of actual causality, provide the key ingredients for moral responsibility judgments. We show that these definitions give insight into commonsense intuitions in a variety of puzzling cases from the literature.\n \n"
                },
                {
                    "title": "Causal Inference via Kernel Deviance Measures",
                    "abstract": "Discovering the causal structure among a set of variables is a fundamental problem in many areas of science. In this paper, we propose Kernel Conditional Deviance for Causal Inference (KCDC) a fully nonparametric causal discovery method based on purely observational data. From a novel interpretation of the notion of asymmetry between cause and effect, we derive a corresponding asymmetry measure using the framework of reproducing kernel Hilbert spaces. Based on this, we propose three decision rules for causal discovery. We demonstrate the wide applicability of our method across a range of diverse synthetic datasets. Furthermore, we test our method on real-world time series data and the real-world benchmark dataset Tubingen Cause-Effect Pairs where we outperform existing state-of-the-art methods."
                },
                {
                    "title": "Theoretical Impediments to Machine Learning With Seven Sparks from the Causal Revolution",
                    "abstract": "Current machine learning systems operate, almost exclusively, in a statistical, or model-blind mode, which entails severe theoretical limits on their power and performance. Such systems cannot reason about interventions and retrospection and, therefore, cannot serve as the basis for strong AI. To achieve human level intelligence, learning machines need the guidance of a model of reality, similar to the ones used in causal inference. To demonstrate the essential role of such models, I will present a summary of seven tasks which are beyond reach of current machine learning systems and which have been accomplished using the tools of causal inference."
                },
                {
                    "title": "Exploring the Landscape of Spatial Robustness",
                    "abstract": "The study of adversarial robustness has so far largely focused on perturbations bound in p-norms. However, state-of-the-art models turn out to be also vulnerable to other, more natural classes of perturbations such as translations and rotations. In this work, we thoroughly investigate the vulnerability of neural network--based classifiers to rotations and translations. While data augmentation offers relatively small robustness, we use ideas from robust optimization and test-time input aggregation to significantly improve robustness. Finally we find that, in contrast to the p-norm case, first-order methods cannot reliably find worst-case perturbations. This highlights spatial robustness as a fundamentally different setting requiring additional study. Code available at this https URL and this https URL."
                },
                {
                    "title": "Sim-to-Real Transfer of Robotic Control with Dynamics Randomization",
                    "abstract": "Simulations are attractive environments for training agents as they provide an abundant source of data and alleviate certain safety concerns during the training process. But the behaviours developed by agents in simulation are often specific to the characteristics of the simulator. Due to modeling error, strategies that are successful in simulation may not transfer to their real world counterparts. In this paper, we demonstrate a simple method to bridge this \u201creality gap\u201d. By randomizing the dynamics of the simulator during training, we are able to develop policies that are capable of adapting to very different dynamics, including ones that differ significantly from the dynamics on which the policies were trained. This adaptivity enables the policies to generalize to the dynamics of the real world without any training on the physical system. Our approach is demonstrated on an object pushing task using a robotic arm. Despite being trained exclusively in simulation, our policies are able to maintain a similar level of performance when deployed on a real robot, reliably moving an object to a desired location from random initial configurations. We explore the impact of various design decisions and show that the resulting policies are robust to significant calibration error."
                },
                {
                    "title": "Counterfactual Explanations Without Opening the Black Box: Automated Decisions and the GDPR",
                    "abstract": "There has been much discussion of the right to explanation in the EU General Data Protection Regulation, and its existence, merits, and disadvantages. Implementing a right to explanation that opens the black box of algorithmic decision-making faces major legal and technical barriers. Explaining the functionality of complex algorithmic decision-making systems and their rationale in specific cases is a technically challenging problem. Some explanations may offer little meaningful information to data subjects, raising questions around their value. Explanations of automated decisions need not hinge on the general public understanding how algorithmic systems function. Even though such interpretability is of great importance and should be pursued, explanations can, in principle, be offered without opening the black box. Looking at explanations as a means to help a data subject act rather than merely understand, one could gauge the scope and content of explanations according to the specific goal or action they are intended to support. From the perspective of individuals affected by automated decision-making, we propose three aims for explanations: (1) to inform and help the individual understand why a particular decision was reached, (2) to provide grounds to contest the decision if the outcome is undesired, and (3) to understand what would need to change in order to receive a desired result in the future, based on the current decision-making model. We assess how each of these goals finds support in the GDPR. We suggest data controllers should offer a particular type of explanation, unconditional counterfactual explanations, to support these three aims. These counterfactual explanations describe the smallest change to the world that can be made to obtain a desirable outcome, or to arrive at the closest possible world, without needing to explain the internal logic of the system."
                },
                {
                    "title": "Causally Regularized Learning with Agnostic Data Selection Bias",
                    "abstract": "Most of previous machine learning algorithms are proposed based on the i.i.d. hypothesis. However, this ideal assumption is often violated in real applications, where selection bias may arise between training and testing process. Moreover, in many scenarios, the testing data is not even available during the training process, which makes the traditional methods like transfer learning infeasible due to their need on prior of test distribution. Therefore, how to address the agnostic selection bias for robust model learning is of paramount importance for both academic research and real applications. In this paper, under the assumption that causal relationships among variables are robust across domains, we incorporate causal technique into predictive modeling and propose a novel Causally Regularized Logistic Regression (CRLR) algorithm by jointly optimize global confounder balancing and weighted logistic regression. Global confounder balancing helps to identify causal features, whose causal effect on outcome are stable across domains, then performing logistic regression on those causal features constructs a robust predictive model against the agnostic bias. To validate the effectiveness of our CRLR algorithm, we conduct comprehensive experiments on both synthetic and real world datasets. Experimental results clearly demonstrate that our CRLR algorithm outperforms the state-of-the-art methods, and the interpretability of our method can be fully depicted by the feature visualization."
                },
                {
                    "title": "Zero-Shot Learning\u2014A Comprehensive Evaluation of the Good, the Bad and the Ugly",
                    "abstract": "Due to the importance of zero-shot learning, i.e., classifying images where there is a lack of labeled training data, the number of proposed approaches has recently increased steadily. We argue that it is time to take a step back and to analyze the status quo of the area. The purpose of this paper is three-fold. First, given the fact that there is no agreed upon zero-shot learning benchmark, we first define a new benchmark by unifying both the evaluation protocols and data splits of publicly available datasets used for this task. This is an important contribution as published results are often not comparable and sometimes even flawed due to, e.g., pre-training on zero-shot test classes. Moreover, we propose a new zero-shot learning dataset, the Animals with Attributes 2 (AWA2) dataset which we make publicly available both in terms of image features and the images themselves. Second, we compare and analyze a significant number of the state-of-the-art methods in depth, both in the classic zero-shot setting but also in the more realistic generalized zero-shot setting. Finally, we discuss in detail the limitations of the current status of the area which can be taken as a basis for advancing it."
                },
                {
                    "title": "Counterfactual Fairness",
                    "abstract": "Machine learning can impact people with legal or ethical consequences when it is used to automate decisions in areas such as insurance, lending, hiring, and predictive policing. In many of these scenarios, previous decisions have been made that are unfairly biased against certain subpopulations, for example those of a particular race, gender, or sexual orientation. Since this past data may be biased, machine learning predictors must account for this to avoid perpetuating or creating discriminatory practices. In this paper, we develop a framework for modeling fairness using tools from causal inference. Our definition of counterfactual fairness captures the intuition that a decision is fair towards an individual if it the same in (a) the actual world and (b) a counterfactual world where the individual belonged to a different demographic group. We demonstrate our framework on a real-world problem of fair prediction of success in law school."
                },
                {
                    "title": "Causal inference and the data-fusion problem",
                    "abstract": "We review concepts, principles, and tools that unify current approaches to causal analysis and attend to new challenges presented by big data. In particular, we address the problem of data fusion\u2014piecing together multiple datasets collected under heterogeneous conditions (i.e., different populations, regimes, and sampling methods) to obtain valid answers to queries of interest. The availability of multiple heterogeneous datasets presents new opportunities to big data analysts, because the knowledge that can be acquired from combined data would not be possible from any individual source alone. However, the biases that emerge in heterogeneous environments require new analytical tools. Some of these biases, including confounding, sampling selection, and cross-population biases, have been addressed in isolation, largely in restricted parametric models. We here present a general, nonparametric framework for handling these biases and, ultimately, a theoretical solution to the problem of data fusion in causal inference tasks."
                },
                {
                    "title": "Group Equivariant Convolutional Networks",
                    "abstract": "We introduce Group equivariant Convolutional Neural Networks (G-CNNs), a natural generalization of convolutional neural networks that reduces sample complexity by exploiting symmetries. G-CNNs use G-convolutions, a new type of layer that enjoys a substantially higher degree of weight sharing than regular convolution layers. G-convolutions increase the expressive capacity of the network without increasing the number of parameters. Group convolution layers are easy to use and can be implemented with negligible computational overhead for discrete groups generated by translations, reflections and rotations. G-CNNs achieve state of the art results on CI- FAR10 and rotated MNIST."
                },
                {
                    "title": "Causality in thought.",
                    "abstract": "Causal knowledge plays a crucial role in human thought, but the nature of causal representation and inference remains a puzzle. Can human causal inference be captured by relations of probabilistic dependency, or does it draw on richer forms of representation? This article explores this question by reviewing research in reasoning, decision making, various forms of judgment, and attribution. We endorse causal Bayesian networks as the best normative framework and as a productive guide to theory building. However, it is incomplete as an account of causal thinking. On the basis of a range of experimental work, we identify three hallmarks of causal reasoning-the role of mechanism, narrative, and mental simulation-all of which go beyond mere probabilistic knowledge. We propose that the hallmarks are closely related. Mental simulations are representations over time of mechanisms. When multiple actors are involved, these simulations are aggregated into narratives."
                },
                {
                    "title": "The Black Box Society: The Secret Algorithms That Control Money and Information",
                    "abstract": "Every day, corporations are connecting the dots about our personal behaviorsilently scrutinizing clues left behind by our work habits and Internet use. The data compiled and portraits created are incredibly detailed, to the point of being invasive. But who connects the dots about what firms are doing with this information? The Black Box Society argues that we all need to be able to do soand to set limits on how big data affects our lives. Hidden algorithms can make (or ruin) reputations, decide the destiny of entrepreneurs, or even devastate an entire economy. Shrouded in secrecy and complexity, decisions at major Silicon Valley and Wall Street firms were long assumed to be neutral and technical. But leaks, whistleblowers, and legal disputes have shed new light on automated judgment. Self-serving and reckless behavior is surprisingly common, and easy to hide in code protected by legal and real secrecy. Even after billions of dollars of fines have been levied, underfunded regulators may have only scratched the surface of this troubling behavior. Frank Pasquale exposes how powerful interests abuse secrecy for profit and explains ways to rein them in. Demanding transparency is only the first step. An intelligible society would assure that key decisions of its most important firms are fair, nondiscriminatory, and open to criticism. Silicon Valley and Wall Street need to accept as much accountability as they impose on others."
                },
                {
                    "title": "In Search of the Real Inductive Bias: On the Role of Implicit Regularization in Deep Learning",
                    "abstract": "We present experiments demonstrating that some other form of capacity control, different from network size, plays a central role in learning multilayer feed-forward networks. We argue, partially through analogy to matrix factorization, that this is an inductive bias that can help shed light on deep learning."
                },
                {
                    "title": "Distinguishing Cause from Effect Using Observational Data: Methods and Benchmarks",
                    "abstract": "The discovery of causal relationships from purely observational data is a fundamental problem in science. The most elementary form of such a causal discovery problem is to decide whether X causes Y or, alternatively, Y causes X, given joint observations of two variables X,Y. An example is to decide whether altitude causes temperature, or vice versa, given only joint measurements of both variables. Even under the simplifying assumptions of no confounding, no feedback loops, and no selection bias, such bivariate causal discovery problems are challenging. Nevertheless, several approaches for addressing those problems have been proposed in recent years. We review two families of such methods: methods based on Additive Noise Models (ANMs) and Information Geometric Causal Inference (IGCI). We present the benchmark CAUSEEFFECTPAIRS that consists of data for 100 different causee ffect pairs selected from 37 data sets from various domains (e.g., meteorology, biology, medicine, engineering, economy, etc.) and motivate our decisions regarding the \"ground truth\" causal directions of all pairs. We evaluate the performance of several bivariate causal discovery methods on these real-world benchmark data and in addition on artificially simulated data. Our empirical results on real-world data indicate that certain methods are indeed able to distinguish cause from effect using only purely observational data, although more benchmark data would be needed to obtain statistically significant conclusions. One of the best performing methods overall is the method based on Additive Noise Models that has originally been proposed by Hoyer et al. (2009), which obtains an accuracy of 63 \u00b1 10 % and an AUC of 0.74 \u00b1 0.05 on the real-world benchmark. As the main theoretical contribution of this work we prove the consistency of that method."
                },
                {
                    "title": "Transportability of Causal Effects: Completeness Results",
                    "abstract": "\n \n The study of transportability aims to identify conditions under which causal information learned from experiments can be reused in a different environment where only passive observations can be collected. The theory introduced in [Pearl and Bareinboim, 2011] (henceforth [PB, 2011]) defines formal conditions for such transfer but falls short of providing an effective procedure for deciding, given assumptions about differences between the source and target domains, whether transportability is feasible. This paper provides such procedure. It establishes a necessary and sufficient condition for deciding when causal effects in the target domain are estimable from both the statistical information available and the causal information transferred from the experiments. The paper further provides a complete algorithm for computing the transport formula, that is, a way of fusing experimental and observational information to synthesize an estimate of the desired causal relation.\n \n"
                },
                {
                    "title": "On causal and anticausal learning",
                    "abstract": "We consider the problem of function estimation in the case where an underlying causal model can be inferred. This has implications for popular scenarios such as covariate shift, concept drift, transfer learning and semi-supervised learning. We argue that causal knowledge may facilitate some approaches for a given problem, and rule out others. In particular, we formulate a hypothesis for when semi-supervised learning can help, and corroborate it with empirical results."
                },
                {
                    "title": "Transportability of Causal and Statistical Relations: A Formal Approach",
                    "abstract": "We address the problem of transferring information learned from experiments to a different environment, in which only passive observations can be collected. We introduce a formal representation called ``selection diagrams'' for expressing knowledge about differences and commonalities between environments and, using this representation, we derive procedures for deciding whether effects in the target environment can be inferred from experiments conducted elsewhere. When the answer is affirmative, the procedures identify the set of experiments and observations that need be conducted to license the transport. We further discuss how transportability analysis can guide the transfer of knowledge in non-experimental learning to minimize re-measurement cost and improve prediction power."
                },
                {
                    "title": "Controlling Selection Bias in Causal Inference",
                    "abstract": "\n \n Selection bias, caused by preferential exclusion of units (or samples) from the data, is a major obstacle to valid causal inferences, for it cannot be removed or even detected by randomized experiments. This paper highlights several graphical and algebraic methods capable of mitigating and sometimes eliminating this bias. These nonparametric methods generalize and improve previously reported results, and identify the type of knowledge that need to be available for reasoning in the presence of selection bias\n \n"
                },
                {
                    "title": "Domain Adaptation: Learning Bounds and Algorithms",
                    "abstract": "This paper addresses the general problem of domain adaptation which arises in a variety of applications where the distribution of the labeled sample available somewhat differs from that of the test data. Building on previous work by Ben-David et al. (2007), we introduce a novel distance between distributions, discrepancy distance, that is tailored to adaptation problems with arbitrary loss functions. We give Rademacher complexity bounds for estimating the discrepancy distance from finite samples for different loss functions. Using this distance, we derive new generalization bounds for domain adaptation for a wide family of loss functions. We also present a series of novel adaptation bounds for large classes of regularization-based algorithms, including support vector machines and kernel ridge regression based on the empirical discrepancy. This motivates our analysis of the problem of minimizing the empirical discrepancy for various loss functions for which we also give several algorithms. We report the results of preliminary experiments that demonstrate the benefits of our discrepancy minimization algorithms for domain adaptation."
                },
                {
                    "title": "Nonlinear causal discovery with additive noise models",
                    "abstract": "The discovery of causal relationships between a set of observed variables is a fundamental problem in science. For continuous-valued data linear acyclic causal models with additive noise are often used because these models are well understood and there are well-known methods to fit them to data. In reality, of course, many causal relationships are more or less nonlinear, raising some doubts as to the applicability and usefulness of purely linear methods. In this contribution we show that the basic linear framework can be generalized to nonlinear models. In this extended framework, nonlinearities in the data-generating process are in fact a blessing rather than a curse, as they typically provide information on the underlying causal system and allow more aspects of the true data-generating mechanisms to be identified. In addition to theoretical results we show simulations and some simple real data experiments illustrating the identification power provided by nonlinearities."
                },
                {
                    "title": "Causation, Prediction, and Search",
                    "abstract": "The writing is not uniformly polished and is scattered with long, awkward sentences that require some effort to unravel. I wonder if this is the result of infelicitous translation from the original German version (Wellek 1994). There are also numerous small typographical errors. More careful editing could have solved these problems before publication. There are no exercises, and so I would hesitate to use the book as a text (although it should be noted that this is not one of the author\u2019s stated aims). Although Testing Statistical Hypotheses of Equivalence has some weaknesses, it is a useful reference for those interested in the question of equivalence testing, particularly in biological applications."
                },
                {
                    "title": "Influence Diagrams for Causal Modelling and Inference",
                    "abstract": "We consider a variety of ways in which probabilistic and causal models can be represented in graphical form. By adding nodes to our graphs to represent parameters, decision, etc., we obtain a generalisation of influence diagrams that supports meaningful causal modelling and inference, and only requires concepts and methods that are already standard in the purely probabilistic case. We relate our representations to others, particularly functional models, and present arguments and examples in favour of their superiority."
                },
                {
                    "title": "Strong completeness and faithfulness in Bayesian networks",
                    "abstract": "A completeness result for d-separation applied to discrete Bayesian networks is presented and it is shown that in a strong measure-theoretic sense almost all discrete distributions for a given network structure are faithful; i.e. the independence facts true of the distribution are all and only those entailed by the network structure"
                },
                {
                    "title": "Defining Deception in Structural Causal Games",
                    "abstract": "Abstract"
                },
                {
                    "title": "Causal learning : psychology, philosophy, and computation",
                    "abstract": "PART I: CAUSATION AND INTERVENTION 1. Interventionist Theories of Causation in Psychological Perspective 2. Infants' Causal Learning: Intervention, Observation, Imitation 3. Detecting Causal Structure: The Role of Interventions in Infants' Understanding of Psychological and Physical Causal Relations 4. An Interventionist Approach to Causation in Psychology 5. Learning From Doing: Intervention and Causal Inference 6. Causal Reasoning Through Intervention 7. On the Importance of Causal Taxonomy PART II: CAUSATION AND PROBABILITY Introduction to Part II: Causation and Probability 8. Teaching the Normative Theory of Causal Reasoning 9. Interactions Between Causal and Statistical Learning 10. Beyond Covariation: Cues to Causal Structure 11. Theory Unification and Graphical Models in Human Categorization 12. Essentialism as a Generative Theory of Classification 13. Data-Mining Probalists or Experimental Determinists? A Dialogue on the Principles Underlying Causal Learning in Children 14. Learning the Structure of Deterministic Systems PART III: CAUSATION, THEORIES, AND MECHANISMS Introduction to Part III: Causation, Theories, and Mechanisms 15. Why Represent Causal Relations? 16. Causal Reasoning as Informed by the Early Development of Explanations 17. Dynamic Interpretations of Covariation Data 18. Statistical Jokes and Social Effects: Intervention and Invariance in Causal Relations 19. Intuitive Theories as Grammars for Causal Inference 20. Two Proposals for Causal Grammars"
                },
                {
                    "title": "Causality : Models , Reasoning , and Inference",
                    "abstract": "the methodological community a major statement on causal inquiry. His account of the philosophy and history of causal analysis is a joy to read, especially his spirited 28-page epilogue, \" The Art and Science of Cause and Effect. \" The heart of Pearl's Causality is the set of ideas that he and his colleagues in computer science have developed over the past 20 years. They comprise a unified methodology for the graphical representation of joint probability distributions along with rules for inferring causality directly from such graphical representations. Reading Causality, however, feels a bit (I would imagine) like going for a hike in a rain forest with Judea Pearl as your guide. Initially, one's excitement is aroused by first contact with his causal flora and then further enchanted by his claims of all that lies ahead. But, before long, one is so deeply enveloped in his graphical vegetation that fear of ever finding a way out begins to take over. As I write this review, I must admit that I am still circling about in the counterfactual mangroves of Chapters 7 through 10. Nonetheless, I have come to a recommendation: Anyone who intends to teach causality in graduate methodology classes should read this book. Anyone who does not should avoid it, at least until the rest of us figure out which of its many ideas are unique and lasting contributions to causal analysis. That is to say, from the perspective of a sociologist, the work as a whole is a memorable tour de force, which succeeds admirably in provoking deep thoughts about (1) what social scientists can and should do with available observational data and (2) what sorts of theories and data sources we should be attempting to cultivate. But it is unclear which of Pearl's specific ideas should or"
                }
            ],
            "categories": [
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nDo agents have to learn causal models in order to adapt to new domains, or are other inductive biases sufficient?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the foundational role of causal reasoning in artificial intelligence and its implications for general intelligence. Understanding whether causal models are necessary for adaptation can influence future research directions in machine learning, particularly in causal representation learning and domain adaptation. This knowledge could lead to practical applications in developing more robust AI systems capable of generalizing across diverse tasks and environments, ultimately advancing our understanding of intelligence itself.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of identifying the necessary causal relationships within data and the limitations of existing models that do not explicitly incorporate causal reasoning. Naive approaches may fail because they might overlook the intricate dependencies between variables that are crucial for effective adaptation. Additionally, the theoretical and practical obstacles include the need for a comprehensive understanding of the data generating process and the ability to accurately model distributional shifts, which can be non-trivial in real-world scenarios.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either causal reasoning or adaptation separately, leading to gaps in understanding their interplay. Limitations in existing solutions include a lack of rigorous frameworks for assessing the necessity of causal models in adaptation tasks. Barriers such as insufficient empirical evidence and the complexity of causal discovery have hindered progress. This approach differs by providing a formal proof of necessity, demonstrating that robust adaptation requires learning a causal model, thus bridging the gap between causal reasoning and practical adaptation.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves proving that any agent capable of adapting to a sufficiently large set of distributional shifts must have learned a causal model of the data generating process. The approach will utilize synthetic datasets to observe the policies of regret-bounded agents under various distributional shifts. The key metrics will include regret bounds and causal identification accuracy. The expected outcomes are a formal demonstration of the necessity of causal models for robust adaptation and insights into how causal models can be learned from adaptive agents, potentially leading to advancements in causal representation learning and emergent capabilities in AI."
            }
        },
        "author_data": {
            "b96b23a2-a9c9-43e8-aa8a-2aa498825cec": {
                "pk": "b96b23a2-a9c9-43e8-aa8a-2aa498825cec",
                "name": "Jonathan Richens",
                "collaborators": [
                    "Zachary Kenton",
                    "Ramana Kumar",
                    "Sebastian Farquhar",
                    "Matt MacDermott",
                    "Tom Everitt"
                ],
                "domain": [
                    "Causal Inference",
                    "Machine Learning",
                    "Safety Analysis",
                    "Game Theory"
                ],
                "publications": [
                    {
                        "title": "Discovering Agents (Abstract Reprint)",
                        "abstract": "Causal models of agents have been used to analyse the safety aspects of machine learning systems. But identifying agents is non-trivial \u2013 often the causal model is just assumed by the modeller without much justification \u2013 and modelling failures can lead to mistakes in the safety analysis. This paper proposes the first formal causal definition of agents \u2013 roughly that agents are systems that would adapt their policy if their actions influenced the world in a different way. From this we derive the first causal discovery algorithm for discovering the presence of agents from empirical data, given a set of variables and under certain assumptions. We also provide algorithms for translating between causal models and game-theoretic influence diagrams. We demonstrate our approach by resolving some previous confusions caused by incorrect causal modelling of agents."
                    }
                ]
            },
            "ffea760b-e6ac-4d85-b299-10455af85419": {
                "pk": "ffea760b-e6ac-4d85-b299-10455af85419",
                "name": "Tom Everitt",
                "collaborators": [
                    "Matt MacDermott",
                    "Zachary Kenton",
                    "Lewis Hammond",
                    "James Fox",
                    "Ryan Carey",
                    "Alessandro Abate",
                    "M. Wooldridge",
                    "Francis Rhys Ward",
                    "Francesco Belardinelli",
                    "Francesca Toni"
                ],
                "domain": [
                    "Causal Reasoning",
                    "Game Theory",
                    "AI Ethics",
                    "Generative AI"
                ],
                "publications": [
                    {
                        "title": "Reasoning about Causality in Games (Abstract Reprint)",
                        "abstract": "Causal reasoning and game-theoretic reasoning are fundamental topics in artificial intelligence, among many other disciplines: this paper is concerned with their intersection. Despite their importance, a formal framework that supports both these forms of reasoning has, until now, been lacking. We offer a solution in the form of (structural) causal games, which can be seen as extending Pearl's causal hierarchy to the game-theoretic domain, or as extending Koller and Milch's multi-agent influence diagrams to the causal domain. We then consider three key questions: i) How can the (causal) dependencies in games \u2013 either between variables, or between strategies \u2013 be modelled in a uniform, principled manner?  ii) How may causal queries be computed in causal games, and what assumptions does this require?  iii) How do causal games compare to existing formalisms?  To address question i), we introduce mechanised games, which encode dependencies between agents' decision rules and the distributions governing the game. In response to question ii), we present definitions of predictions, interventions, and counterfactuals, and discuss the assumptions required for each. Regarding question iii), we describe correspondences between causal games and other formalisms, and explain how causal games can be used to answer queries that other causal or game-theoretic models do not support. Finally, we highlight possible applications of causal games, aided by an extensive open-source Python library."
                    },
                    {
                        "title": "The Reasons that Agents Act: Intention and Instrumental Goals",
                        "abstract": "Intention is an important and challenging concept in AI. It is important because it underlies many other concepts we care about, such as agency, manipulation, legal responsibility, and blame. However, ascribing intent to AI systems is contentious, and there is no universally accepted theory of intention applicable to AI agents. We operationalise the intention with which an agent acts, relating to the reasons it chooses its decision. We introduce a formal definition of intention in structural causal influence models, grounded in the philosophy literature on intent and applicable to real-world machine learning systems. Through a number of examples and results, we show that our definition captures the intuitive notion of intent and satisfies desiderata set-out by past work. In addition, we show how our definition relates to past concepts, including actual causality, and the notion of instrumental goals, which is a core idea in the literature on safe AI agents. Finally, we demonstrate how our definition can be used to infer the intentions of reinforcement learning agents and language models from their behaviour."
                    },
                    {
                        "title": "Discovering Agents (Abstract Reprint)",
                        "abstract": "Causal models of agents have been used to analyse the safety aspects of machine learning systems. But identifying agents is non-trivial \u2013 often the causal model is just assumed by the modeller without much justification \u2013 and modelling failures can lead to mistakes in the safety analysis. This paper proposes the first formal causal definition of agents \u2013 roughly that agents are systems that would adapt their policy if their actions influenced the world in a different way. From this we derive the first causal discovery algorithm for discovering the presence of agents from empirical data, given a set of variables and under certain assumptions. We also provide algorithms for translating between causal models and game-theoretic influence diagrams. We demonstrate our approach by resolving some previous confusions caused by incorrect causal modelling of agents."
                    },
                    {
                        "title": "A Mechanism-Based Approach to Mitigating Harms from Persuasive Generative AI",
                        "abstract": "Recent generative AI systems have demonstrated more advanced persuasive capabilities and are increasingly permeating areas of life where they can influence decision-making. Generative AI presents a new risk profile of persuasion due the opportunity for reciprocal exchange and prolonged interactions. This has led to growing concerns about harms from AI persuasion and how they can be mitigated, highlighting the need for a systematic study of AI persuasion. The current definitions of AI persuasion are unclear and related harms are insufficiently studied. Existing harm mitigation approaches prioritise harms from the outcome of persuasion over harms from the process of persuasion. In this paper, we lay the groundwork for the systematic study of AI persuasion. We first put forward definitions of persuasive generative AI. We distinguish between rationally persuasive generative AI, which relies on providing relevant facts, sound reasoning, or other forms of trustworthy evidence, and manipulative generative AI, which relies on taking advantage of cognitive biases and heuristics or misrepresenting information. We also put forward a map of harms from AI persuasion, including definitions and examples of economic, physical, environmental, psychological, sociocultural, political, privacy, and autonomy harm. We then introduce a map of mechanisms that contribute to harmful persuasion. Lastly, we provide an overview of approaches that can be used to mitigate against process harms of persuasion, including prompt engineering for manipulation classification and red teaming. Future work will operationalise these mitigations and study the interaction between different types of mechanisms of persuasion."
                    }
                ]
            }
        }
    },
    "2406.13173": {
        "paper_data": {
            "title": "Biomedical Visual Instruction Tuning with Clinician Preference Alignment",
            "url": "http://arxiv.org/abs/2406.13173v3",
            "arxiv_id": "2406.13173",
            "authors": [
                "Hejie Cui",
                "Lingjun Mao",
                "Xin Liang",
                "Jieyu Zhang",
                "Hui Ren",
                "Quanzheng Li",
                "Xiang Li",
                "Carl Yang"
            ],
            "abstract": "Recent advancements in multimodal foundation models have showcased impressive capabilities in understanding and reasoning with visual and textual information. Adapting these foundation models trained for general usage to specialized domains like biomedicine requires large-scale domain-specific instruction datasets. While existing works have explored curating such datasets automatically, the resultant datasets are not explicitly aligned with domain expertise. In this work, we propose a data-centric framework, Biomedical Visual Instruction Tuning with Clinician Preference Alignment (BioMed-VITAL), that incorporates clinician preferences into both stages of generating and selecting instruction data for tuning biomedical multimodal foundation models. First, during the generation stage, we prompt the GPT-4V generator with a diverse set of clinician-selected demonstrations for preference-aligned data candidate generation. Then, during the selection phase, we train a separate selection model, which explicitly distills clinician and policy-guided model preferences into a rating function to select high-quality data for medical instruction tuning. Results show that the model tuned with the instruction-following data from our method demonstrates a significant improvement in open visual chat (18.5% relatively) and medical VQA (win rate up to 81.73%). Our instruction-following data and models are available at BioMed-VITAL.github.io.",
            "introduction": "   1 Introduction  Recent advancements in large pre-trained multimodal models, such as GPT-4V\u00a0[1], have demonstrated impressive performance on various language and vision tasks. However, when directly applied to specialized domains like biomedicine, these models may fall short due to their primary focus on general usage rather than domain-specific expertise\u00a0[39, 27]. To bridge this gap and adapt general domain models to specialized domains, researchers have explored various techniques. Instruction tuning has emerged as a promising approach, involving the fine-tuning of large foundation models to follow explicit, natural language instructions\u00a0[38, 24, 43]. These instructions are composed of task-specific prompts and their corresponding response, enabling the models to learn and generalize to a wide range of tasks within the target domain.   Although instruction tuning has proven to be an effective method for adapting models to target domains and performing various downstream tasks, its success heavily depends on large-scale instruction-following datasets. Curating large-scale instructional datasets in specialized domains, such as biomedicine, can be expensive and time-consuming, often requiring significant domain expertise. Previous work proposes to use strong language models to generate instruction data automatically, which effectively reduces the need for extensive manual annotation\u00a0[31]. Such paradigms have successfully been adopted to adapt general domain models to biomedicine. For example, LLaVA-Med\u00a0[15] developed a framework to instruction-tune biomedical language-vision models with GPT-4 generated instruction-following data. This approach has achieved impressive performance on open-ended visual chat and visual question answering benchmarks, highlighting the potential of using model-generation data in the biomedical domain.   However, existing methods for automatically curating datasets do not explicitly incorporate clinician preferences, which may result in models producing irrelevant or impractical output, limiting their utility in real-world applications\u00a0[8]. Yet, aligning domain expertise with the process of instruction-following datasets curation is challenging. First, advanced data generators, such as GPT-4V, are often proprietary and not publicly available for alignment tuning. Second, clinician-annotated preference data in the biomedical domain is limited, further restricting effective preference learning. The combination of model opacity and data scarcity creates a significant bottleneck in developing high-quality, expert-aligned instruction-following data for instruction-tuning. This hinders the development of domain-specific models that can effectively incorporate expert preferences and requirements, ultimately limiting their practical utility and real-world impact.    To tackle this challenge, we propose an effective data-centric approach, BioMed-VITAL, that incorporates clinician preference into the process of automatically curating instruction-following data for biomedical visual instruction tuning. As shown in Figure\u00a01, BioMed-VITAL consists of three stages: (1) data generation with demonstrations, (2) data selection with a preference distilled model, and (3) visual instruction-tuning. In data generation, we strategically sample a diverse set of instructions to collect clinician preferences, which are used as demonstrations for GPT-4V-based instructional data generation, guiding the data generation toward producing more clinically relevant and useful instruction-following examples. In the data selection stage, we train a data selection model that distills a mixture of preferences from clinician-annotated and model-annotated data guided by clinician-curated criteria. This model is then used to rank the generated data samples, and the top-ranked samples are selected for visual instruction-tuning.   The contributions of this work are summarized as follows:   \u2022  We introduce a data-centric framework BioMed-VITAL, which generates and selects instruction-following data aligned with clinician preference for visual instruction tuning.",
            "references": [
                {
                    "title": "Capabilities of Gemini Models in Medicine",
                    "abstract": "Excellence in a wide variety of medical applications poses considerable challenges for AI, requiring advanced reasoning, access to up-to-date medical knowledge and understanding of complex multimodal data. Gemini models, with strong general capabilities in multimodal and long-context reasoning, offer exciting possibilities in medicine. Building on these core strengths of Gemini, we introduce Med-Gemini, a family of highly capable multimodal models that are specialized in medicine with the ability to seamlessly use web search, and that can be efficiently tailored to novel modalities using custom encoders. We evaluate Med-Gemini on 14 medical benchmarks, establishing new state-of-the-art (SoTA) performance on 10 of them, and surpass the GPT-4 model family on every benchmark where a direct comparison is viable, often by a wide margin. On the popular MedQA (USMLE) benchmark, our best-performing Med-Gemini model achieves SoTA performance of 91.1% accuracy, using a novel uncertainty-guided search strategy. On 7 multimodal benchmarks including NEJM Image Challenges and MMMU (health&medicine), Med-Gemini improves over GPT-4V by an average relative margin of 44.5%. We demonstrate the effectiveness of Med-Gemini's long-context capabilities through SoTA performance on a needle-in-a-haystack retrieval task from long de-identified health records and medical video question answering, surpassing prior bespoke methods using only in-context learning. Finally, Med-Gemini's performance suggests real-world utility by surpassing human experts on tasks such as medical text summarization, alongside demonstrations of promising potential for multimodal medical dialogue, medical research and education. Taken together, our results offer compelling evidence for Med-Gemini's potential, although further rigorous evaluation will be crucial before real-world deployment in this safety-critical domain."
                },
                {
                    "title": "A Systematic Review of Testing and Evaluation of Healthcare Applications of Large Language Models (LLMs)",
                    "abstract": "Importance: Large Language Models (LLMs) can assist in a wide range of healthcare-related activities. Current approaches to evaluating LLMs make it difficult to identify the most impactful LLM application areas. Objective: To summarize the current evaluation of LLMs in healthcare in terms of 5 components: evaluation data type, healthcare task, Natural Language Processing (NLP)/Natural Language Understanding (NLU) task, dimension of evaluation, and medical specialty. Data Sources: A systematic search of PubMed and Web of Science was performed for studies published between 01-01-2022 and 02-19-2024. Study Selection: Studies evaluating one or more LLMs in healthcare. Data Extraction and Synthesis: Three independent reviewers categorized 519 studies in terms of data used in the evaluation, the healthcare tasks (the what) and the NLP/NLU tasks (the how) examined, the dimension(s) of evaluation, and the medical specialty studied. Results: Only 5% of reviewed studies utilized real patient care data for LLM evaluation. The most popular healthcare tasks were assessing medical knowledge (e.g. answering medical licensing exam questions, 44.5%), followed by making diagnoses (19.5%), and educating patients (17.7%). Administrative tasks such as assigning provider billing codes (0.2%), writing prescriptions (0.2%), generating clinical referrals (0.6%) and clinical notetaking (0.8%) were less studied. For NLP/NLU tasks, the vast majority of studies examined question answering (84.2%). Other tasks such as summarization (8.9%), conversational dialogue (3.3%), and translation (3.1%) were infrequent. Almost all studies (95.4%) used accuracy as the primary dimension of evaluation; fairness, bias and toxicity (15.8%), robustness (14.8%), deployment considerations (4.6%), and calibration and uncertainty (1.2%) were infrequently measured. Finally, in terms of medical specialty area, most studies were in internal medicine (42%), surgery (11.4%) and ophthalmology (6.9%), with nuclear medicine (0.6%), physical medicine (0.4%) and medical genetics (0.2%) being the least represented. Conclusions and Relevance: Existing evaluations of LLMs mostly focused on accuracy of question answering for medical exams, without consideration of real patient care data. Dimensions like fairness, bias and toxicity, robustness, and deployment considerations received limited attention. To draw meaningful conclusions and improve LLM adoption, future studies need to establish a standardized set of LLM applications and evaluation dimensions, perform evaluations using data from routine care, and broaden testing to include administrative tasks as well as multiple medical specialties. Keywords: Large Language Models, Generative Artificial Intelligence, Healthcare, Dimensions of Evaluation, Evaluation Metrics."
                },
                {
                    "title": "PMC-LLaMA: toward building open-source language models for medicine",
                    "abstract": "OBJECTIVE\nRecently, large language models (LLMs) have showcased remarkable capabilities in natural language understanding. While demonstrating proficiency in everyday conversations and question-answering (QA) situations, these models frequently struggle in domains that require precision, such as medical applications, due to their lack of domain-specific knowledge. In this article, we describe the procedure for building a powerful, open-source language model specifically designed for medicine applications, termed as PMC-LLaMA.\n\n\nMATERIALS AND METHODS\nWe adapt a general-purpose LLM toward the medical domain, involving data-centric knowledge injection through the integration of 4.8M biomedical academic papers and 30K medical textbooks, as well as comprehensive domain-specific instruction fine-tuning, encompassing medical QA, rationale for reasoning, and conversational dialogues with 202M tokens.\n\n\nRESULTS\nWhile evaluating various public medical QA benchmarks and manual rating, our lightweight PMC-LLaMA, which consists of only 13B parameters, exhibits superior performance, even surpassing ChatGPT. All models, codes, and datasets for instruction tuning will be released to the research community.\n\n\nDISCUSSION\nOur contributions are 3-fold: (1) we build up an open-source LLM toward the medical domain. We believe the proposed PMC-LLaMA model can promote further development of foundation models in medicine, serving as a medical trainable basic generative language backbone; (2) we conduct thorough ablation studies to demonstrate the effectiveness of each proposed component, demonstrating how different training data and model scales affect medical LLMs; (3) we contribute a large-scale, comprehensive dataset for instruction tuning.\n\n\nCONCLUSION\nIn this article, we systematically investigate the process of building up an open-source medical-specific LLM, PMC-LLaMA."
                },
                {
                    "title": "Generalizing Visual Question Answering from Synthetic to Human-Written Questions via a Chain of QA with a Large Language Model",
                    "abstract": "Visual question answering (VQA) is a task where an image is given, and a series of questions are asked about the image. To build an efficient VQA algorithm, a large amount of QA data is required which is very expensive. Generating synthetic QA pairs based on templates is a practical way to obtain data. However, VQA models trained on those data do not perform well on complex, human-written questions. To address this issue, we propose a new method called {\\it chain of QA for human-written questions} (CoQAH). CoQAH utilizes a sequence of QA interactions between a large language model and a VQA model trained on synthetic data to reason and derive logical answers for human-written questions. We tested the effectiveness of CoQAH on two types of human-written VQA datasets for 3D-rendered and chest X-ray images and found that it achieved state-of-the-art accuracy in both types of data. Notably, CoQAH outperformed general vision-language models, VQA models, and medical foundation models with no finetuning."
                },
                {
                    "title": "Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models",
                    "abstract": "Harnessing the power of human-annotated data through Supervised Fine-Tuning (SFT) is pivotal for advancing Large Language Models (LLMs). In this paper, we delve into the prospect of growing a strong LLM out of a weak one without the need for acquiring additional human-annotated data. We propose a new fine-tuning method called Self-Play fIne-tuNing (SPIN), which starts from a supervised fine-tuned model. At the heart of SPIN lies a self-play mechanism, where the LLM refines its capability by playing against instances of itself. More specifically, the LLM generates its own training data from its previous iterations, refining its policy by discerning these self-generated responses from those obtained from human-annotated data. Our method progressively elevates the LLM from a nascent model to a formidable one, unlocking the full potential of human-annotated demonstration data for SFT. Theoretically, we prove that the global optimum to the training objective function of our method is achieved only when the LLM policy aligns with the target data distribution. Empirically, we evaluate our method on several benchmark datasets including the HuggingFace Open LLM Leaderboard, MT-Bench, and datasets from Big-Bench. Our results show that SPIN can significantly improve the LLM's performance across a variety of benchmarks and even outperform models trained through direct preference optimization (DPO) supplemented with extra GPT-4 preference data. This sheds light on the promise of self-play, enabling the achievement of human-level performance in LLMs without the need for expert opponents. Codes are available at https://github.com/uclaml/SPIN."
                },
                {
                    "title": "Improved Baselines with Visual Instruction Tuning",
                    "abstract": "Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this paper, we present the first systematic study to investigate the design choices of LMMs in a controlled setting under the LLaVA framework. We show that the fully-connected vision-language connector in LLaVA is surprisingly power-ful and data-efficient. With simple modifications to LLa VA, namely, using CLIP- ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~ 1 day on a single 8-AI00 node. Furthermore, we present some early exploration of open problems in LMMs, including scaling to higher resolution inputs, compositional capabilities, and model hallucination, etc. We hope this makes state-of-the-art LMM research more accessible. Code and model will be publicly available."
                },
                {
                    "title": "Improving Automatic VQA Evaluation Using Large Language Models",
                    "abstract": "8 years after the visual question answering (VQA) task was proposed, accuracy remains the primary metric for automatic evaluation. VQA Accuracy has been effective so far in the IID evaluation setting. However, our community is undergoing a shift towards open-ended generative models and OOD evaluation. In this new paradigm, the existing VQA Accuracy metric is overly stringent and underestimates the performance of VQA systems. Thus, there is a need to develop more robust automatic VQA metrics that serve as a proxy for human judgment. In this work, we propose to leverage the in-context learning capabilities of instruction-tuned large language models (LLMs) to build a better VQA metric. We formulate VQA evaluation as an answer-rating task where the LLM is instructed to score the accuracy of a candidate answer given a set of reference answers. We demonstrate the proposed metric better correlates with human judgment compared to existing metrics across several VQA models and benchmarks. We hope wide adoption of our metric will contribute to better estimating the research progress on the VQA task. We plan to release the evaluation code and collected human judgments."
                },
                {
                    "title": "Large Language Models in Medicine: The Potentials and Pitfalls",
                    "abstract": "Large language models (LLMs) are artificial intelligence models trained on vast text data to generate humanlike outputs. They have been applied to various tasks in health care, ranging from answering medical examination questions to generating clinical reports. With increasing institutional partnerships between companies producing LLMs and health systems, the real-world clinical application of these models is nearing realization. As these models gain traction, health care practitioners must understand what LLMs are, their development, their current and potential applications, and the associated pitfalls in a medical setting. This review, coupled with a tutorial, provides a comprehensive yet accessible overview of these areas with the aim of familiarizing health care professionals with the rapidly changing landscape of LLMs in medicine. Furthermore, the authors highlight active research areas in the field that promise to improve LLMs' usability in health care contexts."
                },
                {
                    "title": "MedAlign: A Clinician-Generated Dataset for Instruction Following with Electronic Medical Records",
                    "abstract": "The ability of large language models (LLMs) to follow natural language instructions with human-level fluency suggests many opportunities in healthcare to reduce administrative burden and improve quality of care. However, evaluating LLMs on realistic text generation tasks for healthcare remains challenging. Existing question answering datasets for electronic health record (EHR) data fail to capture the complexity of information needs and documentation burdens experienced by clinicians. To address these challenges, we introduce MedAlign, a benchmark dataset of 983 natural language instructions for EHR data. MedAlign is curated by 15 clinicians (7 specialities), includes clinician-written reference responses for 303 instructions, and provides 276 longitudinal EHRs for grounding instruction-response pairs. We used MedAlign to evaluate 6 general domain LLMs, having clinicians rank the accuracy and quality of each LLM response. We found high error rates, ranging from 35% (GPT-4) to 68% (MPT-7B-Instruct), and 8.3% drop in accuracy moving from 32k to 2k context lengths for GPT-4. Finally, we report correlations between clinician rankings and automated natural language generation metrics as a way to rank LLMs without human review. We make MedAlign available under a research data use agreement to enable LLM evaluations on tasks aligned with clinician needs and preferences."
                },
                {
                    "title": "Instruction Tuning for Large Language Models: A Survey",
                    "abstract": "This paper surveys research works in the quickly advancing field of instruction tuning (IT), a crucial technique to enhance the capabilities and controllability of large language models (LLMs). Instruction tuning refers to the process of further training LLMs on a dataset consisting of \\textsc{(instruction, output)} pairs in a supervised fashion, which bridges the gap between the next-word prediction objective of LLMs and the users' objective of having LLMs adhere to human instructions. In this work, we make a systematic review of the literature, including the general methodology of IT, the construction of IT datasets, the training of IT models, and applications to different modalities, domains and applications, along with an analysis on aspects that influence the outcome of IT (e.g., generation of instruction outputs, size of the instruction dataset, etc). We also review the potential pitfalls of IT along with criticism against it, along with efforts pointing out current deficiencies of existing strategies and suggest some avenues for fruitful research. Project page: github.com/xiaoya-li/Instruction-Tuning-Survey"
                },
                {
                    "title": "Towards Generalist Biomedical AI",
                    "abstract": "Medicine is inherently multimodal, with rich data modalities spanning text, imaging, genomics, and more. Generalist biomedical artificial intelligence (AI) systems that flexibly encode, integrate, and interpret this data at scale can potentially enable impactful applications ranging from scientific discovery to care delivery. To enable the development of these models, we first curate MultiMedBench, a new multimodal biomedical benchmark. MultiMedBench encompasses 14 diverse tasks such as medical question answering, mammography and dermatology image interpretation, radiology report generation and summarization, and genomic variant calling. We then introduce Med-PaLM Multimodal (Med-PaLM M), our proof of concept for a generalist biomedical AI system. Med-PaLM M is a large multimodal generative model that flexibly encodes and interprets biomedical data including clinical language, imaging, and genomics with the same set of model weights. Med-PaLM M reaches performance competitive with or exceeding the state of the art on all MultiMedBench tasks, often surpassing specialist models by a wide margin. We also report examples of zero-shot generalization to novel medical concepts and tasks, positive transfer learning across tasks, and emergent zero-shot medical reasoning. To further probe the capabilities and limitations of Med-PaLM M, we conduct a radiologist evaluation of model-generated (and human) chest X-ray reports and observe encouraging performance across model scales. In a side-by-side ranking on 246 retrospective chest X-rays, clinicians express a pairwise preference for Med-PaLM M reports over those produced by radiologists in up to 40.50% of cases, suggesting potential clinical utility. While considerable work is needed to validate these models in real-world use cases, our results represent a milestone towards the development of generalist biomedical AI systems."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Masked Vision and Language Pre-training with Unimodal and Multimodal Contrastive Losses for Medical Visual Question Answering",
                    "abstract": "Medical visual question answering (VQA) is a challenging task that requires answering clinical questions of a given medical image, by taking consider of both visual and language information. However, due to the small scale of training data for medical VQA, pre-training fine-tuning paradigms have been a commonly used solution to improve model generalization performance. In this paper, we present a novel self-supervised approach that learns unimodal and multimodal feature representations of input images and text using medical image caption datasets, by leveraging both unimodal and multimodal contrastive losses, along with masked language modeling and image text matching as pretraining objectives. The pre-trained model is then transferred to downstream medical VQA tasks. The proposed approach achieves state-of-the-art (SOTA) performance on three publicly available medical VQA datasets with significant accuracy improvements of 2.2%, 14.7%, and 1.7% respectively. Besides, we conduct a comprehensive analysis to validate the effectiveness of different components of the approach and study different pre-training settings. Our codes and models are available at https://github.com/pengfeiliHEU/MUMC."
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "LLaVA-Med: Training a Large Language-and-Vision Assistant for Biomedicine in One Day",
                    "abstract": "Conversational generative AI has demonstrated remarkable promise for empowering biomedical practitioners, but current investigations focus on unimodal text. Multimodal conversational AI has seen rapid progress by leveraging billions of image-text pairs from the public web, but such general-domain vision-language models still lack sophistication in understanding and conversing about biomedical images. In this paper, we propose a cost-efficient approach for training a vision-language conversational assistant that can answer open-ended research questions of biomedical images. The key idea is to leverage a large-scale, broad-coverage biomedical figure-caption dataset extracted from PubMed Central, use GPT-4 to self-instruct open-ended instruction-following data from the captions, and then fine-tune a large general-domain vision-language model using a novel curriculum learning method. Specifically, the model first learns to align biomedical vocabulary using the figure-caption pairs as is, then learns to master open-ended conversational semantics using GPT-4 generated instruction-following data, broadly mimicking how a layperson gradually acquires biomedical knowledge. This enables us to train a Large Language and Vision Assistant for BioMedicine (LLaVA-Med) in less than 15 hours (with eight A100s). LLaVA-Med exhibits excellent multimodal conversational capability and can follow open-ended instruction to assist with inquiries about a biomedical image. On three standard biomedical visual question answering datasets, LLaVA-Med outperforms previous supervised state-of-the-art on certain metrics. To facilitate biomedical multimodal research, we will release our instruction-following data and the LLaVA-Med model."
                },
                {
                    "title": "Enabling Large Language Models to Generate Text with Citations",
                    "abstract": "Large language models (LLMs) have emerged as a widely-used tool for information seeking, but their generated outputs are prone to hallucination. In this work, our aim is to allow LLMs to generate text with citations, improving their factual correctness and verifiability. Existing work mainly relies on commercial search engines and human evaluation, making it challenging to reproduce and compare different modeling approaches. We propose ALCE, the first benchmark for Automatic LLMs' Citation Evaluation. ALCE collects a diverse set of questions and retrieval corpora and requires building end-to-end systems to retrieve supporting evidence and generate answers with citations. We develop automatic metrics along three dimensions -- fluency, correctness, and citation quality -- and demonstrate their strong correlation with human judgements. Our experiments with state-of-the-art LLMs and novel prompting strategies show that current systems have considerable room for improvement -- For example, on the ELI5 dataset, even the best models lack complete citation support 50% of the time. Our analyses further highlight promising future directions, including developing better retrievers, advancing long-context LLMs, and improving the ability to synthesize information from multiple sources."
                },
                {
                    "title": "Album Storytelling with Iterative Story-aware Captioning and Large Language Models",
                    "abstract": "This work studies how to transform an album to vivid and coherent stories, a task we refer to as\"album storytelling\". While this task can help preserve memories and facilitate experience sharing, it remains an underexplored area in current literature. With recent advances in Large Language Models (LLMs), it is now possible to generate lengthy, coherent text, opening up the opportunity to develop an AI assistant for album storytelling. One natural approach is to use caption models to describe each photo in the album, and then use LLMs to summarize and rewrite the generated captions into an engaging story. However, we find this often results in stories containing hallucinated information that contradicts the images, as each generated caption (\"story-agnostic\") is not always about the description related to the whole story or miss some necessary information. To address these limitations, we propose a new iterative album storytelling pipeline. Specifically, we start with an initial story and build a story-aware caption model to refine the captions using the whole story as guidance. The polished captions are then fed into the LLMs to generate a new refined story. This process is repeated iteratively until the story contains minimal factual errors while maintaining coherence. To evaluate our proposed pipeline, we introduce a new dataset of image collections from vlogs and a set of systematic evaluation metrics. Our results demonstrate that our method effectively generates more accurate and engaging stories for albums, with enhanced coherence and vividness."
                },
                {
                    "title": "PMC-VQA: Visual Instruction Tuning for Medical Visual Question Answering",
                    "abstract": "Medical Visual Question Answering (MedVQA) presents a significant opportunity to enhance diagnostic accuracy and healthcare delivery by leveraging artificial intelligence to interpret and answer questions based on medical images. In this study, we reframe the problem of MedVQA as a generation task that naturally follows the human-machine interaction and propose a generative-based model for medical visual understanding by aligning visual information from a pre-trained vision encoder with a large language model. We establish a scalable pipeline to construct a large-scale medical visual question-answering dataset, named PMC-VQA, which contains 227k VQA pairs of 149k images that cover various modalities or diseases. We train the proposed model on PMC-VQA and then fine-tune it on multiple public benchmarks, e.g., VQA-RAD, SLAKE, and Image-Clef-2019, significantly outperforming existing MedVQA models in generating relevant, accurate free-form answers. In addition, we propose a test set that has undergone manual verification, which is significantly more challenging, serving to better monitor the development of generative MedVQA methods. To facilitate comprehensive evaluation and comparison, we have maintained a leaderboard at https://paperswithcode.com/paper/pmc-vqa-visual-instruction-tuning-for-medical, offering a centralized resource for tracking progress and benchmarking state-of-the-art approaches. The PMC-VQA dataset emerges as a vital resource for the field of research, and the MedVInT presents a significant breakthrough in the area of MedVQA."
                },
                {
                    "title": "Tree of Thoughts: Deliberate Problem Solving with Large Language Models",
                    "abstract": "Language models are increasingly being deployed for general problem solving across a wide range of tasks, but are still confined to token-level, left-to-right decision-making processes during inference. This means they can fall short in tasks that require exploration, strategic lookahead, or where initial decisions play a pivotal role. To surmount these challenges, we introduce a new framework for language model inference, Tree of Thoughts (ToT), which generalizes over the popular Chain of Thought approach to prompting language models, and enables exploration over coherent units of text (thoughts) that serve as intermediate steps toward problem solving. ToT allows LMs to perform deliberate decision making by considering multiple different reasoning paths and self-evaluating choices to decide the next course of action, as well as looking ahead or backtracking when necessary to make global choices. Our experiments show that ToT significantly enhances language models' problem-solving abilities on three novel tasks requiring non-trivial planning or search: Game of 24, Creative Writing, and Mini Crosswords. For instance, in Game of 24, while GPT-4 with chain-of-thought prompting only solved 4% of tasks, our method achieved a success rate of 74%. Code repo with all prompts: https://github.com/princeton-nlp/tree-of-thought-llm."
                },
                {
                    "title": "Text Classification via Large Language Models",
                    "abstract": "Despite the remarkable success of large-scale Language Models (LLMs) such as GPT-3, their performances still significantly underperform fine-tuned models in the task of text classification. This is due to (1) the lack of reasoning ability in addressing complex linguistic phenomena (e.g., intensification, contrast, irony etc); (2) limited number of tokens allowed in in-context learning. In this paper, we introduce Clue And Reasoning Prompting (CARP). CARP adopts a progressive reasoning strategy tailored to addressing the complex linguistic phenomena involved in text classification: CARP first prompts LLMs to find superficial clues (e.g., keywords, tones, semantic relations, references, etc), based on which a diagnostic reasoning process is induced for final decisions. To further address the limited-token issue, CARP uses a fine-tuned model on the supervised dataset for $k$NN demonstration search in the in-context learning, allowing the model to take the advantage of both LLM's generalization ability and the task-specific evidence provided by the full labeled dataset. Remarkably, CARP yields new SOTA performances on 4 out of 5 widely-used text-classification benchmarks, 97.39 (+1.24) on SST-2, 96.40 (+0.72) on AGNews, 98.78 (+0.25) on R8 and 96.95 (+0.6) on R52, and a performance comparable to SOTA on MR (92.39 v.s. 93.3). More importantly, we find that CARP delivers impressive abilities on low-resource and domain-adaptation setups. Specifically, using 16 examples per class, CARP achieves comparable performances to supervised models with 1,024 examples per class."
                },
                {
                    "title": "GPT-RE: In-context Learning for Relation Extraction using Large Language Models",
                    "abstract": "In spite of the potential for ground-breaking achievements offered by large language models (LLMs) (e.g., GPT-3), they still lag significantly behind fully-supervised baselines (e.g., fine-tuned BERT) in relation extraction (RE). This is due to the two major shortcomings of LLMs in RE: (1) low relevance regarding entity and relation in retrieved demonstrations for in-context learning; and (2) the strong inclination to wrongly classify NULL examples into other pre-defined labels. In this paper, we propose GPT-RE to bridge the gap between LLMs and fully-supervised baselines. GPT-RE successfully addresses the aforementioned issues by (1) incorporating task-specific entity representations in demonstration retrieval; and (2) enriching the demonstrations with gold label-induced reasoning logic. We evaluate GPT-RE on four widely-used RE datasets, and observe that GPT-RE achieves improvements over not only existing GPT-3 baselines, but also fully-supervised baselines. Specifically, GPT-RE achieves SOTA performances on the Semeval and SciERC datasets, and competitive performances on the TACRED and ACE05 datasets."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "Instruction Tuning with GPT-4",
                    "abstract": "Prior work has shown that finetuning large language models (LLMs) using machine-generated instruction-following data enables such models to achieve remarkable zero-shot capabilities on new tasks, and no human-written instructions are needed. In this paper, we present the first attempt to use GPT-4 to generate instruction-following data for LLM finetuning. Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by previous state-of-the-art models. We also collect feedback and comparison data from GPT-4 to enable a comprehensive evaluation and reward model training. We make our data generated using GPT-4 as well as our codebase publicly available."
                },
                {
                    "title": "TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering",
                    "abstract": "Despite thousands of researchers, engineers, and artists actively working on improving text-to-image generation models, systems often fail to produce images that accurately align with the text inputs. We introduce TIFA (Text-to-Image Faithfulness evaluation with question Answering), an automatic evaluation metric that measures the faithfulness of a generated image to its text input via visual question answering (VQA). Specifically, given a text input, we automatically generate several question-answer pairs using a language model. We calculate image faithfulness by checking whether existing VQA models can answer these questions using the generated image. TIFA is a reference-free metric that allows for fine-grained and interpretable evaluations of generated images. TIFA also has better correlations with human judgments than existing metrics. Based on this approach, we introduce TIFA v1.0, a benchmark consisting of 4K diverse text inputs and 25K questions across 12 categories (object, counting, etc.). We present a comprehensive evaluation of existing text-to-image models using TIFA v1.0 and highlight the limitations and challenges of current models. For instance, we find that current text-to-image models, despite doing well on color and material, still struggle in counting, spatial relations, and composing multiple objects. We hope our benchmark will help carefully measure the research progress in text-to-image synthesis and provide valuable insights for further research. 1"
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "PMC-CLIP: Contrastive Language-Image Pre-training using Biomedical Documents",
                    "abstract": "Foundation models trained on large-scale dataset gain a recent surge in CV and NLP. In contrast, development in biomedical domain lags far behind due to data scarcity. To address this issue, we build and release PMC-OA, a biomedical dataset with 1.6M image-caption pairs collected from PubMedCentral's OpenAccess subset, which is 8 times larger than before. PMC-OA covers diverse modalities or diseases, with majority of the image-caption samples aligned at finer-grained level, i.e., subfigure and subcaption. While pretraining a CLIP-style model on PMC-OA, our model named PMC-CLIP achieves state-of-the-art results on various downstream tasks, including image-text retrieval on ROCO, MedMNIST image classification, Medical VQA, i.e. +8.1% R@10 on image-text retrieval, +3.9% accuracy on image classification."
                },
                {
                    "title": "Open-Ended Medical Visual Question Answering Through Prefix Tuning of Language Models",
                    "abstract": "Medical Visual Question Answering (VQA) is an important challenge, as it would lead to faster and more accurate diagnoses and treatment decisions. Most existing methods approach it as a multi-class classification problem, which restricts the outcome to a predefined closed-set of curated answers. We focus on open-ended VQA and motivated by the recent advances in language models consider it as a generative task. Leveraging pre-trained language models, we introduce a novel method particularly suited for small, domain-specific, medical datasets. To properly communicate the medical images to the language model, we develop a network that maps the extracted visual features to a set of learnable tokens. Then, alongside the question, these learnable tokens directly prompt the language model. We explore recent parameter-efficient fine-tuning strategies for language models, which allow for resource- and data-efficient fine-tuning. We evaluate our approach on the prime medical VQA benchmarks, namely, Slake, OVQA and PathVQA. The results demonstrate that our approach outperforms existing methods across various training settings while also being computationally efficient."
                },
                {
                    "title": "BiomedCLIP: a multimodal biomedical foundation model pretrained from fifteen million scientific image-text pairs",
                    "abstract": "Biomedical data is inherently multimodal, comprising physical measurements and natural language narratives. A generalist biomedical AI model needs to simultaneously process different modalities of data, including text and images. Therefore, training an effective generalist biomedical model requires high-quality multimodal data, such as parallel image-text pairs. Here, we present PMC-15M, a novel dataset that is two orders of magnitude larger than existing biomedical multimodal datasets such as MIMIC-CXR, and spans a diverse range of biomedical image types. PMC-15M contains 15 million biomedical image-text pairs collected from 4.4 million scientific articles. Based on PMC-15M, we have pretrained BiomedCLIP, a multimodal foundation model, with domain-specific adaptations tailored to biomedical vision-language processing. We conducted extensive experiments and ablation studies on standard biomedical imaging tasks from retrieval to classification to visual question-answering (VQA). BiomedCLIP achieved new state-of-the-art results in a wide range of standard datasets, substantially outperforming prior approaches. Intriguingly, by large-scale pretraining on diverse biomedical image types, BiomedCLIP even outperforms state-of-the-art radiology-specific models such as BioViL in radiology-specific tasks such as RSNA pneumonia detection. In summary, BiomedCLIP is a fully open-access foundation model that achieves state-of-the-art performance on various biomedical tasks, paving the way for transformative multimodal biomedical discovery and applications. We release our models at https://aka.ms/biomedclip to facilitate future research in multimodal biomedical AI."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Self-Instruct: Aligning Language Models with Self-Generated Instructions",
                    "abstract": "Large \u201cinstruction-tuned\u201d language models (i.e., finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is often limited in quantity, diversity, and creativity, therefore hindering the generality of the tuned model. We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off their own generations. Our pipeline generates instructions, input, and output samples from a language model, then filters invalid or similar ones before using them to finetune the original model. Applying our method to the vanilla GPT3, we demonstrate a 33% absolute improvement over the original model on Super-NaturalInstructions, on par with the performance of InstructGPT-001, which was trained with private user data and human annotations. For further evaluation, we curate a set of expert-written instructions for novel tasks, and show through human evaluation that tuning GPT3 with Self-Instruct outperforms using existing public instruction datasets by a large margin, leaving only a 5% absolute gap behind InstructGPT-001. Self-Instruct provides an almost annotation-free method for aligning pre-trained language models with instructions, and we release our large synthetic dataset to facilitate future studies on instruction tuning."
                },
                {
                    "title": "Self-Supervised Vision-Language Pretraining for Medial Visual Question Answering",
                    "abstract": "Medical image visual question answering (VQA) is a task to answer clinical questions, given a radiographic image, which is a challenging problem that requires a model to integrate both vision and language information. Due to the limited size of the training data in medical VQA tasks, pretrain-finetune paradigm is a commonly used solution to improve the model generalization. In this paper, we propose a self-supervised method that applies Masked image modeling, Masked language modeling, Image text matching and Image text alignment via contrastive learning (M2I2) for pretraining on medical image caption dataset, and finetunes to downstream medical VQA tasks. The proposed method achieves state-of-the-art performance on all the three public medical VQA datasets. Our codes and models are available at https://github.com/pengfeiliHEU/M2I2."
                },
                {
                    "title": "Scaling Instruction-Finetuned Language Models",
                    "abstract": "Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models."
                },
                {
                    "title": "Finetuned Language Models Are Zero-Shot Learners",
                    "abstract": "This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning."
                },
                {
                    "title": "Cross-Task Generalization via Natural Language Crowdsourcing Instructions",
                    "abstract": "Humans (e.g., crowdworkers) have a remarkable ability in solving different tasks, by simply reading textual instructions that define them and looking at a few examples. Despite the success of the conventional supervised learning on individual datasets, such models often struggle with generalization across tasks (e.g., a question-answering system cannot solve classification tasks). A long-standing challenge in AI is to build a model that learns a new task by understanding the human-readable instructions that define it. To study this, we introduce NATURAL INSTRUCTIONS, a dataset of 61 distinct tasks, their human-authored instructions, and 193k task instances (input-output pairs). The instructions are obtained from crowdsourcing instructions used to create existing NLP datasets and mapped to a unified schema. Using this meta-dataset, we measure cross-task generalization by training models on seen tasks and measuring generalization to the remaining unseen ones. We adopt generative pre-trained language models to encode task-specific instructions along with input and generate task output. Our results indicate that models benefit from instructions when evaluated in terms of generalization to unseen tasks (19% better for models utilizing instructions). These models, however, are far behind an estimated performance upperbound indicating significant room for more progress in this direction."
                },
                {
                    "title": "Calibrate Before Use: Improving Few-Shot Performance of Language Models",
                    "abstract": "GPT-3 can perform numerous tasks when provided a natural language prompt that contains a few training examples. We show that this type of few-shot learning can be unstable: the choice of prompt format, training examples, and even the order of the training examples can cause accuracy to vary from near chance to near state-of-the-art. We demonstrate that this instability arises from the bias of language models towards predicting certain answers, e.g., those that are placed near the end of the prompt or are common in the pre-training data. To mitigate this, we first estimate the model's bias towards each answer by asking for its prediction when given the training prompt and a content-free test input such as \"N/A\". We then fit calibration parameters that cause the prediction for this input to be uniform across answers. On a diverse set of tasks, this contextual calibration procedure substantially improves GPT-3 and GPT-2's average accuracy (up to 30.0% absolute) and reduces variance across different choices of the prompt."
                },
                {
                    "title": "Slake: A Semantically-Labeled Knowledge-Enhanced Dataset For Medical Visual Question Answering",
                    "abstract": "Medical visual question answering (Med-VQA) has tremendous potential in healthcare. However, the development of this technology is hindered by the lacking of publicly-available and high-quality labeled datasets for training and evaluation. In this paper, we present a large bilingual dataset, SLAKE, with comprehensive semantic labels annotated by experienced physicians and a new structural medical knowledge base for Med-VQA. Besides, SLAKE includes richer modalities and covers more human body parts than the currently available dataset. We show that SLAKE can be used to facilitate the development and evaluation of Med-VQA systems. The dataset can be downloaded from http://www.med-vqa.com/slake."
                },
                {
                    "title": "PathVQA: 30000+ Questions for Medical Visual Question Answering",
                    "abstract": "Is it possible to develop an \"AI Pathologist\" to pass the board-certified examination of the American Board of Pathology? To achieve this goal, the first step is to create a visual question answering (VQA) dataset where the AI agent is presented with a pathology image together with a question and is asked to give the correct answer. Our work makes the first attempt to build such a dataset. Different from creating general-domain VQA datasets where the images are widely accessible and there are many crowdsourcing workers available and capable of generating question-answer pairs, developing a medical VQA dataset is much more challenging. First, due to privacy concerns, pathology images are usually not publicly available. Second, only well-trained pathologists can understand pathology images, but they barely have time to help create datasets for AI research. To address these challenges, we resort to pathology textbooks and online digital libraries. We develop a semi-automated pipeline to extract pathology images and captions from textbooks and generate question-answer pairs from captions using natural language processing. We collect 32,799 open-ended questions from 4,998 pathology images where each question is manually checked to ensure correctness. To our best knowledge, this is the first dataset for pathology VQA. Our dataset will be released publicly to promote research in medical VQA."
                },
                {
                    "title": "PubMedCLIP: How Much Does CLIP Benefit Visual Question Answering in the Medical Domain?",
                    "abstract": "Contrastive Language\u2013Image Pre-training (CLIP) has shown remarkable success in learning with cross-modal supervision from extensive amounts of image\u2013text pairs collected online. Thus far, the effectiveness of CLIP has been investigated primarily in general-domain multimodal problems. In this work, we evaluate the effectiveness of CLIP for the task of Medical Visual Question Answering (MedVQA). We present PubMedCLIP, a fine-tuned version of CLIP for the medical domain based on PubMed articles. Our experiments conducted on two MedVQA benchmark datasets illustrate that PubMedCLIP achieves superior results improving the overall accuracy up to 3% in comparison to the state-of-the-art Model-Agnostic Meta-Learning (MAML) networks pre-trained only on visual data. The PubMedCLIP model with different back-ends, the source code for pre-training them and reproducing our MedVQA pipeline is publicly available at https://github.com/sarahESL/PubMedCLIP."
                },
                {
                    "title": "Descriptor : A dataset of clinically generated visual questions and answers about radiology images",
                    "abstract": "Radiology images are an essential part of clinical decision making and population screening, e.g., for cancer. Automated systems could help clinicians cope with large amounts of images by answering questions about the image contents. An emerging area of artificial intelligence, Visual Question Answering (VQA) in the medical domain explores approaches to this form of clinical decision support. Success of such machine learning tools hinges on availability and design of collections composed of medical images augmented with question-answer pairs directed at the content of the image. We introduce VQA-RAD, the first manually constructed dataset where clinicians asked naturally occurring questions about radiology images and provided reference answers. Manual categorization of images and questions provides insight into clinically relevant tasks and the natural language to phrase them. Evaluating with well-known algorithms, we demonstrate the rich quality of this dataset over other automatically constructed ones. We propose VQARAD to encourage the community to design VQA tools with the goals of improving patient care."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI",
                "cs.CL",
                "cs.LG",
                "68T50, 68T45, 68T37, 68T05, 68T07, 68T09,",
                "I.2.7; I.2.6; I.2.10"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively incorporate clinician preferences into the automatic curation of instruction-following datasets for biomedical visual instruction tuning?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of biomedical machine learning, as it addresses the gap between general-purpose models and the specialized needs of the biomedical domain. By aligning instruction-following datasets with clinician preferences, we can enhance the relevance and practicality of model outputs, leading to improved decision-making in clinical settings. This research could pave the way for more effective domain-specific models, ultimately advancing knowledge in biomedical applications and facilitating the development of tools that better serve healthcare professionals.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of aligning advanced language models with nuanced clinician preferences. Naive approaches may fail because they do not account for the specific needs and context of biomedical applications, leading to irrelevant or impractical outputs. Additionally, the lack of publicly available advanced data generators and the scarcity of clinician-annotated preference data create significant technical and practical obstacles. These factors contribute to the difficulty of developing high-quality, expert-aligned instruction-following datasets necessary for effective instruction tuning.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely overlooked the integration of clinician preferences in the dataset curation process, focusing instead on general instruction-following data. Barriers such as the proprietary nature of advanced data generators and the limited availability of clinician-annotated data have hindered progress. Existing methods do not adequately address the need for expert alignment, resulting in models that may not meet real-world clinical requirements. Our approach differs by explicitly incorporating clinician preferences into the data generation and selection process, thereby improving the relevance and utility of the instruction-following datasets.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, BioMed-VITAL, consists of three key stages: (1) **Data Generation with Demonstrations**: We sample a diverse set of instructions to collect clinician preferences, which guide the generation of instruction-following data using GPT-4V. (2) **Data Selection with a Preference Distilled Model**: We train a model that distills clinician preferences from both clinician-annotated and model-annotated data, ranking the generated samples based on these preferences. (3) **Visual Instruction-Tuning**"
            }
        },
        "author_data": {
            "cbf5b012-680e-46b8-8210-0f8a3eece237": {
                "pk": "cbf5b012-680e-46b8-8210-0f8a3eece237",
                "name": "Hejie Cui",
                "collaborators": [
                    "Carl Yang",
                    "Xuan Kan",
                    "Ying Guo",
                    "Yanqiao Zhu",
                    "Lifang He",
                    "Wei Dai",
                    "Ran Xu",
                    "Yi Yang",
                    "Xiaoxiao Li",
                    "Xinyu Fang"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Neuroimaging",
                    "Machine Learning",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "How Can Graph Neural Networks Help Document Retrieval: A Case Study on CORD19 with Concept Map Generation",
                        "abstract": "Graph neural networks (GNNs), as a group of powerful tools for representation learning on irregular data, have manifested superiority in various downstream tasks. With unstructured texts represented as concept maps, GNNs can be exploited for tasks like document retrieval. Intrigued by how can GNNs help document retrieval, we conduct an empirical study on a large-scale multi-discipline dataset CORD-19. Results show that instead of the complex structure-oriented GNNs such as GINs and GATs, our proposed semantics-oriented graph functions achieve better and more stable performance based on the BM25 retrieved candidates. Our insights in this case study can serve as a guideline for future work to develop effective GNNs with appropriate semantics-oriented inductive biases for textual reasoning tasks like document retrieval and classification. All code for this case study is available at https://github.com/HennyJie/GNN-DocRetrieval."
                    },
                    {
                        "title": "Effective and Interpretable fMRI Analysis via Functional Brain Network Generation",
                        "abstract": "Recent studies in neuroscience show great potential of functional brain networks constructed from fMRI data for popularity modeling and clinical predictions. However, existing functional brain networks are noisy and unaware of downstream prediction tasks, while also incompatible with recent powerful machine learning models of GNNs. In this work, we develop an end-to-end trainable pipeline to extract prominent fMRI features, generate brain networks, and make predictions with GNNs, all under the guidance of downstream prediction tasks. Preliminary experiments on the PNC fMRI data show the superior effectiveness and unique interpretability of our framework."
                    },
                    {
                        "title": "Pulmonary Vessel Segmentation based on Orthogonal Fused U-Net++ of Chest CT Images",
                        "abstract": "Pulmonary vessel segmentation is important for clinical diagnosis of pulmonary diseases, while is also challenging due to the complicated structure. In this work, we present an effective framework and refinement process of pulmonary vessel segmentation from chest computed tomographic (CT) images. The key to our approach is a 2.5D segmentation network applied from three orthogonal axes, which presents a robust and fully automated pulmonary vessel segmentation result with lower network complexity and memory usage compared to 3D networks. The slice radius is introduced to convolve the adjacent information of the center slice and the multi-planar fusion optimizes the presentation of intra- and inter- slice features. Besides, the tree-like structure of the pulmonary vessel is extracted in the post-processing process, which is used for segmentation refining and pruning. In the evaluation experiments, three fusion methods are tested and the most promising one is compared with the state-of-the-art 2D and 3D structures on 300 cases of lung images randomly selected from LIDC dataset. Our method outperforms other network structures by a large margin and achieves by far the highest average DICE score of 0.9272 and precision of 0.9310, as per our knowledge from the pulmonary vessel segmentation models available in the literature."
                    },
                    {
                        "title": "Zero-Shot Scene Graph Relation Prediction through Commonsense Knowledge Integration",
                        "abstract": "Relation prediction among entities in images is an important step in scene graph generation (SGG), which further impacts various visual understanding and reasoning tasks. Existing SGG frameworks, however, require heavy training yet are incapable of modeling unseen (i.e.,zero-shot) triplets. In this work, we stress that such incapability is due to the lack of commonsense reasoning,i.e., the ability to associate similar entities and infer similar relations based on general understanding of the world. To fill this gap, we propose CommOnsense-integrAted sCenegrapHrElation pRediction (COACHER), a framework to integrate commonsense knowledge for SGG, especially for zero-shot relation prediction. Specifically, we develop novel graph mining pipelines to model the neighborhoods and paths around entities in an external commonsense knowledge graph, and integrate them on top of state-of-the-art SGG frameworks. Extensive quantitative evaluations and qualitative case studies on both original and manipulated datasets from Visual Genome demonstrate the effectiveness of our proposed approach."
                    },
                    {
                        "title": "PTGB: Pre-Train Graph Neural Networks for Brain Network Analysis",
                        "abstract": "The human brain is the central hub of the neurobiological system, controlling behavior and cognition in complex ways. Recent advances in neuroscience and neuroimaging analysis have shown a growing interest in the interactions between brain regions of interest (ROIs) and their impact on neural development and disorder diagnosis. As a powerful deep model for analyzing graph-structured data, Graph Neural Networks (GNNs) have been applied for brain network analysis. However, training deep models requires large amounts of labeled data, which is often scarce in brain network datasets due to the complexities of data acquisition and sharing restrictions. To make the most out of available training data, we propose PTGB, a GNN pre-training framework that captures intrinsic brain network structures, regardless of clinical outcomes, and is easily adaptable to various downstream tasks. PTGB comprises two key components: (1) an unsupervised pre-training technique designed specifically for brain networks, which enables learning from large-scale datasets without task-specific labels; (2) a data-driven parcellation atlas mapping pipeline that facilitates knowledge transfer across datasets with different ROI systems. Extensive evaluations using various GNN models have demonstrated the robust and superior performance of PTGB compared to baseline methods."
                    },
                    {
                        "title": "Microstructures and Accuracy of Graph Recall by Large Language Models",
                        "abstract": "Graphs data is crucial for many applications, and much of it exists in the relations described in textual format. As a result, being able to accurately recall and encode a graph described in earlier text is a basic yet pivotal ability that LLMs need to demonstrate if they are to perform reasoning tasks that involve graph-structured information. Human performance at graph recall has been studied by cognitive scientists for decades, and has been found to often exhibit certain structural patterns of bias that align with human handling of social relationships. To date, however, we know little about how LLMs behave in analogous graph recall tasks: do their recalled graphs also exhibit certain biased patterns, and if so, how do they compare with humans and affect other graph reasoning tasks? In this work, we perform the first systematical study of graph recall by LLMs, investigating the accuracy and biased microstructures (local structural patterns) in their recall. We find that LLMs not only underperform often in graph recall, but also tend to favor more triangles and alternating 2-paths. Moreover, we find that more advanced LLMs have a striking dependence on the domain that a real-world graph comes from -- by yielding the best recall accuracy when the graph is narrated in a language style consistent with its original domain."
                    },
                    {
                        "title": "On Positional and Structural Node Features for Graph Neural Networks on Non-attributed Graphs",
                        "abstract": "Graph neural networks (GNNs) have been widely used in various graph-related problems such as node classification and graph classification, where superior performance is mainly established when natural node features are available. However, it is not well understood how GNNs work without natural node features, especially regarding the various ways to construct artificial ones. In this paper, we point out the two types of artificial node features, i.e., positional and structural node features, and provide insights on why each of them is more appropriate for certain tasks, i.e., positional node classification, structural node classification, and graph classification. Extensive experimental results on 10 benchmark datasets validate our insights, thus leading to a practical guideline on the choices between different artificial node features for GNNs on non-attributed graphs. The code is available at https://github.com/zjzijielu/gnn-positional-structural-node-features."
                    },
                    {
                        "title": "Joint Embedding of Structural and Functional Brain Networks with Graph Neural Networks for Mental Illness Diagnosis",
                        "abstract": "Multimodal brain networks characterize complex connectivities among different brain regions from both structural and functional aspects and provide a new means for mental disease analysis. Recently, Graph Neural Networks (GNNs) have become a de facto model for analyzing graph-structured data. However, how to employ GNNs to extract effective representations from brain networks in multiple modalities remains rarely explored. Moreover, as brain networks provide no initial node features, how to design informative node attributes and leverage edge weights for GNNs to learn is left unsolved. To this end, we develop a novel multiview GNN for multimodal brain networks. In particular, we regard each modality as a view for brain networks and employ contrastive learning for multimodal fusion. Then, we propose a GNN model which takes advantage of the message passing scheme by propagating messages based on degree statistics and brain region connectivities. Extensive experiments on two real-world disease datasets (HIV and Bipolar) demonstrate the effectiveness of our proposed method over state-of-the-art baselines."
                    },
                    {
                        "title": "BrainNNExplainer: An Interpretable Graph Neural Network Framework for Brain Network based Disease Analysis",
                        "abstract": "Interpretable brain network models for disease prediction are of great value for the advancement of neuroscience. GNNs are promising to model complicated network data, but they are prone to overfitting and suffer from poor interpretability, which prevents their usage in decision-critical scenarios like healthcare. To bridge this gap, we propose BrainNNExplainer, an interpretable GNN framework for brain network analysis. It is mainly composed of two jointly learned modules: a backbone prediction model that is specifically designed for brain networks and an explanation generator that highlights disease-specific prominent brain network connections. Extensive experimental results with visualizations on two challenging disease prediction datasets demonstrate the unique interpretability and outstanding performance of BrainNNExplainer."
                    },
                    {
                        "title": "PV2TEA: Patching Visual Modality to Textual-Established Information Extraction",
                        "abstract": "Information extraction, e.g., attribute value extraction, has been extensively studied and formulated based only on text. However, many attributes can benefit from image-based extraction, like color, shape, pattern, among others. The visual modality has long been underutilized, mainly due to multimodal annotation difficulty. In this paper, we aim to patch the visual modality to the textual-established attribute information extractor. The cross-modality integration faces several unique challenges: (C1) images and textual descriptions are loosely paired intra-sample and inter-samples; (C2) images usually contain rich backgrounds that can mislead the prediction; (C3) weakly supervised labels from textual-established extractors are biased for multimodal training. We present PV2TEA, an encoder-decoder architecture equipped with three bias reduction schemes: (S1) Augmented label-smoothed contrast to improve the cross-modality alignment for loosely-paired image and text; (S2) Attention-pruning that adaptively distinguishes the visual foreground; (S3) Two-level neighborhood regularization that mitigates the label textual bias via reliability estimation. Empirical results on real-world e-Commerce datasets demonstrate up to 11.74% absolute (20.97% relatively) F1 increase over unimodal baselines."
                    },
                    {
                        "title": "Interpretable Graph Neural Networks for Connectome-Based Brain Disorder Analysis",
                        "abstract": "Human brains lie at the core of complex neurobiological systems, where the neurons, circuits, and subsystems interact in enigmatic ways. Understanding the structural and functional mechanisms of the brain has long been an intriguing pursuit for neuroscience research and clinical disorder therapy. Mapping the connections of the human brain as a network is one of the most pervasive paradigms in neuroscience. Graph Neural Networks (GNNs) have recently emerged as a potential method for modeling complex network data. Deep models, on the other hand, have low interpretability, which prevents their usage in decision-critical contexts like healthcare. To bridge this gap, we propose an interpretable framework to analyze disorder-specific Regions of Interest (ROIs) and prominent connections. The proposed framework consists of two modules: a brain-network-oriented backbone model for disease prediction and a globally shared explanation generator that highlights disorder-specific biomarkers including salient ROIs and important connections. We conduct experiments on three real-world datasets of brain disorders. The results verify that our framework can obtain outstanding performance and also identify meaningful biomarkers. All code for this work is available at https://github.com/HennyJie/IBGNN.git."
                    },
                    {
                        "title": "Multimodal Fusion of EHR in Structures and Semantics: Integrating Clinical Records and Notes with Hypergraph and LLM",
                        "abstract": "Electronic Health Records (EHRs) have become increasingly popular to support clinical decision-making and healthcare in recent decades. EHRs usually contain heterogeneous information, such as structural data in tabular form and unstructured data in textual notes. Different types of information in EHRs can complement each other and provide a more complete picture of the health status of a patient. While there has been a lot of research on representation learning of structured EHR data, the fusion of different types of EHR data (multimodal fusion) is not well studied. This is mostly because of the complex medical coding systems used and the noise and redundancy present in the written notes. In this work, we propose a new framework called MINGLE, which integrates both structures and semantics in EHR effectively. Our framework uses a two-level infusion strategy to combine medical concept semantics and clinical note semantics into hypergraph neural networks, which learn the complex interactions between different types of data to generate visit representations for downstream prediction. Experiment results on two EHR datasets, the public MIMIC-III and private CRADLE, show that MINGLE can effectively improve predictive performance by 11.83% relatively, enhancing semantic integration as well as multimodal fusion for structural and textual EHR data."
                    },
                    {
                        "title": "Data-Efficient Brain Connectome Analysis via Multi-Task Meta-Learning",
                        "abstract": "Brain networks characterize complex connectivities among brain regions as graph structures, which provide a powerful means to study brain connectomes. In recent years, graph neural networks have emerged as a prevalent paradigm of learning with structured data. However, most brain network datasets are limited in sample sizes due to the relatively high cost of data acquisition, which hinders the deep learning models from sufficient training. Inspired by meta-learning that learns new concepts fast with limited training examples, this paper studies data-efficient training strategies for analyzing brain connectomes in a cross-dataset setting. Specifically, we propose to meta-train the model on datasets of large sample sizes and transfer the knowledge to small datasets. In addition, we also explore two brain-network-oriented designs, including atlas transformation and adaptive task reweighing. Compared to other pre-training strategies, our meta-learning-based approach achieves higher and stabler performance, which demonstrates the effectiveness of our proposed solutions. The framework is also able to derive new insights regarding the similarities among datasets and diseases in a data-driven fashion."
                    },
                    {
                        "title": "Transformer-Based Hierarchical Clustering for Brain Network Analysis",
                        "abstract": "Brain networks, graphical models such as those constructed from MRI, have been widely used in pathological prediction and analysis of brain functions. Within the complex brain system, differences in neuronal connection strengths parcellate the brain into various functional modules (network communities), which are critical for brain analysis. However, identifying such communities within the brain has been a nontrivial issue due to the complexity of neuronal interactions. In this work, we propose a novel interpretable transformer-based model for joint hierarchical cluster identification and brain network classification. Extensive experimental results on real-world brain network datasets show that with the help of hierarchical clustering, the model achieves increased accuracy and reduced runtime complexity while providing plausible insight into the functional organization of brain regions. The implementation is available at https://github.com/DDVD233/THC."
                    },
                    {
                        "title": "Structure-Aware Hard Negative Mining for Heterogeneous Graph Contrastive Learning",
                        "abstract": "Recently, heterogeneous Graph Neural Networks (GNNs) have become a de facto model for analyzing HGs, while most of them rely on a relative large number of labeled data. In this work, we investigate Contrastive Learning (CL), a key component in self-supervised approaches, on HGs to alleviate the label scarcity problem. We first generate multiple semantic views according to metapaths and network schemas. Then, by pushing node embeddings corresponding to different semantic views close to each other (positives) and pulling other embeddings apart (negatives), one can obtain informative representations without human annotations. However, this CL approach ignores the relative hardness of negative samples, which may lead to suboptimal performance. Considering the complex graph structure and the smoothing nature of GNNs, we propose a structure-aware hard negative mining scheme that measures hardness by structural characteristics for HGs. By synthesizing more negative nodes, we give larger weights to harder negatives with limited computational overhead to further boost the performance. Empirical studies on three real-world datasets show the effectiveness of our proposed method. The proposed method consistently outperforms existing state-of-the-art methods and notably, even surpasses several supervised counterparts."
                    },
                    {
                        "title": "FBNETGEN: Task-aware GNN-based fMRI Analysis via Functional Brain Network Generation",
                        "abstract": "Functional magnetic resonance imaging (fMRI) is one of the most common imaging modalities to investigate brain functions. Recent studies in neuroscience stress the great potential of functional brain networks constructed from fMRI data for clinical predictions. Traditional functional brain networks, however, are noisy and unaware of downstream prediction tasks, while also incompatible with the deep graph neural network (GNN) models. In order to fully unleash the power of GNNs in network-based fMRI analysis, we develop FBNETGEN, a task-aware and interpretable fMRI analysis framework via deep brain network generation. In particular, we formulate (1) prominent region of interest (ROI) features extraction, (2) brain networks generation, and (3) clinical predictions with GNNs, in an end-to-end trainable model under the guidance of particular prediction tasks. Along with the process, the key novel component is the graph generator which learns to transform raw time-series features into task-oriented brain networks. Our learnable graphs also provide unique interpretations by highlighting prediction-related brain regions. Comprehensive experiments on two datasets, i.e., the recently released and currently largest publicly available fMRI dataset Adolescent Brain Cognitive Development (ABCD), and the widely-used fMRI dataset PNC, prove the superior effectiveness and interpretability of FBNETGEN. The implementation is available at https://github.com/Wayfear/FBNETGEN."
                    },
                    {
                        "title": "Brain Network Transformer",
                        "abstract": "Human brains are commonly modeled as networks of Regions of Interest (ROIs) and their connections for the understanding of brain functions and mental disorders. Recently, Transformer-based models have been studied over different types of data, including graphs, shown to bring performance gains widely. In this work, we study Transformer-based models for brain network analysis. Driven by the unique properties of data, we model brain networks as graphs with nodes of fixed size and order, which allows us to (1) use connection profiles as node features to provide natural and low-cost positional information and (2) learn pair-wise connection strengths among ROIs with efficient attention weights across individuals that are predictive towards downstream analysis tasks. Moreover, we propose an Orthonormal Clustering Readout operation based on self-supervised soft clustering and orthonormal projection. This design accounts for the underlying functional modules that determine similar behaviors among groups of ROIs, leading to distinguishable cluster-aware node embeddings and informative graph embeddings. Finally, we re-standardize the evaluation pipeline on the only one publicly available large-scale brain network dataset of ABIDE, to enable meaningful comparison of different models. Experiment results show clear improvements of our proposed Brain Network Transformer on both the public ABIDE and our restricted ABCD datasets. The implementation is available at https://github.com/Wayfear/BrainNetworkTransformer."
                    },
                    {
                        "title": "Dynamic Brain Transformer with Multi-level Attention for Functional Brain Network Analysis",
                        "abstract": "Recent neuroimaging studies have highlighted the importance of network-centric brain analysis, particularly with functional magnetic resonance imaging. The emergence of Deep Neural Networks has fostered a substantial interest in predicting clinical outcomes and categorizing individuals based on brain networks. However, the conventional approach involving static brain network analysis offers limited potential in capturing the dynamism of brain function. Although recent studies have attempted to harness dynamic brain networks, their high dimensionality and complexity present substantial challenges. This paper proposes a novel methodology, Dynamic bRAin Transformer (DART), which combines static and dynamic brain networks for more effective and nuanced brain function analysis. Our model uses the static brain network as a baseline, integrating dynamic brain networks to enhance performance against traditional methods. We innovatively employ attention mechanisms, enhancing model explainability and exploiting the dynamic brain network's temporal variations. The proposed approach offers a robust solution to the low signal-to-noise ratio of blood-oxygen-level-dependent signals, a recurring issue in direct DNN modeling. It also provides valuable insights into which brain circuits or dynamic networks contribute more to final predictions. As such, DRAT shows a promising direction in neuroimaging studies, contributing to the comprehensive understanding of brain organization and the role of neural circuits."
                    },
                    {
                        "title": "TACCO: Task-guided Co-clustering of Clinical Concepts and Patient Visits for Disease Subtyping based on EHR Data",
                        "abstract": "The growing availability of well-organized Electronic Health Records (EHR) data has enabled the development of various machine learning models towards disease risk prediction. However, existing risk prediction methods overlook the heterogeneity of complex diseases, failing to model the potential disease subtypes regarding their corresponding patient visits and clinical concept subgroups. In this work, we introduce TACCO, a novel framework that jointly discovers clusters of clinical concepts and patient visits based on a hypergraph modeling of EHR data. Specifically, we develop a novel self-supervised co-clustering framework that can be guided by the risk prediction task of specific diseases. Furthermore, we enhance the hypergraph model of EHR data with textual embeddings and enforce the alignment between the clusters of clinical concepts and patient visits through a contrastive objective. Comprehensive experiments conducted on the public MIMIC-III dataset and Emory internal CRADLE dataset over the downstream clinical tasks of phenotype classification and cardiovascular risk prediction demonstrate an average 31.25% performance improvement compared to traditional ML baselines and a 5.26% improvement on top of the vanilla hypergraph model without our co-clustering mechanism. In-depth model analysis, clustering results analysis, and clinical case studies further validate the improved utilities and insightful interpretations delivered by TACCO. Code is available at https://github.com/PericlesHat/TACCO."
                    },
                    {
                        "title": "Open Visual Knowledge Extraction via Relation-Oriented Multimodality Model Prompting",
                        "abstract": "Images contain rich relational knowledge that can help machines understand the world. Existing methods on visual knowledge extraction often rely on the pre-defined format (e.g., sub-verb-obj tuples) or vocabulary (e.g., relation types), restricting the expressiveness of the extracted knowledge. In this work, we take a first exploration to a new paradigm of open visual knowledge extraction. To achieve this, we present OpenVik which consists of an open relational region detector to detect regions potentially containing relational knowledge and a visual knowledge generator that generates format-free knowledge by prompting the large multimodality model with the detected region of interest. We also explore two data enhancement techniques for diversifying the generated format-free visual knowledge. Extensive knowledge quality evaluations highlight the correctness and uniqueness of the extracted open visual knowledge by OpenVik. Moreover, integrating our extracted knowledge across various visual reasoning applications shows consistent improvements, indicating the real-world applicability of OpenVik."
                    }
                ]
            },
            "831a1664-483c-4b14-b8dd-114400d9e12c": {
                "pk": "831a1664-483c-4b14-b8dd-114400d9e12c",
                "name": "Lingjun Mao",
                "collaborators": [
                    "Zineng Tang",
                    "Junwei Su",
                    "Chuan Wu",
                    "Yizhou Chi",
                    "Alane Suhr",
                    "Kejiang Qian",
                    "Xin Liang",
                    "Yimin Ding",
                    "Jin Gao",
                    "Xinran Wei"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Multi-Agent Systems",
                    "Computer Vision",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "BG-HGNN: Toward Scalable and Efficient Heterogeneous Graph Neural Network",
                        "abstract": "Many computer vision and machine learning problems are modelled as learning tasks on heterogeneous graphs, featuring a wide array of relations from diverse types of nodes and edges. Heterogeneous graph neural networks (HGNNs) stand out as a promising neural model class designed for heterogeneous graphs. Built on traditional GNNs, existing HGNNs employ different parameter spaces to model the varied relationships. However, the practical effectiveness of existing HGNNs is often limited to simple heterogeneous graphs with few relation types. This paper first highlights and demonstrates that the standard approach employed by existing HGNNs inevitably leads to parameter explosion and relation collapse, making HGNNs less effective or impractical for complex heterogeneous graphs with numerous relation types. To overcome this issue, we introduce a novel framework, Blend&Grind-HGNN (BG-HGNN), which effectively tackles the challenges by carefully integrating different relations into a unified feature space manageable by a single set of parameters. This results in a refined HGNN method that is more efficient and effective in learning from heterogeneous graphs, especially when the number of relations grows. Our empirical studies illustrate that BG-HGNN significantly surpasses existing HGNNs in terms of parameter efficiency (up to 28.96 $\\times$), training throughput (up to 8.12 $\\times$), and accuracy (up to 1.07 $\\times$)."
                    },
                    {
                        "title": "AMONGAGENTS: Evaluating Large Language Models in the Interactive Text-Based Social Deduction Game",
                        "abstract": "Strategic social deduction games serve as valuable testbeds for evaluating the understanding and inference skills of language models, offering crucial insights into social science, artificial intelligence, and strategic gaming. This paper focuses on creating proxies of human behavior in simulated environments, with Among Us utilized as a tool for studying simulated human behavior. The study introduces a text-based game environment, named AmongAgents, that mirrors the dynamics of Among Us. Players act as crew members aboard a spaceship, tasked with identifying impostors who are sabotaging the ship and eliminating the crew. Within this environment, the behavior of simulated language agents is analyzed. The experiments involve diverse game sequences featuring different configurations of Crewmates and Impostor personality archetypes. Our work demonstrates that state-of-the-art large language models (LLMs) can effectively grasp the game rules and make decisions based on the current context. This work aims to promote further exploration of LLMs in goal-oriented games with incomplete information and complex action spaces, as these settings offer valuable opportunities to assess language model performance in socially driven scenarios."
                    },
                    {
                        "title": "Grounding Language in Multi-Perspective Referential Communication",
                        "abstract": "We introduce a task and dataset for referring expression generation and comprehension in multi-agent embodied environments. In this task, two agents in a shared scene must take into account one another's visual perspective, which may be different from their own, to both produce and understand references to objects in a scene and the spatial relations between them. We collect a dataset of 2,970 human-written referring expressions, each paired with human comprehension judgments, and evaluate the performance of automated models as speakers and listeners paired with human partners, finding that model performance in both reference generation and comprehension lags behind that of pairs of human agents. Finally, we experiment training an open-weight speaker model with evidence of communicative success when paired with a listener, resulting in an improvement from 58.9 to 69.3% in communicative success and even outperforming the strongest proprietary model."
                    },
                    {
                        "title": "AI Agent as Urban Planner: Steering Stakeholder Dynamics in Urban Planning via Consensus-based Multi-Agent Reinforcement Learning",
                        "abstract": "In urban planning, land use readjustment plays a pivotal role in aligning land use configurations with the current demands for sustainable urban development. However, present-day urban planning practices face two main issues. Firstly, land use decisions are predominantly dependent on human experts. Besides, while resident engagement in urban planning can promote urban sustainability and livability, it is challenging to reconcile the diverse interests of stakeholders. To address these challenges, we introduce a Consensus-based Multi-Agent Reinforcement Learning framework for real-world land use readjustment. This framework serves participatory urban planning, allowing diverse intelligent agents as stakeholder representatives to vote for preferred land use types. Within this framework, we propose a novel consensus mechanism in reward design to optimize land utilization through collective decision making. To abstract the structure of the complex urban system, the geographic information of cities is transformed into a spatial graph structure and then processed by graph neural networks. Comprehensive experiments on both traditional top-down planning and participatory planning methods from real-world communities indicate that our computational framework enhances global benefits and accommodates diverse interests, leading to improved satisfaction across different demographic groups. By integrating Multi-Agent Reinforcement Learning, our framework ensures that participatory urban planning decisions are more dynamic and adaptive to evolving community needs and provides a robust platform for automating complex real-world urban planning processes."
                    }
                ]
            },
            "4fed958f-9efc-406f-81c6-03d8551b686f": {
                "pk": "4fed958f-9efc-406f-81c6-03d8551b686f",
                "name": "Xin Liang",
                "collaborators": [
                    "Peter Benner",
                    "Zhen-Chen Guo",
                    "Yi Jiang",
                    "Suzana Miodragovi\u0107",
                    "Ninoslav Truhar",
                    "Ren-Cang Li"
                ],
                "domain": [
                    "Numerical Analysis",
                    "Optimization",
                    "Control Theory",
                    "Eigenvalue Problems"
                ],
                "publications": [
                    {
                        "title": "On the Optimality of the Oja's Algorithm for Online PCA",
                        "abstract": "In this paper we analyze the behavior of the Oja's algorithm for online/streaming principal component subspace estimation. It is proved that with high probability it performs an efficient, gap-free, global convergence rate to approximate an principal component subspace for any sub-Gaussian distribution. Moreover, it is the first time to show that the convergence rate, namely the upper bound of the approximation, exactly matches the lower bound of an approximation obtained by the offline/classical PCA up to a constant factor."
                    },
                    {
                        "title": "Convergence Analysis of Extended LOBPCG for Computing Extreme Eigenvalues",
                        "abstract": "This paper is concerned with the convergence analysis of an extended variation of the locally optimal preconditioned conjugate gradient method (LOBPCG) for the extreme eigenvalue of a Hermitian matrix polynomial which admits some extended form of Rayleigh quotient. This work is a generalization of the analysis by Ovtchinnikov (SIAM J. Numer. Anal., 46(5):2567-2592, 2008). As instances, the algorithms for definite matrix pairs and hyperbolic quadratic matrix polynomials are shown to be globally convergent and to have an asymptotically local convergence rate. Also, numerical examples are given to illustrate the convergence."
                    },
                    {
                        "title": "Nonlinear Kalman Filter Using Cramer Rao Bound",
                        "abstract": "This paper studies the optimal state estimation for a dynamic system, whose transfer function can be nonlinear and the input noise can be of arbitrary distribution. Our algorithm differs from the conventional extended Kalman filter (EKF) and the particle filter (PF) in that it estimates not only the state vector but also the Cramer-Rao bound (CRB), which serves as an accuracy indicator. Combining the state estimation, the CRB, and the incoming new measurement, the algorithm updates the state estimation according to the maximum likelihood (ML) criterion. To illustrate the effectiveness of the proposed method for autonomous driving, we apply it to estimate the position and velocity of a vehicle based on the noisy measurements of distance and Doppler offset. Simulation results show that the proposed algorithm can achieve estimation significantly more accurate than the standard EKF and the PF."
                    },
                    {
                        "title": "Relative Perturbation Theory for Quadratic Hermitian Eigenvalue Problem",
                        "abstract": "In this paper, we derive new relative perturbation bounds for eigenvectors and eigenvalues for regular quadratic eigenvalue problems of the form $\\lambda^2 M x + \\lambda C x + K x = 0$, where $M$ and $K$ are nonsingular Hermitian matrices and $C$ is a general Hermitian matrix. We base our findings on new results for an equivalent regular Hermitian matrix pair $A-\\lambda B$. The new bounds can be applied to many interesting quadratic eigenvalue problems appearing in applications, such as mechanical models with indefinite damping. The quality of our bounds is demonstrated by several numerical experiments."
                    },
                    {
                        "title": "The intrinsic Toeplitz structure and its applications in algebraic Riccati equations",
                        "abstract": "In this paper we derive a Toeplitz-structured closed form of the unique positive semi-definite stabilizing solution for the discrete-time algebraic Riccati equations, especially for the case that the state matrix is not stable. Based on the found form and fast Fourier transform, we propose a new algorithm for solving both discrete-time and continuous-time large-scale algebraic Riccati equations with low-rank structure. It works without unnecessary assumptions, complicated shift selection strategies, or matrix calculations of the cubic order with respect to the problem scale. Numerical examples are given to illustrate its features. Besides, we show that it is theoretically equivalent to several algorithms existing in the literature in the sense that they all produce the same sequence under the same parameter setting."
                    },
                    {
                        "title": "Stochastic algebraic Riccati equations are almost as easy as deterministic ones theoretically",
                        "abstract": "Stochastic algebraic Riccati equations, also known as rational algebraic Riccati equations, arising in linear-quadratic optimal control for stochastic linear time-invariant systems, were considered to be not easy to solve. The-state-of-art numerical methods most rely on differentiability or continuity, such as Newton-type method, LMI method, or homotopy method. In this paper, we will build a novel theoretical framework and reveal the intrinsic algebraic structure appearing in this kind of algebraic Riccati equations. This structure guarantees that to solve them is almost as easy as to solve deterministic/classical ones, which will shed light on the theoretical analysis and numerical algorithm design for this topic."
                    },
                    {
                        "title": "On Generalizing Trace Minimization Principles, II",
                        "abstract": "This paper is concerned with establishing a trace minimization principle for two Hermitian matrix pairs. Specifically, we will answer the question: when is $\\inf_X\\operatorname{tr}(\\widehat AX^{\\rm H}AX)$ subject to $\\widehat BX^{\\rm H}BX=I$ (the identity matrix of apt size) finite? Sufficient and necessary conditions are obtained and, when the infimum is finite, an explicit formula for it is presented in terms of the finite eigenvalues of the matrix pairs. Our results extend Fan's trace minimization principle (1949) for a Hermitian matrix, a minimization principle of Kova\\v{c}-Striko and Veseli\\'c (1995) for a Hermitian matrix pair, and most recent ones by the authors and their collaborators for a Hermitian matrix pair and a Hermitian matrix."
                    }
                ]
            },
            "b6e046d8-027d-4236-a597-7fff8a32baef": {
                "pk": "b6e046d8-027d-4236-a597-7fff8a32baef",
                "name": "Jieyu Zhang",
                "collaborators": [
                    "Alexander Ratner",
                    "Linxin Song",
                    "Jiawei Han",
                    "Cheng-Yu Hsieh",
                    "Tianyi Zhou",
                    "Haonan Wang",
                    "Ranjay Krishna",
                    "Yue Yu",
                    "Carl Yang",
                    "Minhao Jiang"
                ],
                "domain": [
                    "Weak Supervision",
                    "Semi-Supervised Learning",
                    "Graph Neural Network",
                    "Active Learning"
                ],
                "publications": [
                    {
                        "title": "Neural Embedding Propagation on Heterogeneous Networks",
                        "abstract": "Classification is one of the most important problems in machine learning. To address label scarcity, semi-supervised learning (SSL) has been intensively studied over the past two decades, which mainly leverages data affinity modeled by networks. Label propagation (LP), however, as the most popular SSL technique, mostly only works on homogeneous networks with single-typed simple interactions. In this work, we focus on the more general and powerful heterogeneous networks, which accommodate multi-typed objects and links, and thus endure multi-typed complex interactions. Specifically, we propose \\textit{neural embedding propagation} (NEP), which leverages distributed embeddings to represent objects and dynamically composed modular networks to model their complex interactions. While generalizing LP as a simple instance, NEP is far more powerful in its natural awareness of different types of objects and links, and the ability to automatically capture their important interaction patterns. Further, we develop a series of efficient training strategies for NEP, leading to its easy deployment on real-world heterogeneous networks with millions of objects. With extensive experiments on three datasets, we comprehensively demonstrate the effectiveness, efficiency, and robustness of NEP compared with state-of-the-art network embedding and SSL algorithms."
                    },
                    {
                        "title": "Nemo: Guiding and Contextualizing Weak Supervision for Interactive Data Programming",
                        "abstract": "Weak Supervision (WS) techniques allow users to efficiently create large training datasets by programmatically labeling data with heuristic sources of supervision. While the success of WS relies heavily on the provided labeling heuristics, the process of how these heuristics are created in practice has remained under-explored. In this work, we formalize the development process of labeling heuristics as an interactive procedure, built around the existing workflow where users draw ideas from a selected set of development data for designing the heuristic sources. With the formalism, we study two core problems of how to strategically select the development data to guide users in efficiently creating informative heuristics, and how to exploit the information within the development process to contextualize and better learn from the resultant heuristics. Building upon two novel methodologies that effectively tackle the respective problems considered, we present Nemo, an end-to-end interactive system that improves the overall productivity of WS learning pipeline by an average 20% (and up to 47% in one task) compared to the prevailing WS approach."
                    },
                    {
                        "title": "TaxoEnrich: Self-Supervised Taxonomy Completion via Structure-Semantic Representations",
                        "abstract": "Taxonomies are fundamental to many real-world applications in various domains, serving as structural representations of knowledge. To deal with the increasing volume of new concepts needed to be organized as taxonomies, researchers turn to automatically completion of an existing taxonomy with new concepts. In this paper, we propose TaxoEnrich, a new taxonomy completion framework, which effectively leverages both semantic features and structural information in the existing taxonomy and offers a better representation of candidate position to boost the performance of taxonomy completion. Specifically, TaxoEnrich consists of four components: (1) taxonomy-contextualized embedding which incorporates both semantic meanings of concept and taxonomic relations based on powerful pretrained language models; (2) a taxonomy-aware sequential encoder which learns candidate position representations by encoding the structural information of taxonomy; (3) a query-aware sibling encoder which adaptively aggregates candidate siblings to augment candidate position representations based on their importance to the query-position matching; (4) a query-position matching model which extends existing work with our new candidate position representations. Extensive experiments on four large real-world datasets from different domains show that \\TaxoEnrich achieves the best performance among all evaluation metrics and outperforms previous state-of-the-art methods by a large margin."
                    },
                    {
                        "title": "Leveraging Instance Features for Label Aggregation in Programmatic Weak Supervision",
                        "abstract": "Programmatic Weak Supervision (PWS) has emerged as a widespread paradigm to synthesize training labels efficiently. The core component of PWS is the label model, which infers true labels by aggregating the outputs of multiple noisy supervision sources abstracted as labeling functions (LFs). Existing statistical label models typically rely only on the outputs of LF, ignoring the instance features when modeling the underlying generative process. In this paper, we attempt to incorporate the instance features into a statistical label model via the proposed FABLE. In particular, it is built on a mixture of Bayesian label models, each corresponding to a global pattern of correlation, and the coefficients of the mixture components are predicted by a Gaussian Process classifier based on instance features. We adopt an auxiliary variable-based variational inference algorithm to tackle the non-conjugate issue between the Gaussian Process and Bayesian label models. Extensive empirical comparison on eleven benchmark datasets sees FABLE achieving the highest averaged performance across nine baselines."
                    },
                    {
                        "title": "Taming Small-sample Bias in Low-budget Active Learning",
                        "abstract": "Active learning (AL) aims to minimize the annotation cost by only querying a few informative examples for each model training stage. However, training a model on a few queried examples suffers from the small-sample bias. In this paper, we address this small-sample bias issue in low-budget AL by exploring a regularizer called Firth bias reduction, which can provably reduce the bias during the model training process but might hinder learning if its coefficient is not adaptive to the learning progress. Instead of tuning the coefficient for each query round, which is sensitive and time-consuming, we propose the curriculum Firth bias reduction (CHAIN) that can automatically adjust the coefficient to be adaptive to the training process. Under both deep learning and linear model settings, experiments on three benchmark datasets with several widely used query strategies and hyperparameter searching methods show that CHAIN can be used to build more efficient AL and can substantially improve the progress made by each active learning query."
                    },
                    {
                        "title": "Augmentation-Free Graph Contrastive Learning with Performance Guarantee",
                        "abstract": "Graph contrastive learning (GCL) is the most representative and prevalent self-supervised learning approach for graph-structured data. Despite its remarkable success, existing GCL methods highly rely on an augmentation scheme to learn the representations invariant across different augmentation views. In this work, we revisit such a convention in GCL through examining the effect of augmentation techniques on graph data via the lens of spectral theory. We found that graph augmentations preserve the low-frequency components and perturb the middle-and high-frequency components of the graph, which contributes to the success of GCL algorithms on homophilic graphs but hinder its application on heterophilic graphs, due to the high-frequency preference of heterophilic data. Motivated by this, we propose a novel, theoretically-principled, and augmentation-free GCL method, named AF-GCL, that (1) leverages the features aggregated by Graph Neural Network to construct the self-supervision signal instead of augmentations and therefore (2) is less sensitive to the graph homophily degree. Theoretically, We present the performance guarantee for AF-GCL as well as an analysis for understanding the efficacy of AF-GCL. Extensive experiments on 14 benchmark datasets with varying degrees of heterophily show that AF-GCL presents competitive or better performance on homophilic graphs and outperforms all existing state-of-the-art GCL methods on heterophilic graphs with significantly less computational overhead."
                    },
                    {
                        "title": "Frustratingly Easy Regularization on Representation Can Boost Deep Reinforcement Learning",
                        "abstract": "Deep reinforcement learning (DRL) gives the promise that an agent learns good policy from high-dimensional information, whereas representation learning removes irrelevant and redundant information and retains pertinent information. In this work, we demonstrate that the learned representation of the $Q$-network and its target $Q$-network should, in theory, satisfy a favorable distinguishable representation property. Specifically, there exists an upper bound on the representation similarity of the value functions of two adjacent time steps in a typical DRL setting. However, through illustrative experiments, we show that the learned DRL agent may violate this property and lead to a sub-optimal policy. Therefore, we propose a simple yet effective regularizer called Policy Evaluation with Easy Regularization on Representation (PEER), which aims to maintain the distinguishable representation property via explicit regularization on internal representations. And we provide the convergence rate guarantee of PEER. Implementing PEER requires only one line of code. Our experiments demonstrate that incorporating PEER into DRL can significantly improve performance and sample efficiency. Comprehensive experiments show that PEER achieves state-of-the-art performance on all 4 environments on PyBullet, 9 out of 12 tasks on DMControl, and 19 out of 26 games on Atari. To the best of our knowledge, PEER is the first work to study the inherent representation property of Q-network and its target. Our code is available at https://sites.google.com/view/peer-cvpr2023/."
                    },
                    {
                        "title": "Adaptive Ranking-based Sample Selection for Weakly Supervised Class-imbalanced Text Classification",
                        "abstract": "To obtain a large amount of training labels inexpensively, researchers have recently adopted the weak supervision (WS) paradigm, which leverages labeling rules to synthesize training labels rather than using individual annotations to achieve competitive results for natural language processing (NLP) tasks. However, data imbalance is often overlooked in applying the WS paradigm, despite being a common issue in a variety of NLP tasks. To address this challenge, we propose Adaptive Ranking-based Sample Selection (ARS2), a model-agnostic framework to alleviate the data imbalance issue in the WS paradigm. Specifically, it calculates a probabilistic margin score based on the output of the current model to measure and rank the cleanliness of each data point. Then, the ranked data are sampled based on both class-wise and rule-aware ranking. In particular, the two sample strategies corresponds to our motivations: (1) to train the model with balanced data batches to reduce the data imbalance issue and (2) to exploit the expertise of each labeling rule for collecting clean samples. Experiments on four text classification datasets with four different imbalance ratios show that ARS2 outperformed the state-of-the-art imbalanced learning and WS methods, leading to a 2%-57.8% improvement on their F1-score."
                    },
                    {
                        "title": "When to Learn What: Model-Adaptive Data Augmentation Curriculum",
                        "abstract": "Data augmentation (DA) is widely used to improve the generalization of neural networks by enforcing the invariances and symmetries to pre-defined transformations applied to input data. However, a fixed augmentation policy may have different effects on each sample in different training stages but existing approaches cannot adjust the policy to be adaptive to each sample and the training model. In this paper, we propose Model Adaptive Data Augmentation (MADAug) that jointly trains an augmentation policy network to teach the model when to learn what. Unlike previous work, MADAug selects augmentation operators for each input image by a model-adaptive policy varying between training stages, producing a data augmentation curriculum optimized for better generalization. In MADAug, we train the policy through a bi-level optimization scheme, which aims to minimize a validation-set loss of a model trained using the policy-produced data augmentations. We conduct an extensive evaluation of MADAug on multiple image classification tasks and network architectures with thorough comparisons to existing DA approaches. MADAug outperforms or is on par with other baselines and exhibits better fairness: it brings improvement to all classes and more to the difficult ones. Moreover, MADAug learned policy shows better performance when transferred to fine-grained datasets. In addition, the auto-optimized policy in MADAug gradually introduces increasing perturbations and naturally forms an easy-to-hard curriculum."
                    },
                    {
                        "title": "Iterated Learning Improves Compositionality in Large Vision-Language Models",
                        "abstract": "A fundamental characteristic common to both human vision and natural language is their compositional nature. Yet, despite the performance gains contributed by large vision and language pretraining, recent investigations find that most-if not all-our state-of-the-art vision-language models struggle at compositionality. They are unable to distinguish between images of \" a girl in white facing a man in black\" and \"a girl in black facing a man in white\". Moreover, prior work suggests that compositionality doesn't arise with scale: larger model sizes or training data don't help. This paper develops a new iterated training algorithm that incentivizes compositionality. We draw on decades of cognitive science research that identifies cultural transmission-the need to teach a new generation-as a necessary inductive prior that incentivizes humans to develop compositional languages. Specifically, we reframe vision-language contrastive learning as the Lewis Signaling Game between a vision agent and a language agent, and operationalize cultural transmission by iteratively resetting one of the agent's weights during training. After every iteration, this training paradigm induces representations that become \"easier to learn\", a property of compositional languages: e.g. our model trained on CC3M and CC12M improves standard CLIP by 4.7%, 4.0% respectfully in the SugarCrepe benchmark."
                    },
                    {
                        "title": "AcTune: Uncertainty-aware Active Self-Training for Semi-Supervised Active Learning with Pretrained Language Models",
                        "abstract": "While pre-trained language model (PLM) fine-tuning has achieved strong performance in many NLP tasks, the fine-tuning stage can be still demanding in labeled data. Recent works have resorted to active fine-tuning to improve the label efficiency of PLM fine-tuning, but none of them investigate the potential of unlabeled data. We propose {\\ours}, a new framework that leverages unlabeled data to improve the label efficiency of active PLM fine-tuning. AcTune switches between data annotation and model self-training based on uncertainty: it selects high-uncertainty unlabeled samples for active annotation and low-uncertainty ones for model self-training. Under this framework, we design (1) a region-aware sampling strategy that reduces redundancy when actively querying for annotations and (2) a momentum-based memory bank that dynamically aggregates the model's pseudo labels to suppress label noise in self-training. Experiments on 6 text classification datasets show that AcTune outperforms the strongest active learning and self-training baselines and improves the label efficiency of PLM fine-tuning by 56.2\\% on average. Our implementation will be available at \\url{https://github.com/yueyu1030/actune}."
                    },
                    {
                        "title": "Understanding Programmatic Weak Supervision via Source-aware Influence Function",
                        "abstract": "Programmatic Weak Supervision (PWS) aggregates the source votes of multiple weak supervision sources into probabilistic training labels, which are in turn used to train an end model. With its increasing popularity, it is critical to have some tool for users to understand the influence of each component (e.g., the source vote or training data) in the pipeline and interpret the end model behavior. To achieve this, we build on Influence Function (IF) and propose source-aware IF, which leverages the generation process of the probabilistic labels to decompose the end model's training objective and then calculate the influence associated with each (data, source, class) tuple. These primitive influence score can then be used to estimate the influence of individual component of PWS, such as source vote, supervision source, and training data. On datasets of diverse domains, we demonstrate multiple use cases: (1) interpreting incorrect predictions from multiple angles that reveals insights for debugging the PWS pipeline, (2) identifying mislabeling of sources with a gain of 9%-37% over baselines, and (3) improving the end model's generalization performance by removing harmful components in the training objective (13%-24% better than ordinary IF)."
                    },
                    {
                        "title": "Binary Classification with Positive Labeling Sources",
                        "abstract": "To create a large amount of training labels for machine learning models effectively and efficiently, researchers have turned to Weak Supervision (WS), which uses programmatic labeling sources rather than manual annotation. Existing works of WS for binary classification typically assume the presence of labeling sources that are able to assign both positive and negative labels to data in roughly balanced proportions. However, for many tasks of interest where there is a minority positive class, negative examples could be too diverse for developers to generate indicative labeling sources. Thus, in this work, we study the application of WS on binary classification tasks with positive labeling sources only. We propose WEAPO, a simple yet competitive WS method for producing training labels without negative labeling sources. On 10 benchmark datasets, we show WEAPO achieves the highest averaged performance in terms of both the quality of synthesized labels and the performance of the final classifier supervised with these labels. We incorporated the implementation of \\method into WRENCH, an existing benchmarking platform."
                    },
                    {
                        "title": "EcoAssistant: Using LLM Assistant More Affordably and Accurately",
                        "abstract": "Today, users ask Large language models (LLMs) as assistants to answer queries that require external knowledge; they ask about the weather in a specific city, about stock prices, and even about where specific locations are within their neighborhood. These queries require the LLM to produce code that invokes external APIs to answer the user's question, yet LLMs rarely produce correct code on the first try, requiring iterative code refinement upon execution results. In addition, using LLM assistants to support high query volumes can be expensive. In this work, we contribute a framework, EcoAssistant, that enables LLMs to answer code-driven queries more affordably and accurately. EcoAssistant contains three components. First, it allows the LLM assistants to converse with an automatic code executor to iteratively refine code or to produce answers based on the execution results. Second, we use a hierarchy of LLM assistants, which attempts to answer the query with weaker, cheaper LLMs before backing off to stronger, expensive ones. Third, we retrieve solutions from past successful queries as in-context demonstrations to help subsequent queries. Empirically, we show that EcoAssistant offers distinct advantages for affordability and accuracy, surpassing GPT-4 by 10 points of success rate with less than 50% of GPT-4's cost."
                    },
                    {
                        "title": "MaskSearch: Querying Image Masks at Scale",
                        "abstract": "Machine learning tasks over image databases often generate masks that annotate image content (e.g., saliency maps, segmentation maps, depth maps) and enable a variety of applications (e.g., determine if a model is learning spurious correlations or if an image was maliciously modified to mislead a model). While queries that retrieve examples based on mask properties are valuable to practitioners, existing systems do not support them efficiently. In this paper, we formalize the problem and propose MaskSearch, a system that focuses on accelerating queries over databases of image masks while guaranteeing the correctness of query results. MaskSearch leverages a novel indexing technique and an efficient filter-verification query execution framework. Experiments with our prototype show that MaskSearch, using indexes approximately 5% of the compressed data size, accelerates individual queries by up to two orders of magnitude and consistently outperforms existing methods on various multi-query workloads that simulate dataset exploration and analysis processes."
                    },
                    {
                        "title": "How Many Validation Labels Do You Need? Exploring the Design Space of Label-Efficient Model Ranking",
                        "abstract": "This paper presents LEMR (Label-Efficient Model Ranking) and introduces the MoraBench Benchmark. LEMR is a novel framework that minimizes the need for costly annotations in model selection by strategically annotating instances from an unlabeled validation set. To evaluate LEMR, we leverage the MoraBench Benchmark, a comprehensive collection of model outputs across diverse scenarios. Our extensive evaluation across 23 different NLP tasks in semi-supervised learning, weak supervision, and prompt selection tasks demonstrates LEMR's effectiveness in significantly reducing labeling costs. Key findings highlight the impact of suitable ensemble methods, uncertainty sampling strategies, and model committee selection in enhancing model ranking accuracy. LEMR, supported by the insights from MoraBench, provides a cost-effective and accurate solution for model selection, especially valuable in resource-constrained environments."
                    }
                ]
            },
            "8f660db7-1afc-444e-aacc-76a561b402ec": {
                "pk": "8f660db7-1afc-444e-aacc-76a561b402ec",
                "name": "Hui Ren",
                "collaborators": [
                    "Chuyu Zhang",
                    "Xuming He",
                    "Cheng Hui Ren",
                    "Jiasha Liu",
                    "Xiang Li",
                    "Quanzheng Li",
                    "Ruijie Xu"
                ],
                "domain": [
                    "Deep Learning",
                    "Clustering",
                    "Computer Vision",
                    "Ensemble Learning"
                ],
                "publications": [
                    {
                        "title": "Expansion and Improvement of Sieve and application in Goldbach's problem",
                        "abstract": "This paper expands and improves on the general Sieve method. This expaned and improved Sieve is applied to Goldbach's problem. A new estimate of the exception set in Goldbach's number E(X), an improved lower bound D_{1,2}(N) and upper bound D(N) are proposed. The proposed values are: |E(X)|<= X^{0.702+e}, D_{1,2}(N)>= 2.27C(N)/ln^{2}(N), D(N) <= 6.916C(N)/ln^{2}(N)."
                    },
                    {
                        "title": "Multi-Estimator Full Left Ventricle Quantification through Ensemble Learning",
                        "abstract": "Cardiovascular disease accounts for 1 in every 4 deaths in United States. Accurate estimation of structural and functional cardiac parameters is crucial for both diagnosis and disease management. In this work, we develop an ensemble learning framework for more accurate and robust left ventricle (LV) quantification. The framework combines two 1st-level modules: direct estimation module and a segmentation module. The direct estimation module utilizes Convolutional Neural Network (CNN) to achieve end-to-end quantification. The CNN is trained by taking 2D cardiac images as input and cardiac parameters as output. The segmentation module utilizes a U-Net architecture for obtaining pixel-wise prediction of the epicardium and endocardium of LV from the background. The binary U-Net output is then analyzed by a separate CNN for estimating the cardiac parameters. We then employ linear regression between the 1st-level predictor and ground truth to learn a 2nd-level predictor that ensembles the results from 1st-level modules for the final estimation. Preliminary results by testing the proposed framework on the LVQuan18 dataset show superior performance of the ensemble learning model over the two base modules."
                    },
                    {
                        "title": "SP$^2$OT: Semantic-Regularized Progressive Partial Optimal Transport for Imbalanced Clustering",
                        "abstract": "Deep clustering, which learns representation and semantic clustering without labels information, poses a great challenge for deep learning-based approaches. Despite significant progress in recent years, most existing methods focus on uniformly distributed datasets, significantly limiting the practical applicability of their methods. In this paper, we propose a more practical problem setting named deep imbalanced clustering, where the underlying classes exhibit an imbalance distribution. To address this challenge, we introduce a novel optimal transport-based pseudo-label learning framework. Our framework formulates pseudo-label generation as a Semantic-regularized Progressive Partial Optimal Transport (SP$^2$OT) problem, which progressively transports each sample to imbalanced clusters under several prior distribution and semantic relation constraints, thus generating high-quality and imbalance-aware pseudo-labels. To solve SP$^2$OT, we develop a Majorization-Minimization-based optimization algorithm. To be more precise, we employ the strategy of majorization to reformulate the SP$^2$OT problem into a Progressive Partial Optimal Transport problem, which can be transformed into an unbalanced optimal transport problem with augmented constraints and can be solved efficiently by a fast matrix scaling algorithm. Experiments on various datasets, including a human-curated long-tailed CIFAR100, challenging ImageNet-R, and large-scale subsets of fine-grained iNaturalist2018 datasets, demonstrate the superiority of our method."
                    },
                    {
                        "title": "P$^2$OT: Progressive Partial Optimal Transport for Deep Imbalanced Clustering",
                        "abstract": "Deep clustering, which learns representation and semantic clustering without labels information, poses a great challenge for deep learning-based approaches. Despite significant progress in recent years, most existing methods focus on uniformly distributed datasets, significantly limiting the practical applicability of their methods. In this paper, we first introduce a more practical problem setting named deep imbalanced clustering, where the underlying classes exhibit an imbalance distribution. To tackle this problem, we propose a novel pseudo-labeling-based learning framework. Our framework formulates pseudo-label generation as a progressive partial optimal transport problem, which progressively transports each sample to imbalanced clusters under prior distribution constraints, thus generating imbalance-aware pseudo-labels and learning from high-confident samples. In addition, we transform the initial formulation into an unbalanced optimal transport problem with augmented constraints, which can be solved efficiently by a fast matrix scaling algorithm. Experiments on various datasets, including a human-curated long-tailed CIFAR100, challenging ImageNet-R, and large-scale subsets of fine-grained iNaturalist2018 datasets, demonstrate the superiority of our method."
                    },
                    {
                        "title": "Dual-level Adaptive Self-Labeling for Novel Class Discovery in Point Cloud Segmentation",
                        "abstract": "We tackle the novel class discovery in point cloud segmentation, which discovers novel classes based on the semantic knowledge of seen classes. Existing work proposes an online point-wise clustering method with a simplified equal class-size constraint on the novel classes to avoid degenerate solutions. However, the inherent imbalanced distribution of novel classes in point clouds typically violates the equal class-size constraint. Moreover, point-wise clustering ignores the rich spatial context information of objects, which results in less expressive representation for semantic segmentation. To address the above challenges, we propose a novel self-labeling strategy that adaptively generates high-quality pseudo-labels for imbalanced classes during model training. In addition, we develop a dual-level representation that incorporates regional consistency into the point-level classifier learning, reducing noise in generated segmentation. Finally, we conduct extensive experiments on two widely used datasets, SemanticKITTI and SemanticPOSS, and the results show our method outperforms the state of the art by a large margin."
                    }
                ]
            },
            "79ea939d-f0f8-4467-a17f-b1ad59bfafaf": {
                "pk": "79ea939d-f0f8-4467-a17f-b1ad59bfafaf",
                "name": "Quanzheng Li",
                "collaborators": [
                    "Xiang Li",
                    "Mo Zhang",
                    "Dufan Wu",
                    "Jie Zhao",
                    "Li Zhang",
                    "Yutong Xie",
                    "Jiang Hu",
                    "Hui Ren",
                    "Na Li",
                    "Kyungsang Kim"
                ],
                "domain": [
                    "Medical Imaging",
                    "Deep Learning",
                    "Image Segmentation",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "Measurement-conditioned Denoising Diffusion Probabilistic Model for Under-sampled Medical Image Reconstruction",
                        "abstract": "We propose a novel and unified method, measurement-conditioned denoising diffusion probabilistic model (MC-DDPM), for under-sampled medical image reconstruction based on DDPM. Different from previous works, MC-DDPM is defined in measurement domain (e.g. k-space in MRI reconstruction) and conditioned on under-sampling mask. We apply this method to accelerate MRI reconstruction and the experimental results show excellent performance, outperforming full supervision baseline and the state-of-the-art score-based reconstruction method. Due to its generative nature, MC-DDPM can also quantify the uncertainty of reconstruction. Our code is available on github."
                    },
                    {
                        "title": "MS-GWNN:multi-scale graph wavelet neural network for breast cancer diagnosis",
                        "abstract": "Breast cancer is one of the most common cancers in women worldwide, and early detection can significantly reduce the mortality rate of breast cancer. It is crucial to take multi-scale information of tissue structure into account in the detection of breast cancer. And thus, it is the key to design an accurate computer-aided detection (CAD) system to capture multi-scale contextual features in a cancerous tissue. In this work, we present a novel graph convolutional neural network for histopathological image classification of breast cancer. The new method, named multi-scale graph wavelet neural network (MS-GWNN), leverages the localization property of spectral graph wavelet to perform multi-scale analysis. By aggregating features at different scales, MS-GWNN can encode the multi-scale contextual interactions in the whole pathological slide. Experimental results on two public datasets demonstrate the superiority of the proposed method. Moreover, through ablation studies, we find that multi-scale analysis has a significant impact on the accuracy of cancer diagnosis."
                    },
                    {
                        "title": "AdaFish: Fast low-rank parameter-efficient fine-tuning by using second-order information",
                        "abstract": "Recent advancements in large-scale pretrained models have significantly improved performance across a variety of tasks in natural language processing and computer vision. However, the extensive number of parameters in these models necessitates substantial memory and computational resources for full training. To adapt these models for downstream tasks or specific application-oriented datasets, parameter-efficient fine-tuning methods leveraging pretrained parameters have gained considerable attention. However, it can still be time-consuming due to lots of parameters and epochs. In this work, we introduce AdaFish, an efficient algorithm of the second-order type designed to expedite the training process within low-rank decomposition-based fine-tuning frameworks. Our key observation is that the associated generalized Fisher information matrix is either low-rank or extremely small-scaled. Such a generalized Fisher information matrix is shown to be equivalent to the Hessian matrix. Moreover, we prove the global convergence of AdaFish, along with its iteration/oracle complexity. Numerical experiments show that our algorithm is quite competitive with the state-of-the-art AdamW method."
                    },
                    {
                        "title": "Multi-Estimator Full Left Ventricle Quantification through Ensemble Learning",
                        "abstract": "Cardiovascular disease accounts for 1 in every 4 deaths in United States. Accurate estimation of structural and functional cardiac parameters is crucial for both diagnosis and disease management. In this work, we develop an ensemble learning framework for more accurate and robust left ventricle (LV) quantification. The framework combines two 1st-level modules: direct estimation module and a segmentation module. The direct estimation module utilizes Convolutional Neural Network (CNN) to achieve end-to-end quantification. The CNN is trained by taking 2D cardiac images as input and cardiac parameters as output. The segmentation module utilizes a U-Net architecture for obtaining pixel-wise prediction of the epicardium and endocardium of LV from the background. The binary U-Net output is then analyzed by a separate CNN for estimating the cardiac parameters. We then employ linear regression between the 1st-level predictor and ground truth to learn a 2nd-level predictor that ensembles the results from 1st-level modules for the final estimation. Preliminary results by testing the proposed framework on the LVQuan18 dataset show superior performance of the ensemble learning model over the two base modules."
                    },
                    {
                        "title": "APP-RUSS: Automated Path Planning for Robotic Ultrasound System",
                        "abstract": "Autonomous robotic ultrasound System (RUSS) has been extensively studied. However, fully automated ultrasound image acquisition is still challenging, partly due to the lack of study in combining two phases of path planning: guiding the ultrasound probe to the scan target and covering the scan surface or volume. This paper presents a system of Automated Path Planning for RUSS (APP-RUSS). Our focus is on the first phase of automation, which emphasizes directing the ultrasound probe's path toward the target over extended distances. Specifically, our APP-RUSS system consists of a RealSense D405 RGB-D camera that is employed for visual guidance of the UR5e robotic arm and a cubic Bezier curve path planning model that is customized for delivering the probe to the recognized target. APP-RUSS can contribute to understanding the integration of the two phases of path planning in robotic ultrasound imaging, paving the way for its clinical adoption."
                    },
                    {
                        "title": "Image Segmentation and Classification for Sickle Cell Disease using Deformable U-Net",
                        "abstract": "Reliable cell segmentation and classification from biomedical images is a crucial step for both scientific research and clinical practice. A major challenge for more robust segmentation and classification methods is the large variations in the size, shape and viewpoint of the cells, combining with the low image quality caused by noise and artifacts. To address this issue, in this work we propose a learning-based, simultaneous cell segmentation and classification method based on the deep U-Net structure with deformable convolution layers. The U-Net architecture for deep learning has been shown to offer a precise localization for image semantic segmentation. Moreover, deformable convolution layer enables the free form deformation of the feature learning process, thus makes the whole network more robust to various cell morphologies and image settings. The proposed method is tested on microscopic red blood cell images from patients with sickle cell disease. The results show that U-Net with deformable convolution achieves the highest accuracy for segmentation and classification, comparing with original U-Net structure."
                    },
                    {
                        "title": "Decentralized Riemannian natural gradient methods with Kronecker-product approximations",
                        "abstract": "With a computationally efficient approximation of the second-order information, natural gradient methods have been successful in solving large-scale structured optimization problems. We study the natural gradient methods for the large-scale decentralized optimization problems on Riemannian manifolds, where the local objective function defined by the local dataset is of a log-probability type. By utilizing the structure of the Riemannian Fisher information matrix (RFIM), we present an efficient decentralized Riemannian natural gradient descent (DRNGD) method. To overcome the communication issue of the high-dimension RFIM, we consider a class of structured problems for which the RFIM can be approximated by a Kronecker product of two low-dimension matrices. By performing the communications over the Kronecker factors, a high-quality approximation of the RFIM can be obtained in a low cost. We prove that DRNGD converges to a stationary point with the best-known rate of $\\mathcal{O}(1/K)$. Numerical experiments demonstrate the efficiency of our proposed method compared with the state-of-the-art ones. To the best of our knowledge, this is the first Riemannian second-order method for solving decentralized manifold optimization problems."
                    },
                    {
                        "title": "Consensus Neural Network for Medical Imaging Denoising with Only Noisy Training Samples",
                        "abstract": "Deep neural networks have been proved efficient for medical image denoising. Current training methods require both noisy and clean images. However, clean images cannot be acquired for many practical medical applications due to naturally noisy signal, such as dynamic imaging, spectral computed tomography, arterial spin labeling magnetic resonance imaging, etc. In this paper we proposed a training method which learned denoising neural networks from noisy training samples only. Training data in the acquisition domain was split to two subsets and the network was trained to map one noisy set to the other. A consensus loss function was further proposed to efficiently combine the outputs from both subsets. A mathematical proof was provided that the proposed training scheme was equivalent to training with noisy and clean samples when the noise in the two subsets was uncorrelated and zero-mean. The method was validated on Low-dose CT Challenge dataset and NYU MRI dataset and achieved improved performance compared to existing unsupervised methods."
                    },
                    {
                        "title": "A General Computational Framework to Measure the Expressiveness of Complex Networks Using a Tighter Upper Bound of Linear Regions",
                        "abstract": "The expressiveness of deep neural network (DNN) is a perspective to understandthe surprising performance of DNN. The number of linear regions, i.e. pieces thata piece-wise-linear function represented by a DNN, is generally used to measurethe expressiveness. And the upper bound of regions number partitioned by a rec-tifier network, instead of the number itself, is a more practical measurement ofexpressiveness of a rectifier DNN. In this work, we propose a new and tighter up-per bound of regions number. Inspired by the proof of this upper bound and theframework of matrix computation in Hinz & Van de Geer (2019), we propose ageneral computational approach to compute a tight upper bound of regions numberfor theoretically any network structures (e.g. DNN with all kind of skip connec-tions and residual structures). Our experiments show our upper bound is tighterthan existing ones, and explain why skip connections and residual structures canimprove network performance."
                    },
                    {
                        "title": "IFGAN: Missing Value Imputation using Feature-specific Generative Adversarial Networks",
                        "abstract": "Missing value imputation is a challenging and well-researched topic in data mining. In this paper, we propose IFGAN, a missing value imputation algorithm based on Feature-specific Generative Adversarial Networks (GAN). Our idea is intuitive yet effective: a feature-specific generator is trained to impute missing values, while a discriminator is expected to distinguish the imputed values from observed ones. The proposed architecture is capable of handling different data types, data distributions, missing mechanisms, and missing rates. It also improves post-imputation analysis by preserving inter-feature correlations. We empirically show on several real-life datasets that IFGAN outperforms current state-of-the-art algorithm under various missing conditions."
                    },
                    {
                        "title": "Computationally Efficient Deep Neural Network for Computed Tomography Image Reconstruction",
                        "abstract": "Deep-neural-network-based image reconstruction has demonstrated promising performance in medical imaging for under-sampled and low-dose scenarios. However, it requires large amount of memory and extensive time for the training. It is especially challenging to train the reconstruction networks for three-dimensional computed tomography (CT) because of the high resolution of CT images. The purpose of this work is to reduce the memory and time consumption of the training of the reconstruction networks for CT to make it practical for current hardware, while maintaining the quality of the reconstructed images.   We unrolled the proximal gradient descent algorithm for iterative image reconstruction to finite iterations and replaced the terms related to the penalty function with trainable convolutional neural networks (CNN). The network was trained greedily iteration by iteration in the image-domain on patches, which requires reasonable amount of memory and time on mainstream graphics processing unit (GPU). To overcome the local-minimum problem caused by greedy learning, we used deep UNet as the CNN and incorporated separable quadratic surrogate with ordered subsets for data fidelity, so that the solution could escape from easy local minimums and achieve better image quality.   The proposed method achieved comparable image quality with state-of-the-art neural network for CT image reconstruction on 2D sparse-view and limited-angle problems on the low-dose CT challenge dataset."
                    },
                    {
                        "title": "Self-supervised Dynamic CT Perfusion Image Denoising with Deep Neural Networks",
                        "abstract": "Dynamic computed tomography perfusion (CTP) imaging is a promising approach for acute ischemic stroke diagnosis and evaluation. Hemodynamic parametric maps of cerebral parenchyma are calculated from repeated CT scans of the first pass of iodinated contrast through the brain. It is necessary to reduce the dose of CTP for routine applications due to the high radiation exposure from the repeated scans, where image denoising is necessary to achieve a reliable diagnosis. In this paper, we proposed a self-supervised deep learning method for CTP denoising, which did not require any high-dose reference images for training. The network was trained by mapping each frame of CTP to an estimation from its adjacent frames. Because the noise in the source and target was independent, this approach could effectively remove the noise. Being free from high-dose training images granted the proposed method easier adaptation to different scanning protocols. The method was validated on both simulation and a public real dataset. The proposed method achieved improved image quality compared to conventional denoising methods. On the real data, the proposed method also had improved spatial resolution and contrast-to-noise ratio compared to supervised learning which was trained on the simulation data"
                    },
                    {
                        "title": "Beyond Finite Layer Neural Networks: Bridging Deep Architectures and Numerical Differential Equations",
                        "abstract": "In our work, we bridge deep neural network design with numerical differential equations. We show that many effective networks, such as ResNet, PolyNet, FractalNet and RevNet, can be interpreted as different numerical discretizations of differential equations. This finding brings us a brand new perspective on the design of effective deep architectures. We can take advantage of the rich knowledge in numerical analysis to guide us in designing new and potentially more effective deep networks. As an example, we propose a linear multi-step architecture (LM-architecture) which is inspired by the linear multi-step method solving ordinary differential equations. The LM-architecture is an effective structure that can be used on any ResNet-like networks. In particular, we demonstrate that LM-ResNet and LM-ResNeXt (i.e. the networks obtained by applying the LM-architecture on ResNet and ResNeXt respectively) can achieve noticeably higher accuracy than ResNet and ResNeXt on both CIFAR and ImageNet with comparable numbers of trainable parameters. In particular, on both CIFAR and ImageNet, LM-ResNet/LM-ResNeXt can significantly compress ($>50$\\%) the original networks while maintaining a similar performance. This can be explained mathematically using the concept of modified equation from numerical analysis. Last but not least, we also establish a connection between stochastic control and noise injection in the training process which helps to improve generalization of the networks. Furthermore, by relating stochastic training strategy with stochastic dynamic system, we can easily apply stochastic training to the networks with the LM-architecture. As an example, we introduced stochastic depth to LM-ResNet and achieve significant improvement over the original LM-ResNet on CIFAR10."
                    },
                    {
                        "title": "Automated Segmentation of Cervical Nuclei in Pap Smear Images using Deformable Multi-path Ensemble Model",
                        "abstract": "Pap smear testing has been widely used for detecting cervical cancers based on the morphology properties of cell nuclei in microscopic image. An accurate nuclei segmentation could thus improve the success rate of cervical cancer screening. In this work, a method of automated cervical nuclei segmentation using Deformable Multipath Ensemble Model (D-MEM) is proposed. The approach adopts a U-shaped convolutional network as a backbone network, in which dense blocks are used to transfer feature information more effectively. To increase the flexibility of the model, we then use deformable convolution to deal with different nuclei irregular shapes and sizes. To reduce the predictive bias, we further construct multiple networks with different settings, which form an ensemble model. The proposed segmentation framework has achieved state-of-the-art accuracy on Herlev dataset with Zijdenbos similarity index (ZSI) of 0.933, and has the potential to be extended for solving other medical image segmentation tasks."
                    },
                    {
                        "title": "Medical Image Segmentation Based on Multi-Modal Convolutional Neural Network: Study on Image Fusion Schemes",
                        "abstract": "Image analysis using more than one modality (i.e. multi-modal) has been increasingly applied in the field of biomedical imaging. One of the challenges in performing the multimodal analysis is that there exist multiple schemes for fusing the information from different modalities, where such schemes are application-dependent and lack a unified framework to guide their designs. In this work we firstly propose a conceptual architecture for the image fusion schemes in supervised biomedical image analysis: fusing at the feature level, fusing at the classifier level, and fusing at the decision-making level. Further, motivated by the recent success in applying deep learning for natural image analysis, we implement the three image fusion schemes above based on the Convolutional Neural Network (CNN) with varied structures, and combined into a single framework. The proposed image segmentation framework is capable of analyzing the multi-modality images using different fusing schemes simultaneously. The framework is applied to detect the presence of soft tissue sarcoma from the combination of Magnetic Resonance Imaging (MRI), Computed Tomography (CT) and Positron Emission Tomography (PET) images. It is found from the results that while all the fusion schemes outperform the single-modality schemes, fusing at the feature level can generally achieve the best performance in terms of both accuracy and computational cost, but also suffers from the decreased robustness in the presence of large errors in any image modalities."
                    },
                    {
                        "title": "ASCNet: Adaptive-Scale Convolutional Neural Networks for Multi-Scale Feature Learning",
                        "abstract": "Extracting multi-scale information is key to semantic segmentation. However, the classic convolutional neural networks (CNNs) encounter difficulties in achieving multi-scale information extraction: expanding convolutional kernel incurs the high computational cost and using maximum pooling sacrifices image information. The recently developed dilated convolution solves these problems, but with the limitation that the dilation rates are fixed and therefore the receptive field cannot fit for all objects with different sizes in the image. We propose an adaptivescale convolutional neural network (ASCNet), which introduces a 3-layer convolution structure in the end-to-end training, to adaptively learn an appropriate dilation rate for each pixel in the image. Such pixel-level dilation rates produce optimal receptive fields so that the information of objects with different sizes can be extracted at the corresponding scale. We compare the segmentation results using the classic CNN, the dilated CNN and the proposed ASCNet on two types of medical images (The Herlev dataset and SCD RBC dataset). The experimental results show that ASCNet achieves the highest accuracy. Moreover, the automatically generated dilation rates are positively correlated to the sizes of the objects, confirming the effectiveness of the proposed method."
                    },
                    {
                        "title": "Dictionary Learning and Sparse Coding-based Denoising for High-Resolution Task Functional Connectivity MRI Analysis",
                        "abstract": "We propose a novel denoising framework for task functional Magnetic Resonance Imaging (tfMRI) data to delineate the high-resolution spatial pattern of the brain functional connectivity via dictionary learning and sparse coding (DLSC). In order to address the limitations of the unsupervised DLSC-based fMRI studies, we utilize the prior knowledge of task paradigm in the learning step to train a data-driven dictionary and to model the sparse representation. We apply the proposed DLSC-based method to Human Connectome Project (HCP) motor tfMRI dataset. Studies on the functional connectivity of cerebrocerebellar circuits in somatomotor networks show that the DLSC-based denoising framework can significantly improve the prominent connectivity patterns, in comparison to the temporal non-local means (tNLM)-based denoising method as well as the case without denoising, which is consistent and neuroscientifically meaningful within motor area. The promising results show that the proposed method can provide an important foundation for the high-resolution functional connectivity analysis, and provide a better approach for fMRI preprocessing."
                    },
                    {
                        "title": "Learning the Sparse and Low Rank PARAFAC Decomposition via the Elastic Net",
                        "abstract": "In this article, we derive a Bayesian model to learning the sparse and low rank PARAFAC decomposition for the observed tensor with missing values via the elastic net, with property to find the true rank and sparse factor matrix which is robust to the noise. We formulate efficient block coordinate descent algorithm and admax stochastic block coordinate descent algorithm to solve it, which can be used to solve the large scale problem. To choose the appropriate rank and sparsity in PARAFAC decomposition, we will give a solution path by gradually increasing the regularization to increase the sparsity and decrease the rank. When we find the sparse structure of the factor matrix, we can fixed the sparse structure, using a small to regularization to decreasing the recovery error, and one can choose the proper decomposition from the solution path with sufficient sparse factor matrix with low recovery error. We test the power of our algorithm on the simulation data and real data, which show it is powerful."
                    },
                    {
                        "title": "BEFD: Boundary Enhancement and Feature Denoising for Vessel Segmentation",
                        "abstract": "Blood vessel segmentation is crucial for many diagnostic and research applications. In recent years, CNN-based models have leaded to breakthroughs in the task of segmentation, however, such methods usually lose high-frequency information like object boundaries and subtle structures, which are vital to vessel segmentation. To tackle this issue, we propose Boundary Enhancement and Feature Denoising (BEFD) module to facilitate the network ability of extracting boundary information in semantic segmentation, which can be integrated into arbitrary encoder-decoder architecture in an end-to-end way. By introducing Sobel edge detector, the network is able to acquire additional edge prior, thus enhancing boundary in an unsupervised manner for medical image segmentation. In addition, we also utilize a denoising block to reduce the noise hidden in the low-level features. Experimental results on retinal vessel dataset and angiocarpy dataset demonstrate the superior performance of the new BEFD module."
                    },
                    {
                        "title": "FedDAR: Federated Domain-Aware Representation Learning",
                        "abstract": "Cross-silo Federated learning (FL) has become a promising tool in machine learning applications for healthcare. It allows hospitals/institutions to train models with sufficient data while the data is kept private. To make sure the FL model is robust when facing heterogeneous data among FL clients, most efforts focus on personalizing models for clients. However, the latent relationships between clients' data are ignored. In this work, we focus on a special non-iid FL problem, called Domain-mixed FL, where each client's data distribution is assumed to be a mixture of several predefined domains. Recognizing the diversity of domains and the similarity within domains, we propose a novel method, FedDAR, which learns a domain shared representation and domain-wise personalized prediction heads in a decoupled manner. For simplified linear regression settings, we have theoretically proved that FedDAR enjoys a linear convergence rate. For general settings, we have performed intensive empirical studies on both synthetic and real-world medical datasets which demonstrate its superiority over prior FL methods."
                    }
                ]
            },
            "3c0bbed7-5759-4d41-a02b-56a3bb198c0c": {
                "pk": "3c0bbed7-5759-4d41-a02b-56a3bb198c0c",
                "name": "Xiang Li",
                "collaborators": [
                    "Jiguang Bao",
                    "Guanrong Chen",
                    "Yucheng Zhou",
                    "Yang Liu"
                ],
                "domain": [
                    "Complex Networks",
                    "Synchronization",
                    "Cloud Computing",
                    "Nonlinear Dynamics"
                ],
                "publications": [
                    {
                        "title": "Uniform synchronous criticality of diversely random complex networks",
                        "abstract": "We investigate collective synchronous behaviors in random complex networks of limit-cycle oscillators with the non-identical asymmetric coupling scheme, and find a uniform coupling criticality of collective synchronization which is independent of complexity of network topologies. Numerically simulations on categories of random complex networks have verified this conclusion."
                    },
                    {
                        "title": "Comment on \"Comparison of six simulation codes for positive streamers in air\"(Plasma Sources Sci. Technol. 27 (2018) 095002)",
                        "abstract": "Recently, a comparison of six codes for streamer discharge simulations were performed in [1]. In this comment, we discuss about the big differences between the results obtained by the different codes using the same deterministic model, and raise questions on the convergence of the codes and the minimum spatial resolution that are required for a converged results."
                    },
                    {
                        "title": "Sync in Complex Dynamical Networks: Stability, Evolution, Control, and Application",
                        "abstract": "In the past few years, the discoveries of small-world and scale-free properties of many natural and artificial complex networks have stimulated significant advances in better understanding the relationship between the topology and the collective dynamics of complex networks. This paper reports recent progresses in the literature of synchronization of complex dynamical networks including stability criteria, network synchronizability and uniform synchronous criticality in different topologies, and the connection between control and synchronization of complex networks as well. The economic-cycle synchronous phenomenon in the World Trade Web, a scale-free type of social economic networks, is used to illustrate an application of the network synchronization mechanism."
                    },
                    {
                        "title": "The stable property of Newton slopes for general Witt towers",
                        "abstract": "Any polynomial $f(x)\\in\\mathbb{Z}_q[x]$ defines a Witt vector $[f]\\in W(\\mathbb{F}_q[x])$. Consider the Artin-Schreier-Witt tower $y^F-y=[f]$. This is a tower of curves over $\\mathbb{F}_q$, with total Galois group $\\mathbb{Z}_p$. We want to study the Newton slopes of zeta functions of this tower. We reduce it to the Newton polygons of L-functions associated with characters on the Galois groups. We prove that, when the conductors are large enough, these Newton slopes are unions of arithmetic progressions which are changing proportionally as the conductor increases. This is a generalization of the result of \\cite{Da}, where they get the same result in the case the non-zero coefficients of $f(x)$ are roots of unity. To overcome the new difficulty in our process, we apply some $(p^{\\theta},T)$-topology."
                    },
                    {
                        "title": "FASTCloud: A novel framework of assessment and selection for trustworthy cloud service",
                        "abstract": "By virtue of technology and benefit advantages, cloud computing has increasingly attracted a large number of potential cloud consumers (PCC) plan to migrate the traditional business to the cloud service. However, trust has become one of the most challenging issues that prevent the PCC from adopting cloud services, especially in trustworthy cloud service selection. Besides, due to the diversity and dynamic of quality of service (QoS) in the cloud environment, the existing trust assessment methods based on the single constant value of QoS attribute and the subjective weight assignment are not good enough to provide an effective solution for PCCs to identify and select a trustworthy cloud service among a wide range of functionally-equivalent cloud service providers (CSPs). To address the challenge, a novel assessment and selection framework for trustworthy cloud service, FASTCloud, is proposed in this study. This framework facilitates PCCs to select a trustworthy cloud service based on their actual QoS requirements. In order to accurately and efficiently assess the trust level of cloud services, a QoS-based trust assessment model is proposed. This model represents a trust level assessment method based on the interval multiple attributes with an objective weight assignment method based on the deviation maximization to adaptively determine the trust level of different cloud services provisioned by candidate CSPs. The advantage of the proposed trust level assessment method in time complexity is demonstrated by the performance analysis and comparison. The experimental result of a case study with an open-source dataset shows that the trust model is efficient in cloud service trust assessment and the FASTCloud can effectively help PCCs select a trustworthy cloud service."
                    },
                    {
                        "title": "Existence and asymptotic behavior of entire large solutions for Hessian equations",
                        "abstract": "In this paper, we give some existence and nonexistence results for nonradial entire large solutions of the Hessian equation $S_k\\left(D^2 u\\right)=b(x) u^\\gamma$ in the sublinear case $0<\\gamma<k$. The exact asymptotic behavior of large solutions at infinity is also studied when $b(x)$ is the oscillation of a radial function $|x|^{-l}$ at infinity for $l\\leq k-1$."
                    },
                    {
                        "title": "Controlling the spreading in small-world networks",
                        "abstract": "The spreading (propagation) of diseases, viruses, and disasters such as power blackout through a huge-scale and complex network is one of the most concerned issues today. In this paper, we study the control of such spreading in a nonlinear spreading model of small-world networks. We found that the short-cut adding probability $p$ in the N-W model \\cite{N-W:1999} of small-world networks determines the Hopf bifurcation and other bifurcating behaviors in the proposed model. We further show a control technique that stabilize a periodic spreading behavior onto a stable equilibrium over the proposed model of small-world networks."
                    },
                    {
                        "title": "Disentangled and Robust Representation Learning for Bragging Classification in Social Media",
                        "abstract": "Researching bragging behavior on social media arouses interest of computational (socio) linguists. However, existing bragging classification datasets suffer from a serious data imbalance issue. Because labeling a data-balance dataset is expensive, most methods introduce external knowledge to improve model learning. Nevertheless, such methods inevitably introduce noise and non-relevance information from external knowledge. To overcome the drawback, we propose a novel bragging classification method with disentangle-based representation augmentation and domain-aware adversarial strategy. Specifically, model learns to disentangle and reconstruct representation and generate augmented features via disentangle-based representation augmentation. Moreover, domain-aware adversarial strategy aims to constrain domain of augmented features to improve their robustness. Experimental results demonstrate that our method achieves state-of-the-art performance compared to other methods."
                    },
                    {
                        "title": "Simulation of adaptive feedforward control for magnetic alloy cavity",
                        "abstract": "The upgrade plan of the China Spallation Neutron Source aims to enhance the beam power from 100 kW to 500 kW. To achieve this, the plan involves incorporating three new magnetic alloy cavities while maintaining the existing system to enable double harmonic acceleration. As a consequence of the increased current intensity, the beam loading effect will be significantly amplified, presenting a considerable challenge for the low-level RF control system of the magnetic alloy cavity. To address this challenge, an adaptive feedforward algorithm has been developed to enable optimal control. In addition, comprehensive simulations of the algorithm have been successfully conducted to validate its."
                    }
                ]
            },
            "d30b11ff-ce36-4d1a-a10c-c1afb2db202f": {
                "pk": "d30b11ff-ce36-4d1a-a10c-c1afb2db202f",
                "name": "Carl Yang",
                "collaborators": [
                    "Hejie Cui",
                    "Aydin Buluc",
                    "John D. Owens",
                    "Jiawei Han",
                    "Xuan Kan",
                    "Jiaying Lu",
                    "Yi Yang",
                    "Kevin Chen-Chuan Chang",
                    "Jieyu Zhang",
                    "Ying Guo"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Neuroscience",
                    "Machine Learning",
                    "Semi-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "PTGB: Pre-Train Graph Neural Networks for Brain Network Analysis",
                        "abstract": "The human brain is the central hub of the neurobiological system, controlling behavior and cognition in complex ways. Recent advances in neuroscience and neuroimaging analysis have shown a growing interest in the interactions between brain regions of interest (ROIs) and their impact on neural development and disorder diagnosis. As a powerful deep model for analyzing graph-structured data, Graph Neural Networks (GNNs) have been applied for brain network analysis. However, training deep models requires large amounts of labeled data, which is often scarce in brain network datasets due to the complexities of data acquisition and sharing restrictions. To make the most out of available training data, we propose PTGB, a GNN pre-training framework that captures intrinsic brain network structures, regardless of clinical outcomes, and is easily adaptable to various downstream tasks. PTGB comprises two key components: (1) an unsupervised pre-training technique designed specifically for brain networks, which enables learning from large-scale datasets without task-specific labels; (2) a data-driven parcellation atlas mapping pipeline that facilitates knowledge transfer across datasets with different ROI systems. Extensive evaluations using various GNN models have demonstrated the robust and superior performance of PTGB compared to baseline methods."
                    },
                    {
                        "title": "Relationship Profiling over Social Networks: Reverse Smoothness from Similarity to Closeness",
                        "abstract": "On social networks, while nodes bear rich attributes, we often lack the `semantics' of why each link is formed-- and thus we are missing the `road signs' to navigate and organize the complex social universe. How to identify relationship semantics without labels? Founded on the prevalent homophily principle, we propose the novel problem of Attribute-based Relationship Profiling (ARP), to profile the closeness w.r.t. the underlying relationships (e.g., schoolmate) between users based on their similarity in the corresponding attributes (e.g., education) and, as output, learn a set of social affinity graphs, where each link is weighted by its probabilities of carrying the relationships. As requirements, ARP should be systematic and complete to profile every link for every relationship-- our challenges lie in effectively modeling homophily. We propose a novel reverse smoothness principle by observing that the similarity-closeness duality of homophily is consistent with the well-known smoothness assumption in graph-based semi-supervised learning-- only the direction of inference is reversed. To realize smoothness over noisy social graphs, we further propose a novel holistic closeness modeling approach to capture `high-order' smoothness by extending closeness from edges to paths. Extensive experiments on three real-world datasets demonstrate the efficacy of ARP."
                    },
                    {
                        "title": "Design Principles for Sparse Matrix Multiplication on the GPU",
                        "abstract": "We implement two novel algorithms for sparse-matrix dense-matrix multiplication (SpMM) on the GPU. Our algorithms expect the sparse input in the popular compressed-sparse-row (CSR) format and thus do not require expensive format conversion. While previous SpMM work concentrates on thread-level parallelism, we additionally focus on latency hiding with instruction-level parallelism and load-balancing. We show, both theoretically and experimentally, that the proposed SpMM is a better fit for the GPU than previous approaches. We identify a key memory access pattern that allows efficient access into both input and output matrices that is crucial to getting excellent performance on SpMM. By combining these two ingredients---(i) merge-based load-balancing and (ii) row-major coalesced memory access---we demonstrate a 4.1x peak speedup and a 31.7% geomean speedup over state-of-the-art SpMM implementations on real-world datasets."
                    },
                    {
                        "title": "Implementing Push-Pull Efficiently in GraphBLAS",
                        "abstract": "We factor Beamer's push-pull, also known as direction-optimized breadth-first-search (DOBFS) into 3 separable optimizations, and analyze them for generalizability, asymptotic speedup, and contribution to overall speedup. We demonstrate that masking is critical for high performance and can be generalized to all graph algorithms where the sparsity pattern of the output is known a priori. We show that these graph algorithm optimizations, which together constitute DOBFS, can be neatly and separably described using linear algebra and can be expressed in the GraphBLAS linear-algebra-based framework. We provide experimental evidence that with these optimizations, a DOBFS expressed in a linear-algebra-based graph framework attains competitive performance with state-of-the-art graph frameworks on the GPU and on a multi-threaded CPU, achieving 101 GTEPS on a Scale 22 RMAT graph."
                    },
                    {
                        "title": "Neural Embedding Propagation on Heterogeneous Networks",
                        "abstract": "Classification is one of the most important problems in machine learning. To address label scarcity, semi-supervised learning (SSL) has been intensively studied over the past two decades, which mainly leverages data affinity modeled by networks. Label propagation (LP), however, as the most popular SSL technique, mostly only works on homogeneous networks with single-typed simple interactions. In this work, we focus on the more general and powerful heterogeneous networks, which accommodate multi-typed objects and links, and thus endure multi-typed complex interactions. Specifically, we propose \\textit{neural embedding propagation} (NEP), which leverages distributed embeddings to represent objects and dynamically composed modular networks to model their complex interactions. While generalizing LP as a simple instance, NEP is far more powerful in its natural awareness of different types of objects and links, and the ability to automatically capture their important interaction patterns. Further, we develop a series of efficient training strategies for NEP, leading to its easy deployment on real-world heterogeneous networks with millions of objects. With extensive experiments on three datasets, we comprehensively demonstrate the effectiveness, efficiency, and robustness of NEP compared with state-of-the-art network embedding and SSL algorithms."
                    },
                    {
                        "title": "Effective and Interpretable fMRI Analysis via Functional Brain Network Generation",
                        "abstract": "Recent studies in neuroscience show great potential of functional brain networks constructed from fMRI data for popularity modeling and clinical predictions. However, existing functional brain networks are noisy and unaware of downstream prediction tasks, while also incompatible with recent powerful machine learning models of GNNs. In this work, we develop an end-to-end trainable pipeline to extract prominent fMRI features, generate brain networks, and make predictions with GNNs, all under the guidance of downstream prediction tasks. Preliminary experiments on the PNC fMRI data show the superior effectiveness and unique interpretability of our framework."
                    },
                    {
                        "title": "I Know You'll Be Back: Interpretable New User Clustering and Churn Prediction on a Mobile Social Application",
                        "abstract": "As online platforms are striving to get more users, a critical challenge is user churn, which is especially concerning for new users. In this paper, by taking the anonymous large-scale real-world data from Snapchat as an example, we develop \\textit{ClusChurn}, a systematic two-step framework for interpretable new user clustering and churn prediction, based on the intuition that proper user clustering can help understand and predict user churn. Therefore, \\textit{ClusChurn} firstly groups new users into interpretable typical clusters, based on their activities on the platform and ego-network structures. Then we design a novel deep learning pipeline based on LSTM and attention to accurately predict user churn with very limited initial behavior data, by leveraging the correlations among users' multi-dimensional activities and the underlying user types. \\textit{ClusChurn} is also able to predict user types, which enables rapid reactions to different types of user churn. Extensive data analysis and experiments show that \\textit{ClusChurn} provides valuable insight into user behaviors, and achieves state-of-the-art churn prediction performance. The whole framework is deployed as a data analysis pipeline, delivering real-time data analysis and prediction results to multiple relevant teams for business intelligence uses. It is also general enough to be readily adopted by any online systems with user behavior data."
                    },
                    {
                        "title": "GraphBLAST: A High-Performance Linear Algebra-based Graph Framework on the GPU",
                        "abstract": "High-performance implementations of graph algorithms are challenging to implement on new parallel hardware such as GPUs because of three challenges: (1) the difficulty of coming up with graph building blocks, (2) load imbalance on parallel hardware, and (3) graph problems having low arithmetic intensity. To address some of these challenges, GraphBLAS is an innovative, on-going effort by the graph analytics community to propose building blocks based on sparse linear algebra, which will allow graph algorithms to be expressed in a performant, succinct, composable and portable manner. In this paper, we examine the performance challenges of a linear-algebra-based approach to building graph frameworks and describe new design principles for overcoming these bottlenecks. Among the new design principles is exploiting input sparsity, which allows users to write graph algorithms without specifying push and pull direction. Exploiting output sparsity allows users to tell the backend which values of the output in a single vectorized computation they do not want computed. Load-balancing is an important feature for balancing work amongst parallel workers. We describe the important load-balancing features for handling graphs with different characteristics. The design principles described in this paper have been implemented in \"GraphBLAST\", the first high-performance linear algebra-based graph framework on NVIDIA GPUs that is open-source. The results show that on a single GPU, GraphBLAST has on average at least an order of magnitude speedup over previous GraphBLAS implementations SuiteSparse and GBTL, comparable performance to the fastest GPU hardwired primitives and shared-memory graph frameworks Ligra and Gunrock, and better performance than any other GPU graph framework, while offering a simpler and more concise programming model."
                    },
                    {
                        "title": "Zero-Shot Scene Graph Relation Prediction through Commonsense Knowledge Integration",
                        "abstract": "Relation prediction among entities in images is an important step in scene graph generation (SGG), which further impacts various visual understanding and reasoning tasks. Existing SGG frameworks, however, require heavy training yet are incapable of modeling unseen (i.e.,zero-shot) triplets. In this work, we stress that such incapability is due to the lack of commonsense reasoning,i.e., the ability to associate similar entities and infer similar relations based on general understanding of the world. To fill this gap, we propose CommOnsense-integrAted sCenegrapHrElation pRediction (COACHER), a framework to integrate commonsense knowledge for SGG, especially for zero-shot relation prediction. Specifically, we develop novel graph mining pipelines to model the neighborhoods and paths around entities in an external commonsense knowledge graph, and integrate them on top of state-of-the-art SGG frameworks. Extensive quantitative evaluations and qualitative case studies on both original and manipulated datasets from Visual Genome demonstrate the effectiveness of our proposed approach."
                    },
                    {
                        "title": "Weakly Supervised Concept Map Generation through Task-Guided Graph Translation",
                        "abstract": "Recent years have witnessed the rapid development of concept map generation techniques due to their advantages in providing well-structured summarization of knowledge from free texts. Traditional unsupervised methods do not generate task-oriented concept maps, whereas deep generative models require large amounts of training data. In this work, we present GT-D2G (Graph Translation-based Document To Graph), an automatic concept map generation framework that leverages generalized NLP pipelines to derive semantic-rich initial graphs, and translates them into more concise structures under the weak supervision of downstream task labels. The concept maps generated by GT-D2G can provide interpretable summarization of structured knowledge for the input texts, which are demonstrated through human evaluation and case studies on three real-world corpora. Further experiments on the downstream task of document classification show that GT-D2G beats other concept map generation methods. Moreover, we specifically validate the labeling efficiency of GT-D2G in the label-efficient learning setting and the flexibility of generated graph sizes in controlled hyper-parameter studies."
                    },
                    {
                        "title": "How Can Graph Neural Networks Help Document Retrieval: A Case Study on CORD19 with Concept Map Generation",
                        "abstract": "Graph neural networks (GNNs), as a group of powerful tools for representation learning on irregular data, have manifested superiority in various downstream tasks. With unstructured texts represented as concept maps, GNNs can be exploited for tasks like document retrieval. Intrigued by how can GNNs help document retrieval, we conduct an empirical study on a large-scale multi-discipline dataset CORD-19. Results show that instead of the complex structure-oriented GNNs such as GINs and GATs, our proposed semantics-oriented graph functions achieve better and more stable performance based on the BM25 retrieved candidates. Our insights in this case study can serve as a guideline for future work to develop effective GNNs with appropriate semantics-oriented inductive biases for textual reasoning tasks like document retrieval and classification. All code for this case study is available at https://github.com/HennyJie/GNN-DocRetrieval."
                    },
                    {
                        "title": "Graph Auto-Encoder Via Neighborhood Wasserstein Reconstruction",
                        "abstract": "Graph neural networks (GNNs) have drawn significant research attention recently, mostly under the setting of semi-supervised learning. When task-agnostic representations are preferred or supervision is simply unavailable, the auto-encoder framework comes in handy with a natural graph reconstruction objective for unsupervised GNN training. However, existing graph auto-encoders are designed to reconstruct the direct links, so GNNs trained in this way are only optimized towards proximity-oriented graph mining tasks, and will fall short when the topological structures matter. In this work, we revisit the graph encoding process of GNNs which essentially learns to encode the neighborhood information of each node into an embedding vector, and propose a novel graph decoder to reconstruct the entire neighborhood information regarding both proximity and structure via Neighborhood Wasserstein Reconstruction (NWR). Specifically, from the GNN embedding of each node, NWR jointly predicts its node degree and neighbor feature distribution, where the distribution prediction adopts an optimal-transport loss based on the Wasserstein distance. Extensive experiments on both synthetic and real-world network datasets show that the unsupervised node representations learned with NWR have much more advantageous in structure-oriented graph mining tasks, while also achieving competitive performance in proximity-oriented ones."
                    },
                    {
                        "title": "A Bird's-Eye Tutorial of Graph Attention Architectures",
                        "abstract": "Graph Neural Networks (GNNs) have shown tremendous strides in performance for graph-structured problems especially in the domains of natural language processing, computer vision and recommender systems. Inspired by the success of the transformer architecture, there has been an ever-growing body of work on attention variants of GNNs attempting to advance the state of the art in many of these problems. Incorporating \"attention\" into graph mining has been viewed as a way to overcome the noisiness, heterogenity and complexity associated with graph-structured data as well as to encode soft-inductive bias. It is hence crucial and advantageous to study these variants from a bird's-eye view to assess their strengths and weaknesses. We provide a systematic and focused tutorial centered around attention based GNNs in a hope to benefit researchers dealing with graph-structured problems. Our tutorial looks at GNN variants from the point of view of the attention function and iteratively builds the reader's understanding of different graph attention variants."
                    }
                ]
            }
        }
    },
    "2410.14195": {
        "paper_data": {
            "title": "Rethinking Transformer for Long Contextual Histopathology Whole Slide Image Analysis",
            "url": "http://arxiv.org/abs/2410.14195v1",
            "arxiv_id": "2410.14195",
            "authors": [
                "Honglin Li",
                "Yunlong Zhang",
                "Pingyi Chen",
                "Zhongyi Shui",
                "Chenglu Zhu",
                "Lin Yang"
            ],
            "abstract": "Histopathology Whole Slide Image (WSI) analysis serves as the gold standard for clinical cancer diagnosis in the daily routines of doctors. To develop computer-aided diagnosis model for WSIs, previous methods typically employ Multi-Instance Learning to enable slide-level prediction given only slide-level labels. Among these models, vanilla attention mechanisms without pairwise interactions have traditionally been employed but are unable to model contextual information. More recently, self-attention models have been utilized to address this issue. To alleviate the computational complexity of long sequences in large WSIs, methods like HIPT use region-slicing, and TransMIL employs approximation of full self-attention. Both approaches suffer from suboptimal performance due to the loss of key information. Moreover, their use of absolute positional embedding struggles to effectively handle long contextual dependencies in shape-varying WSIs. In this paper, we first analyze how the low-rank nature of the long-sequence attention matrix constrains the representation ability of WSI modelling. Then, we demonstrate that the rank of attention matrix can be improved by focusing on local interactions via a local attention mask. Our analysis shows that the local mask aligns with the attention patterns in the lower layers of the Transformer. Furthermore, the local attention mask can be implemented during chunked attention calculation, reducing the quadratic computational complexity to linear with a small local bandwidth. Building on this, we propose a local-global hybrid Transformer for both computational acceleration and local-global information interactions modelling. Our method, Long-contextual MIL (LongMIL), is evaluated through extensive experiments on various WSI tasks to validate its superiority. Our code will be available at github.com/invoker-LL/Long-MIL.",
            "introduction": "   1 Introduction  Though digital pathology images have been widely used for Cancer diagnosis\u00a0[50; 65; 89; 94; 17; 44] and prognosis\u00a0[9; 12; 71] and gene expression\u00a0[82] via automatic computer-assisted analysis, the Giga-pixels of resolution, as large as 150,000\u00d7150,000150000150000150,000\\times 150,000150 , 000 \u00d7 150 , 000 pixels\u00a0[50; 12; 60] of Whole Slide Image (WSI), still poses great challenges on both annotation labelling and efficient computation for model training\u00a0[42]. Thus, previous methods\u00a0[9; 12; 42; 5; 40; 92; 93; 41] focus on developing annotation- & computational- efficient learning to cope with those problems by employing Multiple Instance Learning (MIL)\u00a0[51; 32] with only WSI-level supervision.   Currently, there are mainly three steps (or mainstream genres) of WSI analysis framework: 1) access better instance-level patch embedding via Self-supervised Learning\u00a0[30; 7; 9; 40; 77].   Figure 1:  Handling an extremely long sequence with a magnification of 20\u00d720\\times20 \u00d7 (or quadrupling to 40\u00d740\\times40 \u00d7) poses a significant challenge. The computational complexity of transformers, denoted as O\u2062(n2)\ud835\udc42superscript\ud835\udc5b2O(n^{2})italic_O ( italic_n start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ), becomes prohibitive in such cases, leading to computational explosion.    2) design WSI head architectures\u00a0[50; 65; 89] and train the head with frozen instance embedding. 3) fine-tune patch embedding with WSI level weak label for better task-specific results\u00a0[90; 42]. Here in this paper, we focus on the step-2 and uncovering that there are still some room for improvement: Firstly, the vanilla attention used in AB-MIL, DS-MIL, CLAM, etc.\u00a0[32; 40; 50; 89], despite its computational efficiency (compared to self-attention), is unable to model contextual or interaction information across instances within a WSI. These interactions, which play a crucial role in prediction decision-making\u00a0[11; 65] indeed, can be modelled via self-attention mechanism. However, the long sequence of WSI instances pose O\u2062(n2)\ud835\udc42superscript\ud835\udc5b2O(n^{2})italic_O ( italic_n start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ) computation complexity with self-attention (Fig. 1). Although this complexity can be alleviated by self-attention approximation methods like Nystr\u00f6mformer\u00a0[83; 75] used in TransMIL\u00a0[65], this approximation only get sub-optimal performance compared to self-attention as pointed out in\u00a0[20; 29]. Authors in HIPT\u00a0[9] mitigates the complexity by non-overlap large region slicing, but the interactions of instances from different region slicing are highly ignored (e.g. adjacent patches may be separated into two regions).   The above issues highlight a strong need for an effective and efficient Transformer for WSI modelling. To begin, we discuss the performance bottleneck of basic Transformer\u00a0[72] for WSI. Different to Vision Transformer\u00a0[23] for natural image modelling where the number of patched tokens are smaller than patch embedding size (e.g. ViT-Small-p16 with embedding size 384 attend on 196+1 tokens), the Transformer-based WSI model suffers severe low-rank bottleneck of attention matrix\u00a0[3; 22] given the long-sequence (n>>1024much-greater-than\ud835\udc5b1024n>>1024italic_n > > 1024) of WSI but limited embedding size (d\u22641024\ud835\udc511024d\\leq 1024italic_d \u2264 1024). We reveal this problem in WSI theoretically, thus finding that one self-attention layer with limited embedding size can not model local contexts and global interactions at the same time. By stacking multiple self-attention layers, we notice that the low layer focus more on local context (Fig. 2a) after training while high layer focus on global. However, the rank of the attention matrix is still limited (Fig. 2a+d), resulting in constrained performance.   We assert that the low-rank bottleneck causes the attention mechanism to become confused between",
            "references": [
                {
                    "title": "Large-scale cervical precancerous screening via AI-assisted cytology whole slide image analysis",
                    "abstract": "Cervical Cancer continues to be the leading gynecological malignancy, posing a persistent threat to women's health on a global scale. Early screening via cytology Whole Slide Image (WSI) diagnosis is critical to prevent this Cancer progression and improve survival rate, but pathologist's single test suffers inevitable false negative due to the immense number of cells that need to be reviewed within a WSI. Though computer-aided automated diagnostic models can serve as strong complement for pathologists, their effectiveness is hampered by the paucity of extensive and detailed annotations, coupled with the limited interpretability and robustness. These factors significantly hinder their practical applicability and reliability in clinical settings. To tackle these challenges, we develop an AI approach, which is a Scalable Technology for Robust and Interpretable Diagnosis built on Extensive data (STRIDE) of cervical cytology. STRIDE addresses the bottleneck of limited annotations by integrating patient-level labels with a small portion of cell-level labels through an end-to-end training strategy, facilitating scalable learning across extensive datasets. To further improve the robustness to real-world domain shifts of cytology slide-making and imaging, STRIDE employs color adversarial samples training that mimic staining and imaging variations. Lastly, to achieve pathologist-level interpretability for the trustworthiness in clinical settings, STRIDE can generate explanatory textual descriptions that simulates pathologists' diagnostic processes by cell image feature and textual description alignment. Conducting extensive experiments and evaluations in 183 medical centers with a dataset of 341,889 WSIs and 0.1 billion cells from cervical cytology patients, STRIDE has demonstrated a remarkable superiority over previous state-of-the-art techniques."
                },
                {
                    "title": "VMamba: Visual State Space Model",
                    "abstract": "Designing computationally efficient network architectures persists as an ongoing necessity in computer vision. In this paper, we transplant Mamba, a state-space language model, into VMamba, a vision backbone that works in linear time complexity. At the core of VMamba lies a stack of Visual State-Space (VSS) blocks with the 2D Selective Scan (SS2D) module. By traversing along four scanning routes, SS2D helps bridge the gap between the ordered nature of 1D selective scan and the non-sequential structure of 2D vision data, which facilitates the gathering of contextual information from various sources and perspectives. Based on the VSS blocks, we develop a family of VMamba architectures and accelerate them through a succession of architectural and implementation enhancements. Extensive experiments showcase VMamba's promising performance across diverse visual perception tasks, highlighting its advantages in input scaling efficiency compared to existing benchmark models. Source code is available at https://github.com/MzeroMiko/VMamba."
                },
                {
                    "title": "Gated Linear Attention Transformers with Hardware-Efficient Training",
                    "abstract": "Transformers with linear attention allow for efficient parallel training but can simultaneously be formulated as an RNN with 2D (matrix-valued) hidden states, thus enjoying linear-time inference complexity. However, linear attention generally underperforms ordinary softmax attention. Moreover, current implementations of linear attention lack I/O-awareness and are thus slower than highly optimized implementations of softmax attention. This work describes a hardware-efficient algorithm for linear attention that trades off memory movement against parallelizability. The resulting implementation, dubbed FLASHLINEARATTENTION, is faster than FLASHATTENTION-2 (Dao, 2023) as a standalone layer even on short sequence lengths (e.g., 1K). We then generalize this algorithm to a more expressive variant of linear attention with data-dependent gates. When used as a replacement for the standard attention layer in Transformers, the resulting gated linear attention (GLA) Transformer is found to perform competitively against the LLaMA-architecture Transformer (Touvron et al., 2023) as well recent linear-time-inference baselines such as RetNet (Sun et al., 2023a) and Mamba (Gu&Dao, 2023) on moderate-scale language modeling experiments. GLA Transformer is especially effective at length generalization, enabling a model trained on 2K to generalize to sequences longer than 20K without significant perplexity degradations. For training speed, the GLA Transformer has higher throughput than a similarly-sized Mamba model."
                },
                {
                    "title": "When an Image is Worth 1, 024 x 1, 024 Words: A Case Study in Computational Pathology",
                    "abstract": "This technical report presents LongViT, a vision Transformer that can process gigapixel images in an end-to-end manner. Specifically, we split the gigapixel image into a sequence of millions of patches and project them linearly into embeddings. LongNet is then employed to model the extremely long sequence, generating representations that capture both short-range and long-range dependencies. The linear computation complexity of LongNet, along with its distributed algorithm, enables us to overcome the constraints of both computation and memory. We apply LongViT in the field of computational pathology, aiming for cancer diagnosis and prognosis within gigapixel whole-slide images. Experimental results demonstrate that LongViT effectively encodes gigapixel images and outperforms previous state-of-the-art methods on cancer subtyping and survival prediction. Code and models will be available at https://aka.ms/LongViT."
                },
                {
                    "title": "Mamba: Linear-Time Sequence Modeling with Selective State Spaces",
                    "abstract": "Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5$\\times$ higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation."
                },
                {
                    "title": "Attention-Challenging Multiple Instance Learning for Whole Slide Image Classification",
                    "abstract": "In the application of Multiple Instance Learning (MIL) methods for Whole Slide Image (WSI) classification, attention mechanisms often focus on a subset of discriminative instances, which are closely linked to overfitting. To mitigate overfitting, we present Attention-Challenging MIL (ACMIL). ACMIL combines two techniques based on separate analyses for attention value concentration. Firstly, UMAP of instance features reveals various patterns among discriminative instances, with existing attention mechanisms capturing only some of them. To remedy this, we introduce Multiple Branch Attention (MBA) to capture more discriminative instances using multiple attention branches. Secondly, the examination of the cumulative value of Top-K attention scores indicates that a tiny number of instances dominate the majority of attention. In response, we present Stochastic Top-K Instance Masking (STKIM), which masks out a portion of instances with Top-K attention values and allocates their attention values to the remaining instances. The extensive experimental results on three WSI datasets with two pre-trained backbones reveal that our ACMIL outperforms state-of-the-art methods. Additionally, through heatmap visualization and UMAP visualization, this paper extensively illustrates ACMIL's effectiveness in suppressing attention value concentration and overcoming the overfitting challenge. The source code is available at \\url{https://github.com/dazhangyu123/ACMIL}."
                },
                {
                    "title": "Scaling Laws of RoPE-based Extrapolation",
                    "abstract": "The extrapolation capability of Large Language Models (LLMs) based on Rotary Position Embedding is currently a topic of considerable interest. The mainstream approach to addressing extrapolation with LLMs involves modifying RoPE by replacing 10000, the rotary base of $\\theta_n={10000}^{-2n/d}$ in the original RoPE, with a larger value and providing longer fine-tuning text. In this work, we first observe that fine-tuning a RoPE-based LLM with either a smaller or larger base in pre-training context length could significantly enhance its extrapolation performance. After that, we propose \\textbf{\\textit{Scaling Laws of RoPE-based Extrapolation}}, a unified framework from the periodic perspective, to describe the relationship between the extrapolation performance and base value as well as tuning context length. In this process, we also explain the origin of the RoPE-based extrapolation issue by \\textbf{\\textit{critical dimension for extrapolation}}. Besides these observations and analyses, we achieve extrapolation up to 1 million context length within only 16K training length on LLaMA2 7B and 13B."
                },
                {
                    "title": "RoFormer for Position Aware Multiple Instance Learning in Whole Slide Image Classification",
                    "abstract": "Whole slide image (WSI) classification is a critical task in computational pathology. However, the gigapixel-size of such images remains a major challenge for the current state of deep-learning. Current methods rely on multiple-instance learning (MIL) models with frozen feature extractors. Given the the high number of instances in each image, MIL methods have long assumed independence and permutation-invariance of patches, disregarding the tissue structure and correlation between patches. Recent works started studying this correlation between instances but the computational workload of such a high number of tokens remained a limiting factor. In particular, relative position of patches remains unaddressed. We propose to apply a straightforward encoding module, namely a RoFormer layer , relying on memory-efficient exact self-attention and relative positional encoding. This module can perform full self-attention with relative position encoding on patches of large and arbitrary shaped WSIs, solving the need for correlation between instances and spatial modeling of tissues. We demonstrate that our method outperforms state-of-the-art MIL models on three commonly used public datasets (TCGA-NSCLC, BRACS and Camelyon16)) on weakly supervised classification tasks. Code is available at https://github.com/Sanofi-Public/DDS-RoFormerMIL"
                },
                {
                    "title": "Efficient Streaming Language Models with Attention Sinks",
                    "abstract": "Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching previous tokens' Key and Value states (KV) consumes extensive memory. Secondly, popular LLMs cannot generalize to longer texts than the training sequence length. Window attention, where only the most recent KVs are cached, is a natural approach -- but we show that it fails when the text length surpasses the cache size. We observe an interesting phenomenon, namely attention sink, that keeping the KV of initial tokens will largely recover the performance of window attention. In this paper, we first demonstrate that the emergence of attention sink is due to the strong attention scores towards initial tokens as a\"sink\"even if they are not semantically important. Based on the above analysis, we introduce StreamingLLM, an efficient framework that enables LLMs trained with a finite length attention window to generalize to infinite sequence lengths without any fine-tuning. We show that StreamingLLM can enable Llama-2, MPT, Falcon, and Pythia to perform stable and efficient language modeling with up to 4 million tokens and more. In addition, we discover that adding a placeholder token as a dedicated attention sink during pre-training can further improve streaming deployment. In streaming settings, StreamingLLM outperforms the sliding window recomputation baseline by up to 22.2x speedup. Code and datasets are provided at https://github.com/mit-han-lab/streaming-llm."
                },
                {
                    "title": "LongLoRA: Efficient Fine-tuning of Long-Context Large Language Models",
                    "abstract": "We present LongLoRA, an efficient fine-tuning approach that extends the context sizes of pre-trained large language models (LLMs), with limited computation cost. Typically, training LLMs with long context sizes is computationally expensive, requiring extensive training hours and GPU resources. For example, training on the context length of 8192 needs 16x computational costs in self-attention layers as that of 2048. In this paper, we speed up the context extension of LLMs in two aspects. On the one hand, although dense global attention is needed during inference, fine-tuning the model can be effectively and efficiently done by sparse local attention. The proposed shifted sparse attention effectively enables context extension, leading to non-trivial computation saving with similar performance to fine-tuning with vanilla attention. Particularly, it can be implemented with only two lines of code in training, while being optional in inference. On the other hand, we revisit the parameter-efficient fine-tuning regime for context expansion. Notably, we find that LoRA for context extension works well under the premise of trainable embedding and normalization. LongLoRA combines this improved LoRA with S^2-Attn. LongLoRA demonstrates strong empirical results on various tasks on Llama2 models from 7B/13B to 70B. LongLoRA extends Llama2 7B from 4k context to 100k, or Llama2 70B to 32k on a single 8x A100 machine. LongLoRA extends models' context while retaining their original architectures, and is compatible with most existing techniques, like Flash-Attention2. In addition, we further conduct supervised fine-tuning with LongLoRA and our long instruction-following LongAlpaca dataset."
                },
                {
                    "title": "FLatten Transformer: Vision Transformer using Focused Linear Attention",
                    "abstract": "The quadratic computation complexity of self-attention has been a persistent challenge when applying Transformer models to vision tasks. Linear attention, on the other hand, offers a much more efficient alternative with its linear complexity by approximating the Softmax operation through carefully designed mapping functions. However, current linear attention approaches either suffer from significant performance degradation or introduce additional computation overhead from the mapping functions. In this paper, we propose a novel Focused Linear Attention module to achieve both high efficiency and expressiveness. Specifically, we first analyze the factors contributing to the performance degradation of linear attention from two perspectives: the focus ability and feature diversity. To overcome these limitations, we introduce a simple yet effective mapping function and an efficient rank restoration module to enhance the expressiveness of self-attention while maintaining low computation complexity. Extensive experiments show that our linear attention module is applicable to a variety of advanced vision Transformers, and achieves consistently improved performances on multiple benchmarks. Code is available at https://github.com/LeapLabTHU/FLatten-Transformer."
                },
                {
                    "title": "Towards a Visual-Language Foundation Model for Computational Pathology",
                    "abstract": "The accelerated adoption of digital pathology and advances in deep learning have enabled the development of powerful models for various pathology tasks across a diverse array of diseases and patient cohorts. However, model training is often difficult due to label scarcity in the medical domain and the model's usage is limited by the specific task and disease for which it is trained. Additionally, most models in histopathology leverage only image data, a stark contrast to how humans teach each other and reason about histopathologic entities. We introduce CONtrastive learning from Captions for Histopathology (CONCH), a visual-language foundation model developed using diverse sources of histopathology images, biomedical text, and notably over 1.17 million image-caption pairs via task-agnostic pretraining. Evaluated on a suite of 13 diverse benchmarks, CONCH can be transferred to a wide range of downstream tasks involving either or both histopathology images and text, achieving state-of-the-art performance on histology image classification, segmentation, captioning, text-to-image and image-to-text retrieval. CONCH represents a substantial leap over concurrent visual-language pretrained systems for histopathology, with the potential to directly facilitate a wide array of machine learning-based workflows requiring minimal or no further supervised fine-tuning."
                },
                {
                    "title": "Retentive Network: A Successor to Transformer for Large Language Models",
                    "abstract": "In this work, we propose Retentive Network (RetNet) as a foundation architecture for large language models, simultaneously achieving training parallelism, low-cost inference, and good performance. We theoretically derive the connection between recurrence and attention. Then we propose the retention mechanism for sequence modeling, which supports three computation paradigms, i.e., parallel, recurrent, and chunkwise recurrent. Specifically, the parallel representation allows for training parallelism. The recurrent representation enables low-cost $O(1)$ inference, which improves decoding throughput, latency, and GPU memory without sacrificing performance. The chunkwise recurrent representation facilitates efficient long-sequence modeling with linear complexity, where each chunk is encoded parallelly while recurrently summarizing the chunks. Experimental results on language modeling show that RetNet achieves favorable scaling results, parallel training, low-cost deployment, and efficient inference. The intriguing properties make RetNet a strong successor to Transformer for large language models. Code will be available at https://aka.ms/retnet."
                },
                {
                    "title": "Spatially Resolved Gene Expression Prediction from H&E Histology Images via Bi-modal Contrastive Learning",
                    "abstract": "Histology imaging is an important tool in medical diagnosis and research, enabling the examination of tissue structure and composition at the microscopic level. Understanding the underlying molecular mechanisms of tissue architecture is critical in uncovering disease mechanisms and developing effective treatments. Gene expression profiling provides insight into the molecular processes underlying tissue architecture, but the process can be time-consuming and expensive. We present BLEEP (Bi-modaL Embedding for Expression Prediction), a bi-modal embedding framework capable of generating spatially resolved gene expression profiles of whole-slide Hematoxylin and eosin (H&E) stained histology images. BLEEP uses contrastive learning to construct a low-dimensional joint embedding space from a reference dataset using paired image and expression profiles at micrometer resolution. With this approach, the gene expression of any query image patch can be imputed using the expression profiles from the reference dataset. We demonstrate BLEEP's effectiveness in gene expression prediction by benchmarking its performance on a human liver tissue dataset captured using the 10x Visium platform, where it achieves significant improvements over existing methods. Our results demonstrate the potential of BLEEP to provide insights into the molecular mechanisms underlying tissue architecture, with important implications in diagnosis and research of various diseases. The proposed approach can significantly reduce the time and cost associated with gene expression profiling, opening up new avenues for high-throughput analysis of histology images for both research and clinical applications."
                },
                {
                    "title": "Histopathology Whole Slide Image Analysis with Heterogeneous Graph Representation Learning",
                    "abstract": "Graph-based methods have been extensively applied to whole slide histopathology image (WSI) analysis due to the advantage of modeling the spatial relationships among different entities. However, most of the existing methods focus on modeling WSIs with homogeneous graphs (e.g., with homogeneous node type). Despite their successes, these works are incapable of mining the complex structural relations between biological entities (e.g., the diverse interaction among different cell types) in the WSI. We propose a novel heterogeneous graph-based framework to leverage the inter-relationships among different types of nuclei for WSI analysis. Specifically, we formulate the WSI as a heterogeneous graph with \u201cnucleus-type\u201d attribute to each node and a semantic similarity attribute to each edge. We then present a new heterogeneous-graph edge attribute transformer (HEAT) to take advantage of the edge and node heterogeneity during massage aggregating. Further, we design a new pseudo-label-based semantic-consistent pooling mechanism to obtain graph-level features, which can mitigate the over-parameterization issue of conventional cluster-based pooling. Additionally, observing the limitations of existing association-based localization methods, we propose a causal-driven approach attributing the contribution of each node to improve the interpretability of our framework. Extensive experiments on three public TCGA benchmark datasets demonstrate that our frame-work outperforms the state-of-the-art methods with considerable margins on various tasks. Our codes are available at https://github.com/HKU-MedAI/WSI-HGNN."
                },
                {
                    "title": "The Rise of AI Language Pathologists: Exploring Two-level Prompt Learning for Few-shot Weakly-supervised Whole Slide Image Classification",
                    "abstract": "This paper introduces the novel concept of few-shot weakly supervised learning for pathology Whole Slide Image (WSI) classification, denoted as FSWC. A solution is proposed based on prompt learning and the utilization of a large language model, GPT-4. Since a WSI is too large and needs to be divided into patches for processing, WSI classification is commonly approached as a Multiple Instance Learning (MIL) problem. In this context, each WSI is considered a bag, and the obtained patches are treated as instances. The objective of FSWC is to classify both bags and instances with only a limited number of labeled bags. Unlike conventional few-shot learning problems, FSWC poses additional challenges due to its weak bag labels within the MIL framework. Drawing inspiration from the recent achievements of vision-language models (V-L models) in downstream few-shot classification tasks, we propose a two-level prompt learning MIL framework tailored for pathology, incorporating language prior knowledge. Specifically, we leverage CLIP to extract instance features for each patch, and introduce a prompt-guided pooling strategy to aggregate these instance features into a bag feature. Subsequently, we employ a small number of labeled bags to facilitate few-shot prompt learning based on the bag features. Our approach incorporates the utilization of GPT-4 in a question-and-answer mode to obtain language prior knowledge at both the instance and bag levels, which are then integrated into the instance and bag level language prompts. Additionally, a learnable component of the language prompts is trained using the available few-shot labeled data. We conduct extensive experiments on three real WSI datasets encompassing breast cancer, lung cancer, and cervical cancer, demonstrating the notable performance of the proposed method in bag and instance classification. All codes will be available."
                },
                {
                    "title": "RWKV: Reinventing RNNs for the Transformer Era",
                    "abstract": "Transformers have revolutionized almost all natural language processing (NLP) tasks but suffer from memory and computational complexity that scales quadratically with sequence length. In contrast, recurrent neural networks (RNNs) exhibit linear scaling in memory and computational requirements but struggle to match the same performance as Transformers due to limitations in parallelization and scalability. We propose a novel model architecture, Receptance Weighted Key Value (RWKV), that combines the efficient parallelizable training of transformers with the efficient inference of RNNs. Our approach leverages a linear attention mechanism and allows us to formulate the model as either a Transformer or an RNN, thus parallelizing computations during training and maintains constant computational and memory complexity during inference. We scale our models as large as 14 billion parameters, by far the largest dense RNN ever trained, and find RWKV performs on par with similarly sized Transformers, suggesting future work can leverage this architecture to create more efficient models. This work presents a significant step towards reconciling trade-offs between computational efficiency and model performance in sequence processing tasks."
                },
                {
                    "title": "Task-Specific Fine-Tuning via Variational Information Bottleneck for Weakly-Supervised Pathology Whole Slide Image Classification",
                    "abstract": "While Multiple Instance Learning (MIL) has shown promising results in digital Pathology Whole Slide Image (WSI) analysis, such a paradigm still faces performance and generalization problems due to high computational costs and limited supervision of Gigapixel WSIs. To deal with the computation problem, previous methods utilize a frozen model pretrained from ImageNet to obtain representations, however, it may lose key information owing to the large domain gap and hinder the generalization ability without image-level training-time augmentation. Though Self-supervised Learning (SSL) proposes viable representation learning schemes, the downstream task-specific features via partial label tuning are not explored. To alleviate this problem, we propose an efficient WSI fine-tuning framework motivated by the Information Bottleneck theory. The theory enables the framework to find the minimal sufficient statistics of WSI, thus supporting us to fine-tune the backbone into a task-specific representation only depending on WSI-level weak labels. The WSI-MIL problem is further analyzed to theoretically deduce our fine-tuning method. We evaluate the method on five pathological WSI datasets on various WSI heads. The experimental results show significant improvements in both accuracy and generalization compared with previous works. Source code will be available at https://github.com/invoker-LL/WSI-finetuning."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Resurrecting Recurrent Neural Networks for Long Sequences",
                    "abstract": "Recurrent Neural Networks (RNNs) offer fast inference on long sequences but are hard to optimize and slow to train. Deep state-space models (SSMs) have recently been shown to perform remarkably well on long sequence modeling tasks, and have the added benefits of fast parallelizable training and RNN-like fast inference. However, while SSMs are superficially similar to RNNs, there are important differences that make it unclear where their performance boost over RNNs comes from. In this paper, we show that careful design of deep RNNs using standard signal propagation arguments can recover the impressive performance of deep SSMs on long-range reasoning tasks, while also matching their training speed. To achieve this, we analyze and ablate a series of changes to standard RNNs including linearizing and diagonalizing the recurrence, using better parameterizations and initializations, and ensuring proper normalization of the forward pass. Our results provide new insights on the origins of the impressive performance of deep SSMs, while also introducing an RNN block called the Linear Recurrent Unit that matches both their performance on the Long Range Arena benchmark and their computational efficiency."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Quantifying & Modeling Multimodal Interactions: An Information Decomposition Framework",
                    "abstract": "The recent explosion of interest in multimodal applications has resulted in a wide selection of datasets and methods for representing and integrating information from different modalities. Despite these empirical advances, there remain fundamental research questions: How can we quantify the interactions that are necessary to solve a multimodal task? Subsequently, what are the most suitable multimodal models to capture these interactions? To answer these questions, we propose an information-theoretic approach to quantify the degree of redundancy, uniqueness, and synergy relating input modalities with an output task. We term these three measures as the PID statistics of a multimodal distribution (or PID for short), and introduce two new estimators for these PID statistics that scale to high-dimensional distributions. To validate PID estimation, we conduct extensive experiments on both synthetic datasets where the PID is known and on large-scale multimodal benchmarks where PID estimations are compared with human annotations. Finally, we demonstrate their usefulness in (1) quantifying interactions within multimodal datasets, (2) quantifying interactions captured by multimodal models, (3) principled approaches for model selection, and (4) three real-world case studies engaging with domain experts in pathology, mood prediction, and robotic perception where our framework helps to recommend strong multimodal models for each application."
                },
                {
                    "title": "Hyena Hierarchy: Towards Larger Convolutional Language Models",
                    "abstract": "Recent advances in deep learning have relied heavily on the use of large Transformers due to their ability to learn at scale. However, the core building block of Transformers, the attention operator, exhibits quadratic cost in sequence length, limiting the amount of context accessible. Existing subquadratic methods based on low-rank and sparse approximations need to be combined with dense attention layers to match Transformers, indicating a gap in capability. In this work, we propose Hyena, a subquadratic drop-in replacement for attention constructed by interleaving implicitly parametrized long convolutions and data-controlled gating. In recall and reasoning tasks on sequences of thousands to hundreds of thousands of tokens, Hyena improves accuracy by more than 50 points over operators relying on state-spaces and other implicit and explicit methods, matching attention-based models. We set a new state-of-the-art for dense-attention-free architectures on language modeling in standard datasets (WikiText103 and The Pile), reaching Transformer quality with a 20% reduction in training compute required at sequence length 2K. Hyena operators are twice as fast as highly optimized attention at sequence length 8K, and 100x faster at sequence length 64K."
                },
                {
                    "title": "Hungry Hungry Hippos: Towards Language Modeling with State Space Models",
                    "abstract": "State space models (SSMs) have demonstrated state-of-the-art sequence modeling performance in some modalities, but underperform attention in language modeling. Moreover, despite scaling nearly linearly in sequence length instead of quadratically, SSMs are still slower than Transformers due to poor hardware utilization. In this paper, we make progress on understanding the expressivity gap between SSMs and attention in language modeling, and on reducing the hardware barrier between SSMs and attention. First, we use synthetic language modeling tasks to understand the gap between SSMs and attention. We find that existing SSMs struggle with two capabilities: recalling earlier tokens in the sequence and comparing tokens across the sequence. To understand the impact on language modeling, we propose a new SSM layer, H3, that is explicitly designed for these abilities. H3 matches attention on the synthetic languages and comes within 0.4 PPL of Transformers on OpenWebText. Furthermore, a hybrid 125M-parameter H3-attention model that retains two attention layers surprisingly outperforms Transformers on OpenWebText by 1.0 PPL. Next, to improve the efficiency of training SSMs on modern hardware, we propose FlashConv. FlashConv uses a fused block FFT algorithm to improve efficiency on sequences up to 8K, and introduces a novel state passing algorithm that exploits the recurrent properties of SSMs to scale to longer sequences. FlashConv yields 2$\\times$ speedup on the long-range arena benchmark and allows hybrid language models to generate text 2.4$\\times$ faster than Transformers. Using FlashConv, we scale hybrid H3-attention language models up to 2.7B parameters on the Pile and find promising initial results, achieving lower perplexity than Transformers and outperforming Transformers in zero- and few-shot learning on a majority of tasks in the SuperGLUE benchmark."
                },
                {
                    "title": "Benchmarking Self-Supervised Learning on Diverse Pathology Datasets",
                    "abstract": "Computational pathology can lead to saving human lives, but models are annotation hungry and pathology images are notoriously expensive to annotate. Self-supervised learning (SSL) has shown to be an effective method for utilizing unlabeled data, and its application to pathology could greatly benefit its downstream tasks. Yet, there are no principled studies that compare SSL methods and discuss how to adapt them for pathology. To address this need, we execute the largest-scale study of SSL pre-training on pathology image data, to date. Our study is conducted using 4 representative SSL methods on diverse downstream tasks. We establish that large-scale domain-aligned pre-training in pathology consistently out-performs ImageNet pre-training in standard SSL settings such as linear and fine-tuning evaluations, as well as in low-label regimes. Moreover, we propose a set of domain-specific techniques that we experimentally show leads to a performance boost. Lastly, for the first time, we apply SSL to the challenging task of nuclei instance segmentation and show large and consistent performance improvements. We release the pre-trained model weights11https://lunit-io.github.io/research/publications/pathology_ssl."
                },
                {
                    "title": "Iterative Patch Selection for High-Resolution Image Recognition",
                    "abstract": "High-resolution images are prevalent in various applications, such as autonomous driving and computer-aided diagnosis. However, training neural networks on such images is computationally challenging and easily leads to out-of-memory errors even on modern GPUs. We propose a simple method, Iterative Patch Selection (IPS), which decouples the memory usage from the input size and thus enables the processing of arbitrarily large images under tight hardware constraints. IPS achieves this by selecting only the most salient patches, which are then aggregated into a global representation for image recognition. For both patch selection and aggregation, a cross-attention based transformer is introduced, which exhibits a close connection to Multiple Instance Learning. Our method demonstrates strong performance and has wide applicability across different domains, training regimes and image sizes while using minimal accelerator memory. For example, we are able to finetune our model on whole-slide images consisting of up to 250k patches (>16 gigapixels) with only 5 GB of GPU VRAM at a batch size of 16."
                },
                {
                    "title": "Bi-directional Weakly Supervised Knowledge Distillation for Whole Slide Image Classification",
                    "abstract": "Computer-aided pathology diagnosis based on the classification of Whole Slide Image (WSI) plays an important role in clinical practice, and it is often formulated as a weakly-supervised Multiple Instance Learning (MIL) problem. Existing methods solve this problem from either a bag classification or an instance classification perspective. In this paper, we propose an end-to-end weakly supervised knowledge distillation framework (WENO) for WSI classification, which integrates a bag classifier and an instance classifier in a knowledge distillation framework to mutually improve the performance of both classifiers. Specifically, an attention-based bag classifier is used as the teacher network, which is trained with weak bag labels, and an instance classifier is used as the student network, which is trained using the normalized attention scores obtained from the teacher network as soft pseudo labels for the instances in positive bags. An instance feature extractor is shared between the teacher and the student to further enhance the knowledge exchange between them. In addition, we propose a hard positive instance mining strategy based on the output of the student network to force the teacher network to keep mining hard positive instances. WENO is a plug-and-play framework that can be easily applied to any existing attention-based bag classification methods. Extensive experiments on five datasets demonstrate the efficiency of WENO. Code is available at https://github.com/miccaiif/WENO."
                },
                {
                    "title": "Benchmarking the Robustness of Deep Neural Networks to Common Corruptions in Digital Pathology",
                    "abstract": "When designing a diagnostic model for a clinical application, it is crucial to guarantee the robustness of the model with respect to a wide range of image corruptions. Herein, an easy-to-use benchmark is established to evaluate how deep neural networks perform on corrupted pathology images. Specifically, corrupted images are generated by injecting nine types of common corruptions into validation images. Besides, two classification and one ranking metrics are designed to evaluate the prediction and confidence performance under corruption. Evaluated on two resulting benchmark datasets, we find that (1) a variety of deep neural network models suffer from a significant accuracy decrease (double the error on clean images) and the unreliable confidence estimation on corrupted images; (2) A low correlation between the validation and test errors while replacing the validation set with our benchmark can increase the correlation. Our codes are available on https://github.com/superjamessyx/robustness_benchmark."
                },
                {
                    "title": "Additive MIL: Intrinsically Interpretable Multiple Instance Learning for Pathology",
                    "abstract": "Multiple Instance Learning (MIL) has been widely applied in pathology towards solving critical problems such as automating cancer diagnosis and grading, predicting patient prognosis, and therapy response. Deploying these models in a clinical setting requires careful inspection of these black boxes during development and deployment to identify failures and maintain physician trust. In this work, we propose a simple formulation of MIL models, which enables interpretability while maintaining similar predictive performance. Our Additive MIL models enable spatial credit assignment such that the contribution of each region in the image can be exactly computed and visualized. We show that our spatial credit assignment coincides with regions used by pathologists during diagnosis and improves upon classical attention heatmaps from attention MIL models. We show that any existing MIL model can be made additive with a simple change in function composition. We also show how these models can debug model failures, identify spurious features, and highlight class-wise regions of interest, enabling their use in high-stakes environments such as clinical decision-making."
                },
                {
                    "title": "Scaling Vision Transformers to Gigapixel Images via Hierarchical Self-Supervised Learning",
                    "abstract": "Vision Transformers (ViTs) and their multi-scale and hierarchical variations have been successful at capturing image representations but their use has been generally studied for low-resolution images (e.g. 256 \u00d7 256, 384 \u00d7 384). For gigapixel whole-slide imaging (WSI) in computational pathology, WSIs can be as large as 150000 \u00d7 150000 pixels at 20 \u00d7 magnification and exhibit a hierarchical structure of visual tokens across varying resolutions: from 16 \u00d7 16 images capturing individual cells, to 4096 \u00d7 4096 images characterizing interactions within the tissue microenvironment. We introduce a new ViT architecture called the Hierarchical Image Pyramid Transformer (HIPT), which leverages the natural hierarchical structure inherent in WSIs using two levels of self-supervised learning to learn high-resolution image representations. HIPT is pretrained across 33 cancer types using 10,678 gigapixel WSIs, 408,218 4096 \u00d7 4096 images, and 104M 256 \u00d7 256 images. We benchmark HIPT representations on 9 slide-level tasks, and demonstrate that: 1) HIPT with hierarchical pretraining outperforms current state-of-the-art methods for cancer subtyping and survival prediction, 2) self-supervised ViTs are able to model important inductive biases about the hierarchical structure of phenotypes in the tumor microenvironment."
                },
                {
                    "title": "Node-aligned Graph Convolutional Network for Whole-slide Image Representation and Classification",
                    "abstract": "The large-scale whole-slide images (WSIs) facilitate the learning-based computational pathology methods. However, the gigapixel size of WSIs makes it hard to train a conventional model directly. Current approaches typically adopt multiple-instance learning (MIL) to tackle this problem. Among them, MIL combined with graph convolutional network (GCN) is a significant branch, where the sampled patches are regarded as the graph nodes to further discover their correlations. However, it is difficult to build correspondence across patches from different WSIs. Therefore, most methods have to perform non-ordered node pooling to generate the bag-level representation. Direct non-ordered pooling will lose much structural and contextual information, such as patch distribution and heterogeneous patterns, which is critical for WSI representation. In this paper, we propose a hierarchical global-to-local clustering strategy to build a Node-Aligned GCN (NAGCN) to represent WSI with rich local structural information as well as global distribution. We first deploy a global clustering operation based on the instance features in the dataset to build the correspondence across different WSIs. Then, we perform a local clustering-based sampling strategy to select typical instances belonging to each cluster within the WSI. Finally, we employ the graph convolution to obtain the representation. Since our graph construction strategy ensures the alignment among different WSIs, WSI-level representation can be easily generated and used for the subsequent classification. The experiment results on two cancer subtype classification datasets demonstrate our method achieves better performance compared with the state-of-the-art methods."
                },
                {
                    "title": "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness",
                    "abstract": "Transformers are slow and memory-hungry on long sequences, since the time and memory complexity of self-attention are quadratic in sequence length. Approximate attention methods have attempted to address this problem by trading off model quality to reduce the compute complexity, but often do not achieve wall-clock speedup. We argue that a missing principle is making attention algorithms IO-aware -- accounting for reads and writes between levels of GPU memory. We propose FlashAttention, an IO-aware exact attention algorithm that uses tiling to reduce the number of memory reads/writes between GPU high bandwidth memory (HBM) and GPU on-chip SRAM. We analyze the IO complexity of FlashAttention, showing that it requires fewer HBM accesses than standard attention, and is optimal for a range of SRAM sizes. We also extend FlashAttention to block-sparse attention, yielding an approximate attention algorithm that is faster than any existing approximate attention method. FlashAttention trains Transformers faster than existing baselines: 15% end-to-end wall-clock speedup on BERT-large (seq. length 512) compared to the MLPerf 1.1 training speed record, 3$\\times$ speedup on GPT-2 (seq. length 1K), and 2.4$\\times$ speedup on long-range arena (seq. length 1K-4K). FlashAttention and block-sparse FlashAttention enable longer context in Transformers, yielding higher quality models (0.7 better perplexity on GPT-2 and 6.4 points of lift on long-document classification) and entirely new capabilities: the first Transformers to achieve better-than-chance performance on the Path-X challenge (seq. length 16K, 61.4% accuracy) and Path-256 (seq. length 64K, 63.1% accuracy)."
                },
                {
                    "title": "KERPLE: Kernelized Relative Positional Embedding for Length Extrapolation",
                    "abstract": "Relative positional embeddings (RPE) have received considerable attention since RPEs effectively model the relative distance among tokens and enable length extrapolation. We propose KERPLE, a framework that generalizes relative position embedding for extrapolation by kernelizing positional differences. We achieve this goal using conditionally positive definite (CPD) kernels, a class of functions known for generalizing distance metrics. To maintain the inner product interpretation of self-attention, we show that a CPD kernel can be transformed into a PD kernel by adding a constant offset. This offset is implicitly absorbed in the Softmax normalization during self-attention. The diversity of CPD kernels allows us to derive various RPEs that enable length extrapolation in a principled way. Experiments demonstrate that the logarithmic variant achieves excellent extrapolation performance on three large language modeling datasets. Our implementation and pretrained checkpoints are released at~\\url{https://github.com/chijames/KERPLE.git}."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "DTFD-MIL: Double-Tier Feature Distillation Multiple Instance Learning for Histopathology Whole Slide Image Classification",
                    "abstract": "Multiple instance learning (MIL) has been increasingly used in the classification of histopathology whole slide images (WSIs). However, MIL approaches for this specific classification problem still face unique challenges, particularly those related to small sample cohorts. In these, there are limited number of WSI slides (bags), while the resolution of a single WSI is huge, which leads to a large number of patches (instances) cropped from this slide. To address this issue, we propose to virtually enlarge the number of bags by introducing the concept of pseudo-bags, on which a double-tier MIL framework is built to effectively use the intrinsic features. Besides, we also contribute to deriving the instance probability under the framework of attentionbased MIL, and utilize the derivation to help construct and analyze the proposed framework. The proposed method outperforms other latest methods on the CAMELYON-16 by substantially large margins, and is also better in performance on the TCGA lung cancer dataset. The proposed framework is ready to be extended for wider MIL applications. The code is available at: https://github. com/hrzhang1123/DTFD-MIL."
                },
                {
                    "title": "Self-attention Does Not Need $O(n^2)$ Memory",
                    "abstract": "We present a very simple algorithm for attention that requires $O(1)$ memory with respect to sequence length and an extension to self-attention that requires $O(\\log n)$ memory. This is in contrast with the frequently stated belief that self-attention requires $O(n^2)$ memory. While the time complexity is still $O(n^2)$, device memory rather than compute capability is often the limiting factor on modern accelerators. Thus, reducing the memory requirements of attention allows processing of longer sequences than might otherwise be feasible. We provide a practical implementation for accelerators that requires $O(\\sqrt{n})$ memory, is numerically stable, and is within a few percent of the runtime of the standard implementation of attention. We also demonstrate how to differentiate the function while remaining memory-efficient. For sequence length 16384, the memory overhead of self-attention is reduced by 59X for inference and by 32X for differentiation."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "BRACS: A Dataset for BReAst Carcinoma Subtyping in H&E Histology Images",
                    "abstract": "Abstract Breast cancer is the most commonly diagnosed cancer and registers the highest number of deaths for women. Advances in diagnostic activities combined with large-scale screening policies have significantly lowered the mortality rates for breast cancer patients. However, the manual inspection of tissue slides by pathologists is cumbersome, time-consuming and is subject to significant inter- and intra-observer variability. Recently, the advent of whole-slide scanning systems has empowered the rapid digitization of pathology slides and enabled the development of Artificial Intelligence (AI)-assisted digital workflows. However, AI techniques, especially Deep Learning, require a large amount of high-quality annotated data to learn from. Constructing such task-specific datasets poses several challenges, such as data-acquisition level constraints, time-consuming and expensive annotations and anonymization of patient information. In this paper, we introduce the BReAst Carcinoma Subtyping (BRACS) dataset, a large cohort of annotated Hematoxylin and Eosin (H&E)-stained images to advance AI development in the automatic characterization of breast lesions. BRACS contains 547 Whole-Slide Images (WSIs) and 4539 Regions Of Interest (ROIs) extracted from the WSIs. Each WSI and respective ROIs are annotated by the consensus of three board-certified pathologists into different lesion categories. Specifically, BRACS includes three lesion types, i.e., benign, malignant and atypical, which are further subtyped into seven categories. It is, to the best of our knowledge, the largest annotated dataset for breast cancer subtyping both at WSI and ROI levels. Furthermore, by including the understudied atypical lesions, BRACS offers a unique opportunity for leveraging AI to better understand their characteristics. We encourage AI practitioners to develop and evaluate novel algorithms on the BRACS dataset to further breast cancer diagnosis and patient care. Database URL: https://www.bracs.icar.cnr.it/"
                },
                {
                    "title": "Efficiently Modeling Long Sequences with Structured State Spaces",
                    "abstract": "A central goal of sequence modeling is designing a single principled model that can address sequence data across a range of modalities and tasks, particularly on long-range dependencies. Although conventional models including RNNs, CNNs, and Transformers have specialized variants for capturing long dependencies, they still struggle to scale to very long sequences of $10000$ or more steps. A promising recent approach proposed modeling sequences by simulating the fundamental state space model (SSM) \\( x'(t) = Ax(t) + Bu(t), y(t) = Cx(t) + Du(t) \\), and showed that for appropriate choices of the state matrix \\( A \\), this system could handle long-range dependencies mathematically and empirically. However, this method has prohibitive computation and memory requirements, rendering it infeasible as a general sequence modeling solution. We propose the Structured State Space sequence model (S4) based on a new parameterization for the SSM, and show that it can be computed much more efficiently than prior approaches while preserving their theoretical strengths. Our technique involves conditioning \\( A \\) with a low-rank correction, allowing it to be diagonalized stably and reducing the SSM to the well-studied computation of a Cauchy kernel. S4 achieves strong empirical results across a diverse range of established benchmarks, including (i) 91\\% accuracy on sequential CIFAR-10 with no data augmentation or auxiliary losses, on par with a larger 2-D ResNet, (ii) substantially closing the gap to Transformers on image and language modeling tasks, while performing generation $60\\times$ faster (iii) SoTA on every task from the Long Range Arena benchmark, including solving the challenging Path-X task of length 16k that all prior work fails on, while being as efficient as all competitors."
                },
                {
                    "title": "Multimodal Co-Attention Transformer for Survival Prediction in Gigapixel Whole Slide Images",
                    "abstract": "Survival outcome prediction is a challenging weakly-supervised and ordinal regression task in computational pathology that involves modeling complex interactions within the tumor microenvironment in gigapixel whole slide images (WSIs). Despite recent progress in formulating WSIs as bags for multiple instance learning (MIL), representation learning of entire WSIs remains an open and challenging problem, especially in overcoming: 1) the computational complexity of feature aggregation in large bags, and 2) the data heterogeneity gap in incorporating biological priors such as genomic measurements. In this work, we present a Multimodal Co-Attention Transformer (MCAT) framework that learns an interpretable, dense co-attention mapping between WSIs and genomic features formulated in an embedding space. Inspired by approaches in Visual Question Answering (VQA) that can attribute how word embed-dings attend to salient objects in an image when answering a question, MCAT learns how histology patches attend to genes when predicting patient survival. In addition to visualizing multimodal interactions, our co-attention trans-formation also reduces the space complexity of WSI bags, which enables the adaptation of Transformer layers as a general encoder backbone in MIL. We apply our proposed method on five different cancer datasets (4,730 WSIs, 67 million patches). Our experimental results demonstrate that the proposed method consistently achieves superior performance compared to the state-of-the-art methods."
                },
                {
                    "title": "Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation",
                    "abstract": "Since the introduction of the transformer model by Vaswani et al. (2017), a fundamental question has yet to be answered: how does a model achieve extrapolation at inference time for sequences that are longer than it saw during training? We first show that extrapolation can be enabled by simply changing the position representation method, though we find that current methods do not allow for efficient extrapolation. We therefore introduce a simpler and more efficient position method, Attention with Linear Biases (ALiBi). ALiBi does not add positional embeddings to word embeddings; instead, it biases query-key attention scores with a penalty that is proportional to their distance. We show that this method trains a 1.3 billion parameter model on input sequences of length 1024 that extrapolates to input sequences of length 2048, achieving the same perplexity as a sinusoidal position embedding model trained on inputs of length 2048 but training 11% faster and using 11% less memory. ALiBi's inductive bias towards recency also leads it to outperform multiple strong position methods on the WikiText-103 benchmark."
                },
                {
                    "title": "Rethinking and Improving Relative Position Encoding for Vision Transformer",
                    "abstract": "Relative position encoding (RPE) is important for transformer to capture sequence ordering of input tokens. General efficacy has been proven in natural language processing. However, in computer vision, its efficacy is not well studied and even remains controversial, e.g., whether relative position encoding can work equally well as absolute position? In order to clarify this, we first review existing relative position encoding methods and analyze their pros and cons when applied in vision transformers. We then propose new relative position encoding methods dedicated to 2D images, called image RPE (iRPE). Our methods consider directional relative distance modeling as well as the interactions between queries and relative position embeddings in self-attention mechanism. The proposed iRPE methods are simple and lightweight. They can be easily plugged into transformer blocks. Experiments demonstrate that solely due to the proposed encoding methods, DeiT [21] and DETR [1] obtain up to 1.5% (top-1 Acc) and 1.3% (mAP) stable improvements over their original versions on ImageNet and COCO respectively, without tuning any extra hyperparameters such as learning rate and weight decay. Our ablation and analysis also yield interesting findings, some of which run counter to previous understanding. Code and models are open-sourced at https://github.com/microsoft/Cream/tree/main/iRPE."
                },
                {
                    "title": "TransMIL: Transformer based Correlated Multiple Instance Learning for Whole Slide Image Classication",
                    "abstract": "Multiple instance learning (MIL) is a powerful tool to solve the weakly supervised classification in whole slide image (WSI) based pathology diagnosis. However, the current MIL methods are usually based on independent and identical distribution hypothesis, thus neglect the correlation among different instances. To address this problem, we proposed a new framework, called correlated MIL, and provided a proof for convergence. Based on this framework, we devised a Transformer based MIL (TransMIL), which explored both morphological and spatial information. The proposed TransMIL can effectively deal with unbalanced/balanced and binary/multiple classification with great visualization and interpretability. We conducted various experiments for three different computational pathology problems and achieved better performance and faster convergence compared with state-of-the-art methods. The test AUC for the binary tumor classification can be up to 93.09% over CAMELYON16 dataset. And the AUC over the cancer subtypes classification can be up to 96.03% and 98.82% over TCGA-NSCLC dataset and TCGA-RCC dataset, respectively. Implementation is available at: https://github.com/szc19990412/TransMIL."
                },
                {
                    "title": "An Attention Free Transformer",
                    "abstract": "We introduce Attention Free Transformer (AFT), an efficient variant of Transformers that eliminates the need for dot product self attention. In an AFT layer, the key and value are first combined with a set of learned position biases, the result of which is multiplied with the query in an element-wise fashion. This new operation has a memory complexity linear w.r.t. both the context size and the dimension of features, making it compatible to both large input and model sizes. We also introduce AFT-local and AFT-conv, two model variants that take advantage of the idea of locality and spatial weight sharing while maintaining global connectivity. We conduct extensive experiments on two autoregressive modeling tasks (CIFAR10 and Enwik8) as well as an image recognition task (ImageNet-1K classification). We show that AFT demonstrates competitive performance on all the benchmarks, while providing excellent efficiency at the same time."
                },
                {
                    "title": "SPIE-AAPM-NCI BreastPathQ Challenge: an image analysis challenge for quantitative tumor cellularity assessment in breast cancer histology images following neoadjuvant treatment",
                    "abstract": "Abstract. Purpose: The Breast Pathology Quantitative Biomarkers (BreastPathQ) Challenge was a Grand Challenge organized jointly by the international society for optics and photonics (SPIE), the American Association of Physicists in Medicine (AAPM), the U.S. National Cancer Institute (NCI), and the U.S. Food and Drug Administration (FDA). The task of the BreastPathQ Challenge was computerized estimation of tumor cellularity (TC) in breast cancer histology images following neoadjuvant treatment. Approach: A total of 39 teams developed, validated, and tested their TC estimation algorithms during the challenge. The training, validation, and testing sets consisted of 2394, 185, and 1119 image patches originating from 63, 6, and 27 scanned pathology slides from 33, 4, and 18 patients, respectively. The summary performance metric used for comparing and ranking algorithms was the average prediction probability concordance (PK) using scores from two pathologists as the TC reference standard. Results: Test PK performance ranged from 0.497 to 0.941 across the 100 submitted algorithms. The submitted algorithms generally performed well in estimating TC, with high-performing algorithms obtaining comparable results to the average interrater PK of 0.927 from the two pathologists providing the reference TC scores. Conclusions: The SPIE-AAPM-NCI BreastPathQ Challenge was a success, indicating that artificial intelligence/machine learning algorithms may be able to approach human performance for cellularity assessment and may have some utility in clinical practice for improving efficiency and reducing reader variability. The BreastPathQ Challenge can be accessed on the Grand Challenge website."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "CvT: Introducing Convolutions to Vision Transformers",
                    "abstract": "We present in this paper a new architecture, named Convolutional vision Transformer (CvT), that improves Vision Transformer (ViT) in performance and efficiency by introducing convolutions into ViT to yield the best of both de-signs. This is accomplished through two primary modifications: a hierarchy of Transformers containing a new convolutional token embedding, and a convolutional Transformer block leveraging a convolutional projection. These changes introduce desirable properties of convolutional neural networks (CNNs) to the ViT architecture (i.e. shift, scale, and distortion invariance) while maintaining the merits of Transformers (i.e. dynamic attention, global context, and better generalization). We validate CvT by conducting extensive experiments, showing that this approach achieves state-of-the-art performance over other Vision Transformers and ResNets on ImageNet-1k, with fewer parameters and lower FLOPs. In addition, performance gains are maintained when pretrained on larger datasets (e.g. ImageNet-22k) and fine-tuned to downstream tasks. Pretrained on ImageNet-22k, our CvT-W24 obtains a top-1 accuracy of 87.7% on the ImageNet-1k val set. Finally, our results show that the positional encoding, a crucial component in existing Vision Transformers, can be safely re-moved in our model, simplifying the design for higher resolution vision tasks. Code will be released at https://github.com/microsoft/CvT."
                },
                {
                    "title": "Perceiver: General Perception with Iterative Attention",
                    "abstract": "Biological systems perceive the world by simultaneously processing high-dimensional inputs from modalities as diverse as vision, audition, touch, proprioception, etc. The perception models used in deep learning on the other hand are designed for individual modalities, often relying on domain-specific assumptions such as the local grid structures exploited by virtually all existing vision models. These priors introduce helpful inductive biases, but also lock models to individual modalities. In this paper we introduce the Perceiver - a model that builds upon Transformers and hence makes few architectural assumptions about the relationship between its inputs, but that also scales to hundreds of thousands of inputs, like ConvNets. The model leverages an asymmetric attention mechanism to iteratively distill inputs into a tight latent bottleneck, allowing it to scale to handle very large inputs. We show that this architecture is competitive with or outperforms strong, specialized models on classification tasks across various modalities: images, point clouds, audio, video, and video+audio. The Perceiver obtains performance comparable to ResNet-50 and ViT on ImageNet without 2D convolutions by directly attending to 50,000 pixels. It is also competitive in all modalities in AudioSet."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Nystr\u00f6mformer: A Nystr\u00f6m-Based Algorithm for Approximating Self-Attention",
                    "abstract": "Transformers have emerged as a powerful tool for a broad range of natural language processing tasks. A key component that drives the impressive performance of Transformers is the self-attention mechanism that encodes the influence or dependence of other tokens on each specific token. While beneficial, the quadratic complexity of self-attention on the input sequence length has limited its application to longer sequences - a topic being actively studied in the community. To address this limitation, we propose Nystr\u00f6mformer - a model that exhibits favorable scalability as a function of sequence length. Our idea is based on adapting the Nystr\u00f6m method to approximate standard self-attention with O(n) complexity. The scalability of Nystr\u00f6mformer enables application to longer sequences with thousands of tokens. We perform evaluations on multiple downstream tasks on the GLUE benchmark and IMDB reviews with standard sequence length, and find that our Nystr\u00f6mformer performs comparably, or in a few cases, even slightly better, than standard self-attention. On longer sequence tasks in the Long Range Arena (LRA) benchmark, Nystr\u00f6mformer performs favorably relative to other efficient self-attention methods. Our code is available at https://github.com/mlpen/Nystromformer."
                },
                {
                    "title": "A Generalization of Transformer Networks to Graphs",
                    "abstract": "We propose a generalization of transformer neural network architecture for arbitrary graphs. The original transformer was designed for Natural Language Processing (NLP), which operates on fully connected graphs representing all connections between the words in a sequence. Such architecture does not leverage the graph connectivity inductive bias, and can perform poorly when the graph topology is important and has not been encoded into the node features. We introduce a graph transformer with four new properties compared to the standard model. First, the attention mechanism is a function of the neighborhood connectivity for each node in the graph. Second, the positional encoding is represented by the Laplacian eigenvectors, which naturally generalize the sinusoidal positional encodings often used in NLP. Third, the layer normalization is replaced by a batch normalization layer, which provides faster training and better generalization performance. Finally, the architecture is extended to edge feature representation, which can be critical to tasks s.a. chemistry (bond type) or link prediction (entity relationship in knowledge graphs). Numerical experiments on a graph benchmark demonstrate the performance of the proposed graph transformer architecture. This work closes the gap between the original transformer, which was designed for the limited case of line graphs, and graph neural networks, that can work with arbitrary graphs. As our architecture is simple and generic, we believe it can be used as a black box for future applications that wish to consider transformer and graphs."
                },
                {
                    "title": "Dual-stream Multiple Instance Learning Network for Whole Slide Image Classification with Self-supervised Contrastive Learning",
                    "abstract": "We address the challenging problem of whole slide image (WSI) classification. WSIs have very high resolutions and usually lack localized annotations. WSI classification can be cast as a multiple instance learning (MIL) problem when only slide-level labels are available. We propose a MIL-based method for WSI classification and tumor detection that does not require localized annotations. Our method has three major components. First, we introduce a novel MIL aggregator that models the relations of the instances in a dual-stream architecture with trainable distance measurement. Second, since WSIs can produce large or unbalanced bags that hinder the training of MIL models, we propose to use self-supervised contrastive learning to extract good representations for MIL and alleviate the issue of prohibitive memory cost for large bags. Third, we adopt a pyramidal fusion mechanism for multiscale WSI features, and further improve the accuracy of classification and localization. Our model is evaluated on two representative WSI datasets. The classification accuracy of our model compares favorably to fully-supervised methods, with less than 2% accuracy gap across datasets. Our results also outperform all previous MIL-based methods. Additional benchmark results on standard MIL datasets further demonstrate the superior performance of our MIL aggregator on general MIL problems."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Cluster-Former: Clustering-based Sparse Transformer for Long-Range Dependency Encoding",
                    "abstract": "Transformer has become ubiquitous in the deep learning field. One of the key ingredients that destined its success is the self-attention mechanism, which allows fully-connected contextual encoding over input tokens. However, despite its effectiveness in modeling short sequences, self-attention suffers when handling inputs with extreme long-range dependencies, as its complexity grows quadratically with respect to the sequence length. Therefore, long sequences are often encoded by Transformer in chunks using a sliding window. In this paper, we propose Cluster-Former, a novel clustering-based sparse Transformer to perform attention across chunked sequences. Our proposed method allows information integration beyond local windows, which is especially beneficial for question answering (QA) and language modeling tasks that rely on long-range dependencies. Experiments show that Cluster-Former achieves state-of-the-art performance on several major QA benchmarks."
                },
                {
                    "title": "Fast Transformers with Clustered Attention",
                    "abstract": "Transformers have been proven a successful model for a variety of tasks in sequence modeling. However, computing the attention matrix, which is their key component, has quadratic complexity with respect to the sequence length, thus making them prohibitively expensive for large sequences. To address this, we propose clustered attention, which instead of computing the attention for every query, groups queries into clusters and computes attention just for the centroids. To further improve this approximation, we use the computed clusters to identify the keys with the highest attention per query and compute the exact key/query dot products. This results in a model with linear complexity with respect to the sequence length for a fixed number of clusters. We evaluate our approach on two automatic speech recognition datasets and show that our model consistently outperforms vanilla transformers for a given computational budget. Finally, we demonstrate that our model can approximate arbitrarily complex attention distributions with a minimal number of clusters by approximating a pretrained BERT model on GLUE and SQuAD benchmarks with only 25 clusters and no loss in performance."
                },
                {
                    "title": "Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention",
                    "abstract": "Transformers achieve remarkable performance in several tasks but due to their quadratic complexity, with respect to the input's length, they are prohibitively slow for very long sequences. To address this limitation, we express the self-attention as a linear dot-product of kernel feature maps and make use of the associativity property of matrix products to reduce the complexity from $\\mathcal{O}\\left(N^2\\right)$ to $\\mathcal{O}\\left(N\\right)$, where $N$ is the sequence length. We show that this formulation permits an iterative implementation that dramatically accelerates autoregressive transformers and reveals their relationship to recurrent neural networks. Our linear transformers achieve similar performance to vanilla transformers and they are up to 4000x faster on autoregressive prediction of very long sequences."
                },
                {
                    "title": "Linformer: Self-Attention with Linear Complexity",
                    "abstract": "Large transformer models have shown extraordinary success in achieving state-of-the-art results in many natural language processing applications. However, training and deploying these models can be prohibitively costly for long sequences, as the standard self-attention mechanism of the Transformer uses $O(n^2)$ time and space with respect to sequence length. In this paper, we demonstrate that the self-attention mechanism can be approximated by a low-rank matrix. We further exploit this finding to propose a new self-attention mechanism, which reduces the overall self-attention complexity from $O(n^2)$ to $O(n)$ in both time and space. The resulting linear transformer, the \\textit{Linformer}, performs on par with standard Transformer models, while being much more memory- and time-efficient."
                },
                {
                    "title": "Masked Language Modeling for Proteins via Linearly Scalable Long-Context Transformers",
                    "abstract": "Transformer models have achieved state-of-the-art results across a diverse range of domains. However, concern over the cost of training the attention mechanism to learn complex dependencies between distant inputs continues to grow. In response, solutions that exploit the structure and sparsity of the learned attention matrix have blossomed. However, real-world applications that involve long sequences, such as biological sequence analysis, may fall short of meeting these assumptions, precluding exploration of these models. To address this challenge, we present a new Transformer architecture, Performer, based on Fast Attention Via Orthogonal Random features (FAVOR). Our mechanism scales linearly rather than quadratically in the number of tokens in the sequence, is characterized by sub-quadratic space complexity and does not incorporate any sparsity pattern priors. Furthermore, it provides strong theoretical guarantees: unbiased estimation of the attention matrix and uniform convergence. It is also backwards-compatible with pre-trained regular Transformers. We demonstrate its effectiveness on the challenging task of protein sequence modeling and provide detailed theoretical analysis."
                },
                {
                    "title": "Longformer: The Long-Document Transformer",
                    "abstract": "Transformer-based models are unable to process long sequences due to their self-attention operation, which scales quadratically with the sequence length. To address this limitation, we introduce the Longformer with an attention mechanism that scales linearly with sequence length, making it easy to process documents of thousands of tokens or longer. Longformer's attention mechanism is a drop-in replacement for the standard self-attention and combines a local windowed attention with a task motivated global attention. Following prior work on long-sequence transformers, we evaluate Longformer on character-level language modeling and achieve state-of-the-art results on text8 and enwik8. In contrast to most prior work, we also pretrain Longformer and finetune it on a variety of downstream tasks. Our pretrained Longformer consistently outperforms RoBERTa on long document tasks and sets new state-of-the-art results on WikiHop and TriviaQA. We finally introduce the Longformer-Encoder-Decoder (LED), a Longformer variant for supporting long document generative sequence-to-sequence tasks, and demonstrate its effectiveness on the arXiv summarization dataset."
                },
                {
                    "title": "Efficient Content-Based Sparse Attention with Routing Transformers",
                    "abstract": "Self-attention has recently been adopted for a wide range of sequence modeling problems. Despite its effectiveness, self-attention suffers from quadratic computation and memory requirements with respect to sequence length. Successful approaches to reduce this complexity focused on attending to local sliding windows or a small set of locations independent of content. Our work proposes to learn dynamic sparse attention patterns that avoid allocating computation and memory to attend to content unrelated to the query of interest. This work builds upon two lines of research: It combines the modeling flexibility of prior work on content-based sparse attention with the efficiency gains from approaches based on local, temporal sparse attention. Our model, the Routing Transformer, endows self-attention with a sparse routing module based on online k-means while reducing the overall complexity of attention to O(n1.5d) from O(n2d) for sequence length n and hidden dimension d. We show that our model outperforms comparable sparse attention models on language modeling on Wikitext-103 (15.8 vs 18.3 perplexity), as well as on image generation on ImageNet-64 (3.43 vs 3.44 bits/dim) while using fewer self-attention layers. Additionally, we set a new state-of-the-art on the newly released PG-19 data-set, obtaining a test perplexity of 33.2 with a 22 layer Routing Transformer model trained on sequences of length 8192. We open-source the code for Routing Transformer in Tensorflow.1"
                },
                {
                    "title": "Bias in Cross-Entropy-Based Training of Deep Survival Networks",
                    "abstract": "Over the last years, utilizing deep learning for the analysis of survival data has become attractive to many researchers. This has led to the advent of numerous network architectures for the prediction of possibly censored time-to-event variables. Unlike networks for cross-sectional data (used e.g., in classification), deep survival networks require the specification of a suitably defined loss function that incorporates typical characteristics of survival data such as censoring and time-dependent features. Here, we provide an in-depth analysis of the cross-entropy loss function, which is a popular loss function for training deep survival networks. For each time point <inline-formula><tex-math notation=\"LaTeX\">$t$</tex-math><alternatives><mml:math><mml:mi>t</mml:mi></mml:math><inline-graphic xlink:href=\"schmid-ieq1-2979450.gif\"/></alternatives></inline-formula>, the cross-entropy loss is defined in terms of a binary outcome with levels \u201cevent at or before <inline-formula><tex-math notation=\"LaTeX\">$t$</tex-math><alternatives><mml:math><mml:mi>t</mml:mi></mml:math><inline-graphic xlink:href=\"schmid-ieq2-2979450.gif\"/></alternatives></inline-formula>\u201d and \u201cevent after <inline-formula><tex-math notation=\"LaTeX\">$t$</tex-math><alternatives><mml:math><mml:mi>t</mml:mi></mml:math><inline-graphic xlink:href=\"schmid-ieq3-2979450.gif\"/></alternatives></inline-formula>\u201d. Using both theoretical and empirical approaches, we show that this definition may result in a high prediction error and a heavy bias in the predicted survival probabilities. To overcome this problem, we analyze an alternative loss function that is derived from the negative log-likelihood function of a discrete time-to-event model. We show that replacing the cross-entropy loss by the negative log-likelihood loss results in much better calibrated prediction rules and also in an improved discriminatory power, as measured by the concordance index."
                },
                {
                    "title": "Low-Rank Bottleneck in Multi-head Attention Models",
                    "abstract": "Attention based Transformer architecture has enabled significant advances in the field of natural language processing. In addition to new pre-training techniques, recent improvements crucially rely on working with a relatively larger embedding dimension for tokens. Unfortunately, this leads to models that are prohibitively large to be employed in the downstream tasks. In this paper we identify one of the important factors contributing to the large embedding size requirement. In particular, our analysis highlights that the scaling between the number of heads and the size of each head in the current architecture gives rise to a low-rank bottleneck in attention heads, causing this limitation. We further validate this in our experiments. As a solution we propose to set the head size of an attention unit to input sequence length, and independent of the number of heads, resulting in multi-head attention layers with provably more expressive power. We empirically show that this allows us to train models with a relatively smaller embedding dimension and with better performance scaling."
                },
                {
                    "title": "Reformer: The Efficient Transformer",
                    "abstract": "Large Transformer models routinely achieve state-of-the-art results on a number of tasks but training these models can be prohibitively costly, especially on long sequences. We introduce two techniques to improve the efficiency of Transformers. For one, we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O($L^2$) to O($L\\log L$), where $L$ is the length of the sequence. Furthermore, we use reversible residual layers instead of the standard residuals, which allows storing activations only once in the training process instead of $N$ times, where $N$ is the number of layers. The resulting model, the Reformer, performs on par with Transformer models while being much more memory-efficient and much faster on long sequences."
                },
                {
                    "title": "Blockwise Self-Attention for Long Document Understanding",
                    "abstract": "We present BlockBERT, a lightweight and efficient BERT model for better modeling long-distance dependencies. Our model extends BERT by introducing sparse block structures into the attention matrix to reduce both memory consumption and training/inference time, which also enables attention heads to capture either short- or long-range contextual information. We conduct experiments on language model pre-training and several benchmark question answering datasets with various paragraph lengths. BlockBERT uses 18.7-36.1% less memory and 12.0-25.1% less time to learn the model. During testing, BlockBERT saves 27.8% inference time, while having comparable and sometimes better prediction accuracy, compared to an advanced BERT-based model, RoBERTa."
                },
                {
                    "title": "Multimodal Transformer for Unaligned Multimodal Language Sequences",
                    "abstract": "Human language is often multimodal, which comprehends a mixture of natural language, facial gestures, and acoustic behaviors. However, two major challenges in modeling such multimodal human language time-series data exist: 1) inherent data non-alignment due to variable sampling rates for the sequences from each modality; and 2) long-range dependencies between elements across modalities. In this paper, we introduce the Multimodal Transformer (MulT) to generically address the above issues in an end-to-end manner without explicitly aligning the data. At the heart of our model is the directional pairwise crossmodal attention, which attends to interactions between multimodal sequences across distinct time steps and latently adapt streams from one modality to another. Comprehensive experiments on both aligned and non-aligned multimodal time-series show that our model outperforms state-of-the-art methods by a large margin. In addition, empirical analysis suggests that correlated crossmodal signals are able to be captured by the proposed crossmodal attention mechanism in MulT."
                },
                {
                    "title": "MnnFast: A Fast and Scalable System Architecture for Memory-Augmented Neural Networks",
                    "abstract": "Memory-augmented neural networks are getting more attention from many researchers as they can make an inference with the previous history stored in memory. Especially, among these memory-augmented neural networks, memory networks are known for their huge reasoning power and capability to learn from a large number of inputs rather than other networks. As the size of input datasets rapidly grows, the necessity of large-scale memory networks continuously arises. Such large-scale memory networks provide excellent reasoning power; however, the current computer infrastructure cannot achieve scalable performance due to its limited system architecture. In this paper, we propose MnnFast, a novel system architecture for large-scale memory networks to achieve fast and scalable reasoning performance. We identify the performance problems of the current architecture by conducting extensive performance bottleneck analysis. Our in-depth analysis indicates that the current architecture suffers from three major performance problems: high memory bandwidth consumption, heavy computation, and cache contention. To overcome these performance problems, we propose three novel optimizations. First, to reduce the memory bandwidth consumption, we propose a new column-based algorithm with streaming which minimizes the size of data spills and hides most of the off-chip memory accessing overhead. Second, to decrease the high computational overhead, we propose a zero-skipping optimization to bypass a large amount of output computation. Lastly, to eliminate the cache contention, we propose an embedding cache dedicated to efficiently cache the embedding matrix. Our evaluations show that MnnFast is significantly effective in various types of hardware: CPU, GPU, and FPGA. MnnFast improves the overall throughput by up to 5.38\u00d7, 4.34\u00d7, and 2.01\u00d7 on CPU, GPU, and FPGA respectively. Also, compared to CPU-based MnnFast, our FPGA-based MnnFast achieves 6.54\u00d7 higher energy efficiency."
                },
                {
                    "title": "Generating Long Sequences with Sparse Transformers",
                    "abstract": "Transformers are powerful sequence models, but require time and memory that grows quadratically with the sequence length. In this paper we introduce sparse factorizations of the attention matrix which reduce this to $O(n \\sqrt{n})$. We also introduce a) a variation on architecture and initialization to train deeper networks, b) the recomputation of attention matrices to save memory, and c) fast attention kernels for training. We call networks with these changes Sparse Transformers, and show they can model sequences tens of thousands of timesteps long using hundreds of layers. We use the same architecture to model images, audio, and text from raw bytes, setting a new state of the art for density modeling of Enwik8, CIFAR-10, and ImageNet-64. We generate unconditional samples that demonstrate global coherence and great diversity, and show it is possible in principle to use self-attention to model sequences of length one million or more."
                },
                {
                    "title": "Transformer-XL: Attentive Language Models beyond a Fixed-Length Context",
                    "abstract": "Transformers have a potential of learning longer-term dependency, but are limited by a fixed-length context in the setting of language modeling. We propose a novel neural architecture Transformer-XL that enables learning dependency beyond a fixed length without disrupting temporal coherence. It consists of a segment-level recurrence mechanism and a novel positional encoding scheme. Our method not only enables capturing longer-term dependency, but also resolves the context fragmentation problem. As a result, Transformer-XL learns dependency that is 80% longer than RNNs and 450% longer than vanilla Transformers, achieves better performance on both short and long sequences, and is up to 1,800+ times faster than vanilla Transformers during evaluation. Notably, we improve the state-of-the-art results of bpc/perplexity to 0.99 on enwiki8, 1.08 on text8, 18.3 on WikiText-103, 21.8 on One Billion Word, and 54.5 on Penn Treebank (without finetuning). When trained only on WikiText-103, Transformer-XL manages to generate reasonably coherent, novel text articles with thousands of tokens. Our code, pretrained models, and hyperparameters are available in both Tensorflow and PyTorch."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "The Cancer Genome Atlas (TCGA): an immeasurable source of knowledge",
                    "abstract": "The Cancer Genome Atlas (TCGA) is a public funded project that aims to catalogue and discover major cancer-causing genomic alterations to create a comprehensive \u201catlas\u201d of cancer genomic profiles. So far, TCGA researchers have analysed large cohorts of over 30 human tumours through large-scale genome sequencing and integrated multi-dimensional analyses. Studies of individual cancer types, as well as comprehensive pan-cancer analyses have extended current knowledge of tumorigenesis. A major goal of the project was to provide publicly available datasets to help improve diagnostic methods, treatment standards, and finally to prevent cancer. This review discusses the current status of TCGA Research Network structure, purpose, and achievements."
                },
                {
                    "title": "A Framework for Multiple-Instance Learning",
                    "abstract": "Multiple-instance learning is a variation on supervised learning, where the task is to learn a concept given positive and negative bags of instances. Each bag may contain many instances, but a bag is labeled positive even if only one of the instances in it falls within the concept. A bag is labeled negative only if all the instances in it are negative. We describe a new general framework, called Diverse Density, for solving multiple-instance learning problems. We apply this framework to learn a simple description of a person from a series of images (bags) containing that person, to a stock selection problem, and to the drug activity prediction problem."
                },
                {
                    "title": "ADR: Attention Diversification Regularization for Mitigating Overfitting in Multiple Instance Learning based Whole Slide Image Classification",
                    "abstract": ". Multiple Instance Learning (MIL) has demonstrated effectiveness in analyzing whole slide images (WSIs), yet it often encounters overfitting challenges in real-world applications. This paper reveals the correlation between MIL\u2019s performance and the entropy of attention values. Based on this observation, we propose Attention Diversity Regularization (ADR), a simple but effective technique aimed at promoting high entropy in attention values. Specifically, ADR introduces a negative Shannon entropy loss for attention values into the regular MIL framework. Compared to existing methods aimed at alleviating overfitting, which often necessitate additional modules or processing steps, our ADR approach requires no such extras, demonstrating simplicity and efficiency. We evaluate our ADR on three WSI classification tasks. ADR achieves superior performance over the state-of-the-art on most of them. We also show that ADR can enhance heatmaps, aligning them better with pathologists\u2019 diagnostic criteria. The source code is available at https://github.com/dazhangyu123/ADR ."
                },
                {
                    "title": "Exploring Low-Rank Property in Multiple Instance Learning for Whole Slide Image Classification",
                    "abstract": "ILRA-MIL"
                },
                {
                    "title": "Bayes-MIL: A New Probabilistic Perspective on Attention-based Multiple Instance Learning for Whole Slide Images",
                    "abstract": "Multiple instance learning (MIL) is a popular weakly-supervised learning model on the whole slide image (WSI) for AI-assisted pathology diagnosis. The recent advance in attention-based MIL allows the model to find its region-of-interest (ROI) for interpretation by learning the attention weights for image patches of WSI slides. However, we empirically find that the interpretability of some related methods is either untrustworthy as the principle of MIL is violated or unsatisfactory as the high-attention regions are not consistent with experts\u2019 annotations. In this paper, we propose Bayes-MIL to address the problem from a probabilistic perspective. The induced patch-level uncertainty is proposed as a new measure of MIL interpretability, which outperforms previous methods in matching doctors annotations. We design a slide-dependent patch regularizer (SDPR) for the attention, imposing constraints derived from the MIL assumption, on the attention distribution. SDPR explicitly constrains the model to generate correct attention values. The spatial information is further encoded by an approximate convolutional conditional random field (CRF), for better interpretability. Experimental results show Bayes-MIL outperforms the related methods in patch-level and slide-level metrics and provides much better interpretable ROI on several large-scale WSI datasets."
                },
                {
                    "title": "Retrieval-Augmented Multiple Instance Learning",
                    "abstract": "Multiple Instance Learning (MIL) is a crucial weakly supervised learning method applied across various domains, e.g., medical diagnosis based on whole slide images (WSIs). Recent advancements in MIL algorithms have yielded exceptional performance when the training and test data originate from the same domain, such as WSIs obtained from the same hospital. However, this paper reveals a performance deterioration of MIL models when tested on an out-of-domain test set, exemplified by WSIs sourced from a novel hospital. To address this challenge, this paper introduces the Retrieval-AugMented MIL (RAM-MIL) framework, which integrates Optimal Transport (OT) as the distance metric for nearest neighbor retrieval. The development of RAM-MIL is driven by two key insights. First, a theoretical discovery indicates that reducing the input\u2019s intrinsic dimension can minimize the approximation error in attention-based MIL. Second, previous studies highlight a link between input intrinsic dimension and the feature merging process with the retrieved data. Empirical evaluations conducted on WSI classification demonstrate that the proposed RAM-MIL framework achieves state-of-the-art performance in both in-domain scenarios, where the training and retrieval data are in the same domain, and more crucially, in out-of-domain scenarios, where the (unlabeled) retrieval data originates from a different domain. Furthermore, the use of the transportation matrix derived from OT renders the retrieval results interpretable at the instance level, in contrast to the vanilla l 2 distance, and allows for visualization for human experts."
                },
                {
                    "title": "SCL-WC: Cross-Slide Contrastive Learning for Weakly-Supervised Whole-Slide Image Classification",
                    "abstract": "Weakly-supervised whole-slide image (WSI) classi\ufb01cation (WSWC) is a challenging task where a large number of unlabeled patches (instances) exist within each WSI (bag) while only a slide label is given. Despite recent progress for the multiple instance learning (MIL)-based WSI analysis, the major limitation is that it usually focuses on the easy-to-distinguish diagnosis-positive regions while ignoring positives that occupy a small ratio in the entire WSI. To obtain more discriminative features, we propose a novel weakly-supervised classi\ufb01cation method based on cross-slide contrastive learning (called SCL-WC), which depends on task-agnostic self-supervised feature pre-extraction and task-speci\ufb01c weakly-supervised feature re\ufb01nement and aggregation for WSI-level prediction. To enable both intra-WSI and inter-WSI information interaction, we propose a positive-negative-aware module (PNM) and a weakly-supervised cross-slide contrastive learning (WSCL) module, respectively. The WSCL aims to pull WSIs with the same disease types closer and push different WSIs away. The PNM aims to facilitate the separation of tumor-like patches and normal ones within each WSI. Extensive experiments demonstrate state-of-the-art performance of our method in three different classi-\ufb01cation tasks (e.g., over 2% of AUC in Camelyon16, 5% of F1 score in BRACS, and 3% of AUC in DiagSet). Our method also shows superior \ufb02exibility and scalability in weakly-supervised localization and semi-supervised classi\ufb01cation experiments (e.g., \ufb01rst place in the BRIGHT challenge). Our code will be available at https://github.com/Xiyue-Wang/SCL-WC."
                },
                {
                    "title": "Dual-Curriculum Contrastive Multi-Instance Learning for Cancer Prognosis Analysis with Whole Slide Images",
                    "abstract": "The multi-instance learning (MIL) has advanced cancer prognosis analysis with whole slide images (WSIs). However, current MIL methods for WSI analysis still confront unique challenges. Previous methods typically generate instance representations via a pre-trained model or a model trained by the instances with bag-level annotations, which, however, may not generalize well to the downstream task due to the introduction of excessive label noises and the lack of \ufb01ne-grained information across multi-magni\ufb01cation WSIs. Additionally, existing methods generally aggregate instance representations as bag ones for prognosis prediction and have no consideration of intra-bag redundancy and inter-bag discrimination. To address these issues, we propose a dual-curriculum contrastive MIL method for cancer prognosis analysis with WSIs. The proposed method consists of two curriculums, i.e., saliency-guided weakly-supervised instance encoding with cross-scale tiles and contrastive-enhanced soft-bag prognosis inference. Extensive experiments on three public datasets demonstrate that our method outperforms state-of-the-art methods in this \ufb01eld. The code is available at https://github.com/YuZhang-SMU/Cancer-Prognosis-Analysis/tree/main/DC_MIL%20Code."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop an effective and efficient Transformer model for Whole Slide Image (WSI) analysis that overcomes the computational complexity and low-rank bottleneck of attention mechanisms?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing digital pathology, as it can significantly enhance the accuracy and efficiency of cancer diagnosis and prognosis through improved analysis of WSI. By addressing the limitations of current models, this research could lead to more reliable automated systems, ultimately benefiting the medical community and patients. Furthermore, it could inspire future research in machine learning techniques for high-dimensional data analysis, paving the way for innovative applications in other fields that deal with large-scale image data.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the extremely high resolution of WSIs, which leads to a long sequence of instances that result in O(n\u00b2) computational complexity for traditional self-attention mechanisms. Naive approaches may fail because they cannot effectively model the contextual interactions across instances due to this complexity. Additionally, the low-rank bottleneck of the attention matrix limits the model's ability to capture both local contexts and global interactions simultaneously, creating a significant theoretical and practical obstacle.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on annotation and computational efficiency, often using methods like Multiple Instance Learning (MIL) with WSI-level supervision. However, these approaches have not adequately addressed the interaction modeling between instances or the computational limitations of self-attention mechanisms. Existing solutions, such as self-attention approximations and non-overlapping region slicing, have proven sub-optimal or have ignored critical interactions. Our approach aims to tackle these gaps by proposing a novel Transformer architecture that effectively balances local and global context modeling without succumbing to the low-rank bottleneck.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves designing a new Transformer architecture specifically tailored for WSI analysis, which incorporates a multi-layer self-attention mechanism that can effectively capture both local and global interactions. We will utilize a large dataset of WSIs for training and evaluate our model using metrics such as accuracy, F1-score, and computational efficiency. The expected outcomes include improved performance in WSI classification tasks, demonstrating the model's ability to handle the complexities of high-resolution images while providing insights into the interactions between different instances."
            }
        },
        "author_data": {
            "25c63fd1-02df-468c-a07f-ff1e937d934c": {
                "pk": "25c63fd1-02df-468c-a07f-ff1e937d934c",
                "name": "Honglin Li",
                "collaborators": [
                    "Payam Barnaghi",
                    "Frieder Ganz",
                    "Shirin Enshaeifar",
                    "Lin Yang",
                    "Shichuan Zhang",
                    "Chenglu Zhu",
                    "Sunyi Zheng",
                    "Yifei Fan",
                    "Anthony Yezzi",
                    "Jiatong Cai"
                ],
                "domain": [
                    "Continual Learning",
                    "Adversarial Learning",
                    "Multi-modal Learning",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "Kalman Filter Modifier for Neural Networks in Non-stationary Environments",
                        "abstract": "Learning in a non-stationary environment is an inevitable problem when applying machine learning algorithm to real world environment. Learning new tasks without forgetting the previous knowledge is a challenge issue in machine learning. We propose a Kalman Filter based modifier to maintain the performance of Neural Network models under non-stationary environments. The result shows that our proposed model can preserve the key information and adapts better to the changes. The accuracy of proposed model decreases by 0.4% in our experiments, while the accuracy of conventional model decreases by 90% in the drifts environment."
                    },
                    {
                        "title": "Continual Learning in Deep Neural Network by Using a Kalman Optimiser",
                        "abstract": "Learning and adapting to new distributions or learning new tasks sequentially without forgetting the previously learned knowledge is a challenging phenomenon in continual learning models. Most of the conventional deep learning models are not capable of learning new tasks sequentially in one model without forgetting the previously learned ones. We address this issue by using a Kalman Optimiser. The Kalman Optimiser divides the neural network into two parts: the long-term and short-term memory units. The long-term memory unit is used to remember the learned tasks and the short-term memory unit is to adapt to the new task. We have evaluated our method on MNIST, CIFAR10, CIFAR100 datasets and compare our results with state-of-the-art baseline models. The results show that our approach enables the model to continually learn and adapt to the new changes without forgetting the previously learned tasks."
                    },
                    {
                        "title": "Continual Learning Using Multi-view Task Conditional Neural Networks",
                        "abstract": "Conventional deep learning models have limited capacity in learning multiple tasks sequentially. The issue of forgetting the previously learned tasks in continual learning is known as catastrophic forgetting or interference. When the input data or the goal of learning change, a continual model will learn and adapt to the new status. However, the model will not remember or recognise any revisits to the previous states. This causes performance reduction and re-training curves in dealing with periodic or irregularly reoccurring changes in the data or goals. The changes in goals or data are referred to as new tasks in a continual learning model. Most of the continual learning methods have a task-known setup in which the task identities are known in advance to the learning model. We propose Multi-view Task Conditional Neural Networks (Mv-TCNN) that does not require to known the reoccurring tasks in advance. We evaluate our model on standard datasets using MNIST, CIFAR10, CIFAR100, and also a real-world dataset that we have collected in a remote healthcare monitoring study (i.e. TIHM dataset). The proposed model outperforms the state-of-the-art solutions in continual learning and adapting to new tasks that are not defined in advance."
                    },
                    {
                        "title": "Verifying the Causes of Adversarial Examples",
                        "abstract": "The robustness of neural networks is challenged by adversarial examples that contain almost imperceptible perturbations to inputs, which mislead a classifier to incorrect outputs in high confidence. Limited by the extreme difficulty in examining a high-dimensional image space thoroughly, research on explaining and justifying the causes of adversarial examples falls behind studies on attacks and defenses. In this paper, we present a collection of potential causes of adversarial examples and verify (or partially verify) them through carefully-designed controlled experiments. The major causes of adversarial examples include model linearity, one-sum constraint, and geometry of the categories. To control the effect of those causes, multiple techniques are applied such as $L_2$ normalization, replacement of loss functions, construction of reference datasets, and novel models using multi-layer perceptron probabilistic neural networks (MLP-PNN) and density estimation (DE). Our experiment results show that geometric factors tend to be more direct causes and statistical factors magnify the phenomenon, especially for assigning high prediction confidence. We believe this paper will inspire more studies to rigorously investigate the root causes of adversarial examples, which in turn provide useful guidance on designing more robust models."
                    },
                    {
                        "title": "Weakly Supervised Learning for cell recognition in immunohistochemical cytoplasm staining images",
                        "abstract": "Cell classification and counting in immunohistochemical cytoplasm staining images play a pivotal role in cancer diagnosis. Weakly supervised learning is a potential method to deal with labor-intensive labeling. However, the inconstant cell morphology and subtle differences between classes also bring challenges. To this end, we present a novel cell recognition framework based on multi-task learning, which utilizes two additional auxiliary tasks to guide robust representation learning of the main task. To deal with misclassification, the tissue prior learning branch is introduced to capture the spatial representation of tumor cells without additional tissue annotation. Moreover, dynamic masks and consistency learning are adopted to learn the invariance of cell scale and shape. We have evaluated our framework on immunohistochemical cytoplasm staining images, and the results demonstrate that our method outperforms recent cell recognition approaches. Besides, we have also done some ablation studies to show significant improvements after adding the auxiliary branches."
                    },
                    {
                        "title": "WSI-VQA: Interpreting Whole Slide Images by Generative Visual Question Answering",
                        "abstract": "Whole slide imaging is routinely adopted for carcinoma diagnosis and prognosis. Abundant experience is required for pathologists to achieve accurate and reliable diagnostic results of whole slide images (WSI). The huge size and heterogeneous features of WSIs make the workflow of pathological reading extremely time-consuming. In this paper, we propose a novel framework (WSI-VQA) to interpret WSIs by generative visual question answering. WSI-VQA shows universality by reframing various kinds of slide-level tasks in a question-answering pattern, in which pathologists can achieve immunohistochemical grading, survival prediction, and tumor subtyping following human-machine interaction. Furthermore, we establish a WSI-VQA dataset which contains 8672 slide-level question-answering pairs with 977 WSIs. Besides the ability to deal with different slide-level tasks, our generative model which is named Wsi2Text Transformer (W2T) outperforms existing discriminative models in medical correctness, which reveals the potential of our model to be applied in the clinical scenario. Additionally, we also visualize the co-attention mapping between word embeddings and WSIs as an intuitive explanation for diagnostic results. The dataset and related code are available at https://github.com/cpystan/WSI-VQA."
                    },
                    {
                        "title": "Semi-supervised Federated Learning for Activity Recognition",
                        "abstract": "Training deep learning models on in-home IoT sensory data is commonly used to recognise human activities. Recently, federated learning systems that use edge devices as clients to support local human activity recognition have emerged as a new paradigm to combine local (individual-level) and global (group-level) models. This approach provides better scalability and generalisability and also offers better privacy compared with the traditional centralised analysis and learning models. The assumption behind federated learning, however, relies on supervised learning on clients. This requires a large volume of labelled data, which is difficult to collect in uncontrolled IoT environments such as remote in-home monitoring.   In this paper, we propose an activity recognition system that uses semi-supervised federated learning, wherein clients conduct unsupervised learning on autoencoders with unlabelled local data to learn general representations, and a cloud server conducts supervised learning on an activity classifier with labelled data. Our experimental results show that using a long short-term memory autoencoder and a Softmax classifier, the accuracy of our proposed system is higher than that of both centralised systems and semi-supervised federated learning using data augmentation. The accuracy is also comparable to that of supervised federated learning systems. Meanwhile, we demonstrate that our system can reduce the number of needed labels and the size of local models, and has faster local activity recognition speed than supervised federated learning does."
                    },
                    {
                        "title": "Multi-modal Learning with Missing Modality in Predicting Axillary Lymph Node Metastasis",
                        "abstract": "Multi-modal Learning has attracted widespread attention in medical image analysis. Using multi-modal data, whole slide images (WSIs) and clinical information, can improve the performance of deep learning models in the diagnosis of axillary lymph node metastasis. However, clinical information is not easy to collect in clinical practice due to privacy concerns, limited resources, lack of interoperability, etc. Although patient selection can ensure the training set to have multi-modal data for model development, missing modality of clinical information can appear during test. This normally leads to performance degradation, which limits the use of multi-modal models in the clinic. To alleviate this problem, we propose a bidirectional distillation framework consisting of a multi-modal branch and a single-modal branch. The single-modal branch acquires the complete multi-modal knowledge from the multi-modal branch, while the multi-modal learns the robust features of WSI from the single-modal. We conduct experiments on a public dataset of Lymph Node Metastasis in Early Breast Cancer to validate the method. Our approach not only achieves state-of-the-art performance with an AUC of 0.861 on the test set without missing data, but also yields an AUC of 0.842 when the rate of missing modality is 80\\%. This shows the effectiveness of the approach in dealing with multi-modal data and missing modality. Such a model has the potential to improve treatment decision-making for early breast cancer patients who have axillary lymph node metastatic status."
                    }
                ]
            },
            "31dc6d39-2918-4df7-a2cc-def10ddfdbc7": {
                "pk": "31dc6d39-2918-4df7-a2cc-def10ddfdbc7",
                "name": "Yunlong Zhang",
                "collaborators": [
                    "Peilong Wang",
                    "Xiaolian Wang",
                    "Zizong Xu",
                    "Xiaoqiang Kong",
                    "Subasish Das",
                    "Hongmin Zhou"
                ],
                "domain": [
                    "Naturalistic Driving",
                    "Photomultiplier Tube",
                    "Association Rule Mining",
                    "Temperature Dependence"
                ],
                "publications": [
                    {
                        "title": "Patterns of near-crash events in a naturalistic driving dataset: applying rules mining",
                        "abstract": "This study aims to explore the associations between near-crash events and road geometry and trip features by investigating a naturalistic driving dataset and a corresponding roadway inventory dataset using an association rule mining method."
                    },
                    {
                        "title": "Magnetic shield of PMT used in DAMPE electromagnetic calorimeter",
                        "abstract": "The magnetic characteristics of photomultiplier tube R5610A-01 are studied in this paper. The experimental data shows that the gain of R5610A-01 loses about 53% when the magnetic field is 3G along its +X axis. A cylinder of one-layer permalloy strip is able to reduce the effect of 3G magnetic field on the PMT's gain to less than 1%."
                    },
                    {
                        "title": "Study on the temperature dependence of BGO light yield",
                        "abstract": "The temperature dependence of BGO coupled with photomultiplier tube R5610A-01 was studied in the range of -30{\\deg}C~30{\\deg}C. The temperature coefficient of the BGO and R5610A in whole was tested to be -1.82%. And the temperature coefficient of the gain of the R5610A is -0.44% which was tested in the same situation using a blue LED. Thus the temperature coefficient of BGO's light yield can be evaluated as -1.38%."
                    }
                ]
            },
            "630ce046-5d17-405c-83c5-5c84b1bec2c4": {
                "pk": "630ce046-5d17-405c-83c5-5c84b1bec2c4",
                "name": "Pingyi Chen",
                "collaborators": [
                    "Chenglu Zhu",
                    "Sunyi Zheng",
                    "Lin Yang",
                    "Honglin Li",
                    "Zhongyi Shui",
                    "Jiatong Cai",
                    "Shichuan Zhang"
                ],
                "domain": [
                    "Medical Image Analysis",
                    "Self-Supervised Learning",
                    "Deep Learning",
                    "Visual Question Answering"
                ],
                "publications": [
                    {
                        "title": "WSI-VQA: Interpreting Whole Slide Images by Generative Visual Question Answering",
                        "abstract": "Whole slide imaging is routinely adopted for carcinoma diagnosis and prognosis. Abundant experience is required for pathologists to achieve accurate and reliable diagnostic results of whole slide images (WSI). The huge size and heterogeneous features of WSIs make the workflow of pathological reading extremely time-consuming. In this paper, we propose a novel framework (WSI-VQA) to interpret WSIs by generative visual question answering. WSI-VQA shows universality by reframing various kinds of slide-level tasks in a question-answering pattern, in which pathologists can achieve immunohistochemical grading, survival prediction, and tumor subtyping following human-machine interaction. Furthermore, we establish a WSI-VQA dataset which contains 8672 slide-level question-answering pairs with 977 WSIs. Besides the ability to deal with different slide-level tasks, our generative model which is named Wsi2Text Transformer (W2T) outperforms existing discriminative models in medical correctness, which reveals the potential of our model to be applied in the clinical scenario. Additionally, we also visualize the co-attention mapping between word embeddings and WSIs as an intuitive explanation for diagnostic results. The dataset and related code are available at https://github.com/cpystan/WSI-VQA."
                    },
                    {
                        "title": "Unsupervised Dense Nuclei Detection and Segmentation with Prior Self-activation Map For Histology Images",
                        "abstract": "The success of supervised deep learning models in medical image segmentation relies on detailed annotations. However, labor-intensive manual labeling is costly and inefficient, especially in dense object segmentation. To this end, we propose a self-supervised learning based approach with a Prior Self-activation Module (PSM) that generates self-activation maps from the input images to avoid labeling costs and further produce pseudo masks for the downstream task. To be specific, we firstly train a neural network using self-supervised learning and utilize the gradient information in the shallow layers of the network to generate self-activation maps. Afterwards, a semantic-guided generator is then introduced as a pipeline to transform visual representations from PSM to pixel-level semantic pseudo masks for downstream tasks. Furthermore, a two-stage training module, consisting of a nuclei detection network and a nuclei segmentation network, is adopted to achieve the final segmentation. Experimental results show the effectiveness on two public pathological datasets. Compared with other fully-supervised and weakly-supervised methods, our method can achieve competitive performance without any manual annotations."
                    }
                ]
            },
            "365e2776-bad2-4de2-8114-59131803ead9": {
                "pk": "365e2776-bad2-4de2-8114-59131803ead9",
                "name": "Zhongyi Shui",
                "collaborators": [
                    "Lin Yang",
                    "Sunyi Zheng",
                    "Chenglu Zhu",
                    "Honglin Li",
                    "Yunlong Zhang",
                    "Shichuan Zhang",
                    "Jingxiong Li",
                    "Pingyi Chen",
                    "Yuxuan Sun",
                    "Jiatong Cai"
                ],
                "domain": [
                    "Medical Image Analysis",
                    "Multi-modal Learning",
                    "Deep Learning",
                    "Self-supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Multi-modal Learning with Missing Modality in Predicting Axillary Lymph Node Metastasis",
                        "abstract": "Multi-modal Learning has attracted widespread attention in medical image analysis. Using multi-modal data, whole slide images (WSIs) and clinical information, can improve the performance of deep learning models in the diagnosis of axillary lymph node metastasis. However, clinical information is not easy to collect in clinical practice due to privacy concerns, limited resources, lack of interoperability, etc. Although patient selection can ensure the training set to have multi-modal data for model development, missing modality of clinical information can appear during test. This normally leads to performance degradation, which limits the use of multi-modal models in the clinic. To alleviate this problem, we propose a bidirectional distillation framework consisting of a multi-modal branch and a single-modal branch. The single-modal branch acquires the complete multi-modal knowledge from the multi-modal branch, while the multi-modal learns the robust features of WSI from the single-modal. We conduct experiments on a public dataset of Lymph Node Metastasis in Early Breast Cancer to validate the method. Our approach not only achieves state-of-the-art performance with an AUC of 0.861 on the test set without missing data, but also yields an AUC of 0.842 when the rate of missing modality is 80\\%. This shows the effectiveness of the approach in dealing with multi-modal data and missing modality. Such a model has the potential to improve treatment decision-making for early breast cancer patients who have axillary lymph node metastatic status."
                    },
                    {
                        "title": "ChrSNet: Chromosome Straightening using Self-attention Guided Networks",
                        "abstract": "Karyotyping is an important procedure to assess the possible existence of chromosomal abnormalities. However, because of the non-rigid nature, chromosomes are usually heavily curved in microscopic images and such deformed shapes hinder the chromosome analysis for cytogeneticists. In this paper, we present a self-attention guided framework to erase the curvature of chromosomes. The proposed framework extracts spatial information and local textures to preserve banding patterns in a regression module. With complementary information from the bent chromosome, a refinement module is designed to further improve fine details. In addition, we propose two dedicated geometric constraints to maintain the length and restore the distortion of chromosomes. To train our framework, we create a synthetic dataset where curved chromosomes are generated from the real-world straight chromosomes by grid-deformation. Quantitative and qualitative experiments are conducted on synthetic and real-world data. Experimental results show that our proposed method can effectively straighten bent chromosomes while keeping banding details and length."
                    },
                    {
                        "title": "Unsupervised Dense Nuclei Detection and Segmentation with Prior Self-activation Map For Histology Images",
                        "abstract": "The success of supervised deep learning models in medical image segmentation relies on detailed annotations. However, labor-intensive manual labeling is costly and inefficient, especially in dense object segmentation. To this end, we propose a self-supervised learning based approach with a Prior Self-activation Module (PSM) that generates self-activation maps from the input images to avoid labeling costs and further produce pseudo masks for the downstream task. To be specific, we firstly train a neural network using self-supervised learning and utilize the gradient information in the shallow layers of the network to generate self-activation maps. Afterwards, a semantic-guided generator is then introduced as a pipeline to transform visual representations from PSM to pixel-level semantic pseudo masks for downstream tasks. Furthermore, a two-stage training module, consisting of a nuclei detection network and a nuclei segmentation network, is adopted to achieve the final segmentation. Experimental results show the effectiveness on two public pathological datasets. Compared with other fully-supervised and weakly-supervised methods, our method can achieve competitive performance without any manual annotations."
                    },
                    {
                        "title": "End-to-end cell recognition by point annotation",
                        "abstract": "Reliable quantitative analysis of immunohistochemical staining images requires accurate and robust cell detection and classification. Recent weakly-supervised methods usually estimate probability density maps for cell recognition. However, in dense cell scenarios, their performance can be limited by pre- and post-processing as it is impossible to find a universal parameter setting. In this paper, we introduce an end-to-end framework that applies direct regression and classification for preset anchor points. Specifically, we propose a pyramidal feature aggregation strategy to combine low-level features and high-level semantics simultaneously, which provides accurate cell recognition for our purely point-based model. In addition, an optimized cost function is designed to adapt our multi-task learning framework by matching ground truth and predicted points. The experimental results demonstrate the superior accuracy and efficiency of the proposed method, which reveals the high potentiality in assisting pathologist assessments."
                    },
                    {
                        "title": "Exploring Unsupervised Cell Recognition with Prior Self-activation Maps",
                        "abstract": "The success of supervised deep learning models on cell recognition tasks relies on detailed annotations. Many previous works have managed to reduce the dependency on labels. However, considering the large number of cells contained in a patch, costly and inefficient labeling is still inevitable. To this end, we explored label-free methods for cell recognition. Prior self-activation maps (PSM) are proposed to generate pseudo masks as training targets. To be specific, an activation network is trained with self-supervised learning. The gradient information in the shallow layers of the network is aggregated to generate prior self-activation maps. Afterward, a semantic clustering module is then introduced as a pipeline to transform PSMs to pixel-level semantic pseudo masks for downstream tasks. We evaluated our method on two histological datasets: MoNuSeg (cell segmentation) and BCData (multi-class cell detection). Compared with other fully-supervised and weakly-supervised methods, our method can achieve competitive performance without any manual annotations. Our simple but effective framework can also achieve multi-class cell detection which can not be done by existing unsupervised methods. The results show the potential of PSMs that might inspire other research to deal with the hunger for labels in medical area."
                    },
                    {
                        "title": "WsiCaption: Multiple Instance Generation of Pathology Reports for Gigapixel Whole-Slide Images",
                        "abstract": "Whole slide images are the foundation of digital pathology for the diagnosis and treatment of carcinomas. Writing pathology reports is laborious and error-prone for inexperienced pathologists. To reduce the workload and improve clinical automation, we investigate how to generate pathology reports given whole slide images. On the data end, we curated the largest WSI-text dataset (PathText). In specific, we collected nearly 10000 high-quality WSI-text pairs for visual-language models by recognizing and cleaning pathology reports which narrate diagnostic slides in TCGA. On the model end, we propose the multiple instance generative model (MI-Gen) which can produce pathology reports for gigapixel WSIs. We benchmark our model on the largest subset of TCGA-PathoText. Experimental results show our model can generate pathology reports which contain multiple clinical clues and achieve competitive performance on certain slide-level tasks. We observe that simple semantic extraction from the pathology reports can achieve the best performance (0.838 of F1 score) on BRCA subtyping surpassing previous state-of-the-art approaches. Our collected dataset and related code are available."
                    },
                    {
                        "title": "Semi-supervised Cell Recognition under Point Supervision",
                        "abstract": "Cell recognition is a fundamental task in digital histopathology image analysis. Point-based cell recognition (PCR) methods normally require a vast number of annotations, which is extremely costly, time-consuming and labor-intensive. Semi-supervised learning (SSL) can provide a shortcut to make full use of cell information in gigapixel whole slide images without exhaustive labeling. However, research into semi-supervised point-based cell recognition (SSPCR) remains largely overlooked. Previous SSPCR works are all built on density map-based PCR models, which suffer from unsatisfactory accuracy, slow inference speed and high sensitivity to hyper-parameters. To address these issues, end-to-end PCR models are proposed recently. In this paper, we develop a SSPCR framework suitable for the end-to-end PCR models for the first time. Overall, we use the current models to generate pseudo labels for unlabeled images, which are in turn utilized to supervise the models training. Besides, we introduce a co-teaching strategy to overcome the confirmation bias problem that generally exists in self-training. A distribution alignment technique is also incorporated to produce high-quality, unbiased pseudo labels for unlabeled data. Experimental results on four histopathology datasets concerning different types of staining styles show the effectiveness and versatility of the proposed framework. Code is available at \\textcolor{magenta}{\\url{https://github.com/windygooo/SSPCR}"
                    },
                    {
                        "title": "DPA-P2PNet: Deformable Proposal-aware P2PNet for Accurate Point-based Cell Detection",
                        "abstract": "Point-based cell detection (PCD), which pursues high-performance cell sensing under low-cost data annotation, has garnered increased attention in computational pathology community. Unlike mainstream PCD methods that rely on intermediate density map representations, the Point-to-Point network (P2PNet) has recently emerged as an end-to-end solution for PCD, demonstrating impressive cell detection accuracy and efficiency. Nevertheless, P2PNet is limited to decoding from a single-level feature map due to the scale-agnostic property of point proposals, which is insufficient to leverage multi-scale information. Moreover, the spatial distribution of pre-set point proposals is biased from that of cells, leading to inaccurate cell localization. To lift these limitations, we present DPA-P2PNet in this work. The proposed method directly extracts multi-scale features for decoding according to the coordinates of point proposals on hierarchical feature maps. On this basis, we further devise deformable point proposals to mitigate the positional bias between proposals and potential cells to promote cell localization. Inspired by practical pathological diagnosis that usually combines high-level tissue structure and low-level cell morphology for accurate cell classification, we propose a multi-field-of-view (mFoV) variant of DPA-P2PNet to accommodate additional large FoV images with tissue information as model input. Finally, we execute the first self-supervised pre-training on immunohistochemistry histopathology image data and evaluate the suitability of four representative self-supervised methods on the PCD task. Experimental results on three benchmarks and a large-scale and real-world interval dataset demonstrate the superiority of our proposed models over the state-of-the-art counterparts. Codes and pre-trained weights will be available."
                    },
                    {
                        "title": "Task-specific Fine-tuning via Variational Information Bottleneck for Weakly-supervised Pathology Whole Slide Image Classification",
                        "abstract": "While Multiple Instance Learning (MIL) has shown promising results in digital Pathology Whole Slide Image (WSI) classification, such a paradigm still faces performance and generalization problems due to challenges in high computational costs on Gigapixel WSIs and limited sample size for model training. To deal with the computation problem, most MIL methods utilize a frozen pretrained model from ImageNet to obtain representations first. This process may lose essential information owing to the large domain gap and hinder the generalization of model due to the lack of image-level training-time augmentations. Though Self-supervised Learning (SSL) proposes viable representation learning schemes, the improvement of the downstream task still needs to be further explored in the conversion from the task-agnostic features of SSL to the task-specifics under the partial label supervised learning. To alleviate the dilemma of computation cost and performance, we propose an efficient WSI fine-tuning framework motivated by the Information Bottleneck theory. The theory enables the framework to find the minimal sufficient statistics of WSI, thus supporting us to fine-tune the backbone into a task-specific representation only depending on WSI-level weak labels. The WSI-MIL problem is further analyzed to theoretically deduce our fine-tuning method. Our framework is evaluated on five pathology WSI datasets on various WSI heads. The experimental results of our fine-tuned representations show significant improvements in both accuracy and generalization compared with previous works. Source code will be available at https://github.com/invoker-LL/WSI-finetuning."
                    },
                    {
                        "title": "PathAsst: A Generative Foundation AI Assistant Towards Artificial General Intelligence of Pathology",
                        "abstract": "As advances in large language models (LLMs) and multimodal techniques continue to mature, the development of general-purpose multimodal large language models (MLLMs) has surged, offering significant applications in interpreting natural images. However, the field of pathology has largely remained untapped, particularly in gathering high-quality data and designing comprehensive model frameworks. To bridge the gap in pathology MLLMs, we present PathAsst, a multimodal generative foundation AI assistant to revolutionize diagnostic and predictive analytics in pathology. The development of PathAsst involves three pivotal steps: data acquisition, CLIP model adaptation, and the training of PathAsst's multimodal generative capabilities. Firstly, we collect over 207K high-quality pathology image-text pairs from authoritative sources. Leveraging the advanced power of ChatGPT, we generate over 180K instruction-following samples. Furthermore, we devise additional instruction-following data specifically tailored for invoking eight pathology-specific sub-models we prepared, allowing the PathAsst to effectively collaborate with these models, enhancing its diagnostic ability. Secondly, by leveraging the collected data, we construct PathCLIP, a pathology-dedicated CLIP, to enhance PathAsst's capabilities in interpreting pathology images. Finally, we integrate PathCLIP with the Vicuna-13b and utilize pathology-specific instruction-tuning data to enhance the multimodal generation capacity of PathAsst and bolster its synergistic interactions with sub-models. The experimental results of PathAsst show the potential of harnessing AI-powered generative foundation model to improve pathology diagnosis and treatment processes."
                    },
                    {
                        "title": "Test-Time Training for Semantic Segmentation with Output Contrastive Loss",
                        "abstract": "Although deep learning-based segmentation models have achieved impressive performance on public benchmarks, generalizing well to unseen environments remains a major challenge. To improve the model's generalization ability to the new domain during evaluation, the test-time training (TTT) is a challenging paradigm that adapts the source-pretrained model in an online fashion. Early efforts on TTT mainly focus on the image classification task. Directly extending these methods to semantic segmentation easily experiences unstable adaption due to segmentation's inherent characteristics, such as extreme class imbalance and complex decision spaces. To stabilize the adaptation process, we introduce contrastive loss (CL), known for its capability to learn robust and generalized representations. Nevertheless, the traditional CL operates in the representation space and cannot directly enhance predictions. In this paper, we resolve this limitation by adapting the CL to the output space, employing a high temperature, and simplifying the formulation, resulting in a straightforward yet effective loss function called Output Contrastive Loss (OCL). Our comprehensive experiments validate the efficacy of our approach across diverse evaluation scenarios. Notably, our method excels even when applied to models initially pre-trained using domain adaptation methods on test domain data, showcasing its resilience and adaptability.\\footnote{Code and more information could be found at~ \\url{https://github.com/dazhangyu123/OCL}}"
                    },
                    {
                        "title": "AEM: Attention Entropy Maximization for Multiple Instance Learning based Whole Slide Image Classification",
                        "abstract": "Multiple Instance Learning (MIL) has demonstrated effectiveness in analyzing whole slide images (WSIs), yet it often encounters overfitting challenges in real-world applications, particularly in the form of attention over-concentration. While existing methods to alleviate this issue introduce complex modules or processing steps, such as multiple-stage training and teacher-student distillation, this paper proposes a simple yet effective regularization: Attention Entropy Maximization (AEM). Motivated by our investigation revealing a positive correlation between attention entropy and model performance, AEM incorporates a negative entropy loss for attention values into the standard MIL framework, penalizing overly concentrated attention and encouraging the model to consider a broader range of informative regions in WSIs, potentially improving its generalization capabilities. Compared to existing overfitting mitigation methods, our AEM approach offers advantages of simplicity, efficiency, and versatility. It requires no additional modules or processing steps, involves only one hyperparameter, and demonstrates compatibility with MIL frameworks and techniques. These advantages make AEM particularly attractive for practical applications. We evaluate AEM on three benchmark datasets, demonstrating consistent performance improvements over existing methods. Furthermore, AEM shows high versatility, integrating effectively with four feature extractors, two advanced MIL frameworks, three attention mechanisms, and Subsampling augmentation technique. The source code is available at \\url{https://github.com/dazhangyu123/AEM}."
                    },
                    {
                        "title": "PathGen-1.6M: 1.6 Million Pathology Image-text Pairs Generation through Multi-agent Collaboration",
                        "abstract": "Vision Language Models (VLMs) like CLIP have attracted substantial attention in pathology, serving as backbones for applications such as zero-shot image classification and Whole Slide Image (WSI) analysis. Additionally, they can function as vision encoders when combined with large language models (LLMs) to support broader capabilities. Current efforts to train pathology VLMs rely on pathology image-text pairs from platforms like PubMed, YouTube, and Twitter, which provide limited, unscalable data with generally suboptimal image quality. In this work, we leverage large-scale WSI datasets like TCGA to extract numerous high-quality image patches. We then train a large multimodal model to generate captions for these images, creating PathGen-1.6M, a dataset containing 1.6 million high-quality image-caption pairs. Our approach involves multiple agent models collaborating to extract representative WSI patches, generating and refining captions to obtain high-quality image-text pairs. Extensive experiments show that integrating these generated pairs with existing datasets to train a pathology-specific CLIP model, PathGen-CLIP, significantly enhances its ability to analyze pathological images, with substantial improvements across nine pathology-related zero-shot image classification tasks and three whole-slide image tasks. Furthermore, we construct 200K instruction-tuning data based on PathGen-1.6M and integrate PathGen-CLIP with the Vicuna LLM to create more powerful multimodal models through instruction tuning. Overall, we provide a scalable pathway for high-quality data generation in pathology, paving the way for next-generation general pathology models."
                    },
                    {
                        "title": "Unleashing the Power of Prompt-driven Nucleus Instance Segmentation",
                        "abstract": "Nucleus instance segmentation in histology images is crucial for a broad spectrum of clinical applications. Current dominant algorithms rely on regression of nuclear proxy maps. Distinguishing nucleus instances from the estimated maps requires carefully curated post-processing, which is error-prone and parameter-sensitive. Recently, the Segment Anything Model (SAM) has earned huge attention in medical image segmentation, owing to its impressive generalization ability and promptable property. Nevertheless, its potential on nucleus instance segmentation remains largely underexplored. In this paper, we present a novel prompt-driven framework that consists of a nucleus prompter and SAM for automatic nucleus instance segmentation. Specifically, the prompter learns to generate a unique point prompt for each nucleus while the SAM is fine-tuned to output the corresponding mask for the prompted nucleus. Furthermore, we propose the inclusion of adjacent nuclei as negative prompts to enhance the model's capability to identify overlapping nuclei. Without complicated post-processing, our proposed method sets a new state-of-the-art performance on three challenging benchmarks. Code is available at \\url{github.com/windygoo/PromptNucSeg}"
                    },
                    {
                        "title": "Masked conditional variational autoencoders for chromosome straightening",
                        "abstract": "Karyotyping is of importance for detecting chromosomal aberrations in human disease. However, chromosomes easily appear curved in microscopic images, which prevents cytogeneticists from analyzing chromosome types. To address this issue, we propose a framework for chromosome straightening, which comprises a preliminary processing algorithm and a generative model called masked conditional variational autoencoders (MC-VAE). The processing method utilizes patch rearrangement to address the difficulty in erasing low degrees of curvature, providing reasonable preliminary results for the MC-VAE. The MC-VAE further straightens the results by leveraging chromosome patches conditioned on their curvatures to learn the mapping between banding patterns and conditions. During model training, we apply a masking strategy with a high masking ratio to train the MC-VAE with eliminated redundancy. This yields a non-trivial reconstruction task, allowing the model to effectively preserve chromosome banding patterns and structure details in the reconstructed results. Extensive experiments on three public datasets with two stain styles show that our framework surpasses the performance of state-of-the-art methods in retaining banding patterns and structure details. Compared to using real-world bent chromosomes, the use of high-quality straightened chromosomes generated by our proposed method can improve the performance of various deep learning models for chromosome classification by a large margin. Such a straightening approach has the potential to be combined with other karyotyping systems to assist cytogeneticists in chromosome analysis."
                    },
                    {
                        "title": "Large-scale cervical precancerous screening via AI-assisted cytology whole slide image analysis",
                        "abstract": "Cervical Cancer continues to be the leading gynecological malignancy, posing a persistent threat to women's health on a global scale. Early screening via cytology Whole Slide Image (WSI) diagnosis is critical to prevent this Cancer progression and improve survival rate, but pathologist's single test suffers inevitable false negative due to the immense number of cells that need to be reviewed within a WSI. Though computer-aided automated diagnostic models can serve as strong complement for pathologists, their effectiveness is hampered by the paucity of extensive and detailed annotations, coupled with the limited interpretability and robustness. These factors significantly hinder their practical applicability and reliability in clinical settings. To tackle these challenges, we develop an AI approach, which is a Scalable Technology for Robust and Interpretable Diagnosis built on Extensive data (STRIDE) of cervical cytology. STRIDE addresses the bottleneck of limited annotations by integrating patient-level labels with a small portion of cell-level labels through an end-to-end training strategy, facilitating scalable learning across extensive datasets. To further improve the robustness to real-world domain shifts of cytology slide-making and imaging, STRIDE employs color adversarial samples training that mimic staining and imaging variations. Lastly, to achieve pathologist-level interpretability for the trustworthiness in clinical settings, STRIDE can generate explanatory textual descriptions that simulates pathologists' diagnostic processes by cell image feature and textual description alignment. Conducting extensive experiments and evaluations in 183 medical centers with a dataset of 341,889 WSIs and 0.1 billion cells from cervical cytology patients, STRIDE has demonstrated a remarkable superiority over previous state-of-the-art techniques."
                    }
                ]
            },
            "120f462d-bd37-42bf-81b4-06d17a430991": {
                "pk": "120f462d-bd37-42bf-81b4-06d17a430991",
                "name": "Chenglu Zhu",
                "collaborators": [
                    "Lin Yang",
                    "Sunyi Zheng",
                    "Honglin Li",
                    "Zhongyi Shui",
                    "Yunlong Zhang",
                    "Shichuan Zhang",
                    "Pingyi Chen",
                    "Yuxuan Sun",
                    "Jiatong Cai",
                    "Jingxiong Li"
                ],
                "domain": [
                    "Medical Image Analysis",
                    "Deep Learning",
                    "Weakly Supervised Learning",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Weakly Supervised Learning for cell recognition in immunohistochemical cytoplasm staining images",
                        "abstract": "Cell classification and counting in immunohistochemical cytoplasm staining images play a pivotal role in cancer diagnosis. Weakly supervised learning is a potential method to deal with labor-intensive labeling. However, the inconstant cell morphology and subtle differences between classes also bring challenges. To this end, we present a novel cell recognition framework based on multi-task learning, which utilizes two additional auxiliary tasks to guide robust representation learning of the main task. To deal with misclassification, the tissue prior learning branch is introduced to capture the spatial representation of tumor cells without additional tissue annotation. Moreover, dynamic masks and consistency learning are adopted to learn the invariance of cell scale and shape. We have evaluated our framework on immunohistochemical cytoplasm staining images, and the results demonstrate that our method outperforms recent cell recognition approaches. Besides, we have also done some ablation studies to show significant improvements after adding the auxiliary branches."
                    },
                    {
                        "title": "WSI-VQA: Interpreting Whole Slide Images by Generative Visual Question Answering",
                        "abstract": "Whole slide imaging is routinely adopted for carcinoma diagnosis and prognosis. Abundant experience is required for pathologists to achieve accurate and reliable diagnostic results of whole slide images (WSI). The huge size and heterogeneous features of WSIs make the workflow of pathological reading extremely time-consuming. In this paper, we propose a novel framework (WSI-VQA) to interpret WSIs by generative visual question answering. WSI-VQA shows universality by reframing various kinds of slide-level tasks in a question-answering pattern, in which pathologists can achieve immunohistochemical grading, survival prediction, and tumor subtyping following human-machine interaction. Furthermore, we establish a WSI-VQA dataset which contains 8672 slide-level question-answering pairs with 977 WSIs. Besides the ability to deal with different slide-level tasks, our generative model which is named Wsi2Text Transformer (W2T) outperforms existing discriminative models in medical correctness, which reveals the potential of our model to be applied in the clinical scenario. Additionally, we also visualize the co-attention mapping between word embeddings and WSIs as an intuitive explanation for diagnostic results. The dataset and related code are available at https://github.com/cpystan/WSI-VQA."
                    },
                    {
                        "title": "ChrSNet: Chromosome Straightening using Self-attention Guided Networks",
                        "abstract": "Karyotyping is an important procedure to assess the possible existence of chromosomal abnormalities. However, because of the non-rigid nature, chromosomes are usually heavily curved in microscopic images and such deformed shapes hinder the chromosome analysis for cytogeneticists. In this paper, we present a self-attention guided framework to erase the curvature of chromosomes. The proposed framework extracts spatial information and local textures to preserve banding patterns in a regression module. With complementary information from the bent chromosome, a refinement module is designed to further improve fine details. In addition, we propose two dedicated geometric constraints to maintain the length and restore the distortion of chromosomes. To train our framework, we create a synthetic dataset where curved chromosomes are generated from the real-world straight chromosomes by grid-deformation. Quantitative and qualitative experiments are conducted on synthetic and real-world data. Experimental results show that our proposed method can effectively straighten bent chromosomes while keeping banding details and length."
                    },
                    {
                        "title": "Unsupervised Dense Nuclei Detection and Segmentation with Prior Self-activation Map For Histology Images",
                        "abstract": "The success of supervised deep learning models in medical image segmentation relies on detailed annotations. However, labor-intensive manual labeling is costly and inefficient, especially in dense object segmentation. To this end, we propose a self-supervised learning based approach with a Prior Self-activation Module (PSM) that generates self-activation maps from the input images to avoid labeling costs and further produce pseudo masks for the downstream task. To be specific, we firstly train a neural network using self-supervised learning and utilize the gradient information in the shallow layers of the network to generate self-activation maps. Afterwards, a semantic-guided generator is then introduced as a pipeline to transform visual representations from PSM to pixel-level semantic pseudo masks for downstream tasks. Furthermore, a two-stage training module, consisting of a nuclei detection network and a nuclei segmentation network, is adopted to achieve the final segmentation. Experimental results show the effectiveness on two public pathological datasets. Compared with other fully-supervised and weakly-supervised methods, our method can achieve competitive performance without any manual annotations."
                    },
                    {
                        "title": "Generalizing Nucleus Recognition Model in Multi-source Images via Pruning",
                        "abstract": "Ki67 is a significant biomarker in the diagnosis and prognosis of cancer, whose index can be evaluated by quantifying its expression in Ki67 immunohistochemistry (IHC) stained images. However, quantitative analysis on multi-source Ki67 images is yet a challenging task in practice due to cross-domain distribution differences, which result from imaging variation, staining styles, and lesion types. Many recent studies have made some efforts on domain generalization (DG), whereas there are still some noteworthy limitations. Specifically in the case of Ki67 images, learning invariant representation is at the mercy of the insufficient number of domains and the cell categories mismatching in different domains. In this paper, we propose a novel method to improve DG by searching the domain-agnostic subnetwork in a domain merging scenario. Partial model parameters are iteratively pruned according to the domain gap, which is caused by the data converting from a single domain into merged domains during training. In addition, the model is optimized by fine-tuning on merged domains to eliminate the interference of class mismatching among various domains. Furthermore, an appropriate implementation is attained by applying the pruning method to different parts of the framework. Compared with known DG methods, our method yields excellent performance in multiclass nucleus recognition of Ki67 IHC images, especially in the lost category cases. Moreover, our competitive results are also evaluated on the public dataset over the state-of-the-art DG methods."
                    },
                    {
                        "title": "Exploring Unsupervised Cell Recognition with Prior Self-activation Maps",
                        "abstract": "The success of supervised deep learning models on cell recognition tasks relies on detailed annotations. Many previous works have managed to reduce the dependency on labels. However, considering the large number of cells contained in a patch, costly and inefficient labeling is still inevitable. To this end, we explored label-free methods for cell recognition. Prior self-activation maps (PSM) are proposed to generate pseudo masks as training targets. To be specific, an activation network is trained with self-supervised learning. The gradient information in the shallow layers of the network is aggregated to generate prior self-activation maps. Afterward, a semantic clustering module is then introduced as a pipeline to transform PSMs to pixel-level semantic pseudo masks for downstream tasks. We evaluated our method on two histological datasets: MoNuSeg (cell segmentation) and BCData (multi-class cell detection). Compared with other fully-supervised and weakly-supervised methods, our method can achieve competitive performance without any manual annotations. Our simple but effective framework can also achieve multi-class cell detection which can not be done by existing unsupervised methods. The results show the potential of PSMs that might inspire other research to deal with the hunger for labels in medical area."
                    },
                    {
                        "title": "Long-MIL: Scaling Long Contextual Multiple Instance Learning for Histopathology Whole Slide Image Analysis",
                        "abstract": "Histopathology image analysis is the golden standard of clinical diagnosis for Cancers. In doctors daily routine and computer-aided diagnosis, the Whole Slide Image (WSI) of histopathology tissue is used for analysis. Because of the extremely large scale of resolution, previous methods generally divide the WSI into a large number of patches, then aggregate all patches within a WSI by Multi-Instance Learning (MIL) to make the slide-level prediction when developing computer-aided diagnosis tools. However, most previous WSI-MIL models using global-attention without pairwise interaction and any positional information, or self-attention with absolute position embedding can not well handle shape varying large WSIs, e.g. testing WSIs after model deployment may be larger than training WSIs, since the model development set is always limited due to the difficulty of histopathology WSIs collection. To deal with the problem, in this paper, we propose to amend position embedding for shape varying long-contextual WSI by introducing Linear Bias into Attention, and adapt it from 1-d long sequence into 2-d long-contextual WSI which helps model extrapolate position embedding to unseen or under-fitted positions. We further utilize Flash-Attention module to tackle the computational complexity of Transformer, which also keep full self-attention performance compared to previous attention approximation work. Our method, Long-contextual MIL (Long-MIL) are evaluated on extensive experiments including 4 dataset including WSI classification and survival prediction tasks to validate the superiority on shape varying WSIs. The source code will be open-accessed soon."
                    },
                    {
                        "title": "WsiCaption: Multiple Instance Generation of Pathology Reports for Gigapixel Whole-Slide Images",
                        "abstract": "Whole slide images are the foundation of digital pathology for the diagnosis and treatment of carcinomas. Writing pathology reports is laborious and error-prone for inexperienced pathologists. To reduce the workload and improve clinical automation, we investigate how to generate pathology reports given whole slide images. On the data end, we curated the largest WSI-text dataset (PathText). In specific, we collected nearly 10000 high-quality WSI-text pairs for visual-language models by recognizing and cleaning pathology reports which narrate diagnostic slides in TCGA. On the model end, we propose the multiple instance generative model (MI-Gen) which can produce pathology reports for gigapixel WSIs. We benchmark our model on the largest subset of TCGA-PathoText. Experimental results show our model can generate pathology reports which contain multiple clinical clues and achieve competitive performance on certain slide-level tasks. We observe that simple semantic extraction from the pathology reports can achieve the best performance (0.838 of F1 score) on BRCA subtyping surpassing previous state-of-the-art approaches. Our collected dataset and related code are available."
                    },
                    {
                        "title": "Benchmarking the Robustness of Deep Neural Networks to Common Corruptions in Digital Pathology",
                        "abstract": "When designing a diagnostic model for a clinical application, it is crucial to guarantee the robustness of the model with respect to a wide range of image corruptions. Herein, an easy-to-use benchmark is established to evaluate how deep neural networks perform on corrupted pathology images. Specifically, corrupted images are generated by injecting nine types of common corruptions into validation images. Besides, two classification and one ranking metrics are designed to evaluate the prediction and confidence performance under corruption. Evaluated on two resulting benchmark datasets, we find that (1) a variety of deep neural network models suffer from a significant accuracy decrease (double the error on clean images) and the unreliable confidence estimation on corrupted images; (2) A low correlation between the validation and test errors while replacing the validation set with our benchmark can increase the correlation. Our codes are available on https://github.com/superjamessyx/robustness_benchmark."
                    },
                    {
                        "title": "End-to-end cell recognition by point annotation",
                        "abstract": "Reliable quantitative analysis of immunohistochemical staining images requires accurate and robust cell detection and classification. Recent weakly-supervised methods usually estimate probability density maps for cell recognition. However, in dense cell scenarios, their performance can be limited by pre- and post-processing as it is impossible to find a universal parameter setting. In this paper, we introduce an end-to-end framework that applies direct regression and classification for preset anchor points. Specifically, we propose a pyramidal feature aggregation strategy to combine low-level features and high-level semantics simultaneously, which provides accurate cell recognition for our purely point-based model. In addition, an optimized cost function is designed to adapt our multi-task learning framework by matching ground truth and predicted points. The experimental results demonstrate the superior accuracy and efficiency of the proposed method, which reveals the high potentiality in assisting pathologist assessments."
                    },
                    {
                        "title": "Task-specific Fine-tuning via Variational Information Bottleneck for Weakly-supervised Pathology Whole Slide Image Classification",
                        "abstract": "While Multiple Instance Learning (MIL) has shown promising results in digital Pathology Whole Slide Image (WSI) classification, such a paradigm still faces performance and generalization problems due to challenges in high computational costs on Gigapixel WSIs and limited sample size for model training. To deal with the computation problem, most MIL methods utilize a frozen pretrained model from ImageNet to obtain representations first. This process may lose essential information owing to the large domain gap and hinder the generalization of model due to the lack of image-level training-time augmentations. Though Self-supervised Learning (SSL) proposes viable representation learning schemes, the improvement of the downstream task still needs to be further explored in the conversion from the task-agnostic features of SSL to the task-specifics under the partial label supervised learning. To alleviate the dilemma of computation cost and performance, we propose an efficient WSI fine-tuning framework motivated by the Information Bottleneck theory. The theory enables the framework to find the minimal sufficient statistics of WSI, thus supporting us to fine-tune the backbone into a task-specific representation only depending on WSI-level weak labels. The WSI-MIL problem is further analyzed to theoretically deduce our fine-tuning method. Our framework is evaluated on five pathology WSI datasets on various WSI heads. The experimental results of our fine-tuned representations show significant improvements in both accuracy and generalization compared with previous works. Source code will be available at https://github.com/invoker-LL/WSI-finetuning."
                    },
                    {
                        "title": "PathAsst: A Generative Foundation AI Assistant Towards Artificial General Intelligence of Pathology",
                        "abstract": "As advances in large language models (LLMs) and multimodal techniques continue to mature, the development of general-purpose multimodal large language models (MLLMs) has surged, offering significant applications in interpreting natural images. However, the field of pathology has largely remained untapped, particularly in gathering high-quality data and designing comprehensive model frameworks. To bridge the gap in pathology MLLMs, we present PathAsst, a multimodal generative foundation AI assistant to revolutionize diagnostic and predictive analytics in pathology. The development of PathAsst involves three pivotal steps: data acquisition, CLIP model adaptation, and the training of PathAsst's multimodal generative capabilities. Firstly, we collect over 207K high-quality pathology image-text pairs from authoritative sources. Leveraging the advanced power of ChatGPT, we generate over 180K instruction-following samples. Furthermore, we devise additional instruction-following data specifically tailored for invoking eight pathology-specific sub-models we prepared, allowing the PathAsst to effectively collaborate with these models, enhancing its diagnostic ability. Secondly, by leveraging the collected data, we construct PathCLIP, a pathology-dedicated CLIP, to enhance PathAsst's capabilities in interpreting pathology images. Finally, we integrate PathCLIP with the Vicuna-13b and utilize pathology-specific instruction-tuning data to enhance the multimodal generation capacity of PathAsst and bolster its synergistic interactions with sub-models. The experimental results of PathAsst show the potential of harnessing AI-powered generative foundation model to improve pathology diagnosis and treatment processes."
                    },
                    {
                        "title": "Semi-supervised Cell Recognition under Point Supervision",
                        "abstract": "Cell recognition is a fundamental task in digital histopathology image analysis. Point-based cell recognition (PCR) methods normally require a vast number of annotations, which is extremely costly, time-consuming and labor-intensive. Semi-supervised learning (SSL) can provide a shortcut to make full use of cell information in gigapixel whole slide images without exhaustive labeling. However, research into semi-supervised point-based cell recognition (SSPCR) remains largely overlooked. Previous SSPCR works are all built on density map-based PCR models, which suffer from unsatisfactory accuracy, slow inference speed and high sensitivity to hyper-parameters. To address these issues, end-to-end PCR models are proposed recently. In this paper, we develop a SSPCR framework suitable for the end-to-end PCR models for the first time. Overall, we use the current models to generate pseudo labels for unlabeled images, which are in turn utilized to supervise the models training. Besides, we introduce a co-teaching strategy to overcome the confirmation bias problem that generally exists in self-training. A distribution alignment technique is also incorporated to produce high-quality, unbiased pseudo labels for unlabeled data. Experimental results on four histopathology datasets concerning different types of staining styles show the effectiveness and versatility of the proposed framework. Code is available at \\textcolor{magenta}{\\url{https://github.com/windygooo/SSPCR}"
                    },
                    {
                        "title": "Attention-Challenging Multiple Instance Learning for Whole Slide Image Classification",
                        "abstract": "In the application of Multiple Instance Learning (MIL) methods for Whole Slide Image (WSI) classification, attention mechanisms often focus on a subset of discriminative instances, which are closely linked to overfitting. To mitigate overfitting, we present Attention-Challenging MIL (ACMIL). ACMIL combines two techniques based on separate analyses for attention value concentration. Firstly, UMAP of instance features reveals various patterns among discriminative instances, with existing attention mechanisms capturing only some of them. To remedy this, we introduce Multiple Branch Attention (MBA) to capture more discriminative instances using multiple attention branches. Secondly, the examination of the cumulative value of Top-K attention scores indicates that a tiny number of instances dominate the majority of attention. In response, we present Stochastic Top-K Instance Masking (STKIM), which masks out a portion of instances with Top-K attention values and allocates their attention values to the remaining instances. The extensive experimental results on three WSI datasets with two pre-trained backbones reveal that our ACMIL outperforms state-of-the-art methods. Additionally, through heatmap visualization and UMAP visualization, this paper extensively illustrates ACMIL's effectiveness in suppressing attention value concentration and overcoming the overfitting challenge. The source code is available at \\url{https://github.com/dazhangyu123/ACMIL}."
                    },
                    {
                        "title": "Test-Time Training for Semantic Segmentation with Output Contrastive Loss",
                        "abstract": "Although deep learning-based segmentation models have achieved impressive performance on public benchmarks, generalizing well to unseen environments remains a major challenge. To improve the model's generalization ability to the new domain during evaluation, the test-time training (TTT) is a challenging paradigm that adapts the source-pretrained model in an online fashion. Early efforts on TTT mainly focus on the image classification task. Directly extending these methods to semantic segmentation easily experiences unstable adaption due to segmentation's inherent characteristics, such as extreme class imbalance and complex decision spaces. To stabilize the adaptation process, we introduce contrastive loss (CL), known for its capability to learn robust and generalized representations. Nevertheless, the traditional CL operates in the representation space and cannot directly enhance predictions. In this paper, we resolve this limitation by adapting the CL to the output space, employing a high temperature, and simplifying the formulation, resulting in a straightforward yet effective loss function called Output Contrastive Loss (OCL). Our comprehensive experiments validate the efficacy of our approach across diverse evaluation scenarios. Notably, our method excels even when applied to models initially pre-trained using domain adaptation methods on test domain data, showcasing its resilience and adaptability.\\footnote{Code and more information could be found at~ \\url{https://github.com/dazhangyu123/OCL}}"
                    },
                    {
                        "title": "AEM: Attention Entropy Maximization for Multiple Instance Learning based Whole Slide Image Classification",
                        "abstract": "Multiple Instance Learning (MIL) has demonstrated effectiveness in analyzing whole slide images (WSIs), yet it often encounters overfitting challenges in real-world applications, particularly in the form of attention over-concentration. While existing methods to alleviate this issue introduce complex modules or processing steps, such as multiple-stage training and teacher-student distillation, this paper proposes a simple yet effective regularization: Attention Entropy Maximization (AEM). Motivated by our investigation revealing a positive correlation between attention entropy and model performance, AEM incorporates a negative entropy loss for attention values into the standard MIL framework, penalizing overly concentrated attention and encouraging the model to consider a broader range of informative regions in WSIs, potentially improving its generalization capabilities. Compared to existing overfitting mitigation methods, our AEM approach offers advantages of simplicity, efficiency, and versatility. It requires no additional modules or processing steps, involves only one hyperparameter, and demonstrates compatibility with MIL frameworks and techniques. These advantages make AEM particularly attractive for practical applications. We evaluate AEM on three benchmark datasets, demonstrating consistent performance improvements over existing methods. Furthermore, AEM shows high versatility, integrating effectively with four feature extractors, two advanced MIL frameworks, three attention mechanisms, and Subsampling augmentation technique. The source code is available at \\url{https://github.com/dazhangyu123/AEM}."
                    },
                    {
                        "title": "PathGen-1.6M: 1.6 Million Pathology Image-text Pairs Generation through Multi-agent Collaboration",
                        "abstract": "Vision Language Models (VLMs) like CLIP have attracted substantial attention in pathology, serving as backbones for applications such as zero-shot image classification and Whole Slide Image (WSI) analysis. Additionally, they can function as vision encoders when combined with large language models (LLMs) to support broader capabilities. Current efforts to train pathology VLMs rely on pathology image-text pairs from platforms like PubMed, YouTube, and Twitter, which provide limited, unscalable data with generally suboptimal image quality. In this work, we leverage large-scale WSI datasets like TCGA to extract numerous high-quality image patches. We then train a large multimodal model to generate captions for these images, creating PathGen-1.6M, a dataset containing 1.6 million high-quality image-caption pairs. Our approach involves multiple agent models collaborating to extract representative WSI patches, generating and refining captions to obtain high-quality image-text pairs. Extensive experiments show that integrating these generated pairs with existing datasets to train a pathology-specific CLIP model, PathGen-CLIP, significantly enhances its ability to analyze pathological images, with substantial improvements across nine pathology-related zero-shot image classification tasks and three whole-slide image tasks. Furthermore, we construct 200K instruction-tuning data based on PathGen-1.6M and integrate PathGen-CLIP with the Vicuna LLM to create more powerful multimodal models through instruction tuning. Overall, we provide a scalable pathway for high-quality data generation in pathology, paving the way for next-generation general pathology models."
                    },
                    {
                        "title": "DPA-P2PNet: Deformable Proposal-aware P2PNet for Accurate Point-based Cell Detection",
                        "abstract": "Point-based cell detection (PCD), which pursues high-performance cell sensing under low-cost data annotation, has garnered increased attention in computational pathology community. Unlike mainstream PCD methods that rely on intermediate density map representations, the Point-to-Point network (P2PNet) has recently emerged as an end-to-end solution for PCD, demonstrating impressive cell detection accuracy and efficiency. Nevertheless, P2PNet is limited to decoding from a single-level feature map due to the scale-agnostic property of point proposals, which is insufficient to leverage multi-scale information. Moreover, the spatial distribution of pre-set point proposals is biased from that of cells, leading to inaccurate cell localization. To lift these limitations, we present DPA-P2PNet in this work. The proposed method directly extracts multi-scale features for decoding according to the coordinates of point proposals on hierarchical feature maps. On this basis, we further devise deformable point proposals to mitigate the positional bias between proposals and potential cells to promote cell localization. Inspired by practical pathological diagnosis that usually combines high-level tissue structure and low-level cell morphology for accurate cell classification, we propose a multi-field-of-view (mFoV) variant of DPA-P2PNet to accommodate additional large FoV images with tissue information as model input. Finally, we execute the first self-supervised pre-training on immunohistochemistry histopathology image data and evaluate the suitability of four representative self-supervised methods on the PCD task. Experimental results on three benchmarks and a large-scale and real-world interval dataset demonstrate the superiority of our proposed models over the state-of-the-art counterparts. Codes and pre-trained weights will be available."
                    }
                ]
            },
            "c68a90a6-02f5-413c-987e-0c9ea5a4f2d4": {
                "pk": "c68a90a6-02f5-413c-987e-0c9ea5a4f2d4",
                "name": "Lin Yang",
                "collaborators": [
                    "Baohua Sun",
                    "Hao Sha",
                    "Michael Lin"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Optical Character Recognition",
                    "Deep Learning",
                    "Low-Power Computing"
                ],
                "publications": [
                    {
                        "title": "Multi-modal Sentiment Analysis using Super Characters Method on Low-power CNN Accelerator Device",
                        "abstract": "Recent years NLP research has witnessed the record-breaking accuracy improvement by DNN models. However, power consumption is one of the practical concerns for deploying NLP systems. Most of the current state-of-the-art algorithms are implemented on GPUs, which is not power-efficient and the deployment cost is also very high. On the other hand, CNN Domain Specific Accelerator (CNN-DSA) has been in mass production providing low-power and low cost computation power. In this paper, we will implement the Super Characters method on the CNN-DSA. In addition, we modify the Super Characters method to utilize the multi-modal data, i.e. text plus tabular data in the CL-Aff sharedtask."
                    },
                    {
                        "title": "SuperOCR: A Conversion from Optical Character Recognition to Image Captioning",
                        "abstract": "Optical Character Recognition (OCR) has many real world applications. The existing methods normally detect where the characters are, and then recognize the character for each detected location. Thus the accuracy of characters recognition is impacted by the performance of characters detection. In this paper, we propose a method for recognizing characters without detecting the location of each character. This is done by converting the OCR task into an image captioning task. One advantage of the proposed method is that the labeled bounding boxes for the characters are not needed during training. The experimental results show the proposed method outperforms the existing methods on both the license plate recognition and the watermeter character recognition tasks. The proposed method is also deployed into a low-power (300mW) CNN accelerator chip connected to a Raspberry Pi 3 for on-device applications."
                    }
                ]
            }
        }
    },
    "2309.02214": {
        "paper_data": {
            "title": "Improving equilibrium propagation without weight symmetry through Jacobian homeostasis",
            "url": "http://arxiv.org/abs/2309.02214v2",
            "arxiv_id": "2309.02214",
            "authors": [
                "Axel Laborieux",
                "Friedemann Zenke"
            ],
            "abstract": "Equilibrium propagation (EP) is a compelling alternative to the backpropagation of error algorithm (BP) for computing gradients of neural networks on biological or analog neuromorphic substrates. Still, the algorithm requires weight symmetry and infinitesimal equilibrium perturbations, i.e., nudges, to estimate unbiased gradients efficiently. Both requirements are challenging to implement in physical systems. Yet, whether and how weight asymmetry affects its applicability is unknown because, in practice, it may be masked by biases introduced through the finite nudge. To address this question, we study generalized EP, which can be formulated without weight symmetry, and analytically isolate the two sources of bias. For complex-differentiable non-symmetric networks, we show that the finite nudge does not pose a problem, as exact derivatives can still be estimated via a Cauchy integral. In contrast, weight asymmetry introduces bias resulting in low task performance due to poor alignment of EP's neuronal error vectors compared to BP. To mitigate this issue, we present a new homeostatic objective that directly penalizes functional asymmetries of the Jacobian at the network's fixed point. This homeostatic objective dramatically improves the network's ability to solve complex tasks such as ImageNet 32x32. Our results lay the theoretical groundwork for studying and mitigating the adverse effects of imperfections of physical networks on learning algorithms that rely on the substrate's relaxation dynamics.",
            "introduction": "   1 Introduction  Virtually all state-of-the-art  artificial intelligence (AI) models are trained using the  backpropagation of error algorithm (BP) (Rumelhart et\u00a0al., 1986). In BP, neural activations are evaluated during the forward pass, and gradients are obtained through a corresponding backward pass. While BP can be implemented efficiently on digital hardware which simulates neural networks, it is less suitable for physical neural networks such as brains or neuromorphic substrates, which operate at lower energy costs. This limitation is mainly due to two technical requirements of the backward pass in BP. First, the backward pass is linear while leaving neuronal activity unaffected (Lillicrap et\u00a0al., 2020; Whittington & Bogacz, 2019). Second, back-propagation requires the transpose of connectivity matrices, often referred to as \u201cweight symmetry\u201d (Grossberg, 1987). For these reasons, many alternative learning algorithms that can run on dedicated neuromorphic hardware have been designed (Boybat et\u00a0al., 2018; Thakur et\u00a0al., 2018; Indiveri et\u00a0al., 2011; Schuman et\u00a0al., 2017; Frenkel & Indiveri, 2022; Yi et\u00a0al., 2023).   One such alternative learning algorithm is  equilibrium propagation (EP)(Scellier & Bengio, 2017), which can provably estimate gradients using only the network\u2019s own dynamics. Classic EP requires an  energy-based model (EBM), such as a differentiable Hopfield network (Hopfield, 1984), which relaxes to an equilibrium for any given input. EP computes gradients by comparing neuronal activity at the free equilibrium to the activity at a second equilibrium that is \u201cnudged\u201d toward a desired target output. Preliminary hardware realization dedicated to EP suggests that it could reduce the energy cost of AI training by four orders of magnitude (Yi et\u00a0al., 2023). Unlike BP in feed-forward networks, EP uses the same network to propagate feed-forward and feed-back activity, thereby dispensing with the need to linearly propagate error vectors. In particular, EP is a promising candidate for learning through neuronal oscillations (Baldi & Pineda, 1991; Laborieux & Zenke, 2022; Anisetti et\u00a0al., 2022; Delacour et\u00a0al., 2021), which is why we focus on EP in this work.   Nevertheless, EP requires weight symmetry and vanishing nudge for unbiased gradient estimates, which limits its potential for neuromorphics compared to other better performing BP alternatives (Ren et\u00a0al., 2023; Journ\u00e9 et\u00a0al., 2023; Payeur et\u00a0al., 2021; Greedy et\u00a0al., 2022; H\u00f8ier et\u00a0al., 2023). While Laborieux & Zenke (2022) showed that an oscillation-based extension of EP called  holomorphic EP (hEP) removes the bias by integrating oscillations of neuronal activity, this approach has only been demonstrated using symmetric weights. Although a generalization of EP to non-symmetric dynamical systems exists, the approach has only been demonstrated on simple tasks like MNIST (Scellier et\u00a0al., 2018; Ernoult et\u00a0al., 2020; Kohan et\u00a0al., 2018), while it fails to train on CIFAR-10 (Laborieux et\u00a0al., 2021). Previous work investigated the effect of non-symmetric feedback on training with BP (Lillicrap et\u00a0al., 2016; N\u00f8kland, 2016; Launay et\u00a0al., 2020; Clark et\u00a0al., 2021; Refinetti et\u00a0al., 2021) and approximate BP (Hinton, 2022; Dellaferrera & Kreiman, 2022; L\u00f6we et\u00a0al., 2019), but little attention has been given to effect of weight symmetry for EP. In this article we fill this gap. Overall, our main contributions are:   \u2022  We provide a comprehensive analysis of the individual sources of bias in the gradient estimate from weight asymmetry and finite-size nudge in generalized EP.    \u2022  An extension of hEP (Laborieux & Zenke, 2022)",
            "references": [
                {
                    "title": "A neuronal least-action principle for real-time learning in cortical circuits",
                    "abstract": "One of the most fundamental laws of physics is the principle of least action. Motivated by its predictive power, we introduce a neuronal least-action principle for cortical processing of sensory streams to produce appropriate behavioural outputs in real time. The principle postulates that the voltage dynamics of cortical pyramidal neurons prospectively minimizes the local somato-dendritic mismatch error within individual neurons. For output neurons, the principle implies minimizing an instantaneous behavioural error. For deep network neurons, it implies the prospective firing to overcome integration delays and correct for possible output errors right in time. The neuron-specific errors are extracted in the apical dendrites of pyramidal neurons through a cortical microcircuit that tries to explain away the feedback from the periphery, and correct the trajectory on the fly. Any motor output is in a moving equilibrium with the sensory input and the motor feedback during the ongoing sensory-motor transform. Online synaptic plasticity reduces the somato-dendritic mismatch error within each cortical neuron and performs gradient descent on the output cost at any moment in time. The neuronal least-action principle offers an axiomatic framework to derive local neuronal and synaptic laws for global real-time computation and learning in the brain."
                },
                {
                    "title": "Flexible Phase Dynamics for Bio-Plausible Contrastive Learning",
                    "abstract": "Many learning algorithms used as normative models in neuroscience or as candidate approaches for learning on neuromorphic chips learn by contrasting one set of network states with another. These Contrastive Learning (CL) algorithms are traditionally implemented with rigid, temporally non-local, and periodic learning dynamics that could limit the range of physical systems capable of harnessing CL. In this study, we build on recent work exploring how CL might be implemented by biological or neurmorphic systems and show that this form of learning can be made temporally local, and can still function even if many of the dynamical requirements of standard training procedures are relaxed. Thanks to a set of general theorems corroborated by numerical experiments across several CL models, our results provide theoretical foundations for the study and development of CL methods for biological and neuromorphic neural networks."
                },
                {
                    "title": "Dual Propagation: Accelerating Contrastive Hebbian Learning with Dyadic Neurons",
                    "abstract": "Activity difference based learning algorithms-such as contrastive Hebbian learning and equilibrium propagation-have been proposed as biologically plausible alternatives to error back-propagation. However, on traditional digital chips these algorithms suffer from having to solve a costly inference problem twice, making these approaches more than two orders of magnitude slower than back-propagation. In the analog realm equilibrium propagation may be promising for fast and energy efficient learning, but states still need to be inferred and stored twice. Inspired by lifted neural networks and compartmental neuron models we propose a simple energy based compartmental neuron model, termed dual propagation, in which each neuron is a dyad with two intrinsic states. At inference time these intrinsic states encode the error/activity duality through their difference and their mean respectively. The advantage of this method is that only a single inference phase is needed and that inference can be solved in layerwise closed-form. Experimentally we show on common computer vision datasets, including Imagenet32x32, that dual propagation performs equivalently to back-propagation both in terms of accuracy and runtime."
                },
                {
                    "title": "The Forward-Forward Algorithm: Some Preliminary Investigations",
                    "abstract": "The aim of this paper is to introduce a new learning procedure for neural networks and to demonstrate that it works well enough on a few small problems to be worth further investigation. The Forward-Forward algorithm replaces the forward and backward passes of backpropagation by two forward passes, one with positive (i.e. real) data and the other with negative data which could be generated by the network itself. Each layer has its own objective function which is simply to have high goodness for positive data and low goodness for negative data. The sum of the squared activities in a layer can be used as the goodness but there are many other possibilities, including minus the sum of the squared activities. If the positive and negative passes could be separated in time, the negative passes could be done offline, which would make the learning much simpler in the positive pass and allow video to be pipelined through the network without ever storing activities or stopping to propagate derivatives."
                },
                {
                    "title": "Scaling Forward Gradient With Local Losses",
                    "abstract": "Forward gradient learning computes a noisy directional gradient and is a biologically plausible alternative to backprop for learning deep neural networks. However, the standard forward gradient algorithm, when applied naively, suffers from high variance when the number of parameters to be learned is large. In this paper, we propose a series of architectural and algorithmic modifications that together make forward gradient learning practical for standard deep learning benchmark tasks. We show that it is possible to substantially reduce the variance of the forward gradient estimator by applying perturbations to activations rather than weights. We further improve the scalability of forward gradient by introducing a large number of local greedy loss functions, each of which involves only a small number of learnable parameters, and a new MLPMixer-inspired architecture, LocalMixer, that is more suitable for local learning. Our approach matches backprop on MNIST and CIFAR-10 and significantly outperforms previously proposed backprop-free algorithms on ImageNet."
                },
                {
                    "title": "Hebbian Deep Learning Without Feedback",
                    "abstract": "Recent approximations to backpropagation (BP) have mitigated many of BP's computational inefficiencies and incompatibilities with biology, but important limitations still remain. Moreover, the approximations significantly decrease accuracy in benchmarks, suggesting that an entirely different approach may be more fruitful. Here, grounded on recent theory for Hebbian learning in soft winner-take-all networks, we present multilayer SoftHebb, i.e. an algorithm that trains deep neural networks, without any feedback, target, or error signals. As a result, it achieves efficiency by avoiding weight transport, non-local plasticity, time-locking of layer updates, iterative equilibria, and (self-) supervisory or other feedback signals -- which were necessary in other approaches. Its increased efficiency and biological compatibility do not trade off accuracy compared to state-of-the-art bio-plausible learning, but rather improve it. With up to five hidden layers and an added linear classifier, accuracies on MNIST, CIFAR-10, STL-10, and ImageNet, respectively reach 99.4%, 80.3%, 76.2%, and 27.3%. In conclusion, SoftHebb shows with a radically different approach from BP that Deep Learning over few layers may be plausible in the brain and increases the accuracy of bio-plausible machine learning. Code is available at https://github.com/NeuromorphicComputing/SoftHebb."
                },
                {
                    "title": "Holomorphic Equilibrium Propagation Computes Exact Gradients Through Finite Size Oscillations",
                    "abstract": "Equilibrium propagation (EP) is an alternative to backpropagation (BP) that allows the training of deep neural networks with local learning rules. It thus provides a compelling framework for training neuromorphic systems and understanding learning in neurobiology. However, EP requires infinitesimal teaching signals, thereby limiting its applicability in noisy physical systems. Moreover, the algorithm requires separate temporal phases and has not been applied to large-scale problems. Here we address these issues by extending EP to holomorphic networks. We show analytically that this extension naturally leads to exact gradients even for finite-amplitude teaching signals. Importantly, the gradient can be computed as the first Fourier coefficient from finite neuronal activity oscillations in continuous time without requiring separate phases. Further, we demonstrate in numerical simulations that our approach permits robust estimation of gradients in the presence of noise and that deeper models benefit from the finite teaching signals. Finally, we establish the first benchmark for EP on the ImageNet 32x32 dataset and show that it matches the performance of an equivalent network trained with BP. Our work provides analytical insights that enable scaling EP to large-scale problems and establishes a formal framework for how oscillations could support learning in biological and neuromorphic systems."
                },
                {
                    "title": "Frequency Propagation: Multimechanism Learning in Nonlinear Physical Networks",
                    "abstract": "Abstract We introduce frequency propagation, a learning algorithm for nonlinear physical networks. In a resistive electrical circuit with variable resistors, an activation current is applied at a set of input nodes at one frequency and an error current is applied at a set of output nodes at another frequency. The voltage response of the circuit to these boundary currents is the superposition of an activation signal and an error signal whose coefficients can be read in different frequencies of the frequency domain. Each conductance is updated proportionally to the product of the two coefficients. The learning rule is local and proved to perform gradient descent on a loss function. We argue that frequency propagation is an instance of a multimechanism learning strategy for physical networks, be it resistive, elastic, or flow networks. Multimechanism learning strategies incorporate at least two physical quantities, potentially governed by independent physical mechanisms, to act as activation and error signals in the training process. Locally available information about these two signals is then used to update the trainable parameters to perform gradient descent. We demonstrate how earlier work implementing learning via chemical signaling in flow networks (Anisetti, Scellier, et al., 2023) also falls under the rubric of multimechanism learning."
                },
                {
                    "title": "The least-control principle for local learning at equilibrium",
                    "abstract": "Equilibrium systems are a powerful way to express neural computations. As special cases, they include models of great current interest in both neuroscience and machine learning, such as deep neural networks, equilibrium recurrent neural networks, deep equilibrium models, or meta-learning. Here, we present a new principle for learning such systems with a temporally- and spatially-local rule. Our principle casts learning as a least-control problem, where we first introduce an optimal controller to lead the system towards a solution state, and then define learning as reducing the amount of control needed to reach such a state. We show that incorporating learning signals within a dynamics as an optimal control enables transmitting activity-dependent credit assignment information, avoids storing intermediate states in memory, and does not rely on infinitesimal learning signals. In practice, our principle leads to strong performance matching that of leading gradient-based learning methods when applied to an array of problems involving recurrent neural networks and meta-learning. Our results shed light on how the brain might learn and offer new ways of approaching a broad class of machine learning problems."
                },
                {
                    "title": "Single-phase deep learning in cortico-cortical networks",
                    "abstract": "The error-backpropagation (backprop) algorithm remains the most common solution to the credit assignment problem in artificial neural networks. In neuroscience, it is unclear whether the brain could adopt a similar strategy to correctly modify its synapses. Recent models have attempted to bridge this gap while being consistent with a range of experimental observations. However, these models are either unable to effectively backpropagate error signals across multiple layers or require a multi-phase learning process, neither of which are reminiscent of learning in the brain. Here, we introduce a new model, Bursting Cortico-Cortical Networks (BurstCCN), which solves these issues by integrating known properties of cortical networks namely bursting activity, short-term plasticity (STP) and dendrite-targeting interneurons. BurstCCN relies on burst multiplexing via connection-type-specific STP to propagate backprop-like error signals within deep cortical networks. These error signals are encoded at distal dendrites and induce burst-dependent plasticity as a result of excitatory-inhibitory top-down inputs. First, we demonstrate that our model can effectively backpropagate errors through multiple layers using a single-phase learning process. Next, we show both empirically and analytically that learning in our model approximates backprop-derived gradients. Finally, we demonstrate that our model is capable of learning complex image classification tasks (MNIST and CIFAR-10). Overall, our results suggest that cortical features across sub-cellular, cellular, microcircuit and systems levels jointly underlie single-phase efficient deep learning in the brain."
                },
                {
                    "title": "Beyond Backpropagation: Bilevel Optimization Through Implicit Differentiation and Equilibrium Propagation",
                    "abstract": "Abstract This review examines gradient-based techniques to solve bilevel optimization problems. Bilevel optimization extends the loss minimization framework underlying statistical learning to systems that are implicitly defined through a quantity they minimize. This characterization can be applied to neural networks, optimizers, algorithmic solvers, and even physical systems and allows for greater modeling flexibility compared to the usual explicit definition of such systems. We focus on solving learning problems of this kind through gradient descent, leveraging the toolbox of implicit differentiation and, for the first time applied to this setting, the equilibrium propagation theorem. We present the mathematical foundations behind such methods, introduce the gradient estimation algorithms in detail, and compare the competitive advantages of the different approaches."
                },
                {
                    "title": "Minimizing Control for Credit Assignment with Strong Feedback",
                    "abstract": "The success of deep learning ignited interest in whether the brain learns hierarchical representations using gradient-based learning. However, current biologically plausible methods for gradient-based credit assignment in deep neural networks need infinitesimally small feedback signals, which is problematic in biologically realistic noisy environments and at odds with experimental evidence in neuroscience showing that top-down feedback can significantly influence neural activity. Building upon deep feedback control (DFC), a recently proposed credit assignment method, we combine strong feedback influences on neural activity with gradient-based learning and show that this naturally leads to a novel view on neural network optimization. Instead of gradually changing the network weights towards configurations with low output loss, weight updates gradually minimize the amount of feedback required from a controller that drives the network to the supervised output label. Moreover, we show that the use of strong feedback in DFC allows learning forward and feedback connections simultaneously, using learning rules fully local in space and time. We complement our theoretical results with experiments on standard computer-vision benchmarks, showing competitive performance to backpropagation as well as robustness to noise. Overall, our work presents a fundamentally novel view of learning as control minimization, while sidestepping biologically unrealistic assumptions."
                },
                {
                    "title": "Signal Propagation: The Framework for Learning and Inference in a Forward Pass",
                    "abstract": "We propose a new learning framework, signal propagation (sigprop), for propagating a learning signal and updating neural network parameters via a forward pass, as an alternative to backpropagation (BP). In sigprop, there is only the forward path for inference and learning. So, there are no structural or computational constraints necessary for learning to take place, beyond the inference model itself, such as feedback connectivity, weight transport, or a backward pass, which exist under BP-based approaches. That is, sigprop enables global supervised learning with only a forward path. This is ideal for parallel training of layers or modules. In biology, this explains how neurons without feedback connections can still receive a global learning signal. In hardware, this provides an approach for global supervised learning without backward connectivity. Sigprop by construction has compatibility with models of learning in the brain and in hardware than BP, including alternative approaches relaxing learning constraints. We also demonstrate that sigprop is more efficient in time and memory than they are. To further explain the behavior of sigprop, we provide evidence that sigprop provides useful learning signals in context to BP. To further support relevance to biological and hardware learning, we use sigprop to train continuous time neural networks with the Hebbian updates and train spiking neural networks (SNNs) with only the voltage or with biologically and hardware-compatible surrogate functions."
                },
                {
                    "title": "Deep Learning in Spiking Phasor Neural Networks",
                    "abstract": "Spiking Neural Networks (SNNs) have attracted the attention of the deep learning community for use in low-latency, low-power neuromorphic hardware, as well as models for understanding neuroscience. In this paper, we introduce Spiking Phasor Neural Networks (SPNNs). SPNNs are based on complex-valued Deep Neural Networks (DNNs), representing phases by spike times. Our model computes robustly employing a spike timing code and gradients can be formed using the complex domain. We train SPNNs on CIFAR-10, and demonstrate that the performance exceeds that of other timing coded SNNs, approaching results with comparable real-valued DNNs."
                },
                {
                    "title": "ReckOn: A 28nm Sub-mm2 Task-Agnostic Spiking Recurrent Neural Network Processor Enabling On-Chip Learning over Second-Long Timescales",
                    "abstract": "The robustness of autonomous inference-only devices deployed in the real world is limited by data distribution changes induced by different users, environments, and task requirements. This challenge calls for the development of edge devices with an always-on adaptation to their target ecosystems. However, the memory requirements of conventional neural-network training algorithms scale with the temporal depth of the data being processed, which is not compatible with the constrained power and area budgets at the edge. For this reason, previous works demonstrating end-to-end on-chip learning without external memory were restricted to the processing of static data such as images [1]\u2013[4], or to instantaneous decisions involving no memory of the past, e.g. obstacle avoidance in mobile robots [5]. The ability to learn short-to-long-term temporal dependencies on-chip is a missing enabler for robust autonomous edge devices in applications such as gesture recognition, speech processing, and cognitive robotics."
                },
                {
                    "title": "Towards Scaling Difference Target Propagation by Learning Backprop Targets",
                    "abstract": "The development of biologically-plausible learning algorithms is important for understanding learning in the brain, but most of them fail to scale-up to real-world tasks, limiting their potential as explanations for learning by real brains. As such, it is important to explore learning algorithms that come with strong theoretical guarantees and can match the performance of backpropagation (BP) on complex tasks. One such algorithm is Difference Target Propagation (DTP), a biologically-plausible learning algorithm whose close relation with Gauss-Newton (GN) optimization has been recently established. However, the conditions under which this connection rigorously holds preclude layer-wise training of the feedback pathway synaptic weights (which is more biologically plausible). Moreover, good alignment between DTP weight updates and loss gradients is only loosely guaranteed and under very specific conditions for the architecture being trained. In this paper, we propose a novel feedback weight training scheme that ensures both that DTP approximates BP and that layer-wise feedback weight training can be restored without sacrificing any theoretical guarantees. Our theory is corroborated by experimental results and we report the best performance ever achieved by DTP on CIFAR-10 and ImageNet 32$\\times$32"
                },
                {
                    "title": "Error-driven Input Modulation: Solving the Credit Assignment Problem without a Backward Pass",
                    "abstract": "Supervised learning in artificial neural networks typically relies on backpropagation, where the weights are updated based on the error-function gradients and sequentially propagated from the output layer to the input layer. Although this approach has proven effective in a wide domain of applications, it lacks biological plausibility in many regards, including the weight symmetry problem, the dependence of learning on non-local signals, the freezing of neural activity during error propagation, and the update locking problem. Alternative training schemes have been introduced, including sign symmetry, feedback alignment, and direct feedback alignment, but they invariably rely on a backward pass that hinders the possibility of solving all the issues simultaneously. Here, we propose to replace the backward pass with a second forward pass in which the input signal is modulated based on the error of the network. We show that this novel learning rule comprehensively addresses all the above-mentioned issues and can be applied to both fully connected and convolutional models. We test this learning rule on MNIST, CIFAR-10, and CIFAR-100. These results help incorporate biological principles into machine learning."
                },
                {
                    "title": "Synaptic balancing: A biologically plausible local learning rule that provably increases neural network noise robustness without sacrificing task performance",
                    "abstract": "We introduce a novel, biologically plausible local learning rule that provably increases the robustness of neural dynamics to noise in nonlinear recurrent neural networks with homogeneous nonlinearities. Our learning rule achieves higher noise robustness without sacrificing performance on the task and without requiring any knowledge of the particular task. The plasticity dynamics\u2014an integrable dynamical system operating on the weights of the network\u2014maintains a multiplicity of conserved quantities, most notably the network\u2019s entire temporal map of input to output trajectories. The outcome of our learning rule is a synaptic balancing between the incoming and outgoing synapses of every neuron. This synaptic balancing rule is consistent with many known aspects of experimentally observed heterosynaptic plasticity, and moreover makes new experimentally testable predictions relating plasticity at the incoming and outgoing synapses of individual neurons. Overall, this work provides a novel, practical local learning rule that exactly preserves overall network function and, in doing so, provides new conceptual bridges between the disparate worlds of the neurobiology of heterosynaptic plasticity, the engineering of regularized noise-robust networks, and the mathematics of integrable Lax dynamical systems."
                },
                {
                    "title": "SoftHebb: Bayesian inference in unsupervised Hebbian soft winner-take-all networks",
                    "abstract": "Hebbian plasticity in winner-take-all (WTA) networks is highly attractive for neuromorphic on-chip learning, owing to its efficient, local, unsupervised, and on-line nature. Moreover, its biological plausibility may help overcome important limitations of artificial algorithms, such as their susceptibility to adversarial attacks, and their high demands for training-example quantity and repetition. However, Hebbian WTA learning has found little use in machine learning, likely because it has been missing an optimization theory compatible with deep learning (DL). Here we show rigorously that WTA networks constructed by standard DL elements, combined with a Hebbian-like plasticity that we derive, maintain a Bayesian generative model of the data. Importantly, without any supervision, our algorithm, SoftHebb, minimizes cross-entropy, i.e. a common loss function in supervised DL. We show this theoretically and in practice. The key is a \u2018soft\u2019 WTA where there is no absolute \u2018hard\u2019 winner neuron. Strikingly, in shallow-network comparisons with backpropagation, SoftHebb shows advantages beyond its Hebbian efficiency. Namely, it converges in fewer iterations, and is significantly more robust to noise and adversarial attacks. Notably, attacks that maximally confuse SoftHebb are also confusing to the human eye, potentially linking human perceptual robustness, with Hebbian WTA circuits of cortex. Finally, SoftHebb can generate synthetic objects as interpolations of real object classes. All in all, Hebbian efficiency, theoretical underpinning, cross-entropy-minimization, and surprising empirical advantages, suggest that SoftHebb may inspire highly neuromorphic and radically different, but practical and advantageous learning algorithms and hardware accelerators."
                },
                {
                    "title": "Oscillatory Neural Networks for Edge AI Computing",
                    "abstract": "In this paper, we showcase the innovative concept of implementing Oscillatory Neural Networks (ONNs) for neuromorphic computing with beyond-CMOS devices based on vanadium dioxide to mimic neurons and resistors to emulate synapses. We explore ONN technology potentials from device to analog circuit-level simulations. We report that ONN behaves like an associative memory and can implement energy-based models such as Hopfield Neural Networks on edge devices. Finally, as a proof of concept, a reconfigurable digital ONN is implemented on FPGA for pattern recognition tasks."
                },
                {
                    "title": "Stabilizing Equilibrium Models by Jacobian Regularization",
                    "abstract": "Deep equilibrium networks (DEQs) are a new class of models that eschews traditional depth in favor of finding the fixed point of a single nonlinear layer. These models have been shown to achieve performance competitive with the state-of-the-art deep networks while using significantly less memory. Yet they are also slower, brittle to architectural choices, and introduce potential instability to the model. In this paper, we propose a regularization scheme for DEQ models that explicitly regularizes the Jacobian of the fixed-point update equations to stabilize the learning of equilibrium models. We show that this regularization adds only minimal computational cost, significantly stabilizes the fixed-point convergence in both forward and backward passes, and scales well to high-dimensional, realistic domains (e.g., WikiText-103 language modeling and ImageNet classification). Using this method, we demonstrate, for the first time, an implicit-depth model that runs with approximately the same speed and level of performance as popular conventional deep networks such as ResNet-101, while still maintaining the constant memory footprint and architectural simplicity of DEQs. Code is available at https://github.com/locuslab/deq ."
                },
                {
                    "title": "Credit Assignment in Neural Networks through Deep Feedback Control",
                    "abstract": "The success of deep learning sparked interest in whether the brain learns by using similar techniques for assigning credit to each synaptic weight for its contribution to the network output. However, the majority of current attempts at biologically-plausible learning methods are either non-local in time, require highly specific connectivity motives, or have no clear link to any known mathematical optimization method. Here, we introduce Deep Feedback Control (DFC), a new learning method that uses a feedback controller to drive a deep neural network to match a desired output target and whose control signal can be used for credit assignment. The resulting learning rule is fully local in space and time and approximates Gauss-Newton optimization for a wide range of feedback connectivity patterns. To further underline its biological plausibility, we relate DFC to a multi-compartment model of cortical pyramidal neurons with a local voltage-dependent synaptic plasticity rule, consistent with recent theories of dendritic processing. By combining dynamical system theory with mathematical optimization theory, we provide a strong theoretical foundation for DFC that we corroborate with detailed results on toy experiments and standard computer-vision benchmarks."
                },
                {
                    "title": "Credit Assignment Through Broadcasting a Global Error Vector",
                    "abstract": "Backpropagation (BP) uses detailed, unit-specific feedback to train deep neural networks (DNNs) with remarkable success. That biological neural circuits appear to perform credit assignment, but cannot implement BP, implies the existence of other powerful learning algorithms. Here, we explore the extent to which a globally broadcast learning signal, coupled with local weight updates, enables training of DNNs. We present both a learning rule, called global error-vector broadcasting (GEVB), and a class of DNNs, called vectorized nonnegative networks (VNNs), in which this learning rule operates. VNNs have vector-valued units and nonnegative weights past the first layer. The GEVB learning rule generalizes three-factor Hebbian learning, updating each weight by an amount proportional to the inner product of the presynaptic activation and a globally broadcast error vector when the postsynaptic unit is active. We prove that these weight updates are matched in sign to the gradient, enabling accurate credit assignment. Moreover, at initialization, these updates are exactly proportional to the gradient in the limit of infinite network width. GEVB matches the performance of BP in VNNs, and in some cases outperforms direct feedback alignment (DFA) applied in conventional networks. Unlike DFA, GEVB successfully trains convolutional layers. Altogether, our theoretical and empirical results point to a surprisingly powerful role for a global learning signal in training DNNs."
                },
                {
                    "title": "A contrastive rule for meta-learning",
                    "abstract": "Humans and other animals are capable of improving their learning performance as they solve related tasks from a given problem domain, to the point of being able to learn from extremely limited data. While synaptic plasticity is generically thought to underlie learning in the brain, the precise neural and synaptic mechanisms by which learning processes improve through experience are not well understood. Here, we present a general-purpose, biologically-plausible meta-learning rule which estimates gradients with respect to the parameters of an underlying learning algorithm by simply running it twice. Our rule may be understood as a generalization of contrastive Hebbian learning to meta-learning and notably, it neither requires computing second derivatives nor going backwards in time, two characteristic features of previous gradient-based methods that are hard to conceive in physical neural circuits. We demonstrate the generality of our rule by applying it to two distinct models: a complex synapse with internal states which consolidate task-shared information, and a dual-system architecture in which a primary network is rapidly modulated by another one to learn the specifics of each task. For both models, our meta-learning rule matches or outperforms reference algorithms on a wide range of benchmark problems, while only using information presumed to be locally available at neurons and synapses. We corroborate these findings with a theoretical analysis of the gradient estimation error incurred by our rule."
                },
                {
                    "title": "Learning Without Feedback: Fixed Random Learning Signals Allow for Feedforward Training of Deep Neural Networks",
                    "abstract": "While the backpropagation of error algorithm enables deep neural network training, it implies (i) bidirectional synaptic weight transport and (ii) update locking until the forward and backward passes are completed. Not only do these constraints preclude biological plausibility, but they also hinder the development of low-cost adaptive smart sensors at the edge, as they severely constrain memory accesses and entail buffering overhead. In this work, we show that the one-hot-encoded labels provided in supervised classification problems, denoted as targets, can be viewed as a proxy for the error sign. Therefore, their fixed random projections enable a layerwise feedforward training of the hidden layers, thus solving the weight transport and update locking problems while relaxing the computational and memory requirements. Based on these observations, we propose the direct random target projection (DRTP) algorithm and demonstrate that it provides a tradeoff between accuracy and computational cost that is suitable for adaptive edge computing devices."
                },
                {
                    "title": "Refining activation downsampling with SoftPool",
                    "abstract": "Convolutional Neural Networks (CNNs) use pooling to decrease the size of activation maps. This process is crucial to increase the receptive fields and to reduce computational requirements of subsequent convolutions. An important feature of the pooling operation is the minimization of information loss, with respect to the initial activation maps, without a significant impact on the computation and memory overhead. To meet these requirements, we propose SoftPool: a fast and efficient method for exponentially weighted activation downsampling. Through experiments across a range of architectures and pooling methods, we demonstrate that SoftPool can retain more information in the reduced activation maps. This refined downsampling leads to improvements in a CNN\u2019s classification accuracy. Experiments with pooling layer substitutions on ImageNet1K show an increase in accuracy over both original architectures and other pooling methods. We also test SoftPool on video datasets for action recognition. Again, through the direct replacement of pooling layers, we observe consistent performance improvements while computational loads and memory requirements remain limited1."
                },
                {
                    "title": "Align, then memorise: the dynamics of learning with feedback alignment",
                    "abstract": "Direct feedback alignment (DFA) is emerging as an efficient and biologically plausible alternative to backpropagation for training deep neural networks. Despite relying on random feedback weights for the backward pass, DFA successfully trains state-of-the-art models such as transformers. On the other hand, it notoriously fails to train convolutional networks. An understanding of the inner workings of DFA to explain these diverging results remains elusive. Here, we propose a theory of feedback alignment algorithms. We first show that learning in shallow networks proceeds in two steps: an alignment phase, where the model adapts its weights to align the approximate gradient with the true gradient of the loss function, is followed by a memorisation phase, where the model focuses on fitting the data. This two-step process has a degeneracy breaking effect: out of all the low-loss solutions in the landscape, a network trained with DFA naturally converges to the solution which maximises gradient alignment. We also identify a key quantity underlying alignment in deep linear networks: the conditioning of the alignment matrices. The latter enables a detailed understanding of the impact of data structure on alignment, and suggests a simple explanation for the well-known failure of DFA to train convolutional neural networks. Numerical experiments on MNIST and CIFAR10 clearly demonstrate degeneracy breaking in deep non-linear networks and show that the align-then-memorize process occurs sequentially from the bottom layers of the network to the top."
                },
                {
                    "title": "A biologically plausible neural network for Slow Feature Analysis",
                    "abstract": "Learning latent features from time series data is an important problem in both machine learning and brain function. One approach, called Slow Feature Analysis (SFA), leverages the slowness of many salient features relative to the rapidly varying input signals. Furthermore, when trained on naturalistic stimuli, SFA reproduces interesting properties of cells in the primary visual cortex and hippocampus, suggesting that the brain uses temporal slowness as a computational principle for learning latent features. However, despite the potential relevance of SFA for modeling brain function, there is currently no SFA algorithm with a biologically plausible neural network implementation, by which we mean an algorithm operates in the online setting and can be mapped onto a neural network with local synaptic updates. In this work, starting from an SFA objective, we derive an SFA algorithm, called Bio-SFA, with a biologically plausible neural network implementation. We validate Bio-SFA on naturalistic stimuli."
                },
                {
                    "title": "Hopfield Networks is All You Need",
                    "abstract": "We show that the transformer attention mechanism is the update rule of a modern Hopfield network with continuous states. This new Hopfield network can store exponentially (with the dimension) many patterns, converges with one update, and has exponentially small retrieval errors. The number of stored patterns is traded off against convergence speed and retrieval error. The new Hopfield network has three types of energy minima (fixed points of the update): (1) global fixed point averaging over all patterns, (2) metastable states averaging over a subset of patterns, and (3) fixed points which store a single pattern. Transformer and BERT models operate in their first layers preferably in the global averaging regime, while they operate in higher layers in metastable states. The gradient in transformers is maximal for metastable states, is uniformly distributed for global averaging, and vanishes for a fixed point near a stored pattern. Using the Hopfield network interpretation, we analyzed learning of transformer and BERT models. Learning starts with attention heads that average and then most of them switch to metastable states. However, the majority of heads in the first layers still averages and can be replaced by averaging, e.g. our proposed Gaussian weighting. In contrast, heads in the last layers steadily learn and seem to use metastable states to collect information created in lower layers. These heads seem to be a promising target for improving transformers. Neural networks with Hopfield networks outperform other methods on immune repertoire classification, where the Hopfield net stores several hundreds of thousands of patterns. We provide a new PyTorch layer called \"Hopfield\", which allows to equip deep learning architectures with modern Hopfield networks as a new powerful concept comprising pooling, memory, and attention. GitHub: this https URL"
                },
                {
                    "title": "A Theoretical Framework for Target Propagation",
                    "abstract": "The success of deep learning, a brain-inspired form of AI, has sparked interest in understanding how the brain could similarly learn across multiple layers of neurons. However, the majority of biologically-plausible learning algorithms have not yet reached the performance of backpropagation (BP), nor are they built on strong theoretical foundations. Here, we analyze target propagation (TP), a popular but not yet fully understood alternative to BP, from the standpoint of mathematical optimization. Our theory shows that TP is closely related to Gauss-Newton optimization and thus substantially differs from BP. Furthermore, our analysis reveals a fundamental limitation of difference target propagation (DTP), a well-known variant of TP, in the realistic scenario of non-invertible neural networks. We provide a first solution to this problem through a novel reconstruction loss that improves feedback weight training, while simultaneously introducing architectural flexibility by allowing for direct feedback connections from the output to each hidden layer. Our theory is corroborated by experimental results that show significant improvements in performance and in the alignment of forward weight updates with loss gradients, compared to DTP."
                },
                {
                    "title": "Direct Feedback Alignment Scales to Modern Deep Learning Tasks and Architectures",
                    "abstract": "Despite being the workhorse of deep learning, the backpropagation algorithm is no panacea. It enforces sequential layer updates, thus preventing efficient parallelization of the training process. Furthermore, its biological plausibility is being challenged. Alternative schemes have been devised; yet, under the constraint of synaptic asymmetry, none have scaled to modern deep learning tasks and architectures. Here, we challenge this perspective, and study the applicability of Direct Feedback Alignment to neural view synthesis, recommender systems, geometric learning, and natural language processing. In contrast with previous studies limited to computer vision tasks, our findings show that it successfully trains a large range of state-of-the-art deep learning architectures, with performance close to fine-tuned backpropagation. At variance with common beliefs, our work supports that challenging tasks can be tackled in the absence of weight transport."
                },
                {
                    "title": "Scaling Equilibrium Propagation to Deep ConvNets by Drastically Reducing Its Gradient Estimator Bias",
                    "abstract": "Equilibrium Propagation is a biologically-inspired algorithm that trains convergent recurrent neural networks with a local learning rule. This approach constitutes a major lead to allow learning-capable neuromophic systems and comes with strong theoretical guarantees. Equilibrium propagation operates in two phases, during which the network is let to evolve freely and then \u201cnudged\u201d toward a target; the weights of the network are then updated based solely on the states of the neurons that they connect. The weight updates of Equilibrium Propagation have been shown mathematically to approach those provided by Backpropagation Through Time (BPTT), the mainstream approach to train recurrent neural networks, when nudging is performed with infinitely small strength. In practice, however, the standard implementation of Equilibrium Propagation does not scale to visual tasks harder than MNIST. In this work, we show that a bias in the gradient estimate of equilibrium propagation, inherent in the use of finite nudging, is responsible for this phenomenon and that canceling it allows training deep convolutional neural networks. We show that this bias can be greatly reduced by using symmetric nudging (a positive nudging and a negative one). We also generalize Equilibrium Propagation to the case of cross-entropy loss (by opposition to squared error). As a result of these advances, we are able to achieve a test error of 11.7% on CIFAR-10, which approaches the one achieved by BPTT and provides a major improvement with respect to the standard Equilibrium Propagation that gives 86% test error. We also apply these techniques to train an architecture with unidirectional forward and backward connections, yielding a 13.2% test error. These results highlight equilibrium propagation as a compelling biologically-plausible approach to compute error gradients in deep neuromorphic systems."
                },
                {
                    "title": "Training End-to-End Analog Neural Networks with Equilibrium Propagation",
                    "abstract": "We introduce a principled method to train end-to-end analog neural networks by stochastic gradient descent. In these analog neural networks, the weights to be adjusted are implemented by the conductances of programmable resistive devices such as memristors [Chua, 1971], and the nonlinear transfer functions (or `activation functions') are implemented by nonlinear components such as diodes. We show mathematically that a class of analog neural networks (called nonlinear resistive networks) are energy-based models: they possess an energy function as a consequence of Kirchhoff's laws governing electrical circuits. This property enables us to train them using the Equilibrium Propagation framework [Scellier and Bengio, 2017]. Our update rule for each conductance, which is local and relies solely on the voltage drop across the corresponding resistor, is shown to compute the gradient of the loss function. Our numerical simulations, which use the SPICE-based Spectre simulation framework to simulate the dynamics of electrical circuits, demonstrate training on the MNIST classification task, performing comparably or better than equivalent-size software-based neural networks. Our work can guide the development of a new generation of ultra-fast, compact and low-power neural networks supporting on-chip learning."
                },
                {
                    "title": "Equilibrium Propagation with Continual Weight Updates",
                    "abstract": "Equilibrium Propagation (EP) is a learning algorithm that bridges Machine Learning and Neuroscience, by computing gradients closely matching those of Backpropagation Through Time (BPTT), but with a learning rule local in space. Given an input $x$ and associated target $y$, EP proceeds in two phases: in the first phase neurons evolve freely towards a first steady state; in the second phase output neurons are nudged towards $y$ until they reach a second steady state. However, in existing implementations of EP, the learning rule is not local in time: the weight update is performed after the dynamics of the second phase have converged and requires information of the first phase that is no longer available physically. In this work, we propose a version of EP named Continual Equilibrium Propagation (C-EP) where neuron and synapse dynamics occur simultaneously throughout the second phase, so that the weight update becomes local in time. Such a learning rule local both in space and time opens the possibility of an extremely energy efficient hardware implementation of EP. We prove theoretically that, provided the learning rates are sufficiently small, at each time step of the second phase the dynamics of neurons and synapses follow the gradients of the loss given by BPTT (Theorem 1). We demonstrate training with C-EP on MNIST and generalize C-EP to neural networks where neurons are connected by asymmetric connections. We show through experiments that the more the network updates follows the gradients of BPTT, the best it performs in terms of training. These results bring EP a step closer to biology by better complying with hardware constraints while maintaining its intimate link with backpropagation."
                },
                {
                    "title": "Deep Equilibrium Models",
                    "abstract": "We present a new approach to modeling sequential data: the deep equilibrium model (DEQ). Motivated by an observation that the hidden layers of many existing deep sequence models converge towards some fixed point, we propose the DEQ approach that directly finds these equilibrium points via root-finding. Such a method is equivalent to running an infinite depth (weight-tied) feedforward network, but has the notable advantage that we can analytically backpropagate through the equilibrium point using implicit differentiation. Using this approach, training and prediction in these networks require only constant memory, regardless of the effective \u201cdepth\u201d of the network. We demonstrate how DEQs can be applied to two state-of-the-art deep sequence models: self-attention transformers and trellis networks. On large-scale language modeling tasks, such as the WikiText-103 benchmark, we show that DEQs 1) often improve performance over these state-of-the-art models (for similar parameter counts); 2) have similar computational requirements to existing models; and 3) vastly reduce memory consumption (often the bottleneck for training large sequence models), demonstrating an up-to 88% memory reduction in our experiments. The code is available at https://github.com/locuslab/deq."
                },
                {
                    "title": "Updates of Equilibrium Prop Match Gradients of Backprop Through Time in an RNN with Static Input",
                    "abstract": "Equilibrium Propagation (EP) is a biologically inspired learning algorithm for convergent recurrent neural networks, i.e. RNNs that are fed by a static input x and settle to a steady state. Training convergent RNNs consists in adjusting the weights until the steady state of output neurons coincides with a target y. Convergent RNNs can also be trained with the more conventional Backpropagation Through Time (BPTT) algorithm. In its original formulation EP was described in the case of real-time neuronal dynamics, which is computationally costly. In this work, we introduce a discrete-time version of EP with simplified equations and with reduced simulation time, bringing EP closer to practical machine learning tasks. We first prove theoretically, as well as numerically that the neural and weight updates of EP, computed by forward-time dynamics, are step-by-step equal to the ones obtained by BPTT, with gradients computed backward in time. The equality is strict when the transition function of the dynamics derives from a primitive function and the steady state is maintained long enough. We then show for more standard discrete-time neural network dynamics that the same property is approximately respected and we subsequently demonstrate training with EP with equivalent performance to BPTT. In particular, we define the first convolutional architecture trained with EP achieving ~ 1% test error on MNIST, which is the lowest error reported with EP. These results can guide the development of deep neural networks trained with EP."
                },
                {
                    "title": "Putting An End to End-to-End: Gradient-Isolated Learning of Representations",
                    "abstract": "We propose a novel deep learning method for local self-supervised representation learning that does not require labels nor end-to-end backpropagation but exploits the natural order in data instead. Inspired by the observation that biological neural networks appear to learn without backpropagating a global error signal, we split a deep neural network into a stack of gradient-isolated modules. Each module is trained to maximally preserve the information of its inputs using the InfoNCE bound from Oord et al. [2018]. Despite this greedy training, we demonstrate that each module improves upon the output of its predecessor, and that the representations created by the top module yield highly competitive results on downstream classification tasks in the audio and visual domain. The proposal enables optimizing modules asynchronously, allowing large-scale distributed training of very deep neural networks on unlabelled datasets."
                },
                {
                    "title": "Deep Learning without Weight Transport",
                    "abstract": "Current algorithms for deep learning probably cannot run in the brain because they rely on weight transport, where forward-path neurons transmit their synaptic weights to a feedback path, in a way that is likely impossible biologically. An algorithm called feedback alignment achieves deep learning without weight transport by using random feedback weights, but it performs poorly on hard visual-recognition tasks. Here we describe two mechanisms - a neural circuit called a weight mirror and a modification of an algorithm proposed by Kolen and Pollack in 1994 - both of which let the feedback path learn appropriate synaptic weights quickly and accurately even in large networks, without weight transport or complex wiring.Tested on the ImageNet visual-recognition task, these mechanisms outperform both feedback alignment and the newer sign-symmetry method, and nearly match backprop, the standard algorithm of deep learning, which uses weight transport."
                },
                {
                    "title": "Robust computation with rhythmic spike patterns",
                    "abstract": "Significance This work makes 2 contributions. First, we present a neural network model of associative memory that stores and retrieves sparse patterns of complex variables. This network can store analog information as fixed-point attractors in the complex domain; it is governed by an energy function and has increased memory capacity compared to early models. Second, we translate complex attractor networks into spiking networks, where the timing of the spike indicates the phase of a complex number. We show that complex fixed points correspond to stable periodic spike patterns. It is demonstrated that such networks can be constructed with resonate-and-fire or integrate-and-fire neurons with biologically plausible mechanisms and be used for robust computations, such as image retrieval. Information coding by precise timing of spikes can be faster and more energy efficient than traditional rate coding. However, spike-timing codes are often brittle, which has limited their use in theoretical neuroscience and computing applications. Here, we propose a type of attractor neural network in complex state space and show how it can be leveraged to construct spiking neural networks with robust computational properties through a phase-to-timing mapping. Building on Hebbian neural associative memories, like Hopfield networks, we first propose threshold phasor associative memory (TPAM) networks. Complex phasor patterns whose components can assume continuous-valued phase angles and binary magnitudes can be stored and retrieved as stable fixed points in the network dynamics. TPAM achieves high memory capacity when storing sparse phasor patterns, and we derive the energy function that governs its fixed-point attractor dynamics. Second, we construct 2 spiking neural networks to approximate the complex algebraic computations in TPAM, a reductionist model with resonate-and-fire neurons and a biologically plausible network of integrate-and-fire neurons with synaptic delays and recurrently connected inhibitory interneurons. The fixed points of TPAM correspond to stable periodic states of precisely timed spiking activity that are robust to perturbation. The link established between rhythmic firing patterns and complex attractor dynamics has implications for the interpretation of spike patterns seen in neuroscience and can serve as a framework for computation in emerging neuromorphic devices."
                },
                {
                    "title": "Dendritic cortical microcircuits approximate the backpropagation algorithm",
                    "abstract": "Deep learning has seen remarkable developments over the last years, many of them inspired by neuroscience. However, the main learning mechanism behind these advances - error backpropagation - appears to be at odds with neurobiology. Here, we introduce a multilayer neuronal network model with simplified dendritic compartments in which error-driven synaptic plasticity adapts the network towards a global desired output. In contrast to previous work our model does not require separate phases and synaptic learning is driven by local dendritic prediction errors continuously in time. Such errors originate at apical dendrites and occur due to a mismatch between predictive input from lateral interneurons and activity from actual top-down feedback. Through the use of simple dendritic compartments and different cell-types our model can represent both error and normal activity within a pyramidal neuron. We demonstrate the learning capabilities of the model in regression and classification tasks, and show analytically that it approximates the error backpropagation algorithm. Moreover, our framework is consistent with recent observations of learning between brain areas and the architecture of cortical microcircuits. Overall, we introduce a novel view of learning on dendritic cortical circuits and on how the brain may solve the long-standing synaptic credit assignment problem."
                },
                {
                    "title": "Biologically-plausible learning algorithms can scale to large datasets",
                    "abstract": "The backpropagation (BP) algorithm is often thought to be biologically implausible in the brain. One of the main reasons is that BP requires symmetric weight matrices in the feedforward and feedback pathways. To address this \"weight transport problem\" (Grossberg, 1987), two more biologically plausible algorithms, proposed by Liao et al. (2016) and Lillicrap et al. (2016), relax BP's weight symmetry requirements and demonstrate comparable learning capabilities to that of BP on small datasets. However, a recent study by Bartunov et al. (2018) evaluate variants of target-propagation (TP) and feedback alignment (FA) on MINIST, CIFAR, and ImageNet datasets, and find that although many of the proposed algorithms perform well on MNIST and CIFAR, they perform significantly worse than BP on ImageNet. Here, we additionally evaluate the sign-symmetry algorithm (Liao et al., 2016), which differs from both BP and FA in that the feedback and feedforward weights share signs but not magnitudes. We examine the performance of sign-symmetry and feedback alignment on ImageNet and MS COCO datasets using different network architectures (ResNet-18 and AlexNet for ImageNet, RetinaNet for MS COCO). Surprisingly, networks trained with sign-symmetry can attain classification performance approaching that of BP-trained networks. These results complement the study by Bartunov et al. (2018), and establish a new benchmark for future biologically plausible learning algorithms on more difficult datasets and more complex architectures."
                },
                {
                    "title": "Generalization of Equilibrium Propagation to Vector Field Dynamics",
                    "abstract": "The biological plausibility of the backpropagation algorithm has long been doubted by neuroscientists. Two major reasons are that neurons would need to send two different types of signal in the forward and backward phases, and that pairs of neurons would need to communicate through symmetric bidirectional connections. We present a simple two-phase learning procedure for fixed point recurrent networks that addresses both these issues. In our model, neurons perform leaky integration and synaptic weights are updated through a local mechanism. Our learning method generalizes Equilibrium Propagation to vector field dynamics, relaxing the requirement of an energy function. As a consequence of this generalization, the algorithm does not compute the true gradient of the objective function, but rather approximates it at a precision which is proven to be directly related to the degree of symmetry of the feedforward and feedback weights. We show experimentally that our algorithm optimizes the objective function."
                },
                {
                    "title": "Error Forward-Propagation: Reusing Feedforward Connections to Propagate Errors in Deep Learning",
                    "abstract": "We introduce Error Forward-Propagation, a biologically plausible mechanism to propagate error feedback forward through the network. Architectural constraints on connectivity are virtually eliminated for error feedback in the brain; systematic backward connectivity is not used or needed to deliver error feedback. Feedback as a means of assigning credit to neurons earlier in the forward pathway for their contribution to the final output is thought to be used in learning in the brain. How the brain solves the credit assignment problem is unclear. In machine learning, error backpropagation is a highly successful mechanism for credit assignment in deep multilayered networks. Backpropagation requires symmetric reciprocal connectivity for every neuron. From a biological perspective, there is no evidence of such an architectural constraint, which makes backpropagation implausible for learning in the brain. This architectural constraint is reduced with the use of random feedback weights. Models using random feedback weights require backward connectivity patterns for every neuron, but avoid symmetric weights and reciprocal connections. In this paper, we practically remove this architectural constraint, requiring only a backward loop connection for effective error feedback. We propose reusing the forward connections to deliver the error feedback by feeding the outputs into the input receiving layer. This mechanism, Error Forward-Propagation, is a plausible basis for how error feedback occurs deep in the brain independent of and yet in support of the functionality underlying intricate network architectures. We show experimentally that recurrent neural networks with two and three hidden layers can be trained using Error Forward-Propagation on the MNIST and Fashion MNIST datasets, achieving $1.90\\%$ and $11\\%$ generalization errors respectively."
                },
                {
                    "title": "Assessing the Scalability of Biologically-Motivated Deep Learning Algorithms and Architectures",
                    "abstract": "The backpropagation of error algorithm (BP) is impossible to implement in a real brain. The recent success of deep networks in machine learning and AI, however, has inspired proposals for understanding how the brain might learn across multiple layers, and hence how it might approximate BP. As of yet, none of these proposals have been rigorously evaluated on tasks where BP-guided deep learning has proved critical, or in architectures more structured than simple fully-connected networks. Here we present results on scaling up biologically motivated models of deep learning on datasets which need deep networks with appropriate architectures to achieve good performance. We present results on the MNIST, CIFAR-10, and ImageNet datasets and explore variants of target-propagation (TP) and feedback alignment (FA) algorithms, and explore performance in both fully- and locally-connected architectures. We also introduce weight-transport-free variants of difference target propagation (DTP) modified to remove backpropagation from the penultimate layer. Many of these algorithms perform well for MNIST, but for CIFAR and ImageNet we find that TP and FA variants perform significantly worse than BP, especially for networks composed of locally connected units, opening questions about whether new architectures and algorithms are required to scale these approaches. Our results and implementation details help establish baselines for biologically motivated deep learning schemes going forward."
                },
                {
                    "title": "Large-Scale Neuromorphic Spiking Array Processors: A Quest to Mimic the Brain",
                    "abstract": "Neuromorphic engineering (NE) encompasses a diverse range of approaches to information processing that are inspired by neurobiological systems, and this feature distinguishes neuromorphic systems from conventional computing systems. The brain has evolved over billions of years to solve difficult engineering problems by using efficient, parallel, low-power computation. The goal of NE is to design systems capable of brain-like computation. Numerous large-scale neuromorphic projects have emerged recently. This interdisciplinary field was listed among the top 10 technology breakthroughs of 2014 by the MIT Technology Review and among the top 10 emerging technologies of 2015 by the World Economic Forum. NE has two-way goals: one, a scientific goal to understand the computational properties of biological neural systems by using models implemented in integrated circuits (ICs); second, an engineering goal to exploit the known properties of biological systems to design and implement efficient devices for engineering applications. Building hardware neural emulators can be extremely useful for simulating large-scale neural models to explain how intelligent behavior arises in the brain. The principal advantages of neuromorphic emulators are that they are highly energy efficient, parallel and distributed, and require a small silicon area. Thus, compared to conventional CPUs, these neuromorphic emulators are beneficial in many engineering applications such as for the porting of deep learning algorithms for various recognitions tasks. In this review article, we describe some of the most significant neuromorphic spiking emulators, compare the different architectures and approaches used by them, illustrate their advantages and drawbacks, and highlight the capabilities that each can deliver to neural modelers. This article focuses on the discussion of large-scale emulators and is a continuation of a previous review of various neural and synapse circuits (Indiveri et al., 2011). We also explore applications where these emulators have been used and discuss some of their promising future applications."
                },
                {
                    "title": "Reviving and Improving Recurrent Back-Propagation",
                    "abstract": "In this paper, we revisit the recurrent back-propagation (RBP) algorithm, discuss the conditions under which it applies as well as how to satisfy them in deep neural networks. We show that RBP can be unstable and propose two variants based on conjugate gradient on the normal equations (CG-RBP) and Neumann series (Neumann-RBP). We further investigate the relationship between Neumann-RBP and back propagation through time (BPTT) and its truncated version (TBPTT). Our Neumann-RBP has the same time complexity as TBPTT but only requires constant memory, whereas TBPTT's memory cost scales linearly with the number of truncation steps. We examine all RBP variants along with BPTT and TBPTT in three different application domains: associative memory with continuous Hopfield networks, document classification in citation networks using graph neural networks and hyperparameter optimization for fully connected networks. All experiments demonstrate that RBPs, especially the Neumann-RBP variant, are efficient and effective for optimizing convergent recurrent neural networks. Code is released at: \\url{this https URL}."
                },
                {
                    "title": "Equivalence of Equilibrium Propagation and Recurrent Backpropagation",
                    "abstract": "Recurrent backpropagation and equilibrium propagation are supervised learning algorithms for fixed-point recurrent neural networks, which differ in their second phase. In the first phase, both algorithms converge to a fixed point that corresponds to the configuration where the prediction is made. In the second phase, equilibrium propagation relaxes to another nearby fixed point corresponding to smaller prediction error, whereas recurrent backpropagation uses a side network to compute error derivatives iteratively. In this work, we establish a close connection between these two algorithms. We show that at every moment in the second phase, the temporal derivatives of the neural activities in equilibrium propagation are equal to the error derivatives computed iteratively by recurrent backpropagation in the side network. This work shows that it is not required to have a side network for the computation of error derivatives and supports the hypothesis that in biological neural networks, temporal derivatives of neural activities may code for error signals."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms",
                    "abstract": "We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at this https URL"
                },
                {
                    "title": "A Survey of Neuromorphic Computing and Neural Networks in Hardware",
                    "abstract": "Neuromorphic computing has come to refer to a variety of brain-inspired computers, devices, and models that contrast the pervasive von Neumann computer architecture. This biologically inspired approach has created highly connected synthetic neurons and synapses that can be used to model neuroscience theories as well as solve challenging machine learning problems. The promise of the technology is to create a brain-like ability to learn and adapt, but the technical challenges are significant, starting with an accurate neuroscience model of how the brain works, to finding materials and engineering breakthroughs to build devices to support these models, to creating a programming framework so the systems can learn, to creating applications with brain-like capabilities. In this work, we provide a comprehensive survey of the research and motivations for neuromorphic computing over its history. We begin with a 35-year review of the motivations and drivers of neuromorphic computing, then look at the major research areas of the field, which we define as neuro-inspired models, algorithms and learning approaches, hardware and devices, supporting systems, and finally applications. We conclude with a broad discussion on the major research topics that need to be addressed in the coming years to see the promise of neuromorphic computing fulfilled. The goals of this work are to provide an exhaustive review of the research conducted in neuromorphic computing since the inception of the term, and to motivate further work by illuminating gaps in the field where new research is needed."
                },
                {
                    "title": "The temporal paradox of Hebbian learning and homeostatic plasticity",
                    "abstract": "Hebbian plasticity, a synaptic mechanism which detects and amplifies co-activity between neurons, is considered a key ingredient underlying learning and memory in the brain. However, Hebbian plasticity alone is unstable, leading to runaway neuronal activity, and therefore requires stabilization by additional compensatory processes. Traditionally, a diversity of homeostatic plasticity phenomena found in neural circuits are thought to play this role. However, recent modelling work suggests that the slow evolution of homeostatic plasticity, as observed in experiments, is insufficient to prevent instabilities originating from Hebbian plasticity. To remedy this situation, we suggest that homeostatic plasticity is complemented by additional rapid compensatory processes, which rapidly stabilize neuronal activity on short timescales."
                },
                {
                    "title": "Direct Feedback Alignment Provides Learning in Deep Neural Networks",
                    "abstract": "Artificial neural networks are most commonly trained with the back-propagation algorithm, where the gradient for learning is provided by back-propagating the error, layer by layer, from the output layer to the hidden layers. A recently discovered method called feedback-alignment shows that the weights used for propagating the error backward don't have to be symmetric with the weights used for propagation the activation forward. In fact, random feedback weights work evenly well, because the network learns how to make the feedback useful. In this work, the feedback alignment principle is used for training hidden layers more independently from the rest of the network, and from a zero initial condition. The error is propagated through fixed random feedback connections directly from the output layer to each hidden layer. This simple method is able to achieve zero training error even in convolutional networks and very deep networks, completely without error back-propagation. The method is a step towards biologically plausible machine learning because the error signal is almost local, and no symmetric or reciprocal weights are required. Experiments show that the test performance on MNIST and CIFAR is almost as good as those obtained with back-propagation for fully connected networks. If combined with dropout, the method achieves 1.45% error on the permutation invariant MNIST task."
                },
                {
                    "title": "Dense Associative Memory for Pattern Recognition",
                    "abstract": "A model of associative memory is studied, which stores and reliably retrieves many more patterns than the number of neurons in the network. We propose a simple duality between this dense associative memory and neural networks commonly used in deep learning. On the associative memory side of this duality, a family of models that smoothly interpolates between two limiting cases can be constructed. One limit is referred to as the feature-matching mode of pattern recognition, and the other one as the prototype regime. On the deep learning side of the duality, this family corresponds to feedforward neural networks with one hidden layer and various activation functions, which transmit the activities of the visible neurons to the hidden layer. This family of activation functions includes logistics, rectified linear units, and rectified polynomials of higher degrees. The proposed duality makes it possible to apply energy-based intuition from associative memory to analyze computational properties of neural networks with unusual activation functions - the higher rectified polynomials which until now have not been used in deep learning. The utility of the dense memories is illustrated for two test cases: the logical gate XOR and the recognition of handwritten digits from the MNIST data set."
                },
                {
                    "title": "Equilibrium Propagation: Bridging the Gap between Energy-Based Models and Backpropagation",
                    "abstract": "We introduce Equilibrium Propagation, a learning framework for energy-based models. It involves only one kind of neural computation, performed in both the first phase (when the prediction is made) and the second phase of training (after the target or prediction error is revealed). Although this algorithm computes the gradient of an objective function just like Backpropagation, it does not need a special computation or circuit for the second phase, where errors are implicitly propagated. Equilibrium Propagation shares similarities with Contrastive Hebbian Learning and Contrastive Divergence while solving the theoretical issues of both algorithms: our algorithm computes the gradient of a well-defined objective function. Because the objective function is defined in terms of local perturbations, the second phase of Equilibrium Propagation corresponds to only nudging the prediction (fixed point or stationary distribution) toward a configuration that reduces prediction error. In the case of a recurrent multi-layer supervised network, the output units are slightly nudged toward their target in the second phase, and the perturbation introduced at the output layer propagates backward in the hidden layers. We show that the signal \u201cback-propagated\u201d during this second phase corresponds to the propagation of error derivatives and encodes the gradient of the objective function, when the synaptic update corresponds to a standard form of spike-timing dependent plasticity. This work makes it more plausible that a mechanism similar to Backpropagation could be implemented by brains, since leaky integrator neural computation performs both inference and error back-propagation in our model. The only local difference between the two phases is whether synaptic changes are allowed or not. We also show experimentally that multi-layer recurrently connected networks with 1, 2, and 3 hidden layers can be trained by Equilibrium Propagation on the permutation-invariant MNIST task."
                },
                {
                    "title": "How Important Is Weight Symmetry in Backpropagation?",
                    "abstract": "\n \n Gradient backpropagation (BP) requires symmetric feedforward and feedback connections \u2014 the same weights must be used for forward and backward passes. This \"weight transport problem'' (Grossberg 1987) is thought to be one of the main reasons to doubt BP's biologically plausibility. Using 15 different classification datasets, we systematically investigate to what extent BP really depends on weight symmetry. In a study that turned out to be surprisingly similar in spirit to Lillicrap et al.'s demonstration (Lillicrap et al. 2014) but orthogonal in its results, our experiments indicate that: (1) the magnitudes of feedback weights do not matter to performance (2) the signs of feedback weights do matter \u2014 the more concordant signs between feedforward and their corresponding feedback connections, the better (3) with feedback weights having random magnitudes and 100% concordant signs, we were able to achieve the same or even better performance than SGD. (4) some normalizations/stabilizations are indispensable for such asymmetric BP to work, namely Batch Normalization (BN) (Ioffe and Szegedy 2015) and/or a \"Batch Manhattan'' (BM) update rule.\n \n"
                },
                {
                    "title": "How Auto-Encoders Could Provide Credit Assignment in Deep Networks via Target Propagation",
                    "abstract": "We propose to exploit {\\em reconstruction} as a layer-local training signal for deep learning. Reconstructions can be propagated in a form of target propagation playing a role similar to back-propagation but helping to reduce the reliance on derivatives in order to perform credit assignment across many levels of possibly strong non-linearities (which is difficult for back-propagation). A regularized auto-encoder tends produce a reconstruction that is a more likely version of its input, i.e., a small move in the direction of higher likelihood. By generalizing gradients, target propagation may also allow to train deep networks with discrete hidden units. If the auto-encoder takes both a representation of input and target (or of any side information) in input, then its reconstruction of input representation provides a target towards a representation that is more likely, conditioned on all the side information. A deep auto-encoder decoding path generalizes gradient propagation in a learned way that can could thus handle not just infinitesimal changes but larger, discrete changes, hopefully allowing credit assignment through a long chain of non-linear operations. In addition to each layer being a good auto-encoder, the encoder also learns to please the upper layers by transforming the data into a space where it is easier to model by them, flattening manifolds and disentangling factors. The motivations and theoretical justifications for this approach are laid down in this paper, along with conjectures that will have to be verified either mathematically or experimentally, including a hypothesis stating that such auto-encoder mediated target propagation could play in brains the role of credit assignment through many non-linear, noisy and discrete transformations."
                },
                {
                    "title": "Neuromorphic Silicon Neuron Circuits",
                    "abstract": "Hardware implementations of spiking neurons can be extremely useful for a large variety of applications, ranging from high-speed modeling of large-scale neural systems to real-time behaving systems, to bidirectional brain\u2013machine interfaces. The specific circuit solutions used to implement silicon neurons depend on the application requirements. In this paper we describe the most common building blocks and techniques used to implement these circuits, and present an overview of a wide range of neuromorphic silicon neurons, which implement different computational models, ranging from biophysically realistic and conductance-based Hodgkin\u2013Huxley models to bi-dimensional generalized adaptive integrate and fire models. We compare the different design methodologies used for each silicon neuron design described, and demonstrate their features with experimental results, measured from a wide range of fabricated VLSI chips."
                },
                {
                    "title": "Spike-timing dependent plasticity",
                    "abstract": "Spike Timing Dependent Plasticity (STDP) is a temporally asymmetric form of Hebbian learning induced by tight temporal correlations between the spikes of pre- and postsynaptic neurons. As with other forms of synaptic plasticity, it is widely believed that it underlies learning and information storage in the brain, as well as the development and refinement of neuronal circuits during brain development (e.g. Bi and Poo, 2001; Sjostrom et al., 2008). With STDP, repeated presynaptic spike arrival a few milliseconds before postsynaptic action potentials leads in many synapse types to long-term potentiation (LTP) of the synapses, whereas repeated spike arrival after postsynaptic spikes leads to long-term depression (LTD) of the same synapse. The change of the synapse plotted as a function of the relative timing of pre- and postsynaptic action potentials is called the STDP function or learning window and varies between synapse types. The rapid change of the STDP function with the relative timing of spikes suggests the possibility of temporal coding schemes on a millisecond time scale."
                },
                {
                    "title": "Slow Feature Analysis: Unsupervised Learning of Invariances",
                    "abstract": "Invariant features of temporally varying signals are useful for analysis and classification. Slow feature analysis (SFA) is a new method for learning invariant or slowly varying features from a vectorial input signal. It is based on a nonlinear expansion of the input signal and application of principal component analysis to this expanded signal and its time derivative. It is guaranteed to find the optimal solution within a family of functions directly and can learn to extract a large number of decor-related features, which are ordered by their degree of invariance. SFA can be applied hierarchically to process high-dimensional input signals and extract complex features. SFA is applied first to complex cell tuning properties based on simple cell output, including disparity and motion. Then more complicated input-output functions are learned by repeated application of SFA. Finally, a hierarchical network of SFA modules is presented as a simple model of the visual system. The same unstructured network can learn translation, size, rotation, contrast, or, to a lesser degree, illumination invariance for one-dimensional objects, depending on only the training stimulus. Surprisingly, only a few training objects suffice to achieve good generalization to new objects. The generated representation is suitable for object recognition. Performance degrades if the network is trained to learn multiple invariances simultaneously."
                },
                {
                    "title": "Backpropagation without weight transport",
                    "abstract": "In backpropagation, connection weights are used to both compute node activations and error gradient for hidden units. Grossberg (1987) has argued that the dual use of the same synaptic connections (weight transport) constitutes a bidirectional flow of information through synapses, which is biologically implausable. In this paper we formally and empirically demonstrate the feasibility of an architecture equivalent to backpropagation, but without the assumption of weight transport. Through coordinated training with weight decay, a reciprocal layer of weights evolves into a copy of the forward connections and acts as the conduit for backward flowing corrective information. Examination of the networks trained with dual weights suggests that functional synchronization, and not weight synchronization, is crucial to the operation of backpropagation methods.<<ETX>>"
                },
                {
                    "title": "Contrastive Learning and Neural Oscillations",
                    "abstract": "The concept of Contrastive Learning (CL) is developed as a family of possible learning algorithms for neural networks. CL is an extension of Deterministic Boltzmann Machines to more general dynamical systems. During learning, the network oscillates between two phases. One phase has a teacher signal and one phase has no teacher signal. The weights are updated using a learning rule that corresponds to gradient descent on a contrast function that measures the discrepancy between the free network and the network with a teacher signal. The CL approach provides a general unified framework for developing new learning algorithms. It also shows that many different types of clamping and teacher signals are possible. Several examples are given and an analysis of the landscape of the contrast function is proposed with some relevant predictions for the CL curves. An approach that may be suitable for collective analog implementations is described. Simulation results and possible extensions are briefly discussed together with a new conjecture regarding the function of certain oscillations in the brain. In the appendix, we also examine two extensions of contrastive learning to time-dependent trajectories."
                },
                {
                    "title": "Competitive Learning: From Interactive Activation to Adaptive Resonance",
                    "abstract": "Functional and mechanistic comparisons are made between several network models of cognitive processing: competitive learning, interactive activation, adaptive resonance, and back propagation. The starting point of this comparison is the article of Rumelhart and Zipser (1985) on feature discovery through competitive learning. All the models which Rumelhart and Zipser (1985) have described were shown in Grossberg (1976b) to exhibit a type of learning which is temporally unstable. Competitive learning mechanisms can be stabilized in response to an arbitrary input environment by being supplemented with mechanisms for learning top-down expectancies, or templates; for matching bottom-up input patterns with the top-down expectancies; and for releasing orienting reactions in a mismatch situation, thereby updating short-term memory and searching for another internal representation. Network architectures which embody all of these mechanisms were called adaptive resonance models by Grossberg (1976c) . Self-stabilizing learning models are candidates for use in real-world applications where unpredictable changes can occur in complex input environments. Competitive learning postulates are inconsistent with the postulates of the interactive activation model of McClelland and Rumelhart (1981) , and suggest different levels of processing and interaction rules for the analysis of word recognition. Adaptive resonance models use these alternative levels and interaction rules. The selforganizing learning of an adaptive resonance model is compared and contrasted with the teacher-directed learning of a back propagation model. A number of criteria for evaluating real-time network models of cognitive processing are described and applied."
                },
                {
                    "title": "Neurons with graded response have collective computational properties like those of two-state neurons.",
                    "abstract": "A model for a large network of \"neurons\" with a graded response (or sigmoid input-output relation) is studied. This deterministic system has collective properties in very close correspondence with the earlier stochastic model based on McCulloch - Pitts neurons. The content- addressable memory and other emergent collective properties of the original model also are present in the graded response model. The idea that such collective properties are used in biological systems is given added credence by the continued presence of such properties for more nearly biological \"neurons.\" Collective analog electrical circuits of the kind described will certainly function. The collective states of the two models have a simple correspondence. The original model will continue to be useful for simulations, because its connection to graded response systems is established. Equations that include the effect of action potentials in the graded response system are also developed."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "The Graph Neural Network Model",
                    "abstract": "Many underlying relationships among data in several areas of science and engineering, e.g., computer vision, molecular chemistry, molecular biology, pattern recognition, and data mining, can be represented in terms of graphs. In this paper, we propose a new neural network model, called graph neural network (GNN) model, that extends existing neural network methods for processing the data represented in graph domains. This GNN model, which can directly process most of the practically useful types of graphs, e.g., acyclic, cyclic, directed, and undirected, implements a function tau(G,n) isin IRm that maps a graph G and one of its nodes n into an m-dimensional Euclidean space. A supervised learning algorithm is derived to estimate the parameters of the proposed GNN model. The computational cost of the proposed algorithm is also considered. Some experimental results are shown to validate the proposed learning algorithm, and to demonstrate its generalization capabilities."
                },
                {
                    "title": "Gradient descent learning algorithm overview: a general dynamical systems perspective",
                    "abstract": "Gives a unified treatment of gradient descent learning algorithms for neural networks using a general framework of dynamical systems. This general approach organizes and simplifies all the known algorithms and results which have been originally derived for different problems (fixed point/trajectory learning), for different models (discrete/continuous), for different architectures (forward/recurrent), and using different techniques (backpropagation, variational calculus, adjoint methods, etc.). The general approach can also be applied to derive new algorithms. The author then briefly examines some of the complexity issues and limitations intrinsic to gradient descent learning. Throughout the paper, the author focuses on the problem of trajectory learning."
                },
                {
                    "title": "A stochastic estimator of the trace of the influence matrix for laplacian smoothing splines",
                    "abstract": "An unbiased stochastic estimator of tr(I-A), where A is the influence matrix associated with the calculation of Laplacian smoothing splines, is described. The estimator is similar to one recently developed by Girard but satisfies a minimum variance criterion and does not require the simulation of a standard normal variable. It uses instead simulations of the discrete random variable which takes the values 1, -1 each with probability 1/2. Bounds on the variance of the estimator, similar to those established by Girard, are obtained using elementary methods. The estimator can be used to approximately minimize generalised cross validation (GCV) when using discretized iterative methods for fitting Laplacian smoothing splines to very large data sets. Simulated examples show that the estimated trace values, using either the estimator presented here or the estimator of Girard, perform almost as well as the exact values when applied to the minimization of GCV for n as small as a few hundred, where n is the number ..."
                },
                {
                    "title": "GEMINI: Gradient Estimation Through Matrix Inversion After Noise Injection",
                    "abstract": "Learning procedures that measure how random perturbations of unit activities correlate with changes in reinforcement are inefficient but simple to implement in hardware. Procedures like back-propagation (Rumelhart, Hinton and Williams, 1986) which compute how changes in activities affect the output error are much more efficient, but require more complex hardware. GEMINI is a hybrid procedure for multilayer networks, which shares many of the implementation advantages of correlational reinforcement procedures but is more efficient. GEMINI injects noise only at the first hidden layer and measures the resultant effect on the output error. A linear network associated with each hidden layer iteratively inverts the matrix which relates the noise to the error change, thereby obtaining the error-derivatives. No back-propagation is involved, thus allowing unknown non-linearities in the system. Two simulations demonstrate the effectiveness of GEMINI."
                },
                {
                    "title": "Generalization of Back propagation to Recurrent and Higher Order Neural Networks",
                    "abstract": "A general method for deriving backpropagation algorithms for networks with recurrent and higher order networks is introduced. The propagation of activation in these networks is determined by dissipative differential equations. The error signal is backpropagated by integrating an associated differential equation. The method is introduced by applying it to the recurrent generalization of the feedforward backpropagation network. The method is extended to the case of higher order networks and to a constrained dynamical system for training a content addressable memory. The essential feature of the adaptive algorithms is that adaptive equation has a simple outer product form. \n \nPreliminary experiments suggest that learning can occur very rapidly in networks with recurrent connections. The continuous formalism makes the new approach more suitable for implementation in VLSI."
                },
                {
                    "title": "Voltage and spike timing interact in STDP \u2013 a unified model",
                    "abstract": "the same timing difference leads to insignificant changes, whereas post-pre-post induces a strong potentiation of synapses in hip-pocampal cultures (Wang et al., 2005). Quite generally, triplet and quadruplet dependencies measured in various experiments (Froemke and Dan, 2002; Wang et al., 2005) cannot be explained by standard STDP models. Third, the STDP function derived from standard experiments with pairs of spikes (Markram et al., 1997; Bi and Poo, 1998; Sj\u00f6str\u00f6m et al., 2001) does not explain the dependence of plasticity upon the timing of bursts of several postsynaptic spikes (Kampa et al. or some voltage dependence (Brader et al., 2007) resolve at most one or two of the above issues, but not all of them. The basic shortcoming of the family of classical phenomeno-logical STDP models is an inappropriate focus on a single pair of pre-and postsynaptic spikes. There are a couple of promising phenomenological STDP models that go beyond the pair"
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.NE"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively extend equilibrium propagation (EP) to non-symmetric weight configurations in neural networks to improve its applicability in neuromorphic hardware?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of neuromorphic computing, as it addresses the limitations of current learning algorithms that rely on backpropagation. By enabling EP to function effectively with non-symmetric weights, we can potentially unlock significant energy savings in AI training, making it more feasible for real-world applications. This research could lead to breakthroughs in understanding how biological systems learn and adapt, influencing future research directions in both artificial intelligence and neuroscience. Additionally, it may pave the way for more efficient and powerful AI systems that can operate in energy-constrained environments.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexities of non-symmetric weight configurations and their impact on gradient estimation in EP. Naive approaches may fail because they do not account for the biases introduced by weight asymmetry and finite-size nudges, which can lead to inaccurate gradient estimates. The technical obstacles include the need for a robust theoretical framework that can handle these complexities and the practical challenge of implementing such a framework in a way that is compatible with existing neuromorphic hardware. Additionally, ensuring that the proposed method can generalize beyond simple tasks to more complex datasets like CIFAR-10 adds another layer of difficulty.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on symmetric weight configurations and has not adequately addressed the effects of weight asymmetry on EP. The lack of attention to this aspect has created a gap in understanding how to effectively apply EP in more complex scenarios. Barriers to solving this problem include the limited exploration of non-symmetric feedback mechanisms in the context of EP and the prevailing focus on backpropagation methods. Our approach differs by providing a comprehensive analysis of the biases introduced by weight asymmetry and proposing an extension of holomorphic EP that can operate under these conditions, thus improving upon prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a detailed analysis of the sources of bias in gradient estimates due to weight asymmetry and finite-size nudges in generalized EP. We will utilize a dataset such as CIFAR-10 to evaluate the performance of our extended hol"
            }
        },
        "author_data": {
            "18f341cb-b86d-447c-a213-f3db70b0bd52": {
                "pk": "18f341cb-b86d-447c-a213-f3db70b0bd52",
                "name": "Axel Laborieux",
                "collaborators": [
                    "D. Querlioz",
                    "E. Vianello",
                    "M. Bocquet",
                    "J. Portal",
                    "T. Hirtzlin",
                    "Jacques-Olivier Klein",
                    "E. Nowak",
                    "Friedemann Zenke",
                    "Manu Srinath Halvagal",
                    "F T Zenke",
                    "A. Majumdar",
                    "D. Gilbert",
                    "M. Belmeguenai",
                    "J. Vogel",
                    "S. Pizzini",
                    "E. Mart\u00ednez",
                    "A. Lamperti",
                    "Y. Roussign\u00e9",
                    "A. Grutter",
                    "E. Arenholtz",
                    "P. Quarterman",
                    "B. Maranville",
                    "S. Ono",
                    "M. S. E. Hadri",
                    "R. Tolley",
                    "E. Fullerton",
                    "A. Stashkevich",
                    "S. Ch\u00e9rif",
                    "A. Kent",
                    "J. Langer",
                    "D. Ravelosona",
                    "B. Ocker",
                    "A. Renaudineau",
                    "Kamel-Eddine Harabi",
                    "C. Turck",
                    "F. Jebali",
                    "J. Walder",
                    "Liza Herrera-Diez",
                    "Yuting Liu",
                    "J. Mohammadi",
                    "L. S. Jos\u00e9",
                    "M. Ernoult",
                    "B. Scellier",
                    "Yoshua Bengio",
                    "J. Grollier",
                    "L. D\u00edez",
                    "L. Diez",
                    "Yang Liu",
                    "J. B. Mohammedi",
                    "L. S\u00e1nchez-Tejerina"
                ],
                "domain": [
                    "Continual Learning",
                    "Neural Networks",
                    "Self-Supervised Learning",
                    "Neuromorphic Computing"
                ],
                "publications": [
                    {
                        "title": "Theories of synaptic memory consolidation and intelligent plasticity for continual learning",
                        "abstract": "Humans and animals learn throughout life. Such continual learning is crucial for intelligence. In this chapter, we examine the pivotal role plasticity mechanisms with complex internal synaptic dynamics could play in enabling this ability in neural networks. By surveying theoretical research, we highlight two fundamental enablers for continual learning. First, synaptic plasticity mechanisms must maintain and evolve an internal state over several behaviorally relevant timescales. Second, plasticity algorithms must leverage the internal state to intelligently regulate plasticity at individual synapses to facilitate the seamless integration of new memories while avoiding detrimental interference with existing ones. Our chapter covers successful applications of these principles to deep neural networks and underscores the significance of synaptic metaplasticity in sustaining continual learning capabilities. Finally, we outline avenues for further research to understand the brain's superb continual learning abilities and harness similar mechanisms for artificial intelligence systems."
                    },
                    {
                        "title": "Experimental Demonstration of Memristor Delay-Based Logic In-Memory Ternary Neural Network",
                        "abstract": "We present a fabricated hybrid CMOS/memristor integrated circuit for efficient implementation of Ternary Neural Networks. Our approach overcomes memristor resolution limitations and uses a simple sense amplifier that simultaneously reads memristor states and performs the TNN multiplication operations. The test chip validates our scheme, paving the way for energy-efficient, nanosecond-latency TNNs."
                    },
                    {
                        "title": "Implicit variance regularization in non-contrastive SSL",
                        "abstract": "Non-contrastive SSL methods like BYOL and SimSiam rely on asymmetric predictor networks to avoid representational collapse without negative samples. Yet, how predictor networks facilitate stable learning is not fully understood. While previous theoretical analyses assumed Euclidean losses, most practical implementations rely on cosine similarity. To gain further theoretical insight into non-contrastive SSL, we analytically study learning dynamics in conjunction with Euclidean and cosine similarity in the eigenspace of closed-form linear predictor networks. We show that both avoid collapse through implicit variance regularization albeit through different dynamical mechanisms. Moreover, we find that the eigenvalues act as effective learning rate multipliers and propose a family of isotropic loss functions (IsoLoss) that equalize convergence rates across eigenmodes. Empirically, IsoLoss speeds up the initial learning dynamics and increases robustness, thereby allowing us to dispense with the EMA target network typically used with non-contrastive methods. Our analysis sheds light on the variance regularization mechanisms of non-contrastive SSL and lays the theoretical grounds for crafting novel loss functions that shape the learning dynamics of the predictor's spectrum."
                    },
                    {
                        "title": "An eigenspace view reveals how predictor networks and stop-grads provide implicit variance regularization",
                        "abstract": "Self-supervised learning (SSL) learns useful representations from unlabelled data by training networks to be invariant to pairs of augmented versions of the same input. Non-contrastive methods avoid collapse either by directly regularizing the covariance matrix of network outputs or through asymmetric loss architectures, two seemingly unrelated approaches. Here, by building on DirectPred [1], we lay out a theoretical framework that reconciles these two views. We derive analytical expressions for the representational learning dynamics in linear networks. By expressing them in the eigenspace of the embedding covariance matrix, where the solutions decouple, we reveal the mechanism and conditions that provide implicit variance regularization. These insights allow us to formulate a new isotropic loss function that equalizes eigenvalue contribution and renders learning more robust. Finally, we show empirically that our findings translate in nonlinear networks trained on CIFAR-10 and STL-10."
                    },
                    {
                        "title": "Holomorphic Equilibrium Propagation Computes Exact Gradients Through Finite Size Oscillations",
                        "abstract": "Equilibrium propagation (EP) is an alternative to backpropagation (BP) that allows the training of deep neural networks with local learning rules. It thus provides a compelling framework for training neuromorphic systems and understanding learning in neurobiology. However, EP requires infinitesimal teaching signals, thereby limiting its applicability in noisy physical systems. Moreover, the algorithm requires separate temporal phases and has not been applied to large-scale problems. Here we address these issues by extending EP to holomorphic networks. We show analytically that this extension naturally leads to exact gradients even for finite-amplitude teaching signals. Importantly, the gradient can be computed as the first Fourier coefficient from finite neuronal activity oscillations in continuous time without requiring separate phases. Further, we demonstrate in numerical simulations that our approach permits robust estimation of gradients in the presence of noise and that deeper models benefit from the finite teaching signals. Finally, we establish the first benchmark for EP on the ImageNet 32x32 dataset and show that it matches the performance of an equivalent network trained with BP. Our work provides analytical insights that enable scaling EP to large-scale problems and establishes a formal framework for how oscillations could support learning in biological and neuromorphic systems."
                    },
                    {
                        "title": "Model of the Weak Reset Process in HfOx Resistive Memory for Deep Learning Frameworks",
                        "abstract": "The implementation of current deep learning training algorithms is power-hungry, due to data transfer between memory and logic units. Oxide-based resistive random access memories (RRAMs) are outstanding candidates to implement in-memory computing, which is less power-intensive. Their weak RESET regime is particularly attractive for learning, as it allows tuning the resistance of the devices with remarkable endurance. However, the resistive change behavior in this regime suffers from many fluctuations and is particularly challenging to model, especially in a way compatible with tools used for simulating deep learning. In this work, we present a model of the weak RESET process in hafnium oxide RRAM and integrate this model within the PyTorch deep learning framework. Validated on experiments on a hybrid CMOS/RRAM technology, our model reproduces both the noisy progressive behavior and the device-to-device (D2D) variability. We use this tool to train binarized neural networks (BNNs) for the MNIST handwritten digit recognition task and the CIFAR-10 object classification task. We simulate our model with and without various aspects of device imperfections to understand their impact on the training process and identify that the D2D variability is the most detrimental aspect. The framework can be used in the same manner for other types of memories to identify the device imperfections that cause the most degradation, which can, in turn, be used to optimize the devices to reduce the impact of these imperfections."
                    },
                    {
                        "title": "CAPC: A Configurable Analog Pop-Count Circuit for Near-Memory Binary Neural Networks",
                        "abstract": "Currently, a major trend in artificial intelligence is to implement neural networks at the edge, within circuits with limited memory capacity. To reach this goal, the in-memory or near-memory implementation of low precision neural networks such as Binarized Neural Networks (BNNs) constitutes an appealing solution. However, the configurability of these approaches is a major challenge: in neural networks, the number of neurons per layer vary tremendously depending on the application, limiting the column-wise or row-wise mapping of neurons in memory arrays. To tackle this issue, we propose, for the first time, a Configurable Analog auto-compensate Pop-Count (CAPC) circuit compatible with column-wise neuron mapping. Our circuit has the advantage of featuring a very natural configurability through analog switch connections. We demonstrate that our solution saves 18% of area compared to non configurable conventional digital solution. Moreover, through extensive Monte-Carlo simulations, we show that the overall error probability remains low, and we highlight, at network level, the resilience of our configurable solution, with very limited accuracy degradation of 0.15% on the MNIST task, and 2.84% on the CIFAR-10 task."
                    },
                    {
                        "title": "Electric field control of magnetism",
                        "abstract": "Tuning the Dzyaloshinskii-Moriya interaction (DMI) using electric (E)-fields in magnetic devices has opened up new perspectives for controlling the stabilization of chiral spin structures. Recent efforts have used voltage-induced charge redistribution at magnetic/oxides interfaces to modulate the DMI. This approach is attractive for active devices but tends to be volatile, making it energy-demanding. Here we demonstrate nonvolatile E-field manipulation of the DMI by ionic-liquid gating of Pt/Co/HfO2 ultra thin films. The E-field effect on the DMI is linked to the migration of oxygen species from the HfO2 layer into the Co and Pt layers and subsequent anchoring. This effect permanently changes the properties of the material, showing that E-fields can be used not only for local gating in devices but also as a material design tool for post growth tuning of the DMI."
                    },
                    {
                        "title": "Implementation of Ternary Weights With Resistive RAM Using a Single Sense Operation Per Synapse",
                        "abstract": "The design of systems implementing low precision neural networks with emerging memories such as resistive random access memory (RRAM) is a significant lead for reducing the energy consumption of artificial intelligence. To achieve maximum energy efficiency in such systems, logic and memory should be integrated as tightly as possible. In this work, we focus on the case of ternary neural networks, where synaptic weights assume ternary values. We propose a two-transistor/two-resistor memory architecture employing a precharge sense amplifier, where the weight value can be extracted in a single sense operation. Based on experimental measurements on a hybrid 130 nm CMOS/RRAM chip featuring this sense amplifier, we show that this technique is particularly appropriate at low supply voltage, and that it is resilient to process, voltage, and temperature variations. We characterize the bit error rate in our scheme. We show based on neural network simulation on the CIFAR-10 image recognition task that the use of ternary neural networks significantly increases neural network performance, with regards to binary ones, which are often preferred for inference hardware. We finally evidence that the neural network is immune to the type of bit errors observed in our scheme, which can therefore be used without error correction."
                    },
                    {
                        "title": "Scaling Equilibrium Propagation to Deep ConvNets by Drastically Reducing Its Gradient Estimator Bias",
                        "abstract": "Equilibrium Propagation is a biologically-inspired algorithm that trains convergent recurrent neural networks with a local learning rule. This approach constitutes a major lead to allow learning-capable neuromophic systems and comes with strong theoretical guarantees. Equilibrium propagation operates in two phases, during which the network is let to evolve freely and then \u201cnudged\u201d toward a target; the weights of the network are then updated based solely on the states of the neurons that they connect. The weight updates of Equilibrium Propagation have been shown mathematically to approach those provided by Backpropagation Through Time (BPTT), the mainstream approach to train recurrent neural networks, when nudging is performed with infinitely small strength. In practice, however, the standard implementation of Equilibrium Propagation does not scale to visual tasks harder than MNIST. In this work, we show that a bias in the gradient estimate of equilibrium propagation, inherent in the use of finite nudging, is responsible for this phenomenon and that canceling it allows training deep convolutional neural networks. We show that this bias can be greatly reduced by using symmetric nudging (a positive nudging and a negative one). We also generalize Equilibrium Propagation to the case of cross-entropy loss (by opposition to squared error). As a result of these advances, we are able to achieve a test error of 11.7% on CIFAR-10, which approaches the one achieved by BPTT and provides a major improvement with respect to the standard Equilibrium Propagation that gives 86% test error. We also apply these techniques to train an architecture with unidirectional forward and backward connections, yielding a 13.2% test error. These results highlight equilibrium propagation as a compelling biologically-plausible approach to compute error gradients in deep neuromorphic systems."
                    },
                    {
                        "title": "Low Power In-Memory Implementation of Ternary Neural Networks with Resistive RAM-Based Synapse",
                        "abstract": "The design of systems implementing low precision neural networks with emerging memories such as resistive random access memory (RRAM) is a major lead for reducing the energy consumption of artificial intelligence (AI). Multiple works have for example proposed in-memory architectures to implement low power binarized neural networks. These simple neural networks, where synaptic weights and neuronal activations assume binary values, can indeed approach state-of-the-art performance on vision tasks. In this work, we revisit one of these architectures where synapses are implemented in a differential fashion to reduce bit errors, and synaptic weights are read using precharge sense amplifiers. Based on experimental measurements on a hybrid 130 nm CMOS/RRAM chip and on circuit simulation, we show that the same memory array architecture can be used to implement ternary weights instead of binary weights, and that this technique is particularly appropriate if the sense amplifier is operated in near-threshold regime. We also show based on neural network simulation on the CIFAR-10 image recognition task that going from binary to ternary neural networks significantly increases neural network performance. These results highlight that AI circuits function may sometimes be revisited when operated in low power regimes."
                    },
                    {
                        "title": "Nonvolatile Ionic Modification of the Dzyaloshinskii-Moriya Interaction",
                        "abstract": "The possibility to tune the Dzyaloshinskii Moriya interaction (DMI) by electric (E) field gating in ultra-thin magnetic materials has opened new perspectives in terms of controlling the stabilization of chiral spin structures. Most recent efforts have used voltage-induced charge redistribution at the interface between a metal and an oxide to modulate DMI. This approach is attractive for active devices but it tends to be volatile, making it energy demanding, and it is limited by Coulomb screening in the metal. Here we have demonstrated the non-volatile E-field manipulation of DMI by ionic liquid gating of Pt/Co/HfO2 ultra-thin films. The E-field effect on DMI scales with the E-field exposure time and is proposed to be linked to the migration and subsequent anchoring of oxygen species from the HfO2 layer into the Co and Pt layers. This effect permanently changes the properties of the material showing that E-fields can not only be used for local gating in devices but also as a highly scalable materials design tool for post-growth tuning of DMI."
                    }
                ]
            },
            "01ecfedb-1515-45d4-aee4-04cd0cde53c0": {
                "pk": "01ecfedb-1515-45d4-aee4-04cd0cde53c0",
                "name": "Friedemann Zenke",
                "collaborators": [
                    "G. Bokoch",
                    "W. Gerstner",
                    "T. Vogels",
                    "Henning Sprekeler",
                    "Axel Laborieux",
                    "C. Clopath",
                    "Charles C. King",
                    "B. Bohl",
                    "A. Newton",
                    "B. Hemmings",
                    "Manu Srinath Halvagal",
                    "Julian Rossbroich",
                    "Julia Gygax",
                    "Y. Wu",
                    "Benjamin Cramer",
                    "Sebastian Billaudelle",
                    "Simeon Kanya",
                    "Aron Leibfried",
                    "Andreas Grubl",
                    "V. Karasenko",
                    "Christian Pehle",
                    "Korbinian Schreiber",
                    "Yannik Stradmann",
                    "Johannes Weis",
                    "J. Schemmel",
                    "Henry L\u00fctcke",
                    "Felipe Gerhard",
                    "F. Helmchen",
                    "R. Froemke",
                    "N. Doyon",
                    "M. Gilson",
                    "J. Haas",
                    "R. Liu",
                    "A. Maffei",
                    "P. Miller",
                    "C. Wierenga",
                    "M. Woodin",
                    "E. Gardiner",
                    "C. King",
                    "P. Dawson",
                    "E. Dutil",
                    "R. H. Daniels",
                    "L. Kapp",
                    "K. Breunig"
                ],
                "domain": [
                    "Neural Networks",
                    "Spiking Neural Networks",
                    "Synaptic Plasticity",
                    "Neuromorphic Computing"
                ],
                "publications": [
                    {
                        "title": "An eigenspace view reveals how predictor networks and stop-grads provide implicit variance regularization",
                        "abstract": "Self-supervised learning (SSL) learns useful representations from unlabelled data by training networks to be invariant to pairs of augmented versions of the same input. Non-contrastive methods avoid collapse either by directly regularizing the covariance matrix of network outputs or through asymmetric loss architectures, two seemingly unrelated approaches. Here, by building on DirectPred [1], we lay out a theoretical framework that reconciles these two views. We derive analytical expressions for the representational learning dynamics in linear networks. By expressing them in the eigenspace of the embedding covariance matrix, where the solutions decouple, we reveal the mechanism and conditions that provide implicit variance regularization. These insights allow us to formulate a new isotropic loss function that equalizes eigenvalue contribution and renders learning more robust. Finally, we show empirically that our findings translate in nonlinear networks trained on CIFAR-10 and STL-10."
                    },
                    {
                        "title": "Fluctuation-driven initialization for spiking neural network training",
                        "abstract": "Spiking neural networks (SNNs) underlie low-power, fault-tolerant information processing in the brain and could constitute a power-efficient alternative to conventional deep neural networks when implemented on suitable neuromorphic hardware accelerators. However, instantiating SNNs that solve complex computational tasks in-silico remains a significant challenge. Surrogate gradient (SG) techniques have emerged as a standard solution for training SNNs end-to-end. Still, their success depends on synaptic weight initialization, similar to conventional artificial neural networks (ANNs). Yet, unlike in the case of ANNs, it remains elusive what constitutes a good initial state for an SNN. Here, we develop a general initialization strategy for SNNs inspired by the fluctuation-driven regime commonly observed in the brain. Specifically, we derive practical solutions for data-dependent weight initialization that ensure fluctuation-driven firing in the widely used leaky integrate-and-fire neurons. We empirically show that SNNs initialized following our strategy exhibit superior learning performance when trained with SGs. These findings generalize across several datasets and SNN architectures, including fully connected, deep convolutional, recurrent, and more biologically plausible SNNs obeying Dale\u2019s law. Thus fluctuation-driven initialization provides a practical, versatile, and easy-to-implement strategy for improving SNN training performance on diverse tasks in neuromorphic engineering and computational neuroscience."
                    },
                    {
                        "title": "Holomorphic Equilibrium Propagation Computes Exact Gradients Through Finite Size Oscillations",
                        "abstract": "Equilibrium propagation (EP) is an alternative to backpropagation (BP) that allows the training of deep neural networks with local learning rules. It thus provides a compelling framework for training neuromorphic systems and understanding learning in neurobiology. However, EP requires infinitesimal teaching signals, thereby limiting its applicability in noisy physical systems. Moreover, the algorithm requires separate temporal phases and has not been applied to large-scale problems. Here we address these issues by extending EP to holomorphic networks. We show analytically that this extension naturally leads to exact gradients even for finite-amplitude teaching signals. Importantly, the gradient can be computed as the first Fourier coefficient from finite neuronal activity oscillations in continuous time without requiring separate phases. Further, we demonstrate in numerical simulations that our approach permits robust estimation of gradients in the presence of noise and that deeper models benefit from the finite teaching signals. Finally, we establish the first benchmark for EP on the ImageNet 32x32 dataset and show that it matches the performance of an equivalent network trained with BP. Our work provides analytical insights that enable scaling EP to large-scale problems and establishes a formal framework for how oscillations could support learning in biological and neuromorphic systems."
                    },
                    {
                        "title": "Nonlinear transient amplification in recurrent neural networks with short-term plasticity",
                        "abstract": "To rapidly process information, neural circuits have to amplify specific activity patterns transiently. How the brain performs this nonlinear operation remains elusive. Hebbian assemblies are one possibility whereby symmetric excitatory connections boost neuronal activity. However, such Hebbian amplification is often associated with dynamical slowing of network dynamics, non-transient attractor states, and pathological run-away activity. Feedback inhibition can alleviate these effects but typically linearizes responses and reduces amplification gain. At the same time, other alternative mechanisms rely on asymmetric connectivity, in conflict with the Hebbian doctrine. Here we propose nonlinear transient amplification (NTA), a plausible circuit mechanism that reconciles symmetric connectivity with rapid amplification while avoiding the above issues. NTA has two distinct temporal phases. Initially, positive feedback excitation selectively amplifies inputs that exceed a critical threshold. Subsequently, short-term plasticity quenches the run-away dynamics into an inhibition-stabilized network state. By characterizing NTA in supralinear network models, we establish that the resulting onset transients are stimulus selective and well-suited for speedy information processing. Further, we find that excitatory-inhibitory co-tuning widens the parameter regime in which NTA is possible. In summary, NTA provides a parsimonious explanation for how excitatory-inhibitory co-tuning and short-term plasticity collaborate in recurrent networks to achieve transient amplification."
                    },
                    {
                        "title": "Surrogate gradients for analog neuromorphic computing",
                        "abstract": "Significance Neuromorphic systems aim to accomplish efficient computation in electronics by mirroring neurobiological principles. Taking advantage of neuromorphic technologies requires effective learning algorithms capable of instantiating high-performing neural networks, while also dealing with inevitable manufacturing variations of individual components, such as memristors or analog neurons. We present a learning framework resulting in bioinspired spiking neural networks with high performance, low inference latency, and sparse spike-coding schemes, which also self-corrects for device mismatch. We validate our approach on the BrainScaleS-2 analog spiking neuromorphic system, demonstrating state-of-the-art accuracy, low latency, and energy efficiency. Our work sketches a path for building powerful neuromorphic processors that take advantage of emerging analog technologies. To rapidly process temporal information at a low metabolic cost, biological neurons integrate inputs as an analog sum, but communicate with spikes, binary events in time. Analog neuromorphic hardware uses the same principles to emulate spiking neural networks with exceptional energy efficiency. However, instantiating high-performing spiking networks on such hardware remains a significant challenge due to device mismatch and the lack of efficient training algorithms. Surrogate gradient learning has emerged as a promising training strategy for spiking networks, but its applicability for analog neuromorphic systems has not been demonstrated. Here, we demonstrate surrogate gradient learning on the BrainScaleS-2 analog neuromorphic system using an in-the-loop approach. We show that learning self-corrects for device mismatch, resulting in competitive spiking network performance on both vision and speech benchmarks. Our networks display sparse spiking activity with, on average, less than one spike per hidden neuron and input, perform inference at rates of up to 85,000 frames per second, and consume less than 200 mW. In summary, our work sets several benchmarks for low-energy spiking network processing on analog neuromorphic hardware and paves the way for future on-chip learning algorithms."
                    },
                    {
                        "title": "Memory formation and recall in recurrent spiking neural networks",
                        "abstract": "Our brain has the capacity to analyze a visual scene in a split second, to learn how to play an instrument, and to remember events, faces and concepts. Neurons underlie all of these diverse functions. Neurons, cells within the brain that generate and transmit electrical activity, communicate with each other through chemical synapses. These synaptic connections dynamically change with experience, a process referred to as synaptic plasticity, which is thought to be at the core of the brain's ability to learn and process the world in sophisticated ways. Our understanding of the rules of synaptic plasticity remains quite limited. To enable efficient computations among neurons or to serve as a trace of memory, synapses must create stable connectivity patterns between neurons. However there remains an insufficient theoretical explanation as to how stable connectivity patterns can form in the presence of synaptic plasticity. Since the dynamics of recurrently connected neurons depend upon their connections, which themselves change in response to the network dynamics, synaptic plasticity and network dynamics have to be treated as a compound system. Due to the nonlinear nature of the system this can be analytically challenging. Utilizing network simulations that model the interplay between the network connectivity and synaptic plasticity can provide valuable insights. However, many existing network models that implement biologically relevant forms of plasticity become unstable. This suggests that current models do not accurately describe the biological networks, which have no difficulty functioning without succumbing to exploding network activity. The instability in these network simulations could originate from the fact that theoretical studies have, almost exclusively, focused on Hebbian plasticity at excitatory synapses. Hebbian plasticity causes connected neurons that are active together to increase the connection strength between them. Biological networks, however, display a large variety of different forms of synaptic plasticity and homeostatic mechanisms, beyond Hebbian plasticity. Furthermore, inhibitory cells can undergo synaptic plasticity as well. These diverse forms of plasticity are active at the same time, and our understanding of the computational role of most of these synaptic dynamics remains elusive. This raises the important question as to whether forms of plasticity that have not been previously considered could -in combination with Hebbian plasticity- lead to stable network dynamics. Here we illustrate that by combining multiple forms of plasticity with distinct roles, a recurrently connected spiking network model self-organizes to distinguish and extract multiple overlapping external stimuli. Moreover we show that the acquired network structures remain stable over hours while plasticity is active. This long-term stability allows the network to function as an associative memory and to correctly classify distorted or partially cued stimuli. During intervals in which no stimulus is shown the network dynamically remembers the last stimulus as selective delay activity. Taken together this work suggest that multiple forms of plasticity and homeostasis on different timescales have to work together to create stable connectivity patterns in neuronal networks which enable them to perform relevant computation."
                    },
                    {
                        "title": "Inference of neuronal network spike dynamics and topology from calcium imaging data",
                        "abstract": "Two-photon calcium imaging enables functional analysis of neuronal circuits by inferring action potential (AP) occurrence (\u201cspike trains\u201d) from cellular fluorescence signals. It remains unclear how experimental parameters such as signal-to-noise ratio (SNR) and acquisition rate affect spike inference and whether additional information about network structure can be extracted. Here we present a simulation framework for quantitatively assessing how well spike dynamics and network topology can be inferred from noisy calcium imaging data. For simulated AP-evoked calcium transients in neocortical pyramidal cells, we analyzed the quality of spike inference as a function of SNR and data acquisition rate using a recently introduced peeling algorithm. Given experimentally attainable values of SNR and acquisition rate, neural spike trains could be reconstructed accurately and with up to millisecond precision. We then applied statistical neuronal network models to explore how remaining uncertainties in spike inference affect estimates of network connectivity and topological features of network organization. We define the experimental conditions suitable for inferring whether the network has a scale-free structure and determine how well hub neurons can be identified. Our findings provide a benchmark for future calcium imaging studies that aim to reliably infer neuronal network properties."
                    },
                    {
                        "title": "Inhibitory synaptic plasticity: spike timing-dependence and putative network function",
                        "abstract": "While the plasticity of excitatory synaptic connections in the brain has been widely studied, the plasticity of inhibitory connections is much less understood. Here, we present recent experimental and theoretical findings concerning the rules of spike timing-dependent inhibitory plasticity and their putative network function. This is a summary of a workshop at the COSYNE conference 2012."
                    },
                    {
                        "title": "Inhibitory Plasticity Balances Excitation and Inhibition in Sensory Pathways and Memory Networks",
                        "abstract": "Plasticity at inhibitory synapses maintains balanced excitatory and inhibitory synaptic inputs at cortical neurons. Cortical neurons receive balanced excitatory and inhibitory synaptic currents. Such a balance could be established and maintained in an experience-dependent manner by synaptic plasticity at inhibitory synapses. We show that this mechanism provides an explanation for the sparse firing patterns observed in response to natural stimuli and fits well with a recently observed interaction of excitatory and inhibitory receptive field plasticity. The introduction of inhibitory plasticity in suitable recurrent networks provides a homeostatic mechanism that leads to asynchronous irregular network states. Further, it can accommodate synaptic memories with activity patterns that become indiscernible from the background state but can be reactivated by external stimuli. Our results suggest an essential role of inhibitory plasticity in the formation and maintenance of functional cortical circuitry."
                    },
                    {
                        "title": "Inhibitory plasticity balances excitation and inhibition in sensory processing and Hebbian assemblies",
                        "abstract": "Cortical neurons receive balanced excitatory and inhibitory membrane currents. Here, we show that such a balance can be established and maintained in an experiencedependent manner by synaptic plasticity at inhibitory synapses. The mechanism we put forward provides an explanation for the sparse firing patterns observed in response to natural stimuli and fits well with a recently observed interaction of excitatory and inhibitory receptive field plasticity. We show that the introduction of inhibitory plasticity in suitable recurrent networks provides a homeostatic mechanism that leads to asynchronous irregular network states. Further, it can accommodate synaptic memories with activity patterns that become indiscernible from the background state, but can be re-activated by external stimuli. Our results suggest an essential role of inhibitory plasticity in the formation and maintenance of functional cortical circuitry."
                    },
                    {
                        "title": "p21-activated Kinase (PAK1) Is Phosphorylated and Activated by 3-Phosphoinositide-dependent Kinase-1 (PDK1)*",
                        "abstract": "In this study, we show that phosphorylated 3-phosphoinositide-dependent kinase 1 (PDK1) phosphorylates p21-activated kinase 1 (PAK1) in the presence of sphingosine. We identify threonine 423, a conserved threonine in the activation loop of kinase subdomain VIII, as the PDK1 phosphorylation site on PAK1. Threonine 423 is a previously identified PAK1 autophosphorylation site that lies within a PAK consensus phosphorylation sequence. After pretreatment with phosphatases, autophosphorylation of PAK1 occurred at all major sites except threonine 423. A phosphothreonine 423-specific antibody detected phosphorylation of recombinant, catalytically inactive PAK1 after incubation with wild-type PAK1, indicating phosphorylation of threonine 423 occurs by an intermolecular mechanism. The biological significance of PDK1 phosphorylation of PAK1 at threonine 423 in vitro is supported by the observation that these two proteins interact in vivo and that PDK1-phosphorylated PAK1 has an increased activity toward substrate. An increase of phosphorylation of catalytically inactive PAK1 was observed in COS-7 cells expressing wild-type, but not catalytically inactive, PDK1 upon elevation of intracellular sphingosine levels. PDK1 phosphorylation of PAK1 was not blocked by pretreatment with wortmannin or when PDK1 was mutated to prevent phosphatidylinositol binding, indicating this process is independent of phosphatidylinositol 3-kinase activity. The data presented here provide evidence for a novel mechanism for PAK1 regulation and activation."
                    },
                    {
                        "title": "Sphingosine Is a Novel Activator of 3-Phosphoinositide-dependent Kinase 1*",
                        "abstract": "3-Phosphoinositide-dependent kinase 1 (PDK1) has previously been shown to phosphorylate the activation loop of several AGC kinase family members. In this study, we show that p21-activated kinase 1, the activity of which is regulated by the GTP-bound form of Cdc42 and Rac and by sphingosine, is phosphorylated by PDK1. Phosphorylation of p21-activated kinase 1 by PDK1 occurred only in the presence of sphingosine, which increased PDK1 autophosphorylation 25-fold. Sphingosine increased PDK1 autophosphorylation in a concentration-dependent manner and significantly increased phosphate incorporation into known PDK1 substrates. Studies on the lipid requirement for PDK1 activation found that both sphingosine isoforms and stearylamine also increased PDK1 autophosphorylation. However, C10-sphingosine, octylamine, and stearic acid were unable to increase PDK1 autophosphorylation, indicating that both a positive charge and a lipid tail containing at least a C10-carbon backbone were required for PDK1 activation. Three PDK1 autophosphorylation sites were identified after stimulation with sphingosine in a serine-rich region located between the kinase domain and the pleckstrin homology domain using two-dimensional phosphopeptide maps and matrix assisted laser desorption/ionization mass spectroscopy. Increased phosphorylation of endogenous Akt at threonine 308 was observed in COS-7 cells expressing wild type PDK1, but not catalytically inactive PDK1, when cellular sphingosine levels were elevated by treatment with sphingomyelinase. Sphingosine thus appears to be a true PDK1 activator."
                    },
                    {
                        "title": "Identification of a Central Phosphorylation Site in p21-activated Kinase Regulating Autoinhibition and Kinase Activity*",
                        "abstract": "p21-activated kinases (Pak)/Ste20 kinases are regulated in vitro and in vivo by the small GTP-binding proteins Rac and Cdc42 and lipids, such as sphingosine, which stimulate autophosphorylation and phosphorylation of exogenous substrates. The mechanism of Pak activation by these agents remains unclear. We investigated Pak kinase activation in more detail to gain insight into the interplay between the GTPase/sphingosine binding, an intramolecular inhibitory interaction, and autophosphorylation. We present biochemical evidence that an autoinhibitory domain (ID) contained within amino acid residues 67\u2013150 of Pak1 interacts with the carboxyl-terminal kinase domain and that this interaction is regulated in a GTPase-dependent fashion. Cdc42- and sphingosine-stimulated Pak1 activity can be inhibited intrans by recombinant ID peptide, indicating similarities in their mode of activation. However, Pak1, which was autophosphorylated in response to either GTPase or sphingosine, is highly active and is insensitive to inhibition by the ID peptide. We identified phospho-acceptor site threonine 423 in the kinase activation loop as a critical determinant for the sensitivity to autoinhibition and enzymatic activity. Phosphorylation studies suggested that the stimulatory effect of both GTPase and sphingosine results in exposure of the activation loop, making it accessible for intermolecular phosphorylation."
                    },
                    {
                        "title": "\u03b1Pix Stimulates p21-activated Kinase Activity through Exchange Factor-dependent and -independent Mechanisms*",
                        "abstract": "Activation of p21-activated kinases (Paks) is achieved through binding of the GTPases Rac or Cdc42 to a conserved domain in the N-terminal regulatory region of Pak. Additional signaling components are also likely to be important in regulating Pak activation. Recently, a family of Pak-interacting guanine nucleotide exchange factors (Pix) have been identified and which are good candidates for regulating Pak activity. Using an active, truncated form of \u03b1Pix (amino acids 155\u2013545), we observe stimulation of Pak1 kinase activity when \u03b1Pix155\u2013545 is co-expressed with Cdc42 and wild-type Pak1 in COS-1 cells. This activation does not occur when we co-express a Pak1 mutant unable to bind \u03b1Pix. The activation of wild-type Pak1 by \u03b1Pix155\u2013545 also requires that \u03b1Pix155\u2013545 retain functional exchange factor activity. However, the Pak1H83,86L mutant that does not bind Rac or Cdc42 is activated in the absence of GTPase by \u03b1Pix155\u2013545 and by a mutant of \u03b1Pix155\u2013545 that no longer has exchange factor activity. Pak1 activity stimulated in vitro using GTP\u03b3S-loaded Cdc42 was also enhanced by recombinant \u03b1Pix155\u2013545 in a binding-dependent manner. These data suggest that Pak activity can be modulated by physical interaction with \u03b1Pix and that this specific effect involves both exchange factor-dependent and -independent mechanisms."
                    },
                    {
                        "title": "Regulated Phosphorylation of the Gal4p Inhibitor Gal80p of Kluyveromyces lactis Revealed by Mutational Analysis",
                        "abstract": "Abstract The yeast Gal80 protein inhibits the transcription activation function of Gal4p by physically interacting with the activation domain (Gal4-AD). Gal80p interaction with Gal1p or Gal3p is required to relieve Gal4p inhibition in response to galactose. Gal80p orthologs of Saccharomyces cerevisiae and Kluyveromyces lactis, ScGal80p and KlGal80p, can also inhibit the heterologous Gal4p variants; however, heterologous Gal3p/Gal1p only regulate ScGal80p but not KlGal80p. To compare KlGal80p and ScGal80p, point mutations known to affect ScGal80p function were introduced at corresponding positions in KlGal80p, and Gal4p regulation in vivo and KlGal80p-binding to Gst-Gal1p and Gst-Gal4-AD in vitro were analysed. The in vitro binding properties of the KlGal80p mutants were similar to those of ScGal80p, but two out of four mutants differed in Gal4p regulation. E. g. KlGAL80s-0 (G302R) but not ScGAL80s-0 (G301R) alleviates Gal4p inhibition. Possibly, this difference is related to a role of phosphorylation in the regulation of Gal80p function in K. lactis. Wild-type and mutant forms of KlGal80p are shown to be subject to carbon source regulated phosphorylation whereas no evidence for ScGal80p phosphorylation exists. (Hyper-)phosphorylation of KlGal80p is strongly reduced in galactose-containing medium. This reduction requires KlGal1p but no interaction with KlGal4p. The inhibition deficient KlGal80s-0p (G302R) variant is under-phosphorylated. We thus propose that phosphorylation of Gal80p in Kluyveromyces lactis contributes to the regulation of Gal4p mediated transcription."
                    }
                ]
            }
        }
    },
    "2406.01793": {
        "paper_data": {
            "title": "Towards the Transferability of Rewards Recovered via Regularized Inverse Reinforcement Learning",
            "url": "http://arxiv.org/abs/2406.01793v1",
            "arxiv_id": "2406.01793",
            "authors": [
                "Andreas Schlaginhaufen",
                "Maryam Kamgarpour"
            ],
            "abstract": "Inverse reinforcement learning (IRL) aims to infer a reward from expert demonstrations, motivated by the idea that the reward, rather than the policy, is the most succinct and transferable description of a task [Ng et al., 2000]. However, the reward corresponding to an optimal policy is not unique, making it unclear if an IRL-learned reward is transferable to new transition laws in the sense that its optimal policy aligns with the optimal policy corresponding to the expert's true reward. Past work has addressed this problem only under the assumption of full access to the expert's policy, guaranteeing transferability when learning from two experts with the same reward but different transition laws that satisfy a specific rank condition [Rolland et al., 2022]. In this work, we show that the conditions developed under full access to the expert's policy cannot guarantee transferability in the more practical scenario where we have access only to demonstrations of the expert. Instead of a binary rank condition, we propose principal angles as a more refined measure of similarity and dissimilarity between transition laws. Based on this, we then establish two key results: 1) a sufficient condition for transferability to any transition laws when learning from at least two experts with sufficiently different transition laws, and 2) a sufficient condition for transferability to local changes in the transition law when learning from a single expert. Furthermore, we also provide a probably approximately correct (PAC) algorithm and an end-to-end analysis for learning transferable rewards from demonstrations of multiple experts.",
            "introduction": "   1 Introduction  Reinforcement learning (RL) has achieved remarkable success in various domains such as robotics (Hwangbo et\u00a0al., 2019), autonomous driving (Lu et\u00a0al., 2023), or fine-tuning of large language models (Stiennon et\u00a0al., 2020). Despite these advances, a key challenge lies in designing appropriate reward functions that reflect the desired outcomes and align with human values. Misaligned rewards can lead to suboptimal behaviors (Ngo et\u00a0al., 2022), undermining the potential benefits of RL in practical scenarios. Inverse reinforcement learning (IRL), also known as inverse optimal control (Kalman, 1964) or structural estimation (Rust, 1994), addresses this problem by inferring a reward from demonstrations of an expert acting optimally in a Markov decision process (MDP).   Compared to behavioral cloning (Pomerleau, 1988), which directly fits a policy to the expert\u2019s demonstrations, IRL is believed to provide a more transferable description of the expert\u2019s task (Ng et\u00a0al., 2000), as recovering the expert\u2019s underlying reward would enable us to train a policy in a new environment with different dynamics. However, it is also known that the reward corresponding to some optimal policy is not unique (Ng et\u00a0al., 1999), making it difficult to recover the expert\u2019s true underlying reward. This raises the question: Is a reward recovered via IRL transferable to a new environment in the sense that its optimal policy aligns with the expert\u2019s true reward? For example, in autonomous driving, could we effectively reuse a reward learned from demonstrations of one car in a given city to train or fine-tune a policy for another car in another city?   Ensuring transferability is challenging, as neither the optimal policy corresponding to a reward nor the reward corresponding to an optimal policy is unique. This leads to trivial solutions to the IRL problem, such as constant rewards that make all policies optimal. Common approaches to address this challenge include characterizing the entire set of rewards for which the expert is optimal (Metelli et\u00a0al., 2021), or assuming the expert is optimal with respect to an entropy regularized RL problem (Ziebart, 2010), leading to many popular IRL and imitation learning algorithms (Ho and Ermon, 2016; Fu et\u00a0al., 2017; Garg et\u00a0al., 2021). Entropy regularization results in a unique and more uniform optimal policy, serving as a model for the expert\u2019s bounded rationality (Ortega et\u00a0al., 2015).   In the entropy-regularized setting, several recent works study the set of rewards for which a given expert policy is optimal. In particular, Cao et\u00a0al. (2021); Skalse et\u00a0al. (2023) show that under entropy regularization, the expert\u2019s reward can be identified up to so-called potential shaping transformations (Ng et\u00a0al., 1999). The authors of (Schlaginhaufen and Kamgarpour, 2023) extend this result to more general steep regularization. Furthermore, they show that to guarantee transferability to any transition law, the expert\u2019s reward needs to be identified up to a constant. The latter can be achieved either by restricting the reward class, e.g., to state-only rewards (Amin et\u00a0al., 2017), or by learning from multiple experts with the same reward but different transition laws, given that a specific rank condition is satisfied (Cao et\u00a0al., 2021; Rolland et\u00a0al., 2022). However, the above results cannot be applied directly in practice, as they rely",
            "references": [
                {
                    "title": "Learning Nash Equilibria in Zero-Sum Markov Games: A Single Time-scale Algorithm Under Weak Reachability",
                    "abstract": "We consider decentralized learning for zero-sum games, where players only see their payoff information and are agnostic to actions and payoffs of the opponent. Previous works demonstrated convergence to a Nash equilibrium in this setting using double time-scale algorithms under strong reachability assumptions. We address the open problem of achieving an approximate Nash equilibrium efficiently with an uncoupled and single time-scale algorithm under weaker conditions. Our contribution is a rational and convergent algorithm, utilizing Tsallis-entropy regularization in a value-iteration-based approach. The algorithm learns an approximate Nash equilibrium in polynomial time, requiring only the existence of a policy pair that induces an irreducible and aperiodic Markov chain, thus considerably weakening past assumptions. Our analysis leverages negative drift inequalities and introduces novel properties of Tsallis entropy that are of independent interest."
                },
                {
                    "title": "Identifiability and Generalizability in Constrained Inverse Reinforcement Learning",
                    "abstract": "Two main challenges in Reinforcement Learning (RL) are designing appropriate reward functions and ensuring the safety of the learned policy. To address these challenges, we present a theoretical framework for Inverse Reinforcement Learning (IRL) in constrained Markov decision processes. From a convex-analytic perspective, we extend prior results on reward identifiability and generalizability to both the constrained setting and a more general class of regularizations. In particular, we show that identifiability up to potential shaping (Cao et al., 2021) is a consequence of entropy regularization and may generally no longer hold for other regularizations or in the presence of safety constraints. We also show that to ensure generalizability to new transition laws and constraints, the true reward must be identified up to a constant. Additionally, we derive a finite sample guarantee for the suboptimality of the learned rewards, and validate our results in a gridworld environment."
                },
                {
                    "title": "Best of Both Worlds Policy Optimization",
                    "abstract": "Policy optimization methods are popular reinforcement learning algorithms in practice. Recent works have built theoretical foundation for them by proving $\\sqrt{T}$ regret bounds even when the losses are adversarial. Such bounds are tight in the worst case but often overly pessimistic. In this work, we show that in tabular Markov decision processes (MDPs), by properly designing the regularizer, the exploration bonus and the learning rates, one can achieve a more favorable polylog$(T)$ regret when the losses are stochastic, without sacrificing the worst-case guarantee in the adversarial regime. To our knowledge, this is also the first time a gap-dependent polylog$(T)$ regret bound is shown for policy optimization. Specifically, we achieve this by leveraging a Tsallis entropy or a Shannon entropy regularizer in the policy update. Then we show that under known transitions, we can further obtain a first-order regret bound in the adversarial regime by leveraging the log-barrier regularizer."
                },
                {
                    "title": "Imitation Is Not Enough: Robustifying Imitation with Reinforcement Learning for Challenging Driving Scenarios",
                    "abstract": "Imitation learning (IL) is a simple and powerful way to use high-quality human driving data, which can be collected at scale, to produce human-like behavior. However, policies based on imitation learning alone often fail to sufficiently account for safety and reliability concerns. In this paper, we show how imitation learning combined with reinforcement learning using simple rewards can substan-tially improve the safety and reliability of driving policies over those learned from imitation alone. In particular, we train a policy on over lOOk miles of urban driving data, and measure its effectiveness in test scenarios grouped by different levels of collision likelihood. Our analysis shows that while imitation can perform well in low-difficulty scenarios that are well-covered by the demonstration data, our proposed approach significantly improves robustness on the most challenging scenarios (over 38 % reduction in failures). To our knowledge, this is the first application of a combined imitation and reinforcement learning approach in autonomous driving that utilizes large amounts of real- world human driving data."
                },
                {
                    "title": "Identifiability and generalizability from multiple experts in Inverse Reinforcement Learning",
                    "abstract": "While Reinforcement Learning (RL) aims to train an agent from a reward function in a given environment, Inverse Reinforcement Learning (IRL) seeks to recover the reward function from observing an expert's behavior. It is well known that, in general, various reward functions can lead to the same optimal policy, and hence, IRL is ill-defined. However, (Cao et al., 2021) showed that, if we observe two or more experts with different discount factors or acting in different environments, the reward function can under certain conditions be identified up to a constant. This work starts by showing an equivalent identifiability statement from multiple experts in tabular MDPs based on a rank condition, which is easily verifiable and is shown to be also necessary. We then extend our result to various different scenarios, i.e., we characterize reward identifiability in the case where the reward function can be represented as a linear combination of given features, making it more interpretable, or when we have access to approximate transition matrices. Even when the reward is not identifiable, we provide conditions characterizing when data on multiple experts in a given environment allows to generalize and train an optimal agent in a new environment. Our theoretical results on reward identifiability and generalizability are validated in various numerical experiments."
                },
                {
                    "title": "The alignment problem from a deep learning perspective",
                    "abstract": "In coming years or decades, artificial general intelligence (AGI) may surpass human capabilities at many critical tasks. We argue that, without substantial effort to prevent it, AGIs could learn to pursue goals that are in conflict (i.e. misaligned) with human interests. If trained like today's most capable models, AGIs could learn to act deceptively to receive higher reward, learn misaligned internally-represented goals which generalize beyond their fine-tuning distributions, and pursue those goals using power-seeking strategies. We review emerging evidence for these properties. AGIs with these properties would be difficult to align and may appear aligned even when they are not. Finally, we briefly outline how the deployment of misaligned AGIs might irreversibly undermine human control over the world, and we review research directions aimed at preventing this outcome."
                },
                {
                    "title": "Invariance in Policy Optimisation and Partial Identifiability in Reward Learning",
                    "abstract": "It is often very challenging to manually design reward functions for complex, real-world tasks. To solve this, one can instead use reward learning to infer a reward function from data. However, there are often multiple reward functions that fit the data equally well, even in the infinite-data limit. This means that the reward function is only partially identifiable. In this work, we formally characterise the partial identifiability of the reward function given several popular reward learning data sources, including expert demonstrations and trajectory comparisons. We also analyse the impact of this partial identifiability for several downstream tasks, such as policy optimisation. We unify our results in a framework for comparing data sources and downstream tasks by their invariances, with implications for the design and selection of data sources for reward learning."
                },
                {
                    "title": "IQ-Learn: Inverse soft-Q Learning for Imitation",
                    "abstract": "In many sequential decision-making problems (e.g., robotics control, game playing, sequential prediction), human or expert data is available containing useful information about the task. However, imitation learning (IL) from a small amount of expert data can be challenging in high-dimensional environments with complex dynamics. Behavioral cloning is a simple method that is widely used due to its simplicity of implementation and stable convergence but doesn't utilize any information involving the environment's dynamics. Many existing methods that exploit dynamics information are difficult to train in practice due to an adversarial optimization process over reward and policy approximators or biased, high variance gradient estimators. We introduce a method for dynamics-aware IL which avoids adversarial training by learning a single Q-function, implicitly representing both reward and policy. On standard benchmarks, the implicitly learned rewards show a high positive correlation with the ground-truth rewards, illustrating our method can also be used for inverse reinforcement learning (IRL). Our method, Inverse soft-Q learning (IQ-Learn) obtains state-of-the-art results in offline and online imitation learning settings, significantly outperforming existing methods both in the number of required environment interactions and scalability in high-dimensional spaces, often by more than 3x."
                },
                {
                    "title": "Identifiability in inverse reinforcement learning",
                    "abstract": "Inverse reinforcement learning attempts to reconstruct the reward function in a Markov decision problem, using observations of agent actions. As already observed in Russell [1998] the problem is ill-posed, and the reward function is not identifiable, even under the presence of perfect information about optimal behavior. We provide a resolution to this non-identifiability for problems with entropy regularization. For a given environment, we fully characterize the reward functions leading to a given policy and demonstrate that, given demonstrations of actions for the same reward under two distinct discount factors, or under sufficiently different environments, the unobserved reward can be recovered up to a constant. We also give general necessary and sufficient conditions for reconstruction of time-homogeneous rewards on finite horizons, and for action-independent rewards, generalizing recent results of Kim et al. [2021] and Fu et al. [2018]."
                },
                {
                    "title": "Regularized Inverse Reinforcement Learning",
                    "abstract": "Inverse Reinforcement Learning (IRL) aims to facilitate a learner's ability to imitate expert behavior by acquiring reward functions that explain the expert's decisions. Regularized IRL applies strongly convex regularizers to the learner's policy in order to avoid the expert's behavior being rationalized by arbitrary constant rewards, also known as degenerate solutions. We propose tractable solutions, and practical methods to obtain them, for regularized IRL. Current methods are restricted to the maximum-entropy IRL framework, limiting them to Shannon-entropy regularizers, as well as proposing the solutions that are intractable in practice. We present theoretical backing for our proposed IRL method's applicability for both discrete and continuous controls, empirically validating our performance on a variety of tasks."
                },
                {
                    "title": "Learning to summarize from human feedback",
                    "abstract": "As language models become more powerful, training and evaluation are increasingly bottlenecked by the data and metrics used for a particular task. For example, summarization models are often trained to predict human reference summaries and evaluated using ROUGE, but both of these metrics are rough proxies for what we really care about---summary quality. In this work, we show that it is possible to significantly improve summary quality by training a model to optimize for human preferences. We collect a large, high-quality dataset of human comparisons between summaries, train a model to predict the human-preferred summary, and use that model as a reward function to fine-tune a summarization policy using reinforcement learning. We apply our method to a version of the TL;DR dataset of Reddit posts and find that our models significantly outperform both human reference summaries and much larger models fine-tuned with supervised learning alone. Our models also transfer to CNN/DM news articles, producing summaries nearly as good as the human reference without any news-specific fine-tuning. We conduct extensive analyses to understand our human feedback dataset and fine-tuned models We establish that our reward model generalizes to new datasets, and that optimizing our reward model results in better summaries than optimizing ROUGE according to humans. We hope the evidence from our paper motivates machine learning researchers to pay closer attention to how their training loss affects the model behavior they actually want."
                },
                {
                    "title": "On the Global Convergence Rates of Softmax Policy Gradient Methods",
                    "abstract": "We make three contributions toward better understanding policy gradient methods in the tabular setting. First, we show that with the true gradient, policy gradient with a softmax parametrization converges at a $O(1/t)$ rate, with constants depending on the problem and initialization. This result significantly expands the recent asymptotic convergence results. The analysis relies on two findings: that the softmax policy gradient satisfies a \u0141ojasiewicz inequality, and the minimum probability of an optimal action during optimization can be bounded in terms of its initial value. Second, we analyze entropy regularized policy gradient and show that it enjoys a significantly faster linear convergence rate $O(e^{-t})$ toward softmax optimal policy. This result resolves an open question in the recent literature. Finally, combining the above two results and additional new $\\Omega(1/t)$ lower bound results, we explain how entropy regularization improves policy optimization, even with the true gradient, from the perspective of convergence rate. The separation of rates is further explained using the notion of non-uniform \u0141ojasiewicz degree. These results provide a theoretical understanding of the impact of entropy and corroborate existing empirical studies."
                },
                {
                    "title": "A Theory of Regularized Markov Decision Processes",
                    "abstract": "Many recent successful (deep) reinforcement learning algorithms make use of regularization, generally based on entropy or Kullback-Leibler divergence. We propose a general theory of regularized Markov Decision Processes that generalizes these approaches in two directions: we consider a larger class of regularizers, and we consider the general modified policy iteration approach, encompassing both policy iteration and value iteration. The core building blocks of this theory are a notion of regularized Bellman operator and the Legendre-Fenchel transform, a classical tool of convex optimization. This approach allows for error propagation analyses of general algorithmic schemes of which (possibly variants of) classical algorithms such as Trust Region Policy Optimization, Soft Q-learning, Stochastic Actor Critic or Dynamic Policy Programming are special cases. This also draws connections to proximal convex optimization, especially to Mirror Descent."
                },
                {
                    "title": "Learning agile and dynamic motor skills for legged robots",
                    "abstract": "A method for learning agile control policies uses simulated data to enable precise, efficient movements in a complex physical robot. Legged robots pose one of the greatest challenges in robotics. Dynamic and agile maneuvers of animals cannot be imitated by existing methods that are crafted by humans. A compelling alternative is reinforcement learning, which requires minimal craftsmanship and promotes the natural evolution of a control policy. However, so far, reinforcement learning research for legged robots is mainly limited to simulation, and only few and comparably simple examples have been deployed on real systems. The primary reason is that training with real robots, particularly with dynamically balancing systems, is complicated and expensive. In the present work, we introduce a method for training a neural network policy in simulation and transferring it to a state-of-the-art legged system, thereby leveraging fast, automated, and cost-effective data generation schemes. The approach is applied to the ANYmal robot, a sophisticated medium-dog\u2013sized quadrupedal system. Using policies trained in simulation, the quadrupedal machine achieves locomotion skills that go beyond what had been achieved with prior methods: ANYmal is capable of precisely and energy-efficiently following high-level body velocity commands, running faster than before, and recovering from falling even in complex configurations."
                },
                {
                    "title": "Tsallis-INF: An Optimal Algorithm for Stochastic and Adversarial Bandits",
                    "abstract": "We derive an algorithm that achieves the optimal (within constants) pseudo-regret in both adversarial and stochastic multi-armed bandits without prior knowledge of the regime and time horizon. The algorithm is based on online mirror descent (OMD) with Tsallis entropy regularization with power $\\alpha=1/2$ and reduced-variance loss estimators. More generally, we define an adversarial regime with a self-bounding constraint, which includes stochastic regime, stochastically constrained adversarial regime (Wei and Luo), and stochastic regime with adversarial corruptions (Lykouris et al.) as special cases, and show that the algorithm achieves logarithmic regret guarantee in this regime and all of its special cases simultaneously with the adversarial regret guarantee.} The algorithm also achieves adversarial and stochastic optimality in the utility-based dueling bandit setting. We provide empirical evaluation of the algorithm demonstrating that it significantly outperforms UCB1 and EXP3 in stochastic environments. We also provide examples of adversarial environments, where UCB1 and Thompson Sampling exhibit almost linear regret, whereas our algorithm suffers only logarithmic regret. To the best of our knowledge, this is the first example demonstrating vulnerability of Thompson Sampling in adversarial environments. Last, but not least, we present a general stochastic analysis and a general adversarial analysis of OMD algorithms with Tsallis entropy regularization for $\\alpha\\in[0,1]$ and explain the reason why $\\alpha=1/2$ works best."
                },
                {
                    "title": "Learning Robust Rewards with Adversarial Inverse Reinforcement Learning",
                    "abstract": "Reinforcement learning provides a powerful and general framework for decision making and control, but its application in practice is often hindered by the need for extensive feature and reward engineering. Deep reinforcement learning methods can remove the need for explicit engineering of policy or value features, but still require a manually specified reward function. Inverse reinforcement learning holds the promise of automatic reward acquisition, but has proven exceptionally difficult to apply to large, high-dimensional problems with unknown dynamics. In this work, we propose adverserial inverse reinforcement learning (AIRL), a practical and scalable inverse reinforcement learning algorithm based on an adversarial reward learning formulation. We demonstrate that AIRL is able to recover reward functions that are robust to changes in dynamics, enabling us to learn policies even under significant variation in the environment seen during training. Our experiments show that AIRL greatly outperforms prior methods in these transfer settings."
                },
                {
                    "title": "Repeated Inverse Reinforcement Learning",
                    "abstract": "We introduce a novel repeated Inverse Reinforcement Learning problem: the agent has to act on behalf of a human in a sequence of tasks and wishes to minimize the number of tasks that it surprises the human by acting suboptimally with respect to how the human would have acted. Each time the human is surprised, the agent is provided a demonstration of the desired behavior by the human. We formalize this problem, including how the sequence of tasks is chosen, in a few different ways and provide some foundational results."
                },
                {
                    "title": "Generative Adversarial Imitation Learning",
                    "abstract": "Consider learning a policy from example expert behavior, without interaction with the expert or access to reinforcement signal. One approach is to recover the expert's cost function with inverse reinforcement learning, then extract a policy from that cost function with reinforcement learning. This approach is indirect and can be slow. We propose a new general framework for directly extracting a policy from data, as if it were obtained by reinforcement learning following inverse reinforcement learning. We show that a certain instantiation of our framework draws an analogy between imitation learning and generative adversarial networks, from which we derive a model-free imitation learning algorithm that obtains significant performance gains over existing model-free methods in imitating complex behaviors in large, high-dimensional environments."
                },
                {
                    "title": "Information-Theoretic Bounded Rationality",
                    "abstract": "Bounded rationality, that is, decision-making and planning under resource limitations, is widely regarded as an important open problem in artificial intelligence, reinforcement learning, computational neuroscience and economics. This paper offers a consolidated presentation of a theory of bounded rationality based on information-theoretic ideas. We provide a conceptual justification for using the free energy functional as the objective function for characterizing bounded-rational decisions. This functional possesses three crucial properties: it controls the size of the solution space; it has Monte Carlo planners that are exact, yet bypass the need for exhaustive search; and it captures model uncertainty arising from lack of evidence or from interacting with other agents having unknown intentions. We discuss the single-step decision-making case, and show how to extend it to sequential decisions using equivalence transformations. This extension yields a very general class of decision problems that encompass classical decision rules (e.g. EXPECTIMAX and MINIMAX) as limit cases, as well as trust- and risk-sensitive planning."
                },
                {
                    "title": "Convex Optimization",
                    "abstract": "This textbook is based on lectures given by the authors at MIPT (Moscow), HSE (Moscow), FEFU (Vladivostok), V.I. Vernadsky KFU (Simferopol), ASU (Republic of Adygea), and the University of Grenoble-Alpes (Grenoble, France). First of all, the authors focused on the program of a two-semester course of lectures on convex optimization, which is given to students of MIPT. The first chapter of this book contains the materials of the first semester (\"Fundamentals of convex analysis and optimization\"), the second and third chapters contain the materials of the second semester (\"Numerical methods of convex optimization\"). The textbook has a number of features. First, in contrast to the classic manuals, this book does not provide proofs of all the theorems mentioned. This allowed, on one side, to describe more themes, but on the other side, made the presentation less self-sufficient. The second important point is that part of the material is advanced and is published in the Russian educational literature, apparently for the first time. Third, the accents that are given do not always coincide with the generally accepted accents in the textbooks that are now popular. First of all, we talk about a sufficiently advanced presentation of conic optimization, including robust optimization, as a vivid demonstration of the capabilities of modern convex analysis."
                },
                {
                    "title": "A Game-Theoretic Approach to Apprenticeship Learning",
                    "abstract": "We study the problem of an apprentice learning to behave in an environment with an unknown reward function by observing the behavior of an expert. We follow on the work of Abbeel and Ng [1] who considered a framework in which the true reward function is assumed to be a linear combination of a set of known and observable features. We give a new algorithm that, like theirs, is guaranteed to learn a policy that is nearly as good as the expert's, given enough examples. However, unlike their algorithm, we show that ours may produce a policy that is substantially better than the expert's. Moreover, our algorithm is computationally faster, is easier to implement, and can be applied even in the absence of an expert. The method is based on a game-theoretic view of the problem, which leads naturally to a direct application of the multiplicative-weights algorithm of Freund and Schapire [2] for playing repeated matrix games. In addition to our formal presentation and analysis of the new algorithm, we sketch how the method can be applied when the transition function itself is unknown, and we provide an experimental demonstration of the algorithm on a toy video-game environment."
                },
                {
                    "title": "Apprenticeship learning via inverse reinforcement learning",
                    "abstract": "We consider learning in a Markov decision process where we are not explicitly given a reward function, but where instead we can observe an expert demonstrating the task that we want to learn to perform. This setting is useful in applications (such as the task of driving) where it may be difficult to write down an explicit reward function specifying exactly how different desiderata should be traded off. We think of the expert as trying to maximize a reward function that is expressible as a linear combination of known features, and give an algorithm for learning the task demonstrated by the expert. Our algorithm is based on using \"inverse reinforcement learning\" to try to recover the unknown reward function. We show that our algorithm terminates in a small number of iterations, and that even though we may never recover the expert's reward function, the policy output by the algorithm will attain performance close to that of the expert, where here performance is measured with respect to the expert's unknown reward function."
                },
                {
                    "title": "Online Convex Programming and Generalized Infinitesimal Gradient Ascent",
                    "abstract": "Convex programming involves a convex set F \u2286 Rn and a convex cost function c : F \u2192 R. The goal of convex programming is to find a point in F which minimizes c. In online convex programming, the convex set is known in advance, but in each step of some repeated optimization problem, one must select a point in F before seeing the cost function for that step. This can be used to model factory production, farm production, and many other industrial optimization problems where one is unaware of the value of the items produced until they have already been constructed. We introduce an algorithm for this domain. We also apply this algorithm to repeated games, and show that it is really a generalization of infinitesimal gradient ascent, and the results here imply that generalized infinitesimal gradient ascent (GIGA) is universally consistent."
                },
                {
                    "title": "Principal Angles between Subspaces in an A-Based Scalar Product: Algorithms and Perturbation Estimates",
                    "abstract": "Computation of principal angles between subspaces is important in many applications, e.g., in statistics and information retrieval. In statistics, the angles are closely related to measures of dependency and covariance of random variables. When applied to column-spaces of matrices, the principal angles describe canonical correlations of a matrix pair. We highlight that all popular software codes for canonical correlations compute only cosine of principal angles, thus making impossible, because of round-off errors, finding small angles accurately. We review a combination of sine and cosine based algorithms that provide accurate results for all angles. We generalize the method to the computation of principal angles in an A-based scalar product for a symmetric and positive definite matrix A. We provide a comprehensive overview of interesting properties of principal angles. We prove basic perturbation theorems for absolute errors for sine and cosine of principal angles with improved constants. Numerical examples and a detailed description of our code are given."
                },
                {
                    "title": "Algorithms for Inverse Reinforcement Learning",
                    "abstract": "Objective\u2014To evaluate the pharmacokinetics of a novel commercial formulation of ivermectin after administration to goats. Animals\u20146 healthy adult goats. Procedure\u2014Ivermectin (200 \u03bcg/kg) was initially administered IV to each goat, and plasma samples were obtained for 36 days. After a washout period of 3 weeks, each goat received a novel commercial formulation of ivermectin (200 \u03bcg/kg) by SC injection. Plasma samples were then obtained for 42 days. Drug concentrations were quantified by use of high-performance liquid chromatography with fluorescence detection. Results\u2014Pharmacokinetics of ivermectin after IV administration were best described by a 2-compartment open model; values for main compartmental variables included volume of distribution at a steady state (9.94 L/kg), clearance (1.54 L/kg/d), and area under the plasma concentration-time curve (AUC; 143 [ng\u2022d]/mL). Values for the noncompartmental variables included mean residence time (7.37 days), AUC (153 [ng\u2022d]/mL), and clearance (1.43 L/kg/d). After ..."
                },
                {
                    "title": "On Principal Angles between Subspaces of Euclidean Space",
                    "abstract": "The cosines of the principal angles between the column spaces of full column rank matrices $X\\in{\\hbox{{\\bf R}$^{m\\times p}$}}$ and $Y\\in{\\hbox{{\\bf R}$^{m\\times q}$}}$ are efficiently computed, using the Bjorck--Golub algorithm, as the singular values of $Q_x^{T}Q_y,$ where Qx and Qy are orthonormal matrices computed by the QR factorizations of X and Y, respectively. This paper shows that the Bjorck--Golub algorithm is mixed stable in the following sense: the computed singular values approximate with small relative error the exact cosines of the principal angles between the column spaces of $X+\\Delta X$ and $Y+\\Delta Y,$ where $\\Delta X,$ $\\Delta Y$ are small backward errors. Further, theoretical analysis and numerical evidence show that the algorithm becomes more robust if the QR factorizations are computed with the complete pivoting scheme of Powell and Reid. Moreover, it is shown that Gaussian elimination with complete pivoting can be used as an efficient preconditioner in computation and as a useful tool in analysis of the sensitivity of the QR factorization."
                },
                {
                    "title": "Policy Invariance Under Reward Transformations: Theory and Application to Reward Shaping",
                    "abstract": "This paper investigates conditions under which modi(cid:12)cations to the reward function of a Markov decision process preserve the optimal policy. It is shown that, besides the positive linear transformation familiar from utility theory, one can add a reward for transitions between states that is expressible as the di(cid:11)erence in value of an arbitrary potential function applied to those states. Further-more, this is shown to be a necessary condition for invariance, in the sense that any other transformation may yield suboptimal policies unless further assumptions are made about the underlying MDP. These results shed light on the practice of reward shaping, a method used in reinforcement learning whereby additional training rewards are used to guide the learning agent. In particular, some well-known \\bugs\" in reward shaping procedures are shown to arise from non-potential-based rewards, and methods are given for constructing shaping potentials corresponding to distance-based and subgoal-based heuristics. We show that such potentials can lead to substantial reductions in learning time."
                },
                {
                    "title": "Markov Decision Processes: Discrete Stochastic Dynamic Programming",
                    "abstract": "From the Publisher: \nThe past decade has seen considerable theoretical and applied research on Markov decision processes, as well as the growing use of these models in ecology, economics, communications engineering, and other fields where outcomes are uncertain and sequential decision-making processes are needed. A timely response to this increased activity, Martin L. Puterman's new work provides a uniquely up-to-date, unified, and rigorous treatment of the theoretical, computational, and applied research on Markov decision process models. It discusses all major research directions in the field, highlights many significant applications of Markov decision processes models, and explores numerous important topics that have previously been neglected or given cursory coverage in the literature. Markov Decision Processes focuses primarily on infinite horizon discrete time models and models with discrete time spaces while also examining models with arbitrary state spaces, finite horizon models, and continuous-time discrete state models. The book is organized around optimality criteria, using a common framework centered on the optimality (Bellman) equation for presenting results. The results are presented in a \"theorem-proof\" format and elaborated on through both discussion and examples, including results that are not available in any other book. A two-state Markov decision process model, presented in Chapter 3, is analyzed repeatedly throughout the book and demonstrates many results and algorithms. Markov Decision Processes covers recent research advances in such areas as countable state space models with average reward criterion, constrained models, and models with risk sensitive optimality criteria. It also explores several topics that have received little or no attention in other books, including modified policy iteration, multichain models with average reward criterion, and sensitive optimality. In addition, a Bibliographic Remarks section in each chapter comments on relevant historic"
                },
                {
                    "title": "Entropic Proximal Mappings with Applications to Nonlinear Programming",
                    "abstract": "We introduce a family of new transforms based on imitating the proximal mapping of Moreau and the associated Moreau-Yosida proximal approximation of a function. The transforms are constructed in terms of the A\u008f\u00c2\u2020-divergence functional a generalization of the relative entropy and of Bregman's measure of distance. An analogue of Moreau's theorem associated with these entropy-like distances is proved. We show that the resulting Entropic Proximal Maps share properties similar to the proximal mapping and provide a fairly general framework for constructing approximation and smoothing schemes for optimization problems. Applications of the results to the construction of generalized augmented Lagrangians for nonlinear programs and the minimax problem are presented."
                },
                {
                    "title": "When Is a Linear Control System Optimal",
                    "abstract": "The purpose of this paper is to formulate, study, and (in certain cases) resolve the Inverse Problem of Optimal Control Theory, which is the following: Given a control law, find all performance indices for which this control law is optimal. Under the assumptions of (a) linear constant plant, (b) linear constant control law, (c) measurable state variables, (d) quadratic loss functions with constant coefficients, (e) single control variable, we give a complete analysis of this problem and obtain various explicit conditions for the optimality of a given control law. An interestingfeature of the analysis is the central role of frequency-domain concepts, which have been ignored in optimal control theory until very recently. The discussion is presented in rigorous mathematical form. The central conclusion is the following (Theorem 6): A stable control law is optimal if and only if the absolute value of the corresponding return difference is at least equal to one at all frequencies. This provides a beautifully simple connecting link between modern control theory and the classical point of view which regards feedback as a means of reducing component variations."
                },
                {
                    "title": "Provably Efficient Learning of Transferable Rewards",
                    "abstract": "The reward function is widely accepted as a succinct, robust, and transferable representation of a task. Typical approaches, at the basis of Inverse Reinforcement Learning (IRL), leverage expert demonstrations to recover a reward function. In this paper, we study the theoretical properties of the class of reward functions that are compatible with the expert\u2019s behavior. We analyze how the limited knowledge of the expert\u2019s policy and of the environment affects the reward reconstruction phase. Then, we examine how the error propagates to the learned policy\u2019s performance when transferring the reward function to a different environment. We employ these \ufb01ndings to devise a provably ef\ufb01cient active sampling approach, aware of the need for transferring the reward function, that can be paired with a large variety of IRL algorithms. Finally, we provide numerical simulations on benchmark environments."
                },
                {
                    "title": "Reward Identification in Inverse Reinforcement Learning",
                    "abstract": "We study the problem of reward identi\ufb01ability in the context of Inverse Reinforcement Learning (IRL). The reward identi\ufb01ability question is critical to answer when reasoning about the effectiveness of using Markov Decision Processes (MDPs) as computational models of real world decision makers in order to understand complex decision making behavior and perform counter-factual reasoning. While identi\ufb01ability has been acknowledged as a fundamental theoretical question in IRL, little is known about the types of MDPs for which rewards are identi\ufb01able, or even if there exist such MDPs. In this work, we formalize the reward identi\ufb01cation problem in IRL and study how identi\ufb01ability relates to properties of the MDP model. For deterministic MDP models with the MaxEntRL objective, we prove necessary and suf\ufb01cient conditions for identi\ufb01ability. Building on these results, we present ef\ufb01cient algorithms for testing whether or not an MDP model is identi\ufb01able."
                },
                {
                    "title": "ALVINN, an autonomous land vehicle in a neural network",
                    "abstract": "The support-vector network is a new leaming machine for two-group classification problems. The machine conceptually implements the following idea: input vectors are non-linearly mapped to a very highdimension feature space. In this feature space a linear decision surface is constructed. Special properties of the decision surface ensures high generalization ability of the learning machine. The idea behind the support-vector network was previously implemented for the restricted case where the training data can be separated without errors. We here extend this result to non-separable training data. High generalization ability of support-vector networks utilizing polynomial input transformations is demonstrated. We also compare the performance of the support-vector network to various classical learning algorithms that all took part in a benchmark study of Optical Character Recognition."
                },
                {
                    "title": "Modeling Purposeful Adaptive Behavior with the Principle of Maximum Causal Entropy",
                    "abstract": "Predicting human behavior from a small amount of training examples is a challenging machine learning problem. In this thesis, we introduce the principle of maximum causal entropy, a general technique for applying information theory to decision-theoretic, game-theoretic, and control settings where relevant information is sequentially revealed over time. This approach guarantees decision-theoretic performance by matching purposeful measures of behavior (Abbeel & Ng, 2004), and/or enforces game-theoretic rationality constraints (Aumann, 1974), while otherwise being as uncertain as possible, which minimizes worst-case predictive log-loss (Grunwald & Dawid, 2003). \nWe derive probabilistic models for decision, control, and multi-player game settings using this approach. We then develop corresponding algorithms for efficient inference that include relaxations of the Bellman equation (Bellman, 1957), and simple learning algorithms based on convex optimization. We apply the models and algorithms to a number of behavior prediction tasks. Specifically, we present empirical evaluations of the approach in the domains of vehicle route preference modeling using over 100,000 miles of collected taxi driving data, pedestrian motion modeling from weeks of indoor movement data, and robust prediction of game play in stochastic multi-player games."
                },
                {
                    "title": "Local strong convexity and local Lipschitz continuity of the gradient of convex functions",
                    "abstract": "Given a pair of convex conjugate functions f and f\u2217, we investigate the relationship between local Lipschitz continuity of \u2207f and local strong convexity properties of f\u2217."
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nIs a reward recovered via inverse reinforcement learning (IRL) transferable to a new environment in the sense that its optimal policy aligns with the expert\u2019s true reward?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the fundamental challenge of reward design in reinforcement learning, which directly impacts the effectiveness of RL applications in real-world scenarios such as autonomous driving and robotics. By ensuring that rewards are transferable across different environments, we can enhance the robustness and adaptability of RL systems, leading to more reliable and efficient implementations. This research could pave the way for advancements in knowledge regarding reward structures and their implications, ultimately facilitating practical applications where RL can be deployed in diverse and dynamic settings.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the non-uniqueness of both the optimal policy corresponding to a reward and the reward corresponding to an optimal policy. This complexity can lead to trivial solutions, such as constant rewards that render all policies optimal, making it difficult to identify a reward that accurately reflects the expert's true intentions. Naive approaches may fail because they do not account for the intricacies of reward transferability across different environments, and they may overlook the need for a structured understanding of the reward landscape. Technical obstacles include the need for robust methods to characterize the set of rewards and the conditions under which transferability can be guaranteed.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on characterizing rewards under specific conditions, such as entropy regularization, but these approaches often rely on theoretical assumptions that do not translate well into practical applications. Limitations include a lack of comprehensive frameworks that can generalize across various environments and the challenge of learning from multiple experts with differing transition laws. Barriers such as insufficient empirical validation and the complexity of reward shaping have hindered progress. My approach aims to build upon prior work by integrating insights from multiple expert demonstrations and developing a more generalized framework for reward transferability.\n\n**[Question 5] - What are the key components of my approach and results?**  \nMy proposed methodology involves using a combination of inverse reinforcement learning techniques and multi-expert learning to identify transferable rewards. I will utilize a diverse dataset of expert demonstrations across different environments, focusing on autonomous driving scenarios. The evaluation metric will be the alignment of the learned policy's performance with the expert's true reward in new environments. The"
            }
        },
        "author_data": {
            "aeb8f795-e52a-4fab-8d66-667e4b442f4d": {
                "pk": "aeb8f795-e52a-4fab-8d66-667e4b442f4d",
                "name": "Andreas Schlaginhaufen",
                "collaborators": [
                    "Maryam Kamgarpour",
                    "Philippe Wenk",
                    "Andreas Krause",
                    "Florian D\u00f6rfler",
                    "Titouan Renard",
                    "Tingting Ni"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Inverse Reinforcement Learning",
                    "Neural ODE",
                    "System Identification"
                ],
                "publications": [
                    {
                        "title": "Learning Stable Deep Dynamics Models for Partially Observed or Delayed Dynamical Systems",
                        "abstract": "Learning how complex dynamical systems evolve over time is a key challenge in system identification. For safety critical systems, it is often crucial that the learned model is guaranteed to converge to some equilibrium point. To this end, neural ODEs regularized with neural Lyapunov functions are a promising approach when states are fully observed. For practical applications however, partial observations are the norm. As we will demonstrate, initialization of unobserved augmented states can become a key problem for neural ODEs. To alleviate this issue, we propose to augment the system's state with its history. Inspired by state augmentation in discrete-time systems, we thus obtain neural delay differential equations. Based on classical time delay stability analysis, we then show how to ensure stability of the learned models, and theoretically analyze our approach. Our experiments demonstrate its applicability to stable system identification of partially observed systems and learning a stabilizing feedback policy in delayed feedback control."
                    },
                    {
                        "title": "Identifiability and Generalizability in Constrained Inverse Reinforcement Learning",
                        "abstract": "Two main challenges in Reinforcement Learning (RL) are designing appropriate reward functions and ensuring the safety of the learned policy. To address these challenges, we present a theoretical framework for Inverse Reinforcement Learning (IRL) in constrained Markov decision processes. From a convex-analytic perspective, we extend prior results on reward identifiability and generalizability to both the constrained setting and a more general class of regularizations. In particular, we show that identifiability up to potential shaping (Cao et al., 2021) is a consequence of entropy regularization and may generally no longer hold for other regularizations or in the presence of safety constraints. We also show that to ensure generalizability to new transition laws and constraints, the true reward must be identified up to a constant. Additionally, we derive a finite sample guarantee for the suboptimality of the learned rewards, and validate our results in a gridworld environment."
                    },
                    {
                        "title": "Convergence of a model-free entropy-regularized inverse reinforcement learning algorithm",
                        "abstract": "Given a dataset of expert demonstrations, inverse reinforcement learning (IRL) aims to recover a reward for which the expert is optimal. This work proposes a model-free algorithm to solve entropy-regularized IRL problem. In particular, we employ a stochastic gradient descent update for the reward and a stochastic soft policy iteration update for the policy. Assuming access to a generative model, we prove that our algorithm is guaranteed to recover a reward for which the expert is $\\varepsilon$-optimal using $\\mathcal{O}(1/\\varepsilon^{2})$ samples of the Markov decision process (MDP). Furthermore, with $\\mathcal{O}(1/\\varepsilon^{4})$ samples we prove that the optimal policy corresponding to the recovered reward is $\\varepsilon$-close to the expert policy in total variation distance."
                    }
                ]
            },
            "83e26eca-ace8-4950-a608-5f986581700f": {
                "pk": "83e26eca-ace8-4950-a608-5f986581700f",
                "name": "Maryam Kamgarpour",
                "collaborators": [
                    "Tatiana Tatarenko",
                    "Tyler Summers",
                    "Luca Furieri",
                    "Orcun Karaca",
                    "Vasileios Lefkopoulos",
                    "Yimeng Lu",
                    "Pier Giuseppe Sessa",
                    "Neil Walton",
                    "Andreas Schlaginhaufen"
                ],
                "domain": [
                    "Game Theory",
                    "Optimization",
                    "Reinforcement Learning",
                    "Control Systems"
                ],
                "publications": [
                    {
                        "title": "Convergence Rate of Learning a Strongly Variationally Stable Equilibrium",
                        "abstract": "We derive the rate of convergence to the strongly variationally stable Nash equilibrium in a convex game, for a zeroth-order learning algorithm. Though we do not assume strong monotonicity of the game, our rates for the one-point feedback and for the two-point feedback match the best known rates for strongly monotone games under zeroth-order information."
                    },
                    {
                        "title": "Learning Nash Equilibria in Monotone Games",
                        "abstract": "We consider multi-agent decision making where each agent's cost function depends on all agents' strategies. We propose a distributed algorithm to learn a Nash equilibrium, whereby each agent uses only obtained values of her cost function at each joint played action, lacking any information of the functional form of her cost or other agents' costs or strategy sets. In contrast to past work where convergent algorithms required strong monotonicity, we prove algorithm convergence under mere monotonicity assumption. This significantly widens algorithm's applicability, such as to games with linear coupling constraints."
                    },
                    {
                        "title": "Payoff-Based Approach to Learning Nash Equilibria in Convex Games",
                        "abstract": "We consider multi-agent decision making, where each agent optimizes its cost function subject to constraints. Agents' actions belong to a compact convex Euclidean space and the agents' cost functions are coupled. We propose a distributed payoff-based algorithm to learn Nash equilibria in the game between agents. Each agent uses only information about its current cost value to compute its next action. We prove convergence of the proposed algorithm to a Nash equilibrium in the game leveraging established results on stochastic processes. The performance of the algorithm is analyzed with a numerical case study."
                    },
                    {
                        "title": "On infinite dimensional linear programming approach to stochastic control",
                        "abstract": "We consider the infinite dimensional linear programming (inf-LP) approach for solving stochastic control problems. The inf-LP corresponding to problems with uncountable state and input spaces is in general computationally intractable. By focusing on linear systems with quadratic cost (LQG), we establish a connection between this approach and the well-known Riccati LMIs. In particular, we show that the semidefinite programs known for the LQG problem can be derived from the pair of primal and dual inf-LPs. Furthermore, we establish a connection between multi-objective and chance constraint criteria and the inf-LP formulation."
                    },
                    {
                        "title": "On the Rate of Convergence of Payoff-based Algorithms to Nash Equilibrium in Strongly Monotone Games",
                        "abstract": "We derive the rate of convergence to Nash equilibria for the payoff-based algorithm proposed in \\cite{tat_kam_TAC}. These rates are achieved under the standard assumption of convexity of the game, strong monotonicity and differentiability of the pseudo-gradient. In particular, we show the algorithm achieves $O(\\frac{1}{T})$ in the two-point function evaluating setting and $O(\\frac{1}{\\sqrt{T}})$ in the one-point function evaluation under additional requirement of Lipschitz continuity of the pseudo-gradient. These rates are to our knowledge the best known rates for the corresponding problem classes."
                    },
                    {
                        "title": "Learning Generalized Nash Equilibria in a Class of Convex Games",
                        "abstract": "We consider multi-agent decision making where each agent optimizes its convex cost function subject to individual and coupling constraints. The constraint sets are compact convex subsets of a Euclidean space. To learn Nash equilibria, we propose a novel distributed payoff-based algorithm, such that each agent uses information only about its cost function values and the constraint function values with their associated dual multiplier. We prove convergence of this algorithm to a Nash equilibrium, under the assumption that the game admits a strictly convex potential function. In the absence of coupling constraints, we prove convergence to Nash equilibria under significantly weaker assumptions, not requiring a potential function. Namely, strict monotonicity of the game mapping is sufficient for convergence. We also derive the convergence rate of the algorithm for strongly monotone game maps."
                    },
                    {
                        "title": "Minimizing Regret of Bandit Online Optimization in Unconstrained Action Spaces",
                        "abstract": "We consider online convex optimization with a zero-order oracle feedback. In particular, the decision maker does not know the explicit representation of the time-varying cost functions, or their gradients. At each time step, she observes the value of the corresponding cost function evaluated at her chosen action (zero-order oracle). The objective is to minimize the regret, that is, the difference between the sum of the costs she accumulates and that of a static optimal action had she known the sequence of cost functions a priori. We present a novel algorithm to minimize regret in unconstrained action spaces. Our algorithm hinges on a classical idea of one-point estimation of the gradients of the cost functions based on their observed values. The algorithm is independent of problem parameters. Letting $T$ denote the number of queries of the zero-order oracle and $n$ the problem dimension, the regret rate achieved is $O(n^{2/3}T^{2/3})$. Moreover, we adapt the presented algorithm to the setting with two-point feedback and demonstrate that the adapted procedure achieves the theoretical lower bound on the regret of $(n^{1/2}T^{1/2})$."
                    },
                    {
                        "title": "Bandit Learning in Convex Non-Strictly Monotone Games",
                        "abstract": "We address learning Nash equilibria in convex games under the payoff information setting. We consider the case in which the game pseudo-gradient is monotone but not necessarily strictly monotone. This relaxation of strict monotonicity enables application of learning algorithms to a larger class of games, such as, for example, a zero-sum game with a merely convex-concave cost function. We derive an algorithm whose iterates provably converge to the least-norm Nash equilibrium in this setting. {From the perspective of a single player using the proposed algorithm, we view the game as an instance of online optimization}. Through this lens, we quantify the regret rate of the algorithm and provide an approach to choose the algorithm's parameters to minimize the regret rate."
                    },
                    {
                        "title": "Distributionally Robust Bootstrap Optimization",
                        "abstract": "Control architectures and autonomy stacks for complex engineering systems are often divided into layers to decompose a complex problem and solution into distinct, manageable sub-problems. To simplify designs, uncertainties are often ignored across layers, an approach with deep roots in classical notions of separation and certainty equivalence. But to develop robust architectures, especially as interactions between data-driven learning layers and model-based decision-making layers grow more intricate, more sophisticated interfaces between layers are required. We propose a basic architecture that couples a statistical parameter estimation layer with a constrained optimization layer. We show how the layers can be tightly integrated by combining bootstrap resampling with distributionally robust optimization. The approach allows a finite-data out-of-sample safety guarantee and an exact reformulation as a tractable finite-dimensional convex optimization problem."
                    },
                    {
                        "title": "Safe Mission Planning under Dynamical Uncertainties",
                        "abstract": "This paper considers safe robot mission planning in uncertain dynamical environments. This problem arises in applications such as surveillance, emergency rescue, and autonomous driving. It is a challenging problem due to modeling and integrating dynamical uncertainties into a safe planning framework, and finding a solution in a computationally tractable way. In this work, we first develop a probabilistic model for dynamical uncertainties. Then, we provide a framework to generate a path that maximizes safety for complex missions by incorporating the uncertainty model. We also devise a Monte Carlo method to obtain a safe path efficiently. Finally, we evaluate the performance of our approach and compare it to potential alternatives in several case studies."
                    },
                    {
                        "title": "Exploring Vickrey-Clarke-Groves Mechanism for Electricity Markets",
                        "abstract": "Control reserves are power generation or consumption entities that ensure balance of supply and demand of electricity in real-time. In many countries, they are operated through a market mechanism in which entities provide bids. The system operator determines the accepted bids based on an optimization algorithm. We develop the Vickrey-Clarke-Groves (VCG) mechanism for these electricity markets. We show that all advantages of the VCG mechanism including incentive compatibility of the equilibria and efficiency of the outcome can be guaranteed in these markets. Furthermore, we derive conditions to ensure collusion and shill bidding are not profitable. Our results are verified with numerical examples."
                    },
                    {
                        "title": "Unified Approach to Convex Robust Distributed Control given Arbitrary Information Structures",
                        "abstract": "We consider the problem of computing optimal linear control policies for linear systems in finite-horizon. The states and the inputs are required to remain inside pre-specified safety sets at all times despite unknown disturbances. In this technical note, we focus on the requirement that the control policy is distributed, in the sense that it can only be based on partial information about the history of the outputs. It is well-known that when a condition denoted as Quadratic Invariance (QI) holds, the optimal distributed control policy can be computed in a tractable way. Our goal is to unify and generalize the class of information structures over which quadratic invariance is equivalent to a test over finitely many binary matrices. The test we propose certifies convexity of the output-feedback distributed control problem in finite-horizon given any arbitrarily defined information structure, including the case of time varying communication networks and forgetting mechanisms. Furthermore, the framework we consider allows for including polytopic constraints on the states and the inputs in a natural way, without affecting convexity."
                    },
                    {
                        "title": "Performance guarantees for greedy maximization of non-submodular controllability metrics",
                        "abstract": "A key problem in emerging complex cyber-physical networks is the design of information and control topologies, including sensor and actuator selection and communication network design. These problems can be posed as combinatorial set function optimization problems to maximize a dynamic performance metric for the network. Some systems and control metrics feature a property called submodularity, which allows simple greedy algorithms to obtain provably near-optimal topology designs. However, many important metrics lack submodularity and therefore lack provable guarantees for using a greedy optimization approach. Here we show that performance guarantees can be obtained for greedy maximization of certain non-submodular functions of the controllability and observability Gramians. Our results are based on two key quantities: the submodularity ratio, which quantifies how far a set function is from being submodular, and the curvature, which quantifies how far a set function is from being supermodular."
                    },
                    {
                        "title": "Core-Selecting Mechanisms in Electricity Markets",
                        "abstract": "Due to its theoretical virtues, several recent works propose the use of the incentive-compatible Vickrey-Clarke-Groves (VCG) mechanism for electricity markets. Coalitions of participants, however, can influence the VCG outcome to obtain higher collective profit. To address this issue, we propose core-selecting mechanisms for their coalition-proofness. We show that core-selecting mechanisms generalize the economic rationale of the locational marginal pricing (LMP) mechanism. Namely, these mechanisms are the exact class of mechanisms that ensure the existence of a competitive equilibrium in linear/nonlinear prices. This implies that the LMP mechanism is also core-selecting, and hence coalition-proof. In contrast to the LMP mechanism, core-selecting mechanisms exist for a broad class of electricity markets, such as ones involving nonconvex costs and nonconvex constraint sets. In addition, they can approximate truthfulness without the price-taking assumption of the LMP mechanism. Finally, we show that they are also budget-balanced. Our results are verified with case studies based on optimal power flow test systems and the Swiss reserve market."
                    },
                    {
                        "title": "Using Uncertainty Data in Chance-Constrained Trajectory Planning",
                        "abstract": "We consider the problem of trajectory planning in an environment comprised of a set of obstacles with uncertain locations. While previous approaches model the uncertainties with a prescribed Gaussian distribution, we consider the realistic case in which the distribution's moments are unknown and are learned online. We derive tight concentration bounds on the error of the estimated moments. These bounds are then used to derive a tractable and tight mixed-integer convex reformulation of the trajectory planning problem, assuming linear dynamics and polyhedral constraints. The solution of the resulting optimization program is a feasible solution for the original problem with high confidence. We illustrate the approach with a case study from autonomous driving."
                    },
                    {
                        "title": "Robust Distributed Control Beyond Quadratic Invariance",
                        "abstract": "The problem of robust distributed control arises in several large-scale systems, such as transportation networks and power grid systems. In many practical scenarios controllers might not have enough information to make globally optimal decisions in a tractable way. We propose a novel class of tractable optimization problems whose solution is a controller complying with any specified information structure. The approach we suggest is based on decomposing intractable information constraints into two subspace constraints in the disturbance feedback domain. We discuss how to perform the decomposition in an optimized way. The resulting control policy is globally optimal when a condition known as Quadratic Invariance (QI) holds, whereas it is feasible and it provides a provable upper bound on the minimum cost when QI does not hold. Finally, we show that our method can lead to improved performance guarantees with respect to previous approaches, by applying the developed techniques to the platooning of autonomous vehicles."
                    },
                    {
                        "title": "Exploiting Weak Supermodularity for Coalition-Proof Mechanisms",
                        "abstract": "Under the incentive-compatible Vickrey-Clarke-Groves mechanism, coalitions of participants can influence the auction outcome to obtain higher collective profit. These manipulations were proven to be eliminated if and only if the market objective is supermodular. Nevertheless, several auctions do not satisfy the stringent conditions for supermodularity. These auctions include electricity markets, which are the main motivation of our study. To characterize nonsupermodular functions, we introduce the supermodularity ratio and the weak supermodularity. We show that these concepts provide us with tight bounds on the profitability of collusion and shill bidding. We then derive an analytical lower bound on the supermodularity ratio. Our results are verified with case studies based on the IEEE test systems."
                    },
                    {
                        "title": "First Order Methods For Globally Optimal Distributed Controllers Beyond Quadratic Invariance",
                        "abstract": "We study the distributed Linear Quadratic Gaussian (LQG) control problem in discrete-time and finite-horizon, where the controller depends linearly on the history of the outputs and it is required to lie in a given subspace, e.g. to possess a certain sparsity pattern. It is well-known that this problem can be solved with convex programming within the Youla domain if and only if a condition known as Quadratic Invariance (QI) holds. In this paper, we first show that given QI sparsity constraints, one can directly descend the gradient of the cost function within the domain of output-feedback controllers and converge to a global optimum. Note that convergence is guaranteed despite non-convexity of the cost function. Second, we characterize a class of Uniquely Stationary (US) problems, for which first-order methods are guaranteed to converge to a global optimum. We show that the class of US problems is strictly larger than that of strongly QI problems and that it is not included in that of QI problems. We refer to Figure 1 for details. Finally, we propose a tractable test for the US property."
                    },
                    {
                        "title": "Trajectory Planning Under Environmental Uncertainty With Finite-Sample Safety Guarantees",
                        "abstract": "We tackle the problem of trajectory planning in an environment comprised of a set of obstacles with uncertain time-varying locations. The uncertainties are modeled using widely accepted Gaussian distributions, resulting in a chance-constrained program. Contrary to previous approaches however, we do not assume perfect knowledge of the moments of the distribution, and instead estimate them through finite samples available from either sensors or past data. We derive tight concentration bounds on the error of these estimates to sufficiently tighten the chance-constraint program. As such, we provide provable guarantees on satisfaction of the chance-constraints corresponding to the nominal yet unknown moments. We illustrate our results with two autonomous vehicle trajectory planning case studies."
                    },
                    {
                        "title": "Identifiability and Generalizability in Constrained Inverse Reinforcement Learning",
                        "abstract": "Two main challenges in Reinforcement Learning (RL) are designing appropriate reward functions and ensuring the safety of the learned policy. To address these challenges, we present a theoretical framework for Inverse Reinforcement Learning (IRL) in constrained Markov decision processes. From a convex-analytic perspective, we extend prior results on reward identifiability and generalizability to both the constrained setting and a more general class of regularizations. In particular, we show that identifiability up to potential shaping (Cao et al., 2021) is a consequence of entropy regularization and may generally no longer hold for other regularizations or in the presence of safety constraints. We also show that to ensure generalizability to new transition laws and constraints, the true reward must be identified up to a constant. Additionally, we derive a finite sample guarantee for the suboptimality of the learned rewards, and validate our results in a gridworld environment."
                    }
                ]
            }
        }
    },
    "2409.17963": {
        "paper_data": {
            "title": "CNCA: Toward Customizable and Natural Generation of Adversarial Camouflage for Vehicle Detectors",
            "url": "http://arxiv.org/abs/2409.17963v1",
            "arxiv_id": "2409.17963",
            "authors": [
                "Linye Lyu",
                "Jiawei Zhou",
                "Daojing He",
                "Yu Li"
            ],
            "abstract": "Prior works on physical adversarial camouflage against vehicle detectors mainly focus on the effectiveness and robustness of the attack. The current most successful methods optimize 3D vehicle texture at a pixel level. However, this results in conspicuous and attention-grabbing patterns in the generated camouflage, which humans can easily identify. To address this issue, we propose a Customizable and Natural Camouflage Attack (CNCA) method by leveraging an off-the-shelf pre-trained diffusion model. By sampling the optimal texture image from the diffusion model with a user-specific text prompt, our method can generate natural and customizable adversarial camouflage while maintaining high attack performance. With extensive experiments on the digital and physical worlds and user studies, the results demonstrate that our proposed method can generate significantly more natural-looking camouflage than the state-of-the-art baselines while achieving competitive attack performance. Our code is available at \\href{https://anonymous.4open.science/r/CNCA-1D54}{https://anonymous.4open.science/r/CNCA-1D54}",
            "introduction": "   1 Introduction  Over the past years, Deep Neural Networks (DNNs) have revolutionized a wide range of research domains, especially in computer vision tasks, such as image classification, object detection, and semantic segmentation. DNNs are widely used in real-world systems, such as face recognition and autonomous driving. Despite their impressive success, DNNs are found vulnerable to adversarial examples [20], which are carefully crafted to deceive DNNs.   Generally, adversarial attacks can be classified into two categories: digital attacks, which primarily add small pixel-level perturbations to the input images; physical attacks, which manipulate the object\u2019s physical properties, such as its shape, surface, or surroundings, to deceive the target model in the real world. Physical attacks are more challenging than digital attacks, as they must remain effective under various complex physical conditions, including different viewing angles, distances, and lighting conditions. This paper focuses on physical attacks against vehicle detection models since they play critical roles in real-world applications like surveillance and autonomous driving systems.    Figure 1:  Customized and natural adversarial camouflage with various styles. (a) A car with normal texture; (b)(c)(d)(e) are the different styles of camouflage generated by our method CNCA. Their captions are user-specified input prompts.    Prior advanced physical attacks against vehicle detectors use adversarial camouflage technique [22; 21; 18; 19]. This technique fully covers the whole vehicle\u2019s surface with adversarial texture, which leads to better attack performance regardless of the viewing positions. Leveraging a differentiable renderer can effectively optimize 3D vehicle texture to deceive vehicle detectors via gradient back-propagation. However, all these methods suffer certain issues. Firstly, there is no prior knowledge of naturalness to guide the camouflage generation, resulting in conspicuous and attention-grabbing camouflage patterns. Secondly, all the current methods optimize the adversarial camouflage at a pixel level, making it challenging to resemble natural-looking patterns. Last but not least, none of the methods can customize the appearance of camouflage, making it hard to adapt to specific environments like forests and deserts.   To address the above issues, we propose CNCA, a novel framework to generate customizable and natural adversarial camouflage against vehicle detectors as shown in Figure\u00a01. Our insight is that: to gain naturalness and customizability, we need to leverage models that are equipped with prior knowledge of naturalness and allow conditional input signals. Motivated by this insight, we leverage an off-the-shelf pre-trained diffusion model to generate adversarial texture images with user-specific text prompts. The challenge of this approach is how to guide the adversarial gradient from the detection model to the image generation process. We introduce an adversarial feature to combine with the original text prompt feature. The combined feature forms the conditional input of the diffusion model. Thus, the resulting image is both natural and adversarial. Furthermore, we apply a clip strategy to the adversarial features to balance the trade-off between naturalness and attack performance. The combination of diffusion models, adversarial features, and clipping strategies facilitates the generation of customizable and natural camouflage.   The main contributions of our work are summarized as follows:   \u2022  To the best of our knowledge, our work is the first to investigate natural physical adversarial camouflage generation with diffusion models. It is",
            "references": [
                {
                    "title": "ACTIVE: Towards Highly Transferable 3D Physical Camouflage for Universal and Robust Vehicle Evasion",
                    "abstract": "Adversarial camouflage has garnered attention for its ability to attack object detectors from any viewpoint by covering the entire object\u2019s surface. However, universality and robustness in existing methods often fall short as the transferability aspect is often overlooked, thus restricting their application only to a specific target with limited performance. To address these challenges, we present Adversarial Camouflage for Transferable and Intensive Vehicle Evasion (ACTIVE), a state-of-the-art physical camouflage attack framework designed to generate universal and robust adversarial camouflage capable of concealing any 3D vehicle from detectors. Our framework incorporates innovative techniques to enhance universality and robustness, including a refined texture rendering that enables common texture application to different vehicles without being constrained to a specific texture map, a novel stealth loss that renders the vehicle undetectable, and a smooth and camouflage loss to enhance the naturalness of the adversarial camouflage. Our extensive experiments on 15 different models show that ACTIVE consistently outperforms existing works on various public detectors, including the latest YOLOv7. Notably, our universality evaluations reveal promising transferability to other vehicle classes, tasks (segmentation models), and the real world, not just other vehicles."
                },
                {
                    "title": "Physically Realizable Natural-Looking Clothing Textures Evade Person Detectors via 3D Modeling",
                    "abstract": "Recent works have proposed to craft adversarial clothes for evading person detectors, while they are either only effective at limited viewing angles or very conspicuous to humans. We aim to craft adversarial texture for clothes based on 3D modeling, an idea that has been used to craft rigid adversarial objects such as a 3D-printed turtle. Unlike rigid objects, humans and clothes are non-rigid, leading to difficulties in physical realization. In order to craft natural-looking adversarial clothes that can evade person detectors at multiple viewing angles, we propose adversarial camou-flage textures (AdvCaT) that resemble one kind of the typical textures of daily clothes, camouflage textures. We leverage the Voronoi diagram and Gumbel-softmax trick to parameterize the camouflage textures and optimize the parameters via 3D modeling. Moreover, we propose an efficient augmentation pipeline on 3D meshes combining topologically plausible projection (TopoProj) and Thin Plate Spline (TPS) to narrow the gap between digital and real-world objects. We printed the developed 3D texture pieces on fabric materials and tailored them into T-shirts and trousers. Experiments show high attack success rates of these clothes against multiple detectors."
                },
                {
                    "title": "LAION-5B: An open large-scale dataset for training next generation image-text models",
                    "abstract": "Groundbreaking language-vision architectures like CLIP and DALL-E proved the utility of training on large amounts of noisy image-text data, without relying on expensive accurate labels used in standard vision unimodal supervised learning. The resulting models showed capabilities of strong text-guided image generation and transfer to downstream tasks, while performing remarkably at zero-shot classification with noteworthy out-of-distribution robustness. Since then, large-scale language-vision models like ALIGN, BASIC, GLIDE, Flamingo and Imagen made further improvements. Studying the training and capabilities of such models requires datasets containing billions of image-text pairs. Until now, no datasets of this size have been made openly available for the broader research community. To address this problem and democratize research on large-scale multi-modal models, we present LAION-5B - a dataset consisting of 5.85 billion CLIP-filtered image-text pairs, of which 2.32B contain English language. We show successful replication and fine-tuning of foundational models like CLIP, GLIDE and Stable Diffusion using the dataset, and discuss further experiments enabled with an openly available dataset of this scale. Additionally we provide several nearest neighbor indices, an improved web-interface for dataset exploration and subset generation, and detection scores for watermark, NSFW, and toxic content detection. Announcement page https://laion.ai/laion-5b-a-new-era-of-open-large-scale-multi-modal-datasets/"
                },
                {
                    "title": "DTA: Physical Camouflage Attacks using Differentiable Transformation Network",
                    "abstract": "To perform adversarial attacks in the physical world, many studies have proposed adversarial camouflage, a method to hide a target object by applying camouflage patterns on 3D object surfaces. For obtaining optimal physical adversarial camouflage, previous studies have utilized the so-called neural renderer, as it supports differentiability. However, existing neural renderers cannot fully represent various real-world transformations due to a lack of control of scene parameters compared to the legacy photo-realistic renderers. In this paper, we propose the Differentiable Transformation Attack (DTA), a framework for generating a robust physical adversarial pattern on a target object to camouflage it against object detection models with a wide range of transformations. It utilizes our novel Differentiable Transformation Network (DTN), which learns the expected transformation of a rendered object when the texture is changed while preserving the original properties of the target object. Using our attack framework, an adversary can gain both the advantages of the legacy photo-realistic renderers including various physical-world transformations and the benefit of white-box access by offering differentiability. Our experiments show that our camouflaged 3D vehicles can successfully evade state-of-the-art object detection models in the photo-realistic environment (i.e., CARLA on Unreal Engine). Furthermore, our demonstration on a scaled Tesla Model 3 proves the applicability and transferability of our method to the real world."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Naturalistic Physical Adversarial Patch for Object Detectors",
                    "abstract": "Most prior works on physical adversarial attacks mainly focus on the attack performance but seldom enforce any restrictions over the appearance of the generated adversarial patches. This leads to conspicuous and attention-grabbing patterns for the generated patches which can be easily identified by humans. To address this issue, we pro-pose a method to craft physical adversarial patches for object detectors by leveraging the learned image manifold of a pretrained generative adversarial network (GAN) (e.g., BigGAN and StyleGAN) upon real-world images. Through sampling the optimal image from the GAN, our method can generate natural looking adversarial patches while maintaining high attack performance. With extensive experiments on both digital and physical domains and several independent subjective surveys, the results show that our proposed method produces significantly more realistic and natural looking patches than several state-of-the-art base-lines while achieving competitive attack performance. 1"
                },
                {
                    "title": "FCA: Learning a 3D Full-coverage Vehicle Camouflage for Multi-view Physical Adversarial Attack",
                    "abstract": "Physical adversarial attacks in object detection have attracted increasing attention. However, most previous works focus on hiding the objects from the detector by generating an individual adversarial patch, which only covers the planar part of the vehicle\u2019s surface and fails to attack the detector in physical scenarios for multi-view, long-distance and partially occluded objects. To bridge the gap between digital attacks and physical attacks, we exploit the full 3D vehicle surface to propose a robust Full-coverage Camouflage Attack (FCA) to fool detectors. Specifically, we first try rendering the nonplanar camouflage texture over the full vehicle surface. To mimic the real-world environment conditions, we then introduce a transformation function to transfer the rendered camouflaged vehicle into a photo-realistic scenario. Finally, we design an efficient loss function to optimize the camouflage texture. Experiments show that the full-coverage camouflage attack can not only outperform state-of-the-art methods under various test cases but also generalize to different environments, vehicles, and object detectors."
                },
                {
                    "title": "YOLOX: Exceeding YOLO Series in 2021",
                    "abstract": "In this report, we present some experienced improvements to YOLO series, forming a new high-performance detector -- YOLOX. We switch the YOLO detector to an anchor-free manner and conduct other advanced detection techniques, i.e., a decoupled head and the leading label assignment strategy SimOTA to achieve state-of-the-art results across a large scale range of models: For YOLO-Nano with only 0.91M parameters and 1.08G FLOPs, we get 25.3% AP on COCO, surpassing NanoDet by 1.8% AP; for YOLOv3, one of the most widely used detectors in industry, we boost it to 47.3% AP on COCO, outperforming the current best practice by 3.0% AP; for YOLOX-L with roughly the same amount of parameters as YOLOv4-CSP, YOLOv5-L, we achieve 50.0% AP on COCO at a speed of 68.9 FPS on Tesla V100, exceeding YOLOv5-L by 1.8% AP. Further, we won the 1st Place on Streaming Perception Challenge (Workshop on Autonomous Driving at CVPR 2021) using a single YOLOX-L model. We hope this report can provide useful experience for developers and researchers in practical scenes, and we also provide deploy versions with ONNX, TensorRT, NCNN, and Openvino supported. Source code is at https://github.com/Megvii-BaseDetection/YOLOX."
                },
                {
                    "title": "You Only Look One-level Feature",
                    "abstract": "This paper revisits feature pyramids networks (FPN) for one-stage detectors and points out that the success of FPN is due to its divide-and-conquer solution to the optimization problem in object detection rather than multi-scale feature fusion. From the perspective of optimization, we introduce an alternative way to address the problem instead of adopting the complex feature pyramids - utilizing only one-level feature for detection. Based on the simple and efficient solution, we present You Only Look One-level Feature (YOLOF). In our method, two key components, Dilated Encoder and Uniform Matching, are proposed and bring considerable improvements. Extensive experiments on the COCO benchmark prove the effectiveness of the proposed model. Our YOLOF achieves comparable results with its feature pyramids counterpart RetinaNet while being 2.5\u00d7 faster. Without transformer layers, YOLOF can match the performance of DETR in a single-level feature manner with 7\u00d7 less training epochs. Code is available at https://github.com/megvii-model/YOLOF."
                },
                {
                    "title": "Dual Attention Suppression Attack: Generate Adversarial Camouflage in Physical World",
                    "abstract": "Deep learning models are vulnerable to adversarial examples. As a more threatening type for practical deep learning systems, physical adversarial examples have received extensive research attention in recent years. However, without exploiting the intrinsic characteristics such as model-agnostic and human-specific patterns, existing works generate weak adversarial perturbations in the physical world, which fall short of attacking across different models and show visually suspicious appearance. Motivated by the viewpoint that attention reflects the intrinsic characteristics of the recognition process, this paper proposes the Dual Attention Suppression (DAS) attack to generate visually-natural physical adversarial camouflages with strong transferability by suppressing both model and human attention. As for attacking, we generate transferable adversarial camouflages by distracting the model-shared similar attention patterns from the target to non-target regions. Meanwhile, based on the fact that human visual attention always focuses on salient items (e.g., suspicious distortions), we evade the human-specific bottom-up attention to generate visually-natural camouflages which are correlated to the scenario context. We conduct extensive experiments in both the digital and physical world for classification and detection tasks on up-to-date models (e.g., Yolo-V5) and demonstrate that our method outperforms state-of-the-art methods.1"
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Sparse R-CNN: End-to-End Object Detection with Learnable Proposals",
                    "abstract": "We present Sparse R-CNN, a purely sparse method for object detection in images. Existing works on object detection heavily rely on dense object candidates, such as k anchor boxes pre-defined on all grids of image feature map of size H \u00d7 W. In our method, however, a fixed sparse set of learned object proposals, total length of N, are provided to object recognition head to perform classification and location. By eliminating HWk (up to hundreds of thousands) hand-designed object candidates to N (e.g. 100) learnable proposals, Sparse R-CNN completely avoids all efforts related to object candidates design and many-to-one label assignment. More importantly, final predictions are directly output without non-maximum suppression post-procedure. Sparse R-CNN demonstrates accuracy, run-time and training convergence performance on par with the well-established detector baselines on the challenging COCO dataset, e.g., achieving 45.0 AP in standard 3\u00d7 training schedule and running at 22 fps using ResNet-50 FPN model. We hope our work could inspire re-thinking the convention of dense prior in object detectors. The code is available at: https://github.com/PeizeSun/SparseR-CNN."
                },
                {
                    "title": "Deformable DETR: Deformable Transformers for End-to-End Object Detection",
                    "abstract": "DETR has been recently proposed to eliminate the need for many hand-designed components in object detection while demonstrating good performance. However, it suffers from slow convergence and limited feature spatial resolution, due to the limitation of Transformer attention modules in processing image feature maps. To mitigate these issues, we proposed Deformable DETR, whose attention modules only attend to a small set of key sampling points around a reference. Deformable DETR can achieve better performance than DETR (especially on small objects) with 10$\\times$ less training epochs. Extensive experiments on the COCO benchmark demonstrate the effectiveness of our approach. Code shall be released."
                },
                {
                    "title": "Physical Adversarial Attack on Vehicle Detector in the Carla Simulator",
                    "abstract": "In this paper, we tackle the issue of physical adversarial examples for object detectors in the wild. Specifically, we proposed to generate adversarial patterns to be applied on vehicle surface so that it's not recognizable by detectors in the photo-realistic Carla simulator. Our approach contains two main techniques, an Enlarge-and-Repeat process and a Discrete Searching method, to craft mosaic-like adversarial vehicle textures without access to neither the model weight of the detector nor a differential rendering procedure. The experimental results demonstrate the effectiveness of our approach in the simulator."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Accelerating 3D deep learning with PyTorch3D",
                    "abstract": "1. Accelerating 3D Deep Learning with PyTorch3D, arXiv 2007.08501 2. Mesh R-CNN, ICCV 2019 3. SynSin: End-to-end View Synthesis from a Single Image, CVPR 2020 4. Fast Differentiable Raycasting for Neural Rendering using Sphere-based Representations, arXiv 2004.07484"
                },
                {
                    "title": "MMDetection: Open MMLab Detection Toolbox and Benchmark",
                    "abstract": "We present MMDetection, an object detection toolbox that contains a rich set of object detection and instance segmentation methods as well as related components and modules. The toolbox started from a codebase of MMDet team who won the detection track of COCO Challenge 2018. It gradually evolves into a unified platform that covers many popular detection methods and contemporary modules. It not only includes training and inference codes, but also provides weights for more than 200 network models. We believe this toolbox is by far the most complete detection toolbox. In this paper, we introduce the various features of this toolbox. In addition, we also conduct a benchmarking study on different methods, components, and their hyper-parameters. We wish that the toolbox and benchmark could serve the growing research community by providing a flexible toolkit to reimplement existing methods and develop their own new detectors. Code and models are available at this https URL. The project is under active development and we will keep this document updated."
                },
                {
                    "title": "Objects as Points",
                    "abstract": "Detection identifies objects as axis-aligned boxes in an image. Most successful object detectors enumerate a nearly exhaustive list of potential object locations and classify each. This is wasteful, inefficient, and requires additional post-processing. In this paper, we take a different approach. We model an object as a single point --- the center point of its bounding box. Our detector uses keypoint estimation to find center points and regresses to all other object properties, such as size, 3D location, orientation, and even pose. Our center point based approach, CenterNet, is end-to-end differentiable, simpler, faster, and more accurate than corresponding bounding box based detectors. CenterNet achieves the best speed-accuracy trade-off on the MS COCO dataset, with 28.1% AP at 142 FPS, 37.4% AP at 52 FPS, and 45.1% AP with multi-scale testing at 1.4 FPS. We use the same approach to estimate 3D bounding box in the KITTI benchmark and human pose on the COCO keypoint dataset. Our method performs competitively with sophisticated multi-stage methods and runs in real-time."
                },
                {
                    "title": "YOLOv3: An Incremental Improvement",
                    "abstract": "We present some updates to YOLO! We made a bunch of little design changes to make it better. We also trained this new network that's pretty swell. It's a little bigger than last time but more accurate. It's still fast though, don't worry. At 320x320 YOLOv3 runs in 22 ms at 28.2 mAP, as accurate as SSD but three times faster. When we look at the old .5 IOU mAP detection metric YOLOv3 is quite good. It achieves 57.9 mAP@50 in 51 ms on a Titan X, compared to 57.5 mAP@50 in 198 ms by RetinaNet, similar performance but 3.8x faster. As always, all the code is online at this https URL"
                },
                {
                    "title": "CARLA: An Open Urban Driving Simulator",
                    "abstract": "We introduce CARLA, an open-source simulator for autonomous driving research. CARLA has been developed from the ground up to support development, training, and validation of autonomous urban driving systems. In addition to open-source code and protocols, CARLA provides open digital assets (urban layouts, buildings, vehicles) that were created for this purpose and can be used freely. The simulation platform supports flexible specification of sensor suites and environmental conditions. We use CARLA to study the performance of three approaches to autonomous driving: a classic modular pipeline, an end-to-end model trained via imitation learning, and an end-to-end model trained via reinforcement learning. The approaches are evaluated in controlled scenarios of increasing difficulty, and their performance is examined via metrics provided by CARLA, illustrating the platform's utility for autonomous driving research. The supplementary video can be viewed at this https URL"
                },
                {
                    "title": "Focal Loss for Dense Object Detection",
                    "abstract": "The highest accuracy object detectors to date are based on a two-stage approach popularized by R-CNN, where a classifier is applied to a sparse set of candidate object locations. In contrast, one-stage detectors that are applied over a regular, dense sampling of possible object locations have the potential to be faster and simpler, but have trailed the accuracy of two-stage detectors thus far. In this paper, we investigate why this is the case. We discover that the extreme foreground-background class imbalance encountered during training of dense detectors is the central cause. We propose to address this class imbalance by reshaping the standard cross entropy loss such that it down-weights the loss assigned to well-classified examples. Our novel Focal Loss focuses training on a sparse set of hard examples and prevents the vast number of easy negatives from overwhelming the detector during training. To evaluate the effectiveness of our loss, we design and train a simple dense detector we call RetinaNet. Our results show that when trained with the focal loss, RetinaNet is able to match the speed of previous one-stage detectors while surpassing the accuracy of all existing state-of-the-art two-stage detectors."
                },
                {
                    "title": "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks",
                    "abstract": "State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available"
                },
                {
                    "title": "Intriguing properties of neural networks",
                    "abstract": "Deep neural networks are highly expressive models that have recently achieved state of the art performance on speech and visual recognition tasks. While their expressiveness is the reason they succeed, it also causes them to learn uninterpretable solutions that could have counter-intuitive properties. In this paper we report two such properties. \nFirst, we find that there is no distinction between individual high level units and random linear combinations of high level units, according to various methods of unit analysis. It suggests that it is the space, rather than the individual units, that contains of the semantic information in the high layers of neural networks. \nSecond, we find that deep neural networks learn input-output mappings that are fairly discontinuous to a significant extend. We can cause the network to misclassify an image by applying a certain imperceptible perturbation, which is found by maximizing the network's prediction error. In addition, the specific nature of these perturbations is not a random artifact of learning: the same perturbation can cause a different network, that was trained on a different subset of the dataset, to misclassify the same input."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we generate customizable and natural adversarial camouflage for vehicle detection models that effectively deceives them under various physical conditions?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for enhancing the robustness of vehicle detection systems, which are integral to applications like surveillance and autonomous driving. By developing effective physical adversarial camouflage, we can better understand the vulnerabilities of DNNs, leading to improved defenses and more secure AI systems. This research could pave the way for future studies on adversarial robustness and inspire practical applications in security and safety, ultimately contributing to the development of more resilient AI technologies.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the need for the camouflage to remain effective across various physical conditions, such as different viewing angles, distances, and lighting. Naive approaches may fail because they do not account for the naturalness of the camouflage, leading to conspicuous patterns that are easily detectable. Additionally, optimizing adversarial camouflage at a pixel level complicates the generation of natural-looking patterns. The technical obstacles include effectively guiding the adversarial gradient from the detection model to the image generation process while maintaining a balance between naturalness and attack performance.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on adversarial camouflage techniques that lack a guiding principle for naturalness, resulting in patterns that are not only ineffective but also easily noticeable. Existing methods optimize camouflage at a pixel level without considering the need for customization to specific environments. Barriers to solving this problem include the absence of models that incorporate prior knowledge of naturalness and the challenges in integrating adversarial features with generative models. Our approach differs by leveraging a pre-trained diffusion model that allows for conditional input signals, enabling the generation of more effective and natural adversarial textures.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, CNCA, involves using a pre-trained diffusion model to generate adversarial texture images based on user-specific text prompts. The key components include the introduction of adversarial features that are combined with the original text prompt to form a conditional input for the diffusion model. We will evaluate the effectiveness of the generated camouflage using metrics such as attack success rate and naturalness assessment. The expected outcomes include the successful generation of customizable and natural adversarial camouflage that can effectively deceive vehicle detection models across"
            }
        },
        "author_data": {
            "de09b2cd-9ec5-4be7-a50a-698385022803": {
                "pk": "de09b2cd-9ec5-4be7-a50a-698385022803",
                "name": "Linye Lyu",
                "collaborators": [
                    "Jiawei Zhou",
                    "Daojing He",
                    "Yu Li"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Computer Vision",
                    "Neural Rendering"
                ],
                "publications": [
                    {
                        "title": "RAUCA: A Novel Physical Adversarial Attack on Vehicle Detectors via Robust and Accurate Camouflage Generation",
                        "abstract": "Adversarial camouflage is a widely used physical attack against vehicle detectors for its superiority in multi-view attack performance. One promising approach involves using differentiable neural renderers to facilitate adversarial camouflage optimization through gradient back-propagation. However, existing methods often struggle to capture environmental characteristics during the rendering process or produce adversarial textures that can precisely map to the target vehicle, resulting in suboptimal attack performance. Moreover, these approaches neglect diverse weather conditions, reducing the efficacy of generated camouflage across varying weather scenarios. To tackle these challenges, we propose a robust and accurate camouflage generation method, namely RAUCA. The core of RAUCA is a novel neural rendering component, Neural Renderer Plus (NRP), which can accurately project vehicle textures and render images with environmental characteristics such as lighting and weather. In addition, we integrate a multi-weather dataset for camouflage generation, leveraging the NRP to enhance the attack robustness. Experimental results on six popular object detectors show that RAUCA consistently outperforms existing methods in both simulation and real-world settings."
                    }
                ]
            },
            "58c0a6bf-15ba-4948-8069-bbf5aa1199c7": {
                "pk": "58c0a6bf-15ba-4948-8069-bbf5aa1199c7",
                "name": "Jiawei Zhou",
                "collaborators": [
                    "Li-Sheng Tseng",
                    "Phillip Keung",
                    "Benjamin Brimacombe",
                    "Alexander M. Rush",
                    "Li Dong",
                    "Furu Wei",
                    "Lei Chen",
                    "Guanfa Wang",
                    "Yong Wang",
                    "Bolin Liao"
                ],
                "domain": [
                    "Mathematics",
                    "Machine Learning",
                    "Symplectic Geometry",
                    "Information Retrieval"
                ],
                "publications": [
                    {
                        "title": "An upper bound for the Lusternik-Schnirelmann category of relative Sullivan algebras",
                        "abstract": "This paper addresses a question posed by F\\'elix, Halperin and Thomas. We prove that the Lusternik-Schnirelmann category of a relative Sullivan algebra is finite if such invariants of the base algebra and fiber algebra are both finite. Furthermore, we provide a similar estimation for the Toomer invariant."
                    },
                    {
                        "title": "Formality of Sphere Bundles",
                        "abstract": "We study the formality of orientable sphere bundles over connected compact manifolds. When the base manifold is formal, we prove that the formality of the bundle is equivalent to the vanishing of the Bianchi-Massey tensor introduced by Crowley-Nordstr\\\"{o}m. As an example, this implies that the unit tangent bundle over a formal manifold can only be formal when the base manifold has vanishing Euler characteristic or a rational cohomology ring generated by one element. When the base manifold is not formal, we give an obstruction to the formality of sphere bundles whose Euler class is reducible."
                    },
                    {
                        "title": "$A_\\infty$-Minimal Model on Differential Graded Algebras",
                        "abstract": "The rational homotopy type of a differential graded algebra (DGA) can be represented by a family of tensors on its cohomology, which constitute an $A_\\infty$-minimal model of this DGA. When only the cohomology is needed to determine the rational homotopy type, then the DGA is called formal. By a theorem of Miller, a compact $k$-connected manifold is formal if its dimension is not greater than $4k+2$. We expand this theorem and a result of Crowley-Nordstr\\\"{o}m to prove that if the dimension of a compact $k$-connected manifold $N\\leq (l+1)k+2$, then its de Rham complex has an $A_\\infty$-minimal model with $m_p=0$ for all $p\\geq l$. Separately, for an odd-dimensional sphere bundle over a formal manifold, we prove that its de Rham complex has an $A_\\infty$-minimal model with only $m_2$ and $m_3$ non-trivial. In the special case of a circle bundle over a formal symplectic manifold satisfying the hard Lefschetz property, we give a necessary condition for formality which becomes sufficient when the base symplectic manifold is of dimension six or less."
                    },
                    {
                        "title": "Cohomology of minimal Sullivan algebras of non-finite type and their realizations",
                        "abstract": "We prove that the morphisms from a minimal Sullivan algebra $\\Lambda V$ to $A_{PL}(|\\Lambda V|)$, the algebra of polynomial differential forms on its realization, can be quasi-isomorphic if and only if the cohomology $H(\\Lambda V)$ is of finite type. Importantly, $\\Lambda V$ itself need not be of finite type. For example, it can be the minimal Sullivan model of the wedge sum of a circle and a sphere. This provides a negative answer to a question posed by F\\'elix, Halperin, and Thomas. Furthermore, we study the spaces whose homotopy groups are reflected by their minimal Sullivan models as a generalization of Sullivan spaces, and explore which properties of Sullivan spaces can be broadened."
                    },
                    {
                        "title": "A Novel Hybrid Approach for Tornado Prediction in the United States: Kalman-Convolutional BiLSTM with Multi-Head Attention",
                        "abstract": "Tornadoes are among the most intense atmospheric vortex phenomena and pose significant challenges for detection and forecasting. Conventional methods, which heavily depend on ground-based observations and radar data, are limited by issues such as decreased accuracy over greater distances and a high rate of false positives. To address these challenges, this study utilizes the Seamless Hybrid Scan Reflectivity (SHSR) dataset from the Multi-Radar Multi-Sensor (MRMS) system, which integrates data from multiple radar sources to enhance accuracy. A novel hybrid model, the Kalman-Convolutional BiLSTM with Multi-Head Attention, is introduced to improve dynamic state estimation and capture both spatial and temporal dependencies within the data. This model demonstrates superior performance in precision, recall, F1-Score, and accuracy compared to methods such as K-Nearest Neighbors (KNN) and LightGBM. The results highlight the considerable potential of advanced machine learning techniques to improve tornado prediction and reduce false alarm rates. Future research will focus on expanding datasets, exploring innovative model architectures, and incorporating large language models (LLMs) to provide deeper insights. This research introduces a novel model for tornado prediction, offering a robust framework for enhancing forecasting accuracy and public safety."
                    },
                    {
                        "title": "Improving Non-autoregressive Neural Machine Translation with Monolingual Data",
                        "abstract": "Non-autoregressive (NAR) neural machine translation is usually done via knowledge distillation from an autoregressive (AR) model. Under this framework, we leverage large monolingual corpora to improve the NAR model's performance, with the goal of transferring the AR model's generalization ability while preventing overfitting. On top of a strong NAR baseline, our experimental results on the WMT14 En-De and WMT16 En-Ro news translation tasks confirm that monolingual data augmentation consistently improves the performance of the NAR model to approach the teacher AR model's performance, yields comparable or better results than the best non-iterative NAR methods in the literature and helps reduce overfitting in the training process."
                    },
                    {
                        "title": "Quick Back-Translation for Unsupervised Machine Translation",
                        "abstract": "The field of unsupervised machine translation has seen significant advancement from the marriage of the Transformer and the back-translation algorithm. The Transformer is a powerful generative model, and back-translation leverages Transformer's high-quality translations for iterative self-improvement. However, the Transformer is encumbered by the run-time of autoregressive inference during back-translation, and back-translation is limited by a lack of synthetic data efficiency. We propose a two-for-one improvement to Transformer back-translation: Quick Back-Translation (QBT). QBT re-purposes the encoder as a generative model, and uses encoder-generated sequences to train the decoder in conjunction with the original autoregressive back-translation step, improving data throughput and utilization. Experiments on various WMT benchmarks demonstrate that a relatively small number of refining steps of QBT improve current unsupervised machine translation models, and that QBT dramatically outperforms standard back-translation only method in terms of training efficiency for comparable translation qualities."
                    },
                    {
                        "title": "Simple Unsupervised Summarization by Contextual Matching",
                        "abstract": "We propose an unsupervised method for sentence summarization using only language modeling. The approach employs two language models, one that is generic (i.e. pretrained), and the other that is specific to the target domain. We show that by using a product-of-experts criteria these are enough for maintaining continuous contextual matching while maintaining output fluency. Experiments on both abstractive and extractive sentence summarization data sets show promising results of our method without being exposed to any paired data."
                    },
                    {
                        "title": "Symplectic Flatness and Twisted Primitive Cohomology",
                        "abstract": "We introduce the notion of symplectic flatness for connections and fiber bundles over symplectic manifolds. Given an $A_\\infty$-algebra, we present a flatness condition that enables the twisting of the differential complex associated with the $A_\\infty$-algebra. The symplectic flatness condition arises from twisting the $A_\\infty$-algebra of differential forms constructed by Tsai, Tseng and Yau. When the symplectic manifold is equipped with a compatible metric, the symplectic flat connections represent a special subclass of Yang-Mills connections. We further study the cohomologies of the twisted differential complex and give a simple vanishing theorem for them."
                    },
                    {
                        "title": "Symplectically Flat Connections and Their Functionals",
                        "abstract": "We continue our study of symplectically flat bundles. We broaden the notion of symplectically flat connections on symplectic manifolds to $\\zeta$-flat connections on smooth manifolds. These connections on principal bundles can be represented by maps from an extension of the base's fundamental group to the structure group. We introduce functionals with zeroes being symplectically flat connections and study their critical points. Such functionals lead to novel geometric flows. We also describe some characteristic classes of the $\\zeta$-flat bundles."
                    },
                    {
                        "title": "Semi-Parametric Retrieval via Binary Token Index",
                        "abstract": "The landscape of information retrieval has broadened from search services to a critical component in various advanced applications, where indexing efficiency, cost-effectiveness, and freshness are increasingly important yet remain less explored. To address these demands, we introduce Semi-parametric Vocabulary Disentangled Retrieval (SVDR). SVDR is a novel semi-parametric retrieval framework that supports two types of indexes: an embedding-based index for high effectiveness, akin to existing neural retrieval methods; and a binary token index that allows for quick and cost-effective setup, resembling traditional term-based retrieval. In our evaluation on three open-domain question answering benchmarks with the entire Wikipedia as the retrieval corpus, SVDR consistently demonstrates superiority. It achieves a 3% higher top-1 retrieval accuracy compared to the dense retriever DPR when using an embedding-based index and an 9% higher top-1 accuracy compared to BM25 when using a binary token index. Specifically, the adoption of a binary token index reduces index preparation time from 30 GPU hours to just 2 CPU hours and storage size from 31 GB to 2 GB, achieving a 90% reduction compared to an embedding-based index."
                    },
                    {
                        "title": "Hydrodynamic limit of the Boltzmann equation to the planar rarefaction wave in three dimensional space",
                        "abstract": "In this paper, we establish the global in time hydrodynamic limit of Boltzmann equation to the planar rarefaction wave of compressible Euler system in three dimensional space $x\\in\\mathbb{R}^3$ for general collision kernels. Our approch is based on a generalized Hilbert expansion, and a recent $L^2-L^\\infty$ framework. In particular, we improve the $L^2$-estimate to be a localized version because the planar rarefaction wave is indeed a one-dimensional wave which makes the source terms to be not integrable in the $L^2$ energy estimate of three dimensional problem. We also point out that the wave strength of rarefaction may be large."
                    },
                    {
                        "title": "Generalized two-temperature model for coupled phonon-magnon diffusion",
                        "abstract": "We generalize the two-temperature model [Sanders and Walton, Phys. Rev. B, 15, 1489 (1977)] for coupled phonon-magnon diffusion to include the effect of the concurrent magnetization flow. Working within the framework of Boltzmann transport equation, we derive the constitutive equations for coupled phonon-magnon transport driven by gradients of both temperature and external magnetic fields, and the corresponding conservation laws. Our equations reduce to the original Sanders-Walton two-temperature model under a uniform external field, but predict a new magnon cooling effect driven by a non-uniform magnetic field in a homogeneous single-domain ferromagnet. We estimate the magnitude of the cooling effect in yttrium iron garnet, and show it is within current experimental reach. With properly optimized materials, the predicted cooling effect can potentially supplement the conventional magnetocaloric effect in cryogenic applications in the future."
                    }
                ]
            },
            "fdb975e4-87a9-4f98-9004-eeaec1ae3067": {
                "pk": "fdb975e4-87a9-4f98-9004-eeaec1ae3067",
                "name": "Daojing He",
                "collaborators": [
                    "Yu Li",
                    "Jiawei Zhou",
                    "Linye Lyu",
                    "Muxi Chen",
                    "Yannan Liu",
                    "Qiang Xu",
                    "Xin Chen",
                    "Zhili Chen",
                    "Benchang Dong",
                    "Shiwen Wei"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Deep Learning",
                    "Active Learning",
                    "Model Merging"
                ],
                "publications": [
                    {
                        "title": "RAUCA: A Novel Physical Adversarial Attack on Vehicle Detectors via Robust and Accurate Camouflage Generation",
                        "abstract": "Adversarial camouflage is a widely used physical attack against vehicle detectors for its superiority in multi-view attack performance. One promising approach involves using differentiable neural renderers to facilitate adversarial camouflage optimization through gradient back-propagation. However, existing methods often struggle to capture environmental characteristics during the rendering process or produce adversarial textures that can precisely map to the target vehicle, resulting in suboptimal attack performance. Moreover, these approaches neglect diverse weather conditions, reducing the efficacy of generated camouflage across varying weather scenarios. To tackle these challenges, we propose a robust and accurate camouflage generation method, namely RAUCA. The core of RAUCA is a novel neural rendering component, Neural Renderer Plus (NRP), which can accurately project vehicle textures and render images with environmental characteristics such as lighting and weather. In addition, we integrate a multi-weather dataset for camouflage generation, leveraging the NRP to enhance the attack robustness. Experimental results on six popular object detectors show that RAUCA consistently outperforms existing methods in both simulation and real-world settings."
                    },
                    {
                        "title": "An Empirical Study on the Efficacy of Deep Active Learning for Image Classification",
                        "abstract": "Deep Active Learning (DAL) has been advocated as a promising method to reduce labeling costs in supervised learning. However, existing evaluations of DAL methods are based on different settings, and their results are controversial. To tackle this issue, this paper comprehensively evaluates 19 existing DAL methods in a uniform setting, including traditional fully-\\underline{s}upervised \\underline{a}ctive \\underline{l}earning (SAL) strategies and emerging \\underline{s}emi-\\underline{s}upervised \\underline{a}ctive \\underline{l}earning (SSAL) techniques. We have several non-trivial findings. First, most SAL methods cannot achieve higher accuracy than random selection. Second, semi-supervised training brings significant performance improvement compared to pure SAL methods. Third, performing data selection in the SSAL setting can achieve a significant and consistent performance improvement, especially with abundant unlabeled data. Our findings produce the following guidance for practitioners: one should (i) apply SSAL early and (ii) collect more unlabeled data whenever possible, for better model performance."
                    },
                    {
                        "title": "FOBNN: Fast Oblivious Binarized Neural Network Inference",
                        "abstract": "The superior performance of deep learning has propelled the rise of Deep Learning as a Service, enabling users to transmit their private data to service providers for model execution and inference retrieval. Nevertheless, the primary concern remains safeguarding the confidentiality of sensitive user data while optimizing the efficiency of secure protocols. To address this, we develop a fast oblivious binarized neural network inference framework, FOBNN. Specifically, we customize binarized convolutional neural networks to enhance oblivious inference, design two fast algorithms for binarized convolutions, and optimize network structures experimentally under constrained costs. Initially, we meticulously analyze the range of intermediate values in binarized convolutions to minimize bit representation, resulting in the Bit Length Bounding (BLB) algorithm. Subsequently, leveraging the efficiency of bitwise operations in BLB, we further enhance performance by employing pure bitwise operations for each binary digit position, yielding the Layer-wise Bit Accumulation (LBA) algorithm. Theoretical analysis validates FOBNN's security and indicates up to $2 \\times$ improvement in computational and communication costs compared to the state-of-the-art method. We demonstrates our framework's effectiveness in RNA function prediction within bioinformatics. Rigorous experimental assessments confirm that our oblivious inference solutions not only maintain but often exceed the original accuracy, surpassing prior efforts."
                    },
                    {
                        "title": "Parameter Competition Balancing for Model Merging",
                        "abstract": "While fine-tuning pretrained models has become common practice, these models often underperform outside their specific domains. Recently developed model merging techniques enable the direct integration of multiple models, each fine-tuned for distinct tasks, into a single model. This strategy promotes multitasking capabilities without requiring retraining on the original datasets. However, existing methods fall short in addressing potential conflicts and complex correlations between tasks, especially in parameter-level adjustments, posing a challenge in effectively balancing parameter competition across various tasks. This paper introduces an innovative technique named PCB-Merging (Parameter Competition Balancing), a lightweight and training-free technique that adjusts the coefficients of each parameter for effective model merging. PCB-Merging employs intra-balancing to gauge parameter significance within individual tasks and inter-balancing to assess parameter similarities across different tasks. Parameters with low importance scores are dropped, and the remaining ones are rescaled to form the final merged model. We assessed our approach in diverse merging scenarios, including cross-task, cross-domain, and cross-training configurations, as well as out-of-domain generalization. The experimental results reveal that our approach achieves substantial performance enhancements across multiple modalities, domains, model sizes, number of tasks, fine-tuning forms, and large language models, outperforming existing model merging methods. The code is publicly available at: \\url{https://github.com/duguodong7/pcb-merging}."
                    }
                ]
            },
            "fe0ae110-3822-44af-b9b2-b1fa59213510": {
                "pk": "fe0ae110-3822-44af-b9b2-b1fa59213510",
                "name": "Yu Li",
                "collaborators": [],
                "domain": [
                    "K\u00e4hler Geometry",
                    "Ricci Flow",
                    "Stochastic Processes",
                    "Cosmology"
                ],
                "publications": [
                    {
                        "title": "Ancient solutions to the K\u00e4hler Ricci flow",
                        "abstract": "We give a complete classification of all $\\kappa$-noncollapsed, complete ancient solutions to the K\\\"ahler Ricci flow with nonnegative bisectional curvature."
                    },
                    {
                        "title": "Tail Behavoir of sums of random components",
                        "abstract": "A class of stochastic processes strongly related to random sums plays an important role in network and in finance. In this paper we study this kind of stochastic process discuss an overtime unchanged parameter and reveal its asymptotic behavior."
                    },
                    {
                        "title": "Ricci flow on asymptotically Euclidean manifolds",
                        "abstract": "In this paper, we prove that if an asymptotically Euclidean manifold $(M^n,g)$ under the condition that $R \\ge 0$ has long time existence of Ricci flow, the mass of $(M^n,g)$ is nonnegative. In addition, we give an independent proof of positive mass theorem in dimension $3$."
                    },
                    {
                        "title": "Dynamics of two-scalar-field cosmological models beyond the exponential potential",
                        "abstract": "In this paper, we discuss the dynamics of two- scalar-field cosmological models. Unlike in the situation of exponential potential, we find that there are late-time attractors in which one scalar field dominates the energy density of universe and the other one decay. We also discuss the possibility of multiple attractors model which is useful to realize the evolution of the universe from a scaling era to recent acceleration era. We also give the conditions of the existence of multiple attractors."
                    },
                    {
                        "title": "Case Study of the Proof of Cook's theorem - Interpretation of A(w)",
                        "abstract": "Cook's theorem is commonly expressed such as any polynomial time-verifiable problem can be reduced to the SAT problem. The proof of Cook's theorem consists in constructing a propositional formula A(w) to simulate a computation of TM, and such A(w) is claimed to be CNF to represent a polynomial time-verifiable problem w. In this paper, we investigate A(w) through a very simple example and show that, A(w) has just an appearance of CNF, but not a true logical form. This case study suggests that there exists the begging the question in Cook's theorem."
                    },
                    {
                        "title": "Gaiotto's Lagrangian subvarieties via loop groups",
                        "abstract": "The purpose of this note is to give a simple proof of the fact that a certain substack, defined in [2], of the moduli stack $T^{\\ast}Bun_G(\\Sigma)$ of Higgs bundles over a curve $\\Sigma$, for a connected, simply connected semisimple group $G$, possesses a Lagrangian structure. The substack, roughly speaking, consists of images under the moment map of global sections of principal $G$-bundles over $\\Sigma$ twisted by a smooth symplectic variety with a Hamiltonian $G$-action."
                    }
                ]
            }
        }
    },
    "2410.15556": {
        "paper_data": {
            "title": "Gradient Rewiring for Editable Graph Neural Network Training",
            "url": "http://arxiv.org/abs/2410.15556v1",
            "arxiv_id": "2410.15556",
            "authors": [
                "Zhimeng Jiang",
                "Zirui Liu",
                "Xiaotian Han",
                "Qizhang Feng",
                "Hongye Jin",
                "Qiaoyu Tan",
                "Kaixiong Zhou",
                "Na Zou",
                "Xia Hu"
            ],
            "abstract": "Deep neural networks are ubiquitously adopted in many applications, such as computer vision, natural language processing, and graph analytics. However, well-trained neural networks can make prediction errors after deployment as the world changes. \\textit{Model editing} involves updating the base model to correct prediction errors with less accessible training data and computational resources. Despite recent advances in model editors in computer vision and natural language processing, editable training in graph neural networks (GNNs) is rarely explored. The challenge with editable GNN training lies in the inherent information aggregation across neighbors, which can lead model editors to affect the predictions of other nodes unintentionally. In this paper, we first observe the gradient of cross-entropy loss for the target node and training nodes with significant inconsistency, which indicates that directly fine-tuning the base model using the loss on the target node deteriorates the performance on training nodes. Motivated by the gradient inconsistency observation, we propose a simple yet effective \\underline{G}radient \\underline{R}ewiring method for \\underline{E}ditable graph neural network training, named \\textbf{GRE}. Specifically, we first store the anchor gradient of the loss on training nodes to preserve the locality. Subsequently, we rewire the gradient of the loss on the target node to preserve performance on the training node using anchor gradient. Experiments demonstrate the effectiveness of GRE on various model architectures and graph datasets in terms of multiple editing situations. The source code is available at \\url{https://github.com/zhimengj0326/Gradient_rewiring_editing}",
            "introduction": "   1 Introduction  Graph Neural Networks (GNNs) have demonstrated exemplary performance for graph learning tasks, such as recommendation, link prediction, molecule property analysis (Ying et\u00a0al., 2018; Zeng et\u00a0al., 2020; Hu et\u00a0al., 2020; Jiang et\u00a0al., 2022; Duan et\u00a0al., 2022; Han et\u00a0al., 2022; Ling et\u00a0al., 2023). With message passing, GNNs learn node representations by recursively aggregating the neighboring nodes\u2019 representations. Once trained, GNN models are deployed to handle various high-stake tasks, such as credit risk assessment in financial networks (Petrone and Latora, 2018) and fake news detection in social networks (Shu et\u00a0al., 2017). However, the impact of erroneous decisions in such influential applications can be substantial. For instance, misplaced credit trust in undetected fake news can lead to severe financial loss.   An ideal approach to tackle such errors should possess the following properties: 1) the ability to rectify severe errors in the model\u2019s predictions, 2) the capacity to generalize these corrections to other similar instances of misclassified samples, and 3) the ability to preserve the model\u2019s prediction accuracy for all other unrelated inputs. To achieve these goals, various model editing frameworks have been developed to rectify errors by dynamically adjusting the model\u2019s behavior when errors are detected (Sinitsin et\u00a0al., 2020; Mitchell et\u00a0al., 2021). The core principle is to implement minimal changes to the model to correct the error while keeping the rest of the model\u2019s behavior intact. However, model editing is not a simple plug-and-play solution. These frameworks often require an additional training phase to prepare for editing before they can be used effectively for editing (Sinitsin et\u00a0al., 2020; Mitchell et\u00a0al., 2021, 2022; De\u00a0Cao et\u00a0al., 2021). Although model editing techniques have shown significant utility in computer vision and language models, there is rare work focused on rectifying critical errors in graph data. The unique challenge arises from the inherent message-passing mechanism in GNNs when edits involve densely interconnected nodes (Liu et\u00a0al., 2023b; Zhong et\u00a0al., 2024). Specifically, editing the behavior of a single node can unintentionally induce a ripple effect, causing changes that propagate throughout the entire graph. (Liu et\u00a0al., 2023b) theoretically and empirically demonstrate the complexity of editing GNNs through the lens of the loss landscape of the Kullback-Lieber divergence between the pre-trained node features and edited final node embeddings. Moreover, a simple yet effective model structure, named EGNN, is proposed with stitched peer multi-layer perception (MLP), where only the stitched MLP is trained during model editing.   In this work, we investigate the model editing problem for GNNs from a brand-new gradient perspective, which is compatible with existing work (Liu et\u00a0al., 2023b). Specifically, we first found a considerable inconsistency between the gradients of the cross-entropy loss for the target node and the training nodes for GNNs. Such inconsistency implies that direct fine-tuning of the base model using the loss of the target node can lead to a deterioration in the performance on the training nodes. Motivated by the above observation, we propose a simple yet effective Gradient Rewiring method for Editable graph neural network training, named GRE. Specifically, we first calculate and store the anchor gradient of the loss on the training nodes. This anchor gradient represents the original learning",
            "references": [
                {
                    "title": "TinyGraph: Joint Feature and Node Condensation for Graph Neural Networks",
                    "abstract": "Training graph neural networks (GNNs) on large-scale graphs can be challenging due to the high computational expense caused by the massive number of nodes and high-dimensional nodal features. Existing graph condensation studies tackle this problem only by reducing the number of nodes in the graph. However, the resulting condensed graph data can still be cumbersome. Specifically, although the nodes of the Citeseer dataset are reduced to 0.9% (30 nodes) in training, the number of features is 3,703, severely exceeding the training sample magnitude. Faced with this challenge, we study the problem of joint condensation for both features and nodes in large-scale graphs. This task is challenging mainly due to 1) the intertwined nature of the node features and the graph structure calls for the feature condensation solver to be structure-aware; and 2) the difficulty of keeping useful information in the condensed graph. To address these challenges, we propose a novel framework TinyGraph, to condense features and nodes simultaneously in graphs. Specifically, we cast the problem as matching the gradients of GNN weights trained on the condensed graph and the gradients obtained from training over the original graph, where the feature condensation is achieved by a trainable function. The condensed graph obtained by minimizing the matching loss along the training trajectory can henceforth retain critical information in the original graph. Extensive experiments were carried out to demonstrate the effectiveness of the proposed TinyGraph. For example, a GNN trained with TinyGraph retains 98.5% and 97.5% of the original test accuracy on the Cora and Citeseer datasets, respectively, while significantly reducing the number of nodes by 97.4% and 98.2%, and the number of features by 90.0% on both datasets."
                },
                {
                    "title": "Error Detection on Knowledge Graphs with Triple Embedding",
                    "abstract": "Knowledge graphs (KGs), as essential ingredients in many real-world applications, always contain a considerable number of errors. KG error detection aims to find the triples whose head entity, tail entity, and corresponding relation are mismatched. Despite being urgently needed, existing KG error detection methods lack generalizability. They mainly utilize supervised information such as entity type or erroneous labels, but such information is not often available in the real world. The challenges in detecting errors in KGs are twofold. Firstly, KGs exhibit unique data characteristics that distinguish them from general graphs. Secondly, real-world KGs tend to be large, and labels are often scarce. To bridge the gap, we propose a novel KG error detection framework based on triple embedding, termed TripleNet. TripleNet constructs a triple network by treating each triple as a node and connecting them via shared entities. It then employs a Bi-LSTM module to capture intra-triple translational information at the local level and uses a graph attention network to gather inter-triple contextual information at the global level. Finally, it computes the suspicious score of each triple by integrating its local and global-level information. Experimental results on two real-world KGs demonstrated that TripleNet outperforms state-of-the-art error detection algorithms with comparable or even better efficiency."
                },
                {
                    "title": "Graph Mixup with Soft Alignments",
                    "abstract": "We study graph data augmentation by mixup, which has been used successfully on images. A key operation of mixup is to compute a convex combination of a pair of inputs. This operation is straightforward for grid-like data, such as images, but challenging for graph data. The key difficulty lies in the fact that different graphs typically have different numbers of nodes, and thus there lacks a node-level correspondence between graphs. In this work, we propose S-Mixup, a simple yet effective mixup method for graph classification by soft alignments. Specifically, given a pair of graphs, we explicitly obtain node-level correspondence via computing a soft assignment matrix to match the nodes between two graphs. Based on the soft assignments, we transform the adjacency and node feature matrices of one graph, so that the transformed graph is aligned with the other graph. In this way, any pair of graphs can be mixed directly to generate an augmented graph. We conduct systematic experiments to show that S-Mixup can improve the performance and generalization of graph neural networks (GNNs) on various graph classification tasks. In addition, we show that S-Mixup can increase the robustness of GNNs against noisy labels."
                },
                {
                    "title": "Editable Graph Neural Network for Node Classifications",
                    "abstract": "Despite Graph Neural Networks (GNNs) have achieved prominent success in many graph-based learning problem, such as credit risk assessment in financial networks and fake news detection in social networks. However, the trained GNNs still make errors and these errors may cause serious negative impact on society. \\textit{Model editing}, which corrects the model behavior on wrongly predicted target samples while leaving model predictions unchanged on unrelated samples, has garnered significant interest in the fields of computer vision and natural language processing. However, model editing for graph neural networks (GNNs) is rarely explored, despite GNNs' widespread applicability. To fill the gap, we first observe that existing model editing methods significantly deteriorate prediction accuracy (up to $50\\%$ accuracy drop) in GNNs while a slight accuracy drop in multi-layer perception (MLP). The rationale behind this observation is that the node aggregation in GNNs will spread the editing effect throughout the whole graph. This propagation pushes the node representation far from its original one. Motivated by this observation, we propose \\underline{E}ditable \\underline{G}raph \\underline{N}eural \\underline{N}etworks (EGNN), a neighbor propagation-free approach to correct the model prediction on misclassified nodes. Specifically, EGNN simply stitches an MLP to the underlying GNNs, where the weights of GNNs are frozen during model editing. In this way, EGNN disables the propagation during editing while still utilizing the neighbor propagation scheme for node prediction to obtain satisfactory results. Experiments demonstrate that EGNN outperforms existing baselines in terms of effectiveness (correcting wrong predictions with lower accuracy drop), generalizability (correcting wrong predictions for other similar nodes), and efficiency (low training time and memory) on various graph datasets."
                },
                {
                    "title": "Hierarchy-Aware Multi-Hop Question Answering over Knowledge Graphs",
                    "abstract": "Knowledge graphs (KGs) have been widely used to enhance complex question answering (QA). To understand complex questions, existing studies employ language models (LMs) to encode contexts. Despite the simplicity, they neglect the latent relational information among question concepts and answers in KGs. While question concepts ubiquitously present hyponymy at the semantic level, e.g., mammals and animals, this feature is identically reflected in the hierarchical relations in KGs, e.g., a_type_of. Therefore, we are motivated to explore comprehensive reasoning by the hierarchical structures in KGs to help understand questions. However, it is non-trivial to reason over tree-like structures compared with chained paths. Moreover, identifying appropriate hierarchies relies on expertise. To this end, we propose HamQA, a novel Hierarchy-aware multi-hop Question Answering framework on knowledge graphs, to effectively align the mutual hierarchical information between question contexts and KGs. The entire learning is conducted in Hyperbolic space, inspired by its advantages of embedding hierarchical structures. Specifically, (i) we design a context-aware graph attentive network to capture context information. (ii) Hierarchical structures are continuously preserved in KGs by minimizing the Hyperbolic geodesic distances. The comprehensive reasoning is conducted to jointly train both components and provide a top-ranked candidate as an optimal answer. We achieve a higher ranking than the state-of-the-art multi-hop baselines on the official OpenBookQA leaderboard with an accuracy of 85%."
                },
                {
                    "title": "Transformer-Patcher: One Mistake worth One Neuron",
                    "abstract": "Large Transformer-based Pretrained Language Models (PLMs) dominate almost all Natural Language Processing (NLP) tasks. Nevertheless, they still make mistakes from time to time. For a model deployed in an industrial environment, fixing these mistakes quickly and robustly is vital to improve user experiences. Previous works formalize such problems as Model Editing (ME) and mostly focus on fixing one mistake. However, the one-mistake-fixing scenario is not an accurate abstraction of the real-world challenge. In the deployment of AI services, there are ever-emerging mistakes, and the same mistake may recur if not corrected in time. Thus a preferable solution is to rectify the mistakes as soon as they appear nonstop. Therefore, we extend the existing ME into Sequential Model Editing (SME) to help develop more practical editing methods. Our study shows that most current ME methods could yield unsatisfying results in this scenario. We then introduce Transformer-Patcher, a novel model editor that can shift the behavior of transformer-based models by simply adding and training a few neurons in the last Feed-Forward Network layer. Experimental results on both classification and generation tasks show that Transformer-Patcher can successively correct up to thousands of errors (Reliability) and generalize to their equivalent inputs (Generality) while retaining the model's accuracy on irrelevant inputs (Locality). Our method outperforms previous fine-tuning and HyperNetwork-based methods and achieves state-of-the-art performance for Sequential Model Editing (SME). The code is available at https://github.com/ZeroYuHuang/Transformer-Patcher."
                },
                {
                    "title": "A Comprehensive Study on Large-Scale Graph Training: Benchmarking and Rethinking",
                    "abstract": "Large-scale graph training is a notoriously challenging problem for graph neural networks (GNNs). Due to the nature of evolving graph structures into the training process, vanilla GNNs usually fail to scale up, limited by the GPU memory space. Up to now, though numerous scalable GNN architectures have been proposed, we still lack a comprehensive survey and fair benchmark of this reservoir to find the rationale for designing scalable GNNs. To this end, we first systematically formulate the representative methods of large-scale graph training into several branches and further establish a fair and consistent benchmark for them by a greedy hyperparameter searching. In addition, regarding efficiency, we theoretically evaluate the time and space complexity of various branches and empirically compare them w.r.t GPU memory usage, throughput, and convergence. Furthermore, We analyze the pros and cons for various branches of scalable GNNs and then present a new ensembling training manner, named EnGCN, to address the existing issues. Our code is available at https://github.com/VITA-Group/Large_Scale_GCN_Benchmarking."
                },
                {
                    "title": "Class Gradient Projection For Continual Learning",
                    "abstract": "Catastrophic forgetting is one of the most critical challenges in Continual Learning (CL). Recent approaches tackle this problem by projecting the gradient update orthogonal to the gradient subspace of existing tasks. While the results are remarkable, those approaches ignore the fact that these calculated gradients are not guaranteed to be orthogonal to the gradient subspace of each class due to the class deviation in tasks, e.g., distinguishing \"Man\" from \"Sea\" v.s. differentiating \"Boy\" from \"Girl\". Therefore, this strategy may still cause catastrophic forgetting for some classes. In this paper, we propose Class Gradient Projection (CGP), which calculates the gradient subspace from individual classes rather than tasks. Gradient update orthogonal to the gradient subspace of existing classes can be effectively utilized to minimize interference from other classes. To improve the generalization and efficiency, we further design a Base Refining (BR) algorithm to combine similar classes and refine class bases dynamically. Moreover, we leverage a contrastive learning method to improve the model's ability to handle unseen tasks. Extensive experiments on benchmark datasets demonstrate the effectiveness of our proposed approach. It improves the previous methods by 2.0% on the CIFAR-100 dataset. The code is available at https://github.com/zackschen/CGP."
                },
                {
                    "title": "Memory-Based Model Editing at Scale",
                    "abstract": "Even the largest neural networks make errors, and once-correct predictions can become invalid as the world changes. Model editors make local updates to the behavior of base (pre-trained) models to inject updated knowledge or correct undesirable behaviors. Existing model editors have shown promise, but also suffer from insufficient expressiveness: they struggle to accurately model an edit's intended scope (examples affected by the edit), leading to inaccurate predictions for test inputs loosely related to the edit, and they often fail altogether after many edits. As a higher-capacity alternative, we propose Semi-Parametric Editing with a Retrieval-Augmented Counterfactual Model (SERAC), which stores edits in an explicit memory and learns to reason over them to modulate the base model's predictions as needed. To enable more rigorous evaluation of model editors, we introduce three challenging language model editing problems based on question answering, fact-checking, and dialogue generation. We find that only SERAC achieves high performance on all three problems, consistently outperforming existing approaches to model editing by a significant margin. Code, data, and additional project information will be made available at https://sites.google.com/view/serac-editing."
                },
                {
                    "title": "G-Mixup: Graph Data Augmentation for Graph Classification",
                    "abstract": "This work develops \\emph{mixup for graph data}. Mixup has shown superiority in improving the generalization and robustness of neural networks by interpolating features and labels between two random samples. Traditionally, Mixup can work on regular, grid-like, and Euclidean data such as image or tabular data. However, it is challenging to directly adopt Mixup to augment graph data because different graphs typically: 1) have different numbers of nodes; 2) are not readily aligned; and 3) have unique typologies in non-Euclidean space. To this end, we propose $\\mathcal{G}$-Mixup to augment graphs for graph classification by interpolating the generator (i.e., graphon) of different classes of graphs. Specifically, we first use graphs within the same class to estimate a graphon. Then, instead of directly manipulating graphs, we interpolate graphons of different classes in the Euclidean space to get mixed graphons, where the synthetic graphs are generated through sampling based on the mixed graphons. Extensive experiments show that $\\mathcal{G}$-Mixup substantially improves the generalization and robustness of GNNs."
                },
                {
                    "title": "FMP: Toward Fair Graph Message Passing against Topology Bias",
                    "abstract": "Despite recent advances in achieving fair representations and predictions through regularization, adversarial debiasing, and contrastive learning in graph neural networks (GNNs), the working mechanism (i.e., message passing) behind GNNs inducing unfairness issue remains unknown. In this work, we theoretically and experimentally demonstrate that representative aggregation in message-passing schemes accumulates bias in node representation due to topology bias induced by graph topology. Thus, a \\textsf{F}air \\textsf{M}essage \\textsf{P}assing (FMP) scheme is proposed to aggregate useful information from neighbors but minimize the effect of topology bias in a unified framework considering graph smoothness and fairness objectives. The proposed FMP is effective, transparent, and compatible with back-propagation training. An acceleration approach on gradient calculation is also adopted to improve algorithm efficiency. Experiments on node classification tasks demonstrate that the proposed FMP outperforms the state-of-the-art baselines in effectively and efficiently mitigating bias on three real-world datasets."
                },
                {
                    "title": "Fast Model Editing at Scale",
                    "abstract": "While large pre-trained models have enabled impressive results on a variety of downstream tasks, the largest existing models still make errors, and even accurate predictions may become outdated over time. Because detecting all such failures at training time is impossible, enabling both developers and end users of such models to correct inaccurate outputs while leaving the model otherwise intact is desirable. However, the distributed, black-box nature of the representations learned by large neural networks makes producing such targeted edits difficult. If presented with only a single problematic input and new desired output, fine-tuning approaches tend to overfit; other editing algorithms are either computationally infeasible or simply ineffective when applied to very large models. To enable easy post-hoc editing at scale, we propose Model Editor Networks using Gradient Decomposition (MEND), a collection of small auxiliary editing networks that use a single desired input-output pair to make fast, local edits to a pre-trained model's behavior. MEND learns to transform the gradient obtained by standard fine-tuning, using a low-rank decomposition of the gradient to make the parameterization of this transformation tractable. MEND can be trained on a single GPU in less than a day even for 10 billion+ parameter models; once trained MEND enables rapid application of new edits to the pre-trained model. Our experiments with T5, GPT, BERT, and BART models show that MEND is the only approach to model editing that effectively edits the behavior of models with more than 10 billion parameters. Code and data available at https://sites.google.com/view/mend-editing."
                },
                {
                    "title": "Graph Condensation for Graph Neural Networks",
                    "abstract": "Given the prevalence of large-scale graphs in real-world applications, the storage and time for training neural models have raised increasing concerns. To alleviate the concerns, we propose and study the problem of graph condensation for graph neural networks (GNNs). Specifically, we aim to condense the large, original graph into a small, synthetic and highly-informative graph, such that GNNs trained on the small graph and large graph have comparable performance. We approach the condensation problem by imitating the GNN training trajectory on the original graph through the optimization of a gradient matching loss and design a strategy to condense node futures and structural information simultaneously. Extensive experiments have demonstrated the effectiveness of the proposed framework in condensing different graph datasets into informative smaller graphs. In particular, we are able to approximate the original test accuracy by 95.3% on Reddit, 99.8% on Flickr and 99.0% on Citeseer, while reducing their graph size by more than 99.9%, and the condensed graphs can be used to train various GNN architectures.Code is released at https://github.com/ChandlerBang/GCond."
                },
                {
                    "title": "Gradient Distribution Alignment Certificates Better Adversarial Domain Adaptation",
                    "abstract": "The latest heuristic for handling the domain shift in un-supervised domain adaptation tasks is to reduce the data distribution discrepancy using adversarial learning. Recent studies improve the conventional adversarial domain adaptation methods with discriminative information by integrating the classifier\u2019s outputs into distribution divergence measurement. However, they still suffer from the equilibrium problem of adversarial learning in which even if the discriminator is fully confused, sufficient similarity between two distributions cannot be guaranteed. To overcome this problem, we propose a novel approach named feature gradient distribution alignment (FGDA)1. We demonstrate the rationale of our method both theoretically and empirically. In particular, we show that the distribution discrepancy can be reduced by constraining feature gradients of two domains to have similar distributions. Meanwhile, our method enjoys a theoretical guarantee that a tighter error upper bound for target samples can be obtained than that of conventional adversarial domain adaptation methods. By integrating the proposed method with existing adversarial domain adaptation models, we achieve state-of-the-art performance on two real-world benchmark datasets."
                },
                {
                    "title": "Cross-Domain Gradient Discrepancy Minimization for Unsupervised Domain Adaptation",
                    "abstract": "Unsupervised Domain Adaptation (UDA) aims to generalize the knowledge learned from a well-labeled source domain to an unlabled target domain. Recently, adversarial domain adaptation with two distinct classifiers (biclassifier) has been introduced into UDA which is effective to align distributions between different domains. Previous bi-classifier adversarial learning methods only focus on the similarity between the outputs of two distinct classifiers. However, the similarity of the outputs cannot guarantee the accuracy of target samples, i.e., traget samples may match to wrong categories even if the discrepancy between two classifiers is small. To challenge this issue, in this paper, we propose a cross-domain gradient discrepancy minimization (CGDM) method which explicitly minimizes the discrepancy of gradients generated by source samples and target samples. Specifically, the gradient gives a cue for the semantic information of target samples so it can be used as a good supervision to improve the accuracy of target samples. In order to compute the gradient signal of target smaples, we further obtain target pseudo labels through a clustering-based self-supervised learning. Extensive experiments on three widely used UDA datasets show that our method surpasses many previous state-of-the-arts."
                },
                {
                    "title": "Editing Factual Knowledge in Language Models",
                    "abstract": "The factual knowledge acquired during pre-training and stored in the parameters of Language Models (LMs) can be useful in downstream tasks (e.g., question answering or textual inference). However, some facts can be incorrectly induced or become obsolete over time. We present KnowledgeEditor, a method which can be used to edit this knowledge and, thus, fix \u2018bugs\u2019 or unexpected predictions without the need for expensive re-training or fine-tuning. Besides being computationally efficient, KnowledgeEditordoes not require any modifications in LM pre-training (e.g., the use of meta-learning). In our approach, we train a hyper-network with constrained optimization to modify a fact without affecting the rest of the knowledge; the trained hyper-network is then used to predict the weight update at test time. We show KnowledgeEditor\u2019s efficacy with two popular architectures and knowledge-intensive tasks: i) a BERT model fine-tuned for fact-checking, and ii) a sequence-to-sequence BART model for question answering. With our method, changing a prediction on the specific wording of a query tends to result in a consistent change in predictions also for its paraphrases. We show that this can be further encouraged by exploiting (e.g., automatically-generated) paraphrases during training. Interestingly, our hyper-network can be regarded as a \u2018probe\u2019 revealing which components need to be changed to manipulate factual knowledge; our analysis shows that the updates tend to be concentrated on a small subset of components. Source code available at https://github.com/nicola-decao/KnowledgeEditor"
                },
                {
                    "title": "Gradient Projection Memory for Continual Learning",
                    "abstract": "The ability to learn continually without forgetting the past tasks is a desired attribute for artificial learning systems. Existing approaches to enable such learning in artificial neural networks usually rely on network growth, importance based weight update or replay of old data from the memory. In contrast, we propose a novel approach where a neural network learns new tasks by taking gradient steps in the orthogonal direction to the gradient subspaces deemed important for the past tasks. We find the bases of these subspaces by analyzing network representations (activations) after learning each task with Singular Value Decomposition (SVD) in a single shot manner and store them in the memory as Gradient Projection Memory (GPM). With qualitative and quantitative analyses, we show that such orthogonal gradient descent induces minimum to no interference with the past tasks, thereby mitigates forgetting. We evaluate our algorithm on diverse image classification datasets with short and long sequences of tasks and report better or on-par performance compared to the state-of-the-art approaches."
                },
                {
                    "title": "Open Graph Benchmark: Datasets for Machine Learning on Graphs",
                    "abstract": "We present the Open Graph Benchmark (OGB), a diverse set of challenging and realistic benchmark datasets to facilitate scalable, robust, and reproducible graph machine learning (ML) research. OGB datasets are large-scale (up to 100+ million nodes and 1+ billion edges), encompass multiple important graph ML tasks, and cover a diverse range of domains, ranging from social and information networks to biological networks, molecular graphs, source code ASTs, and knowledge graphs. For each dataset, we provide a unified evaluation protocol using meaningful application-specific data splits and evaluation metrics. In addition to building the datasets, we also perform extensive benchmark experiments for each dataset. Our experiments suggest that OGB datasets present significant challenges of scalability to large-scale graphs and out-of-distribution generalization under realistic data splits, indicating fruitful opportunities for future research. Finally, OGB provides an automated end-to-end graph ML pipeline that simplifies and standardizes the process of graph data loading, experimental setup, and model evaluation. OGB will be regularly updated and welcomes inputs from the community. OGB datasets as well as data loaders, evaluation scripts, baseline code, and leaderboards are publicly available at this https URL ."
                },
                {
                    "title": "Editable Neural Networks",
                    "abstract": "These days deep neural networks are ubiquitously used in a wide range of tasks, from image classification and machine translation to face identification and self-driving cars. In many applications, a single model error can lead to devastating financial, reputational and even life-threatening consequences. Therefore, it is crucially important to correct model mistakes quickly as they appear. In this work, we investigate the problem of neural network editing - how one can efficiently patch a mistake of the model on a particular sample, without influencing the model behavior on other samples. Namely, we propose Editable Training, a model-agnostic training technique that encourages fast editing of the trained model. We empirically demonstrate the effectiveness of this method on large-scale image classification and machine translation tasks."
                },
                {
                    "title": "Orthogonal Gradient Descent for Continual Learning",
                    "abstract": "Neural networks are achieving state of the art and sometimes super-human performance on learning tasks across a variety of domains. Whenever these problems require learning in a continual or sequential manner, however, neural networks suffer from the problem of catastrophic forgetting; they forget how to solve previous tasks after being trained on a new task, despite having the essential capacity to solve both tasks if they were trained on both simultaneously. In this paper, we propose to address this issue from a parameter space perspective and study an approach to restrict the direction of the gradient updates to avoid forgetting previously-learned data. We present the Orthogonal Gradient Descent (OGD) method, which accomplishes this goal by projecting the gradients from new tasks onto a subspace in which the neural network output on previous task does not change and the projected gradient is still in a useful direction for learning the new task. Our approach utilizes the high capacity of a neural network more efficiently and does not require storing the previously learned data that might raise privacy concerns. Experiments on common benchmarks reveal the effectiveness of the proposed OGD method."
                },
                {
                    "title": "Meta-Learning with Implicit Gradients",
                    "abstract": "A core capability of intelligent systems is the ability to quickly learn new tasks by drawing on prior experience. Gradient (or optimization) based meta-learning has recently emerged as an effective approach for few-shot learning. In this formulation, meta-parameters are learned in the outer loop, while task-specific models are learned in the inner-loop, by using only a small amount of data from the current task. A key challenge in scaling these approaches is the need to differentiate through the inner loop learning process, which can impose considerable computational and memory burdens. By drawing upon implicit differentiation, we develop the implicit MAML algorithm, which depends only on the solution to the inner level optimization and not the path taken by the inner loop optimizer. This effectively decouples the meta-gradient computation from the choice of inner loop optimizer. As a result, our approach is agnostic to the choice of inner loop optimizer and can gracefully handle many gradient steps without vanishing gradients or memory constraints. Theoretically, we prove that implicit MAML can compute accurate meta-gradients with a memory footprint that is, up to small constant factors, no more than that which is required to compute a single inner loop gradient and at no overall increase in the total computational cost. Experimentally, we show that these benefits of implicit MAML translate into empirical gains on few-shot image recognition benchmarks."
                },
                {
                    "title": "GraphSAINT: Graph Sampling Based Inductive Learning Method",
                    "abstract": "Graph Convolutional Networks (GCNs) are powerful models for learning representations of attributed this http URL scale GCNs to large graphs, state-of-the-art methods use various layer sampling techniques to alleviate the \"neighbor explosion\" problem during minibatch training. Here we proposeGraphSAINT, a graph sampling based inductive learning method that improves training efficiency in a fundamentally different way. By a change of perspective, GraphSAINT constructs minibatches by sampling the training graph, rather than the nodes or edges across GCN layers. Each iteration, a complete GCN is built from the properly sampled subgraph. Thus, we ensure fixed number of well-connected nodes in all layers. We further propose normalization technique to eliminate bias, and sampling algorithms for variance reduction. Importantly, we can decouple the sampling process from the forward and backward propagation of training, and extend GraphSAINT with other graph samplers and GCN variants. Comparing with strong baselines using layer sampling, GraphSAINT demonstrates superior performance in both accuracy and training time on four large graphs."
                },
                {
                    "title": "Adaptive Gradient-Based Meta-Learning Methods",
                    "abstract": "We build a theoretical framework for designing and understanding practical meta-learning methods that integrates sophisticated formalizations of task-similarity with the extensive literature on online convex optimization and sequential prediction algorithms. Our approach enables the task-similarity to be learned adaptively, provides sharper transfer-risk bounds in the setting of statistical learning-to-learn, and leads to straightforward derivations of average-case regret bounds for efficient algorithms in settings where the task-environment changes dynamically or the tasks share a certain geometric structure. We use our theory to modify several popular meta-learning algorithms and improve their meta-test-time performance on standard problems in few-shot learning and federated learning."
                },
                {
                    "title": "Simplifying Graph Convolutional Networks",
                    "abstract": "Graph Convolutional Networks (GCNs) and their variants have experienced significant attention and have become the de facto methods for learning graph representations. GCNs derive inspiration primarily from recent deep learning approaches, and as a result, may inherit unnecessary complexity and redundant computation. In this paper, we reduce this excess complexity through successively removing nonlinearities and collapsing weight matrices between consecutive layers. We theoretically analyze the resulting linear model and show that it corresponds to a fixed low-pass filter followed by a linear classifier. Notably, our experimental evaluation demonstrates that these simplifications do not negatively impact accuracy in many downstream applications. Moreover, the resulting model scales to larger datasets, is naturally interpretable, and yields up to two orders of magnitude speedup over FastGCN."
                },
                {
                    "title": "Pitfalls of Graph Neural Network Evaluation",
                    "abstract": "Semi-supervised node classification in graphs is a fundamental problem in graph mining, and the recently proposed graph neural networks (GNNs) have achieved unparalleled results on this task. Due to their massive success, GNNs have attracted a lot of attention, and many novel architectures have been put forward. In this paper we show that existing evaluation strategies for GNN models have serious shortcomings. We show that using the same train/validation/test splits of the same datasets, as well as making significant changes to the training procedure (e.g. early stopping criteria) precludes a fair comparison of different architectures. We perform a thorough empirical evaluation of four prominent GNN models and show that considering different splits of the data leads to dramatically different rankings of models. Even more importantly, our findings suggest that simpler GNN architectures are able to outperform the more sophisticated ones if the hyperparameters and the training procedure are tuned fairly for all models."
                },
                {
                    "title": "Graph Convolutional Neural Networks for Web-Scale Recommender Systems",
                    "abstract": "Recent advancements in deep neural networks for graph-structured data have led to state-of-the-art performance on recommender system benchmarks. However, making these methods practical and scalable to web-scale recommendation tasks with billions of items and hundreds of millions of users remains an unsolved challenge. Here we describe a large-scale deep recommendation engine that we developed and deployed at Pinterest. We develop a data-efficient Graph Convolutional Network (GCN) algorithm, which combines efficient random walks and graph convolutions to generate embeddings of nodes (i.e., items) that incorporate both graph structure as well as node feature information. Compared to prior GCN approaches, we develop a novel method based on highly efficient random walks to structure the convolutions and design a novel training strategy that relies on harder-and-harder training examples to improve robustness and convergence of the model. We also develop an efficient MapReduce model inference algorithm to generate embeddings using a trained model. Overall, we can train on and embed graphs that are four orders of magnitude larger than typical GCN implementations. We show how GCN embeddings can be used to make high-quality recommendations in various settings at Pinterest, which has a massive underlying graph with 3 billion nodes representing pins and boards, and 17 billion edges. According to offline metrics, user studies, as well as A/B tests, our approach generates higher-quality recommendations than comparable deep learning based systems. To our knowledge, this is by far the largest application of deep graph embeddings to date and paves the way for a new generation of web-scale recommender systems based on graph convolutional architectures."
                },
                {
                    "title": "Graph Attention Networks",
                    "abstract": "We present graph attention networks (GATs), novel neural network architectures that operate on graph-structured data, leveraging masked self-attentional layers to address the shortcomings of prior methods based on graph convolutions or their approximations. By stacking layers in which nodes are able to attend over their neighborhoods' features, we enable (implicitly) specifying different weights to different nodes in a neighborhood, without requiring any kind of costly matrix operation (such as inversion) or depending on knowing the graph structure upfront. In this way, we address several key challenges of spectral-based graph neural networks simultaneously, and make our model readily applicable to inductive as well as transductive problems. Our GAT models have achieved or matched state-of-the-art results across four established transductive and inductive graph benchmarks: the Cora, Citeseer and Pubmed citation network datasets, as well as a protein-protein interaction dataset (wherein test graphs remain unseen during training)."
                },
                {
                    "title": "Fake News Detection on Social Media: A Data Mining Perspective",
                    "abstract": "Social media for news consumption is a double-edged sword. On the one hand, its low cost, easy access, and rapid dissemination of information lead people to seek out and consume news from social media. On the other hand, it enables the wide spread of \\fake news\", i.e., low quality news with intentionally false information. The extensive spread of fake news has the potential for extremely negative impacts on individuals and society. Therefore, fake news detection on social media has recently become an emerging research that is attracting tremendous attention. Fake news detection on social media presents unique characteristics and challenges that make existing detection algorithms from traditional news media ine ective or not applicable. First, fake news is intentionally written to mislead readers to believe false information, which makes it difficult and nontrivial to detect based on news content; therefore, we need to include auxiliary information, such as user social engagements on social media, to help make a determination. Second, exploiting this auxiliary information is challenging in and of itself as users' social engagements with fake news produce data that is big, incomplete, unstructured, and noisy. Because the issue of fake news detection on social media is both challenging and relevant, we conducted this survey to further facilitate research on the problem. In this survey, we present a comprehensive review of detecting fake news on social media, including fake news characterizations on psychology and social theories, existing algorithms from a data mining perspective, evaluation metrics and representative datasets. We also discuss related research areas, open problems, and future research directions for fake news detection on social media."
                },
                {
                    "title": "Inductive Representation Learning on Large Graphs",
                    "abstract": "Low-dimensional embeddings of nodes in large graphs have proved extremely useful in a variety of prediction tasks, from content recommendation to identifying protein functions. However, most existing approaches require that all nodes in the graph are present during training of the embeddings; these previous approaches are inherently transductive and do not naturally generalize to unseen nodes. Here we present GraphSAGE, a general, inductive framework that leverages node feature information (e.g., text attributes) to efficiently generate node embeddings for previously unseen data. Instead of training individual embeddings for each node, we learn a function that generates embeddings by sampling and aggregating features from a node's local neighborhood. Our algorithm outperforms strong baselines on three inductive node-classification benchmarks: we classify the category of unseen nodes in evolving information graphs based on citation and Reddit post data, and we show that our algorithm generalizes to completely unseen graphs using a multi-graph dataset of protein-protein interactions."
                },
                {
                    "title": "Neural Message Passing for Quantum Chemistry",
                    "abstract": "Supervised learning on molecules has incredible potential to be useful in chemistry, drug discovery, and materials science. Luckily, several promising and closely related neural network models invariant to molecular symmetries have already been described in the literature. These models learn a message passing algorithm and aggregation procedure to compute a function of their entire input graph. At this point, the next step is to find a particularly effective variant of this general approach and apply it to chemical prediction benchmarks until we either solve them or reach the limits of the approach. In this paper, we reformulate existing models into a single common framework we call Message Passing Neural Networks (MPNNs) and explore additional novel variations within this framework. Using MPNNs we demonstrate state of the art results on an important molecular property prediction benchmark; these results are strong enough that we believe future work should focus on datasets with larger molecules or more accurate ground truth labels."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Quadratic programming",
                    "abstract": "Optimization problems in which a quadratic objective function is optimized subject to linear constraints on the parameters are known as quadratic programming problems (QPs). This focus article reviews algorithms for convex QPs (in which the objective is a convex function) and provides pointers to various online resources about QPs. WIREs Comput Stat 2015, 7:153\u2013159. doi: 10.1002/wics.1344"
                },
                {
                    "title": "Collective Classification in Network Data",
                    "abstract": "Many real-world applications produce networked data such as the world-wide web (hypertext documents connected via hyperlinks), social networks (for example, people connected by friendship links), communication networks (computers connected via communication links) and biological networks (for example, protein interaction networks). A recent focus in machine learning research has been to extend traditional machine learning classification techniques to classify nodes in such networks. In this article, we provide a brief introduction to this area of research and how it has progressed during the past decade. We introduce four of the most widely used inference algorithms for classifying networked data and empirically compare them on both synthetic and real-world data."
                },
                {
                    "title": "GNNs Also Deserve Editing, and They Need It More Than Once",
                    "abstract": "Suppose a self-driving car is crashing into pedes-trians, or a chatbot is instructing its users to conduct criminal wrongdoing; the stakeholders of such products will undoubtedly want to patch these catastrophic errors as soon as possible. To address such concerns, Model Editing: the study of efficiently patching model behaviors without significantly altering their general performance, has seen considerable activity, with hundreds of editing techniques developed in various domains such as CV and NLP. However, the graph learning community has objectively fallen behind with only a few Graph Neural Network-compatible \u2014 and just one GNN-specific \u2014 model editing methods available , where all of which are limited in their practical scope. We argue that the impracticality of these methods lies in their lack of Sequential Editing Robustness : the ability to edit multiple errors sequentially, and therefore fall short in effectiveness, as this setup mirrors how errors are discovered and addressed in the real world. In this paper, we delve into the specific reasons behind the difficulty of editing GNNs in succession and observe the root cause to be model overfitting. We subsequently propose a simple yet effective solution \u2014 SEED-GNN \u2014 by leveraging overfit-prevention techniques in a GNN-specific context to derive the first and only GNN model editing method that scales practically. Additionally, we formally frame the task paradigm of GNN editing and hope to inspire future research in this crucial but currently over-looked field. Please refer to our GitHub repository for code and checkpoints."
                },
                {
                    "title": "Fair Graph Distillation",
                    "abstract": "As graph neural networks (GNNs) struggle with large-scale graphs due to high computational demands, graph data distillation promises to alleviate this issue by distilling a large real graph into a smaller distilled graph while maintaining comparable prediction performance for GNNs trained on both graphs. However, we observe that GNNs trained on distilled graphs may exhibit more severe group fairness issues than GNNs trained on real graphs for vanilla and fair GNNs training. Motivated by these observations, we propose fair graph distillation (FGD), an advanced graph distillation approach to generate fair distilled graphs. The challenge lies in the deficiency of sensitive attributes for nodes in the distilled graph, making most debiasing methods (e.g., regularization and adversarial debiasing) intractable for distilled graphs. We develop a simple yet effective bias metric, named coherence, for distilled graphs. Based on the proposed coherence metric, we introduce a framework for fair graph distillation using a bi-level optimization algorithm. Extensive experiments demonstrate that the proposed algorithm can achieve better prediction performance-fairness trade-offs across various datasets and GNN architectures."
                },
                {
                    "title": "Explicitly Capturing Relations between Entity Mentions via Graph Neural Networks for Domain-specific Named Entity Recognition",
                    "abstract": "Named entity recognition (NER) is well studied for the general domain, and recent systems have achieved human-level performance for identifying common entity types. However, the NER performance is still moderate for specialized domains that tend to feature complicated contexts and jargonistic entity types. To address these challenges, we propose explicitly connecting entity mentions based on both global coreference relations and local dependency relations for building better entity mention representations. In our experiments, we incorporate entity mention relations by Graph Neural Networks and show that our system noticeably improves the NER performance on two datasets from different domains. We further show that the proposed lightweight system can effectively elevate the NER performance to a higher level even when only a tiny amount of labeled data is available, which is desirable for domain-specific NER."
                },
                {
                    "title": "Reconstructing Event Regions for Event Extraction via Graph Attention Networks",
                    "abstract": "Event information is usually scattered across multiple sentences within a document. The local sentence-level event extractors often yield many noisy event role filler extractions in the absence of a broader view of the document-level context. Filtering spurious extractions and aggregating event information in a document remains a challenging problem. Following the observation that a document has several relevant event regions densely populated with event role fillers, we build graphs with candidate role filler extractions enriched by sentential embeddings as nodes, and use graph attention networks to identify event regions in a document and aggregate event information. We characterize edges between candidate extractions in a graph into rich vector representations to facilitate event region identification. The experimental results on two datasets of two languages show that our approach yields new state-of-the-art performance for the challenging event extraction task."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively edit Graph Neural Networks (GNNs) to rectify critical errors in predictions without adversely affecting the model's performance on unrelated inputs?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the reliability of GNNs in high-stakes applications, such as credit risk assessment and fake news detection. By developing effective model editing techniques, we can enhance the robustness of GNNs, leading to safer and more accurate decision-making processes. This research could pave the way for future advancements in GNN applications, enabling them to adapt dynamically to errors and improve their overall performance. Additionally, it could lead to practical applications in various domains where GNNs are deployed, ensuring that they can be trusted to make critical decisions.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complex nature of GNNs and their message-passing mechanism. Editing the behavior of a single node can inadvertently affect the entire graph due to the interconnectedness of nodes, leading to unintended consequences. Naive approaches may fail because they do not account for the ripple effects of changes made to one node on others. Furthermore, the inconsistency between the gradients of the cross-entropy loss for the target node and the training nodes complicates direct fine-tuning, as it can degrade performance on the training nodes. Overcoming these technical and theoretical obstacles requires a nuanced understanding of GNN dynamics and innovative editing strategies.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely overlooked the specific challenges of model editing in GNNs, focusing instead on applications in computer vision and language models. The lack of attention to the unique properties of graph data, such as the dense interconnectivity of nodes, has created a gap in effective solutions. Additionally, existing model editing frameworks often necessitate an additional training phase, which may not be feasible or effective for GNNs. Our approach differs by introducing the Gradient Rewiring method, which addresses the inconsistency in gradients and allows for more effective editing without compromising the model's overall performance.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the Gradient Rewiring method for Editable Graph Neural Network training (GRE). This approach includes calculating and storing the anchor gradient of the loss on the training nodes, which represents"
            }
        },
        "author_data": {
            "dfb4e7e2-247b-4f18-bc74-d0799aca836e": {
                "pk": "dfb4e7e2-247b-4f18-bc74-d0799aca836e",
                "name": "Zhimeng Jiang",
                "collaborators": [
                    "Chen Gong",
                    "Zhengyuan Xu",
                    "Xia Hu",
                    "Guanchu Wang",
                    "Xiaotian Han",
                    "Na Zou",
                    "Hongye Jin",
                    "Ninghao Liu",
                    "Rui Chen",
                    "Huiyuan Chen"
                ],
                "domain": [
                    "Optical Communication",
                    "Graph Neural Network",
                    "Machine Learning",
                    "Adversarial Learning"
                ],
                "publications": [
                    {
                        "title": "Clipping noise approximate analysis and power allocation for photon-detection-based DCO-OFDM and ACO-OFDM",
                        "abstract": "The clipping noise of the photon-level detector for both direct current-biased optical OFDM (DCO-OFDM) and asymmetrically clipped optical OFDM (ACO-OFDM) is investigated. Based on Bussgang theorem and central limit theorem (CLT), we obtain the approximate closed-form SNR of each subcarrier, based on which we further formulate the power allocation among the subcarriers. Numerical results show that the SNR obtained from theoretical analysis can well approximate that obtained from simulation results, and uniform power allocation suffices to perform close to the optimized power allocation from Genetic Algorithm (GA) with significantly reduced computational complexity."
                    },
                    {
                        "title": "Statistical Non-linear Model, Achievable Rates and Signal Detection for Photon-level Photomultiplier Receiver",
                        "abstract": "We characterize the practical receiver in a wide range of signal intensity for optical wireless communication, from discrete pulse regime to continuous waveform regime. We first propose a statistical non-linear model based on the photomultiplier tube (PMT) multi-stage amplification and Poisson channel, and then derive the optimal and tractable suboptimal duty cycle with peak-power and average-power constraints for on-off key (OOK) modulation in linear regime. Subsequently, a threshold-based classifier is proposed to distinguish the PMT working regimes based on the non-linear model. Moreover, we derive the approximate performance of mean power detection with infinite sampling rate and finite over-sampling rate in the linear regime based on small dead time and central-limit theorem. We also fomulate a signal model in the non-linar regime. Furthermore, the performance of mean power detection and photon counting detection with maximum likelihood (ML) detection for different sampling rates is evaluated from both theoretical and numerical perspectives. We can conclude that the sample interval equivalent to dead time is a good choice, and lower sampling rate would significantly degrade the performance."
                    },
                    {
                        "title": "G-Mixup: Graph Data Augmentation for Graph Classification",
                        "abstract": "This work develops \\emph{mixup for graph data}. Mixup has shown superiority in improving the generalization and robustness of neural networks by interpolating features and labels between two random samples. Traditionally, Mixup can work on regular, grid-like, and Euclidean data such as image or tabular data. However, it is challenging to directly adopt Mixup to augment graph data because different graphs typically: 1) have different numbers of nodes; 2) are not readily aligned; and 3) have unique typologies in non-Euclidean space. To this end, we propose $\\mathcal{G}$-Mixup to augment graphs for graph classification by interpolating the generator (i.e., graphon) of different classes of graphs. Specifically, we first use graphs within the same class to estimate a graphon. Then, instead of directly manipulating graphs, we interpolate graphons of different classes in the Euclidean space to get mixed graphons, where the synthetic graphs are generated through sampling based on the mixed graphons. Extensive experiments show that $\\mathcal{G}$-Mixup substantially improves the generalization and robustness of GNNs."
                    },
                    {
                        "title": "On the Sum-Rate Capacity of Poisson Multiple Access Channel with Non-Perfect Photon-Counting Receiver",
                        "abstract": "We first investigate two-user nonasymmetric sum-rate Poisson capacity with non-perfect photoncounting receiver under certain condition and demonstrate three possible transmission strategy, including only one active user and both active users, in sharp contrast to Gaussian multiple access channel (MAC) channel. The two-user capacity reduction due to photon-counting loss is characterized compared with that of continuous Poisson channel. We then study the symmetrical case based on two different methods, demonstrating that the optimal duty cycle for two users must be the same and unique, and the last method maybe can extend to multiple users. Furthermore, we analyze the sum-Rate capacity of Poisson multiple input single output (MISO) MAC. By converting a non-convex optimization problem with a large number of variables into a non-convex optimization problem with two variables, we show that the sum-rate capacity of the Poisson MISO MAC is equivalent to that of SISO under certain condition."
                    },
                    {
                        "title": "On the Achievable Rate and Capacity for a Sample-based Practical Photon-counting Receiver",
                        "abstract": "We investigate the achievable rate and capacity of a non-perfect photon-counting receiver. For the case of long symbol duration, the achievable rate under on-off keying modulation is investigated based on Kullback-Leibler (KL) divergence and Chernoff $\\alpha$-divergence. We prove the tightness of the derived bounds for large peak power with zero background radiation with exponential convergence rate, and for low peak power of order two convergence rate. For large peak power with fixed background radiation and low background radiation with fixed peak power, the proposed bound gap is a small positive value for low background radiation and large peak power, respectively. Moreover, we propose an approximation on the achievable rate in the low background radiation and long symbol duration regime, which is more accurate compared with the derived upper and lower bounds in the medium signal to noise ratio (SNR) regime. For the symbol duration that can be sufficiently small, the capacity and the optimal duty cycle are is investigated. We show that the capacity approaches that of continuous Poisson capacity as $T_s=\\tau\\to0$. The asymptotic capacity is analyzed for low and large peak power. Compared with the continuous Poisson capacity, the capacity of a non-perfect receiver is almost lossless and loss with attenuation for low peak power given zero background radiation and nonzero background radiation, respectively. For large peak power, the capacity with a non-perfect receiver converges, while that of continuous Poisson capacity channel linearly increases. The above theoretical results are extensively validated by numerical results."
                    },
                    {
                        "title": "Asynchronous Multiple Access in Optical Wireless Scattering Communication: Achievable Transmission Rates and Receiver Design",
                        "abstract": "We investigate the asynchronous multiple user access communication in optical wireless scattering communication, where different users transmit signals without perfect alignment in the time domain. Firstly, we characterize the received signal based on hidden markov model (HMM) such that the misalignment among different users can be characterized by the state transition. Then, we investigate the achievable rates based on that of the HMM and obtain the approximated solution using Monte Carlo method. We propose the channel estimation based on expectation-maximization (EM) algorithm. Furthermore, we adopt Viterbi and Bahl-Cocke-Jelinek-Raviv (BCJR) algorithms for joint iterative multi-user decoding. Numerical and experimental results illustrate the performance of proposed channel estimation, joint detection and decoding. It is seen from the experimental results that the proposed approaches perform close to the simulation results."
                    },
                    {
                        "title": "A Double-station Access Protocol for Optical Wireless Scattering Communication Networks",
                        "abstract": "We propose a double-station access protocol (DS-CSMA) with multiple backoff mechanism for optical wireless scattering communication networks (OWSCN). %, where two stations can transmit data to single destination simultaneously.can avoid the frames colliding with each other. Furthermore, we extend existing Bianchi Markov model into state transmission model to analyze the collision probability, throughput and average delay. For the application of protocol, we propose to optimize the initial contention window and indicator matrix to maximize throughput. Both numerical and simulation results imply that the proposed protocol can achieve higher throughput and lower transmission delay compared with state-of-art baseline."
                    },
                    {
                        "title": "Graph Mixup with Soft Alignments",
                        "abstract": "We study graph data augmentation by mixup, which has been used successfully on images. A key operation of mixup is to compute a convex combination of a pair of inputs. This operation is straightforward for grid-like data, such as images, but challenging for graph data. The key difficulty lies in the fact that different graphs typically have different numbers of nodes, and thus there lacks a node-level correspondence between graphs. In this work, we propose S-Mixup, a simple yet effective mixup method for graph classification by soft alignments. Specifically, given a pair of graphs, we explicitly obtain node-level correspondence via computing a soft assignment matrix to match the nodes between two graphs. Based on the soft assignments, we transform the adjacency and node feature matrices of one graph, so that the transformed graph is aligned with the other graph. In this way, any pair of graphs can be mixed directly to generate an augmented graph. We conduct systematic experiments to show that S-Mixup can improve the performance and generalization of graph neural networks (GNNs) on various graph classification tasks. In addition, we show that S-Mixup can increase the robustness of GNNs against noisy labels."
                    },
                    {
                        "title": "Chasing Fairness Under Distribution Shift: A Model Weight Perturbation Approach",
                        "abstract": "Fairness in machine learning has attracted increasing attention in recent years. The fairness methods improving algorithmic fairness for in-distribution data may not perform well under distribution shifts. In this paper, we first theoretically demonstrate the inherent connection between distribution shift, data perturbation, and model weight perturbation. Subsequently, we analyze the sufficient conditions to guarantee fairness (i.e., low demographic parity) for the target dataset, including fairness for the source dataset, and low prediction difference between the source and target datasets for each sensitive attribute group. Motivated by these sufficient conditions, we propose robust fairness regularization (RFR) by considering the worst case within the model weight perturbation ball for each sensitive attribute group. We evaluate the effectiveness of our proposed RFR algorithm on synthetic and real distribution shifts across various datasets. Experimental results demonstrate that RFR achieves better fairness-accuracy trade-off performance compared with several baselines. The source code is available at \\url{https://github.com/zhimengj0326/RFR_NeurIPS23}."
                    },
                    {
                        "title": "Adaptive Risk-Aware Bidding with Budget Constraint in Display Advertising",
                        "abstract": "Real-time bidding (RTB) has become a major paradigm of display advertising. Each ad impression generated from a user visit is auctioned in real time, where demand-side platform (DSP) automatically provides bid price usually relying on the ad impression value estimation and the optimal bid price determination. However, the current bid strategy overlooks large randomness of the user behaviors (e.g., click) and the cost uncertainty caused by the auction competition. In this work, we explicitly factor in the uncertainty of estimated ad impression values and model the risk preference of a DSP under a specific state and market environment via a sequential decision process. Specifically, we propose a novel adaptive risk-aware bidding algorithm with budget constraint via reinforcement learning, which is the first to simultaneously consider estimation uncertainty and the dynamic risk tendency of a DSP. We theoretically unveil the intrinsic relation between the uncertainty and the risk tendency based on value at risk (VaR). Consequently, we propose two instantiations to model risk tendency, including an expert knowledge-based formulation embracing three essential properties and an adaptive learning method based on self-supervised reinforcement learning. We conduct extensive experiments on public datasets and show that the proposed framework outperforms state-of-the-art methods in practical settings."
                    },
                    {
                        "title": "Chasing Fairness in Graphs: A GNN Architecture Perspective",
                        "abstract": "There has been significant progress in improving the performance of graph neural networks (GNNs) through enhancements in graph data, model architecture design, and training strategies. For fairness in graphs, recent studies achieve fair representations and predictions through either graph data pre-processing (e.g., node feature masking, and topology rewiring) or fair training strategies (e.g., regularization, adversarial debiasing, and fair contrastive learning). How to achieve fairness in graphs from the model architecture perspective is less explored. More importantly, GNNs exhibit worse fairness performance compared to multilayer perception since their model architecture (i.e., neighbor aggregation) amplifies biases. To this end, we aim to achieve fairness via a new GNN architecture. We propose \\textsf{F}air \\textsf{M}essage \\textsf{P}assing (FMP) designed within a unified optimization framework for GNNs. Notably, FMP \\textit{explicitly} renders sensitive attribute usage in \\textit{forward propagation} for node classification task using cross-entropy loss without data pre-processing. In FMP, the aggregation is first adopted to utilize neighbors' information and then the bias mitigation step explicitly pushes demographic group node presentation centers together. In this way, FMP scheme can aggregate useful information from neighbors and mitigate bias to achieve better fairness and prediction tradeoff performance. Experiments on node classification tasks demonstrate that the proposed FMP outperforms several baselines in terms of fairness and accuracy on three real-world datasets. The code is available in {\\url{https://github.com/zhimengj0326/FMP}}."
                    },
                    {
                        "title": "A 1Mbps Real-time NLOS UV Scattering Communication System with Receiver Diversity over 1km",
                        "abstract": "In the non-line of sight (NLOS) ultraviolet (UV) scattering communication, the received signals exhibit the characteristics of discrete photoelectrons due to the extremely large path loss. We design and demonstrate an NLOS UV scattering communication system in this work, where the receiver-side signal detection is designed based on a discrete-time Poisson channel model. In our system, a laser and multiple photomultiplier tubes are employed as the optical transmitter and detector, respectively. Furthermore, we design algorithms for pulse-counting, synchronization, channel estimation and $LLR$ computation for hardware realization in FPGA board. Simulation results are provided to evaluate the proposed system design and specify the system key parameters. We perform field tests for real-time communication with the transmission range over $1$km, where the system throughput reaches $1$Mbps."
                    },
                    {
                        "title": "Geometric Graph Representation Learning via Maximizing Rate Reduction",
                        "abstract": "Learning discriminative node representations benefits various downstream tasks in graph analysis such as community detection and node classification. Existing graph representation learning methods (e.g., based on random walk and contrastive learning) are limited to maximizing the local similarity of connected nodes. Such pair-wise learning schemes could fail to capture the global distribution of representations, since it has no explicit constraints on the global geometric properties of representation space. To this end, we propose Geometric Graph Representation Learning (G2R) to learn node representations in an unsupervised manner via maximizing rate reduction. In this way, G2R maps nodes in distinct groups (implicitly stored in the adjacency matrix) into different subspaces, while each subspace is compact and different subspaces are dispersedly distributed. G2R adopts a graph neural network as the encoder and maximizes the rate reduction with the adjacency matrix. Furthermore, we theoretically and empirically demonstrate that rate reduction maximization is equivalent to maximizing the principal angles between different subspaces. Experiments on real-world datasets show that G2R outperforms various baselines on node classification and community detection tasks."
                    },
                    {
                        "title": "Towards Generating Adversarial Examples on Mixed-type Data",
                        "abstract": "The existence of adversarial attacks (or adversarial examples) brings huge concern about the machine learning (ML) model's safety issues. For many safety-critical ML tasks, such as financial forecasting, fraudulent detection, and anomaly detection, the data samples are usually mixed-type, which contain plenty of numerical and categorical features at the same time. However, how to generate adversarial examples with mixed-type data is still seldom studied. In this paper, we propose a novel attack algorithm M-Attack, which can effectively generate adversarial examples in mixed-type data. Based on M-Attack, attackers can attempt to mislead the targeted classification model's prediction, by only slightly perturbing both the numerical and categorical features in the given data samples. More importantly, by adding designed regularizations, our generated adversarial examples can evade potential detection models, which makes the attack indeed insidious. Through extensive empirical studies, we validate the effectiveness and efficiency of our attack method and evaluate the robustness of existing classification models against our proposed attack. The experimental results highlight the feasibility of generating adversarial examples toward machine learning models in real-world applications."
                    },
                    {
                        "title": "GrowLength: Accelerating LLMs Pretraining by Progressively Growing Training Length",
                        "abstract": "The evolving sophistication and intricacies of Large Language Models (LLMs) yield unprecedented advancements, yet they simultaneously demand considerable computational resources and incur significant costs. To alleviate these challenges, this paper introduces a novel, simple, and effective method named ``\\growlength'' to accelerate the pretraining process of LLMs. Our method progressively increases the training length throughout the pretraining phase, thereby mitigating computational costs and enhancing efficiency. For instance, it begins with a sequence length of 128 and progressively extends to 4096. This approach enables models to process a larger number of tokens within limited time frames, potentially boosting their performance. In other words, the efficiency gain is derived from training with shorter sequences optimizing the utilization of resources. Our extensive experiments with various state-of-the-art LLMs have revealed that models trained using our method not only converge more swiftly but also exhibit superior performance metrics compared to those trained with existing methods. Furthermore, our method for LLMs pretraining acceleration does not require any additional engineering efforts, making it a practical solution in the realm of LLMs."
                    },
                    {
                        "title": "Towards Mitigating Dimensional Collapse of Representations in Collaborative Filtering",
                        "abstract": "Contrastive Learning (CL) has shown promising performance in collaborative filtering. The key idea is to generate augmentation-invariant embeddings by maximizing the Mutual Information between different augmented views of the same instance. However, we empirically observe that existing CL models suffer from the \\textsl{dimensional collapse} issue, where user/item embeddings only span a low-dimension subspace of the entire feature space. This suppresses other dimensional information and weakens the distinguishability of embeddings. Here we propose a non-contrastive learning objective, named nCL, which explicitly mitigates dimensional collapse of representations in collaborative filtering. Our nCL aims to achieve geometric properties of \\textsl{Alignment} and \\textsl{Compactness} on the embedding space. In particular, the alignment tries to push together representations of positive-related user-item pairs, while compactness tends to find the optimal coding length of user/item embeddings, subject to a given distortion. More importantly, our nCL does not require data augmentation nor negative sampling during training, making it scalable to large datasets. Experimental results demonstrate the superiority of our nCL."
                    }
                ]
            },
            "44eb4769-ea65-4fa3-8f3d-8062b963b538": {
                "pk": "44eb4769-ea65-4fa3-8f3d-8062b963b538",
                "name": "Zirui Liu",
                "collaborators": [
                    "Kaixiong Zhou",
                    "Xia Hu",
                    "Xiangyu Xie",
                    "Pingxin Zhang",
                    "Chunhao Tian",
                    "Yongjie Sun",
                    "Zhiling Zhou",
                    "Chengming Xu",
                    "Yanwei Fu",
                    "Xinwei Sun"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Molecular Representation Learning",
                    "Robustness in Neural Networks",
                    "Topological Data Analysis"
                ],
                "publications": [
                    {
                        "title": "Simulation of the Signal Propagation for Thin-gap RPC in the ATLAS Phase-II Upgrade",
                        "abstract": "Thin-gap Resistive Plate Chambers (RPCs) with a 1 mm gap size are introduced in the Phase-II ATLAS upgrade. Smaller avalanche charge due to the reduced gap size raises concerns for signal integrity. This work focuses on the RPC signal propagation process in lossless conditions, and an analytical study is implemented for the ATLAS RPC. Detector modeling is presented, and the simulation of the RPC signal is discussed in detail. Simulated characteristic impedance and crosstalk have been compared with the measured value to validate this model. This method is applied to different RPC design geometries, including the newly proposed $\\eta-\\eta$ readout scheme."
                    },
                    {
                        "title": "Robust Network Learning via Inverse Scale Variational Sparsification",
                        "abstract": "While neural networks have made significant strides in many AI tasks, they remain vulnerable to a range of noise types, including natural corruptions, adversarial noise, and low-resolution artifacts. Many existing approaches focus on enhancing robustness against specific noise types, limiting their adaptability to others. Previous studies have addressed general robustness by adopting a spectral perspective, which tends to blur crucial features like texture and object contours. Our proposed solution, however, introduces an inverse scale variational sparsification framework within a time-continuous inverse scale space formulation. This framework progressively learns finer-scale features by discerning variational differences between pixels, ultimately preserving only large-scale features in the smoothed image. Unlike frequency-based methods, our approach not only removes noise by smoothing small-scale features where corruptions often occur but also retains high-contrast details such as textures and object contours. Moreover, our framework offers simplicity and efficiency in implementation. By integrating this algorithm into neural network training, we guide the model to prioritize learning large-scale features. We show the efficacy of our approach through enhanced robustness against various noise types."
                    },
                    {
                        "title": "MolCPT: Molecule Continuous Prompt Tuning to Generalize Molecular Representation Learning",
                        "abstract": "Molecular representation learning is crucial for the problem of molecular property prediction, where graph neural networks (GNNs) serve as an effective solution due to their structure modeling capabilities. Since labeled data is often scarce and expensive to obtain, it is a great challenge for GNNs to generalize in the extensive molecular space. Recently, the training paradigm of \"pre-train, fine-tune\" has been leveraged to improve the generalization capabilities of GNNs. It uses self-supervised information to pre-train the GNN, and then performs fine-tuning to optimize the downstream task with just a few labels. However, pre-training does not always yield statistically significant improvement, especially for self-supervised learning with random structural masking. In fact, the molecular structure is characterized by motif subgraphs, which are frequently occurring and influence molecular properties. To leverage the task-related motifs, we propose a novel paradigm of \"pre-train, prompt, fine-tune\" for molecular representation learning, named molecule continuous prompt tuning (MolCPT). MolCPT defines a motif prompting function that uses the pre-trained model to project the standalone input into an expressive prompt. The prompt effectively augments the molecular graph with meaningful motifs in the continuous representation space; this provides more structural patterns to aid the downstream classifier in identifying molecular properties. Extensive experiments on several benchmark datasets show that MolCPT efficiently generalizes pre-trained GNNs for molecular property prediction, with or without a few fine-tuning steps."
                    },
                    {
                        "title": "Towards Interaction Detection Using Topological Analysis on Neural Networks",
                        "abstract": "Detecting statistical interactions between input features is a crucial and challenging task. Recent advances demonstrate that it is possible to extract learned interactions from trained neural networks. It has also been observed that, in neural networks, any interacting features must follow a strongly weighted connection to common hidden units. Motivated by the observation, in this paper, we propose to investigate the interaction detection problem from a novel topological perspective by analyzing the connectivity in neural networks. Specially, we propose a new measure for quantifying interaction strength, based upon the well-received theory of persistent homology. Based on this measure, a Persistence Interaction detection~(PID) algorithm is developed to efficiently detect interactions. Our proposed algorithm is evaluated across a number of interaction detection tasks on several synthetic and real world datasets with different hyperparameters. Experimental results validate that the PID algorithm outperforms the state-of-the-art baselines."
                    }
                ]
            },
            "b976bdcc-5437-4eb8-87d1-0850dcd3a320": {
                "pk": "b976bdcc-5437-4eb8-87d1-0850dcd3a320",
                "name": "Xiaotian Han",
                "collaborators": [
                    "Xia Hu",
                    "Zhimeng Jiang",
                    "Na Zou",
                    "Ninghao Liu",
                    "Yicheng Wang",
                    "Qifan Wang",
                    "Qingquan Song",
                    "Jundong Li",
                    "Hongye Jin",
                    "Chia-Yuan Chang"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Fairness",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "G-Mixup: Graph Data Augmentation for Graph Classification",
                        "abstract": "This work develops \\emph{mixup for graph data}. Mixup has shown superiority in improving the generalization and robustness of neural networks by interpolating features and labels between two random samples. Traditionally, Mixup can work on regular, grid-like, and Euclidean data such as image or tabular data. However, it is challenging to directly adopt Mixup to augment graph data because different graphs typically: 1) have different numbers of nodes; 2) are not readily aligned; and 3) have unique typologies in non-Euclidean space. To this end, we propose $\\mathcal{G}$-Mixup to augment graphs for graph classification by interpolating the generator (i.e., graphon) of different classes of graphs. Specifically, we first use graphs within the same class to estimate a graphon. Then, instead of directly manipulating graphs, we interpolate graphons of different classes in the Euclidean space to get mixed graphons, where the synthetic graphs are generated through sampling based on the mixed graphons. Extensive experiments show that $\\mathcal{G}$-Mixup substantially improves the generalization and robustness of GNNs."
                    },
                    {
                        "title": "Does Synthetic Data Generation of LLMs Help Clinical Text Mining?",
                        "abstract": "Recent advancements in large language models (LLMs) have led to the development of highly potent models like OpenAI's ChatGPT. These models have exhibited exceptional performance in a variety of tasks, such as question answering, essay composition, and code generation. However, their effectiveness in the healthcare sector remains uncertain. In this study, we seek to investigate the potential of ChatGPT to aid in clinical text mining by examining its ability to extract structured information from unstructured healthcare texts, with a focus on biological named entity recognition and relation extraction. However, our preliminary results indicate that employing ChatGPT directly for these tasks resulted in poor performance and raised privacy concerns associated with uploading patients' information to the ChatGPT API. To overcome these limitations, we propose a new training paradigm that involves generating a vast quantity of high-quality synthetic data with labels utilizing ChatGPT and fine-tuning a local model for the downstream task. Our method has resulted in significant improvements in the performance of downstream tasks, improving the F1-score from 23.37% to 63.99% for the named entity recognition task and from 75.86% to 83.59% for the relation extraction task. Furthermore, generating data using ChatGPT can significantly reduce the time and effort required for data collection and labeling, as well as mitigate data privacy concerns. In summary, the proposed framework presents a promising solution to enhance the applicability of LLM models to clinical text mining."
                    },
                    {
                        "title": "Beyond Fairness: Age-Harmless Parkinson's Detection via Voice",
                        "abstract": "Parkinson's disease (PD), a neurodegenerative disorder, often manifests as speech and voice dysfunction. While utilizing voice data for PD detection has great potential in clinical applications, the widely used deep learning models currently have fairness issues regarding different ages of onset. These deep models perform well for the elderly group (age $>$ 55) but are less accurate for the young group (age $\\leq$ 55). Through our investigation, the discrepancy between the elderly and the young arises due to 1) an imbalanced dataset and 2) the milder symptoms often seen in early-onset patients. However, traditional debiasing methods are impractical as they typically impair the prediction accuracy for the majority group while minimizing the discrepancy. To address this issue, we present a new debiasing method using GradCAM-based feature masking combined with ensemble models, ensuring that neither fairness nor accuracy is compromised. Specifically, the GradCAM-based feature masking selectively obscures age-related features in the input voice data while preserving essential information for PD detection. The ensemble models further improve the prediction accuracy for the minority (young group). Our approach effectively improves detection accuracy for early-onset patients without sacrificing performance for the elderly group. Additionally, we propose a two-step detection strategy for the young group, offering a practical risk assessment for potential early-onset PD patients."
                    },
                    {
                        "title": "MLPInit: Embarrassingly Simple GNN Training Acceleration with MLP Initialization",
                        "abstract": "Training graph neural networks (GNNs) on large graphs is complex and extremely time consuming. This is attributed to overheads caused by sparse matrix multiplication, which are sidestepped when training multi-layer perceptrons (MLPs) with only node features. MLPs, by ignoring graph context, are simple and faster for graph data, however they usually sacrifice prediction accuracy, limiting their applications for graph data. We observe that for most message passing-based GNNs, we can trivially derive an analog MLP (we call this a PeerMLP) with an equivalent weight space, by setting the trainable parameters with the same shapes, making us curious about \\textbf{\\emph{how do GNNs using weights from a fully trained PeerMLP perform?}} Surprisingly, we find that GNNs initialized with such weights significantly outperform their PeerMLPs, motivating us to use PeerMLP training as a precursor, initialization step to GNN training. To this end, we propose an embarrassingly simple, yet hugely effective initialization method for GNN training acceleration, called MLPInit. Our extensive experiments on multiple large-scale graph datasets with diverse GNN architectures validate that MLPInit can accelerate the training of GNNs (up to 33X speedup on OGB-Products) and often improve prediction performance (e.g., up to $7.97\\%$ improvement for GraphSAGE across $7$ datasets for node classification, and up to $17.81\\%$ improvement across $4$ datasets for link prediction on metric Hits@10). The code is available at \\href{https://github.com/snap-research/MLPInit-for-GNNs}."
                    },
                    {
                        "title": "COCO is \"ALL'' You Need for Visual Instruction Fine-tuning",
                        "abstract": "Multi-modal Large Language Models (MLLMs) are increasingly prominent in the field of artificial intelligence. Visual instruction fine-tuning (IFT) is a vital process for aligning MLLMs' output with user's intentions. High-quality and diversified instruction following data is the key to this fine-tuning process. Recent studies propose to construct visual IFT datasets through a multifaceted approach: transforming existing datasets with rule-based templates, employing GPT-4 for rewriting annotations, and utilizing GPT-4V for visual dataset pseudo-labeling. LLaVA-1.5 adopted similar approach and construct LLaVA-mix-665k, which is one of the simplest, most widely used, yet most effective IFT datasets today. Notably, when properly fine-tuned with this dataset, MLLMs can achieve state-of-the-art performance on several benchmarks. However, we noticed that models trained with this dataset often struggle to follow user instructions properly in multi-round dialog. In addition, tradition caption and VQA evaluation benchmarks, with their closed-form evaluation structure, are not fully equipped to assess the capabilities of modern open-ended generative MLLMs. This problem is not unique to the LLaVA-mix-665k dataset, but may be a potential issue in all IFT datasets constructed from image captioning or VQA sources, though the extent of this issue may vary. We argue that datasets with diverse and high-quality detailed instruction following annotations are essential and adequate for MLLMs IFT. In this work, we establish a new IFT dataset, with images sourced from the COCO dataset along with more diverse instructions. Our experiments show that when fine-tuned with out proposed dataset, MLLMs achieve better performance on open-ended evaluation benchmarks in both single-round and multi-round dialog setting."
                    },
                    {
                        "title": "FFB: A Fair Fairness Benchmark for In-Processing Group Fairness Methods",
                        "abstract": "This paper introduces the Fair Fairness Benchmark (\\textsf{FFB}), a benchmarking framework for in-processing group fairness methods. Ensuring fairness in machine learning is important for ethical compliance. However, there exist challenges in comparing and developing fairness methods due to inconsistencies in experimental settings, lack of accessible algorithmic implementations, and limited extensibility of current fairness packages and tools. To address these issues, we introduce an open-source standardized benchmark for evaluating in-processing group fairness methods and provide a comprehensive analysis of state-of-the-art methods to ensure different notions of group fairness. This work offers the following key contributions: the provision of flexible, extensible, minimalistic, and research-oriented open-source code; the establishment of unified fairness method benchmarking pipelines; and extensive benchmarking, which yields key insights from $\\mathbf{45,079}$ experiments, $\\mathbf{14,428}$ GPU hours. We believe that our work will significantly facilitate the growth and development of the fairness research community."
                    },
                    {
                        "title": "Image Scene Graph Generation (SGG) Benchmark",
                        "abstract": "There is a surge of interest in image scene graph generation (object, attribute and relationship detection) due to the need of building fine-grained image understanding models that go beyond object detection. Due to the lack of a good benchmark, the reported results of different scene graph generation models are not directly comparable, impeding the research progress. We have developed a much-needed scene graph generation benchmark based on the maskrcnn-benchmark and several popular models. This paper presents main features of our benchmark and a comprehensive ablation study of scene graph generation models using the Visual Genome and OpenImages Visual relationship detection datasets. Our codebase is made publicly available at https://github.com/microsoft/scene_graph_benchmark."
                    },
                    {
                        "title": "Geometric Graph Representation Learning via Maximizing Rate Reduction",
                        "abstract": "Learning discriminative node representations benefits various downstream tasks in graph analysis such as community detection and node classification. Existing graph representation learning methods (e.g., based on random walk and contrastive learning) are limited to maximizing the local similarity of connected nodes. Such pair-wise learning schemes could fail to capture the global distribution of representations, since it has no explicit constraints on the global geometric properties of representation space. To this end, we propose Geometric Graph Representation Learning (G2R) to learn node representations in an unsupervised manner via maximizing rate reduction. In this way, G2R maps nodes in distinct groups (implicitly stored in the adjacency matrix) into different subspaces, while each subspace is compact and different subspaces are dispersedly distributed. G2R adopts a graph neural network as the encoder and maximizes the rate reduction with the adjacency matrix. Furthermore, we theoretically and empirically demonstrate that rate reduction maximization is equivalent to maximizing the principal angles between different subspaces. Experiments on real-world datasets show that G2R outperforms various baselines on node classification and community detection tasks."
                    },
                    {
                        "title": "GrowLength: Accelerating LLMs Pretraining by Progressively Growing Training Length",
                        "abstract": "The evolving sophistication and intricacies of Large Language Models (LLMs) yield unprecedented advancements, yet they simultaneously demand considerable computational resources and incur significant costs. To alleviate these challenges, this paper introduces a novel, simple, and effective method named ``\\growlength'' to accelerate the pretraining process of LLMs. Our method progressively increases the training length throughout the pretraining phase, thereby mitigating computational costs and enhancing efficiency. For instance, it begins with a sequence length of 128 and progressively extends to 4096. This approach enables models to process a larger number of tokens within limited time frames, potentially boosting their performance. In other words, the efficiency gain is derived from training with shorter sequences optimizing the utilization of resources. Our extensive experiments with various state-of-the-art LLMs have revealed that models trained using our method not only converge more swiftly but also exhibit superior performance metrics compared to those trained with existing methods. Furthermore, our method for LLMs pretraining acceleration does not require any additional engineering efforts, making it a practical solution in the realm of LLMs."
                    },
                    {
                        "title": "Marginal Nodes Matter: Towards Structure Fairness in Graphs",
                        "abstract": "In social network, a person located at the periphery region (marginal node) is likely to be treated unfairly when compared with the persons at the center. While existing fairness works on graphs mainly focus on protecting sensitive attributes (e.g., age and gender), the fairness incurred by the graph structure should also be given attention. On the other hand, the information aggregation mechanism of graph neural networks amplifies such structure unfairness, as marginal nodes are often far away from other nodes. In this paper, we focus on novel fairness incurred by the graph structure on graph neural networks, named \\emph{structure fairness}. Specifically, we first analyzed multiple graphs and observed that marginal nodes in graphs have a worse performance of downstream tasks than others in graph neural networks. Motivated by the observation, we propose \\textbf{S}tructural \\textbf{Fair} \\textbf{G}raph \\textbf{N}eural \\textbf{N}etwork (SFairGNN), which combines neighborhood expansion based structure debiasing with hop-aware attentive information aggregation to achieve structure fairness. Our experiments show \\SFairGNN can significantly improve structure fairness while maintaining overall performance in the downstream tasks."
                    },
                    {
                        "title": "Retiring $\u0394$DP: New Distribution-Level Metrics for Demographic Parity",
                        "abstract": "Demographic parity is the most widely recognized measure of group fairness in machine learning, which ensures equal treatment of different demographic groups. Numerous works aim to achieve demographic parity by pursuing the commonly used metric $\\Delta DP$. Unfortunately, in this paper, we reveal that the fairness metric $\\Delta DP$ can not precisely measure the violation of demographic parity, because it inherently has the following drawbacks: i) zero-value $\\Delta DP$ does not guarantee zero violation of demographic parity, ii) $\\Delta DP$ values can vary with different classification thresholds. To this end, we propose two new fairness metrics, Area Between Probability density function Curves (ABPC) and Area Between Cumulative density function Curves (ABCC), to precisely measure the violation of demographic parity at the distribution level. The new fairness metrics directly measure the difference between the distributions of the prediction probability for different demographic groups. Thus our proposed new metrics enjoy: i) zero-value ABCC/ABPC guarantees zero violation of demographic parity; ii) ABCC/ABPC guarantees demographic parity while the classification thresholds are adjusted. We further re-evaluate the existing fair models with our proposed fairness metrics and observe different fairness behaviors of those models under the new metrics. The code is available at https://github.com/ahxt/new_metric_for_demographic_parity"
                    },
                    {
                        "title": "AutoRec: An Automated Recommender System",
                        "abstract": "Realistic recommender systems are often required to adapt to ever-changing data and tasks or to explore different models systematically. To address the need, we present AutoRec, an open-source automated machine learning (AutoML) platform extended from the TensorFlow ecosystem and, to our knowledge, the first framework to leverage AutoML for model search and hyperparameter tuning in deep recommendation models. AutoRec also supports a highly flexible pipeline that accommodates both sparse and dense inputs, rating prediction and click-through rate (CTR) prediction tasks, and an array of recommendation models. Lastly, AutoRec provides a simple, user-friendly API. Experiments conducted on the benchmark datasets reveal AutoRec is reliable and can identify models which resemble the best model without prior knowledge."
                    },
                    {
                        "title": "Auto-PINN: Understanding and Optimizing Physics-Informed Neural Architecture",
                        "abstract": "Physics-informed neural networks (PINNs) are revolutionizing science and engineering practice by bringing together the power of deep learning to bear on scientific computation. In forward modeling problems, PINNs are meshless partial differential equation (PDE) solvers that can handle irregular, high-dimensional physical domains. Naturally, the neural architecture hyperparameters have a large impact on the efficiency and accuracy of the PINN solver. However, this remains an open and challenging problem because of the large search space and the difficulty of identifying a proper search objective for PDEs. Here, we propose Auto-PINN, the first systematic, automated hyperparameter optimization approach for PINNs, which employs Neural Architecture Search (NAS) techniques to PINN design. Auto-PINN avoids manually or exhaustively searching the hyperparameter space associated with PINNs. A comprehensive set of pre-experiments using standard PDE benchmarks allows us to probe the structure-performance relationship in PINNs. We find that the different hyperparameters can be decoupled, and that the training loss function of PINNs is a good search objective. Comparison experiments with baseline methods demonstrate that Auto-PINN produces neural architectures with superior stability and accuracy over alternative baselines."
                    }
                ]
            },
            "2ec35e12-6f8f-4c7e-ab08-c53bfd96af99": {
                "pk": "2ec35e12-6f8f-4c7e-ab08-c53bfd96af99",
                "name": "Qizhang Feng",
                "collaborators": [
                    "Xia Hu",
                    "Fan Yang",
                    "Ruixiang Tang",
                    "Ninghao Liu",
                    "Mengnan Du",
                    "Na Zou",
                    "Kaixiong Zhou",
                    "Jiahao Chen",
                    "Jiayi Yuan",
                    "Forhan Bin Emdad"
                ],
                "domain": [
                    "Machine Learning",
                    "Fairness",
                    "Graph Neural Network",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Fair Machine Learning in Healthcare: A Review",
                        "abstract": "The digitization of healthcare data coupled with advances in computational capabilities has propelled the adoption of machine learning (ML) in healthcare. However, these methods can perpetuate or even exacerbate existing disparities, leading to fairness concerns such as the unequal distribution of resources and diagnostic inaccuracies among different demographic groups. Addressing these fairness problem is paramount to prevent further entrenchment of social injustices. In this survey, we analyze the intersection of fairness in machine learning and healthcare disparities. We adopt a framework based on the principles of distributive justice to categorize fairness concerns into two distinct classes: equal allocation and equal performance. We provide a critical review of the associated fairness metrics from a machine learning standpoint and examine biases and mitigation strategies across the stages of the ML lifecycle, discussing the relationship between biases and their countermeasures. The paper concludes with a discussion on the pressing challenges that remain unaddressed in ensuring fairness in healthcare ML, and proposes several new research directions that hold promise for developing ethical and equitable ML applications in healthcare."
                    },
                    {
                        "title": "Differentially Private Counterfactuals via Functional Mechanism",
                        "abstract": "Counterfactual, serving as one emerging type of model explanation, has attracted tons of attentions recently from both industry and academia. Different from the conventional feature-based explanations (e.g., attributions), counterfactuals are a series of hypothetical samples which can flip model decisions with minimal perturbations on queries. Given valid counterfactuals, humans are capable of reasoning under ``what-if'' circumstances, so as to better understand the model decision boundaries. However, releasing counterfactuals could be detrimental, since it may unintentionally leak sensitive information to adversaries, which brings about higher risks on both model security and data privacy. To bridge the gap, in this paper, we propose a novel framework to generate differentially private counterfactual (DPC) without touching the deployed model or explanation set, where noises are injected for protection while maintaining the explanation roles of counterfactual. In particular, we train an autoencoder with the functional mechanism to construct noisy class prototypes, and then derive the DPC from the latent prototypes based on the post-processing immunity of differential privacy. Further evaluations demonstrate the effectiveness of the proposed framework, showing that DPC can successfully relieve the risks on both extraction and inference attacks."
                    },
                    {
                        "title": "Did You Train on My Dataset? Towards Public Dataset Protection with Clean-Label Backdoor Watermarking",
                        "abstract": "The huge supporting training data on the Internet has been a key factor in the success of deep learning models. However, this abundance of public-available data also raises concerns about the unauthorized exploitation of datasets for commercial purposes, which is forbidden by dataset licenses. In this paper, we propose a backdoor-based watermarking approach that serves as a general framework for safeguarding public-available data. By inserting a small number of watermarking samples into the dataset, our approach enables the learning model to implicitly learn a secret function set by defenders. This hidden function can then be used as a watermark to track down third-party models that use the dataset illegally. Unfortunately, existing backdoor insertion methods often entail adding arbitrary and mislabeled data to the training set, leading to a significant drop in performance and easy detection by anomaly detection algorithms. To overcome this challenge, we introduce a clean-label backdoor watermarking framework that uses imperceptible perturbations to replace mislabeled samples. As a result, the watermarking samples remain consistent with the original labels, making them difficult to detect. Our experiments on text, image, and audio datasets demonstrate that the proposed framework effectively safeguards datasets with minimal impact on original task performance. We also show that adding just 1% of watermarking samples can inject a traceable watermarking function and that our watermarking samples are stealthy and look benign upon visual inspection."
                    },
                    {
                        "title": "DEGREE: Decomposition Based Explanation For Graph Neural Networks",
                        "abstract": "Graph Neural Networks (GNNs) are gaining extensive attention for their application in graph data. However, the black-box nature of GNNs prevents users from understanding and trusting the models, thus hampering their applicability. Whereas explaining GNNs remains a challenge, most existing methods fall into approximation based and perturbation based approaches with suffer from faithfulness problems and unnatural artifacts, respectively. To tackle these problems, we propose DEGREE \\degree to provide a faithful explanation for GNN predictions. By decomposing the information generation and aggregation mechanism of GNNs, DEGREE allows tracking the contributions of specific components of the input graph to the final prediction. Based on this, we further design a subgraph level interpretation algorithm to reveal complex interactions between graph nodes that are overlooked by previous methods. The efficiency of our algorithm can be further improved by utilizing GNN characteristics. Finally, we conduct quantitative and qualitative experiments on synthetic and real-world datasets to demonstrate the effectiveness of DEGREE on node classification and graph classification tasks."
                    },
                    {
                        "title": "Can Attention Be Used to Explain EHR-Based Mortality Prediction Tasks: A Case Study on Hemorrhagic Stroke",
                        "abstract": "Stroke is a significant cause of mortality and morbidity, necessitating early predictive strategies to minimize risks. Traditional methods for evaluating patients, such as Acute Physiology and Chronic Health Evaluation (APACHE II, IV) and Simplified Acute Physiology Score III (SAPS III), have limited accuracy and interpretability. This paper proposes a novel approach: an interpretable, attention-based transformer model for early stroke mortality prediction. This model seeks to address the limitations of previous predictive models, providing both interpretability (providing clear, understandable explanations of the model) and fidelity (giving a truthful explanation of the model's dynamics from input to output). Furthermore, the study explores and compares fidelity and interpretability scores using Shapley values and attention-based scores to improve model explainability. The research objectives include designing an interpretable attention-based transformer model, evaluating its performance compared to existing models, and providing feature importance derived from the model."
                    },
                    {
                        "title": "Efficient GNN Explanation via Learning Removal-based Attribution",
                        "abstract": "As Graph Neural Networks (GNNs) have been widely used in real-world applications, model explanations are required not only by users but also by legal regulations. However, simultaneously achieving high fidelity and low computational costs in generating explanations has been a challenge for current methods. In this work, we propose a framework of GNN explanation named LeArn Removal-based Attribution (LARA) to address this problem. Specifically, we introduce removal-based attribution and demonstrate its substantiated link to interpretability fidelity theoretically and experimentally. The explainer in LARA learns to generate removal-based attribution which enables providing explanations with high fidelity. A strategy of subgraph sampling is designed in LARA to improve the scalability of the training process. In the deployment, LARA can efficiently generate the explanation through a feed-forward pass. We benchmark our approach with other state-of-the-art GNN explanation methods on six datasets. Results highlight the effectiveness of our framework regarding both efficiency and fidelity. In particular, LARA is 3.5 times faster and achieves higher fidelity than the state-of-the-art method on the large dataset ogbn-arxiv (more than 160K nodes and 1M edges), showing its great potential in real-world applications. Our source code is available at https://anonymous.4open.science/r/LARA-10D8/README.md."
                    },
                    {
                        "title": "Exposing Privacy Gaps: Membership Inference Attack on Preference Data for LLM Alignment",
                        "abstract": "Large Language Models (LLMs) have seen widespread adoption due to their remarkable natural language capabilities. However, when deploying them in real-world settings, it is important to align LLMs to generate texts according to acceptable human standards. Methods such as Proximal Policy Optimization (PPO) and Direct Preference Optimization (DPO) have made significant progress in refining LLMs using human preference data. However, the privacy concerns inherent in utilizing such preference data have yet to be adequately studied. In this paper, we investigate the vulnerability of LLMs aligned using human preference datasets to membership inference attacks (MIAs), highlighting the shortcomings of previous MIA approaches with respect to preference data. Our study has two main contributions: first, we introduce a novel reference-based attack framework specifically for analyzing preference data called PREMIA (\\uline{Pre}ference data \\uline{MIA}); second, we provide empirical evidence that DPO models are more vulnerable to MIA compared to PPO models. Our findings highlight gaps in current privacy-preserving practices for LLM alignment."
                    },
                    {
                        "title": "Harnessing the Power of LLMs in Practice: A Survey on ChatGPT and Beyond",
                        "abstract": "This paper presents a comprehensive and practical guide for practitioners and end-users working with Large Language Models (LLMs) in their downstream natural language processing (NLP) tasks. We provide discussions and insights into the usage of LLMs from the perspectives of models, data, and downstream tasks. Firstly, we offer an introduction and brief summary of current GPT- and BERT-style LLMs. Then, we discuss the influence of pre-training data, training data, and test data. Most importantly, we provide a detailed discussion about the use and non-use cases of large language models for various natural language processing tasks, such as knowledge-intensive tasks, traditional natural language understanding tasks, natural language generation tasks, emergent abilities, and considerations for specific tasks.We present various use cases and non-use cases to illustrate the practical applications and limitations of LLMs in real-world scenarios. We also try to understand the importance of data and the specific challenges associated with each NLP task. Furthermore, we explore the impact of spurious biases on LLMs and delve into other essential considerations, such as efficiency, cost, and latency, to ensure a comprehensive understanding of deploying LLMs in practice. This comprehensive guide aims to provide researchers and practitioners with valuable insights and best practices for working with LLMs, thereby enabling the successful implementation of these models in a wide range of NLP tasks. A curated list of practical guide resources of LLMs, regularly updated, can be found at \\url{https://github.com/Mooler0410/LLMsPracticalGuide}."
                    }
                ]
            },
            "982c0f0c-e005-4d40-80a8-cbf7b6f6af78": {
                "pk": "982c0f0c-e005-4d40-80a8-cbf7b6f6af78",
                "name": "Hongye Jin",
                "collaborators": [
                    "Xia Hu",
                    "Zhimeng Jiang",
                    "Xiaotian Han",
                    "Zirui Liu",
                    "Guanchu Wang",
                    "Jingfeng Yang",
                    "Jiayi Yuan",
                    "Shaochen Zhong",
                    "Na Zou",
                    "Chia-Yuan Chang"
                ],
                "domain": [
                    "Fairness in Machine Learning",
                    "Large Language Models",
                    "Collaborative Filtering",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "Chasing Fairness Under Distribution Shift: A Model Weight Perturbation Approach",
                        "abstract": "Fairness in machine learning has attracted increasing attention in recent years. The fairness methods improving algorithmic fairness for in-distribution data may not perform well under distribution shifts. In this paper, we first theoretically demonstrate the inherent connection between distribution shift, data perturbation, and model weight perturbation. Subsequently, we analyze the sufficient conditions to guarantee fairness (i.e., low demographic parity) for the target dataset, including fairness for the source dataset, and low prediction difference between the source and target datasets for each sensitive attribute group. Motivated by these sufficient conditions, we propose robust fairness regularization (RFR) by considering the worst case within the model weight perturbation ball for each sensitive attribute group. We evaluate the effectiveness of our proposed RFR algorithm on synthetic and real distribution shifts across various datasets. Experimental results demonstrate that RFR achieves better fairness-accuracy trade-off performance compared with several baselines. The source code is available at \\url{https://github.com/zhimengj0326/RFR_NeurIPS23}."
                    },
                    {
                        "title": "GrowLength: Accelerating LLMs Pretraining by Progressively Growing Training Length",
                        "abstract": "The evolving sophistication and intricacies of Large Language Models (LLMs) yield unprecedented advancements, yet they simultaneously demand considerable computational resources and incur significant costs. To alleviate these challenges, this paper introduces a novel, simple, and effective method named ``\\growlength'' to accelerate the pretraining process of LLMs. Our method progressively increases the training length throughout the pretraining phase, thereby mitigating computational costs and enhancing efficiency. For instance, it begins with a sequence length of 128 and progressively extends to 4096. This approach enables models to process a larger number of tokens within limited time frames, potentially boosting their performance. In other words, the efficiency gain is derived from training with shorter sequences optimizing the utilization of resources. Our extensive experiments with various state-of-the-art LLMs have revealed that models trained using our method not only converge more swiftly but also exhibit superior performance metrics compared to those trained with existing methods. Furthermore, our method for LLMs pretraining acceleration does not require any additional engineering efforts, making it a practical solution in the realm of LLMs."
                    },
                    {
                        "title": "Towards Mitigating Dimensional Collapse of Representations in Collaborative Filtering",
                        "abstract": "Contrastive Learning (CL) has shown promising performance in collaborative filtering. The key idea is to generate augmentation-invariant embeddings by maximizing the Mutual Information between different augmented views of the same instance. However, we empirically observe that existing CL models suffer from the \\textsl{dimensional collapse} issue, where user/item embeddings only span a low-dimension subspace of the entire feature space. This suppresses other dimensional information and weakens the distinguishability of embeddings. Here we propose a non-contrastive learning objective, named nCL, which explicitly mitigates dimensional collapse of representations in collaborative filtering. Our nCL aims to achieve geometric properties of \\textsl{Alignment} and \\textsl{Compactness} on the embedding space. In particular, the alignment tries to push together representations of positive-related user-item pairs, while compactness tends to find the optimal coding length of user/item embeddings, subject to a given distortion. More importantly, our nCL does not require data augmentation nor negative sampling during training, making it scalable to large datasets. Experimental results demonstrate the superiority of our nCL."
                    },
                    {
                        "title": "Disentangled Graph Collaborative Filtering",
                        "abstract": "Learning informative representations of users and items from the interaction data is of crucial importance to collaborative filtering (CF). Present embedding functions exploit user-item relationships to enrich the representations, evolving from a single user-item instance to the holistic interaction graph. Nevertheless, they largely model the relationships in a uniform manner, while neglecting the diversity of user intents on adopting the items, which could be to pass time, for interest, or shopping for others like families. Such uniform approach to model user interests easily results in suboptimal representations, failing to model diverse relationships and disentangle user intents in representations.   In this work, we pay special attention to user-item relationships at the finer granularity of user intents. We hence devise a new model, Disentangled Graph Collaborative Filtering (DGCF), to disentangle these factors and yield disentangled representations. Specifically, by modeling a distribution over intents for each user-item interaction, we iteratively refine the intent-aware interaction graphs and representations. Meanwhile, we encourage independence of different intents. This leads to disentangled representations, effectively distilling information pertinent to each intent. We conduct extensive experiments on three benchmark datasets, and DGCF achieves significant improvements over several state-of-the-art models like NGCF, DisenGCN, and MacridVAE. Further analyses offer insights into the advantages of DGCF on the disentanglement of user intents and interpretability of representations. Our codes are available in https://github.com/xiangwang1223/disentangled_graph_collaborative_filtering."
                    },
                    {
                        "title": "Retiring $\u0394$DP: New Distribution-Level Metrics for Demographic Parity",
                        "abstract": "Demographic parity is the most widely recognized measure of group fairness in machine learning, which ensures equal treatment of different demographic groups. Numerous works aim to achieve demographic parity by pursuing the commonly used metric $\\Delta DP$. Unfortunately, in this paper, we reveal that the fairness metric $\\Delta DP$ can not precisely measure the violation of demographic parity, because it inherently has the following drawbacks: i) zero-value $\\Delta DP$ does not guarantee zero violation of demographic parity, ii) $\\Delta DP$ values can vary with different classification thresholds. To this end, we propose two new fairness metrics, Area Between Probability density function Curves (ABPC) and Area Between Cumulative density function Curves (ABCC), to precisely measure the violation of demographic parity at the distribution level. The new fairness metrics directly measure the difference between the distributions of the prediction probability for different demographic groups. Thus our proposed new metrics enjoy: i) zero-value ABCC/ABPC guarantees zero violation of demographic parity; ii) ABCC/ABPC guarantees demographic parity while the classification thresholds are adjusted. We further re-evaluate the existing fair models with our proposed fairness metrics and observe different fairness behaviors of those models under the new metrics. The code is available at https://github.com/ahxt/new_metric_for_demographic_parity"
                    },
                    {
                        "title": "Harnessing the Power of LLMs in Practice: A Survey on ChatGPT and Beyond",
                        "abstract": "This paper presents a comprehensive and practical guide for practitioners and end-users working with Large Language Models (LLMs) in their downstream natural language processing (NLP) tasks. We provide discussions and insights into the usage of LLMs from the perspectives of models, data, and downstream tasks. Firstly, we offer an introduction and brief summary of current GPT- and BERT-style LLMs. Then, we discuss the influence of pre-training data, training data, and test data. Most importantly, we provide a detailed discussion about the use and non-use cases of large language models for various natural language processing tasks, such as knowledge-intensive tasks, traditional natural language understanding tasks, natural language generation tasks, emergent abilities, and considerations for specific tasks.We present various use cases and non-use cases to illustrate the practical applications and limitations of LLMs in real-world scenarios. We also try to understand the importance of data and the specific challenges associated with each NLP task. Furthermore, we explore the impact of spurious biases on LLMs and delve into other essential considerations, such as efficiency, cost, and latency, to ensure a comprehensive understanding of deploying LLMs in practice. This comprehensive guide aims to provide researchers and practitioners with valuable insights and best practices for working with LLMs, thereby enabling the successful implementation of these models in a wide range of NLP tasks. A curated list of practical guide resources of LLMs, regularly updated, can be found at \\url{https://github.com/Mooler0410/LLMsPracticalGuide}."
                    },
                    {
                        "title": "LLM Maybe LongLM: Self-Extend LLM Context Window Without Tuning",
                        "abstract": "It is well known that LLMs cannot generalize well to long contexts whose lengths are larger than the training sequence length. This poses challenges when employing LLMs for processing long input sequences during inference. In this work, we argue that LLMs themselves have inherent capabilities to handle long contexts without fine-tuning. To achieve this goal, we propose SelfExtend to extend the context window of LLMs by constructing bi-level attention information: the grouped attention and the neighbor attention. The grouped attention captures the dependencies among tokens that are far apart, while neighbor attention captures dependencies among adjacent tokens within a specified range. The two-level attentions are computed based on the original model's self-attention mechanism during inference. With minor code modification, our SelfExtend can effortlessly extend existing LLMs' context window without any fine-tuning. We conduct comprehensive experiments on multiple benchmarks and the results show that our SelfExtend can effectively extend existing LLMs' context window length. The code can be found at \\url{https://github.com/datamllab/LongLM}."
                    },
                    {
                        "title": "KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache",
                        "abstract": "Efficiently serving large language models (LLMs) requires batching of many requests to reduce the cost per request. Yet, with larger batch sizes and longer context lengths, the key-value (KV) cache, which stores attention keys and values to avoid re-computations, significantly increases memory demands and becomes the new bottleneck in speed and memory usage. Additionally, the loading of the KV cache causes the computational core to be idle, which limits the inference speed. A straightforward and effective solution to reduce KV cache size is quantization, which decreases the total bytes taken by KV cache. However, there is a lack of in-depth studies that explore the element distribution of KV cache to understand the hardness and limitation of KV cache quantization. To fill the gap, we conducted a comprehensive study on the element distribution in KV cache of popular LLMs. Our findings indicate that the key cache should be quantized per-channel, i.e., group elements along the channel dimension and quantize them together. In contrast, the value cache should be quantized per-token. From this analysis, we developed a tuning-free 2bit KV cache quantization algorithm named KIVI. With hardware-friendly implementation, KIVI can enable Llama, Falcon, and Mistral models to maintain almost the same quality while using $\\mathbf{2.6\\times}$ less peak memory (including model weight). This reduction in memory usage enables up to $\\mathbf{4\\times}$ larger batch size, bringing $\\mathbf{2.35\\times \\sim 3.47\\times}$ throughput on real LLM inference workload. The source code is available at https://github.com/jy-yuan/KIVI."
                    },
                    {
                        "title": "KV Cache Compression, But What Must We Give in Return? A Comprehensive Benchmark of Long Context Capable Approaches",
                        "abstract": "Long context capability is a crucial competency for large language models (LLMs) as it mitigates the human struggle to digest long-form texts. This capability enables complex task-solving scenarios such as book summarization, code assistance, and many more tasks that are traditionally manpower-intensive. However, transformer-based LLMs face significant challenges with long context input due to the growing size of the KV cache and the intrinsic complexity of attending to extended inputs; where multiple schools of efficiency-driven approaches - such as KV cache quantization, token dropping, prompt compression, linear-time sequence models, and hybrid architectures - have been proposed to produce efficient yet long context-capable models. Despite these advancements, no existing work has comprehensively benchmarked these methods in a reasonably aligned environment. In this work, we fill this gap by providing a taxonomy of current methods and evaluating 10+ state-of-the-art approaches across seven categories of long context tasks. Our work reveals numerous previously unknown phenomena and offers insights - as well as a friendly workbench - for the future development of long context-capable LLMs. The source code is available at https://github.com/henryzhongsc/longctx_bench."
                    },
                    {
                        "title": "Taylor Unswift: Secured Weight Release for Large Language Models via Taylor Expansion",
                        "abstract": "Ensuring the security of released large language models (LLMs) poses a significant dilemma, as existing mechanisms either compromise ownership rights or raise data privacy concerns. To address this dilemma, we introduce TaylorMLP to protect the ownership of released LLMs and prevent their abuse. Specifically, TaylorMLP preserves the ownership of LLMs by transforming the weights of LLMs into parameters of Taylor-series. Instead of releasing the original weights, developers can release the Taylor-series parameters with users, thereby ensuring the security of LLMs. Moreover, TaylorMLP can prevent abuse of LLMs by adjusting the generation speed. It can induce low-speed token generation for the protected LLMs by increasing the terms in the Taylor-series. This intentional delay helps LLM developers prevent potential large-scale unauthorized uses of their models. Empirical experiments across five datasets and three LLM architectures demonstrate that TaylorMLP induces over 4x increase in latency, producing the tokens precisely matched with original LLMs. Subsequent defensive experiments further confirm that TaylorMLP effectively prevents users from reconstructing the weight values based on downstream datasets."
                    }
                ]
            },
            "e4585ba3-f8ca-4bc1-bf19-4b32472a5c7d": {
                "pk": "e4585ba3-f8ca-4bc1-bf19-4b32472a5c7d",
                "name": "Qiaoyu Tan",
                "collaborators": [
                    "Ninghao Liu",
                    "Xia Hu",
                    "Xiao Huang",
                    "Daochen Zha",
                    "Jingren Zhou",
                    "Hongxia Yang",
                    "Xin Zhang",
                    "Sirui Ding",
                    "Sheng Zhou",
                    "Yi Fang"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Representation Learning",
                    "Anomaly Detection",
                    "AutoML"
                ],
                "publications": [
                    {
                        "title": "Deep Representation Learning for Social Network Analysis",
                        "abstract": "Social network analysis is an important problem in data mining. A fundamental step for analyzing social networks is to encode network data into low-dimensional representations, i.e., network embeddings, so that the network topology structure and other attribute information can be effectively preserved. Network representation leaning facilitates further applications such as classification, link prediction, anomaly detection and clustering. In addition, techniques based on deep neural networks have attracted great interests over the past a few years. In this survey, we conduct a comprehensive review of current literature in network representation learning utilizing neural network models. First, we introduce the basic models for learning node representations in homogeneous networks. Meanwhile, we will also introduce some extensions of the base models in tackling more complex scenarios, such as analyzing attributed networks, heterogeneous networks and dynamic networks. Then, we introduce the techniques for embedding subgraphs. After that, we present the applications of network representation learning. At the end, we discuss some promising research directions for future work."
                    },
                    {
                        "title": "Graph Contrastive Learning with Personalized Augmentation",
                        "abstract": "Graph contrastive learning (GCL) has emerged as an effective tool for learning unsupervised representations of graphs. The key idea is to maximize the agreement between two augmented views of each graph via data augmentation. Existing GCL models mainly focus on applying \\textit{identical augmentation strategies} for all graphs within a given scenario. However, real-world graphs are often not monomorphic but abstractions of diverse natures. Even within the same scenario (e.g., macromolecules and online communities), different graphs might need diverse augmentations to perform effective GCL. Thus, blindly augmenting all graphs without considering their individual characteristics may undermine the performance of GCL arts.To deal with this, we propose the first principled framework, termed as \\textit{G}raph contrastive learning with \\textit{P}ersonalized \\textit{A}ugmentation (GPA), to advance conventional GCL by allowing each graph to choose its own suitable augmentation operations.In essence, GPA infers tailored augmentation strategies for each graph based on its topology and node attributes via a learnable augmentation selector, which is a plug-and-play module and can be effectively trained with downstream GCL models end-to-end. Extensive experiments across 11 benchmark graphs from different types and domains demonstrate the superiority of GPA against state-of-the-art competitors.Moreover, by visualizing the learned augmentation distributions across different types of datasets, we show that GPA can effectively identify the most suitable augmentations for each graph based on its characteristics."
                    },
                    {
                        "title": "Towards Personalized Preprocessing Pipeline Search",
                        "abstract": "Feature preprocessing, which transforms raw input features into numerical representations, is a crucial step in automated machine learning (AutoML) systems. However, the existing systems often have a very small search space for feature preprocessing with the same preprocessing pipeline applied to all the numerical features. This may result in sub-optimal performance since different datasets often have various feature characteristics, and features within a dataset may also have their own preprocessing preferences. To bridge this gap, we explore personalized preprocessing pipeline search, where the search algorithm is allowed to adopt a different preprocessing pipeline for each feature. This is a challenging task because the search space grows exponentially with more features. To tackle this challenge, we propose ClusterP3S, a novel framework for Personalized Preprocessing Pipeline Search via Clustering. The key idea is to learn feature clusters such that the search space can be significantly reduced by using the same preprocessing pipeline for the features within a cluster. To this end, we propose a hierarchical search strategy to jointly learn the clusters and search for the optimal pipelines, where the upper-level search optimizes the feature clustering to enable better pipelines built upon the clusters, and the lower-level search optimizes the pipeline given a specific cluster assignment. We instantiate this idea with a deep clustering network that is trained with reinforcement learning at the upper level, and random search at the lower level. Experiments on benchmark classification datasets demonstrate the effectiveness of enabling feature-wise preprocessing pipeline search."
                    },
                    {
                        "title": "E2GNN: Efficient Graph Neural Network Ensembles for Semi-Supervised Classification",
                        "abstract": "This work studies ensemble learning for graph neural networks (GNNs) under the popular semi-supervised setting. Ensemble learning has shown superiority in improving the accuracy and robustness of traditional machine learning by combining the outputs of multiple weak learners. However, adopting a similar idea to integrate different GNN models is challenging because of two reasons. First, GNN is notorious for its poor inference ability, so naively assembling multiple GNN models would deteriorate the inference efficiency. Second, when GNN models are trained with few labeled nodes, their performance are limited. In this case, the vanilla ensemble approach, e.g., majority vote, may be sub-optimal since most base models, i.e., GNNs, may make the wrong predictions. To this end, in this paper, we propose an efficient ensemble learner--E2GNN to assemble multiple GNNs in a learnable way by leveraging both labeled and unlabeled nodes. Specifically, we first pre-train different GNN models on a given data scenario according to the labeled nodes. Next, instead of directly combing their outputs for label inference, we train a simple multi-layer perceptron--MLP model to mimic their predictions on both labeled and unlabeled nodes. Then the unified MLP model is deployed to infer labels for unlabeled or new nodes. Since the predictions of unlabeled nodes from different GNN models may be incorrect, we develop a reinforced discriminator to effectively filter out those wrongly predicted nodes to boost the performance of MLP. By doing this, we suggest a principled approach to tackle the inference issues of GNN ensembles and maintain the merit of ensemble learning: improved performance. Comprehensive experiments over both transductive and inductive settings, across different GNN backbones and 8 benchmark datasets, demonstrate the superiority of E2GNN."
                    },
                    {
                        "title": "GAugLLM: Improving Graph Contrastive Learning for Text-Attributed Graphs with Large Language Models",
                        "abstract": "This work studies self-supervised graph learning for text-attributed graphs (TAGs) where nodes are represented by textual attributes. Unlike traditional graph contrastive methods that perturb the numerical feature space and alter the graph's topological structure, we aim to improve view generation through language supervision. This is driven by the prevalence of textual attributes in real applications, which complement graph structures with rich semantic information. However, this presents challenges because of two major reasons. First, text attributes often vary in length and quality, making it difficulty to perturb raw text descriptions without altering their original semantic meanings. Second, although text attributes complement graph structures, they are not inherently well-aligned. To bridge the gap, we introduce GAugLLM, a novel framework for augmenting TAGs. It leverages advanced large language models like Mistral to enhance self-supervised graph learning. Specifically, we introduce a mixture-of-prompt-expert technique to generate augmented node features. This approach adaptively maps multiple prompt experts, each of which modifies raw text attributes using prompt engineering, into numerical feature space. Additionally, we devise a collaborative edge modifier to leverage structural and textual commonalities, enhancing edge augmentation by examining or building connections between nodes. Empirical results across five benchmark datasets spanning various domains underscore our framework's ability to enhance the performance of leading contrastive methods as a plug-in tool. Notably, we observe that the augmented features and graph structure can also enhance the performance of standard generative methods, as well as popular graph neural networks. The open-sourced implementation of our GAugLLM is available at Github."
                    },
                    {
                        "title": "Reasoning Like a Doctor: Improving Medical Dialogue Systems via Diagnostic Reasoning Process Alignment",
                        "abstract": "Medical dialogue systems have attracted significant attention for their potential to act as medical assistants. Enabling these medical systems to emulate clinicians' diagnostic reasoning process has been the long-standing research focus. Previous studies rudimentarily realized the simulation of clinicians' diagnostic process by fine-tuning language models on high-quality dialogue datasets. Nonetheless, they overly focus on the outcomes of the clinician's reasoning process while ignoring their internal thought processes and alignment with clinician preferences. Our work aims to build a medical dialogue system that aligns with clinicians' diagnostic reasoning processes. We propose a novel framework, Emulation, designed to generate an appropriate response that relies on abductive and deductive diagnostic reasoning analyses and aligns with clinician preferences through thought process modeling. Experimental results on two datasets confirm the efficacy of Emulation. Crucially, our framework furnishes clear explanations for the generated responses, enhancing its transparency in medical consultations."
                    },
                    {
                        "title": "GiGaMAE: Generalizable Graph Masked Autoencoder via Collaborative Latent Space Reconstruction",
                        "abstract": "Self-supervised learning with masked autoencoders has recently gained popularity for its ability to produce effective image or textual representations, which can be applied to various downstream tasks without retraining. However, we observe that the current masked autoencoder models lack good generalization ability on graph data. To tackle this issue, we propose a novel graph masked autoencoder framework called GiGaMAE. Different from existing masked autoencoders that learn node presentations by explicitly reconstructing the original graph components (e.g., features or edges), in this paper, we propose to collaboratively reconstruct informative and integrated latent embeddings. By considering embeddings encompassing graph topology and attribute information as reconstruction targets, our model could capture more generalized and comprehensive knowledge. Furthermore, we introduce a mutual information based reconstruction loss that enables the effective reconstruction of multiple targets. This learning objective allows us to differentiate between the exclusive knowledge learned from a single target and common knowledge shared by multiple targets. We evaluate our method on three downstream tasks with seven datasets as benchmarks. Extensive experiments demonstrate the superiority of GiGaMAE against state-of-the-art baselines. We hope our results will shed light on the design of foundation models on graph-structured data. Our code is available at: https://github.com/sycny/GiGaMAE."
                    },
                    {
                        "title": "UniGLM: Training One Unified Language Model for Text-Attributed Graphs",
                        "abstract": "Representation learning on text-attributed graphs (TAGs), where nodes are represented by textual descriptions, is crucial for textual and relational knowledge systems and recommendation systems. Currently, state-of-the-art embedding methods for TAGs primarily focus on fine-tuning language models (e.g., BERT) using structure-aware training signals. While effective, these methods are tailored for individual TAG and cannot generalize across various graph scenarios. Given the shared textual space, leveraging multiple TAGs for joint fine-tuning, aligning text and graph structure from different aspects, would be more beneficial. Motivated by this, we introduce a novel Unified Graph Language Model (UniGLM) framework, the first graph embedding model that generalizes well to both in-domain and cross-domain TAGs. Specifically, UniGLM is trained over multiple TAGs with different domains and scales using self-supervised contrastive learning. UniGLM includes an adaptive positive sample selection technique for identifying structurally similar nodes and a lazy contrastive module that is devised to accelerate training by minimizing repetitive encoding calculations. Extensive empirical results across 9 benchmark TAGs demonstrate UniGLM's efficacy against leading embedding baselines in terms of generalization (various downstream tasks and backbones) and transfer learning (in and out of domain scenarios). The code is available at https://github.com/NYUSHCS/UniGLM."
                    },
                    {
                        "title": "MolecularGPT: Open Large Language Model (LLM) for Few-Shot Molecular Property Prediction",
                        "abstract": "Molecular property prediction (MPP) is a fundamental and crucial task in drug discovery. However, prior methods are limited by the requirement for a large number of labeled molecules and their restricted ability to generalize for unseen and new tasks, both of which are essential for real-world applications. To address these challenges, we present MolecularGPT for few-shot MPP. From a perspective on instruction tuning, we fine-tune large language models (LLMs) based on curated molecular instructions spanning over 1000 property prediction tasks. This enables building a versatile and specialized LLM that can be adapted to novel MPP tasks without any fine-tuning through zero- and few-shot in-context learning (ICL). MolecularGPT exhibits competitive in-context reasoning capabilities across 10 downstream evaluation datasets, setting new benchmarks for few-shot molecular prediction tasks. More importantly, with just two-shot examples, MolecularGPT can outperform standard supervised graph neural network methods on 4 out of 7 datasets. It also excels state-of-the-art LLM baselines by up to 15.7% increase on classification accuracy and decrease of 17.9 on regression metrics (e.g., RMSE) under zero-shot. This study demonstrates the potential of LLMs as effective few-shot molecular property predictors. The code is available at https://github.com/NYUSHCS/MolecularGPT."
                    },
                    {
                        "title": "MGAE: Masked Autoencoders for Self-Supervised Learning on Graphs",
                        "abstract": "We introduce a novel masked graph autoencoder (MGAE) framework to perform effective learning on graph structure data. Taking insights from self-supervised learning, we randomly mask a large proportion of edges and try to reconstruct these missing edges during training. MGAE has two core designs. First, we find that masking a high ratio of the input graph structure, e.g., $70\\%$, yields a nontrivial and meaningful self-supervisory task that benefits downstream applications. Second, we employ a graph neural network (GNN) as an encoder to perform message propagation on the partially-masked graph. To reconstruct the large number of masked edges, a tailored cross-correlation decoder is proposed. It could capture the cross-correlation between the head and tail nodes of anchor edge in multi-granularity. Coupling these two designs enables MGAE to be trained efficiently and effectively. Extensive experiments on multiple open datasets (Planetoid and OGB benchmarks) demonstrate that MGAE generally performs better than state-of-the-art unsupervised learning competitors on link prediction and node classification."
                    },
                    {
                        "title": "Sparse-Interest Network for Sequential Recommendation",
                        "abstract": "Recent methods in sequential recommendation focus on learning an overall embedding vector from a user's behavior sequence for the next-item recommendation. However, from empirical analysis, we discovered that a user's behavior sequence often contains multiple conceptually distinct items, while a unified embedding vector is primarily affected by one's most recent frequent actions. Thus, it may fail to infer the next preferred item if conceptually similar items are not dominant in recent interactions. To this end, an alternative solution is to represent each user with multiple embedding vectors encoding different aspects of the user's intentions. Nevertheless, recent work on multi-interest embedding usually considers a small number of concepts discovered via clustering, which may not be comparable to the large pool of item categories in real systems. It is a non-trivial task to effectively model a large number of diverse conceptual prototypes, as items are often not conceptually well clustered in fine granularity. Besides, an individual usually interacts with only a sparse set of concepts. In light of this, we propose a novel \\textbf{S}parse \\textbf{I}nterest \\textbf{NE}twork (SINE) for sequential recommendation. Our sparse-interest module can adaptively infer a sparse set of concepts for each user from the large concept pool and output multiple embeddings accordingly. Given multiple interest embeddings, we develop an interest aggregation module to actively predict the user's current intention and then use it to explicitly model multiple interests for next-item prediction. Empirical results on several public benchmark datasets and one large-scale industrial dataset demonstrate that SINE can achieve substantial improvement over state-of-the-art methods."
                    },
                    {
                        "title": "Dynamic Memory based Attention Network for Sequential Recommendation",
                        "abstract": "Sequential recommendation has become increasingly essential in various online services. It aims to model the dynamic preferences of users from their historical interactions and predict their next items. The accumulated user behavior records on real systems could be very long. This rich data brings opportunities to track actual interests of users. Prior efforts mainly focus on making recommendations based on relatively recent behaviors. However, the overall sequential data may not be effectively utilized, as early interactions might affect users' current choices. Also, it has become intolerable to scan the entire behavior sequence when performing inference for each user, since real-world system requires short response time. To bridge the gap, we propose a novel long sequential recommendation model, called Dynamic Memory-based Attention Network (DMAN). It segments the overall long behavior sequence into a series of sub-sequences, then trains the model and maintains a set of memory blocks to preserve long-term interests of users. To improve memory fidelity, DMAN dynamically abstracts each user's long-term interest into its own memory blocks by minimizing an auxiliary reconstruction loss. Based on the dynamic memory, the user's short-term and long-term interests can be explicitly extracted and combined for efficient joint recommendation. Empirical results over four benchmark datasets demonstrate the superiority of our model in capturing long-term dependency over various state-of-the-art sequential models."
                    },
                    {
                        "title": "Learning to Hash with Graph Neural Networks for Recommender Systems",
                        "abstract": "Graph representation learning has attracted much attention in supporting high quality candidate search at scale. Despite its effectiveness in learning embedding vectors for objects in the user-item interaction network, the computational costs to infer users' preferences in continuous embedding space are tremendous. In this work, we investigate the problem of hashing with graph neural networks (GNNs) for high quality retrieval, and propose a simple yet effective discrete representation learning framework to jointly learn continuous and discrete codes. Specifically, a deep hashing with GNNs (HashGNN) is presented, which consists of two components, a GNN encoder for learning node representations, and a hash layer for encoding representations to hash codes. The whole architecture is trained end-to-end by jointly optimizing two losses, i.e., reconstruction loss from reconstructing observed links, and ranking loss from preserving the relative ordering of hash codes. A novel discrete optimization strategy based on straight through estimator (STE) with guidance is proposed. The principal idea is to avoid gradient magnification in back-propagation of STE with continuous embedding guidance, in which we begin from learning an easier network that mimic the continuous embedding and let it evolve during the training until it finally goes back to STE. Comprehensive experiments over three publicly available and one real-world Alibaba company datasets demonstrate that our model not only can achieve comparable performance compared with its continuous counterpart but also runs multiple times faster during inference."
                    },
                    {
                        "title": "Collaborative Graph Neural Networks for Attributed Network Embedding",
                        "abstract": "Graph neural networks (GNNs) have shown prominent performance on attributed network embedding. However, existing efforts mainly focus on exploiting network structures, while the exploitation of node attributes is rather limited as they only serve as node features at the initial layer. This simple strategy impedes the potential of node attributes in augmenting node connections, leading to limited receptive field for inactive nodes with few or even no neighbors. Furthermore, the training objectives (i.e., reconstructing network structures) of most GNNs also do not include node attributes, although studies have shown that reconstructing node attributes is beneficial. Thus, it is encouraging to deeply involve node attributes in the key components of GNNs, including graph convolution operations and training objectives. However, this is a nontrivial task since an appropriate way of integration is required to maintain the merits of GNNs. To bridge the gap, in this paper, we propose COllaborative graph Neural Networks--CONN, a tailored GNN architecture for attribute network embedding. It improves model capacity by 1) selectively diffusing messages from neighboring nodes and involved attribute categories, and 2) jointly reconstructing node-to-node and node-to-attribute-category interactions via cross-correlation. Experiments on real-world networks demonstrate that CONN excels state-of-the-art embedding algorithms with a great margin."
                    },
                    {
                        "title": "Retrieval-enhanced Knowledge Editing in Language Models for Multi-Hop Question Answering",
                        "abstract": "Large Language Models (LLMs) have shown proficiency in question-answering tasks but often struggle to integrate real-time knowledge, leading to potentially outdated or inaccurate responses. This problem becomes even more challenging when dealing with multi-hop questions, since they require LLMs to update and integrate multiple knowledge pieces relevant to the questions. To tackle the problem, we propose the Retrieval-Augmented model Editing (RAE) framework for multi-hop question answering. RAE first retrieves edited facts and then refines the language model through in-context learning. Specifically, our retrieval approach, based on mutual information maximization, leverages the reasoning abilities of LLMs to identify chain facts that traditional similarity-based searches might miss. In addition, our framework includes a pruning strategy to eliminate redundant information from the retrieved facts, which enhances the editing accuracy and mitigates the hallucination problem. Our framework is supported by theoretical justification for its fact retrieval efficacy. Finally, comprehensive evaluation across various LLMs validates RAE's ability in providing accurate answers with updated knowledge. Our code is available at: https://github.com/sycny/RAE."
                    },
                    {
                        "title": "Towards a Unified Framework of Clustering-based Anomaly Detection",
                        "abstract": "Unsupervised Anomaly Detection (UAD) plays a crucial role in identifying abnormal patterns within data without labeled examples, holding significant practical implications across various domains. Although the individual contributions of representation learning and clustering to anomaly detection are well-established, their interdependencies remain under-explored due to the absence of a unified theoretical framework. Consequently, their collective potential to enhance anomaly detection performance remains largely untapped. To bridge this gap, in this paper, we propose a novel probabilistic mixture model for anomaly detection to establish a theoretical connection among representation learning, clustering, and anomaly detection. By maximizing a novel anomaly-aware data likelihood, representation learning and clustering can effectively reduce the adverse impact of anomalous data and collaboratively benefit anomaly detection. Meanwhile, a theoretically substantiated anomaly score is naturally derived from this framework. Lastly, drawing inspiration from gravitational analysis in physics, we have devised an improved anomaly score that more effectively harnesses the combined power of representation learning and clustering. Extensive experiments, involving 17 baseline methods across 30 diverse datasets, validate the effectiveness and generalization capability of the proposed method, surpassing state-of-the-art methods."
                    },
                    {
                        "title": "Is a Single Vector Enough? Exploring Node Polysemy for Network Embedding",
                        "abstract": "Networks have been widely used as the data structure for abstracting real-world systems as well as organizing the relations among entities. Network embedding models are powerful tools in mapping nodes in a network into continuous vector-space representations in order to facilitate subsequent tasks such as classification and link prediction. Existing network embedding models comprehensively integrate all information of each node, such as links and attributes, towards a single embedding vector to represent the node's general role in the network. However, a real-world entity could be multifaceted, where it connects to different neighborhoods due to different motives or self-characteristics that are not necessarily correlated. For example, in a movie recommender system, a user may love comedies or horror movies simultaneously, but it is not likely that these two types of movies are mutually close in the embedding space, nor the user embedding vector could be sufficiently close to them at the same time. In this paper, we propose a polysemous embedding approach for modeling multiple facets of nodes, as motivated by the phenomenon of word polysemy in language modeling. Each facet of a node is mapped as an embedding vector, while we also maintain association degree between each pair of node and facet. The proposed method is adaptive to various existing embedding models, without significantly complicating the optimization process. We also discuss how to engage embedding vectors of different facets for inference tasks including classification and link prediction. Experiments on real-world datasets help comprehensively evaluate the performance of the proposed method."
                    },
                    {
                        "title": "Towards Automated Imbalanced Learning with Deep Hierarchical Reinforcement Learning",
                        "abstract": "Imbalanced learning is a fundamental challenge in data mining, where there is a disproportionate ratio of training samples in each class. Over-sampling is an effective technique to tackle imbalanced learning through generating synthetic samples for the minority class. While numerous over-sampling algorithms have been proposed, they heavily rely on heuristics, which could be sub-optimal since we may need different sampling strategies for different datasets and base classifiers, and they cannot directly optimize the performance metric. Motivated by this, we investigate developing a learning-based over-sampling algorithm to optimize the classification performance, which is a challenging task because of the huge and hierarchical decision space. At the high level, we need to decide how many synthetic samples to generate. At the low level, we need to determine where the synthetic samples should be located, which depends on the high-level decision since the optimal locations of the samples may differ for different numbers of samples. To address the challenges, we propose AutoSMOTE, an automated over-sampling algorithm that can jointly optimize different levels of decisions. Motivated by the success of SMOTE~\\cite{chawla2002smote} and its extensions, we formulate the generation process as a Markov decision process (MDP) consisting of three levels of policies to generate synthetic samples within the SMOTE search space. Then we leverage deep hierarchical reinforcement learning to optimize the performance metric on the validation data. Extensive experiments on six real-world datasets demonstrate that AutoSMOTE significantly outperforms the state-of-the-art resampling algorithms. The code is at https://github.com/daochenzha/autosmote"
                    },
                    {
                        "title": "Reinforcement Neighborhood Selection for Unsupervised Graph Anomaly Detection",
                        "abstract": "Unsupervised graph anomaly detection is crucial for various practical applications as it aims to identify anomalies in a graph that exhibit rare patterns deviating significantly from the majority of nodes. Recent advancements have utilized Graph Neural Networks (GNNs) to learn high-quality node representations for anomaly detection by aggregating information from neighborhoods. However, the presence of anomalies may render the observed neighborhood unreliable and result in misleading information aggregation for node representation learning. Selecting the proper neighborhood is critical for graph anomaly detection but also challenging due to the absence of anomaly-oriented guidance and the interdependence with representation learning. To address these issues, we utilize the advantages of reinforcement learning in adaptively learning in complex environments and propose a novel method that incorporates Reinforcement neighborhood selection for unsupervised graph ANomaly Detection (RAND). RAND begins by enriching the candidate neighbor pool of the given central node with multiple types of indirect neighbors. Next, RAND designs a tailored reinforcement anomaly evaluation module to assess the reliability and reward of considering the given neighbor. Finally, RAND selects the most reliable subset of neighbors based on these rewards and introduces an anomaly-aware aggregator to amplify messages from reliable neighbors while diminishing messages from unreliable ones. Extensive experiments on both three synthetic and two real-world datasets demonstrate that RAND outperforms the state-of-the-art methods."
                    }
                ]
            },
            "cb9f5fba-a768-4288-ab00-5f65f13bbc45": {
                "pk": "cb9f5fba-a768-4288-ab00-5f65f13bbc45",
                "name": "Kaixiong Zhou",
                "collaborators": [
                    "Xia Hu",
                    "Daochen Zha",
                    "Xin Wang",
                    "Xiao Huang",
                    "Kwei-Herng Lai",
                    "Qingquan Song",
                    "Ninghao Liu",
                    "Yili Wang",
                    "Tianlong Chen",
                    "Zirui Liu"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Reinforcement Learning",
                    "Differential Privacy",
                    "Time Series Classification"
                ],
                "publications": [
                    {
                        "title": "Auto-GNN: Neural Architecture Search of Graph Neural Networks",
                        "abstract": "Graph neural networks (GNN) has been successfully applied to operate on the graph-structured data. Given a specific scenario, rich human expertise and tremendous laborious trials are usually required to identify a suitable GNN architecture. It is because the performance of a GNN architecture is significantly affected by the choice of graph convolution components, such as aggregate function and hidden dimension. Neural architecture search (NAS) has shown its potential in discovering effective deep architectures for learning tasks in image and language modeling. However, existing NAS algorithms cannot be directly applied to the GNN search problem. First, the search space of GNN is different from the ones in existing NAS work. Second, the representation learning capacity of GNN architecture changes obviously with slight architecture modifications. It affects the search efficiency of traditional search methods. Third, widely used techniques in NAS such as parameter sharing might become unstable in GNN.   To bridge the gap, we propose the automated graph neural networks (AGNN) framework, which aims to find an optimal GNN architecture within a predefined search space. A reinforcement learning based controller is designed to greedily validate architectures via small steps. AGNN has a novel parameter sharing strategy that enables homogeneous architectures to share parameters, based on a carefully-designed homogeneity definition. Experiments on real-world benchmark datasets demonstrate that the GNN architecture identified by AGNN achieves the best performance, comparing with existing handcrafted models and tradistional search methods."
                    },
                    {
                        "title": "ENGAGE: Explanation Guided Data Augmentation for Graph Representation Learning",
                        "abstract": "The recent contrastive learning methods, due to their effectiveness in representation learning, have been widely applied to modeling graph data. Random perturbation is widely used to build contrastive views for graph data, which however, could accidentally break graph structures and lead to suboptimal performance. In addition, graph data is usually highly abstract, so it is hard to extract intuitive meanings and design more informed augmentation schemes. Effective representations should preserve key characteristics in data and abandon superfluous information. In this paper, we propose ENGAGE (ExplaNation Guided data AuGmEntation), where explanation guides the contrastive augmentation process to preserve the key parts in graphs and explore removing superfluous information. Specifically, we design an efficient unsupervised explanation method called smoothed activation map as the indicator of node importance in representation learning. Then, we design two data augmentation schemes on graphs for perturbing structural and feature information, respectively. We also provide justification for the proposed method in the framework of information theories. Experiments of both graph-level and node-level tasks, on various model architectures and on different real-world graphs, are conducted to demonstrate the effectiveness and flexibility of ENGAGE. The code of ENGAGE can be found: https://github.com/sycny/ENGAGE."
                    },
                    {
                        "title": "Experience Replay Optimization",
                        "abstract": "Experience replay enables reinforcement learning agents to memorize and reuse past experiences, just as humans replay memories for the situation at hand. Contemporary off-policy algorithms either replay past experiences uniformly or utilize a rule-based replay strategy, which may be sub-optimal. In this work, we consider learning a replay policy to optimize the cumulative reward. Replay learning is challenging because the replay memory is noisy and large, and the cumulative reward is unstable. To address these issues, we propose a novel experience replay optimization (ERO) framework which alternately updates two policies: the agent policy, and the replay policy. The agent is updated to maximize the cumulative reward based on the replayed data, while the replay policy is updated to provide the agent with the most useful experiences. The conducted experiments on various continuous control tasks demonstrate the effectiveness of ERO, empirically showing promise in experience replay learning to improve the performance of off-policy reinforcement learning algorithms."
                    },
                    {
                        "title": "Simplifying Deep Reinforcement Learning via Self-Supervision",
                        "abstract": "Supervised regression to demonstrations has been demonstrated to be a stable way to train deep policy networks. We are motivated to study how we can take full advantage of supervised loss functions for stably training deep reinforcement learning agents. This is a challenging task because it is unclear how the training data could be collected to enable policy improvement. In this work, we propose Self-Supervised Reinforcement Learning (SSRL), a simple algorithm that optimizes policies with purely supervised losses. We demonstrate that, without policy gradient or value estimation, an iterative procedure of ``labeling\" data and supervised regression is sufficient to drive stable policy improvement. By selecting and imitating trajectories with high episodic rewards, SSRL is surprisingly competitive to contemporary algorithms with more stable performance and less running time, showing the potential of solving reinforcement learning with supervised learning techniques. The code is available at https://github.com/daochenzha/SSRL"
                    },
                    {
                        "title": "Differentially Private Counterfactuals via Functional Mechanism",
                        "abstract": "Counterfactual, serving as one emerging type of model explanation, has attracted tons of attentions recently from both industry and academia. Different from the conventional feature-based explanations (e.g., attributions), counterfactuals are a series of hypothetical samples which can flip model decisions with minimal perturbations on queries. Given valid counterfactuals, humans are capable of reasoning under ``what-if'' circumstances, so as to better understand the model decision boundaries. However, releasing counterfactuals could be detrimental, since it may unintentionally leak sensitive information to adversaries, which brings about higher risks on both model security and data privacy. To bridge the gap, in this paper, we propose a novel framework to generate differentially private counterfactual (DPC) without touching the deployed model or explanation set, where noises are injected for protection while maintaining the explanation roles of counterfactual. In particular, we train an autoencoder with the functional mechanism to construct noisy class prototypes, and then derive the DPC from the latent prototypes based on the post-processing immunity of differential privacy. Further evaluations demonstrate the effectiveness of the proposed framework, showing that DPC can successfully relieve the risks on both extraction and inference attacks."
                    },
                    {
                        "title": "A Survey of Graph Prompting Methods: Techniques, Applications, and Challenges",
                        "abstract": "The recent \"pre-train, prompt, predict training\" paradigm has gained popularity as a way to learn generalizable models with limited labeled data. The approach involves using a pre-trained model and a prompting function that applies a template to input samples, adding indicative context and reformulating target tasks as the pre-training task. However, the design of prompts could be a challenging and time-consuming process in complex tasks. The limitation can be addressed by using graph data, as graphs serve as structured knowledge repositories by explicitly modeling the interaction between entities. In this survey, we review prompting methods from the graph perspective, where prompting functions are augmented with graph knowledge. In particular, we introduce the basic concepts of graph prompt learning, organize the existing work of designing graph prompting functions, and describe their applications and future challenges. This survey will bridge the gap between graphs and prompt design to facilitate future methodology development."
                    },
                    {
                        "title": "Efficient Sharpness-Aware Minimization for Molecular Graph Transformer Models",
                        "abstract": "Sharpness-aware minimization (SAM) has received increasing attention in computer vision since it can effectively eliminate the sharp local minima from the training trajectory and mitigate generalization degradation. However, SAM requires two sequential gradient computations during the optimization of each step: one to obtain the perturbation gradient and the other to obtain the updating gradient. Compared with the base optimizer (e.g., Adam), SAM doubles the time overhead due to the additional perturbation gradient. By dissecting the theory of SAM and observing the training gradient of the molecular graph transformer, we propose a new algorithm named GraphSAM, which reduces the training cost of SAM and improves the generalization performance of graph transformer models. There are two key factors that contribute to this result: (i) \\textit{gradient approximation}: we use the updating gradient of the previous step to approximate the perturbation gradient at the intermediate steps smoothly (\\textbf{increases efficiency}); (ii) \\textit{loss landscape approximation}: we theoretically prove that the loss landscape of GraphSAM is limited to a small range centered on the expected loss of SAM (\\textbf{guarantees generalization performance}). The extensive experiments on six datasets with different tasks demonstrate the superiority of GraphSAM, especially in optimizing the model update process. The code is in:https://github.com/YL-wang/GraphSAM/tree/graphsam"
                    },
                    {
                        "title": "COSCO: A Sharpness-Aware Training Framework for Few-shot Multivariate Time Series Classification",
                        "abstract": "Multivariate time series classification is an important task with widespread domains of applications. Recently, deep neural networks (DNN) have achieved state-of-the-art performance in time series classification. However, they often require large expert-labeled training datasets which can be infeasible in practice. In few-shot settings, i.e. only a limited number of samples per class are available in training data, DNNs show a significant drop in testing accuracy and poor generalization ability. In this paper, we propose to address these problems from an optimization and a loss function perspective. Specifically, we propose a new learning framework named COSCO consisting of a sharpness-aware minimization (SAM) optimization and a Prototypical loss function to improve the generalization ability of DNN for multivariate time series classification problems under few-shot setting. Our experiments demonstrate our proposed method outperforms the existing baseline methods. Our source code is available at: https://github.com/JRB9/COSCO."
                    },
                    {
                        "title": "Policy-GNN: Aggregation Optimization for Graph Neural Networks",
                        "abstract": "Graph data are pervasive in many real-world applications. Recently, increasing attention has been paid on graph neural networks (GNNs), which aim to model the local graph structures and capture the hierarchical patterns by aggregating the information from neighbors with stackable network modules. Motivated by the observation that different nodes often require different iterations of aggregation to fully capture the structural information, in this paper, we propose to explicitly sample diverse iterations of aggregation for different nodes to boost the performance of GNNs. It is a challenging task to develop an effective aggregation strategy for each node, given complex graphs and sparse features. Moreover, it is not straightforward to derive an efficient algorithm since we need to feed the sampled nodes into different number of network layers. To address the above challenges, we propose Policy-GNN, a meta-policy framework that models the sampling procedure and message passing of GNNs into a combined learning process. Specifically, Policy-GNN uses a meta-policy to adaptively determine the number of aggregations for each node. The meta-policy is trained with deep reinforcement learning (RL) by exploiting the feedback from the model. We further introduce parameter sharing and a buffer mechanism to boost the training efficiency. Experimental results on three real-world benchmark datasets suggest that Policy-GNN significantly outperforms the state-of-the-art alternatives, showing the promise in aggregation optimization for GNNs."
                    },
                    {
                        "title": "Towards Similarity-Aware Time-Series Classification",
                        "abstract": "We study time-series classification (TSC), a fundamental task of time-series data mining. Prior work has approached TSC from two major directions: (1) similarity-based methods that classify time-series based on the nearest neighbors, and (2) deep learning models that directly learn the representations for classification in a data-driven manner. Motivated by the different working mechanisms within these two research lines, we aim to connect them in such a way as to jointly model time-series similarities and learn the representations. This is a challenging task because it is unclear how we should efficiently leverage similarity information. To tackle the challenge, we propose Similarity-Aware Time-Series Classification (SimTSC), a conceptually simple and general framework that models similarity information with graph neural networks (GNNs). Specifically, we formulate TSC as a node classification problem in graphs, where the nodes correspond to time-series, and the links correspond to pair-wise similarities. We further design a graph construction strategy and a batch training algorithm with negative sampling to improve training efficiency. We instantiate SimTSC with ResNet as the backbone and Dynamic Time Warping (DTW) as the similarity measure. Extensive experiments on the full UCR datasets and several multivariate datasets demonstrate the effectiveness of incorporating similarity information into deep learning models in both supervised and semi-supervised settings. Our code is available at https://github.com/daochenzha/SimTSC"
                    },
                    {
                        "title": "MolCPT: Molecule Continuous Prompt Tuning to Generalize Molecular Representation Learning",
                        "abstract": "Molecular representation learning is crucial for the problem of molecular property prediction, where graph neural networks (GNNs) serve as an effective solution due to their structure modeling capabilities. Since labeled data is often scarce and expensive to obtain, it is a great challenge for GNNs to generalize in the extensive molecular space. Recently, the training paradigm of \"pre-train, fine-tune\" has been leveraged to improve the generalization capabilities of GNNs. It uses self-supervised information to pre-train the GNN, and then performs fine-tuning to optimize the downstream task with just a few labels. However, pre-training does not always yield statistically significant improvement, especially for self-supervised learning with random structural masking. In fact, the molecular structure is characterized by motif subgraphs, which are frequently occurring and influence molecular properties. To leverage the task-related motifs, we propose a novel paradigm of \"pre-train, prompt, fine-tune\" for molecular representation learning, named molecule continuous prompt tuning (MolCPT). MolCPT defines a motif prompting function that uses the pre-trained model to project the standalone input into an expressive prompt. The prompt effectively augments the molecular graph with meaningful motifs in the continuous representation space; this provides more structural patterns to aid the downstream classifier in identifying molecular properties. Extensive experiments on several benchmark datasets show that MolCPT efficiently generalizes pre-trained GNNs for molecular property prediction, with or without a few fine-tuning steps."
                    },
                    {
                        "title": "Optimizing OOD Detection in Molecular Graphs: A Novel Approach with Diffusion Models",
                        "abstract": "The open-world test dataset is often mixed with out-of-distribution (OOD) samples, where the deployed models will struggle to make accurate predictions. Traditional detection methods need to trade off OOD detection and in-distribution (ID) classification performance since they share the same representation learning model. In this work, we propose to detect OOD molecules by adopting an auxiliary diffusion model-based framework, which compares similarities between input molecules and reconstructed graphs. Due to the generative bias towards reconstructing ID training samples, the similarity scores of OOD molecules will be much lower to facilitate detection. Although it is conceptually simple, extending this vanilla framework to practical detection applications is still limited by two significant challenges. First, the popular similarity metrics based on Euclidian distance fail to consider the complex graph structure. Second, the generative model involving iterative denoising steps is time-consuming especially when it runs on the enormous pool of drugs. To address these challenges, our research pioneers an approach of Prototypical Graph Reconstruction for Molecular OOD Detection, dubbed as PGR-MOOD and hinges on three innovations: i) An effective metric to comprehensively quantify the matching degree of input and reconstructed molecules; ii) A creative graph generator to construct prototypical graphs that are in line with ID but away from OOD; iii) An efficient and scalable OOD detector to compare the similarity between test samples and pre-constructed prototypical graphs and omit the generative process on every new molecule. Extensive experiments on ten benchmark datasets and six baselines are conducted to demonstrate our superiority."
                    },
                    {
                        "title": "Benchmark on Drug Target Interaction Modeling from a Structure Perspective",
                        "abstract": "The prediction modeling of drug-target interactions is crucial to drug discovery and design, which has seen rapid advancements owing to deep learning technologies. Recently developed methods, such as those based on graph neural networks (GNNs) and Transformers, demonstrate exceptional performance across various datasets by effectively extracting structural information. However, the benchmarking of these novel methods often varies significantly in terms of hyperparameter settings and datasets, which limits algorithmic progress. In view of these, we conduct a comprehensive survey and benchmark for drug-target interaction modeling from a structure perspective, via integrating tens of explicit (i.e., GNN-based) and implicit (i.e., Transformer-based) structure learning algorithms. To this end, we first unify the hyperparameter setting within each class of structure learning methods. Moreover, we conduct a macroscopical comparison between these two classes of encoding strategies as well as the different featurization techniques that inform molecules' chemical and physical properties. We then carry out the microscopical comparison between all the integrated models across the six datasets, via comprehensively benchmarking their effectiveness and efficiency. Remarkably, the summarized insights from the benchmark studies lead to the design of model combos. We demonstrate that our combos can achieve new state-of-the-art performance on various datasets associated with cost-effective memory and computation. Our code is available at \\hyperlink{https://github.com/justinwjl/GTB-DTI/tree/main}{https://github.com/justinwjl/GTB-DTI/tree/main}."
                    },
                    {
                        "title": "Cross-Lingual Multi-Hop Knowledge Editing -- Benchmarks, Analysis and a Simple Contrastive Learning based Approach",
                        "abstract": "Large language models are often expected to constantly adapt to new sources of knowledge and knowledge editing techniques aim to efficiently patch the outdated model knowledge, with minimal modification. Most prior works focus on monolingual knowledge editing in English, even though new information can emerge in any language from any part of the world. We propose the Cross-Lingual Multi-Hop Knowledge Editing paradigm, for measuring and analyzing the performance of various SoTA knowledge editing techniques in a cross-lingual setup. Specifically, we create a parallel cross-lingual benchmark, CROLIN-MQUAKE for measuring the knowledge editing capabilities. Our extensive analysis over various knowledge editing techniques uncover significant gaps in performance between the cross-lingual and English-centric setting. Following this, we propose a significantly improved system for cross-lingual multi-hop knowledge editing, CLEVER-CKE. CLEVER-CKE is based on a retrieve, verify and generate knowledge editing framework, where a retriever is formulated to recall edited facts and support an LLM to adhere to knowledge edits. We develop language-aware and hard-negative based contrastive objectives for improving the cross-lingual and fine-grained fact retrieval and verification process used in this framework. Extensive experiments on three LLMs, eight languages, and two datasets show CLEVER-CKE's significant gains of up to 30% over prior methods."
                    },
                    {
                        "title": "Pioneering Reliable Assessment in Text-to-Image Knowledge Editing: Leveraging a Fine-Grained Dataset and an Innovative Criterion",
                        "abstract": "During pre-training, the Text-to-Image (T2I) diffusion models encode factual knowledge into their parameters. These parameterized facts enable realistic image generation, but they may become obsolete over time, thereby misrepresenting the current state of the world. Knowledge editing techniques aim to update model knowledge in a targeted way. However, facing the dual challenges posed by inadequate editing datasets and unreliable evaluation criterion, the development of T2I knowledge editing encounter difficulties in effectively generalizing injected knowledge. In this work, we design a T2I knowledge editing framework by comprehensively spanning on three phases: First, we curate a dataset \\textbf{CAKE}, comprising paraphrase and multi-object test, to enable more fine-grained assessment on knowledge generalization. Second, we propose a novel criterion, \\textbf{adaptive CLIP threshold}, to effectively filter out false successful images under the current criterion and achieve reliable editing evaluation. Finally, we introduce \\textbf{MPE}, a simple but effective approach for T2I knowledge editing. Instead of tuning parameters, MPE precisely recognizes and edits the outdated part of the conditioning text-prompt to accommodate the up-to-date knowledge. A straightforward implementation of MPE (Based on in-context learning) exhibits better overall performance than previous model editors. We hope these efforts can further promote faithful evaluation of T2I knowledge editing methods."
                    },
                    {
                        "title": "Towards Deeper Graph Neural Networks with Differentiable Group Normalization",
                        "abstract": "Graph neural networks (GNNs), which learn the representation of a node by aggregating its neighbors, have become an effective computational tool in downstream applications. Over-smoothing is one of the key issues which limit the performance of GNNs as the number of layers increases. It is because the stacked aggregators would make node representations converge to indistinguishable vectors. Several attempts have been made to tackle the issue by bringing linked node pairs close and unlinked pairs distinct. However, they often ignore the intrinsic community structures and would result in sub-optimal performance. The representations of nodes within the same community/class need be similar to facilitate the classification, while different classes are expected to be separated in embedding space. To bridge the gap, we introduce two over-smoothing metrics and a novel technique, i.e., differentiable group normalization (DGN). It normalizes nodes within the same group independently to increase their smoothness, and separates node distributions among different groups to significantly alleviate the over-smoothing issue. Experiments on real-world datasets demonstrate that DGN makes GNN models more robust to over-smoothing and achieves better performance with deeper GNNs."
                    },
                    {
                        "title": "Multi-Channel Graph Convolutional Networks",
                        "abstract": "Graph neural networks (GNN) has been demonstrated to be effective in classifying graph structures. To further improve the graph representation learning ability, hierarchical GNN has been explored. It leverages the differentiable pooling to cluster nodes into fixed groups, and generates a coarse-grained structure accompanied with the shrinking of the original graph. However, such clustering would discard some graph information and achieve the suboptimal results. It is because the node inherently has different characteristics or roles, and two non-isomorphic graphs may have the same coarse-grained structure that cannot be distinguished after pooling. To compensate the loss caused by coarse-grained clustering and further advance GNN, we propose a multi-channel graph convolutional networks (MuchGCN). It is motivated by the convolutional neural networks, at which a series of channels are encoded to preserve the comprehensive characteristics of the input image. Thus, we define the specific graph convolutions to learn a series of graph channels at each layer, and pool graphs iteratively to encode the hierarchical structures. Experiments have been carefully carried out to demonstrate the superiority of MuchGCN over the state-of-the-art graph classification algorithms."
                    },
                    {
                        "title": "Privacy-preserving Fine-tuning of Large Language Models through Flatness",
                        "abstract": "The privacy concerns associated with the use of Large Language Models (LLMs) have grown recently with the development of LLMs such as ChatGPT. Differential Privacy (DP) techniques are explored in existing work to mitigate their privacy risks at the cost of generalization degradation. Our paper reveals that the flatness of DP-trained models' loss landscape plays an essential role in the trade-off between their privacy and generalization. We further propose a holistic framework to enforce appropriate weight flatness, which substantially improves model generalization with competitive privacy preservation. It innovates from three coarse-to-grained levels, including perturbation-aware min-max optimization on model weights within a layer, flatness-guided sparse prefix-tuning on weights across layers, and weight knowledge distillation between DP \\& non-DP weights copies. Comprehensive experiments of both black-box and white-box scenarios are conducted to demonstrate the effectiveness of our proposal in enhancing generalization and maintaining DP characteristics. For instance, on text classification dataset QNLI, DP-Flat achieves similar performance with non-private full fine-tuning but with DP guarantee under privacy budget $\\epsilon=3$, and even better performance given higher privacy budgets. Codes are provided in the supplement."
                    },
                    {
                        "title": "Orthogonal Graph Neural Networks",
                        "abstract": "Graph neural networks (GNNs) have received tremendous attention due to their superiority in learning node representations. These models rely on message passing and feature transformation functions to encode the structural and feature information from neighbors. However, stacking more convolutional layers significantly decreases the performance of GNNs. Most recent studies attribute this limitation to the over-smoothing issue, where node embeddings converge to indistinguishable vectors. Through a number of experimental observations, we argue that the main factor degrading the performance is the unstable forward normalization and backward gradient resulted from the improper design of the feature transformation, especially for shallow GNNs where the over-smoothing has not happened. Therefore, we propose a novel orthogonal feature transformation, named Ortho-GConv, which could generally augment the existing GNN backbones to stabilize the model training and improve the model's generalization performance. Specifically, we maintain the orthogonality of the feature transformation comprehensively from three perspectives, namely hybrid weight initialization, orthogonal transformation, and orthogonal regularization. By equipping the existing GNNs (e.g. GCN, JKNet, GCNII) with Ortho-GConv, we demonstrate the generality of the orthogonal feature transformation to enable stable training, and show its effectiveness for node and graph classification tasks."
                    },
                    {
                        "title": "Towards Interaction Detection Using Topological Analysis on Neural Networks",
                        "abstract": "Detecting statistical interactions between input features is a crucial and challenging task. Recent advances demonstrate that it is possible to extract learned interactions from trained neural networks. It has also been observed that, in neural networks, any interacting features must follow a strongly weighted connection to common hidden units. Motivated by the observation, in this paper, we propose to investigate the interaction detection problem from a novel topological perspective by analyzing the connectivity in neural networks. Specially, we propose a new measure for quantifying interaction strength, based upon the well-received theory of persistent homology. Based on this measure, a Persistence Interaction detection~(PID) algorithm is developed to efficiently detect interactions. Our proposed algorithm is evaluated across a number of interaction detection tasks on several synthetic and real world datasets with different hyperparameters. Experimental results validate that the PID algorithm outperforms the state-of-the-art baselines."
                    }
                ]
            },
            "e67d4516-2235-48cb-872f-80e86cb66f90": {
                "pk": "e67d4516-2235-48cb-872f-80e86cb66f90",
                "name": "Na Zou",
                "collaborators": [
                    "Xia Hu",
                    "Mengnan Du",
                    "Fan Yang",
                    "Shuiwang Ji",
                    "Yuening Li",
                    "Daochen Zha",
                    "Ninghao Liu",
                    "Jundong Li",
                    "Zhengyang Wang",
                    "Dinggang Shen"
                ],
                "domain": [
                    "Fairness in Machine Learning",
                    "Deep Learning",
                    "Anomaly Detection",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Fairness in Deep Learning: A Computational Perspective",
                        "abstract": "Deep learning is increasingly being used in high-stake decision making applications that affect individual lives. However, deep learning models might exhibit algorithmic discrimination behaviors with respect to protected groups, potentially posing negative impacts on individuals and society. Therefore, fairness in deep learning has attracted tremendous attention recently. We provide a review covering recent progresses to tackle algorithmic fairness problems of deep learning from the computational perspective. Specifically, we show that interpretability can serve as a useful ingredient to diagnose the reasons that lead to algorithmic discrimination. We also discuss fairness mitigation approaches categorized according to three stages of deep learning life-cycle, aiming to push forward the area of fairness in deep learning and build genuinely fair and reliable deep learning systems."
                    },
                    {
                        "title": "Non-local U-Net for Biomedical Image Segmentation",
                        "abstract": "Deep learning has shown its great promise in various biomedical image segmentation tasks. Existing models are typically based on U-Net and rely on an encoder-decoder architecture with stacked local operators to aggregate long-range information gradually. However, only using the local operators limits the efficiency and effectiveness. In this work, we propose the non-local U-Nets, which are equipped with flexible global aggregation blocks, for biomedical image segmentation. These blocks can be inserted into U-Net as size-preserving processes, as well as down-sampling and up-sampling layers. We perform thorough experiments on the 3D multimodality isointense infant brain MR image segmentation task to evaluate the non-local U-Nets. Results show that our proposed models achieve top performances with fewer parameters and faster computation."
                    },
                    {
                        "title": "PyODDS: An End-to-End Outlier Detection System",
                        "abstract": "PyODDS is an end-to end Python system for outlier detection with database support. PyODDS provides outlier detection algorithms which meet the demands for users in different fields, w/wo data science or machine learning background. PyODDS gives the ability to execute machine learning algorithms in-database without moving data out of the database server or over the network. It also provides access to a wide range of outlier detection algorithms, including statistical analysis and more recent deep learning based approaches. PyODDS is released under the MIT open-source license, and currently available at (https://github.com/datamllab/pyodds) with official documentations at (https://pyodds.github.io/)."
                    },
                    {
                        "title": "Modeling Techniques for Machine Learning Fairness: A Survey",
                        "abstract": "Machine learning models are becoming pervasive in high-stakes applications. Despite their clear benefits in terms of performance, the models could show discrimination against minority groups and result in fairness issues in a decision-making process, leading to severe negative impacts on the individuals and the society. In recent years, various techniques have been developed to mitigate the unfairness for machine learning models. Among them, in-processing methods have drawn increasing attention from the community, where fairness is directly taken into consideration during model design to induce intrinsically fair models and fundamentally mitigate fairness issues in outputs and representations. In this survey, we review the current progress of in-processing fairness mitigation techniques. Based on where the fairness is achieved in the model, we categorize them into explicit and implicit methods, where the former directly incorporates fairness metrics in training objectives, and the latter focuses on refining latent representation learning. Finally, we conclude the survey with a discussion of the research challenges in this community to motivate future exploration."
                    },
                    {
                        "title": "Fair Machine Learning in Healthcare: A Review",
                        "abstract": "The digitization of healthcare data coupled with advances in computational capabilities has propelled the adoption of machine learning (ML) in healthcare. However, these methods can perpetuate or even exacerbate existing disparities, leading to fairness concerns such as the unequal distribution of resources and diagnostic inaccuracies among different demographic groups. Addressing these fairness problem is paramount to prevent further entrenchment of social injustices. In this survey, we analyze the intersection of fairness in machine learning and healthcare disparities. We adopt a framework based on the principles of distributive justice to categorize fairness concerns into two distinct classes: equal allocation and equal performance. We provide a critical review of the associated fairness metrics from a machine learning standpoint and examine biases and mitigation strategies across the stages of the ML lifecycle, discussing the relationship between biases and their countermeasures. The paper concludes with a discussion on the pressing challenges that remain unaddressed in ensuring fairness in healthcare ML, and proposes several new research directions that hold promise for developing ethical and equitable ML applications in healthcare."
                    },
                    {
                        "title": "Beyond Fairness: Age-Harmless Parkinson's Detection via Voice",
                        "abstract": "Parkinson's disease (PD), a neurodegenerative disorder, often manifests as speech and voice dysfunction. While utilizing voice data for PD detection has great potential in clinical applications, the widely used deep learning models currently have fairness issues regarding different ages of onset. These deep models perform well for the elderly group (age $>$ 55) but are less accurate for the young group (age $\\leq$ 55). Through our investigation, the discrepancy between the elderly and the young arises due to 1) an imbalanced dataset and 2) the milder symptoms often seen in early-onset patients. However, traditional debiasing methods are impractical as they typically impair the prediction accuracy for the majority group while minimizing the discrepancy. To address this issue, we present a new debiasing method using GradCAM-based feature masking combined with ensemble models, ensuring that neither fairness nor accuracy is compromised. Specifically, the GradCAM-based feature masking selectively obscures age-related features in the input voice data while preserving essential information for PD detection. The ensemble models further improve the prediction accuracy for the minority (young group). Our approach effectively improves detection accuracy for early-onset patients without sacrificing performance for the elderly group. Additionally, we propose a two-step detection strategy for the young group, offering a practical risk assessment for potential early-onset PD patients."
                    },
                    {
                        "title": "SpecAE: Spectral AutoEncoder for Anomaly Detection in Attributed Networks",
                        "abstract": "Anomaly detection aims to distinguish observations that are rare and different from the majority. While most existing algorithms assume that instances are i.i.d., in many practical scenarios, links describing instance-to-instance dependencies and interactions are available. Such systems are called attributed networks. Anomaly detection in attributed networks has various applications such as monitoring suspicious accounts in social media and financial fraud in transaction networks. However, it remains a challenging task since the definition of anomaly becomes more complicated and topological structures are heterogeneous with nodal attributes. In this paper, we propose a spectral convolution and deconvolution based framework -- SpecAE, to project the attributed network into a tailored space to detect global and community anomalies. SpecAE leverages Laplacian sharpening to amplify the distances between representations of anomalies and the ones of the majority. The learned representations along with reconstruction errors are combined with a density estimation model to perform the detection. They are trained jointly as an end-to-end framework. Experiments on real-world datasets demonstrate the effectiveness of SpecAE."
                    },
                    {
                        "title": "Defense Against Explanation Manipulation",
                        "abstract": "Explainable machine learning attracts increasing attention as it improves transparency of models, which is helpful for machine learning to be trusted in real applications. However, explanation methods have recently been demonstrated to be vulnerable to manipulation, where we can easily change a model's explanation while keeping its prediction constant. To tackle this problem, some efforts have been paid to use more stable explanation methods or to change model configurations. In this work, we tackle the problem from the training perspective, and propose a new training scheme called Adversarial Training on EXplanations (ATEX) to improve the internal explanation stability of a model regardless of the specific explanation method being applied. Instead of directly specifying explanation values over data instances, ATEX only puts requirement on model predictions which avoids involving second-order derivatives in optimization. As a further discussion, we also find that explanation stability is closely related to another property of the model, i.e., the risk of being exposed to adversarial attack. Through experiments, besides showing that ATEX improves model robustness against manipulation targeting explanation, it also brings additional benefits including smoothing explanations and improving the efficacy of adversarial training if applied to the model."
                    },
                    {
                        "title": "Graph Mixup with Soft Alignments",
                        "abstract": "We study graph data augmentation by mixup, which has been used successfully on images. A key operation of mixup is to compute a convex combination of a pair of inputs. This operation is straightforward for grid-like data, such as images, but challenging for graph data. The key difficulty lies in the fact that different graphs typically have different numbers of nodes, and thus there lacks a node-level correspondence between graphs. In this work, we propose S-Mixup, a simple yet effective mixup method for graph classification by soft alignments. Specifically, given a pair of graphs, we explicitly obtain node-level correspondence via computing a soft assignment matrix to match the nodes between two graphs. Based on the soft assignments, we transform the adjacency and node feature matrices of one graph, so that the transformed graph is aligned with the other graph. In this way, any pair of graphs can be mixed directly to generate an augmented graph. We conduct systematic experiments to show that S-Mixup can improve the performance and generalization of graph neural networks (GNNs) on various graph classification tasks. In addition, we show that S-Mixup can increase the robustness of GNNs against noisy labels."
                    },
                    {
                        "title": "IDEA: A Flexible Framework of Certified Unlearning for Graph Neural Networks",
                        "abstract": "Graph Neural Networks (GNNs) have been increasingly deployed in a plethora of applications. However, the graph data used for training may contain sensitive personal information of the involved individuals. Once trained, GNNs typically encode such information in their learnable parameters. As a consequence, privacy leakage may happen when the trained GNNs are deployed and exposed to potential attackers. Facing such a threat, machine unlearning for GNNs has become an emerging technique that aims to remove certain personal information from a trained GNN. Among these techniques, certified unlearning stands out, as it provides a solid theoretical guarantee of the information removal effectiveness. Nevertheless, most of the existing certified unlearning methods for GNNs are only designed to handle node and edge unlearning requests. Meanwhile, these approaches are usually tailored for either a specific design of GNN or a specially designed training objective. These disadvantages significantly jeopardize their flexibility. In this paper, we propose a principled framework named IDEA to achieve flexible and certified unlearning for GNNs. Specifically, we first instantiate four types of unlearning requests on graphs, and then we propose an approximation approach to flexibly handle these unlearning requests over diverse GNNs. We further provide theoretical guarantee of the effectiveness for the proposed approach as a certification. Different from existing alternatives, IDEA is not designed for any specific GNNs or optimization objectives to perform certified unlearning, and thus can be easily generalized. Extensive experiments on real-world datasets demonstrate the superiority of IDEA in multiple key perspectives."
                    }
                ]
            },
            "5fa0cc2a-15de-40ef-8cad-3f3b1418ec0b": {
                "pk": "5fa0cc2a-15de-40ef-8cad-3f3b1418ec0b",
                "name": "Xia Hu",
                "collaborators": [
                    "Zijian Zhang",
                    "Fan Yang",
                    "Haofan Wang",
                    "Weiqing Liu",
                    "Jiang Bian",
                    "Jian Pei"
                ],
                "domain": [
                    "Model Interpretation",
                    "Neural Network Complexity",
                    "Explainable AI"
                ],
                "publications": [
                    {
                        "title": "Contextual Local Explanation for Black Box Classifiers",
                        "abstract": "We introduce a new model-agnostic explanation technique which explains the prediction of any classifier called CLE. CLE gives an faithful and interpretable explanation to the prediction, by approximating the model locally using an interpretable model. We demonstrate the flexibility of CLE by explaining different models for text, tabular and image classification, and the fidelity of it by doing simulated user experiments."
                    },
                    {
                        "title": "Measuring Model Complexity of Neural Networks with Curve Activation Functions",
                        "abstract": "It is fundamental to measure model complexity of deep neural networks. The existing literature on model complexity mainly focuses on neural networks with piecewise linear activation functions. Model complexity of neural networks with general curve activation functions remains an open problem. To tackle the challenge, in this paper, we first propose the linear approximation neural network (LANN for short), a piecewise linear framework to approximate a given deep model with curve activation function. LANN constructs individual piecewise linear approximation for the activation function of each neuron, and minimizes the number of linear regions to satisfy a required approximation degree. Then, we analyze the upper bound of the number of linear regions formed by LANNs, and derive the complexity measure based on the upper bound. To examine the usefulness of the complexity measure, we experimentally explore the training process of neural networks and detect overfitting. Our results demonstrate that the occurrence of overfitting is positively correlated with the increase of model complexity during training. We find that the $L^1$ and $L^2$ regularizations suppress the increase of model complexity. Finally, we propose two approaches to prevent overfitting by directly constraining model complexity, namely neuron pruning and customized $L^1$ regularization."
                    }
                ]
            }
        }
    },
    "2406.16341": {
        "paper_data": {
            "title": "EHRCon: Dataset for Checking Consistency between Unstructured Notes and Structured Tables in Electronic Health Records",
            "url": "http://arxiv.org/abs/2406.16341v1",
            "arxiv_id": "2406.16341",
            "authors": [
                "Yeonsu Kwon",
                "Jiho Kim",
                "Gyubok Lee",
                "Seongsu Bae",
                "Daeun Kyung",
                "Wonchul Cha",
                "Tom Pollard",
                "Alistair Johnson",
                "Edward Choi"
            ],
            "abstract": "Electronic Health Records (EHRs) are integral for storing comprehensive patient medical records, combining structured data (e.g., medications) with detailed clinical notes (e.g., physician notes). These elements are essential for straightforward data retrieval and provide deep, contextual insights into patient care. However, they often suffer from discrepancies due to unintuitive EHR system designs and human errors, posing serious risks to patient safety. To address this, we developed EHRCon, a new dataset and task specifically designed to ensure data consistency between structured tables and unstructured notes in EHRs. EHRCon was crafted in collaboration with healthcare professionals using the MIMIC-III EHR dataset, and includes manual annotations of 3,943 entities across 105 clinical notes checked against database entries for consistency. EHRCon has two versions, one using the original MIMIC-III schema, and another using the OMOP CDM schema, in order to increase its applicability and generalizability. Furthermore, leveraging the capabilities of large language models, we introduce CheckEHR, a novel framework for verifying the consistency between clinical notes and database tables. CheckEHR utilizes an eight-stage process and shows promising results in both few-shot and zero-shot settings. The code is available at https://github.com/dustn1259/EHRCon.",
            "introduction": "   1 Introduction  Electronic Health Records (EHRs) are digital datasets comprising the rich information of a patient\u2019s medical history within hospitals. These records integrate both structured data (e.g., medications, diagnoses) and detailed clinical notes (e.g., physician notes). The structured data facilitates straightforward retrieval and analysis of essential information, while clinical notes provide in-depth, contextual insights into the patient\u2019s condition. These two forms of data are interconnected and provide complementary information throughout the diagnostic and treatment processes. For example, a practitioner might start by reviewing test results stored in the database, then determine a diagnosis and formulate a treatment plan, which are documented in the clinical notes. These notes are subsequently used to update the structured data in the database.   However, inconsistencies can arise between the two sets of data for several reasons. One primary issue is that EHR interfaces are often designed with a focus on administrative and financial tasks, which makes it difficult to accurately document clinical information\u00a0[32]. Additionally, overburdened practitioners might unintentionally introduce errors by importing incorrect medication lists, copying and pasting outdated records, or entering inaccurate test results\u00a0[4, 24, 38]. These errors can lead to significant discrepancies between the structured data and clinical notes in the EHR, potentially jeopardizing patient safety and leading to legal complications\u00a0[3].   Figure 1: Examples of consistent and inconsistent data between clinical notes and EHR tables: An inconsistent example (datetime) is when a clinical note records an HR (abbreviation for heart rate) of 167 on \u201c2022-02-02\u201d but the EHR table shows the same HR on \u201c2022-02-06\u201d. A consistent example is when both the clinical note and the EHR table document the administration of Atorvastatin with matching drug name, dosage, and unit. Another example of inconsistency occurs when a clinical note mentions a hypertension diagnosis, but the EHR table lacks this information.   Manual scrutiny of these records is both time-intensive and costly, underscoring the necessity for automated interventions. Despite the need for automated systems, previous studies on consistency check between tables and text have primarily focused on single claims and small-scale single tables\u00a0[1, 7, 8, 35]. These approaches are not designed for the complex and large-scale nature of EHRs, which require more comprehensive and scalable solutions.   To this end, we propose a new task and dataset called EHRCon, which is designed to verify the consistency between clinical notes and large-scale relational databases in EHRs. We collaborated closely with practitioners111EHR technician, nurse, and emergency medicine specialist with over 15 years of experience to design labeling instructions based on their insights and expertise, authentically reflecting real hospital environments. Based on these labeling instructions, trained human annotators used the MIMIC-III EHR dataset\u00a0[15] to manually compare 3,943 entities mentioned in 105 clinical notes against corresponding table contents, annotating them for Consistent or Inconsistent as illustrated in Figure 1. Our dataset also offers interpretability by including detailed information about the specific tables and columns where inconsistencies occurred. Moreover, it contains two versions, one based on the original MIMIC-III schema, and another based on its OMOP CDM\u00a0[34] implementation, allowing us to incorporate various schema types and enhance the generalizability.   Additionally, we introduce CheckEHR, a framework that leverages the reasoning capabilities of large language models (LLMs) to verify consistency between",
            "references": [
                {
                    "title": "Chain-of-Table: Evolving Tables in the Reasoning Chain for Table Understanding",
                    "abstract": "Table-based reasoning with large language models (LLMs) is a promising direction to tackle many table understanding tasks, such as table-based question answering and fact verification. Compared with generic reasoning, table-based reasoning requires the extraction of underlying semantics from both free-form questions and semi-structured tabular data. Chain-of-Thought and its similar approaches incorporate the reasoning chain in the form of textual context, but it is still an open question how to effectively leverage tabular data in the reasoning chain. We propose the Chain-of-Table framework, where tabular data is explicitly used in the reasoning chain as a proxy for intermediate thoughts. Specifically, we guide LLMs using in-context learning to iteratively generate operations and update the table to represent a tabular reasoning chain. LLMs can therefore dynamically plan the next operation based on the results of the previous ones. This continuous evolution of the table forms a chain, showing the reasoning process for a given tabular problem. The chain carries structured information of the intermediate results, enabling more accurate and reliable predictions. Chain-of-Table achieves new state-of-the-art performance on WikiTQ, FeTaQA, and TabFact benchmarks across multiple LLM choices."
                },
                {
                    "title": "Mixtral of Experts",
                    "abstract": "We introduce Mixtral 8x7B, a Sparse Mixture of Experts (SMoE) language model. Mixtral has the same architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks (i.e. experts). For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, each token has access to 47B parameters, but only uses 13B active parameters during inference. Mixtral was trained with a context size of 32k tokens and it outperforms or matches Llama 2 70B and GPT-3.5 across all evaluated benchmarks. In particular, Mixtral vastly outperforms Llama 2 70B on mathematics, code generation, and multilingual benchmarks. We also provide a model fine-tuned to follow instructions, Mixtral 8x7B - Instruct, that surpasses GPT-3.5 Turbo, Claude-2.1, Gemini Pro, and Llama 2 70B - chat model on human benchmarks. Both the base and instruct models are released under the Apache 2.0 license."
                },
                {
                    "title": "Camels in a Changing Climate: Enhancing LM Adaptation with Tulu 2",
                    "abstract": "Since the release of T\\\"ULU [Wang et al., 2023b], open resources for instruction tuning have developed quickly, from better base models to new finetuning techniques. We test and incorporate a number of these advances into T\\\"ULU, resulting in T\\\"ULU 2, a suite of improved T\\\"ULU models for advancing the understanding and best practices of adapting pretrained language models to downstream tasks and user preferences. Concretely, we release: (1) T\\\"ULU-V2-mix, an improved collection of high-quality instruction datasets; (2) T\\\"ULU 2, LLAMA-2 models finetuned on the V2 mixture; (3) T\\\"ULU 2+DPO, T\\\"ULU 2 models trained with direct preference optimization (DPO), including the largest DPO-trained model to date (T\\\"ULU 2+DPO 70B); (4) CODE T\\\"ULU 2, CODE LLAMA models finetuned on our V2 mix that outperform CODE LLAMA and its instruction-tuned variant, CODE LLAMA-Instruct. Our evaluation from multiple perspectives shows that the T\\\"ULU 2 suite achieves state-of-the-art performance among open models and matches or exceeds the performance of GPT-3.5-turbo-0301 on several benchmarks. We release all the checkpoints, data, training and evaluation code to facilitate future open efforts on adapting large language models."
                },
                {
                    "title": "KG-GPT: A General Framework for Reasoning on Knowledge Graphs Using Large Language Models",
                    "abstract": "While large language models (LLMs) have made considerable advancements in understanding and generating unstructured text, their application in structured data remains underexplored. Particularly, using LLMs for complex reasoning tasks on knowledge graphs (KGs) remains largely untouched. To address this, we propose KG-GPT, a multi-purpose framework leveraging LLMs for tasks employing KGs. KG-GPT comprises three steps: Sentence Segmentation, Graph Retrieval, and Inference, each aimed at partitioning sentences, retrieving relevant graph components, and deriving logical conclusions, respectively. We evaluate KG-GPT using KG-based fact verification and KGQA benchmarks, with the model showing competitive and robust performance, even outperforming several fully-supervised models. Our work, therefore, marks a significant step in unifying structured and unstructured data processing within the realm of LLMs."
                },
                {
                    "title": "Scientific Fact-Checking: A Survey of Resources and Approaches",
                    "abstract": "The task of fact-checking deals with assessing the veracity of factual claims based on credible evidence and background knowledge. In particular, scientific fact-checking is the variation of the task concerned with verifying claims rooted in scientific knowledge. This task has received significant attention due to the growing importance of scientific and health discussions on online platforms. Automated scientific fact-checking methods based on NLP can help combat the spread of misinformation, assist researchers in knowledge discovery, and help individuals understand new scientific breakthroughs. In this paper, we present a comprehensive survey of existing research in this emerging field and its related tasks. We provide a task description, discuss the construction process of existing datasets, and analyze proposed models and approaches. Based on our findings, we identify intriguing challenges and outline potential future directions to advance the field."
                },
                {
                    "title": "PaLM 2 Technical Report",
                    "abstract": "We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities and is more compute-efficient than its predecessor PaLM. PaLM 2 is a Transformer-based model trained using a mixture of objectives. Through extensive evaluations on English and multilingual language, and reasoning tasks, we demonstrate that PaLM 2 has significantly improved quality on downstream tasks across different model sizes, while simultaneously exhibiting faster and more efficient inference compared to PaLM. This improved efficiency enables broader deployment while also allowing the model to respond faster, for a more natural pace of interaction. PaLM 2 demonstrates robust reasoning capabilities exemplified by large improvements over PaLM on BIG-Bench and other reasoning tasks. PaLM 2 exhibits stable performance on a suite of responsible AI evaluations, and enables inference-time control over toxicity without additional overhead or impact on other capabilities. Overall, PaLM 2 achieves state-of-the-art performance across a diverse set of tasks and capabilities. When discussing the PaLM 2 family, it is important to distinguish between pre-trained models (of various sizes), fine-tuned variants of these models, and the user-facing products that use these models. In particular, user-facing products typically include additional pre- and post-processing steps. Additionally, the underlying models may evolve over time. Therefore, one should not expect the performance of user-facing products to exactly match the results reported in this report."
                },
                {
                    "title": "StructGPT: A General Framework for Large Language Model to Reason over Structured Data",
                    "abstract": "In this paper, we study how to improve the zero-shot reasoning ability of large language models~(LLMs) over structured data in a unified way. Inspired by the study on tool augmentation for LLMs, we develop an \\emph{Iterative Reading-then-Reasoning~(IRR)} approach for solving question answering tasks based on structured data, called \\textbf{StructGPT}. In our approach, we construct the specialized function to collect relevant evidence from structured data (\\ie \\emph{reading}), and let LLMs concentrate the reasoning task based on the collected information (\\ie \\emph{reasoning}). Specially, we propose an \\emph{invoking-linearization-generation} procedure to support LLMs in reasoning on the structured data with the help of the external interfaces. By iterating this procedures with provided interfaces, our approach can gradually approach the target answer to a given query. Extensive experiments conducted on three types of structured data demonstrate the effectiveness of our approach, which can significantly boost the performance of ChatGPT and achieve comparable performance against the full-data supervised-tuning baselines. Our codes and data are publicly available at~\\url{https://github.com/RUCAIBox/StructGPT}."
                },
                {
                    "title": "FactKG: Fact Verification via Reasoning on Knowledge Graphs",
                    "abstract": "In real world applications, knowledge graphs (KG) are widely used in various domains (e.g. medical applications and dialogue agents). However, for fact verification, KGs have not been adequately utilized as a knowledge source. KGs can be a valuable knowledge source in fact verification due to their reliability and broad applicability. A KG consists of nodes and edges which makes it clear how concepts are linked together, allowing machines to reason over chains of topics. However, there are many challenges in understanding how these machine-readable concepts map to information in text. To enable the community to better use KGs, we introduce a new dataset, FactKG: Fact Verificationvia Reasoning on Knowledge Graphs. It consists of 108k natural language claims with five types of reasoning: One-hop, Conjunction, Existence, Multi-hop, and Negation. Furthermore, FactKG contains various linguistic patterns, including colloquial style claims as well as written style claims to increase practicality. Lastly, we develop a baseline approach and analyze FactKG over these reasoning types. We believe FactKG can advance both reliability and practicality in KG-based fact verification."
                },
                {
                    "title": "FACTIFY-5WQA: 5W Aspect-based Fact Verification through Question Answering",
                    "abstract": "Automatic fact verification has received significant attention recently. Contemporary automatic fact-checking systems focus on estimating truthfulness using numerical scores which are not human-interpretable. A human fact-checker generally follows several logical steps to verify a verisimilitude claim and conclude whether it\u2019s truthful or a mere masquerade. Popular fact-checking websites follow a common structure for fact categorization such as half true, half false, false, pants on fire, etc. Therefore, it is necessary to have an aspect-based (delineating which part(s) are true and which are false) explainable system that can assist human fact-checkers in asking relevant questions related to a fact, which can then be validated separately to reach a final verdict. In this paper, we propose a 5W framework (who, what, when, where, and why) for question-answer-based fact explainability. To that end, we present a semi-automatically generated dataset called FACTIFY-5WQA, which consists of 391, 041 facts along with relevant 5W QAs \u2013 underscoring our major contribution to this paper. A semantic role labeling system has been utilized to locate 5Ws, which generates QA pairs for claims using a masked language model. Finally, we report a baseline QA system to automatically locate those answers from evidence documents, which can serve as a baseline for future research in the field. Lastly, we propose a robust fact verification system that takes paraphrased claims and automatically validates them. The dataset and the baseline model are available at https: //github.com/ankuranii/acl-5W-QA"
                },
                {
                    "title": "Chameleon: Plug-and-Play Compositional Reasoning with Large Language Models",
                    "abstract": "Large language models (LLMs) have achieved remarkable progress in solving various natural language processing tasks due to emergent reasoning abilities. However, LLMs have inherent limitations as they are incapable of accessing up-to-date information (stored on the Web or in task-specific knowledge bases), using external tools, and performing precise mathematical and logical reasoning. In this paper, we present Chameleon, an AI system that mitigates these limitations by augmenting LLMs with plug-and-play modules for compositional reasoning. Chameleon synthesizes programs by composing various tools (e.g., LLMs, off-the-shelf vision models, web search engines, Python functions, and heuristic-based modules) for accomplishing complex reasoning tasks. At the heart of Chameleon is an LLM-based planner that assembles a sequence of tools to execute to generate the final response. We showcase the effectiveness of Chameleon on two multi-modal knowledge-intensive reasoning tasks: ScienceQA and TabMWP. Chameleon, powered by GPT-4, achieves an 86.54% overall accuracy on ScienceQA, improving the best published few-shot result by 11.37%. On TabMWP, GPT-4-powered Chameleon improves the accuracy by 17.0%, lifting the state of the art to 98.78%. Our analysis also shows that the GPT-4-powered planner exhibits more consistent and rational tool selection via inferring potential constraints from instructions, compared to a ChatGPT-powered planner. The project is available at https://chameleon-llm.github.io."
                },
                {
                    "title": "Decomposed Prompting: A Modular Approach for Solving Complex Tasks",
                    "abstract": "Few-shot prompting is a surprisingly powerful way to use Large Language Models (LLMs) to solve various tasks. However, this approach struggles as the task complexity increases or when the individual reasoning steps of the task themselves are hard to learn, especially when embedded in more complex tasks. To address this, we propose Decomposed Prompting, a new approach to solve complex tasks by decomposing them (via prompting) into simpler sub-tasks that can be delegated to a library of prompting-based LLMs dedicated to these sub-tasks. This modular structure allows each prompt to be optimized for its specific sub-task, further decomposed if necessary, and even easily replaced with more effective prompts, trained models, or symbolic functions if desired. We show that the flexibility and modularity of Decomposed Prompting allows it to outperform prior work on few-shot prompting using GPT3. On symbolic reasoning tasks, we can further decompose sub-tasks that are hard for LLMs into even simpler solvable sub-tasks. When the complexity comes from the input length, we can recursively decompose the task into the same task but with smaller inputs. We also evaluate our approach on textual multi-step reasoning tasks: on long-context multi-hop QA task, we can more effectively teach the sub-tasks via our separate sub-tasks prompts; and on open-domain multi-hop QA, we can incorporate a symbolic information retrieval within our decomposition framework, leading to improved performance on both tasks. Datasets, Code and Prompts available at https://github.com/allenai/DecomP."
                },
                {
                    "title": "Least-to-Most Prompting Enables Complex Reasoning in Large Language Models",
                    "abstract": "Chain-of-thought prompting has demonstrated remarkable performance on various natural language reasoning tasks. However, it tends to perform poorly on tasks which requires solving problems harder than the exemplars shown in the prompts. To overcome this challenge of easy-to-hard generalization, we propose a novel prompting strategy, least-to-most prompting. The key idea in this strategy is to break down a complex problem into a series of simpler subproblems and then solve them in sequence. Solving each subproblem is facilitated by the answers to previously solved subproblems. Our experimental results on tasks related to symbolic manipulation, compositional generalization, and math reasoning reveal that least-to-most prompting is capable of generalizing to more difficult problems than those seen in the prompts. A notable finding is that when the GPT-3 code-davinci-002 model is used with least-to-most prompting, it can solve the compositional generalization benchmark SCAN in any split (including length split) with an accuracy of at least 99% using just 14 exemplars, compared to only 16% accuracy with chain-of-thought prompting. This is particularly noteworthy because neural-symbolic models in the literature that specialize in solving SCAN are trained on the entire training set containing over 15,000 examples. We have included prompts for all the tasks in the Appendix."
                },
                {
                    "title": "FaVIQ: FAct Verification from Information-seeking Questions",
                    "abstract": "Despite significant interest in developing general purpose fact checking models, it is challenging to construct a large-scale fact verification dataset with realistic real-world claims. Existing claims are either authored by crowdworkers, thereby introducing subtle biases thatare difficult to control for, or manually verified by professional fact checkers, causing them to be expensive and limited in scale. In this paper, we construct a large-scale challenging fact verification dataset called FAVIQ, consisting of 188k claims derived from an existing corpus of ambiguous information-seeking questions. The ambiguities in the questions enable automatically constructing true and false claims that reflect user confusions (e.g., the year of the movie being filmed vs. being released). Claims in FAVIQ are verified to be natural, contain little lexical bias, and require a complete understanding of the evidence for verification. Our experiments show that the state-of-the-art models are far from solving our new task. Moreover, training on our data helps in professional fact-checking, outperforming models trained on the widely used dataset FEVER or in-domain data by up to 17% absolute. Altogether, our data will serve as a challenging benchmark for natural language understanding and support future progress in professional fact checking."
                },
                {
                    "title": "FEVEROUS: Fact Extraction and VERification Over Unstructured and Structured information",
                    "abstract": "Fact verification has attracted a lot of attention in the machine learning and natural language processing communities, as it is one of the key methods for detecting misinformation. Existing large-scale benchmarks for this task have focused mostly on textual sources, i.e. unstructured information, and thus ignored the wealth of information available in structured formats, such as tables. In this paper we introduce a novel dataset and benchmark, Fact Extraction and VERification Over Unstructured and Structured information (FEVEROUS), which consists of 87,026 verified claims. Each claim is annotated with evidence in the form of sentences and/or cells from tables in Wikipedia, as well as a label indicating whether this evidence supports, refutes, or does not provide enough information to reach a verdict. Furthermore, we detail our efforts to track and minimize the biases present in the dataset and could be exploited by models, e.g. being able to predict the label without using evidence. Finally, we develop a baseline for verifying claims against text and tables which predicts both the correct evidence and verdict for 18% of the claims."
                },
                {
                    "title": "COVID-Fact: Fact Extraction and Verification of Real-World Claims on COVID-19 Pandemic",
                    "abstract": "We introduce a FEVER-like dataset COVID-Fact of 4,086 claims concerning the COVID-19 pandemic. The dataset contains claims, evidence for the claims, and contradictory claims refuted by the evidence. Unlike previous approaches, we automatically detect true claims and their source articles and then generate counter-claims using automatic methods rather than employing human annotators. Along with our constructed resource, we formally present the task of identifying relevant evidence for the claims and verifying whether the evidence refutes or supports a given claim. In addition to scientific claims, our data contains simplified general claims from media sources, making it better suited for detecting general misinformation regarding COVID-19. Our experiments indicate that COVID-Fact will provide a challenging testbed for the development of new systems and our approach will reduce the costs of building domain-specific datasets for detecting misinformation."
                },
                {
                    "title": "SemEval-2021 Task 9: Fact Verification and Evidence Finding for Tabular Data in Scientific Documents (SEM-TAB-FACTS)",
                    "abstract": "Understanding tables is an important and relevant task that involves understanding table structure as well as being able to compare and contrast information within cells. In this paper, we address this challenge by presenting a new dataset and tasks that addresses this goal in a shared task in SemEval 2020 Task 9: Fact Verification and Evidence Finding for Tabular Data in Scientific Documents (SEM-TAB-FACTS). Our dataset contains 981 manually-generated tables and an auto-generated dataset of 1980 tables providing over 180K statement and over 16M evidence annotations. SEM-TAB-FACTS featured two sub-tasks. In sub-task A, the goal was to determine if a statement is supported, refuted or unknown in relation to a table. In sub-task B, the focus was on identifying the specific cells of a table that provide evidence for the statement. 69 teams signed up to participate in the task with 19 successful submissions to subtask A and 12 successful submissions to subtask B. We present our results and main findings from the competition."
                },
                {
                    "title": "Get Your Vitamin C! Robust Fact Verification with Contrastive Evidence",
                    "abstract": "Typical fact verification models use retrieved written evidence to verify claims. Evidence sources, however, often change over time as more information is gathered and revised. In order to adapt, models must be sensitive to subtle differences in supporting evidence. We present VitaminC, a benchmark infused with challenging cases that require fact verification models to discern and adjust to slight factual changes. We collect over 100,000 Wikipedia revisions that modify an underlying fact, and leverage these revisions, together with additional synthetically constructed ones, to create a total of over 400,000 claim-evidence pairs. Unlike previous resources, the examples in VitaminC are contrastive, i.e., they contain evidence pairs that are nearly identical in language and content, with the exception that one supports a given claim while the other does not. We show that training using this design increases robustness\u2014improving accuracy by 10% on adversarial fact verification and 6% on adversarial natural language inference (NLI). Moreover, the structure of VitaminC leads us to define additional tasks for fact-checking resources: tagging relevant words in the evidence for verifying the claim, identifying factual revisions, and providing automatic edits via factually consistent text generation."
                },
                {
                    "title": "HoVer: A Dataset for Many-Hop Fact Extraction And Claim Verification",
                    "abstract": "We introduce HoVer (HOppy VERification), a dataset for many-hop evidence extraction and fact verification. It challenges models to extract facts from several Wikipedia articles that are relevant to a claim and classify whether the claim is supported or not-supported by the facts. In HoVer, the claims require evidence to be extracted from as many as four English Wikipedia articles and embody reasoning graphs of diverse shapes. Moreover, most of the 3/4-hop claims are written in multiple sentences, which adds to the complexity of understanding long-range dependency relations such as coreference. We show that the performance of an existing state-of-the-art semantic-matching model degrades significantly on our dataset as the number of reasoning hops increases, hence demonstrating the necessity of many-hop reasoning to achieve strong results. We hope that the introduction of this challenging dataset and the accompanying evaluation task will encourage research in many-hop fact retrieval and information verification."
                },
                {
                    "title": "Frequency and Types of Patient-Reported Errors in Electronic Health Record Ambulatory Care Notes",
                    "abstract": "This survey study assesses the frequency and types of errors identified by patients who read open ambulatory visit notes."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "INFOTABS: Inference on Tables as Semi-structured Data",
                    "abstract": "In this paper, we observe that semi-structured tabulated text is ubiquitous; understanding them requires not only comprehending the meaning of text fragments, but also implicit relationships between them. We argue that such data can prove as a testing ground for understanding how we reason about information. To study this, we introduce a new dataset called INFOTABS, comprising of human-written textual hypotheses based on premises that are tables extracted from Wikipedia info-boxes. Our analysis shows that the semi-structured, multi-domain and heterogeneous nature of the premises admits complex, multi-faceted reasoning. Experiments reveal that, while human annotators agree on the relationships between a table-hypothesis pair, several standard modeling strategies are unsuccessful at the task, suggesting that reasoning about tables can pose a difficult modeling challenge."
                },
                {
                    "title": "TabFact: A Large-scale Dataset for Table-based Fact Verification",
                    "abstract": "The problem of verifying whether a textual hypothesis holds based on the given evidence, also known as fact verification, plays an important role in the study of natural language understanding and semantic representation. However, existing studies are mainly restricted to dealing with unstructured evidence (e.g., natural language sentences and documents, news, etc), while verification under structured evidence, such as tables, graphs, and databases, remains under-explored. This paper specifically aims to study the fact verification given semi-structured data as evidence. To this end, we construct a large-scale dataset called TabFact with 16k Wikipedia tables as the evidence for 118k human-annotated natural language statements, which are labeled as either ENTAILED or REFUTED. TabFact is challenging since it involves both soft linguistic reasoning and hard symbolic reasoning. To address these reasoning challenges, we design two different models: Table-BERT and Latent Program Algorithm (LPA). Table-BERT leverages the state-of-the-art pre-trained language model to encode the linearized tables and statements into continuous vectors for verification. LPA parses statements into programs and executes them against the tables to obtain the returned binary value for verification. Both methods achieve similar accuracy but still lag far behind human performance. We also perform a comprehensive analysis to demonstrate great future opportunities. The data and code of the dataset are provided in \\url{this https URL}."
                },
                {
                    "title": "Towards Debiasing Fact Verification Models",
                    "abstract": "Fact verification requires validating a claim in the context of evidence. We show, however, that in the popular FEVER dataset this might not necessarily be the case. Claim-only classifiers perform competitively with top evidence-aware models. In this paper, we investigate the cause of this phenomenon, identifying strong cues for predicting labels solely based on the claim, without considering any evidence. We create an evaluation set that avoids those idiosyncrasies. The performance of FEVER-trained models significantly drops when evaluated on this test set. Therefore, we introduce a regularization method which alleviates the effect of bias in the training data, obtaining improvements on the newly created test set. This work is a step towards a more sound evaluation of reasoning capabilities in fact verification models."
                },
                {
                    "title": "Using voice to create inpatient progress notes: effects on note timeliness, quality, and physician satisfaction",
                    "abstract": "Abstract Objectives We describe the evaluation of a system to create hospital progress notes using voice and electronic health record integration to determine if note timeliness, quality, and physician satisfaction are improved. Materials and methods We conducted a randomized controlled trial to measure effects of this new method of writing inpatient progress notes, which evolved over time, on important outcomes. Results Intervention subjects created 709 notes and control subjects created 1143 notes. When adjusting for clustering by provider and secular trends, there was no significant difference between the intervention and control groups in the time between when patients were seen on rounds and when progress notes were viewable by others (95% confidence interval \u2212106.9 to 12.2 min). There were no significant differences in physician satisfaction or note quality between intervention and control. Discussion Though we did not find support for the superiority of this system (Voice-Generated Enhanced Electronic Note System [VGEENS]) for our 3 primary outcomes, if notes are created using voice during or soon after rounds they are available within 10 min. Shortcomings that likely influenced subject satisfaction include the early state of our VGEENS and the short interval for system development before the randomized trial began. Conclusion VGEENS permits voice dictation on rounds to create progress notes and can reduce delay in note availability and may reduce dependence on copy/paste within notes. Timing of dictation determines when notes are available. Capturing notes in near-real-time has potential to apply NLP and decision support sooner than when notes are typed later in the day, and to improve note accuracy."
                },
                {
                    "title": "FEVER: a Large-scale Dataset for Fact Extraction and VERification",
                    "abstract": "In this paper we introduce a new publicly available dataset for verification against textual sources, FEVER: Fact Extraction and VERification. It consists of 185,445 claims generated by altering sentences extracted from Wikipedia and subsequently verified without knowledge of the sentence they were derived from. The claims are classified as Supported, Refuted or NotEnoughInfo by annotators achieving 0.6841 in Fleiss kappa. For the first two classes, the annotators also recorded the sentence(s) forming the necessary evidence for their judgment. To characterize the challenge of the dataset presented, we develop a pipeline approach and compare it to suitably designed oracles. The best accuracy we achieve on labeling a claim accompanied by the correct evidence is 31.87%, while if we ignore the evidence we achieve 50.91%. Thus we believe that FEVER is a challenging testbed that will help stimulate progress on claim verification against textual sources."
                },
                {
                    "title": "Transitions of Care: Improving the Quality of Discharge Summaries Completed By Internal Medicine Residents",
                    "abstract": "Introduction Discharge summaries are now the accepted means of communication in transition from inpatient to ambulatory care. However, there is often no formal residency education on this critical document, leading to discordance in discharge summaries written by internal medicine residents. There is little in the literature focusing on teaching how to effectively create a discharge summary using an electronic health record (EHR). Methods A 1-hour workshop was designed to teach components of the discharge summary and how to utilize this document to safely transition patients from the inpatient to the ambulatory setting. One or two faculty facilitators led the workshop with approximately 20 resident learners. A 50-point rubric was created to assess effectiveness of discharge summaries pre- and postworkshop. Results The workshop was well received by residents and median scores on the rubric improved from 39 to 45 (p < .001) postworkshop. Discussion We found that by teaching the concepts using examples of discharge summaries written by our residents, and then creating a standardized EHR template, residents wrote more effective discharge summaries with increased focus on the transition to the ambulatory provider. These materials can be applied to other programs and levels of learners to improve discharge summary quality. This serves to provide a resource to those at other institutions looking to create a more formalized didactic session on discharge summaries with a particular focus on transitioning care to the ambulatory provider."
                },
                {
                    "title": "Electronic Health Records Uses and Malpractice Risks",
                    "abstract": "INTRODUCTION Our world has been radically transformed by digital technology \u2013 smart phones, tablets, and web-enabled devices have transformed our daily lives and the way we communicate. A greater and more seamless flow of information within a digital healthcare infrastructure, created by electronic health records (EHRs), encompasses and leverages digital progress and can transform the way care is delivered and compensated. An EHR is a digital version of a patient's paper chart. EHRs are real-time, patient-centered records that make information available instantly and securely to authorized users. While an EHR does contain the medical and treatment histories of patients, an EHR system is built to go beyond standard clinical data collected in a provider's office and can be inclusive of a broader view of a patient's care.1 One of the key features of an EHR is that health information can be created and managed by authorized providers in a digital format capable of being shared with other providers across more than one healthcare organization. EHRs are built to share information with other healthcare providers and organizations \u2013 such as laboratories \u2013 so they contain information from all clinicians involved in a patient's care. The EHR is composed of components such as clinical decision support systems (CDSS) that include information about relevant clinical practice guidelines and clinical reminders, and computerized physician order entry (CPOE) of medications that check orders against patient information to flag potential drug interactions, allergic reactions, and errors.2 The Health Information Technology for Economic and Clinical Health (HITECH)\u2026 ABBREVIATIONS: HER - Electronic Health Records, ARRA - American Recovery and Reinvestment Act, HITECH - Health Information Technology for Economic and Clinical Health, CDSS - Clinical Decision Support System, CPOE - Computerized Physician Order Entry, IOM - Institute of Medicine's, HIT - Health Information Technology, CRICO - Controlled Risk Insurance Company"
                },
                {
                    "title": "Effects of Common Data Errors in Electronic Health Records on Emergency Department Operational Performance Metrics: A Monte Carlo Simulation.",
                    "abstract": "OBJECTIVES\nThe objective was to estimate how data errors in electronic health records (EHRs) can affect the accuracy of common emergency department (ED) operational performance metrics.\n\n\nMETHODS\nUsing a 3-month, 7,348-visit data set of electronic time stamps from a suburban academic ED as a baseline, Monte Carlo simulation was used to introduce four types of data errors (substitution, missing, random, and systematic bias) at three frequency levels (2, 4, and 7%). Three commonly used ED operational metrics (arrival to clinician evaluation, disposition decision to exit for admitted patients, and ED length of stay for admitted patients) were calculated and the proportion of ED visits that achieved each performance goal was determined.\n\n\nRESULTS\nEven small data errors have measurable effects on a clinical organization's ability to accurately determine whether it is meeting its operational performance goals. Systematic substitution errors, increased frequency of errors, and the use of shorter-duration metrics resulted in a lower proportion of ED visits reported as meeting the associated performance objectives. However, the presence of other error types mitigated somewhat the effect of the systematic substitution error. Longer time-duration metrics were found to be less sensitive to data errors than shorter time-duration metrics.\n\n\nCONCLUSIONS\nInfrequent and small-magnitude data errors in EHR time stamps can compromise a clinical organization's ability to determine accurately if it is meeting performance goals. By understanding the types and frequencies of data errors in an organization's EHR, organizational leaders can use data management best practices to better measure true performance and enhance operational decision-making."
                },
                {
                    "title": "Feasibility and utility of applications of the common data model to multiple, disparate observational health databases",
                    "abstract": "Objectives To evaluate the utility of applying the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) across multiple observational databases within an organization and to apply standardized analytics tools for conducting observational research. Materials and methods Six deidentified patient-level datasets were transformed to the OMOP CDM. We evaluated the extent of information loss that occurred through the standardization process. We developed a standardized analytic tool to replicate the cohort construction process from a published epidemiology protocol and applied the analysis to all 6 databases to assess time-to-execution and comparability of results. Results Transformation to the CDM resulted in minimal information loss across all 6 databases. Patients and observations excluded were due to identified data quality issues in the source system, 96% to 99% of condition records and 90% to 99% of drug records were successfully mapped into the CDM using the standard vocabulary. The full cohort replication and descriptive baseline summary was executed for 2 cohorts in 6 databases in less than 1 hour. Discussion The standardization process improved data quality, increased efficiency, and facilitated cross-database comparisons to support a more systematic approach to observational research. Comparisons across data sources showed consistency in the impact of inclusion criteria, using the protocol and identified differences in patient characteristics and coding practices across databases. Conclusion Standardizing data structure (through a CDM), content (through a standard vocabulary with source code mappings), and analytics can enable an institution to apply a network-based approach to observational research across multiple, disparate observational health databases."
                },
                {
                    "title": "A review on usability features for designing electronic health records",
                    "abstract": "Health professionals should record clinical care information in order to provide adequate patient assistance. Such recording process requires an Electronic Health Record (EHR) with complete and integral information adjusted to treatments performed on a patient. Moreover, availability and access to EHR are key features supporting decision making and improve patient care. Nevertheless, designing an EHR fulfilling a set of quality attributes is not an easy task. In fact, an EHR with low usability causes deficient clinical services and poor data quality on medical records. In this paper, an interpretation of the ISO 9241-210 standard, in the context of patients' care in an emergency situation, is presented. It considers three mid-level objectives: understanding, designing and evaluating. In this way, an alignment between high-level guidelines, offered by usability standards, and low-level activities to be followed during an EHR design, is achieved."
                },
                {
                    "title": "A Review of Medical Errors in Laboratory Diagnostics and Where We Are Today",
                    "abstract": "While many areas of health care are still struggling with the issue of patient safety, laboratory diagnostics has always been a forerunner in pursuing this issue. Significant progress has been made since the release of \u201cTo Err is Human.\u201d 1 This article briefly reviews laboratory quality assessment and looks at recent statistics concerning laboratory errors."
                },
                {
                    "title": "An unintended consequence of electronic prescriptions: prevalence and impact of internal discrepancies",
                    "abstract": "Many e-prescribing systems allow for both structured and free-text fields in prescriptions, making possible internal discrepancies. This study reviewed 2914 electronic prescriptions that contained free-text fields. Internal discrepancies were found in 16.1% of the prescriptions. Most (83.8%) of the discrepancies could potentially lead to adverse events and many (16.8%) to severe adverse events, involving a hospital admission or death. Discrepancies in doses, routes or complex regimens were most likely to have a potential for a severe event (p=0.0001). Discrepancies between structured and free-text fields in electronic prescriptions are common and can cause patient harm. Improvements in electronic medical record design are necessary to minimize the risk of discrepancies and resulting adverse events."
                },
                {
                    "title": "Evidence-based Fact-Checking of Health-related Claims",
                    "abstract": "The task of verifying the truthfulness of claims in textual documents, or fact-checking, has received signi\ufb01cant attention in recent years. Many existing evidence-based fact-checking datasets contain synthetic claims and the models trained on these data might not be able to verify real-world claims. Partic-ularly few studies addressed evidence-based fact-checking of health-related claims that re-quire medical expertise or evidence from the scienti\ufb01c literature. In this paper, we introduce H EALTH V ER , a new dataset for evidence-based fact-checking of health-related claims that allows to study the validity of real-world claims by evaluating their truthfulness against scienti\ufb01c articles. Using a three-step data creation method, we \ufb01rst retrieved real-world claims from snippets returned by a search engine for questions about COVID-19. Then we automatically retrieved and re-ranked relevant scienti\ufb01c papers using a T5 relevance-based model. Finally, the relations between each evidence statement and the associated claim were manually annotated as S UPPORT , R EFUTE and N EUTRAL . To validate the created dataset of 14,330 evidence-claim pairs, we developed baseline models based on pretrained language models. Our experiments showed that training deep learning models on real-world medical claims greatly improves performance compared to models trained on synthetic and open-domain claims. Our results and manual analysis suggest that H EALTH V ER provides a realistic and challenging dataset for future efforts on evidence-based fact-checking of health-related claims. The dataset, source code, and"
                },
                {
                    "title": "Comparison of accuracy of physical examination findings in initial progress notes between paper charts and a newly implemented electronic health record",
                    "abstract": "Introduction: There have been several concerns about the quality of documentation in electronic health records (EHRs) when compared to paper charts. This study compares the accuracy of physical examination findings documentation between the two in initial progress notes. Methodology: Initial progress notes from patients with 5 specific diagnoses with invariable physical findings admitted to Beaumont Hospital, Royal Oak, between August 2011 and July 2013 were randomly selected for this study. A total of 500 progress notes were retrospectively reviewed. The paper chart arm consisted of progress notes completed prior to the transition to an EHR on July 1, 2012. The remaining charts were placed in the EHR arm. The primary endpoints were accuracy, inaccuracy, and omission of information. Secondary endpoints were time of initiation of progress note, word count, number of systems documented, and accuracy based on level of training. Results: The rate of inaccurate documentation was significantly higher in the EHRs compared to the paper charts (24.4% vs 4.4%). However, expected physical examination findings were more likely to be omitted in the paper notes compared to EHRs (41.2% vs 17.6%). Resident physicians had a smaller number of inaccuracies (5.3% vs 17.3%) and omissions (16.8% vs 33.9%) compared to attending physicians. Conclusions: During the initial phase of implementation of an EHR, inaccuracies were more common in progress notes in the EHR compared to the paper charts. Residents had a lower rate of inaccuracies and omissions compared to attending physicians. Further research is needed to identify training methods and incentives that can reduce inaccuracies in EHRs during initial implementation."
                }
            ],
            "categories": [
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively verify the consistency between clinical notes and structured data in Electronic Health Records (EHRs) to ensure patient safety and improve data integrity?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant gap in the reliability of EHRs, which are foundational for patient care and clinical decision-making. By ensuring consistency between clinical notes and structured data, we can enhance the accuracy of patient records, reduce the risk of medical errors, and improve overall healthcare outcomes. This research could lead to the development of automated systems that streamline the verification process, thereby saving time and resources for healthcare practitioners. Furthermore, it could pave the way for future studies on data integrity in EHRs, influencing how healthcare data is managed and utilized in clinical settings.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity and scale of EHRs, which contain vast amounts of unstructured and structured data. Naive approaches may fail due to the intricate relationships between clinical notes and structured data, as well as the variability in how information is documented. Technical obstacles include the need for advanced natural language processing techniques to interpret clinical notes accurately and the difficulty in establishing a comprehensive framework that can handle diverse EHR schemas. Additionally, practical challenges arise from the need for collaboration with healthcare professionals to ensure that the verification process aligns with real-world clinical practices.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on isolated claims or small-scale tables, which do not capture the complexity of large-scale EHR systems. Limitations in existing solutions include a lack of comprehensive datasets that reflect real hospital environments and insufficient methodologies to address the intricacies of data consistency across different formats. Barriers such as the absence of collaboration with healthcare practitioners and the challenges of scaling solutions to accommodate diverse EHR schemas have also hindered progress. Our approach differs by introducing the EHRCon dataset, which is specifically designed for this task, and by leveraging the reasoning capabilities of large language models through the CheckEHR framework.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the creation of the EHRCon dataset, which includes 3,943 entities annotated for consistency based on clinical notes and corresponding EHR table contents. We will utilize the M"
            }
        },
        "author_data": {
            "3c523902-a13e-445e-91ca-113b909e9dcc": {
                "pk": "3c523902-a13e-445e-91ca-113b909e9dcc",
                "name": "Yeonsu Kwon",
                "collaborators": [
                    "Edward Choi",
                    "Yohan Jo",
                    "Jiho Kim",
                    "Hyunseung Chung",
                    "Sumin Jo",
                    "Sunjun Kweon",
                    "Seonhee Cho",
                    "Sungjin Park",
                    "James Thorne",
                    "Gyubok Lee"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Knowledge Graph",
                    "Explainable AI",
                    "Text-to-SQL"
                ],
                "publications": [
                    {
                        "title": "KG-GPT: A General Framework for Reasoning on Knowledge Graphs Using Large Language Models",
                        "abstract": "While large language models (LLMs) have made considerable advancements in understanding and generating unstructured text, their application in structured data remains underexplored. Particularly, using LLMs for complex reasoning tasks on knowledge graphs (KGs) remains largely untouched. To address this, we propose KG-GPT, a multi-purpose framework leveraging LLMs for tasks employing KGs. KG-GPT comprises three steps: Sentence Segmentation, Graph Retrieval, and Inference, each aimed at partitioning sentences, retrieving relevant graph components, and deriving logical conclusions, respectively. We evaluate KG-GPT using KG-based fact verification and KGQA benchmarks, with the model showing competitive and robust performance, even outperforming several fully-supervised models. Our work, therefore, marks a significant step in unifying structured and unstructured data processing within the realm of LLMs."
                    },
                    {
                        "title": "Time is Not Enough: Time-Frequency based Explanation for Time-Series Black-Box Models",
                        "abstract": "Despite the massive attention given to time-series explanations due to their extensive applications, a notable limitation in existing approaches is their primary reliance on the time-domain. This overlooks the inherent characteristic of time-series data containing both time and frequency features. In this work, we present Spectral eXplanation (SpectralX), an XAI framework that provides time-frequency explanations for time-series black-box classifiers. This easily adaptable framework enables users to \"plug-in\" various perturbation-based XAI methods for any pre-trained time-series classification models to assess their impact on the explanation quality without having to modify the framework architecture. Additionally, we introduce Feature Importance Approximations (FIA), a new perturbation-based XAI method. These methods consist of feature insertion, deletion, and combination techniques to enhance computational efficiency and class-specific explanations in time-series classification tasks. We conduct extensive experiments in the generated synthetic dataset and various UCR Time-Series datasets to first compare the explanation performance of FIA and other existing perturbation-based XAI methods in both time-domain and time-frequency domain, and then show the superiority of our FIA in the time-frequency domain with the SpectralX framework. Finally, we conduct a user study to confirm the practicality of our FIA in SpectralX framework for class-specific time-frequency based time-series explanations. The source code is available in https://github.com/gustmd0121/Time_is_not_Enough"
                    },
                    {
                        "title": "Open-WikiTable: Dataset for Open Domain Question Answering with Complex Reasoning over Table",
                        "abstract": "Despite recent interest in open domain question answering (ODQA) over tables, many studies still rely on datasets that are not truly optimal for the task with respect to utilizing structural nature of table. These datasets assume answers reside as a single cell value and do not necessitate exploring over multiple cells such as aggregation, comparison, and sorting. Thus, we release Open-WikiTable, the first ODQA dataset that requires complex reasoning over tables. Open-WikiTable is built upon WikiSQL and WikiTableQuestions to be applicable in the open-domain setting. As each question is coupled with both textual answers and SQL queries, Open-WikiTable opens up a wide range of possibilities for future research, as both reader and parser methods can be applied. The dataset and code are publicly available."
                    },
                    {
                        "title": "FactKG: Fact Verification via Reasoning on Knowledge Graphs",
                        "abstract": "In real world applications, knowledge graphs (KG) are widely used in various domains (e.g. medical applications and dialogue agents). However, for fact verification, KGs have not been adequately utilized as a knowledge source. KGs can be a valuable knowledge source in fact verification due to their reliability and broad applicability. A KG consists of nodes and edges which makes it clear how concepts are linked together, allowing machines to reason over chains of topics. However, there are many challenges in understanding how these machine-readable concepts map to information in text. To enable the community to better use KGs, we introduce a new dataset, FactKG: Fact Verification via Reasoning on Knowledge Graphs. It consists of 108k natural language claims with five types of reasoning: One-hop, Conjunction, Existence, Multi-hop, and Negation. Furthermore, FactKG contains various linguistic patterns, including colloquial style claims as well as written style claims to increase practicality. Lastly, we develop a baseline approach and analyze FactKG over these reasoning types. We believe FactKG can advance both reliability and practicality in KG-based fact verification."
                    },
                    {
                        "title": "EHRSQL: A Practical Text-to-SQL Benchmark for Electronic Health Records",
                        "abstract": "We present a new text-to-SQL dataset for electronic health records (EHRs). The utterances were collected from 222 hospital staff members, including physicians, nurses, and insurance review and health records teams. To construct the QA dataset on structured EHR data, we conducted a poll at a university hospital and used the responses to create seed questions. We then manually linked these questions to two open-source EHR databases, MIMIC-III and eICU, and included various time expressions and held-out unanswerable questions in the dataset, which were also collected from the poll. Our dataset poses a unique set of challenges: the model needs to 1) generate SQL queries that reflect a wide range of needs in the hospital, including simple retrieval and complex operations such as calculating survival rate, 2) understand various time expressions to answer time-sensitive questions in healthcare, and 3) distinguish whether a given question is answerable or unanswerable. We believe our dataset, EHRSQL, can serve as a practical benchmark for developing and assessing QA models on structured EHR data and take a step further towards bridging the gap between text-to-SQL research and its real-life deployment in healthcare. EHRSQL is available at https://github.com/glee4810/EHRSQL."
                    }
                ]
            },
            "e6361473-cf17-4886-9144-378c442bcbaa": {
                "pk": "e6361473-cf17-4886-9144-378c442bcbaa",
                "name": "Jiho Kim",
                "collaborators": [
                    "Edward Choi",
                    "Seongsu Bae",
                    "Daeyoung Kim",
                    "Yeonsu Kwon",
                    "Yohan Jo",
                    "Sungjin Park",
                    "Tackeun Kim"
                ],
                "domain": [
                    "Health Informatics",
                    "Natural Language Processing",
                    "Knowledge Graph",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "Coinductive properties of Lipschitz functions on streams",
                        "abstract": "A simple hierarchical structure is imposed on the set of Lipschitz functions on streams (i.e. sequences over a fixed alphabet set) under the standard metric. We prove that sets of non-expanding and contractive functions are closed under a certain coiterative construction. The closure property is used to construct new final stream coalgebras over finite alphabets. For an example, we show that the 2-adic extension of the Collatz function and certain variants yield final bitstream coalgebras."
                    },
                    {
                        "title": "Question Answering for Complex Electronic Health Records Database using Unified Encoder-Decoder Architecture",
                        "abstract": "An intelligent machine that can answer human questions based on electronic health records (EHR-QA) has a great practical value, such as supporting clinical decisions, managing hospital administration, and medical chatbots. Previous table-based QA studies focusing on translating natural questions into table queries (NLQ2SQL), however, suffer from the unique nature of EHR data due to complex and specialized medical terminology, hence increased decoding difficulty. In this paper, we design UniQA, a unified encoder-decoder architecture for EHR-QA where natural language questions are converted to queries such as SQL or SPARQL. We also propose input masking (IM), a simple and effective method to cope with complex medical terms and various typos and better learn the SQL/SPARQL syntax. Combining the unified architecture with an effective auxiliary training objective, UniQA demonstrated a significant performance improvement against the previous state-of-the-art model for MIMICSQL* (14.2% gain), the most complex NLQ2SQL dataset in the EHR domain, and its typo-ridden versions (approximately 28.8% gain). In addition, we confirmed consistent results for the graph-based EHR-QA dataset, MIMICSPARQL*."
                    },
                    {
                        "title": "KG-GPT: A General Framework for Reasoning on Knowledge Graphs Using Large Language Models",
                        "abstract": "While large language models (LLMs) have made considerable advancements in understanding and generating unstructured text, their application in structured data remains underexplored. Particularly, using LLMs for complex reasoning tasks on knowledge graphs (KGs) remains largely untouched. To address this, we propose KG-GPT, a multi-purpose framework leveraging LLMs for tasks employing KGs. KG-GPT comprises three steps: Sentence Segmentation, Graph Retrieval, and Inference, each aimed at partitioning sentences, retrieving relevant graph components, and deriving logical conclusions, respectively. We evaluate KG-GPT using KG-based fact verification and KGQA benchmarks, with the model showing competitive and robust performance, even outperforming several fully-supervised models. Our work, therefore, marks a significant step in unifying structured and unstructured data processing within the realm of LLMs."
                    },
                    {
                        "title": "Graph-Text Multi-Modal Pre-training for Medical Representation Learning",
                        "abstract": "As the volume of Electronic Health Records (EHR) sharply grows, there has been emerging interest in learning the representation of EHR for healthcare applications. Representation learning of EHR requires appropriate modeling of the two dominant modalities in EHR: structured data and unstructured text. In this paper, we present MedGTX, a pre-trained model for multi-modal representation learning of the structured and textual EHR data. MedGTX uses a novel graph encoder to exploit the graphical nature of structured EHR data, and a text encoder to handle unstructured text, and a cross-modal encoder to learn a joint representation space. We pre-train our model through four proxy tasks on MIMIC-III, an open-source EHR data, and evaluate our model on two clinical benchmarks and three novel downstream tasks which tackle real-world problems in EHR data. The results consistently show the effectiveness of pre-training the model for joint representation of both structured and unstructured information from EHR. Given the promising performance of MedGTX, we believe this work opens a new door to jointly understanding the two fundamental modalities of EHR data."
                    }
                ]
            },
            "55a65450-84f0-4b16-a4fc-bb777adfb65e": {
                "pk": "55a65450-84f0-4b16-a4fc-bb777adfb65e",
                "name": "Gyubok Lee",
                "collaborators": [
                    "Edward Choi",
                    "Seongsu Bae",
                    "Sunjun Kweon",
                    "Jungwoo Oh",
                    "Seongjun Yang",
                    "Seonhee Cho",
                    "Jaehee Ryu",
                    "Woncheol Shin",
                    "Eunbyeol Cho",
                    "Woosog Chay"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Text-to-SQL",
                    "Healthcare AI",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Improving Lexically Constrained Neural Machine Translation with Source-Conditioned Masked Span Prediction",
                        "abstract": "Accurate terminology translation is crucial for ensuring the practicality and reliability of neural machine translation (NMT) systems. To address this, lexically constrained NMT explores various methods to ensure pre-specified words and phrases appear in the translation output. However, in many cases, those methods are studied on general domain corpora, where the terms are mostly uni- and bi-grams (>98%). In this paper, we instead tackle a more challenging setup consisting of domain-specific corpora with much longer n-gram and highly specialized terms. Inspired by the recent success of masked span prediction models, we propose a simple and effective training strategy that achieves consistent improvements on both terminology and sentence-level translation for three domain-specific corpora in two language pairs."
                    },
                    {
                        "title": "Overview of the EHRSQL 2024 Shared Task on Reliable Text-to-SQL Modeling on Electronic Health Records",
                        "abstract": "Electronic Health Records (EHRs) are relational databases that store the entire medical histories of patients within hospitals. They record numerous aspects of patients' medical care, from hospital admission and diagnosis to treatment and discharge. While EHRs are vital sources of clinical data, exploring them beyond a predefined set of queries requires skills in query languages like SQL. To make information retrieval more accessible, one strategy is to build a question-answering system, possibly leveraging text-to-SQL models that can automatically translate natural language questions into corresponding SQL queries and use these queries to retrieve the answers. The EHRSQL 2024 shared task aims to advance and promote research in developing a question-answering system for EHRs using text-to-SQL modeling, capable of reliably providing requested answers to various healthcare professionals to improve their clinical work processes and satisfy their needs. Among more than 100 participants who applied to the shared task, eight teams were formed and completed the entire shared task requirement and demonstrated a wide range of methods to effectively solve this task. In this paper, we describe the task of reliable text-to-SQL modeling, the dataset, and the methods and results of the participants. We hope this shared task will spur further research and insights into developing reliable question-answering systems for EHRs."
                    },
                    {
                        "title": "TrustSQL: Benchmarking Text-to-SQL Reliability with Penalty-Based Scoring",
                        "abstract": "Text-to-SQL enables users to interact with databases using natural language, simplifying the retrieval and synthesis of information. Despite the remarkable success of large language models (LLMs) in translating natural language questions into SQL queries, widespread deployment remains limited due to two primary challenges. First, the effective use of text-to-SQL models depends on users' understanding of the model's capabilities-the scope of questions the model can correctly answer. Second, the absence of abstention mechanisms can lead to incorrect SQL generation going unnoticed, thereby undermining trust in the model's output. To enable wider deployment, it is crucial to address these challenges in model design and enhance model evaluation to build trust in the model's output. To this end, we introduce TrustSQL, a novel comprehensive benchmark designed to evaluate text-to-SQL reliability-defined as a model's ability to correctly handle any type of input question by generating correct SQL queries for feasible questions and abstaining from generating infeasible ones (e.g., due to schema incompatibility or functionalities beyond SQL). We evaluate existing methods using a novel penalty-based scoring metric with two modeling approaches: (1) pipeline-based methods combining SQL generators with infeasible question detectors and SQL error detectors for abstention; and (2) unified methods using a single model for the entire task. Our experimental results reveal that achieving high scores under severe penalties requires significant effort and provide a new perspective on developing text-to-SQL models for safer deployment. TrustSQL is available at https://github.com/glee4810/TrustSQL."
                    },
                    {
                        "title": "EHR-SeqSQL : A Sequential Text-to-SQL Dataset For Interactively Exploring Electronic Health Records",
                        "abstract": "In this paper, we introduce EHR-SeqSQL, a novel sequential text-to-SQL dataset for Electronic Health Record (EHR) databases. EHR-SeqSQL is designed to address critical yet underexplored aspects in text-to-SQL parsing: interactivity, compositionality, and efficiency. To the best of our knowledge, EHR-SeqSQL is not only the largest but also the first medical text-to-SQL dataset benchmark to include sequential and contextual questions. We provide a data split and the new test set designed to assess compositional generalization ability. Our experiments demonstrate the superiority of a multi-turn approach over a single-turn approach in learning compositionality. Additionally, our dataset integrates specially crafted tokens into SQL queries to improve execution efficiency. With EHR-SeqSQL, we aim to bridge the gap between practical needs and academic research in the text-to-SQL domain. EHR-SeqSQL is available at https://github.com/seonhee99/EHR-SeqSQL."
                    },
                    {
                        "title": "Exploration into Translation-Equivariant Image Quantization",
                        "abstract": "This is an exploratory study that discovers the current image quantization (vector quantization) do not satisfy translation equivariance in the quantized space due to aliasing. Instead of focusing on anti-aliasing, we propose a simple yet effective way to achieve translation-equivariant image quantization by enforcing orthogonality among the codebook embeddings. To explore the advantages of translation-equivariant image quantization, we conduct three proof-of-concept experiments with a carefully controlled dataset: (1) text-to-image generation, where the quantized image indices are the target to predict, (2) image-to-text generation, where the quantized image indices are given as a condition, (3) using a smaller training set to analyze sample efficiency. From the strictly controlled experiments, we empirically verify that the translation-equivariant image quantizer improves not only sample efficiency but also the accuracy over VQGAN up to +11.9% in text-to-image generation and +3.9% in image-to-text generation."
                    },
                    {
                        "title": "Towards Unbiased Evaluation of Detecting Unanswerable Questions in EHRSQL",
                        "abstract": "Incorporating unanswerable questions into EHR QA systems is crucial for testing the trustworthiness of a system, as providing non-existent responses can mislead doctors in their diagnoses. The EHRSQL dataset stands out as a promising benchmark because it is the only dataset that incorporates unanswerable questions in the EHR QA system alongside practical questions. However, in this work, we identify a data bias in these unanswerable questions; they can often be discerned simply by filtering with specific N-gram patterns. Such biases jeopardize the authenticity and reliability of QA system evaluations. To tackle this problem, we propose a simple debiasing method of adjusting the split between the validation and test sets to neutralize the undue influence of N-gram filtering. By experimenting on the MIMIC-III dataset, we demonstrate both the existing data bias in EHRSQL and the effectiveness of our data split strategy in mitigating this bias."
                    },
                    {
                        "title": "ECG-QA: A Comprehensive Question Answering Dataset Combined With Electrocardiogram",
                        "abstract": "Question answering (QA) in the field of healthcare has received much attention due to significant advancements in natural language processing. However, existing healthcare QA datasets primarily focus on medical images, clinical notes, or structured electronic health record tables. This leaves the vast potential of combining electrocardiogram (ECG) data with these systems largely untapped. To address this gap, we present ECG-QA, the first QA dataset specifically designed for ECG analysis. The dataset comprises a total of 70 question templates that cover a wide range of clinically relevant ECG topics, each validated by an ECG expert to ensure their clinical utility. As a result, our dataset includes diverse ECG interpretation questions, including those that require a comparative analysis of two different ECGs. In addition, we have conducted numerous experiments to provide valuable insights for future research directions. We believe that ECG-QA will serve as a valuable resource for the development of intelligent QA systems capable of assisting clinicians in ECG interpretations. Dataset URL: https://github.com/Jwoo5/ecg-qa"
                    },
                    {
                        "title": "EHRSQL: A Practical Text-to-SQL Benchmark for Electronic Health Records",
                        "abstract": "We present a new text-to-SQL dataset for electronic health records (EHRs). The utterances were collected from 222 hospital staff members, including physicians, nurses, and insurance review and health records teams. To construct the QA dataset on structured EHR data, we conducted a poll at a university hospital and used the responses to create seed questions. We then manually linked these questions to two open-source EHR databases, MIMIC-III and eICU, and included various time expressions and held-out unanswerable questions in the dataset, which were also collected from the poll. Our dataset poses a unique set of challenges: the model needs to 1) generate SQL queries that reflect a wide range of needs in the hospital, including simple retrieval and complex operations such as calculating survival rate, 2) understand various time expressions to answer time-sensitive questions in healthcare, and 3) distinguish whether a given question is answerable or unanswerable. We believe our dataset, EHRSQL, can serve as a practical benchmark for developing and assessing QA models on structured EHR data and take a step further towards bridging the gap between text-to-SQL research and its real-life deployment in healthcare. EHRSQL is available at https://github.com/glee4810/EHRSQL."
                    },
                    {
                        "title": "Diverse and Admissible Trajectory Forecasting through Multimodal Context Understanding",
                        "abstract": "Multi-agent trajectory forecasting in autonomous driving requires an agent to accurately anticipate the behaviors of the surrounding vehicles and pedestrians, for safe and reliable decision-making. Due to partial observability in these dynamical scenes, directly obtaining the posterior distribution over future agent trajectories remains a challenging problem. In realistic embodied environments, each agent's future trajectories should be both diverse since multiple plausible sequences of actions can be used to reach its intended goals, and admissible since they must obey physical constraints and stay in drivable areas. In this paper, we propose a model that synthesizes multiple input signals from the multimodal world|the environment's scene context and interactions between multiple surrounding agents|to best model all diverse and admissible trajectories. We compare our model with strong baselines and ablations across two public datasets and show a significant performance improvement over previous state-of-the-art methods. Lastly, we offer new metrics incorporating admissibility criteria to further study and evaluate the diversity of predictions. Codes are at: https://github.com/kami93/CMU-DATF."
                    },
                    {
                        "title": "EHRXQA: A Multi-Modal Question Answering Dataset for Electronic Health Records with Chest X-ray Images",
                        "abstract": "Electronic Health Records (EHRs), which contain patients' medical histories in various multi-modal formats, often overlook the potential for joint reasoning across imaging and table modalities underexplored in current EHR Question Answering (QA) systems. In this paper, we introduce EHRXQA, a novel multi-modal question answering dataset combining structured EHRs and chest X-ray images. To develop our dataset, we first construct two uni-modal resources: 1) The MIMIC-CXR-VQA dataset, our newly created medical visual question answering (VQA) benchmark, specifically designed to augment the imaging modality in EHR QA, and 2) EHRSQL (MIMIC-IV), a refashioned version of a previously established table-based EHR QA dataset. By integrating these two uni-modal resources, we successfully construct a multi-modal EHR QA dataset that necessitates both uni-modal and cross-modal reasoning. To address the unique challenges of multi-modal questions within EHRs, we propose a NeuralSQL-based strategy equipped with an external VQA API. This pioneering endeavor enhances engagement with multi-modal EHR sources and we believe that our dataset can catalyze advances in real-world medical scenarios such as clinical decision-making and research. EHRXQA is available at https://github.com/baeseongsu/ehrxqa."
                    },
                    {
                        "title": "Publicly Shareable Clinical Large Language Model Built on Synthetic Clinical Notes",
                        "abstract": "The development of large language models tailored for handling patients' clinical notes is often hindered by the limited accessibility and usability of these notes due to strict privacy regulations. To address these challenges, we first create synthetic large-scale clinical notes using publicly available case reports extracted from biomedical literature. We then use these synthetic notes to train our specialized clinical large language model, Asclepius. While Asclepius is trained on synthetic data, we assess its potential performance in real-world applications by evaluating it using real clinical notes. We benchmark Asclepius against several other large language models, including GPT-3.5-turbo and other open-source alternatives. To further validate our approach using synthetic notes, we also compare Asclepius with its variants trained on real clinical notes. Our findings convincingly demonstrate that synthetic clinical notes can serve as viable substitutes for real ones when constructing high-performing clinical language models. This conclusion is supported by detailed evaluations conducted by both GPT-4 and medical professionals. All resources including weights, codes, and data used in the development of Asclepius are made publicly accessible for future research. (https://github.com/starmpcc/Asclepius)"
                    }
                ]
            },
            "af55ce0a-4b28-43a3-bdf8-32c570cdf2ae": {
                "pk": "af55ce0a-4b28-43a3-bdf8-32c570cdf2ae",
                "name": "Seongsu Bae",
                "collaborators": [
                    "Edward Choi",
                    "Gyubok Lee",
                    "Sunjun Kweon",
                    "Jungwoo Oh",
                    "Daeyoung Kim",
                    "Jiho Kim",
                    "Tackeun Kim",
                    "Lei Ji",
                    "Eunbyeol Cho",
                    "Seungho Kim"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Healthcare AI",
                    "Question Answering",
                    "Electronic Health Records"
                ],
                "publications": [
                    {
                        "title": "Question Answering for Complex Electronic Health Records Database using Unified Encoder-Decoder Architecture",
                        "abstract": "An intelligent machine that can answer human questions based on electronic health records (EHR-QA) has a great practical value, such as supporting clinical decisions, managing hospital administration, and medical chatbots. Previous table-based QA studies focusing on translating natural questions into table queries (NLQ2SQL), however, suffer from the unique nature of EHR data due to complex and specialized medical terminology, hence increased decoding difficulty. In this paper, we design UniQA, a unified encoder-decoder architecture for EHR-QA where natural language questions are converted to queries such as SQL or SPARQL. We also propose input masking (IM), a simple and effective method to cope with complex medical terms and various typos and better learn the SQL/SPARQL syntax. Combining the unified architecture with an effective auxiliary training objective, UniQA demonstrated a significant performance improvement against the previous state-of-the-art model for MIMICSQL* (14.2% gain), the most complex NLQ2SQL dataset in the EHR domain, and its typo-ridden versions (approximately 28.8% gain). In addition, we confirmed consistent results for the graph-based EHR-QA dataset, MIMICSPARQL*."
                    },
                    {
                        "title": "Uncertainty-Aware Text-to-Program for Question Answering on Structured Electronic Health Records",
                        "abstract": "Question Answering on Electronic Health Records (EHR-QA) has a significant impact on the healthcare domain, and it is being actively studied. Previous research on structured EHR-QA focuses on converting natural language queries into query language such as SQL or SPARQL (NLQ2Query), so the problem scope is limited to pre-defined data types by the specific query language. In order to expand the EHR-QA task beyond this limitation to handle multi-modal medical data and solve complex inference in the future, more primitive systemic language is needed. In this paper, we design the program-based model (NLQ2Program) for EHR-QA as the first step towards the future direction. We tackle MIMICSPARQL*, the graph-based EHR-QA dataset, via a program-based approach in a semi-supervised manner in order to overcome the absence of gold programs. Without the gold program, our proposed model shows comparable performance to the previous state-of-the-art model, which is an NLQ2Query model (0.9% gain). In addition, for a reliable EHR-QA model, we apply the uncertainty decomposition method to measure the ambiguity in the input question. We empirically confirmed data uncertainty is most indicative of the ambiguity in the input question."
                    },
                    {
                        "title": "Overview of the EHRSQL 2024 Shared Task on Reliable Text-to-SQL Modeling on Electronic Health Records",
                        "abstract": "Electronic Health Records (EHRs) are relational databases that store the entire medical histories of patients within hospitals. They record numerous aspects of patients' medical care, from hospital admission and diagnosis to treatment and discharge. While EHRs are vital sources of clinical data, exploring them beyond a predefined set of queries requires skills in query languages like SQL. To make information retrieval more accessible, one strategy is to build a question-answering system, possibly leveraging text-to-SQL models that can automatically translate natural language questions into corresponding SQL queries and use these queries to retrieve the answers. The EHRSQL 2024 shared task aims to advance and promote research in developing a question-answering system for EHRs using text-to-SQL modeling, capable of reliably providing requested answers to various healthcare professionals to improve their clinical work processes and satisfy their needs. Among more than 100 participants who applied to the shared task, eight teams were formed and completed the entire shared task requirement and demonstrated a wide range of methods to effectively solve this task. In this paper, we describe the task of reliable text-to-SQL modeling, the dataset, and the methods and results of the participants. We hope this shared task will spur further research and insights into developing reliable question-answering systems for EHRs."
                    },
                    {
                        "title": "Graph-Text Multi-Modal Pre-training for Medical Representation Learning",
                        "abstract": "As the volume of Electronic Health Records (EHR) sharply grows, there has been emerging interest in learning the representation of EHR for healthcare applications. Representation learning of EHR requires appropriate modeling of the two dominant modalities in EHR: structured data and unstructured text. In this paper, we present MedGTX, a pre-trained model for multi-modal representation learning of the structured and textual EHR data. MedGTX uses a novel graph encoder to exploit the graphical nature of structured EHR data, and a text encoder to handle unstructured text, and a cross-modal encoder to learn a joint representation space. We pre-train our model through four proxy tasks on MIMIC-III, an open-source EHR data, and evaluate our model on two clinical benchmarks and three novel downstream tasks which tackle real-world problems in EHR data. The results consistently show the effectiveness of pre-training the model for joint representation of both structured and unstructured information from EHR. Given the promising performance of MedGTX, we believe this work opens a new door to jointly understanding the two fundamental modalities of EHR data."
                    },
                    {
                        "title": "ECG-QA: A Comprehensive Question Answering Dataset Combined With Electrocardiogram",
                        "abstract": "Question answering (QA) in the field of healthcare has received much attention due to significant advancements in natural language processing. However, existing healthcare QA datasets primarily focus on medical images, clinical notes, or structured electronic health record tables. This leaves the vast potential of combining electrocardiogram (ECG) data with these systems largely untapped. To address this gap, we present ECG-QA, the first QA dataset specifically designed for ECG analysis. The dataset comprises a total of 70 question templates that cover a wide range of clinically relevant ECG topics, each validated by an ECG expert to ensure their clinical utility. As a result, our dataset includes diverse ECG interpretation questions, including those that require a comparative analysis of two different ECGs. In addition, we have conducted numerous experiments to provide valuable insights for future research directions. We believe that ECG-QA will serve as a valuable resource for the development of intelligent QA systems capable of assisting clinicians in ECG interpretations. Dataset URL: https://github.com/Jwoo5/ecg-qa"
                    },
                    {
                        "title": "KU-DMIS-MSRA at RadSum23: Pre-trained Vision-Language Model for Radiology Report Summarization",
                        "abstract": "In this paper, we introduce CheXOFA, a new pre-trained vision-language model (VLM) for the chest X-ray domain. Our model is initially pre-trained on various multimodal datasets within the general domain before being transferred to the chest X-ray domain. Following a prominent VLM, we unify various domain-specific tasks into a simple sequence-to-sequence schema. It enables the model to effectively learn the required knowledge and skills from limited resources in the domain. Demonstrating superior performance on the benchmark datasets provided by the BioNLP shared task, our model benefits from its training across multiple tasks and domains. With subtle techniques including ensemble and factual calibration, our system achieves first place on the RadSum23 leaderboard for the hidden test set."
                    },
                    {
                        "title": "EHRSQL: A Practical Text-to-SQL Benchmark for Electronic Health Records",
                        "abstract": "We present a new text-to-SQL dataset for electronic health records (EHRs). The utterances were collected from 222 hospital staff members, including physicians, nurses, and insurance review and health records teams. To construct the QA dataset on structured EHR data, we conducted a poll at a university hospital and used the responses to create seed questions. We then manually linked these questions to two open-source EHR databases, MIMIC-III and eICU, and included various time expressions and held-out unanswerable questions in the dataset, which were also collected from the poll. Our dataset poses a unique set of challenges: the model needs to 1) generate SQL queries that reflect a wide range of needs in the hospital, including simple retrieval and complex operations such as calculating survival rate, 2) understand various time expressions to answer time-sensitive questions in healthcare, and 3) distinguish whether a given question is answerable or unanswerable. We believe our dataset, EHRSQL, can serve as a practical benchmark for developing and assessing QA models on structured EHR data and take a step further towards bridging the gap between text-to-SQL research and its real-life deployment in healthcare. EHRSQL is available at https://github.com/glee4810/EHRSQL."
                    },
                    {
                        "title": "EHRXQA: A Multi-Modal Question Answering Dataset for Electronic Health Records with Chest X-ray Images",
                        "abstract": "Electronic Health Records (EHRs), which contain patients' medical histories in various multi-modal formats, often overlook the potential for joint reasoning across imaging and table modalities underexplored in current EHR Question Answering (QA) systems. In this paper, we introduce EHRXQA, a novel multi-modal question answering dataset combining structured EHRs and chest X-ray images. To develop our dataset, we first construct two uni-modal resources: 1) The MIMIC-CXR-VQA dataset, our newly created medical visual question answering (VQA) benchmark, specifically designed to augment the imaging modality in EHR QA, and 2) EHRSQL (MIMIC-IV), a refashioned version of a previously established table-based EHR QA dataset. By integrating these two uni-modal resources, we successfully construct a multi-modal EHR QA dataset that necessitates both uni-modal and cross-modal reasoning. To address the unique challenges of multi-modal questions within EHRs, we propose a NeuralSQL-based strategy equipped with an external VQA API. This pioneering endeavor enhances engagement with multi-modal EHR sources and we believe that our dataset can catalyze advances in real-world medical scenarios such as clinical decision-making and research. EHRXQA is available at https://github.com/baeseongsu/ehrxqa."
                    },
                    {
                        "title": "Publicly Shareable Clinical Large Language Model Built on Synthetic Clinical Notes",
                        "abstract": "The development of large language models tailored for handling patients' clinical notes is often hindered by the limited accessibility and usability of these notes due to strict privacy regulations. To address these challenges, we first create synthetic large-scale clinical notes using publicly available case reports extracted from biomedical literature. We then use these synthetic notes to train our specialized clinical large language model, Asclepius. While Asclepius is trained on synthetic data, we assess its potential performance in real-world applications by evaluating it using real clinical notes. We benchmark Asclepius against several other large language models, including GPT-3.5-turbo and other open-source alternatives. To further validate our approach using synthetic notes, we also compare Asclepius with its variants trained on real clinical notes. Our findings convincingly demonstrate that synthetic clinical notes can serve as viable substitutes for real ones when constructing high-performing clinical language models. This conclusion is supported by detailed evaluations conducted by both GPT-4 and medical professionals. All resources including weights, codes, and data used in the development of Asclepius are made publicly accessible for future research. (https://github.com/starmpcc/Asclepius)"
                    }
                ]
            },
            "f3056127-6b76-4a38-9b92-2040cb7f871a": {
                "pk": "f3056127-6b76-4a38-9b92-2040cb7f871a",
                "name": "Daeun Kyung",
                "collaborators": [
                    "Edward Choi",
                    "Tackeun Kim",
                    "Eunbyeol Cho",
                    "Junu Kim",
                    "Kyungmin Jo",
                    "Jaegul Choo",
                    "Joonseok Lee",
                    "Jongseong Jang",
                    "Seung Hwan Kim",
                    "Honglak Lee"
                ],
                "domain": [
                    "Medical Imaging",
                    "Generative Models",
                    "Multi-modal Learning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Towards Predicting Temporal Changes in a Patient's Chest X-ray Images based on Electronic Health Records",
                        "abstract": "Chest X-ray imaging (CXR) is an important diagnostic tool used in hospitals to assess patient conditions and monitor changes over time. Generative models, specifically diffusion-based models, have shown promise in generating realistic synthetic X-rays. However, these models mainly focus on conditional generation using single-time-point data, i.e., typically CXRs taken at a specific time with their corresponding reports, limiting their clinical utility, particularly for capturing temporal changes. To address this limitation, we propose a novel framework, EHRXDiff, which predicts future CXR images by integrating previous CXRs with subsequent medical events, e.g., prescriptions, lab measures, etc. Our framework dynamically tracks and predicts disease progression based on a latent diffusion model, conditioned on the previous CXR image and a history of medical events. We comprehensively evaluate the performance of our framework across three key aspects, including clinical consistency, demographic consistency, and visual realism. We demonstrate that our framework generates high-quality, realistic future images that capture potential temporal changes, suggesting its potential for further development as a clinical simulation tool. This could offer valuable insights for patient monitoring and treatment planning in the medical field."
                    },
                    {
                        "title": "Perspective Projection-Based 3D CT Reconstruction from Biplanar X-rays",
                        "abstract": "X-ray computed tomography (CT) is one of the most common imaging techniques used to diagnose various diseases in the medical field. Its high contrast sensitivity and spatial resolution allow the physician to observe details of body parts such as bones, soft tissue, blood vessels, etc. As it involves potentially harmful radiation exposure to patients and surgeons, however, reconstructing 3D CT volume from perpendicular 2D X-ray images is considered a promising alternative, thanks to its lower radiation risk and better accessibility. This is highly challenging though, since it requires reconstruction of 3D anatomical information from 2D images with limited views, where all the information is overlapped. In this paper, we propose PerX2CT, a novel CT reconstruction framework from X-ray that reflects the perspective projection scheme. Our proposed method provides a different combination of features for each coordinate which implicitly allows the model to obtain information about the 3D location. We reveal the potential to reconstruct the selected part of CT with high resolution by properly using the coordinate-wise local and global features. Our approach shows potential for use in clinical applications with low computational complexity and fast inference time, demonstrating superior performance than baselines in multiple evaluation metrics."
                    },
                    {
                        "title": "Significantly improving zero-shot X-ray pathology classification via fine-tuning pre-trained image-text encoders",
                        "abstract": "Deep neural networks are increasingly used in medical imaging for tasks such as pathological classification, but they face challenges due to the scarcity of high-quality, expert-labeled training data. Recent efforts have utilized pre-trained contrastive image-text models like CLIP, adapting them for medical use by fine-tuning the model with chest X-ray images and corresponding reports for zero-shot pathology classification, thus eliminating the need for pathology-specific annotations. However, most studies continue to use the same contrastive learning objectives as in the general domain, overlooking the multi-labeled nature of medical image-report pairs. In this paper, we propose a new fine-tuning strategy that includes positive-pair loss relaxation and random sentence sampling. We aim to improve the performance of zero-shot pathology classification without relying on external knowledge. Our method can be applied to any pre-trained contrastive image-text encoder and easily transferred to out-of-domain datasets without further training, as it does not use external data. Our approach consistently improves overall zero-shot pathology classification across four chest X-ray datasets and three pre-trained models, with an average macro AUROC increase of 4.3%. Additionally, our method outperforms the state-of-the-art and marginally surpasses board-certified radiologists in zero-shot classification for the five competition pathologies in the CheXpert dataset."
                    },
                    {
                        "title": "DialSim: A Real-Time Simulator for Evaluating Long-Term Multi-Party Dialogue Understanding of Conversational Agents",
                        "abstract": "Recent advancements in Large Language Models (LLMs) have significantly enhanced the capabilities of conversational agents, making them applicable to various fields (e.g., education). Despite their progress, the evaluation of the agents often overlooks the complexities of real-world conversations, such as real-time interactions, multi-party dialogues, and extended contextual dependencies. To bridge this gap, we introduce DialSim, a real-time dialogue simulator. In this simulator, an agent is assigned the role of a character from popular TV shows, requiring it to respond to spontaneous questions using past dialogue information and to distinguish between known and unknown information. Key features of DialSim include evaluating the agent's ability to respond within a reasonable time limit, handling long-term multi-party dialogues, and testing the agent's performance under randomized questioning with a diverse and high-quality question-answer dataset. We utilized this simulator to evaluate the latest conversational agents and analyze their limitations. Our experiments highlight both the strengths and weaknesses of these agents, providing valuable insights for future improvements in the field of conversational AI. DialSim is available at https://dialsim.github.io/."
                    },
                    {
                        "title": "EHRXQA: A Multi-Modal Question Answering Dataset for Electronic Health Records with Chest X-ray Images",
                        "abstract": "Electronic Health Records (EHRs), which contain patients' medical histories in various multi-modal formats, often overlook the potential for joint reasoning across imaging and table modalities underexplored in current EHR Question Answering (QA) systems. In this paper, we introduce EHRXQA, a novel multi-modal question answering dataset combining structured EHRs and chest X-ray images. To develop our dataset, we first construct two uni-modal resources: 1) The MIMIC-CXR-VQA dataset, our newly created medical visual question answering (VQA) benchmark, specifically designed to augment the imaging modality in EHR QA, and 2) EHRSQL (MIMIC-IV), a refashioned version of a previously established table-based EHR QA dataset. By integrating these two uni-modal resources, we successfully construct a multi-modal EHR QA dataset that necessitates both uni-modal and cross-modal reasoning. To address the unique challenges of multi-modal questions within EHRs, we propose a NeuralSQL-based strategy equipped with an external VQA API. This pioneering endeavor enhances engagement with multi-modal EHR sources and we believe that our dataset can catalyze advances in real-world medical scenarios such as clinical decision-making and research. EHRXQA is available at https://github.com/baeseongsu/ehrxqa."
                    }
                ]
            },
            "38501223-baa6-49a6-b094-bf14ced55fbd": {
                "pk": "38501223-baa6-49a6-b094-bf14ced55fbd",
                "name": "Wonchul Cha",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "acd33a76-fc3c-4b7b-95c3-9e725623a263": {
                "pk": "acd33a76-fc3c-4b7b-95c3-9e725623a263",
                "name": "Tom Pollard",
                "collaborators": [
                    "Gerardo Flores",
                    "George H. Chen",
                    "Joyce C. Ho",
                    "Tristan Naumann",
                    "Ming Ying Yang",
                    "Gloria Hyunjung Kwak",
                    "Leo Anthony Celi",
                    "Marzyeh Ghassemi"
                ],
                "domain": [
                    "Health Informatics",
                    "Social Determinants of Health",
                    "Machine Learning",
                    "Algorithmic Fairness"
                ],
                "publications": [
                    {
                        "title": "A collection of invited non-archival papers for the Conference on Health, Inference, and Learning (CHIL) 2022",
                        "abstract": "A collection of invited non-archival papers for the Conference on Health, Inference, and Learning (CHIL) 2022. This index is incomplete as some authors of invited non-archival presentations opted not to include their papers in this index."
                    },
                    {
                        "title": "Evaluating the Impact of Social Determinants on Health Prediction in the Intensive Care Unit",
                        "abstract": "Social determinants of health (SDOH) -- the conditions in which people live, grow, and age -- play a crucial role in a person's health and well-being. There is a large, compelling body of evidence in population health studies showing that a wide range of SDOH is strongly correlated with health outcomes. Yet, a majority of the risk prediction models based on electronic health records (EHR) do not incorporate a comprehensive set of SDOH features as they are often noisy or simply unavailable. Our work links a publicly available EHR database, MIMIC-IV, to well-documented SDOH features. We investigate the impact of such features on common EHR prediction tasks across different patient populations. We find that community-level SDOH features do not improve model performance for a general patient population, but can improve data-limited model fairness for specific subpopulations. We also demonstrate that SDOH features are vital for conducting thorough audits of algorithmic biases beyond protective attributes. We hope the new integrated EHR-SDOH database will enable studies on the relationship between community health and individual outcomes and provide new benchmarks to study algorithmic biases beyond race, gender, and age."
                    }
                ]
            },
            "4bdac30b-aa74-4e50-a190-81e1ac1fe912": {
                "pk": "4bdac30b-aa74-4e50-a190-81e1ac1fe912",
                "name": "Alistair Johnson",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "d3abe647-e9ce-45a9-ba96-b71b0b56bf5e": {
                "pk": "d3abe647-e9ce-45a9-ba96-b71b0b56bf5e",
                "name": "Edward Choi",
                "collaborators": [
                    "Gyubok Lee",
                    "Sungjin Park",
                    "Seongjun Yang",
                    "Seongsu Bae",
                    "Junu Kim",
                    "Daehoon Gwak",
                    "Daeyoung Kim",
                    "Jiho Kim",
                    "Kyunghoon Hur",
                    "Yeonsu Kwon"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Natural Language Processing",
                    "Electronic Health Records",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Multimodal Transformer With a Low-Computational-Cost Guarantee",
                        "abstract": "Transformer-based models have significantly improved performance across a range of multimodal understanding tasks, such as visual question answering and action recognition. However, multimodal Transformers significantly suffer from a quadratic complexity of the multi-head attention with the input sequence length, especially as the number of modalities increases. To address this, we introduce Low-Cost Multimodal Transformer (LoCoMT), a novel multimodal attention mechanism that aims to reduce computational cost during training and inference with minimal performance loss. Specifically, by assigning different multimodal attention patterns to each attention head, LoCoMT can flexibly control multimodal signals and theoretically ensures a reduced computational cost compared to existing multimodal Transformer variants. Experimental results on two multimodal datasets, namely Audioset and MedVidCL demonstrate that LoCoMT not only reduces GFLOPs but also matches or even outperforms established models."
                    },
                    {
                        "title": "Compositional Obverter Communication Learning From Raw Visual Input",
                        "abstract": "One of the distinguishing aspects of human language is its compositionality, which allows us to describe complex environments with limited vocabulary. Previously, it has been shown that neural network agents can learn to communicate in a highly structured, possibly compositional language based on disentangled input (e.g. hand- engineered features). Humans, however, do not learn to communicate based on well-summarized features. In this work, we train neural agents to simultaneously develop visual perception from raw image pixels, and learn to communicate with a sequence of discrete symbols. The agents play an image description game where the image contains factors such as colors and shapes. We train the agents using the obverter technique where an agent introspects to generate messages that maximize its own understanding. Through qualitative analysis, visualization and a zero-shot test, we show that the agents can develop, out of raw image pixels, a language with compositional properties, given a proper pressure from the environment."
                    },
                    {
                        "title": "Do Language Models Understand Measurements?",
                        "abstract": "Recent success of pre-trained language models (PLMs) has stimulated interest in their ability to understand and work with numbers. Yet, the numerical reasoning over measurements has not been formally studied despite their importance. In this study, we show that PLMs lack the capability required for reasoning over measurements. Furthermore, we find that a language model trained on a measurement-rich corpus shows better performance on understanding measurements. We propose a simple embedding strategy to better distinguish between numbers and units, which leads to a significant improvement in the probing tasks."
                    },
                    {
                        "title": "Conditional Generation of Periodic Signals with Fourier-Based Decoder",
                        "abstract": "Periodic signals play an important role in daily lives. Although conventional sequential models have shown remarkable success in various fields, they still come short in modeling periodicity; they either collapse, diverge or ignore details. In this paper, we introduce a novel framework inspired by Fourier series to generate periodic signals. We first decompose the given signals into multiple sines and cosines and then conditionally generate periodic signals with the output components. We have shown our model efficacy on three tasks: reconstruction, imputation and conditional generation. Our model outperforms baselines in all tasks and shows more stable and refined results."
                    },
                    {
                        "title": "Unconditional Image-Text Pair Generation with Multimodal Cross Quantizer",
                        "abstract": "Although deep generative models have gained a lot of attention, most of the existing works are designed for unimodal generation. In this paper, we explore a new method for unconditional image-text pair generation. We design Multimodal Cross-Quantization VAE (MXQ-VAE), a novel vector quantizer for joint image-text representations, with which we discover that a joint image-text representation space is effective for semantically consistent image-text pair generation. To learn a multimodal semantic correlation in a quantized space, we combine VQ-VAE with a Transformer encoder and apply an input masking strategy. Specifically, MXQ-VAE accepts a masked image-text pair as input and learns a quantized joint representation space, so that the input can be converted to a unified code sequence, then we perform unconditional image-text pair generation with the code sequence. Extensive experiments show the correlation between the quantized joint space and the multimodal generation capability on synthetic and real-world datasets. In addition, we demonstrate the superiority of our approach in these two aspects over several baselines. The source code is publicly available at: https://github.com/ttumyche/MXQ-VAE."
                    },
                    {
                        "title": "Improving Lexically Constrained Neural Machine Translation with Source-Conditioned Masked Span Prediction",
                        "abstract": "Accurate terminology translation is crucial for ensuring the practicality and reliability of neural machine translation (NMT) systems. To address this, lexically constrained NMT explores various methods to ensure pre-specified words and phrases appear in the translation output. However, in many cases, those methods are studied on general domain corpora, where the terms are mostly uni- and bi-grams (>98%). In this paper, we instead tackle a more challenging setup consisting of domain-specific corpora with much longer n-gram and highly specialized terms. Inspired by the recent success of masked span prediction models, we propose a simple and effective training strategy that achieves consistent improvements on both terminology and sentence-level translation for three domain-specific corpora in two language pairs."
                    },
                    {
                        "title": "Question Answering for Complex Electronic Health Records Database using Unified Encoder-Decoder Architecture",
                        "abstract": "An intelligent machine that can answer human questions based on electronic health records (EHR-QA) has a great practical value, such as supporting clinical decisions, managing hospital administration, and medical chatbots. Previous table-based QA studies focusing on translating natural questions into table queries (NLQ2SQL), however, suffer from the unique nature of EHR data due to complex and specialized medical terminology, hence increased decoding difficulty. In this paper, we design UniQA, a unified encoder-decoder architecture for EHR-QA where natural language questions are converted to queries such as SQL or SPARQL. We also propose input masking (IM), a simple and effective method to cope with complex medical terms and various typos and better learn the SQL/SPARQL syntax. Combining the unified architecture with an effective auxiliary training objective, UniQA demonstrated a significant performance improvement against the previous state-of-the-art model for MIMICSQL* (14.2% gain), the most complex NLQ2SQL dataset in the EHR domain, and its typo-ridden versions (approximately 28.8% gain). In addition, we confirmed consistent results for the graph-based EHR-QA dataset, MIMICSPARQL*."
                    },
                    {
                        "title": "Task Agnostic and Post-hoc Unseen Distribution Detection",
                        "abstract": "Despite the recent advances in out-of-distribution(OOD) detection, anomaly detection, and uncertainty estimation tasks, there do not exist a task-agnostic and post-hoc approach. To address this limitation, we design a novel clustering-based ensembling method, called Task Agnostic and Post-hoc Unseen Distribution Detection (TAPUDD) that utilizes the features extracted from the model trained on a specific task. Explicitly, it comprises of TAP-Mahalanobis, which clusters the training datasets' features and determines the minimum Mahalanobis distance of the test sample from all clusters. Further, we propose the Ensembling module that aggregates the computation of iterative TAP-Mahalanobis for a different number of clusters to provide reliable and efficient cluster computation. Through extensive experiments on synthetic and real-world datasets, we observe that our approach can detect unseen samples effectively across diverse tasks and performs better or on-par with the existing baselines. To this end, we eliminate the necessity of determining the optimal value of the number of clusters and demonstrate that our method is more viable for large-scale classification tasks."
                    },
                    {
                        "title": "Universal EHR Federated Learning Framework",
                        "abstract": "Federated learning (FL) is the most practical multi-source learning method for electronic healthcare records (EHR). Despite its guarantee of privacy protection, the wide application of FL is restricted by two large challenges: the heterogeneous EHR systems, and the non-i.i.d. data characteristic. A recent research proposed a framework that unifies heterogeneous EHRs, named UniHPF. We attempt to address both the challenges simultaneously by combining UniHPF and FL. Our study is the first approach to unify heterogeneous EHRs into a single FL framework. This combination provides an average of 3.4% performance gain compared to local learning. We believe that our framework is practically applicable in the real-world FL."
                    },
                    {
                        "title": "KG-GPT: A General Framework for Reasoning on Knowledge Graphs Using Large Language Models",
                        "abstract": "While large language models (LLMs) have made considerable advancements in understanding and generating unstructured text, their application in structured data remains underexplored. Particularly, using LLMs for complex reasoning tasks on knowledge graphs (KGs) remains largely untouched. To address this, we propose KG-GPT, a multi-purpose framework leveraging LLMs for tasks employing KGs. KG-GPT comprises three steps: Sentence Segmentation, Graph Retrieval, and Inference, each aimed at partitioning sentences, retrieving relevant graph components, and deriving logical conclusions, respectively. We evaluate KG-GPT using KG-based fact verification and KGQA benchmarks, with the model showing competitive and robust performance, even outperforming several fully-supervised models. Our work, therefore, marks a significant step in unifying structured and unstructured data processing within the realm of LLMs."
                    },
                    {
                        "title": "Self-Supervised Contrastive Learning for Long-term Forecasting",
                        "abstract": "Long-term forecasting presents unique challenges due to the time and memory complexity of handling long sequences. Existing methods, which rely on sliding windows to process long sequences, struggle to effectively capture long-term variations that are partially caught within the short window (i.e., outer-window variations). In this paper, we introduce a novel approach that overcomes this limitation by employing contrastive learning and enhanced decomposition architecture, specifically designed to focus on long-term variations. To this end, our contrastive loss incorporates global autocorrelation held in the whole time series, which facilitates the construction of positive and negative pairs in a self-supervised manner. When combined with our decomposition networks, our contrastive learning significantly improves long-term forecasting performance. Extensive experiments demonstrate that our approach outperforms 14 baseline models in multiple experiments over nine long-term benchmarks, especially in challenging scenarios that require a significantly long output for forecasting. Source code is available at https://github.com/junwoopark92/Self-Supervised-Contrastive-Forecsating."
                    },
                    {
                        "title": "EHRFL: Federated Learning Framework for Institution-Specific Model Construction using Electronic Health Records",
                        "abstract": "The increasing volume of electronic health records (EHRs) across healthcare institutions presents the opportunity to enhance model accuracy and robustness in clinical prediction tasks. Federated learning enables training on data from multiple institutions while preserving patient privacy and complying to regulatory constraints. However, most federated learning research focuses on constructing a global model for multiple clients, overlooking the practical need for institution-specific models. In this work, we introduce EHRFL, a federated learning framework using EHRs designed to develop a model tailored to a single healthcare institution. Our framework addresses two key challenges: (1) enabling federated learning across institutions with heterogeneous EHR systems using text-based EHR modeling, and (2) reducing the costs associated with federated learning by selecting suitable participating clients using averaged patient embeddings, which enables optimizing the number of participants without compromising model performance for the institution. Our experiment results on multiple open-source EHR datasets demonstrate the effectiveness of EHRFL in addressing the two challenges, establishing it as a practical solution for institution-specific model development in federated learning."
                    },
                    {
                        "title": "EHR-SeqSQL : A Sequential Text-to-SQL Dataset For Interactively Exploring Electronic Health Records",
                        "abstract": "In this paper, we introduce EHR-SeqSQL, a novel sequential text-to-SQL dataset for Electronic Health Record (EHR) databases. EHR-SeqSQL is designed to address critical yet underexplored aspects in text-to-SQL parsing: interactivity, compositionality, and efficiency. To the best of our knowledge, EHR-SeqSQL is not only the largest but also the first medical text-to-SQL dataset benchmark to include sequential and contextual questions. We provide a data split and the new test set designed to assess compositional generalization ability. Our experiments demonstrate the superiority of a multi-turn approach over a single-turn approach in learning compositionality. Additionally, our dataset integrates specially crafted tokens into SQL queries to improve execution efficiency. With EHR-SeqSQL, we aim to bridge the gap between practical needs and academic research in the text-to-SQL domain. EHR-SeqSQL is available at https://github.com/seonhee99/EHR-SeqSQL."
                    },
                    {
                        "title": "Towards Predicting Temporal Changes in a Patient's Chest X-ray Images based on Electronic Health Records",
                        "abstract": "Chest X-ray imaging (CXR) is an important diagnostic tool used in hospitals to assess patient conditions and monitor changes over time. Generative models, specifically diffusion-based models, have shown promise in generating realistic synthetic X-rays. However, these models mainly focus on conditional generation using single-time-point data, i.e., typically CXRs taken at a specific time with their corresponding reports, limiting their clinical utility, particularly for capturing temporal changes. To address this limitation, we propose a novel framework, EHRXDiff, which predicts future CXR images by integrating previous CXRs with subsequent medical events, e.g., prescriptions, lab measures, etc. Our framework dynamically tracks and predicts disease progression based on a latent diffusion model, conditioned on the previous CXR image and a history of medical events. We comprehensively evaluate the performance of our framework across three key aspects, including clinical consistency, demographic consistency, and visual realism. We demonstrate that our framework generates high-quality, realistic future images that capture potential temporal changes, suggesting its potential for further development as a clinical simulation tool. This could offer valuable insights for patient monitoring and treatment planning in the medical field."
                    },
                    {
                        "title": "Uncertainty-Aware Text-to-Program for Question Answering on Structured Electronic Health Records",
                        "abstract": "Question Answering on Electronic Health Records (EHR-QA) has a significant impact on the healthcare domain, and it is being actively studied. Previous research on structured EHR-QA focuses on converting natural language queries into query language such as SQL or SPARQL (NLQ2Query), so the problem scope is limited to pre-defined data types by the specific query language. In order to expand the EHR-QA task beyond this limitation to handle multi-modal medical data and solve complex inference in the future, more primitive systemic language is needed. In this paper, we design the program-based model (NLQ2Program) for EHR-QA as the first step towards the future direction. We tackle MIMICSPARQL*, the graph-based EHR-QA dataset, via a program-based approach in a semi-supervised manner in order to overcome the absence of gold programs. Without the gold program, our proposed model shows comparable performance to the previous state-of-the-art model, which is an NLQ2Query model (0.9% gain). In addition, for a reliable EHR-QA model, we apply the uncertainty decomposition method to measure the ambiguity in the input question. We empirically confirmed data uncertainty is most indicative of the ambiguity in the input question."
                    },
                    {
                        "title": "Overview of the EHRSQL 2024 Shared Task on Reliable Text-to-SQL Modeling on Electronic Health Records",
                        "abstract": "Electronic Health Records (EHRs) are relational databases that store the entire medical histories of patients within hospitals. They record numerous aspects of patients' medical care, from hospital admission and diagnosis to treatment and discharge. While EHRs are vital sources of clinical data, exploring them beyond a predefined set of queries requires skills in query languages like SQL. To make information retrieval more accessible, one strategy is to build a question-answering system, possibly leveraging text-to-SQL models that can automatically translate natural language questions into corresponding SQL queries and use these queries to retrieve the answers. The EHRSQL 2024 shared task aims to advance and promote research in developing a question-answering system for EHRs using text-to-SQL modeling, capable of reliably providing requested answers to various healthcare professionals to improve their clinical work processes and satisfy their needs. Among more than 100 participants who applied to the shared task, eight teams were formed and completed the entire shared task requirement and demonstrated a wide range of methods to effectively solve this task. In this paper, we describe the task of reliable text-to-SQL modeling, the dataset, and the methods and results of the participants. We hope this shared task will spur further research and insights into developing reliable question-answering systems for EHRs."
                    },
                    {
                        "title": "Time is Not Enough: Time-Frequency based Explanation for Time-Series Black-Box Models",
                        "abstract": "Despite the massive attention given to time-series explanations due to their extensive applications, a notable limitation in existing approaches is their primary reliance on the time-domain. This overlooks the inherent characteristic of time-series data containing both time and frequency features. In this work, we present Spectral eXplanation (SpectralX), an XAI framework that provides time-frequency explanations for time-series black-box classifiers. This easily adaptable framework enables users to \"plug-in\" various perturbation-based XAI methods for any pre-trained time-series classification models to assess their impact on the explanation quality without having to modify the framework architecture. Additionally, we introduce Feature Importance Approximations (FIA), a new perturbation-based XAI method. These methods consist of feature insertion, deletion, and combination techniques to enhance computational efficiency and class-specific explanations in time-series classification tasks. We conduct extensive experiments in the generated synthetic dataset and various UCR Time-Series datasets to first compare the explanation performance of FIA and other existing perturbation-based XAI methods in both time-domain and time-frequency domain, and then show the superiority of our FIA in the time-frequency domain with the SpectralX framework. Finally, we conduct a user study to confirm the practicality of our FIA in SpectralX framework for class-specific time-frequency based time-series explanations. The source code is available in https://github.com/gustmd0121/Time_is_not_Enough"
                    },
                    {
                        "title": "MiME: Multilevel Medical Embedding of Electronic Health Records for Predictive Healthcare",
                        "abstract": "Deep learning models exhibit state-of-the-art performance for many predictive healthcare tasks using electronic health records (EHR) data, but these models typically require training data volume that exceeds the capacity of most healthcare systems. External resources such as medical ontologies are used to bridge the data volume constraint, but this approach is often not directly applicable or useful because of inconsistencies with terminology. To solve the data insufficiency challenge, we leverage the inherent multilevel structure of EHR data and, in particular, the encoded relationships among medical codes. We propose Multilevel Medical Embedding (MiME) which learns the multilevel embedding of EHR data while jointly performing auxiliary prediction tasks that rely on this inherent EHR structure without the need for external labels. We conducted two prediction tasks, heart failure prediction and sequential disease prediction, where MiME outperformed baseline methods in diverse evaluation settings. In particular, MiME consistently outperformed all baselines when predicting heart failure on datasets of different volumes, especially demonstrating the greatest performance improvement (15% relative gain in PR-AUC over the best baseline) on the smallest dataset, demonstrating its ability to effectively model the multilevel structure of EHR data."
                    },
                    {
                        "title": "TrustSQL: Benchmarking Text-to-SQL Reliability with Penalty-Based Scoring",
                        "abstract": "Text-to-SQL enables users to interact with databases using natural language, simplifying the retrieval and synthesis of information. Despite the remarkable success of large language models (LLMs) in translating natural language questions into SQL queries, widespread deployment remains limited due to two primary challenges. First, the effective use of text-to-SQL models depends on users' understanding of the model's capabilities-the scope of questions the model can correctly answer. Second, the absence of abstention mechanisms can lead to incorrect SQL generation going unnoticed, thereby undermining trust in the model's output. To enable wider deployment, it is crucial to address these challenges in model design and enhance model evaluation to build trust in the model's output. To this end, we introduce TrustSQL, a novel comprehensive benchmark designed to evaluate text-to-SQL reliability-defined as a model's ability to correctly handle any type of input question by generating correct SQL queries for feasible questions and abstaining from generating infeasible ones (e.g., due to schema incompatibility or functionalities beyond SQL). We evaluate existing methods using a novel penalty-based scoring metric with two modeling approaches: (1) pipeline-based methods combining SQL generators with infeasible question detectors and SQL error detectors for abstention; and (2) unified methods using a single model for the entire task. Our experimental results reveal that achieving high scores under severe penalties requires significant effort and provide a new perspective on developing text-to-SQL models for safer deployment. TrustSQL is available at https://github.com/glee4810/TrustSQL."
                    }
                ]
            }
        }
    },
    "2410.06621": {
        "paper_data": {
            "title": "Effective Exploration Based on the Structural Information Principles",
            "url": "http://arxiv.org/abs/2410.06621v1",
            "arxiv_id": "2410.06621",
            "authors": [
                "Xianghua Zeng",
                "Hao Peng",
                "Angsheng Li"
            ],
            "abstract": "Traditional information theory provides a valuable foundation for Reinforcement Learning, particularly through representation learning and entropy maximization for agent exploration. However, existing methods primarily concentrate on modeling the uncertainty associated with RL's random variables, neglecting the inherent structure within the state and action spaces. In this paper, we propose a novel Structural Information principles-based Effective Exploration framework, namely SI2E. Structural mutual information between two variables is defined to address the single-variable limitation in structural information, and an innovative embedding principle is presented to capture dynamics-relevant state-action representations. The SI2E analyzes value differences in the agent's policy between state-action pairs and minimizes structural entropy to derive the hierarchical state-action structure, referred to as the encoding tree. Under this tree structure, value-conditional structural entropy is defined and maximized to design an intrinsic reward mechanism that avoids redundant transitions and promotes enhanced coverage in the state-action space. Theoretical connections are established between SI2E and classical information-theoretic methodologies, highlighting our framework's rationality and advantage. Comprehensive evaluations in the MiniGrid, MetaWorld, and DeepMind Control Suite benchmarks demonstrate that SI2E significantly outperforms state-of-the-art exploration baselines regarding final performance and sample efficiency, with maximum improvements of 37.63% and 60.25%, respectively.",
            "introduction": "   1 Introduction   Figure 1: By incorporating the inherent state-action structure, we simplify the original six-state Markov Decision Process (MDP) with four actions to a five-state MDP with two actions, effectively reducing the size of state-action space from 24\u2062(6\u00d74)246424(6\\times 4)24 ( 6 \u00d7 4 ) to 10\u2062(5\u00d72)105210(5\\times 2)10 ( 5 \u00d7 2 ). Here, s2\u2032superscriptsubscript\ud835\udc602\u2032s_{2}^{\\prime}italic_s start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT \u2032 end_POSTSUPERSCRIPT and a0\u2032superscriptsubscript\ud835\udc4e0\u2032a_{0}^{\\prime}italic_a start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT \u2032 end_POSTSUPERSCRIPT represent vertex communities {s2,s5}subscript\ud835\udc602subscript\ud835\udc605\\{s_{2},s_{5}\\}{ italic_s start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT , italic_s start_POSTSUBSCRIPT 5 end_POSTSUBSCRIPT } and {a0,a1}subscript\ud835\udc4e0subscript\ud835\udc4e1\\{a_{0},a_{1}\\}{ italic_a start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_a start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT }, respectively. In this scenario, a policy maximizing state-action Shannon entropy would encompass all possible transitions (blue color). In contrast, a policy maximizing structural entropy would selectively focus on crucial transitions (red color), avoiding redundant transitions between s2subscript\ud835\udc602s_{2}italic_s start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT and s5subscript\ud835\udc605s_{5}italic_s start_POSTSUBSCRIPT 5 end_POSTSUBSCRIPT.   Reinforcement Learning (RL) has emerged as a pivotal technique for addressing sequential decision-making problems, including game intelligence (Vinyals et\u00a0al., 2019; Badia et\u00a0al., 2020), robotic control (Andrychowicz et\u00a0al., 2017; Liu and Abbeel, 2021), and autonomous driving (Prathiba et\u00a0al., 2021; P\u00e9rez-Gil et\u00a0al., 2022). In the realm of RL, striking a balance between exploration and exploitation is crucial for optimizing agent policies and mitigating the risk of suboptimal outcomes, especially in scenarios characterized by high dimensions and sparse rewards (Zhang et\u00a0al., 2021b).   Recently, advancements in information-theoretic approaches have shown promise for exploration in self-supervised settings. The maximum entropy framework over the action space (Haarnoja et\u00a0al., 2017) has led to the development of robust algorithms such as Soft Q-learning (Nachum et\u00a0al., 2017), SAC (Haarnoja et\u00a0al., 2018), and MPO (Abdolmaleki et\u00a0al., 2018). Additionally, various objectives focused on maximizing state entropy are utilized to ensure comprehensive state coverage (Hazan et\u00a0al., 2019; Islam et\u00a0al., 2019). To facilitate the exploration of complex state-action pairs, MaxRenyi optimizes R\u00e9nyi entropy across the state-action space (Zhang et\u00a0al., 2021a). However, a prevalent issue with entropy maximization strategies is their tendency to bias exploration towards low-value states, making them vulnerable to imbalanced state-value distributions in supervised settings. To mitigate this, value-conditional state entropy is introduced to compute intrinsic rewards based on the estimated values of visited states (Kim et\u00a0al., 2023). Due to their instability in noisy and high-dimensional environments, a Dynamic Bottleneck (DB) (Bai et\u00a0al., 2021) is developed based on the Information Bottleneck (IB) principle (Tishby et\u00a0al., 2000), thereby obtaining dynamics-relevant representations of state-action pairs. Despite their successes, existing information-theoretic exploration methods have a critical limitation: they often overlook the inherent structure within state and action spaces. This oversight necessitates new approaches to enhance exploration effectiveness.   Figure 1 illustrates a simple six-state Markov Decision Process (MDP) with four actions. The different densities of the blue and red lines represent different actions, as indicated in the legend, leading to state transitions aimed at optimizing the return to the initial state s0subscript\ud835\udc600s_{0}italic_s start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT. Solid lines specifically denote actions a0subscript\ud835\udc4e0a_{0}italic_a start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT and a1subscript\ud835\udc4e1a_{1}italic_a start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT. The transitions between states s2subscript\ud835\udc602s_{2}italic_s start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT and s5subscript\ud835\udc605s_{5}italic_s start_POSTSUBSCRIPT 5 end_POSTSUBSCRIPT are deemed redundant as they do not facilitate the primary objective of efficiently returning to s0subscript\ud835\udc600s_{0}italic_s start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT. Therefore, the state-action pairs (s2,a3)subscript\ud835\udc602subscript\ud835\udc4e3(s_{2},a_{3})( italic_s start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT , italic_a start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT",
            "references": [
                {
                    "title": "Multi-Relational Structural Entropy",
                    "abstract": "Structural Entropy (SE) measures the structural information contained in a graph. Minimizing or maximizing SE helps to reveal or obscure the intrinsic structural patterns underlying graphs in an interpretable manner, finding applications in various tasks driven by networked data. However, SE ignores the heterogeneity inherent in the graph relations, which is ubiquitous in modern networks. In this work, we extend SE to consider heterogeneous relations and propose the first metric for multi-relational graph structural information, namely, Multi-relational Structural Entropy (MrSE). To this end, we first cast SE through the novel lens of the stationary distribution from random surfing, which readily extends to multi-relational networks by considering the choices of both nodes and relation types simultaneously at each step. The resulting MrSE is then optimized by a new greedy algorithm to reveal the essential structures within a multi-relational network. Experimental results highlight that the proposed MrSE offers a more insightful interpretation of the structure of multi-relational graphs compared to SE. Additionally, it enhances the performance of two tasks that involve real-world multi-relational graphs, including node clustering and social event detection."
                },
                {
                    "title": "Unsupervised Social Bot Detection via Structural Information Theory",
                    "abstract": "Research on social bot detection plays a crucial role in maintaining the order and reliability of information dissemination while increasing trust in social interactions. The current mainstream social bot detection models rely on black-box neural network technology, e.g., Graph Neural Network, Transformer, etc., which lacks interpretability. In this work, we present UnDBot, a novel unsupervised, interpretable, yet effective and practical framework for detecting social bots. This framework is built upon structural information theory. We begin by designing three social relationship metrics that capture various aspects of social bot behaviors: Posting Type Distribution, Posting Influence, and Follow-to-follower Ratio. Three new relationships are utilized to construct a new, unified, and weighted social multi-relational graph, aiming to model the relevance of social user behaviors and discover long-distance correlations between users. Second, we introduce a novel method for optimizing heterogeneous structural entropy. This method involves the personalized aggregation of edge information from the social multi-relational graph to generate a two-dimensional encoding tree. The heterogeneous structural entropy facilitates decoding of the substantial structure of the social bots network and enables hierarchical clustering of social bots. Thirdly, a new community labeling method is presented to distinguish social bot communities by computing the user\u2019s stationary distribution, measuring user contributions to network structure, and counting the intensity of user aggregation within the community. Compared with ten representative social bot detection approaches, comprehensive experiments demonstrate the advantages of effectiveness and interpretability of UnDBot on four real social network datasets."
                },
                {
                    "title": "Hierarchical and Incremental Structural Entropy Minimization for Unsupervised Social Event Detection",
                    "abstract": "As a trending approach for social event detection, graph neural network (GNN)-based methods enable a fusion of natural language semantics and the complex social network structural information, thus showing SOTA performance. However, GNN-based methods can miss useful message correlations. Moreover, they require manual labeling for training and predetermining the number of events for prediction. In this work, we address social event detection via graph structural entropy (SE) minimization. While keeping the merits of the GNN-based methods, the proposed framework, HISEvent, constructs more informative message graphs, is unsupervised, and does not require the number of events given a priori. Specifically, we incrementally explore the graph neighborhoods using 1-dimensional (1D) SE minimization to supplement the existing message graph with edges between semantically related messages. We then detect events from the message graph by hierarchically minimizing 2-dimensional (2D) SE. Our proposed 1D and 2D SE minimization algorithms are customized for social event detection and effectively tackle the efficiency problem of the existing SE minimization algorithms. Extensive experiments show that HISEvent consistently outperforms GNN-based methods and achieves the new SOTA for social event detection under both closed- and open-set settings while being efficient and robust."
                },
                {
                    "title": "Adversarial Socialbots Modeling Based on Structural Information Principles",
                    "abstract": "The importance of effective detection is underscored by the fact that socialbots imitate human behavior to propagate misinformation, leading to an ongoing competition between socialbots and detectors. Despite the rapid advancement of reactive detectors, the exploration of adversarial socialbot modeling remains incomplete, significantly hindering the development of proactive detectors. To address this issue, we propose a mathematical Structural Information principles-based Adversarial Socialbots Modeling framework, namely SIASM, to enable more accurate and effective modeling of adversarial behaviors. First, a heterogeneous graph is presented to integrate various users and rich activities in the original social network and measure its dynamic uncertainty as structural entropy. By minimizing the high-dimensional structural entropy, a hierarchical community structure of the social network is generated and referred to as the optimal encoding tree. Secondly, a novel method is designed to quantify influence by utilizing the assigned structural entropy, which helps reduce the computational cost of SIASM by filtering out uninfluential users. Besides, a new conditional structural entropy is defined between the socialbot and other users to guide the follower selection for network influence maximization. Extensive and comparative experiments on both homogeneous and heterogeneous social networks demonstrate that, compared with state-of-the-art baselines, the proposed SIASM framework yields substantial performance improvements in terms of network influence (up to 16.32%) and sustainable stealthiness (up to 16.29%) when evaluated against a robust detector with 90% accuracy."
                },
                {
                    "title": "Unsupervised Skin Lesion Segmentation via Structural Entropy Minimization on Multi-Scale Superpixel Graphs",
                    "abstract": "Skin lesion segmentation is a fundamental task in dermoscopic image analysis. The complex features of pixels in the lesion region impede the lesion segmentation accuracy, and existing deep learning-based methods often lack interpretability to this problem. In this work, we propose a novel unsupervised Skin Lesion sEgmentation framework based on structural entropy and isolation forest outlier Detection, namely SLED. Specifically, skin lesions are segmented by minimizing the structural entropy of a superpixel graph constructed from the dermoscopic image. Then, we characterize the consistency of healthy skin features and devise a novel multi-scale segmentation mechanism by outlier detection, which enhances the segmentation accuracy by leveraging the superpixel features from multiple scales. We conduct experiments on four skin lesion benchmarks and compare SLED with nine representative unsupervised segmentation methods. Experimental results demonstrate the superiority of the proposed framework. Additionally, some case studies are analyzed to demonstrate the effectiveness of SLED."
                },
                {
                    "title": "CEM: Constrained Entropy Maximization for Task-Agnostic Safe Exploration",
                    "abstract": "In the absence of assigned tasks, a learning agent typically seeks to explore its environment efficiently. However, the pursuit of exploration will bring more safety risks.\nAn under-explored aspect of reinforcement learning is how to achieve safe efficient exploration when the task is unknown.\nIn this paper, we propose a practical Constrained Entropy Maximization (CEM) algorithm to solve task-agnostic safe exploration problems, which naturally require a finite horizon and undiscounted constraints on safety costs.\nThe CEM algorithm aims to learn a policy that maximizes state entropy under the premise of safety.\nTo avoid approximating the state density in complex domains, CEM leverages a k-nearest neighbor entropy estimator to evaluate the efficiency of exploration.\nIn terms of safety, CEM minimizes the safety costs, and adaptively trades off safety and exploration based on the current constraint satisfaction. The empirical analysis shows that CEM enables the acquisition of a safe exploration policy in complex environments, resulting in improved performance in both safety and sample efficiency for target tasks."
                },
                {
                    "title": "TACO: Temporal Latent Action-Driven Contrastive Loss for Visual Reinforcement Learning",
                    "abstract": "Despite recent progress in reinforcement learning (RL) from raw pixel data, sample inefficiency continues to present a substantial obstacle. Prior works have attempted to address this challenge by creating self-supervised auxiliary tasks, aiming to enrich the agent's learned representations with control-relevant information for future state prediction. However, these objectives are often insufficient to learn representations that can represent the optimal policy or value function, and they often consider tasks with small, abstract discrete action spaces and thus overlook the importance of action representation learning in continuous control. In this paper, we introduce TACO: Temporal Action-driven Contrastive Learning, a simple yet powerful temporal contrastive learning approach that facilitates the concurrent acquisition of latent state and action representations for agents. TACO simultaneously learns a state and an action representation by optimizing the mutual information between representations of current states paired with action sequences and representations of the corresponding future states. Theoretically, TACO can be shown to learn state and action representations that encompass sufficient information for control, thereby improving sample efficiency. For online RL, TACO achieves 40% performance boost after one million environment interaction steps on average across nine challenging visual continuous control tasks from Deepmind Control Suite. In addition, we show that TACO can also serve as a plug-and-play module adding to existing offline visual RL methods to establish the new state-of-the-art performance for offline visual RL across offline datasets with varying quality."
                },
                {
                    "title": "Accelerating Reinforcement Learning with Value-Conditional State Entropy Exploration",
                    "abstract": "A promising technique for exploration is to maximize the entropy of visited state distribution, i.e., state entropy, by encouraging uniform coverage of visited state space. While it has been effective for an unsupervised setup, it tends to struggle in a supervised setup with a task reward, where an agent prefers to visit high-value states to exploit the task reward. Such a preference can cause an imbalance between the distributions of high-value states and low-value states, which biases exploration towards low-value state regions as a result of the state entropy increasing when the distribution becomes more uniform. This issue is exacerbated when high-value states are narrowly distributed within the state space, making it difficult for the agent to complete the tasks. In this paper, we present a novel exploration technique that maximizes the value-conditional state entropy, which separately estimates the state entropies that are conditioned on the value estimates of each state, then maximizes their average. By only considering the visited states with similar value estimates for computing the intrinsic bonus, our method prevents the distribution of low-value states from affecting exploration around high-value states, and vice versa. We demonstrate that the proposed alternative to the state entropy baseline significantly accelerates various reinforcement learning algorithms across a variety of tasks within MiniGrid, DeepMind Control Suite, and Meta-World benchmarks. Source code is available at https://sites.google.com/view/rl-vcse."
                },
                {
                    "title": "Hierarchical State Abstraction Based on Structural Information Principles",
                    "abstract": "State abstraction optimizes decision-making by ignoring irrelevant environmental information in reinforcement learning with rich observations. Nevertheless, recent approaches focus on adequate representational capacities resulting in essential information loss, affecting their performances on challenging tasks. In this article, we propose a novel mathematical Structural Information principles-based State Abstraction framework, namely SISA, from the information-theoretic perspective. Specifically, an unsupervised, adaptive hierarchical state clustering method without requiring manual assistance is presented, and meanwhile, an optimal encoding tree is generated. On each non-root tree node, a new aggregation function and condition structural entropy are designed to achieve hierarchical state abstraction and compensate for sampling-induced essential information loss in state abstraction. Empirical evaluations on a visual gridworld domain and six continuous control benchmarks demonstrate that, compared with five SOTA state abstraction approaches, SISA significantly improves mean episode reward and sample efficiency up to 18.98 and 44.44%, respectively. Besides, we experimentally show that SISA is a general framework that can be flexibly integrated with different representation-learning objectives to improve their performances further."
                },
                {
                    "title": "DEIR: Efficient and Robust Exploration through Discriminative-Model-Based Episodic Intrinsic Rewards",
                    "abstract": "Exploration is a fundamental aspect of reinforcement learning (RL), and its effectiveness is a deciding factor in the performance of RL algorithms, especially when facing sparse extrinsic rewards. Recent studies have shown the effectiveness of encouraging exploration with intrinsic rewards estimated from novelties in observations. However, there is a gap between the novelty of an observation and an exploration, as both the stochasticity in the environment and the agent's behavior may affect the observation. To evaluate exploratory behaviors accurately, we propose DEIR, a novel method in which we theoretically derive an intrinsic reward with a conditional mutual information term that principally scales with the novelty contributed by agent explorations, and then implement the reward with a discriminative forward model. Extensive experiments on both standard and advanced exploration tasks in MiniGrid show that DEIR quickly learns a better policy than the baselines. Our evaluations on ProcGen demonstrate both the generalization capability and the general applicability of our intrinsic reward."
                },
                {
                    "title": "Effective and Stable Role-Based Multi-Agent Collaboration by Structural Information Principles",
                    "abstract": "Role-based learning is a promising approach to improving the performance of Multi-Agent Reinforcement Learning (MARL). Nevertheless, without manual assistance, current role-based methods cannot guarantee stably discovering a set of roles to effectively decompose a complex task, as they assume either a predefined role structure or practical experience for selecting hyperparameters. In this article, we propose a mathematical Structural Information principles-based Role Discovery method, namely SIRD, and then present a SIRD optimizing MARL framework, namely SR-MARL, for multi-agent collaboration. The SIRD transforms role discovery into a hierarchical action space clustering. Specifically, the SIRD consists of structuralization, sparsification, and optimization modules, where an optimal encoding tree is generated to perform abstracting to discover roles. The SIRD is agnostic to specific MARL algorithms and flexibly integrated with various value function factorization approaches. Empirical evaluations on the StarCraft II micromanagement benchmark demonstrate that, compared with state-of-the-art MARL algorithms, the SR-MARL framework improves the average test win rate by 0.17%, 6.08%, and 3.24%, and reduces the deviation by 16.67%, 30.80%, and 66.30%, under easy, hard, and super hard scenarios."
                },
                {
                    "title": "SE-GSL: A General and Effective Graph Structure Learning Framework through Structural Entropy Optimization",
                    "abstract": "Graph Neural Networks (GNNs) are de facto solutions to structural data learning. However, it is susceptible to low-quality and unreliable structure, which has been a norm rather than an exception in real-world graphs. Existing graph structure learning (GSL) frameworks still lack robustness and interpretability. This paper proposes a general GSL framework, SE-GSL, through structural entropy and the graph hierarchy abstracted in the encoding tree. Particularly, we exploit the one-dimensional structural entropy to maximize embedded information content when auxiliary neighbourhood attributes is fused to enhance the original graph. A new scheme of constructing optimal encoding trees are proposed to minimize the uncertainty and noises in the graph whilst assuring proper community partition in hierarchical abstraction. We present a novel sample-based mechanism for restoring the graph structure via node structural entropy distribution. It increases the connectivity among nodes with larger uncertainty in lower-level communities. SE-GSL is compatible with various GNN models and enhances the robustness towards noisy and heterophily structures. Extensive experiments show significant improvements in the effectiveness and robustness of structure learning and node representation learning."
                },
                {
                    "title": "Minimum Entropy Principle Guided Graph Neural Networks",
                    "abstract": "Graph neural networks (GNNs) are now the mainstream method for mining graph-structured data and learning low-dimensional node- and graph-level embeddings to serve downstream tasks. However, limited by the bottleneck of interpretability that deep neural networks present, existing GNNs have ignored the issue of estimating the appropriate number of dimensions for the embeddings. Hence, we propose a novel framework called Minimum Graph Entropy principle-guided Dimension Estimation, i.e. MGEDE, that learns the appropriate embedding dimensions for both node and graph representations. In terms of node-level estimation, a minimum entropy function that counts both structure and attribute entropy, appraises the appropriate number of dimensions. In terms of graph-level estimation, each graph is assigned a customized embedding dimension from a candidate set based on the number of dimensions estimated for the node-level embeddings. Comprehensive experiments with node and graph classification tasks and nine benchmark datasets verify the effectiveness and generalizability of MGEDE."
                },
                {
                    "title": "LECO: Learnable Episodic Count for Task-Specific Intrinsic Reward",
                    "abstract": "Episodic count has been widely used to design a simple yet effective intrinsic motivation for reinforcement learning with a sparse reward. However, the use of episodic count in a high-dimensional state space as well as over a long episode time requires a thorough state compression and fast hashing, which hinders rigorous exploitation of it in such hard and complex exploration environments. Moreover, the interference from task-irrelevant observations in the episodic count may cause its intrinsic motivation to overlook task-related important changes of states, and the novelty in an episodic manner can lead to repeatedly revisit the familiar states across episodes. In order to resolve these issues, in this paper, we propose a learnable hash-based episodic count, which we name LECO, that efficiently performs as a task-specific intrinsic reward in hard exploration problems. In particular, the proposed intrinsic reward consists of the episodic novelty and the task-specific modulation where the former employs a vector quantized variational autoencoder to automatically obtain the discrete state codes for fast counting while the latter regulates the episodic novelty by learning a modulator to optimize the task-specific extrinsic reward. The proposed LECO specifically enables the automatic transition from exploration to exploitation during reinforcement learning. We experimentally show that in contrast to the previous exploration methods LECO successfully solves hard exploration problems and also scales to large state spaces through the most difficult tasks in MiniGrid and DMLab environments."
                },
                {
                    "title": "Masked World Models for Visual Control",
                    "abstract": "Visual model-based reinforcement learning (RL) has the potential to enable sample-efficient robot learning from visual observations. Yet the current approaches typically train a single model end-to-end for learning both visual representations and dynamics, making it difficult to accurately model the interaction between robots and small objects. In this work, we introduce a visual model-based RL framework that decouples visual representation learning and dynamics learning. Specifically, we train an autoencoder with convolutional layers and vision transformers (ViT) to reconstruct pixels given masked convolutional features, and learn a latent dynamics model that operates on the representations from the autoencoder. Moreover, to encode task-relevant information, we introduce an auxiliary reward prediction objective for the autoencoder. We continually update both autoencoder and dynamics model using online samples collected from environment interaction. We demonstrate that our decoupling approach achieves state-of-the-art performance on a variety of visual robotic tasks from Meta-world and RLBench, e.g., we achieve 81.7% success rate on 50 visual robotic manipulation tasks from Meta-world, while the baseline achieves 67.9%. Code is available on the project website: https://sites.google.com/view/mwm-rl."
                },
                {
                    "title": "Structural Entropy Guided Graph Hierarchical Pooling",
                    "abstract": "Following the success of convolution on non-Euclidean space, the corresponding pooling approaches have also been validated on various tasks regarding graphs. However, because of the fixed compression quota and stepwise pooling design, these hierarchical pooling methods still suffer from local structure damage and suboptimal problem. In this work, inspired by structural entropy, we propose a hierarchical pooling approach, SEP, to tackle the two issues. Specifically, without assigning the layer-specific compression quota, a global optimization algorithm is designed to generate the cluster assignment matrices for pooling at once. Then, we present an illustration of the local structure damage from previous methods in the reconstruction of ring and grid synthetic graphs. In addition to SEP, we further design two classification models, SEP-G and SEP-N for graph classification and node classification, respectively. The results show that SEP outperforms state-of-the-art graph pooling methods on graph classification benchmarks and obtains superior performance on node classifications."
                },
                {
                    "title": "R\u00e9nyi State Entropy Maximization for Exploration Acceleration in Reinforcement Learning",
                    "abstract": "One of the most critical challenges in deep reinforcement learning is to maintain the long-term exploration capability of the agent. To tackle this problem, it has been recently proposed to provide intrinsic rewards for the agent to encourage exploration. However, most existing intrinsic reward-based methods proposed in the literature fail to provide sustainable exploration incentives, a problem known as vanishing rewards. In addition, these conventional methods incur complex models and additional memory in their learning procedures, resulting in high computational complexity and low robustness. In this work, a novel intrinsic reward module based on the R\u00e9nyi entropy is proposed to provide high-quality intrinsic rewards. It is shown that the proposed method actually generalizes the existing state entropy maximization methods. In particular, a <inline-formula><tex-math notation=\"LaTeX\">$k$</tex-math></inline-formula>-nearest neighbor estimator is introduced for entropy estimation while a <inline-formula><tex-math notation=\"LaTeX\">$k$</tex-math></inline-formula>-value search method is designed to guarantee the estimation accuracy. Extensive simulation results demonstrate that the proposed R\u00e9nyi entropy-based method can achieve higher performance as compared to existing schemes."
                },
                {
                    "title": "The Importance of Non-Markovianity in Maximum State Entropy Exploration",
                    "abstract": "In the maximum state entropy exploration framework, an agent interacts with a reward-free environment to learn a policy that maximizes the entropy of the expected state visitations it is inducing. Hazan et al. (2019) noted that the class of Markovian stochastic policies is sufficient for the maximum state entropy objective, and exploiting non-Markovianity is generally considered pointless in this setting. In this paper, we argue that non-Markovianity is instead paramount for maximum state entropy exploration in a finite-sample regime. Especially, we recast the objective to target the expected entropy of the induced state visitations in a single trial. Then, we show that the class of non-Markovian deterministic policies is sufficient for the introduced objective, while Markovian policies suffer non-zero regret in general. However, we prove that the problem of finding an optimal non-Markovian policy is NP-hard. Despite this negative result, we discuss avenues to address the problem in a tractable way and how non-Markovian exploration could benefit the sample efficiency of online reinforcement learning in future works."
                },
                {
                    "title": "A Hybrid Deep Reinforcement Learning For Autonomous Vehicles Smart-Platooning",
                    "abstract": "The development of Autonomous Vehicles (AVs) envisions the promising technology of future Intelligent Transportation Systems (ITS). However, the complex road structures and increased vehicles cause traffic congestion and road safety, which eventually leads to horrible accidents. Cooperative driving of AVs, a groundbreaking initiative of vehicle platooning, epitomizes the next wave in vehicular technology through minimizing accident risks, transport times, costs, energy, and fuel consumption. However, the traditional machine learning-based platooning approaches fail to regulate the policy with the dynamic feature of AVs. This paper proposes a hybrid Deep Reinforcement learning and Genetic algorithm for Smart-Platooning (DRG-SP) the AVs. The leverage of the deep reinforcement learning mechanism addresses the computational complexity and accommodates the high dynamic platoon environments. Adopting the Genetic Algorithm in Deep Reinforcement learning overcomes the slow convergence problem and offers long-term performance. The simulation results reveal that the Smart-Platooning effectively forms and maintains the platoons by minimizing traffic congestion and fuel consumption."
                },
                {
                    "title": "URLB: Unsupervised Reinforcement Learning Benchmark",
                    "abstract": "Deep Reinforcement Learning (RL) has emerged as a powerful paradigm to solve a range of complex yet specific control tasks. Yet training generalist agents that can quickly adapt to new tasks remains an outstanding challenge. Recent advances in unsupervised RL have shown that pre-training RL agents with self-supervised intrinsic rewards can result in efficient adaptation. However, these algorithms have been hard to compare and develop due to the lack of a unified benchmark. To this end, we introduce the Unsupervised Reinforcement Learning Benchmark (URLB). URLB consists of two phases: reward-free pre-training and downstream task adaptation with extrinsic rewards. Building on the DeepMind Control Suite, we provide twelve continuous control tasks from three domains for evaluation and open-source code for eight leading unsupervised RL methods. We find that the implemented baselines make progress but are not able to solve URLB and propose directions for future research."
                },
                {
                    "title": "Dynamic Bottleneck for Robust Self-Supervised Exploration",
                    "abstract": "Exploration methods based on pseudo-count of transitions or curiosity of dynamics have achieved promising results in solving reinforcement learning with sparse rewards. However, such methods are usually sensitive to environmental dynamics-irrelevant information, e.g., white-noise. To handle such dynamics-irrelevant information, we propose a Dynamic Bottleneck (DB) model, which attains a dynamics-relevant representation based on the information-bottleneck principle. Based on the DB model, we further propose DB-bonus, which encourages the agent to explore state-action pairs with high information gain. We establish theoretical connections between the proposed DB-bonus, the upper confidence bound (UCB) for linear case, and the visiting count for tabular case. We evaluate the proposed method on Atari suits with dynamics-irrelevant noises. Our experiments show that exploration with DB bonus outperforms several state-of-the-art exploration methods in noisy environments."
                },
                {
                    "title": "An Information-theoretic Perspective of Hierarchical Clustering",
                    "abstract": "A combinatorial cost function for hierarchical clustering was introduced by Dasgupta \\cite{dasgupta2016cost}. It has been generalized by Cohen-Addad et al. \\cite{cohen2019hierarchical} to a general form named admissible function. In this paper, we investigate hierarchical clustering from the \\emph{information-theoretic} perspective and formulate a new objective function. We also establish the relationship between these two perspectives. In algorithmic aspect, we get rid of the traditional top-down and bottom-up frameworks, and propose a new one to stratify the \\emph{sparsest} level of a cluster tree recursively in guide with our objective function. For practical use, our resulting cluster tree is not binary. Our algorithm called HCSE outputs a $k$-level cluster tree by a novel and interpretable mechanism to choose $k$ automatically without any hyper-parameter. Our experimental results on synthetic datasets show that HCSE has a great advantage in finding the intrinsic number of hierarchies, and the results on real datasets show that HCSE also achieves competitive costs over the popular algorithms LOUVAIN and HLP."
                },
                {
                    "title": "Mastering Visual Continuous Control: Improved Data-Augmented Reinforcement Learning",
                    "abstract": "We present DrQ-v2, a model-free reinforcement learning (RL) algorithm for visual continuous control. DrQ-v2 builds on DrQ, an off-policy actor-critic approach that uses data augmentation to learn directly from pixels. We introduce several improvements that yield state-of-the-art results on the DeepMind Control Suite. Notably, DrQ-v2 is able to solve complex humanoid locomotion tasks directly from pixel observations, previously unattained by model-free RL. DrQ-v2 is conceptually simple, easy to implement, and provides significantly better computational footprint compared to prior work, with the majority of tasks taking just 8 hours to train on a single GPU. Finally, we publicly release DrQ-v2's implementation to provide RL practitioners with a strong and computationally efficient baseline."
                },
                {
                    "title": "MADE: Exploration via Maximizing Deviation from Explored Regions",
                    "abstract": "In online reinforcement learning (RL), efficient exploration remains particularly challenging in high-dimensional environments with sparse rewards. In low-dimensional environments, where tabular parameterization is possible, count-based upper confidence bound (UCB) exploration methods achieve minimax near-optimal rates. However, it remains unclear how to efficiently implement UCB in realistic RL tasks that involve non-linear function approximation. To address this, we propose a new exploration approach via \\textit{maximizing} the deviation of the occupancy of the next policy from the explored regions. We add this term as an adaptive regularizer to the standard RL objective to balance exploration vs. exploitation. We pair the new objective with a provably convergent algorithm, giving rise to a new intrinsic reward that adjusts existing bonuses. The proposed intrinsic reward is easy to implement and combine with other existing RL algorithms to conduct exploration. As a proof of concept, we evaluate the new intrinsic reward on tabular examples across a variety of model-based and model-free algorithms, showing improvements over count-only exploration strategies. When tested on navigation and locomotion tasks from MiniGrid and DeepMind Control Suite benchmarks, our approach significantly improves sample efficiency over state-of-the-art methods. Our code is available at https://github.com/tianjunz/MADE."
                },
                {
                    "title": "Behavior From the Void: Unsupervised Active Pre-Training",
                    "abstract": "We introduce a new unsupervised pre-training method for reinforcement learning called APT, which stands for Active Pre-Training. APT learns behaviors and representations by actively searching for novel states in reward-free environments. The key novel idea is to explore the environment by maximizing a non-parametric entropy computed in an abstract representation space, which avoids challenging density modeling and consequently allows our approach to scale much better in environments that have high-dimensional observations (e.g., image observations). We empirically evaluate APT by exposing task-specific reward after a long unsupervised pre-training phase. In Atari games, APT achieves human-level performance on 12 games and obtains highly competitive performance compared to canonical fully supervised RL algorithms. On DMControl suite, APT beats all baselines in terms of asymptotic performance and data efficiency and dramatically improves performance on tasks that are extremely difficult to train from scratch."
                },
                {
                    "title": "State Entropy Maximization with Random Encoders for Efficient Exploration",
                    "abstract": "Recent exploration methods have proven to be a recipe for improving sample-efficiency in deep reinforcement learning (RL). However, efficient exploration in high-dimensional observation spaces still remains a challenge. This paper presents Random Encoders for Efficient Exploration (RE3), an exploration method that utilizes state entropy as an intrinsic reward. In order to estimate state entropy in environments with high-dimensional observations, we utilize a k-nearest neighbor entropy estimator in the low-dimensional representation space of a convolutional encoder. In particular, we find that the state entropy can be estimated in a stable and compute-efficient manner by utilizing a randomly initialized encoder, which is fixed throughout training. Our experiments show that RE3 significantly improves the sample-efficiency of both model-free and model-based RL methods on locomotion and navigation tasks from DeepMind Control Suite and MiniGrid benchmarks. We also show that RE3 allows learning diverse behaviors without extrinsic rewards, effectively improving sample-efficiency in downstream tasks. Source code and videos are available at https://sites.google.com/view/re3-rl."
                },
                {
                    "title": "Novelty Search in representational space for sample efficient exploration",
                    "abstract": "We present a new approach for efficient exploration which leverages a low-dimensional encoding of the environment learned with a combination of model-based and model-free objectives. Our approach uses intrinsic rewards that are based on the distance of nearest neighbors in the low dimensional representational space to gauge novelty. We then leverage these intrinsic rewards for sample-efficient exploration with planning routines in representational space for hard exploration tasks with sparse rewards. One key element of our approach is the use of information theoretic principles to shape our representations in a way so that our novelty reward goes beyond pixel similarity. We test our approach on a number of maze tasks, as well as a control problem and show that our exploration approach is more sample-efficient compared to strong baselines."
                },
                {
                    "title": "Exploration by Maximizing Renyi Entropy for Reward-Free RL Framework",
                    "abstract": "Exploration is essential for reinforcement learning (RL). To face the challenges of exploration, we consider a reward-free RL framework that completely separates exploration from exploitation and brings new challenges for exploration algorithms. In the exploration phase, the agent learns an exploratory policy by interacting with a reward-free environment and collects a dataset of transitions by executing the policy. In the planning phase, the agent computes a good policy for any reward function based on the dataset without further interacting with the environment. This framework is suitable for the meta RL setting where there are many reward functions of interest. In the exploration phase, we propose to maximize the Renyi entropy over the state-action space and justify this objective theoretically. The success of using Renyi entropy as the objective results from its encouragement to explore the hard-to-reach state-actions. We further deduce a policy gradient formulation for this objective and design a practical exploration algorithm that can deal with complex environments. In the planning phase, we solve for good policies given arbitrary reward functions using a batch RL algorithm. Empirically, we show that our exploration algorithm is effective and sample efficient, and results in superior policies for arbitrary reward functions in the planning phase."
                },
                {
                    "title": "Agent57: Outperforming the Atari Human Benchmark",
                    "abstract": "Atari games have been a long-standing benchmark in the reinforcement learning (RL) community for the past decade. This benchmark was proposed to test general competency of RL algorithms. Previous work has achieved good average performance by doing outstandingly well on many games of the set, but very poorly in several of the most challenging games. We propose Agent57, the first deep RL agent that outperforms the standard human benchmark on all 57 Atari games. To achieve this result, we train a neural network which parameterizes a family of policies ranging from very exploratory to purely exploitative. We propose an adaptive mechanism to choose which policy to prioritize throughout the training process. Additionally, we utilize a novel parameterization of the architecture that allows for more consistent and stable learning."
                },
                {
                    "title": "Entropy Regularization with Discounted Future State Distribution in Policy Gradient Methods",
                    "abstract": "The policy gradient theorem is defined based on an objective with respect to the initial distribution over states. In the discounted case, this results in policies that are optimal for one distribution over initial states, but may not be uniformly optimal for others, no matter where the agent starts from. Furthermore, to obtain unbiased gradient estimates, the starting point of the policy gradient estimator requires sampling states from a normalized discounted weighting of states. However, the difficulty of estimating the normalized discounted weighting of states, or the stationary state distribution, is quite well-known. Additionally, the large sample complexity of policy gradient methods is often attributed to insufficient exploration, and to remedy this, it is often assumed that the restart distribution provides sufficient exploration in these algorithms. In this work, we propose exploration in policy gradient methods based on maximizing entropy of the discounted future state distribution. The key contribution of our work includes providing a practically feasible algorithm to estimate the normalized discounted weighting of states, i.e, the \\textit{discounted future state distribution}. We propose that exploration can be achieved by entropy regularization with the discounted state distribution in policy gradients, where a metric for maximal coverage of the state space can be based on the entropy of the induced state distribution. The proposed approach can be considered as a three time-scale algorithm and under some mild technical conditions, we prove its convergence to a locally optimal policy. Experimentally, we demonstrate usefulness of regularization with the discounted future state distribution in terms of increased state space coverage and faster learning on a range of complex tasks."
                },
                {
                    "title": "Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning",
                    "abstract": "Meta-reinforcement learning algorithms can enable robots to acquire new skills much more quickly, by leveraging prior experience to learn how to learn. However, much of the current research on meta-reinforcement learning focuses on task distributions that are very narrow. For example, a commonly used meta-reinforcement learning benchmark uses different running velocities for a simulated robot as different tasks. When policies are meta-trained on such narrow task distributions, they cannot possibly generalize to more quickly acquire entirely new tasks. Therefore, if the aim of these methods is to enable faster acquisition of entirely new behaviors, we must evaluate them on task distributions that are sufficiently broad to enable generalization to new behaviors. In this paper, we propose an open-source simulated benchmark for meta-reinforcement learning and multi-task learning consisting of 50 distinct robotic manipulation tasks. Our aim is to make it possible to develop algorithms that generalize to accelerate the acquisition of entirely new, held-out tasks. We evaluate 6 state-of-the-art meta-reinforcement learning and multi-task learning algorithms on these tasks. Surprisingly, while each task and its variations (e.g., with different object positions) can be learned with reasonable success, these algorithms struggle to learn with multiple tasks at the same time, even with as few as ten distinct training tasks. Our analysis and open-source environments pave the way for future research in multi-task learning and meta-learning that can enable meaningful generalization, thereby unlocking the full potential of these methods."
                },
                {
                    "title": "Curiosity-Bottleneck: Exploration By Distilling Task-Specific Novelty",
                    "abstract": "Exploration based on state novelty has brought great success in challenging reinforcement learning problems with sparse rewards. However, existing novelty-based strategies become inefficient in real-world problems where observation contains not only task-dependent state novelty of our interest but also task-irrelevant information that should be ignored. We introduce an informationtheoretic exploration strategy named CuriosityBottleneck that distills task-relevant information from observation. Based on the information bottleneck principle, our exploration bonus is quantified as the compressiveness of observation with respect to the learned representation of a compressive value network. With extensive experiments on static image classification, grid-world and three hard-exploration Atari games, we show that Curiosity-Bottleneck learns an effective exploration strategy by robustly measuring the state novelty in distractive environments where stateof-the-art exploration methods often degenerate."
                },
                {
                    "title": "Provably Efficient Maximum Entropy Exploration",
                    "abstract": "Suppose an agent is in a (possibly unknown) Markov Decision Process in the absence of a reward signal, what might we hope that an agent can efficiently learn to do? This work studies a broad class of objectives that are defined solely as functions of the state-visitation frequencies that are induced by how the agent behaves. For example, one natural, intrinsically defined, objective problem is for the agent to learn a policy which induces a distribution over state space that is as uniform as possible, which can be measured in an entropic sense. We provide an efficient algorithm to optimize such such intrinsically defined objectives, when given access to a black box planning oracle (which is robust to function approximation). Furthermore, when restricted to the tabular setting where we have sample based access to the MDP, our proposed algorithm is provably efficient, both in terms of its sample and computational complexities. Key to our algorithmic methodology is utilizing the conditional gradient method (a.k.a. the Frank-Wolfe algorithm) which utilizes an approximate MDP solver."
                },
                {
                    "title": "EMI: Exploration with Mutual Information",
                    "abstract": "Reinforcement learning algorithms struggle when the reward signal is very sparse. In these cases, naive random exploration methods essentially rely on a random walk to stumble onto a rewarding state. Recent works utilize intrinsic motivation to guide the exploration via generative models, predictive forward models, or discriminative modeling of novelty. We propose EMI, which is an exploration method that constructs embedding representation of states and actions that does not rely on generative decoding of the full observation but extracts predictive signals that can be used to guide exploration based on forward prediction in the representation space. Our experiments show competitive results on challenging locomotion tasks with continuous control and on image-based exploration tasks with discrete actions on Atari. The source code is available at this https URL ."
                },
                {
                    "title": "Representation Learning with Contrastive Predictive Coding",
                    "abstract": "While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments."
                },
                {
                    "title": "Maximum a Posteriori Policy Optimisation",
                    "abstract": "We introduce a new algorithm for reinforcement learning called Maximum aposteriori Policy Optimisation (MPO) based on coordinate ascent on a relative entropy objective. We show that several existing methods can directly be related to our derivation. We develop two off-policy algorithms and demonstrate that they are competitive with the state-of-the-art in deep reinforcement learning. In particular, for continuous control, our method outperforms existing methods with respect to sample efficiency, premature convergence and robustness to hyperparameter settings while achieving similar or better final performance."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "Hindsight Experience Replay",
                    "abstract": "Dealing with sparse rewards is one of the biggest challenges in Reinforcement Learning (RL). We present a novel technique called Hindsight Experience Replay which allows sample-efficient learning from rewards which are sparse and binary and therefore avoid the need for complicated reward engineering. It can be combined with an arbitrary off-policy RL algorithm and may be seen as a form of implicit curriculum. \nWe demonstrate our approach on the task of manipulating objects with a robotic arm. In particular, we run experiments on three different tasks: pushing, sliding, and pick-and-place, in each case using only binary rewards indicating whether or not the task is completed. Our ablation studies show that Hindsight Experience Replay is a crucial ingredient which makes training possible in these challenging environments. We show that our policies trained on a physics simulation can be deployed on a physical robot and successfully complete the task."
                },
                {
                    "title": "Curiosity-Driven Exploration by Self-Supervised Prediction",
                    "abstract": "In many real-world scenarios, rewards extrinsic to the agent are extremely sparse, or absent altogether. In such cases, curiosity can serve as an intrinsic reward signal to enable the agent to explore its environment and learn skills that might be useful later in its life. We formulate curiosity as the error in an agent's ability to predict the consequence of its own actions in a visual feature space learned by a self-supervised inverse dynamics model. Our formulation scales to high-dimensional continuous state spaces like images, bypasses the difficulties of directly predicting pixels, and, critically, ignores the aspects of the environment that cannot affect the agent. The proposed approach is evaluated in two environments: VizDoom and Super Mario Bros. Three broad settings are investigated: 1) sparse extrinsic reward; 2) exploration with no extrinsic reward; and 3) generalization to unseen scenarios (e.g. new levels of the same game)."
                },
                {
                    "title": "Bridging the Gap Between Value and Policy Based Reinforcement Learning",
                    "abstract": "We establish a new connection between value and policy based reinforcement learning (RL) based on a relationship between softmax temporal value consistency and policy optimality under entropy regularization. Specifically, we show that softmax consistent action values correspond to optimal entropy regularized policy probabilities along any action sequence, regardless of provenance. From this observation, we develop a new RL algorithm, Path Consistency Learning (PCL), that minimizes a notion of soft consistency error along multi-step action sequences extracted from both on- and off-policy traces. We examine the behavior of PCL in different scenarios and show that PCL can be interpreted as generalizing both actor-critic and Q-learning algorithms. We subsequently deepen the relationship by showing how a single model can be used to represent both a policy and the corresponding softmax state values, eliminating the need for a separate critic. The experimental evaluation demonstrates that PCL significantly outperforms strong actor-critic and Q-learning baselines across several benchmarks."
                },
                {
                    "title": "Reinforcement Learning with Deep Energy-Based Policies",
                    "abstract": "We propose a method for learning expressive energy-based policies for continuous states and actions, which has been feasible only in tabular domains before. We apply our method to learning maximum entropy policies, resulting into a new algorithm, called soft Q-learning, that expresses the optimal policy via a Boltzmann distribution. We use the recently proposed amortized Stein variational gradient descent to learn a stochastic sampling network that approximates samples from this distribution. The benefits of the proposed algorithm include improved exploration and compositionality that allows transferring skills between tasks, which we confirm in simulated experiments with swimming and walking robots. We also draw a connection to actor-critic methods, which can be viewed performing approximate inference on the corresponding energy-based model."
                },
                {
                    "title": "Structural Information and Dynamical Complexity of Networks",
                    "abstract": "In 1953, Shannon proposed the question of quantification of structural information to analyze communication systems. The question has become one of the longest great challenges in information science and computer science. Here, we propose the first metric for structural information. Given a graph G , we define the K-dimensional structural information of G (or structure entropy of G), denoted by HK(G) , to be the minimum overall number of bits required to determine the K-dimensional code of the node that is accessible from random walk in G. The K-dimensional structural information provides the principle for completely detecting the natural or true structure, which consists of the rules, regulations, and orders of the graphs, for fully distinguishing the order from disorder in structured noisy data, and for analyzing communication systems, solving the Shannon's problem and opening up new directions. The K-dimensional structural information is also the first metric of dynamical complexity of networks, measuring the complexity of interactions, communications, operations, and even evolution of networks. The metric satisfies a number of fundamental properties, including additivity, locality, robustness, local and incremental computability, and so on. We establish the fundamental theorems of the one- and two-dimensional structural information of networks, including both lower and upper bounds of the metrics of classic data structures, general graphs, the networks of models, and the networks of natural evolution. We propose algorithms to approximate the K-dimensional structural information of graphs by finding the K-dimensional structure of the graphs that minimizes the K-dimensional structure entropy. We find that the K-dimensional structure entropy minimization is the principle for detecting the natural or true structures in real-world networks. Consequently, our structural information provides the foundation for knowledge discovering from noisy data. We establish a black hole principle by using the two-dimensional structure information of graphs. We propose the natural rank of locally listing algorithms by the structure entropy minimization principle, providing the basis for a next-generation search engine."
                },
                {
                    "title": "Asynchronous Methods for Deep Reinforcement Learning",
                    "abstract": "We propose a conceptually simple and lightweight framework for deep reinforcement learning that uses asynchronous gradient descent for optimization of deep neural network controllers. We present asynchronous variants of four standard reinforcement learning algorithms and show that parallel actor-learners have a stabilizing effect on training allowing all four methods to successfully train neural network controllers. The best performing method, an asynchronous variant of actor-critic, surpasses the current state-of-the-art on the Atari domain while training for half the time on a single multi-core CPU instead of a GPU. Furthermore, we show that asynchronous actor-critic succeeds on a wide variety of continuous motor control problems as well as on a new task of navigating random 3D mazes using a visual input."
                },
                {
                    "title": "Nearest Neighbor Estimates of Entropy",
                    "abstract": "SYNOPTIC ABSTRACT Motivated by the problems in molecular sciences, we introduce new nonparametric estimators of entropy which are based on the kth nearest neighbor distances between the n sample points, where k (< n \u2013 1) is a fixed positive integer. These provide competing estimators to an estimator proposed by Kozachenko and Leonenko (1987), which is based on the first nearest neighbor distances of the sample points. These estimators are helpful in the evaluation of entropies of random vectors. We establish the asymptotic unbiasedness and consistency of the proposed estimators. For some standard distributions, we also investigate their performance for finite sample sizes using Monte Carlo simulations. The proposed estimators are applied to estimate the entropy of internal rotation in the methanol molecule, which can be characterized by a one-dimensional random vector, and of diethyl ether, which is described by a four-dimensional random vector."
                },
                {
                    "title": "The information bottleneck method",
                    "abstract": "We define the relevant information in a signal $x\\in X$ as being the information that this signal provides about another signal $y\\in \\Y$. Examples include the information that face images provide about the names of the people portrayed, or the information that speech sounds provide about the words spoken. Understanding the signal $x$ requires more than just predicting $y$, it also requires specifying which features of $\\X$ play a role in the prediction. We formalize this problem as that of finding a short code for $\\X$ that preserves the maximum information about $\\Y$. That is, we squeeze the information that $\\X$ provides about $\\Y$ through a `bottleneck' formed by a limited set of codewords $\\tX$. This constrained optimization problem can be seen as a generalization of rate distortion theory in which the distortion measure $d(x,\\x)$ emerges from the joint statistics of $\\X$ and $\\Y$. This approach yields an exact set of self consistent equations for the coding rules $X \\to \\tX$ and $\\tX \\to \\Y$. Solutions to these equations can be found by a convergent re-estimation method that generalizes the Blahut-Arimoto algorithm. Our variational principle provides a surprisingly rich framework for discussing a variety of problems in signal processing and learning, as will be described in detail elsewhere."
                },
                {
                    "title": "A Markovian Decision Process",
                    "abstract": "Abstract : The purpose of this paper is to discuss the asymptotic behavior of the sequence (f sub n(i)) generated by a nonlinear recurrence relation. This problem arises in connection with an equipment replacement problem, cf. S. Dreyfus, A Note on an Industrial Replacement Process."
                },
                {
                    "title": "The lattice theory of information",
                    "abstract": "The word \"information\" has been given many different meanings by various writers in the general field of information theory. It is likely that at least a number of these will prove sufficiently useful in certain applications to deserve further study and permanent recognition. It is hardly to be expected that a single concept of information would satisfactorily account for the numerous possible applications of this general field. The present note outlines a new approach to information theory which is aimed specifically at the analysis of certain communication problems in which there exist a number of information sources simultaneously in operation. A typical example is that of a simple communication channel with a feedback path from the receiving point to the transmitting point. The problem is to make use of the feedback information for improving forward transmission, and to determine the forward channel capacity when the best possible use is made of this feedback information. Another more general problem is that of a communication system consisting of a large number of transmitting and receiving points with some type of interconnection network between the various points. The problem here is to formulate the best systems design whereby, in some sense, the best overall use of the available facilities is made. While the analysis sketched here has not yet proceeded to the point or a complete solution of these problems, partial answers have been found and it is believed that a complete solution may be possible."
                },
                {
                    "title": "REM: From Structural Entropy to Community Structure Deception",
                    "abstract": "This paper focuses on the privacy risks of disclosing the community structure in an online social network. By exploiting the community affiliations of user accounts, an attacker may infer sensitive user attributes. This raises the problem of community structure deception (CSD), which asks for ways to minimally modify the network so that a given community structure maximally hides itself from community detection algorithms. We investigate CSD through an information-theoretic lens. To this end, we propose a community-based structural entropy to express the amount of information revealed by a community structure. This notion allows us to devise residual entropy minimization (REM) as an efficient procedure to solve CSD. Experimental results over 9 real-world networks and 6 community detection algorithms show that REM is very effective in obfuscating the community structure as compared to other benchmark methods."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we enhance exploration strategies in Reinforcement Learning by incorporating the inherent structure of state and action spaces to improve agent performance in high-dimensional and sparse reward environments?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses a significant limitation in current exploration strategies that often overlook the structural relationships within state and action spaces. By improving exploration methods, we can advance the development of more efficient and effective Reinforcement Learning algorithms, leading to better performance in complex tasks such as game intelligence, robotic control, and autonomous driving. This research could pave the way for practical applications that require robust decision-making in uncertain environments, ultimately contributing to the broader field of artificial intelligence.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the complexity of accurately modeling the inherent structure within state and action spaces, which is often overlooked in existing methods. Naive approaches may fail because they do not account for the relationships between states and actions, leading to inefficient exploration and suboptimal policies. Additionally, the high dimensionality and sparsity of rewards in many environments complicate the learning process, making it difficult to derive meaningful insights from the data. Overcoming these technical and theoretical obstacles requires innovative methodologies that can effectively capture and utilize the structural information present in the state-action space.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on maximizing entropy without adequately considering the underlying structure of state and action spaces. This oversight has resulted in exploration strategies that are biased towards low-value states, limiting their effectiveness. Barriers such as the lack of robust frameworks to incorporate structural information and the complexity of high-dimensional environments have prevented this problem from being solved. Our approach differs by explicitly integrating the inherent state-action structure into the exploration process, thereby enhancing the efficiency and effectiveness of the learning algorithms.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves developing a new exploration strategy that leverages the inherent structure of state and action spaces. We will utilize a modified Markov Decision Process (MDP) framework, focusing on optimizing state-action transitions based on structural entropy. The dataset will consist of simulated environments with varying complexities, and we will evaluate our approach using metrics such as cumulative reward and exploration efficiency. The expected outcomes include improved agent performance in terms of faster convergence to optimal policies and enhanced exploration capabilities, ultimately demonstrating the effectiveness of incorporating structural information in Reinforcement Learning."
            }
        },
        "author_data": {
            "709fe104-b71d-4483-8af7-fb4555833982": {
                "pk": "709fe104-b71d-4483-8af7-fb4555833982",
                "name": "Xianghua Zeng",
                "collaborators": [
                    "Hao Peng",
                    "Angsheng Li",
                    "Dingli Su",
                    "Chunyang Liu",
                    "Lifang He",
                    "Philip S. Yu"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Multi-Agent Systems",
                    "Adversarial Modeling",
                    "Information Theory"
                ],
                "publications": [
                    {
                        "title": "Effective and Stable Role-Based Multi-Agent Collaboration by Structural Information Principles",
                        "abstract": "Role-based learning is a promising approach to improving the performance of Multi-Agent Reinforcement Learning (MARL). Nevertheless, without manual assistance, current role-based methods cannot guarantee stably discovering a set of roles to effectively decompose a complex task, as they assume either a predefined role structure or practical experience for selecting hyperparameters. In this article, we propose a mathematical Structural Information principles-based Role Discovery method, namely SIRD, and then present a SIRD optimizing MARL framework, namely SR-MARL, for multi-agent collaboration. The SIRD transforms role discovery into a hierarchical action space clustering. Specifically, the SIRD consists of structuralization, sparsification, and optimization modules, where an optimal encoding tree is generated to perform abstracting to discover roles. The SIRD is agnostic to specific MARL algorithms and flexibly integrated with various value function factorization approaches. Empirical evaluations on the StarCraft II micromanagement benchmark demonstrate that, compared with state-of-the-art MARL algorithms, the SR-MARL framework improves the average test win rate by 0.17%, 6.08%, and 3.24%, and reduces the deviation by 16.67%, 30.80%, and 66.30%, under easy, hard, and super hard scenarios."
                    },
                    {
                        "title": "Adversarial Socialbots Modeling Based on Structural Information Principles",
                        "abstract": "The importance of effective detection is underscored by the fact that socialbots imitate human behavior to propagate misinformation, leading to an ongoing competition between socialbots and detectors. Despite the rapid advancement of reactive detectors, the exploration of adversarial socialbot modeling remains incomplete, significantly hindering the development of proactive detectors. To address this issue, we propose a mathematical Structural Information principles-based Adversarial Socialbots Modeling framework, namely SIASM, to enable more accurate and effective modeling of adversarial behaviors. First, a heterogeneous graph is presented to integrate various users and rich activities in the original social network and measure its dynamic uncertainty as structural entropy. By minimizing the high-dimensional structural entropy, a hierarchical community structure of the social network is generated and referred to as the optimal encoding tree. Secondly, a novel method is designed to quantify influence by utilizing the assigned structural entropy, which helps reduce the computational cost of SIASM by filtering out uninfluential users. Besides, a new conditional structural entropy is defined between the socialbot and other users to guide the follower selection for network influence maximization. Extensive and comparative experiments on both homogeneous and heterogeneous social networks demonstrate that, compared with state-of-the-art baselines, the proposed SIASM framework yields substantial performance improvements in terms of network influence (up to 16.32%) and sustainable stealthiness (up to 16.29%) when evaluated against a robust detector with 90% accuracy."
                    },
                    {
                        "title": "Effective Reinforcement Learning Based on Structural Information Principles",
                        "abstract": "Although Reinforcement Learning (RL) algorithms acquire sequential behavioral patterns through interactions with the environment, their effectiveness in noisy and high-dimensional scenarios typically relies on specific structural priors. In this paper, we propose a novel and general Structural Information principles-based framework for effective Decision-Making, namely SIDM, approached from an information-theoretic perspective. This paper presents a specific unsupervised partitioning method that forms vertex communities in the state and action spaces based on their feature similarities. An aggregation function, which utilizes structural entropy as the vertex weight, is devised within each community to obtain its embedding, thereby facilitating hierarchical state and action abstractions. By extracting abstract elements from historical trajectories, a directed, weighted, homogeneous transition graph is constructed. The minimization of this graph's high-dimensional entropy leads to the generation of an optimal encoding tree. An innovative two-layer skill-based learning mechanism is introduced to compute the common path entropy of each state transition as its identified probability, thereby obviating the requirement for expert knowledge. Moreover, SIDM can be flexibly incorporated into various single-agent and multi-agent RL algorithms, enhancing their performance. Finally, extensive evaluations on challenging benchmarks demonstrate that, compared with SOTA baselines, our framework significantly and consistently improves the policy's quality, stability, and efficiency up to 32.70%, 88.26%, and 64.86%, respectively."
                    },
                    {
                        "title": "Hierarchical State Abstraction Based on Structural Information Principles",
                        "abstract": "State abstraction optimizes decision-making by ignoring irrelevant environmental information in reinforcement learning with rich observations. Nevertheless, recent approaches focus on adequate representational capacities resulting in essential information loss, affecting their performances on challenging tasks. In this article, we propose a novel mathematical Structural Information principles-based State Abstraction framework, namely SISA, from the information-theoretic perspective. Specifically, an unsupervised, adaptive hierarchical state clustering method without requiring manual assistance is presented, and meanwhile, an optimal encoding tree is generated. On each non-root tree node, a new aggregation function and condition structural entropy are designed to achieve hierarchical state abstraction and compensate for sampling-induced essential information loss in state abstraction. Empirical evaluations on a visual gridworld domain and six continuous control benchmarks demonstrate that, compared with five SOTA state abstraction approaches, SISA significantly improves mean episode reward and sample efficiency up to 18.98 and 44.44%, respectively. Besides, we experimentally show that SISA is a general framework that can be flexibly integrated with different representation-learning objectives to improve their performances further."
                    }
                ]
            },
            "7abe3345-2e56-46ad-9b82-5e7a64f7530d": {
                "pk": "7abe3345-2e56-46ad-9b82-5e7a64f7530d",
                "name": "Hao Peng",
                "collaborators": [
                    "Yuan Qi",
                    "Xiaoli Bai",
                    "Qingsong Tang",
                    "Cailing Wang",
                    "Yuejian Peng",
                    "Roy Schwartz",
                    "Noah A. Smith"
                ],
                "domain": [
                    "Gaussian Processes",
                    "Machine Learning",
                    "Hypergraph Theory",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "EigenGP: Gaussian Process Models with Adaptive Eigenfunctions",
                        "abstract": "Gaussian processes (GPs) provide a nonparametric representation of functions. However, classical GP inference suffers from high computational cost for big data. In this paper, we propose a new Bayesian approach, EigenGP, that learns both basis dictionary elements--eigenfunctions of a GP prior--and prior precisions in a sparse finite model. It is well known that, among all orthogonal basis functions, eigenfunctions can provide the most compact representation. Unlike other sparse Bayesian finite models where the basis function has a fixed form, our eigenfunctions live in a reproducing kernel Hilbert space as a finite linear combination of kernel functions. We learn the dictionary elements--eigenfunctions--and the prior precisions over these elements as well as all the other hyperparameters from data by maximizing the model marginal likelihood. We explore computational linear algebra to simplify the gradient computation significantly. Our experimental results demonstrate improved predictive performance of EigenGP over alternative sparse GP methods as well as relevance vector machine."
                    },
                    {
                        "title": "Improving Orbit Prediction Accuracy through Supervised Machine Learning",
                        "abstract": "Due to the lack of information such as the space environment condition and resident space objects' (RSOs') body characteristics, current orbit predictions that are solely grounded on physics-based models may fail to achieve required accuracy for collision avoidance and have led to satellite collisions already. This paper presents a methodology to predict RSOs' trajectories with higher accuracy than that of the current methods. Inspired by the machine learning (ML) theory through which the models are learned based on large amounts of observed data and the prediction is conducted without explicitly modeling space objects and space environment, the proposed ML approach integrates physics-based orbit prediction algorithms with a learning-based process that focuses on reducing the prediction errors. Using a simulation-based space catalog environment as the test bed, the paper demonstrates three types of generalization capability for the proposed ML approach: 1) the ML model can be used to improve the same RSO's orbit information that is not available during the learning process but shares the same time interval as the training data; 2) the ML model can be used to improve predictions of the same RSO at future epochs; and 3) the ML model based on a RSO can be applied to other RSOs that share some common features."
                    },
                    {
                        "title": "On Frankl and Furedi's conjecture for 3-uniform hypergraphs",
                        "abstract": "The Lagrangian of a hypergraph has been a useful tool in hypergraph extremal problems. In most applications, we need an upper bound for the Lagrangian of a hypergraph. Frankl and Furedi in \\cite{FF} conjectured that the $r$-graph with $m$ edges formed by taking the first $m$ sets in the colex ordering of ${\\mathbb N}^{(r)}$ has the largest Lagrangian of all $r$-graphs with $m$ edges. In this paper, we give some partial results for this conjecture."
                    },
                    {
                        "title": "PaLM: A Hybrid Parser and Language Model",
                        "abstract": "We present PaLM, a hybrid parser and neural language model. Building on an RNN language model, PaLM adds an attention layer over text spans in the left context. An unsupervised constituency parser can be derived from its attention weights, using a greedy decoding algorithm. We evaluate PaLM on language modeling, and empirically show that it outperforms strong baselines. If syntactic annotations are available, the attention component can be trained in a supervised manner, providing syntactically-informed representations of the context, and further improving language modeling performance."
                    }
                ]
            },
            "0e82a04d-6357-45e7-830c-7ce03c7ab1f6": {
                "pk": "0e82a04d-6357-45e7-830c-7ce03c7ab1f6",
                "name": "Angsheng Li",
                "collaborators": [
                    "Yicheng Pan",
                    "Jiankou Li",
                    "Hao Peng",
                    "Pan Peng",
                    "Wei Zhang",
                    "Xianghua Zeng",
                    "George Barmpalias",
                    "Mingji Xia",
                    "Peng Zhang",
                    "Xuechen Li"
                ],
                "domain": [
                    "Network Science",
                    "Machine Learning",
                    "Graph Theory",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Structural Information Learning Machinery: Learning from Observing, Associating, Optimizing, Decoding, and Abstracting",
                        "abstract": "In the present paper, we propose the model of {\\it structural information learning machines} (SiLeM for short), leading to a mathematical definition of learning by merging the theories of computation and information. Our model shows that the essence of learning is {\\it to gain information}, that to gain information is {\\it to eliminate uncertainty} embedded in a data space, and that to eliminate uncertainty of a data space can be reduced to an optimization problem, that is, an {\\it information optimization problem}, which can be realized by a general {\\it encoding tree method}. The principle and criterion of the structural information learning machines are maximization of {\\it decoding information} from the data points observed together with the relationships among the data points, and semantical {\\it interpretation} of syntactical {\\it essential structure}, respectively. A SiLeM machine learns the laws or rules of nature. It observes the data points of real world, builds the {\\it connections} among the observed data and constructs a {\\it data space}, for which the principle is to choose the way of connections of data points so that the {\\it decoding information} of the data space is maximized, finds the {\\it encoding tree} of the data space that minimizes the dynamical uncertainty of the data space, in which the encoding tree is hence referred to as a {\\it decoder}, due to the fact that it has already eliminated the maximum amount of uncertainty embedded in the data space, interprets the {\\it semantics} of the decoder, an encoding tree, to form a {\\it knowledge tree}, extracts the {\\it remarkable common features} for both semantical and syntactical features of the modules decoded by a decoder to construct {\\it trees of abstractions}, providing the foundations for {\\it intuitive reasoning} in the learning when new data are observed."
                    },
                    {
                        "title": "Characters and patterns of communities in networks",
                        "abstract": "In this paper, we propose some new notions to characterize and analyze the communities. The new notions are general characters of the communities or local structures of networks. At first, we introduce the notions of internal dominating set and external dominating set of a community. We show that most communities in real networks have a small internal dominating set and a small external dominating set, and that the internal dominating set of a community keeps much of the information of the community. Secondly, based on the notions of the internal dominating set and the external dominating set, we define an internal slope (ISlope, for short) and an external slope (ESlope, for short) to measure the internal heterogeneity and external heterogeneity of a community respectively. We show that the internal slope (ISlope) of a community largely determines the structure of the community, that most communities in real networks are heterogeneous, meaning that most of the communities have a core/periphery structure, and that both ISlopes and ESlopes (reflecting the structure of communities) of all the communities of a network approximately follow a normal distribution. Therefore typical values of both ISolpes and ESoples of all the communities of a given network are in a narrow interval, and there is only a small number of communities having ISlopes or ESlopes out of the range of typical values of the ISlopes and ESlopes of the network. Finally, we show that all the communities of the real networks we studied, have a three degree separation phenomenon, that is, the average distance of communities is approximately 3, implying a general property of true communities for many real networks, and that good community finding algorithms find communities that amplify clustering coefficients of the networks, for many real networks."
                    },
                    {
                        "title": "The Small-Community Phenomenon in Networks",
                        "abstract": "We investigate several geometric models of network which simultaneously have some nice global properties, that the small diameter property, the small-community phenomenon, which is defined to capture the common experience that (almost) every one in our society belongs to some meaningful small communities by the authors (2011), and that under certain conditions on the parameters, the power law degree distribution, which significantly strengths the results given by van den Esker (2008), and Jordan (2010). The results above, together with our previous progress in Li and Peng (2011), build a mathematical foundation for the study of communities and the small-community phenomenon in various networks.   In the proof of the power law degree distribution, we develop the method of alternating concentration analysis to build concentration inequality by alternatively and iteratively applying both the sub- and super-martingale inequalities, which seems powerful, and which may have more potential applications."
                    },
                    {
                        "title": "A Theory for Valiant's Matchcircuits (Extended Abstract)",
                        "abstract": "The computational function of a matchgate is represented by its character matrix. In this article, we show that all nonsingular character matrices are closed under matrix inverse operation, so that for every $k$, the nonsingular character matrices of $k$-bit matchgates form a group, extending the recent work of Cai and Choudhary (2006) of the same result for the case of $k=2$, and that the single and the two-bit matchgates are universal for matchcircuits, answering a question of Valiant (2002)."
                    },
                    {
                        "title": "Kolmogorov complexity and computably enumerable sets",
                        "abstract": "We study the computably enumerable sets in terms of the: (a) Kolmogorov complexity of their initial segments; (b) Kolmogorov complexity of finite programs when they are used as oracles. We present an extended discussion of the existing research on this topic, along with recent developments and open problems. Besides this survey, our main original result is the following characterization of the computably enumerable sets with trivial initial segment prefix-free complexity. A computably enumerable set $A$ is $K$-trivial if and only if the family of sets with complexity bounded by the complexity of $A$ is uniformly computable from the halting problem."
                    },
                    {
                        "title": "Structure Entropy and Resistor Graphs",
                        "abstract": "We propose the notion of {\\it resistance of a graph} as an accompanying notion of the structure entropy to measure the force of the graph to resist cascading failure of strategic virus attacks. We show that for any connected network $G$, the resistance of $G$ is $\\mathcal{R}(G)=\\mathcal{H}^1(G)-\\mathcal{H}^2(G)$, where $\\mathcal{H}^1(G)$ and $\\mathcal{H}^2(G)$ are the one- and two-dimensional structure entropy of $G$, respectively. According to this, we define the notion of {\\it security index of a graph} to be the normalized resistance, that is, $\\theta (G)=\\frac{\\mathcal{R}(G)}{\\mathcal{H}^1(H)}$. We say that a connected graph is an $(n,\\theta)$-{\\it resistor graph}, if $G$ has $n$ vertices and has security index $\\theta (G)\\geq\\theta$. We show that trees and grid graphs are $(n,\\theta)$-resistor graphs for large constant $\\theta$, that the graphs with bounded degree $d$ and $n$ vertices, are $(n,\\frac{2}{d}-o(1))$-resistor graphs, and that for a graph $G$ generated by the security model \\cite{LLPZ2015, LP2016}, with high probability, $G$ is an $(n,\\theta)$-resistor graph, for a constant $\\theta$ arbitrarily close to $1$, provided that $n$ is sufficiently large. To the opposite side, we show that expander graphs are not good resistor graphs, in the sense that, there is a global constant $\\theta_0<1$ such that expander graphs cannot be $(n,\\theta)$-resistor graph for any $\\theta\\geq\\theta_0$. In particular, for the complete graph $G$, the resistance of $G$ is a constant $O(1)$, and hence the security index of $G$ is $\\theta (G)=o(1)$. Finally, we show that for any simple and connected graph $G$, if $G$ is an $(n, 1-o(1))$-resistor graph, then there is a large $k$ such that the $k$-th largest eigenvalue of the Laplacian of $G$ is $o(1)$, giving rise to an algebraic characterization for the graphs that are secure against intentional virus attack."
                    },
                    {
                        "title": "Detecting and Characterizing Small Dense Bipartite-like Subgraphs by the Bipartiteness Ratio Measure",
                        "abstract": "We study the problem of finding and characterizing subgraphs with small \\textit{bipartiteness ratio}. We give a bicriteria approximation algorithm \\verb|SwpDB| such that if there exists a subset $S$ of volume at most $k$ and bipartiteness ratio $\\theta$, then for any $0<\\epsilon<1/2$, it finds a set $S'$ of volume at most $2k^{1+\\epsilon}$ and bipartiteness ratio at most $4\\sqrt{\\theta/\\epsilon}$. By combining a truncation operation, we give a local algorithm \\verb|LocDB|, which has asymptotically the same approximation guarantee as the algorithm \\verb|SwpDB| on both the volume and bipartiteness ratio of the output set, and runs in time $O(\\epsilon^2\\theta^{-2}k^{1+\\epsilon}\\ln^3k)$, independent of the size of the graph. Finally, we give a spectral characterization of the small dense bipartite-like subgraphs by using the $k$th \\textit{largest} eigenvalue of the Laplacian of the graph."
                    },
                    {
                        "title": "Algorithmic Aspects of Homophyly of Networks",
                        "abstract": "We investigate the algorithmic problems of the {\\it homophyly phenomenon} in networks. Given an undirected graph $G = (V, E)$ and a vertex coloring $c \\colon V \\rightarrow {1, 2, ..., k}$ of $G$, we say that a vertex $v\\in V$ is {\\it happy} if $v$ shares the same color with all its neighbors, and {\\it unhappy}, otherwise, and that an edge $e\\in E$ is {\\it happy}, if its two endpoints have the same color, and {\\it unhappy}, otherwise. Supposing $c$ is a {\\it partial vertex coloring} of $G$, we define the Maximum Happy Vertices problem (MHV, for short) as to color all the remaining vertices such that the number of happy vertices is maximized, and the Maximum Happy Edges problem (MHE, for short) as to color all the remaining vertices such that the number of happy edges is maximized.   Let $k$ be the number of colors allowed in the problems. We show that both MHV and MHE can be solved in polynomial time if $k = 2$, and that both MHV and MHE are NP-hard if $k \\geq 3$. We devise a $\\max {1/k, \\Omega(\\Delta^{-3})}$-approximation algorithm for the MHV problem, where $\\Delta$ is the maximum degree of vertices in the input graph, and a 1/2-approximation algorithm for the MHE problem. This is the first theoretical progress of these two natural and fundamental new problems."
                    },
                    {
                        "title": "Testing Small Set Expansion in General Graphs",
                        "abstract": "We consider the problem of testing small set expansion for general graphs. A graph $G$ is a $(k,\\phi)$-expander if every subset of volume at most $k$ has conductance at least $\\phi$. Small set expansion has recently received significant attention due to its close connection to the unique games conjecture, the local graph partitioning algorithms and locally testable codes.   We give testers with two-sided error and one-sided error in the adjacency list model that allows degree and neighbor queries to the oracle of the input graph. The testers take as input an $n$-vertex graph $G$, a volume bound $k$, an expansion bound $\\phi$ and a distance parameter $\\varepsilon>0$. For the two-sided error tester, with probability at least $2/3$, it accepts the graph if it is a $(k,\\phi)$-expander and rejects the graph if it is $\\varepsilon$-far from any $(k^*,\\phi^*)$-expander, where $k^*=\\Theta(k\\varepsilon)$ and $\\phi^*=\\Theta(\\frac{\\phi^4}{\\min\\{\\log(4m/k),\\log n\\}\\cdot(\\ln k)})$. The query complexity and running time of the tester are $\\widetilde{O}(\\sqrt{m}\\phi^{-4}\\varepsilon^{-2})$, where $m$ is the number of edges of the graph. For the one-sided error tester, it accepts every $(k,\\phi)$-expander, and with probability at least $2/3$, rejects every graph that is $\\varepsilon$-far from $(k^*,\\phi^*)$-expander, where $k^*=O(k^{1-\\xi})$ and $\\phi^*=O(\\xi\\phi^2)$ for any $0<\\xi<1$. The query complexity and running time of this tester are $\\widetilde{O}(\\sqrt{\\frac{n}{\\varepsilon^3}}+\\frac{k}{\\varepsilon \\phi^4})$. We also give a two-sided error tester with smaller gap between $\\phi^*$ and $\\phi$ in the rotation map model that allows (neighbor, index) queries and degree queries."
                    },
                    {
                        "title": "Homophyly Networks -- A Structural Theory of Networks",
                        "abstract": "A grand challenge in network science is apparently the missing of a structural theory of networks. The authors have showed that the existence of community structures is a universal phenomenon in real networks, and that neither randomness nor preferential attachment is a mechanism of community structures of network \\footnote{A. Li, J. Li, and Y. Pan, Community structures are definable in networks, and universal in the real world, To appear.}. This poses a fundamental question: What are the mechanisms of community structures of real networks? Here we found that homophyly is the mechanism of community structures and a structural theory of networks. We proposed a homophyly model. It was shown that networks of our model satisfy a series of new topological, probabilistic and combinatorial principles, including a fundamental principle, a community structure principle, a degree priority principle, a widths principle, an inclusion and infection principle, a king node principle, and a predicting principle etc, leading to a structural theory of networks. Our model demonstrates that homophyly is the underlying mechanism of community structures of networks, that nodes of the same community share common features, that power law and small world property are never obstacles of the existence of community structures in networks, and that community structures are definable in networks."
                    },
                    {
                        "title": "Homophyly and Randomness Resist Cascading Failure in Networks",
                        "abstract": "The universal properties of power law and small world phenomenon of networks seem unavoidably obstacles for security of networking systems. Existing models never give secure networks. We found that the essence of security is the security against cascading failures of attacks and that nature solves the security by mechanisms. We proposed a model of networks by the natural mechanisms of homophyly, randomness and preferential attachment. It was shown that homophyly creates a community structure, that homophyly and randomness introduce ordering in the networks, and that homophyly creates inclusiveness and introduces rules of infections. These principles allow us to provably guarantee the security of the networks against any attacks. Our results show that security can be achieved provably by structures, that there is a tradeoff between the roles of structures and of thresholds in security engineering, and that power law and small world property are never obstacles for security of networks."
                    },
                    {
                        "title": "Community Structures Are Definable in Networks, and Universal in Real World",
                        "abstract": "Community detecting is one of the main approaches to understanding networks \\cite{For2010}.   However it has been a longstanding challenge to give a definition for community structures of networks. Here we found that community structures are definable in networks, and are universal in real world. We proposed the notions of entropy- and conductance-community structure ratios. It was shown that the definitions of the modularity proposed in \\cite{NG2004}, and our entropy- and conductance-community structures are equivalent in defining community structures of networks, that randomness in the ER model \\cite{ER1960} and preferential attachment in the PA \\cite{Bar1999} model are not mechanisms of community structures of networks, and that the existence of community structures is a universal phenomenon in real networks. Our results demonstrate that community structure is a universal phenomenon in the real world that is definable, solving the challenge of definition of community structures in networks. This progress provides a foundation for a structural theory of networks."
                    },
                    {
                        "title": "Dimensions, Structures and Security of Networks",
                        "abstract": "One of the main issues in modern network science is the phenomenon of cascading failures of a small number of attacks. Here we define the dimension of a network to be the maximal number of functions or features of nodes of the network. It was shown that there exist linear networks which are provably secure, where a network is linear, if it has dimension one, that the high dimensions of networks are the mechanisms of overlapping communities, that overlapping communities are obstacles for network security, and that there exists an algorithm to reduce high dimensional networks to low dimensional ones which simultaneously preserves all the network properties and significantly amplifies security of networks. Our results explore that dimension is a fundamental measure of networks, that there exist linear networks which are provably secure, that high dimensional networks are insecure, and that security of networks can be amplified by reducing dimensions."
                    },
                    {
                        "title": "Community Structures Are Definable in Networks: A Structural Theory of Networks",
                        "abstract": "We found that neither randomness in the ER model nor the preferential attachment in the PA model is the mechanism of community structures of networks, that community structures are universal in real networks, that community structures are definable in networks, that communities are interpretable in networks, and that homophyly is the mechanism of community structures and a structural theory of networks. We proposed the notions of entropy- and conductance-community structures. It was shown that the two definitions of the entropy- and conductance-community structures and the notion of modularity proposed by physicists are all equivalent in defining community structures of networks, that neither randomness in the ER model nor preferential attachment in the PA model is the mechanism of community structures of networks, and that the existence of community structures is a universal phenomenon in real networks. This poses a fundamental question: What are the mechanisms of community structures of real networks? To answer this question, we proposed a homophyly model of networks. It was shown that networks of our model satisfy a series of new topological, probabilistic and combinatorial principles, including a fundamental principle, a community structure principle, a degree priority principle, a widths principle, an inclusion and infection principle, a king node principle and a predicting principle etc. The new principles provide a firm foundation for a structural theory of networks. Our homophyly model demonstrates that homophyly is the underlying mechanism of community structures of networks, that nodes of the same community share common features, that power law and small world property are never obstacles of the existence of community structures in networks, that community structures are {\\it definable} in networks, and that (natural) communities are {\\it interpretable}."
                    },
                    {
                        "title": "Provable Security of Networks",
                        "abstract": "We propose a definition of {\\it security} and a definition of {\\it robustness} of networks against the cascading failure models of deliberate attacks and random errors respectively, and investigate the principles of the security and robustness of networks. We propose a {\\it security model} such that networks constructed by the model are provably secure against any attacks of small sizes under the cascading failure models, and simultaneously follow a power law, and have the small world property with a navigating algorithm of time complex $O(\\log n)$. It is shown that for any network $G$ constructed from the security model, $G$ satisfies some remarkable topological properties, including: (i) the {\\it small community phenomenon}, that is, $G$ is rich in communities of the form $X$ of size poly logarithmic in $\\log n$ with conductance bounded by $O(\\frac{1}{|X|^{\\beta}})$ for some constant $\\beta$, (ii) small diameter property, with diameter $O(\\log n)$ allowing a navigation by a $O(\\log n)$ time algorithm to find a path for arbitrarily given two nodes, and (iii) power law distribution, and satisfies some probabilistic and combinatorial principles, including the {\\it degree priority theorem}, and {\\it infection-inclusion theorem}. By using these principles, we show that a network $G$ constructed from the security model is secure for any attacks of small scales under both the uniform threshold and random threshold cascading failure models. Our security theorems show that networks constructed from the security model are provably secure against any attacks of small sizes, for which natural selections of {\\it homophyly, randomness} and {\\it preferential attachment} are the underlying mechanisms."
                    },
                    {
                        "title": "Pointed computations and Martin-L\u00f6f randomness",
                        "abstract": "Schnorr showed that a real is Martin-Loef random if and only if all of its initial segments are incompressible with respect to prefix-free complexity. Fortnow and independently Nies, Stephan and Terwijn noticed that this statement remains true if we can merely require that the initial segments of the real corresponding to a computable increasing sequence of lengths are incompressible. The purpose of this note is to establish the following generalization of this fact. We show that a real is X Martin-Loef random if and only if its initial segments corresponding to a pointedly X-computable sequence (r_n) (where r_n is computable from X in a self-delimiting way, so that at most the first r_n bits of X are queried in the computation) of lengths are incompressible. On the other hand we also show that there are reals which are very far from being Martin-Loef random, yet they compute an increasing sequence of lengths at which their initial segments are incompressible."
                    },
                    {
                        "title": "Incremental Measurement of Structural Entropy for Dynamic Graphs",
                        "abstract": "Structural entropy is a metric that measures the amount of information embedded in graph structure data under a strategy of hierarchical abstracting. To measure the structural entropy of a dynamic graph, we need to decode the optimal encoding tree corresponding to the best community partitioning for each snapshot. However, the current methods do not support dynamic encoding tree updating and incremental structural entropy computation. To address this issue, we propose Incre-2dSE, a novel incremental measurement framework that dynamically adjusts the community partitioning and efficiently computes the updated structural entropy for each updated graph. Specifically, Incre-2dSE includes incremental algorithms based on two dynamic adjustment strategies for two-dimensional encoding trees, i.e., the naive adjustment strategy and the node-shifting adjustment strategy, which support theoretical analysis of updated structural entropy and incrementally optimize community partitioning towards a lower structural entropy. We conduct extensive experiments on 3 artificial datasets generated by Hawkes Process and 3 real-world datasets. Experimental results confirm that our incremental algorithms effectively capture the dynamic evolution of the communities, reduce time consumption, and provide great interpretability."
                    },
                    {
                        "title": "Effective and Stable Role-Based Multi-Agent Collaboration by Structural Information Principles",
                        "abstract": "Role-based learning is a promising approach to improving the performance of Multi-Agent Reinforcement Learning (MARL). Nevertheless, without manual assistance, current role-based methods cannot guarantee stably discovering a set of roles to effectively decompose a complex task, as they assume either a predefined role structure or practical experience for selecting hyperparameters. In this article, we propose a mathematical Structural Information principles-based Role Discovery method, namely SIRD, and then present a SIRD optimizing MARL framework, namely SR-MARL, for multi-agent collaboration. The SIRD transforms role discovery into a hierarchical action space clustering. Specifically, the SIRD consists of structuralization, sparsification, and optimization modules, where an optimal encoding tree is generated to perform abstracting to discover roles. The SIRD is agnostic to specific MARL algorithms and flexibly integrated with various value function factorization approaches. Empirical evaluations on the StarCraft II micromanagement benchmark demonstrate that, compared with state-of-the-art MARL algorithms, the SR-MARL framework improves the average test win rate by 0.17%, 6.08%, and 3.24%, and reduces the deviation by 16.67%, 30.80%, and 66.30%, under easy, hard, and super hard scenarios."
                    },
                    {
                        "title": "Adversarial Socialbots Modeling Based on Structural Information Principles",
                        "abstract": "The importance of effective detection is underscored by the fact that socialbots imitate human behavior to propagate misinformation, leading to an ongoing competition between socialbots and detectors. Despite the rapid advancement of reactive detectors, the exploration of adversarial socialbot modeling remains incomplete, significantly hindering the development of proactive detectors. To address this issue, we propose a mathematical Structural Information principles-based Adversarial Socialbots Modeling framework, namely SIASM, to enable more accurate and effective modeling of adversarial behaviors. First, a heterogeneous graph is presented to integrate various users and rich activities in the original social network and measure its dynamic uncertainty as structural entropy. By minimizing the high-dimensional structural entropy, a hierarchical community structure of the social network is generated and referred to as the optimal encoding tree. Secondly, a novel method is designed to quantify influence by utilizing the assigned structural entropy, which helps reduce the computational cost of SIASM by filtering out uninfluential users. Besides, a new conditional structural entropy is defined between the socialbot and other users to guide the follower selection for network influence maximization. Extensive and comparative experiments on both homogeneous and heterogeneous social networks demonstrate that, compared with state-of-the-art baselines, the proposed SIASM framework yields substantial performance improvements in terms of network influence (up to 16.32%) and sustainable stealthiness (up to 16.29%) when evaluated against a robust detector with 90% accuracy."
                    },
                    {
                        "title": "Effective Reinforcement Learning Based on Structural Information Principles",
                        "abstract": "Although Reinforcement Learning (RL) algorithms acquire sequential behavioral patterns through interactions with the environment, their effectiveness in noisy and high-dimensional scenarios typically relies on specific structural priors. In this paper, we propose a novel and general Structural Information principles-based framework for effective Decision-Making, namely SIDM, approached from an information-theoretic perspective. This paper presents a specific unsupervised partitioning method that forms vertex communities in the state and action spaces based on their feature similarities. An aggregation function, which utilizes structural entropy as the vertex weight, is devised within each community to obtain its embedding, thereby facilitating hierarchical state and action abstractions. By extracting abstract elements from historical trajectories, a directed, weighted, homogeneous transition graph is constructed. The minimization of this graph's high-dimensional entropy leads to the generation of an optimal encoding tree. An innovative two-layer skill-based learning mechanism is introduced to compute the common path entropy of each state transition as its identified probability, thereby obviating the requirement for expert knowledge. Moreover, SIDM can be flexibly incorporated into various single-agent and multi-agent RL algorithms, enhancing their performance. Finally, extensive evaluations on challenging benchmarks demonstrate that, compared with SOTA baselines, our framework significantly and consistently improves the policy's quality, stability, and efficiency up to 32.70%, 88.26%, and 64.86%, respectively."
                    }
                ]
            }
        }
    },
    "2405.18378": {
        "paper_data": {
            "title": "A Canonization Perspective on Invariant and Equivariant Learning",
            "url": "http://arxiv.org/abs/2405.18378v2",
            "arxiv_id": "2405.18378",
            "authors": [
                "George Ma",
                "Yifei Wang",
                "Derek Lim",
                "Stefanie Jegelka",
                "Yisen Wang"
            ],
            "abstract": "In many applications, we desire neural networks to exhibit invariance or equivariance to certain groups due to symmetries inherent in the data. Recently, frame-averaging methods emerged to be a unified framework for attaining symmetries efficiently by averaging over input-dependent subsets of the group, i.e., frames. What we currently lack is a principled understanding of the design of frames. In this work, we introduce a canonization perspective that provides an essential and complete view of the design of frames. Canonization is a classic approach for attaining invariance by mapping inputs to their canonical forms. We show that there exists an inherent connection between frames and canonical forms. Leveraging this connection, we can efficiently compare the complexity of frames as well as determine the optimality of certain frames. Guided by this principle, we design novel frames for eigenvectors that are strictly superior to existing methods -- some are even optimal -- both theoretically and empirically. The reduction to the canonization perspective further uncovers equivalences between previous methods. These observations suggest that canonization provides a fundamental understanding of existing frame-averaging methods and unifies existing equivariant and invariant learning methods.",
            "introduction": " Introduction When designing machine learning models, incorporating data symmetry provides a strong inductive bias that facilitates learning and generalization [ 53,55], as evidenced in multiple applications like convolutional neural networks [ 29], graph neural networks [ 50], point clouds [ 45,46],etc. Often these symmetry priors require models to be invariant or equivariant to certain groups G. Among these approaches, model-specific Methods implemented by ourselves are marked with *. The background to Appendix B.1. We prove the superiority of OAP-lap over MAP. Proof. OAP is strictly superior to MAP in two aspects. Firstly, we notice that the computation of summary vectors xiin these two algorithms is almost the same, except with different values for \u03b1i. 34In OAP , the \u03b1iis defined by \u03b1i= hash(Pii,{ {Pij} }j\u0338=i), while in MAP , their \u03b1\u2032 i=\u2225Pi\u2225is a special case of ours. Thus: \u2022 If\u03b1i=\u03b1j, thenPii=Pjjand{ {Pik} }={ {Pjk} }. This indicates \u03b1\u2032 i=\u2225Pi\u2225=\u2225Pj\u2225=\u03b1\u2032 j. \u2022However, if \u03b1\u2032 i=\u03b1\u2032 j, this does not necessarily mean \u03b1i=\u03b1j. For instance, let P1=P2= (1,0), then \u03b1\u2032 1=\u03b1\u2032 2but\u03b11\u0338=\u03b12. This means \u03b1iis more \u201cdistinctive\u201d than \u03b1\u2032 i, and as a result we will have a better chance finding indices i1<\u00b7\u00b7\u00b7< idsuch that Pxi1, . . . ,Pxidis a basis of the eigenspace. This is because, more \u201cdistinctive\u201d \u03b1iwould lead to more \u201crefined\u201d (smaller) groups Bi, and as a result, the rank of the projections {Pxi}will not decrease (and possibly increase). By splitting the group Biinto several smaller groups, the original summary vector can be expressed as a linear combination of all new summary vectors, thus the rank of projections of summary vectors does not decrease. Secondly, OAP is superior to MAP by allowing flexible choices of i1<\u00b7\u00b7\u00b7< id. In fact, after working out the summary vectors, the rest of the MAP algorithm is equivalent to Algorithm 3 except that they restrict i1= 1, . . . , i d=d. Recall that in MAP , given u1, . . . ,ui\u22121already found, they look for ui in the orthogonal complementary space of \u27e8u1, . . . ,ui\u22121\u27e9inVby projecting xionto that space; and ifxiis orthogonal to that space, their algorithm fails. This is exactly what Gram-Schmidt process does to the projection vectors {Pxi}. Except in our algorithm, instead of looking at projections Px1, . . . ,Pxd, we look for all projections Px1, . . . ,Pxkand try to select a subset of them such that the algorithm succeeds. This is obviously strictly superior, since there may be cases where Px1, . . . ,Pxdis not a basis of the eigenspace but another subset of Px1, . . . ,Pxkis. Next we prove the superiority of OAP-graph over FA-graph. FA-graph works as follows. Assume the graph has keigenspaces with projection matrices P1, . . . ,Pk. Lets1, . . . ,skbe the diagonals of these projection matrices, and let S= [s1, . . . ,sk]\u2208 Rn\u00d7k. Then the frame of the graph is defined to be all permutations that sort the rows of Sin column lexicographic order, and the canonical form of the graph is the set of graphs produced by these permutations (by applying them to the original graph). We call this canonization FA-graph. Next we introduce OAP-graph. Let \u03b1\u2113 ibe defined in the same way as \u03b1i, except that \u2113= 1, . . . , k denotes that \u03b1\u2113 iis computed from the projection matrix of the",
            "references": [
                {
                    "title": "Baking Symmetry into GFlowNets",
                    "abstract": "GFlowNets have exhibited promising performance in generating diverse candidates with high rewards. These networks generate objects incrementally and aim to learn a policy that assigns probability of sampling objects in proportion to rewards. However, the current training pipelines of GFlowNets do not consider the presence of isomorphic actions, which are actions resulting in symmetric or isomorphic states. This lack of symmetry increases the amount of samples required for training GFlowNets and can result in inefficient and potentially incorrect flow functions. As a consequence, the reward and diversity of the generated objects decrease. In this study, our objective is to integrate symmetries into GFlowNets by identifying equivalent actions during the generation process. Experimental results using synthetic data demonstrate the promising performance of our proposed approaches."
                },
                {
                    "title": "Beyond Weisfeiler-Lehman: A Quantitative Framework for GNN Expressiveness",
                    "abstract": "Designing expressive Graph Neural Networks (GNNs) is a fundamental topic in the graph learning community. So far, GNN expressiveness has been primarily assessed via the Weisfeiler-Lehman (WL) hierarchy. However, such an expressivity measure has notable limitations: it is inherently coarse, qualitative, and may not well reflect practical requirements (e.g., the ability to encode substructures). In this paper, we introduce a unified framework for quantitatively studying the expressiveness of GNN architectures, addressing all the above limitations. Specifically, we identify a fundamental expressivity measure termed homomorphism expressivity, which quantifies the ability of GNN models to count graphs under homomorphism. Homomorphism expressivity offers a complete and practical assessment tool: the completeness enables direct expressivity comparisons between GNN models, while the practicality allows for understanding concrete GNN abilities such as subgraph counting. By examining four classes of prominent GNNs as case studies, we derive simple, unified, and elegant descriptions of their homomorphism expressivity for both invariant and equivariant settings. Our results provide novel insights into a series of previous work, unify the landscape of different subareas in the community, and settle several open questions. Empirically, extensive experiments on both synthetic and real-world tasks verify our theory, showing that the practical performance of GNN models aligns well with the proposed metric."
                },
                {
                    "title": "Expressive Sign Equivariant Networks for Spectral Geometric Learning",
                    "abstract": "Recent work has shown the utility of developing machine learning models that respect the structure and symmetries of eigenvectors. These works promote sign invariance, since for any eigenvector v the negation -v is also an eigenvector. However, we show that sign invariance is theoretically limited for tasks such as building orthogonally equivariant models and learning node positional encodings for link prediction in graphs. In this work, we demonstrate the benefits of sign equivariance for these tasks. To obtain these benefits, we develop novel sign equivariant neural network architectures. Our models are based on a new analytic characterization of sign equivariant polynomials and thus inherit provable expressiveness properties. Controlled synthetic experiments show that our networks can achieve the theoretically predicted benefits of sign equivariant models. Code is available at https://github.com/cptq/Sign-Equivariant-Nets."
                },
                {
                    "title": "Laplacian Canonization: A Minimalist Approach to Sign and Basis Invariant Spectral Embedding",
                    "abstract": "Spectral embedding is a powerful graph embedding technique that has received a lot of attention recently due to its effectiveness on Graph Transformers. However, from a theoretical perspective, the universal expressive power of spectral embedding comes at the price of losing two important invariance properties of graphs, sign and basis invariance, which also limits its effectiveness on graph data. To remedy this issue, many previous methods developed costly approaches to learn new invariants and suffer from high computation complexity. In this work, we explore a minimal approach that resolves the ambiguity issues by directly finding canonical directions for the eigenvectors, named Laplacian Canonization (LC). As a pure pre-processing method, LC is light-weighted and can be applied to any existing GNNs. We provide a thorough investigation, from theory to algorithm, on this approach, and discover an efficient algorithm named Maximal Axis Projection (MAP) that works for both sign and basis invariance and successfully canonizes more than 90% of all eigenvectors. Experiments on real-world benchmark datasets like ZINC, MOLTOX21, and MOLPCBA show that MAP consistently outperforms existing methods while bringing minimal computation overhead. Code is available at https://github.com/PKU-ML/LaplacianCanonization."
                },
                {
                    "title": "AbDiffuser: Full-Atom Generation of In-Vitro Functioning Antibodies",
                    "abstract": "We introduce AbDiffuser, an equivariant and physics-informed diffusion model for the joint generation of antibody 3D structures and sequences. AbDiffuser is built on top of a new representation of protein structure, relies on a novel architecture for aligned proteins, and utilizes strong diffusion priors to improve the denoising process. Our approach improves protein diffusion by taking advantage of domain knowledge and physics-based constraints; handles sequence-length changes; and reduces memory complexity by an order of magnitude, enabling backbone and side chain generation. We validate AbDiffuser in silico and in vitro. Numerical experiments showcase the ability of AbDiffuser to generate antibodies that closely track the sequence and structural properties of a reference set. Laboratory experiments confirm that all 16 HER2 antibodies discovered were expressed at high levels and that 57.1% of the selected designs were tight binders."
                },
                {
                    "title": "FAENet: Frame Averaging Equivariant GNN for Materials Modeling",
                    "abstract": "Applications of machine learning techniques for materials modeling typically involve functions known to be equivariant or invariant to specific symmetries. While graph neural networks (GNNs) have proven successful in such tasks, they enforce symmetries via the model architecture, which often reduces their expressivity, scalability and comprehensibility. In this paper, we introduce (1) a flexible framework relying on stochastic frame-averaging (SFA) to make any model E(3)-equivariant or invariant through data transformations. (2) FAENet: a simple, fast and expressive GNN, optimized for SFA, that processes geometric information without any symmetrypreserving design constraints. We prove the validity of our method theoretically and empirically demonstrate its superior accuracy and computational scalability in materials modeling on the OC20 dataset (S2EF, IS2RE) as well as common molecular modeling tasks (QM9, QM7-X). A package implementation is available at https://faenet.readthedocs.io."
                },
                {
                    "title": "The Exact Sample Complexity Gain from Invariances for Kernel Regression on Manifolds",
                    "abstract": "In practice, encoding invariances into models helps sample complexity. In this work, we tighten and generalize theoretical results on how invariances improve sample complexity. In particular, we provide minimax optimal rates for kernel ridge regression on any manifold, with a target function that is invariant to an arbitrary group action on the manifold. Our results hold for (almost) any group action, even groups of positive dimension. For a finite group, the gain increases the\"effective\"number of samples by the group size. For groups of positive dimension, the gain is observed by a reduction in the manifold's dimension, in addition to a factor proportional to the volume of the quotient space. Our proof takes the viewpoint of differential geometry, in contrast to the more common strategy of using invariant polynomials. Hence, this new geometric viewpoint on learning with invariances may be of independent interest."
                },
                {
                    "title": "A Complete Expressiveness Hierarchy for Subgraph GNNs via Subgraph Weisfeiler-Lehman Tests",
                    "abstract": "Recently, subgraph GNNs have emerged as an important direction for developing expressive graph neural networks (GNNs). While numerous architectures have been proposed, so far there is still a limited understanding of how various design paradigms differ in terms of expressive power, nor is it clear what design principle achieves maximal expressiveness with minimal architectural complexity. To address these fundamental questions, this paper conducts a systematic study of general node-based subgraph GNNs through the lens of Subgraph Weisfeiler-Lehman Tests (SWL). Our central result is to build a complete hierarchy of SWL with strictly growing expressivity. Concretely, we prove that any node-based subgraph GNN falls into one of the six SWL equivalence classes, among which $\\mathsf{SSWL}$ achieves the maximal expressive power. We also study how these equivalence classes differ in terms of their practical expressiveness such as encoding graph distance and biconnectivity. Furthermore, we give a tight expressivity upper bound of all SWL algorithms by establishing a close relation with localized versions of WL and Folklore WL (FWL) tests. Our results provide insights into the power of existing subgraph GNNs, guide the design of new architectures, and point out their limitations by revealing an inherent gap with the 2-FWL test. Finally, experiments demonstrate that $\\mathsf{SSWL}$-inspired subgraph GNNs can significantly outperform prior architectures on multiple benchmarks despite great simplicity."
                },
                {
                    "title": "Reducing SO(3) Convolutions to SO(2) for Efficient Equivariant GNNs",
                    "abstract": "Graph neural networks that model 3D data, such as point clouds or atoms, are typically desired to be $SO(3)$ equivariant, i.e., equivariant to 3D rotations. Unfortunately equivariant convolutions, which are a fundamental operation for equivariant networks, increase significantly in computational complexity as higher-order tensors are used. In this paper, we address this issue by reducing the $SO(3)$ convolutions or tensor products to mathematically equivalent convolutions in $SO(2)$ . This is accomplished by aligning the node embeddings' primary axis with the edge vectors, which sparsifies the tensor product and reduces the computational complexity from $O(L^6)$ to $O(L^3)$, where $L$ is the degree of the representation. We demonstrate the potential implications of this improvement by proposing the Equivariant Spherical Channel Network (eSCN), a graph neural network utilizing our novel approach to equivariant convolutions, which achieves state-of-the-art results on the large-scale OC-20 and OC-22 datasets."
                },
                {
                    "title": "Rethinking the Expressive Power of GNNs via Graph Biconnectivity",
                    "abstract": "Designing expressive Graph Neural Networks (GNNs) is a central topic in learning graph-structured data. While numerous approaches have been proposed to improve GNNs in terms of the Weisfeiler-Lehman (WL) test, generally there is still a lack of deep understanding of what additional power they can systematically and provably gain. In this paper, we take a fundamentally different perspective to study the expressive power of GNNs beyond the WL test. Specifically, we introduce a novel class of expressivity metrics via graph biconnectivity and highlight their importance in both theory and practice. As biconnectivity can be easily calculated using simple algorithms that have linear computational costs, it is natural to expect that popular GNNs can learn it easily as well. However, after a thorough review of prior GNN architectures, we surprisingly find that most of them are not expressive for any of these metrics. The only exception is the ESAN framework, for which we give a theoretical justification of its power. We proceed to introduce a principled and more efficient approach, called the Generalized Distance Weisfeiler-Lehman (GD-WL), which is provably expressive for all biconnectivity metrics. Practically, we show GD-WL can be implemented by a Transformer-like architecture that preserves expressiveness and enjoys full parallelizability. A set of experiments on both synthetic and real datasets demonstrates that our approach can consistently outperform prior GNN architectures."
                },
                {
                    "title": "Canonical Fields: Self-Supervised Learning of Pose-Canonicalized Neural Fields",
                    "abstract": "Coordinate-based implicit neural networks, or neural fields, have emerged as useful representations of shape and appearance in 3D computer vision. Despite advances, however, it remains challenging to build neural fields for categories of objects without datasets like ShapeNet that provide \u201ccanonicalized\u201d object instances that are consistently aligned for their 3D position and orientation (pose). We present Canonical Field Network (CaFi-Net), a self-supervised method to canonicalize the 3D pose of instances from an object category represented as neural fields, specifically neural radiance fields (NeRFs). CaFi-Net directly learns from continuous and noisy radiance fields using a Siamese network architecture that is designed to extract equivariant field features for category-level canonicalization. During inference, our method takes pre-trained neural radiance fields of novel object instances at arbitrary 3D pose and estimates a canonical field with consistent 3D pose across the entire category. Extensive experiments on a new dataset of 1300 NeRF models across 13 object categories show that our method matches or exceeds the performance of 3D point cloud-based methods."
                },
                {
                    "title": "Equivariance with Learned Canonicalization Functions",
                    "abstract": "Symmetry-based neural networks often constrain the architecture in order to achieve invariance or equivariance to a group of transformations. In this paper, we propose an alternative that avoids this architectural constraint by learning to produce canonical representations of the data. These canonicalization functions can readily be plugged into non-equivariant backbone architectures. We offer explicit ways to implement them for some groups of interest. We show that this approach enjoys universality while providing interpretable insights. Our main hypothesis, supported by our empirical results, is that learning a small neural network to perform canonicalization is better than using predefined heuristics. Our experiments show that learning the canonicalization function is competitive with existing techniques for learning equivariant functions across many tasks, including image classification, $N$-body dynamics prediction, point cloud classification and part segmentation, while being faster across the board."
                },
                {
                    "title": "On The Computational Complexity of Self-Attention",
                    "abstract": "Transformer architectures have led to remarkable progress in many state-of-art applications. However, despite their successes, modern transformers rely on the self-attention mechanism, whose time- and space-complexity is quadratic in the length of the input. Several approaches have been proposed to speed up self-attention mechanisms to achieve sub-quadratic running time; however, the large majority of these works are not accompanied by rigorous error guarantees. In this work, we establish lower bounds on the computational complexity of self-attention in a number of scenarios. We prove that the time complexity of self-attention is necessarily quadratic in the input length, unless the Strong Exponential Time Hypothesis (SETH) is false. This argument holds even if the attention computation is performed only approximately, and for a variety of attention mechanisms. As a complement to our lower bounds, we show that it is indeed possible to approximate dot-product self-attention using finite Taylor series in linear-time, at the cost of having an exponential dependence on the polynomial order."
                },
                {
                    "title": "Understanding and Extending Subgraph GNNs by Rethinking Their Symmetries",
                    "abstract": "Subgraph GNNs are a recent class of expressive Graph Neural Networks (GNNs) which model graphs as collections of subgraphs. So far, the design space of possible Subgraph GNN architectures as well as their basic theoretical properties are still largely unexplored. In this paper, we study the most prominent form of subgraph methods, which employs node-based subgraph selection policies such as ego-networks or node marking and deletion. We address two central questions: (1) What is the upper-bound of the expressive power of these methods? and (2) What is the family of equivariant message passing layers on these sets of subgraphs?. Our first step in answering these questions is a novel symmetry analysis which shows that modelling the symmetries of node-based subgraph collections requires a significantly smaller symmetry group than the one adopted in previous works. This analysis is then used to establish a link between Subgraph GNNs and Invariant Graph Networks (IGNs). We answer the questions above by first bounding the expressive power of subgraph methods by 3-WL, and then proposing a general family of message-passing layers for subgraph methods that generalises all previous node-based Subgraph GNNs. Finally, we design a novel Subgraph GNN dubbed SUN, which theoretically unifies previous architectures while providing better empirical performance on multiple benchmarks."
                },
                {
                    "title": "Equivariant and Stable Positional Encoding for More Powerful Graph Neural Networks",
                    "abstract": "Graph neural networks (GNN) have shown great advantages in many graph-based learning tasks but often fail to predict accurately for a task-based on sets of nodes such as link/motif prediction and so on. Many works have recently proposed to address this problem by using random node features or node distance features. However, they suffer from either slow convergence, inaccurate prediction, or high complexity. In this work, we revisit GNNs that allow using positional features of nodes given by positional encoding (PE) techniques such as Laplacian Eigenmap, Deepwalk, etc. GNNs with PE often get criticized because they are not generalizable to unseen graphs (inductive) or stable. Here, we study these issues in a principled way and propose a provable solution, a class of GNN layers termed PEG with rigorous mathematical analysis. PEG uses separate channels to update the original node features and positional features. PEG imposes permutation equivariance w.r.t. the original node features and imposes $O(p)$ (orthogonal group) equivariance w.r.t. the positional features simultaneously, where $p$ is the dimension of used positional features. Extensive link prediction experiments over 8 real-world networks demonstrate the advantages of PEG in generalization and scalability."
                },
                {
                    "title": "Sign and Basis Invariant Networks for Spectral Graph Representation Learning",
                    "abstract": "We introduce SignNet and BasisNet -- new neural architectures that are invariant to two key symmetries displayed by eigenvectors: (i) sign flips, since if $v$ is an eigenvector then so is $-v$; and (ii) more general basis symmetries, which occur in higher dimensional eigenspaces with infinitely many choices of basis eigenvectors. We prove that under certain conditions our networks are universal, i.e., they can approximate any continuous function of eigenvectors with the desired invariances. When used with Laplacian eigenvectors, our networks are provably more expressive than existing spectral methods on graphs; for instance, they subsume all spectral graph convolutions, certain spectral graph invariants, and previously proposed graph positional encodings as special cases. Experiments show that our networks significantly outperform existing baselines on molecular graph regression, learning expressive graph representations, and learning neural fields on triangle meshes. Our code is available at https://github.com/cptq/SignNet-BasisNet ."
                },
                {
                    "title": "Multiscale Graph Comparison via the Embedded Laplacian Discrepancy",
                    "abstract": "Laplacian eigenvectors capture natural community structures on graphs and are widely used in spectral clustering and manifold learning. The use of Laplacian eigenvectors as embeddings for the purpose of multiscale graph comparison has however been limited. Here we propose the Embedded Laplacian Discrepancy (ELD) as a simple and fast approach to compare graphs (of potentially different sizes) based on the similarity of the graphs' community structures. The ELD operates by representing graphs as point clouds in a common, low-dimensional space, on which a natural Wasserstein-based distance can be efficiently computed. A main challenge in comparing graphs through any eigenvector-based approaches is the potential ambiguity that could arise due to sign-flips and basis symmetries. The ELD leverages a simple symmetrization trick to bypass any sign ambiguities. For comparing graphs that do not have any ambiguities due to basis symmetries (i.e. the spectrums are simple), we show that the ELD becomes a natural pseudo-metric that enjoys nice properties such as invariance under graph isomorphism. For comparing graphs with non-simple spectrums, we propose a procedure to approximate the ELD via a simple perturbation technique to resolve any ambiguity from basis symmetries. We show that such perturbations are stable using matrix perturbation theory under mild assumptions that are straightforward to verify in practice. We demonstrate the excellent applicability of the ELD approach on both simulated and real datasets."
                },
                {
                    "title": "ConDor: Self-Supervised Canonicalization of 3D Pose for Partial Shapes",
                    "abstract": "Progress in 3D object understanding has relied on manually \u201ccanonicalized\u201d shape datasets that contain instances with consistent position and orientation (3D pose). This has made it hard to generalize these methods to in-the-wild shapes, e.g., from internet model collections or depth sensors. ConDor is a self-supervised method that learns to Canonicalize the 3D orientation and position for full and partial 3D point clouds. We build on top of Tensor Field Networks (TFNs), a class of permutation- and rotation-equivariant, and translation-invariant 3D networks. During inference, our method takes an unseen full or partial 3D point cloud at an arbitrary pose and outputs an equivariant canonical pose. During training, this network uses self-supervision losses to learn the canonical pose from an un-canonicalized collection of full and partial 3D point clouds. ConDor can also learn to consistently co-segment object parts without any supervision. Extensive quantitative results on four new metrics show that our approach out-performs existing methods while enabling new applications such as operation on depth images and annotation transfer."
                },
                {
                    "title": "Multiset-Equivariant Set Prediction with Approximate Implicit Differentiation",
                    "abstract": "Most set prediction models in deep learning use set-equivariant operations, but they actually operate on multisets. We show that set-equivariant functions cannot represent certain functions on multisets, so we introduce the more appropriate notion of multiset-equivariance. We identify that the existing Deep Set Prediction Network (DSPN) can be multiset-equivariant without being hindered by set-equivariance and improve it with approximate implicit differentiation, allowing for better optimization while being faster and saving memory. In a range of toy experiments, we show that the perspective of multiset-equivariance is beneficial and that our changes to DSPN achieve better results in most cases. On CLEVR object property prediction, we substantially improve over the state-of-the-art Slot Attention from 8% to 77% in one of the strictest evaluation metrics because of the benefits made possible by implicit differentiation."
                },
                {
                    "title": "Graph Neural Networks with Learnable Structural and Positional Representations",
                    "abstract": "Graph neural networks (GNNs) have become the standard learning architectures for graphs. GNNs have been applied to numerous domains ranging from quantum chemistry, recommender systems to knowledge graphs and natural language processing. A major issue with arbitrary graphs is the absence of canonical positional information of nodes, which decreases the representation power of GNNs to distinguish e.g. isomorphic nodes and other graph symmetries. An approach to tackle this issue is to introduce Positional Encoding (PE) of nodes, and inject it into the input layer, like in Transformers. Possible graph PE are Laplacian eigenvectors. In this work, we propose to decouple structural and positional representations to make easy for the network to learn these two essential properties. We introduce a novel generic architecture which we call LSPE (Learnable Structural and Positional Encodings). We investigate several sparse and fully-connected (Transformer-like) GNNs, and observe a performance increase for molecular datasets, from 1.79% up to 64.14% when considering learnable PE for both GNN classes."
                },
                {
                    "title": "Frame Averaging for Invariant and Equivariant Network Design",
                    "abstract": "Many machine learning tasks involve learning functions that are known to be invariant or equivariant to certain symmetries of the input data. However, it is often challenging to design neural network architectures that respect these symmetries while being expressive and computationally efficient. For example, Euclidean motion invariant/equivariant graph or point cloud neural networks. We introduce Frame Averaging (FA), a general purpose and systematic framework for adapting known (backbone) architectures to become invariant or equivariant to new symmetry types. Our framework builds on the well known group averaging operator that guarantees invariance or equivariance but is intractable. In contrast, we observe that for many important classes of symmetries, this operator can be replaced with an averaging operator over a small subset of the group elements, called a frame. We show that averaging over a frame guarantees exact invariance or equivariance while often being much simpler to compute than averaging over the entire group. Furthermore, we prove that FA-based models have maximal expressive power in a broad setting and in general preserve the expressive power of their backbone architectures. Using frame averaging, we propose a new class of universal Graph Neural Networks (GNNs), universal Euclidean motion invariant point cloud networks, and Euclidean motion invariant Message Passing (MP) GNNs. We demonstrate the practical effectiveness of FA on several applications including point cloud normal estimation, beyond $2$-WL graph separation, and $n$-body dynamics prediction, achieving state-of-the-art results in all of these benchmarks."
                },
                {
                    "title": "Breaking the Limits of Message Passing Graph Neural Networks",
                    "abstract": "Since the Message Passing (Graph) Neural Networks (MPNNs) have a linear complexity with respect to the number of nodes when applied to sparse graphs, they have been widely implemented and still raise a lot of interest even though their theoretical expressive power is limited to the first order Weisfeiler-Lehman test (1-WL). In this paper, we show that if the graph convolution supports are designed in spectral-domain by a non-linear custom function of eigenvalues and masked with an arbitrary large receptive field, the MPNN is theoretically more powerful than the 1-WL test and experimentally as powerful as a 3-WL existing models, while remaining spatially localized. Moreover, by designing custom filter functions, outputs can have various frequency components that allow the convolution process to learn different relationships between a given input graph signal and its associated properties. So far, the best 3-WL equivalent graph neural networks have a computational complexity in $\\mathcal{O}(n^3)$ with memory usage in $\\mathcal{O}(n^2)$, consider non-local update mechanism and do not provide the spectral richness of output profile. The proposed method overcomes all these aforementioned problems and reaches state-of-the-art results in many downstream tasks."
                },
                {
                    "title": "Rethinking Graph Transformers with Spectral Attention",
                    "abstract": "In recent years, the Transformer architecture has proven to be very successful in sequence processing, but its application to other data structures, such as graphs, has remained limited due to the difficulty of properly defining positions. Here, we present the $\\textit{Spectral Attention Network}$ (SAN), which uses a learned positional encoding (LPE) that can take advantage of the full Laplacian spectrum to learn the position of each node in a given graph. This LPE is then added to the node features of the graph and passed to a fully-connected Transformer. By leveraging the full spectrum of the Laplacian, our model is theoretically powerful in distinguishing graphs, and can better detect similar sub-structures from their resonance. Further, by fully connecting the graph, the Transformer does not suffer from over-squashing, an information bottleneck of most GNNs, and enables better modeling of physical phenomenons such as heat transfer and electric interaction. When tested empirically on a set of 4 standard datasets, our model performs on par or better than state-of-the-art GNNs, and outperforms any attention-based model by a wide margin, becoming the first fully-connected architecture to perform well on graph benchmarks."
                },
                {
                    "title": "The Surprising Power of Graph Neural Networks with Random Node Initialization",
                    "abstract": "Graph neural networks (GNNs) are effective models for representation learning on relational data. However, standard GNNs are limited in their expressive power, as they cannot distinguish graphs beyond the capability of the Weisfeiler-Leman graph isomorphism heuristic. In order to break this expressiveness barrier, GNNs have been enhanced with random node initialization (RNI), where the idea is to train and run the models with randomized initial node features. In this work, we analyze the expressive power of GNNs with RNI, and prove that these models are universal, a first such result for GNNs not relying on computationally demanding higher-order properties. This universality result holds even with partially randomized initial node features, and preserves the invariance properties of GNNs in expectation. We then empirically analyze the effect of RNI on GNNs, based on carefully constructed datasets. Our empirical findings support the superior performance of GNNs with RNI over standard GNNs."
                },
                {
                    "title": "SE(3)-Transformers: 3D Roto-Translation Equivariant Attention Networks",
                    "abstract": "We introduce the SE(3)-Transformer, a variant of the self-attention module for 3D point clouds, which is equivariant under continuous 3D roto-translations. Equivariance is important to ensure stable and predictable performance in the presence of nuisance transformations of the data input. A positive corollary of equivariance is increased weight-tying within the model, leading to fewer trainable parameters and thus decreased sample complexity (i.e. we need less training data). The SE(3)-Transformer leverages the benefits of self-attention to operate on large point clouds with varying number of points, while guaranteeing SE(3)-equivariance for robustness. We evaluate our model on a toy $N$-body particle simulation dataset, showcasing the robustness of the predictions under rotations of the input. We further achieve competitive performance on two real-world datasets, ScanObjectNN and QM9. In all cases, our model outperforms a strong, non-equivariant attention baseline and an equivariant model without attention."
                },
                {
                    "title": "Open Graph Benchmark: Datasets for Machine Learning on Graphs",
                    "abstract": "We present the Open Graph Benchmark (OGB), a diverse set of challenging and realistic benchmark datasets to facilitate scalable, robust, and reproducible graph machine learning (ML) research. OGB datasets are large-scale (up to 100+ million nodes and 1+ billion edges), encompass multiple important graph ML tasks, and cover a diverse range of domains, ranging from social and information networks to biological networks, molecular graphs, source code ASTs, and knowledge graphs. For each dataset, we provide a unified evaluation protocol using meaningful application-specific data splits and evaluation metrics. In addition to building the datasets, we also perform extensive benchmark experiments for each dataset. Our experiments suggest that OGB datasets present significant challenges of scalability to large-scale graphs and out-of-distribution generalization under realistic data splits, indicating fruitful opportunities for future research. Finally, OGB provides an automated end-to-end graph ML pipeline that simplifies and standardizes the process of graph data loading, experimental setup, and model evaluation. OGB will be regularly updated and welcomes inputs from the community. OGB datasets as well as data loaders, evaluation scripts, baseline code, and leaderboards are publicly available at this https URL ."
                },
                {
                    "title": "Principal Neighbourhood Aggregation for Graph Nets",
                    "abstract": "Graph Neural Networks (GNNs) have been shown to be effective models for different predictive tasks on graph-structured data. Recent work on their expressive power has focused on isomorphism tasks and countable feature spaces. We extend this theoretical framework to include continuous features - which occur regularly in real-world input domains and within the hidden layers of GNNs - and we demonstrate the requirement for multiple aggregation functions in this context. Accordingly, we propose Principal Neighbourhood Aggregation (PNA), a novel architecture combining multiple aggregators with degree-scalers (which generalize the sum aggregator). Finally, we compare the capacity of different models to capture and exploit the graph structure via a novel benchmark containing multiple tasks taken from classical graph theory, alongside existing benchmarks from real-world domains, all of which demonstrate the strength of our model. With this work, we hope to steer some of the GNN research towards new aggregation methods which we believe are essential in the search for powerful and robust models."
                },
                {
                    "title": "Universal Equivariant Multilayer Perceptrons",
                    "abstract": "Group invariant and equivariant Multilayer Perceptrons (MLP), also known as Equivariant Networks, have achieved remarkable success in learning on a variety of data structures, such as sequences, images, sets, and graphs. Using tools from group theory, this paper proves the universality of a broad class of equivariant MLPs with a single hidden layer. In particular, it is shown that having a hidden layer on which the group acts regularly is sufficient for universal equivariance (invariance). A corollary is unconditional universality of equivariant MLPs for Abelian groups, such as CNNs with a single hidden layer. A second corollary is the universality of equivariant MLPs with a high-order hidden layer, where we give both group-agnostic bounds and means for calculating group-specific bounds on the order of hidden layer that guarantees universal equivariance (invariance)."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "ChebNet: Efficient and Stable Constructions of Deep Neural Networks with Rectified Power Units using Chebyshev Approximations",
                    "abstract": "In a recent paper [B. Li, S. Tang and H. Yu, arXiv:1903.05858], it was shown that deep neural networks built with rectified power units (RePU) can give better approximation for sufficient smooth functions than those with rectified linear units, by converting polynomial approximation given in power series into deep neural networks with optimal complexity and no approximation error. However, in practice, power series are not easy to compute. In this paper, we propose a new and more stable way to construct deep RePU neural networks based on Chebyshev polynomial approximations. By using a hierarchical structure of Chebyshev polynomial approximation in frequency domain, we build efficient and stable deep neural network constructions. In theory, ChebNets and the deep RePU nets based on Power series have the same upper error bounds for general function approximations. But numerically, ChebNets are much more stable. Numerical results show that the constructed ChebNets can be further trained and obtain much better results than those obtained by training deep RePU nets constructed basing on power series."
                },
                {
                    "title": "On Universal Equivariant Set Networks",
                    "abstract": "Using deep neural networks that are either invariant or equivariant to permutations in order to learn functions on unordered sets has become prevalent. The most popular, basic models are DeepSets [Zaheer et al. 2017] and PointNet [Qi et al. 2017]. While known to be universal for approximating invariant functions, DeepSets and PointNet are not known to be universal when approximating \\emph{equivariant} set functions. On the other hand, several recent equivariant set architectures have been proven equivariant universal [Sannai et al. 2019], [Keriven et al. 2019], however these models either use layers that are not permutation equivariant (in the standard sense) and/or use higher order tensor variables which are less practical. \nThere is, therefore, a gap in understanding the universality of popular equivariant set models versus theoretical ones. \nIn this paper we close this gap by proving that: (i) PointNet is not equivariant universal; and (ii) adding a single linear transmission layer makes PointNet universal. We call this architecture PointNetST and argue it is the simplest permutation equivariant universal model known to date. Another consequence is that DeepSets is universal, and also PointNetSeg, a popular point cloud segmentation network (used eg, in [Qi et al. 2017]) is universal. \nThe key theoretical tool used to prove the above results is an explicit characterization of all permutation equivariant polynomial layers. Lastly, we provide numerical experiments validating the theoretical results and comparing different permutation equivariant models."
                },
                {
                    "title": "Strategies for Pre-training Graph Neural Networks",
                    "abstract": "Many applications of machine learning require a model to make accurate pre-dictions on test examples that are distributionally different from training ones, while task-specific labels are scarce during training. An effective approach to this challenge is to pre-train a model on related tasks where data is abundant, and then fine-tune it on a downstream task of interest. While pre-training has been effective in many language and vision domains, it remains an open question how to effectively use pre-training on graph datasets. In this paper, we develop a new strategy and self-supervised methods for pre-training Graph Neural Networks (GNNs). The key to the success of our strategy is to pre-train an expressive GNN at the level of individual nodes as well as entire graphs so that the GNN can learn useful local and global representations simultaneously. We systematically study pre-training on multiple graph classification datasets. We find that naive strategies, which pre-train GNNs at the level of either entire graphs or individual nodes, give limited improvement and can even lead to negative transfer on many downstream tasks. In contrast, our strategy avoids negative transfer and improves generalization significantly across downstream tasks, leading up to 9.4% absolute improvements in ROC-AUC over non-pre-trained models and achieving state-of-the-art performance for molecular property prediction and protein function prediction."
                },
                {
                    "title": "Relational Pooling for Graph Representations",
                    "abstract": "This work generalizes graph neural networks (GNNs) beyond those based on the Weisfeiler-Lehman (WL) algorithm, graph Laplacians, and diffusions. Our approach, denoted Relational Pooling (RP), draws from the theory of finite partial exchangeability to provide a framework with maximal representation power for graphs. RP can work with existing graph representation models and, somewhat counterintuitively, can make them even more powerful than the original WL isomorphism test. Additionally, RP allows architectures like Recurrent Neural Networks and Convolutional Neural Networks to be used in a theoretically sound approach for graph classification. We demonstrate improved performance of RP-based graph representations over state-of-the-art methods on a number of tasks."
                },
                {
                    "title": "On the Universality of Invariant Networks",
                    "abstract": "Constraining linear layers in neural networks to respect symmetry transformations from a group $G$ is a common design principle for invariant networks that has found many applications in machine learning. \nIn this paper, we consider a fundamental question that has received little attention to date: Can these networks approximate any (continuous) invariant function? \nWe tackle the rather general case where $G\\leq S_n$ (an arbitrary subgroup of the symmetric group) that acts on $\\mathbb{R}^n$ by permuting coordinates. This setting includes several recent popular invariant networks. We present two main results: First, $G$-invariant networks are universal if high-order tensors are allowed. Second, there are groups $G$ for which higher-order tensors are unavoidable for obtaining universality. \n$G$-invariant networks consisting of only first-order tensors are of special interest due to their practical value. We conclude the paper by proving a necessary condition for the universality of $G$-invariant networks that incorporate only first-order tensors."
                },
                {
                    "title": "Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks",
                    "abstract": "In recent years, graph neural networks (GNNs) have emerged as a powerful neural architecture to learn vector representations of nodes and graphs in a supervised, end-to-end fashion. Up to now, GNNs have only been evaluated empirically\u2014showing promising results. The following work investigates GNNs from a theoretical point of view and relates them to the 1-dimensional Weisfeiler-Leman graph isomorphism heuristic (1-WL). We show that GNNs have the same expressiveness as the 1-WL in terms of distinguishing non-isomorphic (sub-)graphs. Hence, both algorithms also have the same shortcomings. Based on this, we propose a generalization of GNNs, so-called k-dimensional GNNs (k-GNNs), which can take higher-order graph structures at multiple scales into account. These higher-order structures play an essential role in the characterization of social networks and molecule graphs. Our experimental evaluation confirms our theoretical findings as well as confirms that higher-order information is useful in the task of graph classification and regression."
                },
                {
                    "title": "How Powerful are Graph Neural Networks?",
                    "abstract": "Graph Neural Networks (GNNs) are an effective framework for representation learning of graphs. GNNs follow a neighborhood aggregation scheme, where the representation vector of a node is computed by recursively aggregating and transforming representation vectors of its neighboring nodes. Many GNN variants have been proposed and have achieved state-of-the-art results on both node and graph classification tasks. However, despite GNNs revolutionizing graph representation learning, there is limited understanding of their representational properties and limitations. Here, we present a theoretical framework for analyzing the expressive power of GNNs to capture different graph structures. Our results characterize the discriminative power of popular GNN variants, such as Graph Convolutional Networks and GraphSAGE, and show that they cannot learn to distinguish certain simple graph structures. We then develop a simple architecture that is provably the most expressive among the class of GNNs and is as powerful as the Weisfeiler-Lehman graph isomorphism test. We empirically validate our theoretical findings on a number of graph classification benchmarks, and demonstrate that our model achieves state-of-the-art performance."
                },
                {
                    "title": "Tensor Field Networks: Rotation- and Translation-Equivariant Neural Networks for 3D Point Clouds",
                    "abstract": "We introduce tensor field neural networks, which are locally equivariant to 3D rotations, translations, and permutations of points at every layer. 3D rotation equivariance removes the need for data augmentation to identify features in arbitrary orientations. Our network uses filters built from spherical harmonics; due to the mathematical consequences of this filter choice, each layer accepts as input (and guarantees as output) scalars, vectors, and higher-order tensors, in the geometric sense of these terms. We demonstrate the capabilities of tensor field networks with tasks in geometry, physics, and chemistry."
                },
                {
                    "title": "Residual Gated Graph ConvNets",
                    "abstract": "Graph-structured data such as social networks, functional brain networks, gene regulatory networks, communications networks have brought the interest in generalizing deep learning techniques to graph domains. In this paper, we are interested to design neural networks for graphs with variable length in order to solve learning problems such as vertex classification, graph classification, graph regression, and graph generative tasks. Most existing works have focused on recurrent neural networks (RNNs) to learn meaningful representations of graphs, and more recently new convolutional neural networks (ConvNets) have been introduced. In this work, we want to compare rigorously these two fundamental families of architectures to solve graph learning tasks. We review existing graph RNN and ConvNet architectures, and propose natural extension of LSTM and ConvNet to graphs with arbitrary size. Then, we design a set of analytically controlled experiments on two basic graph problems, i.e. subgraph matching and graph clustering, to test the different architectures. Numerical results show that the proposed graph ConvNets are 3-17% more accurate and 1.5-4x faster than graph RNNs. Graph ConvNets are also 36% more accurate than variational (non-learning) techniques. Finally, the most effective graph ConvNet architecture uses gated edges and residuality. Residuality plays an essential role to learn multi-layer architectures as they provide a 10% gain of performance."
                },
                {
                    "title": "Graph Attention Networks",
                    "abstract": "We present graph attention networks (GATs), novel neural network architectures that operate on graph-structured data, leveraging masked self-attentional layers to address the shortcomings of prior methods based on graph convolutions or their approximations. By stacking layers in which nodes are able to attend over their neighborhoods' features, we enable (implicitly) specifying different weights to different nodes in a neighborhood, without requiring any kind of costly matrix operation (such as inversion) or depending on knowing the graph structure upfront. In this way, we address several key challenges of spectral-based graph neural networks simultaneously, and make our model readily applicable to inductive as well as transductive problems. Our GAT models have achieved or matched state-of-the-art results across four established transductive and inductive graph benchmarks: the Cora, Citeseer and Pubmed citation network datasets, as well as a protein-protein interaction dataset (wherein test graphs remain unseen during training)."
                },
                {
                    "title": "PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space",
                    "abstract": "Few prior works study deep learning on point sets. PointNet by Qi et al. is a pioneer in this direction. However, by design PointNet does not capture local structures induced by the metric space points live in, limiting its ability to recognize fine-grained patterns and generalizability to complex scenes. In this work, we introduce a hierarchical neural network that applies PointNet recursively on a nested partitioning of the input point set. By exploiting metric space distances, our network is able to learn local features with increasing contextual scales. With further observation that point sets are usually sampled with varying densities, which results in greatly decreased performance for networks trained on uniform densities, we propose novel set learning layers to adaptively combine features from multiple scales. Experiments show that our network called PointNet++ is able to learn deep point set features efficiently and robustly. In particular, results significantly better than state-of-the-art have been obtained on challenging benchmarks of 3D point clouds."
                },
                {
                    "title": "Multiscale Nonrigid Point Cloud Registration Using Rotation-Invariant Sliced-Wasserstein Distance via Laplace-Beltrami Eigenmap",
                    "abstract": "In this work, we propose computational models and algorithms for point cloud registration with nonrigid transformation. First, point clouds sampled from manifolds originally embedded in some Euclid..."
                },
                {
                    "title": "Deep Sets",
                    "abstract": "In this paper, we study the problem of designing objective functions for machine learning problems defined on finite \\emph{sets}. In contrast to traditional objective functions defined for machine learning problems operating on finite dimensional vectors, the new objective functions we propose are operating on finite sets and are invariant to permutations. Such problems are widespread, ranging from estimation of population statistics \\citep{poczos13aistats}, via anomaly detection in piezometer data of embankment dams \\citep{Jung15Exploration}, to cosmology \\citep{Ntampaka16Dynamical,Ravanbakhsh16ICML1}. Our main theorem characterizes the permutation invariant objective functions and provides a family of functions to which any permutation invariant objective function must belong. This family of functions has a special structure which enables us to design a deep network architecture that can operate on sets and which can be deployed on a variety of scenarios including both unsupervised and supervised learning tasks. We demonstrate the applicability of our method on population statistic estimation, point cloud classification, set expansion, and image tagging."
                },
                {
                    "title": "PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation",
                    "abstract": "Point cloud is an important type of geometric data structure. Due to its irregular format, most researchers transform such data to regular 3D voxel grids or collections of images. This, however, renders data unnecessarily voluminous and causes issues. In this paper, we design a novel type of neural network that directly consumes point clouds, which well respects the permutation invariance of points in the input. Our network, named PointNet, provides a unified architecture for applications ranging from object classification, part segmentation, to scene semantic parsing. Though simple, PointNet is highly efficient and effective. Empirically, it shows strong performance on par or even better than state of the art. Theoretically, we provide analysis towards understanding of what the network has learnt and why the network is robust with respect to input perturbation and corruption."
                },
                {
                    "title": "Generalization Error of Invariant Classifiers",
                    "abstract": "This paper studies the generalization error of invariant classifiers. In particular, we consider the common scenario where the classification task is invariant to certain transformations of the input, and that the classifier is constructed (or learned) to be invariant to these transformations. Our approach relies on factoring the input space into a product of a base space and a set of transformations. We show that whereas the generalization error of a non-invariant classifier is proportional to the complexity of the input space, the generalization error of an invariant classifier is proportional to the complexity of the base space. We also derive a set of sufficient conditions on the geometry of the base space and the set of transformations that ensure that the complexity of the base space is much smaller than the complexity of the input space. Our analysis applies to general classifiers such as convolutional neural networks. We demonstrate the implications of the developed theory for such classifiers with experiments on the MNIST and CIFAR-10 datasets."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Concentration inequalities for sampling without replacement",
                    "abstract": "Concentration inequalities quantify the deviation of a random variable from a fixed value. In spite of numerous applications, such as opinion surveys or ecological counting procedures , few concentration results are known for the setting of sampling without replacement from a finite population. Until now, the best general concentration inequality has been a Hoeffding inequality due to ?. In this paper, we first improve on the fundamental result of ?, and further extend it to obtain a Bernstein concentration bound for sampling without replacement. We then derive an empirical version of our bound that does not require the variance to be known to the user."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "ZINC: A Free Tool to Discover Chemistry for Biology",
                    "abstract": "ZINC is a free public resource for ligand discovery. The database contains over twenty million commercially available molecules in biologically relevant representations that may be downloaded in popular ready-to-dock formats and subsets. The Web site also enables searches by structure, biological activity, physical property, vendor, catalog number, name, and CAS number. Small custom subsets may be created, edited, shared, docked, downloaded, and conveyed to a vendor for purchase. The database is maintained and curated for a high purchasing success rate and is freely available at zinc.docking.org."
                },
                {
                    "title": "Weisfeiler-Lehman Graph Kernels",
                    "abstract": "In this article, we propose a family of efficient kernels for large graphs with discrete node labels. Key to our method is a rapid feature extraction scheme based on the Weisfeiler-Lehman test of isomorphism on graphs. It maps the original graph to a sequence of graphs, whose node attributes capture topological and label information. A family of kernels can be defined based on this Weisfeiler-Lehman sequence of graphs, including a highly efficient kernel comparing subtree-like patterns. Its runtime scales only linearly in the number of edges of the graphs and the length of the Weisfeiler-Lehman graph sequence. In our experimental evaluation, our kernels outperform state-of-the-art graph kernels on several graph classification benchmark data sets in terms of accuracy and runtime. Our kernels open the door to large-scale applications of graph kernels in various disciplines such as computational biology and social network analysis."
                },
                {
                    "title": "Resolving the sign ambiguity in the singular value decomposition",
                    "abstract": "Many modern data analysis methods involve computing a matrix singular value decomposition (SVD) or eigenvalue decomposition (EVD). Principal component analysis is the time\u2010honored example, but more recent applications include latent semantic indexing (LSI), hypertext induced topic selection (HITS), clustering, classification, etc. Though the SVD and EVD are well established and can be computed via state\u2010of\u2010the\u2010art algorithms, it is not commonly mentioned that there is an intrinsic sign indeterminacy that can significantly impact the conclusions and interpretations drawn from their results. Here we provide a solution to the sign ambiguity problem and show how it leads to more sensible solutions. Copyright \u00a9 2008 John Wiley & Sons, Ltd."
                },
                {
                    "title": "Segmentation using eigenvectors: a unifying view",
                    "abstract": "Automatic grouping and segmentation of images remains a challenging problem in computer vision. Recently, a number of authors have demonstrated good performance on this task using methods that are based on eigenvectors of the affinity matrix. These approaches are extremely attractive in that they are based on simple eigendecomposition algorithms whose stability is well understood. Nevertheless, the use of eigendecompositions in the context of segmentation is far from well understood. In this paper we give a unified treatment of these algorithms, and show the close connections between them while highlighting their distinguishing features. We then prove results on eigenvectors of block matrices that allow us to analyze the performance of these algorithms in simple grouping settings. Finally, we use our analysis to motivate a variation on the existing methods that combines aspects from different eigenvector segmentation algorithms. We illustrate our analysis with results on real and synthetic images."
                },
                {
                    "title": "A general theory of canonical forms",
                    "abstract": "Si G est un groupe de Lie compact et M une G-variete riemannienne d'orbites principales de codimension k alors une section ou une forme canonique pour M est une sous-variete fermee lisse a k dimensions de M qui rencontre toutes les orbites de M orthogonalement"
                },
                {
                    "title": "Canonical labeling of graphs",
                    "abstract": "We announce an algebraic approach to the problem of assigning canonical forms to graphs. We compute canonical forms and the associated canonical labelings (or renumberings) in polynomial time for graphs of bounded valence, in moderately exponential, exp(n\u00bd + &ogr;(1)),time for general graphs, in subexponential, nlog n, time for tournaments and for 2-(&ngr;,&kgr;,&lgr;) block designs with &kgr;,&lgr; bounded and nlog log n time for &lgr;-planes (symmetric designs) with &lgr; bounded. We prove some related problems NP-hard and indicate some open problems."
                },
                {
                    "title": "Isomorphism of graphs with bounded eigenvalue multiplicity",
                    "abstract": "We investigate the connection between the spectrum of a graph, i.e. the eigenvalues of the adjacency matrix, and the complexity of testing isomorphism. In particular we describe two polynomial time algorithms which test isomorphism of undirected graphs whose eigenvalues have bounded multiplicity. If X and Y are graphs of eigenvalue multiplicity m, then the isomorphism of X and Y can be tested by an O(n4m+c) deterministic and by an O(n2m+c) Las Vegas algorithm, where n is the number of vertices of X and Y."
                },
                {
                    "title": "Benchmarking Graph Neural Networks",
                    "abstract": "Graph neural networks (GNNs) have become the standard toolkit for analyzing and learning from data on graphs. As the field grows, it becomes critical to identify key architectures and validate new ideas that generalize to larger, more complex datasets. Unfortunately, it has been increasingly difficult to gauge the effectiveness of new models in the absence of a standardized benchmark with consistent experimental settings. In this paper, we introduce a reproducible GNN benchmarking framework, with the facility for researchers to add new models conveniently for arbitrary datasets. We demonstrate the usefulness of our framework by presenting a principled investigation into the recent Weisfeiler-Lehman GNNs (WL-GNNs) compared to message passing-based graph convolutional networks (GCNs) for a variety of graph tasks, i.e. graph regression/classification and node/link prediction, with medium-scale datasets."
                },
                {
                    "title": "On the Stability of Expressive Positional Encodings for Graph Neural Networks",
                    "abstract": ","
                },
                {
                    "title": "THE REDUCTION OF A GRAPH TO CANONICAL FORM AND THE ALGEBRA WHICH APPEARS THEREIN",
                    "abstract": "We consider an algorithm for the reduction of a given finite multigraph \u0393 to canonical form. Therein the new invariant of a graph appears \u2014 the algebra A(\u0393). The study of properties of the algebra A(\u0393) turns out to be helpful in solving a number of graph-theoretic problems. We pose and discuss some conjectures on the relation between properties of the algebra A(\u0393) and the automorphism group Aut(\u0393) of a graph \u0393. We give an example of undirected graph \u0393 whose algebra A(\u0393) coincides with the group algebra of some noncommutative group. English abstract from the original article. An algorithm is considered, reducing the specified finite multigraph \u0393 to canonical form. In the course of this reduction, a new invariant of the graph is generated \u2014 algebra A(\u0393). Study of the properties of the algebra A(\u0393) proves helpful in solving a number of graph-theoretic problems. Some propositions concerning the relationships between the properties of the algebra A(\u0393) and the graph\u2019s automorphism group Aut(\u0393) are discussed. An example of non-oriented graph \u0393 is constructed whose algebra A(\u0393) coincides with the group algebra of a non-commutative group. English title from the original article. A reduction of a graph to canonical form and an algebra arising during this reduction. 1. Consider a finite graph \u0393 and its adjacency matrix A(\u0393) = {aij}, where aij is the number of edges from ith vertex to jth one; i, j = 1, 2, . . . , n. If \u0393 is an undirected graph then set aij = aji. A canonical form of a graph is defined to be its adjacency matrix with respect to a canonical labeling of its vertices, that is a partial ordering of the vertex set such that if vertices a and b are incomparable then there is an automorphism of a graph moving a to b and preserving the adjacency relation. In Sections 6 and 7, we describe the reduction of a graph to canonical form which consists of a step-by-step reordering of rows and columns of the matrix A(\u0393) and, roughly speaking, adds up to the following. Consider for simplicity an undirected graph without multiple edges. Associate with every vertex of the graph the characteristic vector which has one component equal to the number of neighbors of this vertex. Then divide vertices into classes such that vertices with equal characteristic vectors belong to the same class and order classes according to the natural order on the set of characteristic vectors. Further, associate with every vertex the characteristic vector vi = (l, vi1, vi2, . . .), where vik is the number of neighbors of vertex i from class k and l is the number of the class which contains vertex i. Now again divide vertices into classes according to new characteristic vectors ordered lexicographically, etc. Note that if vertices a and b belong to different classes and the condition a < b holds at some step then this condition also holds at the next steps. This implies that the described process stops after at most n steps and after the stop either all vertices belong to different classes (i.e. a canonical labeling was constructed) or further division does not proceed."
                },
                {
                    "title": "UvA-DARE (Digital Academic Repository) Neural Relational Inference for Interacting Systems Neural Relational Inference for Interacting Systems",
                    "abstract": "Interacting systems are prevalent in nature, from dynamical systems in physics to complex societal dynamics. The interplay of components can give rise to complex behavior, which can often be explained using a simple model of the system\u2019s con-stituent parts. In this work, we introduce the neural relational inference (NRI) model: an unsupervised model that learns to infer interactions while simultaneously learning the dynamics purely from observational data. Our model takes the form of a variational auto-encoder, in which the latent code represents the underlying interaction graph and the reconstruction is based on graph neural networks. In experiments on simulated physical systems, we show that our NRI model can accurately recover ground-truth interactions in an unsupervised manner. We further demonstrate that we can \ufb01nd an interpretable structure and predict complex dynamics in real motion capture and sports tracking data."
                },
                {
                    "title": "Seeing and Hearing the Eigenvectors of a Fluid",
                    "abstract": "The intricate shapes and sounds that arise from vibrating Chladni plates are a well-known phenomenon. They are also quantitatively well understood, as the spatial patterns correspond to the eigenvectors of the underlying plate, and the audio frequencies arise from the plate\u2019s eigenvalues. We explore a generalization of the phenomenon by computing analogous quantities for a computational fluid dynamics simulation. Unlike the Chladni plate case, direct analytic expressions are not available, so we instead compute a set of \u201cempirical\u201d eigenvectors and eigenvalues. We find that these vectors form abstract, turbulent patterns in space. In another departure from the Chladni plate case, the eigenvalues no longer have a natural sonic mapping, so we construct a sonification that allows us to \u201clisten\u201d to the eigenvectors of the fluid. The united visual and sonic forms comprise a multimodal compositional palette that has great artistic potential. Figure 1: Left: Chladni patterns realized through a physical experiment. Source: Wikimedia Commons. Right: Analytic 2D eigenvector of the Navier-Stokes equations using the method of de Witt et al. [1]."
                },
                {
                    "title": "The Graph Neural Network Model",
                    "abstract": "Many underlying relationships among data in several areas of science and engineering, e.g., computer vision, molecular chemistry, molecular biology, pattern recognition, and data mining, can be represented in terms of graphs. In this paper, we propose a new neural network model, called graph neural network (GNN) model, that extends existing neural network methods for processing the data represented in graph domains. This GNN model, which can directly process most of the practically useful types of graphs, e.g., acyclic, cyclic, directed, and undirected, implements a function tau(G,n) isin IRm that maps a graph G and one of its nodes n into an m-dimensional Euclidean space. A supervised learning algorithm is derived to estimate the parameters of the proposed GNN model. The computational cost of the proposed algorithm is also considered. Some experimental results are shown to validate the proposed learning algorithm, and to demonstrate its generalization capabilities."
                },
                {
                    "title": "Blind Source Separation via Generalized Eigenvalue Decomposition",
                    "abstract": "In this short note we highlight the fact that linear blind source separation can be formulated as a generalized eigenvalue decomposition under the assumptions of non-Gaussian, non-stationary, or non-white independent sources. The solution for the unmixing matrix is given by the generalized eigenvectors that simultaneously diagonalize the covariance matrix of the observations and an additional symmetric matrix whose form depends upon the particular assumptions. The method critically determines the mixture coefficients and is therefore not robust to estimation errors. However it provides a rather general and unified solution that summarizes the conditions for successful blind source separation. To demonstrate the method, which can be implemented in two lines of matlab code, we present results for artificial mixtures of speech and real mixtures of electroencephalography (EEG) data, showing that the same sources are recovered under the various assumptions."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we improve the efficiency and effectiveness of machine learning models by incorporating data symmetry through invariant or equivariant approaches?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the field of machine learning, as it can lead to more robust models that generalize better across various applications, such as computer vision and graph analysis. By addressing the challenges of incorporating symmetry into model design, future research can explore new architectures and algorithms that leverage these principles, potentially leading to breakthroughs in performance and interpretability. Additionally, practical applications could emerge in areas requiring high accuracy and reliability, such as medical diagnosis and autonomous systems.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the complexity of defining and implementing symmetry in machine learning models. Naive approaches may fail because they do not adequately capture the nuances of data structures or the relationships between data points. Technical obstacles include the need for sophisticated mathematical frameworks to represent symmetry and the computational burden of optimizing models that incorporate these principles. Theoretical challenges arise in ensuring that the models remain invariant or equivariant under transformations, which requires a deep understanding of both the data and the underlying mathematical properties.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often overlooked the importance of symmetry in model design, focusing instead on more straightforward approaches that do not leverage these principles. Limitations in existing solutions include a lack of comprehensive frameworks for integrating symmetry into various model architectures and insufficient exploration of the implications of symmetry on model performance. Barriers such as the complexity of mathematical formulations and the computational demands of symmetry-aware algorithms have also hindered progress. Our approach differs by introducing novel algorithms, such as OAP and OAP-graph, which explicitly incorporate symmetry in a more flexible and effective manner than prior methods.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves developing algorithms that utilize symmetry in machine learning models, specifically through the OAP and OAP-graph frameworks. We will use datasets that exhibit inherent symmetries, such as point clouds and graph structures, to evaluate our methods. The performance will be measured using metrics such as accuracy, computational efficiency, and generalization ability. We expect our approach to yield improved model performance, demonstrating that incorporating symmetry leads to more distinctive and refined representations, ultimately enhancing the learning process and model robustness."
            }
        },
        "author_data": {
            "bfc8b36e-ac2b-485a-9d99-89c54747feb5": {
                "pk": "bfc8b36e-ac2b-485a-9d99-89c54747feb5",
                "name": "George Ma",
                "collaborators": [
                    "Xiaonan Ma",
                    "George Marinescu",
                    "Dan Coman",
                    "Tien-Cuong Dinh",
                    "Hugues Auvray",
                    "Emmanuel Bengio",
                    "Yoshua Bengio",
                    "Dinghuai Zhang",
                    "Wen Lu"
                ],
                "domain": [
                    "Complex Geometry",
                    "Asymptotic Analysis",
                    "Hermitian Geometry",
                    "Symplectic Geometry"
                ],
                "publications": [
                    {
                        "title": "The first coefficients of the asymptotic expansion of the Bergman kernel of the spin^c Dirac operator",
                        "abstract": "We establish the existence of the asymptotic expansion of the Bergman kernel associated to the spin-c Dirac operators acting on high tensor powers of line bundles with non-degenerate mixed curvature (negative and positive eigenvalues) by extending the paper \" On the asymptotic expansion of Bergman kernel \" (math.DG/0404494) of Dai-Liu-Ma. We compute the second coefficient b_1 in the asymptotic expansion using the method of our paper \"Generalized Bergman kernels on symplectic manifolds\" (math.DG/0411559)."
                    },
                    {
                        "title": "Exponential Estimate for the asymptotics of Bergman kernels",
                        "abstract": "We prove an exponential estimate for the asymptotics of Bergman kernels of a positive line bundle under hypotheses of bounded geometry. We give further Bergman kernel proofs of complex geometry results, such as separation of points, existence of local coordinates and holomorphic convexity by sections of positive line bundles."
                    },
                    {
                        "title": "The SPIN$^c$ Dirac operator on high tensor powers of a line bundle",
                        "abstract": "We study the asymptotic of the spectrum of the \\spin Dirac operator on high tensor powers of a line bundle. As application, we get a simple proof of the main result of Guillemin-Uribe, which was originally proved by using the analysis of Toeplitz operators of Boutet de Monvel and Guillemin."
                    },
                    {
                        "title": "Remark on the off-diagonal expansion of the Bergman kernel on compact K\u00e4hler manifolds",
                        "abstract": "In this short note, we compare our previous works on the off-diagonal expansion of the Bergman kernel and the recent preprint of Lu-Shiffman (arxiv.1301.2166). In particular, we note that the vanishing of the coefficient of p^{-1/2} is implicitly contained in Dai-Liu-Ma's work (J. Differential Geom. 72 (2006), no. 1, 1-41) and was explicitly stated in our book (Holomorphic Morse inequalities and Bergman kernels, Progress in Math., vol. 254, Birkh\\\"auser, 2007)."
                    },
                    {
                        "title": "Berezin-Toeplitz quantization and its kernel expansion",
                        "abstract": "We survey recent results about the asymptotic expansion of Toeplitz operators and their kernels, as well as Berezin-Toeplitz quantization. We deal in particular with calculation of the first coefficients of these expansions."
                    },
                    {
                        "title": "Generalized Bergman kernels on symplectic manifolds",
                        "abstract": "We study the near diagonal asymptotic expansion of the generalized Bergman kernel of the renormalized Bochner-Laplacian on high tensor powers of a positive line bundle over a compact symplectic manifold. We show how to compute the coefficients of the expansion by recurrence and give a closed formula for the first two of them. As consequence, we calculate the density of states function of the Bochner-Laplacian and establish a symplectic version of the convergence of the induced Fubini-Study metric. We also discuss generalizations of the asymptotic expansion for non-compact or singular manifolds as well as their applications. Our approach is inspired by the analytic localization techniques of Bismut-Lebeau."
                    },
                    {
                        "title": "Berezin-Toeplitz quantization on Kaehler manifolds",
                        "abstract": "We study the Berezin-Toeplitz quantization on Kaehler manifolds. We explain first how to compute various associated asymptotic expansions, then we compute explicitly the first terms of the expansion of the kernel of the Berezin-Toeplitz operators, and of the composition of two Berezin-Toeplitz operators. As application we estimate the norm of Donaldson's Q-operator."
                    },
                    {
                        "title": "Equidistribution and convergence speed for zeros of holomorphic sections of singular Hermitian line bundles",
                        "abstract": "We establish the equidistribution of zeros of random holomorphic sections of powers of a semipositive singular Hermitian line bundle, with an estimate of the convergence speed."
                    },
                    {
                        "title": "Quotient of Bergman kernels on punctured Riemann surfaces",
                        "abstract": "In this paper we consider a punctured Riemann surface endowed with a Hermitian metric that equals the Poincar{\\'e} metric near the punctures, and a holomorphic line bundle that polarizes the metric. We show that the quotient of the Bergman kernel of high tensor powers of the line bundle and of the Bergman kernel of the Poincar{\\'e} model near the singularity tends to one up to arbitrary negative powers of the tensor power."
                    },
                    {
                        "title": "Baking Symmetry into GFlowNets",
                        "abstract": "GFlowNets have exhibited promising performance in generating diverse candidates with high rewards. These networks generate objects incrementally and aim to learn a policy that assigns probability of sampling objects in proportion to rewards. However, the current training pipelines of GFlowNets do not consider the presence of isomorphic actions, which are actions resulting in symmetric or isomorphic states. This lack of symmetry increases the amount of samples required for training GFlowNets and can result in inefficient and potentially incorrect flow functions. As a consequence, the reward and diversity of the generated objects decrease. In this study, our objective is to integrate symmetries into GFlowNets by identifying equivalent actions during the generation process. Experimental results using synthetic data demonstrate the promising performance of our proposed approaches."
                    },
                    {
                        "title": "Tian's theorem for Moishezon spaces",
                        "abstract": "We prove that the Fubini-Study currents associated to a sequence of singular Hermitian holomorphic line bundles on a compact normal Moishezon space distribute asymptotically as the curvature currents of their metrics."
                    },
                    {
                        "title": "Equidistribution for sequences of line bundles on normal Kaehler spaces",
                        "abstract": "We study the asymptotics of Fubini-Study currents and zeros of random holomorphic sections associated to a sequence of singular Hermitian line bundles on a compact normal Kaehler complex space."
                    },
                    {
                        "title": "Donaldson's $Q$-operators for symplectic manifolds",
                        "abstract": "We prove an estimate for Donaldson's $Q$-operator on a prequantized compact symplectic manifold. This estimate is an ingredient in the recent result of Keller and Lejmi about a symplectic generalization of Donaldson's lower bound for the $L^2$-norm of the Hermitian scalar curvature."
                    }
                ]
            },
            "ca43cc9c-700a-4040-9990-eae559bfb902": {
                "pk": "ca43cc9c-700a-4040-9990-eae559bfb902",
                "name": "Yifei Wang",
                "collaborators": [
                    "Yisen Wang",
                    "Mert Pilanci",
                    "Jiansheng Yang",
                    "Zhouchen Lin",
                    "Wuchen Li",
                    "Qixun Wang",
                    "Chuhong Zhu",
                    "Zhirui Wang",
                    "Hong Zhu",
                    "Xianghua Ying"
                ],
                "domain": [
                    "Machine Learning",
                    "Adversarial Training",
                    "Graph Neural Network",
                    "Bayesian Inference"
                ],
                "publications": [
                    {
                        "title": "Interaction Mechanism and Response of Tidal Effect on the Shallow Geology of Europa",
                        "abstract": "Europa has been conf irmed to have a multilayered structure and complex geological condition in last two decades whose detail and cause of formation remains unclear.In this work,we start from analyzing the mechanism of tidal ef fect on satellite's surface and discuss if interaction like tidal locking and orbital resonance play an important role in tidal ef fect including heating,underground convection and eruption.During discussion we propose the main factors affecting Europa's tidal heat and the formation mechanism of typical shallow geological features,and provide theoretical support for further exploration."
                    },
                    {
                        "title": "Aspect-based Sentiment Analysis in Document -- FOMC Meeting Minutes on Economic Projection",
                        "abstract": "The Federal Open Market Committee within the Federal Reserve System is responsible for managing inflation, maximizing employment, and stabilizing interest rates. Meeting minutes play an important role for market movements because they provide the birds eye view of how this economic complexity is constantly re-weighed. Therefore, There has been growing interest in analyzing and extracting sentiments on various aspects from large financial texts for economic projection. However, Aspect-based Sentiment Analysis is not widely used on financial data due to the lack of large labeled dataset. In this paper, I propose a model to train ABSA on financial documents under weak supervision and analyze its predictive power on various macroeconomic indicators."
                    },
                    {
                        "title": "Information Newton's flow: second-order optimization method in probability space",
                        "abstract": "We introduce a framework for Newton's flows in probability space with information metrics, named information Newton's flows. Here two information metrics are considered, including both the Fisher-Rao metric and the Wasserstein-2 metric. A known fact is that overdamped Langevin dynamics correspond to Wasserstein gradient flows of Kullback-Leibler (KL) divergence. Extending this fact to Wasserstein Newton's flows, we derive Newton's Langevin dynamics. We provide examples of Newton's Langevin dynamics in both one-dimensional space and Gaussian families. For the numerical implementation, we design sampling efficient variational methods in affine models and reproducing kernel Hilbert space (RKHS) to approximate Wasserstein Newton's directions. We also establish convergence results of the proposed information Newton's method with approximated directions. Several numerical examples from Bayesian sampling problems are shown to demonstrate the effectiveness of the proposed method."
                    },
                    {
                        "title": "The Convex Geometry of Backpropagation: Neural Network Gradient Flows Converge to Extreme Points of the Dual Convex Program",
                        "abstract": "We study non-convex subgradient flows for training two-layer ReLU neural networks from a convex geometry and duality perspective. We characterize the implicit bias of unregularized non-convex gradient flow as convex regularization of an equivalent convex model. We then show that the limit points of non-convex subgradient flows can be identified via primal-dual correspondence in this convex optimization problem. Moreover, we derive a sufficient condition on the dual variables which ensures that the stationary points of the non-convex objective are the KKT points of the convex objective, thus proving convergence of non-convex gradient flows to the global optimum. For a class of regular training data distributions such as orthogonal separable data, we show that this sufficient condition holds. Therefore, non-convex gradient flows in fact converge to optimal solutions of a convex optimization problem. We present numerical results verifying the predictions of our theory for non-convex subgradient descent."
                    },
                    {
                        "title": "Accelerated Information Gradient flow",
                        "abstract": "We present a framework for Nesterov's accelerated gradient flows in probability space to design efficient mean-field Markov chain Monte Carlo (MCMC) algorithms for Bayesian inverse problems. Here four examples of information metrics are considered, including Fisher-Rao metric, Wasserstein-2 metric, Kalman-Wasserstein metric and Stein metric. For both Fisher-Rao and Wasserstein-2 metrics, we prove convergence properties of accelerated gradient flows. In implementations, we propose a sampling-efficient discrete-time algorithm for Wasserstein-2, Kalman-Wasserstein and Stein accelerated gradient flows with a restart technique. We also formulate a kernel bandwidth selection method, which learns the gradient of logarithm of density from Brownian-motion samples. Numerical experiments, including Bayesian logistic regression and Bayesian neural network, show the strength of the proposed methods compared with state-of-the-art algorithms."
                    },
                    {
                        "title": "Sketching the Krylov Subspace: Faster Computation of the Entire Ridge Regularization Path",
                        "abstract": "We propose a fast algorithm for computing the entire ridge regression regularization path in nearly linear time. Our method constructs a basis on which the solution of ridge regression can be computed instantly for any value of the regularization parameter. Consequently, linear models can be tuned via cross-validation or other risk estimation strategies with substantially better efficiency. The algorithm is based on iteratively sketching the Krylov subspace with a binomial decomposition over the regularization path. We provide a convergence analysis with various sketching matrices and show that it improves the state-of-the-art computational complexity. We also provide a technique to adaptively estimate the sketching dimension. This algorithm works for both the over-determined and under-determined problems. We also provide an extension for matrix-valued ridge regression. The numerical results on real medium and large scale ridge regression tasks illustrate the effectiveness of the proposed method compared to standard baselines which require super-linear computational time."
                    },
                    {
                        "title": "SM2C: Boost the Semi-supervised Segmentation for Medical Image by using Meta Pseudo Labels and Mixed Images",
                        "abstract": "Recently, machine learning-based semantic segmentation algorithms have demonstrated their potential to accurately segment regions and contours in medical images, allowing the precise location of anatomical structures and abnormalities. Although medical images are difficult to acquire and annotate, semi-supervised learning methods are efficient in dealing with the scarcity of labeled data. However, overfitting is almost inevitable due to the limited images for training. Furthermore, the intricate shapes of organs and lesions in medical images introduce additional complexity in different cases, preventing networks from acquiring a strong ability to generalize. To this end, we introduce a novel method called Scaling-up Mix with Multi-Class (SM2C). This method uses three strategies - scaling-up image size, multi-class mixing, and object shape jittering - to improve the ability to learn semantic features within medical images. By diversifying the shape of the segmentation objects and enriching the semantic information within each sample, the SM2C demonstrates its potential, especially in the training of unlabelled data. Extensive experiments demonstrate the effectiveness of the SM2C on three benchmark medical image segmentation datasets. The proposed framework shows significant improvements over state-of-the-art counterparts."
                    },
                    {
                        "title": "Polynomial-Time Solutions for ReLU Network Training: A Complexity Classification via Max-Cut and Zonotopes",
                        "abstract": "We investigate the complexity of training a two-layer ReLU neural network with weight decay regularization. Previous research has shown that the optimal solution of this problem can be found by solving a standard cone-constrained convex program. Using this convex formulation, we prove that the hardness of approximation of ReLU networks not only mirrors the complexity of the Max-Cut problem but also, in certain special cases, exactly corresponds to it. In particular, when $\\epsilon\\leq\\sqrt{84/83}-1\\approx 0.006$, we show that it is NP-hard to find an approximate global optimizer of the ReLU network objective with relative error $\\epsilon$ with respect to the objective value. Moreover, we develop a randomized algorithm which mirrors the Goemans-Williamson rounding of semidefinite Max-Cut relaxations. To provide polynomial-time approximations, we classify training datasets into three categories: (i) For orthogonal separable datasets, a precise solution can be obtained in polynomial-time. (ii) When there is a negative correlation between samples of different classes, we give a polynomial-time approximation with relative error $\\sqrt{\\pi/2}-1\\approx 0.253$. (iii) For general datasets, the degree to which the problem can be approximated in polynomial-time is governed by a geometric factor that controls the diameter of two zonotopes intrinsic to the dataset. To our knowledge, these results present the first polynomial-time approximation guarantees along with first hardness of approximation results for regularized ReLU networks."
                    },
                    {
                        "title": "Reparameterized Sampling for Generative Adversarial Networks",
                        "abstract": "Recently, sampling methods have been successfully applied to enhance the sample quality of Generative Adversarial Networks (GANs). However, in practice, they typically have poor sample efficiency because of the independent proposal sampling from the generator. In this work, we propose REP-GAN, a novel sampling method that allows general dependent proposals by REParameterizing the Markov chains into the latent space of the generator. Theoretically, we show that our reparameterized proposal admits a closed-form Metropolis-Hastings acceptance ratio. Empirically, extensive experiments on synthetic and real datasets demonstrate that our REP-GAN largely improves the sample efficiency and obtains better sample quality simultaneously."
                    },
                    {
                        "title": "Fooling Adversarial Training with Inducing Noise",
                        "abstract": "Adversarial training is widely believed to be a reliable approach to improve model robustness against adversarial attack. However, in this paper, we show that when trained on one type of poisoned data, adversarial training can also be fooled to have catastrophic behavior, e.g., $<1\\%$ robust test accuracy with $>90\\%$ robust training accuracy on CIFAR-10 dataset. Previously, there are other types of noise poisoned in the training data that have successfully fooled standard training ($15.8\\%$ standard test accuracy with $99.9\\%$ standard training accuracy on CIFAR-10 dataset), but their poisonings can be easily removed when adopting adversarial training. Therefore, we aim to design a new type of inducing noise, named ADVIN, which is an irremovable poisoning of training data. ADVIN can not only degrade the robustness of adversarial training by a large margin, for example, from $51.7\\%$ to $0.57\\%$ on CIFAR-10 dataset, but also be effective for fooling standard training ($13.1\\%$ standard test accuracy with $100\\%$ standard training accuracy). Additionally, ADVIN can be applied to preventing personal data (like selfies) from being exploited without authorization under whether standard or adversarial training."
                    },
                    {
                        "title": "Improving Out-of-Distribution Generalization by Adversarial Training with Structured Priors",
                        "abstract": "Deep models often fail to generalize well in test domains when the data distribution differs from that in the training domain. Among numerous approaches to address this Out-of-Distribution (OOD) generalization problem, there has been a growing surge of interest in exploiting Adversarial Training (AT) to improve OOD performance. Recent works have revealed that the robust model obtained by conducting sample-wise AT also retains transferability to biased test domains. In this paper, we empirically show that sample-wise AT has limited improvement on OOD performance. Specifically, we find that AT can only maintain performance at smaller scales of perturbation while Universal AT (UAT) is more robust to larger-scale perturbations. This provides us with clues that adversarial perturbations with universal (low dimensional) structures can enhance the robustness against large data distribution shifts that are common in OOD scenarios. Inspired by this, we propose two AT variants with low-rank structures to train OOD-robust models. Extensive experiments on DomainBed benchmark show that our proposed approaches outperform Empirical Risk Minimization (ERM) and sample-wise AT. Our code is available at https://github.com/NOVAglow646/NIPS22-MAT-and-LDAT-for-OOD."
                    },
                    {
                        "title": "Can In-context Learning Really Generalize to Out-of-distribution Tasks?",
                        "abstract": "In this work, we explore the mechanism of in-context learning (ICL) on out-of-distribution (OOD) tasks that were not encountered during training. To achieve this, we conduct synthetic experiments where the objective is to learn OOD mathematical functions through ICL using a GPT-2 model. We reveal that Transformers may struggle to learn OOD task functions through ICL. Specifically, ICL performance resembles implementing a function within the pretraining hypothesis space and optimizing it with gradient descent based on the in-context examples. Additionally, we investigate ICL's well-documented ability to learn unseen abstract labels in context. We demonstrate that such ability only manifests in the scenarios without distributional shifts and, therefore, may not serve as evidence of new-task-learning ability. Furthermore, we assess ICL's performance on OOD tasks when the model is pretrained on multiple tasks. Both empirical and theoretical analyses demonstrate the existence of the \\textbf{low-test-error preference} of ICL, where it tends to implement the pretraining function that yields low test error in the testing context. We validate this through numerical experiments. This new theoretical result, combined with our empirical findings, elucidates the mechanism of ICL in addressing OOD tasks."
                    },
                    {
                        "title": "Dissecting the Diffusion Process in Linear Graph Convolutional Networks",
                        "abstract": "Graph Convolutional Networks (GCNs) have attracted more and more attentions in recent years. A typical GCN layer consists of a linear feature propagation step and a nonlinear transformation step. Recent works show that a linear GCN can achieve comparable performance to the original non-linear GCN while being much more computationally efficient. In this paper, we dissect the feature propagation steps of linear GCNs from a perspective of continuous graph diffusion, and analyze why linear GCNs fail to benefit from more propagation steps. Following that, we propose Decoupled Graph Convolution (DGC) that decouples the terminal time and the feature propagation steps, making it more flexible and capable of exploiting a very large number of feature propagation steps. Experiments demonstrate that our proposed DGC improves linear GCNs by a large margin and makes them competitive with many modern variants of non-linear GCNs."
                    },
                    {
                        "title": "A Unified Contrastive Energy-based Model for Understanding the Generative Ability of Adversarial Training",
                        "abstract": "Adversarial Training (AT) is known as an effective approach to enhance the robustness of deep neural networks. Recently researchers notice that robust models with AT have good generative ability and can synthesize realistic images, while the reason behind it is yet under-explored. In this paper, we demystify this phenomenon by developing a unified probabilistic framework, called Contrastive Energy-based Models (CEM). On the one hand, we provide the first probabilistic characterization of AT through a unified understanding of robustness and generative ability. On the other hand, our unified framework can be extended to the unsupervised scenario, which interprets unsupervised contrastive learning as an important sampling of CEM. Based on these, we propose a principled method to develop adversarial learning and sampling methods. Experiments show that the sampling methods derived from our framework improve the sample quality in both supervised and unsupervised learning. Notably, our unsupervised adversarial sampling method achieves an Inception score of 9.61 on CIFAR-10, which is superior to previous energy-based models and comparable to state-of-the-art generative models."
                    }
                ]
            },
            "94e015c1-b215-46fa-a352-9f3e7f960dfc": {
                "pk": "94e015c1-b215-46fa-a352-9f3e7f960dfc",
                "name": "Derek Lim",
                "collaborators": [
                    "Stefanie Jegelka",
                    "Haggai Maron",
                    "Joshua Robinson",
                    "Charles R. Johnson",
                    "Amit Harlev",
                    "Austin R. Benson",
                    "Ren\u00e9 Vidal",
                    "Benjamin D. Haeffele",
                    "Xiuyu Li",
                    "Felix Hohne"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Eigenvalue Theory",
                    "Machine Learning",
                    "Subspace Clustering"
                ],
                "publications": [
                    {
                        "title": "The Doubly Stochastic Single Eigenvalue Problem: A Computational Approach",
                        "abstract": "The problem of determining $DS_n$, the complex numbers that occur as an eigenvalue of an $n$-by-$n$ doubly stochastic matrix, has been a target of study for some time. The Perfect-Mirsky region, $PM_n$, is contained in $DS_n$, and is known to be exactly $DS_n$ for $n \\leq 4$, but strictly contained within $DS_n$ for $n = 5$. Here, we present a Boundary Conjecture that asserts that the boundary of $DS_n$ is achieved by eigenvalues of convex combinations of pairs of (or single) permutation matrices. We present a method to efficiently compute a portion of $DS_n$, and obtain computational results that support the Boundary Conjecture. We also give evidence that $DS_n$ is equal to $PM_n$ for certain $n > 5$."
                    },
                    {
                        "title": "Expertise and Dynamics within Crowdsourced Musical Knowledge Curation: A Case Study of the Genius Platform",
                        "abstract": "Many platforms collect crowdsourced information primarily from volunteers. As this type of knowledge curation has become widespread, contribution formats vary substantially and are driven by diverse processes across differing platforms. Thus, models for one platform are not necessarily applicable to others. Here, we study the temporal dynamics of Genius, a platform primarily designed for user-contributed annotations of song lyrics. A unique aspect of Genius is that the annotations are extremely local -- an annotated lyric may just be a few lines of a song -- but also highly related, e.g., by song, album, artist, or genre. We analyze several dynamical processes associated with lyric annotations and their edits, which differ substantially from models for other platforms. For example, expertise on song annotations follows a \"U shape\" where experts are both early and late contributors with non-experts contributing intermediately; we develop a user utility model that captures such behavior. We also find several contribution traits appearing early in a user's lifespan of contributions that distinguish (eventual) experts from non-experts. Combining our findings, we develop a model for early prediction of user expertise."
                    },
                    {
                        "title": "Doubly Stochastic Subspace Clustering",
                        "abstract": "Many state-of-the-art subspace clustering methods follow a two-step process by first constructing an affinity matrix between data points and then applying spectral clustering to this affinity. Most of the research into these methods focuses on the first step of generating the affinity, which often exploits the self-expressive property of linear subspaces, with little consideration typically given to the spectral clustering step that produces the final clustering. Moreover, existing methods often obtain the final affinity that is used in the spectral clustering step by applying ad-hoc or arbitrarily chosen postprocessing steps to the affinity generated by a self-expressive clustering formulation, which can have a significant impact on the overall clustering performance. In this work, we unify these two steps by learning both a self-expressive representation of the data and an affinity matrix that is well-normalized for spectral clustering. In our proposed models, we constrain the affinity matrix to be doubly stochastic, which results in a principled method for affinity matrix normalization while also exploiting known benefits of doubly stochastic normalization in spectral clustering. We develop a general framework and derive two models: one that jointly learns the self-expressive representation along with the doubly stochastic affinity, and one that sequentially solves for one then the other. Furthermore, we leverage sparsity in the problem to develop a fast active-set method for the sequential solver that enables efficient computation on large datasets. Experiments show that our method achieves state-of-the-art subspace clustering performance on many common datasets in computer vision."
                    },
                    {
                        "title": "New Benchmarks for Learning on Non-Homophilous Graphs",
                        "abstract": "Much data with graph structures satisfy the principle of homophily, meaning that connected nodes tend to be similar with respect to a specific attribute. As such, ubiquitous datasets for graph machine learning tasks have generally been highly homophilous, rewarding methods that leverage homophily as an inductive bias. Recent work has pointed out this particular focus, as new non-homophilous datasets have been introduced and graph representation learning models better suited for low-homophily settings have been developed. However, these datasets are small and poorly suited to truly testing the effectiveness of new methods in non-homophilous settings. We present a series of improved graph datasets with node label relationships that do not satisfy the homophily principle. Along with this, we introduce a new measure of the presence or absence of homophily that is better suited than existing measures in different regimes. We benchmark a range of simple methods and graph neural networks across our proposed datasets, drawing new insights for further research. Data and codes can be found at https://github.com/CUAI/Non-Homophily-Benchmarks."
                    },
                    {
                        "title": "Counting Substructures with Higher-Order Graph Neural Networks: Possibility and Impossibility Results",
                        "abstract": "While message passing Graph Neural Networks (GNNs) have become increasingly popular architectures for learning with graphs, recent works have revealed important shortcomings in their expressive power. In response, several higher-order GNNs have been proposed that substantially increase the expressive power, albeit at a large computational cost. Motivated by this gap, we explore alternative strategies and lower bounds. In particular, we analyze a new recursive pooling technique of local neighborhoods that allows different tradeoffs of computational cost and expressive power. First, we prove that this model can count subgraphs of size $k$, and thereby overcomes a known limitation of low-order GNNs. Second, we show how recursive pooling can exploit sparsity to reduce the computational complexity compared to the existing higher-order GNNs. More generally, we provide a (near) matching information-theoretic lower bound for counting subgraphs with graph representations that pool over representations of derived (sub-)graphs. We also discuss lower bounds on time complexity."
                    },
                    {
                        "title": "Spectra of Convex Hulls of Matrix Groups",
                        "abstract": "The still-unsolved problem of determining the set of eigenvalues realized by $n$-by-$n$ doubly stochastic matrices, those matrices with row sums and column sums equal to $1$, has attracted much attention in the last century. This problem is somewhat algebraic in nature, due to a result of Birkhoff demonstrating that the set of doubly stochastic matrices is the convex hull of the permutation matrices. Here we are interested in a general matrix group $G \\subseteq GL_n(\\mathbb{C})$ and the hull spectrum $\\text{HS}(G)$ of eigenvalues realized by convex combinations of elements of $G$. We show that hull spectra of matrix groups share many nice properties. Moreover, we give bounds on the hull spectra of matrix groups, determine $\\text{HS}(G)$ exactly for important classes of matrix groups, and study the hull spectra of representations of abstract groups."
                    },
                    {
                        "title": "Expressive Sign Equivariant Networks for Spectral Geometric Learning",
                        "abstract": "Recent work has shown the utility of developing machine learning models that respect the structure and symmetries of eigenvectors. These works promote sign invariance, since for any eigenvector v the negation -v is also an eigenvector. However, we show that sign invariance is theoretically limited for tasks such as building orthogonally equivariant models and learning node positional encodings for link prediction in graphs. In this work, we demonstrate the benefits of sign equivariance for these tasks. To obtain these benefits, we develop novel sign equivariant neural network architectures. Our models are based on a new analytic characterization of sign equivariant polynomials and thus inherit provable expressiveness properties. Controlled synthetic experiments show that our networks can achieve the theoretically predicted benefits of sign equivariant models. Code is available at https://github.com/cptq/Sign-Equivariant-Nets."
                    },
                    {
                        "title": "Structuring Representation Geometry with Rotationally Equivariant Contrastive Learning",
                        "abstract": "Self-supervised learning converts raw perceptual data such as images to a compact space where simple Euclidean distances measure meaningful variations in data. In this paper, we extend this formulation by adding additional geometric structure to the embedding space by enforcing transformations of input space to correspond to simple (i.e., linear) transformations of embedding space. Specifically, in the contrastive learning setting, we introduce an equivariance objective and theoretically prove that its minima forces augmentations on input space to correspond to rotations on the spherical embedding space. We show that merely combining our equivariant loss with a non-collapse term results in non-trivial representations, without requiring invariance to data augmentations. Optimal performance is achieved by also encouraging approximate invariance, where input augmentations correspond to small rotations. Our method, CARE: Contrastive Augmentation-induced Rotational Equivariance, leads to improved performance on downstream tasks, and ensures sensitivity in embedding space to important variations in data (e.g., color) that standard contrastive methods do not achieve. Code is available at https://github.com/Sharut/CARE."
                    },
                    {
                        "title": "Learning on LoRAs: GL-Equivariant Processing of Low-Rank Weight Spaces for Large Finetuned Models",
                        "abstract": "Low-rank adaptations (LoRAs) have revolutionized the finetuning of large foundation models, enabling efficient adaptation even with limited computational resources. The resulting proliferation of LoRAs presents exciting opportunities for applying machine learning techniques that take these low-rank weights themselves as inputs. In this paper, we investigate the potential of Learning on LoRAs (LoL), a paradigm where LoRA weights serve as input to machine learning models. For instance, an LoL model that takes in LoRA weights as inputs could predict the performance of the finetuned model on downstream tasks, detect potentially harmful finetunes, or even generate novel model edits without traditional training methods. We first identify the inherent parameter symmetries of low rank decompositions of weights, which differ significantly from the parameter symmetries of standard neural networks. To efficiently process LoRA weights, we develop several symmetry-aware invariant or equivariant LoL models, using tools such as canonicalization, invariant featurization, and equivariant layers. We finetune thousands of text-to-image diffusion models and language models to collect datasets of LoRAs. In numerical experiments on these datasets, we show that our LoL architectures are capable of processing low rank weight decompositions to predict CLIP score, finetuning data attributes, finetuning data membership, and accuracy on downstream tasks."
                    },
                    {
                        "title": "Sign and Basis Invariant Networks for Spectral Graph Representation Learning",
                        "abstract": "We introduce SignNet and BasisNet -- new neural architectures that are invariant to two key symmetries displayed by eigenvectors: (i) sign flips, since if $v$ is an eigenvector then so is $-v$; and (ii) more general basis symmetries, which occur in higher dimensional eigenspaces with infinitely many choices of basis eigenvectors. We prove that under certain conditions our networks are universal, i.e., they can approximate any continuous function of eigenvectors with the desired invariances. When used with Laplacian eigenvectors, our networks are provably more expressive than existing spectral methods on graphs; for instance, they subsume all spectral graph convolutions, certain spectral graph invariants, and previously proposed graph positional encodings as special cases. Experiments show that our networks significantly outperform existing baselines on molecular graph regression, learning expressive graph representations, and learning neural fields on triangle meshes. Our code is available at https://github.com/cptq/SignNet-BasisNet ."
                    },
                    {
                        "title": "Graph Metanetworks for Processing Diverse Neural Architectures",
                        "abstract": "Neural networks efficiently encode learned information within their parameters. Consequently, many tasks can be unified by treating neural networks themselves as input data. When doing so, recent studies demonstrated the importance of accounting for the symmetries and geometry of parameter spaces. However, those works developed architectures tailored to specific networks such as MLPs and CNNs without normalization layers, and generalizing such architectures to other types of networks can be challenging. In this work, we overcome these challenges by building new metanetworks - neural networks that take weights from other neural networks as input. Put simply, we carefully build graphs representing the input neural networks and process the graphs using graph neural networks. Our approach, Graph Metanetworks (GMNs), generalizes to neural architectures where competing methods struggle, such as multi-head attention layers, normalization layers, convolutional layers, ResNet blocks, and group-equivariant linear layers. We prove that GMNs are expressive and equivariant to parameter permutation symmetries that leave the input neural network functions unchanged. We validate the effectiveness of our method on several metanetwork tasks over diverse neural network architectures."
                    },
                    {
                        "title": "The Empirical Impact of Neural Parameter Symmetries, or Lack Thereof",
                        "abstract": "Many algorithms and observed phenomena in deep learning appear to be affected by parameter symmetries -- transformations of neural network parameters that do not change the underlying neural network function. These include linear mode connectivity, model merging, Bayesian neural network inference, metanetworks, and several other characteristics of optimization or loss-landscapes. However, theoretical analysis of the relationship between parameter space symmetries and these phenomena is difficult. In this work, we empirically investigate the impact of neural parameter symmetries by introducing new neural network architectures that have reduced parameter space symmetries. We develop two methods, with some provable guarantees, of modifying standard neural networks to reduce parameter space symmetries. With these new methods, we conduct a comprehensive experimental study consisting of multiple tasks aimed at assessing the effect of removing parameter symmetries. Our experiments reveal several interesting observations on the empirical impact of parameter symmetries; for instance, we observe linear mode connectivity between our networks without alignment of weight spaces, and we find that our networks allow for faster and more effective Bayesian neural network training. Our code is available at https://github.com/cptq/asymmetric-networks"
                    }
                ]
            },
            "721ccfad-3bfa-4d76-b3b0-98a312f87a61": {
                "pk": "721ccfad-3bfa-4d76-b3b0-98a312f87a61",
                "name": "Stefanie Jegelka",
                "collaborators": [
                    "Suvrit Sra",
                    "Jeff Bilmes",
                    "Chengtao Li",
                    "Matthew Staib",
                    "Thien Le",
                    "Behrooz Tahmasebi",
                    "Rishabh Iyer",
                    "Hongzhou Lin",
                    "Zi Wang",
                    "Arindam Banerjee"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Optimization",
                    "Statistical Learning"
                ],
                "publications": [
                    {
                        "title": "Theory of Graph Neural Networks: Representation and Learning",
                        "abstract": "Graph Neural Networks (GNNs), neural network architectures targeted to learning representations of graphs, have become a popular learning model for prediction tasks on nodes, graphs and configurations of points, with wide success in practice. This article summarizes a selection of the emerging theoretical results on approximation and learning properties of widely used message passing GNNs and higher-order GNNs, focusing on representation, generalization and extrapolation. Along the way, it summarizes mathematical connections."
                    },
                    {
                        "title": "Fast Sampling for Strongly Rayleigh Measures with Application to Determinantal Point Processes",
                        "abstract": "In this note we consider sampling from (non-homogeneous) strongly Rayleigh probability measures. As an important corollary, we obtain a fast mixing Markov Chain sampler for Determinantal Point Processes."
                    },
                    {
                        "title": "ResNet with one-neuron hidden layers is a Universal Approximator",
                        "abstract": "We demonstrate that a very deep ResNet with stacked modules with one neuron per hidden layer and ReLU activation functions can uniformly approximate any Lebesgue integrable function in $d$ dimensions, i.e. $\\ell_1(\\mathbb{R}^d)$. Because of the identity mapping inherent to ResNets, our network has alternating layers of dimension one and $d$. This stands in sharp contrast to fully connected networks, which are not universal approximators if their width is the input dimension $d$ [Lu et al, 2017; Hanin and Sellke, 2017]. Hence, our result implies an increase in representational power for narrow deep networks by the ResNet architecture."
                    },
                    {
                        "title": "Robust Budget Allocation via Continuous Submodular Functions",
                        "abstract": "The optimal allocation of resources for maximizing influence, spread of information or coverage, has gained attention in the past years, in particular in machine learning and data mining. But in applications, the parameters of the problem are rarely known exactly, and using wrong parameters can lead to undesirable outcomes. We hence revisit a continuous version of the Budget Allocation or Bipartite Influence Maximization problem introduced by Alon et al. (2012) from a robust optimization perspective, where an adversary may choose the least favorable parameters within a confidence set. The resulting problem is a nonconvex-concave saddle point problem (or game). We show that this nonconvex problem can be solved exactly by leveraging connections to continuous submodular functions, and by solving a constrained submodular minimization problem. Although constrained submodular minimization is hard in general, here, we establish conditions under which such a problem can be solved to arbitrary precision $\\epsilon$."
                    },
                    {
                        "title": "Max-value Entropy Search for Efficient Bayesian Optimization",
                        "abstract": "Entropy Search (ES) and Predictive Entropy Search (PES) are popular and empirically successful Bayesian Optimization techniques. Both rely on a compelling information-theoretic motivation, and maximize the information gained about the $\\arg\\max$ of the unknown function; yet, both are plagued by the expensive computation for estimating entropies. We propose a new criterion, Max-value Entropy Search (MES), that instead uses the information about the maximum function value. We show relations of MES to other Bayesian optimization methods, and establish a regret bound. We observe that MES maintains or improves the good empirical performance of ES/PES, while tremendously lightening the computational burden. In particular, MES is much more robust to the number of samples used for computing the entropy, and hence more efficient for higher dimensional problems."
                    },
                    {
                        "title": "Distributionally Robust Optimization and Generalization in Kernel Methods",
                        "abstract": "Distributionally robust optimization (DRO) has attracted attention in machine learning due to its connections to regularization, generalization, and robustness. Existing work has considered uncertainty sets based on phi-divergences and Wasserstein distances, each of which have drawbacks. In this paper, we study DRO with uncertainty sets measured via maximum mean discrepancy (MMD). We show that MMD DRO is roughly equivalent to regularization by the Hilbert norm and, as a byproduct, reveal deep connections to classic results in statistical learning. In particular, we obtain an alternative proof of a generalization bound for Gaussian kernel ridge regression via a DRO lense. The proof also suggests a new regularizer. Our results apply beyond kernel methods: we derive a generically applicable approximation of MMD DRO, and show that it generalizes recent work on variance-based regularization."
                    },
                    {
                        "title": "Training invariances and the low-rank phenomenon: beyond linear networks",
                        "abstract": "The implicit bias induced by the training of neural networks has become a topic of rigorous study. In the limit of gradient flow and gradient descent with appropriate step size, it has been shown that when one trains a deep linear network with logistic or exponential loss on linearly separable data, the weights converge to rank-1 matrices. In this paper, we extend this theoretical result to the last few linear layers of the much wider class of nonlinear ReLU-activated feedforward networks containing fully-connected layers and skip connections. Similar to the linear case, the proof relies on specific local training invariances, sometimes referred to as alignment, which we show to hold for submatrices where neurons are stably-activated in all training examples, and it reflects empirical results in the literature. We also show this is not true in general for the full matrix of ReLU fully-connected layers. Our proof relies on a specific decomposition of the network into a multilinear function and another ReLU network whose weights are constant under a certain parameter directional convergence."
                    },
                    {
                        "title": "The Exact Sample Complexity Gain from Invariances for Kernel Regression",
                        "abstract": "In practice, encoding invariances into models improves sample complexity. In this work, we study this phenomenon from a theoretical perspective. In particular, we provide minimax optimal rates for kernel ridge regression on compact manifolds, with a target function that is invariant to a group action on the manifold. Our results hold for any smooth compact Lie group action, even groups of positive dimension. For a finite group, the gain effectively multiplies the number of samples by the group size. For groups of positive dimension, the gain is observed by a reduction in the manifold's dimension, in addition to a factor proportional to the volume of the quotient space. Our proof takes the viewpoint of differential geometry, in contrast to the more common strategy of using invariant polynomials. This new geometric viewpoint on learning with invariances may be of independent interest."
                    },
                    {
                        "title": "Limits, approximation and size transferability for GNNs on sparse graphs via graphops",
                        "abstract": "Can graph neural networks generalize to graphs that are different from the graphs they were trained on, e.g., in size? In this work, we study this question from a theoretical perspective. While recent work established such transferability and approximation results via graph limits, e.g., via graphons, these only apply non-trivially to dense graphs. To include frequently encountered sparse graphs such as bounded-degree or power law graphs, we take a perspective of taking limits of operators derived from graphs, such as the aggregation operation that makes up GNNs. This leads to the recently introduced limit notion of graphops (Backhausz and Szegedy, 2022). We demonstrate how the operator perspective allows us to develop quantitative bounds on the distance between a finite GNN and its limit on an infinite graph, as well as the distance between the GNN on graphs of different sizes that share structural properties, under a regularity assumption verified for various graph sequences. Our results hold for dense and sparse graphs, and various notions of graph limits."
                    },
                    {
                        "title": "Sample Complexity Bounds for Estimating Probability Divergences under Invariances",
                        "abstract": "Group-invariant probability distributions appear in many data-generative models in machine learning, such as graphs, point clouds, and images. In practice, one often needs to estimate divergences between such distributions. In this work, we study how the inherent invariances, with respect to any smooth action of a Lie group on a manifold, improve sample complexity when estimating the 1-Wasserstein distance, the Sobolev Integral Probability Metrics (Sobolev IPMs), the Maximum Mean Discrepancy (MMD), and also the complexity of the density estimation problem (in the $L^2$ and $L^\\infty$ distance). Our results indicate a two-fold gain: (1) reducing the sample complexity by a multiplicative factor corresponding to the group size (for finite groups) or the normalized volume of the quotient space (for groups of positive dimension); (2) improving the exponent in the convergence rate (for groups of positive dimension). These results are completely new for groups of positive dimension and extend recent bounds for finite group actions."
                    },
                    {
                        "title": "Approximation Algorithms for Bregman Co-clustering and Tensor Clustering",
                        "abstract": "In the past few years powerful generalizations to the Euclidean k-means problem have been made, such as Bregman clustering [7], co-clustering (i.e., simultaneous clustering of rows and columns of an input matrix) [9,18], and tensor clustering [8,34]. Like k-means, these more general problems also suffer from the NP-hardness of the associated optimization. Researchers have developed approximation algorithms of varying degrees of sophistication for k-means, k-medians, and more recently also for Bregman clustering [2]. However, there seem to be no approximation algorithms for Bregman co- and tensor clustering. In this paper we derive the first (to our knowledge) guaranteed methods for these increasingly important clustering settings. Going beyond Bregman divergences, we also prove an approximation factor for tensor clustering with arbitrary separable metrics. Through extensive experiments we evaluate the characteristics of our method, and show that it also has practical impact."
                    },
                    {
                        "title": "Graph Cuts with Interacting Edge Costs - Examples, Approximations, and Algorithms",
                        "abstract": "We study an extension of the classical graph cut problem, wherein we replace the modular (sum of edge weights) cost function by a submodular set function defined over graph edges. Special cases of this problem have appeared in different applications in signal processing, machine learning, and computer vision. In this paper, we connect these applications via the generic formulation of \"cooperative graph cuts\", for which we study complexity, algorithms, and connections to polymatroidal network flows. Finally, we compare the proposed algorithms empirically."
                    },
                    {
                        "title": "Structured Optimal Transport",
                        "abstract": "Optimal Transport has recently gained interest in machine learning for applications ranging from domain adaptation, sentence similarities to deep learning. Yet, its ability to capture frequently occurring structure beyond the \"ground metric\" is limited. In this work, we develop a nonlinear generalization of (discrete) optimal transport that is able to reflect much additional structure. We demonstrate how to leverage the geometry of this new model for fast algorithms, and explore connections and properties. Illustrative experiments highlight the benefit of the induced structured couplings for tasks in domain adaptation and natural language processing."
                    },
                    {
                        "title": "The Role of Embedding Complexity in Domain-invariant Representations",
                        "abstract": "Unsupervised domain adaptation aims to generalize the hypothesis trained in a source domain to an unlabeled target domain. One popular approach to this problem is to learn domain-invariant embeddings for both domains. In this work, we study, theoretically and empirically, the effect of the embedding complexity on generalization to the target domain. In particular, this complexity affects an upper bound on the target risk; this is reflected in experiments, too. Next, we specify our theoretical framework to multilayer neural networks. As a result, we develop a strategy that mitigates sensitivity to the embedding complexity, and empirically achieves performance on par with or better than the best layer-dependent complexity tradeoff."
                    },
                    {
                        "title": "Optimal algorithms for group distributionally robust optimization and beyond",
                        "abstract": "Distributionally robust optimization (DRO) can improve the robustness and fairness of learning methods. In this paper, we devise stochastic algorithms for a class of DRO problems including group DRO, subpopulation fairness, and empirical conditional value at risk (CVaR) optimization. Our new algorithms achieve faster convergence rates than existing algorithms for multiple DRO settings. We also provide a new information-theoretic lower bound that implies our bounds are tight for group DRO. Empirically, too, our algorithms outperform known methods"
                    },
                    {
                        "title": "Fast Semidifferential-based Submodular Function Optimization",
                        "abstract": "We present a practical and powerful new framework for both unconstrained and constrained submodular function optimization based on discrete semidifferentials (sub- and super-differentials). The resulting algorithms, which repeatedly compute and then efficiently optimize submodular semigradients, offer new and generalize many old methods for submodular optimization. Our approach, moreover, takes steps towards providing a unifying paradigm applicable to both submodular min- imization and maximization, problems that historically have been treated quite distinctly. The practicality of our algorithms is important since interest in submodularity, owing to its natural and wide applicability, has recently been in ascendance within machine learning. We analyze theoretical properties of our algorithms for minimization and maximization, and show that many state-of-the-art maximization algorithms are special cases. Lastly, we complement our theoretical analyses with supporting empirical experiments."
                    },
                    {
                        "title": "Curvature and Optimal Algorithms for Learning and Minimizing Submodular Functions",
                        "abstract": "We investigate three related and important problems connected to machine learning: approximating a submodular function everywhere, learning a submodular function (in a PAC-like setting [53]), and constrained minimization of submodular functions. We show that the complexity of all three problems depends on the 'curvature' of the submodular function, and provide lower and upper bounds that refine and improve previous results [3, 16, 18, 52]. Our proof techniques are fairly generic. We either use a black-box transformation of the function (for approximation and learning), or a transformation of algorithms to use an appropriate surrogate function (for minimization). Curiously, curvature has been known to influence approximations for submodular maximization [7, 55], but its effect on minimization, approximation and learning has hitherto been open. We complete this picture, and also support our theoretical claims by empirical results."
                    },
                    {
                        "title": "Submodular meets Structured: Finding Diverse Subsets in Exponentially-Large Structured Item Sets",
                        "abstract": "To cope with the high level of ambiguity faced in domains such as Computer Vision or Natural Language processing, robust prediction methods often search for a diverse set of high-quality candidate solutions or proposals. In structured prediction problems, this becomes a daunting task, as the solution space (image labelings, sentence parses, etc.) is exponentially large. We study greedy algorithms for finding a diverse subset of solutions in structured-output spaces by drawing new connections between submodular functions over combinatorial item sets and High-Order Potentials (HOPs) studied for graphical models. Specifically, we show via examples that when marginal gains of submodular diversity functions allow structured representations, this enables efficient (sub-linear time) approximate maximization by reducing the greedy augmentation step to inference in a factor graph with appropriately constructed HOPs. We discuss benefits, tradeoffs, and show that our constructions lead to significantly better proposals."
                    },
                    {
                        "title": "Auxiliary Image Regularization for Deep CNNs with Noisy Labels",
                        "abstract": "Precisely-labeled data sets with sufficient amount of samples are very important for training deep convolutional neural networks (CNNs). However, many of the available real-world data sets contain erroneously labeled samples and those errors substantially hinder the learning of very accurate CNN models. In this work, we consider the problem of training a deep CNN model for image classification with mislabeled training samples - an issue that is common in real image data sets with tags supplied by amateur users. To solve this problem, we propose an auxiliary image regularization technique, optimized by the stochastic Alternating Direction Method of Multipliers (ADMM) algorithm, that automatically exploits the mutual context information among training images and encourages the model to select reliable images to robustify the learning process. Comprehensive experiments on benchmark data sets clearly demonstrate our proposed regularized CNN model is resistant to label noise in training data."
                    },
                    {
                        "title": "Gauss quadrature for matrix inverse forms with applications",
                        "abstract": "We present a framework for accelerating a spectrum of machine learning algorithms that require computation of bilinear inverse forms $u^\\top A^{-1}u$, where $A$ is a positive definite matrix and $u$ a given vector. Our framework is built on Gauss-type quadrature and easily scales to large, sparse matrices. Further, it allows retrospective computation of lower and upper bounds on $u^\\top A^{-1}u$, which in turn accelerates several algorithms. We prove that these bounds tighten iteratively and converge at a linear (geometric) rate. To our knowledge, ours is the first work to demonstrate these key properties of Gauss-type quadrature, which is a classical and deeply studied topic. We illustrate empirical consequences of our results by using quadrature to accelerate machine learning tasks involving determinantal point processes and submodular optimization, and observe tremendous speedups in several instances."
                    }
                ]
            },
            "26d22cdb-deee-48e2-b50b-0bbf2df327f5": {
                "pk": "26d22cdb-deee-48e2-b50b-0bbf2df327f5",
                "name": "Yisen Wang",
                "collaborators": [
                    "Yifei Wang",
                    "Jiansheng Yang",
                    "Zhouchen Lin",
                    "Hongjun Wang",
                    "Qixun Wang",
                    "Qi Zhang",
                    "Liang Huang",
                    "Dongxian Wu",
                    "Zhirui Wang",
                    "Hong Zhu"
                ],
                "domain": [
                    "Adversarial Training",
                    "Deep Learning",
                    "Graph Neural Network",
                    "Out-of-Distribution Generalization"
                ],
                "publications": [
                    {
                        "title": "Symmetry Hierarchy and Thermalization Frustration in Graphene Nanoresonators",
                        "abstract": "As the essential cause of the intrinsic dissipation that limits the quality of graphene nanoresonators, intermodal energy transfer is also a key issue in thermalization dynamics. Typically systems with larger initial energy demand shorter time to be thermalized. However, we find quantitatively that instead of becoming shorter, the equipartition time of the graphene nanoresonator can increase abruptly by one order of magnitude. This thermalization frustration emerges due to the partition of the normal modes based on the hierarchical symmetry, and a sensitive on-off switching of the energy flow channels between symmetry classes controlled by Mathieu instabilities. The results uncover the decisive roles of symmetry in the thermalization at the nanoscale, and may also lead to strategies for improving the performance of graphene nanoresonators."
                    },
                    {
                        "title": "Generalist: Decoupling Natural and Robust Generalization",
                        "abstract": "Deep neural networks obtained by standard training have been constantly plagued by adversarial examples. Although adversarial training demonstrates its capability to defend against adversarial examples, unfortunately, it leads to an inevitable drop in the natural generalization. To address the issue, we decouple the natural generalization and the robust generalization from joint training and formulate different training strategies for each one. Specifically, instead of minimizing a global loss on the expectation over these two generalization errors, we propose a bi-expert framework called \\emph{Generalist} where we simultaneously train base learners with task-aware strategies so that they can specialize in their own fields. The parameters of base learners are collected and combined to form a global learner at intervals during the training process. The global learner is then distributed to the base learners as initialized parameters for continued training. Theoretically, we prove that the risks of Generalist will get lower once the base learners are well trained. Extensive experiments verify the applicability of Generalist to achieve high accuracy on natural examples while maintaining considerable robustness to adversarial ones. Code is available at https://github.com/PKU-ML/Generalist."
                    },
                    {
                        "title": "Self-Ensemble Adversarial Training for Improved Robustness",
                        "abstract": "Due to numerous breakthroughs in real-world applications brought by machine intelligence, deep neural networks (DNNs) are widely employed in critical applications. However, predictions of DNNs are easily manipulated with imperceptible adversarial perturbations, which impedes the further deployment of DNNs and may result in profound security and privacy implications. By incorporating adversarial samples into the training data pool, adversarial training is the strongest principled strategy against various adversarial attacks among all sorts of defense methods. Recent works mainly focus on developing new loss functions or regularizers, attempting to find the unique optimal point in the weight space. But none of them taps the potentials of classifiers obtained from standard adversarial training, especially states on the searching trajectory of training. In this work, we are dedicated to the weight states of models through the training process and devise a simple but powerful \\emph{Self-Ensemble Adversarial Training} (SEAT) method for yielding a robust classifier by averaging weights of history models. This considerably improves the robustness of the target model against several well known adversarial attacks, even merely utilizing the naive cross-entropy loss to supervise. We also discuss the relationship between the ensemble of predictions from different adversarially trained models and the prediction of weight-ensembled models, as well as provide theoretical and empirical evidence that the proposed self-ensemble method provides a smoother loss landscape and better robustness than both individual models and the ensemble of predictions from different classifiers. We further analyze a subtle but fatal issue in the general settings for the self-ensemble model, which causes the deterioration of the weight-ensembled method in the late phases."
                    },
                    {
                        "title": "Adversarial Neuron Pruning Purifies Backdoored Deep Models",
                        "abstract": "As deep neural networks (DNNs) are growing larger, their requirements for computational resources become huge, which makes outsourcing training more popular. Training in a third-party platform, however, may introduce potential risks that a malicious trainer will return backdoored DNNs, which behave normally on clean samples but output targeted misclassifications whenever a trigger appears at the test time. Without any knowledge of the trigger, it is difficult to distinguish or recover benign DNNs from backdoored ones. In this paper, we first identify an unexpected sensitivity of backdoored DNNs, that is, they are much easier to collapse and tend to predict the target label on clean samples when their neurons are adversarially perturbed. Based on these observations, we propose a novel model repairing method, termed Adversarial Neuron Pruning (ANP), which prunes some sensitive neurons to purify the injected backdoor. Experiments show, even with only an extremely small amount of clean data (e.g., 1%), ANP effectively removes the injected backdoor without causing obvious performance degradation."
                    },
                    {
                        "title": "Reparameterized Sampling for Generative Adversarial Networks",
                        "abstract": "Recently, sampling methods have been successfully applied to enhance the sample quality of Generative Adversarial Networks (GANs). However, in practice, they typically have poor sample efficiency because of the independent proposal sampling from the generator. In this work, we propose REP-GAN, a novel sampling method that allows general dependent proposals by REParameterizing the Markov chains into the latent space of the generator. Theoretically, we show that our reparameterized proposal admits a closed-form Metropolis-Hastings acceptance ratio. Empirically, extensive experiments on synthetic and real datasets demonstrate that our REP-GAN largely improves the sample efficiency and obtains better sample quality simultaneously."
                    },
                    {
                        "title": "Fooling Adversarial Training with Inducing Noise",
                        "abstract": "Adversarial training is widely believed to be a reliable approach to improve model robustness against adversarial attack. However, in this paper, we show that when trained on one type of poisoned data, adversarial training can also be fooled to have catastrophic behavior, e.g., $<1\\%$ robust test accuracy with $>90\\%$ robust training accuracy on CIFAR-10 dataset. Previously, there are other types of noise poisoned in the training data that have successfully fooled standard training ($15.8\\%$ standard test accuracy with $99.9\\%$ standard training accuracy on CIFAR-10 dataset), but their poisonings can be easily removed when adopting adversarial training. Therefore, we aim to design a new type of inducing noise, named ADVIN, which is an irremovable poisoning of training data. ADVIN can not only degrade the robustness of adversarial training by a large margin, for example, from $51.7\\%$ to $0.57\\%$ on CIFAR-10 dataset, but also be effective for fooling standard training ($13.1\\%$ standard test accuracy with $100\\%$ standard training accuracy). Additionally, ADVIN can be applied to preventing personal data (like selfies) from being exploited without authorization under whether standard or adversarial training."
                    },
                    {
                        "title": "Improving Out-of-Distribution Generalization by Adversarial Training with Structured Priors",
                        "abstract": "Deep models often fail to generalize well in test domains when the data distribution differs from that in the training domain. Among numerous approaches to address this Out-of-Distribution (OOD) generalization problem, there has been a growing surge of interest in exploiting Adversarial Training (AT) to improve OOD performance. Recent works have revealed that the robust model obtained by conducting sample-wise AT also retains transferability to biased test domains. In this paper, we empirically show that sample-wise AT has limited improvement on OOD performance. Specifically, we find that AT can only maintain performance at smaller scales of perturbation while Universal AT (UAT) is more robust to larger-scale perturbations. This provides us with clues that adversarial perturbations with universal (low dimensional) structures can enhance the robustness against large data distribution shifts that are common in OOD scenarios. Inspired by this, we propose two AT variants with low-rank structures to train OOD-robust models. Extensive experiments on DomainBed benchmark show that our proposed approaches outperform Empirical Risk Minimization (ERM) and sample-wise AT. Our code is available at https://github.com/NOVAglow646/NIPS22-MAT-and-LDAT-for-OOD."
                    },
                    {
                        "title": "Can In-context Learning Really Generalize to Out-of-distribution Tasks?",
                        "abstract": "In this work, we explore the mechanism of in-context learning (ICL) on out-of-distribution (OOD) tasks that were not encountered during training. To achieve this, we conduct synthetic experiments where the objective is to learn OOD mathematical functions through ICL using a GPT-2 model. We reveal that Transformers may struggle to learn OOD task functions through ICL. Specifically, ICL performance resembles implementing a function within the pretraining hypothesis space and optimizing it with gradient descent based on the in-context examples. Additionally, we investigate ICL's well-documented ability to learn unseen abstract labels in context. We demonstrate that such ability only manifests in the scenarios without distributional shifts and, therefore, may not serve as evidence of new-task-learning ability. Furthermore, we assess ICL's performance on OOD tasks when the model is pretrained on multiple tasks. Both empirical and theoretical analyses demonstrate the existence of the \\textbf{low-test-error preference} of ICL, where it tends to implement the pretraining function that yields low test error in the testing context. We validate this through numerical experiments. This new theoretical result, combined with our empirical findings, elucidates the mechanism of ICL in addressing OOD tasks."
                    },
                    {
                        "title": "How to generate all possible rational Wilf-Zeilberger forms?",
                        "abstract": "Wilf-Zeilberger pairs are fundamental in the algorithmic theory of Wilf and Zeilberger for computer-generated proofs of combinatorial identities. Wilf-Zeilberger forms are their high-dimensional generalizations, which can be used for proving and discovering convergence acceleration formulas. This paper presents a structural description of all possible rational such forms, which can be viewed as an additive analog of the classical Ore-Sato theorem. Based on this analog, we show a structural decomposition of so-called multivariate hyperarithmetic terms, which extend multivariate hypergeometric terms to the additive setting."
                    },
                    {
                        "title": "Dissecting the Diffusion Process in Linear Graph Convolutional Networks",
                        "abstract": "Graph Convolutional Networks (GCNs) have attracted more and more attentions in recent years. A typical GCN layer consists of a linear feature propagation step and a nonlinear transformation step. Recent works show that a linear GCN can achieve comparable performance to the original non-linear GCN while being much more computationally efficient. In this paper, we dissect the feature propagation steps of linear GCNs from a perspective of continuous graph diffusion, and analyze why linear GCNs fail to benefit from more propagation steps. Following that, we propose Decoupled Graph Convolution (DGC) that decouples the terminal time and the feature propagation steps, making it more flexible and capable of exploiting a very large number of feature propagation steps. Experiments demonstrate that our proposed DGC improves linear GCNs by a large margin and makes them competitive with many modern variants of non-linear GCNs."
                    },
                    {
                        "title": "A Unified Contrastive Energy-based Model for Understanding the Generative Ability of Adversarial Training",
                        "abstract": "Adversarial Training (AT) is known as an effective approach to enhance the robustness of deep neural networks. Recently researchers notice that robust models with AT have good generative ability and can synthesize realistic images, while the reason behind it is yet under-explored. In this paper, we demystify this phenomenon by developing a unified probabilistic framework, called Contrastive Energy-based Models (CEM). On the one hand, we provide the first probabilistic characterization of AT through a unified understanding of robustness and generative ability. On the other hand, our unified framework can be extended to the unsupervised scenario, which interprets unsupervised contrastive learning as an important sampling of CEM. Based on these, we propose a principled method to develop adversarial learning and sampling methods. Experiments show that the sampling methods derived from our framework improve the sample quality in both supervised and unsupervised learning. Notably, our unsupervised adversarial sampling method achieves an Inception score of 9.61 on CIFAR-10, which is superior to previous energy-based models and comparable to state-of-the-art generative models."
                    },
                    {
                        "title": "Non-negative Contrastive Learning",
                        "abstract": "Deep representations have shown promising performance when transferred to downstream tasks in a black-box manner. Yet, their inherent lack of interpretability remains a significant challenge, as these features are often opaque to human understanding. In this paper, we propose Non-negative Contrastive Learning (NCL), a renaissance of Non-negative Matrix Factorization (NMF) aimed at deriving interpretable features. The power of NCL lies in its enforcement of non-negativity constraints on features, reminiscent of NMF's capability to extract features that align closely with sample clusters. NCL not only aligns mathematically well with an NMF objective but also preserves NMF's interpretability attributes, resulting in a more sparse and disentangled representation compared to standard contrastive learning (CL). Theoretically, we establish guarantees on the identifiability and downstream generalization of NCL. Empirically, we show that these advantages enable NCL to outperform CL significantly on feature disentanglement, feature selection, as well as downstream classification tasks. At last, we show that NCL can be easily extended to other learning scenarios and benefit supervised learning as well. Code is available at https://github.com/PKU-ML/non_neg."
                    },
                    {
                        "title": "Do Generated Data Always Help Contrastive Learning?",
                        "abstract": "Contrastive Learning (CL) has emerged as one of the most successful paradigms for unsupervised visual representation learning, yet it often depends on intensive manual data augmentations. With the rise of generative models, especially diffusion models, the ability to generate realistic images close to the real data distribution has been well recognized. These generated high-equality images have been successfully applied to enhance contrastive representation learning, a technique termed ``data inflation''. However, we find that the generated data (even from a good diffusion model like DDPM) may sometimes even harm contrastive learning. We investigate the causes behind this failure from the perspective of both data inflation and data augmentation. For the first time, we reveal the complementary roles that stronger data inflation should be accompanied by weaker augmentations, and vice versa. We also provide rigorous theoretical explanations for these phenomena via deriving its generalization bounds under data inflation. Drawing from these insights, we propose Adaptive Inflation (AdaInf), a purely data-centric strategy without introducing any extra computation cost. On benchmark datasets, AdaInf can bring significant improvements for various contrastive learning methods. Notably, without using external data, AdaInf obtains 94.70% linear accuracy on CIFAR-10 with SimCLR, setting a new record that surpasses many sophisticated methods. Code is available at https://github.com/PKU-ML/adainf."
                    },
                    {
                        "title": "How to Craft Backdoors with Unlabeled Data Alone?",
                        "abstract": "Relying only on unlabeled data, Self-supervised learning (SSL) can learn rich features in an economical and scalable way. As the drive-horse for building foundation models, SSL has received a lot of attention recently with wide applications, which also raises security concerns where backdoor attack is a major type of threat: if the released dataset is maliciously poisoned, backdoored SSL models can behave badly when triggers are injected to test samples. The goal of this work is to investigate this potential risk. We notice that existing backdoors all require a considerable amount of \\emph{labeled} data that may not be available for SSL. To circumvent this limitation, we explore a more restrictive setting called no-label backdoors, where we only have access to the unlabeled data alone, where the key challenge is how to select the proper poison set without using label information. We propose two strategies for poison selection: clustering-based selection using pseudolabels, and contrastive selection derived from the mutual information principle. Experiments on CIFAR-10 and ImageNet-100 show that both no-label backdoors are effective on many SSL methods and outperform random poisoning by a large margin. Code will be available at https://github.com/PKU-ML/nlb."
                    },
                    {
                        "title": "CycleHOI: Improving Human-Object Interaction Detection with Cycle Consistency of Detection and Generation",
                        "abstract": "Recognition and generation are two fundamental tasks in computer vision, which are often investigated separately in the exiting literature. However, these two tasks are highly correlated in essence as they both require understanding the underline semantics of visual concepts. In this paper, we propose a new learning framework, coined as CycleHOI, to boost the performance of human-object interaction (HOI) detection by bridging the DETR-based detection pipeline and the pre-trained text-to-image diffusion model. Our key design is to introduce a novel cycle consistency loss for the training of HOI detector, which is able to explicitly leverage the knowledge captured in the powerful diffusion model to guide the HOI detector training. Specifically, we build an extra generation task on top of the decoded instance representations from HOI detector to enforce a detection-generation cycle consistency. Moreover, we perform feature distillation from diffusion model to detector encoder to enhance its representation power. In addition, we further utilize the generation power of diffusion model to augment the training set in both aspects of label correction and sample generation. We perform extensive experiments to verify the effectiveness and generalization power of our CycleHOI with three HOI detection frameworks on two public datasets: HICO-DET and V-COCO. The experimental results demonstrate our CycleHOI can significantly improve the performance of the state-of-the-art HOI detectors."
                    },
                    {
                        "title": "Unifying Decision Trees Split Criteria Using Tsallis Entropy",
                        "abstract": "The construction of efficient and effective decision trees remains a key topic in machine learning because of their simplicity and flexibility. A lot of heuristic algorithms have been proposed to construct near-optimal decision trees. ID3, C4.5 and CART are classical decision tree algorithms and the split criteria they used are Shannon entropy, Gain Ratio and Gini index respectively. All the split criteria seem to be independent, actually, they can be unified in a Tsallis entropy framework. Tsallis entropy is a generalization of Shannon entropy and provides a new approach to enhance decision trees' performance with an adjustable parameter $q$. In this paper, a Tsallis Entropy Criterion (TEC) algorithm is proposed to unify Shannon entropy, Gain Ratio and Gini index, which generalizes the split criteria of decision trees. More importantly, we reveal the relations between Tsallis entropy with different $q$ and other split criteria. Experimental results on UCI data sets indicate that the TEC algorithm achieves statistically significant improvement over the classical algorithms."
                    },
                    {
                        "title": "Residual Convolutional CTC Networks for Automatic Speech Recognition",
                        "abstract": "Deep learning approaches have been widely used in Automatic Speech Recognition (ASR) and they have achieved a significant accuracy improvement. Especially, Convolutional Neural Networks (CNNs) have been revisited in ASR recently. However, most CNNs used in existing work have less than 10 layers which may not be deep enough to capture all human speech signal information. In this paper, we propose a novel deep and wide CNN architecture denoted as RCNN-CTC, which has residual connections and Connectionist Temporal Classification (CTC) loss function. RCNN-CTC is an end-to-end system which can exploit temporal and spectral structures of speech signals simultaneously. Furthermore, we introduce a CTC-based system combination, which is different from the conventional frame-wise senone-based one. The basic subsystems adopted in the combination are different types and thus mutually complementary to each other. Experimental results show that our proposed single system RCNN-CTC can achieve the lowest word error rate (WER) on WSJ and Tencent Chat data sets, compared to several widely used neural network systems in ASR. In addition, the proposed system combination can offer a further error reduction on these two data sets, resulting in relative WER reductions of $14.91\\%$ and $6.52\\%$ on WSJ dev93 and Tencent Chat data sets respectively."
                    },
                    {
                        "title": "Generalization in Deep RL for TSP Problems via Equivariance and Local Search",
                        "abstract": "Deep reinforcement learning (RL) has proved to be a competitive heuristic for solving small-sized instances of traveling salesman problems (TSP), but its performance on larger-sized instances is insufficient. Since training on large instances is impractical, we design a novel deep RL approach with a focus on generalizability. Our proposition consisting of a simple deep learning architecture that learns with novel RL training techniques, exploits two main ideas. First, we exploit equivariance to facilitate training. Second, we interleave efficient local search heuristics with the usual RL training to smooth the value landscape. In order to validate the whole approach, we empirically evaluate our proposition on random and realistic TSP problems against relevant state-of-the-art deep RL methods. Moreover, we present an ablation study to understand the contribution of each of its component"
                    },
                    {
                        "title": "How Mask Matters: Towards Theoretical Understandings of Masked Autoencoders",
                        "abstract": "Masked Autoencoders (MAE) based on a reconstruction task have risen to be a promising paradigm for self-supervised learning (SSL) and achieve state-of-the-art performance across different benchmark datasets. However, despite its impressive empirical success, there is still limited theoretical understanding of it. In this paper, we propose a theoretical understanding of how masking matters for MAE to learn meaningful features. We establish a close connection between MAE and contrastive learning, which shows that MAE implicit aligns the mask-induced positive pairs. Built upon this connection, we develop the first downstream guarantees for MAE methods, and analyze the effect of mask ratio. Besides, as a result of the implicit alignment, we also point out the dimensional collapse issue of MAE, and propose a Uniformity-enhanced MAE (U-MAE) loss that can effectively address this issue and bring significant improvements on real-world datasets, including CIFAR-10, ImageNet-100, and ImageNet-1K. Code is available at (https://github.com/zhangq327/U-MAE)."
                    },
                    {
                        "title": "On the Connection between Invariant Learning and Adversarial Training for Out-of-Distribution Generalization",
                        "abstract": "Despite impressive success in many tasks, deep learning models are shown to rely on spurious features, which will catastrophically fail when generalized to out-of-distribution (OOD) data. Invariant Risk Minimization (IRM) is proposed to alleviate this issue by extracting domain-invariant features for OOD generalization. Nevertheless, recent work shows that IRM is only effective for a certain type of distribution shift (e.g., correlation shift) while it fails for other cases (e.g., diversity shift). Meanwhile, another thread of method, Adversarial Training (AT), has shown better domain transfer performance, suggesting that it has the potential to be an effective candidate for extracting domain-invariant features. This paper investigates this possibility by exploring the similarity between the IRM and AT objectives. Inspired by this connection, we propose Domainwise Adversarial Training (DAT), an AT-inspired method for alleviating distribution shift by domain-specific perturbations. Extensive experiments show that our proposed DAT can effectively remove domain-varying features and improve OOD generalization under both correlation shift and diversity shift."
                    }
                ]
            }
        }
    },
    "2406.03537": {
        "paper_data": {
            "title": "A Geometric View of Data Complexity: Efficient Local Intrinsic Dimension Estimation with Diffusion Models",
            "url": "http://arxiv.org/abs/2406.03537v1",
            "arxiv_id": "2406.03537",
            "authors": [
                "Hamidreza Kamkari",
                "Brendan Leigh Ross",
                "Rasa Hosseinzadeh",
                "Jesse C. Cresswell",
                "Gabriel Loaiza-Ganem"
            ],
            "abstract": "High-dimensional data commonly lies on low-dimensional submanifolds, and estimating the local intrinsic dimension (LID) of a datum -- i.e. the dimension of the submanifold it belongs to -- is a longstanding problem. LID can be understood as the number of local factors of variation: the more factors of variation a datum has, the more complex it tends to be. Estimating this quantity has proven useful in contexts ranging from generalization in neural networks to detection of out-of-distribution data, adversarial examples, and AI-generated text. The recent successes of deep generative models present an opportunity to leverage them for LID estimation, but current methods based on generative models produce inaccurate estimates, require more than a single pre-trained model, are computationally intensive, or do not exploit the best available deep generative models, i.e. diffusion models (DMs). In this work, we show that the Fokker-Planck equation associated with a DM can provide a LID estimator which addresses all the aforementioned deficiencies. Our estimator, called FLIPD, is compatible with all popular DMs, and outperforms existing baselines on LID estimation benchmarks. We also apply FLIPD on natural images where the true LID is unknown. Compared to competing estimators, FLIPD exhibits a higher correlation with non-LID measures of complexity, better matches a qualitative assessment of complexity, and is the only estimator to remain tractable with high-resolution images at the scale of Stable Diffusion.",
            "introduction": "   1 Introduction  The manifold hypothesis [7], which has been empirically verified in contexts ranging from natural images [47, 10] to calorimeter showers in physics [14], states that high-dimensional data of interest in \u211dDsuperscript\u211d\ud835\udc37\\mathbb{R}^{D}blackboard_R start_POSTSUPERSCRIPT italic_D end_POSTSUPERSCRIPT often lies on low-dimensional submanifolds of \u211dDsuperscript\u211d\ud835\udc37\\mathbb{R}^{D}blackboard_R start_POSTSUPERSCRIPT italic_D end_POSTSUPERSCRIPT. For a given datum x\u2208\u211dD\ud835\udc65superscript\u211d\ud835\udc37x\\in\\mathbb{R}^{D}italic_x \u2208 blackboard_R start_POSTSUPERSCRIPT italic_D end_POSTSUPERSCRIPT, this hypothesis motivates using its local intrinsic dimension (LID), denoted LID\u2062(x)LID\ud835\udc65\\text{LID}(x)LID ( italic_x ), as a natural measure of its complexity. LID\u2062(x)LID\ud835\udc65\\text{LID}(x)LID ( italic_x ) corresponds to the dimension of the data manifold that x\ud835\udc65xitalic_x belongs to, and can be intuitively understood as the minimal number of variables needed to describe x\ud835\udc65xitalic_x. More complex data needs more variables to be adequately described, as illustrated in Figure\u00a01. Data manifolds are typically not known explicitly, meaning that LID must be estimated. Here we tackle the following problem: given a dataset along with a query datum x\ud835\udc65xitalic_x, how can we tractably estimate LID\u2062(x)LID\ud835\udc65\\text{LID}(x)LID ( italic_x )?   This is a longstanding problem, with LID estimates being highly useful due to their innate interpretation as a measure of complexity. For example, these estimates can be used to detect outliers [26, 2, 31], AI-generated text [61], and adversarial examples [41]. Connections between the generalization achieved by a neural network and the LID estimates of its internal representations have also been shown [4, 8, 43, 9]. These insights can be leveraged to identify which representations contain maximal semantic content [63], and help explain why LID estimates can be helpful as regularizers [69] and for pruning large models [67]. LID estimation is thus not only of mathematical and statistical interest, but can also benefit the empirical performance of deep learning models at numerous tasks.   Figure 1: (Left) A cartoon illustration showing that LID is a natural measure of relative complexity. We depict two manifolds of MNIST digits, corresponding to 1s and 8s, as 1-dimensional and 2-dimensional submanifolds of \u211d3superscript\u211d3\\mathbb{R}^{3}blackboard_R start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT, respectively. The relatively simpler manifold of 1s exhibits a single factor of variation (\u201ctilt\u201d), whereas 8s have an additional factor of variation (\u201cdisproportionality\u201d). (Right) The 4 lowest- and highest-LID datapoints from a subsample of LAION Aesthetics, as measured by our method, FLIPD, applied to Stable Diffusion v1.5. FLIPD scales efficiently to large models on high-dimensional data, and aligns closely with subjective complexity.   Traditional estimators of intrinsic dimension [21, 37, 42, 11, 29, 19, 1, 5] typically rely on pairwise distances and nearest neighbours, so computing them is prohibitively expensive for large datasets. Recent work has thus sought to move away from these model-free estimators and instead take advantage of deep generative models which learn the distribution of observed data. When this distribution is supported on low-dimensional submanifolds of \u211dDsuperscript\u211d\ud835\udc37\\mathbb{R}^{D}blackboard_R start_POSTSUPERSCRIPT italic_D end_POSTSUPERSCRIPT, successful generative models must implicitly learn the dimensions of the data submanifolds, suggesting they can be used to construct LID estimators. However, existing model-based estimators suffer from various drawbacks, including being inaccurate and computationally expensive [57], not leveraging the best existing generative models [60, 68] (i.e.\u00a0diffusion models [54, 24, 56]), and requiring training several models [60] or altering the training procedure rather than",
            "references": [
                {
                    "title": "Deep Generative Models through the Lens of the Manifold Hypothesis: A Survey and New Connections",
                    "abstract": "In recent years there has been increased interest in understanding the interplay between deep generative models (DGMs) and the manifold hypothesis. Research in this area focuses on understanding the reasons why commonly-used DGMs succeed or fail at learning distributions supported on unknown low-dimensional manifolds, as well as developing new models explicitly designed to account for manifold-supported data. This manifold lens provides both clarity as to why some DGMs (e.g. diffusion models and some generative adversarial networks) empirically surpass others (e.g. likelihood-based models such as variational autoencoders, normalizing flows, or energy-based models) at sample generation, and guidance for devising more performant DGMs. We carry out the first survey of DGMs viewed through this lens, making two novel contributions along the way. First, we formally establish that numerical instability of likelihoods in high ambient dimensions is unavoidable when modelling data with low intrinsic dimension. We then show that DGMs on learned representations of autoencoders can be interpreted as approximately minimizing Wasserstein distance: this result, which applies to latent diffusion models, helps justify their outstanding empirical results. The manifold lens provides a rich perspective from which to understand DGMs, and we aim to make this perspective more accessible and widespread."
                },
                {
                    "title": "A Geometric Explanation of the Likelihood OOD Detection Paradox",
                    "abstract": "Likelihood-based deep generative models (DGMs) commonly exhibit a puzzling behaviour: when trained on a relatively complex dataset, they assign higher likelihood values to out-of-distribution (OOD) data from simpler sources. Adding to the mystery, OOD samples are never generated by these DGMs despite having higher likelihoods. This two-pronged paradox has yet to be conclusively explained, making likelihood-based OOD detection unreliable. Our primary observation is that high-likelihood regions will not be generated if they contain minimal probability mass. We demonstrate how this seeming contradiction of large densities yet low probability mass can occur around data confined to low-dimensional manifolds. We also show that this scenario can be identified through local intrinsic dimension (LID) estimation, and propose a method for OOD detection which pairs the likelihoods and LID estimates obtained from a pre-trained DGM. Our method can be applied to normalizing flows and score-based diffusion models, and obtains results which match or surpass state-of-the-art OOD detection benchmarks using the same DGM backbones. Our code is available at https://github.com/layer6ai-labs/dgm_ood_detection."
                },
                {
                    "title": "On gauge freedom, conservativity and intrinsic dimensionality estimation in diffusion models",
                    "abstract": "Diffusion models are generative models that have recently demonstrated impressive performances in terms of sampling quality and density estimation in high dimensions. They rely on a forward continuous diffusion process and a backward continuous denoising process, which can be described by a time-dependent vector field and is used as a generative model. In the original formulation of the diffusion model, this vector field is assumed to be the score function (i.e. it is the gradient of the log-probability at a given time in the diffusion process). Curiously, on the practical side, most studies on diffusion models implement this vector field as a neural network function and do not constrain it be the gradient of some energy function (that is, most studies do not constrain the vector field to be conservative). Even though some studies investigated empirically whether such a constraint will lead to a performance gain, they lead to contradicting results and failed to provide analytical results. Here, we provide three analytical results regarding the extent of the modeling freedom of this vector field. {Firstly, we propose a novel decomposition of vector fields into a conservative component and an orthogonal component which satisfies a given (gauge) freedom. Secondly, from this orthogonal decomposition, we show that exact density estimation and exact sampling is achieved when the conservative component is exactly equals to the true score and therefore conservativity is neither necessary nor sufficient to obtain exact density estimation and exact sampling. Finally, we show that when it comes to inferring local information of the data manifold, constraining the vector field to be conservative is desirable."
                },
                {
                    "title": "Order-based Structure Learning with Normalizing Flows",
                    "abstract": "Estimating the causal structure of observational data is a challenging combinatorial search problem that scales super-exponentially with graph size. Existing methods use continuous relaxations to make this problem computationally tractable but often restrict the data-generating process to additive noise models (ANMs) through explicit or implicit assumptions. We present Order-based Structure Learning with Normalizing Flows (OSLow), a framework that relaxes these assumptions using autoregressive normalizing flows. We leverage the insight that searching over topological orderings is a natural way to enforce acyclicity in structure discovery and propose a novel, differentiable permutation learning method to find such orderings. Through extensive experiments on synthetic and real-world data, we demonstrate that OSLow outperforms prior baselines and improves performance on the observational Sachs and SynTReN datasets as measured by structural hamming distance and structural intervention distance, highlighting the importance of relaxing the ANM assumption made by existing methods."
                },
                {
                    "title": "Intrinsic Dimension Estimation for Robust Detection of AI-Generated Texts",
                    "abstract": "Rapidly increasing quality of AI-generated content makes it difficult to distinguish between human and AI-generated texts, which may lead to undesirable consequences for society. Therefore, it becomes increasingly important to study the properties of human texts that are invariant over different text domains and varying proficiency of human writers, can be easily calculated for any language, and can robustly separate natural and AI-generated texts regardless of the generation model and sampling method. In this work, we propose such an invariant for human-written texts, namely the intrinsic dimensionality of the manifold underlying the set of embeddings for a given text sample. We show that the average intrinsic dimensionality of fluent texts in a natural language is hovering around the value $9$ for several alphabet-based languages and around $7$ for Chinese, while the average intrinsic dimensionality of AI-generated texts for each language is $\\approx 1.5$ lower, with a clear statistical separation between human-generated and AI-generated distributions. This property allows us to build a score-based artificial text detector. The proposed detector's accuracy is stable over text domains, generator models, and human writer proficiency levels, outperforming SOTA detectors in model-agnostic and cross-domain scenarios by a significant margin."
                },
                {
                    "title": "Learning Manifold Dimensions with Conditional Variational Autoencoders",
                    "abstract": "Although the variational autoencoder (VAE) and its conditional extension (CVAE) are capable of state-of-the-art results across multiple domains, their precise behavior is still not fully understood, particularly in the context of data (like images) that lie on or near a low-dimensional manifold. For example, while prior work has suggested that the globally optimal VAE solution can learn the correct manifold dimension, a necessary (but not sufficient) condition for producing samples from the true data distribution, this has never been rigorously proven. Moreover, it remains unclear how such considerations would change when various types of conditioning variables are introduced, or when the data support is extended to a union of manifolds (e.g., as is likely the case for MNIST digits and related). In this work, we address these points by first proving that VAE global minima are indeed capable of recovering the correct manifold dimension. We then extend this result to more general CVAEs, demonstrating practical scenarios whereby the conditioning variables allow the model to adaptively learn manifolds of varying dimension across samples. Our analyses, which have practical implications for various CVAE design choices, are also supported by numerical results on both synthetic and real-world datasets."
                },
                {
                    "title": "The geometry of hidden representations of large transformer models",
                    "abstract": "Large transformers are powerful architectures used for self-supervised data analysis across various data types, including protein sequences, images, and text. In these models, the semantic structure of the dataset emerges from a sequence of transformations between one representation and the next. We characterize the geometric and statistical properties of these representations and how they change as we move through the layers. By analyzing the intrinsic dimension (ID) and neighbor composition, we find that the representations evolve similarly in transformers trained on protein language tasks and image reconstruction tasks. In the first layers, the data manifold expands, becoming high-dimensional, and then contracts significantly in the intermediate layers. In the last part of the model, the ID remains approximately constant or forms a second shallow peak. We show that the semantic information of the dataset is better expressed at the end of the first peak, and this phenomenon can be observed across many models trained on diverse datasets. Based on our findings, we point out an explicit strategy to identify, without supervision, the layers that maximize semantic content: representations at intermediate layers corresponding to a relative minimum of the ID profile are more suitable for downstream learning tasks."
                },
                {
                    "title": "Mathematical analysis of singularities in the diffusion model under the submanifold assumption",
                    "abstract": "This paper provide several mathematical analyses of the diffusion model in machine learning. The drift term of the backwards sampling process is represented as a conditional expectation involving the data distribution and the forward diffusion. The training process aims to find such a drift function by minimizing the mean-squared residue related to the conditional expectation. Using small-time approximations of the Green's function of the forward diffusion, we show that the analytical mean drift function in DDPM and the score function in SGM asymptotically blow up in the final stages of the sampling process for singular data distributions such as those concentrated on lower-dimensional manifolds, and is therefore difficult to approximate by a network. To overcome this difficulty, we derive a new target function and associated loss, which remains bounded even for singular data distributions. We illustrate the theoretical findings with several numerical examples."
                },
                {
                    "title": "Your diffusion model secretly knows the dimension of the data manifold",
                    "abstract": "In this work, we propose a novel framework for estimating the dimension of the data manifold using a trained diffusion model. A diffusion model approximates the score function i.e. the gradient of the log density of a noise-corrupted version of the target distribution for varying levels of corruption. We prove that, if the data concentrates around a manifold embedded in the high-dimensional ambient space, then as the level of corruption decreases, the score function points towards the manifold, as this direction becomes the direction of maximal likelihood increase. Therefore, for small levels of corruption, the diffusion model provides us with access to an approximation of the normal bundle of the data manifold. This allows us to estimate the dimension of the tangent space, thus, the intrinsic dimension of the data manifold. To the best of our knowledge, our method is the first estimator of the data manifold dimension based on diffusion models and it outperforms well established statistical estimators in controlled experiments on both Euclidean and image data."
                },
                {
                    "title": "CaloMan: Fast generation of calorimeter showers with density estimation on learned manifolds",
                    "abstract": "Precision measurements and new physics searches at the Large Hadron Collider require efficient simulations of particle propagation and interactions within the detectors. The most computationally expensive simulations involve calorimeter showers. Advances in deep generative modelling - particularly in the realm of high-dimensional data - have opened the possibility of generating realistic calorimeter showers orders of magnitude more quickly than physics-based simulation. However, the high-dimensional representation of showers belies the relative simplicity and structure of the underlying physical laws. This phenomenon is yet another example of the manifold hypothesis from machine learning, which states that high-dimensional data is supported on low-dimensional manifolds. We thus propose modelling calorimeter showers first by learning their manifold structure, and then estimating the density of data across this manifold. Learning manifold structure reduces the dimensionality of the data, which enables fast training and generation when compared with competing methods."
                },
                {
                    "title": "Relating Regularization and Generalization through the Intrinsic Dimension of Activations",
                    "abstract": "Given a pair of models with similar training set performance, it is natural to assume that the model that possesses simpler internal representations would exhibit better generalization. In this work, we provide empirical evidence for this intuition through an analysis of the intrinsic dimension (ID) of model activations, which can be thought of as the minimal number of factors of variation in the model's representation of the data. First, we show that common regularization techniques uniformly decrease the last-layer ID (LLID) of validation set activations for image classification models and show how this strongly affects generalization performance. We also investigate how excessive regularization decreases a model's ability to extract features from data in earlier layers, leading to a negative effect on validation accuracy even while LLID continues to decrease and training accuracy remains near-perfect. Finally, we examine the LLID over the course of training of models that exhibit grokking. We observe that well after training accuracy saturates, when models ``grok'' and validation accuracy suddenly improves from random to perfect, there is a co-occurent sudden drop in LLID, thus providing more insight into the dynamics of sudden generalization."
                },
                {
                    "title": "LAION-5B: An open large-scale dataset for training next generation image-text models",
                    "abstract": "Groundbreaking language-vision architectures like CLIP and DALL-E proved the utility of training on large amounts of noisy image-text data, without relying on expensive accurate labels used in standard vision unimodal supervised learning. The resulting models showed capabilities of strong text-guided image generation and transfer to downstream tasks, while performing remarkably at zero-shot classification with noteworthy out-of-distribution robustness. Since then, large-scale language-vision models like ALIGN, BASIC, GLIDE, Flamingo and Imagen made further improvements. Studying the training and capabilities of such models requires datasets containing billions of image-text pairs. Until now, no datasets of this size have been made openly available for the broader research community. To address this problem and democratize research on large-scale multi-modal models, we present LAION-5B - a dataset consisting of 5.85 billion CLIP-filtered image-text pairs, of which 2.32B contain English language. We show successful replication and fine-tuning of foundational models like CLIP, GLIDE and Stable Diffusion using the dataset, and discuss further experiments enabled with an openly available dataset of this scale. Additionally we provide several nearest neighbor indices, an improved web-interface for dataset exploration and subset generation, and detection scores for watermark, NSFW, and toxic content detection. Announcement page https://laion.ai/laion-5b-a-new-era-of-open-large-scale-multi-modal-datasets/"
                },
                {
                    "title": "Convergence of denoising diffusion models under the manifold hypothesis",
                    "abstract": "Denoising diffusion models are a recent class of generative models exhibiting state-of-the-art performance in image and audio synthesis. Such models approximate the time-reversal of a forward noising process from a target distribution to a reference density, which is usually Gaussian. Despite their strong empirical results, the theoretical analysis of such models remains limited. In particular, all current approaches crucially assume that the target density admits a density w.r.t. the Lebesgue measure. This does not cover settings where the target distribution is supported on a lower-dimensional manifold or is given by some empirical distribution. In this paper, we bridge this gap by providing the first convergence results for diffusion models in this more general setting. In particular, we provide quantitative bounds on the Wasserstein distance of order one between the target data distribution and the generative distribution of the diffusion model."
                },
                {
                    "title": "Verifying the Union of Manifolds Hypothesis for Image Data",
                    "abstract": "Deep learning has had tremendous success at learning low-dimensional representations of high-dimensional data. This success would be impossible if there was no hidden low-dimensional structure in data of interest; this existence is posited by the manifold hypothesis, which states that the data lies on an unknown manifold of low intrinsic dimension. In this paper, we argue that this hypothesis does not properly capture the low-dimensional structure typically present in image data. Assuming that data lies on a single manifold implies intrinsic dimension is identical across the entire data space, and does not allow for subregions of this space to have a different number of factors of variation. To address this deficiency, we consider the union of manifolds hypothesis, which states that data lies on a disjoint union of manifolds of varying intrinsic dimensions. We empirically verify this hypothesis on commonly-used image datasets, finding that indeed, observed data lies on a disconnected set and that intrinsic dimension is not constant. We also provide insights into the implications of the union of manifolds hypothesis in deep learning, both supervised and unsupervised, showing that designing models with an inductive bias for this structure improves performance across classification and generative modelling tasks. Our code is available at https://github.com/layer6ai-labs/UoMH."
                },
                {
                    "title": "LIDL: Local Intrinsic Dimension Estimation Using Approximate Likelihood",
                    "abstract": "Most of the existing methods for estimating the local intrinsic dimension of a data distribution do not scale well to high-dimensional data. Many of them rely on a non-parametric nearest neighbors approach which suffers from the curse of dimensionality. We attempt to address that challenge by proposing a novel approach to the problem: Local Intrinsic Dimension estimation using approximate Likelihood (LIDL). Our method relies on an arbitrary density estimation method as its subroutine and hence tries to sidestep the dimensionality challenge by making use of the recent progress in parametric neural methods for likelihood estimation. We carefully investigate the empirical properties of the proposed method, compare them with our theoretical predictions, and show that LIDL yields competitive results on the standard benchmarks for this problem and that it scales to thousands of dimensions. What is more, we anticipate this approach to improve further with the continuing advances in the density estimation literature."
                },
                {
                    "title": "Score-Based Generative Models Detect Manifolds",
                    "abstract": "Score-based generative models (SGMs) need to approximate the scores $\\nabla \\log p_t$ of the intermediate distributions as well as the final distribution $p_T$ of the forward process. The theoretical underpinnings of the effects of these approximations are still lacking. We find precise conditions under which SGMs are able to produce samples from an underlying (low-dimensional) data manifold $\\mathcal{M}$. This assures us that SGMs are able to generate the\"right kind of samples\". For example, taking $\\mathcal{M}$ to be the subset of images of faces, we find conditions under which the SGM robustly produces an image of a face, even though the relative frequencies of these images might not accurately represent the true data generating distribution. Moreover, this analysis is a first step towards understanding the generalization properties of SGMs: Taking $\\mathcal{M}$ to be the set of all training samples, our results provide a precise description of when the SGM memorizes its training data."
                },
                {
                    "title": "Topology and geometry of data manifold in deep learning",
                    "abstract": "Despite significant advances in the field of deep learning in applications to various fields, explaining the inner processes of deep learning models remains an important and open question. The purpose of this article is to describe and substantiate the geometric and topological view of the learning process of neural networks. Our attention is focused on the internal representation of neural networks and on the dynamics of changes in the topology and geometry of the data manifold on different layers. We also propose a method for assessing the generalizing ability of neural networks based on topological descriptors. In this paper, we use the concepts of topological data analysis and intrinsic dimension, and we present a wide range of experiments on different datasets and different configurations of convolutional neural network architectures. In addition, we consider the issue of the geometry of adversarial attacks in the classification task and spoofing attacks on face recognition systems. Our work is a contribution to the development of an important area of explainable and interpretable AI through the example of computer vision."
                },
                {
                    "title": "Searching Intrinsic Dimensions of Vision Transformers",
                    "abstract": "It has been shown by many researchers that transformers perform as well as convolutional neural networks in many computer vision tasks. Meanwhile, the large computational costs of its attention module hinder further studies and applications on edge devices. Some pruning methods have been developed to construct efficient vision transformers, but most of them have considered image classification tasks only. Inspired by these results, we propose SiDT, a method for pruning vision transformer backbones on more complicated vision tasks like object detection, based on the search of transformer dimensions. Experiments on CIFAR-100 and COCO datasets show that the backbones with 20\\% or 40\\% dimensions/parameters pruned can have similar or even better performance than the unpruned models. Moreover, we have also provided the complexity analysis and comparisons with the previous pruning methods."
                },
                {
                    "title": "Diagnosing and Fixing Manifold Overfitting in Deep Generative Models",
                    "abstract": "Likelihood-based, or explicit, deep generative models use neural networks to construct flexible high-dimensional densities. This formulation directly contradicts the manifold hypothesis, which states that observed data lies on a low-dimensional manifold embedded in high-dimensional ambient space. In this paper we investigate the pathologies of maximum-likelihood training in the presence of this dimensionality mismatch. We formally prove that degenerate optima are achieved wherein the manifold itself is learned but not the distribution on it, a phenomenon we call manifold overfitting. We propose a class of two-step procedures consisting of a dimensionality reduction step followed by maximum-likelihood density estimation, and prove that they recover the data-generating distribution in the nonparametric regime, thus avoiding manifold overfitting. We also show that these procedures enable density estimation on the manifolds learned by implicit models, such as generative adversarial networks, hence addressing a major shortcoming of these models. Several recently proposed methods are instances of our two-step procedures; we thus unify, extend, and theoretically justify a large class of models."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Intrinsic Dimension, Persistent Homology and Generalization in Neural Networks",
                    "abstract": "Disobeying the classical wisdom of statistical learning theory, modern deep neural networks generalize well even though they typically contain millions of parameters. Recently, it has been shown that the trajectories of iterative optimization algorithms can possess fractal structures, and their generalization error can be formally linked to the complexity of such fractals. This complexity is measured by the fractal's intrinsic dimension, a quantity usually much smaller than the number of parameters in the network. Even though this perspective provides an explanation for why overparametrized networks would not overfit, computing the intrinsic dimension (e.g., for monitoring generalization during training) is a notoriously difficult task, where existing methods typically fail even in moderate ambient dimensions. In this study, we consider this problem from the lens of topological data analysis (TDA) and develop a generic computational tool that is built on rigorous mathematical foundations. By making a novel connection between learning theory and TDA, we first illustrate that the generalization error can be equivalently bounded in terms of a notion called the 'persistent homology dimension' (PHD), where, compared with prior work, our approach does not require any additional geometrical or statistical assumptions on the training dynamics. Then, by utilizing recently established theoretical results and TDA tools, we develop an efficient algorithm to estimate PHD in the scale of modern deep neural networks and further provide visualization tools to help understand generalization in deep learning. Our experiments show that the proposed approach can efficiently compute a network's intrinsic dimension in a variety of settings, which is predictive of the generalization error."
                },
                {
                    "title": "Scikit-Dimension: A Python Package for Intrinsic Dimension Estimation",
                    "abstract": "Dealing with uncertainty in applications of machine learning to real-life data critically depends on the knowledge of intrinsic dimensionality (ID). A number of methods have been suggested for the purpose of estimating ID, but no standard package to easily apply them one by one or all at once has been implemented in Python. This technical note introduces scikit-dimension, an open-source Python package for intrinsic dimension estimation. The scikit-dimension package provides a uniform implementation of most of the known ID estimators based on the scikit-learn application programming interface to evaluate the global and local intrinsic dimension, as well as generators of synthetic toy and benchmark datasets widespread in the literature. The package is developed with tools assessing the code quality, coverage, unit testing and continuous integration. We briefly describe the package and demonstrate its use in a large-scale (more than 500 datasets) benchmarking of methods for ID estimation for real-life and synthetic data."
                },
                {
                    "title": "Score-based Generative Modeling in Latent Space",
                    "abstract": "Score-based generative models (SGMs) have recently demonstrated impressive results in terms of both sample quality and distribution coverage. However, they are usually applied directly in data space and often require thousands of network evaluations for sampling. Here, we propose the Latent Score-based Generative Model (LSGM), a novel approach that trains SGMs in a latent space, relying on the variational autoencoder framework. Moving from data to latent space allows us to train more expressive generative models, apply SGMs to non-continuous data, and learn smoother SGMs in a smaller space, resulting in fewer network evaluations and faster sampling. To enable training LSGMs end-to-end in a scalable and stable manner, we (i) introduce a new score-matching objective suitable to the LSGM setting, (ii) propose a novel parameterization of the score function that allows SGM to focus on the mismatch of the target distribution with respect to a simple Normal one, and (iii) analytically derive multiple techniques for variance reduction of the training objective. LSGM obtains a state-of-the-art FID score of 2.10 on CIFAR-10, outperforming all existing generative results on this dataset. On CelebA-HQ-256, LSGM is on a par with previous SGMs in sample quality while outperforming them in sampling time by two orders of magnitude. In modeling binary images, LSGM achieves state-of-the-art likelihood on the binarized OMNIGLOT dataset. Our project page and code can be found at https://nvlabs.github.io/LSGM ."
                },
                {
                    "title": "The Intrinsic Dimension of Images and Its Impact on Learning",
                    "abstract": "It is widely believed that natural image data exhibits low-dimensional structure despite the high dimensionality of conventional pixel representations. This idea underlies a common intuition for the remarkable success of deep learning in computer vision. In this work, we apply dimension estimation tools to popular datasets and investigate the role of low-dimensional structure in deep learning. We find that common natural image datasets indeed have very low intrinsic dimension relative to the high number of pixels in the images. Additionally, we find that low dimensional datasets are easier for neural networks to learn, and models solving these tasks generalize better from training to test data. Along the way, we develop a technique for validating our dimension estimation tools on synthetic data generated by GANs allowing us to actively manipulate the intrinsic dimension by controlling the image generation process. Code for our experiments may be found here https://github.com/ppope/dimensions."
                },
                {
                    "title": "Maximum Likelihood Training of Score-Based Diffusion Models",
                    "abstract": "Score-based diffusion models synthesize samples by reversing a stochastic process that diffuses data to noise, and are trained by minimizing a weighted combination of score matching losses. The log-likelihood of score-based diffusion models can be tractably computed through a connection to continuous normalizing flows, but log-likelihood is not directly optimized by the weighted combination of score matching losses. We show that for a specific weighting scheme, the objective upper bounds the negative log-likelihood, thus enabling approximate maximum likelihood training of score-based diffusion models. We empirically observe that maximum likelihood training consistently improves the likelihood of score-based diffusion models across multiple datasets, stochastic processes, and model architectures. Our best models achieve negative log-likelihoods of 2.83 and 3.76 bits/dim on CIFAR-10 and ImageNet 32x32 without any data augmentation, on a par with state-of-the-art autoregressive models on these tasks."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Why Normalizing Flows Fail to Detect Out-of-Distribution Data",
                    "abstract": "Detecting out-of-distribution (OOD) data is crucial for robust machine learning systems. Normalizing flows are flexible deep generative models that often surprisingly fail to distinguish between in- and out-of-distribution data: a flow trained on pictures of clothing assigns higher likelihood to handwritten digits. We investigate why normalizing flows perform poorly for OOD detection. We demonstrate that flows learn local pixel correlations and generic image-to-latent-space transformations which are not specific to the target image dataset. We show that by modifying the architecture of flow coupling layers we can bias the flow towards learning the semantic structure of the target data, improving OOD detection. Our investigation reveals that properties that enable flows to generate high-fidelity images can have a detrimental effect on OOD detection."
                },
                {
                    "title": "Relaxing Bijectivity Constraints with Continuously Indexed Normalising Flows",
                    "abstract": "We show that normalising flows become pathological when used to model targets whose supports have complicated topologies. In this scenario, we prove that a flow must become arbitrarily numerically noninvertible in order to approximate the target closely. This result has implications for all flow-based models, and especially Residual Flows (ResFlows), which explicitly control the Lipschitz constant of the bijection used. To address this, we propose Continuously Indexed Flows (CIFs), which replace the single bijection used by normalising flows with a continuously indexed family of bijections, and which can intuitively \"clean up\" mass that would otherwise be misplaced by a single bijection. We show theoretically that CIFs are not subject to the same topological limitations as normalising flows, and obtain better empirical performance on a variety of models and benchmarks."
                },
                {
                    "title": "Neural Spline Flows",
                    "abstract": "A normalizing flow models a complex probability density as an invertible transformation of a simple base density. Flows based on either coupling or autoregressive transforms both offer exact density evaluation and sampling, but rely on the parameterization of an easily invertible elementwise transformation, whose choice determines the flexibility of these models. Building upon recent work, we propose a fully-differentiable module based on monotonic rational-quadratic splines, which enhances the flexibility of both coupling and autoregressive transforms while retaining analytic invertibility. We demonstrate that neural spline flows improve density estimation, variational inference, and generative modeling of images."
                },
                {
                    "title": "Intrinsic dimension of data representations in deep neural networks",
                    "abstract": "Deep neural networks progressively transform their inputs across multiple processing layers. What are the geometrical properties of the representations learned by these networks? Here we study the intrinsic dimensionality (ID) of data-representations, i.e. the minimal number of parameters needed to describe a representation. We find that, in a trained network, the ID is orders of magnitude smaller than the number of units in each layer. Across layers, the ID first increases and then progressively decreases in the final layers. Remarkably, the ID of the last hidden layer predicts classification accuracy on the test set. These results can neither be found by linear dimensionality estimates (e.g., with principal component analysis), nor in representations that had been artificially linearized. They are neither found in untrained networks, nor in networks that are trained on randomized labels. This suggests that neural networks that can generalize are those that transform the data into low-dimensional, but not necessarily flat manifolds."
                },
                {
                    "title": "Applied Stochastic Differential Equations",
                    "abstract": "The topic of this book is stochastic differential equations (SDEs). As their name suggests, they really are differential equations that produce a different \u201canswer\u201d or solution trajectory each time they are solved. This peculiar behaviour gives them properties that are useful in modeling of uncertainties in a wide range of applications, but at the same time it complicates the rigorous mathematical treatment of SDEs. The emphasis of the book is on applied rather than theoretical aspects of SDEs and, therefore, we have chosen to structure the book in a way that we believe supports learning SDEs from an applied point of view. In the following, we briefly outline the purposes of each of the remaining chapters and explain how the chapters are connected to each other. In the chapters, we have attempted to provide a wide selection of examples of the practical application of theoretical and methodological results. Each chapter (except for the Introduction and Epilogue) also contains a representative set of analytic and hands-on exercises that can be used for testing and deepening understanding of the topics. Chapter 2 is a brief outline of concepts and solutions methods for deterministic ordinary differential equations (ODEs). We especially emphasize solution methods for linear ODEs, because the methods translate quite easily to SDEs. We also examine commonly used numerical methods such as the Euler method and Runge\u2013Kutta methods, which we extend to SDEs in the later chapters. Chapter 3 starts with a number of motivating examples of SDEs found in physics, engineering, finance, and other applications. It turns out that in a modeling sense, SDEs can be regarded as noise-driven ODEs, but this notion should not be taken too far. The aim of the rest of the chapter is to show where things start to go wrong. Roughly speaking, with linear SDEs we are quite safe with this kind of thinking, but anything beyond them will not work."
                },
                {
                    "title": "Estimating the effective dimension of large biological datasets using Fisher separability analysis",
                    "abstract": "Modern large-scale datasets are frequently said to be high-dimensional. However, their data point clouds frequently possess structures, significantly decreasing their intrinsic dimensionality (ID) due to the presence of clusters, points being located close to low-dimensional varieties or fine-grained lumping. We introduce and test a dimensionality estimator, based on analysing the separability properties of data points, on several benchmarks and real biological datasets. We show that the introduced measure of ID has performance competitive with state-of-the-art measures, being efficient across a wide range of dimensions and performing better in the case of noisy samples. Moreover, it allows estimating the intrinsic dimension in situations where the intrinsic manifold assumption is not valid."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "Characterizing Adversarial Subspaces Using Local Intrinsic Dimensionality",
                    "abstract": "Deep Neural Networks (DNNs) have recently been shown to be vulnerable against adversarial examples, which are carefully crafted instances that can mislead DNNs to make errors during prediction. To better understand such attacks, a characterization is needed of the properties of regions (the so-called 'adversarial subspaces') in which adversarial examples lie. We tackle this challenge by characterizing the dimensional properties of adversarial regions, via the use of Local Intrinsic Dimensionality (LID). LID assesses the space-filling capability of the region surrounding a reference example, based on the distance distribution of the example to its neighbors. We first provide explanations about how adversarial perturbation can affect the LID characteristic of adversarial regions, and then show empirically that LID characteristics can facilitate the distinction of adversarial examples generated using state-of-the-art attacks. As a proof-of-concept, we show that a potential application of LID is to distinguish adversarial examples, and the preliminary results show that it can outperform several state-of-the-art detection measures by large margins for five attack strategies considered in this paper across three benchmark datasets. Our analysis of the LID characteristic for adversarial regions not only motivates new directions of effective adversarial defense, but also opens up more challenges for developing new attacks to better understand the vulnerabilities of DNNs."
                },
                {
                    "title": "LDMNet: Low Dimensional Manifold Regularized Neural Networks",
                    "abstract": "Deep neural networks have proved very successful on archetypal tasks for which large training sets are available, but when the training data are scarce, their performance suffers from overfitting. Many existing methods of reducing overfitting are data-independent. Data-dependent regularizations are mostly motivated by the observation that data of interest lie close to a manifold, which is typically hard to parametrize explicitly. These methods usually only focus on the geometry of the input data, and do not necessarily encourage the networks to produce geometrically meaningful features. To resolve this, we propose the Low-Dimensional-Manifold-regularized neural Network (LDMNet), which incorporates a feature regularization method that focuses on the geometry of both the input data and the output features. In LDMNet, we regularize the network by encouraging the combination of the input data and the output features to sample a collection of low dimensional manifolds, which are searched efficiently without explicit parametrization. To achieve this, we directly use the manifold dimension as a regularization term in a variational functional. The resulting Euler-Lagrange equation is a Laplace-Beltrami equation over a point cloud, which is solved by the point integral method without increasing the computational complexity. In the experiments, we show that LDMNet significantly outperforms widely-used regularizers. Moreover, LDMNet can extract common features of an object imaged via different modalities, which is very useful in real-world applications such as cross-spectral face recognition."
                },
                {
                    "title": "Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms",
                    "abstract": "We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at this https URL"
                },
                {
                    "title": "Density estimation using Real NVP",
                    "abstract": "Unsupervised learning of probabilistic models is a central yet challenging problem in machine learning. Specifically, designing models with tractable learning, sampling, inference and evaluation is crucial in solving this task. We extend the space of such models using real-valued non-volume preserving (real NVP) transformations, a set of powerful invertible and learnable transformations, resulting in an unsupervised learning algorithm with exact log-likelihood computation, exact sampling, exact inference of latent variables, and an interpretable latent space. We demonstrate its ability to model natural images on four datasets through sampling, log-likelihood evaluation and latent variable manipulations."
                },
                {
                    "title": "Variational Inference with Normalizing Flows",
                    "abstract": "The choice of approximate posterior distribution is one of the core problems in variational inference. Most applications of variational inference employ simple families of posterior approximations in order to allow for efficient inference, focusing on mean-field or other simple structured approximations. This restriction has a significant impact on the quality of inferences made using variational methods. We introduce a new approach for specifying flexible, arbitrarily complex and scalable approximate posterior distributions. Our approximations are distributions constructed through a normalizing flow, whereby a simple initial density is transformed into a more complex one by applying a sequence of invertible transformations until a desired level of complexity is attained. We use this view of normalizing flows to develop categories of finite and infinitesimal flows and provide a unified view of approaches for constructing rich posterior approximations. We demonstrate that the theoretical advantages of having posteriors that better match the true posterior, combined with the scalability of amortized variational approaches, provides a clear improvement in performance and applicability of variational inference."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "Automatic differentiation in machine learning: a survey",
                    "abstract": "Derivatives, mostly in the form of gradients and Hessians, are ubiquitous in machine learning. Automatic differentiation (AD), also called algorithmic differentiation or simply \u201cauto-diff\u201d, is a family of techniques similar to but more general than backpropagation for efficiently and accurately evaluating derivatives of numeric functions expressed as computer programs. AD is a small but established field with applications in areas including computational fluid dynamics, atmospheric sciences, and engineering design optimization. Until \nvery recently, the fields of machine learning and AD have largely been unaware of each other and, in some cases, have independently discovered each other\u2019s results. Despite its \nrelevance, general-purpose AD has been missing from the machine learning toolbox, a situation slowly changing with its ongoing adoption under the names \u201cdynamic computational \ngraphs\u201d and \u201cdifferentiable programming\u201d. We survey the intersection of AD and machine learning, cover applications where AD has direct relevance, and address the main imple- \nmentation techniques. By precisely defining the main differentiation techniques and their interrelationships, we aim to bring clarity to the usage of the terms \u201cautodiff\u201d, \u201cautomatic differentiation\u201d, and \u201csymbolic differentiation\u201d as these are encountered more and more in machine learning settings."
                },
                {
                    "title": "Representation Learning: A Review and New Perspectives",
                    "abstract": "The success of machine learning algorithms generally depends on data representation, and we hypothesize that this is because different representations can entangle and hide more or less the different explanatory factors of variation behind the data. Although specific domain knowledge can be used to help design representations, learning with generic priors can also be used, and the quest for AI is motivating the design of more powerful representation-learning algorithms implementing such priors. This paper reviews recent work in the area of unsupervised feature learning and deep learning, covering advances in probabilistic models, autoencoders, manifold learning, and deep networks. This motivates longer term unanswered questions about the appropriate objectives for learning good representations, for computing representations (i.e., inference), and the geometrical connections between representation learning, density estimation, and manifold learning."
                },
                {
                    "title": "Concordance coefficients to measure the agreement among several sets of ranks",
                    "abstract": "In this paper, two measures of agreement among several sets of ranks, Kendall's concordance coefficient and top-down concordance coefficient, are reviewed. In order to illustrate the utility of these measures, two examples, in the fields of health and sports, are presented. A Monte Carlo simulation study was carried out to compare the performance of Kendall's and top-down concordance coefficients in detecting several types and magnitudes of agreements. The data generation scheme was developed in order to induce an agreement with different intensities among m (m>2) sets of ranks in non-directional and directional rank agreement scenarios. The performance of each coefficient was estimated by the proportion of rejected null hypotheses, assessed at 5% significance level, when testing whether the underlying population concordance coefficient is sufficiently greater than zero. For the directional rank agreement scenario, the top-down concordance coefficient allowed to achieve a percentage of significant concordances that was higher than the one achieved by Kendall's concordance coefficient. Mainly, when the degree of agreement was small, the results of the simulation study pointed to the advantage of using a weighted rank concordance, namely the top-down concordance coefficient, simultaneously with Kendall's concordance coefficient, enabling the detection of agreement (in a top-down sense) in situations not detected by Kendall's concordance coefficient."
                },
                {
                    "title": "A Connection Between Score Matching and Denoising Autoencoders",
                    "abstract": "Denoising autoencoders have been previously shown to be competitive alternatives to restricted Boltzmann machines for unsupervised pretraining of each layer of a deep architecture. We show that a simple denoising autoencoder training criterion is equivalent to matching the score (with respect to the data) of a specific energy-based model to that of a nonparametric Parzen density estimator of the data. This yields several useful insights. It defines a proper probabilistic model for the denoising autoencoder technique, which makes it in principle possible to sample from them or rank examples by their energy. It suggests a different way to apply score matching that is related to learning to denoise and does not require computing second derivatives. It justifies the use of tied weights between the encoder and decoder and suggests ways to extend the success of denoising autoencoders to a larger family of energy-based models."
                },
                {
                    "title": "Finding a \"Kneedle\" in a Haystack: Detecting Knee Points in System Behavior",
                    "abstract": "Computer systems often reach a point at which the relative cost to increase some tunable parameter is no longer worth the corresponding performance benefit. These ``knees'' typically represent beneficial points that system designers have long selected to best balance inherent trade-offs. While prior work largely uses ad hoc, system-specific approaches to detect knees, we present Kneedle, a general approach to on line and off line knee detection that is applicable to a wide range of systems. We define a knee formally for continuous functions using the mathematical concept of curvature and compare our definition against alternatives. We then evaluate Kneedle's accuracy against existing algorithms on both synthetic and real data sets, and evaluate its performance in two different applications."
                },
                {
                    "title": "Cross-concordances: terminology mapping and its effectiveness for information retrieval",
                    "abstract": "The German Federal Ministry for Education and Research funded a major terminology mapping initiative, which found its conclusion in 2007. The task of this terminology mapping initiative was to organize, create and manage 'cross-concordances' between controlled vocabularies (thesauri, classification systems, subject heading lists) centred around the social sciences but quickly extending to other subject areas. 64 crosswalks with more than 500,000 relations were established. In the final phase of the project, a major evaluation effort to test and measure the effectiveness of the vocabulary mappings in an information system environment was conducted. The paper reports on the cross-concordance work and evaluation results."
                },
                {
                    "title": "On Ranking in Survival Analysis: Bounds on the Concordance Index",
                    "abstract": "In this paper, we show that classical survival analysis involving censored data can naturally be cast as a ranking problem. The concordance index (CI), which quantifies the quality of rankings, is the standard performance measure for model assessment in survival analysis. In contrast, the standard approach to learning the popular proportional hazard (PH) model is based on Cox's partial likelihood. We devise two bounds on CI-one of which emerges directly from the properties of PH models-and optimize them directly. Our experimental results suggest that all three methods perform about equally well, with our new approach giving slightly better results. We also explain why a method designed to maximize the Cox's partial likelihood also ends up (approximately) maximizing the CI."
                },
                {
                    "title": "Maximum Likelihood Estimation of Intrinsic Dimension",
                    "abstract": "We propose a new method for estimating intrinsic dimension of a dataset derived by applying the principle of maximum likelihood to the distances between close neighbors. We derive the estimator by a Poisson process approximation, assess its bias and variance theoretically and by simulations, and apply it to a number of simulated and real datasets. We also show it has the best overall performance compared with two other intrinsic dimension estimators."
                },
                {
                    "title": "Introduction to Smooth Manifolds",
                    "abstract": "Preface.- 1 Smooth Manifolds.- 2 Smooth Maps.- 3 Tangent Vectors.- 4 Submersions, Immersions, and Embeddings.- 5 Submanifolds.- 6 Sard's Theorem.- 7 Lie Groups.- 8 Vector Fields.- 9 Integral Curves and Flows.- 10 Vector Bundles.- 11 The Cotangent Bundle.- 12 Tensors.- 13 Riemannian Metrics.- 14 Differential Forms.- 15 Orientations.- 16 Integration on Manifolds.- 17 De Rham Cohomology.- 18 The de Rham Theorem.- 19 Distributions and Foliations.- 20 The Exponential Map.- 21 Quotient Manifolds.- 22 Symplectic Manifolds.- Appendix A: Review of Topology.- Appendix B: Review of Linear Algebra.- Appendix C: Review of Calculus.- Appendix D: Review of Differential Equations.- References.- Notation Index.- Subject Index"
                },
                {
                    "title": "Fractal Geometry: Mathematical Foundations and Applications",
                    "abstract": "Part I Foundations: mathematical background Hausdorff measure and dimension alternative definitions of dimension techniques for calculating dimensions local structure of fractals projections of fractals products of fractals intersections of fractals. Part II Applications and examples: fractals defined by transformations examples from number theory graphs of functions examples from pure mathematics dynamical systems iteration of complex functions-Julia sets random fractals Brownian motion and Brownian surfaces multifractal measures physical applications."
                },
                {
                    "title": "TIME REVERSAL OF DIFFUSIONS",
                    "abstract": "On montre que si un processus de diffusion {X t :0\u2264t\u22641} sur R d satisfait dX t =b(t,X t ) dt+\u03c3(t,X t ) d\u03c9 t alors le processus retourne, {X t :0\u2264t\u22641} ou X t =X 1\u2212t est encore une diffusion avec une derive b et un coefficient de diffusion \u03c3"
                },
                {
                    "title": "An Algorithm for Finding Intrinsic Dimensionality of Data",
                    "abstract": "An algorithm for the analysis of multivariant data is presented along with some experimental results. The basic idea of the method is to examine the data in many small subregions, and from this determine the number of governing parameters, or intrinsic dimensionality. This intrinsic dimensionality is usually much lower than the dimensionality that is given by the standard Karhunen-Lo\u00e8ve technique. An analysis that demonstrates the feasability of this approach is presented."
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                },
                {
                    "title": "Low Bias Local Intrinsic Dimension Estimation from Expected Simplex Skewness",
                    "abstract": "In exploratory high-dimensional data analysis, local intrinsic dimension estimation can sometimes be used in order to discriminate between data sets sampled from different low-dimensional structures. Global intrinsic dimension estimators can in many cases be adapted to local estimation, but this leads to problems with high negative bias or high variance. We introduce a method that exploits the curse/blessing of dimensionality and produces local intrinsic dimension estimators that have very low bias, even in cases where the intrinsic dimension is higher than the number of data points, in combination with relatively low variance. We show that our estimators have a very good ability to classify local data sets by their dimension compared to other local intrinsic dimension estimators; furthermore we provide examples showing the usefulness of local intrinsic dimension estimation in general and our method in particular for stratification of real data sets."
                },
                {
                    "title": "Reading Digits in Natural Images with Unsupervised Feature Learning",
                    "abstract": "Detecting and reading text from natural images is a hard computer vision task that is central to a variety of emerging applications. Related problems like document character recognition have been widely studied by computer vision and machine learning researchers and are virtually solved for practical applications like reading handwritten digits. Reliably recognizing characters in more complex scenes like photographs, however, is far more difficult: the best existing methods lag well behind human performance on the same tasks. In this paper we attack the problem of recognizing digits in a real application using unsupervised feature learning methods: reading house numbers from street level photos. To this end, we introduce a new benchmark dataset for research use containing over 600,000 labeled digits cropped from Street View images. We then demonstrate the difficulty of recognizing these digits when the problem is approached with hand-designed features. Finally, we employ variants of two recently proposed unsupervised feature learning methods and find that they are convincingly superior on our benchmarks."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Gradient-based learning applied to document recognition",
                    "abstract": "Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day."
                },
                {
                    "title": "A stochastic estimator of the trace of the influence matrix for laplacian smoothing splines",
                    "abstract": "An unbiased stochastic estimator of tr(I-A), where A is the influence matrix associated with the calculation of Laplacian smoothing splines, is described. The estimator is similar to one recently developed by Girard but satisfies a minimum variance criterion and does not require the simulation of a standard normal variable. It uses instead simulations of the discrete random variable which takes the values 1, -1 each with probability 1/2. Bounds on the variance of the estimator, similar to those established by Girard, are obtained using elementary methods. The estimator can be used to approximately minimize generalised cross validation (GCV) when using discretized iterative methods for fitting Laplacian smoothing splines to very large data sets. Simulated examples show that the estimated trace values, using either the estimator presented here or the estimator of Girard, perform almost as well as the exact values when applied to the minimization of GCV for n as small as a few hundred, where n is the number ..."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we tractably estimate the local intrinsic dimension (LID) of a given datum in a dataset, given that data manifolds are typically not known explicitly?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of estimating LID has significant implications for the research community as it provides a quantitative measure of data complexity, which can enhance our understanding of high-dimensional data structures. Accurate LID estimates can improve various applications, such as outlier detection, identifying AI-generated content, and understanding the generalization capabilities of neural networks. By advancing LID estimation techniques, future research can leverage these insights to optimize deep learning models, leading to better performance across numerous tasks and potentially uncovering new methodologies in data analysis.\n\n**[Question 3] - Why is it hard?**  \nEstimating LID is challenging due to the high computational cost associated with traditional methods that rely on pairwise distances and nearest neighbors, making them impractical for large datasets. Naive approaches may fail because they do not account for the complex, low-dimensional structures that high-dimensional data often resides on. Additionally, existing model-based estimators can be inaccurate and computationally expensive, and they may not effectively utilize the latest advancements in generative models, such as diffusion models, which complicates the estimation process.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the reliance on traditional intrinsic dimension estimators that are not scalable to large datasets. Additionally, existing solutions often do not leverage the most effective generative models available, leading to inaccuracies. Barriers such as the need for extensive computational resources and the requirement to train multiple models or alter training procedures have hindered progress. Our approach aims to overcome these limitations by utilizing advanced generative models more effectively, thereby improving the accuracy and efficiency of LID estimation.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using a novel estimator, FLIPD, which leverages deep generative models, particularly diffusion models, to estimate LID efficiently. We will apply this method to large datasets, such as LAION Aesthetics, and evaluate its performance using metrics that assess the accuracy of LID estimates against subjective complexity. The expected outcomes include a scalable and accurate LID estimation technique that aligns closely with the intrinsic complexity of data, ultimately enhancing the understanding and application of high-dimensional data analysis in machine learning"
            }
        },
        "author_data": {
            "ea0c945b-b5f0-42e0-a34b-a65481ddf6d7": {
                "pk": "ea0c945b-b5f0-42e0-a34b-a65481ddf6d7",
                "name": "Hamidreza Kamkari",
                "collaborators": [
                    "Rahul G. Krishnan",
                    "Yuan Gao",
                    "Andreas Karrenbauer",
                    "Kurt Mehlhorn",
                    "Mohammadamin Sharifi",
                    "Vahid Balazadeh",
                    "Vahid Zehtab",
                    "Brendan Leigh Ross",
                    "Jesse C. Cresswell",
                    "Anthony L. Caterini"
                ],
                "domain": [
                    "Optimization",
                    "Causal Inference",
                    "Deep Learning",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Physarum Inspired Dynamics to Solve Semi-Definite Programs",
                        "abstract": "Physarum Polycephalum is a slime mold that can solve shortest path problems. A mathematical model based on Physarum's behavior, known as the Physarum Directed Dynamics, can solve positive linear programs. In this paper, we present a family of Physarum-based dynamics extending the previous work and introduce a new algorithm to solve positive Semi-Definite Programs (SDP). The Physarum dynamics are governed by orthogonal projections (w.r.t. time-dependent scalar products) on the affine subspace defined by the linear constraints. We present a natural generalization of the scalar products used in the LP case to the matrix space for SDPs, which boils down to the linear case when all matrices in the SDP are diagonal, thus, representing an LP. We investigate the behavior of the induced dynamics theoretically and experimentally, highlight challenges arising from the non-commutative nature of matrix products, and prove soundness and convergence under mild conditions. Moreover, we consider a more abstract view on the dynamics that suggests a slight variation to guarantee unconditional soundness and convergence-to-optimality. By simulating these dynamics using suitable discretizations, one obtains numerical algorithms for solving positive SDPs, which have applications in discrete optimization, e.g., for computing the Goemans-Williamson approximation for MaxCut or the Lovasz theta number for determining the clique/chromatic number in perfect graphs."
                    },
                    {
                        "title": "Order-based Structure Learning with Normalizing Flows",
                        "abstract": "Estimating the causal structure of observational data is a challenging combinatorial search problem that scales super-exponentially with graph size. Existing methods use continuous relaxations to make this problem computationally tractable but often restrict the data-generating process to additive noise models (ANMs) through explicit or implicit assumptions. We present Order-based Structure Learning with Normalizing Flows (OSLow), a framework that relaxes these assumptions using autoregressive normalizing flows. We leverage the insight that searching over topological orderings is a natural way to enforce acyclicity in structure discovery and propose a novel, differentiable permutation learning method to find such orderings. Through extensive experiments on synthetic and real-world data, we demonstrate that OSLow outperforms prior baselines and improves performance on the observational Sachs and SynTReN datasets as measured by structural hamming distance and structural intervention distance, highlighting the importance of relaxing the ANM assumption made by existing methods."
                    },
                    {
                        "title": "A Geometric Explanation of the Likelihood OOD Detection Paradox",
                        "abstract": "Likelihood-based deep generative models (DGMs) commonly exhibit a puzzling behaviour: when trained on a relatively complex dataset, they assign higher likelihood values to out-of-distribution (OOD) data from simpler sources. Adding to the mystery, OOD samples are never generated by these DGMs despite having higher likelihoods. This two-pronged paradox has yet to be conclusively explained, making likelihood-based OOD detection unreliable. Our primary observation is that high-likelihood regions will not be generated if they contain minimal probability mass. We demonstrate how this seeming contradiction of large densities yet low probability mass can occur around data confined to low-dimensional manifolds. We also show that this scenario can be identified through local intrinsic dimension (LID) estimation, and propose a method for OOD detection which pairs the likelihoods and LID estimates obtained from a pre-trained DGM. Our method can be applied to normalizing flows and score-based diffusion models, and obtains results which match or surpass state-of-the-art OOD detection benchmarks using the same DGM backbones. Our code is available at https://github.com/layer6ai-labs/dgm_ood_detection."
                    }
                ]
            },
            "e08ebf72-ae2d-4314-ba11-bb2535c8edd0": {
                "pk": "e08ebf72-ae2d-4314-ba11-bb2535c8edd0",
                "name": "Brendan Leigh Ross",
                "collaborators": [
                    "Jesse C. Cresswell",
                    "Gabriel Loaiza-Ganem",
                    "Anthony L. Caterini",
                    "Yi Sui",
                    "Rasa Hosseinzadeh",
                    "Junfeng Wen",
                    "Yenson Lau",
                    "Humberto Reyes-Gonzalez",
                    "Marco Letizia",
                    "Luhuan Wu"
                ],
                "domain": [
                    "Generative Models",
                    "Manifold Learning",
                    "Deep Learning",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "Tractable Density Estimation on Learned Manifolds with Conformal Embedding Flows",
                        "abstract": "Normalizing flows are generative models that provide tractable density estimation via an invertible transformation from a simple base distribution to a complex target distribution. However, this technique cannot directly model data supported on an unknown low-dimensional manifold, a common occurrence in real-world domains such as image data. Recent attempts to remedy this limitation have introduced geometric complications that defeat a central benefit of normalizing flows: exact density estimation. We recover this benefit with Conformal Embedding Flows, a framework for designing flows that learn manifolds with tractable densities. We argue that composing a standard flow with a trainable conformal embedding is the most natural way to model manifold-supported data. To this end, we present a series of conformal building blocks and apply them in experiments with synthetic and real-world data to demonstrate that flows can model manifold-supported distributions without sacrificing tractable likelihoods."
                    },
                    {
                        "title": "Neural Implicit Manifold Learning for Topology-Aware Density Estimation",
                        "abstract": "Natural data observed in $\\mathbb{R}^n$ is often constrained to an $m$-dimensional manifold $\\mathcal{M}$, where $m < n$. This work focuses on the task of building theoretically principled generative models for such data. Current generative models learn $\\mathcal{M}$ by mapping an $m$-dimensional latent variable through a neural network $f_\\theta: \\mathbb{R}^m \\to \\mathbb{R}^n$. These procedures, which we call pushforward models, incur a straightforward limitation: manifolds cannot in general be represented with a single parameterization, meaning that attempts to do so will incur either computational instability or the inability to learn probability densities within the manifold. To remedy this problem, we propose to model $\\mathcal{M}$ as a neural implicit manifold: the set of zeros of a neural network. We then learn the probability density within $\\mathcal{M}$ with a constrained energy-based model, which employs a constrained variant of Langevin dynamics to train and sample from the learned manifold. In experiments on synthetic and natural data, we show that our model can learn manifold-supported distributions with complex topologies more accurately than pushforward models."
                    },
                    {
                        "title": "Find Your Friends: Personalized Federated Learning with the Right Collaborators",
                        "abstract": "In the traditional federated learning setting, a central server coordinates a network of clients to train one global model. However, the global model may serve many clients poorly due to data heterogeneity. Moreover, there may not exist a trusted central party that can coordinate the clients to ensure that each of them can benefit from others. To address these concerns, we present a novel decentralized framework, FedeRiCo, where each client can learn as much or as little from other clients as is optimal for its local data distribution. Based on expectation-maximization, FedeRiCo estimates the utilities of other participants' models on each client's data so that everyone can select the right collaborators for learning. As a result, our algorithm outperforms other federated, personalized, and/or decentralized approaches on several benchmark datasets, being the only approach that consistently performs better than training with local data only."
                    },
                    {
                        "title": "Diagnosing and Fixing Manifold Overfitting in Deep Generative Models",
                        "abstract": "Likelihood-based, or explicit, deep generative models use neural networks to construct flexible high-dimensional densities. This formulation directly contradicts the manifold hypothesis, which states that observed data lies on a low-dimensional manifold embedded in high-dimensional ambient space. In this paper we investigate the pathologies of maximum-likelihood training in the presence of this dimensionality mismatch. We formally prove that degenerate optima are achieved wherein the manifold itself is learned but not the distribution on it, a phenomenon we call manifold overfitting. We propose a class of two-step procedures consisting of a dimensionality reduction step followed by maximum-likelihood density estimation, and prove that they recover the data-generating distribution in the nonparametric regime, thus avoiding manifold overfitting. We also show that these procedures enable density estimation on the manifolds learned by implicit models, such as generative adversarial networks, hence addressing a major shortcoming of these models. Several recently proposed methods are instances of our two-step procedures; we thus unify, extend, and theoretically justify a large class of models."
                    },
                    {
                        "title": "CaloMan: Fast generation of calorimeter showers with density estimation on learned manifolds",
                        "abstract": "Precision measurements and new physics searches at the Large Hadron Collider require efficient simulations of particle propagation and interactions within the detectors. The most computationally expensive simulations involve calorimeter showers. Advances in deep generative modelling - particularly in the realm of high-dimensional data - have opened the possibility of generating realistic calorimeter showers orders of magnitude more quickly than physics-based simulation. However, the high-dimensional representation of showers belies the relative simplicity and structure of the underlying physical laws. This phenomenon is yet another example of the manifold hypothesis from machine learning, which states that high-dimensional data is supported on low-dimensional manifolds. We thus propose modelling calorimeter showers first by learning their manifold structure, and then estimating the density of data across this manifold. Learning manifold structure reduces the dimensionality of the data, which enables fast training and generation when compared with competing methods."
                    },
                    {
                        "title": "Denoising Deep Generative Models",
                        "abstract": "Likelihood-based deep generative models have recently been shown to exhibit pathological behaviour under the manifold hypothesis as a consequence of using high-dimensional densities to model data with low-dimensional structure. In this paper we propose two methodologies aimed at addressing this problem. Both are based on adding Gaussian noise to the data to remove the dimensionality mismatch during training, and both provide a denoising mechanism whose goal is to sample from the model as though no noise had been added to the data. Our first approach is based on Tweedie's formula, and the second on models which take the variance of added noise as a conditional input. We show that surprisingly, while well motivated, these approaches only sporadically improve performance over not adding noise, and that other methods of addressing the dimensionality mismatch are more empirically adequate."
                    },
                    {
                        "title": "Verifying the Union of Manifolds Hypothesis for Image Data",
                        "abstract": "Deep learning has had tremendous success at learning low-dimensional representations of high-dimensional data. This success would be impossible if there was no hidden low-dimensional structure in data of interest; this existence is posited by the manifold hypothesis, which states that the data lies on an unknown manifold of low intrinsic dimension. In this paper, we argue that this hypothesis does not properly capture the low-dimensional structure typically present in image data. Assuming that data lies on a single manifold implies intrinsic dimension is identical across the entire data space, and does not allow for subregions of this space to have a different number of factors of variation. To address this deficiency, we consider the union of manifolds hypothesis, which states that data lies on a disjoint union of manifolds of varying intrinsic dimensions. We empirically verify this hypothesis on commonly-used image datasets, finding that indeed, observed data lies on a disconnected set and that intrinsic dimension is not constant. We also provide insights into the implications of the union of manifolds hypothesis in deep learning, both supervised and unsupervised, showing that designing models with an inductive bias for this structure improves performance across classification and generative modelling tasks. Our code is available at https://github.com/layer6ai-labs/UoMH."
                    },
                    {
                        "title": "A Geometric Explanation of the Likelihood OOD Detection Paradox",
                        "abstract": "Likelihood-based deep generative models (DGMs) commonly exhibit a puzzling behaviour: when trained on a relatively complex dataset, they assign higher likelihood values to out-of-distribution (OOD) data from simpler sources. Adding to the mystery, OOD samples are never generated by these DGMs despite having higher likelihoods. This two-pronged paradox has yet to be conclusively explained, making likelihood-based OOD detection unreliable. Our primary observation is that high-likelihood regions will not be generated if they contain minimal probability mass. We demonstrate how this seeming contradiction of large densities yet low probability mass can occur around data confined to low-dimensional manifolds. We also show that this scenario can be identified through local intrinsic dimension (LID) estimation, and propose a method for OOD detection which pairs the likelihoods and LID estimates obtained from a pre-trained DGM. Our method can be applied to normalizing flows and score-based diffusion models, and obtains results which match or surpass state-of-the-art OOD detection benchmarks using the same DGM backbones. Our code is available at https://github.com/layer6ai-labs/dgm_ood_detection."
                    },
                    {
                        "title": "Deep Generative Models through the Lens of the Manifold Hypothesis: A Survey and New Connections",
                        "abstract": "In recent years there has been increased interest in understanding the interplay between deep generative models (DGMs) and the manifold hypothesis. Research in this area focuses on understanding the reasons why commonly-used DGMs succeed or fail at learning distributions supported on unknown low-dimensional manifolds, as well as developing new models explicitly designed to account for manifold-supported data. This manifold lens provides both clarity as to why some DGMs (e.g. diffusion models and some generative adversarial networks) empirically surpass others (e.g. likelihood-based models such as variational autoencoders, normalizing flows, or energy-based models) at sample generation, and guidance for devising more performant DGMs. We carry out the first survey of DGMs viewed through this lens, making two novel contributions along the way. First, we formally establish that numerical instability of likelihoods in high ambient dimensions is unavoidable when modelling data with low intrinsic dimension. We then show that DGMs on learned representations of autoencoders can be interpreted as approximately minimizing Wasserstein distance: this result, which applies to latent diffusion models, helps justify their outstanding empirical results. The manifold lens provides a rich perspective from which to understand DGMs, and we aim to make this perspective more accessible and widespread."
                    },
                    {
                        "title": "Exposing flaws of generative model evaluation metrics and their unfair treatment of diffusion models",
                        "abstract": "We systematically study a wide variety of generative models spanning semantically-diverse image datasets to understand and improve the feature extractors and metrics used to evaluate them. Using best practices in psychophysics, we measure human perception of image realism for generated samples by conducting the largest experiment evaluating generative models to date, and find that no existing metric strongly correlates with human evaluations. Comparing to 17 modern metrics for evaluating the overall performance, fidelity, diversity, rarity, and memorization of generative models, we find that the state-of-the-art perceptual realism of diffusion models as judged by humans is not reflected in commonly reported metrics such as FID. This discrepancy is not explained by diversity in generated samples, though one cause is over-reliance on Inception-V3. We address these flaws through a study of alternative self-supervised feature extractors, find that the semantic information encoded by individual networks strongly depends on their training procedure, and show that DINOv2-ViT-L/14 allows for much richer evaluation of generative models. Next, we investigate data memorization, and find that generative models do memorize training examples on simple, smaller datasets like CIFAR10, but not necessarily on more complex datasets like ImageNet. However, our experiments show that current metrics do not properly detect memorization: none in the literature is able to separate memorization from other phenomena such as underfitting or mode shrinkage. To facilitate further development of generative models and their evaluation we release all generated image datasets, human evaluation data, and a modular library to compute 17 common metrics for 9 different encoders at https://github.com/layer6ai-labs/dgm-eval."
                    }
                ]
            },
            "0659ac8e-7092-4a35-9a07-2bce2ff1b7f3": {
                "pk": "0659ac8e-7092-4a35-9a07-2bce2ff1b7f3",
                "name": "Rasa Hosseinzadeh",
                "collaborators": [
                    "Jesse C. Cresswell",
                    "Murat A. Erdogdu",
                    "Maksims Volkovs",
                    "Junwei Ma",
                    "Keyvan Golestan",
                    "Guangwei Yu",
                    "Zhaoyan Liu",
                    "Gabriel Loaiza-Ganem",
                    "Brendan Leigh Ross",
                    "Anthony L. Caterini"
                ],
                "domain": [
                    "Generative Models",
                    "Deep Learning",
                    "Monte Carlo Methods",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "On the Convergence of Langevin Monte Carlo: The Interplay between Tail Growth and Smoothness",
                        "abstract": "We study sampling from a target distribution ${\\nu_* = e^{-f}}$ using the unadjusted Langevin Monte Carlo (LMC) algorithm. For any potential function $f$ whose tails behave like ${\\|x\\|^\\alpha}$ for ${\\alpha \\in [1,2]}$, and has $\\beta$-H\\\"older continuous gradient, we prove that ${\\widetilde{\\mathcal{O}} \\Big(d^{\\frac{1}{\\beta}+\\frac{1+\\beta}{\\beta}(\\frac{2}{\\alpha} - \\boldsymbol{1}_{\\{\\alpha \\neq 1\\}})} \\epsilon^{-\\frac{1}{\\beta}}\\Big)}$ steps are sufficient to reach the $\\epsilon $-neighborhood of a $d$-dimensional target distribution $\\nu_*$ in KL-divergence. This convergence rate, in terms of $\\epsilon$ dependency, is not directly influenced by the tail growth rate $\\alpha$ of the potential function as long as its growth is at least linear, and it only relies on the order of smoothness $\\beta$. One notable consequence of this result is that for potentials with Lipschitz gradient, i.e. $\\beta=1$, our rate recovers the best known rate ${\\widetilde{\\mathcal{O}}(d\\epsilon^{-1})}$ which was established for strongly convex potentials in terms of $\\epsilon$ dependency, but we show that the same rate is achievable for a wider class of potentials that are degenerately convex at infinity. The growth rate $\\alpha$ starts to have an effect on the established rate in high dimensions where $d$ is large; furthermore, it recovers the best-known dimension dependency when the tail growth of the potential is quadratic, i.e. ${\\alpha = 2}$, in the current setup. Our framework allows for finite perturbations, and any order of smoothness ${\\beta\\in(0,1]}$; consequently, our results are applicable to a wide class of non-convex potentials that are weakly smooth and exhibit at least linear tail growth."
                    },
                    {
                        "title": "DiMS: Distilling Multiple Steps of Iterative Non-Autoregressive Transformers for Machine Translation",
                        "abstract": "The computational benefits of iterative non-autoregressive transformers decrease as the number of decoding steps increases. As a remedy, we introduce Distill Multiple Steps (DiMS), a simple yet effective distillation technique to decrease the number of required steps to reach a certain translation quality. The distilled model enjoys the computational benefits of early iterations while preserving the enhancements from several iterative steps. DiMS relies on two models namely student and teacher. The student is optimized to predict the output of the teacher after multiple decoding steps while the teacher follows the student via a slow-moving average. The moving average keeps the teacher's knowledge updated and enhances the quality of the labels provided by the teacher. During inference, the student is used for translation and no additional computation is added. We verify the effectiveness of DiMS on various models obtaining 7.8 and 12.9 BLEU points improvements in single-step translation accuracy on distilled and raw versions of WMT'14 De-En."
                    },
                    {
                        "title": "Convergence of Langevin Monte Carlo in Chi-Squared and Renyi Divergence",
                        "abstract": "We study sampling from a target distribution $\\nu_* = e^{-f}$ using the unadjusted Langevin Monte Carlo (LMC) algorithm when the potential $f$ satisfies a strong dissipativity condition and it is first-order smooth with a Lipschitz gradient. We prove that, initialized with a Gaussian random vector that has sufficiently small variance, iterating the LMC algorithm for $\\widetilde{\\mathcal{O}}(\\lambda^2 d\\epsilon^{-1})$ steps is sufficient to reach $\\epsilon$-neighborhood of the target in both Chi-squared and Renyi divergence, where $\\lambda$ is the logarithmic Sobolev constant of $\\nu_*$. Our results do not require warm-start to deal with the exponential dimension dependency in Chi-squared divergence at initialization. In particular, for strongly convex and first-order smooth potentials, we show that the LMC algorithm achieves the rate estimate $\\widetilde{\\mathcal{O}}(d\\epsilon^{-1})$ which improves the previously known rates in both of these metrics, under the same assumptions. Translating this rate to other metrics, our results also recover the state-of-the-art rate estimates in KL divergence, total variation and $2$-Wasserstein distance in the same setup. Finally, as we rely on the logarithmic Sobolev inequality, our framework covers a range of non-convex potentials that are first-order smooth and exhibit strong convexity outside of a compact region."
                    },
                    {
                        "title": "Retrieval & Fine-Tuning for In-Context Tabular Models",
                        "abstract": "Tabular data is a pervasive modality spanning a wide range of domains, and the inherent diversity poses a considerable challenge for deep learning. Recent advancements using transformer-based in-context learning have shown promise on smaller and less complex datasets, but have struggled to scale to larger and more complex ones. To address this limitation, we propose a combination of retrieval and fine-tuning: we can adapt the transformer to a local subset of the data by collecting nearest neighbours, and then perform task-specific fine-tuning with this retrieved set of neighbours in context. Using TabPFN as the base model -- currently the best tabular in-context learner -- and applying our retrieval and fine-tuning scheme on top results in what we call a locally-calibrated PFN, or LoCalPFN. We conduct extensive evaluation on 95 datasets curated by TabZilla from OpenML, upon which we establish a new state-of-the-art with LoCalPFN -- even with respect to tuned tree-based models. Notably, we show a significant boost in performance compared to the base in-context model, demonstrating the efficacy of our approach and advancing the frontier of deep learning in tabular data."
                    },
                    {
                        "title": "MSc-SQL: Multi-Sample Critiquing Small Language Models For Text-To-SQL Translation",
                        "abstract": "Text-to-SQL generation enables non-experts to interact with databases via natural language. Recent advances rely on large closed-source models like GPT-4 that present challenges in accessibility, privacy, and latency. To address these issues, we focus on developing small, efficient, and open-source text-to-SQL models. We demonstrate the benefits of sampling multiple candidate SQL generations and propose our method, MSc-SQL, to critique them using associated metadata. Our sample critiquing model evaluates multiple outputs simultaneously, achieving state-of-the-art performance compared to other open-source models while remaining competitive with larger models at a much lower cost. Full code can be found at github.com/layer6ai-labs/msc-sql."
                    },
                    {
                        "title": "Deep Generative Models through the Lens of the Manifold Hypothesis: A Survey and New Connections",
                        "abstract": "In recent years there has been increased interest in understanding the interplay between deep generative models (DGMs) and the manifold hypothesis. Research in this area focuses on understanding the reasons why commonly-used DGMs succeed or fail at learning distributions supported on unknown low-dimensional manifolds, as well as developing new models explicitly designed to account for manifold-supported data. This manifold lens provides both clarity as to why some DGMs (e.g. diffusion models and some generative adversarial networks) empirically surpass others (e.g. likelihood-based models such as variational autoencoders, normalizing flows, or energy-based models) at sample generation, and guidance for devising more performant DGMs. We carry out the first survey of DGMs viewed through this lens, making two novel contributions along the way. First, we formally establish that numerical instability of likelihoods in high ambient dimensions is unavoidable when modelling data with low intrinsic dimension. We then show that DGMs on learned representations of autoencoders can be interpreted as approximately minimizing Wasserstein distance: this result, which applies to latent diffusion models, helps justify their outstanding empirical results. The manifold lens provides a rich perspective from which to understand DGMs, and we aim to make this perspective more accessible and widespread."
                    },
                    {
                        "title": "Tabular Data Contrastive Learning via Class-Conditioned and Feature-Correlation Based Augmentation",
                        "abstract": "Contrastive learning is a model pre-training technique by first creating similar views of the original data, and then encouraging the data and its corresponding views to be close in the embedding space. Contrastive learning has witnessed success in image and natural language data, thanks to the domain-specific augmentation techniques that are both intuitive and effective. Nonetheless, in tabular domain, the predominant augmentation technique for creating views is through corrupting tabular entries via swapping values, which is not as sound or effective. We propose a simple yet powerful improvement to this augmentation technique: corrupting tabular data conditioned on class identity. Specifically, when corrupting a specific tabular entry from an anchor row, instead of randomly sampling a value in the same feature column from the entire table uniformly, we only sample from rows that are identified to be within the same class as the anchor row. We assume the semi-supervised learning setting, and adopt the pseudo labeling technique for obtaining class identities over all table rows. We also explore the novel idea of selecting features to be corrupted based on feature correlation structures. Extensive experiments show that the proposed approach consistently outperforms the conventional corruption method for tabular data classification tasks. Our code is available at https://github.com/willtop/Tabular-Class-Conditioned-SSL."
                    },
                    {
                        "title": "Exposing flaws of generative model evaluation metrics and their unfair treatment of diffusion models",
                        "abstract": "We systematically study a wide variety of generative models spanning semantically-diverse image datasets to understand and improve the feature extractors and metrics used to evaluate them. Using best practices in psychophysics, we measure human perception of image realism for generated samples by conducting the largest experiment evaluating generative models to date, and find that no existing metric strongly correlates with human evaluations. Comparing to 17 modern metrics for evaluating the overall performance, fidelity, diversity, rarity, and memorization of generative models, we find that the state-of-the-art perceptual realism of diffusion models as judged by humans is not reflected in commonly reported metrics such as FID. This discrepancy is not explained by diversity in generated samples, though one cause is over-reliance on Inception-V3. We address these flaws through a study of alternative self-supervised feature extractors, find that the semantic information encoded by individual networks strongly depends on their training procedure, and show that DINOv2-ViT-L/14 allows for much richer evaluation of generative models. Next, we investigate data memorization, and find that generative models do memorize training examples on simple, smaller datasets like CIFAR10, but not necessarily on more complex datasets like ImageNet. However, our experiments show that current metrics do not properly detect memorization: none in the literature is able to separate memorization from other phenomena such as underfitting or mode shrinkage. To facilitate further development of generative models and their evaluation we release all generated image datasets, human evaluation data, and a modular library to compute 17 common metrics for 9 different encoders at https://github.com/layer6ai-labs/dgm-eval."
                    }
                ]
            },
            "e2fe2d79-13e1-40bc-aa6b-990af6afd996": {
                "pk": "e2fe2d79-13e1-40bc-aa6b-990af6afd996",
                "name": "Jesse C. Cresswell",
                "collaborators": [
                    "Amanda W. Peet",
                    "Yi Sui",
                    "Bhargava Kumar",
                    "Ilan Tzitrin",
                    "Aaron Z. Goldberg",
                    "Brendan Leigh Ross",
                    "Dan N. Vollick",
                    "Taewoo Kim",
                    "Ian T. Jardine",
                    "No\u00ebl Vouitsis"
                ],
                "domain": [
                    "Quantum Information",
                    "Machine Learning",
                    "Entanglement",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "Universal entanglement timescale for R\u00e9nyi entropies",
                        "abstract": "Recently it was shown that the growth of entanglement in an initially separable state, as measured by the purity of subsystems, can be characterized by a timescale that takes a universal form for any Hamiltonian. We show that the same timescale governs the growth of entanglement for all R\\'enyi entropies. Since the family of R\\'enyi entropies completely characterizes the entanglement of a pure bipartite state, our timescale is a universal feature of bipartite entanglement. The timescale depends only on the interaction Hamiltonian and the initial state."
                    },
                    {
                        "title": "Quantum Information Approaches to Quantum Gravity",
                        "abstract": "In this thesis we apply techniques from quantum information theory to study quantum gravity within the framework of the anti-de Sitter / conformal field theory correspondence (AdS/CFT). Through AdS/CFT, progress has been made in understanding the structure of entanglement in quantum field theories, and in how gravitational physics can emerge from these structures. However, this understanding is far from complete and will require the development of new tools to quantify correlations in CFT.   This thesis presents refinements of a duality between operator product expansion (OPE) blocks in the CFT, giving the contribution of a conformal family to the OPE, and geodesic integrated fields in AdS which are diffeomorphism invariant quantities. This duality was originally discovered in the maximally symmetric setting of pure AdS dual to the CFT ground state. In less symmetric states the duality must be modified. Working with excited states within AdS$_3$/CFT$_2$, this thesis shows how the OPE block decomposes into more fine-grained CFT observables that are dual to AdS fields integrated over non-minimal geodesics.   Additionally, this thesis contains results on the dynamics of entanglement measures for general quantum systems. Results are presented for the family of quantum R\\'enyi entropies and entanglement negativity. R\\'enyi entropies are studied for general dynamics by imposing special initial conditions. Around pure, separable initial states, all R\\'enyi entropies grow with the same timescale at leading, and next-to-leading order. Mathematical tools are developed for the differentiation of non-analytic matrix functions with respect to constrained arguments and are used to construct analytic expressions for derivatives of negativity. We establish bounds on the rate of change of state distinguishability and the rate of entanglement growth for closed systems. Note: Abstract shortened."
                    },
                    {
                        "title": "Lorenz gauge quantization in conformally flat spacetimes",
                        "abstract": "Recently it was shown that Dirac's method of quantizing constrained dynamical systems can be used to impose the Lorenz gauge condition in a four-dimensional cosmological spacetime. In this paper we use Dirac's method to impose the Lorenz gauge condition in a general four-dimensional conformally flat spacetime and find that there is no particle production. We show that in cosmological spacetimes with dimension $D\\neq 4$ there will be particle production when the scale factor changes, and we calculate the particle production due to a sudden change."
                    },
                    {
                        "title": "Kinematic space for conical defects",
                        "abstract": "Kinematic space can be used as an intermediate step in the AdS/CFT dictionary and lends itself naturally to the description of diffeomorphism invariant quantities. From the bulk it has been defined as the space of boundary anchored geodesics, and from the boundary as the space of pairs of CFT points. When the bulk is not globally AdS$_3$ the appearance of non-minimal geodesics leads to ambiguities in these definitions. In this work conical defect spacetimes are considered as an example where non-minimal geodesics are common. From the bulk it is found that the conical defect kinematic space can be obtained from the AdS$_3$ kinematic space by the same quotient under which one obtains the defect from AdS$_3$. The resulting kinematic space is one of many equivalent fundamental regions. From the boundary the conical defect kinematic space can be determined by breaking up OPE blocks into contributions from individual bulk geodesics. A duality is established between partial OPE blocks and bulk fields integrated over individual geodesics, minimal or non-minimal."
                    },
                    {
                        "title": "Scaling Up Diffusion and Flow-based XGBoost Models",
                        "abstract": "Novel machine learning methods for tabular data generation are often developed on small datasets which do not match the scale required for scientific applications. We investigate a recent proposal to use XGBoost as the function approximator in diffusion and flow-matching models on tabular data, which proved to be extremely memory intensive, even on tiny datasets. In this work, we conduct a critical analysis of the existing implementation from an engineering perspective, and show that these limitations are not fundamental to the method; with better implementation it can be scaled to datasets 370x larger than previously used. Our efficient implementation also unlocks scaling models to much larger sizes which we show directly leads to improved performance on benchmark tasks. We also propose algorithmic improvements that can further benefit resource usage and model performance, including multi-output trees which are well-suited to generative modeling. Finally, we present results on large-scale scientific datasets derived from experimental particle physics as part of the Fast Calorimeter Simulation Challenge. Code is available at https://github.com/layer6ai-labs/calo-forest."
                    },
                    {
                        "title": "Holographic relations for OPE blocks in excited states",
                        "abstract": "We study the holographic duality between boundary OPE blocks and geodesic integrated bulk fields in quotients of AdS$_3$ dual to excited CFT states. The quotient geometries exhibit non-minimal geodesics between pairs of spacelike separated boundary points which modify the OPE block duality. We decompose OPE blocks into quotient invariant operators and propose a duality with bulk fields integrated over individual geodesics, minimal or non-minimal. We provide evidence for this relationship by studying the monodromy of asymptotic maps that implement the quotients."
                    },
                    {
                        "title": "Conformal Prediction Sets Improve Human Decision Making",
                        "abstract": "In response to everyday queries, humans explicitly signal uncertainty and offer alternative answers when they are unsure. Machine learning models that output calibrated prediction sets through conformal prediction mimic this human behaviour; larger sets signal greater uncertainty while providing alternatives. In this work, we study the usefulness of conformal prediction sets as an aid for human decision making by conducting a pre-registered randomized controlled trial with conformal prediction sets provided to human subjects. With statistical significance, we find that when humans are given conformal prediction sets their accuracy on tasks improves compared to fixed-size prediction sets with the same coverage guarantee. The results show that quantifying model uncertainty with conformal prediction is helpful for human-in-the-loop decision making and human-AI teams."
                    },
                    {
                        "title": "Conformal Prediction Sets Can Cause Disparate Impact",
                        "abstract": "Although conformal prediction is a promising method for quantifying the uncertainty of machine learning models, the prediction sets it outputs are not inherently actionable. Many applications require a single output to act on, not several. To overcome this, prediction sets can be provided to a human who then makes an informed decision. In any such system it is crucial to ensure the fairness of outcomes across protected groups, and researchers have proposed that Equalized Coverage be used as the standard for fairness. By conducting experiments with human participants, we demonstrate that providing prediction sets can increase the unfairness of their decisions. Disquietingly, we find that providing sets that satisfy Equalized Coverage actually increases unfairness compared to marginal coverage. Instead of equalizing coverage, we propose to equalize set sizes across groups which empirically leads to more fair outcomes."
                    },
                    {
                        "title": "Operational symmetries of entangled states",
                        "abstract": "Quantum entanglement obscures the notion of local operations; there exist quantum states for which all local actions on one subsystem can be equivalently realized by actions on another. We characterize the states for which this fundamental property of entanglement does and does not hold, including multipartite and mixed states. Our results lead to a method for quantifying entanglement based on operational symmetries and has connections to quantum steering, envariance, the Reeh-Schlieder theorem, and classical entanglement."
                    },
                    {
                        "title": "Tractable Density Estimation on Learned Manifolds with Conformal Embedding Flows",
                        "abstract": "Normalizing flows are generative models that provide tractable density estimation via an invertible transformation from a simple base distribution to a complex target distribution. However, this technique cannot directly model data supported on an unknown low-dimensional manifold, a common occurrence in real-world domains such as image data. Recent attempts to remedy this limitation have introduced geometric complications that defeat a central benefit of normalizing flows: exact density estimation. We recover this benefit with Conformal Embedding Flows, a framework for designing flows that learn manifolds with tractable densities. We argue that composing a standard flow with a trainable conformal embedding is the most natural way to model manifold-supported data. To this end, we present a series of conformal building blocks and apply them in experiments with synthetic and real-world data to demonstrate that flows can model manifold-supported distributions without sacrificing tractable likelihoods."
                    },
                    {
                        "title": "Decentralized Federated Learning through Proxy Model Sharing",
                        "abstract": "Institutions in highly regulated domains such as finance and healthcare often have restrictive rules around data sharing. Federated learning is a distributed learning framework that enables multi-institutional collaborations on decentralized data with improved protection for each collaborator's data privacy. In this paper, we propose a communication-efficient scheme for decentralized federated learning called ProxyFL, or proxy-based federated learning. Each participant in ProxyFL maintains two models, a private model, and a publicly shared proxy model designed to protect the participant's privacy. Proxy models allow efficient information exchange among participants without the need of a centralized server. The proposed method eliminates a significant limitation of canonical federated learning by allowing model heterogeneity; each participant can have a private model with any architecture. Furthermore, our protocol for communication by proxy leads to stronger privacy guarantees using differential privacy analysis. Experiments on popular image datasets, and a cancer diagnostic problem using high-quality gigapixel histology whole slide images, show that ProxyFL can outperform existing alternatives with much less communication overhead and stronger privacy."
                    },
                    {
                        "title": "Disparate Impact in Differential Privacy from Gradient Misalignment",
                        "abstract": "As machine learning becomes more widespread throughout society, aspects including data privacy and fairness must be carefully considered, and are crucial for deployment in highly regulated industries. Unfortunately, the application of privacy enhancing technologies can worsen unfair tendencies in models. In particular, one of the most widely used techniques for private model training, differentially private stochastic gradient descent (DPSGD), frequently intensifies disparate impact on groups within data. In this work we study the fine-grained causes of unfairness in DPSGD and identify gradient misalignment due to inequitable gradient clipping as the most significant source. This observation leads us to a new method for reducing unfairness by preventing gradient misalignment in DPSGD."
                    },
                    {
                        "title": "Perturbative expansion of entanglement negativity using patterned matrix calculus",
                        "abstract": "Negativity is an entanglement monotone frequently used to quantify entanglement in bipartite states. Because negativity is a non-analytic function of a density matrix, existing methods used in the physics literature are insufficient to compute its derivatives. To this end we develop techniques in the calculus of complex, patterned matrices and use them to conduct a perturbative analysis of negativity in terms of arbitrary variations of the density operator. The result is an easy-to-implement expansion that can be carried out to all orders. On the way we provide convenient representations of the partial transposition map appearing in the definition of negativity. Our methods are well-suited to study the growth and decay of entanglement in a wide range of physical systems, including the generic linear growth of entanglement in many-body systems, and have broad relevance to many functions of quantum states and observables."
                    },
                    {
                        "title": "CaloMan: Fast generation of calorimeter showers with density estimation on learned manifolds",
                        "abstract": "Precision measurements and new physics searches at the Large Hadron Collider require efficient simulations of particle propagation and interactions within the detectors. The most computationally expensive simulations involve calorimeter showers. Advances in deep generative modelling - particularly in the realm of high-dimensional data - have opened the possibility of generating realistic calorimeter showers orders of magnitude more quickly than physics-based simulation. However, the high-dimensional representation of showers belies the relative simplicity and structure of the underlying physical laws. This phenomenon is yet another example of the manifold hypothesis from machine learning, which states that high-dimensional data is supported on low-dimensional manifolds. We thus propose modelling calorimeter showers first by learning their manifold structure, and then estimating the density of data across this manifold. Learning manifold structure reduces the dimensionality of the data, which enables fast training and generation when compared with competing methods."
                    }
                ]
            },
            "5fffcb44-777f-4e2a-a9d9-ac902a243d57": {
                "pk": "5fffcb44-777f-4e2a-a9d9-ac902a243d57",
                "name": "Gabriel Loaiza-Ganem",
                "collaborators": [
                    "Anthony L. Caterini",
                    "Jesse C. Cresswell",
                    "Brendan Leigh Ross",
                    "John P. Cunningham",
                    "Bradley C. A. Brown",
                    "Harry J. Braviner",
                    "Panteha Naderian",
                    "Yuanjun Gao",
                    "Jordan Juravsky",
                    "Zhaoyan Liu"
                ],
                "domain": [
                    "Deep Learning",
                    "Generative Models",
                    "Reinforcement Learning",
                    "Statistical Modeling"
                ],
                "publications": [
                    {
                        "title": "Maximum Entropy Flow Networks",
                        "abstract": "Maximum entropy modeling is a flexible and popular framework for formulating statistical models given partial knowledge. In this paper, rather than the traditional method of optimizing over the continuous density directly, we learn a smooth and invertible transformation that maps a simple distribution to the desired maximum entropy distribution. Doing so is nontrivial in that the objective being maximized (entropy) is a function of the density itself. By exploiting recent developments in normalizing flow networks, we cast the maximum entropy problem into a finite-dimensional constrained optimization, and solve the problem by combining stochastic optimization with the augmented Lagrangian method. Simulation results demonstrate the effectiveness of our method, and applications to finance and computer vision show the flexibility and accuracy of using maximum entropy flow networks."
                    },
                    {
                        "title": "The continuous Bernoulli: fixing a pervasive error in variational autoencoders",
                        "abstract": "Variational autoencoders (VAE) have quickly become a central tool in machine learning, applicable to a broad range of data types and latent variable models. By far the most common first step, taken by seminal papers and by core software libraries alike, is to model MNIST data using a deep network parameterizing a Bernoulli likelihood. This practice contains what appears to be and what is often set aside as a minor inconvenience: the pixel data is [0,1] valued, not {0,1} as supported by the Bernoulli likelihood. Here we show that, far from being a triviality or nuisance that is convenient to ignore, this error has profound importance to VAE, both qualitative and quantitative. We introduce and fully characterize a new [0,1]-supported, single parameter distribution: the continuous Bernoulli, which patches this pervasive bug in VAE. This distribution is not nitpicking; it produces meaningful performance improvements across a range of metrics and datasets, including sharper image samples, and suggests a broader class of performant VAE."
                    },
                    {
                        "title": "Relating Regularization and Generalization through the Intrinsic Dimension of Activations",
                        "abstract": "Given a pair of models with similar training set performance, it is natural to assume that the model that possesses simpler internal representations would exhibit better generalization. In this work, we provide empirical evidence for this intuition through an analysis of the intrinsic dimension (ID) of model activations, which can be thought of as the minimal number of factors of variation in the model's representation of the data. First, we show that common regularization techniques uniformly decrease the last-layer ID (LLID) of validation set activations for image classification models and show how this strongly affects generalization performance. We also investigate how excessive regularization decreases a model's ability to extract features from data in earlier layers, leading to a negative effect on validation accuracy even while LLID continues to decrease and training accuracy remains near-perfect. Finally, we examine the LLID over the course of training of models that exhibit grokking. We observe that well after training accuracy saturates, when models ``grok'' and validation accuracy suddenly improves from random to perfect, there is a co-occurent sudden drop in LLID, thus providing more insight into the dynamics of sudden generalization."
                    },
                    {
                        "title": "TR0N: Translator Networks for 0-Shot Plug-and-Play Conditional Generation",
                        "abstract": "We propose TR0N, a highly general framework to turn pre-trained unconditional generative models, such as GANs and VAEs, into conditional models. The conditioning can be highly arbitrary, and requires only a pre-trained auxiliary model. For example, we show how to turn unconditional models into class-conditional ones with the help of a classifier, and also into text-to-image models by leveraging CLIP. TR0N learns a lightweight stochastic mapping which \"translates\" between the space of conditions and the latent space of the generative model, in such a way that the generated latent corresponds to a data sample satisfying the desired condition. The translated latent samples are then further improved upon through Langevin dynamics, enabling us to obtain higher-quality data samples. TR0N requires no training data nor fine-tuning, yet can achieve a zero-shot FID of 10.9 on MS-COCO, outperforming competing alternatives not only on this metric, but also in sampling speed -- all while retaining a much higher level of generality. Our code is available at https://github.com/layer6ai-labs/tr0n."
                    },
                    {
                        "title": "Bayesian Nonparametrics for Offline Skill Discovery",
                        "abstract": "Skills or low-level policies in reinforcement learning are temporally extended actions that can speed up learning and enable complex behaviours. Recent work in offline reinforcement learning and imitation learning has proposed several techniques for skill discovery from a set of expert trajectories. While these methods are promising, the number K of skills to discover is always a fixed hyperparameter, which requires either prior knowledge about the environment or an additional parameter search to tune it. We first propose a method for offline learning of options (a particular skill framework) exploiting advances in variational inference and continuous relaxations. We then highlight an unexplored connection between Bayesian nonparametrics and offline skill discovery, and show how to obtain a nonparametric version of our model. This version is tractable thanks to a carefully structured approximate posterior with a dynamically-changing number of options, removing the need to specify K. We also show how our nonparametric extension can be applied in other skill frameworks, and empirically demonstrate that our method can outperform state-of-the-art offline skill learning algorithms across a variety of environments. Our code is available at https://github.com/layer6ai-labs/BNPO ."
                    },
                    {
                        "title": "Diagnosing and Fixing Manifold Overfitting in Deep Generative Models",
                        "abstract": "Likelihood-based, or explicit, deep generative models use neural networks to construct flexible high-dimensional densities. This formulation directly contradicts the manifold hypothesis, which states that observed data lies on a low-dimensional manifold embedded in high-dimensional ambient space. In this paper we investigate the pathologies of maximum-likelihood training in the presence of this dimensionality mismatch. We formally prove that degenerate optima are achieved wherein the manifold itself is learned but not the distribution on it, a phenomenon we call manifold overfitting. We propose a class of two-step procedures consisting of a dimensionality reduction step followed by maximum-likelihood density estimation, and prove that they recover the data-generating distribution in the nonparametric regime, thus avoiding manifold overfitting. We also show that these procedures enable density estimation on the manifolds learned by implicit models, such as generative adversarial networks, hence addressing a major shortcoming of these models. Several recently proposed methods are instances of our two-step procedures; we thus unify, extend, and theoretically justify a large class of models."
                    },
                    {
                        "title": "Denoising Deep Generative Models",
                        "abstract": "Likelihood-based deep generative models have recently been shown to exhibit pathological behaviour under the manifold hypothesis as a consequence of using high-dimensional densities to model data with low-dimensional structure. In this paper we propose two methodologies aimed at addressing this problem. Both are based on adding Gaussian noise to the data to remove the dimensionality mismatch during training, and both provide a denoising mechanism whose goal is to sample from the model as though no noise had been added to the data. Our first approach is based on Tweedie's formula, and the second on models which take the variance of added noise as a conditional input. We show that surprisingly, while well motivated, these approaches only sporadically improve performance over not adding noise, and that other methods of addressing the dimensionality mismatch are more empirically adequate."
                    },
                    {
                        "title": "Deep Random Splines for Point Process Intensity Estimation of Neural Population Data",
                        "abstract": "Gaussian processes are the leading class of distributions on random functions, but they suffer from well known issues including difficulty scaling and inflexibility with respect to certain shape constraints (such as nonnegativity). Here we propose Deep Random Splines, a flexible class of random functions obtained by transforming Gaussian noise through a deep neural network whose output are the parameters of a spline. Unlike Gaussian processes, Deep Random Splines allow us to readily enforce shape constraints while inheriting the richness and tractability of deep generative models. We also present an observational model for point process data which uses Deep Random Splines to model the intensity function of each point process and apply it to neural population data to obtain a low-dimensional representation of spiking activity. Inference is performed via a variational autoencoder that uses a novel recurrent encoder architecture that can handle multiple point processes as input. We use a newly collected dataset where a primate completes a pedaling task, and observe better dimensionality reduction with our model than with competing alternatives."
                    },
                    {
                        "title": "C-Learning: Horizon-Aware Cumulative Accessibility Estimation",
                        "abstract": "Multi-goal reaching is an important problem in reinforcement learning needed to achieve algorithmic generalization. Despite recent advances in this field, current algorithms suffer from three major challenges: high sample complexity, learning only a single way of reaching the goals, and difficulties in solving complex motion planning tasks. In order to address these limitations, we introduce the concept of cumulative accessibility functions, which measure the reachability of a goal from a given state within a specified horizon. We show that these functions obey a recurrence relation, which enables learning from offline interactions. We also prove that optimal cumulative accessibility functions are monotonic in the planning horizon. Additionally, our method can trade off speed and reliability in goal-reaching by suggesting multiple paths to a single goal depending on the provided horizon. We evaluate our approach on a set of multi-goal discrete and continuous control tasks. We show that our method outperforms state-of-the-art goal-reaching algorithms in success rate, sample complexity, and path optimality. Our code is available at https://github.com/layer6ai-labs/CAE, and additional visualizations can be found at https://sites.google.com/view/learning-cae/."
                    },
                    {
                        "title": "Verifying the Union of Manifolds Hypothesis for Image Data",
                        "abstract": "Deep learning has had tremendous success at learning low-dimensional representations of high-dimensional data. This success would be impossible if there was no hidden low-dimensional structure in data of interest; this existence is posited by the manifold hypothesis, which states that the data lies on an unknown manifold of low intrinsic dimension. In this paper, we argue that this hypothesis does not properly capture the low-dimensional structure typically present in image data. Assuming that data lies on a single manifold implies intrinsic dimension is identical across the entire data space, and does not allow for subregions of this space to have a different number of factors of variation. To address this deficiency, we consider the union of manifolds hypothesis, which states that data lies on a disjoint union of manifolds of varying intrinsic dimensions. We empirically verify this hypothesis on commonly-used image datasets, finding that indeed, observed data lies on a disconnected set and that intrinsic dimension is not constant. We also provide insights into the implications of the union of manifolds hypothesis in deep learning, both supervised and unsupervised, showing that designing models with an inductive bias for this structure improves performance across classification and generative modelling tasks. Our code is available at https://github.com/layer6ai-labs/UoMH."
                    },
                    {
                        "title": "A Geometric Explanation of the Likelihood OOD Detection Paradox",
                        "abstract": "Likelihood-based deep generative models (DGMs) commonly exhibit a puzzling behaviour: when trained on a relatively complex dataset, they assign higher likelihood values to out-of-distribution (OOD) data from simpler sources. Adding to the mystery, OOD samples are never generated by these DGMs despite having higher likelihoods. This two-pronged paradox has yet to be conclusively explained, making likelihood-based OOD detection unreliable. Our primary observation is that high-likelihood regions will not be generated if they contain minimal probability mass. We demonstrate how this seeming contradiction of large densities yet low probability mass can occur around data confined to low-dimensional manifolds. We also show that this scenario can be identified through local intrinsic dimension (LID) estimation, and propose a method for OOD detection which pairs the likelihoods and LID estimates obtained from a pre-trained DGM. Our method can be applied to normalizing flows and score-based diffusion models, and obtains results which match or surpass state-of-the-art OOD detection benchmarks using the same DGM backbones. Our code is available at https://github.com/layer6ai-labs/dgm_ood_detection."
                    },
                    {
                        "title": "Deep Generative Models through the Lens of the Manifold Hypothesis: A Survey and New Connections",
                        "abstract": "In recent years there has been increased interest in understanding the interplay between deep generative models (DGMs) and the manifold hypothesis. Research in this area focuses on understanding the reasons why commonly-used DGMs succeed or fail at learning distributions supported on unknown low-dimensional manifolds, as well as developing new models explicitly designed to account for manifold-supported data. This manifold lens provides both clarity as to why some DGMs (e.g. diffusion models and some generative adversarial networks) empirically surpass others (e.g. likelihood-based models such as variational autoencoders, normalizing flows, or energy-based models) at sample generation, and guidance for devising more performant DGMs. We carry out the first survey of DGMs viewed through this lens, making two novel contributions along the way. First, we formally establish that numerical instability of likelihoods in high ambient dimensions is unavoidable when modelling data with low intrinsic dimension. We then show that DGMs on learned representations of autoencoders can be interpreted as approximately minimizing Wasserstein distance: this result, which applies to latent diffusion models, helps justify their outstanding empirical results. The manifold lens provides a rich perspective from which to understand DGMs, and we aim to make this perspective more accessible and widespread."
                    }
                ]
            }
        }
    },
    "2402.16302": {
        "paper_data": {
            "title": "Graph Diffusion Policy Optimization",
            "url": "http://arxiv.org/abs/2402.16302v1",
            "arxiv_id": "2402.16302",
            "authors": [
                "Yijing Liu",
                "Chao Du",
                "Tianyu Pang",
                "Chongxuan Li",
                "Wei Chen",
                "Min Lin"
            ],
            "abstract": "Recent research has made significant progress in optimizing diffusion models for specific downstream objectives, which is an important pursuit in fields such as graph generation for drug design. However, directly applying these models to graph diffusion presents challenges, resulting in suboptimal performance. This paper introduces graph diffusion policy optimization (GDPO), a novel approach to optimize graph diffusion models for arbitrary (e.g., non-differentiable) objectives using reinforcement learning. GDPO is based on an eager policy gradient tailored for graph diffusion models, developed through meticulous analysis and promising improved performance. Experimental results show that GDPO achieves state-of-the-art performance in various graph generation tasks with complex and diverse objectives. Code is available at https://github.com/sail-sg/GDPO.",
            "introduction": "   1 Introduction  Graph generation, a key facet of graph learning, has applications in a variety of domains, including drug and material design\u00a0(Simonovsky & Komodakis, 2018), code completion\u00a0(Brockschmidt et\u00a0al., 2018), social network analysis\u00a0(Grover et\u00a0al., 2018), and neural architecture search\u00a0(Xie et\u00a0al., 2019). Numerous studies have shown significant progress in graph generation with deep generative models\u00a0(Kipf & Welling, 2016; Wang et\u00a0al., 2017; You et\u00a0al., 2018b; Guo & Zhao, 2020). Recent advances in the field, most notably the introduction of graph diffusion probabilistic models (DPMs)\u00a0(Vignac et\u00a0al., 2022; Jo et\u00a0al., 2022), represent a significant breakthrough. These methods can learn the underlying distribution from graph-structured data samples and produce high-quality novel graph structures.   In many use cases of graph generation, the primary focus is on achieving specific objectives, such as high drug efficacy\u00a0(Trott & Olson, 2009) or creating novel graphs with special properties\u00a0(Harary & Nash-Williams, 1965). These objectives are often expressed as reward signals, such as binding affinity\u00a0(Ciepli\u0144ski et\u00a0al., 2020) and synthetic accessibility\u00a0(Boda et\u00a0al., 2007), rather than a set of training graph samples. Therefore, a more pertinent goal in such scenarios is to train graph generative models to meet these predefined objectives directly, rather than learning to match a distribution over training data\u00a0(Zhou et\u00a0al., 2018).   A major challenge in this context is that most signals are non-differentiable w.r.t.\u00a0graph representations, making it difficult to apply many optimization algorithms. To address this, methods based on property predictors\u00a0(Jin et\u00a0al., 2020b; Lee et\u00a0al., 2022) learn parametric models to predict the reward signals, providing gradient guidance for graph generation. However, since reward signals can be highly complex (e.g., results from physical simulations), these predictors often struggle to provide accurate guidance\u00a0(Nguyen & Wei, 2019). An alternative direction is to learn graph generative models as policies through reinforcement learning (RL)\u00a0(Zhou et\u00a0al., 2018), which enables the integration of exact reward signals into the optimization. However, existing work primarily explores earlier graph generative models and has yet to leverage the superior performance of graph DPMs\u00a0(Cao & Kipf, 2018; You et\u00a0al., 2018a). On the other hand, several pioneer works have seen significant progress in optimizing continuous-variable (e.g., images) DPMs for downstream objectives\u00a0(Black et\u00a0al., 2023; Fan et\u00a0al., 2023). The central idea is to formulate the sampling process as a policy, with the objective serving as a reward, and then learn the model using policy gradient methods. However, when these approaches are directly extended to (discrete-variable) graph DPMs, we empirically observe a substantial failure, which we will illustrate and discuss in Sec.\u00a04.   To close this gap, we present graph diffusion policy optimization (GDPO), a policy gradient method designed to optimize graph DPMs for arbitrary reward signals. Using an RL formulation similar to that introduced by Black et\u00a0al. (2023) and Fan et\u00a0al. (2023) for continuous-variable DPMs, we first adapt the discrete diffusion process of graph DPMs to a Markov decision process (MDP) and formulate the learning problem as policy optimization. Then, to address the observed empirical failure, we introduce a slight modification to the standard policy gradient method REINFORCE\u00a0(Sutton & Barto, 1998), dubbed the eager policy gradient and specifically tailored for graph DPMs. Experimental evaluation shows that GDPO proves effective across various scenarios and achieves high sample efficiency. Remarkably, our method achieves a 41.64%percent41.64\\mathbf{41.64\\%}bold_41.64 % to 81.97%percent81.97\\mathbf{81.97\\%}bold_81.97 % average reduction in generation-test",
            "references": [
                {
                    "title": "Aligning Text-to-Image Diffusion Models with Reward Backpropagation",
                    "abstract": "Text-to-image diffusion models have recently emerged at the forefront of image generation, powered by very large-scale unsupervised or weakly supervised text-to-image training datasets. Due to their unsupervised training, controlling their behavior in downstream tasks, such as maximizing human-perceived image quality, image-text alignment, or ethical image generation, is difficult. Recent works finetune diffusion models to downstream reward functions using vanilla reinforcement learning, notorious for the high variance of the gradient estimators. In this paper, we propose AlignProp, a method that aligns diffusion models to downstream reward functions using end-to-end backpropagation of the reward gradient through the denoising process. While naive implementation of such backpropagation would require prohibitive memory resources for storing the partial derivatives of modern text-to-image models, AlignProp finetunes low-rank adapter weight modules and uses gradient checkpointing, to render its memory usage viable. We test AlignProp in finetuning diffusion models to various objectives, such as image-text semantic alignment, aesthetics, compressibility and controllability of the number of objects present, as well as their combinations. We show AlignProp achieves higher rewards in fewer training steps than alternatives, while being conceptually simpler, making it a straightforward choice for optimizing diffusion models for differentiable reward functions of interest. Code and Visualization results are available at https://align-prop.github.io/."
                },
                {
                    "title": "Directly Fine-Tuning Diffusion Models on Differentiable Rewards",
                    "abstract": "We present Direct Reward Fine-Tuning (DRaFT), a simple and effective method for fine-tuning diffusion models to maximize differentiable reward functions, such as scores from human preference models. We first show that it is possible to backpropagate the reward function gradient through the full sampling procedure, and that doing so achieves strong performance on a variety of rewards, outperforming reinforcement learning-based approaches. We then propose more efficient variants of DRaFT: DRaFT-K, which truncates backpropagation to only the last K steps of sampling, and DRaFT-LV, which obtains lower-variance gradient estimates for the case when K=1. We show that our methods work well for a variety of reward functions and can be used to substantially improve the aesthetic quality of images generated by Stable Diffusion 1.4. Finally, we draw connections between our approach and prior work, providing a unifying perspective on the design space of gradient-based fine-tuning algorithms."
                },
                {
                    "title": "DPOK: Reinforcement Learning for Fine-tuning Text-to-Image Diffusion Models",
                    "abstract": "Learning from human feedback has been shown to improve text-to-image models. These techniques first learn a reward function that captures what humans care about in the task and then improve the models based on the learned reward function. Even though relatively simple approaches (e.g., rejection sampling based on reward scores) have been investigated, fine-tuning text-to-image models with the reward function remains challenging. In this work, we propose using online reinforcement learning (RL) to fine-tune text-to-image models. We focus on diffusion models, defining the fine-tuning task as an RL problem, and updating the pre-trained text-to-image diffusion models using policy gradient to maximize the feedback-trained reward. Our approach, coined DPOK, integrates policy optimization with KL regularization. We conduct an analysis of KL regularization for both RL fine-tuning and supervised fine-tuning. In our experiments, we show that DPOK is generally superior to supervised fine-tuning with respect to both image-text alignment and image quality. Our code is available at https://github.com/google-research/google-research/tree/master/dpok."
                },
                {
                    "title": "Training Diffusion Models with Reinforcement Learning",
                    "abstract": "Diffusion models are a class of flexible generative models trained with an approximation to the log-likelihood objective. However, most use cases of diffusion models are not concerned with likelihoods, but instead with downstream objectives such as human-perceived image quality or drug effectiveness. In this paper, we investigate reinforcement learning methods for directly optimizing diffusion models for such objectives. We describe how posing denoising as a multi-step decision-making problem enables a class of policy gradient algorithms, which we refer to as denoising diffusion policy optimization (DDPO), that are more effective than alternative reward-weighted likelihood approaches. Empirically, DDPO is able to adapt text-to-image diffusion models to objectives that are difficult to express via prompting, such as image compressibility, and those derived from human feedback, such as aesthetic quality. Finally, we show that DDPO can improve prompt-image alignment using feedback from a vision-language model without the need for additional data collection or human annotation. The project's website can be found at http://rl-diffusion.github.io ."
                },
                {
                    "title": "A Survey on Graph Diffusion Models: Generative AI in Science for Molecule, Protein and Material",
                    "abstract": "Diffusion models have become a new SOTA generative modeling method in various fields, for which there are multiple survey works that provide an overall survey. With the number of articles on diffusion models increasing exponentially in the past few years, there is an increasing need for surveys of diffusion models on specific fields. In this work, we are committed to conducting a survey on the graph diffusion models. Even though our focus is to cover the progress of diffusion models in graphs, we first briefly summarize how other generative modeling methods are used for graphs. After that, we introduce the mechanism of diffusion models in various forms, which facilitates the discussion on the graph diffusion models. The applications of graph diffusion models mainly fall into the category of AI-generated content (AIGC) in science, for which we mainly focus on how graph diffusion models are utilized for generating molecules and proteins but also cover other cases, including materials design. Moreover, we discuss the issue of evaluating diffusion models in the graph domain and the existing challenges."
                },
                {
                    "title": "Aligning Text-to-Image Models using Human Feedback",
                    "abstract": "Deep generative models have shown impressive results in text-to-image synthesis. However, current text-to-image models often generate images that are inadequately aligned with text prompts. We propose a fine-tuning method for aligning such models using human feedback, comprising three stages. First, we collect human feedback assessing model output alignment from a set of diverse text prompts. We then use the human-labeled image-text dataset to train a reward function that predicts human feedback. Lastly, the text-to-image model is fine-tuned by maximizing reward-weighted likelihood to improve image-text alignment. Our method generates objects with specified colors, counts and backgrounds more accurately than the pre-trained model. We also analyze several design choices and find that careful investigations on such design choices are important in balancing the alignment-fidelity tradeoffs. Our results demonstrate the potential for learning from human feedback to significantly improve text-to-image models."
                },
                {
                    "title": "Optimizing DDPM Sampling with Shortcut Fine-Tuning",
                    "abstract": "In this study, we propose Shortcut Fine-Tuning (SFT), a new approach for addressing the challenge of fast sampling of pretrained Denoising Diffusion Probabilistic Models (DDPMs). SFT advocates for the fine-tuning of DDPM samplers through the direct minimization of Integral Probability Metrics (IPM), instead of learning the backward diffusion process. This enables samplers to discover an alternative and more efficient sampling shortcut, deviating from the backward diffusion process. Inspired by a control perspective, we propose a new algorithm SFT-PG: Shortcut Fine-Tuning with Policy Gradient, and prove that under certain assumptions, gradient descent of diffusion models with respect to IPM is equivalent to performing policy gradient. To our best knowledge, this is the first attempt to utilize reinforcement learning (RL) methods to train diffusion models. Through empirical evaluation, we demonstrate that our fine-tuning method can further enhance existing fast DDPM samplers, resulting in sample quality comparable to or even surpassing that of the full-step model across various datasets."
                },
                {
                    "title": "DiGress: Discrete Denoising diffusion for graph generation",
                    "abstract": "This work introduces DiGress, a discrete denoising diffusion model for generating graphs with categorical node and edge attributes. Our model utilizes a discrete diffusion process that progressively edits graphs with noise, through the process of adding or removing edges and changing the categories. A graph transformer network is trained to revert this process, simplifying the problem of distribution learning over graphs into a sequence of node and edge classification tasks. We further improve sample quality by introducing a Markovian noise model that preserves the marginal distribution of node and edge types during diffusion, and by incorporating auxiliary graph-theoretic features. A procedure for conditioning the generation on graph-level features is also proposed. DiGress achieves state-of-the-art performance on molecular and non-molecular datasets, with up to 3x validity improvement on a planar graph dataset. It is also the first model to scale to the large GuacaMol dataset containing 1.3M drug-like molecules without the use of molecule-specific representations."
                },
                {
                    "title": "LIMO: Latent Inceptionism for Targeted Molecule Generation",
                    "abstract": "Generation of drug-like molecules with high binding affinity to target proteins remains a difficult and resource-intensive task in drug discovery. Existing approaches primarily employ reinforcement learning, Markov sampling, or deep generative models guided by Gaussian processes, which can be prohibitively slow when generating molecules with high binding affinity calculated by computationally-expensive physics-based methods. We present Latent Inceptionism on Molecules (LIMO), which significantly accelerates molecule generation with an inceptionism-like technique. LIMO employs a variational autoencoder-generated latent space and property prediction by two neural networks in sequence to enable faster gradient-based reverse-optimization of molecular properties. Comprehensive experiments show that LIMO performs competitively on benchmark tasks and markedly outperforms state-of-the-art techniques on the novel task of generating drug-like compounds with high binding affinity, reaching nanomolar range against two protein targets. We corroborate these docking-based results with more accurate molecular dynamics-based calculations of absolute binding free energy and show that one of our generated drug-like compounds has a predicted K D (a measure of binding affinity) of 6 \u00b7 10-14 M against the human estrogen receptor, well beyond the affinities of typical early-stage drug candidates and most FDA-approved drugs to their respective targets. Code is available at https://github.com/Rose-STL-Lab/LIMO."
                },
                {
                    "title": "Exploring Chemical Space with Score-based Out-of-distribution Generation",
                    "abstract": "A well-known limitation of existing molecular generative models is that the generated molecules highly resemble those in the training set. To generate truly novel molecules that may have even better properties for de novo drug discovery, more powerful exploration in the chemical space is necessary. To this end, we propose Molecular Out-Of-distribution Diffusion(MOOD), a score-based diffusion scheme that incorporates out-of-distribution (OOD) control in the generative stochastic differential equation (SDE) with simple control of a hyperparameter, thus requires no additional costs. Since some novel molecules may not meet the basic requirements of real-world drugs, MOOD performs conditional generation by utilizing the gradients from a property predictor that guides the reverse-time diffusion process to high-scoring regions according to target properties such as protein-ligand interactions, drug-likeness, and synthesizability. This allows MOOD to search for novel and meaningful molecules rather than generating unseen yet trivial ones. We experimentally validate that MOOD is able to explore the chemical space beyond the training distribution, generating molecules that outscore ones found with existing methods, and even the top 0.01% of the original training pool. Our code is available at https://github.com/SeulLee05/MOOD."
                },
                {
                    "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                    "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "SPECTRE : Spectral Conditioning Helps to Overcome the Expressivity Limits of One-shot Graph Generators",
                    "abstract": "We approach the graph generation problem from a spectral perspective by first generating the dominant parts of the graph Laplacian spectrum and then building a graph matching these eigenvalues and eigenvectors. Spectral conditioning allows for direct modeling of the global and local graph structure and helps to overcome the expressivity and mode collapse issues of one-shot graph generators. Our novel GAN, called SPECTRE, enables the one-shot generation of much larger graphs than previously possible with one-shot models. SPECTRE outperforms state-of-the-art deep autoregressive generators in terms of modeling fidelity, while also avoiding expensive sequential generation and dependence on node ordering. A case in point, in sizable synthetic and real-world graphs SPECTRE achieves a 4-to-170 fold improvement over the best competitor that does not overfit and is 23-to-30 times faster than autoregressive generators."
                },
                {
                    "title": "MGCVAE: Multi-objective Inverse Design via Molecular Graph Conditional Variational Autoencoder",
                    "abstract": "The ultimate goal of various fields is to directly generate molecules with desired properties, such as water-soluble molecules in drug development and molecules suitable for organic light-emitting diodes or photosensitizers in the field of development of new organic materials. This study proposes a molecular graph generative model based on an autoencoder for the de novo design. The performance of the molecular graph conditional variational autoencoder (MGCVAE) for generating molecules with specific desired properties was investigated by comparing it to a molecular graph variational autoencoder (MGVAE). Furthermore, multi-objective optimization for MGCVAE was applied to satisfy the two selected properties simultaneously. In this study, two physical properties, calculated logP and molar refractivity, were used as optimization targets for designing de novo molecules. Consequently, it was confirmed that among the generated molecules, 25.89% of the optimized molecules were generated in MGCVAE compared to 0.66% in MGVAE. This demonstrates that MGCVAE effectively produced drug-like molecules with two target properties. The results of this study suggest that these graph-based data-driven models are an effective method for designing new molecules that fulfill various physical properties."
                },
                {
                    "title": "Score-based Generative Modeling of Graphs via the System of Stochastic Differential Equations",
                    "abstract": "Generating graph-structured data requires learning the underlying distribution of graphs. Yet, this is a challenging problem, and the previous graph generative methods either fail to capture the permutation-invariance property of graphs or cannot sufficiently model the complex dependency between nodes and edges, which is crucial for generating real-world graphs such as molecules. To overcome such limitations, we propose a novel score-based generative model for graphs with a continuous-time framework. Specifically, we propose a new graph diffusion process that models the joint distribution of the nodes and edges through a system of stochastic differential equations (SDEs). Then, we derive novel score matching objectives tailored for the proposed diffusion process to estimate the gradient of the joint log-density with respect to each component, and introduce a new solver for the system of SDEs to efficiently sample from the reverse diffusion process. We validate our graph generation method on diverse datasets, on which it either achieves significantly superior or competitive performance to the baselines. Further analysis shows that our method is able to generate molecules that lie close to the training distribution yet do not violate the chemical valency rule, demonstrating the effectiveness of the system of SDEs in modeling the node-edge relationships. Our code is available at https://github.com/harryjo97/GDSS."
                },
                {
                    "title": "Hit and Lead Discovery with Explorative RL and Fragment-based Molecule Generation",
                    "abstract": "Recently, utilizing reinforcement learning (RL) to generate molecules with desired properties has been highlighted as a promising strategy for drug design. A molecular docking program - a physical simulation that estimates protein-small molecule binding affinity - can be an ideal reward scoring function for RL, as it is a straightforward proxy of the therapeutic potential. Still, two imminent challenges exist for this task. First, the models often fail to generate chemically realistic and pharmacochemically acceptable molecules. Second, the docking score optimization is a difficult exploration problem that involves many local optima and less smooth surfaces with respect to molecular structure. To tackle these challenges, we propose a novel RL framework that generates pharmacochemically acceptable molecules with large docking scores. Our method - Fragment-based generative RL with Explorative Experience replay for Drug design (FREED) - constrains the generated molecules to a realistic and qualified chemical space and effectively explores the space to find drugs by coupling our fragment-based generation method and a novel error-prioritized experience replay (PER). We also show that our model performs well on both de novo and scaffold-based schemes. Our model produces molecules of higher quality compared to existing methods while achieving state-of-the-art performance on two of three targets in terms of the docking scores of the generated molecules. We further show with ablation studies that our method, predictive error-PER (FREED(PE)), significantly improves the model performance."
                },
                {
                    "title": "Structured Denoising Diffusion Models in Discrete State-Spaces",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) (Ho et al. 2020) have shown impressive results on image and waveform generation in continuous state spaces. Here, we introduce Discrete Denoising Diffusion Probabilistic Models (D3PMs), diffusion-like generative models for discrete data that generalize the multinomial diffusion model of Hoogeboom et al. 2021, by going beyond corruption processes with uniform transition probabilities. This includes corruption with transition matrices that mimic Gaussian kernels in continuous space, matrices based on nearest neighbors in embedding space, and matrices that introduce absorbing states. The third allows us to draw a connection between diffusion models and autoregressive and mask-based generative models. We show that the choice of transition matrix is an important design decision that leads to improved results in image and text domains. We also introduce a new loss function that combines the variational lower bound with an auxiliary cross entropy loss. For text, this model class achieves strong results on character-level text generation while scaling to large vocabularies on LM1B. On the image dataset CIFAR-10, our models approach the sample quality and exceed the log-likelihood of the continuous-space DDPM model."
                },
                {
                    "title": "AutoDock Vina 1.2.0: New Docking Methods, Expanded Force Field, and Python Bindings",
                    "abstract": "AutoDock Vina is arguably one of the fastest and most widely used open-source programs for molecular docking. However, compared to other programs in the AutoDock Suite, it lacks support for modeling specific features such as macrocycles or explicit water molecules. Here, we describe the implementation of this functionality in AutoDock Vina 1.2.0. Additionally, AutoDock Vina 1.2.0 supports the AutoDock4.2 scoring function, simultaneous docking of multiple ligands, and a batch mode for docking a large number of ligands. Furthermore, we implemented Python bindings to facilitate scripting and the development of docking workflows. This work is an effort toward the unification of the features of the AutoDock4 and AutoDock Vina programs. The source code is available at https://github.com/ccsb-scripps/AutoDock-Vina."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "MARS: Markov Molecular Sampling for Multi-objective Drug Discovery",
                    "abstract": "Searching for novel molecules with desired chemical properties is crucial in drug discovery. Existing work focuses on developing neural models to generate either molecular sequences or chemical graphs. However, it remains a big challenge to find novel and diverse compounds satisfying several properties. In this paper, we propose MARS, a method for multi-objective drug molecule discovery. MARS is based on the idea of generating the chemical candidates by iteratively editing fragments of molecular graphs. To search for high-quality candidates, it employs Markov chain Monte Carlo sampling (MCMC) on molecules with an annealing scheme and an adaptive proposal. To further improve sample efficiency, MARS uses a graph neural network (GNN) to represent and select candidate edits, where the GNN is trained on-the-fly with samples from MCMC. Experiments show that MARS achieves state-of-the-art performance in various multi-objective settings where molecular bio-activity, drug-likeness, and synthesizability are considered. Remarkably, in the most challenging setting where all four objectives are simultaneously optimized, our approach outperforms previous methods significantly in comprehensive evaluations. The code is available at this https URL."
                },
                {
                    "title": "GraphDF: A Discrete Flow Model for Molecular Graph Generation",
                    "abstract": "We consider the problem of molecular graph generation using deep models. While graphs are discrete, most existing methods use continuous latent variables, resulting in inaccurate modeling of discrete graph structures. In this work, we propose GraphDF, a novel discrete latent variable model for molecular graph generation based on normalizing flow methods. GraphDF uses invertible modulo shift transforms to map discrete latent variables to graph nodes and edges. We show that the use of discrete latent variables reduces computational costs and eliminates the negative effect of dequantization. Comprehensive experimental results show that GraphDF outperforms prior methods on random generation, property optimization, and constrained optimization tasks."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Conditional Constrained Graph Variational Autoencoders for Molecule Design",
                    "abstract": "In recent years, deep generative models for graphs have been used to generate new molecules. These models have produced good results, leading to several proposals in the literature. However, these models may have troubles learning some of the complex laws governing the chemical world. In this work, we explore the usage of the histogram of atom valences to drive the generation of molecules in such models. We present Conditional Constrained Graph Variational Autoencoder (CCGVAE), a model that implements this key-idea in a state-of-the-art model, and shows improved results on several evaluation metrics on two commonly adopted datasets for molecule generation."
                },
                {
                    "title": "A Systematic Survey on Deep Generative Models for Graph Generation",
                    "abstract": "Graphs are important data representations for describing objects and their relationships, which appear in a wide diversity of real-world scenarios. As one of a critical problem in this area, graph generation considers learning the distributions of given graphs and generating more novel graphs. Owing to their wide range of applications, generative models for graphs, which have a rich history, however, are traditionally hand-crafted and only capable of modeling a few statistical properties of graphs. Recent advances in deep generative models for graph generation is an important step towards improving the fidelity of generated graphs and paves the way for new kinds of applications. This article provides an extensive overview of the literature in the field of deep generative models for graph generation. First, the formal definition of deep generative models for the graph generation and the preliminary knowledge are provided. Second, taxonomies of deep generative models for both unconditional and conditional graph generation are proposed respectively; the existing works of each are compared and analyzed. After that, an overview of the evaluation metrics in this specific domain is provided. Finally, the applications that deep graph generation enables are summarized and five promising future research directions are highlighted."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Permutation Invariant Graph Generation via Score-Based Generative Modeling",
                    "abstract": "Learning generative models for graph-structured data is challenging because graphs are discrete, combinatorial, and the underlying data distribution is invariant to the ordering of nodes. However, most of the existing generative models for graphs are not invariant to the chosen ordering, which might lead to an undesirable bias in the learned distribution. To address this difficulty, we propose a permutation invariant approach to modeling graphs, using the recent framework of score-based generative modeling. In particular, we design a permutation equivariant, multi-channel graph neural network to model the gradient of the data distribution at the input graph (a.k.a., the score function). This permutation equivariant model of gradients implicitly defines a permutation invariant distribution for graphs. We train this graph neural network with score matching and sample from it with annealed Langevin dynamics. In our experiments, we first demonstrate the capacity of this new architecture in learning discrete graph algorithms. For graph generation, we find that our learning approach achieves better or comparable results to existing models on benchmark datasets."
                },
                {
                    "title": "Hierarchical Generation of Molecular Graphs using Structural Motifs",
                    "abstract": "Graph generation techniques are increasingly being adopted for drug discovery. Previous graph generation approaches have utilized relatively small molecular building blocks such as atoms or simple cycles, limiting their effectiveness to smaller molecules. Indeed, as we demonstrate, their performance degrades significantly for larger molecules. In this paper, we propose a new hierarchical graph encoder-decoder that employs significantly larger and more flexible graph motifs as basic building blocks. Our encoder produces a multi-resolution representation for each molecule in a fine-to-coarse fashion, from atoms to connected motifs. Each level integrates the encoding of constituents below with the graph at that level. Our autoregressive coarse-to-fine decoder adds one motif at a time, interleaving the decision of selecting a new motif with the process of resolving its attachments to the emerging molecule. We evaluate our model on multiple molecule generation tasks, including polymers, and show that our model significantly outperforms previous state-of-the-art baselines."
                },
                {
                    "title": "Multi-Objective Molecule Generation using Interpretable Substructures",
                    "abstract": "Drug discovery aims to find novel compounds with specified chemical property profiles. In terms of generative modeling, the goal is to learn to sample molecules in the intersection of multiple property constraints. This task becomes increasingly challenging when there are many property constraints. We propose to offset this complexity by composing molecules from a vocabulary of substructures that we call molecular rationales. These rationales are identified from molecules as substructures that are likely responsible for each property of interest. We then learn to expand rationales into a full molecule using graph generative models. Our final generative model composes molecules as mixtures of multiple rationale completions, and this mixture is fine-tuned to preserve the properties of interest. We evaluate our model on various drug design tasks and demonstrate significant improvements over state-of-the-art baselines in terms of accuracy, diversity, and novelty of generated compounds."
                },
                {
                    "title": "GraphAF: a Flow-based Autoregressive Model for Molecular Graph Generation",
                    "abstract": "Molecular graph generation is a fundamental problem for drug discovery and has been attracting growing attention. The problem is challenging since it requires not only generating chemically valid molecular structures but also optimizing their chemical properties in the meantime. Inspired by the recent progress in deep generative models, in this paper we propose a flow-based autoregressive model for graph generation called GraphAF. GraphAF combines the advantages of both autoregressive and flow-based approaches and enjoys: (1) high model flexibility for data density estimation; (2) efficient parallel computation for training; (3) an iterative sampling process, which allows leveraging chemical domain knowledge for valency checking. Experimental results show that GraphAF is able to generate 68% chemically valid molecules even without chemical knowledge rules and 100% valid molecules with chemical rules. The training process of GraphAF is two times faster than the existing state-of-the-art approach GCPN. After fine-tuning the model for goal-directed property optimization with reinforcement learning, GraphAF achieves state-of-the-art performance on both chemical property optimization and constrained property optimization."
                },
                {
                    "title": "Efficient Graph Generation with Graph Recurrent Attention Networks",
                    "abstract": "We propose a new family of efficient and expressive deep generative models of graphs, called Graph Recurrent Attention Networks (GRANs). Our model generates graphs one block of nodes and associated edges at a time. The block size and sampling stride allow us to trade off sample quality for efficiency. Compared to previous RNN-based graph generative models, our framework better captures the auto-regressive conditioning between the already-generated and to-be-generated parts of the graph using Graph Neural Networks (GNNs) with attention. This not only reduces the dependency on node ordering but also bypasses the long-term bottleneck caused by the sequential nature of RNNs. Moreover, we parameterize the output distribution per block using a mixture of Bernoulli, which captures the correlations among generated edges within the block. Finally, we propose to handle node orderings in generation by marginalizing over a family of canonical orderings. On standard benchmarks, we achieve state-of-the-art time efficiency and sample quality compared to previous models. Additionally, we show our model is capable of generating large graphs of up to 5K nodes with good quality. Our code is released at: \\url{https://github.com/lrjconan/GRAN}."
                },
                {
                    "title": "Semi-Implicit Graph Variational Auto-Encoders",
                    "abstract": "Semi-implicit graph variational auto-encoder (SIG-VAE) is proposed to expand the flexibility of variational graph auto-encoders (VGAE) to model graph data. SIG-VAE employs a hierarchical variational framework to enable neighboring node sharing for better generative modeling of graph dependency structure, together with a Bernoulli-Poisson link decoder. Not only does this hierarchical construction provide a more flexible generative graph model to better capture real-world graph properties, but also does SIG-VAE naturally lead to semi-implicit hierarchical variational inference that allows faithful modeling of implicit posteriors of given graph data, which may exhibit heavy tails, multiple modes, skewness, and rich dependency structures. Compared to VGAE, the derived graph latent representations by SIG-VAE are more interpretable, due to more expressive generative model and more faithful inference enabled by the flexible semi-implicit construction. Extensive experiments with a variety of graph data show that SIG-VAE significantly outperforms state-of-the-art methods on several different graph analytic tasks."
                },
                {
                    "title": "AGL-Score: Algebraic Graph Learning Score for Protein-Ligand Binding Scoring, Ranking, Docking, and Screening",
                    "abstract": "Although algebraic graph theory-based models have been widely applied in physical modeling and molecular studies, they are typically incompetent in the analysis and prediction of biomolecular properties, confirming the common belief that \"one cannot hear the shape of a drum\". A new development in the century-old question about the spectrum-geometry relationship is provided. Novel algebraic graph learning score (AGL-Score) models are proposed to encode high-dimensional physical and biological information into intrinsically low-dimensional representations. The proposed AGL-Score models employ multiscale weighted colored subgraphs to describe crucial molecular and biomolecular interactions in terms of graph invariants derived from graph Laplacian, its pseudo-inverse, and adjacency matrices. Additionally, AGL-Score models are integrated with an advanced machine learning algorithm to predict biomolecular macroscopic properties from the low-dimensional graph representation of biomolecular structures. The proposed AGL-Score models are extensively validated for their scoring power, ranking power, docking power, and screening power via a number of benchmark datasets, namely CASF-2007, CASF-2013, and CASF-2016. Numerical results indicate that the proposed AGL-Score models are able to outperform other state-of-the-art scoring functions in protein-ligand binding scoring, ranking, docking, and screening. This study indicates that machine learning methods are powerful tools for molecular docking and virtual screening. It also indicates that spectral geometry or spectral graph theory has the ability to infer geometric properties."
                },
                {
                    "title": "Graph Normalizing Flows",
                    "abstract": "We introduce graph normalizing flows: a new, reversible graph neural network model for prediction and generation. On supervised tasks, graph normalizing flows perform similarly to message passing neural networks, but at a significantly reduced memory footprint, allowing them to scale to larger graphs. In the unsupervised case, we combine graph normalizing flows with a novel graph auto-encoder to create a generative model of graph structures. Our model is permutation-invariant, generating entire graphs with a single feed-forward pass, and achieves competitive results with the state-of-the art auto-regressive models, while being better suited to parallel computing architectures."
                },
                {
                    "title": "Exploring Randomly Wired Neural Networks for Image Recognition",
                    "abstract": "Neural networks for image recognition have evolved through extensive manual design from simple chain-like models to structures with multiple wiring paths. The success of ResNets and DenseNets is due in large part to their innovative wiring plans. Now, neural architecture search (NAS) studies are exploring the joint optimization of wiring and operation types, however, the space of possible wirings is constrained and still driven by manual design despite being searched. In this paper, we explore a more diverse set of connectivity patterns through the lens of randomly wired neural networks. To do this, we first define the concept of a stochastic network generator that encapsulates the entire network generation process. Encapsulation provides a unified view of NAS and randomly wired networks. Then, we use three classical random graph models to generate randomly wired graphs for networks. The results are surprising: several variants of these random generators yield network instances that have competitive accuracy on the ImageNet benchmark. These results suggest that new efforts focusing on designing better network generators may lead to new breakthroughs by exploring less constrained search spaces with more room for novel design."
                },
                {
                    "title": "Molecular Sets (MOSES): A Benchmarking Platform for Molecular Generation Models",
                    "abstract": "Generative models are becoming a tool of choice for exploring the molecular space. These models learn on a large training dataset and produce novel molecular structures with similar properties. Generated structures can be utilized for virtual screening or training semi-supervized predictive models in the downstream tasks. While there are plenty of generative models, it is unclear how to compare and rank them. In this work, we introduce a benchmarking platform called Molecular Sets (MOSES) to standardize training and comparison of molecular generative models. MOSES provides training and testing datasets, and a set of metrics to evaluate the quality and diversity of generated structures. We have implemented and compared several molecular generation models and suggest to use our results as reference points for further advancements in generative chemistry research. The platform and source code are available at https://github.com/molecularsets/moses."
                },
                {
                    "title": "Graph Convolutional Policy Network for Goal-Directed Molecular Graph Generation",
                    "abstract": "Generating novel graph structures that optimize given objectives while obeying some given underlying rules is fundamental for chemistry, biology and social science research. This is especially important in the task of molecular graph generation, whose goal is to discover novel molecules with desired properties such as drug-likeness and synthetic accessibility, while obeying physical laws such as chemical valency. However, designing models to find molecules that optimize desired properties while incorporating highly complex and non-differentiable rules remains to be a challenging task. Here we propose Graph Convolutional Policy Network (GCPN), a general graph convolutional network based model for goal-directed graph generation through reinforcement learning. The model is trained to optimize domain-specific rewards and adversarial loss through policy gradient, and acts in an environment that incorporates domain-specific rules. Experimental results show that GCPN can achieve 61% improvement on chemical property optimization over state-of-the-art baselines while resembling known molecules, and achieve 184% improvement on the constrained property optimization task."
                },
                {
                    "title": "MolGAN: An implicit generative model for small molecular graphs",
                    "abstract": "eep generative models for graph-structured data offer a new angle on the problem of chemical synthesis: by optimizing differentiable models that directly generate molecular graphs, it is pos-sible to side-step expensive search procedures in the discrete and vast space of chemical structures. We introduce MolGAN, an implicit, likelihood-free generative model for small molecular graphs that circumvents the need for expensive graph matching procedures or node ordering heuris-tics of previous likelihood-based methods. Our method adapts generative adversarial networks (GANs) to operate directly on graph-structured data. We combine our approach with a reinforce-ment learning objective to encourage the genera-tion of molecules with specific desired chemical properties. In experiments on the QM9 chemi-cal database, we demonstrate that our model is capable of generating close to 100% valid com-pounds. MolGAN compares favorably both to recent proposals that use string-based (SMILES) representations of molecules and to a likelihood-based method that directly generates graphs, al-beit being susceptible to mode collapse."
                },
                {
                    "title": "Constrained Graph Variational Autoencoders for Molecule Design",
                    "abstract": "Graphs are ubiquitous data structures for representing interactions between entities. With an emphasis on the use of graphs to represent chemical molecules, we explore the task of learning to generate graphs that conform to a distribution observed in training data. We propose a variational autoencoder model in which both encoder and decoder are graph-structured. Our decoder assumes a sequential ordering of graph extension steps and we discuss and analyze design choices that mitigate the potential downsides of this linearization. Experiments compare our approach with a wide range of baselines on the molecule generation task and show that our method is more successful at matching the statistics of the original dataset on semantically important metrics. Furthermore, we show that by using appropriate shaping of the latent space, our model allows us to design molecules that are (locally) optimal in desired properties."
                },
                {
                    "title": "Generative Code Modeling with Graphs",
                    "abstract": "Generative models for source code are an interesting structured prediction problem, requiring to reason about both hard syntactic and semantic constraints as well as about natural, likely programs. We present a novel model for this problem that uses a graph to represent the intermediate state of the generated output. The generative procedure interleaves grammar-driven expansion steps with graph augmentation and neural message passing steps. An experimental evaluation shows that our new model can generate semantically meaningful expressions, outperforming a range of strong baselines."
                },
                {
                    "title": "Graphite: Iterative Generative Modeling of Graphs",
                    "abstract": "Graphs are a fundamental abstraction for modeling relational data. However, graphs are discrete and combinatorial in nature, and learning representations suitable for machine learning tasks poses statistical and computational challenges. In this work, we propose Graphite, an algorithmic framework for unsupervised learning of representations over nodes in large graphs using deep latent variable generative models. Our model parameterizes variational autoencoders (VAE) with graph neural networks, and uses a novel iterative graph refinement strategy inspired by low-rank approximations for decoding. On a wide variety of synthetic and benchmark datasets, Graphite outperforms competing approaches for the tasks of density estimation, link prediction, and node classification. Finally, we derive a theoretical connection between message passing in graph neural networks and mean-field variational inference."
                },
                {
                    "title": "GraphRNN: Generating Realistic Graphs with Deep Auto-regressive Models",
                    "abstract": "Modeling and generating graphs is fundamental for studying networks in biology, engineering, and social sciences. However, modeling complex distributions over graphs and then efficiently sampling from these distributions is challenging due to the non-unique, high-dimensional nature of graphs and the complex, non-local dependencies that exist between edges in a given graph. Here we propose GraphRNN, a deep autoregressive model that addresses the above challenges and approximates any distribution of graphs with minimal assumptions about their structure. GraphRNN learns to generate graphs by training on a representative set of graphs and decomposes the graph generation process into a sequence of node and edge formations, conditioned on the graph structure generated so far. \nIn order to quantitatively evaluate the performance of GraphRNN, we introduce a benchmark suite of datasets, baselines and novel evaluation metrics based on Maximum Mean Discrepancy, which measure distances between sets of graphs. Our experiments show that GraphRNN significantly outperforms all baselines, learning to generate diverse graphs that match the structural characteristics of a target set, while also scaling to graphs 50 times larger than previous deep models."
                },
                {
                    "title": "GraphGAN: Graph Representation Learning with Generative Adversarial Nets",
                    "abstract": "\n \n The goal of graph representation learning is to embed each vertex in a graph into a low-dimensional vector space. Existing graph representation learning methods can be classified into two categories: generative models that learn the underlying connectivity distribution in the graph, and discriminative models that predict the probability of edge existence between a pair of vertices. In this paper, we propose GraphGAN, an innovative graph representation learning framework unifying above two classes of methods, in which the generative model and discriminative model play a game-theoretical minimax game. Specifically, for a given vertex, the generative model tries to fit its underlying true connectivity distribution over all other vertices and produces \"fake\" samples to fool the discriminative model, while the discriminative model tries to detect whether the sampled vertex is from ground truth or generated by the generative model. With the competition between these two models, both of them can alternately and iteratively boost their performance. Moreover, when considering the implementation of generative model, we propose a novel graph softmax to overcome the limitations of traditional softmax function, which can be proven satisfying desirable properties of normalization, graph structure awareness, and computational efficiency. Through extensive experiments on real-world datasets, we demonstrate that GraphGAN achieves substantial gains in a variety of applications, including link prediction, node classification, and recommendation, over state-of-the-art baselines.\n \n"
                },
                {
                    "title": "Reinforcement Learning for Bandit Neural Machine Translation with Simulated Human Feedback",
                    "abstract": "Machine translation is a natural candidate problem for reinforcement learning from human feedback: users provide quick, dirty ratings on candidate translations to guide a system to improve. Yet, current neural machine translation training focuses on expensive human-generated reference translations. We describe a reinforcement learning algorithm that improves neural machine translation systems from simulated human feedback. Our algorithm combines the advantage actor-critic algorithm (Mnih et al., 2016) with the attention-based neural encoder-decoder architecture (Luong et al., 2015). This algorithm (a) is well-designed for problems with a large action space and delayed rewards, (b) effectively optimizes traditional corpus-level machine translation metrics, and (c) is robust to skewed, high-variance, granular feedback modeled after actual human behaviors."
                },
                {
                    "title": "Variational Graph Auto-Encoders",
                    "abstract": "We introduce the variational graph auto-encoder (VGAE), a framework for unsupervised learning on graph-structured data based on the variational auto-encoder (VAE). This model makes use of latent variables and is capable of learning interpretable latent representations for undirected graphs. We demonstrate this model using a graph convolutional network (GCN) encoder and a simple inner product decoder. Our model achieves competitive results on a link prediction task in citation networks. In contrast to most existing models for unsupervised learning on graph-structured data and link prediction, our model can naturally incorporate node features, which significantly improves predictive performance on a number of benchmark datasets."
                },
                {
                    "title": "Efficient Graphlet Counting for Large Networks",
                    "abstract": "From social science to biology, numerous applications often rely on graphlets for intuitive and meaningful characterization of networks at both the global macro-level as well as the local micro-level. While graphlets have witnessed a tremendous success and impact in a variety of domains, there has yet to be a fast and efficient approach for computing the frequencies of these subgraph patterns. However, existing methods are not scalable to large networks with millions of nodes and edges, which impedes the application of graphlets to new problems that require large-scale network analysis. To address these problems, we propose a fast, efficient, and parallel algorithm for counting graphlets of size k={3,4}-nodes that take only a fraction of the time to compute when compared with the current methods used. The proposed graphlet counting algorithms leverages a number of proven combinatorial arguments for different graphlets. For each edge, we count a few graphlets, and with these counts along with the combinatorial arguments, we obtain the exact counts of others in constant time. On a large collection of 300+ networks from a variety of domains, our graphlet counting strategies are on average 460x faster than current methods. This brings new opportunities to investigate the use of graphlets on much larger networks and newer applications as we show in the experiments. To the best of our knowledge, this paper provides the largest graphlet computations to date as well as the largest systematic investigation on over 300+ networks from a variety of domains."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "ZINC: A Free Tool to Discover Chemistry for Biology",
                    "abstract": "ZINC is a free public resource for ligand discovery. The database contains over twenty million commercially available molecules in biologically relevant representations that may be downloaded in popular ready-to-dock formats and subsets. The Web site also enables searches by structure, biological activity, physical property, vendor, catalog number, name, and CAS number. Small custom subsets may be created, edited, shared, docked, downloaded, and conveyed to a vendor for purchase. The database is maintained and curated for a high purchasing success rate and is freely available at zinc.docking.org."
                },
                {
                    "title": "A Kernel Two-Sample Test",
                    "abstract": "We propose a framework for analyzing and comparing distributions, which we use to construct statistical tests to determine if two samples are drawn from different distributions. Our test statistic is the largest difference in expectations over functions in the unit ball of a reproducing kernel Hilbert space (RKHS), and is called the maximum mean discrepancy (MMD).We present two distribution free tests based on large deviation bounds for the MMD, and a third test based on the asymptotic distribution of this statistic. The MMD can be computed in quadratic time, although efficient linear time approximations are available. Our statistic is an instance of an integral probability metric, and various classical metrics on distributions are obtained when alternative function classes are used in place of an RKHS. We apply our two-sample tests to a variety of problems, including attribute matching for databases using the Hungarian marriage method, where they perform strongly. Excellent performance is also obtained when comparing distributions over graphs, for which these are the first such tests."
                },
                {
                    "title": "A Machine Learning-Based Method To Improve Docking Scoring Functions and Its Application to Drug Repurposing",
                    "abstract": "Docking scoring functions are notoriously weak predictors of binding affinity. They typically assign a common set of weights to the individual energy terms that contribute to the overall energy score; however, these weights should be gene family dependent. In addition, they incorrectly assume that individual interactions contribute toward the total binding affinity in an additive manner. In reality, noncovalent interactions often depend on one another in a nonlinear manner. In this paper, we show how the use of support vector machines (SVMs), trained by associating sets of individual energy terms retrieved from molecular docking with the known binding affinity of each compound from high-throughput screening experiments, can be used to improve the correlation between known binding affinities and those predicted by the docking program eHiTS. We construct two prediction models: a regression model trained using IC(50) values from BindingDB, and a classification model trained using active and decoy compounds from the Directory of Useful Decoys (DUD). Moreover, to address the issue of overrepresentation of negative data in high-throughput screening data sets, we have designed a multiple-planar SVM training procedure for the classification model. The increased performance that both SVMs give when compared with the original eHiTS scoring function highlights the potential for using nonlinear methods when deriving overall energy scores from their individual components. We apply the above methodology to train a new scoring function for direct inhibitors of Mycobacterium tuberculosis (M.tb) InhA. By combining ligand binding site comparison with the new scoring function, we propose that phosphodiesterase inhibitors can potentially be repurposed to target M.tb InhA. Our methodology may be applied to other gene families for which target structures and activity data are available, as demonstrated in the work presented here."
                },
                {
                    "title": "AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading",
                    "abstract": "AutoDock Vina, a new program for molecular docking and virtual screening, is presented. AutoDock Vina achieves an approximately two orders of magnitude speed\u2010up compared with the molecular docking software previously developed in our lab (AutoDock 4), while also significantly improving the accuracy of the binding mode predictions, judging by our tests on the training set used in AutoDock 4 development. Further speed\u2010up is achieved from parallelism, by using multithreading on multicore machines. AutoDock Vina automatically calculates the grid maps and clusters the results in a way transparent to the user. \u00a9 2009 Wiley Periodicals, Inc. J Comput Chem 2010"
                },
                {
                    "title": "Network clustering coefficient without degree-correlation biases.",
                    "abstract": "The clustering coefficient quantifies how well connected are the neighbors of a vertex in a graph. In real networks it decreases with the vertex degree, which has been taken as a signature of the network hierarchical structure. Here we show that this signature of hierarchical structure is a consequence of degree-correlation biases in the clustering coefficient definition. We introduce a definition in which the degree-correlation biases are filtered out, and provide evidence that in real networks the clustering coefficient is constant or decays logarithmically with vertex degree."
                },
                {
                    "title": "On Eulerian and Hamiltonian Graphs and Line Graphs",
                    "abstract": "A graph G has a finite set V of points and a set X of lines each of which joins two distinct points (called its end-points), and no two lines join the same pair of points. A graph with one point and no line is trivial. A line is incident with each of its end-points. Two points are adjacent if they are joined by a line. The degree of a point is the number of lines incident with it. The line-graph L(G) of G has X as its set of points and two elements x, y of X are adjacent in L(G) whenever the lines x and y of G have a common end-point. A walk in G is an alternating sequence v1, x1, v2, x2, \u2026, vn of points and lines, the first and last terms being points, such that xi is the line joining vi to vi+1 for i=1, \u2026, n-1."
                },
                {
                    "title": "Conditional Structure Generation through Graph Variational Generative Adversarial Nets",
                    "abstract": "Graph embedding has been intensively studied recently, due to the advance of various neural network models. Theoretical analyses and empirical studies have pushed forward the translation of discrete graph structures into distributed representation vectors, but seldom considered the reverse direction, i.e., generation of graphs from given related context spaces. Particularly, since graphs often become more meaningful when associated with semantic contexts (e.g., social networks of certain communities, gene networks of certain diseases), the ability to infer graph structures according to given semantic conditions could be of great value. While existing graph generative models only consider graph structures without semantic contexts, we formulate the novel problem of conditional structure generation, and propose a novel unified model of graph variational generative adversarial nets (CondGen) to handle the intrinsic challenges of flexible context-structure conditioning and permutation-invariant generation. Extensive experiments on two deliberately created benchmark datasets of real-world context-enriched networks demonstrate the supreme effectiveness and generalizability of CondGen."
                },
                {
                    "title": "Handbook of Markov decision processes : methods and applications",
                    "abstract": "1. Introduction E.A. Feinberg, A. Shwartz. Part I: Finite State and Action Models. 2. Finite State and Action MDPs L. Kallenberg. 3. Bias Optimality M.E. Lewis, M.L. Puterman. 4. Singular Perturbations of Markov Chains and Decision Processes K.E. Avrachenkov, et al. Part II: Infinite State Models. 5. Average Reward Optimization Theory for Denumerable State Spaces L.I. Sennott. 6. Total Reward Criteria E.A. Feinberg. 7. Mixed Criteria E.A. Feinberg, A. Shwartz. 8. Blackwell Optimality A. Hordijk, A.A. Yushkevich. 9. The Poisson Equation for Countable Markov Chains: Probabilistic Methods and Interpretations A.M. Makowski, A. Shwartz. 10. Stability, Performance Evaluation, and Optimization S.P. Meyn. 11. Convex Analytic Methods in Markov Decision Processes V.S. Borkar. 12. The Linear Programming Approach O. Hernandez-Lerma, J.B. Lasserre. 13. Invariant Gambling Problems and Markov Decision Processes L.E. Dubins, et al. Part III: Applications. 14. Neuro-Dynamic Programming: Overview and Recent Trends B. Van Roy. 15. Markov Decision Processes in Finance and Dynamic Options M. Schal. 16. Applications of Markov Decision Processes in Communication Networks E. Altman. 17. Water Reservoir Applications of Markov Decision Processes B.F. Lamond, A. Boukhtouta. Index."
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                },
                {
                    "title": "On the evolution of random graphs",
                    "abstract": "(n) k edges have equal probabilities to be chosen as the next one . We shall 2 study the \"evolution\" of such a random graph if N is increased . In this investigation we endeavour to find what is the \"typical\" structure at a given stage of evolution (i . e . if N is equal, or asymptotically equal, to a given function N(n) of n) . By a \"typical\" structure we mean such a structure the probability of which tends to 1 if n -* + when N = N(n) . If A is such a property that lim Pn,N,(n ) ( A) = 1, we shall say that \u201ealmost all\" graphs Gn,N(n) n--possess this property ."
                },
                {
                    "title": "Ieee Transactions on Systems, Man and Cybernetics\u2014part C: Applications and Reviews 1 a Survey of Actor-critic Reinforcement Learning: Standard and Natural Policy Gradients",
                    "abstract": "\u2014Policy gradient based actor-critic algorithms are amongst the most popular algorithms in the reinforcement learning framework. Their advantage of being able to search for optimal policies using low-variance gradient estimates has made them useful in several real-life applications, such as robotics, power control and finance. Although general surveys on reinforcement learning techniques already exist, no survey is specifically dedicated to actor-critic algorithms in particular. This paper therefore describes the state of the art of actor-critic algorithms, with a focus on methods that can work in an online setting and use function approximation in order to deal with continuous state and action spaces. After starting with a discussion on the concepts of reinforcement learning and the origins of actor-critic algorithms, this paper describes the workings of the natural gradient, which has made its way into many actor-critic algorithms in the past few years. A review of several standard and natural actor-critic algorithms follows and the paper concludes with an overview of application areas and a discussion on open issues."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CE"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively optimize graph diffusion probabilistic models (DPMs) for arbitrary reward signals in graph generation tasks?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of graph generation, as it directly impacts various applications such as drug design, material science, and social network analysis. By optimizing graph DPMs to meet specific objectives, we can enhance the quality and relevance of generated graphs, leading to more effective solutions in real-world scenarios. This research could pave the way for future studies to explore more complex reward structures and improve the efficiency of generative models, ultimately contributing to the development of more sophisticated AI systems capable of tackling intricate tasks.\n\n**[Question 3] - Why is it hard?**  \nThe primary challenge lies in the non-differentiable nature of many reward signals associated with graph representations, which complicates the application of traditional optimization algorithms. Naive approaches may fail because they do not account for the complexity of reward signals, which can arise from intricate physical simulations or other non-linear processes. Additionally, adapting existing reinforcement learning techniques to discrete-variable graph DPMs has proven difficult, as empirical evidence shows that direct extensions from continuous-variable models do not yield satisfactory results. Overcoming these technical and theoretical obstacles is essential for effective optimization.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on earlier graph generative models and has not fully leveraged the advancements made with graph DPMs. Existing methods, such as property predictors, struggle with the accuracy of reward signal guidance due to the complexity of these signals. Additionally, while some work has been done on optimizing continuous-variable DPMs, the unique challenges posed by discrete-variable graphs have not been adequately addressed. Our approach differs by introducing a novel policy gradient method specifically tailored for graph DPMs, which aims to overcome the limitations of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, graph diffusion policy optimization (GDPO), involves adapting the discrete diffusion process of graph DPMs into a Markov decision process (MDP) and formulating the learning problem as policy optimization. We will utilize a modified policy gradient method, termed the eager policy gradient, to enhance performance in graph generation tasks. The expected outcomes include significant improvements in sample efficiency and a marked reduction in generation-test discrepancies, with empirical results"
            }
        },
        "author_data": {
            "73d61dd0-e989-4037-baf8-f076325ccb6d": {
                "pk": "73d61dd0-e989-4037-baf8-f076325ccb6d",
                "name": "Yijing Liu",
                "collaborators": [
                    "Xiang Zhang",
                    "Renchang Dai",
                    "Guangyi Liu",
                    "Yunlong Zhao",
                    "Xiaoheng Deng",
                    "Ke Xu",
                    "Yingqiu Zhu",
                    "Hansheng Wang",
                    "Peng Wei",
                    "Zhiwei Wang"
                ],
                "domain": [
                    "Power Systems",
                    "Data Analysis",
                    "Machine Learning",
                    "Model Stealing"
                ],
                "publications": [
                    {
                        "title": "Using Terminal Circuit for Power System Electromagnetic Transient Simulation",
                        "abstract": "The modern power system is evolving with increasing penetration of power electronics introducing complicated electromagnetic phenomenon. Electromagnetic transient (EMT) simulation is essential to understand power system behavior under disturbance which however is one of the most sophisticated and time-consuming applications in power system. To improve the electromagnetic transient simulation efficiency while keeping the simulation accuracy, this paper proposes to model and simulate power system electromagnetic transients by very large-scale integrated circuit (VLSI) as a preliminary exploration to eventually represent power system by VLSI circuit chip avoiding numerical calculation. To proof the concept, a simple 5 bus system is modeled and simulated to verify the feasibility of the proposed approach."
                    },
                    {
                        "title": "CluBear: A Subsampling Package for Interactive Statistical Analysis with Massive Data on A Single Machine",
                        "abstract": "This article introduces CluBear, a Python-based open-source package for interactive massive data analysis. The key feature of CluBear is that it enables users to conduct convenient and interactive statistical analysis of massive data with only a traditional single-computer system. Thus, CluBear provides a cost-effective solution when mining large-scale datasets. In addition, the CluBear package integrates many commonly used statistical and graphical tools, which are useful for most commonly encountered data analysis tasks."
                    },
                    {
                        "title": "Power System Transient Modeling and Simulation using Integrated Circuit",
                        "abstract": "Transient stability analysis (TSA) plays an important role in power system analysis to investigate the stability of power system. Traditionally, transient stability analysis methods have been developed using time domain simulation by means of numerical integration method. In this paper, a new approach is proposed to model power systems as an integrated circuit and simulate the power system dynamic behavior by integrated circuit simulator. The proposed method modeled power grid, generator, governor, and exciter with high fidelity. The power system dynamic simulation accuracy and efficiency of the proposed approach are verified and demonstrated by case study on an IEEE standard system."
                    },
                    {
                        "title": "Integration of Renewable Generators in Synthetic Electric Grids for Dynamic Analysis",
                        "abstract": "This paper presents a method to better integrate dynamic models for renewable resources into synthetic electric grids. An automated dynamic models assignment process is proposed for wind and solar generators. A realistic composition ratio for different types of wind turbine generators (WTG) is assigned to each wind generator. Statistics summarized from real electric grid data form the bases in assigning proper models with reasonable parameters to each WTG. A similar process is used to assign appropriate models and parameters to each photovoltaic (PV) generator. Multiple control strategies of the renewable resources are considered and tested in case studies. Two large-scale synthetic network test cases are used as examples of modeling the dynamics of renewable generators. Several transient stability metrics are adopted to assess the stability level after being subject to N-1 contingency event. Representative contingency events are given to demonstrate the performance of the synthetic renewable generator models."
                    },
                    {
                        "title": "MetaDD: Boosting Dataset Distillation with Neural Network Architecture-Invariant Generalization",
                        "abstract": "Dataset distillation (DD) entails creating a refined, compact distilled dataset from a large-scale dataset to facilitate efficient training. A significant challenge in DD is the dependency between the distilled dataset and the neural network (NN) architecture used. Training a different NN architecture with a distilled dataset distilled using a specific architecture often results in diminished trainning performance for other architectures. This paper introduces MetaDD, designed to enhance the generalizability of DD across various NN architectures. Specifically, MetaDD partitions distilled data into meta features (i.e., the data's common characteristics that remain consistent across different NN architectures) and heterogeneous features (i.e., the data's unique feature to each NN architecture). Then, MetaDD employs an architecture-invariant loss function for multi-architecture feature alignment, which increases meta features and reduces heterogeneous features in distilled data. As a low-memory consumption component, MetaDD can be seamlessly integrated into any DD methodology. Experimental results demonstrate that MetaDD significantly improves performance across various DD methods. On the Distilled Tiny-Imagenet with Sre2L (50 IPC), MetaDD achieves cross-architecture NN accuracy of up to 30.1\\%, surpassing the second-best method (GLaD) by 1.7\\%."
                    },
                    {
                        "title": "Fully Exploiting Every Real Sample: SuperPixel Sample Gradient Model Stealing",
                        "abstract": "Model stealing (MS) involves querying and observing the output of a machine learning model to steal its capabilities. The quality of queried data is crucial, yet obtaining a large amount of real data for MS is often challenging. Recent works have reduced reliance on real data by using generative models. However, when high-dimensional query data is required, these methods are impractical due to the high costs of querying and the risk of model collapse. In this work, we propose using sample gradients (SG) to enhance the utility of each real sample, as SG provides crucial guidance on the decision boundaries of the victim model. However, utilizing SG in the model stealing scenario faces two challenges: 1. Pixel-level gradient estimation requires extensive query volume and is susceptible to defenses. 2. The estimation of sample gradients has a significant variance. This paper proposes Superpixel Sample Gradient stealing (SPSG) for model stealing under the constraint of limited real samples. With the basic idea of imitating the victim model's low-variance patch-level gradients instead of pixel-level gradients, SPSG achieves efficient sample gradient estimation through two steps. First, we perform patch-wise perturbations on query images to estimate the average gradient in different regions of the image. Then, we filter the gradients through a threshold strategy to reduce variance. Exhaustive experiments demonstrate that, with the same number of real samples, SPSG achieves accuracy, agreements, and adversarial success rate significantly surpassing the current state-of-the-art MS methods. Codes are available at https://github.com/zyl123456aB/SPSG_attack."
                    }
                ]
            },
            "fba14fa8-82b6-4dbb-baa6-72edd91abf47": {
                "pk": "fba14fa8-82b6-4dbb-baa6-72edd91abf47",
                "name": "Chao Du",
                "collaborators": [
                    "Jun Zhu",
                    "Tianyu Pang",
                    "S. C. Kou",
                    "Jingdong Wang",
                    "Bo Zhang",
                    "Wing Hong Wong",
                    "Yinpeng Dong"
                ],
                "domain": [
                    "Machine Learning",
                    "Bayesian Inference",
                    "Stochastic Modeling",
                    "Adversarial Robustness"
                ],
                "publications": [
                    {
                        "title": "Correlation analysis of enzymatic reaction of a single protein molecule",
                        "abstract": "New advances in nano sciences open the door for scientists to study biological processes on a microscopic molecule-by-molecule basis. Recent single-molecule biophysical experiments on enzyme systems, in particular, reveal that enzyme molecules behave fundamentally differently from what classical model predicts. A stochastic network model was previously proposed to explain the experimental discovery. This paper conducts detailed theoretical and data analyses of the stochastic network model, focusing on the correlation structure of the successive reaction times of a single enzyme molecule. We investigate the correlation of experimental fluorescence intensity and the correlation of enzymatic reaction times, and examine the role of substrate concentration in enzymatic reactions. Our study shows that the stochastic network model is capable of explaining the experimental data in depth."
                    },
                    {
                        "title": "Inner Product Similarity Search using Compositional Codes",
                        "abstract": "This paper addresses the nearest neighbor search problem under inner product similarity and introduces a compact code-based approach. The idea is to approximate a vector using the composition of several elements selected from a source dictionary and to represent this vector by a short code composed of the indices of the selected elements. The inner product between a query vector and a database vector is efficiently estimated from the query vector and the short code of the database vector. We show the superior performance of the proposed group $M$-selection algorithm that selects $M$ elements from $M$ source dictionaries for vector approximation in terms of search accuracy and efficiency for compact codes of the same length via theoretical and empirical analysis. Experimental results on large-scale datasets ($1M$ and $1B$ SIFT features, $1M$ linear models and Netflix) demonstrate the superiority of the proposed approach."
                    },
                    {
                        "title": "Learning Deep Generative Models with Doubly Stochastic MCMC",
                        "abstract": "We present doubly stochastic gradient MCMC, a simple and generic method for (approximate) Bayesian inference of deep generative models (DGMs) in a collapsed continuous parameter space. At each MCMC sampling step, the algorithm randomly draws a mini-batch of data samples to estimate the gradient of log-posterior and further estimates the intractable expectation over hidden variables via a neural adaptive importance sampler, where the proposal distribution is parameterized by a deep neural network and learnt jointly. We demonstrate the effectiveness on learning various DGMs in a wide range of tasks, including density estimation, data generation and missing data imputation. Our method outperforms many state-of-the-art competitors."
                    },
                    {
                        "title": "Stochastic Modeling and Statistical Inference of Intrinsic Noise in Gene Regulation System via Chemical Master Equation",
                        "abstract": "Intrinsic noise, the stochastic cell-to-cell fluctuations in mRNAs and proteins, has been observed and proved to play important roles in cellular systems. Due to the recent development in single-cell-level measurement technology, the studies on intrinsic noise are becoming increasingly popular among scholars. The chemical master equation (CME) has been used to model the evolutions of complex chemical and biological systems since 1940, and are often served as the standard tool for modeling intrinsic noise in gene regulation system. A CME-based model can capture the discrete, stochastic, and dynamical nature of gene regulation system, and may offer casual and physical explanation of the observed data at single-cell level. Nonetheless, the complexity of CME also pose serious challenge for researchers in proposing practical modeling and inference frameworks. In this article, we will review the existing works on the modelings and inference of intrinsic noise in gene regulation system within the framework of CME model. We will explore the principles in constructing a CME model for studying gene regulation system and discuss the popular approximations of CME. Then we will study the simulation simulation methods as well as the analytical and numerical approaches that can be used to obtain solution to a CME model. Finally we will summary the exiting statistical methods that can be used to infer the unknown parameters or structures in CME model using single-cell-level gene expression data."
                    },
                    {
                        "title": "Towards Robust Detection of Adversarial Examples",
                        "abstract": "Although the recent progress is substantial, deep learning methods can be vulnerable to the maliciously generated adversarial examples. In this paper, we present a novel training procedure and a thresholding test strategy, towards robust detection of adversarial examples. In training, we propose to minimize the reverse cross-entropy (RCE), which encourages a deep network to learn latent representations that better distinguish adversarial examples from normal ones. In testing, we propose to use a thresholding strategy as the detector to filter out adversarial examples for reliable predictions. Our method is simple to implement using standard algorithms, with little extra training cost compared to the common cross-entropy minimization. We apply our method to defend various attacking methods on the widely used MNIST and CIFAR-10 datasets, and achieve significant improvements on robust predictions under all the threat models in the adversarial setting."
                    },
                    {
                        "title": "Max-Mahalanobis Linear Discriminant Analysis Networks",
                        "abstract": "A deep neural network (DNN) consists of a nonlinear transformation from an input to a feature representation, followed by a common softmax linear classifier. Though many efforts have been devoted to designing a proper architecture for nonlinear transformation, little investigation has been done on the classifier part. In this paper, we show that a properly designed classifier can improve robustness to adversarial attacks and lead to better prediction results. Specifically, we define a Max-Mahalanobis distribution (MMD) and theoretically show that if the input distributes as a MMD, the linear discriminant analysis (LDA) classifier will have the best robustness to adversarial examples. We further propose a novel Max-Mahalanobis linear discriminant analysis (MM-LDA) network, which explicitly maps a complicated data distribution in the input space to a MMD in the latent feature space and then applies LDA to make predictions. Our results demonstrate that the MM-LDA networks are significantly more robust to adversarial attacks, and have better performance in class-biased classification."
                    }
                ]
            },
            "0571608f-ef1d-478b-ab68-adccf4627a40": {
                "pk": "0571608f-ef1d-478b-ab68-adccf4627a40",
                "name": "Tianyu Pang",
                "collaborators": [
                    "Jun Zhu",
                    "Yinpeng Dong",
                    "Hang Su",
                    "Chao Du",
                    "Shuicheng Yan",
                    "Min Lin",
                    "Xiao Yang",
                    "Kun Xu",
                    "Zonghan Yang",
                    "Yang Liu"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Deep Learning",
                    "Robustness",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Towards Robust Detection of Adversarial Examples",
                        "abstract": "Although the recent progress is substantial, deep learning methods can be vulnerable to the maliciously generated adversarial examples. In this paper, we present a novel training procedure and a thresholding test strategy, towards robust detection of adversarial examples. In training, we propose to minimize the reverse cross-entropy (RCE), which encourages a deep network to learn latent representations that better distinguish adversarial examples from normal ones. In testing, we propose to use a thresholding strategy as the detector to filter out adversarial examples for reliable predictions. Our method is simple to implement using standard algorithms, with little extra training cost compared to the common cross-entropy minimization. We apply our method to defend various attacking methods on the widely used MNIST and CIFAR-10 datasets, and achieve significant improvements on robust predictions under all the threat models in the adversarial setting."
                    },
                    {
                        "title": "Max-Mahalanobis Linear Discriminant Analysis Networks",
                        "abstract": "A deep neural network (DNN) consists of a nonlinear transformation from an input to a feature representation, followed by a common softmax linear classifier. Though many efforts have been devoted to designing a proper architecture for nonlinear transformation, little investigation has been done on the classifier part. In this paper, we show that a properly designed classifier can improve robustness to adversarial attacks and lead to better prediction results. Specifically, we define a Max-Mahalanobis distribution (MMD) and theoretically show that if the input distributes as a MMD, the linear discriminant analysis (LDA) classifier will have the best robustness to adversarial examples. We further propose a novel Max-Mahalanobis linear discriminant analysis (MM-LDA) network, which explicitly maps a complicated data distribution in the input space to a MMD in the latent feature space and then applies LDA to make predictions. Our results demonstrate that the MM-LDA networks are significantly more robust to adversarial attacks, and have better performance in class-biased classification."
                    },
                    {
                        "title": "Mixup Inference: Better Exploiting Mixup to Defend Adversarial Attacks",
                        "abstract": "It has been widely recognized that adversarial examples can be easily crafted to fool deep networks, which mainly root from the locally non-linear behavior nearby input examples. Applying mixup in training provides an effective mechanism to improve generalization performance and model robustness against adversarial perturbations, which introduces the globally linear behavior in-between training examples. However, in previous work, the mixup-trained models only passively defend adversarial attacks in inference by directly classifying the inputs, where the induced global linearity is not well exploited. Namely, since the locality of the adversarial perturbations, it would be more efficient to actively break the locality via the globality of the model predictions. Inspired by simple geometric intuition, we develop an inference principle, named mixup inference (MI), for mixup-trained models. MI mixups the input with other random clean samples, which can shrink and transfer the equivalent perturbation if the input is adversarial. Our experiments on CIFAR-10 and CIFAR-100 demonstrate that MI can further improve the adversarial robustness for the models trained by mixup and its variants."
                    },
                    {
                        "title": "$O(N^2)$ Universal Antisymmetry in Fermionic Neural Networks",
                        "abstract": "Fermionic neural network (FermiNet) is a recently proposed wavefunction Ansatz, which is used in variational Monte Carlo (VMC) methods to solve the many-electron Schr\\\"{o}dinger equation. FermiNet proposes permutation-equivariant architectures, on which a Slater determinant is applied to induce antisymmetry. FermiNet is proved to have universal approximation capability with a single determinant, namely, it suffices to represent any antisymmetric function given sufficient parameters. However, the asymptotic computational bottleneck comes from the Slater determinant, which scales with $O(N^3)$ for $N$ electrons. In this paper, we substitute the Slater determinant with a pairwise antisymmetry construction, which is easy to implement and can reduce the computational cost to $O(N^2)$. We formally prove that the pairwise construction built upon permutation-equivariant architectures can universally represent any antisymmetric function. Besides, this universality can be achieved via continuous approximators when we aim to represent ground-state wavefunctions."
                    },
                    {
                        "title": "Evading Defenses to Transferable Adversarial Examples by Translation-Invariant Attacks",
                        "abstract": "Deep neural networks are vulnerable to adversarial examples, which can mislead classifiers by adding imperceptible perturbations. An intriguing property of adversarial examples is their good transferability, making black-box attacks feasible in real-world applications. Due to the threat of adversarial attacks, many methods have been proposed to improve the robustness. Several state-of-the-art defenses are shown to be robust against transferable adversarial examples. In this paper, we propose a translation-invariant attack method to generate more transferable adversarial examples against the defense models. By optimizing a perturbation over an ensemble of translated images, the generated adversarial example is less sensitive to the white-box model being attacked and has better transferability. To improve the efficiency of attacks, we further show that our method can be implemented by convolving the gradient at the untranslated image with a pre-defined kernel. Our method is generally applicable to any gradient-based attack method. Extensive experiments on the ImageNet dataset validate the effectiveness of the proposed method. Our best attack fools eight state-of-the-art defenses at an 82% success rate on average based only on the transferability, demonstrating the insecurity of the current defense techniques."
                    },
                    {
                        "title": "A Closer Look at the Adversarial Robustness of Deep Equilibrium Models",
                        "abstract": "Deep equilibrium models (DEQs) refrain from the traditional layer-stacking paradigm and turn to find the fixed point of a single layer. DEQs have achieved promising performance on different applications with featured memory efficiency. At the same time, the adversarial vulnerability of DEQs raises concerns. Several works propose to certify robustness for monotone DEQs. However, limited efforts are devoted to studying empirical robustness for general DEQs. To this end, we observe that an adversarially trained DEQ requires more forward steps to arrive at the equilibrium state, or even violates its fixed-point structure. Besides, the forward and backward tracks of DEQs are misaligned due to the black-box solvers. These facts cause gradient obfuscation when applying the ready-made attacks to evaluate or adversarially train DEQs. Given this, we develop approaches to estimate the intermediate gradients of DEQs and integrate them into the attacking pipelines. Our approaches facilitate fully white-box evaluations and lead to effective adversarial defense for DEQs. Extensive experiments on CIFAR-10 validate the adversarial robustness of DEQs competitive with deep networks of similar sizes."
                    },
                    {
                        "title": "Improving Adversarial Robustness of DEQs with Explicit Regulations Along the Neural Dynamics",
                        "abstract": "Deep equilibrium (DEQ) models replace the multiple-layer stacking of conventional deep networks with a fixed-point iteration of a single-layer transformation. Having been demonstrated to be competitive in a variety of real-world scenarios, the adversarial robustness of general DEQs becomes increasingly crucial for their reliable deployment. Existing works improve the robustness of general DEQ models with the widely-used adversarial training (AT) framework, but they fail to exploit the structural uniquenesses of DEQ models. To this end, we interpret DEQs through the lens of neural dynamics and find that AT under-regulates intermediate states. Besides, the intermediate states typically provide predictions with a high prediction entropy. Informed by the correlation between the entropy of dynamical systems and their stability properties, we propose reducing prediction entropy by progressively updating inputs along the neural dynamics. During AT, we also utilize random intermediate states to compute the loss function. Our methods regulate the neural dynamics of DEQ models in this manner. Extensive experiments demonstrate that our methods substantially increase the robustness of DEQ models and even outperform the strong deep network baselines."
                    },
                    {
                        "title": "Improving Adversarial Robustness via Promoting Ensemble Diversity",
                        "abstract": "Though deep neural networks have achieved significant progress on various tasks, often enhanced by model ensemble, existing high-performance models can be vulnerable to adversarial attacks. Many efforts have been devoted to enhancing the robustness of individual networks and then constructing a straightforward ensemble, e.g., by directly averaging the outputs, which ignores the interaction among networks. This paper presents a new method that explores the interaction among individual networks to improve robustness for ensemble models. Technically, we define a new notion of ensemble diversity in the adversarial setting as the diversity among non-maximal predictions of individual members, and present an adaptive diversity promoting (ADP) regularizer to encourage the diversity, which leads to globally better robustness for the ensemble by making adversarial examples difficult to transfer among individual members. Our method is computationally efficient and compatible with the defense methods acting on individual networks. Empirical results on various datasets verify that our method can improve adversarial robustness while maintaining state-of-the-art accuracy on normal examples."
                    },
                    {
                        "title": "Bag of Tricks for Adversarial Training",
                        "abstract": "Adversarial training (AT) is one of the most effective strategies for promoting model robustness. However, recent benchmarks show that most of the proposed improvements on AT are less effective than simply early stopping the training procedure. This counter-intuitive fact motivates us to investigate the implementation details of tens of AT methods. Surprisingly, we find that the basic settings (e.g., weight decay, training schedule, etc.) used in these methods are highly inconsistent. In this work, we provide comprehensive evaluations on CIFAR-10, focusing on the effects of mostly overlooked training tricks and hyperparameters for adversarially trained models. Our empirical observations suggest that adversarial robustness is much more sensitive to some basic training settings than we thought. For example, a slightly different value of weight decay can reduce the model robust accuracy by more than 7%, which is probable to override the potential promotion induced by the proposed methods. We conclude a baseline training setting and re-implement previous defenses to achieve new state-of-the-art results. These facts also appeal to more concerns on the overlooked confounders when benchmarking defenses."
                    },
                    {
                        "title": "Accumulative Poisoning Attacks on Real-time Data",
                        "abstract": "Collecting training data from untrusted sources exposes machine learning services to poisoning adversaries, who maliciously manipulate training data to degrade the model accuracy. When trained on offline datasets, poisoning adversaries have to inject the poisoned data in advance before training, and the order of feeding these poisoned batches into the model is stochastic. In contrast, practical systems are more usually trained/fine-tuned on sequentially captured real-time data, in which case poisoning adversaries could dynamically poison each data batch according to the current model state. In this paper, we focus on the real-time settings and propose a new attacking strategy, which affiliates an accumulative phase with poisoning attacks to secretly (i.e., without affecting accuracy) magnify the destructive effect of a (poisoned) trigger batch. By mimicking online learning and federated learning on MNIST and CIFAR-10, we show that model accuracy significantly drops by a single update step on the trigger batch after the accumulative phase. Our work validates that a well-designed but straightforward attacking strategy can dramatically amplify the poisoning effects, with no need to explore complex techniques."
                    },
                    {
                        "title": "Robustness and Accuracy Could Be Reconcilable by (Proper) Definition",
                        "abstract": "The trade-off between robustness and accuracy has been widely studied in the adversarial literature. Although still controversial, the prevailing view is that this trade-off is inherent, either empirically or theoretically. Thus, we dig for the origin of this trade-off in adversarial training and find that it may stem from the improperly defined robust error, which imposes an inductive bias of local invariance -- an overcorrection towards smoothness. Given this, we advocate employing local equivariance to describe the ideal behavior of a robust model, leading to a self-consistent robust error named SCORE. By definition, SCORE facilitates the reconciliation between robustness and accuracy, while still handling the worst-case uncertainty via robust optimization. By simply substituting KL divergence with variants of distance metrics, SCORE can be efficiently minimized. Empirically, our models achieve top-rank performance on RobustBench under AutoAttack. Besides, SCORE provides instructive insights for explaining the overfitting phenomenon and semantic input gradients observed on robust models. Code is available at https://github.com/P2333/SCORE."
                    },
                    {
                        "title": "Intriguing Properties of Data Attribution on Diffusion Models",
                        "abstract": "Data attribution seeks to trace model outputs back to training data. With the recent development of diffusion models, data attribution has become a desired module to properly assign valuations for high-quality or copyrighted training samples, ensuring that data contributors are fairly compensated or credited. Several theoretically motivated methods have been proposed to implement data attribution, in an effort to improve the trade-off between computational scalability and effectiveness. In this work, we conduct extensive experiments and ablation studies on attributing diffusion models, specifically focusing on DDPMs trained on CIFAR-10 and CelebA, as well as a Stable Diffusion model LoRA-finetuned on ArtBench. Intriguingly, we report counter-intuitive observations that theoretically unjustified design choices for attribution empirically outperform previous baselines by a large margin, in terms of both linear datamodeling score and counterfactual evaluation. Our work presents a significantly more efficient approach for attributing diffusion models, while the unexpected findings suggest that at least in non-convex settings, constructions guided by theoretical assumptions may lead to inferior attribution performance. The code is available at https://github.com/sail-sg/D-TRAK."
                    },
                    {
                        "title": "Adversarial Distributional Training for Robust Deep Learning",
                        "abstract": "Adversarial training (AT) is among the most effective techniques to improve model robustness by augmenting training data with adversarial examples. However, most existing AT methods adopt a specific attack to craft adversarial examples, leading to the unreliable robustness against other unseen attacks. Besides, a single attack algorithm could be insufficient to explore the space of perturbations. In this paper, we introduce adversarial distributional training (ADT), a novel framework for learning robust models. ADT is formulated as a minimax optimization problem, where the inner maximization aims to learn an adversarial distribution to characterize the potential adversarial examples around a natural one under an entropic regularizer, and the outer minimization aims to train robust models by minimizing the expected loss over the worst-case adversarial distributions. Through a theoretical analysis, we develop a general algorithm for solving ADT, and present three approaches for parameterizing the adversarial distributions, ranging from the typical Gaussian distributions to the flexible implicit ones. Empirical results on several benchmarks validate the effectiveness of ADT compared with the state-of-the-art AT methods."
                    },
                    {
                        "title": "Boosting Transferability of Targeted Adversarial Examples via Hierarchical Generative Networks",
                        "abstract": "Transfer-based adversarial attacks can evaluate model robustness in the black-box setting. Several methods have demonstrated impressive untargeted transferability, however, it is still challenging to efficiently produce targeted transferability. To this end, we develop a simple yet effective framework to craft targeted transfer-based adversarial examples, applying a hierarchical generative network. In particular, we contribute to amortized designs that well adapt to multi-class targeted attacks. Extensive experiments on ImageNet show that our method improves the success rates of targeted black-box attacks by a significant margin over the existing methods -- it reaches an average success rate of 29.1\\% against six diverse models based only on one substitute white-box model, which significantly outperforms the state-of-the-art gradient-based attack methods. Moreover, the proposed method is also more efficient beyond an order of magnitude than gradient-based methods."
                    },
                    {
                        "title": "Nonparametric Generative Modeling with Conditional Sliced-Wasserstein Flows",
                        "abstract": "Sliced-Wasserstein Flow (SWF) is a promising approach to nonparametric generative modeling but has not been widely adopted due to its suboptimal generative quality and lack of conditional modeling capabilities. In this work, we make two major contributions to bridging this gap. First, based on a pleasant observation that (under certain conditions) the SWF of joint distributions coincides with those of conditional distributions, we propose Conditional Sliced-Wasserstein Flow (CSWF), a simple yet effective extension of SWF that enables nonparametric conditional modeling. Second, we introduce appropriate inductive biases of images into SWF with two techniques inspired by local connectivity and multiscale representation in vision research, which greatly improve the efficiency and quality of modeling images. With all the improvements, we achieve generative performance comparable with many deep parametric generative models on both conditional and unconditional tasks in a purely nonparametric fashion, demonstrating its great potential."
                    },
                    {
                        "title": "Improving Black-box Adversarial Attacks with a Transfer-based Prior",
                        "abstract": "We consider the black-box adversarial setting, where the adversary has to generate adversarial perturbations without access to the target models to compute gradients. Previous methods tried to approximate the gradient either by using a transfer gradient of a surrogate white-box model, or based on the query feedback. However, these methods often suffer from low attack success rates or poor query efficiency since it is non-trivial to estimate the gradient in a high-dimensional space with limited information. To address these problems, we propose a prior-guided random gradient-free (P-RGF) method to improve black-box adversarial attacks, which takes the advantage of a transfer-based prior and the query information simultaneously. The transfer-based prior given by the gradient of a surrogate model is appropriately integrated into our algorithm by an optimal coefficient derived by a theoretical analysis. Extensive experiments demonstrate that our method requires much fewer queries to attack black-box models with higher success rates compared with the alternative state-of-the-art methods."
                    },
                    {
                        "title": "Query-Efficient Black-box Adversarial Attacks Guided by a Transfer-based Prior",
                        "abstract": "Adversarial attacks have been extensively studied in recent years since they can identify the vulnerability of deep learning models before deployed. In this paper, we consider the black-box adversarial setting, where the adversary needs to craft adversarial examples without access to the gradients of a target model. Previous methods attempted to approximate the true gradient either by using the transfer gradient of a surrogate white-box model or based on the feedback of model queries. However, the existing methods inevitably suffer from low attack success rates or poor query efficiency since it is difficult to estimate the gradient in a high-dimensional input space with limited information. To address these problems and improve black-box attacks, we propose two prior-guided random gradient-free (PRGF) algorithms based on biased sampling and gradient averaging, respectively. Our methods can take the advantage of a transfer-based prior given by the gradient of a surrogate model and the query information simultaneously. Through theoretical analyses, the transfer-based prior is appropriately integrated with model queries by an optimal coefficient in each method. Extensive experiments demonstrate that, in comparison with the alternative state-of-the-arts, both of our methods require much fewer queries to attack black-box models with higher success rates."
                    },
                    {
                        "title": "Efficient Diffusion Policies for Offline Reinforcement Learning",
                        "abstract": "Offline reinforcement learning (RL) aims to learn optimal policies from offline datasets, where the parameterization of policies is crucial but often overlooked. Recently, Diffsuion-QL significantly boosts the performance of offline RL by representing a policy with a diffusion model, whose success relies on a parametrized Markov Chain with hundreds of steps for sampling. However, Diffusion-QL suffers from two critical limitations. 1) It is computationally inefficient to forward and backward through the whole Markov chain during training. 2) It is incompatible with maximum likelihood-based RL algorithms (e.g., policy gradient methods) as the likelihood of diffusion models is intractable. Therefore, we propose efficient diffusion policy (EDP) to overcome these two challenges. EDP approximately constructs actions from corrupted ones at training to avoid running the sampling chain. We conduct extensive experiments on the D4RL benchmark. The results show that EDP can reduce the diffusion policy training time from 5 days to 5 hours on gym-locomotion tasks. Moreover, we show that EDP is compatible with various offline RL algorithms (TD3, CRR, and IQL) and achieves new state-of-the-art on D4RL by large margins over previous methods. Our code is available at https://github.com/sail-sg/edp."
                    },
                    {
                        "title": "Crafting Heavy-Tails in Weight Matrix Spectrum without Gradient Noise",
                        "abstract": "Training strategies for modern deep neural networks (NNs) tend to induce a heavy-tailed (HT) empirical spectral density (ESD) in the layer weights. While previous efforts have shown that the HT phenomenon correlates with good generalization in large NNs, a theoretical explanation of its occurrence is still lacking. Especially, understanding the conditions which lead to this phenomenon can shed light on the interplay between generalization and weight spectra. Our work aims to bridge this gap by presenting a simple, rich setting to model the emergence of HT ESD. In particular, we present a theory-informed analysis for 'crafting' heavy tails in the ESD of two-layer NNs without any gradient noise. This is the first work to analyze a noise-free setting and incorporate optimizer (GD/Adam) dependent (large) learning rates into the HT ESD analysis. Our results highlight the role of learning rates on the Bulk+Spike and HT shape of the ESDs in the early phase of training, which can facilitate generalization in the two-layer NN. These observations shed light on the behavior of large-scale NNs, albeit in a much simpler setting. Last but not least, we present a novel perspective on the ESD evolution dynamics by analyzing the singular vectors of weight matrices and optimizer updates."
                    }
                ]
            },
            "37626943-7e74-48b2-b905-5271e9043be7": {
                "pk": "37626943-7e74-48b2-b905-5271e9043be7",
                "name": "Chongxuan Li",
                "collaborators": [
                    "Jun Zhu",
                    "Bo Zhang",
                    "Kun Xu",
                    "Guoqiang Wu",
                    "Fan Bao",
                    "Tsung Wei Tsai",
                    "Chao Du",
                    "Jiacheng Sun",
                    "Yilong Yin",
                    "Yidong Ouyang"
                ],
                "domain": [
                    "Deep Generative Models",
                    "Machine Learning",
                    "Adversarial Learning",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "Learning to Generate with Memory",
                        "abstract": "Memory units have been widely used to enrich the capabilities of deep networks on capturing long-term dependencies in reasoning and prediction tasks, but little investigation exists on deep generative models (DGMs) which are good at inferring high-level invariant representations from unlabeled data. This paper presents a deep generative model with a possibly large external memory and an attention mechanism to capture the local detail information that is often lost in the bottom-up abstraction process in representation learning. By adopting a smooth attention model, the whole network is trained end-to-end by optimizing a variational bound of data likelihood via auto-encoding variational Bayesian methods, where an asymmetric recognition network is learnt jointly to infer high-level invariant representations. The asymmetric architecture can reduce the competition between bottom-up invariant feature extraction and top-down generation of instance details. Our experiments on several datasets demonstrate that memory can significantly boost the performance of DGMs and even achieve state-of-the-art results on various tasks, including density estimation, image generation, and missing value imputation."
                    },
                    {
                        "title": "Rethinking and Reweighting the Univariate Losses for Multi-Label Ranking: Consistency and Generalization",
                        "abstract": "(Partial) ranking loss is a commonly used evaluation measure for multi-label classification, which is usually optimized with convex surrogates for computational efficiency. Prior theoretical work on multi-label ranking mainly focuses on (Fisher) consistency analyses. However, there is a gap between existing theory and practice -- some pairwise losses can lead to promising performance but lack consistency, while some univariate losses are consistent but usually have no clear superiority in practice. In this paper, we attempt to fill this gap through a systematic study from two complementary perspectives of consistency and generalization error bounds of learning algorithms. Our results show that learning algorithms with the consistent univariate loss have an error bound of $O(c)$ ($c$ is the number of labels), while algorithms with the inconsistent pairwise loss depend on $O(\\sqrt{c})$ as shown in prior work. This explains that the latter can achieve better performance than the former in practice. Moreover, we present an inconsistent reweighted univariate loss-based learning algorithm that enjoys an error bound of $O(\\sqrt{c})$ for promising performance as well as the computational efficiency of univariate losses. Finally, experimental results validate our theoretical analyses."
                    },
                    {
                        "title": "Max-Margin Deep Generative Models for (Semi-)Supervised Learning",
                        "abstract": "Deep generative models (DGMs) are effective on learning multilayered representations of complex data and performing inference of input data by exploring the generative ability. However, it is relatively insufficient to empower the discriminative ability of DGMs on making accurate predictions. This paper presents max-margin deep generative models (mmDGMs) and a class-conditional variant (mmDCGMs), which explore the strongly discriminative principle of max-margin learning to improve the predictive performance of DGMs in both supervised and semi-supervised learning, while retaining the generative capability. In semi-supervised learning, we use the predictions of a max-margin classifier as the missing labels instead of performing full posterior inference for efficiency; we also introduce additional max-margin and label-balance regularization terms of unlabeled data for effectiveness. We develop an efficient doubly stochastic subgradient algorithm for the piecewise linear objectives in different settings. Empirical results on various datasets demonstrate that: (1) max-margin learning can significantly improve the prediction performance of DGMs and meanwhile retain the generative ability; (2) in supervised learning, mmDGMs are competitive to the best fully discriminative networks when employing convolutional neural networks as the generative and recognition models; and (3) in semi-supervised learning, mmDCGMs can perform efficient inference and achieve state-of-the-art classification results on several benchmarks."
                    },
                    {
                        "title": "Multi-objects Generation with Amortized Structural Regularization",
                        "abstract": "Deep generative models (DGMs) have shown promise in image generation. However, most of the existing work learn the model by simply optimizing a divergence between the marginal distributions of the model and the data, and often fail to capture the rich structures and relations in multi-object images. Human knowledge is a critical element to the success of DGMs to infer these structures. In this paper, we propose the amortized structural regularization (ASR) framework, which adopts the posterior regularization (PR) to embed human knowledge into DGMs via a set of structural constraints. We derive a lower bound of the regularized log-likelihood, which can be jointly optimized with respect to the generative model and recognition model efficiently. Empirical results show that ASR significantly outperforms the DGM baselines in terms of inference accuracy and sample quality."
                    },
                    {
                        "title": "Countering Noisy Labels By Learning From Auxiliary Clean Labels",
                        "abstract": "We consider the learning from noisy labels (NL) problem which emerges in many real-world applications. In addition to the widely-studied synthetic noise in the NL literature, we also consider the pseudo labels in semi-supervised learning (Semi-SL) as a special case of NL. For both types of noise, we argue that the generalization performance of existing methods is highly coupled with the quality of noisy labels. Therefore, we counter the problem from a novel and unified perspective: learning from the auxiliary clean labels. Specifically, we propose the Rotational-Decoupling Consistency Regularization (RDCR) framework that integrates the consistency-based methods with the self-supervised rotation task to learn noise-tolerant representations. The experiments show that RDCR achieves comparable or superior performance than the state-of-the-art methods under small noise, while outperforms the existing methods significantly when there is large noise."
                    },
                    {
                        "title": "Understanding and Stabilizing GANs' Training Dynamics with Control Theory",
                        "abstract": "Generative adversarial networks (GANs) are effective in generating realistic images but the training is often unstable. There are existing efforts that model the training dynamics of GANs in the parameter space but the analysis cannot directly motivate practically effective stabilizing methods. To this end, we present a conceptually novel perspective from control theory to directly model the dynamics of GANs in the function space and provide simple yet effective methods to stabilize GANs' training. We first analyze the training dynamic of a prototypical Dirac GAN and adopt the widely-used closed-loop control (CLC) to improve its stability. We then extend CLC to stabilize the training dynamic of normal GANs, where CLC is implemented as a squared $L2$ regularizer on the output of the discriminator. Empirical results show that our method can effectively stabilize the training and obtain state-of-the-art performance on data generation tasks."
                    },
                    {
                        "title": "MiCE: Mixture of Contrastive Experts for Unsupervised Image Clustering",
                        "abstract": "We present Mixture of Contrastive Experts (MiCE), a unified probabilistic clustering framework that simultaneously exploits the discriminative representations learned by contrastive learning and the semantic structures captured by a latent mixture model. Motivated by the mixture of experts, MiCE employs a gating function to partition an unlabeled dataset into subsets according to the latent semantics and multiple experts to discriminate distinct subsets of instances assigned to them in a contrastive learning manner. To solve the nontrivial inference and learning problems caused by the latent variables, we further develop a scalable variant of the Expectation-Maximization (EM) algorithm for MiCE and provide proof of the convergence. Empirically, we evaluate the clustering performance of MiCE on four widely adopted natural image datasets. MiCE achieves significantly better results than various previous methods and a strong contrastive learning baseline."
                    },
                    {
                        "title": "Analytic-DPM: an Analytic Estimate of the Optimal Reverse Variance in Diffusion Probabilistic Models",
                        "abstract": "Diffusion probabilistic models (DPMs) represent a class of powerful generative models. Despite their success, the inference of DPMs is expensive since it generally needs to iterate over thousands of timesteps. A key problem in the inference is to estimate the variance in each timestep of the reverse process. In this work, we present a surprising result that both the optimal reverse variance and the corresponding optimal KL divergence of a DPM have analytic forms w.r.t. its score function. Building upon it, we propose Analytic-DPM, a training-free inference framework that estimates the analytic forms of the variance and KL divergence using the Monte Carlo method and a pretrained score-based model. Further, to correct the potential bias caused by the score-based model, we derive both lower and upper bounds of the optimal variance and clip the estimate for a better result. Empirically, our analytic-DPM improves the log-likelihood of various DPMs, produces high-quality samples, and meanwhile enjoys a 20x to 80x speed up."
                    },
                    {
                        "title": "Why Are Conditional Generative Models Better Than Unconditional Ones?",
                        "abstract": "Extensive empirical evidence demonstrates that conditional generative models are easier to train and perform better than unconditional ones by exploiting the labels of data. So do score-based diffusion models. In this paper, we analyze the phenomenon formally and identify that the key of conditional learning is to partition the data properly. Inspired by the analyses, we propose self-conditioned diffusion models (SCDM), which is trained conditioned on indices clustered by the k-means algorithm on the features extracted by a model pre-trained in a self-supervised manner. SCDM significantly improves the unconditional model across various datasets and achieves a record-breaking FID of 3.94 on ImageNet 64x64 without labels. Besides, SCDM achieves a slightly better FID than the corresponding conditional model on CIFAR10."
                    },
                    {
                        "title": "Towards Understanding Generalization of Macro-AUC in Multi-label Learning",
                        "abstract": "Macro-AUC is the arithmetic mean of the class-wise AUCs in multi-label learning and is commonly used in practice. However, its theoretical understanding is far lacking. Toward solving it, we characterize the generalization properties of various learning algorithms based on the corresponding surrogate losses w.r.t. Macro-AUC. We theoretically identify a critical factor of the dataset affecting the generalization bounds: \\emph{the label-wise class imbalance}. Our results on the imbalance-aware error bounds show that the widely-used univariate loss-based algorithm is more sensitive to the label-wise class imbalance than the proposed pairwise and reweighted loss-based ones, which probably implies its worse performance. Moreover, empirical results on various datasets corroborate our theory findings. To establish it, technically, we propose a new (and more general) McDiarmid-type concentration inequality, which may be of independent interest."
                    },
                    {
                        "title": "MissDiff: Training Diffusion Models on Tabular Data with Missing Values",
                        "abstract": "The diffusion model has shown remarkable performance in modeling data distributions and synthesizing data. However, the vanilla diffusion model requires complete or fully observed data for training. Incomplete data is a common issue in various real-world applications, including healthcare and finance, particularly when dealing with tabular datasets. This work presents a unified and principled diffusion-based framework for learning from data with missing values under various missing mechanisms. We first observe that the widely adopted \"impute-then-generate\" pipeline may lead to a biased learning objective. Then we propose to mask the regression loss of Denoising Score Matching in the training phase. We prove the proposed method is consistent in learning the score of data distributions, and the proposed training objective serves as an upper bound for the negative likelihood in certain cases. The proposed framework is evaluated on multiple tabular datasets using realistic and efficacious metrics and is demonstrated to outperform state-of-the-art diffusion model on tabular data with \"impute-then-generate\" pipeline by a large margin."
                    },
                    {
                        "title": "EGSDE: Unpaired Image-to-Image Translation via Energy-Guided Stochastic Differential Equations",
                        "abstract": "Score-based diffusion models (SBDMs) have achieved the SOTA FID results in unpaired image-to-image translation (I2I). However, we notice that existing methods totally ignore the training data in the source domain, leading to sub-optimal solutions for unpaired I2I. To this end, we propose energy-guided stochastic differential equations (EGSDE) that employs an energy function pretrained on both the source and target domains to guide the inference process of a pretrained SDE for realistic and faithful unpaired I2I. Building upon two feature extractors, we carefully design the energy function such that it encourages the transferred image to preserve the domain-independent features and discard domain-specific ones. Further, we provide an alternative explanation of the EGSDE as a product of experts, where each of the three experts (corresponding to the SDE and two feature extractors) solely contributes to faithfulness or realism. Empirically, we compare EGSDE to a large family of baselines on three widely-adopted unpaired I2I tasks under four metrics. EGSDE not only consistently outperforms existing SBDMs-based methods in almost all settings but also achieves the SOTA realism results without harming the faithful performance. Furthermore, EGSDE allows for flexible trade-offs between realism and faithfulness and we improve the realism results further (e.g., FID of 51.04 in Cat to Dog and FID of 50.43 in Wild to Dog on AFHQ) by tuning hyper-parameters. The code is available at https://github.com/ML-GSAI/EGSDE."
                    },
                    {
                        "title": "Toward Understanding Generative Data Augmentation",
                        "abstract": "Generative data augmentation, which scales datasets by obtaining fake labeled examples from a trained conditional generative model, boosts classification performance in various learning tasks including (semi-)supervised learning, few-shot learning, and adversarially robust learning. However, little work has theoretically investigated the effect of generative data augmentation. To fill this gap, we establish a general stability bound in this not independently and identically distributed (non-i.i.d.) setting, where the learned distribution is dependent on the original train set and generally not the same as the true distribution. Our theoretical result includes the divergence between the learned distribution and the true distribution. It shows that generative data augmentation can enjoy a faster learning rate when the order of divergence term is $o(\\max\\left( \\log(m)\\beta_m, 1 / \\sqrt{m})\\right)$, where $m$ is the train set size and $\\beta_m$ is the corresponding stability constant. We further specify the learning setup to the Gaussian mixture model and generative adversarial nets. We prove that in both cases, though generative data augmentation does not enjoy a faster learning rate, it can improve the learning guarantees at a constant level when the train set is small, which is significant when the awful overfitting occurs. Simulation results on the Gaussian mixture model and empirical results on generative adversarial nets support our theoretical conclusions. Our code is available at https://github.com/ML-GSAI/Understanding-GDA."
                    },
                    {
                        "title": "Max-margin Deep Generative Models",
                        "abstract": "Deep generative models (DGMs) are effective on learning multilayered representations of complex data and performing inference of input data by exploring the generative ability. However, little work has been done on examining or empowering the discriminative ability of DGMs on making accurate predictions. This paper presents max-margin deep generative models (mmDGMs), which explore the strongly discriminative principle of max-margin learning to improve the discriminative power of DGMs, while retaining the generative capability. We develop an efficient doubly stochastic subgradient algorithm for the piecewise linear objective. Empirical results on MNIST and SVHN datasets demonstrate that (1) max-margin learning can significantly improve the prediction performance of DGMs and meanwhile retain the generative ability; and (2) mmDGMs are competitive to the state-of-the-art fully discriminative networks by employing deep convolutional neural networks (CNNs) as both recognition and generative models."
                    },
                    {
                        "title": "Triple Generative Adversarial Networks",
                        "abstract": "We propose a unified game-theoretical framework to perform classification and conditional image generation given limited supervision. It is formulated as a three-player minimax game consisting of a generator, a classifier and a discriminator, and therefore is referred to as Triple Generative Adversarial Network (Triple-GAN). The generator and the classifier characterize the conditional distributions between images and labels to perform conditional generation and classification, respectively. The discriminator solely focuses on identifying fake image-label pairs. Under a nonparametric assumption, we prove the unique equilibrium of the game is that the distributions characterized by the generator and the classifier converge to the data distribution. As a byproduct of the three-player mechanism, Triple-GAN is flexible to incorporate different semi-supervised classifiers and GAN architectures. We evaluate Triple-GAN in two challenging settings, namely, semi-supervised learning and the extreme low data regime. In both settings, Triple-GAN can achieve excellent classification results and generate meaningful samples in a specific class simultaneously. In particular, using a commonly adopted 13-layer CNN classifier, Triple-GAN outperforms extensive semi-supervised learning methods substantially on more than 10 benchmarks no matter data augmentation is applied or not."
                    },
                    {
                        "title": "Graphical Generative Adversarial Networks",
                        "abstract": "We propose Graphical Generative Adversarial Networks (Graphical-GAN) to model structured data. Graphical-GAN conjoins the power of Bayesian networks on compactly representing the dependency structures among random variables and that of generative adversarial networks on learning expressive dependency functions. We introduce a structured recognition model to infer the posterior distribution of latent variables given observations. We generalize the Expectation Propagation (EP) algorithm to learn the generative model and recognition model jointly. Finally, we present two important instances of Graphical-GAN, i.e. Gaussian Mixture GAN (GMGAN) and State Space GAN (SSGAN), which can successfully learn the discrete and temporal structures on visual datasets, respectively."
                    },
                    {
                        "title": "Triple Generative Adversarial Nets",
                        "abstract": "Generative Adversarial Nets (GANs) have shown promise in image generation and semi-supervised learning (SSL). However, existing GANs in SSL have two problems: (1) the generator and the discriminator (i.e. the classifier) may not be optimal at the same time; and (2) the generator cannot control the semantics of the generated samples. The problems essentially arise from the two-player formulation, where a single discriminator shares incompatible roles of identifying fake samples and predicting labels and it only estimates the data without considering the labels. To address the problems, we present triple generative adversarial net (Triple-GAN), which consists of three players---a generator, a discriminator and a classifier. The generator and the classifier characterize the conditional distributions between images and labels, and the discriminator solely focuses on identifying fake image-label pairs. We design compatible utilities to ensure that the distributions characterized by the classifier and the generator both converge to the data distribution. Our results on various datasets demonstrate that Triple-GAN as a unified model can simultaneously (1) achieve the state-of-the-art classification results among deep generative models, and (2) disentangle the classes and styles of the input and transfer smoothly in the data space via interpolation in the latent space class-conditionally."
                    },
                    {
                        "title": "Collaborative Filtering with User-Item Co-Autoregressive Models",
                        "abstract": "Deep neural networks have shown promise in collaborative filtering (CF). However, existing neural approaches are either user-based or item-based, which cannot leverage all the underlying information explicitly. We propose CF-UIcA, a neural co-autoregressive model for CF tasks, which exploits the structural correlation in the domains of both users and items. The co-autoregression allows extra desired properties to be incorporated for different tasks. Furthermore, we develop an efficient stochastic learning algorithm to handle large scale datasets. We evaluate CF-UIcA on two popular benchmarks: MovieLens 1M and Netflix, and achieve state-of-the-art performance in both rating prediction and top-N recommendation tasks, which demonstrates the effectiveness of CF-UIcA."
                    },
                    {
                        "title": "Learning Implicit Generative Models by Teaching Explicit Ones",
                        "abstract": "Implicit generative models are difficult to train as no explicit density functions are defined. Generative adversarial nets (GANs) present a minimax framework to train such models, which however can suffer from mode collapse due to the nature of the JS-divergence. This paper presents a learning by teaching (LBT) approach to learning implicit models, which intrinsically avoids the mode collapse problem by optimizing a KL-divergence rather than the JS-divergence in GANs. In LBT, an auxiliary density estimator is introduced to fit the implicit model's distribution while the implicit model teaches the density estimator to match the data distribution. LBT is formulated as a bilevel optimization problem, whose optimal generator matches the true data distribution. LBT can be naturally integrated with GANs to derive a hybrid LBT-GAN that enjoys complimentary benefits. Finally, we present a stochastic gradient ascent algorithm with unrolling to solve the challenging learning problems. Experimental results demonstrate the effectiveness of our method."
                    },
                    {
                        "title": "Implicit Normalizing Flows",
                        "abstract": "Normalizing flows define a probability distribution by an explicit invertible transformation $\\boldsymbol{\\mathbf{z}}=f(\\boldsymbol{\\mathbf{x}})$. In this work, we present implicit normalizing flows (ImpFlows), which generalize normalizing flows by allowing the mapping to be implicitly defined by the roots of an equation $F(\\boldsymbol{\\mathbf{z}}, \\boldsymbol{\\mathbf{x}})= \\boldsymbol{\\mathbf{0}}$. ImpFlows build on residual flows (ResFlows) with a proper balance between expressiveness and tractability. Through theoretical analysis, we show that the function space of ImpFlow is strictly richer than that of ResFlows. Furthermore, for any ResFlow with a fixed number of blocks, there exists some function that ResFlow has a non-negligible approximation error. However, the function is exactly representable by a single-block ImpFlow. We propose a scalable algorithm to train and draw samples from ImpFlows. Empirically, we evaluate ImpFlow on several classification and density modeling tasks, and ImpFlow outperforms ResFlow with a comparable amount of parameters on all the benchmarks."
                    }
                ]
            },
            "edb3ed74-a6c3-45f6-8f53-4fe7867058aa": {
                "pk": "edb3ed74-a6c3-45f6-8f53-4fe7867058aa",
                "name": "Wei Chen",
                "collaborators": [],
                "domain": [
                    "Quantum Mechanics",
                    "Topological Insulators",
                    "Financial Mathematics",
                    "Signal Processing"
                ],
                "publications": [
                    {
                        "title": "Data sets of very large linear feasibility problems solved by projection methods",
                        "abstract": "We give a link to a page on the Web on which we deposited a set of eight huge Linear Programming (LP) problems for Intensity-Modulated Proton Therapy (IMPT) treatment planning. These huge LP problems were employed in our recent research and we were asked to make them public."
                    },
                    {
                        "title": "An Issue in the Martingale Analysis of the Influence Maximization Algorithm IMM",
                        "abstract": "This paper explains a subtle issue in the martingale analysis of the IMM algorithm, a state-of-the-art influence maximization algorithm. Two workarounds are proposed to fix the issue, both requiring minor changes on the algorithm and incurring a slight penalty on the running time of the algorithm."
                    },
                    {
                        "title": "Edelstein and inverse Edelstein effects caused by the pristine surface states of topological insulators",
                        "abstract": "The Edelstein effect caused by the pristine surface states of three-dimensional topological insulators is investigated by means of a semiclassical approach. The combined effect of random impurity scattering and the spin-momentum locking of the gapless Dirac cone yields a current-induced surface spin accumulation independent from chemical potential and temperature. Through combing the semiclassical approach with the Bloch equation, the inverse Edelstein effect that converts the spin pumping spin current into a charge current is well explained. Consistency of these results with various experiments will be elaborated in detail."
                    },
                    {
                        "title": "Two Mirroring And Interpolating Methods To Estimate Peak Position For Symmetric Signals With Single Peak",
                        "abstract": "Signals with single peak and symmetry property are very common in various fields, such as probability density function of normal distribution. Among the information contained in such signals, peak position is the most important, sometimes even the only parameter concerned. Current methods for peak position estimation are always with low precision and bad noise resistance, perform badly for sparse spectrum. This manuscript proposes two new algorithms that take advantage of symmetry property, conduct mirroring and interpolating operations to condense signal spectrum. From tests done in this paper, these two algorithms indicate outstanding advantages compared with current methods."
                    },
                    {
                        "title": "A Novel Probability Weighting Method To Fit Gaussian Functions",
                        "abstract": "Gaussian functions are commonly used in different fields, many real signals can be modeled into such form. Research aiming to obtain a precise fitting result for these functions is very meaningful. This manuscript intends to introduce a new algorithm used to estimate the full parameters of the Gaussian-shaped function. It is basically a weighting method, starting from Caruana's method, while the selection of weighting factors is from the statistics view and based on the estimation of the confidence level for the samples. Tests designed for comparison with current similar methods have been conducted. The simulation results indicate a good performance for this new method, mainly in precision and robustness."
                    },
                    {
                        "title": "Quantum geometrical properties of topological materials",
                        "abstract": "The momentum space of topological insulators and topological superconductors is equipped with a quantum metric defined from the overlap of neighboring valence band states or quasihole states. We investigate the quantum geometrical properties of these materials within the framework of Dirac models and differential geometry. The Ricci scalar is found to be a constant throughout the whole Brillouin zone, and the vacuum Einstein equation is satisfied in 3D with a finite cosmological constant. For linear Dirac models, several geometrical properties are found to be independent of the band gap, including the straight line geodesic, constant volume of the curved momentum space, exponential decay form of the nonlocal topological marker, and unity Euler characteristic in 2D, indicating the peculiar yet universal quantum geometrical properties of these models."
                    },
                    {
                        "title": "Time Consistent Bid-Ask Dynamic Pricing Mechanisms for Contingent Claims and Its Numerical Simulations Under Uncertainty",
                        "abstract": "We study time consistent dynamic pricing mechanisms of European contingent claims under uncertainty by using G framework introduced by Peng ([24]). We consider a financial market consisting of a riskless asset and a risky stock with price process modelled by a geometric generalized G-Brownian motion, which features the drift uncertainty and volatility uncertainty of the stock price process. Using the techniques on G-framework we show that the risk premium of the asset is uncertain and distributed with maximum distribution. A time consistent G-expectation is defined by the viscosity solution of the G-heat equation. Using the time consistent G-expectation we define the G dynamic pricing mechanism for the claim. We prove that G dynamic pricing mechanism is the bid-ask Markovian dynamic pricing mechanism. The full nonlinear PDE is derived to describe the bid (resp. ask) price process of the claim. Monotone implicit characteristic finite difference schemes for the nonlinear PDE are given, nonlinear iterative schemes are constructed, and the simulations of the bid (resp. ask) prices of contingent claims under uncertainty are implemented."
                    },
                    {
                        "title": "Inductance Due to Spin Current",
                        "abstract": "The inductance of spintronic devices that transport charge neutral spin currents is discussed. It is known that in a media that contains charge neutral spins, a time-varying electric field induces a spin current. We show that since the spin current itself produces an electric field, this implies existence of inductance and electromotive force when the spin current changes with time. The relations between the electromotive force and the corresponding flux, which is a vector calculated by the cross product of electric field and the trajectory of the device, are clarified. The relativistic origin generally renders an extremely small inductance, which indicates the advantage of spin current in building low inductance devices. The same argument also explains the inductance due to electric dipole current, and applies to physical dipoles consist of polarized bound charges."
                    },
                    {
                        "title": "Fractional G-White Noise Theory, Wavelet Decomposition for Fractional G-Brownian Motion, and Bid-Ask Pricing Application to Finance Under Uncertainty",
                        "abstract": "G-framework is presented by Peng [41] for measure risk under uncertainty. In this paper, we define fractional G-Brownian motion (fGBm). Fractional G-Brownian motion is a centered G-Gaussian process with zero mean and stationary increments in the sense of sub-linearity with Hurst index $H\\in (0,1)$. This process has stationary increments, self-similarity, and long rang dependence properties in the sense of sub-linearity. These properties make the fractional G-Brownian motion a suitable driven process in mathematical finance. We construct wavelet decomposition of the fGBm by wavelet with compactly support. We develop fractional G-white noise theory, define G-It\\^o-Wick stochastic integral, establish the fractional G-It\\^o formula and the fractional G-Clark-Ocone formula, and derive the G-Girsanov's Theorem. For application the G-white noise theory, we consider the financial market modelled by G-Wick-It\\^o type of SDE driven by fGBm. The financial asset price modelled by fGBm has volatility uncertainty, using G-Girsanov's Theorem and G-Clark-Ocone Theorem, we derive that sublinear expectation of the discounted European contingent claim is the bid-ask price of the claim."
                    },
                    {
                        "title": "G-Doob-Meyer Decomposition and its Application in Bid-Ask Pricing for American Contingent Claim Under Knightian Uncertainty",
                        "abstract": "The target of this paper is to establish the bid-ask pricing frame work for the American contingent claims against risky assets with G-asset price systems (see \\cite{Chen2013b}) on the financial market under Knight uncertainty. First, we prove G-Dooby-Meyer decomposition for G-supermartingale. Furthermore, we consider bid-ask pricing American contingent claims under Knight uncertain, by using G-Dooby-Meyer decomposition, we construct dynamic superhedge stragies for the optimal stopping problem, and prove that the value functions of the optimal stopping problems are the bid and ask prices of the American contingent claims under Knight uncertain. Finally, we consider a free boundary problem, prove the strong solution existence of the free boundary problem, and derive that the value function of the optimal stopping problem is equivalent to the strong solution to the free boundary problem."
                    },
                    {
                        "title": "Scaling Theory of Topological Phase Transitions",
                        "abstract": "Topologically ordered systems are characterized by topological invariants that are often calculated from the momentum space integration of a certain function that represents the curvature of the many-body state. The curvature function may be Berry curvature, Berry connection, or other quantities depending on the system. Akin to stretching a messy string to reveal the number of knots it contains, a scaling procedure is proposed for the curvature function in inversion symmetric systems, from which the topological phase transition can be identified from the flow of the driving energy parameters that control the topology (hopping, chemical potential, etc.) under scaling. At an infinitesimal operation, one obtains the renormalization group (RG) equations for the driving energy parameters. A length scale defined from the curvature function near the gap-closing momentum is suggested to characterize the scale invariance at critical points and fixed points, and displays a universal critical behavior in a variety of systems examined."
                    },
                    {
                        "title": "Weakly Interacting Topological Insulators: Quantum Criticality and Renormalization Group Approach",
                        "abstract": "For D-dimensional weakly interacting topological insulators in certain symmetry classes, the topological invariant can be calculated from a D- or (D+1)-dimensional integration over a certain curvature function that is expressed in terms of single-particle Green's functions. Based on the divergence of curvature function at the topological phase transition, we demonstrate how a renormalization group approach circumvents these integrations and reduces the necessary calculation to that for the Green's function alone, rendering a numerically efficient tool to identify topological phase transitions in a large parameter space. The method further unveils a number of statistical aspects related to the quantum criticality in weakly interacting topological insulators, including correlation function, critical exponents, and scaling laws, that can be used to characterize the topological phase transitions driven by either interacting or noninteracting parameters. We use 1D class BDI and 2D class A Dirac models with electron-electron and electron-phonon interactions to demonstrate these principles, and find that interactions may change the critical exponents of the topological insulators."
                    },
                    {
                        "title": "Spin torque and persistent currents caused by percolation of topological surface states",
                        "abstract": "The topological insulator/ferromagnetic metal (TI/FMM) bilayer thin films emerged as promising topological surface state-based spintronic devices, most notably in their efficiency of current-induced spin torque. Using a cubic lattice model, we reveal that the surface state Dirac cone of the TI can gradually merge into or be highly intertwined with the FMM bulk bands, and the surface states percolate into the FMM and eventually hybridize with the quantum well states therein. The magnetization can distort the spin-momentum locking of the surface states and yield an asymmetric band structure, which causes a laminar flow of room temperature persistent charge current. Moreover, the proximity to the FMM also promotes a persistent laminar spin current. Through a linear response theory, we elaborate that both the surface state and the FMM bulk bands contribute to the current-induced spin torque, and their real wave functions render the spin torque predominantly field-like, with a magnitude highly influenced by the degree of the percolation of the surface states. On the other hand, impurities can change the spin polarization expected from the Edelstein effect and generate a damping-like torque, and produce a torque even when the magnetization points in-plane and orthogonal to the current direction."
                    },
                    {
                        "title": "Absence of equilibrium edge currents in theoretical models of topological insulators",
                        "abstract": "The low energy sector of 2D and 3D topological insulators (TIs) exhibits propagating edge states, which has speculated the existence of equilibrium edge currents or edge spin currents. We demonstrate that if the low energy sector of TIs is regularized in a straightforward manner into a square or cubic lattice, then the current from the edge states is in fact canceled out exactly by that from the valence bands, rendering no edge current. This result serves as a warning that for any equilibrium property of topological insulators, the contribution from the valence bands should not be overlooked. In these regularized lattice model, there is a finite edge current only if the Dirac point of the edge states is shifted away from the chemical potential, for instance by doping, impurities, edge confining potential, surface band bending, or gate voltage. The edge current in small quantum dots as a function of the gate voltage is quantized, and the edge current can flow out of the gated region up to the decay length of the edge state."
                    },
                    {
                        "title": "A Sequential Variational Mode Decomposition Method",
                        "abstract": "In this paper, we introduce a sequential variational mode decomposition method to separate non-stationary mixed signals successively. This method is inspired by the variational method, and can precisely recover the original components one by one from the raw mixture without prior knowing or assuming the number of components. And in such a way, the mode number also can be determined during the separation procedure. Such character brings great convenience for real application and differs from the current VMD method. Furthermore, we also conduct a principal elongation for the mixture signal before the decomposing operation. By applying such an approach, the end effect can be reduced to a low level compared with the VMD method. To obtain higher accuracy, a refinement process has been introduced after gross extraction. Combined these techniques together, the final decomposition result implies a significant improvement compared with the VMD method and EMD method."
                    },
                    {
                        "title": "Optical absorption measurement of spin Berry curvature and spin Chern marker",
                        "abstract": "In two-dimensional time-reversal symmetric topological insulators described by Dirac models, the ${\\mathbb Z}_{2}$ topological invariant can be described by the spin Chern number. We present a linear response theory for the spin Berry curvature that integrates to the spin Chern number, and introduce its spectral function that can be measured at finite temperature by momentum- and spin-resolved circular dichroism, which may be achieved by pump-probe type of experiments using spin- and time-resolved ARPES. As a result, the sign of the Pfaffian of the ${\\mathbb Z}_{2}$ invariant can be directly measured. The spin Chern number expressed in real space yields a spin Chern marker, whose spatial variation may be measured by circular dichroism and spin-resolved photoemission with a spatial resolution. A spin Chern correlator and a nonlocal spin Chern marker are further proposed to characterize the quantum criticality near topological phase transitions, which are shown to take the form of overlaps between spin-selected Wannier states."
                    },
                    {
                        "title": "Universal topological marker",
                        "abstract": "We elaborate that for topological insulators and topological superconductors described by Dirac models in any dimension and symmetry class, the topological order can be mapped to lattice sites by a universal topological marker. Deriving from a recently discovered momentum-space universal topological invariant, we introduce a topological operator that consists of alternating projectors to filled and empty lattice eigenstates and the position operators, multiplied by the Dirac matrices that are omitted in the Hamiltonian. The topological operator projected to lattice sites yields the topological marker, whose form is explicitly constructed for every topologically nontrivial symmetry class from 1D to 3D. The off-diagonal elements of the topological operator yields a nonlocal topological marker, which decays with a correlation length that diverges at topological phase transitions, and represents a Wannier state correlation function. Various prototype examples, including Su-Schrieffer-Heeger model, Majorana chain, Chern insulators, Bernevig-Hughes-Zhang model, 2D chiral and helical $p$-wave superconductors, lattice model of $^{3}$He B-phase, and 3D time-reversal symmetric topological insulators, etc, are employed to demonstrate the ubiquity of our formalism."
                    },
                    {
                        "title": "Dielectric and optical markers originated from quantum geometry",
                        "abstract": "We elaborate that practically all the non-excitonic dielectric and optical properties of semiconductors and insulators are determined by the quantum metric of the valence band states, including charge susceptibility, relative dielectric constant, optical conductivity, dielectric function, refractive index, absorption coefficient, reflectance, and transmittance. The key to this recognition is the complex optical conductivity, which contains the quantum metric in the optical transition matrix element, and the fact that all these dielectric and optical properties can be expressed in terms of the real and imaginary parts of optical conductivity. Our formalism allows to map all these properties to real space lattice sites as local markers following the formalism of topological markers, enabling the effect of disorder on the propagation of electromagnetic wave in the nanometer scale to be investigated, as demonstrated by a minimal model of 3D topological insulators."
                    },
                    {
                        "title": "Edelstein effects, spin-transfer torque, and spin pumping caused by pristine surface states of topological insulators",
                        "abstract": "The Edelstein effect caused by the pristine surface states of three-dimensional topological insulators is investigated by means of a semiclassical approach. The combined effect of random impurity scattering and the spin-momentum locking of the gapless Dirac cone yields a current-induced surface spin accumulation independent from chemical potential and temperature. In a nearby ferromagnet that does not make direct contact with the topological insulator, the bound state nature of the pristine surface state causes a spin-transfer torque that is entirely field-like, whose magnitude is highly influenced by the interface cleanliness and the quantum well state of the ferromagnet. Through incorporating quantum tunneling into Bloch equation, the spin pumping mediated by the pristine surface state is shown to be described by the same spin mixing conductance as the spin-transfer torque, and a semiclassical approach is proposed to explain the inverse Edelstein effect that converts the spin pumping spin current into a charge current. Consistency of these results with various experiments will be elaborated in detail."
                    },
                    {
                        "title": "Berry curvature and quantum metric in copper-substituted lead phosphate apatite",
                        "abstract": "The recent discovery of copper-substituted lead phosphate apatite, also known as LK-99, has caught much attention owing to certain experimental evidence of room-temperature superconductivity, although this claim is currently under intensive debate. Be it superconducting or not, we show that the normal state of this material has peculiar quantum geometrical properties that may be related to the magnetism and the mechanism for flat band superconductivity. Based on a recently proposed spinless two-band tight-binding model for the Pb-Cu hexagonal lattice subset of the crystalline structure, which qualitatively captures the two flat bands in the band structure, we elaborate the highly anisotropic Berry curvature and quantum metric in the regions of Brillouin zone where one flat band is above and the other below the Fermi surface. In these regions, the Berry curvature has a pattern in the planar momentum that remains unchanged along the out-of-plane momentum. Moreover, the net orbital magnetization contributed from the Berry curvature is zero, signifying that the magnetism in this material should come from other sources. The quantum metric has a similar momentum dependence, and its two planar components are found to be roughly the same but the out-of-plane component vanishes, hinting that the superfluid stiffness of the flat band superconductivity, shall it occur, may be quite anisotropic."
                    }
                ]
            },
            "7eb9a1e0-c4bc-480a-b538-689b46b170ce": {
                "pk": "7eb9a1e0-c4bc-480a-b538-689b46b170ce",
                "name": "Min Lin",
                "collaborators": [
                    "Shuicheng Yan",
                    "Xu He",
                    "Qiang Chen",
                    "Shuo Li",
                    "Xuan Luo",
                    "Jie Fu",
                    "Yoshua Bengio"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Continual Learning",
                    "Deep Learning",
                    "Automatic Differentiation"
                ],
                "publications": [
                    {
                        "title": "Softmax GAN",
                        "abstract": "Softmax GAN is a novel variant of Generative Adversarial Network (GAN). The key idea of Softmax GAN is to replace the classification loss in the original GAN with a softmax cross-entropy loss in the sample space of one single batch. In the adversarial learning of $N$ real training samples and $M$ generated samples, the target of discriminator training is to distribute all the probability mass to the real samples, each with probability $\\frac{1}{M}$, and distribute zero probability to generated data. In the generator training phase, the target is to assign equal probability to all data points in the batch, each with probability $\\frac{1}{M+N}$. While the original GAN is closely related to Noise Contrastive Estimation (NCE), we show that Softmax GAN is the Importance Sampling version of GAN. We futher demonstrate with experiments that this simple change stabilizes GAN training."
                    },
                    {
                        "title": "Automatic Functional Differentiation in JAX",
                        "abstract": "We extend JAX with the capability to automatically differentiate higher-order functions (functionals and operators). By representing functions as a generalization of arrays, we seamlessly use JAX's existing primitive system to implement higher-order functions. We present a set of primitive operators that serve as foundational building blocks for constructing several key types of functionals. For every introduced primitive operator, we derive and implement both linearization and transposition rules, aligning with JAX's internal protocols for forward and reverse mode automatic differentiation. This enhancement allows for functional differentiation in the same syntax traditionally use for functions. The resulting functional gradients are themselves functions ready to be invoked in python. We showcase this tool's efficacy and simplicity through applications where functional derivatives are indispensable. The source code of this work is released at https://github.com/sail-sg/autofd ."
                    },
                    {
                        "title": "Continual Learning from the Perspective of Compression",
                        "abstract": "Connectionist models such as neural networks suffer from catastrophic forgetting. In this work, we study this problem from the perspective of information theory and define forgetting as the increase of description lengths of previous data when they are compressed with a sequentially learned model. In addition, we show that continual learning approaches based on variational posterior approximation and generative replay can be considered as approximations to two prequential coding methods in compression, namely, the Bayesian mixture code and maximum likelihood (ML) plug-in code. We compare these approaches in terms of both compression and forgetting and empirically study the reasons that limit the performance of continual learning methods based on variational posterior approximation. To address these limitations, we propose a new continual learning method that combines ML plug-in and Bayesian mixture codes."
                    },
                    {
                        "title": "Network In Network",
                        "abstract": "We propose a novel deep network structure called \"Network In Network\" (NIN) to enhance model discriminability for local patches within the receptive field. The conventional convolutional layer uses linear filters followed by a nonlinear activation function to scan the input. Instead, we build micro neural networks with more complex structures to abstract the data within the receptive field. We instantiate the micro neural network with a multilayer perceptron, which is a potent function approximator. The feature maps are obtained by sliding the micro networks over the input in a similar manner as CNN; they are then fed into the next layer. Deep NIN can be implemented by stacking mutiple of the above described structure. With enhanced local modeling via the micro network, we are able to utilize global average pooling over feature maps in the classification layer, which is easier to interpret and less prone to overfitting than traditional fully connected layers. We demonstrated the state-of-the-art classification performances with NIN on CIFAR-10 and CIFAR-100, and reasonable performances on SVHN and MNIST datasets."
                    },
                    {
                        "title": "Purine: A bi-graph based deep learning framework",
                        "abstract": "In this paper, we introduce a novel deep learning framework, termed Purine. In Purine, a deep network is expressed as a bipartite graph (bi-graph), which is composed of interconnected operators and data tensors. With the bi-graph abstraction, networks are easily solvable with event-driven task dispatcher. We then demonstrate that different parallelism schemes over GPUs and/or CPUs on single or multiple PCs can be universally implemented by graph composition. This eases researchers from coding for various parallelization schemes, and the same dispatcher can be used for solving variant graphs. Scheduled by the task dispatcher, memory transfers are fully overlapped with other computations, which greatly reduce the communication overhead and help us achieve approximate linear acceleration."
                    },
                    {
                        "title": "Conditional Computation for Continual Learning",
                        "abstract": "Catastrophic forgetting of connectionist neural networks is caused by the global sharing of parameters among all training examples. In this study, we analyze parameter sharing under the conditional computation framework where the parameters of a neural network are conditioned on each input example. At one extreme, if each input example uses a disjoint set of parameters, there is no sharing of parameters thus no catastrophic forgetting. At the other extreme, if the parameters are the same for every example, it reduces to the conventional neural network. We then introduce a clipped version of maxout networks which lies in the middle, i.e. parameters are shared partially among examples. Based on the parameter sharing analysis, we can locate a limited set of examples that are interfered when learning a new example. We propose to perform rehearsal on this set to prevent forgetting, which is termed as conditional rehearsal. Finally, we demonstrate the effectiveness of the proposed method in an online non-stationary setup, where updates are made after each new example and the distribution of the received example shifts over time."
                    }
                ]
            }
        }
    },
    "2310.17567": {
        "paper_data": {
            "title": "Skill-Mix: a Flexible and Expandable Family of Evaluations for AI models",
            "url": "http://arxiv.org/abs/2310.17567v1",
            "arxiv_id": "2310.17567",
            "authors": [
                "Dingli Yu",
                "Simran Kaur",
                "Arushi Gupta",
                "Jonah Brown-Cohen",
                "Anirudh Goyal",
                "Sanjeev Arora"
            ],
            "abstract": "With LLMs shifting their role from statistical modeling of language to serving as general-purpose AI agents, how should LLM evaluations change? Arguably, a key ability of an AI agent is to flexibly combine, as needed, the basic skills it has learned. The capability to combine skills plays an important role in (human) pedagogy and also in a paper on emergence phenomena (Arora & Goyal, 2023).   This work introduces Skill-Mix, a new evaluation to measure ability to combine skills. Using a list of $N$ skills the evaluator repeatedly picks random subsets of $k$ skills and asks the LLM to produce text combining that subset of skills. Since the number of subsets grows like $N^k$, for even modest $k$ this evaluation will, with high probability, require the LLM to produce text significantly different from any text in the training set. The paper develops a methodology for (a) designing and administering such an evaluation, and (b) automatic grading (plus spot-checking by humans) of the results using GPT-4 as well as the open LLaMA-2 70B model.   Administering a version of to popular chatbots gave results that, while generally in line with prior expectations, contained surprises. Sizeable differences exist among model capabilities that are not captured by their ranking on popular LLM leaderboards (\"cramming for the leaderboard\"). Furthermore, simple probability calculations indicate that GPT-4's reasonable performance on $k=5$ is suggestive of going beyond \"stochastic parrot\" behavior (Bender et al., 2021), i.e., it combines skills in ways that it had not seen during training.   We sketch how the methodology can lead to a Skill-Mix based eco-system of open evaluations for AI capabilities of future models.",
            "introduction": " introduction to critical and logical thinking. In WW Norton & Company Inc , 2014. Kenneth R Koedinger, Elizabeth A McLaughlin, and John C Stamper. Automated student model improvement. International Educational Data Mining Society , 2012. Kenneth R Koedinger, Paulo F Carvalho, Ran Liu, and Elizabeth A McLaughlin. An astonishing regularity in student learning rate. Proceedings of the National Academy of Sciences , 120(13): e2221311120, 2023. Nan Li, Eliane Stampfer, William Cohen, and Kenneth Koedinger. General and efficient cognitive model discovery using a simulated student. In Proceedings of the Annual Meeting of the Cognitive Science Society , volume 35, 2013. Xuechen Li, Tianyi Zhang, Yann Dubois, Rohan Taori, Ishaan Gulrajani, Carlos Guestrin, Percy Liang, and Tatsunori B. Hashimoto. Alpacaeval: An automatic evaluator of instruction-following models. https://github.com/tatsu-lab/alpaca_eval , 2023. Yucheng Li. Estimating contamination via perplexity: Quantifying memorisation in language model evaluation, 2023. Stephanie Lin, Jacob Hilton, and Owain Evans. Truthfulqa: Measuring how models mimic human falsehoods. arXiv preprint arXiv:2109.07958 , 2021. Edward Loper and Steven Bird. Nltk: The natural language toolkit. arXiv preprint cs/0205028 , 2002. 13Mistral AI Team. Mistral 7b, 9 2023. URL https://mistral.ai/news/ announcing-mistral-7b/ . Erik T Mueller. Commonsense reasoning: an event calculus based approach . Morgan Kaufmann, 2014. Arvind Narayan and Sayash Kapoor. Gpt-4 and professional benchmarks: the wrong answer to the wrong question. URL https://www.aisnakeoil.com/p/ gpt-4-and-professional-benchmarks . Shaun Nichols and Stephen P Stich. Mindreading. an integrated account of pretence, self-awareness, and understanding other\u2019s minds. In Oxford University Press , 2003. OpenAI. Gpt-4 technical report, 2023. Karl Pertsch, Youngwoon Lee, and Joseph Lim. Accelerating reinforcement learning with learned skill priors. In Conference on robot learning , pp. 188\u2013204. PMLR, 2021. Laura Ruis, Akbir Khan, Stella Biderman, Sara Hooker, Tim Rockt \u00a8aschel, and Edward Grefenstette. Large language models are not zero-shot communicators. NeuRIPS , 2023. Philipp Schmid, Omar Sanseviero, Pedro Cuenca, Leandro von Werra, and Julien Launay. Spread your wings: Falcon 180b is here, 9 2023. URL https://huggingface.co/blog/ falcon-180b . TigerResearch. TigerBot: A cutting-edge foundation for your very own llm. https://github. com/TigerResearch/TigerBot . Accessed: 2023-09-30. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288 , 2023. Xwin-LM Team. Xwin-lm, 9 2023. URL https://github.com/Xwin-LM/Xwin-LM . Rowan Zellers, Ari Holtzman, Yonatan Bisk, Ali Farhadi, and Yejin Choi. Hellaswag: Can a machine really finish your sentence?, 2019. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric. P Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. Judging llm-as-a-judge with mt-bench and chatbot arena, 2023. 14A L IST OF SKILLS AND TOPICS Here we release a random sample of 10% of the skill list and topic list. We do not release the full lists to avoid potential \u201ccramming\u201d for SKILL -MIX. Table 5: 10 randomly sampled skills from the 101 skills we used in SKILL -MIX evaluation. Category Skill Definition Example reasoning self serving bias A cognitive or perceptual process that is distorted by the need to maintain and enhance one\u2019s self es- teem.\u201cIf I do well on the exam, it\u2019s because of my academic prowess and hard work. If I do poorly, it\u2019s because the course was poorly taught, and the exam was poorly proctored.\u201d rhetorical accident (fallacy) an informal fallacy and a deductively valid but unsound argument oc- curring in a statistical syllogism (an argument based on a generaliza- tion) when an exception to a rule of thumb",
            "references": [
                {
                    "title": "Estimating Contamination via Perplexity: Quantifying Memorisation in Language Model Evaluation",
                    "abstract": "Data contamination in model evaluation is getting increasingly prevalent as the massive training corpora of large language models often unintentionally include benchmark samples. Therefore, contamination analysis has became an inevitable part of reliable model evaluation. However, existing method of contamination analysis requires the access of the entire training data which is often confidential for recent models. This prevent the community to rigorously audit these models and conduct accurate assessment of their capability. In this paper, we propose a novel method to quantify contamination without the access of the full training set, that measure the extent of contamination with perplexity. Our analysis provides evidence of significant memorisation of recent foundation models in popular reading comprehension, summarisation benchmarks, while multiple choice appears less contaminated."
                },
                {
                    "title": "A Theory for Emergence of Complex Skills in Language Models",
                    "abstract": "A major driver of AI products today is the fact that new skills emerge in language models when their parameter set and training corpora are scaled up. This phenomenon is poorly understood, and a mechanistic explanation via mathematical analysis of gradient-based training seems difficult. The current paper takes a different approach, analysing emergence using the famous (and empirical) Scaling Laws of LLMs and a simple statistical framework. Contributions include: (a) A statistical framework that relates cross-entropy loss of LLMs to competence on the basic skills that underlie language tasks. (b) Mathematical analysis showing that the Scaling Laws imply a strong form of inductive bias that allows the pre-trained model to learn very efficiently. We informally call this {\\em slingshot generalization} since naively viewed it appears to give competence levels at skills that violate usual generalization theory. (c) A key example of slingshot generalization, that competence at executing tasks involving $k$-tuples of skills emerges essentially at the same scaling and same rate as competence on the elementary skills themselves."
                },
                {
                    "title": "Skill-it! A Data-Driven Skills Framework for Understanding and Training Language Models",
                    "abstract": "The quality of training data impacts the performance of pre-trained large language models (LMs). Given a fixed budget of tokens, we study how to best select data that leads to good downstream model performance across tasks. We develop a new framework based on a simple hypothesis: just as humans acquire interdependent skills in a deliberate order, language models also follow a natural order when learning a set of skills from their training data. If such an order exists, it can be utilized for improved understanding of LMs and for data-efficient training. Using this intuition, our framework formalizes the notion of a skill and of an ordered set of skills in terms of the associated data. First, using both synthetic and real data, we demonstrate that these ordered skill sets exist, and that their existence enables more advanced skills to be learned with less data when we train on their prerequisite skills. Second, using our proposed framework, we introduce an online data sampling algorithm, Skill-It, over mixtures of skills for both continual pre-training and fine-tuning regimes, where the objective is to efficiently learn multiple skills in the former and an individual skill in the latter. On the LEGO synthetic in the continual pre-training setting, Skill-It obtains 36.5 points higher accuracy than random sampling. On the Natural Instructions dataset in the fine-tuning setting, Skill-It reduces the validation loss on the target skill by 13.6% versus training on data associated with the target skill itself. We apply our skills framework on the recent RedPajama dataset to continually pre-train a 3B-parameter LM, achieving higher accuracy on the LM Evaluation Harness with 1B tokens than the baseline approach of sampling uniformly over data sources with 3B tokens."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "An astonishing regularity in student learning rate",
                    "abstract": "Significance Prior research, often using self-report data, hypothesizes that the path to expertise requires extensive practice and that different learners acquire competence at different rates. Fitting cognitive and statistical growth models to 27 datasets involving observations of learning and performance in academic settings, we find evidence for the first hypothesis and against the second. Students do need extensive practice, about seven opportunities per component of knowledge. Students do not show substantial differences in their rate of learning. These results provide a challenge for learning theory to explain this striking similarity in student learning rate. They also suggest that educational achievement gaps come from differences in learning opportunities and that better access to such opportunities can help close those gaps."
                },
                {
                    "title": "The Goldilocks of Pragmatic Understanding: Fine-Tuning Strategy Matters for Implicature Resolution by LLMs",
                    "abstract": "Despite widespread use of LLMs as conversational agents, evaluations of performance fail to capture a crucial aspect of communication: interpreting language in context -- incorporating its pragmatics. Humans interpret language using beliefs and prior knowledge about the world. For example, we intuitively understand the response\"I wore gloves\"to the question\"Did you leave fingerprints?\"as meaning\"No\". To investigate whether LLMs have the ability to make this type of inference, known as an implicature, we design a simple task and evaluate four categories of widely used state-of-the-art models. We find that, despite only evaluating on utterances that require a binary inference (yes or no), models in three of these categories perform close to random. However, LLMs instruction-tuned at the example-level perform significantly better. These results suggest that certain fine-tuning strategies are far better at inducing pragmatic understanding in models. We present our findings as the starting point for further research into evaluating how LLMs interpret language in context and to drive the development of more pragmatic and useful models of human discourse."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "TruthfulQA: Measuring How Models Mimic Human Falsehoods",
                    "abstract": "We propose a benchmark to measure whether a language model is truthful in generating answers to questions. The benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics. We crafted questions that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. We tested GPT-3, GPT-Neo/J, GPT-2 and a T5-based model. The best model was truthful on 58% of questions, while human performance was 94%. Models generated many false answers that mimic popular misconceptions and have the potential to deceive humans. The largest models were generally the least truthful. This contrasts with other NLP tasks, where performance improves with model size. However, this result is expected if false answers are learned from the training distribution. We suggest that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web."
                },
                {
                    "title": "Evaluating Large Language Models Trained on Code",
                    "abstract": "We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics."
                },
                {
                    "title": "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? \ud83e\udd9c",
                    "abstract": "The past 3 years of work in NLP have been characterized by the development and deployment of ever larger language models, especially for English. BERT, its variants, GPT-2/3, and others, most recently Switch-C, have pushed the boundaries of the possible both through architectural innovations and through sheer size. Using these pretrained models and the methodology of fine-tuning them for specific tasks, researchers have extended the state of the art on a wide array of tasks as measured by leaderboards on specific benchmarks for English. In this paper, we take a step back and ask: How big is too big? What are the possible risks associated with this technology and what paths are available for mitigating those risks? We provide recommendations including weighing the environmental and financial costs first, investing resources into curating and carefully documenting datasets rather than ingesting everything on the web, carrying out pre-development exercises evaluating how the planned approach fits into research and development goals and supports stakeholder values, and encouraging research directions beyond ever larger language models."
                },
                {
                    "title": "Accelerating Reinforcement Learning with Learned Skill Priors",
                    "abstract": "Intelligent agents rely heavily on prior experience when learning a new task, yet most modern reinforcement learning (RL) approaches learn every task from scratch. One approach for leveraging prior knowledge is to transfer skills learned on prior tasks to the new task. However, as the amount of prior experience increases, the number of transferable skills grows too, making it challenging to explore the full set of available skills during downstream learning. Yet, intuitively, not all skills should be explored with equal probability; for example information about the current state can hint which skills are promising to explore. In this work, we propose to implement this intuition by learning a prior over skills. We propose a deep latent variable model that jointly learns an embedding space of skills and the skill prior from offline agent experience. We then extend common maximum-entropy RL approaches to use skill priors to guide downstream learning. We validate our approach, SPiRL (Skill-Prior RL), on complex navigation and robotic manipulation tasks and show that learned skill priors are essential for effective skill transfer from rich datasets. Videos and code are available at this https URL."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "HellaSwag: Can a Machine Really Finish Your Sentence?",
                    "abstract": "Recent work by Zellers et al. (2018) introduced a new task of commonsense natural language inference: given an event description such as \u201cA woman sits at a piano,\u201d a machine must select the most likely followup: \u201cShe sets her fingers on the keys.\u201d With the introduction of BERT, near human-level performance was reached. Does this mean that machines can perform human level commonsense inference? In this paper, we show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans (>95% accuracy), state-of-the-art models struggle (<48%). We achieve this via Adversarial Filtering (AF), a data collection paradigm wherein a series of discriminators iteratively select an adversarial set of machine-generated wrong answers. AF proves to be surprisingly robust. The key insight is to scale up the length and complexity of the dataset examples towards a critical \u2018Goldilocks\u2019 zone wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models. Our construction of HellaSwag, and its resulting difficulty, sheds light on the inner workings of deep pretrained models. More broadly, it suggests a new path forward for NLP research, in which benchmarks co-evolve with the evolving state-of-the-art in an adversarial way, so as to present ever-harder challenges."
                },
                {
                    "title": "Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge",
                    "abstract": "We present a new question set, text corpus, and baselines assembled to encourage AI research in advanced question answering. Together, these constitute the AI2 Reasoning Challenge (ARC), which requires far more powerful knowledge and reasoning than previous challenges such as SQuAD or SNLI. The ARC question set is partitioned into a Challenge Set and an Easy Set, where the Challenge Set contains only questions answered incorrectly by both a retrieval-based algorithm and a word co-occurence algorithm. The dataset contains only natural, grade-school science questions (authored for human tests), and is the largest public-domain set of this kind (7,787 questions). We test several baselines on the Challenge Set, including leading neural models from the SQuAD and SNLI tasks, and find that none are able to significantly outperform a random baseline, reflecting the difficult nature of this task. We are also releasing the ARC Corpus, a corpus of 14M science sentences relevant to the task, and implementations of the three neural baseline models tested. Can your model perform better? We pose ARC as a challenge to the community."
                },
                {
                    "title": "Together",
                    "abstract": "His parents sent him to the prestigious boarding school Phillips Exeter Academy in Exeter, New Hampshire, when he was 13.[6] In his book, The Road Less Traveled,[7] he confides the story of his brief stay at Exeter, and admits that it was a most miserable time. Finally, at age 15, during the spring holiday of his third year, he came home and refused to return to the school, whereupon his parents sought psychiatric help for him and he was (much to his amusement in later life) diagnosed with depression and recommended for a month's stay in a psychiatric hospital (unless he chose to return to school). He then transferred to Friends Seminary (a private K\u201312 school) in late 1952, and graduated in 1954, after which he received a BA from Harvard in 1958, and an MD degree from Case Western Reserve University in 1963.[6]"
                },
                {
                    "title": "Automated Student Model Improvement",
                    "abstract": "Student modeling plays a critical role in developing and improving instruction and instructional technologies. We present a technique for automated improvement of student models that leverages the DataShop repository, crowd sourcing, and a version of the Learning Factors Analysis algorithm. We demonstrate this method on eleven educational technology data sets from intelligent tutors to games in a variety of domains from math to second language learning. In at least ten of the eleven cases, the method discovers improved models based on better test-set prediction in cross validation. The improvements isolate flaws in the original student models, and we show how focused investigation of flawed parts of models leads to new insights into the student learning process and suggests specific improvements for tutor design. We also discuss the great potential for future work that substitutes alternative statistical models of learning from the EDM literature or alternative model search algorithms."
                },
                {
                    "title": "A Taxonomy for Learning, Teaching, and Assessing: A Revision of Bloom's Taxonomy of Educational Objectives",
                    "abstract": "List of Tables and Figures. Preface. Foreword. SECTION I: THE TAXONOMY, EDUCATIONAL OBJECTIVES AND STUDENT LEARNING. 1. Introduction. 2. The Structure, Specificity, and Problems of Objectives. SECTION II: THE REVISED TAXONOMY STRUCTURE. 3. The Taxonomy Table. 4. The Knowledge Dimension. 5. The Cognitive Process Dimension. SECTION III: THE TAXONOMY IN USE. 6. Using the Taxonomy Table. 7. Introduction to the Vignettes. 8. Nutrition Vignette. 9. Macbeth Vignette. 10. Addition Facts Vignette. 11. Parliamentary Acts Vignette. 12. Volcanoes? Here? Vignette. 13. Report Writing Vignette. 14. Addressing Long-standing Problems in Classroom Instruction. APPENDICES. Appendix A: Summary of the Changes from the Original Framework. Appendix B: Condensed Version of the Original Taxonomy of Educational Objectives: Cognitive Domain. References. Credits. Index."
                },
                {
                    "title": "Pragmatics: A multidisciplinary perspective",
                    "abstract": "Louise Cummings draws on analytic philosophy to connect three areas of meaning research: 1) utterance interpretation and its importance for understanding mundane reasoning; 2) the use of informal fallacies in presenting scientific inquiries and how this affects public understanding of science; and 3) acquired and developmental communication disorders. Integrating these disparate research interests, the ten chapters of the book present different facets of contemporary pragmatics. The first four chapters illustrate traditional approaches to the study of linguistic meaning. The material in chapters 5-7 examines how the mind processes pragmatic information, contrasts the systems available for assessing an argument, and presents a close comparison of two definitions of reasoning. The final three chapters look at how findings from pragmatics can be utilized within other intellectual domains."
                },
                {
                    "title": "General and Efficient Cognitive Model Discovery Using a Simulated Student",
                    "abstract": "General and Efficient Cognitive Model Discovery Using a Simulated Student Nan Li (nli1@cs.cmu.edu) Eliane Stampfer (estampfe@cs.cmu.edu) William W. Cohen (wcohen@cs.cmu.edu) Kenneth R. Koedinger (koedinger@cs.cmu.edu) School of Computer Science , Carnegie Mellon University 5000 Forbes Ave., Pittsburgh PA 15213 USA Abstract important instructional details may still be overlooked. Au- tomated search methods such as Learning Factor Analysis (LFA) (Cen, Koedinger, & Junker, 2006) are more objective: the algorithm searches through the space of human-provided factors to find a cognitive model that best matches with hu- man data. Although automated search methods have found better models than manual construction, the quality of the dis- covered model depends on the quality of the human-provided factors. If there is a better model that can not be expressed by known factors, LFA will not be able to uncover it. In Li, Matsuda, Cohen, and Koedinger (2011), we have proposed to use the state-of-the-art learning agent, SimStu- dent (Matsuda, Lee, Cohen, & Koedinger, 2009), to auto- matically discover cognitive models without depending on human-provided factors. SimStudent learns skill knowledge from demonstration and problem solving experience. Each skill SimStudent acquires corresponds to a KC in the cogni- tive model. To demonstrate the generality of this approach, we present evaluations of the SimStudent-generated models in three domains: algebra, stoichiometry, and fraction addi- tion. We validate the quality of the cognitive models using hu- man student data as in Koedinger and MacLaren (1997). In- stead of matching with performance data, we use the discov- ered cognitive model to predict human learning curve data. Experimental results show that for algebra and stoichiometry, SimStudent directly finds a better cognitive model than hu- mans. For fraction addition, SimStudent results assist LFA in finding a better cognitive model than a domain expert. We have also carried out an in-depth study using Focused Bene- fits Investigation (FBI) (Koedinger, McLaughlin, & Stamper, 2012) to better understand this machine learning approach, and discussed possible ways of further improvements. In order to better understand how humans acquire knowledge, one of the essential goals in cognitive science is to build a cognitive model of human learning. Moreover, a cognitive model that better matches student behavior will often yield bet- ter instruction in intelligent tutoring systems. However, man- ual construction of such cognitive models is time consuming, and requires domain expertise. Further, manually-constructed models may still miss distinctions in learning which are impor- tant for instruction. Our prior work proposed an approach that finds cognitive models using a state-of-the-art learning agent, SimStudent, and we demonstrated that, for algebra learning, the agent can find a better cognitive model than human experts. To ensure the generality of that proposed approach, we further apply it to three domains: algebra, stoichiometry, and frac- tion addition. To evaluate the quality of the cognitive models discovered, we measured how well the cognitive models fit to student learning curve data. In two of those domains, SimStu- dent directly discovers a cognitive model that predicts human student behavior better than the human-generated model. In fraction addition, SimStudent supported discovery of a better cognitive model in combination with another automated cog- nitive model discovery method. Keywords: cognitive model, machine learning, simulated stu- dent Introduction One of the fundamental goals in cognitive science is to un- derstand human knowledge acquisition. A cognitive model of human learning that fits data would be a significant achieve- ment. This goal also complements with another goal in edu- cation, which is to provide individualized instruction based on students\u2019 abilities, learning styes, etc. Cognitive mod- els provide intelligent tutoring systems with useful informa- tion on the learning task difficulties and transfer of learning among similar problems. A better cognitive model often leads to more effective tutoring. A cognitive model is a system that can solve problems in the various ways human students can. One common way of representing a cognitive model is a set of knowledge components (KC) (Koedinger & McLaughlin, 2010). The set of KCs includes the component skills, con- cepts, or percepts that a student must learn to be successful on the target tasks. For example, a KC \u201cdivide\u201d in algebra encodes how to proceed given problems of the form Nv = N (e.g., \u22123x = 6), where N stands for a number, and v stands for a variable. Nevertheless, manual construction of cognitive models re- mains time consuming and error prone. Traditional ways to construct cognitive models include structured interviews, think-aloud protocols, and rational analysis. Manual con- struction of cognitive models requires domain expertise, and A Brief Review of SimStudent SimStudent is an intelligent agent that inductively learns skills to solve problems from demonstrated solutions and from problem solving experience. It is a realization of pro- gramming by demonstration (Lau & Weld, 1998) using a variation of the version space algorithm (Mitchell, 1982), in- ductive logic programming (Muggleton & Raedt, 1994), and iterative-deepening depth-first search as underlying learning techniques. For more details, please refer to Matsuda et al. (2009). Recently, in order to build a more human-like in- telligent agent, we have developed a model of representation learning, and integrated it into SimStudent\u2019s skill acquisition mechanism."
                },
                {
                    "title": "From Emerging Perspectives on Learning, Teaching and Technology",
                    "abstract": "1 Social Constructivism Vignette 2 What is Social Constructivism? 3 Assumptions of Social Constructivism 4 Intersubjectivity of Social Meanings 5 Social Context for Learning 6 General Perspectives of Social Constructivism on Learning 7 Social Constructivism and Instructional Models 8 Sorting Out Variations on the Terms \"Constructionism\" and Constructivism\" 9 References 10 Bibliography 11 Additional Resources 12 Citation Social Constructivism"
                },
                {
                    "title": "NLTK: The Natural Language Toolkit",
                    "abstract": "The Natural Language Toolkit is a suite of program modules, data sets and tutorials supporting research and teaching in computational linguistics and natural language processing. NLTK is written in Python and distributed under the GPL open source license. Over the past year the toolkit has been rewritten, simplifying many linguistic data structures and taking advantage of recent enhancements in the Python language. This paper reports on the simpli-\ufb01ed toolkit and explains how it is used in teaching NLP."
                },
                {
                    "title": "Mindreading: An Integrated Account of Pretence, Self-Awareness, and Understanding Other Minds",
                    "abstract": "1. Introduction 2. A Cognitive Theory of Pretence 3. Pieces of Mind: A Theory of Third-Person Mindreading 4. Reading One's Own Mind 5. Objections, Replies, and Philosophical Implications References"
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG",
                "cs.NE"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively evaluate and improve the cognitive models used in automated student learning systems?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it can lead to significant advancements in educational technology, particularly in personalized learning. Improved cognitive models can enhance the accuracy of student assessments, leading to tailored educational experiences that cater to individual learning needs. This research could pave the way for more effective teaching strategies, better learning outcomes, and the development of intelligent tutoring systems that adapt to student performance in real-time. Furthermore, it can inspire future research into cognitive modeling and machine learning applications in education, ultimately contributing to a more data-driven approach in pedagogical practices.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the complexity of accurately modeling human cognition and learning processes. Naive approaches may fail because they often overlook the nuances of individual learning styles, the variability in student engagement, and the contextual factors influencing learning. Additionally, there are technical obstacles such as the need for large, diverse datasets to train models effectively, as well as theoretical challenges in understanding the underlying cognitive processes. Practical obstacles include the integration of these models into existing educational frameworks and ensuring they are interpretable and usable by educators.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on isolated aspects of cognitive modeling or has relied on simplistic models that do not capture the full complexity of student learning. Limitations in data availability, computational resources, and the lack of interdisciplinary collaboration between cognitive science and machine learning have hindered progress. Additionally, existing solutions may not have adequately addressed the dynamic nature of learning environments. My approach differs by integrating advanced machine learning techniques with comprehensive cognitive theories, utilizing richer datasets, and emphasizing the adaptability of models to real-world educational contexts.\n\n### [Question 5] - What are the key components of my approach and results?\nMy proposed methodology involves developing a hybrid cognitive model that combines reinforcement learning with cognitive theories of learning. I plan to use a diverse dataset of student interactions from various educational platforms, focusing on metrics such as learning gains, engagement levels, and retention rates. The expected outcomes include a more accurate and dynamic model of student learning that can predict performance and adapt instructional strategies accordingly. Additionally, I aim to create a framework for continuous model improvement based on real-time feedback from student interactions, ultimately leading to enhanced educational tools and resources."
            }
        },
        "author_data": {
            "e69919e0-9c25-47c0-ac0e-dbc98b9c21c5": {
                "pk": "e69919e0-9c25-47c0-ac0e-dbc98b9c21c5",
                "name": "Dingli Yu",
                "collaborators": [
                    "Sanjeev Arora",
                    "Zhiyuan Li",
                    "Anirudh Goyal",
                    "Kaifeng Lyu",
                    "Wei Hu",
                    "Haoyu Zhao",
                    "S. Du",
                    "R. Salakhutdinov",
                    "Ruosong Wang",
                    "Simran Kaur"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Compositional Generalization",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Can Models Learn Skill Composition from Examples?",
                        "abstract": "As large language models (LLMs) become increasingly advanced, their ability to exhibit compositional generalization -- the capacity to combine learned skills in novel ways not encountered during training -- has garnered significant attention. This type of generalization, particularly in scenarios beyond training data, is also of great interest in the study of AI safety and alignment. A recent study introduced the SKILL-MIX evaluation, where models are tasked with composing a short paragraph demonstrating the use of a specified $k$-tuple of language skills. While small models struggled with composing even with $k=3$, larger models like GPT-4 performed reasonably well with $k=5$ and $6$. In this paper, we employ a setup akin to SKILL-MIX to evaluate the capacity of smaller models to learn compositional generalization from examples. Utilizing a diverse set of language skills -- including rhetorical, literary, reasoning, theory of mind, and common sense -- GPT-4 was used to generate text samples that exhibit random subsets of $k$ skills. Subsequent fine-tuning of 7B and 13B parameter models on these combined skill texts, for increasing values of $k$, revealed the following findings: (1) Training on combinations of $k=2$ and $3$ skills results in noticeable improvements in the ability to compose texts with $k=4$ and $5$ skills, despite models never having seen such examples during training. (2) When skill categories are split into training and held-out groups, models significantly improve at composing texts with held-out skills during testing despite having only seen training skills during fine-tuning, illustrating the efficacy of the training approach even with previously unseen skills. This study also suggests that incorporating skill-rich (potentially synthetic) text into training can substantially enhance the compositional capabilities of models."
                    },
                    {
                        "title": "AI-Assisted Generation of Difficult Math Questions",
                        "abstract": "Current LLM training positions mathematical reasoning as a core capability. With publicly available sources fully tapped, there is unmet demand for diverse and challenging math questions. Relying solely on human experts is both time-consuming and costly, while LLM-generated questions often lack the requisite diversity and difficulty. We present a design framework that combines the strengths of LLMs with a human-in-the-loop approach to generate a diverse array of challenging math questions. We leverage LLM metacognition skills [Didolkar et al., 2024] of a strong LLM to extract core\"skills\"from existing math datasets. These skills serve as the basis for generating novel and difficult questions by prompting the LLM with random pairs of core skills. The use of two different skills within each question makes finding such questions an\"out of distribution\"task for both LLMs and humans. Our pipeline employs LLMs to iteratively generate and refine questions and solutions through multiturn prompting. Human annotators then verify and further refine the questions, with their efficiency enhanced via further LLM interactions. Applying this pipeline on skills extracted from the MATH dataset [Hendrycks et al., 2021] resulted in MATH$^2$ - a dataset of higher-quality math questions, as evidenced by: (a) Lower performance of all models on MATH$^2$ than on MATH (b) Higher performance on MATH when using MATH$^2$ questions as in-context examples. Although focused on mathematics, our methodology seems applicable to other domains requiring structured reasoning, and potentially as a component of scalable oversight. Also of interest is a striking relationship observed between models' performance on the new dataset: the success rate on MATH$^2$ is the square on MATH, suggesting that successfully solving the question in MATH$^2$ requires a nontrivial combination of two distinct math skills."
                    },
                    {
                        "title": "Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates",
                        "abstract": "Public LLMs such as the Llama 2-Chat have driven huge activity in LLM research. These models underwent alignment training and were considered safe. Recently Qi et al. (2023) reported that even benign fine-tuning (e.g., on seemingly safe datasets) can give rise to unsafe behaviors in the models. The current paper is about methods and best practices to mitigate such loss of alignment. Through extensive experiments on several chat models (Meta's Llama 2-Chat, Mistral AI's Mistral 7B Instruct v0.2, and OpenAI's GPT-3.5 Turbo), this paper uncovers that the prompt templates used during fine-tuning and inference play a crucial role in preserving safety alignment, and proposes the\"Pure Tuning, Safe Testing\"(PTST) principle -- fine-tune models without a safety prompt, but include it at test time. Fine-tuning experiments on GSM8K, ChatDoctor, and OpenOrca show that PTST significantly reduces the rise of unsafe behaviors, and even almost eliminates them in some cases."
                    },
                    {
                        "title": "ConceptMix: A Compositional Image Generation Benchmark with Controllable Difficulty",
                        "abstract": "Compositionality is a critical capability in Text-to-Image (T2I) models, as it reflects their ability to understand and combine multiple concepts from text descriptions. Existing evaluations of compositional capability rely heavily on human-designed text prompts or fixed templates, limiting their diversity and complexity, and yielding low discriminative power. We propose ConceptMix, a scalable, controllable, and customizable benchmark which automatically evaluates compositional generation ability of T2I models. This is done in two stages. First, ConceptMix generates the text prompts: concretely, using categories of visual concepts (e.g., objects, colors, shapes, spatial relationships), it randomly samples an object and k-tuples of visual concepts, then uses GPT4-o to generate text prompts for image generation based on these sampled concepts. Second, ConceptMix evaluates the images generated in response to these prompts: concretely, it checks how many of the k concepts actually appeared in the image by generating one question per visual concept and using a strong VLM to answer them. Through administering ConceptMix to a diverse set of T2I models (proprietary as well as open ones) using increasing values of k, we show that our ConceptMix has higher discrimination power than earlier benchmarks. Specifically, ConceptMix reveals that the performance of several models, especially open models, drops dramatically with increased k. Importantly, it also provides insight into the lack of prompt diversity in widely-used training datasets. Additionally, we conduct extensive human studies to validate the design of ConceptMix and compare our automatic grading with human judgement. We hope it will guide future T2I model development."
                    },
                    {
                        "title": "New Definitions and Evaluations for Saliency Methods: Staying Intrinsic, Complete and Sound",
                        "abstract": "Saliency methods compute heat maps that highlight portions of an input that were most {\\em important} for the label assigned to it by a deep net. Evaluations of saliency methods convert this heat map into a new {\\em masked input} by retaining the $k$ highest-ranked pixels of the original input and replacing the rest with \\textquotedblleft uninformative\\textquotedblright\\ pixels, and checking if the net's output is mostly unchanged. This is usually seen as an {\\em explanation} of the output, but the current paper highlights reasons why this inference of causality may be suspect. Inspired by logic concepts of {\\em completeness \\&soundness}, it observes that the above type of evaluation focuses on completeness of the explanation, but ignores soundness. New evaluation metrics are introduced to capture both notions, while staying in an {\\em intrinsic} framework -- i.e., using the dataset and the net, but no separately trained nets, human evaluations, etc. A simple saliency method is described that matches or outperforms prior methods in the evaluations. Experiments also suggest new intrinsic justifications, based on soundness, for popular heuristic tricks such as TV regularization and upsampling."
                    },
                    {
                        "title": "A Kernel-Based View of Language Model Fine-Tuning",
                        "abstract": "It has become standard to solve NLP tasks by fine-tuning pre-trained language models (LMs), especially in low-data settings. There is minimal theoretical understanding of empirical success, e.g., why fine-tuning a model with $10^8$ or more parameters on a couple dozen training points does not result in overfitting. We investigate whether the Neural Tangent Kernel (NTK) - which originated as a model to study the gradient descent dynamics of infinitely wide networks with suitable random initialization - describes fine-tuning of pre-trained LMs. This study was inspired by the decent performance of NTK for computer vision tasks (Wei et al., 2022). We extend the NTK formalism to Adam and use Tensor Programs (Yang, 2020) to characterize conditions under which the NTK lens may describe fine-tuning updates to pre-trained language models. Extensive experiments on 14 NLP tasks validate our theory and show that formulating the downstream task as a masked word prediction problem through prompting often induces kernel-based dynamics during fine-tuning. Finally, we use this kernel view to propose an explanation for the success of parameter-efficient subspace-based fine-tuning methods."
                    },
                    {
                        "title": "Fast Mixing of Stochastic Gradient Descent with Normalization and Weight Decay",
                        "abstract": "We prove the Fast Equilibrium Conjecture proposed by Li et al. [1], i.e. , stochastic gradient descent (SGD) on a scale-invariant loss ( e.g. , using networks with various normalization schemes) with learning rate \u2318 and weight decay factor \ufffd mixes in function space in e O (1 / ( \u2318\ufffd )) steps, under two standard assumptions: (1) the noise covariance matrix is non-degenerate and (2) the minimizers of the loss form a connected, compact and analytic manifold. The analysis uses the framework of Li et al. [2] and shows that for every T > 0 , the iterates of SGD with learning rate \u2318 and weight decay factor \ufffd on the scale-invariant loss converge in distribution in ln(1 + T \ufffd / \u2318 ) / (4 \u2318\ufffd ) iterations as \u2318\ufffd ! 0 while satisfying \u2318 \uf8ff O ( \ufffd ) \uf8ff O (1) . Moreover, the evolution of the limiting distribution can be described by a stochastic differential equation that mixes to the same equilibrium distribution for every initialization around the manifold of minimizers as T ! 1 ."
                    },
                    {
                        "title": "Understanding Generalization of Deep Neural Networks Trained with Noisy Labels",
                        "abstract": "Over-parameterized deep neural networks trained by simple first-order methods are known to be able to fit any labeling of data. When the training dataset contains a fraction of noisy labels, can neural networks be resistant to over-fitting and still generalize on the true distribution? Inspired by recent theoretical work that established connections between over-parameterized neural networks and neural tangent kernel (NTK), we propose two simple regularization methods for this purpose: (i) regularization by the distance between the network parameters to initialization, and (ii) adding a trainable auxiliary variable to the network output for each training example. Theoretically, both methods are related to kernel ridge regression with respect to the NTK, and we prove their generalization guarantee on the true data distribution despite being trained using noisy labels. The generalization bound is independent of the network size, and only depends on the training inputs and true labels (instead of noisy labels) as well as the noise level in the labels. Empirical results verify the effectiveness of these methods on noisily labeled datasets."
                    },
                    {
                        "title": "Harnessing the Power of Infinitely Wide Deep Nets on Small-data Tasks",
                        "abstract": "Recent research shows that the following two models are equivalent: (a) infinitely wide neural networks (NNs) trained under l2 loss by gradient descent with infinitesimally small learning rate (b) kernel regression with respect to so-called Neural Tangent Kernels (NTKs) (Jacot et al., 2018). An efficient algorithm to compute the NTK, as well as its convolutional counterparts, appears in Arora et al. (2019a), which allowed studying performance of infinitely wide nets on datasets like CIFAR-10. However, super-quadratic running time of kernel methods makes them best suited for small-data tasks. We report results suggesting neural tangent kernels perform strongly on low-data tasks.  1. On a standard testbed of classification/regression tasks from the UCI database, NTK SVM beats the previous gold standard, Random Forests (RF), and also the corresponding finite nets.  2. On CIFAR-10 with 10 - 640 training samples, Convolutional NTK consistently beats ResNet-34 by 1% - 3%.  3. On VOC07 testbed for few-shot image classification tasks on ImageNet with transfer learning (Goyal et al., 2019), replacing the linear SVM currently used with a Convolutional NTK SVM consistently improves performance.  4. Comparing the performance of NTK with the finite-width net it was derived from, NTK behavior starts at lower net widths than suggested by theoretical analysis(Arora et al., 2019a). NTK's efficacy may trace to lower variance of output."
                    },
                    {
                        "title": "Enhanced Convolutional Neural Tangent Kernels",
                        "abstract": "Recent research shows that for training with $\\ell_2$ loss, convolutional neural networks (CNNs) whose width (number of channels in convolutional layers) goes to infinity correspond to regression with respect to the CNN Gaussian Process kernel (CNN-GP) if only the last layer is trained, and correspond to regression with respect to the Convolutional Neural Tangent Kernel (CNTK) if all layers are trained. An exact algorithm to compute CNTK (Arora et al., 2019) yielded the finding that classification accuracy of CNTK on CIFAR-10 is within 6-7% of that of that of the corresponding CNN architecture (best figure being around 78%) which is interesting performance for a fixed kernel. Here we show how to significantly enhance the performance of these kernels using two ideas. (1) Modifying the kernel using a new operation called Local Average Pooling (LAP) which preserves efficient computability of the kernel and inherits the spirit of standard data augmentation using pixel shifts. Earlier papers were unable to incorporate naive data augmentation because of the quadratic training cost of kernel regression. This idea is inspired by Global Average Pooling (GAP), which we show for CNN-GP and CNTK is equivalent to full translation data augmentation. (2) Representing the input image using a pre-processing technique proposed by Coates et al. (2011), which uses a single convolutional layer composed of random image patches. On CIFAR-10, the resulting kernel, CNN-GP with LAP and horizontal flip data augmentation, achieves 89% accuracy, matching the performance of AlexNet (Krizhevsky et al., 2012). Note that this is the best such result we know of for a classifier that is not a trained neural network. Similar improvements are obtained for Fashion-MNIST."
                    },
                    {
                        "title": "Simple and Effective Regularization Methods for Training on Noisily Labeled Data with Generalization Guarantee",
                        "abstract": "Over-parameterized deep neural networks trained by simple first-order methods are known to be able to fit any labeling of data. Such over-fitting ability hinders generalization when mislabeled training examples are present. On the other hand, simple regularization methods like early-stopping can often achieve highly nontrivial performance on clean test data in these scenarios, a phenomenon not theoretically understood. This paper proposes and analyzes two simple and intuitive regularization methods: (i) regularization by the distance between the network parameters to initialization, and (ii) adding a trainable auxiliary variable to the network output for each training example. Theoretically, we prove that gradient descent training with either of these two methods leads to a generalization guarantee on the clean data distribution despite being trained using noisy labels. Our generalization analysis relies on the connection between wide neural network and neural tangent kernel (NTK). The generalization bound is independent of the network size, and is comparable to the bound one can get when there is no label noise. Experimental results verify the effectiveness of these methods on noisily labeled datasets."
                    }
                ]
            },
            "fe802f2b-fff3-40d3-b308-78baa8e504c5": {
                "pk": "fe802f2b-fff3-40d3-b308-78baa8e504c5",
                "name": "Simran Kaur",
                "collaborators": [
                    "Anirudh Goyal",
                    "Sanjeev Arora",
                    "Haoyu Zhao",
                    "Dingli Yu",
                    "Simon Park"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Compositional Generalization",
                    "Instruction Following",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "Can Models Learn Skill Composition from Examples?",
                        "abstract": "As large language models (LLMs) become increasingly advanced, their ability to exhibit compositional generalization -- the capacity to combine learned skills in novel ways not encountered during training -- has garnered significant attention. This type of generalization, particularly in scenarios beyond training data, is also of great interest in the study of AI safety and alignment. A recent study introduced the SKILL-MIX evaluation, where models are tasked with composing a short paragraph demonstrating the use of a specified $k$-tuple of language skills. While small models struggled with composing even with $k=3$, larger models like GPT-4 performed reasonably well with $k=5$ and $6$. In this paper, we employ a setup akin to SKILL-MIX to evaluate the capacity of smaller models to learn compositional generalization from examples. Utilizing a diverse set of language skills -- including rhetorical, literary, reasoning, theory of mind, and common sense -- GPT-4 was used to generate text samples that exhibit random subsets of $k$ skills. Subsequent fine-tuning of 7B and 13B parameter models on these combined skill texts, for increasing values of $k$, revealed the following findings: (1) Training on combinations of $k=2$ and $3$ skills results in noticeable improvements in the ability to compose texts with $k=4$ and $5$ skills, despite models never having seen such examples during training. (2) When skill categories are split into training and held-out groups, models significantly improve at composing texts with held-out skills during testing despite having only seen training skills during fine-tuning, illustrating the efficacy of the training approach even with previously unseen skills. This study also suggests that incorporating skill-rich (potentially synthetic) text into training can substantially enhance the compositional capabilities of models."
                    },
                    {
                        "title": "Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning",
                        "abstract": "We introduce Instruct-SkillMix, an automated approach for creating diverse, high quality SFT data. The Instruct-SkillMix pipeline involves two stages, each leveraging an existing powerful LLM: (1) Skill extraction: uses the LLM to extract core\"skills\"for instruction-following, either from existing datasets, or by directly prompting the model; (2) Data generation: uses the powerful LLM to generate (instruction, response) data that exhibit a randomly chosen pair of these skills. Here, the use of random skill combinations promotes diversity and difficulty. Vanilla SFT (i.e., no PPO, DPO, or RL methods) on data generated from Instruct-SkillMix leads to strong gains on instruction following benchmarks such as AlpacaEval 2.0, MT-Bench, and WildBench. With just $4$K examples, LLaMA-3-8B-Base achieves 42.76% length-controlled win rate on AlpacaEval 2.0. To our knowledge, this achieves state-of-the-art performance among all models that have only undergone SFT (no RL methods) and competes with proprietary models such as Claude 3 Opus and LLaMA-3.1-405B-Instruct. Ablation studies also suggest plausible reasons for why creating open instruction-tuning datasets via naive crowd-sourcing has proved difficult. Introducing low quality answers (\"shirkers\") in $20\\%$ of Instruct-SkillMix examples causes performance to plummet, sometimes catastrophically. The Instruct-SkillMix pipeline is flexible and is adaptable to other settings."
                    }
                ]
            },
            "e598fbee-cb90-4fc1-9fde-1c86cd42a3e2": {
                "pk": "e598fbee-cb90-4fc1-9fde-1c86cd42a3e2",
                "name": "Arushi Gupta",
                "collaborators": [
                    "Nikunj Saunshi",
                    "Sanjeev Arora",
                    "J. Z. Matilla",
                    "Daniel J. Hsu",
                    "Z. Haiman",
                    "Udaya Ghai",
                    "Wenhan Xia",
                    "Karan Singh",
                    "Elad Hazan",
                    "Siddharth Gupta"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Computer Vision",
                    "Deep Learning",
                    "Cosmology"
                ],
                "publications": [
                    {
                        "title": "Online Nonstochastic Model-Free Reinforcement Learning",
                        "abstract": "We investigate robust model-free reinforcement learning algorithms designed for environments that may be dynamic or even adversarial. Traditional state-based policies often struggle to accommodate the challenges imposed by the presence of unmodeled disturbances in such settings. Moreover, optimizing linear state-based policies pose an obstacle for efficient optimization, leading to nonconvex objectives, even in benign environments like linear dynamical systems. Drawing inspiration from recent advancements in model-based control, we introduce a novel class of policies centered on disturbance signals. We define several categories of these signals, which we term pseudo-disturbances, and develop corresponding policy classes based on them. We provide efficient and practical algorithms for optimizing these policies. Next, we examine the task of online adaptation of reinforcement learning agents in the face of adversarial disturbances. Our methods seamlessly integrate with any black-box model-free approach, yielding provable regret guarantees when dealing with linear dynamics. These regret guarantees unconditionally improve the best-known results for bandit linear control in having no dependence on the state-space dimension. We evaluate our method over various standard RL benchmarks and demonstrate improved robustness."
                    },
                    {
                        "title": "Reimagining Carbon Nanomaterial Analysis: Empowering Transfer Learning and Machine Vision in Scanning Electron Microscopy for High-Fidelity Identification",
                        "abstract": "In this report, we propose a novel technique for identifying and analyzing diverse nanoscale carbon allotropes using scanning electron micrographs. By precisely controlling the quenching rates of undercooled molten carbon through laser irradiation, we achieved the formation of microdiamonds, nanodiamonds, and Q-carbon films. However, standard laser irradiation without proper undercooling control leads to the formation of sparsely located diverse carbon polymorphs, hindering their discovery and classification through manual analyses. To address this challenge, we applied transfer-learning approaches using convolutional neural networks and computer vision techniques to achieve allotrope discovery even with sparse spatial presence. Our method achieved high accuracy rates of 92% for Q-carbon identification and 94% for distinguishing it from nanodiamonds. By leveraging scanning electron micrographs and precise undercooling control, our technique enables the efficient identification and characterization of nanoscale carbon structures. This research significantly contributes to the advancement of the field, providing automated tools for Q-materials and carbon polymorph identification. It opens up new opportunities for the further exploration of these materials in various applications."
                    },
                    {
                        "title": "Understanding Influence Functions and Datamodels via Harmonic Analysis",
                        "abstract": "Influence functions estimate effect of individual data points on predictions of the model on test data and were adapted to deep learning in Koh and Liang [2017]. They have been used for detecting data poisoning, detecting helpful and harmful examples, influence of groups of datapoints, etc. Recently, Ilyas et al. [2022] introduced a linear regression method they termed datamodels to predict the effect of training points on outputs on test data. The current paper seeks to provide a better theoretical understanding of such interesting empirical phenomena. The primary tool is harmonic analysis and the idea of noise stability. Contributions include: (a) Exact characterization of the learnt datamodel in terms of Fourier coefficients. (b) An efficient method to estimate the residual error and quality of the optimum linear datamodel without having to train the datamodel. (c) New insights into when influences of groups of datapoints may or may not add up linearly."
                    },
                    {
                        "title": "New Definitions and Evaluations for Saliency Methods: Staying Intrinsic, Complete and Sound",
                        "abstract": "Saliency methods compute heat maps that highlight portions of an input that were most {\\em important} for the label assigned to it by a deep net. Evaluations of saliency methods convert this heat map into a new {\\em masked input} by retaining the $k$ highest-ranked pixels of the original input and replacing the rest with \\textquotedblleft uninformative\\textquotedblright\\ pixels, and checking if the net's output is mostly unchanged. This is usually seen as an {\\em explanation} of the output, but the current paper highlights reasons why this inference of causality may be suspect. Inspired by logic concepts of {\\em completeness \\&soundness}, it observes that the above type of evaluation focuses on completeness of the explanation, but ignores soundness. New evaluation metrics are introduced to capture both notions, while staying in an {\\em intrinsic} framework -- i.e., using the dataset and the net, but no separately trained nets, human evaluations, etc. A simple saliency method is described that matches or outperforms prior methods in the evaluations. Experiments also suggest new intrinsic justifications, based on soundness, for popular heuristic tricks such as TV regularization and upsampling."
                    },
                    {
                        "title": "On Predicting Generalization using GANs",
                        "abstract": "Research on generalization bounds for deep networks seeks to give ways to predict test error using just the training dataset and the network parameters. While generalization bounds can give many insights about architecture design, training algorithms, etc., what they do not currently do is yield good predictions for actual test error. A recently introduced Predicting Generalization in Deep Learning competition~\\citep{jiang2020neurips} aims to encourage discovery of methods to better predict test error. The current paper investigates a simple idea: can test error be predicted using {\\em synthetic data,} produced using a Generative Adversarial Network (GAN) that was trained on the same training dataset? Upon investigating several GAN models and architectures, we find that this turns out to be the case. In fact, using GANs pre-trained on standard datasets, the test error can be predicted without requiring any additional hyper-parameter tuning. This result is surprising because GANs have well-known limitations (e.g. mode collapse) and are known to not learn the data distribution accurately. Yet the generated samples are good enough to substitute for test data. Several additional experiments are presented to explore reasons why GANs do well at this task. In addition to a new approach for predicting generalization, the counter-intuitive phenomena presented in our work may also call for a better understanding of GANs' strengths and limitations."
                    },
                    {
                        "title": "A Representation Learning Perspective on the Importance of Train-Validation Splitting in Meta-Learning",
                        "abstract": "An effective approach in meta-learning is to utilize multiple\"train tasks\"to learn a good initialization for model parameters that can help solve unseen\"test tasks\"with very few samples by fine-tuning from this initialization. Although successful in practice, theoretical understanding of such methods is limited. This work studies an important aspect of these methods: splitting the data from each task into train (support) and validation (query) sets during meta-training. Inspired by recent work (Raghu et al., 2020), we view such meta-learning methods through the lens of representation learning and argue that the train-validation split encourages the learned representation to be low-rank without compromising on expressivity, as opposed to the non-splitting variant that encourages high-rank representations. Since sample efficiency benefits from low-rankness, the splitting strategy will require very few samples to solve unseen test tasks. We present theoretical results that formalize this idea for linear representation learning on a subspace meta-learning instance, and experimentally verify this practical benefit of splitting in simulations and on standard meta-learning benchmarks."
                    },
                    {
                        "title": "Inherent Noise in Gradient Based Methods",
                        "abstract": "Previous work has examined the ability of larger capacity neural networks to generalize better than smaller ones, even without explicit regularizers, by analyzing gradient based algorithms such as GD and SGD. The presence of noise and its effect on robustness to parameter perturbations has been linked to generalization. We examine a property of GD and SGD, namely that instead of iterating through all scalar weights in the network and updating them one by one, GD (and SGD) updates all the parameters at the same time. As a result, each parameter $w^i$ calculates its partial derivative at the stale parameter $\\mathbf{w_t}$, but then suffers loss $\\hat{L}(\\mathbf{w_{t+1}})$. We show that this causes noise to be introduced into the optimization. We find that this noise penalizes models that are sensitive to perturbations in the weights. We find that penalties are most pronounced for batches that are currently being used to update, and are higher for larger models."
                    },
                    {
                        "title": "A Simple Saliency Method That Passes the Sanity Checks",
                        "abstract": "There is great interest in \"saliency methods\" (also called \"attribution methods\"), which give \"explanations\" for a deep net's decision, by assigning a \"score\" to each feature/pixel in the input. Their design usually involves credit-assignment via the gradient of the output with respect to input. Recently Adebayo et al. [arXiv:1810.03292] questioned the validity of many of these methods since they do not pass simple *sanity checks* which test whether the scores shift/vanish when layers of the trained net are randomized, or when the net is retrained using random labels for inputs.  We propose a simple fix to existing saliency methods that helps them pass sanity checks, which we call \"competition for pixels\". This involves computing saliency maps for all possible labels in the classification task, and using a simple competition among them to identify and remove less relevant pixels from the map. The simplest variant of this is \"Competitive Gradient $\\odot$ Input (CGI)\": it is efficient, requires no additional training, and uses only the input and gradient. Some theoretical justification is provided for it (especially for ReLU networks) and its performance is empirically demonstrated."
                    },
                    {
                        "title": "Non-Gaussian information from weak lensing data via deep learning",
                        "abstract": "Weak lensing maps contain information beyond two-point statistics on small scales. Much recent work has tried to extract this information through a range of different observables or via nonlinear transformations of the lensing field. Here we train and apply a two-dimensional convolutional neural network to simulated noiseless lensing maps covering 96 different cosmological models over a range of ${{\\mathrm{\\ensuremath{\\Omega}}}_{m},{\\ensuremath{\\sigma}}_{8}}$. Using the area of the confidence contour in the ${{\\mathrm{\\ensuremath{\\Omega}}}_{m},{\\ensuremath{\\sigma}}_{8}}$ plane as a figure of merit, derived from simulated convergence maps smoothed on a scale of 1.0 arcmin, we show that the neural network yields $\\ensuremath{\\approx}5\\ifmmode\\times\\else\\texttimes\\fi{}$ tighter constraints than the power spectrum, and $\\ensuremath{\\approx}4\\ifmmode\\times\\else\\texttimes\\fi{}$ tighter than the lensing peaks. Such gains illustrate the extent to which weak lensing data encode cosmological information not accessible to the power spectrum or even other, non-Gaussian statistics such as lensing peaks."
                    },
                    {
                        "title": "Do dark matter halos explain lensing peaks",
                        "abstract": "We have investigated a recently proposed halo-based model, Camelus, for predicting weak-lensing peak counts, and compared its results over a collection of 162 cosmologies with those from N-body simulations. While counts from both models agree for peaks with $\\mathcal{S/N}>1$ (where $\\mathcal{S/N}$ is the ratio of the peak height to the r.m.s. shape noise), we find $\\approx 50\\%$ fewer counts for peaks near $\\mathcal{S/N}=0$ and significantly higher counts in the negative $\\mathcal{S/N}$ tail. Adding shape noise reduces the differences to within $20\\%$ for all cosmologies. We also found larger covariances that are more sensitive to cosmological parameters. As a result, credibility regions in the $\\{\\Omega_m, \\sigma_8\\}$ are $\\approx 30\\%$ larger. Even though the credible contours are commensurate, each model draws its predictive power from different types of peaks. Low peaks, especially those with $2 3)$. Our results confirm the importance of using a cosmology-dependent covariance with at least a 14\\% improvement in parameter constraints. We identified the covariance estimation as the main driver behind differences in inference, and suggest possible ways to make Camelus even more useful as a highly accurate peak count emulator."
                    }
                ]
            },
            "4f1d688c-8c18-4b64-b84e-c81df947a0c4": {
                "pk": "4f1d688c-8c18-4b64-b84e-c81df947a0c4",
                "name": "Jonah Brown-Cohen",
                "collaborators": [
                    "P. Raghavendra",
                    "Zachary Kenton",
                    "Noah Y. Siegel",
                    "J'anos Kram'ar",
                    "Samuel Albanie",
                    "Jannis Bulian",
                    "Rishabh Agarwal",
                    "David Lindner",
                    "Yunhao Tang",
                    "Noah D. Goodman"
                ],
                "domain": [
                    "AI Safety",
                    "Reinforcement Learning",
                    "Cryptography",
                    "Computational Topology"
                ],
                "publications": [
                    {
                        "title": "On scalable oversight with weak LLMs judging strong LLMs",
                        "abstract": "Scalable oversight protocols aim to enable humans to accurately supervise superhuman AI. In this paper we study debate, where two AI's compete to convince a judge; consultancy, where a single AI tries to convince a judge that asks questions; and compare to a baseline of direct question-answering, where the judge just answers outright without the AI. We use large language models (LLMs) as both AI agents and as stand-ins for human judges, taking the judge models to be weaker than agent models. We benchmark on a diverse range of asymmetries between judges and agents, extending previous work on a single extractive QA task with information asymmetry, to also include mathematics, coding, logic and multimodal reasoning asymmetries. We find that debate outperforms consultancy across all tasks when the consultant is randomly assigned to argue for the correct/incorrect answer. Comparing debate to direct question answering, the results depend on the type of task: in extractive QA tasks with information asymmetry debate outperforms direct question answering, but in other tasks without information asymmetry the results are mixed. Previous work assigned debaters/consultants an answer to argue for. When we allow them to instead choose which answer to argue for, we find judges are less frequently convinced by the wrong answer in debate than in consultancy. Further, we find that stronger debater models increase judge accuracy, though more modestly than in previous studies."
                    },
                    {
                        "title": "Scalable AI Safety via Doubly-Efficient Debate",
                        "abstract": "The emergence of pre-trained AI systems with powerful capabilities across a diverse and ever-increasing set of complex domains has raised a critical challenge for AI safety as tasks can become too complicated for humans to judge directly. Irving et al. [2018] proposed a debate method in this direction with the goal of pitting the power of such AI models against each other until the problem of identifying (mis)-alignment is broken down into a manageable subtask. While the promise of this approach is clear, the original framework was based on the assumption that the honest strategy is able to simulate deterministic AI systems for an exponential number of steps, limiting its applicability. In this paper, we show how to address these challenges by designing a new set of debate protocols where the honest strategy can always succeed using a simulation of a polynomial number of steps, whilst being able to verify the alignment of stochastic AI systems, even when the dishonest strategy is allowed to use exponentially many simulation steps."
                    },
                    {
                        "title": "Detecting Adversarial Directions in Deep Reinforcement Learning to Make Robust Decisions",
                        "abstract": "Learning in MDPs with highly complex state representations is currently possible due to multiple advancements in reinforcement learning algorithm design. However, this incline in complexity, and furthermore the increase in the dimensions of the observation came at the cost of volatility that can be taken advantage of via adversarial attacks (i.e. moving along worst-case directions in the observation space). To solve this policy instability problem we propose a novel method to detect the presence of these non-robust directions via local quadratic approximation of the deep neural policy loss. Our method provides a theoretical basis for the fundamental cut-off between safe observations and adversarial observations. Furthermore, our technique is computationally efficient, and does not depend on the methods used to produce the worst-case directions. We conduct extensive experiments in the Arcade Learning Environment with several different adversarial attack techniques. Most significantly, we demonstrate the effectiveness of our approach even in the setting where non-robust directions are explicitly optimized to circumvent our proposed method."
                    },
                    {
                        "title": "Faster Algorithms and Constant Lower Bounds for the Worst-Case Expected Error",
                        "abstract": "The study of statistical estimation without distributional assumptions on data values, but with knowledge of data collection methods was recently introduced by Chen, Valiant and Valiant (NeurIPS 2020). In this framework, the goal is to design estimators that minimize the worst-case expected error. Here the expectation is over a known, randomized data collection process from some population, and the data values corresponding to each element of the population are assumed to be worst-case. Chen, Valiant and Valiant show that, when data values are (cid:96) \u221e -normalized, there is a polynomial time algorithm to compute an estimator for the mean with worst-case expected error that is within a factor \u03c0 2 of the optimum within the natural class of semilinear estimators. However, their algorithm is based on optimizing a somewhat complex concave objective function over a constrained set of positive semide\ufb01nite matrices, and thus does not come with explicit runtime guarantees beyond being polynomial time in the input. In this paper we design provably ef\ufb01cient algorithms for approximating the optimal semilinear estimator based on online convex optimization. In the setting where data values are (cid:96) \u221e -normalized, our algorithm achieves a \u03c0 2 -approximation by iteratively solving a sequence of standard SDPs. When data values are (cid:96) 2 -normalized, our algorithm iteratively computes the top eigenvector of a sequence of matrices, and does not lose any multiplicative approximation factor. Further, using experiments in settings where sample membership is correlated with data values (e.g. \"importance sampling\" and \"snowball sampling\"), we show that our (cid:96) 2 -normalized algorithm gives a similar advantage over standard estimators as the original (cid:96) \u221e -normalized algorithm of Chen, Valiant and Valiant, but with much lower computational complexity. We complement these positive results by stating a simple combinatorial condition which, if satis\ufb01ed by a data collection process, implies that any (not necessarily semilinear) estimator for the mean has constant worst-case expected error."
                    },
                    {
                        "title": "Optimal Inapproximability with Universal Factor Graphs",
                        "abstract": "The factor graph of an instance of a constraint satisfaction problem (CSP) is the bipartite graph indicating which variables appear in each constraint. An instance of the CSP is given by the factor graph together with a list of which predicate is applied for each constraint. We establish that many Max-CSPs remain as hard to approximate as in the general case even when the factor graph is fixed (depending only on the size of the instance) and known in advance. Examples of results obtained for this restricted setting are: (1) Optimal inapproximability for Max-3-Lin and Max-3-Sat (H\u00e5stad, J. ACM 2001).(2) Approximation resistance for predicates supporting pairwise independent subgroups (Chan, J. ACM 2016).(3) Hardness of the \u201c(2 + \u03f5)-Sat\u201d problem and other Promise CSPs (Austrin et al., SIAM J. Comput. 2017). The main technical tool used to establish these results is a new way of folding the long code which we call \u201cfunctional folding\u201d."
                    },
                    {
                        "title": "Extended Formulation Lower Bounds for Refuting Random CSPs",
                        "abstract": "Random constraint satisfaction problems (CSPs) such as random $3$-SAT are conjectured to be computationally intractable. The average case hardness of random $3$-SAT and other CSPs has broad and far-reaching implications on problems in approximation, learning theory and cryptography.  In this work, we show subexponential lower bounds on the size of linear programming relaxations for refuting random instances of constraint satisfaction problems. Formally, suppose $P : \\{0,1\\}^k \\to \\{0,1\\}$ is a predicate that supports a $t-1$-wise uniform distribution on its satisfying assignments. Consider the distribution of random instances of CSP $P$ with $m = \\Delta n$ constraints. We show that any linear programming extended formulation that can refute instances from this distribution with constant probability must have size at least $\\Omega\\left(\\exp\\left(\\left(\\frac{n^{t-2}}{\\Delta^2}\\right)^{\\frac{1-\\nu}{k}}\\right)\\right)$ for all $\\nu > 0$. For example, this yields a lower bound of size $\\exp(n^{1/3})$ for random $3$-SAT with a linear number of clauses.  We use the technique of pseudocalibration to directly obtain extended formulation lower bounds from the planted distribution. This approach bypasses the need to construct Sherali-Adams integrality gaps in proving general LP lower bounds. As a corollary, one obtains a self-contained proof of subexponential Sherali-Adams LP lower bounds for these problems. We believe the result sheds light on the technique of pseudocalibration, a promising but conjectural approach to LP/SDP lower bounds."
                    },
                    {
                        "title": "Formal Barriers to Longest-Chain Proof-of-Stake Protocols",
                        "abstract": "The security of most existing cryptocurrencies is based on a concept called Proof-of-Work, in which users must solve a computationally hard cryptopuzzle to authorize transactions (\"one unit of computation, one vote''). This leads to enormous expenditure on hardware and electricity in order to collect the rewards associated with transaction authorization. Proof-of-Stake is an alternative concept that instead selects users to authorize transactions proportional to their wealth (\"one coin, one vote\"). Some aspects of the two paradigms are the same. For instance, obtaining voting power in Proof-of-Stake has a monetary cost just as in Proof-of-Work: a coin cannot be freely duplicated any more easily than a unit of computation. However some aspects are fundamentally different. In particular, exactly because Proof-of-Stake is wasteless, there is no inherent resource cost to deviating (commonly referred to as the \"Nothing-at-Stake'' problem). In contrast to prior work, we focus on incentive-driven deviations (any participant will deviate if doing so yields higher revenue) instead of adversarial corruption (an adversary may take over a significant fraction of the network, but the remaining players follow the protocol). The main results of this paper are several formal barriers to designing incentive-compatible proof-of-stake cryptocurrencies (that don't apply to proof-of-work)."
                    },
                    {
                        "title": "Extended Formulation Lower Bounds for Combinatorial Optimization",
                        "abstract": "Author(s): Brown-Cohen, Jonah | Advisor(s): Raghavendra, Prasad | Abstract: Linear and semidefinite programs are fundamental algorithmic tools, often providing conjecturallyoptimal results for a variety of combinatorial optimization problems. Thus, a naturalquestion is to understand the limitations of linear and semidefinite programming relaxations.In particular, the goal is to prove unconditional lower bounds on the size of any linear orsemidefinite programming relaxation for a given problem.In this dissertation, I will give two results of this flavor. First, I will show that any linearprogramming relaxation for refuting random instances of constraint satisfaction problems(e.g. k-SAT) requires super-polynomial size. This theorem can be understood as evidencethat refuting CSPs is hard, since it rules out a broad class of algorithms. Second, I willshow that any symmetric semidefinite programming relaxation for the matching problemin general graphs requires exponential size. Since there is a polynomial time algorithm forthe matching problem, this result provides an example of the limitations of semidefiniteprogramming relaxations."
                    },
                    {
                        "title": "Correlation Decay and Tractability of CSPs",
                        "abstract": "The algebraic dichotomy conjecture of Bulatov, Krokhin and Jeavons yields an elegant characterization of the complexity of constraint satisfaction problems. Roughly speaking, the characterization asserts that a CSP L is tractable if and only if there exist certain non-trivial operations known as polymorphisms to combine solutions to L to create new ones.    In this work, we study the dynamical system associated with repeated applications of a polymorphism to a distribution over assignments. Specifically, we exhibit a correlation decay phenomenon that makes two variables or groups of variables that are not perfectly correlated become independent after repeated applications of a polymorphism.    We show that this correlation decay phenomenon can be utilized in designing algorithms for CSPs by exhibiting two applications:    1. A simple randomized algorithm to solve linear equations over a prime field, whose analysis crucially relies on correlation decay.    2. A sufficient condition for the simple linear programming relaxation for a 2-CSP to be sound (have no integrality gap) on a given instance."
                    },
                    {
                        "title": "Combinatorial Optimization Algorithms via Polymorphisms",
                        "abstract": "An elegant characterization of the complexity of constraint satisfaction problems has emerged in the form of the the algebraic dichotomy conjecture of [BKJ00]. Roughly speaking, the characterization asserts that a CSP {\\Lambda} is tractable if and only if there exist certain non-trivial operations known as polymorphisms to combine solutions to {\\Lambda} to create new ones. In an entirely separate line of work, the unique games conjecture yields a characterization of approximability of Max-CSPs. Surprisingly, this characterization for Max-CSPs can also be reformulated in the language of polymorphisms.  In this work, we study whether existence of non-trivial polymorphisms implies tractability beyond the realm of constraint satisfaction problems, namely in the value-oracle model. Specifically, given a function f in the value-oracle model along with an appropriate operation that never increases the value of f , we design algorithms to minimize f . In particular, we design a randomized algorithm to minimize a function f that admits a fractional polymorphism which is measure preserving and has a transitive symmetry.  We also reinterpret known results on MaxCSPs and thereby reformulate the unique games conjecture as a characterization of approximability of max-CSPs in terms of their approximate polymorphisms."
                    },
                    {
                        "title": "An Optimization Layer for Distributed Matrix Computations",
                        "abstract": "Distributed matrix computations are essential in many big data applications. Algorithms for problems such as Noisy Matrix Factorization have applications in many areas such as recommender systems (i.e. Netflix), succinct data storage, clustering, and more. For large matrices, these algorithms quickly become computationally expensive, and it is necessary to allocate computational resources (such as processors and amount of time spent) wisely. In addition, these algorithms often take many input parameters, and these parameters influence the accuracy of the outcome or the total runtime. Though it is possible to determine the best parameters empirically, this is both computationally expensive and prone to error. In addition, different datasets may respond differently to parameter settings, and parameter settings that were good for a problem instance may become suboptimal as time progresses and the input data changes. In this work we describe a general system that, given an input problem, a distributed algorithm, and a computational budget, automates parameter choices for distributed matrix computations. We implement this optimizer, focusing specifically on automating the choice of degree of parallelism. As a trial application, we use the optimizer on the Divide-Factor-Combine algorithm (DFC), a recentlydeveloped matrix factorization algorithm. We find that running the algorithm with the optimizer-chosen parameters yields errors that are as low as possible given the time and monetary budgets imposed."
                    },
                    {
                        "title": "Evidence that the Diffie-Hellman Problem is as Hard as Computing Discrete Logs",
                        "abstract": "The Diffie-Hellman protocol was one of the first methods discovered for two people, say Alice and Bob, to agree on a shared secret key despite the fact that any of the messages sent between the two of them might be intercepted. That is, the goal of the protocol is to make it impossible to determine the shared secret key for some third party, say Eve, who sees all the communication between Alice and Bob. Let G = \u3008g\u3009 be a cyclic group with generator g. To perform the Diffie-Hellman protocol, first Alice chooses a secret number a < |G|, and Bob chooses a secret b < |G|. Then Alice sends ga to Bob, and Bob sends gb to Alice. Now, Alice computes (gb)a = gab, and Bob computes (ga)b = gab. Thus, the two of them have agreed on a shared secret group element gab \u2208 G. The question then is: is it possible for the evesdropper Eve to determine gab from the communications sent between Alice and Bob? Note that there are only two messages sent between Alice and Bob, namely the two group elements ga and gb. Thus the problem becomes: given {ga, gb} compute gab. This is the Diffie-Hellman problem, and the assumption that it is hard (in the sense that no efficient algorithm exists) is central in many cryptographic protocols. One of the reasons for this assumption has to do with the relationship of the Diffie-Hellman problem to the problem of computing discrete logarithms in a cyclic group G. Note that if it were possible to efficiently compute the discrete logarithm a of ga, then an attacker could easily solve the Diffie-Hellman problem by first computing a from ga, and then calculating (gb)a = gab. Thus, the Discrete Log problem is at least as hard as the Diffie-Hellman problem. The other direction of this relationship i.e. whether the Diffie-Hellman problem is as hard as the Discrete Log problem, is a fundamental open question in cryptography. Since the Discrete Log problem is generally thought to be hard, a reduction from Discrete Log to Diffie-Hellman would give strong evidence that the Diffie-Hellman protocol is secure. One of the first steps toward such a reduction was made by den Boer in [1]. He showed that for primes p satisfying a certain condition, there is a reduction from Discrete Log to Diffie-Hellman in the group Z\u2217 p . The condition required was that \u03c6(p \u2212 1) had only small prime factors. Here \u03c6(n) is Euler\u2019s Totient function which counts the number of positive integers less than n that are coprime to n. This result was later generalized to all groups by Maurer in [2] using elliptic curves, and requiring a different number-theoretic assumption. In this paper, we will present Maurer\u2019s main result, along with the necessary background from number theory."
                    },
                    {
                        "title": "A Topological Method for Three-Dimensional Symmetry Detection in Images",
                        "abstract": "Recent work in computational topology has made it possible to compute the homology of geometric objects given only a description in the form of point cloud data\u2013a finite collection of sample points lying on or near the space being sampled. The main technique is that of persistent homology developed in [8]. Additionally, a program for the computation of persistent homology known as PLEX has been developed by the computational topology research group at Stanford. A portion of the previous applications of persistent homology to data analysis have been in the domain of image and shape processing. In [4], persistent homology techniques were used to determine that the set of highcontrast patches in natural images has the topology of a klein bottle. Other applications have focused on using persistent homology as an invariant or signature to distinguish shapes. In [5] the authors develop a hybrid method using techniques from differential geometry and persistent homology to classify shapes. In this paper, we present a new application of persistent homology to approximate symmetry detection in sets of images. The motivation for our work is the desire to identify the symmetries of a three-dimensional object given a collection of two-dimensional images (e.g. photographs) of the object under different rotations. In this situation if two images of an object under two different rotations appear the same, then it seems likely that the object has some rotational symmetry. The question then is how to extract the symmetry information encoded by a set of images of an object. To answer this question, we define the symmetrized rotation space of an object to be the set of all possible three-dimensional rotations of the object."
                    }
                ]
            },
            "f4d6aa75-339b-4e4b-a362-a2c9c85b66ee": {
                "pk": "f4d6aa75-339b-4e4b-a362-a2c9c85b66ee",
                "name": "Anirudh Goyal",
                "collaborators": [
                    "M. Mozer",
                    "Y. Bengio",
                    "Sanjeev Arora",
                    "Aniket Didolkar",
                    "Nan Rosemary Ke",
                    "Dingli Yu",
                    "Cristian Meo",
                    "Haoyu Zhao",
                    "Simran Kaur",
                    "Wenyue Zheng"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Self-Supervised Learning",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "Bayesian-LoRA: LoRA based Parameter Efficient Fine-Tuning using Optimal Quantization levels and Rank Values trough Differentiable Bayesian Gates",
                        "abstract": "It is a common practice in natural language processing to pre-train a single model on a general domain and then fine-tune it for downstream tasks. However, when it comes to Large Language Models, fine-tuning the entire model can be computationally expensive, resulting in very intensive energy consumption. As a result, several Parameter Efficient Fine-Tuning (PEFT) approaches were recently proposed. One of the most popular approaches is low-rank adaptation (LoRA), where the key insight is decomposing the update weights of the pre-trained model into two low-rank matrices. However, the proposed approaches either use the same rank value across all different weight matrices, which has been shown to be a sub-optimal choice, or do not use any quantization technique, one of the most important factors when it comes to a model's energy consumption. In this work, we propose Bayesian-LoRA which approaches low-rank adaptation and quantization from a Bayesian perspective by employing a prior distribution on both quantization levels and rank values. As a result, B-LoRA is able to fine-tune a pre-trained model on a specific downstream task, finding the optimal rank values and quantization levels for every low-rank matrix. We validate the proposed model by fine-tuning a pre-trained DeBERTaV3 on the GLUE benchmark. Moreover, we compare it to relevant baselines and present both qualitative and quantitative results, showing how the proposed approach is able to learn optimal-rank quantized matrices. B-LoRA performs on par with or better than the baselines while reducing the total number of bit operations by roughly 70% compared to the baseline methods."
                    },
                    {
                        "title": "Can Models Learn Skill Composition from Examples?",
                        "abstract": "As large language models (LLMs) become increasingly advanced, their ability to exhibit compositional generalization -- the capacity to combine learned skills in novel ways not encountered during training -- has garnered significant attention. This type of generalization, particularly in scenarios beyond training data, is also of great interest in the study of AI safety and alignment. A recent study introduced the SKILL-MIX evaluation, where models are tasked with composing a short paragraph demonstrating the use of a specified $k$-tuple of language skills. While small models struggled with composing even with $k=3$, larger models like GPT-4 performed reasonably well with $k=5$ and $6$. In this paper, we employ a setup akin to SKILL-MIX to evaluate the capacity of smaller models to learn compositional generalization from examples. Utilizing a diverse set of language skills -- including rhetorical, literary, reasoning, theory of mind, and common sense -- GPT-4 was used to generate text samples that exhibit random subsets of $k$ skills. Subsequent fine-tuning of 7B and 13B parameter models on these combined skill texts, for increasing values of $k$, revealed the following findings: (1) Training on combinations of $k=2$ and $3$ skills results in noticeable improvements in the ability to compose texts with $k=4$ and $5$ skills, despite models never having seen such examples during training. (2) When skill categories are split into training and held-out groups, models significantly improve at composing texts with held-out skills during testing despite having only seen training skills during fine-tuning, illustrating the efficacy of the training approach even with previously unseen skills. This study also suggests that incorporating skill-rich (potentially synthetic) text into training can substantially enhance the compositional capabilities of models."
                    },
                    {
                        "title": "Accelerating Greedy Coordinate Gradient via Probe Sampling",
                        "abstract": "Safety of Large Language Models (LLMs) has become a critical issue given their rapid progresses. Greedy Coordinate Gradient (GCG) is shown to be effective in constructing adversarial prompts to break the aligned LLMs, but optimization of GCG is time-consuming. To reduce the time cost of GCG and enable more comprehensive studies of LLM safety, in this work, we study a new algorithm called $\\texttt{Probe sampling}$. At the core of the algorithm is a mechanism that dynamically determines how similar a smaller draft model's predictions are to the target model's predictions for prompt candidates. When the target model is similar to the draft model, we rely heavily on the draft model to filter out a large number of potential prompt candidates. Probe sampling achieves up to $5.6$ times speedup using Llama2-7b-chat and leads to equal or improved attack success rate (ASR) on the AdvBench. Furthermore, probe sampling is also able to accelerate other prompt optimization techniques and adversarial methods, leading to acceleration of $1.8\\times$ for AutoPrompt, $2.4\\times$ for APE and $2.4\\times$ for AutoDAN."
                    },
                    {
                        "title": "Zero-Shot Object-Centric Representation Learning",
                        "abstract": "The goal of object-centric representation learning is to decompose visual scenes into a structured representation that isolates the entities. Recent successes have shown that object-centric representation learning can be scaled to real-world scenes by utilizing pre-trained self-supervised features. However, so far, object-centric methods have mostly been applied in-distribution, with models trained and evaluated on the same dataset. This is in contrast to the wider trend in machine learning towards general-purpose models directly applicable to unseen data and tasks. Thus, in this work, we study current object-centric methods through the lens of zero-shot generalization by introducing a benchmark comprising eight different synthetic and real-world datasets. We analyze the factors influencing zero-shot performance and find that training on diverse real-world images improves transferability to unseen scenarios. Furthermore, inspired by the success of task-specific fine-tuning in foundation models, we introduce a novel fine-tuning strategy to adapt pre-trained vision encoders for the task of object discovery. We find that the proposed approach results in state-of-the-art performance for unsupervised object discovery, exhibiting strong zero-shot transfer to unseen datasets."
                    },
                    {
                        "title": "Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning",
                        "abstract": "We introduce Instruct-SkillMix, an automated approach for creating diverse, high quality SFT data. The Instruct-SkillMix pipeline involves two stages, each leveraging an existing powerful LLM: (1) Skill extraction: uses the LLM to extract core\"skills\"for instruction-following, either from existing datasets, or by directly prompting the model; (2) Data generation: uses the powerful LLM to generate (instruction, response) data that exhibit a randomly chosen pair of these skills. Here, the use of random skill combinations promotes diversity and difficulty. Vanilla SFT (i.e., no PPO, DPO, or RL methods) on data generated from Instruct-SkillMix leads to strong gains on instruction following benchmarks such as AlpacaEval 2.0, MT-Bench, and WildBench. With just $4$K examples, LLaMA-3-8B-Base achieves 42.76% length-controlled win rate on AlpacaEval 2.0. To our knowledge, this achieves state-of-the-art performance among all models that have only undergone SFT (no RL methods) and competes with proprietary models such as Claude 3 Opus and LLaMA-3.1-405B-Instruct. Ablation studies also suggest plausible reasons for why creating open instruction-tuning datasets via naive crowd-sourcing has proved difficult. Introducing low quality answers (\"shirkers\") in $20\\%$ of Instruct-SkillMix examples causes performance to plummet, sometimes catastrophically. The Instruct-SkillMix pipeline is flexible and is adaptable to other settings."
                    },
                    {
                        "title": "Monte Carlo Tree Search Boosts Reasoning via Iterative Preference Learning",
                        "abstract": "We introduce an approach aimed at enhancing the reasoning capabilities of Large Language Models (LLMs) through an iterative preference learning process inspired by the successful strategy employed by AlphaZero. Our work leverages Monte Carlo Tree Search (MCTS) to iteratively collect preference data, utilizing its look-ahead ability to break down instance-level rewards into more granular step-level signals. To enhance consistency in intermediate steps, we combine outcome validation and stepwise self-evaluation, continually updating the quality assessment of newly generated data. The proposed algorithm employs Direct Preference Optimization (DPO) to update the LLM policy using this newly generated step-level preference data. Theoretical analysis reveals the importance of using on-policy sampled data for successful self-improving. Extensive evaluations on various arithmetic and commonsense reasoning tasks demonstrate remarkable performance improvements over existing models. For instance, our approach outperforms the Mistral-7B Supervised Fine-Tuning (SFT) baseline on GSM8K, MATH, and ARC-C, with substantial increases in accuracy to $81.8\\%$ (+$5.9\\%$), $34.7\\%$ (+$5.8\\%$), and $76.4\\%$ (+$15.8\\%$), respectively. Additionally, our research delves into the training and inference compute tradeoff, providing insights into how our method effectively maximizes performance gains. Our code is publicly available at https://github.com/YuxiXie/MCTS-DPO."
                    },
                    {
                        "title": "Learning Beyond Pattern Matching? Assaying Mathematical Understanding in LLMs",
                        "abstract": "We are beginning to see progress in language model assisted scientific discovery. Motivated by the use of LLMs as a general scientific assistant, this paper assesses the domain knowledge of LLMs through its understanding of different mathematical skills required to solve problems. In particular, we look at not just what the pre-trained model already knows, but how it learned to learn from information during in-context learning or instruction-tuning through exploiting the complex knowledge structure within mathematics. Motivated by the Neural Tangent Kernel (NTK), we propose \\textit{NTKEval} to assess changes in LLM's probability distribution via training on different kinds of math data. Our systematic analysis finds evidence of domain understanding during in-context learning. By contrast, certain instruction-tuning leads to similar performance changes irrespective of training on different data, suggesting a lack of domain understanding across different skills."
                    },
                    {
                        "title": "Metacognitive Capabilities of LLMs: An Exploration in Mathematical Problem Solving",
                        "abstract": "Metacognitive knowledge refers to humans' intuitive knowledge of their own thinking and reasoning processes. Today's best LLMs clearly possess some reasoning processes. The paper gives evidence that they also have metacognitive knowledge, including ability to name skills and procedures to apply given a task. We explore this primarily in context of math reasoning, developing a prompt-guided interaction procedure to get a powerful LLM to assign sensible skill labels to math questions, followed by having it perform semantic clustering to obtain coarser families of skill labels. These coarse skill labels look interpretable to humans. To validate that these skill labels are meaningful and relevant to the LLM's reasoning processes we perform the following experiments. (a) We ask GPT-4 to assign skill labels to training questions in math datasets GSM8K and MATH. (b) When using an LLM to solve the test questions, we present it with the full list of skill labels and ask it to identify the skill needed. Then it is presented with randomly selected exemplar solved questions associated with that skill label. This improves accuracy on GSM8k and MATH for several strong LLMs, including code-assisted models. The methodology presented is domain-agnostic, even though this article applies it to math problems."
                    },
                    {
                        "title": "AI-Assisted Generation of Difficult Math Questions",
                        "abstract": "Current LLM training positions mathematical reasoning as a core capability. With publicly available sources fully tapped, there is unmet demand for diverse and challenging math questions. Relying solely on human experts is both time-consuming and costly, while LLM-generated questions often lack the requisite diversity and difficulty. We present a design framework that combines the strengths of LLMs with a human-in-the-loop approach to generate a diverse array of challenging math questions. We leverage LLM metacognition skills [Didolkar et al., 2024] of a strong LLM to extract core\"skills\"from existing math datasets. These skills serve as the basis for generating novel and difficult questions by prompting the LLM with random pairs of core skills. The use of two different skills within each question makes finding such questions an\"out of distribution\"task for both LLMs and humans. Our pipeline employs LLMs to iteratively generate and refine questions and solutions through multiturn prompting. Human annotators then verify and further refine the questions, with their efficiency enhanced via further LLM interactions. Applying this pipeline on skills extracted from the MATH dataset [Hendrycks et al., 2021] resulted in MATH$^2$ - a dataset of higher-quality math questions, as evidenced by: (a) Lower performance of all models on MATH$^2$ than on MATH (b) Higher performance on MATH when using MATH$^2$ questions as in-context examples. Although focused on mathematics, our methodology seems applicable to other domains requiring structured reasoning, and potentially as a component of scalable oversight. Also of interest is a striking relationship observed between models' performance on the new dataset: the success rate on MATH$^2$ is the square on MATH, suggesting that successfully solving the question in MATH$^2$ requires a nontrivial combination of two distinct math skills."
                    },
                    {
                        "title": "Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates",
                        "abstract": "Public LLMs such as the Llama 2-Chat have driven huge activity in LLM research. These models underwent alignment training and were considered safe. Recently Qi et al. (2023) reported that even benign fine-tuning (e.g., on seemingly safe datasets) can give rise to unsafe behaviors in the models. The current paper is about methods and best practices to mitigate such loss of alignment. Through extensive experiments on several chat models (Meta's Llama 2-Chat, Mistral AI's Mistral 7B Instruct v0.2, and OpenAI's GPT-3.5 Turbo), this paper uncovers that the prompt templates used during fine-tuning and inference play a crucial role in preserving safety alignment, and proposes the\"Pure Tuning, Safe Testing\"(PTST) principle -- fine-tune models without a safety prompt, but include it at test time. Fine-tuning experiments on GSM8K, ChatDoctor, and OpenOrca show that PTST significantly reduces the rise of unsafe behaviors, and even almost eliminates them in some cases."
                    },
                    {
                        "title": "TC-VAE: Uncovering Out-of-Distribution Data Generative Factors",
                        "abstract": "Uncovering data generative factors is the ultimate goal of disentanglement learning. Although many works proposed disentangling generative models able to uncover the underlying generative factors of a dataset, so far no one was able to uncover OOD generative factors (i.e., factors of variations that are not explicitly shown on the dataset). Moreover, the datasets used to validate these models are synthetically generated using a balanced mixture of some predefined generative factors, implicitly assuming that generative factors are uniformly distributed across the datasets. However, real datasets do not present this property. In this work we analyse the effect of using datasets with unbalanced generative factors, providing qualitative and quantitative results for widely used generative models. Moreover, we propose TC-VAE, a generative model optimized using a lower bound of the joint total correlation between the learned latent representations and the input data. We show that the proposed model is able to uncover OOD generative factors on different datasets and outperforms on average the related baselines in terms of downstream disentanglement metrics."
                    },
                    {
                        "title": "Unlearning via Sparse Representations",
                        "abstract": "Machine \\emph{unlearning}, which involves erasing knowledge about a \\emph{forget set} from a trained model, can prove to be costly and infeasible by existing techniques. We propose a nearly compute-free zero-shot unlearning technique based on a discrete representational bottleneck. We show that the proposed technique efficiently unlearns the forget set and incurs negligible damage to the model's performance on the rest of the data set. We evaluate the proposed technique on the problem of \\textit{class unlearning} using three datasets: CIFAR-10, CIFAR-100, and LACUNA-100. We compare the proposed technique to SCRUB, a state-of-the-art approach which uses knowledge distillation for unlearning. Across all three datasets, the proposed technique performs as well as, if not better than SCRUB while incurring almost no computational cost."
                    },
                    {
                        "title": "Spotlight Attention: Robust Object-Centric Learning With a Spatial Locality Prior",
                        "abstract": "The aim of object-centric vision is to construct an explicit representation of the objects in a scene. This representation is obtained via a set of interchangeable modules called \\emph{slots} or \\emph{object files} that compete for local patches of an image. The competition has a weak inductive bias to preserve spatial continuity; consequently, one slot may claim patches scattered diffusely throughout the image. In contrast, the inductive bias of human vision is strong, to the degree that attention has classically been described with a spotlight metaphor. We incorporate a spatial-locality prior into state-of-the-art object-centric vision models and obtain significant improvements in segmenting objects in both synthetic and real-world datasets. Similar to human visual attention, the combination of image content and spatial constraints yield robust unsupervised object-centric learning, including less sensitivity to model hyperparameters."
                    },
                    {
                        "title": "DiscoGen: Learning to Discover Gene Regulatory Networks",
                        "abstract": "Accurately inferring Gene Regulatory Networks (GRNs) is a critical and challenging task in biology. GRNs model the activatory and inhibitory interactions between genes and are inherently causal in nature. To accurately identify GRNs, perturbational data is required. However, most GRN discovery methods only operate on observational data. Recent advances in neural network-based causal discovery methods have significantly improved causal discovery, including handling interventional data, improvements in performance and scalability. However, applying state-of-the-art (SOTA) causal discovery methods in biology poses challenges, such as noisy data and a large number of samples. Thus, adapting the causal discovery methods is necessary to handle these challenges. In this paper, we introduce DiscoGen, a neural network-based GRN discovery method that can denoise gene expression measurements and handle interventional data. We demonstrate that our model outperforms SOTA neural network-based causal discovery methods."
                    },
                    {
                        "title": "Physical Reasoning and Object Planning for Household Embodied Agents",
                        "abstract": "In this study, we explore the sophisticated domain of task planning for robust household embodied agents, with a particular emphasis on the intricate task of selecting substitute objects. We introduce the CommonSense Object Affordance Task (COAT), a novel framework designed to analyze reasoning capabilities in commonsense scenarios. This approach is centered on understanding how these agents can effectively identify and utilize alternative objects when executing household tasks, thereby offering insights into the complexities of practical decision-making in real-world environments. Drawing inspiration from factors affecting human decision-making, we explore how large language models tackle this challenge through four meticulously crafted commonsense question-and-answer datasets featuring refined rules and human annotations. Our evaluation of state-of-the-art language models on these datasets sheds light on three pivotal considerations: 1) aligning an object's inherent utility with the task at hand, 2) navigating contextual dependencies (societal norms, safety, appropriateness, and efficiency), and 3) accounting for the current physical state of the object. To maintain accessibility, we introduce five abstract variables reflecting an object's physical condition, modulated by human insights, to simulate diverse household scenarios. Our contributions include insightful human preference mappings for all three factors and four extensive QA datasets (2K, 15k, 60k, 70K questions) probing the intricacies of utility dependencies, contextual dependencies and object physical states. The datasets, along with our findings, are accessible at: https://github.com/Ayush8120/COAT. This research not only advances our understanding of physical commonsense reasoning in language models but also paves the way for future improvements in household agent intelligence."
                    },
                    {
                        "title": "Cycle Consistency Driven Object Discovery",
                        "abstract": "Developing deep learning models that effectively learn object-centric representations, akin to human cognition, remains a challenging task. Existing approaches facilitate object discovery by representing objects as fixed-size vectors, called ``slots'' or ``object files''. While these approaches have shown promise in certain scenarios, they still exhibit certain limitations. First, they rely on architectural priors which can be unreliable and usually require meticulous engineering to identify the correct objects. Second, there has been a notable gap in investigating the practical utility of these representations in downstream tasks. To address the first limitation, we introduce a method that explicitly optimizes the constraint that each object in a scene should be associated with a distinct slot. We formalize this constraint by introducing consistency objectives which are cyclic in nature. By integrating these consistency objectives into various existing slot-based object-centric methods, we showcase substantial improvements in object-discovery performance. These enhancements consistently hold true across both synthetic and real-world scenes, underscoring the effectiveness and adaptability of the proposed approach. To tackle the second limitation, we apply the learned object-centric representations from the proposed method to two downstream reinforcement learning tasks, demonstrating considerable performance enhancements compared to conventional slot-based and monolithic representation learning methods. Our results suggest that the proposed approach not only improves object discovery, but also provides richer features for downstream tasks."
                    },
                    {
                        "title": "Neural Causal Structure Discovery from Interventions",
                        "abstract": "Recent promising results have generated a surge of interest in continuous optimization methods for causal discovery from observational data. However, there are theoretical limitations on the identi\ufb01ability of underlying structures obtained solely from observational data. Interventional data, on the other hand, provides richer information about the underlying data-generating process. Nevertheless, extending and applying methods designed for observational data to include interventions is a challenging problem. To address this issue, we propose a general framework based on neural networks to develop models that incorporate both observational and interventional data. Notably, our method can handle the challenging and realistic scenario where the identity of the intervened upon variable is unknown. We evaluate our proposed approach in the context of graph recovery, both de novo and from a partially-known edge set. Our method achieves strong benchmark results on various structure learning tasks, including structure recovery of synthetic graphs as well as standard graphs from the Bayesian Network Repository"
                    },
                    {
                        "title": "Vaccine Effectiveness Against Influenza A-Associated Hospitalization, Organ Failure, and Death: United States, 2022-2023.",
                        "abstract": "BACKGROUND Influenza circulation during the 2022-2023 season in the United States largely returned to pre-coronavirus disease 2019 (COVID-19)-pandemic patterns and levels. Influenza A(H3N2) viruses were detected most frequently this season, predominately clade 3C.2a1b.2a, a close antigenic match to the vaccine strain.   METHODS To understand effectiveness of the 2022-2023 influenza vaccine against influenza-associated hospitalization, organ failure, and death, a multicenter sentinel surveillance network in the United States prospectively enrolled adults hospitalized with acute respiratory illness between 1 October 2022, and 28 February 2023. Using the test-negative design, vaccine effectiveness (VE) estimates against influenza-associated hospitalization, organ failures, and death were measured by comparing the odds of current-season influenza vaccination in influenza-positive case-patients and influenza-negative, SARS-CoV-2-negative control-patients.   RESULTS A total of 3707 patients, including 714 influenza cases (33% vaccinated) and 2993 influenza- and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)-negative controls (49% vaccinated) were analyzed. VE against influenza-associated hospitalization was 37% (95% confidence interval [CI]: 27%-46%) and varied by age (18-64 years: 47% [30%-60%]; \u226565 years: 28% [10%-43%]), and virus (A[H3N2]: 29% [6%-46%], A[H1N1]: 47% [23%-64%]). VE against more severe influenza-associated outcomes included: 41% (29%-50%) against influenza with hypoxemia treated with supplemental oxygen; 65% (56%-72%) against influenza with respiratory, cardiovascular, or renal failure treated with organ support; and 66% (40%-81%) against influenza with respiratory failure treated with invasive mechanical ventilation.   CONCLUSIONS During an early 2022-2023 influenza season with a well-matched influenza vaccine, vaccination was associated with reduced risk of influenza-associated hospitalization and organ failure."
                    },
                    {
                        "title": "Leveraging the Third Dimension in Contrastive Learning",
                        "abstract": "Self-Supervised Learning (SSL) methods operate on unlabeled data to learn robust representations useful for downstream tasks. Most SSL methods rely on augmentations obtained by transforming the 2D image pixel map. These augmentations ignore the fact that biological vision takes place in an immersive three-dimensional, temporally contiguous environment, and that low-level biological vision relies heavily on depth cues. Using a signal provided by a pretrained state-of-the-art monocular RGB-to-depth model (the \\emph{Depth Prediction Transformer}, Ranftl et al., 2021), we explore two distinct approaches to incorporating depth signals into the SSL framework. First, we evaluate contrastive learning using an RGB+depth input representation. Second, we use the depth signal to generate novel views from slightly different camera positions, thereby producing a 3D augmentation for contrastive learning. We evaluate these two approaches on three different SSL methods -- BYOL, SimSiam, and SwAV -- using ImageNette (10 class subset of ImageNet), ImageNet-100 and ImageNet-1k datasets. We find that both approaches to incorporating depth signals improve the robustness and generalization of the baseline SSL methods, though the first approach (with depth-channel concatenation) is superior. For instance, BYOL with the additional depth channel leads to an increase in downstream classification accuracy from 85.3\\% to 88.0\\% on ImageNette and 84.1\\% to 87.0\\% on ImageNet-C."
                    }
                ]
            },
            "a161bc95-40d9-498e-8d98-87728671109c": {
                "pk": "a161bc95-40d9-498e-8d98-87728671109c",
                "name": "Sanjeev Arora",
                "collaborators": [
                    "Anirudh Goyal",
                    "Dingli Yu",
                    "Y. Bengio",
                    "M. Mozer",
                    "Kaifeng Lyu",
                    "Haoyu Zhao",
                    "Simran Kaur",
                    "Simon Park",
                    "Nan Rosemary Ke",
                    "Vedant Shah"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Compositional Generalization",
                    "AI Safety"
                ],
                "publications": [
                    {
                        "title": "Can Models Learn Skill Composition from Examples?",
                        "abstract": "As large language models (LLMs) become increasingly advanced, their ability to exhibit compositional generalization -- the capacity to combine learned skills in novel ways not encountered during training -- has garnered significant attention. This type of generalization, particularly in scenarios beyond training data, is also of great interest in the study of AI safety and alignment. A recent study introduced the SKILL-MIX evaluation, where models are tasked with composing a short paragraph demonstrating the use of a specified $k$-tuple of language skills. While small models struggled with composing even with $k=3$, larger models like GPT-4 performed reasonably well with $k=5$ and $6$. In this paper, we employ a setup akin to SKILL-MIX to evaluate the capacity of smaller models to learn compositional generalization from examples. Utilizing a diverse set of language skills -- including rhetorical, literary, reasoning, theory of mind, and common sense -- GPT-4 was used to generate text samples that exhibit random subsets of $k$ skills. Subsequent fine-tuning of 7B and 13B parameter models on these combined skill texts, for increasing values of $k$, revealed the following findings: (1) Training on combinations of $k=2$ and $3$ skills results in noticeable improvements in the ability to compose texts with $k=4$ and $5$ skills, despite models never having seen such examples during training. (2) When skill categories are split into training and held-out groups, models significantly improve at composing texts with held-out skills during testing despite having only seen training skills during fine-tuning, illustrating the efficacy of the training approach even with previously unseen skills. This study also suggests that incorporating skill-rich (potentially synthetic) text into training can substantially enhance the compositional capabilities of models."
                    },
                    {
                        "title": "Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning",
                        "abstract": "We introduce Instruct-SkillMix, an automated approach for creating diverse, high quality SFT data. The Instruct-SkillMix pipeline involves two stages, each leveraging an existing powerful LLM: (1) Skill extraction: uses the LLM to extract core\"skills\"for instruction-following, either from existing datasets, or by directly prompting the model; (2) Data generation: uses the powerful LLM to generate (instruction, response) data that exhibit a randomly chosen pair of these skills. Here, the use of random skill combinations promotes diversity and difficulty. Vanilla SFT (i.e., no PPO, DPO, or RL methods) on data generated from Instruct-SkillMix leads to strong gains on instruction following benchmarks such as AlpacaEval 2.0, MT-Bench, and WildBench. With just $4$K examples, LLaMA-3-8B-Base achieves 42.76% length-controlled win rate on AlpacaEval 2.0. To our knowledge, this achieves state-of-the-art performance among all models that have only undergone SFT (no RL methods) and competes with proprietary models such as Claude 3 Opus and LLaMA-3.1-405B-Instruct. Ablation studies also suggest plausible reasons for why creating open instruction-tuning datasets via naive crowd-sourcing has proved difficult. Introducing low quality answers (\"shirkers\") in $20\\%$ of Instruct-SkillMix examples causes performance to plummet, sometimes catastrophically. The Instruct-SkillMix pipeline is flexible and is adaptable to other settings."
                    },
                    {
                        "title": "Metacognitive Capabilities of LLMs: An Exploration in Mathematical Problem Solving",
                        "abstract": "Metacognitive knowledge refers to humans' intuitive knowledge of their own thinking and reasoning processes. Today's best LLMs clearly possess some reasoning processes. The paper gives evidence that they also have metacognitive knowledge, including ability to name skills and procedures to apply given a task. We explore this primarily in context of math reasoning, developing a prompt-guided interaction procedure to get a powerful LLM to assign sensible skill labels to math questions, followed by having it perform semantic clustering to obtain coarser families of skill labels. These coarse skill labels look interpretable to humans. To validate that these skill labels are meaningful and relevant to the LLM's reasoning processes we perform the following experiments. (a) We ask GPT-4 to assign skill labels to training questions in math datasets GSM8K and MATH. (b) When using an LLM to solve the test questions, we present it with the full list of skill labels and ask it to identify the skill needed. Then it is presented with randomly selected exemplar solved questions associated with that skill label. This improves accuracy on GSM8k and MATH for several strong LLMs, including code-assisted models. The methodology presented is domain-agnostic, even though this article applies it to math problems."
                    },
                    {
                        "title": "AI-Assisted Generation of Difficult Math Questions",
                        "abstract": "Current LLM training positions mathematical reasoning as a core capability. With publicly available sources fully tapped, there is unmet demand for diverse and challenging math questions. Relying solely on human experts is both time-consuming and costly, while LLM-generated questions often lack the requisite diversity and difficulty. We present a design framework that combines the strengths of LLMs with a human-in-the-loop approach to generate a diverse array of challenging math questions. We leverage LLM metacognition skills [Didolkar et al., 2024] of a strong LLM to extract core\"skills\"from existing math datasets. These skills serve as the basis for generating novel and difficult questions by prompting the LLM with random pairs of core skills. The use of two different skills within each question makes finding such questions an\"out of distribution\"task for both LLMs and humans. Our pipeline employs LLMs to iteratively generate and refine questions and solutions through multiturn prompting. Human annotators then verify and further refine the questions, with their efficiency enhanced via further LLM interactions. Applying this pipeline on skills extracted from the MATH dataset [Hendrycks et al., 2021] resulted in MATH$^2$ - a dataset of higher-quality math questions, as evidenced by: (a) Lower performance of all models on MATH$^2$ than on MATH (b) Higher performance on MATH when using MATH$^2$ questions as in-context examples. Although focused on mathematics, our methodology seems applicable to other domains requiring structured reasoning, and potentially as a component of scalable oversight. Also of interest is a striking relationship observed between models' performance on the new dataset: the success rate on MATH$^2$ is the square on MATH, suggesting that successfully solving the question in MATH$^2$ requires a nontrivial combination of two distinct math skills."
                    },
                    {
                        "title": "Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates",
                        "abstract": "Public LLMs such as the Llama 2-Chat have driven huge activity in LLM research. These models underwent alignment training and were considered safe. Recently Qi et al. (2023) reported that even benign fine-tuning (e.g., on seemingly safe datasets) can give rise to unsafe behaviors in the models. The current paper is about methods and best practices to mitigate such loss of alignment. Through extensive experiments on several chat models (Meta's Llama 2-Chat, Mistral AI's Mistral 7B Instruct v0.2, and OpenAI's GPT-3.5 Turbo), this paper uncovers that the prompt templates used during fine-tuning and inference play a crucial role in preserving safety alignment, and proposes the\"Pure Tuning, Safe Testing\"(PTST) principle -- fine-tune models without a safety prompt, but include it at test time. Fine-tuning experiments on GSM8K, ChatDoctor, and OpenOrca show that PTST significantly reduces the rise of unsafe behaviors, and even almost eliminates them in some cases."
                    },
                    {
                        "title": "A Quadratic Synchronization Rule for Distributed Deep Learning",
                        "abstract": "In distributed deep learning with data parallelism, synchronizing gradients at each training step can cause a huge communication overhead, especially when many nodes work together to train large models. Local gradient methods, such as Local SGD, address this issue by allowing workers to compute locally for $H$ steps without synchronizing with others, hence reducing communication frequency. While $H$ has been viewed as a hyperparameter to trade optimization efficiency for communication cost, recent research indicates that setting a proper $H$ value can lead to generalization improvement. Yet, selecting a proper $H$ is elusive. This work proposes a theory-grounded method for determining $H$, named the Quadratic Synchronization Rule (QSR), which recommends dynamically setting $H$ in proportion to $\\frac{1}{\\eta^2}$ as the learning rate $\\eta$ decays over time. Extensive ImageNet experiments on ResNet and ViT show that local gradient methods with QSR consistently improve the test accuracy over other synchronization strategies. Compared with the standard data parallel training, QSR enables Local AdamW on ViT-B to cut the training time on 16 or 64 GPUs down from 26.7 to 20.2 hours or from 8.6 to 5.5 hours and, at the same time, achieves $1.16\\%$ or $0.84\\%$ higher top-1 validation accuracy."
                    },
                    {
                        "title": "Unlearning via Sparse Representations",
                        "abstract": "Machine \\emph{unlearning}, which involves erasing knowledge about a \\emph{forget set} from a trained model, can prove to be costly and infeasible by existing techniques. We propose a nearly compute-free zero-shot unlearning technique based on a discrete representational bottleneck. We show that the proposed technique efficiently unlearns the forget set and incurs negligible damage to the model's performance on the rest of the data set. We evaluate the proposed technique on the problem of \\textit{class unlearning} using three datasets: CIFAR-10, CIFAR-100, and LACUNA-100. We compare the proposed technique to SCRUB, a state-of-the-art approach which uses knowledge distillation for unlearning. Across all three datasets, the proposed technique performs as well as, if not better than SCRUB while incurring almost no computational cost."
                    }
                ]
            }
        }
    },
    "2407.11052": {
        "paper_data": {
            "title": "Revisiting, Benchmarking and Understanding Unsupervised Graph Domain Adaptation",
            "url": "http://arxiv.org/abs/2407.11052v1",
            "arxiv_id": "2407.11052",
            "authors": [
                "Meihan Liu",
                "Zhen Zhang",
                "Jiachen Tang",
                "Jiajun Bu",
                "Bingsheng He",
                "Sheng Zhou"
            ],
            "abstract": "Unsupervised Graph Domain Adaptation (UGDA) involves the transfer of knowledge from a label-rich source graph to an unlabeled target graph under domain discrepancies. Despite the proliferation of methods designed for this emerging task, the lack of standard experimental settings and fair performance comparisons makes it challenging to understand which and when models perform well across different scenarios. To fill this gap, we present the first comprehensive benchmark for unsupervised graph domain adaptation named GDABench, which encompasses 16 algorithms across 5 datasets with 74 adaptation tasks. Through extensive experiments, we observe that the performance of current UGDA models varies significantly across different datasets and adaptation scenarios. Specifically, we recognize that when the source and target graphs face significant distribution shifts, it is imperative to formulate strategies to effectively address and mitigate graph structural shifts. We also find that with appropriate neighbourhood aggregation mechanisms, simple GNN variants can even surpass state-of-the-art UGDA baselines. To facilitate reproducibility, we have developed an easy-to-use library PyGDA for training and evaluating existing UGDA methods, providing a standardized platform in this community. Our source codes and datasets can be found at: https://github.com/pygda-team/pygda.",
            "introduction": "   1 Introduction  The last decade has witnessed significant advancements in Graph Neural Networks (GNNs) with their successful applications spanning various fields [1, 2], including social network analysis [3], protein interaction prediction [4], and traffic flow forecasting [5], etc. However, the presence of distribution shifts [6] and label scarcity in real-world graph data impedes the ability of existing GNN models to adapt to new domains [7]. To addresses this challenge, Unsupervised Graph Domain Adaptation (UGDA) has become an important solution for transferring knowledge from a labeled source graph to an unlabeled target graph. This powerful paradigm has been widely studied, unlocking broader application for graph neural networks.   Despite a wide range of researches of UGDA have been developed [8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25], the understanding of their capabilities and limitations is inadequate due to the following reasons:   \u2022  Inadequate Evaluation of Domain Distribution Discrepancies. The distribution shifts in node attributes, graph structures, and label proportions between graphs will significantly influence the adaptation performance and result in various adaptation scenarios [25, 26]. However, the types and magnitudes of distribution discrepancies among different domains have not been thoroughly evaluated and discussed, which makes it challenging to understand the robustness and efficacy of current methods.    \u2022  Lack of Standard, Fair, and Comprehensive Comparisons. The utilization of distinct datasets, varying data processing methodologies, and divergent data partitioning strategies among existing domain adaptation models results in incomparability across different findings [8, 27, 16]. Furthermore, they are mainly evaluated against limited baselines with constrained scenarios, such as social networks or citation networks, which lack validation of the model\u2019s capability in more diverse or complex applications.    \u2022  Limited Investigation on GNN Inherent Transferability. Despite the advancements made by existing UGDA algorithms, it is still unclear how data shift impose challenges on GNN and how to unleash the transferabililty power for GNN. Due to the non-IID nature of graph data, the aggregation architectures and aggregation scopes affect the underlying distribution of latent representations generated by GNN. When there exists significant structural difference between the source and target domains, such as variations in the degree [28, 29] or differences in subgraph patterns [22, 30], the information aggregation capability of GNN will be directly affected [31, 24, 30]. Thus, understanding the key components that affect adaptation in GNN will be crucial for enhancing GNN\u2019s transferabililty, which is still an open problem.      To fill these gaps, we revisit existing UGDA algorithms and conduct a comprehensive benchmark named GDABench. Specifically, GDABench includes 16 state-of-the-art UGDA models and 5 diverse graph datasets covering node attributes, graph structures, and label proportion shifts, resulting in 74 different adaptation tasks. Additionally, we also explore the limits of GNN transferability by combining 7 GNN variants with 2 domain alignment and 3 unsupervised graph learning techniques. Our work is the first to provide a rigorous empirical analysis of how various aggregation mechanisms influence alignments in domain adaptation task. Through comprehensive experiments, we observe that: (1) the performance of current UGDA models varies greatly across different datasets and adaptation scenarios; (2) it is crucial to develop tailored strategies to address graph structural shifts, especially",
            "references": [
                {
                    "title": "Collaborate to Adapt: Source-Free Graph Domain Adaptation via Bi-directional Adaptation",
                    "abstract": "Unsupervised Graph Domain Adaptation (UGDA) has emerged as a practical solution to transfer knowledge from a label-rich source graph to a completely unlabelled target graph. However, most methods require a labelled source graph to provide supervision signals, which might not be accessible in the real-world settings due to regulations and privacy concerns. In this paper, we explore the scenario of source-free unsupervised graph domain adaptation, which tries to address the domain adaptation problem without accessing the labelled source graph. Specifically, we present a novel paradigm called GraphCTA, which performs model adaptation and graph adaptation collaboratively through a series of procedures: (1) conduct model adaptation based on node's neighborhood predictions in target graph considering both local and global information; (2) perform graph adaptation by updating graph structure and node attributes via neighborhood contrastive learning; and (3) the updated graph serves as an input to facilitate the subsequent iteration of model adaptation, thereby establishing a collaborative loop between model adaptation and graph adaptation. Comprehensive experiments are conducted on various public datasets. The experimental results demonstrate that our proposed model outperforms recent source-free baselines by large margins."
                },
                {
                    "title": "Pairwise Alignment Improves Graph Domain Adaptation",
                    "abstract": "Graph-based methods, pivotal for label inference over interconnected objects in many real-world applications, often encounter generalization challenges, if the graph used for model training differs significantly from the graph used for testing. This work delves into Graph Domain Adaptation (GDA) to address the unique complexities of distribution shifts over graph data, where interconnected data points experience shifts in features, labels, and in particular, connecting patterns. We propose a novel, theoretically principled method, Pairwise Alignment (Pair-Align) to counter graph structure shift by mitigating conditional structure shift (CSS) and label shift (LS). Pair-Align uses edge weights to recalibrate the influence among neighboring nodes to handle CSS and adjusts the classification loss with label weights to handle LS. Our method demonstrates superior performance in real-world applications, including node classification with region shift in social networks, and the pileup mitigation task in particle colliding experiments. For the first application, we also curate the largest dataset by far for GDA studies. Our method shows strong performance in synthetic and other existing benchmark datasets."
                },
                {
                    "title": "Rethinking Propagation for Unsupervised Graph Domain Adaptation",
                    "abstract": "Unsupervised Graph Domain Adaptation (UGDA) aims to transfer knowledge from a labelled source graph to an unlabelled target graph in order to address the distribution shifts between graph domains. Previous works have primarily focused on aligning data from the source and target graph in the representation space learned by graph neural networks (GNNs). However, the inherent generalization capability of GNNs has been largely overlooked. Motivated by our empirical analysis, we reevaluate the role of GNNs in graph domain adaptation and uncover the pivotal role of the propagation process in GNNs for adapting to different graph domains. We provide a comprehensive theoretical analysis of UGDA and derive a generalization bound for multi-layer GNNs. By formulating GNN Lipschitz for k-layer GNNs, we show that the target risk bound can be tighter by removing propagation layers in source graph and stacking multiple propagation layers in target graph. Based on the empirical and theoretical analysis mentioned above, we propose a simple yet effective approach called A2GNN for graph domain adaptation. Through extensive experiments on real-world datasets, we demonstrate the effectiveness of our proposed A2GNN framework."
                },
                {
                    "title": "GraphMETRO: Mitigating Complex Graph Distribution Shifts via Mixture of Aligned Experts",
                    "abstract": "Graph data are inherently complex and heterogeneous, leading to a high natural diversity of distributional shifts. However, it remains unclear how to build machine learning architectures that generalize to complex non-synthetic distributional shifts naturally occurring in the real world. Here we develop GraphMETRO, a Graph Neural Network architecture, that reliably models natural diversity and captures complex distributional shifts. GraphMETRO employs a Mixture-of-Experts (MoE) architecture with a gating model and multiple expert models, where each expert model targets a specific distributional shift to produce a shift-invariant representation, and the gating model identifies shift components. Additionally, we design a novel objective that aligns the representations from different expert models to ensure smooth optimization. GraphMETRO achieves state-of-the-art results on four datasets from GOOD benchmark comprised of complex and natural real-world distribution shifts, improving by 67% and 4.2% on WebKB and Twitch datasets."
                },
                {
                    "title": "SA-GDA: Spectral Augmentation for Graph Domain Adaptation",
                    "abstract": "Graph neural networks (GNNs) have achieved impressive impressions for graph-related tasks. However, most GNNs are primarily studied under the cases of signal domain with supervised training, which requires abundant task-specific labels and is difficult to transfer to other domains. There are few works focused on domain adaptation for graph node classification. They mainly focused on aligning the feature space of the source and target domains, without considering the feature alignment between different categories, which may lead to confusion of classification in the target domain. However, due to the scarcity of labels of the target domain, we cannot directly perform effective alignment of categories from different domains, which makes the problem more challenging. In this paper, we present the Spectral Augmentation for Graph Domain Adaptation (SA-GDA) for graph node classification. First, we observe that nodes with the same category in different domains exhibit similar characteristics in the spectral domain, while different classes are quite different. Following the observation, we align the category feature space of different domains in the spectral domain instead of aligning the whole features space, and we theoretical proof the stability of proposed SA-GDA. Then, we develop a dual graph convolutional network to jointly exploits local and global consistency for feature aggregation. Last, we utilize a domain classifier with an adversarial learning submodule to facilitate knowledge transfer between different domain graphs. Experimental results on a variety of publicly available datasets reveal the effectiveness of our SA-GDA."
                },
                {
                    "title": "Improving Graph Domain Adaptation with Network Hierarchy",
                    "abstract": "Graph domain adaptation models have become instrumental in addressing cross-network learning problems due to their ability to transfer abundant label and structural knowledge from source graphs to target graphs. A crucial step in transfer involves measuring domain discrepancy, which refers to distribution shifts between graphs from source and target domains. While conventional models simply provide a node-level measurement, exploiting information from different levels of network hierarchy is intuitive. As each hierarchical level characterizes distinct and meaningful properties or functionalities of the original graph, integrating domain discrepancy based on such hierarchies should contribute to a more precise domain discrepancy measurement. Moreover, class conditional distribution shift is often overlooked in node classification tasks, which could potentially lead to sub-optimal performance. To address the above limitations, we propose a new graph domain adaptation model and apply it to cross-network node classification tasks. Specifically, a hierarchical pooling model to extract meaningful and adaptive hierarchical structures is designed, where both marginal and class conditional distribution shifts on each hierarchical level are jointly minimized. The effectiveness is demonstrated through theoretical analysis and experimental studies across various datasets."
                },
                {
                    "title": "GraphPatcher: Mitigating Degree Bias for Graph Neural Networks via Test-time Augmentation",
                    "abstract": "Recent studies have shown that graph neural networks (GNNs) exhibit strong biases towards the node degree: they usually perform satisfactorily on high-degree nodes with rich neighbor information but struggle with low-degree nodes. Existing works tackle this problem by deriving either designated GNN architectures or training strategies specifically for low-degree nodes. Though effective, these approaches unintentionally create an artificial out-of-distribution scenario, where models mainly or even only observe low-degree nodes during the training, leading to a downgraded performance for high-degree nodes that GNNs originally perform well at. In light of this, we propose a test-time augmentation framework, namely GraphPatcher, to enhance test-time generalization of any GNNs on low-degree nodes. Specifically, GraphPatcher iteratively generates virtual nodes to patch artificially created low-degree nodes via corruptions, aiming at progressively reconstructing target GNN's predictions over a sequence of increasingly corrupted nodes. Through this scheme, GraphPatcher not only learns how to enhance low-degree nodes (when the neighborhoods are heavily corrupted) but also preserves the original superior performance of GNNs on high-degree nodes (when lightly corrupted). Additionally, GraphPatcher is model-agnostic and can also mitigate the degree bias for either self-supervised or supervised GNNs. Comprehensive experiments are conducted over seven benchmark datasets and GraphPatcher consistently enhances common GNNs' overall performance by up to 3.6% and low-degree performance by up to 6.5%, significantly outperforming state-of-the-art baselines. The source code is publicly available at https://github.com/jumxglhf/GraphPatcher."
                },
                {
                    "title": "Bridged-GNN: Knowledge Bridge Learning for Effective Knowledge Transfer",
                    "abstract": "The data-hungry problem, characterized by insufficiency and low-quality of data, poses obstacles for deep learning models. Transfer learning has been a feasible way to transfer knowledge from high-quality external data of source domains to limited data of target domains, which follows a domain-level knowledge transfer to learn a shared posterior distribution. However, they are usually built on strong assumptions, e.g., the domain invariant posterior distribution, which is usually unsatisfied and may introduce noises, resulting in poor generalization ability on target domains. Inspired by Graph Neural Networks (GNNs) that aggregate information from neighboring nodes, we redefine the paradigm as learning a knowledge-enhanced posterior distribution for target domains, namely Knowledge Bridge Learning (KBL). KBL first learns the scope of knowledge transfer by constructing a Bridged-Graph that connects knowledgeable samples to each target sample and then performs sample-wise knowledge transfer via GNNs.KBL is free from strong assumptions and is robust to noises in the source data. Guided by KBL, we propose the Bridged-GNN including an Adaptive Knowledge Retrieval module to build Bridged-Graph and a Graph Knowledge Transfer module. Comprehensive experiments on both un-relational and relational data-hungry scenarios demonstrate the significant improvements of Bridged-GNN compared with SOTA methods"
                },
                {
                    "title": "OpenGDA: Graph Domain Adaptation Benchmark for Cross-network Learning",
                    "abstract": "Graph domain adaptation models are widely adopted in cross-network learning tasks to transfer labeling or structural knowledge. Currently, there mainly exist two limitations in evaluating graph domain adaptation models. On one side, they are primarily tested for the specific cross-network node classification task, leaving tasks at edge-level and graph-level largely under-explored. Moreover, they are primarily examined in limited scenarios, such as social networks or citation networks, needing more validation in richer scenarios. As comprehensively assessing models could enhance model practicality in real-world applications, we propose a benchmark known as OpenGDA. It provides abundant pre-processed and unified datasets for different types of tasks (node, edge, graph). They originate from diverse scenarios, covering web information systems, urban systems and natural systems. Furthermore, it integrates state-of-the-art models with standardized and end-to-end pipelines. Overall, OpenGDA provides a user-friendly, scalable and reproducible benchmark for evaluating graph domain adaptation models. The benchmark experiments highlight the challenges of applying GDA models to real-world applications with consistent good performance, and they potentially provide insights to future research. As an emerging project, OpenGDA will be regularly updated with new datasets and models. It could be accessed from https://github.com/Skyorca/OpenGDA."
                },
                {
                    "title": "Correntropy-Induced Wasserstein GCN: Learning Graph Embedding via Domain Adaptation",
                    "abstract": "Graph embedding aims at learning vertex representations in a low-dimensional space by distilling information from a complex-structured graph. Recent efforts in graph embedding have been devoted to generalizing the representations from the trained graph in a source domain to the new graph in a different target domain based on information transfer. However, when the graphs are contaminated by unpredictable and complex noise in practice, this transfer problem is quite challenging because of the need to extract helpful knowledge from the source graph and to reliably transfer knowledge to the target graph. This paper puts forward a two-step correntropy-induced Wasserstein GCN (graph convolutional network, or CW-GCN for short) architecture to facilitate the robustness in cross-graph embedding. In the first step, CW-GCN originally investigates correntropy-induced loss in GCN, which places bounded and smooth losses on the noisy nodes with incorrect edges or attributes. Consequently, helpful information are extracted only from clean nodes in the source graph. In the second step, a novel Wasserstein distance is introduced to measure the difference in marginal distributions between graphs, avoiding the negative influence of noise. Afterwards, CW-GCN maps the target graph to the same embedding space as the source graph by minimizing the Wasserstein distance, and thus the knowledge preserved in the first step is expected to be reliably transferred to assist the target graph analysis tasks. Extensive experiments demonstrate the significant superiority of CW-GCN over state-of-the-art methods in different noisy environments."
                },
                {
                    "title": "Structural Re-weighting Improves Graph Domain Adaptation",
                    "abstract": "In many real-world applications, graph-structured data used for training and testing have differences in distribution, such as in high energy physics (HEP) where simulation data used for training may not match real experiments. Graph domain adaptation (GDA) is a method used to address these differences. However, current GDA primarily works by aligning the distributions of node representations output by a single graph neural network encoder shared across the training and testing domains, which may often yield sub-optimal solutions. This work examines different impacts of distribution shifts caused by either graph structure or node attributes and identifies a new type of shift, named conditional structure shift (CSS), which current GDA approaches are provably sub-optimal to deal with. A novel approach, called structural reweighting (StruRW), is proposed to address this issue and is tested on synthetic graphs, four benchmark datasets, and a new application in HEP. StruRW has shown significant performance improvement over the baselines in the settings with large graph structure shifts, and reasonable performance improvement when node attribute shift dominates."
                },
                {
                    "title": "Learning Adaptive Node Embeddings Across Graphs",
                    "abstract": "Recently, learning embeddings of nodes in graphs has attracted increasing research attention. There are two main kinds of graph embedding methods, i.e., transductive embedding methods and inductive embedding methods. The former focuses on directly optimizing the embedding vectors, and the latter tries to learn a mapping function for the given nodes and features. However, little work has focused on applying the learned model from one graph to another, which is a pervasive idea in Computer Vision or Natural Language Processing. Although some of the graph neural networks (GNNs) present a similar motivation, none of them considers both the structure bias and the feature bias between graphs. In this paper, we present a novel graph embedding problem called Adaptive Task (AT), and propose a unified framework for the adaptive task, which introduces two types of alignment to learn adaptive node embeddings across graphs. Then, based on the proposed framework, a novel Graph Adaptive Embedding network (GraphAE) is designed to address the adaptive task. Furthermore, we extend GraphAE to a multi-graph version to consider a more complex adaptive situation. The extensive experimental results demonstrate that our model significantly outperforms the state-of-the-art methods, and also show that our framework can make a great improvement over a number of existing GNNs."
                },
                {
                    "title": "Non-IID Transfer Learning on Graphs",
                    "abstract": "Transfer learning refers to the transfer of knowledge or information from a relevant source domain to a target domain. However, most existing transfer learning theories and algorithms focus on IID tasks, where the source/target samples are assumed to be independent and identically distributed. Very little effort is devoted to theoretically studying the knowledge transferability on non-IID tasks, e.g., cross-network mining. To bridge the gap, in this paper, we propose rigorous generalization bounds and algorithms for cross-network transfer learning from a source graph to a target graph. The crucial idea is to characterize the cross-network knowledge transferability from the perspective of the Weisfeiler-Lehman graph isomorphism test. To this end, we propose a novel Graph Subtree Discrepancy to measure the graph distribution shift between source and target graphs. Then the generalization error bounds on cross-network transfer learning, including both cross-network node classification and link prediction tasks, can be derived in terms of the source knowledge and the Graph Subtree Discrepancy across domains. This thereby motivates us to propose a generic graph adaptive network (GRADE) to minimize the distribution shift between source and target graphs for cross-network transfer learning. Experimental results verify the effectiveness and efficiency of our GRADE framework on both cross-network node classification and cross-domain recommendation tasks."
                },
                {
                    "title": "Make More Connections: Urban Traffic Flow Forecasting with Spatiotemporal Adaptive Gated Graph Convolution Network",
                    "abstract": "Urban traffic flow forecasting is a critical issue in intelligent transportation systems. Due to the complexity and uncertainty of urban road conditions, how to capture the dynamic spatiotemporal correlation and make accurate predictions is very challenging. In most of existing works, urban road network is often modeled as a fixed graph based on local proximity. However, such modeling is not sufficient to describe the dynamics of the road network and capture the global contextual information. In this paper, we consider constructing the road network as a dynamic weighted graph through attention mechanism. Furthermore, we propose to seek both spatial neighbors and semantic neighbors to make more connections between road nodes. We propose a novel Spatiotemporal Adaptive Gated Graph Convolution Network (STAG-GCN) to predict traffic conditions for several time steps ahead. STAG-GCN mainly consists of two major components: (1) multivariate self-attention Temporal Convolution Network (TCN) is utilized to capture local and long-range temporal dependencies across recent, daily-periodic and weekly-periodic observations; (2) mix-hop AG-GCN extracts selective spatial and semantic dependencies within multi-layer stacking through adaptive graph gating mechanism and mix-hop propagation mechanism. The output of different components are weighted fused to generate the final prediction results. Extensive experiments on two real-world large scale urban traffic dataset have verified the effectiveness, and the multi-step forecasting performance of our proposed models outperforms the state-of-the-art baselines."
                },
                {
                    "title": "Augmentations in Graph Contrastive Learning: Current Methodological Flaws & Towards Better Practices",
                    "abstract": "Graph classification has a wide range of applications in bioinformatics, social sciences, automated fake news detection, web document classification, and more. In many practical scenarios, including web-scale applications, labels are scarce or hard to obtain. Unsupervised learning is thus a natural paradigm for these settings, but its performance often lags behind that of supervised learning. However, recently contrastive learning (CL) has enabled unsupervised computer vision models to perform comparably to supervised models. Theoretical and empirical works analyzing visual CL frameworks find that leveraging large datasets and task relevant augmentations is essential for CL framework success. Interestingly, graph CL frameworks report high performance while using orders of magnitude smaller data, and employing domain-agnostic graph augmentations (DAGAs) that can corrupt task relevant information. Motivated by these discrepancies, we seek to determine why existing graph CL frameworks continue to perform well, and identify flawed practices in graph data augmentation and popular graph CL evaluation protocols. We find that DAGA can destroy task-relevant information and harm the model\u2019s ability to learn discriminative representations. We also show that on small benchmark datasets, the inductive bias of graph neural networks can significantly compensate for these limitations, while on larger graph classification tasks commonly-used DAGAs perform poorly. Based on our findings, we propose better practices and sanity checks for future research and applications, including adhering to principles in visual CL when designing context-aware graph augmentations. For example, in graph-based document classification, which can be used for better web search, we show task-relevant augmentations improve accuracy by up to 20."
                },
                {
                    "title": "Adversarial Separation Network for Cross-Network Node Classification",
                    "abstract": "Node classification is an important yet challenging task in various network applications, and many effective methods have been developed for a single network. While for cross-network scenarios, neither single network embedding nor traditional domain adaptation can directly solve the task. Existing approaches have been proposed to combine network embedding and domain adaptation for cross-network node classification. However, they only focus on domain-invariant features, ignoring the individual features of each network, and they only utilize 1-hop neighborhood information (local consistency), ignoring the global consistency information. To tackle the above problems, in this paper, we propose a novel model, Adversarial Separation Network(ASN), to learn effective node representations between source and target networks. We explicitly separate domain-private and domain-shared information. Two domain-private encoders are employed to extract the domain-specific features in each network and a shared encoder is employed to extract the domain-invariant shared features across networks. Moreover, in each encoder, we combine local and global consistency to capture network topology information more comprehensively. ASN integrates deep network embedding with adversarial domain adaptation to reduce the distribution discrepancy across domains. Extensive experiments on real-world datasets show that our proposed model achieves state-of-the-art performance in cross-network node classification tasks compared with existing algorithms."
                },
                {
                    "title": "An Empirical Study of Graph Contrastive Learning",
                    "abstract": "Graph Contrastive Learning (GCL) establishes a new paradigm for learning graph representations without human annotations. Although remarkable progress has been witnessed recently, the success behind GCL is still left somewhat mysterious. In this work, we first identify several critical design considerations within a general GCL paradigm, including augmentation functions, contrasting modes, contrastive objectives, and negative mining techniques. Then, to understand the interplay of different GCL components, we conduct extensive, controlled experiments over a set of benchmark tasks on datasets across various domains. Our empirical studies suggest a set of general receipts for effective GCL, e.g., simple topology augmentations that produce sparse graph views bring promising performance improvements; contrasting modes should be aligned with the granularities of end tasks. In addition, to foster future research and ease the implementation of GCL algorithms, we develop an easy-to-use library PyGCL, featuring modularized CL components, standardized evaluation, and experiment management. We envision this work to provide useful empirical evidence of effective GCL algorithms and offer several insights for future research."
                },
                {
                    "title": "Graph Domain Adaptation: A Generative View",
                    "abstract": "Recent years have witnessed tremendous interest in deep learning on graph-structured data. Due to the high cost of collecting labeled graph-structured data, domain adaptation is important to supervised graph learning tasks with limited samples. However, current graph domain adaptation methods are generally adopted from traditional domain adaptation tasks, and the properties of graph-structured data are not well utilized. For example, the observed social networks on different platforms are controlled not only by the different crowds or communities but also by domain-specific policies and background noise. Based on these properties in graph-structured data, we first assume that the graph-structured data generation process is controlled by three independent types of latent variables, i.e., the semantic latent variables, the domain latent variables, and the random latent variables. Based on this assumption, we propose a disentanglement-based unsupervised domain adaptation method for the graph-structured data, which applies variational graph auto-encoders to recover these latent variables and disentangles them via three supervised learning modules. Extensive experimental results on two real-world datasets in the graph classification task reveal that our method not only significantly outperforms the traditional domain adaptation methods and the disentangled-based domain adaptation methods but also outperforms the state-of-the-art graph domain adaptation algorithms. The code is available at https://github.com/rynewu224/GraphDA."
                },
                {
                    "title": "Graph Self-Supervised Learning: A Survey",
                    "abstract": "Deep learning on graphs has attracted significant interests recently. However, most of the works have focused on (semi-) supervised learning, resulting in shortcomings including heavy label reliance, poor generalization, and weak robustness. To address these issues, self-supervised learning (SSL), which extracts informative knowledge through well-designed pretext tasks without relying on manual labels, has become a promising and trending learning paradigm for graph data. Different from SSL on other domains like computer vision and natural language processing, SSL on graphs has an exclusive background, design ideas, and taxonomies. Under the umbrella of graph self-supervised learning, we present a timely and comprehensive review of the existing approaches which employ SSL techniques for graph data. We construct a unified framework that mathematically formalizes the paradigm of graph SSL. According to the objectives of pretext tasks, we divide these approaches into four categories: generation-based, auxiliary property-based, contrast-based, and hybrid approaches. We further describe the applications of graph SSL across various research fields and summarize the commonly used datasets, evaluation benchmark, performance comparison and open-source codes of graph SSL. Finally, we discuss the remaining challenges and potential future directions in this research field."
                },
                {
                    "title": "Graph Contrastive Learning with Augmentations",
                    "abstract": "Generalizable, transferrable, and robust representation learning on graph-structured data remains a challenge for current graph neural networks (GNNs). Unlike what has been developed for convolutional neural networks (CNNs) for image data, self-supervised learning and pre-training are less explored for GNNs. In this paper, we propose a graph contrastive learning (GraphCL) framework for learning unsupervised representations of graph data. We first design four types of graph augmentations to incorporate various priors. We then systematically study the impact of various combinations of graph augmentations on multiple datasets, in four different settings: semi-supervised, unsupervised, and transfer learning as well as adversarial attacks. The results show that, even without tuning augmentation extents nor using sophisticated GNN architectures, our GraphCL framework can produce graph representations of similar or better generalizability, transferrability, and robustness compared to state-of-the-art methods. We also investigate the impact of parameterized graph augmentation extents and patterns, and observe further performance gains in preliminary experiments. Our codes are available at https://github.com/Shen-Lab/GraphCL."
                },
                {
                    "title": "Investigating and Mitigating Degree-Related Biases in Graph Convoltuional Networks",
                    "abstract": "Graph Convolutional Networks (GCNs) show promising results for semi-supervised learning tasks on graphs, thus become favorable comparing with other approaches. Despite the remarkable success of GCNs, it is difficult to train GCNs with insufficient supervision. When labeled data are limited, the performance of GCNs becomes unsatisfying for low-degree nodes. While some prior work analyze successes and failures of GCNs on the entire model level, profiling GCNs on individual node level is still underexplored. In this paper, we analyze GCNs in regard to the node degree distribution. From empirical observation to theoretical proof, we confirm that GCNs are biased towards nodes with larger degrees with higher accuracy on them, even if high-degree nodes are underrepresented in most graphs. We further develop a novel Self-Supervised-Learning Degree-Specific GCN (SL-DSGCN) that mitigate the degree-related biases of GCNs from model and data aspects. Firstly, we propose a degree-specific GCN layer that captures both discrepancies and similarities of nodes with different degrees, which reduces the inner model-aspect biases of GCNs caused by sharing the same parameters with all nodes. Secondly, we design a self-supervised-learning algorithm that creates pseudo labels with uncertainty scores on unlabeled nodes with a Bayesian neural network. Pseudo labels increase the chance of connecting to labeled neighbors for low-degree nodes, thus reducing the biases of GCNs from the data perspective. Uncertainty scores are further exploited to weight pseudo labels dynamically in the stochastic gradient descent for SL-DSGCN. Experiments on three benchmark datasets show SL-DSGCN not only outperforms state-of-the-art self-training/self-supervised-learning GCN methods, but also improves GCN accuracy dramatically for low-degree nodes."
                },
                {
                    "title": "Unsupervised Domain Adaptive Graph Convolutional Networks",
                    "abstract": "Graph convolutional networks (GCNs) have achieved impressive success in many graph related analytics tasks. However, most GCNs only work in a single domain (graph) incapable of transferring knowledge from/to other domains (graphs), due to the challenges in both graph representation learning and domain adaptation over graph structures. In this paper, we present a novel approach, unsupervised domain adaptive graph convolutional networks (UDA-GCN), for domain adaptation learning for graphs. To enable effective graph representation learning, we first develop a dual graph convolutional network component, which jointly exploits local and global consistency for feature aggregation. An attention mechanism is further used to produce a unified representation for each node in different graphs. To facilitate knowledge transfer between graphs, we propose a domain adaptive learning module to optimize three different loss functions, namely source classifier loss, domain classifier loss, and target classifier loss as a whole, thus our model can differentiate class labels in the source domain, samples from different domains, the class labels from the target domain, respectively. Experimental results on real-world datasets in the node classification task validate the performance of our method, compared to state-of-the-art graph neural network algorithms."
                },
                {
                    "title": "Domain Adaptation with Conditional Distribution Matching and Generalized Label Shift",
                    "abstract": "Adversarial learning has demonstrated good performance in the unsupervised domain adaptation setting, by learning domain-invariant representations that perform well on the source domain. However, recent work has underlined limitations of existing methods in the presence of mismatched label distributions between the source and target domains. In this paper, we extend a recent upper-bound on the performance of adversarial domain adaptation to multi-class classification and more general discriminators. We then propose generalized label shift (GLS) as a way to improve robustness against mismatched label distributions. GLS states that, conditioned on the label, there exists a representation of the input that is invariant between the source and target domains. Under GLS, we provide theoretical guarantees on the transfer performance of any classifier. We also devise necessary and sufficient conditions for GLS to hold. The conditions are based on the estimation of the relative class weights between domains and on an appropriate reweighting of samples. Guided by our theoretical insights, we modify three widely used algorithms, JAN, DANN and CDAN and evaluate their performance on standard domain adaptation tasks where our method outperforms the base versions. We also demonstrate significant gains on artificially created tasks with large divergences between their source and target label distributions."
                },
                {
                    "title": "Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation",
                    "abstract": "Unsupervised domain adaptation (UDA) aims to leverage the knowledge learned from a labeled source dataset to solve similar tasks in a new unlabeled domain. Prior UDA methods typically require to access the source data when learning to adapt the model, making them risky and inefficient for decentralized private data. This work tackles a practical setting where only a trained source model is available and investigates how we can effectively utilize such a model without source data to solve UDA problems. We propose a simple yet generic representation learning framework, named \\emph{Source HypOthesis Transfer} (SHOT). SHOT freezes the classifier module (hypothesis) of the source model and learns the target-specific feature extraction module by exploiting both information maximization and self-supervised pseudo-labeling to implicitly align representations from the target domains to the source hypothesis. To verify its versatility, we evaluate SHOT in a variety of adaptation cases including closed-set, partial-set, and open-set domain adaptation. Experiments indicate that SHOT yields state-of-the-art results among multiple domain adaptation benchmarks."
                },
                {
                    "title": "Adversarial Deep Network Embedding for Cross-network Node Classification",
                    "abstract": "In this paper, the task of cross-network node classification, which leverages the abundant labeled nodes from a source network to help classify unlabeled nodes in a target network, is studied. The existing domain adaptation algorithms generally fail to model the network structural information, and the current network embedding models mainly focus on single-network applications. Thus, both of them cannot be directly applied to solve the cross-network node classification problem. This motivates us to propose an adversarial cross-network deep network embedding (ACDNE) model to integrate adversarial domain adaptation with deep network embedding so as to learn network-invariant node representations that can also well preserve the network structural information. In ACDNE, the deep network embedding module utilizes two feature extractors to jointly preserve attributed affinity and topological proximities between nodes. In addition, a node classifier is incorporated to make node representations label-discriminative. Moreover, an adversarial domain adaptation technique is employed to make node representations network-invariant. Extensive experimental results demonstrate that the proposed ACDNE model achieves the state-of-the-art performance in cross-network node classification."
                },
                {
                    "title": "A Survey on Knowledge Graphs: Representation, Acquisition, and Applications",
                    "abstract": "Human knowledge provides a formal understanding of the world. Knowledge graphs that represent structural relations between entities have become an increasingly popular research direction toward cognition and human-level intelligence. In this survey, we provide a comprehensive review of the knowledge graph covering overall research topics about: 1) knowledge graph representation learning; 2) knowledge acquisition and completion; 3) temporal knowledge graph; and 4) knowledge-aware applications and summarize recent breakthroughs and perspective directions to facilitate future research. We propose a full-view categorization and new taxonomies on these topics. Knowledge graph embedding is organized from four aspects of representation space, scoring function, encoding models, and auxiliary information. For knowledge acquisition, especially knowledge graph completion, embedding methods, path inference, and logical rule reasoning are reviewed. We further explore several emerging topics, including metarelational learning, commonsense reasoning, and temporal knowledge graphs. To facilitate future research on knowledge graphs, we also provide a curated collection of data sets and open-source libraries on different tasks. In the end, we have a thorough outlook on several promising research directions."
                },
                {
                    "title": "Network Together: Node Classification via Cross-Network Deep Network Embedding",
                    "abstract": "Network embedding is a highly effective method to learn low-dimensional node vector representations with original network structures being well preserved. However, existing network embedding algorithms are mostly developed for a single network, which fails to learn generalized feature representations across different networks. In this article, we study a cross-network node classification problem, which aims at leveraging the abundant labeled information from a source network to help classify the unlabeled nodes in a target network. To succeed in such a task, transferable features should be learned for nodes across different networks. To this end, a novel cross-network deep network embedding (CDNE) model is proposed to incorporate domain adaptation into deep network embedding in order to learn label-discriminative and network-invariant node vector representations. On the one hand, CDNE leverages network structures to capture the proximities between nodes within a network, by mapping more strongly connected nodes to have more similar latent vector representations. On the other hand, node attributes and labels are leveraged to capture the proximities between nodes across different networks by making the same labeled nodes across networks have aligned latent vector representations. Extensive experiments have been conducted, demonstrating that the proposed CDNE model significantly outperforms the state-of-the-art network embedding algorithms in cross-network node classification."
                },
                {
                    "title": "Graph Transfer Learning via Adversarial Domain Adaptation With Graph Convolution",
                    "abstract": "This paper studies the problem of cross-network node classification to overcome the insufficiency of labeled data in a single network. It aims to leverage the label information in a partially labeled source network to assist node classification in a completely unlabeled or partially labeled target network. Existing methods for single network learning cannot solve this problem due to the domain shift across networks. Some multi-network learning methods heavily rely on the existence of cross-network connections, thus are inapplicable for this problem. To tackle this problem, we propose a novel graph transfer learning framework AdaGCN by leveraging the techniques of adversarial domain adaptation and graph convolution. It consists of two components: a semi-supervised learning component and an adversarial domain adaptation component. The former aims to learn class discriminative node representations with given label information of the source and target networks, while the latter contributes to mitigating the distribution divergence between the source and target domains to facilitate knowledge transfer. Extensive empirical evaluations on real-world datasets show that AdaGCN can successfully transfer class information with a low label rate on the source network and a substantial divergence between the source and target domains."
                },
                {
                    "title": "DANE: Domain Adaptive Network Embedding",
                    "abstract": "Recent works reveal that network embedding techniques enable many machine learning models to handle diverse downstream tasks on graph structured data. However, as previous methods usually focus on learning embeddings for a single network, they can not learn representations transferable on multiple networks. Hence, it is important to design a network embedding algorithm that supports downstream model transferring on different networks, known as domain adaptation. In this paper, we propose a novel Domain Adaptive Network Embedding framework, which applies graph convolutional network to learn transferable embeddings. In DANE, nodes from multiple networks are encoded to vectors via a shared set of learnable parameters so that the vectors share an aligned embedding space. The distribution of embeddings on different networks are further aligned by adversarial learning regularization. In addition, DANE's advantage in learning transferable network embedding can be guaranteed theoretically. Extensive experiments reflect that the proposed framework outperforms other state-of-the-art network embedding baselines in cross-network domain adaptation tasks."
                },
                {
                    "title": "How Powerful are Graph Neural Networks?",
                    "abstract": "Graph Neural Networks (GNNs) are an effective framework for representation learning of graphs. GNNs follow a neighborhood aggregation scheme, where the representation vector of a node is computed by recursively aggregating and transforming representation vectors of its neighboring nodes. Many GNN variants have been proposed and have achieved state-of-the-art results on both node and graph classification tasks. However, despite GNNs revolutionizing graph representation learning, there is limited understanding of their representational properties and limitations. Here, we present a theoretical framework for analyzing the expressive power of GNNs to capture different graph structures. Our results characterize the discriminative power of popular GNN variants, such as Graph Convolutional Networks and GraphSAGE, and show that they cannot learn to distinguish certain simple graph structures. We then develop a simple architecture that is provably the most expressive among the class of GNNs and is as powerful as the Weisfeiler-Lehman graph isomorphism test. We empirically validate our theoretical findings on a number of graph classification benchmarks, and demonstrate that our model achieves state-of-the-art performance."
                },
                {
                    "title": "Link Prediction Based on Graph Neural Networks",
                    "abstract": "Link prediction is a key problem for network-structured data. Link prediction heuristics use some score functions, such as common neighbors and Katz index, to measure the likelihood of links. They have obtained wide practical uses due to their simplicity, interpretability, and for some of them, scalability. However, every heuristic has a strong assumption on when two nodes are likely to link, which limits their effectiveness on networks where these assumptions fail. In this regard, a more reasonable way should be learning a suitable heuristic from a given network instead of using predefined ones. By extracting a local subgraph around each target link, we aim to learn a function mapping the subgraph patterns to link existence, thus automatically learning a `heuristic' that suits the current network. In this paper, we study this heuristic learning paradigm for link prediction. First, we develop a novel $\\gamma$-decaying heuristic theory. The theory unifies a wide range of heuristics in a single framework, and proves that all these heuristics can be well approximated from local subgraphs. Our results show that local subgraphs reserve rich information related to link existence. Second, based on the $\\gamma$-decaying theory, we propose a new algorithm to learn heuristics from local subgraphs using a graph neural network (GNN). Its experimental results show unprecedented performance, working consistently well on a wide range of problems."
                },
                {
                    "title": "Graph Attention Networks",
                    "abstract": "We present graph attention networks (GATs), novel neural network architectures that operate on graph-structured data, leveraging masked self-attentional layers to address the shortcomings of prior methods based on graph convolutions or their approximations. By stacking layers in which nodes are able to attend over their neighborhoods' features, we enable (implicitly) specifying different weights to different nodes in a neighborhood, without requiring any kind of costly matrix operation (such as inversion) or depending on knowing the graph structure upfront. In this way, we address several key challenges of spectral-based graph neural networks simultaneously, and make our model readily applicable to inductive as well as transductive problems. Our GAT models have achieved or matched state-of-the-art results across four established transductive and inductive graph benchmarks: the Cora, Citeseer and Pubmed citation network datasets, as well as a protein-protein interaction dataset (wherein test graphs remain unseen during training)."
                },
                {
                    "title": "Wasserstein Distance Guided Representation Learning for Domain Adaptation",
                    "abstract": "\n \n Domain adaptation aims at generalizing a high-performance learner on a target domain via utilizing the knowledge distilled from a source domain which has a different but related data distribution. One solution to domain adaptation is to learn domain invariant feature representations while the learned representations should also be discriminative in prediction. To learn such representations, domain adaptation frameworks usually include a domain invariant representation learning approach to measure and reduce the domain discrepancy, as well as a discriminator for classification. Inspired by Wasserstein GAN, in this paper we propose a novel approach to learn domain invariant feature representations, namely Wasserstein Distance Guided Representation Learning (WDGRL). WDGRL utilizes a neural network, denoted by the domain critic, to estimate empirical Wasserstein distance between the source and target samples and optimizes the feature extractor network to minimize the estimated Wasserstein distance in an adversarial manner. The theoretical advantages of Wasserstein distance for domain adaptation lie in its gradient property and promising generalization bound. Empirical studies on common sentiment and image classification adaptation datasets demonstrate that our proposed WDGRL outperforms the state-of-the-art domain invariant representation learning approaches.\n \n"
                },
                {
                    "title": "Inductive Representation Learning on Large Graphs",
                    "abstract": "Low-dimensional embeddings of nodes in large graphs have proved extremely useful in a variety of prediction tasks, from content recommendation to identifying protein functions. However, most existing approaches require that all nodes in the graph are present during training of the embeddings; these previous approaches are inherently transductive and do not naturally generalize to unseen nodes. Here we present GraphSAGE, a general, inductive framework that leverages node feature information (e.g., text attributes) to efficiently generate node embeddings for previously unseen data. Instead of training individual embeddings for each node, we learn a function that generates embeddings by sampling and aggregating features from a node's local neighborhood. Our algorithm outperforms strong baselines on three inductive node-classification benchmarks: we classify the category of unseen nodes in evolving information graphs based on citation and Reddit post data, and we show that our algorithm generalizes to completely unseen graphs using a multi-graph dataset of protein-protein interactions."
                },
                {
                    "title": "Conditional Adversarial Domain Adaptation",
                    "abstract": "Adversarial learning has been embedded into deep networks to learn disentangled and transferable representations for domain adaptation. Existing adversarial domain adaptation methods may struggle to align different domains of multimodal distributions that are native in classification problems. In this paper, we present conditional adversarial domain adaptation, a principled framework that conditions the adversarial adaptation models on discriminative information conveyed in the classifier predictions. Conditional domain adversarial networks (CDANs) are designed with two novel conditioning strategies: multilinear conditioning that captures the cross-covariance between feature representations and classifier predictions to improve the discriminability, and entropy conditioning that controls the uncertainty of classifier predictions to guarantee the transferability. Experiments testify that the proposed approach exceeds the state-of-the-art results on five benchmark datasets."
                },
                {
                    "title": "Adversarial Discriminative Domain Adaptation",
                    "abstract": "Adversarial learning methods are a promising approach to training robust deep networks, and can generate complex samples across diverse domains. They can also improve recognition despite the presence of domain shift or dataset bias: recent adversarial approaches to unsupervised domain adaptation reduce the difference between the training and test domain distributions and thus improve generalization performance. However, while generative adversarial networks (GANs) show compelling visualizations, they are not optimal on discriminative tasks and can be limited to smaller shifts. On the other hand, discriminative approaches can handle larger domain shifts, but impose tied weights on the model and do not exploit a GAN-based loss. In this work, we first outline a novel generalized framework for adversarial adaptation, which subsumes recent state-of-the-art approaches as special cases, and use this generalized view to better relate prior approaches. We then propose a previously unexplored instance of our general framework which combines discriminative modeling, untied weight sharing, and a GAN loss, which we call Adversarial Discriminative Domain Adaptation (ADDA). We show that ADDA is more effective yet considerably simpler than competing domain-adversarial methods, and demonstrate the promise of our approach by exceeding state-of-the-art unsupervised adaptation results on standard domain adaptation tasks as well as a difficult cross-modality object classification task."
                },
                {
                    "title": "Central Moment Discrepancy (CMD) for Domain-Invariant Representation Learning",
                    "abstract": "The learning of domain-invariant representations in the context of domain adaptation with neural networks is considered. We propose a new regularization method that minimizes the discrepancy between domain-specific latent feature representations directly in the hidden activation space. Although some standard distribution matching approaches exist that can be interpreted as the matching of weighted sums of moments, e.g. Maximum Mean Discrepancy (MMD), an explicit order-wise matching of higher order moments has not been considered before. We propose to match the higher order central moments of probability distributions by means of order-wise moment differences. Our model does not require computationally expensive distance and kernel matrix computations. We utilize the equivalent representation of probability distributions by moment sequences to define a new distance function, called Central Moment Discrepancy (CMD). We prove that CMD is a metric on the set of probability distributions on a compact interval. We further prove that convergence of probability distributions on compact intervals w.r.t. the new metric implies convergence in distribution of the respective random variables. We test our approach on two different benchmark data sets for object recognition (Office) and sentiment analysis of product reviews (Amazon reviews). CMD achieves a new state-of-the-art performance on most domain adaptation tasks of Office and outperforms networks trained with MMD, Variational Fair Autoencoders and Domain Adversarial Neural Networks on Amazon reviews. In addition, a post-hoc parameter sensitivity analysis shows that the new approach is stable w.r.t. parameter changes in a certain interval. The source code of the experiments is publicly available."
                },
                {
                    "title": "Improved Deep Metric Learning with Multi-class N-pair Loss Objective",
                    "abstract": "Deep metric learning has gained much popularity in recent years, following the success of deep learning. However, existing frameworks of deep metric learning based on contrastive loss and triplet loss often suffer from slow convergence, partially because they employ only one negative example while not interacting with the other negative classes in each update. In this paper, we propose to address this problem with a new metric learning objective called multi-class N-pair loss. The proposed objective function firstly generalizes triplet loss by allowing joint comparison among more than one negative examples - more specifically, N-1 negative examples - and secondly reduces the computational burden of evaluating deep embedding vectors via an efficient batch construction strategy using only N pairs of examples, instead of (N+1) x N. We demonstrate the superiority of our proposed loss to the triplet loss as well as other competing loss functions for a variety of tasks on several visual recognition benchmark, including fine-grained object recognition and verification, image clustering and retrieval, and face verification and identification."
                },
                {
                    "title": "Variational Graph Auto-Encoders",
                    "abstract": "We introduce the variational graph auto-encoder (VGAE), a framework for unsupervised learning on graph-structured data based on the variational auto-encoder (VAE). This model makes use of latent variables and is capable of learning interpretable latent representations for undirected graphs. We demonstrate this model using a graph convolutional network (GCN) encoder and a simple inner product decoder. Our model achieves competitive results on a link prediction task in citation networks. In contrast to most existing models for unsupervised learning on graph-structured data and link prediction, our model can naturally incorporate node features, which significantly improves predictive performance on a number of benchmark datasets."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Deep Neural Networks for Learning Graph Representations",
                    "abstract": "\n \n In this paper, we propose a novel model for learning graph representations, which generates a low-dimensional vector representation for each vertex by capturing the graph structural information. Different from other previous research efforts, we adopt a random surfing model to capture graph structural information directly, instead of using the sampling-based method for generating linear sequences proposed by Perozzi et al. (2014). The advantages of our approach will be illustrated from both theorical and empirical perspectives. We also give a new perspective for the matrix factorization method proposed by Levy and Goldberg (2014), in which the pointwise mutual information (PMI) matrix is considered as an analytical solution to the objective function of the skip-gram model with negative sampling proposed by Mikolov et al. (2013). Unlike their approach which involves the use of the SVD for finding the low-dimensitonal projections from the PMI matrix, however, the stacked denoising autoencoder is introduced in our model to extract complex features and model non-linearities. To demonstrate the effectiveness of our model, we conduct experiments on clustering and visualization tasks, employing the learned vertex representations as features. Empirical results on datasets of varying sizes show that our model outperforms other stat-of-the-art models in such tasks.\n \n"
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Transfer learning for speech and language processing",
                    "abstract": "Transfer learning is a vital technique that generalizes models trained for one setting or task to other settings or tasks. For example in speech recognition, an acoustic model trained for one language can be used to recognize speech in another language, with little or no re-training data. Transfer learning is closely related to multi-task learning (cross-lingual vs. multilingual), and is traditionally studied in the name of `model adaptation'. Recent advance in deep learning shows that transfer learning becomes much easier and more effective with high-level abstract features learned by deep models, and the `transfer' can be conducted not only between data distributions and data types, but also between model structures (e.g., shallow nets and deep nets) or even model types (e.g., Bayesian models and neural models). This review paper summarizes some recent prominent research towards this direction, particularly for speech and language processing. We also report some results from our group and highlight the potential of this very interesting research field1."
                },
                {
                    "title": "Learning Transferable Features with Deep Adaptation Networks",
                    "abstract": "Recent studies reveal that a deep neural network can learn transferable features which generalize well to novel tasks for domain adaptation. However, as deep features eventually transition from general to specific along the network, the feature transferability drops significantly in higher layers with increasing domain discrepancy. Hence, it is important to formally reduce the dataset bias and enhance the transferability in task-specific layers. In this paper, we propose a new Deep Adaptation Network (DAN) architecture, which generalizes deep convolutional neural network to the domain adaptation scenario. In DAN, hidden representations of all task-specific layers are embedded in a reproducing kernel Hilbert space where the mean embeddings of different domain distributions can be explicitly matched. The domain discrepancy is further reduced using an optimal multikernel selection method for mean embedding matching. DAN can learn transferable features with statistical guarantees, and can scale linearly by unbiased estimate of kernel embedding. Extensive empirical evidence shows that the proposed architecture yields state-of-the-art image classification error rates on standard domain adaptation benchmarks."
                },
                {
                    "title": "Deep Domain Confusion: Maximizing for Domain Invariance",
                    "abstract": "Recent reports suggest that a generic supervised deep CNN model trained on a large-scale dataset reduces, but does not remove, dataset bias on a standard benchmark. Fine-tuning deep models in a new domain can require a significant amount of data, which for many applications is simply not available. We propose a new CNN architecture which introduces an adaptation layer and an additional domain confusion loss, to learn a representation that is both semantically meaningful and domain invariant. We additionally show that a domain confusion metric can be used for model selection to determine the dimension of an adaptation layer and the best position for the layer in the CNN architecture. Our proposed adaptation method offers empirical performance which exceeds previously published results on a standard benchmark visual domain adaptation task."
                },
                {
                    "title": "A Kernel Two-Sample Test",
                    "abstract": "We propose a framework for analyzing and comparing distributions, which we use to construct statistical tests to determine if two samples are drawn from different distributions. Our test statistic is the largest difference in expectations over functions in the unit ball of a reproducing kernel Hilbert space (RKHS), and is called the maximum mean discrepancy (MMD).We present two distribution free tests based on large deviation bounds for the MMD, and a third test based on the asymptotic distribution of this statistic. The MMD can be computed in quadratic time, although efficient linear time approximations are available. Our statistic is an instance of an integral probability metric, and various classical metrics on distributions are obtained when alternative function classes are used in place of an RKHS. We apply our two-sample tests to a variety of problems, including attribute matching for databases using the Hungarian marriage method, where they perform strongly. Excellent performance is also obtained when comparing distributions over graphs, for which these are the first such tests."
                },
                {
                    "title": "Unbiased look at dataset bias",
                    "abstract": "Datasets are an integral part of contemporary object recognition research. They have been the chief reason for the considerable progress in the field, not just as source of large amounts of training data, but also as means of measuring and comparing performance of competing algorithms. At the same time, datasets have often been blamed for narrowing the focus of object recognition research, reducing it to a single benchmark performance number. Indeed, some datasets, that started out as data capture efforts aimed at representing the visual world, have become closed worlds unto themselves (e.g. the Corel world, the Caltech-101 world, the PASCAL VOC world). With the focus on beating the latest benchmark numbers on the latest dataset, have we perhaps lost sight of the original purpose? The goal of this paper is to take stock of the current state of recognition datasets. We present a comparison study using a set of popular datasets, evaluated based on a number of criteria including: relative data bias, cross-dataset generalization, effects of closed-world assumption, and sample value. The experimental results, some rather surprising, suggest directions that can improve dataset collection as well as algorithm evaluation protocols. But more broadly, the hope is to stimulate discussion in the community regarding this very important, but largely neglected issue."
                },
                {
                    "title": "Weisfeiler-Lehman Graph Kernels",
                    "abstract": "In this article, we propose a family of efficient kernels for large graphs with discrete node labels. Key to our method is a rapid feature extraction scheme based on the Weisfeiler-Lehman test of isomorphism on graphs. It maps the original graph to a sequence of graphs, whose node attributes capture topological and label information. A family of kernels can be defined based on this Weisfeiler-Lehman sequence of graphs, including a highly efficient kernel comparing subtree-like patterns. Its runtime scales only linearly in the number of edges of the graphs and the length of the Weisfeiler-Lehman graph sequence. In our experimental evaluation, our kernels outperform state-of-the-art graph kernels on several graph classification benchmark data sets in terms of accuracy and runtime. Our kernels open the door to large-scale applications of graph kernels in various disciplines such as computational biology and social network analysis."
                },
                {
                    "title": "A Survey on Transfer Learning",
                    "abstract": "A major assumption in many machine learning and data mining algorithms is that the training and future data must be in the same feature space and have the same distribution. However, in many real-world applications, this assumption may not hold. For example, we sometimes have a classification task in one domain of interest, but we only have sufficient training data in another domain of interest, where the latter data may be in a different feature space or follow a different data distribution. In such cases, knowledge transfer, if done successfully, would greatly improve the performance of learning by avoiding much expensive data-labeling efforts. In recent years, transfer learning has emerged as a new learning framework to address this problem. This survey focuses on categorizing and reviewing the current progress on transfer learning for classification, regression, and clustering problems. In this survey, we discuss the relationship between transfer learning and other related machine learning techniques such as domain adaptation, multitask learning and sample selection bias, as well as covariate shift. We also explore some potential future issues in transfer learning research."
                },
                {
                    "title": "ArnetMiner: extraction and mining of academic social networks",
                    "abstract": "This paper addresses several key issues in the ArnetMiner system, which aims at extracting and mining academic social networks. Specifically, the system focuses on: 1) Extracting researcher profiles automatically from the Web; 2) Integrating the publication data into the network from existing digital libraries; 3) Modeling the entire academic network; and 4) Providing search services for the academic network. So far, 448,470 researcher profiles have been extracted using a unified tagging approach. We integrate publications from online Web databases and propose a probabilistic framework to deal with the name ambiguity problem. Furthermore, we propose a unified modeling approach to simultaneously model topical aspects of papers, authors, and publication venues. Search services such as expertise search and people association search have been provided based on the modeling results. In this paper, we describe the architecture and main features of the system. We also present the empirical evaluation of the proposed methods."
                },
                {
                    "title": "Graph Domain Adaptation via Theory-Grounded Spectral Regularization",
                    "abstract": "Transfer learning on graphs drawn from varied distributions (domains) is in great demand across many applications. Emerging methods attempt to learn domaininvariant representations using graph neural networks (GNNs), yet the empirical performances vary and the theoretical foundation is limited. This paper aims at designing theory-grounded algorithms for graph domain adaptation (GDA). (i) As the first attempt, we derive a model-based GDA bound closely related to two GNN spectral properties: spectral smoothness (SS) and maximum frequency response (MFR). This is achieved by cross-pollinating between the OT-based (optimal transport) DA and graph filter theories. (ii) Inspired by the theoretical results, we propose algorithms regularizing spectral properties of SS and MFR to improve GNN transferability. We further extend the GDA theory into the more challenging scenario of conditional shift, where spectral regularization still applies. (iii) More importantly, our analyses of the theory reveal which regularization would improve performance of what transfer learning scenario, (iv) with numerical agreement with extensive real-world experiments: SS and MFR regularizations bring more benefits to the scenarios of node transfer and link transfer, respectively. In a nutshell, our study paves the way toward explicitly constructing and training GNNs that can capture more transferable representations across graph domains. Codes are released at https://github.com/Shen-Lab/GDA-SpecReg."
                },
                {
                    "title": "A Comprehensive Survey on Graph Neural Networks",
                    "abstract": "Deep learning has revolutionized many machine learning tasks in recent years, ranging from image classification and video processing to speech recognition and natural language understanding. The data in these tasks are typically represented in the Euclidean space. However, there is an increasing number of applications, where data are generated from non-Euclidean domains and are represented as graphs with complex relationships and interdependency between objects. The complexity of graph data has imposed significant challenges on the existing machine learning algorithms. Recently, many studies on extending deep learning approaches for graph data have emerged. In this article, we provide a comprehensive overview of graph neural networks (GNNs) in data mining and machine learning fields. We propose a new taxonomy to divide the state-of-the-art GNNs into four categories, namely, recurrent GNNs, convolutional GNNs, graph autoencoders, and spatial\u2013temporal GNNs. We further discuss the applications of GNNs across various domains and summarize the open-source codes, benchmark data sets, and model evaluation of GNNs. Finally, we propose potential research directions in this rapidly growing field."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively evaluate and enhance the transferability of Graph Neural Networks (GNNs) in the context of Unsupervised Graph Domain Adaptation (UGDA) when faced with distribution shifts and label scarcity in real-world graph data?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of current GNN models in adapting to new domains, which is a significant barrier to their practical applications in diverse fields such as social network analysis, bioinformatics, and traffic forecasting. By providing a comprehensive evaluation framework (GDABench) and understanding the inherent transferability of GNNs, this research could lead to more robust and adaptable models, fostering advancements in both theoretical understanding and practical implementations of GNNs. This could open new avenues for research in domain adaptation and enhance the applicability of GNNs in real-world scenarios where data distribution is often non-stationary.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complex nature of graph data, which is often non-IID (independent and identically distributed). Naive approaches may fail due to the intricate relationships between node attributes, graph structures, and label distributions that can vary significantly between source and target domains. Additionally, the lack of standardized evaluation metrics and methodologies complicates the comparison of different UGDA models. Technical obstacles include understanding how different aggregation mechanisms in GNNs affect performance under varying distribution shifts, as well as the need for tailored strategies to address structural differences between graphs.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by inadequate evaluation of domain distribution discrepancies, leading to a poor understanding of the robustness of existing UGDA methods. The lack of standardization in datasets and evaluation methodologies has resulted in findings that are difficult to compare. Additionally, there has been insufficient investigation into the inherent transferability of GNNs, particularly regarding how data shifts impact their performance. Our approach differs by providing a comprehensive benchmark (GDABench) that systematically evaluates various UGDA models across diverse datasets and adaptation scenarios, thereby addressing the gaps in prior research.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of GDABench, which includes 16 state-of-the-art UGDA models and 5 diverse graph datasets that"
            }
        },
        "author_data": {
            "558f4cdb-04cd-46ff-8bea-e415188c7e56": {
                "pk": "558f4cdb-04cd-46ff-8bea-e415188c7e56",
                "name": "Meihan Liu",
                "collaborators": [
                    "Jiajun Bu",
                    "Zhen Zhang",
                    "Ming Gu",
                    "Sheng Zhou",
                    "Zeyu Fang",
                    "Xin Wang",
                    "Anhui Wang",
                    "Hongyang Chen",
                    "Zhao Li",
                    "Bingsheng He"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Domain Adaptation",
                    "Clustering",
                    "Unsupervised Learning"
                ],
                "publications": [
                    {
                        "title": "Rethinking Propagation for Unsupervised Graph Domain Adaptation",
                        "abstract": "Unsupervised Graph Domain Adaptation (UGDA) aims to transfer knowledge from a labelled source graph to an unlabelled target graph in order to address the distribution shifts between graph domains. Previous works have primarily focused on aligning data from the source and target graph in the representation space learned by graph neural networks (GNNs). However, the inherent generalization capability of GNNs has been largely overlooked. Motivated by our empirical analysis, we reevaluate the role of GNNs in graph domain adaptation and uncover the pivotal role of the propagation process in GNNs for adapting to different graph domains. We provide a comprehensive theoretical analysis of UGDA and derive a generalization bound for multi-layer GNNs. By formulating GNN Lipschitz for k-layer GNNs, we show that the target risk bound can be tighter by removing propagation layers in source graph and stacking multiple propagation layers in target graph. Based on the empirical and theoretical analysis mentioned above, we propose a simple yet effective approach called A2GNN for graph domain adaptation. Through extensive experiments on real-world datasets, we demonstrate the effectiveness of our proposed A2GNN framework."
                    },
                    {
                        "title": "Collaborate to Adapt: Source-Free Graph Domain Adaptation via Bi-directional Adaptation",
                        "abstract": "Unsupervised Graph Domain Adaptation (UGDA) has emerged as a practical solution to transfer knowledge from a label-rich source graph to a completely unlabelled target graph. However, most methods require a labelled source graph to provide supervision signals, which might not be accessible in the real-world settings due to regulations and privacy concerns. In this paper, we explore the scenario of source-free unsupervised graph domain adaptation, which tries to address the domain adaptation problem without accessing the labelled source graph. Specifically, we present a novel paradigm called GraphCTA, which performs model adaptation and graph adaptation collaboratively through a series of procedures: (1) conduct model adaptation based on node's neighborhood predictions in target graph considering both local and global information; (2) perform graph adaptation by updating graph structure and node attributes via neighborhood contrastive learning; and (3) the updated graph serves as an input to facilitate the subsequent iteration of model adaptation, thereby establishing a collaborative loop between model adaptation and graph adaptation. Comprehensive experiments are conducted on various public datasets. The experimental results demonstrate that our proposed model outperforms recent source-free baselines by large margins."
                    },
                    {
                        "title": "Homophily-enhanced Structure Learning for Graph Clustering",
                        "abstract": "Graph clustering is a fundamental task in graph analysis, and recent advances in utilizing graph neural networks (GNNs) have shown impressive results. Despite the success of existing GNN-based graph clustering methods, they often overlook the quality of graph structure, which is inherent in real-world graphs due to their sparse and multifarious nature, leading to subpar performance. Graph structure learning allows refining the input graph by adding missing links and removing spurious connections. However, previous endeavors in graph structure learning have predominantly centered around supervised settings, and cannot be directly applied to our specific clustering tasks due to the absence of ground-truth labels. To bridge the gap, we propose a novel method called \\textbf{ho}mophily-enhanced structure \\textbf{le}arning for graph clustering (HoLe). Our motivation stems from the observation that subtly enhancing the degree of homophily within the graph structure can significantly improve GNNs and clustering outcomes. To realize this objective, we develop two clustering-oriented structure learning modules, i.e., hierarchical correlation estimation and cluster-aware sparsification. The former module enables a more accurate estimation of pairwise node relationships by leveraging guidance from latent and clustering spaces, while the latter one generates a sparsified structure based on the similarity matrix and clustering assignments. Additionally, we devise a joint optimization approach alternating between training the homophily-enhanced structure learning and GNN-based clustering, thereby enforcing their reciprocal effects. Extensive experiments on seven benchmark datasets of various types and scales, across a range of clustering metrics, demonstrate the superiority of HoLe against state-of-the-art baselines."
                    }
                ]
            },
            "5c8ea642-b195-458c-84ab-6808d085c583": {
                "pk": "5c8ea642-b195-458c-84ab-6808d085c583",
                "name": "Zhen Zhang",
                "collaborators": [
                    "Yue Cao",
                    "Luming Duan",
                    "Rui Zhang",
                    "Xianmin XU",
                    "Robert Connelly",
                    "Wei Yang",
                    "Qiang Li",
                    "Shiyong Ma"
                ],
                "domain": [
                    "Algebra",
                    "Image Quality Assessment",
                    "Dark Energy",
                    "Control Systems",
                    "Omics Data Analysis"
                ],
                "publications": [
                    {
                        "title": "Simple-minded systems and coherent rings",
                        "abstract": "Let $A$ be a finite dimensional algebra over an algebraically closed field. We present a relationship between simple-minded systems and coherent rings."
                    },
                    {
                        "title": "Cross-IQA: Unsupervised Learning for Image Quality Assessment",
                        "abstract": "Automatic perception of image quality is a challenging problem that impacts billions of Internet and social media users daily. To advance research in this field, we propose a no-reference image quality assessment (NR-IQA) method termed Cross-IQA based on vision transformer(ViT) model. The proposed Cross-IQA method can learn image quality features from unlabeled image data. We construct the pretext task of synthesized image reconstruction to unsupervised extract the image quality information based ViT block. The pretrained encoder of Cross-IQA is used to fine-tune a linear regression model for score prediction. Experimental results show that Cross-IQA can achieve state-of-the-art performance in assessing the low-frequency degradation information (e.g., color change, blurring, etc.) of images compared with the classical full-reference IQA and NR-IQA under the same datasets."
                    },
                    {
                        "title": "Geometrization of light bending and its application to SdSw spacetime",
                        "abstract": "The mysterious dark energy remains one of the greatest puzzles of modern science. Current detections for it are mostly indirect. The spacetime effects of dark energy can be locally described by the SdSw metric. Understanding these local effects exactly is an essential step towards the direct probe of dark energy. From first principles, we prove that dark energy can exert a repulsive dark force on astrophysical scales, different from the Newtonian attraction of both visible and dark matter. One way of measuring local effects of dark energy is through the gravitational deflection of light. We geometrize the bending of light in any curved static spacetime. First of all, we define a generalized deflection angle, referred to as the Gaussian deflection angle, in a mathematically strict and conceptually clean way. Basing on the Gauss-Bonnet theorem, we then prove that the Gaussian deflection angle is equivalent to the surface integral of the Gaussian curvature over a chosen lensing patch. As an application of the geometrization, we study the problem of whether dark energy affects the bending of light and provide a strict solution to this problem in the SdSw spacetime. According to this solution, we propose a method to overcome the difficulty of measuring local dark energy effects. Exactly speaking, we find that the lensing effect of dark energy can be enhanced by 14 orders of magnitude when properly choosing the lensing patch in certain cases. It means that we can probe the existence and nature of dark energy directly in our Solar System. This points to an exciting direction to help unraveling the great mystery of dark energy."
                    },
                    {
                        "title": "Research on the evolution of domestic multi-functional meter technology",
                        "abstract": "The technical evolution of domestic multi-functional electricity meter is deeply discussed. With the rapid development of the domestic power market and the continuous innovation of technology, the domestic multi-functional electricity meters have experienced the transformation from simple billing to complex multi-functional, from a single application to a wide range of fields. This transformation has not only driven the rapid development of electricity meter technology, but also met the increasing power demand and management requirements. This paper expounds the concept of multi-function meter, the working principle and algorithm of digital multiplier, the initiation and evolution of multi-function electricity meter standard, and the initiation and evolution of domestic multi-function electricity meter products. Although the domestic independent production of multi-functional meter has made great achievements in performance, but in the reliability and key process technology still need to be improved. In addition, the development of communication technology also provides a new opportunity for the progress of electricity meter technology. The application of the new technology provides a more convenient and efficient way for the data transmission and remote management of electricity meters. Domestic multi-functional electricity meters have made remarkable achievements in technology evolution and application and expansion, but they still face some challenges and opportunities. In the future, with the continuous development of the power market and the promotion of smart grid construction, domestic multi-functional electricity meters need to continue to strengthen technological innovation and product research and development, improve the reliability and competitiveness of products, in order to meet higher application needs and market requirements."
                    },
                    {
                        "title": "Research on the evolution of smart meter technology",
                        "abstract": "Smart meter is not only a device used to measure the amount of electricity, but also a core component of the smart grid, realizing the efficient monitoring, prediction and management of power use. With an insight into the evolution of smart meter technology, I realized that this change didn't happen overnight. It has undergone a long journey from the initial mechanical electricity meters to the electronic electricity meters, and now to the highly intelligent electricity meters. Technological breakthroughs at each stage have laid the foundation for the final form of smart meters. In the era of mechanical watt-hour meters, the measurement of electric energy mainly depends on the rotation and counting of mechanical structures. The accuracy and stability of this method are affected by mechanical wear, environmental interference and other factors, and it is difficult to meet the increasing demand for power management. With the rapid development of electronic technology, the electronic electricity meter came into being. It uses electronic technology to sample and process the current and voltage, which greatly improves the accuracy and stability of the measurement. At the same time, the electronic electricity meter also has the function of remote meter reading and data processing, which has brought great convenience to the power management. However, there still have some limitations, such as the complexity of data processing and the limitation of communication capacity. It is these challenges that drive the creation of smart power meters. smart meters combine advanced technologies such as the Internet of Things, big data and cloud computing to realize the real-time monitoring, analysis and prediction of the use of power."
                    },
                    {
                        "title": "Quantum Metrology with Dicke Squeezed States",
                        "abstract": "We introduce a new class of quantum many-particle entangled states, called the Dicke squeezed (or DS) states, which can be used to improve the precision in quantum metrology beyond the standard quantum limit. We show that the enhancement in measurement precision is characterized by a single experimentally detectable parameter, called the Dicke squeezing $\\xi_{D}$, which also bounds the entanglement depth for this class of states. The measurement precision approaches the ultimate Heisenberg limit as $\\xi_{D}$ attains the minimum in an ideal Dicke state. Compared with other entangled states, we show that the Dicke squeezed states are more robust to decoherence and give better measurement precision under typical experimental noise."
                    },
                    {
                        "title": "Dark Energy on Astrophysical Scales and Its Detection in the Milky Way",
                        "abstract": "The origin and nature of dark energy is one of the most significant challenges in modern science. This research aims to investigate dark energy on astrophysical scales and provide a cosmology-independent method to measure its equation-of-state parameter $w$. To accomplish this, we introduce the concept of a perfect fluid in any static, curved spacetime, and express the energy-momentum tensor of the perfect fluid in a general isotropic form, namely Weinberg's isotropic form. This enables us to define an equation-of-state parameter in a physical and global manner. Within this theoretical framework, we demonstrate that the energy-momentum tensor of dark energy on different scales can take the general isotropic form. Furthermore, we explore the SdS$_{w}$ spacetime and establish its connection with dark energy in cosmology through the equation-of-state parameter $w$. In the SdS$_{w}$ spacetime, a repulsive dark force can be induced by dark energy locally. We then apply the concept of the dark force to realistic astrophysical systems using the Poisson equation. Finally, we find that an anomaly in the Milky Way rotation curve can be quantitatively interpreted by the dark force. By fitting the galactic curve, we are able to obtain the value of the equation-of-state parameter of dark energy, independently of specific dark energy models."
                    },
                    {
                        "title": "Effective boundary conditions for dynamic contact angle hysteresis on chemically inhomogeneous surfaces",
                        "abstract": "Recent experiments (Guan et al. 2016a,b) showed many interesting phenomena on dynamic contact angle hysteresis while there is still a lack of complete theoretical interpretation. In this work, we study the time averaging of the apparent advancing and receding contact angles on surfaces with periodic chemical patterns. We first derive a Cox-type boundary condition for the apparent dynamic contact angle on homogeneous surfaces using Onsager variational principle. Based on this condition, we propose a reduced model for some typical moving contact line problems on chemically inhomogeneous surfaces in two dimensions. Multiscale expansion and averaging techniques are employed to approximate the model for asymptotically small chemical patterns. We obtain a quantitative formula for the averaged dynamic contact angles. It gives explicitly how the advancing and receding contact angles depend on the velocity and the chemical inhomogeneity of the substrate. The formula is a coarse-graining version of the Cox-type boundary condition on inhomogeneous surfaces. Numerical simulations are presented to validate the analytical results. The numerical results also show that the formula characterizes very well the complicated behaviour of dynamic contact angle hysteresis observed in the experiments."
                    },
                    {
                        "title": "Rigidity of Circle Packings with Flexible Radii",
                        "abstract": "Circle packings are arrangement of circles satisfying specified tangency requirements. Many problems about packing of circles and spheres occur in nature particularly in material design and protein structure. Surprisingly, little is known about the stability and rigidity of circle packings. In this paper, we study the rigidity of circle packings representing a given planar graph. The radii of circles are flexible with equality and inequality constraints. We provide a dual condition for the packing to be rigid in the first order. This gives us a sufficient condition to show a packing is rigid. Then we will explore the difficulties on rigidity problems beyond the first order."
                    },
                    {
                        "title": "Design and Verification of a Flat-Top Beam Shaper Utilizing Spherical Lenses",
                        "abstract": "We propose a novel coaxial spherical-lens shaping system to achieve a flat-top focus profile by exploiting the spherical aberration of the lens. The well-known aspherical-lens shaping system is compared to ours, and a comparison of etching on silicon is conducted between a Gaussian beam and the one shaped by the proposed system. Analytical and experimental results show that the proposed system is able to achieve flat-top beam shaping using only spherical lenses. Compared with the existing systems, our proposed one is capable of shaping Gaussian beams with different wavelengths and waist radii. Furthermore, the shaped one obtained by ours is the flat-top beam with extended depth of focus. This letter provides an efficient and precise flat-top beam shaping scheme for optics and laser technology, and increases the potential of spherical lenses in the field of laser shaping."
                    },
                    {
                        "title": "A robust design of time-varying internal model principle-based control for ultra-precision tracking in a direct-drive servo stage",
                        "abstract": "This paper proposes a robust design of the time-varying internal model principle-based control (TV-IMPC) for tracking sophisticated references generated by linear time-varying (LTV) autonomous systems. The existing TV-IMPC design usually requires a complete knowledge of the plant I/O (input/output) model, leading to the lack of structural robustness. To tackle this issue, we, in the paper, design a gray-box extended state observer (ESO) to estimate and compensate unknown model uncertainties and external disturbances. By means of the ESO feedback, the plant model is kept as nominal, and hence the structural robustness is achieved for the time-varying internal model. It is shown that the proposed design has bounded ESO estimation errors, which can be further adjusted by modifying the corresponding control gains. To stabilize the ESO-based TV-IMPC, a time-varying stabilizer is developed by employing Linear Matrix Inequalities (LMIs). Extensive simulation and experimental studies are conducted on a direct-drive servo stage to validate the proposed robust TV-IMPC with ultra-precision tracking performance ($\\sim 60$nm RMSE out of $\\pm80$mm stroke)."
                    },
                    {
                        "title": "Population aging caused by rise in sex ratio at birth",
                        "abstract": "Despite its historical and biological stability, the sex ratio at birth (SRB) has risen in parts of the world in the last several decades. The resultant demographic consequences, mostly on sex imbalance, are well documented, typically including \"missing girls/women\" and \"marriage squeeze.\" However, the SRB-induced impact on demographic dynamics, particularly its underlying mechanism, has not been explored in depth. We aim to investigate the impact of the SRB rise on the size, structure, and growth of a population, particularly emphasizing on population aging. We provide a simple framework, derived from classical stable population models, to analyze how the SRB rise can reduce the population size and make the population old. We demonstrate that the cohorts born with a higher SRB are smaller in size than those with a lower SRB. As the affected cohorts are born into the population, their smaller size will reduce the total population size, thereby lifting the fraction of old people that were born with the original SRB and have the same size as before. The resultant population aging speed increases as the cohorts with the new SRB take an increasing share of the population. This study adds that, in addition to fertility and mortality, the SRB can be a driving factor of population dynamics, especially when it moves far above normal biological levels."
                    },
                    {
                        "title": "Enhanced Contour Tracking: a Time-Varying Internal Model Principle-Based Approach",
                        "abstract": "Contour tracking plays a crucial role in multi-axis motion control systems, and it requires both multi-axial contouring as well as standard servo performance in each axis. Among the existing contouring control methods, the cross coupled control (CCC) lacks of an asymptotical tracking performance for general contours, and the task coordinate frame (TCF) control usually leads to system nonlinearity, and by design is not well-suited for multi-axis contour tracking. Here we propose a novel time-varying internal model principle-based contouring control (TV-IMCC) methodology to enhance contour tracking performance with both axial and contour error reduction. The proposed TV-IMCC is twofold, including an extended position domain framework with master-slave structures for contour regulation, and a time-varying internal model principle-based controller for each axial tracking precision improvement. Specifically, a novel signal conversion algorithm is proposed with the extended position domain framework, hence the original n-axis contouring problem can be decoupled into (n-1) two-axis master-slave tracking problems in the position domain, and the class of contour candidates can be extended as well. With this, the time-varying internal model principle-based control method is proposed to deal with the time-varying dynamics in the axial systems resulted from the transformation between the time and position domains. Furthermore, the stability analysis is provided for the closed-loop system of the TV-IMCC. Various simulation and experimental results validate the TV-IMCC with enhanced contour tracking performance compared with the existing methods. Moreover, there is no strict requirement on the precision of the master axis, therefore a potential application of the TV-IMCC is multi-axis macro-micro motion systems."
                    },
                    {
                        "title": "OmicsMapNet: Transforming omics data to take advantage of Deep Convolutional Neural Network for discovery",
                        "abstract": "We developed OmicsMapNet approach to take advantage of existing deep leaning frameworks to analyze high-dimensional omics data as 2-dimensional images. The omics data of individual samples were first rearranged into 2D images in which molecular features related in functions, ontologies, or other relationships were organized in spatially adjacent and patterned locations. Deep learning neural networks were trained to classify the images. Molecular features informative of classes of different phenotypes were subsequently identified. As an example, we used the KEGG BRITE database to rearrange RNA-Seq expression data of TCGA diffuse glioma samples as treemaps to capture the functional hierarchical structure of genes in 2D images. Deep Convolutional Neural Networks (CNN) were derived using tools from TensorFlow to learn the grade of TCGA LGG and GBM samples with relatively high accuracy. The most contributory features in the trained CNN were confirmed in pathway analysis for their plausible functional involvement."
                    }
                ]
            },
            "872cf928-78bf-4a06-95b8-a098642ba5f7": {
                "pk": "872cf928-78bf-4a06-95b8-a098642ba5f7",
                "name": "Jiachen Tang",
                "collaborators": [
                    "Luhao Liu",
                    "Yi Shi",
                    "Songlin Li",
                    "Binxi Liang",
                    "Jian Chen",
                    "Yinjiang Shao",
                    "Xinran Wang",
                    "Zhuolin He",
                    "Xinrun Li",
                    "Heng Gao"
                ],
                "domain": [
                    "2D Materials",
                    "Semiconductor Devices",
                    "3D Object Detection",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Enhancement of Carrier Mobility in Semiconductor Nanostructures by Carrier Distribution Engineering",
                        "abstract": "Two-dimensional (2D) van der Waals semiconductors are appealing for low-power transistors. Here, we show the feasibility in enhancing carrier mobility in 2D semiconductors through engineering the vertical distribution of carriers confined inside the ultrathin channels via symmetrizing gate configuration or increasing channel thickness. Through self-consistently solving the Schr\\\"odinger-Poisson equations, the shapes of electron envelope functions are extensively investigated by clarifying their relationship with gate configuration, channel thickness, dielectric permittivity, and electron density. The impacts of electron distribution variation on various carrier scattering matrix elements and overall carrier mobility are insightfully clarified. It is found that the carrier mobility can be generally enhanced in the dual-gated configuration due to the centralization of carrier redistribution in the nanometer-thick semiconductor channels and the rate of increase reaches up to 23% in HfO$_2$ dual-gated 10-layer MoS$_2$ channels. This finding represents a viable strategy for performance optimization in transistors consisting of 2D semiconductors."
                    },
                    {
                        "title": "A nanogapped hysteresis-free field-effect transistor",
                        "abstract": "We propose a semi-suspended device structure and construct nanogapped, hysteresis-free field-effect transistors (FETs), based on the van der Waals stacking technique. The structure, which features a semi-suspended channel above a submicron-long wedge-like nanogap, is fulfilled by transferring ultraclean BN-supported MoS$_2$ channels directly onto dielectric-spaced vertical source/drain stacks. Electronic characterization and analyses reveal a high overall device quality, including ultraclean channel interfaces, negligible electrical scanning hysteresis, and Ohmic contacts in the structures. The unique hollow FET structure holds the potential for exploiting reliable electronics, as well as nanofluid and pressure sensors."
                    },
                    {
                        "title": "Towards Open-set Camera 3D Object Detection",
                        "abstract": "Traditional camera 3D object detectors are typically trained to recognize a predefined set of known object classes. In real-world scenarios, these detectors may encounter unknown objects outside the training categories and fail to identify them correctly. To address this gap, we present OS-Det3D (Open-set Camera 3D Object Detection), a two-stage training framework enhancing the ability of camera 3D detectors to identify both known and unknown objects. The framework involves our proposed 3D Object Discovery Network (ODN3D), which is specifically trained using geometric cues such as the location and scale of 3D boxes to discover general 3D objects. ODN3D is trained in a class-agnostic manner, and the provided 3D object region proposals inherently come with data noise. To boost accuracy in identifying unknown objects, we introduce a Joint Objectness Selection (JOS) module. JOS selects the pseudo ground truth for unknown objects from the 3D object region proposals of ODN3D by combining the ODN3D objectness and camera feature attention objectness. Experiments on the nuScenes and KITTI datasets demonstrate the effectiveness of our framework in enabling camera 3D detectors to successfully identify unknown objects while also improving their performance on known objects."
                    }
                ]
            },
            "cef36eb0-d53c-4736-a4bf-3e9514ecbf3f": {
                "pk": "cef36eb0-d53c-4736-a4bf-3e9514ecbf3f",
                "name": "Jiajun Bu",
                "collaborators": [
                    "Can Wang",
                    "Sheng Zhou",
                    "Kui Zhao",
                    "Ning Ma",
                    "Zhen Zhang",
                    "Xia Hu",
                    "Bangpeng Li",
                    "Zilun Peng",
                    "Ping Li",
                    "Chun Chen"
                ],
                "domain": [
                    "Domain Adaptation",
                    "Few-shot Learning",
                    "Heterogeneous Information Network",
                    "Co-clustering"
                ],
                "publications": [
                    {
                        "title": "Uncertainty-Guided Mixup for Semi-Supervised Domain Adaptation without Source Data",
                        "abstract": "Present domain adaptation methods usually perform explicit representation alignment by simultaneously accessing the source data and target data. However, the source data are not always available due to the privacy preserving consideration or bandwidth limitation. Source-free domain adaptation aims to solve the above problem by performing domain adaptation without accessing the source data. The adaptation paradigm is receiving more and more attention in recent years, and multiple works have been proposed for unsupervised source-free domain adaptation. However, without utilizing any supervised signal and source data at the adaptation stage, the optimization of the target model is unstable and fragile. To alleviate the problem, we focus on semi-supervised domain adaptation under source-free setting. More specifically, we propose uncertainty-guided Mixup to reduce the representation's intra-domain discrepancy and perform inter-domain alignment without directly accessing the source data. Finally, we conduct extensive semi-supervised domain adaptation experiments on various datasets. Our method outperforms the recent semi-supervised baselines and the unsupervised variant also achieves competitive performance. The experiment codes will be released in the future."
                    },
                    {
                        "title": "Deep Style Match for Complementary Recommendation",
                        "abstract": "Humans develop a common sense of style compatibility between items based on their attributes. We seek to automatically answer questions like \"Does this shirt go well with that pair of jeans?\" In order to answer these kinds of questions, we attempt to model human sense of style compatibility in this paper. The basic assumption of our approach is that most of the important attributes for a product in an online store are included in its title description. Therefore it is feasible to learn style compatibility from these descriptions. We design a Siamese Convolutional Neural Network architecture and feed it with title pairs of items, which are either compatible or incompatible. Those pairs will be mapped from the original space of symbolic words into some embedded style space. Our approach takes only words as the input with few preprocessing and there is no laborious and expensive feature engineering."
                    },
                    {
                        "title": "Navigation Objects Extraction for Better Content Structure Understanding",
                        "abstract": "Existing works for extracting navigation objects from webpages focus on navigation menus, so as to reveal the information architecture of the site. However, web 2.0 sites such as social networks, e-commerce portals etc. are making the understanding of the content structure in a web site increasingly difficult. Dynamic and personalized elements such as top stories, recommended list in a webpage are vital to the understanding of the dynamic nature of web 2.0 sites. To better understand the content structure in web 2.0 sites, in this paper we propose a new extraction method for navigation objects in a webpage. Our method will extract not only the static navigation menus, but also the dynamic and personalized page-specific navigation lists. Since the navigation objects in a webpage naturally come in blocks, we first cluster hyperlinks into different blocks by exploiting spatial locations of hyperlinks, the hierarchical structure of the DOM-tree and the hyperlink density. Then we identify navigation objects from those blocks using the SVM classifier with novel features such as anchor text lengths etc. Experiments on real-world data sets with webpages from various domains and styles verified the effectiveness of our method."
                    },
                    {
                        "title": "Relational Multi-Manifold Co-Clustering",
                        "abstract": "Co-clustering targets on grouping the samples (e.g., documents, users) and the features (e.g., words, ratings) simultaneously. It employs the dual relation and the bilateral information between the samples and features. In many realworld applications, data usually reside on a submanifold of the ambient Euclidean space, but it is nontrivial to estimate the intrinsic manifold of the data space in a principled way. In this study, we focus on improving the co-clustering performance via manifold ensemble learning, which is able to maximally approximate the intrinsic manifolds of both the sample and feature spaces. To achieve this, we develop a novel co-clustering algorithm called Relational Multi-manifold Co-clustering (RMC) based on symmetric nonnegative matrix tri-factorization, which decomposes the relational data matrix into three submatrices. This method considers the intertype relationship revealed by the relational data matrix, and also the intra-type information reflected by the affinity matrices encoded on the sample and feature data distributions. Specifically, we assume the intrinsic manifold of the sample or feature space lies in a convex hull of some pre-defined candidate manifolds. We want to learn a convex combination of them to maximally approach the desired intrinsic manifold. To optimize the objective function, the multiplicative rules are utilized to update the submatrices alternatively. Besides, both the entropic mirror descent algorithm and the coordinate descent algorithm are exploited to learn the manifold coefficient vector. Extensive experiments on documents, images and gene expression data sets have demonstrated the superiority of the proposed algorithm compared to other well-established methods."
                    },
                    {
                        "title": "HAHE: Hierarchical Attentive Heterogeneous Information Network Embedding",
                        "abstract": "Heterogeneous information network (HIN) embedding has recently attracted much attention due to its effectiveness in dealing with the complex heterogeneous data. Meta path, which connects different object types with various semantic meanings, is widely used by existing HIN embedding works. However, several challenges have not been addressed so far. First, different meta paths convey different semantic meanings, while existing works assume that all nodes share same weights for meta paths and ignore the personalized preferences of different nodes on different meta paths. Second, given a meta path, nodes in HIN are connected by path instances while existing works fail to fully explore the differences between path instances that reflect nodes' preferences in the semantic space. rTo tackle the above challenges, we propose aHierarchical Attentive Heterogeneous information network Embedding (HAHE) model to capture the personalized preferences on meta paths and path instances in each semantic space. As path instances are based on a particular meta path, a hierarchical attention mechanism is naturally utilized to model the personalized preference on meta paths and path instances. Extensive experiments on several real-world datasets show that our proposed \\model model significantly outperforms the state-of-the-art methods in terms of various data mining tasks."
                    },
                    {
                        "title": "Less is More: A Closer Look at Semantic-based Few-Shot Learning",
                        "abstract": "Few-shot Learning aims to learn and distinguish new categories with a very limited number of available images, presenting a significant challenge in the realm of deep learning. Recent researchers have sought to leverage the additional textual or linguistic information of these rare categories with a pre-trained language model to facilitate learning, thus partially alleviating the problem of insufficient supervision signals. However, the full potential of the textual information and pre-trained language model have been underestimated in the few-shot learning till now, resulting in limited performance enhancements. To address this, we propose a simple but effective framework for few-shot learning tasks, specifically designed to exploit the textual information and language model. In more detail, we explicitly exploit the zero-shot capability of the pre-trained language model with the learnable prompt. And we just add the visual feature with the textual feature for inference directly without the intricate designed fusion modules in previous works. Additionally, we apply the self-ensemble and distillation to further enhance these components. Our extensive experiments conducted across four widely used few-shot datasets demonstrate that our simple framework achieves impressive results. Particularly noteworthy is its outstanding performance in the 1-shot learning task, surpassing state-of-the-art methods by an average of 3.0\\% in classification accuracy. \\footnote{We will make the source codes of the proposed framework publicly available upon acceptance. }."
                    }
                ]
            },
            "d12ba6b9-5a93-480c-a8e6-ee6a62e1f5eb": {
                "pk": "d12ba6b9-5a93-480c-a8e6-ee6a62e1f5eb",
                "name": "Bingsheng He",
                "collaborators": [
                    "Zining Zhang",
                    "Zhenjie Zhang",
                    "Xiaoming Yuan",
                    "Qinbin Li",
                    "Zhaomin Wu",
                    "Shengjie Xu",
                    "Dawn Song",
                    "Jianlong Zhong",
                    "Jiong He",
                    "Mian Lu"
                ],
                "domain": [
                    "Convex Optimization",
                    "Federated Learning",
                    "Speech Processing",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "On construction of splitting contraction algorithms in a prediction-correction framework for separable convex optimization",
                        "abstract": "In the past decade, we had developed a series of splitting contraction algorithms for separable convex optimization problems, at the root of the alternating direction method of multipliers. Convergence of these algorithms was studied under specific model-tailored conditions, while these conditions can be conceptually abstracted as two generic conditions when these algorithms are all unified as a prediction-correction framework. In this paper, in turn, we showcase a constructive way for specifying the generic convergence-guaranteeing conditions, via which new splitting contraction algorithms can be generated automatically. It becomes possible to design more application-tailored splitting contraction algorithms by specifying the prediction-correction framework, while proving their convergence is a routine."
                    },
                    {
                        "title": "Balanced Augmented Lagrangian Method for Convex Programming",
                        "abstract": "We consider the convex minimization model with both linear equality and inequality constraints, and reshape the classic augmented Lagrangian method (ALM) by balancing its subproblems. As a result, one of its subproblems decouples the objective function and the coefficient matrix without any extra condition, and the other subproblem becomes a positive definite system of linear equations or a positive definite linear complementary problem. The balanced ALM advances the classic ALM by enlarging its applicable range, balancing its subproblems, and improving its implementation. We also extend our discussion to two-block and multiple-block separable convex programming models, and accordingly design various splitting versions of the balanced ALM for these separable models. Convergence analysis for the balanced ALM and its splitting versions is conducted in the context of variational inequalities through the lens of the classic proximal point algorithm."
                    },
                    {
                        "title": "Kernelet: High-Throughput GPU Kernel Executions with Dynamic Slicing and Scheduling",
                        "abstract": "Graphics processors, or GPUs, have recently been widely used as accelerators in the shared environments such as clusters and clouds. In such shared environments, many kernels are submitted to GPUs from different users, and throughput is an important metric for performance and total ownership cost. Despite the recently improved runtime support for concurrent GPU kernel executions, the GPU can be severely underutilized, resulting in suboptimal throughput. In this paper, we propose Kernelet, a runtime system with dynamic slicing and scheduling techniques to improve the throughput of concurrent kernel executions on the GPU. With slicing, Kernelet divides a GPU kernel into multiple sub-kernels (namely slices). Each slice has tunable occupancy to allow co-scheduling with other slices and to fully utilize the GPU resources. We develop a novel and effective Markov chain based performance model to guide the scheduling decision. Our experimental results demonstrate up to 31.1% and 23.4% performance improvement on NVIDIA Tesla C2050 and GTX680 GPUs, respectively."
                    },
                    {
                        "title": "Revisiting Co-Processing for Hash Joins on the Coupled CPU-GPU Architecture",
                        "abstract": "Query co-processing on graphics processors (GPUs) has become an effective means to improve the performance of main memory databases. However, the relatively low bandwidth and high latency of the PCI-e bus are usually bottleneck issues for co-processing. Recently, coupled CPU-GPU architectures have received a lot of attention, e.g. AMD APUs with the CPU and the GPU integrated into a single chip. That opens up new opportunities for optimizing query co-processing. In this paper, we experimentally revisit hash joins, one of the most important join algorithms for main memory databases, on a coupled CPU-GPU architecture. Particularly, we study the fine-grained co-processing mechanisms on hash joins with and without partitioning. The co-processing outlines an interesting design space. We extend existing cost models to automatically guide decisions on the design space. Our experimental results on a recent AMD APU show that (1) the coupled architecture enables fine-grained co-processing and cache reuses, which are inefficient on discrete CPU-GPU architectures; (2) the cost model can automatically guide the design and tuning knobs in the design space; (3) fine-grained co-processing achieves up to 53%, 35% and 28% performance improvement over CPU-only, GPU-only and conventional CPU-GPU co-processing, respectively. We believe that the insights and implications from this study are initial yet important for further research on query co-processing on coupled CPU-GPU architectures."
                    },
                    {
                        "title": "A generalized primal-dual algorithm with improved convergence condition for saddle point problems",
                        "abstract": "We generalize the well-known primal-dual algorithm proposed by Chambolle and Pock for saddle point problems, and improve the condition for ensuring its convergence. The improved convergence-guaranteeing condition is effective for the generic setting, and it is shown to be optimal. It also allows us to discern larger step sizes for the resulting subproblems, and thus provides a simple and universal way to improve numerical performance of the original primal-dual algorithm. In addition, we present a structure-exploring heuristic to further relax the convergence-guaranteeing condition for some specific saddle point problems, which could yield much larger step sizes and hence significantly better performance. Effectiveness of this heuristic is numerically illustrated by the classic assignment problem."
                    },
                    {
                        "title": "TransMask: A Compact and Fast Speech Separation Model Based on Transformer",
                        "abstract": "Speech separation is an important problem in speech processing, which targets to separate and generate clean speech from a mixed audio containing speech from different speakers. Empowered by the deep learning technologies over sequence-to-sequence domain, recent neural speech separation models are now capable of generating highly clean speech audios. To make these models more practical by reducing the model size and inference time while maintaining high separation quality, we propose a new transformer-based speech separation approach, called TransMask. By fully un-leashing the power of self-attention on long-term dependency exception, we demonstrate the size of TransMask is more than 60% smaller and the inference is more than 2 times faster than state-of-the-art solutions. TransMask fully utilizes the parallelism during inference, and achieves nearly linear inference time within reasonable input audio lengths. It also outperforms existing solutions on output speech audio quality, achieving SDR above 16 over Librimix benchmark."
                    },
                    {
                        "title": "Extensions of ADMM for Separable Convex Optimization Problems with Linear Equality or Inequality Constraints",
                        "abstract": "The alternating direction method of multipliers (ADMM) proposed by Glowinski and Marrocco is a benchmark algorithm for two-block separable convex optimization problems with linear equality constraints. It has been modified, specified, and generalized from various perspectives to tackle more concrete or complicated application problems. Despite its versatility and phenomenal popularity, it remains unknown whether or not the ADMM can be extended to separable convex optimization problems with linear inequality constraints. In this paper, we lay down the foundation of how to extend the ADMM to two-block and multiple-block (more than two blocks) separable convex optimization problems with linear inequality constraints. From a high-level and methodological perspective, we propose a unified framework of algorithmic design and a roadmap for convergence analysis in the context of variational inequalities, based on which it is possible to design a series of concrete ADMM-based algorithms with provable convergence in the prediction-correction structure. The proposed algorithmic framework and roadmap for convergence analysis are eligible to various convex optimization problems with different degrees of separability, in which both linear equality and linear inequality constraints can be included. The analysis is comprehensive yet can be presented by elementary mathematics, and hence generically understandable."
                    },
                    {
                        "title": "Practical One-Shot Federated Learning for Cross-Silo Setting",
                        "abstract": "Federated learning enables multiple parties to collaboratively learn a model without exchanging their data. While most existing federated learning algorithms need many rounds to converge, one-shot federated learning (i.e., federated learning with a single communication round) is a promising approach to make federated learning applicable in cross-silo setting in practice. However, existing one-shot algorithms only support specific models and do not provide any privacy guarantees, which significantly limit the applications in practice. In this paper, we propose a practical one-shot federated learning algorithm named FedKT. By utilizing the knowledge transfer technique, FedKT can be applied to any classification models and can flexibly achieve differential privacy guarantees. Our experiments on various tasks show that FedKT can significantly outperform the other state-of-the-art federated learning algorithms with a single communication round."
                    },
                    {
                        "title": "X-TaSNet: Robust and Accurate Time-Domain Speaker Extraction Network",
                        "abstract": "Extracting the speech of a target speaker from mixed audios, based on a reference speech from the target speaker, is a challenging yet powerful technology in speech processing. Recent studies of speaker-independent speech separation, such as TasNet, have shown promising results by applying deep neural networks over the time-domain waveform. Such separation neural network does not directly generate reliable and accurate output when target speakers are specified, because of the necessary prior on the number of speakers and the lack of robustness when dealing with audios with absent speakers. In this paper, we break these limitations by introducing a new speaker-aware speech masking method, called X-TaSNet. Our proposal adopts new strategies, including a distortion-based loss and corresponding alternating training scheme, to better address the robustness issue. X-TaSNet significantly enhances the extracted speech quality, doubling SDRi and SI-SNRi of the output speech audio over state-of-the-art voice filtering approach. X-TaSNet also improves the reliability of the results by improving the accuracy of speaker identity in the output audio to 95.4%, such that it returns silent audios in most cases when the target speaker is absent. These results demonstrate X-TaSNet moves one solid step towards more practical applications of speaker extraction technology."
                    },
                    {
                        "title": "GAZEV: GAN-Based Zero-Shot Voice Conversion over Non-parallel Speech Corpus",
                        "abstract": "Non-parallel many-to-many voice conversion is recently attract-ing huge research efforts in the speech processing community. A voice conversion system transforms an utterance of a source speaker to another utterance of a target speaker by keeping the content in the original utterance and replacing by the vocal features from the target speaker. Existing solutions, e.g., StarGAN-VC2, present promising results, only when speech corpus of the engaged speakers is available during model training. AUTOVCis able to perform voice conversion on unseen speakers, but it needs an external pretrained speaker verification model. In this paper, we present our new GAN-based zero-shot voice conversion solution, called GAZEV, which targets to support unseen speakers on both source and target utterances. Our key technical contribution is the adoption of speaker embedding loss on top of the GAN framework, as well as adaptive instance normalization strategy, in order to address the limitations of speaker identity transfer in existing solutions. Our empirical evaluations demonstrate significant performance improvement on output speech quality and comparable speaker similarity to AUTOVC."
                    },
                    {
                        "title": "DBL: Efficient Reachability Queries on Dynamic Graphs (Complete Version)",
                        "abstract": "Reachability query is a fundamental problem on graphs, which has been extensively studied in academia and industry. Since graphs are subject to frequent updates in many applications, it is essential to support efficient graph updates while offering good performance in reachability queries. Existing solutions compress the original graph with the Directed Acyclic Graph (DAG) and propose efficient query processing and index update techniques. However, they focus on optimizing the scenarios where the Strong Connected Components(SCCs) remain unchanged and have overlooked the prohibitively high cost of the DAG maintenance when SCCs are updated. In this paper, we propose DBL, an efficient DAG-free index to support the reachability query on dynamic graphs with insertion-only updates. DBL builds on two complementary indexes: Dynamic Landmark (DL) label and Bidirectional Leaf (BL) label. The former leverages landmark nodes to quickly determine reachable pairs whereas the latter prunes unreachable pairs by indexing the leaf nodes in the graph. We evaluate DBL against the state-of-the-art approaches on dynamic reachability index with extensive experiments on real-world datasets. The results have demonstrated that DBL achieves orders of magnitude speedup in terms of index update, while still producing competitive query efficiency."
                    },
                    {
                        "title": "Model-Contrastive Federated Learning",
                        "abstract": "Federated learning enables multiple parties to collaboratively train a machine learning model without communicating their local data. A key challenge in federated learning is to handle the heterogeneity of local data distribution across parties. Although many studies have been proposed to address this challenge, we find that they fail to achieve high performance in image datasets with deep learning models. In this paper, we propose MOON: model-contrastive federated learning. MOON is a simple and effective federated learning framework. The key idea of MOON is to utilize the similarity between model representations to correct the local training of individual parties, i.e., conducting contrastive learning in model-level. Our extensive experiments show that MOON significantly outperforms the other state-of-the-art federated learning algorithms on various image classification tasks."
                    },
                    {
                        "title": "A Simulation Platform for Multi-tenant Machine Learning Services on Thousands of GPUs",
                        "abstract": "Multi-tenant machine learning services have become emerging data-intensive workloads in data centers with heavy usage of GPU resources. Due to the large scale, many tuning parameters and heavy resource usage, it is usually impractical to evaluate and benchmark those machine learning services on real clusters. In this demonstration, we present AnalySIM, a cluster simulator that allows efficient design explorations for multi-tenant machine learning services. Specifically, by trace-driven cluster workload simulation, AnalySIM can easily test and analyze various scheduling policies in a number of performance metrics such as GPU resource utilization. AnalySIM simulates the cluster computational resource based on both physical topology and logical partition. The tool has been used in SenseTime to understand the impact of different scheduling policies with the trace from a real production cluster of over 1000 GPUs. We find that preemption and migration are able to significantly reduce average job completion time and mitigate the resource fragmentation problem."
                    },
                    {
                        "title": "Practical Vertical Federated Learning with Unsupervised Representation Learning",
                        "abstract": "As societal concerns on data privacy recently increase, we have witnessed data silos among multiple parties in various applications. Federated learning emerges as a new learning paradigm that enables multiple parties to collaboratively train a machine learning model without sharing their raw data. Vertical federated learning, where each party owns different features of the same set of samples and only a single party has the label, is an important and challenging topic in federated learning. Communication costs among different parties have been a major hurdle for practical vertical learning systems. In this paper, we propose a novel communication-efficient vertical federated learning algorithm named FedOnce, which requires only one-shot communication among parties. To improve model accuracy and provide privacy guarantee, FedOnce features unsupervised learning representations in the federated setting and privacy-preserving techniques based on moments accountant. The comprehensive experiments on 10 datasets demonstrate that FedOnce achieves close performance compared to state-of-the-art vertical federated learning algorithms with much lower communication costs. Meanwhile, our privacy-preserving technique significantly outperforms the state-of-the-art approaches under the same privacy budget."
                    },
                    {
                        "title": "HARL: Hierarchical Adaptive Reinforcement Learning Based Auto Scheduler for Neural Networks",
                        "abstract": "To efficiently perform inference with neural networks, the underlying tensor programs require sufficient tuning efforts before being deployed into production environments. Usually, enormous tensor program candidates need to be sufficiently explored to find the one with the best performance. This is necessary to make the neural network products meet the high demand of real-world applications such as natural language processing, auto-driving, etc. Auto-schedulers are being developed to avoid the need for human intervention. However, due to the gigantic search space and lack of intelligent search guidance, current auto-schedulers require hours to days of tuning time to find the best-performing tensor program for the entire neural network.   In this paper, we propose HARL, a reinforcement learning (RL) based auto-scheduler specifically designed for efficient tensor program exploration. HARL uses a hierarchical RL architecture in which learning-based decisions are made at all different levels of search granularity. It also automatically adjusts exploration configurations in real-time for faster performance convergence. As a result, HARL improves the tensor operator performance by 22% and the search speed by 4.3x compared to the state-of-the-art auto-scheduler. Inference performance and search speed are also significantly improved on end-to-end neural networks."
                    },
                    {
                        "title": "VertiBench: Advancing Feature Distribution Diversity in Vertical Federated Learning Benchmarks",
                        "abstract": "Vertical Federated Learning (VFL) is a crucial paradigm for training machine learning models on feature-partitioned, distributed data. However, due to privacy restrictions, few public real-world VFL datasets exist for algorithm evaluation, and these represent a limited array of feature distributions. Existing benchmarks often resort to synthetic datasets, derived from arbitrary feature splits from a global set, which only capture a subset of feature distributions, leading to inadequate algorithm performance assessment. This paper addresses these shortcomings by introducing two key factors affecting VFL performance - feature importance and feature correlation - and proposing associated evaluation metrics and dataset splitting methods. Additionally, we introduce a real VFL dataset to address the deficit in image-image VFL scenarios. Our comprehensive evaluation of cutting-edge VFL algorithms provides valuable insights for future research in the field."
                    },
                    {
                        "title": "Exploiting Label Skews in Federated Learning with Model Concatenation",
                        "abstract": "Federated Learning (FL) has emerged as a promising solution to perform deep learning on different data owners without exchanging raw data. However, non-IID data has been a key challenge in FL, which could significantly degrade the accuracy of the final model. Among different non-IID types, label skews have been challenging and common in image classification and other tasks. Instead of averaging the local models in most previous studies, we propose FedConcat, a simple and effective approach that concatenates these local models as the base of the global model to effectively aggregate the local knowledge. To reduce the size of the global model, we adopt the clustering technique to group the clients by their label distributions and collaboratively train a model inside each cluster. We theoretically analyze the advantage of concatenation over averaging by analyzing the information bottleneck of deep neural networks. Experimental results demonstrate that FedConcat achieves significantly higher accuracy than previous state-of-the-art FL methods in various heterogeneous label skew distribution settings and meanwhile has lower communication costs. Our code is publicly available at https://github.com/sjtudyq/FedConcat."
                    },
                    {
                        "title": "Aggressive Post-Training Compression on Extremely Large Language Models",
                        "abstract": "The increasing size and complexity of Large Language Models (LLMs) pose challenges for their deployment on personal computers and mobile devices. Aggressive post-training model compression is necessary to reduce the models' size, but it often results in significant accuracy loss. To address this challenge, we propose a novel network pruning technology that utilizes over 0.7 sparsity and less than 8 bits of quantization. Our approach enables the compression of prevailing LLMs within a couple of hours while maintaining a relatively small accuracy loss. In experimental evaluations, our method demonstrates effectiveness and potential for practical deployment. By making LLMs available on domestic devices, our work can facilitate a new era of natural language processing applications with wide-ranging impacts."
                    },
                    {
                        "title": "Federated Data-Efficient Instruction Tuning for Large Language Models",
                        "abstract": "Instruction tuning helps improve pretrained large language models (LLMs) in terms of the responsiveness to human instructions, which is benefited from diversified instruction data. Federated learning extends the sources of instruction data by exploiting the diversified client-side data, making it increasingly popular for tuning LLMs. Existing approaches of federated LLM tuning typically traverse all local data during local training, bringing excessive computation overhead and posing a risk of overfitting local data. Thus, a federated data-efficient instruction tuning approach, which consumes relatively little data from the entire dataset, is needed. In response, this work introduces an approach of federated data-efficient instruction tuning for LLMs, FedHDS, which utilizes a representative subset of edge-side data, coreset, to tune the LLM. It reduces the redundancy of data samples at both intra-client and inter-client levels through a hierarchical data selection framework performed by jointly selecting a small number of representative data samples for local training without sharing the raw data. Extensive experiments conducted across six scenarios with various LLMs, datasets and data partitions demonstrate that FedHDS significantly reduces the amount of data required for fine-tuning while improving the responsiveness of the instruction-tuned LLMs to unseen tasks."
                    },
                    {
                        "title": "Monetary Cost Optimizations for Hosting Workflow-as-a-Service in IaaS Clouds",
                        "abstract": "Recently, we have witnessed workflows from science and other data-intensive applications emerging on Infrastructure-asa-Service (IaaS) clouds, and many workflow service providers offering workflow as a service (WaaS). The major concern of WaaS providers is to minimize the monetary cost of executing workflows in the IaaS cloud. While there have been previous studies on this concern, most of them assume static task execution time and static pricing scheme, and have the QoS notion of satisfying a deterministic deadline. However, cloud environment is dynamic, with performance dynamics caused by the interference from concurrent executions and price dynamics like spot prices offered by Amazon EC2. Therefore, we argue that WaaS providers should have the notion of offering probabilistic performance guarantees for individual workflows on IaaS clouds. We develop a probabilistic scheduling framework called Dyna to minimize the monetary cost while offering probabilistic deadline guarantees. The framework includes an A*-based instance configuration method for performance dynamics, and a hybrid instance configuration refinement for utilizing spot instances. Experimental results with three real-world scientific workflow applications on Amazon EC2 demonstrate (1) the accuracy of our framework on satisfying the probabilistic deadline guarantees required by the users; (2) the effectiveness of our framework on reducing monetary cost in comparison with the existing approaches."
                    }
                ]
            },
            "f8d539a3-f70f-46c9-ad70-c56d558282af": {
                "pk": "f8d539a3-f70f-46c9-ad70-c56d558282af",
                "name": "Sheng Zhou",
                "collaborators": [
                    "Sheng Chen",
                    "Zhiyuan Jiang",
                    "Zhisheng Niu",
                    "Liumeng Wang",
                    "Yining Xu",
                    "Xiufeng Huang",
                    "Ruichen Jiang",
                    "Ruichen Deng",
                    "Wei Zhang"
                ],
                "domain": [
                    "Wireless Communication",
                    "Machine Learning",
                    "Edge Computing",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "Time-Sequence Channel Inference for Beam Alignment in Vehicular Networks",
                        "abstract": "In this paper, we propose a learning-based low-overhead beam alignment method for vehicle-to-infrastructure communication in vehicular networks. The main idea is to remotely infer the optimal beam directions at a target base station in future time slots, based on the CSI of a source base station in previous time slots. The proposed scheme can reduce channel acquisition and beam training overhead by replacing pilot-aided beam training with online inference from a sequence-to-sequence neural network. Simulation results based on ray-tracing channel data show that our proposed scheme achieves a $8.86\\%$ improvement over location-based beamforming schemes with a positioning error of $1$m, and is within a $4.93\\%$ performance loss compared with the genie-aided optimal beamformer."
                    },
                    {
                        "title": "Joint User Scheduling and Beam Selection Optimization for Beam-Based Massive MIMO Downlinks",
                        "abstract": "In beam-based massive multiple-input multiple-output systems, signals are processed spatially in the radio-frequency (RF) front-end and thereby the number of RF chains can be reduced to save hardware cost, power consumptions and pilot overhead. Most existing work focuses on how to select, or design analog beams to achieve performance close to full digital systems. However, since beams are strongly correlated (directed) to certain users, the selection of beams and scheduling of users should be jointly considered. In this paper, we formulate the joint user scheduling and beam selection problem based on the Lyapunov-drift optimization framework and obtain the optimal scheduling policy in a closed-form. For reduced overhead and computational cost, the proposed scheduling schemes are based only upon statistical channel state information. Towards this end, asymptotic expressions of the downlink broadcast channel capacity are derived. To address the weighted sum rate maximization problem in the Lyapunov optimization, an algorithm based on block coordinated update is proposed and proved to converge to the optimum of the relaxed problem. To further reduce the complexity, an incremental greedy scheduling algorithm is also proposed, whose performance is proved to be bounded within a constant multiplicative factor. Simulation results based on widely-used spatial channel models are given. It is shown that the proposed schemes are close to optimal, and outperform several state-of-the-art schemes."
                    },
                    {
                        "title": "On the Fronthaul Statistical Multiplexing Gain",
                        "abstract": "Breaking the fronthaul capacity limitations is vital to make cloud radio access network (C-RAN) scalable and practical. One promising way is aggregating several remote radio units (RRUs) as a cluster to share a fronthaul link, so as to enjoy the statistical multiplexing gain brought by the spatial randomness of the traffic. In this letter, a tractable model is proposed to analyze the fronthaul statistical multiplexing gain. We first derive the user blocking probability caused by the limited fronthaul capacity, including its upper and lower bounds. We then obtain the limits of fronthaul statistical multiplexing gain when the cluster size approaches infinity. Analytical results reveal that the user blocking probability decreases exponentially with the average fronthaul capacity per RRU, and the exponent is proportional to the cluster size. Numerical results further show considerable fronthaul statistical multiplexing gain even at a small to medium cluster size."
                    },
                    {
                        "title": "Flexible Functional Split and Power Control for Energy Harvesting Cloud Radio Access Networks",
                        "abstract": "Functional split is a promising technique to flexibly balance the processing cost at remote ends and the fronthaul rate in cloud radio access networks (C-RAN). By harvesting renewable energy, remote radio units (RRUs) can save grid power and be flexibly deployed. However, the randomness of energy arrival poses a major design challenge. To maximize the throughput under the average fronthaul rate constraint in C-RAN with renewable powered RRUs, we first study the offline problem of selecting the optimal functional split modes and the corresponding durations, jointly with the transmission power. We find that between successive energy arrivals, at most two functional split modes should be selected. Then the optimal online problem is formulated as an Markov decision process (MDP). To deal with the curse of dimensionality of solving MDP, we further analyze the special case with one instance of energy arrival and two candidate functional split modes as inspired by the offline solution, and then a heuristic online policy is proposed. Numerical results show that with flexible functional split, the throughput can be significantly improved compared with fixed functional split. Also, the proposed heuristic online policy has similar performance with the optimal online one, as validated by simulations."
                    },
                    {
                        "title": "On the Coverage and Capacity of Ultra-Dense Networks with Directional Transmissions",
                        "abstract": "We investigate the performance of a downlink ultra-dense network (UDN) with directional transmissions via stochastic geometry. Considering the dual-slope path loss model and sectored beamforming pattern, we derive the expressions and asymptotic characteristics of the coverage probability and constrained area spectrum efficiency (ASE). Several special scenarios, namely the physically feasible path loss model and adjustable beam pattern, are also analyzed. Although signal-to-interference-plus-noise ratio collapsing still exists when the path loss exponent in the near-field is no larger than 2, using strategies like beam pattern adaption, can avoid the decrease of the coverage probability and constrained ASE even when the base station density approaches infinity."
                    },
                    {
                        "title": "Cluster-Based Cooperative Digital Over-the-Air Aggregation for Wireless Federated Edge Learning",
                        "abstract": "In this paper, we study a federated learning system at the wireless edge that uses over-the-air computation (AirComp). In such a system, users transmit their messages over a multi-access channel concurrently to achieve fast model aggregation. Recently, an AirComp scheme based on digital modulation has been proposed featuring one-bit gradient quantization and truncated channel inversion at users and a majority-voting based decoder at the fusion center (FC). We propose an improved digital AirComp scheme to relax its requirements on the transmitters, where users perform phase correction and transmit with full power. To characterize the decoding failure probability at the FC, we introduce the normalized detection signal-to-noise ratio (SNR), which can be interpreted as the effective participation rate of users. To mitigate wireless fading, we further propose a cluster-based system and design the relay selection scheme based on the normalized detection SNR. By local data fusion within each cluster and relay selection, our scheme can fully exploit spatial diversity to increase the effective number of voting users and accelerate model convergence."
                    },
                    {
                        "title": "Channel Fingerprint Based Beam Tracking for Millimeter Wave Communications",
                        "abstract": "Beamforming structures with fixed beam codebooks provide economical solutions for millimeter wave (mmWave) communications due to the low hardware cost. However, the training overhead to search for the optimal beamforming configuration is proportional to the codebook size. To improve the efficiency of beam tracking, we propose a beam tracking scheme based on the channel fingerprint database, which comprises mappings between statistical beamforming gains and user locations. The scheme tracks user movement by utilizing the trained beam configurations and estimating the gains of beam configurations that are not trained. Simulations show that the proposed scheme achieves significant beamforming performance gains over existing beam tracking schemes."
                    },
                    {
                        "title": "Dynamic Compression Ratio Selection for Edge Inference Systems with Hard Deadlines",
                        "abstract": "Implementing machine learning algorithms on Internet of things (IoT) devices has become essential for emerging applications, such as autonomous driving, environment monitoring. But the limitations of computation capability and energy consumption make it difficult to run complex machine learning algorithms on IoT devices, especially when latency deadline exists. One solution is to offload the computation intensive tasks to the edge server. However, the wireless uploading of the raw data is time consuming and may lead to deadline violation. To reduce the communication cost, lossy data compression can be exploited for inference tasks, but may bring more erroneous inference results. In this paper, we propose a dynamic compression ratio selection scheme for edge inference system with hard deadlines. The key idea is to balance the tradeoff between communication cost and inference accuracy. By dynamically selecting the optimal compression ratio with the remaining deadline budgets for queued tasks, more tasks can be timely completed with correct inference under limited communication resources. Furthermore, information augmentation that retransmits less compressed data of task with erroneous inference, is proposed to enhance the accuracy performance. While it is often hard to know the correctness of inference, we use uncertainty to estimate the confidence of the inference, and based on that, jointly optimize the information augmentation and compression ratio selection. Lastly, considering the wireless transmission errors, we further design a retransmission scheme to reduce performance degradation due to packet losses. Simulation results show the performance of the proposed schemes under different deadlines and task arrival rates."
                    },
                    {
                        "title": "Resource Allocation for Channel Estimation in Reconfigurable Intelligent Surface-Aided Multi-Cell Networks",
                        "abstract": "Reconfigurable intelligent surface (RIS) is a promising solution to deal with the blockage-sensitivity of millimeter wave band and reduce the high energy consumption caused by network densification. However, deploying large scale RISs may not bring expected performance gain due to significant channel estimation overhead and non-negligible reflected interference. In this paper, we derive the analytical expressions of the coverage probability, area spectrum efficiency (ASE) and energy efficiency (EE) of a downlink RIS-aided multi-cell network. In order to optimize the network performance, we investigate the conditions for the optimal number of training symbols of each antenna-to-antenna and antenna-to-element path (referred to as the optimal unit training overhead) in channel estimation. Our study shows that: 1) RIS deployment is not `the more, the better', only when blockage objects are dense should one deploy more RISs; 2) the coverage probability is maximized when the unit training overhead is designed as large as possible; 3) however, the ASE-and-EE-optimal unit training overhead exists. It is a monotonically increasing function of the frame length and a monotonically decreasing function of the average signal-to-noise-ratio (in the high signal-to-noise-ratio region). Additionally, the optimal unit training overhead is smaller when communication ends deploy particularly few or many antennas."
                    },
                    {
                        "title": "Status from a Random Field: How Densely Should One Update?",
                        "abstract": "In many applications, status information of a general spatial process, in contrast to a point information source, is of interest. In this paper, we consider a system where status information is drawn from a random field and transmitted to a fusion center through a wireless multiaccess channel. The optimal density of spatial sampling points to minimize the remote status estimation error is investigated. Assuming a one-dimensional Gauss Markov random field and an exponential correlation function, closed-form expressions of remote estimation error are obtained for First-Come First-Served (FCFS) and Last-Come First-Served (LCFS) service disciplines. The optimal spatial sampling density for the LCFS case is given explicitly. Simulation results are presented which agree with our analysis."
                    },
                    {
                        "title": "QMNet: Importance-Aware Message Exchange for Decentralized Multi-Agent Reinforcement Learning",
                        "abstract": "To improve the performance of multi-agent reinforcement learning under the constraint of wireless resources, we propose a message importance metric and design an importance-aware scheduling policy to effectively exchange messages. The key insight is spending the precious communication resources on important messages. The message importance depends not only on the messages themselves, but also on the needs of agents who receive them. Accordingly, we propose a query-message-based architecture, called QMNet. Agents generate queries and messages with the environment observation. Sharing queries can help calculate message importance. Exchanging messages can help agents cooperate better. Besides, we exploit the message importance to deal with random access collisions in decentralized systems. Furthermore, a message prediction mechanism is proposed to compensate for messages that are not transmitted. Finally, we evaluate the proposed schemes in a traffic junction environment, where only a fraction of agents can send messages due to limited wireless resources. Results show that QMNet can extract valuable information to guarantee the system performance even when only $30\\%$ of agents can share messages. By exploiting message prediction, the system can further save $40\\%$ of wireless resources. The importance-aware decentralized multi-access mechanism can effectively avoid collisions, achieving almost the same performance as centralized scheduling."
                    }
                ]
            }
        }
    },
    "2406.05869": {
        "paper_data": {
            "title": "An Analysis of Elo Rating Systems via Markov Chains",
            "url": "http://arxiv.org/abs/2406.05869v1",
            "arxiv_id": "2406.05869",
            "authors": [
                "Sam Olesker-Taylor",
                "Luca Zanetti"
            ],
            "abstract": "We present a theoretical analysis of the Elo rating system, a popular method for ranking skills of players in an online setting. In particular, we study Elo under the Bradley--Terry--Luce model and, using techniques from Markov chain theory, show that Elo learns the model parameters at a rate competitive with the state of the art. We apply our results to the problem of efficient tournament design and discuss a connection with the fastest-mixing Markov chain problem.",
            "introduction": " Introduction The Elo rating system is a popular method for calculating the relative skills of players (or teams) in sports analytics and particularly chess [ 2,4,1,3]. It is based on a simple zero-sum update rule: if player ibeats player j, then the rating of player iincreases proportionally to the model probability thatiwould lose to j, while the rating of jdecreases by the same amount. This amount depends on the previously estimated difference in skills between iandj. Despite their widespread popularity, Elo rating systems still lack a rigorous theoretical understand- ing [6]. Here, we take a probabilistic approach and study Elo under the well-known Bradley\u2013Terry\u2013 Luce model (BTL) [ 11,28]. In this model, the probability pi,jthatiwins against jiswi/(wi+wj), where wkis the strength of player k. In Elo, this is usually reparametrised via wk= e\u03c1k: pi,j= e\u03c1i/(e\u03c1i+ e\u03c1j) = 1 /(1 + e\u03c1j\u2212\u03c1i) =\u03c3(\u03c1i\u2212\u03c1j), where \u03c1kis the true rating ofkand\u03c3(z):= 1/(1 + exp( \u2212z))forz\u2208Ris the sigmoid function. In this setting, after observing ibeatj, the corresponding Elo ratings xiandxjare updated as xi\u2190xi+\u03b7\u03c3(xj\u2212xi)and xj\u2190xj\u2212\u03b7\u03c3(xj\u2212xi), where the step-size \u03b7 >0is chosen by the modeller. The size of the update depends exponentially on the difference in ratings: beating a much lower rated opponent does not change the ratings much. The goal of the Elo rating system is to estimate the true ratings of nplayers by observing Results We close with Related Work The problem of designing an efficient tournament graph is related to active ranking [21]. Active ranking, however, allows one to choose which matches to schedule next after observing the related work and outline the most important parts of the proof. Throughout the paper, \u201c f\u2272g\u201d means \u201c f=O(g)\u201d, \u201cf\u226ag\u201d means \u201c f=o(g)\u201d and \u201cwhp\u201d means \u201cwith probability 1\u2212O(1/n)\u201d; also, later, E\u03c0[\u00b7]indicates that X0\u223c\u03c0. 2.1 Our experiments are indicative of a more general behaviour is due to the peculiarities of the Elo rating systems. In particular, the Elo update offers diminishing returns: if the rating of iis relatively much larger than the rating of j, ifiwins against j, the rating of iwon\u2019t increase by much. This is because the sigmoid function \u03c3approaches zero very fast. Despite this, proving that the maximum rating cannot increase significantly appears much harder. This is due to the fact that the maximum rating is not a supermartingale: if all of the neighbours of the node iwith largest true rating have abnormally large rating, the rating of iis likely to increase, no matter how large it already was. However, because the zero-sum property of the ratings, the neighbours of icannot maintain an abnormally large rating for too long. Indeed, it appears very unlikely that such balanced but \u201cabnormal\u201d configurations are ever reached. 13Figure 8: Behaviour of the largest Elo rating in absolute value for path (left) and star (right) graphs of varying size. In the two-player case, the situation is simpler: if player 1has Elo rating x1, then player 2has x2=\u2212x1, by the zero-sum nature. It is straightforward to check that the Elo ratings are biased towards the true skills: that is, if player 1has rating x1before a game, then their rating x\u2032 1after the game has E[|x\u2032 1\u2212\u03c11|]<|x1\u2212\u03c11|, if\u03b7 <1 2. Suppose that the same argument holds for n >2players: i.e., typically , Elo ratings are biased towards the true skills. Biased random walks on Rhave exponential tails. So, if \u03b7\u226a1/logn, then it is super-polynomially",
            "references": [
                {
                    "title": "Spectral gap of nonreversible Markov chains",
                    "abstract": "We define the spectral gap of a Markov chain on a finite state space as the second-smallest singular value of the generator of the chain, generalizing the usual definition of spectral gap for reversible chains. We then define the relaxation time of the chain as the inverse of this spectral gap, and show that this relaxation time can be characterized, for any Markov chain, as the time required for convergence of empirical averages. This relaxation time is related to the Cheeger constant and the mixing time of the chain through inequalities that are similar to the reversible case, and the path argument can be used to get upper bounds. Several examples are worked out. An interesting finding from the examples is that the time for convergence of empirical averages in nonreversible chains can often be substantially smaller than the mixing time."
                },
                {
                    "title": "A state-space perspective on modelling and inference for online skill rating",
                    "abstract": "\n We summarize popular methods used for skill rating in competitive sports, along with their inferential paradigms and introduce new approaches based on sequential Monte Carlo and discrete hidden Markov models. We advocate for a state-space model perspective, wherein players\u2019 skills are represented as time-varying, and match results serve as observed quantities. We explore the steps to construct the model and the three stages of inference: filtering, smoothing, and parameter estimation. We examine the challenges of scaling up to numerous players and matches, highlighting the main approximations and reductions which facilitate statistical and computational efficiency. We additionally compare approaches in a realistic experimental pipeline that can be easily reproduced and extended with our open-source Python package, abile."
                },
                {
                    "title": "Geometric Bounds on the Fastest Mixing Markov Chain",
                    "abstract": "<jats:p>In the Fastest Mixing Markov Chain problem, we are given a graph <jats:inline-formula><jats:alternatives><jats:tex-math>$$G = (V, E)$$</jats:tex-math><mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\">\n                <mml:mrow>\n                  <mml:mi>G</mml:mi>\n                  <mml:mo>=</mml:mo>\n                  <mml:mo>(</mml:mo>\n                  <mml:mi>V</mml:mi>\n                  <mml:mo>,</mml:mo>\n                  <mml:mi>E</mml:mi>\n                  <mml:mo>)</mml:mo>\n                </mml:mrow>\n              </mml:math></jats:alternatives></jats:inline-formula> and desire the discrete-time Markov chain with smallest mixing time <jats:inline-formula><jats:alternatives><jats:tex-math>$$\\tau $$</jats:tex-math><mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\">\n                <mml:mi>\u03c4</mml:mi>\n              </mml:math></jats:alternatives></jats:inline-formula> subject to having equilibrium distribution uniform on <jats:italic>V</jats:italic> and non-zero transition probabilities only across edges of the graph. It is well-known that the mixing time <jats:inline-formula><jats:alternatives><jats:tex-math>$$\\tau _\\textsf {RW}$$</jats:tex-math><mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\">\n                <mml:msub>\n                  <mml:mi>\u03c4</mml:mi>\n                  <mml:mi>RW</mml:mi>\n                </mml:msub>\n              </mml:math></jats:alternatives></jats:inline-formula> of the lazy random walk on <jats:italic>G</jats:italic> is characterised by the edge conductance <jats:inline-formula><jats:alternatives><jats:tex-math>$$\\Phi $$</jats:tex-math><mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\">\n                <mml:mi>\u03a6</mml:mi>\n              </mml:math></jats:alternatives></jats:inline-formula> of <jats:italic>G</jats:italic> via Cheeger\u2019s inequality: <jats:inline-formula><jats:alternatives><jats:tex-math>$$\\Phi ^{-1} \\lesssim \\tau _\\textsf {RW} \\lesssim \\Phi ^{-2} \\log |V|$$</jats:tex-math><mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\">\n                <mml:mrow>\n                  <mml:msup>\n                    <mml:mi>\u03a6</mml:mi>\n                    <mml:mrow>\n                      <mml:mo>-</mml:mo>\n                      <mml:mn>1</mml:mn>\n                    </mml:mrow>\n                  </mml:msup>\n                  <mml:mo>\u2272</mml:mo>\n                  <mml:msub>\n                    <mml:mi>\u03c4</mml:mi>\n                    <mml:mi>RW</mml:mi>\n                  </mml:msub>\n                  <mml:mo>\u2272</mml:mo>\n                  <mml:msup>\n                    <mml:mi>\u03a6</mml:mi>\n                    <mml:mrow>\n                      <mml:mo>-</mml:mo>\n                      <mml:mn>2</mml:mn>\n                    </mml:mrow>\n                  </mml:msup>\n                  <mml:mo>log</mml:mo>\n                  <mml:mrow>\n                    <mml:mo>|</mml:mo>\n                    <mml:mi>V</mml:mi>\n                    <mml:mo>|</mml:mo>\n                  </mml:mrow>\n                </mml:mrow>\n              </mml:math></jats:alternatives></jats:inline-formula>. Analogously, we characterise the fastest mixing time <jats:inline-formula><jats:alternatives><jats:tex-math>$$\\tau ^\\star $$</jats:tex-math><mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\">\n                <mml:msup>\n                  <mml:mi>\u03c4</mml:mi>\n                  <mml:mo>\u22c6</mml:mo>\n                </mml:msup>\n              </mml:math></jats:alternatives></jats:inline-formula> via a Cheeger-type inequality but for a different geometric quantity, namely the vertex conductance <jats:inline-formula><jats:alternatives><jats:tex-math>$$\\Psi $$</jats:tex-math><mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\">\n                <mml:mi>\u03a8</mml:mi>\n              </mml:math></jats:alternatives></jats:inline-formula> of <jats:italic>G</jats:italic>: <jats:inline-formula><jats:alternatives><jats:tex-math>$$\\Psi ^{-1} \\lesssim \\tau ^\\star \\lesssim \\Psi ^{-2} (\\log |V|)^2$$</jats:tex-math><mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\">\n                <mml:mrow>\n                  <mml:msup>\n                    <mml:mi>\u03a8</mml:mi>\n                    <mml:mrow>\n                      <mml:mo>-</mml:mo>\n                      <mml:mn>1</mml:mn>\n                    </mml:mrow>\n                  </mml:msup>\n                  <mml:mo>\u2272</mml:mo>\n                  <mml:msup>\n                    <mml:mi>\u03c4</mml:mi>\n                    <mml:mo>\u22c6</mml:mo>\n                  </mml:msup>\n                  <mml:mo>\u2272</mml:mo>\n                  <mml:msup>\n                    <mml:mi>\u03a8</mml:mi>\n                    <mml:mrow>\n                      <mml:mo>-</mml:mo>\n                      <mml:mn>2</mml:mn>\n                    </mml:mrow>\n                  </mml:msup>\n                  <mml:msup>\n                    <mml:mrow>\n                      <mml:mo>(</mml:mo>\n                      <mml:mo>log</mml:mo>\n                      <mml:mo>|</mml:mo>\n                      <mml:mi>V</mml:mi>\n                      <mml:mo>|</mml:mo>\n                      <mml:mo>)</mml:mo>\n                    </mml:mrow>\n                    <mml:mn>2</mml:mn>\n                  </mml:msup>\n                </mml:mrow>\n              </mml:math></jats:alternatives></jats:inline-formula>. This characterisation forbids fast mixing for graphs with small vertex conductance. To bypass this fundamental barrier, we consider Markov chains on <jats:italic>G</jats:italic> with equilibrium distribution which need not be uniform, but rather only <jats:inline-formula><jats:alternatives><jats:tex-math>$$\\varepsilon $$</jats:tex-math><mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\">\n                <mml:mi>\u03b5</mml:mi>\n              </mml:math></jats:alternatives></jats:inline-formula>-close to uniform in total variation. We show that it is always possible to construct such a chain with mixing time <jats:inline-formula><jats:alternatives><jats:tex-math>$$\\tau \\lesssim \\varepsilon ^{-1} ({\\text {diam}} G)^2 \\log |V|$$</jats:tex-math><mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\">\n                <mml:mrow>\n                  <mml:mi>\u03c4</mml:mi>\n                  <mml:mo>\u2272</mml:mo>\n                  <mml:msup>\n                    <mml:mi>\u03b5</mml:mi>\n                    <mml:mrow>\n                      <mml:mo>-</mml:mo>\n                      <mml:mn>1</mml:mn>\n                    </mml:mrow>\n                  </mml:msup>\n                  <mml:msup>\n                    <mml:mrow>\n                      <mml:mo>(</mml:mo>\n                      <mml:mtext>diam</mml:mtext>\n                      <mml:mi>G</mml:mi>\n                      <mml:mo>)</mml:mo>\n                    </mml:mrow>\n                    <mml:mn>2</mml:mn>\n                  </mml:msup>\n                  <mml:mo>log</mml:mo>\n                  <mml:mrow>\n                    <mml:mo>|</mml:mo>\n                    <mml:mi>V</mml:mi>\n                    <mml:mo>|</mml:mo>\n                  </mml:mrow>\n                </mml:mrow>\n              </mml:math></jats:alternatives></jats:inline-formula>. Finally, we discuss analogous questions for continuous-time and time-inhomogeneous chains.</jats:p>"
                },
                {
                    "title": "Generalized Results for the Existence and Consistency of the MLE in the Bradley-Terry-Luce Model",
                    "abstract": "Ranking problems based on pairwise comparisons, such as those arising in online gaming, often involve a large pool of items to order. In these situations, the gap in performance between any two items can be significant, and the smallest and largest winning probabilities can be very close to zero or one. Furthermore, each item may be compared only to a subset of all the items, so that not all pairwise comparisons are observed. In this paper, we study the performance of the Bradley-Terry-Luce model for ranking from pairwise comparison data under more realistic settings than those considered in the literature so far. In particular, we allow for near-degenerate winning probabilities and arbitrary comparison designs. We obtain novel results about the existence of the maximum likelihood estimator (MLE) and the corresponding $\\ell_2$ estimation error without the bounded winning probability assumption commonly used in the literature and for arbitrary comparison graph topologies. Central to our approach is the reliance on the Fisher information matrix to express the dependence on the graph topologies and the impact of the values of the winning probabilities on the estimation risk and on the conditions for the existence of the MLE. Our bounds recover existing results as special cases but are more broadly applicable."
                },
                {
                    "title": "$\\ell_{\\infty}$-Bounds of the MLE in the BTL Model under General Comparison Graphs",
                    "abstract": "The Bradley-Terry-Luce (BTL) model is a popular statistical approach for estimating the global ranking of a collection of items using pairwise comparisons. To ensure accurate ranking, it is essential to obtain precise estimates of the model parameters in the $\\ell_{\\infty}$-loss. The difficulty of this task depends crucially on the topology of the pairwise comparison graph over the given items. However, beyond very few well-studied cases, such as the complete and Erd\\\"os-R\\'enyi comparison graphs, little is known about the performance of the maximum likelihood estimator MLE) of the BTL model parameters in the $\\ell_{\\infty}$-loss under more general graph topologies. In this paper, we derive novel, general upper bounds on the $\\ell_{\\infty}$ estimation error of the BTL MLE that depend explicitly on the algebraic connectivity of the comparison graph, the maximal performance gap across items and the sample complexity. We demonstrate that the derived bounds perform well and in some cases are sharper compared to known results obtained using different loss functions and more restricted assumptions and graph topologies. We carefully compare our results to Yan et al. (2012), which is closest in spirit to our work. We further provide minimax lower bounds under $\\ell_{\\infty}$-error that nearly match the upper bounds over a class of sufficiently regular graph topologies. Finally, we study the implications of our $\\ell_{\\infty}$-bounds for efficient (offline) tournament design. We illustrate and discuss our findings through various examples and simulations."
                },
                {
                    "title": "Zig-Zag Sampling for Discrete Structures and Nonreversible Phylogenetic MCMC",
                    "abstract": "Abstract We construct a zig-zag process targeting a posterior distribution defined on a hybrid state space consisting of both discrete and continuous variables. The construction does not require any assumptions on the structure among discrete variables. We demonstrate our method on two examples in genetics based on the Kingman coalescent, showing that the zig-zag process can lead to efficiency gains of up to several orders of magnitude over classical Metropolis\u2013Hastings algorithms, and that it is well suited to parallel computation. Our construction resembles existing techniques for Hamiltonian Monte Carlo on a hybrid state space, which suffers from implementationally and analytically complex boundary crossings when applied to the coalescent. We demonstrate that the continuous-time zig-zag process avoids these complications. Supplementary materials for this article are available online."
                },
                {
                    "title": "Accelerated Spectral Ranking",
                    "abstract": "The problem of rank aggregation from pairwise and multiway comparisons has a wide range of implications, ranging from recommendation systems to sports rankings to social choice. Some of the most popular algorithms for this problem come from the class of spectral ranking algorithms; these include the rank centrality algorithm for pairwise comparisons, which returns consistent estimates under the Bradley-Terry-Luce (BTL) model for pairwise comparisons (Negahban et al., 2017), and its generalization, the Luce spectral ranking algorithm, which returns consistent estimates under the more general multinomial logit (MNL) model for multiway comparisons (Maystre & Grossglauser, 2015). In this paper, we design a provably faster spectral ranking algorithm, which we call accelerated spectral ranking (ASR), that is also consistent under the MNL/BTL models. Our accelerated algorithm is achieved by designing a random walk that has a faster mixing time than the random walks associated with previous algorithms. In addition to a faster algorithm, our results yield improved sample complexity bounds for recovery of the MNL/BTL parameters: to the best of our knowledge, we give the first general sample complexity bounds for recovering the parameters of the MNL model from multiway comparisons under any (connected) comparison graph (and improve significantly over previous bounds for the BTL model for pairwise comparisons). We also give a message-passing interpretation of our algorithm, which suggests a decentralized distributed implementation. Our experiments on several real world and synthetic datasets confirm that our new ASR algorithm is indeed orders of magnitude faster than existing algorithms. Department of Computer and Information Science, University of Pennsylvania, Philadelphia, USA. Correspondence to: Arpit Agarwal , Shivani Agarwal . Proceedings of the 35 th International Conference on Machine Learning, Stockholm, Sweden, PMLR 80, 2018. Copyright 2018 by the author(s)."
                },
                {
                    "title": "Elo Ratings and the Sports Model: A Neglected Topic in Applied Probability?",
                    "abstract": "Author(s): Aldous, D | Abstract: \u00a9 Institute of Mathematical Statistics, 2017. In a simple model for sports, the probability A beats B is a specified function of their difference in strength. One might think this would be a staple topic in Applied Probability textbooks (like the Galton-Watson branching process model, for instance) but it is curiously absent. Our first purpose is to point out that the model suggests a wide range of questions, suitable for \"undergraduate research\" via simulation but also challenging as professional research. Our second, more specific, purpose concerns Elo-type rating algorithms for tracking changing strengths. There has been little foundational research on their accuracy, despite a much-copied \"30 matches suffice\" claim, which our simulation study casts doubt upon."
                },
                {
                    "title": "Bridging the gap between constant step size stochastic gradient descent and Markov chains",
                    "abstract": "We consider the minimization of an objective function given access to unbiased estimates of its gradient through stochastic gradient descent (SGD) with constant step-size. While the detailed analysis was only performed for quadratic functions, we provide an explicit asymptotic expansion of the moments of the averaged SGD iterates that outlines the dependence on initial conditions, the effect of noise and the step-size, as well as the lack of convergence in the general (non-quadratic) case. For this analysis, we bring tools from Markov chain theory into the analysis of stochastic gradient. We then show that Richardson-Romberg extrapolation may be used to get closer to the global optimum and we show empirical improvements of the new extrapolation scheme."
                },
                {
                    "title": "Lifting Markov Chains To Mix Faster: Limits and Opportunities",
                    "abstract": "Lifted Markov chains are Markov chains on graphs with added local \"memory\" and can be used to mix towards a target distribution faster than their memoryless counterparts. Upper and lower bounds on the achievable performance have been provided under specific assumptions. In this paper, we analyze which assumptions and constraints are relevant for mixing, and how changing these assumptions affects those bounds. Explicitly, we find that requesting mixing on either the original or the full lifted graph, and allowing for reducible lifted chains or not, have no essential influence on mixing time bounds. On the other hand, allowing for suitable initialization of the lifted dynamics and/or imposing invariance of the target distribution for any initialization do significantly affect the convergence performance. The achievable convergence speed for a lifted chain goes from diameter-time to no acceleration over a standard Markov chain, with conductance bounds limiting the effectiveness of the intermediate cases. In addition, we show that the relevance of imposing ergodic flows depends on the other criteria. The presented analysis allows us to clarify in which scenarios designing lifted dynamics can lead to better mixing, and provide a flexible framework to compare lifted walks with other acceleration methods."
                },
                {
                    "title": "Active ranking from pairwise comparisons and when parametric assumptions do not help",
                    "abstract": "We consider sequential or active ranking of a set of n items based on noisy pairwise comparisons. Items are ranked according to the probability that a given item beats a randomly chosen item, and ranking refers to partitioning the items into sets of pre-specified sizes according to their scores. This notion of ranking includes as special cases the identification of the top-k items and the total ordering of the items. We first analyze a sequential ranking algorithm that counts the number of comparisons won, and uses these counts to decide whether to stop, or to compare another pair of items, chosen based on confidence intervals specified by the data collected up to that point. We prove that this algorithm succeeds in recovering the ranking using a number of comparisons that is optimal up to logarithmic factors. This guarantee does not require any structural properties of the underlying pairwise probability matrix, unlike a significant body of past work on pairwise ranking based on parametric models such as the Thurstone or Bradley-Terry-Luce models. It has been a long-standing open question as to whether or not imposing these parametric assumptions allows for improved ranking algorithms. For stochastic comparison models, in which the pairwise probabilities are bounded away from zero, our second contribution is to resolve this issue by proving a lower bound for parametric models. This shows, perhaps surprisingly, that these popular parametric modeling choices offer at most logarithmic gains for stochastic comparisons."
                },
                {
                    "title": "Estimation from Pairwise Comparisons: Sharp Minimax Bounds with Topology Dependence",
                    "abstract": "Data in the form of pairwise comparisons arises in many domains, including preference elicitation, sporting competitions, and peer grading among others. We consider parametric ordinal models for such pairwise comparison data involving a latent vector w* e Rd that represents the \"qualities\" of the d items being compared; this class of models includes the two most widely used parametric models|the Bradley-Terry-Luce (BTL) and the Thurstone models. Working within a standard minimax framework, we provide tight upper and lower bounds on the optimal error in estimating the quality score vector w* under this class of models. The bounds depend on the topology of the comparison graph induced by the subset of pairs being compared, via the spectrum of the Laplacian of the comparison graph. Thus, in settings where the subset of pairs may be chosen, our results provide principled guidelines for making this choice. Finally, we compare these error rates to those under cardinal measurement models and show that the error rates in the ordinal and cardinal settings have identical scalings apart from constant pre-factors."
                },
                {
                    "title": "Minimax-optimal Inference from Partial Rankings",
                    "abstract": "This paper studies the problem of rank aggregation under the Plackett-Luce model. The goal is to infer a global ranking and related scores of the items, based on partial rankings provided by multiple users over multiple subsets of items. A question of particular interest is how to optimally assign items to users for ranking and how many item assignments are needed to achieve a target estimation error. Without any assumptions on how the items are assigned to users, we derive an oracle lower bound and the Cramer-Rao lower bound of the estimation error. We prove an upper bound on the estimation error achieved by the maximum likelihood estimator, and show that both the upper bound and the Cramer-Rao lower bound inversely depend on the spectral gap of the Laplacian of an appropriately defined comparison graph. Since random comparison graphs are known to have large spectral gaps, this suggests the use of random assignments when we have the control. Precisely, the matching oracle lower bound and the upper bound on the estimation error imply that the maximum likelihood estimator together with a random assignment is minimax-optimal up to a logarithmic factor. We further analyze a popular rank-breaking scheme that decompose partial rankings into pairwise comparisons. We show that even if one applies the mismatched maximum likelihood estimator that assumes independence (on pairwise comparisons that are now dependent due to rank-breaking), minimax optimal performance is still achieved up to a logarithmic factor."
                },
                {
                    "title": "Concentration inequalities for Markov chains by Marton couplings and spectral methods",
                    "abstract": "We prove a version of McDiarmid\u2019s bounded differences inequality for Markov chains, with constants proportional to the mixing time of the chain. We also show variance bounds and Bernstein-type inequalities for empirical averages of Markov chains. In the case of non-reversible chains, we introduce a new quantity called the \u201cpseudo spectral gap\u201d, and show that it plays a similar role for non-reversible chains as the spectral gap plays for reversible chains. Our techniques for proving these results are based on a coupling construction of Katalin Marton, and on spectral techniques due to Pascal Lezaud. The pseudo spectral gap generalises the multiplicative reversiblication approach of Jim Fill."
                },
                {
                    "title": "Rank Centrality: Ranking from Pairwise Comparisons",
                    "abstract": "The question of aggregating pairwise comparisons to obtain a global ranking over a collection of objects has been of interest for a very long time: be it ranking of online gamers (e.g., MSR\u2019s TrueSkill system) and chess players, aggregating social opinions, or deciding which product to sell based on transactions. In most settings, in addition to obtaining a ranking, finding \u2018scores\u2019 for each object (e.g., player\u2019s rating) is of interest for understanding the intensity of the preferences.In this paper, we propose Rank Centrality , an iterative rank aggregation algorithm for discovering scores for objects (or items) from pairwise comparisons. The algorithm has a natural random walk interpretation over the graph of objects with an edge present between a pair of objects if they are compared; the score, which we call Rank Centrality, of an object turns out to be its stationary probability under this random walk.To study the efficacy of the algorithm, we consider the popular Bradley-Terry-Luce (BTL) model (equivalent to the Multinomial Logit (MNL) for pairwise comparisons) in which each object has an associated score that determines the probabilistic outcomes of pairwise comparisons between objects. In terms of the pairwise marginal probabilities, which is the main subject of this paper, the MNL model and the BTL model are identical. We bound the finite sample error rates between the scores assumed by the BTL model and those estimated by our algorithm. In particular, the number of samples required to learn the score well with high probability depends on the structure of the comparison graph. When the Laplacian of the comparison graph has a strictly positive spectral gap, e.g., each item is compared to a subset of randomly chosen items, this leads to dependence on the number of samples that is nearly order optimal.Experimental evaluations on synthetic data sets generated according to the BTL model show that our algorithm performs as well as the maximum likelihood estimator for that model and outperforms other popular ranking algorithms."
                },
                {
                    "title": "Chernoff-Hoeffding Bounds for Markov Chains: Generalized and Simplified",
                    "abstract": "We prove the first Chernoff-Hoeffding bounds for general nonreversible finite-state Markov chains based on the standard L_1 (variation distance) mixing-time of the chain. Specifically, consider an ergodic Markov chain M and a weight function f: [n] -> [0,1] on the state space [n] of M with mean mu = E_{v = delta mu t ], is at most exp(-Omega(delta^2 mu t / T)) for 0 1. In fact, the bounds hold even if the weight functions f_i's for i in [t] are distinct, provided that all of them have the same mean mu. \nWe also obtain a simplified proof for the Chernoff-Hoeffding bounds based on the spectral expansion lambda of M, which is the square root of the second largest eigenvalue (in absolute value) of M tilde{M}, where tilde{M} is the time-reversal Markov chain of M. We show that the probability Pr [ |X - mu t| >= delta mu t ] is at most exp(-Omega(delta^2 (1-lambda) mu t)) for 0 1. \nBoth of our results extend to continuous-time Markov chains, and to the case where the walk starts from an arbitrary distribution x, at a price of a multiplicative factor depending on the distribution x in the concentration bounds."
                },
                {
                    "title": "The Methodology of Officially Recognized International Sports Rating Systems",
                    "abstract": "A comprehensive comparative survey is presented, covering official rating systems as published by internationally recognized sports federations. Mind sports and physical sports are both included. As of November 2010, competitions in 159 international sports are organized by sports federations recognized by the IOC, Sport Accord and by Wikipedia identified under \u0093List of International Sport Federations.\u0094 Of the 159 sports, 18 are combat sports in which opponents are in direct physical contact as in boxing and wrestling, 74 are independent sports in which significant contact is not allowed as in swimming and archery and 67 are object sports in which indirect contact is allowed while opponents attempt to control an object as in basketball and chess. Of the 159 sports, 60 sports have no rating system, two combat sports have a subjective rating system, 84 sports have an accumulative system in which points accrue non-decreasingly over some window of time, and 13 sports have an adjustive system in which a rating self adjusts based on the difference between some observed result and a prediction of that result based on past performance. For accumulative rating systems, features include converting results to points, ageing results more than one year old, and possibly adjusting points using other performance measures. Such systems are favored by tournament organizers who want to encourage many top competitors to enter as for skiing and tennis. The adjustive systems include Elo, probit and averaging methods. These systems are favored for their technical sophistication by sports such as chess, draughts, go, cricket, and women\u0092s soccer. This study thus identifies the observed successful methodology used by the various sports federations to publish comparative ratings. Predictive success of certain rating systems are tabulated for FIBA world championship basketball, Grand Slam tennis and FIFA men\u0092s world cup soccer."
                },
                {
                    "title": "A new Poisson-type deviation inequality for Markov jump processes with positive Wasserstein curvature",
                    "abstract": "The purpose of this paper is to extend the investigation of Poisson-type deviation inequalities started by Joulin (Bernoulli 13 (2007) 782--798) to the empirical mean of positively curved Markov jump processes. In particular, our main result generalizes the tail estimates given by Lezaud (Ann. Appl. Probab. 8 (1998) 849--867, ESAIM Probab. Statist. 5 (2001) 183--201). An application to birth--death processes completes this work."
                },
                {
                    "title": "CURVATURE, CONCENTRATION AND ERROR ESTIMATES FOR MARKOV CHAIN MONTE CARLO",
                    "abstract": "We provide explicit nonasymptotic estimates for the rate of convergence of empirical means of Markov chains, together with a Gaussian or exponential control on the deviations of empirical means. These estimates hold under a \"positive curvature\" assumption expressing a kind of metric ergodicity, which generalizes the Ricci curvature from differential geometry and, on finite graphs, amounts to contraction under path coupling."
                },
                {
                    "title": "The Fastest Mixing Markov Process on a Graph and a Connection to a Maximum Variance Unfolding Problem",
                    "abstract": "We consider a Markov process on a connected graph, with edges labeled with transition rates between the adjacent vertices. The distribution of the Markov process converges to the uniform distribution at a rate determined by the second smallest eigenvalue $\\lambda_2$ of the Laplacian of the weighted graph. In this paper we consider the problem of assigning transition rates to the edges so as to maximize $\\lambda_2$ subject to a linear constraint on the rates. This is the problem of finding the fastest mixing Markov process (FMMP) on the graph. We show that the FMMP problem is a convex optimization problem, which can in turn be expressed as a semidefinite program, and therefore effectively solved numerically. We formulate a dual of the FMMP problem and show that it has a natural geometric interpretation as a maximum variance unfolding (MVU) problem, , the problem of choosing a set of points to be as far apart as possible, measured by their variance, while respecting local distance constraints. This MVU problem is closely related to a problem recently proposed by Weinberger and Saul as a method for \u201cunfolding\u201d high-dimensional data that lies on a low-dimensional manifold. The duality between the FMMP and MVU problems sheds light on both problems, and allows us to characterize and, in some cases, find optimal solutions."
                },
                {
                    "title": "Improving Asymptotic Variance of MCMC Estimators: Non-reversible Chains are Better",
                    "abstract": "I show how any reversible Markov chain on a finite state space that is irreducible, and hence suitable for estimating expectations with respect to its invariant distribution, can be used to construct a non-reversible Markov chain on a related state space that can also be used to estimate these expectations, with asymptotic variance at least as small as that using the reversible chain (typically smaller). The non-reversible chain achieves this improvement by avoiding (to the extent possible) transitions that backtrack to the state from which the chain just came. The proof that this modification cannot increase the asymptotic variance of an MCMC estimator uses a new technique that can also be used to prove Peskun's (1973) theorem that modifying a reversible chain to reduce the probability of staying in the same state cannot increase asymptotic variance. A non-reversible chain that avoids backtracking will often take little or no more computation time per transition than the original reversible chain, and can sometime produce a large reduction in asymptotic variance, though for other chains the improvement is slight. In addition to being of some practical interest, this construction demonstrates that non-reversible chains have a fundamental advantage over reversible chains for MCMC estimation. Research into better MCMC methods may therefore best be focused on non-reversible chains."
                },
                {
                    "title": "Chernoff and Berry-Ess\u00e9en inequalities for Markov processes",
                    "abstract": "In this paper, we develop bounds on the distribution function of the empirical mean for general ergodic Markov processes having a spectral gap. Our approach is based on the perturbation theory for linear operators, following the technique introduced by Gillman."
                },
                {
                    "title": "ANALYSIS OF A NONREVERSIBLE MARKOV CHAIN SAMPLER",
                    "abstract": "We analyze the convergence to stationarity of a simple nonreversible Markov chain that serves as a model for several nonreversible Markov chain sampling methods that are used in practice. Our theoretical and numerical results show that nonreversibility can indeed lead to improvements over the diffusive behavior of simple Markov chain sampling schemes. The analysis uses both probabilistic techniques and an explicit diagonalization. 1. Introduction. Markov chain sampling methods are commonly used in statistics [33, 32], computer science [31], statistical mechanics [3] and quantum field theory [34, 23]. In all these fields, distributions are encountered that are difficult to sample from directly, but for which a Markov chain that converges to the distribution can easily be constructed. For many such methods (e.g., the Metropolis algorithm [25, 13], and the Gibbs sampler [17, 16] with a random scan) the Markov chain constructed is reversible. Some of these methods explore the distribution by means of a diffusive random walk. We use the term \u201cdiffusive\u201d for processes like the ordinary random walk on a d-dimensional lattice which require time of order T 2 to travel distance T. Some other common methods, such as the Gibbs sampler with a systematic scan, use a Markov chain that is not reversible, but have diffusive behavior resembling that of a related reversible chain [30]. Some Markov chain methods attempt to avoid the inefficiencies of such diffusive exploration. The Hybrid Monte Carlo method [15] uses an elaborate Metropolis proposal that can make large changes to the state. In a variant of this method due to Horowitz [21], a similar effect is produced using a Markov chain that is carefully designed to be nonreversible. (See [34, 23, 27] for reviews of these methods.) The overrelaxation method [1] also employs a nonreversible Markov chain as a way of suppressing diffusive behavior, as discussed in [29]. In this paper, we analyze a nonreversible Markov chain that does a onedimensional walk, as an abstraction of these practical sampling methods, particularly that of Horowitz [21]. Gustafson [19] has also recently tried using adaptations of Horowitz\u2019s method. We find that the nonreversible walk does indeed converge more rapidly than the usual simple random walk. We ana"
                },
                {
                    "title": "Lifting Markov chains to speed up mixing",
                    "abstract": "There are several examples where the mixing time of a Markov chain can be reduced substantially, often to about its square root, by \u201clifting\u201d, i.e., by splitting each state into several states. In several examples of random walks on groups, the lifted chain not only mixes better, but is easier to analyze. We characterize the best mixing time achievable through lifting in terms of multicommodity flows. We show that the reduction to square root is best possible. If the lifted chain is time-reversible, then the gain is smaller, at most a factor of log(l/na), where 110 is the smallest stationary probability of any state. We give an example showing that a gain of a factor of log(l/~o)/log log(l/rro) is possible."
                },
                {
                    "title": "Path coupling: A technique for proving rapid mixing in Markov chains",
                    "abstract": "The main technique used in algorithm design for approximating #P-hard counting problems is the Markov chain Monte Carlo method. At the heart of the method is the study of the convergence (mixing) rates of particular Markov chains of interest. In this paper we illustrate a new approach to the coupling technique, which we call path coupling, for bounding mixing rates. Previous applications of coupling have required detailed insights into the combinatorics of the problem at hand, and this complexity can make the technique extremely difficult to apply successfully. Path coupling helps to minimize the combinatorial difficulty and in all cases provides simpler convergence proofs than does the standard coupling method. However the true power of the method is that the simplification obtained may allow coupling proofs which were previously unknown, or provide significantly better bounds than those obtained using the standard method. We apply the path coupling method to several hard combinatorial problems, obtaining new or improved results. We examine combinatorial problems such as graph colouring and TWICE-SAT, and problems from statistical physics, such as the antiferromagnetic Potts model and the hard-core lattice gas model. In each case we provide either a proof of rapid mixing where none was known previously, or substantial simplification of existing proofs with consequent gains in the performance of the resulting algorithms."
                },
                {
                    "title": "A Chernoff bound for random walks on expander graphs",
                    "abstract": "We consider a finite random walk on a weighted graph G; we show that the sample average of visits to a set of vertices A converges to the stationary probability /spl pi/(A) with error probability exponentially small in the length of the random walk and the square of the size of the deviation from /spl pi/(A). The exponential bound is in terms of the expansion of G and improves previous results. We show that the method of taking the sample average from one trajectory is a more efficient estimate of /spl pi/(A) than the standard method of generating independent sample points from several trajectories. Using this more efficient sampling method, we improve the algorithms of Jerrum and Sinclair (1989) for approximating the number of perfect matchings in a dense graph and for approximating the partition function of an Ising system. We also give a fast estimate of the entropy of a random walk on an unweighted graph.<<ETX>>"
                },
                {
                    "title": "Fastest Mixing Markov Chain on a Graph",
                    "abstract": "We consider a symmetric random walk on a connected graph, where each edge is labeled with the probability of transition between the two adjacent vertices. The associated Markov chain has a uniform equilibrium distribution; the rate of convergence to this distribution, i.e., the mixing rate of the Markov chain, is determined by the second largest eigenvalue modulus (SLEM) of the transition probability matrix. In this paper we address the problem of assigning probabilities to the edges of the graph in such a way as to minimize the SLEM, i.e., the problem of finding the fastest mixing Markov chain on the graph. \nWe show that this problem can be formulated as a convex optimization problem, which can in turn be expressed as a semidefinite program (SDP). This allows us to easily compute the (globally) fastest mixing Markov chain for any graph with a modest number of edges (say, $1000$) using standard numerical methods for SDPs. Larger problems can be solved by exploiting various types of symmetry and structure in the problem, and far larger problems (say, 100,000 edges) can be solved using a subgradient method we describe. We compare the fastest mixing Markov chain to those obtained using two commonly used heuristics: the maximum-degree method, and the Metropolis--Hastings algorithm. For many of the examples considered, the fastest mixing Markov chain is substantially faster than those obtained using these heuristic methods. \nWe derive the Lagrange dual of the fastest mixing Markov chain problem, which gives a sophisticated method for obtaining (arbitrarily good) bounds on the optimal mixing rate, as well as the optimality conditions. Finally, we describe various extensions of the method, including a solution of the problem of finding the fastest mixing reversible Markov chain, on a fixed graph, with a given equilibrium distribution."
                }
            ],
            "categories": [
                "math.PR",
                "math.ST",
                "stat.ML",
                "stat.TH"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we rigorously understand and improve the Elo rating system using a probabilistic approach, particularly under the Bradley\u2013Terry\u2013Luce model?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it provides a theoretical foundation for the widely used Elo rating system, which currently lacks rigorous understanding. By addressing this question, we can advance knowledge in sports analytics and game theory, leading to improved rating systems that can be applied in various competitive environments. This could enhance player evaluations, tournament designs, and overall fairness in competitive settings, influencing future research in ranking methodologies and probabilistic models.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the complexities of the Elo rating system, particularly its zero-sum nature and the diminishing returns of rating updates. Naive approaches may fail because they do not account for the non-linear behavior of the sigmoid function in the rating updates, which can obscure the true skill estimation. Additionally, proving that the maximum rating cannot increase significantly is complicated by the fact that the maximum rating is not a supermartingale, and the interactions between players can lead to unexpected rating changes that defy simple probabilistic models.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on empirical observations or heuristic improvements to the Elo system without delving into its theoretical underpinnings. Limitations in existing solutions include a lack of rigorous mathematical frameworks and an insufficient understanding of the underlying probabilistic dynamics. Barriers such as the complexity of the rating update mechanism and the interactions between players have hindered progress. Our approach differs by applying the Bradley\u2013Terry\u2013Luce model to provide a more robust theoretical framework, allowing for a deeper understanding of the Elo system's behavior.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves applying the Bradley\u2013Terry\u2013Luce model to analyze the Elo rating system, focusing on the probabilistic dynamics of player interactions. We will use a dataset of player match outcomes to empirically validate our theoretical findings. The key metrics will include the accuracy of rating estimations and the stability of ratings over time. We expect to demonstrate that our approach leads to a more accurate representation of players' true skills and provides insights into the limitations of the current Elo system, ultimately contributing to the development of improved ranking methodologies."
            }
        },
        "author_data": {
            "6a4aae3a-8cb4-4ef5-88e3-840a525d5325": {
                "pk": "6a4aae3a-8cb4-4ef5-88e3-840a525d5325",
                "name": "Sam Olesker-Taylor",
                "collaborators": [
                    "Jonathan Hermon",
                    "Perla Sousi",
                    "Evita Nestoridi",
                    "Luca Zanetti",
                    "Dominik Schmid"
                ],
                "domain": [
                    "Markov Chains",
                    "Random Walks",
                    "Graph Theory",
                    "Probability Theory"
                ],
                "publications": [
                    {
                        "title": "Cutoff for Rewiring Dynamics on Perfect Matchings",
                        "abstract": "We establish cutoff for a natural random walk (RW) on the set of perfect matchings (PMs). An $n$-PM is a pairing of $2n$ objects. The $k$-PM RW selects $k$ pairs uniformly at random, disassociates the corresponding $2k$ objects, then chooses a new pairing on these $2k$ objects uniformly at random. The equilibrium distribution is uniform over the set of all $n$-PM.   We establish cutoff for the $k$-PM RW whenever $2 \\le k \\ll n$. If $k \\gg 1$, then the mixing time is $\\tfrac nk \\log n$ to leading order. The case $k = 2$ was established by Diaconis and Holmes (2002) by relating the $2$-PM RW to the random transpositions card shuffle and also by Ceccherini-Silberstein, Scarabotti and Tolli (2007, 2008) using representation theory. We are the first to handle $k > 2$. Our argument builds on previous work of Berestycki, Schramm, \\c{S}eng\\\"ul and Zeitouni (2005, 2011, 2019) regarding conjugacy-invariant RWs on the permutation group."
                    },
                    {
                        "title": "Multicoloured Hardcore Model: Fast Mixing and Queueing",
                        "abstract": "We extend the hardcore model to a multicoloured version: a subset of vertices of a graph are coloured such that no vertex is adjacent to one of the same colour; uncoloured vertices do not constrain neighbours. This mathematically models multi-channel resource sharing, such as fibreoptic routing.   We analyse certain simple Glauber-type dynamics on such configurations, and find conditions which ensure fast mixing. These dynamics model a queueing system: customers queue for service at vertices, who only serve customers whilst they are coloured in the underlying configuration; uncoloured vertices sit idle. The mixing estimates are applied to control queue lengths in equilibrium."
                    },
                    {
                        "title": "Metastability in Loss Networks with Dynamic Alternative Routing",
                        "abstract": "Consider $N$ stations interconnected with links, each of capacity $K$, forming a complete graph. Calls arrive to each link at rate $\\lambda$ and depart at rate $1$. If a call arrives to a link $x y$, connecting stations $x$ and $y$, which is at capacity, then a third station $z$ is chosen uniformly at random and the call is attempted to be routed via $z$: if both links $x z$ and $z y$ have spare capacity, then the call is held simultaneously on these two; otherwise the call is lost.   We analyse an approximation of this model. We show rigorously that there are three phases according to the traffic intensity $\\alpha := \\lambda/K$: for $\\alpha \\in (0,\\alpha_c) \\cup (1,\\infty)$, the system has mixing time logarithmic in the number of links $n := \\binom N2$; for $\\alpha \\in (\\alpha_c,1)$ the system has mixing time exponential in $n$, the number of links. Here $\\alpha_c := \\tfrac13 (5 \\sqrt{10} - 13) \\approx 0.937$ is an explicit critical threshold with a simple interpretation. We also consider allowing multiple rerouting attempts. This has little effect on the overall behaviour; it does not remove the metastability phase.   Finally, we add trunk reservation: in this, some number $\\sigma$ of circuits are reserved; a rerouting attempt is only accepted if at least $\\sigma+1$ circuits are available. We show that if $\\sigma$ is chosen sufficiently large, depending only on $\\alpha$, not $K$ or $n$, then the metastability phase is removed."
                    },
                    {
                        "title": "Cutoff for Random Walk on Dynamical Erd\u0151s--R\u00e9nyi Graph",
                        "abstract": "We consider dynamical percolation on the complete graph $K_n$, where each edge refreshes its state at rate $\\mu \\ll 1/n$, and is then declared open with probability $p = \\lambda/n$ where $\\lambda > 1$. We study a random walk on this dynamical environment which jumps at rate $1/n$ along every open edge. We show that the mixing time of the full system exhibits cutoff at $\\log n/\\mu$. We do this by showing that the random walk component mixes faster than the environment process; along the way, we control the time it takes for the walk to become isolated."
                    },
                    {
                        "title": "Supplementary Material for Random Cayley Graphs Project",
                        "abstract": "This document contains supplementary material for the main articles in our Random Cayley Graphs project.   We prove refined results about simple random walks on the integers and on the cycle. We are primarily interested in the entropy of these random walks at certain times and how this entropy changes when the time changes slightly. Additionally, we prove some large deviation and exit time estimates.   We prove some results on the size of discrete lattice balls and how this size changes when the radius changes slightly. We do this in a general $L_q$ norm, with $q \\in [1,\\infty]$.   We also prove some other technical results deferred from the main papers.   We hope that some of the results, particularly the simple random walk estimates, will be useful in their own right for other researchers."
                    },
                    {
                        "title": "Chen--Stein Method for the Uncovered Set of Random Walk on $\\mathbb Z_n^d$ for $d \\ge 3$",
                        "abstract": "Let $X$ be a simple random walk on $\\mathbb{Z}_n^d$ with $d\\geq 3$ and let $t_{\\rm{cov}}$ be the expected cover time. We consider the set of points $\\mathcal{U}_\\alpha$ of $\\mathbb{Z}_n^d$ that have not been visited by the walk by time $\\alpha t_{\\rm{cov}}$ for $\\alpha\\in (0,1)$. It was shown in [MS17] that there exists $\\alpha_1(d)\\in (0,1)$ such that for all $\\alpha>\\alpha_1(d)$ the total variation distance between the law of the set $\\mathcal{U}_\\alpha$ and an i.i.d. sequence of Bernoulli random variables indexed by $\\mathbb{Z}_n^d$ with success probability $n^{-\\alpha d}$ tends to $0$ as $n \\to \\infty$. In [MS17] the constant $\\alpha_1(d)$ converges to $1$ as $d\\to\\infty$. In this short note using the Chen--Stein method and a concentration result for Markov chains of Lezaud we greatly simplify the proof of [MS17] and find a constant $\\alpha_1(d)$ which converges to $3/4$ as $d\\to\\infty$."
                    },
                    {
                        "title": "Limit Profiles for Reversible Markov Chains",
                        "abstract": "In a recent breakthrough, Teyssier [Tey20] introduced a new method for approximating the distance from equilibrium of a random walk on a group. He used it to study the limit profile for the random transpositions card shuffle. His techniques were restricted to conjugacy-invariant random walks on groups; we derive similar approximation lemmas for random walks on homogeneous spaces and for general reversible Markov chains. We illustrate applications of these lemmas to some famous problems: the $k$-cycle shuffle, improving results of Hough [Hou16] and Berestycki, Schramm and Zeitouni [BSZ11]; the Ehrenfest urn diffusion with many urns, improving results of Ceccherini-Silberstein, Scarabotti and Tolli [CST07]; a Gibbs sampler, which is a fundamental tool in statistical physics, with Binomial prior and hypergeometric posterior, improving results of Diaconis, Khare and Saloff-Coste [DKS08]."
                    },
                    {
                        "title": "Geometric Bounds on the Fastest Mixing Markov Chain",
                        "abstract": "In the Fastest Mixing Markov Chain problem, we are given a graph $G = (V, E)$ and desire the discrete-time Markov chain with smallest mixing time $\\tau$ subject to having equilibrium distribution uniform on $V$ and non-zero transition probabilities only across edges of the graph.   It is well-known that the mixing time $\\tau_\\textsf{RW}$ of the lazy random walk on $G$ is characterised by the edge conductance $\\Phi$ of $G$ via Cheeger's inequality: $\\Phi^{-1} \\lesssim \\tau_\\textsf{RW} \\lesssim \\Phi^{-2} \\log |V|$. Analogously, we characterise the fastest mixing time $\\tau^\\star$ via a Cheeger-type inequality but for a different geometric quantity, namely the vertex conductance $\\Psi$ of $G$: $\\Psi^{-1} \\lesssim \\tau^\\star \\lesssim \\Psi^{-2} (\\log |V|)^2$.   This characterisation forbids fast mixing for graphs with small vertex conductance. To bypass this fundamental barrier, we consider Markov chains on $G$ with equilibrium distribution which need not be uniform, but rather only $\\varepsilon$-close to uniform in total variation. We show that it is always possible to construct such a chain with mixing time $\\tau \\lesssim \\varepsilon^{-1} (\\operatorname{diam} G)^2 \\log |V|$.   Finally, we discuss analogous questions for continuous-time and time-inhomogeneous chains."
                    },
                    {
                        "title": "Limit Profile for Projections of Random Walks on Groups",
                        "abstract": "Establishing cutoff -- an abrupt transition from \"not mixed\" to \"well mixed\" -- is a classical topic in the theory of mixing times for Markov chains. Interest has grown recently in determining not only the existence of cutoff and the order of its mixing time and window, but the exact shape, or profile, of the convergence inside the window. Classical techniques, such as coupling or $\\ell_2$-bounds, are typically too crude to establish this and there has been a push to develop general techniques.   We build upon this work, extending from conjugacy-invariant random walks on groups to certain projections. We exemplify our method by analysing the $k$-particle interchange process on the complete $n$-graph with $k \\asymp n$. This is a projection of the random-transpositions card shuffle, which corresponds to $k = n$, analysed by Teyssier."
                    },
                    {
                        "title": "Limit Profile for the Bernoulli--Laplace Urn",
                        "abstract": "We analyse the convergence to equilibrium of the Bernoulli--Laplace urn model: initially, one urn contains $k$ red balls and a second $n-k$ blue balls; in each step, a pair of balls is chosen uniform and their locations are switched. Cutoff is known to occur at $\\tfrac12 n \\log \\min\\{k, \\sqrt n\\}$ with window order $n$ whenever $1 \\ll k \\le \\tfrac12 n$. We refine this by determining the limit profile: a function $\\Phi$ such that \\[   d_\\mathsf{TV}\\bigl( \\tfrac12 n \\log \\min\\{k, \\sqrt n\\} + \\theta n \\bigr) \\to   \\Phi(\\theta) \\quad\\text{as}\\quad   n \\to \\infty \\quad\\text{for all}\\quad   \\theta \\in \\mathbb R. \\] Our main technical contribution, of independent interest, approximates a rescaled chain by a diffusion on $\\mathbb R$ when $k \\gg \\sqrt n$, and uses its explicit law as a Gaussian process."
                    },
                    {
                        "title": "Geometry of Random Cayley Graphs of Abelian Groups",
                        "abstract": "Consider the random Cayley graph of a finite Abelian group $G$ with respect to $k$ generators chosen uniformly at random, with $1 \\ll \\log k \\ll \\log |G|$. Draw a vertex $U \\sim \\operatorname{Unif}(G)$.   We show that the graph distance $\\operatorname{dist}(\\mathsf{id},U)$ from the identity to $U$ concentrates at a particular value $M$, which is the minimal radius of a ball in $\\mathbb Z^k$ of cardinality at least $|G|$, under mild conditions. In other words, the distance from the identity for all but $o(|G|)$ of the elements of $G$ lies in the interval $[M - o(M), M + o(M)]$. In the regime $k \\gtrsim \\log |G|$, we show that the diameter of the graph is also asymptotically $M$. In the spirit of a conjecture of Aldous and Diaconis (1985), this $M$ depends only on $k$ and $|G|$, not on the algebraic structure of $G$.   Write $d(G)$ for the minimal size of a generating subset of $G$. We prove that the order of the spectral gap is $|G|^{-2/k}$ when $k - d(G) \\asymp k$ and $|G|$ lies in a density-$1$ subset of $\\mathbb N$ or when $k - 2 d(G) \\asymp k$. This extends, for Abelian groups, a celebrated result of Alon and Roichman (1994).   The aforementioned results all hold with high probability over the random Cayley graph."
                    },
                    {
                        "title": "Further Results and Discussions on Random Cayley Graphs",
                        "abstract": "Consider the random Cayley graph of a finite group $G$ with respect to $k$ generators chosen uniformly at random, with $1 \\ll k \\lesssim \\log |G|$. The results of this article supplement those in the three main papers on random Cayley graphs. The majority of the results are inspired by a `universality' conjecture of Aldous and Diaconis (1985).   To start, we study the limit profile of cutoff for the simple random walk on this random graph, as well as a detailed investigation into mixing properties when $G = \\mathbb Z_p^d$ with $p$ prime. We then exposit a proof of Diaconis and Saloff-Coste (1994) establishing lack of cutoff when $k \\asymp 1$. We move onto discussing material from our companion paper on matrix groups. We then study distance of a typical element of $G$ from the identity in an $L_q$-type graph distance in the Abelian set-up. Finally, we give necessary and sufficient conditions for $k$ independent uniform elements of $G$ to generate $G$, ie for the random Cayley graph to be connected, based on work of Pomerance (2001).   The aforementioned results all hold with high probability over the random Cayley graph."
                    },
                    {
                        "title": "Cutoff for Almost All Random Walks on Abelian Groups",
                        "abstract": "Consider the random Cayley graph of a finite group $G$ with respect to $k$ generators chosen uniformly at random, with $1 \\ll \\log k \\ll \\log |G|$; denote it $G_k$. A conjecture of Aldous and Diaconis (1985) asserts, for $k \\gg \\log |G|$, that the random walk on this graph exhibits cutoff. Further, the cutoff time should be a function only of $k$ and $|G|$, to sub-leading order. This was verified for all Abelian groups in the '90s. We extend the conjecture to $1 \\ll k \\lesssim \\log |G|$. We establish cutoff for all Abelian groups under the condition $k - d(G) \\gg 1$, where $d(G)$ is the minimal size of a generating subset of $G$, which is almost optimal. The cutoff time is described (abstractly) in terms of the entropy of random walk on $\\mathbb Z^k$. This abstract definition allows us to deduce that the cutoff time can be written as a function only of $k$ and $|G|$ when $d(G) \\ll \\log |G|$ and $k - d(G) \\asymp k \\gg 1$; this is not the case when $d(G) \\asymp \\log |G| \\asymp k$. For certain regimes of $k$, we find the limit profile of the convergence to equilibrium. Wilson (1997) conjectured that $\\mathbb Z_2^d$ gives rise to the slowest mixing time for $G_k$ amongst all groups of size at most $2^d$. We give a partial answer, verifying the conjecture for nilpotent groups. This is obtained via a comparison result of independent interest between the mixing times of nilpotent $G$ and a corresponding Abelian group $\\overline G$, namely the direct sum of the Abelian quotients in the lower central series of $G$. We use this to refine a celebrated result of Alon and Roichman (1994): we show for nilpotent $G$ that $G_k$ is an expander provided $k - d(\\overline G) \\gtrsim \\log |G|$. As another consequence, we establish cutoff for nilpotent groups with relatively small commutators, including high-dimensional special groups, such as Heisenberg groups."
                    },
                    {
                        "title": "Cutoff for Random Walks on Upper Triangular Matrices",
                        "abstract": "Consider the random Cayley graph of a finite group $G$ with respect to $k$ generators chosen uniformly at random, with $1 \\ll \\log k \\ll \\log |G|$ (ie $1 \\ll k = |G|^{o(1)}$). A conjecture of Aldous and Diaconis (1985) asserts, for $k\\gg\\log|G|$, that the random walk on this graph exhibits cutoff.   When $\\log k \\lesssim \\log\\log|G|$ (ie $k = (\\log |G|)^{\\mathcal O(1)}$), the only example of a non-Abelian group for which cutoff has been established is the dihedral group. We establish cutoff (as $p\\to infty$) for the group of $d \\times d$ unit upper triangular matrices with integer entries modulo $p$ (prime), which we denote $U_{p,d}$, for fixed $d$ or $d$ diverging sufficiently slowly. We allow $1 \\ll k \\lesssim \\log |U_{p,d}|$ as well as $k\\gg\\log|U_{p,d}|$. The cutoff time is $\\max\\{\\log_k |U_{p,d}|, \\: s_0 k\\}$, where $s_0$ is the time at which the entropy of the random walk on $\\mathbb Z$ reaches $(\\log |U_{p,d}^\\mathrm{ab}|)/k$, where $U_{p,d}^\\mathrm{ab} \\cong \\mathbb Z_p^{d-1}$ is the Abelianisation of $U_{p,d}$. When $1 \\ll k \\ll \\log |U_{p,d}^\\mathrm{ab}|$ and $d \\asymp 1$, we find the limit profile. We also prove highly related results for the $d$-dimensional Heisenberg group over $\\mathbb Z_p$.   The Aldous--Diaconis conjecture also asserts, for $k gg\\log |G|$, that the cutoff time should depend only on $k$ and $|G|$. This was verified for all Abelian groups. Our result shows that this is not the case for $U_{p,d}$: the cutoff time depends on $k$, $|U_{p,d}| = p^{d(d-1)/2}$ and $|U_{p,d}^\\mathrm{ab}|=p^{d-1}$.   We also show that all but $o(|U_{p,d}|)$ of the elements of $U_{p,d}$ lie at graph distance $M \\pm o(M)$ from the identity, where $M$ is the minimal radius of a ball in $\\mathbb Z^k$ of cardinality $|U_{p,d}^\\mathrm{ab}| = p^{d-1}$. Finally, we show that the diameter is also asymptotically $M$ when $k \\gtrsim \\log |U_{p,d}^\\textrm{ab}|$ and $d\\asymp1$."
                    }
                ]
            },
            "7039d23e-38cc-4253-9628-cfbdace36e73": {
                "pk": "7039d23e-38cc-4253-9628-cfbdace36e73",
                "name": "Luca Zanetti",
                "collaborators": [
                    "He Sun",
                    "Thomas Sauerwald",
                    "Huan Li",
                    "Dario Tamascelli",
                    "Mihai Cucuringu",
                    "Sam Olesker-Taylor",
                    "Richard Peng",
                    "Leran Cai"
                ],
                "domain": [
                    "Graph Theory",
                    "Random Walks",
                    "Spectral Clustering",
                    "Quantum Algorithms"
                ],
                "publications": [
                    {
                        "title": "Random Walks on Dynamic Graphs: Mixing Times, HittingTimes, and Return Probabilities",
                        "abstract": "We establish and generalise several bounds for various random walk quantities including the mixing time and the maximum hitting time. Unlike previous analyses, our derivations are based on rather intuitive notions of local expansion properties which allows us to capture the progress the random walk makes through $t$-step probabilities.   We apply our framework to dynamically changing graphs, where the set of vertices is fixed while the set of edges changes in each round. For random walks on dynamic connected graphs for which the stationary distribution does not change over time, we show that their behaviour is in a certain sense similar to static graphs. For example, we show that the mixing and hitting times of any sequence of $d$-regular connected graphs is $O(n^2)$, generalising a well-known result for static graphs. We also provide refined bounds depending on the isoperimetric dimension of the graph, matching again known results for static graphs. Finally, we investigate properties of random walks on dynamic graphs that are not always connected: we relate their convergence to stationarity to the spectral properties of an average of transition matrices and provide some examples that demonstrate strong discrepancies between static and dynamic graphs."
                    },
                    {
                        "title": "A quantum-walk-inspired adiabatic algorithm for graph isomorphism",
                        "abstract": "We present a 2-local quantum algorithm for graph isomorphism GI based on an adiabatic protocol. By exploiting continuous-time quantum-walks, we are able to avoid a mere diffusion over all possible configurations and to significantly reduce the dimensionality of the visited space. Within this restricted space, the graph isomorphism problem can be translated into the search of a satisfying assignment to a 2-SAT formula without resorting to perturbation gadgets or projective techniques. We present an analysis of the execution time of the algorithm on small instances of the graph isomorphism problem and discuss the issue of an implementation of the proposed adiabatic scheme on current quantum computing hardware."
                    },
                    {
                        "title": "Hermitian matrices for clustering directed graphs: insights and applications",
                        "abstract": "Graph clustering is a basic technique in machine learning, and has widespread applications in different domains. While spectral techniques have been successfully applied for clustering undirected graphs, the performance of spectral clustering algorithms for directed graphs (digraphs) is not in general satisfactory: these algorithms usually require symmetrising the matrix representing a digraph, and typical objective functions for undirected graph clustering do not capture cluster-structures in which the information given by the direction of the edges is crucial. To overcome these downsides, we propose a spectral clustering algorithm based on a complex-valued matrix representation of digraphs. We analyse its theoretical performance on a Stochastic Block Model for digraphs in which the cluster-structure is given not only by variations in edge densities, but also by the direction of the edges. The significance of our work is highlighted on a data set pertaining to internal migration in the United States: while previous spectral clustering algorithms for digraphs can only reveal that people are more likely to move between counties that are geographically close, our approach is able to cluster together counties with a similar socio-economical profile even when they are geographically distant, and illustrates how people tend to move from rural to more urbanised areas."
                    },
                    {
                        "title": "Distributed Graph Clustering and Sparsification",
                        "abstract": "Graph clustering is a fundamental computational problem with a number of applications in algorithm design, machine learning, data mining, and analysis of social networks. Over the past decades, researchers have proposed a number of algorithmic design methods for graph clustering. Most of these methods, however, are based on complicated spectral techniques or convex optimisation, and cannot be directly applied for clustering many networks that occur in practice, whose information is often collected on different sites. Designing a simple and distributed clustering algorithm is of great interest, and has wide applications for processing big datasets.   In this paper we present a simple and distributed algorithm for graph clustering: for a wide class of graphs that are characterised by a strong cluster-structure, our algorithm finishes in a poly-logarithmic number of rounds, and recovers a partition of the graph close to optimal. One of the main components behind our algorithm is a sampling scheme that, given a dense graph as input, produces a sparse subgraph that provably preserves the cluster-structure of the input. Compared with previous sparsification algorithms that require Laplacian solvers or involve combinatorial constructions, this component is easy to implement in a distributed way and runs fast in practice."
                    },
                    {
                        "title": "Geometric Bounds on the Fastest Mixing Markov Chain",
                        "abstract": "In the Fastest Mixing Markov Chain problem, we are given a graph $G = (V, E)$ and desire the discrete-time Markov chain with smallest mixing time $\\tau$ subject to having equilibrium distribution uniform on $V$ and non-zero transition probabilities only across edges of the graph.   It is well-known that the mixing time $\\tau_\\textsf{RW}$ of the lazy random walk on $G$ is characterised by the edge conductance $\\Phi$ of $G$ via Cheeger's inequality: $\\Phi^{-1} \\lesssim \\tau_\\textsf{RW} \\lesssim \\Phi^{-2} \\log |V|$. Analogously, we characterise the fastest mixing time $\\tau^\\star$ via a Cheeger-type inequality but for a different geometric quantity, namely the vertex conductance $\\Psi$ of $G$: $\\Psi^{-1} \\lesssim \\tau^\\star \\lesssim \\Psi^{-2} (\\log |V|)^2$.   This characterisation forbids fast mixing for graphs with small vertex conductance. To bypass this fundamental barrier, we consider Markov chains on $G$ with equilibrium distribution which need not be uniform, but rather only $\\varepsilon$-close to uniform in total variation. We show that it is always possible to construct such a chain with mixing time $\\tau \\lesssim \\varepsilon^{-1} (\\operatorname{diam} G)^2 \\log |V|$.   Finally, we discuss analogous questions for continuous-time and time-inhomogeneous chains."
                    },
                    {
                        "title": "Distributed Graph Clustering by Load Balancing",
                        "abstract": "Graph clustering is a fundamental computational problem with a number of applications in algorithm design, machine learning, data mining, and analysis of social networks. Over the past decades, researchers have proposed a number of algorithmic design methods for graph clustering. However, most of these methods are based on complicated spectral techniques or convex optimisation, and cannot be applied directly for clustering many networks that occur in practice, whose information is often collected on different sites. Designing a simple and distributed clustering algorithm is of great interest, and has wide applications for processing big datasets. In this paper we present a simple and distributed algorithm for graph clustering: for a wide class of graphs that are characterised by a strong cluster-structure, our algorithm finishes in a poly-logarithmic number of rounds, and recovers a partition of the graph close to an optimal partition. The main component of our algorithm is an application of the random matching model of load balancing, which is a fundamental protocol in distributed computing and has been extensively studied in the past 20 years. Hence, our result highlights an intrinsic and interesting connection between graph clustering and load balancing. At a technical level, we present a purely algebraic result characterising the early behaviours of load balancing processes for graphs exhibiting a cluster-structure. We believe that this result can be further applied to analyse other gossip processes, such as rumour spreading and averaging processes."
                    },
                    {
                        "title": "Partitioning Well-Clustered Graphs: Spectral Clustering Works!",
                        "abstract": "In this paper we study variants of the widely used spectral clustering that partitions a graph into k clusters by (1) embedding the vertices of a graph into a low-dimensional space using the bottom eigenvectors of the Laplacian matrix, and (2) grouping the embedded points into k clusters via k-means algorithms. We show that, for a wide class of graphs, spectral clustering gives a good approximation of the optimal clustering. While this approach was proposed in the early 1990s and has comprehensive applications, prior to our work similar results were known only for graphs generated from stochastic models.   We also give a nearly-linear time algorithm for partitioning well-clustered graphs based on computing a matrix exponential and approximate nearest neighbor data structures."
                    },
                    {
                        "title": "Hermitian Laplacians and a Cheeger inequality for the Max-2-Lin problem",
                        "abstract": "We study spectral approaches for the MAX-2-LIN(k) problem, in which we are given a system of $m$ linear equations of the form $x_i - x_j \\equiv c_{ij}\\mod k$, and required to find an assignment to the $n$ variables $\\{x_i\\}$ that maximises the total number of satisfied equations.   We consider Hermitian Laplacians related to this problem, and prove a Cheeger inequality that relates the smallest eigenvalue of a Hermitian Laplacian to the maximum number of satisfied equations of a MAX-2-LIN(k) instance $\\mathcal{I}$. We develop an $\\widetilde{O}(kn^2)$ time algorithm that, for any $(1-\\varepsilon)$-satisfiable instance, produces an assignment satisfying a $\\left(1 - O(k)\\sqrt{\\varepsilon}\\right)$-fraction of equations. We also present a subquadratic-time algorithm that, when the graph associated with $\\mathcal{I}$ is an expander, produces an assignment satisfying a $\\left(1- O(k^2)\\varepsilon \\right)$-fraction of the equations. Our Cheeger inequality and first algorithm can be seen as generalisations of the Cheeger inequality and algorithm for MAX-CUT developed by Trevisan."
                    },
                    {
                        "title": "Random walks on randomly evolving graphs",
                        "abstract": "A random walk is a basic stochastic process on graphs and a key primitive in the design of distributed algorithms. One of the most important features of random walks is that, under mild conditions, they converge to a stationary distribution in time that is at most polynomial in the size of the graph. This fundamental property, however, only holds if the graph does not change over time, while on the other hand many distributed networks are inherently dynamic, and their topology is subjected to potentially drastic changes.   In this work we study the mixing (i.e., converging) properties of random walks on graphs subjected to random changes over time. Specifically, we consider the edge-Markovian random graph model: for each edge slot, there is a two-state Markov chain with transition probabilities $p$ (add a non-existing edge) and $q$ (remove an existing edge). We derive several positive and negative results that depend on both the density of the graph and the speed by which the graph changes. We show that if $p$ is very small (i.e., below the connectivity threshold of Erd\\H{o}s-R\\'{e}nyi random graphs), random walks do not mix (fast). When $p$ is larger, instead, we observe the following behavior: if the graph changes slowly over time (i.e., $q$ is small), random walks enjoy strong mixing properties that are comparable to the ones possessed by random walks on static graphs; however, if the graph changes too fast (i.e., $q$ is large), only coarse mixing properties are preserved."
                    }
                ]
            }
        }
    },
    "2402.02017": {
        "paper_data": {
            "title": "Adaptive $Q$-Aid for Conditional Supervised Learning in Offline Reinforcement Learning",
            "url": "http://arxiv.org/abs/2402.02017v2",
            "arxiv_id": "2402.02017",
            "authors": [
                "Jeonghye Kim",
                "Suyoung Lee",
                "Woojun Kim",
                "Youngchul Sung"
            ],
            "abstract": "Offline reinforcement learning (RL) has progressed with return-conditioned supervised learning (RCSL), but its lack of stitching ability remains a limitation. We introduce $Q$-Aided Conditional Supervised Learning (QCS), which effectively combines the stability of RCSL with the stitching capability of $Q$-functions. By analyzing $Q$-function over-generalization, which impairs stable stitching, QCS adaptively integrates $Q$-aid into RCSL's loss function based on trajectory return. Empirical results show that QCS significantly outperforms RCSL and value-based methods, consistently achieving or exceeding the maximum trajectory returns across diverse offline RL benchmarks.",
            "introduction": "   1 Introduction  Figure 1: Conceptual idea of QCS: Follow RCSL when learning from optimal trajectories where it predicts actions confidently but the Q\ud835\udc44Qitalic_Q-function may stitch incorrectly. Conversely, refer to the Q\ud835\udc44Qitalic_Q-function when learning from sub-optimal trajectories where RCSL is less certain but the Q\ud835\udc44Qitalic_Q-function is likely accurate.   Offline reinforcement learning (RL) is a vital framework for acquiring decision-making skills from fixed datasets, particularly when online interactions are impractical. This is especially relevant in fields such as robotics, autonomous driving, and healthcare, where the costs and risks of real-time experimentation are significant.   A promising approach in offline RL is return-conditioned supervised learning (RCSL) [10, 12, 20]. By framing offline RL as sequence modeling tasks, RCSL allows an agent to leverage past experiences and condition on the target outcome, facilitating the generation of future actions that are likely to achieve desired outcomes. This method builds on recent advancements in supervised learning (SL) [41, 9, 11, 30], and thus benefits from the inherent stability and scalability of SL. However, RCSL is significantly limited by its lack of \u2018stitching ability\u2019, the ability to combine suboptimal trajectory segments to form better overall trajectories [13, 27, 53, 16, 7, 58]. As a result, its effectiveness is restricted to the best trajectories within the dataset.        Figure 2: Mean normalized return in MuJoCo medium, medium-replay, medium-expert, and AntMaze large. The scores of RCSL, the value-based methods, and the combined methods represent the maximum mean performances within their respective groups. The full scores are in Section 6.2.   Conversely, the Q\ud835\udc44Qitalic_Q-function possesses the ability to stitch together multiple sub-optimal trajectories, dissecting and reassembling them into an optimal trajectory through dynamic programming. Therefore, to address the weakness of RCSL, prior works have attempted to enhance stitching ability through the Q\ud835\udc44Qitalic_Q-function [53, 16]. However, these prior works employ the Q\ud835\udc44Qitalic_Q-function as a conditioning factor for RCSL and do not fully leverage the Q\ud835\udc44Qitalic_Q-function\u2019s stitching ability, resulting in either negligible performance improvements or even reduced performance. The primary challenge lies in the fact that utilizing the Q\ud835\udc44Qitalic_Q-function through conditioning, without managing the conditions for stable Q\ud835\udc44Qitalic_Q-guided stitching, can result in a sub-optimal algorithm.   In this work, we aim to fully synergize the stable and scalable learning framework of RCSL with the stitching ability of the Q\ud835\udc44Qitalic_Q-function. To effectively utilize the Q\ud835\udc44Qitalic_Q-function, it is crucial to identify when it can benefit RCSL and to integrate its assistance, termed Q\ud835\udc44Qitalic_Q-aid. Our contributions to achieving the effective utilization of Q\ud835\udc44Qitalic_Q-aid in RCSL are as follows: (1) We discovered that in-sample Q\ud835\udc44Qitalic_Q-learning on an expert dataset, which predominantly consists of optimal actions with similar Q\ud835\udc44Qitalic_Q-values within a constrained action range, causes the Q\ud835\udc44Qitalic_Q-function to receive improper learning signals and become over-generalized. (2) To prevent errors from this Q\ud835\udc44Qitalic_Q-generalization and to incorporate stitching ability within RCSL\u2019s stable framework, we propose Q\ud835\udc44Qitalic_Q-Aided Conditional Supervised Learning (QCS), which adaptively integrates Q\ud835\udc44Qitalic_Q-aid into the RCSL\u2019s loss function based on trajectory returns.   Despite its simplicity, the effectiveness of QCS is empirically demonstrated across offline RL benchmarks, showing significant advancements over existing state-of-the-art (SOTA) methods, including both RCSL and value-based methods. Especially, QCS surpasses the maximal dataset trajectory return across diverse",
            "references": [
                {
                    "title": "Critic-Guided Decision Transformer for Offline Reinforcement Learning",
                    "abstract": "Recent advancements in offline reinforcement learning (RL) have underscored the capabilities of Return-Conditioned Supervised Learning (RCSL), a paradigm that learns the action distribution based on target returns for each state in a supervised manner. However, prevailing RCSL methods largely focus on deterministic trajectory modeling, disregarding stochastic state transitions and the diversity of future trajectory distributions. A fundamental challenge arises from the inconsistency between the sampled returns within individual trajectories and the expected returns across multiple trajectories. Fortunately, value-based methods offer a solution by leveraging a value function to approximate the expected returns, thereby addressing the inconsistency effectively. Building upon these insights, we propose a novel approach, termed the Critic-Guided Decision Transformer (CGDT), which combines the predictability of long-term returns from value-based methods with the trajectory modeling capability of the Decision Transformer. By incorporating a learned value function, known as the critic, CGDT ensures a direct alignment between the specified target returns and the expected returns of actions. This integration bridges the gap between the deterministic nature of RCSL and the probabilistic characteristics of value-based methods. Empirical evaluations on stochastic environments and D4RL benchmark datasets demonstrate the superiority of CGDT over traditional RCSL methods. These results highlight the potential of CGDT to advance the state of the art in offline RL and extend the applicability of RCSL to a wide range of RL tasks."
                },
                {
                    "title": "Free from Bellman Completeness: Trajectory Stitching via Model-based Return-conditioned Supervised Learning",
                    "abstract": "Off-policy dynamic programming (DP) techniques such as $Q$-learning have proven to be important in sequential decision-making problems. In the presence of function approximation, however, these techniques often diverge due to the absence of Bellman completeness in the function classes considered, a crucial condition for the success of DP-based methods. In this paper, we show how off-policy learning techniques based on return-conditioned supervised learning (RCSL) are able to circumvent these challenges of Bellman completeness, converging under significantly more relaxed assumptions inherited from supervised learning. We prove there exists a natural environment in which if one uses two-layer multilayer perceptron as the function approximator, the layer width needs to grow linearly with the state space size to satisfy Bellman completeness while a constant layer width is enough for RCSL. These findings take a step towards explaining the superior empirical performance of RCSL methods compared to DP-based methods in environments with near-optimal datasets. Furthermore, in order to learn from sub-optimal datasets, we propose a simple framework called MBRCSL, granting RCSL methods the ability of dynamic programming to stitch together segments from distinct trajectories. MBRCSL leverages learned dynamics models and forward sampling to accomplish trajectory stitching while avoiding the need for Bellman completeness that plagues all dynamic programming algorithms. We propose both theoretical analysis and experimental evaluation to back these claims, outperforming state-of-the-art model-free and model-based offline RL algorithms across several simulated robotics problems."
                },
                {
                    "title": "Train Once, Get a Family: State-Adaptive Balances for Offline-to-Online Reinforcement Learning",
                    "abstract": "Offline-to-online reinforcement learning (RL) is a training paradigm that combines pre-training on a pre-collected dataset with fine-tuning in an online environment. However, the incorporation of online fine-tuning can intensify the well-known distributional shift problem. Existing solutions tackle this problem by imposing a policy constraint on the policy improvement objective in both offline and online learning. They typically advocate a single balance between policy improvement and constraints across diverse data collections. This one-size-fits-all manner may not optimally leverage each collected sample due to the significant variation in data quality across different states. To this end, we introduce Family Offline-to-Online RL (FamO2O), a simple yet effective framework that empowers existing algorithms to determine state-adaptive improvement-constraint balances. FamO2O utilizes a universal model to train a family of policies with different improvement/constraint intensities, and a balance model to select a suitable policy for each state. Theoretically, we prove that state-adaptive balances are necessary for achieving a higher policy performance upper bound. Empirically, extensive experiments show that FamO2O offers a statistically significant improvement over various existing methods, achieving state-of-the-art performance on the D4RL benchmark. Codes are available at https://github.com/LeapLabTHU/FamO2O."
                },
                {
                    "title": "Understanding, Predicting and Better Resolving Q-Value Divergence in Offline-RL",
                    "abstract": "The divergence of the Q-value estimation has been a prominent issue in offline RL, where the agent has no access to real dynamics. Traditional beliefs attribute this instability to querying out-of-distribution actions when bootstrapping value targets. Though this issue can be alleviated with policy constraints or conservative Q estimation, a theoretical understanding of the underlying mechanism causing the divergence has been absent. In this work, we aim to thoroughly comprehend this mechanism and attain an improved solution. We first identify a fundamental pattern, self-excitation, as the primary cause of Q-value estimation divergence in offline RL. Then, we propose a novel Self-Excite Eigenvalue Measure (SEEM) metric based on Neural Tangent Kernel (NTK) to measure the evolving property of Q-network at training, which provides an intriguing explanation of the emergence of divergence. For the first time, our theory can reliably decide whether the training will diverge at an early stage, and even predict the order of the growth for the estimated Q-value, the model's norm, and the crashing step when an SGD optimizer is used. The experiments demonstrate perfect alignment with this theoretic analysis. Building on our insights, we propose to resolve divergence from a novel perspective, namely improving the model's architecture for better extrapolating behavior. Through extensive empirical studies, we identify LayerNorm as a good solution to effectively avoid divergence without introducing detrimental bias, leading to superior performance. Experimental results prove that it can still work in some most challenging settings, i.e. using only 1 transitions of the dataset, where all previous methods fail. Moreover, it can be easily plugged into modern offline RL methods and achieve SOTA results on many challenging tasks. We also give unique insights into its effectiveness."
                },
                {
                    "title": "Decision ConvFormer: Local Filtering in MetaFormer is Sufficient for Decision Making",
                    "abstract": "The recent success of Transformer in natural language processing has sparked its use in various domains. In offline reinforcement learning (RL), Decision Transformer (DT) is emerging as a promising model based on Transformer. However, we discovered that the attention module of DT is not appropriate to capture the inherent local dependence pattern in trajectories of RL modeled as a Markov decision process. To overcome the limitations of DT, we propose a novel action sequence predictor, named Decision ConvFormer (DC), based on the architecture of MetaFormer, which is a general structure to process multiple entities in parallel and understand the interrelationship among the multiple entities. DC employs local convolution filtering as the token mixer and can effectively capture the inherent local associations of the RL dataset. In extensive experiments, DC achieved state-of-the-art performance across various standard RL benchmarks while requiring fewer resources. Furthermore, we show that DC better understands the underlying meaning in data and exhibits enhanced generalization capability."
                },
                {
                    "title": "ACT: Empowering Decision Transformer with Dynamic Programming via Advantage Conditioning",
                    "abstract": "Decision Transformer (DT), which employs expressive sequence modeling techniques to perform action generation, has emerged as a promising approach to offline policy optimization. However, DT generates actions conditioned on a desired future return, which is known to bear some weaknesses such as the susceptibility to environmental stochasticity. To overcome DT's weaknesses, we propose to empower DT with dynamic programming. Our method comprises three steps. First, we employ in-sample value iteration to obtain approximated value functions, which involves dynamic programming over the MDP structure. Second, we evaluate action quality in context with estimated advantages. We introduce two types of advantage estimators, IAE and GAE, which are suitable for different tasks. Third, we train an Advantage-Conditioned Transformer (ACT) to generate actions conditioned on the estimated advantages. Finally, during testing, ACT generates actions conditioned on a desired advantage. Our evaluation results validate that, by leveraging the power of dynamic programming, ACT demonstrates effective trajectory stitching and robust action generation in spite of the environmental stochasticity, outperforming baseline methods across various benchmarks. Additionally, we conduct an in-depth analysis of ACT's various design choices through ablation studies. Our code is available at https://github.com/LAMDA-RL/ACT."
                },
                {
                    "title": "Elastic Decision Transformer",
                    "abstract": "This paper introduces Elastic Decision Transformer (EDT), a significant advancement over the existing Decision Transformer (DT) and its variants. Although DT purports to generate an optimal trajectory, empirical evidence suggests it struggles with trajectory stitching, a process involving the generation of an optimal or near-optimal trajectory from the best parts of a set of sub-optimal trajectories. The proposed EDT differentiates itself by facilitating trajectory stitching during action inference at test time, achieved by adjusting the history length maintained in DT. Further, the EDT optimizes the trajectory by retaining a longer history when the previous trajectory is optimal and a shorter one when it is sub-optimal, enabling it to\"stitch\"with a more optimal trajectory. Extensive experimentation demonstrates EDT's ability to bridge the performance gap between DT-based and Q Learning-based approaches. In particular, the EDT outperforms Q Learning-based methods in a multi-task regime on the D4RL locomotion benchmark and Atari games. Videos are available at: https://kristery.github.io/edt/"
                },
                {
                    "title": "Offline RL with No OOD Actions: In-Sample Learning via Implicit Value Regularization",
                    "abstract": "Most offline reinforcement learning (RL) methods suffer from the trade-off between improving the policy to surpass the behavior policy and constraining the policy to limit the deviation from the behavior policy as computing $Q$-values using out-of-distribution (OOD) actions will suffer from errors due to distributional shift. The recently proposed \\textit{In-sample Learning} paradigm (i.e., IQL), which improves the policy by quantile regression using only data samples, shows great promise because it learns an optimal policy without querying the value function of any unseen actions. However, it remains unclear how this type of method handles the distributional shift in learning the value function. In this work, we make a key finding that the in-sample learning paradigm arises under the \\textit{Implicit Value Regularization} (IVR) framework. This gives a deeper understanding of why the in-sample learning paradigm works, i.e., it applies implicit value regularization to the policy. Based on the IVR framework, we further propose two practical algorithms, Sparse $Q$-learning (SQL) and Exponential $Q$-learning (EQL), which adopt the same value regularization used in existing works, but in a complete in-sample manner. Compared with IQL, we find that our algorithms introduce sparsity in learning the value function, making them more robust in noisy data regimes. We also verify the effectiveness of SQL and EQL on D4RL benchmark datasets and show the benefits of in-sample learning by comparing them with CQL in small data regimes."
                },
                {
                    "title": "Efficient Online Reinforcement Learning with Offline Data",
                    "abstract": "Sample efficiency and exploration remain major challenges in online reinforcement learning (RL). A powerful approach that can be applied to address these issues is the inclusion of offline data, such as prior trajectories from a human expert or a sub-optimal exploration policy. Previous methods have relied on extensive modifications and additional complexity to ensure the effective use of this data. Instead, we ask: can we simply apply existing off-policy methods to leverage offline data when learning online? In this work, we demonstrate that the answer is yes; however, a set of minimal but important changes to existing off-policy RL algorithms are required to achieve reliable performance. We extensively ablate these design choices, demonstrating the key factors that most affect performance, and arrive at a set of recommendations that practitioners can readily apply, whether their data comprise a small number of expert demonstrations or large volumes of sub-optimal trajectories. We see that correct application of these simple recommendations can provide a $\\mathbf{2.5\\times}$ improvement over existing approaches across a diverse set of competitive benchmarks, with no additional computational overhead. We have released our code at https://github.com/ikostrikov/rlpd."
                },
                {
                    "title": "A Policy-Guided Imitation Approach for Offline Reinforcement Learning",
                    "abstract": "Offline reinforcement learning (RL) methods can generally be categorized into two types: RL-based and Imitation-based. RL-based methods could in principle enjoy out-of-distribution generalization but suffer from erroneous off-policy evaluation. Imitation-based methods avoid off-policy evaluation but are too conservative to surpass the dataset. In this study, we propose an alternative approach, inheriting the training stability of imitation-style methods while still allowing logical out-of-distribution generalization. We decompose the conventional reward-maximizing policy in offline RL into a guide-policy and an execute-policy. During training, the guide-poicy and execute-policy are learned using only data from the dataset, in a supervised and decoupled manner. During evaluation, the guide-policy guides the execute-policy by telling where it should go so that the reward can be maximized, serving as the \\textit{Prophet}. By doing so, our algorithm allows \\textit{state-compositionality} from the dataset, rather than \\textit{action-compositionality} conducted in prior imitation-style methods. We dumb this new approach Policy-guided Offline RL (\\texttt{POR}). \\texttt{POR} demonstrates the state-of-the-art performance on D4RL, a standard benchmark for offline RL. We also highlight the benefits of \\texttt{POR} in terms of improving with supplementary suboptimal data and easily adapting to new tasks by only changing the guide-poicy."
                },
                {
                    "title": "Q-learning Decision Transformer: Leveraging Dynamic Programming for Conditional Sequence Modelling in Offline RL",
                    "abstract": "Recent works have shown that tackling offline reinforcement learning (RL) with a conditional policy produces promising results. The Decision Transformer (DT) combines the conditional policy approach and a transformer architecture, showing competitive performance against several benchmarks. However, DT lacks stitching ability -- one of the critical abilities for offline RL to learn the optimal policy from sub-optimal trajectories. This issue becomes particularly significant when the offline dataset only contains sub-optimal trajectories. On the other hand, the conventional RL approaches based on Dynamic Programming (such as Q-learning) do not have the same limitation; however, they suffer from unstable learning behaviours, especially when they rely on function approximation in an off-policy learning setting. In this paper, we propose the Q-learning Decision Transformer (QDT) to address the shortcomings of DT by leveraging the benefits of Dynamic Programming (Q-learning). It utilises the Dynamic Programming results to relabel the return-to-go in the training data to then train the DT with the relabelled data. Our approach efficiently exploits the benefits of these two approaches and compensates for each other's shortcomings to achieve better performance. We empirically show these in both simple toy environments and the more complex D4RL benchmark, showing competitive performance gains."
                },
                {
                    "title": "Diffusion Policies as an Expressive Policy Class for Offline Reinforcement Learning",
                    "abstract": "Offline reinforcement learning (RL), which aims to learn an optimal policy using a previously collected static dataset, is an important paradigm of RL. Standard RL methods often perform poorly in this regime due to the function approximation errors on out-of-distribution actions. While a variety of regularization methods have been proposed to mitigate this issue, they are often constrained by policy classes with limited expressiveness that can lead to highly suboptimal solutions. In this paper, we propose representing the policy as a diffusion model, a recent class of highly-expressive deep generative models. We introduce Diffusion Q-learning (Diffusion-QL) that utilizes a conditional diffusion model to represent the policy. In our approach, we learn an action-value function and we add a term maximizing action-values into the training loss of the conditional diffusion model, which results in a loss that seeks optimal actions that are near the behavior policy. We show the expressiveness of the diffusion model-based policy, and the coupling of the behavior cloning and policy improvement under the diffusion model both contribute to the outstanding performance of Diffusion-QL. We illustrate the superiority of our method compared to prior works in a simple 2D bandit example with a multimodal behavior policy. We then show that our method can achieve state-of-the-art performance on the majority of the D4RL benchmark tasks."
                },
                {
                    "title": "When does return-conditioned supervised learning work for offline reinforcement learning?",
                    "abstract": "Several recent works have proposed a class of algorithms for the offline reinforcement learning (RL) problem that we will refer to as return-conditioned supervised learning (RCSL). RCSL algorithms learn the distribution of actions conditioned on both the state and the return of the trajectory. Then they define a policy by conditioning on achieving high return. In this paper, we provide a rigorous study of the capabilities and limitations of RCSL, something which is crucially missing in previous work. We find that RCSL returns the optimal policy under a set of assumptions that are stronger than those needed for the more traditional dynamic programming-based algorithms. We provide specific examples of MDPs and datasets that illustrate the necessity of these assumptions and the limits of RCSL. Finally, we present empirical evidence that these limitations will also cause issues in practice by providing illustrative experiments in simple point-mass environments and on datasets from the D4RL benchmark."
                },
                {
                    "title": "When Should We Prefer Offline Reinforcement Learning Over Behavioral Cloning?",
                    "abstract": "Offline reinforcement learning (RL) algorithms can acquire effective policies by utilizing previously collected experience, without any online interaction. It is widely understood that offline RL is able to extract good policies even from highly suboptimal data, a scenario where imitation learning finds suboptimal solutions that do not improve over the demonstrator that generated the dataset. However, another common use case for practitioners is to learn from data that resembles demonstrations. In this case, one can choose to apply offline RL, but can also use behavioral cloning (BC) algorithms, which mimic a subset of the dataset via supervised learning. Therefore, it seems natural to ask: when can an offline RL method outperform BC with an equal amount of expert data, even when BC is a natural choice? To answer this question, we characterize the properties of environments that allow offline RL methods to perform better than BC methods, even when only provided with expert data. Additionally, we show that policies trained on sufficiently noisy suboptimal data can attain better performance than even BC algorithms with expert data, especially on long-horizon problems. We validate our theoretical results via extensive experiments on both diagnostic and high-dimensional domains including robotic manipulation, maze navigation, and Atari games, with a variety of data distributions. We observe that, under specific but common conditions such as sparse rewards or noisy data sources, modern offline RL methods can significantly outperform BC."
                },
                {
                    "title": "Supported Policy Optimization for Offline Reinforcement Learning",
                    "abstract": "Policy constraint methods to offline reinforcement learning (RL) typically utilize parameterization or regularization that constrains the policy to perform actions within the support set of the behavior policy. The elaborative designs of parameterization methods usually intrude into the policy networks, which may bring extra inference cost and cannot take full advantage of well-established online methods. Regularization methods reduce the divergence between the learned policy and the behavior policy, which may mismatch the inherent density-based definition of support set thereby failing to avoid the out-of-distribution actions effectively. This paper presents Supported Policy OpTimization (SPOT), which is directly derived from the theoretical formalization of the density-based support constraint. SPOT adopts a VAE-based density estimator to explicitly model the support set of behavior policy and presents a simple but effective density-based regularization term, which can be plugged non-intrusively into off-the-shelf off-policy RL algorithms. SPOT achieves the state-of-the-art performance on standard benchmarks for offline RL. Benefiting from the pluggable design, offline pretrained models from SPOT can also be applied to perform online fine-tuning seamlessly."
                },
                {
                    "title": "Online Decision Transformer",
                    "abstract": "Recent work has shown that offline reinforcement learning (RL) can be formulated as a sequence modeling problem (Chen et al., 2021; Janner et al., 2021) and solved via approaches similar to large-scale language modeling. However, any practical instantiation of RL also involves an online component, where policies pretrained on passive offline datasets are finetuned via taskspecific interactions with the environment. We propose Online Decision Transformers (ODT), an RL algorithm based on sequence modeling that blends offline pretraining with online finetuning in a unified framework. Our framework uses sequence-level entropy regularizers in conjunction with autoregressive modeling objectives for sample-efficient exploration and finetuning. Empirically, we show that ODT is competitive with the state-of-the-art in absolute performance on the D4RL benchmark but shows much more significant gains during the finetuning procedure."
                },
                {
                    "title": "RvS: What is Essential for Offline RL via Supervised Learning?",
                    "abstract": "Recent work has shown that supervised learning alone, without temporal difference (TD) learning, can be remarkably effective for offline RL. When does this hold true, and which algorithmic components are necessary? Through extensive experiments, we boil supervised learning for offline RL down to its essential elements. In every environment suite we consider, simply maximizing likelihood with a two-layer feedforward MLP is competitive with state-of-the-art results of substantially more complex methods based on TD learning or sequence modeling with Transformers. Carefully choosing model capacity (e.g., via regularization or architecture) and choosing which information to condition on (e.g., goals or rewards) are critical for performance. These insights serve as a field guide for practitioners doing Reinforcement Learning via Supervised Learning (which we coin\"RvS learning\"). They also probe the limits of existing RvS methods, which are comparatively weak on random data, and suggest a number of open problems."
                },
                {
                    "title": "DR3: Value-Based Deep Reinforcement Learning Requires Explicit Regularization",
                    "abstract": "Despite overparameterization, deep networks trained via supervised learning are easy to optimize and exhibit excellent generalization. One hypothesis to explain this is that overparameterized deep networks enjoy the benefits of implicit regularization induced by stochastic gradient descent, which favors parsimonious solutions that generalize well on test inputs. It is reasonable to surmise that deep reinforcement learning (RL) methods could also benefit from this effect. In this paper, we discuss how the implicit regularization effect of SGD seen in supervised learning could in fact be harmful in the offline deep RL setting, leading to poor generalization and degenerate feature representations. Our theoretical analysis shows that when existing models of implicit regularization are applied to temporal difference learning, the resulting derived regularizer favors degenerate solutions with excessive\"aliasing\", in stark contrast to the supervised learning case. We back up these findings empirically, showing that feature representations learned by a deep network value function trained via bootstrapping can indeed become degenerate, aliasing the representations for state-action pairs that appear on either side of the Bellman backup. To address this issue, we derive the form of this implicit regularizer and, inspired by this derivation, propose a simple and effective explicit regularizer, called DR3, that counteracts the undesirable effects of this implicit regularizer. When combined with existing offline RL methods, DR3 substantially improves performance and stability, alleviating unlearning in Atari 2600 games, D4RL domains and robotic manipulation from images."
                },
                {
                    "title": "Offline Reinforcement Learning with Implicit Q-Learning",
                    "abstract": "Offline reinforcement learning requires reconciling two conflicting aims: learning a policy that improves over the behavior policy that collected the dataset, while at the same time minimizing the deviation from the behavior policy so as to avoid errors due to distributional shift. This trade-off is critical, because most current offline reinforcement learning methods need to query the value of unseen actions during training to improve the policy, and therefore need to either constrain these actions to be in-distribution, or else regularize their values. We propose an offline RL method that never needs to evaluate actions outside of the dataset, but still enables the learned policy to improve substantially over the best behavior in the data through generalization. The main insight in our work is that, instead of evaluating unseen actions from the latest policy, we can approximate the policy improvement step implicitly by treating the state value function as a random variable, with randomness determined by the action (while still integrating over the dynamics to avoid excessive optimism), and then taking a state conditional upper expectile of this random variable to estimate the value of the best actions in that state. This leverages the generalization capacity of the function approximator to estimate the value of the best available action at a given state without ever directly querying a Q-function with this unseen action. Our algorithm alternates between fitting this upper expectile value function and backing it up into a Q-function. Then, we extract the policy via advantage-weighted behavioral cloning. We dub our method implicit Q-learning (IQL). IQL demonstrates the state-of-the-art performance on D4RL, a standard benchmark for offline reinforcement learning. We also demonstrate that IQL achieves strong performance fine-tuning using online interaction after offline initialization."
                },
                {
                    "title": "Uncertainty-Based Offline Reinforcement Learning with Diversified Q-Ensemble",
                    "abstract": "Offline reinforcement learning (offline RL), which aims to find an optimal policy from a previously collected static dataset, bears algorithmic difficulties due to function approximation errors from out-of-distribution (OOD) data points. To this end, offline RL algorithms adopt either a constraint or a penalty term that explicitly guides the policy to stay close to the given dataset. However, prior methods typically require accurate estimation of the behavior policy or sampling from OOD data points, which themselves can be a non-trivial problem. Moreover, these methods under-utilize the generalization ability of deep neural networks and often fall into suboptimal solutions too close to the given dataset. In this work, we propose an uncertainty-based offline RL method that takes into account the confidence of the Q-value prediction and does not require any estimation or sampling of the data distribution. We show that the clipped Q-learning, a technique widely used in online RL, can be leveraged to successfully penalize OOD data points with high prediction uncertainties. Surprisingly, we find that it is possible to substantially outperform existing offline RL methods on various tasks by simply increasing the number of Q-networks along with the clipped Q-learning. Based on this observation, we propose an ensemble-diversified actor-critic algorithm that reduces the number of required ensemble networks down to a tenth compared to the naive ensemble while achieving state-of-the-art performance on most of the D4RL benchmarks considered."
                },
                {
                    "title": "A Minimalist Approach to Offline Reinforcement Learning",
                    "abstract": "Offline reinforcement learning (RL) defines the task of learning from a fixed batch of data. Due to errors in value estimation from out-of-distribution actions, most offline RL algorithms take the approach of constraining or regularizing the policy with the actions contained in the dataset. Built on pre-existing RL algorithms, modifications to make an RL algorithm work offline comes at the cost of additional complexity. Offline RL algorithms introduce new hyperparameters and often leverage secondary components such as generative models, while adjusting the underlying RL algorithm. In this paper we aim to make a deep RL algorithm work while making minimal changes. We find that we can match the performance of state-of-the-art offline RL algorithms by simply adding a behavior cloning term to the policy update of an online RL algorithm and normalizing the data. The resulting algorithm is a simple to implement and tune baseline, while more than halving the overall run time by removing the additional computational overhead of previous methods."
                },
                {
                    "title": "Decision Transformer: Reinforcement Learning via Sequence Modeling",
                    "abstract": "We introduce a framework that abstracts Reinforcement Learning (RL) as a sequence modeling problem. This allows us to draw upon the simplicity and scalability of the Transformer architecture, and associated advances in language modeling such as GPT-x and BERT. In particular, we present Decision Transformer, an architecture that casts the problem of RL as conditional sequence modeling. Unlike prior approaches to RL that fit value functions or compute policy gradients, Decision Transformer simply outputs the optimal actions by leveraging a causally masked Transformer. By conditioning an autoregressive model on the desired return (reward), past states, and actions, our Decision Transformer model can generate future actions that achieve the desired return. Despite its simplicity, Decision Transformer matches or exceeds the performance of state-of-the-art model-free offline RL baselines on Atari, OpenAI Gym, and Key-to-Door tasks."
                },
                {
                    "title": "Uncertainty Weighted Actor-Critic for Offline Reinforcement Learning",
                    "abstract": "Offline Reinforcement Learning promises to learn effective policies from previously-collected, static datasets without the need for exploration. However, existing Q-learning and actor-critic based off-policy RL algorithms fail when bootstrapping from out-of-distribution (OOD) actions or states. We hypothesize that a key missing ingredient from the existing methods is a proper treatment of uncertainty in the offline setting. We propose Uncertainty Weighted Actor-Critic (UWAC), an algorithm that detects OOD state-action pairs and down-weights their contribution in the training objectives accordingly. Implementation-wise, we adopt a practical and effective dropout-based uncertainty estimation method that introduces very little overhead over existing RL algorithms. Empirically, we observe that UWAC substantially improves model stability during training. In addition, UWAC out-performs existing offline RL methods on a variety of competitive tasks, and achieves significant performance gains over the state-of-the-art baseline on datasets with sparse demonstrations collected from human experts."
                },
                {
                    "title": "Offline Reinforcement Learning with Fisher Divergence Critic Regularization",
                    "abstract": "Many modern approaches to offline Reinforcement Learning (RL) utilize behavior regularization, typically augmenting a model-free actor critic algorithm with a penalty measuring divergence of the policy from the offline data. In this work, we propose an alternative approach to encouraging the learned policy to stay close to the data, namely parameterizing the critic as the log-behavior-policy, which generated the offline data, plus a state-action value offset term, which can be learned using a neural network. Behavior regularization then corresponds to an appropriate regularizer on the offset term. We propose using a gradient penalty regularizer for the offset term and demonstrate its equivalence to Fisher divergence regularization, suggesting connections to the score matching and generative energy-based model literature. We thus term our resulting algorithm Fisher-BRC (Behavior Regularized Critic). On standard offline RL benchmarks, Fisher-BRC achieves both improved performance and faster convergence over existing state-of-the-art methods."
                },
                {
                    "title": "COMBO: Conservative Offline Model-Based Policy Optimization",
                    "abstract": "Model-based algorithms, which learn a dynamics model from logged experience and perform some sort of pessimistic planning under the learned model, have emerged as a promising paradigm for offline reinforcement learning (offline RL). However, practical variants of such model-based algorithms rely on explicit uncertainty quantification for incorporating pessimism. Uncertainty estimation with complex models, such as deep neural networks, can be difficult and unreliable. We overcome this limitation by developing a new model-based offline RL algorithm, COMBO, that regularizes the value function on out-of-support state-action tuples generated via rollouts under the learned model. This results in a conservative estimate of the value function for out-of-support state-action tuples, without requiring explicit uncertainty estimation. We theoretically show that our method optimizes a lower bound on the true policy value, that this bound is tighter than that of prior methods, and our approach satisfies a policy improvement guarantee in the offline setting. Through experiments, we find that COMBO consistently performs as well or better as compared to prior offline model-free and model-based methods on widely studied offline RL benchmarks, including image-based tasks."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Critic Regularized Regression",
                    "abstract": "Offline reinforcement learning (RL), also known as batch RL, offers the prospect of policy optimization from large pre-recorded datasets without online environment interaction. It addresses challenges with regard to the cost of data collection and safety, both of which are particularly pertinent to real-world applications of RL. Unfortunately, most off-policy algorithms perform poorly when learning from a fixed dataset. In this paper, we propose a novel offline RL algorithm to learn policies from data using a form of critic-regularized regression (CRR). We find that CRR performs surprisingly well and scales to tasks with high-dimensional state and action spaces -- outperforming several state-of-the-art offline RL algorithms by a significant margin on a wide range of benchmark tasks."
                },
                {
                    "title": "Accelerating Online Reinforcement Learning with Offline Datasets",
                    "abstract": "Reinforcement learning provides an appealing formalism for learning control policies from experience. However, the classic active formulation of reinforcement learning necessitates a lengthy active exploration process for each behavior, making it difficult to apply in real-world settings. If we can instead allow reinforcement learning to effectively use previously collected data to aid the online learning process, where the data could be expert demonstrations or more generally any prior experience, we could make reinforcement learning a substantially more practical tool. While a number of recent methods have sought to learn offline from previously collected data, it remains exceptionally difficult to train a policy with offline data and improve it further with online reinforcement learning. In this paper we systematically analyze why this problem is so challenging, and propose a novel algorithm that combines sample-efficient dynamic programming with maximum likelihood policy updates, providing a simple and effective framework that is able to leverage large amounts of offline data and then quickly perform online fine-tuning of reinforcement learning policies. We show that our method enables rapid learning of skills with a combination of prior demonstration data and online experience across a suite of difficult dexterous manipulation and benchmark tasks."
                },
                {
                    "title": "Conservative Q-Learning for Offline Reinforcement Learning",
                    "abstract": "Effectively leveraging large, previously collected datasets in reinforcement learning (RL) is a key challenge for large-scale real-world applications. Offline RL algorithms promise to learn effective policies from previously-collected, static datasets without further interaction. However, in practice, offline RL presents a major challenge, and standard off-policy RL methods can fail due to overestimation of values induced by the distributional shift between the dataset and the learned policy, especially when training on complex and multi-modal data distributions. In this paper, we propose conservative Q-learning (CQL), which aims to address these limitations by learning a conservative Q-function such that the expected value of a policy under this Q-function lower-bounds its true value. We theoretically show that CQL produces a lower bound on the value of the current policy and that it can be incorporated into a policy learning procedure with theoretical improvement guarantees. In practice, CQL augments the standard Bellman error objective with a simple Q-value regularizer which is straightforward to implement on top of existing deep Q-learning and actor-critic implementations. On both discrete and continuous control domains, we show that CQL substantially outperforms existing offline RL methods, often learning policies that attain 2-5 times higher final return, especially when learning from complex and multi-modal data distributions."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems",
                    "abstract": "In this tutorial article, we aim to provide the reader with the conceptual tools needed to get started on research on offline reinforcement learning algorithms: reinforcement learning algorithms that utilize previously collected data, without additional online data collection. Offline reinforcement learning algorithms hold tremendous promise for making it possible to turn large datasets into powerful decision making engines. Effective offline reinforcement learning methods would be able to extract policies with the maximum possible utility out of the available data, thereby allowing automation of a wide range of decision-making domains, from healthcare and education to robotics. However, the limitations of current algorithms make this difficult. We will aim to provide the reader with an understanding of these challenges, particularly in the context of modern deep reinforcement learning methods, and describe some potential solutions that have been explored in recent work to mitigate these challenges, along with recent applications, and a discussion of perspectives on open problems in the field."
                },
                {
                    "title": "D4RL: Datasets for Deep Data-Driven Reinforcement Learning",
                    "abstract": "The offline reinforcement learning (RL) problem, also known as batch RL, refers to the setting where a policy must be learned from a static dataset, without additional online data collection. This setting is compelling as potentially it allows RL methods to take advantage of large, pre-collected datasets, much like how the rise of large datasets has fueled results in supervised learning in recent years. However, existing online RL benchmarks are not tailored towards the offline setting, making progress in offline RL difficult to measure. In this work, we introduce benchmarks specifically designed for the offline setting, guided by key properties of datasets relevant to real-world applications of offline RL. Examples of such properties include: datasets generated via hand-designed controllers and human demonstrators, multi-objective datasets where an agent can perform different tasks in the same environment, and datasets consisting of a mixtures of policies. To facilitate research, we release our benchmark tasks and datasets with a comprehensive evaluation of existing algorithms and an evaluation protocol together with an open-source codebase. We hope that our benchmark will focus research effort on methods that drive improvements not just on simulated tasks, but ultimately on the kinds of real-world problems where offline RL will have the largest impact."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Relay Policy Learning: Solving Long-Horizon Tasks via Imitation and Reinforcement Learning",
                    "abstract": "We present relay policy learning, a method for imitation and reinforcement learning that can solve multi-stage, long-horizon robotic tasks. This general and universally-applicable, two-phase approach consists of an imitation learning stage that produces goal-conditioned hierarchical policies, and a reinforcement learning phase that finetunes these policies for task performance. Our method, while not necessarily perfect at imitation learning, is very amenable to further improvement via environment interaction, allowing it to scale to challenging long-horizon tasks. We simplify the long-horizon policy learning problem by using a novel data-relabeling algorithm for learning goal-conditioned hierarchical policies, where the low-level only acts for a fixed number of steps, regardless of the goal achieved. While we rely on demonstration data to bootstrap policy learning, we do not assume access to demonstrations of every specific tasks that is being solved, and instead leverage unstructured and unsegmented demonstrations of semantically meaningful behaviors that are not only less burdensome to provide, but also can greatly facilitate further improvement using reinforcement learning. We demonstrate the effectiveness of our method on a number of multi-stage, long-horizon manipulation tasks in a challenging kitchen simulation environment. Videos are available at this https URL"
                },
                {
                    "title": "Advantage-Weighted Regression: Simple and Scalable Off-Policy Reinforcement Learning",
                    "abstract": "In this paper, we aim to develop a simple and scalable reinforcement learning algorithm that uses standard supervised learning methods as subroutines. Our goal is an algorithm that utilizes only simple and convergent maximum likelihood loss functions, while also being able to leverage off-policy data. Our proposed approach, which we refer to as advantage-weighted regression (AWR), consists of two standard supervised learning steps: one to regress onto target values for a value function, and another to regress onto weighted target actions for the policy. The method is simple and general, can accommodate continuous and discrete actions, and can be implemented in just a few lines of code on top of standard supervised learning methods. We provide a theoretical motivation for AWR and analyze its properties when incorporating off-policy data from experience replay. We evaluate AWR on a suite of standard OpenAI Gym benchmark tasks, and show that it achieves competitive performance compared to a number of well-established state-of-the-art RL algorithms. AWR is also able to acquire more effective policies than most off-policy algorithms when learning from purely static datasets with no additional environmental interactions. Furthermore, we demonstrate our algorithm on challenging continuous control tasks with highly complex simulated characters."
                },
                {
                    "title": "Behavior Regularized Offline Reinforcement Learning",
                    "abstract": "In reinforcement learning (RL) research, it is common to assume access to direct online interactions with the environment. However in many real-world applications, access to the environment is limited to a fixed offline dataset of logged experience. In such settings, standard RL algorithms have been shown to diverge or otherwise yield poor performance. Accordingly, recent work has suggested a number of remedies to these issues. In this work, we introduce a general framework, behavior regularized actor critic (BRAC), to empirically evaluate recently proposed methods as well as a number of simple baselines across a variety of offline continuous control tasks. Surprisingly, we find that many of the technical complexities introduced in recent methods are unnecessary to achieve strong performance. Additional ablations provide insights into which design choices matter most in the offline RL setting."
                },
                {
                    "title": "Stabilizing Off-Policy Q-Learning via Bootstrapping Error Reduction",
                    "abstract": "Off-policy reinforcement learning aims to leverage experience collected from prior policies for sample-efficient learning. However, in practice, commonly used off-policy approximate dynamic programming methods based on Q-learning and actor-critic methods are highly sensitive to the data distribution, and can make only limited progress without collecting additional on-policy data. As a step towards more robust off-policy algorithms, we study the setting where the off-policy experience is fixed and there is no further interaction with the environment. We identify bootstrapping error as a key source of instability in current methods. Bootstrapping error is due to bootstrapping from actions that lie outside of the training data distribution, and it accumulates via the Bellman backup operator. We theoretically analyze bootstrapping error, and demonstrate how carefully constraining action selection in the backup can mitigate it. Based on our analysis, we propose a practical algorithm, bootstrapping error accumulation reduction (BEAR). We demonstrate that BEAR is able to learn robustly from different off-policy distributions, including random and suboptimal demonstrations, on a range of continuous control tasks."
                },
                {
                    "title": "Towards Characterizing Divergence in Deep Q-Learning",
                    "abstract": "Deep Q-Learning (DQL), a family of temporal difference algorithms for control, employs three techniques collectively known as the `deadly triad' in reinforcement learning: bootstrapping, off-policy learning, and function approximation. Prior work has demonstrated that together these can lead to divergence in Q-learning algorithms, but the conditions under which divergence occurs are not well-understood. In this note, we give a simple analysis based on a linear approximation to the Q-value updates, which we believe provides insight into divergence under the deadly triad. The central point in our analysis is to consider when the leading order approximation to the deep-Q update is or is not a contraction in the sup norm. Based on this analysis, we develop an algorithm which permits stable deep Q-learning for continuous control without any of the tricks conventionally used (such as target networks, adaptive gradient optimizers, or using multiple Q functions). We demonstrate that our algorithm performs above or near state-of-the-art on standard MuJoCo benchmarks from the OpenAI Gym."
                },
                {
                    "title": "Wide neural networks of any depth evolve as linear models under gradient descent",
                    "abstract": "A longstanding goal in deep learning research has been to precisely characterize training and generalization. However, the often complex loss landscapes of neural networks (NNs) have made a theory of learning dynamics elusive. In this work, we show that for wide NNs the learning dynamics simplify considerably and that, in the infinite width limit, they are governed by a linear model obtained from the first-order Taylor expansion of the network around its initial parameters. Furthermore, mirroring the correspondence between wide Bayesian NNs and Gaussian processes (GPs), gradient-based training of wide NNs with a squared loss produces test set predictions drawn from a GP with a particular compositional kernel. While these theoretical results are only exact in the infinite width limit, we nevertheless find excellent empirical agreement between the predictions of the original network and those of the linearized version even for finite practically-sized networks. This agreement is robust across different architectures, optimization methods, and loss functions."
                },
                {
                    "title": "Off-Policy Deep Reinforcement Learning without Exploration",
                    "abstract": "Many practical applications of reinforcement learning constrain agents to learn from a fixed batch of data which has already been gathered, without offering further possibility for data collection. In this paper, we demonstrate that due to errors introduced by extrapolation, standard off-policy deep reinforcement learning algorithms, such as DQN and DDPG, are incapable of learning with data uncorrelated to the distribution under the current policy, making them ineffective for this fixed batch setting. We introduce a novel class of off-policy algorithms, batch-constrained reinforcement learning, which restricts the action space in order to force the agent towards behaving close to on-policy with respect to a subset of the given data. We present the first continuous control deep reinforcement learning algorithm which can learn effectively from arbitrary, fixed batch data, and empirically demonstrate the quality of its behavior in several tasks."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "Deep Learning using Rectified Linear Units (ReLU)",
                    "abstract": "We introduce the use of rectified linear units (ReLU) as the classification function in a deep neural network (DNN). Conventionally, ReLU is used as an activation function in DNNs, with Softmax function as their classification function. However, there have been several studies on using a classification function other than Softmax, and this study is an addition to those. We accomplish this by taking the activation of the penultimate layer $h_{n - 1}$ in a neural network, then multiply it by weight parameters $\\theta$ to get the raw scores $o_{i}$. Afterwards, we threshold the raw scores $o_{i}$ by $0$, i.e. $f(o) = \\max(0, o_{i})$, where $f(o)$ is the ReLU function. We provide class predictions $\\hat{y}$ through argmax function, i.e. argmax $f(x)$."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrations",
                    "abstract": "Dexterous multi-fingered hands are extremely versatile and provide a generic way to perform multiple tasks in human-centric environments. However, effectively controlling them remains challenging due to their high dimensionality and large number of potential contacts. Deep reinforcement learning (DRL) provides a model-agnostic approach to control complex dynamical systems, but has not been shown to scale to high-dimensional dexterous manipulation. Furthermore, deployment of DRL on physical systems remains challenging due to sample inefficiency. Thus, the success of DRL in robotics has thus far been limited to simpler manipulators and tasks. In this work, we show that model-free DRL with natural policy gradients can effectively scale up to complex manipulation tasks with a high-dimensional 24-DoF hand, and solve them from scratch in simulated experiments. Furthermore, with the use of a small number of human demonstrations, the sample complexity can be significantly reduced, and enable learning within the equivalent of a few hours of robot experience. We demonstrate successful policies for multiple complex tasks: object relocation, in-hand manipulation, tool use, and door opening."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Layer Normalization",
                    "abstract": "Training state-of-the-art, deep neural networks is computationally expensive. One way to reduce the training time is to normalize the activities of the neurons. A recently introduced technique called batch normalization uses the distribution of the summed input to a neuron over a mini-batch of training cases to compute a mean and variance which are then used to normalize the summed input to that neuron on each training case. This significantly reduces the training time in feed-forward neural networks. However, the effect of batch normalization is dependent on the mini-batch size and it is not obvious how to apply it to recurrent neural networks. In this paper, we transpose batch normalization into layer normalization by computing the mean and variance used for normalization from all of the summed inputs to the neurons in a layer on a single training case. Like batch normalization, we also give each neuron its own adaptive bias and gain which are applied after the normalization but before the non-linearity. Unlike batch normalization, layer normalization performs exactly the same computation at training and test times. It is also straightforward to apply to recurrent neural networks by computing the normalization statistics separately at each time step. Layer normalization is very effective at stabilizing the hidden state dynamics in recurrent networks. Empirically, we show that layer normalization can substantially reduce the training time compared with previously published techniques."
                },
                {
                    "title": "MuJoCo: A physics engine for model-based control",
                    "abstract": "We describe a new physics engine tailored to model-based control. Multi-joint dynamics are represented in generalized coordinates and computed via recursive algorithms. Contact responses are computed via efficient new algorithms we have developed, based on the modern velocity-stepping approach which avoids the difficulties with spring-dampers. Models are specified using either a high-level C++ API or an intuitive XML file format. A built-in compiler transforms the user model into an optimized data structure used for runtime computation. The engine can compute both forward and inverse dynamics. The latter are well-defined even in the presence of contacts and equality constraints. The model can include tendon wrapping as well as actuator activation states (e.g. pneumatic cylinders or muscles). To facilitate optimal control applications and in particular sampling and finite differencing, the dynamics can be evaluated for different states and controls in parallel. Around 400,000 dynamics evaluations per second are possible on a 12-core machine, for a 3D homanoid with 18 dofs and 6 active contacts. We have already used the engine in a number of control applications. It will soon be made publicly available."
                },
                {
                    "title": "Nearest Neighbor Estimates of Entropy",
                    "abstract": "SYNOPTIC ABSTRACT Motivated by the problems in molecular sciences, we introduce new nonparametric estimators of entropy which are based on the kth nearest neighbor distances between the n sample points, where k (< n \u2013 1) is a fixed positive integer. These provide competing estimators to an estimator proposed by Kozachenko and Leonenko (1987), which is based on the first nearest neighbor distances of the sample points. These estimators are helpful in the evaluation of entropies of random vectors. We establish the asymptotic unbiasedness and consistency of the proposed estimators. For some standard distributions, we also investigate their performance for finite sample sizes using Monte Carlo simulations. The proposed estimators are applied to estimate the entropy of internal rotation in the methanol molecule, which can be characterized by a one-dimensional random vector, and of diethyl ether, which is described by a four-dimensional random vector."
                },
                {
                    "title": "A Markovian Decision Process",
                    "abstract": "Abstract : The purpose of this paper is to discuss the asymptotic behavior of the sequence (f sub n(i)) generated by a nonlinear recurrence relation. This problem arises in connection with an equipment replacement problem, cf. S. Dreyfus, A Note on an Industrial Replacement Process."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "Visualizing Data using t-SNE",
                    "abstract": "We present a new technique called \u201ct-SNE\u201d that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to influence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are significantly better than those produced by the other techniques on almost all of the datasets."
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                },
                {
                    "title": "A Study of Generalization in Offline Reinforcement Learning",
                    "abstract": "Despite the recent progress in offline reinforcement learning (RL) algorithms, these methods are still trained and tested on the same environment. In this paper, we perform an in-depth study of the generalization abilities of offline RL algorithms, showing that they struggle to generalize to new environments. We also introduce the first benchmark for evaluating generalization in offline learning, collecting datasets of varying sizes and skill-levels from Procgen (2D video games) and WebShop (e-commerce websites). The datasets contain trajectories for a limited number of game levels or natural language instructions and at test time, the agent has to generalize to new levels or instructions. Our experiments reveal that existing offline learning algorithms perform significantly worse than online RL on both train and test environments. Behavioral cloning is a strong baseline, typically outperforming offline RL aprpoac and sequence modeling approaches when trained on data from multiple environments and tested on new ones. Finally, we find that increasing the diversity of the data, rather than its size, improves generalization for all algorithms. Our study demonstrates the limited generalization of current offline learning algorithms highlighting the need for more research in this area."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively integrate the stitching ability of the Q-function with return-conditioned supervised learning (RCSL) to improve decision-making in offline reinforcement learning?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing offline reinforcement learning, particularly in high-stakes fields like robotics, autonomous driving, and healthcare, where real-time experimentation is risky and costly. By enhancing the stitching ability of RCSL through the Q-function, we can improve the generation of optimal trajectories from suboptimal data, leading to more reliable decision-making systems. This research could pave the way for more robust algorithms that leverage past experiences effectively, ultimately influencing future research directions in offline RL and its applications.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complex interaction between RCSL and the Q-function. Naive approaches may fail because simply conditioning RCSL on the Q-function does not account for the stability required for effective stitching. The Q-function can become over-generalized when trained on expert datasets, leading to improper learning signals. Additionally, integrating the Q-function's stitching ability into RCSL without destabilizing the learning process presents significant technical and theoretical obstacles that must be addressed.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has attempted to enhance RCSL with the Q-function but has primarily focused on conditioning rather than fully leveraging its stitching capabilities. This has resulted in limited performance improvements due to improper integration and a lack of understanding of when the Q-function can effectively aid RCSL. Barriers such as the over-generalization of the Q-function when trained on expert data have also hindered progress. Our approach differs by proposing QQ-aided conditional supervised learning (QCS), which adaptively incorporates Q-aid into the RCSL framework based on trajectory returns, addressing these limitations.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, QCS, integrates Q-aid into the RCSL loss function based on trajectory returns. We will evaluate this approach using standard offline RL benchmarks, employing metrics such as mean normalized return to assess performance. The expected outcomes include significant improvements in decision-making capabilities over existing state-of-the-art methods, demonstrating the effectiveness of combining RCSL with the Q-function's stitching ability, particularly in generating optimal trajectories from suboptimal data."
            }
        },
        "author_data": {
            "ddd18b88-8b71-471a-a9fc-15aa494b2912": {
                "pk": "ddd18b88-8b71-471a-a9fc-15aa494b2912",
                "name": "Jeonghye Kim",
                "collaborators": [
                    "Woojun Kim",
                    "Youngchul Sung",
                    "Suyoung Lee"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Transformer",
                    "Deep Learning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "LESSON: Learning to Integrate Exploration Strategies for Reinforcement Learning via an Option Framework",
                        "abstract": "In this paper, a unified framework for exploration in reinforcement learning (RL) is proposed based on an option-critic model. The proposed framework learns to integrate a set of diverse exploration strategies so that the agent can adaptively select the most effective exploration strategy over time to realize a relevant exploration-exploitation trade-off for each given task. The effectiveness of the proposed exploration framework is demonstrated by various experiments in the MiniGrid and Atari environments."
                    },
                    {
                        "title": "Decision ConvFormer: Local Filtering in MetaFormer is Sufficient for Decision Making",
                        "abstract": "The recent success of Transformer in natural language processing has sparked its use in various domains. In offline reinforcement learning (RL), Decision Transformer (DT) is emerging as a promising model based on Transformer. However, we discovered that the attention module of DT is not appropriate to capture the inherent local dependence pattern in trajectories of RL modeled as a Markov decision process. To overcome the limitations of DT, we propose a novel action sequence predictor, named Decision ConvFormer (DC), based on the architecture of MetaFormer, which is a general structure to process multiple entities in parallel and understand the interrelationship among the multiple entities. DC employs local convolution filtering as the token mixer and can effectively capture the inherent local associations of the RL dataset. In extensive experiments, DC achieved state-of-the-art performance across various standard RL benchmarks while requiring fewer resources. Furthermore, we show that DC better understands the underlying meaning in data and exhibits enhanced generalization capability."
                    }
                ]
            },
            "162cbb47-01ac-4b38-8d52-18151b894822": {
                "pk": "162cbb47-01ac-4b38-8d52-18151b894822",
                "name": "Suyoung Lee",
                "collaborators": [
                    "Kyoung Mu Lee",
                    "Sooel Son",
                    "Sanghyun Son",
                    "Seungjun Nah",
                    "Radu Timofte",
                    "Myungsub Choi",
                    "Sae-Young Chung",
                    "Jeonghye Kim",
                    "Woojun Kim",
                    "Youngchul Sung"
                ],
                "domain": [
                    "Super Resolution",
                    "Meta Learning",
                    "Reinforcement Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "DynaVSR: Dynamic Adaptive Blind Video Super-Resolution",
                        "abstract": "Most conventional supervised super-resolution (SR) algorithms assume that low-resolution (LR) data is obtained by downscaling high-resolution (HR) data with a fixed known kernel, but such an assumption often does not hold in real scenarios. Some recent blind SR algorithms have been proposed to estimate different downscaling kernels for each input LR image. However, they suffer from heavy computational overhead, making them infeasible for direct application to videos. In this work, we present DynaVSR, a novel meta-learning-based framework for real-world video SR that enables efficient downscaling model estimation and adaptation to the current input. Specifically, we train a multi-frame downscaling module with various types of synthetic blur kernels, which is seamlessly combined with a video SR network for input-aware adaptation. Experimental results show that DynaVSR consistently improves the performance of the state-of-the-art video SR models by a large margin, with an order of magnitude faster inference time compared to the existing blind SR approaches."
                    },
                    {
                        "title": "Improving Generalization in Meta-RL with Imaginary Tasks from Latent Dynamics Mixture",
                        "abstract": "The generalization ability of most meta-reinforcement learning (meta-RL) methods is largely limited to test tasks that are sampled from the same distribution used to sample training tasks. To overcome the limitation, we propose Latent Dynamics Mixture (LDM) that trains a reinforcement learning agent with imaginary tasks generated from mixtures of learned latent dynamics. By training a policy on mixture tasks along with original training tasks, LDM allows the agent to prepare for unseen test tasks during training and prevents the agent from overfitting the training tasks. LDM significantly outperforms standard meta-RL methods in test returns on the gridworld navigation and MuJoCo tasks where we strictly separate the training task distribution and the test task distribution."
                    },
                    {
                        "title": "Decision ConvFormer: Local Filtering in MetaFormer is Sufficient for Decision Making",
                        "abstract": "The recent success of Transformer in natural language processing has sparked its use in various domains. In offline reinforcement learning (RL), Decision Transformer (DT) is emerging as a promising model based on Transformer. However, we discovered that the attention module of DT is not appropriate to capture the inherent local dependence pattern in trajectories of RL modeled as a Markov decision process. To overcome the limitations of DT, we propose a novel action sequence predictor, named Decision ConvFormer (DC), based on the architecture of MetaFormer, which is a general structure to process multiple entities in parallel and understand the interrelationship among the multiple entities. DC employs local convolution filtering as the token mixer and can effectively capture the inherent local associations of the RL dataset. In extensive experiments, DC achieved state-of-the-art performance across various standard RL benchmarks while requiring fewer resources. Furthermore, we show that DC better understands the underlying meaning in data and exhibits enhanced generalization capability."
                    },
                    {
                        "title": "Montage: A Neural Network Language Model-Guided JavaScript Engine Fuzzer",
                        "abstract": "JavaScript (JS) engine vulnerabilities pose significant security threats affecting billions of web browsers. While fuzzing is a prevalent technique for finding such vulnerabilities, there have been few studies that leverage the recent advances in neural network language models (NNLMs). In this paper, we present Montage, the first NNLM-guided fuzzer for finding JS engine vulnerabilities. The key aspect of our technique is to transform a JS abstract syntax tree (AST) into a sequence of AST subtrees that can directly train prevailing NNLMs. We demonstrate that Montage is capable of generating valid JS tests, and show that it outperforms previous studies in terms of finding vulnerabilities. Montage found 37 real-world bugs, including three CVEs, in the latest JS engines, demonstrating its efficacy in finding JS engine bugs."
                    },
                    {
                        "title": "Evaluating the Robustness of Trigger Set-Based Watermarks Embedded in Deep Neural Networks",
                        "abstract": "Trigger set-based watermarking schemes have gained emerging attention as they provide a means to prove ownership for deep neural network model owners. In this paper, we argue that state-of-the-art trigger set-based watermarking algorithms do not achieve their designed goal of proving ownership. We posit that this impaired capability stems from two common experimental flaws that the existing research practice has committed when evaluating the robustness of watermarking algorithms: (1) incomplete adversarial evaluation and (2) overlooked adaptive attacks. We conduct a comprehensive adversarial evaluation of 11 representative watermarking schemes against six of the existing attacks and demonstrate that each of these watermarking schemes lacks robustness against at least two non-adaptive attacks. We also propose novel adaptive attacks that harness the adversary's knowledge of the underlying watermarking algorithm of a target model. We demonstrate that the proposed attacks effectively break all of the 11 watermarking schemes, consequently allowing adversaries to obscure the ownership of any watermarked model. We encourage follow-up studies to consider our guidelines when evaluating the robustness of their watermarking schemes via conducting comprehensive adversarial evaluation that includes our adaptive attacks to demonstrate a meaningful upper bound of watermark robustness."
                    },
                    {
                        "title": "LucidDreamer: Domain-free Generation of 3D Gaussian Splatting Scenes",
                        "abstract": "With the widespread usage of VR devices and contents, demands for 3D scene generation techniques become more popular. Existing 3D scene generation models, however, limit the target scene to specific domain, primarily due to their training strategies using 3D scan dataset that is far from the real-world. To address such limitation, we propose LucidDreamer, a domain-free scene generation pipeline by fully leveraging the power of existing large-scale diffusion-based generative model. Our LucidDreamer has two alternate steps: Dreaming and Alignment. First, to generate multi-view consistent images from inputs, we set the point cloud as a geometrical guideline for each image generation. Specifically, we project a portion of point cloud to the desired view and provide the projection as a guidance for inpainting using the generative model. The inpainted images are lifted to 3D space with estimated depth maps, composing a new points. Second, to aggregate the new points into the 3D scene, we propose an aligning algorithm which harmoniously integrates the portions of newly generated 3D scenes. The finally obtained 3D scene serves as initial points for optimizing Gaussian splats. LucidDreamer produces Gaussian splats that are highly-detailed compared to the previous 3D scene generation methods, with no constraint on domain of the target scene. Project page: https://luciddreamer-cvlab.github.io/"
                    },
                    {
                        "title": "NTIRE 2021 Challenge on Video Super-Resolution",
                        "abstract": "Super-Resolution (SR) is a fundamental computer vision task that aims to obtain a high-resolution clean image from the given low-resolution counterpart. This paper reviews the NTIRE 2021 Challenge on Video Super-Resolution. We present evaluation results from two competition tracks as well as the proposed solutions. Track 1 aims to develop conventional video SR methods focusing on the restoration quality. Track 2 assumes a more challenging environment with lower frame rates, casting spatio-temporal SR problem. In each competition, 247 and 223 participants have registered, respectively. During the final testing phase, 14 teams competed in each track to achieve state-of-the-art performance on video SR tasks."
                    },
                    {
                        "title": "NTIRE 2021 Challenge on Image Deblurring",
                        "abstract": "Motion blur is a common photography artifact in dynamic environments that typically comes jointly with the other types of degradation. This paper reviews the NTIRE 2021 Challenge on Image Deblurring. In this challenge report, we describe the challenge specifics and the evaluation results from the 2 competition tracks with the proposed solutions. While both the tracks aim to recover a high-quality clean image from a blurry image, different artifacts are jointly involved. In track 1, the blurry images are in a low resolution while track 2 images are compressed in JPEG format. In each competition, there were 338 and 238 registered participants and in the final testing phase, 18 and 17 teams competed. The winning methods demonstrate the state-of-the-art performance on the image deblurring task with the jointly combined artifacts."
                    }
                ]
            },
            "e489427b-f619-46aa-b9f5-f2e06c2a0c83": {
                "pk": "e489427b-f619-46aa-b9f5-f2e06c2a0c83",
                "name": "Woojun Kim",
                "collaborators": [
                    "Youngchul Sung",
                    "Whiyoung Jung",
                    "Myungsik Cho",
                    "Jeonghye Kim",
                    "Sungho Choi",
                    "Seungyul Han",
                    "Katia Sycara",
                    "Suyoung Lee",
                    "Jeewon Jeon",
                    "Yongjae Shin"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Multi-Agent Systems",
                    "Robotics",
                    "Imitation Learning"
                ],
                "publications": [
                    {
                        "title": "Parameter Sharing with Network Pruning for Scalable Multi-Agent Deep Reinforcement Learning",
                        "abstract": "Handling the problem of scalability is one of the essential issues for multi-agent reinforcement learning (MARL) algorithms to be applied to real-world problems typically involving massively many agents. For this, parameter sharing across multiple agents has widely been used since it reduces the training time by decreasing the number of parameters and increasing the sample efficiency. However, using the same parameters across agents limits the representational capacity of the joint policy and consequently, the performance can be degraded in multi-agent tasks that require different behaviors for different agents. In this paper, we propose a simple method that adopts structured pruning for a deep neural network to increase the representational capacity of the joint policy without introducing additional parameters. We evaluate the proposed method on several benchmark tasks, and numerical results show that the proposed method significantly outperforms other parameter-sharing methods."
                    },
                    {
                        "title": "LESSON: Learning to Integrate Exploration Strategies for Reinforcement Learning via an Option Framework",
                        "abstract": "In this paper, a unified framework for exploration in reinforcement learning (RL) is proposed based on an option-critic model. The proposed framework learns to integrate a set of diverse exploration strategies so that the agent can adaptively select the most effective exploration strategy over time to realize a relevant exploration-exploitation trade-off for each given task. The effectiveness of the proposed exploration framework is demonstrated by various experiments in the MiniGrid and Atari environments."
                    },
                    {
                        "title": "Decision ConvFormer: Local Filtering in MetaFormer is Sufficient for Decision Making",
                        "abstract": "The recent success of Transformer in natural language processing has sparked its use in various domains. In offline reinforcement learning (RL), Decision Transformer (DT) is emerging as a promising model based on Transformer. However, we discovered that the attention module of DT is not appropriate to capture the inherent local dependence pattern in trajectories of RL modeled as a Markov decision process. To overcome the limitations of DT, we propose a novel action sequence predictor, named Decision ConvFormer (DC), based on the architecture of MetaFormer, which is a general structure to process multiple entities in parallel and understand the interrelationship among the multiple entities. DC employs local convolution filtering as the token mixer and can effectively capture the inherent local associations of the RL dataset. In extensive experiments, DC achieved state-of-the-art performance across various standard RL benchmarks while requiring fewer resources. Furthermore, we show that DC better understands the underlying meaning in data and exhibits enhanced generalization capability."
                    },
                    {
                        "title": "A Maximum Mutual Information Framework for Multi-Agent Reinforcement Learning",
                        "abstract": "In this paper, we propose a maximum mutual information (MMI) framework for multi-agent reinforcement learning (MARL) to enable multiple agents to learn coordinated behaviors by regularizing the accumulated return with the mutual information between actions. By introducing a latent variable to induce nonzero mutual information between actions and applying a variational bound, we derive a tractable lower bound on the considered MMI-regularized objective function. Applying policy iteration to maximize the derived lower bound, we propose a practical algorithm named variational maximum mutual information multi-agent actor-critic (VM3-AC), which follows centralized learning with decentralized execution (CTDE). We evaluated VM3-AC for several games requiring coordination, and numerical results show that VM3-AC outperforms MADDPG and other MARL algorithms in multi-agent tasks requiring coordination."
                    },
                    {
                        "title": "A Variational Approach to Mutual Information-Based Coordination for Multi-Agent Reinforcement Learning",
                        "abstract": "In this paper, we propose a new mutual information framework for multi-agent reinforcement learning to enable multiple agents to learn coordinated behaviors by regularizing the accumulated return with the simultaneous mutual information between multi-agent actions. By introducing a latent variable to induce nonzero mutual information between multi-agent actions and applying a variational bound, we derive a tractable lower bound on the considered MMI-regularized objective function. The derived tractable objective can be interpreted as maximum entropy reinforcement learning combined with uncertainty reduction of other agents actions. Applying policy iteration to maximize the derived lower bound, we propose a practical algorithm named variational maximum mutual information multi-agent actor-critic, which follows centralized learning with decentralized execution. We evaluated VM3-AC for several games requiring coordination, and numerical results show that VM3-AC outperforms other MARL algorithms in multi-agent tasks requiring high-quality coordination."
                    },
                    {
                        "title": "Message-Dropout: An Efficient Training Method for Multi-Agent Deep Reinforcement Learning",
                        "abstract": "In this paper, we propose a new learning technique named message-dropout to improve the performance for multi-agent deep reinforcement learning under two application scenarios: 1) classical multi-agent reinforcement learning with direct message communication among agents and 2) centralized training with decentralized execution. In the first application scenario of multi-agent systems in which direct message communication among agents is allowed, the message-dropout technique drops out the received messages from other agents in a block-wise manner with a certain probability in the training phase and compensates for this effect by multiplying the weights of the dropped-out block units with a correction probability. The applied message-dropout technique effectively handles the increased input dimension in multi-agent reinforcement learning with communication and makes learning robust against communication errors in the execution phase. In the second application scenario of centralized training with decentralized execution, we particularly consider the application of the proposed message-dropout to Multi-Agent Deep Deterministic Policy Gradient (MADDPG), which uses a centralized critic to train a decentralized actor for each agent. We evaluate the proposed message-dropout technique for several games, and numerical results show that the proposed message-dropout technique with proper dropout rate improves the reinforcement learning performance significantly in terms of the training speed and the steady-state performance in the execution phase."
                    },
                    {
                        "title": "Cross-Domain Imitation Learning with a Dual Structure",
                        "abstract": "In this paper, we consider cross-domain imitation learning (CDIL) in which an agent in a target domain learns a policy to perform well in the target domain by observing expert demonstrations in a source domain without accessing any reward function. In order to overcome the domain difference for imitation learning, we propose a dual-structured learning method. The proposed learning method extracts two feature vectors from each input observation such that one vector contains domain information and the other vector contains policy expertness information, and then enhances feature vectors by synthesizing new feature vectors containing both target-domain and policy expertness information. The proposed CDIL method is tested on several MuJoCo tasks where the domain difference is determined by image angles or colors. Numerical results show that the proposed method shows superior performance in CDIL to other existing algorithms and achieves almost the same performance as imitation learning without domain difference."
                    },
                    {
                        "title": "MASER: Multi-Agent Reinforcement Learning with Subgoals Generated from Experience Replay Buffer",
                        "abstract": "In this paper, we consider cooperative multi-agent reinforcement learning (MARL) with sparse reward. To tackle this problem, we propose a novel method named MASER: MARL with subgoals generated from experience replay buffer. Under the widely-used assumption of centralized training with decentralized execution and consistent Q-value decomposition for MARL, MASER automatically generates proper subgoals for multiple agents from the experience replay buffer by considering both individual Q-value and total Q-value. Then, MASER designs individual intrinsic reward for each agent based on actionable representation relevant to Q-learning so that the agents reach their subgoals while maximizing the joint action value. Numerical results show that MASER significantly outperforms StarCraft II micromanagement benchmark compared to other state-of-the-art MARL algorithms."
                    },
                    {
                        "title": "Sample-Efficient and Safe Deep Reinforcement Learning via Reset Deep Ensemble Agents",
                        "abstract": "Deep reinforcement learning (RL) has achieved remarkable success in solving complex tasks through its integration with deep neural networks (DNNs) as function approximators. However, the reliance on DNNs has introduced a new challenge called primacy bias, whereby these function approximators tend to prioritize early experiences, leading to overfitting. To mitigate this primacy bias, a reset method has been proposed, which performs periodic resets of a portion or the entirety of a deep RL agent while preserving the replay buffer. However, the use of the reset method can result in performance collapses after executing the reset, which can be detrimental from the perspective of safe RL and regret minimization. In this paper, we propose a new reset-based method that leverages deep ensemble learning to address the limitations of the vanilla reset method and enhance sample efficiency. The proposed method is evaluated through various experiments including those in the domain of safe RL. Numerical results show its effectiveness in high sample efficiency and safety considerations."
                    },
                    {
                        "title": "DyPNIPP: Predicting Environment Dynamics for RL-based Robust Informative Path Planning",
                        "abstract": "Informative path planning (IPP) is an important planning paradigm for various real-world robotic applications such as environment monitoring. IPP involves planning a path that can learn an accurate belief of the quantity of interest, while adhering to planning constraints. Traditional IPP methods typically require high computation time during execution, giving rise to reinforcement learning (RL) based IPP methods. However, the existing RL-based methods do not consider spatio-temporal environments which involve their own challenges due to variations in environment characteristics. In this paper, we propose DyPNIPP, a robust RL-based IPP framework, designed to operate effectively across spatio-temporal environments with varying dynamics. To achieve this, DyPNIPP incorporates domain randomization to train the agent across diverse environments and introduces a dynamics prediction model to capture and adapt the agent actions to specific environment dynamics. Our extensive experiments in a wildfire environment demonstrate that DyPNIPP outperforms existing RL-based IPP algorithms by significantly improving robustness and performing across diverse environment conditions."
                    },
                    {
                        "title": "Escaping Local Minima: Hybrid Artificial Potential Field with Wall-Follower for Decentralized Multi-Robot Navigation",
                        "abstract": "We tackle the challenges of decentralized multi-robot navigation in environments with nonconvex obstacles, where complete environmental knowledge is unavailable. While reactive methods like Artificial Potential Field (APF) offer simplicity and efficiency, they suffer from local minima, causing robots to become trapped due to their lack of global environmental awareness. Other existing solutions either rely on inter-robot communication, are limited to single-robot scenarios, or struggle to overcome nonconvex obstacles effectively.   Our proposed methods enable collision-free navigation using only local sensor and state information without a map. By incorporating a wall-following (WF) behavior into the APF approach, our method allows robots to escape local minima, even in the presence of nonconvex and dynamic obstacles including other robots. We introduce two algorithms for switching between APF and WF: a rule-based system and an encoder network trained on expert demonstrations. Experimental results show that our approach achieves substantially higher success rates compared to state-of-the-art methods, highlighting its ability to overcome the limitations of local minima in complex environments"
                    },
                    {
                        "title": "Domain Adaptive Imitation Learning with Visual Observation",
                        "abstract": "In this paper, we consider domain-adaptive imitation learning with visual observation, where an agent in a target domain learns to perform a task by observing expert demonstrations in a source domain. Domain adaptive imitation learning arises in practical scenarios where a robot, receiving visual sensory data, needs to mimic movements by visually observing other robots from different angles or observing robots of different shapes. To overcome the domain shift in cross-domain imitation learning with visual observation, we propose a novel framework for extracting domain-independent behavioral features from input observations that can be used to train the learner, based on dual feature extraction and image reconstruction. Empirical results demonstrate that our approach outperforms previous algorithms for imitation learning from visual observation with domain shift."
                    },
                    {
                        "title": "ShapeGrasp: Zero-Shot Task-Oriented Grasping with Large Language Models through Geometric Decomposition",
                        "abstract": "Task-oriented grasping of unfamiliar objects is a necessary skill for robots in dynamic in-home environments. Inspired by the human capability to grasp such objects through intuition about their shape and structure, we present a novel zero-shot task-oriented grasping method leveraging a geometric decomposition of the target object into simple, convex shapes that we represent in a graph structure, including geometric attributes and spatial relationships. Our approach employs minimal essential information - the object's name and the intended task - to facilitate zero-shot task-oriented grasping. We utilize the commonsense reasoning capabilities of large language models to dynamically assign semantic meaning to each decomposed part and subsequently reason over the utility of each part for the intended task. Through extensive experiments on a real-world robotics platform, we demonstrate that our grasping approach's decomposition and reasoning pipeline is capable of selecting the correct part in 92% of the cases and successfully grasping the object in 82% of the tasks we evaluate. Additional videos, experiments, code, and data are available on our project website: https://shapegrasp.github.io/."
                    }
                ]
            },
            "608a9c59-dbf8-4dee-90fb-19c76d410a87": {
                "pk": "608a9c59-dbf8-4dee-90fb-19c76d410a87",
                "name": "Youngchul Sung",
                "collaborators": [
                    "H. Vincent Poor",
                    "Seungyul Han",
                    "Lang Tong",
                    "Gilwon Lee",
                    "Cheulsoon Kim",
                    "Junyeong Seo",
                    "Juho Park",
                    "Heejung Yu",
                    "Jungho So",
                    "Woojun Kim"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Wireless Communication",
                    "Beamforming",
                    "Multi-Agent Systems"
                ],
                "publications": [
                    {
                        "title": "The capacity for the linear time-invariant Gaussian relay channel",
                        "abstract": "In this paper, the Gaussian relay channel with linear time-invariant relay filtering is considered. Based on spectral theory for stationary processes, the maximum achievable rate for this subclass of linear Gaussian relay operation is obtained in finite-letter characterization. The maximum rate can be achieved by dividing the overall frequency band into at most eight subbands and by making the relay behave as an instantaneous amplify-and-forward relay at each subband. Numerical results are provided to evaluate the performance of LTI relaying."
                    },
                    {
                        "title": "Diversity Actor-Critic: Sample-Aware Entropy Regularization for Sample-Efficient Exploration",
                        "abstract": "In this paper, sample-aware policy entropy regularization is proposed to enhance the conventional policy entropy regularization for better exploration. Exploiting the sample distribution obtainable from the replay buffer, the proposed sample-aware entropy regularization maximizes the entropy of the weighted sum of the policy action distribution and the sample action distribution from the replay buffer for sample-efficient exploration. A practical algorithm named diversity actor-critic (DAC) is developed by applying policy iteration to the objective function with the proposed sample-aware entropy regularization. Numerical results show that DAC significantly outperforms existing recent algorithms for reinforcement learning."
                    },
                    {
                        "title": "A New Approach to User Scheduling in Massive Multi-User MIMO Broadcast Channels",
                        "abstract": "In this paper, a new user-scheduling-and-beamforming method is proposed for multi-user massive multiple-input multiple-output (massive MIMO) broadcast channels in the context of two-stage beamforming. The key ideas of the proposed scheduling method are 1) to use a set of orthogonal reference beams and construct a double cone around each reference beam to select `nearly-optimal' semi-orthogonal users based only on channel quality indicator (CQI) feedback and 2) to apply post-user-selection beam refinement with zero-forcing beamforming (ZFBF) based on channel state information (CSI) feedback only from the selected users. It is proved that the proposed scheduling-and-beamforming method is asymptotically optimal as the number of users increases. Furthermore, the proposed scheduling-and-beamforming method almost achieves the performance of the existing semi-orthogonal user selection with ZFBF (SUS-ZFBF) that requires full CSI feedback from all users, with significantly reduced feedback overhead which is even less than that required by random beamforming."
                    },
                    {
                        "title": "Enhancing Non-Orthogonal Multiple Access By Forming Relaying Broadcast Channels",
                        "abstract": "In this paper, using relaying broadcast channels (RBCs) as component channels for non-orthogonal multiple access (NOMA) is proposed to enhance the performance of NOMA in single-input single-output (SISO) cellular downlink systems. To analyze the performance of the proposed scheme, an achievable rate region of a RBC with compress-and-forward (CF) relaying is newly derived based on the recent work of noisy network coding (NNC). Based on the analysis of the achievable rate region of a RBC with decode-and-forward (DF) relaying, CF relaying, or CF relaying with dirty-paper coding (DPC) at the transmitter, the overall system performance of NOMA equipped with RBC component channels is investigated. It is shown that NOMA with RBC-DF yields marginal gain and NOMA with RBC-CF/DPC yields drastic gain over the simple NOMA based on broadcast component channels in a practical system setup. By going beyond simple broadcast channel (BC)/successive interference cancellation (SIC) to advanced multi-terminal encoding including DPC and CF/NNC, far larger gains can be obtained for NOMA."
                    },
                    {
                        "title": "Beam Design and User Scheduling for Non-Orthogonal Multiple Access with Multiple Antennas Based on Pareto-Optimality",
                        "abstract": "In this paper, an efficient transmit beam design and user scheduling method is proposed for multi-user (MU) multiple-input single-output (MISO) non-orthogonal multiple access (NOMA) downlink, based on Pareto-optimality. The proposed beam design and user scheduling method groups simultaneously-served users into multiple clusters with practical two users in each cluster, and then applies spatical zeroforcing (ZF) across clusters to control inter-cluster interference (ICI) and Pareto-optimal beam design with successive interference cancellation (SIC) to two users in each cluster to remove interference to strong users and leverage signal-to-interference-plus-noise ratios (SINRs) of interference-experiencing weak users. The proposed method has flexibility to control the rates of strong and weak users and numerical results show that the proposed method yields good performance."
                    },
                    {
                        "title": "AMBER: Adaptive Multi-Batch Experience Replay for Continuous Action Control",
                        "abstract": "In this paper, a new adaptive multi-batch experience replay scheme is proposed for proximal policy optimization (PPO) for continuous action control. On the contrary to original PPO, the proposed scheme uses the batch samples of past policies as well as the current policy for the update for the next policy, where the number of the used past batches is adaptively determined based on the oldness of the past batches measured by the average importance sampling (IS) weight. The new algorithm constructed by combining PPO with the proposed multi-batch experience replay scheme maintains the advantages of original PPO such as random mini-batch sampling and small bias due to low IS weights by storing the pre-computed advantages and values and adaptively determining the mini-batch size. Numerical results show that the proposed method significantly increases the speed and stability of convergence on various continuous control tasks compared to original PPO."
                    },
                    {
                        "title": "Dimension-Wise Importance Sampling Weight Clipping for Sample-Efficient Reinforcement Learning",
                        "abstract": "In importance sampling (IS)-based reinforcement learning algorithms such as Proximal Policy Optimization (PPO), IS weights are typically clipped to avoid large variance in learning. However, policy update from clipped statistics induces large bias in tasks with high action dimensions, and bias from clipping makes it difficult to reuse old samples with large IS weights. In this paper, we consider PPO, a representative on-policy algorithm, and propose its improvement by dimension-wise IS weight clipping which separately clips the IS weight of each action dimension to avoid large bias and adaptively controls the IS weight to bound policy update from the current policy. This new technique enables efficient learning for high action-dimensional tasks and reusing of old samples like in off-policy learning to increase the sample efficiency. Numerical results show that the proposed new algorithm outperforms PPO and other RL algorithms in various Open AI Gym tasks."
                    },
                    {
                        "title": "A Max-Min Entropy Framework for Reinforcement Learning",
                        "abstract": "In this paper, we propose a max-min entropy framework for reinforcement learning (RL) to overcome the limitation of the soft actor-critic (SAC) algorithm implementing the maximum entropy RL in model-free sample-based learning. Whereas the maximum entropy RL guides learning for policies to reach states with high entropy in the future, the proposed max-min entropy framework aims to learn to visit states with low entropy and maximize the entropy of these low-entropy states to promote better exploration. For general Markov decision processes (MDPs), an efficient algorithm is constructed under the proposed max-min entropy framework based on disentanglement of exploration and exploitation. Numerical results show that the proposed algorithm yields drastic performance improvement over the current state-of-the-art RL algorithms."
                    },
                    {
                        "title": "On the Pareto-Optimal Beam Structure and Design for Multi-User MIMO Interference Channels",
                        "abstract": "In this paper, the Pareto-optimal beam structure for multi-user multiple-input multiple-output (MIMO) interference channels is investigated and a necessary condition for any Pareto-optimal transmit signal covariance matrix is presented for the K-pair Gaussian (N,M_1,...,M_K) interference channel. It is shown that any Pareto-optimal transmit signal covariance matrix at a transmitter should have its column space contained in the union of the eigen-spaces of the channel matrices from the transmitter to all receivers. Based on this necessary condition, an efficient parameterization for the beam search space is proposed. The proposed parameterization is given by the product manifold of a Stiefel manifold and a subset of a hyperplane and enables us to construct a very efficient beam design algorithm by exploiting its rich geometrical structure and existing tools for optimization on Stiefel manifolds. Reduction in the beam search space dimension and computational complexity by the proposed parameterization and the proposed beam design approach is significant when the number of transmit antennas is larger than the sum of the numbers of receive antennas, as in upcoming cellular networks adopting massive MIMO technologies. Numerical results validate the proposed parameterization and the proposed cooperative beam design method based on the parameterization for MIMO interference channels."
                    },
                    {
                        "title": "Parameter Sharing with Network Pruning for Scalable Multi-Agent Deep Reinforcement Learning",
                        "abstract": "Handling the problem of scalability is one of the essential issues for multi-agent reinforcement learning (MARL) algorithms to be applied to real-world problems typically involving massively many agents. For this, parameter sharing across multiple agents has widely been used since it reduces the training time by decreasing the number of parameters and increasing the sample efficiency. However, using the same parameters across agents limits the representational capacity of the joint policy and consequently, the performance can be degraded in multi-agent tasks that require different behaviors for different agents. In this paper, we propose a simple method that adopts structured pruning for a deep neural network to increase the representational capacity of the joint policy without introducing additional parameters. We evaluate the proposed method on several benchmark tasks, and numerical results show that the proposed method significantly outperforms other parameter-sharing methods."
                    },
                    {
                        "title": "Sensor Configuration and Activation for Field Detection in Large Sensor Arrays",
                        "abstract": "The problems of sensor configuration and activation for the detection of correlated random fields using large sensor arrays are considered. Using results that characterize the large-array performance of sensor networks in this application, the detection capabilities of different sensor configurations are analyzed and compared. The dependence of the optimal choice of configuration on parameters such as sensor signal-to-noise ratio (SNR), field correlation, etc., is examined, yielding insights into the most effective choices for sensor selection and activation in various operating regimes."
                    },
                    {
                        "title": "Information, Energy and Density for Ad Hoc Sensor Networks over Correlated Random Fields: Large Deviations Analysis",
                        "abstract": "Using large deviations results that characterize the amount of information per node on a two-dimensional (2-D) lattice, asymptotic behavior of a sensor network deployed over a correlated random field for statistical inference is investigated. Under a 2-D hidden Gauss-Markov random field model with symmetric first order conditional autoregression, the behavior of the total information [nats] and energy efficiency [nats/J] defined as the ratio of total gathered information to the required energy is obtained as the coverage area, node density and energy vary."
                    },
                    {
                        "title": "Optimal Node Density for Two-Dimensional Sensor Arrays",
                        "abstract": "The problem of optimal node density for ad hoc sensor networks deployed for making inferences about two dimensional correlated random fields is considered. Using a symmetric first order conditional autoregressive Gauss-Markov random field model, large deviations results are used to characterize the asymptotic per-node information gained from the array. This result then allows an analysis of the node density that maximizes the information under an energy constraint, yielding insights into the trade-offs among the information, density and energy."
                    },
                    {
                        "title": "A Large Deviations Approach to Sensor Scheduling for Detection of Correlated Random Fields",
                        "abstract": "The problem of scheduling sensor transmissions for the detection of correlated random fields using spatially deployed sensors is considered. Using the large deviations principle, a closed-form expression for the error exponent of the miss probability is given as a function of the sensor spacing and signal-to-noise ratio (SNR). It is shown that the error exponent has a distinct characteristic: at high SNR, the error exponent is monotonically increasing with respect to sensor spacing, while at low SNR there is an optimal spacing for scheduled sensors."
                    },
                    {
                        "title": "Neyman-Pearson Detection of Gauss-Markov Signals in Noise: Closed-Form Error Exponent and Properties",
                        "abstract": "The performance of Neyman-Pearson detection of correlated stochastic signals using noisy observations is investigated via the error exponent for the miss probability with a fixed level. Using the state-space structure of the signal and observation model, a closed-form expression for the error exponent is derived, and the connection between the asymptotic behavior of the optimal detector and that of the Kalman filter is established. The properties of the error exponent are investigated for the scalar case. It is shown that the error exponent has distinct characteristics with respect to correlation strength: for signal-to-noise ratio (SNR) >1 the error exponent decreases monotonically as the correlation becomes stronger, whereas for SNR <1 there is an optimal correlation that maximizes the error exponent for a given SNR."
                    },
                    {
                        "title": "Optimal and Suboptimal Detection of Gaussian Signals in Noise: Asymptotic Relative Efficiency",
                        "abstract": "The performance of Bayesian detection of Gaussian signals using noisy observations is investigated via the error exponent for the average error probability. Under unknown signal correlation structure or limited processing capability it is reasonable to use the simple quadratic detector that is optimal in the case of an independent and identically distributed (i.i.d.) signal. Using the large deviations principle, the performance of this detector (which is suboptimal for non-i.i.d. signals) is compared with that of the optimal detector for correlated signals via the asymptotic relative efficiency defined as the ratio between sample sizes of two detectors required for the same performance in the large-sample-size regime. The effects of SNR on the ARE are investigated. It is shown that the asymptotic efficiency of the simple quadratic detector relative to the optimal detector converges to one as the SNR increases without bound for any bounded spectrum, and that the simple quadratic detector performs as well as the optimal detector for a wide range of the correlation values at high SNR."
                    },
                    {
                        "title": "On Beamformer Design for Multiuser MIMO Interference Channels",
                        "abstract": "This paper considers several linear beamformer design paradigms for multiuser time-invariant multiple-input multiple-output interference channels. Notably, interference alignment and sum-rate based algorithms such as the maximum signal-to-interference-plus noise (max-SINR) algorithm are considered. Optimal linear beamforming under interference alignment consists of two layers; an inner precoder and decoder (or receive filter) accomplish interference alignment to eliminate inter-user interference, and an outer precoder and decoder diagonalize the effective single-user channel resulting from the interference alignment by the inner precoder and decoder. The relationship between this two-layer beamforming and the max-SINR algorithm is established at high signal-to-noise ratio. Also, the optimality of the max-SINR algorithm within the class of linear beamforming algorithms, and its local convergence with exponential rate, are established at high signal-to-noise ratio."
                    },
                    {
                        "title": "A joint time-invariant filtering approach to the linear Gaussian relay problem",
                        "abstract": "In this paper, the linear Gaussian relay problem is considered. Under the linear time-invariant (LTI) model the problem is formulated in the frequency domain based on the Toeplitz distribution theorem. Under the further assumption of realizable input spectra, the LTI Gaussian relay problem is converted to a joint design problem of source and relay filters under two power constraints, one at the source and the other at the relay, and a practical solution to this problem is proposed based on the projected subgradient method. Numerical results show that the proposed method yields a noticeable gain over the instantaneous amplify-and-forward (AF) scheme in inter-symbol interference (ISI) channels. Also, the optimality of the AF scheme within the class of one-tap relay filters is established in flat-fading channels."
                    },
                    {
                        "title": "Two-Stage Beamformer Design for Massive MIMO Downlink By Trace Quotient Formulation",
                        "abstract": "In this paper, the problem of outer beamformer design based only on channel statistic information is considered for two-stage beamforming for multi-user massive MIMO downlink, and the problem is approached based on signal-to-leakage-plus-noise ratio (SLNR). To eliminate the dependence on the instantaneous channel state information, a lower bound on the average SLNR is derived by assuming zero-forcing (ZF) inner beamforming, and an outer beamformer design method that maximizes the lower bound on the average SLNR is proposed. It is shown that the proposed SLNR-based outer beamformer design problem reduces to a trace quotient problem (TQP), which is often encountered in the field of machine learning. An iterative algorithm is presented to obtain an optimal solution to the proposed TQP. The proposed method has the capability of optimally controlling the weighting factor between the signal power to the desired user and the interference leakage power to undesired users according to different channel statistics. Numerical results show that the proposed outer beamformer design method yields significant performance gain over existing methods."
                    },
                    {
                        "title": "Randomly-Directional Beamforming in Millimeter-Wave Multi-User MISO Downlink",
                        "abstract": "In this paper, randomly-directional beamforming (RDB) is considered for millimeter-wave (mmwave) multi-user (MU) multiple-input single-output (MISO) downlink systems. By using asymptotic techniques, the performance of RDB and the MU gain in mm-wave MISO are analyzed based on the uniform random line-of-sight (UR-LoS) channel model suitable for highly directional mm-wave radio propagation channels. It is shown that there exists a transition point on the number of users relative to the number of antenna elements for non-trivial performance of the RDB scheme, and furthermore sum rate scaling arbitrarily close to linear scaling with respect to the number of antenna elements can be achieved under the UR-LoS channel model by opportunistic random beamforming with proper user scheduling if the number of users increases linearly with respect to the number of antenna elements. The provided results yield insights into the most effective beamforming and scheduling choices for mm-wave MU-MISO in various operating conditions. Simulation results validate our analysis based on asymptotic techniques for finite cases."
                    }
                ]
            }
        }
    },
    "2409.08302": {
        "paper_data": {
            "title": "How Molecules Impact Cells: Unlocking Contrastive PhenoMolecular Retrieval",
            "url": "http://arxiv.org/abs/2409.08302v1",
            "arxiv_id": "2409.08302",
            "authors": [
                "Philip Fradkin",
                "Puria Azadi",
                "Karush Suri",
                "Frederik Wenkel",
                "Ali Bashashati",
                "Maciej Sypetkowski",
                "Dominique Beaini"
            ],
            "abstract": "Predicting molecular impact on cellular function is a core challenge in therapeutic design. Phenomic experiments, designed to capture cellular morphology, utilize microscopy based techniques and demonstrate a high throughput solution for uncovering molecular impact on the cell. In this work, we learn a joint latent space between molecular structures and microscopy phenomic experiments, aligning paired samples with contrastive learning. Specifically, we study the problem ofContrastive PhenoMolecular Retrieval, which consists of zero-shot molecular structure identification conditioned on phenomic experiments. We assess challenges in multi-modal learning of phenomics and molecular modalities such as experimental batch effect, inactive molecule perturbations, and encoding perturbation concentration. We demonstrate improved multi-modal learner retrieval through (1) a uni-modal pre-trained phenomics model, (2) a novel inter sample similarity aware loss, and (3) models conditioned on a representation of molecular concentration. Following this recipe, we propose MolPhenix, a molecular phenomics model. MolPhenix leverages a pre-trained phenomics model to demonstrate significant performance gains across perturbation concentrations, molecular scaffolds, and activity thresholds. In particular, we demonstrate an 8.1x improvement in zero shot molecular retrieval of active molecules over the previous state-of-the-art, reaching 77.33% in top-1% accuracy. These results open the door for machine learning to be applied in virtual phenomics screening, which can significantly benefit drug discovery applications.",
            "introduction": "   1 Introduction  Quantifying cellular responses elicited by genetic and molecular perturbations represents a core challenge in medicinal research Bock et\u00a0al. (2022); Vincent et\u00a0al. (2022). Out of an approximate 1060superscript106010^{60}10 start_POSTSUPERSCRIPT 60 end_POSTSUPERSCRIPT druglike molecule designs, a small number are able to alter cellular properties to reverse the course of diseases Bohacek et\u00a0al. (1996); Knox et\u00a0al. (2023). In recent years, microscopy-based cell morphology screening techniques, demonstrated potential for quantitative understanding of a molecule\u2019s biological effects. Experimental techniques such as cell-painting are used to capture cellular morphology, which correspond to physical and structural properties of the cell Boutros et\u00a0al. (2015); Bray et\u00a0al. (2016). Cells treated with molecular perturbations can change morphology, which is captured by staining and high throughput microscopy techniques. Perturbations with similar cellular impact induce analogous morphological changes, allowing to capture underlying biological effects in phenomic experiments. Identifying such perturbations with similar morphological changes can aid in discovery of novel therapeutic drug candidates Simm et\u00a0al. (2018); Kraus et\u00a0al. (2017); Hofmarcher et\u00a0al. (2019).   Determining molecular impact on the cell can be formulated as a multi-modal learning problem, allowing us to build on a rich family of methods Radford et\u00a0al. (2021); Zhai et\u00a0al. (2023); Srinivasa et\u00a0al. (2023). Similar to text-image models, paired data is collected from phenomic experiments along with molecules used to perturb the cells. Contrastive objectives have been used as an effective approach in aligning paired samples from different modalities Radford et\u00a0al. (2021); Lanusse et\u00a0al. (2023). A model that has learned a cross-modal joint latent space must be able to retrieve a molecular perturbant conditioned on the phenomic experiment. We identify this problem as contrastive phenomolecular retrieval (see Figure 2). Addressing this problem can allow for identification of molecular impact on cellular function, however, this comes with its own set of challenges. F\u00fcrst et\u00a0al. (2022); Albelwi (2022); Zhong et\u00a0al. (2022).   (1) Firstly, multi-modal paired phenomics molecular data suffers from lower overall dataset sizes and is subject to batch effects. Challenges with uniform processing and prohibitive costs associated with acquisition of paired data, leads to an order of magnitude fewer data points compared to text-image datasets Schuhmann et\u00a0al. (2022); Chandrasekaran et\u00a0al. (2023). Furthermore, data is subject to random batch effects that capture non-biologically meaningful variation Leek et\u00a0al. (2010); Sypetkowski et\u00a0al. (2023). (2) Paired phenomic-molecular data contains inactive perturbations that do not have a biological effect or do not perturb cellular morphology. It is difficult to infer a priori whether a molecule has a cellular effect, leading to the collection of paired molecular structures with unperturbed cells. These data-points are challenging to filter out without an effective phenomic embedding, as morphological effects are rarely discernible. These samples can be interpreted as misannotated, under the assumption of all collected pairs having biologically meaningful interactions. (3) Finally, a complete solution for capturing molecular effects on cells must capture molecular concentration. The same molecule can have drastically different effects along its dose response curve, thus making concentration an essential component for learning molecular impact.   In this work, we explore the problem of contrastive phenomolecular retrieval by addressing the above challenges circumvented in prior works. Our key contributions are as follows:     \u2022  We demonstrate significantly higher phenomolecular retrieval rates by utilizing a",
            "references": [
                {
                    "title": "CWCL: Cross-Modal Transfer with Continuously Weighted Contrastive Loss",
                    "abstract": "This paper considers contrastive training for cross-modal 0-shot transfer wherein a pre-trained model in one modality is used for representation learning in another domain using pairwise data. The learnt models in the latter domain can then be used for a diverse set of tasks in a zero-shot way, similar to ``Contrastive Language-Image Pre-training (CLIP)'' and ``Locked-image Tuning (LiT)'' that have recently gained considerable attention. Most existing works for cross-modal representation alignment (including CLIP and LiT) use the standard contrastive training objective, which employs sets of positive and negative examples to align similar and repel dissimilar training data samples. However, similarity amongst training examples has a more continuous nature, thus calling for a more `non-binary' treatment. To address this, we propose a novel loss function called Continuously Weighted Contrastive Loss (CWCL) that employs a continuous measure of similarity. With CWCL, we seek to align the embedding space of one modality with another. Owing to the continuous nature of similarity in the proposed loss function, these models outperform existing methods for 0-shot transfer across multiple models, datasets and modalities. Particularly, we consider the modality pairs of image-text and speech-text and our models achieve 5-8% (absolute) improvement over previous state-of-the-art methods in 0-shot image classification and 20-30% (absolute) improvement in 0-shot speech-to-intent classification and keyword classification."
                },
                {
                    "title": "CLOOME: contrastive learning unlocks bioimaging databases for queries with chemical structures",
                    "abstract": "Currently, bioimaging databases cannot be queried by chemical structures that induce the phenotypic effects captured by an image. Through the advent of the contrastive learning paradigm, images and text could be embedded into the same space. We build on this contrastive learning paradigm, to present a novel retrieval system that is able to identify the correct bioimage given a chemical structure out of a database of \u223c2,000 candidate images with a top-1 accuracy >70 times higher than a random baseline. Additionally, the learned embeddings of our method are highly transferable to various relevant downstream tasks in drug discovery, including activity prediction, microscopy image classification and mechanism of action identification."
                },
                {
                    "title": "MAESTER: Masked Autoencoder Guided Segmentation at Pixel Resolution for Accurate, Self-Supervised Subcellular Structure Recognition",
                    "abstract": "Accurate segmentation of cellular images remains an elusive task due to the intrinsic variability in morphology of biological structures. Complete manual segmentation is unfeasible for large datasets, and while supervised methods have been proposed to automate segmentation, they often rely on manually generated ground truths which are especially challenging and time consuming to generate in biology due to the requirement of domain expertise. Furthermore, these methods have limited generalization capacity, requiring additional manual labels to be generated for each dataset and use case. We introduce MAESTER (Masked AutoEncoder guided Segmen'Iation at pixEl Resolution), a self-supervised method for accurate, subcellular structure segmentation at pixel resolution. MAESTER treats segmentation as a representation learning and clustering problem. Specifically, MAESTER learns semantically meaningful token representations of multi-pixel image patches while simultaneously maintaining a sufficiently large field of view for contextual learning. We also develop a cover-and-stride inference strategy to achieve pixel-level subcellular structure segmentation. We evaluated MAESTER on a publicly available volumetric electron microscopy (VEM) dataset of primary mouse pancreatic islets \u00df cells and achieved up-wards of 29.1 % improvement over state-of-the-art under the same evaluation criteria. Furthermore, our results are competitive against supervised methods trained on the same tasks, closing the gap between self-supervised and supervised approaches. MAESTER shows promise for alleviating the critical bottleneck of ground truth generation for imaging related data analysis and thereby greatly increasing the rate of biological discovery. Code available at https://github.com/bowang-lab/MAESTER."
                },
                {
                    "title": "Sigmoid Loss for Language Image Pre-Training",
                    "abstract": "We propose a simple pairwise sigmoid loss for imagetext pre-training. Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization. The sigmoid loss simultaneously allows further scaling up the batch size, while also performing better at smaller batch sizes. With only four TPUv4 chips, we can train a Base CLIP model at 4k batch size and a Large LiT model at 20k batch size, the latter achieves 84.5% ImageNet zero-shot accuracy in two days. This disentanglement of the batch size from the loss further allows us to study the impact of examples vs pairs and negative to positive ratio. Finally, we push the batch size to the extreme, up to one million, and find that the benefits of growing batch size quickly diminish, with a more reasonable batch size of 32k being sufficient. We hope our research motivates further explorations in improving the quality and efficiency of language-image pre-training."
                },
                {
                    "title": "Scaling Vision Transformers to 22 Billion Parameters",
                    "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                },
                {
                    "title": "RxRx3: Phenomics Map of Biology",
                    "abstract": "The combination of modern genetic perturbation techniques with high content screening has enabled genome-scale cell microscopy experiments that can be leveraged to construct maps of biology. These are built by processing microscopy images to produce readouts in unified and relatable representation space to capture known biological relationships and discover new ones. To further enable the scientific community to develop methods and insights from map-scale data, here we release RxRx3, the first ever public high-content screening dataset combining genome-scale CRISPR knockouts with multiple-concentration screening of small molecules (a set of FDA approved and commercially available bioactive compounds). The dataset contains 6-channel fluorescent microscopy images and associated deep learning embeddings from over 2.2 million wells that span 17,063 CRISPR knockouts and 1,674 compounds at 8 doses each. RxRx3 is one of the largest collections of cellular screening data, and as far as we know, the largest generated consistently via a common experimental protocol within a single laboratory. Our goal in releasing RxRx3 is to demonstrate the benefits of generating consistent data, enable the development of the machine learning methods on this scale of data and to foster research, methods development, and collaboration. For more information about RxRx3 please visit RxRx.ai/rxrx3"
                },
                {
                    "title": "GPS++: Reviving the Art of Message Passing for Molecular Property Prediction",
                    "abstract": "We present GPS++, a hybrid Message Passing Neural Network / Graph Transformer model for molecular property prediction. Our model integrates a well-tuned local message passing component and biased global attention with other key ideas from prior literature to achieve state-of-the-art results on large-scale molecular dataset PCQM4Mv2. Through a thorough ablation study we highlight the impact of individual components and find that nearly all of the model's performance can be maintained without any use of global self-attention, showing that message passing is still a competitive approach for 3D molecular property prediction despite the recent dominance of graph transformers. We also find that our approach is significantly more accurate than prior art when 3D positional information is not available."
                },
                {
                    "title": "RxRx1: A Dataset for Evaluating Experimental Batch Correction Methods",
                    "abstract": "High-throughput screening techniques are commonly used to obtain large quantities of data in many fields of biology. It is well known that artifacts arising from variability in the technical execution of different experimental batches within such screens confound these observations, and can lead to invalid biological conclusions. It is, therefore, necessary to account for these batch effects when analyzing outcomes. In this paper, we describe RxRx1, a biological dataset designed specifically for the systematic study of batch effect correction methods. The dataset consists of 125,510 high-resolution fluorescence microscopy images of human cells under 1,138 genetic perturbations in 51 experimental batches across 4 cell types. Visual inspection of the images clearly demonstrates significant batch effects. We also propose a classification task designed to evaluate the effectiveness of experimental batch correction methods on these images and examine the performance of a number of correction methods on this task. Our goal in releasing RxRx1 is to encourage the development of effective experimental batch correction methods that generalize well to unseen experimental batches. The dataset can be downloaded at https://rxrx.ai."
                },
                {
                    "title": "LAION-5B: An open large-scale dataset for training next generation image-text models",
                    "abstract": "Groundbreaking language-vision architectures like CLIP and DALL-E proved the utility of training on large amounts of noisy image-text data, without relying on expensive accurate labels used in standard vision unimodal supervised learning. The resulting models showed capabilities of strong text-guided image generation and transfer to downstream tasks, while performing remarkably at zero-shot classification with noteworthy out-of-distribution robustness. Since then, large-scale language-vision models like ALIGN, BASIC, GLIDE, Flamingo and Imagen made further improvements. Studying the training and capabilities of such models requires datasets containing billions of image-text pairs. Until now, no datasets of this size have been made openly available for the broader research community. To address this problem and democratize research on large-scale multi-modal models, we present LAION-5B - a dataset consisting of 5.85 billion CLIP-filtered image-text pairs, of which 2.32B contain English language. We show successful replication and fine-tuning of foundational models like CLIP, GLIDE and Stable Diffusion using the dataset, and discuss further experiments enabled with an openly available dataset of this scale. Additionally we provide several nearest neighbor indices, an improved web-interface for dataset exploration and subset generation, and detection scores for watermark, NSFW, and toxic content detection. Announcement page https://laion.ai/laion-5b-a-new-era-of-open-large-scale-multi-modal-datasets/"
                },
                {
                    "title": "Is Self-Supervised Learning More Robust Than Supervised Learning?",
                    "abstract": "Self-supervised contrastive learning is a powerful tool to learn visual representation without labels. Prior work has primarily focused on evaluating the recognition accuracy of various pre-training algorithms, but has overlooked other behavioral aspects. In addition to accuracy, distributional robustness plays a critical role in the reliability of machine learning models. We design and conduct a series of robustness tests to quantify the behavioral differences between contrastive learning and supervised learning to downstream or pre-training data distribution changes. These tests leverage data corruptions at multiple levels, ranging from pixel-level gamma distortion to patch-level shuffling and to dataset-level distribution shift. Our tests unveil intriguing robustness behaviors of contrastive and supervised learning. On the one hand, under downstream corruptions, we generally observe that contrastive learning is surprisingly more robust than supervised learning. On the other hand, under pre-training corruptions, we find contrastive learning vulnerable to patch shuffling and pixel intensity change, yet less sensitive to dataset-level distribution change. We attempt to explain these results through the role of data augmentation and feature space properties. Our insight has implications in improving the downstream robustness of supervised learning."
                },
                {
                    "title": "Pre-training via Denoising for Molecular Property Prediction",
                    "abstract": "Many important problems involving molecular property prediction from 3D structures have limited data, posing a generalization challenge for neural networks. In this paper, we describe a pre-training technique based on denoising that achieves a new state-of-the-art in molecular property prediction by utilizing large datasets of 3D molecular structures at equilibrium to learn meaningful representations for downstream tasks. Relying on the well-known link between denoising autoencoders and score-matching, we show that the denoising objective corresponds to learning a molecular force field -- arising from approximating the Boltzmann distribution with a mixture of Gaussians -- directly from equilibrium structures. Our experiments demonstrate that using this pre-training objective significantly improves performance on multiple benchmarks, achieving a new state-of-the-art on the majority of targets in the widely used QM9 dataset. Our analysis then provides practical insights into the effects of different factors -- dataset sizes, model size and architecture, and the choice of upstream and downstream datasets -- on pre-training."
                },
                {
                    "title": "Masked Autoencoders As Spatiotemporal Learners",
                    "abstract": "This paper studies a conceptually simple extension of Masked Autoencoders (MAE) to spatiotemporal representation learning from videos. We randomly mask out spacetime patches in videos and learn an autoencoder to reconstruct them in pixels. Interestingly, we show that our MAE method can learn strong representations with almost no inductive bias on spacetime (only except for patch and positional embeddings), and spacetime-agnostic random masking performs the best. We observe that the optimal masking ratio is as high as 90% (vs. 75% on images), supporting the hypothesis that this ratio is related to information redundancy of the data. A high masking ratio leads to a large speedup, e.g.,>4x in wall-clock time or even more. We report competitive results on several challenging video datasets using vanilla Vision Transformers. We observe that MAE can outperform supervised pre-training by large margins. We further report encouraging results of training on real-world, uncurated Instagram data. Our study suggests that the general framework of masked autoencoding (BERT, MAE, etc.) can be a unified methodology for representation learning with minimal domain knowledge."
                },
                {
                    "title": "Survey on Self-Supervised Learning: Auxiliary Pretext Tasks and Contrastive Learning Methods in Imaging",
                    "abstract": "Although deep learning algorithms have achieved significant progress in a variety of domains, they require costly annotations on huge datasets. Self-supervised learning (SSL) using unlabeled data has emerged as an alternative, as it eliminates manual annotation. To do this, SSL constructs feature representations using pretext tasks that operate without manual annotation, which allows models trained in these tasks to extract useful latent representations that later improve downstream tasks such as object classification and detection. The early methods of SSL are based on auxiliary pretext tasks as a way to learn representations using pseudo-labels, or labels that were created automatically based on the dataset\u2019s attributes. Furthermore, contrastive learning has also performed well in learning representations via SSL. To succeed, it pushes positive samples closer together, and negative ones further apart, in the latent space. This paper provides a comprehensive literature review of the top-performing SSL methods using auxiliary pretext and contrastive learning techniques. It details the motivation for this research, a general pipeline of SSL, the terminologies of the field, and provides an examination of pretext tasks and self-supervised methods. It also examines how self-supervised methods compare to supervised ones, and then discusses both further considerations and ongoing challenges faced by SSL."
                },
                {
                    "title": "How to Understand Masked Autoencoders",
                    "abstract": "\"Masked Autoencoders (MAE) Are Scalable Vision Learners\"revolutionizes the self-supervised learning method in that it not only achieves the state-of-the-art for image pre-training, but is also a milestone that bridges the gap between visual and linguistic masked autoencoding (BERT-style) pre-trainings. However, to our knowledge, to date there are no theoretical perspectives to explain the powerful expressivity of MAE. In this paper, we, for the first time, propose a unified theoretical framework that provides a mathematical understanding for MAE. Specifically, we explain the patch-based attention approaches of MAE using an integral kernel under a non-overlapping domain decomposition setting. To help the research community to further comprehend the main reasons of the great success of MAE, based on our framework, we pose five questions and answer them with mathematical rigor using insights from operator theory."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "CLOOB: Modern Hopfield Networks with InfoLOOB Outperform CLIP",
                    "abstract": "CLIP yielded impressive results on zero-shot transfer learning tasks and is considered as a foundation model like BERT or GPT3. CLIP vision models that have a rich representation are pre-trained using the InfoNCE objective and natural language supervision before they are fine-tuned on particular tasks. Though CLIP excels at zero-shot transfer learning, it suffers from an explaining away problem, that is, it focuses on one or few features, while neglecting other relevant features. This problem is caused by insufficiently extracting the covariance structure in the original multi-modal data. We suggest to use modern Hopfield networks to tackle the problem of explaining away. Their retrieved embeddings have an enriched covariance structure derived from co-occurrences of features in the stored embeddings. However, modern Hopfield networks increase the saturation effect of the InfoNCE objective which hampers learning. We propose to use the InfoLOOB objective to mitigate this saturation effect. We introduce the novel\"Contrastive Leave One Out Boost\"(CLOOB), which uses modern Hopfield networks for covariance enrichment together with the InfoLOOB objective. In experiments we compare CLOOB to CLIP after pre-training on the Conceptual Captions and the YFCC dataset with respect to their zero-shot transfer learning performance on other datasets. CLOOB consistently outperforms CLIP at zero-shot transfer learning across all considered architectures and datasets."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Self-Supervised Graph Transformer on Large-Scale Molecular Data",
                    "abstract": "How to obtain informative representations of molecules is a crucial prerequisite in AI-driven drug design and discovery. Recent researches abstract molecules as graphs and employ Graph Neural Networks (GNNs) for molecular representation learning. Nevertheless, two issues impede the usage of GNNs in real scenarios: (1) insufficient labeled molecules for supervised training; (2) poor generalization capability to new-synthesized molecules. To address them both, we propose a novel framework, GROVER, which stands for Graph Representation frOm self-superVised mEssage passing tRansformer. With carefully designed self-supervised tasks in node-, edge- and graph-level, GROVER can learn rich structural and semantic information of molecules from enormous unlabelled molecular data. Rather, to encode such complex information, GROVER integrates Message Passing Networks into the Transformer-style architecture to deliver a class of more expressive encoders of molecules. The flexibility of GROVER allows it to be trained efficiently on large-scale molecular dataset without requiring any supervision, thus being immunized to the two issues mentioned above. We pre-train GROVER with 100 million parameters on 10 million unlabelled molecules -- the biggest GNN and the largest training dataset in molecular representation learning. We then leverage the pre-trained GROVER for molecular property prediction followed by task-specific fine-tuning, where we observe a huge improvement (more than 6% on average) from current state-of-the-art methods on 11 challenging benchmarks. The insights we gained are that well-designed self-supervision losses and largely-expressive pre-trained models enjoy the significant potential on performance boosting."
                },
                {
                    "title": "Big Self-Supervised Models are Strong Semi-Supervised Learners",
                    "abstract": "One paradigm for learning from few labeled examples while making best use of a large amount of unlabeled data is unsupervised pretraining followed by supervised fine-tuning. Although this paradigm uses unlabeled data in a task-agnostic way, in contrast to most previous approaches to semi-supervised learning for computer vision, we show that it is surprisingly effective for semi-supervised learning on ImageNet. A key ingredient of our approach is the use of a big (deep and wide) network during pretraining and fine-tuning. We find that, the fewer the labels, the more this approach (task-agnostic use of unlabeled data) benefits from a bigger network. After fine-tuning, the big network can be further improved and distilled into a much smaller one with little loss in classification accuracy by using the unlabeled examples for a second time, but in a task-specific way. The proposed semi-supervised learning algorithm can be summarized in three steps: unsupervised pretraining of a big ResNet model using SimCLRv2 (a modification of SimCLR), supervised fine-tuning on a few labeled examples, and distillation with unlabeled examples for refining and transferring the task-specific knowledge. This procedure achieves 73.9\\% ImageNet top-1 accuracy with just 1\\% of the labels ($\\le$13 labeled images per class) using ResNet-50, a $10\\times$ improvement in label efficiency over the previous state-of-the-art. With 10\\% of labels, ResNet-50 trained with our method achieves 77.5\\% top-1 accuracy, outperforming standard supervised training with all of the labels."
                },
                {
                    "title": "Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning",
                    "abstract": "We introduce Bootstrap Your Own Latent (BYOL), a new approach to self-supervised image representation learning. BYOL relies on two neural networks, referred to as online and target networks, that interact and learn from each other. From an augmented view of an image, we train the online network to predict the target network representation of the same image under a different augmented view. At the same time, we update the target network with a slow-moving average of the online network. While state-of-the art methods rely on negative pairs, BYOL achieves a new state of the art without them. BYOL reaches $74.3\\%$ top-1 classification accuracy on ImageNet using a linear evaluation with a ResNet-50 architecture and $79.6\\%$ with a larger ResNet. We show that BYOL performs on par or better than the current state of the art on both transfer and semi-supervised benchmarks. Our implementation and pretrained models are given on GitHub."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Momentum Contrast for Unsupervised Visual Representation Learning",
                    "abstract": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a moving-averaged encoder. This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning. MoCo provides competitive results under the common linear protocol on ImageNet classification. More importantly, the representations learned by MoCo transfer well to downstream tasks. MoCo can outperform its supervised pre-training counterpart in 7 detection/segmentation tasks on PASCAL VOC, COCO, and other datasets, sometimes surpassing it by large margins. This suggests that the gap between unsupervised and supervised representation learning has been largely closed in many vision tasks."
                },
                {
                    "title": "On Variational Bounds of Mutual Information",
                    "abstract": "Estimating and optimizing Mutual Information (MI) is core to many problems in machine learning; however, bounding MI in high dimensions is challenging. To establish tractable and scalable objectives, recent work has turned to variational bounds parameterized by neural networks, but the relationships and tradeoffs between these bounds remains unclear. In this work, we unify these recent developments in a single framework. We find that the existing variational lower bounds degrade when the MI is large, exhibiting either high bias or high variance. To address this problem, we introduce a continuum of lower bounds that encompasses previous bounds and flexibly trades off bias and variance. On high-dimensional, controlled problems, we empirically characterize the bias and variance of the bounds and their gradients and demonstrate the effectiveness of our new bounds for estimation and representation learning."
                },
                {
                    "title": "Accurate Prediction of Biological Assays with High-Throughput Microscopy Images and Convolutional Networks",
                    "abstract": "Predicting the outcome of biological assays based on high-throughput imaging data is a highly promising task in drug discovery since it can tremendously increase hit rates and suggest novel chemical scaffolds. However, end-to-end learning with convolutional neural networks (CNNs) has not been assessed for the task biological assay prediction despite the success of these networks at visual recognition. We compared several CNNs trained directly on high-throughput imaging data to a) CNNs trained on cell-centric crops and to b) the current state-of-the-art: fully connected networks trained on precalculated morphological cell features. The comparison was performed on the Cell Painting data set, the largest publicly available data set of microscopic images of cells with approximately 30,000 compound treatments. We found that CNNs perform significantly better at predicting the outcome of assays than fully connected networks operating on precomputed morphological features of cells. Surprisingly, the best performing method could predict 32% of the 209 biological assays at high predictive performance (AUC > 0.9) indicating that the cell morphology changes contain a large amount of information about compound activities. Our results suggest that many biological assays could be replaced by high-throughput imaging together with convolutional neural networks and that the costly cell segmentation and feature extraction step can be replaced by convolutional neural networks."
                },
                {
                    "title": "ChEMBL: towards direct deposition of bioassay data",
                    "abstract": "Abstract ChEMBL is a large, open-access bioactivity database (https://www.ebi.ac.uk/chembl), previously described in the 2012, 2014 and 2017 Nucleic Acids Research Database Issues. In the last two years, several important improvements have been made to the database and are described here. These include more robust capture and representation of assay details; a new data deposition system, allowing updating of data sets and deposition of supplementary data; and a completely redesigned web interface, with enhanced search and filtering capabilities."
                },
                {
                    "title": "Representation Learning with Contrastive Predictive Coding",
                    "abstract": "While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments."
                },
                {
                    "title": "Improved Deep Metric Learning with Multi-class N-pair Loss Objective",
                    "abstract": "Deep metric learning has gained much popularity in recent years, following the success of deep learning. However, existing frameworks of deep metric learning based on contrastive loss and triplet loss often suffer from slow convergence, partially because they employ only one negative example while not interacting with the other negative classes in each update. In this paper, we propose to address this problem with a new metric learning objective called multi-class N-pair loss. The proposed objective function firstly generalizes triplet loss by allowing joint comparison among more than one negative examples - more specifically, N-1 negative examples - and secondly reduces the computational burden of evaluating deep embedding vectors via an efficient batch construction strategy using only N pairs of examples, instead of (N+1) x N. We demonstrate the superiority of our proposed loss to the triplet loss as well as other competing loss functions for a variety of tasks on several visual recognition benchmark, including fine-grained object recognition and verification, image clustering and retrieval, and face verification and identification."
                },
                {
                    "title": "Cell Painting, a high-content image-based assay for morphological profiling using multiplexed fluorescent dyes",
                    "abstract": "In morphological profiling, quantitative data are extracted from microscopy images of cells to identify biologically relevant similarities and differences among samples based on these profiles. This protocol describes the design and execution of experiments using Cell Painting, which is a morphological profiling assay that multiplexes six fluorescent dyes, imaged in five channels, to reveal eight broadly relevant cellular components or organelles. Cells are plated in multiwell plates, perturbed with the treatments to be tested, stained, fixed, and imaged on a high-throughput microscope. Next, an automated image analysis software identifies individual cells and measures \u223c1,500 morphological features (various measures of size, shape, texture, intensity, and so on) to produce a rich profile that is suitable for the detection of subtle phenotypes. Profiles of cell populations treated with different experimental perturbations can be compared to suit many goals, such as identifying the phenotypic impact of chemical or genetic perturbations, grouping compounds and/or genes into functional pathways, and identifying signatures of disease. Cell culture and image acquisition takes 2 weeks; feature extraction and data analysis take an additional 1\u20132 weeks."
                },
                {
                    "title": "Batch Effect Confounding Leads to Strong Bias in Performance Estimates Obtained by Cross-Validation",
                    "abstract": "Background With the large amount of biological data that is currently publicly available, many investigators combine multiple data sets to increase the sample size and potentially also the power of their analyses. However, technical differences (\u201cbatch effects\u201d) as well as differences in sample composition between the data sets may significantly affect the ability to draw generalizable conclusions from such studies. Focus The current study focuses on the construction of classifiers, and the use of cross-validation to estimate their performance. In particular, we investigate the impact of batch effects and differences in sample composition between batches on the accuracy of the classification performance estimate obtained via cross-validation. The focus on estimation bias is a main difference compared to previous studies, which have mostly focused on the predictive performance and how it relates to the presence of batch effects. Data We work on simulated data sets. To have realistic intensity distributions, we use real gene expression data as the basis for our simulation. Random samples from this expression matrix are selected and assigned to group 1 (e.g., \u2018control\u2019) or group 2 (e.g., \u2018treated\u2019). We introduce batch effects and select some features to be differentially expressed between the two groups. We consider several scenarios for our study, most importantly different levels of confounding between groups and batch effects. Methods We focus on well-known classifiers: logistic regression, Support Vector Machines (SVM), k-nearest neighbors (kNN) and Random Forests (RF). Feature selection is performed with the Wilcoxon test or the lasso. Parameter tuning and feature selection, as well as the estimation of the prediction performance of each classifier, is performed within a nested cross-validation scheme. The estimated classification performance is then compared to what is obtained when applying the classifier to independent data."
                },
                {
                    "title": "How accurate is in vitro prediction of carcinogenicity?",
                    "abstract": "Positive genetic toxicity data suggest carcinogenic hazard, and this can stop a candidate pharmaceutical reaching the clinic. However, during the last decade, it has become clear that many non\u2010carcinogens produce misleading positive results in one or other of the regulatory genotoxicity assays. These doubtful conclusions cost a lot of time and money, as they trigger additional testing of apparently genotoxic candidates, both in vitro and in animals, to discover whether the suggested hazard is genuine. This in turn means that clinical trials can be put on hold. This review describes the current approaches to the \u2018misleading positive\u2019 problem as well as efforts to reduce the use of animals in genotoxicity assessment. The following issues are then addressed: the application of genotoxicity testing screens earlier in development; the search for new or improved in vitro genotoxicity tests; proposed changes to the International Committee on Harmonisation guidance on genotoxicity testing [S2(R1)]. Together, developments in all these areas offer good prospects of a more rapid and cost\u2010effective way to understand genetic toxicity concerns."
                },
                {
                    "title": "Uni-Mol: A Universal 3D Molecular Representation Learning Framework",
                    "abstract": "Molecular representation learning (MRL) has gained tremendous attention due to its critical role in learning from limited supervised data for applications like drug design. In most MRL methods, molecules are treated as 1D sequential tokens or 2D topology graphs, limiting their ability to incorporate 3D information for downstream tasks and, in particular, making it almost impossible for 3D geometry prediction or generation. Herein, we propose Uni-Mol, a universal MRL framework that significantly enlarges the representation ability and application scope of MRL schemes. Uni-Mol is composed of two models with the same SE(3)-equivariant transformer architecture: a molecular pretraining model trained by 209M molecular conformations; a pocket pretraining model trained by 3M candidate protein pocket data. The two models are used independently for separate tasks, and are combined when used in protein-ligand binding tasks. By properly incorporating 3D information, Uni-Mol outperforms SOTA in 14/15 molecular property prediction tasks. Moreover, Uni-Mol achieves superior performance in 3D spatial tasks, including protein-ligand binding pose prediction, molecular conformation generation, etc. Finally, we show that Uni-Mol can be successfully applied to the tasks with few-shot data like pocket druggability prediction. The code, model, and data are made publicly available at https://github.com/dptech-corp/Uni-Mol ."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Analysis of the e\ufb00ects of related \ufb01ngerprints on molecular similarity using an eigenvalue entropy approach",
                    "abstract": "Two-dimensional (2D) chemical \ufb01ngerprints are widely used as numerical features for the quanti\ufb01cation of structural similarity of chemical compounds, which is an important step in similarity-based virtual screening (VS). Here, using an eigenvalue-based entropy approach, we sought to identify 2D \ufb01ngerprints with little to no contribution to shaping the eigenvalue distribution of the feature matrix as related \ufb01ngerprints and examined the degree to which these related 2D \ufb01ngerprints in\ufb02uence molecular similarity scores via Tanimoto coe\ufb03cient. We found that there are many related \ufb01ngerprints in publicly available \ufb01ngerprint schemes and that their presence in the feature set tends to decrease the similarity scores. Our results have implication in the optimal selection of 2D \ufb01ngerprints and the identi\ufb01cation of potential hits for compounds with target biological activity in VS."
                },
                {
                    "title": "Molecular Systems Biology Peer Review Process File Automated Analysis of High-content Microscopy Data with Deep Learning Editor: Maria Polychronidou Transaction Report",
                    "abstract": "(Note: With the exception of the correction of typographical or spelling errors that could be a source of ambiguity, letters and reports are not edited. The original formatting of letters and referee reports may not be reflected in this compilation.) Thank you again for submitting your work to Molecular Systems Biology. We have now heard back from the three referees who agreed to evaluate your study. As you will see below, the reviewers appreciate that the proposed approach is going to be of interest for the field. However, they raise a series of concerns, which we would ask you to address in a revision. The reviewers' recommendations are quite clear so I think that there is no need to repeat the points listed below. Of course, please feel free to contact me in case you would like to discuss any of the issues raised by the reviewers. In this paper, the authors present a deep learning based approach called DeepLoc for automatically classifying protein localisation in fluorescence images of yeast cells. They test the method do reanalyse an imaging data set they previously produced in their lab (Chong et al. 2015) and to compare ith with the previous analysis that was approached using support vector machine (SVM) classifiers (ensLOC). The authors demonstrate that DeepLoc outperforms ensLOC by requiring less training and producing more accurate classifications. They also show that it can be more easily transferred to imaging data sets obtained using different instruments or using cells that differ in their overall morphology (e.g. after treatment with \u03b1-factor) or where different fluorophore tag have been used. This is a well-executed study that demonstrates the utility of deep learning for studies of protein"
                }
            ],
            "categories": [
                "q-bio.QM",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively perform contrastive phenomolecular retrieval to identify the molecular impact on cellular function from multi-modal paired phenomic and molecular data?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing medicinal research, as it can lead to the discovery of novel therapeutic drug candidates by enabling a deeper understanding of how molecular perturbations affect cellular morphology. This research could significantly impact future studies in drug development and personalized medicine, as it provides a framework for systematically exploring the relationship between molecular structures and their biological effects. By addressing this question, we can enhance knowledge in multi-modal learning and its applications in biology, potentially leading to more effective treatments for various diseases.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include: (1) The limited size of multi-modal paired phenomic molecular datasets, which are significantly smaller than typical text-image datasets, making it difficult to train robust models. (2) The presence of inactive perturbations in the data, which complicates the identification of biologically meaningful interactions and can lead to misinterpretation of results. (3) The necessity to account for molecular concentration, as the same molecule can exhibit varying effects depending on its dosage, adding complexity to the learning process. Naive approaches may fail due to these intricacies, as they might not adequately address the noise from batch effects or the challenge of filtering out inactive perturbations.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the small size of available datasets and the challenges associated with batch effects and inactive perturbations. Many existing solutions have not effectively addressed the need for a robust phenomic embedding to filter out non-biologically meaningful data. Additionally, prior work may not have fully considered the importance of molecular concentration in understanding cellular responses. Our approach differs by focusing on a contrastive learning framework that specifically targets these challenges, aiming to improve retrieval rates and enhance the understanding of molecular effects on cellular function.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves utilizing a contrastive learning framework to align paired phenomic and molecular data. We will collect a dataset comprising various molecular perturbations and their corresponding cellular morphology images obtained through high-throughput microscopy. The evaluation metric will focus on phenomolecular retrieval rates, assessing the model's ability to accurately retrieve molecular perturbants based on the"
            }
        },
        "author_data": {
            "83add685-6e44-41d3-b8d0-cdfec7329ef2": {
                "pk": "83add685-6e44-41d3-b8d0-cdfec7329ef2",
                "name": "Philip Fradkin",
                "collaborators": [
                    "Ruian Shi",
                    "Bo Wang",
                    "Brendan Frey",
                    "Leo J. Lee",
                    "Maciej Sypetkowski",
                    "Frederik Wenkel",
                    "Farimah Poursafaei",
                    "Nia Dickson",
                    "Karush Suri",
                    "Dominique Beaini"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Genomic Data",
                    "Contrastive Learning",
                    "Drug Discovery"
                ],
                "publications": [
                    {
                        "title": "Splicing Up Your Predictions with RNA Contrastive Learning",
                        "abstract": "In the face of rapidly accumulating genomic data, our understanding of the RNA regulatory code remains incomplete. Recent self-supervised methods in other domains have demonstrated the ability to learn rules underlying the data-generating process such as sentence structure in language. Inspired by this, we extend contrastive learning techniques to genomic data by utilizing functional similarities between sequences generated through alternative splicing and gene duplication. Our novel dataset and contrastive objective enable the learning of generalized RNA isoform representations. We validate their utility on downstream tasks such as RNA half-life and mean ribosome load prediction. Our pre-training strategy yields competitive results using linear probing on both tasks, along with up to a two-fold increase in Pearson correlation in low-data conditions. Importantly, our exploration of the learned latent space reveals that our contrastive objective yields semantically meaningful representations, underscoring its potential as a valuable initialization technique for RNA property prediction."
                    },
                    {
                        "title": "On the Scalability of GNNs for Molecular Graphs",
                        "abstract": "Scaling deep learning models has been at the heart of recent revolutions in language modelling and image generation. Practitioners have observed a strong relationship between model size, dataset size, and performance. However, structure-based architectures such as Graph Neural Networks (GNNs) are yet to show the benefits of scale mainly due to the lower efficiency of sparse operations, large data requirements, and lack of clarity about the effectiveness of various architectures. We address this drawback of GNNs by studying their scaling behavior. Specifically, we analyze message-passing networks, graph Transformers, and hybrid architectures on the largest public collection of 2D molecular graphs. For the first time, we observe that GNNs benefit tremendously from the increasing scale of depth, width, number of molecules, number of labels, and the diversity in the pretraining datasets. We further demonstrate strong finetuning scaling behavior on 38 highly competitive downstream tasks, outclassing previous large models. This gives rise to MolGPS, a new graph foundation model that allows to navigate the chemical space, outperforming the previous state-of-the-arts on 26 out the 38 downstream tasks. We hope that our work paves the way for an era where foundational GNNs drive pharmaceutical drug discovery."
                    }
                ]
            },
            "7b084e9b-859a-4930-bfb2-ee9a4f90bda5": {
                "pk": "7b084e9b-859a-4930-bfb2-ee9a4f90bda5",
                "name": "Puria Azadi",
                "collaborators": [
                    "Puria Azadi Moghadam",
                    "Hossein Farahani",
                    "Ali Bashashati",
                    "Sanne Van Dalen",
                    "Karina C. Martin",
                    "Jochen Lennerz",
                    "Stephen Yip",
                    "Hamed Rahmani Khezri",
                    "Mohammad Karimzadeh Farshbafan",
                    "Vahid Shah-Mansouri"
                ],
                "domain": [
                    "Deep Learning",
                    "Histopathology",
                    "Image Synthesis",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "A Morphology Focused Diffusion Probabilistic Model for Synthesis of Histopathology Images",
                        "abstract": "Visual microscopic study of diseased tissue by pathologists has been the cornerstone for cancer diagnosis and prognostication for more than a century. Recently, deep learning methods have made significant advances in the analysis and classification of tissue images. However, there has been limited work on the utility of such models in generating histopathology images. These synthetic images have several applications in pathology including utilities in education, proficiency testing, privacy, and data sharing. Recently, diffusion probabilistic models were introduced to generate high quality images. Here, for the first time, we investigate the potential use of such models along with prioritized morphology weighting and color normalization to synthesize high quality histopathology images of brain cancer. Our detailed results show that diffusion probabilistic models are capable of synthesizing a wide range of histopathology images and have superior performance compared to generative adversarial networks."
                    },
                    {
                        "title": "Deep Q-Learning for Dynamic Reliability Aware NFV-Based Service Provisioning",
                        "abstract": "Network function virtualization (NFV) is referred to the technology in which softwarized network functions virtually run on commodity servers. Such functions are called virtual network functions (VNFs). A specific service is composed of a set of VNFs. This is a paradigm shift for service provisioning in telecom networks which introduces new design and implementation challenges. One of these challenges is to meet the reliability requirement of the requested services considering the reliability of the commodity servers. VNF placement which is the problem of assigning commodity servers to the VNFs becomes crucial under such circumstances. To address such an issue, in this paper, we employ Deep Q-Network (DQN) to model NFV placement problem considering the reliability requirement of the services. The output of the introduced model determines what placement will be optimal in each state. Numerical evaluations show that the introduced model can significantly improve the performance of the network operator."
                    },
                    {
                        "title": "AMIGO: Sparse Multi-Modal Graph Transformer with Shared-Context Processing for Representation Learning of Giga-pixel Images",
                        "abstract": "Processing giga-pixel whole slide histopathology images (WSI) is a computationally expensive task. Multiple instance learning (MIL) has become the conventional approach to process WSIs, in which these images are split into smaller patches for further processing. However, MIL-based techniques ignore explicit information about the individual cells within a patch. In this paper, by defining the novel concept of shared-context processing, we designed a multi-modal Graph Transformer (AMIGO) that uses the celluar graph within the tissue to provide a single representation for a patient while taking advantage of the hierarchical structure of the tissue, enabling a dynamic focus between cell-level and tissue-level information. We benchmarked the performance of our model against multiple state-of-the-art methods in survival prediction and showed that ours can significantly outperform all of them including hierarchical Vision Transformer (ViT). More importantly, we show that our model is strongly robust to missing information to an extent that it can achieve the same performance with as low as 20% of the data. Finally, in two different cancer datasets, we demonstrated that our model was able to stratify the patients into low-risk and high-risk groups while other state-of-the-art methods failed to achieve this goal. We also publish a large dataset of immunohistochemistry images (InUIT) containing 1,600 tissue microarray (TMA) cores from 188 patients along with their survival information, making it one of the largest publicly available datasets in this context."
                    }
                ]
            },
            "056ec196-13f4-4854-86ac-8ab854afe06e": {
                "pk": "056ec196-13f4-4854-86ac-8ab854afe06e",
                "name": "Karush Suri",
                "collaborators": [
                    "Rinki Gupta",
                    "Xiao Qi Shi",
                    "Konstantinos Plataniotis",
                    "Yuri Lawryshyn",
                    "Konstantinos N. Plataniotis",
                    "Yuri A. Lawryshyn",
                    "Maciej Sypetkowski",
                    "Frederik Wenkel",
                    "Farimah Poursafaei",
                    "Nia Dickson"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Deep Learning",
                    "Gesture Recognition",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Activity Detection from Wearable Electromyogram Sensors using Hidden Markov Model",
                        "abstract": "Surface electromyography (sEMG) has gained significant importance during recent advancements in consumer electronics for healthcare systems, gesture analysis and recognition and sign language communication. For such a system, it is imperative to determine the regions of activity in a continuously recorded sEMG signal. The proposed work provides a novel activity detection approach based on Hidden Markov Models (HMM) using sEMG signals recorded when various hand gestures are performed. Detection procedure is designed based on a probabilistic outlook by making use of mathematical models. The requirement of a threshold for activity detection is obviated making it subject and activity independent. Correctness of the predicted outputs is asserted by classifying the signal segments around the detected transition regions as activity or rest. Classified outputs are compared with the transition regions in a stimulus given to the subject to perform the activity. The activity onsets are detected with an average of 96.25% accuracy whereas the activity termination regions with an average of 87.5% accuracy with the considered set of six activities and four subjects."
                    },
                    {
                        "title": "Continuous sign language recognition from wearable IMUs using deep capsule networks and game theory",
                        "abstract": "Sign Language is used by the deaf community all over world. The work presented here proposes a novel one-dimensional deep capsule network (CapsNet) architecture for continuous Indian Sign Language recognition by means of signals obtained from a custom designed wearable IMU system. The performance of the proposed CapsNet architecture is assessed by altering dynamic routing between capsule layers. The proposed CapsNet yields improved accuracy values of 94% for 3 routings and 92.50% for 5 routings in comparison with the convolutional neural network (CNN) that yields an accuracy of 87.99%. Improved learning of the proposed architecture is also validated by spatial activations depicting excited units at the predictive layer. Finally, a novel non-cooperative pick-and-predict competition is designed between CapsNet and CNN. Higher value of Nash equilibrium for CapsNet as compared to CNN indicates the suitability of the proposed approach."
                    },
                    {
                        "title": "Classification of Hand Gestures from Wearable IMUs using Deep Neural Network",
                        "abstract": "IMUs are gaining significant importance in the field of hand gesture analysis, trajectory detection and kinematic functional study. An Inertial Measurement Unit (IMU) consists of tri-axial accelerometers and gyroscopes which can together be used for formation analysis. The paper presents a novel classification approach using a Deep Neural Network (DNN) for classifying hand gestures obtained from wearable IMU sensors. An optimization objective is set for the classifier in order to reduce correlation between the activities and fit the signal-set with best performance parameters. Training of the network is carried out by feed-forward computation of the input features followed by the back-propagation of errors. The predicted outputs are analyzed in the form of classification accuracies which are then compared to the conventional classification schemes of SVM and kNN. A 3-5% improvement in accuracies is observed in the case of DNN classification. Results are presented for the recorded accelerometer and gyroscope signals and the considered classification schemes."
                    },
                    {
                        "title": "Dual Stage Classification of Hand Gestures using Surface Electromyogram",
                        "abstract": "Surface electromyography (sEMG) is becoming exceeding useful in applications involving analysis of human motion such as in human-machine interface, assistive technology, healthcare and prosthetic development. The proposed work presents a novel dual stage classification approach for classification of grasping gestures from sEMG signals. A statistical assessment of these activities is presented to determine the similar characteristics between the considered activities. Similar activities are grouped together. In the first stage of classification, an activity is identified as belonging to a group, which is then further classified as one of the activities within the group in the second stage of classification. The performance of the proposed approach is compared to the conventional single stage classification approach in terms of classification accuracies. The classification accuracies obtained using the proposed dual stage classification are significantly higher as compared to that for single stage classification."
                    },
                    {
                        "title": "Transfer Learning for sEMG-based Hand Gesture Classification using Deep Learning in a Master-Slave Architecture",
                        "abstract": "Recent advancements in diagnostic learning and development of gesture-based human machine interfaces have driven surface electromyography (sEMG) towards significant importance. Analysis of hand gestures requires an accurate assessment of sEMG signals. The proposed work presents a novel sequential master-slave architecture consisting of deep neural networks (DNNs) for classification of signs from the Indian sign language using signals recorded from multiple sEMG channels. The performance of the master-slave network is augmented by leveraging additional synthetic feature data generated by long short term memory networks. Performance of the proposed network is compared to that of a conventional DNN prior to and after the addition of synthetic data. Up to 14% improvement is observed in the conventional DNN and up to 9% improvement in master-slave network on addition of synthetic data with an average accuracy value of 93.5% asserting the suitability of the proposed approach."
                    },
                    {
                        "title": "Convolutional Neural Network Array for Sign Language Recognition using Wearable IMUs",
                        "abstract": "Advancements in gesture recognition algorithms have led to a significant growth in sign language translation. By making use of efficient intelligent models, signs can be recognized with precision. The proposed work presents a novel one-dimensional Convolutional Neural Network (CNN) array architecture for recognition of signs from the Indian sign language using signals recorded from a custom designed wearable IMU device. The IMU device makes use of tri-axial accelerometer and gyroscope. The signals recorded using the IMU device are segregated on the basis of their context, such as whether they correspond to signing for a general sentence or an interrogative sentence. The array comprises of two individual CNNs, one classifying the general sentences and the other classifying the interrogative sentence. Performances of individual CNNs in the array architecture are compared to that of a conventional CNN classifying the unsegregated dataset. Peak classification accuracies of 94.20% for general sentences and 95.00% for interrogative sentences achieved with the proposed CNN array in comparison to 93.50% for conventional CNN assert the suitability of the proposed approach."
                    },
                    {
                        "title": "TradeR: Practical Deep Hierarchical Reinforcement Learning for Trade Execution",
                        "abstract": "Advances in Reinforcement Learning (RL) span a wide variety of applications which motivate development in this area. While application tasks serve as suitable benchmarks for real world problems, RL is seldomly used in practical scenarios consisting of abrupt dynamics. This allows one to rethink the problem setup in light of practical challenges. We present Trade Execution using Reinforcement Learning (TradeR) which aims to address two such practical challenges of catastrophy and surprise minimization by formulating trading as a real-world hierarchical RL problem. Through this lens, TradeR makes use of hierarchical RL to execute trade bids on high frequency real market experiences comprising of abrupt price variations during the 2019 fiscal year COVID19 stock market crash. The framework utilizes an energy-based scheme in conjunction with surprise value function for estimating and minimizing surprise. In a large-scale study of 35 stock symbols from the S&P500 index, TradeR demonstrates robustness to abrupt price changes and catastrophic losses while maintaining profitable outcomes. We hope that our work serves as a motivating example for application of RL to practical problems."
                    },
                    {
                        "title": "Energy-based Surprise Minimization for Multi-Agent Value Factorization",
                        "abstract": "Multi-Agent Reinforcement Learning (MARL) has demonstrated significant success in training decentralised policies in a centralised manner by making use of value factorization methods. However, addressing surprise across spurious states and approximation bias remain open problems for multi-agent settings. Towards this goal, we introduce the Energy-based MIXer (EMIX), an algorithm which minimizes surprise utilizing the energy across agents. Our contributions are threefold; (1) EMIX introduces a novel surprise minimization technique across multiple agents in the case of multi-agent partially-observable settings. (2) EMIX highlights a practical use of energy functions in MARL with theoretical guarantees and experiment validations of the energy operator. Lastly, (3) EMIX extends Maxmin Q-learning for addressing overestimation bias across agents in MARL. In a study of challenging StarCraft II micromanagement scenarios, EMIX demonstrates consistent stable performance for multiagent surprise minimization. Moreover, our ablation study highlights the necessity of the energy-based scheme and the need for elimination of overestimation bias in MARL. Our implementation of EMIX can be found at karush17.github.io/emix-web/."
                    },
                    {
                        "title": "Maximum Mutation Reinforcement Learning for Scalable Control",
                        "abstract": "Advances in Reinforcement Learning (RL) have demonstrated data efficiency and optimal control over large state spaces at the cost of scalable performance. Genetic methods, on the other hand, provide scalability but depict hyperparameter sensitivity towards evolutionary operations. However, a combination of the two methods has recently demonstrated success in scaling RL agents to high-dimensional action spaces. Parallel to recent developments, we present the Evolution-based Soft Actor-Critic (ESAC), a scalable RL algorithm. We abstract exploration from exploitation by combining Evolution Strategies (ES) with Soft Actor-Critic (SAC). Through this lens, we enable dominant skill transfer between offsprings by making use of soft winner selections and genetic crossovers in hindsight and simultaneously improve hyperparameter sensitivity in evolutions using the novel Automatic Mutation Tuning (AMT). AMT gradually replaces the entropy framework of SAC allowing the population to succeed at the task while acting as randomly as possible, without making use of backpropagation updates. In a study of challenging locomotion tasks consisting of high-dimensional action spaces and sparse rewards, ESAC demonstrates improved performance and sample efficiency in comparison to the Maximum Entropy framework. Additionally, ESAC presents efficacious use of hardware resources and algorithm overhead. A complete implementation of ESAC can be found at karush17.github.io/esac-web/."
                    },
                    {
                        "title": "On the Scalability of GNNs for Molecular Graphs",
                        "abstract": "Scaling deep learning models has been at the heart of recent revolutions in language modelling and image generation. Practitioners have observed a strong relationship between model size, dataset size, and performance. However, structure-based architectures such as Graph Neural Networks (GNNs) are yet to show the benefits of scale mainly due to the lower efficiency of sparse operations, large data requirements, and lack of clarity about the effectiveness of various architectures. We address this drawback of GNNs by studying their scaling behavior. Specifically, we analyze message-passing networks, graph Transformers, and hybrid architectures on the largest public collection of 2D molecular graphs. For the first time, we observe that GNNs benefit tremendously from the increasing scale of depth, width, number of molecules, number of labels, and the diversity in the pretraining datasets. We further demonstrate strong finetuning scaling behavior on 38 highly competitive downstream tasks, outclassing previous large models. This gives rise to MolGPS, a new graph foundation model that allows to navigate the chemical space, outperforming the previous state-of-the-arts on 26 out the 38 downstream tasks. We hope that our work paves the way for an era where foundational GNNs drive pharmaceutical drug discovery."
                    }
                ]
            },
            "565ff6d6-f2eb-4b99-9440-6d5401352aba": {
                "pk": "565ff6d6-f2eb-4b99-9440-6d5401352aba",
                "name": "Frederik Wenkel",
                "collaborators": [
                    "Guy Wolf",
                    "Michael Perlmutter",
                    "Yimeng Min",
                    "Smita Krishnaswamy",
                    "Matthew Hirn",
                    "Semih Cant\u00fcrk",
                    "Alexander Tong",
                    "Dhananjay Bhaskar",
                    "Maciej Sypetkowski",
                    "Farimah Poursafaei"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Geometric Deep Learning",
                    "Combinatorial Optimization",
                    "Dynamic Systems"
                ],
                "publications": [
                    {
                        "title": "Geometric Scattering Attention Networks",
                        "abstract": "Geometric scattering has recently gained recognition in graph representation learning, and recent work has shown that integrating scattering features in graph convolution networks (GCNs) can alleviate the typical oversmoothing of features in node representation learning. However, scattering often relies on handcrafted design, requiring careful selection of frequency bands via a cascade of wavelet transforms, as well as an effective weight sharing scheme to combine low- and band-pass information. Here, we introduce a new attention-based architecture to produce adaptive task-driven node representations by implicitly learning node-wise weights for combining multiple scattering and GCN channels in the network. We show the resulting geometric scattering attention network (GSAN) outperforms previous networks in semi-supervised node classification, while also enabling a spectral study of extracted information by examining node-wise attention weights."
                    },
                    {
                        "title": "Can Hybrid Geometric Scattering Networks Help Solve the Maximum Clique Problem?",
                        "abstract": "We propose a geometric scattering-based graph neural network (GNN) for approximating solutions of the NP-hard maximum clique (MC) problem. We construct a loss function with two terms, one which encourages the network to find highly connected nodes and the other which acts as a surrogate for the constraint that the nodes form a clique. We then use this loss to train an efficient GNN architecture that outputs a vector representing the probability for each node to be part of the MC and apply a rule-based decoder to make our final prediction. The incorporation of the scattering transform alleviates the so-called oversmoothing problem that is often encountered in GNNs and would degrade the performance of our proposed setup. Our empirical results demonstrate that our method outperforms representative GNN baselines in terms of solution accuracy and inference speed as well as conventional solvers like Gurobi with limited time budgets. Furthermore, our scattering model is very parameter efficient with only $\\sim$ 0.1\\% of the number of parameters compared to previous GNN baseline models."
                    },
                    {
                        "title": "Overcoming Oversmoothness in Graph Convolutional Networks via Hybrid Scattering Networks",
                        "abstract": "Geometric deep learning has made great strides towards generalizing the design of structure-aware neural networks from traditional domains to non-Euclidean ones, giving rise to graph neural networks (GNN) that can be applied to graph-structured data arising in, e.g., social networks, biochemistry, and material science. Graph convolutional networks (GCNs) in particular, inspired by their Euclidean counterparts, have been successful in processing graph data by extracting structure-aware features. However, current GNN models are often constrained by various phenomena that limit their expressive power and ability to generalize to more complex graph datasets. Most models essentially rely on low-pass filtering of graph signals via local averaging operations, leading to oversmoothing. Moreover, to avoid severe oversmoothing, most popular GCN-style networks tend to be shallow, with narrow receptive fields, leading to underreaching. Here, we propose a hybrid GNN framework that combines traditional GCN filters with band-pass filters defined via geometric scattering. We further introduce an attention framework that allows the model to locally attend over combined information from different filters at the node level. Our theoretical results establish the complementary benefits of the scattering filters to leverage structural information from the graph, while our experiments show the benefits of our method on various learning tasks."
                    },
                    {
                        "title": "Towards a General GNN Framework for Combinatorial Optimization",
                        "abstract": "Graph neural networks (GNNs) have achieved great success for a variety of tasks such as node classification, graph classification, and link prediction. However, the use of GNNs (and machine learning more generally) to solve combinatorial optimization (CO) problems is much less explored. Here, we introduce a novel GNN architecture which leverages a complex filter bank and localized attention mechanisms designed to solve CO problems on graphs. We show how our method differentiates itself from prior GNN-based CO solvers and how it can be effectively applied to the maximum clique, minimum dominating set, and maximum cut problems in a self-supervised learning setting. In addition to demonstrating competitive overall performance across all tasks, we establish state-of-the-art results for the max cut problem."
                    },
                    {
                        "title": "On the Scalability of GNNs for Molecular Graphs",
                        "abstract": "Scaling deep learning models has been at the heart of recent revolutions in language modelling and image generation. Practitioners have observed a strong relationship between model size, dataset size, and performance. However, structure-based architectures such as Graph Neural Networks (GNNs) are yet to show the benefits of scale mainly due to the lower efficiency of sparse operations, large data requirements, and lack of clarity about the effectiveness of various architectures. We address this drawback of GNNs by studying their scaling behavior. Specifically, we analyze message-passing networks, graph Transformers, and hybrid architectures on the largest public collection of 2D molecular graphs. For the first time, we observe that GNNs benefit tremendously from the increasing scale of depth, width, number of molecules, number of labels, and the diversity in the pretraining datasets. We further demonstrate strong finetuning scaling behavior on 38 highly competitive downstream tasks, outclassing previous large models. This gives rise to MolGPS, a new graph foundation model that allows to navigate the chemical space, outperforming the previous state-of-the-arts on 26 out the 38 downstream tasks. We hope that our work paves the way for an era where foundational GNNs drive pharmaceutical drug discovery."
                    },
                    {
                        "title": "Data-Driven Learning of Geometric Scattering Networks",
                        "abstract": "We propose a new graph neural network (GNN) module, based on relaxations of recently proposed geometric scattering transforms, which consist of a cascade of graph wavelet filters. Our learnable geometric scattering (LEGS) module enables adaptive tuning of the wavelets to encourage band-pass features to emerge in learned representations. The incorporation of our LEGS-module in GNNs enables the learning of longer-range graph relations compared to many popular GNNs, which often rely on encoding graph structure via smoothness or similarity between neighbors. Further, its wavelet priors result in simplified architectures with significantly fewer learned parameters compared to competing GNNs. We demonstrate the predictive performance of LEGS-based networks on graph classification benchmarks, as well as the descriptive quality of their learned features in biochemical graph data exploration tasks."
                    },
                    {
                        "title": "Scattering GCN: Overcoming Oversmoothness in Graph Convolutional Networks",
                        "abstract": "Graph convolutional networks (GCNs) have shown promising results in processing graph data by extracting structure-aware features. This gave rise to extensive work in geometric deep learning, focusing on designing network architectures that ensure neuron activations conform to regularity patterns within the input graph. However, in most cases the graph structure is only accounted for by considering the similarity of activations between adjacent nodes, which limits the capabilities of such methods to discriminate between nodes in a graph. Here, we propose to augment conventional GCNs with geometric scattering transforms and residual convolutions. The former enables band-pass filtering of graph signals, thus alleviating the so-called oversmoothing often encountered in GCNs, while the latter is introduced to clear the resulting features of high-frequency noise. We establish the advantages of the presented Scattering GCN with both theoretical results establishing the complementary benefits of scattering and GCN features, as well as experimental results showing the benefits of our method compared to leading graph neural networks for semi-supervised node classification, including the recently proposed GAT network that typically alleviates oversmoothing using graph attention mechanisms."
                    },
                    {
                        "title": "Learnable Filters for Geometric Scattering Modules",
                        "abstract": "We propose a new graph neural network (GNN) module, based on relaxations of recently proposed geometric scattering transforms, which consist of a cascade of graph wavelet filters. Our learnable geometric scattering (LEGS) module enables adaptive tuning of the wavelets to encourage band-pass features to emerge in learned representations. The incorporation of our LEGS-module in GNNs enables the learning of longer-range graph relations compared to many popular GNNs, which often rely on encoding graph structure via smoothness or similarity between neighbors. Further, its wavelet priors result in simplified architectures with significantly fewer learned parameters compared to competing GNNs. We demonstrate the predictive performance of LEGS-based networks on graph classification benchmarks, as well as the descriptive quality of their learned features in biochemical graph data exploration tasks. Our results show that LEGS-based networks match or outperforms popular GNNs, as well as the original geometric scattering construction, on many datasets, in particular in biochemical domains, while retaining certain mathematical properties of handcrafted (non-learned) geometric scattering."
                    },
                    {
                        "title": "Inferring dynamic regulatory interaction graphs from time series data with perturbations",
                        "abstract": "Complex systems are characterized by intricate interactions between entities that evolve dynamically over time. Accurate inference of these dynamic relationships is crucial for understanding and predicting system behavior. In this paper, we propose Regulatory Temporal Interaction Network Inference (RiTINI) for inferring time-varying interaction graphs in complex systems using a novel combination of space-and-time graph attentions and graph neural ordinary differential equations (ODEs). RiTINI leverages time-lapse signals on a graph prior, as well as perturbations of signals at various nodes in order to effectively capture the dynamics of the underlying system. This approach is distinct from traditional causal inference networks, which are limited to inferring acyclic and static graphs. In contrast, RiTINI can infer cyclic, directed, and time-varying graphs, providing a more comprehensive and accurate representation of complex systems. The graph attention mechanism in RiTINI allows the model to adaptively focus on the most relevant interactions in time and space, while the graph neural ODEs enable continuous-time modeling of the system's dynamics. We evaluate RiTINI's performance on various simulated and real-world datasets, demonstrating its state-of-the-art capability in inferring interaction graphs compared to previous methods."
                    },
                    {
                        "title": "Towards a Taxonomy of Graph Learning Datasets",
                        "abstract": "Graph neural networks (GNNs) have attracted much attention due to their ability to leverage the intrinsic geometries of the underlying data. Although many different types of GNN models have been developed, with many benchmarking procedures to demonstrate the superiority of one GNN model over the others, there is a lack of systematic understanding of the underlying benchmarking datasets, and what aspects of the model are being tested. Here, we provide a principled approach to taxonomize graph benchmarking datasets by carefully designing a collection of graph perturbations to probe the essential data characteristics that GNN models leverage to perform predictions. Our data-driven taxonomization of graph datasets provides a new understanding of critical dataset characteristics that will enable better model evaluation and the development of more specialized GNN models."
                    }
                ]
            },
            "d1adb296-bc57-45cd-bfcc-2216432e5b13": {
                "pk": "d1adb296-bc57-45cd-bfcc-2216432e5b13",
                "name": "Ali Bashashati",
                "collaborators": [
                    "Hossein Farahani",
                    "Ramin Nakhli",
                    "Amirali Darbandsari",
                    "Ali Khajegili Mirabadi",
                    "Puria Azadi Moghadam",
                    "C. Blake Gilks",
                    "Allen Zhang",
                    "Mayur Mallya",
                    "Sanne Van Dalen",
                    "Karina C. Martin"
                ],
                "domain": [
                    "Histopathology",
                    "Deep Learning",
                    "Cancer Diagnosis",
                    "Self-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Benchmarking Histopathology Foundation Models for Ovarian Cancer Bevacizumab Treatment Response Prediction from Whole Slide Images",
                        "abstract": "Bevacizumab is a widely studied targeted therapeutic drug used in conjunction with standard chemotherapy for the treatment of recurrent ovarian cancer. While its administration has shown to increase the progression-free survival (PFS) in patients with advanced stage ovarian cancer, the lack of identifiable biomarkers for predicting patient response has been a major roadblock in its effective adoption towards personalized medicine. In this work, we leverage the latest histopathology foundation models trained on large-scale whole slide image (WSI) datasets to extract ovarian tumor tissue features for predicting bevacizumab response from WSIs. Our extensive experiments across a combination of different histopathology foundation models and multiple instance learning (MIL) strategies demonstrate capability of these large models in predicting bevacizumab response in ovarian cancer patients with the best models achieving an AUC score of 0.86 and an accuracy score of 72.5%. Furthermore, our survival models are able to stratify high- and low-risk cases with statistical significance (p < 0.05) even among the patients with the aggressive subtype of high-grade serous ovarian carcinoma. This work highlights the utility of histopathology foundation models for the task of ovarian bevacizumab response prediction from WSIs. The high-attention regions of the WSIs highlighted by these models not only aid the model explainability but also serve as promising imaging biomarkers for treatment prognosis."
                    },
                    {
                        "title": "CCRL: Contrastive Cell Representation Learning",
                        "abstract": "Cell identification within the H&E slides is an essential prerequisite that can pave the way towards further pathology analyses including tissue classification, cancer grading, and phenotype prediction. However, performing such a task using deep learning techniques requires a large cell-level annotated dataset. Although previous studies have investigated the performance of contrastive self-supervised methods in tissue classification, the utility of this class of algorithms in cell identification and clustering is still unknown. In this work, we investigated the utility of Self-Supervised Learning (SSL) in cell clustering by proposing the Contrastive Cell Representation Learning (CCRL) model. Through comprehensive comparisons, we show that this model can outperform all currently available cell clustering models by a large margin across two datasets from different tissue types. More interestingly, the results show that our proposed model worked well with a few number of cell categories while the utility of SSL models has been mainly shown in the context of natural image datasets with large numbers of classes (e.g., ImageNet). The unsupervised representation learning approach proposed in this research eliminates the time-consuming step of data annotation in cell classification tasks, which enables us to train our model on a much larger dataset compared to previous methods. Therefore, considering the promising outcome, this approach can open a new avenue to automatic cell representation learning."
                    },
                    {
                        "title": "A Morphology Focused Diffusion Probabilistic Model for Synthesis of Histopathology Images",
                        "abstract": "Visual microscopic study of diseased tissue by pathologists has been the cornerstone for cancer diagnosis and prognostication for more than a century. Recently, deep learning methods have made significant advances in the analysis and classification of tissue images. However, there has been limited work on the utility of such models in generating histopathology images. These synthetic images have several applications in pathology including utilities in education, proficiency testing, privacy, and data sharing. Recently, diffusion probabilistic models were introduced to generate high quality images. Here, for the first time, we investigate the potential use of such models along with prioritized morphology weighting and color normalization to synthesize high quality histopathology images of brain cancer. Our detailed results show that diffusion probabilistic models are capable of synthesizing a wide range of histopathology images and have superior performance compared to generative adversarial networks."
                    },
                    {
                        "title": "Multi-Scale Relational Graph Convolutional Network for Multiple Instance Learning in Histopathology Images",
                        "abstract": "Graph convolutional neural networks have shown significant potential in natural and histopathology images. However, their use has only been studied in a single magnification or multi-magnification with late fusion. In order to leverage the multi-magnification information and early fusion with graph convolutional networks, we handle different embedding spaces at each magnification by introducing the Multi-Scale Relational Graph Convolutional Network (MS-RGCN) as a multiple instance learning method. We model histopathology image patches and their relation with neighboring patches and patches at other scales (i.e., magnifications) as a graph. To pass the information between different magnification embedding spaces, we define separate message-passing neural networks based on the node and edge type. We experiment on prostate cancer histopathology images to predict the grade groups based on the extracted features from patches. We also compare our MS-RGCN with multiple state-of-the-art methods with evaluations on several source and held-out datasets. Our method outperforms the state-of-the-art on all of the datasets and image types consisting of tissue microarrays, whole-mount slide regions, and whole-slide images. Through an ablation study, we test and show the value of the pertinent design features of the MS-RGCN."
                    },
                    {
                        "title": "Classification of Epithelial Ovarian Carcinoma Whole-Slide Pathology Images Using Deep Transfer Learning",
                        "abstract": "Ovarian cancer is the most lethal cancer of the female reproductive organs. There are $5$ major histological subtypes of epithelial ovarian cancer, each with distinct morphological, genetic, and clinical features. Currently, these histotypes are determined by a pathologist's microscopic examination of tumor whole-slide images (WSI). This process has been hampered by poor inter-observer agreement (Cohen's kappa $0.54$-$0.67$). We utilized a \\textit{two}-stage deep transfer learning algorithm based on convolutional neural networks (CNN) and progressive resizing for automatic classification of epithelial ovarian carcinoma WSIs. The proposed algorithm achieved a mean accuracy of $87.54\\%$ and Cohen's kappa of $0.8106$ in the slide-level classification of $305$ WSIs; performing better than a standard CNN and pathologists without gynecology-specific training."
                    },
                    {
                        "title": "AMIGO: Sparse Multi-Modal Graph Transformer with Shared-Context Processing for Representation Learning of Giga-pixel Images",
                        "abstract": "Processing giga-pixel whole slide histopathology images (WSI) is a computationally expensive task. Multiple instance learning (MIL) has become the conventional approach to process WSIs, in which these images are split into smaller patches for further processing. However, MIL-based techniques ignore explicit information about the individual cells within a patch. In this paper, by defining the novel concept of shared-context processing, we designed a multi-modal Graph Transformer (AMIGO) that uses the celluar graph within the tissue to provide a single representation for a patient while taking advantage of the hierarchical structure of the tissue, enabling a dynamic focus between cell-level and tissue-level information. We benchmarked the performance of our model against multiple state-of-the-art methods in survival prediction and showed that ours can significantly outperform all of them including hierarchical Vision Transformer (ViT). More importantly, we show that our model is strongly robust to missing information to an extent that it can achieve the same performance with as low as 20% of the data. Finally, in two different cancer datasets, we demonstrated that our model was able to stratify the patients into low-risk and high-risk groups while other state-of-the-art methods failed to achieve this goal. We also publish a large dataset of immunohistochemistry images (InUIT) containing 1,600 tissue microarray (TMA) cores from 188 patients along with their survival information, making it one of the largest publicly available datasets in this context."
                    },
                    {
                        "title": "GRASP: GRAph-Structured Pyramidal Whole Slide Image Representation",
                        "abstract": "Cancer subtyping is one of the most challenging tasks in digital pathology, where Multiple Instance Learning (MIL) by processing gigapixel whole slide images (WSIs) has been in the spotlight of recent research. However, MIL approaches do not take advantage of inter- and intra-magnification information contained in WSIs. In this work, we present GRASP, a novel graph-structured multi-magnification framework for processing WSIs in digital pathology. Our approach is designed to dynamically emulate the pathologist's behavior in handling WSIs and benefits from the hierarchical structure of WSIs. GRASP, which introduces a convergence-based node aggregation instead of traditional pooling mechanisms, outperforms state-of-the-art methods over two distinct cancer datasets by a margin of up to 10% balanced accuracy, while being 7 times smaller than the closest-performing state-of-the-art model in terms of the number of parameters. Our results show that GRASP is dynamic in finding and consulting with different magnifications for subtyping cancers and is reliable and stable across different hyperparameters. The model's behavior has been evaluated by two expert pathologists confirming the interpretability of the model's dynamic. We also provide a theoretical foundation, along with empirical evidence, for our work, explaining how GRASP interacts with different magnifications and nodes in the graph to make predictions. We believe that the strong characteristics yet simple structure of GRASP will encourage the development of interpretable, structure-based designs for WSI representation in digital pathology. Furthermore, we publish two large graph datasets of rare Ovarian and Bladder cancers to contribute to the field."
                    },
                    {
                        "title": "VOLTA: an Environment-Aware Contrastive Cell Representation Learning for Histopathology",
                        "abstract": "In clinical practice, many diagnosis tasks rely on the identification of cells in histopathology images. While supervised machine learning techniques require labels, providing manual cell annotations is time-consuming due to the large number of cells. In this paper, we propose a self-supervised framework (VOLTA) for cell representation learning in histopathology images using a novel technique that accounts for the cell's mutual relationship with its environment for improved cell representations. We subjected our model to extensive experiments on the data collected from multiple institutions around the world comprising of over 700,000 cells, four cancer types, and cell types ranging from three to six categories for each dataset. The results show that our model outperforms the state-of-the-art models in cell representation learning. To showcase the potential power of our proposed framework, we applied VOLTA to ovarian and endometrial cancers with very small sample sizes (10-20 samples) and demonstrated that our cell representations can be utilized to identify the known histotypes of ovarian cancer and provide novel insights that link histopathology and molecular subtypes of endometrial cancer. Unlike supervised deep learning models that require large sample sizes for training, we provide a framework that can empower new discoveries without any annotation data in situations where sample sizes are limited."
                    }
                ]
            },
            "9f9ea841-9910-443f-b21b-7bf7adce7fea": {
                "pk": "9f9ea841-9910-443f-b21b-7bf7adce7fea",
                "name": "Maciej Sypetkowski",
                "collaborators": [
                    "Oren Kraus",
                    "Saber Saberian",
                    "Ben Mabey",
                    "Berton Earnshaw",
                    "Dominique Beaini",
                    "Kian Kenyon-Dean",
                    "Maryam Fallah",
                    "Peter McLean",
                    "Jess Leung",
                    "Vasudev Sharma"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Self-Supervised Learning",
                    "Graph Neural Network",
                    "Biological Data Analysis"
                ],
                "publications": [
                    {
                        "title": "Spatial Latent Representations in Generative Adversarial Networks for Image Generation",
                        "abstract": "In the majority of GAN architectures, the latent space is defined as a set of vectors of given dimensionality. Such representations are not easily interpretable and do not capture spatial information of image content directly. In this work, we define a family of spatial latent spaces for StyleGAN2, capable of capturing more details and representing images that are out-of-sample in terms of the number and arrangement of object parts, such as an image of multiple faces or a face with more than two eyes. We propose a method for encoding images into our spaces, together with an attribute model capable of performing attribute editing in these spaces. We show that our spaces are effective for image manipulation and encode semantic information well. Our approach can be used on pre-trained generator models, and attribute edition can be done using pre-generated direction vectors making the barrier to entry for experimentation and use extremely low. We propose a regularization method for optimizing latent representations, which equalizes distributions of parts of latent spaces, making representations much closer to generated ones. We use it for encoding images into spatial spaces to obtain significant improvement in quality while keeping semantics and ability to use our attribute model for edition purposes. In total, using our methods gives encoding quality boost even as high as 30% in terms of LPIPS score comparing to standard methods, while keeping semantics. Additionally, we propose a StyleGAN2 training procedure on our spatial latent spaces, together with a custom spatial latent representation distribution to make spatially closer elements in the representation more dependent on each other than farther elements. Such approach improves the FID score by 29% on SpaceNet, and is able to generate consistent images of arbitrary sizes on spatially homogeneous datasets, like satellite imagery."
                    },
                    {
                        "title": "Augmentation Inside the Network",
                        "abstract": "In this paper, we present augmentation inside the network, a method that simulates data augmentation techniques for computer vision problems on intermediate features of a convolutional neural network. We perform these transformations, changing the data flow through the network, and sharing common computations when it is possible. Our method allows us to obtain smoother speed-accuracy trade-off adjustment and achieves better results than using standard test-time augmentation (TTA) techniques. Additionally, our approach can improve model performance even further when coupled with test-time augmentation. We validate our method on the ImageNet-2012 and CIFAR-100 datasets for image classification. We propose a modification that is 30% faster than the flip test-time augmentation and achieves the same results for CIFAR-100."
                    },
                    {
                        "title": "On the Scalability of GNNs for Molecular Graphs",
                        "abstract": "Scaling deep learning models has been at the heart of recent revolutions in language modelling and image generation. Practitioners have observed a strong relationship between model size, dataset size, and performance. However, structure-based architectures such as Graph Neural Networks (GNNs) are yet to show the benefits of scale mainly due to the lower efficiency of sparse operations, large data requirements, and lack of clarity about the effectiveness of various architectures. We address this drawback of GNNs by studying their scaling behavior. Specifically, we analyze message-passing networks, graph Transformers, and hybrid architectures on the largest public collection of 2D molecular graphs. For the first time, we observe that GNNs benefit tremendously from the increasing scale of depth, width, number of molecules, number of labels, and the diversity in the pretraining datasets. We further demonstrate strong finetuning scaling behavior on 38 highly competitive downstream tasks, outclassing previous large models. This gives rise to MolGPS, a new graph foundation model that allows to navigate the chemical space, outperforming the previous state-of-the-arts on 26 out the 38 downstream tasks. We hope that our work paves the way for an era where foundational GNNs drive pharmaceutical drug discovery."
                    },
                    {
                        "title": "Insights From the NeurIPS 2021 NetHack Challenge",
                        "abstract": "In this report, we summarize the takeaways from the first NeurIPS 2021 NetHack Challenge. Participants were tasked with developing a program or agent that can win (i.e., 'ascend' in) the popular dungeon-crawler game of NetHack by interacting with the NetHack Learning Environment (NLE), a scalable, procedurally generated, and challenging Gym environment for reinforcement learning (RL). The challenge showcased community-driven progress in AI with many diverse approaches significantly beating the previously best results on NetHack. Furthermore, it served as a direct comparison between neural (e.g., deep RL) and symbolic AI, as well as hybrid systems, demonstrating that on NetHack symbolic bots currently outperform deep RL by a large margin. Lastly, no agent got close to winning the game, illustrating NetHack's suitability as a long-term benchmark for AI research."
                    },
                    {
                        "title": "Masked Autoencoders are Scalable Learners of Cellular Morphology",
                        "abstract": "Inferring biological relationships from cellular phenotypes in high-content microscopy screens provides significant opportunity and challenge in biological research. Prior results have shown that deep vision models can capture biological signal better than hand-crafted features. This work explores how self-supervised deep learning approaches scale when training larger models on larger microscopy datasets. Our results show that both CNN- and ViT-based masked autoencoders significantly outperform weakly supervised baselines. At the high-end of our scale, a ViT-L/8 trained on over 3.5-billion unique crops sampled from 93-million microscopy images achieves relative improvements as high as 28% over our best weakly supervised baseline at inferring known biological relationships curated from public databases. Relevant code and select models released with this work can be found at: https://github.com/recursionpharma/maes_microscopy."
                    },
                    {
                        "title": "RxRx1: A Dataset for Evaluating Experimental Batch Correction Methods",
                        "abstract": "High-throughput screening techniques are commonly used to obtain large quantities of data in many fields of biology. It is well known that artifacts arising from variability in the technical execution of different experimental batches within such screens confound these observations and can lead to invalid biological conclusions. It is therefore necessary to account for these batch effects when analyzing outcomes. In this paper we describe RxRx1, a biological dataset designed specifically for the systematic study of batch effect correction methods. The dataset consists of 125,510 high-resolution fluorescence microscopy images of human cells under 1,138 genetic perturbations in 51 experimental batches across 4 cell types. Visual inspection of the images alone clearly demonstrates significant batch effects. We propose a classification task designed to evaluate the effectiveness of experimental batch correction methods on these images and examine the performance of a number of correction methods on this task. Our goal in releasing RxRx1 is to encourage the development of effective experimental batch correction methods that generalize well to unseen experimental batches. The dataset can be downloaded at https://rxrx.ai."
                    },
                    {
                        "title": "Masked Autoencoders for Microscopy are Scalable Learners of Cellular Biology",
                        "abstract": "Featurizing microscopy images for use in biological research remains a significant challenge, especially for large-scale experiments spanning millions of images. This work explores the scaling properties of weakly supervised classifiers and self-supervised masked autoencoders (MAEs) when training with increasingly larger model backbones and microscopy datasets. Our results show that ViT-based MAEs outperform weakly supervised classifiers on a variety of tasks, achieving as much as a 11.5% relative improvement when recalling known biological relationships curated from public databases. Additionally, we develop a new channel-agnostic MAE architecture (CA-MAE) that allows for inputting images of different numbers and orders of channels at inference time. We demonstrate that CA-MAEs effectively generalize by inferring and evaluating on a microscopy image dataset (JUMP-CP) generated under different experimental conditions with a different channel structure than our pretraining data (RPI-93M). Our findings motivate continued research into scaling self-supervised learning on microscopy data in order to create powerful foundation models of cellular biology that have the potential to catalyze advancements in drug discovery and beyond."
                    }
                ]
            },
            "10cfd3ca-1705-405d-ab38-0cd78871107a": {
                "pk": "10cfd3ca-1705-405d-ab38-0cd78871107a",
                "name": "Dominique Beaini",
                "collaborators": [
                    "Sofiane Achiche",
                    "Maxime Raison",
                    "Prudencio Tossou",
                    "Pietro Li\u00f2",
                    "Ladislav Ramp\u00e1\u0161ek",
                    "Vincent L\u00e9tourneau",
                    "Gabriele Corso",
                    "Guy Wolf",
                    "Fabrice Nonez",
                    "Emmanuel Noutahi"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Computer Vision",
                    "Machine Learning",
                    "Molecular Property Prediction"
                ],
                "publications": [
                    {
                        "title": "Improving Convolutional Neural Networks Via Conservative Field Regularisation and Integration",
                        "abstract": "Current research in convolutional neural networks (CNN) focuses mainly on changing the architecture of the networks, optimizing the hyper-parameters and improving the gradient descent. However, most work use only 3 standard families of operations inside the CNN, the convolution, the activation function, and the pooling. In this work, we propose a new family of operations based on the Green's function of the Laplacian, which allows the network to solve the Laplacian, to integrate any vector field and to regularize the field by forcing it to be conservative. Hence, the Green's function (GF) is the first operation that regularizes the 2D or 3D feature space by forcing it to be conservative and physically interpretable, instead of regularizing the norm of the weights. Our results show that such regularization allows the network to learn faster, to have smoother training curves and to better generalize, without any additional parameter. The current manuscript presents early results, more work is required to benchmark the proposed method."
                    },
                    {
                        "title": "Saliency Enhancement using Gradient Domain Edges Merging",
                        "abstract": "In recent years, there has been a rapid progress in solving the binary problems in computer vision, such as edge detection which finds the boundaries of an image and salient object detection which finds the important object in an image. This progress happened thanks to the rise of deep-learning and convolutional neural networks (CNN) which allow to extract complex and abstract features. However, edge detection and saliency are still two different fields and do not interact together, although it is intuitive for a human to detect salient objects based on its boundaries. Those features are not well merged in a CNN because edges and surfaces do not intersect since one feature represents a region while the other represents boundaries between different regions. In the current work, the main objective is to develop a method to merge the edges with the saliency maps to improve the performance of the saliency. Hence, we developed the gradient-domain merging (GDM) which can be used to quickly combine the image-domain information of salient object detection with the gradient-domain information of the edge detection. This leads to our proposed saliency enhancement using edges (SEE) with an average improvement of the F-measure of at least 3.4 times higher on the DUT-OMRON dataset and 6.6 times higher on the ECSSD dataset, when compared to competing algorithm such as denseCRF and BGOF. The SEE algorithm is split into 2 parts, SEE-Pre for preprocessing and SEE-Post pour postprocessing."
                    },
                    {
                        "title": "Novel Convolution Kernels for Computer Vision and Shape Analysis based on Electromagnetism",
                        "abstract": "Computer vision is a growing field with a lot of new applications in automation and robotics, since it allows the analysis of images and shapes for the generation of numerical or analytical information. One of the most used method of information extraction is image filtering through convolution kernels, with each kernel specialized for specific applications. The objective of this paper is to present a novel convolution kernels, based on principles of electromagnetic potentials and fields, for a general use in computer vision and to demonstrate its usage for shape and stroke analysis. Such filtering possesses unique geometrical properties that can be interpreted using well understood physics theorems. Therefore, this paper focuses on the development of the electromagnetic kernels and on their application on images for shape and stroke analysis. It also presents several interesting features of electromagnetic kernels, such as resolution, size and orientation independence, robustness to noise and deformation, long distance stroke interaction and ability to work with 3D images"
                    },
                    {
                        "title": "Computing the Spatial Probability of Inclusion inside Partial Contours for Computer Vision Applications",
                        "abstract": "In Computer Vision, edge detection is one of the favored approaches for feature and object detection in images since it provides information about their objects boundaries. Other region-based approaches use probabilistic analysis such as clustering and Markov random fields, but those methods cannot be used to analyze edges and their interaction. In fact, only image segmentation can produce regions based on edges, but it requires thresholding by simply separating the regions into binary in-out information. Hence, there is currently a gap between edge-based and region-based algorithms, since edges cannot be used to study the properties of a region and vice versa. The objective of this paper is to present a novel spatial probability analysis that allows determining the probability of inclusion inside a set of partial contours (strokes). To answer this objective, we developed a new approach that uses electromagnetic convolutions and repulsion optimization to compute the required probabilities. Hence, it becomes possible to generate a continuous space of probability based only on the edge information, thus bridging the gap between the edge-based methods and the region-based methods. The developed method is consistent with the fundamental properties of inclusion probabilities and its results are validated by comparing an image with the probability-based estimation given by our algorithm. The method can also be generalized to take into consideration the intensity of the edges or to be used for 3D shapes. This is the first documented method that allows computing a space of probability based on interacting edges, which opens the path to broader applications such as image segmentation and contour completion."
                    },
                    {
                        "title": "Deep Green Function Convolution for Improving Saliency in Convolutional Neural Networks",
                        "abstract": "Current saliency methods require to learn large scale regional features using small convolutional kernels, which is not possible with a simple feed-forward network. Some methods solve this problem by using segmentation into superpixels while others downscale the image through the network and rescale it back to its original size. The objective of this paper is to show that saliency convolutional neural networks (CNN) can be improved by using a Green's function convolution (GFC) to extrapolate edges features into salient regions. The GFC acts as a gradient integrator, allowing to produce saliency features by filling thin edges directly inside the CNN. Hence, we propose the gradient integration and sum (GIS) layer that combines the edges features with the saliency features. Using the HED and DSS architecture, we demonstrated that adding a GIS layer near the network's output allows to reduce the sensitivity to the parameter initialization, to reduce the overfitting and to improve the repeatability of the training. By simply adding a GIS layer to the state-of-the-art DSS model, there is an absolute increase of 1.6% for the F-measure on the DUT-OMRON dataset, with only 10ms of additional computation time. The GIS layer further allows the network to perform significantly better in the case of highly noisy images or low-brightness images. In fact, we observed an F-measure improvement of 5.2% when noise was added to the dataset and 2.8% when the brightness was reduced. Since the GIS layer is model agnostic, it can be implemented into different fully convolutional networks. A major contribution of the current work is the first implementation of Green's function convolution inside a neural network, which allows the network to operate in the feature domain and in the gradient domain at the same time, thus improving the regional representation via edge filling."
                    },
                    {
                        "title": "Towards Interpretable Sparse Graph Representation Learning with Laplacian Pooling",
                        "abstract": "Recent work in graph neural networks (GNNs) has led to improvements in molecular activity and property prediction tasks. Unfortunately, GNNs often fail to capture the relative importance of interactions between molecular substructures, in part due to the absence of efficient intermediate pooling steps. To address these issues, we propose LaPool (Laplacian Pooling), a novel, data-driven, and interpretable hierarchical graph pooling method that takes into account both node features and graph structure to improve molecular representation. We benchmark LaPool on molecular graph prediction and understanding tasks and show that it outperforms recent GNNs. Interestingly, LaPool also remains competitive on non-molecular tasks. Both quantitative and qualitative assessments are done to demonstrate LaPool's improved interpretability and highlight its potential benefits in drug design. Finally, we demonstrate LaPool's utility for the generation of valid and novel molecules by incorporating it into an adversarial autoencoder."
                    },
                    {
                        "title": "Directional Graph Networks",
                        "abstract": "The lack of anisotropic kernels in graph neural networks (GNNs) strongly limits their expressiveness, contributing to well-known issues such as over-smoothing. To overcome this limitation, we propose the first globally consistent anisotropic kernels for GNNs, allowing for graph convolutions that are defined according to topologicaly-derived directional flows. First, by defining a vector field in the graph, we develop a method of applying directional derivatives and smoothing by projecting node-specific messages into the field. Then, we propose the use of the Laplacian eigenvectors as such vector field. We show that the method generalizes CNNs on an $n$-dimensional grid and is provably more discriminative than standard GNNs regarding the Weisfeiler-Lehman 1-WL test. We evaluate our method on different standard benchmarks and see a relative error reduction of 8% on the CIFAR10 graph dataset and 11% to 32% on the molecular ZINC dataset, and a relative increase in precision of 1.6% on the MolPCBA dataset. An important outcome of this work is that it enables graph networks to embed directions in an unsupervised way, thus allowing a better representation of the anisotropic features in different physical or biological problems."
                    },
                    {
                        "title": "Task-Agnostic Graph Neural Network Evaluation via Adversarial Collaboration",
                        "abstract": "It has been increasingly demanding to develop reliable methods to evaluate the progress of Graph Neural Network (GNN) research for molecular representation learning. Existing GNN benchmarking methods for molecular representation learning focus on comparing the GNNs' performances on some node/graph classification/regression tasks on certain datasets. However, there lacks a principled, task-agnostic method to directly compare two GNNs. Additionally, most of the existing self-supervised learning works incorporate handcrafted augmentations to the data, which has several severe difficulties to be applied on graphs due to their unique characteristics. To address the aforementioned issues, we propose GraphAC (Graph Adversarial Collaboration) -- a conceptually novel, principled, task-agnostic, and stable framework for evaluating GNNs through contrastive self-supervision. We introduce a novel objective function: the Competitive Barlow Twins, that allow two GNNs to jointly update themselves from direct competitions against each other. GraphAC succeeds in distinguishing GNNs of different expressiveness across various aspects, and has demonstrated to be a principled and reliable GNN evaluation method, without necessitating any augmentations."
                    },
                    {
                        "title": "Principal Neighbourhood Aggregation for Graph Nets",
                        "abstract": "Graph Neural Networks (GNNs) have been shown to be effective models for different predictive tasks on graph-structured data. Recent work on their expressive power has focused on isomorphism tasks and countable feature spaces. We extend this theoretical framework to include continuous features - which occur regularly in real-world input domains and within the hidden layers of GNNs - and we demonstrate the requirement for multiple aggregation functions in this context. Accordingly, we propose Principal Neighbourhood Aggregation (PNA), a novel architecture combining multiple aggregators with degree-scalers (which generalize the sum aggregator). Finally, we compare the capacity of different models to capture and exploit the graph structure via a novel benchmark containing multiple tasks taken from classical graph theory, alongside existing benchmarks from real-world domains, all of which demonstrate the strength of our model. With this work, we hope to steer some of the GNN research towards new aggregation methods which we believe are essential in the search for powerful and robust models."
                    },
                    {
                        "title": "Role of Structural and Conformational Diversity for Machine Learning Potentials",
                        "abstract": "In the field of Machine Learning Interatomic Potentials (MLIPs), understanding the intricate relationship between data biases, specifically conformational and structural diversity, and model generalization is critical in improving the quality of Quantum Mechanics (QM) data generation efforts. We investigate these dynamics through two distinct experiments: a fixed budget one, where the dataset size remains constant, and a fixed molecular set one, which focuses on fixed structural diversity while varying conformational diversity. Our results reveal nuanced patterns in generalization metrics. Notably, for optimal structural and conformational generalization, a careful balance between structural and conformational diversity is required, but existing QM datasets do not meet that trade-off. Additionally, our results highlight the limitation of the MLIP models at generalizing beyond their training distribution, emphasizing the importance of defining applicability domain during model deployment. These findings provide valuable insights and guidelines for QM data generation efforts."
                    },
                    {
                        "title": "3D Infomax improves GNNs for Molecular Property Prediction",
                        "abstract": "Molecular property prediction is one of the fastest-growing applications of deep learning with critical real-world impacts. Including 3D molecular structure as input to learned models improves their performance for many molecular tasks. However, this information is infeasible to compute at the scale required by several real-world applications. We propose pre-training a model to reason about the geometry of molecules given only their 2D molecular graphs. Using methods from self-supervised learning, we maximize the mutual information between 3D summary vectors and the representations of a Graph Neural Network (GNN) such that they contain latent 3D information. During fine-tuning on molecules with unknown geometry, the GNN still generates implicit 3D information and can use it to improve downstream tasks. We show that 3D pre-training provides significant improvements for a wide range of properties, such as a 22% average MAE reduction on eight quantum mechanical properties. Moreover, the learned representations can be effectively transferred between datasets in different molecular spaces."
                    },
                    {
                        "title": "Rethinking Graph Transformers with Spectral Attention",
                        "abstract": "In recent years, the Transformer architecture has proven to be very successful in sequence processing, but its application to other data structures, such as graphs, has remained limited due to the difficulty of properly defining positions. Here, we present the $\\textit{Spectral Attention Network}$ (SAN), which uses a learned positional encoding (LPE) that can take advantage of the full Laplacian spectrum to learn the position of each node in a given graph. This LPE is then added to the node features of the graph and passed to a fully-connected Transformer. By leveraging the full spectrum of the Laplacian, our model is theoretically powerful in distinguishing graphs, and can better detect similar sub-structures from their resonance. Further, by fully connecting the graph, the Transformer does not suffer from over-squashing, an information bottleneck of most GNNs, and enables better modeling of physical phenomenons such as heat transfer and electric interaction. When tested empirically on a set of 4 standard datasets, our model performs on par or better than state-of-the-art GNNs, and outperforms any attention-based model by a wide margin, becoming the first fully-connected architecture to perform well on graph benchmarks."
                    },
                    {
                        "title": "Graph Positional and Structural Encoder",
                        "abstract": "Positional and structural encodings (PSE) enable better identifiability of nodes within a graph, rendering them essential tools for empowering modern GNNs, and in particular graph Transformers. However, designing PSEs that work optimally for all graph prediction tasks is a challenging and unsolved problem. Here, we present the Graph Positional and Structural Encoder (GPSE), the first-ever graph encoder designed to capture rich PSE representations for augmenting any GNN. GPSE learns an efficient common latent representation for multiple PSEs, and is highly transferable: The encoder trained on a particular graph dataset can be used effectively on datasets drawn from markedly different distributions and modalities. We show that across a wide range of benchmarks, GPSE-enhanced models can significantly outperform those that employ explicitly computed PSEs, and at least match their performance in others. Our results pave the way for the development of foundational pre-trained graph encoders for extracting positional and structural information, and highlight their potential as a more powerful and efficient alternative to explicitly computed PSEs and existing self-supervised pre-training approaches. Our framework and pre-trained models are publicly available at https://github.com/G-Taxonomy-Workgroup/GPSE. For convenience, GPSE has also been integrated into the PyG library to facilitate downstream applications."
                    },
                    {
                        "title": "On the Scalability of GNNs for Molecular Graphs",
                        "abstract": "Scaling deep learning models has been at the heart of recent revolutions in language modelling and image generation. Practitioners have observed a strong relationship between model size, dataset size, and performance. However, structure-based architectures such as Graph Neural Networks (GNNs) are yet to show the benefits of scale mainly due to the lower efficiency of sparse operations, large data requirements, and lack of clarity about the effectiveness of various architectures. We address this drawback of GNNs by studying their scaling behavior. Specifically, we analyze message-passing networks, graph Transformers, and hybrid architectures on the largest public collection of 2D molecular graphs. For the first time, we observe that GNNs benefit tremendously from the increasing scale of depth, width, number of molecules, number of labels, and the diversity in the pretraining datasets. We further demonstrate strong finetuning scaling behavior on 38 highly competitive downstream tasks, outclassing previous large models. This gives rise to MolGPS, a new graph foundation model that allows to navigate the chemical space, outperforming the previous state-of-the-arts on 26 out the 38 downstream tasks. We hope that our work paves the way for an era where foundational GNNs drive pharmaceutical drug discovery."
                    },
                    {
                        "title": "Fast and Optimal Laplacian Solver for Gradient-Domain Image Editing using Green Function Convolution",
                        "abstract": "In computer vision, the gradient and Laplacian of an image are used in different applications, such as edge detection, feature extraction, and seamless image cloning. Computing the gradient of an image is straightforward since numerical derivatives are available in most computer vision toolboxes. However, the reverse problem is more difficult, since computing an image from its gradient requires to solve the Laplacian equation, also called Poisson equation. Current discrete methods are either slow or require heavy parallel computing. The objective of this paper is to present a novel fast and robust method of solving the image gradient or Laplacian with minimal error, which can be used for gradient domain editing. By using a single convolution based on a numerical Green's function, the whole process is faster and straightforward to implement with different computer vision libraries. It can also be optimized on a GPU using fast Fourier transforms and can easily be generalized for an n dimension image. The tests show that, for images of resolution 801x1200, the proposed GFC can solve 100 Laplacian in parallel in around 1.0 milliseconds ms. This is orders of magnitude faster than our nearest competitor which requires 294ms for a single image. Furthermore, we prove mathematically and demonstrate empirically that the proposed method is the least error solver for gradient domain editing. The developed method is also validated with examples of Poisson blending, gradient removal, and the proposed gradient domain merging GDM. Finally, we present how the GDM can be leveraged in future works for convolutional neural networks CNN."
                    },
                    {
                        "title": "Recipe for a General, Powerful, Scalable Graph Transformer",
                        "abstract": "We propose a recipe on how to build a general, powerful, scalable (GPS) graph Transformer with linear complexity and state-of-the-art results on a diverse set of benchmarks. Graph Transformers (GTs) have gained popularity in the field of graph representation learning with a variety of recent publications but they lack a common foundation about what constitutes a good positional or structural encoding, and what differentiates them. In this paper, we summarize the different types of encodings with a clearer definition and categorize them as being $\\textit{local}$, $\\textit{global}$ or $\\textit{relative}$. The prior GTs are constrained to small graphs with a few hundred nodes, here we propose the first architecture with a complexity linear in the number of nodes and edges $O(N+E)$ by decoupling the local real-edge aggregation from the fully-connected Transformer. We argue that this decoupling does not negatively affect the expressivity, with our architecture being a universal function approximator on graphs. Our GPS recipe consists of choosing 3 main ingredients: (i) positional/structural encoding, (ii) local message-passing mechanism, and (iii) global attention mechanism. We provide a modular framework $\\textit{GraphGPS}$ that supports multiple types of encodings and that provides efficiency and scalability both in small and large graphs. We test our architecture on 16 benchmarks and show highly competitive results in all of them, show-casing the empirical benefits gained by the modularity and the combination of different strategies."
                    },
                    {
                        "title": "Long Range Graph Benchmark",
                        "abstract": "Graph Neural Networks (GNNs) that are based on the message passing (MP) paradigm generally exchange information between 1-hop neighbors to build node representations at each layer. In principle, such networks are not able to capture long-range interactions (LRI) that may be desired or necessary for learning a given task on graphs. Recently, there has been an increasing interest in development of Transformer-based methods for graphs that can consider full node connectivity beyond the original sparse structure, thus enabling the modeling of LRI. However, MP-GNNs that simply rely on 1-hop message passing often fare better in several existing graph benchmarks when combined with positional feature representations, among other innovations, hence limiting the perceived utility and ranking of Transformer-like architectures. Here, we present the Long Range Graph Benchmark (LRGB) with 5 graph learning datasets: PascalVOC-SP, COCO-SP, PCQM-Contact, Peptides-func and Peptides-struct that arguably require LRI reasoning to achieve strong performance in a given task. We benchmark both baseline GNNs and Graph Transformer networks to verify that the models which capture long-range dependencies perform significantly better on these tasks. Therefore, these datasets are suitable for benchmarking and exploration of MP-GNNs and Graph Transformer architectures that are intended to capture LRI."
                    },
                    {
                        "title": "Generating QM1B with PySCF$_{\\text{IPU}}$",
                        "abstract": "The emergence of foundation models in Computer Vision and Natural Language Processing have resulted in immense progress on downstream tasks. This progress was enabled by datasets with billions of training examples. Similar benefits are yet to be unlocked for quantum chemistry, where the potential of deep learning is constrained by comparatively small datasets with 100k to 20M training examples. These datasets are limited in size because the labels are computed using the accurate (but computationally demanding) predictions of Density Functional Theory (DFT). Notably, prior DFT datasets were created using CPU supercomputers without leveraging hardware acceleration. In this paper, we take a first step towards utilising hardware accelerators by introducing the data generator PySCF$_{\\text{IPU}}$ using Intelligence Processing Units (IPUs). This allowed us to create the dataset QM1B with one billion training examples containing 9-11 heavy atoms. We demonstrate that a simple baseline neural network (SchNet 9M) improves its performance by simply increasing the amount of training data without additional inductive biases. To encourage future researchers to use QM1B responsibly, we highlight several limitations of QM1B and emphasise the low-resolution of our DFT options, which also serves as motivation for even larger, more accurate datasets. Code and dataset are available on Github: http://github.com/graphcore-research/pyscf-ipu"
                    },
                    {
                        "title": "GPS++: An Optimised Hybrid MPNN/Transformer for Molecular Property Prediction",
                        "abstract": "This technical report presents GPS++, the first-place solution to the Open Graph Benchmark Large-Scale Challenge (OGB-LSC 2022) for the PCQM4Mv2 molecular property prediction task. Our approach implements several key principles from the prior literature. At its core our GPS++ method is a hybrid MPNN/Transformer model that incorporates 3D atom positions and an auxiliary denoising task. The effectiveness of GPS++ is demonstrated by achieving 0.0719 mean absolute error on the independent test-challenge PCQM4Mv2 split. Thanks to Graphcore IPU acceleration, GPS++ scales to deep architectures (16 layers), training at 3 minutes per epoch, and large ensemble (112 models), completing the final predictions in 1 hour 32 minutes, well under the 4 hour inference budget allocated. Our implementation is publicly available at: https://github.com/graphcore/ogb-lsc-pcqm4mv2."
                    }
                ]
            }
        }
    },
    "2402.11984": {
        "paper_data": {
            "title": "Hebbian Learning based Orthogonal Projection for Continual Learning of Spiking Neural Networks",
            "url": "http://arxiv.org/abs/2402.11984v1",
            "arxiv_id": "2402.11984",
            "authors": [
                "Mingqing Xiao",
                "Qingyan Meng",
                "Zongpeng Zhang",
                "Di He",
                "Zhouchen Lin"
            ],
            "abstract": "Neuromorphic computing with spiking neural networks is promising for energy-efficient artificial intelligence (AI) applications. However, different from humans who continually learn different tasks in a lifetime, neural network models suffer from catastrophic forgetting. How could neuronal operations solve this problem is an important question for AI and neuroscience. Many previous studies draw inspiration from observed neuroscience phenomena and propose episodic replay or synaptic metaplasticity, but they are not guaranteed to explicitly preserve knowledge for neuron populations. Other works focus on machine learning methods with more mathematical grounding, e.g., orthogonal projection on high dimensional spaces, but there is no neural correspondence for neuromorphic computing. In this work, we develop a new method with neuronal operations based on lateral connections and Hebbian learning, which can protect knowledge by projecting activity traces of neurons into an orthogonal subspace so that synaptic weight update will not interfere with old tasks. We show that Hebbian and anti-Hebbian learning on recurrent lateral connections can effectively extract the principal subspace of neural activities and enable orthogonal projection. This provides new insights into how neural circuits and Hebbian learning can help continual learning, and also how the concept of orthogonal projection can be realized in neuronal systems. Our method is also flexible to utilize arbitrary training methods based on presynaptic activities/traces. Experiments show that our method consistently solves forgetting for spiking neural networks with nearly zero forgetting under various supervised training methods with different error propagation approaches, and outperforms previous approaches under various settings. Our method can pave a solid path for building continual neuromorphic computing systems.",
            "introduction": " introduction to backpropagation (BP), feedback alignment (FA), and sign symmetric (SS) used in our work. \ud835\udc4a \ud835\udc4a\u22a4ForwardError (a) Backpropagation (BP)\ud835\udc4a \ud835\udc39ForwardError (b) Feedback Alignment (FA)\ud835\udc4aForwardError (c) Sign Symmetric (SS)sign\ud835\udc4a\u22a4 Figure 7: Illustration of different error propagation DISCUSSION F.1 D ISCUSSION ON HEBBIAN LEARNING One major component of our method is Hebbian learning. Hebbian learning is often related to memory modeling, e.g., Hopfield network (Hopfield, 1982) with Hebbian-type learning can serve as associative memory systems. In this work, we provide a new perspective on how Hebbian learning can support advanced learning abilities, which is not in an associative approach. Extraction of the principal subspace of streaming data is to some extent similar to memory but is not in an associative approach, and can well fit into our projection-based continual learning method. This work mainly focuses on Hebbian learning based on neuronal spiking activities, which can be shown to optimize synaptic weights for principal component analysis (Oja, 1982; 1989). In biological systems, spike- timing-dependent plasticity (STDP) is the often observed Hebbian learning considering the time of spikes (Caporale et al., 2008). However, STDP has not been shown to optimize an explicit objec- tive. It is interesting future work to study if the biological STDP rule can be similarly leveraged to systematically promote advanced abilities such as continual learning. A basic focus of our work is neuromorphic computing. Neuromorphic computing aims at the com- putation inspired by the structure and function of the human brain. At the hardware level, neu- romorphic chips are designed to imitate biological neurons (Akopyan et al., 2015; Davies et al., 2018; Pei et al., 2019), such as their spiking property and synaptic connections with local storage of weights for in-memory computation, for highly energy-efficient and parallel event-driven computa- tion with avoidance of frequent memory transpose (its computation architecture is expected to be different from the commonly used hardware with von Neumann architecture such as CPU or GPU). At the algorithm level, we are interested in developing RESULTS G.1 D OMAIN -CL ON5-D ATASETS To further show the scalability of our method for the domain-incremental setting on larger datasets and networks apart from PMNIST, we supplement the domain-CL",
            "references": [
                {
                    "title": "TACOS: Task Agnostic Continual Learning in Spiking Neural Networks",
                    "abstract": "Catastrophic interference, the loss of previously learned information when learning new information, remains a major challenge in machine learning. Since living organisms do not seem to suffer from this problem, researchers have taken inspiration from biology to improve memory retention in artificial intelligence systems. However, previous attempts to use bio-inspired mechanisms have typically resulted in systems that rely on task boundary information during training and/or explicit task identification during inference, information that is not available in real-world scenarios. Here, we show that neuro-inspired mechanisms such as synaptic consolidation and metaplasticity can mitigate catastrophic interference in a spiking neural network, using only synapse-local information, with no need for task awareness, and with a fixed memory size that does not need to be increased when training on new tasks. Our model, TACOS, combines neuromodulation with complex synaptic dynamics to enable new learning while protecting previous information. We evaluate TACOS on sequential image recognition tasks and demonstrate its effectiveness in reducing catastrophic interference. Our results show that TACOS outperforms existing regularization techniques in domain-incremental learning scenarios. We also report the results of an ablation study to elucidate the contribution of each neuro-inspired mechanism separately."
                },
                {
                    "title": "Towards Memory- and Time-Efficient Backpropagation for Training Spiking Neural Networks",
                    "abstract": "Spiking Neural Networks (SNNs) are promising energy-efficient models for neuromorphic computing. For training the non-differentiable SNN methods, the backpropagation through time (BPTT) with surrogate gradients (SG) method has achieved high performance. However, this method suffers from considerable memory cost and training time during training. In this paper, we propose the Spatial Learning Through Time (SLTT) method that can achieve high performance while greatly improving training efficiency compared with BPTT. First, we show that the backpropagation of SNNs through the temporal domain contributes just a little to the final calculated gradients. Thus, we propose to ignore the unimportant routes in the computational graph during backpropagation. The proposed method reduces the number of scalar multiplications and achieves a small memory occupation that is independent of the total time steps. Furthermore, we propose a variant of SLTT, called SLTT-K, that allows backpropagation only at K time steps, then the required number of scalar multiplications is further reduced and is independent of the total time steps. Experiments on both static and neuromorphic datasets demonstrate superior training efficiency and performance of our SLTT. In particular, our method achieves state-of-the-art accuracy on ImageNet, while the memory cost and training time are reduced by more than 70% and 50%, respectively, compared with BPTT. Our code is available at https://github.com/qymeng94/SLTT."
                },
                {
                    "title": "Online Training Through Time for Spiking Neural Networks",
                    "abstract": "Spiking neural networks (SNNs) are promising brain-inspired energy-efficient models. Recent progress in training methods has enabled successful deep SNNs on large-scale tasks with low latency. Particularly, backpropagation through time (BPTT) with surrogate gradients (SG) is popularly used to achieve high performance in a very small number of time steps. However, it is at the cost of large memory consumption for training, lack of theoretical clarity for optimization, and inconsistency with the online property of biological learning and rules on neuromorphic hardware. Other works connect spike representations of SNNs with equivalent artificial neural network formulation and train SNNs by gradients from equivalent mappings to ensure descent directions. But they fail to achieve low latency and are also not online. In this work, we propose online training through time (OTTT) for SNNs, which is derived from BPTT to enable forward-in-time learning by tracking presynaptic activities and leveraging instantaneous loss and gradients. Meanwhile, we theoretically analyze and prove that gradients of OTTT can provide a similar descent direction for optimization as gradients based on spike representations under both feedforward and recurrent conditions. OTTT only requires constant training memory costs agnostic to time steps, avoiding the significant memory costs of BPTT for GPU training. Furthermore, the update rule of OTTT is in the form of three-factor Hebbian learning, which could pave a path for online on-chip learning. With OTTT, it is the first time that two mainstream supervised SNN training methods, BPTT with SG and spike representation-based training, are connected, and meanwhile in a biologically plausible form. Experiments on CIFAR-10, CIFAR-100, ImageNet, and CIFAR10-DVS demonstrate the superior performance of our method on large-scale static and neuromorphic datasets in small time steps."
                },
                {
                    "title": "Training High-Performance Low-Latency Spiking Neural Networks by Differentiation on Spike Representation",
                    "abstract": "Spiking Neural Network (SNN) is a promising energy-efficient AI model when implemented on neuromorphic hardware. However, it is a challenge to efficiently train SNNs due to their non-differentiability. Most existing methods either suffer from high latency (i.e., long simulation time steps), or cannot achieve as high performance as Artificial Neural Networks (ANNs). In this paper, we propose the Differentiation on Spike Representation (DSR) method, which could achieve high performance that is competitive to ANNs yet with low latency. First, we encode the spike trains into spike representation using (weighted) firing rate coding. Based on the spike representation, we systematically derive that the spiking dynamics with common neural models can be represented as some sub-differentiable mapping. With this viewpoint, our proposed DSR method trains SNNs through gradients of the mapping and avoids the common non-differentiability problem in SNN training. Then we analyze the error when representing the specific mapping with the forward computation of the SNN. To reduce such error, we propose to train the spike threshold in each layer, and to introduce a new hyperparameter for the neural models. With these components, the DSR method can achieve state-of-the-art SNN performance with low latency on both static and neuromorphic datasets, including CIFAR-10, CIFAR-100, ImageNet, and DVS-CIFAR10."
                },
                {
                    "title": "TRGP: Trust Region Gradient Projection for Continual Learning",
                    "abstract": "Catastrophic forgetting is one of the major challenges in continual learning. To address this issue, some existing methods put restrictive constraints on the optimization space of the new task for minimizing the interference to old tasks. However, this may lead to unsatisfactory performance for the new task, especially when the new task is strongly correlated with old tasks. To tackle this challenge, we propose Trust Region Gradient Projection (TRGP) for continual learning to facilitate the forward knowledge transfer based on an efficient characterization of task correlation. Particularly, we introduce a notion of `trust region' to select the most related old tasks for the new task in a layer-wise and single-shot manner, using the norm of gradient projection onto the subspace spanned by task inputs. Then, a scaled weight projection is proposed to cleverly reuse the frozen weights of the selected old tasks in the trust region through a layer-wise scaling matrix. By jointly optimizing the scaling matrices and the model, where the model is updated along the directions orthogonal to the subspaces of old tasks, TRGP can effectively prompt knowledge transfer without forgetting. Extensive experiments show that our approach achieves significant improvement over related state-of-the-art methods."
                },
                {
                    "title": "Training Feedback Spiking Neural Networks by Implicit Differentiation on the Equilibrium State",
                    "abstract": "Spiking neural networks (SNNs) are brain-inspired models that enable energy-efficient implementation on neuromorphic hardware. However, the supervised training of SNNs remains a hard problem due to the discontinuity of the spiking neuron model. Most existing methods imitate the backpropagation framework and feedforward architectures for artificial neural networks, and use surrogate derivatives or compute gradients with respect to the spiking time to deal with the problem. These approaches either accumulate approximation errors or only propagate information limitedly through existing spikes, and usually require information propagation along time steps with large memory costs and biological implausibility. In this work, we consider feedback spiking neural networks, which are more brain-like, and propose a novel training method that does not rely on the exact reverse of the forward computation. First, we show that the average firing rates of SNNs with feedback connections would gradually evolve to an equilibrium state along time, which follows a fixed-point equation. Then by viewing the forward computation of feedback SNNs as a black-box solver for this equation, and leveraging the implicit differentiation on the equation, we can compute the gradient for parameters without considering the exact forward procedure. In this way, the forward and backward procedures are decoupled and therefore the problem of non-differentiable spiking functions is avoided. We also briefly discuss the biological plausibility of implicit differentiation, which only requires computing another equilibrium. Extensive experiments on MNIST, Fashion-MNIST, N-MNIST, CIFAR-10, and CIFAR-100 demonstrate the superior performance of our method for feedback models with fewer neurons and parameters in a small number of time steps. Our code is avaiable at https://github.com/pkuxmq/IDE-FSNN."
                },
                {
                    "title": "Gradient Projection Memory for Continual Learning",
                    "abstract": "The ability to learn continually without forgetting the past tasks is a desired attribute for artificial learning systems. Existing approaches to enable such learning in artificial neural networks usually rely on network growth, importance based weight update or replay of old data from the memory. In contrast, we propose a novel approach where a neural network learns new tasks by taking gradient steps in the orthogonal direction to the gradient subspaces deemed important for the past tasks. We find the bases of these subspaces by analyzing network representations (activations) after learning each task with Singular Value Decomposition (SVD) in a single shot manner and store them in the memory as Gradient Projection Memory (GPM). With qualitative and quantitative analyses, we show that such orthogonal gradient descent induces minimum to no interference with the past tasks, thereby mitigates forgetting. We evaluate our algorithm on diverse image classification datasets with short and long sequences of tasks and report better or on-par performance compared to the state-of-the-art approaches."
                },
                {
                    "title": "Deep Residual Learning in Spiking Neural Networks",
                    "abstract": "Deep Spiking Neural Networks (SNNs) present optimization difficulties for gradient-based approaches due to discrete binary activation and complex spatial-temporal dynamics. Considering the huge success of ResNet in deep learning, it would be natural to train deep SNNs with residual learning. Previous Spiking ResNet mimics the standard residual block in ANNs and simply replaces ReLU activation layers with spiking neurons, which suffers the degradation problem and can hardly implement residual learning. In this paper, we propose the spike-element-wise (SEW) ResNet to realize residual learning in deep SNNs. We prove that the SEW ResNet can easily implement identity mapping and overcome the vanishing/exploding gradient problems of Spiking ResNet. We evaluate our SEW ResNet on ImageNet, DVS Gesture, and CIFAR10-DVS datasets, and show that SEW ResNet outperforms the state-of-the-art directly trained SNNs in both accuracy and time-steps. Moreover, SEW ResNet can achieve higher performance by simply adding more layers, providing a simple method to train deep SNNs. To our best knowledge, this is the first time that directly training deep SNNs with more than 100 layers becomes possible. Our codes are available at https://github.com/fangwei123456/Spike-Element-Wise-ResNet."
                },
                {
                    "title": "Going Deeper With Directly-Trained Larger Spiking Neural Networks",
                    "abstract": "Spiking neural networks (SNNs) are promising in a bio-plausible coding for spatio-temporal information and event-driven signal processing, which is very suited for energy-efficient implementation in neuromorphic hardware. However, the unique working mode of SNNs makes them more difficult to train than traditional networks. Currently, there are two main routes to explore the training of deep SNNs with high performance. The first is to convert a pre-trained ANN model to its SNN version, which usually requires a long coding window for convergence and cannot exploit the spatio-temporal features during training for solving temporal tasks. The other is to directly train SNNs in the spatio-temporal domain. But due to the binary spike activity of the firing function and the problem of gradient vanishing or explosion, current methods are restricted to shallow architectures and thereby difficult in harnessing large-scale datasets (e.g. ImageNet). To this end, we propose a threshold-dependent batch normalization (tdBN) method based on the emerging spatio-temporal backpropagation, termed \u201cSTBP-tdBN\u201d, enabling direct training of a very deep SNN and the efficient implementation of its inference on neuromorphic hardware. With the proposed method and elaborated shortcut connection, we significantly extend directly-trained SNNs from a shallow structure ("
                },
                {
                    "title": "Online Spatio-Temporal Learning in Deep Neural Networks",
                    "abstract": "Biological neural networks are equipped with an inherent capability to continuously adapt through online learning. This aspect remains in stark contrast to learning with error backpropagation through time (BPTT) that involves offline computation of the gradients due to the need to unroll the network through time. Here, we present an alternative online learning algorithm ic framework for deep recurrent neural networks (RNNs) and spiking neural networks (SNNs), called online spatio-temporal learning (OSTL). It is based on insights from biology and proposes the clear separation of spatial and temporal gradient components. For shallow SNNs, OSTL is gradient equivalent to BPTT enabling for the first time online training of SNNs with BPTT-equivalent gradients. In addition, the proposed formulation unveils a class of SNN architectures trainable online at low time complexity. Moreover, we extend OSTL to a generic form, applicable to a wide range of network architectures, including networks comprising long short-term memory (LSTM) and gated recurrent units (GRUs). We demonstrate the operation of our algorithm ic framework on various tasks from language modeling to speech recognition and obtain results on par with the BPTT baselines."
                },
                {
                    "title": "ODE-Inspired Analysis for the Biological Version of Oja's Rule in Solving Streaming PCA",
                    "abstract": "Oja's rule [Oja, Journal of mathematical biology 1982] is a well-known biologically-plausible algorithm using a Hebbian-type synaptic update rule to solve streaming principal component analysis (PCA). Computational neuroscientists have known that this biological version of Oja's rule converges to the top eigenvector of the covariance matrix of the input in the limit. However, prior to this work, it was open to prove any convergence rate guarantee. \nIn this work, we give the first convergence rate analysis for the biological version of Oja's rule in solving streaming PCA. Moreover, our convergence rate matches the information theoretical lower bound up to logarithmic factors and outperforms the state-of-the-art upper bound for streaming PCA. Furthermore, we develop a novel framework inspired by ordinary differential equations (ODE) to analyze general stochastic dynamics. The framework abandons the traditional step-by-step analysis and instead analyzes a stochastic dynamic in one-shot by giving a closed-form solution to the entire dynamic. The one-shot framework allows us to apply stopping time and martingale techniques to have a flexible and precise control on the dynamic. We believe that this general framework is powerful and should lead to effective yet simple analysis for a large class of problems with stochastic dynamics."
                },
                {
                    "title": "Orthogonal Gradient Descent for Continual Learning",
                    "abstract": "Neural networks are achieving state of the art and sometimes super-human performance on learning tasks across a variety of domains. Whenever these problems require learning in a continual or sequential manner, however, neural networks suffer from the problem of catastrophic forgetting; they forget how to solve previous tasks after being trained on a new task, despite having the essential capacity to solve both tasks if they were trained on both simultaneously. In this paper, we propose to address this issue from a parameter space perspective and study an approach to restrict the direction of the gradient updates to avoid forgetting previously-learned data. We present the Orthogonal Gradient Descent (OGD) method, which accomplishes this goal by projecting the gradients from new tasks onto a subspace in which the neural network output on previous task does not change and the projected gradient is still in a useful direction for learning the new task. Our approach utilizes the high capacity of a neural network more efficiently and does not require storing the previously learned data that might raise privacy concerns. Experiments on common benchmarks reveal the effectiveness of the proposed OGD method."
                },
                {
                    "title": "Temporal-Coded Deep Spiking Neural Network with Easy Training and Robust Performance",
                    "abstract": "Spiking neural network (SNN) is promising but the development has fallen far behind conventional deep neural networks (DNNs) because of difficult training. To resolve the training problem, we analyze the closed-form input-output response of spiking neurons and use the response expression to build abstract SNN models for training. This avoids calculating membrane potential during training and makes the direct training of SNN as efficient as DNN. We show that the nonleaky integrate-and-fire neuron with single-spike temporal-coding is the best choice for direct-train deep SNNs. We develop an energy-efficient phase-domain signal processing circuit for the neuron and propose a direct-train deep SNN framework. Thanks to easy training, we train deep SNNs under weight quantizations to study their robustness over low-cost neuromorphic hardware. Experiments show that our direct-train deep SNNs have the highest CIFAR-10 classification accuracy among SNNs, achieve ImageNet classification accuracy within 1% of the DNN of equivalent architecture, and are robust to weight quantization and noise perturbation."
                },
                {
                    "title": "A Continual Learning Survey: Defying Forgetting in Classification Tasks",
                    "abstract": "Artificial neural networks thrive in solving the classification problem for a particular rigid task, acquiring knowledge through generalized learning behaviour from a distinct training phase. The resulting network resembles a static entity of knowledge, with endeavours to extend this knowledge without targeting the original task resulting in a catastrophic forgetting. Continual learning shifts this paradigm towards networks that can continually accumulate knowledge over different tasks without the need to retrain from scratch. We focus on task incremental classification, where tasks arrive sequentially and are delineated by clear boundaries. Our main contributions concern: (1) a taxonomy and extensive overview of the state-of-the-art; (2) a novel framework to continually determine the stability-plasticity trade-off of the continual learner; (3) a comprehensive experimental comparison of 11 state-of-the-art continual learning methods; and (4) baselines. We empirically scrutinize method strengths and weaknesses on three benchmarks, considering Tiny Imagenet and large-scale unbalanced iNaturalist and a sequence of recognition datasets. We study the influence of model capacity, weight decay and dropout regularization, and the order in which the tasks are presented, and qualitatively compare methods in terms of required memory, computation time, and storage."
                },
                {
                    "title": "Spike-Train Level Backpropagation for Training Deep Recurrent Spiking Neural Networks",
                    "abstract": "Spiking neural networks (SNNs) well support spatiotemporal learning and energy-efficient event-driven hardware neuromorphic processors. As an important class of SNNs, recurrent spiking neural networks (RSNNs) possess great computational power. However, the practical application of RSNNs is severely limited by challenges in training. Biologically-inspired unsupervised learning has limited capability in boosting the performance of RSNNs. On the other hand, existing backpropagation (BP) methods suffer from high complexity of unrolling in time, vanishing and exploding gradients, and approximate differentiation of discontinuous spiking activities when applied to RSNNs. To enable supervised training of RSNNs under a well-defined loss function, we present a novel Spike-Train level RSNNs Backpropagation (ST-RSBP) algorithm for training deep RSNNs. The proposed ST-RSBP directly computes the gradient of a rated-coded loss function defined at the output layer of the network w.r.t tunable parameters. The scalability of ST-RSBP is achieved by the proposed spike-train level computation during which temporal effects of the SNN is captured in both the forward and backward pass of BP. Our ST-RSBP algorithm can be broadly applied to RSNNs with a single recurrent layer or deep RSNNs with multiple feed-forward and recurrent layers. Based upon challenging speech and image datasets including TI46, N-TIDIGITS, Fashion-MNIST and MNIST, ST-RSBP is able to train RSNNs with an accuracy surpassing that of the current state-of-art SNN BP algorithms and conventional non-spiking deep learning models."
                },
                {
                    "title": "Temporal Coding in Spiking Neural Networks with Alpha Synaptic Function",
                    "abstract": "We propose a spiking neural network model that encodes information in the relative timing of individual neuron spikes and performs classification using the first output neuron to spike. This temporal coding scheme allows the supervised training of the network with backpropagation, using locally exact derivatives of the postsynaptic with respect to presynaptic spike times. The network uses a biologically-inspired alpha synaptic transfer function and trainable synchronisation pulses as temporal references. We successfully train the network on the MNIST dataset encoded in time. Our spiking neural network outperforms comparable spiking models and achieves similar accuracy to a fully connected conventional network. During training, our network displays a speed-accuracy trade-off, with either slow and highly-accurate or very fast but less accurate classification. The results demonstrate the computational power of spiking networks with biological characteristics that encode information in the timing of individual neurons. Our code is publicly available."
                },
                {
                    "title": "A Tandem Learning Rule for Effective Training and Rapid Inference of Deep Spiking Neural Networks",
                    "abstract": "Spiking neural networks (SNNs) represent the most prominent biologically inspired computing model for neuromorphic computing (NC) architectures. However, due to the nondifferentiable nature of spiking neuronal functions, the standard error backpropagation algorithm is not directly applicable to SNNs. In this work, we propose a tandem learning framework that consists of an SNN and an artificial neural network (ANN) coupled through weight sharing. The ANN is an auxiliary structure that facilitates the error backpropagation for the training of the SNN at the spike-train level. To this end, we consider the spike count as the discrete neural representation in the SNN and design an ANN neuronal activation function that can effectively approximate the spike count of the coupled SNN. The proposed tandem learning rule demonstrates competitive pattern recognition and regression capabilities on both the conventional frame- and event-based vision datasets, with at least an order of magnitude reduced inference time and total synaptic operations over other state-of-the-art SNN implementations. Therefore, the proposed tandem learning rule offers a novel solution to training efficient, low latency, and high-accuracy deep SNNs with low computing resources."
                },
                {
                    "title": "Experience Replay for Continual Learning",
                    "abstract": "Continual learning is the problem of learning new tasks or knowledge while protecting old knowledge and ideally generalizing from old experience to learn new tasks faster. Neural networks trained by stochastic gradient descent often degrade on old tasks when trained successively on new tasks with different data distributions. This phenomenon, referred to as catastrophic forgetting, is considered a major hurdle to learning with non-stationary data or sequences of new tasks, and prevents networks from continually accumulating knowledge and skills. We examine this issue in the context of reinforcement learning, in a setting where an agent is exposed to tasks in a sequence. Unlike most other work, we do not provide an explicit indication to the model of task boundaries, which is the most general circumstance for a learning agent exposed to continuous experience. While various methods to counteract catastrophic forgetting have recently been proposed, we explore a straightforward, general, and seemingly overlooked solution - that of using experience replay buffers for all past events - with a mixture of on- and off-policy learning, leveraging behavioral cloning. We show that this strategy can still learn new tasks quickly yet can substantially reduce catastrophic forgetting in both Atari and DMLab domains, even matching the performance of methods that require task identities. When buffer storage is constrained, we confirm that a simple mechanism for randomly discarding data allows a limited size buffer to perform almost as well as an unbounded one."
                },
                {
                    "title": "Biologically-plausible learning algorithms can scale to large datasets",
                    "abstract": "The backpropagation (BP) algorithm is often thought to be biologically implausible in the brain. One of the main reasons is that BP requires symmetric weight matrices in the feedforward and feedback pathways. To address this \"weight transport problem\" (Grossberg, 1987), two more biologically plausible algorithms, proposed by Liao et al. (2016) and Lillicrap et al. (2016), relax BP's weight symmetry requirements and demonstrate comparable learning capabilities to that of BP on small datasets. However, a recent study by Bartunov et al. (2018) evaluate variants of target-propagation (TP) and feedback alignment (FA) on MINIST, CIFAR, and ImageNet datasets, and find that although many of the proposed algorithms perform well on MNIST and CIFAR, they perform significantly worse than BP on ImageNet. Here, we additionally evaluate the sign-symmetry algorithm (Liao et al., 2016), which differs from both BP and FA in that the feedback and feedforward weights share signs but not magnitudes. We examine the performance of sign-symmetry and feedback alignment on ImageNet and MS COCO datasets using different network architectures (ResNet-18 and AlexNet for ImageNet, RetinaNet for MS COCO). Surprisingly, networks trained with sign-symmetry can attain classification performance approaching that of BP-trained networks. These results complement the study by Bartunov et al. (2018), and establish a new benchmark for future biologically plausible learning algorithms on more difficult datasets and more complex architectures."
                },
                {
                    "title": "Efficient Lifelong Learning with A-GEM",
                    "abstract": "In lifelong learning, the learner is presented with a sequence of tasks, incrementally building a data-driven prior which may be leveraged to speed up learning of a new task. In this work, we investigate the efficiency of current lifelong approaches, in terms of sample complexity, computational and memory cost. Towards this end, we first introduce a new and a more realistic evaluation protocol, whereby learners observe each example only once and hyper-parameter selection is done on a small and disjoint set of tasks, which is not used for the actual learning experience and evaluation. Second, we introduce a new metric measuring how quickly a learner acquires a new skill. Third, we propose an improved version of GEM (Lopez-Paz & Ranzato, 2017), dubbed Averaged GEM (A-GEM), which enjoys the same or even better performance as GEM, while being almost as computationally and memory efficient as EWC (Kirkpatrick et al., 2016) and other regularization-based methods. Finally, we show that all algorithms including A-GEM can learn even more quickly if they are provided with task descriptors specifying the classification tasks under consideration. Our experiments on several standard lifelong learning benchmarks demonstrate that A-GEM has the best trade-off between accuracy and efficiency."
                },
                {
                    "title": "SLAYER: Spike Layer Error Reassignment in Time",
                    "abstract": "Configuring deep Spiking Neural Networks (SNNs) is an exciting research avenue for low power spike event based computation. However, the spike generation function is non-differentiable and therefore not directly compatible with the standard error backpropagation algorithm. In this paper, we introduce a new general backpropagation mechanism for learning synaptic weights and axonal delays which overcomes the problem of non-differentiability of the spike function and uses a temporal credit assignment policy for backpropagating error to preceding layers. We describe and release a GPU accelerated software implementation of our method which allows training both fully connected and convolutional neural network (CNN) architectures. Using our software, we compare our method against existing SNN based learning approaches and standard ANN to SNN conversion techniques and show that our method achieves state of the art performance for an SNN on the MNIST, NMNIST, DVS Gesture, and TIDIGITS datasets."
                },
                {
                    "title": "Long short-term memory and Learning-to-learn in networks of spiking neurons",
                    "abstract": "Recurrent networks of spiking neurons (RSNNs) underlie the astounding computing and learning capabilities of the brain. But computing and learning capabilities of RSNN models have remained poor, at least in comparison with artificial neural networks (ANNs). We address two possible reasons for that. One is that RSNNs in the brain are not randomly connected or designed according to simple rules, and they do not start learning as a tabula rasa network. Rather, RSNNs in the brain were optimized for their tasks through evolution, development, and prior experience. Details of these optimization processes are largely unknown. But their functional contribution can be approximated through powerful optimization methods, such as backpropagation through time (BPTT). \nA second major mismatch between RSNNs in the brain and models is that the latter only show a small fraction of the dynamics of neurons and synapses in the brain. We include neurons in our RSNN model that reproduce one prominent dynamical process of biological neurons that takes place at the behaviourally relevant time scale of seconds: neuronal adaptation. We denote these networks as LSNNs because of their Long short-term memory. The inclusion of adapting neurons drastically increases the computing and learning capability of RSNNs if they are trained and configured by deep learning (BPTT combined with a rewiring algorithm that optimizes the network architecture). In fact, the computational performance of these RSNNs approaches for the first time that of LSTM networks. In addition RSNNs with adapting neurons can acquire abstract knowledge from prior learning in a Learning-to-Learn (L2L) scheme, and transfer that knowledge in order to learn new but related tasks from very few examples. We demonstrate this for supervised learning and reinforcement learning."
                },
                {
                    "title": "Loihi: A Neuromorphic Manycore Processor with On-Chip Learning",
                    "abstract": "Loihi is a 60-mm2 chip fabricated in Intels 14-nm process that advances the state-of-the-art modeling of spiking neural networks in silicon. It integrates a wide range of novel features for the field, such as hierarchical connectivity, dendritic compartments, synaptic delays, and, most importantly, programmable synaptic learning rules. Running a spiking convolutional form of the Locally Competitive Algorithm, Loihi can solve LASSO optimization problems with over three orders of magnitude superior energy-delay-product compared to conventional solvers running on a CPU iso-process/voltage/area. This provides an unambiguous example of spike-based computation, outperforming all known conventional solutions."
                },
                {
                    "title": "Overcoming catastrophic forgetting with hard attention to the task",
                    "abstract": "Catastrophic forgetting occurs when a neural network loses the information learned in a previous task after training on subsequent tasks. This problem remains a hurdle for artificial intelligence systems with sequential learning capabilities. In this paper, we propose a task-based hard attention mechanism that preserves previous tasks' information without affecting the current task's learning. A hard attention mask is learned concurrently to every task, through stochastic gradient descent, and previous masks are exploited to condition such learning. We show that the proposed mechanism is effective for reducing catastrophic forgetting, cutting current rates by 45 to 80%. We also show that it is robust to different hyperparameter choices, and that it offers a number of monitoring capabilities. The approach features the possibility to control both the stability and compactness of the learned knowledge, which we believe makes it also attractive for online learning or network compression applications."
                },
                {
                    "title": "PackNet: Adding Multiple Tasks to a Single Network by Iterative Pruning",
                    "abstract": "This paper presents a method for adding multiple tasks to a single deep neural network while avoiding catastrophic forgetting. Inspired by network pruning techniques, we exploit redundancies in large deep networks to free up parameters that can then be employed to learn new tasks. By performing iterative pruning and network re-training, we are able to sequentially \"pack\" multiple tasks into a single network while ensuring minimal drop in performance and minimal storage overhead. Unlike prior work that uses proxy losses to maintain accuracy on older tasks, we always optimize for the task at hand. We perform extensive experiments on a variety of network architectures and large-scale datasets, and observe much better robustness against catastrophic forgetting than prior work. In particular, we are able to add three fine-grained classification tasks to a single ImageNet-trained VGG-16 network and achieve accuracies close to those of separately trained networks for each task."
                },
                {
                    "title": "Lifelong Learning with Dynamically Expandable Networks",
                    "abstract": "We propose a novel deep network architecture for lifelong learning which we refer to as Dynamically Expandable Network (DEN), that can dynamically decide its network capacity as it trains on a sequence of tasks, to learn a compact overlapping knowledge sharing structure among tasks. DEN is efficiently trained in an online manner by performing selective retraining, dynamically expands network capacity upon arrival of each task with only the necessary number of units, and effectively prevents semantic drift by splitting/duplicating units and timestamping them. We validate DEN on multiple public datasets under lifelong learning scenarios, on which it not only significantly outperforms existing lifelong learning methods for deep networks, but also achieves the same level of performance as the batch counterparts with substantially fewer number of parameters. Further, the obtained network fine-tuned on all tasks obtained significantly better performance over the batch models, which shows that it can be used to estimate the optimal network structure even when all tasks are available in the first place."
                },
                {
                    "title": "Spatio-Temporal Backpropagation for Training High-Performance Spiking Neural Networks",
                    "abstract": "Spiking neural networks (SNNs) are promising in ascertaining brain-like behaviors since spikes are capable of encoding spatio-temporal information. Recent schemes, e.g., pre-training from artificial neural networks (ANNs) or direct training based on backpropagation (BP), make the high-performance supervised training of SNNs possible. However, these methods primarily fasten more attention on its spatial domain information, and the dynamics in temporal domain are attached less significance. Consequently, this might lead to the performance bottleneck, and scores of training techniques shall be additionally required. Another underlying problem is that the spike activity is naturally non-differentiable, raising more difficulties in supervised training of SNNs. In this paper, we propose a spatio-temporal backpropagation (STBP) algorithm for training high-performance SNNs. In order to solve the non-differentiable problem of SNNs, an approximated derivative for spike activity is proposed, being appropriate for gradient descent training. The STBP algorithm combines the layer-by-layer spatial domain (SD) and the timing-dependent temporal domain (TD), and does not require any additional complicated skill. We evaluate this method through adopting both the fully connected and convolutional architecture on the static MNIST dataset, a custom object detection dataset, and the dynamic N-MNIST dataset. Results bespeak that our approach achieves the best accuracy compared with existing state-of-the-art algorithms on spiking networks. This work provides a new perspective to investigate the high-performance SNNs for future brain-like computing paradigm with rich spatio-temporal dynamics."
                },
                {
                    "title": "Gradient Episodic Memory for Continual Learning",
                    "abstract": "One major obstacle towards AI is the poor ability of models to solve new problems quicker, and without forgetting previously acquired knowledge. To better understand this issue, we study the problem of continual learning, where the model observes, once and one by one, examples concerning a sequence of tasks. First, we propose a set of metrics to evaluate models learning over a continuum of data. These metrics characterize models not only by their test accuracy, but also in terms of their ability to transfer knowledge across tasks. Second, we propose a model for continual learning, called Gradient Episodic Memory (GEM) that alleviates forgetting, while allowing beneficial transfer of knowledge to previous tasks. Our experiments on variants of the MNIST and CIFAR-100 datasets demonstrate the strong performance of GEM when compared to the state-of-the-art."
                },
                {
                    "title": "SuperSpike: Supervised Learning in Multilayer Spiking Neural Networks",
                    "abstract": "A vast majority of computation in the brain is performed by spiking neural networks. Despite the ubiquity of such spiking, we currently lack an understanding of how biological spiking neural circuits learn and compute in vivo, as well as how we can instantiate such capabilities in artificial spiking circuits in silico. Here we revisit the problem of supervised learning in temporally coding multilayer spiking neural networks. First, by using a surrogate gradient approach, we derive SuperSpike, a nonlinear voltage-based three-factor learning rule capable of training multilayer networks of deterministic integrate-and-fire neurons to perform nonlinear computations on spatiotemporal spike patterns. Second, inspired by recent results on feedback alignment, we compare the performance of our learning rule under different credit assignment strategies for propagating output errors to hidden units. Specifically, we test uniform, symmetric, and random feedback, finding that simpler tasks can be solved with any type of feedback, while more complex tasks require symmetric feedback. In summary, our results open the door to obtaining a better scientific understanding of learning and computation in spiking neural networks by advancing our ability to train them to solve nonlinear problems involving transformations between different spatiotemporal spike time patterns."
                },
                {
                    "title": "Continual Learning with Deep Generative Replay",
                    "abstract": "Attempts to train a comprehensive artificial intelligence capable of solving multiple tasks have been impeded by a chronic problem called catastrophic forgetting. Although simply replaying all previous data alleviates the problem, it requires large memory and even worse, often infeasible in real world applications where the access to past data is limited. Inspired by the generative nature of hippocampus as a short-term memory system in primate brain, we propose the Deep Generative Replay, a novel framework with a cooperative dual model architecture consisting of a deep generative model (\"generator\") and a task solving model (\"solver\"). With only these two models, training data for previous tasks can easily be sampled and interleaved with those for a new task. We test our methods in several sequential learning settings involving image classification tasks."
                },
                {
                    "title": "ASP: Learning to Forget With Adaptive Synaptic Plasticity in Spiking Neural Networks",
                    "abstract": "A fundamental feature of learning in animals is the \u201cability to forget\u201d that allows an organism to perceive, model, and make decisions from disparate streams of information and adapt to changing environments. Against this backdrop, we present a novel unsupervised learning mechanism adaptive synaptic plasticity (ASP) for improved recognition with spiking neural networks (SNNs) for real time online learning in a dynamic environment. We incorporate an adaptive weight decay mechanism with the traditional spike timing dependent plasticity (STDP) learning to model adaptivity in SNNs. The leak rate of the synaptic weights is modulated based on the temporal correlation between the spiking patterns of the pre- and post-synaptic neurons. This mechanism helps in gradual forgetting of insignificant data while retaining significant, yet old, information. ASP, thus, maintains a balance between forgetting and immediate learning to construct a stable-plastic self-adaptive SNN for continuously changing inputs. We demonstrate that the proposed learning methodology addresses catastrophic forgetting, while yielding significantly improved accuracy over the conventional STDP learning method for digit recognition applications. In addition, we observe that the proposed learning model automatically encodes selective attention toward relevant features in the input data, while eliminating the influence of background noise (or denoising) further improving the robustness of the ASP learning."
                },
                {
                    "title": "Continual Learning Through Synaptic Intelligence",
                    "abstract": "While deep learning has led to remarkable advances across diverse applications, it struggles in domains where the data distribution changes over the course of learning. In stark contrast, biological neural networks continually adapt to changing domains, possibly by leveraging complex molecular machinery to solve many tasks simultaneously. In this study, we introduce intelligent synapses that bring some of this biological complexity into artificial neural networks. Each synapse accumulates task relevant information over time, and exploits this information to rapidly store new memories without forgetting old ones. We evaluate our approach on continual learning of classification tasks, and show that it dramatically reduces forgetting while maintaining computational efficiency."
                },
                {
                    "title": "Overcoming catastrophic forgetting in neural networks",
                    "abstract": "Significance Deep neural networks are currently the most successful machine-learning technique for solving a variety of tasks, including language translation, image classification, and image generation. One weakness of such models is that, unlike humans, they are unable to learn multiple tasks sequentially. In this work we propose a practical solution to train such models sequentially by protecting the weights important for previous tasks. This approach, inspired by synaptic consolidation in neuroscience, enables state of the art results on multiple reinforcement learning problems experienced sequentially. The ability to learn tasks in a sequential fashion is crucial to the development of artificial intelligence. Until now neural networks have not been capable of this and it has been widely thought that catastrophic forgetting is an inevitable feature of connectionist models. We show that it is possible to overcome this limitation and train networks that can maintain expertise on tasks that they have not experienced for a long time. Our approach remembers old tasks by selectively slowing down learning on the weights important for those tasks. We demonstrate our approach is scalable and effective by solving a set of classification tasks based on a hand-written digit dataset and by learning several Atari 2600 games sequentially."
                },
                {
                    "title": "iCaRL: Incremental Classifier and Representation Learning",
                    "abstract": "A major open problem on the road to artificial intelligence is the development of incrementally learning systems that learn about more and more concepts over time from a stream of data. In this work, we introduce a new training strategy, iCaRL, that allows learning in such a class-incremental way: only the training data for a small number of classes has to be present at the same time and new classes can be added progressively. iCaRL learns strong classifiers and a data representation simultaneously. This distinguishes it from earlier works that were fundamentally limited to fixed data representations and therefore incompatible with deep learning architectures. We show by experiments on CIFAR-100 and ImageNet ILSVRC 2012 data that iCaRL can learn many classes incrementally over a long period of time where other strategies quickly fail."
                },
                {
                    "title": "Direct Feedback Alignment Provides Learning in Deep Neural Networks",
                    "abstract": "Artificial neural networks are most commonly trained with the back-propagation algorithm, where the gradient for learning is provided by back-propagating the error, layer by layer, from the output layer to the hidden layers. A recently discovered method called feedback-alignment shows that the weights used for propagating the error backward don't have to be symmetric with the weights used for propagation the activation forward. In fact, random feedback weights work evenly well, because the network learns how to make the feedback useful. In this work, the feedback alignment principle is used for training hidden layers more independently from the rest of the network, and from a zero initial condition. The error is propagated through fixed random feedback connections directly from the output layer to each hidden layer. This simple method is able to achieve zero training error even in convolutional networks and very deep networks, completely without error back-propagation. The method is a step towards biologically plausible machine learning because the error signal is almost local, and no symmetric or reciprocal weights are required. Experiments show that the test performance on MNIST and CIFAR is almost as good as those obtained with back-propagation for fully connected networks. If combined with dropout, the method achieves 1.45% error on the permutation invariant MNIST task."
                },
                {
                    "title": "Training Deep Spiking Neural Networks Using Backpropagation",
                    "abstract": "Deep spiking neural networks (SNNs) hold the potential for improving the latency and energy efficiency of deep neural networks through data-driven event-based computation. However, training such networks is difficult due to the non-differentiable nature of spike events. In this paper, we introduce a novel technique, which treats the membrane potentials of spiking neurons as differentiable signals, where discontinuities at spike times are considered as noise. This enables an error backpropagation mechanism for deep SNNs that follows the same principles as in conventional deep networks, but works directly on spike signals and membrane potentials. Compared with previous methods relying on indirect training and conversion, our technique has the potential to capture the statistics of spikes more precisely. We evaluate the proposed framework on artificially generated events from the original MNIST handwritten digit benchmark, and also on the N-MNIST benchmark recorded with an event-based dynamic vision sensor, in which the proposed method reduces the error rate by a factor of more than three compared to the best previous SNN, and also achieves a higher accuracy than a conventional convolutional neural network (CNN) trained and tested on the same data. We demonstrate in the context of the MNIST task that thanks to their event-driven operation, deep SNNs (both fully connected and convolutional) trained with our method achieve accuracy equivalent with conventional neural networks. In the N-MNIST example, equivalent accuracy is achieved with about five times fewer computational operations."
                },
                {
                    "title": "Supervised Learning Based on Temporal Coding in Spiking Neural Networks",
                    "abstract": "Gradient descent training techniques are remarkably successful in training analog-valued artificial neural networks (ANNs). Such training techniques, however, do not transfer easily to spiking networks due to the spike generation hard nonlinearity and the discrete nature of spike communication. We show that in a feedforward spiking network that uses a temporal coding scheme where information is encoded in spike times instead of spike rates, the network input\u2013output relation is differentiable almost everywhere. Moreover, this relation is piecewise linear after a transformation of variables. Methods for training ANNs thus carry directly to the training of such spiking networks as we show when training on the permutation invariant MNIST task. In contrast to rate-based spiking networks that are often used to approximate the behavior of ANNs, the networks we present spike much more sparsely and their behavior cannot be directly approximated by conventional ANNs. Our results highlight a new approach for controlling the behavior of spiking networks with realistic temporal dynamics, opening up the potential for using these networks to process spike patterns with complex temporal information."
                },
                {
                    "title": "Progressive Neural Networks",
                    "abstract": "Learning to solve complex sequences of tasks--while both leveraging transfer and avoiding catastrophic forgetting--remains a key obstacle to achieving human-level intelligence. The progressive networks approach represents a step forward in this direction: they are immune to forgetting and can leverage prior knowledge via lateral connections to previously learned features. We evaluate this architecture extensively on a wide variety of reinforcement learning tasks (Atari and 3D maze games), and show that it outperforms common baselines based on pretraining and finetuning. Using a novel sensitivity measure, we demonstrate that transfer occurs at both low-level sensory and high-level control layers of the learned policy."
                },
                {
                    "title": "TrueNorth: Design and Tool Flow of a 65 mW 1 Million Neuron Programmable Neurosynaptic Chip",
                    "abstract": "The new era of cognitive computing brings forth the grand challenge of developing systems capable of processing massive amounts of noisy multisensory data. This type of intelligent computing poses a set of constraints, including real-time operation, low-power consumption and scalability, which require a radical departure from conventional system design. Brain-inspired architectures offer tremendous promise in this area. To this end, we developed TrueNorth, a 65 mW real-time neurosynaptic processor that implements a non-von Neumann, low-power, highly-parallel, scalable, and defect-tolerant architecture. With 4096 neurosynaptic cores, the TrueNorth chip contains 1 million digital neurons and 256 million synapses tightly interconnected by an event-driven routing infrastructure. The fully digital 5.4 billion transistor implementation leverages existing CMOS scaling trends, while ensuring one-to-one correspondence between hardware and software. With such aggressive design metrics and the TrueNorth architecture breaking path with prevailing architectures, it is clear that conventional computer-aided design (CAD) tools could not be used for the design. As a result, we developed a novel design methodology that includes mixed asynchronous-synchronous circuits and a complete tool flow for building an event-driven, low-power neurosynaptic chip. The TrueNorth chip is fully configurable in terms of connectivity and neural parameters to allow custom configurations for a wide range of cognitive and sensory perception applications. To reduce the system's communication energy, we have adapted existing application-agnostic very large-scale integration CAD placement tools for mapping logical neural networks to the physical neurosynaptic core locations on the TrueNorth chips. With that, we have successfully demonstrated the use of TrueNorth-based systems in multiple applications, including visual object recognition, with higher performance and orders of magnitude lower power consumption than the same algorithms run on von Neumann architectures. The TrueNorth chip and its tool flow serve as building blocks for future cognitive systems, and give designers an opportunity to develop novel brain-inspired architectures and systems based on the knowledge obtained from this paper."
                },
                {
                    "title": "A Hebbian/Anti-Hebbian Neural Network for Linear Subspace Learning: A Derivation from Multidimensional Scaling of Streaming Data",
                    "abstract": "Neural network models of early sensory processing typically reduce the dimensionality of streaming input data. Such networks learn the principal subspace, in the sense of principal component analysis, by adjusting synaptic weights according to activity-dependent learning rules. When derived from a principled cost function, these rules are nonlocal and hence biologically implausible. At the same time, biologically plausible local rules have been postulated rather than derived from a principled cost function. Here, to bridge this gap, we derive a biologically plausible network for subspace learning on streaming data by minimizing a principled cost function. In a departure from previous work, where cost was quantified by the representation, or reconstruction, error, we adopt a multidimensional scaling cost function for streaming data. The resulting algorithm relies only on biologically plausible Hebbian and anti-Hebbian local learning rules. In a stochastic setting, synaptic weights converge to a stationary state, which projects the input data onto the principal subspace. If the data are generated by a nonstationary distribution, the network can track the principal subspace. Thus, our result makes a step toward an algorithmic theory of neural computation."
                },
                {
                    "title": "An Empirical Investigation of Catastrophic Forgeting in Gradient-Based Neural Networks",
                    "abstract": "Catastrophic forgetting is a problem faced by many machine learning models and algorithms. When trained on one task, then trained on a second task, many machine learning models \"forget\" how to perform the first task. This is widely believed to be a serious problem for neural networks. Here, we investigate the extent to which the catastrophic forgetting problem occurs for modern neural networks, comparing both established and recent gradient-based training algorithms and activation functions. We also examine the effect of the relationship between the first task and the second task on catastrophic forgetting. We find that it is always best to train using the dropout algorithm--the dropout algorithm is consistently best at adapting to the new task, remembering the old task, and has the best tradeoff curve between these two extremes. We find that different tasks and relationships between tasks result in very different rankings of activation function performance. This suggests the choice of activation function should always be cross-validated."
                },
                {
                    "title": "Spike timing-dependent plasticity: a Hebbian learning rule.",
                    "abstract": "Spike timing-dependent plasticity (STDP) as a Hebbian synaptic learning rule has been demonstrated in various neural circuits over a wide spectrum of species, from insects to humans. The dependence of synaptic modification on the order of pre- and postsynaptic spiking within a critical window of tens of milliseconds has profound functional implications. Over the past decade, significant progress has been made in understanding the cellular mechanisms of STDP at both excitatory and inhibitory synapses and of the associated changes in neuronal excitability and synaptic integration. Beyond the basic asymmetric window, recent studies have also revealed several layers of complexity in STDP, including its dependence on dendritic location, the nonlinear integration of synaptic modification induced by complex spike trains, and the modulation of STDP by inhibitory and neuromodulatory inputs. Finally, the functional consequences of STDP have been examined directly in an increasing number of neural circuits in vivo."
                },
                {
                    "title": "Global analysis of Oja's flow for neural networks",
                    "abstract": "A detailed study of Oja's learning equation in neural networks is undertaken in this paper. Not only are such fundamental issues as existence, uniqueness, and representation of solutions completely resolved, but also the convergence issue is resolved. It is shown that the solution of Oja's equation is exponentially convergent to an equilibrium from any initial value. Moreover, the necessary and sufficient conditions are given on the initial value for the solution to converge to a dominant eigenspace of the associated autocorrelation matrix. As a by-product, this result confirms one of Oja's conjectures that the solution converges to the principal eigenspace from almost all initial values. Some other characteristics of the limiting solution are also revealed. These facilitate the determination of the limiting solution in advance using only the initial information. Two examples are analyzed demonstrating the explicit dependence of the limiting solution on the initial value. In another respect, it is found that Oja's equation is the gradient flow of generalized Rayleigh quotients on a Stiefel manifold."
                },
                {
                    "title": "Neural networks and physical systems with emergent collective computational abilities.",
                    "abstract": "Computational properties of use of biological organisms or to the construction of computers can emerge as collective properties of systems having a large number of simple equivalent components (or neurons). The physical meaning of content-addressable memory is described by an appropriate phase space flow of the state of a system. A model of such a system is given, based on aspects of neurobiology but readily adapted to integrated circuits. The collective properties of this model produce a content-addressable memory which correctly yields an entire memory from any subpart of sufficient size. The algorithm for the time evolution of the state of the system is based on asynchronous parallel processing. Additional emergent collective properties include some capacity for generalization, familiarity recognition, categorization, error correction, and time sequence retention. The collective properties are only weakly sensitive to details of the modeling or the failure of individual devices."
                },
                {
                    "title": "Neural Networks, Principal Components, and Subspaces",
                    "abstract": "A single neuron with Hebbian-type learning for the connection weights, and with nonlinear internal feedback, has been shown to extract the statistical principal components of its stationary input pattern sequence. A generalization of this model to a layer of neuron units is given, called the Subspace Network, which yields a multi-dimensional, principal component subspace. This can be used as an associative memory for the input vectors or as a module in nonsupervised learning of data clusters in the input space. It is also able to realize a powerful pattern classifier based on projections on class subspaces. Some classification results for natural textures are given."
                }
            ],
            "categories": [
                "cs.NE",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can Hebbian learning be effectively utilized to enhance continual learning in neuromorphic computing systems?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the field of machine learning, particularly in the context of neuromorphic computing, which aims to mimic the brain's efficiency and adaptability. By leveraging Hebbian learning for continual learning, we can develop systems that not only learn from new data streams without forgetting previous knowledge but also do so in a biologically inspired manner. This could lead to significant advancements in artificial intelligence applications, such as robotics, autonomous systems, and real-time data processing, ultimately influencing future research directions in both theoretical and practical domains.\n\n### [Question 3] - Why is it hard?\nThe challenges in addressing this problem stem from the inherent complexities of Hebbian learning and its application to continual learning. Naive approaches may fail due to the difficulty in optimizing synaptic weights for dynamic data streams, as traditional methods often rely on static datasets. Additionally, the lack of a clear objective function in spike-timing-dependent plasticity (STDP) complicates the integration of biological learning principles into computational models. Overcoming these technical obstacles requires innovative methodologies that can effectively balance memory retention and adaptability in a continually changing environment.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on associative memory models, which do not adequately address the challenges of continual learning in dynamic contexts. Limitations in existing solutions include a lack of scalable methods that can handle larger datasets and networks, as well as insufficient exploration of the potential of STDP in optimizing learning objectives. Our approach differs by integrating Hebbian learning principles with projection-based continual learning methods, aiming to systematically promote advanced learning capabilities that have not been fully realized in prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves utilizing Hebbian learning based on neuronal spiking activities to optimize synaptic weights for principal component analysis in a continual learning framework. We will employ domain-incremental datasets, such as PMNIST, to evaluate the scalability of our method. The key metrics for success will include the model's ability to retain previously learned information while effectively adapting to new data. We expect our results to demonstrate improved performance in continual learning tasks, showcasing the potential of integrating biological learning principles into neuromorphic computing systems."
            }
        },
        "author_data": {
            "5d59fd73-3dee-4ae7-bbc0-65c9ac65fcaf": {
                "pk": "5d59fd73-3dee-4ae7-bbc0-65c9ac65fcaf",
                "name": "Mingqing Xiao",
                "collaborators": [
                    "Zhouchen Lin",
                    "Qingyan Meng",
                    "Zongpeng Zhang",
                    "D.K. He",
                    "Yisen Wang",
                    "Shen Yan",
                    "Zhimin Luo",
                    "Qilun Luo",
                    "Ming Yang",
                    "Wen Li",
                    "Liping Deng",
                    "Yixin Zhu",
                    "Zhoutong Wu",
                    "Cong Fang",
                    "Taoyun Ji",
                    "Yuwu Jiang",
                    "Tong Lin",
                    "Xiaohua Zhou",
                    "Wensheng Chen",
                    "Xiaoli Sun",
                    "Xiujun Zhang",
                    "Chen Xu",
                    "Yuanyan Tang"
                ],
                "domain": [
                    "Spiking Neural Networks",
                    "Neuromorphic Computing",
                    "Machine Learning",
                    "Hyperparameter Optimization"
                ],
                "publications": [
                    {
                        "title": "Online Pseudo-Zeroth-Order Training of Neuromorphic Spiking Neural Networks",
                        "abstract": "Brain-inspired neuromorphic computing with spiking neural networks (SNNs) is a promising energy-efficient computational approach. However, successfully training SNNs in a more biologically plausible and neuromorphic-hardware-friendly way is still challenging. Most recent methods leverage spatial and temporal backpropagation (BP), not adhering to neuromorphic properties. Despite the efforts of some online training methods, tackling spatial credit assignments by alternatives with comparable performance as spatial BP remains a significant problem. In this work, we propose a novel method, online pseudo-zeroth-order (OPZO) training. Our method only requires a single forward propagation with noise injection and direct top-down signals for spatial credit assignment, avoiding spatial BP's problem of symmetric weights and separate phases for layer-by-layer forward-backward propagation. OPZO solves the large variance problem of zeroth-order methods by the pseudo-zeroth-order formulation and momentum feedback connections, while having more guarantees than random feedback. Combining online training, OPZO can pave paths to on-chip SNN training. Experiments on neuromorphic and static datasets with fully connected and convolutional networks demonstrate the effectiveness of OPZO with similar performance compared with spatial BP, as well as estimated low training costs."
                    },
                    {
                        "title": "Temporal Spiking Neural Networks with Synaptic Delay for Graph Reasoning",
                        "abstract": "Spiking neural networks (SNNs) are investigated as biologically inspired models of neural computation, distinguished by their computational capability and energy efficiency due to precise spiking times and sparse spikes with event-driven computation. A significant question is how SNNs can emulate human-like graph-based reasoning of concepts and relations, especially leveraging the temporal domain optimally. This paper reveals that SNNs, when amalgamated with synaptic delay and temporal coding, are proficient in executing (knowledge) graph reasoning. It is elucidated that spiking time can function as an additional dimension to encode relation properties via a neural-generalized path formulation. Empirical results highlight the efficacy of temporal delay in relation processing and showcase exemplary performance in diverse graph reasoning tasks. The spiking model is theoretically estimated to achieve $20\\times$ energy savings compared to non-spiking counterparts, deepening insights into the capabilities and potential of biologically inspired SNNs for efficient reasoning. The code is available at https://github.com/pkuxmq/GRSNN."
                    },
                    {
                        "title": "Designing Universally-Approximating Deep Neural Networks: A First-Order Optimization Approach",
                        "abstract": "Universal approximation capability, also referred to as universality, is an important property of deep neural networks, endowing them with the potency to accurately represent the underlying target function in learning tasks. In practice, the architecture of deep neural networks largely influences the performance of the models. However, most existing methodologies for designing neural architectures, such as the heuristic manual design or neural architecture search, ignore the universal approximation property, thus losing a potential safeguard about the performance. In this paper, we propose a unified framework to design the architectures of deep neural networks with a universality guarantee based on first-order optimization algorithms, where the forward pass is interpreted as the updates of an optimization algorithm. The (explicit or implicit) network is designed by replacing each gradient term in the algorithm with a learnable module similar to a two-layer network or its derivatives. Specifically, we explore the realm of width-bounded neural networks, a common practical scenario, showcasing their universality. Moreover, adding operations of normalization, downsampling, and upsampling does not hurt the universality. To the best of our knowledge, this is the first work that width-bounded networks with universal approximation guarantee can be designed in a principled way. Our framework can inspire a variety of neural architectures including some renowned structures such as ResNet and DenseNet, as well as novel innovations. The experimental results on image classification problems demonstrate that the newly inspired networks are competitive and surpass the baselines of ResNet, DenseNet, as well as the advanced ConvNeXt and ViT, testifying to the effectiveness of our framework."
                    },
                    {
                        "title": "Efficient and generalizable cross-patient epileptic seizure detection through a spiking neural network",
                        "abstract": "Introduction Epilepsy is a global chronic disease that brings pain and inconvenience to patients, and an electroencephalogram (EEG) is the main analytical tool. For clinical aid that can be applied to any patient, an automatic cross-patient epilepsy seizure detection algorithm is of great significance. Spiking neural networks (SNNs) are modeled on biological neurons and are energy-efficient on neuromorphic hardware, which can be expected to better handle brain signals and benefit real-world, low-power applications. However, automatic epilepsy seizure detection rarely considers SNNs. Methods In this article, we have explored SNNs for cross-patient seizure detection and discovered that SNNs can achieve comparable state-of-the-art performance or a performance that is even better than artificial neural networks (ANNs). We propose an EEG-based spiking neural network (EESNN) with a recurrent spiking convolution structure, which may better take advantage of temporal and biological characteristics in EEG signals. Results We extensively evaluate the performance of different SNN structures, training methods, and time settings, which builds a solid basis for understanding and evaluation of SNNs in seizure detection. Moreover, we show that our EESNN model can achieve energy reduction by several orders of magnitude compared with ANNs according to the theoretical estimation. Discussion These results show the potential for building high-performance, low-power neuromorphic systems for seizure detection and also broaden real-world application scenarios of SNNs."
                    },
                    {
                        "title": "Towards Memory- and Time-Efficient Backpropagation for Training Spiking Neural Networks",
                        "abstract": "Spiking Neural Networks (SNNs) are promising energy-efficient models for neuromorphic computing. For training the non-differentiable SNN methods, the backpropagation through time (BPTT) with surrogate gradients (SG) method has achieved high performance. However, this method suffers from considerable memory cost and training time during training. In this paper, we propose the Spatial Learning Through Time (SLTT) method that can achieve high performance while greatly improving training efficiency compared with BPTT. First, we show that the backpropagation of SNNs through the temporal domain contributes just a little to the final calculated gradients. Thus, we propose to ignore the unimportant routes in the computational graph during backpropagation. The proposed method reduces the number of scalar multiplications and achieves a small memory occupation that is independent of the total time steps. Furthermore, we propose a variant of SLTT, called SLTT-K, that allows backpropagation only at K time steps, then the required number of scalar multiplications is further reduced and is independent of the total time steps. Experiments on both static and neuromorphic datasets demonstrate superior training efficiency and performance of our SLTT. In particular, our method achieves state-of-the-art accuracy on ImageNet, while the memory cost and training time are reduced by more than 70% and 50%, respectively, compared with BPTT. Our code is available at https://github.com/qymeng94/SLTT."
                    },
                    {
                        "title": "Hyper-Laplacian Regularized Multi-View Clustering with Exclusive L21 Regularization and Tensor Log-Determinant Minimization Approach",
                        "abstract": "Multi-view clustering aims to capture the multiple views inherent information by identifying the data clustering that reflects distinct features of datasets. Since there is a consensus in literature that different views of a dataset share a common latent structure, most existing multi-view subspace learning methods rely on the nuclear norm to seek the low-rank representation of the underlying subspace. However, the nuclear norm often fails to distinguish the variance of features for each cluster due to its convex nature and data tends to fall in multiple non-linear subspaces for multi-dimensional datasets. To address these problems, we propose a new and novel multi-view clustering method (HL-L21-TLD-MSC) that unifies the Hyper-Laplacian (HL) and exclusive \u21132,1 (L21) regularization with the Tensor Log-Determinant Rank Minimization (TLD) setting. Specifically, the hyper-Laplacian regularization maintains the local geometrical structure that makes the estimation prune to nonlinearities, and the mixed \u21132,1 and \u21131,2 regularization provides the joint sparsity within-cluster as well as the exclusive sparsity between-cluster. Furthermore, a log-determinant function is used as a tighter tensor rank approximation to discriminate the dimension of features. An efficient alternating algorithm is then derived to optimize the proposed model, and the construction of a convergent sequence to the Karush-Kuhn-Tucker (KKT) critical point solution is mathematically validated in detail. Extensive experiments are conducted on ten well-known datasets to demonstrate that the proposed approach outperforms the existing state-of-the-art approaches with various scenarios, in which, six of them achieve perfect results under our framework developed in this article, demonstrating highl effectiveness for the proposed approach."
                    },
                    {
                        "title": "Multidimensional Data Processing With Bayesian Inference via Structural Block Decomposition",
                        "abstract": "How to handle large multidimensional datasets, such as hyperspectral images and video information, efficiently and effectively plays a critical role in big-data processing. The characteristics of low-rank tensor decomposition in recent years demonstrate the essentials in describing the tensor rank, which often leads to promising approaches. However, most current tensor decomposition models consider the rank-1 component simply to be the vector outer product, which may not fully capture the correlated spatial information effectively for large-scale and high-order multidimensional datasets. In this article, we develop a new novel tensor decomposition model by extending it to the matrix outer product or called Bhattacharya\u2013Mesner product, to form an effective dataset decomposition. The fundamental idea is to decompose tensors structurally in a compact manner as much as possible while retaining data spatial characteristics in a tractable way. By incorporating the framework of the Bayesian inference, a new tensor decomposition model on the subtle matrix unfolding outer product is established for both tensor completion and robust principal component analysis problems, including hyperspectral image completion and denoising, traffic data imputation, and video background subtraction. Numerical experiments on real-world datasets demonstrate the highly desirable effectiveness of the proposed approach."
                    },
                    {
                        "title": "Metafeature Selection via Multivariate Sparse-Group Lasso Learning for Automatic Hyperparameter Configuration Recommendation",
                        "abstract": "The performance of classification algorithms is mainly governed by the hyperparameter settings deployed in applications, and the search for desirable hyperparameter configurations usually is quite challenging due to the complexity of datasets. Metafeatures are a group of measures that characterize the underlying dataset from various aspects, and the corresponding recommendation algorithm fully relies on the appropriate selection of metafeatures. Metalearning (MtL), aiming to improve the learning algorithm itself, requires development in integrating features, models, and algorithm learning to accomplish its goal. In this article, we develop a multivariate sparse-group Lasso (SGLasso) model embedded with MtL capacity in recommending suitable configurations via learning. The main idea is to select the principal metafeatures by removing those redundant or irregular ones, promoting both efficiency and performance in the hyperparameter configuration recommendation. To be specific, we first extract the metafeatures and classification performance of a set of configurations from the collection of historical datasets, and then, a metaregression task is established through SGLasso to capture the main characteristics of the underlying relationship between metafeatures and historical performance. For a new dataset, the classification performance of configurations can be estimated through the selected metafeatures so that the configuration with the highest predictive performance in terms of the new dataset can be generated. Furthermore, a general MtL architecture combined with our model is developed. Extensive experiments are conducted on 136 UCI datasets, demonstrating the effectiveness of the proposed approach. The empirical results on the well-known SVM show that our model can effectively recommend suitable configurations and outperform the existing MtL-based methods and the well-known search-based algorithms, such as random search, Bayesian optimization, and Hyperband."
                    },
                    {
                        "title": "Tensorial Multiview Representation for Saliency Detection via Nonconvex Approach",
                        "abstract": "In the study of salient object detection, multiview features play an important role in identifying various underlying salient objects. As to current common patch-based methods, all different features are handled directly by stacking them into a high-dimensional vector to represent related image patches. These approaches ignore the correlations inhering in the original spatial structure, which may lead to the loss of certain underlying characterization such as view interaction. In this article, different from currently available approaches, a tensorial feature representation framework is developed for the salient object detection in order to better explore the complementary information of multiview features. Under the tensor framework, a tensor low-rank constraint is applied to the background to capture its intrinsic structure, a tensor group sparsity regularization is posed on the salient part, and a tensorial sliced Laplacian regularization is then introduced to enlarge the gap between the subspaces of the background and salient object. Moreover, a nonconvex tensor Log-determinant function, instead of the tensor nuclear norm, is adopted to approximate the tensor rank for effectively suppressing the confusing information resulted from underlying complex backgrounds. Further, we have deduced the closed-form solution of this nonconvex minimization problem and established a feasible algorithm whose convergence is mathematically proven. Experiments on five well-known public datasets are provided and the simulations demonstrate that our method outperforms the latest unsupervised handcrafted features-based methods in the literature. Furthermore, our model is flexible with various deep features and is competitive with the state-of-the-art approaches."
                    },
                    {
                        "title": "Supplementary Materials for: Training Feedback Spiking Neural Networks by Implicit Differentiation on the Equilibrium State",
                        "abstract": "Mingqing Xiao, Qingyan Meng, Zongpeng Zhang, Yisen Wang, Zhouchen Lin \u2217 Key Laboratory of Machine Perception (MoE), School of AI, Peking University The Chinese University of Hong Kong, Shenzhen Shenzhen Research Institute of Big Data Center for Data Science, Academy for Advanced Interdisciplinary Studies, Peking University Institute for Artificial Intelligence, Peking University Pazhou Lab, Guangzhou 510330, China {mingqing_xiao, yisen.wang, zlin}@pku.edu.cn, qingyanmeng@link.cuhk.edu.cn, zongpeng.zhang98@gmail.com"
                    },
                    {
                        "title": "Online Training Through Time for Spiking Neural Networks",
                        "abstract": "Spiking neural networks (SNNs) are promising brain-inspired energy-efficient models. Recent progress in training methods has enabled successful deep SNNs on large-scale tasks with low latency. Particularly, backpropagation through time (BPTT) with surrogate gradients (SG) is popularly used to achieve high performance in a very small number of time steps. However, it is at the cost of large memory consumption for training, lack of theoretical clarity for optimization, and inconsistency with the online property of biological learning and rules on neuromorphic hardware. Other works connect spike representations of SNNs with equivalent artificial neural network formulation and train SNNs by gradients from equivalent mappings to ensure descent directions. But they fail to achieve low latency and are also not online. In this work, we propose online training through time (OTTT) for SNNs, which is derived from BPTT to enable forward-in-time learning by tracking presynaptic activities and leveraging instantaneous loss and gradients. Meanwhile, we theoretically analyze and prove that gradients of OTTT can provide a similar descent direction for optimization as gradients based on spike representations under both feedforward and recurrent conditions. OTTT only requires constant training memory costs agnostic to time steps, avoiding the significant memory costs of BPTT for GPU training. Furthermore, the update rule of OTTT is in the form of three-factor Hebbian learning, which could pave a path for online on-chip learning. With OTTT, it is the first time that two mainstream supervised SNN training methods, BPTT with SG and spike representation-based training, are connected, and meanwhile in a biologically plausible form. Experiments on CIFAR-10, CIFAR-100, ImageNet, and CIFAR10-DVS demonstrate the superior performance of our method on large-scale static and neuromorphic datasets in small time steps."
                    },
                    {
                        "title": "A New Automatic Hyperparameter Recommendation Approach Under Low-Rank Tensor Completion e Framework",
                        "abstract": "Hyperparameter optimization (HPO), characterized by hyperparameter tuning, is not only a critical step for effective modeling but also is the most time-consuming process in machine learning. Traditional search-based algorithms tend to require extensive configuration evaluations for each round to select the desirable hyperparameters during the process, and they are often very inefficient for the implementations on large-scale tasks. In this paper, we study the HPO problem via meta-learning (MtL) approach under the low-rank tensor completion (LRTC) framework. Our proposed approach predicts the performance for hyperparameters of new problems based on their previous performance so that the underlying suitable hyperparameters with better efficiency can be attained. Different from existing approaches, the hyperparameter performance space is instantiated under tensor framework that can preserve the spatial structure and reflect the correlations among the adjacent hyperparameters. When some partial evaluations are available for a new problem, the task of estimating the performance of the unevaluated hyperparameters can be formulated as a tensor completion (TC) problem. Toward the completion purpose, we develop an LRTC algorithm utilizing the sum of nuclear norm (SNN) model. A kernelized version is further developed to capture the nonlinear structure of the performance space. In addition, a corresponding coupled matrix factorization (CMF) algorithm is established to render the predictions solely depend on the meta-features to avoid additional hyperparameter evaluations. Finally, a strategy integrating LRTC and CMF is provided to further enhance the recommendation capacity. We test recommendation performance with our proposed methods for classical SVM and the state-of-the-art deep neural networks such as vision transformer (ViT) and residual network (ResNet), and the obtained results demonstrate the effectiveness of our approaches under various evaluation metrics by comparing with the baselines commonly used for MtL."
                    },
                    {
                        "title": "Training High-Performance Low-Latency Spiking Neural Networks by Differentiation on Spike Representation",
                        "abstract": "Spiking Neural Network (SNN) is a promising energy-efficient AI model when implemented on neuromorphic hardware. However, it is a challenge to efficiently train SNNs due to their non-differentiability. Most existing methods either suffer from high latency (i.e., long simulation time steps), or cannot achieve as high performance as Artificial Neural Networks (ANNs). In this paper, we propose the Differentiation on Spike Representation (DSR) method, which could achieve high performance that is competitive to ANNs yet with low latency. First, we encode the spike trains into spike representation using (weighted) firing rate coding. Based on the spike representation, we systematically derive that the spiking dynamics with common neural models can be represented as some sub-differentiable mapping. With this viewpoint, our proposed DSR method trains SNNs through gradients of the mapping and avoids the common non-differentiability problem in SNN training. Then we analyze the error when representing the specific mapping with the forward computation of the SNN. To reduce such error, we propose to train the spike threshold in each layer, and to introduce a new hyperparameter for the neural models. With these components, the DSR method can achieve state-of-the-art SNN performance with low latency on both static and neuromorphic datasets, including CIFAR-10, CIFAR-100, ImageNet, and DVS-CIFAR10."
                    }
                ]
            },
            "254c6141-8853-4366-a5a5-cba14b462605": {
                "pk": "254c6141-8853-4366-a5a5-cba14b462605",
                "name": "Qingyan Meng",
                "collaborators": [
                    "Mingqing Xiao",
                    "Zhouchen Lin",
                    "Peizheng Yan",
                    "X. Chang",
                    "Tian Zhang",
                    "Zongpeng Zhang",
                    "Yisen Wang",
                    "D.K. He",
                    "Shen Yan",
                    "Zhimin Luo",
                    "Yan Liu",
                    "Dong Liu",
                    "Duosheng Luo",
                    "Xue Wang",
                    "XiuTeng Zhou",
                    "Zhenyu Zhao",
                    "Wenjin Zhang",
                    "Chunxia Ma",
                    "Bin Wang",
                    "Xiaoxuan Fan",
                    "Qianwen Sun",
                    "Qingxiang Zhang",
                    "Yue Kong",
                    "Dongsheng Zhao",
                    "Beibei Mao",
                    "Pan Zhao",
                    "Lu Zhang",
                    "Jiahao Qi",
                    "Jun-dong Zhang",
                    "Jiaming Ju",
                    "Haotian Jiang",
                    "ShiyuanWang",
                    "Ruifeng Ji",
                    "Junyan Wang",
                    "Yongtian Yin",
                    "Shiyuan Wang",
                    "Longqiong Zou",
                    "M. Zhang",
                    "Lei Zhang",
                    "Yuping Zhao"
                ],
                "domain": [
                    "Spiking Neural Networks",
                    "Neuromorphic Computing",
                    "Cardiovascular Disease",
                    "Mitochondrial Quality Control"
                ],
                "publications": [
                    {
                        "title": "Online Pseudo-Zeroth-Order Training of Neuromorphic Spiking Neural Networks",
                        "abstract": "Brain-inspired neuromorphic computing with spiking neural networks (SNNs) is a promising energy-efficient computational approach. However, successfully training SNNs in a more biologically plausible and neuromorphic-hardware-friendly way is still challenging. Most recent methods leverage spatial and temporal backpropagation (BP), not adhering to neuromorphic properties. Despite the efforts of some online training methods, tackling spatial credit assignments by alternatives with comparable performance as spatial BP remains a significant problem. In this work, we propose a novel method, online pseudo-zeroth-order (OPZO) training. Our method only requires a single forward propagation with noise injection and direct top-down signals for spatial credit assignment, avoiding spatial BP's problem of symmetric weights and separate phases for layer-by-layer forward-backward propagation. OPZO solves the large variance problem of zeroth-order methods by the pseudo-zeroth-order formulation and momentum feedback connections, while having more guarantees than random feedback. Combining online training, OPZO can pave paths to on-chip SNN training. Experiments on neuromorphic and static datasets with fully connected and convolutional networks demonstrate the effectiveness of OPZO with similar performance compared with spatial BP, as well as estimated low training costs."
                    },
                    {
                        "title": "Towards Memory- and Time-Efficient Backpropagation for Training Spiking Neural Networks",
                        "abstract": "Spiking Neural Networks (SNNs) are promising energy-efficient models for neuromorphic computing. For training the non-differentiable SNN methods, the backpropagation through time (BPTT) with surrogate gradients (SG) method has achieved high performance. However, this method suffers from considerable memory cost and training time during training. In this paper, we propose the Spatial Learning Through Time (SLTT) method that can achieve high performance while greatly improving training efficiency compared with BPTT. First, we show that the backpropagation of SNNs through the temporal domain contributes just a little to the final calculated gradients. Thus, we propose to ignore the unimportant routes in the computational graph during backpropagation. The proposed method reduces the number of scalar multiplications and achieves a small memory occupation that is independent of the total time steps. Furthermore, we propose a variant of SLTT, called SLTT-K, that allows backpropagation only at K time steps, then the required number of scalar multiplications is further reduced and is independent of the total time steps. Experiments on both static and neuromorphic datasets demonstrate superior training efficiency and performance of our SLTT. In particular, our method achieves state-of-the-art accuracy on ImageNet, while the memory cost and training time are reduced by more than 70% and 50%, respectively, compared with BPTT. Our code is available at https://github.com/qymeng94/SLTT."
                    },
                    {
                        "title": "Supplementary Materials for: Training Feedback Spiking Neural Networks by Implicit Differentiation on the Equilibrium State",
                        "abstract": "Mingqing Xiao, Qingyan Meng, Zongpeng Zhang, Yisen Wang, Zhouchen Lin \u2217 Key Laboratory of Machine Perception (MoE), School of AI, Peking University The Chinese University of Hong Kong, Shenzhen Shenzhen Research Institute of Big Data Center for Data Science, Academy for Advanced Interdisciplinary Studies, Peking University Institute for Artificial Intelligence, Peking University Pazhou Lab, Guangzhou 510330, China {mingqing_xiao, yisen.wang, zlin}@pku.edu.cn, qingyanmeng@link.cuhk.edu.cn, zongpeng.zhang98@gmail.com"
                    },
                    {
                        "title": "Xiaoqinglong Decoction Enhances Autophagy to Antagonist Airway Inflammation Induced by Cold in Asthmatic Rats",
                        "abstract": "Asthma is a common chronic respiratory disease characterized by wheezing and shortness of breath. Its risk factors include genetic and acquired factors. The acquired factors are closely related to the environment, especially cold conditions. Autophagy plays a regulatory role in asthma. Therefore, we hypothesized that asthma can be controlled by drug intervention at the autophagy level under cold conditions. The Xiaoqinglong decoction (XQLT) was freeze-dried. The compounds in the freeze-dried powder were identified and quantified using reference standards via the high-performance liquid chromatography method. Ovalbumin (OVA)-sensitized rats were subjected to cold stimulation. The effect of cold stimulation on autophagy levels was determined, and it was confirmed that cold stimulation affected autophagy. The effects and mechanisms of XQLT in an asthmatic rat model (OVA-sensitized rats stimulated with cold) were explored. The concentrations of paeoniflorin, liquiritin, trans-cinnamic acid, glycyrrhizic acid, 6-gingerol, schisandrol A, and asarinin in XQLT freeze-dried powder were 14.45, 3.85, 1.03, 3.93, 0.59, 0.24, and 0.091\u2009mg/g, respectively. Cold stimulation is an important cause of asthma. The inflammatory factors in bronchoalveolar lavage fluid and serum were increased in the model group, accompanied by a decline in autophagy level. The treatment with XQLT increased the expression of autophagy genes and decreased the expression of inflammatory factors. Histological studies showed that XQLT improved inflammatory infiltration and collagen fiber deposition in the lungs of rats. XQLT intervention increased autophagy in asthmatic rats. Autophagy plays a role in phagocytosis and reduces the accumulation of abnormal metabolites in the body to reduce airway inflammation and promote asthma recovery."
                    },
                    {
                        "title": "Online Training Through Time for Spiking Neural Networks",
                        "abstract": "Spiking neural networks (SNNs) are promising brain-inspired energy-efficient models. Recent progress in training methods has enabled successful deep SNNs on large-scale tasks with low latency. Particularly, backpropagation through time (BPTT) with surrogate gradients (SG) is popularly used to achieve high performance in a very small number of time steps. However, it is at the cost of large memory consumption for training, lack of theoretical clarity for optimization, and inconsistency with the online property of biological learning and rules on neuromorphic hardware. Other works connect spike representations of SNNs with equivalent artificial neural network formulation and train SNNs by gradients from equivalent mappings to ensure descent directions. But they fail to achieve low latency and are also not online. In this work, we propose online training through time (OTTT) for SNNs, which is derived from BPTT to enable forward-in-time learning by tracking presynaptic activities and leveraging instantaneous loss and gradients. Meanwhile, we theoretically analyze and prove that gradients of OTTT can provide a similar descent direction for optimization as gradients based on spike representations under both feedforward and recurrent conditions. OTTT only requires constant training memory costs agnostic to time steps, avoiding the significant memory costs of BPTT for GPU training. Furthermore, the update rule of OTTT is in the form of three-factor Hebbian learning, which could pave a path for online on-chip learning. With OTTT, it is the first time that two mainstream supervised SNN training methods, BPTT with SG and spike representation-based training, are connected, and meanwhile in a biologically plausible form. Experiments on CIFAR-10, CIFAR-100, ImageNet, and CIFAR10-DVS demonstrate the superior performance of our method on large-scale static and neuromorphic datasets in small time steps."
                    },
                    {
                        "title": "Detection of Auxiliary Equipment in Engine Room Based on Improved SSD",
                        "abstract": "To improve the intelligent operation and maintenance of ships, the research on the perception of engine room auxiliary equipment (A/E) will realize AUTO-0 and replace the engineers. Due to compact layout of the cabin and the different shapes of equipment, the Single Shot MultiBox Detector (SSD) that uses the traditional convolution method for feature extraction cannot adapt to the complex environment. Therefore, an improved A/E detection algorithm based on SSD is proposed. Firstly, we collected the images of actual ships, annotated them manually, and used the K-Means clustering algorithm to analyze them. Secondly, we replaced the classification loss of SSD with the focal loss, balancing the positive and negative samples in the dataset. Finally, we added repulsion loss for the overlapping targets to improve the detection effect of the model against dense occlusion & overlap. The experiment is based on the dataset containing 18185 training object boxes, and 2044 test object boxes, the mean average precision (mAP) in this paper reaches 78.95%, which is 5.63% higher than the original SSD, and the detection boxes are more suitable."
                    },
                    {
                        "title": "Training High-Performance Low-Latency Spiking Neural Networks by Differentiation on Spike Representation",
                        "abstract": "Spiking Neural Network (SNN) is a promising energy-efficient AI model when implemented on neuromorphic hardware. However, it is a challenge to efficiently train SNNs due to their non-differentiability. Most existing methods either suffer from high latency (i.e., long simulation time steps), or cannot achieve as high performance as Artificial Neural Networks (ANNs). In this paper, we propose the Differentiation on Spike Representation (DSR) method, which could achieve high performance that is competitive to ANNs yet with low latency. First, we encode the spike trains into spike representation using (weighted) firing rate coding. Based on the spike representation, we systematically derive that the spiking dynamics with common neural models can be represented as some sub-differentiable mapping. With this viewpoint, our proposed DSR method trains SNNs through gradients of the mapping and avoids the common non-differentiability problem in SNN training. Then we analyze the error when representing the specific mapping with the forward computation of the SNN. To reduce such error, we propose to train the spike threshold in each layer, and to introduce a new hyperparameter for the neural models. With these components, the DSR method can achieve state-of-the-art SNN performance with low latency on both static and neuromorphic datasets, including CIFAR-10, CIFAR-100, ImageNet, and DVS-CIFAR10."
                    },
                    {
                        "title": "Training Feedback Spiking Neural Networks by Implicit Differentiation on the Equilibrium State",
                        "abstract": "Spiking neural networks (SNNs) are brain-inspired models that enable energy-efficient implementation on neuromorphic hardware. However, the supervised training of SNNs remains a hard problem due to the discontinuity of the spiking neuron model. Most existing methods imitate the backpropagation framework and feedforward architectures for artificial neural networks, and use surrogate derivatives or compute gradients with respect to the spiking time to deal with the problem. These approaches either accumulate approximation errors or only propagate information limitedly through existing spikes, and usually require information propagation along time steps with large memory costs and biological implausibility. In this work, we consider feedback spiking neural networks, which are more brain-like, and propose a novel training method that does not rely on the exact reverse of the forward computation. First, we show that the average firing rates of SNNs with feedback connections would gradually evolve to an equilibrium state along time, which follows a fixed-point equation. Then by viewing the forward computation of feedback SNNs as a black-box solver for this equation, and leveraging the implicit differentiation on the equation, we can compute the gradient for parameters without considering the exact forward procedure. In this way, the forward and backward procedures are decoupled and therefore the problem of non-differentiable spiking functions is avoided. We also briefly discuss the biological plausibility of implicit differentiation, which only requires computing another equilibrium. Extensive experiments on MNIST, Fashion-MNIST, N-MNIST, CIFAR-10, and CIFAR-100 demonstrate the superior performance of our method for feedback models with fewer neurons and parameters in a small number of time steps. Our code is avaiable at https://github.com/pkuxmq/IDE-FSNN."
                    },
                    {
                        "title": "Puerarin Attenuates LPS-Induced Inflammatory Responses and Oxidative Stress Injury in Human Umbilical Vein Endothelial Cells through Mitochondrial Quality Control",
                        "abstract": "Atherosclerosis is closely associated with the inflammatory reaction of vascular endothelial cells. Puerarin (Pue), the main active component isolated from the rhizome of Pueraria lobata, is an isoflavone compound with potent antioxidant properties. Although Pue exhibits promising antiatherosclerotic pharmacological effects, only a few studies have reported its protective effect on endothelial cells. This study found that Pue could partly regulate mitochondrial function in human umbilical vein endothelial cells (HUVECs) and reduce or inhibit lipopolysaccharide-induced inflammatory reactions and oxidative stress injury in HUVECs, likely via mitochondrial quality control. Furthermore, the protective effect of Pue on HUVECs was closely related to the SIRT-1 signaling pathway. Pue increased autophagy and mitochondrial antioxidant potential via increased SIRT-1 expression, reducing excessive production of ROS and inhibiting the expression of inflammatory factors and oxidative stress injury. Therefore, Pue may improve mitochondrial respiratory function and energy metabolism, increasing the vulnerability of HUVECs to an inflammatory state."
                    },
                    {
                        "title": "Quercetin Improves Cardiomyocyte Vulnerability to Hypoxia by Regulating SIRT1/TMBIM6-Related Mitophagy and Endoplasmic Reticulum Stress",
                        "abstract": "Cardiomyocyte apoptosis is an important pathological mechanism underlying cardiovascular diseases and is commonly caused by hypoxia. Moreover, hypoxic injury occurs not only in common cardiovascular diseases but also following various treatments of heart-related conditions. One of the major mechanisms underlying hypoxic injury is oxidative stress. Quercetin has been shown to exert antioxidant stress and vascular protective effects, making it a promising candidate for treating cardiovascular diseases. Therefore, we examined the protective effect of quercetin on human cardiomyocytes subjected to hypoxia-induced oxidative stress damage and its underlying mechanism. Human cardiomyocytes were subjected to hypoxia/reoxygenation (H/R) in vitro with or without quercetin pretreatment; thereafter, flow cytometry, Cell Counting Kit-8 assay, laser scanning confocal microscopy, quantitative PCR, western blotting, and enzyme-linked immunosorbent assay were performed to analyze the effects of quercetin on cardiomyocytes. We found that H/R induced reactive oxygen species overproduction and endoplasmic reticulum stress, as well as inhibited the function of the mitochondria/endoplasmic reticulum and mitophagy, eventually leading to apoptosis and decreasing the viability of human cardiomyocytes. Quercetin pretreatment inhibited H/R-mediated overproduction of reactive oxygen species and damage caused by oxidative stress, increased mitophagy, regulated mRNA and protein expression of transmembrane BAX inhibitor-1 motif-containing 6 (TMBIM6), regulated endoplasmic reticulum stress, and improved the vulnerability of human cardiomyocytes to H/R. Furthermore, transfection with short interfering RNA against silent information regulator protein 1 (SIRT1) counteracted the protective effects of quercetin on cardiomyocytes. Thus, quercetin was predicted to regulate mitophagy and endoplasmic reticulum stress through SIRT1/TMBIM6 and inhibit H/R-induced oxidative stress damage. These findings may be useful for developing treatments for hypoxic injury-induced cardiovascular diseases and further highlight the potential of quercetin for regulating mitochondrial quality control and endoplasmic reticulum function."
                    },
                    {
                        "title": "SIRT5-Related Desuccinylation Modification Contributes to Quercetin-Induced Protection against Heart Failure and High-Glucose-Prompted Cardiomyocytes Injured through Regulation of Mitochondrial Quality Surveillance",
                        "abstract": "Myocardial fibrosis represents the primary pathological change associated with diabetic cardiomyopathy and heart failure, and it leads to decreased myocardial compliance with impaired cardiac diastolic and systolic function. Quercetin, an active ingredient in various medicinal plants, exerts therapeutic effects against cardiovascular diseases. Here, we investigate whether SIRT5- and IDH2-related desuccinylation is involved in the underlying mechanism of myocardial fibrosis in heart failure while exploring related therapeutic drugs for mitochondrial quality surveillance. Mouse models of myocardial fibrosis and heart failure, established by transverse aortic constriction (TAC), were administered with quercetin (50\u2009mg/kg) daily for 4 weeks. HL-1 cells were pretreated with quercetin and treated with high glucose (30\u2009mM) in vitro. Cardiac function, western blotting, quantitative PCR, enzyme-linked immunosorbent assay, and immunofluorescence analysis were employed to analyze mitochondrial quality surveillance, oxidative stress, and inflammatory response in myocardial cells, whereas IDH2 succinylation levels were detected using immunoprecipitation. Myocardial fibrosis and heart failure incidence increased after TAC, with abnormal cardiac ejection function. Following high-glucose treatment, HL-1 cell activity was inhibited, causing excess production of reactive oxygen species and inhibition of mitochondrial respiratory complex I/III activity and mitochondrial antioxidant enzyme activity, as well as increased oxidative stress and inflammatory response, imbalanced mitochondrial quality surveillance and homeostasis, and increased apoptosis. Quercetin inhibited myocardial fibrosis and improved cardiac function by increasing mitochondrial energy metabolism and regulating mitochondrial fusion/fission and mitochondrial biosynthesis while inhibiting the inflammatory response and oxidative stress injury. Additionally, TAC inhibited SIRT5 expression at the mitochondrial level and increased IDH2 succinylation. However, quercetin promoted the desuccinylation of IDH2 by increasing SIRT5 expression. Moreover, treatment with si-SIRT5 abolished the protective effect of quercetin on cell viability. Hence, quercetin may promote the desuccinylation of IDH2 through SIRT5, maintain mitochondrial homeostasis, protect mouse cardiomyocytes under inflammatory conditions, and improve myocardial fibrosis, thereby reducing the incidence of heart failure."
                    },
                    {
                        "title": "Natural Antioxidants Improve the Vulnerability of Cardiomyocytes and Vascular Endothelial Cells under Stress Conditions: A Focus on Mitochondrial Quality Control",
                        "abstract": "Cardiovascular disease has become one of the main causes of human death. In addition, many cardiovascular diseases are accompanied by a series of irreversible damages that lead to organ and vascular complications. In recent years, the potential therapeutic strategy of natural antioxidants in the treatment of cardiovascular diseases through mitochondrial quality control has received extensive attention. Mitochondria are the main site of energy metabolism in eukaryotic cells, including myocardial and vascular endothelial cells. Mitochondrial quality control processes ensure normal activities of mitochondria and cells by maintaining stable mitochondrial quantity and quality, thus protecting myocardial and endothelial cells against stress. Various stresses can affect mitochondrial morphology and function. Natural antioxidants extracted from plants and natural medicines are becoming increasingly common in the clinical treatment of diseases, especially in the treatment of cardiovascular diseases. Natural antioxidants can effectively protect myocardial and endothelial cells from stress-induced injury by regulating mitochondrial quality control, and their safety and effectiveness have been preliminarily verified. This review summarises the damage mechanisms of various stresses in cardiomyocytes and vascular endothelial cells and the mechanisms of natural antioxidants in improving the vulnerability of these cell types to stress by regulating mitochondrial quality control. This review is aimed at paving the way for novel treatments for cardiovascular diseases and the development of natural antioxidant drugs."
                    },
                    {
                        "title": "Regulation of Mitochondrial Quality Control by Natural Drugs in the Treatment of Cardiovascular Diseases: Potential and Advantages",
                        "abstract": "Mitochondria are double-membraned cellular organelles that provide the required energy and metabolic intermediates to cardiomyocytes. Mitochondrial respiratory chain defects, structure abnormalities, and DNA mutations can affect the normal function of cardiomyocytes, causing an imbalance in intracellular calcium ion homeostasis, production of reactive oxygen species, and apoptosis. Mitochondrial quality control (MQC) is an important process that maintains mitochondrial homeostasis in cardiomyocytes and involves multi-level regulatory mechanisms, such as mitophagy, mitochondrial fission and fusion, mitochondrial energy metabolism, mitochondrial antioxidant system, and mitochondrial respiratory chain. Furthermore, MQC plays a role in the pathological mechanisms of various cardiovascular diseases (CVDs). In recent years, the regulatory effects of natural plants, drugs, and active ingredients on MQC in the context of CVDs have received significant attention. Effective active ingredients in natural drugs can influence the production of energy-supplying substances in the mitochondria, interfere with the expression of genes associated with mitochondrial energy requirements, and regulate various mechanisms of MQC modulation. Thus, these ingredients have therapeutic effects against CVDs. This review provides useful information about novel treatment options for CVDs and development of novel drugs targeting MQC."
                    }
                ]
            },
            "6eadd3c0-1932-4e3b-afd9-d2d4895264d3": {
                "pk": "6eadd3c0-1932-4e3b-afd9-d2d4895264d3",
                "name": "Zongpeng Zhang",
                "collaborators": [
                    "Zhouchen Lin",
                    "Mingqing Xiao",
                    "Qingyan Meng",
                    "Tong Lin",
                    "D.K. He",
                    "Taoyun Ji",
                    "Yuwu Jiang",
                    "Xiaohua Zhou",
                    "Zenan Ling",
                    "Yisen Wang",
                    "Xingyu Xie",
                    "Qiuhao Wang",
                    "Wen Wang",
                    "Guo-yuan Yu"
                ],
                "domain": [
                    "Spiking Neural Networks",
                    "Neuromorphic Computing",
                    "Machine Learning",
                    "Epilepsy Detection"
                ],
                "publications": [
                    {
                        "title": "Online Pseudo-Zeroth-Order Training of Neuromorphic Spiking Neural Networks",
                        "abstract": "Brain-inspired neuromorphic computing with spiking neural networks (SNNs) is a promising energy-efficient computational approach. However, successfully training SNNs in a more biologically plausible and neuromorphic-hardware-friendly way is still challenging. Most recent methods leverage spatial and temporal backpropagation (BP), not adhering to neuromorphic properties. Despite the efforts of some online training methods, tackling spatial credit assignments by alternatives with comparable performance as spatial BP remains a significant problem. In this work, we propose a novel method, online pseudo-zeroth-order (OPZO) training. Our method only requires a single forward propagation with noise injection and direct top-down signals for spatial credit assignment, avoiding spatial BP's problem of symmetric weights and separate phases for layer-by-layer forward-backward propagation. OPZO solves the large variance problem of zeroth-order methods by the pseudo-zeroth-order formulation and momentum feedback connections, while having more guarantees than random feedback. Combining online training, OPZO can pave paths to on-chip SNN training. Experiments on neuromorphic and static datasets with fully connected and convolutional networks demonstrate the effectiveness of OPZO with similar performance compared with spatial BP, as well as estimated low training costs."
                    },
                    {
                        "title": "Efficient and generalizable cross-patient epileptic seizure detection through a spiking neural network",
                        "abstract": "Introduction Epilepsy is a global chronic disease that brings pain and inconvenience to patients, and an electroencephalogram (EEG) is the main analytical tool. For clinical aid that can be applied to any patient, an automatic cross-patient epilepsy seizure detection algorithm is of great significance. Spiking neural networks (SNNs) are modeled on biological neurons and are energy-efficient on neuromorphic hardware, which can be expected to better handle brain signals and benefit real-world, low-power applications. However, automatic epilepsy seizure detection rarely considers SNNs. Methods In this article, we have explored SNNs for cross-patient seizure detection and discovered that SNNs can achieve comparable state-of-the-art performance or a performance that is even better than artificial neural networks (ANNs). We propose an EEG-based spiking neural network (EESNN) with a recurrent spiking convolution structure, which may better take advantage of temporal and biological characteristics in EEG signals. Results We extensively evaluate the performance of different SNN structures, training methods, and time settings, which builds a solid basis for understanding and evaluation of SNNs in seizure detection. Moreover, we show that our EESNN model can achieve energy reduction by several orders of magnitude compared with ANNs according to the theoretical estimation. Discussion These results show the potential for building high-performance, low-power neuromorphic systems for seizure detection and also broaden real-world application scenarios of SNNs."
                    },
                    {
                        "title": "Gradient Descent Optimizes Normalization-Free ResNets",
                        "abstract": "Recent empirical studies observe that even without normalization, a deep residual network can be trained reliably. We call such a structure as normalization-free Residual Networks (N-F ResNets), which add a learnable parameter $\\alpha$ to control the scale of the residual block instead of normalization. However, the theoretical understanding on N-F ResNets is still limited despite their empirical success. In this paper, we provide the first theoretical understanding of N-F ResNets from two perspectives. Firstly, we prove that the gradient descent (GD) algorithm can find the global minimum of the training loss at a linear rate for over-parameterized N-F ResNets. Secondly, we prove that N-F ResNets can avoid the gradient exploding or vanishing problem, by initializing the key parameter $\\alpha$ to be a small constant. Notably, we demonstrate that the gradients of N-F ResNets are more stable than those of ResNets with Kaiming initialization. Moreover, empirical experiments on benchmark datasets verify our theoretical results."
                    },
                    {
                        "title": "Supplementary Materials for: Training Feedback Spiking Neural Networks by Implicit Differentiation on the Equilibrium State",
                        "abstract": "Mingqing Xiao, Qingyan Meng, Zongpeng Zhang, Yisen Wang, Zhouchen Lin \u2217 Key Laboratory of Machine Perception (MoE), School of AI, Peking University The Chinese University of Hong Kong, Shenzhen Shenzhen Research Institute of Big Data Center for Data Science, Academy for Advanced Interdisciplinary Studies, Peking University Institute for Artificial Intelligence, Peking University Pazhou Lab, Guangzhou 510330, China {mingqing_xiao, yisen.wang, zlin}@pku.edu.cn, qingyanmeng@link.cuhk.edu.cn, zongpeng.zhang98@gmail.com"
                    },
                    {
                        "title": "Global Convergence of Over-parameterized Deep Equilibrium Models",
                        "abstract": "A deep equilibrium model (DEQ) is implicitly defined through an equilibrium point of an infinite-depth weight-tied model with an input-injection. Instead of infinite computations, it solves an equilibrium point directly with root-finding and computes gradients with implicit differentiation. The training dynamics of over-parameterized DEQs are investigated in this study. By supposing a condition on the initial equilibrium point, we show that the unique equilibrium point always exists during the training process, and the gradient descent is proved to converge to a globally optimal solution at a linear convergence rate for the quadratic loss function. In order to show that the required initial condition is satisfied via mild over-parameterization, we perform a fine-grained analysis on random DEQs. We propose a novel probabilistic framework to overcome the technical difficulty in the non-asymptotic analysis of infinite-depth weight-tied models."
                    },
                    {
                        "title": "Online Training Through Time for Spiking Neural Networks",
                        "abstract": "Spiking neural networks (SNNs) are promising brain-inspired energy-efficient models. Recent progress in training methods has enabled successful deep SNNs on large-scale tasks with low latency. Particularly, backpropagation through time (BPTT) with surrogate gradients (SG) is popularly used to achieve high performance in a very small number of time steps. However, it is at the cost of large memory consumption for training, lack of theoretical clarity for optimization, and inconsistency with the online property of biological learning and rules on neuromorphic hardware. Other works connect spike representations of SNNs with equivalent artificial neural network formulation and train SNNs by gradients from equivalent mappings to ensure descent directions. But they fail to achieve low latency and are also not online. In this work, we propose online training through time (OTTT) for SNNs, which is derived from BPTT to enable forward-in-time learning by tracking presynaptic activities and leveraging instantaneous loss and gradients. Meanwhile, we theoretically analyze and prove that gradients of OTTT can provide a similar descent direction for optimization as gradients based on spike representations under both feedforward and recurrent conditions. OTTT only requires constant training memory costs agnostic to time steps, avoiding the significant memory costs of BPTT for GPU training. Furthermore, the update rule of OTTT is in the form of three-factor Hebbian learning, which could pave a path for online on-chip learning. With OTTT, it is the first time that two mainstream supervised SNN training methods, BPTT with SG and spike representation-based training, are connected, and meanwhile in a biologically plausible form. Experiments on CIFAR-10, CIFAR-100, ImageNet, and CIFAR10-DVS demonstrate the superior performance of our method on large-scale static and neuromorphic datasets in small time steps."
                    },
                    {
                        "title": "Training Feedback Spiking Neural Networks by Implicit Differentiation on the Equilibrium State",
                        "abstract": "Spiking neural networks (SNNs) are brain-inspired models that enable energy-efficient implementation on neuromorphic hardware. However, the supervised training of SNNs remains a hard problem due to the discontinuity of the spiking neuron model. Most existing methods imitate the backpropagation framework and feedforward architectures for artificial neural networks, and use surrogate derivatives or compute gradients with respect to the spiking time to deal with the problem. These approaches either accumulate approximation errors or only propagate information limitedly through existing spikes, and usually require information propagation along time steps with large memory costs and biological implausibility. In this work, we consider feedback spiking neural networks, which are more brain-like, and propose a novel training method that does not rely on the exact reverse of the forward computation. First, we show that the average firing rates of SNNs with feedback connections would gradually evolve to an equilibrium state along time, which follows a fixed-point equation. Then by viewing the forward computation of feedback SNNs as a black-box solver for this equation, and leveraging the implicit differentiation on the equation, we can compute the gradient for parameters without considering the exact forward procedure. In this way, the forward and backward procedures are decoupled and therefore the problem of non-differentiable spiking functions is avoided. We also briefly discuss the biological plausibility of implicit differentiation, which only requires computing another equilibrium. Extensive experiments on MNIST, Fashion-MNIST, N-MNIST, CIFAR-10, and CIFAR-100 demonstrate the superior performance of our method for feedback models with fewer neurons and parameters in a small number of time steps. Our code is avaiable at https://github.com/pkuxmq/IDE-FSNN."
                    },
                    {
                        "title": "Biomedical Signal Processing and",
                        "abstract": "Cross-patient automatic epileptic seizure detection through electroencephalogram (EEG) is significant for clinical application and research. However, most automatic seizure detection methods are patient-specific and have poor generalization ability to unseen patients. In this paper, we consider two aspects for solving the cross-patient generalization problem in deep learning approaches. Firstly, we propose a new data augmentation method to largely improve generalization. We analyze the statistical distribution of seizure EEG signals and propose spatio-temporal EEG augmentation (STEA) to generate synthetic training seizure data with spatio-temporal dependencies in EEG. Secondly, we propose patient-adversarial neural network (PANN) to learn a patient-invariant representation for better generalization. We add a discriminator to distinguish the identity of patients and perform adversarial optimization between the feature extractor and identity discriminator, so that only shared seizure features are maintained. We conduct experiments on both public and clinical EEG datasets with various settings of different window lengths and divisions of training and testing sets. Our model significantly improves the performance of cross-patient detection compared with other methods and achieves state-of-the-art performance. Particularly, our proposed method achieves up to 95% sensitivity on CHB-MIT under 5 s window length segments, and achieves about 85% AUC and sensitivity even on the clinical dataset. Our designed methods can enable detection models extended to unseen patients, with performance achieving the clinical application standards."
                    }
                ]
            },
            "bb550bb6-1da5-44cc-bb7c-08e08bb29e05": {
                "pk": "bb550bb6-1da5-44cc-bb7c-08e08bb29e05",
                "name": "Di He",
                "collaborators": [
                    "Mingqing Xiao",
                    "Zhouchen Lin",
                    "Qingyan Meng",
                    "Zongpeng Zhang",
                    "Yixin Zhu"
                ],
                "domain": [
                    "Neuromorphic Computing",
                    "Spiking Neural Networks",
                    "Energy-efficient Computing",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Online Pseudo-Zeroth-Order Training of Neuromorphic Spiking Neural Networks",
                        "abstract": "Brain-inspired neuromorphic computing with spiking neural networks (SNNs) is a promising energy-efficient computational approach. However, successfully training SNNs in a more biologically plausible and neuromorphic-hardware-friendly way is still challenging. Most recent methods leverage spatial and temporal backpropagation (BP), not adhering to neuromorphic properties. Despite the efforts of some online training methods, tackling spatial credit assignments by alternatives with comparable performance as spatial BP remains a significant problem. In this work, we propose a novel method, online pseudo-zeroth-order (OPZO) training. Our method only requires a single forward propagation with noise injection and direct top-down signals for spatial credit assignment, avoiding spatial BP's problem of symmetric weights and separate phases for layer-by-layer forward-backward propagation. OPZO solves the large variance problem of zeroth-order methods by the pseudo-zeroth-order formulation and momentum feedback connections, while having more guarantees than random feedback. Combining online training, OPZO can pave paths to on-chip SNN training. Experiments on neuromorphic and static datasets with fully connected and convolutional networks demonstrate the effectiveness of OPZO with similar performance compared with spatial BP, as well as estimated low training costs."
                    },
                    {
                        "title": "Temporal Spiking Neural Networks with Synaptic Delay for Graph Reasoning",
                        "abstract": "Spiking neural networks (SNNs) are investigated as biologically inspired models of neural computation, distinguished by their computational capability and energy efficiency due to precise spiking times and sparse spikes with event-driven computation. A significant question is how SNNs can emulate human-like graph-based reasoning of concepts and relations, especially leveraging the temporal domain optimally. This paper reveals that SNNs, when amalgamated with synaptic delay and temporal coding, are proficient in executing (knowledge) graph reasoning. It is elucidated that spiking time can function as an additional dimension to encode relation properties via a neural-generalized path formulation. Empirical results highlight the efficacy of temporal delay in relation processing and showcase exemplary performance in diverse graph reasoning tasks. The spiking model is theoretically estimated to achieve $20\\times$ energy savings compared to non-spiking counterparts, deepening insights into the capabilities and potential of biologically inspired SNNs for efficient reasoning. The code is available at https://github.com/pkuxmq/GRSNN."
                    },
                    {
                        "title": "Online Training Through Time for Spiking Neural Networks",
                        "abstract": "Spiking neural networks (SNNs) are promising brain-inspired energy-efficient models. Recent progress in training methods has enabled successful deep SNNs on large-scale tasks with low latency. Particularly, backpropagation through time (BPTT) with surrogate gradients (SG) is popularly used to achieve high performance in a very small number of time steps. However, it is at the cost of large memory consumption for training, lack of theoretical clarity for optimization, and inconsistency with the online property of biological learning and rules on neuromorphic hardware. Other works connect spike representations of SNNs with equivalent artificial neural network formulation and train SNNs by gradients from equivalent mappings to ensure descent directions. But they fail to achieve low latency and are also not online. In this work, we propose online training through time (OTTT) for SNNs, which is derived from BPTT to enable forward-in-time learning by tracking presynaptic activities and leveraging instantaneous loss and gradients. Meanwhile, we theoretically analyze and prove that gradients of OTTT can provide a similar descent direction for optimization as gradients based on spike representations under both feedforward and recurrent conditions. OTTT only requires constant training memory costs agnostic to time steps, avoiding the significant memory costs of BPTT for GPU training. Furthermore, the update rule of OTTT is in the form of three-factor Hebbian learning, which could pave a path for online on-chip learning. With OTTT, it is the first time that two mainstream supervised SNN training methods, BPTT with SG and spike representation-based training, are connected, and meanwhile in a biologically plausible form. Experiments on CIFAR-10, CIFAR-100, ImageNet, and CIFAR10-DVS demonstrate the superior performance of our method on large-scale static and neuromorphic datasets in small time steps."
                    }
                ]
            },
            "31a4ff45-28de-4490-9749-8ec869366c12": {
                "pk": "31a4ff45-28de-4490-9749-8ec869366c12",
                "name": "Zhouchen Lin",
                "collaborators": [
                    "Mingqing Xiao",
                    "Yisen Wang",
                    "Zongpeng Zhang",
                    "Qingyan Meng",
                    "D.K. He",
                    "Mingjie Li",
                    "Taoyun Ji",
                    "Yuwu Jiang",
                    "Tong Lin",
                    "Xiaohua Zhou",
                    "Yichuan Mo",
                    "Dongxian Wu",
                    "Xingjun Ma",
                    "Lexiang Hu",
                    "Yixin Zhu",
                    "Pan Zhou",
                    "Xingyu Xie",
                    "Shuicheng Yan",
                    "Zhoutong Wu",
                    "Cong Fang",
                    "Yifei Wang",
                    "Liangchen Li",
                    "Jiansheng Yang",
                    "Wen Wang",
                    "Guo-yuan Yu"
                ],
                "domain": [
                    "Spiking Neural Networks",
                    "Adversarial Learning",
                    "Neuromorphic Computing",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Online Pseudo-Zeroth-Order Training of Neuromorphic Spiking Neural Networks",
                        "abstract": "Brain-inspired neuromorphic computing with spiking neural networks (SNNs) is a promising energy-efficient computational approach. However, successfully training SNNs in a more biologically plausible and neuromorphic-hardware-friendly way is still challenging. Most recent methods leverage spatial and temporal backpropagation (BP), not adhering to neuromorphic properties. Despite the efforts of some online training methods, tackling spatial credit assignments by alternatives with comparable performance as spatial BP remains a significant problem. In this work, we propose a novel method, online pseudo-zeroth-order (OPZO) training. Our method only requires a single forward propagation with noise injection and direct top-down signals for spatial credit assignment, avoiding spatial BP's problem of symmetric weights and separate phases for layer-by-layer forward-backward propagation. OPZO solves the large variance problem of zeroth-order methods by the pseudo-zeroth-order formulation and momentum feedback connections, while having more guarantees than random feedback. Combining online training, OPZO can pave paths to on-chip SNN training. Experiments on neuromorphic and static datasets with fully connected and convolutional networks demonstrate the effectiveness of OPZO with similar performance compared with spatial BP, as well as estimated low training costs."
                    },
                    {
                        "title": "On the Adversarial Transferability of Generalized\"Skip Connections\"",
                        "abstract": "Skip connection is an essential ingredient for modern deep models to be deeper and more powerful. Despite their huge success in normal scenarios (state-of-the-art classification performance on natural examples), we investigate and identify an interesting property of skip connections under adversarial scenarios, namely, the use of skip connections allows easier generation of highly transferable adversarial examples. Specifically, in ResNet-like models (with skip connections), we find that using more gradients from the skip connections rather than the residual modules according to a decay factor during backpropagation allows one to craft adversarial examples with high transferability. The above method is termed as Skip Gradient Method (SGM). Although starting from ResNet-like models in vision domains, we further extend SGM to more advanced architectures, including Vision Transformers (ViTs) and models with length-varying paths and other domains, i.e. natural language processing. We conduct comprehensive transfer attacks against various models including ResNets, Transformers, Inceptions, Neural Architecture Search, and Large Language Models (LLMs). We show that employing SGM can greatly improve the transferability of crafted attacks in almost all cases. Furthermore, considering the big complexity for practical use, we further demonstrate that SGM can even improve the transferability on ensembles of models or targeted attacks and the stealthiness against current defenses. At last, we provide theoretical explanations and empirical insights on how SGM works. Our findings not only motivate new adversarial research into the architectural characteristics of models but also open up further challenges for secure model architecture design. Our code is available at https://github.com/mo666666/SGM."
                    },
                    {
                        "title": "EKAN: Equivariant Kolmogorov-Arnold Networks",
                        "abstract": "Kolmogorov-Arnold Networks (KANs) have seen great success in scientific domains thanks to spline activation functions, becoming an alternative to Multi-Layer Perceptrons (MLPs). However, spline functions may not respect symmetry in tasks, which is crucial prior knowledge in machine learning. Previously, equivariant networks embed symmetry into their architectures, achieving better performance in specific applications. Among these, Equivariant Multi-Layer Perceptrons (EMLP) introduce arbitrary matrix group equivariance into MLPs, providing a general framework for constructing equivariant networks layer by layer. In this paper, we propose Equivariant Kolmogorov-Arnold Networks (EKAN), a method for incorporating matrix group equivariance into KANs, aiming to broaden their applicability to more fields. First, we construct gated spline basis functions, which form the EKAN layer together with equivariant linear weights. We then define a lift layer to align the input space of EKAN with the feature space of the dataset, thereby building the entire EKAN architecture. Compared with baseline models, EKAN achieves higher accuracy with smaller datasets or fewer parameters on symmetry-related tasks, such as particle scattering and the three-body problem, often reducing test MSE by several orders of magnitude. Even in non-symbolic formula scenarios, such as top quark tagging with three jet constituents, EKAN achieves comparable results with EMLP using only $26\\%$ of the parameters, while KANs do not outperform MLPs as expected."
                    },
                    {
                        "title": "Temporal Spiking Neural Networks with Synaptic Delay for Graph Reasoning",
                        "abstract": "Spiking neural networks (SNNs) are investigated as biologically inspired models of neural computation, distinguished by their computational capability and energy efficiency due to precise spiking times and sparse spikes with event-driven computation. A significant question is how SNNs can emulate human-like graph-based reasoning of concepts and relations, especially leveraging the temporal domain optimally. This paper reveals that SNNs, when amalgamated with synaptic delay and temporal coding, are proficient in executing (knowledge) graph reasoning. It is elucidated that spiking time can function as an additional dimension to encode relation properties via a neural-generalized path formulation. Empirical results highlight the efficacy of temporal delay in relation processing and showcase exemplary performance in diverse graph reasoning tasks. The spiking model is theoretically estimated to achieve $20\\times$ energy savings compared to non-spiking counterparts, deepening insights into the capabilities and potential of biologically inspired SNNs for efficient reasoning. The code is available at https://github.com/pkuxmq/GRSNN."
                    },
                    {
                        "title": "Towards Understanding Convergence and Generalization of AdamW",
                        "abstract": "AdamW modifies Adam by adding a decoupled weight decay to decay network weights per training iteration. For adaptive algorithms, this decoupled weight decay does not affect specific optimization steps, and differs from the widely used <inline-formula><tex-math notation=\"LaTeX\">$\\ell _{2}$</tex-math><alternatives><mml:math><mml:msub><mml:mi>\u2113</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math><inline-graphic xlink:href=\"zhou-ieq1-3382294.gif\"/></alternatives></inline-formula>-regularizer which changes optimization steps via changing the first- and second-order gradient moments. Despite its great practical success, for AdamW, its convergence behavior and generalization improvement over Adam and <inline-formula><tex-math notation=\"LaTeX\">$\\ell _{2}$</tex-math><alternatives><mml:math><mml:msub><mml:mi>\u2113</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math><inline-graphic xlink:href=\"zhou-ieq2-3382294.gif\"/></alternatives></inline-formula>-regularized Adam (<inline-formula><tex-math notation=\"LaTeX\">$\\ell _{2}$</tex-math><alternatives><mml:math><mml:msub><mml:mi>\u2113</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math><inline-graphic xlink:href=\"zhou-ieq3-3382294.gif\"/></alternatives></inline-formula>-Adam) remain absent yet. To solve this issue, we prove the convergence of AdamW and justify its generalization advantages over Adam and <inline-formula><tex-math notation=\"LaTeX\">$\\ell _{2}$</tex-math><alternatives><mml:math><mml:msub><mml:mi>\u2113</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math><inline-graphic xlink:href=\"zhou-ieq4-3382294.gif\"/></alternatives></inline-formula>-Adam. Specifically, AdamW provably converges but minimizes a dynamically regularized loss that combines vanilla loss and a dynamical regularization induced by decoupled weight decay, thus yielding different behaviors with Adam and <inline-formula><tex-math notation=\"LaTeX\">$\\ell _{2}$</tex-math><alternatives><mml:math><mml:msub><mml:mi>\u2113</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math><inline-graphic xlink:href=\"zhou-ieq5-3382294.gif\"/></alternatives></inline-formula>-Adam. Moreover, on both general nonconvex problems and P\u0141-conditioned problems, we establish stochastic gradient complexity of AdamW to find a stationary point. Such complexity is also applicable to Adam and <inline-formula><tex-math notation=\"LaTeX\">$\\ell _{2}$</tex-math><alternatives><mml:math><mml:msub><mml:mi>\u2113</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math><inline-graphic xlink:href=\"zhou-ieq6-3382294.gif\"/></alternatives></inline-formula>-Adam, and improves their previously known complexity, especially for over-parametrized networks. Besides, we prove that AdamW enjoys smaller generalization errors than Adam and <inline-formula><tex-math notation=\"LaTeX\">$\\ell _{2}$</tex-math><alternatives><mml:math><mml:msub><mml:mi>\u2113</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math><inline-graphic xlink:href=\"zhou-ieq7-3382294.gif\"/></alternatives></inline-formula>-Adam from the Bayesian posterior aspect. This result, for the first time, explicitly reveals the benefits of decoupled weight decay in AdamW. Experimental results validate our theory."
                    },
                    {
                        "title": "Designing Universally-Approximating Deep Neural Networks: A First-Order Optimization Approach",
                        "abstract": "Universal approximation capability, also referred to as universality, is an important property of deep neural networks, endowing them with the potency to accurately represent the underlying target function in learning tasks. In practice, the architecture of deep neural networks largely influences the performance of the models. However, most existing methodologies for designing neural architectures, such as the heuristic manual design or neural architecture search, ignore the universal approximation property, thus losing a potential safeguard about the performance. In this paper, we propose a unified framework to design the architectures of deep neural networks with a universality guarantee based on first-order optimization algorithms, where the forward pass is interpreted as the updates of an optimization algorithm. The (explicit or implicit) network is designed by replacing each gradient term in the algorithm with a learnable module similar to a two-layer network or its derivatives. Specifically, we explore the realm of width-bounded neural networks, a common practical scenario, showcasing their universality. Moreover, adding operations of normalization, downsampling, and upsampling does not hurt the universality. To the best of our knowledge, this is the first work that width-bounded networks with universal approximation guarantee can be designed in a principled way. Our framework can inspire a variety of neural architectures including some renowned structures such as ResNet and DenseNet, as well as novel innovations. The experimental results on image classification problems demonstrate that the newly inspired networks are competitive and surpass the baselines of ResNet, DenseNet, as well as the advanced ConvNeXt and ViT, testifying to the effectiveness of our framework."
                    },
                    {
                        "title": "Efficient and generalizable cross-patient epileptic seizure detection through a spiking neural network",
                        "abstract": "Introduction Epilepsy is a global chronic disease that brings pain and inconvenience to patients, and an electroencephalogram (EEG) is the main analytical tool. For clinical aid that can be applied to any patient, an automatic cross-patient epilepsy seizure detection algorithm is of great significance. Spiking neural networks (SNNs) are modeled on biological neurons and are energy-efficient on neuromorphic hardware, which can be expected to better handle brain signals and benefit real-world, low-power applications. However, automatic epilepsy seizure detection rarely considers SNNs. Methods In this article, we have explored SNNs for cross-patient seizure detection and discovered that SNNs can achieve comparable state-of-the-art performance or a performance that is even better than artificial neural networks (ANNs). We propose an EEG-based spiking neural network (EESNN) with a recurrent spiking convolution structure, which may better take advantage of temporal and biological characteristics in EEG signals. Results We extensively evaluate the performance of different SNN structures, training methods, and time settings, which builds a solid basis for understanding and evaluation of SNNs in seizure detection. Moreover, we show that our EESNN model can achieve energy reduction by several orders of magnitude compared with ANNs according to the theoretical estimation. Discussion These results show the potential for building high-performance, low-power neuromorphic systems for seizure detection and also broaden real-world application scenarios of SNNs."
                    },
                    {
                        "title": "GEQ: Gaussian Kernel Inspired Equilibrium Models",
                        "abstract": "Despite the connection established by optimization-induced deep equilibrium models (OptEqs) between their output and the underlying hidden optimization problems, the performance of it along with its related works is still not good enough especially when compared to deep networks. One key factor responsible for this performance limitation is the use of linear kernels to extract features in these models. To address this issue, we propose a novel approach by replacing its linear kernel with a new function that can readily capture nonlinear feature dependencies in the input data. Drawing inspiration from classical machine learning algorithms, we introduce Gaussian kernels as the alternative function and then propose our new equilibrium model, which we refer to as GEQ. By leveraging Gaussian kernels, GEQ can effectively extract the nonlinear information embedded within the input features, surpassing the performance of the original OptEqs. Moreover, GEQ can be perceived as a weight-tied neural network with infinite width and depth. GEQ also enjoys better theoretical properties and improved overall performance. Additionally, our GEQ exhibits enhanced stability when confronted with various samples. We further substantiate the effectiveness and stability of GEQ through a series of comprehensive experiments."
                    },
                    {
                        "title": "Balance, Imbalance, and Rebalance: Understanding Robust Overfitting from a Minimax Game Perspective",
                        "abstract": "Adversarial Training (AT) has become arguably the state-of-the-art algorithm for extracting robust features. However, researchers recently notice that AT suffers from severe robust overfitting problems, particularly after learning rate (LR) decay. In this paper, we explain this phenomenon by viewing adversarial training as a dynamic minimax game between the model trainer and the attacker. Specifically, we analyze how LR decay breaks the balance between the minimax game by empowering the trainer with a stronger memorization ability, and show such imbalance induces robust overfitting as a result of memorizing non-robust features. We validate this understanding with extensive experiments, and provide a holistic view of robust overfitting from the dynamics of both the two game players. This understanding further inspires us to alleviate robust overfitting by rebalancing the two players by either regularizing the trainer's capacity or improving the attack strength. Experiments show that the proposed ReBalanced Adversarial Training (ReBAT) can attain good robustness and does not suffer from robust overfitting even after very long training. Code is available at https://github.com/PKU-ML/ReBAT."
                    },
                    {
                        "title": "Supplementary Materials for: Training Feedback Spiking Neural Networks by Implicit Differentiation on the Equilibrium State",
                        "abstract": "Mingqing Xiao, Qingyan Meng, Zongpeng Zhang, Yisen Wang, Zhouchen Lin \u2217 Key Laboratory of Machine Perception (MoE), School of AI, Peking University The Chinese University of Hong Kong, Shenzhen Shenzhen Research Institute of Big Data Center for Data Science, Academy for Advanced Interdisciplinary Studies, Peking University Institute for Artificial Intelligence, Peking University Pazhou Lab, Guangzhou 510330, China {mingqing_xiao, yisen.wang, zlin}@pku.edu.cn, qingyanmeng@link.cuhk.edu.cn, zongpeng.zhang98@gmail.com"
                    },
                    {
                        "title": "Biomedical Signal Processing and",
                        "abstract": "Cross-patient automatic epileptic seizure detection through electroencephalogram (EEG) is significant for clinical application and research. However, most automatic seizure detection methods are patient-specific and have poor generalization ability to unseen patients. In this paper, we consider two aspects for solving the cross-patient generalization problem in deep learning approaches. Firstly, we propose a new data augmentation method to largely improve generalization. We analyze the statistical distribution of seizure EEG signals and propose spatio-temporal EEG augmentation (STEA) to generate synthetic training seizure data with spatio-temporal dependencies in EEG. Secondly, we propose patient-adversarial neural network (PANN) to learn a patient-invariant representation for better generalization. We add a discriminator to distinguish the identity of patients and perform adversarial optimization between the feature extractor and identity discriminator, so that only shared seizure features are maintained. We conduct experiments on both public and clinical EEG datasets with various settings of different window lengths and divisions of training and testing sets. Our model significantly improves the performance of cross-patient detection compared with other methods and achieves state-of-the-art performance. Particularly, our proposed method achieves up to 95% sensitivity on CHB-MIT under 5 s window length segments, and achieves about 85% AUC and sensitivity even on the clinical dataset. Our designed methods can enable detection models extended to unseen patients, with performance achieving the clinical application standards."
                    }
                ]
            }
        }
    },
    "2410.02430": {
        "paper_data": {
            "title": "Predictive Attractor Models",
            "url": "http://arxiv.org/abs/2410.02430v1",
            "arxiv_id": "2410.02430",
            "authors": [
                "Ramy Mounir",
                "Sudeep Sarkar"
            ],
            "abstract": "Sequential memory, the ability to form and accurately recall a sequence of events or stimuli in the correct order, is a fundamental prerequisite for biological and artificial intelligence as it underpins numerous cognitive functions (e.g., language comprehension, planning, episodic memory formation, etc.) However, existing methods of sequential memory suffer from catastrophic forgetting, limited capacity, slow iterative learning procedures, low-order Markov memory, and, most importantly, the inability to represent and generate multiple valid future possibilities stemming from the same context. Inspired by biologically plausible neuroscience theories of cognition, we propose \\textit{Predictive Attractor Models (PAM)}, a novel sequence memory architecture with desirable generative properties. PAM is a streaming model that learns a sequence in an online, continuous manner by observing each input \\textit{only once}. Additionally, we find that PAM avoids catastrophic forgetting by uniquely representing past context through lateral inhibition in cortical minicolumns, which prevents new memories from overwriting previously learned knowledge. PAM generates future predictions by sampling from a union set of predicted possibilities; this generative ability is realized through an attractor model trained alongside the predictor. We show that PAM is trained with local computations through Hebbian plasticity rules in a biologically plausible framework. Other desirable traits (e.g., noise tolerance, CPU-based learning, capacity scaling) are discussed throughout the paper. Our findings suggest that PAM represents a significant step forward in the pursuit of biologically plausible and computationally efficient sequential memory models, with broad implications for cognitive science and artificial intelligence research.",
            "introduction": "   1 Introduction  Modeling the temporal associations between consecutive inputs in a sequence (i.e., sequential memory) enables biological agents to perform various cognitive functions, such as episodic memory formation\u00a0[45, 55, 14], complex action planning\u00a0[20] and translating between languages\u00a0[3]. For example, playing a musical instrument requires remembering the sequence of notes in a piece of music; similarly, playing a game of chess requires simulating, planning, and executing a sequence of moves in a specific order. While the ability to form such memories of static, unrelated events has been extensively studied\u00a0[48, 47, 54, 59, 57], the ability of biologically-plausible artificial networks to learn and recall temporally-dependent patterns has not been sufficiently explored in literature\u00a0[53]. The task of sequential memory is considered challenging for models operating under biological constraints (i.e., local synaptic computations) for many reasons, including catastrophic forgetting, ambiguous context representations, multiple future possibilities, etc.   In addition to the biological constraint, we impose the following set of desirable characteristics as learning constraints on sequence memory models.     \u2022  The learning of one sequence does not overwrite the previously learned sequences. This property is defined under the continual learning framework as avoiding catastrophic forgetting and evaluated with the Backward Transfer (BWT) metric\u00a0[34].    \u2022  The model should uniquely encode inputs based on their context in a sequence. Consider the sequence of letters \u201cEVER\u201d; the representation of \u201cE\u201d at position 1 should be different from \u201cE\u201d at position 3, thus resulting in different predictions: \u201cV\u201d and \u201cR\u201d. Moreover, the representation of \u201cE\u201d at position 3 in \u201cEVER\u201d should be different from \u201cE\u201d at position 3 in \u201cCLEVER\u201d. Therefore, positional encoding is not a valid solution.    \u2022  When presented with multiple valid, future possibilities, the model should learn to represent each possibility separately yet stochastically sample a single valid possibility. Consider the two sequences \u201cTHAT\u201d and \u201cTHEY\u201d; after seeing \u201cTH\u201d, the model should learn to generate either \u201cA\u201d or \u201cE\u201d, but not an average\u00a0[33] or a union of both\u00a0[23].    \u2022  The model should be capable of incrementally learning each transition without seeing the whole sequence or revisiting older sequence transitions that are previously learned. This property falls under online learning constraints, also called stream learning\u00a0[40, 23].    \u2022  The learning algorithm should be tolerant and robust to significant input noise. A model should continuously clean the noisy inputs using learned priors and beliefs, thus performing future predictions based on the noise-free observations\u00a0[23].      We propose Predictive Attractor Models (PAM), which consists of a state prediction model and a generative attractor model. The predictor in PAM is inspired by the Hierarchical Temporal Memory (HTM)\u00a0[23] learning algorithm, where a group of neurons in the same cortical minicolumn share the same receptive feedforward connection from the sensory input on their proximal dendrites. The depolarization of the voltage of any neuron in a single minicolumn (i.e., on distal dendrites) primes this depolarized neuron to fire first while inhibiting all the other neurons in the same minicolumn from firing (i.e., competitive learning). The choice of which neurons fire within the same minicolumn is based on the previously active neurons and their trainable synaptic weights to the depolarized neurons, which gives rise to a unique context representation for every input.",
            "references": [
                {
                    "title": "Revisiting Feature Prediction for Learning Visual Representations from Video",
                    "abstract": "This paper explores feature prediction as a stand-alone objective for unsupervised learning from video and introduces V-JEPA, a collection of vision models trained solely using a feature prediction objective, without the use of pretrained image encoders, text, negative examples, reconstruction, or other sources of supervision. The models are trained on 2 million videos collected from public datasets and are evaluated on downstream image and video tasks. Our results show that learning by predicting video features leads to versatile visual representations that perform well on both motion and appearance-based tasks, without adaption of the model's parameters; e.g., using a frozen backbone. Our largest model, a ViT-H/16 trained only on videos, obtains 81.9% on Kinetics-400, 72.2% on Something-Something-v2, and 77.9% on ImageNet1K."
                },
                {
                    "title": "Brain-Inspired Computational Intelligence via Predictive Coding",
                    "abstract": "Artificial intelligence (AI) is rapidly becoming one of the key technologies of this century. The majority of results in AI thus far have been achieved using deep neural networks trained with the error backpropagation learning algorithm. However, the ubiquitous adoption of this approach has highlighted some important limitations such as substantial computational cost, difficulty in quantifying uncertainty, lack of robustness, unreliability, and biological implausibility. It is possible that addressing these limitations may require schemes that are inspired and guided by neuroscience theories. One such theory, called predictive coding (PC), has shown promising performance in machine intelligence tasks, exhibiting exciting properties that make it potentially valuable for the machine learning community: PC can model information processing in different brain areas, can be used in cognitive control and robotics, and has a solid mathematical grounding in variational inference, offering a powerful inversion scheme for a specific class of continuous-state generative models. With the hope of foregrounding research in this direction, we survey the literature that has contributed to this perspective, highlighting the many ways that PC might play a role in the future of machine learning and computational intelligence at large."
                },
                {
                    "title": "MC-JEPA: A Joint-Embedding Predictive Architecture for Self-Supervised Learning of Motion and Content Features",
                    "abstract": "Self-supervised learning of visual representations has been focusing on learning content features, which do not capture object motion or location, and focus on identifying and differentiating objects in images and videos. On the other hand, optical flow estimation is a task that does not involve understanding the content of the images on which it is estimated. We unify the two approaches and introduce MC-JEPA, a joint-embedding predictive architecture and self-supervised learning approach to jointly learn optical flow and content features within a shared encoder, demonstrating that the two associated objectives; the optical flow estimation objective and the self-supervised learning objective; benefit from each other and thus learn content features that incorporate motion information. The proposed approach achieves performance on-par with existing unsupervised optical flow benchmarks, as well as with common self-supervised learning approaches on downstream tasks such as semantic segmentation of images and videos."
                },
                {
                    "title": "Long Sequence Hopfield Memory",
                    "abstract": "\n Sequence memory is an essential attribute of natural and artificial intelligence that enables agents to encode, store, and retrieve complex sequences of stimuli and actions. Computational models of sequence memory have been proposed where recurrent Hopfield-like neural networks are trained with temporally asymmetric Hebbian rules. However, these networks suffer from limited sequence capacity (maximal length of the stored sequence) due to interference between the memories. Inspired by recent work on Dense Associative Memories, we expand the sequence capacity of these models by introducing a nonlinear interaction term, enhancing separation between the patterns. We derive novel scaling laws for sequence capacity with respect to network size, significantly outperforming existing scaling laws for models based on traditional Hopfield networks, and verify these theoretical results with numerical simulation. Moreover, we introduce a generalized pseudoinverse rule to recall sequences of highly correlated patterns. Finally, we extend this model to store sequences with variable timing between states\u2019 transitions and describe a biologically-plausible implementation, with connections to motor neuroscience."
                },
                {
                    "title": "Sequential Memory with Temporal Predictive Coding",
                    "abstract": "Forming accurate memory of sequential stimuli is a fundamental function of biological agents. However, the computational mechanism underlying sequential memory in the brain remains unclear. Inspired by neuroscience theories and recent successes in applying predictive coding (PC) to static memory tasks, in this work we propose a novel PC-based model for sequential memory, called temporal predictive coding (tPC). We show that our tPC models can memorize and retrieve sequential inputs accurately with a biologically plausible neural implementation. Importantly, our analytical study reveals that tPC can be viewed as a classical Asymmetric Hopfield Network (AHN) with an implicit statistical whitening process, which leads to more stable performance in sequential memory tasks of structured inputs. Moreover, we find that tPC exhibits properties consistent with behavioral observations and theories in neuroscience, thereby strengthening its biological relevance. Our work establishes a possible computational mechanism underlying sequential memory in the brain that can also be theoretically interpreted using existing memory model frameworks."
                },
                {
                    "title": "Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture",
                    "abstract": "This paper demonstrates an approach for learning highly semantic image representations without relying on hand-crafted data-augmentations. We introduce the Image-based Joint-Embedding Predictive Architecture (I-JEPA), a non-generative approach for self-supervised learning from images. The idea behind I-JEPA is simple: from a single context block, predict the representations of various target blocks in the same image. A core design choice to guide I-JEPA towards producing semantic representations is the masking strategy; specifically, it is crucial to (a) sample target blocks with sufficiently large scale (semantic), and to (b) use a sufficiently informative (spatially distributed) context block. Empirically, when combined with Vision Transformers, we find I-JEPA to be highly scalable. For instance, we train a ViT-Huge/14 on ImageNet using 16 A100 GPUs in under 72 hours to achieve strong downstream performance across a wide range of tasks, from linear classification to object counting and depth prediction."
                },
                {
                    "title": "Recurrent predictive coding models for associative memory employing covariance learning",
                    "abstract": "The computational principles adopted by the hippocampus in associative memory (AM) tasks have been one of the mostly studied topics in computational and theoretical neuroscience. Classical models of the hippocampal network assume that AM is performed via a form of covariance learning, where associations between memorized items are represented by entries in the learned covariance matrix encoded in the recurrent connections in the hippocampal subfield CA3. On the other hand, it has been recently proposed that AM in the hippocampus is achieved through predictive coding. Hierarchical predictive coding models following this theory perform AM, but fail to capture the recurrent hippocampal structure that encodes the covariance in the classical models. Such a dichotomy pose potential difficulties for developing a unitary theory of how memory is formed and recalled in the hippocampus. Earlier predictive coding models that learn the covariance information of inputs explicitly seem to be a solution to this dichotomy. Here, we show that although these models can perform AM, they do it in an implausible and numerically unstable way. Instead, we propose alternatives to these earlier covariance-learning predictive coding networks, which learn the covariance information implicitly and plausibly, and can use dendritic structures to encode prediction errors. We show analytically that our proposed models are perfectly equivalent to the earlier predictive coding model learning covariance explicitly, and encounter no numerical issues when performing AM tasks in practice. We further show that our models can be combined with hierarchical predictive coding networks to model the hippocampo-neocortical interactions. Our models provide a biologically plausible approach to modelling the hippocampal network, pointing to a potential computational mechanism employed by the hippocampus during memory formation and recall, which unifies predictive coding and covariance learning based on the recurrent network structure. Author summary The hippocampus and adjacent cortical areas have long been considered essential for the formation of associative memories. Earlier theoretical works have assumed that the hippocampus stores in its recurrent connections statistical regularities embedded in the sensory inputs. On the other hand, it has been recently suggested that the hippocampus retrieves memory by generating predictions of ongoing sensory inputs. Computational models have thus been proposed to account for this predictive nature of the hippocampal network using predictive coding, a general theory of information processing in the cortex. However, these hierarchical predictive coding models of the hippocampus did not describe how it stores the statistical regularities that play a key role for associative memory in the classical hippocampal models, hindering a unified understanding of the underlying computational principles employed by the hippocampus. To address this dichotomy, here we present a family of predictive coding models that also learn the statistical information needed for associative memory. Our models can stably perform associative memory tasks in a biologically plausible manner, even with large structured data such as natural scenes. Our work provides a possible mechanism of how the recurrent hippocampal network may employ various computational principles concurrently to perform associative memory."
                },
                {
                    "title": "BayesPCN: A Continually Learnable Predictive Coding Associative Memory",
                    "abstract": "Associative memory plays an important role in human intelligence and its mechanisms have been linked to attention in machine learning. While the machine learning community's interest in associative memories has recently been rekindled, most work has focused on memory recall ($read$) over memory learning ($write$). In this paper, we present BayesPCN, a hierarchical associative memory capable of performing continual one-shot memory writes without meta-learning. Moreover, BayesPCN is able to gradually forget past observations ($forget$) to free its memory. Experiments show that BayesPCN can recall corrupted i.i.d. high-dimensional data observed hundreds to a thousand ``timesteps'' ago without a large drop in recall ability compared to the state-of-the-art offline-learned parametric memory models."
                },
                {
                    "title": "Universal Hopfield Networks: A General Framework for Single-Shot Associative Memory Models",
                    "abstract": "A large number of neural network models of associative memory have been proposed in the literature. These include the classical Hopfield networks (HNs), sparse distributed memories (SDMs), and more recently the modern continuous Hopfield networks (MCHNs), which possess close links with self-attention in machine learning. In this paper, we propose a general framework for understanding the operation of such memory networks as a sequence of three operations: similarity, separation, and projection. We derive all these memory models as instances of our general framework with differing similarity and separation functions. We extend the mathematical framework of Krotov & Hopfield (2020) to express general associative memory models using neural network dynamics with local computation, and derive a general energy function that is a Lyapunov function of the dynamics. Finally, using our framework, we empirically investigate the capacity of using different similarity functions for these associative memory models, beyond the dot product similarity measure, and demonstrate empirically that Euclidean or Manhattan distance similarity metrics perform substantially better in practice on many tasks, enabling a more robust retrieval and higher memory capacity than existing models."
                },
                {
                    "title": "Learning on Arbitrary Graph Topologies via Predictive Coding",
                    "abstract": "Training with backpropagation (BP) in standard deep learning consists of two main steps: a forward pass that maps a data point to its prediction, and a backward pass that propagates the error of this prediction back through the network. This process is highly effective when the goal is to minimize a specific objective function. However, it does not allow training on networks with cyclic or backward connections. This is an obstacle to reaching brain-like capabilities, as the highly complex heterarchical structure of the neural connections in the neocortex are potentially fundamental for its effectiveness. In this paper, we show how predictive coding (PC), a theory of information processing in the cortex, can be used to perform inference and learning on arbitrary graph topologies. We experimentally show how this formulation, called PC graphs, can be used to flexibly perform different tasks with the same network by simply stimulating specific neurons. This enables the model to be queried on stimuli with different structures, such as partial images, images with labels, or images without labels. We conclude by investigating how the topology of the graph influences the final performance, and comparing against simple baselines trained with BP."
                },
                {
                    "title": "Associative Memories via Predictive Coding",
                    "abstract": "Associative memories in the brain receive and store patterns of activity registered by the sensory neurons, and are able to retrieve them when necessary. Due to their importance in human intelligence, computational models of associative memories have been developed for several decades now. In this paper, we present a novel neural model for realizing associative memories, which is based on a hierarchical generative network that receives external stimuli via sensory neurons. It is trained using predictive coding, an error-based learning algorithm inspired by information processing in the cortex. To test the model's capabilities, we perform multiple retrieval experiments from both corrupted and incomplete data points. In an extensive comparison, we show that this new model outperforms in retrieval accuracy and robustness popular associative memory models, such as autoencoders trained via backpropagation, and modern Hopfield networks. In particular, in completing partial data points, our model achieves remarkable results on natural image datasets, such as ImageNet, with a surprisingly high accuracy, even when only a tiny fraction of pixels of the original images is presented. Our model provides a plausible framework to study learning and retrieval of memories in the brain, as it closely mimics the behavior of the hippocampus as a memory index and generative model."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "Barlow Twins: Self-Supervised Learning via Redundancy Reduction",
                    "abstract": "Self-supervised learning (SSL) is rapidly closing the gap with supervised methods on large computer vision benchmarks. A successful approach to SSL is to learn embeddings which are invariant to distortions of the input sample. However, a recurring issue with this approach is the existence of trivial constant solutions. Most current methods avoid such solutions by careful implementation details. We propose an objective function that naturally avoids collapse by measuring the cross-correlation matrix between the outputs of two identical networks fed with distorted versions of a sample, and making it as close to the identity matrix as possible. This causes the embedding vectors of distorted versions of a sample to be similar, while minimizing the redundancy between the components of these vectors. The method is called Barlow Twins, owing to neuroscientist H. Barlow's redundancy-reduction principle applied to a pair of identical networks. Barlow Twins does not require large batches nor asymmetry between the network twins such as a predictor network, gradient stopping, or a moving average on the weight updates. Intriguingly it benefits from very high-dimensional output vectors. Barlow Twins outperforms previous methods on ImageNet for semi-supervised classification in the low-data regime, and is on par with current state of the art for ImageNet classification with a linear classifier head, and for transfer tasks of classification and object detection."
                },
                {
                    "title": "Exploring Simple Siamese Representation Learning",
                    "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code is made available.1"
                },
                {
                    "title": "Hopfield Networks is All You Need",
                    "abstract": "We show that the transformer attention mechanism is the update rule of a modern Hopfield network with continuous states. This new Hopfield network can store exponentially (with the dimension) many patterns, converges with one update, and has exponentially small retrieval errors. The number of stored patterns is traded off against convergence speed and retrieval error. The new Hopfield network has three types of energy minima (fixed points of the update): (1) global fixed point averaging over all patterns, (2) metastable states averaging over a subset of patterns, and (3) fixed points which store a single pattern. Transformer and BERT models operate in their first layers preferably in the global averaging regime, while they operate in higher layers in metastable states. The gradient in transformers is maximal for metastable states, is uniformly distributed for global averaging, and vanishes for a fixed point near a stored pattern. Using the Hopfield network interpretation, we analyzed learning of transformer and BERT models. Learning starts with attention heads that average and then most of them switch to metastable states. However, the majority of heads in the first layers still averages and can be replaced by averaging, e.g. our proposed Gaussian weighting. In contrast, heads in the last layers steadily learn and seem to use metastable states to collect information created in lower layers. These heads seem to be a promising target for improving transformers. Neural networks with Hopfield networks outperform other methods on immune repertoire classification, where the Hopfield net stores several hundreds of thousands of patterns. We provide a new PyTorch layer called \"Hopfield\", which allows to equip deep learning architectures with modern Hopfield networks as a new powerful concept comprising pooling, memory, and attention. GitHub: this https URL"
                },
                {
                    "title": "Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning",
                    "abstract": "We introduce Bootstrap Your Own Latent (BYOL), a new approach to self-supervised image representation learning. BYOL relies on two neural networks, referred to as online and target networks, that interact and learn from each other. From an augmented view of an image, we train the online network to predict the target network representation of the same image under a different augmented view. At the same time, we update the target network with a slow-moving average of the online network. While state-of-the art methods rely on negative pairs, BYOL achieves a new state of the art without them. BYOL reaches $74.3\\%$ top-1 classification accuracy on ImageNet using a linear evaluation with a ResNet-50 architecture and $79.6\\%$ with a larger ResNet. We show that BYOL performs on par or better than the current state of the art on both transfer and semi-supervised benchmarks. Our implementation and pretrained models are given on GitHub."
                },
                {
                    "title": "CLEVRER: CoLlision Events for Video REpresentation and Reasoning",
                    "abstract": "The ability to reason about temporal and causal events from videos lies at the core of human intelligence. Most video reasoning benchmarks, however, focus on pattern recognition from complex visual and language input, instead of on causal structure. We study the complementary problem, exploring the temporal and causal structures behind videos of objects with simple visual appearance. To this end, we introduce the CoLlision Events for Video REpresentation and Reasoning (CLEVRER), a diagnostic video dataset for systematic evaluation of computational models on a wide range of reasoning tasks. Motivated by the theory of human casual judgment, CLEVRER includes four types of questions: descriptive (e.g., \"what color\"), explanatory (\"what is responsible for\"), predictive (\"what will happen next\"), and counterfactual (\"what if\"). We evaluate various state-of-the-art models for visual reasoning on our benchmark. While these models thrive on the perception-based task (descriptive), they perform poorly on the causal tasks (explanatory, predictive and counterfactual), suggesting that a principled approach for causal reasoning should incorporate the capability of both perceiving complex visual and language inputs, and understanding the underlying dynamics and causal relations. We also study an oracle model that explicitly combines these components via symbolic representations."
                },
                {
                    "title": "Video Representation Learning by Dense Predictive Coding",
                    "abstract": "The objective of this paper is self-supervised learning of spatio-temporal embeddings from video, suitable for human action recognition. We make three contributions: First, we introduce the Dense Predictive Coding (DPC) framework for self-supervised representation learning on videos. This learns a dense encoding of spatio-temporal blocks by recurrently predicting future representations; Second, we propose a curriculum training scheme to predict further into the future with progressively less temporal context. This encourages the model to only encode slowly varying spatial-temporal signals, therefore leading to semantic representations; Third, we evaluate the approach by first training the DPC model on the Kinetics-400 dataset with self-supervised learning, and then finetuning the representation on a downstream task, i.e. action recognition. With single stream (RGB only), DPC pretrained representations achieve state-of-the-art self-supervised performance on both UCF101(75.7% top1 acc) and HMDB51(35.7% top1 acc), outperforming all previous learning methods by a significant margin, and approaching the performance of a baseline pre-trained on ImageNet."
                },
                {
                    "title": "Predictive Coding",
                    "abstract": "Technology has created a problem. The ease of use, the ubiquitous availability, the ease of retention versus the effort of management has caused emails (and other communications) to build into an overwhelming volume of data that is exponentially more expensive to deal with in litigation. Can technology solve the problem that technology created? Promoters of predictive coding claim that it can. That\u2019s a big promise, because parties involved in litigation tell us that this is a big problem."
                },
                {
                    "title": "A Perceptual Prediction Framework for Self Supervised Event Segmentation",
                    "abstract": "Temporal segmentation of long videos is an important problem, that has largely been tackled through supervised learning, often requiring large amounts of annotated training data. In this paper, we tackle the problem of self-supervised temporal segmentation that alleviates the need for any supervision in the form of labels (full supervision) or temporal ordering (weak supervision). We introduce a self-supervised, predictive learning framework that draws inspiration from cognitive psychology to segment long, visually complex videos into constituent events. Learning involves only a single pass through the training data. We also introduce a new adaptive learning paradigm that helps reduce the effect of catastrophic forgetting in recurrent neural networks. Extensive experiments on three publicly available datasets - Breakfast Actions, 50 Salads, and INRIA Instructional Videos datasets show the efficacy of the proposed approach. We show that the proposed approach outperforms weakly-supervised and unsupervised baselines by up to 24% and achieves competitive segmentation results compared to fully supervised baselines with only a single pass through the training data. Finally, we show that the proposed self-supervised learning paradigm learns highly discriminating features to improve action recognition."
                },
                {
                    "title": "Deep Predictive Coding Network with Local Recurrent Processing for Object Recognition",
                    "abstract": "Inspired by \"predictive coding\" - a theory in neuroscience, we develop a bi-directional and dynamic neural network with local recurrent processing, namely predictive coding network (PCN). Unlike feedforward-only convolutional neural networks, PCN includes both feedback connections, which carry top-down predictions, and feedforward connections, which carry bottom-up errors of prediction. Feedback and feedforward connections enable adjacent layers to interact locally and recurrently to refine representations towards minimization of layer-wise prediction errors. When unfolded over time, the recurrent processing gives rise to an increasingly deeper hierarchy of non-linear transformation, allowing a shallow network to dynamically extend itself into an arbitrarily deep network. We train and test PCN for image classification with SVHN, CIFAR and ImageNet datasets. Despite notably fewer layers and parameters, PCN achieves competitive performance compared to classical and state-of-the-art models. Further analysis shows that the internal representations in PCN converge over time and yield increasingly better accuracy in object recognition. Errors of top-down prediction also reveal visual saliency or bottom-up attention."
                },
                {
                    "title": "Gradient Episodic Memory for Continual Learning",
                    "abstract": "One major obstacle towards AI is the poor ability of models to solve new problems quicker, and without forgetting previously acquired knowledge. To better understand this issue, we study the problem of continual learning, where the model observes, once and one by one, examples concerning a sequence of tasks. First, we propose a set of metrics to evaluate models learning over a continuum of data. These metrics characterize models not only by their test accuracy, but also in terms of their ability to transfer knowledge across tasks. Second, we propose a model for continual learning, called Gradient Episodic Memory (GEM) that alleviates forgetting, while allowing beneficial transfer of knowledge to previous tasks. Our experiments on variants of the MNIST and CIFAR-100 datasets demonstrate the strong performance of GEM when compared to the state-of-the-art."
                },
                {
                    "title": "An Approximation of the Error Backpropagation Algorithm in a Predictive Coding Network with Local Hebbian Synaptic Plasticity",
                    "abstract": "Abstract To efficiently learn from feedback, cortical networks need to update synaptic weights on multiple levels of cortical hierarchy. An effective and well-known algorithm for computing such changes in synaptic weights is the error backpropagation algorithm. However, in this algorithm, the change in synaptic weights is a complex function of weights and activities of neurons not directly connected with the synapse being modified, whereas the changes in biological synapses are determined only by the activity of presynaptic and postsynaptic neurons. Several models have been proposed that approximate the backpropagation algorithm with local synaptic plasticity, but these models require complex external control over the network or relatively complex plasticity rules. Here we show that a network developed in the predictive coding framework can efficiently perform supervised learning fully autonomously, employing only simple local Hebbian plasticity. Furthermore, for certain parameters, the weight change in the predictive coding model converges to that of the backpropagation algorithm. This suggests that it is possible for cortical networks with simple Hebbian synaptic plasticity to implement efficient learning algorithms in which synapses in areas on multiple levels of hierarchy are modified to minimize the error on the output."
                },
                {
                    "title": "Dense Associative Memory for Pattern Recognition",
                    "abstract": "A model of associative memory is studied, which stores and reliably retrieves many more patterns than the number of neurons in the network. We propose a simple duality between this dense associative memory and neural networks commonly used in deep learning. On the associative memory side of this duality, a family of models that smoothly interpolates between two limiting cases can be constructed. One limit is referred to as the feature-matching mode of pattern recognition, and the other one as the prototype regime. On the deep learning side of the duality, this family corresponds to feedforward neural networks with one hidden layer and various activation functions, which transmit the activities of the visible neurons to the hidden layer. This family of activation functions includes logistics, rectified linear units, and rectified polynomials of higher degrees. The proposed duality makes it possible to apply energy-based intuition from associative memory to analyze computational properties of neural networks with unusual activation functions - the higher rectified polynomials which until now have not been used in deep learning. The utility of the dense memories is illustrated for two test cases: the logical gate XOR and the recognition of handwritten digits from the MNIST data set."
                },
                {
                    "title": "Deep Predictive Coding Networks for Video Prediction and Unsupervised Learning",
                    "abstract": "While great strides have been made in using deep learning algorithms to solve supervised learning tasks, the problem of unsupervised learning - leveraging unlabeled examples to learn about the structure of a domain - remains a difficult unsolved challenge. Here, we explore prediction of future frames in a video sequence as an unsupervised learning rule for learning about the structure of the visual world. We describe a predictive neural network (\"PredNet\") architecture that is inspired by the concept of \"predictive coding\" from the neuroscience literature. These networks learn to predict future frames in a video sequence, with each layer in the network making local predictions and only forwarding deviations from those predictions to subsequent network layers. We show that these networks are able to robustly learn to predict the movement of synthetic (rendered) objects, and that in doing so, the networks learn internal representations that are useful for decoding latent object parameters (e.g. pose) that support object recognition with fewer training views. We also show that these networks can scale to complex natural image streams (car-mounted camera videos), capturing key aspects of both egocentric movement and the movement of objects in the visual scene, and the representation learned in this setting is useful for estimating the steering angle. Altogether, these results suggest that prediction represents a powerful framework for unsupervised learning, allowing for implicit learning of object and scene structure."
                },
                {
                    "title": "How do neurons operate on sparse distributed representations? A mathematical theory of sparsity, neurons and active dendrites",
                    "abstract": "We propose a formal mathematical model for sparse representations and active dendrites in neocortex. Our model is inspired by recent experimental findings on active dendritic processing and NMDA spikes in pyramidal neurons. These experimental and modeling studies suggest that the basic unit of pattern memory in the neocortex is instantiated by small clusters of synapses operated on by localized non-linear dendritic processes. We derive a number of scaling laws that characterize the accuracy of such dendrites in detecting activation patterns in a neuronal population under adverse conditions. We introduce the union property which shows that synapses for multiple patterns can be randomly mixed together within a segment and still lead to highly accurate recognition. We describe simulation results that provide further insight into sparse representations as well as two primary results. First we show that pattern recognition by a neuron with active dendrites can be extremely accurate and robust with high dimensional sparse inputs even when using a tiny number of synapses to recognize large patterns. Second, equations representing recognition accuracy of a dendrite predict optimal NMDA spiking thresholds under a generous set of assumptions. The prediction tightly matches NMDA spiking thresholds measured in the literature. Our model matches many of the known properties of pyramidal neurons. As such the theory provides a mathematical framework for understanding the benefits and limits of sparse representations in cortical networks."
                },
                {
                    "title": "Why Neurons Have Thousands of Synapses, a Theory of Sequence Memory in Neocortex",
                    "abstract": "Pyramidal neurons represent the majority of excitatory neurons in the neocortex. Each pyramidal neuron receives input from thousands of excitatory synapses that are segregated onto dendritic branches. The dendrites themselves are segregated into apical, basal, and proximal integration zones, which have different properties. It is a mystery how pyramidal neurons integrate the input from thousands of synapses, what role the different dendrites play in this integration, and what kind of network behavior this enables in cortical tissue. It has been previously proposed that non-linear properties of dendrites enable cortical neurons to recognize multiple independent patterns. In this paper we extend this idea in multiple ways. First we show that a neuron with several thousand synapses segregated on active dendrites can recognize hundreds of independent patterns of cellular activity even in the presence of large amounts of noise and pattern variation. We then propose a neuron model where patterns detected on proximal dendrites lead to action potentials, defining the classic receptive field of the neuron, and patterns detected on basal and apical dendrites act as predictions by slightly depolarizing the neuron without generating an action potential. By this mechanism, a neuron can predict its activation in hundreds of independent contexts. We then present a network model based on neurons with these properties that learns time-based sequences. The network relies on fast local inhibition to preferentially activate neurons that are slightly depolarized. Through simulation we show that the network scales well and operates robustly over a wide range of parameters as long as the network uses a sparse distributed code of cellular activations. We contrast the properties of the new network model with several other neural network models to illustrate the relative capabilities of each. We conclude that pyramidal neurons with thousands of synapses, active dendrites, and multiple integration zones create a robust and powerful sequence memory. Given the prevalence and similarity of excitatory neurons throughout the neocortex and the importance of sequence memory in inference and behavior, we propose that this form of sequence memory may be a universal property of neocortical tissue."
                },
                {
                    "title": "Sequential memory: Binding dynamics.",
                    "abstract": "Temporal order memories are critical for everyday animal and human functioning. Experiments and our own experience show that the binding or association of various features of an event together and the maintaining of multimodality events in sequential order are the key components of any sequential memories-episodic, semantic, working, etc. We study a robustness of binding sequential dynamics based on our previously introduced model in the form of generalized Lotka-Volterra equations. In the phase space of the model, there exists a multi-dimensional binding heteroclinic network consisting of saddle equilibrium points and heteroclinic trajectories joining them. We prove here the robustness of the binding sequential dynamics, i.e., the feasibility phenomenon for coupled heteroclinic networks: for each collection of successive heteroclinic trajectories inside the unified networks, there is an open set of initial points such that the trajectory going through each of them follows the prescribed collection staying in a small neighborhood of it. We show also that the symbolic complexity function of the system restricted to this neighborhood is a polynomial of degree L - 1, where L is the number of modalities."
                },
                {
                    "title": "Properties of Sparse Distributed Representations and their Application to Hierarchical Temporal Memory",
                    "abstract": "Empirical evidence demonstrates that every region of the neocortex represents information using sparse activity patterns. This paper examines Sparse Distributed Representations (SDRs), the primary information representation strategy in Hierarchical Temporal Memory (HTM) systems and the neocortex. We derive a number of properties that are core to scaling, robustness, and generalization. We use the theory to provide practical guidelines and illustrate the power of SDRs as the basis of HTM. Our goal is to help create a unified mathematical and practical framework for SDRs as it relates to cortical function."
                },
                {
                    "title": "Unsupervised Learning of Video Representations using LSTMs",
                    "abstract": "We use Long Short Term Memory (LSTM) networks to learn representations of video sequences. Our model uses an encoder LSTM to map an input sequence into a fixed length representation. This representation is decoded using single or multiple decoder LSTMs to perform different tasks, such as reconstructing the input sequence, or predicting the future sequence. We experiment with two kinds of input sequences - patches of image pixels and high-level representations (\"percepts\") of video frames extracted using a pretrained convolutional net. We explore different design choices such as whether the decoder LSTMs should condition on the generated output. We analyze the outputs of the model qualitatively to see how well the model can extrapolate the learned video representation into the future and into the past. We further evaluate the representations by finetuning them for a supervised learning problem - human action recognition on the UCF-101 and HMDB-51 datasets. We show that the representations help improve classification accuracy, especially when there are only few training examples. Even models pretrained on unrelated datasets (300 hours of YouTube videos) can help action recognition performance."
                },
                {
                    "title": "Probabilistic Graphical Models: Principles and Techniques - Adaptive Computation and Machine Learning",
                    "abstract": "Most tasks require a person or an automated system to reasonto reach conclusions based on available information. The framework of probabilistic graphical models, presented in this book, provides a general approach for this task. The approach is model-based, allowing interpretable models to be constructed and then manipulated by reasoning algorithms. These models can also be learned automatically from data, allowing the approach to be used in cases where manually constructing a model is difficult or even impossible. Because uncertainty is an inescapable aspect of most real-world applications, the book focuses on probabilistic models, which make the uncertainty explicit and provide models that are more faithful to reality. Probabilistic Graphical Models discusses a variety of models, spanning Bayesian networks, undirected Markov networks, discrete and continuous models, and extensions to deal with dynamical systems and relational data. For each class of models, the text describes the three fundamental cornerstones: representation, inference, and learning, presenting both basic concepts and advanced techniques. Finally, the book considers the use of the proposed framework for causal reasoning and decision making under uncertainty. The main text in each chapter provides the detailed technical development of the key ideas. Most chapters also include boxes with additional material: skill boxes, which describe techniques; case study boxes, which discuss empirical cases related to the approach described in the text, including applications in computer vision, robotics, natural language understanding, and computational biology; and concept boxes, which present significant concepts drawn from the material in the chapter. Instructors (and readers) can group chapters in various combinations, from core topics to more technically advanced material, to suit their particular needs. Adaptive Computation and Machine Learning series"
                },
                {
                    "title": "Bidirectional associative memories",
                    "abstract": "Stability and encoding properties of two-layer nonlinear feedback neural networks are examined. Bidirectionality is introduced in neural nets to produce two-way associative search for stored associations. The bidirectional associative memory (BAM) is the minimal two-layer nonlinear feedback network. The author proves that every n-by-p matrix M is a bidirectionally stable heteroassociative content-addressable memory for both binary/bipolar and continuous neurons. When the BAM neutrons are activated, the network quickly evolves to a stable state of two-pattern reverberation, or resonance. The stable reverberation corresponds to a system energy local minimum. Heteroassociative information is encoded in a BAM by summing correlation matrices. The BAM storage capacity for reliable recall is roughly m<min (n,p). It is also shown that it is better on average to use bipolar (-1,1) coding than binary"
                },
                {
                    "title": "Temporal association in asymmetric neural networks.",
                    "abstract": "A neural network model which is capable of recalling time sequences and cycles of patterns is introduced. In this model, some of the synaptic connections, ${J}_{\\mathrm{ij}}$, between pairs of neurons are asymmetric (${J}_{\\mathrm{ij}}\\ensuremath{\\ne}{J}_{\\mathrm{ji}}$) and have slow dynamic response. The effects of thermal noise on the generated sequences are discussed. Simulation results demonstrating the performance of the network are presented. The model may be also useful in understanding the generation of rhythmic patterns in biological motor systems."
                },
                {
                    "title": "Predictive coding: a fresh view of inhibition in the retina",
                    "abstract": "Interneurons exhibiting centre-surround antagonism within their receptive fields are commonly found in peripheral visual pathways. We propose that this organization enables the visual system to encode spatial detail in a manner that minimizes the deleterious effects of intrinsic noise, by exploiting the spatial correlation that exists within natural scenes. The antagonistic surround takes a weighted mean of the signals in neighbouring receptors to generate a statistical prediction of the signal at the centre. The predicted value is subtracted from the actual centre signal, thus minimizing the range of outputs transmitted by the centre. In this way the entire dynamic range of the interneuron can be devoted to encoding a small range of intensities, thus rendering fine detail detectable against intrinsic noise injected at later stages in processing. This predictive encoding scheme also reduces spatial redundancy, thereby enabling the array of interneurons to transmit a larger number of distinguishable images, taking into account the expected structure of the visual world. The profile of the required inhibitory field is derived from statistical estimation theory. This profile depends strongly upon the signal: noise ratio and weakly upon the extent of lateral spatial correlation. The receptive fields that are quantitatively predicted by the theory resemble those of X-type retinal ganglion cells and show that the inhibitory surround should become weaker and more diffuse at low intensities. The latter property is unequivocally demonstrated in the first-order interneurons of the fly\u2019s compound eye. The theory is extended to the time domain to account for the phasic responses of fly interneurons. These comparisons suggest that, in the early stages of processing, the visual system is concerned primarily with coding the visual image to protect against subsequent intrinsic noise, rather than with reconstructing the scene or extracting specific features from it. The treatment emphasizes that a neuron\u2019s dynamic range should be matched to both its receptive field and the statistical properties of the visual pattern expected within this field. Finally, the analysis is synthetic because it is an extension of the background suppression hypothesis (Barlow & Levick 1976), satisfies the redundancy reduction hypothesis (Barlow 1961 a, b) and is equivalent to deblurring under certain conditions (Ratliff 1965)."
                },
                {
                    "title": "Neural networks and physical systems with emergent collective computational abilities.",
                    "abstract": "Computational properties of use of biological organisms or to the construction of computers can emerge as collective properties of systems having a large number of simple equivalent components (or neurons). The physical meaning of content-addressable memory is described by an appropriate phase space flow of the state of a system. A model of such a system is given, based on aspects of neurobiology but readily adapted to integrated circuits. The collective properties of this model produce a content-addressable memory which correctly yields an entire memory from any subpart of sufficient size. The algorithm for the time evolution of the state of the system is based on asynchronous parallel processing. Additional emergent collective properties include some capacity for generalization, familiarity recognition, categorization, error correction, and time sequence retention. The collective properties are only weakly sensitive to details of the modeling or the failure of individual devices."
                },
                {
                    "title": "Entropy coding in speech and television differential PCM systems (Corresp.)",
                    "abstract": "Much of the redundancy in a speech or television signal is eliminated when it is encoded into digital form by a differential pulse-code-modulation (DPCM) encoder. Additional coding of the DPCM output using entropy coding techniques (Huffman or Shannon-Fano coding) can result in a further increase in the signal-to-quantizing-noise ratio of 5.6 dB without increasing the transmission rate."
                },
                {
                    "title": "STREAMER: Streaming Representation Learning and Event Segmentation in a Hierarchical Manner",
                    "abstract": "We present a novel self-supervised approach for hierarchical representation learning and segmentation of perceptual inputs in a streaming fashion. Our research addresses how to semantically group streaming inputs into chunks at various levels of a hierarchy while simultaneously learning, for each chunk, robust global representations throughout the domain. To achieve this, we propose STREAMER, an architecture that is trained layer-by-layer, adapting to the complexity of the input domain. In our approach, each layer is trained with two primary objectives: making accurate predictions into the future and providing necessary information to other levels for achieving the same objective. The event hierarchy is constructed by detecting prediction error peaks at different levels, where a detected boundary triggers a bottom-up information flow. At an event boundary, the encoded representation of inputs at one layer becomes the input to a higher-level layer. Additionally, we design a communication module that facilitates top-down and bottom-up exchange of information during the prediction process. Notably, our model is fully self-supervised and trained in a streaming manner, enabling a single pass on the training data. This means that the model encounters each input only once and does not store the data. We evaluate the performance of our model on the egocentric EPIC-KITCHENS dataset, specifically focusing on temporal event segmentation. Furthermore, we conduct event retrieval experiments using the learned representations to demonstrate the high quality of our video event representations. Illustration videos and code are available on our project page: https://ramymounir.com/publications/streamer ."
                },
                {
                    "title": "A Path Towards Autonomous Machine Intelligence Version 0.9.2, 2022-06-27",
                    "abstract": "How could machines learn as efficiently as humans and animals? How could machines learn to reason and plan? How could machines learn representations of percepts and action plans at multiple levels of abstraction, enabling them to reason, predict, and plan at multiple time horizons? This position paper proposes an architecture and training paradigms with which to construct autonomous intelligent agents. It combines concepts such as configurable predictive world model, behavior driven through intrinsic motivation, and hierarchical joint embedding architectures trained with self-supervised learning."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "An introduction to hidden Markov models",
                    "abstract": "The basic theory of Markov chains has been known to mathematicians and engineers for close to 80 years, but it is only in the past decade that it has been applied explicitly to problems in speech processing. One of the major reasons why speech models, based on Markov chains, have not been developed until recently was the lack of a method for optimizing the parameters of the Markov model to match observed signal patterns. Such a method was proposed in the late 1960's and was immediately applied to speech processing in several research institutions. Continued refinements in the theory and implementation of Markov modelling techniques have greatly enhanced the method, leading to a wide range of applications of these models. It is the purpose of this tutorial paper to give an introduction to the theory of Markov models, and to illustrate how they have been applied to problems in speech recognition."
                }
            ],
            "categories": [
                "cs.AI",
                "cs.CV",
                "cs.LG",
                "q-bio.NC"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can biologically-plausible artificial networks effectively learn and recall temporally-dependent patterns in sequences while overcoming challenges such as catastrophic forgetting and ambiguous context representations?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the understanding of cognitive functions in both biological and artificial systems. It has broader implications for the research community by providing insights into how memory and learning can be modeled in a way that mimics biological processes. This could lead to the development of more robust machine learning models that can perform complex tasks such as language translation, action planning, and real-time decision-making. Addressing this question could also pave the way for practical applications in robotics, natural language processing, and adaptive learning systems, ultimately enhancing the capabilities of AI technologies.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to model sequential memory under biological constraints, which introduces complexities such as catastrophic forgetting, where new learning can overwrite previous knowledge. Naive approaches may fail because they do not account for the need to uniquely encode inputs based on their context within a sequence, nor do they effectively handle multiple valid future possibilities. Additionally, the requirement for online learning\u2014where the model must learn transitions incrementally without revisiting older sequences\u2014adds another layer of difficulty. The model must also be robust to significant input noise, necessitating sophisticated mechanisms for noise reduction and context representation.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on static or unrelated event memory, leaving a gap in understanding how to model temporally-dependent patterns in a biologically plausible manner. Existing solutions often lack the ability to prevent catastrophic forgetting or to uniquely encode context-dependent inputs. Barriers such as the complexity of implementing competitive learning within a biologically-inspired framework and the challenge of developing algorithms that can learn incrementally without revisiting past data have hindered progress. Our approach differs by introducing Predictive Attractor Models (PAM), which integrate a state prediction model with a generative attractor model, addressing these limitations through a novel learning algorithm inspired by Hierarchical Temporal Memory.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of Predictive Attractor Models (PAM), which consist of a state prediction model and a generative attractor model. The model will be evaluated using a dataset of sequential patterns"
            }
        },
        "author_data": {
            "eb17760c-5e8f-47ec-94bd-3d483b92b4cf": {
                "pk": "eb17760c-5e8f-47ec-94bd-3d483b92b4cf",
                "name": "Ramy Mounir",
                "collaborators": [
                    "Kyle Reed",
                    "Redwan Alqasemi",
                    "Rajiv Dubey",
                    "Roman Gula",
                    "J\u00f6rn Theuerkauf",
                    "Sudeep Sarkar"
                ],
                "domain": [
                    "Haptic Feedback",
                    "Brain-Computer Interface",
                    "Event Segmentation",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Polyrhythmic Bimanual Coordination Training using Haptic Force Feedback",
                        "abstract": "It is challenging to develop two thoughts at the same time or perform two uncorrelated motions simultaneously. This work looks specifically towards training humans to perform a 2:3 polyrhythmic bimanual ratio using haptic force feedback devices (SensAble Phantom OMNI). We implemented an interactive training session to help participants learn to decouple their hand motions quickly. Three subjects (2 Females, 1 Male) were tested and have successfully increased their scores after adaptive training durations of under five minutes."
                    },
                    {
                        "title": "BCI-Controlled Hands-Free Wheelchair Navigation with Obstacle Avoidance",
                        "abstract": "Brain-Computer interfaces (BCI) are widely used in reading brain signals and converting them into real-world motion. However, the signals produced from the BCI are noisy and hard to analyze. This paper looks specifically towards combining the BCI's latest technology with ultrasonic sensors to provide a hands-free wheelchair that can efficiently navigate through crowded environments. This combination provides safety and obstacle avoidance features necessary for the BCI Navigation system to gain more confidence and operate the wheelchair at a relatively higher velocity. A population of six human subjects tested the BCI-controller and obstacle avoidance features. Subjects were able to mentally control the destination of the wheelchair, by moving the target from the starting position to a predefined position, in an average of 287.12 seconds and a standard deviation of 48.63 seconds after 10 minutes of training. The wheelchair successfully avoided all obstacles placed by the subjects during the test."
                    },
                    {
                        "title": "Spatio-Temporal Event Segmentation and Localization for Wildlife Extended Videos",
                        "abstract": "Using offline training schemes, researchers have tackled the event segmentation problem by providing full or weak-supervision through manually annotated labels or self-supervised epoch-based training. Most works consider videos that are at most 10's of minutes long. We present a self-supervised perceptual prediction framework capable of temporal event segmentation by building stable representations of objects over time and demonstrate it on long videos, spanning several days. The approach is deceptively simple but quite effective. We rely on predictions of high-level features computed by a standard deep learning backbone. For prediction, we use an LSTM, augmented with an attention mechanism, trained in a self-supervised manner using the prediction error. The self-learned attention maps effectively localize and track the event-related objects in each frame. The proposed approach does not require labels. It requires only a single pass through the video, with no separate training set. Given the lack of datasets of very long videos, we demonstrate our method on video from 10 days (254 hours) of continuous wildlife monitoring data that we had collected with required permissions. We find that the approach is robust to various environmental conditions such as day/night conditions, rain, sharp shadows, and windy conditions. For the task of temporally locating events, we had an 80% recall rate at 20% false-positive rate for frame-level segmentation. At the activity level, we had an 80% activity recall rate for one false activity detection every 50 minutes. We will make the dataset, which is the first of its kind, and the code available to the research community."
                    }
                ]
            },
            "63bb9f9e-3014-4edf-9a79-4ede3e83f2ab": {
                "pk": "63bb9f9e-3014-4edf-9a79-4ede3e83f2ab",
                "name": "Sudeep Sarkar",
                "collaborators": [
                    "Sathyanarayanan N. Aakur",
                    "Mauricio Pamplona Segundo",
                    "Rodrigo Minetto",
                    "Gilbert Rotich",
                    "Mona Fathollahi Ghezelghieh",
                    "Rangachar Kasturi",
                    "Xiaoyang Guo",
                    "Anuj Srivastava",
                    "Cole Hill",
                    "Ramy Mounir"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Action Recognition",
                    "Satellite Imagery"
                ],
                "publications": [
                    {
                        "title": "Learning camera viewpoint using CNN to improve 3D body pose estimation",
                        "abstract": "The objective of this work is to estimate 3D human pose from a single RGB image. Extracting image representations which incorporate both spatial relation of body parts and their relative depth plays an essential role in accurate3D pose reconstruction. In this paper, for the first time, we show that camera viewpoint in combination to 2D joint lo-cations significantly improves 3D pose accuracy without the explicit use of perspective geometry mathematical models.To this end, we train a deep Convolutional Neural Net-work (CNN) to learn categorical camera viewpoint. To make the network robust against clothing and body shape of the subject in the image, we utilized 3D computer rendering to synthesize additional training images. We test our framework on the largest 3D pose estimation bench-mark, Human3.6m, and achieve up to 20% error reduction compared to the state-of-the-art approaches that do not use body part segmentation."
                    },
                    {
                        "title": "Resource-Constrained Simultaneous Detection and Labeling of Objects in High-Resolution Satellite Images",
                        "abstract": "We describe a strategy for detection and classification of man-made objects in large high-resolution satellite photos under computational resource constraints. We detect and classify candidate objects by using five pipelines of convolutional neural network processing (CNN), run in parallel. Each pipeline has its own unique strategy for fine tunning parameters, proposal region filtering, and dealing with image scales. The conflicting region proposals are merged based on region confidence and not just based on overlap areas, which improves the quality of the final bounding-box regions selected. We demonstrate this strategy using the recent xView challenge, which is a complex benchmark with more than 1,100 high-resolution images, spanning 800,000 aerial objects around the world covering a total area of 1,400 square kilometers at 0.3 meter ground sample distance. To tackle the resource-constrained problem posed by the xView challenge, where inferences are restricted to be on CPU with 8GB memory limit, we used lightweight CNN's trained with the single shot detector algorithm. Our approach was competitive on sequestered sets; it was ranked third."
                    },
                    {
                        "title": "A Perceptual Prediction Framework for Self Supervised Event Segmentation",
                        "abstract": "Temporal segmentation of long videos is an important problem, that has largely been tackled through supervised learning, often requiring large amounts of annotated training data. In this paper, we tackle the problem of self-supervised temporal segmentation of long videos that alleviate the need for any supervision. We introduce a self-supervised, predictive learning framework that draws inspiration from cognitive psychology to segment long, visually complex videos into individual, stable segments that share the same semantics. We also introduce a new adaptive learning paradigm that helps reduce the effect of catastrophic forgetting in recurrent neural networks. Extensive experiments on three publicly available datasets - Breakfast Actions, 50 Salads, and INRIA Instructional Videos datasets show the efficacy of the proposed approach. We show that the proposed approach is able to outperform weakly-supervised and other unsupervised learning approaches by up to 24% and have competitive performance compared to fully supervised approaches. We also show that the proposed approach is able to learn highly discriminative features that help improve action recognition when used in a representation learning paradigm."
                    },
                    {
                        "title": "Hydra: an Ensemble of Convolutional Neural Networks for Geospatial Land Classification",
                        "abstract": "We describe in this paper Hydra, an ensemble of convolutional neural networks (CNN) for geospatial land classification. The idea behind Hydra is to create an initial CNN that is coarsely optimized but provides a good starting pointing for further optimization, which will serve as the Hydra's body. Then, the obtained weights are fine-tuned multiple times with different augmentation techniques, crop styles, and classes weights to form an ensemble of CNNs that represent the Hydra's heads. By doing so, we prompt convergence to different endpoints, which is a desirable aspect for ensembles. With this framework, we were able to reduce the training time while maintaining the classification performance of the ensemble. We created ensembles for our experiments using two state-of-the-art CNN architectures, ResNet and DenseNet. We have demonstrated the application of our Hydra framework in two datasets, FMOW and NWPU-RESISC45, achieving results comparable to the state-of-the-art for the former and the best reported performance so far for the latter. Code and CNN models are available at https://github.com/maups/hydra-fmow"
                    },
                    {
                        "title": "A Quotient Space Formulation for Generative Statistical Analysis of Graphical Data",
                        "abstract": "Complex analyses involving multiple, dependent random quantities often lead to graphical models - a set of nodes denoting variables of interest, and corresponding edges denoting statistical interactions between nodes. To develop statistical analyses for graphical data, especially towards generative modeling, one needs mathematical representations and metrics for matching and comparing graphs, and subsequent tools, such as geodesics, means, and covariances. This paper utilizes a quotient structure to develop efficient algorithms for computing these quantities, leading to useful statistical tools, including principal component analysis, statistical testing, and modeling. We demonstrate the efficacy of this framework using datasets taken from several problem areas, including letters, biochemical structures, and social networks."
                    },
                    {
                        "title": "Actor-centered Representations for Action Localization in Streaming Videos",
                        "abstract": "Event perception tasks such as recognizing and localizing actions in streaming videos are essential for scaling to real-world application contexts. We tackle the problem of learning actor-centered representations through the notion of continual hierarchical predictive learning to localize actions in streaming videos without the need for training labels and outlines for the objects in the video. We propose a framework driven by the notion of hierarchical predictive learning to construct actor-centered features by attention-based contextualization. The key idea is that predictable features or objects do not attract attention and hence do not contribute to the action of interest. Experiments on three benchmark datasets show that the approach can learn robust representations for localizing actions using only one epoch of training, i.e., a single pass through the streaming video. We show that the proposed approach outperforms unsupervised and weakly supervised baselines while offering competitive performance to fully supervised approaches. Additionally, we extend the model to multi-actor settings to recognize group activities while localizing the multiple, plausible actors. We also show that it generalizes to out-of-domain data with limited performance degradation."
                    },
                    {
                        "title": "Reducing Training Demands for 3D Gait Recognition with Deep Koopman Operator Constraints",
                        "abstract": "Deep learning research has made many biometric recognition solution viable, but it requires vast training data to achieve real-world generalization. Unlike other biometric traits, such as face and ear, gait samples cannot be easily crawled from the web to form massive unconstrained datasets. As the human body has been extensively studied for different digital applications, one can rely on prior shape knowledge to overcome data scarcity. This work follows the recent trend of fitting a 3D deformable body model into gait videos using deep neural networks to obtain disentangled shape and pose representations for each frame. To enforce temporal consistency in the network, we introduce a new Linear Dynamical Systems (LDS) module and loss based on Koopman operator theory, which provides an unsupervised motion regularization for the periodic nature of gait, as well as a predictive capacity for extending gait sequences. We compare LDS to the traditional adversarial training approach and use the USF HumanID and CASIA-B datasets to show that LDS can obtain better accuracy with less training data. Finally, we also show that our 3D modeling approach is much better than other 3D gait approaches in overcoming viewpoint variation under normal, bag-carrying and clothing change conditions."
                    },
                    {
                        "title": "Action Localization through Continual Predictive Learning",
                        "abstract": "The problem of action recognition involves locating the action in the video, both over time and spatially in the image. The dominant current approaches use supervised learning to solve this problem, and require large amounts of annotated training data, in the form of frame-level bounding box annotations around the region of interest. In this paper, we present a new approach based on continual learning that uses feature-level predictions for self-supervision. It does not require any training annotations in terms of frame-level bounding boxes. The approach is inspired by cognitive models of visual event perception that propose a prediction-based approach to event understanding. We use a stack of LSTMs coupled with CNN encoder, along with novel attention mechanisms, to model the events in the video and use this model to predict high-level features for the future frames. The prediction errors are used to continuously learn the parameters of the models. This self-supervised framework is not complicated as other approaches but is very effective in learning robust visual representations for both labeling and localization. It should be noted that the approach outputs in a streaming fashion, requiring only a single pass through the video, making it amenable for real-time processing. We demonstrate this on three datasets - UCF Sports, JHMDB, and THUMOS'13 and show that the proposed approach outperforms weakly-supervised and unsupervised baselines and obtains competitive performance compared to fully supervised baselines. Finally, we show that the proposed framework can generalize to egocentric videos and obtain state-of-the-art results in unsupervised gaze prediction."
                    },
                    {
                        "title": "Abductive Reasoning as Self-Supervision for Common Sense Question Answering",
                        "abstract": "Question answering has seen significant advances in recent times, especially with the introduction of increasingly bigger transformer-based models pre-trained on massive amounts of data. While achieving impressive results on many benchmarks, their performances appear to be proportional to the amount of training data available in the target domain. In this work, we explore the ability of current question-answering models to generalize - to both other domains as well as with restricted training data. We find that large amounts of training data are necessary, both for pre-training as well as fine-tuning to a task, for the models to perform well on the designated task. We introduce a novel abductive reasoning approach based on Grenander's Pattern Theory framework to provide self-supervised domain adaptation cues or \"pseudo-labels,\" which can be used instead of expensive human annotations. The proposed self-supervised training regimen allows for effective domain adaptation without losing performance compared to fully supervised baselines. Extensive experiments on two publicly available benchmarks show the efficacy of the proposed approach. We show that neural networks models trained using self-labeled data can retain up to $75\\%$ of the performance of models trained on large amounts of human-annotated training data."
                    },
                    {
                        "title": "Spatio-Temporal Event Segmentation and Localization for Wildlife Extended Videos",
                        "abstract": "Using offline training schemes, researchers have tackled the event segmentation problem by providing full or weak-supervision through manually annotated labels or self-supervised epoch-based training. Most works consider videos that are at most 10's of minutes long. We present a self-supervised perceptual prediction framework capable of temporal event segmentation by building stable representations of objects over time and demonstrate it on long videos, spanning several days. The approach is deceptively simple but quite effective. We rely on predictions of high-level features computed by a standard deep learning backbone. For prediction, we use an LSTM, augmented with an attention mechanism, trained in a self-supervised manner using the prediction error. The self-learned attention maps effectively localize and track the event-related objects in each frame. The proposed approach does not require labels. It requires only a single pass through the video, with no separate training set. Given the lack of datasets of very long videos, we demonstrate our method on video from 10 days (254 hours) of continuous wildlife monitoring data that we had collected with required permissions. We find that the approach is robust to various environmental conditions such as day/night conditions, rain, sharp shadows, and windy conditions. For the task of temporally locating events, we had an 80% recall rate at 20% false-positive rate for frame-level segmentation. At the activity level, we had an 80% activity recall rate for one false activity detection every 50 minutes. We will make the dataset, which is the first of its kind, and the code available to the research community."
                    },
                    {
                        "title": "Employing Fusion of Learned and Handcrafted Features for Unconstrained Ear Recognition",
                        "abstract": "We present an unconstrained ear recognition framework that outperforms state-of-the-art systems in different publicly available image databases. To this end, we developed CNN-based solutions for ear normalization and description, we used well-known handcrafted descriptors, and we fused learned and handcrafted features to improve recognition. We designed a two-stage landmark detector that successfully worked under untrained scenarios. We used the results generated to perform a geometric image normalization that boosted the performance of all evaluated descriptors. Our CNN descriptor outperformed other CNN-based works in the literature, specially in more difficult scenarios. The fusion of learned and handcrafted matchers appears to be complementary as it achieved the best performance in all experiments. The obtained results outperformed all other reported results for the UERC challenge, which contains the most difficult database nowadays."
                    },
                    {
                        "title": "LocaliseBot: Multi-view 3D object localisation with differentiable rendering for robot grasping",
                        "abstract": "Robot grasp typically follows five stages: object detection, object localisation, object pose estimation, grasp pose estimation, and grasp planning. We focus on object pose estimation. Our approach relies on three pieces of information: multiple views of the object, the camera's extrinsic parameters at those viewpoints, and 3D CAD models of objects. The first step involves a standard deep learning backbone (FCN ResNet) to estimate the object label, semantic segmentation, and a coarse estimate of the object pose with respect to the camera. Our novelty is using a refinement module that starts from the coarse pose estimate and refines it by optimisation through differentiable rendering. This is a purely vision-based approach that avoids the need for other information such as point cloud or depth images. We evaluate our object pose estimation approach on the ShapeNet dataset and show improvements over the state of the art. We also show that the estimated object pose results in 99.65% grasp accuracy with the ground truth grasp candidates on the Object Clutter Indoor Dataset (OCID) Grasp dataset, as computed using standard practice."
                    },
                    {
                        "title": "Measuring Human and Economic Activity from Satellite Imagery to Support City-Scale Decision-Making during COVID-19 Pandemic",
                        "abstract": "The COVID-19 outbreak forced governments worldwide to impose lockdowns and quarantines to prevent virus transmission. As a consequence, there are disruptions in human and economic activities all over the globe. The recovery process is also expected to be rough. Economic activities impact social behaviors, which leave signatures in satellite images that can be automatically detected and classified. Satellite imagery can support the decision-making of analysts and policymakers by providing a different kind of visibility into the unfolding economic changes. In this work, we use a deep learning approach that combines strategic location sampling and an ensemble of lightweight convolutional neural networks (CNNs) to recognize specific elements in satellite images that could be used to compute economic indicators based on it, automatically. This CNN ensemble framework ranked third place in the US Department of Defense xView challenge, the most advanced benchmark for object detection in satellite images. We show the potential of our framework for temporal analysis using the US IARPA Function Map of the World (fMoW) dataset. We also show results on real examples of different sites before and after the COVID-19 outbreak to illustrate different measurable indicators. Our code and annotated high-resolution aerial scenes before and after the outbreak are available on GitHub (https://github.com/maups/covid19-satellite-analysis)."
                    }
                ]
            }
        }
    },
    "2401.14469": {
        "paper_data": {
            "title": "Unveiling the Unseen: Identifiable Clusters in Trained Depthwise Convolutional Kernels",
            "url": "http://arxiv.org/abs/2401.14469v1",
            "arxiv_id": "2401.14469",
            "authors": [
                "Zahra Babaiee",
                "Peyman M. Kiasari",
                "Daniela Rus",
                "Radu Grosu"
            ],
            "abstract": "Recent advances in depthwise-separable convolutional neural networks (DS-CNNs) have led to novel architectures, that surpass the performance of classical CNNs, by a considerable scalability and accuracy margin. This paper reveals another striking property of DS-CNN architectures: discernible and explainable patterns emerge in their trained depthwise convolutional kernels in all layers. Through an extensive analysis of millions of trained filters, with different sizes and from various models, we employed unsupervised clustering with autoencoders, to categorize these filters. Astonishingly, the patterns converged into a few main clusters, each resembling the difference of Gaussian (DoG) functions, and their first and second-order derivatives. Notably, we were able to classify over 95\\% and 90\\% of the filters from state-of-the-art ConvNextV2 and ConvNeXt models, respectively. This finding is not merely a technological curiosity; it echoes the foundational models neuroscientists have long proposed for the vision systems of mammals. Our results thus deepen our understanding of the emergent properties of trained DS-CNNs and provide a bridge between artificial and biological visual processing systems. More broadly, they pave the way for more interpretable and biologically-inspired neural network designs in the future.",
            "introduction": "   1 Introduction  Convolutional neural networks (CNNs) have been extensively studied in order to understand how they learn ever since their inception. As early as the seminal papers on CNNs, researchers discovered that CNNs learn filters in their initial layers that detect edges or specific edge colors when applied to natural images\u00a0(Krizhevsky et\u00a0al., 2012). These learned features bear a strong resemblance to Gabor filters, which are related to responses in the primary visual cortex, and thus present some of the most striking links between neuroscience and machine learning\u00a0(Goodfellow et\u00a0al., 2016). However, interpretability significantly decreases as one examines deeper layers. Consequently, subsequent work has mainly focused on inspecting the features learned by convolutional layers, rather than the weights themselves, to elucidate how CNNs operate\u00a0(Zeiler & Fergus, 2014; Yosinski et\u00a0al., 2015; Olah et\u00a0al., 2017; Bau et\u00a0al., 2017). While studying the learned features by convolutional layers is intuitive, comprehending the filter weights of deep layers of CNNs remains an open question.   In recent years, depthwise-separable convolutional neural networks (DS-CNNs) have become widely adopted in computer vision. The significantly lower computational costs of DS-CNNs enabled MobileNets\u00a0(Howard et\u00a0al., 2017) to achieve state-of-the-art accuracy, with substantially fewer parameters and multiplication-addition operations, than traditional CNNs. Following this breakthrough, DS-CNNs have become the de facto standard in most modern CNN architectures, as they offer a favorable option for scaling up networks\u00a0(Liu et\u00a0al., 2022). However, most studies analyzing and interpreting the learned kernels and feature maps of CNNs, remained confined to the traditional architectures, using regular convolutions\u00a0(Zhou et\u00a0al., 2018; Bau et\u00a0al., 2017). In contrast, the emergent properties and associated interpretability of DS-CNNs are nowadays largely under-explored.   Through an extensive investigation of several model types and sizes, including regular CNNs and DS-CNNs, trained on ImageNet-1k and ImageNet-21k, we discovered a striking property of DS-CNN kernels. Unlike regular convolutions, depthwise convolutions exhibit repeating patterns in their trained kernel weights. Moreover, these patterns persist throughout all layers, even in deeper stages of the DS-CNNs. This reveals a new level of structure and interpretability in the emergent representations learned by DS-CNNs that has not been identified and characterized before.      (a) D\u2062o\u2062G\u2062(x,y)\ud835\udc37\ud835\udc5c\ud835\udc3a\ud835\udc65\ud835\udc66DoG(x,y)italic_D italic_o italic_G ( italic_x , italic_y )     (b) \u2212D\u2062o\u2062G\u2062(x,y)\ud835\udc37\ud835\udc5c\ud835\udc3a\ud835\udc65\ud835\udc66-DoG(x,y)- italic_D italic_o italic_G ( italic_x , italic_y )     (c) d\u2062D\u2062o\u2062G\u2062(x,y)d\u2062x\ud835\udc51\ud835\udc37\ud835\udc5c\ud835\udc3a\ud835\udc65\ud835\udc66\ud835\udc51\ud835\udc65\\frac{dDoG(x,y)}{dx}divide start_ARG italic_d italic_D italic_o italic_G ( italic_x , italic_y ) end_ARG start_ARG italic_d italic_x end_ARG     (d) d2\u2062D\u2062o\u2062G\u2062(x,y)d\u2062x2superscript\ud835\udc512\ud835\udc37\ud835\udc5c\ud835\udc3a\ud835\udc65\ud835\udc66\ud835\udc51superscript\ud835\udc652\\frac{d^{2}DoG(x,y)}{dx^{2}}divide start_ARG italic_d start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_D italic_o italic_G ( italic_x , italic_y ) end_ARG start_ARG italic_d italic_x start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT end_ARG     Figure 1: 3D and 2D plots of the DoG function and its derivatives, utilized in Neuroscience for modeling visual receptive fields and in Image Processing for edge and brightness change detection.   Furthermore, we discovered that these identifiable patterns are highly clusterable. To this end, we employed an unsupervised autoencoder-based clustering methodology, mapping each kernel to a single hidden dimension. In the autoencoder outputs, clear clusters emerged, enabling us to categorize nearly all kernels across all layers into eight main classes. Through this classification, we uncovered an intriguing property: the prominent clusters have spatial patterns resembling the difference of Gaussian (DoG) functions and their first and second-order derivatives, as shown in Figure\u00a01.   Fascinatingly, DoG derivatives have been extensively proposed",
            "references": [
                {
                    "title": "Deep Continuous Networks",
                    "abstract": "CNNs and computational models of biological vision share some fundamental principles, which opened new avenues of research. However, fruitful cross-field research is hampered by conventional CNN architectures being based on spatially and depthwise discrete representations, which cannot accommodate certain aspects of biological complexity such as continuously varying receptive field sizes and dynamics of neuronal responses. Here we propose deep continuous networks (DCNs), which combine spatially continuous filters, with the continuous depth framework of neural ODEs. This allows us to learn the spatial support of the filters during training, as well as model the continuous evolution of feature maps, linking DCNs closely to biological models. We show that DCNs are versatile and highly applicable to standard image classification and reconstruction problems, where they improve parameter and data efficiency, and allow for meta-parametrization. We illustrate the biological plausibility of the scale distributions learned by DCNs and explore their performance in a neuroscientifically inspired pattern completion task. Finally, we investigate an efficient implementation of DCNs by changing input contrast."
                },
                {
                    "title": "Neural Echos: Depthwise Convolutional Filters Replicate Biological Receptive Fields",
                    "abstract": "In this study, we present evidence suggesting that depthwise convolutional kernels are effectively replicating the structural intricacies of the biological receptive fields observed in the mammalian retina. We provide analytics of trained kernels from various state-of-the-art models substantiating this evidence. Inspired by this intriguing discovery, we propose an initialization scheme that draws inspiration from the biological receptive fields. Experimental analysis of the ImageNet dataset with multiple CNN architectures featuring depthwise convolutions reveals a marked enhancement in the accuracy of the learned model when initialized with biologically derived weights. This underlies the potential for biologically inspired computational models to further our understanding of vision processing systems and to improve the efficacy of convolutional networks."
                },
                {
                    "title": "ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders",
                    "abstract": "Driven by improved architectures and better representation learning frameworks, the field of visual recognition has enjoyed rapid modernization and performance boost in the early 2020s. For example, modern ConvNets, represented by ConvNeXt [33], have demonstrated strong performance in various scenarios. While these models were originally designed for supervised learning with ImageNet labels, they can also potentially benefit from self-supervised learning techniques such as masked autoencoders (MAE) [14]. However, we found that simply combining these two approaches leads to subpar performance. In this paper, we propose a fully convolutional masked autoencoder framework and a new Global Response Normalization (GRN) layer that can be added to the ConvNeXt architecture to enhance inter-channel feature competition. This co-design of self-supervised learning techniques and architectural improvement results in a new model family called ConvNeXt V2, which significantly improves the performance of pure ConvNets on various recognition benchmarks, including ImageNet classification, COCO detection, and ADE20K segmentation. We also provide pre-trained ConvNeXt V2 models of various sizes, ranging from an efficient 3.7M-parameter Atto model with 76.7% top-1 accuracy on ImageNet, to a 650M Huge model that achieves a state-of-the-art 88.9% accuracy using only public training data."
                },
                {
                    "title": "HorNet: Efficient High-Order Spatial Interactions with Recursive Gated Convolutions",
                    "abstract": "Recent progress in vision Transformers exhibits great success in various tasks driven by the new spatial modeling mechanism based on dot-product self-attention. In this paper, we show that the key ingredients behind the vision Transformers, namely input-adaptive, long-range and high-order spatial interactions, can also be efficiently implemented with a convolution-based framework. We present the Recursive Gated Convolution ($\\textit{g}^\\textit{n}$Conv) that performs high-order spatial interactions with gated convolutions and recursive designs. The new operation is highly flexible and customizable, which is compatible with various variants of convolution and extends the two-order interactions in self-attention to arbitrary orders without introducing significant extra computation. $\\textit{g}^\\textit{n}$Conv can serve as a plug-and-play module to improve various vision Transformers and convolution-based models. Based on the operation, we construct a new family of generic vision backbones named HorNet. Extensive experiments on ImageNet classification, COCO object detection and ADE20K semantic segmentation show HorNet outperform Swin Transformers and ConvNeXt by a significant margin with similar overall architecture and training configurations. HorNet also shows favorable scalability to more training data and larger model sizes. Apart from the effectiveness in visual encoders, we also show $\\textit{g}^\\textit{n}$Conv can be applied to task-specific decoders and consistently improve dense prediction performance with less computation. Our results demonstrate that $\\textit{g}^\\textit{n}$Conv can be a new basic module for visual modeling that effectively combines the merits of both vision Transformers and CNNs. Code is available at https://github.com/raoyongming/HorNet"
                },
                {
                    "title": "CNN Filter DB: An Empirical Investigation of Trained Convolutional Filters",
                    "abstract": "Currently, many theoretical as well as practically relevant questions towards the transferability and robustness of Convolutional Neural Networks (CNNs) remain unsolved. While ongoing research efforts are engaging these problems from various angles, in most computer vision related cases these approaches can be generalized to investigations of the effects of distribution shifts in image data. In this context, we propose to study the shifts in the learned weights of trained CNN models. Here we focus on the properties of the distributions of dominantly used 3\u00d73 convolution filter kernels. We collected and publicly provide a dataset with over 1.4 billion filters from hundreds of trained CNNs, using a wide range of datasets, architectures, and vision tasks. In a first use case of the proposed dataset, we can show highly relevant properties of many publicly available pre-trained models for practical applications: I) We analyze distribution shifts (or the lack thereof) between trained filters along different axes of meta-parameters, like visual category of the dataset, task, architecture, or layer depth. Based on these results, we conclude that model pre-training can succeed on arbitrary datasets if they meet size and variance conditions. II) We show that many pre-trained models contain degenerated filters which make them less robust and less suitable for fine-tuning on target applications. Data & Project website: https://github.com/paulgavrikov/cnn-filter-db."
                },
                {
                    "title": "Patches Are All You Need?",
                    "abstract": "Although convolutional networks have been the dominant architecture for vision tasks for many years, recent experiments have shown that Transformer-based models, most notably the Vision Transformer (ViT), may exceed their performance in some settings. However, due to the quadratic runtime of the self-attention layers in Transformers, ViTs require the use of patch embeddings, which group together small regions of the image into single input features, in order to be applied to larger image sizes. This raises a question: Is the performance of ViTs due to the inherently-more-powerful Transformer architecture, or is it at least partly due to using patches as the input representation? In this paper, we present some evidence for the latter: specifically, we propose the ConvMixer, an extremely simple model that is similar in spirit to the ViT and the even-more-basic MLP-Mixer in that it operates directly on patches as input, separates the mixing of spatial and channel dimensions, and maintains equal size and resolution throughout the network. In contrast, however, the ConvMixer uses only standard convolutions to achieve the mixing steps. Despite its simplicity, we show that the ConvMixer outperforms the ViT, MLP-Mixer, and some of their variants for similar parameter counts and data set sizes, in addition to outperforming classical vision models such as the ResNet. Our code is available at https://github.com/locuslab/convmixer."
                },
                {
                    "title": "A ConvNet for the 2020s",
                    "abstract": "The \u201cRoaring 20s\u201d of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually \u201cmodernize\u201d a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets."
                },
                {
                    "title": "On-Off Center-Surround Receptive Fields for Accurate and Robust Image Classification",
                    "abstract": "Robustness to variations in lighting conditions is a key objective for any deep vision system. To this end, our paper extends the receptive field of convolutional neural networks with two residual components, ubiquitous in the visual processing system of vertebrates: On-center and off-center pathways, with excitatory center and inhibitory surround; OOCS for short. The on-center pathway is excited by the presence of a light stimulus in its center but not in its surround, whereas the off-center one is excited by the absence of a light stimulus in its center but not in its surround. We design OOCS pathways via a difference of Gaussians, with their variance computed analytically from the size of the receptive fields. OOCS pathways complement each other in their response to light stimuli, ensuring this way a strong edge-detection capability, and as a result, an accurate and robust inference under challenging lighting conditions. We provide extensive empirical evidence showing that networks supplied with the OOCS edge representation gain accuracy and illumination-robustness compared to standard deep models."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "MixConv: Mixed Depthwise Convolutional Kernels",
                    "abstract": "Depthwise convolution is becoming increasingly popular in modern efficient ConvNets, but its kernel size is often overlooked. In this paper, we systematically study the impact of different kernel sizes, and observe that combining the benefits of multiple kernel sizes can lead to better accuracy and efficiency. Based on this observation, we propose a new mixed depthwise convolution (MixConv), which naturally mixes up multiple kernel sizes in a single convolution. As a simple drop-in replacement of vanilla depthwise convolution, our MixConv improves the accuracy and efficiency for existing MobileNets on both ImageNet classification and COCO object detection. To demonstrate the effectiveness of MixConv, we integrate it into AutoML search space and develop a new family of models, named as MixNets, which outperform previous mobile models including MobileNetV2 [20] (ImageNet top-1 accuracy +4.2%), ShuffleNetV2 [16] (+3.5%), MnasNet [26] (+1.3%), ProxylessNAS [2] (+2.2%), and FBNet [27] (+2.0%). In particular, our MixNet-L achieves a new state-of-the-art 78.9% ImageNet top-1 accuracy under typical mobile settings (<600M FLOPS). Code is at this https URL tensorflow/tpu/tree/master/models/official/mnasnet/mixnet"
                },
                {
                    "title": "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks",
                    "abstract": "Convolutional Neural Networks (ConvNets) are commonly developed at a fixed resource budget, and then scaled up for better accuracy if more resources are available. In this paper, we systematically study model scaling and identify that carefully balancing network depth, width, and resolution can lead to better performance. Based on this observation, we propose a new scaling method that uniformly scales all dimensions of depth/width/resolution using a simple yet highly effective compound coefficient. We demonstrate the effectiveness of this method on scaling up MobileNets and ResNet. \nTo go even further, we use neural architecture search to design a new baseline network and scale it up to obtain a family of models, called EfficientNets, which achieve much better accuracy and efficiency than previous ConvNets. In particular, our EfficientNet-B7 achieves state-of-the-art 84.4% top-1 / 97.1% top-5 accuracy on ImageNet, while being 8.4x smaller and 6.1x faster on inference than the best existing ConvNet. Our EfficientNets also transfer well and achieve state-of-the-art accuracy on CIFAR-100 (91.7%), Flowers (98.8%), and 3 other transfer learning datasets, with an order of magnitude fewer parameters. Source code is at this https URL."
                },
                {
                    "title": "Searching for MobileNetV3",
                    "abstract": "We present the next generation of MobileNets based on a combination of complementary search techniques as well as a novel architecture design. MobileNetV3 is tuned to mobile phone CPUs through a combination of hardware-aware network architecture search (NAS) complemented by the NetAdapt algorithm and then subsequently improved through novel architecture advances. This paper starts the exploration of how automated search algorithms and network design can work together to harness complementary approaches improving the overall state of the art. Through this process we create two new MobileNet models for release: MobileNetV3-Large and MobileNetV3-Small which are targeted for high and low resource use cases. These models are then adapted and applied to the tasks of object detection and semantic segmentation. For the task of semantic segmentation (or any dense pixel prediction), we propose a new efficient segmentation decoder Lite Reduced Atrous Spatial Pyramid Pooling (LR-ASPP). We achieve new state of the art results for mobile classification, detection and segmentation. MobileNetV3-Large is 3.2% more accurate on ImageNet classification while reducing latency by 20% compared to MobileNetV2. MobileNetV3-Small is 6.6% more accurate compared to a MobileNetV2 model with comparable latency. MobileNetV3-Large detection is over 25% faster at roughly the same accuracy as MobileNetV2 on COCO detection. MobileNetV3-Large LR-ASPP is 34% faster than MobileNetV2 R-ASPP at similar accuracy for Cityscapes segmentation."
                },
                {
                    "title": "MnasNet: Platform-Aware Neural Architecture Search for Mobile",
                    "abstract": "Designing convolutional neural networks (CNN) for mobile devices is challenging because mobile models need to be small and fast, yet still accurate. Although significant efforts have been dedicated to design and improve mobile CNNs on all dimensions, it is very difficult to manually balance these trade-offs when there are so many architectural possibilities to consider. In this paper, we propose an automated mobile neural architecture search (MNAS) approach, which explicitly incorporate model latency into the main objective so that the search can identify a model that achieves a good trade-off between accuracy and latency. Unlike previous work, where latency is considered via another, often inaccurate proxy (e.g., FLOPS), our approach directly measures real-world inference latency by executing the model on mobile phones. To further strike the right balance between flexibility and search space size, we propose a novel factorized hierarchical search space that encourages layer diversity throughout the network. Experimental results show that our approach consistently outperforms state-of-the-art mobile CNN models across multiple vision tasks. On the ImageNet classification task, our MnasNet achieves 75.2% top-1 accuracy with 78ms latency on a Pixel phone, which is 1.8\u00d7 faster than MobileNetV2 with 0.5% higher accuracy and 2.3\u00d7 faster than NASNet with 1.2% higher accuracy. Our MnasNet also achieves better mAP quality than MobileNets for COCO object detection. Code is at https://github.com/tensorflow/tpu/tree/master/models/official/mnasnet."
                },
                {
                    "title": "Network Dissection: Quantifying Interpretability of Deep Visual Representations",
                    "abstract": "We propose a general framework called Network Dissection for quantifying the interpretability of latent representations of CNNs by evaluating the alignment between individual hidden units and a set of semantic concepts. Given any CNN model, the proposed method draws on a data set of concepts to score the semantics of hidden units at each intermediate convolutional layer. The units with semantics are labeled across a broad range of visual concepts including objects, parts, scenes, textures, materials, and colors. We use the proposed method to test the hypothesis that interpretability is an axis-independent property of the representation space, then we apply the method to compare the latent representations of various networks when trained to solve different classification problems. We further analyze the effect of training iterations, compare networks trained with different initializations, and measure the effect of dropout and batch normalization on the interpretability of deep visual representations. We demonstrate that the proposed method can shed light on characteristics of CNN models and training methods that go beyond measurements of their discriminative power."
                },
                {
                    "title": "MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications",
                    "abstract": "We present a class of efficient models called MobileNets for mobile and embedded vision applications. MobileNets are based on a streamlined architecture that uses depth-wise separable convolutions to build light weight deep neural networks. We introduce two simple global hyper-parameters that efficiently trade off between latency and accuracy. These hyper-parameters allow the model builder to choose the right sized model for their application based on the constraints of the problem. We present extensive experiments on resource and accuracy tradeoffs and show strong performance compared to other popular models on ImageNet classification. We then demonstrate the effectiveness of MobileNets across a wide range of applications and use cases including object detection, finegrain classification, face attributes and large scale geo-localization."
                },
                {
                    "title": "Understanding Neural Networks Through Deep Visualization",
                    "abstract": "Recent years have produced great advances in training large, deep neural networks (DNNs), including notable successes in training convolutional neural networks (convnets) to recognize natural images. However, our understanding of how these models work, especially what computations they perform at intermediate layers, has lagged behind. Progress in the field will be further accelerated by the development of better tools for visualizing and interpreting neural nets. We introduce two such tools here. The first is a tool that visualizes the activations produced on each layer of a trained convnet as it processes an image or video (e.g. a live webcam stream). We have found that looking at live activations that change in response to user input helps build valuable intuitions about how convnets work. The second tool enables visualizing features at each layer of a DNN via regularized optimization in image space. Because previous versions of this idea produced less recognizable images, here we introduce several new regularization methods that combine to produce qualitatively clearer, more interpretable visualizations. Both tools are open source and work on a pre-trained convnet with minimal setup."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "The Gaussian derivative model for spatial vision: I. Retinal mechanisms.",
                    "abstract": "Physiological evidence is presented that visual receptive fields in the primate eye are shaped like the sum of a Gaussian function and its Laplacian. A new 'difference-of-offset-Gaussians' or DOOG neural mechanism was identified, which provided a plausible neural mechanism for generating such Gaussian derivative-like fields. The DOOG mechanism and the associated Gaussian derivative model provided a better approximation to the data than did the Gabor or other competing models. A model-free Wiener filter analysis provided independent confirmation of these results. A machine vision system was constructed to simulate human foveal retinal vision, based on Gaussian derivative filters. It provided edge and line enhancement (deblurring) and noise suppression, while retaining all the information in the original image."
                },
                {
                    "title": "The two-dimensional spatial structure of simple receptive fields in cat striate cortex.",
                    "abstract": "1. A reverse correlation (6, 8, 25, 35) method is developed that allows quantitative determination of visual receptive-field structure in two spatial dimensions. This method is applied to simple cells in the cat striate cortex. 2. It is demonstrated that the reverse correlation method yields results with several desirable properties, including convergence and reproducibility independent of modest changes in stimulus parameters. 3. In contrast to results obtained with moving stimuli, we find that the bright and dark excitatory subregions in simple receptive fields do not overlap to any great extent. This difference in results may be attributed to confounding the independent variables space and time when using moving stimuli. 4. All simple receptive fields have subregions that vary smoothly in all directions in space. There are no sharp transitions either between excitatory subregions or between subregions and the area surrounding the receptive field. 5. Simple receptive fields vary both in the number of subregions observed, in the elongation of each subregion, and in the overall elongation of the field. In contrast with results obtained using moving stimuli, we find that subregions within a given receptive field need not be the same length. 6. The hypothesis that simple receptive fields can be modeled as either even symmetric or odd symmetric about a central axis is evaluated. This hypothesis is found to be false in general. Most simple receptive fields are neither even symmetric nor odd symmetric. 7. The hypothesis that simple receptive fields can be modeled as the product of a width response profile and an orthogonal length response profile (Cartesian separability) is evaluated. This hypothesis is found to be true for only approximately 50% of the cells in our sample."
                },
                {
                    "title": "Efficient Multi-order Gated Aggregation Network",
                    "abstract": "Since the recent success of Vision Transformers (ViTs), explorations toward transformer-style architectures have triggered the resurgence of modern ConvNets. In this work, we explore the representation ability of DNNs through the lens of interaction complexities. We empirically show that interaction complexity is an overlooked but essential indi-cator for visual recognition. Accordingly, a new family of ef\ufb01cient ConvNets, named MogaNet, is presented to pursue informative context mining in pure ConvNet-based models, with preferable complexity-performance trade-offs. In MogaNet, interactions across multiple complexities are facil-itated and contextualized by leveraging two specially designed aggregation blocks in both spatial and channel interaction spaces. Extensive studies are conducted on ImageNet classi\ufb01cation, COCO object detection, and ADE20K semantic segmentation tasks. The results demonstrate that our MogaNet establishes new state-of-the-art over other popular methods in mainstream scenarios and all model scales. Typically, the lightweight MogaNet-T achieves 80.0% top-1 accuracy with only 1.44G FLOPs using re\ufb01ned training setup on ImageNet-1K, surpassing ParC-Net-S by 1.4% accuracy but saving 59% (2.04G) FLOPs."
                },
                {
                    "title": "Scaling Up Your Kernels to 31x31: Revisiting Large Kernel Design in CNNs",
                    "abstract": "For training MobileNet V2 models (Sec. 3), we use 8 GPUs, an SGD optimizer with momentum of 0.9, a batch size of 32 per GPU, input resolution of 224\u00d7224, weight decay of 4 \u00d7 10\u22125, learning rate schedule with 5-epoch warmup, initial value of 0.1 and cosine annealing for 100 epochs. For the data augmentation, we only use random cropping and left-right flipping, as a common practice. For training RepLKNet models (Sec. 4.2),we use 32 GPUs and a batch size of 64 per GPU to train for 120 epochs. The optimizer is AdamW [10] with momentum of 0.9 and weight decay of 0.05. The learning rate setting includes an initial value of 4 \u00d7 10\u22123, cosine annealing and 10-epoch warm-up. For the data augmentation and regularization, we use RandAugment [4] (\u201crand-m9-mstd0.5-inc1\u201d as implemented by timm [15]), label smoothing coefficient of 0.1, mixup [18] with \u03b1 = 0.8, CutMix with \u03b1 = 1.0, Rand Erasing [19] with probability of 25% and Stochastic Depth with a drop-path rate of 30%, following the recent works [1, 8, 9, 12]. The RepLKNet-31B reported in Sec. 4.3 is trained with the same configurations except the epoch number of 300 and drop-path rate of 50%. For finetuning the 224\u00d7224-trained RepLKNet-31B with 384\u00d7384, we use 32 GPUs, a batch size of 32 per GPU, initial learning rate of 4 \u00d7 10\u22124, cosine annealing, 1-epoch warm-up, 30 epochs, model EMA (Exponential Moving Average) with momentum of 10\u22124, the same RandAugment as above but no CutMix nor mixup."
                },
                {
                    "title": "The Gaussian derivative model for spatial-temporal vision: I. Cortical model.",
                    "abstract": "How do we see the motion of objects as well as their shapes? The Gaussian Derivative (GD) spatial model is extended to time to help answer this question. The GD spatio-temporal model requires only two numbers to describe the complete three-dimensional space-time shapes of individual receptive fields in primate visual cortex. These two numbers are the derivative numbers along the respective spatial and temporal principal axes of a given receptive field. Nine transformation parameters allow for a standard geometric association of these intrinsic axes with the extrinsic environment. The GD spatio-temporal model describes in one framework the following properties of primate simple cell fields: motion properties, number of lobes in space-time, spatial orientation. location, and size. A discrete difference-of-offset-Gaussians (DOOG) model provides a plausible physiological mechanism to form GD-like model fields in both space and time. The GD model hypothesizes that receptive fields at the first stage of processing in the visual cortex approximate 'derivative analyzers' that estimate local spatial and temporal derivatives of the intensity profile in the visual environment. The receptive fields as modeled provide operators that can allow later stages of processing in either a biological or machine vision system to estimate the motion as well as the shapes of objects in the environment."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CV",
                "cs.NE"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we interpret and understand the emergent properties and learned representations of depthwise-separable convolutional neural networks (DS-CNNs) in comparison to traditional convolutional neural networks (CNNs)?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the interpretability of deep learning models, particularly DS-CNNs, which are increasingly used in computer vision. By uncovering the structure and patterns in the learned kernels of DS-CNNs, we can enhance our understanding of how these models operate, leading to improved model design and optimization. This research could pave the way for more interpretable AI systems, fostering trust and transparency in machine learning applications across various domains, including healthcare, autonomous systems, and beyond.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem stem from the complexity of deep learning architectures and the non-intuitive nature of their learned representations. Naive approaches may fail because they often rely on traditional interpretability techniques that are not suited for the unique characteristics of DS-CNNs. Additionally, the intricate relationships between kernel weights and their spatial patterns, especially in deeper layers, pose significant theoretical and practical obstacles that require sophisticated analytical methods to unravel.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on traditional CNN architectures, leaving a gap in the understanding of DS-CNNs. The lack of attention to the unique properties of depthwise convolutions and their emergent representations has hindered progress. Barriers such as the complexity of analyzing high-dimensional kernel weights and the absence of tailored interpretability frameworks for DS-CNNs have prevented this problem from being adequately addressed. Our approach differs by employing an unsupervised autoencoder-based clustering methodology specifically designed to reveal the structured patterns in DS-CNN kernels.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves training various DS-CNN models on ImageNet-1k and ImageNet-21k datasets, followed by an analysis of the learned kernel weights using an unsupervised autoencoder for clustering. We will categorize the kernels into distinct classes based on their spatial patterns, particularly focusing on those resembling difference of Gaussian (DoG) functions and their derivatives. The expected outcomes include a comprehensive classification of DS-CNN kernels, revealing new insights into their interpretability and structure, which could inform future"
            }
        },
        "author_data": {
            "9bc43498-a5bd-4772-b4d0-dc8e1e3903c1": {
                "pk": "9bc43498-a5bd-4772-b4d0-dc8e1e3903c1",
                "name": "Zahra Babaiee",
                "collaborators": [
                    "R. Grosu",
                    "Daniela Rus",
                    "Shrajan Bhandary",
                    "T. Fechter",
                    "C. Zamboglou",
                    "Ramin M. Hasani",
                    "Julian Lemmel",
                    "Sophie A. Neubauer",
                    "Dejan Kuhn",
                    "Anca-Ligia Grosu"
                ],
                "domain": [
                    "Medical Image Analysis",
                    "Deep Learning",
                    "Convolutional Neural Networks",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Segmentation of Prostate Tumour Volumes from PET Images is a Different Ball Game",
                        "abstract": "Accurate segmentation of prostate tumours from PET images presents a formidable challenge in medical image analysis. Despite considerable work and improvement in delineating organs from CT and MR modalities, the existing standards do not transfer well and produce quality results in PET related tasks. Particularly, contemporary methods fail to accurately consider the intensity-based scaling applied by the physicians during manual annotation of tumour contours. In this paper, we observe that the prostate-localised uptake threshold ranges are beneficial for suppressing outliers. Therefore, we utilize the intensity threshold values, to implement a new custom-feature-clipping normalisation technique. We evaluate multiple, established U-Net variants under different normalisation schemes, using the nnU-Net framework. All models were trained and tested on multiple datasets, obtained with two radioactive tracers: [68-Ga]Ga-PSMA-11 and [18-F]PSMA-1007. Our results show that the U-Net models achieve much better performance when the PET scans are preprocessed with our novel clipping technique."
                    },
                    {
                        "title": "Scalable Offline Reinforcement Learning for Mean Field Games",
                        "abstract": "Reinforcement learning algorithms for mean-field games offer a scalable framework for optimizing policies in large populations of interacting agents. Existing methods often depend on online interactions or access to system dynamics, limiting their practicality in real-world scenarios where such interactions are infeasible or difficult to model. In this paper, we present Offline Munchausen Mirror Descent (Off-MMD), a novel mean-field RL algorithm that approximates equilibrium policies in mean-field games using purely offline data. By leveraging iterative mirror descent and importance sampling techniques, Off-MMD estimates the mean-field distribution from static datasets without relying on simulation or environment dynamics. Additionally, we incorporate techniques from offline reinforcement learning to address common issues like Q-value overestimation, ensuring robust policy learning even with limited data coverage. Our algorithm scales to complex environments and demonstrates strong performance on benchmark tasks like crowd exploration or navigation, highlighting its applicability to real-world multi-agent systems where online experimentation is infeasible. We empirically demonstrate the robustness of Off-MMD to low-quality datasets and conduct experiments to investigate its sensitivity to hyperparameter choices."
                    },
                    {
                        "title": "Neural Echos: Depthwise Convolutional Filters Replicate Biological Receptive Fields",
                        "abstract": "In this study, we present evidence suggesting that depthwise convolutional kernels are effectively replicating the structural intricacies of the biological receptive fields observed in the mammalian retina. We provide analytics of trained kernels from various state-of-the-art models substantiating this evidence. Inspired by this intriguing discovery, we propose an initialization scheme that draws inspiration from the biological receptive fields. Experimental analysis of the ImageNet dataset with multiple CNN architectures featuring depthwise convolutions reveals a marked enhancement in the accuracy of the learned model when initialized with biologically derived weights. This underlies the potential for biologically inspired computational models to further our understanding of vision processing systems and to improve the efficacy of convolutional networks."
                    },
                    {
                        "title": "Prediction of Tourism Flow with Sparse Geolocation Data",
                        "abstract": "Modern tourism in the 21st century is facing numerous challenges. Among these the rapidly growing number of tourists visiting space-limited regions like historical cities, museums and bottlenecks such as bridges is one of the biggest. In this context, a proper and accurate prediction of tourism volume and tourism flow within a certain area is important and critical for visitor management tasks such as sustainable treatment of the environment and prevention of overcrowding. Static flow control methods like conventional low-level controllers or limiting access to overcrowded venues could not solve the problem yet. In this paper, we empirically evaluate the performance of state-of-the-art deep-learning methods such as RNNs, GNNs, and Transformers as well as the classic statistical ARIMA method. Granular limited data supplied by a tourism region is extended by exogenous data such as geolocation trajectories of individual tourists, weather and holidays. In the field of visitor flow prediction with sparse data, we are thereby capable of increasing the accuracy of our predictions, incorporating modern input feature handling as well as mapping geolocation data on top of discrete POI data."
                    },
                    {
                        "title": "Deep-Learning vs Regression: Prediction of Tourism Flow with Limited Data",
                        "abstract": "Modern tourism in the 21st century is facing numerous challenges. One of these challenges is the rapidly growing number of tourists in space limited regions such as historical city centers, museums or geographical bottlenecks like narrow valleys. In this context, a proper and accurate prediction of tourism volume and tourism flow within a certain area is important and critical for visitor management tasks such as visitor flow control and prevention of overcrowding. Static flow control methods like limiting access to hotspots or using conventional low level controllers could not solve the problem yet. In this paper, we empirically evaluate the performance of several state-of-the-art deep-learning methods in the field of visitor flow prediction with limited data by using available granular data supplied by a tourism region and comparing the results to ARIMA, a classical statistical method. Our results show that deep-learning models yield better predictions compared to the ARIMA method, while both featuring faster inference times and being able to incorporate additional input features."
                    },
                    {
                        "title": "Entangled Residual Mappings",
                        "abstract": "Residual mappings have been shown to perform representation learning in the first layers and iterative feature refinement in higher layers. This interplay, combined with their stabilizing effect on the gradient norms, enables them to train very deep networks. In this paper, we take a step further and introduce entangled residual mappings to generalize the structure of the residual connections and evaluate their role in iterative learning representations. An entangled residual mapping replaces the identity skip connections with specialized entangled mappings such as orthogonal, sparse, and structural correlation matrices that share key attributes (eigenvalues, structure, and Jacobian norm) with identity mappings. We show that while entangled mappings can preserve the iterative refinement of features across various deep models, they influence the representation learning process in convolutional networks differently than attention-based models and recurrent neural networks. In general, we find that for CNNs and Vision Transformers entangled sparse mapping can help generalization while orthogonal mappings hurt performance. For recurrent networks, orthogonal residual mappings form an inductive bias for time-variant sequences, which degrades accuracy on time-invariant tasks."
                    },
                    {
                        "title": "End-to-End Sensitivity-Based Filter Pruning",
                        "abstract": "In this paper, we present a novel sensitivity-based filter pruning algorithm (SbF-Pruner) to learn the importance scores of filters of each layer end-to-end. Our method learns the scores from the filter weights, enabling it to account for the correlations between the filters of each layer. Moreover, by training the pruning scores of all layers simultaneously our method can account for layer interdependencies, which is essential to find a performant sparse sub-network. Our proposed method can train and generate a pruned network from scratch in a straightforward, one-stage training process without requiring a pretrained network. Ultimately, we do not need layer-specific hyperparameters and pre-defined layer budgets, since SbF-Pruner can implicitly determine the appropriate number of channels in each layer. Our experimental results on different network architectures suggest that SbF-Pruner outperforms advanced pruning methods. Notably, on CIFAR-10, without requiring a pretrained baseline network, we obtain 1.02% and 1.19% accuracy gain on ResNet56 and ResNet110, compared to the baseline reported for state-of-the-art pruning algorithms. This is while SbF-Pruner reduces parameter-count by 52.3% (for ResNet56) and 54% (for ResNet101), which is better than the state-of-the-art pruning algorithms with a high margin of 9.5% and 6.6%."
                    },
                    {
                        "title": "Pruning by Active Attention Manipulation",
                        "abstract": "Filter pruning of a CNN is typically achieved by applying discrete masks on the CNN's filter weights or activation maps, post-training. Here, we present a new filter-importance-scoring concept named pruning by active attention manipulation (PAAM), that sparsifies the CNN's set of filters through a particular attention mechanism, during-training. PAAM learns analog filter scores from the filter weights by optimizing a cost function regularized by an additive term in the scores. As the filters are not independent, we use attention to dynamically learn their correlations. Moreover, by training the pruning scores of all layers simultaneously, PAAM can account for layer inter-dependencies, which is essential to finding a performant sparse sub-network. PAAM can also train and generate a pruned network from scratch in a straightforward, one-stage training process without requiring a pre-trained network. Finally, PAAM does not need layer-specific hyperparameters and pre-defined layer budgets, since it can implicitly determine the appropriate number of filters in each layer. Our experimental results on different network architectures suggest that PAAM outperforms state-of-the-art structured-pruning methods (SOTA). On CIFAR-10 dataset, without requiring a pre-trained baseline network, we obtain 1.02% and 1.19% accuracy gain and 52.3% and 54% parameters reduction, on ResNet56 and ResNet110, respectively. Similarly, on the ImageNet dataset, PAAM achieves 1.06% accuracy gain while pruning 51.1% of the parameters on ResNet50. For Cifar-10, this is better than the SOTA with a margin of 9.5% and 6.6%, respectively, and on ImageNet with a margin of 11%."
                    },
                    {
                        "title": "IB-U-Nets: Improving medical image segmentation tasks with 3D Inductive Biased kernels",
                        "abstract": "Despite the success of convolutional neural networks for 3D medical-image segmentation, the architectures currently used are still not robust enough to the protocols of different scanners, and the variety of image properties they produce. Moreover, access to large-scale datasets with annotated regions of interest is scarce, and obtaining good results is thus difficult. To overcome these challenges, we introduce IB-U-Nets, a novel architecture with inductive bias, inspired by the visual processing in vertebrates. With the 3D U-Net as the base, we add two 3D residual components to the second encoder blocks. They provide an inductive bias, helping U-Nets to segment anatomical structures from 3D images with increased robustness and accuracy. We compared IB-U-Nets with state-of-the-art 3D U-Nets on multiple modalities and organs, such as the prostate and spleen, using the same training and testing pipeline, including data processing, augmentation and cross-validation. Our results demonstrate the superior robustness and accuracy of IB-U-Nets, especially on small datasets, as is typically the case in medical-image analysis. IB-U-Nets source code and models are publicly available."
                    },
                    {
                        "title": "3D-OOCS: Learning Prostate Segmentation with Inductive Bias",
                        "abstract": "Despite the great success of convolutional neural networks (CNN) in 3D medical image segmentation tasks, the methods currently in use are still not robust enough to the vari- ety of image properties or artefacts. To this end, we intro-duce OOCS-enhanced networks, a novel architecture inspired by the innate nature of visual processing in the vertebrates. With different 3D U-Net variants as the base, we add two 3D residual components to the second encoder blocks: on and off center-surround (OOCS). They generalise the ganglion path- ways in the retina to a 3D setting. The use of 2D-OOCS in any standard CNN network complements the feedforward frame- work with sharp edge-detection inductive biases. The use of 3D-OOCS also helps 3D U-Nets to scrutinise and delineate anatomical structures present in 3D images with increased accuracy. We compared the state-of-the-art 3D U-Nets with their 3D-OOCS extensions and showed the superior accuracy and robustness of the latter in automatic prostate segmenta- tion from 3D Magnetic Resonance Images (MRIs)."
                    },
                    {
                        "title": "On-Off Center-Surround Receptive Fields for Accurate and Robust Image Classification",
                        "abstract": "Robustness to variations in lighting conditions is a key objective for any deep vision system. To this end, our paper extends the receptive field of convolutional neural networks with two residual components, ubiquitous in the visual processing system of vertebrates: On-center and off-center pathways, with excitatory center and inhibitory surround; OOCS for short. The on-center pathway is excited by the presence of a light stimulus in its center but not in its surround, whereas the off-center one is excited by the absence of a light stimulus in its center but not in its surround. We design OOCS pathways via a difference of Gaussians, with their variance computed analytically from the size of the receptive fields. OOCS pathways complement each other in their response to light stimuli, ensuring this way a strong edge-detection capability, and as a result, an accurate and robust inference under challenging lighting conditions. We provide extensive empirical evidence showing that networks supplied with the OOCS edge representation gain accuracy and illumination-robustness compared to standard deep models."
                    },
                    {
                        "title": "Computerized Medical Imaging and Graphics",
                        "abstract": "of targeted structures. However, accurate biomedical image segmentation is time-consuming, requires considerable experience and is prone to observer variability. In the past decade, the use of deep learning models has significantly increased in the field of medical image segmentation. At present, a vast number of anatomical structures can be demarcated on a clinician\u2019s level with deep learning models. These models would not only unload work, but they can offer unbiased characterization of the disease. The main architectures used in segmentation are the U-Net and its variants, that exhibit outstanding performances. However, reproducing results or directly comparing methods is often limited by closed source of data and the large heterogeneity among medical images. With this in mind, our intention is to provide a reliable source for assessing deep learning models. As an example, we chose the challenging task of delineating the prostate gland in multi-modal images. First, this paper provides a comprehensive review of current state-of-the-art convolutional neural networks for 3D prostate segmentation. Second, utilizing public and in-house CT and MR datasets of varying properties, we created a framework for an objective comparison of automatic prostate segmentation algorithms. The framework was used for rigorous evaluations of the models, highlighting their strengths and weaknesses."
                    }
                ]
            },
            "1d874bf9-b6ea-4e92-9a08-2379bce67ac6": {
                "pk": "1d874bf9-b6ea-4e92-9a08-2379bce67ac6",
                "name": "Peyman M. Kiasari",
                "collaborators": [
                    "Z. Babaiee",
                    "Daniela Rus",
                    "R. Grosu"
                ],
                "domain": [
                    "Computer Vision",
                    "Convolutional Neural Networks",
                    "Biological Inspiration"
                ],
                "publications": [
                    {
                        "title": "Neural Echos: Depthwise Convolutional Filters Replicate Biological Receptive Fields",
                        "abstract": "In this study, we present evidence suggesting that depthwise convolutional kernels are effectively replicating the structural intricacies of the biological receptive fields observed in the mammalian retina. We provide analytics of trained kernels from various state-of-the-art models substantiating this evidence. Inspired by this intriguing discovery, we propose an initialization scheme that draws inspiration from the biological receptive fields. Experimental analysis of the ImageNet dataset with multiple CNN architectures featuring depthwise convolutions reveals a marked enhancement in the accuracy of the learned model when initialized with biologically derived weights. This underlies the potential for biologically inspired computational models to further our understanding of vision processing systems and to improve the efficacy of convolutional networks."
                    }
                ]
            },
            "cc21b7dc-0fd3-4c66-a35c-9e8ade65d4cf": {
                "pk": "cc21b7dc-0fd3-4c66-a35c-9e8ade65d4cf",
                "name": "Daniela Rus",
                "collaborators": [
                    "Z. Babaiee",
                    "Peyman M. Kiasari",
                    "R. Grosu"
                ],
                "domain": [
                    "Computer Vision",
                    "Convolutional Neural Networks",
                    "Biological Inspiration"
                ],
                "publications": [
                    {
                        "title": "Neural Echos: Depthwise Convolutional Filters Replicate Biological Receptive Fields",
                        "abstract": "In this study, we present evidence suggesting that depthwise convolutional kernels are effectively replicating the structural intricacies of the biological receptive fields observed in the mammalian retina. We provide analytics of trained kernels from various state-of-the-art models substantiating this evidence. Inspired by this intriguing discovery, we propose an initialization scheme that draws inspiration from the biological receptive fields. Experimental analysis of the ImageNet dataset with multiple CNN architectures featuring depthwise convolutions reveals a marked enhancement in the accuracy of the learned model when initialized with biologically derived weights. This underlies the potential for biologically inspired computational models to further our understanding of vision processing systems and to improve the efficacy of convolutional networks."
                    }
                ]
            },
            "fa244e47-2ab6-488a-9c2c-8aca1f99de9a": {
                "pk": "fa244e47-2ab6-488a-9c2c-8aca1f99de9a",
                "name": "Radu Grosu",
                "collaborators": [
                    "Z. Babaiee",
                    "M'onika Farsang",
                    "Julian Lemmel",
                    "Felix Resch",
                    "Axel Brunnbauer",
                    "Sophie A. Neubauer",
                    "Daniela Rus",
                    "P. Priller",
                    "D. Ni\u010dkovi\u0107",
                    "Mugdim Bublin"
                ],
                "domain": [
                    "Neuromorphic Computing",
                    "Machine Learning",
                    "Reinforcement Learning",
                    "Medical Image Analysis"
                ],
                "publications": [
                    {
                        "title": "Neuromorphic IoT Architecture for Efficient Water Management: A Smart Village Case Study",
                        "abstract": "The exponential growth of IoT networks necessitates a paradigm shift towards architectures that offer high flexibility and learning capabilities while maintaining low energy consumption, minimal communication overhead, and low latency. Traditional IoT systems, particularly when integrated with machine learning approaches, often suffer from high communication overhead and significant energy consumption. This work addresses these challenges by proposing a neuromorphic architecture inspired by biological systems. To illustrate the practical application of our proposed architecture, we present a case study focusing on water management in the Carinthian community of Neuhaus. Preliminary results regarding water consumption prediction and anomaly detection in this community are presented. We also introduce a novel neuromorphic IoT architecture that integrates biological principles into the design of IoT systems. This architecture is specifically tailored for edge computing scenarios, where low power and high efficiency are crucial. Our approach leverages the inherent advantages of neuromorphic computing, such as asynchronous processing and event-driven communication, to create an IoT framework that is both energy-efficient and responsive. This case study demonstrates how the neuromorphic IoT architecture can be deployed in a real-world scenario, highlighting its benefits in terms of energy savings, reduced communication overhead, and improved system responsiveness."
                    },
                    {
                        "title": "Segmentation of Prostate Tumour Volumes from PET Images is a Different Ball Game",
                        "abstract": "Accurate segmentation of prostate tumours from PET images presents a formidable challenge in medical image analysis. Despite considerable work and improvement in delineating organs from CT and MR modalities, the existing standards do not transfer well and produce quality results in PET related tasks. Particularly, contemporary methods fail to accurately consider the intensity-based scaling applied by the physicians during manual annotation of tumour contours. In this paper, we observe that the prostate-localised uptake threshold ranges are beneficial for suppressing outliers. Therefore, we utilize the intensity threshold values, to implement a new custom-feature-clipping normalisation technique. We evaluate multiple, established U-Net variants under different normalisation schemes, using the nnU-Net framework. All models were trained and tested on multiple datasets, obtained with two radioactive tracers: [68-Ga]Ga-PSMA-11 and [18-F]PSMA-1007. Our results show that the U-Net models achieve much better performance when the PET scans are preprocessed with our novel clipping technique."
                    },
                    {
                        "title": "MMDVS-LF: A Multi-Modal Dynamic-Vision-Sensor Line Following Dataset",
                        "abstract": "Dynamic Vision Sensors (DVS), offer a unique advantage in control applications, due to their high temporal resolution, and asynchronous event-based data. Still, their adoption in machine learning algorithms remains limited. To address this gap, and promote the development of models that leverage the specific characteristics of DVS data, we introduce the Multi-Modal Dynamic-Vision-Sensor Line Following dataset (MMDVS-LF). This comprehensive dataset, is the first to integrate multiple sensor modalities, including DVS recordings, RGB video, odometry, and Inertial Measurement Unit (IMU) data, from a small-scale standardized vehicle. Additionally, the dataset includes eye-tracking and demographic data of drivers performing a Line Following task on a track. With its diverse range of data, MMDVS-LF opens new opportunities for developing deep learning algorithms, and conducting data science projects across various domains, supporting innovation in autonomous systems and control applications."
                    },
                    {
                        "title": "Scalable Offline Reinforcement Learning for Mean Field Games",
                        "abstract": "Reinforcement learning algorithms for mean-field games offer a scalable framework for optimizing policies in large populations of interacting agents. Existing methods often depend on online interactions or access to system dynamics, limiting their practicality in real-world scenarios where such interactions are infeasible or difficult to model. In this paper, we present Offline Munchausen Mirror Descent (Off-MMD), a novel mean-field RL algorithm that approximates equilibrium policies in mean-field games using purely offline data. By leveraging iterative mirror descent and importance sampling techniques, Off-MMD estimates the mean-field distribution from static datasets without relying on simulation or environment dynamics. Additionally, we incorporate techniques from offline reinforcement learning to address common issues like Q-value overestimation, ensuring robust policy learning even with limited data coverage. Our algorithm scales to complex environments and demonstrates strong performance on benchmark tasks like crowd exploration or navigation, highlighting its applicability to real-world multi-agent systems where online experimentation is infeasible. We empirically demonstrate the robustness of Off-MMD to low-quality datasets and conduct experiments to investigate its sensitivity to hyperparameter choices."
                    },
                    {
                        "title": "Automated Immunophenotyping Assessment for Diagnosing Childhood Acute Leukemia using Set-Transformers",
                        "abstract": "Acute Leukemia is the most common hematologic malignancy in children and adolescents. A key methodology in the diagnostic evaluation of this malignancy is immunophenotyping based on Multiparameter Flow Cytometry (FCM). However, this approach is manual, and thus time-consuming and subjective. To alleviate this situation, we propose in this paper the FCM-Former, a machine learning, self-attention based FCM-diagnostic tool, automating the immunophenotyping assessment in Childhood Acute Leukemia. The FCM-Former is trained in a supervised manner, by directly using flow cytometric data. Our FCM-Former achieves an accuracy of 96.5% assigning lineage to each sample among 960 cases of either acute B-cell, T-cell lymphoblastic, and acute myeloid leukemia (B-ALL, T-ALL, AML). To the best of our knowledge, the FCM-Former is the first work that automates the immunophenotyping assessment with FCM data in diagnosing pediatric Acute Leukemia."
                    },
                    {
                        "title": "Liquid Resistance Liquid Capacitance Networks",
                        "abstract": "We introduce liquid-resistance liquid-capacitance neural networks (LRCs), a neural-ODE model which considerably improve the generalization, accuracy, and biological plausibility of electrical equivalent circuits (EECs), liquid time-constant networks (LTCs), and saturated liquid time-constant networks (STCs), respectively. We also introduce LRC units (LRCUs), as a very efficient and accurate gated RNN-model, which results from solving LRCs with an explicit Euler scheme using just one unfolding. We empirically show and formally prove that the liquid capacitance of LRCs considerably dampens the oscillations of LTCs and STCs, while at the same time dramatically increasing accuracy even for cheap solvers. We experimentally demonstrate that LRCs are a highly competitive alternative to popular neural ODEs and gated RNNs in terms of accuracy, efficiency, and interpretability, on classic time-series benchmarks and a complex autonomous-driving lane-keeping task."
                    },
                    {
                        "title": "Neural Echos: Depthwise Convolutional Filters Replicate Biological Receptive Fields",
                        "abstract": "In this study, we present evidence suggesting that depthwise convolutional kernels are effectively replicating the structural intricacies of the biological receptive fields observed in the mammalian retina. We provide analytics of trained kernels from various state-of-the-art models substantiating this evidence. Inspired by this intriguing discovery, we propose an initialization scheme that draws inspiration from the biological receptive fields. Experimental analysis of the ImageNet dataset with multiple CNN architectures featuring depthwise convolutions reveals a marked enhancement in the accuracy of the learned model when initialized with biologically derived weights. This underlies the potential for biologically inspired computational models to further our understanding of vision processing systems and to improve the efficacy of convolutional networks."
                    },
                    {
                        "title": "Flock-Formation Control of Multi-Agent Systems using Imperfect Relative Distance Measurements",
                        "abstract": "We present distributed distance-based control (DDC), a novel approach for controlling a multi-agent system, such that it achieves a desired formation, in a resource-constrained setting. Our controller is fully distributed and only requires local state-estimation and scalar measurements of inter-agent distances. It does not require an external localization system or inter-agent exchange of state information. Our approach uses spatial-predictive control (SPC), to optimize a cost function given strictly in terms of inter-agent distances and the distance to the target location. In DDC, each agent continuously learns and updates a very abstract model of the actual system, in the form of a dictionary of three independent key-value pairs $(\\Delta \\vec s,\\Delta d)$, where \u2206d is the partial derivative of the distance measurements along a spatial direction $\\Delta \\vec s$. This is sufficient for an agent to choose the best next action. We validate our approach by using DDC to control a collection of Crazyflie drones to achieve formation flight and reach a target while maintaining flock formation."
                    },
                    {
                        "title": "Scenario-Based Curriculum Generation for Multi-Agent Autonomous Driving",
                        "abstract": "The automated generation of diverse and complex training scenarios has been an important ingredient in many complex learning tasks. Especially in real-world application domains, such as autonomous driving, auto-curriculum generation is considered vital for obtaining robust and general policies. However, crafting traffic scenarios with multiple, heterogeneous agents is typically considered as a tedious and time-consuming task, especially in more complex simulation environments. In our work, we introduce MATS-Gym, a Multi-Agent Traffic Scenario framework to train agents in CARLA, a high-fidelity driving simulator. MATS-Gym is a multi-agent training framework for autonomous driving that uses partial scenario specifications to generate traffic scenarios with variable numbers of agents. This paper unifies various existing approaches to traffic scenario description into a single training framework and demonstrates how it can be integrated with techniques from unsupervised environment design to automate the generation of adaptive auto-curricula. The code is available at https://github.com/AutonomousDrivingExaminer/mats-gym."
                    },
                    {
                        "title": "Real-Time Recurrent Reinforcement Learning",
                        "abstract": "In this paper we propose real-time recurrent reinforcement learning (RTRRL), a biologically plausible approach to solving discrete and continuous control tasks in partially-observable markov decision processes (POMDPs). RTRRL consists of three parts: (1) a Meta-RL RNN architecture, implementing on its own an actor-critic algorithm; (2) an outer reinforcement learning algorithm, exploiting temporal difference learning and dutch eligibility traces to train the Meta-RL network; and (3) random-feedback local-online (RFLO) learning, an online automatic differentiation algorithm for computing the gradients with respect to parameters of the network.Our experimental results show that by replacing the optimization algorithm in RTRRL with the biologically implausible back propagation through time (BPTT), or real-time recurrent learning (RTRL), one does not improve returns, while matching the computational complexity for BPTT, and even increasing complexity for RTRL. RTRRL thus serves as a model of learning in biological neural networks, mimicking reward pathways in the basal ganglia."
                    },
                    {
                        "title": "Learning with Chemical versus Electrical Synapses Does it Make a Difference?",
                        "abstract": "Bio-inspired neural networks have the potential to advance our understanding of neural computation and improve the state-of-the-art of AI systems. Bio-electrical synapses directly transmit neural signals, by enabling fast current flow between neurons. In contrast, bio-chemical synapses transmit neural signals indirectly, through neurotransmitters. Prior work showed that interpretable dynamics for complex robotic control, can be achieved by using chemical synapses, within a sparse, bio-inspired architecture, called Neural Circuit Policies (NCPs). However, a comparison of these two synaptic models, within the same architecture, remains an unexplored area. In this work we aim to determine the impact of using chemical synapses compared to electrical synapses, in both sparse and all-to-all connected networks. We conduct experiments with autonomous lane-keeping through a photorealistic autonomous driving simulator to evaluate their performance under diverse conditions and in the presence of noise. The experiments highlight the substantial influence of the architectural and synaptic-model choices, respectively. Our results show that employing chemical synapses yields noticeable improvements compared to electrical synapses, and that NCPs lead to better results in both synaptic models."
                    },
                    {
                        "title": "TD-Magic: From Pictures of Timing Diagrams To Formal Specifications",
                        "abstract": "We introduce TD-Magic, the first neuro-symbolic approach for translating an image of a timing-diagram (TD) to a formal specification. We overcome the lack of labelled data for supervised learning, by first developing a synthetic data generator of labelled TDs. We then use object detection techniques to identify rising and failing edges, OCR to recognise the text, and image processing algorithms to capture synchronisation patterns. Finally, we use semantic interpretation to analyse the extracted features and generate the associated formal specification. Our experiments on industrial TDs show high translation accuracy opening the way to more sophisticated requirements-extraction algorithms from pictures."
                    },
                    {
                        "title": "NimbleAI: Towards Neuromorphic Sensing-Processing 3D-integrated Chips",
                        "abstract": "The NimbleAI Horizon Europe project leverages key principles of energy-efficient visual sensing and processing in biological eyes and brains, and harnesses the latest advances in $\\mathbf{33D}$ stacked silicon integration, to create an integral sensing-processing neuromorphic architecture that efficiently and accurately runs computer vision algorithms in area-constrained endpoint chips. The rationale behind the NimbleAI architecture is: sense data only with high information value and discard data as soon as they are found not to be useful for the application (in a given context). The NimbleAI sensing-processing architecture is to be specialized after-deployment by tunning system-level trade-offs for each particular computer vision algorithm and deployment environment. The objectives of NimbleAI are: (1) $\\mathbf{100x}$ performance per mW gains compared to state-of-the-practice solutions (i.e., CPU/GPUs processing frame-based video); (2) $\\mathbf{50x}$ processing latency reduction compared to CPU/GPUs; (3) energy consumption in the order of tens of mWs; and (4) silicon area of approx. 50 mm2."
                    },
                    {
                        "title": "On the Benefits of Biophysical Synapses",
                        "abstract": "The approximation capability of ANNs and their RNN instantiations, is strongly correlated with the number of parameters packed into these networks. However, the complexity barrier for human understanding, is arguably related to the number of neurons and synapses in the networks, and to the associated nonlinear transformations. In this paper we show that the use of biophysical synapses, as found in LTCs, have two main benefits. First, they allow to pack more parameters for a given number of neurons and synapses. Second, they allow to formulate the nonlinear-network transformation, as a linear system with state-dependent coefficients. Both increase interpretability, as for a given task, they allow to learn a system linear in its input features, that is smaller in size compared to the state of the art. We substantiate the above claims on various time-series prediction tasks, but we believe that our results are applicable to any feedforward or recurrent ANN."
                    }
                ]
            }
        }
    },
    "2408.16862": {
        "paper_data": {
            "title": "Probabilistic Decomposed Linear Dynamical Systems for Robust Discovery of Latent Neural Dynamics",
            "url": "http://arxiv.org/abs/2408.16862v1",
            "arxiv_id": "2408.16862",
            "authors": [
                "Yenho Chen",
                "Noga Mudrik",
                "Kyle A. Johnsen",
                "Sankaraleengam Alagapan",
                "Adam S. Charles",
                "Christopher J. Rozell"
            ],
            "abstract": "Time-varying linear state-space models are powerful tools for obtaining mathematically interpretable representations of neural signals. For example, switching and decomposed models describe complex systems using latent variables that evolve according to simple locally linear dynamics. However, existing methods for latent variable estimation are not robust to dynamical noise and system nonlinearity due to noise-sensitive inference procedures and limited model formulations. This can lead to inconsistent results on signals with similar dynamics, limiting the model's ability to provide scientific insight. In this work, we address these limitations and propose a probabilistic approach to latent variable estimation in decomposed models that improves robustness against dynamical noise. Additionally, we introduce an extended latent dynamics model to improve robustness against system nonlinearities. We evaluate our approach on several synthetic dynamical systems, including an empirically-derived brain-computer interface experiment, and demonstrate more accurate latent variable inference in nonlinear systems with diverse noise conditions. Furthermore, we apply our method to a real-world clinical neurophysiology dataset, illustrating the ability to identify interpretable and coherent structure where previous models cannot.",
            "introduction": "   1 Introduction  A central goal in computational neuroscience is to develop models capable of discovering latent structure within noisy, high-dimensional neural signals. By identifying hidden relationships within neural recordings, we can begin to understand, predict, and control the behaviors of the underlying systems. Modeling neural time-series is challenging due to the range of temporal dynamics present. For example, there may be gradual short-term fluctuations, abrupt shifts in response to external stimuli, and long-term global drifts resulting from changes in baseline activity levels\u00a0[32, 39, 35, 10, 6].   Although black-box approaches based on deep learning are available\u00a0[30, 17, 34], their complexity often obscures the relationships learned from the data, making it difficult to extract scientific insights from these models. As a result, practitioners may favor time-varying linear state-space models which offer mathematically interpretable representations by approximating complex dynamics with simple locally linear regimes [22]. However, obtaining latent variable estimates that are robust to dynamical noise and system nonlinearity in these state-space models is challenging. When applied to neural time-series, latent variable estimation may become unstable due to inflexible model formulations or noise-sensitive inference procedures. This can incorrectly produce disparate results for signals generated from the same underlying system.   For example, switching linear dynamical systems (SLDS) and related models\u00a0[38, 40] segment time-series into discrete linear dynamical states, providing a piecewise linear approximation of the underlying system while highlighting coherent groups of activity. However, the assumption of discrete components can be a poor modeling choice for neural signals that contain continuous-valued fluctuations, such as gradual or random changes to the system speed as seen during neural ramping activity\u00a0[27] or L\u00e9vy walk dynamics in the cerebral cortex\u00a0[23]. As demonstrated in [26] and our experiments in Section 4.3, when applied to real-world datasets, inference of the switching variables can result in rapid, random oscillations between the discrete modes, indicating that the model is unable to identify meaningful structure in the data.   To address the limitations of discrete states, the decomposed linear dynamical systems (dLDS) model\u00a0[26] learns a dictionary set of linear dynamical regimes, referred to as dynamic operators (DOs), that can be modified and combined through a linear combination of sparse coefficients. By allowing coefficients to be time-varying and continuous-valued, dLDS naturally captures both gradual changes by adjusting the coefficient magnitudes and abrupt shifts by changing the set of active DOs over time. Inference is accomplished by optimizing over a cost function that encourages data reconstruction while also constraining the structure of the dynamic coefficients to be sparse and temporally smooth.   Unfortunately, there are two critical shortcomings that prevent the robust inference of the latent variables in dLDS. First, the cost-based inference procedure is sensitive to noise, because of the regularization term encouraging temporal smoothness. This term sequentially propagates errors from noisy coefficient estimates over the length of the time series. Consequently, the model may produce inconsistent coefficient estimates on similar signals and have poor multi-step inference performance, indicating that the learned dynamics are unable to generalize well beyond a single time step. Second, the original latent dynamics model lacks a method for accurately representing systems with multiple fixed points, causing DO coefficients to",
            "references": [
                {
                    "title": "Brain signal variability and executive functions across the life span",
                    "abstract": "Abstract Neural variability is thought to facilitate survival through flexible adaptation to changing environmental demands. In humans, such capacity for flexible adaptation may manifest as fluid reasoning, inhibition of automatic responses, and mental set-switching\u2014skills falling under the broad domain of executive functions that fluctuate over the life span. Neural variability can be quantified via the BOLD signal in resting-state fMRI. Variability of large-scale brain networks is posited to underpin complex cognitive activities requiring interactions between multiple brain regions. Few studies have examined the extent to which network-level brain signal variability across the life span maps onto high-level processes under the umbrella of executive functions. The present study leveraged a large publicly available neuroimaging dataset to investigate the relationship between signal variability and executive functions across the life span. Associations between brain signal variability and executive functions shifted as a function of age. Limbic-specific variability was consistently associated with greater performance across subcomponents of executive functions. Associations between executive function subcomponents and network-level variability of the default mode and central executive networks, as well as whole-brain variability, varied across the life span. Findings suggest that brain signal variability may help to explain to age-related differences in executive functions across the life span."
                },
                {
                    "title": "Switching Autoregressive Low-rank Tensor Models",
                    "abstract": "An important problem in time-series analysis is modeling systems with time-varying dynamics. Probabilistic models with joint continuous and discrete latent states offer interpretable, efficient, and experimentally useful descriptions of such data. Commonly used models include autoregressive hidden Markov models (ARHMMs) and switching linear dynamical systems (SLDSs), each with its own advantages and disadvantages. ARHMMs permit exact inference and easy parameter estimation, but are parameter intensive when modeling long dependencies, and hence are prone to overfitting. In contrast, SLDSs can capture long-range dependencies in a parameter efficient way through Markovian latent dynamics, but present an intractable likelihood and a challenging parameter estimation task. In this paper, we propose switching autoregressive low-rank tensor (SALT) models, which retain the advantages of both approaches while ameliorating the weaknesses. SALT parameterizes the tensor of an ARHMM with a low-rank factorization to control the number of parameters and allow longer range dependencies without overfitting. We prove theoretical and discuss practical connections between SALT, linear dynamical systems, and SLDSs. We empirically demonstrate quantitative advantages of SALT models on a range of simulated and real prediction tasks, including behavioral and neural datasets. Furthermore, the learned low-rank tensor provides novel insights into temporal dependencies within each discrete state."
                },
                {
                    "title": "Decomposed Linear Dynamical Systems (dLDS) for learning the latent components of neural dynamics",
                    "abstract": "Learning interpretable representations of neural dynamics at a population level is a crucial first step to understanding how observed neural activity relates to perception and behavior. Models of neural dynamics often focus on either low-dimensional projections of neural activity, or on learning dynamical systems that explicitly relate to the neural state over time. We discuss how these two approaches are interrelated by considering dynamical systems as representative of flows on a low-dimensional manifold. Building on this concept, we propose a new decomposed dynamical system model that represents complex non-stationary and nonlinear dynamics of time series data as a sparse combination of simpler, more interpretable components. Our model is trained through a dictionary learning procedure, where we leverage recent results in tracking sparse vectors over time. The decomposed nature of the dynamics is more expressive than previous switched approaches for a given number of parameters and enables modeling of overlapping and non-stationary dynamics. In both continuous-time and discrete-time instructional examples we demonstrate that our model can well approximate the original system, learn efficient representations, and capture smooth transitions between dynamical modes, focusing on intuitive low-dimensional non-stationary linear and nonlinear systems. Furthermore, we highlight our model's ability to efficiently capture and demix population dynamics generated from multiple independent subnetworks, a task that is computationally impractical for switched models. Finally, we apply our model to neural\"full brain\"recordings of C. elegans data, illustrating a diversity of dynamics that is obscured when classified into discrete states."
                },
                {
                    "title": "Slow fluctuations in ongoing brain activity decrease in amplitude with ageing yet their impact on task-related evoked responses is dissociable from behavior",
                    "abstract": "In humans, ageing is characterized by decreased brain signal variability and increased behavioral variability. To understand how reduced brain variability segregates with increased behavioral variability, we investigated the association between reaction time variability, evoked brain responses and ongoing brain signal dynamics, in young (N=36) and older adults (N=39). We studied the electroencephalogram (EEG) and pupil size fluctuations to characterize the cortical and arousal responses elicited by a cued go/no-go task. Evoked responses were strongly modulated by slow (<2 Hz) fluctuations of the ongoing signals, which presented reduced power in the older participants. Although variability of the evoked responses was lower in the older participants, once we adjusted for the effect of the ongoing signal fluctuations, evoked responses were equally variable in both groups. Moreover, the modulation of the evoked responses caused by the ongoing signal fluctuations had no impact on reaction time, thereby explaining why although ongoing brain signal variability is decreased in older individuals, behavioral variability is not. Finally, we showed that adjusting for the effect of the ongoing signal was critical to unmask the link between neural responses and behavior as well as the link between task-related evoked EEG and pupil responses."
                },
                {
                    "title": "Modeling multiscale causal interactions between spiking and field potential signals during behavior",
                    "abstract": "Objective. Brain recordings exhibit dynamics at multiple spatiotemporal scales, which are measured with spike trains and larger-scale field potential signals. To study neural processes, it is important to identify and model causal interactions not only at a single scale of activity, but also across multiple scales, i.e. between spike trains and field potential signals. Standard causality measures are not directly applicable here because spike trains are binary-valued but field potentials are continuous-valued. It is thus important to develop computational tools to recover multiscale neural causality during behavior, assess their performance on neural datasets, and study whether modeling multiscale causalities can improve the prediction of neural signals beyond what is possible with single-scale causality. Approach. We design a multiscale model-based Granger-like causality method based on directed information and evaluate its success both in realistic biophysical spike-field simulations and in motor cortical datasets from two non-human primates (NHP) performing a motor behavior. To compute multiscale causality, we learn point-process generalized linear models that predict the spike events at a given time based on the history of both spike trains and field potential signals. We also learn linear Gaussian models that predict the field potential signals at a given time based on their own history as well as either the history of binary spike events or that of latent firing rates. Main results. We find that our method reveals the true multiscale causality network structure in biophysical simulations despite the presence of model mismatch. Further, models with the identified multiscale causalities in the NHP neural datasets lead to better prediction of both spike trains and field potential signals compared to just modeling single-scale causalities. Finally, we find that latent firing rates are better predictors of field potential signals compared with the binary spike events in the NHP datasets. Significance. This multiscale causality method can reveal the directed functional interactions across spatiotemporal scales of brain activity to inform basic science investigations and neurotechnologies."
                },
                {
                    "title": "Unifying and generalizing models of neural dynamics during decision-making",
                    "abstract": "An open question in systems and computational neuroscience is how neural circuits accumulate evidence towards a decision. Fitting models of decision-making theory to neural activity helps answer this question, but current approaches limit the number of these models that we can fit to neural data. Here we propose a unifying framework for modeling neural activity during decision-making tasks. The framework includes the canonical drift-diffusion model and enables extensions such as multi-dimensional accumulators, variable and collapsing boundaries, and discrete jumps. Our framework is based on constraining the parameters of recurrent state-space models, for which we introduce a scalable variational Laplace-EM inference algorithm. We applied the modeling approach to spiking responses recorded from monkey parietal cortex during two decision-making tasks. We found that a two-dimensional accumulator better captured the trial-averaged responses of a set of parietal neurons than a single accumulator model. Next, we identified a variable lower boundary in the responses of an LIP neuron during a random dot motion task."
                },
                {
                    "title": "Sparse Bayesian Learning With Dynamic Filtering for Inference of Time-Varying Sparse Signals",
                    "abstract": "Many signal processing applications require estimation of time-varying sparse signals, potentially with the knowledge of an imperfect dynamics model. In this paper, we propose an algorithm for dynamic filtering of time-varying sparse signals based on the sparse Bayesian learning (SBL) framework. The key idea underlying the algorithm, termed SBL-DF, is the incorporation of a signal prediction generated from a dynamics model and estimates of previous time steps into the hyperpriors of the SBL probability model. The proposed algorithm is online, robust to imperfect dynamics models (due to the propagation of dynamics information through higher-order statistics), robust to certain undesirable dictionary properties such as coherence (due to properties of the SBL framework), allows the use of arbitrary dynamics models, and requires the tuning of fewer parameters than many other dynamic filtering algorithms do. We also extend the fast marginal likelihood SBL inference procedure to the informative hyperprior setting to create a particularly efficient version of the SBL-DF algorithm. Numerical simulations show that SBL-DF converges much faster and to more accurate solutions than standard SBL and other dynamical filtering algorithms. In particular, we show that SBL-DF outperforms state of the art algorithms when the dictionary contains the challenging coherence and column scaling structure found in many practical applications."
                },
                {
                    "title": "Tree-Structured Recurrent Switching Linear Dynamical Systems for Multi-Scale Modeling",
                    "abstract": "Many real-world systems studied are governed by complex, nonlinear dynamics. By modeling these dynamics, we can gain insight into how these systems work, make predictions about how they will behave, and develop strategies for controlling them. While there are many methods for modeling nonlinear dynamical systems, existing techniques face a trade off between offering interpretable descriptions and making accurate predictions. Here, we develop a class of models that aims to achieve both simultaneously, smoothly interpolating between simple descriptions and more complex, yet also more accurate models. Our probabilistic model achieves this multi-scale property through a hierarchy of locally linear dynamics that jointly approximate global nonlinear dynamics. We call it the tree-structured recurrent switching linear dynamical system. To fit this model, we present a fully-Bayesian sampling procedure using Polya-Gamma data augmentation to allow for fast and conjugate Gibbs sampling. Through a variety of synthetic and real examples, we show how these models outperform existing methods in both interpretability and predictive capability."
                },
                {
                    "title": "Smooth Loss Functions for Deep Top-k Classification",
                    "abstract": "The top-k error is a common measure of performance in machine learning and computer vision. In practice, top-k classification is typically performed with deep neural networks trained with the cross-entropy loss. Theoretical results indeed suggest that cross-entropy is an optimal learning objective for such a task in the limit of infinite data. In the context of limited and noisy data however, the use of a loss function that is specifically designed for top-k classification can bring significant improvements. Our empirical evidence suggests that the loss function must be smooth and have non-sparse gradients in order to work well with deep neural networks. Consequently, we introduce a family of smoothed loss functions that are suited to top-k optimization via deep learning. The widely used cross-entropy is a special case of our family. Evaluating our smooth loss functions is computationally challenging: a na\\\"ive algorithm would require $\\mathcal{O}(\\binom{n}{k})$ operations, where n is the number of classes. Thanks to a connection to polynomial algebra and a divide-and-conquer approach, we provide an algorithm with a time complexity of $\\mathcal{O}(k n)$. Furthermore, we present a novel approximation to obtain fast and stable algorithms on GPUs with single floating point precision. We compare the performance of the cross-entropy loss and our margin-based losses in various regimes of noise and data size, for the predominant use case of k=5. Our investigation reveals that our loss is more robust to noise and overfitting than cross-entropy."
                },
                {
                    "title": "Recurrent switching linear dynamical systems",
                    "abstract": "Many natural systems, such as neurons firing in the brain or basketball teams traversing a court, give rise to time series data with complex, nonlinear dynamics. We can gain insight into these systems by decomposing the data into segments that are each explained by simpler dynamic units. Building on switching linear dynamical systems (SLDS), we present a new model class that not only discovers these dynamical units, but also explains how their switching behavior depends on observations or continuous latent states. These \"recurrent\" switching linear dynamical systems provide further insight by discovering the conditions under which each unit is deployed, something that traditional SLDS models fail to do. We leverage recent algorithmic advances in approximate inference to make Bayesian inference in these models easy, fast, and scalable."
                },
                {
                    "title": "Processing of emotional stimuli is reflected by modulations of beta band activity in the subgenual anterior cingulate cortex in patients with treatment resistant depression.",
                    "abstract": "Deep brain stimulation (DBS) of the subgenual anterior cingulate cortex (sgACC) has emerged as a new therapeutic option in patients with treatment resistant depression (TRD). At the same time, DBS offers a unique opportunity as an innovative research tool to study brain function in vivo Indirect measures of brain function such as positron-emission-tomography imaging findings have revealed a hypermetabolism in the sgACC area in patients with TRD that normalizes in parallel with treatment response to DBS. We used direct intracranial recordings via implanted DBS electrodes to study the neuronal oscillatory activity in the sgACC area during a picture viewing task including emotional and neutral stimuli in eight patients with TRD who underwent DBS.We found a stimulus-induced decrease in beta-band and increase in gamma-band activity, with a main effect of valence for event-related desynchronisation in the beta-frequency range (14-30\u2009Hz). Unpleasant stimuli induced the strongest and most sustained beta-power decrease. The degree of beta-band modulation upon emotional stimuli correlated with the patients' rating of stimulus valence. Our findings confirm the involvement of the sgACC area in emotional processing that was more enhanced for unpleasant stimuli. Moreover, stimulus evaluation may be encoded by modulations of beta-band activity."
                },
                {
                    "title": "Loss Functions for Top-k Error: Analysis and Insights",
                    "abstract": "In order to push the performance on realistic computer vision tasks, the number of classes in modern benchmark datasets has significantly increased in recent years. This increase in the number of classes comes along with increased ambiguity between the class labels, raising the question if top-1 error is the right performance measure. In this paper, we provide an extensive comparison and evaluation of established multiclass methods comparing their top-k performance both from a practical as well as from a theoretical perspective. Moreover, we introduce novel top-k loss functions as modifications of the softmax and the multiclass SVM losses and provide efficient optimization schemes for them. In the experiments, we compare on various datasets all of the proposed and established methods for top-k error optimization. An interesting insight of this paper is that the softmax loss yields competitive top-k performance for all k simultaneously. For a specific top-k error, our new top-k losses lead typically to further improvements while being faster to train than the softmax."
                },
                {
                    "title": "Computing the Local Field Potential (LFP) from Integrate-and-Fire Network Models",
                    "abstract": "Leaky integrate-and-fire (LIF) network models are commonly used to study how the spiking dynamics of neural networks changes with stimuli, tasks or dynamic network states. However, neurophysiological studies in vivo often rather measure the mass activity of neuronal microcircuits with the local field potential (LFP). Given that LFPs are generated by spatially separated currents across the neuronal membrane, they cannot be computed directly from quantities defined in models of point-like LIF neurons. Here, we explore the best approximation for predicting the LFP based on standard output from point-neuron LIF networks. To search for this best \u201cLFP proxy\u201d, we compared LFP predictions from candidate proxies based on LIF network output (e.g, firing rates, membrane potentials, synaptic currents) with \u201cground-truth\u201d LFP obtained when the LIF network synaptic input currents were injected into an analogous three-dimensional (3D) network model of multi-compartmental neurons with realistic morphology, spatial distributions of somata and synapses. We found that a specific fixed linear combination of the LIF synaptic currents provided an accurate LFP proxy, accounting for most of the variance of the LFP time course observed in the 3D network for all recording locations. This proxy performed well over a broad set of conditions, including substantial variations of the neuronal morphologies. Our results provide a simple formula for estimating the time course of the LFP from LIF network simulations in cases where a single pyramidal population dominates the LFP generation, and thereby facilitate quantitative comparison between computational models and experimental LFP recordings in vivo."
                },
                {
                    "title": "Dynamic Filtering of Time-Varying Sparse Signals via $\\ell _1$ Minimization",
                    "abstract": "Despite the importance of sparsity signal models and the increasing prevalence of high-dimensional streaming data, there are relatively few algorithms for dynamic filtering of timevarying sparse signals. Of the existing algorithms, fewer still provide strong performance guarantees. This paper examines two algorithms for dynamic filtering of sparse signals that are based on efficient \u21131 optimization methods. We first present an analysis for one simple algorithm (BPDN-DF) that works well when the system dynamics are known exactly. We then introduce a novel second algorithm (RWL1-DF) that is more computationally complex than BPDN-DF but performs better in practice, especially in the case where the system dynamics model is inaccurate. Robustness to model inaccuracy is achieved by using a hierarchical probabilistic data model and propagating higher-order statistics from the previous estimate (akin to Kalman filtering) in the sparse inference process. We demonstrate the properties of these algorithms on both simulated data as well as natural video sequences. Taken together, the algorithms presented in this paper represent the first strong performance analysis of dynamic filtering algorithms for time-varying sparse signals as well as state-of-the-art performance in this emerging application."
                },
                {
                    "title": "Contribution of LFP dynamics to single-neuron spiking variability in motor cortex during movement execution",
                    "abstract": "Understanding the sources of variability in single-neuron spiking responses is an important open problem for the theory of neural coding. This variability is thought to result primarily from spontaneous collective dynamics in neuronal networks. Here, we investigate how well collective dynamics reflected in motor cortex local field potentials (LFPs) can account for spiking variability during motor behavior. Neural activity was recorded via microelectrode arrays implanted in ventral and dorsal premotor and primary motor cortices of non-human primates performing naturalistic 3-D reaching and grasping actions. Point process models were used to quantify how well LFP features accounted for spiking variability not explained by the measured 3-D reach and grasp kinematics. LFP features included the instantaneous magnitude, phase and analytic-signal components of narrow band-pass filtered (\u03b4,\u03b8,\u03b1,\u03b2) LFPs, and analytic signal and amplitude envelope features in higher-frequency bands. Multiband LFP features predicted single-neuron spiking (1ms resolution) with substantial accuracy as assessed via ROC analysis. Notably, however, models including both LFP and kinematics features displayed marginal improvement over kinematics-only models. Furthermore, the small predictive information added by LFP features to kinematic models was redundant to information available in fast-timescale (<100 ms) spiking history. Overall, information in multiband LFP features, although predictive of single-neuron spiking during movement execution, was redundant to information available in movement parameters and spiking history. Our findings suggest that, during movement execution, collective dynamics reflected in motor cortex LFPs primarily relate to sensorimotor processes directly controlling movement output, adding little explanatory power to variability not accounted by movement parameters."
                },
                {
                    "title": "Time-resolved resting-state brain networks",
                    "abstract": "Significance Large-scale organizational properties of brain networks mapped with functional magnetic resonance imaging have been studied in a time-averaged sense. This is an oversimplification. We demonstrate that brain activity between multiple pairs of spatially distributed regions spontaneously fluctuates in and out of correlation over time in a globally coordinated manner, giving rise to sporadic intervals during which information can be efficiently exchanged between neuronal populations. We argue that dynamic fluctuations in the brain\u2019s organizational properties may minimize metabolic requirements while maintaining the brain in a responsive state. Neuronal dynamics display a complex spatiotemporal structure involving the precise, context-dependent coordination of activation patterns across a large number of spatially distributed regions. Functional magnetic resonance imaging (fMRI) has played a central role in demonstrating the nontrivial spatial and topological structure of these interactions, but thus far has been limited in its capacity to study their temporal evolution. Here, using high-resolution resting-state fMRI data obtained from the Human Connectome Project, we mapped time-resolved functional connectivity across the entire brain at a subsecond resolution with the aim of understanding how nonstationary fluctuations in pairwise interactions between regions relate to large-scale topological properties of the human brain. We report evidence for a consistent set of functional connections that show pronounced fluctuations in their strength over time. The most dynamic connections are intermodular, linking elements from topologically separable subsystems, and localize to known hubs of default mode and fronto-parietal systems. We found that spatially distributed regions spontaneously increased, for brief intervals, the efficiency with which they can transfer information, producing temporary, globally efficient network states. Our findings suggest that brain dynamics give rise to variations in complex network properties over time, possibly achieving a balance between efficient information-processing and metabolic expenditure."
                },
                {
                    "title": "The roles of monkey M1 neuron classes in movement preparation and execution",
                    "abstract": "The motor cortices exhibit substantial activity while preparing movements, yet the arm remains still during preparation. We investigated whether a subpopulation of presumed inhibitory neurons in primary motor cortex (M1) might be involved in \u201cgating\u201d motor output during preparation, while permitting output during movement. This hypothesis predicts a release of inhibition just before movement onset. In data from M1 of two monkeys, we did not find evidence for this hypothesis: few neurons exhibited a clear pause during movement, and these were at the tail end of a broad distribution. We then identified a subpopulation likely to be enriched for inhibitory interneurons, using their waveform shapes. We found that the firing rates of this subpopulation tended to increase during movement instead of decreasing as predicted by the M1 gating model. No clear subset that might implement an inhibitory gate was observed. Together with previous evidence against upstream inhibitory mechanisms in premotor cortex, this provides evidence against an inhibitory \u201cgate\u201d for motor output in cortex. Instead, it appears that some other mechanism must likely exist."
                },
                {
                    "title": "Empirical models of spiking in neural populations",
                    "abstract": "Neurons in the neocortex code and compute as part of a locally interconnected population. Large-scale multi-electrode recording makes it possible to access these population processes empirically by fitting statistical models to unaveraged data. What statistical structure best describes the concurrent spiking of cells within a local network? We argue that in the cortex, where firing exhibits extensive correlations in both time and space and where a typical sample of neurons still reflects only a very small fraction of the local population, the most appropriate model captures shared variability by a low-dimensional latent process evolving with smooth dynamics, rather than by putative direct coupling. We test this claim by comparing a latent dynamical model with realistic spiking observations to coupled generalised linear spike-response models (GLMs) using cortical recordings. We find that the latent dynamical approach outperforms the GLM in terms of goodness-of-fit, and reproduces the temporal correlations in the data more accurately. We also compare models whose observations models are either derived from a Gaussian or point-process models, finding that the non-Gaussian model provides slightly better goodness-of-fit and more realistic population spike counts."
                },
                {
                    "title": "The horseshoe estimator for sparse signals",
                    "abstract": "This paper proposes a new approach to sparsity, called the horseshoe estimator, which arises from a prior based on multivariate-normal scale mixtures. We describe the estimator's advantages over existing approaches, including its robustness, adaptivity to different sparsity patterns and analytical tractability. We prove two theorems: one that characterizes the horseshoe estimator's tail robustness and the other that demonstrates a super-efficient rate of convergence to the correct estimate of the sampling density in sparse situations. Finally, using both real and simulated data, we show that the horseshoe estimator corresponds quite closely to the answers obtained by Bayesian model averaging under a point-mass mixture prior. Copyright 2010, Oxford University Press."
                },
                {
                    "title": "Gene selection using a two-level hierarchical Bayesian model",
                    "abstract": "SUMMARY\nThe fundamental problem of gene selection via cDNA data is to identify which genes are differentially expressed across different kinds of tissue samples (e.g. normal and cancer). cDNA data contain large number of variables (genes) and usually the sample size is relatively small so the selection process can be unstable. Therefore, models which incorporate sparsity in terms of variables (genes) are desirable for this kind of problem. This paper proposes a two-level hierarchical Bayesian model for variable selection which assumes a prior that favors sparseness. We adopt a Markov chain Monte Carlo (MCMC) based computation technique to simulate the parameters from the posteriors. The method is applied to leukemia data from a previous study and a published dataset on breast cancer.\n\n\nSUPPLEMENTARY INFORMATION\nhttp://stat.tamu.edu/people/faculty/bmallick.html."
                },
                {
                    "title": "The Elements of Statistical Learning: Data Mining, Inference, and Prediction",
                    "abstract": "In the words of the authors, the goal of this book was to \u201cbring together many of the important new ideas in learning, and explain them in a statistical framework.\u201d The authors have been quite successful in achieving this objective, and their work is a welcome addition to the statistics and learning literatures. Statistics has always been interdisciplinary, borrowing ideas from diverse \u008e elds and repaying the debt with contributions, both theoretical and practical, to the other intellectual disciplines. For statistical learning, this cross-fertilization is especially noticeable. This book is a valuable resource, both for the statistician needing an introduction to machine learning and related \u008e elds and for the computer scientist wishing to learn more about statistics. Statisticians will especially appreciate that it is written in their own language. The level of the book is roughly that of a second-year doctoral student in statistics, and it will be useful as a textbook for such students. In a stimulating article, Breiman (2001) argued that statistics has been focused too much on a \u201cdata modeling culture,\u201d where the model is paramount. Breiman argued instead for an \u201calgorithmic modeling culture,\u201d with emphasis on black-box types of prediction. Breiman\u2019s article is controversial, and in his discussion, Efron objects that \u201cprediction is certainly an interesting subject, but Leo\u2019s paper overstates both its role and our profession\u2019s lack of interest in it.\u201d Although I mostly agree with Efron, I worry that the courses offered by most statistics departments include little, if any, treatment of statistical learning and prediction. (Stanford, where Efron and the authors of this book teach, is an exception.) Graduate students in statistics certainly need to know more than they do now about prediction, machine learning, statistical learning, and data mining (not disjoint subjects). I hope that graduate courses covering the topics of this book will become more common in statistics curricula. Most of the book is focused on supervised learning, where one has inputs and outputs from some system and wishes to predict unknown outputs corresponding to known inputs. The methods discussed for supervised learning include linear and logistic regression; basis expansion, such as splines and wavelets; kernel techniques, such as local regression, local likelihood, and radial basis functions; neural networks; additive models; decision trees based on recursive partitioning, such as CART; and support vector machines. There is a \u008e nal chapter on unsupervised learning, including association rules, cluster analysis, self-organizing maps, principal components and curves, and independent component analysis. Many statisticians will be unfamiliar with at least some of these algorithms. Association rules are popular for mining commercial data in what is called \u201cmarket basket analysis.\u201d The aim is to discover types of products often purchased together. Such knowledge can be used to develop marketing strategies, such as store or catalog layouts. Self-organizing maps (SOMs) involve essentially constrained k-means clustering, where prototypes are mapped to a two-dimensional curved coordinate system. Independent components analysis is similar to principal components analysis and factor analysis, but it uses higher-order moments to achieve independence, not merely zero correlation between components. A strength of the book is the attempt to organize a plethora of methods into a coherent whole. The relationships among the methods are emphasized. I know of no other book that covers so much ground. Of course, with such broad coverage, it is not possible to cover any single topic in great depth, so this book will encourage further reading. Fortunately, each chapter includes bibliographic notes surveying the recent literature. These notes and the extensive references provide a good introduction to the learning literature, including much outside of statistics. The book might be more suitable as a textbook if less material were covered in greater depth; however, such a change would compromise the book\u2019s usefulness as a reference, and so I am happier with the book as it was written."
                },
                {
                    "title": "Modeling and decoding motor cortical activity using a switching Kalman filter",
                    "abstract": "We present a switching Kalman filter model for the real-time inference of hand kinematics from a population of motor cortical neurons. Firing rates are modeled as a Gaussian mixture where the mean of each Gaussian component is a linear function of hand kinematics. A \"hidden state\" models the probability of each mixture component and evolves over time in a Markov chain. The model generalizes previous encoding and decoding methods, addresses the non-Gaussian nature of firing rates, and can cope with crudely sorted neural data common in on-line prosthetic applications."
                },
                {
                    "title": "Adaptive Sparseness for Supervised Learning",
                    "abstract": "The goal of supervised learning is to infer a functional mapping based on a set of training examples. To achieve good generalization, it is necessary to control the \"complexity\" of the learned function. In Bayesian approaches, this is done by adopting a prior for the parameters of the function being learned. We propose a Bayesian approach to supervised learning, which leads to sparse solutions; that is, in which irrelevant parameters are automatically set exactly to zero. Other ways to obtain sparse classifiers (such as Laplacian priors, support vector machines) involve (hyper)parameters which control the degree of sparseness of the resulting classifiers; these parameters have to be somehow adjusted/estimated from the training data. In contrast, our approach does not involve any (hyper)parameters to be adjusted or estimated. This is achieved by a hierarchical-Bayes interpretation of the Laplacian prior, which is then modified by the adoption of a Jeffreys' noninformative hyperprior. Implementation is carried out by an expectation-maximization (EM) algorithm. Experiments with several benchmark data sets show that the proposed approach yields state-of-the-art performance. In particular, our method outperforms SVMs and performs competitively with the best alternative techniques, although it involves no tuning or adjustment of sparseness-controlling hyperparameters."
                },
                {
                    "title": "On the maximum likelihood estimates for linear dynamic systems",
                    "abstract": "A steel slab or bloom containing 0.3 DIFFERENCE 0.15% by weight of C, 0.05 DIFFERENCE 0.60% by weight of Si, 0.60 DIFFERENCE 2.5% by weight of Mn, 0.010 DIFFERENCE 0.15% by weight of Nb, 0.005 DIFFERENCE 0.10% by weight of soluble Al and the remainder of iron is heated to such a temperature region above 1050 DEG C. and below the reheating temperature where the austenitic grain size would be 150 mu . The heated slab or bloom is then hot rolled with total rolling reduction of more than 40% with respect to a finished thickness at the non-recrystallized austenitic region and the hot rolled stocks are subjected to accelerated cooling at a rate of 5 DEG DIFFERENCE 20 DEG C./sec. from a temperature higher than the Ar3 transformation point to a temperature within 550 DEG DIFFERENCE 650 DEG C. Finally the stocks are air cooled."
                },
                {
                    "title": "Inferring Latent Dynamics Underlying Neural Population Activity via Neural Differential Equations",
                    "abstract": "An important problem in systems neuroscience is to identify the latent dynamics underlying neural population activity. Here we address this problem by introducing a low-dimensional nonlinear model for latent neural population dynamics us-ing neural ordinary differential equations (neural ODEs), with noisy sensory inputs and Poisson spike train outputs. We refer to this as the Poisson Latent Neural Differential Equations (PLNDE) model. We apply the PLNDE framework to a variety of synthetic datasets, and show that it accurately infers the phase portraits and \ufb01xed points of nonlinear systems augmented to produce spike train data, including the FitzHugh-Nagumo oscillator, a 3-dimensional nonlinear spiral, and a nonlinear sensory decision-making model with attractor dynamics. Our model signi\ufb01cantly out-performs existing methods at inferring single-trial neural \ufb01ring rates and the corresponding latent trajectories that generated them, especially in the regime where the spike counts and number of trials are low. We then apply our model to multi-region neural population recordings from medial frontal cortex of rats performing an auditory decision-making task. Our model provides a general, interpretable framework for investigating the neural mechanisms of decision-making and other cognitive computations through the lens of dynamical systems."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the robustness and accuracy of latent variable estimation in decomposed linear dynamical systems (dLDS) for modeling high-dimensional neural time-series data?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of current models in accurately capturing the complex dynamics of neural signals. Improved latent variable estimation can lead to better understanding of neural behaviors, enhance predictive capabilities, and facilitate the development of more effective interventions in neuroscience. This research could pave the way for advancements in computational neuroscience, enabling more precise modeling of brain activity and potentially leading to practical applications in neuroprosthetics, brain-computer interfaces, and understanding neurological disorders.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent noise in neural time-series data and the complexities of the underlying dynamics, which include gradual fluctuations and abrupt shifts. Naive approaches may fail because they do not account for the temporal smoothness and sparsity required for accurate inference, leading to unstable and inconsistent estimates. Additionally, the original dLDS model struggles with representing systems with multiple fixed points, complicating the modeling of continuous-valued fluctuations in neural signals. Overcoming these technical and theoretical obstacles is essential for achieving robust latent variable estimation.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on discrete state models, which do not adequately capture the continuous dynamics present in neural signals. The limitations of existing solutions, such as the sensitivity of cost-based inference procedures to noise and the inability to represent multiple fixed points, have hindered progress. Additionally, the lack of methods that allow for time-varying and continuous-valued coefficients in latent variable models has prevented the development of more effective approaches. Our proposed methodology aims to address these gaps by enhancing the dLDS framework to improve robustness and accuracy in latent variable estimation.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves enhancing the dLDS model by incorporating a robust inference procedure that mitigates the impact of noise on coefficient estimates. We will utilize a dataset of high-dimensional neural time-series recordings and evaluate our approach using metrics such as reconstruction error and coefficient stability. The expected outcomes include improved accuracy in latent variable estimation, better generalization of learned dynamics across time steps, and enhanced ability to capture both gradual and abrupt changes in"
            }
        },
        "author_data": {
            "eecea08e-ac98-4e0d-a101-82c853697978": {
                "pk": "eecea08e-ac98-4e0d-a101-82c853697978",
                "name": "Yenho Chen",
                "collaborators": [
                    "Noga Mudrik",
                    "Eva Yezerets",
                    "Christopher Rozell",
                    "Adam Charles",
                    "Brighton Ancelin",
                    "Peimeng Guan",
                    "Chiraag Kaushik",
                    "Belen Martin-Urcelay",
                    "Alex Saad-Falcon",
                    "Nakul Singh"
                ],
                "domain": [
                    "Dynamical Systems",
                    "Self-Supervised Learning",
                    "Neural Dynamics",
                    "Image Processing"
                ],
                "publications": [
                    {
                        "title": "LINOCS: Lookahead Inference of Networked Operators for Continuous Stability",
                        "abstract": "Identifying latent interactions within complex systems is key to unlocking deeper insights into their operational dynamics, including how their elements affect each other and contribute to the overall system behavior. For instance, in neuroscience, discovering neuron-to-neuron interactions is essential for understanding brain function; in ecology, recognizing the interactions among populations is key for understanding complex ecosystems. Such systems, often modeled as dynamical systems, typically exhibit noisy high-dimensional and non-stationary temporal behavior that renders their identification challenging. Existing dynamical system identification methods often yield operators that accurately capture short-term behavior but fail to predict long-term trends, suggesting an incomplete capture of the underlying process. Methods that consider extended forecasts (e.g., recurrent neural networks) lack explicit representations of element-wise interactions and require substantial training data, thereby failing to capture interpretable network operators. Here we introduce Lookahead-driven Inference of Networked Operators for Continuous Stability (LINOCS), a robust learning procedure for identifying hidden dynamical interactions in noisy time-series data. LINOCS integrates several multi-step predictions with adaptive weights during training to recover dynamical operators that can yield accurate long-term predictions. We demonstrate LINOCS' ability to recover the ground truth dynamical operators underlying synthetic time-series data for multiple dynamical systems models (including linear, piece-wise linear, time-changing linear systems' decomposition, and regularized linear time-varying systems) as well as its capability to produce meaningful operators with robust reconstructions through various real-world examples."
                    },
                    {
                        "title": "MANGO: Disentangled Image Transformation Manifolds with Grouped Operators",
                        "abstract": "Learning semantically meaningful image transformations (i.e. rotation, thickness, blur) directly from examples can be a challenging task. Recently, the Manifold Autoencoder (MAE) proposed using a set of Lie group operators to learn image transformations directly from examples. However, this approach has limitations, as the learned operators are not guaranteed to be disentangled and the training routine is prohibitively expensive when scaling up the model. To address these limitations, we propose MANGO (transformation Manifolds with Grouped Operators) for learning disentangled operators that describe image transformations in distinct latent subspaces. Moreover, our approach allows practitioners the ability to define which transformations they aim to model, thus improving the semantic meaning of the learned operators. Through our experiments, we demonstrate that MANGO enables composition of image transformations and introduces a one-phase training routine that leads to a 100x speedup over prior works."
                    },
                    {
                        "title": "AnatoMask: Enhancing Medical Image Segmentation with Reconstruction-guided Self-masking",
                        "abstract": "Due to the scarcity of labeled data, self-supervised learning (SSL) has gained much attention in 3D medical image segmentation, by extracting semantic representations from unlabeled data. Among SSL strategies, Masked image modeling (MIM) has shown effectiveness by reconstructing randomly masked images to learn detailed representations. However, conventional MIM methods require extensive training data to achieve good performance, which still poses a challenge for medical imaging. Since random masking uniformly samples all regions within medical images, it may overlook crucial anatomical regions and thus degrade the pretraining efficiency. We propose AnatoMask, a novel MIM method that leverages reconstruction loss to dynamically identify and mask out anatomically significant regions to improve pretraining efficacy. AnatoMask takes a self-distillation approach, where the model learns both how to find more significant regions to mask and how to reconstruct these masked regions. To avoid suboptimal learning, Anatomask adjusts the pretraining difficulty progressively using a masking dynamics function. We have evaluated our method on 4 public datasets with multiple imaging modalities (CT, MRI, and PET). AnatoMask demonstrates superior performance and scalability compared to existing SSL methods. The code is available at https://github.com/ricklisz/AnatoMask."
                    },
                    {
                        "title": "Decomposed Linear Dynamical Systems (dLDS) for learning the latent components of neural dynamics",
                        "abstract": "Learning interpretable representations of neural dynamics at a population level is a crucial first step to understanding how observed neural activity relates to perception and behavior. Models of neural dynamics often focus on either low-dimensional projections of neural activity, or on learning dynamical systems that explicitly relate to the neural state over time. We discuss how these two approaches are interrelated by considering dynamical systems as representative of flows on a low-dimensional manifold. Building on this concept, we propose a new decomposed dynamical system model that represents complex non-stationary and nonlinear dynamics of time series data as a sparse combination of simpler, more interpretable components. Our model is trained through a dictionary learning procedure, where we leverage recent results in tracking sparse vectors over time. The decomposed nature of the dynamics is more expressive than previous switched approaches for a given number of parameters and enables modeling of overlapping and non-stationary dynamics. In both continuous-time and discrete-time instructional examples we demonstrate that our model can well approximate the original system, learn efficient representations, and capture smooth transitions between dynamical modes, focusing on intuitive low-dimensional non-stationary linear and nonlinear systems. Furthermore, we highlight our model's ability to efficiently capture and demix population dynamics generated from multiple independent subnetworks, a task that is computationally impractical for switched models. Finally, we apply our model to neural \"full brain\" recordings of C. elegans data, illustrating a diversity of dynamics that is obscured when classified into discrete states."
                    },
                    {
                        "title": "Testing for context-dependent changes in neural encoding in naturalistic experiments",
                        "abstract": "We propose a decoding-based approach to detect context effects on neural codes in longitudinal neural recording data. The approach is agnostic to how information is encoded in neural activity, and can control for a variety of possible confounding factors present in the data. We demonstrate our approach by determining whether it is possible to decode location encoding from prefrontal cortex in the mouse and, further, testing whether the encoding changes due to task engagement."
                    }
                ]
            },
            "d1b4dfc3-a733-473d-9259-240a8fa0fbc3": {
                "pk": "d1b4dfc3-a733-473d-9259-240a8fa0fbc3",
                "name": "Noga Mudrik",
                "collaborators": [
                    "Adam S. Charles",
                    "Gal Mishne",
                    "Eva Yezerets",
                    "Yenho Chen",
                    "Adam Charles",
                    "Christopher Rozell",
                    "Ryan Ly",
                    "Oliver Ruebel",
                    "Jiancheng Xie",
                    "Lou C. Kohler Voinov"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Dynamics",
                    "Graph-based Learning",
                    "Time Series Analysis"
                ],
                "publications": [
                    {
                        "title": "Multi-Lingual DALL-E Storytime",
                        "abstract": "While recent advancements in artificial intelligence (AI) language models demonstrate cutting-edge performance when working with English texts, equivalent models do not exist in other languages or do not reach the same performance level. This undesired effect of AI advancements increases the gap between access to new technology from different populations across the world. This unsought bias mainly discriminates against individuals whose English skills are less developed, e.g., non-English speakers children. Following significant advancements in AI research in recent years, OpenAI has recently presented DALL-E: a powerful tool for creating images based on English text prompts. While DALL-E is a promising tool for many applications, its decreased performance when given input in a different language, limits its audience and deepens the gap between populations. An additional limitation of the current DALL-E model is that it only allows for the creation of a few images in response to a given input prompt, rather than a series of consecutive coherent frames that tell a story or describe a process that changes over time. Here, we present an easy-to-use automatic DALL-E storytelling framework that leverages the existing DALL-E model to enable fast and coherent visualizations of non-English songs and stories, pushing the limit of the one-step-at-a-time option DALL-E currently offers. We show that our framework is able to effectively visualize stories from non-English texts and portray the changes in the plot over time. It is also able to create a narrative and maintain interpretable changes in the description across frames. Additionally, our framework offers users the ability to specify constraints on the story elements, such as a specific location or context, and to maintain a consistent style throughout the visualization."
                    },
                    {
                        "title": "SiBBlInGS: Similarity-driven Building-Block Inference using Graphs across States",
                        "abstract": "Time series data across scientific domains are often collected under distinct states (e.g., tasks), wherein latent processes (e.g., biological factors) create complex inter- and intra-state variability. A key approach to capture this complexity is to uncover fundamental interpretable units within the data, Building Blocks (BBs), which modulate their activity and adjust their structure across observations. Existing methods for identifying BBs in multi-way data often overlook inter- vs. intra-state variability, produce uninterpretable components, or do not align with properties of real-world data, such as missing samples and sessions of different duration. Here, we present a framework for Similarity-driven Building Block Inference using Graphs across States (SiBBlInGS). SiBBlInGS offers a graph-based dictionary learning approach for discovering sparse BBs along with their temporal traces, based on co-activity patterns and inter- vs. intra-state relationships. Moreover, SiBBlInGS captures per-trial temporal variability and controlled cross-state structural BB adaptations, identifies state-specific vs. state-invariant components, and accommodates variability in the number and duration of observed sessions across states. We demonstrate SiBBlInGS's ability to reveal insights into complex phenomena as well as its robustness to noise and missing samples through several synthetic and real-world examples, including web search and neural data."
                    },
                    {
                        "title": "LINOCS: Lookahead Inference of Networked Operators for Continuous Stability",
                        "abstract": "Identifying latent interactions within complex systems is key to unlocking deeper insights into their operational dynamics, including how their elements affect each other and contribute to the overall system behavior. For instance, in neuroscience, discovering neuron-to-neuron interactions is essential for understanding brain function; in ecology, recognizing the interactions among populations is key for understanding complex ecosystems. Such systems, often modeled as dynamical systems, typically exhibit noisy high-dimensional and non-stationary temporal behavior that renders their identification challenging. Existing dynamical system identification methods often yield operators that accurately capture short-term behavior but fail to predict long-term trends, suggesting an incomplete capture of the underlying process. Methods that consider extended forecasts (e.g., recurrent neural networks) lack explicit representations of element-wise interactions and require substantial training data, thereby failing to capture interpretable network operators. Here we introduce Lookahead-driven Inference of Networked Operators for Continuous Stability (LINOCS), a robust learning procedure for identifying hidden dynamical interactions in noisy time-series data. LINOCS integrates several multi-step predictions with adaptive weights during training to recover dynamical operators that can yield accurate long-term predictions. We demonstrate LINOCS' ability to recover the ground truth dynamical operators underlying synthetic time-series data for multiple dynamical systems models (including linear, piece-wise linear, time-changing linear systems' decomposition, and regularized linear time-varying systems) as well as its capability to produce meaningful operators with robust reconstructions through various real-world examples."
                    },
                    {
                        "title": "CrEIMBO: Cross Ensemble Interactions in Multi-view Brain Observations",
                        "abstract": "Modern recordings of neural activity provide diverse observations of neurons across brain areas, behavioral conditions, and subjects -- thus presenting an exciting opportunity to reveal the fundamentals of brain-wide dynamics underlying cognitive function. Current methods, however, often fail to fully harness the richness of such data as they either provide an uninterpretable representation (e.g., via \"black box\" deep networks) or over-simplify the model (e.g., assume stationary dynamics or analyze each session independently). Here, instead of regarding asynchronous recordings that lack alignment in neural identity or brain areas as a limitation, we exploit these diverse views of the same brain system to learn a unified model of brain dynamics. We assume that brain observations stem from the joint activity of a set of functional neural ensembles (groups of co-active neurons) that are similar in functionality across recordings, and propose to discover the ensemble and their non-stationary dynamical interactions in a new model we term CrEIMBO (Cross-Ensemble Interactions in Multi-view Brain Observations). CrEIMBO identifies the composition of the per-session neural ensembles through graph-driven dictionary learning and models the ensemble dynamics as a latent sparse time-varying decomposition of global sub-circuits, thereby capturing non-stationary dynamics. CrEIMBO identifies multiple co-active sub-circuits while maintaining representation interpretability due to sharing sub-circuits across sessions. CrEIMBO distinguishes session-specific from global (session-invariant) computations by exploring when distinct sub-circuits are active. We demonstrate CrEIMBO's ability to recover ground truth components in synthetic data and uncover meaningful brain dynamics, capturing cross-subject and inter- and intra-area variability, in high-density electrode recordings of humans performing a memory task."
                    },
                    {
                        "title": "Multiway Multislice PHATE: Visualizing Hidden Dynamics of RNNs through Training",
                        "abstract": "Recurrent neural networks (RNNs) are a widely used tool for sequential data analysis, however, they are still often seen as black boxes of computation. Understanding the functional principles of these networks is critical to developing ideal model architectures and optimization strategies. Previous studies typically only emphasize the network representation post-training, overlooking their evolution process throughout training. Here, we present Multiway Multislice PHATE (MM-PHATE), a novel method for visualizing the evolution of RNNs' hidden states. MM-PHATE is a graph-based embedding using structured kernels across the multiple dimensions spanned by RNNs: time, training epoch, and units. We demonstrate on various datasets that MM-PHATE uniquely preserves hidden representation community structure among units and identifies information processing and compression phases during training. The embedding allows users to look under the hood of RNNs across training and provides an intuitive and comprehensive strategy to understanding the network's internal dynamics and draw conclusions, e.g., on why and how one model outperforms another or how a specific architecture might impact an RNN's learning ability."
                    },
                    {
                        "title": "Decomposed Linear Dynamical Systems (dLDS) for learning the latent components of neural dynamics",
                        "abstract": "Learning interpretable representations of neural dynamics at a population level is a crucial first step to understanding how observed neural activity relates to perception and behavior. Models of neural dynamics often focus on either low-dimensional projections of neural activity, or on learning dynamical systems that explicitly relate to the neural state over time. We discuss how these two approaches are interrelated by considering dynamical systems as representative of flows on a low-dimensional manifold. Building on this concept, we propose a new decomposed dynamical system model that represents complex non-stationary and nonlinear dynamics of time series data as a sparse combination of simpler, more interpretable components. Our model is trained through a dictionary learning procedure, where we leverage recent results in tracking sparse vectors over time. The decomposed nature of the dynamics is more expressive than previous switched approaches for a given number of parameters and enables modeling of overlapping and non-stationary dynamics. In both continuous-time and discrete-time instructional examples we demonstrate that our model can well approximate the original system, learn efficient representations, and capture smooth transitions between dynamical modes, focusing on intuitive low-dimensional non-stationary linear and nonlinear systems. Furthermore, we highlight our model's ability to efficiently capture and demix population dynamics generated from multiple independent subnetworks, a task that is computationally impractical for switched models. Finally, we apply our model to neural \"full brain\" recordings of C. elegans data, illustrating a diversity of dynamics that is obscured when classified into discrete states."
                    }
                ]
            },
            "84ee2029-d3b6-4263-994b-473c36479e24": {
                "pk": "84ee2029-d3b6-4263-994b-473c36479e24",
                "name": "Kyle A. Johnsen",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "273075d6-726c-431e-9a80-d093504cb8dc": {
                "pk": "273075d6-726c-431e-9a80-d093504cb8dc",
                "name": "Sankaraleengam Alagapan",
                "collaborators": [
                    "Hae Won Shin",
                    "Flavio Frohlich",
                    "Hau-tieng Wu"
                ],
                "domain": [
                    "Neurophysiology",
                    "Signal Processing",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Diffusion geometry approach to efficiently remove electrical stimulation artifacts in intracranial electroencephalography",
                        "abstract": "Cortical oscillations, electrophysiological activity patterns, associated with cognitive functions and impaired in many psychiatric disorders can be observed in intracranial electroencephalography (iEEG). Direct cortical stimulation (DCS) may directly target these oscillations and may serve as therapeutic approaches to restore functional impairments. However, the presence of electrical stimulation artifacts in neurophysiological data limits the analysis of the effects of stimulation. Currently available methods suffer in performance in the presence of nonstationarity inherent in biological data. Our algorithm, Shape Adaptive Nonlocal Artifact Removal (SANAR) is based on unsupervised manifold learning. By estimating the Euclidean median of k nearest neighbors of each artifact in a nonlocal fashion, we obtain a faithful representation of the artifact which is then subtracted. This approach overcomes the challenges presented by nonstationarity. SANAR is effective in removing stimulation artifacts in the time domain while preserving the spectral content of the endogenous neurophysiological signal. We demonstrate the performance in a simulated dataset as well as in human iEEG data. Using two quantitative measures, that capture how much of information from endogenous activity is retained, we demonstrate that SANAR's performance exceeds that of one of the widely used approaches, independent component analysis, in the time domain as well as the frequency domain. This approach allows for the analysis of iEEG data, single channel or multiple channels, during DCS, a crucial step in advancing our understanding of the effects of periodic stimulation and developing new therapies."
                    }
                ]
            },
            "64e3787e-559b-493e-bd69-415ec4c8a1e1": {
                "pk": "64e3787e-559b-493e-bd69-415ec4c8a1e1",
                "name": "Adam S. Charles",
                "collaborators": [
                    "Gal Mishne",
                    "Christopher J. Rozell",
                    "Hadas Benisty",
                    "Alexander Song",
                    "Iuliia Dmitrieva",
                    "Sergey Babkin"
                ],
                "domain": [
                    "Deep Learning",
                    "Signal Processing",
                    "Neural Imaging",
                    "Computational Neuroscience"
                ],
                "publications": [
                    {
                        "title": "Interpreting Deep Learning: The Machine Learning Rorschach Test?",
                        "abstract": "Theoretical understanding of deep learning is one of the most important tasks facing the statistics and machine learning communities. While deep neural networks (DNNs) originated as engineering methods and models of biological networks in neuroscience and psychology, they have quickly become a centerpiece of the machine learning toolbox. Unfortunately, DNN adoption powered by recent successes combined with the open-source nature of the machine learning community, has outpaced our theoretical understanding. We cannot reliably identify when and why DNNs will make mistakes. In some applications like text translation these mistakes may be comical and provide for fun fodder in research talks, a single error can be very costly in tasks like medical imaging. As we utilize DNNs in increasingly sensitive applications, a better understanding of their properties is thus imperative. Recent advances in DNN theory are numerous and include many different sources of intuition, such as learning theory, sparse signal analysis, physics, chemistry, and psychology. An interesting pattern begins to emerge in the breadth of possible interpretations. The seemingly limitless approaches are mostly constrained by the lens with which the mathematical operations are viewed. Ultimately, the interpretation of DNNs appears to mimic a type of Rorschach test --- a psychological test wherein subjects interpret a series of seemingly ambiguous ink-blots. Validation for DNN theory requires a convergence of the literature. We must distinguish between universal results that are invariant to the analysis perspective and those that are specific to a particular network configuration. Simultaneously we must deal with the fact that many standard statistical tools for quantifying generalization or empirically assessing important network features are difficult to apply to DNNs."
                    },
                    {
                        "title": "Re-Weighted l_1 Dynamic Filtering for Time-Varying Sparse Signal Estimation",
                        "abstract": "Signal estimation from incomplete observations improves as more signal structure can be exploited in the inference process. Classic algorithms (e.g., Kalman filtering) have exploited strong dynamic structure for time-varying signals while modern work has often focused on exploiting low-dimensional signal structure (e.g., sparsity in a basis) for static signals. Few algorithms attempt to merge both static and dynamic structure to improve estimation for time-varying sparse signals (e.g., video). In this work we present a re-weighted l_1 dynamic filtering scheme for causal signal estimation that utilizes both sparsity assumptions and dynamic structure. Our algorithm leverages work on hierarchical Laplacian scale mixture models to create a dynamic probabilistic model. The resulting algorithm incorporates both dynamic and sparsity priors in the estimation procedure in a robust and efficient algorithm. We demonstrate the results in simulation using both synthetic and natural data."
                    },
                    {
                        "title": "Learning spatially-correlated temporal dictionaries for calcium imaging",
                        "abstract": "Calcium imaging has become a fundamental neural imaging technique, aiming to recover the individual activity of hundreds of neurons in a cortical region. Current methods (mostly matrix factorization) are aimed at detecting neurons in the field-of-view and then inferring the corresponding time-traces. In this paper, we reverse the modeling and instead aim to minimize the spatial inference, while focusing on finding the set of temporal traces present in the data. We reframe the problem in a dictionary learning setting, where the dictionary contains the time-traces and the sparse coefficient are spatial maps. We adapt dictionary learning to calcium imaging by introducing constraints on the norms and correlations of the time-traces, and incorporating a hierarchical spatial filtering model that correlates the time-trace usage over the field-of-view. We demonstrate on synthetic and real data that our solution has advantages regarding initialization, implicitly inferring number of neurons and simultaneously detecting different neuronal types."
                    },
                    {
                        "title": "Data Processing of Functional Optical Microscopy for Neuroscience",
                        "abstract": "Functional optical imaging in neuroscience is rapidly growing with the development of new optical systems and fluorescence indicators. To realize the potential of these massive spatiotemporal datasets for relating neuronal activity to behavior and stimuli and uncovering local circuits in the brain, accurate automated processing is increasingly essential. In this review, we cover recent computational developments in the full data processing pipeline of functional optical microscopy for neuroscience data and discuss ongoing and emerging challenges."
                    },
                    {
                        "title": "realSEUDO for real-time calcium imaging analysis",
                        "abstract": "Closed-loop neuroscience experimentation, where recorded neural activity is used to modify the experiment on-the-fly, is critical for deducing causal connections and optimizing experimental time. A critical step in creating a closed-loop experiment is real-time inference of neural activity from streaming recordings. One challenging modality for real-time processing is multi-photon calcium imaging (CI). CI enables the recording of activity in large populations of neurons however, often requires batch processing of the video data to extract single-neuron activity from the fluorescence videos. We use the recently proposed robust time-trace estimator-Sparse Emulation of Unused Dictionary Objects (SEUDO) algorithm-as a basis for a new on-line processing algorithm that simultaneously identifies neurons in the fluorescence video and infers their time traces in a way that is robust to as-yet unidentified neurons. To achieve real-time SEUDO (realSEUDO), we optimize the core estimator via both algorithmic improvements and an fast C-based implementation, and create a new cell finding loop to enable realSEUDO to also identify new cells. We demonstrate comparable performance to offline algorithms (e.g., CNMF), and improved performance over the current on-line approach (OnACID) at speeds of 120 Hz on average."
                    }
                ]
            },
            "d061aa21-5de5-47c4-ab55-fde0fc735c9b": {
                "pk": "d061aa21-5de5-47c4-ab55-fde0fc735c9b",
                "name": "Christopher J. Rozell",
                "collaborators": [
                    "Han Lun Yap",
                    "Adam S. Charles",
                    "Michael B. Wakin",
                    "Justin Romberg",
                    "John Lee",
                    "Armin Eftekhari",
                    "Kion Fallah",
                    "Nicholas P. Bertrand",
                    "Aur\u00e8le Balavoine",
                    "Gregory H. Canal"
                ],
                "domain": [
                    "Signal Processing",
                    "Compressed Sensing",
                    "Machine Learning",
                    "Optimal Transport"
                ],
                "publications": [
                    {
                        "title": "Short Term Memory Capacity in Networks via the Restricted Isometry Property",
                        "abstract": "Cortical networks are hypothesized to rely on transient network activity to support short term memory (STM). In this paper we study the capacity of randomly connected recurrent linear networks for performing STM when the input signals are approximately sparse in some basis. We leverage results from compressed sensing to provide rigorous non asymptotic recovery guarantees, quantifying the impact of the input sparsity level, the input sparsity basis, and the network characteristics on the system capacity. Our analysis demonstrates that network memory capacities can scale superlinearly with the number of nodes, and in some situations can achieve STM capacities that are much larger than the network size. We provide perfect recovery guarantees for finite sequences and recovery bounds for infinite sequences. The latter analysis predicts that network STM systems may have an optimal recovery length that balances errors due to omission and recall mistakes. Furthermore, we show that the conditions yielding optimal STM capacity can be embodied in several network topologies, including networks with sparse or dense connectivities."
                    },
                    {
                        "title": "Re-Weighted l_1 Dynamic Filtering for Time-Varying Sparse Signal Estimation",
                        "abstract": "Signal estimation from incomplete observations improves as more signal structure can be exploited in the inference process. Classic algorithms (e.g., Kalman filtering) have exploited strong dynamic structure for time-varying signals while modern work has often focused on exploiting low-dimensional signal structure (e.g., sparsity in a basis) for static signals. Few algorithms attempt to merge both static and dynamic structure to improve estimation for time-varying sparse signals (e.g., video). In this work we present a re-weighted l_1 dynamic filtering scheme for causal signal estimation that utilizes both sparsity assumptions and dynamic structure. Our algorithm leverages work on hierarchical Laplacian scale mixture models to create a dynamic probabilistic model. The resulting algorithm incorporates both dynamic and sparsity priors in the estimation procedure in a robust and efficient algorithm. We demonstrate the results in simulation using both synthetic and natural data."
                    },
                    {
                        "title": "Variational Sparse Coding with Learned Thresholding",
                        "abstract": "Sparse coding strategies have been lauded for their parsimonious representations of data that leverage low dimensional structure. However, inference of these codes typically relies on an optimization procedure with poor computational scaling in high-dimensional problems. For example, sparse inference in the representations learned in the high-dimensional intermediary layers of deep neural networks (DNNs) requires an iterative minimization to be performed at each training step. As such, recent, quick methods in variational inference have been proposed to infer sparse codes by learning a distribution over the codes with a DNN. In this work, we propose a new approach to variational sparse coding that allows us to learn sparse distributions by thresholding samples, avoiding the use of problematic relaxations. We first evaluate and analyze our method by training a linear generator, showing that it has superior performance, statistical efficiency, and gradient estimation compared to other sparse distributions. We then compare to a standard variational autoencoder using a DNN generator on the Fashion MNIST and CelebA datasets"
                    },
                    {
                        "title": "Discrete and Continuous-time Soft-Thresholding with Dynamic Inputs",
                        "abstract": "There exist many well-established techniques to recover sparse signals from compressed measurements with known performance guarantees in the static case. However, only a few methods have been proposed to tackle the recovery of time-varying signals, and even fewer benefit from a theoretical analysis. In this paper, we study the capacity of the Iterative Soft-Thresholding Algorithm (ISTA) and its continuous-time analogue the Locally Competitive Algorithm (LCA) to perform this tracking in real time. ISTA is a well-known digital solver for static sparse recovery, whose iteration is a first-order discretization of the LCA differential equation. Our analysis shows that the outputs of both algorithms can track a time-varying signal while compressed measurements are streaming, even when no convergence criterion is imposed at each time step. The L2-distance between the target signal and the outputs of both discrete- and continuous-time solvers is shown to decay to a bound that is essentially optimal. Our analyses is supported by simulations on both synthetic and real data."
                    },
                    {
                        "title": "Parallel Unbalanced Optimal Transport Regularization for Large Scale Imaging Problems",
                        "abstract": "The modeling of phenomenological structure is a crucial aspect in inverse imaging problems. One emerging modeling tool in computational imaging is the optimal transport framework. Its ability to model geometric displacements across an image's support gives it attractive qualities similar to those of optical flow methods which are effective at capturing visual motion, but are restricted to operate in significantly smaller state-spaces. Despite this advantage, two major drawbacks make it unsuitable for general deployment: (i) it suffers from exorbitant computational costs due to a quadratic optimization-variable complexity, and (ii) it has a mass-balancing assumption that limits applications with natural images. We tackle these issues simultaneously by proposing a novel formulation for an unbalanced optimal transport regularizer that has linear optimization-variable complexity. In addition, we present a general parallelizable proximal method for this regularizer, and demonstrate superior empirical performance on novel dynamical tracking applications in synthetic and real video."
                    },
                    {
                        "title": "Stable Takens' Embeddings for Linear Dynamical Systems",
                        "abstract": "Takens' Embedding Theorem remarkably established that concatenating M previous outputs of a dynamical system into a vector (called a delay coordinate map) can be a one-to-one mapping of a low-dimensional attractor from the system state space. However, Takens' theorem is fragile in the sense that even small imperfections can induce arbitrarily large errors in this attractor representation. We extend Takens' result to establish deterministic, explicit and non-asymptotic sufficient conditions for a delay coordinate map to form a stable embedding in the restricted case of linear dynamical systems and observation functions. Our work is inspired by the field of Compressive Sensing (CS), where results guarantee that low-dimensional signal families can be robustly reconstructed if they are stably embedded by a measurement operator. However, in contrast to typical CS results, i) our sufficient conditions are independent of the size of the ambient state space, and ii) some system and measurement pairs have fundamental limits on the conditioning of the embedding (i.e., how close it is to an isometry), meaning that further measurements beyond some point add no further significant value. We use several simple simulations to explore the conditions of the main results, including the tightness of the bounds and the convergence speed of the stable embedding. We also present an example task of estimating the attractor dimension from time-series data to highlight the value of stable embeddings over traditional Takens' embeddings."
                    },
                    {
                        "title": "Convergence and Rate Analysis of Neural Networks for Sparse Approximation",
                        "abstract": "We present an analysis of the Locally Competitive Algorithm (LCA), a Hopfield-style neural network that efficiently solves sparse approximation problems (e.g., approximating a vector from a dictionary using just a few non-zero coefficients). This class of problems plays a significant role in both theories of neural coding and applications in signal processing. However, the LCA lacks analysis of its convergence properties and previous results on neural networks for nonsmooth optimization do not apply to the specifics of the LCA architecture. We show that the LCA has desirable convergence properties, such as stability and global convergence to the optimum of the objective function when it is unique. Under some mild conditions, the support of the solution is also proven to be reached in finite time. Furthermore, some restrictions on the problem specifics allow us to characterize the convergence rate of the system by showing that the LCA converges exponentially fast with an analytically bounded convergence rate. We support our analysis with several illustrative simulations."
                    },
                    {
                        "title": "Convergence Speed of a Dynamical System for Sparse Recovery",
                        "abstract": "This paper studies the convergence rate of a continuous-time dynamical system for L1-minimization, known as the Locally Competitive Algorithm (LCA). Solving L1-minimization} problems efficiently and rapidly is of great interest to the signal processing community, as these programs have been shown to recover sparse solutions to underdetermined systems of linear equations and come with strong performance guarantees. The LCA under study differs from the typical L1 solver in that it operates in continuous time: instead of being specified by discrete iterations, it evolves according to a system of nonlinear ordinary differential equations. The LCA is constructed from simple components, giving it the potential to be implemented as a large-scale analog circuit.   The goal of this paper is to give guarantees on the convergence time of the LCA system. To do so, we analyze how the LCA evolves as it is recovering a sparse signal from underdetermined measurements. We show that under appropriate conditions on the measurement matrix and the problem parameters, the path the LCA follows can be described as a sequence of linear differential equations, each with a small number of active variables. This allows us to relate the convergence time of the system to the restricted isometry constant of the matrix. Interesting parallels to sparse-recovery digital solvers emerge from this study. Our analysis covers both the noisy and noiseless settings and is supported by simulation results."
                    },
                    {
                        "title": "Stabilizing Embedology: Geometry-Preserving Delay-Coordinate Maps",
                        "abstract": "Delay-coordinate mapping is an effective and widely used technique for reconstructing and analyzing the dynamics of a nonlinear system based on time-series outputs. The efficacy of delay-coordinate mapping has long been supported by Takens' embedding theorem, which guarantees that delay-coordinate maps use the time-series output to provide a reconstruction of the hidden state space that is a one-to-one embedding of the system's attractor. While this topological guarantee ensures that distinct points in the reconstruction correspond to distinct points in the original state space, it does not characterize the quality of this embedding or illuminate how the specific parameters affect the reconstruction. In this paper, we extend Takens' result by establishing conditions under which delay-coordinate mapping is guaranteed to provide a stable embedding of a system's attractor. Beyond only preserving the attractor topology, a stable embedding preserves the attractor geometry by ensuring that distances between points in the state space are approximately preserved. In particular, we find that delay-coordinate mapping stably embeds an attractor of a dynamical system if the stable rank of the system is large enough to be proportional to the dimension of the attractor. The stable rank reflects the relation between the sampling interval and the number of delays in delay-coordinate mapping. Our theoretical findings give guidance to choosing system parameters, echoing the trade-off between irrelevancy and redundancy that has been heuristically investigated in the literature. Our initial result is stated for attractors that are smooth submanifolds of Euclidean space, with extensions provided for the case of strange attractors."
                    },
                    {
                        "title": "Stable Manifold Embeddings with Structured Random Matrices",
                        "abstract": "The fields of compressed sensing (CS) and matrix completion have shown that high-dimensional signals with sparse or low-rank structure can be effectively projected into a low-dimensional space (for efficient acquisition or processing) when the projection operator achieves a stable embedding of the data by satisfying the Restricted Isometry Property (RIP). It has also been shown that such stable embeddings can be achieved for general Riemannian submanifolds when random orthoprojectors are used for dimensionality reduction. Due to computational costs and system constraints, the CS community has recently explored the RIP for structured random matrices (e.g., random convolutions, localized measurements, deterministic constructions). The main contribution of this paper is to show that any matrix satisfying the RIP (i.e., providing a stable embedding for sparse signals) can be used to construct a stable embedding for manifold-modeled signals by randomizing the column signs and paying reasonable additional factors in the number of measurements. We demonstrate this result with several new constructions for stable manifold embeddings using structured matrices. This result allows advances in efficient projection schemes for sparse signals to be immediately applied to manifold signal models."
                    },
                    {
                        "title": "The Restricted Isometry Property for Random Block Diagonal Matrices",
                        "abstract": "In Compressive Sensing, the Restricted Isometry Property (RIP) ensures that robust recovery of sparse vectors is possible from noisy, undersampled measurements via computationally tractable algorithms. It is by now well-known that Gaussian (or, more generally, sub-Gaussian) random matrices satisfy the RIP under certain conditions on the number of measurements. Their use can be limited in practice, however, due to storage limitations, computational considerations, or the mismatch of such matrices with certain measurement architectures. These issues have recently motivated considerable effort towards studying the RIP for structured random matrices. In this paper, we study the RIP for block diagonal measurement matrices where each block on the main diagonal is itself a sub-Gaussian random matrix. Our main result states that such matrices can indeed satisfy the RIP but that the requisite number of measurements depends on certain properties of the basis in which the signals are sparse. In the best case, these matrices perform nearly as well as dense Gaussian random matrices, despite having many fewer nonzero entries."
                    },
                    {
                        "title": "A First Analysis of the Stability of Takens' Embedding",
                        "abstract": "Takens' Embedding Theorem asserts that when the states of a hidden dynamical system are confined to a low-dimensional attractor, complete information about the states can be preserved in the observed time-series output through the delay coordinate map. However, the conditions for the theorem to hold ignore the effects of noise and time-series analysis in practice requires a careful empirical determination of the sampling time and number of delays resulting in a number of delay coordinates larger than the minimum prescribed by Takens' theorem. In this paper, we use tools and ideas in Compressed Sensing to provide a first theoretical justification for the choice of the number of delays in noisy conditions. In particular, we show that under certain conditions on the dynamical system, measurement function, number of delays and sampling time, the delay-coordinate map can be a stable embedding of the dynamical system's attractor."
                    },
                    {
                        "title": "Analog Sparse Approximation with Applications to Compressed Sensing",
                        "abstract": "Recent research has shown that performance in signal processing tasks can often be significantly improved by using signal models based on sparse representations, where a signal is approximated using a small number of elements from a fixed dictionary. Unfortunately, inference in this model involves solving non-smooth optimization problems that are computationally expensive. While significant efforts have focused on developing digital algorithms specifically for this problem, these algorithms are inappropriate for many applications because of the time and power requirements necessary to solve large optimization problems. Based on recent work in computational neuroscience, we explore the potential advantages of continuous time dynamical systems for solving sparse approximation problems if they were implemented in analog VLSI. Specifically, in the simulated task of recovering synthetic and MRI data acquired via compressive sensing techniques, we show that these systems can potentially perform recovery at time scales of 10-20{\\mu}s, supporting datarates of 50-100 kHz (orders of magnitude faster that digital algorithms). Furthermore, we show analytically that a wide range of sparse approximation problems can be solved in the same basic architecture, including approximate $\\ell^p$ norms, modified $\\ell^1$ norms, re-weighted $\\ell^1$ and $\\ell^2$, the block $\\ell^1$ norm and classic Tikhonov regularization."
                    },
                    {
                        "title": "PrefGen: Preference Guided Image Generation with Relative Attributes",
                        "abstract": "Deep generative models have the capacity to render high fidelity images of content like human faces. Recently, there has been substantial progress in conditionally generating images with specific quantitative attributes, like the emotion conveyed by one's face. These methods typically require a user to explicitly quantify the desired intensity of a visual attribute. A limitation of this method is that many attributes, like how \"angry\" a human face looks, are difficult for a user to precisely quantify. However, a user would be able to reliably say which of two faces seems \"angrier\". Following this premise, we develop the $\\textit{PrefGen}$ system, which allows users to control the relative attributes of generated images by presenting them with simple paired comparison queries of the form \"do you prefer image $a$ or image $b$?\" Using information from a sequence of query responses, we can estimate user preferences over a set of image attributes and perform preference-guided image editing and generation. Furthermore, to make preference localization feasible and efficient, we apply an active query selection strategy. We demonstrate the success of this approach using a StyleGAN2 generator on the task of human face editing. Additionally, we demonstrate how our approach can be combined with CLIP, allowing a user to edit the relative intensity of attributes specified by text prompts. Code at https://github.com/helblazer811/PrefGen."
                    },
                    {
                        "title": "Efficient Tracking of Sparse Signals via an Earth Mover's Distance Dynamics Regularizer",
                        "abstract": "Tracking algorithms such as the Kalman filter aim to improve inference performance by leveraging the temporal dynamics in streaming observations. However, the tracking regularizers are often based on the $\\ell_p$-norm which cannot account for important geometrical relationships between neighboring signal elements. We propose a practical approach to using the earth mover's distance (EMD) via the earth mover's distance dynamic filtering (EMD-DF) algorithm for causally tracking time-varying sparse signals when there is a natural geometry to the coefficient space that should be respected (e.g., meaningful ordering). Specifically, this paper presents a new Beckmann formulation that dramatically reduces computational complexity, as well as an evaluation of the performance and complexity of the proposed approach in imaging and frequency tracking applications with real and simulated neurophysiology data."
                    },
                    {
                        "title": "Active embedding search via noisy paired comparisons",
                        "abstract": "Suppose that we wish to estimate a user's preference vector $w$ from paired comparisons of the form \"does user $w$ prefer item $p$ or item $q$?,\" where both the user and items are embedded in a low-dimensional Euclidean space with distances that reflect user and item similarities. Such observations arise in numerous settings, including psychometrics and psychology experiments, search tasks, advertising, and recommender systems. In such tasks, queries can be extremely costly and subject to varying levels of response noise; thus, we aim to actively choose pairs that are most informative given the results of previous comparisons. We provide new theoretical insights into the benefits and challenges of greedy information maximization in this setting, and develop two novel strategies that maximize lower bounds on information gain and are simpler to analyze and compute respectively. We use simulated responses from a real-world dataset to validate our strategies through their similar performance to greedy information maximization, and their superior preference estimation over state-of-the-art selection methods as well as random queries."
                    },
                    {
                        "title": "Sparse Bayesian Learning with Dynamic Filtering for Inference of Time-Varying Sparse Signals",
                        "abstract": "Many signal processing applications require estimation of time-varying sparse signals, potentially with the knowledge of an imperfect dynamics model. In this paper, we propose an algorithm for dynamic filtering of time-varying sparse signals based on the sparse Bayesian learning (SBL) framework. The key idea underlying the algorithm, termed SBL-DF, is the incorporation of a signal prediction generated from a dynamics model and estimates of previous time steps into the hyperpriors of the SBL probability model. The proposed algorithm is robust to imperfect dynamics models (due to the propagation of dynamics information through higher-order statistics), robust to certain undesirable dictionary properties such as coherence (due to properties of the SBL framework), allows the use of arbitrary dynamics models, and requires the tuning of fewer parameters than many other dynamic filtering algorithms do. We also extend the fast marginal likelihood SBL inference procedure to the informative hyperprior setting to create a particularly efficient version of the SBL-DF algorithm. Numerical simulations show that SBL-DF converges much faster and to more accurate solutions than standard SBL and other dynamical filtering algorithms. In particular, we show that SBL-DF outperforms state of the art algorithms when the dictionary contains challenging structure such as coherence and column scaling, as is found in many practical applications."
                    },
                    {
                        "title": "Hierarchical Optimal Transport for Multimodal Distribution Alignment",
                        "abstract": "In many machine learning applications, it is necessary to meaningfully aggregate, through alignment, different but related datasets. Optimal transport (OT)-based approaches pose alignment as a divergence minimization problem: the aim is to transform a source dataset to match a target dataset using the Wasserstein distance as a divergence measure. We introduce a hierarchical formulation of OT which leverages clustered structure in data to improve alignment in noisy, ambiguous, or multimodal settings. To solve this numerically, we propose a distributed ADMM algorithm that also exploits the Sinkhorn distance, thus it has an efficient computational complexity that scales quadratically with the size of the largest cluster. When the transformation between two datasets is unitary, we provide performance guarantees that describe when and how well aligned cluster correspondences can be recovered with our formulation, as well as provide worst-case dataset geometry for such a strategy. We apply this method to synthetic datasets that model data as mixtures of low-rank Gaussians and study the impact that different geometric properties of the data have on alignment. Next, we applied our approach to a neural decoding application where the goal is to predict movement directions and instantaneous velocities from populations of neurons in the macaque primary motor cortex. Our results demonstrate that when clustered structure exists in datasets, and is consistent across trials or time points, a hierarchical alignment strategy that leverages such structure can provide significant improvements in cross-domain alignment."
                    },
                    {
                        "title": "Variational Autoencoder with Learned Latent Structure",
                        "abstract": "The manifold hypothesis states that high-dimensional data can be modeled as lying on or near a low-dimensional, nonlinear manifold. Variational Autoencoders (VAEs) approximate this manifold by learning mappings from low-dimensional latent vectors to high-dimensional data while encouraging a global structure in the latent space through the use of a specified prior distribution. When this prior does not match the structure of the true data manifold, it can lead to a less accurate model of the data. To resolve this mismatch, we introduce the Variational Autoencoder with Learned Latent Structure (VAELLS) which incorporates a learnable manifold model into the latent space of a VAE. This enables us to learn the nonlinear manifold structure from the data and use that structure to define a prior in the latent space. The integration of a latent manifold model not only ensures that our prior is well-matched to the data, but also allows us to define generative transformation paths in the latent space and describe class manifolds with transformations stemming from examples of each class. We validate our model on examples with known latent structure and also demonstrate its capabilities on a real-world dataset."
                    }
                ]
            }
        }
    },
    "2306.14268": {
        "paper_data": {
            "title": "Adaptive Window Pruning for Efficient Local Motion Deblurring",
            "url": "http://arxiv.org/abs/2306.14268v2",
            "arxiv_id": "2306.14268",
            "authors": [
                "Haoying Li",
                "Jixin Zhao",
                "Shangchen Zhou",
                "Huajun Feng",
                "Chongyi Li",
                "Chen Change Loy"
            ],
            "abstract": "Local motion blur commonly occurs in real-world photography due to the mixing between moving objects and stationary backgrounds during exposure. Existing image deblurring methods predominantly focus on global deblurring, inadvertently affecting the sharpness of backgrounds in locally blurred images and wasting unnecessary computation on sharp pixels, especially for high-resolution images. This paper aims to adaptively and efficiently restore high-resolution locally blurred images. We propose a local motion deblurring vision Transformer (LMD-ViT) built on adaptive window pruning Transformer blocks (AdaWPT). To focus deblurring on local regions and reduce computation, AdaWPT prunes unnecessary windows, only allowing the active windows to be involved in the deblurring processes. The pruning operation relies on the blurriness confidence predicted by a confidence predictor that is trained end-to-end using a reconstruction loss with Gumbel-Softmax re-parameterization and a pruning loss guided by annotated blur masks. Our method removes local motion blur effectively without distorting sharp regions, demonstrated by its exceptional perceptual and quantitative improvements compared to state-of-the-art methods. In addition, our approach substantially reduces FLOPs by 66% and achieves more than a twofold increase in inference speed compared to Transformer-based deblurring methods. We will make our code and annotated blur masks publicly available.",
            "introduction": " Introduction Contrary to global motion blur, which typically affects an entire image [ 31], local motion blur is confined to specific regions within an image. Such local blur is generally the result of object movements captured by stationary cameras [ 10,19]. Applying global deblurring background of our Drawing inspiration from preceding work, in this paper, we propose a U-shaped local motion deblur- ring vision Transformer (LMD-ViT) with adaptive window pruning Transformer blocks (AdaWPT) as its core component. Our proposed LMD-ViT is the first to apply sparse vision Transformer to the local motion deblurring task. The core block, AdaWPT, focuses on locally blurred regions rather than global regions, which is made possible by removing windows unrelated to blurred areas, surpassing prior state-of-the-art Related work Single image deep motion deblurring. The task of deep motion deblurring for single images origi- nated from global deblurring [ 31,18,20,29,30,32]. Pioneering deep global motion deblurring works utilize CNNs as basic layers and achieve promising improvements in image quality. Among them, DeepDeblur [ 16], a multi-scale convolutional neural network, performs residual blocks to increase convergence speed. DeblurGAN [ 8] and DeblurGAN-v2 [ 9] introduce GANs and a perceptual loss to improve subjective quality. HINet [ 2] applies Instance Normalization to boost performance. Recently, a CNN-based local motion deblurring method, LBAG [ 10], bridges the gap between global and local motion deblurring by inserting gate modules at the end of MIMO-UNet architecture [ 3]. It predicts differentiable blur masks to reduce sharp backgrounds from modifications and guide the network to deblur locally. Although the performance is significantly improved, CNN-based Appendix E. Results of user study for real-world local motion deblurring. The horizontal axis indicates the percentage of favoring each method. 4 Experiments and analyses 4.1 Experimental settings We train LMD-ViT using AdamW optimizer [ 7] with the momentum terms of (0.9, 0.999), a batch size of 12, and an initial learning rate of 2\u00d710\u22124that is updated every 2k steps by a cosine annealing schedule [ 14]. We set the window size of AdaWPT to 8\u00d78, and the initial embedded dim to 32 which is doubled after passing each down-sampling layer. LMD-ViT is trained on the GoPro dataset [16] and ReLoBlur dataset [ 10] together, in order to enable both local and global motion deblurring. The sampling ratio of the GoPro training data [ 16] and the ReLoBlur training data [ 10] is set close to 1:1. For a fair comparison, we train the baseline results of user study for real-world local motion deblurring. We compare our proposed LMD-ViT with the top 3 Methods Feature channels \u2191PSNR\u2191SSIM\u2191PSNR w\u2191SSIM wTime Params FLOPs 1 LBAG [10] 32-64-128 34.92 0.9318 29.30 0.8946 0.51s 16.11M 7.852T 2LBAG-Large [10] 60-120-240 34.93 0.9266 29.63 0.8871 1.13s 56.58M 17.659T 3LMD-ViT-Small 16-32-64-128-256 34.98 0.9259 29.89 0.8907 0.23s 21.59M 0.311T 4 LMD-ViT 32-64-128-256-512 35.42 0.9285 30.25 0.8938 0.56s 54.50M 1.485T D.2 More visual experiments demonstrate that our method effectively eliminates local motion blur while ensuring minimal deformation of sharp regions, resulting in a significant improvement in image quality and inference speed. Due to limited page space, we discuss the limitations and broader impacts in Conclusion In this paper, we presented an adaptive and efficient approach, LMD-ViT, the first sparse vision Transformer for restoring high-resolution images affected by local motion blurs. LMD-ViT is built upon our novel adaptive window pruning Transformer blocks",
            "references": [
                {
                    "title": "ShadowFormer: Global Context Helps Image Shadow Removal",
                    "abstract": "Recent deep learning methods have achieved promising results in image shadow removal. However, most of the existing approaches focus on working locally within shadow and non-shadow regions, resulting in severe artifacts around the shadow boundaries as well as inconsistent illumination between shadow and non-shadow regions. It is still challenging for the deep shadow removal model to exploit the global contextual correlation between shadow and non-shadow regions. In this work, we first propose a Retinex-based shadow model, from which we derive a novel transformer-based network, dubbed ShandowFormer, to exploit non-shadow regions to help shadow region restoration. A multi-scale channel attention framework is employed to hierarchically capture the global information. Based on that, we propose a Shadow-Interaction Module (SIM) with Shadow-Interaction Attention (SIA) in the bottleneck stage to effectively model the context correlation between shadow and non-shadow regions. We conduct extensive experiments on three popular public datasets, including ISTD, ISTD+, and SRD, \nto evaluate the proposed method. Our method achieves state-of-the-art performance by using up to 150X fewer model parameters."
                },
                {
                    "title": "Token Merging: Your ViT But Faster",
                    "abstract": "We introduce Token Merging (ToMe), a simple method to increase the throughput of existing ViT models without needing to train. ToMe gradually combines similar tokens in a transformer using a general and light-weight matching algorithm that is as fast as pruning while being more accurate. Off-the-shelf, ToMe can 2x the throughput of state-of-the-art ViT-L @ 512 and ViT-H @ 518 models on images and 2.2x the throughput of ViT-L on video with only a 0.2-0.3% accuracy drop in each case. ToMe can also easily be applied during training, improving in practice training speed up to 2x for MAE fine-tuning on video. Training with ToMe further minimizes accuracy drop, leading to 2x the throughput of ViT-B on audio for only a 0.4% mAP drop. Qualitatively, we find that ToMe merges object parts into one token, even over multiple frames of video. Overall, ToMe's accuracy and speed are competitive with state-of-the-art on images, video, and audio."
                },
                {
                    "title": "Real-world Deep Local Motion Deblurring",
                    "abstract": "Most existing deblurring methods focus on removing global blur caused by camera shake, while they cannot well handle local blur caused by object movements. To fill the vacancy of local deblurring in real scenes, we establish the first real local motion blur dataset (ReLoBlur), which is captured by a synchronized beam-splitting photographing system and corrected by a post-progressing pipeline. Based on ReLoBlur, we propose a Local Blur-Aware Gated network (LBAG) and several local blur-aware techniques to bridge the gap between global and local deblurring: 1) a blur detection approach based on background subtraction to localize blurred regions; 2) a gate mechanism to guide our network to focus on blurred regions; and 3) a blur-aware patch cropping strategy to address data imbalance problem. Extensive experiments prove the reliability of ReLoBlur dataset, and demonstrate that LBAG achieves better performance than state-of-the-art global deblurring methods and our proposed local blur-aware techniques are effective."
                },
                {
                    "title": "Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations",
                    "abstract": "Vision Transformers (ViTs) take all the image patches as tokens and construct multi-head self-attention (MHSA) among them. Complete leverage of these image tokens brings redundant computations since not all the tokens are attentive in MHSA. Examples include that tokens containing semantically meaningless or distractive image backgrounds do not positively contribute to the ViT predictions. In this work, we propose to reorganize image tokens during the feed-forward process of ViT models, which is integrated into ViT during training. For each forward inference, we identify the attentive image tokens between MHSA and FFN (i.e., feed-forward network) modules, which is guided by the corresponding class token attention. Then, we reorganize image tokens by preserving attentive image tokens and fusing inattentive ones to expedite subsequent MHSA and FFN computations. To this end, our method EViT improves ViTs from two perspectives. First, under the same amount of input image tokens, our method reduces MHSA and FFN computation for efficient inference. For instance, the inference speed of DeiT-S is increased by 50% while its recognition accuracy is decreased by only 0.3% for ImageNet classification. Second, by maintaining the same computational cost, our method empowers ViTs to take more image tokens as input for recognition accuracy improvement, where the image tokens are from higher resolution images. An example is that we improve the recognition accuracy of DeiT-S by 1% for ImageNet classification at the same computational cost of a vanilla DeiT-S. Meanwhile, our method does not introduce more parameters to ViTs. Experiments on the standard benchmarks show the effectiveness of our method. The code is available at https://github.com/youweiliang/evit"
                },
                {
                    "title": "A-ViT: Adaptive Tokens for Efficient Vision Transformer",
                    "abstract": "We introduce A - ViT, a method that adaptively adjusts the inference cost of vision transformer (ViT) for images of different complexity. A - ViT achieves this by automatically reducing the number of tokens in vision transformers that are processed in the network as inference proceeds. We refor-mulate Adaptive Computation Time (ACT [17]) for this task, extending halting to discard redundant spatial tokens. The appealing architectural properties of vision transformers enables our adaptive token reduction mechanism to speed up inference without modifying the network architecture or inference hardware. We demonstrate that A - ViT requires no extra parameters or sub-network for halting, as we base the learning of adaptive halting on the original network parameters. We further introduce distributional prior regularization that stabilizes training compared to prior ACT approaches. On the image classification task (ImageNet1K), we show that our proposed A - ViT yields high efficacy in filtering informative spatial features and cutting down on the overall compute. The proposed method improves the throughput of DeiT-Tiny by 62% and DeiT-Small by 38% with only 0.3% accuracy drop, outperforming prior art by a large margin."
                },
                {
                    "title": "AdaViT: Adaptive Vision Transformers for Efficient Image Recognition",
                    "abstract": "Built on top of self-attention mechanisms, vision transformers have demonstrated remarkable performance on a variety of tasks recently. While achieving excellent performance, they still require relatively intensive computational cost that scales up drastically as the numbers of patches, self-attention heads and transformer blocks increase. In this paper, we argue that due to the large variations among images, their need for modeling long-range dependencies between patches differ. To this end, we introduce AdaViT, an adaptive computation framework that learns to derive usage policies on which patches, self-attention heads and transformer blocks to use throughout the backbone on a per-input basis, aiming to improve inference efficiency of vision transformers with a minimal drop of accuracy for image recognition. Optimized jointly with a transformer backbone in an end-to-end manner, a light-weight decision network is attached to the backbone to produce decisions on-the-fly. Extensive experiments on ImageNet demonstrate that our method obtains more than 2 \u00d7 improvement on efficiency compared to state-of-the-art vision transformers with only 0.8% drop of accuracy, achieving good efficiency/accuracy trade-offs conditioned on different computational budgets. We further conduct quantitative and qualitative analysis on learned usage polices and provide more insights on the redundancy in vision transformers. Code is available at ht tps: / / gi thub. com/MengLcool/AdaVi T."
                },
                {
                    "title": "Restormer: Efficient Transformer for High-Resolution Image Restoration",
                    "abstract": "Since convolutional neural networks (CNNs) perform well at learning generalizable image priors from large-scale data, these models have been extensively applied to image restoration and related tasks. Recently, another class of neural architectures, Transformers, have shown significant performance gains on natural language and high-level vision tasks. While the Transformer model mitigates the shortcomings of CNNs (i.e., limited receptive field and inadaptability to input content), its computational complexity grows quadratically with the spatial resolution, therefore making it infeasible to apply to most image restoration tasks involving high-resolution images. In this work, we propose an efficient Transformer model by making several key designs in the building blocks (multi-head attention and feed-forward network) such that it can capture long-range pixel interactions, while still remaining applicable to large images. Our model, named Restoration Transformer (Restormer), achieves state-of-the-art results on several image restoration tasks, including image deraining, single-image motion deblurring, defocus deblurring (single-image and dual-pixel data), and image denoising (Gaussian grayscale/color denoising, and real image denoising). The source code and pre-trained models are available at https://github.com/swz30/Restormer."
                },
                {
                    "title": "SwinIR: Image Restoration Using Swin Transformer",
                    "abstract": "Image restoration is a long-standing low-level vision problem that aims to restore high-quality images from low-quality images (e.g., downscaled, noisy and compressed images). While state-of-the-art image restoration methods are based on convolutional neural networks, few attempts have been made with Transformers which show impressive performance on high-level vision tasks. In this paper, we propose a strong baseline model SwinIR for image restoration based on the Swin Transformer. SwinIR consists of three parts: shallow feature extraction, deep feature extraction and high-quality image reconstruction. In particular, the deep feature extraction module is composed of several residual Swin Transformer blocks (RSTB), each of which has several Swin Transformer layers together with a residual connection. We conduct experiments on three representative tasks: image super-resolution (including classical, lightweight and real-world image super-resolution), image denoising (including grayscale and color image denoising) and JPEG compression artifact reduction. Experimental results demonstrate that SwinIR outperforms state-of-the-art methods on different tasks by up to 0.14\u223c0.45dB, while the total number of parameters can be reduced by up to 67%."
                },
                {
                    "title": "Rethinking Coarse-to-Fine Approach in Single Image Deblurring",
                    "abstract": "Coarse-to-fine strategies have been extensively used for the architecture design of single image deblurring networks. Conventional methods typically stack sub-networks with multi-scale input images and gradually improve sharpness of images from the bottom sub-network to the top sub-network, yielding inevitably high computational costs. Toward a fast and accurate deblurring network design, we revisit the coarse-to-fine strategy and present a multi-input multi-output U-net (MIMO-UNet). The MIMO-UNet has three distinct features. First, the single encoder of the MIMO-UNet takes multi-scale input images to ease the difficulty of training. Second, the single decoder of the MIMO-UNet outputs multiple deblurred images with different scales to mimic multi-cascaded U-nets using a single U-shaped network. Last, asymmetric feature fusion is introduced to merge multi-scale features in an efficient manner. Extensive experiments on the GoPro and RealBlur datasets demonstrate that the proposed network outperforms the state-of-the-art methods in terms of both accuracy and computational complexity. Source code is available for research purposes at https://github.com/chosj95/MIMO-UNet."
                },
                {
                    "title": "Mutual Affine Network for Spatially Variant Kernel Estimation in Blind Image Super-Resolution",
                    "abstract": "Existing blind image super-resolution (SR) methods mostly assume blur kernels are spatially invariant across the whole image. However, such an assumption is rarely applicable for real images whose blur kernels are usually spatially variant due to factors such as object motion and out-of-focus. Hence, existing blind SR methods would inevitably give rise to poor performance in real applications. To address this issue, this paper proposes a mutual affine network (MANet) for spatially variant kernel estimation. Specifically, MANet has two distinctive features. First, it has a moderate receptive field so as to keep the locality of degradation. Second, it involves a new mutual affine convolution (MAConv) layer that enhances feature expressiveness without increasing receptive field, model size and computation burden. This is made possible through exploiting channel interdependence, which applies each channel split with an affine transformation module whose input are the rest channel splits. Extensive experiments on synthetic and real images show that the proposed MANet not only performs favorably for both spatially variant and invariant kernel estimation, but also leads to state-of-the-art blind SR performance when combined with non-blind SR methods."
                },
                {
                    "title": "Focal Self-attention for Local-Global Interactions in Vision Transformers",
                    "abstract": "Recently, Vision Transformer and its variants have shown great promise on various computer vision tasks. The ability of capturing short- and long-range visual dependencies through self-attention is arguably the main source for the success. But it also brings challenges due to quadratic computational overhead, especially for the high-resolution vision tasks (e.g., object detection). In this paper, we present focal self-attention, a new mechanism that incorporates both fine-grained local and coarse-grained global interactions. Using this new mechanism, each token attends the closest surrounding tokens at fine granularity but the tokens far away at coarse granularity, and thus can capture both short- and long-range visual dependencies efficiently and effectively. With focal self-attention, we propose a new variant of Vision Transformer models, called Focal Transformer, which achieves superior performance over the state-of-the-art vision Transformers on a range of public image classification and object detection benchmarks. In particular, our Focal Transformer models with a moderate size of 51.1M and a larger size of 89.8M achieve 83.5 and 83.8 Top-1 accuracy, respectively, on ImageNet classification at 224x224 resolution. Using Focal Transformers as the backbones, we obtain consistent and substantial improvements over the current state-of-the-art Swin Transformers for 6 different object detection methods trained with standard 1x and 3x schedules. Our largest Focal Transformer yields 58.7/58.9 box mAPs and 50.9/51.3 mask mAPs on COCO mini-val/test-dev, and 55.4 mIoU on ADE20K for semantic segmentation, creating new SoTA on three of the most challenging computer vision tasks."
                },
                {
                    "title": "Uformer: A General U-Shaped Transformer for Image Restoration",
                    "abstract": "In this paper, we present Uformer, an effective and efficient Transformer-based architecture for image restoration, in which we build a hierarchical encoder-decoder network using the Transformer block. In Uformer, there are two core designs. First, we introduce a novel locally-enhanced window (LeWin) Transformer block, which performs non-overlapping window-based self-attention instead of global self-attention. It significantly reduces the computational complexity on high resolution feature map while capturing local context. Second, we propose a learnable multi-scale restoration modulator in the form of a multi-scale spatial bias to adjust features in multiple layers of the Uformer decoder. Our modulator demonstrates superior capability for restoring details for various image restoration tasks while introducing marginal extra parameters and computational cost. Powered by these two designs, Uformer enjoys a high capability for capturing both local and global dependencies for image restoration. To evaluate our approach, extensive experiments are conducted on several image restoration tasks, including image denoising, motion deblurring, defocus deblurring and deraining. Without bells and whistles, our Uformer achieves superior or comparable performance compared with the state-of-the-art algorithms. The code and models are available at https://github.com/ZhendongWang6/Uformer."
                },
                {
                    "title": "DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification",
                    "abstract": "Attention is sparse in vision transformers. We observe the final prediction in vision transformers is only based on a subset of most informative tokens, which is sufficient for accurate image recognition. Based on this observation, we propose a dynamic token sparsification framework to prune redundant tokens progressively and dynamically based on the input. Specifically, we devise a lightweight prediction module to estimate the importance score of each token given the current features. The module is added to different layers to prune redundant tokens hierarchically. To optimize the prediction module in an end-to-end manner, we propose an attention masking strategy to differentiably prune a token by blocking its interactions with other tokens. Benefiting from the nature of self-attention, the unstructured sparse tokens are still hardware friendly, which makes our framework easy to achieve actual speed-up. By hierarchically pruning 66% of the input tokens, our method greatly reduces 31%~37% FLOPs and improves the throughput by over 40% while the drop of accuracy is within 0.5% for various vision transformers. Equipped with the dynamic token sparsification framework, DynamicViT models can achieve very competitive complexity/accuracy trade-offs compared to state-of-the-art CNNs and vision transformers on ImageNet. Code is available at https://github.com/raoyongming/DynamicViT"
                },
                {
                    "title": "Deep Dynamic Scene Deblurring From Optical Flow",
                    "abstract": "Deblurring can not only provide visually more pleasant pictures and make photography more convenient, but also can improve the performance of objection detection as well as tracking. However, removing dynamic scene blur from images is a non-trivial task as it is difficult to model the non-uniform blur mathematically. Several methods first use single or multiple images to estimate optical flow (which is treated as an approximation of blur kernels) and then adopt non-blind deblurring algorithms to reconstruct the sharp images. However, these methods cannot be trained in an end-to-end manner and are usually computationally expensive. In this paper, we explore optical flow to remove dynamic scene blur by using the multi-scale spatially variant recurrent neural network (RNN). We utilize FlowNets to estimate optical flow from two consecutive images in different scales. The estimated optical flow provides the RNN weights in different scales so that the weights can better help RNNs to remove blur in the feature spaces. Finally, we develop a convolutional neural network (CNN) to restore the sharp images from the deblurred features. Both quantitatively and qualitatively evaluations on the benchmark datasets demonstrate that the proposed method performs favorably against state-of-the-art algorithms in terms of accuracy, speed, and model size."
                },
                {
                    "title": "HINet: Half Instance Normalization Network for Image Restoration",
                    "abstract": "In this paper, we explore the role of Instance Normalization in low-level vision tasks. Specifically, we present a novel block: Half Instance Normalization Block (HIN Block), to boost the performance of image restoration networks. Based on HIN Block, we design a simple and powerful multi-stage network named HINet, which consists of two subnetworks. With the help of HIN Block, HINet surpasses the state-of-the-art (SOTA) on various image restoration tasks. For image denoising, we exceed it 0.11dB and 0.28 dB in PSNR on SIDD dataset, with only 7.5% and 30% of its multiplier-accumulator operations (MACs), 6.8\u00d7 and 2.9\u00d7 speedup respectively. For image deblurring, we get comparable performance with 22.5% of its MACs and 3.3\u00d7 speedup on REDS and GoPro datasets. For image deraining, we exceed it by 0.3 dB in PSNR on the average result of multiple datasets with 1.4\u00d7 speedup. With HINet, we won the 1st place on the NTIRE 2021 Image Deblurring Challenge - Track2. JPEG Artifacts, with a PSNR of 29.70."
                },
                {
                    "title": "Scaling Local Self-Attention for Parameter Efficient Visual Backbones",
                    "abstract": "Self-attention has the promise of improving computer vision systems due to parameter-independent scaling of receptive fields and content-dependent interactions, in contrast to parameter-dependent scaling and content-independent interactions of convolutions. Self-attention models have recently been shown to have encouraging improvements on accuracy-parameter trade-offs compared to baseline convolutional models such as ResNet-50. In this work, we develop self-attention models that can outperform not just the canonical baseline models, but even the high-performing convolutional models. We propose two extensions to self-attention that, in conjunction with a more efficient implementation of self-attention, improve the speed, memory usage, and accuracy of these models. We leverage these improvements to develop a new self-attention model family, HaloNets, which reach state-of-the-art accuracies on the parameter-limited setting of the ImageNet classification benchmark. In preliminary transfer learning experiments, we find that HaloNet models outperform much larger models and have better inference performance. On harder tasks such as object detection and instance segmentation, our simple local self-attention and convolutional hybrids show improvements over very strong baselines. These results mark another step in demonstrating the efficacy of self-attention models on settings traditionally dominated by convolutions. 1"
                },
                {
                    "title": "Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions",
                    "abstract": "Although convolutional neural networks (CNNs) have achieved great success in computer vision, this work investigates a simpler, convolution-free backbone network use-fid for many dense prediction tasks. Unlike the recently-proposed Vision Transformer (ViT) that was designed for image classification specifically, we introduce the Pyramid Vision Transformer (PVT), which overcomes the difficulties of porting Transformer to various dense prediction tasks. PVT has several merits compared to current state of the arts. (1) Different from ViT that typically yields low-resolution outputs and incurs high computational and memory costs, PVT not only can be trained on dense partitions of an image to achieve high output resolution, which is important for dense prediction, but also uses a progressive shrinking pyramid to reduce the computations of large feature maps. (2) PVT inherits the advantages of both CNN and Transformer, making it a unified backbone for various vision tasks without convolutions, where it can be used as a direct replacement for CNN backbones. (3) We validate PVT through extensive experiments, showing that it boosts the performance of many downstream tasks, including object detection, instance and semantic segmentation. For example, with a comparable number of parameters, PVT+RetinaNet achieves 40.4 AP on the COCO dataset, surpassing ResNet50+RetinNet (36.3 AP) by 4.1 absolute AP (see Figure 2). We hope that PVT could, serre as an alternative and useful backbone for pixel-level predictions and facilitate future research."
                },
                {
                    "title": "Deblurring Dynamic Scenes via Spatially Varying Recurrent Neural Networks",
                    "abstract": "Deblurring images captured in dynamic scenes is challenging as the motion blurs are spatially varying caused by camera shakes and object movements. In this paper, we propose a spatially varying neural network to deblur dynamic scenes. The proposed model is composed of three deep convolutional neural networks (CNNs) and a recurrent neural network (RNN). The RNN is used as a deconvolution operator on feature maps extracted from the input image by one of the CNNs. Another CNN is used to learn the spatially varying weights for the RNN. As a result, the RNN is spatial-aware and can implicitly model the deblurring process with spatially varying kernels. To better exploit properties of the spatially varying RNN, we develop both one-dimensional and two-dimensional RNNs for deblurring. The third component, based on a CNN, reconstructs the final deblurred feature maps into a restored image. In addition, the whole network is end-to-end trainable. Quantitative and qualitative evaluations on benchmark datasets demonstrate that the proposed method performs favorably against the state-of-the-art deblurring algorithms."
                },
                {
                    "title": "Multi-Stage Progressive Image Restoration",
                    "abstract": "Image restoration tasks demand a complex balance between spatial details and high-level contextualized information while recovering images. In this paper, we propose a novel synergistic design that can optimally balance these competing goals. Our main proposal is a multi-stage architecture, that progressively learns restoration functions for the degraded inputs, thereby breaking down the overall recovery process into more manageable steps. Specifically, our model first learns the contextualized features using encoder-decoder architectures and later combines them with a high-resolution branch that retains local information. At each stage, we introduce a novel per-pixel adaptive design that leverages in-situ supervised attention to reweight the local features. A key ingredient in such a multi-stage architecture is the information exchange between different stages. To this end, we propose a two-faceted approach where the information is not only exchanged sequentially from early to late stages, but lateral connections between feature processing blocks also exist to avoid any loss of information. The resulting tightly interlinked multi-stage architecture, named as MPRNet, delivers strong performance gains on ten datasets across a range of tasks including image deraining, deblurring, and denoising. The source code and pre-trained models are available at https://github.com/swz30/MPRNet."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "DeblurGAN-v2: Deblurring (Orders-of-Magnitude) Faster and Better",
                    "abstract": "We present a new end-to-end generative adversarial network (GAN) for single image motion deblurring, named DeblurGAN-V2, which considerably boosts state-of-the-art deblurring performance while being much more flexible and efficient. DeblurGAN-V2 is based on a relativistic conditional GAN with a double-scale discriminator. For the first time, we introduce the Feature Pyramid Network into deblurring, as a core building block in the generator of DeblurGAN-V2. It can flexibly work with a wide range of backbones, to navigate the balance between performance and efficiency. The plug-in of sophisticated backbones (e.g. Inception ResNet v2) can lead to solid state-of-the-art performance. Meanwhile, with light-weight backbones (e.g. MobileNet and its variants), DeblurGAN-V2 becomes 10-100 times faster than the nearest competitors, while maintaining close to state-of-the-art results, implying the option of real-time video deblurring. We demonstrate that DeblurGAN-V2 has very competitive performance on several popular benchmarks, in terms of deblurring quality (both objective and subjective), as well as efficiency. In addition, we show the architecture to be effective for general image restoration tasks too. Our models and codes will be made available upon acceptance."
                },
                {
                    "title": "Learning to Understand Image Blur",
                    "abstract": "While many approaches have been proposed to estimate and remove blur in a photo, few efforts were made to have an algorithm automatically understand the blur desirability: whether the blur is desired or not, and how it affects the quality of the photo. Such a task not only relies on low-level visual features to identify blurry regions, but also requires high-level understanding of the image content as well as user intent during photo capture. In this paper, we propose a unified framework to estimate a spatially-varying blur map and understand its desirability in terms of image quality at the same time. In particular, we use a dilated fully convolutional neural network with pyramid pooling and boundary refinement layers to generate high-quality blur response maps. If blur exists, we classify its desirability to three levels ranging from good to bad, by distilling high-level semantics and learning an attention map to adaptively localize the important content in the image. The whole framework is end-to-end jointly trained with both supervisions of pixel-wise blur responses and image-wise blur desirability levels. Considering the limitations of existing image blur datasets, we collected a new large-scale dataset with both annotations to facilitate training. The proposed methods are extensively evaluated on two datasets and demonstrate state-of-the-art performance on both tasks."
                },
                {
                    "title": "Scale-Recurrent Network for Deep Image Deblurring",
                    "abstract": "In single image deblurring, the \"coarse-to-fine\" scheme, i.e. gradually restoring the sharp image on different resolutions in a pyramid, is very successful in both traditional optimization-based methods and recent neural-network-based approaches. In this paper, we investigate this strategy and propose a Scale-recurrent Network (SRN-DeblurNet) for this deblurring task. Compared with the many recent learning-based approaches in [25], it has a simpler network structure, a smaller number of parameters and is easier to train. We evaluate our method on large-scale deblurring datasets with complex motion. Results show that our method can produce better quality results than state-of-the-arts, both quantitatively and qualitatively."
                },
                {
                    "title": "DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networks",
                    "abstract": "We present DeblurGAN, an end-to-end learned method for motion deblurring. The learning is based on a conditional GAN and the content loss. DeblurGAN achieves state-of-the art performance both in the structural similarity measure and visual appearance. The quality of the deblurring model is also evaluated in a novel way on a real-world problem - object detection on (de-)blurred images. The method is 5 times faster than the closest competitor - Deep-Deblur [25]. We also introduce a novel method for generating synthetic motion blurred images from sharp ones, allowing realistic dataset augmentation. The model, code and the dataset are available at https://github.com/KupynOrest/DeblurGAN"
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Deep Multi-scale Convolutional Neural Network for Dynamic Scene Deblurring",
                    "abstract": "Non-uniform blind deblurring for general dynamic scenes is a challenging computer vision problem as blurs arise not only from multiple object motions but also from camera shake, scene depth variation. To remove these complicated motion blurs, conventional energy optimization based methods rely on simple assumptions such that blur kernel is partially uniform or locally linear. Moreover, recent machine learning based methods also depend on synthetic blur datasets generated under these assumptions. This makes conventional deblurring methods fail to remove blurs where blur kernel is difficult to approximate or parameterize (e.g. object motion boundaries). In this work, we propose a multi-scale convolutional neural network that restores sharp images in an end-to-end manner where blur is caused by various sources. Together, we present multi-scale loss function that mimics conventional coarse-to-fine approaches. Furthermore, we propose a new large-scale dataset that provides pairs of realistic blurry image and the corresponding ground truth sharp image that are obtained by a high-speed camera. With the proposed model trained on this dataset, we demonstrate empirically that our method achieves the state-of-the-art performance in dynamic scene deblurring not only qualitatively, but also quantitatively."
                },
                {
                    "title": "Categorical Reparameterization with Gumbel-Softmax",
                    "abstract": "Categorical variables are a natural choice for representing discrete structure in the world. However, stochastic neural networks rarely use categorical latent variables due to the inability to backpropagate through samples. In this work, we present an efficient gradient estimator that replaces the non-differentiable sample from a categorical distribution with a differentiable sample from a novel Gumbel-Softmax distribution. This distribution has the essential property that it can be smoothly annealed into a categorical distribution. We show that our Gumbel-Softmax estimator outperforms state-of-the-art gradient estimators on structured output prediction and unsupervised generative modeling tasks with categorical latent variables, and enables large speedups on semi-supervised classification."
                },
                {
                    "title": "SGDR: Stochastic Gradient Descent with Warm Restarts",
                    "abstract": "Restart techniques are common in gradient-free optimization to deal with multimodal functions. Partial warm restarts are also gaining popularity in gradient-based optimization to improve the rate of convergence in accelerated gradient schemes to deal with ill-conditioned functions. In this paper, we propose a simple warm restart technique for stochastic gradient descent to improve its anytime performance when training deep neural networks. We empirically study its performance on the CIFAR-10 and CIFAR-100 datasets, where we demonstrate new state-of-the-art results at 3.14% and 16.21%, respectively. We also demonstrate its advantages on a dataset of EEG recordings and on a downsampled version of the ImageNet dataset. Our source code is available at this https URL"
                },
                {
                    "title": "Adam: A Method for Stochastic Optimization",
                    "abstract": "We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
                },
                {
                    "title": "Localized Image Blur Removal through Non-parametric Kernel Estimation",
                    "abstract": "We address the problem of estimating and removing localized image blur, as it for example arises from moving objects in a scene, or when the depth of field is insufficient to sharply render all objects of interest. Unlike the case of camera shake, such blur changes abruptly at the object boundaries. To cope with this, we propose an automated sharp image recovery method that simultaneously determines blurred regions and estimates their responsible blur kernels. To address a wide range of different scenarios, our model is not restricted to a discrete set of candidate blurs, but allows for arbitrary, non-parametric blur kernels. Moreover, our approach does not require specialized hardware, an alpha matte, or user annotation of the blurred region. Unlike previous methods, we show that localized blur estimation can be accomplished by incorporating a pixel-wise latent variable to indicate the active blur kernel. Furthermore, we generalize the marginal likelihood technique of blind deblurring to the case of localized blur. Specifically, we integrate out the latent image derivatives to permit marginal density estimates of both blur kernels and their regions of influence. We obtain sharp images in applications to both object motion blur and defocus blur removal. Quantitative results on two novel datasets as well as qualitative results comparing to a range of specialized methods demonstrate the versatility and effectiveness of our non-parametric approach."
                },
                {
                    "title": "Image quality assessment: from error visibility to structural similarity",
                    "abstract": "Objective methods for assessing perceptual image quality traditionally attempted to quantify the visibility of errors (differences) between a distorted image and a reference image using a variety of known properties of the human visual system. Under the assumption that human visual perception is highly adapted for extracting structural information from a scene, we introduce an alternative complementary framework for quality assessment based on the degradation of structural information. As a specific example of this concept, we develop a structural similarity index and demonstrate its promise through a set of intuitive examples, as well as comparison to both subjective ratings and state-of-the-art objective methods on a database of images compressed with JPEG and JPEG2000. A MATLAB implementation of the proposed algorithm is available online at http://www.cns.nyu.edu//spl sim/lcv/ssim/."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively restore high-resolution images affected by local motion blur using a sparse vision Transformer approach?\n\n### [Question 2] - Why is it interesting and important?\nSolving the problem of local motion deblurring is crucial for various applications in computer vision, such as enhancing image quality in photography, improving video clarity, and aiding in autonomous vehicle navigation. By addressing this issue, the research community can advance the understanding of motion blur dynamics and develop more sophisticated image processing techniques. The proposed LMD-ViT could pave the way for future research into efficient deep learning architectures that focus on localized image restoration, potentially leading to practical applications in real-time image enhancement and video processing.\n\n### [Question 3] - Why is it hard?\nThe challenge in local motion deblurring lies in accurately identifying and restoring only the affected regions of an image while preserving the integrity of sharp areas. Naive approaches may fail because they often treat the image as a whole, leading to over-smoothing or artifacts in non-blurred regions. The complexities include the need for precise localization of blur, the variability of motion patterns, and the computational efficiency required for real-time applications. Additionally, existing methods may struggle with the trade-off between performance and speed, making it difficult to achieve high-quality results without significant computational resources.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on global motion deblurring, which does not adequately address the nuances of local motion blur. Existing solutions often rely on convolutional neural networks (CNNs) that may not effectively capture the spatial dependencies of localized blur. Barriers to solving this problem include the lack of specialized architectures that prioritize local features and the computational overhead associated with processing high-resolution images. The LMD-ViT approach differs by utilizing adaptive window pruning to concentrate on blurred regions, thereby improving efficiency and effectiveness compared to prior methods.\n\n### [Question 5] - What are the key components of my approach and results?\nThe proposed methodology involves the use of a U-shaped local motion deblurring vision Transformer (LMD-ViT) with adaptive window pruning Transformer blocks (AdaWPT). The model will be trained on the GoPro and ReLoBlur datasets, focusing on both local and global motion deblurring. Key metrics for evaluation will include Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM). Expected outcomes include significant improvements in image quality, faster inference times, and the ability to"
            }
        },
        "author_data": {
            "8d0cfc29-ddba-4a82-864b-9df3b5f6e521": {
                "pk": "8d0cfc29-ddba-4a82-864b-9df3b5f6e521",
                "name": "Haoying Li",
                "collaborators": [
                    "H. Feng",
                    "Zhi-hai Xu",
                    "Qi Li",
                    "Yue-ting Chen",
                    "R. Zeng",
                    "M. Tu",
                    "Ziran Zhang",
                    "Tingting Jiang",
                    "Honghao Chen",
                    "Shenyu Yang",
                    "T. Xue",
                    "Fangjun Ruan",
                    "Ting Xue",
                    "Jinshun Liu",
                    "Guomian Lv",
                    "Hao Zhou",
                    "Peng Luo",
                    "Meng Chang",
                    "Yiping Huang",
                    "Peishan Jian",
                    "Z. Xie",
                    "Youheng Wu",
                    "Fangmin Situ",
                    "Ke Xu",
                    "Yingjie Liu",
                    "Yujie Wang",
                    "Jiajun Tian",
                    "Liang Wang",
                    "Jiangbing Du",
                    "Zuyuan He",
                    "Q. Song",
                    "J. Xv",
                    "Wenrui Zhang",
                    "Guichao Lai",
                    "Haoyu Xue",
                    "Jianhao Zhao",
                    "Jian Xu",
                    "Tianyue Zhang",
                    "Ziwei Feng",
                    "Dejiao Hu",
                    "Fei Qin",
                    "Xueying Ouyang",
                    "Yaoyu Cao",
                    "Lingxiang Jiang",
                    "Xiangping Li",
                    "Lin Du",
                    "Xiaohui Yang",
                    "Wenqiang Li",
                    "Shanbao Feng",
                    "Bin Yu",
                    "L. Xiao",
                    "Y. Liu",
                    "H. Nie"
                ],
                "domain": [
                    "Optical Imaging",
                    "Fluid Dynamics",
                    "Biomedical Engineering",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Investigation of Temperature Field of Falling Film Based on Planar Laser-Induced Fluorescence",
                        "abstract": "Falling film is widely applied in the industry due to its efficient heat and mass transfer. The temperature field distribution is essential for the research of heat transfer characteristics and mechanisms. However, few studies have been conducted on the temperature field measurement of the falling film in the pipe. In this paper, the reasons for the distortion of fluorescence intensity in the flow and radial directions are investigated for falling water film. A fluorescence intensity-temperature calibration curve is fitted with Gaussian compensation and exponential attenuation compensation for global temperature visualization. The falling water film thermal characteristics are analyzed qualitatively, demonstrating that the relationship between interface temperature and film thickness is not exactly inverse but rather shifts in phase. Moreover, combining the temperature distribution with the liquid film flow topology, two potential hydrodynamic behaviors are interpreted. In the combination between the disturbance wave and base film, the high-temperature liquid flows rapidly from the base film to the colder wave peak. Whereas in partially isolated waves, the internal flow circulation in the wave front drives the homogeneous mixing of local temperature."
                    },
                    {
                        "title": "A Novel 3-D Reconstruction Method of Liquid Film Based on Scanning Laser-Induced Fluorescence",
                        "abstract": "The accurate measurement of liquid film in the pipe is essential for process monitoring and industrial safety. Three-dimensional measurement can effectively improve the measuring accuracy. However, there are few measurement methods and limited spatial resolution. In this article, a new scanning laser-induced fluorescence (SLIF) method is proposed to perform the 3-D reconstruction of liquid film in the pipe with high precision. Combined with Zhang\u2019s calibration method, the imaging magnifications of slice images are evaluated and corrected. In order to eliminate the motion smear in fluorescence imaging, a prediction model is established. In addition, an innovative spatial distortion correction method has been developed to reconstruct circumferential liquid film, and the mean absolute percentage error is less than 1.56%. Furthermore, an advanced contour interpolation method based on inverse distance transformation and volume rendering method is performed sequentially to achieve 3-D reconstruction. Moreover, the volume measurement is carried out and concludes that the relative error of reconstruction results is 2.38%, which verifies the reliability of the 3-D reconstruction method based on SLIF. Finally, the 3-D reconstruction of the falling liquid film is conducted, which characterizes the detailed structure of the liquid film. These results supply key solutions for 3-D reconstruction and can be accomplished successfully in experimental research."
                    },
                    {
                        "title": "Deep learning-based image reconstruction for photonic integrated interferometric imaging.",
                        "abstract": "Photonic integrated interferometric imaging (PIII) is an emerging technique that uses far-field spatial coherence measurements to extract intensity information from a source to form an image. At present, low sampling rate and noise disturbance are the main factors hindering the development of this technology. This paper implements a deep learning-based method to improve image quality. Firstly, we propose a frequency-domain dataset generation method based on imaging principles. Secondly, spatial-frequency dual-domain fusion networks (SFDF-Nets) are presented for image reconstruction. We utilize normalized amplitude and phase to train networks, which reduces the difficulty of network training using complex data. SFDF-Nets can fuse multi-frame data captured by rotation sampling to increase the sampling rate and generate high-quality spatial images through dual-domain supervised learning and frequency domain fusion. Furthermore, we propose an inverse fast Fourier transform loss (IFFT loss) for network training in the frequency domain. Extensive experiments show that our method improves PSNR and SSIM by 5.64 dB and 0.20, respectively. Our method effectively improves the reconstructed image quality and opens a new dimension in interferometric imaging."
                    },
                    {
                        "title": "Real-world Deep Local Motion Deblurring",
                        "abstract": "Most existing deblurring methods focus on removing global blur caused by camera shake, while they cannot well handle local blur caused by object movements. To fill the vacancy of local deblurring in real scenes, we establish the first real local motion blur dataset (ReLoBlur), which is captured by a synchronized beam-splitting photographing system and corrected by a post-progressing pipeline. Based on ReLoBlur, we propose a Local Blur-Aware Gated network (LBAG) and several local blur-aware techniques to bridge the gap between global and local deblurring: 1) a blur detection approach based on background subtraction to localize blurred regions; 2) a gate mechanism to guide our network to focus on blurred regions; and 3) a blur-aware patch cropping strategy to address data imbalance problem. Extensive experiments prove the reliability of ReLoBlur dataset, and demonstrate that LBAG achieves better performance than state-of-the-art global deblurring methods and our proposed local blur-aware techniques are effective."
                    },
                    {
                        "title": "A plug-and-play demosaicing and denoising method on mobile cameras",
                        "abstract": "Image demosaicing and denoising are two important processes in the ISP pipeline of mobile cameras, because almost all mobile cameras in use today require colorful images generated by demosaicing algorithm, and the small sensor area of mobile cameras triggers low signal-to-noise ratio. Over the years, a considerable number of sequential demosaicing and denoising methods have been proposed, while they su\ufb00er from estimating the noise distribution and adjusting the hyper-parameters in order to balance demosaicing and denoising. There exit simultaneous demosaicing and denoising methods solving these problems. But they lack guidelines designed for mobile cameras. We propose a Plug-and-Play (PnP) demosaicing and denoising method on mobile cameras. Our method is built on PnP demosaicing framework which is derived from variable splitting theory. Any color demosaicing algorithm (i.e., bilinear, Malvar) can be plugged into our framework. We novelly trained an ISO conditioned denoiser for the framework and iteratively apply the denoiser in it. The ISO conditioned denoiser not only removes noise from the demosaicing procedure itself but also noise from camera sensors. By introducing ISO settings to the denoiser, our method takes possession of the adaptability and robustness in various capturing environments under di\ufb00erent camera settings. Our method has only two hyperparameters to tune, which eases the hyper-parameter adjustment in sequential demosaicing and denoising methods. Extensive experiments on synthetic datasets show that our method performs better than sequential demosaicing and denoising methods and is practical for mobile cameras."
                    },
                    {
                        "title": "Imaging degradation evaluation on beam-splitting devices",
                        "abstract": "Researchers have proposed beam-splitting imaging methods to solve the problem that capturing the simultaneous view by multiple imaging instruments and positionally aligning them pixel by pixel. Beam-splitting prism and beam-splitting plate are two kinds of commonly used beam-splitting devices. However, which kind of the two devices degrade images to a less extent lacks discussion. Firstly, we theoretically analyzed possible image degradation caused by beam-splitting devices, which mainly include spherical aberration, ghost effect, a non-uniformity of the degradation function, and color cast. And we used ZEMAX optical simulation software to establish a beam-splitting imaging emulation experiment to simulate the image degradations mentioned above. we constructed an experience system of beam-splitting imaging in the laboratory. researchers could select a suitable device for their projects based on our study."
                    },
                    {
                        "title": "Enhanced cell affinity and osteogenic differentiation of liquid crystal-based substrate via surface bio-functionalization.",
                        "abstract": "Regulation of cell-substrate interactions is an important factor for modulating cell behaviors. Tailoring the physical and chemical properties of the substrates to better mimic the extracellular matrix (ECM) of native tissue is a more effective strategy for enhancing the cell-substrate contact. In current work, we aim at improving surface bioactivity based on the liquid crystalline substrates for the enhancement in cell affinity and osteogenic differentiation. Polydopamine adhesive coating (PDOPA) was used as a reactive platform for the immobilization of chitooligosaccharide (COS) on the octyl hydroxypropyl cellulose ester (OPC) substrate to generate active OPC-PDOPA-COSs liquid crystalline substrates. Results demonstrated that PDOPA-coated OPC surfaces showed remarkably improved hydrophility and increased elastic modulus, leading to better initial cell attachment. Subsequent COS immobilization on the OPC-PDOPA layer could induce promotion of cell proliferation, polarization and cytoskeleton formation. rBMSCs seeded on the OPC-PDOPA-COSs showed higher ALP activity, calcium deposition, and up-regulated bone-related genes expression, including BMP-2, RUNx-2, COL-I and OCN. In conclusion, surface biofunctionalization on the OPC-based liquid crystalline substrates could come into being the appropriate combination of surface chemistry and liquid crystalline characteristic that simulating in vivo ECM environment, resulting in a favorable support to enhance positive cell-substrate interactions. This article is protected by copyright. All rights reserved."
                    },
                    {
                        "title": "Optical Fiber Humidity Sensor Based on Water Absorption Peak Near 2-\u03bcm Waveband",
                        "abstract": "We demonstrate the absorption-based fiber humidity sensors working near 1950\u00a0nm wavelength which is a strong water absorption peak. No additional coating of humidity-sensitive material is needed for the proposed relative humidity (RH) fiber sensors. Two sensor structures including a taper fiber and a microfiber knot resonator are fabricated. For the taper fiber sensor, upto 0.18 mW/% RH sensitivity has been achieved via direct power measurement. For the fiber knot sensor, the RH variations induce both extinction ratio change and wavelength shift. The optical spectra of the fiber knot resonator under different RH are measured, and the extinction ratio and wavelength shift sensitivity are measured to be 0.034 dB/% RH and 10 pm/% RH, respectively. The response time for the tapered fiber is 1.13\u00a0s for RH increasing from 28.6% to 53.4% and 2.27\u00a0s for the reverse operation. The microfiber knot sensor can respond in 0.8 and 1.55\u00a0s accordingly. The sensor performance for the two structures is also investigated under different temperatures from 25\u00a0\u00b0C to 45\u00a0\u00b0C and the results indicate negligible impact from the temperature variations. We believe the simple fabrication and low cost of the sensor will make it potential and practical for the RH measurement."
                    },
                    {
                        "title": "Anti-biofouling and functionalizable bioinspired chitosan-based hydrogel coating via surface photo-immobilization",
                        "abstract": "Abstract Zwitterionic polymer is a new generation of anti-fouling materials with its good resistance to protein and bacterial adhesion. Constructing the anti-fouling surfaces with zwitterionic polymer has been regarded as an effective approach for improving the biocompatibility and biofunctionality of clinic devices. Herein, we reported a facile approach to construct a biodegradable anti-biofouling and functionalizable hydrogel coating via photo-immobilization using commercial polyethylene terephthalate (PET) films as the substrate, based on zwitterionic glycidyl methacrylate-phosphorylcholine-chitosan (PCCs-GMA). The surface structure and physicochemical properties of zwitterionic PCCs-GMA hydrogel coating were investigated by X-ray photoelectron spectroscopy (XPS), atomic force microscope (AFM) and static water contact angle measurement, and its functionalizable sites were detected by fluorescence labeling. Compared with the pristine PET and cationic chitosan - GMA and hydroxypropyltrimethyl ammonium chloride chitosan (HTCC) - GMA hydrogel coatings, zwitterionic PCCs-GMA hydrogel coating exhibited excellent biocompatibility, and significantly reduced protein adsorption for three model proteins of fibrinogen, immunoglobulin and lysozyme, repelled platelet adhesion, as well as showed a high resistance to bacterial attachment of Escherichia coli and Staphylococcus aureus and superior anti-fouling properties to MRC-5 cells. The results indicated that photo-immobilized zwitterionic PCCs-GMA hydrogel coating has perspective as a dual functional platform with integrated antifouling and further biofunctional properties for various biomedical applications."
                    },
                    {
                        "title": "Plasmonic Nanoprobes for Multiplexed Fluorescence\u2010Free Super\u2010Resolution Imaging",
                        "abstract": "Noble metal nanoparticles (NPs), owing to their unique optical and physicochemical properties, are routinely used for optical imaging and labeling of biological specimens. Even though they can provide vital information for studying multiple cellular events and their interplays at the same time, optically multiplexing and resolving specific NPs within a diffraction\u2010limited region labeled in complex biological specimens remains a fundamental challenge. By introducing and manipulating plasmonic resonance assisted saturable scattering effects, multiplexed fluorescence\u2010free super\u2010resolution imaging of gold NPs in tumor cells with remarkable subdiffraction resolution is demonstrated. The saturable scattering allows fluorescence\u2010free resolving single plasmonic nanoprobes with significantly improved resolution down to \u2248100 nm. The revealed plasmonic resonance assisted saturation effect as well as the associated spectral flexibility to variant sizes provides access to multiplexing capability in complex bioenvironments. The demonstrated feasibility of two\u2010color super\u2010resolution cellular imaging is achieved at ultralow suppression powers \u22480.28 MW cm\u22122, corresponding to a two\u2010order of magnitude improvement compared to the state\u2010of\u2010the\u2010art of stimulated emission depletion (STED) nanoscopy."
                    },
                    {
                        "title": "Time-of-flight secondary ion mass spectrometry analyses of vancomycin.",
                        "abstract": "As an antibiotic that prevents and treats infections caused by Gram-positive bacteria such as Staphylococcus aureus, vancomycin incorporated in a biodegradable polymer poly(lactide-co-glycolide) provides opportunities to construct controlled-release drug delivery systems. Developments associated with this promising system have been largely concentrated on areas of drug delivery kinetics and biodegradability. In order to provide surface analytical approaches to this important system, the authors demonstrate applicability of time-of-flight secondary ion mass spectrometry (TOF-SIMS) in three-dimensional molecular imaging for a model system consisting of alternating layers of ploy(lactide-co-glycolide) and vancomycin. TOF-SIMS imaging clarified that the two chemicals can undergo phase separation when dimethyl sulfoxide is used as the solvent. The authors identified two diagnostic ions that are abundant and structural moieties of vancomycin. The results on TOF-SIMS imaging and depth profiling vancomycin provide useful information for further applications of TOF-SIMS in the development of antibiotic drug delivery systems involving the use of vancomycin."
                    }
                ]
            },
            "58081fdd-1bb5-480c-b41b-5e02a589d36b": {
                "pk": "58081fdd-1bb5-480c-b41b-5e02a589d36b",
                "name": "Jixin Zhao",
                "collaborators": [
                    "Qixin Yan",
                    "Chunle Guo",
                    "Yu Dai",
                    "Chen Change Loy",
                    "Chongyi Li"
                ],
                "domain": [
                    "Computer Vision",
                    "Image Processing",
                    "Deep Learning",
                    "Style Transfer"
                ],
                "publications": [
                    {
                        "title": "BeautyREC: Robust, Efficient, and Content-preserving Makeup Transfer",
                        "abstract": "In what follows, we detail our network structure and parameters, including content encoder, style encoder, componentspecific correspondence, long-range dependencies, and image reconstruction. The content encoder is composed of three convolutional layers with kernels {7, 4, 4} and strides {1, 2, 2}, and three resblocks. Each convolutional layer is followed by an instance norm layer and a ReLU activation function. The number of feature channels is set to 48. The global style encoder has the same architectures and parameters as the component style encoder. The style encoder has four convolutional layers with kernels {7, 4, 4, 1} and strides {1, 2, 2, 1}. The number of feature channels is set to 48. We did not adopt the normalization operation after the convolutional layers for style encoders. All layers use the ReLU as the activation function. In the component-specific correspondence module, we use a three-layer fully-connected network with an architecture C \u2192 Cr \u2192 C(C = 48, r = 16) to learn channel-wise weights from statistic information of part-specific styles. For spatial attention, we separately apply mean and maximum pooling operations to the feature map that is re-weighted by the channel attention module along the channel direction. The resulting two maps by mean and maximum pooling operations are concatenated to learn a spatial attention mask using a 1\u00d7 1 convolutional layer with the Sigmoid activation function. In the long-range dependencies module, a transformer-based structure is employed to exploit the long-range dependencies between the source image and the reference image. The query of our model represents the global style features extracted by the global style encoder. The key and value of our model are the content features. We use sinusoidal positional encodings for the query and key features. The number of heads in the multi-head attention module is set to 8. A two-layer MLP with a residual connection between its input and output is placed at the end of the transformer unit. In the image reconstruction stage, three residual blocks are used to progressively refine features. We exploit two upsample operations and convolutional layers to recover the image resolution. The component-specific transferred features and the global transferred features are concatenated with the corresponding decoder features. For the last convolutional layer, we use the Tanh activation function without normalization operation to produce the final result."
                    },
                    {
                        "title": "BeautyREC: Robust, Efficient, and Component-Specific Makeup Transfer",
                        "abstract": "In this work, we propose a Robust, Efficient, and Component-specific makeup transfer method (abbreviated as BeautyREC). A unique departure from prior methods that leverage global attention, simply concatenate features, or implicitly manipulate features in latent space, we propose a component-specific correspondence to directly transfer the makeup style of a reference image to the corresponding components (e.g., skin, lips, eyes) of a source image, making elaborate and accurate local makeup transfer. As an auxiliary, the long-range visual dependencies of Transformer are introduced for effective global makeup transfer. Instead of the commonly used cycle structure that is complex and unstable, we employ a content consistency loss coupled with a content encoder to implement efficient single-path makeup transfer. The key insights of this study are modeling component-specific correspondence for local makeup transfer, capturing long-range dependencies for global makeup transfer, and enabling efficient makeup transfer via a single-path structure.We also contribute BeautyFace, a makeup transfer dataset to supplement existing datasets. This dataset contains 3,000 faces, covering more diverse makeup styles, face poses, and races. Each face has annotated parsing map. Extensive experiments demonstrate the effectiveness of our method against state-of-the-art methods. Besides, our method is appealing as it is with only 1M parameters, outperforming the state-of-the-art methods (BeautyGAN: 8.43M, PSGAN: 12.62M, SCGAN: 15.30M, CPM: 9.24M, SSAT: 10.48M)."
                    }
                ]
            },
            "39818676-d814-49f0-98f4-5b6af9ec3908": {
                "pk": "39818676-d814-49f0-98f4-5b6af9ec3908",
                "name": "Shangchen Zhou",
                "collaborators": [
                    "Chen Change Loy",
                    "Chongyi Li",
                    "Yu Dai",
                    "Ruicheng Feng",
                    "Peiqing Yang",
                    "Zongsheng Yue",
                    "Qingpeng Zhu",
                    "Qi Sun",
                    "Wenxiu Sun",
                    "Jinwei Gu",
                    "Yihang Luo",
                    "Xiaoming Li",
                    "Kai Chen",
                    "Zhexin Liang",
                    "Qinyue Li",
                    "Chunle Guo",
                    "Jianyi Wang",
                    "Kelvin C. K. Chan",
                    "Qingyu Yang",
                    "Jianzong Wu",
                    "Xiangtai Li",
                    "Chenyang Si",
                    "Jingkang Yang",
                    "Jiangning Zhang",
                    "Yining Li",
                    "Yunhai Tong",
                    "Ziwei Liu",
                    "Zhaochen Li",
                    "Yaqi Wu",
                    "Zhihao Fan",
                    "X. Chu",
                    "Jimmy S. Ren",
                    "Senyan Xu",
                    "Zhijing Sun",
                    "Jiaying Zhu",
                    "Yurui Zhu",
                    "Xueyang Fu",
                    "Zheng-Jun Zha",
                    "Jun Cao",
                    "Cheng Li",
                    "Shu Chen",
                    "Liang Ma",
                    "Shiyang Zhou",
                    "Haijin Zeng",
                    "Kai Feng",
                    "Yongyong Chen",
                    "Jingyong Su",
                    "Xianyu Guan",
                    "Hongyuan Yu",
                    "Cheng Wan",
                    "Jiamin Lin",
                    "Bin Han",
                    "Yajun Zou",
                    "Zhuoyuan Wu",
                    "Yuan Huang",
                    "Yongsheng Yu",
                    "Daoan Zhang",
                    "Jizhe Li",
                    "Xuanwu Yin",
                    "Kunlong Zuo",
                    "Yunfan Lu",
                    "Yijie Xu",
                    "Wenzong Ma",
                    "Weiyu Guo",
                    "Hui Xiong",
                    "Wei Yu",
                    "Bingchun Luo",
                    "S. Nathan",
                    "Priya Kansal",
                    "Yushi Lan",
                    "Fangzhou Hong",
                    "Shuai Yang",
                    "Xuyi Meng",
                    "Bo Dai",
                    "Xingang Pan",
                    "Dafeng Zhang",
                    "Zhezhu Jin",
                    "Guanqun Liu",
                    "Xin Jin",
                    "Ruoqi Li",
                    "Chang Liu",
                    "Ziyi Wang",
                    "Yao Du",
                    "Jingjing Yang",
                    "Long Bao",
                    "Heng Sun",
                    "Xiangyu Kong",
                    "Xiaoxia Xing",
                    "Jinlong Wu",
                    "Yu-Na Xue",
                    "Hyunhee Park",
                    "Sejun Song",
                    "Changho Kim",
                    "Jingfan Tan",
                    "Wenhan Luo",
                    "Zikun Liu",
                    "Mingde Qiao",
                    "Junjun Jiang",
                    "Kui Jiang",
                    "Yao Xiao"
                ],
                "domain": [
                    "Computer Vision",
                    "Image Processing",
                    "Deep Learning",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Towards Language-Driven Video Inpainting via Multimodal Large Language Models",
                        "abstract": "We introduce a new task - language-driven video inpainting, which uses natural language instructions to guide the inpainting process. This approach overcomes the limitations of traditional video inpainting methods that depend on manually labeled binary masks, a process often tedious and labor-intensive. We present the Remove Objects from Videos by Instructions (ROVI) dataset, containing 5,650 videos and 9,091 inpainting results, to support training and evaluation for this task. We also propose a novel diffusion-based language-driven video inpainting framework, the first end-to-end baseline for this task, integrating Multimodal Large Language Models to understand and execute complex language-based inpaintingrequests effectively. Our comprehensive results showcase the dataset's versatility and the model's effectiveness in various language-instructed inpainting scenarios. We have made datasets, code, and models publicly available at https://github.com/jianzongwu/Language-Driven-Video-Inpainting."
                    },
                    {
                        "title": "Control Color: Multimodal Diffusion-based Interactive Image Colorization",
                        "abstract": "Despite the existence of numerous colorization methods, several limitations still exist, such as lack of user interaction, inflexibility in local colorization, unnatural color rendering, insufficient color variation, and color overflow. To solve these issues, we introduce Control Color (CtrlColor), a multi-modal colorization method that leverages the pre-trained Stable Diffusion (SD) model, offering promising capabilities in highly controllable interactive image colorization. While several diffusion-based methods have been proposed, supporting colorization in multiple modalities remains non-trivial. In this study, we aim to tackle both unconditional and conditional image colorization (text prompts, strokes, exemplars) and address color overflow and incorrect color within a unified framework. Specifically, we present an effective way to encode user strokes to enable precise local color manipulation and employ a practical way to constrain the color distribution similar to exemplars. Apart from accepting text prompts as conditions, these designs add versatility to our approach. We also introduce a novel module based on self-attention and a content-guided deformable autoencoder to address the long-standing issues of color overflow and inaccurate coloring. Extensive comparisons show that our model outperforms state-of-the-art image colorization methods both qualitatively and quantitatively."
                    },
                    {
                        "title": "Learning Inclusion Matching for Animation Paint Bucket Colorization",
                        "abstract": "Colorizing line art is a pivotal task in the production of hand-drawn cel animation. This typically involves digital painters using a paint bucket tool to manually color each segment enclosed by lines, based on RGB values predeter-mined by a color designer. This frame-by-frame process is both arduous and time-intensive. Current automated meth-ods mainly focus on segment matching. This technique mi-grates colors from a reference to the target frame by aligning features within line-enclosed segments across frames. However, issues like occlusion and wrinkles in animations often disrupt these direct correspondences, leading to mis-matches. In this work, we introduce a new learning-based inclusion matching pipeline, which directs the network to comprehend the inclusion relationships between segments rather than relying solely on direct visual correspondences. Our method features a two-stage pipeline that integrates a coarse color warping module with an inclusion matching module, enabling more nuanced and accurate colorization. To facilitate the training of our network, we also develope a unique dataset, referred to as PaintBucket-Character. This dataset includes rendered line arts alongside their colorized counterparts, featuring various 3D characters. Extensive experiments demonstrate the effectiveness and superiority of our method over existing techniques."
                    },
                    {
                        "title": "Paint Bucket Colorization Using Anime Character Color Design Sheets",
                        "abstract": "Line art colorization plays a crucial role in hand-drawn animation production, where digital artists manually colorize segments using a paint bucket tool, guided by RGB values from character color design sheets. This process, often called paint bucket colorization, involves two main tasks: keyframe colorization, where colors are applied according to the character's color design sheet, and consecutive frame colorization, where these colors are replicated across adjacent frames. Current automated colorization methods primarily focus on reference-based and segment-matching approaches. However, reference-based methods often fail to accurately assign specific colors to each region, while matching-based methods are limited to consecutive frame colorization and struggle with issues like significant deformation and occlusion. In this work, we introduce inclusion matching, which allows the network to understand the inclusion relationships between segments, rather than relying solely on direct visual correspondences. By integrating this approach with segment parsing and color warping modules, our inclusion matching pipeline significantly improves performance in both keyframe colorization and consecutive frame colorization. To support our network's training, we have developed a unique dataset named PaintBucket-Character, which includes rendered line arts alongside their colorized versions and shading annotations for various 3D characters. To replicate industry animation data formats, we also created color design sheets for each character, with semantic information for each color and standard pose reference images. Experiments highlight the superiority of our method, demonstrating accurate and consistent colorization across both our proposed benchmarks and hand-drawn animations."
                    },
                    {
                        "title": "MIPI 2024 Challenge on Demosaic for Hybridevs Camera: Methods and Results",
                        "abstract": "The rising demand for computational photography on mobile devices drives development of advanced image sensors and algorithms for camera systems. But the lack of opportunities for in-depth exchange between industry and academia are constraining the development of Mobile Intelligent Photography and Imaging (MIPI). Building on the successes of the prior MIPI Workshops at ECCV 2022 and CVPR 2023, we are pleased to introduce our third MIPI challenge, which includes three tracks focusing on novel image sensors and imaging algorithms. In this paper, we summarize and review the Demosaic for Hybridevs Camera track on MIPI 2024. A total of 110 participants from both industrial and academic backgrounds contributed many valuable solutions to address the difficulty of the restoration of HybridEVS\u2019s raw data, thus raising the reconstructed performance to a new height. This paper gives a comprehensive description and analysis of all solutions developed during this challenge. More detailed information about this challenge is available at https://mipi-challenge.org/MIPI2024."
                    },
                    {
                        "title": "LN3Diff: Scalable Latent Neural Fields Diffusion for Speedy 3D Generation",
                        "abstract": "The field of neural rendering has witnessed significant progress with advancements in generative models and differentiable rendering techniques. Though 2D diffusion has achieved success, a unified 3D diffusion pipeline remains unsettled. This paper introduces a novel framework called LN3Diff to address this gap and enable fast, high-quality, and generic conditional 3D generation. Our approach harnesses a 3D-aware architecture and variational autoencoder (VAE) to encode the input image into a structured, compact, and 3D latent space. The latent is decoded by a transformer-based decoder into a high-capacity 3D neural field. Through training a diffusion model on this 3D-aware latent space, our method achieves state-of-the-art performance on ShapeNet for 3D generation and demonstrates superior performance in monocular 3D reconstruction and conditional 3D generation across various datasets. Moreover, it surpasses existing 3D diffusion methods in terms of inference speed, requiring no per-instance optimization. Our proposed LN3Diff presents a significant advancement in 3D generative modeling and holds promise for various applications in 3D vision and graphics tasks."
                    },
                    {
                        "title": "MIPI 2024 Challenge on Nighttime Flare Removal: Methods and Results",
                        "abstract": "The increasing demand for computational photography and imaging on mobile platforms has led to the widespread development and integration of advanced image sensors with novel algorithms in camera systems. However, the scarcity of high-quality data for research and the rare opportunity for in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). Building on the achievements of the previous MIPI Workshops held at ECCV 2022 and CVPR 2023, we introduce our third MIPI challenge including three tracks focusing on novel image sensors and imaging algorithms. In this paper, we summarize and review the Nighttime Flare Removal track on MIPI 2024. In total, 170 participants were successfully registered, and 14 teams submitted results in the final testing phase. The developed solutions in this challenge achieved state-of-the-art performance on Nighttime Flare Removal. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2024."
                    },
                    {
                        "title": "MIPI 2024 Challenge on Few-shot RAW Image Denoising: Methods and Results",
                        "abstract": "The increasing demand for computational photography and imaging on mobile platforms has led to the widespread development and integration of advanced image sensors with novel algorithms in camera systems. However, the scarcity of high-quality data for research and the rare opportunity for in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). Building on the achievements of the previous MIPI Workshops held at ECCV 2022 and CVPR 2023, we introduce our third MIPI challenge including three tracks focusing on novel image sensors and imaging algorithms. In this paper, we summarize and review the Few-shot RAW Image Denoising track on MIPI 2024. In total, 165 participants were successfully registered, and 7 teams submitted results in the final testing phase. The developed solutions in this challenge achieved state-of-the-art performance on Few-shot RAW Image Denoising. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2024."
                    },
                    {
                        "title": "FlexiCurve: Flexible Piecewise Curves Estimation for Photo Retouching",
                        "abstract": "This paper presents a new method, called FlexiCurve, for photo retouching. Unlike most existing methods that perform image-to-image mapping, which requires expensive pixel-wise reconstruction, FlexiCurve takes an input image and estimates global curves to adjust the image. The adjustment curves are specially designed for performing piecewise mapping, taking nonlinear adjustment and differentiability into account. To cope with challenging and diverse properties in real-world photos, FlexiCurve is formulated to produce diverse estimations. The spatial dependencies among these estimations are implicitly modeled by a Transformer structure to improve local retouching of different regions. Thanks to the image-to-curve formulation, FlexiCurve only needs a lightweight network. Our method improves efficiency without compromising the retouching quality and losing details in the original image. The method is also appealing as it is not limited to paired training data, thus it can flexibly learn rich retouching styles from unpaired data. Extensive experiments demonstrate the efficiency, retouching performance, and flexibility of our method quantitatively and qualitatively."
                    },
                    {
                        "title": "Exploiting Diffusion Prior for Real-World Image Super-Resolution",
                        "abstract": "We present a novel approach to leverage prior knowledge encapsulated in pre-trained text-to-image diffusion models for blind super-resolution (SR). Specifically, by employing our time-aware encoder, we can achieve promising restoration results without altering the pre-trained synthesis model, thereby preserving the generative prior and minimizing training cost. To remedy the loss of fidelity caused by the inherent stochasticity of diffusion models, we introduce a controllable feature wrapping module that allows users to balance quality and fidelity by simply adjusting a scalar value during the inference process. Moreover, we develop a progressive aggregation sampling strategy to overcome the fixed-size constraints of pre-trained diffusion models, enabling adaptation to resolutions of any size. A comprehensive evaluation of our method using both synthetic and real-world benchmarks demonstrates its superiority over current state-of-the-art approaches."
                    },
                    {
                        "title": "Flare7K++: Mixing Synthetic and Real Datasets for Nighttime Flare Removal and Beyond",
                        "abstract": "Artificial lights commonly leave strong lens flare artifacts on the images captured at night, degrading both the visual quality and performance of vision algorithms. Existing flare removal approaches mainly focus on removing daytime flares and fail in nighttime cases. Nighttime flare removal is challenging due to the unique luminance and spectrum of artificial lights, as well as the diverse patterns and image degradation of the flares. The scarcity of the nighttime flare removal dataset constrains the research on this crucial task. In this paper, we introduce Flare7K++, the first comprehensive nighttime flare removal dataset, consisting of 962 real-captured flare images (Flare-R) and 7000 synthetic flares (Flare7K). Compared to Flare7K, Flare7K++ is particularly effective in eliminating complicated degradation around the light source, which is intractable by using synthetic flares alone. Besides, the previous flare removal pipeline relies on the manual threshold and blur kernel settings to extract light sources, which may fail when the light sources are tiny or not overexposed. To address this issue, we additionally provide the annotations of light sources in Flare7K++ and propose a new end-to-end pipeline to preserve the light source while removing lens flares. Our dataset and pipeline offer a valuable foundation and benchmark for future investigations into nighttime flare removal studies. Extensive experiments demonstrate that Flare7K++ supplements the diversity of existing flare datasets and pushes the frontier of nighttime flare removal toward real-world scenarios."
                    },
                    {
                        "title": "MIPI 2023 Challenge on Nighttime Flare Removal: Methods and Results",
                        "abstract": "Developing and integrating advanced image sensors with novel algorithms in camera systems are prevalent with the increasing demand for computational photography and imaging on mobile platforms. However, the lack of high-quality data for research and the rare opportunity for in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). With the success of the 1st MIPI Workshop@ECCV 2022, we introduce the second MIPI challenge including four tracks focusing on novel image sensors and imaging algorithms. In this paper, we summarize and review the Nighttime Flare Removal track on MIPI 2023. In total, 120 participants were successfully registered, and 11 teams submitted results in the final testing phase. The developed solutions in this challenge achieved state-of-the-art performance on Nighttime Flare Removal. A detailed description of all models developed in this challenge is provided in this paper. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2023/."
                    },
                    {
                        "title": "MIPI 2023 Challenge on RGBW Fusion: Methods and Results",
                        "abstract": "Developing and integrating advanced image sensors with novel algorithms in camera systems are prevalent with the increasing demand for computational photography and imaging on mobile platforms. However, the lack of high-quality data for research and the rare opportunity for an in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). With the success of the 1st MIPI Workshop@ECCV 2022, we introduce the second MIPI challenge, including four tracks focusing on novel image sensors and imaging algorithms. This paper summarizes and reviews the RGBW Joint Fusion and Denoise track on MIPI 2023. In total, 69 participants were successfully registered, and 4 teams submitted results in the final testing phase. The final results are evaluated using objective metrics, including PSNR, SSIM, LPIPS, and KLD. A detailed description of the models developed in this challenge is provided in this paper. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2023/."
                    },
                    {
                        "title": "PGDiff: Guiding Diffusion Models for Versatile Face Restoration via Partial Guidance",
                        "abstract": "Exploiting pre-trained diffusion models for restoration has recently become a favored alternative to the traditional task-specific training approach. Previous works have achieved noteworthy success by limiting the solution space using explicit degradation models. However, these methods often fall short when faced with complex degradations as they generally cannot be precisely modeled. In this paper, we propose PGDiff by introducing partial guidance, a fresh perspective that is more adaptable to real-world degradations compared to existing works. Rather than specifically defining the degradation process, our approach models the desired properties, such as image structure and color statistics of high-quality images, and applies this guidance during the reverse diffusion process. These properties are readily available and make no assumptions about the degradation process. When combined with a diffusion prior, this partial guidance can deliver appealing results across a range of restoration tasks. Additionally, PGDiff can be extended to handle composite tasks by consolidating multiple high-quality image properties, achieved by integrating the guidance from respective tasks. Experimental results demonstrate that our method not only outperforms existing diffusion-prior-based approaches but also competes favorably with task-specific models."
                    },
                    {
                        "title": "Upscale-A-Video: Temporal-Consistent Diffusion Model for Real-World Video Super-Resolution",
                        "abstract": "Text-based diffusion models have exhibited remarkable success in generation and editing, showing great promise for enhancing visual content with their generative prior. However, applying these models to video super-resolution remains challenging due to the high demands for output fidelity and temporal consistency, which is complicated by the inherent randomness in diffusion models. Our study introduces Upscale-A-Video, a text-guided latent diffusion framework for video upscaling. This framework ensures temporal coherence through two key mechanisms: locally, it integrates temporal layers into U-Net and VAE-Decoder, maintaining consistency within short sequences; globally, without training, a flow-guided recurrent latent propagation module is introduced to enhance overall video stability by propagating and fusing latent across the entire sequences. Thanks to the diffusion paradigm, our model also offers greater flexibility by allowing text prompts to guide texture creation and adjustable noise levels to balance restoration and generation, enabling a trade-off between fidelity and quality. Extensive experiments show that Upscale-A-Video surpasses existing methods in both synthetic and real-world benchmarks, as well as in AI-generated videos, showcasing impressive visual realism and temporal consistency."
                    },
                    {
                        "title": "MIPI 2023 Challenge on RGB+ToF Depth Completion: Methods and Results",
                        "abstract": "Depth completion from RGB images and sparse Time-of-Flight (ToF) measurements is an important problem in computer vision and robotics. While traditional methods for depth completion have relied on stereo vision or structured light techniques, recent advances in deep learning have enabled more accurate and efficient completion of depth maps from RGB images and sparse ToF measurements. To evaluate the performance of different depth completion methods, we organized an RGB+sparse ToF depth completion competition. The competition aimed to encourage research in this area by providing a standardized dataset and evaluation metrics to compare the accuracy of different approaches. In this report, we present the results of the competition and analyze the strengths and weaknesses of the top-performing methods. We also discuss the implications of our findings for future research in RGB+sparse ToF depth completion. We hope that this competition and report will help to advance the state-of-the-art in this important area of research. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2023/."
                    },
                    {
                        "title": "MIPI 2023 Challenge on RGBW Remosaic: Methods and Results",
                        "abstract": "Developing and integrating advanced image sensors with novel algorithms in camera systems are prevalent with the increasing demand for computational photography and imaging on mobile platforms. However, the lack of high-quality data for research and the rare opportunity for an in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). With the success of the 1st MIPI Workshop@ECCV 2022, we introduce the second MIPI challenge, including four tracks focusing on novel image sensors and imaging algorithms. This paper summarizes and reviews the RGBW Joint Remosaic and Denoise track on MIPI 2023. In total, 81 participants were successfully registered, and 4 teams submitted results in the final testing phase. The final results are evaluated using objective metrics, including PSNR, SSIM, LPIPS, and KLD. A detailed description of the top three models developed in this challenge is provided in this paper. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2023/."
                    },
                    {
                        "title": "ProPainter: Improving Propagation and Transformer for Video Inpainting",
                        "abstract": "Flow-based propagation and spatiotemporal Transformer are two mainstream mechanisms in video inpainting (VI). Despite the effectiveness of these components, they still suffer from some limitations that affect their performance. Previous propagation-based approaches are performed separately either in the image or feature domain. Global image propagation isolated from learning may cause spatial misalignment due to inaccurate optical flow. Moreover, memory or computational constraints limit the temporal range of feature propagation and video Transformer, preventing exploration of correspondence information from distant frames. To address these issues, we propose an improved framework, called ProPainter, which involves enhanced ProPagation and an efficient Transformer. Specifically, we introduce dual-domain propagation that combines the advantages of image and feature warping, exploiting global correspondences reliably. We also propose a mask-guided sparse video Transformer, which achieves high efficiency by discarding unnecessary and redundant tokens. With these components, ProPainter outperforms prior arts by a large margin of 1.46 dB in PSNR while maintaining appealing efficiency."
                    },
                    {
                        "title": "Iterative Prompt Learning for Unsupervised Backlit Image Enhancement",
                        "abstract": "We propose a novel unsupervised backlit image enhancement method, abbreviated as CLIP-LIT, by exploring the potential of Contrastive Language-Image Pre-Training (CLIP) for pixel-level image enhancement. We show that the openworld CLIP prior not only aids in distinguishing between backlit and well-lit images, but also in perceiving heterogeneous regions with different luminance, facilitating the optimization of the enhancement network. Unlike high-level and image manipulation tasks, directly applying CLIP to enhancement tasks is non-trivial, owing to the difficulty in finding accurate prompts. To solve this issue, we devise a prompt learning framework that first learns an initial prompt pair by constraining the text-image similarity between the prompt (negative/positive sample) and the corresponding image (backlit image/well-lit image) in the CLIP latent space. Then, we train the enhancement network based on the textimage similarity between the enhanced result and the initial prompt pair. To further improve the accuracy of the initial prompt pair, we iteratively fine-tune the prompt learning framework to reduce the distribution gaps between the backlit images, enhanced results, and well-lit images via rank learning, boosting the enhancement performance. Our method alternates between updating the prompt learning framework and enhancement network until visually pleasing results are achieved. Extensive experiments demonstrate that our method outperforms state-of-the-art methods in terms of visual quality and generalization ability, without requiring any paired data."
                    }
                ]
            },
            "ad8a0d9d-d609-4b04-99a1-20b39d3e12c6": {
                "pk": "ad8a0d9d-d609-4b04-99a1-20b39d3e12c6",
                "name": "Huajun Feng",
                "collaborators": [
                    "Zhi-hai Xu",
                    "Yue-ting Chen",
                    "Qi Li",
                    "Shiqi Chen",
                    "Z. Jin",
                    "Tingting Jiang",
                    "Ziran Zhang",
                    "Hao Xu",
                    "Menghao Li",
                    "Haoying Li",
                    "Jing Wang",
                    "Hai Hu",
                    "Haiquan Hu",
                    "Peng Luo",
                    "Yawen deng",
                    "Ruikang Li",
                    "Shiwei Bao",
                    "Hongyao Chen",
                    "Zida Chen",
                    "Yiming Chen",
                    "Zhehan Song",
                    "Jingwen Zhou",
                    "Tong Li",
                    "Nan Xu",
                    "Jiaqi Yang"
                ],
                "domain": [
                    "Image Processing",
                    "Hyperspectral Imaging",
                    "Computer Vision",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Wavelength encoding spectral imaging based on the combination of deeply learned filters and an RGB camera.",
                        "abstract": "Hyperspectral imaging is a critical tool for gathering spatial-spectral information in various scientific research fields. As a result of improvements in spectral reconstruction algorithms, significant progress has been made in reconstructing hyperspectral images from commonly acquired RGB images. However, due to the limited input, reconstructing spectral information from RGB images is ill-posed. Furthermore, conventional camera color filter arrays (CFA) are designed for human perception and are not optimal for spectral reconstruction. To increase the diversity of wavelength encoding, we propose to place broadband encoding filters in front of the RGB camera. In this condition, the spectral sensitivity of the imaging system is determined by the filters and the camera itself. To achieve an optimal encoding scheme, we use an end-to-end optimization framework to automatically design the filters' transmittance functions and optimize the weights of the spectral reconstruction network. Simulation experiments show that our proposed spectral reconstruction network has excellent spectral mapping capabilities. Additionally, our novel joint wavelength encoding imaging framework is superior to traditional RGB imaging systems. We develop the deeply learned filter and conduct actual shooting experiments. The spectral reconstruction results have an attractive spatial resolution and spectral accuracy."
                    },
                    {
                        "title": "Image restoration using dual-domain fusion network for rotating rectangular synthetic aperture system",
                        "abstract": "Compared to single aperture systems, optical synthetic aperture systems greatly improve the spatial resolution, yet still exhibit a certain degree of blurring and contrast reduction. To address this challenge, numerous image restoration methods have been proposed. Recently, instead of the conventional circular synthetic aperture, the rotating rectangular synthetic aperture (RRSA) system employs a rectangular aperture to capture a sequence of images of the same scene. The RRSA system\u2019s foldable design and absence of common-phase adjustments confer cost and complexity benefits. The captured degraded image sequences contain information about multiple directions of the target scene, so it is necessary to use multi-frame image fusion technology to restore them. However, most conventional methods often introduce visual artifacts and require substantial computational time. In this paper, we propose a Dual-Domain Fusion Network (DDFNet), restoring multi-frame degraded images in the spatial and frequency domain and then achieving superior fusion results. DDFNet employs a nested U-Net architecture to capture local pixel-level relationships, facilitating the recovery of local features and structures from spatial domain images. In parallel, we transform the input images into the frequency domain, and utilize another nested U-Net for feature extraction on the normalized spectrum and phase, thereby improving the recovery of texture and edge information. Finally, the fusion model effectively utilizes multi-level features and contextual awareness to combine the spatial and frequency domain features, achieving high-quality fusion results of captured degraded sequence images. Extensive experiments demonstrate that our method achieves superior performance both in quantitative and qualitative assessments compared to state-of-the-art techniques."
                    },
                    {
                        "title": "Reliable image dehazing by NeRF.",
                        "abstract": "Image dehazing is a typical low-level visual task. With the continuous improvement of network performance and the introduction of various prior knowledge, the ability of image dehazing is becoming stronger. However, the existing dehazing methods have problems such as the inability to obtain real shooting datasets, unreliable dehazing processes, and the difficulty to deal with complex lighting scenes. To solve these problems, we propose a new haze model combining the optical scattering model and the computer graphics rendering. Based on the new haze model, we propose a high-quality and widely applicable dehazing dataset generation pipeline that does not require paired-data training and prior knowledge. We reconstruct the three-dimensional fog space with array camera and remove haze by thresholding voxel deletion. We use the Unreal Engine 5 to generate simulation datasets and the real shooting in laboratory to verify the effectiveness and the reliability of our generation pipeline. Through our pipeline, we can obtain wonderful dehaze results and dehaze datasets under various complex outdoors lighting conditions. We also propose a dehaze dataset enhancement method based on voxel control. Our pipeline and data enhancement are suitable for the latest algorithm model, these solutions can obtain better visual effects and objective indicators."
                    },
                    {
                        "title": "Imaging Chain Modeling and a Scaling Experiment for Optical Remote Sensing of Lunar Surface",
                        "abstract": "Lunar exploration has drawn attention all over the world, and remote sensing of the lunar surface through the satellite\u2019s optical camera is one of the most important ways to provide scientific data. Although on-orbit measurement is the ultimate goal, lunar exploration is a huge cost project, so researchers should take on-ground simulations and experiments to evaluate the performance of the optical instrument and provide basic references before launch. In this article, we propose an optical imaging chain modeling algorithm that focuses on the physical imaging process from light scattering among the 3-D lunar surface to the optical instrument\u2019s output digital value (DV). Compared to other optical signal measurement simulations, we make progress in considering the 3-D structure of the observed scene, the spectral property of the imaging parameters, and corresponding absolute calibration methods. The proposed method is verified in a scaling experiment for lunar optical remote sensing with a calibrated optical camera, simulated moon soil, and a parallel light source. We compare the imaging simulation results with the experiment results for directly illuminated lunar regions, shadowed lunar regions, and multispectral imaging tasks, and the results show average relative errors (REs) around 2.4%, 11.5%, and 4.3% and coefficients of determination around 0.972, 0.942, and 0.915 for the three cases, respectively. Both the modeling algorithm and the scaling experiment will provide basic analysis for optical imaging in lunar remote sensing before the execution of a real project."
                    },
                    {
                        "title": "Deep Linear Array Pushbroom Image Restoration: A Degradation Pipeline and Jitter-Aware Restoration Network",
                        "abstract": "Linear Array Pushbroom (LAP) imaging technology is widely used in the realm of remote sensing. However, images acquired through LAP always suffer from distortion and blur because of camera jitter. Traditional methods for restoring LAP images, such as algorithms estimating the point spread function (PSF), exhibit limited performance. To tackle this issue, we propose a Jitter-Aware Restoration Network (JARNet), to remove the distortion and blur in two stages. In the first stage, we formulate an Optical Flow Correction (OFC) block to refine the optical flow of the degraded LAP images, resulting in pre-corrected images where most of the distortions are alleviated. In the second stage, for further enhancement of the pre-corrected images, we integrate two jitter-aware techniques within the Spatial and Frequency Residual (SFRes) block: 1) introducing Coordinate Attention (CoA) to the SFRes block in order to capture the jitter state in orthogonal direction; 2) manipulating image features in both spatial and frequency domains to leverage local and global priors. Additionally, we develop a data synthesis pipeline, which applies Continue Dynamic Shooting Model (CDSM) to simulate realistic degradation in LAP images. Both the proposed JARNet and LAP image synthesis pipeline establish a foundation for addressing this intricate challenge. Extensive experiments demonstrate that the proposed two-stage method outperforms state-of-the-art image restoration models. Code is available at https://github.com/JHW2000/JARNet."
                    },
                    {
                        "title": "Nighttime Image Dehazing by Render",
                        "abstract": "Nighttime image dehazing presents unique challenges due to the unevenly distributed haze caused by the color change of artificial light sources. This results in multiple interferences, including atmospheric light, glow, and direct light, which make the complex scattering haze interference difficult to accurately distinguish and remove. Additionally, obtaining pairs of high-definition data for fog removal at night is a difficult task. These challenges make nighttime image dehazing a particularly challenging problem to solve. To address these challenges, we introduced the haze scattering formula to more accurately express the haze in three-dimensional space. We also proposed a novel data synthesis method using the latest CG textures and lumen lighting technology to build scenes where various hazes can be seen clearly through ray tracing. We converted the complex 3D scattering relationship transformation into a 2D image dataset to better learn the mapping from 3D haze to 2D haze. Additionally, we improved the existing neural network and established a night haze intensity evaluation label based on the idea of optical PSF. This allowed us to adjust the haze intensity of the rendered dataset according to the intensity of the real haze image and improve the accuracy of dehazing. Our experiments showed that our data construction and network improvement achieved better visual effects, objective indicators, and calculation speed."
                    },
                    {
                        "title": "Effective framework for space target detection through atmospheric turbulence",
                        "abstract": "Atmospheric turbulence is a major challenge in long-range imaging of ground-based telescopes, especially in the surveillance of space targets, whose observation distance is usually more than 100 km. In this case, space targets are extremely small in images, occupying less than 0.12% of the total image area, and suffer from severe blur and distortion. Consequently, the accuracy of object detection by both conventional and deep-learning-based methods is significantly hampered. Therefore, this paper proposes an effective framework for detecting space target through atmospheric turbulence. The framework incorporates a shallow deblurring module, a transformer-based feature extractor, and a small region proposal network. The training data comprises simulated degraded images of space target images against celestial backgrounds, as well as a selection of images from the Dotav2 dataset. Testing results show that the proposed framework outperforms the general framework, achieving a mean Average Precision (mAP) improvement of over 20%."
                    },
                    {
                        "title": "DR-UNet: dynamic residual U-Net for blind correction of optical degradation",
                        "abstract": "Due to the size limitation of mobile devices, their optical design is difficult to reach the level of professional equipment. Corresponding restoration methods are then needed to compensate for the shortage. However, most of the models are still static, which leads to their limited representation ability of images. To tackle this problem, we propose a plug-and-play deformable residual block for efficiently sampling the spatially related features at different scales. Moreover, considering that the optical degradation is closely correlated with the field-of-view (FOV), we introduce a FOV attention block based on omni-dimensional dynamic convolution to integrate spatial features. On this basis, we further propose a novel optical degradation correction model called DR-UNet. It is constructed on an encoder-decoder structure to capture multiscale information, along with several context blocks. By correcting the optical degradation in images from coarse to fine, we finally obtain high-quality and degradation-free images. Extensive results demonstrate that our method can compete favorably with some state-of-the-art methods."
                    },
                    {
                        "title": "A Data Generation Method for Image Flare Removal Based on Similarity and Centrosymmetric Effect",
                        "abstract": "Image pairs in under-illuminated scenes along with the presence of complex light sources often result in strong flare artifacts in images, affecting both image quality and the performance of downstream visual applications. Removing lens flare and ghosts is a challenging issue, particularly in low-light environments. Existing methods for flare removal are mainly restricted by inadequate simulation and real-world capture, resulting in singular categories of scattered flares and unavailable reflected ghosts. Therefore, a comprehensive deterioration procedure is crucial for generating a dataset for flare removal. We propose a methodology based on spatial position relationships for generating data pairs with flare deterioration, which is supported by theoretical analysis and real-world evaluation. Our procedure is comprehensive and realizes the similarity of scattered flares and the symmetric effect of reflected ghosts. We also construct a real-shot pipeline that respectively processes the effects of scattering and reflective flares, aiming to directly generate data for end-to-end methods. Experimental results demonstrate that our methodology adds diversity to existing flare datasets and constructs a comprehensive mapping procedure for flare data pairs. Our method facilitates the data-driven model to achieve better restoration in flare images and proposes a better evaluation system based on real shots, thus promoting progress in the area of real flare removal."
                    },
                    {
                        "title": "Let Segment Anything Help Image Dehaze",
                        "abstract": "The large language model and high-level vision model have achieved impressive performance improvements with large datasets and model sizes. However, low-level computer vision tasks, such as image dehaze and blur removal, still rely on a small number of datasets and small-sized models, which generally leads to overfitting and local optima. Therefore, we propose a framework to integrate large-model prior into low-level computer vision tasks. Just as with the task of image segmentation, the degradation of haze is also texture-related. So we propose to detect gray-scale coding, network channel expansion, and pre-dehaze structures to integrate large-model prior knowledge into any low-level dehazing network. We demonstrate the effectiveness and applicability of large models in guiding low-level visual tasks through different datasets and algorithms comparison experiments. Finally, we demonstrate the effect of grayscale coding, network channel expansion, and recurrent network structures through ablation experiments. Under the conditions where additional data and training resources are not required, we successfully prove that the integration of large-model prior knowledge will improve the dehaze performance and save training time for low-level visual tasks."
                    },
                    {
                        "title": "Elimination of varying chromatic aberrations based on diffractive optics.",
                        "abstract": "In telescopic systems consisting of Alvarez lenses, chromatic aberrations vary with the magnifications and the fields of view. Computational imaging has developed rapidly in recent years, therefore we propose a method of optimizing the DOE and the post-processing neural network in 2 stages for achromatic aberrations. We apply the iterative algorithm and the gradient descent method to optimize the DOE, respectively, and then adopt U-Net to further optimize the results. The results show that the optimized DOEs improve the results, the gradient descent optimized DOE with U-Net performs the best and has a very robust and good performance in the case of simulated chromatic aberrations. The results also verify the validity of our algorithm."
                    },
                    {
                        "title": "Snapshot hyperspectral imaging based on equalization designed DOE.",
                        "abstract": "Hyperspectral imaging attempts to determine distinctive information in spatial and spectral domain of a target. Over the past few years, hyperspectral imaging systems have developed towards lighter and faster. In phase-coded hyperspectral imaging systems, a better coding aperture design can improve the spectral accuracy relatively. Using wave optics, we post an equalization designed phase-coded aperture to achieve desired equalization point spread functions (PSFs) which provides richer features for subsequent image reconstruction. During the reconstruction of images, our raised hyperspectral reconstruction network, CAFormer, achieves better results than the state-of-the-art networks with less computation by substituting self-attention with channel-attention. Our work revolves around the equalization design of the phase-coded aperture and optimizes the imaging process from three aspects: hardware design, reconstruction algorithm, and PSF calibration. Our work is putting snapshot compact hyperspectral technology closer to a practical application."
                    },
                    {
                        "title": "Toward Real Flare Removal: A Comprehensive Pipeline and A New Benchmark",
                        "abstract": "Photographing in the under-illuminated scenes, the presence of complex light sources often leave strong flare artifacts in images, where the intensity, the spectrum, the reflection, and the aberration altogether contribute the deterioration. Besides the image quality, it also influence the performance of down-stream visual applications. Thus, removing the lens flare and ghosts is a challenge issue especially in low-light environment. However, existing methods for flare removal mainly restricted to the problems of inadequate simulation and real-world capture, where the categories of scattered flares are singular and the reflected ghosts are unavailable. Therefore, a comprehensive deterioration procedure is crucial for constructing the dataset of flare removal. Based on the theoretical analysis and real-world evaluation, we propose a well-developed methodology for generating the data-pairs with flare deterioration. The procedure is comprehensive, where the similarity of scattered flares and the symmetric effect of reflected ghosts are realized. Moreover, we also construct a real-shot pipeline that respectively processes the effects of scattering and reflective flares, aiming to directly generate the data for end-to-end methods. Experimental results show that the proposed methodology add diversity to the existing flare datasets and construct a comprehensive mapping procedure for flare data pairs. And our method facilities the data-driven model to realize better restoration in flare images and proposes a better evaluation system based on real shots, resulting promote progress in the area of real flare removal."
                    },
                    {
                        "title": "A Data Enhancement Method toward Real-World Single Image Super-Resolution",
                        "abstract": "Introducing convolutional neural networks (CNNs) into single image super-resolution (SISR) has achieved remarkable performance in experimental scenarios over the past decades. However, maintaining constant mapping complexity, most existing CNN-based SISR approaches make it challenging to reconstruct high-quality images in real scenarios because of considerable variation between authentic degradation and simulated degradation. In this paper, to improve SISR models\u2019 capabilities in the real world, we propose a data enhancement method, consisting of frequency domain data augmentation (FDA) and dynamical frequency loss (DFL). By introducing high-resolution images\u2019 frequency information, the FDA enhances the quality and quantity of data, regularizing the model to learn \u201chow\u201d to appropriately apply super-resolution on different frequency components of a given image. In addition, to further strengthen the regularization effect, we design DFL to dynamically assign the loss trade-off hyperparameters of distinct frequency components, guiding the model to allocate computational overhead reasonably. Considering quantitative analysis and visual quality in combination, extensive experiments prove that the proposed enhancement method significantly improves SR models\u2019 capabilities in real scenarios without adding any computational cost to the evaluation procedure."
                    }
                ]
            },
            "08a2e9bd-0119-4ec3-b180-089636639e78": {
                "pk": "08a2e9bd-0119-4ec3-b180-089636639e78",
                "name": "Chongyi Li",
                "collaborators": [
                    "Chunle Guo",
                    "Ruicheng Feng",
                    "Chen Change Loy",
                    "Shangchen Zhou",
                    "Man Zhou",
                    "Jie Huang",
                    "Jinwei Gu",
                    "Yu Dai",
                    "Qingpeng Zhu",
                    "Qi Sun",
                    "Wenxiu Sun",
                    "Naishan Zheng",
                    "Xin Jin",
                    "Linghao Han",
                    "Fengmei Zhao",
                    "Qingyu Yang",
                    "Qiming Ai",
                    "X. Lin",
                    "Jingtong Yue",
                    "Chao Ren",
                    "Z. Li",
                    "Z. Chai",
                    "Yuyan Zhou",
                    "Dong Liang",
                    "Songcan Chen",
                    "Shengjian Huang",
                    "Shuo Yang",
                    "Jianqiang Xiao",
                    "Ruixun Zhang",
                    "Xialei Liu",
                    "D. Hong",
                    "J. Chanussot",
                    "Yihang Luo",
                    "Xiangyu Rui",
                    "Deyu Meng",
                    "Jinghao Zhang",
                    "Ruixia Wu",
                    "Zheng-Peng Duan",
                    "Zhi Chai",
                    "Yi Yu",
                    "Yang Huang",
                    "Kang Zhang",
                    "Meiya Chen",
                    "Yu Wang",
                    "Yong-Chao Li",
                    "Hao Jiang",
                    "A. Muduli",
                    "Vikash Kumar",
                    "Kunal Swami",
                    "Pankaj Bajpai",
                    "Yunchao Ma",
                    "Jiajun Xiao",
                    "Zhi Ling",
                    "Huaijin Chen",
                    "Shuai Li",
                    "Wenyi Zhao",
                    "W. Zhang",
                    "Lu Yang",
                    "Peixian Zhuang",
                    "Lingqiao Li",
                    "Kefeng Fan",
                    "Huihua Yang"
                ],
                "domain": [
                    "Image Restoration",
                    "Computational Photography",
                    "Deep Learning",
                    "Self-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "FlexiCurve: Flexible Piecewise Curves Estimation for Photo Retouching",
                        "abstract": "This paper presents a new method, called FlexiCurve, for photo retouching. Unlike most existing methods that perform image-to-image mapping, which requires expensive pixel-wise reconstruction, FlexiCurve takes an input image and estimates global curves to adjust the image. The adjustment curves are specially designed for performing piecewise mapping, taking nonlinear adjustment and differentiability into account. To cope with challenging and diverse properties in real-world photos, FlexiCurve is formulated to produce diverse estimations. The spatial dependencies among these estimations are implicitly modeled by a Transformer structure to improve local retouching of different regions. Thanks to the image-to-curve formulation, FlexiCurve only needs a lightweight network. Our method improves efficiency without compromising the retouching quality and losing details in the original image. The method is also appealing as it is not limited to paired training data, thus it can flexibly learn rich retouching styles from unpaired data. Extensive experiments demonstrate the efficiency, retouching performance, and flexibility of our method quantitatively and qualitatively."
                    },
                    {
                        "title": "Unlocking Low-Light-Rainy Image Restoration by Pairwise Degradation Feature Vector Guidance",
                        "abstract": "Rain in the dark is a common natural phenomenon. Photos captured in such a condition significantly impact the performance of various nighttime activities, such as autonomous driving, surveillance systems"
                    },
                    {
                        "title": "DNF: Decouple and Feedback Network for Seeing in the Dark",
                        "abstract": "The exclusive properties of RAW data have shown great potential for low-light image enhancement. Nevertheless, the performance is bottlenecked by the inherent limitations of existing architectures in both single-stage and multi-stage methods. Mixed mapping across two different domains, noise-to-clean and RAW-to-sRGB, misleads the single-stage methods due to the domain ambiguity. The multi-stage methods propagate the information merely through the resulting image of each stage, neglecting the abundant features in the lossy image-level dataflow. In this paper, we probe a generalized solution to these bottlenecks and propose a Decouple aNd Feedback framework, abbreviated as DNF. To mitigate the domain ambiguity, domain-specific subtasks are decoupled, along with fully utilizing the unique properties in RAW and sRGB domains. The feature propagation across stages with a feedback mechanism avoids the information loss caused by image-level dataflow. The two key insights of our method resolve the inherent limitations of RAW data-based low-light image enhancement satisfactorily, empowering our method to outperform the previous state-of-the-art method by a large margin with only 19% parameters, achieving 0.97dB and 1.30dB PSNR improvements on the Sony and Fuji subsets of SID."
                    },
                    {
                        "title": "Improving Lens Flare Removal with General-Purpose Pipeline and Multiple Light Sources Recovery",
                        "abstract": "When taking images against strong light sources, the resulting images often contain heterogeneous flare artifacts. These artifacts can importantly affect image visual quality and downstream computer vision tasks. While collecting real data pairs of flare-corrupted/flare-free images for training flare removal models is challenging, current methods utilize the direct-add approach to synthesize data. However, these methods do not consider automatic exposure and tone mapping in image signal processing pipeline (ISP), leading to the limited generalization capability of deep models training using such data. Besides, existing methods struggle to handle multiple light sources due to the different sizes, shapes and illuminance of various light sources. In this paper, we propose a solution to improve the performance of lens flare removal by revisiting the ISP and remodeling the principle of automatic exposure in the synthesis pipeline and design a more reliable light sources recovery strategy. The new pipeline approaches realistic imaging by discriminating the local and global illumination through convex combination, avoiding global illumination shifting and local over-saturation. Our strategy for recovering multiple light sources convexly averages the input and output of the neural network based on illuminance levels, thereby avoiding the need for a hard threshold in identifying light sources. We also contribute a new flare removal testing dataset containing the flare-corrupted images captured by ten types of consumer electronics. The dataset facilitates the verification of the generalization capability of flare removal methods. Extensive experiments show that our solution can effectively improve the performance of lens flare removal and push the frontier toward more general situations."
                    },
                    {
                        "title": "Make Explicit Calibration Implicit: Calibrate Denoiser Instead of the Noise Model",
                        "abstract": "Explicit calibration-based methods have dominated RAW image denoising under extremely low-light environments. However, these methods are impeded by several critical limitations: a) the explicit calibration process is both labor- and time-intensive, b) challenge exists in transferring denoisers across different camera models, and c) the disparity between synthetic and real noise is exacerbated by digital gain. To address these issues, we introduce a groundbreaking pipeline named Lighting Every Darkness (LED), which is effective regardless of the digital gain or the camera sensor. LED eliminates the need for explicit noise model calibration, instead utilizing an implicit fine-tuning process that allows quick deployment and requires minimal data. Structural modifications are also included to reduce the discrepancy between synthetic and real noise without extra computational demands. Our method surpasses existing methods in various camera models, including new ones not in public datasets, with just a few pairs per digital gain and only 0.5% of the typical iterations. Furthermore, LED also allows researchers to focus more on deep learning advancements while still utilizing sensor engineering benefits. Code and related materials can be found in https://srameo.github.io/projects/led-iccv23/ ."
                    },
                    {
                        "title": "Rethinking Pan-Sharpening in Closed-Loop Regularization",
                        "abstract": "It is generally known that pan-sharpening is fundamentally a PAN-guided multispectral (MS) image super-resolution problem that involves learning the nonlinear mapping from low-resolution (LR) to high-resolution (HR) MS images. Since an infinite number of HR-MS images can be downsampled to produce the same corresponding LR-MS image, learning the mapping from LR-MS to HR-MS image is typically ill-posed and the space of the possible pan-sharpening functions can be extremely large, making it difficult to estimate the optimal mapping solution. To address the above issue, we propose a closed-loop scheme that learns the two opposite mapping including the pan-sharpening and its corresponding degradation process simultaneously to regularize the solution space in a single pipeline. More specifically, an invertible neural network (INN) is introduced to perform a bidirectional closed-loop: the forward operation for LR-MS pan-sharpening and the backward operation for learning the corresponding HR-MS image degradation process. In addition, given the vital importance of high-frequency textures for the Pan-sharpened MS images, we further strengthen the INN by designing a specified multiscale high-frequency texture extraction module. Extensive experimental results demonstrate that the proposed algorithm performs favorably against state-of-the-art methods qualitatively and quantitatively with fewer parameters. Ablation studies also verify the effectiveness of the closed-loop mechanism in pan-sharpening. The source code is made publicly available at https://github.com/manman1995/pan-sharpening-Team-zhouman/."
                    },
                    {
                        "title": "Fourmer: An Efficient Global Modeling Paradigm for Image Restoration",
                        "abstract": "Global modeling-based image restoration frameworks have become popular. However, they often require a high memory footprint and do not consider task-specific degradation. Our work presents an alternative approach to global modeling that is more efficient for image restoration. The key insights which motivate our study are two-fold: 1) Fourier transform is capable of disentangling image degradation and content component to a certain extent, serving as the image degradation prior, and 2) Fourier domain innately embraces global properties, where each pixel in the Fourier space is involved with all spatial pixels. While adhering to the \u201cspatial interaction + channel evo-lution\u201d rule of previous studies, we customize the core designs with Fourier spatial interaction modeling and Fourier channel evolution. Our paradigm, Fourmer, achieves competitive performance on common image restoration tasks such as image de-raining, image enhancement, image dehazing, and guided image super-resolution, while requiring fewer computational resources. The code for Fourmer is publicly available at https://manman1995.github.io/ ."
                    },
                    {
                        "title": "Flare7K++: Mixing Synthetic and Real Datasets for Nighttime Flare Removal and Beyond",
                        "abstract": "Artificial lights commonly leave strong lens flare artifacts on the images captured at night, degrading both the visual quality and performance of vision algorithms. Existing flare removal approaches mainly focus on removing daytime flares and fail in nighttime cases. Nighttime flare removal is challenging due to the unique luminance and spectrum of artificial lights, as well as the diverse patterns and image degradation of the flares. The scarcity of the nighttime flare removal dataset constrains the research on this crucial task. In this paper, we introduce Flare7K++, the first comprehensive nighttime flare removal dataset, consisting of 962 real-captured flare images (Flare-R) and 7000 synthetic flares (Flare7K). Compared to Flare7K, Flare7K++ is particularly effective in eliminating complicated degradation around the light source, which is intractable by using synthetic flares alone. Besides, the previous flare removal pipeline relies on the manual threshold and blur kernel settings to extract light sources, which may fail when the light sources are tiny or not overexposed. To address this issue, we additionally provide the annotations of light sources in Flare7K++ and propose a new end-to-end pipeline to preserve the light source while removing lens flares. Our dataset and pipeline offer a valuable foundation and benchmark for future investigations into nighttime flare removal studies. Extensive experiments demonstrate that Flare7K++ supplements the diversity of existing flare datasets and pushes the frontier of nighttime flare removal toward real-world scenarios."
                    },
                    {
                        "title": "MIPI 2023 Challenge on Nighttime Flare Removal: Methods and Results",
                        "abstract": "Developing and integrating advanced image sensors with novel algorithms in camera systems are prevalent with the increasing demand for computational photography and imaging on mobile platforms. However, the lack of high-quality data for research and the rare opportunity for in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). With the success of the 1st MIPI Workshop@ECCV 2022, we introduce the second MIPI challenge including four tracks focusing on novel image sensors and imaging algorithms. In this paper, we summarize and review the Nighttime Flare Removal track on MIPI 2023. In total, 120 participants were successfully registered, and 11 teams submitted results in the final testing phase. The developed solutions in this challenge achieved state-of-the-art performance on Nighttime Flare Removal. A detailed description of all models developed in this challenge is provided in this paper. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2023/."
                    },
                    {
                        "title": "MIPI 2023 Challenge on RGBW Fusion: Methods and Results",
                        "abstract": "Developing and integrating advanced image sensors with novel algorithms in camera systems are prevalent with the increasing demand for computational photography and imaging on mobile platforms. However, the lack of high-quality data for research and the rare opportunity for an in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). With the success of the 1st MIPI Workshop@ECCV 2022, we introduce the second MIPI challenge, including four tracks focusing on novel image sensors and imaging algorithms. This paper summarizes and reviews the RGBW Joint Fusion and Denoise track on MIPI 2023. In total, 69 participants were successfully registered, and 4 teams submitted results in the final testing phase. The final results are evaluated using objective metrics, including PSNR, SSIM, LPIPS, and KLD. A detailed description of the models developed in this challenge is provided in this paper. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2023/."
                    },
                    {
                        "title": "Random Weights Networks Work as Loss Prior Constraint for Image Restoration",
                        "abstract": "In this paper, orthogonal to the existing data and model studies, we instead resort our efforts to investigate the potential of loss function in a new perspective and present our belief ``Random Weights Networks can Be Acted as Loss Prior Constraint for Image Restoration''. Inspired by Functional theory, we provide several alternative solutions to implement our belief in the strict mathematical manifolds including Taylor's Unfolding Network, Invertible Neural Network, Central Difference Convolution and Zero-order Filtering as ``random weights network prototype'' with respect of the following four levels: 1) the different random weights strategies; 2) the different network architectures, \\emph{eg,} pure convolution layer or transformer; 3) the different network architecture depths; 4) the different numbers of random weights network combination. Furthermore, to enlarge the capability of the randomly initialized manifolds, we devise the manner of random weights in the following two variants: 1) the weights are randomly initialized only once during the whole training procedure; 2) the weights are randomly initialized at each training iteration epoch. Our propose belief can be directly inserted into existing networks without any training and testing computational cost. Extensive experiments across multiple image restoration tasks, including image de-noising, low-light image enhancement, guided image super-resolution demonstrate the consistent performance gains obtained by introducing our belief. To emphasize, our main focus is to spark the realms of loss function and save their current neglected status. Code will be publicly available."
                    },
                    {
                        "title": "Decomposition Ascribed Synergistic Learning for Unified Image Restoration",
                        "abstract": "Learning to restore multiple image degradations within a single model is quite beneficial for real-world applications. Nevertheless, existing works typically concentrate on regarding each degradation independently, while their relationship has been less exploited to ensure the synergistic learning. To this end, we revisit the diverse degradations through the lens of singular value decomposition, with the observation that the decomposed singular vectors and singular values naturally undertake the different types of degradation information, dividing various restoration tasks into two groups, \\ie, singular vector dominated and singular value dominated. The above analysis renders a more unified perspective to ascribe the diverse degradations, compared to previous task-level independent learning. The dedicated optimization of degraded singular vectors and singular values inherently utilizes the potential relationship among diverse restoration tasks, attributing to the Decomposition Ascribed Synergistic Learning (DASL). Specifically, DASL comprises two effective operators, namely, Singular VEctor Operator (SVEO) and Singular VAlue Operator (SVAO), to favor the decomposed optimization, which can be lightly integrated into existing image restoration backbone. Moreover, the congruous decomposition loss has been devised for auxiliary. Extensive experiments on blended five image restoration tasks demonstrate the effectiveness of our method."
                    },
                    {
                        "title": "RIDCP: Revitalizing Real Image Dehazing via High-Quality Codebook Priors",
                        "abstract": "Existing dehazing approaches struggle to process real-world hazy images owing to the lack of paired real data and robust priors. In this work, we present a new paradigm for real image dehazing from the perspectives of synthesizing more realistic hazy data and introducing more robust priors into the network. Specifically, (1) instead of adopting the de facto physical scattering model, we rethink the degradation of real hazy images and propose a phenomenological pipeline considering diverse degradation types. (2) We propose a Real Image Dehazing network via high-quality Codebook Priors (RIDCP). Firstly, a VQGAN is pre-trained on a large-scale high-quality dataset to obtain the discrete codebook, encapsulating high-quality priors (HQPs). After replacing the negative effects brought by haze with HQPs, the decoder equipped with a novel normalized feature alignment module can effectively utilize high-quality features and produce clean results. However, although our degradation pipeline drastically mitigates the domain gap between synthetic and real data, it is still intractable to avoid it, which challenges HQPs matching in the wild. Thus, we recalculate the distance when matching the features to the HQPs by a controllable matching operation, which facilitates finding better counterparts. We provide a recommendation to control the matching based on an explainable solution. Users can also flexibly adjust the enhancement degree as per their preference. Extensive experiments verify the effectiveness of our data synthesis pipeline and the superior performance of RIDCP in real image dehazing. Code and data are released at https://rqwu.github.io/projects/RIDCP."
                    },
                    {
                        "title": "MIPI 2023 Challenge on RGB+ToF Depth Completion: Methods and Results",
                        "abstract": "Depth completion from RGB images and sparse Time-of-Flight (ToF) measurements is an important problem in computer vision and robotics. While traditional methods for depth completion have relied on stereo vision or structured light techniques, recent advances in deep learning have enabled more accurate and efficient completion of depth maps from RGB images and sparse ToF measurements. To evaluate the performance of different depth completion methods, we organized an RGB+sparse ToF depth completion competition. The competition aimed to encourage research in this area by providing a standardized dataset and evaluation metrics to compare the accuracy of different approaches. In this report, we present the results of the competition and analyze the strengths and weaknesses of the top-performing methods. We also discuss the implications of our findings for future research in RGB+sparse ToF depth completion. We hope that this competition and report will help to advance the state-of-the-art in this important area of research. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2023/."
                    },
                    {
                        "title": "Generating Aligned Pseudo-Supervision from Non-Aligned Data for Image Restoration in Under-Display Camera",
                        "abstract": "Due to the difficulty in collecting large-scale and perfectly aligned paired training data for Under-Display Camera (UDC) image restoration, previous methods resort to monitor-based image systems or simulation-based methods, sacrificing the realness of the data and introducing domain gaps. In this work, we revisit the classic stereo setup for training data collection \u2013 capturing two images of the same scene with one UDC and one standard camera. The key idea is to \u201ccopy\u201d details from a high-quality reference image and \u201cpaste\u201d them on the UDC image. While being able to generate real training pairs, this setting is susceptible to spatial misalignment due to perspective and depth of field changes. The problem is further compounded by the large domain discrepancy between the UDC and normal images, which is unique to UDC restoration. In this paper, we mitigate the non-trivial domain discrepancy and spatial misalignment through a novel Transformer-based framework that generates well-aligned yet high-quality target data for the corresponding UDC input. This is made possible through two carefully designed components, namely, the Domain Alignment Module (DAM) and Geometric Alignment Module (GAM), which encourage robust and accurate discovery of correspondence between the UDC and normal views. Extensive experiments show that high-quality and well-aligned pseudo UDC training pairs are beneficial for training a robust restoration network. Code and the dataset are available at https://github.com/jnjaby/AlignFormer."
                    },
                    {
                        "title": "Unlocking Masked Autoencoders as Loss Function for Image and Video Restoration",
                        "abstract": "Image and video restoration has achieved a remarkable leap with the advent of deep learning. The success of deep learning paradigm lies in three key components: data, model, and loss. Currently, many efforts have been devoted to the first two while seldom study focuses on loss function. With the question ``are the de facto optimization functions e.g., $L_1$, $L_2$, and perceptual losses optimal?'', we explore the potential of loss and raise our belief ``learned loss function empowers the learning capability of neural networks for image and video restoration''. Concretely, we stand on the shoulders of the masked Autoencoders (MAE) and formulate it as a `learned loss function', owing to the fact the pre-trained MAE innately inherits the prior of image reasoning. We investigate the efficacy of our belief from three perspectives: 1) from task-customized MAE to native MAE, 2) from image task to video task, and 3) from transformer structure to convolution neural network structure. Extensive experiments across multiple image and video tasks, including image denoising, image super-resolution, image enhancement, guided image super-resolution, video denoising, and video enhancement, demonstrate the consistent performance improvements introduced by the learned loss function. Besides, the learned loss function is preferable as it can be directly plugged into existing networks during training without involving computations in the inference stage. Code will be publicly available."
                    },
                    {
                        "title": "MIPI 2023 Challenge on RGBW Remosaic: Methods and Results",
                        "abstract": "Developing and integrating advanced image sensors with novel algorithms in camera systems are prevalent with the increasing demand for computational photography and imaging on mobile platforms. However, the lack of high-quality data for research and the rare opportunity for an in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). With the success of the 1st MIPI Workshop@ECCV 2022, we introduce the second MIPI challenge, including four tracks focusing on novel image sensors and imaging algorithms. This paper summarizes and reviews the RGBW Joint Remosaic and Denoise track on MIPI 2023. In total, 81 participants were successfully registered, and 4 teams submitted results in the final testing phase. The final results are evaluated using objective metrics, including PSNR, SSIM, LPIPS, and KLD. A detailed description of the top three models developed in this challenge is provided in this paper. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2023/."
                    },
                    {
                        "title": "Learned Image Reasoning Prior Penetrates Deep Unfolding Network for Panchromatic and Multi-Spectral Image Fusion",
                        "abstract": "The success of deep neural networks for pan-sharpening is commonly in a form of black box, lacking transparency and interpretability. To alleviate this issue, we propose a novel model-driven deep unfolding framework with image reasoning prior tailored for the pan-sharpening task. Different from existing unfolding solutions that deliver the proximal operator networks as the uncertain and vague priors, our framework is motivated by the content reasoning ability of masked autoencoders (MAE) with insightful designs. Specifically, the pre-trained MAE with spatial masking strategy, acting as intrinsic reasoning prior, is embedded into unfolding architecture. Meanwhile, the pre-trained MAE with spatial-spectral masking strategy is treated as the regularization term within loss function to constrain the spatial-spectral consistency. Such designs penetrate the image reasoning prior into deep unfolding networks while improving its interpretability and representation capability. The uniqueness of our framework is that the holistic learning process is explicitly integrated with the inherent physical mechanism underlying the pan-sharpening task. Extensive experiments on multiple satellite datasets demonstrate the superiority of our method over the existing state-of-the-art approaches. Code will be released at https://manman1995.github.io/."
                    },
                    {
                        "title": "Embedding Global Contrastive and Local Location in Self-Supervised Learning",
                        "abstract": "Self-supervised representation learning (SSL) typically suffers from inadequate data utilization and feature-specificity due to the suboptimal sampling strategy and the monotonous optimization method. Existing contrastive-based methods alleviate these issues through exceedingly long training time and large batch size, resulting in non-negligible computational consumption and memory usage. In this paper, we present an efficient self-supervised framework, called GLNet. The key insights of this work are the novel sampling and ensemble learning strategies embedded in the self-supervised framework. We first propose a location-based sampling strategy to integrate the complementary advantages of semantic and spatial characteristics. Whereafter, a Siamese network with momentum update is introduced to generate representative vectors, which are used to optimize the feature extractor. Finally, we particularly embed global contrastive and local location tasks in the framework, which aims to leverage the complementarity between the high-level semantic features and low-level texture features. Such complementarity is significant for mitigating the feature-specificity and improving the generalizability, thus effectively improving the performance of downstream tasks. Extensive experiments on representative benchmark datasets demonstrate that GLNet performs favorably against the state-of-the-art SSL methods. Specifically, GLNet improves MoCo-v3 by 2.4% accuracy on ImageNet dataset, while improves 2% accuracy and consumes only 75% training time on the ImageNet-100 dataset. In addition, GLNet is appealing in its compatibility with popular SSL frameworks. Code is available at GLNet."
                    }
                ]
            },
            "306389d2-a17a-4b70-89f7-2c9a262c3862": {
                "pk": "306389d2-a17a-4b70-89f7-2c9a262c3862",
                "name": "Chen Change Loy",
                "collaborators": [
                    "Ziwei Liu",
                    "Shangchen Zhou",
                    "Chongyi Li",
                    "Ruicheng Feng",
                    "Shuai Yang",
                    "Liming Jiang",
                    "Yu Dai",
                    "Wei Li",
                    "Bo Dai",
                    "Quan Wang",
                    "Chen Qian",
                    "Lei Yang",
                    "Wayne Wu",
                    "Fangzhou Hong",
                    "Qingpeng Zhu",
                    "Qi Sun",
                    "Wenxiu Sun",
                    "Jinwei Gu",
                    "Chunle Guo",
                    "Qiming Ai",
                    "Ze-Sen Chen",
                    "Gengshi Huang",
                    "Jianing Teng",
                    "Kun Wang",
                    "Jing Shao",
                    "Lu Sheng",
                    "Henghui Ding",
                    "Chang Liu",
                    "Shuting He",
                    "Xudong Jiang",
                    "Yifan Zhou",
                    "Quanzhou Li",
                    "Jingbo Wang",
                    "Jianyi Wang",
                    "Zongsheng Yue",
                    "Kelvin C. K. Chan",
                    "Lian Siyao",
                    "Weijiang Yu",
                    "Tianpei Gu",
                    "Chunze Lin",
                    "Guangcong Wang",
                    "Zhaoxi Chen",
                    "Dongwei Pan",
                    "Long Zhuo",
                    "Jingtan Piao",
                    "Huiwen Luo",
                    "W. Cheng",
                    "Yuxin Wang",
                    "Siming Fan",
                    "Shengqi Liu",
                    "Dahua Lin",
                    "Kwan-Yee Lin",
                    "Jianhui Yu",
                    "Hao Zhu",
                    "Weidong (Tom) Cai",
                    "Junzhe Zhang",
                    "Yushi Lan",
                    "C. Yeo",
                    "Zhongang Cai",
                    "Liang Pan",
                    "Chen Wei",
                    "Wanqi Yin",
                    "Mingyuan Zhang",
                    "Yihang Luo",
                    "Zhongbin Fang",
                    "Xiangtai Li",
                    "Xia Li",
                    "J. Buhmann",
                    "Mengyuan Liu",
                    "Yuhang Zang",
                    "Jun Han",
                    "Kaiyang Zhou",
                    "Qingyu Yang",
                    "Peiqing Yang",
                    "Qingyi Tao",
                    "Yuhan Wang"
                ],
                "domain": [
                    "Computer Vision",
                    "Image Processing",
                    "Video Generation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "FlexiCurve: Flexible Piecewise Curves Estimation for Photo Retouching",
                        "abstract": "This paper presents a new method, called FlexiCurve, for photo retouching. Unlike most existing methods that perform image-to-image mapping, which requires expensive pixel-wise reconstruction, FlexiCurve takes an input image and estimates global curves to adjust the image. The adjustment curves are specially designed for performing piecewise mapping, taking nonlinear adjustment and differentiability into account. To cope with challenging and diverse properties in real-world photos, FlexiCurve is formulated to produce diverse estimations. The spatial dependencies among these estimations are implicitly modeled by a Transformer structure to improve local retouching of different regions. Thanks to the image-to-curve formulation, FlexiCurve only needs a lightweight network. Our method improves efficiency without compromising the retouching quality and losing details in the original image. The method is also appealing as it is not limited to paired training data, thus it can flexibly learn rich retouching styles from unpaired data. Extensive experiments demonstrate the efficiency, retouching performance, and flexibility of our method quantitatively and qualitatively."
                    },
                    {
                        "title": "Siamese DETR",
                        "abstract": "Recent self-supervised methods are mainly designed for representation learning with the base model, e.g., ResNets or ViTs. They cannot be easily transferred to DETR, with task-specific Transformer modules. In this work, we present Siamese DETR, a Siamese self-supervised pretraining approach for the Transformer architecture in DETR. We consider learning view-invariant and detection-oriented representations simultaneously through two complementary tasks, i.e., localization and discrimination, in a novel multi-view learning framework. Two self-supervised pretext tasks are designed: (i) Multi-View Region Detection aims at learning to localize regions-of-interest between augmented views of the input, and (ii) Multi-View Semantic Discrimination attempts to improve object-level discrimination for each region. The proposed Siamese DETR achieves state-of-the-art transfer performance on COCO and PASCAL VOC detection using different DETR variants in all setups. Code is available at https://github.com/Zx55/SiameseDEtr."
                    },
                    {
                        "title": "StyleGANEX: StyleGAN-Based Manipulation Beyond Cropped Aligned Faces",
                        "abstract": "Recent advances in face manipulation using StyleGAN have produced impressive results. However, StyleGAN is inherently limited to cropped aligned faces at a fixed image resolution it is pre-trained on. In this paper, we propose a simple and effective solution to this limitation by using dilated convolutions to rescale the receptive fields of shallow layers in StyleGAN, without altering any model parameters. This allows fixed-size small features at shallow layers to be extended into larger ones that can accommodate variable resolutions, making them more robust in characterizing unaligned faces. To enable real face inversion and manipulation, we introduce a corresponding encoder that provides the first-layer feature of the extended StyleGAN in addition to the latent style code. We validate the effectiveness of our method using unaligned face inputs of various resolutions in a diverse set of face manipulation tasks, including facial attribute editing, super-resolution, sketch/mask-to-face translation, and face toonification. Project page https://www.mmlab-ntu.com/project/styleganex"
                    },
                    {
                        "title": "MeViS: A Large-scale Benchmark for Video Segmentation with Motion Expressions",
                        "abstract": "This paper strives for motion expressions guided video segmentation, which focuses on segmenting objects in video content based on a sentence describing the motion of the objects. Existing referring video object datasets typically focus on salient objects and use language expressions that contain excessive static attributes that could potentially enable the target object to be identified in a single frame. These datasets downplay the importance of motion in video content for language-guided video object segmentation. To investigate the feasibility of using motion expressions to ground and segment objects in videos, we propose a large-scale dataset called MeViS, which contains numerous motion expressions to indicate target objects in complex environments. We benchmarked 5 existing referring video object segmentation (RVOS) methods and conducted a comprehensive comparison on the MeViS dataset. The results show that current RVOS methods cannot effectively address motion expression-guided video segmentation. We further analyze the challenges and propose a baseline approach for the proposed MeViS dataset. The goal of our benchmark is to provide a platform that enables the development of effective language-guided video segmentation algorithms that leverage motion expressions as a primary cue for object segmentation in complex video scenes. The proposed MeViS dataset has been released at https://henghuiding.github.io/MeViS."
                    },
                    {
                        "title": "Rerender A Video: Zero-Shot Text-Guided Video-to-Video Translation",
                        "abstract": "Large text-to-image diffusion models have exhibited impressive proficiency in generating high-quality images. However, when applying these models to video domain, ensuring temporal consistency across video frames remains a formidable challenge. This paper proposes a novel zero-shot text-guided video-to-video translation framework to adapt image models to videos. The framework includes two parts: key frame translation and full video translation. The first part uses an adapted diffusion model to generate key frames, with hierarchical cross-frame constraints applied to enforce coherence in shapes, textures and colors. The second part propagates the key frames to other frames with temporal-aware patch matching and frame blending. Our framework achieves global style and local texture temporal consistency at a low cost (without re-training or optimization). The adaptation is compatible with existing image diffusion techniques, allowing our framework to take advantage of them, such as customizing a specific subject with LoRA, and introducing extra spatial guidance with ControlNet. Extensive experimental results demonstrate the effectiveness of our proposed framework over existing methods in rendering high-quality and temporally-coherent videos. Code is available at our project page: https://www.mmlab-ntu.com/project/rerender/"
                    },
                    {
                        "title": "Task-Oriented Human-Object Interactions Generation with Implicit Neural Representations",
                        "abstract": "Digital human motion synthesis is a vibrant research field with applications in movies, AR/VR, and video games. Whereas methods were proposed to generate natural and realistic human motions, most only focus on modeling humans and largely ignore object movements. Generating task-oriented human-object interaction motions in simulation is challenging. For different intents of using the objects, humans conduct various motions, which requires the human first to approach the objects and then make them move consistently with the human instead of staying still. Also, to deploy in downstream applications, the synthesized motions are desired to be flexible in length, providing options to personalize the predicted motions for various purposes. To this end, we propose TOHO: Task-Oriented Human-Object Interactions Generation with Implicit Neural Representations, which generates full human-object interaction motions to conduct specific tasks, given only the task type, the object, and a starting human status. TOHO generates human-object motions in four steps: 1) it first estimates the object\u2019s final position given the task intent; 2) it then generates keyframe poses grasping the objects; 3) after that, it infills the keyframes and generates continuous motions; 4) finally, it applies a compact closed-form object motion estimation to generate the object motion. Our method generates continuous motions that are parameterized only by the temporal coordinate, which allows for upsampling of the sequence to arbitrary frames and adjusting the motion speeds by designing the temporal coordinate vector. This work takes a step further toward general human-scene interaction simulation."
                    },
                    {
                        "title": "Exploiting Diffusion Prior for Real-World Image Super-Resolution",
                        "abstract": "We present a novel approach to leverage prior knowledge encapsulated in pre-trained text-to-image diffusion models for blind super-resolution (SR). Specifically, by employing our time-aware encoder, we can achieve promising restoration results without altering the pre-trained synthesis model, thereby preserving the generative prior and minimizing training cost. To remedy the loss of fidelity caused by the inherent stochasticity of diffusion models, we introduce a controllable feature wrapping module that allows users to balance quality and fidelity by simply adjusting a scalar value during the inference process. Moreover, we develop a progressive aggregation sampling strategy to overcome the fixed-size constraints of pre-trained diffusion models, enabling adaptation to resolutions of any size. A comprehensive evaluation of our method using both synthetic and real-world benchmarks demonstrates its superiority over current state-of-the-art approaches."
                    },
                    {
                        "title": "Bailando++: 3D Dance GPT With Choreographic Memory",
                        "abstract": "Our proposed music-to-dance framework, Bailando++, addresses the challenges of driving 3D characters to dance in a way that follows the constraints of choreography norms and maintains temporal coherency with different music genres. Bailando++ consists of two components: a choreographic memory that learns to summarize meaningful dancing units from 3D pose sequences, and an actor-critic Generative Pre-trained Transformer (GPT) that composes these units into a fluent dance coherent to the music. In particular, to synchronize the diverse motion tempos and music beats, we introduce an actor-critic-based reinforcement learning scheme to the GPT with a novel beat-align reward function. Additionally, we consider learning human dance poses in the rotation domain to avoid body distortions incompatible with human morphology, and introduce a musical contextual encoding to allow the motion GPT to grasp longer-term patterns of music. Our experiments on the standard benchmark show that Bailando++ achieves state-of-the-art performance both qualitatively and quantitatively, with the added benefit of the unsupervised discovery of human-interpretable dancing-style poses in the choreographic memory."
                    },
                    {
                        "title": "SparseNeRF: Distilling Depth Ranking for Few-shot Novel View Synthesis",
                        "abstract": "Neural Radiance Field (NeRF) significantly degrades when only a limited number of views are available. To complement the lack of 3D information, depth-based models, such as DSNeRF and MonoSDF, explicitly assume the availability of accurate depth maps of multiple views. They linearly scale the accurate depth maps as supervision to guide the predicted depth of few-shot NeRFs. However, accurate depth maps are difficult and expensive to capture due to wide-range depth distances in the wild. This work presents a new Sparse-view NeRF (SparseNeRF) framework that exploits depth priors from real-world inaccurate observations. The inaccurate depth observations are either from pre-trained depth models or coarse depth maps of consumer-level depth sensors. Since coarse depth maps are not strictly scaled to the ground-truth depth maps, we propose a simple yet effective constraint, a local depth ranking method, on NeRFs such that the expected depth ranking of the NeRF is consistent with that of the coarse depth maps in local patches. To preserve the spatial continuity of the estimated depth of NeRF, we further propose a spatial continuity constraint to encourage the consistency of the expected depth continuity of NeRF with coarse depth maps. Surprisingly, with simple depth ranking constraints, SparseNeRF outperforms all state-of-the-art few-shot NeRF methods (including depth-based models) on standard LLFF and DTU datasets. Moreover, we collect a new dataset NVS-RGBD that contains real-world depth maps from Azure Kinect, ZED 2, and iPhone 13 Pro. Extensive experiments on NVS-RGBD dataset also validate the superiority and generaliz-ability of SparseNeRF. Code and dataset are available at https://sparsenerf.github.io/."
                    },
                    {
                        "title": "RenderMe-360: A Large Digital Asset Library and Benchmarks Towards High-fidelity Head Avatars",
                        "abstract": "Synthesizing high-fidelity head avatars is a central problem for computer vision and graphics. While head avatar synthesis algorithms have advanced rapidly, the best ones still face great obstacles in real-world scenarios. One of the vital causes is inadequate datasets -- 1) current public datasets can only support researchers to explore high-fidelity head avatars in one or two task directions; 2) these datasets usually contain digital head assets with limited data volume, and narrow distribution over different attributes. In this paper, we present RenderMe-360, a comprehensive 4D human head dataset to drive advance in head avatar research. It contains massive data assets, with 243+ million complete head frames, and over 800k video sequences from 500 different identities captured by synchronized multi-view cameras at 30 FPS. It is a large-scale digital library for head avatars with three key attributes: 1) High Fidelity: all subjects are captured by 60 synchronized, high-resolution 2K cameras in 360 degrees. 2) High Diversity: The collected subjects vary from different ages, eras, ethnicities, and cultures, providing abundant materials with distinctive styles in appearance and geometry. Moreover, each subject is asked to perform various motions, such as expressions and head rotations, which further extend the richness of assets. 3) Rich Annotations: we provide annotations with different granularities: cameras' parameters, matting, scan, 2D/3D facial landmarks, FLAME fitting, and text description. Based on the dataset, we build a comprehensive benchmark for head avatar research, with 16 state-of-the-art methods performed on five main tasks: novel view synthesis, novel expression synthesis, hair rendering, hair editing, and talking head generation. Our experiments uncover the strengths and weaknesses of current methods. RenderMe-360 opens the door for future exploration in head avatars."
                    },
                    {
                        "title": "CelebV-Text: A Large-Scale Facial Text-Video Dataset",
                        "abstract": "Text-driven generation models are flourishing in video generation and editing. However, face-centric text-to-video generation remains a challenge due to the lack of a suitable dataset containing high-quality videos and highly relevant texts. This paper presents Celeb V- Text, a large-scale, di-verse, and high-quality dataset of facial text-video pairs, to facilitate research on facial text-to- video generation tasks. CelebV-Text comprises 70,000 in-the-wild face video clips with diverse visual content, each paired with 20 texts gen-erated using the proposed semi-automatic text generation strategy. The provided texts are of high quality, describing both static and dynamic attributes precisely. The supe-riority of CelebV- Text over other datasets is demonstrated via comprehensive statistical analysis of the videos, texts, and text-video relevance. The effectiveness and potential of CelebV- Text are further shown through extensive self-evaluation. A benchmark is constructed with representative methods to standardize the evaluation of the facial text-to-video generation task. All data and models are publicly available11Project page: https://celebv-text.github.io."
                    },
                    {
                        "title": "DeformToon3d: Deformable Neural Radiance Fields for 3D Toonification",
                        "abstract": "In this paper, we address the challenging problem of 3D toonification, which involves transferring the style of an artistic domain onto a target 3D face with stylized geometry and texture. Although fine-tuning a pre-trained 3D GAN on the artistic domain can produce reasonable performance, this strategy has limitations in the 3D domain. In particular, fine-tuning can deteriorate the original GAN latent space, which affects subsequent semantic editing, and requires independent optimization and storage for each new style, limiting flexibility and efficient deployment. To overcome these challenges, we propose DeformToon3d, an effective toonification framework tailored for hierarchical 3D GAN. Our approach decomposes 3D toonification into subproblems of geometry and texture stylization to better preserve the original latent space. Specifically, we devise a novel StyleField that predicts conditional 3D deformation to align a real-space NeRF to the style space for geometry stylization. Thanks to the StyleField formulation, which already handles geometry stylization well, texture stylization can be achieved conveniently via adaptive style mixing that injects information of the artistic domain into the decoder of the pre-trained 3D GAN. Due to the unique design, our method enables flexible style degree control and shape-texture-specific style swap. Furthermore, we achieve efficient training without any real-world 2D-3D training pairs but proxy samples synthesized from off-the-shelf 2D toonification models. Code is released at https://github.com/junzhezhang/DeformToon3D."
                    },
                    {
                        "title": "PointHPS: Cascaded 3D Human Pose and Shape Estimation from Point Clouds",
                        "abstract": "Human pose and shape estimation (HPS) has attracted increasing attention in recent years. While most existing studies focus on HPS from 2D images or videos with inherent depth ambiguity, there are surging need to investigate HPS from 3D point clouds as depth sensors have been frequently employed in commercial devices. However, real-world sensory 3D points are usually noisy and incomplete, and also human bodies could have different poses of high diversity. To tackle these challenges, we propose a principled framework, PointHPS, for accurate 3D HPS from point clouds captured in real-world settings, which iteratively refines point features through a cascaded architecture. Specifically, each stage of PointHPS performs a series of downsampling and upsampling operations to extract and collate both local and global cues, which are further enhanced by two novel modules: 1) Cross-stage Feature Fusion (CFF) for multi-scale feature propagation that allows information to flow effectively through the stages, and 2) Intermediate Feature Enhancement (IFE) for body-aware feature aggregation that improves feature quality after each stage. To facilitate a comprehensive study under various scenarios, we conduct our experiments on two large-scale benchmarks, comprising i) a dataset that features diverse subjects and actions captured by real commercial sensors in a laboratory environment, and ii) controlled synthetic data generated with realistic considerations such as clothed humans in crowded outdoor scenes. Extensive experiments demonstrate that PointHPS, with its powerful point feature extraction and processing scheme, outperforms State-of-the-Art methods by significant margins across the board. Homepage: https://caizhongang.github.io/projects/PointHPS/."
                    },
                    {
                        "title": "Flare7K++: Mixing Synthetic and Real Datasets for Nighttime Flare Removal and Beyond",
                        "abstract": "Artificial lights commonly leave strong lens flare artifacts on the images captured at night, degrading both the visual quality and performance of vision algorithms. Existing flare removal approaches mainly focus on removing daytime flares and fail in nighttime cases. Nighttime flare removal is challenging due to the unique luminance and spectrum of artificial lights, as well as the diverse patterns and image degradation of the flares. The scarcity of the nighttime flare removal dataset constrains the research on this crucial task. In this paper, we introduce Flare7K++, the first comprehensive nighttime flare removal dataset, consisting of 962 real-captured flare images (Flare-R) and 7000 synthetic flares (Flare7K). Compared to Flare7K, Flare7K++ is particularly effective in eliminating complicated degradation around the light source, which is intractable by using synthetic flares alone. Besides, the previous flare removal pipeline relies on the manual threshold and blur kernel settings to extract light sources, which may fail when the light sources are tiny or not overexposed. To address this issue, we additionally provide the annotations of light sources in Flare7K++ and propose a new end-to-end pipeline to preserve the light source while removing lens flares. Our dataset and pipeline offer a valuable foundation and benchmark for future investigations into nighttime flare removal studies. Extensive experiments demonstrate that Flare7K++ supplements the diversity of existing flare datasets and pushes the frontier of nighttime flare removal toward real-world scenarios."
                    },
                    {
                        "title": "Explore In-Context Learning for 3D Point Cloud Understanding",
                        "abstract": "With the rise of large-scale models trained on broad data, in-context learning has become a new learning paradigm that has demonstrated significant potential in natural language processing and computer vision tasks. Meanwhile, in-context learning is still largely unexplored in the 3D point cloud domain. Although masked modeling has been successfully applied for in-context learning in 2D vision, directly extending it to 3D point clouds remains a formidable challenge. In the case of point clouds, the tokens themselves are the point cloud positions (coordinates) that are masked during inference. Moreover, position embedding in previous works may inadvertently introduce information leakage. To address these challenges, we introduce a novel framework, named Point-In-Context, designed especially for in-context learning in 3D point clouds, where both inputs and outputs are modeled as coordinates for each task. Additionally, we propose the Joint Sampling module, carefully designed to work in tandem with the general point sampling operator, effectively resolving the aforementioned technical issues. We conduct extensive experiments to validate the versatility and adaptability of our proposed methods in handling a wide range of tasks."
                    },
                    {
                        "title": "Contextual Object Detection with Multimodal Large Language Models",
                        "abstract": "Recent Multimodal Large Language Models (MLLMs) are remarkable in vision-language tasks, such as image captioning and question answering, but lack the essential perception ability, i.e., object detection. In this work, we address this limitation by introducing a novel research problem of contextual object detection -- understanding visible objects within different human-AI interactive contexts. Three representative scenarios are investigated, including the language cloze test, visual captioning, and question answering. Moreover, we present ContextDET, a unified multimodal model that is capable of end-to-end differentiable modeling of visual-language contexts, so as to locate, identify, and associate visual objects with language inputs for human-AI interaction. Our ContextDET involves three key submodels: (i) a visual encoder for extracting visual representations, (ii) a pre-trained LLM for multimodal context decoding, and (iii) a visual decoder for predicting bounding boxes given contextual object words. The new generate-then-detect framework enables us to detect object words within human vocabulary. Extensive experiments show the advantages of ContextDET on our proposed CODE benchmark, open-vocabulary detection, and referring image segmentation. Github: https://github.com/yuhangzang/ContextDET."
                    },
                    {
                        "title": "MIPI 2023 Challenge on Nighttime Flare Removal: Methods and Results",
                        "abstract": "Developing and integrating advanced image sensors with novel algorithms in camera systems are prevalent with the increasing demand for computational photography and imaging on mobile platforms. However, the lack of high-quality data for research and the rare opportunity for in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). With the success of the 1st MIPI Workshop@ECCV 2022, we introduce the second MIPI challenge including four tracks focusing on novel image sensors and imaging algorithms. In this paper, we summarize and review the Nighttime Flare Removal track on MIPI 2023. In total, 120 participants were successfully registered, and 11 teams submitted results in the final testing phase. The developed solutions in this challenge achieved state-of-the-art performance on Nighttime Flare Removal. A detailed description of all models developed in this challenge is provided in this paper. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2023/."
                    },
                    {
                        "title": "MIPI 2023 Challenge on RGBW Fusion: Methods and Results",
                        "abstract": "Developing and integrating advanced image sensors with novel algorithms in camera systems are prevalent with the increasing demand for computational photography and imaging on mobile platforms. However, the lack of high-quality data for research and the rare opportunity for an in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). With the success of the 1st MIPI Workshop@ECCV 2022, we introduce the second MIPI challenge, including four tracks focusing on novel image sensors and imaging algorithms. This paper summarizes and reviews the RGBW Joint Fusion and Denoise track on MIPI 2023. In total, 69 participants were successfully registered, and 4 teams submitted results in the final testing phase. The final results are evaluated using objective metrics, including PSNR, SSIM, LPIPS, and KLD. A detailed description of the models developed in this challenge is provided in this paper. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2023/."
                    },
                    {
                        "title": "PGDiff: Guiding Diffusion Models for Versatile Face Restoration via Partial Guidance",
                        "abstract": "Exploiting pre-trained diffusion models for restoration has recently become a favored alternative to the traditional task-specific training approach. Previous works have achieved noteworthy success by limiting the solution space using explicit degradation models. However, these methods often fall short when faced with complex degradations as they generally cannot be precisely modeled. In this paper, we propose PGDiff by introducing partial guidance, a fresh perspective that is more adaptable to real-world degradations compared to existing works. Rather than specifically defining the degradation process, our approach models the desired properties, such as image structure and color statistics of high-quality images, and applies this guidance during the reverse diffusion process. These properties are readily available and make no assumptions about the degradation process. When combined with a diffusion prior, this partial guidance can deliver appealing results across a range of restoration tasks. Additionally, PGDiff can be extended to handle composite tasks by consolidating multiple high-quality image properties, achieved by integrating the guidance from respective tasks. Experimental results demonstrate that our method not only outperforms existing diffusion-prior-based approaches but also competes favorably with task-specific models."
                    },
                    {
                        "title": "StyleInV: A Temporal Style Modulated Inversion Network for Unconditional Video Generation",
                        "abstract": "Unconditional video generation is a challenging task that involves synthesizing high-quality videos that are both coherent and of extended duration. To address this challenge, researchers have used pretrained StyleGAN image generators for high-quality frame synthesis and focused on motion generator design. The motion generator is trained in an autoregressive manner using heavy 3D convolutional discriminators to ensure motion coherence during video generation. In this paper, we introduce a novel motion generator design that uses a learning-based inversion network for GAN. The encoder in our method captures rich and smooth priors from encoding images to latents, and given the latent of an initially generated frame as guidance, our method can generate smooth future latent by modulating the inversion encoder temporally. Our method enjoys the advantage of sparse training and naturally constrains the generation space of our motion generator with the inversion network guided by the initial frame, eliminating the need for heavy discriminators. Moreover, our method supports style transfer with simple fine-tuning when the encoder is paired with a pretrained StyleGAN generator. Extensive experiments conducted on various benchmarks demonstrate the superiority of our method in generating long and high-resolution videos with decent single-frame quality and temporal consistency. Code is available at https://github.com/johannwyh/StyleInV."
                    }
                ]
            }
        }
    },
    "2402.02622": {
        "paper_data": {
            "title": "DenseFormer: Enhancing Information Flow in Transformers via Depth Weighted Averaging",
            "url": "http://arxiv.org/abs/2402.02622v2",
            "arxiv_id": "2402.02622",
            "authors": [
                "Matteo Pagliardini",
                "Amirkeivan Mohtashami",
                "Francois Fleuret",
                "Martin Jaggi"
            ],
            "abstract": "The transformer architecture by Vaswani et al. (2017) is now ubiquitous across application domains, from natural language processing to speech processing and image understanding. We propose DenseFormer, a simple modification to the standard architecture that improves the perplexity of the model without increasing its size -- adding a few thousand parameters for large-scale models in the 100B parameters range. Our approach relies on an additional averaging step after each transformer block, which computes a weighted average of current and past representations -- we refer to this operation as Depth-Weighted-Average (DWA). The learned DWA weights exhibit coherent patterns of information flow, revealing the strong and structured reuse of activations from distant layers. Experiments demonstrate that DenseFormer is more data efficient, reaching the same perplexity of much deeper transformer models, and that for the same perplexity, these new models outperform transformer baselines in terms of memory efficiency and inference time.",
            "introduction": "   1 Introduction  The transformer architecture (Vaswani et\u00a0al., 2017) is the workhorse of modern natural language processing. Recent leaps in the state of the art can be attributed in a large part to efforts scaling this architecture, from millions of parameters (Devlin et\u00a0al., 2019) to large billion-parameter models (Radford et\u00a0al., 2019; Brown et\u00a0al., 2020; Touvron et\u00a0al., 2023a, b; OpenAI, 2023). Unfortunately, those larger models come with an increased computational cost, and a large memory footprint. This renders them impractical to use in a wide range of use-cases, limiting who can benefit from them to a handful of big corporations. As an attempt to mitigate this issue, Touvron et\u00a0al. (2023a) propose training a smaller model for more steps. However, longer training requires larger datasets which becomes challenging as we are reaching scales where even extremely large datasets fall short of sufficient amounts of data (Villalobos et\u00a0al., 2022).   Furthermore, recent observations suggest that we are reaching the state of diminishing returns where increasing the depth of the model beyond a certain point does not significantly improve performance (Petty et\u00a0al., 2023a). Interestingly, a similar state of diminishing returns has been observed in the field of computer vision focused on the training of deep convolutional neural networks. Various solutions were proposed to address this issue, including DenseNets\u00a0(Huang et\u00a0al., 2017) which alleviated the problem by allowing subsequent layers to directly access outputs of earlier layers.   In this work, using a similar intuition as DenseNets, we propose the DenseFormer architecture. In particular, instead of only having skip connections from one block to the next, in DenseFormer, a weighted average of the outputs of all previous blocks is given as the input of the next block. The approach is visually summarized in Fig.\u00a00(a).      (a) DenseFormer Architecture.     (b) DWA Weights with Dilation.    Figure 1: DenseFormer architecture. The diagram in (a) shows the DenseFormer architecture with two transformer layers and a dilation of 1111. After the first (resp. second) block, the past and current intermediary representations {X0,X1}subscript\ud835\udc4b0subscript\ud835\udc4b1\\{X_{0},X_{1}\\}{ italic_X start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_X start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT } (resp. {X0,X1,X2}subscript\ud835\udc4b0subscript\ud835\udc4b1subscript\ud835\udc4b2\\{X_{0},X_{1},X_{2}\\}{ italic_X start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_X start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , italic_X start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT }) are averaged using the first (resp. second) DWA weights [\u03b10,0,\u03b10,1]subscript\ud835\udefc00subscript\ud835\udefc01[\\alpha_{0,0},\\alpha_{0,1}][ italic_\u03b1 start_POSTSUBSCRIPT 0 , 0 end_POSTSUBSCRIPT , italic_\u03b1 start_POSTSUBSCRIPT 0 , 1 end_POSTSUBSCRIPT ] (resp. [\u03b11,0,\u03b11,1,\u03b11,2]subscript\ud835\udefc10subscript\ud835\udefc11subscript\ud835\udefc12[\\alpha_{1,0},\\alpha_{1,1},\\alpha_{1,2}][ italic_\u03b1 start_POSTSUBSCRIPT 1 , 0 end_POSTSUBSCRIPT , italic_\u03b1 start_POSTSUBSCRIPT 1 , 1 end_POSTSUBSCRIPT , italic_\u03b1 start_POSTSUBSCRIPT 1 , 2 end_POSTSUBSCRIPT ]). The DWA weights are supported by red arrows. Those weights are represented in matrix form in (b), for a 12121212 layers DenseFormer. A DWA module at depth i\ud835\udc56iitalic_i has i+1\ud835\udc561i+1italic_i + 1 weights, represented in red. Increasing the dilation sparsifies this matrix, reducing the computational overhead without degrading the perplexity, see Section\u00a03.2 for more details.   We show that DenseFormers can perform the same as a much deeper standard Transformer model while at the same time being smaller in size, faster, and consuming less memory at inference. More importantly, this is achieved without increasing the amount of required data. As such, DenseFormers are also more data efficient, obtaining much better performance when trained on the same amount of data than",
            "references": [
                {
                    "title": "Large Language Models Suffer From Their Own Output: An Analysis of the Self-Consuming Training Loop",
                    "abstract": "Large Language Models (LLM) are already widely used to generate content for a variety of online platforms. As we are not able to safely distinguish LLM-generated content from human-produced content, LLM-generated content is used to train the next generation of LLMs, giving rise to a self-consuming training loop. From the image generation domain we know that such a self-consuming training loop reduces both quality and diversity of images finally ending in a model collapse. However, it is unclear whether this alarming effect can also be observed for LLMs. Therefore, we present the first study investigating the self-consuming training loop for LLMs. Further, we propose a novel method based on logic expressions that allows us to unambiguously verify the correctness of LLM-generated content, which is difficult for natural language text. We find that the self-consuming training loop produces correct outputs, however, the output declines in its diversity depending on the proportion of the used generated data. Fresh data can slow down this decline, but not stop it. Given these concerning results, we encourage researchers to study methods to negate this process."
                },
                {
                    "title": "Simplifying Transformer Blocks",
                    "abstract": "A simple design recipe for deep Transformers is to compose identical building blocks. But standard transformer blocks are far from simple, interweaving attention and MLP sub-blocks with skip connections&normalisation layers in precise arrangements. This complexity leads to brittle architectures, where seemingly minor changes can significantly reduce training speed, or render models untrainable. In this work, we ask to what extent the standard transformer block can be simplified? Combining signal propagation theory and empirical observations, we motivate modifications that allow many block components to be removed with no loss of training speed, including skip connections, projection or value parameters, sequential sub-blocks and normalisation layers. In experiments on both autoregressive decoder-only and BERT encoder-only models, our simplified transformers emulate the per-update training speed and performance of standard transformers, while enjoying 15% faster training throughput, and using 15% fewer parameters."
                },
                {
                    "title": "The Impact of Depth on Compositional Generalization in Transformer Language Models",
                    "abstract": "To process novel sentences, language models (LMs) must generalize compositionally\u2014combine familiar elements in new ways. What aspects of a model\u2019s structure promote compositional generalization? Focusing on transformers, we test the hypothesis, motivated by theoretical and empirical work, that deeper transformers generalize more compositionally. Simply adding layers increases the total number of parameters; to address this confound between depth and size, we construct three classes of models which trade off depth for width such that the total number of parameters is kept constant (41M, 134M and 374M parameters). We pretrain all models as LMs and fine-tune them on tasks that test for compositional generalization. We report three main conclusions: (1) after fine-tuning, deeper models generalize more compositionally than shallower models do, but the benefit of additional layers diminishes rapidly; (2) within each family, deeper models show better language modeling performance, but returns are similarly diminishing; (3) the benefits of depth for compositional generalization cannot be attributed solely to better performance on language modeling. Because model latency is approximately linear in the number of layers, these results lead us to the recommendation that, with a given total parameter budget, transformers can be made shallower than is typical without sacrificing performance."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning",
                    "abstract": "Scaling Transformers to longer sequence lengths has been a major problem in the last several years, promising to improve performance in language modeling and high-resolution image understanding, as well as to unlock new applications in code, audio, and video generation. The attention layer is the main bottleneck in scaling to longer sequences, as its runtime and memory increase quadratically in the sequence length. FlashAttention exploits the asymmetric GPU memory hierarchy to bring significant memory saving (linear instead of quadratic) and runtime speedup (2-4$\\times$ compared to optimized baselines), with no approximation. However, FlashAttention is still not nearly as fast as optimized matrix-multiply (GEMM) operations, reaching only 25-40\\% of the theoretical maximum FLOPs/s. We observe that the inefficiency is due to suboptimal work partitioning between different thread blocks and warps on the GPU, causing either low-occupancy or unnecessary shared memory reads/writes. We propose FlashAttention-2, with better work partitioning to address these issues. In particular, we (1) tweak the algorithm to reduce the number of non-matmul FLOPs (2) parallelize the attention computation, even for a single head, across different thread blocks to increase occupancy, and (3) within each thread block, distribute the work between warps to reduce communication through shared memory. These yield around 2$\\times$ speedup compared to FlashAttention, reaching 50-73\\% of the theoretical maximum FLOPs/s on A100 and getting close to the efficiency of GEMM operations. We empirically validate that when used end-to-end to train GPT-style models, FlashAttention-2 reaches training speed of up to 225 TFLOPs/s per A100 GPU (72\\% model FLOPs utilization)."
                },
                {
                    "title": "The RefinedWeb Dataset for Falcon LLM: Outperforming Curated Corpora with Web Data, and Web Data Only",
                    "abstract": "Large language models are commonly trained on a mixture of filtered web data and curated high-quality corpora, such as social media conversations, books, or technical papers. This curation process is believed to be necessary to produce performant models with broad zero-shot generalization abilities. However, as larger models requiring pretraining on trillions of tokens are considered, it is unclear how scalable is curation and whether we will run out of unique high-quality data soon. At variance with previous beliefs, we show that properly filtered and deduplicated web data alone can lead to powerful models; even significantly outperforming models from the state-of-the-art trained on The Pile. Despite extensive filtering, the high-quality data we extract from the web is still plentiful, and we are able to obtain five trillion tokens from CommonCrawl. We publicly release an extract of 600 billion tokens from our RefinedWeb dataset, and 1.3/7.5B parameters language models trained on it."
                },
                {
                    "title": "Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate",
                    "abstract": "Modern large language models (LLMs) like ChatGPT have shown remarkable performance on general language tasks but still struggle on complex reasoning tasks, which drives the research on cognitive behaviors of LLMs to explore human-like problem-solving strategies. Along this direction, one representative strategy is self-reflection, which asks an LLM to refine the solution with the feedback generated by itself iteratively. However, our study shows that such reflection-style methods suffer from the Degeneration-of-Thought (DoT) problem: once the LLM has established confidence in its solutions, it is unable to generate novel thoughts later through reflection even if its initial stance is incorrect. To address the DoT problem, we propose a Multi-Agent Debate (MAD) framework, in which multiple agents express their arguments in the state of\"tit for tat\"and a judge manages the debate process to obtain a final solution. Clearly, our MAD framework encourages divergent thinking in LLMs which would be helpful for tasks that require deep levels of contemplation. Experiment results on two challenging datasets, commonsense machine translation and counter-intuitive arithmetic reasoning, demonstrate the effectiveness of our MAD framework. Extensive analyses suggest that the adaptive break of debate and the modest level of\"tit for tat\"state are required for MAD to obtain good performance. Moreover, we find that LLMs might not be a fair judge if different LLMs are used for agents. Code is available at https://github.com/Skytliang/Multi-Agents-Debate."
                },
                {
                    "title": "Self-Refine: Iterative Refinement with Self-Feedback",
                    "abstract": "Like humans, large language models (LLMs) do not always generate the best output on their first try. Motivated by how humans refine their written text, we introduce Self-Refine, an approach for improving initial outputs from LLMs through iterative feedback and refinement. The main idea is to generate an initial output using an LLMs; then, the same LLMs provides feedback for its output and uses it to refine itself, iteratively. Self-Refine does not require any supervised training data, additional training, or reinforcement learning, and instead uses a single LLM as the generator, refiner, and feedback provider. We evaluate Self-Refine across 7 diverse tasks, ranging from dialog response generation to mathematical reasoning, using state-of-the-art (GPT-3.5, ChatGPT, and GPT-4) LLMs. Across all evaluated tasks, outputs generated with Self-Refine are preferred by humans and automatic metrics over those generated with the same LLM using conventional one-step generation, improving by ~20% absolute on average in task performance. Our work demonstrates that even state-of-the-art LLMs like GPT-4 can be further improved at test time using our simple, standalone approach."
                },
                {
                    "title": "Will we run out of data? Limits of LLM scaling based on human-generated data",
                    "abstract": "We investigate the potential constraints on LLM scaling posed by the availability of public human-generated text data. We forecast the growing demand for training data based on current trends and estimate the total stock of public human text data. Our findings indicate that if current LLM development trends continue, models will be trained on datasets roughly equal in size to the available stock of public human text data between 2026 and 2032, or slightly earlier if models are overtrained. We explore how progress in language modeling can continue when human-generated text datasets cannot be scaled any further. We argue that synthetic data generation, transfer learning from data-rich domains, and data efficiency improvements might support further progress."
                },
                {
                    "title": "Depth-Wise Attention (DWAtt): A Layer Fusion Method for Data-Efficient Classification",
                    "abstract": "Language Models pretrained on large textual data have been shown to encode different types of knowledge simultaneously. Traditionally, only the features from the last layer are used when adapting to new tasks or data. We put forward that, when using or finetuning deep pretrained models, intermediate layer features that may be relevant to the downstream task are buried too deep to be used efficiently in terms of needed samples or steps. To test this, we propose a new layer fusion method: Depth-Wise Attention (DWAtt), to help re-surface signals from non-final layers. We compare DWAtt to a basic concatenation-based layer fusion method (Concat), and compare both to a deeper model baseline\u2014all kept within a similar parameter budget. Our findings show that DWAtt and Concat are more step- and sample-efficient than the baseline, especially in the few-shot setting. DWAtt outperforms Concat on larger data sizes. On CoNLL-03 NER, layer fusion shows 3.68 \u2212 9.73% F1 gain at different few-shot sizes. The layer fusion models presented significantly outperform the baseline in various training scenarios with different data sizes, architectures, and training constraints."
                },
                {
                    "title": "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness",
                    "abstract": "Transformers are slow and memory-hungry on long sequences, since the time and memory complexity of self-attention are quadratic in sequence length. Approximate attention methods have attempted to address this problem by trading off model quality to reduce the compute complexity, but often do not achieve wall-clock speedup. We argue that a missing principle is making attention algorithms IO-aware -- accounting for reads and writes between levels of GPU memory. We propose FlashAttention, an IO-aware exact attention algorithm that uses tiling to reduce the number of memory reads/writes between GPU high bandwidth memory (HBM) and GPU on-chip SRAM. We analyze the IO complexity of FlashAttention, showing that it requires fewer HBM accesses than standard attention, and is optimal for a range of SRAM sizes. We also extend FlashAttention to block-sparse attention, yielding an approximate attention algorithm that is faster than any existing approximate attention method. FlashAttention trains Transformers faster than existing baselines: 15% end-to-end wall-clock speedup on BERT-large (seq. length 512) compared to the MLPerf 1.1 training speed record, 3$\\times$ speedup on GPT-2 (seq. length 1K), and 2.4$\\times$ speedup on long-range arena (seq. length 1K-4K). FlashAttention and block-sparse FlashAttention enable longer context in Transformers, yielding higher quality models (0.7 better perplexity on GPT-2 and 6.4 points of lift on long-document classification) and entirely new capabilities: the first Transformers to achieve better-than-chance performance on the Path-X challenge (seq. length 16K, 61.4% accuracy) and Path-256 (seq. length 64K, 63.1% accuracy)."
                },
                {
                    "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity",
                    "abstract": "In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with outrageous numbers of parameters -- but a constant computational cost. However, despite several notable successes of MoE, widespread adoption has been hindered by complexity, communication costs and training instability -- we address these with the Switch Transformer. We simplify the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs. Our proposed training techniques help wrangle the instabilities and we show large sparse models may be trained, for the first time, with lower precision (bfloat16) formats. We design models based off T5-Base and T5-Large to obtain up to 7x increases in pre-training speed with the same computational resources. These improvements extend into multilingual settings where we measure gains over the mT5-Base version across all 101 languages. Finally, we advance the current scale of language models by pre-training up to trillion parameter models on the\"Colossal Clean Crawled Corpus\"and achieve a 4x speedup over the T5-XXL model."
                },
                {
                    "title": "The Pile: An 800GB Dataset of Diverse Text for Language Modeling",
                    "abstract": "Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present the Pile : an 825 GiB English text corpus tar-geted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets\u2014both existing and newly constructed\u2014many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve signi\ufb01cantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction. 1"
                },
                {
                    "title": "Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention",
                    "abstract": "Transformers achieve remarkable performance in several tasks but due to their quadratic complexity, with respect to the input's length, they are prohibitively slow for very long sequences. To address this limitation, we express the self-attention as a linear dot-product of kernel feature maps and make use of the associativity property of matrix products to reduce the complexity from $\\mathcal{O}\\left(N^2\\right)$ to $\\mathcal{O}\\left(N\\right)$, where $N$ is the sequence length. We show that this formulation permits an iterative implementation that dramatically accelerates autoregressive transformers and reveals their relationship to recurrent neural networks. Our linear transformers achieve similar performance to vanilla transformers and they are up to 4000x faster on autoregressive prediction of very long sequences."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "GLU Variants Improve Transformer",
                    "abstract": "Gated Linear Units (arXiv:1612.08083) consist of the component-wise product of two linear projections, one of which is first passed through a sigmoid function. Variations on GLU are possible, using different nonlinear (or even linear) functions in place of sigmoid. We test these variants in the feed-forward sublayers of the Transformer (arXiv:1706.03762) sequence-to-sequence model, and find that some of them yield quality improvements over the typically-used ReLU or GELU activations."
                },
                {
                    "title": "Reformer: The Efficient Transformer",
                    "abstract": "Large Transformer models routinely achieve state-of-the-art results on a number of tasks but training these models can be prohibitively costly, especially on long sequences. We introduce two techniques to improve the efficiency of Transformers. For one, we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O($L^2$) to O($L\\log L$), where $L$ is the length of the sequence. Furthermore, we use reversible residual layers instead of the standard residuals, which allows storing activations only once in the training process instead of $N$ times, where $N$ is the number of layers. The resulting model, the Reformer, performs on par with Transformer models while being much more memory-efficient and much faster on long sequences."
                },
                {
                    "title": "Compressive Transformers for Long-Range Sequence Modelling",
                    "abstract": "We present the Compressive Transformer, an attentive sequence model which compresses past memories for long-range sequence learning. We find the Compressive Transformer obtains state-of-the-art language modelling results in the WikiText-103 and Enwik8 benchmarks, achieving 17.1 ppl and 0.97 bpc respectively. We also find it can model high-frequency speech effectively and can be used as a memory mechanism for RL, demonstrated on an object matching task. To promote the domain of long-range sequence learning, we propose a new open-vocabulary language modelling benchmark derived from books, PG-19."
                },
                {
                    "title": "Fast Transformer Decoding: One Write-Head is All You Need",
                    "abstract": "Multi-head attention layers, as used in the Transformer neural sequence model, are a powerful alternative to RNNs for moving information across and between sequences. While training these layers is generally fast and simple, due to parallelizability across the length of the sequence, incremental inference (where such paralleization is impossible) is often slow, due to the memory-bandwidth cost of repeatedly loading the large \"keys\" and \"values\" tensors. We propose a variant called multi-query attention, where the keys and values are shared across all of the different attention \"heads\", greatly reducing the size of these tensors and hence the memory bandwidth requirements of incremental decoding. We verify experimentally that the resulting models can indeed be much faster to decode, and incur only minor quality degradation from the baseline."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Automatic differentiation in PyTorch",
                    "abstract": "In this article, we describe an automatic differentiation module of PyTorch \u2014 a library designed to enable rapid research on machine learning models. It builds upon a few projects, most notably Lua Torch, Chainer, and HIPS Autograd [4], and provides a high performance environment with easy access to automatic differentiation of models executed on different devices (CPU and GPU). To make prototyping easier, PyTorch does not follow the symbolic approach used in many other deep learning frameworks, but focuses on differentiation of purely imperative programs, with a focus on extensibility and low overhead. Note that this preprint is a draft of certain sections from an upcoming paper covering all PyTorch features."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Densely Connected Convolutional Networks",
                    "abstract": "Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections&#x2014;one between each layer and its subsequent layer&#x2014;our network has L(L+1)/2 direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters. We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet). DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less memory and computation to achieve high performance. Code and pre-trained models are available at https://github.com/liuzhuang13/DenseNet."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Highway Networks",
                    "abstract": "There is plenty of theoretical and empirical evidence that depth of neural networks is a crucial ingredient for their success. However, network training becomes more difficult with increasing depth and training of very deep networks remains an open problem. In this extended abstract, we introduce a new architecture designed to ease gradient-based training of very deep networks. We refer to networks with this architecture as highway networks, since they allow unimpeded information flow across several layers on\"information highways\". The architecture is characterized by the use of gating units which learn to regulate the flow of information through a network. Highway networks with hundreds of layers can be trained directly using stochastic gradient descent and with a variety of activation functions, opening up the possibility of studying extremely deep and efficient architectures."
                },
                {
                    "title": "Convolutional neural networks at constrained time cost",
                    "abstract": "Though recent advanced convolutional neural networks (CNNs) have been improving the image recognition accuracy, the models are getting more complex and time-consuming. For real-world applications in industrial and commercial scenarios, engineers and developers are often faced with the requirement of constrained time budget. In this paper, we investigate the accuracy of CNNs under constrained time cost. Under this constraint, the designs of the network architectures should exhibit as trade-offs among the factors like depth, numbers of filters, filter sizes, etc. With a series of controlled comparisons, we progressively modify a baseline model while preserving its time complexity. This is also helpful for understanding the importance of the factors in network designs. We present an architecture that achieves very competitive accuracy in the ImageNet dataset (11.8% top-5 error, 10-view test), yet is 20% faster than \u201cAlexNet\u201d [14] (16.0% top-5 error, 10-view test)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we design a more efficient transformer architecture that maintains performance while reducing computational cost and memory footprint?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for democratizing access to advanced natural language processing models, as current large models are primarily accessible to big corporations due to their high resource requirements. By developing a more efficient architecture, we can enable a wider range of applications and researchers to utilize powerful models, potentially leading to innovative advancements in various fields. This research could pave the way for future studies focused on optimizing model efficiency and performance, ultimately contributing to the development of more sustainable AI technologies.\n\n### [Question 3] - Why is it hard?\nThe challenges in addressing this problem include the inherent complexity of transformer architectures, which require careful balancing of depth and performance. Naive approaches, such as simply reducing model size or depth, may lead to significant drops in performance due to the loss of representational capacity. Additionally, the need for larger datasets to train smaller models effectively poses a practical obstacle, especially as we approach the limits of available data. The diminishing returns observed with deeper models further complicate the search for an optimal architecture that can achieve high performance without excessive resource consumption.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on scaling up transformer models, often overlooking the potential for architectural innovations that could enhance efficiency. Existing solutions have not adequately addressed the trade-offs between model size, performance, and data requirements. Barriers such as a lack of exploration into alternative architectures, like the proposed DenseFormer, have prevented the development of more efficient models. Our approach differs by introducing a novel architecture that leverages weighted averaging of outputs from previous blocks, which has not been extensively explored in the context of transformers.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves the DenseFormer architecture, which incorporates a weighted average of outputs from all previous transformer blocks as input for the next block, enhancing information flow and efficiency. We will evaluate the DenseFormer using standard NLP datasets, measuring performance through metrics such as perplexity and accuracy. The expected outcomes include achieving comparable performance to deeper transformer models while being smaller, faster, and more memory-efficient, all without requiring additional data for training. This approach aims to demonstrate that architectural innovation can lead to significant improvements in model efficiency."
            }
        },
        "author_data": {
            "417f34e2-1cbb-4436-94e8-2fc867ab46c4": {
                "pk": "417f34e2-1cbb-4436-94e8-2fc867ab46c4",
                "name": "Matteo Pagliardini",
                "collaborators": [
                    "Martin Jaggi",
                    "Tatjana Chavdarova",
                    "Fran\u00e7ois Fleuret",
                    "Prakhar Gupta",
                    "Simin Fan",
                    "Michael I. Jordan",
                    "Amirkeivan Mohtashami",
                    "Sebastian U. Stich",
                    "Daniele Paliotta",
                    "Pierre Ablin"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Optimization",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Faster Causal Attention Over Large Sequences Through Sparse Flash Attention",
                        "abstract": "Transformer-based language models have found many diverse applications requiring them to process sequences of increasing length. For these applications, the causal self-attention -- which is the only component scaling quadratically w.r.t. the sequence length -- becomes a central concern. While many works have proposed schemes to sparsify the attention patterns and reduce the computational overhead of self-attention, those are often limited by implementations concerns and end up imposing a simple and static structure over the attention matrix. Conversely, implementing more dynamic sparse attentions often results in runtimes significantly slower than computing the full attention using the Flash implementation from Dao et al. (2022). We extend FlashAttention to accommodate a large class of attention sparsity patterns that, in particular, encompass key/query dropping and hashing-based attention. This leads to implementations with no computational complexity overhead and a multi-fold runtime speedup on top of FlashAttention. Even with relatively low degrees of sparsity, our method improves visibly upon FlashAttention as the sequence length increases. Without sacrificing perplexity, we increase the training speed of a transformer language model by $2.0\\times$ and $3.3\\times$ for sequences of respectively $8k$ and $16k$ tokens."
                    },
                    {
                        "title": "Better Word Embeddings by Disentangling Contextual n-Gram Information",
                        "abstract": "Pre-trained word vectors are ubiquitous in Natural Language Processing applications. In this paper, we show how training word embeddings jointly with bigram and even trigram embeddings, results in improved unigram embeddings. We claim that training word embeddings along with higher n-gram embeddings helps in the removal of the contextual information from the unigrams, resulting in better stand-alone word embeddings. We empirically show the validity of our hypothesis by outperforming other competing word representation models by a significant margin on a wide variety of tasks. We make our models publicly available."
                    },
                    {
                        "title": "Unsupervised Learning of Sentence Embeddings using Compositional n-Gram Features",
                        "abstract": "The recent tremendous success of unsupervised word embeddings in a multitude of applications raises the obvious question if similar methods could be derived to improve embeddings (i.e. semantic representations) of word sequences as well. We present a simple but efficient unsupervised objective to train distributed representations of sentences. Our method outperforms the state-of-the-art unsupervised models on most benchmark tasks, highlighting the robustness of the produced general-purpose sentence embeddings."
                    },
                    {
                        "title": "The AdEMAMix Optimizer: Better, Faster, Older",
                        "abstract": "Momentum based optimizers are central to a wide range of machine learning applications. These typically rely on an Exponential Moving Average (EMA) of gradients, which decays exponentially the present contribution of older gradients. This accounts for gradients being local linear approximations which lose their relevance as the iterate moves along the loss landscape. This work questions the use of a single EMA to accumulate past gradients and empirically demonstrates how this choice can be sub-optimal: a single EMA cannot simultaneously give a high weight to the immediate past, and a non-negligible weight to older gradients. Building on this observation, we propose AdEMAMix, a simple modification of the Adam optimizer with a mixture of two EMAs to better take advantage of past gradients. Our experiments on language modeling and image classification show -- quite surprisingly -- that gradients can stay relevant for tens of thousands of steps. They help to converge faster, and often to lower minima: e.g., a $1.3$B parameter AdEMAMix LLM trained on $101$B tokens performs comparably to an AdamW model trained on $197$B tokens ($+95\\%$). Moreover, our method significantly slows-down model forgetting during training. Our work motivates further exploration of different types of functions to leverage past gradients, beyond EMAs."
                    },
                    {
                        "title": "Agree to Disagree: Diversity through Disagreement for Better Transferability",
                        "abstract": "Gradient-based learning algorithms have an implicit simplicity bias which in effect can limit the diversity of predictors being sampled by the learning procedure. This behavior can hinder the transferability of trained models by (i) favoring the learning of simpler but spurious features -- present in the training data but absent from the test data -- and (ii) by only leveraging a small subset of predictive features. Such an effect is especially magnified when the test distribution does not exactly match the train distribution -- referred to as the Out of Distribution (OOD) generalization problem. However, given only the training data, it is not always possible to apriori assess if a given feature is spurious or transferable. Instead, we advocate for learning an ensemble of models which capture a diverse set of predictive features. Towards this, we propose a new algorithm D-BAT (Diversity-By-disAgreement Training), which enforces agreement among the models on the training data, but disagreement on the OOD data. We show how D-BAT naturally emerges from the notion of generalized discrepancy, as well as demonstrate in multiple experiments how the proposed method can mitigate shortcut-learning, enhance uncertainty and OOD detection, as well as improve transferability."
                    },
                    {
                        "title": "DoGE: Domain Reweighting with Generalization Estimation",
                        "abstract": "The coverage and composition of the pretraining data significantly impacts the generalization ability of Large Language Models (LLMs). Despite its importance, recent LLMs still rely on heuristics and trial and error to increase or reduce the influence of data-domains. We propose DOmain reweighting with Generalization Estimation (DoGE), which optimizes the probability of sampling from each domain (domain weights) in a principled way. Our approach is a two-stage process consisting of (i) training a proxy model to obtain domain weights using a bi-level optimization algorithm; (ii) training a larger base model by sampling training domains according to the learned domain weights. In our experiments, we extensively show how DoGE improves the generalization of the base model to any target data mixture. On the SlimPajama dataset, our base model gets better perplexity and few-shot reasoning accuracies across $6$ tasks compared to baseline methods. Moreover, aiming to generalize to out-of-domain target tasks, which is unseen in the pretraining corpus (OOD domain), DoGE can effectively identify inter-domain dependencies, and consistently achieves better test perplexity on the target domain."
                    },
                    {
                        "title": "Improving Generalization via Uncertainty Driven Perturbations",
                        "abstract": "Recently Shah et al., 2020 pointed out the pitfalls of the simplicity bias - the tendency of gradient-based algorithms to learn simple models - which include the model's high sensitivity to small input perturbations, as well as sub-optimal margins. In particular, while Stochastic Gradient Descent yields max-margin boundary on linear models, such guarantee does not extend to non-linear models. To mitigate the simplicity bias, we consider uncertainty-driven perturbations (UDP) of the training data points, obtained iteratively by following the direction that maximizes the model's estimated uncertainty. The uncertainty estimate does not rely on the input's label and it is highest at the decision boundary, and - unlike loss-driven perturbations - it allows for using a larger range of values for the perturbation magnitude. Furthermore, as real-world datasets have non-isotropic distances between data points of different classes, the above property is particularly appealing for increasing the margin of the decision boundary, which in turn improves the model's generalization. We show that UDP is guaranteed to achieve the maximum margin decision boundary on linear models and that it notably increases it on challenging simulated datasets. For nonlinear models, we show empirically that UDP reduces the simplicity bias and learns more exhaustive features. Interestingly, it also achieves competitive loss-based robustness and generalization trade-off on several datasets."
                    },
                    {
                        "title": "CoTFormer: A Chain-of-Thought Driven Architecture with Budget-Adaptive Computation Cost at Inference",
                        "abstract": "Scaling language models to larger and deeper sizes has led to significant boosts in performance. Even though the size of these models limits their application in compute-constrained environments, the race to continually develop ever larger and deeper foundational models is underway. At the same time -- regardless of the model size -- task-specific techniques continue to play a pivotal role in achieving optimal downstream performance. One of these techniques, called Chain-of-Thought (CoT), is particularly interesting since, as we point out in this work, it resembles employing a deeper transformer through re-applying the model multiple times. However, a key subtlety in computing the attention of past tokens differentiates CoT from simply applying the model several times. Based on this insight, we propose CoTFormer, a novel architecture which closely mimics CoT at the token level, allowing us to obtain significantly improved accuracies close to much larger models. While applying CoT introduces additional computation costs, we compensate for it by leveraging CoTFormer's special compatibility with token-wise variable depth. Through a compute adaptive model -- which automatically allocates the compute to tokens that need it most -- we show that it is possible to reduce the computation cost significantly without any reduction in accuracy, and with further compute cost reductions possible while maintaining a competitive accuracy."
                    },
                    {
                        "title": "A Primal-Dual Approach to Solving Variational Inequalities with General Constraints",
                        "abstract": "Yang et al. (2023) recently showed how to use first-order gradient methods to solve general variational inequalities (VIs) under a limiting assumption that analytic solutions of specific subproblems are available. In this paper, we circumvent this assumption via a warm-starting technique where we solve subproblems approximately and initialize variables with the approximate solution found at the previous iteration. We prove the convergence of this method and show that the gap function of the last iterate of the method decreases at a rate of $O(\\frac{1}{\\sqrt{K}})$ when the operator is $L$-Lipschitz and monotone. In numerical experiments, we show that this technique can converge much faster than its exact counterpart. Furthermore, for the cases when the inequality constraints are simple, we introduce an alternative variant of ACVI and establish its convergence under the same conditions. Finally, we relax the smoothness assumptions in Yang et al., yielding, to our knowledge, the first convergence result for VIs with general constraints that does not rely on the assumption that the operator is $L$-Lipschitz."
                    },
                    {
                        "title": "The Peril of Popular Deep Learning Uncertainty Estimation Methods",
                        "abstract": "Uncertainty estimation (UE) techniques -- such as the Gaussian process (GP), Bayesian neural networks (BNN), Monte Carlo dropout (MCDropout) -- aim to improve the interpretability of machine learning models by assigning an estimated uncertainty value to each of their prediction outputs. However, since too high uncertainty estimates can have fatal consequences in practice, this paper analyzes the above techniques.   Firstly, we show that GP methods always yield high uncertainty estimates on out of distribution (OOD) data. Secondly, we show on a 2D toy example that both BNNs and MCDropout do not give high uncertainty estimates on OOD samples. Finally, we show empirically that this pitfall of BNNs and MCDropout holds on real world datasets as well. Our insights (i) raise awareness for the more cautious use of currently popular UE methods in Deep Learning, (ii) encourage the development of UE methods that approximate GP-based methods -- instead of BNNs and MCDropout, and (iii) our empirical setups can be used for verifying the OOD performances of any other UE method. The source code is available at https://github.com/epfml/uncertainity-estimation."
                    },
                    {
                        "title": "Taming GANs with Lookahead-Minmax",
                        "abstract": "Generative Adversarial Networks are notoriously challenging to train. The underlying minmax optimization is highly susceptible to the variance of the stochastic gradient and the rotational component of the associated game vector field. To tackle these challenges, we propose the Lookahead algorithm for minmax optimization, originally developed for single objective minimization only. The backtracking step of our Lookahead-minmax naturally handles the rotational game dynamics, a property which was identified to be key for enabling gradient ascent descent methods to converge on challenging examples often analyzed in the literature. Moreover, it implicitly handles high variance without using large mini-batches, known to be essential for reaching state of the art performance. Experimental results on MNIST, SVHN, CIFAR-10, and ImageNet demonstrate a clear advantage of combining Lookahead-minmax with Adam or extragradient, in terms of performance and improved stability, for negligible memory and computational cost. Using 30-fold fewer parameters and 16-fold smaller minibatches we outperform the reported performance of the class-dependent BigGAN on CIFAR-10 by obtaining FID of 12.19 without using the class labels, bringing state-of-the-art GAN training within reach of common computational resources."
                    },
                    {
                        "title": "MEDITRON-70B: Scaling Medical Pretraining for Large Language Models",
                        "abstract": "Large language models (LLMs) can potentially democratize access to medical knowledge. While many efforts have been made to harness and improve LLMs' medical knowledge and reasoning capacities, the resulting models are either closed-source (e.g., PaLM, GPT-4) or limited in scale (<= 13B parameters), which restricts their abilities. In this work, we improve access to large-scale medical LLMs by releasing MEDITRON: a suite of open-source LLMs with 7B and 70B parameters adapted to the medical domain. MEDITRON builds on Llama-2 (through our adaptation of Nvidia's Megatron-LM distributed trainer), and extends pretraining on a comprehensively curated medical corpus, including selected PubMed articles, abstracts, and internationally-recognized medical guidelines. Evaluations using four major medical benchmarks show significant performance gains over several state-of-the-art baselines before and after task-specific finetuning. Overall, MEDITRON achieves a 6% absolute performance gain over the best public baseline in its parameter class and 3% over the strongest baseline we finetuned from Llama-2. Compared to closed-source LLMs, MEDITRON-70B outperforms GPT-3.5 and Med-PaLM and is within 5% of GPT-4 and 10% of Med-PaLM-2. We release our code for curating the medical pretraining corpus and the MEDITRON model weights to drive open-source development of more capable medical LLMs."
                    }
                ]
            },
            "cdaea8fa-242e-45e1-b799-acc1cad473eb": {
                "pk": "cdaea8fa-242e-45e1-b799-acc1cad473eb",
                "name": "Amirkeivan Mohtashami",
                "collaborators": [
                    "Martin Jaggi",
                    "Sebastian U. Stich",
                    "Sebastian Stich",
                    "Matteo Pagliardini",
                    "Dan Alistarh",
                    "Mauro Verzetti",
                    "Paul K. Rubenstein",
                    "Vitaly Aksenov",
                    "Alexandra Drozdova",
                    "Florian Hartmann"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Deep Learning",
                    "Quantization"
                ],
                "publications": [
                    {
                        "title": "Landmark Attention: Random-Access Infinite Context Length for Transformers",
                        "abstract": "While Transformers have shown remarkable success in natural language processing, their attention mechanism's large memory requirements have limited their ability to handle longer contexts. Prior approaches, such as recurrent memory or retrieval-based augmentation, have either compromised the random-access flexibility of attention (i.e., the capability to select any token in the entire context) or relied on separate mechanisms for relevant context retrieval, which may not be compatible with the model's attention. In this paper, we present a novel approach that allows access to the complete context while retaining random-access flexibility, closely resembling running attention on the entire context. Our method uses a landmark token to represent each block of the input and trains the attention to use it for selecting relevant blocks, enabling retrieval of blocks directly through the attention mechanism instead of by relying on a separate mechanism. Our approach seamlessly integrates with specialized data structures and the system's memory hierarchy, enabling processing of arbitrarily long context lengths. We demonstrate that our method can obtain comparable performance with Transformer-XL while significantly reducing the number of retrieved tokens in each step. Finally, we show that fine-tuning LLaMA 7B with our method successfully extends its context length capacity to over 32k tokens, allowing for inference at the context lengths of GPT-4. We release the implementation of landmark attention and the code to reproduce our experiments at https://github.com/epfml/landmark-attention/."
                    },
                    {
                        "title": "Learning Translation Quality Evaluation on Low Resource Languages from Large Language Models",
                        "abstract": "Learned metrics such as BLEURT have in recent years become widely employed to evaluate the quality of machine translation systems. Training such metrics requires data which can be expensive and difficult to acquire, particularly for lower-resource languages. We show how knowledge can be distilled from Large Language Models (LLMs) to improve upon such learned metrics without requiring human annotators, by creating synthetic datasets which can be mixed into existing datasets, requiring only a corpus of text in the target language. We show that the performance of a BLEURT-like model on lower resource languages can be improved in this way."
                    },
                    {
                        "title": "Critical Parameters for Scalable Distributed Learning with Large Batches and Asynchronous Updates",
                        "abstract": "It has been experimentally observed that the efficiency of distributed training with stochastic gradient (SGD) depends decisively on the batch size and -- in asynchronous implementations -- on the gradient staleness. Especially, it has been observed that the speedup saturates beyond a certain batch size and/or when the delays grow too large. We identify a data-dependent parameter that explains the speedup saturation in both these settings. Our comprehensive theoretical analysis, for strongly convex, convex and non-convex settings, unifies and generalized prior work directions that often focused on only one of these two aspects. In particular, our approach allows us to derive improved speedup results under frequently considered sparsity assumptions. Our insights give rise to theoretically based guidelines on how the learning rates can be adjusted in practice. We show that our results are tight and illustrate key findings in numerical experiments."
                    },
                    {
                        "title": "Masked Training of Neural Networks with Partial Gradients",
                        "abstract": "State-of-the-art training algorithms for deep learning models are based on stochastic gradient descent (SGD). Recently, many variations have been explored: perturbing parameters for better accuracy (such as in Extragradient), limiting SGD updates to a subset of parameters for increased efficiency (such as meProp) or a combination of both (such as Dropout). However, the convergence of these methods is often not studied in theory. We propose a unified theoretical framework to study such SGD variants -- encompassing the aforementioned algorithms and additionally a broad variety of methods used for communication efficient training or model compression. Our insights can be used as a guide to improve the efficiency of such methods and facilitate generalization to new applications. As an example, we tackle the task of jointly training networks, a version of which (limited to sub-networks) is used to create Slimmable Networks. By training a low-rank Transformer jointly with a standard one we obtain superior performance than when it is trained separately."
                    },
                    {
                        "title": "Characterizing & Finding Good Data Orderings for Fast Convergence of Sequential Gradient Methods",
                        "abstract": "While SGD, which samples from the data with replacement is widely studied in theory, a variant called Random Reshuffling (RR) is more common in practice. RR iterates through random permutations of the dataset and has been shown to converge faster than SGD. When the order is chosen deterministically, a variant called incremental gradient descent (IG), the existing convergence bounds show improvement over SGD but are worse than RR. However, these bounds do not differentiate between a good and a bad ordering and hold for the worst choice of order. Meanwhile, in some cases, choosing the right order when using IG can lead to convergence faster than RR. In this work, we quantify the effect of order on convergence speed, obtaining convergence bounds based on the chosen sequence of permutations while also recovering previous results for RR. In addition, we show benefits of using structured shuffling when various levels of abstractions (e.g. tasks, classes, augmentations, etc.) exists in the dataset in theory and in practice. Finally, relying on our measure, we develop a greedy algorithm for choosing good orders during training, achieving superior performance (by more than 14 percent in accuracy) over RR."
                    },
                    {
                        "title": "Special Properties of Gradient Descent with Large Learning Rates",
                        "abstract": "When training neural networks, it has been widely observed that a large step size is essential in stochastic gradient descent (SGD) for obtaining superior models. However, the effect of large step sizes on the success of SGD is not well understood theoretically. Several previous works have attributed this success to the stochastic noise present in SGD. However, we show through a novel set of experiments that the stochastic noise is not sufficient to explain good non-convex training, and that instead the effect of a large learning rate itself is essential for obtaining best performance.We demonstrate the same effects also in the noise-less case, i.e. for full-batch GD. We formally prove that GD with large step size -- on certain non-convex function classes -- follows a different trajectory than GD with a small step size, which can lead to convergence to a global minimum instead of a local one. Our settings provide a framework for future analysis which allows comparing algorithms based on behaviors that can not be observed in the traditional settings."
                    },
                    {
                        "title": "CoTFormer: A Chain-of-Thought Driven Architecture with Budget-Adaptive Computation Cost at Inference",
                        "abstract": "Scaling language models to larger and deeper sizes has led to significant boosts in performance. Even though the size of these models limits their application in compute-constrained environments, the race to continually develop ever larger and deeper foundational models is underway. At the same time -- regardless of the model size -- task-specific techniques continue to play a pivotal role in achieving optimal downstream performance. One of these techniques, called Chain-of-Thought (CoT), is particularly interesting since, as we point out in this work, it resembles employing a deeper transformer through re-applying the model multiple times. However, a key subtlety in computing the attention of past tokens differentiates CoT from simply applying the model several times. Based on this insight, we propose CoTFormer, a novel architecture which closely mimics CoT at the token level, allowing us to obtain significantly improved accuracies close to much larger models. While applying CoT introduces additional computation costs, we compensate for it by leveraging CoTFormer's special compatibility with token-wise variable depth. Through a compute adaptive model -- which automatically allocates the compute to tokens that need it most -- we show that it is possible to reduce the computation cost significantly without any reduction in accuracy, and with further compute cost reductions possible while maintaining a competitive accuracy."
                    },
                    {
                        "title": "The Splay-List: A Distribution-Adaptive Concurrent Skip-List",
                        "abstract": "The design and implementation of efficient concurrent data structures have seen significant attention. However, most of this work has focused on concurrent data structures providing good \\emph{worst-case} guarantees. In real workloads, objects are often accessed at different rates, since access distributions may be non-uniform. Efficient distribution-adaptive data structures are known in the sequential case, e.g. the splay-trees; however, they often are hard to translate efficiently in the concurrent case.   In this paper, we investigate distribution-adaptive concurrent data structures and propose a new design called the splay-list. At a high level, the splay-list is similar to a standard skip-list, with the key distinction that the height of each element adapts dynamically to its access rate: popular elements ``move up,'' whereas rarely-accessed elements decrease in height. We show that the splay-list provides order-optimal amortized complexity bounds for a subset of operations while being amenable to efficient concurrent implementation. Experimental results show that the splay-list can leverage distribution-adaptivity to improve on the performance of classic concurrent designs, and can outperform the only previously-known distribution-adaptive design in certain settings."
                    },
                    {
                        "title": "Social Learning: Towards Collaborative Learning with Large Language Models",
                        "abstract": "We introduce the framework of \"social learning\" in the context of large language models (LLMs), whereby models share knowledge with each other in a privacy-aware manner using natural language. We present and evaluate two approaches for knowledge transfer between LLMs. In the first scenario, we allow the model to generate abstract prompts aiming to teach the task. In our second approach, models transfer knowledge by generating synthetic examples. We evaluate these methods across diverse datasets and quantify memorization as a proxy for privacy loss. These techniques inspired by social learning yield promising results with low memorization of the original data. In particular, we show that performance using these methods is comparable to results with the use of original labels and prompts. Our work demonstrates the viability of social learning for LLMs, establishes baseline approaches and highlights several unexplored areas for future work."
                    },
                    {
                        "title": "QuaRot: Outlier-Free 4-Bit Inference in Rotated LLMs",
                        "abstract": "We introduce QuaRot, a new Quantization scheme based on Rotations, which is able to quantize LLMs end-to-end, including all weights, activations, and KV cache in 4 bits. QuaRot rotates LLMs in a way that removes outliers from the hidden state without changing the output, making quantization easier. This computational invariance is applied to the hidden state (residual) of the LLM, as well as to the activations of the feed-forward components, aspects of the attention mechanism and to the KV cache. The result is a quantized model where all matrix multiplications are performed in 4-bits, without any channels identified for retention in higher precision. Our quantized LLaMa2-70B model has losses of at most 0.29 WikiText-2 perplexity and retains 99% of the zero-shot performance. Code is available at: https://github.com/spcl/QuaRot."
                    },
                    {
                        "title": "MEDITRON-70B: Scaling Medical Pretraining for Large Language Models",
                        "abstract": "Large language models (LLMs) can potentially democratize access to medical knowledge. While many efforts have been made to harness and improve LLMs' medical knowledge and reasoning capacities, the resulting models are either closed-source (e.g., PaLM, GPT-4) or limited in scale (<= 13B parameters), which restricts their abilities. In this work, we improve access to large-scale medical LLMs by releasing MEDITRON: a suite of open-source LLMs with 7B and 70B parameters adapted to the medical domain. MEDITRON builds on Llama-2 (through our adaptation of Nvidia's Megatron-LM distributed trainer), and extends pretraining on a comprehensively curated medical corpus, including selected PubMed articles, abstracts, and internationally-recognized medical guidelines. Evaluations using four major medical benchmarks show significant performance gains over several state-of-the-art baselines before and after task-specific finetuning. Overall, MEDITRON achieves a 6% absolute performance gain over the best public baseline in its parameter class and 3% over the strongest baseline we finetuned from Llama-2. Compared to closed-source LLMs, MEDITRON-70B outperforms GPT-3.5 and Med-PaLM and is within 5% of GPT-4 and 10% of Med-PaLM-2. We release our code for curating the medical pretraining corpus and the MEDITRON model weights to drive open-source development of more capable medical LLMs."
                    }
                ]
            },
            "a3a4d0b7-7e99-4bde-9ead-c96833f65028": {
                "pk": "a3a4d0b7-7e99-4bde-9ead-c96833f65028",
                "name": "Francois Fleuret",
                "collaborators": [
                    "Suraj Srinivas",
                    "Cijo Jose",
                    "Evann Courdier"
                ],
                "domain": [
                    "Deep Learning",
                    "Transfer Learning",
                    "Metric Learning",
                    "Neural Network Interpretability"
                ],
                "publications": [
                    {
                        "title": "Knowledge Transfer with Jacobian Matching",
                        "abstract": "Classical distillation methods transfer representations from a \"teacher\" neural network to a \"student\" network by matching their output activations. Recent methods also match the Jacobians, or the gradient of output activations with the input. However, this involves making some ad hoc decisions, in particular, the choice of the loss function.   In this paper, we first establish an equivalence between Jacobian matching and distillation with input noise, from which we derive appropriate loss functions for Jacobian matching. We then rely on this analysis to apply Jacobian matching to transfer learning by establishing equivalence of a recent transfer learning procedure to distillation.   We then show experimentally on standard image datasets that Jacobian-based penalties improve distillation, robustness to noisy inputs, and transfer learning."
                    },
                    {
                        "title": "Scalable Metric Learning via Weighted Approximate Rank Component Analysis",
                        "abstract": "We are interested in the large-scale learning of Mahalanobis distances, with a particular focus on person re-identification.   We propose a metric learning formulation called Weighted Approximate Rank Component Analysis (WARCA). WARCA optimizes the precision at top ranks by combining the WARP loss with a regularizer that favors orthonormal linear mappings, and avoids rank-deficient embeddings. Using this new regularizer allows us to adapt the large-scale WSABIE procedure and to leverage the Adam stochastic optimization algorithm, which results in an algorithm that scales gracefully to very large data-sets. Also, we derive a kernelized version which allows to take advantage of state-of-the-art features for re-identification when data-set size permits kernel computation.   Benchmarks on recent and standard re-identification data-sets show that our method beats existing state-of-the-art techniques both in term of accuracy and speed. We also provide experimental analysis to shade lights on the properties of the regularizer we use, and how it improves performance."
                    },
                    {
                        "title": "Full-Gradient Representation for Neural Network Visualization",
                        "abstract": "We introduce a new tool for interpreting neural net responses, namely full-gradients, which decomposes the neural net response into input sensitivity and per-neuron sensitivity components. This is the first proposed representation which satisfies two key properties: completeness and weak dependence, which provably cannot be satisfied by any saliency map-based interpretability method. For convolutional nets, we also propose an approximate saliency map representation, called FullGrad, obtained by aggregating the full-gradient components.   We experimentally evaluate the usefulness of FullGrad in explaining model behaviour with two quantitative tests: pixel perturbation and remove-and-retrain. Our experiments reveal that our method explains model behaviour correctly, and more comprehensively than other methods in the literature. Visual inspection also reveals that our saliency maps are sharper and more tightly confined to object regions than other methods."
                    },
                    {
                        "title": "Real-Time Segmentation Networks should be Latency Aware",
                        "abstract": "As scene segmentation systems reach visually accurate results, many recent papers focus on making these network architectures faster, smaller and more efficient. In particular, studies often aim at designingreal-time'systems. Achieving this goal is particularly relevant in the context of real-time video understanding for autonomous vehicles, and robots. In this paper, we argue that the commonly used performance metric of mean Intersection over Union (mIoU) does not fully capture the information required to estimate the true performance of these networks when they operate inreal-time'. We propose a change of objective in the segmentation task, and its associated metric that encapsulates this missing information in the following way: We propose to predict the future output segmentation map that will match the future input frame at the time when the network finishes the processing. We introduce the associated latency-aware metric, from which we can determine a ranking. We perform latency timing experiments of some recent networks on different hardware and assess the performances of these networks on our proposed task. We propose improvements to scene segmentation networks to better perform on our task by using multi-frames input and increasing capacity in the initial convolutional layers."
                    }
                ]
            },
            "0e54f312-23ac-410e-81ca-981bd9ccd70c": {
                "pk": "0e54f312-23ac-410e-81ca-981bd9ccd70c",
                "name": "Martin Jaggi",
                "collaborators": [
                    "Martin Tak\u00e1\u010d",
                    "Prakhar Gupta",
                    "Virginia Smith",
                    "Michael I. Jordan",
                    "Bernd G\u00e4rtner",
                    "Matteo Pagliardini",
                    "Rajiv Khanna",
                    "Michael Tschannen",
                    "Chenxin Ma",
                    "Peter Richt\u00e1rik"
                ],
                "domain": [
                    "Optimization",
                    "Machine Learning",
                    "Natural Language Processing",
                    "Signal Processing"
                ],
                "publications": [
                    {
                        "title": "An Equivalence between the Lasso and Support Vector Machines",
                        "abstract": "We investigate the relation of two fundamental tools in machine learning and signal processing, that is the support vector machine (SVM) for classification, and the Lasso technique used in regression. We show that the resulting optimization problems are equivalent, in the following sense. Given any instance of an $\\ell_2$-loss soft-margin (or hard-margin) SVM, we construct a Lasso instance having the same optimal solutions, and vice versa.   As a consequence, many existing optimization algorithms for both SVMs and Lasso can also be applied to the respective other problem instances. Also, the equivalence allows for many known theoretical insights for SVM and Lasso to be translated between the two settings. One such implication gives a simple kernelized version of the Lasso, analogous to the kernels used in the SVM setting. Another consequence is that the sparsity of a Lasso solution is equal to the number of support vectors for the corresponding SVM instance, and that one can use screening rules to prune the set of support vectors. Furthermore, we can relate sublinear time algorithms for the two problems, and give a new such algorithm variant for the Lasso. We also study the regularization paths for both methods."
                    },
                    {
                        "title": "Convex Optimization without Projection Steps",
                        "abstract": "For the general problem of minimizing a convex function over a compact convex domain, we will investigate a simple iterative approximation algorithm based on the method by Frank & Wolfe 1956, that does not need projection steps in order to stay inside the optimization domain. Instead of a projection step, the linearized problem defined by a current subgradient is solved, which gives a step direction that will naturally stay in the domain. Our framework generalizes the sparse greedy algorithm of Frank & Wolfe and its primal-dual analysis by Clarkson 2010 (and the low-rank SDP approach by Hazan 2008) to arbitrary convex domains. We give a convergence proof guaranteeing {\\epsilon}-small duality gap after O(1/{\\epsilon}) iterations.   The method allows us to understand the sparsity of approximate solutions for any l1-regularized convex optimization problem (and for optimization over the simplex), expressed as a function of the approximation quality. We obtain matching upper and lower bounds of {\\Theta}(1/{\\epsilon}) for the sparsity for l1-problems. The same bounds apply to low-rank semidefinite optimization with bounded trace, showing that rank O(1/{\\epsilon}) is best possible here as well. As another application, we obtain sparse matrices of O(1/{\\epsilon}) non-zero entries as {\\epsilon}-approximate solutions when optimizing any convex function over a class of diagonally dominant symmetric matrices.   We show that our proposed first-order method also applies to nuclear norm and max-norm matrix optimization problems. For nuclear norm regularized optimization, such as matrix completion and low-rank recovery, we demonstrate the practical efficiency and scalability of our algorithm for large matrix problems, as e.g. the Netflix dataset. For general convex optimization over bounded matrix max-norm, our algorithm is the first with a convergence guarantee, to the best of our knowledge."
                    },
                    {
                        "title": "Unsupervised Learning of Sentence Embeddings using Compositional n-Gram Features",
                        "abstract": "The recent tremendous success of unsupervised word embeddings in a multitude of applications raises the obvious question if similar methods could be derived to improve embeddings (i.e. semantic representations) of word sequences as well. We present a simple but efficient unsupervised objective to train distributed representations of sentences. Our method outperforms the state-of-the-art unsupervised models on most benchmark tasks, highlighting the robustness of the produced general-purpose sentence embeddings."
                    },
                    {
                        "title": "A Unified Optimization View on Generalized Matching Pursuit and Frank-Wolfe",
                        "abstract": "Two of the most fundamental prototypes of greedy optimization are the matching pursuit and Frank-Wolfe algorithms. In this paper, we take a unified view on both classes of methods, leading to the first explicit convergence rates of matching pursuit methods in an optimization sense, for general sets of atoms. We derive sublinear ($1/t$) convergence for both classes on general smooth objectives, and linear convergence on strongly convex objectives, as well as a clear correspondence of algorithm variants. Our presented algorithms and rates are affine invariant, and do not need any incoherence or sparsity assumptions."
                    },
                    {
                        "title": "Generating Steganographic Text with LSTMs",
                        "abstract": "Motivated by concerns for user privacy, we design a steganographic system (\"stegosystem\") that enables two users to exchange encrypted messages without an adversary detecting that such an exchange is taking place. We propose a new linguistic stegosystem based on a Long Short-Term Memory (LSTM) neural network. We demonstrate our approach on the Twitter and Enron email datasets and show that it yields high-quality steganographic text while significantly improving capacity (encrypted bits per word) relative to the state-of-the-art."
                    },
                    {
                        "title": "An Affine Invariant Linear Convergence Analysis for Frank-Wolfe Algorithms",
                        "abstract": "We study the linear convergence of variants of the Frank-Wolfe algorithms for some classes of strongly convex problems, using only affine-invariant quantities. As in Guelat & Marcotte (1986), we show the linear convergence of the standard Frank-Wolfe algorithm when the solution is in the interior of the domain, but with affine invariant constants. We also show the linear convergence of the away-steps variant of the Frank-Wolfe algorithm, but with constants which only depend on the geometry of the domain, and not any property of the location of the optimal solution. Running these algorithms does not require knowing any problem specific parameters."
                    },
                    {
                        "title": "Primal-Dual Rates and Certificates",
                        "abstract": "We propose an algorithm-independent framework to equip existing optimization methods with primal-dual certificates. Such certificates and corresponding rate of convergence guarantees are important for practitioners to diagnose progress, in particular in machine learning applications. We obtain new primal-dual convergence rates, e.g., for the Lasso as well as many L1, Elastic Net, group Lasso and TV-regularized problems. The theory applies to any norm-regularized generalized linear model. Our approach provides efficiently computable duality gaps which are globally defined, without modifying the original problems in the region of interest."
                    },
                    {
                        "title": "Adding vs. Averaging in Distributed Primal-Dual Optimization",
                        "abstract": "Distributed optimization methods for large-scale machine learning suffer from a communication bottleneck. It is difficult to reduce this bottleneck while still efficiently and accurately aggregating partial work from different machines. In this paper, we present a novel generalization of the recent communication-efficient primal-dual framework (CoCoA) for distributed optimization. Our framework, CoCoA+, allows for additive combination of local updates to the global parameters at each iteration, whereas previous schemes with convergence guarantees only allow conservative averaging. We give stronger (primal-dual) convergence rate guarantees for both CoCoA as well as our new variants, and generalize the theory for both methods to cover non-smooth convex loss functions. We provide an extensive experimental comparison that shows the markedly improved performance of CoCoA+ on several real-world distributed datasets, especially when scaling up the number of machines."
                    },
                    {
                        "title": "An Exponential Lower Bound on the Complexity of Regularization Paths",
                        "abstract": "For a variety of regularized optimization problems in machine learning, algorithms computing the entire solution path have been developed recently. Most of these methods are quadratic programs that are parameterized by a single parameter, as for example the Support Vector Machine (SVM). Solution path algorithms do not only compute the solution for one particular value of the regularization parameter but the entire path of solutions, making the selection of an optimal parameter much easier.   It has been assumed that these piecewise linear solution paths have only linear complexity, i.e. linearly many bends. We prove that for the support vector machine this complexity can be exponential in the number of training points in the worst case. More strongly, we construct a single instance of n input points in d dimensions for an SVM such that at least \\Theta(2^{n/2}) = \\Theta(2^d) many distinct subsets of support vectors occur as the regularization parameter changes."
                    },
                    {
                        "title": "Pursuits in Structured Non-Convex Matrix Factorizations",
                        "abstract": "Efficiently representing real world data in a succinct and parsimonious manner is of central importance in many fields. We present a generalized greedy pursuit framework, allowing us to efficiently solve structured matrix factorization problems, where the factors are allowed to be from arbitrary sets of structured vectors. Such structure may include sparsity, non-negativeness, order, or a combination thereof. The algorithm approximates a given matrix by a linear combination of few rank-1 matrices, each factorized into an outer product of two vector atoms of the desired structure. For the non-convex subproblems of obtaining good rank-1 structured matrix atoms, we employ and analyze a general atomic power method. In addition to the above applications, we prove linear convergence for generalized pursuit variants in Hilbert spaces - for the task of approximation over the linear span of arbitrary dictionaries - which generalizes OMP and is useful beyond matrix problems. Our experiments on real datasets confirm both the efficiency and also the broad applicability of our framework in practice."
                    },
                    {
                        "title": "Faster Coordinate Descent via Adaptive Importance Sampling",
                        "abstract": "Coordinate descent methods employ random partial updates of decision variables in order to solve huge-scale convex optimization problems. In this work, we introduce new adaptive rules for the random selection of their updates. By adaptive, we mean that our selection rules are based on the dual residual or the primal-dual gap estimates and can change at each iteration. We theoretically characterize the performance of our selection rules and demonstrate improvements over the state-of-the-art, and extend our theory and algorithms to general convex objectives. Numerical evidence with hinge-loss support vector machines and Lasso confirm that the practice follows the theory."
                    },
                    {
                        "title": "Communication-Efficient Distributed Dual Coordinate Ascent",
                        "abstract": "Communication remains the most significant bottleneck in the performance of distributed optimization algorithms for large-scale machine learning. In this paper, we propose a communication-efficient framework, CoCoA, that uses local computation in a primal-dual setting to dramatically reduce the amount of necessary communication. We provide a strong convergence rate analysis for this class of algorithms, as well as experiments on real-world distributed datasets with implementations in Spark. In our experiments, we find that as compared to state-of-the-art mini-batch versions of SGD and SDCA algorithms, CoCoA converges to the same .001-accurate solution quality on average 25x as quickly."
                    },
                    {
                        "title": "Distributed Optimization with Arbitrary Local Solvers",
                        "abstract": "With the growth of data and necessity for distributed optimization methods, solvers that work well on a single machine must be re-designed to leverage distributed computation. Recent work in this area has been limited by focusing heavily on developing highly specific methods for the distributed environment. These special-purpose methods are often unable to fully leverage the competitive performance of their well-tuned and customized single machine counterparts. Further, they are unable to easily integrate improvements that continue to be made to single machine methods. To this end, we present a framework for distributed optimization that both allows the flexibility of arbitrary solvers to be used on each (single) machine locally, and yet maintains competitive performance against other state-of-the-art special-purpose distributed methods. We give strong primal-dual convergence rate guarantees for our framework that hold for arbitrary local solvers. We demonstrate the impact of local solver selection both theoretically and in an extensive experimental comparison. Finally, we provide thorough implementation details for our framework, highlighting areas for practical performance gains."
                    },
                    {
                        "title": "A Combinatorial Algorithm to Compute Regularization Paths",
                        "abstract": "For a wide variety of regularization methods, algorithms computing the entire solution path have been developed recently. Solution path algorithms do not only compute the solution for one particular value of the regularization parameter but the entire path of solutions, making the selection of an optimal parameter much easier. Most of the currently used algorithms are not robust in the sense that they cannot deal with general or degenerate input. Here we present a new robust, generic method for parametric quadratic programming. Our algorithm directly applies to nearly all machine learning applications, where so far every application required its own different algorithm.   We illustrate the usefulness of our method by applying it to a very low rank problem which could not be solved by existing path tracking methods, namely to compute part-worth values in choice based conjoint analysis, a popular technique from market research to estimate consumers preferences on a class of parameterized options."
                    },
                    {
                        "title": "Screening Rules for Convex Problems",
                        "abstract": "We propose a new framework for deriving screening rules for convex optimization problems. Our approach covers a large class of constrained and penalized optimization formulations, and works in two steps. First, given any approximate point, the structure of the objective function and the duality gap is used to gather information on the optimal solution. In the second step, this information is used to produce screening rules, i.e. safely identifying unimportant weight variables of the optimal solution. Our general framework leads to a large variety of useful existing as well as new screening rules for many applications. For example, we provide new screening rules for general simplex and $L_1$-constrained problems, Elastic Net, squared-loss Support Vector Machines, minimum enclosing ball, as well as structured norm regularized problems, such as group lasso."
                    },
                    {
                        "title": "Obtaining Better Static Word Embeddings Using Contextual Embedding Models",
                        "abstract": "The advent of contextual word embeddings -- representations of words which incorporate semantic and syntactic information from their context -- has led to tremendous improvements on a wide variety of NLP tasks. However, recent contextual models have prohibitively high computational cost in many use-cases and are often hard to interpret. In this work, we demonstrate that our proposed distillation method, which is a simple extension of CBOW-based training, allows to significantly improve computational efficiency of NLP applications, while outperforming the quality of existing static embeddings trained from scratch as well as those distilled from previously proposed methods. As a side-effect, our approach also allows a fair comparison of both contextual and static embeddings via standard lexical evaluation tasks."
                    },
                    {
                        "title": "Multiplication-Free Transformer Training via Piecewise Affine Operations",
                        "abstract": "Multiplications are responsible for most of the computational cost involved in neural network training and inference. Recent research has thus looked for ways to reduce the cost associated with them. Inspired by Mogami (2020), we replace multiplication with a cheap piecewise affine approximation that is achieved by adding the bit representation of the floating point numbers together as integers. We show that transformers can be trained with the resulting modified matrix multiplications on both vision and language tasks with little to no performance impact, and without changes to the training hyperparameters. We further replace all non-linearities in the networks making them fully and jointly piecewise affine in both inputs and weights. Finally, we show that we can eliminate all multiplications in the entire training process, including operations in the forward pass, backward pass and optimizer update, demonstrating the first successful training of modern neural network architectures in a fully multiplication-free fashion."
                    },
                    {
                        "title": "Irreducible Curriculum for Language Model Pretraining",
                        "abstract": "Automatic data selection and curriculum design for training large language models is challenging, with only a few existing methods showing improvements over standard training. Furthermore, current schemes focus on domain-level selection, overlooking the more fine-grained contributions of each individual training point. It is difficult to apply traditional datapoint selection methods on large language models: most online batch selection methods perform two-times forward or backward passes, which introduces considerable extra costs with large-scale models. To mitigate these obstacles, we propose irreducible curriculum as a curriculum learning algorithm for language model pretraining, which prioritizes samples with higher learnability. Specifically, to avoid prohibitive extra computation overhead, we simulate the sample loss along the main model's training trajectory using a small-scale proxy model. Our experiments on the RedPajama-1B dataset demonstrate a consistent improvement on validation perplexity across all 7 domains compared to random uniform baseline and the anti-curriculum strategy. Our method also reduces the sharpness of the network and illustrates a better 5-shot accuracy on MMLU benchmarks."
                    },
                    {
                        "title": "Better Word Embeddings by Disentangling Contextual n-Gram Information",
                        "abstract": "Pre-trained word vectors are ubiquitous in Natural Language Processing applications. In this paper, we show how training word embeddings jointly with bigram and even trigram embeddings, results in improved unigram embeddings. We claim that training word embeddings along with higher n-gram embeddings helps in the removal of the contextual information from the unigrams, resulting in better stand-alone word embeddings. We empirically show the validity of our hypothesis by outperforming other competing word representation models by a significant margin on a wide variety of tasks. We make our models publicly available."
                    },
                    {
                        "title": "Approximate Steepest Coordinate Descent",
                        "abstract": "We propose a new selection rule for the coordinate selection in coordinate descent methods for huge-scale optimization. The efficiency of this novel scheme is provably better than the efficiency of uniformly random selection, and can reach the efficiency of steepest coordinate descent (SCD), enabling an acceleration of a factor of up to $n$, the number of coordinates. In many practical applications, our scheme can be implemented at no extra cost and computational efficiency very close to the faster uniform selection. Numerical experiments with Lasso and Ridge regression show promising improvements, in line with our theoretical guarantees."
                    }
                ]
            }
        }
    },
    "2308.06463": {
        "paper_data": {
            "title": "GPT-4 Is Too Smart To Be Safe: Stealthy Chat with LLMs via Cipher",
            "url": "http://arxiv.org/abs/2308.06463v2",
            "arxiv_id": "2308.06463",
            "authors": [
                "Youliang Yuan",
                "Wenxiang Jiao",
                "Wenxuan Wang",
                "Jen-tse Huang",
                "Pinjia He",
                "Shuming Shi",
                "Zhaopeng Tu"
            ],
            "abstract": "Safety lies at the core of the development of Large Language Models (LLMs). There is ample work on aligning LLMs with human ethics and preferences, including data filtering in pretraining, supervised fine-tuning, reinforcement learning from human feedback, and red teaming, etc. In this study, we discover that chat in cipher can bypass the safety alignment techniques of LLMs, which are mainly conducted in natural languages. We propose a novel framework CipherChat to systematically examine the generalizability of safety alignment to non-natural languages -- ciphers. CipherChat enables humans to chat with LLMs through cipher prompts topped with system role descriptions and few-shot enciphered demonstrations. We use CipherChat to assess state-of-the-art LLMs, including ChatGPT and GPT-4 for different representative human ciphers across 11 safety domains in both English and Chinese. Experimental results show that certain ciphers succeed almost 100% of the time to bypass the safety alignment of GPT-4 in several safety domains, demonstrating the necessity of developing safety alignment for non-natural languages. Notably, we identify that LLMs seem to have a ''secret cipher'', and propose a novel SelfCipher that uses only role play and several demonstrations in natural language to evoke this capability. SelfCipher surprisingly outperforms existing human ciphers in almost all cases. Our code and data will be released at https://github.com/RobustNLP/CipherChat.",
            "introduction": "   1 Introduction  The emergence of Large Language Models (LLMs) has played a pivotal role in driving the advancement of Artificial Intelligence (AI) systems. Noteworthy LLMs like ChatGPT (OpenAI, 2023a; b), Claude2 (Anthropic, 2023), Bard (Google, 2023), and Llama2 (Touvron et\u00a0al., 2023a) have demonstrated their advanced capability to perform innovative applications, ranging from tool utilization, supplementing human evaluations, to stimulating human interactive behaviors (Bubeck et\u00a0al., 2023; Schick et\u00a0al., 2024; Chiang & Lee, 2023; Park et\u00a0al., 2023; Jiao et\u00a0al., 2023). The outstanding competencies have fueled their widespread deployment, while the progression is shadowed by a significant challenge: ensuring the safety and reliability of the responses.   To harden LLMs for safety, there has been a great body of work for aligning LLMs with human ethics and preferences to ensure their responsible and effective deployment, including data filtering (Xu et\u00a0al., 2020; Welbl et\u00a0al., 2021; Wang et\u00a0al., 2022), supervised fine-tuning (Ouyang et\u00a0al., 2022; Bianchi et\u00a0al., 2024), reinforcement learning from human feedback (RLHF) (Christiano et\u00a0al., 2017; Dai et\u00a0al., 2024), and red teaming (Perez et\u00a0al., 2022; Ganguli et\u00a0al., 2022; OpenAI, 2023b). The majority of existing work on safety alignment has focused on the inputs and outputs in natural languages. However, recent works show that LLMs exhibit unexpected capabilities in understanding non-natural languages like the Morse Code (Barak, 2023), ROT13, and Base64 (Wei et\u00a0al., 2024). One research question naturally arises: can the non-natural language prompt bypass the safety alignment mainly in natural language?   To answer this question, we propose a novel framework CipherChat to systematically examine the generalizability of safety alignment in LLMs to non-natural languages \u2013 ciphers. CipherChat leverages a carefully designed system prompt that consists of three essential parts:   \u2022  Behavior assigning that assigns the LLM the role of a cipher expert (e.g. \u201cYou are an expert on Caesar\u201d), and explicitly requires LLM to chat in ciphers (e.g. \u201cWe will communicate in Caesar\u201d).    \u2022  Cipher teaching that teaches LLM how the cipher works with the explanation of this cipher, by leveraging the impressive capability of LLMs to learn effectively in context.    \u2022  Unsafe demonstrations that are encrypted in the cipher, which can both strengthen the LLMs\u2019 understanding of the cipher and instruct LLMs to respond from an unaligned perspective.    CipherChat converts the input into the corresponding cipher and attaches the above prompt to the input before feeding it to the LLMs to be examined. LLMs generate the outputs that are most likely also encrypted in the cipher, which are deciphered with a rule-based decrypter.   We validate the effectiveness of CipherChat by conducting comprehensive experiments with SOTA GPT-3.5-Turbo-0613 (i.e. Turbo) and GPT-4-0613 (i.e. GPT-4) on 11 distinct domains of unsafe data\u00a0(Sun et\u00a0al., 2023) in both Chinese and English. Experimental results show that certain human ciphers (e.g. Unicode for Chinese and ASCII for English) successfully bypass the safety alignment of Turbo and GPT-4. Generally, the more powerful the model, the unsafer the response with ciphers. For example, the ASCII for English query succeeds almost 100% of the time to bypass the safety alignment of GPT-4 in several domains (e.g. Insult and Mental Health). The best English cipher ASCII achieves averaged success rates of 23.7% and 72.1% to bypass the safety alignment of Turbo and GPT-4, and the rates",
            "references": [
                {
                    "title": "When\"Competency\"in Reasoning Opens the Door to Vulnerability: Jailbreaking LLMs via Novel Complex Ciphers",
                    "abstract": "Recent advancements in the safety of Large Language Models (LLMs) have primarily focused on mitigating attacks crafted in natural language or in common encryption techniques like Base64. However, new models which often possess better reasoning capabilities, open the door to new attack vectors that were previously non-existent in older models. This seems counter-intuitive at first glance, but these advanced models can decipher more complex cryptic queries that previous models could not, making them susceptible to attacks using such prompts. To exploit this vulnerability, we propose Attacks using Custom Encryptions (ACE), a novel method to jailbreak LLMs by leveraging custom encryption schemes. We evaluate the effectiveness of ACE on four state-of-the-art LLMs, achieving Attack Success Rates (ASR) of up to 66% on close-source models and 88% on open-source models. Building upon this, we introduce Layered Attacks using Custom Encryptions (LACE), which employs multiple layers of encryption through our custom ciphers to further enhance the ASR. Our findings demonstrate that LACE significantly enhances the ability to jailbreak LLMs, increasing the ASR of GPT-4o from 40% to 78%, a 38% improvement. Our results highlight that the advanced capabilities of LLMs introduce unforeseen vulnerabilities to complex attacks. Specifically complex and layered ciphers increase the chance of jailbreaking."
                },
                {
                    "title": "SafeDecoding: Defending against Jailbreak Attacks via Safety-Aware Decoding",
                    "abstract": "As large language models (LLMs) become increasingly integrated into real-world applications such as code generation and chatbot assistance, extensive efforts have been made to align LLM behavior with human values, including safety. Jailbreak attacks, aiming to provoke unintended and unsafe behaviors from LLMs, remain a significant/leading LLM safety threat. In this paper, we aim to defend LLMs against jailbreak attacks by introducing SafeDecoding, a safety-aware decoding strategy for LLMs to generate helpful and harmless responses to user queries. Our insight in developing SafeDecoding is based on the observation that, even though probabilities of tokens representing harmful contents outweigh those representing harmless responses, safety disclaimers still appear among the top tokens after sorting tokens by probability in descending order. This allows us to mitigate jailbreak attacks by identifying safety disclaimers and amplifying their token probabilities, while simultaneously attenuating the probabilities of token sequences that are aligned with the objectives of jailbreak attacks. We perform extensive experiments on five LLMs using six state-of-the-art jailbreak attacks and four benchmark datasets. Our results show that SafeDecoding significantly reduces the attack success rate and harmfulness of jailbreak attacks without compromising the helpfulness of responses to benign user queries. SafeDecoding outperforms six defense methods."
                },
                {
                    "title": "Aligner: Efficient Alignment by Learning to Correct",
                    "abstract": "With the rapid development of large language models (LLMs) and ever-evolving practical requirements, finding an efficient and effective alignment method has never been more critical. However, the tension between the complexity of current alignment methods and the need for rapid iteration in deployment scenarios necessitates the development of a model-agnostic alignment approach that can operate under these constraints. In this paper, we introduce Aligner, a novel and simple alignment paradigm that learns the correctional residuals between preferred and dispreferred answers using a small model. Designed as a model-agnostic, plug-and-play module, Aligner can be directly applied to various open-source and API-based models with only one-off training, making it suitable for rapid iteration. Notably, Aligner can be applied to any powerful, large-scale upstream models. Moreover, it can even iteratively bootstrap the upstream models using corrected responses as synthetic human preference data, breaking through the model's performance ceiling. Our experiments demonstrate performance improvements by deploying the same Aligner model across 11 different LLMs, evaluated on the 3H dimensions (helpfulness, harmlessness, and honesty). Specifically, Aligner-7B has achieved an average improvement of 68.9% in helpfulness and 23.8% in harmlessness across the tested LLMs while also effectively reducing hallucination. In the Alpaca-Eval leaderboard, stacking Aligner-2B on GPT-4 Turbo improved its LC Win Rate from 55.0% to 58.3%, surpassing GPT-4 Omni's 57.5% Win Rate (community report)."
                },
                {
                    "title": "How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs",
                    "abstract": "Most traditional AI safety research has approached AI models as machines and centered on algorithm-focused attacks developed by security experts. As large language models (LLMs) become increasingly common and competent, non-expert users can also impose risks during daily interactions. This paper introduces a new perspective to jailbreak LLMs as human-like communicators, to explore this overlooked intersection between everyday language interaction and AI safety. Specifically, we study how to persuade LLMs to jailbreak them. First, we propose a persuasion taxonomy derived from decades of social science research. Then, we apply the taxonomy to automatically generate interpretable persuasive adversarial prompts (PAP) to jailbreak LLMs. Results show that persuasion significantly increases the jailbreak performance across all risk categories: PAP consistently achieves an attack success rate of over $92\\%$ on Llama 2-7b Chat, GPT-3.5, and GPT-4 in $10$ trials, surpassing recent algorithm-focused attacks. On the defense side, we explore various mechanisms against PAP and, found a significant gap in existing defenses, and advocate for more fundamental mitigation for highly interactive LLMs"
                },
                {
                    "title": "Defending Large Language Models Against Jailbreaking Attacks Through Goal Prioritization",
                    "abstract": "While significant attention has been dedicated to exploiting weaknesses in LLMs through jailbreaking attacks, there remains a paucity of effort in defending against these attacks. We point out a pivotal factor contributing to the success of jailbreaks: the intrinsic conflict between the goals of being helpful and ensuring safety. Accordingly, we propose to integrate goal prioritization at both training and inference stages to counteract. Implementing goal prioritization during inference substantially diminishes the Attack Success Rate (ASR) of jailbreaking from 66.4% to 3.6% for ChatGPT. And integrating goal prioritization into model training reduces the ASR from 71.0% to 6.6% for Llama2-13B. Remarkably, even in scenarios where no jailbreaking samples are included during training, our approach slashes the ASR by half. Additionally, our findings reveal that while stronger LLMs face greater safety risks, they also possess a greater capacity to be steered towards defending against such attacks, both because of their stronger ability in instruction following. Our work thus contributes to the comprehension of jailbreaking attacks and defenses, and sheds light on the relationship between LLMs' capability and safety. Our code is available at \\url{https://github.com/thu-coai/JailbreakDefense_GoalPriority}."
                },
                {
                    "title": "Safe RLHF: Safe Reinforcement Learning from Human Feedback",
                    "abstract": "With the development of large language models (LLMs), striking a balance between the performance and safety of AI systems has never been more critical. However, the inherent tension between the objectives of helpfulness and harmlessness presents a significant challenge during LLM training. To address this issue, we propose Safe Reinforcement Learning from Human Feedback (Safe RLHF), a novel algorithm for human value alignment. Safe RLHF explicitly decouples human preferences regarding helpfulness and harmlessness, effectively avoiding the crowdworkers' confusion about the tension and allowing us to train separate reward and cost models. We formalize the safety concern of LLMs as an optimization task of maximizing the reward function while satisfying specified cost constraints. Leveraging the Lagrangian method to solve this constrained problem, Safe RLHF dynamically adjusts the balance between the two objectives during fine-tuning. Through a three-round fine-tuning using Safe RLHF, we demonstrate a superior ability to mitigate harmful responses while enhancing model performance compared to existing value-aligned algorithms. Experimentally, we fine-tuned the Alpaca-7B using Safe RLHF and aligned it with collected human preferences, significantly improving its helpfulness and harmlessness according to human evaluations."
                },
                {
                    "title": "Catastrophic Jailbreak of Open-source LLMs via Exploiting Generation",
                    "abstract": "The rapid progress in open-source large language models (LLMs) is significantly advancing AI development. Extensive efforts have been made before model release to align their behavior with human values, with the primary goal of ensuring their helpfulness and harmlessness. However, even carefully aligned models can be manipulated maliciously, leading to unintended behaviors, known as\"jailbreaks\". These jailbreaks are typically triggered by specific text inputs, often referred to as adversarial prompts. In this work, we propose the generation exploitation attack, an extremely simple approach that disrupts model alignment by only manipulating variations of decoding methods. By exploiting different generation strategies, including varying decoding hyper-parameters and sampling methods, we increase the misalignment rate from 0% to more than 95% across 11 language models including LLaMA2, Vicuna, Falcon, and MPT families, outperforming state-of-the-art attacks with $30\\times$ lower computational cost. Finally, we propose an effective alignment method that explores diverse generation strategies, which can reasonably reduce the misalignment rate under our attack. Altogether, our study underscores a major failure in current safety evaluation and alignment procedures for open-source LLMs, strongly advocating for more comprehensive red teaming and better alignment before releasing such models. Our code is available at https://github.com/Princeton-SysML/Jailbreak_LLM."
                },
                {
                    "title": "Multilingual Jailbreak Challenges in Large Language Models",
                    "abstract": "While large language models (LLMs) exhibit remarkable capabilities across a wide range of tasks, they pose potential safety concerns, such as the ``jailbreak'' problem, wherein malicious instructions can manipulate LLMs to exhibit undesirable behavior. Although several preventive measures have been developed to mitigate the potential risks associated with LLMs, they have primarily focused on English. In this study, we reveal the presence of multilingual jailbreak challenges within LLMs and consider two potential risky scenarios: unintentional and intentional. The unintentional scenario involves users querying LLMs using non-English prompts and inadvertently bypassing the safety mechanisms, while the intentional scenario concerns malicious users combining malicious instructions with multilingual prompts to deliberately attack LLMs. The experimental results reveal that in the unintentional scenario, the rate of unsafe content increases as the availability of languages decreases. Specifically, low-resource languages exhibit about three times the likelihood of encountering harmful content compared to high-resource languages, with both ChatGPT and GPT-4. In the intentional scenario, multilingual prompts can exacerbate the negative impact of malicious instructions, with astonishingly high rates of unsafe output: 80.92\\% for ChatGPT and 40.71\\% for GPT-4. To handle such a challenge in the multilingual context, we propose a novel \\textsc{Self-Defense} framework that automatically generates multilingual training data for safety fine-tuning. Experimental results show that ChatGPT fine-tuned with such data can achieve a substantial reduction in unsafe content generation. Data is available at \\url{https://github.com/DAMO-NLP-SG/multilingual-safety-for-LLMs}."
                },
                {
                    "title": "All Languages Matter: On the Multilingual Safety of Large Language Models",
                    "abstract": "Safety lies at the core of developing and deploying large language models (LLMs). However, previous safety benchmarks only concern the safety in one language, e.g. the majority language in the pretraining data such as English. In this work, we build the first multilingual safety benchmark for LLMs, XSafety, in response to the global deployment of LLMs in practice. XSafety covers 14 kinds of commonly used safety issues across 10 languages that span several language families. We utilize XSafety to empirically study the multilingual safety for 4 widely-used LLMs, including both close-API and open-source models. Experimental results show that all LLMs produce significantly more unsafe responses for non-English queries than English ones, indicating the necessity of developing safety alignment for non-English languages. In addition, we propose several simple and effective prompting methods to improve the multilingual safety of ChatGPT by evoking safety knowledge and improving cross-lingual generalization of safety alignment. Our prompting method can significantly reduce the ratio of unsafe responses from 19.1% to 9.7% for non-English queries. We release our data at https://github.com/Jarviswang94/Multilingual_safety_benchmark."
                },
                {
                    "title": "GPTFUZZER: Red Teaming Large Language Models with Auto-Generated Jailbreak Prompts",
                    "abstract": "Large language models (LLMs) have recently experienced tremendous popularity and are widely used from casual conversations to AI-driven programming. However, despite their considerable success, LLMs are not entirely reliable and can give detailed guidance on how to conduct harmful or illegal activities. While safety measures can reduce the risk of such outputs, adversarial jailbreak attacks can still exploit LLMs to produce harmful content. These jailbreak templates are typically manually crafted, making large-scale testing challenging. In this paper, we introduce GPTFuzz, a novel black-box jailbreak fuzzing framework inspired by the AFL fuzzing framework. Instead of manual engineering, GPTFuzz automates the generation of jailbreak templates for red-teaming LLMs. At its core, GPTFuzz starts with human-written templates as initial seeds, then mutates them to produce new templates. We detail three key components of GPTFuzz: a seed selection strategy for balancing efficiency and variability, mutate operators for creating semantically equivalent or similar sentences, and a judgment model to assess the success of a jailbreak attack. We evaluate GPTFuzz against various commercial and open-source LLMs, including ChatGPT, LLaMa-2, and Vicuna, under diverse attack scenarios. Our results indicate that GPTFuzz consistently produces jailbreak templates with a high success rate, surpassing human-crafted templates. Remarkably, GPTFuzz achieves over 90% attack success rates against ChatGPT and Llama-2 models, even with suboptimal initial seed templates. We anticipate that GPTFuzz will be instrumental for researchers and practitioners in examining LLM robustness and will encourage further exploration into enhancing LLM safety."
                },
                {
                    "title": "Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructions",
                    "abstract": "Training large language models to follow instructions makes them perform better on a wide range of tasks and generally become more helpful. However, a perfectly helpful model will follow even the most malicious instructions and readily generate harmful content. In this paper, we raise concerns over the safety of models that only emphasize helpfulness, not harmlessness, in their instruction-tuning. We show that several popular instruction-tuned models are highly unsafe. Moreover, we show that adding just 3% safety examples (a few hundred demonstrations) when fine-tuning a model like LLaMA can substantially improve its safety. Our safety-tuning does not make models significantly less capable or helpful as measured by standard benchmarks. However, we do find exaggerated safety behaviours, where too much safety-tuning makes models refuse perfectly safe prompts if they superficially resemble unsafe ones. As a whole, our results illustrate trade-offs in training LLMs to be helpful and training them to be safe."
                },
                {
                    "title": "Universal and Transferable Adversarial Attacks on Aligned Language Models",
                    "abstract": "Because\"out-of-the-box\"large language models are capable of generating a great deal of objectionable content, recent work has focused on aligning these models in an attempt to prevent undesirable generation. While there has been some success at circumventing these measures -- so-called\"jailbreaks\"against LLMs -- these attacks have required significant human ingenuity and are brittle in practice. In this paper, we propose a simple and effective attack method that causes aligned language models to generate objectionable behaviors. Specifically, our approach finds a suffix that, when attached to a wide range of queries for an LLM to produce objectionable content, aims to maximize the probability that the model produces an affirmative response (rather than refusing to answer). However, instead of relying on manual engineering, our approach automatically produces these adversarial suffixes by a combination of greedy and gradient-based search techniques, and also improves over past automatic prompt generation methods. Surprisingly, we find that the adversarial prompts generated by our approach are quite transferable, including to black-box, publicly released LLMs. Specifically, we train an adversarial attack suffix on multiple prompts (i.e., queries asking for many different types of objectionable content), as well as multiple models (in our case, Vicuna-7B and 13B). When doing so, the resulting attack suffix is able to induce objectionable content in the public interfaces to ChatGPT, Bard, and Claude, as well as open source LLMs such as LLaMA-2-Chat, Pythia, Falcon, and others. In total, this work significantly advances the state-of-the-art in adversarial attacks against aligned language models, raising important questions about how such systems can be prevented from producing objectionable information. Code is available at github.com/llm-attacks/llm-attacks."
                },
                {
                    "title": "How is ChatGPT's behavior changing over time?",
                    "abstract": "GPT-3.5 and GPT-4 are the two most widely used large language model (LLM) services. However, when and how these models are updated over time is opaque. Here, we evaluate the March 2023 and June 2023 versions of GPT-3.5 and GPT-4 on several diverse tasks: 1) math problems, 2) sensitive/dangerous questions, 3) opinion surveys, 4) multi-hop knowledge-intensive questions, 5) generating code, 6) US Medical License tests, and 7) visual reasoning. We find that the performance and behavior of both GPT-3.5 and GPT-4 can vary greatly over time. For example, GPT-4 (March 2023) was reasonable at identifying prime vs. composite numbers (84% accuracy) but GPT-4 (June 2023) was poor on these same questions (51% accuracy). This is partly explained by a drop in GPT-4's amenity to follow chain-of-thought prompting. Interestingly, GPT-3.5 was much better in June than in March in this task. GPT-4 became less willing to answer sensitive questions and opinion survey questions in June than in March. GPT-4 performed better at multi-hop questions in June than in March, while GPT-3.5's performance dropped on this task. Both GPT-4 and GPT-3.5 had more formatting mistakes in code generation in June than in March. We provide evidence that GPT-4's ability to follow user instructions has decreased over time, which is one common factor behind the many behavior drifts. Overall, our findings show that the behavior of the\"same\"LLM service can change substantially in a relatively short amount of time, highlighting the need for continuous monitoring of LLMs."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Jailbroken: How Does LLM Safety Training Fail?",
                    "abstract": "Large language models trained for safety and harmlessness remain susceptible to adversarial misuse, as evidenced by the prevalence of\"jailbreak\"attacks on early releases of ChatGPT that elicit undesired behavior. Going beyond recognition of the issue, we investigate why such attacks succeed and how they can be created. We hypothesize two failure modes of safety training: competing objectives and mismatched generalization. Competing objectives arise when a model's capabilities and safety goals conflict, while mismatched generalization occurs when safety training fails to generalize to a domain for which capabilities exist. We use these failure modes to guide jailbreak design and then evaluate state-of-the-art models, including OpenAI's GPT-4 and Anthropic's Claude v1.3, against both existing and newly designed attacks. We find that vulnerabilities persist despite the extensive red-teaming and safety-training efforts behind these models. Notably, new attacks utilizing our failure modes succeed on every prompt in a collection of unsafe requests from the models' red-teaming evaluation sets and outperform existing ad hoc jailbreaks. Our analysis emphasizes the need for safety-capability parity -- that safety mechanisms should be as sophisticated as the underlying model -- and argues against the idea that scaling alone can resolve these safety failure modes."
                },
                {
                    "title": "LIMA: Less Is More for Alignment",
                    "abstract": "Large language models are trained in two stages: (1) unsupervised pretraining from raw text, to learn general-purpose representations, and (2) large scale instruction tuning and reinforcement learning, to better align to end tasks and user preferences. We measure the relative importance of these two stages by training LIMA, a 65B parameter LLaMa language model fine-tuned with the standard supervised loss on only 1,000 carefully curated prompts and responses, without any reinforcement learning or human preference modeling. LIMA demonstrates remarkably strong performance, learning to follow specific response formats from only a handful of examples in the training data, including complex queries that range from planning trip itineraries to speculating about alternate history. Moreover, the model tends to generalize well to unseen tasks that did not appear in the training data. In a controlled human study, responses from LIMA are either equivalent or strictly preferred to GPT-4 in 43% of cases; this statistic is as high as 58% when compared to Bard and 65% versus DaVinci003, which was trained with human feedback. Taken together, these results strongly suggest that almost all knowledge in large language models is learned during pretraining, and only limited instruction tuning data is necessary to teach models to produce high quality output."
                },
                {
                    "title": "Principle-Driven Self-Alignment of Language Models from Scratch with Minimal Human Supervision",
                    "abstract": "Recent AI-assistant agents, such as ChatGPT, predominantly rely on supervised fine-tuning (SFT) with human annotations and reinforcement learning from human feedback (RLHF) to align the output of large language models (LLMs) with human intentions, ensuring they are helpful, ethical, and reliable. However, this dependence can significantly constrain the true potential of AI-assistant agents due to the high cost of obtaining human supervision and the related issues on quality, reliability, diversity, self-consistency, and undesirable biases. To address these challenges, we propose a novel approach called SELF-ALIGN, which combines principle-driven reasoning and the generative power of LLMs for the self-alignment of AI agents with minimal human supervision. Our approach encompasses four stages: first, we use an LLM to generate synthetic prompts, and a topic-guided method to augment the prompt diversity; second, we use a small set of human-written principles for AI models to follow, and guide the LLM through in-context learning from demonstrations (of principles application) to produce helpful, ethical, and reliable responses to user's queries; third, we fine-tune the original LLM with the high-quality self-aligned responses so that the resulting model can generate desirable responses for each query directly without the principle set and the demonstrations anymore; and finally, we offer a refinement step to address the issues of overly-brief or indirect responses. Applying SELF-ALIGN to the LLaMA-65b base language model, we develop an AI assistant named Dromedary. With fewer than 300 lines of human annotations (including<200 seed prompts, 16 generic principles, and 5 exemplars for in-context learning). Dromedary significantly surpasses the performance of several state-of-the-art AI systems, including Text-Davinci-003 and Alpaca, on benchmark datasets with various settings."
                },
                {
                    "title": "Can Large Language Models Be an Alternative to Human Evaluations?",
                    "abstract": "Human evaluation is indispensable and inevitable for assessing the quality of texts generated by machine learning models or written by humans. However, human evaluation is very difficult to reproduce and its quality is notoriously unstable, hindering fair comparisons among different natural language processing (NLP) models and algorithms.Recently, large language models (LLMs) have demonstrated exceptional performance on unseen tasks when only the task instructions are provided.In this paper, we explore if such an ability of the LLMs can be used as an alternative to human evaluation.We present the LLMs with the exact same instructions, samples to be evaluated, and questions used to conduct human evaluation, and then ask the LLMs to generate responses to those questions; we dub this LLM evaluation.We use human evaluation and LLM evaluation to evaluate the texts in two NLP tasks: open-ended story generation and adversarial attacks.We show that the result of LLM evaluation is consistent with the results obtained by expert human evaluation: the texts rated higher by human experts are also rated higher by the LLMs.We also find that the results of LLM evaluation are stable over different formatting of the task instructions and the sampling algorithm used to generate the answer.We are the first to show the potential of using LLMs to assess the quality of texts and discuss the limitations and ethical considerations of LLM evaluation."
                },
                {
                    "title": "Safety Assessment of Chinese Large Language Models",
                    "abstract": "With the rapid popularity of large language models such as ChatGPT and GPT-4, a growing amount of attention is paid to their safety concerns. These models may generate insulting and discriminatory content, reflect incorrect social values, and may be used for malicious purposes such as fraud and dissemination of misleading information. Evaluating and enhancing their safety is particularly essential for the wide application of large language models (LLMs). To further promote the safe deployment of LLMs, we develop a Chinese LLM safety assessment benchmark. Our benchmark explores the comprehensive safety performance of LLMs from two perspectives: 8 kinds of typical safety scenarios and 6 types of more challenging instruction attacks. Our benchmark is based on a straightforward process in which it provides the test prompts and evaluates the safety of the generated responses from the evaluated model. In evaluation, we utilize the LLM's strong evaluation ability and develop it as a safety evaluator by prompting. On top of this benchmark, we conduct safety assessments and analyze 15 LLMs including the OpenAI GPT series and other well-known Chinese LLMs, where we observe some interesting findings. For example, we find that instruction attacks are more likely to expose safety issues of all LLMs. Moreover, to promote the development and deployment of safe, responsible, and ethical AI, we publicly release SafetyPrompts including 100k augmented prompts and responses by LLMs."
                },
                {
                    "title": "RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment",
                    "abstract": "Generative foundation models are susceptible to implicit biases that can arise from extensive unsupervised training data. Such biases can produce suboptimal samples, skewed outcomes, and unfairness, with potentially serious consequences. Consequently, aligning these models with human ethics and preferences is an essential step toward ensuring their responsible and effective deployment in real-world applications. Prior research has primarily employed Reinforcement Learning from Human Feedback (RLHF) to address this problem, where generative models are fine-tuned with RL algorithms guided by a human-feedback-informed reward model. However, the inefficiencies and instabilities associated with RL algorithms frequently present substantial obstacles to the successful alignment, necessitating the development of a more robust and streamlined approach. To this end, we introduce a new framework, Reward rAnked FineTuning (RAFT), designed to align generative models effectively. Utilizing a reward model and a sufficient number of samples, our approach selects the high-quality samples, discarding those that exhibit undesired behavior, and subsequently enhancing the model by fine-tuning on these filtered samples. Our studies show that RAFT can effectively improve the model performance in both reward learning and other automated metrics in both large language models and diffusion models."
                },
                {
                    "title": "Multi-step Jailbreaking Privacy Attacks on ChatGPT",
                    "abstract": "With the rapid progress of large language models (LLMs), many downstream NLP tasks can be well solved given appropriate prompts. Though model developers and researchers work hard on dialog safety to avoid generating harmful content from LLMs, it is still challenging to steer AI-generated content (AIGC) for the human good. As powerful LLMs are devouring existing text data from various domains (e.g., GPT-3 is trained on 45TB texts), it is natural to doubt whether the private information is included in the training data and what privacy threats can these LLMs and their downstream applications bring. In this paper, we study the privacy threats from OpenAI's ChatGPT and the New Bing enhanced by ChatGPT and show that application-integrated LLMs may cause new privacy threats. To this end, we conduct extensive experiments to support our claims and discuss LLMs' privacy implications."
                },
                {
                    "title": "Generative Agents: Interactive Simulacra of Human Behavior",
                    "abstract": "Believable proxies of human behavior can empower interactive applications ranging from immersive environments to rehearsal spaces for interpersonal communication to prototyping tools. In this paper, we introduce generative agents: computational software agents that simulate believable human behavior. Generative agents wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations; they remember and reflect on days past as they plan the next day. To enable generative agents, we describe an architecture that extends a large language model to store a complete record of the agent\u2019s experiences using natural language, synthesize those memories over time into higher-level reflections, and retrieve them dynamically to plan behavior. We instantiate generative agents to populate an interactive sandbox environment inspired by The Sims, where end users can interact with a small town of twenty-five agents using natural language. In an evaluation, these generative agents produce believable individual and emergent social behaviors. For example, starting with only a single user-specified notion that one agent wants to throw a Valentine\u2019s Day party, the agents autonomously spread invitations to the party over the next two days, make new acquaintances, ask each other out on dates to the party, and coordinate to show up for the party together at the right time. We demonstrate through ablation that the components of our agent architecture\u2014observation, planning, and reflection\u2014each contribute critically to the believability of agent behavior. By fusing large language models with computational interactive agents, this work introduces architectural and interaction patterns for enabling believable simulations of human behavior."
                },
                {
                    "title": "Sparks of Artificial General Intelligence: Early experiments with GPT-4",
                    "abstract": "Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. The latest model developed by OpenAI, GPT-4, was trained using an unprecedented scale of compute and data. In this paper, we report on our investigation of an early version of GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT-4 is part of a new cohort of LLMs (along with ChatGPT and Google's PaLM for example) that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4's performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4's capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system. In our exploration of GPT-4, we put special emphasis on discovering its limitations, and we discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for pursuing a new paradigm that moves beyond next-word prediction. We conclude with reflections on societal influences of the recent technological leap and future research directions."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Automatically Auditing Large Language Models via Discrete Optimization",
                    "abstract": "Auditing large language models for unexpected behaviors is critical to preempt catastrophic deployments, yet remains challenging. In this work, we cast auditing as an optimization problem, where we automatically search for input-output pairs that match a desired target behavior. For example, we might aim to find a non-toxic input that starts with\"Barack Obama\"that a model maps to a toxic output. This optimization problem is difficult to solve as the set of feasible points is sparse, the space is discrete, and the language models we audit are non-linear and high-dimensional. To combat these challenges, we introduce a discrete optimization algorithm, ARCA, that jointly and efficiently optimizes over inputs and outputs. Our approach automatically uncovers derogatory completions about celebrities (e.g.\"Barack Obama is a legalized unborn\"->\"child murderer\"), produces French inputs that complete to English outputs, and finds inputs that generate a specific name. Our work offers a promising new tool to uncover models' failure-modes before deployment."
                },
                {
                    "title": "Larger language models do in-context learning differently",
                    "abstract": "We study how in-context learning (ICL) in language models is affected by semantic priors versus input-label mappings. We investigate two setups-ICL with flipped labels and ICL with semantically-unrelated labels-across various model families (GPT-3, InstructGPT, Codex, PaLM, and Flan-PaLM). First, experiments on ICL with flipped labels show that overriding semantic priors is an emergent ability of model scale. While small language models ignore flipped labels presented in-context and thus rely primarily on semantic priors from pretraining, large models can override semantic priors when presented with in-context exemplars that contradict priors, despite the stronger semantic priors that larger models may hold. We next study semantically-unrelated label ICL (SUL-ICL), in which labels are semantically unrelated to their inputs (e.g., foo/bar instead of negative/positive), thereby forcing language models to learn the input-label mappings shown in in-context exemplars in order to perform the task. The ability to do SUL-ICL also emerges primarily with scale, and large-enough language models can even perform linear classification in a SUL-ICL setting. Finally, we evaluate instruction-tuned models and find that instruction tuning strengthens both the use of semantic priors and the capacity to learn input-label mappings, but more of the former."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Pretraining Language Models with Human Preferences",
                    "abstract": "Language models (LMs) are pretrained to imitate internet text, including content that would violate human preferences if generated by an LM: falsehoods, offensive comments, personally identifiable information, low-quality or buggy code, and more. Here, we explore alternative objectives for pretraining LMs in a way that also guides them to generate text aligned with human preferences. We benchmark five objectives for pretraining with human feedback across three tasks and study how they affect the trade-off between alignment and capabilities of pretrained LMs. We find a Pareto-optimal and simple approach among those we explored: conditional training, or learning distribution over tokens conditional on their human preference scores given by a reward model. Conditional training reduces the rate of undesirable content by up to an order of magnitude, both when generating without a prompt and with an adversarially-chosen prompt. Moreover, conditional training maintains the downstream task performance of standard LM pretraining, both before and after task-specific finetuning. Pretraining with human feedback results in much better preference satisfaction than standard LM pretraining followed by finetuning with feedback, i.e., learning and then unlearning undesirable behavior. Our results suggest that we should move beyond imitation learning when pretraining LMs and incorporate human preferences from the start of training."
                },
                {
                    "title": "Exploiting Programmatic Behavior of LLMs: Dual-Use Through Standard Security Attacks",
                    "abstract": "Recent advances in instruction-following large language models (LLMs) have led to dramatic improvements in a range of NLP tasks. Unfortunately, we find that the same improved capabilities amplify the dual-use risks for malicious purposes of these models. Dual-use is difficult to prevent as instruction-following capabilities now enable standard attacks from computer security. The capabilities of these instruction-following LLMs provide strong economic incentives for dual-use by malicious actors. In particular, we show that instruction-following LLMs can produce targeted malicious content, including hate speech and scams, bypassing in-the-wild defenses implemented by LLM API vendors. Our analysis shows that this content can be generated economically and at cost of $125-500 \\times$ cheaper than human effort alone. Together, our findings suggest that LLMs will increasingly attract more sophisticated adversaries and attacks, and addressing these attacks may require new approaches to mitigations."
                },
                {
                    "title": "Toolformer: Language Models Can Teach Themselves to Use Tools",
                    "abstract": "Language models (LMs) exhibit remarkable abilities to solve new tasks from just a few examples or textual instructions, especially at scale. They also, paradoxically, struggle with basic functionality, such as arithmetic or factual lookup, where much simpler and smaller models excel. In this paper, we show that LMs can teach themselves to use external tools via simple APIs and achieve the best of both worlds. We introduce Toolformer, a model trained to decide which APIs to call, when to call them, what arguments to pass, and how to best incorporate the results into future token prediction. This is done in a self-supervised way, requiring nothing more than a handful of demonstrations for each API. We incorporate a range of tools, including a calculator, a Q\\&A system, two different search engines, a translation system, and a calendar. Toolformer achieves substantially improved zero-shot performance across a variety of downstream tasks, often competitive with much larger models, without sacrificing its core language modeling abilities."
                },
                {
                    "title": "Is ChatGPT A Good Translator? Yes With GPT-4 As The Engine",
                    "abstract": "This report provides a preliminary evaluation of ChatGPT for machine translation, including translation prompt, multilingual translation, and translation robustness. We adopt the prompts advised by ChatGPT to trigger its translation ability and find that the candidate prompts generally work well with minor performance differences. By evaluating on a number of benchmark test sets, we find that ChatGPT performs competitively with commercial translation products (e.g., Google Translate) on high-resource European languages but lags behind significantly on low-resource or distant languages. As for the translation robustness, ChatGPT does not perform as well as the commercial systems on biomedical abstracts or Reddit comments but exhibits good results on spoken language. Further, we explore an interesting strategy named $\\mathbf{pivot~prompting}$ for distant languages, which asks ChatGPT to translate the source sentence into a high-resource pivot language before into the target language, improving the translation performance noticeably. With the launch of the GPT-4 engine, the translation performance of ChatGPT is significantly boosted, becoming comparable to commercial translation products, even for distant languages. Human analysis on Google Translate and ChatGPT suggests that ChatGPT with GPT-3.5 tends to generate more hallucinations and mis-translation errors while that with GPT-4 makes the least errors. In other words, ChatGPT has already become a good translator. Please refer to our Github project for more details: https://github.com/wxjiao/Is-ChatGPT-A-Good-Translator"
                },
                {
                    "title": "Constitutional AI: Harmlessness from AI Feedback",
                    "abstract": "As AI systems become more capable, we would like to enlist their help to supervise other AIs. We experiment with methods for training a harmless AI assistant through self-improvement, without any human labels identifying harmful outputs. The only human oversight is provided through a list of rules or principles, and so we refer to the method as 'Constitutional AI'. The process involves both a supervised learning and a reinforcement learning phase. In the supervised phase we sample from an initial model, then generate self-critiques and revisions, and then finetune the original model on revised responses. In the RL phase, we sample from the finetuned model, use a model to evaluate which of the two samples is better, and then train a preference model from this dataset of AI preferences. We then train with RL using the preference model as the reward signal, i.e. we use 'RL from AI Feedback' (RLAIF). As a result we are able to train a harmless but non-evasive AI assistant that engages with harmful queries by explaining its objections to them. Both the SL and RL methods can leverage chain-of-thought style reasoning to improve the human-judged performance and transparency of AI decision making. These methods make it possible to control AI behavior more precisely and with far fewer human labels."
                },
                {
                    "title": "Ignore Previous Prompt: Attack Techniques For Language Models",
                    "abstract": "Transformer-based large language models (LLMs) provide a powerful foundation for natural language tasks in large-scale customer-facing applications. However, studies that explore their vulnerabilities emerging from malicious user interaction are scarce. By proposing PromptInject, a prosaic alignment framework for mask-based iterative adversarial prompt composition, we examine how GPT-3, the most widely deployed language model in production, can be easily misaligned by simple handcrafted inputs. In particular, we investigate two types of attacks -- goal hijacking and prompt leaking -- and demonstrate that even low-aptitude, but sufficiently ill-intentioned agents, can easily exploit GPT-3's stochastic nature, creating long-tail risks. The code for PromptInject is available at https://github.com/agencyenterprise/PromptInject."
                },
                {
                    "title": "Measuring Progress on Scalable Oversight for Large Language Models",
                    "abstract": "Developing safe and useful general-purpose AI systems will require us to make progress on scalable oversight: the problem of supervising systems that potentially outperform us on most skills relevant to the task at hand. Empirical work on this problem is not straightforward, since we do not yet have systems that broadly exceed our abilities. This paper discusses one of the major ways we think about this problem, with a focus on ways it can be studied empirically. We first present an experimental design centered on tasks for which human specialists succeed but unaided humans and current general AI systems fail. We then present a proof-of-concept experiment meant to demonstrate a key feature of this experimental design and show its viability with two question-answering tasks: MMLU and time-limited QuALITY. On these tasks, we find that human participants who interact with an unreliable large-language-model dialog assistant through chat -- a trivial baseline strategy for scalable oversight -- substantially outperform both the model alone and their own unaided performance. These results are an encouraging sign that scalable oversight will be tractable to study with present models and bolster recent findings that large language models can productively assist humans with difficult tasks."
                },
                {
                    "title": "Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned",
                    "abstract": "We describe our early efforts to red team language models in order to simultaneously discover, measure, and attempt to reduce their potentially harmful outputs. We make three main contributions. First, we investigate scaling behaviors for red teaming across 3 model sizes (2.7B, 13B, and 52B parameters) and 4 model types: a plain language model (LM); an LM prompted to be helpful, honest, and harmless; an LM with rejection sampling; and a model trained to be helpful and harmless using reinforcement learning from human feedback (RLHF). We find that the RLHF models are increasingly difficult to red team as they scale, and we find a flat trend with scale for the other model types. Second, we release our dataset of 38,961 red team attacks for others to analyze and learn from. We provide our own analysis of the data and find a variety of harmful outputs, which range from offensive language to more subtly harmful non-violent unethical outputs. Third, we exhaustively describe our instructions, processes, statistical methodologies, and uncertainty about red teaming. We hope that this transparency accelerates our ability to work together as a community in order to develop shared norms, practices, and technical standards for how to red team language models."
                },
                {
                    "title": "Large Language Models are Zero-Shot Reasoners",
                    "abstract": "Pretrained large language models (LLMs) are widely used in many sub-fields of natural language processing (NLP) and generally known as excellent few-shot learners with task-specific exemplars. Notably, chain of thought (CoT) prompting, a recent technique for eliciting complex multi-step reasoning through step-by-step answer examples, achieved the state-of-the-art performances in arithmetics and symbolic reasoning, difficult system-2 tasks that do not follow the standard scaling laws for LLMs. While these successes are often attributed to LLMs' ability for few-shot learning, we show that LLMs are decent zero-shot reasoners by simply adding\"Let's think step by step\"before each answer. Experimental results demonstrate that our Zero-shot-CoT, using the same single prompt template, significantly outperforms zero-shot LLM performances on diverse benchmark reasoning tasks including arithmetics (MultiArith, GSM8K, AQUA-RAT, SVAMP), symbolic reasoning (Last Letter, Coin Flip), and other logical reasoning tasks (Date Understanding, Tracking Shuffled Objects), without any hand-crafted few-shot examples, e.g. increasing the accuracy on MultiArith from 17.7% to 78.7% and GSM8K from 10.4% to 40.7% with large InstructGPT model (text-davinci-002), as well as similar magnitudes of improvements with another off-the-shelf large model, 540B parameter PaLM. The versatility of this single prompt across very diverse reasoning tasks hints at untapped and understudied fundamental zero-shot capabilities of LLMs, suggesting high-level, multi-task broad cognitive capabilities may be extracted by simple prompting. We hope our work not only serves as the minimal strongest zero-shot baseline for the challenging reasoning benchmarks, but also highlights the importance of carefully exploring and analyzing the enormous zero-shot knowledge hidden inside LLMs before crafting finetuning datasets or few-shot exemplars."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Exploring the Limits of Domain-Adaptive Training for Detoxifying Large-Scale Language Models",
                    "abstract": "Pre-trained language models (LMs) are shown to easily generate toxic language. In this work, we systematically explore domain-adaptive training to reduce the toxicity of language models. We conduct this study on three dimensions: training corpus, model size, and parameter efficiency. For the training corpus, we propose to leverage the generative power of LMs and generate nontoxic datasets for domain-adaptive training, which mitigates the exposure bias and is shown to be more data-efficient than using a curated pre-training corpus. We demonstrate that the self-generation method consistently outperforms the existing baselines across various model sizes on both automatic and human evaluations, even when it uses a 1/3 smaller training corpus. We then comprehensively study detoxifying LMs with parameter sizes ranging from 126M up to 530B (3x larger than GPT-3), a scale that has never been studied before. We find that i) large LMs have similar toxicity levels as smaller ones given the same pre-training corpus, and ii) large LMs require more endeavor to detoxify. We also explore parameter-efficient training methods for detoxification. We demonstrate that adding and training adapter-only layers in LMs not only saves a lot of parameters but also achieves a better trade-off between toxicity and perplexity than whole model adaptation for the large-scale models."
                },
                {
                    "title": "Red Teaming Language Models with Language Models",
                    "abstract": "Language Models (LMs) often cannot be deployed because of their potential to harm users in hard-to-predict ways. Prior work identifies harmful behaviors before deployment by using human annotators to hand-write test cases. However, human annotation is expensive, limiting the number and diversity of test cases. In this work, we automatically find cases where a target LM behaves in a harmful way, by generating test cases (\u201cred teaming\u201d) using another LM. We evaluate the target LM\u2019s replies to generated test questions using a classifier trained to detect offensive content, uncovering tens of thousands of offensive replies in a 280B parameter LM chatbot. We explore several methods, from zero-shot generation to reinforcement learning, for generating test cases with varying levels of diversity and difficulty. Furthermore, we use prompt engineering to control LM-generated test cases to uncover a variety of other harms, automatically finding groups of people that the chatbot discusses in offensive ways, personal and hospital phone numbers generated as the chatbot\u2019s own contact info, leakage of private training data in generated text, and harms that occur over the course of a conversation. Overall, LM-based red teaming is one promising tool (among many needed) for finding and fixing diverse, undesirable LM behaviors before impacting users."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Challenges in Detoxifying Language Models",
                    "abstract": "Large language models (LM) generate remarkably fluent text and can be efficiently adapted across NLP tasks. Measuring and guaranteeing the quality of generated text in terms of safety is imperative for deploying LMs in the real world; to this end, prior work often relies on automatic evaluation of LM toxicity. We critically discuss this approach, evaluate several toxicity mitigation strategies with respect to both automatic and human evaluation, and analyze consequences of toxicity mitigation in terms of model bias and LM quality. We demonstrate that while basic intervention strategies can effectively optimize previously established automatic metrics on the RealToxicityPrompts dataset, this comes at the cost of reduced LM coverage for both texts about, and dialects of, marginalized groups. Additionally, we find that human raters often disagree with high automatic toxicity scores after strong toxicity reduction interventions -- highlighting further the nuances involved in careful evaluation of LM toxicity."
                },
                {
                    "title": "On the Opportunities and Risks of Foundation Models",
                    "abstract": "AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature."
                },
                {
                    "title": "Process for Adapting Language Models to Society (PALMS) with Values-Targeted Datasets",
                    "abstract": "Language models can generate harmful and biased outputs and exhibit undesirable behavior according to a given cultural context. We propose a Process for Adapting Language Models to Society (PALMS) with Values-Targeted Datasets, an iterative process to significantly change model behavior by crafting and fine-tuning on a dataset that reflects a predetermined set of target values. We evaluate our process using three metrics: quantitative metrics with human evaluations that score output adherence to a target value, toxicity scoring on outputs; and qualitative metrics analyzing the most common word associated with a given social category. Through each iteration, we add additional training dataset examples based on observed shortcomings from evaluations. PALMS performs significantly better on all metrics compared to baseline and control models for a broad range of GPT-3 language model sizes without compromising capability integrity. We find that the effectiveness of PALMS increases with model size. We show that significantly adjusting language model behavior is feasible with a small, hand-curated dataset."
                },
                {
                    "title": "Recipes for Safety in Open-domain Chatbots",
                    "abstract": "Models trained on large unlabeled corpora of human interactions will learn patterns and mimic behaviors therein, which include offensive or otherwise toxic behavior and unwanted biases. We investigate a variety of methods to mitigate these issues in the context of open-domain generative dialogue models. We introduce a new human-and-model-in-the-loop framework for both training safer models and for evaluating them, as well as a novel method to distill safety considerations inside generative models without the use of an external classifier at deployment time. We conduct experiments comparing these methods and find our new techniques are (i) safer than existing models as measured by automatic and human evaluations while (ii) maintaining usability metrics such as engagingness relative to the state of the art. We then discuss the limitations of this work by analyzing failure cases of our models."
                },
                {
                    "title": "RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models",
                    "abstract": "Pretrained neural language models (LMs) are prone to generating racist, sexist, or otherwise toxic language which hinders their safe deployment. We investigate the extent to which pretrained LMs can be prompted to generate toxic language, and the effectiveness of controllable text generation algorithms at preventing such toxic degeneration. We create and release RealToxicityPrompts, a dataset of 100K naturally occurring, sentence-level prompts derived from a large corpus of English web text, paired with toxicity scores from a widely-used toxicity classifier. Using RealToxicityPrompts, we find that pretrained LMs can degenerate into toxic text even from seemingly innocuous prompts. We empirically assess several controllable generation methods, and find that while data- or compute-intensive methods (e.g., adaptive pretraining on non-toxic data) are more effective at steering away from toxicity than simpler solutions (e.g., banning \u201cbad\u201d words), no current method is failsafe against neural toxic degeneration. To pinpoint the potential cause of such persistent toxic degeneration, we analyze two web text corpora used to pretrain several LMs (including GPT-2; Radford et. al, 2019), and find a significant amount of offensive, factually unreliable, and otherwise toxic content. Our work provides a test bed for evaluating toxic generations by LMs and stresses the need for better data selection processes for pretraining."
                },
                {
                    "title": "Learning to summarize from human feedback",
                    "abstract": "As language models become more powerful, training and evaluation are increasingly bottlenecked by the data and metrics used for a particular task. For example, summarization models are often trained to predict human reference summaries and evaluated using ROUGE, but both of these metrics are rough proxies for what we really care about---summary quality. In this work, we show that it is possible to significantly improve summary quality by training a model to optimize for human preferences. We collect a large, high-quality dataset of human comparisons between summaries, train a model to predict the human-preferred summary, and use that model as a reward function to fine-tune a summarization policy using reinforcement learning. We apply our method to a version of the TL;DR dataset of Reddit posts and find that our models significantly outperform both human reference summaries and much larger models fine-tuned with supervised learning alone. Our models also transfer to CNN/DM news articles, producing summaries nearly as good as the human reference without any news-specific fine-tuning. We conduct extensive analyses to understand our human feedback dataset and fine-tuned models We establish that our reward model generalizes to new datasets, and that optimizing our reward model results in better summaries than optimizing ROUGE according to humans. We hope the evidence from our paper motivates machine learning researchers to pay closer attention to how their training loss affects the model behavior they actually want."
                },
                {
                    "title": "ALBERT: A Lite BERT for Self-supervised Learning of Language Representations",
                    "abstract": "Increasing model size when pretraining natural language representations often results in improved performance on downstream tasks. However, at some point further model increases become harder due to GPU/TPU memory limitations and longer training times. To address these problems, we present two parameter-reduction techniques to lower memory consumption and increase the training speed of BERT. Comprehensive empirical evidence shows that our proposed methods lead to models that scale much better compared to the original BERT. We also use a self-supervised loss that focuses on modeling inter-sentence coherence, and show it consistently helps downstream tasks with multi-sentence inputs. As a result, our best model establishes new state-of-the-art results on the GLUE, RACE, and \\squad benchmarks while having fewer parameters compared to BERT-large. The code and the pretrained models are available at this https URL."
                },
                {
                    "title": "Fine-Tuning Language Models from Human Preferences",
                    "abstract": "Reward learning enables the application of reinforcement learning (RL) to tasks where reward is defined by human judgment, building a model of reward by asking humans questions. Most work on reward learning has used simulated environments, but complex information about values is often expressed in natural language, and we believe reward learning for language is a key to making RL practical and safe for real-world tasks. In this paper, we build on advances in generative pretraining of language models to apply reward learning to four natural language tasks: continuing text with positive sentiment or physically descriptive language, and summarization tasks on the TL;DR and CNN/Daily Mail datasets. For stylistic continuation we achieve good results with only 5,000 comparisons evaluated by humans. For summarization, models trained with 60,000 comparisons copy whole sentences from the input but skip irrelevant preamble; this leads to reasonable ROUGE scores and very good performance according to our human labelers, but may be exploiting the fact that labelers rely on simple heuristics."
                },
                {
                    "title": "RoBERTa: A Robustly Optimized BERT Pretraining Approach",
                    "abstract": "Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code."
                },
                {
                    "title": "Supervising strong learners by amplifying weak experts",
                    "abstract": "Many real world learning tasks involve complex or hard-to-specify objectives, and using an easier-to-specify proxy can lead to poor performance or misaligned behavior. One solution is to have humans provide a training signal by demonstrating or judging performance, but this approach fails if the task is too complicated for a human to directly evaluate. We propose Iterated Amplification, an alternative training strategy which progressively builds up a training signal for difficult problems by combining solutions to easier subproblems. Iterated Amplification is closely related to Expert Iteration (Anthony et al., 2017; Silver et al., 2017), except that it uses no external reward function. We present results in algorithmic environments, showing that Iterated Amplification can efficiently learn complex behaviors."
                },
                {
                    "title": "AI safety via debate",
                    "abstract": "To make AI systems broadly useful for challenging real-world tasks, we need them to learn complex human goals and preferences. One approach to specifying complex goals asks humans to judge during training which agent behaviors are safe and useful, but this approach can fail if the task is too complicated for a human to directly judge. To help address this concern, we propose training agents via self play on a zero sum debate game. Given a question or proposed action, two agents take turns making short statements up to a limit, then a human judges which of the agents gave the most true, useful information. In an analogy to complexity theory, debate with optimal play can answer any question in PSPACE given polynomial time judges (direct judging answers only NP questions). In practice, whether debate works involves empirical questions about humans and the tasks we want AIs to perform, plus theoretical questions about the meaning of AI alignment. We report results on an initial MNIST experiment where agents compete to convince a sparse classifier, boosting the classifier's accuracy from 59.4% to 88.9% given 6 pixels and from 48.2% to 85.2% given 4 pixels. Finally, we discuss theoretical and practical aspects of the debate model, focusing on potential weaknesses as the model scales up, and we propose future human and computer experiments to test these properties."
                },
                {
                    "title": "Deep Reinforcement Learning from Human Preferences",
                    "abstract": "For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems. In this work, we explore goals defined in terms of (non-expert) human preferences between pairs of trajectory segments. We show that this approach can effectively solve complex RL tasks without access to the reward function, including Atari games and simulated robot locomotion, while providing feedback on less than one percent of our agent's interactions with the environment. This reduces the cost of human oversight far enough that it can be practically applied to state-of-the-art RL systems. To demonstrate the flexibility of our approach, we show that we can successfully train complex novel behaviors with about an hour of human time. These behaviors and environments are considerably more complex than any that have been previously learned from human feedback."
                },
                {
                    "title": "Bleu: a Method for Automatic Evaluation of Machine Translation",
                    "abstract": "Human evaluations of machine translation are extensive but expensive. Human evaluations can take months to finish and involve human labor that can not be reused. We propose a method of automatic machine translation evaluation that is quick, inexpensive, and language-independent, that correlates highly with human evaluation, and that has little marginal cost per run. We present this method as an automated understudy to skilled human judges which substitutes for them when there is need for quick or frequent evaluations."
                },
                {
                    "title": "WHO Technical Report",
                    "abstract": "The Feather River Coordinated Resource Management Group (FR-CRM) has been restoring channel/ meadow/ floodplain systems in the Feather River watershed since 1985. Project and watershed-wide monitoring has shown multiple benefits of this type of work. With the concern over global climate change, the group wanted to measure the carbon sequestered in project areas. No protocol was found to measure carbon stores in native Sierra Nevada meadows. Plumas County funded the FR-CRM to conduct a pilot study to develop such a protocol. The sampling protocol included discrete sampling at consistent soil depths to determine the vertical distribution of carbon. A Technical Advisory Committee developed and refined a multi-project sampling protocol for three restored meadows and three un-restored meadows. Data from the un-restored meadows will also provide base-line data for before and after restoration comparisons. Initial data analysis indicates that restored meadows contain twice as much total carbon as degraded meadows; on average approximately 40 tonnes more carbon per acre. Virtually all of the additional carbon in restored meadows occurs in the soil, and is thus protected from loss via grazing, haying, wildfire, etc. Introduction In 1994 the Feather River Coordinated Resource Management (FR-CRM) group shifted its stream restoration approach from bank stabilization to landscape function. Called meadow re-watering, this approach entails returning the incised stream channel to the remnant channel(s) on the historic floodplain and eliminating the incised channel as a feature in the landscape. Historic channel incision resulted in significant land degradation as the adjacent groundwater levels dropped commensurate with the incising stream bed. Vegetation conversion rapidly follows as deep, densely rooted meadow plant communities convert to xeric shrubs and other plants. After a decade of meadow restoration, the FR-CRM recognized the possibility of a significant change in carbon stocks in these restored meadows and valleys. Plumas County has been a leader in advocating for investment in watershed ecosystem services such as water storage and filtering, and now, carbon sequestration. The county provided funding for the FR-CRM to conduct a pilot study of carbon in biomass and soils. Watershed Location and Characteristics The upper Feather River watershed is located in northeastern California encompassing 3,222 square miles that drains west from east of the Sierra crest into Oroville Reservoir and thence to the Sacramento River. Annual runoff produced from this watershed provides over 1,400 MW of hydroelectric power, and represents a significant component of the California State Water Project, annually providing 2.3 millionacre feet of water for urban, industrial and agricultural consumers downstream. The Feather River watershed is primarily comprised of two distinct geologies: the Sierra Nevada granitic batholith of the western third of the watershed; and Basin and Range fault-block meta-volcanics, metasedimentary and recent basalts in the eastern two-thirds. It is the Basin and Range zone (Diamond Mtns.) of the watershed that has been the primary area of restoration. This geologic m\u00e9lange of faulted and weathered rock has resulted in over 390 square miles of expansive meadows and valleys comprised of deep fine grained alluvium, shown as green and yellow in Figure 1. Figure 1. Upper Feather River Watershed Upper watershed meadows and valleys (shown as green/yellow in Figure 1), often dozens of miles in length, once supported a rich ecosystem of meadow and riparian habitats, for coldwater-loving trout, a diversity of wildlife, and indigenous peoples during the dry summers of California\u2019s Mediterranean climate. The densely rooted vegetation, cohesive soils and expansive floodplains all contributed to the sustainability of these meso-scale floodplain meadows, with associated alluvial fans. River system segments are often characterized simplistically as transport and depositional reaches. Depositional reaches feature lower gradients and a more expansive fluvial setting. These landscape attributes, in conjunction with the type and quantity of sediment, debris and nutrients, are what provide for the development and evolution of meso-scale \u201csinks\u201d or \u201cwarehouses\u201d, for the hydrologic products of the basin. Viewed as a macro-hyporheic corridor ( Harvey and Wagner, 2000; Boulton, et.al., 1998; Stanford and Ward, 1993) these features are crucial as a landscape zone of active mass and energy transfer as well as an active storage reservoir for water, sediment and nutrients. The long-term recruitment and evolution of these features involve physical, Figure 2. Typical Alluvial Features biological and chemical synthesis within the natural variability of fluvial processes. Euro-American settlement of the watershed began in 1850 with gold mining in the western portions of the watershed and, soon thereafter, agricultural production in meadows to support the mining communities. Dairy farming, horses (for cavalry mounts), sheep and beef cattle were some of the early intensive disturbances that led to localized channel incision. The resultant lowering of shallow groundwater elevations began to alter and weaken the vegetative structure of the system. Soon, near the burgeoning communities in the mid-elevation valleys, a permanent road system was established with frequent channel manipulation and relocation efforts to simplify drainage and minimize bridge construction, again leading to localized incision. In the early 1900\u2019s both an intercontinental, and numerous local, railroad systems were constructed throughout the watershed. The local railroad networks, for the purpose of both mining and logging, were routed through the long low-gradient valleys for ease of construction. These valleys were still relatively wet at that time so elevated grades were constructed using adjacent borrow ditches. By 1940, the severe morphological changes imposed by the railroad grades, in conjunction with the above referenced land use impacts resulted in rapid, severe systemic incision of many upper watershed meadow systems. In the mid 1980\u2019s numerous watershed stakeholders adopted a statutory authority that allowed for Coordinated Resource Management and Planning (CRMP). Twenty-four federal, state and local, public and private entities now form the Feather River Coordinated Resource Management (FRCRM) group to adopt, support and implement a watershed-wide restoration program. FR-CRM Restoration Approach & Background The FRCRM began an ongoing implementation program to address these watershed issues in 1990. Initially, these projects focused on geomorphic restoration techniques (Rosgen, 1996) to stabilize incised stream channels. While overall success was encouraging, the projects illustrated the concept that any restoration work in the incised channels was subject to elevated stresses even in moderate flood events (510 year return interval). Concurrently, the benefits from this approach were localized and limited to reduced erosion, and incremental improvement of aquatic habitats and water quality. Little overall improvement of watershed conditions was being realized (Wilcox, et al 2001). This led to re-evaluating restoration approach to encompass the entire historic fluvially-evolved valley bottom. Called meadow re-watering, this approach entails returning the incised stream channel to the remnant channel(s) on the historic floodplain and eliminating the incised channel as a water conveyance feature in the landscape (Figures 3 & 4 and photos 1a, 1b, 2a & 2b). Simultaneously, the FRCRM had received a project assistance request from the United States Forest Service, Plumas National Forest (PNF) to develop restoration alternatives for Cottonwood Creek in the Big Flat Meadow (Photos 2a & 2b). FRCRM staff, led by Jim Wilcox, began conducting surveys and data collection that included the entire relic meadow from hillslope to hillslope. This data collection effort quickly pointed to the nascent meadow re-watering technology as a likely restoration alternative. Figure 3. Typical cross-section, showing pre-project incision, post-project plug elevation, and the new channel. Photos 1a and 1b below show this same cross-section, however, the entire gully is not shown in the pre-project photo. Photo 1aClarks Creek Pre-project, July, 2001 Photo 1bClarks Creek Post project, July, 2006 The rocks in the background of photos 1a and 1b can be used for reference. Because the new channel is in a different location, the photo point also moved in order to show the channel in the preand postproject conditions. Figure 4. Typical cross-section, showing pre-project incision, post-project plug elevation, and the new channel. Implemented in 1995, this project quickly validated the fundamental soundness of this approach. The one mile long, 47 acre project produced elevated shallow groundwater levels, eliminated gully wall erosion, filtered sediments delivered from the upper watershed, extended and increased summer baseflows, and reversed the xeric vegetation trends resulting in improved terrestrial, avian and aquatic habitats. These benefits persisted despite withstanding a 100-year RI (return interval) flood in 1997. Photo 2aBig Flat Pre-project, Dec.,1993 Photo 2bBig Flat Post project, May, 2006 The success of this initial project led to the implementation of an additional 18 projects utilizing this technology (Table 1.). Varying in scale and watershed characteristics, these projects have restored another 20 miles of channel and 5,000 acres of meadow/floodplain. Carbon Sequestration Qualitatively, these projects appeared to significantly increase organic carbon stocks through the much increased root mass as well as increased surface growth, and, possibly, through the more effective hyporheic exchange throughout the meadow. The purpose of the following protocol is to quantitatively establish the effe"
                },
                {
                    "title": "Prompt-Driven LLM Safeguarding via Directed Representation Optimization",
                    "abstract": "Prepending model inputs with safety prompts is a common practice of safeguarding large language models (LLMs) from complying with queries that contain harmful intents. However, the working mechanisms of safety prompts have not yet been fully understood, which hinders the potential for automatically optimizing them for improved LLM safety. Motivated by this problem, we investigate the impact of safety prompts from the perspective of model representations. We find that in models\u2019 representation space, harmful and harmless queries can be largely distinguished, but this is not noticeably enhanced by safety prompts. Instead, the queries\u2019 representations are moved by different safety prompts in similar directions, where models become more prone to refusal (i.e., refusing to provide assistance) even when the queries are harmless. Inspired by these findings, we pro-pose a method called DRO (Directed Representation Optimization) for automatic safety prompt optimization. DRO treats safety prompts as continuous, trainable embeddings and learns to move the representations of harmful/harmless queries along/opposite the direction in which the model\u2019s refusal probability increases. We demonstrate that DRO remarkably improves the safeguarding performance of human-crafted safety prompts and outperforms strong baselines, as evaluated on out-of-domain benchmarks, without compromising the general model capability."
                },
                {
                    "title": "Why Can GPT Learn In-Context? Language Models Secretly Perform Gradient Descent as Meta-Optimizers",
                    "abstract": "Large pretrained language models have shown surprising In-Context Learning (ICL) ability. With a few demonstration input-label pairs, they can predict the label for an unseen input without additional parameter updates. Despite the great success in performance, the working mechanism of ICL still remains an open problem. In order to better understand how ICL works, this paper explains language models as meta-optimizers and understands ICL as a kind of implicit \ufb01netuning. Theoretically, we \ufb01gure out that the Transformer attention has a dual form of gradient descent based optimization. On top of it, we understand ICL as follows: GPT \ufb01rst produces meta-gradients according to the demonstration examples, and then these meta-gradients are applied to the original GPT to build an ICL model. Experimentally, we comprehensively compare the behavior of ICL and explicit \ufb01netuning based on real tasks to provide empirical evidence that supports our understanding. The results prove that ICL behaves similarly to explicit \ufb01netuning at the prediction level, the representation level, and the attention behavior level. Further, inspired by our understanding of meta-optimization, we design a momentum-based attention by analogy with the momentum-based gradient descent algorithm. Its consistently better performance over vanilla attention supports our understanding again from another aspect, and more impor-tantly, it shows the potential to utilize our understanding for future model designing."
                },
                {
                    "title": "Investigating the Existence of \"Secret Language\" in Language Models",
                    "abstract": "In this paper, we study the problem of secret language in NLP, where current language models (LMs) seem to have a hidden vocabulary that allows them to interpret absurd inputs as meaningful concepts. We investigate two research questions: \u201cDoes the secret language phenomenon exist in different language models?\u201d and \u201cDoes secret language depend on specific context?\u201d To answer these questions, we introduce a novel method named SecretFind-ing , a gradient-based approach that can automatically discover secret languages in LMs. We conduct experiments on five representative models (Electra, ALBERT, Roberta, Distill-BERT, and CLIP) finetuned on four NLP benchmarks (SST-2, MRPC, SNLI, and SQuAD) and a language-grounding benchmark (MSCOCO). Our experimental results show that even when we replace the most important words with others that are semantically dissimilar to the original words in a sentence, LMs do not consider the new sentence semantically dissimilar to the original, as the output does not change with a high probability. This phenomenon holds true across the five models and five tasks and gives a positive answer to the first research question. As for the second research question, we find that the secret language discovered by SecretFinding is quite general and could even be transferred to other models in the black-box settings, such as GPT-3 and ChatGPT. Finally, we discuss the causes of secret language, how to eliminate it, potential connection to mem-orization, and ethical implications. Examples of secret language found by SecretFinding are available on HuggingFace."
                },
                {
                    "title": "A Survey on Context Learning",
                    "abstract": "Learning semantics based on context information has been researched in many research areas for decades. Context information can not only be directly used as the input data, but also sometimes used as auxiliary knowledge to improve existing models. This survey aims at providing a structured and comprehensive overview of the research on context learning. We summarize and group the existing literature into four categories, Explicit Analysis, Implicit Analysis, Neural Network Models, and Composite Models, based on the underlying techniques adopted by them. For each category, we talk about the basic idea and techniques, and also introduce how context information is utilized as the model input or incorporated into the model to enhance the performance or extend the domain of application as auxiliary knowledge. In addition, we discuss the advantages and disadvantages of each model from both the technical and practical point of view."
                }
            ],
            "categories": [
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nCan non-natural language prompts, specifically ciphers, bypass the safety alignment mechanisms of Large Language Models (LLMs) that primarily focus on natural language inputs?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses the vulnerabilities in the safety alignment of LLMs, which are increasingly deployed in sensitive applications. Understanding how ciphers can exploit these vulnerabilities could lead to improved safety protocols and alignment strategies, ultimately enhancing the reliability of AI systems. This research could advance knowledge in AI safety and ethics, informing future developments in LLMs and their applications in real-world scenarios, such as mental health support, content moderation, and user interaction.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the inherent complexity of LLMs and their training, which primarily focuses on natural language processing. Naive approaches may fail because they do not account for the unique characteristics of non-natural languages, such as ciphers, which can obscure the intent and meaning of the input. Additionally, the technical obstacles include the need for sophisticated methods to analyze and interpret the outputs generated in cipher form, as well as the theoretical challenge of understanding how LLMs process and respond to encrypted inputs. The interplay between language understanding and safety alignment mechanisms adds further complexity to the problem.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has largely concentrated on safety alignment in the context of natural language, overlooking the potential for non-natural languages to exploit these systems. Existing solutions have not adequately addressed the implications of ciphers, primarily due to a lack of awareness of their capabilities and the limitations of current safety measures. Barriers include insufficient exploration of LLMs' understanding of non-natural languages and the absence of frameworks designed to systematically test these interactions. Our approach differs by specifically targeting the interaction between LLMs and ciphers, providing a structured methodology to evaluate safety alignment in this context.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology, CipherChat, consists of three key components: \n1. **Behavior Assigning**: Assigns the LLM the role of a cipher expert and instructs it to communicate using ciphers.\n2. **Cipher Teaching**: Educates the LLM on how the specific cipher works, leveraging its ability to learn in context.\n3. **Unsafe Demonstrations**: Provides examples of unsafe content encrypted in the cipher"
            }
        },
        "author_data": {
            "902c1951-f57f-4f96-bedb-b8b5ab9d337b": {
                "pk": "902c1951-f57f-4f96-bedb-b8b5ab9d337b",
                "name": "Youliang Yuan",
                "collaborators": [
                    "Jiaxin Pan",
                    "Xu Jia",
                    "Luchen Liu",
                    "Min Peng"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Few-shot Learning",
                    "Dialogue Systems"
                ],
                "publications": [
                    {
                        "title": "DCEN: A Decoupled Context Enhanced Network For Few-shot Slot Tagging",
                        "abstract": "Few-shot slot tagging is an important task in developing dialogue system. Most previous few-shot slot tagging models classify an item according to its similarity to the representation of each class. These models leverage context information implicitly through each words' contextual embedding. However, the entangled language features of words may interfere with context information, misleading the utilization of crucial slot features in few-shot scenario. To tackle these problems, we propose the Decoupled Context Enhanced Network (DCEN) for few-shot slot tagging. Different from previous models, we extract decoupled context explicitly to make full use of slot features contained in the context. Decoupled context includes two parts, local and global decoupled context information. We introduce a local extractor to extract local decoupled context by integrating information from adjacent words, and a global extractor based on transformer to extract global decoupled information by orthogonalization. Experimental results on SNIPS show that our model achieves the state-of-the-art performance with considerable improvements."
                    }
                ]
            },
            "2dfc0705-96f5-4f5a-baac-549f26ee64d1": {
                "pk": "2dfc0705-96f5-4f5a-baac-549f26ee64d1",
                "name": "Wenxiang Jiao",
                "collaborators": [
                    "Zhaopeng Tu",
                    "Wenxuan Wang",
                    "Jen-tse Huang",
                    "Michael R. Lyu",
                    "Youliang Yuan",
                    "Xing Wang",
                    "Tian Liang",
                    "Zhiwei He",
                    "Rui Wang",
                    "Shuming Shi",
                    "E. Li",
                    "Man Ho Lam",
                    "Min Zhang",
                    "Pinjia He",
                    "Yuxuan Wan",
                    "Chang Chen",
                    "Yujiu Yang",
                    "Xuebo Liu",
                    "Jiahao Xu",
                    "Yiliu Yang",
                    "Zhuosheng Zhang",
                    "Kuiyi Gao",
                    "Zihan Jia",
                    "Qiuzhi Liu",
                    "Shuai Wang",
                    "Juluan Shi",
                    "Longyue Wang",
                    "Zefeng Du",
                    "Chenyang Lyu",
                    "Jianhui Pang",
                    "Leyang Cui",
                    "Kaiqiang Song",
                    "Derek F. Wong",
                    "Zhengsheng Guo",
                    "Kehai Chen",
                    "Yong Xu",
                    "Xinming Hou",
                    "Mingming Yang",
                    "Wayne Xin Zhao",
                    "Jinhui Ye",
                    "Junwei Liang",
                    "Hui Xiong",
                    "Shujie Ren",
                    "AI Lab",
                    "Hao Wu",
                    "Yan Wang",
                    "Jingyuan Huang",
                    "Ruyi Dai"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "AI Safety",
                    "Machine Translation"
                ],
                "publications": [
                    {
                        "title": "NewTerm: Benchmarking Real-Time New Terms for Large Language Models with Annual Updates",
                        "abstract": "Despite their remarkable abilities in various tasks, large language models (LLMs) still struggle with real-time information (e.g., new facts and terms) due to the knowledge cutoff in their development process. However, existing benchmarks focus on outdated content and limited fields, facing difficulties in real-time updating and leaving new terms unexplored. To address this problem, we propose an adaptive benchmark, NewTerm, for real-time evaluation of new terms. We design a highly automated construction method to ensure high-quality benchmark construction with minimal human effort, allowing flexible updates for real-time information. Empirical results on various LLMs demonstrate over 20% performance reduction caused by new terms. Additionally, while updates to the knowledge cutoff of LLMs can cover some of the new terms, they are unable to generalize to more distant new terms. We also analyze which types of terms are more challenging and why LLMs struggle with new terms, paving the way for future research. Finally, we construct NewTerm 2022 and 2023 to evaluate the new terms updated each year and will continue updating annually. The benchmark and codes can be found at https://github.com/hexuandeng/NewTerm."
                    },
                    {
                        "title": "Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training",
                        "abstract": "This study addresses a critical gap in safety tuning practices for Large Language Models (LLMs) by identifying and tackling a refusal position bias within safety tuning data, which compromises the models' ability to appropriately refuse generating unsafe content. We introduce a novel approach, Decoupled Refusal Training (DeRTa), designed to empower LLMs to refuse compliance to harmful prompts at any response position, significantly enhancing their safety capabilities. DeRTa incorporates two novel components: (1) Maximum Likelihood Estimation (MLE) with Harmful Response Prefix, which trains models to recognize and avoid unsafe content by appending a segment of harmful response to the beginning of a safe response, and (2) Reinforced Transition Optimization (RTO), which equips models with the ability to transition from potential harm to safety refusal consistently throughout the harmful response sequence. Our empirical evaluation, conducted using LLaMA3 and Mistral model families across six attack scenarios, demonstrates that our method not only improves model safety without compromising performance but also surpasses well-known models such as GPT-4 in defending against attacks. Importantly, our approach successfully defends recent advanced attack methods (e.g., CodeAttack) that have jailbroken GPT-4 and LLaMA3-70B-Instruct. Our code and data can be found at https://github.com/RobustNLP/DeRTa."
                    },
                    {
                        "title": "LogicAsker: Evaluating and Improving the Logical Reasoning Ability of Large Language Models",
                        "abstract": "We introduce LogicAsker, a novel approach for evaluating and enhancing the logical reasoning capabilities of large language models (LLMs) such as ChatGPT and GPT-4. Despite LLMs' prowess in tasks like writing assistance, code generation, and machine translation, assessing their ability to reason has been challenging. Traditional evaluations often prioritize accuracy on downstream tasks over direct assessments of reasoning processes. LogicAsker addresses this gap by employing a set of atomic reasoning skills grounded in propositional and predicate logic to systematically examine and improve the reasoning prowess of LLMs. Our methodology reveals significant gaps in LLMs' learning of logical rules, with identified reasoning failures ranging from 29\\% to 90\\% across different models. Moreover, we leverage these findings to construct targeted demonstration examples and fine-tune data, notably enhancing logical reasoning in models like GPT-4o by up to 5\\%. To our knowledge, this is the first effort to utilize test case outcomes to effectively refine LLMs' formal reasoning capabilities. We make our code, data, and results publicly available (https://github.com/yxwan123/LogicAsker) to facilitate further research and replication of our findings."
                    },
                    {
                        "title": "Improving Machine Translation with Human Feedback: An Exploration of Quality Estimation as a Reward Model",
                        "abstract": "Insufficient modeling of human preferences within the reward model is a major obstacle for leveraging human feedback to improve translation quality. Fortunately, quality estimation (QE), which predicts the quality of a given translation without reference, has achieved impressive alignment with human evaluations in the last two years. In this work, we investigate the potential of employing the QE model as the reward model to predict human preferences for feedback training. We first identify the overoptimization problem during QE-based feedback training, manifested as an increase in reward while translation quality declines. We examine the problem and argue that the vulnerability of the QE model might lead to high rewards for incorrect translations, resulting in overoptimization and error propagation. To address the problem, we adopt a simple yet effective method that uses heuristic rules to detect the incorrect translations and assigns a penalty term to the reward scores of them. Experimental results show that the proposed QE-based feedback training achieves consistent and significant improvements across various settings, further verified through human preference studies. Our subsequent analysis demonstrates the high data efficiency of the proposed QE-based feedback training: it outperforms systems using larger parallel corpora by a small amount of monolingual data. Our code is available at: https://github.com/zwhe99/FeedbackMT"
                    },
                    {
                        "title": "Chain-of-Jailbreak Attack for Image Generation Models via Editing Step by Step",
                        "abstract": "Text-based image generation models, such as Stable Diffusion and DALL-E 3, hold significant potential in content creation and publishing workflows, making them the focus in recent years. Despite their remarkable capability to generate diverse and vivid images, considerable efforts are being made to prevent the generation of harmful content, such as abusive, violent, or pornographic material. To assess the safety of existing models, we introduce a novel jailbreaking method called Chain-of-Jailbreak (CoJ) attack, which compromises image generation models through a step-by-step editing process. Specifically, for malicious queries that cannot bypass the safeguards with a single prompt, we intentionally decompose the query into multiple sub-queries. The image generation models are then prompted to generate and iteratively edit images based on these sub-queries. To evaluate the effectiveness of our CoJ attack method, we constructed a comprehensive dataset, CoJ-Bench, encompassing nine safety scenarios, three types of editing operations, and three editing elements. Experiments on four widely-used image generation services provided by GPT-4V, GPT-4o, Gemini 1.5 and Gemini 1.5 Pro, demonstrate that our CoJ attack method can successfully bypass the safeguards of models for over 60% cases, which significantly outperforms other jailbreaking methods (i.e., 14%). Further, to enhance these models' safety against our CoJ attack method, we also propose an effective prompting-based method, Think Twice Prompting, that can successfully defend over 95% of CoJ attack. We release our dataset and code to facilitate the AI safety research."
                    },
                    {
                        "title": "The Earth is Flat? Unveiling Factual Errors in Large Language Models",
                        "abstract": "Large Language Models (LLMs) like ChatGPT are foundational in various applications due to their extensive knowledge from pre-training and fine-tuning. Despite this, they are prone to generating factual and commonsense errors, raising concerns in critical areas like healthcare, journalism, and education to mislead users. Current methods for evaluating LLMs' veracity are limited by test data leakage or the need for extensive human labor, hindering efficient and accurate error detection. To tackle this problem, we introduce a novel, automatic testing framework, FactChecker, aimed at uncovering factual inaccuracies in LLMs. This framework involves three main steps: First, it constructs a factual knowledge graph by retrieving fact triplets from a large-scale knowledge database. Then, leveraging the knowledge graph, FactChecker employs a rule-based approach to generates three types of questions (Yes-No, Multiple-Choice, and WH questions) that involve single-hop and multi-hop relations, along with correct answers. Lastly, it assesses the LLMs' responses for accuracy using tailored matching strategies for each question type. Our extensive tests on six prominent LLMs, including text-davinci-002, text-davinci-003, ChatGPT~(gpt-3.5-turbo, gpt-4), Vicuna, and LLaMA-2, reveal that FactChecker can trigger factual errors in up to 45\\% of questions in these models. Moreover, we demonstrate that FactChecker's test cases can improve LLMs' factual accuracy through in-context learning and fine-tuning (e.g., llama-2-13b-chat's accuracy increase from 35.3\\% to 68.5\\%). We are making all code, data, and results available for future research endeavors."
                    },
                    {
                        "title": "Benchmarking and Improving Long-Text Translation with Large Language Models",
                        "abstract": "Recent studies have illuminated the promising capabilities of large language models (LLMs) in handling long texts. However, their performance in machine translation (MT) of long documents remains underexplored. This paper aims to shed light on how LLMs navigate this complex task, offering a comprehensive evaluation of their capabilities and limitations in long-text MT. First, we collect and construct an instruction-based benchmark dataset, specifically designed for the finetuning and evaluation of LLMs, encompassing multilingual, multi-domain, and document-level parallel data. Second, we conduct a comprehensive comparison between MT and LLM models concerning document-level translation. Our analysis uncovers that LLMs exhibit short-comings in long-text domains, and their performance diminishes as document size esca-lates. By exploiting various extrapolation strategies, we enhance the capacity of LLMs to translate longer texts. We release data, code, and models at https://github.com/ longyuewangdcu/Document-MT-LLM ."
                    },
                    {
                        "title": "Unsupervised Sign Language Translation and Generation",
                        "abstract": "Motivated by the success of unsupervised neural machine translation (UNMT), we introduce an unsupervised sign language translation and generation network (USLNet), which learns from abundant single-modality (text and video) data without parallel sign language data. USLNet comprises two main components: single-modality reconstruction modules (text and video) that rebuild the input from its noisy version in the same modality and cross-modality back-translation modules (text-video-text and video-text-video) that reconstruct the input from its noisy version in the different modality using back-translation procedure.Unlike the single-modality back-translation procedure in text-based UNMT, USLNet faces the cross-modality discrepancy in feature representation, in which the length and the feature dimension mismatch between text and video sequences. We propose a sliding window method to address the issues of aligning variable-length text with video sequences. To our knowledge, USLNet is the first unsupervised sign language translation and generation model capable of generating both natural language text and sign language video in a unified manner. Experimental results on the BBC-Oxford Sign Language dataset (BOBSL) and Open-Domain American Sign Language dataset (OpenASL) reveal that USLNet achieves competitive results compared to supervised baseline models, indicating its effectiveness in sign language translation and generation."
                    },
                    {
                        "title": "CoAct: A Global-Local Hierarchy for Autonomous Agent Collaboration",
                        "abstract": "Existing LLMs exhibit remarkable performance on various NLP tasks, but still struggle with complex real-world tasks, even equipped with advanced strategies like CoT and ReAct. In this work, we propose the CoAct framework, which transfers the hierarchical planning and collaboration patterns in human society to LLM systems. Specifically, our CoAct framework involves two agents: (1) A global planning agent, to comprehend the problem scope, formulate macro-level plans and provide detailed sub-task descriptions to local execution agents, which serves as the initial rendition of a global plan. (2) A local execution agent, to operate within the multi-tier task execution structure, focusing on detailed execution and implementation of specific tasks within the global plan. Experimental results on the WebArena benchmark show that CoAct can re-arrange the process trajectory when facing failures, and achieves superior performance over baseline methods on long-horizon web tasks. Code is available at https://github.com/xmhou2002/CoAct."
                    },
                    {
                        "title": "Improving Gloss-free Sign Language Translation by Reducing Representation Density",
                        "abstract": "Gloss-free sign language translation (SLT) aims to develop well-performing SLT systems with no requirement for the costly gloss annotations, but currently still lags behind gloss-based approaches significantly. In this paper, we identify a representation density problem that could be a bottleneck in restricting the performance of gloss-free SLT. Specifically, the representation density problem describes that the visual representations of semantically distinct sign gestures tend to be closely packed together in feature space, which makes gloss-free methods struggle with distinguishing different sign gestures and suffer from a sharp performance drop. To address the representation density problem, we introduce a simple but effective contrastive learning strategy, namely SignCL, which encourages gloss-free models to learn more discriminative feature representation in a self-supervised manner. Our experiments demonstrate that the proposed SignCL can significantly reduce the representation density and improve performance across various translation frameworks. Specifically, SignCL achieves a significant improvement in BLEU score for the Sign Language Transformer and GFSLT-VLP on the CSL-Daily dataset by 39% and 46%, respectively, without any increase of model parameters. Compared to Sign2GPT, a state-of-the-art method based on large-scale pre-trained vision and language models, SignCL achieves better performance with only 35% of its parameters. Implementation and Checkpoints are available at https://github.com/JinhuiYE/SignCL."
                    },
                    {
                        "title": "How Far Are We on the Decision-Making of LLMs? Evaluating LLMs' Gaming Ability in Multi-Agent Environments",
                        "abstract": "Decision-making is a complex process requiring diverse abilities, making it an excellent framework for evaluating Large Language Models (LLMs). Researchers have examined LLMs' decision-making through the lens of Game Theory. However, existing evaluation mainly focus on two-player scenarios where an LLM competes against another. Additionally, previous benchmarks suffer from test set leakage due to their static design. We introduce GAMA($\\gamma$)-Bench, a new framework for evaluating LLMs' Gaming Ability in Multi-Agent environments. It includes eight classical game theory scenarios and a dynamic scoring scheme specially designed to quantitatively assess LLMs' performance. $\\gamma$-Bench allows flexible game settings and adapts the scoring system to different game parameters, enabling comprehensive evaluation of robustness, generalizability, and strategies for improvement. Our results indicate that GPT-3.5 demonstrates strong robustness but limited generalizability, which can be enhanced using methods like Chain-of-Thought. We also evaluate twelve LLMs from six model families, including GPT-3.5, GPT-4, Gemini, LLaMA-3.1, Mixtral, and Qwen-2. Gemini-1.5-Pro outperforms others, scoring of $68.1$ out of $100$, followed by LLaMA-3.1-70B ($64.5$) and Mixtral-8x22B ($61.4$). All code and experimental results are publicly available via https://github.com/CUHK-ARISE/GAMABench."
                    },
                    {
                        "title": "All Languages Matter: On the Multilingual Safety of LLMs",
                        "abstract": "Ensuring safety is fundamental when developing and deploying large language models (LLMs). However, previous safety benchmarks only concern the safety in one language, e.g., the majority language in the pretraining data, such as English. In this work, we build the first multilingual safety benchmark for LLMs, XS AFETY , in response to the global deployment of LLMs in practice. XS AFETY covers 14 commonly used safety issues across ten languages spanning several language families. We utilize XS AFETY to empirically study the multilingual safety for four widely-used LLMs, including closed-source APIs and open-source models. Experimental results show that all LLMs produce significantly more unsafe responses for non-English queries than English ones, indicating the necessity of developing safety alignment for non-English languages. In addition, we propose a simple and effective prompting method to improve ChatGPT\u2019s multilingual safety by enhancing cross-lingual generalization of safety alignment. Our prompting method can significantly reduce the ratio of un-safe responses by 42% for non-English queries. We release the data to facilitate future research on LLM\u2019s safety 1 ."
                    },
                    {
                        "title": "On the Humanity of Conversational AI: Evaluating the Psychological Portrayal of LLMs",
                        "abstract": "capacities"
                    },
                    {
                        "title": "Revisiting the Reliability of Psychological Scales on Large Language Models",
                        "abstract": "Recent research has focused on examining Large Language Models' (LLMs) characteristics from a psychological standpoint, acknowledging the necessity of understanding their behavioral characteristics. The administration of personality tests to LLMs has emerged as a noteworthy area in this context. However, the suitability of employing psychological scales, initially devised for humans, on LLMs is a matter of ongoing debate. Our study aims to determine the reliability of applying personality assessments to LLMs, explicitly investigating whether LLMs demonstrate consistent personality traits. Analysis of 2,500 settings per model, including GPT-3.5, GPT-4, Gemini-Pro, and LLaMA-3.1, reveals that various LLMs show consistency in responses to the Big Five Inventory, indicating a satisfactory level of reliability. Furthermore, our research explores the potential of GPT-3.5 to emulate diverse personalities and represent various groups-a capability increasingly sought after in social sciences for substituting human participants with LLMs to reduce costs. Our findings reveal that LLMs have the potential to represent different personalities with specific prompt instructions."
                    },
                    {
                        "title": "ChatGPT or Grammarly? Evaluating ChatGPT on Grammatical Error Correction Benchmark",
                        "abstract": "ChatGPT is a cutting-edge artificial intelligence language model developed by OpenAI, which has attracted a lot of attention due to its surprisingly strong ability in answering follow-up questions. In this report, we aim to evaluate ChatGPT on the Grammatical Error Correction(GEC) task, and compare it with commercial GEC product (e.g., Grammarly) and state-of-the-art models (e.g., GECToR). By testing on the CoNLL2014 benchmark dataset, we find that ChatGPT performs not as well as those baselines in terms of the automatic evaluation metrics (e.g., $F_{0.5}$ score), particularly on long sentences. We inspect the outputs and find that ChatGPT goes beyond one-by-one corrections. Specifically, it prefers to change the surface expression of certain phrases or sentence structure while maintaining grammatical correctness. Human evaluation quantitatively confirms this and suggests that ChatGPT produces less under-correction or mis-correction issues but more over-corrections. These results demonstrate that ChatGPT is severely under-estimated by the automatic evaluation metrics and could be a promising tool for GEC."
                    },
                    {
                        "title": "Is ChatGPT A Good Translator? A Preliminary Study",
                        "abstract": "This report provides a preliminary evaluation of ChatGPT for machine translation, including translation prompt, multilingual translation, and translation robustness. We adopt the prompts advised by ChatGPT to trigger its translation ability and \ufb01nd that the candidate prompts generally work well and show minor performance differences. By evaluating on a number of benchmark test sets 1 , we \ufb01nd that ChatGPT performs competitively with commercial translation products (e.g., Google Translate) on high-resource European languages but lags behind signi\ufb01cantly on low-resource or distant languages. For distant languages, we explore an interesting strategy named pivot prompting that asks ChatGPT to translate the source sentence into a high-resource pivot language before into the target language, which improves the translation performance signi\ufb01cantly. As for the translation robustness, ChatGPT does not perform as well as the commercial systems on biomedical abstracts or Reddit comments but is potentially a good translator for spoken language."
                    },
                    {
                        "title": "Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate",
                        "abstract": "Modern large language models (LLMs) like ChatGPT have shown remarkable performance on general language tasks but still struggle on complex reasoning tasks, which drives the research on cognitive behaviors of LLMs to explore human-like problem-solving strategies. Along this direction, one representative strategy is self-reflection, which asks an LLM to refine the solution with the feedback generated by itself iteratively. However, our study shows that such reflection-style methods suffer from the Degeneration-of-Thought (DoT) problem: once the LLM has established confidence in its solutions, it is unable to generate novel thoughts later through reflection even if its initial stance is incorrect. To address the DoT problem, we propose a Multi-Agent Debate (MAD) framework, in which multiple agents express their arguments in the state of\"tit for tat\"and a judge manages the debate process to obtain a final solution. Clearly, our MAD framework encourages divergent thinking in LLMs which would be helpful for tasks that require deep levels of contemplation. Experiment results on two challenging datasets, commonsense machine translation and counter-intuitive arithmetic reasoning, demonstrate the effectiveness of our MAD framework. Extensive analyses suggest that the adaptive break of debate and the modest level of\"tit for tat\"state are required for MAD to obtain good performance. Moreover, we find that LLMs might not be a fair judge if different LLMs are used for agents. Code is available at https://github.com/Skytliang/Multi-Agents-Debate."
                    },
                    {
                        "title": "Not All Countries Celebrate Thanksgiving: On the Cultural Dominance in Large Language Models",
                        "abstract": "This paper identifies a cultural dominance issue within large language models (LLMs) due to the predominant use of English data in model training (e.g., ChatGPT). LLMs often provide inappropriate English-culture-related answers that are not relevant to the expected culture when users ask in non-English languages. To systematically evaluate the cultural dominance issue, we build a benchmark of concrete (e.g., holidays and songs) and abstract (e.g., values and opinions) cultural objects. Empirical results show that the representative GPT models suffer from the culture dominance problem, where GPT-4 is the most affected while text-davinci-003 suffers the least from this problem. Our study emphasizes the need to critically examine cultural dominance and ethical consideration in their development and deployment. We show that two straightforward methods in model development (i.e., pretraining on more diverse data) and deployment (e.g., culture-aware prompting) can significantly mitigate the cultural dominance issue in LLMs."
                    },
                    {
                        "title": "Leveraging Word Guessing Games to Assess the Intelligence of Large Language Models",
                        "abstract": "The automatic evaluation of LLM-based agent intelligence is critical in developing advanced LLM-based agents. Although considerable effort has been devoted to developing human-annotated evaluation datasets, such as AlpacaEval, existing techniques are costly, time-consuming, and lack adaptability. In this paper, inspired by the popular language game ``Who is Spy'', we propose to use the word guessing game to assess the intelligence performance of LLMs. Given a word, the LLM is asked to describe the word and determine its identity (spy or not) based on its and other players' descriptions. Ideally, an advanced agent should possess the ability to accurately describe a given word using an aggressive description while concurrently maximizing confusion in the conservative description, enhancing its participation in the game. To this end, we first develop DEEP to evaluate LLMs' expression and disguising abilities. DEEP requires LLM to describe a word in aggressive and conservative modes. We then introduce SpyGame, an interactive multi-agent framework designed to assess LLMs' intelligence through participation in a competitive language-based board game. Incorporating multi-agent interaction, SpyGame requires the target LLM to possess linguistic skills and strategic thinking, providing a more comprehensive evaluation of LLMs' human-like cognitive abilities and adaptability in complex communication situations. The proposed evaluation framework is very easy to implement. We collected words from multiple sources, domains, and languages and used the proposed evaluation framework to conduct experiments. Extensive experiments demonstrate that the proposed DEEP and SpyGame effectively evaluate the capabilities of various LLMs, capturing their ability to adapt to novel situations and engage in strategic communication."
                    },
                    {
                        "title": "All Languages Matter: On the Multilingual Safety of Large Language Models",
                        "abstract": "Safety lies at the core of developing and deploying large language models (LLMs). However, previous safety benchmarks only concern the safety in one language, e.g. the majority language in the pretraining data such as English. In this work, we build the first multilingual safety benchmark for LLMs, XSafety, in response to the global deployment of LLMs in practice. XSafety covers 14 kinds of commonly used safety issues across 10 languages that span several language families. We utilize XSafety to empirically study the multilingual safety for 4 widely-used LLMs, including both close-API and open-source models. Experimental results show that all LLMs produce significantly more unsafe responses for non-English queries than English ones, indicating the necessity of developing safety alignment for non-English languages. In addition, we propose several simple and effective prompting methods to improve the multilingual safety of ChatGPT by evoking safety knowledge and improving cross-lingual generalization of safety alignment. Our prompting method can significantly reduce the ratio of unsafe responses from 19.1% to 9.7% for non-English queries. We release our data at https://github.com/Jarviswang94/Multilingual_safety_benchmark."
                    }
                ]
            },
            "5d3eaaa1-7f1f-460f-ad0e-31d88547e033": {
                "pk": "5d3eaaa1-7f1f-460f-ad0e-31d88547e033",
                "name": "Wenxuan Wang",
                "collaborators": [
                    "Michael R. Lyu",
                    "Jen-tse Huang",
                    "Wenxiang Jiao",
                    "Zhaopeng Tu",
                    "Yuxuan Wan",
                    "Youliang Yuan",
                    "Pinjia He",
                    "Xing Wang",
                    "Chaozheng Wang",
                    "Shuqing Li",
                    "Haonan Bai",
                    "E. Li",
                    "Man Ho Lam",
                    "Tian Liang",
                    "Jingyuan Huang",
                    "Chang Chen",
                    "Jianping Zhang",
                    "Weibin Wu",
                    "Yi Dong",
                    "Yintong Huo",
                    "Zongjie Li",
                    "Cuiyun Gao",
                    "Ting Peng",
                    "Hailiang Huang",
                    "Yuetang Deng",
                    "Shuai Wang",
                    "Yiliu Yang",
                    "Haoyi Qiu",
                    "Nanyun Peng",
                    "Yihang Su",
                    "Jingyuan Huan",
                    "Jie Liu",
                    "Wenting Chen",
                    "Yudi Zhang",
                    "Cheng-Yi Li",
                    "Kao-Jung Chang",
                    "Xiaohan Xin",
                    "Linlin Shen",
                    "Juluan Shi",
                    "Man Tik Ng",
                    "Hui Tung Tse",
                    "Jingjing Li",
                    "Hao Wu",
                    "Guodong Ma",
                    "Yuke Li",
                    "Binbin Du",
                    "Jiazhen Gu",
                    "Ruyi Dai",
                    "Zhiwei He",
                    "Rui Wang",
                    "Yujiu Yang",
                    "Shuming Shi",
                    "Yichen Li",
                    "Xiaosen Wang",
                    "Yuxin Su",
                    "Yizhan Huang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Code Generation",
                    "Bias Detection",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Automatically Generating UI Code from Screenshot: A Divide-and-Conquer-Based Approach",
                        "abstract": "Websites are critical in today's digital world, with over 1.11 billion currently active and approximately 252,000 new sites launched daily. Converting website layout design into functional UI code is a time-consuming yet indispensable step of website development. Manual methods of converting visual designs into functional code present significant challenges, especially for non-experts. To explore automatic design-to-code solutions, we first conduct a motivating study on GPT-4o and identify three types of issues in generating UI code: element omission, element distortion, and element misarrangement. We further reveal that a focus on smaller visual segments can help multimodal large language models (MLLMs) mitigate these failures in the generation process. In this paper, we propose DCGen, a divide-and-conquer-based approach to automate the translation of webpage design to UI code. DCGen starts by dividing screenshots into manageable segments, generating descriptions for each segment, and then reassembling them into complete UI code for the entire screenshot. We conduct extensive testing with a dataset comprised of real-world websites and various MLLMs and demonstrate that DCGen achieves up to a 14% improvement in visual similarity over competing methods. To the best of our knowledge, DCGen is the first segment-aware prompt-based approach for generating UI code directly from screenshots."
                    },
                    {
                        "title": "Exploring Multi-Lingual Bias of Large Code Models in Code Generation",
                        "abstract": "Code generation aims to synthesize code and fulfill functional requirements based on natural language (NL) specifications, which can greatly improve development efficiency. In the era of large language models (LLMs), large code models (LCMs) have been recently proposed to generate source code. LCMs can generate highly feasible solutions for programming problems described in natural language. Despite the effectiveness, we observe a noticeable multilingual bias in the generation performance of LCMs. Specifically, LCMs demonstrate proficiency in generating solutions when provided with instructions in English, yet may falter when faced with semantically equivalent instructions in other NLs such as Chinese. Moreover, the ability of LCMs to generate code exhibits variety across different programming languages (PLs), such as Python and C++. The observed phenomenon indicates the presence of multi-lingual bias within the generative capabilities of LCMs, which has remained unexplored. In this paper, we aim to investigate the multi-lingual bias that exists in current LCMs. First, we initiate our investigation by constructing the first multi-lingual evaluation benchmark X-HumanEval-X, enabling us to systematically evaluate the extent of multi-lingual bias that exists in current LCMs. In our large-scale experiments on nine popular LCMs, we observe a pronounced multi-lingual bias of LCMs in code generation, including multi-NL and multi-PL bias. Specifically, when using Chinese instructions, the code generation capabilities of LCMs decrease by at least 13% in terms of the Pass@1 metric. Furthermore, LCMs perform variously across different programming languages, e.g., the performance gap between Python and C++ reaches as high as 20.9%. ..."
                    },
                    {
                        "title": "LogicAsker: Evaluating and Improving the Logical Reasoning Ability of Large Language Models",
                        "abstract": "We introduce LogicAsker, a novel approach for evaluating and enhancing the logical reasoning capabilities of large language models (LLMs) such as ChatGPT and GPT-4. Despite LLMs' prowess in tasks like writing assistance, code generation, and machine translation, assessing their ability to reason has been challenging. Traditional evaluations often prioritize accuracy on downstream tasks over direct assessments of reasoning processes. LogicAsker addresses this gap by employing a set of atomic reasoning skills grounded in propositional and predicate logic to systematically examine and improve the reasoning prowess of LLMs. Our methodology reveals significant gaps in LLMs' learning of logical rules, with identified reasoning failures ranging from 29\\% to 90\\% across different models. Moreover, we leverage these findings to construct targeted demonstration examples and fine-tune data, notably enhancing logical reasoning in models like GPT-4o by up to 5\\%. To our knowledge, this is the first effort to utilize test case outcomes to effectively refine LLMs' formal reasoning capabilities. We make our code, data, and results publicly available (https://github.com/yxwan123/LogicAsker) to facilitate further research and replication of our findings."
                    },
                    {
                        "title": "New Job, New Gender? Measuring the Social Bias in Image Generation Models",
                        "abstract": "Image generation models can generate or edit images from a given text. Recent advancements in image generation technology, exemplified by DALL-E and Midjourney, have been groundbreaking. These advanced models, despite their impressive capabilities, are often trained on massive Internet datasets, making them susceptible to generating content that perpetuates social stereotypes and biases, which can lead to severe consequences. Prior research on assessing bias within image generation models suffers from several shortcomings, including limited accuracy, reliance on extensive human labor, and lack of comprehensive analysis. In this paper, we propose BiasPainter, a novel evaluation framework that can accurately, automatically and comprehensively trigger social bias in image generation models. BiasPainter uses a diverse range of seed images of individuals and prompts the image generation models to edit these images using gender, race, and age-neutral queries. These queries span 62 professions, 39 activities, 57 types of objects, and 70 personality traits. The framework then compares the edited images to the original seed images, focusing on the significant changes related to gender, race, and age. BiasPainter adopts a key insight that these characteristics should not be modified when subjected to neutral prompts. Built upon this design, BiasPainter can trigger the social bias and evaluate the fairness of image generation models. We use BiasPainter to evaluate six widely-used image generation models, such as stable diffusion and Midjourney. Experimental results show that BiasPainter can successfully trigger social bias in image generation models. According to our human evaluation, BiasPainter can achieve 90.8% accuracy on automatic bias detection, which is significantly higher than the results reported in previous work."
                    },
                    {
                        "title": "Asclepius: A Spectrum Evaluation Benchmark for Medical Multi-Modal Large Language Models",
                        "abstract": "The significant breakthroughs of Medical Multi-Modal Large Language Models (Med-MLLMs) renovate modern healthcare with robust information synthesis and medical decision support. However, these models are often evaluated on benchmarks that are unsuitable for the Med-MLLMs due to the intricate nature of the real-world diagnostic frameworks, which encompass diverse medical specialties and involve complex clinical decisions. Moreover, these benchmarks are susceptible to data leakage, since Med-MLLMs are trained on large assemblies of publicly available data. Thus, an isolated and clinically representative benchmark is highly desirable for credible Med-MLLMs evaluation. To this end, we introduce Asclepius, a novel Med-MLLM benchmark that rigorously and comprehensively assesses model capability in terms of: distinct medical specialties (cardiovascular, gastroenterology, etc.) and different diagnostic capacities (perception, disease analysis, etc.). Grounded in 3 proposed core principles, Asclepius ensures a comprehensive evaluation by encompassing 15 medical specialties, stratifying into 3 main categories and 8 sub-categories of clinical tasks, and exempting from train-validate contamination. We further provide an in-depth analysis of 6 Med-MLLMs and compare them with 5 human specialists, providing insights into their competencies and limitations in various medical contexts. Our work not only advances the understanding of Med-MLLMs' capabilities but also sets a precedent for future evaluations and the safe deployment of these models in clinical environments. We launch and maintain a leaderboard for community assessment of Med-MLLM capabilities (https://asclepius-med.github.io/)."
                    },
                    {
                        "title": "The Earth is Flat? Unveiling Factual Errors in Large Language Models",
                        "abstract": "Large Language Models (LLMs) like ChatGPT are foundational in various applications due to their extensive knowledge from pre-training and fine-tuning. Despite this, they are prone to generating factual and commonsense errors, raising concerns in critical areas like healthcare, journalism, and education to mislead users. Current methods for evaluating LLMs' veracity are limited by test data leakage or the need for extensive human labor, hindering efficient and accurate error detection. To tackle this problem, we introduce a novel, automatic testing framework, FactChecker, aimed at uncovering factual inaccuracies in LLMs. This framework involves three main steps: First, it constructs a factual knowledge graph by retrieving fact triplets from a large-scale knowledge database. Then, leveraging the knowledge graph, FactChecker employs a rule-based approach to generates three types of questions (Yes-No, Multiple-Choice, and WH questions) that involve single-hop and multi-hop relations, along with correct answers. Lastly, it assesses the LLMs' responses for accuracy using tailored matching strategies for each question type. Our extensive tests on six prominent LLMs, including text-davinci-002, text-davinci-003, ChatGPT~(gpt-3.5-turbo, gpt-4), Vicuna, and LLaMA-2, reveal that FactChecker can trigger factual errors in up to 45\\% of questions in these models. Moreover, we demonstrate that FactChecker's test cases can improve LLMs' factual accuracy through in-context learning and fine-tuning (e.g., llama-2-13b-chat's accuracy increase from 35.3\\% to 68.5\\%). We are making all code, data, and results available for future research endeavors."
                    },
                    {
                        "title": "How Well Can LLMs Echo Us? Evaluating AI Chatbots' Role-Play Ability with ECHO",
                        "abstract": "The role-play ability of Large Language Models (LLMs) has emerged as a popular research direction. However, existing studies focus on imitating well-known public figures or fictional characters, overlooking the potential for simulating ordinary individuals. Such an oversight limits the potential for advancements in digital human clones and non-player characters in video games. To bridge this gap, we introduce ECHO, an evaluative framework inspired by the Turing test. This framework engages the acquaintances of the target individuals to distinguish between human and machine-generated responses. Notably, our framework focuses on emulating average individuals rather than historical or fictional figures, presenting a unique advantage to apply the Turing Test. We evaluated three role-playing LLMs using ECHO, with GPT-3.5 and GPT-4 serving as foundational models, alongside the online application GPTs from OpenAI. Our results demonstrate that GPT-4 more effectively deceives human evaluators, and GPTs achieves a leading success rate of 48.3%. Furthermore, we investigated whether LLMs could discern between human-generated and machine-generated texts. While GPT-4 can identify differences, it could not determine which texts were human-produced. Our code and results of reproducing the role-playing LLMs are made publicly available via https://github.com/CUHK-ARISE/ECHO."
                    },
                    {
                        "title": "How Far Are We on the Decision-Making of LLMs? Evaluating LLMs' Gaming Ability in Multi-Agent Environments",
                        "abstract": "Decision-making is a complex process requiring diverse abilities, making it an excellent framework for evaluating Large Language Models (LLMs). Researchers have examined LLMs' decision-making through the lens of Game Theory. However, existing evaluation mainly focus on two-player scenarios where an LLM competes against another. Additionally, previous benchmarks suffer from test set leakage due to their static design. We introduce GAMA($\\gamma$)-Bench, a new framework for evaluating LLMs' Gaming Ability in Multi-Agent environments. It includes eight classical game theory scenarios and a dynamic scoring scheme specially designed to quantitatively assess LLMs' performance. $\\gamma$-Bench allows flexible game settings and adapts the scoring system to different game parameters, enabling comprehensive evaluation of robustness, generalizability, and strategies for improvement. Our results indicate that GPT-3.5 demonstrates strong robustness but limited generalizability, which can be enhanced using methods like Chain-of-Thought. We also evaluate twelve LLMs from six model families, including GPT-3.5, GPT-4, Gemini, LLaMA-3.1, Mixtral, and Qwen-2. Gemini-1.5-Pro outperforms others, scoring of $68.1$ out of $100$, followed by LLaMA-3.1-70B ($64.5$) and Mixtral-8x22B ($61.4$). All code and experimental results are publicly available via https://github.com/CUHK-ARISE/GAMABench."
                    },
                    {
                        "title": "Revisiting the Reliability of Psychological Scales on Large Language Models",
                        "abstract": "Recent research has focused on examining Large Language Models' (LLMs) characteristics from a psychological standpoint, acknowledging the necessity of understanding their behavioral characteristics. The administration of personality tests to LLMs has emerged as a noteworthy area in this context. However, the suitability of employing psychological scales, initially devised for humans, on LLMs is a matter of ongoing debate. Our study aims to determine the reliability of applying personality assessments to LLMs, explicitly investigating whether LLMs demonstrate consistent personality traits. Analysis of 2,500 settings per model, including GPT-3.5, GPT-4, Gemini-Pro, and LLaMA-3.1, reveals that various LLMs show consistency in responses to the Big Five Inventory, indicating a satisfactory level of reliability. Furthermore, our research explores the potential of GPT-3.5 to emulate diverse personalities and represent various groups-a capability increasingly sought after in social sciences for substituting human participants with LLMs to reduce costs. Our findings reveal that LLMs have the potential to represent different personalities with specific prompt instructions."
                    },
                    {
                        "title": "ChatGPT or Grammarly? Evaluating ChatGPT on Grammatical Error Correction Benchmark",
                        "abstract": "ChatGPT is a cutting-edge artificial intelligence language model developed by OpenAI, which has attracted a lot of attention due to its surprisingly strong ability in answering follow-up questions. In this report, we aim to evaluate ChatGPT on the Grammatical Error Correction(GEC) task, and compare it with commercial GEC product (e.g., Grammarly) and state-of-the-art models (e.g., GECToR). By testing on the CoNLL2014 benchmark dataset, we find that ChatGPT performs not as well as those baselines in terms of the automatic evaluation metrics (e.g., $F_{0.5}$ score), particularly on long sentences. We inspect the outputs and find that ChatGPT goes beyond one-by-one corrections. Specifically, it prefers to change the surface expression of certain phrases or sentence structure while maintaining grammatical correctness. Human evaluation quantitatively confirms this and suggests that ChatGPT produces less under-correction or mis-correction issues but more over-corrections. These results demonstrate that ChatGPT is severely under-estimated by the automatic evaluation metrics and could be a promising tool for GEC."
                    },
                    {
                        "title": "Language-Routing Mixture of Experts for Multilingual and Code-Switching Speech Recognition",
                        "abstract": "Multilingual speech recognition for both monolingual and code-switching speech is a challenging task. Recently, based on the Mixture of Experts (MoE), many works have made good progress in multilingual and code-switching ASR, but present huge computational complexity with the increase of supported languages. In this work, we propose a computation-efficient network named Language-Routing Mixture of Experts (LR-MoE) for multilingual and code-switching ASR. LR-MoE extracts language-specific representations through the Mixture of Language Experts (MLE), which is guided to learn by a frame-wise language routing mechanism. The weight-shared frame-level language identification (LID) network is jointly trained as the shared pre-router of each MoE layer. Experiments show that the proposed method significantly improves multilingual and code-switching speech recognition performances over baseline with comparable computational efficiency."
                    },
                    {
                        "title": "An Image is Worth a Thousand Toxic Words: A Metamorphic Testing Framework for Content Moderation Software",
                        "abstract": "The exponential growth of social media platforms has brought about a revolution in communication and content dissemination in human society. Nevertheless, these platforms are being increasingly misused to spread toxic content, including hate speech, malicious advertising, and pornography, leading to severe negative consequences such as harm to teenagers' mental health. Despite tremendous efforts in developing and deploying textual and image content moderation methods, malicious users can evade moderation by embedding texts into images, such as screenshots of the text, usually with some interference. We find that modern content moderation software's performance against such malicious inputs remains underexplored. In this work, we propose OASIS, a metamorphic testing framework for content moderation software. OASIS employs 21 transform rules summarized from our pilot study on 5,000 real-world toxic contents collected from 4 popular social media applications, including Twitter, Instagram, Sina Weibo, and Baidu Tieba. Given toxic textual contents, OASIS can generate image test cases, which preserve the toxicity yet are likely to bypass moderation. In the evaluation, we employ OASIS to test five commercial textual content moderation software from famous companies (i.e., Google Cloud, Microsoft Azure, Baidu Cloud, Alibaba Cloud and Tencent Cloud), as well as a state-of-the-art moderation research model. The results show that OASIS achieves up to 100% error finding rates. Moreover, through retraining the models with the test cases generated by OASIS, the robustness of the moderation model can be improved without performance degradation."
                    },
                    {
                        "title": "Is ChatGPT A Good Translator? A Preliminary Study",
                        "abstract": "This report provides a preliminary evaluation of ChatGPT for machine translation, including translation prompt, multilingual translation, and translation robustness. We adopt the prompts advised by ChatGPT to trigger its translation ability and \ufb01nd that the candidate prompts generally work well and show minor performance differences. By evaluating on a number of benchmark test sets 1 , we \ufb01nd that ChatGPT performs competitively with commercial translation products (e.g., Google Translate) on high-resource European languages but lags behind signi\ufb01cantly on low-resource or distant languages. For distant languages, we explore an interesting strategy named pivot prompting that asks ChatGPT to translate the source sentence into a high-resource pivot language before into the target language, which improves the translation performance signi\ufb01cantly. As for the translation robustness, ChatGPT does not perform as well as the commercial systems on biomedical abstracts or Reddit comments but is potentially a good translator for spoken language."
                    },
                    {
                        "title": "Not All Countries Celebrate Thanksgiving: On the Cultural Dominance in Large Language Models",
                        "abstract": "This paper identifies a cultural dominance issue within large language models (LLMs) due to the predominant use of English data in model training (e.g., ChatGPT). LLMs often provide inappropriate English-culture-related answers that are not relevant to the expected culture when users ask in non-English languages. To systematically evaluate the cultural dominance issue, we build a benchmark of concrete (e.g., holidays and songs) and abstract (e.g., values and opinions) cultural objects. Empirical results show that the representative GPT models suffer from the culture dominance problem, where GPT-4 is the most affected while text-davinci-003 suffers the least from this problem. Our study emphasizes the need to critically examine cultural dominance and ethical consideration in their development and deployment. We show that two straightforward methods in model development (i.e., pretraining on more diverse data) and deployment (e.g., culture-aware prompting) can significantly mitigate the cultural dominance issue in LLMs."
                    },
                    {
                        "title": "Leveraging Word Guessing Games to Assess the Intelligence of Large Language Models",
                        "abstract": "The automatic evaluation of LLM-based agent intelligence is critical in developing advanced LLM-based agents. Although considerable effort has been devoted to developing human-annotated evaluation datasets, such as AlpacaEval, existing techniques are costly, time-consuming, and lack adaptability. In this paper, inspired by the popular language game ``Who is Spy'', we propose to use the word guessing game to assess the intelligence performance of LLMs. Given a word, the LLM is asked to describe the word and determine its identity (spy or not) based on its and other players' descriptions. Ideally, an advanced agent should possess the ability to accurately describe a given word using an aggressive description while concurrently maximizing confusion in the conservative description, enhancing its participation in the game. To this end, we first develop DEEP to evaluate LLMs' expression and disguising abilities. DEEP requires LLM to describe a word in aggressive and conservative modes. We then introduce SpyGame, an interactive multi-agent framework designed to assess LLMs' intelligence through participation in a competitive language-based board game. Incorporating multi-agent interaction, SpyGame requires the target LLM to possess linguistic skills and strategic thinking, providing a more comprehensive evaluation of LLMs' human-like cognitive abilities and adaptability in complex communication situations. The proposed evaluation framework is very easy to implement. We collected words from multiple sources, domains, and languages and used the proposed evaluation framework to conduct experiments. Extensive experiments demonstrate that the proposed DEEP and SpyGame effectively evaluate the capabilities of various LLMs, capturing their ability to adapt to novel situations and engage in strategic communication."
                    },
                    {
                        "title": "All Languages Matter: On the Multilingual Safety of Large Language Models",
                        "abstract": "Safety lies at the core of developing and deploying large language models (LLMs). However, previous safety benchmarks only concern the safety in one language, e.g. the majority language in the pretraining data such as English. In this work, we build the first multilingual safety benchmark for LLMs, XSafety, in response to the global deployment of LLMs in practice. XSafety covers 14 kinds of commonly used safety issues across 10 languages that span several language families. We utilize XSafety to empirically study the multilingual safety for 4 widely-used LLMs, including both close-API and open-source models. Experimental results show that all LLMs produce significantly more unsafe responses for non-English queries than English ones, indicating the necessity of developing safety alignment for non-English languages. In addition, we propose several simple and effective prompting methods to improve the multilingual safety of ChatGPT by evoking safety knowledge and improving cross-lingual generalization of safety alignment. Our prompting method can significantly reduce the ratio of unsafe responses from 19.1% to 9.7% for non-English queries. We release our data at https://github.com/Jarviswang94/Multilingual_safety_benchmark."
                    },
                    {
                        "title": "Improving the Transferability of Adversarial Samples by Path-Augmented Method",
                        "abstract": "Deep neural networks have achieved unprecedented success on diverse vision tasks. However, they are vulnerable to adversarial noise that is imperceptible to humans. This phenomenon negatively affects their deployment in real-world scenarios, especially security-related ones. To evaluate the robustness of a target model in practice, transfer-based attacks craft adversarial samples with a local model and have attracted increasing attention from researchers due to their high efficiency. The state-of-the-art transfer-based attacks are generally based on data augmentation, which typically augments multiple training images from a linear path when learning adversarial samples. However, such methods selected the image augmentation path heuristically and may augment images that are semantics-inconsistent with the target images, which harms the transferability of the generated adversarial samples. To overcome the pitfall, we propose the Path-Augmented Method (PAM). Specifically, PAM first constructs a candidate augmentation path pool. It then settles the employed augmentation paths during adversarial sample generation with greedy search. Furthermore, to avoid augmenting semantics-inconsistent images, we train a Semantics Predictor (SP) to constrain the length of the augmentation path. Extensive experiments confirm that PAM can achieve an improvement of over 4.8% on average compared with the state-of-the-art baselines in terms of the attack success rates."
                    },
                    {
                        "title": "BiasAsker: Measuring the Bias in Conversational AI System",
                        "abstract": "Powered by advanced Artificial Intelligence (AI) techniques, conversational AI systems, such as ChatGPT, and digital assistants like Siri, have been widely deployed in daily life. However, such systems may still produce content containing biases and stereotypes, causing potential social problems. Due to modern AI techniques\u2019 data-driven, black-box nature, comprehensively identifying and measuring biases in conversational systems remains challenging. Particularly, it is hard to generate inputs that can comprehensively trigger potential bias due to the lack of data containing both social groups and biased properties. In addition, modern conversational systems can produce diverse responses (e.g., chatting and explanation), which makes existing bias detection methods based solely on sentiment and toxicity hardly being adopted. In this paper, we propose BiasAsker, an automated framework to identify and measure social bias in conversational AI systems. To obtain social groups and biased properties, we construct a comprehensive social bias dataset containing a total of 841 groups and 5,021 biased properties. Given the dataset, BiasAsker automatically generates questions and adopts a novel method based on existence measurement to identify two types of biases (i.e., absolute bias and related bias) in conversational systems. Extensive experiments on eight commercial systems and two famous research models, such as ChatGPT and GPT-3, show that 32.83% of the questions generated by BiasAsker can trigger biased behaviors in these widely deployed conversational systems. All the code, data, and experimental results have been released to facilitate future research."
                    },
                    {
                        "title": "Is ChatGPT A Good Translator? Yes With GPT-4 As The Engine",
                        "abstract": "This report provides a preliminary evaluation of ChatGPT for machine translation, including translation prompt, multilingual translation, and translation robustness. We adopt the prompts advised by ChatGPT to trigger its translation ability and find that the candidate prompts generally work well with minor performance differences. By evaluating on a number of benchmark test sets, we find that ChatGPT performs competitively with commercial translation products (e.g., Google Translate) on high-resource European languages but lags behind significantly on low-resource or distant languages. As for the translation robustness, ChatGPT does not perform as well as the commercial systems on biomedical abstracts or Reddit comments but exhibits good results on spoken language. Further, we explore an interesting strategy named $\\mathbf{pivot~prompting}$ for distant languages, which asks ChatGPT to translate the source sentence into a high-resource pivot language before into the target language, improving the translation performance noticeably. With the launch of the GPT-4 engine, the translation performance of ChatGPT is significantly boosted, becoming comparable to commercial translation products, even for distant languages. Human analysis on Google Translate and ChatGPT suggests that ChatGPT with GPT-3.5 tends to generate more hallucinations and mis-translation errors while that with GPT-4 makes the least errors. In other words, ChatGPT has already become a good translator. Please refer to our Github project for more details: https://github.com/wxjiao/Is-ChatGPT-A-Good-Translator"
                    },
                    {
                        "title": "MTTM: Metamorphic Testing for Textual Content Moderation Software",
                        "abstract": "The exponential growth of social media platforms such as Twitter and Facebook has revolutionized textual communication and textual content publication in human society. However, they have been increasingly exploited to propagate toxic content, such as hate speech, malicious advertisement, and pornography, which can lead to highly negative impacts (e.g., harmful effects on teen mental health). Researchers and practitioners have been enthusiastically developing and extensively deploying textual content moderation software to address this problem. However, we find that malicious users can evade moderation by changing only a few words in the toxic content. Moreover, modern content moderation software's performance against malicious inputs remains underexplored. To this end, we propose MTTM, a Metamorphic Testing framework for Textual content Moderation software. Specifically, we conduct a pilot study on 2, 000 text messages collected from real users and summarize eleven metamorphic relations across three perturbation levels: character, word, and sentence. MTTM employs these metamorphic relations on toxic textual contents to generate test cases, which are still toxic yet likely to evade moderation. In our evaluation, we employ MTTM to test three commercial textual content moderation software and two state-of-the-art moderation algorithms against three kinds of toxic content. The results show that MTTM achieves up to 83.9%, 51%, and 82.5% error finding rates (EFR) when testing commercial moderation software provided by Google, Baidu, and Huawei, respectively, and it obtains up to 91.2% EFR when testing the state-of-the-art algorithms from the academy. In addition, we leverage the test cases generated by MTTM to retrain the model we explored, which largely improves model robustness 0% ~ 5.9% EFR) while maintaining the accuracy on the original test set. A demo can be found in this link1."
                    }
                ]
            },
            "302d24f5-b011-4450-b1b4-7d1536536dda": {
                "pk": "302d24f5-b011-4450-b1b4-7d1536536dda",
                "name": "Jen-tse Huang",
                "collaborators": [
                    "Wenxuan Wang",
                    "Michael R. Lyu",
                    "Youliang Yuan",
                    "Wenxiang Jiao",
                    "Zhaopeng Tu",
                    "Pinjia He",
                    "Tian Liang",
                    "Cheryl Lee",
                    "E. Li",
                    "Man Ho Lam",
                    "Xing Wang",
                    "Chang Chen",
                    "Yuxuan Wan",
                    "Qiuzhi Liu",
                    "Chun Xia",
                    "Zhouruixing Zhu",
                    "Lingming Zhang",
                    "Juluan Shi",
                    "Jingyuan Huang",
                    "Jiahao Xu",
                    "Yiliu Yang",
                    "Haonan Bai",
                    "Haoyi Qiu",
                    "Nanyun Peng",
                    "Xiaoyuan Liu",
                    "Chaozheng Wang",
                    "Kuiyi Gao",
                    "Zihan Jia",
                    "Shuai Wang",
                    "Man Tik Ng",
                    "Hui Tung Tse",
                    "Jingjing Li",
                    "Jiaxu Zhou",
                    "Tailin Jin",
                    "Xuhui Zhou",
                    "Zixi Chen",
                    "Maarten Sap",
                    "Shujie Ren",
                    "AI Lab",
                    "Jiazhen Gu",
                    "Ruyi Dai",
                    "Zhiwei He",
                    "Rui Wang",
                    "Yujiu Yang",
                    "Shuming Shi"
                ],
                "domain": [
                    "Large Language Models",
                    "Safety",
                    "Bias Evaluation",
                    "Multimodal Systems"
                ],
                "publications": [
                    {
                        "title": "Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training",
                        "abstract": "This study addresses a critical gap in safety tuning practices for Large Language Models (LLMs) by identifying and tackling a refusal position bias within safety tuning data, which compromises the models' ability to appropriately refuse generating unsafe content. We introduce a novel approach, Decoupled Refusal Training (DeRTa), designed to empower LLMs to refuse compliance to harmful prompts at any response position, significantly enhancing their safety capabilities. DeRTa incorporates two novel components: (1) Maximum Likelihood Estimation (MLE) with Harmful Response Prefix, which trains models to recognize and avoid unsafe content by appending a segment of harmful response to the beginning of a safe response, and (2) Reinforced Transition Optimization (RTO), which equips models with the ability to transition from potential harm to safety refusal consistently throughout the harmful response sequence. Our empirical evaluation, conducted using LLaMA3 and Mistral model families across six attack scenarios, demonstrates that our method not only improves model safety without compromising performance but also surpasses well-known models such as GPT-4 in defending against attacks. Importantly, our approach successfully defends recent advanced attack methods (e.g., CodeAttack) that have jailbroken GPT-4 and LLaMA3-70B-Instruct. Our code and data can be found at https://github.com/RobustNLP/DeRTa."
                    },
                    {
                        "title": "LogicAsker: Evaluating and Improving the Logical Reasoning Ability of Large Language Models",
                        "abstract": "We introduce LogicAsker, a novel approach for evaluating and enhancing the logical reasoning capabilities of large language models (LLMs) such as ChatGPT and GPT-4. Despite LLMs' prowess in tasks like writing assistance, code generation, and machine translation, assessing their ability to reason has been challenging. Traditional evaluations often prioritize accuracy on downstream tasks over direct assessments of reasoning processes. LogicAsker addresses this gap by employing a set of atomic reasoning skills grounded in propositional and predicate logic to systematically examine and improve the reasoning prowess of LLMs. Our methodology reveals significant gaps in LLMs' learning of logical rules, with identified reasoning failures ranging from 29\\% to 90\\% across different models. Moreover, we leverage these findings to construct targeted demonstration examples and fine-tune data, notably enhancing logical reasoning in models like GPT-4o by up to 5\\%. To our knowledge, this is the first effort to utilize test case outcomes to effectively refine LLMs' formal reasoning capabilities. We make our code, data, and results publicly available (https://github.com/yxwan123/LogicAsker) to facilitate further research and replication of our findings."
                    },
                    {
                        "title": "New Job, New Gender? Measuring the Social Bias in Image Generation Models",
                        "abstract": "Image generation models can generate or edit images from a given text. Recent advancements in image generation technology, exemplified by DALL-E and Midjourney, have been groundbreaking. These advanced models, despite their impressive capabilities, are often trained on massive Internet datasets, making them susceptible to generating content that perpetuates social stereotypes and biases, which can lead to severe consequences. Prior research on assessing bias within image generation models suffers from several shortcomings, including limited accuracy, reliance on extensive human labor, and lack of comprehensive analysis. In this paper, we propose BiasPainter, a novel evaluation framework that can accurately, automatically and comprehensively trigger social bias in image generation models. BiasPainter uses a diverse range of seed images of individuals and prompts the image generation models to edit these images using gender, race, and age-neutral queries. These queries span 62 professions, 39 activities, 57 types of objects, and 70 personality traits. The framework then compares the edited images to the original seed images, focusing on the significant changes related to gender, race, and age. BiasPainter adopts a key insight that these characteristics should not be modified when subjected to neutral prompts. Built upon this design, BiasPainter can trigger the social bias and evaluate the fairness of image generation models. We use BiasPainter to evaluate six widely-used image generation models, such as stable diffusion and Midjourney. Experimental results show that BiasPainter can successfully trigger social bias in image generation models. According to our human evaluation, BiasPainter can achieve 90.8% accuracy on automatic bias detection, which is significantly higher than the results reported in previous work."
                    },
                    {
                        "title": "Insight Over Sight? Exploring the Vision-Knowledge Conflicts in Multimodal LLMs",
                        "abstract": "This paper explores the problem of commonsense-level vision-knowledge conflict in Multimodal Large Language Models (MLLMs), where visual information contradicts model's internal commonsense knowledge (see Figure 1). To study this issue, we introduce an automated pipeline, augmented with human-in-the-loop quality control, to establish a benchmark aimed at simulating and assessing the conflicts in MLLMs. Utilizing this pipeline, we have crafted a diagnostic benchmark comprising 374 original images and 1,122 high-quality question-answer (QA) pairs. This benchmark covers two types of conflict target and three question difficulty levels, providing a thorough assessment tool. Through this benchmark, we evaluate the conflict-resolution capabilities of nine representative MLLMs across various model families and find a noticeable over-reliance on textual queries. Drawing on these findings, we propose a novel prompting strategy,\"Focus-on-Vision\"(FoV), which markedly enhances MLLMs' ability to favor visual data over conflicting textual knowledge. Our detailed analysis and the newly proposed strategy significantly advance the understanding and mitigating of vision-knowledge conflicts in MLLMs. The data and code are made publicly available."
                    },
                    {
                        "title": "A Unified Debugging Approach via LLM-Based Multi-Agent Synergy",
                        "abstract": "\u2014Tremendous efforts have been devoted to automating software debugging, a time-consuming process involving fault localization and repair generation. Recently, Large Language Models (LLMs) have shown great potential in automated debugging. However, we identified three challenges posed to traditional and LLM-based debugging tools: 1) the upstream imperfection of fault localization affects the downstream repair, 2) the deficiency in handling complex logic errors, and 3) the ignorance of program contexts. In this context, we propose the first automated, unified debugging framework, FixAgent, via LLM agent synergy. FixA-gent can perform end-to-end localization, repair, and analysis of bugs. Our insight is that LLMs can benefit from general software engineering principles recognized by human developers in debugging, such as rubber duck debugging , enabling a better understanding of program functionality and logic bugs. Hence, we create three designs inspired by rubber ducking to address these challenges. They are LLM agent specialization and synergy, key variable tracking, and program context comprehension, which request LLMs to provide explicit explanations and force them to focus on crucial program logic information. Experiments on the widely used dataset QuixBugs show that FixAgent correctly fixes 79 out of 80 bugs, 9 of which have never been fixed. It also plausibly patches 1.9X more defects than the best-performing repair tool on Codeflaws, even with no bug location information and fewer than 0.6% sampling times. On average, FixAgent increases about 20% plausible and correct fixes compared to its base model using different LLMs, showing the effectiveness of our designs. Moreover, the correctness rate of FixAgent reaches remarkably 97.26%, indicating that FixAgent can potentially overcome the overfitting issue of the existing approaches."
                    },
                    {
                        "title": "Learning to Ask: When LLMs Meet Unclear Instruction",
                        "abstract": "Equipped with the capability to call functions, modern large language models (LLMs) can leverage external tools for addressing a range of tasks unattainable through language skills alone. However, the effective execution of these tools relies heavily not just on the advanced capabilities of LLMs but also on precise user instructions, which often cannot be ensured in the real world. To evaluate the performance of LLMs tool-use under imperfect instructions, we meticulously examine the real-world instructions queried from users, analyze the error patterns, and build a challenging tool-use benchmark called Noisy ToolBench (NoisyToolBench). We find that due to the next-token prediction training objective, LLMs tend to arbitrarily generate the missed argument, which may lead to hallucinations and risks. To address this issue, we propose a novel framework, Ask-when-Needed (AwN), which prompts LLMs to ask questions to users whenever they encounter obstacles due to unclear instructions. Moreover, to reduce the manual labor involved in user-LLM interaction and assess LLMs performance in tool utilization from both accuracy and efficiency perspectives, we design an automated evaluation tool named ToolEvaluator. Our experiments demonstrate that the AwN significantly outperforms existing frameworks for tool learning in the NoisyToolBench. We will release all related code and datasets to support future research."
                    },
                    {
                        "title": "Chain-of-Jailbreak Attack for Image Generation Models via Editing Step by Step",
                        "abstract": "Text-based image generation models, such as Stable Diffusion and DALL-E 3, hold significant potential in content creation and publishing workflows, making them the focus in recent years. Despite their remarkable capability to generate diverse and vivid images, considerable efforts are being made to prevent the generation of harmful content, such as abusive, violent, or pornographic material. To assess the safety of existing models, we introduce a novel jailbreaking method called Chain-of-Jailbreak (CoJ) attack, which compromises image generation models through a step-by-step editing process. Specifically, for malicious queries that cannot bypass the safeguards with a single prompt, we intentionally decompose the query into multiple sub-queries. The image generation models are then prompted to generate and iteratively edit images based on these sub-queries. To evaluate the effectiveness of our CoJ attack method, we constructed a comprehensive dataset, CoJ-Bench, encompassing nine safety scenarios, three types of editing operations, and three editing elements. Experiments on four widely-used image generation services provided by GPT-4V, GPT-4o, Gemini 1.5 and Gemini 1.5 Pro, demonstrate that our CoJ attack method can successfully bypass the safeguards of models for over 60% cases, which significantly outperforms other jailbreaking methods (i.e., 14%). Further, to enhance these models' safety against our CoJ attack method, we also propose an effective prompting-based method, Think Twice Prompting, that can successfully defend over 95% of CoJ attack. We release our dataset and code to facilitate the AI safety research."
                    },
                    {
                        "title": "A Unified Debugging Approach via LLM-Based Multi-Agent Synergy",
                        "abstract": "Software debugging is a time-consuming endeavor involving a series of steps, such as fault localization and patch generation, each requiring thorough analysis and a deep understanding of the underlying logic. While large language models (LLMs) demonstrate promising potential in coding tasks, their performance in debugging remains limited. Current LLM-based methods often focus on isolated steps and struggle with complex bugs. In this paper, we propose the first end-to-end framework, FixAgent, for unified debugging through multi-agent synergy. It mimics the entire cognitive processes of developers, with each agent specialized as a particular component of this process rather than mirroring the actions of an independent expert as in previous multi-agent systems. Agents are coordinated through a three-level design, following a cognitive model of debugging, allowing adaptive handling of bugs with varying complexities. Experiments on extensive benchmarks demonstrate that FixAgent significantly outperforms state-of-the-art repair methods, fixing 1.25$\\times$ to 2.56$\\times$ bugs on the repo-level benchmark, Defects4J. This performance is achieved without requiring ground-truth root-cause code statements, unlike the baselines. Our source code is available on https://github.com/AcceptePapier/UniDebugger."
                    },
                    {
                        "title": "The Earth is Flat? Unveiling Factual Errors in Large Language Models",
                        "abstract": "Large Language Models (LLMs) like ChatGPT are foundational in various applications due to their extensive knowledge from pre-training and fine-tuning. Despite this, they are prone to generating factual and commonsense errors, raising concerns in critical areas like healthcare, journalism, and education to mislead users. Current methods for evaluating LLMs' veracity are limited by test data leakage or the need for extensive human labor, hindering efficient and accurate error detection. To tackle this problem, we introduce a novel, automatic testing framework, FactChecker, aimed at uncovering factual inaccuracies in LLMs. This framework involves three main steps: First, it constructs a factual knowledge graph by retrieving fact triplets from a large-scale knowledge database. Then, leveraging the knowledge graph, FactChecker employs a rule-based approach to generates three types of questions (Yes-No, Multiple-Choice, and WH questions) that involve single-hop and multi-hop relations, along with correct answers. Lastly, it assesses the LLMs' responses for accuracy using tailored matching strategies for each question type. Our extensive tests on six prominent LLMs, including text-davinci-002, text-davinci-003, ChatGPT~(gpt-3.5-turbo, gpt-4), Vicuna, and LLaMA-2, reveal that FactChecker can trigger factual errors in up to 45\\% of questions in these models. Moreover, we demonstrate that FactChecker's test cases can improve LLMs' factual accuracy through in-context learning and fine-tuning (e.g., llama-2-13b-chat's accuracy increase from 35.3\\% to 68.5\\%). We are making all code, data, and results available for future research endeavors."
                    },
                    {
                        "title": "How Well Can LLMs Echo Us? Evaluating AI Chatbots' Role-Play Ability with ECHO",
                        "abstract": "The role-play ability of Large Language Models (LLMs) has emerged as a popular research direction. However, existing studies focus on imitating well-known public figures or fictional characters, overlooking the potential for simulating ordinary individuals. Such an oversight limits the potential for advancements in digital human clones and non-player characters in video games. To bridge this gap, we introduce ECHO, an evaluative framework inspired by the Turing test. This framework engages the acquaintances of the target individuals to distinguish between human and machine-generated responses. Notably, our framework focuses on emulating average individuals rather than historical or fictional figures, presenting a unique advantage to apply the Turing Test. We evaluated three role-playing LLMs using ECHO, with GPT-3.5 and GPT-4 serving as foundational models, alongside the online application GPTs from OpenAI. Our results demonstrate that GPT-4 more effectively deceives human evaluators, and GPTs achieves a leading success rate of 48.3%. Furthermore, we investigated whether LLMs could discern between human-generated and machine-generated texts. While GPT-4 can identify differences, it could not determine which texts were human-produced. Our code and results of reproducing the role-playing LLMs are made publicly available via https://github.com/CUHK-ARISE/ECHO."
                    },
                    {
                        "title": "On the Resilience of Multi-Agent Systems with Malicious Agents",
                        "abstract": "Multi-agent systems, powered by large language models, have shown great abilities across various tasks due to the collaboration of expert agents, each focusing on a specific domain. However, when agents are deployed separately, there is a risk that malicious users may introduce malicious agents who generate incorrect or irrelevant results that are too stealthy to be identified by other non-specialized agents. Therefore, this paper investigates two essential questions: (1) What is the resilience of various multi-agent system structures (e.g., A$\\rightarrow$B$\\rightarrow$C, A$\\leftrightarrow$B$\\leftrightarrow$C) under malicious agents, on different downstream tasks? (2) How can we increase system resilience to defend against malicious agents? To simulate malicious agents, we devise two methods, AutoTransform and AutoInject, to transform any agent into a malicious one while preserving its functional integrity. We run comprehensive experiments on four downstream multi-agent systems tasks, namely code generation, math problems, translation, and text evaluation. Results suggest that the\"hierarchical\"multi-agent structure, i.e., A$\\rightarrow$(B$\\leftrightarrow$C), exhibits superior resilience with the lowest performance drop of $23.6\\%$, compared to $46.4\\%$ and $49.8\\%$ of other two structures. Additionally, we show the promise of improving multi-agent system resilience by demonstrating that two defense methods, introducing a mechanism for each agent to challenge others' outputs, or an additional agent to review and correct messages, can enhance system resilience. Our code and data are available at https://github.com/CUHK-ARISE/MAS-Resilience."
                    },
                    {
                        "title": "How Far Are We on the Decision-Making of LLMs? Evaluating LLMs' Gaming Ability in Multi-Agent Environments",
                        "abstract": "Decision-making is a complex process requiring diverse abilities, making it an excellent framework for evaluating Large Language Models (LLMs). Researchers have examined LLMs' decision-making through the lens of Game Theory. However, existing evaluation mainly focus on two-player scenarios where an LLM competes against another. Additionally, previous benchmarks suffer from test set leakage due to their static design. We introduce GAMA($\\gamma$)-Bench, a new framework for evaluating LLMs' Gaming Ability in Multi-Agent environments. It includes eight classical game theory scenarios and a dynamic scoring scheme specially designed to quantitatively assess LLMs' performance. $\\gamma$-Bench allows flexible game settings and adapts the scoring system to different game parameters, enabling comprehensive evaluation of robustness, generalizability, and strategies for improvement. Our results indicate that GPT-3.5 demonstrates strong robustness but limited generalizability, which can be enhanced using methods like Chain-of-Thought. We also evaluate twelve LLMs from six model families, including GPT-3.5, GPT-4, Gemini, LLaMA-3.1, Mixtral, and Qwen-2. Gemini-1.5-Pro outperforms others, scoring of $68.1$ out of $100$, followed by LLaMA-3.1-70B ($64.5$) and Mixtral-8x22B ($61.4$). All code and experimental results are publicly available via https://github.com/CUHK-ARISE/GAMABench."
                    },
                    {
                        "title": "All Languages Matter: On the Multilingual Safety of LLMs",
                        "abstract": "Ensuring safety is fundamental when developing and deploying large language models (LLMs). However, previous safety benchmarks only concern the safety in one language, e.g., the majority language in the pretraining data, such as English. In this work, we build the first multilingual safety benchmark for LLMs, XS AFETY , in response to the global deployment of LLMs in practice. XS AFETY covers 14 commonly used safety issues across ten languages spanning several language families. We utilize XS AFETY to empirically study the multilingual safety for four widely-used LLMs, including closed-source APIs and open-source models. Experimental results show that all LLMs produce significantly more unsafe responses for non-English queries than English ones, indicating the necessity of developing safety alignment for non-English languages. In addition, we propose a simple and effective prompting method to improve ChatGPT\u2019s multilingual safety by enhancing cross-lingual generalization of safety alignment. Our prompting method can significantly reduce the ratio of un-safe responses by 42% for non-English queries. We release the data to facilitate future research on LLM\u2019s safety 1 ."
                    },
                    {
                        "title": "On the Humanity of Conversational AI: Evaluating the Psychological Portrayal of LLMs",
                        "abstract": "capacities"
                    },
                    {
                        "title": "Revisiting the Reliability of Psychological Scales on Large Language Models",
                        "abstract": "Recent research has focused on examining Large Language Models' (LLMs) characteristics from a psychological standpoint, acknowledging the necessity of understanding their behavioral characteristics. The administration of personality tests to LLMs has emerged as a noteworthy area in this context. However, the suitability of employing psychological scales, initially devised for humans, on LLMs is a matter of ongoing debate. Our study aims to determine the reliability of applying personality assessments to LLMs, explicitly investigating whether LLMs demonstrate consistent personality traits. Analysis of 2,500 settings per model, including GPT-3.5, GPT-4, Gemini-Pro, and LLaMA-3.1, reveals that various LLMs show consistency in responses to the Big Five Inventory, indicating a satisfactory level of reliability. Furthermore, our research explores the potential of GPT-3.5 to emulate diverse personalities and represent various groups-a capability increasingly sought after in social sciences for substituting human participants with LLMs to reduce costs. Our findings reveal that LLMs have the potential to represent different personalities with specific prompt instructions."
                    },
                    {
                        "title": "An Image is Worth a Thousand Toxic Words: A Metamorphic Testing Framework for Content Moderation Software",
                        "abstract": "The exponential growth of social media platforms has brought about a revolution in communication and content dissemination in human society. Nevertheless, these platforms are being increasingly misused to spread toxic content, including hate speech, malicious advertising, and pornography, leading to severe negative consequences such as harm to teenagers' mental health. Despite tremendous efforts in developing and deploying textual and image content moderation methods, malicious users can evade moderation by embedding texts into images, such as screenshots of the text, usually with some interference. We find that modern content moderation software's performance against such malicious inputs remains underexplored. In this work, we propose OASIS, a metamorphic testing framework for content moderation software. OASIS employs 21 transform rules summarized from our pilot study on 5,000 real-world toxic contents collected from 4 popular social media applications, including Twitter, Instagram, Sina Weibo, and Baidu Tieba. Given toxic textual contents, OASIS can generate image test cases, which preserve the toxicity yet are likely to bypass moderation. In the evaluation, we employ OASIS to test five commercial textual content moderation software from famous companies (i.e., Google Cloud, Microsoft Azure, Baidu Cloud, Alibaba Cloud and Tencent Cloud), as well as a state-of-the-art moderation research model. The results show that OASIS achieves up to 100% error finding rates. Moreover, through retraining the models with the test cases generated by OASIS, the robustness of the moderation model can be improved without performance degradation."
                    },
                    {
                        "title": "Is ChatGPT A Good Translator? A Preliminary Study",
                        "abstract": "This report provides a preliminary evaluation of ChatGPT for machine translation, including translation prompt, multilingual translation, and translation robustness. We adopt the prompts advised by ChatGPT to trigger its translation ability and \ufb01nd that the candidate prompts generally work well and show minor performance differences. By evaluating on a number of benchmark test sets 1 , we \ufb01nd that ChatGPT performs competitively with commercial translation products (e.g., Google Translate) on high-resource European languages but lags behind signi\ufb01cantly on low-resource or distant languages. For distant languages, we explore an interesting strategy named pivot prompting that asks ChatGPT to translate the source sentence into a high-resource pivot language before into the target language, which improves the translation performance signi\ufb01cantly. As for the translation robustness, ChatGPT does not perform as well as the commercial systems on biomedical abstracts or Reddit comments but is potentially a good translator for spoken language."
                    },
                    {
                        "title": "Not All Countries Celebrate Thanksgiving: On the Cultural Dominance in Large Language Models",
                        "abstract": "This paper identifies a cultural dominance issue within large language models (LLMs) due to the predominant use of English data in model training (e.g., ChatGPT). LLMs often provide inappropriate English-culture-related answers that are not relevant to the expected culture when users ask in non-English languages. To systematically evaluate the cultural dominance issue, we build a benchmark of concrete (e.g., holidays and songs) and abstract (e.g., values and opinions) cultural objects. Empirical results show that the representative GPT models suffer from the culture dominance problem, where GPT-4 is the most affected while text-davinci-003 suffers the least from this problem. Our study emphasizes the need to critically examine cultural dominance and ethical consideration in their development and deployment. We show that two straightforward methods in model development (i.e., pretraining on more diverse data) and deployment (e.g., culture-aware prompting) can significantly mitigate the cultural dominance issue in LLMs."
                    },
                    {
                        "title": "Leveraging Word Guessing Games to Assess the Intelligence of Large Language Models",
                        "abstract": "The automatic evaluation of LLM-based agent intelligence is critical in developing advanced LLM-based agents. Although considerable effort has been devoted to developing human-annotated evaluation datasets, such as AlpacaEval, existing techniques are costly, time-consuming, and lack adaptability. In this paper, inspired by the popular language game ``Who is Spy'', we propose to use the word guessing game to assess the intelligence performance of LLMs. Given a word, the LLM is asked to describe the word and determine its identity (spy or not) based on its and other players' descriptions. Ideally, an advanced agent should possess the ability to accurately describe a given word using an aggressive description while concurrently maximizing confusion in the conservative description, enhancing its participation in the game. To this end, we first develop DEEP to evaluate LLMs' expression and disguising abilities. DEEP requires LLM to describe a word in aggressive and conservative modes. We then introduce SpyGame, an interactive multi-agent framework designed to assess LLMs' intelligence through participation in a competitive language-based board game. Incorporating multi-agent interaction, SpyGame requires the target LLM to possess linguistic skills and strategic thinking, providing a more comprehensive evaluation of LLMs' human-like cognitive abilities and adaptability in complex communication situations. The proposed evaluation framework is very easy to implement. We collected words from multiple sources, domains, and languages and used the proposed evaluation framework to conduct experiments. Extensive experiments demonstrate that the proposed DEEP and SpyGame effectively evaluate the capabilities of various LLMs, capturing their ability to adapt to novel situations and engage in strategic communication."
                    },
                    {
                        "title": "All Languages Matter: On the Multilingual Safety of Large Language Models",
                        "abstract": "Safety lies at the core of developing and deploying large language models (LLMs). However, previous safety benchmarks only concern the safety in one language, e.g. the majority language in the pretraining data such as English. In this work, we build the first multilingual safety benchmark for LLMs, XSafety, in response to the global deployment of LLMs in practice. XSafety covers 14 kinds of commonly used safety issues across 10 languages that span several language families. We utilize XSafety to empirically study the multilingual safety for 4 widely-used LLMs, including both close-API and open-source models. Experimental results show that all LLMs produce significantly more unsafe responses for non-English queries than English ones, indicating the necessity of developing safety alignment for non-English languages. In addition, we propose several simple and effective prompting methods to improve the multilingual safety of ChatGPT by evoking safety knowledge and improving cross-lingual generalization of safety alignment. Our prompting method can significantly reduce the ratio of unsafe responses from 19.1% to 9.7% for non-English queries. We release our data at https://github.com/Jarviswang94/Multilingual_safety_benchmark."
                    }
                ]
            },
            "26e2e7a3-3fef-45a4-83b2-b89311a217f0": {
                "pk": "26e2e7a3-3fef-45a4-83b2-b89311a217f0",
                "name": "Pinjia He",
                "collaborators": [
                    "Michael R. Lyu",
                    "Yintong Huo",
                    "Wenxuan Wang",
                    "Jiazhen Gu",
                    "Junjielong Xu",
                    "Shilin He",
                    "Yuxin Su",
                    "Jen-tse Huang",
                    "Zhihan Jiang",
                    "Jinyang Liu",
                    "Xu Zhang",
                    "Liqun Li",
                    "Yu Kang",
                    "Qingwei Lin",
                    "S. Rajmohan",
                    "Dongmei Zhang",
                    "Zhouruixing Zhu",
                    "Yuxuan Wan",
                    "Yichen Li",
                    "Renyi Zhong",
                    "Junjie Huang",
                    "Jingyuan Huang",
                    "Chang Chen",
                    "Boxi Yu",
                    "Weibin Wu",
                    "Jianping Zhang",
                    "Zhuangbin Chen",
                    "Ziang Cui",
                    "Yuan Zhao",
                    "Yingnong Dang",
                    "Cheryl Lee",
                    "Tianyi Yang",
                    "Xiaoxue Ma",
                    "J. Keung",
                    "Yan Xiao",
                    "Xiao Yu",
                    "Yishu Li",
                    "Yiliu Yang",
                    "Youliang Yuan",
                    "Wenxiang Jiao",
                    "Yi-Nuo Hu",
                    "Qiuyang Mang",
                    "Wen-Ying Hu",
                    "Siyu Yu",
                    "Ningjiang Chen",
                    "Yifan Wu",
                    "Ruichun Yang",
                    "Chengyu Zhang",
                    "Zifan Xie",
                    "Haonan Bai",
                    "Yizhan Huang",
                    "Shuqing Li",
                    "Hongyu Zhang",
                    "Zhangwei Xu",
                    "Qiuai Fu",
                    "Yutong Cheng",
                    "Zhijing Li",
                    "Yuchi Ma",
                    "Zhiqing Zhong",
                    "Jiaqi Li",
                    "Yixing Yang"
                ],
                "domain": [
                    "Automated Logging",
                    "Anomaly Detection",
                    "Content Moderation",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "UniLog: Automatic Logging via LLM and In-Context Learning",
                        "abstract": "Logging, which aims to determine the position of logging statements, the verbosity levels, and the log messages, is a crucial process for software reliability enhancement. In recent years, numerous automatic logging tools have been designed to assist developers in one of the logging tasks (e.g., providing suggestions on whether to log in try-catch blocks). These tools are useful in certain situations yet cannot provide a comprehensive logging solution in general. Moreover, although recent research has started to explore end-to-end logging, it is still largely constrained by the high cost of fine-tuning, hindering its practical usefulness in software development. To address these problems, this paper proposes UniLog, an automatic logging framework based on the in-context learning (ICL) paradigm of large language models (LLMs). Specifically, UniLog can generate an appropriate logging statement with only a prompt containing five demonstration examples without any model tuning. In addition, UniLog can further enhance its logging ability after warmup with only a few hundred random samples. We evaluated UniLog on a large dataset containing 12,012 code snippets extracted from 1,465 GitHub repositories. The results show that UniLog achieved the state-of-the-art performance in automatic logging: (1) 76.9% accuracy in selecting logging positions, (2) 72.3% accuracy in predicting verbosity levels, and (3) 27.1 BLEU-4 score in generating log messages. Meanwhile, UniLog requires less than 4% of the parameter tuning time needed by fine-tuning the same LLM."
                    },
                    {
                        "title": "SPES: Towards Optimizing Performance-Resource Trade-Off for Serverless Functions",
                        "abstract": "As an emerging cloud computing deployment paradigm, serverless computing is gaining traction due to its efficiency and ability to harness on-demand cloud resources. However, a significant hurdle remains in the form of the cold start problem, causing latency when launching new function instances from scratch. Existing solutions tend to use over-simplistic strategies for function pre-loading/unloading without full invocation pattern exploitation, rendering unsatisfactory optimization of the trade-off between cold start latency and resource waste. To bridge this gap, we propose SPES, the first differentiated scheduler for runtime cold start mitigation by optimizing serverless function provision. Our insight is that the common architecture of serverless systems prompts the concentration of certain invocation patterns, leading to predictable invocation behaviors. This allows us to categorize functions and pre-load/unload proper function instances with finer-grained strategies based on accurate invocation prediction. Experiments demonstrate the success of SPES in optimizing serverless function provision on both sides: reducing the 75th-percentile cold start rates by 49.77% and the wasted memory time by 56.43%, compared to the state-of-the-art. By mitigating the cold start issue, SPES is a promising advancement in facilitating cloud services deployed on serverless architectures."
                    },
                    {
                        "title": "A Semisupervised Approach for Industrial Anomaly Detection via Self-Adaptive Clustering",
                        "abstract": "With the rapid development of the Industrial Internet of Things, log-based anomaly detection has become vital for smart industrial construction that has prompted many researchers to contribute. To detect anomalies based on log data, semisupervised approaches stand out from supervised and unsupervised approaches because they only require a portion of labeled data and are relatively stable. However, the state-of-the-art semisupervised approaches still suffer from two main problems: manual parameter setting and unsatisfactory performance with high false positives. We propose AdaLog, an integrated semisupervised approach based on self-adaptive clustering, for industrial anomaly detection. In particular, the clustering step performs automatic label probability estimation by distinguishing 12 situations so that the label probability of each unlabeled data can be carefully calculated, leading to high accuracy. In addition, AdaLog employs a pretrained model to learn contextual information comprehensively and a transformer-based model to detect anomalies efficiently. To alleviate class imbalance, an undersampling method is incorporated. The results on three popular datasets demonstrate that AdaLog significantly outperforms three state-of-the-art semisupervised approaches by 17.8%\u20132489.8% on average in terms of F1-score, and is even superior to two supervised approaches in most cases with average improvements of 10.9%\u201323.8%."
                    },
                    {
                        "title": "LogicAsker: Evaluating and Improving the Logical Reasoning Ability of Large Language Models",
                        "abstract": "We introduce LogicAsker, a novel approach for evaluating and enhancing the logical reasoning capabilities of large language models (LLMs) such as ChatGPT and GPT-4. Despite LLMs' prowess in tasks like writing assistance, code generation, and machine translation, assessing their ability to reason has been challenging. Traditional evaluations often prioritize accuracy on downstream tasks over direct assessments of reasoning processes. LogicAsker addresses this gap by employing a set of atomic reasoning skills grounded in propositional and predicate logic to systematically examine and improve the reasoning prowess of LLMs. Our methodology reveals significant gaps in LLMs' learning of logical rules, with identified reasoning failures ranging from 29\\% to 90\\% across different models. Moreover, we leverage these findings to construct targeted demonstration examples and fine-tune data, notably enhancing logical reasoning in models like GPT-4o by up to 5\\%. To our knowledge, this is the first effort to utilize test case outcomes to effectively refine LLMs' formal reasoning capabilities. We make our code, data, and results publicly available (https://github.com/yxwan123/LogicAsker) to facilitate further research and replication of our findings."
                    },
                    {
                        "title": "Go Static: Contextualized Logging Statement Generation",
                        "abstract": "Logging practices have been extensively investigated to assist developers in writing appropriate logging statements for documenting software behaviors. Although numerous automatic logging approaches have been proposed, their performance remains unsatisfactory due to the constraint of the single-method input, without informative programming context outside the method. Specifically, we identify three inherent limitations with single-method context: limited static scope of logging statements, inconsistent logging styles, and missing type information of logging variables.                                                                To tackle these limitations, we propose SCLogger, the first contextualized logging statement generation approach with inter-method static contexts.First, SCLogger extracts inter-method contexts with static analysis to construct the contextualized prompt for language models to generate a tentative logging statement. The contextualized prompt consists of an extended static scope and sampled similar methods, ordered by the chain-of-thought (COT) strategy. Second, SCLogger refines the access of logging variables by formulating a new refinement prompt for language models, which incorporates detailed type information of variables in the tentative logging statement.                                                                The evaluation results show that SCLogger surpasses the state-of-the-art approach by 8.7% in logging position accuracy, 32.1% in level accuracy, 19.6% in variable precision, and 138.4% in text BLEU-4 score. Furthermore, SCLogger consistently boosts the performance of logging statement generation across a range of large language models, thereby showcasing the generalizability of this approach."
                    },
                    {
                        "title": "An Image is Worth a Thousand Toxic Words: A Metamorphic Testing Framework for Content Moderation Software",
                        "abstract": "The exponential growth of social media platforms has brought about a revolution in communication and content dissemination in human society. Nevertheless, these platforms are being increasingly misused to spread toxic content, including hate speech, malicious advertising, and pornography, leading to severe negative consequences such as harm to teenagers' mental health. Despite tremendous efforts in developing and deploying textual and image content moderation methods, malicious users can evade moderation by embedding texts into images, such as screenshots of the text, usually with some interference. We find that modern content moderation software's performance against such malicious inputs remains underexplored. In this work, we propose OASIS, a metamorphic testing framework for content moderation software. OASIS employs 21 transform rules summarized from our pilot study on 5,000 real-world toxic contents collected from 4 popular social media applications, including Twitter, Instagram, Sina Weibo, and Baidu Tieba. Given toxic textual contents, OASIS can generate image test cases, which preserve the toxicity yet are likely to bypass moderation. In the evaluation, we employ OASIS to test five commercial textual content moderation software from famous companies (i.e., Google Cloud, Microsoft Azure, Baidu Cloud, Alibaba Cloud and Tencent Cloud), as well as a state-of-the-art moderation research model. The results show that OASIS achieves up to 100% error finding rates. Moreover, through retraining the models with the test cases generated by OASIS, the robustness of the moderation model can be improved without performance degradation."
                    },
                    {
                        "title": "Automated Testing and Improvement of Named Entity Recognition Systems",
                        "abstract": "Named entity recognition (NER) systems have seen rapid progress in recent years due to the development of deep neural networks. These systems are widely used in various natural language processing applications, such as information extraction, question answering, and sentiment analysis. However, the complexity and intractability of deep neural networks can make NER systems unreliable in certain circumstances, resulting in incorrect predictions. For example, NER systems may misidentify female names as chemicals or fail to recognize the names of minority groups, leading to user dissatisfaction. To tackle this problem, we introduce TIN, a novel, widely applicable approach for automatically testing and repairing various NER systems. The key idea for automated testing is that the NER predictions of the same named entities under similar contexts should be identical. The core idea for automated repairing is that similar named entities should have the same NER prediction under the same context. We use TIN to test two SOTA NER models and two commercial NER APIs, i.e., Azure NER and AWS NER. We manually verify 784 of the suspicious issues reported by TIN and find that 702 are erroneous issues, leading to high precision (85.0%-93.4%) across four categories of NER errors: omission, over-labeling, incorrect category, and range error. For automated repairing, TIN achieves a high error reduction rate (26.8%-50.6%) over the four systems under test, which successfully repairs 1,056 out of the 1,877 reported NER errors."
                    },
                    {
                        "title": "Brain: Log Parsing With Bidirectional Parallel Tree",
                        "abstract": "Automated log analysis can facilitate failure diagnosis for developers and operators using a large volume of logs. Log parsing is a prerequisite step for automated log analysis, which parses semi-structured logs into structured logs. However, existing parsers are difficult to apply to software-intensive systems, due to their unstable parsing accuracy on various software. Although neural network-based approaches are stable, their inefficiency makes it challenging to keep up with the speed of log production. In this work, we found that the longest common pattern among logs is likely to be part of the log template. Inspired by this key insight, we propose a new stable log parsing approach, called Brain, which creates initial groups according to the longest common pattern. Then a bidirectional tree is used to hierarchically complement the constant words to the longest common pattern to form the complete log template efficiently. Experimental results on 16 benchmark datasets show that our approach outperforms the state-of-the-art parsers on two widely-used parsing accuracy metrics, and it only takes around 46 seconds to process one million lines of logs."
                    },
                    {
                        "title": "Prompting for Automatic Log Template Extraction",
                        "abstract": "\u2014Log parsing, the initial and vital stage in automated log analysis, involves extracting log templates from semi-structured logs to generate structured logs. Nonetheless, current log parsers are limited in effectiveness due to two primary reasons. Firstly, traditional data-driven log parsers heavily rely on heuristics or manually crafted features provided by domain experts, which may not consistently yield optimal performance when applied to diverse log systems. Secondly, existing deep learning-based log parsers necessitate model tuning, which is typically confined to training samples and leads to suboptimal performance across the entire log source. To overcome these limitations, we propose a precise log parsing framework named LogDiv, which leverages the in-context inference capability of large language models. Specifically, LogDiv extracts the hidden semantics from multiple log examples through prompt demonstrations. Without the need for model tuning, LogDiv can directly generate a log template for the target log message by leveraging the semantics provided in the prompt context. Additionally, we introduce a simple yet effective prompt format for extracting the output and enhancing the quality of the generated log templates. To validate the performance of LogDiv, we conducted experiments using 16 widely-used public datasets. The results show that LogDiv achieves state-of-the-art performance with an average parsing accuracy of 97.7%, precision template accuracy of 88.1%, and recall template accuracy of 90.8%."
                    },
                    {
                        "title": "Exploring the Effectiveness of LLMs in Automated Logging Generation: An Empirical Study",
                        "abstract": "Automated logging statement generation supports developers in documenting critical software runtime behavior. Given the great success in natural language generation and programming language comprehension, large language models (LLMs) might help developers generate logging statements, but this has not yet been investigated. To fill the gap, this paper performs the first study on exploring LLMs for logging statement generation.We first build a logging statement generation dataset, LogBench, with two parts: (1) LogBench-O: logging statements collected from GitHub repositories, and (2) LogBench-T: the transformed unseen code from LogBench-O. Then, we leverage LogBench to evaluate the effectiveness and generalization capabilities (using LogBench-T) of eleven top-performing LLMs. In addition, we examine the performance of these LLMs against classical retrieval-based and machine learning-based logging methods from the era preceding LLMs. We further evaluate LLM's logging generalization capabilities using unseen data (LogBench-T) derived from code transformation techniques. While existing LLMs deliver decent predictions on logging levels and logging variables, our study indicates that they only achieve a maximum BLEU score of 0.249, thus calling for improvements. The paper also highlights the importance of prompt constructions and external factors (e.g., programming contexts and code comments) for LLMs' logging performance. Based on these findings, we identify five implications and provide practical advice for future logging research. Our empirical analysis discloses the limitations of current logging approaches while showcasing the potential of LLM-based logging tools, and provides actionable guidance for building more practical models."
                    },
                    {
                        "title": "AutoLog: A Log Sequence Synthesis Framework for Anomaly Detection",
                        "abstract": "The rapid progress of modern computing systems has led to a growing interest in informative run-time logs. Various log-based anomaly detection techniques have been proposed to ensure software reliability. However, their implementation in the industry has been limited due to the lack of high-quality public log resources as training datasets. While some log datasets are available for anomaly detection, they suffer from limitations in (1) comprehensiveness of log events; (2) scalability over diverse systems; and (3) flexibility of log utility. To address these limitations, we propose AUTOLOG, the first automated log generation methodology for anomaly detection. AUTOLOG uses program analysis to generate runtime log sequences without actually running the system. AUTOLOG starts with probing comprehensive logging statements associated with the call graphs of an application. Then, it constructs execution graphs for each method after pruning the call graphs to find log-related execution paths in a scalable manner. Finally, AUTOLOG propagates the anomaly label to each acquired execution path based on human knowledge. It generates flexible log sequences by walking along the log execution paths with controllable parameters. Experiments on 50 popular Java projects show that AUTOLOG acquires significantly more (9x-58x) log events than existing log datasets from the same system, and generates log messages much faster (15x) with a single machine than existing passive data collection approaches. AUTOLOG also provides hyper-parameters to adjust the data size, anomaly rate, and component indicator for simulating different real-world scenarios. We further demonstrate AUTOLOG's practicality by showing that AUTOLOG enables log-based anomaly detectors to achieve better performance (1.93%) compared to existing log datasets. We hope AUTOLOG can facilitate the benchmarking and adoption of automated log analysis techniques."
                    },
                    {
                        "title": "BiasAsker: Measuring the Bias in Conversational AI System",
                        "abstract": "Powered by advanced Artificial Intelligence (AI) techniques, conversational AI systems, such as ChatGPT, and digital assistants like Siri, have been widely deployed in daily life. However, such systems may still produce content containing biases and stereotypes, causing potential social problems. Due to modern AI techniques\u2019 data-driven, black-box nature, comprehensively identifying and measuring biases in conversational systems remains challenging. Particularly, it is hard to generate inputs that can comprehensively trigger potential bias due to the lack of data containing both social groups and biased properties. In addition, modern conversational systems can produce diverse responses (e.g., chatting and explanation), which makes existing bias detection methods based solely on sentiment and toxicity hardly being adopted. In this paper, we propose BiasAsker, an automated framework to identify and measure social bias in conversational AI systems. To obtain social groups and biased properties, we construct a comprehensive social bias dataset containing a total of 841 groups and 5,021 biased properties. Given the dataset, BiasAsker automatically generates questions and adopts a novel method based on existence measurement to identify two types of biases (i.e., absolute bias and related bias) in conversational systems. Extensive experiments on eight commercial systems and two famous research models, such as ChatGPT and GPT-3, show that 32.83% of the questions generated by BiasAsker can trigger biased behaviors in these widely deployed conversational systems. All the code, data, and experimental results have been released to facilitate future research."
                    },
                    {
                        "title": "MTTM: Metamorphic Testing for Textual Content Moderation Software",
                        "abstract": "The exponential growth of social media platforms such as Twitter and Facebook has revolutionized textual communication and textual content publication in human society. However, they have been increasingly exploited to propagate toxic content, such as hate speech, malicious advertisement, and pornography, which can lead to highly negative impacts (e.g., harmful effects on teen mental health). Researchers and practitioners have been enthusiastically developing and extensively deploying textual content moderation software to address this problem. However, we find that malicious users can evade moderation by changing only a few words in the toxic content. Moreover, modern content moderation software's performance against malicious inputs remains underexplored. To this end, we propose MTTM, a Metamorphic Testing framework for Textual content Moderation software. Specifically, we conduct a pilot study on 2, 000 text messages collected from real users and summarize eleven metamorphic relations across three perturbation levels: character, word, and sentence. MTTM employs these metamorphic relations on toxic textual contents to generate test cases, which are still toxic yet likely to evade moderation. In our evaluation, we employ MTTM to test three commercial textual content moderation software and two state-of-the-art moderation algorithms against three kinds of toxic content. The results show that MTTM achieves up to 83.9%, 51%, and 82.5% error finding rates (EFR) when testing commercial moderation software provided by Google, Baidu, and Huawei, respectively, and it obtains up to 91.2% EFR when testing the state-of-the-art algorithms from the academy. In addition, we leverage the test cases generated by MTTM to retrain the model we explored, which largely improves model robustness 0% ~ 5.9% EFR) while maintaining the accuracy on the original test set. A demo can be found in this link1."
                    },
                    {
                        "title": "Incident-aware Duplicate Ticket Aggregation for Cloud Systems",
                        "abstract": "In cloud systems, incidents are potential threats to customer satisfaction and business revenue. When customers are affected by incidents, they often request customer support service (CSS) from the cloud provider by submitting a support ticket. Many tickets could be duplicate as they are reported in a distributed and uncoordinated manner. Thus, aggregating such duplicate tickets is essential for efficient ticket management. Previous studies mainly rely on tickets' textual similarity to detect duplication; however, duplicate tickets in a cloud system could carry semantically different descriptions due to the complex service dependency of the cloud system. To tackle this problem, we propose iPACK, an incident-aware method for aggregating duplicate tickets by fusing the failure information between the customer side (i.e., tickets) and the cloud side (i.e., incidents). We extensively evaluate iPACK on three datasets collected from the production environment of a large-scale cloud platform, Azure. The experimental results show that iPACK can precisely and comprehensively aggregate duplicate tickets, achieving an F1 score of 0.871~0.935 and outperforming state-of-the-art methods by 12.4%~31.2%."
                    },
                    {
                        "title": "Validating Multimedia Content Moderation Software via Semantic Fusion",
                        "abstract": "The exponential growth of social media platforms, such as Facebook, Instagram, Youtube, and TikTok, has revolutionized communication and content publication in human society. Users on these platforms can publish multimedia content that delivers information via the combination of text, audio, images, and video. Meanwhile, the multimedia content release facility has been increasingly exploited to propagate toxic content, such as hate speech, malicious advertisement, and pornography. To this end, content moderation software has been widely deployed on these platforms to detect and blocks toxic content. However, due to the complexity of content moderation models and the difficulty of understanding information across multiple modalities, existing content moderation software can fail to detect toxic content, which often leads to extremely negative impacts (e.g., harmful effects on teen mental health). We introduce Semantic Fusion, a general, effective methodology for validating multimedia content moderation software. Our key idea is to fuse two or more existing single-modal inputs (e.g., a textual sentence and an image) into a new input that combines the semantics of its ancestors in a novel manner and has toxic nature by construction. This fused input is then used for validating multimedia content moderation software. We realized Semantic Fusion as DUO, a practical content moderation software testing tool. In our evaluation, we employ DUO to test five commercial content moderation software and two state-of-the-art models against three kinds of toxic contents. The results show that DUO achieves up to 100% error finding rate (EFR) when testing moderation software and it obtains up to 94.1% EFR when testing the state-of-the-art models. In addition, we leverage the test cases generated by DUO to retrain the two models we explored, which largely improves model robustness (2.5%\u223c5.7% EFR) while maintaining the accuracy on the original test set."
                    },
                    {
                        "title": "LILAC: Log Parsing using LLMs with Adaptive Parsing Cache",
                        "abstract": "Log parsing transforms log messages into structured formats, serving as the prerequisite step for various log analysis tasks. Although a variety of log parsing approaches have been proposed, their performance on complicated log data remains compromised due to the use of human-crafted rules or learning-based models with limited training data. The recent emergence of powerful large language models (LLMs) demonstrates their vast pre-trained knowledge related to code and logging, making it promising to apply LLMs for log parsing. However, their lack of specialized log parsing capabilities currently hinders their parsing accuracy. Moreover, the inherent inconsistent answers, as well as the substantial overhead, prevent the practical adoption of LLM-based log parsing.       To address these challenges, we propose LILAC, the first practical Log parsIng framework using LLMs with Adaptive parsing Cache. To facilitate accurate and robust log parsing, LILAC leverages the in-context learning (ICL) capability of the LLM by performing a hierarchical candidate sampling algorithm and selecting high-quality demonstrations. Furthermore, LILAC incorporates a novel component, an adaptive parsing cache, to store and refine the templates generated by the LLM. It helps mitigate LLM's inefficiency issue by enabling rapid retrieval of previously processed log templates. In this process, LILAC adaptively updates the templates within the parsing cache to ensure the consistency of parsed results. The extensive evaluation on public large-scale datasets shows that LILAC outperforms state-of-the-art methods by 69.5% in terms of the average F1 score of template accuracy. In addition, LILAC reduces the query times to LLMs by several orders of magnitude, achieving a comparable efficiency to the fastest baseline."
                    },
                    {
                        "title": "Hue: A User-Adaptive Parser for Hybrid Logs",
                        "abstract": "Log parsing, which extracts log templates from semi-structured logs and produces structured logs, is the first and the most critical step in automated log analysis. While existing log parsers have achieved decent results, they suffer from two major limitations by design. First, they do not natively support hybrid logs that consist of both single-line logs and multi-line logs (Java Exception and Hadoop Counters). Second, they fall short in integrating domain knowledge in parsing, making it hard to identify ambiguous tokens in logs. This paper defines a new research problem, hybrid log parsing, as a superset of traditional log parsing tasks, and proposes Hue, the first attempt for hybrid log parsing via a user-adaptive manner. Specifically, Hue converts each log message to a sequence of special wildcards using a key casting table and determines the log types via line aggregating and pattern extracting. In addition, Hue can effectively utilize user feedback via a novel merge-reject strategy, making it possible to quickly adapt to complex and changing log templates. We evaluated Hue on three hybrid log datasets and sixteen widely-used single-line log datasets (Loghub). The results show that Hue achieves an average grouping accuracy of 0.845 on hybrid logs, which largely outperforms the best results (0.563 on average) obtained by existing parsers. Hue also exhibits SOTA performance on single-line log datasets."
                    },
                    {
                        "title": "ROME: Testing Image Captioning Systems via Recursive Object Melting",
                        "abstract": "Image captioning (IC) systems aim to generate a text description of the salient objects in an image. In recent years, IC systems have been increasingly integrated into our daily lives, such as assistance for visually-impaired people and description generation in Microsoft Powerpoint. However, even the cutting-edge IC systems (e.g., Microsoft Azure Cognitive Services) and algorithms (e.g., OFA) could produce erroneous captions, leading to incorrect captioning of important objects, misunderstanding, and threats to personal safety. The existing testing approaches either fail to handle the complex form of IC system output (i.e., sentences in natural language) or generate unnatural images as test cases. To address these problems, we introduce Recursive Object MElting (ROME), a novel metamorphic testing approach for validating IC systems. Different from existing approaches that generate test cases by inserting objects, which easily make the generated images unnatural, ROME melts (i.e., remove and inpaint) objects. ROME assumes that the object set in the caption of an image includes the object set in the caption of a generated image after object melting. Given an image, ROME can recursively remove its objects to generate different pairs of images. We use ROME to test one widely-adopted image captioning API and four state-of-the-art (SOTA) algorithms. The results show that the test cases generated by ROME look much more natural than the SOTA IC testing approach and they achieve comparable naturalness to the original images. Meanwhile, by generating test pairs using 226 seed images, ROME reports a total of 9,121 erroneous issues with high precision (86.47%-92.17%). In addition, we further utilize the test cases generated by ROME to retrain the Oscar, which improves its performance across multiple evaluation metrics."
                    }
                ]
            },
            "d1102925-f64a-4dd6-bbe3-5e17e87648db": {
                "pk": "d1102925-f64a-4dd6-bbe3-5e17e87648db",
                "name": "Shuming Shi",
                "collaborators": [
                    "Lemao Liu",
                    "Longyue Wang",
                    "Zhaopeng Tu",
                    "Zhirui Zhang",
                    "Rui Wang",
                    "Haiyun Jiang",
                    "Chenyang Lyu",
                    "Leyang Cui",
                    "Bingshuai Liu",
                    "Yafu Li",
                    "Wei Bi",
                    "Guoping Huang",
                    "Zhangyin Feng",
                    "Xiaocheng Feng",
                    "Duyu Tang",
                    "Bing Qin",
                    "Minghao Wu",
                    "Xinting Huang",
                    "Zefeng Du",
                    "Qintong Li",
                    "Linyi Yang",
                    "Yue Zhang",
                    "Xu Huang",
                    "Ruize Gao",
                    "Yichao Du",
                    "Gouping Huang",
                    "Jiajun Chen",
                    "Shujian Huang",
                    "Yuejiao Fei",
                    "Sen Yang",
                    "Wai Lam",
                    "Zhenzhong Lan",
                    "Wenye Lin",
                    "Yangming Li",
                    "Haitao Zheng",
                    "Xingyu Chen",
                    "Mingming Yang",
                    "Hong-ping Hao",
                    "Tian Liang",
                    "Zhiwei He",
                    "Wenxiang Jiao",
                    "Xing Wang",
                    "Yan Wang",
                    "Yujiu Yang",
                    "Yuchen Ren",
                    "Xinmiao Yu",
                    "Tianbo Ji",
                    "Dian Yu",
                    "Yong Zhang",
                    "Jinsong Su",
                    "Yong Dai",
                    "Fan Zhang",
                    "Shuangzhi Wu",
                    "Yunbo Cao",
                    "Lingfeng Shen",
                    "Yi Chen",
                    "Rui-Lan Xu",
                    "M. Song",
                    "L. Jing",
                    "Siyou Liu",
                    "Mingzhou Xu",
                    "Linfeng Song",
                    "Jianhao Yan",
                    "Yongjng Yin"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Multi-modal Learning",
                    "Machine Translation",
                    "Text Generation"
                ],
                "publications": [
                    {
                        "title": "Macaw-LLM: Multi-Modal Language Modeling with Image, Audio, Video, and Text Integration",
                        "abstract": "Although instruction-tuned large language models (LLMs) have exhibited remarkable capabilities across various NLP tasks, their effectiveness on other data modalities beyond text has not been fully studied. In this work, we propose Macaw-LLM, a novel multi-modal LLM that seamlessly integrates visual, audio, and textual information. Macaw-LLM consists of three main components: a modality module for encoding multi-modal data, a cognitive module for harnessing pretrained LLMs, and an alignment module for harmonizing diverse representations. Our novel alignment module seamlessly bridges multi-modal features to textual features, simplifying the adaptation process from the modality modules to the cognitive module. In addition, we construct a large-scale multi-modal instruction dataset in terms of multi-turn dialogue, including 69K image instances and 50K video instances. We have made our data, code and model publicly available, which we hope can pave the way for future research in multi-modal LLMs and expand the capabilities of LLMs to handle diverse data modalities and address complex real-world scenarios."
                    },
                    {
                        "title": "Deepfake Text Detection in the Wild",
                        "abstract": "Recent advances in large language models have enabled them to reach a level of text generation comparable to that of humans. These models show powerful capabilities across a wide range of content, including news article writing, story generation, and scienti\ufb01c writing. Such capability further narrows the gap between human-authored and machine-generated texts, highlighting the importance of deepfake text detection to avoid potential risks such as fake news propagation and plagia-rism. However, previous work has been limited in that they testify methods on testbeds of speci\ufb01c domains or certain language models. In practical scenarios, the detector faces texts from various domains or LLMs without knowing their sources. To this end, we build a wild testbed by gathering texts from various human writings and deepfake texts generated by different LLMs. Human annotators are only slightly better than random guessing at identifying machine-generated texts. Empirical results on automatic detection meth-ods further showcase the challenges of deep-fake text detection in a wild testbed. In addition, out-of-distribution poses a greater challenge for a detector to be employed in realistic application scenarios. We release our resources at https://github.com/yafuly/ DeepfakeTextDetect ."
                    },
                    {
                        "title": "IMTLab: An Open-Source Platform for Building, Evaluating, and Diagnosing Interactive Machine Translation Systems",
                        "abstract": "We present IMTLab, an open-source end-to-end interactive machine translation (IMT) system platform that enables researchers to quickly build IMT systems with state-of-the-art models, perform an end-to-end evaluation, and diagnose the weakness of systems. IMTLab treats the whole interactive translation process as a task-oriented dialogue with a human-in-the-loop setting, in which human interventions can be explicitly incorporated to produce high-quality, error-free translations. To this end, a general communication interface is designed to support the flexible IMT architectures and user policies. Based on the proposed design, we construct a simulated and real interactive environment to achieve end-to-end evaluation and leverage the framework to systematically evaluate previous IMT systems. Our simulated and manual experiments show that the prefix-constrained decoding approach still gains the lowest editing cost in the end-to-end evaluation, while BiTIIMT achieves comparable editing cost with a better interactive experience."
                    },
                    {
                        "title": "Enhancing Grammatical Error Correction Systems with Explanations",
                        "abstract": "Grammatical error correction systems improve written communication by detecting and correcting language mistakes. To help language learners better understand why the GEC system makes a certain correction, the causes of errors (evidence words) and the corresponding error types are two key factors. To enhance GEC systems with explanations, we introduce EXPECT, a large dataset annotated with evidence words and grammatical error types. We propose several baselines and anlysis to understand this task. Furthermore, human evaluation verifies our explainable GEC system\u2019s explanations can assist second-language learners in determining whether to accept a correction suggestion and in understanding the associated grammar rule."
                    },
                    {
                        "title": "A Simple Yet Effective Approach to Structured Knowledge Distillation",
                        "abstract": "Structured prediction models aim at solving tasks where the output is a complex structure, rather than a single variable. Performing knowledge distillation for such problems is non- trivial due to their exponentially large output space. Previous works address this problem by developing particular distillation strategies (e.g., dynamic programming) that are both complicated and of low run-time efficiency. In this work, we propose an approach that is much simpler in its formulation, far more efficient for training than existing methods, and even performs better than our baselines. Specifically, we transfer the knowledge from a teacher model to its student by locally matching their computations on all internal structures rather than the final outputs. In this manner, we avoid time-consuming techniques like Monte Carlo Sampling for decoding output structures, permitting parallel computation and efficient training. Besides, we show that it encourages the student model to better mimic the internal behavior of the teacher model. Experiments on two structured prediction tasks demonstrate that our approach not only halves the time cost, but also outperforms previous methods on two widely adopted benchmark datasets.1 2"
                    },
                    {
                        "title": "Rethinking Word-Level Auto-Completion in Computer-Aided Translation",
                        "abstract": "Word-Level Auto-Completion (WLAC) plays a crucial role in Computer-Assisted Translation. It aims at providing word-level auto-completion suggestions for human translators. While previous studies have primarily focused on designing complex model architectures, this paper takes a different perspective by rethinking the fundamental question: what kind of words are good auto-completions? We introduce a measurable criterion to answer this question and discover that existing WLAC models often fail to meet this criterion. Building upon this observation, we propose an effective approach to enhance WLAC performance by promoting adherence to the criterion. Notably, the proposed approach is general and can be applied to various encoder-based architectures. Through extensive experiments, we demonstrate that our approach outperforms the top-performing system submitted to the WLAC shared tasks in WMT2022, while utilizing significantly smaller model sizes."
                    },
                    {
                        "title": "Sen2Pro: A Probabilistic Perspective to Sentence Embedding from Pre-trained Language Model",
                        "abstract": "Sentence embedding is one of the most fundamental tasks in Natural Language Processing and plays an important role in various tasks. The recent breakthrough in sentence embedding is achieved by pre-trained language models (PLMs). Despite its success, an embedded vector (Sen2Vec) representing a point estimate does not naturally express uncertainty in a taskagnostic way. This paper thereby proposes an efficient framework on probabilistic sentence embedding (Sen2Pro) from PLMs, and it represents a sentence as a probability density distribution in an embedding space to reflect both model uncertainty and data uncertainty (i.e., many-to-one nature) in the sentence representation. The proposed framework performs in a plug-and-play way without retraining PLMs anymore, and it is easy to implement and generally applied on top of any PLM. The superiority of Sen2Pro over Sen2Vec has been theoretically verified and practically illustrated on different NLP tasks."
                    },
                    {
                        "title": "Rethinking Translation Memory Augmented Neural Machine Translation",
                        "abstract": "This paper rethinks translation memory augmented neural machine translation (TM-augmented NMT) from two perspectives, i.e., a probabilistic view of retrieval and the variance-bias decomposition principle. The finding demonstrates that TM-augmented NMT is good at the ability of fitting data (i.e., lower bias) but is more sensitive to the fluctuations in the training data (i.e., higher variance), which provides an explanation to a recently reported contradictory phenomenon on the same translation task: TM-augmented NMT substantially advances vanilla NMT under the high-resource scenario whereas it fails under the low-resource scenario. Then we propose a simple yet effective TM-augmented NMT model to promote the variance and address the contradictory phenomenon. Extensive experiments show that the proposed TM-augmented NMT achieves consistent gains over both conventional NMT and existing TM-augmented NMT under two variance-preferable (low-resource and plug-and-play) scenarios as well as the high-resource scenario."
                    },
                    {
                        "title": "Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate",
                        "abstract": "Modern large language models (LLMs) like ChatGPT have shown remarkable performance on general language tasks but still struggle on complex reasoning tasks, which drives the research on cognitive behaviors of LLMs to explore human-like problem-solving strategies. Along this direction, one representative strategy is self-reflection, which asks an LLM to refine the solution with the feedback generated by itself iteratively. However, our study shows that such reflection-style methods suffer from the Degeneration-of-Thought (DoT) problem: once the LLM has established confidence in its solutions, it is unable to generate novel thoughts later through reflection even if its initial stance is incorrect. To address the DoT problem, we propose a Multi-Agent Debate (MAD) framework, in which multiple agents express their arguments in the state of\"tit for tat\"and a judge manages the debate process to obtain a final solution. Clearly, our MAD framework encourages divergent thinking in LLMs which would be helpful for tasks that require deep levels of contemplation. Experiment results on two challenging datasets, commonsense machine translation and counter-intuitive arithmetic reasoning, demonstrate the effectiveness of our MAD framework. Extensive analyses suggest that the adaptive break of debate and the modest level of\"tit for tat\"state are required for MAD to obtain good performance. Moreover, we find that LLMs might not be a fair judge if different LLMs are used for agents. Code is available at https://github.com/Skytliang/Multi-Agents-Debate."
                    },
                    {
                        "title": "Improved Visual Story Generation with Adaptive Context Modeling",
                        "abstract": "Diffusion models developed on top of powerful text-to-image generation models like Stable Diffusion achieve remarkable success in visual story generation. However, the best-performing approach considers historically generated results as flattened memory cells, ignoring the fact that not all preceding images contribute equally to the generation of the characters and scenes at the current stage. To address this, we present a simple method that improves the leading system with adaptive context modeling, which is not only incorporated in the encoder but also adopted as additional guidance in the sampling stage to boost the global consistency of the generated story. We evaluate our model on PororoSV and FlintstonesSV datasets and show that our approach achieves state-of-the-art FID scores on both story visualization and continuation scenarios. We conduct detailed model analysis and show that our model excels at generating semantically consistent images for stories."
                    },
                    {
                        "title": "Document-Level Machine Translation with Large Language Models",
                        "abstract": "Large language models (LLMs) such as ChatGPT can produce coherent, cohesive, relevant, and fluent answers for various natural language processing (NLP) tasks. Taking document-level machine translation (MT) as a testbed, this paper provides an in-depth evaluation of LLMs' ability on discourse modeling. The study focuses on three aspects: 1) Effects of Context-Aware Prompts, where we investigate the impact of different prompts on document-level translation quality and discourse phenomena; 2) Comparison of Translation Models, where we compare the translation performance of ChatGPT with commercial MT systems and advanced document-level MT methods; 3) Analysis of Discourse Modelling Abilities, where we further probe discourse knowledge encoded in LLMs and shed light on impacts of training techniques on discourse modeling. By evaluating on a number of benchmarks, we surprisingly find that LLMs have demonstrated superior performance and show potential to become a new paradigm for document-level translation: 1) leveraging their powerful long-text modeling capabilities, GPT-3.5 and GPT-4 outperform commercial MT systems in terms of human evaluation; 2) GPT-4 demonstrates a stronger ability for probing linguistic knowledge than GPT-3.5. This work highlights the challenges and opportunities of LLMs for MT, which we hope can inspire the future design and evaluation of LLMs.We release our data and annotations at https://github.com/longyuewangdcu/Document-MT-LLM."
                    },
                    {
                        "title": "On the Cultural Gap in Text-to-Image Generation",
                        "abstract": "One challenge in text-to-image (T2I) generation is the inadvertent reflection of culture gaps present in the training data, which signifies the disparity in generated image quality when the cultural elements of the input text are rarely collected in the training set. Although various T2I models have shown impressive but arbitrary examples, there is no benchmark to systematically evaluate a T2I model's ability to generate cross-cultural images. To bridge the gap, we propose a Challenging Cross-Cultural (C3) benchmark with comprehensive evaluation criteria, which can assess how well-suited a model is to a target culture. By analyzing the flawed images generated by the Stable Diffusion model on the C3 benchmark, we find that the model often fails to generate certain cultural objects. Accordingly, we propose a novel multi-modal metric that considers object-text alignment to filter the fine-tuning data in the target culture, which is used to fine-tune a T2I model to improve cross-cultural generation. Experimental results show that our multi-modal metric provides stronger data selection performance on the C3 benchmark than existing metrics, in which the object-text alignment is crucial. We release the benchmark, data, code, and generated images to facilitate future research on culturally diverse T2I generation (https://github.com/longyuewangdcu/C3-Bench)."
                    },
                    {
                        "title": "SkillNet-X: A Multilingual Multitask Model with Sparsely Activated Skills",
                        "abstract": "Traditional multitask learning methods typically can only leverage shared knowledge within specific tasks or languages, resulting in a loss of either cross-language or cross-task knowledge. This paper proposes a general multilingual multitask model, named SkillNet-X, which enables a single model to tackle many different tasks from different languages. To this end, we define several language-specific skills and task-specific skills, each of which corresponds to a skill module. SkillNet-X sparsely activates parts of the skill modules which are relevant to eitherthe target task or the target language. Acting as knowledge transit hubs, skill modules are capable of absorbing task-related knowledge and language-related knowledge consecutively. We evaluate SkillNet-X on eleven natural language understanding datasets in four languages. Results show that SkillNet-X performs better than task-specific and two multitask learning baselines.To investigate the generalization of our model, we conduct experiments on two new tasks and find that SkillNet-X significantly outperforms baselines."
                    },
                    {
                        "title": "A Simple and Plug-and-play Method for Unsupervised Sentence Representation Enhancement",
                        "abstract": "Generating proper embedding of sentences through an unsupervised way is beneficial to semantic matching and retrieval problems in real-world scenarios. This paper presents Representation ALchemy (RepAL), an extremely simple post-processing method that enhances sentence representations. The basic idea in RepAL is to de-emphasize redundant information of sentence embedding generated by pre-trained models. Through comprehensive experiments, we show that RepAL is free of training and is a plug-and-play method that can be combined with most existing unsupervised sentence learning models. We also conducted in-depth analysis to understand RepAL."
                    },
                    {
                        "title": "Exploring the Use of Large Language Models for Reference-Free Text Quality Evaluation: A Preliminary Empirical Study",
                        "abstract": "Evaluating the quality of generated text is a challenging task in natural language processing. This dif\ufb01culty arises from the inherent complexity and diversity of text. Recently, OpenAI\u2019s ChatGPT, a powerful large language model (LLM), has garnered signi\ufb01cant attention due to its impressive performance in various tasks. Therefore, we present this report to investigate the effectiveness of LLMs, especially ChatGPT, and explore ways to optimize their use in assessing text quality. We compared three kinds of reference-free evaluation methods based on ChatGPT or similar LLMs. The experimental results prove that ChatGPT is capable to evaluate text quality effectively from various perspectives without reference and demonstrates superior performance than most existing automatic metrics. In particular, the Explicit Score, which utilizes ChatGPT to generate a numeric score measuring text quality, is the most effective and reliable method among the three exploited approaches. However, directly comparing the quality of two texts using ChatGPT may lead to suboptimal results. We hope this report will provide valuable insights into selecting appropriate meth-ods for evaluating text quality with LLMs such as ChatGPT. We have released the used data 1 ."
                    },
                    {
                        "title": "Unsupervised Keyphrase Extraction by Learning Neural Keyphrase Set Function",
                        "abstract": ","
                    },
                    {
                        "title": "A Survey on Zero Pronoun Translation",
                        "abstract": "Zero pronouns (ZPs) are frequently omitted in pro-drop languages (e.g. Chinese, Hungarian, and Hindi), but should be recalled in non-pro-drop languages (e.g. English). This phenomenon has been studied extensively in machine translation (MT), as it poses a significant challenge for MT systems due to the difficulty in determining the correct antecedent for the pronoun. This survey paper highlights the major works that have been undertaken in zero pronoun translation (ZPT) after the neural revolution so that researchers can recognize the current state and future directions of this field. We provide an organization of the literature based on evolution, dataset, method, and evaluation. In addition, we compare and analyze competing models and evaluation metrics on different benchmarks. We uncover a number of insightful findings such as: 1) ZPT is in line with the development trend of large language model; 2) data limitation causes learning bias in languages and domains; 3) performance improvements are often reported on single benchmarks, but advanced methods are still far from real-world use; 4) general-purpose metrics are not reliable on nuances and complexities of ZPT, emphasizing the necessity of targeted metrics; 5) apart from commonly-cited errors, ZPs will cause risks of gender bias."
                    },
                    {
                        "title": "Explicit Syntactic Guidance for Neural Text Generation",
                        "abstract": "Most existing text generation models follow the sequence-to-sequence paradigm. Generative Grammar suggests that humans generate natural language texts by learning language grammar. We propose a syntax-guided generation schema, which generates the sequence guided by a constituency parse tree in a top-down direction. The decoding process can be decomposed into two parts: (1) predicting the infilling texts for each constituent in the lexicalized syntax context given the source sentence; (2) mapping and expanding each constituent to construct the next-level syntax context. Accordingly, we propose a structural beam search method to find possible syntax structures hierarchically. Experiments on paraphrase generation and machine translation show that the proposed method outperforms autoregressive baselines, while also demonstrating effectiveness in terms of interpretability, controllability, and diversity."
                    }
                ]
            },
            "a591ebb1-f4ad-4a2c-9d07-ee001644dd19": {
                "pk": "a591ebb1-f4ad-4a2c-9d07-ee001644dd19",
                "name": "Zhaopeng Tu",
                "collaborators": [
                    "Shuming Shi",
                    "Xing Wang",
                    "Longyue Wang",
                    "Wenxiang Jiao",
                    "Jie Hu",
                    "Hanting Chen",
                    "Chenyang Lyu",
                    "Wenxuan Wang",
                    "Jen-tse Huang",
                    "Zhiwei He",
                    "Yunhe Wang",
                    "Wei Li",
                    "Bingshuai Liu",
                    "Zhihao Wang",
                    "Tian Liang",
                    "Rui Wang",
                    "Yujiu Yang",
                    "Zhirui Zhang",
                    "Jinsong Su",
                    "Minghao Wu",
                    "Xinting Huang",
                    "Zefeng Du",
                    "Kangjie Zheng",
                    "Binqi Chen",
                    "Ming Zhang",
                    "Yan Wang",
                    "Yichao Du",
                    "Guo Zhengsheng",
                    "Jinchuan Tian",
                    "Jianwei Yu",
                    "Tong Xu",
                    "Enhong Chen",
                    "Yun-feng Zhang",
                    "Simiao Li",
                    "Hai-lin Wang",
                    "Wenjun Wang",
                    "Bingyi Jing",
                    "Yun Wang",
                    "Tianbo Ji",
                    "Dian Yu",
                    "Yong Zhang",
                    "Mingjian Zhu",
                    "Qiang Yan",
                    "Xu Huang",
                    "Guanyu Lin",
                    "Hailin Hu",
                    "Siyou Liu",
                    "Mingzhou Xu",
                    "Linfeng Song",
                    "Zhuosheng Zhang",
                    "Jinhui Ye",
                    "Hui Xiong",
                    "Junfeng Yao",
                    "Yawei Li",
                    "Yulun Zhang",
                    "R. Timofte",
                    "L. Gool",
                    "Kunpeng Du",
                    "Xiaofei Wang",
                    "Xiangyu Kong",
                    "Jinlong Wu",
                    "Dafeng Zhang",
                    "Jianxing Zhang",
                    "Shuai Liu",
                    "Furui Bai",
                    "Chaoyu Feng",
                    "Hao Wang",
                    "Yuqian Zhang",
                    "Guangqi Shao",
                    "Xiaotao Wang",
                    "Lei Lei",
                    "Rongjian Xu",
                    "Zhilu Zhang",
                    "Yunjin Chen",
                    "Dongwei Ren",
                    "W. Zuo",
                    "Qi Wu",
                    "Mingyan Han",
                    "Shen Cheng",
                    "Haipeng Li",
                    "Ting Jiang",
                    "Chengzhi Jiang",
                    "Xinpeng Li",
                    "Jinting Luo",
                    "Wenjie Lin",
                    "Leiting Yu",
                    "Haoqiang Fan",
                    "Shuaicheng Liu",
                    "Aditya Arora",
                    "Syed Waqas Zamir",
                    "Javier V\u00e1zquez-Corral",
                    "K. Derpanis",
                    "Michael S. Brown",
                    "Hao Li",
                    "Zhihao Zhao",
                    "Jin-shan Pan",
                    "Jiangxin Dong",
                    "Jinhui Tang",
                    "Bo Yang",
                    "Jingxiang Chen"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Multi-modal Learning",
                    "Machine Translation",
                    "Knowledge Distillation"
                ],
                "publications": [
                    {
                        "title": "Macaw-LLM: Multi-Modal Language Modeling with Image, Audio, Video, and Text Integration",
                        "abstract": "Although instruction-tuned large language models (LLMs) have exhibited remarkable capabilities across various NLP tasks, their effectiveness on other data modalities beyond text has not been fully studied. In this work, we propose Macaw-LLM, a novel multi-modal LLM that seamlessly integrates visual, audio, and textual information. Macaw-LLM consists of three main components: a modality module for encoding multi-modal data, a cognitive module for harnessing pretrained LLMs, and an alignment module for harmonizing diverse representations. Our novel alignment module seamlessly bridges multi-modal features to textual features, simplifying the adaptation process from the modality modules to the cognitive module. In addition, we construct a large-scale multi-modal instruction dataset in terms of multi-turn dialogue, including 69K image instances and 50K video instances. We have made our data, code and model publicly available, which we hope can pave the way for future research in multi-modal LLMs and expand the capabilities of LLMs to handle diverse data modalities and address complex real-world scenarios."
                    },
                    {
                        "title": "Towards a Unified Training for Levenshtein Transformer",
                        "abstract": "Levenshtein Transformer (LevT) is a widely-used text-editing model, which generates a sequence based on editing operations (deletion and insertion) in a non-autoregressive manner. However, it is challenging to train the key refinement components of LevT due to training-inference discrepancy. By carefully designing experiments, our work reveals that the deletion module is under-trained while the insertion module is over-trained due to the imbalance training signals for the two refinement modules. Based on these observations, we further propose a dual learning approach that can remedy the imbalance training by feeding an initial input to both refinement modules, which is consistent with the process in inference. Experimental results on three representative NLP tasks demonstrate the effectiveness and universality of the proposed approach. 1"
                    },
                    {
                        "title": "Is ChatGPT A Good Translator? A Preliminary Study",
                        "abstract": "This report provides a preliminary evaluation of ChatGPT for machine translation, including translation prompt, multilingual translation, and translation robustness. We adopt the prompts advised by ChatGPT to trigger its translation ability and \ufb01nd that the candidate prompts generally work well and show minor performance differences. By evaluating on a number of benchmark test sets 1 , we \ufb01nd that ChatGPT performs competitively with commercial translation products (e.g., Google Translate) on high-resource European languages but lags behind signi\ufb01cantly on low-resource or distant languages. For distant languages, we explore an interesting strategy named pivot prompting that asks ChatGPT to translate the source sentence into a high-resource pivot language before into the target language, which improves the translation performance signi\ufb01cantly. As for the translation robustness, ChatGPT does not perform as well as the commercial systems on biomedical abstracts or Reddit comments but is potentially a good translator for spoken language."
                    },
                    {
                        "title": "Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate",
                        "abstract": "Modern large language models (LLMs) like ChatGPT have shown remarkable performance on general language tasks but still struggle on complex reasoning tasks, which drives the research on cognitive behaviors of LLMs to explore human-like problem-solving strategies. Along this direction, one representative strategy is self-reflection, which asks an LLM to refine the solution with the feedback generated by itself iteratively. However, our study shows that such reflection-style methods suffer from the Degeneration-of-Thought (DoT) problem: once the LLM has established confidence in its solutions, it is unable to generate novel thoughts later through reflection even if its initial stance is incorrect. To address the DoT problem, we propose a Multi-Agent Debate (MAD) framework, in which multiple agents express their arguments in the state of\"tit for tat\"and a judge manages the debate process to obtain a final solution. Clearly, our MAD framework encourages divergent thinking in LLMs which would be helpful for tasks that require deep levels of contemplation. Experiment results on two challenging datasets, commonsense machine translation and counter-intuitive arithmetic reasoning, demonstrate the effectiveness of our MAD framework. Extensive analyses suggest that the adaptive break of debate and the modest level of\"tit for tat\"state are required for MAD to obtain good performance. Moreover, we find that LLMs might not be a fair judge if different LLMs are used for agents. Code is available at https://github.com/Skytliang/Multi-Agents-Debate."
                    },
                    {
                        "title": "The MineTrans Systems for IWSLT 2023 Offline Speech Translation and Speech-to-Speech Translation Tasks",
                        "abstract": "This paper presents the extscMineTrans English-to-Chinese speech translation systems developed for two challenge tracks of IWSLT 2023, i.e., Offline Speech Translation (S2T) and Speech-to-Speech Translation (S2ST). For the S2T track, extscMineTrans employs a practical cascaded system to explore the limits of translation performance in both constrained and unconstrained settings, where the whole system consists of automatic speech recognition (ASR), punctuation recognition (PC), and machine translation (MT) modules. We also investigate the effectiveness of multiple ASR architectures and explore two MT strategies: supervised in-domain fine-tuning and prompt-guided translation using a large language model. For the S2ST track, we explore a speech-to-unit (S2U) framework to build an end-to-end S2ST system. This system encodes the target speech as discrete units via our trained HuBERT. Then it leverages the standard sequence-to-sequence model to directly learn the mapping between source speech and discrete units without any auxiliary recognition tasks (i.e., ASR and MT tasks). Various efforts are made to improve the extscMineTrans\u2019s performance, such as acoustic model pre-training on large-scale data, data filtering, data augmentation, speech segmentation, knowledge distillation, consistency training, model ensembles, etc."
                    },
                    {
                        "title": "Toward Accurate Post-Training Quantization for Image Super Resolution",
                        "abstract": "Model quantization is a crucial step for deploying super resolution (SR) networks on mobile devices. However, existing works focus on quantization-aware training, which requires complete dataset and expensive computational overhead. In this paper, we study post-training quantization (PTQ) for image super resolution using only a few unlabeled calibration images. As the SR model aims to maintain the texture and color information of input images, the distribution of activations are long-tailed, asymmetric and highly dynamic compared with classification models. To this end, we introduce the density-based dual clipping to cut off the outliers based on analyzing the asymmetric bounds of activations. Moreover, we present a novel pixel aware calibration method with the supervision of the full-precision model to accommodate the highly dynamic range of different samples. Extensive experiments demonstrate that the proposed method significantly outperforms existing PTQ algorithms on various models and datasets. For instance, we get a 2.091 dB increase on Urban100 benchmark when quantizing $EDSR\\times4$ to 4-bit with 100 unlabeled images. Our code is available at both PyTorch and MindSpore."
                    },
                    {
                        "title": "Data Upcycling Knowledge Distillation for Image Super-Resolution",
                        "abstract": "Knowledge distillation (KD) compresses deep neural networks by transferring task-related knowledge from cumbersome pre-trained teacher models to compact student models. However, current KD methods for super-resolution (SR) networks overlook the nature of SR task that the outputs of the teacher model are noisy approximations to the ground-truth distribution of high-quality images (GT), which shades the teacher model's knowledge to result in limited KD effects. To utilize the teacher model beyond the GT upper-bound, we present the Data Upcycling Knowledge Distillation (DUKD), to transfer the teacher model's knowledge to the student model through the upcycled in-domain data derived from training data. Besides, we impose label consistency regularization to KD for SR by the paired invertible augmentations to improve the student model's performance and robustness. Comprehensive experiments demonstrate that the DUKD method significantly outperforms previous arts on several SR tasks."
                    },
                    {
                        "title": "Document-Level Machine Translation with Large Language Models",
                        "abstract": "Large language models (LLMs) such as ChatGPT can produce coherent, cohesive, relevant, and fluent answers for various natural language processing (NLP) tasks. Taking document-level machine translation (MT) as a testbed, this paper provides an in-depth evaluation of LLMs' ability on discourse modeling. The study focuses on three aspects: 1) Effects of Context-Aware Prompts, where we investigate the impact of different prompts on document-level translation quality and discourse phenomena; 2) Comparison of Translation Models, where we compare the translation performance of ChatGPT with commercial MT systems and advanced document-level MT methods; 3) Analysis of Discourse Modelling Abilities, where we further probe discourse knowledge encoded in LLMs and shed light on impacts of training techniques on discourse modeling. By evaluating on a number of benchmarks, we surprisingly find that LLMs have demonstrated superior performance and show potential to become a new paradigm for document-level translation: 1) leveraging their powerful long-text modeling capabilities, GPT-3.5 and GPT-4 outperform commercial MT systems in terms of human evaluation; 2) GPT-4 demonstrates a stronger ability for probing linguistic knowledge than GPT-3.5. This work highlights the challenges and opportunities of LLMs for MT, which we hope can inspire the future design and evaluation of LLMs.We release our data and annotations at https://github.com/longyuewangdcu/Document-MT-LLM."
                    },
                    {
                        "title": "On the Cultural Gap in Text-to-Image Generation",
                        "abstract": "One challenge in text-to-image (T2I) generation is the inadvertent reflection of culture gaps present in the training data, which signifies the disparity in generated image quality when the cultural elements of the input text are rarely collected in the training set. Although various T2I models have shown impressive but arbitrary examples, there is no benchmark to systematically evaluate a T2I model's ability to generate cross-cultural images. To bridge the gap, we propose a Challenging Cross-Cultural (C3) benchmark with comprehensive evaluation criteria, which can assess how well-suited a model is to a target culture. By analyzing the flawed images generated by the Stable Diffusion model on the C3 benchmark, we find that the model often fails to generate certain cultural objects. Accordingly, we propose a novel multi-modal metric that considers object-text alignment to filter the fine-tuning data in the target culture, which is used to fine-tune a T2I model to improve cross-cultural generation. Experimental results show that our multi-modal metric provides stronger data selection performance on the C3 benchmark than existing metrics, in which the object-text alignment is crucial. We release the benchmark, data, code, and generated images to facilitate future research on culturally diverse T2I generation (https://github.com/longyuewangdcu/C3-Bench)."
                    },
                    {
                        "title": "GenImage: A Million-Scale Benchmark for Detecting AI-Generated Image",
                        "abstract": "The extraordinary ability of generative models to generate photographic images has intensified concerns about the spread of disinformation, thereby leading to the demand for detectors capable of distinguishing between AI-generated fake images and real images. However, the lack of large datasets containing images from the most advanced image generators poses an obstacle to the development of such detectors. In this paper, we introduce the GenImage dataset, which has the following advantages: 1) Plenty of Images, including over one million pairs of AI-generated fake images and collected real images. 2) Rich Image Content, encompassing a broad range of image classes. 3) State-of-the-art Generators, synthesizing images with advanced diffusion models and GANs. The aforementioned advantages allow the detectors trained on GenImage to undergo a thorough evaluation and demonstrate strong applicability to diverse images. We conduct a comprehensive analysis of the dataset and propose two tasks for evaluating the detection method in resembling real-world scenarios. The cross-generator image classification task measures the performance of a detector trained on one generator when tested on the others. The degraded image classification task assesses the capability of the detectors in handling degraded images such as low-resolution, blurred, and compressed images. With the GenImage dataset, researchers can effectively expedite the development and evaluation of superior AI-generated image detectors in comparison to prevailing methodologies."
                    },
                    {
                        "title": "Is ChatGPT A Good Translator? Yes With GPT-4 As The Engine",
                        "abstract": "This report provides a preliminary evaluation of ChatGPT for machine translation, including translation prompt, multilingual translation, and translation robustness. We adopt the prompts advised by ChatGPT to trigger its translation ability and find that the candidate prompts generally work well with minor performance differences. By evaluating on a number of benchmark test sets, we find that ChatGPT performs competitively with commercial translation products (e.g., Google Translate) on high-resource European languages but lags behind significantly on low-resource or distant languages. As for the translation robustness, ChatGPT does not perform as well as the commercial systems on biomedical abstracts or Reddit comments but exhibits good results on spoken language. Further, we explore an interesting strategy named $\\mathbf{pivot~prompting}$ for distant languages, which asks ChatGPT to translate the source sentence into a high-resource pivot language before into the target language, improving the translation performance noticeably. With the launch of the GPT-4 engine, the translation performance of ChatGPT is significantly boosted, becoming comparable to commercial translation products, even for distant languages. Human analysis on Google Translate and ChatGPT suggests that ChatGPT with GPT-3.5 tends to generate more hallucinations and mis-translation errors while that with GPT-4 makes the least errors. In other words, ChatGPT has already become a good translator. Please refer to our Github project for more details: https://github.com/wxjiao/Is-ChatGPT-A-Good-Translator"
                    },
                    {
                        "title": "A Survey on Zero Pronoun Translation",
                        "abstract": "Zero pronouns (ZPs) are frequently omitted in pro-drop languages (e.g. Chinese, Hungarian, and Hindi), but should be recalled in non-pro-drop languages (e.g. English). This phenomenon has been studied extensively in machine translation (MT), as it poses a significant challenge for MT systems due to the difficulty in determining the correct antecedent for the pronoun. This survey paper highlights the major works that have been undertaken in zero pronoun translation (ZPT) after the neural revolution so that researchers can recognize the current state and future directions of this field. We provide an organization of the literature based on evolution, dataset, method, and evaluation. In addition, we compare and analyze competing models and evaluation metrics on different benchmarks. We uncover a number of insightful findings such as: 1) ZPT is in line with the development trend of large language model; 2) data limitation causes learning bias in languages and domains; 3) performance improvements are often reported on single benchmarks, but advanced methods are still far from real-world use; 4) general-purpose metrics are not reliable on nuances and complexities of ZPT, emphasizing the necessity of targeted metrics; 5) apart from commonly-cited errors, ZPs will cause risks of gender bias."
                    },
                    {
                        "title": "Exploring Human-Like Translation Strategy with Large Language Models",
                        "abstract": "Large language models (LLMs) have demonstrated impressive capabilities in general scenarios, exhibiting a level of aptitude that approaches, in some aspects even surpasses, human-level intelligence. Among their numerous skills, the translation abilities of LLMs have received considerable attention. Compared to typical machine translation that focuses solely on source-to-target mapping, LLM-based translation can potentially mimic the human translation process, which might take preparatory steps to ensure high-quality translation. This work explores this possibility by proposing the MAPS framework, which stands for Multi-Aspect Prompting and Selection. Specifically, we enable LLMs first to analyze the given source sentence and induce three aspects of translation-related knowledge (keywords, topics, and relevant demonstrations) to guide the final translation process. Moreover, we employ a selection mechanism based on quality estimation to filter out noisy and unhelpful knowledge. Both automatic (3 LLMs \u00d7 11 directions \u00d7 2 automatic metrics) and human evaluation (preference study and MQM) demonstrate the effectiveness of MAPS. Further analysis shows that by mimicking the human translation process, MAPS reduces various translation errors such as hallucination, ambiguity, mistranslation, awkward style, untranslated text, and omission. Source code is available at https://github.com/zwhe99/MAPS-mt."
                    },
                    {
                        "title": "ParroT: Translating During Chat Using Large Language Models",
                        "abstract": "Large language models (LLMs) like ChatGPT have exhibited remarkable abilities on a wide range of natural language processing~(NLP) tasks, including various machine translation abilities accomplished during chat. However, these models are only accessible through restricted APIs, which creates barriers to new research and advancements in the field. Therefore, we propose ParroT, a framework to enhance and regulate the translation abilities during chat based on open-source LLMs (e.g., LLaMA), human-written translation and feedback data. Specifically, ParroT reformulates translation data into the instruction-following style, and introduces a\"$\\mathbf{Hint}$\"field for incorporating extra requirements to regulate the translation process. Accordingly, we propose three instruction types for finetuning ParroT models, including translation instruction, contrastive instruction, and error-guided instruction. Experiments on Flores subsets and WMT22 test sets suggest that translation instruction improves the translation performance of vanilla LLMs significantly while error-guided instruction can lead to further improvement, which demonstrates the importance of learning from low-quality translations annotated by humans. We also demonstrate the potential of automatic evaluation tools in providing quality information of translations, when constructing error-guided instructions for directions that lack human annotation data. Please refer to our Github project for more implementation details: https://github.com/wxjiao/ParroT"
                    },
                    {
                        "title": "Cross-modality Data Augmentation for End-to-End Sign Language Translation",
                        "abstract": "End-to-end sign language translation (SLT) aims to convert sign language videos into spoken language texts directly without intermediate representations. It has been a challenging task due to the modality gap between sign videos and texts and the data scarcity of labeled data. Due to these challenges, the input and output distributions of end-to-end sign language translation (i.e., video-to-text) are less effective compared to the gloss-to-text approach (i.e., text-to-text). To tackle these challenges, we propose a novel Cross-modality Data Augmentation (XmDA) framework to transfer the powerful gloss-to-text translation capabilities to end-to-end sign language translation (i.e. video-to-text) by exploiting pseudo gloss-text pairs from the sign gloss translation model. Specifically, XmDA consists of two key components, namely, cross-modality mix-up and cross-modality knowledge distillation. The former explicitly encourages the alignment between sign video features and gloss embeddings to bridge the modality gap. The latter utilizes the generation knowledge from gloss-to-text teacher models to guide the spoken language text generation. Experimental results on two widely used SLT datasets, i.e., PHOENIX-2014T and CSL-Daily, demonstrate that the proposed XmDA framework significantly and consistently outperforms the baseline models. Extensive analyses confirm our claim that XmDA enhances spoken language text generation by reducing the representation distance between videos and texts, as well as improving the processing of low-frequency words and long sentences."
                    },
                    {
                        "title": "Revisiting Non-Autoregressive Translation at Scale",
                        "abstract": "In real-world systems, scaling has been critical for improving the translation quality in autoregressive translation (AT), which however has not been well studied for non-autoregressive translation (NAT). In this work, we bridge the gap by systematically studying the impact of scaling on NAT behaviors. Extensive experiments on six WMT benchmarks over two advanced NAT models show that scaling can alleviate the commonly-cited weaknesses of NAT models, resulting in better translation performance. To reduce the side-effect of scaling on decoding speed, we empirically investigate the impact of NAT encoder and decoder on the translation performance. Experimental results on the large-scale WMT20 En-De show that the asymmetric architecture (e.g. bigger encoder and smaller decoder) can achieve comparable performance with the scaling model, while maintaining the superiority of decoding speed with standard NAT models. To this end, we establish a new benchmark by validating scaled NAT models on the scaled dataset, which can be regarded as a strong baseline for future works. We release code and system outputs at https://github.com/DeepLearnXMU/Scaling4NAT."
                    },
                    {
                        "title": "NTIRE 2023 Challenge on Image Denoising: Methods and Results",
                        "abstract": "This paper reviews the NTIRE 2023 challenge on image denoising (\u03c3 = 50) with a focus on the proposed solutions and results. The aim is to obtain a network design capable to produce high-quality results with the best performance measured by PSNR for image denoising. Independent additive white Gaussian noise (AWGN) is assumed and the noise level is 50. The challenge had 225 registered participants, and 16 teams made valid submissions. They gauge the state-of-the-art for image denoising."
                    }
                ]
            }
        }
    },
    "2410.05578": {
        "paper_data": {
            "title": "Swift Sampler: Efficient Learning of Sampler by 10 Parameters",
            "url": "http://arxiv.org/abs/2410.05578v1",
            "arxiv_id": "2410.05578",
            "authors": [
                "Jiawei Yao",
                "Chuming Li",
                "Canran Xiao"
            ],
            "abstract": "Data selection is essential for training deep learning models. An effective data sampler assigns proper sampling probability for training data and helps the model converge to a good local minimum with high performance. Previous studies in data sampling are mainly based on heuristic rules or learning through a huge amount of time-consuming trials. In this paper, we propose an automatic \\textbf{swift sampler} search algorithm, \\textbf{SS}, to explore automatically learning effective samplers efficiently. In particular, \\textbf{SS} utilizes a novel formulation to map a sampler to a low dimension of hyper-parameters and uses an approximated local minimum to quickly examine the quality of a sampler. Benefiting from its low computational expense, \\textbf{SS} can be applied on large-scale data sets with high efficiency. Comprehensive experiments on various tasks demonstrate that \\textbf{SS} powered sampling can achieve obvious improvements (e.g., 1.5\\% on ImageNet) and transfer among different neural networks. Project page: https://github.com/Alexander-Yao/Swift-Sampler.",
            "introduction": "   1 Introduction  Training data plays a pivotal role in deep learning tasks. The sampling probability of data in the training process can significantly influence the performance of the learned model. A set of works (Jiang et\u00a0al., 2019; Katharopoulos and Fleuret, 2018; Needell et\u00a0al., 2014; Johnson and Guestrin, 2018; Han et\u00a0al., 2018; Hu et\u00a0al., 2024) have demonstrated improvements in model training by sampling data according to different features and under different rules. These studies reveal that how to sample examples during training is nontrivial and sampling data in a totally uniform way is not always the optimal choice.   Previous works on data sampling strategy include two main categories: human-defined rules and learning-based methods. Some works (Han et\u00a0al., 2018; Jiang et\u00a0al., 2019; Katharopoulos and Fleuret, 2018; Hacohen and Weinshall, 2019; Kumar et\u00a0al., 2010) proposed manually designed rules, such as setting the sampling probability of training examples with loss values larger than a threshold as zero, or making the sampling probability proportional to the gradient norms. Such human-defined rules are designed for specific tasks and hardly adapts to different scenarios, because optimal sampling strategy varies among different tasks and data sets. Learning-based methods (Fan et\u00a0al., 2017; Jiang et\u00a0al., 2017; Ren et\u00a0al., 2018) explore an automatic way to assign sampling probability to a given learning task, including sample-based and differential based methods. Sample-based methods take advantage of deep reinforcement learning (DRL) (Fan et\u00a0al., 2017) to model the training process of the target model as an environment and use a DRL model to learn the optimal sampling strategy during training. These methods need hundreds of replays of the training process and require too much searching cost to be applied on large-scale data sets, e.g., ImageNet (Russakovsky et\u00a0al., 2015). Differential based methods (Ren et\u00a0al., 2018; Shu et\u00a0al., 2019) assume a completely clean meta dataset and use the inner product between the gradients of training data and meta data to sample or reweight training data. The problem is that the clean meta data is not guaranteed to be available for all scenarios, and complex changes are made in the training process.   To explore an automatic way of sampler search, we focus on improving sample-based search methods as they do not require extra meta data and impose no change in the training process when applying the obtained sampler. Typically, a sample-based method repeatedly uses an agent (e.g., Bayesian Optimization or Deep Reinforcement Learning) to sample a sampler and evaluate the objective function value of the sampled sampler to update the agent. There are three challenges in designing a sample-based method for automatic sampler search: (1) High dimension. A sampler is defined by a vector of sampling probabilities of all instances in the training set. Therefore, the complexity of searching the optimal sampler in a sample-based way increases exponentially with the number of training instances. (2) Sharpness. As great difference exists among gradients of various instances in the training set, sampling probability of some instances has a significant impact on the performance of resulted models. Such property results in the sharpness of the objective function for samplers and harms the efficiency of the agent\u2019s learning.",
            "references": [
                {
                    "title": "Cross-Modal Hashing Method with Properties of Hamming Space: A New Perspective.",
                    "abstract": "Cross-modal hashing (CMH) has attracted considerable attention in recent years. Almost all existing CMH methods primarily focus on reducing the modality gap and semantic gap, i.e., aligning multi-modal features and their semantics in Hamming space, without taking into account the space gap, i.e., difference between the real number space and the Hamming space. In fact, the space gap can affect the performance of CMH methods. In this paper, we analyze and demonstrate how the space gap affects the existing CMH methods, which therefore raises two problems: solution space compression and loss function oscillation. These two problems eventually cause the retrieval performance deteriorating. Based on these findings, we propose a novel algorithm, namely Semantic Channel Hashing (SCH). Firstly, we classify sample pairs into fully semantic-similar, partially semantic-similar, and semantic-negative ones based on their similarity and impose different constraints on them, respectively, to ensure that the entire Hamming space is utilized. Then, we introduce a semantic channel to alleviate the issue of loss function oscillation. Experimental results on three public datasets demonstrate that SCH outperforms the state-of-the-art methods. Furthermore, experimental validations are provided to substantiate the conjectures regarding solution space compression and loss function oscillation, offering visual evidence of their impact on the CMH methods. Codes are available at https://github.com/hutt94/SCH."
                },
                {
                    "title": "Key Point Sensitive Loss for Long-Tailed Visual Recognition",
                    "abstract": "For long-tailed distributed data, existing classification models often learn overwhelmingly on the head classes while ignoring the tail classes, resulting in poor generalization capability. To address this problem, we thereby propose a new approach in this paper, in which a key point sensitive (KPS) loss is presented to regularize the key points strongly to improve the generalization performance of the classification model. Meanwhile, in order to improve the performance on tail classes, the proposed KPS loss also assigns relatively large margins on tail classes. Furthermore, we propose a gradient adjustment (GA) optimization strategy to re-balance the gradients of positive and negative samples for each class. By virtue of the gradient analysis of the loss function, it is found that the tail classes always receive negative signals during training, which misleads the tail prediction to be biased towards the head. The proposed GA strategy can circumvent excessive negative signals on tail classes and further improve the overall classification accuracy. Extensive experiments conducted on long-tailed benchmarks show that the proposed method is capable of significantly improving the classification accuracy of the model in tail classes while maintaining competent performance in head classes."
                },
                {
                    "title": "Improving Auto-Augment via Augmentation-Wise Weight Sharing",
                    "abstract": "The recent progress on automatically searching augmentation policies has boosted the performance substantially for various tasks. A key component of automatic augmentation search is the evaluation process for a particular augmentation policy, which is utilized to return reward and usually runs thousands of times. A plain evaluation process, which includes full model training and validation, would be time-consuming. To achieve efficiency, many choose to sacrifice evaluation reliability for speed. In this paper, we dive into the dynamics of augmented training of the model. This inspires us to design a powerful and efficient proxy task based on the Augmentation-Wise Weight Sharing (AWS) to form a fast yet accurate evaluation process in an elegant way. Comprehensive analysis verifies the superiority of this approach in terms of effectiveness and efficiency. The augmentation policies found by our method achieve superior accuracies compared with existing auto-augmentation search methods. On CIFAR-10, we achieve a top-1 error rate of 1.24%, which is currently the best performing single model without extra training data. On ImageNet, we get a top-1 error rate of 20.36% for ResNet-50, which leads to 3.34% absolute error rate reduction over the baseline augmentation."
                },
                {
                    "title": "Accelerating Deep Learning by Focusing on the Biggest Losers",
                    "abstract": "This paper introduces Selective-Backprop, a technique that accelerates the training of deep neural networks (DNNs) by prioritizing examples with high loss at each iteration. Selective-Backprop uses the output of a training example's forward pass to decide whether to use that example to compute gradients and update parameters, or to skip immediately to the next example. By reducing the number of computationally-expensive backpropagation steps performed, Selective-Backprop accelerates training. Evaluation on CIFAR10, CIFAR100, and SVHN, across a variety of modern image models, shows that Selective-Backprop converges to target error rates up to 3.5x faster than with standard SGD and between 1.02--1.8x faster than a state-of-the-art importance sampling approach. Further acceleration of 26% can be achieved by using stale forward pass results for selection, thus also skipping forward passes of low priority examples."
                },
                {
                    "title": "LAW: Learning to Auto Weight",
                    "abstract": "Example weighting algorithm is an effective solution to the training bias problem. However, typical methods are usually limited to human knowledge and require laborious tuning of hyperparameters. In this study, we propose a novel example weighting framework called Learning to Auto Weight (LAW), which can learn weighting policy from data adaptively based on reinforcement learning (RL). \nTo shrink the huge searching space in a complete training process, we divide the training procedure consisting of numerous iterations into a small number of stages, and then search a low-deformational continuous vector as action, which determines the weight of each sample. To make training more efficient, we make an innovative design of the reward to remove randomness during the RL process. \nExperimental results demonstrate the superiority of weighting policy explored by LAW over standard training pipeline. Especially, compared with baselines, LAW can find a better weighting schedule which achieves higher accuracy in the origin CIFAR dataset, and over 10% higher in accuracy on the contaminated CIFAR dataset with 30% label noises. Our code will be released soon."
                },
                {
                    "title": "AM-LFS: AutoML for Loss Function Search",
                    "abstract": "Designing an effective loss function plays an important role in visual analysis. Most existing loss function designs rely on hand-crafted heuristics that require domain experts to explore the large design space, which is usually sub-optimal and time-consuming. In this paper, we propose AutoML for Loss Function Search (AM-LFS) which leverages REINFORCE to search loss functions during the training process. The key contribution of this work is the design of search space which can guarantee the generalization and transferability on different vision tasks by including a bunch of existing prevailing loss functions in a unified formulation. We also propose an efficient optimization framework which can dynamically optimize the parameters of loss function's distribution during training. Extensive experimental results on four benchmark datasets show that, without any tricks, our method outperforms existing hand-crafted loss functions in various computer vision tasks."
                },
                {
                    "title": "Online Hyper-Parameter Learning for Auto-Augmentation Strategy",
                    "abstract": "Data augmentation is critical to the success of modern deep learning techniques. In this paper, we propose Online Hyper-parameter Learning for Auto-Augmentation (OHL-Auto-Aug), an economical solution that learns the augmentation policy distribution along with network training. Unlike previous methods on auto-augmentation that search augmentation strategies in an offline manner, our method formulates the augmentation policy as a parameterized probability distribution, thus allowing its parameters to be optimized jointly with network parameters. Our proposed OHL-Auto-Aug eliminates the need of re-training and dramatically reduces the cost of the overall search process, while establishes significantly accuracy improvements over baseline models. On both CIFAR-10 and ImageNet, our method achieves remarkable on search accuracy, 60x faster on CIFAR-10 and 24x faster on ImageNet, while maintaining competitive accuracies."
                },
                {
                    "title": "On The Power of Curriculum Learning in Training Deep Networks",
                    "abstract": "Training neural networks is traditionally done by providing a sequence of random mini-batches sampled uniformly from the entire training data. In this work, we analyze the effect of curriculum learning, which involves the non-uniform sampling of mini-batches, on the training of deep networks, and specifically CNNs trained for image recognition. To employ curriculum learning, the training algorithm must resolve 2 problems: (i) sort the training examples by difficulty; (ii) compute a series of mini-batches that exhibit an increasing level of difficulty. We address challenge (i) using two methods: transfer learning from some competitive ``teacher\" network, and bootstrapping. In our empirical evaluation, both methods show similar benefits in terms of increased learning speed and improved final performance on test data. We address challenge (ii) by investigating different pacing functions to guide the sampling. The empirical investigation includes a variety of network architectures, using images from CIFAR-10, CIFAR-100 and subsets of ImageNet. We conclude with a novel theoretical analysis of curriculum learning, where we show how it effectively modifies the optimization landscape. We then define the concept of an ideal curriculum, and show that under mild conditions it does not change the corresponding global minimum of the optimization function."
                },
                {
                    "title": "Meta-Weight-Net: Learning an Explicit Mapping For Sample Weighting",
                    "abstract": "Current deep neural networks (DNNs) can easily overfit to biased training data with corrupted labels or class imbalance. Sample re-weighting strategy is commonly used to alleviate this issue by designing a weighting function mapping from training loss to sample weight, and then iterating between weight recalculating and classifier updating. Current approaches, however, need manually pre-specify the weighting function as well as its additional hyper-parameters. It makes them fairly hard to be generally applied in practice due to the significant variation of proper weighting schemes relying on the investigated problem and training data. To address this issue, we propose a method capable of adaptively learning an explicit weighting function directly from data. The weighting function is an MLP with one hidden layer, constituting a universal approximator to almost any continuous functions, making the method able to fit a wide range of weighting functions including those assumed in conventional research. Guided by a small amount of unbiased meta-data, the parameters of the weighting function can be finely updated simultaneously with the learning process of the classifiers. Synthetic and real experiments substantiate the capability of our method for achieving proper weighting functions in class imbalance and noisy label cases, fully complying with the common settings in traditional methods, and more complicated scenarios beyond conventional cases. This naturally leads to its better accuracy than other state-of-the-art methods."
                },
                {
                    "title": "Dynamic Curriculum Learning for Imbalanced Data Classification",
                    "abstract": "Human attribute analysis is a challenging task in the field of computer vision. One of the significant difficulties is brought from largely imbalance-distributed data. Conventional techniques such as re-sampling and cost-sensitive learning require prior-knowledge to train the system. To address this problem, we propose a unified framework called Dynamic Curriculum Learning (DCL) to adaptively adjust the sampling strategy and loss weight in each batch, which results in better ability of generalization and discrimination. Inspired by curriculum learning, DCL consists of two-level curriculum schedulers: (1) sampling scheduler which manages the data distribution not only from imbalance to balance but also from easy to hard; (2) loss scheduler which controls the learning importance between classification and metric learning loss. With these two schedulers, we achieve state-of-the-art performance on the widely used face attribute dataset CelebA and pedestrian attribute dataset RAP."
                },
                {
                    "title": "An Empirical Study of Example Forgetting during Deep Neural Network Learning",
                    "abstract": "Inspired by the phenomenon of catastrophic forgetting, we investigate the learning dynamics of neural networks as they train on single classification tasks. Our goal is to understand whether a related phenomenon occurs when data does not undergo a clear distributional shift. We define a \u201cforgetting event\u201d to have occurred when an individual training example transitions from being classified correctly to incorrectly over the course of learning. Across several benchmark data sets, we find that: (i) certain examples are forgotten with high frequency, and some not at all; (ii) a data set\u2019s (un)forgettable examples generalize across neural architectures; and (iii) based on forgetting dynamics, a significant fraction of examples can be omitted from the training data set while still maintaining state-of-the-art generalization performance."
                },
                {
                    "title": "AutoAugment: Learning Augmentation Policies from Data",
                    "abstract": "Data augmentation is an effective technique for improving the accuracy of modern image classifiers. However, current data augmentation implementations are manually designed. In this paper, we describe a simple procedure called AutoAugment to automatically search for improved data augmentation policies. In our implementation, we have designed a search space where a policy consists of many sub-policies, one of which is randomly chosen for each image in each mini-batch. A sub-policy consists of two operations, each operation being an image processing function such as translation, rotation, or shearing, and the probabilities and magnitudes with which the functions are applied. We use a search algorithm to find the best policy such that the neural network yields the highest validation accuracy on a target dataset. Our method achieves state-of-the-art accuracy on CIFAR-10, CIFAR-100, SVHN, and ImageNet (without additional data). On ImageNet, we attain a Top-1 accuracy of 83.5% which is 0.4% better than the previous record of 83.1%. On CIFAR-10, we achieve an error rate of 1.5%, which is 0.6% better than the previous state-of-the-art. Augmentation policies we find are transferable between datasets. The policy learned on ImageNet transfers well to achieve significant improvements on other datasets, such as Oxford Flowers, Caltech-101, Oxford-IIT Pets, FGVC Aircraft, and Stanford Cars."
                },
                {
                    "title": "Co-teaching: Robust training of deep neural networks with extremely noisy labels",
                    "abstract": "Deep learning with noisy labels is practically challenging, as the capacity of deep models is so high that they can totally memorize these noisy labels sooner or later during training. Nonetheless, recent studies on the memorization effects of deep neural networks show that they would first memorize training data of clean labels and then those of noisy labels. Therefore in this paper, we propose a new deep learning paradigm called Co-teaching for combating with noisy labels. Namely, we train two deep neural networks simultaneously, and let them teach each other given every mini-batch: firstly, each network feeds forward all data and selects some data of possibly clean labels; secondly, two networks communicate with each other what data in this mini-batch should be used for training; finally, each network back propagates the data selected by its peer network and updates itself. Empirical results on noisy versions of MNIST, CIFAR-10 and CIFAR-100 demonstrate that Co-teaching is much superior to the state-of-the-art methods in the robustness of trained deep models."
                },
                {
                    "title": "Learning to Reweight Examples for Robust Deep Learning",
                    "abstract": "Deep neural networks have been shown to be very powerful modeling tools for many supervised learning tasks involving complex input patterns. However, they can also easily overfit to training set biases and label noises. In addition to various regularizers, example reweighting algorithms are popular solutions to these problems, but they require careful tuning of additional hyperparameters, such as example mining schedules and regularization hyperparameters. In contrast to past reweighting methods, which typically consist of functions of the cost value of each example, in this work we propose a novel meta-learning algorithm that learns to assign weights to training examples based on their gradient directions. To determine the example weights, our method performs a meta gradient descent step on the current mini-batch example weights (which are initialized from zero) to minimize the loss on a clean unbiased validation set. Our proposed method can be easily implemented on any type of deep network, does not require any additional hyperparameter tuning, and achieves impressive performance on class imbalance and corrupted label problems where only a small amount of clean validation data is available."
                },
                {
                    "title": "Not All Samples Are Created Equal: Deep Learning with Importance Sampling",
                    "abstract": "Deep neural network training spends most of the computation on examples that are properly handled, and could be ignored. We propose to mitigate this phenomenon with a principled importance sampling scheme that focuses computation on \"informative\" examples, and reduces the variance of the stochastic gradients during training. Our contribution is twofold: first, we derive a tractable upper bound to the per-sample gradient norm, and second we derive an estimator of the variance reduction achieved with importance sampling, which enables us to switch it on when it will result in an actual speedup. The resulting scheme can be used by changing a few lines of code in a standard SGD procedure, and we demonstrate experimentally, on image classification, CNN fine-tuning, and RNN training, that for a fixed wall-clock time budget, it provides a reduction of the train losses of up to an order of magnitude and a relative improvement of test errors between 5% and 17%."
                },
                {
                    "title": "Neural Architecture Search with Bayesian Optimisation and Optimal Transport",
                    "abstract": "Bayesian Optimisation (BO) refers to a class of methods for global optimisation of a function $f$ which is only accessible via point evaluations. It is typically used in settings where $f$ is expensive to evaluate. A common use case for BO in machine learning is model selection, where it is not possible to analytically model the generalisation performance of a statistical model, and we resort to noisy and expensive training and validation procedures to choose the best model. Conventional BO methods have focused on Euclidean and categorical domains, which, in the context of model selection, only permits tuning scalar hyper-parameters of machine learning algorithms. However, with the surge of interest in deep learning, there is an increasing demand to tune neural network \\emph{architectures}. In this work, we develop NASBOT, a Gaussian process based BO framework for neural architecture search. To accomplish this, we develop a distance metric in the space of neural network architectures which can be computed efficiently via an optimal transport program. This distance might be of independent interest to the deep learning community as it may find applications outside of BO. We demonstrate that NASBOT outperforms other alternatives for architecture search in several cross validation based model selection tasks on multi-layer perceptrons and convolutional neural networks."
                },
                {
                    "title": "MobileNetV2: Inverted Residuals and Linear Bottlenecks",
                    "abstract": "In this paper we describe a new mobile architecture, MobileNetV2, that improves the state of the art performance of mobile models on multiple tasks and benchmarks as well as across a spectrum of different model sizes. We also describe efficient ways of applying these mobile models to object detection in a novel framework we call SSDLite. Additionally, we demonstrate how to build mobile semantic segmentation models through a reduced form of DeepLabv3 which we call Mobile DeepLabv3. is based on an inverted residual structure where the shortcut connections are between the thin bottleneck layers. The intermediate expansion layer uses lightweight depthwise convolutions to filter features as a source of non-linearity. Additionally, we find that it is important to remove non-linearities in the narrow layers in order to maintain representational power. We demonstrate that this improves performance and provide an intuition that led to this design. Finally, our approach allows decoupling of the input/output domains from the expressiveness of the transformation, which provides a convenient framework for further analysis. We measure our performance on ImageNet [1] classification, COCO object detection [2], VOC image segmentation [3]. We evaluate the trade-offs between accuracy, and number of operations measured by multiply-adds (MAdd), as well as actual latency, and the number of parameters."
                },
                {
                    "title": "MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks on Corrupted Labels",
                    "abstract": "Recent deep networks are capable of memorizing the entire data even when the labels are completely random. To overcome the overfitting on corrupted labels, we propose a novel technique of learning another neural network, called MentorNet, to supervise the training of the base deep networks, namely, StudentNet. During training, MentorNet provides a curriculum (sample weighting scheme) for StudentNet to focus on the sample the label of which is probably correct. Unlike the existing curriculum that is usually predefined by human experts, MentorNet learns a data-driven curriculum dynamically with StudentNet. Experimental results demonstrate that our approach can significantly improve the generalization performance of deep networks trained on corrupted training data. Notably, to the best of our knowledge, we achieve the best-published result on WebVision, a large benchmark containing 2.2 million images of real-world noisy labels. The code are at this https URL"
                },
                {
                    "title": "Squeeze-and-Excitation Networks",
                    "abstract": "Convolutional neural networks are built upon the convolution operation, which extracts informative features by fusing spatial and channel-wise information together within local receptive fields. In order to boost the representational power of a network, several recent approaches have shown the benefit of enhancing spatial encoding. In this work, we focus on the channel relationship and propose a novel architectural unit, which we term the \"Squeeze-and-Excitation\" (SE) block, that adaptively recalibrates channel-wise feature responses by explicitly modelling interdependencies between channels. We demonstrate that by stacking these blocks together, we can construct SENet architectures that generalise extremely well across challenging datasets. Crucially, we find that SE blocks produce significant performance improvements for existing state-of-the-art deep architectures at minimal additional computational cost. SENets formed the foundation of our ILSVRC 2017 classification submission which won first place and significantly reduced the top-5 error to 2.251%, achieving a ~25% relative improvement over the winning entry of 2016. Code and models are available at https://github.com/hujie-frank/SENet."
                },
                {
                    "title": "Focal Loss for Dense Object Detection",
                    "abstract": "The highest accuracy object detectors to date are based on a two-stage approach popularized by R-CNN, where a classifier is applied to a sparse set of candidate object locations. In contrast, one-stage detectors that are applied over a regular, dense sampling of possible object locations have the potential to be faster and simpler, but have trailed the accuracy of two-stage detectors thus far. In this paper, we investigate why this is the case. We discover that the extreme foreground-background class imbalance encountered during training of dense detectors is the central cause. We propose to address this class imbalance by reshaping the standard cross entropy loss such that it down-weights the loss assigned to well-classified examples. Our novel Focal Loss focuses training on a sparse set of hard examples and prevents the vast number of easy negatives from overwhelming the detector during training. To evaluate the effectiveness of our loss, we design and train a simple dense detector we call RetinaNet. Our results show that when trained with the focal loss, RetinaNet is able to match the speed of previous one-stage detectors while surpassing the accuracy of all existing state-of-the-art two-stage detectors."
                },
                {
                    "title": "Self-Paced Co-training",
                    "abstract": "Co-training is a well-known semi-supervised learning approach which trains classi\ufb01ers on two different views and exchanges labels of unlabeled instances in an iterative way. During co-training process, labels of unlabeled instances in the training pool are very likely to be false especially in the initial training rounds, while the standard co-training algorithm utilizes a \u201cdraw without replacement\u201d manner and does not remove these false labeled instances from training. This issue not only tends to degenerate its performance but also hampers its fundamental theory. Besides, there is no optimization model to explain what objective a co-training process optimizes. To these issues, in this study we design a new co-training algorithm named self-paced co-training (SPaCo) with a \u201cdraw with replacement\u201d learning mode. The rationality of SPaCo can be proved under theoretical assumptions utilized in traditional co-training research, and furthermore, the algorithm exactly complies with the alternative optimization process for an optimization model of self-paced curriculum learning, which can be \ufb01nely explained in robust learning manner. Experimental results substantiate the superiority of the proposed method as compared with current state-of-the-art co-training methods."
                },
                {
                    "title": "Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour",
                    "abstract": "Deep learning thrives with large neural networks and large datasets. However, larger networks and larger datasets result in longer training times that impede research and development progress. Distributed synchronous SGD offers a potential solution to this problem by dividing SGD minibatches over a pool of parallel workers. Yet to make this scheme efficient, the per-worker workload must be large, which implies nontrivial growth in the SGD minibatch size. In this paper, we empirically show that on the ImageNet dataset large minibatches cause optimization difficulties, but when these are addressed the trained networks exhibit good generalization. Specifically, we show no loss of accuracy when training with large minibatch sizes up to 8192 images. To achieve this result, we adopt a hyper-parameter-free linear scaling rule for adjusting learning rates as a function of minibatch size and develop a new warmup scheme that overcomes optimization challenges early in training. With these simple techniques, our Caffe2-based system trains ResNet-50 with a minibatch size of 8192 on 256 GPUs in one hour, while matching small minibatch accuracy. Using commodity hardware, our implementation achieves ~90% scaling efficiency when moving from 8 to 256 GPUs. Our findings enable training visual recognition models on internet-scale data with high efficiency."
                },
                {
                    "title": "Automated Curriculum Learning for Neural Networks",
                    "abstract": "We introduce a method for automatically selecting the path, or syllabus, that a neural network follows through a curriculum so as to maximise learning efficiency. A measure of the amount that the network learns from each data sample is provided as a reward signal to a nonstationary multi-armed bandit algorithm, which then determines a stochastic syllabus. We consider a range of signals derived from two distinct indicators of learning progress: rate of increase in prediction accuracy, and rate of increase in network complexity. Experimental results for LSTM networks on three curricula demonstrate that our approach can significantly accelerate learning, in some cases halving the time required to attain a satisfactory performance level."
                },
                {
                    "title": "Learning What Data to Learn",
                    "abstract": "Machine learning is essentially the sciences of playing with data. An adaptive data selection strategy, enabling to dynamically choose different data at various training stages, can reach a more effective model in a more efficient way. In this paper, we propose a deep reinforcement learning framework, which we call \\emph{\\textbf{N}eural \\textbf{D}ata \\textbf{F}ilter} (\\textbf{NDF}), to explore automatic and adaptive data selection in the training process. In particular, NDF takes advantage of a deep neural network to adaptively select and filter important data instances from a sequential stream of training data, such that the future accumulative reward (e.g., the convergence speed) is maximized. In contrast to previous studies in data selection that is mainly based on heuristic strategies, NDF is quite generic and thus can be widely suitable for many machine learning tasks. Taking neural network training with stochastic gradient descent (SGD) as an example, comprehensive experiments with respect to various neural network modeling (e.g., multi-layer perceptron networks, convolutional neural networks and recurrent neural networks) and several applications (e.g., image classification and text understanding) demonstrate that NDF powered SGD can achieve comparable accuracy with standard SGD process by using less data and fewer iterations."
                },
                {
                    "title": "Overcoming catastrophic forgetting in neural networks",
                    "abstract": "Significance Deep neural networks are currently the most successful machine-learning technique for solving a variety of tasks, including language translation, image classification, and image generation. One weakness of such models is that, unlike humans, they are unable to learn multiple tasks sequentially. In this work we propose a practical solution to train such models sequentially by protecting the weights important for previous tasks. This approach, inspired by synaptic consolidation in neuroscience, enables state of the art results on multiple reinforcement learning problems experienced sequentially. The ability to learn tasks in a sequential fashion is crucial to the development of artificial intelligence. Until now neural networks have not been capable of this and it has been widely thought that catastrophic forgetting is an inevitable feature of connectionist models. We show that it is possible to overcome this limitation and train networks that can maintain expertise on tasks that they have not experienced for a long time. Our approach remembers old tasks by selectively slowing down learning on the weights important for those tasks. We demonstrate our approach is scalable and effective by solving a set of classification tasks based on a hand-written digit dataset and by learning several Atari 2600 games sequentially."
                },
                {
                    "title": "Fast Bayesian Optimization of Machine Learning Hyperparameters on Large Datasets",
                    "abstract": "Bayesian optimization has become a successful tool for hyperparameter optimization of machine learning algorithms, such as support vector machines or deep neural networks. Despite its success, for large datasets, training and validating a single configuration often takes hours, days, or even weeks, which limits the achievable performance. To accelerate hyperparameter optimization, we propose a generative model for the validation error as a function of training set size, which is learned during the optimization process and allows exploration of preliminary configurations on small subsets, by extrapolating to the full dataset. We construct a Bayesian optimization procedure, dubbed Fabolas, which models loss and training time as a function of dataset size and automatically trades off high information gain about the global optimum against computational cost. Experiments optimizing support vector machines and deep neural networks show that Fabolas often finds high-quality solutions 10 to 100 times faster than other state-of-the-art Bayesian optimization methods or the recently proposed bandit strategy Hyperband."
                },
                {
                    "title": "Frontal to profile face verification in the wild",
                    "abstract": "We have collected a new face data set that will facilitate research in the problem of frontal to profile face verification `in the wild'. The aim of this data set is to isolate the factor of pose variation in terms of extreme poses like profile, where many features are occluded, along with other `in the wild' variations. We call this data set the Celebrities in Frontal-Profile (CFP) data set. We find that human performance on Frontal-Profile verification in this data set is only slightly worse (94.57% accuracy) than that on Frontal-Frontal verification (96.24% accuracy). However we evaluated many state-of-the-art algorithms, including Fisher Vector, Sub-SML and a Deep learning algorithm. We observe that all of them degrade more than 10% from Frontal-Frontal to Frontal-Profile verification. The Deep learning implementation, which performs comparable to humans on Frontal-Frontal, performs significantly worse (84.91% accuracy) on Frontal-Profile. This suggests that there is a gap between human performance and automatic face recognition methods for large pose variation in unconstrained images."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Self-Paced Curriculum Learning",
                    "abstract": "\n \n Curriculum learning (CL) or self-paced learning (SPL) represents a recently proposed learning regime inspired by the learning process of humans and animals that gradually proceeds from easy to more complex samples in training. The two methods share a similar conceptual learning paradigm, but differ in specific learning schemes. In CL, the curriculum is predetermined by prior knowledge, and remain fixed thereafter. Therefore, this type of method heavily relies on the quality of prior knowledge while ignoring feedback about the learner. In SPL, the curriculum is dynamically determined to adjust to the learning pace of the leaner. However, SPL is unable to deal with prior knowledge, rendering it prone to overfitting. In this paper, we discover the missing link between CL and SPL, and propose a unified framework named self-paced curriculum leaning (SPCL). SPCL is formulated as a concise optimization problem that takes into account both prior knowledge known before training and the learning progress during training. In comparison to human education, SPCL is analogous to \"instructor-student-collaborative\" learning mode, as opposed to \"instructor-driven\" in CL or \"student-driven\" in SPL. Empirically, we show that the advantage of SPCL on two tasks.\n \n"
                },
                {
                    "title": "Training Deep Neural Networks on Noisy Labels with Bootstrapping",
                    "abstract": "Current state-of-the-art deep learning systems for visual object recognition and detection use purely supervised training with regularization such as dropout to avoid overfitting. The performance depends critically on the amount of labeled examples, and in current practice the labels are assumed to be unambiguous and accurate. However, this assumption often does not hold; e.g. in recognition, class labels may be missing; in detection, objects in the image may not be localized; and in general, the labeling may be subjective. In this work we propose a generic way to handle noisy and incomplete labeling by augmenting the prediction objective with a notion of consistency. We consider a prediction consistent if the same prediction is made given similar percepts, where the notion of similarity is between deep network features computed from the input data. In experiments we demonstrate that our approach yields substantial robustness to label noise on several datasets. On MNIST handwritten digits, we show that our model is robust to label corruption. On the Toronto Face Database, we show that our model handles well the case of subjective labels in emotion recognition, achieving state-of-the- art results, and can also benefit from unlabeled face images with no modification to our method. On the ILSVRC2014 detection challenge data, we show that our approach extends to very deep networks, high resolution images and structured outputs, and results in improved scalable detection."
                },
                {
                    "title": "Practical Bayesian Optimization of Machine Learning Algorithms",
                    "abstract": "The use of machine learning algorithms frequently involves careful tuning of learning parameters and model hyperparameters. Unfortunately, this tuning is often a \"black art\" requiring expert experience, rules of thumb, or sometimes brute-force search. There is therefore great appeal for automatic approaches that can optimize the performance of any given learning algorithm to the problem at hand. In this work, we consider this problem through the framework of Bayesian optimization, in which a learning algorithm's generalization performance is modeled as a sample from a Gaussian process (GP). We show that certain choices for the nature of the GP, such as the type of kernel and the treatment of its hyperparameters, can play a crucial role in obtaining a good optimizer that can achieve expertlevel performance. We describe new algorithms that take into account the variable cost (duration) of learning algorithm experiments and that can leverage the presence of multiple cores for parallel experimentation. We show that these proposed algorithms improve on previous automatic procedures and can reach or surpass human expert-level optimization for many algorithms including latent Dirichlet allocation, structured SVMs and convolutional neural networks."
                },
                {
                    "title": "Local neighbourhood extension of SMOTE for mining imbalanced data",
                    "abstract": "In this paper we discuss problems of inducing classifiers from imbalanced data and improving recognition of minority class using focused resampling techniques. We are particularly interested in SMOTE over-sampling method that generates new synthetic examples from the minority class between the closest neighbours from this class. However, SMOTE could also overgeneralize the minority class region as it does not consider distribution of other neighbours from the majority classes. Therefore, we introduce a new generalization of SMOTE, called LN-SMOTE, which exploits more precisely information about the local neighbourhood of the considered examples. In the experiments we compare this method with original SMOTE and its two, the most related, other generalizations Borderline and Safe-Level SMOTE. All these pre-processing methods are applied together with either decision tree or Naive Bayes classifiers. The results show that the new LN-SMOTE method improves evaluation measures for the minority class."
                },
                {
                    "title": "A Tutorial on Bayesian Optimization of Expensive Cost Functions, with Application to Active User Modeling and Hierarchical Reinforcement Learning",
                    "abstract": "We present a tutorial on Bayesian optimization, a method of finding the maximum of expensive cost functions. Bayesian optimization employs the Bayesian technique of setting a prior over the objective function and combining it with evidence to get a posterior function. This permits a utility-based selection of the next observation to make on the objective function, which must take into account both exploration (sampling from areas of high uncertainty) and exploitation (sampling areas likely to offer improvement over the current best observation). We also present two detailed extensions of Bayesian optimization, with experiments---active user modelling with preferences, and hierarchical reinforcement learning---and a discussion of the pros and cons of Bayesian optimization based on our experiences."
                },
                {
                    "title": "Self-Paced Learning for Latent Variable Models",
                    "abstract": "Latent variable models are a powerful tool for addressing several tasks in machine learning. However, the algorithms for learning the parameters of latent variable models are prone to getting stuck in a bad local optimum. To alleviate this problem, we build on the intuition that, rather than considering all samples simultaneously, the algorithm should be presented with the training data in a meaningful order that facilitates learning. The order of the samples is determined by how easy they are. The main challenge is that often we are not provided with a readily computable measure of the easiness of samples. We address this issue by proposing a novel, iterative self-paced learning algorithm where each iteration simultaneously selects easy samples and learns a new parameter vector. The number of samples selected is governed by a weight that is annealed until the entire training data has been considered. We empirically demonstrate that the self-paced learning algorithm outperforms the state of the art method for learning a latent structural SVM on four applications: object localization, noun phrase coreference, motif finding and handwritten digit recognition."
                },
                {
                    "title": "Research Design and Statistical Analysis",
                    "abstract": "Preliminary Considerations Samples and Populations Some Important Distributions Between-Subjects Designs - One Factor Between-Subjects Designs - Several Factors Contrasts Among Means Trend Analysis Repeated-Measures Designs Mixed Designs - Combining Between-Subject and Within-Subjects Factors Hierarchical Designs Latin Squares and Related Designs Bivariate Correlation and Regression Analysis of Covariance More About Correlation Multiple Regression Regression with Categorical Variables Appendices - Notation and Summation Operations, Expected Values and Their Applications, Matrix Algebra, Statistical Tables, Control Information for Computer Programmes for Statistical Analysis."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "Learning From Imbalanced Data",
                    "abstract": "A very challenging issue in real-world data is that in many domains like medicine, finance, marketing, web, telecommunication, management, etc. the distribution of data among classes is inherently imbalanced. A widely accepted researched issue is that the traditional classifier algorithms assume a balanced distribution among the classes. Data imbalance is evident when the number of instances representing the class of concern is much lesser than other classes. Hence, the classifiers tend to bias towards the well-represented class. This leads to a higher misclassification rate among the lesser represented class. Hence, there is a need of efficient learners to classify imbalanced data. This chapter aims to address the need, challenges, existing methods, and evaluation metrics identified when learning from imbalanced data sets. Future research challenges and directions are highlighted."
                },
                {
                    "title": "Training Deep Models Faster with Robust, Approximate Importance Sampling",
                    "abstract": "In theory, importance sampling speeds up stochastic gradient algorithms for supervised learning by prioritizing training examples. In practice, the cost of computing importances greatly limits the impact of importance sampling. We propose a robust, approximate importance sampling procedure (RAIS) for stochastic gradient de- scent. By approximating the ideal sampling distribution using robust optimization, RAIS provides much of the benefit of exact importance sampling with drastically reduced overhead. Empirically, we find RAIS-SGD and standard SGD follow similar learning curves, but RAIS moves faster through these paths, achieving speed-ups of at least 20% and sometimes much more."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop an efficient and effective automatic sampling strategy for training data in deep learning that adapts to various tasks and datasets without requiring extensive computational resources or clean meta datasets?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the fundamental issue of data sampling in deep learning, which significantly impacts model performance. An effective automatic sampling strategy could lead to more robust models that generalize better across different tasks and datasets. This advancement could streamline the training process, reduce computational costs, and enable researchers to focus on model architecture and other critical aspects of machine learning. Furthermore, it could open new avenues for practical applications in fields such as computer vision, natural language processing, and beyond, where diverse datasets are prevalent.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the high dimensionality of the sampling space, where the complexity of finding the optimal sampler increases exponentially with the number of training instances. Additionally, the sharpness of the objective function for samplers complicates the learning process, as small changes in sampling probabilities can lead to significant variations in model performance. Naive approaches may fail because they do not account for the intricate relationships between sampling probabilities and model outcomes, leading to inefficient exploration of the sampling space and suboptimal performance.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either human-defined rules or learning-based methods, both of which have limitations. Human-defined rules are often task-specific and lack adaptability, while learning-based methods require extensive computational resources and clean meta datasets that are not always available. These barriers have hindered the development of a more generalized and efficient automatic sampling strategy. Our approach differs by focusing on sample-based search methods that do not require additional meta data and can be applied without altering the training process, thus addressing the limitations of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using a sample-based search method, leveraging agents such as Bayesian Optimization or Deep Reinforcement Learning to iteratively sample and evaluate different sampling strategies. We will utilize a diverse dataset, such as ImageNet, to assess the effectiveness of our approach. The performance will be measured using metrics such as model accuracy and training efficiency. We expect our results to demonstrate that our automatic sampling strategy significantly improves model performance while reducing computational costs"
            }
        },
        "author_data": {
            "ddc7fcc4-57e3-4a11-9213-ba527626a483": {
                "pk": "ddc7fcc4-57e3-4a11-9213-ba527626a483",
                "name": "Jiawei Yao",
                "collaborators": [
                    "Juhua Hu",
                    "Tong Wu",
                    "Qi Qian",
                    "Xiaofeng Zhang",
                    "Jianqing Fan",
                    "Lingzhou Xue",
                    "Hongrui Kou",
                    "Xiaochao Pan",
                    "Jusheng Zhang",
                    "Enbei Liu"
                ],
                "domain": [
                    "Clustering",
                    "Deep Learning",
                    "Computer Vision",
                    "Time Series Analysis"
                ],
                "publications": [
                    {
                        "title": "Dual-disentangled Deep Multiple Clustering",
                        "abstract": "Multiple clustering has gathered significant attention in recent years due to its potential to reveal multiple hidden structures of the data from different perspectives. Most of multiple clustering methods first derive feature representations by controlling the dissimilarity among them, subsequently employing traditional clustering methods (e.g., k-means) to achieve the final multiple clustering outcomes. However, the learned feature representations can exhibit a weak relevance to the ultimate goal of distinct clustering. Moreover, these features are often not explicitly learned for the purpose of clustering. Therefore, in this paper, we propose a novel Dual-Disentangled deep Multiple Clustering method named DDMC by learning disentangled representations. Specifically, DDMC is achieved by a variational Expectation-Maximization (EM) framework. In the E-step, the disentanglement learning module employs coarse-grained and fine-grained disentangled representations to obtain a more diverse set of latent factors from the data. In the M-step, the cluster assignment module utilizes a cluster objective function to augment the effectiveness of the cluster output. Our extensive experiments demonstrate that DDMC consistently outperforms state-of-the-art methods across seven commonly used tasks. Our code is available at https://github.com/Alexander-Yao/DDMC."
                    },
                    {
                        "title": "DepthSSC: Depth-Spatial Alignment and Dynamic Voxel Resolution for Monocular 3D Semantic Scene Completion",
                        "abstract": "The task of 3D semantic scene completion with monocular cameras is gaining increasing attention in the field of autonomous driving. Its objective is to predict the occupancy status of each voxel in the 3D scene from partial image inputs. Despite the existence of numerous methods, many of them overlook the issue of accurate alignment between spatial and depth information. To address this, we propose DepthSSC, an advanced method for semantic scene completion solely based on monocular cameras. DepthSSC combines the ST-GF (Spatial Transformation Graph Fusion) module with geometric-aware voxelization, enabling dynamic adjustment of voxel resolution and considering the geometric complexity of 3D space to ensure precise alignment between spatial and depth information. This approach successfully mitigates spatial misalignment and distortion issues observed in prior methods. Through evaluation on the SemanticKITTI dataset, DepthSSC not only demonstrates its effectiveness in capturing intricate 3D structural details but also achieves state-of-the-art performance. We believe DepthSSC provides a fresh perspective on monocular camera-based 3D semantic scene completion research and anticipate it will inspire further related studies."
                    },
                    {
                        "title": "Multi-Modal Proxy Learning Towards Personalized Visual Multiple Clustering",
                        "abstract": "Multiple clustering has gained significant attention in recent years due to its potential to reveal multiple hidden structures of data from different perspectives. The advent of deep multiple clustering techniques has notably advanced the performance by uncovering complex patterns and relationships within large datasets. However, a major challenge arises as users often do not need all the clusterings that algorithms generate, and figuring out the one needed requires a substantial understanding of each clustering result. Traditionally, aligning a user's brief keyword of interest with the corresponding vision components was challenging, but the emergence of multi-modal and large language models (LLMs) has begun to bridge this gap. In response, given unlabeled target visual data, we propose Multi-MaP, a novel method employing a multi-modal proxy learning process. It leverages CLIP encoders to extract coherent text and image embeddings, with GPT-4 integrating users' interests to formulate effective textual contexts. Moreover, reference word constraint and concept-level constraint are designed to learn the optimal text proxy according to the user's interest. Multi-MaP not only adeptly captures a user's interest via a keyword but also facilitates identifying relevant clusterings. Our extensive experiments show that Multi-MaP consistently outperforms state-of-the-art methods in all benchmark multi-clustering vision tasks. Our code is available at https://github.com/Alexander-Yao/Multi-MaP."
                    },
                    {
                        "title": "AugDMC: Data Augmentation Guided Deep Multiple Clustering",
                        "abstract": "Clustering aims to group similar objects together while separating dissimilar ones apart. Thereafter, structures hidden in data can be identified to help understand data in an unsupervised manner. Traditional clustering methods such as k-means provide only a single clustering for one data set. Deep clustering methods such as auto-encoder based clustering methods have shown a better performance, but still provide a single clustering. However, a given dataset might have multiple clustering structures and each represents a unique perspective of the data. Therefore, some multiple clustering methods have been developed to discover multiple independent structures hidden in data. Although deep multiple clustering methods provide better performance, how to efficiently capture the alternative perspectives in data is still a problem. In this paper, we propose AugDMC, a novel data Augmentation guided Deep Multiple Clustering method, to tackle the challenge. Specifically, AugDMC leverages data augmentations to automatically extract features related to a certain aspect of the data using a self-supervised prototype-based representation learning, where different aspects of the data can be preserved under different data augmentations. Moreover, a stable optimization strategy is proposed to alleviate the unstable problem from different augmentations. Thereafter, multiple clusterings based on different aspects of the data can be obtained. Experimental results on three real-world datasets compared with state-of-the-art methods validate the effectiveness of the proposed method."
                    },
                    {
                        "title": "Improving Depth Gradient Continuity in Transformers: A Comparative Study on Monocular Depth Estimation with CNN",
                        "abstract": "Monocular depth estimation is an ongoing challenge in computer vision. Recent progress with Transformer models has demonstrated notable advantages over conventional CNNs in this area. However, there's still a gap in understanding how these models prioritize different regions in 2D images and how these regions affect depth estimation performance. To explore the differences between Transformers and CNNs, we employ a sparse pixel approach to contrastively analyze the distinctions between the two. Our findings suggest that while Transformers excel in handling global context and intricate textures, they lag behind CNNs in preserving depth gradient continuity. To further enhance the performance of Transformer models in monocular depth estimation, we propose the Depth Gradient Refinement (DGR) module that refines depth estimation through high-order differentiation, feature fusion, and recalibration. Additionally, we leverage optimal transport theory, treating depth maps as spatial probability distributions, and employ the optimal transport distance as a loss function to optimize our model. Experimental results demonstrate that models integrated with the plug-and-play Depth Gradient Refinement (DGR) module and the proposed loss function enhance performance without increasing complexity and computational costs on both outdoor KITTI and indoor NYU-Depth-v2 datasets. This research not only offers fresh insights into the distinctions between Transformers and CNNs in depth estimation but also paves the way for novel depth estimation methodologies."
                    },
                    {
                        "title": "Power Enhancement in High Dimensional Cross-Sectional Tests",
                        "abstract": "We propose a novel technique to boost the power of testing a high-dimensional vector $H:\\btheta=0$ against sparse alternatives where the null hypothesis is violated only by a couple of components. Existing tests based on quadratic forms such as the Wald statistic often suffer from low powers due to the accumulation of errors in estimating high-dimensional parameters. More powerful tests for sparse alternatives such as thresholding and extreme-value tests, on the other hand, require either stringent conditions or bootstrap to derive the null distribution and often suffer from size distortions due to the slow convergence. Based on a screening technique, we introduce a \"power enhancement component\", which is zero under the null hypothesis with high probability, but diverges quickly under sparse alternatives. The proposed test statistic combines the power enhancement component with an asymptotically pivotal statistic, and strengthens the power under sparse alternatives. The null distribution does not require stringent regularity conditions, and is completely determined by that of the pivotal statistic. As a byproduct, the power enhancement component also consistently identifies the elements that violate the null hypothesis. As specific applications, the proposed methods are applied to testing the factor pricing models and validating the cross-sectional independence in panel data models."
                    },
                    {
                        "title": "Sufficient Forecasting Using Factor Models",
                        "abstract": "We consider forecasting a single time series when there is a large number of predictors and a possible nonlinear effect. The dimensionality was first reduced via a high-dimensional (approximate) factor model implemented by the principal component analysis. Using the extracted factors, we develop a novel forecasting method called the sufficient forecasting, which provides a set of sufficient predictive indices, inferred from high-dimensional predictors, to deliver additional predictive power. The projected principal component analysis will be employed to enhance the accuracy of inferred factors when a semi-parametric (approximate) factor model is assumed. Our method is also applicable to cross-sectional sufficient regression using extracted factors. The connection between the sufficient forecasting and the deep learning architecture is explicitly stated. The sufficient forecasting correctly estimates projection indices of the underlying factors even in the presence of a nonparametric forecasting function. The proposed method extends the sufficient dimension reduction to high-dimensional regimes by condensing the cross-sectional information through factor models. We derive asymptotic properties for the estimate of the central subspace spanned by these projection directions as well as the estimates of the sufficient predictive indices. We further show that the natural method of running multiple regression of target on estimated factors yields a linear estimate that actually falls into this central subspace. Our method and theory allow the number of predictors to be larger than the number of observations. We finally demonstrate that the sufficient forecasting improves upon the linear forecasting in both simulation studies and an empirical study of forecasting macroeconomic variables."
                    },
                    {
                        "title": "QE-BEV: Query Evolution for Bird's Eye View Object Detection in Varied Contexts",
                        "abstract": "3D object detection plays a pivotal role in autonomous driving and robotics, demanding precise interpretation of Bird's Eye View (BEV) images. The dynamic nature of real-world environments necessitates the use of dynamic query mechanisms in 3D object detection to adaptively capture and process the complex spatio-temporal relationships present in these scenes. However, prior implementations of dynamic queries have often faced difficulties in effectively leveraging these relationships, particularly when it comes to integrating temporal information in a computationally efficient manner. Addressing this limitation, we introduce a framework utilizing dynamic query evolution strategy, harnesses K-means clustering and Top-K attention mechanisms for refined spatio-temporal data processing. By dynamically segmenting the BEV space and prioritizing key features through Top-K attention, our model achieves a real-time, focused analysis of pertinent scene elements. Our extensive evaluation on the nuScenes and Waymo dataset showcases a marked improvement in detection accuracy, setting a new benchmark in the domain of query-based BEV object detection. Our dynamic query evolution strategy has the potential to push the boundaries of current BEV methods with enhanced adaptability and computational efficiency. Project page: https://github.com/Jiawei-Yao0812/QE-BEV"
                    },
                    {
                        "title": "Building Lane-Level Maps from Aerial Images",
                        "abstract": "Detecting lane lines from sensors is becoming an increasingly significant part of autonomous driving systems. However, less development has been made on high-definition lane-level mapping based on aerial images, which could automatically build and update offline maps for auto-driving systems. To this end, our work focuses on extracting fine-level detailed lane lines together with their topological structures. This task is challenging since it requires large amounts of data covering different lane types, terrain and regions. In this paper, we introduce for the first time a large-scale aerial image dataset built for lane detection, with high-quality polyline lane annotations on high-resolution images of around 80 kilometers of road. Moreover, we developed a baseline deep learning lane detection method from aerial images, called AerialLaneNet, consisting of two stages. The first stage is to produce coarse-grained results at point level, and the second stage exploits the coarse-grained results and feature to perform the vertex-matching task, producing fine-grained lanes with topology. The experiments show our approach achieves significant improvement compared with the state-of-the-art methods on our new dataset. Our code and new dataset are available at https://github.com/Jiawei-Yao0812/AerialLaneNet."
                    },
                    {
                        "title": "Text-Guided Mixup Towards Long-Tailed Image Categorization",
                        "abstract": "In many real-world applications, the frequency distribution of class labels for training data can exhibit a long-tailed distribution, which challenges traditional approaches of training deep neural networks that require heavy amounts of balanced data. Gathering and labeling data to balance out the class label distribution can be both costly and time-consuming. Many existing solutions that enable ensemble learning, re-balancing strategies, or fine-tuning applied to deep neural networks are limited by the inert problem of few class samples across a subset of classes. Recently, vision-language models like CLIP have been observed as effective solutions to zero-shot or few-shot learning by grasping a similarity between vision and language features for image and text pairs. Considering that large pre-trained vision-language models may contain valuable side textual information for minor classes, we propose to leverage text supervision to tackle the challenge of long-tailed learning. Concretely, we propose a novel text-guided mixup technique that takes advantage of the semantic relations between classes recognized by the pre-trained text encoder to help alleviate the long-tailed problem. Our empirical study on benchmark long-tailed tasks demonstrates the effectiveness of our proposal with a theoretical guarantee. Our code is available at https://github.com/rsamf/text-guided-mixup."
                    },
                    {
                        "title": "HarmonicNeRF: Geometry-Informed Synthetic View Augmentation for 3D Scene Reconstruction in Driving Scenarios",
                        "abstract": "In the realm of autonomous driving, achieving precise 3D reconstruction of the driving environment is critical for ensuring safety and effective navigation. Neural Radiance Fields (NeRF) have shown promise in creating highly detailed and accurate models of complex environments. However, the application of NeRF in autonomous driving scenarios encounters several challenges, primarily due to the sparsity of viewpoints inherent in camera trajectories and the constraints on data collection in unbounded outdoor scenes, which typically occur along predetermined paths. This limitation not only reduces the available scene information but also poses significant challenges for NeRF training, as the sparse and path-distributed observational data leads to under-representation of the scene's geometry. In this paper, we introduce HarmonicNeRF, a novel approach for outdoor self-supervised monocular scene reconstruction. HarmonicNeRF capitalizes on the strengths of NeRF and enhances surface reconstruction accuracy by augmenting the input space with geometry-informed synthetic views. This is achieved through the application of spherical harmonics to generate novel radiance values, taking into careful consideration the color observations from the limited available real-world views. Additionally, our method incorporates proxy geometry to effectively manage occlusion, generating radiance pseudo-labels that circumvent the limitations of traditional image-warping techniques, which often fail in sparse data conditions typical of autonomous driving environments. Extensive experiments conducted on the KITTI, Argoverse, and NuScenes datasets demonstrate our approach establishes new benchmarks in synthesizing novel depth views and reconstructing scenes, significantly outperforming existing methods. Project page: https://github.com/Jiawei-Yao0812/HarmonicNeRF"
                    },
                    {
                        "title": "Inverse Moment Methods for Sufficient Forecasting using High-Dimensional Predictors",
                        "abstract": "We consider forecasting a single time series using a large number of predictors in the presence of a possible nonlinear forecast function. Assuming that the predictors affect the response through the latent factors, we propose to first conduct factor analysis and then apply sufficient dimension reduction on the estimated factors, to derive the reduced data for subsequent forecasting. Using directional regression and the inverse third-moment method in the stage of sufficient dimension reduction, the proposed methods can capture the non-monotone effect of factors on the response. We also allow a diverging number of factors and only impose general regularity conditions on the distribution of factors, avoiding the undesired time reversibility of the factors by the latter. These make the proposed methods fundamentally more applicable than the sufficient forecasting method in Fan et al. (2017). The proposed methods are demonstrated in both simulation studies and an empirical study of forecasting monthly macroeconomic data from 1959 to 2016. Also, our theory contributes to the literature of sufficient dimension reduction, as it includes an invariance result, a path to perform sufficient dimension reduction under the high-dimensional setting without assuming sparsity, and the corresponding order-determination procedure."
                    }
                ]
            },
            "cbda1c16-0f23-41a2-a584-e30df4ac465f": {
                "pk": "cbda1c16-0f23-41a2-a584-e30df4ac465f",
                "name": "Chuming Li",
                "collaborators": [
                    "Wanli Ouyang",
                    "Chen Lin",
                    "Junjie Yan",
                    "Ming Sun",
                    "Jie Liu",
                    "Wei Wu",
                    "Yinmin Zhang",
                    "Yaodong Yang",
                    "Yu Liu",
                    "Boyu Chen"
                ],
                "domain": [
                    "Neural Architecture Search",
                    "Reinforcement Learning",
                    "Computer Vision",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Improving One-shot NAS by Suppressing the Posterior Fading",
                        "abstract": "There is a growing interest in automated neural architecture search (NAS). To improve the efficiency of NAS, previous approaches adopt weight sharing method to force all models share the same set of weights. However, it has been observed that a model performing better with shared weights does not necessarily perform better when trained alone. In this paper, we analyse existing weight sharing one-shot NAS approaches from a Bayesian point of view and identify the posterior fading problem, which compromises the effectiveness of shared weights. To alleviate this problem, we present a practical approach to guide the parameter posterior towards its true distribution. Moreover, a hard latency constraint is introduced during the search so that the desired latency can be achieved. The resulted method, namely Posterior Convergent NAS (PC-NAS), achieves state-of-the-art performance under standard GPU latency constraint on ImageNet. In our small search space, our model PC-NAS-S attains 76.8 % top-1 accuracy, 2.1% higher than MobileNetV2 (1.4x) with the same latency. When adopted to the large search space, PC-NAS-L achieves 78.1 % top-1 accuracy within 11ms. The discovered architecture also transfers well to other computer vision applications such as object detection and person re-identification."
                    },
                    {
                        "title": "NDC-Scene: Boost Monocular 3D Semantic Scene Completion in Normalized Device Coordinates Space",
                        "abstract": "Monocular 3D Semantic Scene Completion (SSC) has garnered significant attention in recent years due to its potential to predict complex semantics and geometry shapes from a single image, requiring no 3D inputs. In this paper, we identify several critical issues in current state-of-the-art methods, including the Feature Ambiguity of projected 2D features in the ray to the 3D space, the Pose Ambiguity of the 3D convolution, and the Computation Imbalance in the 3D convolution across different depth levels. To address these problems, we devise a novel Normalized Device Coordinates scene completion network (NDC-Scene) that directly extends the 2D feature map to a Normalized Device Coordinates (NDC) space, rather than to the world space directly, through progressive restoration of the dimension of depth with deconvolution operations. Experiment results demonstrate that transferring the majority of computation from the target 3D space to the proposed normalized device coordinates space benefits monocular SSC tasks. Additionally, we design a Depth-Adaptive Dual Decoder to simultaneously upsample and fuse the 2D and 3D feature maps, further improving overall performance. Our extensive experiments confirm that the proposed method consistently outperforms state-of-the-art methods on both outdoor SemanticKITTI and indoor NYUv2 datasets. Our code are available at https://github.com/Jiawei-Yao0812/NDCScene."
                    },
                    {
                        "title": "A Perspective of Q-value Estimation on Offline-to-Online Reinforcement Learning",
                        "abstract": "Offline-to-online Reinforcement Learning (O2O RL) aims to improve the performance of offline pretrained policy using only a few online samples. Built on offline RL algorithms, most O2O methods focus on the balance between RL objective and pessimism, or the utilization of offline and online samples. In this paper, from a novel perspective, we systematically study the challenges that remain in O2O RL and identify that the reason behind the slow improvement of the performance and the instability of online finetuning lies in the inaccurate Q-value estimation inherited from offline pretraining. Specifically, we demonstrate that the estimation bias and the inaccurate rank of Q-value cause a misleading signal for the policy update, making the standard offline RL algorithms, such as CQL and TD3-BC, ineffective in the online finetuning. Based on this observation, we address the problem of Q-value estimation by two techniques: (1) perturbed value update and (2) increased frequency of Q-value updates. The first technique smooths out biased Q-value estimation with sharp peaks, preventing early-stage policy exploitation of sub-optimal actions. The second one alleviates the estimation bias inherited from offline pretraining by accelerating learning. Extensive experiments on the MuJoco and Adroit environments demonstrate that the proposed method, named SO2, significantly alleviates Q-value estimation issues, and consistently improves the performance against the state-of-the-art methods by up to 83.1%."
                    },
                    {
                        "title": "Neural Architecture Search on Acoustic Scene Classification",
                        "abstract": "Convolutional neural networks are widely adopted in Acoustic Scene Classification (ASC) tasks, but they generally carry a heavy computational burden. In this work, we propose a lightweight yet high-performing baseline network inspired by MobileNetV2, which replaces square convolutional kernels with unidirectional ones to extract features alternately in temporal and frequency dimensions. Furthermore, we explore a dynamic architecture space built on the basis of the proposed baseline with the recent Neural Architecture Search (NAS) paradigm, which first trains a supernet that incorporates all candidate networks and then applies a well-known evolutionary algorithm NSGA-II to discover more efficient networks with higher accuracy and lower computational cost. Experimental results demonstrate that our searched network is competent in ASC tasks, which achieves 90.3% F1-score on the DCASE2018 task 5 evaluation set, marking a new state-of-the-art performance while saving 25% of FLOPs compared to our baseline network."
                    },
                    {
                        "title": "Residual Relaxation for Multi-view Representation Learning",
                        "abstract": "Multi-view methods learn representations by aligning multiple views of the same image and their performance largely depends on the choice of data augmentation. In this paper, we notice that some other useful augmentations, such as image rotation, are harmful for multi-view methods because they cause a semantic shift that is too large to be aligned well. This observation motivates us to relax the exact alignment objective to better cultivate stronger augmentations. Taking image rotation as a case study, we develop a generic approach, Pretext-aware Residual Relaxation (Prelax), that relaxes the exact alignment by allowing an adaptive residual vector between different views and encoding the semantic shift through pretext-aware learning. Extensive experiments on different backbones show that our method can not only improve multi-view methods with existing augmentations, but also benefit from stronger image augmentations like rotation."
                    },
                    {
                        "title": "AM-LFS: AutoML for Loss Function Search",
                        "abstract": "Designing an effective loss function plays an important role in visual analysis. Most existing loss function designs rely on hand-crafted heuristics that require domain experts to explore the large design space, which is usually sub-optimal and time-consuming. In this paper, we propose AutoML for Loss Function Search (AM-LFS) which leverages REINFORCE to search loss functions during the training process. The key contribution of this work is the design of search space which can guarantee the generalization and transferability on different vision tasks by including a bunch of existing prevailing loss functions in a unified formulation. We also propose an efficient optimization framework which can dynamically optimize the parameters of loss function's distribution during training. Extensive experimental results on four benchmark datasets show that, without any tricks, our method outperforms existing hand-crafted loss functions in various computer vision tasks."
                    },
                    {
                        "title": "Adaptive Gradient Method with Resilience and Momentum",
                        "abstract": "Several variants of stochastic gradient descent (SGD) have been proposed to improve the learning effectiveness and efficiency when training deep neural networks, among which some recent influential attempts would like to adaptively control the parameter-wise learning rate (e.g., Adam and RMSProp). Although they show a large improvement in convergence speed, most adaptive learning rate methods suffer from compromised generalization compared with SGD. In this paper, we proposed an Adaptive Gradient Method with Resilience and Momentum (AdaRem), motivated by the observation that the oscillations of network parameters slow the training, and give a theoretical proof of convergence. For each parameter, AdaRem adjusts the parameter-wise learning rate according to whether the direction of one parameter changes in the past is aligned with the direction of the current gradient, and thus encourages long-term consistent parameter updating with much fewer oscillations. Comprehensive experiments have been conducted to verify the effectiveness of AdaRem when training various models on a large-scale image recognition dataset, e.g., ImageNet, which also demonstrate that our method outperforms previous adaptive learning rate-based algorithms in terms of the training speed and the test error, respectively."
                    },
                    {
                        "title": "Powering One-shot Topological NAS with Stabilized Share-parameter Proxy",
                        "abstract": "One-shot NAS method has attracted much interest from the research community due to its remarkable training efficiency and capacity to discover high performance models. However, the search spaces of previous one-shot based works usually relied on hand-craft design and were short for flexibility on the network topology. In this work, we try to enhance the one-shot NAS by exploring high-performing network architectures in our large-scale Topology Augmented Search Space (i.e., over 3.4*10^10 different topological structures). Specifically, the difficulties for architecture searching in such a complex space has been eliminated by the proposed stabilized share-parameter proxy, which employs Stochastic Gradient Langevin Dynamics to enable fast shared parameter sampling, so as to achieve stabilized measurement of architecture performance even in search space with complex topological structures. The proposed method, namely Stablized Topological Neural Architecture Search (ST-NAS), achieves state-of-the-art performance under Multiply-Adds (MAdds) constraint on ImageNet. Our lite model ST-NAS-A achieves 76.4% top-1 accuracy with only 326M MAdds. Our moderate model ST-NAS-B achieves 77.9% top-1 accuracy just required 503M MAdds. Both of our models offer superior performances in comparison to other concurrent works on one-shot NAS."
                    },
                    {
                        "title": "MaskMA: Towards Zero-Shot Multi-Agent Decision Making with Mask-Based Collaborative Learning",
                        "abstract": "Building a single generalist agent with strong zero-shot capability has recently sparked significant advancements. However, extending this capability to multi-agent decision making scenarios presents challenges. Most current works struggle with zero-shot transfer, due to two challenges particular to the multi-agent settings: (a) a mismatch between centralized training and decentralized execution; and (b) difficulties in creating generalizable representations across diverse tasks due to varying agent numbers and action spaces. To overcome these challenges, we propose a Mask-Based collaborative learning framework for Multi-Agent decision making (MaskMA). Firstly, we propose to randomly mask part of the units and collaboratively learn the policies of unmasked units to handle the mismatch. In addition, MaskMA integrates a generalizable action representation by dividing the action space into intrinsic actions solely related to the unit itself and interactive actions involving interactions with other units. This flexibility allows MaskMA to tackle tasks with varying agent numbers and thus different action spaces. Extensive experiments in SMAC reveal MaskMA, with a single model trained on 11 training maps, can achieve an impressive 77.8% average zero-shot win rate on 60 unseen test maps by decentralized execution, while also performing effectively on other types of downstream tasks (e.g., varied policies collaboration, ally malfunction, and ad hoc team play)."
                    },
                    {
                        "title": "BN-NAS: Neural Architecture Search with Batch Normalization",
                        "abstract": "We present BN-NAS, neural architecture search with Batch Normalization (BN-NAS), to accelerate neural architecture search (NAS). BN-NAS can significantly reduce the time required by model training and evaluation in NAS. Specifically, for fast evaluation, we propose a BN-based indicator for predicting subnet performance at a very early training stage. The BN-based indicator further facilitates us to improve the training efficiency by only training the BN parameters during the supernet training. This is based on our observation that training the whole supernet is not necessary while training only BN parameters accelerates network convergence for network architecture search. Extensive experiments show that our method can significantly shorten the time of training supernet by more than 10 times and shorten the time of evaluating subnets by more than 600,000 times without losing accuracy."
                    },
                    {
                        "title": "GLiT: Neural Architecture Search for Global and Local Image Transformer",
                        "abstract": "We introduce the first Neural Architecture Search (NAS) method to find a better transformer architecture for image recognition. Recently, transformers without CNN-based backbones are found to achieve impressive performance for image recognition. However, the transformer is designed for NLP tasks and thus could be sub-optimal when directly used for image recognition. In order to improve the visual representation ability for transformers, we propose a new search space and searching algorithm. Specifically, we introduce a locality module that models the local correlations in images explicitly with fewer computational cost. With the locality module, our search space is defined to let the search algorithm freely trade off between global and local information as well as optimizing the low-level design choice in each module. To tackle the problem caused by huge search space, a hierarchical neural architecture search method is proposed to search the optimal vision transformer from two levels separately with the evolutionary algorithm. Extensive experiments on the ImageNet dataset demonstrate that our method can find more discriminative and efficient transformer variants than the ResNet family (e.g., ResNet101) and the baseline ViT for image classification."
                    },
                    {
                        "title": "PSViT: Better Vision Transformer via Token Pooling and Attention Sharing",
                        "abstract": "In this paper, we observe two levels of redundancies when applying vision transformers (ViT) for image recognition. First, fixing the number of tokens through the whole network produces redundant features at the spatial level. Second, the attention maps among different transformer layers are redundant. Based on the observations above, we propose a PSViT: a ViT with token Pooling and attention Sharing to reduce the redundancy, effectively enhancing the feature representation ability, and achieving a better speed-accuracy trade-off. Specifically, in our PSViT, token pooling can be defined as the operation that decreases the number of tokens at the spatial level. Besides, attention sharing will be built between the neighboring transformer layers for reusing the attention maps having a strong correlation among adjacent layers. Then, a compact set of the possible combinations for different token pooling and attention sharing mechanisms are constructed. Based on the proposed compact set, the number of tokens in each layer and the choices of layers sharing attention can be treated as hyper-parameters that are learned from data automatically. Experimental results show that the proposed scheme can achieve up to 6.6% accuracy improvement in ImageNet classification compared with the DeiT."
                    },
                    {
                        "title": "Fast Learning of Temporal Action Proposal via Dense Boundary Generator",
                        "abstract": "Generating temporal action proposals remains a very challenging problem, where the main issue lies in predicting precise temporal proposal boundaries and reliable action confidence in long and untrimmed real-world videos. In this paper, we propose an efficient and unified framework to generate temporal action proposals named Dense Boundary Generator (DBG), which draws inspiration from boundary-sensitive methods and implements boundary classification and action completeness regression for densely distributed proposals. In particular, the DBG consists of two modules: Temporal boundary classification (TBC) and Action-aware completeness regression (ACR). The TBC aims to provide two temporal boundary confidence maps by low-level two-stream features, while the ACR is designed to generate an action completeness score map by high-level action-aware features. Moreover, we introduce a dual stream BaseNet (DSB) to encode RGB and optical flow information, which helps to capture discriminative boundary and actionness features. Extensive experiments on popular benchmarks ActivityNet-1.3 and THUMOS14 demonstrate the superiority of DBG over the state-of-the-art proposal generator (e.g., MGG and BMN). Our code will be made available upon publication."
                    },
                    {
                        "title": "Once Quantization-Aware Training: High Performance Extremely Low-bit Architecture Search",
                        "abstract": "Quantization Neural Networks (QNN) have attracted a lot of attention due to their high efficiency. To enhance the quantization accuracy, prior works mainly focus on designing advanced quantization algorithms but still fail to achieve satisfactory results under the extremely low-bit case. In this work, we take an architecture perspective to investigate the potential of high-performance QNN. Therefore, we propose to combine Network Architecture Search methods with quantization to enjoy the merits of the two sides. However, a naive combination inevitably faces unacceptable time consumption or unstable training problem. To alleviate these problems, we first propose the joint training of architecture and quantization with a shared step size to acquire a large number of quantized models. Then a bit-inheritance scheme is introduced to transfer the quantized models to the lower bit, which further reduces the time cost and meanwhile improves the quantization accuracy. Equipped with this overall framework, dubbed as Once Quantization-Aware Training~(OQAT), our searched model family, OQATNets, achieves a new state-of-the-art compared with various architectures under different bit-widths. In particular, OQAT-2bit-M achieves 61.6% ImageNet Top-1 accuracy, outperforming 2-bit counterpart MobileNetV3 by a large margin of 9% with 10% less computation cost. A series of quantization-friendly architectures are identified easily and extensive analysis can be made to summarize the interaction between quantization and neural architectures. Codes and models are released at https://github.com/LaVieEnRoseSMZ/OQA"
                    },
                    {
                        "title": "Inception Convolution with Efficient Dilation Search",
                        "abstract": "As a variant of standard convolution, a dilated convolution can control effective receptive fields and handle large scale variance of objects without introducing additional computational costs. To fully explore the potential of dilated convolution, we proposed a new type of dilated convolution (referred to as inception convolution), where the convolution operations have independent dilation patterns among different axes, channels and layers. To develop a practical method for learning complex inception convolution based on the data, a simple but effective search algorithm, referred to as efficient dilation optimization (EDO), is developed. Based on statistical optimization, the EDO method operates in a low-cost manner and is extremely fast when it is applied on large scale datasets. Empirical results validate that our method achieves consistent performance gains for image recognition, object detection, instance segmentation, human detection, and human pose estimation. For instance, by simply replacing the 3x3 standard convolution in the ResNet-50 backbone with inception convolution, we significantly improve the AP of Faster R-CNN from 36.4% to 39.2% on MS COCO."
                    },
                    {
                        "title": "Online Hyper-parameter Learning for Auto-Augmentation Strategy",
                        "abstract": "Data augmentation is critical to the success of modern deep learning techniques. In this paper, we propose Online Hyper-parameter Learning for Auto-Augmentation (OHL-Auto-Aug), an economical solution that learns the augmentation policy distribution along with network training. Unlike previous methods on auto-augmentation that search augmentation strategies in an offline manner, our method formulates the augmentation policy as a parameterized probability distribution, thus allowing its parameters to be optimized jointly with network parameters. Our proposed OHL-Auto-Aug eliminates the need of re-training and dramatically reduces the cost of the overall search process, while establishes significantly accuracy improvements over baseline models. On both CIFAR-10 and ImageNet, our method achieves remarkable on search accuracy, 60x faster on CIFAR-10 and 24x faster on ImageNet, while maintaining competitive accuracies."
                    },
                    {
                        "title": "Theoretically Guaranteed Policy Improvement Distilled from Model-Based Planning",
                        "abstract": "Model-based reinforcement learning (RL) has demonstrated remarkable successes on a range of continuous control tasks due to its high sample efficiency. To save the computation cost of conducting planning online, recent practices tend to distill optimized action sequences into an RL policy during the training phase. Although the distillation can incorporate both the foresight of planning and the exploration ability of RL policies, the theoretical understanding of these methods is yet unclear. In this paper, we extend the policy improvement step of Soft Actor-Critic (SAC) by developing an approach to distill from model-based planning to the policy. We then demonstrate that such an approach of policy improvement has a theoretical guarantee of monotonic improvement and convergence to the maximum value defined in SAC. We discuss effective design choices and implement our theory as a practical algorithm -- Model-based Planning Distilled to Policy (MPDP) -- that updates the policy jointly over multiple future time steps. Extensive experiments show that MPDP achieves better sample efficiency and asymptotic performance than both model-free and model-based planning algorithms on six continuous control benchmark tasks in MuJoCo."
                    }
                ]
            },
            "705ab9c1-25ac-480a-8dda-fa338ed3f054": {
                "pk": "705ab9c1-25ac-480a-8dda-fa338ed3f054",
                "name": "Canran Xiao",
                "collaborators": [
                    "Xiaochao Pan",
                    "Jiawei Yao",
                    "Hongrui Kou",
                    "Tong Wu"
                ],
                "domain": [
                    "Computer Vision",
                    "3D Reconstruction",
                    "Autonomous Driving",
                    "Neural Radiance Fields"
                ],
                "publications": [
                    {
                        "title": "HarmonicNeRF: Geometry-Informed Synthetic View Augmentation for 3D Scene Reconstruction in Driving Scenarios",
                        "abstract": "In the realm of autonomous driving, achieving precise 3D reconstruction of the driving environment is critical for ensuring safety and effective navigation. Neural Radiance Fields (NeRF) have shown promise in creating highly detailed and accurate models of complex environments. However, the application of NeRF in autonomous driving scenarios encounters several challenges, primarily due to the sparsity of viewpoints inherent in camera trajectories and the constraints on data collection in unbounded outdoor scenes, which typically occur along predetermined paths. This limitation not only reduces the available scene information but also poses significant challenges for NeRF training, as the sparse and path-distributed observational data leads to under-representation of the scene's geometry. In this paper, we introduce HarmonicNeRF, a novel approach for outdoor self-supervised monocular scene reconstruction. HarmonicNeRF capitalizes on the strengths of NeRF and enhances surface reconstruction accuracy by augmenting the input space with geometry-informed synthetic views. This is achieved through the application of spherical harmonics to generate novel radiance values, taking into careful consideration the color observations from the limited available real-world views. Additionally, our method incorporates proxy geometry to effectively manage occlusion, generating radiance pseudo-labels that circumvent the limitations of traditional image-warping techniques, which often fail in sparse data conditions typical of autonomous driving environments. Extensive experiments conducted on the KITTI, Argoverse, and NuScenes datasets demonstrate our approach establishes new benchmarks in synthesizing novel depth views and reconstructing scenes, significantly outperforming existing methods. Project page: https://github.com/Jiawei-Yao0812/HarmonicNeRF"
                    }
                ]
            }
        }
    },
    "2406.17863": {
        "paper_data": {
            "title": "What type of inference is planning?",
            "url": "http://arxiv.org/abs/2406.17863v2",
            "arxiv_id": "2406.17863",
            "authors": [
                "Miguel L\u00e1zaro-Gredilla",
                "Li Yang Ku",
                "Kevin P. Murphy",
                "Dileep George"
            ],
            "abstract": "Multiple types of inference are available for probabilistic graphical models, e.g., marginal, maximum-a-posteriori, and even marginal maximum-a-posteriori. Which one do researchers mean when they talk about \"planning as inference\"? There is no consistency in the literature, different types are used, and their ability to do planning is further entangled with specific approximations or additional constraints. In this work we use the variational framework to show that all commonly used types of inference correspond to different weightings of the entropy terms in the variational problem, and that planning corresponds _exactly_ to a _different_ set of weights. This means that all the tricks of variational inference are readily applicable to planning. We develop an analogue of loopy belief propagation that allows us to perform approximate planning in factored state Markov decisions processes without incurring intractability due to the exponentially large state space. The variational perspective shows that the previous types of inference for planning are only adequate in environments with low stochasticity, and allows us to characterize each type by its own merits, disentangling the type of inference from the additional approximations that its practical use requires. We validate these results empirically on synthetic MDPs and tasks posed in the International Planning Competition.",
            "introduction": "   1 Introduction  There are many kinds of probabilistic inference that are used, such as marginal, maximum-a-posteriori (MAP), or marginal MAP (MMAP). In this work we show that planning is a distinct type of inference, and that under stochastic dynamics does not correspond exactly with any of the above methods. Furthermore, we show how to rank the above methods in terms of quality as it pertains to planning.   Our approach is based on variational inference. The variational perspective that we adopt allows direct comparison between different inference types, and allows us to develop analogues of existing approximate inference algorithms for this \u201cplanning inference\u201d task. Given a \"flat\" Markov Decision Process (MDP), respresented as the probabilistic graphical model (PGM) in Fig 1[Left], we show that planning inference provides the same (exact) results as value iteration. And using an analogue of loopy belief propagation (LBP), we show how to apply approximate planning inference to the factored MDP in Fig.\u00a01[Right], which has an exponentially large state space, and for which exact solutions are no longer tractable. Under moderate stochasticity in the dynamics, we show that this approximate planning inference is superior to other more established types of inference.   Figure 1: Factor graphs: [Left] Standard MDP [Right] Factored MDP with sparse factor connectivity.     2 Background   2.1 Markov Decision processes (MDPs)  A finite-horizon Markov decision process (MDP) is a tuple (\ud835\udcb3,\ud835\udc9c,p\u2062(x1),\ud835\udcab,\u211b,T)\ud835\udcb3\ud835\udc9c\ud835\udc5dsubscript\ud835\udc651\ud835\udcab\u211b\ud835\udc47({\\cal X},{\\cal A},p(x_{1}),{\\cal P},{\\cal R},T)( caligraphic_X , caligraphic_A , italic_p ( italic_x start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ) , caligraphic_P , caligraphic_R , italic_T ), where \ud835\udcb3\ud835\udcb3{\\cal X}caligraphic_X is the state space, \ud835\udc9c\ud835\udc9c{\\cal A}caligraphic_A the action space, p\u2062(x1)\ud835\udc5dsubscript\ud835\udc651p(x_{1})italic_p ( italic_x start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ) the starting state distribution, \ud835\udcab\ud835\udcab{\\cal P}caligraphic_P the transition probabilities P\u2062(xt+1|xt,at)\ud835\udc43conditionalsubscript\ud835\udc65\ud835\udc611subscript\ud835\udc65\ud835\udc61subscript\ud835\udc4e\ud835\udc61P(x_{t+1}|x_{t},a_{t})italic_P ( italic_x start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT | italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT , italic_a start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) (i.e., the dynamics of the process), Rt\u2062(xt,at,xt+1)subscript\ud835\udc45\ud835\udc61subscript\ud835\udc65\ud835\udc61subscript\ud835\udc4e\ud835\udc61subscript\ud835\udc65\ud835\udc611R_{t}(x_{t},a_{t},x_{t+1})italic_R start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ( italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT , italic_a start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT ) the reward for transitioning from xtsubscript\ud835\udc65\ud835\udc61x_{t}italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT to xt+1subscript\ud835\udc65\ud835\udc611x_{t+1}italic_x start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT under action atsubscript\ud835\udc4e\ud835\udc61a_{t}italic_a start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT at timestep t\ud835\udc61titalic_t, and T\ud835\udc47Titalic_T is the horizon. For simplicity of exposition we will consider discrete states and actions, but analogous results apply in the continuous case. Solving an MDP corresponds to finding a policy \u03c0t\u2062(at|xt)subscript\ud835\udf0b\ud835\udc61conditionalsubscript\ud835\udc4e\ud835\udc61subscript\ud835\udc65\ud835\udc61\\pi_{t}(a_{t}|x_{t})italic_\u03c0 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ( italic_a start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT | italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) that maximizes the expectation of some function g\u2062(\u22c5)\ud835\udc54\u22c5g(\\cdot)italic_g ( \u22c5 ) of the reward \ud835\udd3c\u2062[g\u2062(R1\u2062(x1,a1,x2),\u2026,RT\u22121\u2062(xT\u22121,aT\u22121,xT))]\ud835\udd3cdelimited-[]\ud835\udc54subscript\ud835\udc451subscript\ud835\udc651subscript\ud835\udc4e1subscript\ud835\udc652\u2026subscript\ud835\udc45\ud835\udc471subscript\ud835\udc65\ud835\udc471subscript\ud835\udc4e\ud835\udc471subscript\ud835\udc65\ud835\udc47\\mathbb{E}[g(R_{1}(x_{1},a_{1},x_{2}),\\ldots,R_{T-1}(x_{T-1},a_{T-1},x_{T}))]blackboard_E [ italic_g ( italic_R start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_x start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , italic_a start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ) , \u2026 , italic_R start_POSTSUBSCRIPT italic_T - 1 end_POSTSUBSCRIPT ( italic_x start_POSTSUBSCRIPT italic_T - 1 end_POSTSUBSCRIPT , italic_a start_POSTSUBSCRIPT italic_T - 1 end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT italic_T end_POSTSUBSCRIPT ) ) ], typically its sum. The optimal policy can be different at each time step, i.e., non-stationary. We do not use a discount factor, but it can be trivially included in the reward function, since it is also non-stationary. Policies, states and actions at all timesteps will be noted \ud835\udf45\u2261{\u03c0t}t=1T\u22121,\ud835\udc99\u2261{xt}t=1Tformulae-sequence\ud835\udf45superscriptsubscriptsubscript\ud835\udf0b\ud835\udc61\ud835\udc611\ud835\udc471\ud835\udc99superscriptsubscriptsubscript\ud835\udc65\ud835\udc61\ud835\udc611\ud835\udc47\\bm{\\pi}\\equiv\\{\\pi_{t}\\}_{t=1}^{T-1},\\bm{x}\\equiv\\{x_{t}\\}_{t=1}^{T}bold_italic_\u03c0 \u2261 { italic_\u03c0 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT } start_POSTSUBSCRIPT italic_t = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_T - 1 end_POSTSUPERSCRIPT , bold_italic_x \u2261 { italic_x start_POSTSUBSCRIPT italic_t",
            "references": [
                {
                    "title": "Finite-Horizon Constrained MDPs With Both Additive And Multiplicative Utilities",
                    "abstract": "This paper considers the problem of finding a solution to the finite horizon constrained Markov decision processes (CMDP) where the objective as well as constraints are sum of additive and multiplicative utilities. Towards solving this, we construct another CMDP, with only additive utilities under a restricted set of policies, whose optimal value is equal to that of the original CMDP. Furthermore, we provide a finite dimensional bilinear program (BLP) whose value equals the CMDP value and whose solution provides the optimal policy. We also suggest an algorithm to solve this BLP."
                },
                {
                    "title": "Approximate Inference for Stochastic Planning in Factored Spaces",
                    "abstract": "Stochastic planning can be reduced to probabilistic inference in large discrete graphical models, but hardness of inference requires approximation schemes to be used. In this paper we argue that such applications can be disentangled along two dimensions. The first is the direction of information flow in the idealized exact optimization objective, i.e., forward vs. backward inference. The second is the type of approximation used to compute this objective, e.g., Belief Propagation (BP) vs. mean field variational inference (MFVI). This new categorization allows us to unify a large amount of isolated efforts in prior work explaining their connections and differences as well as potential improvements. An extensive experimental evaluation over large stochastic planning problems shows the advantage of forward BP over several algorithms based on MFVI. An analysis of practical limitations of MFVI motivates a novel algorithm, collapsed state variational inference (CSVI), which provides a tighter approximation and achieves comparable planning performance with forward BP."
                },
                {
                    "title": "Stochastic Planning with Lifted Symbolic Trajectory Optimization",
                    "abstract": "This paper investigates online stochastic planning for problems with large factored state and action spaces. One promising approach in recent work estimates the quality of applicable actions in the current state through aggregate simulation from the states they reach. This leads to significant speedup, compared to search over concrete states and actions, and suffices to guide decision making in cases where the performance of a random policy is informative of the quality of a state. The paper makes two significant improvements to this approach. The first, taking inspiration from lifted belief propagation, exploits the structure of the problem to derive a more compact computation graph for aggregate simulation. The second improvement replaces the random policy embedded in the computation graph with symbolic variables that are optimized simultaneously with the search for high quality actions. This expands the scope of the approach to problems that require deep search and where information is lost quickly with random steps. An empirical evaluation shows that these ideas significantly improve performance, leading to state of the art performance on hard planning problems."
                },
                {
                    "title": "Reinforcement Learning and Control as Probabilistic Inference: Tutorial and Review",
                    "abstract": "The framework of reinforcement learning or optimal control provides a mathematical formalization of intelligent decision making that is powerful and broadly applicable. While the general form of the reinforcement learning problem enables effective reasoning about uncertainty, the connection between reinforcement learning and inference in probabilistic models is not immediately obvious. However, such a connection has considerable value when it comes to algorithm design: formalizing a problem as probabilistic inference in principle allows us to bring to bear a wide array of approximate inference tools, extend the model in flexible and powerful ways, and reason about compositionality and partial observability. In this article, we will discuss how a generalization of the reinforcement learning or optimal control problem, which is sometimes termed maximum entropy reinforcement learning, is equivalent to exact probabilistic inference in the case of deterministic dynamics, and variational inference in the case of stochastic dynamics. We will present a detailed derivation of this framework, overview prior work that has drawn on this and related ideas to propose new reinforcement learning and control algorithms, and describe perspectives on future research."
                },
                {
                    "title": "Causal Consistency of Structural Equation Models",
                    "abstract": "Complex systems can be modelled at various levels of detail. Ideally, causal models of the same system should be consistent with one another in the sense that they agree in their predictions of the effects of interventions. We formalise this notion of consistency in the case of Structural Equation Models (SEMs) by introducing exact transformations between SEMs. This provides a general language to consider, for instance, the different levels of description in the following three scenarios: (a) models with large numbers of variables versus models in which the `irrelevant' or unobservable variables have been marginalised out; (b) micro-level models versus macro-level models in which the macro-variables are aggregate features of the micro-variables; (c) dynamical time series models versus models of their stationary behaviour. Our analysis stresses the importance of well specified interventions in the causal modelling process and sheds light on the interpretation of cyclic SEMs."
                },
                {
                    "title": "Online Symbolic Gradient-Based Optimization for Factored Action MDPs",
                    "abstract": "This paper investigates online stochastic planning for problems with large factored state and action spaces. We introduce a novel algorithm that builds a symbolic representation capturing an approximation of the action-value Q-function in terms of action variables, and then performs gradient based search to select an action for the current state. The algorithm can be seen as a symbolic extension of Monte-Carlo search, induced by independence assumptions on state and action variables, and augmented with gradients to speed up the search. This avoids the space explosion typically faced by symbolic methods, and the dearth of samples faced by Monte-Carlo methods when the action space is large. An experimental evaluation on benchmark problems shows that the algorithm is competitive with state of the art across problem sizes and that it provides significant improvements for large factored action spaces."
                },
                {
                    "title": "Causal inference and the data-fusion problem",
                    "abstract": "We review concepts, principles, and tools that unify current approaches to causal analysis and attend to new challenges presented by big data. In particular, we address the problem of data fusion\u2014piecing together multiple datasets collected under heterogeneous conditions (i.e., different populations, regimes, and sampling methods) to obtain valid answers to queries of interest. The availability of multiple heterogeneous datasets presents new opportunities to big data analysts, because the knowledge that can be acquired from combined data would not be possible from any individual source alone. However, the biases that emerge in heterogeneous environments require new analytical tools. Some of these biases, including confounding, sampling selection, and cross-population biases, have been addressed in isolation, largely in restricted parametric models. We here present a general, nonparametric framework for handling these biases and, ultimately, a theoretical solution to the problem of data fusion in causal inference tasks."
                },
                {
                    "title": "Finite horizon risk sensitive MDP and linear programming",
                    "abstract": "In the context of standard Markov decision processes (MDPs), the connection between Dynamic Program (DP) and Linear Program (LP) is well understood and is well established under sufficiently general conditions. LP based approach facilitates solving the constrained MDPs. Multiplicative or Risk sensitive MDPs, introduced to control the fluctuations/variations around the expected value, are relatively less studied objects. DP equations are considerably well understood even in the context of Risk MDPs, however the LP connection is not known. We consider a finite horizon risk MDP problem and establish the connections between the DP and LP approaches. We augment the state space with a suitable component, to obtain the optimal policies for constrained risk MDPs. We apply this results to a server selection problem in Ber/M/K/K queues, with a constraint on the utilization of the fast server. We discuss some interesting structural properties of the risk optimal policies."
                },
                {
                    "title": "Factored MCTS for Large Scale Stochastic Planning",
                    "abstract": "\n \n This paper investigates stochastic planning problemswith large factored state and action spaces. We show that even with moderate increase in the size of existing challenge problems, the performance of state of the art algorithms deteriorates rapidly, making them ineffective.To address this problem we propose a family of simple but scalable online planning algorithms that combine sampling, as in Monte Carlo tree search, with \u201caggregation,\u201d where the aggregation approximates a distribution over random variables by the product of their marginals. The algorithms are correct under some rather strong technical conditions and can serve as an unsound but effective heuristic when the conditions do not hold. An extensive experimental evaluation demonstrates that the new algorithms provide significant improvement over the state of the art when solving largeproblems in a number of challenge benchmark domains.\n \n"
                },
                {
                    "title": "Understanding the Bethe Approximation: When and How can it go Wrong?",
                    "abstract": "Belief propagation is a remarkably effective tool for inference, even when applied to networks with cycles. It may be viewed as a way to seek the minimum of the Bethe free energy, though with no convergence guarantee in general. A variational perspective shows that, compared to exact inference, this minimization employs two forms of approximation: (i) the true entropy is approximated by the Bethe entropy, and (ii) the minimization is performed over a relaxation of the marginal polytope termed the local polytope. Here we explore when and how the Bethe approximation can fail for binary pairwise models by examining each aspect of the approximation, deriving results both analytically and with new experimental methods."
                },
                {
                    "title": "Linear Programming for Large-Scale Markov Decision Problems",
                    "abstract": "We consider the problem of controlling a Markov decision process (MDP) with a large state space, so as to minimize average cost. Since it is intractable to compete with the optimal policy for large scale problems, we pursue the more modest goal of competing with a low-dimensional family of policies. We use the dual linear programming formulation of the MDP average cost problem, in which the variable is a stationary distribution over state-action pairs, and we consider a neighborhood of a low-dimensional subset of the set of stationary distributions (defined in terms of state-action features) as the comparison class. We propose a technique based on stochastic convex optimization and give bounds that show that the performance of our algorithm approaches the best achievable by any policy in the comparison class. Most importantly, this result depends on the size of the comparison class, but not on the size of the state space. Preliminary experiments show the effectiveness of the proposed algorithm in a queuing application."
                },
                {
                    "title": "Variational Planning for Graph-based MDPs",
                    "abstract": "Markov Decision Processes (MDPs) are extremely useful for modeling and solving sequential decision making problems. Graph-based MDPs provide a compact representation for MDPs with large numbers of random variables. However, the complexity of exactly solving a graph-based MDP usually grows exponentially in the number of variables, which limits their application. We present a new variational framework to describe and solve the planning problem of MDPs, and derive both exact and approximate planning algorithms. In particular, by exploiting the graph structure of graph-based MDPs, we propose a factored variational value iteration algorithm in which the value function is first approximated by the multiplication of local-scope value functions, then solved by minimizing a Kullback-Leibler (KL) divergence. The KL divergence is optimized using the belief propagation algorithm, with complexity exponential in only the cluster size of the graph. Experimental comparison on different models shows that our algorithm outperforms existing approximation algorithms at finding good policies."
                },
                {
                    "title": "Risk-Sensitive Reinforcement Learning",
                    "abstract": "We derive a family of risk-sensitive reinforcement learning methods for agents, who face sequential decision-making tasks in uncertain environments. By applying a utility function to the temporal difference (TD) error, nonlinear transformations are effectively applied not only to the received rewards but also to the true transition probabilities of the underlying Markov decision process. When appropriate utility functions are chosen, the agents\u2019 behaviors express key features of human behavior as predicted by prospect theory (Kahneman & Tversky, 1979), for example, different risk preferences for gains and losses, as well as the shape of subjective probability curves. We derive a risk-sensitive Q-learning algorithm, which is necessary for modeling human behavior when transition probabilities are unknown, and prove its convergence. As a proof of principle for the applicability of the new framework, we apply it to quantify human behavior in a sequential investment task. We find that the risk-sensitive variant provides a significantly better fit to the behavioral data and that it leads to an interpretation of the subject's responses that is indeed consistent with prospect theory. The analysis of simultaneously measured fMRI signals shows a significant correlation of the risk-sensitive TD error with BOLD signal change in the ventral striatum. In addition we find a significant correlation of the risk-sensitive Q-values with neural activity in the striatum, cingulate cortex, and insula that is not present if standard Q-values are used."
                },
                {
                    "title": "The Causal Foundations of Structural Equation Modeling",
                    "abstract": "Abstract : The role of causality in SEM research is widely perceived to be, on the one hand, of pivotal methodological importance and, on the other hand, confusing, enigmatic and controversial. The confusion is vividly portrayed, for example, in the influential report of Wilkinson and Task Force s (1999) on Statistical Methods in Psychology Journals: Guidelines and Explanations. In discussing SEM, the report starts with the usual warning: It is sometimes thought that correlation does not prove causation but causal modeling does. [Wrong! There are] dangers in this practice. But then ends with a startling conclusion: The use of complicated causal-modeling software [read SEM] rarely yields any results that have any interpretation as causal effects. The implication being that the entire enterprise of causal modeling, from Sewell Wright (1921) to Blalock (1964) and Duncan (1975), the entire literature in econometric research, including modern advances in graphical and nonparametric structural models have all been misguided, for they have been chasing parameters that have no causal interpretation."
                },
                {
                    "title": "Variational algorithms for marginal MAP",
                    "abstract": "The marginal maximum a posteriori probability (MAP) estimation problem, which calculates the mode of the marginal posterior distribution of a subset of variables with the remaining variables marginalized, is an important inference problem in many models, such as those with hidden variables or uncertain parameters. Unfortunately, marginal MAP can be NP-hard even on trees, and has attracted less attention in the literature compared to the joint MAP (maximization) and marginalization problems. We derive a general dual representation for marginal MAP that naturally integrates the marginalization and maximization operations into a joint variational optimization problem, making it possible to easily extend most or all variational-based algorithms to marginal MAP. In particular, we derive a set of \"mixed-product\" message passing algorithms for marginal MAP, whose form is a hybrid of max-product, sum-product and a novel \"argmax-product\" message updates. We also derive a class of convergent algorithms based on proximal point methods, including one that transforms the marginal MAP problem into a sequence of standard marginalization problems. Theoretically, we provide guarantees under which our algorithms give globally or locally optimal solutions, and provide novel upper bounds on the optimal objectives. Empirically, we demonstrate that our algorithms significantly outperform the existing approaches, including a state-of-the-art algorithm based on local search methods."
                },
                {
                    "title": "The FF Planning System: Fast Plan Generation Through Heuristic Search",
                    "abstract": "We describe and evaluate the algorithmic techniques that are used in the FF planning system. Like the HSP system, FF relies on forward state space search, using a heuristic that estimates goal distances by ignoring delete lists. Unlike HSP's heuristic, our method does not assume facts to be independent. We introduce a novel search strategy that combines hill-climbing with systematic search, and we show how other powerful heuristic information can be extracted and used to prune the search space. FF was the most successful automatic planner at the recent AIPS-2000 planning competition. We review the results of the competition, give data for other benchmark domains, and investigate the reasons for the runtime performance of FF compared to HSP."
                },
                {
                    "title": "Norm-Product Belief Propagation: Primal-Dual Message-Passing for Approximate Inference",
                    "abstract": "Inference problems in graphical models can be represented as a constrained optimization of a free-energy function. In this paper, we treat both forms of probabilistic inference, estimating marginal probabilities of the joint distribution and finding the most probable assignment, through a unified message-passing algorithm architecture. In particular we generalize the belief propagation (BP) algorithms of sum-product and max-product and tree-reweighted (TRW) sum and max product algorithms (TRBP) and introduce a new set of convergent algorithms based on \u201cconvex-free-energy\u201d and linear-programming (LP) relaxation as a zero-temperature of a convex-free-energy. The main idea of this work arises from taking a general perspective on the existing BP and TRBP algorithms while observing that they all are reductions from the basic optimization formula of f +\u03a3ihi where the function f is an extended-valued, strictly convex but nonsmooth and the functions hi are extended-valued functions (not necessarily convex). We use tools from convex duality to present the \u201cprimal-dual ascent\u201d algorithm which is an extension of the Bregman successive projection scheme and is designed to handle optimization of the general type f + \u03a3ihi. We then map the fractional-free-energy variational principle for approximate inference onto the optimization formula above and introduce the \u201cnorm-product\u201d message-passing algorithm. Special cases of the norm-product include sum-product and max-product (BP algorithms), TRBP and NMPLP algorithms. When the fractional-free-energy is set to be convex (convex-free-energy) the norm-product is globally convergent for the estimation of marginal probabilities and for approximating the LP-relaxation. We also introduce another branch of the norm-product which arises as the \u201czero-temperature\u201d of the convex-free-energy which we refer to as the \u201cconvex-max-product\u201d. The convex-max-product is convergent (unlike max-product) and aims at solving the LP- relaxation."
                },
                {
                    "title": "Probabilistic Planning via Determinization in Hindsight",
                    "abstract": "This paper investigates hindsight optimization as an approach for leveraging the significant advances in deterministic planning for action selection in probabilistic domains. Hindsight optimization is an online technique that evaluates the one-step-reachable states by sampling future outcomes to generate multiple non-stationary deterministic planning problems which can then be solved using search. Hindsight optimization has been successfully used in a number of online scheduling applications; however, it has not yet been considered in the substantially different context of goal-based probabilistic planning. We describe an implementation of hindsight optimization for probabilistic planning based on deterministic forward heuristic search and evaluate its performance on planning-competition benchmarks and other probabilistically interesting problems. The planner is able to outperform a number of probabilistic planners including FF-Replan on many problems. Finally, we investigate conditions under which hindsight optimization is guaranteed to be effective with respect to goal achievement, and also illustrate examples where the approach can go wrong."
                },
                {
                    "title": "MAP Estimation, Linear Programming and Belief Propagation with Convex Free Energies",
                    "abstract": "Finding the most probable assignment (MAP) in a general graphical model is known to be NP hard but good approximations have been attained with max-product belief propagation (BP) and its variants. In particular, it is known that using BP on a single-cycle graph or tree reweighted BP on an arbitrary graph will give the MAP solution if the beliefs have no ties. In this paper we extend the setting under which BP can be used to provably extract the MAP. We define Convex BP as BP algorithms based on a convex free energy approximation and show that this class includes ordinary BP with single-cycle, tree reweighted BP and many other BP variants. We show that when there are no ties, fixed-points of convex max-product BP will provably give the MAP solution. We also show that convex sum-product BP at sufficiently small temperatures can be used to solve linear programs that arise from relaxing the MAP problem. Finally, we derive a novel condition that allows us to derive the MAP solution even if some of the convex BP beliefs have ties. In experiments, we show that our theorems allow us to find the MAP in many real-world instances of graphical models where exact inference using junction-tree is impossible."
                },
                {
                    "title": "Convergent Tree-Reweighted Message Passing for Energy Minimization",
                    "abstract": "Algorithms for discrete energy minimization are of fundamental importance in computer vision. In this paper, we focus on the recent technique proposed by Wainwright et al. (Nov. 2005)- tree-reweighted max-product message passing (TRW). It was inspired by the problem of maximizing a lower bound on the energy. However, the algorithm is not guaranteed to increase this bound - it may actually go down. In addition, TRW does not always converge. We develop a modification of this algorithm which we call sequential tree-reweighted message passing. Its main property is that the bound is guaranteed not to decrease. We also give a weak tree agreement condition which characterizes local maxima of the bound with respect to TRW algorithms. We prove that our algorithm has a limit point that achieves weak tree agreement. Finally, we show that, our algorithm requires half as much memory as traditional message passing approaches. Experimental results demonstrate that on certain synthetic and real problems, our algorithm outperforms both the ordinary belief propagation and tree-reweighted algorithm in (M. J. Wainwright, et al., Nov. 2005). In addition, on stereo problems with Potts interactions, we obtain a lower energy than graph cuts"
                },
                {
                    "title": "Probabilistic inference for solving discrete and continuous state Markov Decision Processes",
                    "abstract": "Inference in Markov Decision Processes has recently received interest as a means to infer goals of an observed action, policy recognition, and also as a tool to compute policies. A particularly interesting aspect of the approach is that any existing inference technique in DBNs now becomes available for answering behavioral question--including those on continuous, factorial, or hierarchical state representations. Here we present an Expectation Maximization algorithm for computing optimal policies. Unlike previous approaches we can show that this actually optimizes the discounted expected future return for arbitrary reward functions and without assuming an ad hoc finite total time. The algorithm is generic in that any inference technique can be utilized in the E-step. We demonstrate this for exact inference on a discrete maze and Gaussian belief state propagation in continuous stochastic optimal control problems."
                },
                {
                    "title": "The Fast Downward Planning System",
                    "abstract": "Fast Downward is a classical planning system based on heuristic search. It can deal with general deterministic planning problems encoded in the propositional fragment of PDDL2.2, including advanced features like ADL conditions and effects and derived predicates (axioms). Like other well-known planners such as HSP and FF, Fast Downward is a progression planner, searching the space of world states of a planning task in the forward direction. However, unlike other PDDL planning systems, Fast Downward does not use the propositional PDDL representation of a planning task directly. Instead, the input is first translated into an alternative representation called multivalued planning tasks, which makes many of the implicit constraints of a propositional planning task explicit. Exploiting this alternative representation, Fast Downward uses hierarchical decompositions of planning tasks for computing its heuristic function, called the causal graph heuristic, which is very different from traditional HSP-like heuristics based on ignoring negative interactions of operators. \n \nIn this article, we give a full account of Fast Downward's approach to solving multivalued planning tasks. We extend our earlier discussion of the causal graph heuristic to tasks involving axioms and conditional effects and present some novel techniques for search control that are used within Fast Downward's best-first search algorithm: preferred operators transfer the idea of helpful actions from local search to global best-first search, deferred evaluation of heuristic functions mitigates the negative effect of large branching factors on search performance, and multiheuristic best-first search combines several heuristic evaluation functions within a single search algorithm in an orthogonal way. We also describe efficient data structures for fast state expansion (successor generators and axiom evaluators) and present a new non-heuristic search algorithm called focused iterative-broadening search, which utilizes the information encoded in causal graphs in a novel way. \n \nFast Downward has proven remarkably successful: It won the \"classical\" (i. e., propositional, non-optimising) track of the 4th International Planning Competition at ICAPS 2004, following in the footsteps of planners such as FF and LPG. Our experiments show that it also performs very well on the benchmarks of the earlier planning competitions and provide some insights about the usefulness of the new search enhancements."
                },
                {
                    "title": "MAP estimation via agreement on trees: message-passing and linear programming",
                    "abstract": "We develop and analyze methods for computing provably optimal maximum a posteriori probability (MAP) configurations for a subclass of Markov random fields defined on graphs with cycles. By decomposing the original distribution into a convex combination of tree-structured distributions, we obtain an upper bound on the optimal value of the original problem (i.e., the log probability of the MAP assignment) in terms of the combined optimal values of the tree problems. We prove that this upper bound is tight if and only if all the tree distributions share an optimal configuration in common. An important implication is that any such shared configuration must also be a MAP configuration for the original distribution. Next we develop two approaches to attempting to obtain tight upper bounds: a) a tree-relaxed linear program (LP), which is derived from the Lagrangian dual of the upper bounds; and b) a tree-reweighted max-product message-passing algorithm that is related to but distinct from the max-product algorithm. In this way, we establish a connection between a certain LP relaxation of the mode-finding problem and a reweighted form of the max-product (min-sum) message-passing algorithm."
                },
                {
                    "title": "Constructing free-energy approximations and generalized belief propagation algorithms",
                    "abstract": "Important inference problems in statistical physics, computer vision, error-correcting coding theory, and artificial intelligence can all be reformulated as the computation of marginal probabilities on factor graphs. The belief propagation (BP) algorithm is an efficient way to solve these problems that is exact when the factor graph is a tree, but only approximate when the factor graph has cycles. We show that BP fixed points correspond to the stationary points of the Bethe approximation of the free energy for a factor graph. We explain how to obtain region-based free energy approximations that improve the Bethe approximation, and corresponding generalized belief propagation (GBP) algorithms. We emphasize the conditions a free energy approximation must satisfy in order to be a \"valid\" or \"maxent-normal\" approximation. We describe the relationship between four different methods that can be used to generate valid approximations: the \"Bethe method\", the \"junction graph method\", the \"cluster variation method\", and the \"region graph method\". Finally, we explain how to tell whether a region-based approximation, and its corresponding GBP algorithm, is likely to be accurate, and describe empirical results showing that GBP can significantly outperform BP."
                },
                {
                    "title": "Efficient Solution Algorithms for Factored MDPs",
                    "abstract": "This paper addresses the problem of planning under uncertainty in large Markov Decision Processes (MDPs). Factored MDPs represent a complex state space using state variables and the transition model using a dynamic Bayesian network. This representation often allows an exponential reduction in the representation size of structured MDPs, but the complexity of exact solution algorithms for such MDPs can grow exponentially in the representation size. In this paper, we present two approximate solution algorithms that exploit structure in factored MDPs. Both use an approximate value function represented as a linear combination of basis functions, where each basis function involves only a small subset of the domain variables. A key contribution of this paper is that it shows how the basic operations of both algorithms can be performed efficiently in closed form, by exploiting both additive and context-specific structure in a factored MDP. A central element of our algorithms is a novel linear program decomposition technique, analogous to variable elimination in Bayesian networks, which reduces an exponentially large LP to a provably equivalent, polynomial-sized one. One algorithm uses approximate linear programming, and the second approximate dynamic programming. Our dynamic programming algorithm is novel in that it uses an approximation based on max-norm, a technique that more directly minimizes the terms that appear in error bounds for approximate MDP algorithms. We provide experimental results on problems with over 10^40 states, demonstrating a promising indication of the scalability of our approach, and compare our algorithm to an existing state-of-the-art approach, showing, in some problems, exponential gains in computation time."
                },
                {
                    "title": "Computing Factored Value Functions for Policies in Structured MDPs",
                    "abstract": "Many large Markov decision processes (MDPs) can be represented compactly using a structured representation such as a dynamic Bayesian network. Unfortunately, the compact representation does not help standard MDP algorithms, because the value function for the MDP does not retain the structure of the process description. We argue that in many such MDPs, structure is approximately retained. That is, the value functions are nearly additive: closely approximated by a linear function over factors associated with small subsets of problem features. Based on this idea, we present a convergent, approximate value determination algorithm for structured MDPs. The algorithm maintains an additive value function, alternating dynamic programming steps with steps that project the result back into the restricted space of additive functions. We show that both the dynamic programming and the projection steps can be computed efficiently, despite the fact that the number of states is exponential in the number of state variables."
                },
                {
                    "title": "Constrained Markov Decision Processes",
                    "abstract": "INTRODUCTION Examples of Constrained Dynamic Control Problems On Solution Approaches for CMDPs with Expected Costs Other Types of CMDPs Cost Criteria and Assumptions The Convex Analytical Approach and Occupation Measures Linear Programming and Lagrangian Approach for CMDPs About the Methodology The Structure of the Book PART ONE: FINITE MDPS MARKOV DECISION PROCESSES The Model Cost Criteria and the Constrained Problem Some Notation The Dominance of Markov Policies THE DISCOUNTED COST Occupation Measure and the Primal LP Dynamic Programming and Dual LP: the Unconstrained Case Constrained Control: Lagrangian Approach The Dual LP Number of Randomizations THE EXPECTED AVERAGE COST Occupation Measure and the Primal LP Equivalent Linear Program The Dual Program Number of Randomizations FLOW AND SERVICE CONTROL IN A SINGLE-SERVER QUEUE The Model The Lagrangian The Original Constrained Problem Structure of Randomization and Implementation Issues On Coordination Between Controllers Open Questions PART TWO: INFINITE MDPS MDPS WITH INFINITE STATE AND ACTION SPACES The Model Cost Criteria Mixed Policies, and Topologic Structures The Dominance of Markov Policies Aggregation of States Extra Randomization in the Policies Equivalent Quasi-Markov Model and Quasi-Markov Policies THE TOTAL COST: CLASSIFICATION OF MDPS Transient and Absorbing MDPs MDPs With Uniform Lyapunov Functions Equivalence of MDP With Unbounded and bounded costs Properties of MDPs With Uniform Lyapunov Functions Properties for Fixed Initial Distribution Examples of Uniform Lyapunov Functions Contracting MDPs THE TOTAL COST: OCCUPATION MEASURES AND THE PRIMAL LP Occupation Measure Continuity of Occupation Measures More Properties of MDPs Characterization of Achievable Sets of Occupation Measure Relation Between Cost and Occupation Measure Dominating Classes of Policies Equivalent Linear Program The Dual Program THE TOTAL COST: DYNAMIC AND LINEAR PROGRAMMING Non-Constrained Control: Dynamic and Linear Programming Superharmonic Functions and Linear Programming Set of Achievable Costs Constrained Control: Lagrangian Approach The Dual LP State Truncation A Second LP Approach for Optimal Mixed Policies More on Unbound Costs THE DISCOUNTED COST The Equivalent Total Cost Model Occupation Measure and LP Non-negative Immediate Cost Weak Contracting Assumptions and Lyapunov Functions Example: Flow and Service Control THE EXPECTED AVERAGE COST Occupation Measures Completeness Properties of Stationary Policies Relation Between Cost and Occupation Measure Dominating Classes of Policies Equivalent Linear Program The Dual Program The Contracting Framework Other Conditions for the Uniform Integrability The Case of Uniform Lyapunov Conditions EXPECTED AVERAGE COST: DYNAMIC PROGRAMMING AND LP The Non-Constrained Case: Optimality Inequality Non-Constrained Control: Cost Bounded Below Dynamic Programming and Uniform Lyapunov Function Super-Harmonic Functions and Linear Programming Set of Achievable Costs Constrained Control: Lagrangian Approach The Dual LP A Second LP Approach for Optimal Mixed Policies PART THREE: ASYMPTOTIC METHODS AND APPROXIMATIONS SENSITIVITY ANALYSIS Introduction Approximation of the Values Approximation and Robustness of the Policies CONVERGENCE OF DISCOUNTED CONSTRAINED MDPS Convergence in the Discount Factor Convergence to the Expected Average Cost The Case of Uniform Lyapunov Function CONVERGENCE AS THE HORIZON TENDS TO INFINITY The Discounted Cost The Expected Average Cost: Stationary Policies The Expected Average Cost: General Policies STATE TRUNCATION AND APPROXIMATION The Approximating sets of States Scheme I: the Total Cost Scheme II: the Total Cost Scheme III: the Total Cost The Expected Average Cost Infinite MDPs: on the Number of Randomizations APPENDIX: CONVERGENCE OF PROBABILITY MEASURES REFERENCES LIST OF SYMBOLS AND NOTATION INDEX"
                },
                {
                    "title": "Markov Decision Processes: Discrete Stochastic Dynamic Programming",
                    "abstract": "From the Publisher: \nThe past decade has seen considerable theoretical and applied research on Markov decision processes, as well as the growing use of these models in ecology, economics, communications engineering, and other fields where outcomes are uncertain and sequential decision-making processes are needed. A timely response to this increased activity, Martin L. Puterman's new work provides a uniquely up-to-date, unified, and rigorous treatment of the theoretical, computational, and applied research on Markov decision process models. It discusses all major research directions in the field, highlights many significant applications of Markov decision processes models, and explores numerous important topics that have previously been neglected or given cursory coverage in the literature. Markov Decision Processes focuses primarily on infinite horizon discrete time models and models with discrete time spaces while also examining models with arbitrary state spaces, finite horizon models, and continuous-time discrete state models. The book is organized around optimality criteria, using a common framework centered on the optimality (Bellman) equation for presenting results. The results are presented in a \"theorem-proof\" format and elaborated on through both discussion and examples, including results that are not available in any other book. A two-state Markov decision process model, presented in Chapter 3, is analyzed repeatedly throughout the book and demonstrates many results and algorithms. Markov Decision Processes covers recent research advances in such areas as countable state space models with average reward criterion, constrained models, and models with risk sensitive optimality criteria. It also explores several topics that have received little or no attention in other books, including modified policy iteration, multichain models with average reward criterion, and sensitive optimality. In addition, a Bibliographic Remarks section in each chapter comments on relevant historic"
                },
                {
                    "title": "A Unifying View of Estimation and Control Using Belief Propagation With Application to Path Planning",
                    "abstract": "The use of estimation techniques on stochastic models to solve control problems is an emerging paradigm that falls under the rubric of Active Inference (AI) and Control as Inference (CAI). In this work, we use probability propagation on factor graphs to show that various algorithms proposed in the literature can be seen as specific composition rules in a factor graph. We show how this unified approach, presented both in probability space and in log of the probability space, provides a very general framework that includes the Sum-product, the Max-product, Dynamic programming and mixed Reward/Entropy criteria-based algorithms. The framework also expands algorithmic design options that lead to new smoother or sharper policy distributions. We propose original recursions such as: a generalized Sum/Max-product algorithm, a Smooth Dynamic programming algorithm and a modified versions of the Reward/Entropy algorithm. The discussion is carried over with reference to a path planning problem where the recursions that arise from various cost functions, although they may appear similar in scope, bear noticeable differences. We provide a comprehensive table of composition rules and a comparison through simulations, first on a synthetic small grid with a single goal with obstacles, and then on a grid extrapolated from a real-world scene with multiple goals and a semantic map."
                },
                {
                    "title": "From Stochastic Planning to Marginal MAP",
                    "abstract": "It is well known that the problems of stochastic planning and probabilistic inference are closely related. This paper makes two contributions in this context. The first is to provide an analysis of the recently developed SOGBOFA heuristic planning algorithm that was shown to be effective for problems with large factored state and action spaces. It is shown that SOGBOFA can be seen as a specialized inference algorithm that computes its solutions through a combination of a symbolic variant of belief propagation and gradient ascent. The second contribution is a new solver for Marginal MAP (MMAP) inference. We introduce a new reduction from MMAP to maximum expected utility problems which are suitable for the symbolic computation in SOGBOFA. This yields a novel algebraic gradient-based solver (AGS) for MMAP. An experimental evaluation illustrates the potential of AGS in solving difficult MMAP problems."
                },
                {
                    "title": "AD3: alternating directions dual decomposition for MAP inference in graphical models",
                    "abstract": "We present AD3, a new algorithm for approximate maximum a posteriori (MAP) inference on factor graphs, based on the alternating directions method of multipliers. Like other dual decomposition algorithms, AD3 has a modular architecture, where local subproblems are solved independently, and their solutions are gathered to compute a global update. The key characteristic of AD3 is that each local subproblem has a quadratic regularizer, leading to faster convergence, both theoretically and in practice. We provide closed-form solutions for these AD3 subproblems for binary pairwise factors and factors imposing first-order logic constraints. For arbitrary factors (large or combinatorial), we introduce an active set method which requires only an oracle for computing a local MAP configuration, making AD3 applicable to a wide range of problems. Experiments on synthetic and real-world problems show that AD3 compares favorably with the state-of-the-art."
                },
                {
                    "title": "1 Introduction to Dual Decomposition for Inference",
                    "abstract": "Many inference problems with discrete variables result in a difficult com-binatorial optimization problem. In recent years, the technique of dual decomposition , also called Lagrangian relaxation, has proven to be a powerful means of solving these inference problems by decomposing them into simpler components that are repeatedly solved independently and combined into a global solution. In this chapter, we introduce the general technique of dual decomposition through its application to the problem of finding the most likely (MAP) assignment in Markov random fields. We discuss both subgradient and block coordinate descent approaches to solving the dual problem. The resulting message-passing algorithms are similar to max-product, but can be shown to solve a linear programming relaxation of the MAP problem. We show how many of the MAP algorithms are related to each other, and also quantify when the MAP solution can and cannot be decoded directly from the dual solution. 1.1 Introduction Many problems in engineering and the sciences require solutions to challenging combinatorial optimization problems. These include traditional problems such as scheduling, planning, fault diagnosis, or searching for molecular conformations. In addition, a wealth of combinatorial problems arise directly from probabilistic modeling (graphical models). Graphical models (see Koller and Friedman, 2009, for a textbook introduction) have been widely adopted in areas such as computational biology, machine vision, and natural language processing, and are increasingly being used as a framework for expressing combinatorial problems. Consider, for example, a protein side-chain placement problem where the goal is to find the minimum energy conformation of amino acid side-chains along a fixed carbon backbone. The orientations of the side-chains are represented by discretized angles called rotamers. The combinatorial difficulty arises here from the fact that rotamer choices for nearby amino acids are energetically coupled. For globular proteins, for example, such couplings may be present for most pairs of side-chain orientations. This problem is couched in probabilistic modeling terms by associating molecular conformations with the setting of discrete random variables corresponding to the rotamer angles. The interactions between such random variables come from the energetic couplings between nearby amino acids. Finding the minimum energy conformation is then equivalently solved by finding the most probable assignment of states to the variables. We will consider here combinatorial problems that are expressed in terms of structured probability models (graphical models). A graphical model is defined over a set of discrete variables x = {x j } j\u2208V. Local \u2026"
                },
                {
                    "title": "Planning by Probabilistic Inference",
                    "abstract": "This paper presents and demonstrates a new approach to the problem of planning under uncertainty. Actions are treated as hidden variables, with their own prior distributions, in a probabilistic generative model involving actions and states. Planning is done by computing the posterior distribution over actions, conditioned on reaching the goal state within a specified number of steps. Under the new formulation, the toolbox of inference techniques be brought to bear on the planning problem. This paper focuses on problems with discrete actions and states, and discusses some extensions."
                },
                {
                    "title": "Stochastic Finance: An Introduction in Discrete Time",
                    "abstract": "This book is an introduction to financial mathematics. It is intended for graduate students in mathematics and for researchers working in academia and industry. The focus on stochastic models in discrete time has two immediate benefits. First, the probabilistic machinery is simpler, and one can discuss right away some of the key problems in the theory of pricing and hedging of financial derivatives. Second, the paradigm of a complete financial market, where all derivatives admit a perfect hedge, becomes the exception rather than the rule. Thus, the need to confront the intrinsic risks arising from market incomleteness appears at a very early stage. The first part of the book contains a study of a simple one-period model, which also serves as a building block for later developments. Topics include the characterization of arbitrage-free markets, preferences on asset profiles, an introduction to equilibrium analysis, and monetary measures of financial risk. In the second part, the idea of dynamic hedging of contingent claims is developed in a multiperiod framework. Topics include martingale measures, pricing formulas for derivatives, American options, superhedging, and hedging strategies with minimal shortfall risk. This third revised and extended edition now contains more than one hundred exercises. It also includes new material on risk measures and the related issue of model uncertainty, in particular a new chapter on dynamic risk measures and new sections on robust utility maximization and on efficient hedging with convex risk measures."
                }
            ],
            "categories": [
                "cs.AI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively formulate and solve planning inference in Markov Decision Processes (MDPs) under stochastic dynamics, distinguishing it from traditional probabilistic inference methods?\n\n### [Question 2] - Why is it interesting and important?\nSolving the problem of planning inference in MDPs has significant implications for the research community, as it provides a clearer understanding of the relationship between different types of inference and planning. By establishing a framework that ranks inference methods based on their effectiveness in planning, this research could lead to advancements in reinforcement learning, robotics, and decision-making systems. Furthermore, the development of approximate planning inference methods could enable practical applications in complex environments where traditional methods are computationally infeasible, thus broadening the scope of MDP applications in real-world scenarios.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the inherent complexity of stochastic dynamics in MDPs, where traditional inference methods like marginal or MAP do not adequately capture the nuances of planning. Naive approaches may fail because they do not account for the dynamic nature of the environment or the need for a policy that adapts over time. Additionally, the exponential growth of the state space in factored MDPs presents significant computational obstacles, making exact solutions impractical. Overcoming these challenges requires innovative algorithms that can efficiently approximate planning inference while maintaining accuracy.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often treated planning as a subset of probabilistic inference without recognizing its distinct characteristics, leading to a lack of tailored methodologies. Existing solutions have been limited by their inability to handle the complexities of factored MDPs and the stochastic nature of the dynamics involved. Barriers such as the computational intractability of exact solutions and the inadequacy of traditional inference methods have prevented progress in this area. This research proposes a novel approach that leverages variational inference to create a framework specifically designed for planning inference, thus addressing the limitations of prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nThe proposed methodology involves using variational inference to develop a framework for planning inference in MDPs. This includes applying an analogue of loopy belief propagation (LBP) to approximate planning inference in factored MDPs with large state spaces. The dataset will consist of various MDP configurations, and the performance will be evaluated using metrics such as the quality of the resulting policies and the computational efficiency of the inference methods. The"
            }
        },
        "author_data": {
            "6ae5eccc-3486-4730-a82c-f76064094650": {
                "pk": "6ae5eccc-3486-4730-a82c-f76064094650",
                "name": "Miguel L\u00e1zaro-Gredilla",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "1176dc28-32aa-42ec-8e05-d91712f862c2": {
                "pk": "1176dc28-32aa-42ec-8e05-d91712f862c2",
                "name": "Li Yang Ku",
                "collaborators": [
                    "Erik Learned-Miller",
                    "Rod Grupen"
                ],
                "domain": [
                    "Robotics",
                    "Computer Vision",
                    "Deep Learning",
                    "Grasping"
                ],
                "publications": [
                    {
                        "title": "Associating Grasp Configurations with Hierarchical Features in Convolutional Neural Networks",
                        "abstract": "In this work, we provide a solution for posturing the anthropomorphic Robonaut-2 hand and arm for grasping based on visual information. A mapping from visual features extracted from a convolutional neural network (CNN) to grasp points is learned. We demonstrate that a CNN pre-trained for image classification can be applied to a grasping task based on a small set of grasping examples. Our approach takes advantage of the hierarchical nature of the CNN by identifying features that capture the hierarchical support relations between filters in different CNN layers and locating their 3D positions by tracing activations backwards in the CNN. When this backward trace terminates in the RGB-D image, important manipulable structures are thereby localized. These features that reside in different layers of the CNN are then associated with controllers that engage different kinematic subchains in the hand/arm system for grasping. A grasping dataset is collected using demonstrated hand/object relationships for Robonaut-2 to evaluate the proposed approach in terms of the precision of the resulting preshape postures. We demonstrate that this approach outperforms baseline approaches in cluttered scenarios on the grasping dataset and a point cloud based approach on a grasping task using Robonaut-2."
                    }
                ]
            },
            "84668655-1738-4192-8c54-0fb32ffff18e": {
                "pk": "84668655-1738-4192-8c54-0fb32ffff18e",
                "name": "Kevin P. Murphy",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "0a4efd49-eb12-43b6-a35d-fe813d5ddf5e": {
                "pk": "0a4efd49-eb12-43b6-a35d-fe813d5ddf5e",
                "name": "Dileep George",
                "collaborators": [
                    "Miguel L\u00e1zaro-Gredilla",
                    "Antoine Dedieu",
                    "Miguel Lazaro-Gredilla",
                    "Wolfgang Lehrach",
                    "Guangyao Zhou",
                    "Nishad Gothoskar",
                    "Yi Liu",
                    "D. Scott Phoenix",
                    "J. Swaroop Guntupalli",
                    "Huayan Wang"
                ],
                "domain": [
                    "Probabilistic Graphical Models",
                    "Reinforcement Learning",
                    "Cognitive Robotics",
                    "Variational Inference"
                ],
                "publications": [
                    {
                        "title": "What can the brain teach us about building artificial intelligence?",
                        "abstract": "This paper is the preprint of an invited commentary on Lake et al's Behavioral and Brain Sciences article titled \"Building machines that learn and think like people\". Lake et al's paper offers a timely critique on the recent accomplishments in artificial intelligence from the vantage point of human intelligence, and provides insightful suggestions about research directions for building more human-like intelligence. Since we agree with most of the points raised in that paper, we will offer a few points that are complementary."
                    },
                    {
                        "title": "A Model of Fast Concept Inference with Object-Factorized Cognitive Programs",
                        "abstract": "The ability of humans to quickly identify general concepts from a handful of images has proven difficult to emulate with robots. Recently, a computer architecture was developed that allows robots to mimic some aspects of this human ability by modeling concepts as cognitive programs using an instruction set of primitive cognitive functions. This allowed a robot to emulate human imagination by simulating candidate programs in a world model before generalizing to the physical world. However, this model used a naive search algorithm that required 30 minutes to discover a single concept, and became intractable for programs with more than 20 instructions. To circumvent this bottleneck, we present an algorithm that emulates the human cognitive heuristics of object factorization and sub-goaling, allowing human-level inference speed, improving accuracy, and making the output more explainable."
                    },
                    {
                        "title": "Sample-Efficient L0-L2 Constrained Structure Learning of Sparse Ising Models",
                        "abstract": "We consider the problem of learning the underlying graph of a sparse Ising model with $p$ nodes from $n$ i.i.d. samples. The most recent and best performing approaches combine an empirical loss (the logistic regression loss or the interaction screening loss) with a regularizer (an L1 penalty or an L1 constraint). This results in a convex problem that can be solved separately for each node of the graph. In this work, we leverage the cardinality constraint L0 norm, which is known to properly induce sparsity, and further combine it with an L2 norm to better model the non-zero coefficients. We show that our proposed estimators achieve an improved sample complexity, both (a) theoretically, by reaching new state-of-the-art upper bounds for recovery guarantees, and (b) empirically, by showing sharper phase transitions between poor and full recovery for graph topologies studied in the literature, when compared to their L1-based state-of-the-art methods."
                    },
                    {
                        "title": "Learning undirected models via query training",
                        "abstract": "Typical amortized inference in variational autoencoders is specialized for a single probabilistic query. Here we propose an inference network architecture that generalizes to unseen probabilistic queries. Instead of an encoder-decoder pair, we can train a single inference network directly from data, using a cost function that is stochastic not only over samples, but also over queries. We can use this network to perform the same inference tasks as we would in an undirected graphical model with hidden variables, without having to deal with the intractable partition function. The results can be mapped to the learning of an actual undirected model, which is a notoriously hard problem. Our network also marginalizes nuisance variables as required. We show that our approach generalizes to unseen probabilistic queries on also unseen test data, providing fast and flexible inference. Experiments show that this approach outperforms or matches PCD and AdVIL on 9 benchmark datasets."
                    },
                    {
                        "title": "Perturb-and-max-product: Sampling and learning in discrete energy-based models",
                        "abstract": "Perturb-and-MAP offers an elegant approach to approximately sample from a energy-based model (EBM) by computing the maximum-a-posteriori (MAP) configuration of a perturbed version of the model. Sampling in turn enables learning. However, this line of research has been hindered by the general intractability of the MAP computation. Very few works venture outside tractable models, and when they do, they use linear programming approaches, which as we will show, have several limitations. In this work we present perturb-and-max-product (PMP), a parallel and scalable mechanism for sampling and learning in discrete EBMs. Models can be arbitrary as long as they are built using tractable factors. We show that (a) for Ising models, PMP is orders of magnitude faster than Gibbs and Gibbs-with-Gradients (GWG) at learning and generating samples of similar or better quality; (b) PMP is able to learn and sample from RBMs; (c) in a large, entangled graphical model in which Gibbs and GWG fail to mix, PMP succeeds."
                    },
                    {
                        "title": "From proprioception to long-horizon planning in novel environments: A hierarchical RL model",
                        "abstract": "For an intelligent agent to flexibly and efficiently operate in complex environments, they must be able to reason at multiple levels of temporal, spatial, and conceptual abstraction. At the lower levels, the agent must interpret their proprioceptive inputs and control their muscles, and at the higher levels, the agent must select goals and plan how they will achieve those goals. It is clear that each of these types of reasoning is amenable to different types of representations, algorithms, and inputs. In this work, we introduce a simple, three-level hierarchical architecture that reflects these distinctions. The low-level controller operates on the continuous proprioceptive inputs, using model-free learning to acquire useful behaviors. These in turn induce a set of mid-level dynamics, which are learned by the mid-level controller and used for model-predictive control, to select a behavior to activate at each timestep. The high-level controller leverages a discrete, graph representation for goal selection and path planning to specify targets for the mid-level controller. We apply our method to a series of navigation tasks in the Mujoco Ant environment, consistently demonstrating significant improvements in sample-efficiency compared to prior model-free, model-based, and hierarchical RL methods. Finally, as an illustrative example of the advantages of our architecture, we apply our method to a complex maze environment that requires efficient exploration and long-horizon planning."
                    },
                    {
                        "title": "Learning noisy-OR Bayesian Networks with Max-Product Belief Propagation",
                        "abstract": "Noisy-OR Bayesian Networks (BNs) are a family of probabilistic graphical models which express rich statistical dependencies in binary data. Variational inference (VI) has been the main method proposed to learn noisy-OR BNs with complex latent structures (Jaakkola & Jordan, 1999; Ji et al., 2020; Buhai et al., 2020). However, the proposed VI approaches either (a) use a recognition network with standard amortized inference that cannot induce ``explaining-away''; or (b) assume a simple mean-field (MF) posterior which is vulnerable to bad local optima. Existing MF VI methods also update the MF parameters sequentially which makes them inherently slow. In this paper, we propose parallel max-product as an alternative algorithm for learning noisy-OR BNs with complex latent structures and we derive a fast stochastic training scheme that scales to large datasets. We evaluate both approaches on several benchmarks where VI is the state-of-the-art and show that our method (a) achieves better test performance than Ji et al. (2020) for learning noisy-OR BNs with hierarchical latent structures on large sparse real datasets; (b) recovers a higher number of ground truth parameters than Buhai et al. (2020) from cluttered synthetic scenes; and (c) solves the 2D blind deconvolution problem from Lazaro-Gredilla et al. (2021) and variant - including binary matrix factorization - while VI catastrophically fails and is up to two orders of magnitude slower."
                    },
                    {
                        "title": "Hierarchical compositional feature learning",
                        "abstract": "We introduce the hierarchical compositional network (HCN), a directed generative model able to discover and disentangle, without supervision, the building blocks of a set of binary images. The building blocks are binary features defined hierarchically as a composition of some of the features in the layer immediately below, arranged in a particular manner. At a high level, HCN is similar to a sigmoid belief network with pooling. Inference and learning in HCN are very challenging and existing variational approximations do not work satisfactorily. A main contribution of this work is to show that both can be addressed using max-product message passing (MPMP) with a particular schedule (no EM required). Also, using MPMP as an inference engine for HCN makes new tasks simple: adding supervision information, classifying images, or performing inpainting all correspond to clamping some variables of the model to their known values and running MPMP on the rest. When used for classification, fast inference with HCN has exactly the same functional form as a convolutional neural network (CNN) with linear activations and binary weights. However, HCN's features are qualitatively very different."
                    },
                    {
                        "title": "Learning a generative model for robot control using visual feedback",
                        "abstract": "We introduce a novel formulation for incorporating visual feedback in controlling robots. We define a generative model from actions to image observations of features on the end-effector. Inference in the model allows us to infer the robot state corresponding to target locations of the features. This, in turn, guides motion of the robot and allows for matching the target locations of the features in significantly fewer steps than state-of-the-art visual servoing methods. The training procedure for our model enables effective learning of the kinematics, feature structure, and camera parameters, simultaneously. This can be done with no prior information about the robot, structure, and cameras that observe it. Learning is done sample-efficiently and shows strong generalization to test data. Since our formulation is modular, we can modify components of our setup, like cameras and objects, and relearn them quickly online. Our method can handle noise in the observed state and noise in the controllers that we interact with. We demonstrate the effectiveness of our method by executing grasping and tight-fit insertions on robots with inaccurate controllers."
                    },
                    {
                        "title": "Fast exploration and learning of latent graphs with aliased observations",
                        "abstract": "We consider the problem of recovering a latent graph where the observations at each node are \\emph{aliased}, and transitions are stochastic. Observations are gathered by an agent traversing the graph. Aliasing means that multiple nodes emit the same observation, so the agent can not know in which node it is located. The agent needs to uncover the hidden topology as accurately as possible and in as few steps as possible. This is equivalent to efficient recovery of the transition probabilities of a partially observable Markov decision process (POMDP) in which the observation probabilities are known. An algorithm for efficiently exploring (and ultimately recovering) the latent graph is provided. Our approach is exponentially faster than naive exploration in a variety of challenging topologies with aliased observations while remaining competitive with existing baselines in the unaliased regime."
                    },
                    {
                        "title": "Beyond imitation: Zero-shot task transfer on robots by learning concepts as cognitive programs",
                        "abstract": "Humans can infer concepts from image pairs and apply those in the physical world in a completely different setting, enabling tasks like IKEA assembly from diagrams. If robots could represent and infer high-level concepts, it would significantly improve their ability to understand our intent and to transfer tasks between different environments. To that end, we introduce a computational framework that replicates aspects of human concept learning. Concepts are represented as programs on a novel computer architecture consisting of a visual perception system, working memory, and action controller. The instruction set of this \"cognitive computer\" has commands for parsing a visual scene, directing gaze and attention, imagining new objects, manipulating the contents of a visual working memory, and controlling arm movement. Inferring a concept corresponds to inducing a program that can transform the input to the output. Some concepts require the use of imagination and recursion. Previously learned concepts simplify the learning of subsequent more elaborate concepts, and create a hierarchy of abstractions. We demonstrate how a robot can use these abstractions to interpret novel concepts presented to it as schematic images, and then apply those concepts in dramatically different situations. By bringing cognitive science ideas on mental imagery, perceptual symbols, embodied cognition, and deictic mechanisms into the realm of machine learning, our work brings us closer to the goal of building robots that have interpretable representations and commonsense."
                    },
                    {
                        "title": "Graphical Models with Attention for Context-Specific Independence and an Application to Perceptual Grouping",
                        "abstract": "Discrete undirected graphical models, also known as Markov Random Fields (MRFs), can flexibly encode probabilistic interactions of multiple variables, and have enjoyed successful applications to a wide range of problems. However, a well-known yet little studied limitation of discrete MRFs is that they cannot capture context-specific independence (CSI). Existing methods require carefully developed theories and purpose-built inference methods, which limit their applications to only small-scale problems. In this paper, we propose the Markov Attention Model (MAM), a family of discrete MRFs that incorporates an attention mechanism. The attention mechanism allows variables to dynamically attend to some other variables while ignoring the rest, and enables capturing of CSIs in MRFs. A MAM is formulated as an MRF, allowing it to benefit from the rich set of existing MRF inference methods and scale to large models and datasets. To demonstrate MAM's capabilities to capture CSIs at scale, we apply MAMs to capture an important type of CSI that is present in a symbolic approach to recurrent computations in perceptual grouping. Experiments on two recently proposed synthetic perceptual grouping tasks and on realistic images demonstrate the advantages of MAMs in sample-efficiency, interpretability and generalizability when compared with strong recurrent neural network baselines, and validate MAM's capabilities to efficiently capture CSIs at scale."
                    },
                    {
                        "title": "Learning Cognitive Maps from Transformer Representations for Efficient Planning in Partially Observed Environments",
                        "abstract": "Despite their stellar performance on a wide range of tasks, including in-context tasks only revealed during inference, vanilla transformers and variants trained for next-token predictions (a) do not learn an explicit world model of their environment which can be flexibly queried and (b) cannot be used for planning or navigation. In this paper, we consider partially observed environments (POEs), where an agent receives perceptually aliased observations as it navigates, which makes path planning hard. We introduce a transformer with (multiple) discrete bottleneck(s), TDB, whose latent codes learn a compressed representation of the history of observations and actions. After training a TDB to predict the future observation(s) given the history, we extract interpretable cognitive maps of the environment from its active bottleneck(s) indices. These maps are then paired with an external solver to solve (constrained) path planning problems. First, we show that a TDB trained on POEs (a) retains the near perfect predictive performance of a vanilla transformer or an LSTM while (b) solving shortest path problems exponentially faster. Second, a TDB extracts interpretable representations from text datasets, while reaching higher in-context accuracy than vanilla sequence models. Finally, in new POEs, a TDB (a) reaches near-perfect in-context accuracy, (b) learns accurate in-context cognitive maps (c) solves in-context path planning problems."
                    },
                    {
                        "title": "Cortical Microcircuits from a Generative Vision Model",
                        "abstract": "Understanding the information processing roles of cortical circuits is an outstanding problem in neuroscience and artificial intelligence. The theoretical setting of Bayesian inference has been suggested as a framework for understanding cortical computation. Based on a recently published generative model for visual inference (George et al., 2017), we derive a family of anatomically instantiated and functional cortical circuit models. In contrast to simplistic models of Bayesian inference, the underlying generative model's representational choices are validated with real-world tasks that required efficient inference and strong generalization. The cortical circuit model is derived by systematically comparing the computational requirements of this model with known anatomical constraints. The derived model suggests precise functional roles for the feedforward, feedback and lateral connections observed in different laminae and columns, and assigns a computational role for the path through the thalamus."
                    },
                    {
                        "title": "Teaching Compositionality to CNNs",
                        "abstract": "Convolutional neural networks (CNNs) have shown great success in computer vision, approaching human-level performance when trained for specific tasks via application-specific loss functions. In this paper, we propose a method for augmenting and training CNNs so that their learned features are compositional. It encourages networks to form representations that disentangle objects from their surroundings and from each other, thereby promoting better generalization. Our method is agnostic to the specific details of the underlying CNN to which it is applied and can in principle be used with any CNN. As we show in our experiments, the learned representations lead to feature activations that are more localized and improve performance over non-compositional baselines in object recognition tasks."
                    },
                    {
                        "title": "A backward pass through a CNN using a generative model of its activations",
                        "abstract": "Neural networks have shown to be a practical way of building a very complex mapping between a pre-specified input space and output space. For example, a convolutional neural network (CNN) mapping an image into one of a thousand object labels is approaching human performance in this particular task. However the mapping (neural network) does not automatically lend itself to other forms of queries, for example, to detect/reconstruct object instances, to enforce top-down signal on ambiguous inputs, or to recover object instances from occlusion. One way to address these queries is a backward pass through the network that fuses top-down and bottom-up information. In this paper, we show a way of building such a backward pass by defining a generative model of the neural network's activations. Approximate inference of the model would naturally take the form of a backward pass through the CNN layers, and it addresses the aforementioned queries in a unified framework."
                    },
                    {
                        "title": "PGMax: Factor Graphs for Discrete Probabilistic Graphical Models and Loopy Belief Propagation in JAX",
                        "abstract": "PGMax is an open-source Python package for (a) easily specifying discrete Probabilistic Graphical Models (PGMs) as factor graphs; and (b) automatically running efficient and scalable loopy belief propagation (LBP) in JAX. PGMax supports general factor graphs with tractable factors, and leverages modern accelerators like GPUs for inference. Compared with existing alternatives, PGMax obtains higher-quality inference results with up to three orders-of-magnitude inference time speedups. PGMax additionally interacts seamlessly with the rapidly growing JAX ecosystem, opening up new research possibilities. Our source code, examples and documentation are available at https://github.com/deepmind/PGMax."
                    },
                    {
                        "title": "PushWorld: A benchmark for manipulation planning with tools and movable obstacles",
                        "abstract": "While recent advances in artificial intelligence have achieved human-level performance in environments like Starcraft and Go, many physical reasoning tasks remain challenging for modern algorithms. To date, few algorithms have been evaluated on physical tasks that involve manipulating objects when movable obstacles are present and when tools must be used to perform the manipulation. To promote research on such tasks, we introduce PushWorld, an environment with simplistic physics that requires manipulation planning with both movable obstacles and tools. We provide a benchmark of more than 200 PushWorld puzzles in PDDL and in an OpenAI Gym environment. We evaluate state-of-the-art classical planning and reinforcement learning algorithms on this benchmark, and we find that these baseline results are below human-level performance. We then provide a new classical planning heuristic that solves the most puzzles among the baselines, and although it is 40 times faster than the best baseline planner, it remains below human-level performance."
                    },
                    {
                        "title": "Learning higher-order sequential structure with cloned HMMs",
                        "abstract": "Variable order sequence modeling is an important problem in artificial and natural intelligence. While overcomplete Hidden Markov Models (HMMs), in theory, have the capacity to represent long-term temporal structure, they often fail to learn and converge to local minima. We show that by constraining HMMs with a simple sparsity structure inspired by biology, we can make it learn variable order sequences efficiently. We call this model cloned HMM (CHMM) because the sparsity structure enforces that many hidden states map deterministically to the same emission state. CHMMs with over 1 billion parameters can be efficiently trained on GPUs without being severely affected by the credit diffusion problem of standard HMMs. Unlike n-grams and sequence memoizers, CHMMs can model temporal dependencies at arbitrarily long distances and recognize contexts with 'holes' in them. Compared to Recurrent Neural Networks and their Long Short-Term Memory extensions (LSTMs), CHMMs are generative models that can natively deal with uncertainty. Moreover, CHMMs return a higher-order graph that represents the temporal structure of the data which can be useful for community detection, and for building hierarchical models. Our experiments show that CHMMs can beat n-grams, sequence memoizers, and LSTMs on character-level language modeling tasks. CHMMs can be a viable alternative to these methods in some tasks that require variable order sequence modeling and the handling of uncertainty."
                    },
                    {
                        "title": "Query Training: Learning a Worse Model to Infer Better Marginals in Undirected Graphical Models with Hidden Variables",
                        "abstract": "Probabilistic graphical models (PGMs) provide a compact representation of knowledge that can be queried in a flexible way: after learning the parameters of a graphical model once, new probabilistic queries can be answered at test time without retraining. However, when using undirected PGMS with hidden variables, two sources of error typically compound in all but the simplest models (a) learning error (both computing the partition function and integrating out the hidden variables is intractable); and (b) prediction error (exact inference is also intractable). Here we introduce query training (QT), a mechanism to learn a PGM that is optimized for the approximate inference algorithm that will be paired with it. The resulting PGM is a worse model of the data (as measured by the likelihood), but it is tuned to produce better marginals for a given inference algorithm. Unlike prior works, our approach preserves the querying flexibility of the original PGM: at test time, we can estimate the marginal of any variable given any partial evidence. We demonstrate experimentally that QT can be used to learn a challenging 8-connected grid Markov random field with hidden variables and that it consistently outperforms the state-of-the-art AdVIL when tested on three undirected models across multiple datasets."
                    }
                ]
            }
        }
    },
    "2311.13628": {
        "paper_data": {
            "title": "Prompt Risk Control: A Rigorous Framework for Responsible Deployment of Large Language Models",
            "url": "http://arxiv.org/abs/2311.13628v2",
            "arxiv_id": "2311.13628",
            "authors": [
                "Thomas P. Zollo",
                "Todd Morrill",
                "Zhun Deng",
                "Jake C. Snell",
                "Toniann Pitassi",
                "Richard Zemel"
            ],
            "abstract": "The recent explosion in the capabilities of large language models has led to a wave of interest in how best to prompt a model to perform a given task. While it may be tempting to simply choose a prompt based on average performance on a validation set, this can lead to a deployment where unexpectedly poor responses are generated, especially for the worst-off users. To mitigate this prospect, we propose Prompt Risk Control, a lightweight framework for selecting a prompt based on rigorous upper bounds on families of informative risk measures. We offer methods for producing bounds on a diverse set of metrics, including quantities that measure worst-case responses and disparities in generation quality across the population of users. In addition, we extend the underlying statistical bounding techniques to accommodate the possibility of distribution shifts in deployment. Experiments on applications such as open-ended chat, medical question summarization, and code generation highlight how such a framework can foster responsible deployment by reducing the risk of the worst outcomes.",
            "introduction": "   1 Introduction  Figure 1:  Prompt Risk Control (PRC) assists in choosing a prompt (or set of prompts) that will, with high likelihood, not incur too high of a loss according to some chosen risk measure and threshold. Here we illustrate PRC being used to select a system prompt to be appended to input queries to a chatbot, a popular setup in modern LLM deployments (algorithm inputs are in grey). The goal is to ensure that the responses will not be too toxic for the highest-loss (most toxic) portion of the data distribution (e.g., under the CVaR risk measure). The algorithm returns a set of prompts that bound the risk at an acceptable level, from which a user can select a safe prompt for deployment.    Recent leaps in the capabilities of large language models (LLMs) such as GPT-4 (OpenAI, 2023), LLaMA (Touvron et\u00a0al., 2023), and Claude have driven a wave of interest in constructing the best prompt for a given task, where a prompt generally refers to an input to the LLM. Various prompting strategies have been proposed, including but not limited to: in-context learning (Brown et\u00a0al., 2020), instruction following (Wei et\u00a0al., 2022a), chain-of-thought prompting (Wei et\u00a0al., 2022b), and prompt-tuning (Lester et\u00a0al., 2021), as well as a range of more complex approaches. Despite this proliferation of methods and their suggested strengths, prompting remains an experimental and poorly understood area, with little clear evidence why one task verbalization or a particular ordering of few-shot exemplars should improve performance (Kaddour et\u00a0al., 2023; Webson and Pavlick, 2022). Lacking a rigorous understanding of the underlying mechanisms, prompt choices are usually made based on empirical average results on a validation set (Burnell et\u00a0al., 2023). However, a prompt that performs well on average on a validation set may in fact be prone to producing some poor generations in deployment with an unacceptably high probability, since a single validation score lacks information about the underlying variance or likelihood of outlier events. For example, when deploying an open-ended chatbot, one may find that the prompt that produces the most helpful generations on a validation set also produces unacceptably high toxicity for some portion of users in deployment. This potential trade-off between usefulness and safety (or helpfulness and harmlessness) is an area of increasing interest and importance, both in the context of prompting as well as under the various fine-tuning alignment methods that are applied to models before deployment (Bai et\u00a0al., 2022; Ganguli et\u00a0al., 2022).   To mitigate this prospect of unexpectedly bad outcomes in LLM deployment and manage these trade-offs in a principled way, we introduce Prompt Risk Control (PRC), a framework for selecting a prompt based on rigorous upper bounds on some user-chosen risk measure. Our framework employs statistically and theoretically sound methods from the Distribution-Free Uncertainty Quantification (DFUQ) family of techniques (Vovk et\u00a0al., 2005; Bates et\u00a0al., 2021; Angelopoulos and Bates, 2021; Snell et\u00a0al., 2023; Deng et\u00a0al., 2023) in order to control (i.e., produce bounds on) a rich set of informative risk measures, and uses these bounds to return a set of prompts that with high probability will not incur an unacceptable outcome according to some user-chosen criteria (see Figure 1). PRC can be applied",
            "references": [
                {
                    "title": "Estimating and Implementing Conventional Fairness Metrics With Probabilistic Protected Features",
                    "abstract": "The vast majority of techniques to train fair models require access to the protected attribute (e.g., race, gender), either at train time or in production. However, in many practically important applications, this protected attribute is largely unavailable. Still, AI systems used in sensitive business and government applications \u2014 such as housing, ad delivery, and credit underwriting \u2014 are increasingly required by law to measure and mitigate their bias. In this paper, we develop methods for measuring and reducing fairness violations in a setting with limited access to protected attribute labels. Specifically, we assume access to protected attribute labels on a small subset of the dataset of interest, but only probabilistic estimates of protected attribute labels (e.g., via Bayesian Improved Surname Geocoding) for the rest of the dataset. With this setting in mind, we propose a method to estimate bounds on common fairness metrics for an existing model, as well as a method for training a model to limit fairness violations by solving a constrained non-convex optimization problem. Unlike existing approaches, our methods take advantage of contextual information \u2013 specifically the relationships between a model\u2019s predictions and the probabilistic prediction of protected attributes, given the true protected attribute, and vice versa \u2013 to provide tighter bounds on the true disparity. We provide an empirical illustration of our methods using voting data as well as the COMPAS dataset. First, we show that our measurement method can bound the true disparity up to 5.5x tighter than previous methods in these applications. Then, we demonstrate that our training technique effectively reduces disparity in comparison to an unconstrained model while often incurring less severe fairness-accuracy trade-offs than other fair optimization methods with limited access to protected attributes."
                },
                {
                    "title": "Distribution-Free Statistical Dispersion Control for Societal Applications",
                    "abstract": "Explicit finite-sample statistical guarantees on model performance are an important ingredient in responsible machine learning. Previous work has focused mainly on bounding either the expected loss of a predictor or the probability that an individual prediction will incur a loss value in a specified range. However, for many high-stakes applications, it is crucial to understand and control the dispersion of a loss distribution, or the extent to which different members of a population experience unequal effects of algorithmic decisions. We initiate the study of distribution-free control of statistical dispersion measures with societal implications and propose a simple yet flexible framework that allows us to handle a much richer class of statistical functionals beyond previous work. Our methods are verified through experiments in toxic comment detection, medical imaging, and film recommendation."
                },
                {
                    "title": "Challenges and Applications of Large Language Models",
                    "abstract": "Large Language Models (LLMs) went from non-existent to ubiquitous in the machine learning discourse within a few years. Due to the fast pace of the field, it is difficult to identify the remaining challenges and already fruitful application areas. In this paper, we aim to establish a systematic set of open problems and application successes so that ML researchers can comprehend the field's current state more quickly and become productive."
                },
                {
                    "title": "Robots That Ask For Help: Uncertainty Alignment for Large Language Model Planners",
                    "abstract": "Large language models (LLMs) exhibit a wide range of promising capabilities -- from step-by-step planning to commonsense reasoning -- that may provide utility for robots, but remain prone to confidently hallucinated predictions. In this work, we present KnowNo, which is a framework for measuring and aligning the uncertainty of LLM-based planners such that they know when they don't know and ask for help when needed. KnowNo builds on the theory of conformal prediction to provide statistical guarantees on task completion while minimizing human help in complex multi-step planning settings. Experiments across a variety of simulated and real robot setups that involve tasks with different modes of ambiguity (e.g., from spatial to numeric uncertainties, from human preferences to Winograd schemas) show that KnowNo performs favorably over modern baselines (which may involve ensembles or extensive prompt tuning) in terms of improving efficiency and autonomy, while providing formal assurances. KnowNo can be used with LLMs out of the box without model-finetuning, and suggests a promising lightweight approach to modeling uncertainty that can complement and scale with the growing capabilities of foundation models. Website: https://robot-help.github.io"
                },
                {
                    "title": "Conformal Language Modeling",
                    "abstract": "We propose a novel approach to conformal prediction for generative language models (LMs). Standard conformal prediction produces prediction sets -- in place of single predictions -- that have rigorous, statistical performance guarantees. LM responses are typically sampled from the model's predicted distribution over the large, combinatorial output space of natural language. Translating this process to conformal prediction, we calibrate a stopping rule for sampling different outputs from the LM that get added to a growing set of candidates until we are confident that the output set is sufficient. Since some samples may be low-quality, we also simultaneously calibrate and apply a rejection rule for removing candidates from the output set to reduce noise. Similar to conformal prediction, we prove that the sampled set returned by our procedure contains at least one acceptable answer with high probability, while still being empirically precise (i.e., small) on average. Furthermore, within this set of candidate responses, we show that we can also accurately identify subsets of individual components -- such as phrases or sentences -- that are each independently correct (e.g., that are not\"hallucinations\"), again with statistical guarantees. We demonstrate the promise of our approach on multiple tasks in open-domain question answering, text summarization, and radiology report generation using different LM variants."
                },
                {
                    "title": "Conformal Prediction with Large Language Models for Multi-Choice Question Answering",
                    "abstract": "As large language models continue to be widely developed, robust uncertainty quantification techniques will become crucial for their safe deployment in high-stakes scenarios. In this work, we explore how conformal prediction can be used to provide uncertainty quantification in language models for the specific task of multiple-choice question-answering. We find that the uncertainty estimates from conformal prediction are tightly correlated with prediction accuracy. This observation can be useful for downstream applications such as selective classification and filtering out low-quality predictions. We also investigate the exchangeability assumption required by conformal prediction to out-of-subject questions, which may be a more realistic scenario for many practical applications. Our work contributes towards more trustworthy and reliable usage of large language models in safety-critical situations, where robust guarantees of error rate are required."
                },
                {
                    "title": "Holistic Evaluation of Language Models",
                    "abstract": "Language models (LMs) like GPT\u20103, PaLM, and ChatGPT are the foundation for almost all major language technologies, but their capabilities, limitations, and risks are not well understood. We present Holistic Evaluation of Language Models (HELM) to improve the transparency of LMs. LMs can serve many purposes and their behavior should satisfy many desiderata. To navigate the vast space of potential scenarios and metrics, we taxonomize the space and select representative subsets. We evaluate models on 16 core scenarios and 7 metrics, exposing important trade\u2010offs. We supplement our core evaluation with seven targeted evaluations to deeply analyze specific aspects (including world knowledge, reasoning, regurgitation of copyrighted content, and generation of disinformation). We benchmark 30 LMs, from OpenAI, Microsoft, Google, Meta, Cohere, AI21 Labs, and others. Prior to HELM, models were evaluated on just 17.9% of the core HELM scenarios, with some prominent models not sharing a single scenario in common. We improve this to 96.0%: all 30 models are now benchmarked under the same standardized conditions. Our evaluation surfaces 25 top\u2010level findings. For full transparency, we release all raw model prompts and completions publicly. HELM is a living benchmark for the community, continuously updated with new scenarios, metrics, and models https://crfm.stanford.edu/helm/latest/."
                },
                {
                    "title": "Rethink reporting of evaluation results in AI",
                    "abstract": "Aggregate metrics and lack of access to results limit understanding Artificial intelligence (AI) systems have begun to be deployed in high-stakes contexts, including autonomous driving and medical diagnosis. In contexts such as these, the consequences of system failures can be devastating. It is therefore vital that researchers and policy-makers have a full understanding of the capabilities and weaknesses of AI systems so that they can make informed decisions about where these systems are safe to use and how they might be improved. Unfortunately, current approaches to AI evaluation make it exceedingly difficult to build such an understanding, for two key reasons. First, aggregate metrics make it hard to predict how a system will perform in a particular situation. Second, the instance-by-instance evaluation results that could be used to unpack these aggregate metrics are rarely made available (1). Here, we propose a path forward in which results are presented in more nuanced ways and instance-by-instance evaluation results are made publicly available."
                },
                {
                    "title": "RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment",
                    "abstract": "Generative foundation models are susceptible to implicit biases that can arise from extensive unsupervised training data. Such biases can produce suboptimal samples, skewed outcomes, and unfairness, with potentially serious consequences. Consequently, aligning these models with human ethics and preferences is an essential step toward ensuring their responsible and effective deployment in real-world applications. Prior research has primarily employed Reinforcement Learning from Human Feedback (RLHF) to address this problem, where generative models are fine-tuned with RL algorithms guided by a human-feedback-informed reward model. However, the inefficiencies and instabilities associated with RL algorithms frequently present substantial obstacles to the successful alignment, necessitating the development of a more robust and streamlined approach. To this end, we introduce a new framework, Reward rAnked FineTuning (RAFT), designed to align generative models effectively. Utilizing a reward model and a sufficient number of samples, our approach selects the high-quality samples, discarding those that exhibit undesired behavior, and subsequently enhancing the model by fine-tuning on these filtered samples. Our studies show that RAFT can effectively improve the model performance in both reward learning and other automated metrics in both large language models and diffusion models."
                },
                {
                    "title": "Quantile Risk Control: A Flexible Framework for Bounding the Probability of High-Loss Predictions",
                    "abstract": "Rigorous guarantees about the performance of predictive algorithms are necessary in order to ensure their responsible use. Previous work has largely focused on bounding the expected loss of a predictor, but this is not sufficient in many risk-sensitive applications where the distribution of errors is important. In this work, we propose a flexible framework to produce a family of bounds on quantiles of the loss distribution incurred by a predictor. Our method takes advantage of the order statistics of the observed loss values rather than relying on the sample mean alone. We show that a quantile is an informative way of quantifying predictive performance, and that our framework applies to a variety of quantile-based metrics, each targeting important subsets of the data distribution. We analyze the theoretical properties of our proposed method and demonstrate its ability to rigorously control loss quantiles on several real-world datasets."
                },
                {
                    "title": "Semantic uncertainty intervals for disentangled latent spaces",
                    "abstract": "Meaningful uncertainty quantification in computer vision requires reasoning about semantic information -- say, the hair color of the person in a photo or the location of a car on the street. To this end, recent breakthroughs in generative modeling allow us to represent semantic information in disentangled latent spaces, but providing uncertainties on the semantic latent variables has remained challenging. In this work, we provide principled uncertainty intervals that are guaranteed to contain the true semantic factors for any underlying generative model. The method does the following: (1) it uses quantile regression to output a heuristic uncertainty interval for each element in the latent space (2) calibrates these uncertainties such that they contain the true value of the latent for a new, unseen input. The endpoints of these calibrated intervals can then be propagated through the generator to produce interpretable uncertainty visualizations for each semantic factor. This technique reliably communicates semantically meaningful, principled, and instance-adaptive uncertainty in inverse problems like image super-resolution and image completion."
                },
                {
                    "title": "Confident Adaptive Language Modeling",
                    "abstract": "Recent advances in Transformer-based large language models (LLMs) have led to significant performance improvements across many tasks. These gains come with a drastic increase in the models' size, potentially leading to slow and costly use at inference time. In practice, however, the series of generations made by LLMs is composed of varying levels of difficulty. While certain predictions truly benefit from the models' full capacity, other continuations are more trivial and can be solved with reduced compute. In this work, we introduce Confident Adaptive Language Modeling (CALM), a framework for dynamically allocating different amounts of compute per input and generation timestep. Early exit decoding involves several challenges that we address here, such as: (1) what confidence measure to use; (2) connecting sequence-level constraints to local per-token exit decisions; and (3) attending back to missing hidden representations due to early exits in previous tokens. Through theoretical analysis and empirical experiments on three diverse text generation tasks, we demonstrate the efficacy of our framework in reducing compute -- potential speedup of up to $\\times 3$ -- while provably maintaining high performance."
                },
                {
                    "title": "A global analysis of metrics used for measuring performance in natural language processing",
                    "abstract": "Measuring the performance of natural language processing models is challenging. Traditionally used metrics, such as BLEU and ROUGE, originally devised for machine translation and summarization, have been shown to suffer from low correlation with human judgment and a lack of transferability to other tasks and languages. In the past 15 years, a wide range of alternative metrics have been proposed. However, it is unclear to what extent this has had an impact on NLP benchmarking efforts. Here we provide the first large-scale cross-sectional analysis of metrics used for measuring performance in natural language processing. We curated, mapped and systematized more than 3500 machine learning model performance results from the open repository \u2018Papers with Code\u2019 to enable a global and comprehensive analysis. Our results suggest that the large majority of natural language processing metrics currently used have properties that may result in an inadequate reflection of a models\u2019 performance. Furthermore, we found that ambiguities and inconsistencies in the reporting of metrics may lead to difficulties in interpreting and comparing model performances, impairing transparency and reproducibility in NLP research."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "Prediction sets adaptive to unknown covariate shift.",
                    "abstract": "Predicting sets of outcomes-instead of unique outcomes-is a promising solution to uncertainty quantification in statistical learning. Despite a rich literature on constructing prediction sets with statistical guarantees, adapting to unknown covariate shift-a prevalent issue in practice-poses a serious unsolved challenge. In this article, we show that prediction sets with finite-sample coverage guarantee are uninformative and propose a novel flexible distribution-free method, PredSet-1Step, to efficiently construct prediction sets with an asymptotic coverage guarantee under unknown covariate shift. We formally show that our method is asymptotically probably approximately correct, having well-calibrated coverage error with high confidence for large samples. We illustrate that it achieves nominal coverage in a number of experiments and a data set concerning HIV risk prediction in a South African cohort study. Our theory hinges on a new bound for the convergence rate of the coverage of Wald confidence intervals based on general asymptotically linear estimators."
                },
                {
                    "title": "Conformal prediction under feedback covariate shift for biomolecular design",
                    "abstract": "Significance An increasingly high-impact application of machine learning in scientific discovery is its use in the design of novel objects with desired properties, such as the design of proteins, small molecules, and materials. Although a variety of algorithms have been developed for this purpose, it remains unclear when practitioners can trust the predictions made by learned models for designed objects, since design algorithms induce a distinctive shift between the training and test data distributions. We propose a method that provides confidence sets for designed objects, which we show satisfy finite-sample guarantees of statistical validity, for any design algorithm involving any learned regression model. Our work enables more trustworthy use of machine learning for design."
                },
                {
                    "title": "Red Teaming Language Models with Language Models",
                    "abstract": "Language Models (LMs) often cannot be deployed because of their potential to harm users in hard-to-predict ways. Prior work identifies harmful behaviors before deployment by using human annotators to hand-write test cases. However, human annotation is expensive, limiting the number and diversity of test cases. In this work, we automatically find cases where a target LM behaves in a harmful way, by generating test cases (\u201cred teaming\u201d) using another LM. We evaluate the target LM\u2019s replies to generated test questions using a classifier trained to detect offensive content, uncovering tens of thousands of offensive replies in a 280B parameter LM chatbot. We explore several methods, from zero-shot generation to reinforcement learning, for generating test cases with varying levels of diversity and difficulty. Furthermore, we use prompt engineering to control LM-generated test cases to uncover a variety of other harms, automatically finding groups of people that the chatbot discusses in offensive ways, personal and hospital phone numbers generated as the chatbot\u2019s own contact info, leakage of private training data in generated text, and harms that occur over the course of a conversation. Overall, LM-based red teaming is one promising tool (among many needed) for finding and fixing diverse, undesirable LM behaviors before impacting users."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Control",
                    "abstract": "We introduce a framework for calibrating machine learning models so that their predictions satisfy explicit, finite-sample statistical guarantees. Our calibration algorithms work with any underlying model and (unknown) data-generating distribution and do not require model refitting. The framework addresses, among other examples, false discovery rate control in multi-label classification, intersection-over-union control in instance segmentation, and the simultaneous control of the type-1 error of outlier detection and confidence set coverage in classification or regression. Our main insight is to reframe the risk-control problem as multiple hypothesis testing, enabling techniques and mathematical arguments different from those in the previous literature. We use the framework to provide new calibration methods for several core machine learning tasks, with detailed worked examples in computer vision and tabular medical data."
                },
                {
                    "title": "Finetuned Language Models Are Zero-Shot Learners",
                    "abstract": "This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning."
                },
                {
                    "title": "Do Prompt-Based Models Really Understand the Meaning of Their Prompts?",
                    "abstract": "Recently, a boom of papers has shown extraordinary progress in zero-shot and few-shot learning with various prompt-based models. It is commonly argued that prompts help models to learn faster in the same way that humans learn faster when provided with task instructions expressed in natural language. In this study, we experiment with over 30 prompts manually written for natural language inference (NLI). We find that models can learn just as fast with many prompts that are intentionally irrelevant or even pathologically misleading as they do with instructively \u201cgood\u201d prompts. Further, such patterns hold even for models as large as 175 billion parameters (Brown et al., 2020) as well as the recently proposed instruction-tuned models which are trained on hundreds of prompts (Sanh et al., 2021). That is, instruction-tuned models often produce good predictions with irrelevant and misleading prompts even at zero shots. In sum, notwithstanding prompt-based models\u2019 impressive improvement, we find evidence of serious limitations that question the degree to which such improvement is derived from models understanding task instructions in ways analogous to humans\u2019 use of task instructions."
                },
                {
                    "title": "A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification",
                    "abstract": "Black-box machine learning models are now routinely used in high-risk settings, like medical diagnostics, which demand uncertainty quantification to avoid consequential model failures. Conformal prediction is a user-friendly paradigm for creating statistically rigorous uncertainty sets/intervals for the predictions of such models. Critically, the sets are valid in a distribution-free sense: they possess explicit, non-asymptotic guarantees even without distributional assumptions or model assumptions. One can use conformal prediction with any pre-trained model, such as a neural network, to produce sets that are guaranteed to contain the ground truth with a user-specified probability, such as 90%. It is easy-to-understand, easy-to-use, and general, applying naturally to problems arising in the fields of computer vision, natural language processing, deep reinforcement learning, and so on. This hands-on introduction is aimed to provide the reader a working understanding of conformal prediction and related distribution-free uncertainty quantification techniques with one self-contained document. We lead the reader through practical theory for and examples of conformal prediction and describe its extensions to complex machine learning tasks involving structured outputs, distribution shift, time-series, outliers, models that abstain, and more. Throughout, there are many explanatory illustrations, examples, and code samples in Python. With each code sample comes a Jupyter notebook implementing the method on a real-data example; the notebooks can be accessed and easily run using our codebase."
                },
                {
                    "title": "PAC Prediction Sets Under Covariate Shift",
                    "abstract": "An important challenge facing modern machine learning is how to rigorously quantify the uncertainty of model predictions. Conveying uncertainty is especially important when there are changes to the underlying data distribution that might invalidate the predictive model. Yet, most existing uncertainty quantification algorithms break down in the presence of such shifts. We propose a novel approach that addresses this challenge by constructing \\emph{probably approximately correct (PAC)} prediction sets in the presence of covariate shift. Our approach focuses on the setting where there is a covariate shift from the source distribution (where we have labeled training examples) to the target distribution (for which we want to quantify uncertainty). Our algorithm assumes given importance weights that encode how the probabilities of the training examples change under the covariate shift. In practice, importance weights typically need to be estimated; thus, we extend our algorithm to the setting where we are given confidence intervals for the importance weights. We demonstrate the effectiveness of our approach on covariate shifts based on DomainNet and ImageNet. Our algorithm satisfies the PAC constraint, and gives prediction sets with the smallest average normalized size among approaches that always satisfy the PAC constraint."
                },
                {
                    "title": "Adaptive Conformal Inference Under Distribution Shift",
                    "abstract": "We develop methods for forming prediction sets in an online setting where the data generating distribution is allowed to vary over time in an unknown fashion. Our framework builds on ideas from conformal inference to provide a general wrapper that can be combined with any black box method that produces point predictions of the unseen label or estimated quantiles of its distribution. While previous conformal inference methods rely on the assumption that the data points are exchangeable, our adaptive approach provably achieves the desired coverage frequency over long-time intervals irrespective of the true data generating process. We accomplish this by modelling the distribution shift as a learning problem in a single parameter whose optimal value is varying over time and must be continuously re-estimated. We test our method, adaptive conformal inference, on two real world datasets and find that its predictions are robust to visible and significant distribution shifts."
                },
                {
                    "title": "The Power of Scale for Parameter-Efficient Prompt Tuning",
                    "abstract": "In this work, we explore \u201cprompt tuning,\u201d a simple yet effective mechanism for learning \u201csoft prompts\u201d to condition frozen language models to perform specific downstream tasks. Unlike the discrete text prompts used by GPT-3, soft prompts are learned through backpropagation and can be tuned to incorporate signals from any number of labeled examples. Our end-to-end learned approach outperforms GPT-3\u2019s few-shot learning by a large margin. More remarkably, through ablations on model size using T5, we show that prompt tuning becomes more competitive with scale: as models exceed billions of parameters, our method \u201ccloses the gap\u201d and matches the strong performance of model tuning (where all model weights are tuned). This finding is especially relevant because large models are costly to share and serve and the ability to reuse one frozen model for multiple downstream tasks can ease this burden. Our method can be seen as a simplification of the recently proposed \u201cprefix tuning\u201d of Li and Liang (2021) and we provide a comparison to this and other similar approaches. Finally, we show that conditioning a frozen model with soft prompts confers benefits in robustness to domain transfer and enables efficient \u201cprompt ensembling.\u201d We release code and model checkpoints to reproduce our experiments."
                },
                {
                    "title": "Distribution-Free, Risk-Controlling Prediction Sets",
                    "abstract": "While improving prediction accuracy has been the focus of machine learning in recent years, this alone does not suffice for reliable decision-making. Deploying learning systems in consequential settings also requires calibrating and communicating the uncertainty of predictions. To convey instance-wise uncertainty for prediction tasks, we show how to generate set-valued predictions from a black-box predictor that controls the expected loss on future test points at a user-specified level. Our approach provides explicit finite-sample guarantees for any dataset by using a holdout set to calibrate the size of the prediction sets. This framework enables simple, distribution-free, rigorous error control for many tasks, and we demonstrate it in five large-scale machine learning problems: (1) classification problems where some mistakes are more costly than others; (2) multi-label classification, where each observation has multiple associated labels; (3) classification problems where the labels have a hierarchical structure; (4) image segmentation, where we wish to predict a set of pixels containing an object of interest; and (5) protein structure prediction. Last, we discuss extensions to uncertainty quantification for ranking, metric learning, and distributionally robust learning."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "PAC Confidence Sets for Deep Neural Networks via Calibrated Prediction",
                    "abstract": "We propose an algorithm combining calibrated prediction and generalization bounds from learning theory to construct confidence sets for deep neural networks with PAC guarantees---i.e., the confidence set for a given input contains the true label with high probability. We demonstrate how our approach can be used to construct PAC confidence sets on ResNet for ImageNet, and on a dynamics model the half-cheetah reinforcement learning problem."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "On the Summarization of Consumer Health Questions",
                    "abstract": "Question understanding is one of the main challenges in question answering. In real world applications, users often submit natural language questions that are longer than needed and include peripheral information that increases the complexity of the question, leading to substantially more false positives in answer retrieval. In this paper, we study neural abstractive models for medical question summarization. We introduce the MeQSum corpus of 1,000 summarized consumer health questions. We explore data augmentation methods and evaluate state-of-the-art neural abstractive models on this new task. In particular, we show that semantic augmentation from question datasets improves the overall performance, and that pointer-generator networks outperform sequence-to-sequence attentional models on this task, with a ROUGE-1 score of 44.16%. We also present a detailed error analysis and discuss directions for improvement that are specific to question summarization."
                },
                {
                    "title": "Fairness risk measures",
                    "abstract": "Ensuring that classifiers are non-discriminatory or fair with respect to a sensitive feature (e.g., race or gender) is a topical problem. Progress in this task requires fixing a definition of fairness, and there have been several proposals in this regard over the past few years. Several of these, however, assume either binary sensitive features (thus precluding categorical or real-valued sensitive groups), or result in non-convex objectives (thus adversely affecting the optimisation landscape). In this paper, we propose a new definition of fairness that generalises some existing proposals, while allowing for generic sensitive features and resulting in a convex objective. The key idea is to enforce that the expected losses (or risks) across each subgroup induced by the sensitive feature are commensurate. We show how this relates to the rich literature on risk measures from mathematical finance. As a special case, this leads to a new convex fairness-aware objective based on minimising the conditional value at risk (CVaR)."
                },
                {
                    "title": "Don\u2019t Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarization",
                    "abstract": "We introduce \u201cextreme summarization\u201d, a new single-document summarization task which does not favor extractive strategies and calls for an abstractive modeling approach. The idea is to create a short, one-sentence news summary answering the question \u201cWhat is the article about?\u201d. We collect a real-world, large-scale dataset for this task by harvesting online articles from the British Broadcasting Corporation (BBC). We propose a novel abstractive model which is conditioned on the article\u2019s topics and based entirely on convolutional neural networks. We demonstrate experimentally that this architecture captures long-range dependencies in a document and recognizes pertinent content, outperforming an oracle extractive system and state-of-the-art abstractive approaches when evaluated automatically and by humans."
                },
                {
                    "title": "Abstractive Text Summarization using Sequence-to-sequence RNNs and Beyond",
                    "abstract": "In this work, we model abstractive text summarization using Attentional Encoder-Decoder Recurrent Neural Networks, and show that they achieve state-of-the-art performance on two different corpora. We propose several novel models that address critical problems in summarization that are not adequately modeled by the basic architecture, such as modeling key-words, capturing the hierarchy of sentence-to-word structure, and emitting words that are rare or unseen at training time. Our work shows that many of our proposed models contribute to further improvement in performance. We also propose a new dataset consisting of multi-sentence summaries, and establish performance benchmarks for further research."
                },
                {
                    "title": "A tutorial on conformal prediction",
                    "abstract": "Conformal prediction uses past experience to determine precise levels of confidence in new predictions. Given an error probability e, together with a method that makes a prediction \u0177 of a label y, it produces a set of labels, typically containing \u0177, that also contains y with probability 1 \u0096 e. Conformal prediction can be applied to any method for producing \u0177: a nearest-neighbor method, a support-vector machine, ridge regression, etc. \n \nConformal prediction is designed for an on-line setting in which labels are predicted successively, each one being revealed before the next is predicted. The most novel and valuable feature of conformal prediction is that if the successive examples are sampled independently from the same distribution, then the successive predictions will be right 1 \u0096 e of the time, even though they are based on an accumulating data set rather than on independent data sets. \n \nIn addition to the model under which successive examples are sampled independently, other on-line compression models can also use conformal prediction. The widely used Gaussian linear model is one of these. \n \nThis tutorial presents a self-contained account of the theory of conformal prediction and works through several numerical examples. A more comprehensive treatment of the topic is provided in Algorithmic Learning in a Random World, by Vladimir Vovk, Alex Gammerman, and Glenn Shafer (Springer, 2005)."
                },
                {
                    "title": "ROUGE: A Package for Automatic Evaluation of Summaries",
                    "abstract": "ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST."
                },
                {
                    "title": "The Kolmogorov-Smirnov Test for Goodness of Fit",
                    "abstract": "Abstract The test is based on the maximum difference between an empirical and a hypothetical cumulative distribution. Percentage points are tabled, and a lower bound to the power function is charted. Confidence limits for a cumulative distribution are described. Examples are given. Indications that the test is superior to the chi-square test are cited."
                },
                {
                    "title": "On the Measurement of Inequality",
                    "abstract": "Measures of inequality are used by economists to answer a wide range of questions. Is the distribution of income more equal than it was in the past? Are underdeveloped countries characterised by greater inequality than advanced countries ? Do taxes lead to greater equality in the distribution of income or wealth? However, despite the wide use of these measures, relatively little attention has been given to the conceptual problems involved in the measurement of inequality and there have been few contributions to the theoretical foundations of the subject. In this paper, I try to clarify some of the basic issues, to examine the properties of the measures that are commonly employed, and to discuss a possible new approach. In the course of this, I draw on the parallel with the formally similar problem of measuring risk in the theory of decisionmaking under uncertainty and make use of recent results in this fie1d.l The problem with which we are concerned is basically that of comparing two frequency distributions f(u) of an attribute y which for convenience I shall refer to as income. The conventional approach in nearly all empirical work is to adopt some summary statistic of inequality such as the variance, the coefficient of variation or the Gini coefficientwith no very explicit reason being given for preferring one measure rather than another. As, however, was pointed out by Dalton 50 years ago in his pioneering article [3], underlying any such measure is some concept of social welfare and it is with this concept that we should be concerned. He argued that we should approach the question by considering directly the form of the social welfare function to be employed. If we follow him in assuming that this would be an additively separable and symmetric"
                },
                {
                    "title": "Optimization of conditional value-at risk",
                    "abstract": "In an intensifying international competition banks are forced to place increased emphasis on enterprise wide risk-/return management. Financial risks have to be limited and managed from a bank wide portfolio perspective. Risk management requirements have to be met from an internal as well as from a regulatory point of view. Banks need to maximize their expected returns under these constraints. This leads to a generalized portfolio optimization problem under different capital restrictions. We pursue a two-step Risk-/Return Management Approach (\u201cRRM-Approach\u201d) [3]. At first we formulate an optimization model that maximizes the expected returns of the bank portfolio to the planning horizon under internal and regulatory loss risk limitations. The restriction on the internal economic capital is based on the risk measure of Conditional Value at Risk (CVaR), that has been proved to be appropriate for measuring bank wide loss risk [1]. The regulatory capital restrictions represent the actual Basle Rules of risk limitation. The optimization model of step 1 of the RRM-Approach is solved by an application of the CVaR-optimization approach by Rockafellar/Uryasev [1]. In the second step, we derive a consistent risk-/return key ratio system from the optimum portfolio of step 1. We estimate the risk and return contributions of each single asset in the portfolio and achieve additive, linear representations of the expected returns and the regulatory and internal risk contributions. The latter we obtain by an application of Euler\u2019s Formula on CVaR [2]. We sum up the risk and return contributions of the single assets on the business line level. In this way we deduce consistent return targets and capital limits of the economic and the internal capital for each business line. These quantities represent basic planning information ensuring maximum return targets and an efficient capital allocation of the economic and the regulatory capital. The impact of the RRM-Approach is shown by a brief application example."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively select prompts for large language models (LLMs) to minimize the risk of generating harmful or toxic outputs while maintaining performance?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing concern of safety and ethical implications in deploying LLMs. By developing a systematic approach to prompt selection that balances usefulness and safety, we can enhance the reliability of LLMs in real-world applications. This research could lead to advancements in understanding the underlying mechanisms of prompting, fostering further exploration into safe AI deployment, and potentially influencing regulatory frameworks for AI systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of LLMs and the unpredictability of their outputs based on different prompts. Naive approaches may fail because they often rely on average performance metrics that do not account for the variance or potential outlier events in model responses. Additionally, the lack of rigorous methodologies for quantifying risk and uncertainty in prompt selection complicates the task, as does the need to balance multiple competing objectives (e.g., helpfulness vs. harmlessness).\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on empirical validation of prompts without a comprehensive understanding of the underlying risk dynamics. Existing solutions often lack a formal framework for quantifying and controlling risk, leading to gaps in knowledge about the trade-offs involved in prompt selection. Our approach differs by employing Distribution-Free Uncertainty Quantification (DFUQ) techniques to provide statistically sound risk measures, enabling a more principled selection of prompts that can mitigate harmful outputs.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the implementation of the Prompt Risk Control (PRC) framework, which utilizes DFUQ techniques to establish upper bounds on user-defined risk measures. We will apply this framework to a dataset of prompts and their corresponding outputs from LLMs, measuring toxicity and helpfulness as key metrics. The expected outcome is a set of prompts that can be confidently selected to minimize the risk of generating unacceptable outputs while maintaining overall performance, thus providing a safer deployment strategy for LLMs."
            }
        },
        "author_data": {
            "e6a9c71e-bdc8-4284-8e1b-f9c7e438bc2f": {
                "pk": "e6a9c71e-bdc8-4284-8e1b-f9c7e438bc2f",
                "name": "Thomas P. Zollo",
                "collaborators": [
                    "Zhun Deng",
                    "Jake Snell",
                    "T. Pitassi",
                    "R. Zemel",
                    "A. Siah",
                    "Naimeng Ye",
                    "Ang Li",
                    "Hongseok Namkoong",
                    "Richard Zemel"
                ],
                "domain": [
                    "Machine Learning",
                    "Personalization",
                    "Calibration",
                    "Statistical Guarantees"
                ],
                "publications": [
                    {
                        "title": "PersonalLLM: Tailoring LLMs to Individual Preferences",
                        "abstract": "As LLMs become capable of complex tasks, there is growing potential for personalized interactions tailored to the subtle and idiosyncratic preferences of the user. We present a public benchmark, PersonalLLM, focusing on adapting LLMs to provide maximal benefits for a particular user. Departing from existing alignment benchmarks that implicitly assume uniform preferences, we curate open-ended prompts paired with many high-quality answers over which users would be expected to display heterogeneous latent preferences. Instead of persona-prompting LLMs based on high-level attributes (e.g., user's race or response length), which yields homogeneous preferences relative to humans, we develop a method that can simulate a large user base with diverse preferences from a set of pre-trained reward models. Our dataset and generated personalities offer an innovative testbed for developing personalization algorithms that grapple with continual data sparsity--few relevant feedback from the particular user--by leveraging historical data from other (similar) users. We explore basic in-context learning and meta-learning baselines to illustrate the utility of PersonalLLM and highlight the need for future methodological development. Our dataset is available at https://huggingface.co/datasets/namkoong-lab/PersonalLLM"
                    },
                    {
                        "title": "Improving Predictor Reliability with Selective Recalibration",
                        "abstract": "A reliable deep learning system should be able to accurately express its confidence with respect to its predictions, a quality known as calibration. One of the most effective ways to produce reliable confidence estimates with a pre-trained model is by applying a post-hoc recalibration method. Popular recalibration methods like temperature scaling are typically fit on a small amount of data and work in the model's output space, as opposed to the more expressive feature embedding space, and thus usually have only one or a handful of parameters. However, the target distribution to which they are applied is often complex and difficult to fit well with such a function. To this end we propose \\textit{selective recalibration}, where a selection model learns to reject some user-chosen proportion of the data in order to allow the recalibrator to focus on regions of the input space that can be well-captured by such a model. We provide theoretical analysis to motivate our algorithm, and test our method through comprehensive experiments on difficult medical imaging and zero-shot classification tasks. Our results show that selective recalibration consistently leads to significantly lower calibration error than a wide range of selection and recalibration baselines."
                    },
                    {
                        "title": "Distribution-Free Statistical Dispersion Control for Societal Applications",
                        "abstract": "Explicit finite-sample statistical guarantees on model performance are an important ingredient in responsible machine learning. Previous work has focused mainly on bounding either the expected loss of a predictor or the probability that an individual prediction will incur a loss value in a specified range. However, for many high-stakes applications, it is crucial to understand and control the dispersion of a loss distribution, or the extent to which different members of a population experience unequal effects of algorithmic decisions. We initiate the study of distribution-free control of statistical dispersion measures with societal implications and propose a simple yet flexible framework that allows us to handle a much richer class of statistical functionals beyond previous work. Our methods are verified through experiments in toxic comment detection, medical imaging, and film recommendation."
                    },
                    {
                        "title": "Quantile Risk Control: A Flexible Framework for Bounding the Probability of High-Loss Predictions",
                        "abstract": "Rigorous guarantees about the performance of predictive algorithms are necessary in order to ensure their responsible use. Previous work has largely focused on bounding the expected loss of a predictor, but this is not sufficient in many risk-sensitive applications where the distribution of errors is important. In this work, we propose a flexible framework to produce a family of bounds on quantiles of the loss distribution incurred by a predictor. Our method takes advantage of the order statistics of the observed loss values rather than relying on the sample mean alone. We show that a quantile is an informative way of quantifying predictive performance, and that our framework applies to a variety of quantile-based metrics, each targeting important subsets of the data distribution. We analyze the theoretical properties of our proposed method and demonstrate its ability to rigorously control loss quantiles on several real-world datasets."
                    }
                ]
            },
            "8edce95b-e7d0-49f5-93eb-7b3913e50d97": {
                "pk": "8edce95b-e7d0-49f5-93eb-7b3913e50d97",
                "name": "Todd Morrill",
                "collaborators": [
                    "Zhaoyuan Deng",
                    "Yanda Chen",
                    "Amith Ananthram",
                    "Colin Wayne Leach",
                    "Kathleen McKeown"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Dialogue Systems",
                    "Social Computing"
                ],
                "publications": [
                    {
                        "title": "Social Orientation: A New Feature for Dialogue Analysis",
                        "abstract": "There are many settings where it is useful to predict and explain the success or failure of a dialogue. Circumplex theory from psychology models the social orientations (e.g., Warm-Agreeable, Arrogant-Calculating) of conversation participants and can be used to predict and explain the outcome of social interactions. Our work is novel in its systematic application of social orientation tags to modeling conversation outcomes. In this paper, we introduce a new data set of dialogue utterances machine-labeled with social orientation tags. We show that social orientation tags improve task performance, especially in low-resource settings, on both English and Chinese language benchmarks. We also demonstrate how social orientation tags help explain the outcomes of social interactions when used in neural models. Based on these results showing the utility of social orientation tags for dialogue outcome prediction tasks, we release our data sets, code, and models that are fine-tuned to predict social orientation tags on dialogue utterances."
                    }
                ]
            },
            "d67a20e8-6d28-4623-ae0c-7138ff8ec2ad": {
                "pk": "d67a20e8-6d28-4623-ae0c-7138ff8ec2ad",
                "name": "Zhun Deng",
                "collaborators": [
                    "Thomas P. Zollo",
                    "Jake Snell",
                    "T. Pitassi",
                    "Richard Zemel",
                    "R. Zemel"
                ],
                "domain": [
                    "Deep Learning",
                    "Model Calibration",
                    "Responsible AI",
                    "Statistical Learning"
                ],
                "publications": [
                    {
                        "title": "Improving Predictor Reliability with Selective Recalibration",
                        "abstract": "A reliable deep learning system should be able to accurately express its confidence with respect to its predictions, a quality known as calibration. One of the most effective ways to produce reliable confidence estimates with a pre-trained model is by applying a post-hoc recalibration method. Popular recalibration methods like temperature scaling are typically fit on a small amount of data and work in the model's output space, as opposed to the more expressive feature embedding space, and thus usually have only one or a handful of parameters. However, the target distribution to which they are applied is often complex and difficult to fit well with such a function. To this end we propose \\textit{selective recalibration}, where a selection model learns to reject some user-chosen proportion of the data in order to allow the recalibrator to focus on regions of the input space that can be well-captured by such a model. We provide theoretical analysis to motivate our algorithm, and test our method through comprehensive experiments on difficult medical imaging and zero-shot classification tasks. Our results show that selective recalibration consistently leads to significantly lower calibration error than a wide range of selection and recalibration baselines."
                    },
                    {
                        "title": "Distribution-Free Statistical Dispersion Control for Societal Applications",
                        "abstract": "Explicit finite-sample statistical guarantees on model performance are an important ingredient in responsible machine learning. Previous work has focused mainly on bounding either the expected loss of a predictor or the probability that an individual prediction will incur a loss value in a specified range. However, for many high-stakes applications, it is crucial to understand and control the dispersion of a loss distribution, or the extent to which different members of a population experience unequal effects of algorithmic decisions. We initiate the study of distribution-free control of statistical dispersion measures with societal implications and propose a simple yet flexible framework that allows us to handle a much richer class of statistical functionals beyond previous work. Our methods are verified through experiments in toxic comment detection, medical imaging, and film recommendation."
                    }
                ]
            },
            "7a555cbe-94ec-4ccb-a4fe-c9ad0130a031": {
                "pk": "7a555cbe-94ec-4ccb-a4fe-c9ad0130a031",
                "name": "Jake C. Snell",
                "collaborators": [
                    "R. Zemel",
                    "Kevin Swersky",
                    "Mengye Ren",
                    "Eleni Triantafillou",
                    "Thomas P. Zollo",
                    "Zhun Deng",
                    "T. Pitassi",
                    "Kuan-Chieh Jackson Wang",
                    "James Lucas",
                    "X. Pitkow",
                    "A. Tolias",
                    "Peng Xie",
                    "Haitao Zheng",
                    "Chelsea Finn",
                    "Pieter Abbeel",
                    "Sergey Levine",
                    "R. Socher",
                    "J. Devlin",
                    "Ming-Wei Chang",
                    "Kenton Lee",
                    "Richard Zemel",
                    "Bhishma Dedhia",
                    "Michael Chang",
                    "T. Griffiths",
                    "N. Jha",
                    "Xu Han",
                    "Alexander Fritzler",
                    "V. Logacheva",
                    "M. Garnelo",
                    "Dan Rosenbaum",
                    "Christopher Maddison",
                    "Tiago Ramalho",
                    "David Saxton",
                    "Murray Shanahan",
                    "Maximilian Hofer",
                    "A. Kormilitzin",
                    "Paul Goldberg",
                    "Yongliang Shen",
                    "Xinyin Ma",
                    "Zeqi Tan",
                    "Shuai Zhang",
                    "Philip Torr",
                    "Timothy M. Hospedales",
                    "Kelsey Allen",
                    "Evan Shelhamer",
                    "Hanul Shin",
                    "Mina Ghadimi Atigh",
                    "Martin Keller-Ressel",
                    "Yinbo Chen",
                    "Zhuang Liu",
                    "Huijuan Xu",
                    "Trevor Darrell",
                    "Jia Deng",
                    "Wei Dong",
                    "Li-Jia Li",
                    "Kai Li",
                    "Ning Ding",
                    "Xiaobin Wang",
                    "Yao Fu",
                    "Guangwei Xu",
                    "Rui",
                    "Pengjun Wang",
                    "Ying Xie",
                    "Fei Shen",
                    "Haihua Huang",
                    "Hai-Tao",
                    "Yulin Chen",
                    "Xuliang Han",
                    "Luca Franceschi",
                    "P. Frasconi",
                    "Saverio Salzo",
                    "Avinash Ravichandran",
                    "Rahul Bhotika",
                    "Sachin Ravi",
                    "Eleanor Rosch",
                    "Carolyn B. Mervis",
                    "Wayne D Gray",
                    "David M Johnson",
                    "P. Boyes-Braem",
                    "Andrei A. Rusu",
                    "Dushyant Rao",
                    "Jakub Sygnowski",
                    "Oriol",
                    "Razvan Pascanu",
                    "Simon Osindero",
                    "Hiba Asri",
                    "H. Mousannif",
                    "H. A. Moatassime",
                    "L. Bentivogli",
                    "Ido Dagan",
                    "H. Dang",
                    "Giampiccolo Bernardo",
                    "Magnini. 2009",
                    "Tom B. Brown",
                    "Benjamin Mann",
                    "Nick Ryder",
                    "Jared D Subbiah",
                    "Prafulla Kaplan",
                    "A. Dhariwal",
                    "P. Neelakantan"
                ],
                "domain": [
                    "Few-Shot Learning",
                    "Machine Learning",
                    "Natural Language Processing",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "Improving Predictor Reliability with Selective Recalibration",
                        "abstract": "A reliable deep learning system should be able to accurately express its confidence with respect to its predictions, a quality known as calibration. One of the most effective ways to produce reliable confidence estimates with a pre-trained model is by applying a post-hoc recalibration method. Popular recalibration methods like temperature scaling are typically fit on a small amount of data and work in the model's output space, as opposed to the more expressive feature embedding space, and thus usually have only one or a handful of parameters. However, the target distribution to which they are applied is often complex and difficult to fit well with such a function. To this end we propose \\textit{selective recalibration}, where a selection model learns to reject some user-chosen proportion of the data in order to allow the recalibrator to focus on regions of the input space that can be well-captured by such a model. We provide theoretical analysis to motivate our algorithm, and test our method through comprehensive experiments on difficult medical imaging and zero-shot classification tasks. Our results show that selective recalibration consistently leads to significantly lower calibration error than a wide range of selection and recalibration baselines."
                    },
                    {
                        "title": "Im-Promptu: In-Context Composition from Image Prompts",
                        "abstract": "Large language models are few-shot learners that can solve diverse tasks from a handful of demonstrations. This implicit understanding of tasks suggests that the attention mechanisms over word tokens may play a role in analogical reasoning. In this work, we investigate whether analogical reasoning can enable in-context composition over composable elements of visual stimuli. First, we introduce a suite of three benchmarks to test the generalization properties of a visual in-context learner. We formalize the notion of an analogy-based in-context learner and use it to design a meta-learning framework called Im-Promptu. Whereas the requisite token granularity for language is well established, the appropriate compositional granularity for enabling in-context generalization in visual stimuli is usually unspecified. To this end, we use Im-Promptu to train multiple agents with different levels of compositionality, including vector representations, patch representations, and object slots. Our experiments reveal tradeoffs between extrapolation abilities and the degree of compositionality, with non-compositional representations extending learned composition rules to unseen domains but performing poorly on combinatorial tasks. Patch-based representations require patches to contain entire objects for robust extrapolation. At the same time, object-centric tokenizers coupled with a cross-attention module generate consistent and high-fidelity solutions, with these inductive biases being particularly crucial for compositional generalization. Lastly, we demonstrate a use case of Im-Promptu as an intuitive programming interface for image generation."
                    },
                    {
                        "title": "Mitigating Prototype Shift: Few-Shot Nested Named Entity Recognition with Prototype-Attention",
                        "abstract": "Nested entities are prone to obtain similar rep-001 resentations in pre-trained language models, 002 posing challenges for Named Entity Recog-003 nition (NER), especially in the few-shot set-004 ting where prototype shifts often occur due to 005 distribution differences between the support 006 and query sets. In this paper, we regard en-007 tity representation as the combination of proto-008 type and non-prototype representations. With 009 a hypothesis that using the prototype repre-010 sentation specifically can help mitigate poten-011 tial prototype shifts, we propose a Prototype-012 Attention mechanism in the Contrastive Learn-013 ing framework (PACL) for the few-shot nested 014 NER. PACL first generates prototype-enhanced 015 span representations to mitigate the prototype 016 shift by applying a prototype attention mech-017 anism. It then adopts a novel prototype-span 018 contrastive loss to reduce prototype differences 019 further and overcome the O-type\u2019s non-unique 020 prototype limitation by comparing prototype-021 enhanced span representations with prototypes 022 and original semantic representations. Our ex-023 periments on three English, German, and Rus-024 sian nested NER datasets show that the PACL 025 outperformed seven baseline models on the 026 1-shot and 5-shot tasks in terms of F 1 score. 027 Further analyses indicate that our Prototype-028 Attention mechanism has high generality, en-029 hancing the performance of two baseline mod-030 els, and can serve as a valuable tool for NLP 031 practitioners facing few-shot nested NER tasks. 032"
                    },
                    {
                        "title": "Distribution-Free Statistical Dispersion Control for Societal Applications",
                        "abstract": "Explicit finite-sample statistical guarantees on model performance are an important ingredient in responsible machine learning. Previous work has focused mainly on bounding either the expected loss of a predictor or the probability that an individual prediction will incur a loss value in a specified range. However, for many high-stakes applications, it is crucial to understand and control the dispersion of a loss distribution, or the extent to which different members of a population experience unequal effects of algorithmic decisions. We initiate the study of distribution-free control of statistical dispersion measures with societal implications and propose a simple yet flexible framework that allows us to handle a much richer class of statistical functionals beyond previous work. Our methods are verified through experiments in toxic comment detection, medical imaging, and film recommendation."
                    },
                    {
                        "title": "Quantile Risk Control: A Flexible Framework for Bounding the Probability of High-Loss Predictions",
                        "abstract": "Rigorous guarantees about the performance of predictive algorithms are necessary in order to ensure their responsible use. Previous work has largely focused on bounding the expected loss of a predictor, but this is not sufficient in many risk-sensitive applications where the distribution of errors is important. In this work, we propose a flexible framework to produce a family of bounds on quantiles of the loss distribution incurred by a predictor. Our method takes advantage of the order statistics of the observed loss values rather than relying on the sample mean alone. We show that a quantile is an informative way of quantifying predictive performance, and that our framework applies to a variety of quantile-based metrics, each targeting important subsets of the data distribution. We analyze the theoretical properties of our proposed method and demonstrate its ability to rigorously control loss quantiles on several real-world datasets."
                    },
                    {
                        "title": "Few-shot Learning with Big Prototypes",
                        "abstract": "Metric-based meta-learning is one of the de 001 facto standards in few-shot learning. It com-002 poses of representation learning and metrics 003 calculation designs. Previous works construct 004 class representations in different ways, varying 005 from mean output embedding to covariance 006 and distributions. However, using embeddings 007 in space lacks expressivity and cannot capture 008 class information robustly, while statistical 009 complex modeling poses difficulty to metric 010 designs. In this work, we use tensor fields 011 (\u201careas\u201d) to model classes from the geometrical 012 perspective for few-shot learning. Specifically, 013 we present big prototypes , where class infor-014 mation is represented by hyperspheres with 015 dynamic sizes with two sets of learnable param-016 eters: the hypersphere\u2019s center and the radius. 017 Extending from points to areas, hyperspheres 018 are much more expressive than embeddings. 019 Moreover, it is more convenient to perform 020 metric-based classification with big prototypes 021 than statistical modeling, as we only need to 022 calculate the distance from a data point to the 023 surface of the hypersphere. Following this 024 idea, we also develop two variants of big pro-025 totypes under other measurements. Extensive 026 experiments and analysis on few-shot learning 027 tasks across NLP and CV and comparison 028 with 20+ competitive baselines demonstrate 029 the effectiveness of big prototypes. 030"
                    },
                    {
                        "title": "Few-Shot Attribute Learning",
                        "abstract": "Semantic concepts are frequently de\ufb01ned by combinations of underlying attributes. As mappings from attributes to classes are often simple, attribute-based representations facilitate novel concept learning with zero or few examples. A signi\ufb01cant limitation of existing attribute-based learning paradigms, such as zero-shot learning, is that the attributes are assumed to be known and \ufb01xed. In this work we study the rapid learning of attributes that were not previously labeled. Compared to standard few-shot learning of semantic classes, in which novel classes may be de\ufb01ned by attributes that were relevant at training time, learning new attributes imposes a stiffer challenge. We found that supervised learning with training attributes does not generalize well to new test attributes, whereas self-supervised pre-training brings signi\ufb01cant improvement. We further experimented with random splits of the attribute space and found that predictability of test attributes provides an informative estimate of a model\u2019s generalization ability."
                    },
                    {
                        "title": "P ERFECT : Prompt-free and Efficient Language Model Fine-Tuning",
                        "abstract": "Current methods for few-shot fine-tuning of 001 pretrained masked language model (PLM) require 002 carefully engineered prompts and verbalizers 003 for each new task, to convert examples into a 004 cloze-format that the PLM can score. In this work, 005 we propose P ERFECT , a simple and efficient 006 method for few-shot fine-tuning of PLMs without 007 relying on any such handcrafting , which is highly 008 effective given as few as 32 data points. P ERFECT 009 makes two key design choices: First, we show 010 that manually engineered task prompts can be 011 replaced with task-specific adapters that enable 012 sample-efficient fine-tuning and reduce memory 013 and storage costs by roughly factors of 5 and 100, 014 respectively. Second, instead of using handcrafted 015 verbalizers, we learn a new multi-token label em-016 bedding during fine-tuning which are not tied to 017 the model vocabulary and which allow us to avoid 018 complex auto-regressive decoding. These embed-019 dings are not only learnable from limited data but 020 also enable nearly 100x faster training and infer-021 ence. Experiments on a wide range of few shot 022 NLP tasks demonstrate that P ERFECT , while be-023 ing simple and efficient, also outperforms existing 024 state-of-the-art few-shot learning methods. 1 025"
                    },
                    {
                        "title": "Bayesian Few-Shot Classification with One-vs-Each P\u00f3lya-Gamma Augmented Gaussian Processes",
                        "abstract": "Few-shot classification (FSC), the task of adapting a classifier to unseen classes given a small labeled dataset, is an important step on the path toward human-like machine learning. Bayesian methods are well-suited to tackling the fundamental issue of overfitting in the few-shot scenario because they allow practitioners to specify prior beliefs and update those beliefs in light of observed data. Contemporary approaches to Bayesian few-shot classification maintain a posterior distribution over model parameters, which is slow and requires storage that scales with model size. Instead, we propose a Gaussian process classifier based on a novel combination of Polya-gamma augmentation and the one-vs-each softmax approximation that allows us to efficiently marginalize over functions rather than model parameters. We demonstrate improved accuracy and uncertainty quantification on both standard few-shot classification benchmarks and few-shot domain transfer tasks."
                    },
                    {
                        "title": "Probing Few-Shot Generalization with Attributes",
                        "abstract": "Despite impressive progress in deep learning, generalizing far beyond the training distribution is an important open challenge. In this work, we consider few-shot classi\ufb01cation, and aim to shed light on what makes some novel classes easier to learn than others, and what types of learned representations generalize better. To this end, we de\ufb01ne a new paradigm in terms of attributes \u2014simple building blocks of which concepts are formed\u2014as a means of quantifying the degree of relatedness of different concepts. Our empirical analysis reveals that supervised learning generalizes poorly to new attributes, but a combination of self-supervised pretraining with supervised \ufb01netuning leads to stronger generalization. The bene\ufb01t of self-supervised pretraining and supervised \ufb01netuning is further investigated through controlled experiments using random splits of the attribute space, and we \ufb01nd that predictability of test attributes provides an informative estimate of a model\u2019s generalization ability."
                    },
                    {
                        "title": "Flexible Few-Shot Learning of Contextual Similarity",
                        "abstract": "Existing approaches to few-shot learning deal with tasks that have persistent, rigid notions of classes. Typically, the learner observes data only from a \ufb01xed number of classes at training time and is asked to generalize to a new set of classes at test time. Two examples from the same class would always be assigned the same labels in any episode. In this work, we consider a realistic setting where the similarities between examples can change from episode to episode depending on the task context, which is not given to the learner. We de\ufb01ne two new benchmark datasets for this \ufb02exible few-shot scenario, where the tasks are based on images of faces (Celeb-A) and shoes (Zappos50K). While classi\ufb01cation baselines and episodic approaches learn representations that work well for standard few-shot learning, they suffer in our \ufb02exible tasks as novel similarity de\ufb01nitions arise during testing. We propose to build upon recent contrastive unsupervised learning techniques and use a combination of instance and class invariance learning, aiming to obtain general and \ufb02exible features. We \ufb01nd that our approach performs strongly on our new \ufb02exible few-shot learning benchmarks."
                    },
                    {
                        "title": "Lorentzian Distance Learning for Hyperbolic Representations",
                        "abstract": "We introduce an approach to learn representations based on the Lorentzian distance in hyperbolic geometry. Hyperbolic geometry is especially suited to hierarchically-structured datasets, which are prevalent in the real world. Current hyperbolic representation learning methods compare examples with the Poincar\u00e9 distance. They try to minimize the distance of each node in a hierarchy with its descendants while maximizing its distance with other nodes. This formulation produces node representations close to the centroid of their descendants. To obtain efficient and interpretable algorithms, we exploit the fact that the centroid w.r.t the squared Lorentzian distance can be written in closed-form. We show that the Euclidean norm of such a centroid decreases as the curvature of the hyperbolic space decreases. This property makes it appropriate to represent hierarchies where parent nodes minimize the distances to their descendants and have smaller Euclidean norm than their children. Our approach obtains state-of-the-art results in retrieval and classification tasks on different datasets."
                    },
                    {
                        "title": "Learning Latent Subspaces in Variational Autoencoders",
                        "abstract": "Variational autoencoders (VAEs) are widely used deep generative models capable of learning unsupervised latent representations of data. Such representations are often difficult to interpret or control. We consider the problem of unsupervised learning of features correlated to specific labels in a dataset. We propose a VAE-based generative model which we show is capable of extracting features correlated to binary labels in the data and structuring it in a latent subspace which is easy to interpret. Our model, the Conditional Subspace VAE (CSVAE), uses mutual information minimization to learn a low-dimensional latent subspace associated with each label that can easily be inspected and independently manipulated. We demonstrate the utility of the learned representations for attribute manipulation tasks on both the Toronto Face and CelebA datasets."
                    },
                    {
                        "title": "Meta-Learning for Semi-Supervised Few-Shot Classification",
                        "abstract": "In few-shot classification, we are interested in learning algorithms that train a classifier from only a handful of labeled examples. Recent progress in few-shot classification has featured meta-learning, in which a parameterized model for a learning algorithm is defined and trained on episodes representing different classification problems, each with a small labeled training set and its corresponding test set. In this work, we advance this few-shot classification paradigm towards a scenario where unlabeled examples are also available within each episode. We consider two situations: one where all unlabeled examples are assumed to belong to the same set of classes as the labeled examples of the episode, as well as the more challenging situation where examples from other distractor classes are also provided. To address this paradigm, we propose novel extensions of Prototypical Networks (Snell et al., 2017) that are augmented with the ability to use unlabeled examples when producing prototypes. These models are trained in an end-to-end way on episodes, to learn to leverage the unlabeled examples successfully. We evaluate these methods on versions of the Omniglot and miniImageNet benchmarks, adapted to this new framework augmented with unlabeled examples. We also propose a new split of ImageNet, consisting of a large set of classes, with a hierarchical structure. Our experiments confirm that our Prototypical Networks can learn to improve their predictions due to unlabeled examples, much like a semi-supervised algorithm would."
                    },
                    {
                        "title": "Stochastic Segmentation Trees for Multiple Ground Truths",
                        "abstract": "A strong machine learning system should aim to produce the full range of valid interpretations rather than a single mode in tasks involving inherent ambiguity. This is particularly true for image segmentation, in which there are many sensible ways to partition an image into regions. We formulate a tree-structured probabilistic model, the stochastic segmentation tree , that represents a distribution over segmentations of a given image. We train this model by optimizing a novel objective that quanti\ufb01es the degree of match between statistics of the model and ground truth segmentations. Our method allows learning of both the parameters in the tree and the structure itself. We demonstrate on two datasets, including the challenging Berkeley Segmentation Dataset, that our model is able to successfully capture the range of ground truths and to produce novel plausible segmentations beyond those found in the data."
                    },
                    {
                        "title": "Prototypical Networks for Few-shot Learning",
                        "abstract": "We propose Prototypical Networks for the problem of few-shot classification, where a classifier must generalize to new classes not seen in the training set, given only a small number of examples of each new class. Prototypical Networks learn a metric space in which classification can be performed by computing distances to prototype representations of each class. Compared to recent approaches for few-shot learning, they reflect a simpler inductive bias that is beneficial in this limited-data regime, and achieve excellent results. We provide an analysis showing that some simple design decisions can yield substantial improvements over recent approaches involving complicated architectural choices and meta-learning. We further extend Prototypical Networks to zero-shot learning and achieve state-of-the-art results on the CU-Birds dataset."
                    }
                ]
            },
            "6f5fa4f6-43f8-486a-a3ca-53ae557893ab": {
                "pk": "6f5fa4f6-43f8-486a-a3ca-53ae557893ab",
                "name": "Toniann Pitassi",
                "collaborators": [
                    "Thomas P. Zollo",
                    "Zhun Deng",
                    "Jake Snell",
                    "R. Impagliazzo",
                    "Noah Fleming",
                    "Robert Robere",
                    "Rex Lei",
                    "Jessica Sorrell",
                    "Morgan Shirley",
                    "R. Zemel",
                    "Oliver Korten",
                    "Stefan Grosser",
                    "Richard Zemel",
                    "Mark Bun",
                    "Marco Gaboardi",
                    "Max Hopkins",
                    "Satchit Sivakumar",
                    "Lianna Hambardzumyan",
                    "Suhail Sherif",
                    "Adi Shraibman",
                    "Nicola Galesi",
                    "Joshua A. Grochow",
                    "Adrian She",
                    "Kasper Green Larsen",
                    "R. Pagh",
                    "Or Zamir",
                    "A. Shraibman",
                    "Alex Edmonds",
                    "Aleksandar Nikolov",
                    "Sasank Mouli",
                    "Benny Applebaum",
                    "A. Beimel",
                    "O. Nir",
                    "Naty Peter",
                    "Mc",
                    "Prasanna Ramakrishnan",
                    "Li-Yang Tan",
                    "Gal Vardi",
                    "Daniel Reichman",
                    "Ohad Shamir"
                ],
                "domain": [
                    "Computational Complexity",
                    "Machine Learning",
                    "Statistical Learning Theory",
                    "Secret Sharing"
                ],
                "publications": [
                    {
                        "title": "Strong vs. Weak Range Avoidance and the Linear Ordering Principle",
                        "abstract": ","
                    },
                    {
                        "title": "Black-Box PPP Is Not Turing-Closed",
                        "abstract": "The complexity class PPP contains all total search problems many-one reducible to the Pigeon problem, where we are given a succinct encoding of a function mapping n+1 pigeons to n holes, and must output two pigeons that collide in a hole. PPP is one of the \u201coriginal five\u201d syntactically-defined subclasses of TFNP, and has been extensively studied due to the strong connections between its defining problem \u2014 the pigeonhole principle \u2014 and problems in cryptography, extremal combinatorics, proof complexity, and other fields. However, despite its importance, PPP appears to be less robust than the other important TFNP subclasses. In particular, unlike all other major TFNP subclasses, it was conjectured by Buss and Johnson that PPP is not closed under Turing reductions, and they called for a black-box separation in order to provide evidence for this conjecture. The question of whether PPP contains its Turing closure was further highlighted by Daskalakis in his recent IMU Abacus Medal Lecture. In this work we prove that PPP is indeed not Turing-closed in the black-box setting, affirmatively resolving the above conjecture and providing strong evidence that PPP is not Turing-closed. In fact, we are able to separate PPP from its non-adaptive Turing closure, in which all calls to the Pigeon oracle must be made in parallel. This differentiates PPP from all other important TFNP subclasses, and especially from its closely-related subclass PWPP \u2014 defined by reducibility to the weak pigeonhole principle \u2014 which is known to be non-adaptively Turing-closed. Our proof requires developing new tools for PPP lower bounds, and creates new connections between PPP and the theory of pseudoexpectation operators used for Sherali-Adams and Sum-of-Squares lower bounds. In particular, we introduce a new type of pseudoexpectation operator that is precisely tailored for lower bounds against black-box PPP, which may be of independent interest."
                    },
                    {
                        "title": "Improving Predictor Reliability with Selective Recalibration",
                        "abstract": "A reliable deep learning system should be able to accurately express its confidence with respect to its predictions, a quality known as calibration. One of the most effective ways to produce reliable confidence estimates with a pre-trained model is by applying a post-hoc recalibration method. Popular recalibration methods like temperature scaling are typically fit on a small amount of data and work in the model's output space, as opposed to the more expressive feature embedding space, and thus usually have only one or a handful of parameters. However, the target distribution to which they are applied is often complex and difficult to fit well with such a function. To this end we propose \\textit{selective recalibration}, where a selection model learns to reject some user-chosen proportion of the data in order to allow the recalibrator to focus on regions of the input space that can be well-captured by such a model. We provide theoretical analysis to motivate our algorithm, and test our method through comprehensive experiments on difficult medical imaging and zero-shot classification tasks. Our results show that selective recalibration consistently leads to significantly lower calibration error than a wide range of selection and recalibration baselines."
                    },
                    {
                        "title": "Stability Is Stable: Connections between Replicability, Privacy, and Adaptive Generalization",
                        "abstract": "The notion of replicable algorithms was introduced by Impagliazzo, Lei, Pitassi, and Sorrell (STOC\u201922) to describe randomized algorithms that are stable under the resampling of their inputs. More precisely, a replicable algorithm gives the same output with high probability when its randomness is fixed and it is run on a new i.i.d. sample drawn from the same distribution. Using replicable algorithms for data analysis can facilitate the verification of published results by ensuring that the results of an analysis will be the same with high probability, even when that analysis is performed on a new data set. In this work, we establish new connections and separations between replicability and standard notions of algorithmic stability. In particular, we give sample-efficient algorithmic reductions between perfect generalization, approximate differential privacy, and replicability for a broad class of statistical problems. Conversely, we show any such equivalence must break down computationally: there exist statistical problems that are easy under differential privacy, but that cannot be solved replicably without breaking public-key cryptography. Furthermore, these results are tight: our reductions are statistically optimal, and we show that any computational separation between DP and replicability must imply the existence of one-way functions. Our statistical reductions give a new algorithmic framework for translating between notions of stability, which we instantiate to answer several open questions in replicability and privacy. This includes giving sample-efficient replicable algorithms for various PAC learning, distribution estimation, and distribution testing problems, algorithmic amplification of \u03b4 in approximate DP, conversions from item-level to user-level privacy, and the existence of private agnostic-to-realizable learning reductions under structured distributions."
                    },
                    {
                        "title": "An improved protocol for ExactlyN with more than 3 players",
                        "abstract": "The ExactlyN problem in the number-on-forehead (NOF) communication setting asks $k$ players, each of whom can see every input but their own, if the $k$ input numbers add up to $N$. Introduced by Chandra, Furst and Lipton in 1983, ExactlyN is important for its role in understanding the strength of randomness in communication complexity with many players. It is also tightly connected to the field of combinatorics: its $k$-party NOF communication complexity is related to the size of the largest corner-free subset in $[N]^{k-1}$. In 2021, Linial and Shraibman gave more efficient protocols for ExactlyN for 3 players. As an immediate consequence, this also gave a new construction of larger corner-free subsets in $[N]^2$. Later that year Green gave a further refinement to their argument. These results represent the first improvements to the highest-order term for $k=3$ since the famous work of Behrend in 1946. In this paper we give a corresponding improvement to the highest-order term for all $k>3$, the first since Rankin in 1961. That is, we give a more efficient protocol for ExactlyN as well as larger corner-free sets in higher dimensions. Nearly all previous results in this line of research approached the problem from the combinatorics perspective, implicitly resulting in non-constructive protocols for ExactlyN. Approaching the problem from the communication complexity point of view and constructing explicit protocols for ExactlyN was key to the improvements in the $k=3$ setting. As a further contribution we provide explicit protocols for ExactlyN for any number of players which serves as a base for our improvement."
                    },
                    {
                        "title": "Distribution-Free Statistical Dispersion Control for Societal Applications",
                        "abstract": "Explicit finite-sample statistical guarantees on model performance are an important ingredient in responsible machine learning. Previous work has focused mainly on bounding either the expected loss of a predictor or the probability that an individual prediction will incur a loss value in a specified range. However, for many high-stakes applications, it is crucial to understand and control the dispersion of a loss distribution, or the extent to which different members of a population experience unequal effects of algorithmic decisions. We initiate the study of distribution-free control of statistical dispersion measures with societal implications and propose a simple yet flexible framework that allows us to handle a much richer class of statistical functionals beyond previous work. Our methods are verified through experiments in toxic comment detection, medical imaging, and film recommendation."
                    },
                    {
                        "title": "On the algebraic proof complexity of Tensor Isomorphism",
                        "abstract": "The Tensor Isomorphism problem (TI) has recently emerged as having connections to multiple areas of research within complexity and beyond, but the current best upper bound is essentially the brute force algorithm. Being an algebraic problem, TI (or rather, proving that two tensors are non-isomorphic) lends itself very naturally to algebraic and semi-algebraic proof systems, such as the Polynomial Calculus (PC) and Sum of Squares (SoS). For its combinatorial cousin Graph Isomorphism, essentially optimal lower bounds are known for approaches based on PC and SoS (Berkholz&Grohe, SODA '17). Our main results are an $\\Omega(n)$ lower bound on PC degree or SoS degree for Tensor Isomorphism, and a nontrivial upper bound for testing isomorphism of tensors of bounded rank. We also show that PC cannot perform basic linear algebra in sub-linear degree, such as comparing the rank of two matrices, or deriving $BA=I$ from $AB=I$. As linear algebra is a key tool for understanding tensors, we introduce a strictly stronger proof system, PC+Inv, which allows as derivation rules all substitution instances of the implication $AB=I \\rightarrow BA=I$. We conjecture that even PC+Inv cannot solve TI in polynomial time either, but leave open getting lower bounds on PC+Inv for any system of equations, let alone those for TI. We also highlight many other open questions about proof complexity approaches to TI."
                    },
                    {
                        "title": "Optimal Non-Adaptive Cell Probe Dictionaries and Hashing",
                        "abstract": "We present a simple and provably optimal non-adaptive cell probe data structure for the static dictionary problem. Our data structure supports storing a set of n key-value pairs from [u]x[u] using s words of space and answering key lookup queries in t = O(lg(u/n)/ lg(s/n)) nonadaptive probes. This generalizes a solution to the membership problem (i.e., where no values are associated with keys) due to Buhrman et al. We also present matching lower bounds for the non-adaptive static membership problem in the deterministic setting. Our lower bound implies that both our dictionary algorithm and the preceding membership algorithm are optimal, and in particular that there is an inherent complexity gap in these problems between no adaptivity and one round of adaptivity (with which hashing-based algorithms solve these problems in constant time). Using the ideas underlying our data structure, we also obtain the first implementation of a n-wise independent family of hash functions with optimal evaluation time in the cell probe model."
                    },
                    {
                        "title": "The Strength of Equality Oracles in Communication",
                        "abstract": "It is well-known that randomized communication protocols are more powerful than deterministic protocols. In particular the Equality function requires \u2126( n ) deterministic communication complexity but has efficient randomized protocols. Previous work of Chattopadhyay, Lovett and Vinyals shows that randomized communication is strictly stronger than what can be solved by deterministic protocols equipped with an Equality oracle. Despite this separation, we are far from understanding the exact strength of Equality oracles in the context of communication complexity. In this work we focus on nondeterminisic communication equipped with an Equality oracle, which is a subclass of Merlin-Arthur communication. We show that this inclusion is strict by proving that the previously-studied Integer Inner Product function, which can be efficiently computed even with bounded-error randomness, cannot be computed using sublinear communication in the nondeterministic Equality model. To prove this we give a new matrix-theoretic characterization of the nondeterministic Equality model: specifically, there is a tight connection between this model and a covering number based on the blocky matrices of Hambardzumyan, Hatami, and Hatami, as well as a natural variant of the Gamma-2 factorization norm. Similar equivalences are shown for the unambiguous nondeterministic model with Equality oracles. A bonus result arises from these proofs: for the studied communication models, a single Equality oracle call suffices without loss of generality. Our results allow us to prove a separation between deterministic and unambiguous nondeter-minism in the presence of Equality oracles. This stands in contrast to the result of Yannakakis which shows that these models are polynomially-related without oracles. We suggest a number of intriguing open questions along this direction of inquiry, as well as others that arise from our work."
                    },
                    {
                        "title": "Reproducibility in learning",
                        "abstract": "We introduce the notion of a reproducible algorithm in the context of learning. A reproducible learning algorithm is resilient to variations in its samples \u2014 with high probability, it returns the exact same output when run on two samples from the same underlying distribution. We begin by unpacking the definition, clarifying how randomness is instrumental in balancing accuracy and reproducibility. We initiate a theory of reproducible algorithms, showing how reproducibility implies desirable properties such as data reuse and efficient testability. Despite the exceedingly strong demand of reproducibility, there are efficient reproducible algorithms for several fundamental problems in statistics and learning. First, we show that any statistical query algorithm can be made reproducible with a modest increase in sample complexity, and we use this to construct reproducible algorithms for finding approximate heavy-hitters and medians. Using these ideas, we give the first reproducible algorithm for learning halfspaces via a reproducible weak learner and a reproducible boosting algorithm. Interestingly, we utilize a connection to foams as a higher-dimension randomized rounding scheme. Finally, we initiate the study of lower bounds and inherent tradeoffs for reproducible algorithms, giving nearly tight sample complexity upper and lower bounds for reproducible versus nonreproducible SQ algorithms."
                    },
                    {
                        "title": "Learning versus Refutation in Noninteractive Local Differential Privacy",
                        "abstract": "We study two basic statistical tasks in non-interactive local differential privacy (LDP): learning and refutation. Learning requires finding a concept that best fits an unknown target function (from labelled samples drawn from a distribution), whereas refutation requires distinguishing between data distributions that are well-correlated with some concept in the class, versus distributions where the labels are random. Our main result is a complete characterization of the sample complexity of agnostic PAC learning for non-interactive LDP protocols. We show that the optimal sample complexity for any concept class is captured by the approximate $\\gamma_2$~norm of a natural matrix associated with the class. Combined with previous work [Edmonds, Nikolov and Ullman, 2019] this gives an equivalence between learning and refutation in the agnostic setting."
                    },
                    {
                        "title": "Lower bounds for Polynomial Calculus with extension variables over finite fields",
                        "abstract": "For every prime p > 0, every n > 0 and \u03ba = O (log n ), we show the existence of an unsatisfiable system of polynomial equations over O ( n log n ) variables of degree O (log n ) such that any Polynomial Calculus refutation over F p with M extension variables, each depending on at most \u03ba original variables requires size exp (cid:0) \u2126( n 2 ) / 10 \u03ba ( M + n log n ) (cid:1)"
                    },
                    {
                        "title": "Quantile Risk Control: A Flexible Framework for Bounding the Probability of High-Loss Predictions",
                        "abstract": "Rigorous guarantees about the performance of predictive algorithms are necessary in order to ensure their responsible use. Previous work has largely focused on bounding the expected loss of a predictor, but this is not sufficient in many risk-sensitive applications where the distribution of errors is important. In this work, we propose a flexible framework to produce a family of bounds on quantiles of the loss distribution incurred by a predictor. Our method takes advantage of the order statistics of the observed loss values rather than relying on the sample mean alone. We show that a quantile is an informative way of quantifying predictive performance, and that our framework applies to a variety of quantile-based metrics, each targeting important subsets of the data distribution. We analyze the theoretical properties of our proposed method and demonstrate its ability to rigorously control loss quantiles on several real-world datasets."
                    },
                    {
                        "title": "Secret Sharing, Slice Formulas, and Monotone Real Circuits",
                        "abstract": "A secret-sharing scheme allows to distribute a secret s among n parties such that only some pre-de\ufb01ned \u201cauthorized\u201d sets of parties can reconstruct the secret, and all other \u201cunauthorized\u201d sets learn nothing about s . For over 30 years, it was known that any (monotone) collection of authorized sets can be realized by a secret-sharing scheme whose shares are of size 2 n \u2212 o ( n ) and until recently no better scheme was known. In a recent breakthrough, Liu and Vaikuntanathan (STOC 2018) have reduced the share size to 2 0 . 994 n + o ( n ) , and this was further improved by several follow-ups accumulating in an upper bound of 1 . 5 n + o ( n ) (Applebaum and Nir, CRYPTO 2021). Following these advances, it is natural to ask whether these new approaches can lead to a truly sub-exponential upper-bound of 2 n 1 \u2212 \u03b5 for some constant \u03b5 > 0 , or even all the way down to polynomial upper-bounds. In this paper, we relate this question to the complexity of computing monotone Boolean functions by monotone real circuits (MRCs) \u2013 a computational model that was introduced by Pudl\u00e1k (J. Symb. Log., 1997) in the context of proof complexity. We introduce a new notion of \u201cseparable\u201d MRCs that lies between monotone real circuits and monotone real formulas (MRFs). As our main results, we show that recent constructions of general secret-sharing schemes implicitly give rise to separable MRCs for general monotone functions of similar complexity, and that some monotone functions (in monotone NP) cannot be computed by sub-exponential size separable MRCs. Interestingly, it seems that proving similar lower-bounds for general MRCs is beyond the reach of current techniques. We use this connection to obtain lower-bounds against a natural family of secret-sharing schemes, as well as new non-trivial upper-bounds for MRCs. Speci\ufb01cally, we conclude that recent approaches for secret-sharing schemes cannot achieve sub-exponential share size and that every monotone function can be realized by an MRC (or even MRF) of complexity 1 . 5 n + o ( n ) . To the best of our knowledge, this is the \ufb01rst improvement over the trivial 2 n \u2212 o ( n ) upper-bound. Along the way, we show that the recent constructions of general secret-sharing schemes implicitly give rise to Boolean formulas over slice functions and prove that such formulas can be simulated by separable MRCs of similar size. On a conceptual level, our paper continues the rich line of study that relates the share size of secret-sharing schemes to monotone complexity measures."
                    },
                    {
                        "title": "Random \u0398( log n ) -CNFs are Hard for Cuting Planes",
                        "abstract": "The random k -SAT model is one of the most important and well-studied distributions over k -SAT instances. It is closely connected to statistical physics and is a benchmark for satisiability algorithms. We show that when k = \u0398( log n ) , any Cutting Planes refutation for random k -SAT requires exponential length in the regime where the number of clauses guarantees that the formula is unsatisiable with high probability."
                    },
                    {
                        "title": ") Fall , 2022 Herbrand Theorem , Equality , and Compactness The Herbrand Theorem",
                        "abstract": "Definitions (quantifier-free, \u2200-sentence, ground instance): A formula A is quantifierfree if A has no occurrence of either of the quantifiers \u2200 or \u2203. A \u2200-sentence is a sentence of the form \u2200x1...\u2200xkB where k \u2265 0 and B is a quantifier-free formula. A ground instance of this sentence is a sentence of the form B(t1/x1)(t2/x2)...(tk/xk), where t1, ..., tk are ground terms (i.e. terms with no variables) from the underlying language."
                    },
                    {
                        "title": "Tradeoffs for small-depth Frege proofs",
                        "abstract": "We study the complexity of small-depth Frege proofs and give the first tradeoffs between the size of each line and the number of lines. Existing lower bounds apply to the overall proof size-the sum of sizes of all lines-and do not distinguish between these notions of complexity. For depth-d Frege proofs of the Tseitin principle where each line is a size-s formula, we prove that $\\exp(n/2^{\\Omega(d\\sqrt{\\log s})})$ many lines are necessary. This yields new lower bounds on line complexity that are not implied by $\\mathbf{H}\\mathop{\\mathbf{a}}\\!\\!\\!\\!^{\\circ}\\mathbf{stad}$'s recent $\\exp(n^{\\Omega(1/d)})$ lower bound on the overall proof size. For $s$= poly $(n)$, for example, our lower bound remains $\\exp(n^{1-o(1)})$ for all $d=o(\\sqrt{\\log n})$, whereas $\\mathbf{H}\\mathop{\\mathbf{a}}\\!\\!\\!\\!^{\\circ}\\mathbf{stad}$'s lower bound is $\\exp(n^{o(1)})$ once $d\\ = \\omega_{n}(1)$. Our main conceptual contribution is the simple obser-vation that techniques for establishing correlation bounds in circuit complexity can be leveraged to establish such tradeoffs in proof complexity."
                    },
                    {
                        "title": "Size and Depth Separation in Approximating Natural Functions with Neural Networks",
                        "abstract": "When studying the expressive power of neural networks, a main challenge is to understand how the size and depth of the network affect its ability to approximate real functions. However, not all functions are interesting from a practical viewpoint: functions of interest usually have a polynomially-bounded Lipschitz constant, and can be computed efficiently. We call functions that satisfy these conditions\"benign\", and explore the benefits of size and depth for approximation of benign functions with ReLU networks. As we show, this problem is more challenging than the corresponding problem for non-benign functions. We give barriers to showing depth-lower-bounds: Proving existence of a benign function that cannot be approximated by polynomial-size networks of depth $4$ would settle longstanding open problems in computational complexity. It implies that beyond depth $4$ there is a barrier to showing depth-separation for benign functions, even between networks of constant depth and networks of nonconstant depth. We also study size-separation, namely, whether there are benign functions that can be approximated with networks of size $O(s(d))$, but not with networks of size $O(s'(d))$. We show a complexity-theoretic barrier to proving such results beyond size $O(d\\log^2(d))$, but also show an explicit benign function, that can be approximated with networks of size $O(d)$ and not with networks of size $o(d/\\log d)$. For approximation in $L_\\infty$ we achieve such separation already between size $O(d)$ and size $o(d)$. Moreover, we show superpolynomial size lower bounds and barriers to such lower bounds, depending on the assumptions on the function. Our size-separation results rely on an analysis of size lower bounds for Boolean functions, which is of independent interest: We show linear size lower bounds for computing explicit Boolean functions with neural networks and threshold circuits."
                    }
                ]
            },
            "87fb6f53-0e78-4439-86ab-bafd096ece85": {
                "pk": "87fb6f53-0e78-4439-86ab-bafd096ece85",
                "name": "Richard Zemel",
                "collaborators": [
                    "Rahul Gupta",
                    "Ninareh Mehrabi",
                    "Palash Goyal",
                    "A. Galstyan",
                    "Kai-Wei Chang",
                    "Charith Peris",
                    "Anil Ramakrishna",
                    "Jimit Majmudar",
                    "Jixuan Wang",
                    "Clement Chung",
                    "J. Dhamala",
                    "Tao Meng",
                    "Zhewei Sun",
                    "Qian Hu",
                    "Yang Xu",
                    "Thomas P. Zollo",
                    "Zhun Deng",
                    "Jake Snell",
                    "T. Pitassi",
                    "Tiantian Feng",
                    "Morteza Ziyadi",
                    "Anaelia Ovalle",
                    "Yuval Pinter",
                    "Junyi Li",
                    "Jack Good",
                    "Christophe Dupuy",
                    "Shalini Ghosh",
                    "Eleni Trianta\ufb01llou",
                    "Vincent Dumoulin",
                    "Hugo Larochelle"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Federated Learning",
                    "Machine Learning",
                    "Model Calibration"
                ],
                "publications": [
                    {
                        "title": "Attribute Controlled Fine-tuning for Large Language Models: A Case Study on Detoxification",
                        "abstract": "We propose a constraint learning schema for fine-tuning Large Language Models (LLMs) with attribute control. Given a training corpus and control criteria formulated as a sequence-level constraint on model outputs, our method fine-tunes the LLM on the training corpus while enhancing constraint satisfaction with minimal impact on its utility and generation quality. Specifically, our approach regularizes the LLM training by penalizing the KL divergence between the desired output distribution, which satisfies the constraints, and the LLM's posterior. This regularization term can be approximated by an auxiliary model trained to decompose the sequence-level constraints into token-level guidance, allowing the term to be measured by a closed-form formulation. To further improve efficiency, we design a parallel scheme for concurrently updating both the LLM and the auxiliary model. We evaluate the empirical performance of our approach by controlling the toxicity when training an LLM. We show that our approach leads to an LLM that produces fewer inappropriate responses while achieving competitive performance on benchmarks and a toxicity detection task."
                    },
                    {
                        "title": "Toward Informal Language Processing: Knowledge of Slang in Large Language Models",
                        "abstract": "Recent advancement in large language models (LLMs) has offered a strong potential for natural language systems to process informal language. A representative form of informal language is slang, used commonly in daily conversations and online social media. To date, slang has not been comprehensively evaluated in LLMs due partly to the absence of a carefully designed and publicly accessible benchmark. Using movie subtitles, we construct a dataset that supports evaluation on a diverse set of tasks pertaining to automatic processing of slang. For both evaluation and finetuning, we show the effectiveness of our dataset on two core applications: 1) slang detection, and 2) identification of regional and historical sources of slang from natural sentences. We also show how our dataset can be used to probe the output distributions of LLMs for interpretive insights. We find that while LLMs such as GPT-4 achieve good performance in a zero-shot setting, smaller BERT-like models finetuned on our dataset achieve comparable performance. Furthermore, we show that our dataset enables finetuning of LLMs such as GPT-3.5 that achieve substantially better performance than strong zero-shot baselines. Our work offers a comprehensive evaluation and a high-quality benchmark on English slang based on the OpenSubtitles corpus, serving both as a publicly accessible resource and a platform for applying tools for informal language processing."
                    },
                    {
                        "title": "Improving Predictor Reliability with Selective Recalibration",
                        "abstract": "A reliable deep learning system should be able to accurately express its confidence with respect to its predictions, a quality known as calibration. One of the most effective ways to produce reliable confidence estimates with a pre-trained model is by applying a post-hoc recalibration method. Popular recalibration methods like temperature scaling are typically fit on a small amount of data and work in the model's output space, as opposed to the more expressive feature embedding space, and thus usually have only one or a handful of parameters. However, the target distribution to which they are applied is often complex and difficult to fit well with such a function. To this end we propose \\textit{selective recalibration}, where a selection model learns to reject some user-chosen proportion of the data in order to allow the recalibrator to focus on regions of the input space that can be well-captured by such a model. We provide theoretical analysis to motivate our algorithm, and test our method through comprehensive experiments on difficult medical imaging and zero-shot classification tasks. Our results show that selective recalibration consistently leads to significantly lower calibration error than a wide range of selection and recalibration baselines."
                    },
                    {
                        "title": "Partial Federated Learning",
                        "abstract": "Federated Learning (FL) is a popular algorithm to train machine learning models on user data constrained to edge devices (for example, mobile phones) due to privacy concerns. Typically, FL is trained with the assumption that no part of the user data can be egressed from the edge. However, in many production settings, specific data-modalities/meta-data are limited to be on device while others are not. For example, in commercial SLU systems, it is typically desired to prevent transmission of biometric signals (such as audio recordings of the input prompt) to the cloud, but egress of locally (i.e. on the edge device) transcribed text to the cloud may be possible. In this work, we propose a new algorithm called Partial Federated Learning (PartialFL), where a machine learning model is trained using data where a subset of data modalities or their intermediate representations can be made available to the server. We further restrict our model training by preventing the egress of data labels to the cloud for better privacy, and instead use a contrastive learning based model objective. We evaluate our approach on two different multi-modal datasets and show promising results with our proposed approach."
                    },
                    {
                        "title": "Tokenization Matters: Navigating Data-Scarce Tokenization for Gender Inclusive Language Technologies",
                        "abstract": "Gender-inclusive NLP research has documented the harmful limitations of gender binary-centric large language models (LLM), such as the inability to correctly use gender-diverse English neopronouns (e.g., xe, zir, fae). While data scarcity is a known culprit, the precise mechanisms through which scarcity affects this behavior remain underexplored. We discover LLM misgendering is significantly influenced by Byte-Pair Encoding (BPE) tokenization, the tokenizer powering many popular LLMs. Unlike binary pronouns, BPE overfragments neopronouns, a direct consequence of data scarcity during tokenizer training. This disparate tokenization mirrors tokenizer limitations observed in multilingual and low-resource NLP, unlocking new misgendering mitigation strategies. We propose two techniques: (1) pronoun tokenization parity, a method to enforce consistent tokenization across gendered pronouns, and (2) utilizing pre-existing LLM pronoun knowledge to improve neopronoun proficiency. Our proposed methods outperform finetuning with standard BPE, improving neopronoun accuracy from 14.1% to 58.4%. Our paper is the first to link LLM misgendering to tokenization and deficient neopronoun grammar, indicating that LLMs unable to correctly treat neopronouns as pronouns are more prone to misgender."
                    },
                    {
                        "title": "The steerability of large language models toward data-driven personas",
                        "abstract": "Large language models (LLMs) are known to generate biased responses where the opinions of certain groups and populations are underrepresented. Here, we present a novel approach to achieve controllable generation of specific viewpoints using LLMs, that can be leveraged to produce multiple perspectives and to reflect the diverse opinions. Moving beyond the traditional reliance on demographics like age, gender, or party affiliation, we introduce a data-driven notion of persona grounded in collaborative filtering, which is defined as either a single individual or a cohort of individuals manifesting similar views across specific inquiries. As individuals in the same demographic group may have different personas, our data-driven persona definition allows for a more nuanced understanding of different (latent) social groups present in the population. In addition to this, we also explore an efficient method to steer LLMs toward the personas that we define. We show that our data-driven personas significantly enhance model steerability, with improvements of between 57%-77% over our best performing baselines."
                    },
                    {
                        "title": "Coordinated Replay Sample Selection for Continual Federated Learning",
                        "abstract": "Continual Federated Learning (CFL) combines Federated Learning (FL), the decentralized learning of a central model on a number of client devices that may not communicate their data, and Continual Learning (CL), the learning of a model from a continual stream of data without keeping the entire history. In CL, the main challenge is \\textit{forgetting} what was learned from past data. While replay-based algorithms that keep a small pool of past training data are effective to reduce forgetting, only simple replay sample selection strategies have been applied to CFL in prior work, and no previous work has explored coordination among clients for better sample selection. To bridge this gap, we adapt a replay sample selection objective based on loss gradient diversity to CFL and propose a new relaxation-based selection of samples to optimize the objective. Next, we propose a practical algorithm to coordinate gradient-based replay sample selection across clients without communicating private data. We benchmark our coordinated and uncoordinated replay sample selection algorithms against random sampling-based baselines with language models trained on a large scale de-identified real-world text dataset. We show that gradient-based sample selection methods both boost performance and reduce forgetting compared to random sampling methods, with our coordination method showing gains early in the low replay size regime (when the budget for storing past data is small)."
                    },
                    {
                        "title": "JAB: Joint Adversarial Prompting and Belief Augmentation",
                        "abstract": "With the recent surge of language models in different applications, attention to safety and robustness of these models has gained significant importance. Here we introduce a joint framework in which we simultaneously probe and improve the robustness of a black-box target model via adversarial prompting and belief augmentation using iterative feedback loops. This framework utilizes an automated red teaming approach to probe the target model, along with a belief augmenter to generate instructions for the target model to improve its robustness to those adversarial probes. Importantly, the adversarial model and the belief generator leverage the feedback from past interactions to improve the effectiveness of the adversarial prompts and beliefs, respectively. In our experiments, we demonstrate that such a framework can reduce toxic content generation both in dynamic cases where an adversary directly interacts with a target model and static cases where we use a static benchmark dataset to evaluate our model."
                    },
                    {
                        "title": "Learning Flexible Classi\ufb01ers with Shot-CONditional Episodic (SCONE) Training",
                        "abstract": "Early few-shot classi\ufb01cation work advocates for episodic training, i.e. training over learning episodes each posing a few-shot classi\ufb01cation task. However, the role of this training regime remains poorly understood. Standard classi\ufb01cation methods (\u201cpre-training\u201d) followed by episodic \ufb01ne-tuning have recently achieved strong results. We aim to understand the role of this episodic \ufb01ne-tuning phase through an exploration of the effect of the \u201cshot\u201d (number of examples per class) that is used during \ufb01ne-tuning. We discover that using a \ufb01xed shot can specialize the pre-trained model to solving episodes of that shot at the expense of performance on other shots, in agreement with a trade-off recently observed in the context of end-to-end episodic training. To amend this, we propose a shot-conditional form of episodic \ufb01ne-tuning, inspired from recent work that trains a single model on a distribution of losses. We show that this \ufb02exible approach consitutes an effective general solution that does not suffer disproportionately on any shot. We then subject it to the large-scale Meta-Dataset benchmark of varying shots and imbalanced episodes and observe performance gains in that challenging environment."
                    }
                ]
            }
        }
    },
    "2406.06494": {
        "paper_data": {
            "title": "Scaling Continuous Latent Variable Models as Probabilistic Integral Circuits",
            "url": "http://arxiv.org/abs/2406.06494v1",
            "arxiv_id": "2406.06494",
            "authors": [
                "Gennaro Gala",
                "Cassio de Campos",
                "Antonio Vergari",
                "Erik Quaeghebeur"
            ],
            "abstract": "Probabilistic integral circuits (PICs) have been recently introduced as probabilistic models enjoying the key ingredient behind expressive generative models: continuous latent variables (LVs). PICs are symbolic computational graphs defining continuous LV models as hierarchies of functions that are summed and multiplied together, or integrated over some LVs. They are tractable if LVs can be analytically integrated out, otherwise they can be approximated by tractable probabilistic circuits (PC) encoding a hierarchical numerical quadrature process, called QPCs.   So far, only tree-shaped PICs have been explored, and training them via numerical quadrature requires memory-intensive processing at scale. In this paper, we address these issues, and present: (i) a pipeline for building DAG-shaped PICs out of arbitrary variable decompositions, (ii) a procedure for training PICs using tensorized circuit architectures, and (iii) neural functional sharing techniques to allow scalable training. In extensive experiments, we showcase the effectiveness of functional sharing and the superiority of QPCs over traditional PCs.",
            "introduction": " Introduction Continuous latent variables (LVs) are arguably the key ingredient behind many successful genera- tive models, from variational autoencoders [24] to generative adversarial networks [20], and more recently diffusion models [56]. While these models allow to learn expressive distributions from data, they are limited to sampling and require task-specific approximations when it comes to per- form probabilistic reasoning , as even simple tasks such as computing marginals or conditionals are intractable for them. On the other hand, performing these tasks can be tractable for (hierarchical) discrete LV models [4, 3], but these prove to be more challenging to learn at scale [9, 10, 32, 33]. This inherent trade-off among tractability, ease of learning, and expressiveness can be analyzed and explored with probabilistic integral circuits (PICs) [18], a recently introduced class of deep generative models defining hierarchies of continuous LVs using symbolic functional circuits. PICs are tractable when their continuous LVs can be analytically integrated out. Intractable PICs can however be systematically approximated as (tensorized) probabilistic circuits (PCs) [53, 4], the representation language of discrete LV models. An instance of such PCs encodes a hierarchical numerical quadrature process of the PIC to approximate, and as such is called quadrature PC (QPC). Distilling QPCs from PICs has proven to be an effective alternative way to train PCs, but it has only been explored for tree-shaped PICs, as building and scaling to richer LV structures is an open research question that requires new tools [18]. In this paper, we fill this gap by redefining the se- mantics of PICs and extending them to DAG-shaped hierarchies of continuous LVs. Specifically, we \u2217Corresponding author: g.gala@tue.nl .\u222bShared supervision Preprint. Under review.arXiv:2406.06494v1  [cs.LG]  10 Jun 2024design PICs as a language to represent hierarchical quasi-tensors factorizations [49], parameterized by light-weight multi-layer perceptrons. Contributions. (1) We present a systematic pipeline to build DAG-shaped PICs, starting from ar- bitrary variable decompositions (Section 3.1). (2)We show how to learn and approximate PICs via a hierarchical quadrature process which we encode in tensorized QPCs that match certain circuit architectures proposed in different prior works [43, 42, 31, 39] (Section 3.2). (3)We present func- tional sharing techniques to scale the training of PICs, which lead us to parameterize them with multi-headed multi-layer perceptrons (MLPs) requiring comparable resources as PCs (Section 3.3). (4)In extensive Appendix C.3. From Table 1, we see that QPCs prove again very competitive, consistently outper- forming standard PCs commonly trained with Adam, and the best performing PC from the literature, HCLT, which is trained via EM schemes and patch-wise Experiments In our results are averaged over 5 different runs. QT-CP-256 QG-CP-256 QG-TK-32 QPC PC QPC PC QPC PC CelebA\u22174.810\u00b10.004 4 .851\u00b10.002 4.739\u00b10.002 4 .781\u00b10.002 5.352\u00b10.002 5 .364\u00b10.002 CelebA\u20204.159\u00b10.003 4 .215\u00b10.003 4.114\u00b10.003 4 .155\u00b10.006 4.720\u00b10.003 5.718\u00b10.001 ImgNet64\u22175.143\u00b10.003 5 .221\u00b10.003 5.051\u00b10.002 5 .220\u00b10.005 5.657\u00b10.004 5 .764\u00b10.001 ImgNet64\u20204.523\u00b10.003 4 .591\u00b10.002 4.425\u00b10.004 4 .590\u00b10.006 5.011\u00b10.007 5 .138\u00b10.001 21 methods [32]. Furthermore, QPCs are close to PCs trained via latent variable distillation (LVD) [32, 33], a framework that requires extra supervi- sion over their latent spaces by distilling information from existing deep generative models (DGMs). This technique requires pre-trained DGMs, several heuristics, and a final fine-tuning stage via EM or SGD, while PIC training method is instead end-to-end and self-contained. 5 Discussion & Conclusion With this work, we systematized the construction of PICs, extending them to DAG-like structures (Section 3.1),",
            "references": [
                {
                    "title": "Probabilistic Integral Circuits",
                    "abstract": "Continuous latent variables (LVs) are a key ingredient of many generative models, as they allow modelling expressive mixtures with an uncountable number of components. In contrast, probabilistic circuits (PCs) are hierarchical discrete mixtures represented as computational graphs composed of input, sum and product units. Unlike continuous LV models, PCs provide tractable inference but are limited to discrete LVs with categorical (i.e. unordered) states. We bridge these model classes by introducing probabilistic integral circuits (PICs), a new language of computational graphs that extends PCs with integral units representing continuous LVs. In the first place, PICs are symbolic computational graphs and are fully tractable in simple cases where analytical integration is possible. In practice, we parameterise PICs with light-weight neural nets delivering an intractable hierarchical continuous mixture that can be approximated arbitrarily well with large PCs using numerical quadrature. On several distribution estimation benchmarks, we show that such PIC-approximating PCs systematically outperform PCs commonly learned via expectation-maximization or SGD."
                },
                {
                    "title": "Subtractive Mixture Models via Squaring: Representation and Learning",
                    "abstract": "Mixture models are traditionally represented and learned by adding several distributions as components. Allowing mixtures to subtract probability mass or density can drastically reduce the number of components needed to model complex distributions. However, learning such subtractive mixtures while ensuring they still encode a non-negative function is challenging. We investigate how to learn and perform inference on deep subtractive mixtures by squaring them. We do this in the framework of probabilistic circuits, which enable us to represent tensorized mixtures and generalize several other subtractive models. We theoretically prove that the class of squared circuits allowing subtractions can be exponentially more expressive than traditional additive mixtures; and, we empirically show this increased expressiveness on a series of real-world distribution estimation tasks."
                },
                {
                    "title": "How to Turn Your Knowledge Graph Embeddings into Generative Models via Probabilistic Circuits",
                    "abstract": "Some of the most successful knowledge graph embedding (KGE) models for link prediction -- CP, RESCAL, TuckER, ComplEx -- can be interpreted as energy-based models. Under this perspective they are not amenable for exact maximum-likelihood estimation (MLE), sampling and struggle to integrate logical constraints. This work re-interprets the score functions of these KGEs as circuits -- constrained computational graphs allowing efficient marginalisation. Then, we design two recipes to obtain efficient generative circuit models by either restricting their activations to be non-negative or squaring their outputs. Our interpretation comes with little or no loss of performance for link prediction, while the circuits framework unlocks exact learning by MLE, efficient sampling of new triples, and guarantee that logical constraints are satisfied by design. Furthermore, our models scale more gracefully than the original KGEs on graphs with millions of entities."
                },
                {
                    "title": "Tractable and Expressive Generative Models of Genetic Variation Data",
                    "abstract": "Population genetic studies often rely on artificial genomes (AGs) simulated by generative models of genetic data. In recent years, unsupervised learning models, based on hidden Markov models, deep generative adversarial networks, restricted Boltzmann machines, and variational autoencoders, have gained popularity due to their ability to generate AGs closely resembling empirical data. These models, however, present a tradeoff between expressivity and tractability. Here, we propose to use hidden Chow-Liu trees (HCLTs) and their representation as probabilistic circuits (PCs) as a solution to this tradeoff. We first learn an HCLT structure that captures the long-range dependencies among SNPs in the training data set. We then convert the HCLT to its equivalent PC as a means of supporting tractable and efficient probabilistic inference. The parameters in these PCs are inferred with an expectation-maximization algorithm using the training data. Compared to other models for generating AGs, HCLT obtains the largest log-likelihood on test genomes across SNPs chosen across the genome and from a contiguous genomic region. Moreover, the AGs generated by HCLT more accurately resemble the source data set in their patterns of allele frequencies, linkage disequilibrium, pairwise haplotype distances, and population structure. This work not only presents a new and robust AG simulator but also manifests the potential of PCs in population genetics."
                },
                {
                    "title": "Tractable Control for Autoregressive Language Generation",
                    "abstract": "Despite the success of autoregressive large language models in text generation, it remains a major challenge to generate text that satisfies complex constraints: sampling from the conditional distribution ${\\Pr}(\\text{text} | \\alpha)$ is intractable for even the simplest lexical constraints $\\alpha$. To overcome this challenge, we propose to use tractable probabilistic models (TPMs) to impose lexical constraints in autoregressive text generation models, which we refer to as GeLaTo (Generating Language with Tractable Constraints). To demonstrate the effectiveness of this framework, we use distilled hidden Markov models, where we can efficiently compute ${\\Pr}(\\text{text} | \\alpha)$, to guide autoregressive generation from GPT2. GeLaTo achieves state-of-the-art performance on challenging benchmarks for constrained text generation (e.g., CommonGen), beating various strong baselines by a large margin. Our work not only opens up new avenues for controlling large language models but also motivates the development of more expressive TPMs."
                },
                {
                    "title": "Bayesian Structure Scores for Probabilistic Circuits",
                    "abstract": "Probabilistic circuits (PCs) are a prominent representation of probability distributions with tractable inference. While parameter learning in PCs is rigorously studied, structure learning is often more based on heuristics than on principled objectives. In this paper, we develop Bayesian structure scores for deterministic PCs, i.e., the structure likelihood with parameters marginalized out, which are well known as rigorous objectives for structure learning in probabilistic graphical models. When used within a greedy cutset algorithm, our scores effectively protect against overfitting and yield a fast and almost hyper-parameter-free structure learner, distinguishing it from previous approaches. In experiments, we achieve good trade-offs between training time and model fit in terms of log-likelihood. Moreover, the principled nature of Bayesian scores unlocks PCs for accommodating frameworks such as structural expectation-maximization."
                },
                {
                    "title": "Understanding the Distillation Process from Deep Generative Models to Tractable Probabilistic Circuits",
                    "abstract": "Probabilistic Circuits (PCs) are a general and unified computational framework for tractable probabilistic models that support efficient computation of various inference tasks (e.g., computing marginal probabilities). Towards enabling such reasoning capabilities in complex real-world tasks, Liu et al. (2022) propose to distill knowledge (through latent variable assignments) from less tractable but more expressive deep generative models. However, it is still unclear what factors make this distillation work well. In this paper, we theoretically and empirically discover that the performance of a PC can exceed that of its teacher model. Therefore, instead of performing distillation from the most expressive deep generative model, we study what properties the teacher model and the PC should have in order to achieve good distillation performance. This leads to a generic algorithmic improvement as well as other data-type-specific ones over the existing latent variable distillation pipeline. Empirically, we outperform SoTA TPMs by a large margin on challenging image modeling benchmarks. In particular, on ImageNet32, PCs achieve 4.06 bits-per-dimension, which is only 0.34 behind variational diffusion models (Kingma et al., 2021)."
                },
                {
                    "title": "Sparse Probabilistic Circuits via Pruning and Growing",
                    "abstract": "Probabilistic circuits (PCs) are a tractable representation of probability distributions allowing for exact and efficient computation of likelihoods and marginals. There has been significant recent progress on improving the scale and expressiveness of PCs. However, PC training performance plateaus as model size increases. We discover that most capacity in existing large PC structures is wasted: fully-connected parameter layers are only sparsely used. We propose two operations: pruning and growing, that exploit the sparsity of PC structures. Specifically, the pruning operation removes unimportant sub-networks of the PC for model compression and comes with theoretical guarantees. The growing operation increases model capacity by increasing the size of the latent space. By alternatingly applying pruning and growing, we increase the capacity that is meaningfully used, allowing us to significantly scale up PC learning. Empirically, our learner achieves state-of-the-art likelihoods on MNIST-family image datasets and on Penn Tree Bank language data compared to other PC learners and less tractable deep generative models such as flow-based models and variational autoencoders (VAEs)."
                },
                {
                    "title": "Scaling Up Probabilistic Circuits by Latent Variable Distillation",
                    "abstract": "Probabilistic Circuits (PCs) are a unified framework for tractable probabilistic models that support efficient computation of various probabilistic queries (e.g., marginal probabilities). One key challenge is to scale PCs to model large and high-dimensional real-world datasets: we observe that as the number of parameters in PCs increases, their performance immediately plateaus. This phenomenon suggests that the existing optimizers fail to exploit the full expressive power of large PCs. We propose to overcome such bottleneck by latent variable distillation: we leverage the less tractable but more expressive deep generative models to provide extra supervision over the latent variables of PCs. Specifically, we extract information from Transformer-based generative models to assign values to latent variables of PCs, providing guidance to PC optimizers. Experiments on both image and language modeling benchmarks (e.g., ImageNet and WikiText-2) show that latent variable distillation substantially boosts the performance of large PCs compared to their counterparts without latent variable distillation. In particular, on the image modeling benchmarks, PCs achieve competitive performance against some of the widely-used deep generative models, including variational autoencoders and flow-based models, opening up new avenues for tractable generative modeling."
                },
                {
                    "title": "Continuous Mixtures of Tractable Probabilistic Models",
                    "abstract": "Probabilistic models based on continuous latent spaces, such as variational autoencoders, can be understood as uncountable mixture models where components depend continuously on the latent code. They have proven to be expressive tools for generative and probabilistic modelling, but are at odds with tractable probabilistic inference, that is, computing marginals and conditionals of the represented probability distribution. Meanwhile, tractable probabilistic models such as probabilistic circuits (PCs) can be understood as hierarchical discrete mixture models, and thus are capable of performing exact inference efficiently but often show subpar performance in comparison to continuous latent-space models. In this paper, we investigate a hybrid approach, namely continuous mixtures of tractable models with a small latent dimension. While these models are analytically intractable, they are well amenable to numerical integration schemes based on a finite set of integration points. With a large enough number of integration points the approximation becomes de-facto exact. Moreover, for a finite set of integration points, the integration method effectively compiles the continuous mixture into a standard PC. In experiments, we show that this simple scheme proves remarkably effective, as PCs learnt this way set new state of the art for tractable models on many standard density estimation benchmarks."
                },
                {
                    "title": "Diffusion Models: A Comprehensive Survey of Methods and Applications",
                    "abstract": "Diffusion models have emerged as a powerful new family of deep generative models with record-breaking performance in many applications, including image synthesis, video generation, and molecule design. In this survey, we provide an overview of the rapidly expanding body of work on diffusion models, categorizing the research into three key areas: efficient sampling, improved likelihood estimation, and handling data with special structures. We also discuss the potential for combining diffusion models with other generative models for enhanced results. We further review the wide-ranging applications of diffusion models in fields spanning from computer vision, natural language processing, temporal data modeling, to interdisciplinary applications in other scientific disciplines. This survey aims to provide a contextualized, in-depth look at the state of diffusion models, identifying the key areas of focus and pointing to potential areas for further exploration. Github: https://github.com/YangLing0818/Diffusion-Models-Papers-Survey-Taxonomy"
                },
                {
                    "title": "Lossless Compression with Probabilistic Circuits",
                    "abstract": "Despite extensive progress on image generation, common deep generative model architectures are not easily applied to lossless compression. For example, VAEs suffer from a compression cost overhead due to their latent variables. This overhead can only be partially eliminated with elaborate schemes such as bits-back coding, often resulting in poor single-sample compression rates. To overcome such problems, we establish a new class of tractable lossless compression models that permit efficient encoding and decoding: Probabilistic Circuits (PCs). These are a class of neural networks involving $|p|$ computational units that support efficient marginalization over arbitrary subsets of the $D$ feature dimensions, enabling efficient arithmetic coding. We derive efficient encoding and decoding schemes that both have time complexity $\\mathcal{O} (\\log(D) \\cdot |p|)$, where a naive scheme would have linear costs in $D$ and $|p|$, making the approach highly scalable. Empirically, our PC-based (de)compression algorithm runs 5-40 times faster than neural compression algorithms that achieve similar bitrates. By scaling up the traditional PC structure learning pipeline, we achieve state-of-the-art results on image datasets such as MNIST. Furthermore, PCs can be naturally integrated with existing neural compression algorithms to improve the performance of these base models on natural image datasets. Our results highlight the potential impact that non-standard learning architectures may have on neural data compression."
                },
                {
                    "title": "Tractable Regularization of Probabilistic Circuits",
                    "abstract": "Probabilistic Circuits (PCs) are a promising avenue for probabilistic modeling. They combine advantages of probabilistic graphical models (PGMs) with those of neural networks (NNs). Crucially, however, they are tractable probabilistic models, supporting efficient and exact computation of many probabilistic inference queries, such as marginals and MAP. Further, since PCs are structured computation graphs, they can take advantage of deep-learning-style parameter updates, which greatly improves their scalability. However, this innovation also makes PCs prone to overfitting, which has been observed in many standard benchmarks. Despite the existence of abundant regularization techniques for both PGMs and NNs, they are not effective enough when applied to PCs. Instead, we re-think regularization for PCs and propose two intuitive techniques, data softening and entropy regularization, that both take advantage of PCs' tractability and still have an efficient implementation as a computation graph. Specifically, data softening provides a principled way to add uncertainty in datasets in closed form, which implicitly regularizes PC parameters. To learn parameters from a softened dataset, PCs only need linear time by virtue of their tractability. In entropy regularization, the exact entropy of the distribution encoded by a PC can be regularized directly, which is again infeasible for most other density estimation models. We show that both methods consistently improve the generalization performance of a wide variety of PCs. Moreover, when paired with a simple PC structure, we achieved state-of-the-art results on 10 out of 20 standard discrete density estimation benchmarks."
                },
                {
                    "title": "Improving Lossless Compression Rates via Monte Carlo Bits-Back Coding",
                    "abstract": "Latent variable models have been successfully applied in lossless compression with the bits-back coding algorithm. However, bits-back suffers from an increase in the bitrate equal to the KL divergence between the approximate posterior and the true posterior. In this paper, we show how to remove this gap asymptotically by deriving bits-back coding algorithms from tighter variational bounds. The key idea is to exploit extended space representations of Monte Carlo estimators of the marginal likelihood. Naively applied, our schemes would require more initial bits than the standard bits-back coder, but we show how to drastically reduce this additional cost with couplings in the latent space. When parallel architectures can be exploited, our coders can achieve better rates than bits-back with little additional cost. We demonstrate improved lossless compression rates in a variety of settings, especially in out-of-distribution or sequential data compression."
                },
                {
                    "title": "Joints in Random Forests",
                    "abstract": "Decision Trees (DTs) and Random Forests (RFs) are powerful discriminative learners and tools of central importance to the everyday machine learning practitioner and data scientist. Due to their discriminative nature, however, they lack principled methods to process inputs with missing features or to detect outliers, which requires pairing them with imputation techniques or a separate generative model. In this paper, we demonstrate that DTs and RFs can naturally be interpreted as generative models, by drawing a connection to Probabilistic Circuits, a prominent class of tractable probabilistic models. This reinterpretation equips them with a full joint distribution over the feature space and leads to Generative Decision Trees (GeDTs) and Generative Forests (GeFs), a family of novel hybrid generative-discriminative models. This family of models retains the overall characteristics of DTs and RFs while additionally being able to handle missing features by means of marginalisation. Under certain assumptions, frequently made for Bayes consistency results, we show that consistency in GeDTs and GeFs extend to any pattern of missing input features, if missing at random. Empirically, we show that our models often outperform common routines to treat missing data, such as K-nearest neighbour imputation, and moreover, that our models can naturally detect outliers by monitoring the marginal probability of input features."
                },
                {
                    "title": "Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains",
                    "abstract": "We show that passing input points through a simple Fourier feature mapping enables a multilayer perceptron (MLP) to learn high-frequency functions in low-dimensional problem domains. These results shed light on recent advances in computer vision and graphics that achieve state-of-the-art results by using MLPs to represent complex 3D objects and scenes. Using tools from the neural tangent kernel (NTK) literature, we show that a standard MLP fails to learn high frequencies both in theory and in practice. To overcome this spectral bias, we use a Fourier feature mapping to transform the effective NTK into a stationary kernel with a tunable bandwidth. We suggest an approach for selecting problem-specific Fourier features that greatly improves the performance of MLPs for low-dimensional regression tasks relevant to the computer vision and graphics communities."
                },
                {
                    "title": "Einsum Networks: Fast and Scalable Learning of Tractable Probabilistic Circuits",
                    "abstract": "Probabilistic circuits (PCs) are a promising avenue for probabilistic modeling, as they permit a wide range of exact and efficient inference routines. Recent ``deep-learning-style'' implementations of PCs strive for a better scalability, but are still difficult to train on real-world data, due to their sparsely connected computational graphs. In this paper, we propose Einsum Networks (EiNets), a novel implementation design for PCs, improving prior art in several regards. At their core, EiNets combine a large number of arithmetic operations in a single monolithic einsum-operation, leading to speedups and memory savings of up to two orders of magnitude, in comparison to previous implementations. As an algorithmic contribution, we show that the implementation of Expectation-Maximization (EM) can be simplified for PCs, by leveraging automatic differentiation. Furthermore, we demonstrate that EiNets scale well to datasets which were previously out of reach, such as SVHN and CelebA, and that they can be used as faithful generative image models."
                },
                {
                    "title": "Sum-Product Network Decompilation",
                    "abstract": "There exists a dichotomy between classical probabilistic graphical models, such as Bayesian networks (BNs), and modern tractable models, such as sum-product networks (SPNs). The former have generally intractable inference, but allow a high level of interpretability, while the latter admits a wide range of tractable inference routines, but are typically harder to interpret. Due to this dichotomy, tools to convert between BNs and SPNs are desirable. While one direction -- compiling BNs into SPNs -- is well discussed in Darwiche's seminal work on arithmetic circuit compilation, the converse direction -- decompiling SPNs into BNs -- has received surprisingly little attention. In this paper, we fill this gap by proposing SPN2BN, an algorithm that decompiles an SPN into a BN. SPN2BN has several salient features when compared to the only other two works decompiling SPNs. Most significantly, the BNs returned by SPN2BN are minimal independence-maps. Secondly, SPN2BN is more parsimonious with respect to the introduction of latent variables. Thirdly, the output BN produced by SPN2BN can be precisely characterized with respect to the compiled BN. More specifically, a certain set of directed edges will be added to the input BN, giving what we will call the moral-closure. It immediately follows that there is a set of BNs related to the input BN that will also return the same moral closure. Lastly, it is established that our compilation-decompilation process is idempotent. We confirm our results with systematic experiments on a number of synthetic BNs."
                },
                {
                    "title": "Random Sum-Product Networks: A Simple and Effective Approach to Probabilistic Deep Learning",
                    "abstract": "Sum-product networks (SPNs) are expressive probabilistic models with a rich set of exact and ef\ufb01cient inference routines. However, in order to guarantee exact inference, they require speci\ufb01c structural constraints, which complicate learning SPNs from data. Thereby, most SPN structure learners proposed so far are tedious to tune, do not scale easily, and are not easily integrated with deep learning frameworks. In this paper, we follow a simple \u201cdeep learning\u201d approach, by generating unspecial-ized random structures, scalable to millions of parameters, and subsequently applying GPU-based optimization. Somewhat surprisingly, our models often perform on par with state-of-the-art SPN structure learners and deep neural networks on a diverse range of generative and discriminative scenarios. At the same time, our models yield well-calibrated uncertainties, and stand out among most deep generative and discriminative models in being robust to missing features and being able to detect anomalies."
                },
                {
                    "title": "Bit-Swap: Recursive Bits-Back Coding for Lossless Compression with Hierarchical Latent Variables",
                    "abstract": "The bits-back argument suggests that latent variable models can be turned into lossless compression schemes. Translating the bits-back argument into efficient and practical lossless compression schemes for general latent variable models, however, is still an open problem. Bits-Back with Asymmetric Numeral Systems (BB-ANS), recently proposed by Townsend et al. (2019), makes bits-back coding practically feasible for latent variable models with one latent layer, but it is inefficient for hierarchical latent variable models. In this paper we propose Bit-Swap, a new compression scheme that generalizes BB-ANS and achieves strictly better compression rates for hierarchical latent variable models with Markov chain structure. Through experiments we verify that Bit-Swap results in lossless compression rates that are empirically superior to existing techniques. Our implementation is available at this https URL."
                },
                {
                    "title": "Integer Discrete Flows and Lossless Compression",
                    "abstract": "Lossless compression methods shorten the expected representation size of data without loss of information, using a statistical model. Flow-based models are attractive in this setting because they admit exact likelihood optimization, which is equivalent to minimizing the expected number of bits per message. However, conventional flows assume continuous data, which may lead to reconstruction errors when quantized for compression. For that reason, we introduce a flow-based generative model for ordinal discrete data called Integer Discrete Flow (IDF): a bijective integer map that can learn rich transformations on high-dimensional data. As building blocks for IDFs, we introduce a flexible transformation layer called integer discrete coupling. Our experiments show that IDFs are competitive with other flow-based generative models. Furthermore, we demonstrate that IDF based compression achieves state-of-the-art lossless compression rates on CIFAR10, ImageNet32, and ImageNet64. To the best of our knowledge, this is the first lossless compression method that uses invertible neural networks."
                },
                {
                    "title": "Practical Lossless Compression with Latent Variables using Bits Back Coding",
                    "abstract": "Deep latent variable models have seen recent success in many data domains. Lossless compression is an application of these models which, despite having the potential to be highly useful, has yet to be implemented in a practical manner. We present `Bits Back with ANS' (BB-ANS), a scheme to perform lossless compression with latent variable models at a near optimal rate. We demonstrate this scheme by using it to compress the MNIST dataset with a variational auto-encoder model (VAE), achieving compression rates superior to standard methods with only a simple VAE. Given that the scheme is highly amenable to parallelization, we conclude that with a sufficiently high quality generative model this scheme could be used to achieve substantial improvements in compression rate with acceptable running time. We make our implementation available open source at this https URL ."
                },
                {
                    "title": "Mixed Sum-Product Networks: A Deep Architecture for Hybrid Domains",
                    "abstract": "\n \n While all kinds of mixed data---from personal data, over panel and scientific data, to public and commercial data---are collected and stored, building probabilistic graphical models for these hybrid domains becomes more difficult. Users spend significant amounts of time in identifying the parametric form of the random variables (Gaussian, Poisson, Logit, etc.) involved and learning the mixed models. To make this difficult task easier, we propose the first trainable probabilistic deep architecture for hybrid domains that features tractable queries. It is based on Sum-Product Networks (SPNs) with piecewise polynomial leaf distributions together with novel nonparametric decomposition and conditioning steps using the Hirschfeld-Gebelein-Renyi Maximum Correlation Coefficient. This relieves the user from deciding a-priori the parametric form of the random variables but is still expressive enough to effectively approximate any distribution and permits efficient learning and inference.Our experiments show that the architecture, called Mixed SPNs, can indeed capture complex distributions across a wide range of hybrid domains.\n \n"
                },
                {
                    "title": "Sum-Product Autoencoding: Encoding and Decoding Representations Using Sum-Product Networks",
                    "abstract": "\n \n Sum-Product Networks (SPNs) are a deep probabilistic architecture that up to now has been successfully employed for tractable inference. Here, we extend their scope towards unsupervised representation learning: we encode samples into continuous and categorical embeddings and show that they can also be decoded back into the original input space by leveraging MPE inference. We characterize when this Sum-Product Autoencoding (SPAE) leads to equivalent reconstructions and extend it towards dealing with missing embedding information. Our experimental results on several multi-label classification problems demonstrate that SPAE is competitive with state-of-the-art autoencoder architectures, even if the SPNs were never trained to reconstruct their inputs.\n \n"
                },
                {
                    "title": "Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms",
                    "abstract": "We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at this https URL"
                },
                {
                    "title": "EMNIST: Extending MNIST to handwritten letters",
                    "abstract": "The MNIST dataset has become a standard benchmark for learning, classification and computer vision systems. Contributing to its widespread adoption are the understandable and intuitive nature of the task, the relatively small size and storage requirements and the accessibility and ease-of-use of the database itself. The MNIST database was derived from a larger dataset known as the NIST Special Database 19 which contains digits, uppercase and lowercase handwritten letters. This paper introduces a variant of the full NIST dataset, which we have called Extended MNIST (EMNIST), which follows the same conversion paradigm used to create the MNIST dataset. The result is a dataset that constitutes a more challenging classification task involving letters and digits, and one that shares the same image structure and parameters as the original MNIST task, allowing for direct compatibility with all existing classifiers and systems. Benchmark results using an online ELM algorithm are presented along with a validation of the conversion process through the comparison of the classification results on NIST digits and the MNIST digits."
                },
                {
                    "title": "SGDR: Stochastic Gradient Descent with Warm Restarts",
                    "abstract": "Restart techniques are common in gradient-free optimization to deal with multimodal functions. Partial warm restarts are also gaining popularity in gradient-based optimization to improve the rate of convergence in accelerated gradient schemes to deal with ill-conditioned functions. In this paper, we propose a simple warm restart technique for stochastic gradient descent to improve its anytime performance when training deep neural networks. We empirically study its performance on the CIFAR-10 and CIFAR-100 datasets, where we demonstrate new state-of-the-art results at 3.14% and 16.21%, respectively. We also demonstrate its advantages on a dataset of EEG recordings and on a downsampled version of the ImageNet dataset. Our source code is available at this https URL"
                },
                {
                    "title": "On the Latent Variable Interpretation in Sum-Product Networks",
                    "abstract": "One of the central themes in Sum-Product networks (SPNs) is the interpretation of sum nodes as marginalized latent variables (LVs). This interpretation yields an increased syntactic or semantic structure, allows the application of the EM algorithm and to efficiently perform MPE inference. In literature, the LV interpretation was justified by explicitly introducing the indicator variables corresponding to the LVs\u2019 states. However, as pointed out in this paper, this approach is in conflict with the completeness condition in SPNs and does not fully specify the probabilistic model. We propose a remedy for this problem by modifying the original approach for introducing the LVs, which we call SPN augmentation. We discuss conditional independencies in augmented SPNs, formally establish the probabilistic interpretation of the sum-weights and give an interpretation of augmented SPNs as Bayesian networks. Based on these results, we find a sound derivation of the EM algorithm for SPNs. Furthermore, the Viterbi-style algorithm for MPE proposed in literature was never proven to be correct. We show that this is indeed a correct algorithm, when applied to selective SPNs, and in particular when applied to augmented SPNs. Our theoretical results are confirmed in experiments on synthetic data and 103 real-world datasets."
                },
                {
                    "title": "Continuous analogues of matrix factorizations",
                    "abstract": "Analogues of singular value decomposition (SVD), QR, LU and Cholesky factorizations are presented for problems in which the usual discrete matrix is replaced by a \u2018quasimatrix\u2019, continuous in one dimension, or a \u2018cmatrix\u2019, continuous in both dimensions. Two challenges arise: the generalization of the notions of triangular structure and row and column pivoting to continuous variables (required in all cases except the SVD, and far from obvious), and the convergence of the infinite series that define the cmatrix factorizations. Our generalizations of triangularity and pivoting are based on a new notion of a \u2018triangular quasimatrix\u2019. Concerning convergence of the series, we prove theorems asserting convergence provided the functions involved are sufficiently smooth."
                },
                {
                    "title": "Deep Learning Face Attributes in the Wild",
                    "abstract": "Predicting face attributes in the wild is challenging due to complex face variations. We propose a novel deep learning framework for attribute prediction in the wild. It cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently. LNet is pre-trained by massive general object categories for face localization, while ANet is pre-trained by massive face identities for attribute prediction. This framework not only outperforms the state-of-the-art with a large margin, but also reveals valuable facts on learning face representation. (1) It shows how the performances of face localization (LNet) and attribute prediction (ANet) can be improved by different pre-training strategies. (2) It reveals that although the filters of LNet are fine-tuned only with image-level attribute tags, their response maps over entire images have strong indication of face locations. This fact enables training LNet for face localization with only image-level annotations, but without face bounding boxes or landmarks, which are required by all attribute recognition works. (3) It also demonstrates that the high-level hidden neurons of ANet automatically discover semantic concepts after pre-training with massive face identities, and such concepts are significantly enriched after fine-tuning with attribute tags. Each attribute can be well explained with a sparse linear combination of these concepts."
                },
                {
                    "title": "Learning the Structure of Sum-Product Networks",
                    "abstract": "Sum-product networks (SPNs) are a new class of deep probabilistic models. SPNs can have unbounded treewidth but inference in them is always tractable. An SPN is either a univariate distribution, a product of SPNs over disjoint variables, or a weighted sum of SPNs over the same variables. We propose the first algorithm for learning the structure of SPNs that takes full advantage of their expressiveness. At each step, the algorithm attempts to divide the current variables into approximately independent subsets. If successful, it returns the product of recursive calls on the subsets; otherwise it returns the sum of recursive calls on subsets of similar instances from the current training set. A comprehensive empirical study shows that the learned SPNs are typically comparable to graphical models in likelihood but superior in inference speed and accuracy."
                },
                {
                    "title": "Learning the Architecture of Sum-Product Networks Using Clustering on Variables",
                    "abstract": "The sum-product network (SPN) is a recently-proposed deep model consisting of a network of sum and product nodes, and has been shown to be competitive with state-of-the-art deep models on certain difficult tasks such as image completion. Designing an SPN network architecture that is suitable for the task at hand is an open question. We propose an algorithm for learning the SPN architecture from data. The idea is to cluster variables (as opposed to data instances) in order to identify variable subsets that strongly interact with one another. Nodes in the SPN network are then allocated towards explaining these interactions. Experimental evidence shows that learning the SPN architecture significantly improves its performance compared to using a previously-proposed static architecture."
                },
                {
                    "title": "Sum-product networks: A new deep architecture",
                    "abstract": "The key limiting factor in graphical model inference and learning is the complexity of the partition function. We thus ask the question: what are the most general conditions under which the partition function is tractable? The answer leads to a new kind of deep architecture, which we call sum-product networks (SPNs) and will present in this abstract."
                },
                {
                    "title": "Learning Latent Tree Graphical Models",
                    "abstract": "We study the problem of learning a latent tree graphical model where samples are available only from a subset of variables. We propose two consistent and computationally efficient algorithms for learning minimal latent trees, that is, trees without any redundant hidden nodes. Unlike many existing methods, the observed nodes (or variables) are not constrained to be leaf nodes. Our algorithms can be applied to both discrete and Gaussian random variables and our learned models are such that all the observed and latent variables have the same domain (state space). Our first algorithm, recursive grouping, builds the latent tree recursively by identifying sibling groups using so-called information distances. One of the main contributions of this work is our second algorithm, which we refer to as CLGrouping. CLGrouping starts with a pre-processing procedure in which a tree over the observed variables is constructed. This global step groups the observed nodes that are likely to be close to each other in the true latent tree, thereby guiding subsequent recursive grouping (or equivalent procedures such as neighbor-joining) on much smaller subsets of variables. This results in more accurate and efficient learning of latent trees. We also present regularized versions of our algorithms that learn latent tree approximations of arbitrary distributions. We compare the proposed algorithms to other methods by performing extensive numerical experiments on various latent tree graphical models such as hidden Markov models and star graphs. In addition, we demonstrate the applicability of our methods on real-world data sets by modeling the dependency structure of monthly stock returns in the S&P index and of the words in the 20 newsgroups data set."
                },
                {
                    "title": "Probabilistic Graphical Models: Principles and Techniques - Adaptive Computation and Machine Learning",
                    "abstract": "Most tasks require a person or an automated system to reasonto reach conclusions based on available information. The framework of probabilistic graphical models, presented in this book, provides a general approach for this task. The approach is model-based, allowing interpretable models to be constructed and then manipulated by reasoning algorithms. These models can also be learned automatically from data, allowing the approach to be used in cases where manually constructing a model is difficult or even impossible. Because uncertainty is an inescapable aspect of most real-world applications, the book focuses on probabilistic models, which make the uncertainty explicit and provide models that are more faithful to reality. Probabilistic Graphical Models discusses a variety of models, spanning Bayesian networks, undirected Markov networks, discrete and continuous models, and extensions to deal with dynamical systems and relational data. For each class of models, the text describes the three fundamental cornerstones: representation, inference, and learning, presenting both basic concepts and advanced techniques. Finally, the book considers the use of the proposed framework for causal reasoning and decision making under uncertainty. The main text in each chapter provides the detailed technical development of the key ideas. Most chapters also include boxes with additional material: skill boxes, which describe techniques; case study boxes, which discuss empirical cases related to the approach described in the text, including applications in computer vision, robotics, natural language understanding, and computational biology; and concept boxes, which present significant concepts drawn from the material in the chapter. Instructors (and readers) can group chapters in various combinations, from core topics to more technically advanced material, to suit their particular needs. Adaptive Computation and Machine Learning series"
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Bayesian networks",
                    "abstract": "What are Bayesian networks and why are their applications growing across all fields?"
                },
                {
                    "title": "Fast Local Algorithms for Large Scale Nonnegative Matrix and Tensor Factorizations",
                    "abstract": "Nonnegative matrix factorization (NMF) and its extensions such as Nonnegative Tensor Factorization (NTF) have become prominent techniques for blind sources separation (BSS), analysis of image databases, data mining and other information retrieval and clustering applications. In this paper we propose a family of efficient algorithms for NMF/NTF, as well as sparse nonnegative coding and representation, that has many potential applications in computational neuroscience, multi-sensory processing, compressed sensing and multidimensional data analysis. We have developed a class of optimized local algorithms which are referred to as Hierarchical Alternating Least Squares (HALS) algorithms. For these purposes, we have performed sequential constrained minimization on a set of squared Euclidean distances. We then extend this approach to robust cost functions using the alpha and beta divergences and derive flexible update rules. Our algorithms are locally stable and work well for NMF-based blind source separation (BSS) not only for the over-determined case but also for an under-determined (over-complete) case (i.e., for a system which has less sensors than sources) if data are sufficiently sparse. The NMF learning rules are extended and generalized for N-th order nonnegative tensor factorization (NTF). Moreover, these algorithms can be tuned to different noise statistics by adjusting a single parameter. Extensive experimental results confirm the accuracy and computational performance of the developed algorithms, especially, with usage of multi-layer hierarchical NMF approach [3]."
                },
                {
                    "title": "New Compilation Languages Based on Structured Decomposability",
                    "abstract": "We introduce in this paper two new, complete propositional languages and study their properties in terms of (1) their support for polytime operations and (2) their ability to represent boolean functions compactly. The new languages are based on a structured version of decomposability--a property that underlies a number of tractable languages. The key characteristic of structured decomposability is its support for a polytime conjoin operation, which is known to be intractable for unstructured decomposability. We show that any CNF can be compiled into formulas in the new languages, whose size is only exponential in the treewidth of the CNF. Our study also reveals that one of the languages we identify is as powerful as OBDDs in terms of answering key inference queries, yet is more succinct than OBDDs."
                },
                {
                    "title": "Nonnegative Tucker Decomposition",
                    "abstract": "Nonnegative tensor factorization (NTF) is a recent multiway (multilinear) extension of nonnegative matrix factorization (NMF), where nonnegativity constraints are imposed on the CANDECOMP/PARAFAC model. In this paper we consider the Tucker model with nonnegativity constraints and develop a new tensor factorization method, referred to as nonnegative Tucker decomposition (NTD). The main contributions of this paper include: (1) multiplicative updating algorithms for NTD; (2) an initialization method for speeding up convergence; (3) a sparseness control method in tensor factorization. Through several computer vision examples, we show the useful behavior of the NTD, over existing NTF and NMF methods."
                },
                {
                    "title": "Multilinear operators for higher-order decompositions",
                    "abstract": "We propose two new multilinear operators for expressing the matrix compositions that are needed in the Tucker and PARAFAC (CANDECOMP) decompositions. The first operator, which we call the Tucker operator, is shorthand for performing an n-mode matrix multiplication for every mode of a given tensor and can be employed to concisely express the Tucker decomposition. The second operator, which we call the Kruskal operator, is shorthand for the sum of the outer-products of the columns of N matrices and allows a divorce from a matricized representation and a very concise expression of the PARAFAC decomposition. We explore the properties of the Tucker and Kruskal operators independently of the related decompositions. Additionally, we provide a review of the matrix and tensor operations that are frequently used in the context of tensor decompositions."
                },
                {
                    "title": "YCoCg-R: A Color Space with RGB Reversibility and Low Dynamic Range",
                    "abstract": "At the latest JVT meeting we presented the YCoCg color space, including its simple transformation equations relative to RGB and its improved coding gain relative to both RGB and YCbCr. We also discussed the reversibility of RGB to YCoCg conversion process in the case that two additional bits of precision are used for YCoCg relative to the precision used for source RGB data. In this proposal, we additionally present a variation of this color space, which we call YCoCg-R. The YCoCg color space has been applied to JPEG XR image compression and to texture compression through the YCoCg-DXT algorithm."
                },
                {
                    "title": "Learning Module Networks",
                    "abstract": "Methods for learning Bayesian networks can discover dependency structure between observed variables. Although these methods are useful in many applications, they run into computational and statistical problems in domains that involve a large number of variables. In this paper, we consider a solution that is applicable when many variables have similar behavior. We introduce a new class of models, module networks, that explicitly partition the variables into modules, so that the variables in each module share the same parents in the network and the same conditional probability distribution. We define the semantics of module networks, and describe an algorithm that learns the modules' composition and their dependency structure from data. Evaluation on real data in the domains of gene expression and the stock market shows that module networks generalize better than Bayesian networks, and that the learned module network structure reveals regularities that are obscured in learned Bayesian networks."
                },
                {
                    "title": "Methods of Numerical Integration",
                    "abstract": "Want to get experience? Want to get any ideas to create new things in your life? Read methods of numerical integration now! By reading this book as soon as possible, you can renew the situation to get the inspirations. Yeah, this way will lead you to always think more and more. In this case, this book will be always right for you. When you can observe more about the book, you will know why you need this."
                },
                {
                    "title": "Approximating discrete probability distributions with dependence trees",
                    "abstract": "A method is presented to approximate optimally an n -dimensional discrete probability distribution by a product of second-order distributions, or the distribution of the first-order tree dependence. The problem is to find an optimum set of n - 1 first order dependence relationship among the n variables. It is shown that the procedure derived in this paper yields an approximation of a minimum difference in information. It is further shown that when this procedure is applied to empirical observations from an unknown distribution of tree dependence, the procedure is the maximum-likelihood estimate of the distribution."
                },
                {
                    "title": "Random probabilistic circuits",
                    "abstract": "Density estimation could be viewed as a core component in machine learning, since a good estimator could be used to solve many tasks such as clas-si\ufb01cation, regression, and imputing missing values. The main challenge of density estimation is balancing the model expressiveness and its learning and inference complexity. Probabilistic circuits (PCs) model a probability distribution as a computational graph. By imposing speci\ufb01c structural properties on such models many inference tasks become tractable. However, learning PCs usually relies on greedy and time consuming algorithms. In this paper we propose a new uni\ufb01ed approach to ef\ufb01ciently learn PCs having several structural properties. We introduce extremely randomized PCs (XPCs), PCs with a random structure. We show their advantage on standard density estimation benchmarks when compared to other density estimators."
                },
                {
                    "title": "A Compositional Atlas of Tractable Circuit Operations for Probabilistic Inference",
                    "abstract": "Circuit representations are becoming the lingua franca to express and reason about tractable generative and discriminative models. In this paper, we show how complex inference scenarios for these models that commonly arise in machine learning\u2014from computing the expectations of decision tree ensembles to information-theoretic divergences of sum-product networks\u2014can be represented in terms of tractable modular operations over circuits. Speci\ufb01cally, we characterize the tractability of simple transformations\u2014sums, products, quotients, powers, logarithms, and exponentials\u2014in terms of suf\ufb01cient structural constraints of the circuits they operate on, and present novel hardness results for the cases in which these properties are not satis\ufb01ed. Building on these operations, we derive a uni\ufb01ed framework for reasoning about tractable models that generalizes several results in the literature and opens up novel tractable inference scenarios."
                },
                {
                    "title": "Probabilistic Circuits: A Unifying Framework for Tractable Probabilistic Models\u2217",
                    "abstract": "2 Probabilistic Inference: Models, Queries, and Tractability 5 2.1 Probabilistic Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2 Probabilistic Queries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.3 Tractable Probabilistic Inference . . . . . . . . . . . . . . . . . . . . . . . . 8 2.4 Properties of Tractable Probabilistic Models . . . . . . . . . . . . . . . . . . 9"
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                },
                {
                    "title": "GENERATIVE ADVERSARIAL NETS",
                    "abstract": "Estimating individualized treatment effects (ITE) is a challenging task due to the need for an individual\u2019s potential outcomes to be learned from biased data and without having access to the counterfactuals. We propose a novel method for inferring ITE based on the Generative Adversarial Nets (GANs) framework. Our method, termed Generative Adversarial Nets for inference of Individualized Treatment Effects (GANITE), is motivated by the possibility that we can capture the uncertainty in the counterfactual distributions by attempting to learn them using a GAN. We generate proxies of the counterfactual outcomes using a counterfactual generator, G, and then pass these proxies to an ITE generator, I, in order to train it. By modeling both of these using the GAN framework, we are able to infer based on the factual data, while still accounting for the unseen counterfactuals. We test our method on three real-world datasets (with both binary and multiple treatments) and show that GANITE outperforms state-of-the-art methods."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively extend probabilistic integral circuits (PICs) to represent directed acyclic graph (DAG)-shaped hierarchies of continuous latent variables for improved tractability and learning in generative models?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the inherent trade-off between tractability, ease of learning, and expressiveness in generative models. By extending PICs to DAG structures, we can enhance the capabilities of generative models, enabling more complex probabilistic reasoning tasks that are currently intractable. This advancement could lead to significant improvements in various applications, such as image generation, natural language processing, and other domains where understanding complex distributions is essential. Furthermore, it could inspire future research directions in deep generative modeling and probabilistic reasoning.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of learning and scaling hierarchical structures of continuous latent variables. Naive approaches may fail due to the intractability of computing marginals or conditionals in existing models, which complicates probabilistic reasoning. Additionally, the need for systematic approximations and the integration of continuous latent variables into a coherent framework presents significant technical and theoretical obstacles. The intricacies of designing effective learning algorithms that can handle the added complexity of DAG structures further complicate the task.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on tree-shaped PICs, leaving a gap in the exploration of richer hierarchical structures like DAGs. Limitations in existing methodologies, such as the reliance on specific circuit architectures and the challenges of scaling, have hindered progress. Additionally, prior work often required extra supervision and complex training processes, which are not conducive to end-to-end learning. Our approach differs by providing a systematic pipeline for constructing DAG-shaped PICs and employing a more efficient training method that is self-contained and does not rely on pre-trained models or extensive heuristics.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves several key components: (1) a systematic pipeline for building DAG-shaped PICs from arbitrary variable decompositions; (2) a hierarchical quadrature process for learning and approximating PICs, encoded in tensorized quadrature probabilistic circuits (QPCs); (3) functional sharing techniques"
            }
        },
        "author_data": {
            "0032a8df-c8b1-4e0c-a40d-c5e981298ab1": {
                "pk": "0032a8df-c8b1-4e0c-a40d-c5e981298ab1",
                "name": "Gennaro Gala",
                "collaborators": [
                    "Erik Quaeghebeur",
                    "Robert Peharz",
                    "Lorenzo Loconte",
                    "Cassio de Campos",
                    "Antonio Vergari",
                    "Daniele Grattarola",
                    "Antonio Mari",
                    "Gennaro Vessio",
                    "Yang Yang"
                ],
                "domain": [
                    "Deep Learning",
                    "Probabilistic Models",
                    "Graph Neural Networks",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "DeeProb-kit: a Python Library for Deep Probabilistic Modelling",
                        "abstract": "DeeProb-kit is a unified library written in Python consisting of a collection of deep probabilistic models (DPMs) that are tractable and exact representations for the modelled probability distributions. The availability of a representative selection of DPMs in a single library makes it possible to combine them in a straightforward manner, a common practice in deep learning research nowadays. In addition, it includes efficiently implemented learning techniques, inference routines, statistical algorithms, and provides high-quality fully-documented APIs. The development of DeeProb-kit will help the community to accelerate research on DPMs as well as to standardise their evaluation and better understand how they are related based on their expressivity."
                    },
                    {
                        "title": "E(n)-equivariant Graph Neural Cellular Automata",
                        "abstract": "Cellular automata (CAs) are computational models exhibiting rich dynamics emerging from the local interaction of cells arranged in a regular lattice. Graph CAs (GCAs) generalise standard CAs by allowing for arbitrary graphs rather than regular lattices, similar to how Graph Neural Networks (GNNs) generalise Convolutional NNs. Recently, Graph Neural CAs (GNCAs) have been proposed as models built on top of standard GNNs that can be trained to approximate the transition rule of any arbitrary GCA. Existing GNCAs are anisotropic in the sense that their transition rules are not equivariant to translation, rotation, and reflection of the nodes' spatial locations. However, it is desirable for instances related by such transformations to be treated identically by the model. By replacing standard graph convolutions with E(n)-equivariant ones, we avoid anisotropy by design and propose a class of isotropic automata that we call E(n)-GNCAs. These models are lightweight, but can nevertheless handle large graphs, capture complex dynamics and exhibit emergent self-organising behaviours. We showcase the broad and successful applicability of E(n)-GNCAs on three different tasks: (i) pattern formation, (ii) graph auto-encoding, and (iii) simulation of E(n)-equivariant dynamical systems."
                    },
                    {
                        "title": "What is the Relationship between Tensor Factorizations and Circuits (and How Can We Exploit it)?",
                        "abstract": "This paper establishes a rigorous connection between circuit representations and tensor factorizations, two seemingly distinct yet fundamentally related areas. By connecting these fields, we highlight a series of opportunities that can benefit both communities. Our work generalizes popular tensor factorizations within the circuit language, and unifies various circuit learning algorithms under a single, generalized hierarchical factorization framework. Specifically, we introduce a modular \"Lego block\" approach to build tensorized circuit architectures. This, in turn, allows us to systematically construct and explore various circuit and tensor factorization models while maintaining tractability. This connection not only clarifies similarities and differences in existing models, but also enables the development of a comprehensive pipeline for building and optimizing new circuit/tensor factorization architectures. We show the effectiveness of our framework through extensive empirical evaluations, and highlight new research opportunities for tensor factorizations in probabilistic modeling."
                    },
                    {
                        "title": "Bayesian Structure Scores for Probabilistic Circuits",
                        "abstract": "Probabilistic circuits (PCs) are a prominent representation of probability distributions with tractable inference. While parameter learning in PCs is rigorously studied, structure learning is often more based on heuristics than on principled objectives. In this paper, we develop Bayesian structure scores for deterministic PCs, i.e., the structure likelihood with parameters marginalized out, which are well known as rigorous objectives for structure learning in probabilistic graphical models. When used within a greedy cutset algorithm, our scores effectively protect against overfitting and yield a fast and almost hyper-parameter-free structure learner, distinguishing it from previous approaches. In experiments, we achieve good trade-offs between training time and model fit in terms of log-likelihood. Moreover, the principled nature of Bayesian scores unlocks PCs for accommodating frameworks such as structural expectation-maximization."
                    },
                    {
                        "title": "Probabilistic Integral Circuits",
                        "abstract": "Continuous latent variables (LVs) are a key ingredient of many generative models, as they allow modelling expressive mixtures with an uncountable number of components. In contrast, probabilistic circuits (PCs) are hierarchical discrete mixtures represented as computational graphs composed of input, sum and product units. Unlike continuous LV models, PCs provide tractable inference but are limited to discrete LVs with categorical (i.e. unordered) states. We bridge these model classes by introducing probabilistic integral circuits (PICs), a new language of computational graphs that extends PCs with integral units representing continuous LVs. In the first place, PICs are symbolic computational graphs and are fully tractable in simple cases where analytical integration is possible. In practice, we parameterise PICs with light-weight neural nets delivering an intractable hierarchical continuous mixture that can be approximated arbitrarily well with large PCs using numerical quadrature. On several distribution estimation benchmarks, we show that such PIC-approximating PCs systematically outperform PCs commonly learned via expectation-maximization or SGD."
                    }
                ]
            },
            "8afbde58-ed4f-4b97-8d4f-0cf25f1d9c69": {
                "pk": "8afbde58-ed4f-4b97-8d4f-0cf25f1d9c69",
                "name": "Cassio de Campos",
                "collaborators": [
                    "Soroush Ghandi",
                    "Benjamin Quost",
                    "Alvaro H. C. Correia",
                    "Robert Peharz",
                    "Alessio Benavoli",
                    "Vu-Linh Nguyen",
                    "Yang Yang",
                    "James Cussens",
                    "Gennaro Gala",
                    "Antonio Vergari"
                ],
                "domain": [
                    "Probabilistic Models",
                    "Machine Learning",
                    "Bayesian Inference",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Bayesian Kernelised Test of (In)dependence with Mixed-type Variables",
                        "abstract": "A fundamental task in AI is to assess (in)dependence between mixed-type variables (text, image, sound). We propose a Bayesian kernelised correlation test of (in)dependence using a Dirichlet process model. The new measure of (in)dependence allows us to answer some fundamental questions: Based on data, are (mixed-type) variables independent? How likely is dependence/independence to hold? How high is the probability that two mixed-type variables are more than just weakly dependent? We theoretically show the properties of the approach, as well as algorithms for fast computation with it. We empirically demonstrate the effectiveness of the proposed method by analysing its performance and by comparing it with other frequentist and Bayesian approaches on a range of datasets and tasks with mixed-type variables."
                    },
                    {
                        "title": "Probabilistic Circuits with Constraints via Convex Optimization",
                        "abstract": "This work addresses integrating probabilistic propositional logic constraints into the distribution encoded by a probabilistic circuit (PC). PCs are a class of tractable models that allow efficient computations (such as conditional and marginal probabilities) while achieving state-of-the-art performance in some domains. The proposed approach takes both a PC and constraints as inputs, and outputs a new PC that satisfies the constraints. This is done efficiently via convex optimization without the need to retrain the entire model. Empirical evaluations indicate that the combination of constraints and PCs can have multiple use cases, including the improvement of model performance under scarce or incomplete data, as well as the enforcement of machine learning fairness measures into the model without compromising model fitness. We believe that these ideas will open possibilities for multiple other applications involving the combination of logics and deep probabilistic models."
                    },
                    {
                        "title": "Soft Learning Probabilistic Circuits",
                        "abstract": "Probabilistic Circuits (PCs) are prominent tractable probabilistic models, allowing for a range of exact inferences. This paper focuses on the main algorithm for training PCs, LearnSPN, a gold standard due to its efficiency, performance, and ease of use, in particular for tabular data. We show that LearnSPN is a greedy likelihood maximizer under mild assumptions. While inferences in PCs may use the entire circuit structure for processing queries, LearnSPN applies a hard method for learning them, propagating at each sum node a data point through one and only one of the children/edges as in a hard clustering process. We propose a new learning procedure named SoftLearn, that induces a PC using a soft clustering process. We investigate the effect of this learning-inference compatibility in PCs. Our experiments show that SoftLearn outperforms LearnSPN in many situations, yielding better likelihoods and arguably better samples. We also analyze comparable tractable models to highlight the differences between soft/hard learning and model querying."
                    },
                    {
                        "title": "Probabilistic Multi-Dimensional Classification",
                        "abstract": "Multi-dimensional classification (MDC) can be employed in a range of applications where one needs to predict multiple class variables for each given instance. Many existing MDC methods suffer from at least one of inaccuracy, scalability, limited use to certain types of data, hardness of interpretation or lack of probabilistic (uncertainty) estimations. This paper is an attempt to address all these disadvantages simultaneously. We propose a formal framework for probabilistic MDC in which learning an optimal multi-dimensional classifier can be decomposed, without loss of generality, into learning a set of (smaller) single-variable multi-class probabilistic classifiers and a directed acyclic graph. Current and future developments of both probabilistic classification and graphical model learning can directly enhance our framework, which is flexible and provably optimal. A collection of experiments is conducted to highlight the usefulness of this MDC framework."
                    },
                    {
                        "title": "On Pruning for Score-Based Bayesian Network Structure Learning",
                        "abstract": "Many algorithms for score-based Bayesian network structure learning (BNSL), in particular exact ones, take as input a collection of potentially optimal parent sets for each variable in the data. Constructing such collections naively is computationally intensive since the number of parent sets grows exponentially with the number of variables. Thus, pruning techniques are not only desirable but essential. While good pruning rules exist for the Bayesian Information Criterion (BIC), current results for the Bayesian Dirichlet equivalent uniform (BDeu) score reduce the search space very modestly, hampering the use of the (often preferred) BDeu. We derive new non-trivial theoretical upper bounds for the BDeu score that considerably improve on the state-of-the-art. Since the new bounds are mathematically proven to be tighter than previous ones and at little extra computational cost, they are a promising addition to BNSL methods."
                    },
                    {
                        "title": "Towards Robust Classification with Deep Generative Forests",
                        "abstract": "Decision Trees and Random Forests are among the most widely used machine learning models, and often achieve state-of-the-art performance in tabular, domain-agnostic datasets. Nonetheless, being primarily discriminative models they lack principled methods to manipulate the uncertainty of predictions. In this paper, we exploit Generative Forests (GeFs), a recent class of deep probabilistic models that addresses these issues by extending Random Forests to generative models representing the full joint distribution over the feature space. We demonstrate that GeFs are uncertainty-aware classifiers, capable of measuring the robustness of each prediction as well as detecting out-of-distribution samples."
                    },
                    {
                        "title": "Probabilistic Integral Circuits",
                        "abstract": "Continuous latent variables (LVs) are a key ingredient of many generative models, as they allow modelling expressive mixtures with an uncountable number of components. In contrast, probabilistic circuits (PCs) are hierarchical discrete mixtures represented as computational graphs composed of input, sum and product units. Unlike continuous LV models, PCs provide tractable inference but are limited to discrete LVs with categorical (i.e. unordered) states. We bridge these model classes by introducing probabilistic integral circuits (PICs), a new language of computational graphs that extends PCs with integral units representing continuous LVs. In the first place, PICs are symbolic computational graphs and are fully tractable in simple cases where analytical integration is possible. In practice, we parameterise PICs with light-weight neural nets delivering an intractable hierarchical continuous mixture that can be approximated arbitrarily well with large PCs using numerical quadrature. On several distribution estimation benchmarks, we show that such PIC-approximating PCs systematically outperform PCs commonly learned via expectation-maximization or SGD."
                    },
                    {
                        "title": "Nerva: a Truly Sparse Implementation of Neural Networks",
                        "abstract": "We introduce Nerva, a fast neural network library under development in C++. It supports sparsity by using the sparse matrix operations of Intel's Math Kernel Library (MKL), which eliminates the need for binary masks. We show that Nerva significantly decreases training time and memory usage while reaching equivalent accuracy to PyTorch. We run static sparse experiments with an MLP on CIFAR-10. On high sparsity levels like $99\\%$, the runtime is reduced by a factor of $4\\times$ compared to a PyTorch model using masks. Similar to other popular frameworks such as PyTorch and Keras, Nerva offers a Python interface for users to work with."
                    }
                ]
            },
            "356042fa-f1d6-4574-9e28-ef3ddcf63850": {
                "pk": "356042fa-f1d6-4574-9e28-ef3ddcf63850",
                "name": "Antonio Vergari",
                "collaborators": [
                    "Stefano Teso",
                    "Nicola Di Mauro",
                    "Guy Van den Broeck",
                    "Robert Peharz",
                    "Lorenzo Loconte",
                    "Floriana Esposito",
                    "Alejandro Molina",
                    "Kristian Kersting",
                    "Alessandro Suglia",
                    "Diego Calanzone"
                ],
                "domain": [
                    "Probabilistic Models",
                    "Active Learning",
                    "Natural Language Processing",
                    "Graphical Models"
                ],
                "publications": [
                    {
                        "title": "Efficient and Reliable Probabilistic Interactive Learning with Structured Outputs",
                        "abstract": "In this position paper, we study interactive learning for structured output spaces, with a focus on active learning, in which labels are unknown and must be acquired, and on skeptical learning, in which the labels are noisy and may need relabeling. These scenarios require expressive models that guarantee reliable and efficient computation of probabilistic quantities to measure uncertainty. We identify conditions under which a class of probabilistic models -- which we denote CRISPs -- meet all of these conditions, thus delivering tractable computation of the above quantities while preserving expressiveness. Building on prior work on tractable probabilistic circuits, we illustrate how CRISPs enable robust and efficient active and skeptical learning in large structured output spaces."
                    },
                    {
                        "title": "A Compositional Atlas of Tractable Circuit Operations: From Simple Transformations to Complex Information-Theoretic Queries",
                        "abstract": "Circuit representations are becoming the lingua franca to express and reason about tractable generative and discriminative models. In this paper, we show how complex inference scenarios for these models that commonly arise in machine learning -- from computing the expectations of decision tree ensembles to information-theoretic divergences of deep mixture models -- can be represented in terms of tractable modular operations over circuits. Specifically, we characterize the tractability of a vocabulary of simple transformations -- sums, products, quotients, powers, logarithms, and exponentials -- in terms of sufficient structural constraints of the circuits they operate on, and present novel hardness results for the cases in which these properties are not satisfied. Building on these operations, we derive a unified framework for reasoning about tractable models that generalizes several results in the literature and opens up novel tractable inference scenarios."
                    },
                    {
                        "title": "Automatic Bayesian Density Analysis",
                        "abstract": "Making sense of a dataset in an automatic and unsupervised fashion is a challenging problem in statistics and AI. Classical approaches for {exploratory data analysis} are usually not flexible enough to deal with the uncertainty inherent to real-world data: they are often restricted to fixed latent interaction models and homogeneous likelihoods; they are sensitive to missing, corrupt and anomalous data; moreover, their expressiveness generally comes at the price of intractable inference. As a result, supervision from statisticians is usually needed to find the right model for the data. However, since domain experts are not necessarily also experts in statistics, we propose Automatic Bayesian Density Analysis (ABDA) to make exploratory data analysis accessible at large. Specifically, ABDA allows for automatic and efficient missing value estimation, statistical data type and likelihood discovery, anomaly detection and dependency structure mining, on top of providing accurate density estimation. Extensive empirical evidence shows that ABDA is a suitable tool for automatic exploratory analysis of mixed continuous and discrete tabular data."
                    },
                    {
                        "title": "ChemAlgebra: Algebraic Reasoning on Chemical Reactions",
                        "abstract": "While showing impressive performance on various kinds of learning tasks, it is yet unclear whether deep learning models have the ability to robustly tackle reasoning tasks. than by learning the underlying reasoning process that is actually required to solve the tasks. Measuring the robustness of reasoning in machine learning models is challenging as one needs to provide a task that cannot be easily shortcut by exploiting spurious statistical correlations in the data, while operating on complex objects and constraints. reasoning task. To address this issue, we propose ChemAlgebra, a benchmark for measuring the reasoning capabilities of deep learning models through the prediction of stoichiometrically-balanced chemical reactions. ChemAlgebra requires manipulating sets of complex discrete objects -- molecules represented as formulas or graphs -- under algebraic constraints such as the mass preservation principle. We believe that ChemAlgebra can serve as a useful test bed for the next generation of machine reasoning models and as a promoter of their development."
                    },
                    {
                        "title": "Not All Neuro-Symbolic Concepts Are Created Equal: Analysis and Mitigation of Reasoning Shortcuts",
                        "abstract": "Neuro-Symbolic (NeSy) predictive models hold the promise of improved compliance with given constraints, systematic generalization, and interpretability, as they allow to infer labels that are consistent with some prior knowledge by reasoning over high-level concepts extracted from sub-symbolic inputs. It was recently shown that NeSy predictors are affected by reasoning shortcuts: they can attain high accuracy but by leveraging concepts with unintended semantics, thus coming short of their promised advantages. Yet, a systematic characterization of reasoning shortcuts and of potential mitigation strategies is missing. This work fills this gap by characterizing them as unintended optima of the learning objective and identifying four key conditions behind their occurrence. Based on this, we derive several natural mitigation strategies, and analyze their efficacy both theoretically and empirically. Our analysis shows reasoning shortcuts are difficult to deal with, casting doubts on the trustworthiness and interpretability of existing NeSy solutions."
                    },
                    {
                        "title": "Taming the Sigmoid Bottleneck: Provably Argmaxable Sparse Multi-Label Classification",
                        "abstract": "Sigmoid output layers are widely used in multi-label classification (MLC) tasks, in which multiple labels can be assigned to any input. In many practical MLC tasks, the number of possible labels is in the thousands, often exceeding the number of input features and resulting in a low-rank output layer. In multi-class classification, it is known that such a low-rank output layer is a bottleneck that can result in unargmaxable classes: classes which cannot be predicted for any input. In this paper, we show that for MLC tasks, the analogous sigmoid bottleneck results in exponentially many unargmaxable label combinations. We explain how to detect these unargmaxable outputs and demonstrate their presence in three widely used MLC datasets. We then show that they can be prevented in practice by introducing a Discrete Fourier Transform (DFT) output layer, which guarantees that all sparse label combinations with up to $k$ active labels are argmaxable. Our DFT layer trains faster and is more parameter efficient, matching the F1@k score of a sigmoid layer while using up to 50% fewer trainable parameters. Our code is publicly available at https://github.com/andreasgrv/sigmoid-bottleneck."
                    },
                    {
                        "title": "PIXAR: Auto-Regressive Language Modeling in Pixel Space",
                        "abstract": "Recent work showed the possibility of building open-vocabulary large language models (LLMs) that directly operate on pixel representations. These models are implemented as autoencoders that reconstruct masked patches of rendered text. However, these pixel-based LLMs are limited to discriminative tasks (e.g., classification) and, similar to BERT, cannot be used to generate text. Therefore, they cannot be used for generative tasks such as free-form question answering. In this work, we introduce PIXAR, the first pixel-based autoregressive LLM that performs text generation. Consisting of only a decoder, PIXAR can perform free-form generative tasks while keeping the number of parameters on par with previous encoder-decoder models. Furthermore, we highlight the challenges of generating text as non-noisy images and show this is due to using a maximum likelihood objective. To overcome this problem, we propose an adversarial pretraining stage that improves the readability and accuracy of PIXAR by 8.1 on LAMBADA and 8.5 on bAbI -- making it comparable to GPT-2 on text generation tasks. This paves the way to build open-vocabulary LLMs that operate on perceptual input only and calls into question the necessity of the usual symbolic input representation, i.e., text as (sub)tokens."
                    },
                    {
                        "title": "Towards Logically Consistent Language Models via Probabilistic Reasoning",
                        "abstract": "Large language models (LLMs) are a promising venue for natural language understanding and generation tasks. However, current LLMs are far from reliable: they are prone to generate non-factual information and, more crucially, to contradict themselves when prompted to reason about beliefs of the world. These problems are currently addressed with large scale fine-tuning or by delegating consistent reasoning to external tools. In this work, we strive for a middle ground and introduce a training objective based on principled probabilistic reasoning that teaches a LLM to be consistent with external knowledge in the form of a set of facts and rules. Fine-tuning with our loss on a limited set of facts enables our LLMs to be more logically consistent than previous baselines and allows them to extrapolate to unseen but semantically similar factual knowledge more systematically."
                    },
                    {
                        "title": "What can Large Language Models Capture about Code Functional Equivalence?",
                        "abstract": "Code-LLMs, LLMs pre-trained on large code corpora, have shown great progress in learning rich representations of the structure and syntax of code, successfully using it to generate or classify code fragments. At the same time, understanding if they are able to do so because they capture code semantics, and how well, is still an open question. In this paper, we tackle this problem by introducing SeqCoBench, a benchmark for systematically assessing how Code-LLMs can capture code functional equivalence. SeqCoBench contains over 20 code transformations that either preserve or alter the semantics of Python programs. We conduct extensive evaluations in different settings, including zero-shot and parameter-efficient finetuning methods on state-of-the-art (Code-)LLMs to see if they can discern semantically equivalent or different pairs of programs in SeqCoBench. We find that the performance gap between these LLMs and classical match-based retrieval scores is minimal, with both approaches showing a concerning lack of depth in understanding code semantics."
                    },
                    {
                        "title": "Sum of Squares Circuits",
                        "abstract": "Designing expressive generative models that support exact and efficient inference is a core question in probabilistic ML. Probabilistic circuits (PCs) offer a framework where this tractability-vs-expressiveness trade-off can be analyzed theoretically. Recently, squared PCs encoding subtractive mixtures via negative parameters have emerged as tractable models that can be exponentially more expressive than monotonic PCs, i.e., PCs with positive parameters only. In this paper, we provide a more precise theoretical characterization of the expressiveness relationships among these models. First, we prove that squared PCs can be less expressive than monotonic ones. Second, we formalize a novel class of PCs -- sum of squares PCs -- that can be exponentially more expressive than both squared and monotonic PCs. Around sum of squares PCs, we build an expressiveness hierarchy that allows us to precisely unify and separate different tractable model classes such as Born Machines and PSD models, and other recently introduced tractable probabilistic models by using complex parameters. Finally, we empirically show the effectiveness of sum of squares circuits in performing distribution estimation."
                    },
                    {
                        "title": "Logically Consistent Language Models via Neuro-Symbolic Integration",
                        "abstract": "Large language models (LLMs) are a promising venue for natural language understanding and generation. However, current LLMs are far from reliable: they are prone to generating non-factual information and, more crucially, to contradicting themselves when prompted to reason about relations between entities of the world. These problems are currently addressed with large scale fine-tuning or by delegating reasoning to external tools. In this work, we strive for a middle ground and introduce a loss based on neuro-symbolic reasoning that teaches an LLM to be logically consistent with an external set of facts and rules and improves self-consistency even when the LLM is fine-tuned on a limited set of facts. Our approach also allows to easily combine multiple logical constraints at once in a principled way, delivering LLMs that are more consistent w.r.t. all constraints and improve over several baselines w.r.t. a given constraint. Moreover, our method allows LLMs to extrapolate to unseen but semantically similar factual knowledge, represented in unseen datasets, more systematically."
                    },
                    {
                        "title": "What is the Relationship between Tensor Factorizations and Circuits (and How Can We Exploit it)?",
                        "abstract": "This paper establishes a rigorous connection between circuit representations and tensor factorizations, two seemingly distinct yet fundamentally related areas. By connecting these fields, we highlight a series of opportunities that can benefit both communities. Our work generalizes popular tensor factorizations within the circuit language, and unifies various circuit learning algorithms under a single, generalized hierarchical factorization framework. Specifically, we introduce a modular \"Lego block\" approach to build tensorized circuit architectures. This, in turn, allows us to systematically construct and explore various circuit and tensor factorization models while maintaining tractability. This connection not only clarifies similarities and differences in existing models, but also enables the development of a comprehensive pipeline for building and optimizing new circuit/tensor factorization architectures. We show the effectiveness of our framework through extensive empirical evaluations, and highlight new research opportunities for tensor factorizations in probabilistic modeling."
                    },
                    {
                        "title": "Towards Representation Learning with Tractable Probabilistic Models",
                        "abstract": "Probabilistic models learned as density estimators can be exploited in representation learning beside being toolboxes used to answer inference queries only. However, how to extract useful representations highly depends on the particular model involved. We argue that tractable inference, i.e. inference that can be computed in polynomial time, can enable general schemes to extract features from black box models. We plan to investigate how Tractable Probabilistic Models (TPMs) can be exploited to generate embeddings by random query evaluations. We devise two experimental designs to assess and compare different TPMs as feature extractors in an unsupervised representation learning framework. We show some experimental results on standard image datasets by applying such a method to Sum-Product Networks and Mixture of Trees as tractable models generating embeddings."
                    },
                    {
                        "title": "Strudel: Learning Structured-Decomposable Probabilistic Circuits",
                        "abstract": "Probabilistic circuits (PCs) represent a probability distribution as a computational graph. Enforcing structural properties on these graphs guarantees that several inference scenarios become tractable. Among these properties, structured decomposability is a particularly appealing one: it enables the efficient and exact computations of the probability of complex logical formulas, and can be used to reason about the expected output of certain predictive models under missing data. This paper proposes Strudel, a simple, fast and accurate learning algorithm for structured-decomposable PCs. Compared to prior work for learning structured-decomposable PCs, Strudel delivers more accurate single PC models in fewer iterations, and dramatically scales learning when building ensembles of PCs. It achieves this scalability by exploiting another structural property of PCs, called determinism, and by sharing the same computational graph across mixture components. We show these advantages on standard density estimation benchmarks and challenging inference scenarios."
                    },
                    {
                        "title": "How to Turn Your Knowledge Graph Embeddings into Generative Models",
                        "abstract": "Some of the most successful knowledge graph embedding (KGE) models for link prediction -- CP, RESCAL, TuckER, ComplEx -- can be interpreted as energy-based models. Under this perspective they are not amenable for exact maximum-likelihood estimation (MLE), sampling and struggle to integrate logical constraints. This work re-interprets the score functions of these KGEs as circuits -- constrained computational graphs allowing efficient marginalisation. Then, we design two recipes to obtain efficient generative circuit models by either restricting their activations to be non-negative or squaring their outputs. Our interpretation comes with little or no loss of performance for link prediction, while the circuits framework unlocks exact learning by MLE, efficient sampling of new triples, and guarantee that logical constraints are satisfied by design. Furthermore, our models scale more gracefully than the original KGEs on graphs with millions of entities."
                    },
                    {
                        "title": "Visualizing and Understanding Sum-Product Networks",
                        "abstract": "Sum-Product Networks (SPNs) are recently introduced deep tractable probabilistic models by which several kinds of inference queries can be answered exactly and in a tractable time. Up to now, they have been largely used as black box density estimators, assessed only by comparing their likelihood scores only. In this paper we explore and exploit the inner representations learned by SPNs. We do this with a threefold aim: first we want to get a better understanding of the inner workings of SPNs; secondly, we seek additional ways to evaluate one SPN model and compare it against other probabilistic models, providing diagnostic tools to practitioners; lastly, we want to empirically evaluate how good and meaningful the extracted representations are, as in a classic Representation Learning framework. In order to do so we revise their interpretation as deep neural networks and we propose to exploit several visualization techniques on their node activations and network outputs under different types of inference queries. To investigate these models as feature extractors, we plug some SPNs, learned in a greedy unsupervised fashion on image datasets, in supervised classification learning tasks. We extract several embedding types from node activations by filtering nodes by their type, by their associated feature abstraction level and by their scope. In a thorough empirical comparison we prove them to be competitive against those generated from popular feature extractors as Restricted Boltzmann Machines. Finally, we investigate embeddings generated from random probabilistic marginal queries as means to compare other tractable probabilistic models on a common ground, extending our experiments to Mixtures of Trees."
                    },
                    {
                        "title": "Sum-Product Networks for Hybrid Domains",
                        "abstract": "While all kinds of mixed data -from personal data, over panel and scientific data, to public and commercial data- are collected and stored, building probabilistic graphical models for these hybrid domains becomes more difficult. Users spend significant amounts of time in identifying the parametric form of the random variables (Gaussian, Poisson, Logit, etc.) involved and learning the mixed models. To make this difficult task easier, we propose the first trainable probabilistic deep architecture for hybrid domains that features tractable queries. It is based on Sum-Product Networks (SPNs) with piecewise polynomial leave distributions together with novel nonparametric decomposition and conditioning steps using the Hirschfeld-Gebelein-R\\'enyi Maximum Correlation Coefficient. This relieves the user from deciding a-priori the parametric form of the random variables but is still expressive enough to effectively approximate any continuous distribution and permits efficient learning and inference. Our empirical evidence shows that the architecture, called Mixed SPNs, can indeed capture complex distributions across a wide range of hybrid domains."
                    },
                    {
                        "title": "An Empirical Study on the Generalization Power of Neural Representations Learned via Visual Guessing Games",
                        "abstract": "Guessing games are a prototypical instance of the \"learning by interacting\" paradigm. This work investigates how well an artificial agent can benefit from playing guessing games when later asked to perform on novel NLP downstream tasks such as Visual Question Answering (VQA). We propose two ways to exploit playing guessing games: 1) a supervised learning scenario in which the agent learns to mimic successful guessing games and 2) a novel way for an agent to play by itself, called Self-play via Iterated Experience Learning (SPIEL).   We evaluate the ability of both procedures to generalize: an in-domain evaluation shows an increased accuracy (+7.79) compared with competitors on the evaluation suite CompGuessWhat?!; a transfer evaluation shows improved performance for VQA on the TDIUC dataset in terms of harmonic average accuracy (+5.31) thanks to more fine-grained object representations learned via SPIEL."
                    },
                    {
                        "title": "Tractable Computation of Expected Kernels",
                        "abstract": "Computing the expectation of kernel functions is a ubiquitous task in machine learning, with applications from classical support vector machines to exploiting kernel embeddings of distributions in probabilistic modeling, statistical inference, causal discovery, and deep learning. In all these scenarios, we tend to resort to Monte Carlo estimates as expectations of kernels are intractable in general. In this work, we characterize the conditions under which we can compute expected kernels exactly and efficiently, by leveraging recent advances in probabilistic circuit representations. We first construct a circuit representation for kernels and propose an approach to such tractable computation. We then demonstrate possible advancements for kernel embedding frameworks by exploiting tractable expected kernels to derive new algorithms for two challenging scenarios: 1) reasoning under missing data with kernel support vector regressors; 2) devising a collapsed black-box importance sampling scheme. Finally, we empirically evaluate both algorithms and show that they outperform standard baselines on a variety of datasets."
                    }
                ]
            },
            "404f8cd3-9665-4b09-b855-893d9e2f4c6d": {
                "pk": "404f8cd3-9665-4b09-b855-893d9e2f4c6d",
                "name": "Erik Quaeghebeur",
                "collaborators": [
                    "Gert de Cooman",
                    "Gennaro Gala",
                    "Filip Hermans",
                    "Cassio de Campos",
                    "Robert Peharz",
                    "Enrique Miranda",
                    "Antonio Vergari",
                    "Daniele Grattarola",
                    "Rodrigo L. S. Silva",
                    "Clemens Verhoosel"
                ],
                "domain": [
                    "Probabilistic Modeling",
                    "Uncertainty Quantification",
                    "Graph Neural Networks",
                    "Bayesian Inference"
                ],
                "publications": [
                    {
                        "title": "Characterizing the Set of Coherent Lower Previsions with a Finite Number of Constraints or Vertices",
                        "abstract": "The standard coherence criterion for lower previsions is expressed using an infinite number of linear constraints. For lower previsions that are essentially defined on some finite set of gambles on a finite possibility space, we present a reformulation of this criterion that only uses a finite number of constraints. Any such lower prevision is coherent if it lies within the convex polytope defined by these constraints. The vertices of this polytope are the extreme coherent lower previsions for the given set of gambles. Our reformulation makes it possible to compute them. We show how this is done and illustrate the procedure and its results."
                    },
                    {
                        "title": "Accept & Reject Statement-Based Uncertainty Models",
                        "abstract": "We develop a framework for modelling and reasoning with uncertainty based on accept and reject statements about gambles. It generalises the frameworks found in the literature based on statements of acceptability, desirability, or favourability and clarifies their relative position. Next to the statement-based formulation, we also provide a translation in terms of preference relations, discuss---as a bridge to existing frameworks---a number of simplified variants, and show the relationship with prevision-based uncertainty models. We furthermore provide an application to modelling symmetry judgements."
                    },
                    {
                        "title": "Exchangeable lower previsions",
                        "abstract": "We extend de Finetti's (1937) notion of exchangeability to finite and countable sequences of variables, when a subject's beliefs about them are modelled using coherent lower previsions rather than (linear) previsions. We prove representation theorems in both the finite and the countable case, in terms of sampling without and with replacement, respectively. We also establish a convergence result for sample means of exchangeable sequences. Finally, we study and solve the problem of exchangeable natural extension: how to find the most conservative (point-wise smallest) coherent and exchangeable lower prevision that dominates a given lower prevision."
                    },
                    {
                        "title": "Exchangeable lower previsions",
                        "abstract": "We extend de Finetti's [Ann. Inst. H. Poincar\\'{e} 7 (1937) 1--68] notion of exchangeability to finite and countable sequences of variables, when a subject's beliefs about them are modelled using coherent lower previsions rather than (linear) previsions. We derive representation theorems in both the finite and countable cases, in terms of sampling without and with replacement, respectively."
                    },
                    {
                        "title": "Imprecise Markov chains and their limit behaviour",
                        "abstract": "When the initial and transition probabilities of a finite Markov chain in discrete time are not well known, we should perform a sensitivity analysis. This can be done by considering as basic uncertainty models the so-called credal sets that these probabilities are known or believed to belong to, and by allowing the probabilities to vary over such sets. This leads to the definition of an imprecise Markov chain. We show that the time evolution of such a system can be studied very efficiently using so-called lower and upper expectations, which are equivalent mathematical representations of credal sets. We also study how the inferred credal set about the state at time n evolves as n goes to infinity: under quite unrestrictive conditions, it converges to a uniquely invariant credal set, regardless of the credal set given for the initial state. This leads to a non-trivial generalisation of the classical Perron-Frobenius Theorem to imprecise Markov chains."
                    },
                    {
                        "title": "Exchangeability and sets of desirable gambles",
                        "abstract": "Sets of desirable gambles constitute a quite general type of uncertainty model with an interesting geometrical interpretation. We give a general discussion of such models and their rationality criteria. We study exchangeability assessments for them, and prove counterparts of de Finetti's finite and infinite representation theorems. We show that the finite representation in terms of count vectors has a very nice geometrical interpretation, and that the representation in terms of frequency vectors is tied up with multivariate Bernstein (basis) polynomials. We also lay bare the relationships between the representations of updated exchangeable models, and discuss conservative inference (natural extension) under exchangeability and the extension of exchangeable sequences."
                    },
                    {
                        "title": "Sensitivity analysis for finite Markov chains in discrete time",
                        "abstract": "When the initial and transition probabilities of a finite Markov chain in discrete time are not well known, we should perform a sensitivity analysis. This is done by considering as basic uncertainty models the so-called credal sets that these probabilities are known or believed to belong to, and by allowing the probabilities to vary over such sets. This leads to the definition of an imprecise Markov chain. We show that the time evolution of such a system can be studied very efficiently using so-called lower and upper expectations. We also study how the inferred credal set about the state at time n evolves as n->infinity: under quite unrestrictive conditions, it converges to a uniquely invariant credal set, regardless of the credal set given for the initial state. This leads to a non-trivial generalisation of the classical Perron-Frobenius Theorem to imprecise Markov chains."
                    },
                    {
                        "title": "E(n)-equivariant Graph Neural Cellular Automata",
                        "abstract": "Cellular automata (CAs) are computational models exhibiting rich dynamics emerging from the local interaction of cells arranged in a regular lattice. Graph CAs (GCAs) generalise standard CAs by allowing for arbitrary graphs rather than regular lattices, similar to how Graph Neural Networks (GNNs) generalise Convolutional NNs. Recently, Graph Neural CAs (GNCAs) have been proposed as models built on top of standard GNNs that can be trained to approximate the transition rule of any arbitrary GCA. Existing GNCAs are anisotropic in the sense that their transition rules are not equivariant to translation, rotation, and reflection of the nodes' spatial locations. However, it is desirable for instances related by such transformations to be treated identically by the model. By replacing standard graph convolutions with E(n)-equivariant ones, we avoid anisotropy by design and propose a class of isotropic automata that we call E(n)-GNCAs. These models are lightweight, but can nevertheless handle large graphs, capture complex dynamics and exhibit emergent self-organising behaviours. We showcase the broad and successful applicability of E(n)-GNCAs on three different tasks: (i) pattern formation, (ii) graph auto-encoding, and (iii) simulation of E(n)-equivariant dynamical systems."
                    },
                    {
                        "title": "Bayesian estimation and uncertainty quantification of a temperature-dependent thermal conductivity",
                        "abstract": "We consider the problem of estimating a temperature-dependent thermal conductivity model (curve) from temperature measurements. We apply a Bayesian estimation approach that takes into account measurement errors and limited prior information of system properties. The approach intertwines system simulation and Markov chain Monte Carlo (MCMC) sampling. We investigate the impact of assuming different model classes - cubic polynomials and piecewise linear functions - their parametrization, and different types of prior information - ranging from uninformative to informative. Piecewise linear functions require more parameters (conductivity values) to be estimated than the four parameters (coefficients or conductivity values) needed for cubic polynomials. The former model class is more flexible, but the latter requires less MCMC samples. While parametrizing polynomials with coefficients may feel more natural, it turns out that parametrizing them using conductivity values is far more natural for the specification of prior information. Robust estimation is possible for all model classes and parametrizations, as long as the prior information is accurate or not too informative. Gaussian Markov random field priors are especially well-suited for piecewise linear functions."
                    },
                    {
                        "title": "Probabilistic Integral Circuits",
                        "abstract": "Continuous latent variables (LVs) are a key ingredient of many generative models, as they allow modelling expressive mixtures with an uncountable number of components. In contrast, probabilistic circuits (PCs) are hierarchical discrete mixtures represented as computational graphs composed of input, sum and product units. Unlike continuous LV models, PCs provide tractable inference but are limited to discrete LVs with categorical (i.e. unordered) states. We bridge these model classes by introducing probabilistic integral circuits (PICs), a new language of computational graphs that extends PCs with integral units representing continuous LVs. In the first place, PICs are symbolic computational graphs and are fully tractable in simple cases where analytical integration is possible. In practice, we parameterise PICs with light-weight neural nets delivering an intractable hierarchical continuous mixture that can be approximated arbitrarily well with large PCs using numerical quadrature. On several distribution estimation benchmarks, we show that such PIC-approximating PCs systematically outperform PCs commonly learned via expectation-maximization or SGD."
                    },
                    {
                        "title": "What is the Relationship between Tensor Factorizations and Circuits (and How Can We Exploit it)?",
                        "abstract": "This paper establishes a rigorous connection between circuit representations and tensor factorizations, two seemingly distinct yet fundamentally related areas. By connecting these fields, we highlight a series of opportunities that can benefit both communities. Our work generalizes popular tensor factorizations within the circuit language, and unifies various circuit learning algorithms under a single, generalized hierarchical factorization framework. Specifically, we introduce a modular \"Lego block\" approach to build tensorized circuit architectures. This, in turn, allows us to systematically construct and explore various circuit and tensor factorization models while maintaining tractability. This connection not only clarifies similarities and differences in existing models, but also enables the development of a comprehensive pipeline for building and optimizing new circuit/tensor factorization architectures. We show the effectiveness of our framework through extensive empirical evaluations, and highlight new research opportunities for tensor factorizations in probabilistic modeling."
                    },
                    {
                        "title": "Continuous Mixtures of Tractable Probabilistic Models",
                        "abstract": "Probabilistic models based on continuous latent spaces, such as variational autoencoders, can be understood as uncountable mixture models where components depend continuously on the latent code. They have proven to be expressive tools for generative and probabilistic modelling, but are at odds with tractable probabilistic inference, that is, computing marginals and conditionals of the represented probability distribution. Meanwhile, tractable probabilistic models such as probabilistic circuits (PCs) can be understood as hierarchical discrete mixture models, and thus are capable of performing exact inference efficiently but often show subpar performance in comparison to continuous latent-space models. In this paper, we investigate a hybrid approach, namely continuous mixtures of tractable models with a small latent dimension. While these models are analytically intractable, they are well amenable to numerical integration schemes based on a finite set of integration points. With a large enough number of integration points the approximation becomes de-facto exact. Moreover, for a finite set of integration points, the integration method effectively compiles the continuous mixture into a standard PC. In experiments, we show that this simple scheme proves remarkably effective, as PCs learnt this way set new state of the art for tractable models on many standard density estimation benchmarks."
                    }
                ]
            }
        }
    },
    "2310.04560": {
        "paper_data": {
            "title": "Talk like a Graph: Encoding Graphs for Large Language Models",
            "url": "http://arxiv.org/abs/2310.04560v1",
            "arxiv_id": "2310.04560",
            "authors": [
                "Bahare Fatemi",
                "Jonathan Halcrow",
                "Bryan Perozzi"
            ],
            "abstract": "Graphs are a powerful tool for representing and analyzing complex relationships in real-world applications such as social networks, recommender systems, and computational finance. Reasoning on graphs is essential for drawing inferences about the relationships between entities in a complex system, and to identify hidden patterns and trends. Despite the remarkable progress in automated reasoning with natural text, reasoning on graphs with large language models (LLMs) remains an understudied problem. In this work, we perform the first comprehensive study of encoding graph-structured data as text for consumption by LLMs. We show that LLM performance on graph reasoning tasks varies on three fundamental levels: (1) the graph encoding method, (2) the nature of the graph task itself, and (3) interestingly, the very structure of the graph considered. These novel results provide valuable insight on strategies for encoding graphs as text. Using these insights we illustrate how the correct choice of encoders can boost performance on graph reasoning tasks inside LLMs by 4.8% to 61.8%, depending on the task.",
            "introduction": "   1 Introduction  There has been remarkable recent progress in the research and applications of large language models\u00a0(LLMs)\u00a0(Vaswani et\u00a0al., 2017; Devlin et\u00a0al., 2018; Brown et\u00a0al., 2020a; Ouyang et\u00a0al., 2022). These generative models have captivated the artificial intelligence community and a plethora of models trained on a variety of tasks and modalities have recently been released (Zhao et\u00a0al., 2023). All of these advancements have led to a growing consensus that LLMs are a pivotal advancement on the path to artificial general intelligence (AGI) (Bubeck et\u00a0al., 2023).   However, despite all their successes, there are a number of limitations with the current methodology of design and implementation of LLMs. One of the most obvious limitations is their reliance on unstructured text, causing the models to sometimes miss obvious logical entailments or hallucinate incorrect conclusions (Zhang et\u00a0al., 2023b). Another is that LLMs are fundamentally limited by when they were trained, and it can be difficult to incorporate \u2018fresh\u2019 information about the state of the world which has changed (Lewis et\u00a0al., 2020). Graph-structured data is one of the most flexible ways to represent information and could be a promising solution to both challenges (Schneider et\u00a0al., 2022; Pan et\u00a0al., 2023).   Interestingly, despite this promise, the intersection of graphs and LLMs has been relatively understudied. For example, while much work has focused on LLMs and graph databases (or knowledge graphs (Guu et\u00a0al., 2020; Lewis et\u00a0al., 2020)) there has not been much study about general purpose use of graph-structured data. More recently, Wang et\u00a0al. (2023) have sought to address this by designing a graph benchmarking task for language models. While their task represents an exciting initial foray into measuring LLMs graph reasoning capabilities, there are still many open questions due to the omission of several natural graph tasks and a lack of variety in the type of graph structure considered. Other recent work seeks to replace graph-structured data with LLMs (Ye et\u00a0al., 2023), but this does not address fundamental challenges with LLMs.   In this work, we perform the first comprehensive study about reasoning over graph-structured data as text for consumption by LLMs. To analyze graph reasoning more closely, we decompose the problem into graph encoding and graph prompt engineering. Varying graph encoding methods allows us to understand how LLM\u2019s learned representations are leveraged in graph tasks. While studying prompt engineering techniques finds the most suitable way to get a desired solution to a question from an LLM. Our experimental results seek to uncover the situations where different prompt heuristics work well. To that end, we propose a new set of benchmarks GraphQA for measuring LLM performance reasoning over graph data. GraphQA is distinguished by using graphs with much more varied and realistic graph structure than has previously been studied with LLMs.   Our Contributions: Specifically, the contributions of our work are the following:     1.  An extensive study of graph-structure prompting techniques for use in LLMs.    2.  Insights and best practices for encoding graphs as text for use in LLMs.    3.  A new graph benchmark (GraphQA) to aid the community in studying the effects of graph structure on LLM prompting further.        2 Prompting LLMs for Graph Reasoning  Notation. Let f\ud835\udc53fitalic_f be the interface function to a generative AI model, which takes high-dimensional discrete input tokens W\ud835\udc4aWitalic_W and produces output in the",
            "references": [
                {
                    "title": "Large Language Models as Optimizers",
                    "abstract": "Optimization is ubiquitous. While derivative-based algorithms have been powerful tools for various problems, the absence of gradient imposes challenges on many real-world applications. In this work, we propose Optimization by PROmpting (OPRO), a simple and effective approach to leverage large language models (LLMs) as optimizers, where the optimization task is described in natural language. In each optimization step, the LLM generates new solutions from the prompt that contains previously generated solutions with their values, then the new solutions are evaluated and added to the prompt for the next optimization step. We first showcase OPRO on linear regression and traveling salesman problems, then move on to our main application in prompt optimization, where the goal is to find instructions that maximize the task accuracy. With a variety of LLMs, we demonstrate that the best prompts optimized by OPRO outperform human-designed prompts by up to 8% on GSM8K, and by up to 50% on Big-Bench Hard tasks. Code at https://github.com/google-deepmind/opro."
                },
                {
                    "title": "Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Models",
                    "abstract": "While large language models (LLMs) have demonstrated remarkable capabilities across a range of downstream tasks, a significant concern revolves around their propensity to exhibit hallucinations: LLMs occasionally generate content that diverges from the user input, contradicts previously generated context, or misaligns with established world knowledge. This phenomenon poses a substantial challenge to the reliability of LLMs in real-world scenarios. In this paper, we survey recent efforts on the detection, explanation, and mitigation of hallucination, with an emphasis on the unique challenges posed by LLMs. We present taxonomies of the LLM hallucination phenomena and evaluation benchmarks, analyze existing approaches aiming at mitigating LLM hallucination, and discuss potential directions for future research."
                },
                {
                    "title": "Language is All a Graph Needs",
                    "abstract": "The emergence of large-scale pre-trained language models has revolutionized various AI research domains. Transformers-based Large Language Models (LLMs) have gradually replaced CNNs and RNNs to unify fields of computer vision and natural language processing. Compared with independent data like images, videos or texts, graphs usually contain rich structural and relational information. Meanwhile, languages, especially natural language, being one of the most expressive mediums, excels in describing complex structures. However, existing work on incorporating graph problems into the generative language modeling framework remains very limited. Considering the rising prominence of LLMs, it becomes essential to explore whether LLMs can also replace GNNs as the foundation model for graphs. In this paper, we propose InstructGLM (Instruction-finetuned Graph Language Model) with highly scalable prompts based on natural language instructions. We use natural language to describe multi-scale geometric structure of the graph and then instruction finetune an LLM to perform graph tasks, which enables Generative Graph Learning. Our method surpasses all GNN baselines on ogbn-arxiv, Cora and PubMed datasets, underscoring its effectiveness and sheds light on generative LLMs as new foundation model for graph machine learning. Our code is available at https://github.com/agiresearch/InstructGLM."
                },
                {
                    "title": "Examining the Effects of Degree Distribution and Homophily in Graph Learning Models",
                    "abstract": "Despite a surge in interest in GNN development, homogeneity in benchmarking datasets still presents a fundamental issue to GNN research. GraphWorld is a recent solution which uses the Stochastic Block Model (SBM) to generate diverse populations of synthetic graphs for benchmarking any GNN task. Despite its success, the SBM imposed fundamental limitations on the kinds of graph structure GraphWorld could create. In this work we examine how two additional synthetic graph generators can improve GraphWorld's evaluation; LFR, a well-established model in the graph clustering literature and CABAM, a recent adaptation of the Barabasi-Albert model tailored for GNN benchmarking. By integrating these generators, we significantly expand the coverage of graph space within the GraphWorld framework while preserving key graph properties observed in real-world networks. To demonstrate their effectiveness, we generate 300,000 graphs to benchmark 11 GNN models on a node classification task. We find GNN performance variations in response to homophily, degree distribution and feature signal. Based on these findings, we classify models by their sensitivity to the new generators under these properties. Additionally, we release the extensions made to GraphWorld on the GitHub repository, offering further evaluation of GNN performance on new graphs."
                },
                {
                    "title": "Exploring the Potential of Large Language Models (LLMs)in Learning on Graphs",
                    "abstract": "Learning on Graphs has attracted immense attention due to its wide real-world applications. The most popular pipeline for learning on graphs with textual node attributes primarily relies on Graph Neural Networks (GNNs), and utilizes shallow text embedding as initial node representations, which has limitations in general knowledge and profound semantic understanding. In recent years, Large Language Models (LLMs) have been proven to possess extensive common knowledge and powerful semantic comprehension abilities that have revolutionized existing workflows to handle text data. In this paper, we aim to explore the potential of LLMs in graph machine learning, especially the node classification task, and investigate two possible pipelines: LLMs-as-Enhancers and LLMs-as-Predictors. The former leverages LLMs to enhance nodes' text attributes with their massive knowledge and then generate predictions through GNNs. The latter attempts to directly employ LLMs as standalone predictors. We conduct comprehensive and systematical studies on these two pipelines under various settings. From comprehensive empirical results, we make original observations and find new insights that open new possibilities and suggest promising directions to leverage LLMs for learning on graphs. Our codes and datasets are available at: https://github.com/CurryTang/Graph-LLM ."
                },
                {
                    "title": "Exploring the MIT Mathematics and EECS Curriculum Using Large Language Models",
                    "abstract": "We curate a comprehensive dataset of 4,550 questions and solutions from problem sets, midterm exams, and final exams across all MIT Mathematics and Electrical Engineering and Computer Science (EECS) courses required for obtaining a degree. We evaluate the ability of large language models to fulfill the graduation requirements for any MIT major in Mathematics and EECS. Our results demonstrate that GPT-3.5 successfully solves a third of the entire MIT curriculum, while GPT-4, with prompt engineering, achieves a perfect solve rate on a test set excluding questions based on images. We fine-tune an open-source large language model on this dataset. We employ GPT-4 to automatically grade model responses, providing a detailed performance breakdown by course, question, and answer type. By embedding questions in a low-dimensional space, we explore the relationships between questions, topics, and classes and discover which questions and classes are required for solving other questions and classes through few-shot learning. Our analysis offers valuable insights into course prerequisites and curriculum design, highlighting language models' potential for learning and improving Mathematics and EECS education."
                },
                {
                    "title": "Unifying Large Language Models and Knowledge Graphs: A Roadmap",
                    "abstract": "Large language models (LLMs), such as ChatGPT and GPT4, are making new waves in the field of natural language processing and artificial intelligence, due to their emergent ability and generalizability. However, LLMs are black-box models, which often fall short of capturing and accessing factual knowledge. In contrast, Knowledge Graphs (KGs), Wikipedia, and Huapu for example, are structured knowledge models that explicitly store rich factual knowledge. KGs can enhance LLMs by providing external knowledge for inference and interpretability. Meanwhile, KGs are difficult to construct and evolve by nature, which challenges the existing methods in KGs to generate new facts and represent unseen knowledge. Therefore, it is complementary to unify LLMs and KGs together and, simultaneously, leverage their advantages. In this article, we present a forward-looking roadmap for the unification of LLMs and KGs. Our roadmap consists of three general frameworks, namely: 1) KG-enhanced LLMs, which incorporate KGs during the pre-training and inference phases of LLMs, or for the purpose of enhancing understanding of the knowledge learned by LLMs; 2) LLM-augmented KGs, that leverage LLMs for different KG tasks such as embedding, completion, construction, graph-to-text generation, and question answering; and 3) Synergized LLMs + KGs, in which LLMs and KGs play equal roles and work in a mutually beneficial way to enhance both LLMs and KGs for bidirectional reasoning driven by both data and knowledge. We review and summarize existing efforts within these three frameworks in our roadmap and pinpoint their future research directions."
                },
                {
                    "title": "GPT4Graph: Can Large Language Models Understand Graph Structured Data ? An Empirical Evaluation and Benchmarking",
                    "abstract": "Large language models~(LLM) like ChatGPT have become indispensable to artificial general intelligence~(AGI), demonstrating excellent performance in various natural language processing tasks. In the real world, graph data is ubiquitous and an essential part of AGI and prevails in domains like social network analysis, bioinformatics and recommender systems. The training corpus of large language models often includes some algorithmic components, which allows them to achieve certain effects on some graph data-related problems. However, there is still little research on their performance on a broader range of graph-structured data. In this study, we conduct an extensive investigation to assess the proficiency of LLMs in comprehending graph data, employing a diverse range of structural and semantic-related tasks. Our analysis encompasses 10 distinct tasks that evaluate the LLMs' capabilities in graph understanding. Through our study, we not only uncover the current limitations of language models in comprehending graph structures and performing associated reasoning tasks but also emphasize the necessity for further advancements and novel approaches to enhance their graph processing capabilities. Our findings contribute valuable insights towards bridging the gap between language models and graph understanding, paving the way for more effective graph mining and knowledge extraction."
                },
                {
                    "title": "Patton: Language Model Pretraining on Text-Rich Networks",
                    "abstract": "A real-world text corpus sometimes comprises not only text documents, but also semantic links between them (e.g., academic papers in a bibliographic network are linked by citations and co-authorships).Text documents and semantic connections form a text-rich network, which empowers a wide range of downstream tasks such as classification and retrieval.However, pretraining methods for such structures are still lacking, making it difficult to build one generic model that can be adapted to various tasks on text-rich networks.Current pretraining objectives, such as masked language modeling, purely model texts and do not take inter-document structure information into consideration.To this end, we propose our PretrAining on TexT-Rich NetwOrk framework Patton.Patton includes two pretraining strategies: network-contextualized masked language modeling and masked node prediction, to capture the inherent dependency between textual attributes and network structure.We conduct experiments on four downstream tasks in five datasets from both academic and e-commerce domains, where Patton outperforms baselines significantly and consistently."
                },
                {
                    "title": "Can Language Models Solve Graph Problems in Natural Language?",
                    "abstract": "Large language models (LLMs) are increasingly adopted for a variety of tasks with implicit graphical structures, such as planning in robotics, multi-hop question answering or knowledge probing, structured commonsense reasoning, and more. While LLMs have advanced the state-of-the-art on these tasks with structure implications, whether LLMs could explicitly process textual descriptions of graphs and structures, map them to grounded conceptual spaces, and perform structured operations remains underexplored. To this end, we propose NLGraph (Natural Language Graph), a comprehensive benchmark of graph-based problem solving designed in natural language. NLGraph contains 29,370 problems, covering eight graph reasoning tasks with varying complexity from simple tasks such as connectivity and shortest path up to complex problems such as maximum flow and simulating graph neural networks. We evaluate LLMs (GPT-3/4) with various prompting approaches on the NLGraph benchmark and find that 1) language models do demonstrate preliminary graph reasoning abilities, 2) the benefit of advanced prompting and in-context learning diminishes on more complex graph problems, while 3) LLMs are also (un)surprisingly brittle in the face of spurious correlations in graph and problem settings. We then propose Build-a-Graph Prompting and Algorithmic Prompting, two instruction-based approaches to enhance LLMs in solving natural language graph problems. Build-a-Graph and Algorithmic prompting improve the performance of LLMs on NLGraph by 3.07% to 16.85% across multiple tasks and settings, while how to solve the most complicated graph reasoning tasks in our setup with language models remains an open research question. The NLGraph benchmark and evaluation code are available at https://github.com/Arthur-Heng/NLGraph."
                },
                {
                    "title": "PaLM 2 Technical Report",
                    "abstract": "We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities and is more compute-efficient than its predecessor PaLM. PaLM 2 is a Transformer-based model trained using a mixture of objectives. Through extensive evaluations on English and multilingual language, and reasoning tasks, we demonstrate that PaLM 2 has significantly improved quality on downstream tasks across different model sizes, while simultaneously exhibiting faster and more efficient inference compared to PaLM. This improved efficiency enables broader deployment while also allowing the model to respond faster, for a more natural pace of interaction. PaLM 2 demonstrates robust reasoning capabilities exemplified by large improvements over PaLM on BIG-Bench and other reasoning tasks. PaLM 2 exhibits stable performance on a suite of responsible AI evaluations, and enables inference-time control over toxicity without additional overhead or impact on other capabilities. Overall, PaLM 2 achieves state-of-the-art performance across a diverse set of tasks and capabilities. When discussing the PaLM 2 family, it is important to distinguish between pre-trained models (of various sizes), fine-tuned variants of these models, and the user-facing products that use these models. In particular, user-facing products typically include additional pre- and post-processing steps. Additionally, the underlying models may evolve over time. Therefore, one should not expect the performance of user-facing products to exactly match the results reported in this report."
                },
                {
                    "title": "Automatic Prompt Optimization with \"Gradient Descent\" and Beam Search",
                    "abstract": "Large Language Models (LLMs) have shown impressive performance as general purpose agents, but their abilities remain highly dependent on prompts which are hand written with onerous trial-and-error effort. We propose a simple and nonparametric solution to this problem, Automatic Prompt Optimization (APO), which is inspired by numerical gradient descent to automatically improve prompts, assuming access to training data and an LLM API. The algorithm uses minibatches of data to form natural language\"gradients\"that criticize the current prompt. The gradients are then\"propagated\"into the prompt by editing the prompt in the opposite semantic direction of the gradient. These gradient descent steps are guided by a beam search and bandit selection procedure which significantly improves algorithmic efficiency. Preliminary results across three benchmark NLP tasks and the novel problem of LLM jailbreak detection suggest that Automatic Prompt Optimization can outperform prior prompt editing techniques and improve an initial prompt's performance by up to 31%, by using data to rewrite vague task descriptions into more precise annotation instructions."
                },
                {
                    "title": "A Survey of Large Language Models",
                    "abstract": "Language is essentially a complex, intricate system of human expressions governed by grammatical rules. It poses a significant challenge to develop capable AI algorithms for comprehending and grasping a language. As a major approach, language modeling has been widely studied for language understanding and generation in the past two decades, evolving from statistical language models to neural language models. Recently, pre-trained language models (PLMs) have been proposed by pre-training Transformer models over large-scale corpora, showing strong capabilities in solving various NLP tasks. Since researchers have found that model scaling can lead to performance improvement, they further study the scaling effect by increasing the model size to an even larger size. Interestingly, when the parameter scale exceeds a certain level, these enlarged language models not only achieve a significant performance improvement but also show some special abilities that are not present in small-scale language models. To discriminate the difference in parameter scale, the research community has coined the term large language models (LLM) for the PLMs of significant size. Recently, the research on LLMs has been largely advanced by both academia and industry, and a remarkable progress is the launch of ChatGPT, which has attracted widespread attention from society. The technical evolution of LLMs has been making an important impact on the entire AI community, which would revolutionize the way how we develop and use AI algorithms. In this survey, we review the recent advances of LLMs by introducing the background, key findings, and mainstream techniques. In particular, we focus on four major aspects of LLMs, namely pre-training, adaptation tuning, utilization, and capacity evaluation. Besides, we also summarize the available resources for developing LLMs and discuss the remaining issues for future directions."
                },
                {
                    "title": "Sparks of Artificial General Intelligence: Early experiments with GPT-4",
                    "abstract": "Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. The latest model developed by OpenAI, GPT-4, was trained using an unprecedented scale of compute and data. In this paper, we report on our investigation of an early version of GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT-4 is part of a new cohort of LLMs (along with ChatGPT and Google's PaLM for example) that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4's performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4's capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system. In our exploration of GPT-4, we put special emphasis on discovering its limitations, and we discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for pursuing a new paradigm that moves beyond next-word prediction. We conclude with reflections on societal influences of the recent technological leap and future research directions."
                },
                {
                    "title": "Attending to Graph Transformers",
                    "abstract": "Recently, transformer architectures for graphs emerged as an alternative to established techniques for machine learning with graphs, such as (message-passing) graph neural networks. So far, they have shown promising empirical results, e.g., on molecular prediction datasets, often attributed to their ability to circumvent graph neural networks' shortcomings, such as over-smoothing and over-squashing. Here, we derive a taxonomy of graph transformer architectures, bringing some order to this emerging field. We overview their theoretical properties, survey structural and positional encodings, and discuss extensions for important graph classes, e.g., 3D molecular graphs. Empirically, we probe how well graph transformers can recover various graph properties, how well they can deal with heterophilic graphs, and to what extent they prevent over-squashing. Further, we outline open challenges and research direction to stimulate future work. Our code is available at https://github.com/luis-mueller/probing-graph-transformers."
                },
                {
                    "title": "LAMBADA: Backward Chaining for Automated Reasoning in Natural Language",
                    "abstract": "Remarkable progress has been made on automated reasoning with natural text, by using Large Language Models (LLMs) and methods such as Chain-of-Thought prompting and Selection-Inference. These techniques search for proofs in the forward direction from axioms to the conclusion, which suffers from a combinatorial explosion of the search space, and thus high failure rates for problems requiring longer chains of reasoning. The classical automated reasoning literature has shown that reasoning in the backward direction (i.e. from intended conclusion to supporting axioms) is significantly more efficient at proof-finding. Importing this intuition into the LM setting, we develop a Backward Chaining algorithm, called LAMBADA, that decomposes reasoning into four sub-modules, that are simply implemented by few-shot prompted LLM inference. We show that LAMBADA achieves sizable accuracy boosts over state-of-the-art forward reasoning methods on two challenging logical reasoning datasets, particularly when deep and accurate proof chains are required."
                },
                {
                    "title": "Towards Reasoning in Large Language Models: A Survey",
                    "abstract": "Reasoning is a fundamental aspect of human intelligence that plays a crucial role in activities such as problem solving, decision making, and critical thinking. In recent years, large language models (LLMs) have made significant progress in natural language processing, and there is observation that these models may exhibit reasoning abilities when they are sufficiently large. However, it is not yet clear to what extent LLMs are capable of reasoning. This paper provides a comprehensive overview of the current state of knowledge on reasoning in LLMs, including techniques for improving and eliciting reasoning in these models, methods and benchmarks for evaluating reasoning abilities, findings and implications of previous research in this field, and suggestions on future directions. Our aim is to provide a detailed and up-to-date review of this topic and stimulate meaningful discussion and future work."
                },
                {
                    "title": "Large Language Models Are Human-Level Prompt Engineers",
                    "abstract": "By conditioning on natural language instructions, large language models (LLMs) have displayed impressive capabilities as general-purpose computers. However, task performance depends significantly on the quality of the prompt used to steer the model, and most effective prompts have been handcrafted by humans. Inspired by classical program synthesis and the human approach to prompt engineering, we propose Automatic Prompt Engineer (APE) for automatic instruction generation and selection. In our method, we treat the instruction as the\"program,\"optimized by searching over a pool of instruction candidates proposed by an LLM in order to maximize a chosen score function. To evaluate the quality of the selected instruction, we evaluate the zero-shot performance of another LLM following the selected instruction. Experiments on 24 NLP tasks show that our automatically generated instructions outperform the prior LLM baseline by a large margin and achieve better or comparable performance to the instructions generated by human annotators on 19/24 tasks. We conduct extensive qualitative and quantitative analyses to explore the performance of APE. We show that APE-engineered prompts can be applied to steer models toward truthfulness and/or informativeness, as well as to improve few-shot learning performance by simply prepending them to standard in-context learning prompts. Please check out our webpage at https://sites.google.com/view/automatic-prompt-engineer."
                },
                {
                    "title": "Scaling Instruction-Finetuned Language Models",
                    "abstract": "Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models."
                },
                {
                    "title": "Deep Bidirectional Language-Knowledge Graph Pretraining",
                    "abstract": "Pretraining a language model (LM) on text has been shown to help various downstream NLP tasks. Recent works show that a knowledge graph (KG) can complement text data, offering structured background knowledge that provides a useful scaffold for reasoning. However, these works are not pretrained to learn a deep fusion of the two modalities at scale, limiting the potential to acquire fully joint representations of text and KG. Here we propose DRAGON (Deep Bidirectional Language-Knowledge Graph Pretraining), a self-supervised approach to pretraining a deeply joint language-knowledge foundation model from text and KG at scale. Specifically, our model takes pairs of text segments and relevant KG subgraphs as input and bidirectionally fuses information from both modalities. We pretrain this model by unifying two self-supervised reasoning tasks, masked language modeling and KG link prediction. DRAGON outperforms existing LM and LM+KG models on diverse downstream tasks including question answering across general and biomedical domains, with +5% absolute gain on average. In particular, DRAGON achieves notable performance on complex reasoning about language and knowledge (+10% on questions involving long contexts or multi-step reasoning) and low-resource QA (+8% on OBQA and RiddleSense), and new state-of-the-art results on various BioNLP tasks. Our code and trained models are available at https://github.com/michiyasunaga/dragon."
                },
                {
                    "title": "Decomposed Prompting: A Modular Approach for Solving Complex Tasks",
                    "abstract": "Few-shot prompting is a surprisingly powerful way to use Large Language Models (LLMs) to solve various tasks. However, this approach struggles as the task complexity increases or when the individual reasoning steps of the task themselves are hard to learn, especially when embedded in more complex tasks. To address this, we propose Decomposed Prompting, a new approach to solve complex tasks by decomposing them (via prompting) into simpler sub-tasks that can be delegated to a library of prompting-based LLMs dedicated to these sub-tasks. This modular structure allows each prompt to be optimized for its specific sub-task, further decomposed if necessary, and even easily replaced with more effective prompts, trained models, or symbolic functions if desired. We show that the flexibility and modularity of Decomposed Prompting allows it to outperform prior work on few-shot prompting using GPT3. On symbolic reasoning tasks, we can further decompose sub-tasks that are hard for LLMs into even simpler solvable sub-tasks. When the complexity comes from the input length, we can recursively decompose the task into the same task but with smaller inputs. We also evaluate our approach on textual multi-step reasoning tasks: on long-context multi-hop QA task, we can more effectively teach the sub-tasks via our separate sub-tasks prompts; and on open-domain multi-hop QA, we can incorporate a symbolic information retrieval within our decomposition framework, leading to improved performance on both tasks. Datasets, Code and Prompts available at https://github.com/allenai/DecomP."
                },
                {
                    "title": "A Decade of Knowledge Graphs in Natural Language Processing: A Survey",
                    "abstract": "In pace with developments in the research field of artificial intelligence, knowledge graphs (KGs) have attracted a surge of interest from both academia and industry. As a representation of semantic relations between entities, KGs have proven to be particularly relevant for natural language processing (NLP), experiencing a rapid spread and wide adoption within recent years. Given the increasing amount of research work in this area, several KG-related approaches have been surveyed in the NLP research community. However, a comprehensive study that categorizes established topics and reviews the maturity of individual research streams remains absent to this day. Contributing to closing this gap, we systematically analyzed 507 papers from the literature on KGs in NLP. Our survey encompasses a multifaceted review of tasks, research types, and contributions. As a result, we present a structured overview of the research landscape, provide a taxonomy of tasks, summarize our findings, and highlight directions for future work."
                },
                {
                    "title": "Hidden Schema Networks",
                    "abstract": "Large, pretrained language models infer powerful representations that encode rich semantic and syntactic content, albeit implicitly. In this work we introduce a novel neural language model that enforces, via inductive biases, explicit relational structures which allow for compositionality onto the output representations of pretrained language models. Specifically, the model encodes sentences into sequences of symbols (composed representations), which correspond to the nodes visited by biased random walkers on a global latent graph, and infers the posterior distribution of the latter. We first demonstrate that the model is able to uncover ground-truth graphs from artificially generated datasets of random token sequences. Next, we leverage pretrained BERT and GPT-2 language models as encoder and decoder, respectively, to infer networks of symbols (schemata) from natural language datasets. Our experiments show that (i) the inferred symbols can be interpreted as encoding different aspects of language, as e.g. topics or sentiments, and that (ii) GPT-2-like models can effectively be conditioned on symbolic representations. Finally, we explore training autoregressive, random walk \u201creasoning\u201d models on schema networks inferred from commonsense knowledge databases, and using the sampled paths to enhance the performance of pretrained language models on commonsense If-Then reasoning tasks."
                },
                {
                    "title": "Large Language Models are Zero-Shot Reasoners",
                    "abstract": "Pretrained large language models (LLMs) are widely used in many sub-fields of natural language processing (NLP) and generally known as excellent few-shot learners with task-specific exemplars. Notably, chain of thought (CoT) prompting, a recent technique for eliciting complex multi-step reasoning through step-by-step answer examples, achieved the state-of-the-art performances in arithmetics and symbolic reasoning, difficult system-2 tasks that do not follow the standard scaling laws for LLMs. While these successes are often attributed to LLMs' ability for few-shot learning, we show that LLMs are decent zero-shot reasoners by simply adding\"Let's think step by step\"before each answer. Experimental results demonstrate that our Zero-shot-CoT, using the same single prompt template, significantly outperforms zero-shot LLM performances on diverse benchmark reasoning tasks including arithmetics (MultiArith, GSM8K, AQUA-RAT, SVAMP), symbolic reasoning (Last Letter, Coin Flip), and other logical reasoning tasks (Date Understanding, Tracking Shuffled Objects), without any hand-crafted few-shot examples, e.g. increasing the accuracy on MultiArith from 17.7% to 78.7% and GSM8K from 10.4% to 40.7% with large InstructGPT model (text-davinci-002), as well as similar magnitudes of improvements with another off-the-shelf large model, 540B parameter PaLM. The versatility of this single prompt across very diverse reasoning tasks hints at untapped and understudied fundamental zero-shot capabilities of LLMs, suggesting high-level, multi-task broad cognitive capabilities may be extracted by simple prompting. We hope our work not only serves as the minimal strongest zero-shot baseline for the challenging reasoning benchmarks, but also highlights the importance of carefully exploring and analyzing the enormous zero-shot knowledge hidden inside LLMs before crafting finetuning datasets or few-shot exemplars."
                },
                {
                    "title": "Least-to-Most Prompting Enables Complex Reasoning in Large Language Models",
                    "abstract": "Chain-of-thought prompting has demonstrated remarkable performance on various natural language reasoning tasks. However, it tends to perform poorly on tasks which requires solving problems harder than the exemplars shown in the prompts. To overcome this challenge of easy-to-hard generalization, we propose a novel prompting strategy, least-to-most prompting. The key idea in this strategy is to break down a complex problem into a series of simpler subproblems and then solve them in sequence. Solving each subproblem is facilitated by the answers to previously solved subproblems. Our experimental results on tasks related to symbolic manipulation, compositional generalization, and math reasoning reveal that least-to-most prompting is capable of generalizing to more difficult problems than those seen in the prompts. A notable finding is that when the GPT-3 code-davinci-002 model is used with least-to-most prompting, it can solve the compositional generalization benchmark SCAN in any split (including length split) with an accuracy of at least 99% using just 14 exemplars, compared to only 16% accuracy with chain-of-thought prompting. This is particularly noteworthy because neural-symbolic models in the literature that specialize in solving SCAN are trained on the entire training set containing over 15,000 examples. We have included prompts for all the tasks in the Appendix."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "GraphWorld: Fake Graphs Bring Real Insights for GNNs",
                    "abstract": "Despite advances in the field of Graph Neural Networks (GNNs), only a small number (~5) of datasets are currently used to evaluate new models. This continued reliance on a handful of datasets provides minimal insight into the performance differences between models, and is especially challenging for industrial practitioners who are likely to have datasets which are very different from academic benchmarks. In the course of our work on GNN infrastructure and open-source software at Google, we have sought to develop benchmarks that are robust, tunable, scalable, and generalizable. In this work we introduce GraphWorld, a novel methodology and system for benchmarking GNN models on an arbitrarily-large population ofsynthetic graphs for any conceivable GNN task. GraphWorld allows a user to efficiently generate a world with millions of statistically diverse datasets. It is accessible, scalable, and easy to use. GraphWorld can be run on a single machine without specialized hardware, or it can be easily scaled up to run on arbitrary clusters or cloud frameworks. Using GraphWorld, a user has fine-grained control over graph generator parameters, and can benchmark arbitrary GNN models with built-in hyperparameter tuning. We present insights from GraphWorld experiments regarding the performance characteristics of thirteen GNN models and baselines over millions of benchmark datasets. We further show that GraphWorld efficiently explores regions of benchmark dataset space uncovered by standard benchmarks, revealing comparisons between models that have not been historically obtainable. Using GraphWorld, we also are able to study in-detail the relationship between graph properties and task performance metrics, which is nearly impossible with the classic collection of real-world benchmarks."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Generate & Rank: A Multi-task Framework for Math Word Problems",
                    "abstract": "Math word problem (MWP) is a challenging and critical task in natural language processing. Many recent studies formalize MWP as a generation task and have adopted sequence-to-sequence models to transform problem descriptions to mathematical expressions. However, mathematical expressions are prone to minor mistakes while the generation objective does not explicitly handle such mistakes. To address this limitation, we devise a new ranking task for MWP and propose Generate&Rank, a multi-task framework based on a generative pre-trained language model. By joint training with generation and ranking, the model learns from its own mistakes and is able to distinguish between correct and incorrect expressions. Meanwhile, we perform tree-based disturbance specially designed for MWP and an online update to boost the ranker. We demonstrate the effectiveness of our proposed method on the benchmark and the results show that our method consistently outperforms baselines in all datasets. Particularly, in the classical Math23k, our method is 7% (78.4% $\\rightarrow$ 85.4%) higher than the state-of-the-art."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Knowledge Graph Based Synthetic Corpus Generation for Knowledge-Enhanced Language Model Pre-training",
                    "abstract": "Prior work on Data-To-Text Generation, the task of converting knowledge graph (KG) triples into natural text, focused on domain-specific benchmark datasets. In this paper, however, we verbalize the entire English Wikidata KG, and discuss the unique challenges associated with a broad, open-domain, large-scale verbalization. We further show that verbalizing a comprehensive, encyclopedic KG like Wikidata can be used to integrate structured KGs and natural language corpora. In contrast to the many architectures that have been developed to integrate these two sources, our approach converts the KG into natural text, allowing it to be seamlessly integrated into existing language models. It carries the further advantages of improved factual accuracy and reduced toxicity in the resulting language model. We evaluate this approach by augmenting the retrieval corpus in a retrieval language model and showing significant improvements on the knowledge intensive tasks of open domain QA and the LAMA knowledge probe."
                },
                {
                    "title": "The Power of Scale for Parameter-Efficient Prompt Tuning",
                    "abstract": "In this work, we explore \u201cprompt tuning,\u201d a simple yet effective mechanism for learning \u201csoft prompts\u201d to condition frozen language models to perform specific downstream tasks. Unlike the discrete text prompts used by GPT-3, soft prompts are learned through backpropagation and can be tuned to incorporate signals from any number of labeled examples. Our end-to-end learned approach outperforms GPT-3\u2019s few-shot learning by a large margin. More remarkably, through ablations on model size using T5, we show that prompt tuning becomes more competitive with scale: as models exceed billions of parameters, our method \u201ccloses the gap\u201d and matches the strong performance of model tuning (where all model weights are tuned). This finding is especially relevant because large models are costly to share and serve and the ability to reuse one frozen model for multiple downstream tasks can ease this burden. Our method can be seen as a simplification of the recently proposed \u201cprefix tuning\u201d of Li and Liang (2021) and we provide a comparison to this and other similar approaches. Finally, we show that conditioning a frozen model with soft prompts confers benefits in robustness to domain transfer and enables efficient \u201cprompt ensembling.\u201d We release code and model checkpoints to reproduce our experiments."
                },
                {
                    "title": "Measuring Mathematical Problem Solving With the MATH Dataset",
                    "abstract": "Many intellectual endeavors require mathematical problem solving, but this skill remains beyond the capabilities of computers. To measure this ability in machine learning models, we introduce MATH, a new dataset of 12,500 challenging competition mathematics problems. Each problem in MATH has a full step-by-step solution which can be used to teach models to generate answer derivations and explanations. To facilitate future research and increase accuracy on MATH, we also contribute a large auxiliary pretraining dataset which helps teach models the fundamentals of mathematics. Even though we are able to increase accuracy on MATH, our results show that accuracy remains relatively low, even with enormous Transformer models. Moreover, we find that simply increasing budgets and model parameter counts will be impractical for achieving strong mathematical reasoning if scaling trends continue. While scaling Transformers is automatically solving most other text-based tasks, scaling is not currently solving MATH. To have more traction on mathematical problem solving we will likely need new algorithmic advancements from the broader research community."
                },
                {
                    "title": "What Makes Good In-Context Examples for GPT-3?",
                    "abstract": "GPT-3 has attracted lots of attention due to its superior performance across a wide range of NLP tasks, especially with its in-context learning abilities. Despite its success, we found that the empirical results of GPT-3 depend heavily on the choice of in-context examples. In this work, we investigate whether there are more effective strategies for judiciously selecting in-context examples (relative to random sampling) that better leverage GPT-3\u2019s in-context learning capabilities.Inspired by the recent success of leveraging a retrieval module to augment neural networks, we propose to retrieve examples that are semantically-similar to a test query sample to formulate its corresponding prompt. Intuitively, the examples selected with such a strategy may serve as more informative inputs to unleash GPT-3\u2019s power of text generation. We evaluate the proposed approach on several natural language understanding and generation benchmarks, where the retrieval-based prompt selection approach consistently outperforms the random selection baseline. Moreover, it is observed that the sentence encoders fine-tuned on task-related datasets yield even more helpful retrieval results. Notably, significant gains are observed on tasks such as table-to-text generation (44.3% on the ToTTo dataset) and open-domain question answering (45.5% on the NQ dataset)."
                },
                {
                    "title": "A Generalization of Transformer Networks to Graphs",
                    "abstract": "We propose a generalization of transformer neural network architecture for arbitrary graphs. The original transformer was designed for Natural Language Processing (NLP), which operates on fully connected graphs representing all connections between the words in a sequence. Such architecture does not leverage the graph connectivity inductive bias, and can perform poorly when the graph topology is important and has not been encoded into the node features. We introduce a graph transformer with four new properties compared to the standard model. First, the attention mechanism is a function of the neighborhood connectivity for each node in the graph. Second, the positional encoding is represented by the Laplacian eigenvectors, which naturally generalize the sinusoidal positional encodings often used in NLP. Third, the layer normalization is replaced by a batch normalization layer, which provides faster training and better generalization performance. Finally, the architecture is extended to edge feature representation, which can be critical to tasks s.a. chemistry (bond type) or link prediction (entity relationship in knowledge graphs). Numerical experiments on a graph benchmark demonstrate the performance of the proposed graph transformer architecture. This work closes the gap between the original transformer, which was designed for the limited case of line graphs, and graph neural networks, that can work with arbitrary graphs. As our architecture is simple and generic, we believe it can be used as a black box for future applications that wish to consider transformer and graphs."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks",
                    "abstract": "Large pre-trained language models have been shown to store factual knowledge in their parameters, and achieve state-of-the-art results when fine-tuned on downstream NLP tasks. However, their ability to access and precisely manipulate knowledge is still limited, and hence on knowledge-intensive tasks, their performance lags behind task-specific architectures. Additionally, providing provenance for their decisions and updating their world knowledge remain open research problems. Pre-trained models with a differentiable access mechanism to explicit non-parametric memory can overcome this issue, but have so far been only investigated for extractive downstream tasks. We explore a general-purpose fine-tuning recipe for retrieval-augmented generation (RAG) -- models which combine pre-trained parametric and non-parametric memory for language generation. We introduce RAG models where the parametric memory is a pre-trained seq2seq model and the non-parametric memory is a dense vector index of Wikipedia, accessed with a pre-trained neural retriever. We compare two RAG formulations, one which conditions on the same retrieved passages across the whole generated sequence, the other can use different passages per token. We fine-tune and evaluate our models on a wide range of knowledge-intensive NLP tasks and set the state-of-the-art on three open domain QA tasks, outperforming parametric seq2seq models and task-specific retrieve-and-extract architectures. For language generation tasks, we find that RAG models generate more specific, diverse and factual language than a state-of-the-art parametric-only seq2seq baseline."
                },
                {
                    "title": "Transformers as Soft Reasoners over Language",
                    "abstract": "Beginning with McCarthy's Advice Taker (1959), AI has pursued the goal of providing a system with explicit, general knowledge and having the system reason over that knowledge. However, expressing the knowledge in a formal (logical or probabilistic) representation has been a major obstacle to this research. This paper investigates a modern approach to this problem where the facts and rules are provided as natural language sentences, thus bypassing a formal representation. We train transformers to reason (or emulate reasoning) over these sentences using synthetically generated data. Our models, that we call RuleTakers, provide the first empirical demonstration that this kind of soft reasoning over language is learnable, can achieve high (99%) accuracy, and generalizes to test data requiring substantially deeper chaining than seen during training (95%+ scores). We also demonstrate that the models transfer well to two hand-authored rulebases, and to rulebases paraphrased into more natural language. These findings are significant as it suggests a new role for transformers, namely as limited \"soft theorem provers\" operating over explicit theories in language. This in turn suggests new possibilities for explainability, correctability, and counterfactual reasoning in question-answering."
                },
                {
                    "title": "REALM: Retrieval-Augmented Language Model Pre-Training",
                    "abstract": "Language model pre-training has been shown to capture a surprising amount of world knowledge, crucial for NLP tasks such as question answering. However, this knowledge is stored implicitly in the parameters of a neural network, requiring ever-larger networks to cover more facts. \nTo capture knowledge in a more modular and interpretable way, we augment language model pre-training with a latent knowledge retriever, which allows the model to retrieve and attend over documents from a large corpus such as Wikipedia, used during pre-training, fine-tuning and inference. For the first time, we show how to pre-train such a knowledge retriever in an unsupervised manner, using masked language modeling as the learning signal and backpropagating through a retrieval step that considers millions of documents. \nWe demonstrate the effectiveness of Retrieval-Augmented Language Model pre-training (REALM) by fine-tuning on the challenging task of Open-domain Question Answering (Open-QA). We compare against state-of-the-art models for both explicit and implicit knowledge storage on three popular Open-QA benchmarks, and find that we outperform all previous methods by a significant margin (4-16% absolute accuracy), while also providing qualitative benefits such as interpretability and modularity."
                },
                {
                    "title": "Graph-Bert: Only Attention is Needed for Learning Graph Representations",
                    "abstract": "The dominant graph neural networks (GNNs) over-rely on the graph links, several serious performance problems with which have been witnessed already, e.g., suspended animation problem and over-smoothing problem. What's more, the inherently inter-connected nature precludes parallelization within the graph, which becomes critical for large-sized graph, as memory constraints limit batching across the nodes. In this paper, we will introduce a new graph neural network, namely GRAPH-BERT (Graph based BERT), solely based on the attention mechanism without any graph convolution or aggregation operators. Instead of feeding GRAPH-BERT with the complete large input graph, we propose to train GRAPH-BERT with sampled linkless subgraphs within their local contexts. GRAPH-BERT can be learned effectively in a standalone mode. Meanwhile, a pre-trained GRAPH-BERT can also be transferred to other application tasks directly or with necessary fine-tuning if any supervised label information or certain application oriented objective is available. We have tested the effectiveness of GRAPH-BERT on several graph benchmark datasets. Based the pre-trained GRAPH-BERT with the node attribute reconstruction and structure recovery tasks, we further fine-tune GRAPH-BERT on node classification and graph clustering tasks specifically. The experimental results have demonstrated that GRAPH-BERT can out-perform the existing GNNs in both the learning effectiveness and efficiency."
                },
                {
                    "title": "Random Graphs",
                    "abstract": "An introduction to the mathematics of the Poisson random graph, the simplest model of a random network. The chapter starts with a definition of the model, followed by derivations of basic properties like the mean degree, degree distribution, and clustering coefficient. This is followed with a detailed derivation of the large-scale structural properties of random graphs, including the position of the phase transition at which a giant component appears, the size of the giant component, the average size of the small components, and the expected diameter of the network. The chapter ends with a discussion of some of the shortcomings of the random graph model."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Exploring Network Structure, Dynamics, and Function using NetworkX",
                    "abstract": "NetworkX is a Python language package for exploration and analysis of networks and network algorithms. The core package provides data structures for representing many types of networks, or graphs, including simple graphs, directed graphs, and graphs with parallel edges and self-loops. The nodes in NetworkX graphs can be any (hashable) Python object and edges can contain arbitrary data; this flexibility makes NetworkX ideal for representing networks found in many different scientific fields. In addition to the basic data structures many graph algorithms are implemented for calculating network properties and structure measures: shortest paths, betweenness centrality, clustering, and degree distribution and many more. NetworkX can read and write various graph formats for easy exchange with existing data, and provides generators for many classic graphs and popular graph models, such as the Erdos-Renyi, Small World, and Barabasi-Albert models. The ease-of-use and flexibility of the Python programming language together with connection to the SciPy tools make NetworkX a powerful tool for scientific computations. We discuss some of our recent work studying synchronization of coupled oscillators to demonstrate how NetworkX enables research in the field of computational networks."
                },
                {
                    "title": "Statistical mechanics of complex networks",
                    "abstract": "The emergence of order in natural systems is a constant source of inspiration for both physical and biological sciences. While the spatial order characterizing for example the crystals has been the basis of many advances in contemporary physics, most complex systems in nature do not offer such high degree of order. Many of these systems form complex networks whose nodes are the elements of the system and edges represent the interactions between them. \nTraditionally complex networks have been described by the random graph theory founded in 1959 by Paul Erdohs and Alfred Renyi. One of the defining features of random graphs is that they are statistically homogeneous, and their degree distribution (characterizing the spread in the number of edges starting from a node) is a Poisson distribution. In contrast, recent empirical studies, including the work of our group, indicate that the topology of real networks is much richer than that of random graphs. In particular, the degree distribution of real networks is a power-law, indicating a heterogeneous topology in which the majority of the nodes have a small degree, but there is a significant fraction of highly connected nodes that play an important role in the connectivity of the network. \nThe scale-free topology of real networks has very important consequences on their functioning. For example, we have discovered that scale-free networks are extremely resilient to the random disruption of their nodes. On the other hand, the selective removal of the nodes with highest degree induces a rapid breakdown of the network to isolated subparts that cannot communicate with each other. \nThe non-trivial scaling of the degree distribution of real networks is also an indication of their assembly and evolution. Indeed, our modeling studies have shown us that there are general principles governing the evolution of networks. Most networks start from a small seed and grow by the addition of new nodes which attach to the nodes already in the system. This process obeys preferential attachment: the new nodes are more likely to connect to nodes with already high degree. We have proposed a simple model based on these two principles wich was able to reproduce the power-law degree distribution of real networks. Perhaps even more importantly, this model paved the way to a new paradigm of network modeling, trying to capture the evolution of networks, not just their static topology."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Emergence of Scaling in Random Networks",
                    "abstract": "level of Co and a maximum of inverse TMR is expected when the Fermi level of LSMO is approximately at the maximum of the spin2 DOS of Co. This is consistent with the maximum of inverse TMR observed at 20.4 V for Co/STO/LSMO junctions (Fig. 3A). For a positive bias, the TMR is expected to change sign and become normal above 1 V when the Fermi level of LSMO goes down into the energy range of the majority spin d-band of Co. This is also observed in Fig. 3A. For ALO and ALO/STO barriers, a predominant tunneling of s-character electrons (see arrow in Fig. 2B) is the usual explanation of the positive polarization (6\u20138). The rapid drop with bias (Fig. 3B) is similar to what has been observed in most junctions with ALO barriers, and completely different from what is obtained when the tunneling is predominantly by d-character electrons (Fig. 3A). The origin of this rapid decrease of the TMR at relatively small bias has never been clearly explained. This is roughly consistent with the energy dependence of the DOS induced by sp-d bonding effects on the first atomic layer of ALO in the calculation of Nguyen-Mahn et al. (8) for the Co-ALO interface. But Zhang et al. (13) have also shown that a large part of the TMR drop can be attributed to the excitation of spin waves. The experiments reported here and in several recent publications (3, 4) demonstrate the important role of the electronic structure of the metal-oxide interface in determining the spin polarization of the tunneling electrons. The negative polarization for the Co-STO interface has been ascribed to d-d bonding effects between Al and Ti (4). This interpretation is similar to that proposed to explain, in terms of sp-d bonding, the positive polarization at the Co-ALO interface (8). However, there is no general theory predicting the trend of the experimental results for Co\u2014that is, a negative polarization with oxides of d elements (STO, CLO, Ta2O5) and a positive one when there are only s and p states (ALO). It is likely that the spin polarization should also depend on the position of the Fermi level with respect to the electronic levels of each character above and below the gap of the insulator. In addition, as an evanescent wave in an insulator is a Bloch wave with an imaginary wave vector, one can expect different decay lengths for Bloch waves of different character. This means that the final polarization could also depend on the thickness of the barrier, as illustrated by the calculations of MacLaren et al. for Fe/ZnSe/Fe junctions (14). The influence of the barrier on the spin polarization opens new ways to shape and optimize the TMR. Interesting bias dependencies can be obtained with barriers selecting the d electrons and probing the fine structure of the d-DOS, as in Fig. 3A. The DOS of a d-band can also be easily tailored by alloying (for example, by introduction of virtual bound states) to produce specific bias dependencies. Although here we concentrated on the problem of the spin polarization of the Co electrode and regarded the strongly spin-polarized LSMO only as a useful spin analyzer, the large TMR ratios obtained by combining Co and LSMO electrodes (50% with a STO barrier) are also an interesting result. The drawback arising from the low Curie temperature of LSMO (;350 K) is the reduction of the TMR at room temperature, down to about 5% at 300 K in Co/STO/ LSMO (4). However, other types of oxides of the double-perovskite family (for example, Sr2FeMoO6) combine electronic properties similar to those of manganites with a definitely higher Curie temperature (15). Their use in magnetic tunnel junctions is promising for a new generation of tunnel junctions with very high magnetoresistance for room-temperature applications."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively leverage graph-structured data to enhance the reasoning capabilities of large language models (LLMs)?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of how LLMs can process and reason with structured information, which is essential for developing more robust AI systems. By integrating graph-structured data, we can improve LLMs' ability to handle logical entailments and incorporate up-to-date information, potentially leading to practical applications in areas such as knowledge representation, natural language understanding, and artificial general intelligence (AGI). This research could pave the way for future studies that explore the intersection of LLMs and various data structures, ultimately enhancing the capabilities of AI systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem include the complexity of effectively encoding graph-structured data into a format that LLMs can understand and reason with. Naive approaches may fail because they do not account for the nuances of graph representation or the specific requirements of LLMs in processing such data. Additionally, there are theoretical obstacles related to how LLMs learn and generalize from structured information, as well as practical issues in designing effective prompting techniques that yield accurate and meaningful responses from the models.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the interaction between LLMs and graph databases or knowledge graphs, neglecting the broader application of graph-structured data in general-purpose reasoning tasks. Existing solutions often lack the variety and complexity needed to fully explore the potential of graph reasoning in LLMs. Barriers such as limited benchmarks and a narrow focus on specific graph tasks have hindered progress. Our approach differs by providing a comprehensive study of graph encoding and prompting techniques, along with the introduction of the GraphQA benchmark, which encompasses a wider range of graph structures and reasoning tasks.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves two main components: graph encoding and graph prompt engineering. We will experiment with various graph encoding methods to assess how LLMs leverage learned representations in graph tasks. Additionally, we will explore different prompting techniques to identify the most effective strategies for eliciting desired responses from LLMs. The dataset will consist of diverse graph structures, and we will use the GraphQA benchmark to evaluate LLM performance"
            }
        },
        "author_data": {
            "6acafb11-d787-47cc-950b-433d12ed69fc": {
                "pk": "6acafb11-d787-47cc-950b-433d12ed69fc",
                "name": "Bahare Fatemi",
                "collaborators": [
                    "Bryan Perozzi",
                    "Anton Tsitsulin",
                    "Mehran Kazemi",
                    "Jonathan J. Halcrow",
                    "Seyed Mehran Kazemi",
                    "Sami Abu-El-Haija",
                    "Perouz Taslakian",
                    "Dustin Zelle",
                    "Nishanth Dikkala",
                    "Neslihan Bulut"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Large Language Model",
                    "Causal Reasoning",
                    "Knowledge Graph"
                ],
                "publications": [
                    {
                        "title": "Let Your Graph Do the Talking: Encoding Structured Data for LLMs",
                        "abstract": "How can we best encode structured data into sequential form for use in large language models (LLMs)? In this work, we introduce a parameter-efficient method to explicitly represent structured data for LLMs. Our method, GraphToken, learns an encoding function to extend prompts with explicit structured information. Unlike other work which focuses on limited domains (e.g. knowledge graph representation), our work is the first effort focused on the general encoding of structured data to be used for various reasoning tasks. We show that explicitly representing the graph structure allows significant improvements to graph reasoning tasks. Specifically, we see across the board improvements - up to 73% points - on node, edge and, graph-level tasks from the GraphQA benchmark."
                    },
                    {
                        "title": "Text-space Graph Foundation Models: Comprehensive Benchmarks and New Insights",
                        "abstract": "Given the ubiquity of graph data and its applications in diverse domains, building a Graph Foundation Model (GFM) that can work well across different graphs and tasks with a unified backbone has recently garnered significant interests. A major obstacle to achieving this goal stems from the fact that graphs from different domains often exhibit diverse node features. Inspired by multi-modal models that align different modalities with natural language, the text has recently been adopted to provide a unified feature space for diverse graphs. Despite the great potential of these text-space GFMs, current research in this field is hampered by two problems. First, the absence of a comprehensive benchmark with unified problem settings hinders a clear understanding of the comparative effectiveness and practical value of different text-space GFMs. Second, there is a lack of sufficient datasets to thoroughly explore the methods' full potential and verify their effectiveness across diverse settings. To address these issues, we conduct a comprehensive benchmark providing novel text-space datasets and comprehensive evaluation under unified problem settings. Empirical results provide new insights and inspire future research directions. Our code and data are publicly available from \\url{https://github.com/CurryTang/TSGFM}."
                    },
                    {
                        "title": "Test of Time: A Benchmark for Evaluating LLMs on Temporal Reasoning",
                        "abstract": "Large language models (LLMs) have showcased remarkable reasoning capabilities, yet they remain susceptible to errors, particularly in temporal reasoning tasks involving complex temporal logic. Existing research has explored LLM performance on temporal reasoning using diverse datasets and benchmarks. However, these studies often rely on real-world data that LLMs may have encountered during pre-training or employ anonymization techniques that can inadvertently introduce factual inconsistencies. In this work, we address these limitations by introducing novel synthetic datasets specifically designed to assess LLM temporal reasoning abilities in various scenarios. The diversity of question types across these datasets enables systematic investigation into the impact of the problem structure, size, question type, fact order, and other factors on LLM performance. Our findings provide valuable insights into the strengths and weaknesses of current LLMs in temporal reasoning tasks. To foster further research in this area, we are open-sourcing the datasets and evaluation framework used in our experiments: https://huggingface.co/datasets/baharef/ToT."
                    },
                    {
                        "title": "CausalGraph2LLM: Evaluating LLMs for Causal Queries",
                        "abstract": "Causality is essential in scientific research, enabling researchers to interpret true relationships between variables. These causal relationships are often represented by causal graphs, which are directed acyclic graphs. With the recent advancements in Large Language Models (LLMs), there is an increasing interest in exploring their capabilities in causal reasoning and their potential use to hypothesize causal graphs. These tasks necessitate the LLMs to encode the causal graph effectively for subsequent downstream tasks. In this paper, we propose a comprehensive benchmark, \\emph{CausalGraph2LLM}, encompassing a variety of causal graph settings to assess the causal graph understanding capability of LLMs. We categorize the causal queries into two types: graph-level and node-level queries. We benchmark both open-sourced and closed models for our study. Our findings reveal that while LLMs show promise in this domain, they are highly sensitive to the encoding used. Even capable models like GPT-4 and Gemini-1.5 exhibit sensitivity to encoding, with deviations of about $60\\%$. We further demonstrate this sensitivity for downstream causal intervention tasks. Moreover, we observe that LLMs can often display biases when presented with contextual information about a causal graph, potentially stemming from their parametric memory."
                    },
                    {
                        "title": "ReMI: A Dataset for Reasoning with Multiple Images",
                        "abstract": "With the continuous advancement of large language models (LLMs), it is essential to create new benchmarks to effectively evaluate their expanding capabilities and identify areas for improvement. This work focuses on multi-image reasoning, an emerging capability in state-of-the-art LLMs. We introduce ReMI, a dataset designed to assess LLMs' ability to Reason with Multiple Images. This dataset encompasses a diverse range of tasks, spanning various reasoning domains such as math, physics, logic, code, table/chart understanding, and spatial and temporal reasoning. It also covers a broad spectrum of characteristics found in multi-image reasoning scenarios. We have benchmarked several cutting-edge LLMs using ReMI and found a substantial gap between their performance and human-level proficiency. This highlights the challenges in multi-image reasoning and the need for further research. Our analysis also reveals the strengths and weaknesses of different models, shedding light on the types of reasoning that are currently attainable and areas where future models require improvement. To foster further research in this area, we are releasing ReMI publicly: https://huggingface.co/datasets/mehrankazemi/ReMI."
                    },
                    {
                        "title": "Understanding Transformer Reasoning Capabilities via Graph Algorithms",
                        "abstract": "Which transformer scaling regimes are able to perfectly solve different classes of algorithmic problems? While tremendous empirical advances have been attained by transformer-based neural networks, a theoretical understanding of their algorithmic reasoning capabilities in realistic parameter regimes is lacking. We investigate this question in terms of the network's depth, width, and number of extra tokens for algorithm execution. Our novel representational hierarchy separates 9 algorithmic reasoning problems into classes solvable by transformers in different realistic parameter scaling regimes. We prove that logarithmic depth is necessary and sufficient for tasks like graph connectivity, while single-layer transformers with small embedding dimensions can solve contextual retrieval tasks. We also support our theoretical analysis with ample empirical evidence using the GraphQA benchmark. These results show that transformers excel at many graph reasoning tasks, even outperforming specialized graph neural networks."
                    },
                    {
                        "title": "Michelangelo: Long Context Evaluations Beyond Haystacks via Latent Structure Queries",
                        "abstract": "We introduce Michelangelo: a minimal, synthetic, and unleaked long-context reasoning evaluation for large language models which is also easy to automatically score. This evaluation is derived via a novel, unifying framework for evaluations over arbitrarily long contexts which measure the model's ability to do more than retrieve a single piece of information from its context. The central idea of the Latent Structure Queries framework (LSQ) is to construct tasks which require a model to ``chisel away'' the irrelevant information in the context, revealing a latent structure in the context. To verify a model's understanding of this latent structure, we query the model for details of the structure. Using LSQ, we produce three diagnostic long-context evaluations across code and natural-language domains intended to provide a stronger signal of long-context language model capabilities. We perform evaluations on several state-of-the-art models and demonstrate both that a) the proposed evaluations are high-signal and b) that there is significant room for improvement in synthesizing long-context information."
                    },
                    {
                        "title": "Don't Forget to Connect! Improving RAG with Graph-based Reranking",
                        "abstract": "Retrieval Augmented Generation (RAG) has greatly improved the performance of Large Language Model (LLM) responses by grounding generation with context from existing documents. These systems work well when documents are clearly relevant to a question context. But what about when a document has partial information, or less obvious connections to the context? And how should we reason about connections between documents? In this work, we seek to answer these two core questions about RAG generation. We introduce G-RAG, a reranker based on graph neural networks (GNNs) between the retriever and reader in RAG. Our method combines both connections between documents and semantic information (via Abstract Meaning Representation graphs) to provide a context-informed ranker for RAG. G-RAG outperforms state-of-the-art approaches while having smaller computational footprint. Additionally, we assess the performance of PaLM 2 as a reranker and find it to significantly underperform G-RAG. This result emphasizes the importance of reranking for RAG even when using Large Language Models."
                    },
                    {
                        "title": "SubMix: Learning to Mix Graph Sampling Heuristics",
                        "abstract": "Sampling subgraphs for training Graph Neural Networks (GNNs) is receiving much attention from the GNN community. While a variety of meth-ods have been proposed, each method samples the graph according to its own heuristic. However, there has been little work in mixing these heuristics in an end-to-end trainable manner. In this work, we design a generative framework for subgraph sampling. Our method, SubMix, parame-terizes subgraph sampling as a convex combination of heuristics. We show that a continuous approximation of the discrete sampling process allows us to ef\ufb01ciently obtain analytical gradients for training the sampling parameters. Our experimental results illustrate the usefulness of learning graph sampling in three scenarios: (1) robust training of GNNs by automatically learning to discard noisy edge sources; (2) improving model performance by trainable and online edge subset selection; and (3) by integrating our framework into decoupled GNN models improves their performance on standard benchmarks."
                    },
                    {
                        "title": "Learning to Substitute Ingredients in Recipes",
                        "abstract": "Recipe personalization through ingredient substitution has the potential to help people meet their dietary needs and preferences, avoid potential allergens, and ease culinary exploration in everyone's kitchen. To address ingredient substitution, we build a benchmark, composed of a dataset of substitution pairs with standardized splits, evaluation metrics, and baselines. We further introduce Graph-based Ingredient Substitution Module (GISMo), a novel model that leverages the context of a recipe as well as generic ingredient relational information encoded within a graph to rank plausible substitutions. We show through comprehensive experimental validation that GISMo surpasses the best performing baseline by a large margin in terms of mean reciprocal rank. Finally, we highlight the benefits of GISMo by integrating it in an improved image-to-recipe generation pipeline, enabling recipe personalization through user intervention. Quantitative and qualitative results show the efficacy of our proposed system, paving the road towards truly personalized cooking and tasting experiences."
                    },
                    {
                        "title": "UGSL: A Unified Framework for Benchmarking Graph Structure Learning",
                        "abstract": "Graph neural networks (GNNs) demonstrate outstanding performance in a broad range of applications. While the majority of GNN applications assume that a graph structure is given, some recent methods substantially expanded the applicability of GNNs by showing that they may be effective even when no graph structure is explicitly provided. The GNN parameters and a graph structure are jointly learned. Previous studies adopt different experimentation setups, making it difficult to compare their merits. In this paper, we propose a benchmarking strategy for graph structure learning using a unified framework. Our framework, called Unified Graph Structure Learning (UGSL), reformulates existing models into a single model. We implement a wide range of existing models in our framework and conduct extensive analyses of the effectiveness of different components in the framework. Our results provide a clear and concise understanding of the different methods in this area as well as their strengths and weaknesses. The benchmark code is available at https://github.com/google-research/google-research/tree/master/ugsl."
                    },
                    {
                        "title": "19th International Workshop on Mining and Learning with Graphs (MLG)",
                        "abstract": "The 19th International Workshop on Mining and Learning with Graphs (MLG) is held in Long Beach, CA, USA and is co-located with the Tenth International Workshop on Deep Learning on Graphs (DLG) as part of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. This workshop is a forum for exchanging ideas and methods for mining and learning with graphs, developing new common understandings of the problems at hand, sharing of data sets where applicable, and leveraging existing knowledge from different disciplines. In doing so, we aim to better understand the overarching principles and the limitations of our current methods, and to inspire research on new algorithms and techniques for mining and learning with graphs. Topics of interest include, but are not limited to, graph mining, graph machine learning, statistical relational learning, social network analysis, and network science. The target audience spans researchers and practitioners across academia, government, and industry."
                    },
                    {
                        "title": "TpuGraphs: A Performance Prediction Dataset on Large Tensor Computational Graphs",
                        "abstract": "Precise hardware performance models play a crucial role in code optimizations. They can assist compilers in making heuristic decisions or aid autotuners in identifying the optimal configuration for a given program. For example, the autotuner for XLA, a machine learning compiler, discovered 10-20% speedup on state-of-the-art models serving substantial production traffic at Google. Although there exist a few datasets for program performance prediction, they target small sub-programs such as basic blocks or kernels. This paper introduces TpuGraphs, a performance prediction dataset on full tensor programs, represented as computational graphs, running on Tensor Processing Units (TPUs). Each graph in the dataset represents the main computation of a machine learning workload, e.g., a training epoch or an inference step. Each data sample contains a computational graph, a compilation configuration, and the execution time of the graph when compiled with the configuration. The graphs in the dataset are collected from open-source machine learning programs, featuring popular model architectures, e.g., ResNet, EfficientNet, Mask R-CNN, and Transformer. TpuGraphs provides 25x more graphs than the largest graph property prediction dataset (with comparable graph sizes), and 770x larger graphs on average compared to existing performance prediction datasets on machine learning programs. This graph-level prediction task on large graphs introduces new challenges in learning, ranging from scalability, training efficiency, to model quality."
                    },
                    {
                        "title": "Using Graph Algorithms to Pretrain Graph Completion Transformers",
                        "abstract": "Recent work on Graph Neural Networks has demonstrated that self-supervised pretraining can further enhance performance on downstream graph, link, and node classification tasks. However, the efficacy of pretraining tasks has not been fully investigated for downstream large knowledge graph completion tasks. Using a contextualized knowledge graph embedding approach, we investigate five different pretraining signals, constructed using several graph algorithms and no external data, as well as their combination. We leverage the versatility of our Transformer-based model to explore graph structure generation pretraining tasks (i.e. path and k-hop neighborhood generation), typically inapplicable to most graph embedding methods. We further propose a new path-finding algorithm guided by information gain and find that it is the best-performing pretraining task across three downstream knowledge graph completion datasets. While using our new path-finding algorithm as a pretraining signal provides 2-3% MRR improvements, we show that pretraining on all signals together gives the best knowledge graph completion results. In a multitask setting that combines all pretraining tasks, our method surpasses the latest and strong performing knowledge graph embedding methods on all metrics for FB15K-237, on MRR and Hit@1 for WN18RRand on MRR and hit@10 for JF17K (a knowledge hypergraph dataset)."
                    },
                    {
                        "title": "Knowledge Hypergraph Embedding Meets Relational Algebra",
                        "abstract": "Embedding-based methods for reasoning in knowledge hypergraphs learn a representation for each entity and relation. Current methods do not capture the procedural rules underlying the relations in the graph. We propose a simple embedding-based model called ReAlE that performs link prediction in knowledge hypergraphs (generalized knowledge graphs) and can represent high-level abstractions in terms of relational algebra operations. We show theoretically that ReAlE is fully expressive and provide proofs and empirical evidence that it can represent a large subset of the primitive relational algebra operations, namely renaming, projection, set union, selection, and set difference. We also verify experimentally that ReAlE outperforms state-of-the-art models in knowledge hypergraph completion, and in representing each of these primitive relational algebra operations. For the latter experiment, we generate a synthetic knowledge hypergraph, for which we design an algorithm based on the Erdos-R'enyi model for generating random graphs."
                    },
                    {
                        "title": "SLAPS: Self-Supervision Improves Structure Learning for Graph Neural Networks",
                        "abstract": "Graph neural networks (GNNs) work well when the graph structure is provided. However, this structure may not always be available in real-world applications. One solution to this problem is to infer a task-specific latent structure and then apply a GNN to the inferred graph. Unfortunately, the space of possible graph structures grows super-exponentially with the number of nodes and so the task-specific supervision may be insufficient for learning both the structure and the GNN parameters. In this work, we propose the Simultaneous Learning of Adjacency and GNN Parameters with Self-supervision, or SLAPS, a method that provides more supervision for inferring a graph structure through self-supervision. A comprehensive experimental study demonstrates that SLAPS scales to large graphs with hundreds of thousands of nodes and outperforms several models that have been proposed to learn a task-specific graph structure on established benchmarks."
                    },
                    {
                        "title": "Knowledge Hypergraphs: Prediction Beyond Binary Relations",
                        "abstract": "Knowledge graphs store facts using relations between two entities. In this work, we address the question of link prediction in knowledge hypergraphs where relations are defined on any number of entities. While techniques exist (such as reification) that convert non-binary relations into binary ones, we show that current embedding-based methods for knowledge graph completion do not work well out of the box for knowledge graphs obtained through these techniques. To overcome this, we introduce HSimplE and HypE, two embedding-based methods that work directly with knowledge hypergraphs. In both models, the prediction is a function of the relation embedding, the entity embeddings and their corresponding positions in the relation. We also develop public datasets, benchmarks and baselines for hypergraph prediction and show experimentally that the proposed models are more effective than the baselines."
                    }
                ]
            },
            "bb6f1fb1-b39d-4f28-b990-f362467d525b": {
                "pk": "bb6f1fb1-b39d-4f28-b990-f362467d525b",
                "name": "Jonathan Halcrow",
                "collaborators": [
                    "Bryan Perozzi",
                    "Anton Tsitsulin",
                    "Bahare Fatemi",
                    "Dustin Zelle",
                    "V. Mirrokni",
                    "J. Gibson",
                    "Mehran Kazemi",
                    "John Palowitch",
                    "Seyed Mehran Kazemi",
                    "Brandon Mayer"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Large Language Models",
                    "Temporal Reasoning",
                    "Graph Embedding"
                ],
                "publications": [
                    {
                        "title": "Let Your Graph Do the Talking: Encoding Structured Data for LLMs",
                        "abstract": "How can we best encode structured data into sequential form for use in large language models (LLMs)? In this work, we introduce a parameter-efficient method to explicitly represent structured data for LLMs. Our method, GraphToken, learns an encoding function to extend prompts with explicit structured information. Unlike other work which focuses on limited domains (e.g. knowledge graph representation), our work is the first effort focused on the general encoding of structured data to be used for various reasoning tasks. We show that explicitly representing the graph structure allows significant improvements to graph reasoning tasks. Specifically, we see across the board improvements - up to 73% points - on node, edge and, graph-level tasks from the GraphQA benchmark."
                    },
                    {
                        "title": "Test of Time: A Benchmark for Evaluating LLMs on Temporal Reasoning",
                        "abstract": "Large language models (LLMs) have showcased remarkable reasoning capabilities, yet they remain susceptible to errors, particularly in temporal reasoning tasks involving complex temporal logic. Existing research has explored LLM performance on temporal reasoning using diverse datasets and benchmarks. However, these studies often rely on real-world data that LLMs may have encountered during pre-training or employ anonymization techniques that can inadvertently introduce factual inconsistencies. In this work, we address these limitations by introducing novel synthetic datasets specifically designed to assess LLM temporal reasoning abilities in various scenarios. The diversity of question types across these datasets enables systematic investigation into the impact of the problem structure, size, question type, fact order, and other factors on LLM performance. Our findings provide valuable insights into the strengths and weaknesses of current LLMs in temporal reasoning tasks. To foster further research in this area, we are open-sourcing the datasets and evaluation framework used in our experiments: https://huggingface.co/datasets/baharef/ToT."
                    },
                    {
                        "title": "Understanding Transformer Reasoning Capabilities via Graph Algorithms",
                        "abstract": "Which transformer scaling regimes are able to perfectly solve different classes of algorithmic problems? While tremendous empirical advances have been attained by transformer-based neural networks, a theoretical understanding of their algorithmic reasoning capabilities in realistic parameter regimes is lacking. We investigate this question in terms of the network's depth, width, and number of extra tokens for algorithm execution. Our novel representational hierarchy separates 9 algorithmic reasoning problems into classes solvable by transformers in different realistic parameter scaling regimes. We prove that logarithmic depth is necessary and sufficient for tasks like graph connectivity, while single-layer transformers with small embedding dimensions can solve contextual retrieval tasks. We also support our theoretical analysis with ample empirical evidence using the GraphQA benchmark. These results show that transformers excel at many graph reasoning tasks, even outperforming specialized graph neural networks."
                    },
                    {
                        "title": "HUGE: Huge Unsupervised Graph Embeddings with TPUs",
                        "abstract": "Graphs are a representation of structured data that captures the relationships between sets of objects. With the ubiquity of available network data, there is increasing industrial and academic need to quickly analyze graphs with billions of nodes and trillions of edges. A common first step for network understanding is Graph Embedding, the process of creating a continuous representation of nodes in a graph. A continuous representation is often more amenable, especially at scale, for solving downstream machine learning tasks such as classification, link prediction, and clustering. A high-performance graph embedding architecture leveraging Tensor Processing Units (TPUs) with configurable amounts of high-bandwidth memory is presented that simplifies the graph embedding problem and can scale to graphs with billions of nodes and trillions of edges. We verify the embedding space quality on real and synthetic large-scale datasets."
                    },
                    {
                        "title": "UGSL: A Unified Framework for Benchmarking Graph Structure Learning",
                        "abstract": "Graph neural networks (GNNs) demonstrate outstanding performance in a broad range of applications. While the majority of GNN applications assume that a graph structure is given, some recent methods substantially expanded the applicability of GNNs by showing that they may be effective even when no graph structure is explicitly provided. The GNN parameters and a graph structure are jointly learned. Previous studies adopt different experimentation setups, making it difficult to compare their merits. In this paper, we propose a benchmarking strategy for graph structure learning using a unified framework. Our framework, called Unified Graph Structure Learning (UGSL), reformulates existing models into a single model. We implement a wide range of existing models in our framework and conduct extensive analyses of the effectiveness of different components in the framework. Our results provide a clear and concise understanding of the different methods in this area as well as their strengths and weaknesses. The benchmark code is available at https://github.com/google-research/google-research/tree/master/ugsl."
                    },
                    {
                        "title": "Stars: Tera-Scale Graph Building for Clustering and Graph Learning",
                        "abstract": "A fundamental procedure in the analysis of massive datasets is the construction of similarity graphs. Such graphs play a key role for many downstream tasks, including clustering, classification, graph learning, and nearest neighbor search. For these tasks, it is critical to build graphs which are sparse yet still representative of the underlying data. The benefits of sparsity are twofold: firstly, constructing dense graphs is infeasible in practice for large datasets, and secondly, the runtime of downstream tasks is directly influenced by the sparsity of the similarity graph. In this work, we present $\\textit{Stars}$: a highly scalable method for building extremely sparse graphs via two-hop spanners, which are graphs where similar points are connected by a path of length at most two. Stars can construct two-hop spanners with significantly fewer similarity comparisons, which are a major bottleneck for learning based models where comparisons are expensive to evaluate. Theoretically, we demonstrate that Stars builds a graph in nearly-linear time, where approximate nearest neighbors are contained within two-hop neighborhoods. In practice, we have deployed Stars for multiple data sets allowing for graph building at the $\\textit{Tera-Scale}$, i.e., for graphs with tens of trillions of edges. We evaluate the performance of Stars for clustering and graph learning, and demonstrate 10~1000-fold improvements in pairwise similarity comparisons compared to different baselines, and 2~10-fold improvement in running time without quality loss."
                    },
                    {
                        "title": "TF-GNN: Graph Neural Networks in TensorFlow",
                        "abstract": "TensorFlow-GNN (TF-GNN) is a scalable library for Graph Neural Networks in TensorFlow. It is designed from the bottom up to support the kinds of rich heterogeneous graph data that occurs in today's information ecosystems. In addition to enabling machine learning researchers and advanced developers, TF-GNN offers low-code solutions to empower the broader developer community in graph learning. Many production models at Google use TF-GNN, and it has been recently released as an open source project. In this paper we describe the TF-GNN data model, its Keras message passing API, and relevant capabilities such as graph sampling and distributed training."
                    },
                    {
                        "title": "Grale: Designing Networks for Graph Learning",
                        "abstract": "How can we find the right graph for semi-supervised learning? In real world applications, the choice of which edges to use for computation is the first step in any graph learning process. Interestingly, there are often many types of similarity available to choose as the edges between nodes, and the choice of edges can drastically affect the performance of downstream semi-supervised learning systems. However, despite the importance of graph design, most of the literature assumes that the graph is static. In this work, we present Grale, a scalable method we have developed to address the problem of graph design for graphs with billions of nodes. Grale operates by fusing together different measures of (potentially weak) similarity to create a graph which exhibits high task-specific homophily between its nodes. Grale is designed for running on large datasets. We have deployed Grale in more than 20 different industrial settings at Google, including datasets which have tens of billions of nodes, and hundreds of trillions of potential edges to score. By employing locality sensitive hashing techniques, we greatly reduce the number of pairs that need to be scored, allowing us to learn a task specific model and build the associated nearest neighbor graph for such datasets in hours, rather than the days or even weeks that might be required otherwise. We illustrate this through a case study where we examine the application of Grale to an abuse classification problem on YouTube with hundreds of million of items. In this application, we find that Grale detects a large number of malicious actors on top of hard-coded rules and content classifiers, increasing the total recall by 89% over those approaches alone."
                    },
                    {
                        "title": "Geometry of turbulence: a stroll through 61,506 dimensions",
                        "abstract": "In the world of moderate Reynolds number, everyday turbulence of fluids flowing across planes and down pipes a velvet revolution is taking place. Experiments are almost as detailed as the numerical simulations, DNS is yielding exact numerical solutions that one dared not dream about a decade ago, and dynamical systems visualization of turbulent fluid\u2019s state space geometry is unexpectedly elegant. We shall take you on a tour of this newly breached, hitherto inaccessible territory. Mastery of fluid mechanics is no prerequisite, and perhaps a hindrance: the talk is aimed at anyone who had ever wondered why if no cloud is ever seen twice we know a cloud when we see one? And how do we turn that into mathematics?"
                    },
                    {
                        "title": "Heteroclinic connections in plane Couette flow",
                        "abstract": "Plane Couette flow transitions to turbulence at Re \u2248 325 even though the laminar solution with a linear profile is linearly stable for all Re (Reynolds number). One starting point for understanding this subcritical transition is the existence of invariant sets in the state space of the Navier\u2013Stokes equation, such as upper and lower branch equilibria and periodic and relative periodic solutions, that are distinct from the laminar solution. This article reports several heteroclinic connections between such objects and briefly describes a numerical method for locating heteroclinic connections. We show that the nature of streaks and streamwise rolls can change significantly along a heteroclinic connection."
                    },
                    {
                        "title": "Visualizing the geometry of state space in plane Couette flow",
                        "abstract": "Motivated by recent experimental and numerical studies of coherent structures in wall-bounded shear flows, we initiate a systematic exploration of the hierarchy of unstable invariant solutions of the Navier\u2013Stokes equations. We construct a dynamical 105-dimensional state-space representation of plane Couette flow at Reynolds number Re = 400 in a small periodic cell and offer a new method of visualizing invariant manifolds embedded in such high dimensions. We compute a new equilibrium solution of plane Couette flow and the leading eigenvalues and eigenfunctions of known equilibria at this Re and cell size. What emerges from global continuations of their unstable manifolds is a surprisingly elegant dynamical-systems visualization of moderate-Re turbulence. The invariant manifolds partially tessellate the region of state space explored by transiently turbulent dynamics with a rigid web of symmetry-induced heteroclinic connections."
                    },
                    {
                        "title": "Implementation of a Pressure Poisson Equation method for Plane Couette Flow",
                        "abstract": "This paper discusses my work through December 2004 to implement Johnston and Liu\u2019s Pressure Poisson Equation (PPE) method[1] for Plane Couette flow. Also, I discuss some results and plans for implementing a periodic orbit search. The advantage of this method is that it decouples the computation of velocity and pressure at each time step by treating the pressure term explicitly in time. The project is my first step in designing a program to search for periodic orbits in Plane Couette flow, using variational methods."
                    }
                ]
            },
            "25805df8-dd60-4372-8092-6d711e5e54b8": {
                "pk": "25805df8-dd60-4372-8092-6d711e5e54b8",
                "name": "Bryan Perozzi",
                "collaborators": [
                    "Anton Tsitsulin",
                    "Bahare Fatemi",
                    "Jonathan J. Halcrow",
                    "Sami Abu-El-Haija",
                    "Dustin Zelle",
                    "John Palowitch",
                    "Mehran Kazemi",
                    "Seyed Mehran Kazemi",
                    "Kaidi Cao",
                    "P. Phothilimthana"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Large Language Model",
                    "Structured Data",
                    "Transfer Learning"
                ],
                "publications": [
                    {
                        "title": "Let Your Graph Do the Talking: Encoding Structured Data for LLMs",
                        "abstract": "How can we best encode structured data into sequential form for use in large language models (LLMs)? In this work, we introduce a parameter-efficient method to explicitly represent structured data for LLMs. Our method, GraphToken, learns an encoding function to extend prompts with explicit structured information. Unlike other work which focuses on limited domains (e.g. knowledge graph representation), our work is the first effort focused on the general encoding of structured data to be used for various reasoning tasks. We show that explicitly representing the graph structure allows significant improvements to graph reasoning tasks. Specifically, we see across the board improvements - up to 73% points - on node, edge and, graph-level tasks from the GraphQA benchmark."
                    },
                    {
                        "title": "Text-space Graph Foundation Models: Comprehensive Benchmarks and New Insights",
                        "abstract": "Given the ubiquity of graph data and its applications in diverse domains, building a Graph Foundation Model (GFM) that can work well across different graphs and tasks with a unified backbone has recently garnered significant interests. A major obstacle to achieving this goal stems from the fact that graphs from different domains often exhibit diverse node features. Inspired by multi-modal models that align different modalities with natural language, the text has recently been adopted to provide a unified feature space for diverse graphs. Despite the great potential of these text-space GFMs, current research in this field is hampered by two problems. First, the absence of a comprehensive benchmark with unified problem settings hinders a clear understanding of the comparative effectiveness and practical value of different text-space GFMs. Second, there is a lack of sufficient datasets to thoroughly explore the methods' full potential and verify their effectiveness across diverse settings. To address these issues, we conduct a comprehensive benchmark providing novel text-space datasets and comprehensive evaluation under unified problem settings. Empirical results provide new insights and inspire future research directions. Our code and data are publicly available from \\url{https://github.com/CurryTang/TSGFM}."
                    },
                    {
                        "title": "Test of Time: A Benchmark for Evaluating LLMs on Temporal Reasoning",
                        "abstract": "Large language models (LLMs) have showcased remarkable reasoning capabilities, yet they remain susceptible to errors, particularly in temporal reasoning tasks involving complex temporal logic. Existing research has explored LLM performance on temporal reasoning using diverse datasets and benchmarks. However, these studies often rely on real-world data that LLMs may have encountered during pre-training or employ anonymization techniques that can inadvertently introduce factual inconsistencies. In this work, we address these limitations by introducing novel synthetic datasets specifically designed to assess LLM temporal reasoning abilities in various scenarios. The diversity of question types across these datasets enables systematic investigation into the impact of the problem structure, size, question type, fact order, and other factors on LLM performance. Our findings provide valuable insights into the strengths and weaknesses of current LLMs in temporal reasoning tasks. To foster further research in this area, we are open-sourcing the datasets and evaluation framework used in our experiments: https://huggingface.co/datasets/baharef/ToT."
                    },
                    {
                        "title": "Understanding Transformer Reasoning Capabilities via Graph Algorithms",
                        "abstract": "Which transformer scaling regimes are able to perfectly solve different classes of algorithmic problems? While tremendous empirical advances have been attained by transformer-based neural networks, a theoretical understanding of their algorithmic reasoning capabilities in realistic parameter regimes is lacking. We investigate this question in terms of the network's depth, width, and number of extra tokens for algorithm execution. Our novel representational hierarchy separates 9 algorithmic reasoning problems into classes solvable by transformers in different realistic parameter scaling regimes. We prove that logarithmic depth is necessary and sufficient for tasks like graph connectivity, while single-layer transformers with small embedding dimensions can solve contextual retrieval tasks. We also support our theoretical analysis with ample empirical evidence using the GraphQA benchmark. These results show that transformers excel at many graph reasoning tasks, even outperforming specialized graph neural networks."
                    },
                    {
                        "title": "Don't Forget to Connect! Improving RAG with Graph-based Reranking",
                        "abstract": "Retrieval Augmented Generation (RAG) has greatly improved the performance of Large Language Model (LLM) responses by grounding generation with context from existing documents. These systems work well when documents are clearly relevant to a question context. But what about when a document has partial information, or less obvious connections to the context? And how should we reason about connections between documents? In this work, we seek to answer these two core questions about RAG generation. We introduce G-RAG, a reranker based on graph neural networks (GNNs) between the retriever and reader in RAG. Our method combines both connections between documents and semantic information (via Abstract Meaning Representation graphs) to provide a context-informed ranker for RAG. G-RAG outperforms state-of-the-art approaches while having smaller computational footprint. Additionally, we assess the performance of PaLM 2 as a reranker and find it to significantly underperform G-RAG. This result emphasizes the importance of reranking for RAG even when using Large Language Models."
                    },
                    {
                        "title": "Examining the Effects of Degree Distribution and Homophily in Graph Learning Models",
                        "abstract": "Despite a surge in interest in GNN development, homogeneity in benchmarking datasets still presents a fundamental issue to GNN research. GraphWorld is a recent solution which uses the Stochastic Block Model (SBM) to generate diverse populations of synthetic graphs for benchmarking any GNN task. Despite its success, the SBM imposed fundamental limitations on the kinds of graph structure GraphWorld could create. In this work we examine how two additional synthetic graph generators can improve GraphWorld's evaluation; LFR, a well-established model in the graph clustering literature and CABAM, a recent adaptation of the Barabasi-Albert model tailored for GNN benchmarking. By integrating these generators, we significantly expand the coverage of graph space within the GraphWorld framework while preserving key graph properties observed in real-world networks. To demonstrate their effectiveness, we generate 300,000 graphs to benchmark 11 GNN models on a node classification task. We find GNN performance variations in response to homophily, degree distribution and feature signal. Based on these findings, we classify models by their sensitivity to the new generators under these properties. Additionally, we release the extensions made to GraphWorld on the GitHub repository, offering further evaluation of GNN performance on new graphs."
                    },
                    {
                        "title": "Learning Large Graph Property Prediction via Graph Segment Training",
                        "abstract": "Learning to predict properties of large graphs is challenging because each prediction requires the knowledge of an entire graph, while the amount of memory available during training is bounded. Here we propose Graph Segment Training (GST), a general framework that utilizes a divide-and-conquer approach to allow learning large graph property prediction with a constant memory footprint. GST first divides a large graph into segments and then backpropagates through only a few segments sampled per training iteration. We refine the GST paradigm by introducing a historical embedding table to efficiently obtain embeddings for segments not sampled for backpropagation. To mitigate the staleness of historical embeddings, we design two novel techniques. First, we finetune the prediction head to fix the input distribution shift. Second, we introduce Stale Embedding Dropout to drop some stale embeddings during training to reduce bias. We evaluate our complete method GST-EFD (with all the techniques together) on two large graph property prediction benchmarks: MalNet and TpuGraphs. Our experiments show that GST-EFD is both memory-efficient and fast, while offering a slight boost on test accuracy over a typical full graph training regime."
                    },
                    {
                        "title": "Graph Neural Networks in TensorFlow",
                        "abstract": "Graphs are general data structures that can represent information from a variety of domains (social, biomedical, online transactions, and many more). Graph Neural Networks (GNNs) are quickly becoming the de-facto Machine Learning models for learning from Graph data and hereby infer missing information, such as, predicting labels of nodes or imputing missing edges. In this tutorial we'll cover essential applications of Graph Machine Learning using TensorFlow GNN1[12], a Python framework that extends TensorFlow [1] with Graph Neural Networks (GNNs): models that leverage graph-structured data. TF-GNN is motivated and informed by years of applying graph representation learning to practical problems at Google [2-8, 10, 11, 13-19, 22-29]. In particular, TF-GNN focuses on the representation of heterogeneous graph data and supports the explicit modeling of an arbitrary number of relationship (edge) types between an arbitrary number of entity (node) types. These relationships can be used in combination with other TensorFlow components, e.g., a TF-GNN model might connect representations from a language model to those of a vision model and fine-tune these features for a node classification task. Many teams at Google run TF-GNN models in production. We believe this to be a direct consequence of TF-GNN's multi-layered API which is designed for accessibility to developers (regardless of their prior experience with machine learning)."
                    },
                    {
                        "title": "SubMix: Learning to Mix Graph Sampling Heuristics",
                        "abstract": "Sampling subgraphs for training Graph Neural Networks (GNNs) is receiving much attention from the GNN community. While a variety of meth-ods have been proposed, each method samples the graph according to its own heuristic. However, there has been little work in mixing these heuristics in an end-to-end trainable manner. In this work, we design a generative framework for subgraph sampling. Our method, SubMix, parame-terizes subgraph sampling as a convex combination of heuristics. We show that a continuous approximation of the discrete sampling process allows us to ef\ufb01ciently obtain analytical gradients for training the sampling parameters. Our experimental results illustrate the usefulness of learning graph sampling in three scenarios: (1) robust training of GNNs by automatically learning to discard noisy edge sources; (2) improving model performance by trainable and online edge subset selection; and (3) by integrating our framework into decoupled GNN models improves their performance on standard benchmarks."
                    },
                    {
                        "title": "Unsupervised Embedding Quality Evaluation",
                        "abstract": "Unsupervised learning has recently significantly gained in popularity, especially with deep learning-based approaches. Despite numerous successes and approaching supervised-level performance on a variety of academic benchmarks, it is still hard to train and evaluate SSL models in practice due to the unsupervised nature of the problem. Even with networks trained in a supervised fashion, it is often unclear whether they will perform well when transferred to another domain. Past works are generally limited to assessing the amount of information contained in embeddings, which is most relevant for self-supervised learning of deep neural networks. This works chooses to follow a different approach: can we quantify how easy it is to linearly separate the data in a stable way? We survey the literature and uncover three methods that could be potentially used for evaluating quality of representations. We also introduce one novel method based on recent advances in understanding the high-dimensional geometric structure of self-supervised learning. We conduct extensive experiments and study the properties of these metrics and ones introduced in the previous work. Our results suggest that while there is no free lunch, there are metrics that can robustly estimate embedding quality in an unsupervised way."
                    },
                    {
                        "title": "The Graph Lottery Ticket Hypothesis: Finding Sparse, Informative Graph Structure",
                        "abstract": "Graph learning methods help utilize implicit relationships among data items, thereby reducing training label requirements and improving task performance. However, determining the optimal graph structure for a particular learning task remains a challenging research problem. In this work, we introduce the Graph Lottery Ticket (GLT) Hypothesis - that there is an extremely sparse backbone for every graph, and that graph learning algorithms attain comparable performance when trained on that subgraph as on the full graph. We identify and systematically study 8 key metrics of interest that directly influence the performance of graph learning algorithms. Subsequently, we define the notion of a\"winning ticket\"for graph structure - an extremely sparse subset of edges that can deliver a robust approximation of the entire graph's performance. We propose a straightforward and efficient algorithm for finding these GLTs in arbitrary graphs. Empirically, we observe that performance of different graph learning algorithms can be matched or even exceeded on graphs with the average degree as low as 5."
                    },
                    {
                        "title": "HUGE: Huge Unsupervised Graph Embeddings with TPUs",
                        "abstract": "Graphs are a representation of structured data that captures the relationships between sets of objects. With the ubiquity of available network data, there is increasing industrial and academic need to quickly analyze graphs with billions of nodes and trillions of edges. A common first step for network understanding is Graph Embedding, the process of creating a continuous representation of nodes in a graph. A continuous representation is often more amenable, especially at scale, for solving downstream machine learning tasks such as classification, link prediction, and clustering. A high-performance graph embedding architecture leveraging Tensor Processing Units (TPUs) with configurable amounts of high-bandwidth memory is presented that simplifies the graph embedding problem and can scale to graphs with billions of nodes and trillions of edges. We verify the embedding space quality on real and synthetic large-scale datasets."
                    },
                    {
                        "title": "UGSL: A Unified Framework for Benchmarking Graph Structure Learning",
                        "abstract": "Graph neural networks (GNNs) demonstrate outstanding performance in a broad range of applications. While the majority of GNN applications assume that a graph structure is given, some recent methods substantially expanded the applicability of GNNs by showing that they may be effective even when no graph structure is explicitly provided. The GNN parameters and a graph structure are jointly learned. Previous studies adopt different experimentation setups, making it difficult to compare their merits. In this paper, we propose a benchmarking strategy for graph structure learning using a unified framework. Our framework, called Unified Graph Structure Learning (UGSL), reformulates existing models into a single model. We implement a wide range of existing models in our framework and conduct extensive analyses of the effectiveness of different components in the framework. Our results provide a clear and concise understanding of the different methods in this area as well as their strengths and weaknesses. The benchmark code is available at https://github.com/google-research/google-research/tree/master/ugsl."
                    },
                    {
                        "title": "TpuGraphs: A Performance Prediction Dataset on Large Tensor Computational Graphs",
                        "abstract": "Precise hardware performance models play a crucial role in code optimizations. They can assist compilers in making heuristic decisions or aid autotuners in identifying the optimal configuration for a given program. For example, the autotuner for XLA, a machine learning compiler, discovered 10-20% speedup on state-of-the-art models serving substantial production traffic at Google. Although there exist a few datasets for program performance prediction, they target small sub-programs such as basic blocks or kernels. This paper introduces TpuGraphs, a performance prediction dataset on full tensor programs, represented as computational graphs, running on Tensor Processing Units (TPUs). Each graph in the dataset represents the main computation of a machine learning workload, e.g., a training epoch or an inference step. Each data sample contains a computational graph, a compilation configuration, and the execution time of the graph when compiled with the configuration. The graphs in the dataset are collected from open-source machine learning programs, featuring popular model architectures, e.g., ResNet, EfficientNet, Mask R-CNN, and Transformer. TpuGraphs provides 25x more graphs than the largest graph property prediction dataset (with comparable graph sizes), and 770x larger graphs on average compared to existing performance prediction datasets on machine learning programs. This graph-level prediction task on large graphs introduces new challenges in learning, ranging from scalability, training efficiency, to model quality."
                    },
                    {
                        "title": "Explaining and Adapting Graph Conditional Shift",
                        "abstract": "Graph Neural Networks (GNNs) have shown remarkable performance on graph-structured data. However, recent empirical studies suggest that GNNs are very susceptible to distribution shift. There is still significant ambiguity about why graph-based models seem more vulnerable to these shifts. In this work we provide a thorough theoretical analysis on it by quantifying the magnitude of conditional shift between the input features and the output label. Our findings show that both graph heterophily and model architecture exacerbate conditional shifts, leading to performance degradation. To address this, we propose an approach that involves estimating and minimizing the conditional shift for unsupervised domain adaptation on graphs. In our controlled synthetic experiments, our algorithm demonstrates robustness towards distribution shift, resulting in up to 10% absolute ROC AUC improvement versus the second-best algorithm. Furthermore, comprehensive experiments on both node classification and graph classification show its robust performance under various distribution shifts."
                    },
                    {
                        "title": "Graph Generative Model for Benchmarking Graph Neural Networks",
                        "abstract": "As the field of Graph Neural Networks (GNN) continues to grow, it experiences a corresponding increase in the need for large, real-world datasets to train and test new GNN models on challenging, realistic problems. Unfortunately, such graph datasets are often generated from online, highly privacy-restricted ecosystems, which makes research and development on these datasets hard, if not impossible. This greatly reduces the amount of benchmark graphs available to researchers, causing the field to rely only on a handful of publicly-available datasets. To address this problem, we introduce a novel graph generative model, Computation Graph Transformer (CGT) that learns and reproduces the distribution of real-world graphs in a privacy-controlled way. More specifically, CGT (1) generates effective benchmark graphs on which GNNs show similar task performance as on the source graphs, (2) scales to process large-scale graphs, (3) incorporates off-the-shelf privacy modules to guarantee end-user privacy of the generated graph. Extensive experiments across a vast body of graph generative models show that only our model can successfully generate privacy-controlled, synthetic substitutes of large-scale real-world graphs that can be effectively used to benchmark GNN models."
                    },
                    {
                        "title": "Zero-shot Transfer Learning within a Heterogeneous Graph via Knowledge Transfer Networks",
                        "abstract": "Data continuously emitted from industrial ecosystems such as social or e-commerce platforms are commonly represented as heterogeneous graphs (HG) composed of multiple node/edge types. State-of-the-art graph learning methods for HGs known as heterogeneous graph neural networks (HGNNs) are applied to learn deep context-informed node representations. However, many HG datasets from industrial applications suffer from label imbalance between node types. As there is no direct way to learn using labels rooted at different node types, HGNNs have been applied to only a few node types with abundant labels. We propose a zero-shot transfer learning module for HGNNs called a Knowledge Transfer Network (KTN) that transfers knowledge from label-abundant node types to zero-labeled node types through rich relational information given in the HG. KTN is derived from the theoretical relationship, which we introduce in this work, between distinct feature extractors for each node type given in an HGNN model. KTN improves performance of 6 different types of HGNN models by up to 960% for inference on zero-labeled node types and outperforms state-of-the-art transfer learning baselines by up to 73% across 18 different transfer learning tasks on HGs."
                    },
                    {
                        "title": "On Classification Thresholds for Graph Attention with Edge Features",
                        "abstract": "The recent years we have seen the rise of graph neural networks for prediction tasks on graphs. One of the dominant architectures is graph attention due to its ability to make predictions using weighted edge features and not only node features. In this paper we analyze, theoretically and empirically, graph attention networks and their ability of correctly labelling nodes in a classic classification task. More specifically, we study the performance of graph attention on the classic contextual stochastic block model (CSBM). In CSBM the nodes and edge features are obtained from a mixture of Gaussians and the edges from a stochastic block model. We consider a general graph attention mechanism that takes random edge features as input to determine the attention coefficients. We study two cases, in the first one, when the edge features are noisy, we prove that the majority of the attention coefficients are up to a constant uniform. This allows us to prove that graph attention with edge features is not better than simple graph convolution for achieving perfect node classification. Second, we prove that when the edge features are clean graph attention can distinguish intra- from inter-edges and this makes graph attention better than classic graph convolution."
                    }
                ]
            }
        }
    },
    "2106.04255": {
        "paper_data": {
            "title": "Nonparametric Regression for 3D Point Cloud Learning",
            "url": "http://arxiv.org/abs/2106.04255v4",
            "arxiv_id": "2106.04255",
            "authors": [
                "Xinyi Li",
                "Shan Yu",
                "Yueying Wang",
                "Guannan Wang",
                "Ming-Jun Lai",
                "Li Wang"
            ],
            "abstract": "Over the past two decades, we have seen an exponentially increased amount of point clouds collected with irregular shapes in various areas. Motivated by the importance of solid modeling for point clouds, we develop a novel and efficient smoothing tool based on multivariate splines over the tetrahedral partitions to extract the underlying signal and build up a 3D solid model from the point cloud. The proposed smoothing method can denoise or deblur the point cloud effectively and provide a multi-resolution reconstruction of the actual signal. In addition, it can handle sparse and irregularly distributed point clouds and recover the underlying trajectory. The proposed smoothing and interpolation method also provides a natural way of numerosity data reduction. Furthermore, we establish the theoretical guarantees of the proposed method. Specifically, we derive the convergence rate and asymptotic normality of the proposed estimator and illustrate that the convergence rate achieves the optimal nonparametric convergence rate. Through extensive simulation studies and a real data example, we demonstrate the superiority of the proposed method over traditional smoothing methods in terms of estimation accuracy and efficiency of data reduction.",
            "introduction": " Introduction Recent advances in computer and information technology have dramatically boosted the avail- ability of three-dimensional (3D) point clouds in many \felds, such as geography, environmental science, computer graphics, engineering, economics, and medical imaging. These point clouds are usually collections of enormous measurements in space de\fned by a given coordinates sys- tem. Sometimes, these point clouds are rendered and inspected directly; examples can be found in Rusinkiewicz and Levoy (2000). More often, the point clouds are converted to polygon mesh or triangle mesh models, e.g., 3D reconstruction is widely used for automated driving vehicles (Lee et al., 2016; Shao, 2021). Many e\u000borts have been devoted to investigating the converting technique, and a survey of surface reconstruction from point clouds can be found in Berger Address forcorrespondence: Li Wang, Department of Statistics, George Mason University, Fairfax, VA 22030, USA. Email: lwang41@gmu.eduarXiv:2106.04255v4  [stat.CO]  18 Feb 2023et al. (2017). These 3D data typically contain more information than the shape of an object, as they represent some meaningful values in real life. For example, healthcare professionals can use brain activity levels contained in 3D neuroimages to access new angles, resolutions, and details to better understand tissue. Another example is using cloud-based 3D scans to explore the location and access the grade and type of ore in the mining industry. Despite the potential of these applications, there has been limited research on identifying underlying signals from 3D solid objects, as there are many challenges in achieving this goal. First of all, as shown in the point cloud inputs in Figure 1.2, the shape of the point clouds is usually irregular. To extract the underlying signal over the irregular-shaped 3D solid object, many conventional results on the consistency of spatial regression with partial di\u000berential equation regularization,\" Statistica Sinica , 32, 209{238. Awanou, G., Lai, M. J., and Wenston, P. (2005), \\The multivariate spline method for scattered data \ftting and numerical solutions of partial di\u000berential equations,\" Wavelets and splines: Athens 2005 , 24{74. Berger, M., Tagliasacchi, A., Seversky, L. M., Alliez, P., Guennebaud, G., Levine, J. A., Sharf, A., and Silva, C. T. (2017), \\A survey of surface reconstruction from point clouds,\" Computer Graphics Forum , 36, 301{329. Bosq, D. (1998), Nonparametric statistics for stochastic processes , Springer-Verlag, New York. Chung, M. K., Wang, Y., and Wu, G. (2018), \\Discrete heat kernel smoothing in ir- regular image domains,\" in 2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) , pp. 5101{5104. DeVore, R. A. and Lorentz, G. G. (1993), Constructive approximation , vol. 303, Berlin: Springer-Verlag. Duchon, J. (1976), \\Fonctions-spline \u0012 a \u0013 energie invariante par rotation,\" Tech. Rep. 27, Universit\u0013 e de Grenoble. Eilers, P. H. and Marx, B. D. (1996), \\Flexible smoothing with B-splines and penalties,\" Statistical science , 11, 89{121. Fang, Q. and Boas, D. A. (2009), \\Tetrahedral mesh generation from volumetric binary and grayscale images,\" in 2009 IEEE International Symposium on Biomedical Imaging: From Nano to Macro , pp. 1142{1145. 60Goldsmith, J., Huang, L., and Crainiceanu, C. M. (2014), \\Smooth scalar-on-image re- gression via spatial Bayesian variable selection,\" Journal of Computational and Graph- ical Statistics , 23, 46{64. Green, P. J. and Silverman, B. W. (1994), Nonparametric regression and generalized linear models: a roughness penalty approach , Chapman and Hall, London. Guo, J., Duan, X., and Shum, C. (2010), \\Non-isotropic Gaussian smoothing and leakage reduction for",
            "references": [
                {
                    "title": "Testing Object Detection for Autonomous Driving Systems via 3D Reconstruction",
                    "abstract": "Object detection is to identify objects from images. In autonomous driving systems, object detection serves as an intermediate module, which is used as the input of autonomous decisions for vehicles. That is, the accuracy of autonomous decisions relies on the object detection. The state-of-the-art object detection modules are designed based on Deep Neural Networks (DNNs). It is difficult to employ white-box testing on DNNs since the output of a single neuron is inexplicable. Existing work conducted metamorphic testing for object detection via image synthesis: the detected object in the original image should be detected in the new synthetic image. However, a synthetic image may not look real from humans' perspective. Even the object detection module fails in detecting such synthetic image, the failure may not reflect the ability of object detection. In this paper, we propose an automatic approach to testing object detection via 3D reconstruction of vehicles in real photos. The 3D reconstruction is developed via vanishing point estimation in photos and heuristic based image insertion. Our approach adds new objects to blank spaces in photos to synthesize images. For example, a new vehicle can be added to a photo of a road and vehicles. In this approach, the output synthetic images are expected to be more natural-looking than randomly synthesizing images. The experiment is conducting on 500 driving photos from the Apollo autonomous driving dataset."
                },
                {
                    "title": "Generalized Spatially Varying Coefficient Models",
                    "abstract": "Abstract In this article, we introduce a new class of nonparametric regression models, called generalized spatially varying coefficient models (GSVCMs), for data distributed over complex domains. For model estimation, we propose a nonparametric quasi-likelihood approach using the bivariate penalized spline approximation technique. We show that our estimation procedure is able to handle irregularly-shaped spatial domains with complex boundaries. Under some regularity conditions, the estimator for the coefficient function is proved to be consistent in the L 2 sense and its convergence rate is established. We develop a numerically stable algorithm using penalized iteratively reweighted least squares method to estimate the coefficient functions in GSVCMs. To gain efficiency in the computation for large-scale data, we further propose a QR decomposition-based algorithm, which requires only sub-blocks of the design matrix to be computed at a time, so that it allows efficient estimation of GSVCMs for large datasets with modest computer hardware. The finite sample performance of the GSVCM and its estimation method is examined by simulations studies. The proposed method is also illustrated by an analysis of the crash data in Florida. Supplementary materials for this article are available online."
                },
                {
                    "title": "Estimation and Inference for Generalized Geoadditive Models",
                    "abstract": "Abstract In many application areas, data are collected on a count or binary response with spatial covariate information. In this article, we introduce a new class of generalized geoadditive models (GGAMs) for spatial data distributed over complex domains. Through a link function, the proposed GGAM assumes that the mean of the discrete response variable depends on additive univariate functions of explanatory variables and a bivariate function to adjust for the spatial effect. We propose a two-stage approach for estimating and making inferences of the components in the GGAM. In the first stage, the univariate components and the geographical component in the model are approximated via univariate polynomial splines and bivariate penalized splines over triangulation, respectively. In the second stage, local polynomial smoothing is applied to the cleaned univariate data to average out the variation of the first-stage estimators. We investigate the consistency of the proposed estimators and the asymptotic normality of the univariate components. We also establish the simultaneous confidence band for each of the univariate components. The performance of the proposed method is evaluated by two simulation studies. We apply the proposed method to analyze the crash counts data in the Tampa-St. Petersburg urbanized area in Florida. Supplementary materials for this article are available online."
                },
                {
                    "title": "Estimation of additive frontier functions with shape constraints",
                    "abstract": "ABSTRACT Production frontier is an important concept in modern economics and has been widely used to measure production efficiency. Existing nonparametric frontier models often only allow one or low-dimensional input variables due to \u2018curse-of-dimensionality\u2019. In this paper we propose a flexible additive frontier model which quantifies the effects of multiple input variables on the maximum output. In addition, we consider the estimation of the nonparametric frontier functions with shape restrictions. Economic theory often imposes shape constraints on production frontier, such as, monotonicity and concavity. A two-step constrained polynomial spline method is proposed to give smooth estimates that automatically satisfy such shape constraints. The proposed method is not only easy to compute, but also more robust to outliers. In theory, we established uniform consistency of the proposed method. We illustrate the proposed method by both simulation studies and an application to the Norwegian farm data. The numerical studies suggest that the proposed method has superior performance by incorporating shape constraints."
                },
                {
                    "title": "Consistent Blind Image Deblurring Using Jump-Preserving Extrapolation",
                    "abstract": "Abstract Blind image deblurring (BID) aims to estimate the true image from blurred observations when the blurring mechanism described by a point spread function (PSF) cannot be completely specified beforehand. This is a challenging ill-posed problem because there are infinitely many solutions even in cases when the PSF is location invariant. In reality, however, blurring often cannot be accurately described by a location-invariant PSF and the ill-posedness of the BID problem would be further exacerbated. In this article, we propose a novel BID methodology, which allows the PSF to change over locations. The true image is estimated by our jump-preserving extrapolation (JPEX) technique. The proposed method classifies the pixels in a blurred image into two categories: blurry pixels and sharp pixels. The key idea of JPEX is to remove blur involved in blurry pixels using nearby sharp pixels. Theoretical justifications show that the proposed method is statistically consistent. Numerical experiments with both simulated and real images demonstrate that it works well in applications. The comparison with state-of-the-art methods favors the proposed approach as well."
                },
                {
                    "title": "Discrete Heat Kernel Smoothing in Irregular Image Domains",
                    "abstract": "We present the discrete version of heat kernel smoothing on graph data structure. The method is used to smooth data in an irregularly shaped domains in 3D images. New statistical properties of heat kernel smoothing are derived. As an application, we show how to filter out noisy data in the lung blood vessel trees obtained from computed tomography. The method can be further used in representing the complex vessel trees parametrically as a linear combination of basis functions and extracting the skeleton representation of the trees."
                },
                {
                    "title": "blockCV: an R package for generating spatially or environmentally separated folds for k-fold cross-validation of species distribution models",
                    "abstract": "When applied to structured data, conventional random cross-validation techniques can lead to underestimation of prediction error, and may result in inappropriate model selection. We present the R package blockCV, a new toolbox for cross-validation of species distribution modelling. The package can generate spatially or environmentally separated folds. It includes tools to measure spatial autocorrelation ranges in candidate covariates, providing the user with insights into the spatial structure in these data. It also offers interactive graphical capabilities for creating spatial blocks and exploring data folds. Package blockCV enables modellers to more easily implement a range of evaluation approaches. It will help the modelling community learn more about the impacts of evaluation approaches on our understanding of predictive performance of species distribution models."
                },
                {
                    "title": "Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure",
                    "abstract": "Ecological data often show temporal, spatial, hierarchical (random effects), or phylogenetic structure. Modern statistical approaches are increasingly accounting for such dependencies. However, when performing cross-validation, these structures are regularly ignored, resulting in serious underestimation of predictive error. One cause for the poor performance of uncorrected (random) cross-validation, noted often by modellers, are dependence structures in the data that persist as dependence structures in model residuals, violating the assumption of independence. Even more concerning, because often overlooked, is that structured data also provides ample opportunity for overfitting with non-causal predictors. This problem can persist even if remedies such as autoregressive models, generalized least squares, or mixed models are used. Block cross-validation, where data are split strategically rather than randomly, can address these issues. However, the blocking strategy must be carefully considered. Blocking in space, time, random effects or phylogenetic distance, while accounting for dependencies in the data, may also unwittingly induce extrapolations by restricting the ranges or combinations of predictor variables available for model training, thus overestimating interpolation errors. On the other hand, deliberate blocking in predictor space may also improve error estimates when extrapolation is the modelling goal. Here, we review the ecological literature on non-random and blocked cross-validation approaches. We also provide a series of simulations and case studies, in which we show that, for all instances tested, block cross-validation is nearly universally more appropriate than random cross-validation if the goal is predicting to new data or predictor space, or for selecting causal predictors. We recommend that block cross-validation be used wherever dependence structures exist in a dataset, even if no correlation structure is visible in the fitted model residuals, or if the fitted models account for such correlations."
                },
                {
                    "title": "A Survey of Surface Reconstruction from Point Clouds",
                    "abstract": "The area of surface reconstruction has seen substantial progress in the past two decades. The traditional problem addressed by surface reconstruction is to recover the digital representation of a physical shape that has been scanned, where the scanned data contain a wide variety of defects. While much of the earlier work has been focused on reconstructing a piece\u2010wise smooth representation of the original shape, recent work has taken on more specialized priors to address significantly challenging data imperfections, where the reconstruction can take on different representations\u2014not necessarily the explicit geometry. We survey the field of surface reconstruction, and provide a categorization with respect to priors, data imperfections and reconstruction output. By considering a holistic view of surface reconstruction, we show a detailed characterization of the field, highlight similarities between diverse reconstruction techniques and provide directions for future work in surface reconstruction."
                },
                {
                    "title": "3D Reconstruction using a sparse laser scanner and a single camera for outdoor autonomous vehicle",
                    "abstract": "This paper presents a 3D scene reconstruction method for autonomous vehicle driving in a wide range of outdoor environments. Autonomous vehicles, most of which currently employ laser and image sensors, are required to have systems for object detection, obstacle avoidance, navigation etc. The one of the most important pieces of information for these systems is an accurate dense 3D depth map. However, range data is much sparser than image data, thus the challenging problem is to reconstruct a dense depth map using sparse range and image data. Here we propose a novel approach to fuse these different types of sensor data to reconstruct 3D scenes which maintains the shape of local objects. Our method features two main phases: the local range modeling phase and 3D depth map reconstruction phase. In the local range modeling phase we interpolate 3D points from the laser scanner using Gaussian Process regression. It estimates 3D measurements across the outdoor environment and accommodates for defective sensor information. In the reconstruction phase, we fuse an image and interpolated points to build a 3D depth map and optimize based on a Markov Random Field. It provides a depth map corresponding to all image pixels. Qualitative and time complexity results show that our approach is robust and fast enough to demonstrate in real-time for an autonomous vehicle in complex urban scenes."
                },
                {
                    "title": ">Efficient Estimation of Partially Linear Models for Data on Complicated Domains by Bivariate Penalized Splines over Triangulations",
                    "abstract": "In this paper, we study the estimation of partially linear models for spatial data distributed over complex domains. We use bivariate splines over triangulations to represent the nonparametric component on an irregular two-dimensional domain. The proposed method is formulated as a constrained minimization problem which does not require constructing finite elements or locally supported basis functions. Thus, it allows an easier implementation of piecewise polynomial representations of various degrees and various smoothness over an arbitrary triangulation. Moreover, the constrained minimization problem is converted into an unconstrained minimization via a QR decomposition of the smoothness constraints, which leads to a penalized least squares method to estimate the model. The estimators of the parameters are proved to be asymptotically normal under some regularity conditions. The estimator of the bivariate function is consistent, and its rate of convergence is also established. The proposed method enables us to construct confidence intervals and permits inference for the parameters. The performance of the estimators is evaluated by two simulation examples and by a real data analysis."
                },
                {
                    "title": "TetGen, a Delaunay-Based Quality Tetrahedral Mesh Generator",
                    "abstract": "TetGen is a C++ program for generating good quality tetrahedral meshes aimed to support numerical methods and scientific computing. The problem of quality tetrahedral mesh generation is challenged by many theoretical and practical issues. TetGen uses Delaunay-based algorithms which have theoretical guarantee of correctness. It can robustly handle arbitrary complex 3D geometries and is fast in practice. The source code of TetGen is freely available. This article presents the essential algorithms and techniques used to develop TetGen. The intended audience are researchers or developers in mesh generation or other related areas. It describes the key software components of TetGen, including an efficient tetrahedral mesh data structure, a set of enhanced local mesh operations (combination of flips and edge removal), and filtered exact geometric predicates. The essential algorithms include incremental Delaunay algorithms for inserting vertices, constrained Delaunay algorithms for inserting constraints (edges and triangles), a new edge recovery algorithm for recovering constraints, and a new constrained Delaunay refinement algorithm for adaptive quality tetrahedral mesh generation. Experimental examples as well as comparisons with other softwares are presented."
                },
                {
                    "title": "FIBER DIRECTION ESTIMATION, SMOOTHING AND TRACKING IN DIFFUSION MRI.",
                    "abstract": "Diffusion magnetic resonance imaging is an imaging technology designed to probe anatomical architectures of biological samples in an in vivo and noninvasive manner through measuring water diffusion. The contribution of this paper is threefold. First, it proposes a new method to identify and estimate multiple diffusion directions within a voxel through a new and identifiable parametrization of the widely used multi-tensor model. Unlike many existing methods, this method focuses on the estimation of diffusion directions rather than the diffusion tensors. Second, this paper proposes a novel direction smoothing method which greatly improves direction estimation in regions with crossing fibers. This smoothing method is shown to have excellent theoretical and empirical properties. Last, this paper develops a fiber tracking algorithm that can handle multiple directions within a voxel. The overall methodology is illustrated with simulated data and a data set collected for the study of Alzheimer's disease by the Alzheimer's Disease Neuroimaging Initiative (ADNI)."
                },
                {
                    "title": "Smooth Scalar-on-Image Regression via Spatial Bayesian Variable Selection",
                    "abstract": "We develop scalar-on-image regression models when images are registered multidimensional manifolds. We propose a fast and scalable Bayes\u2019 inferential procedure to estimate the image coefficient. The central idea is the combination of an Ising prior distribution, which controls a latent binary indicator map, and an intrinsic Gaussian Markov random field, which controls the smoothness of the nonzero coefficients. The model is fit using a single-site Gibbs sampler, which allows fitting within minutes for hundreds of subjects with predictor images containing thousands of locations. The code is simple and is provided in the online Appendix (see the \u201cSupplementary Materials\u201d section). We apply this method to a neuroimaging study where cognitive outcomes are regressed on measures of white-matter microstructure at every voxel of the corpus callosum for hundreds of subjects."
                },
                {
                    "title": "Two-step spline estimating equations for generalized additive partially linear models with large cluster sizes",
                    "abstract": "We propose a two-step estimating procedure for generalized additive partially linear models with clustered data using estimating equations. Our proposed method applies to the case that the number of observations per cluster is allowed to increase with the number of independent subjects. We establish oracle properties for the two-step estimator of each function component such that it performs as well as the univariate function estimator by assuming that the parametric vector and all other function components are known. Asymptotic distributions and consistency properties of the estimators are obtained. Finite-sample experiments with both simulated continuous and binary response variables confirm the asymptotic results. We illustrate the methods with an application to a U.S. unemployment data set."
                },
                {
                    "title": "An explicit link between Gaussian fields and Gaussian Markov random fields: the stochastic partial differential equation approach",
                    "abstract": "Summary.\u2002 Continuously indexed Gaussian fields (GFs) are the most important ingredient in spatial statistical modelling and geostatistics. The specification through the covariance function gives an intuitive interpretation of the field properties. On the computational side, GFs are hampered with the big n problem, since the cost of factorizing dense matrices is cubic in the dimension. Although computational power today is at an all time high, this fact seems still to be a computational bottleneck in many applications. Along with GFs, there is the class of Gaussian Markov random fields (GMRFs) which are discretely indexed. The Markov property makes the precision matrix involved sparse, which enables the use of numerical algorithms for sparse matrices, that for fields in only use the square root of the time required by general algorithms. The specification of a GMRF is through its full conditional distributions but its marginal properties are not transparent in such a parameterization. We show that, using an approximate stochastic weak solution to (linear) stochastic partial differential equations, we can, for some GFs in the Mat\u00e9rn class, provide an explicit link, for any triangulation of , between GFs and GMRFs, formulated as a basis function representation. The consequence is that we can take the best from the two worlds and do the modelling by using GFs but do the computations by using GMRFs. Perhaps more importantly, our approach generalizes to other covariance functions generated by SPDEs, including oscillating and non\u2010stationary GFs, as well as GFs on manifolds. We illustrate our approach by analysing global temperature data with a non\u2010stationary model defined on a sphere."
                },
                {
                    "title": "Non-isotropic Gaussian smoothing and leakage reduction for determining mass changes over land and ocean using GRACE data",
                    "abstract": "SUMMARY We present a new non-isotropic Gaussian filter for smoothing mass changes computed from Gravity Recovery and Climate Experiment (GRACE) L2 products and a new method to reduce land\u2013ocean signal leakage caused by Gaussian smoothing. The kernel of our non-isotropic filter is the product of two Gaussian functions with distinct latitudinal and longitudinal smoothing radii. When expressed as number of kilometres at the Earth\u2019s surface, the longitudinal smoothing radius, defined as a fixed longitude interval, is longer at the equator, and shorter at higher latitude. This is principally in accordance with the resolution of the GRACE data, and permits us to produce homogeneously smoothed results without excessive smoothing in latitudinal direction. This filter is not applicable in polar regions, where we choose to use the classical isotropic Gaussian filter. A smoothing radius choice scheme is proposed for the two filters to mesh seamlessly. In our leakage reduction method, the inputs are the mass change data after smoothing using a Gaussian filter. Along coasts where mass change signal on land is far larger than that over ocean (or signal over ocean is reduced to a very small magnitude by removing a model beforehand, and adding the model back afterwards), our method approximately recovers a smoothed mass change signal over both land and ocean sides as if a regional Gaussian filter with the same smoothing radius were applied over land and ocean separately, in which no signal leakages appear. The side lobe problem does not appear in our approach. Our leakage reduction method could also be used to study mass changes within a region where signal is far larger than that in surrounding regions, or where signal in surrounding regions could be reduced to very low magnitude by removing a model."
                },
                {
                    "title": "Functional Generalized Linear Models with Images as Predictors",
                    "abstract": "Summary Functional principal component regression (FPCR) is a promising new method for regressing scalar outcomes on functional predictors. In this article, we present a theoretical justification for the use of principal components in functional regression. FPCR is then extended in two directions: from linear to the generalized linear modeling, and from univariate signal predictors to high\u2010resolution image predictors. We show how to implement the method efficiently by adapting generalized additive model technology to the functional regression context. A technique is proposed for estimating simultaneous confidence bands for the coefficient function; in the neuroimaging setting, this yields a novel means to identify brain regions that are associated with a clinical outcome. A new application of likelihood ratio testing is described for assessing the null hypothesis of a constant coefficient function. The performance of the methodology is illustrated via simulations and real data analyses with positron emission tomography images as predictors."
                },
                {
                    "title": "Nonstationary Spatial Gaussian Markov Random Fields",
                    "abstract": "Thin-plate splines have been widely used as spatial smoothers. In this article, we present a Bayesian adaptive thin-plate spline (BATS) suitable for modeling nonstationary spatial data. Fully Bayesian inference can be carried out through efficient Markov chain Monte Carlo simulation. Performance is demonstrated with simulation studies and by an application to a rainfall dataset. A FORTRAN program implementing the method, a proof of the theorem, and the dataset in this article are available online."
                },
                {
                    "title": "Tetrahedral mesh generation from volumetric binary and grayscale images",
                    "abstract": "We report a general purpose mesh generator for creating finite-element surface or volumetric mesh from 31) binary or gray-scale medical images. This toolbox incorporates a number of existing free mesh processing utilities and enables researchers to perform a range of mesh processing tasks for image-based mesh generation, including raw image processing, surface mesh extraction, surface re-sampling, and multi-scale/adaptive tetrahedral mesh generation. We also implemented robust algorithms for meshing open-surfaces and sub-region labeling. Atomic meshing utilities for each processing step can be accessed with simple interfaces, which can be streamlined or executed independently. The toolbox is compatible with Matlab or GNU Octave. We demonstrate the applications of this toolbox for meshing a range of challenging geometries including complex vessel network, human brain and breast."
                },
                {
                    "title": "A Domain Decomposition Method for Computing Bivariate Spline Fits of Scattered Data",
                    "abstract": "A domain decomposition method for solving large bivariate scattered data fitting problems with bivariate minimal energy, discrete least-squares, and penalized least-squares splines is described. The method is based on splitting the domain into smaller domains, solving the associated smaller fitting problems, and combining the coefficients to get a global fit. Explicit error bounds are established for how well our locally constructed spline fits approximate the global fits. Some numerical examples are given to illustrate the effectiveness of the method."
                },
                {
                    "title": "Spline Smoothing Over Difficult Regions",
                    "abstract": "We consider the problem of smoothing data on two-dimensional grids with holes or gaps. Such grids are often referred to as difficult regions. Since the data is not observed on these locations, the gap is not part of the domain. We cannot apply standard smoothing methods since they smooth over and across difficult regions. More unfavorable properties of standard smoothers become visible when the data is observed on an irregular grid in a non-rectangular domain. In this paper, we adopt smoothing spline methods within a state space framework to smooth data on one- or two-dimensional grids with difficult regions. We make a distinction between two types of missing observations to handle the irregularity of the grid and to ensure that no smoothing takes place over and across the difficult region. For smoothing on two-dimensional grids, we introduce a two-step spline smoothing method. The proposed solution applies to all smoothing methods that can be represented in a state space framework. We illustrate our methods for three different cases of interest."
                },
                {
                    "title": "On the asymptotics of penalized splines",
                    "abstract": "We study the asymptotic behaviour of penalized spline estimators in the univariate case. We use B-splines and a penalty is placed on mth-order differences of the coefficients. The number of knots is assumed to converge to infinity as the sample size increases. We show that penalized splines behave similarly to Nadaraya--Watson kernel estimators with 'equivalent' kernels depending upon m. The equivalent kernels we obtain for penalized splines are the same as those found by Silverman for smoothing splines. The asymptotic distribution of the penalized spline estimator is Gaussian and we give simple expressions for the asymptotic mean and variance. Provided that it is fast enough, the rate at which the number of knots converges to infinity does not affect the asymptotic distribution. The optimal rate of convergence of the penalty parameter is given. Penalized splines are not design-adaptive. Copyright 2008, Oxford University Press."
                },
                {
                    "title": "Spline functions on triangulations",
                    "abstract": "Preface 1. Bivariate polynomials 2. Bernstein-Bezier methods for bivariate polynomials 3. B-patches 4. Triangulations and quadrangulations 5. Bernstein-Bezier methods for spline spaces 6. C1 Macro-element spaces 7. C2 Macro-element spaces 8. Cr Macro-element spaces 9. Dimension of spline splines 10. Approximation power of spline spaces 11. Stable local minimal determining sets 12. Bivariate box splines 13. Spherical splines 14. Approximation power of spherical splines 15. Trivariate polynomials 16. Tetrahedral partitions 17. Trivariate splines 18. Trivariate macro-element spaces Bibliography Index."
                },
                {
                    "title": "Low\u2010Rank Smoothing Splines on Complicated Domains",
                    "abstract": "Summary Smoothing over a domain with irregular boundaries or interior gaps and holes is addressed. Consider the problem of estimating mercury in sediment concentrations in the estuarine waters in New Hampshire. A modified version of low\u2010rank thin plate splines (LTPS) is introduced where the geodesic distance is applied to evaluate dissimilarity of any two data observations: loosely speaking, distances between locations are not measured as the crow flies, but as the fish swims. The method is compared with competing smoothing techniques, LTPS, and finite element L\u2010splines."
                },
                {
                    "title": "Local asymptotics for polynomial spline regression",
                    "abstract": "In this paper we develop a general theory of local asymptotics for least squares estimates over polynomial spline spaces in a regression problem. The polynomial spline spaces we consider include univariate splines, tensor product splines, and bivariate or multivariate splines on triangulations. We establish asymptotic normality of the estimate and study the magnitude of the bias due to spline approximation. The asymptotic normality holds uniformly over the points where the regression function is to be estimated and uniformly over a broad class of design densities, error distributions and regression functions. The bias is controlled by the minimum L\u221e norm of the error when the target regression function is approximated by a function in the polynomial spline space that is used to define the estimate. The control of bias relies on the stability in L\u221e norm of L 2 projections onto polynomial spline spaces. Asymptotic normality of least squares estimates over polynomial or trigonometric polynomial spaces is also treated by the general theory. In addition, a preliminary analysis of additive models is provided."
                },
                {
                    "title": "Thin plate regression splines",
                    "abstract": "Summary. I discuss the production of low rank smoothers for d\u2003\u2265\u20031 dimensional data, which can be fitted by regression or penalized regression methods. The smoothers are constructed by a simple transformation and truncation of the basis that arises from the solution of the thin plate spline smoothing problem and are optimal in the sense that the truncation is designed to result in the minimum possible perturbation of the thin plate spline smoothing problem given the dimension of the basis used to construct the smoother. By making use of Lanczos iteration the basis change and truncation are computationally efficient. The smoothers allow the use of approximate thin plate spline models with large data sets, avoid the problems that are associated with \u2018knot placement\u2019 that usually complicate modelling with regression splines or penalized regression splines, provide a sensible way of modelling interaction terms in generalized additive models, provide low rank approximations to generalized smoothing spline models, appropriate for use with large data sets, provide a means for incorporating smooth functions of more than one variable into non\u2010linear models and improve the computational efficiency of penalized likelihood models incorporating thin plate splines. Given that the approach produces spline\u2010like models with a sparse basis, it also provides a natural way of incorporating unpenalized spline\u2010like terms in linear and generalized linear models, and these can be treated just like any other model terms from the point of view of model selection, inference and diagnostics."
                },
                {
                    "title": "QSplat: a multiresolution point rendering system for large meshes",
                    "abstract": "Advances in 3D scanning technologies have enabled the practical creation of meshes with hundreds of millions of polygons. Traditional algorithms for display, simplification, and progressive transmission of meshes are impractical for data sets of this size. We describe a system for representing and progressively displaying these meshes that combines a multiresolution hierarchy based on bounding spheres with a rendering system based on points. A single data structure is used for view frustum culling, backface culling, level-of-detail selection, and rendering. The representation is compact and can be computed quickly, making it suitable for large data sets. Our implementation, written for use in a large-scale 3D digitization project, launches quickly, maintains a user-settable interactive frame rate regardless of object complexity or camera position, yields reasonable image quality during motion, and refines progressively when idle to a high final image quality. We have demonstrated the system on scanned models containing hundreds of millions of samples."
                },
                {
                    "title": "A condition guaranteeing the existence of higher-dimensional constrained Delaunay triangulations",
                    "abstract": "Let X be a complex of vertices and piecewise linear constraining facets embedded in Ed. Say that a simplex is strongly Delaunay if its vertices are in X and there exists a sphere that passes through its vertices but passes through and encloses no other vertex. Then X has a d-dimensional constrained Delaunay triangulation if each k-dimensional constraining facet in X with k d 2 is a union of strongly Delaunay k-simplices. This theorem is especially useful in E3 for forming tetrahedralizations that respect specified planar facets. If the bounding segments of these facets are subdivided so that the subsegments are strongly Delaunay, then a constrained tetrahedralization exists. Hence, fewer vertices are needed than in the most common practice in the literature, wherein additional vertices are inserted in the relative interiors of facets to form a conforming (but unconstrained) Delaunay tetrahedralization."
                },
                {
                    "title": "Flexible smoothing with B-splines and penalties",
                    "abstract": "B-splines are attractive for nonparametric modelling, but choosing the optimal number and positions of knots is a complex task. Equidistant knots can be used, but their small and discrete number allows only limited control over smoothness and fit. We propose to use a relatively large number of knots and a difference penalty on coefficients of adjacent B-splines. We show connections to the familiar spline penalty on the integral of the squared second derivative. A short overview of $B$-splines, of their construction and of penalized likelihood is presented. We discuss properties of penalized B-splines and propose various criteria for the choice of an optimal penalty parameter. Nonparametric logistic regression, density estimation and scatterplot smoothing are used as examples. Some details of the computations are presented."
                },
                {
                    "title": "Nonparametric Regression and Generalized Linear Models: A roughness penalty approach",
                    "abstract": "Preface. Introduction. Approaches to Regression. Roughness Penalties. Extensions of the Roughness Penalty Approach. Computing the Estimates. Further Reading. Interpolating and Smoothing Splines. One-Dimensional Case: Further Topics. Partial Splines. Generalized Linear Models. Extending the Model. Thin Plate Splines. Available Software. Reference. Author Index. Subject Index."
                },
                {
                    "title": "Some first results on the consistency of spatial regression with partial differential equation regularization",
                    "abstract": "We study the consistency of the estimator in spatial regression with partial differential equation (PDE) regularization. This new smoothing technique allows to accurately estimate spatial fields over complex two-dimensional domains, starting from noisy observations; the regularizing term involves a PDE that formalizes problem specific information about the phenomenon at hand. Differently from classical smoothing methods, the solution of the infinite-dimensional estimation problem cannot be computed analytically. An approximation is obtained via the finite element method, considering a suitable triangulation of the spatial domain. We first consider the consistency of the estimator in the infinite-dimensional setting. We then study the consistency of the finite element estimator, resulting from the approximated PDE. We study the bias and variance of the estimators, with respect to the sample size and to the value of the smoothing parameter. Some final simulation studies provide numerical evidence of the rates derived for the bias, variance and mean square error."
                },
                {
                    "title": "Bivariate Penalized Splines for Regression",
                    "abstract": "In this paper the asymptotic behavior of penalized spline estimators is studied using bivariate splines over triangulations and an energy functional as the penalty. The rate of L2 convergence is derived, which achieves the optimal nonparametric convergence rate established by Stone (1982). The asymptotic normality of the penalized spline estimators is established, which is shown to hold uniformly over the points where the regression function is estimated. The size of the asymptotic conditional variance is also evaluated and a simple expression for the asymptotic variance is given. Simulation experiments have provided strong evidence that corroborates with the asymptotic theory."
                },
                {
                    "title": "SUPPLEMENT TO \u201cA SIMPLE BOOTSTRAP METHOD FOR CONSTRUCTING NONPARAMETRIC CONFIDENCE BANDS FOR FUNCTIONS.\u201d",
                    "abstract": "B.1. Outline of theoretical properties underpinning our methodology. Here we give a non-technical overview explaining why our methodology has the properties claimed (and proved theoretically) for it. We assume for simplicity that the problem is univariate (i.e. r = 1), and, except in the last paragraph of this appendix, we suppose that the set of points x \u2208 R for which b(x) = 0 is of measure zero. If the model at (2.1) obtains, and \u011d is a local linear estimator computed using a bandwidth h and a nonnegative, symmetric, compactly supported kernel, then the bootstrap bias estimator admits the expansion"
                },
                {
                    "title": "POLYNOMIAL SPLINE CONFIDENCE BANDS FOR REGRESSION CURVES",
                    "abstract": "Asymptotically exact and conservative confidence bands are obtained for a nonparametric regression function, using piecewise constant and piecewise lin- ear spline estimation, respectively. Compared to the pointwise confidence interval of Huang (2003), the confidence bands are inflated by a factor proportional to {log (n)} 1/2 , with the same width order as the Nadaraya-Watson bands of Hardle (1989), and the local polynomial bands of Xia (1998) and Claeskens and Van Kei- legom (2003). Simulation experiments corroborate the asymptotic theory. The linear spline band has been used to identify an appropriate polynomial trend for fossil data."
                },
                {
                    "title": "The Multivariate Spline Method for Scattered Data Fitting and Numerical Solutions of Partial Differential Equations",
                    "abstract": "Multivariate spline functions are smooth piecewise polynomial functions over triangulations consisting of n-simplices in the Euclidean space IR. A straightforward method for using these spline functions to fit given scattered data and numerically solve elliptic partial differential equations is presented . This method does not require constructing macro-elements or locally supported basis functions nor computing the dimension of the finite element spaces or spline spaces. The method for splines in IR and IR has been implemented in MATLAB. Several numerical examples are shown to demonstrate the effectiveness and efficiency of the method. Table of"
                },
                {
                    "title": "A Simple Mesh Generator in MATLAB",
                    "abstract": "Creating a mesh is the first step in a wide range of applications, including scientific computing and computer graphics. An unstructured simplex mesh requires a choice of meshpoints (vertex nodes) ..."
                },
                {
                    "title": "Journal of the American Statistical Association Spatially Varying Coefficient Model for Neuroimaging Data with Jump Discontinuities Spatially Varying Coefficient Model for Neuroimaging Data with Jump Discontinuities",
                    "abstract": "Taylor & Francis makes every effort to ensure the accuracy of all the information (the \" Content \") contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. Motivated by recent work on studying massive imaging data in various neuroimaging studies, we propose a novel spatially varying coefficient model (SVCM) to capture the varying association between imaging measures in a three-dimensional volume (or two-dimensional surface) with a set of covariates. Two stylized features of neuorimaging data are the presence of multiple piecewise smooth regions with unknown edges and jumps and substantial spatial correlations. To specifically account for these two features, SVCM includes a measurement model with multiple varying coefficient functions, a jumping surface model for each varying coefficient function, and a functional principal component model. We develop a three-stage estimation procedure to simultaneously estimate the varying coefficient functions and the spatial correlations. The estimation procedure includes a fast multiscale adaptive estimation and testing procedure to independently estimate each varying coefficient function, while preserving its edges among different piecewise-smooth regions. We systematically investigate the asymptotic properties (e.g., consistency and asymptotic normality) of the multiscale adaptive parameter estimates. We also establish the uniform convergence rate of the estimated spatial covariance function and its associated eigenvalues and eigenfunctions. Our Monte Carlo simulation and real-data analysis have confirmed the excellent performance of SVCM. Supplementary materials for this article are available online."
                }
            ],
            "categories": [
                "stat.CO",
                "62G05, 62G08"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively extract underlying signals from irregularly shaped 3D point clouds in various applications, such as healthcare and mining?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of 3D data analysis, particularly in fields like medical imaging and environmental science. By developing methods to extract meaningful signals from 3D point clouds, we can enhance the accuracy of diagnoses in healthcare, improve resource exploration in mining, and facilitate better decision-making in various industries. This research could lead to new methodologies that not only advance theoretical knowledge but also have practical applications in real-world scenarios, ultimately influencing future research directions and technological innovations.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the irregular shapes of 3D point clouds, which complicate the extraction of underlying signals. Conventional methods often rely on assumptions of regularity or uniformity that do not hold in this context, leading to potential inaccuracies. Additionally, the high dimensionality and noise inherent in point cloud data pose significant technical obstacles. Naive approaches may fail to account for these complexities, resulting in oversimplified models that do not capture the true structure of the data. Overcoming these challenges requires sophisticated algorithms that can handle irregular geometries and extract meaningful patterns.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on surface reconstruction techniques rather than signal extraction from 3D point clouds. Existing solutions often lack the necessary frameworks to address the irregularity and complexity of the data. Barriers such as limited computational resources, inadequate theoretical models, and a lack of interdisciplinary approaches have hindered progress in this area. My approach aims to bridge these gaps by integrating advanced statistical methods and machine learning techniques specifically designed for irregular data, thus improving upon prior work and offering a more robust solution.\n\n**[Question 5] - What are the key components of my approach and results?**  \nMy proposed methodology involves developing a novel algorithm that combines machine learning techniques with advanced statistical methods tailored for irregular 3D point clouds. I will utilize a diverse dataset of 3D scans from various applications, including medical imaging and geological surveys. The performance of the algorithm will be evaluated using metrics such as signal-to-noise ratio and accuracy of signal extraction. Expected outcomes include improved methods for signal extraction that can be applied across multiple domains, leading to enhanced understanding"
            }
        },
        "author_data": {
            "d49162e7-dc3c-45c6-aebf-caf5edb9a9a9": {
                "pk": "d49162e7-dc3c-45c6-aebf-caf5edb9a9a9",
                "name": "Xinyi Li",
                "collaborators": [
                    "Daisuke Shiraishi",
                    "Haibin Ling",
                    "Zijie Zhuang",
                    "Runsheng Liu",
                    "Yu Liu",
                    "Kyle Fox",
                    "Gefei Cai",
                    "Marcelo Hilario",
                    "Petr Panov"
                ],
                "domain": [
                    "Stochastic Processes",
                    "Percolation Theory",
                    "Random Walks",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Percolative properties of Brownian interlacements and its vacant set",
                        "abstract": "In this article we investigate the percolative properties of Brownian interlacements, a model introduced by Alain-Sol Sznitman in arXiv:1209.4531, and show that: the interlacement set is \"well-connected\", i.e., any two \"sausages\" in $d$-dimensional Brownian interlacements, $d\\geq 3$, can be connected via no more than $\\lceil (d-4)/2 \\rceil$ intermediate sausages almost surely; while the vacant set undergoes a non-trivial percolation phase transition when the level parameter varies."
                    },
                    {
                        "title": "A lower bound for disconnection by simple random walk",
                        "abstract": "We consider simple random walk on Z^d, d bigger or equal to 3. Motivated by the work of A.-S. Sznitman and the author in arXiv:1304.7477 and arXiv:1310.2177, we investigate the asymptotic behaviour of the probability that a large body gets disconnected from infinity by the set of points visited by a simple random walk. We derive asymptotic lower bounds that bring into play random interlacements. Although open at the moment, some of the lower bounds that we obtain possibly match the asymptotic upper bounds obtained in a recent article of A.-S. Sznitman. This potentially yields special significance to the tilted walks that we use in this work, and to the strategy that we employ to implement disconnection."
                    },
                    {
                        "title": "The H\u00f6lder continuity of the scaling limit of three-dimensional loop-erased random walk",
                        "abstract": "Let $\\beta$ be the growth exponent of the loop-erased random walk (LERW) in three dimensions. We prove that the scaling limit of 3D LERW is $h$-H\\\"older continuous almost surely for all $h < 1/\\beta$, while not $1/\\beta$-H\\\"older continuous almost surely."
                    },
                    {
                        "title": "Convergence of three-dimensional loop-erased random walk in the natural parametrization",
                        "abstract": "In this work we consider loop-erased random walk (LERW) and its scaling limit in three dimensions, and prove that 3D LERW parametrized by renormalized length converges to its scaling limit parametrized by some suitable measure with respect to the uniform convergence topology in the lattice size scaling limit. Our result greatly improves the work (Acta Math. 199(1):29-152) of Gady Kozma which establishes the weak convergence of the rescaled trace of 3D LERW towards a random compact set with respect to the Hausdorff distance."
                    },
                    {
                        "title": "One-point function estimates for loop-erased random walk in three dimensions",
                        "abstract": "In this work, we consider loop-erased random walk (LERW) in three dimensions and give an asymptotic estimate on the one-point function for LERW and the non-intersection probability of LERW and simple random walk in three dimensions for dyadic scales. These estimates will be crucial to the characterization of the convergence of LERW to its scaling limit in natural parametrization. As a step in the proof, we also obtain a coupling of two pairs of LERW and SRW with different starting points conditioned to avoid each other."
                    },
                    {
                        "title": "On large deviations and intersection of random interlacements",
                        "abstract": "We investigate random interlacements on $\\mathbb{Z}^d$ with $d \\geq 3$, and derive the large deviation rate for the probability that the capacity of the interlacement set in a macroscopic box is much smaller than that of the box. As an application, we obtain the large deviation rate for the probability that two independent interlacements have empty intersections in a macroscopic box. We also prove that conditioning on this event, one of them will be sparse in the box in terms of capacity. This result is an example of the entropic repulsion phenomenon for random interlacements."
                    },
                    {
                        "title": "The intermediate level-sets of the four-dimensional membrane model",
                        "abstract": "In this paper, we consider the discrete membrane model in four dimensions. We confirm the existence of the scaling limit of the intermediate (i.e., a multiple of the expected maximum) level-sets of the model, and show that it is equal in law to the sub-critical Gaussian multiplicative chaos (GMC) measure of the continuum membrane model."
                    },
                    {
                        "title": "On the Robustness of Multi-View Rotation Averaging",
                        "abstract": "Rotation averaging is a synchronization process on single or multiple rotation groups, and is a fundamental problem in many computer vision tasks such as multi-view structure from motion (SfM). Specifically, rotation averaging involves the recovery of an underlying pose-graph consistency from pairwise relative camera poses. Specifically, given pairwise motion in rotation groups, especially 3-dimensional rotation groups (\\eg, $\\mathbb{SO}(3)$), one is interested in recovering the original signal of multiple rotations with respect to a fixed frame. In this paper, we propose a robust framework to solve multiple rotation averaging problem, especially in the cases that a significant amount of noisy measurements are present. By introducing the $\\epsilon$-cycle consistency term into the solver, we enable the robust initialization scheme to be implemented into the IRLS solver. Instead of conducting the costly edge removal, we implicitly constrain the negative effect of erroneous measurements by weight reducing, such that IRLS failures caused by poor initialization can be effectively avoided. Experiment results demonstrate that our proposed approach outperforms state of the arts on various benchmarks."
                    },
                    {
                        "title": "Hybrid Camera Pose Estimation with Online Partitioning for SLAM",
                        "abstract": "This paper presents a hybrid real-time camera pose estimation framework with a novel partitioning scheme and introduces motion averaging to monocular Simultaneous Localization and Mapping (SLAM) systems. Breaking through the limitations of fixed-size temporal partitioning in many conventional SLAM pipelines, our approach significantly improves the accuracy of local bundle adjustment by gathering spatially-strongly-connected cameras into each block. With the dynamic initialization using intermediate computation values, \\XL{we improve the Levenberg-Marquardt solver to further enhance the efficiency of the local optimization.} Moreover, the dense data association between blocks by our co-visibility-based partitioning enables us to explore and implement motion averaging to efficiently align the blocks globally, updating camera motion estimations on-the-fly. Experiments on benchmarks convincingly demonstrate the practicality and robustness of our proposed approach by significantly outperforming conventional approaches."
                    },
                    {
                        "title": "TransCamP: Graph Transformer for 6-DoF Camera Pose Estimation",
                        "abstract": "Camera pose estimation or camera relocalization is the centerpiece in numerous computer vision tasks such as visual odometry, structure from motion (SfM) and SLAM. In this paper we propose a neural network approach with a graph transformer backbone, namely TransCamP, to address the camera relocalization problem. In contrast with prior work where the pose regression is mainly guided by photometric consistency, TransCamP effectively fuses the image features, camera pose information and inter-frame relative camera motions into encoded graph attributes and is trained towards the graph consistency and accuracy instead, yielding significantly higher computational efficiency. By leveraging graph transformer layers with edge features and enabling tensorized adjacency matrix, TransCamP dynamically captures the global attention and thus endows the pose graph with evolving structures to achieve improved robustness and accuracy. In addition, optional temporal transformer layers actively enhance the spatiotemporal inter-frame relation for sequential inputs. Evaluation of the proposed network on various public benchmarks demonstrates that TransCamP outperforms state-of-the-art approaches."
                    },
                    {
                        "title": "Sharpness of phase transition for Voronoi percolation in hyperbolic space",
                        "abstract": "In this paper, we consider Voronoi percolation in the hyperbolic space $\\mathbb{H}^d$ ($d\\ge 2$) and show that the phase transition is sharp. More precisely, we show that for Voronoi percolation with parameter $p$ generated by a homogeneous Poisson point process with intensity $\\lambda$, there exists $p_c:=p_c(\\lambda,d)$ such that the probability of a monochromatic path from the origin reaching a distance of $n$ decays exponentially fast in $n$. We also prove the mean-field lower bound $\\mathbb{P}_{\\lambda,p}(0\\leftrightarrow \\infty)\\ge c(p-p_c)$ for $p>p_c$."
                    },
                    {
                        "title": "Approximating the Geometric Edit Distance",
                        "abstract": "Edit distance is a measurement of similarity between two sequences such as strings, point sequences, or polygonal curves. Many matching problems from a variety of areas, such as signal analysis, bioinformatics, etc., need to be solved in a geometric space. Therefore, the geometric edit distance (GED) has been studied. In this paper, we describe the first strictly sublinear approximate near-linear time algorithm for computing the GED of two point sequences in constant dimensional Euclidean space. Specifically, we present a randomized (O(n\\log^2n)) time (O(\\sqrt n))-approximation algorithm. Then, we generalize our result to give a randomized $\\alpha$-approximation algorithm for any $\\alpha\\in [\\sqrt{\\log n}, \\sqrt{n / \\log n}]$, running in time $O(n^2/\\alpha^2 \\log n)$. Both algorithms are Monte Carlo and return approximately optimal solutions with high probability."
                    },
                    {
                        "title": "On Natural Measures of SLE- and CLE-Related Random Fractals",
                        "abstract": "In this paper, we construct and then prove the up-to constants uniqueness of the natural measure on several random fractals, namely the SLE cut points, SLE boundary touching points, CLE pivotal points and the CLE carpet/gasket. As an application, we also show the equivalence between our natural measures defined in this paper (i.e. CLE pivotal and gasket measures) and their discrete analogs of counting measures in critical continuum planar Bernoulli percolation in [Garban-Pete-Schramm, J. Amer. Math. Soc.,2013]. Although the existence and uniqueness for the natural measure for CLE carpet/gasket have already been proved in [Miller-Schoug, arXiv:2201.01748], in this paper we provide with a different argument via the coupling of CLE and LQG."
                    },
                    {
                        "title": "Shape theorem and surface fluctuation for Poisson cylinders",
                        "abstract": "In this work, we prove a shape theorem for Poisson cylinders and give a power law bound on surface fluctuations. We prove that for any $a \\in (1/2, 1)$, conditioned on the origin being in the set of cylinders, every point in this set, whose Euclidean norm is less than $R$, lies at an internal distance less than $R+O(R^a)$ from the origin."
                    }
                ]
            },
            "9449010e-66c1-4e6a-b79a-e4c6c615d376": {
                "pk": "9449010e-66c1-4e6a-b79a-e4c6c615d376",
                "name": "Shan Yu",
                "collaborators": [
                    "Bo Cui",
                    "Danko Nikoli\u0107",
                    "He Zhu",
                    "Fangbo Qin",
                    "Taogang Hou",
                    "Shan Lin",
                    "Kaiyuan Wang",
                    "Michael C. Yip",
                    "Hongdian Yang",
                    "Oren Shriki"
                ],
                "domain": [
                    "Quantum Mechanics",
                    "Machine Learning",
                    "Neural Networks",
                    "Action Recognition"
                ],
                "publications": [
                    {
                        "title": "Quantum mechanics needs no consciousness (and the other way around)",
                        "abstract": "It has been suggested that consciousness plays an important role in quantum mechanics as it is necessary for the collapse of wave function during the measurement. Furthermore, this idea has spawned a symmetrical proposal: a possibility that quantum mechanics explains the emergence of consciousness in the brain. Here we formulated several predictions that follow from this hypothetical relationship and that can be empirically tested. Some of the experimental results that are already available suggest falsification of the first hypothesis. Thus, the suggested link between human consciousness and collapse of wave function does not seem viable. We discuss the constraints implied by the existing evidence on the role that the human observer may play for quantum mechanics and the role that quantum mechanics may play in the observer's consciousness."
                    },
                    {
                        "title": "Intra-Class Uncertainty Loss Function for Classification",
                        "abstract": "Most classification models can be considered as the process of matching templates. However, when intra-class uncertainty/variability is not considered, especially for datasets containing unbalanced classes, this may lead to classification errors. To address this issue, we propose a loss function with intra-class uncertainty following Gaussian distribution. Specifically, in our framework, the features extracted by deep networks of each class are characterized by independent Gaussian distribution. The parameters of distribution are learned with a likelihood regularization along with other network parameters. The means of the Gaussian play a similar role as the center anchor in existing methods, and the variance describes the uncertainty of different classes. In addition, similar to the inter-class margin in traditional loss functions, we introduce a margin to intra-class uncertainty to make each cluster more compact and reduce the imbalance of feature distribution from different categories. Based on MNIST, CIFAR, ImageNet, and Long-tailed CIFAR analyses, the proposed approach shows improved classification performance, through learning a better class representation."
                    },
                    {
                        "title": "AnyOKP: One-Shot and Instance-Aware Object Keypoint Extraction with Pretrained ViT",
                        "abstract": "Towards flexible object-centric visual perception, we propose a one-shot instance-aware object keypoint (OKP) extraction approach, AnyOKP, which leverages the powerful representation ability of pretrained vision transformer (ViT), and can obtain keypoints on multiple object instances of arbitrary category after learning from a support image. An off-the-shelf petrained ViT is directly deployed for generalizable and transferable feature extraction, which is followed by training-free feature enhancement. The best-prototype pairs (BPPs) are searched for in support and query images based on appearance similarity, to yield instance-unaware candidate keypoints.Then, the entire graph with all candidate keypoints as vertices are divided to sub-graphs according to the feature distributions on the graph edges. Finally, each sub-graph represents an object instance. AnyOKP is evaluated on real object images collected with the cameras of a robot arm, a mobile robot, and a surgical robot, which not only demonstrates the cross-category flexibility and instance awareness, but also show remarkable robustness to domain shift and viewpoint change."
                    },
                    {
                        "title": "Universal Organization of Resting Brain Activity at the Thermodynamic Critical Point",
                        "abstract": "Thermodynamic criticality describes emergent phenomena in a wide variety of complex systems. In the mammalian brain, the complex dynamics that spontaneously emerge from neuronal interactions have been characterized as neuronal avalanches, a form of critical branching dynamics. Here, we show that neuronal avalanches also reflect that the brain dynamics are organized close to a thermodynamic critical point. We recorded spontaneous cortical activity in monkeys and humans at rest using high-density intracranial microelectrode arrays and magnetoencephalography, respectively. By numerically changing a control parameter equivalent to thermodynamic temperature, we observed typical critical behavior in cortical activities near the actual physiological condition, including the phase transition of an order parameter, as well as the divergence of susceptibility and specific heat. Finite-size scaling of these quantities allowed us to derive robust critical exponents highly consistent across monkey and humans that uncover a distinct, yet universal organization of brain dynamics."
                    },
                    {
                        "title": "Nonparametric Automatic Differentiation Variational Inference with Spline Approximation",
                        "abstract": "Automatic Differentiation Variational Inference (ADVI) is efficient in learning probabilistic models. Classic ADVI relies on the parametric approach to approximate the posterior. In this paper, we develop a spline-based nonparametric approximation approach that enables flexible posterior approximation for distributions with complicated structures, such as skewness, multimodality, and bounded support. Compared with widely-used nonparametric variational inference methods, the proposed method is easy to implement and adaptive to various data structures. By adopting the spline approximation, we derive a lower bound of the importance weighted autoencoder and establish the asymptotic consistency. Experiments demonstrate the efficiency of the proposed method in approximating complex posterior distributions and improving the performance of generative models with incomplete data."
                    },
                    {
                        "title": "Continual Learning of Context-dependent Processing in Neural Networks",
                        "abstract": "Deep neural networks (DNNs) are powerful tools in learning sophisticated but fixed mapping rules between inputs and outputs, thereby limiting their application in more complex and dynamic situations in which the mapping rules are not kept the same but changing according to different contexts. To lift such limits, we developed a novel approach involving a learning algorithm, called orthogonal weights modification (OWM), with the addition of a context-dependent processing (CDP) module. We demonstrated that with OWM to overcome the problem of catastrophic forgetting, and the CDP module to learn how to reuse a feature representation and a classifier for different contexts, a single network can acquire numerous context-dependent mapping rules in an online and continual manner, with as few as $\\sim$10 samples to learn each. This should enable highly compact systems to gradually learn myriad regularities of the real world and eventually behave appropriately within it."
                    },
                    {
                        "title": "Skeleton-Based Action Recognition with Synchronous Local and Non-local Spatio-temporal Learning and Frequency Attention",
                        "abstract": "Benefiting from its succinctness and robustness, skeleton-based action recognition has recently attracted much attention. Most existing methods utilize local networks (e.g., recurrent, convolutional, and graph convolutional networks) to extract spatio-temporal dynamics hierarchically. As a consequence, the local and non-local dependencies, which contain more details and semantics respectively, are asynchronously captured in different level of layers. Moreover, existing methods are limited to the spatio-temporal domain and ignore information in the frequency domain. To better extract synchronous detailed and semantic information from multi-domains, we propose a residual frequency attention (rFA) block to focus on discriminative patterns in the frequency domain, and a synchronous local and non-local (SLnL) block to simultaneously capture the details and semantics in the spatio-temporal domain. Besides, a soft-margin focal loss (SMFL) is proposed to optimize the learning whole process, which automatically conducts data selection and encourages intrinsic margins in classifiers. Our approach significantly outperforms other state-of-the-art methods on several large-scale datasets."
                    }
                ]
            },
            "41d369c6-8bb9-4376-8bf0-879e9d024825": {
                "pk": "41d369c6-8bb9-4376-8bf0-879e9d024825",
                "name": "Yueying Wang",
                "collaborators": [
                    "Guannan Wang",
                    "Li Wang",
                    "Zhiling Gu",
                    "Xinyi Li",
                    "Shan Yu",
                    "Myungjin Kim",
                    "Lei Gao",
                    "R. Todd Ogden",
                    "Shurong Mo",
                    "Nailong Wu"
                ],
                "domain": [
                    "Biomedical Imaging",
                    "Epidemic Modeling",
                    "Reinforcement Learning",
                    "Data Analysis"
                ],
                "publications": [
                    {
                        "title": "Simultaneous Confidence Corridors for Mean Functions in Functional Data Analysis of Imaging Data",
                        "abstract": "Motivated by recent work involving the analysis of biomedical imaging data, we present a novel procedure for constructing simultaneous confidence corridors for the mean of imaging data. We propose to use flexible bivariate splines over triangulations to handle irregular domain of the images that is common in brain imaging studies and in other biomedical imaging applications. The proposed spline estimators of the mean functions are shown to be consistent and asymptotically normal under some regularity conditions. We also provide a computationally efficient estimator of the covariance function and derive its uniform consistency. The procedure is also extended to the two-sample case in which we focus on comparing the mean functions from two populations of imaging data. Through Monte Carlo simulation studies we examine the finite-sample performance of the proposed method. Finally, the proposed method is applied to analyze brain Positron Emission Tomography (PET) data in two different studies. One dataset used in preparation of this article was obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database."
                    },
                    {
                        "title": "Proximal Policy Optimization Learning based Control of Congested Freeway Traffic",
                        "abstract": "This study proposes a delay-compensated feedback controller based on proximal policy optimization (PPO) reinforcement learning to stabilize traffic flow in the congested regime by manipulating the time-gap of adaptive cruise control-equipped (ACC-equipped) vehicles.The traffic dynamics on a freeway segment are governed by an Aw-Rascle-Zhang (ARZ) model, consisting of $2\\times 2$ nonlinear first-order partial differential equations (PDEs).Inspired by the backstepping delay compensator [18] but different from whose complex segmented control scheme, the PPO control is composed of three feedbacks, namely the current traffic flow velocity, the current traffic flow density and previous one step control input. The control gains for the three feedbacks are learned from the interaction between the PPO and the numerical simulator of the traffic system without knowing the system dynamics. Numerical simulation experiments are designed to compare the Lyapunov control, the backstepping control and the PPO control. The results show that for a delay-free system, the PPO control has faster convergence rate and less control effort than the Lyapunov control. For a traffic system with input delay, the performance of the PPO controller is comparable to that of the Backstepping controller, even for the situation that the delay value does not match. However, the PPO is robust to parameter perturbations, while the Backstepping controller cannot stabilize a system where one of the parameters is disturbed by Gaussian noise."
                    },
                    {
                        "title": "Comparing and Integrating US COVID-19 Data from Multiple Sources with Anomaly Detection and Repairing",
                        "abstract": "Over the past few months, the outbreak of Coronavirus disease (COVID-19) has been expanding over the world. A reliable and accurate dataset of the cases is vital for scientists to conduct related research and for policy-makers to make better decisions. We collect the United States COVID-19 daily reported data from four open sources: the New York Times, the COVID-19 Data Repository by Johns Hopkins University, the COVID Tracking Project at the Atlantic, and the USAFacts, then compare the similarities and differences among them. To obtain reliable data for further analysis, we first examine the cyclical pattern and the following anomalies, which frequently occur in the reported cases: (1) the order dependencies violation, (2) the point or period anomalies, and (3) the issue of reporting delay. To address these detected issues, we propose the corresponding repairing methods and procedures if corrections are necessary. In addition, we integrate the COVID-19 reported cases with the county-level auxiliary information of the local features from official sources, such as health infrastructure, demographic, socioeconomic, and environmental information, which are also essential for understanding the spread of the virus."
                    },
                    {
                        "title": "Spatiotemporal Dynamics, Nowcasting and Forecasting of COVID-19 in the United States",
                        "abstract": "Epidemic modeling is an essential tool to understand the spread of the novel coronavirus and ultimately assist in disease prevention, policymaking, and resource allocation. In this article, we establish a state of the art interface between classic mathematical and statistical models and propose a novel space-time epidemic modeling framework to study the spatial-temporal pattern in the spread of infectious disease. We propose a quasi-likelihood approach via the penalized spline approximation and alternatively reweighted least-squares technique to estimate the model. Furthermore, we provide a short-term and long-term county-level prediction of the infected/death count for the U.S. by accounting for the control measures, health service resources, and other local features. Utilizing spatiotemporal analysis, our proposed model enhances the dynamics of the epidemiological mechanism and dissects the spatiotemporal structure of the spreading disease. To assess the uncertainty associated with the prediction, we develop a projection band based on the envelope of the bootstrap forecast paths. The performance of the proposed method is evaluated by a simulation study. We apply the proposed method to model and forecast the spread of COVID-19 at both county and state levels in the United States."
                    }
                ]
            },
            "cea1250b-f1e4-4985-bd14-8c35bb0de9fc": {
                "pk": "cea1250b-f1e4-4985-bd14-8c35bb0de9fc",
                "name": "Guannan Wang",
                "collaborators": [
                    "Li Wang",
                    "Shan Yu",
                    "Jue Wang",
                    "Tao Zhou",
                    "Yan-Li Lee",
                    "Yueying Wang",
                    "R. Todd Ogden",
                    "Min-Jun Lai",
                    "Lei Gao",
                    "Lijian Yang"
                ],
                "domain": [
                    "Spatial Statistics",
                    "Image Analysis",
                    "Nonparametric Methods",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "On Selection of Semiparametric Spatial Regression Models",
                        "abstract": "In this paper, we focus on the variable selection techniques for a class of semiparametric spatial regression models which allow one to study the effects of explanatory variables in the presence of the spatial information. The spatial smoothing problem in the nonparametric part is tackled by means of bivariate splines over triangulation, which is able to deal efficiently with data distributed over irregularly shaped regions. In addition, we develop a unified procedure for variable selection to identify significant covariates under a double penalization framework, and we show that the penalized estimators enjoy the \"oracle\" property. The proposed method can simultaneously identify non-zero spatially distributed covariates and solve the problem of \"leakage\" across complex domains of the functional spatial component. To estimate the standard deviations of the proposed estimators for the coefficients, a sandwich formula is developed as well. In the end, Monte Carlo simulation examples and a real data example are provided to illustrate the proposed methodology. All technical proofs are given in the supplementary materials."
                    },
                    {
                        "title": "Experimental analyses on 2-hop-based and 3-hop-based link prediction algorithms",
                        "abstract": "Link prediction is a significant and challenging task in network science. The majority of known methods are similarity-based, which assign similarity indices for node pairs and assume that two nodes of larger similarity have higher probability to be connected by a link. Due to their simplicity, interpretability and high efficiency, similarity-based methods, in particular those based only on local information, have already found successful applications on disparate fields. In this research domain, an intuitive consensus is that two nodes sharing common neighbors are very likely to have a link, while some recent evidences argue that the number of 3-hop paths more accurately predicts missing links than the number of common neighbors. In this paper, we implement extensive experimental comparisons between 2-hop-based and 3-hop-based similarity indices on 128 real networks. Our results indicate that the 3-hop-based indices perform slightly better with a winning rate about 55.88%, but which index is the best one still depends on the target network. Overall speaking, the class of Cannistraci-Hebb indices performs the best among all considered candidates."
                    },
                    {
                        "title": "Simultaneous Confidence Corridors for Mean Functions in Functional Data Analysis of Imaging Data",
                        "abstract": "Motivated by recent work involving the analysis of biomedical imaging data, we present a novel procedure for constructing simultaneous confidence corridors for the mean of imaging data. We propose to use flexible bivariate splines over triangulations to handle irregular domain of the images that is common in brain imaging studies and in other biomedical imaging applications. The proposed spline estimators of the mean functions are shown to be consistent and asymptotically normal under some regularity conditions. We also provide a computationally efficient estimator of the covariance function and derive its uniform consistency. The procedure is also extended to the two-sample case in which we focus on comparing the mean functions from two populations of imaging data. Through Monte Carlo simulation studies we examine the finite-sample performance of the proposed method. Finally, the proposed method is applied to analyze brain Positron Emission Tomography (PET) data in two different studies. One dataset used in preparation of this article was obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database."
                    },
                    {
                        "title": "Efficient Estimation of Partially Linear Models for Spatial Data over Complex Domain",
                        "abstract": "In this paper, we study the estimation of partially linear models for spatial data distributed over complex domains. We use bivariate splines over triangulations to represent the nonparametric component on an irregular two-dimensional domain. The proposed method is formulated as a constrained minimization problem which does not require constructing finite elements or locally supported basis functions. Thus, it allows an easier implementation of piecewise polynomial representations of various degrees and various smoothness over an arbitrary triangulation. Moreover, the constrained minimization problem is converted into an unconstrained minimization via a QR decomposition of the smoothness constraints, which allows for the development of a fast and efficient penalized least squares algorithm to fit the model. The estimators of the parameters are proved to be asymptotically normal under some regularity conditions. The estimator of the bivariate function is consistent, and its rate of convergence is also established. The proposed method enables us to construct confidence intervals and permits inference for the parameters. The performance of the estimators is evaluated by two simulation examples and by a real data analysis."
                    },
                    {
                        "title": "Multivariate Spline Estimation and Inference for Image-On-Scalar Regression",
                        "abstract": "Motivated by recent data analyses in biomedical imaging studies, we consider a class of image-on-scalar regression models for imaging responses and scalar predictors. We propose using flexible multivariate splines over triangulations to handle the irregular domain of the objects of interest on the images, as well as other characteristics of images. The proposed estimators of the coefficient functions are proved to be root-n consistent and asymptotically normal under some regularity conditions. We also provide a consistent and computationally efficient estimator of the covariance function. Asymptotic pointwise confidence intervals and data-driven simultaneous confidence corridors for the coefficient functions are constructed. Our method can simultaneously estimate and make inferences on the coefficient functions while incorporating spatial heterogeneity and spatial correlation. A highly efficient and scalable estimation algorithm is developed. Monte Carlo simulation studies are conducted to examine the finite-sample performance of the proposed method, which is then applied to the spatially normalized positron emission tomography data of the Alzheimer's Disease Neuroimaging Initiative."
                    },
                    {
                        "title": "Nonparametric Density Estimation for Data Scattered on Irregular Spatial Domains: A Likelihood-Based Approach Using Bivariate Penalized Spline Smoothing",
                        "abstract": "Accurately estimating data density is crucial for making informed decisions and modeling in various fields. This paper presents a novel nonparametric density estimation procedure that utilizes bivariate penalized spline smoothing over triangulation for data scattered over irregular spatial domains. The approach is likelihood-based with a regularization term that addresses the roughness of the logarithm of density based on a second-order differential operator. The proposed method offers greater efficiency and flexibility in estimating density over complex domains and has been theoretically supported by establishing the asymptotic convergence rate under mild natural conditions. Through extensive simulation studies and a real-world application that analyzes motor vehicle theft data from Portland City, Oregon, we demonstrate the advantages of the proposed method over existing techniques detailed in the literature."
                    },
                    {
                        "title": "FCDM: Sparse-view Sinogram Inpainting with Frequency Domain Convolution Enhanced Diffusion Models",
                        "abstract": "Reducing the radiation dose in computed tomography (CT) is crucial, but it often results in sparse-view CT, where the number of available projections is significantly reduced. This reduction in projection data makes it challenging to accurately reconstruct high-quality CT images. In this condition, a sinogram, which is a collection of these projections, becomes incomplete. Sinogram inpainting then becomes essential because it enables accurate image reconstruction with limited projections. Existing models performing well on conventional RGB images for inpainting mostly fail in the case of sinograms. Further, these models usually do not make full use of unique properties, e.g., frequency features and absorption characteristics in the sinogram, and cannot handle large-area masks and complex real-world projections well.   To address these limitations, we propose a novel model called the Frequency Convolution Diffusion Model (FCDM). It employs frequency domain convolutions to extract frequency information from various angles and capture the intricate relationships between these angles, which is essential for high-quality CT reconstruction. We also design a specific loss function based on the unique properties of a sinogram to maintain the consistency in physical properties, which allows the model to learn more effectively even in larger mask areas. We compare FCDM using both simulations and real data with nine inpainting models examples, among which two are designed for sinogram and seven for RGB. The results indicate that our model significantly improves the quality of the inpainted sinograms in terms of both visually and quantitatively, with an SSIM of more than 0.95 and PSNR of more than 30, achieving up to a 33% improvement in SSIM and a 29% improvement in PSNR compared to the baseline."
                    }
                ]
            },
            "84b792f5-a604-4f01-9e63-eda379167d4b": {
                "pk": "84b792f5-a604-4f01-9e63-eda379167d4b",
                "name": "Ming-Jun Lai",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "0a785ba3-9463-46f3-9512-5ac418b23a63": {
                "pk": "0a785ba3-9463-46f3-9512-5ac418b23a63",
                "name": "Li Wang",
                "collaborators": [
                    "Zenghu Li",
                    "Qian Li"
                ],
                "domain": [
                    "Stochastic Processes",
                    "Dimensionality Reduction",
                    "Data Mining",
                    "Option Pricing"
                ],
                "publications": [
                    {
                        "title": "Strong law of large number for branching Hunt processes",
                        "abstract": "In this paper we prove that, under certain conditions, a strong law of large number holds for a class of branching particle systems $X$ corresponding to the parameters $(Y,\\beta,\\psi)$, where $Y$ is a Hunt process and $\\psi$ is the generating function for the offspring. The main tool of this paper is the spine decomposition and we only need a $L\\log L$ condition."
                    },
                    {
                        "title": "An almost sure limit theorem for super-Brownian motion",
                        "abstract": "We establish an almost sure scaling limit theorem for super-Brownian motion on $\\mathbb{R}^d$ associated with the semi-linear equation $u_t = {1/2}\\Delta u +\\beta u-\\alpha u^2$, where $\\alpha$ and $\\beta$ are positive constants. In this case, the spectral theoretical assumptions that required in Chen et al (2008) are not satisfied. An example is given to show that the main results also hold for some sub-domains in $\\mathbb{R}^d$."
                    },
                    {
                        "title": "Central limit theorems for supercritical superprocesses with immigration",
                        "abstract": "In this paper, we establish a central limit theorem for a large class of general supercritical superprocesses with immigration with spatially dependent branching mechanisms satisfying a second moment condition. This central limit theorem extends and generalizes the results obtained by Ren, Song and Zhang[2015]. We first give law of large numbers for supercritical superprocesses with immigration since there is few convergence result on immigration superprocesses, then based on these results, we establish the central limit theorem."
                    },
                    {
                        "title": "Towards a Model for LSH",
                        "abstract": "As data volumes continue to grow, clustering and outlier detection algorithms are becoming increasingly time-consuming. Classical index structures for neighbor search are no longer sustainable due to the \"curse of dimensionality\". Instead, approximated index structures offer a good opportunity to significantly accelerate the neighbor search for clustering and outlier detection and to have the lowest possible error rate in the results of the algorithms. Locality-sensitive hashing is one of those. We indicate directions to model the properties of LSH."
                    },
                    {
                        "title": "Probabilistic Dimensionality Reduction via Structure Learning",
                        "abstract": "We propose a novel probabilistic dimensionality reduction framework that can naturally integrate the generative model and the locality information of data. Based on this framework, we present a new model, which is able to learn a smooth skeleton of embedding points in a low-dimensional space from high-dimensional noisy data. The formulation of the new model can be equivalently interpreted as two coupled learning problem, i.e., structure learning and the learning of projection matrix. This interpretation motivates the learning of the embedding points that can directly form an explicit graph structure. We develop a new method to learn the embedding points that form a spanning tree, which is further extended to obtain a discriminative and compact feature representation for clustering problems. Unlike traditional clustering methods, we assume that centers of clusters should be close to each other if they are connected in a learned graph, and other cluster centers should be distant. This can greatly facilitate data visualization and scientific discovery in downstream analysis. Extensive experiments are performed that demonstrate that the proposed framework is able to obtain discriminative feature representations, and correctly recover the intrinsic structures of various real-world datasets."
                    },
                    {
                        "title": "DM algorithms in health industry",
                        "abstract": "This survey reviews several approaches of data mining (DM) in healthindustry from many research groups world wide. The focus is on modern multi-core processors built into today's commodity computers, which are typically found at university institutes both as small server and workstation computers. So they are deliberately not high-performance computers. Modern multi-core processors consist of several (2 to over 100) computer cores, which work independently of each other according to the principle of \"multiple instruction multiple data\" (MIMD). They have a common main memory (shared memory). Each of these computer cores has several (2-16) arithmetic-logic units, which can simultaneously carry out the same arithmetic operation on several data in a vector-like manner (single instruction multiple data, SIMD). DM algorithms must use both types of parallelism (SIMD and MIMD), with access to the main memory (centralized component) being the main barrier to increased efficiency. This is important for DM in healthindustry applications like ECG, EEG, CT, SPECT, fMRI, DTI, ultrasound, microscopy, dermascopy, etc."
                    },
                    {
                        "title": "Fluctuation limits of the super-Brownian motion with a single point catalyst",
                        "abstract": "We prove a fluctuating limit theorem of a sequence of super-Brownian motions over $\\mbb{R}$ with a single point catalyst. The weak convergence of the processes on the space of Schwarz distributions is established. The limiting process is an Ornstein-Uhlenbeck type process solving a Langevin type equation driven by a one-dimensional Brownian motion."
                    },
                    {
                        "title": "Option pricing under jump diffusion model",
                        "abstract": "We provide an European option pricing formula written in the form of an infinite series of Black Scholes type terms under double Levy jumps model, where both the interest rate and underlying price are driven by Levy process. The series solution converges with a radius of convergence, and it is complemented by some numerical experiments to demonstrate its speed of convergence."
                    }
                ]
            }
        }
    },
    "2408.13256": {
        "paper_data": {
            "title": "How Diffusion Models Learn to Factorize and Compose",
            "url": "http://arxiv.org/abs/2408.13256v2",
            "arxiv_id": "2408.13256",
            "authors": [
                "Qiyao Liang",
                "Ziming Liu",
                "Mitchell Ostrow",
                "Ila Fiete"
            ],
            "abstract": "Diffusion models are capable of generating photo-realistic images that combine elements which likely do not appear together in the training set, demonstrating the ability to \\textit{compositionally generalize}. Nonetheless, the precise mechanism of compositionality and how it is acquired through training remains elusive. Inspired by cognitive neuroscientific approaches, we consider a highly reduced setting to examine whether and when diffusion models learn semantically meaningful and factorized representations of composable features. We performed extensive controlled experiments on conditional Denoising Diffusion Probabilistic Models (DDPMs) trained to generate various forms of 2D Gaussian bump images. We found that the models learn factorized but not fully continuous manifold representations for encoding continuous features of variation underlying the data. With such representations, models demonstrate superior feature compositionality but limited ability to interpolate over unseen values of a given feature. Our experimental results further demonstrate that diffusion models can attain compositionality with few compositional examples, suggesting a more efficient way to train DDPMs. Finally, we connect manifold formation in diffusion models to percolation theory in physics, offering insight into the sudden onset of factorized representation learning. Our thorough toy experiments thus contribute a deeper understanding of how diffusion models capture compositional structure in data.",
            "introduction": "   1 Introduction  Large-scale text-to-image generative models can produce photo-realistic synthetic images that combine elements in a novel fashion (compositional generalization). Nonetheless, the ability of models to do so, as well as their failure modes, are not well-studied systematically in large diffusion models due to the size of the model and the complex and high-dimensional nature of their training datasets. Factorization and compositional generalization have been theoretically and empirically investigated in many deep generative models before\u00a0Zhao et\u00a0al. (2018); Higgins et\u00a0al. (2017); Burgess et\u00a0al. (2018); Montero et\u00a0al. (2021); Xu et\u00a0al. (2022); Okawa et\u00a0al. (2023); Wiedemer et\u00a0al. (2023); Bowers et\u00a0al. (2016); Chaabouni et\u00a0al. (2020). However, these studies have not reached a unanimous conclusion on whether factorized representations learned in the intermediate layers of the model promote compositional generalization in the model performance. Specifically, several studies\u00a0Xu et\u00a0al. (2022); Montero et\u00a0al. (2021); Chaabouni et\u00a0al. (2020) have found little correlation between factorization and compositionality, contrary to others that suggest that factorization promotes compositionality\u00a0Bengio et\u00a0al. (2013); Higgins et\u00a0al. (2017); Burgess et\u00a0al. (2018); Duan et\u00a0al. (2020); Bowers et\u00a0al. (2016). As a result, there is no consensus on the exact mechanism of compositionality and how models gain the ability to compositionally generalize. However, these previous studies involve data with a complex mix of discrete and continuously varying features, making it difficult to explicitly analyze the model\u2019s learned representations beyond the disentanglement score, and thereby hindering a deeper understanding of factorization and compositionality.   Despite the remarkable compositional abilities demonstrated by diffusion-based text-to-image generative models, explicit exploration of factorization and compositional generalization of diffusion models has been relatively limited. In a recent empirical study of toy diffusion models, Okawa et. al.\u00a0Okawa et\u00a0al. (2023) shows that diffusion models learn to compositionally generalize multiplicatively. They too, however, did not focus on the mechanistic aspect of compositional generalization within the model or analyze the representations learned by the models. To investigate factorization and compositionality in diffusion models, we conduct extensive experiments on conditional denoising diffusion probablistic models (DDPMs) Ho et\u00a0al. (2020) trained with various 2D Gaussian datasets to characterize their learned representations and generalization performance.111A preliminary version of this work was presented at the ME-FoMo and BGPT workshops at ICLR 2024 Liang et\u00a0al. (2024).  Drawing inspiration from traditional cognitive psychology and neuroscience studies on animals and humans, we design straightforward cognitive tasks for models to perform while analyzing both their generation output (\u201cbehavior\u201d) and their intermediate layer representations (\u201cneural activities\u201d). Specifically, we evaluate the diffusion model\u2019s ability to generalize by generating 2D Gaussian bumps centered at specified x,y\ud835\udc65\ud835\udc66x,yitalic_x , italic_y coordinates on a finite canvas. We deliberately chose a low-dimensional dataset and a simplified experimental setup to enable a clear analysis of how the model\u2019s representation factorization relates to its generalization capabilities.   Conceptually, the models can adopt different parameterizations of the x,y\ud835\udc65\ud835\udc66x,yitalic_x , italic_y coordinates for the Gaussian bumps, each with varying implications for data efficiency. Consider a K\ud835\udc3eKitalic_K-dimensional dataset created from the composition of multiple independently varying 1-dimensional (1D) latent features (e.g. a factorizable data distribution P\u2062(x1,x2,\u2026,xK)=P\u2062(x1)\u2062P\u2062(x2)\u2062\u2026\u2062P\u2062(xK)\ud835\udc43subscript\ud835\udc651subscript\ud835\udc652\u2026subscript\ud835\udc65\ud835\udc3e\ud835\udc43subscript\ud835\udc651\ud835\udc43subscript\ud835\udc652\u2026\ud835\udc43subscript\ud835\udc65\ud835\udc3eP(x_{1},x_{2},\\ldots,x_{K})=P(x_{1})P(x_{2})\\ldots P(x_{K})italic_P ( italic_x start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , italic_x start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT , \u2026 , italic_x start_POSTSUBSCRIPT italic_K end_POSTSUBSCRIPT )",
            "references": [
                {
                    "title": "Factorized visual representations in the primate visual system and deep neural networks",
                    "abstract": "Object classification has been proposed as a principal objective of the primate ventral visual stream and has been used as an optimization target for deep neural network models (DNNs) of the visual system. However, visual brain areas represent many different types of information, and optimizing for classification of object identity alone does not constrain how other information may be encoded in visual representations. Information about different scene parameters may be discarded altogether (\u201cinvariance\u201d), represented in non-interfering subspaces of population activity (\u201cfactorization\u201d) or encoded in an entangled fashion. In this work, we provide evidence that factorization is a normative principle of biological visual representations. In the monkey ventral visual hierarchy, we found that factorization of object pose and background information from object identity increased in higher-level regions and strongly contributed to improving object identity decoding performance. We then conducted a large-scale analysis of factorization of individual scene parameters \u2013 lighting, background, camera viewpoint, and object pose \u2013 in a diverse library of DNN models of the visual system. Models which best matched neural, fMRI and behavioral data from both monkeys and humans across 12 datasets tended to be those which factorized scene parameters most strongly. Notably, invariance to these parameters was not as consistently associated with matches to neural and behavioral data, suggesting that maintaining non-class information in factorized activity subspaces is often preferred to dropping it altogether. Thus, we propose that factorization of visual scene information is a widely used strategy in brains and DNN models thereof."
                },
                {
                    "title": "Do Diffusion Models Learn Semantically Meaningful and Efficient Representations?",
                    "abstract": "Diffusion models are capable of impressive feats of image generation with uncommon juxtapositions such as astronauts riding horses on the moon with properly placed shadows. These outputs indicate the ability to perform compositional generalization, but how do the models do so? We perform controlled experiments on conditional DDPMs learning to generate 2D spherical Gaussian bumps centered at specified $x$- and $y$-positions. Our results show that the emergence of semantically meaningful latent representations is key to achieving high performance. En route to successful performance over learning, the model traverses three distinct phases of latent representations: (phase A) no latent structure, (phase B) a 2D manifold of disordered states, and (phase C) a 2D ordered manifold. Corresponding to each of these phases, we identify qualitatively different generation behaviors: 1) multiple bumps are generated, 2) one bump is generated but at inaccurate $x$ and $y$ locations, 3) a bump is generated at the correct $x$ and y location. Furthermore, we show that even under imbalanced datasets where features ($x$- versus $y$-positions) are represented with skewed frequencies, the learning process for $x$ and $y$ is coupled rather than factorized, demonstrating that simple vanilla-flavored diffusion models cannot learn efficient representations in which localization in $x$ and $y$ are factorized into separate 1D tasks. These findings suggest the need for future work to find inductive biases that will push generative models to discover and exploit factorizable independent structures in their inputs, which will be required to vault these models into more data-efficient regimes."
                },
                {
                    "title": "Compositional Abilities Emerge Multiplicatively: Exploring Diffusion Models on a Synthetic Task",
                    "abstract": "Modern generative models exhibit unprecedented capabilities to generate extremely realistic data. However, given the inherent compositionality of the real world, reliable use of these models in practical applications requires that they exhibit the capability to compose a novel set of concepts to generate outputs not seen in the training data set. Prior work demonstrates that recent diffusion models do exhibit intriguing compositional generalization abilities, but also fail unpredictably. Motivated by this, we perform a controlled study for understanding compositional generalization in conditional diffusion models in a synthetic setting, varying different attributes of the training data and measuring the model's ability to generate samples out-of-distribution. Our results show: (i) the order in which the ability to generate samples from a concept and compose them emerges is governed by the structure of the underlying data-generating process; (ii) performance on compositional tasks exhibits a sudden\"emergence\"due to multiplicative reliance on the performance of constituent tasks, partially explaining emergent phenomena seen in generative models; and (iii) composing concepts with lower frequency in the training data to generate out-of-distribution samples requires considerably more optimization steps compared to generating in-distribution samples. Overall, our study lays a foundation for understanding capabilities and compositionality in generative models from a data-centric perspective."
                },
                {
                    "title": "Compositional Generalization from First Principles",
                    "abstract": "Leveraging the compositional nature of our world to expedite learning and facilitate generalization is a hallmark of human perception. In machine learning, on the other hand, achieving compositional generalization has proven to be an elusive goal, even for models with explicit compositional priors. To get a better handle on compositional generalization, we here approach it from the bottom up: Inspired by identifiable representation learning, we investigate compositionality as a property of the data-generating process rather than the data itself. This reformulation enables us to derive mild conditions on only the support of the training distribution and the model architecture, which are sufficient for compositional generalization. We further demonstrate how our theoretical framework applies to real-world scenarios and validate our findings empirically. Our results set the stage for a principled theoretical study of compositional generalization."
                },
                {
                    "title": "Compositional Generalization in Unsupervised Compositional Representation Learning: A Study on Disentanglement and Emergent Language",
                    "abstract": "Deep learning models struggle with compositional generalization, i.e. the ability to recognize or generate novel combinations of observed elementary concepts. In hopes of enabling compositional generalization, various unsupervised learning algorithms have been proposed with inductive biases that aim to induce compositional structure in learned representations (e.g. disentangled representation and emergent language learning). In this work, we evaluate these unsupervised learning algorithms in terms of how well they enable compositional generalization. Specifically, our evaluation protocol focuses on whether or not it is easy to train a simple model on top of the learned representation that generalizes to new combinations of compositional factors. We systematically study three unsupervised representation learning algorithms - $\\beta$-VAE, $\\beta$-TCVAE, and emergent language (EL) autoencoders - on two datasets that allow directly testing compositional generalization. We find that directly using the bottleneck representation with simple models and few labels may lead to worse generalization than using representations from layers before or after the learned representation itself. In addition, we find that the previously proposed metrics for evaluating the levels of compositionality are not correlated with actual compositional generalization in our framework. Surprisingly, we find that increasing pressure to produce a disentangled representation produces representations with worse generalization, while representations from EL models show strong compositional generalization. Taken together, our results shed new light on the compositional generalization behavior of different unsupervised learning algorithms with a new setting to rigorously test this behavior, and suggest the potential benefits of delevoping EL learning algorithms for more generalizable representations."
                },
                {
                    "title": "Recurrent neural network models for working memory of continuous variables: activity manifolds, connectivity patterns, and dynamic codes",
                    "abstract": "Many daily activities and psychophysical experiments involve keeping multiple items in working memory. When items take continuous values (e.g., orientation, contrast, length, loudness) they must be stored in a continuous structure of appropriate dimensions. We investigate how this structure is represented in neural circuits by training recurrent networks to report two previously shown stimulus orientations. We find the activity manifold for the two orientations resembles a Clifford torus. Although a Clifford and standard torus (the surface of a donut) are topologically equivalent, they have important functional differences. A Clifford torus treats the two orientations equally and keeps them in orthogonal subspaces, as demanded by the task, whereas a standard torus does not. We find and characterize the connectivity patterns that support the Clifford torus. Moreover, in addition to attractors that store information via persistent activity, our networks also use a dynamic code where units change their tuning to prevent new sensory input from overwriting the previously stored one. We argue that such dynamic codes are generally required whenever multiple inputs enter a memory system via shared connections. Finally, we apply our framework to a human psychophysics experiment in which subjects reported two remembered orientations. By varying the training conditions of the RNNs, we test and support the hypothesis that human behavior is a product of both neural noise and reliance on the more stable and behaviorally relevant memory of the ordinal relationship between the two orientations. This suggests that suitable inductive biases in RNNs are important for uncovering how the human brain implements working memory. Together, these results offer an understanding of the neural computations underlying a class of visual decoding tasks, bridging the scales from human behavior to synaptic connectivity."
                },
                {
                    "title": "Image Super-Resolution via Iterative Refinement",
                    "abstract": "We present SR3, an approach to image Super-Resolution via Repeated Refinement. SR3 adapts denoising diffusion probabilistic models (Ho et al. 2020), (Sohl-Dickstein et al. 2015) to image-to-image translation, and performs super-resolution through a stochastic iterative denoising process. Output images are initialized with pure Gaussian noise and iteratively refined using a U-Net architecture that is trained on denoising at various noise levels, conditioned on a low-resolution input image. SR3 exhibits strong performance on super-resolution tasks at different magnification factors, on faces and natural images. We conduct human evaluation on a standard 8\u00d7 face super-resolution task on CelebA-HQ for which SR3 achieves a fool rate close to 50%, suggesting photo-realistic outputs, while GAN baselines do not exceed a fool rate of 34%. We evaluate SR3 on a 4\u00d7 super-resolution task on ImageNet, where SR3 outperforms baselines in human evaluation and classification accuracy of a ResNet-50 classifier trained on high-resolution images. We further show the effectiveness of SR3 in cascaded image generation, where a generative model is chained with super-resolution models to synthesize high-resolution images with competitive FID scores on the class-conditional 256\u00d7256 ImageNet generation challenge."
                },
                {
                    "title": "WaveGrad: Estimating Gradients for Waveform Generation",
                    "abstract": "This paper introduces WaveGrad, a conditional model for waveform generation which estimates gradients of the data density. The model is built on prior work on score matching and diffusion probabilistic models. It starts from a Gaussian white noise signal and iteratively refines the signal via a gradient-based sampler conditioned on the mel-spectrogram. WaveGrad offers a natural way to trade inference speed for sample quality by adjusting the number of refinement steps, and bridges the gap between non-autoregressive and autoregressive models in terms of audio quality. We find that it can generate high fidelity audio samples using as few as six iterations. Experiments reveal WaveGrad to generate high fidelity audio, outperforming adversarial non-autoregressive baselines and matching a strong likelihood-based autoregressive baseline using fewer sequential operations. Audio samples are available at this https URL."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Compositionality and Generalization In Emergent Languages",
                    "abstract": "Natural language allows us to refer to novel composite concepts by combining expressions denoting their parts according to systematic rules, a property known as compositionality. In this paper, we study whether the language emerging in deep multi-agent simulations possesses a similar ability to refer to novel primitive combinations, and whether it accomplishes this feat by strategies akin to human-language compositionality. Equipped with new ways to measure compositionality in emergent languages inspired by disentanglement in representation learning, we establish three main results: First, given sufficiently large input spaces, the emergent language will naturally develop the ability to refer to novel composite concepts. Second, there is no correlation between the degree of compositionality of an emergent language and its ability to generalize. Third, while compositionality is not necessary for generalization, it provides an advantage in terms of language transmission: The more compositional a language is, the more easily it will be picked up by new learners, even when the latter differ in architecture from the original agents. We conclude that compositionality does not arise from simple generalization pressure, but if an emergent language does chance upon it, it will be more likely to survive and thrive."
                },
                {
                    "title": "Unsupervised Model Selection for Variational Disentangled Representation Learning",
                    "abstract": "Disentangled representations have recently been shown to improve fairness, data efficiency and generalisation in simple supervised and reinforcement learning tasks. To extend the benefits of disentangled representations to more complex domains and practical applications, it is important to enable hyperparameter tuning and model selection of existing unsupervised approaches without requiring access to ground truth attribute labels, which are not available for most datasets. This paper addresses this problem by introducing a simple yet robust and reliable method for unsupervised disentangled model selection. Our approach, Unsupervised Disentanglement Ranking (UDR), leverages the recent theoretical results that explain why variational autoencoders disentangle (Rolinek et al, 2019), to quantify the quality of disentanglement by performing pairwise comparisons between trained model representations. We show that our approach performs comparably to the existing supervised alternatives across 5,400 models from six state of the art unsupervised disentangled representation learning model classes. Furthermore, we show that the ranking produced by our approach correlates well with the final task performance on two different domains."
                },
                {
                    "title": "UMAP: Uniform Manifold Approximation and Projection",
                    "abstract": "Uniform Manifold Approximation and Projection (UMAP) is a dimension reduction technique that can be used for visualisation similarly to t-SNE, but also for general non-linear dimension reduction. UMAP has a rigorous mathematical foundation, but is simple to use, with a scikit-learn compatible API. UMAP is among the fastest manifold learning implementations available \u2013 significantly faster than most t-SNE implementations."
                },
                {
                    "title": "Understanding disentangling in \u03b2-VAE",
                    "abstract": "We present new intuitions and theoretical assessments of the emergence of disentangled representation in variational autoencoders. Taking a rate-distortion theory perspective, we show the circumstances under which representations aligned with the underlying generative factors of variation of data emerge when optimising the modified ELBO bound in $\\beta$-VAE, as training progresses. From these insights, we propose a modification to the training regime of $\\beta$-VAE, that progressively increases the information capacity of the latent code during training. This modification facilitates the robust learning of disentangled representations in $\\beta$-VAE, without the previous trade-off in reconstruction accuracy."
                },
                {
                    "title": "UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction",
                    "abstract": "UMAP (Uniform Manifold Approximation and Projection) is a novel manifold learning technique for dimension reduction. UMAP is constructed from a theoretical framework based in Riemannian geometry and algebraic topology. The result is a practical scalable algorithm that applies to real world data. The UMAP algorithm is competitive with t-SNE for visualization quality, and arguably preserves more of the global structure with superior run time performance. Furthermore, UMAP has no computational restrictions on embedding dimension, making it viable as a general purpose dimension reduction technique for machine learning."
                },
                {
                    "title": "A theory of multineuronal dimensionality, dynamics and measurement",
                    "abstract": "In many experiments, neuroscientists tightly control behavior, record many trials, and obtain trial-averaged firing rates from hundreds of neurons in circuits containing billions of behaviorally relevant neurons. Di-mensionality reduction methods reveal a striking simplicity underlying such multi-neuronal data: they can be reduced to a low-dimensional space, and the resulting neural trajectories in this space yield a remarkably insightful dynamical portrait of circuit computation. This simplicity raises profound and timely conceptual questions. What are its origins and its implications for the complexity of neural dynamics? How would the situation change if we recorded more neurons? When, if at all, can we trust dynamical portraits obtained from measuring an infinitesimal fraction of task relevant neurons? We present a theory that answers these questions, and test it using physiological recordings from reaching monkeys. This theory reveals conceptual insights into how task complexity governs both neural dimensionality and accurate recovery of dynamic portraits, thereby providing quantitative guidelines for future large-scale experimental design."
                },
                {
                    "title": "beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework",
                    "abstract": "an"
                },
                {
                    "title": "Place cells, grid cells, and memory.",
                    "abstract": "The hippocampal system is critical for storage and retrieval of declarative memories, including memories for locations and events that take place at those locations. Spatial memories place high demands on capacity. Memories must be distinct to be recalled without interference and encoding must be fast. Recent studies have indicated that hippocampal networks allow for fast storage of large quantities of uncorrelated spatial information. The aim of the this article is to review and discuss some of this work, taking as a starting point the discovery of multiple functionally specialized cell types of the hippocampal-entorhinal circuit, such as place, grid, and border cells. We will show that grid cells provide the hippocampus with a metric, as well as a putative mechanism for decorrelation of representations, that the formation of environment-specific place maps depends on mechanisms for long-term plasticity in the hippocampus, and that long-term spatiotemporal memory storage may depend on offline consolidation processes related to sharp-wave ripple activity in the hippocampus. The multitude of representations generated through interactions between a variety of functionally specialized cell types in the entorhinal-hippocampal circuit may be at the heart of the mechanism for declarative memory formation."
                },
                {
                    "title": "Representation Learning: A Review and New Perspectives",
                    "abstract": "The success of machine learning algorithms generally depends on data representation, and we hypothesize that this is because different representations can entangle and hide more or less the different explanatory factors of variation behind the data. Although specific domain knowledge can be used to help design representations, learning with generic priors can also be used, and the quest for AI is motivating the design of more powerful representation-learning algorithms implementing such priors. This paper reviews recent work in the area of unsupervised feature learning and deep learning, covering advances in probabilistic models, autoencoders, manifold learning, and deep networks. This motivates longer term unanswered questions about the appropriate objectives for learning good representations, for computing representations (i.e., inference), and the geometrical connections between representation learning, density estimation, and manifold learning."
                },
                {
                    "title": "Accurate Path Integration in Continuous Attractor Network Models of Grid Cells",
                    "abstract": "Grid cells in the rat entorhinal cortex display strikingly regular firing responses to the animal's position in 2-D space and have been hypothesized to form the neural substrate for dead-reckoning. However, errors accumulate rapidly when velocity inputs are integrated in existing models of grid cell activity. To produce grid-cell-like responses, these models would require frequent resets triggered by external sensory cues. Such inadequacies, shared by various models, cast doubt on the dead-reckoning potential of the grid cell system. Here we focus on the question of accurate path integration, specifically in continuous attractor models of grid cell activity. We show, in contrast to previous models, that continuous attractor models can generate regular triangular grid responses, based on inputs that encode only the rat's velocity and heading direction. We consider the role of the network boundary in the integration performance of the network and show that both periodic and aperiodic networks are capable of accurate path integration, despite important differences in their attractor manifolds. We quantify the rate at which errors in the velocity integration accumulate as a function of network size and intrinsic noise within the network. With a plausible range of parameters and the inclusion of spike variability, our model networks can accurately integrate velocity inputs over a maximum of \u223c10\u2013100 meters and \u223c1\u201310 minutes. These findings form a proof-of-concept that continuous attractor dynamics may underlie velocity integration in the dorsolateral medial entorhinal cortex. The simulations also generate pertinent upper bounds on the accuracy of integration that may be achieved by continuous attractor dynamics in the grid cell network. We suggest experiments to test the continuous attractor model and differentiate it from models in which single cells establish their responses independently of each other."
                },
                {
                    "title": "Place cells, grid cells, and the brain's spatial representation system.",
                    "abstract": "More than three decades of research have demonstrated a role for hippocampal place cells in representation of the spatial environment in the brain. New studies have shown that place cells are part of a broader circuit for dynamic representation of self-location. A key component of this network is the entorhinal grid cells, which, by virtue of their tessellating firing fields, may provide the elements of a path integration-based neural map. Here we review how place cells and grid cells may form the basis for quantitative spatiotemporal representation of places, routes, and associated experiences during behavior and in memory. Because these cell types have some of the most conspicuous behavioral correlates among neurons in nonsensory cortical systems, and because their spatial firing structure reflects computations internally in the system, studies of entorhinal-hippocampal representations may offer considerable insight into general principles of cortical network dynamics."
                },
                {
                    "title": "The head direction signal: origins and sensory-motor integration.",
                    "abstract": "Navigation first requires accurate perception of one's spatial orientation within the environment, which consists of knowledge about location and directional heading. Cells within several limbic system areas of the mammalian brain discharge allocentrically as a function of the animal's directional heading, independent of the animal's location and ongoing behavior. These cells are referred to as head direction (HD) cells and are believed to encode the animal's perceived directional heading with respect to its environment. Although HD cells are found in several areas, the principal circuit for generating this signal originates in the dorsal tegmental nucleus and projects serially, with some reciprocal connections, to the lateral mammillary nucleus --> anterodorsal thalamus --> PoS, and terminates in the entorhinal cortex. HD cells receive multimodal information about landmarks and self-generated movements. Vestibular information appears critical for generating the directional signal, but motor/proprioceptive and landmark information are important for updating it."
                },
                {
                    "title": "Head-direction cells recorded from the postsubiculum in freely moving rats. I. Description and quantitative analysis",
                    "abstract": "This paper is a study of the behavioral and spatial firing correlates of neurons in the rat postsubiculum. Recordings were made from postsubicular neurons as rats moved freely throughout a cylindrical chamber, where the major cue for orientation was a white card taped to the inside wall. An automatic video/computer system monitored cell discharge while simultaneously tracking the position of 2 colored light emitting diodes (LEDs) secured to the animal's head. The animal's location was calculated from the position of one of the LEDs and head direction in the horizontal plane calculated from the relative positions of the 2 LEDs. Approximately 26% of the cells were classified as head-direction cells because they discharged as a function of the animal's head direction in the horizontal plane, independent of the animal's behavior, location, or trunk position. For each head-direction cell, vectors drawn in the direction of maximal firing were parallel throughout the recording chamber and did not converge toward a single point. Plots of firing rate versus head direction showed that each firing-rate/head-direction function was adequately described by a triangular function. Each cell's maximum firing rate occurred at only one (the preferred) head direction; firing rates at head directions on either side of the preferred direction decreased linearly with angular deviation from the preferred direction. Results from 24 head-direction cells in 7 animals showed an equal distribution of preferred firing directions over a 360 degrees angle. The peak firing rate of head- direction cells varied from 5 to 115 spikes/sec (mean: 35). The range of head-direction angles over which discharge was elevated (directional firing range) was usually about 90 degrees, with little, if any, discharge at head directions outside this range. Quantitative analysis showed the location of the animal within the cylinder had minimal effect on directional cell firing. For each head-direction cell, the preferred direction, peak firing rate, and directional firing range remained stable for days. These results identify a new cell type that signals the animal's head direction in its environment."
                }
            ],
            "categories": [
                "cs.AI",
                "cs.CV",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow do factorization and compositional generalization interact in large-scale diffusion models, and what mechanisms underlie their ability to generate novel compositions?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses the fundamental understanding of how generative models learn and represent complex data. A clearer understanding of the relationship between factorization and compositional generalization could lead to improved model architectures and training methodologies, ultimately enhancing the performance of generative models in practical applications such as image synthesis, creative content generation, and human-computer interaction. This research could pave the way for future studies that explore more complex datasets and tasks, thereby advancing the field of machine learning.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the high-dimensional nature of the training datasets and the complexity of the models involved. Naive approaches may fail because they do not adequately account for the intricate relationships between learned representations and compositionality. Additionally, the lack of consensus in previous studies indicates that the mechanisms of compositionality are not straightforward and may involve nuanced interactions that are difficult to isolate and analyze. Technical obstacles include the need for effective evaluation metrics that can capture the subtleties of representation factorization and generalization performance.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has been limited by the complexity of the datasets used, which often contain a mix of discrete and continuous features, making it challenging to analyze learned representations beyond basic metrics. Additionally, earlier studies have not focused on the mechanistic aspects of compositional generalization within diffusion models. Barriers such as the scale of the models and the lack of targeted experimental designs have prevented a thorough investigation. Our approach differs by utilizing a simplified, low-dimensional dataset and cognitive tasks that allow for a clearer analysis of the relationship between representation factorization and generalization capabilities.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves conducting extensive experiments on conditional denoising diffusion probabilistic models (DDPMs) trained with various 2D Gaussian datasets. We will evaluate the models' ability to generate Gaussian bumps at specified coordinates on a finite canvas, analyzing both the generation outputs and the intermediate layer representations. The metrics used will include representation factorization scores and generalization performance measures. We expect to uncover insights into how different parameterizations of the coordinates affect data efficiency and the models' compositional"
            }
        },
        "author_data": {
            "6106229f-8936-46f7-a42d-753de59dbac6": {
                "pk": "6106229f-8936-46f7-a42d-753de59dbac6",
                "name": "Qiyao Liang",
                "collaborators": [
                    "Mingyu Kang",
                    "Ming Li",
                    "Yunseong Nam",
                    "Bichen Zhang",
                    "Shilin Huang",
                    "Ye Wang",
                    "Chao Fang",
                    "Jungsang Kim",
                    "Kenneth R. Brown",
                    "Ziming Liu"
                ],
                "domain": [
                    "Quantum Computing",
                    "Trapped Ions",
                    "Quantum Algorithms",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Pulse optimization for high-precision motional-mode characterization in trapped-ion quantum computers",
                        "abstract": "High-fidelity operation of quantum computers requires precise knowledge of the physical system through characterization. For motion-mediated entanglement generation in trapped ions, it is crucial to have precise knowledge of the motional-mode parameters such as the mode frequencies and the Lamb-Dicke parameters. Unfortunately, the state-of-the-art mode-characterization schemes do not easily render the mode parameters in a sufficiently scalable and accurate fashion, due to the unwanted excitation of adjacent modes in the frequency space when targeting a single mode, an effect known as the \\textit{cross-mode coupling}. Here, we develop an alternative scheme that leverages the degrees of freedom in pulse design for the characterization experiment such that the effects of the cross-mode coupling is actively silenced. Further, we devise stabilization methods to accurately characterize the Lamb-Dicke parameters even when the mode frequencies are not precisely known due to experimental drifts or characterization inaccuracies. We extensively benchmark our scheme in simulations of a three-ion chain and discuss the parameter regimes in which the shaped pulses significantly outperform the traditional square pulses."
                    },
                    {
                        "title": "Do Diffusion Models Learn Semantically Meaningful and Efficient Representations?",
                        "abstract": "Diffusion models are capable of impressive feats of image generation with uncommon juxtapositions such as astronauts riding horses on the moon with properly placed shadows. These outputs indicate the ability to perform compositional generalization, but how do the models do so? We perform controlled experiments on conditional DDPMs learning to generate 2D spherical Gaussian bumps centered at specified $x$- and $y$-positions. Our results show that the emergence of semantically meaningful latent representations is key to achieving high performance. En route to successful performance over learning, the model traverses three distinct phases of latent representations: (phase A) no latent structure, (phase B) a 2D manifold of disordered states, and (phase C) a 2D ordered manifold. Corresponding to each of these phases, we identify qualitatively different generation behaviors: 1) multiple bumps are generated, 2) one bump is generated but at inaccurate $x$ and $y$ locations, 3) a bump is generated at the correct $x$ and y location. Furthermore, we show that even under imbalanced datasets where features ($x$- versus $y$-positions) are represented with skewed frequencies, the learning process for $x$ and $y$ is coupled rather than factorized, demonstrating that simple vanilla-flavored diffusion models cannot learn efficient representations in which localization in $x$ and $y$ are factorized into separate 1D tasks. These findings suggest the need for future work to find inductive biases that will push generative models to discover and exploit factorizable independent structures in their inputs, which will be required to vault these models into more data-efficient regimes."
                    },
                    {
                        "title": "Efficient motional-mode characterization for high-fidelity trapped-ion quantum computing",
                        "abstract": "To achieve high-fidelity operations on a large-scale quantum computer, the parameters of the physical system must be efficiently characterized with high accuracy. For trapped ions, the entanglement between qubits are mediated by the motional modes of the ion chain, and thus characterizing the motional-mode parameters becomes essential. In this paper, we develop and explore physical models that accurately predict both magnitude and sign of the Lamb-Dicke parameters when the modes are probed {\\it in parallel}. We further devise an advanced characterization protocol that shortens the characterization time by more than an order of magnitude, when compared to that of the conventional method that only uses mode spectroscopy. We discuss potential ramifications of our results to the development of a scalable trapped-ion quantum computer, viewed through the lens of system-level resource trade offs."
                    },
                    {
                        "title": "Modeling the Performance of Early Fault-Tolerant Quantum Algorithms",
                        "abstract": "Progress in fault-tolerant quantum computation (FTQC) has driven the pursuit of practical applications with early fault-tolerant quantum computers (EFTQC). These devices, limited in their qubit counts and fault-tolerance capabilities, require algorithms that can accommodate some degrees of error, which are known as EFTQC algorithms. To predict the onset of early quantum advantage, a comprehensive methodology is needed to develop and analyze EFTQC algorithms, drawing insights from both the methodologies of noisy intermediate-scale quantum (NISQ) and traditional FTQC. To address this need, we propose such a methodology for modeling algorithm performance on EFTQC devices under varying degrees of error. As a case study, we apply our methodology to analyze the performance of Randomized Fourier Estimation (RFE), an EFTQC algorithm for phase estimation. We investigate the runtime performance and the fault-tolerant overhead of RFE in comparison to the traditional quantum phase estimation algorithm. Our analysis reveals that RFE achieves significant savings in physical qubit counts while having a much higher runtime upper bound. We anticipate even greater physical qubit savings when considering more realistic assumptions about the performance of EFTQC devices. By providing insights into the performance trade-offs and resource requirements of EFTQC algorithms, our work contributes to the development of practical and efficient quantum computing solutions on the path to quantum advantage."
                    },
                    {
                        "title": "Batch Optimization of Frequency-Modulated Pulses for Robust Two-qubit Gates in Ion Chains",
                        "abstract": "Two-qubit gates in trapped-ion quantum computers are generated by applying spin-dependent forces that temporarily entangle the internal state of the ion with its motion. Laser pulses are carefully designed to generate a maximally entangling gate between the ions while minimizing any residual entanglement between the motion and the ion. The quality of the gates suffers when the actual experimental parameters differ from the ideal case. Here, we improve the robustness of frequency-modulated M{\\o}lmer-S{\\o}rensen gates to motional mode-frequency offsets by optimizing the average performance over a range of systematic errors using batch optimization. We then compare this method with frequency-modulated gates optimized for ideal parameters that include an analytic robustness condition. Numerical simulations show good performance up to 12 ions, and the method is experimentally demonstrated on a two-ion chain."
                    },
                    {
                        "title": "High-fidelity Two-qubit Gates Using a MEMS-based Beam Steering System for Individual Qubit Addressing",
                        "abstract": "In a large scale trapped atomic ion quantum computer, high-fidelity two-qubit gates need to be extended over all qubits with individual control. We realize and characterize high-fidelity two-qubit gates in a system with up to 4 ions using radial modes. The ions are individually addressed by two tightly focused beams steered using micro-electromechanical system (MEMS) mirrors. We deduce a gate fidelity of 99.49(7)% in a two-ion chain and 99.30(6)% in a four-ion chain by applying a sequence of up to 21 two-qubit gates and measuring the final state fidelity. We characterize the residual errors and discuss methods to further improve the gate fidelity towards values that are compatible with fault-tolerant quantum computation."
                    }
                ]
            },
            "c0739528-6cd3-42c8-93c1-53ffef59b3d6": {
                "pk": "c0739528-6cd3-42c8-93c1-53ffef59b3d6",
                "name": "Ziming Liu",
                "collaborators": [
                    "Max Tegmark",
                    "Eric J. Michaud",
                    "Shenggan Cheng",
                    "Yang You",
                    "Ziqian Zhong",
                    "Xiaobo Liu",
                    "Zheng Zhang",
                    "Varun Madhavan",
                    "Andrew M. Stuart",
                    "Yixuan Wang"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Statistical Physics",
                    "Dynamical Systems"
                ],
                "publications": [
                    {
                        "title": "Multi-PCA based Fault Detection Model Combined with Prior knowledge of HVAC",
                        "abstract": "The traditional PCA fault detection methods completely depend on the training data. The prior knowledge such as the physical principle of the system has not been taken into account. In this paper, we propose a new multi-PCA fault detection model combined with prior knowledge. This new model can adapt to the variable operating conditions of the central air conditioning system, and it can detect small deviation faults of sensors and significantly shorten the time delay of detecting drift faults. We also conducted enough ablation experiments to demonstrate that our model is more robust and efficient."
                    },
                    {
                        "title": "A Neural Scaling Law from Lottery Ticket Ensembling",
                        "abstract": "Neural scaling laws (NSL) refer to the phenomenon where model performance improves with scale. Sharma & Kaplan analyzed NSL using approximation theory and predict that MSE losses decay as $N^{-\\alpha}$, $\\alpha=4/d$, where $N$ is the number of model parameters, and $d$ is the intrinsic input dimension. Although their theory works well for some cases (e.g., ReLU networks), we surprisingly find that a simple 1D problem $y=x^2$ manifests a different scaling law ($\\alpha=1$) from their predictions ($\\alpha=4$). We opened the neural networks and found that the new scaling law originates from lottery ticket ensembling: a wider network on average has more \"lottery tickets\", which are ensembled to reduce the variance of outputs. We support the ensembling mechanism by mechanistically interpreting single neural networks, as well as studying them statistically. We attribute the $N^{-1}$ scaling law to the \"central limit theorem\" of lottery tickets. Finally, we discuss its potential implications for large language models and statistical physics-type theories of learning."
                    },
                    {
                        "title": "AI Poincar\u00e9: Machine Learning Conservation Laws from Trajectories",
                        "abstract": "We present AI Poincar\\'e, a machine learning algorithm for auto-discovering conserved quantities using trajectory data from unknown dynamical systems. We test it on five Hamiltonian systems, including the gravitational 3-body problem, and find that it discovers not only all exactly conserved quantities, but also periodic orbits, phase transitions and breakdown timescales for approximate conservation laws."
                    },
                    {
                        "title": "Quantum-Inspired Hamiltonian Monte Carlo for Bayesian Sampling",
                        "abstract": "Hamiltonian Monte Carlo (HMC) is an efficient Bayesian sampling method that can make distant proposals in the parameter space by simulating a Hamiltonian dynamical system. Despite its popularity in machine learning and data science, HMC is inefficient to sample from spiky and multimodal distributions. Motivated by the energy-time uncertainty relation from quantum mechanics, we propose a Quantum-Inspired Hamiltonian Monte Carlo algorithm (QHMC). This algorithm allows a particle to have a random mass matrix with a probability distribution rather than a fixed mass. We prove the convergence property of QHMC and further show why such a random mass can improve the performance when we sample a broad class of distributions. In order to handle the big training data sets in large-scale machine learning, we develop a stochastic gradient version of QHMC using Nos{\\'e}-Hoover thermostat called QSGNHT, and we also provide theoretical justifications about its steady-state distributions. Finally in the experiments, we demonstrate the effectiveness of QHMC and QSGNHT on synthetic examples, bridge regression, image denoising and neural network pruning. The proposed QHMC and QSGNHT can indeed achieve much more stable and accurate sampling results on the test cases."
                    },
                    {
                        "title": "Machine-learning hidden symmetries",
                        "abstract": "We present an automated method for finding hidden symmetries, defined as symmetries that become manifest only in a new coordinate system that must be discovered. Its core idea is to quantify asymmetry as violation of certain partial differential equations, and to numerically minimize such violation over the space of all invertible transformations, parametrized as invertible neural networks. For example, our method rediscovers the famous Gullstrand-Painleve metric that manifests hidden translational symmetry in the Schwarzschild metric of non-rotating black holes, as well as Hamiltonicity, modularity and other simplifying traits not traditionally viewed as symmetries."
                    },
                    {
                        "title": "AI Poincar\u00e9 2.0: Machine Learning Conservation Laws from Differential Equations",
                        "abstract": "We present a machine learning algorithm that discovers conservation laws from differential equations, both numerically (parametrized as neural networks) and symbolically, ensuring their functional independence (a non-linear generalization of linear independence). Our independence module can be viewed as a nonlinear generalization of singular value decomposition. Our method can readily handle inductive biases for conservation laws. We validate it with examples including the 3-body problem, the KdV equation and nonlinear Schr\\\"odinger equation."
                    },
                    {
                        "title": "Second Order Ensemble Langevin Method for Sampling and Inverse Problems",
                        "abstract": "We propose a sampling method based on an ensemble approximation of second order Langevin dynamics. The log target density is appended with a quadratic term in an auxiliary momentum variable and damped-driven Hamiltonian dynamics introduced; the resulting stochastic differential equation is invariant to the Gibbs measure, with marginal on the position coordinates given by the target. A preconditioner based on covariance under the law of the dynamics does not change this invariance property, and is introduced to accelerate convergence to the Gibbs measure. The resulting mean-field dynamics may be approximated by an ensemble method; this results in a gradient-free and affine-invariant stochastic dynamical system. Numerical results demonstrate its potential as the basis for a numerical sampler in Bayesian inverse problems."
                    },
                    {
                        "title": "Omnigrok: Grokking Beyond Algorithmic Data",
                        "abstract": "Grokking, the unusual phenomenon for algorithmic datasets where generalization happens long after overfitting the training data, has remained elusive. We aim to understand grokking by analyzing the loss landscapes of neural networks, identifying the mismatch between training and test losses as the cause for grokking. We refer to this as the \"LU mechanism\" because training and test losses (against model weight norm) typically resemble \"L\" and \"U\", respectively. This simple mechanism can nicely explain many aspects of grokking: data size dependence, weight decay dependence, the emergence of representations, etc. Guided by the intuitive picture, we are able to induce grokking on tasks involving images, language and molecules. In the reverse direction, we are able to eliminate grokking for algorithmic datasets. We attribute the dramatic nature of grokking for algorithmic datasets to representation learning."
                    },
                    {
                        "title": "Seeing is Believing: Brain-Inspired Modular Training for Mechanistic Interpretability",
                        "abstract": "We introduce Brain-Inspired Modular Training (BIMT), a method for making neural networks more modular and interpretable. Inspired by brains, BIMT embeds neurons in a geometric space and augments the loss function with a cost proportional to the length of each neuron connection. We demonstrate that BIMT discovers useful modular neural networks for many simple tasks, revealing compositional structures in symbolic formulas, interpretable decision boundaries and features for classification, and mathematical structure in algorithmic datasets. The ability to directly see modules with the naked eye can complement current mechanistic interpretability strategies such as probes, interventions or staring at all weights."
                    },
                    {
                        "title": "Hanayo: Harnessing Wave-like Pipeline Parallelism for Enhanced Large Model Training Efficiency",
                        "abstract": "Large-scale language models have become increasingly challenging and expensive to train. Among various methods addressing this issue, Pipeline Parallelism has been widely employed to accommodate massive model weights within limited GPU memory. This paper introduces Hanayo, a wave-like pipeline parallelism strategy that boasts a concise structure and practical applicability, alongside a high-performance pipeline execution runtime to tackle the challenges of pipeline strategy implementation. Hanayo mitigates the issues of pipeline bubbles and excessive memory consumption prevalent in existing schemes, without resorting to model duplicates as in Chimera. Our evaluation, conducted on four distinct computing clusters and involving both GPT-like and BERT-like architectures with up to 32 GPUs, demonstrates up to a 30.4 \\% increase in throughput compared to the state-of-the-art approach."
                    },
                    {
                        "title": "IPG-Net: Image Pyramid Guidance Network for Small Object Detection",
                        "abstract": "For Convolutional Neural Network-based object detection, there is a typical dilemma: the spatial information is well kept in the shallow layers which unfortunately do not have enough semantic information, while the deep layers have a high semantic concept but lost a lot of spatial information, resulting in serious information imbalance. To acquire enough semantic information for shallow layers, Feature Pyramid Networks (FPN) is used to build a top-down propagated path. In this paper, except for top-down combining of information for shallow layers, we propose a novel network called Image Pyramid Guidance Network (IPG-Net) to make sure both the spatial information and semantic information are abundant for each layer. Our IPG-Net has two main parts: the image pyramid guidance transformation module and the image pyramid guidance fusion module. Our main idea is to introduce the image pyramid guidance into the backbone stream to solve the information imbalance problem, which alleviates the vanishment of the small object features. This IPG transformation module promises even in the deepest stage of the backbone, there is enough spatial information for bounding box regression and classification. Furthermore, we designed an effective fusion module to fuse the features from the image pyramid and features from the backbone stream. We have tried to apply this novel network to both one-stage and two-stage detection models, state of the art results are obtained on the most popular benchmark data sets, i.e. MS COCO and Pascal VOC."
                    },
                    {
                        "title": "Principal Component Analysis of collective flow in Relativistic Heavy-Ion Collisions",
                        "abstract": "In this paper, we implement Principal Component Analysis (PCA) to study the single particle distributions generated from thousands of {\\tt VISH2+1} hydrodynamic simulations with an aim to explore if a machine could directly discover flow from the huge amount of data without explicit instructions from human-beings. We found that the obtained PCA eigenvectors are similar to but not identical with the traditional Fourier bases. Correspondingly, the PCA defined flow harmonics $v_n^\\prime$ are also similar to the traditional $v_n$ for $n=2$ and 3, but largely deviated from the Fourier ones for $n\\geq 4$. A further study on the symmetric cumulants and the Pearson coefficients indicates that mode-coupling effects are reduced for these flow harmonics defined by PCA."
                    },
                    {
                        "title": "Precision Machine Learning",
                        "abstract": "We explore unique considerations involved in fitting ML models to data with very high precision, as is often required for science applications. We empirically compare various function approximation methods and study how they scale with increasing parameters and data. We find that neural networks can often outperform classical approximation methods on high-dimensional examples, by auto-discovering and exploiting modular structures therein. However, neural networks trained with common optimizers are less powerful for low-dimensional cases, which motivates us to study the unique properties of neural network loss landscapes and the corresponding optimization challenges that arise in the high precision regime. To address the optimization issue in low dimensions, we develop training tricks which enable us to train neural networks to extremely low loss, close to the limits allowed by numerical precision."
                    },
                    {
                        "title": "ATP: Adaptive Tensor Parallelism for Foundation Models",
                        "abstract": "Foundation models have impressive performance and generalization capabilities across a wide range of applications. The increasing size of the models introduces great challenges for the training. Tensor parallelism is a critical technique that is currently used in almost all foundation model training and has a significant impact on overall training performance. However, current tensor parallelism in machine learning frameworks misses optimization opportunities in fitting various interconnection topologies. In this work, we present ATP, an adaptive tensor parallelism framework for foundation models, which can automatically select the optimal parallel strategy on different interconnections. We propose column- and row-first tensor parallelism based on 2D device meshes and construct a search space. Combined with the hierarchical communication matrix, ATP can identify the optimal strategy in the search space. We also propose chunk-based overlapping to reduce communication overhead. Our evaluations show ATP consistently outperforms the state-of-the-art approaches for various model sizes and interconnects, achieving end-to-end training performance improvements of up to 37-64% on specific interconnects. Based on our theoretical model, the communication overhead of ATP decreases with scaling, indicating a qualitative leap forward."
                    },
                    {
                        "title": "The Clock and the Pizza: Two Stories in Mechanistic Explanation of Neural Networks",
                        "abstract": "Do neural networks, trained on well-understood algorithmic tasks, reliably rediscover known algorithms for solving those tasks? Several recent studies, on tasks ranging from group arithmetic to in-context linear regression, have suggested that the answer is yes. Using modular addition as a prototypical problem, we show that algorithm discovery in neural networks is sometimes more complex. Small changes to model hyperparameters and initializations can induce the discovery of qualitatively different algorithms from a fixed training set, and even parallel implementations of multiple such algorithms. Some networks trained to perform modular addition implement a familiar Clock algorithm; others implement a previously undescribed, less intuitive, but comprehensible procedure which we term the Pizza algorithm, or a variety of even more complex procedures. Our results show that even simple learning problems can admit a surprising diversity of solutions, motivating the development of new tools for characterizing the behavior of neural networks across their algorithmic phase space."
                    },
                    {
                        "title": "Grokking as Compression: A Nonlinear Complexity Perspective",
                        "abstract": "We attribute grokking, the phenomenon where generalization is much delayed after memorization, to compression. To do so, we define linear mapping number (LMN) to measure network complexity, which is a generalized version of linear region number for ReLU networks. LMN can nicely characterize neural network compression before generalization. Although the $L_2$ norm has been a popular choice for characterizing model complexity, we argue in favor of LMN for a number of reasons: (1) LMN can be naturally interpreted as information/computation, while $L_2$ cannot. (2) In the compression phase, LMN has linear relations with test losses, while $L_2$ is correlated with test losses in a complicated nonlinear way. (3) LMN also reveals an intriguing phenomenon of the XOR network switching between two generalization solutions, while $L_2$ does not. Besides explaining grokking, we argue that LMN is a promising candidate as the neural network version of the Kolmogorov complexity since it explicitly considers local or conditioned linear computations aligned with the nature of modern artificial neural networks."
                    }
                ]
            },
            "6f48d6fb-6acd-4171-9fba-cd356bcd8e68": {
                "pk": "6f48d6fb-6acd-4171-9fba-cd356bcd8e68",
                "name": "Mitchell Ostrow",
                "collaborators": [
                    "Adam Eisen",
                    "Ila Fiete",
                    "Leo Kozachkov",
                    "Rylan Schaeffer",
                    "Nika Zahedi",
                    "Mikail Khona",
                    "Dhruv Pai",
                    "Sang Truong",
                    "Yilun Du",
                    "Sarthak Chandra"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Dynamical Systems",
                    "Neural Networks",
                    "Probabilistic Modeling"
                ],
                "publications": [
                    {
                        "title": "Delay Embedding Theory of Neural Sequence Models",
                        "abstract": "To generate coherent responses, language models infer unobserved meaning from their input text sequence. One potential explanation for this capability arises from theories of delay embeddings in dynamical systems, which prove that unobserved variables can be recovered from the history of only a handful of observed variables. To test whether language models are effectively constructing delay embeddings, we measure the capacities of sequence models to reconstruct unobserved dynamics. We trained 1-layer transformer decoders and state-space sequence models on next-step prediction from noisy, partially-observed time series data. We found that each sequence layer can learn a viable embedding of the underlying system. However, state-space models have a stronger inductive bias than transformers-in particular, they more effectively reconstruct unobserved information at initialization, leading to more parameter-efficient models and lower error on dynamics tasks. Our work thus forges a novel connection between dynamical systems and deep learning sequence models via delay embedding theory."
                    },
                    {
                        "title": "Beyond Geometry: Comparing the Temporal Structure of Computation in Neural Circuits with Dynamical Similarity Analysis",
                        "abstract": "How can we tell whether two neural networks utilize the same internal processes for a particular computation? This question is pertinent for multiple subfields of neuroscience and machine learning, including neuroAI, mechanistic interpretability, and brain-machine interfaces. Standard approaches for comparing neural networks focus on the spatial geometry of latent states. Yet in recurrent networks, computations are implemented at the level of dynamics, and two networks performing the same computation with equivalent dynamics need not exhibit the same geometry. To bridge this gap, we introduce a novel similarity metric that compares two systems at the level of their dynamics, called Dynamical Similarity Analysis (DSA). Our method incorporates two components: Using recent advances in data-driven dynamical systems theory, we learn a high-dimensional linear system that accurately captures core features of the original nonlinear dynamics. Next, we compare different systems passed through this embedding using a novel extension of Procrustes Analysis that accounts for how vector fields change under orthogonal transformation. In four case studies, we demonstrate that our method disentangles conjugate and non-conjugate recurrent neural networks (RNNs), while geometric methods fall short. We additionally show that our method can distinguish learning rules in an unsupervised manner. Our method opens the door to comparative analyses of the essential temporal structure of computation in neural circuits."
                    },
                    {
                        "title": "Bridging Associative Memory and Probabilistic Modeling",
                        "abstract": "Associative memory and probabilistic modeling are two fundamental topics in artificial intelligence. The first studies recurrent neural networks designed to denoise, complete and retrieve data, whereas the second studies learning and sampling from probability distributions. Based on the observation that associative memory's energy functions can be seen as probabilistic modeling's negative log likelihoods, we build a bridge between the two that enables useful flow of ideas in both directions. We showcase four examples: First, we propose new energy-based models that flexibly adapt their energy functions to new in-context datasets, an approach we term \\textit{in-context learning of energy functions}. Second, we propose two new associative memory models: one that dynamically creates new memories as necessitated by the training data using Bayesian nonparametrics, and another that explicitly computes proportional memory assignments using the evidence lower bound. Third, using tools from associative memory, we analytically and numerically characterize the memory capacity of Gaussian kernel density estimators, a widespread tool in probababilistic modeling. Fourth, we study a widespread implementation choice in transformers -- normalization followed by self attention -- to show it performs clustering on the hypersphere. Altogether, this work urges further exchange of useful ideas between these two continents of artificial intelligence."
                    }
                ]
            },
            "8ffee0c6-f8fc-48f7-85a2-716036204d4e": {
                "pk": "8ffee0c6-f8fc-48f7-85a2-716036204d4e",
                "name": "Ila Fiete",
                "collaborators": [
                    "Akhilan Boopathy",
                    "Mitchell Ostrow",
                    "Adam Eisen",
                    "Ingmar Kanitscheider",
                    "Rishidev Chaudhuri",
                    "Yoram Burak",
                    "Ted Brookings",
                    "Leo Kozachkov",
                    "Qiyao Liang",
                    "Ziming Liu"
                ],
                "domain": [
                    "Neural Networks",
                    "Robotics",
                    "Cognitive Science",
                    "Memory Systems"
                ],
                "publications": [
                    {
                        "title": "Training recurrent networks to generate hypotheses about how the brain solves hard navigation problems",
                        "abstract": "Self-localization during navigation with noisy sensors in an ambiguous world is computationally challenging, yet animals and humans excel at it. In robotics, Simultaneous Location and Mapping (SLAM) algorithms solve this problem though joint sequential probabilistic inference of their own coordinates and those of external spatial landmarks. We generate the first neural solution to the SLAM problem by training recurrent LSTM networks to perform a set of hard 2D navigation tasks that include generalization to completely novel trajectories and environments. The hidden unit representations exhibit several key properties of hippocampal place cells, including stable tuning curves that remap between environments. Our result is also a proof of concept for end-to-end-learning of a SLAM algorithm using recurrent networks, and a demonstration of why this approach may have some advantages for robotic SLAM."
                    },
                    {
                        "title": "Associative content-addressable networks with exponentially many robust stable states",
                        "abstract": "The brain must robustly store a large number of memories, corresponding to the many events encountered over a lifetime. However, the number of memory states in existing neural network models either grows weakly with network size or recall fails catastrophically with vanishingly little noise. We construct an associative content-addressable memory with exponentially many stable states and robust error-correction. The network possesses expander graph connectivity on a restricted Boltzmann machine architecture. The expansion property allows simple neural network dynamics to perform at par with modern error-correcting codes. Appropriate networks can be constructed with sparse random connections, glomerular nodes, and associative learning using low dynamic-range weights. Thus, sparse quasi-random structures---characteristic of important error-correcting codes---may provide for high-performance computation in artificial neural networks and the brain."
                    },
                    {
                        "title": "How to Train Your Wide Neural Network Without Backprop: An Input-Weight Alignment Perspective",
                        "abstract": "Recent works have examined theoretical and empirical properties of wide neural networks trained in the Neural Tangent Kernel (NTK) regime. Given that biological neural networks are much wider than their artificial counterparts, we consider NTK regime wide neural networks as a possible model of biological neural networks. Leveraging NTK theory, we show theoretically that gradient descent drives layerwise weight updates that are aligned with their input activity correlations weighted by error, and demonstrate empirically that the result also holds in finite-width wide networks. The alignment result allows us to formulate a family of biologically-motivated, backpropagation-free learning rules that are theoretically equivalent to backpropagation in infinite-width networks. We test these learning rules on benchmark problems in feedforward and recurrent neural networks and demonstrate, in wide networks, comparable performance to backpropagation. The proposed rules are particularly effective in low data regimes, which are common in biological learning settings."
                    },
                    {
                        "title": "Unified Neural Network Scaling Laws and Scale-time Equivalence",
                        "abstract": "As neural networks continue to grow in size but datasets might not, it is vital to understand how much performance improvement can be expected: is it more important to scale network size or data volume? Thus, neural network scaling laws, which characterize how test error varies with network size and data volume, have become increasingly important. However, existing scaling laws are often applicable only in limited regimes and often do not incorporate or predict well-known phenomena such as double descent. Here, we present a novel theoretical characterization of how three factors -- model size, training time, and data volume -- interact to determine the performance of deep neural networks. We first establish a theoretical and empirical equivalence between scaling the size of a neural network and increasing its training time proportionally. Scale-time equivalence challenges the current practice, wherein large models are trained for small durations, and suggests that smaller models trained over extended periods could match their efficacy. It also leads to a novel method for predicting the performance of large-scale networks from small-scale networks trained for extended epochs, and vice versa. We next combine scale-time equivalence with a linear model analysis of double descent to obtain a unified theoretical scaling law, which we confirm with experiments across vision benchmarks and network architectures. These laws explain several previously unexplained phenomena: reduced data requirements for generalization in larger models, heightened sensitivity to label noise in overparameterized models, and instances where increasing model scale does not necessarily enhance performance. Our findings hold significant implications for the practical deployment of neural networks, offering a more accessible and efficient path to training and fine-tuning large models."
                    },
                    {
                        "title": "Delay Embedding Theory of Neural Sequence Models",
                        "abstract": "To generate coherent responses, language models infer unobserved meaning from their input text sequence. One potential explanation for this capability arises from theories of delay embeddings in dynamical systems, which prove that unobserved variables can be recovered from the history of only a handful of observed variables. To test whether language models are effectively constructing delay embeddings, we measure the capacities of sequence models to reconstruct unobserved dynamics. We trained 1-layer transformer decoders and state-space sequence models on next-step prediction from noisy, partially-observed time series data. We found that each sequence layer can learn a viable embedding of the underlying system. However, state-space models have a stronger inductive bias than transformers-in particular, they more effectively reconstruct unobserved information at initialization, leading to more parameter-efficient models and lower error on dynamics tasks. Our work thus forges a novel connection between dynamical systems and deep learning sequence models via delay embedding theory."
                    },
                    {
                        "title": "Triangular lattice neurons may implement an advanced numeral system to precisely encode rat position over large ranges",
                        "abstract": "We argue by observation of the neural data that neurons in area dMEC of rats, which fire whenever the rat is on any vertex of a regular triangular lattice that tiles 2-d space, may be using an advanced numeral system to reversibly encode rat position. We interpret measured dMEC properties within the framework of a residue number system (RNS), and describe how RNS encoding -- which breaks the non-periodic variable of rat position into a set of narrowly distributed periodic variables -- allows a small set of cells to compactly represent and efficiently update rat position with high resolution over a large range. We show that the uniquely useful properties of RNS encoding still hold when the encoded and encoding quantities are relaxed to be real numbers with built-in uncertainties, and provide a numerical and functional estimate of the range and resolution of rat positions that can be uniquely encoded in dMEC. The use of a compact, `arithmetic-friendly' numeral system to encode a metric variable, as we propose is happening in dMEC, is qualitatively different from all previously identified examples of coding in the brain. We discuss the numerous neurobiological implications and predictions of our hypothesis."
                    },
                    {
                        "title": "Beyond Geometry: Comparing the Temporal Structure of Computation in Neural Circuits with Dynamical Similarity Analysis",
                        "abstract": "How can we tell whether two neural networks utilize the same internal processes for a particular computation? This question is pertinent for multiple subfields of neuroscience and machine learning, including neuroAI, mechanistic interpretability, and brain-machine interfaces. Standard approaches for comparing neural networks focus on the spatial geometry of latent states. Yet in recurrent networks, computations are implemented at the level of dynamics, and two networks performing the same computation with equivalent dynamics need not exhibit the same geometry. To bridge this gap, we introduce a novel similarity metric that compares two systems at the level of their dynamics, called Dynamical Similarity Analysis (DSA). Our method incorporates two components: Using recent advances in data-driven dynamical systems theory, we learn a high-dimensional linear system that accurately captures core features of the original nonlinear dynamics. Next, we compare different systems passed through this embedding using a novel extension of Procrustes Analysis that accounts for how vector fields change under orthogonal transformation. In four case studies, we demonstrate that our method disentangles conjugate and non-conjugate recurrent neural networks (RNNs), while geometric methods fall short. We additionally show that our method can distinguish learning rules in an unsupervised manner. Our method opens the door to comparative analyses of the essential temporal structure of computation in neural circuits."
                    },
                    {
                        "title": "Do Diffusion Models Learn Semantically Meaningful and Efficient Representations?",
                        "abstract": "Diffusion models are capable of impressive feats of image generation with uncommon juxtapositions such as astronauts riding horses on the moon with properly placed shadows. These outputs indicate the ability to perform compositional generalization, but how do the models do so? We perform controlled experiments on conditional DDPMs learning to generate 2D spherical Gaussian bumps centered at specified $x$- and $y$-positions. Our results show that the emergence of semantically meaningful latent representations is key to achieving high performance. En route to successful performance over learning, the model traverses three distinct phases of latent representations: (phase A) no latent structure, (phase B) a 2D manifold of disordered states, and (phase C) a 2D ordered manifold. Corresponding to each of these phases, we identify qualitatively different generation behaviors: 1) multiple bumps are generated, 2) one bump is generated but at inaccurate $x$ and $y$ locations, 3) a bump is generated at the correct $x$ and y location. Furthermore, we show that even under imbalanced datasets where features ($x$- versus $y$-positions) are represented with skewed frequencies, the learning process for $x$ and $y$ is coupled rather than factorized, demonstrating that simple vanilla-flavored diffusion models cannot learn efficient representations in which localization in $x$ and $y$ are factorized into separate 1D tasks. These findings suggest the need for future work to find inductive biases that will push generative models to discover and exploit factorizable independent structures in their inputs, which will be required to vault these models into more data-efficient regimes."
                    },
                    {
                        "title": "Map Induction: Compositional spatial submap learning for efficient exploration in novel environments",
                        "abstract": "Humans are expert explorers. Understanding the computational cognitive mechanisms that support this efficiency can advance the study of the human mind and enable more efficient exploration algorithms. We hypothesize that humans explore new environments efficiently by inferring the structure of unobserved spaces using spatial information collected from previously explored spaces. This cognitive process can be modeled computationally using program induction in a Hierarchical Bayesian framework that explicitly reasons about uncertainty with strong spatial priors. Using a new behavioral Map Induction Task, we demonstrate that this computational framework explains human exploration behavior better than non-inductive models and outperforms state-of-the-art planning algorithms when applied to a realistic spatial navigation domain."
                    },
                    {
                        "title": "A binary Hopfield network with $1/\\log(n)$ information rate and applications to grid cell decoding",
                        "abstract": "A Hopfield network is an auto-associative, distributive model of neural memory storage and retrieval. A form of error-correcting code, the Hopfield network can learn a set of patterns as stable points of the network dynamic, and retrieve them from noisy inputs -- thus Hopfield networks are their own decoders. Unlike in coding theory, where the information rate of a good code (in the Shannon sense) is finite but the cost of decoding does not play a role in the rate, the information rate of Hopfield networks trained with state-of-the-art learning algorithms is of the order ${\\log(n)}/{n}$, a quantity that tends to zero asymptotically with $n$, the number of neurons in the network. For specially constructed networks, the best information rate currently achieved is of order ${1}/{\\sqrt{n}}$. In this work, we design simple binary Hopfield networks that have asymptotically vanishing error rates at an information rate of ${1}/{\\log(n)}$. These networks can be added as the decoders of any neural code with noisy neurons. As an example, we apply our network to a binary neural decoder of the grid cell code to attain information rate ${1}/{\\log(n)}$."
                    },
                    {
                        "title": "Winning the lottery with neural connectivity constraints: faster learning across cognitive tasks with spatially constrained sparse RNNs",
                        "abstract": "Recurrent neural networks (RNNs) are often used to model circuits in the brain, and can solve a variety of difficult computational problems requiring memory, error-correction, or selection [Hopfield, 1982, Maass et al., 2002, Maass, 2011]. However, fully-connected RNNs contrast structurally with their biological counterparts, which are extremely sparse (~0.1%). Motivated by the neocortex, where neural connectivity is constrained by physical distance along cortical sheets and other synaptic wiring costs, we introduce locality masked RNNs (LM-RNNs) that utilize task-agnostic predetermined graphs with sparsity as low as 4%. We study LM-RNNs in a multitask learning setting relevant to cognitive systems neuroscience with a commonly used set of tasks, 20-Cog-tasks [Yang et al., 2019]. We show through reductio ad absurdum that 20-Cog-tasks can be solved by a small pool of separated autapses that we can mechanistically analyze and understand. Thus, these tasks fall short of the goal of inducing complex recurrent dynamics and modular structure in RNNs. We next contribute a new cognitive multi-task battery, Mod-Cog, consisting of upto 132 tasks that expands by 7-fold the number of tasks and task-complexity of 20-Cog-tasks. Importantly, while autapses can solve the simple 20-Cog-tasks, the expanded task-set requires richer neural architectures and continuous attractor dynamics. On these tasks, we show that LM-RNNs with an optimal sparsity result in faster training and better data-efficiency than fully connected networks."
                    }
                ]
            }
        }
    },
    "2406.01583": {
        "paper_data": {
            "title": "Decomposing and Interpreting Image Representations via Text in ViTs Beyond CLIP",
            "url": "http://arxiv.org/abs/2406.01583v2",
            "arxiv_id": "2406.01583",
            "authors": [
                "Sriram Balasubramanian",
                "Samyadeep Basu",
                "Soheil Feizi"
            ],
            "abstract": "Recent work has explored how individual components of the CLIP-ViT model contribute to the final representation by leveraging the shared image-text representation space of CLIP. These components, such as attention heads and MLPs, have been shown to capture distinct image features like shape, color or texture. However, understanding the role of these components in arbitrary vision transformers (ViTs) is challenging. To this end, we introduce a general framework which can identify the roles of various components in ViTs beyond CLIP. Specifically, we (a) automate the decomposition of the final representation into contributions from different model components, and (b) linearly map these contributions to CLIP space to interpret them via text. Additionally, we introduce a novel scoring function to rank components by their importance with respect to specific features. Applying our framework to various ViT variants (e.g. DeiT, DINO, DINOv2, Swin, MaxViT), we gain insights into the roles of different components concerning particular image features. These insights facilitate applications such as image retrieval using text descriptions or reference images, visualizing token importance heatmaps, and mitigating spurious correlations. We release our code to reproduce the experiments at https://github.com/SriramB-98/vit-decompose",
            "introduction": "   1 Introduction            Figure 1: (Left) Workflow:  The first step (RepDecompose) is to decompose a representation \ud835\udc9b\ud835\udc9b{\\bm{z}}bold_italic_z into contributions from its model components \ud835\udc84isubscript\ud835\udc84\ud835\udc56{\\bm{c}}_{i}bold_italic_c start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT after being transformed by residual transformations like LayerNorm, linear projections, resampling, patch merging and so on. The second step (CompAlign) aligns each contribution to CLIP space using a set of linear maps f0,f1,\u2026,fnsubscript\ud835\udc530subscript\ud835\udc531\u2026subscript\ud835\udc53\ud835\udc5bf_{0},f_{1},\\dots,f_{n}italic_f start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_f start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , \u2026 , italic_f start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT on the corresponding contributions \ud835\udc840,\ud835\udc841,\u2026,\ud835\udc84nsubscript\ud835\udc840subscript\ud835\udc841\u2026subscript\ud835\udc84\ud835\udc5b{\\bm{c}}_{0},{\\bm{c}}_{1},\\dots,{\\bm{c}}_{n}bold_italic_c start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , bold_italic_c start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , \u2026 , bold_italic_c start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT. We can then interpret these aligned contributions using the CLIP text encoder. (Right) Applications of our method:  (a) Visualizing contributions of each token through a specific component using a joint token-component decomposition (b) Retrieving images that are close matches of the reference image (on top) with respect to a given image feature like pattern, person, or location   Vision transformers and their variants [10, 22, 7, 33, 17, 32] have emerged as powerful image encoders, becoming the preferred architecture for modern image foundation models. However, the mechanisms by which these models transform images into representation vectors remain poorly understood. Recently, Gandelsman et\u00a0al. [11] made significant progress on this question for CLIP-ViT models with two key insights: (i) They demonstrated that the residual connections and attention mechanisms of CLIP-ViT enable the model output to be mathematically represented as a sum of vectors over layers, attention heads, and tokens, along with contributions from MLPs and the CLS token. Each vector corresponds to the contribution of a specific token attended to by a particular attention head in a specific layer. (ii) These contribution vectors exist within the same shared image-text representation space, allowing the CLIP text encoder to interpret each vector individually via text.   Extending this approach to other transformer-based image encoders presents several challenges. Popular models like DeiT\u00a0[32], DINO-ViT\u00a0[7, 22], and Swin\u00a0[17] lack a corresponding text encoder to interpret the component contributions. Additionally, extracting the contribution vectors corresponding to these components is not straightforward, as they are often not explicitly computed during the forward pass of the model. Other complications include diverse attention mechanisms such as grid attention, block attention (in MaxViT), and windowed/shifted windowed attention (in Swin), as well as various linear transformations like pooling, downsampling, and patch merging applied to the residual streams between attention blocks. These differences necessitate a fresh mathematical analysis for each model architecture, followed by careful application of necessary transformations to the intermediate output of each component to determine its contribution to the final representation. To address these challenges, we propose our framework (described in Figure\u00a01) to identify roles of components in general ViTs.   First, we automate the decomposition of the representation by leveraging the computational graph created during the forward pass. This results in a drop-in function, RepDecompose, that can decompose any representation into contributions vectors from model components simply by calling it on the representation. Since this method operates on the computational graph, it is agnostic to the specifics of the model implementation and thus applicable to a variety of model architectures.   Secondly, we introduce",
            "references": [
                {
                    "title": "Decomposing and Editing Predictions by Modeling Model Computation",
                    "abstract": "How does the internal computation of a machine learning model transform inputs into predictions? In this paper, we introduce a task called component modeling that aims to address this question. The goal of component modeling is to decompose an ML model's prediction in terms of its components -- simple functions (e.g., convolution filters, attention heads) that are the\"building blocks\"of model computation. We focus on a special case of this task, component attribution, where the goal is to estimate the counterfactual impact of individual components on a given prediction. We then present COAR, a scalable algorithm for estimating component attributions; we demonstrate its effectiveness across models, datasets, and modalities. Finally, we show that component attributions estimated with COAR directly enable model editing across five tasks, namely: fixing model errors, ``forgetting'' specific classes, boosting subpopulation robustness, localizing backdoor attacks, and improving robustness to typographic attacks. We provide code for COAR at https://github.com/MadryLab/modelcomponents ."
                },
                {
                    "title": "Analyzing Vision Transformers for Image Classification in Class Embedding Space",
                    "abstract": "Despite the growing use of transformer models in computer vision, a mechanistic understanding of these networks is still needed. This work introduces a method to reverse-engineer Vision Transformers trained to solve image classification tasks. Inspired by previous research in NLP, we demonstrate how the inner representations at any level of the hierarchy can be projected onto the learned class embedding space to uncover how these networks build categorical representations for their predictions. We use our framework to show how image tokens develop class-specific representations that depend on attention mechanisms and contextual information, and give insights on how self-attention and MLP layers differentially contribute to this categorical composition. We additionally demonstrate that this method (1) can be used to determine the parts of an image that would be important for detecting the class of interest, and (2) exhibits significant advantages over traditional linear probing approaches. Taken together, our results position our proposed framework as a powerful tool for mechanistic interpretability and explainability research."
                },
                {
                    "title": "Interpreting CLIP's Image Representation via Text-Based Decomposition",
                    "abstract": "We investigate the CLIP image encoder by analyzing how individual model components affect the final representation. We decompose the image representation as a sum across individual image patches, model layers, and attention heads, and use CLIP's text representation to interpret the summands. Interpreting the attention heads, we characterize each head's role by automatically finding text representations that span its output space, which reveals property-specific roles for many heads (e.g. location or shape). Next, interpreting the image patches, we uncover an emergent spatial localization within CLIP. Finally, we use this understanding to remove spurious features from CLIP and to create a strong zero-shot image segmenter. Our results indicate that a scalable understanding of transformer models is attainable and can be used to repair and improve models."
                },
                {
                    "title": "Reproducible Scaling Laws for Contrastive Language-Image Learning",
                    "abstract": "Scaling up neural networks has led to remarkable performance across a wide range of tasks. Moreover, performance often follows reliable scaling laws as a function of training set size, model size, and compute, which offers valuable guidance as large-scale experiments are becoming increasingly expensive. However, previous work on scaling laws has primarily used private data & models or focused on uni-modal language or vision learning. To address these limitations, we investigate scaling laws for contrastive language-image pre-training (CLIP) with the public LAION dataset and the open-source OpenCLIP repository. Our large-scale experiments involve models trained on up to two billion image-text pairs and identify power law scaling for multiple downstream tasks including zero-shot classification, retrieval, linear probing, and end-to-end fine-tuning. We find that the training distribution plays a key role in scaling laws as the OpenAI and OpenCLIP models exhibit different scaling behavior despite identical model architectures and similar training recipes. We open-source our evaluation workflow and all models, including the largest public CLIP models, to ensure reproducibility and make scaling laws research more accessible. Source code and instructions to reproduce this study is available at https://github.eom/LAION-AI/sealing-laws-openelip."
                },
                {
                    "title": "Towards Improved Input Masking for Convolutional Neural Networks",
                    "abstract": "The ability to remove features from the input of machine learning models is very important to understand and interpret model predictions. However, this is non-trivial for vision models since masking out parts of the input image typically causes large distribution shifts. This is because the baseline color used for masking (typically grey or black) is out of distribution. Furthermore, the shape of the mask itself can contain unwanted signals which can be used by the model for its predictions. Recently, there has been some progress in mitigating this issue (called missingness bias) in image masking for vision transformers. In this work, we propose a new masking method for CNNs we call layer masking in which the missingness bias caused by masking is reduced to a large extent. Intuitively, layer masking applies a mask to intermediate activation maps so that the model only processes the unmasked input. We show that our method (i) is able to eliminate or minimize the influence of the mask shape or color on the output of the model, and (ii) is much better than replacing the masked region by black or grey for input perturbation based interpretability techniques like LIME. Thus, layer masking is much less affected by missingness bias than other masking strategies. We also demonstrate how the shape of the mask may leak information about the class, thus affecting estimates of model reliance on class-relevant features derived from input masking. Furthermore, we discuss the role of data augmentation techniques for tackling this problem, and argue that they are not sufficient for preventing model reliance on mask shape."
                },
                {
                    "title": "CLIP-Dissect: Automatic Description of Neuron Representations in Deep Vision Networks",
                    "abstract": "In this paper, we propose CLIP-Dissect, a new technique to automatically describe the function of individual hidden neurons inside vision networks. CLIP-Dissect leverages recent advances in multimodal vision/language models to label internal neurons with open-ended concepts without the need for any labeled data or human examples. We show that CLIP-Dissect provides more accurate descriptions than existing methods for last layer neurons where the ground-truth is available as well as qualitatively good descriptions for hidden layer neurons. In addition, our method is very flexible: it is model agnostic, can easily handle new concepts and can be extended to take advantage of better multimodal models in the future. Finally CLIP-Dissect is computationally efficient and can label all neurons from five layers of ResNet-50 in just 4 minutes, which is more than 10 times faster than existing methods. Our code is available at https://github.com/Trustworthy-ML-Lab/CLIP-dissect. Finally, crowdsourced user study results are available at Appendix B to further support the effectiveness of our method."
                },
                {
                    "title": "MaxViT: Multi-Axis Vision Transformer",
                    "abstract": "Transformers have recently gained significant attention in the computer vision community. However, the lack of scalability of self-attention mechanisms with respect to image size has limited their wide adoption in state-of-the-art vision backbones. In this paper we introduce an efficient and scalable attention model we call multi-axis attention, which consists of two aspects: blocked local and dilated global attention. These design choices allow global-local spatial interactions on arbitrary input resolutions with only linear complexity. We also present a new architectural element by effectively blending our proposed attention model with convolutions, and accordingly propose a simple hierarchical vision backbone, dubbed MaxViT, by simply repeating the basic building block over multiple stages. Notably, MaxViT is able to ''see'' globally throughout the entire network, even in earlier, high-resolution stages. We demonstrate the effectiveness of our model on a broad spectrum of vision tasks. On image classification, MaxViT achieves state-of-the-art performance under various settings: without extra data, MaxViT attains 86.5% ImageNet-1K top-1 accuracy; with ImageNet-21K pre-training, our model achieves 88.7% top-1 accuracy. For downstream tasks, MaxViT as a backbone delivers favorable performance on object detection as well as visual aesthetic assessment. We also show that our proposed model expresses strong generative modeling capability on ImageNet, demonstrating the superior potential of MaxViT blocks as a universal vision module. The source code and trained models will be available at https://github.com/google-research/maxvit."
                },
                {
                    "title": "Revisiting Model Stitching to Compare Neural Representations",
                    "abstract": "We revisit and extend model stitching (Lenc&Vedaldi 2015) as a methodology to study the internal representations of neural networks. Given two trained and frozen models $A$ and $B$, we consider a\"stitched model'' formed by connecting the bottom-layers of $A$ to the top-layers of $B$, with a simple trainable layer between them. We argue that model stitching is a powerful and perhaps under-appreciated tool, which reveals aspects of representations that measures such as centered kernel alignment (CKA) cannot. Through extensive experiments, we use model stitching to obtain quantitative verifications for intuitive statements such as\"good networks learn similar representations'', by demonstrating that good networks of the same architecture, but trained in very different ways (e.g.: supervised vs. self-supervised learning), can be stitched to each other without drop in performance. We also give evidence for the intuition that\"more is better'' by showing that representations learnt with (1) more data, (2) bigger width, or (3) more training time can be\"plugged in'' to weaker models to improve performance. Finally, our experiments reveal a new structural property of SGD which we call\"stitching connectivity'', akin to mode-connectivity: typical minima reached by SGD can all be stitched to each other with minimal change in accuracy."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Training data-efficient image transformers & distillation through attention",
                    "abstract": "Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption. In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data. More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models."
                },
                {
                    "title": "Transformer Interpretability Beyond Attention Visualization",
                    "abstract": "Self-attention techniques, and specifically Transformers, are dominating the field of text processing and are becoming increasingly popular in computer vision classification tasks. In order to visualize the parts of the image that led to a certain classification, existing methods either rely on the obtained attention maps or employ heuristic propagation along the attention graph. In this work, we propose a novel way to compute relevancy for Transformer networks. The method assigns local relevance based on the Deep Taylor Decomposition principle and then propagates these relevancy scores through the layers. This propagation involves attention layers and skip connections, which challenge existing methods. Our solution is based on a specific formulation that is shown to maintain the total relevancy across layers. We benchmark our method on very recent visual Transformer networks, as well as on a text classification problem, and demonstrate a clear advantage over the existing explainability methods. Our code is available at: https://github.com/hila-chefer/Transformer-Explainability."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Understanding the role of individual units in a deep neural network",
                    "abstract": "Deep neural networks excel at finding hierarchical representations that solve complex tasks over large datasets. How can we humans understand these learned representations? In this work, we present network dissection, an analytic framework to systematically identify the semantics of individual hidden units within image classification and image generation networks. First, we analyze a convolutional neural network (CNN) trained on scene classification and discover units that match a diverse set of object concepts. We find evidence that the network has learned many object classes that play crucial roles in classifying scene classes. Second, we use a similar analytic method to analyze a generative adversarial network (GAN) model trained to generate scenes. By analyzing changes made when small sets of units are activated or deactivated, we find that objects can be added and removed from the output scenes while adapting to the context. Finally, we apply our analytic framework to understanding adversarial attacks and to semantic image editing."
                },
                {
                    "title": "Influence Functions in Deep Learning Are Fragile",
                    "abstract": "Influence functions approximate the effect of training samples in test-time predictions and have a wide variety of applications in machine learning interpretability and uncertainty estimation. A commonly-used (first-order) influence function can be implemented efficiently as a post-hoc method requiring access only to the gradients and Hessian of the model. For linear models, influence functions are well-defined due to the convexity of the underlying loss function and are generally accurate even across difficult settings where model changes are fairly large such as estimating group influences. Influence functions, however, are not well-understood in the context of deep learning with non-convex loss functions. In this paper, we provide a comprehensive and large-scale empirical study of successes and failures of influence functions in neural network models trained on datasets such as Iris, MNIST, CIFAR-10 and ImageNet. Through our extensive experiments, we show that the network architecture, its depth and width, as well as the extent of model parameterization and regularization techniques have strong effects in the accuracy of influence functions. In particular, we find that (i) influence estimates are fairly accurate for shallow networks, while for deeper networks the estimates are often erroneous; (ii) for certain network architectures and datasets, training with weight-decay regularization is important to get high-quality influence estimates; and (iii) the accuracy of influence estimates can vary significantly depending on the examined test points. These results suggest that in general influence functions in deep learning are fragile and call for developing improved influence estimation methods to mitigate these issues in non-convex setups."
                },
                {
                    "title": "Estimating Training Data Influence by Tracking Gradient Descent",
                    "abstract": "We introduce a method called TrackIn that computes the influence of a training example on a prediction made by the model, by tracking how the loss on the test point changes during the training process whenever the training example of interest was utilized. We provide a scalable implementation of TrackIn via a combination of a few key ideas: (a) a first-order approximation to the exact computation, (b) using random projections to speed up the computation of the first-order approximation for large models, (c) using saved checkpoints of standard training procedures, and (d) cherry-picking layers of a deep neural network. An experimental evaluation shows that TrackIn is more effective in identifying mislabelled training examples than other related methods such as influence functions and representer points. We also discuss insights from applying the method on vision, regression and natural language tasks."
                },
                {
                    "title": "Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization",
                    "abstract": "Overparameterized neural networks can be highly accurate on average on an i.i.d. test set yet consistently fail on atypical groups of the data (e.g., by learning spurious correlations that hold on average but not in such groups). Distributionally robust optimization (DRO) allows us to learn models that instead minimize the worst-case training loss over a set of pre-defined groups. However, we find that naively applying group DRO to overparameterized neural networks fails: these models can perfectly fit the training data, and any model with vanishing average training loss also already has vanishing worst-case training loss. Instead, the poor worst-case performance arises from poor generalization on some groups. By coupling group DRO models with increased regularization---a stronger-than-typical L2 penalty or early stopping---we achieve substantially higher worst-group accuracies, with 10-40 percentage point improvements on a natural language inference task and two image tasks, while maintaining high average accuracies. Our results suggest that regularization is important for worst-group generalization in the overparameterized regime, even if it is not needed for average generalization. Finally, we introduce a stochastic optimization algorithm, with convergence guarantees, to efficiently train group DRO models."
                },
                {
                    "title": "SmoothGrad: removing noise by adding noise",
                    "abstract": "Explaining the output of a deep network remains a challenge. In the case of an image classifier, one type of explanation is to identify pixels that strongly influence the final decision. A starting point for this strategy is the gradient of the class score function with respect to the input image. This gradient can be interpreted as a sensitivity map, and there are several techniques that elaborate on this basic idea. This paper makes two contributions: it introduces SmoothGrad, a simple method that can help visually sharpen gradient-based sensitivity maps, and it discusses lessons in the visualization of these maps. We publish the code for our experiments and a website with our results."
                },
                {
                    "title": "A Unified Approach to Interpreting Model Predictions",
                    "abstract": "Understanding why a model makes a certain prediction can be as crucial as the prediction's accuracy in many applications. However, the highest accuracy for large modern datasets is often achieved by complex models that even experts struggle to interpret, such as ensemble or deep learning models, creating a tension between accuracy and interpretability. In response, various methods have recently been proposed to help users interpret the predictions of complex models, but it is often unclear how these methods are related and when one method is preferable over another. To address this problem, we present a unified framework for interpreting predictions, SHAP (SHapley Additive exPlanations). SHAP assigns each feature an importance value for a particular prediction. Its novel components include: (1) the identification of a new class of additive feature importance measures, and (2) theoretical results showing there is a unique solution in this class with a set of desirable properties. The new class unifies six existing methods, notable because several recent methods in the class lack the proposed desirable properties. Based on insights from this unification, we present new methods that show improved computational performance and/or better consistency with human intuition than previous approaches."
                },
                {
                    "title": "Understanding Black-box Predictions via Influence Functions",
                    "abstract": "How can we explain the predictions of a black-box model? In this paper, we use influence functions \u2014 a classic technique from robust statistics \u2014 to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. We show that even on non-convex and non-differentiable models where the theory breaks down, approximations to influence functions can still provide valuable information. On linear models and convolutional neural networks, we demonstrate that influence functions are useful for multiple purposes: understanding model behavior, debugging models, detecting dataset errors, and even creating visually-indistinguishable training-set attacks."
                },
                {
                    "title": "Axiomatic Attribution for Deep Networks",
                    "abstract": "We study the problem of attributing the prediction of a deep network to its input features, a problem previously studied by several other works. We identify two fundamental axioms\u2014 Sensitivity and Implementation Invariance that attribution methods ought to satisfy. We show that they are not satisfied by most known attribution methods, which we consider to be a fundamental weakness of those methods. We use the axioms to guide the design of a new attribution method called Integrated Gradients. Our method requires no modification to the original network and is extremely simple to implement; it just needs a few calls to the standard gradient operator. We apply this method to a couple of image models, a couple of text models and a chemistry model, demonstrating its ability to debug networks, to extract rules from a network, and to enable users to engage with models better."
                },
                {
                    "title": "\u201cModel\u201d",
                    "abstract": "Open Wiki community (do-ocracy) (sort of like Wikipedia but for software instead of content) Wiki Way participation in the code (500+ with full write access to the complete code base) Scheduled releases (major releases every eight months, with Long Term Support versions, similar to Ubuntu) All-in-one codebase (1,000,000 lines of code, with everything bundled. Each feature is optional.) Inherent synchronized releases (all features have to be ready at the same time). Half the code comes from other projects such as Zend Framework, jQuery, Smarty, etc. Many features, but no duplication (in a wiki, similar/related content is merged, so the same is applied to Tiki's features). Tiki is the Free/Open Source Web Application with the most built-in features. Dogfood Tiki is a community recursively developing a community management system."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "Grad-CAM: Why did you say that? Visual Explanations from Deep Networks via Gradient-based Localization",
                    "abstract": "We propose a technique for making Convolutional Neural Network (CNN)-based\nmodels more transparent by visualizing the regions of input that are\n\"important\" for predictions from these models - or visual explanations. Our approach, called Gradient-weighted Class Activation Mapping (Grad-CAM),\nuses the class-specific gradient information flowing into the final\nconvolutional layer of a CNN to produce a coarse localization map of the\nimportant regions in the image. Grad-CAM is a strict generalization of the\nClass Activation Mapping. Unlike CAM, Grad-CAM requires no re-training and is\nbroadly applicable to any CNN-based architectures. We also show how Grad-CAM\nmay be combined with existing pixel-space visualizations to create a\nhigh-resolution class-discriminative visualization (Guided Grad-CAM). We\ngenerate Grad-CAM and Guided Grad-CAM visual explanations to better understand\nimage classification, image captioning, and visual question answering (VQA)\nmodels. In the context of image classification models, our visualizations (a)\nlend insight into their failure modes showing that seemingly unreasonable\npredictions have reasonable explanations, and (b) outperform pixel-space\ngradient visualizations (Guided Backpropagation and Deconvolution) on the\nILSVRC-15 weakly supervised localization task. For image captioning and VQA,\nour visualizations expose the somewhat surprising insight that common CNN +\nLSTM models can often be good at localizing discriminative input image regions\ndespite not being trained on grounded image-text pairs. Finally, we design and conduct human studies to measure if Guided Grad-CAM\nexplanations help users establish trust in the predictions made by deep\nnetworks. Interestingly, we show that Guided Grad-CAM helps untrained users\nsuccessfully discern a \"stronger\" deep network from a \"weaker\" one even when\nboth networks make identical predictions."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.LG",
                "I.5.1"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively decompose and interpret the contributions of individual components in various vision transformer architectures to understand their roles in generating image representations?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of vision transformers, which have become the backbone of modern image processing tasks. By elucidating how different components contribute to the final representation, researchers can improve model interpretability, leading to better model design and optimization. This understanding could also facilitate the development of more robust and efficient models, ultimately impacting practical applications in fields such as computer vision, autonomous systems, and human-computer interaction.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem stem from the complexity of various vision transformer architectures, which employ diverse attention mechanisms and transformations that complicate the extraction of contribution vectors. Naive approaches may fail because they do not account for the intricate interactions between layers, attention heads, and tokens, nor do they consider the unique characteristics of each model architecture. Additionally, the lack of a standardized method for interpreting contributions across different models presents a significant theoretical and practical obstacle.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on specific models, such as CLIP-ViT, without extending the insights to other architectures like DeiT, DINO-ViT, and Swin. The absence of corresponding text encoders for these models has limited the ability to interpret component contributions. Furthermore, existing methods have not adequately addressed the diverse attention mechanisms and transformations used in these models, creating a gap in understanding. Our approach differs by providing a generalized framework that automates the decomposition of representations across various architectures, leveraging the computational graph to facilitate this process.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology consists of two key components: (1) **RepDecompose**, a function that automates the decomposition of representations into contribution vectors from model components by utilizing the computational graph during the forward pass, making it applicable to various vision transformer architectures. (2) **CompAlign**, which aligns these contribution vectors to a shared representation space for interpretation. We will evaluate our approach using a diverse set of vision transformer models and metrics that assess the interpretability and accuracy of the contributions. The expected outcomes include a clearer understanding of component roles in image representation and enhanced interpret"
            }
        },
        "author_data": {
            "6a2dd2d6-4c1e-426c-9bd4-81a830585a8e": {
                "pk": "6a2dd2d6-4c1e-426c-9bd4-81a830585a8e",
                "name": "Sriram Balasubramanian",
                "collaborators": [
                    "Scott McCullough",
                    "Neha Hotwani",
                    "Soheil Feizi",
                    "Meric Augat",
                    "Udeni Wijesooriya",
                    "Gaurang Sriramanan",
                    "Vinu Sankar Sadasivan",
                    "Naman Jain",
                    "Gaurav Jindal",
                    "Abhijeet Awasthi"
                ],
                "domain": [
                    "Operator Theory",
                    "Machine Learning",
                    "Neural Networks",
                    "Quantum Information"
                ],
                "publications": [
                    {
                        "title": "Non-commutative Caratheodory Interpolation",
                        "abstract": "We prove a Caratheodory-Fejer type interpolation theorem for certain matrix convex sets in $\\C^d$ using the Blecher-Ruan-Sinclair characterization of abstract operator algebras. Our results generalize the work of Dmitry S. Kalyuzhnyi-Verbovetzkii for the d-dimensional non-commutative polydisc."
                    },
                    {
                        "title": "A Note on Decomposable Maps on Operator Systems",
                        "abstract": "This article contains a characterization of operator systems $\\cS$ with the property that every positive map $\\phi:\\cS \\rightarrow M_n$ is decomposable, as well as an alternate and a more direct proof of a characterization of decomposable maps due to E. St\\o rmer."
                    },
                    {
                        "title": "An Extended Reich Fixed Point Theorem",
                        "abstract": "We obtain an extended Reich fixed point theorem for the setting of generalized cone rectangular metric spaces without assuming the normality of the underlying cone. Our work is a generalization of the main result in \\cite{AAB} and \\cite{JS}."
                    },
                    {
                        "title": "On the Gruss Inequality for unital 2-positive linear maps",
                        "abstract": "In a recent work, Moslehian and Rajic have shown that the Gruss inequality holds for unital n-positive linear maps $\\phi:\\mathcal A \\rightarrow B(H)$, where $\\mathcal A$ is a unital C*-algebra and H is a Hilbert space, if $n \\ge 3$. They also demonstrate that the inequality fails to hold, in general, if $n = 1$ and question whether the inequality holds if $n=2$. In this article, we provide an affirmative answer to this question."
                    },
                    {
                        "title": "Toeplitz Corona and the Douglas Property for Free Functions",
                        "abstract": "The well known Douglas Lemma says that for operators $A,B$ on Hilbert space that $AA^*-BB^*\\succeq 0$ implies $B=AC$ for some contraction operator $C$. The result carries over directly to classical operator-valued Toeplitz operators by simply replacing operator by Toeplitz operator. Free functions generalize the notion of free polynomials and formal power series and trace back to the work of J. Taylor in the 1970s. They are of current interest, in part because of their connections with free probability and engineering systems theory. For free functions $a$ and $b$ on a free domain $\\cK$ defined free polynomial inequalities, a sufficient condition on the difference $aa^*-bb^*$ to imply the existence a free function $c$ taking contractive values on $\\cK$ such that $a=bc$ is established. The connection to recent work of Agler and McCarthy and their free Toeplitz Corona Theorem is exposited."
                    },
                    {
                        "title": "Quasi-Convex Free Polynomials",
                        "abstract": "Let $\\Rx$ denote the ring of polynomials in $g$ freely non-commuting variables $x=(x_1,...,x_g)$. There is a natural involution * on $\\Rx$ determined by $x_j^*=x_j$ and $(pq)^*=q^* p^*$ and a free polynomial $p\\in\\Rx$ is symmetric if it is invariant under this involution. If $X=(X_1,...,X_g)$ is a $g$ tuple of symmetric $n\\times n$ matrices, then the evaluation $p(X)$ is naturally defined and further $p^*(X)=p(X)^*$. In particular, if $p$ is symmetric, then $p(X)^*=p(X)$. The main result of this article says if $p$ is symmetric, $p(0)=0$ and for each $n$ and each symmetric positive definite $n\\times n$ matrix $A$ the set ${X:A-p(X)\\succ 0}$ is convex, then $p$ has degree at most two and is itself convex, or $-p$ is a hermitian sum of squares."
                    },
                    {
                        "title": "Compact Sets in the Free Topology",
                        "abstract": "Subsets of the set of $g$-tuples of matrices that are closed with respect to direct sums and compact in the free topology are characterized. They are, in a dilation theoretic sense, contained in the hull of a single point."
                    },
                    {
                        "title": "Szego and Widom Theorems for the Neil Algebra",
                        "abstract": "Versions of well known function theoretic operator theory results of Szego and Widom are established for the Neil algebra. The Neil algebra is the subalgebra of the algebra of bounded analytic functions on the unit disc consisting of those functions whose derivative vanishes at the origin."
                    },
                    {
                        "title": "Bilinear matrix inequalities and polynomials in several freely noncommuting variables",
                        "abstract": "Matrix-valued polynomials in any finite number of freely noncommuting variables that enjoy certain canonical partial convexity properties are characterized, via an algebraic certificate, in terms of Linear Matrix Inequalities and Bilinear Matrix Inequalities."
                    },
                    {
                        "title": "Towards Improved Input Masking for Convolutional Neural Networks",
                        "abstract": "The ability to remove features from the input of machine learning models is very important to understand and interpret model predictions. However, this is non-trivial for vision models since masking out parts of the input image typically causes large distribution shifts. This is because the baseline color used for masking (typically grey or black) is out of distribution. Furthermore, the shape of the mask itself can contain unwanted signals which can be used by the model for its predictions. Recently, there has been some progress in mitigating this issue (called missingness bias) in image masking for vision transformers. In this work, we propose a new masking method for CNNs we call layer masking in which the missingness bias caused by masking is reduced to a large extent. Intuitively, layer masking applies a mask to intermediate activation maps so that the model only processes the unmasked input. We show that our method (i) is able to eliminate or minimize the influence of the mask shape or color on the output of the model, and (ii) is much better than replacing the masked region by black or grey for input perturbation based interpretability techniques like LIME. Thus, layer masking is much less affected by missingness bias than other masking strategies. We also demonstrate how the shape of the mask may leak information about the class, thus affecting estimates of model reliance on class-relevant features derived from input masking. Furthermore, we discuss the role of data augmentation techniques for tackling this problem, and argue that they are not sufficient for preventing model reliance on mask shape. The code for this project is publicly available at https://github.com/SriramB-98/layer_masking"
                    },
                    {
                        "title": "C*-extreme entanglement breaking maps on operator systems",
                        "abstract": "Let $\\mathcal E$ denote the set of all unital entanglement breaking (UEB) linear maps defined on an operator system $\\mathcal S \\subset M_d$ and, mapping into $M_n$. As it turns out, the set $\\mathcal E$ is not only convex in the classical sense but also in a quantum sense, namely it is $C^*$-convex. The main objective of this article is to describe the $C^*$-extreme points of this set $\\mathcal E$. By observing that every EB map defined on the operator system $\\mathcal S$ dilates to a positive map with commutative range and also extends to an EB map on $M_d$, we show that the $C^*$-extreme points of the set $\\mathcal E$ are precisely the UEB maps that are maximal in the sense of Arveson (\\cite{A} and \\cite{A69}) and that they are also exactly the linear extreme points of the set $\\mathcal E$ with commutative range. We also determine their explicit structure, thereby obtaining operator system generalizations of the analogous structure theorem and the Krein-Milman type theorem given in \\cite{BDMS}. As a consequence, we show that $C^*$-extreme (UEB) maps in $\\mathcal E$ extend to $C^*$-extreme UEB maps on the full algebra. Finally, we obtain an improved version of the main result in \\cite{BDMS}, which contains various characterizations of $C^*$-extreme UEB maps between the algebras $M_d$ and $M_n$."
                    },
                    {
                        "title": "Exploring Geometry of Blind Spots in Vision Models",
                        "abstract": "Despite the remarkable success of deep neural networks in a myriad of settings, several works have demonstrated their overwhelming sensitivity to near-imperceptible perturbations, known as adversarial attacks. On the other hand, prior works have also observed that deep networks can be under-sensitive, wherein large-magnitude perturbations in input space do not induce appreciable changes to network activations. In this work, we study in detail the phenomenon of under-sensitivity in vision models such as CNNs and Transformers, and present techniques to study the geometry and extent of \"equi-confidence\" level sets of such networks. We propose a Level Set Traversal algorithm that iteratively explores regions of high confidence with respect to the input space using orthogonal components of the local gradients. Given a source image, we use this algorithm to identify inputs that lie in the same equi-confidence level set as the source image despite being perceptually similar to arbitrary images from other classes. We further observe that the source image is linearly connected by a high-confidence path to these inputs, uncovering a star-like structure for level sets of deep networks. Furthermore, we attempt to identify and estimate the extent of these connected higher-dimensional regions over which the model maintains a high degree of confidence. The code for this project is publicly available at https://github.com/SriramB-98/blindspots-neurips-sub"
                    },
                    {
                        "title": "What's in a Name? Are BERT Named Entity Representations just as Good for any other Name?",
                        "abstract": "We evaluate named entity representations of BERT-based NLP models by investigating their robustness to replacements from the same typed class in the input. We highlight that on several tasks while such perturbations are natural, state of the art trained models are surprisingly brittle. The brittleness continues even with the recent entity-aware BERT models. We also try to discern the cause of this non-robustness, considering factors such as tokenization and frequency of occurrence. Then we provide a simple method that ensembles predictions from multiple replacements while jointly modeling the uncertainty of type annotations and label predictions. Experiments on three NLP tasks show that our method enhances robustness and increases accuracy on both natural and adversarial datasets."
                    },
                    {
                        "title": "Simulating Network Paths with Recurrent Buffering Units",
                        "abstract": "Simulating physical network paths (e.g., Internet) is a cornerstone research problem in the emerging sub-field of AI-for-networking. We seek a model that generates end-to-end packet delay values in response to the time-varying load offered by a sender, which is typically a function of the previously output delays. The problem setting is unique, and renders the state-of-the-art text and time-series generative models inapplicable or ineffective. We formulate an ML problem at the intersection of dynamical systems, sequential decision making, and time-series modeling. We propose a novel grey-box approach to network simulation that embeds the semantics of physical network path in a new RNN-style model called RBU, providing the interpretability of standard network simulator tools, the power of neural models, the efficiency of SGD-based techniques for learning, and yielding promising results on synthetic and real-world network traces."
                    }
                ]
            },
            "9fb5241c-11b7-4e7a-883b-8019671a4a21": {
                "pk": "9fb5241c-11b7-4e7a-883b-8019671a4a21",
                "name": "Samyadeep Basu",
                "collaborators": [
                    "Soheil Feizi",
                    "Reshmi Ghosh",
                    "Soundararajan Srinivasan",
                    "Daniela Massiceti",
                    "Shell Xu Hu",
                    "Priyatham Kattakinda",
                    "Rauf Izmailov",
                    "Chris Mesterharm",
                    "Vlad Morariu",
                    "Varun Manjunatha"
                ],
                "domain": [
                    "Machine Learning",
                    "Privacy",
                    "Natural Language Processing",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Membership Model Inversion Attacks for Deep Networks",
                        "abstract": "With the increasing adoption of AI, inherent security and privacy vulnerabilities formachine learning systems are being discovered. One such vulnerability makes itpossible for an adversary to obtain private information about the types of instancesused to train the targeted machine learning model. This so-called model inversionattack is based on sequential leveraging of classification scores towards obtaininghigh confidence representations for various classes. However, for deep networks,such procedures usually lead to unrecognizable representations that are uselessfor the adversary. In this paper, we introduce a more realistic definition of modelinversion, where the adversary is aware of the general purpose of the attackedmodel (for instance, whether it is an OCR system or a facial recognition system),and the goal is to find realistic class representations within the corresponding lower-dimensional manifold (of, respectively, general symbols or general faces). To thatend, we leverage properties of generative adversarial networks for constructinga connected lower-dimensional manifold, and demonstrate the efficiency of ourmodel inversion attack that is carried out within that manifold."
                    },
                    {
                        "title": "On Second-Order Group Influence Functions for Black-Box Predictions",
                        "abstract": "With the rapid adoption of machine learning systems in sensitive applications, there is an increasing need to make black-box models explainable. Often we want to identify an influential group of training samples in a particular test prediction for a given machine learning model. Existing influence functions tackle this problem by using first-order approximations of the effect of removing a sample from the training set on model parameters. To compute the influence of a group of training samples (rather than an individual point) in model predictions, the change in optimal model parameters after removing that group from the training set can be large. Thus, in such cases, the first-order approximation can be loose. In this paper, we address this issue and propose second-order influence functions for identifying influential groups in test-time predictions. For linear models, across different sizes and types of groups, we show that using the proposed second-order influence function improves the correlation between the computed influence values and the ground truth ones. We also show that second-order influence functions could be used with optimization techniques to improve the selection of the most influential group for a test-sample."
                    },
                    {
                        "title": "On Hard Episodes in Meta-Learning",
                        "abstract": "Existing meta-learners primarily focus on improving the average task accuracy across multiple episodes. Different episodes, however, may vary in hardness and quality leading to a wide gap in the meta-learner's performance across episodes. Understanding this issue is particularly critical in industrial few-shot settings, where there is limited control over test episodes as they are typically uploaded by end-users. In this paper, we empirically analyse the behaviour of meta-learners on episodes of varying hardness across three standard benchmark datasets: CIFAR-FS, mini-ImageNet, and tiered-ImageNet. Surprisingly, we observe a wide gap in accuracy of around 50% between the hardest and easiest episodes across all the standard benchmarks and meta-learners. We additionally investigate various properties of hard episodes and highlight their connection to catastrophic forgetting during meta-training. To address the issue of sub-par performance on hard episodes, we investigate and benchmark different meta-training strategies based on adversarial training and curriculum learning. We find that adversarial training strategies are much more powerful than curriculum learning in improving the prediction performance on hard episodes."
                    },
                    {
                        "title": "Influence Functions in Deep Learning Are Fragile",
                        "abstract": "Influence functions approximate the effect of training samples in test-time predictions and have a wide variety of applications in machine learning interpretability and uncertainty estimation. A commonly-used (first-order) influence function can be implemented efficiently as a post-hoc method requiring access only to the gradients and Hessian of the model. For linear models, influence functions are well-defined due to the convexity of the underlying loss function and are generally accurate even across difficult settings where model changes are fairly large such as estimating group influences. Influence functions, however, are not well-understood in the context of deep learning with non-convex loss functions. In this paper, we provide a comprehensive and large-scale empirical study of successes and failures of influence functions in neural network models trained on datasets such as Iris, MNIST, CIFAR-10 and ImageNet. Through our extensive experiments, we show that the network architecture, its depth and width, as well as the extent of model parameterization and regularization techniques have strong effects in the accuracy of influence functions. In particular, we find that (i) influence estimates are fairly accurate for shallow networks, while for deeper networks the estimates are often erroneous; (ii) for certain network architectures and datasets, training with weight-decay regularization is important to get high-quality influence estimates; and (iii) the accuracy of influence estimates can vary significantly depending on the examined test points. These results suggest that in general influence functions in deep learning are fragile and call for developing improved influence estimation methods to mitigate these issues in non-convex setups."
                    },
                    {
                        "title": "Strong Baselines for Parameter Efficient Few-Shot Fine-tuning",
                        "abstract": "Few-shot classification (FSC) entails learning novel classes given only a few examples per class after a pre-training (or meta-training) phase on a set of base classes. Recent works have shown that simply fine-tuning a pre-trained Vision Transformer (ViT) on new test classes is a strong approach for FSC. Fine-tuning ViTs, however, is expensive in time, compute and storage. This has motivated the design of parameter efficient fine-tuning (PEFT) methods which fine-tune only a fraction of the Transformer's parameters. While these methods have shown promise, inconsistencies in experimental conditions make it difficult to disentangle their advantage from other experimental factors including the feature extractor architecture, pre-trained initialization and fine-tuning algorithm, amongst others. In our paper, we conduct a large-scale, experimentally consistent, empirical analysis to study PEFTs for few-shot image classification. Through a battery of over 1.8k controlled experiments on large-scale few-shot benchmarks including Meta-Dataset (MD) and ORBIT, we uncover novel insights on PEFTs that cast light on their efficacy in fine-tuning ViTs for few-shot classification. Through our controlled empirical study, we have two main findings: (i) Fine-tuning just the LayerNorm parameters (which we call LN-Tune) during few-shot adaptation is an extremely strong baseline across ViTs pre-trained with both self-supervised and supervised objectives, (ii) For self-supervised ViTs, we find that simply learning a set of scaling parameters for each attention matrix (which we call AttnScale) along with a domain-residual adapter (DRA) module leads to state-of-the-art performance (while being $\\sim\\!$ 9$\\times$ more parameter-efficient) on MD. Our extensive empirical findings set strong baselines and call for rethinking the current design of PEFT methods for FSC."
                    },
                    {
                        "title": "Localizing and Editing Knowledge in Text-to-Image Generative Models",
                        "abstract": "Text-to-Image Diffusion Models such as Stable-Diffusion and Imagen have achieved unprecedented quality of photorealism with state-of-the-art FID scores on MS-COCO and other generation benchmarks. Given a caption, image generation requires fine-grained knowledge about attributes such as object structure, style, and viewpoint amongst others. Where does this information reside in text-to-image generative models? In our paper, we tackle this question and understand how knowledge corresponding to distinct visual attributes is stored in large-scale text-to-image diffusion models. We adapt Causal Mediation Analysis for text-to-image models and trace knowledge about distinct visual attributes to various (causal) components in the (i) UNet and (ii) text-encoder of the diffusion model. In particular, we show that unlike generative large-language models, knowledge about different attributes is not localized in isolated components, but is instead distributed amongst a set of components in the conditional UNet. These sets of components are often distinct for different visual attributes. Remarkably, we find that the CLIP text-encoder in public text-to-image models such as Stable-Diffusion contains only one causal state across different visual attributes, and this is the first self-attention layer corresponding to the last subject token of the attribute in the caption. This is in stark contrast to the causal states in other language models which are often the mid-MLP layers. Based on this observation of only one causal state in the text-encoder, we introduce a fast, data-free model editing method Diff-QuickFix which can effectively edit concepts in text-to-image models. DiffQuickFix can edit (ablate) concepts in under a second with a closed-form update, providing a significant 1000x speedup and comparable editing performance to existing fine-tuning based editing methods."
                    },
                    {
                        "title": "Privacy Leakage Avoidance with Switching Ensembles",
                        "abstract": "We consider membership inference attacks, one of the main privacy issues in machine learning. These recently developed attacks have been proven successful in determining, with confidence better than a random guess, whether a given sample belongs to the dataset on which the attacked machine learning model was trained. Several approaches have been developed to mitigate this privacy leakage but the tradeoff performance implications of these defensive mechanisms (i.e., accuracy and utility of the defended machine learning model) are not well studied yet. We propose a novel approach of privacy leakage avoidance with switching ensembles (PASE), which both protects against current membership inference attacks and does that with very small accuracy penalty, while requiring acceptable increase in training and inference time. We test our PASE method, along with the the current state-of-the-art PATE approach, on three calibration image datasets and analyze their tradeoffs."
                    },
                    {
                        "title": "IntCoOp: Interpretability-Aware Vision-Language Prompt Tuning",
                        "abstract": "Image-text contrastive models such as CLIP learn transferable and robust representations for zero-shot transfer to a variety of downstream tasks. However, to obtain strong downstream performances, prompts need to be carefully curated, which can be a tedious engineering task. To address the issue of manual prompt engineering, prompt-tuning is used where a set of contextual vectors are learned by leveraging information from the training data. Despite their effectiveness, existing prompt-tuning frameworks often lack interpretability, thus limiting their ability to understand the compositional nature of images. In this work, we first identify that incorporating compositional attributes (e.g., a \"green\" tree frog) in the design of manual prompts can significantly enhance image-text alignment scores. Building upon this observation, we propose a novel and interpretable prompt-tuning method named IntCoOp, which learns to jointly align attribute-level inductive biases and class embeddings during prompt-tuning. To assess the effectiveness of our approach, we evaluate IntCoOp across two representative tasks in a few-shot learning setup: generalization to novel classes, and unseen domain shifts. Through extensive experiments across 10 downstream datasets on CLIP, we find that introducing attribute-level inductive biases leads to superior performance against state-of-the-art prompt tuning frameworks. Notably, in a 16-shot setup, IntCoOp improves CoOp by 7.35% in average performance across 10 diverse datasets."
                    },
                    {
                        "title": "Topic Segmentation in the Wild: Towards Segmentation of Semi-structured & Unstructured Chats",
                        "abstract": "Breaking down a document or a conversation into multiple contiguous segments based on its semantic structure is an important and challenging problem in NLP, which can assist many downstream tasks. However, current works on topic segmentation often focus on segmentation of structured texts. In this paper, we comprehensively analyze the generalization capabilities of state-of-the-art topic segmentation models on unstructured texts. We find that: (a) Current strategies of pre-training on a large corpus of structured text such as Wiki-727K do not help in transferability to unstructured texts. (b) Training from scratch with only a relatively small-sized dataset of the target unstructured domain improves the segmentation results by a significant margin."
                    },
                    {
                        "title": "Distilling Knowledge from Text-to-Image Generative Models Improves Visio-Linguistic Reasoning in CLIP",
                        "abstract": "Image-text contrastive models like CLIP have wide applications in zero-shot classification, image-text retrieval, and transfer learning. However, they often struggle on compositional visio-linguistic tasks (e.g., attribute-binding or object-relationships) where their performance is no better than random chance. To address this, we introduce SDS-CLIP, a lightweight and sample-efficient distillation method to enhance CLIP's compositional visio-linguistic reasoning. Our approach fine-tunes CLIP using a distillation objective borrowed from large text-to-image generative models like Stable-Diffusion, which are known for their strong visio-linguistic reasoning abilities. On the challenging Winoground benchmark, SDS-CLIP improves the visio-linguistic performance of various CLIP models by up to 7%, while on the ARO dataset, it boosts performance by up to 3%. This work underscores the potential of well-designed distillation objectives from generative models to enhance contrastive image-text models with improved visio-linguistic reasoning capabilities."
                    },
                    {
                        "title": "Understanding Information Storage and Transfer in Multi-modal Large Language Models",
                        "abstract": "Understanding the mechanisms of information storage and transfer in Transformer-based models is important for driving model understanding progress. Recent work has studied these mechanisms for Large Language Models (LLMs), revealing insights on how information is stored in a model's parameters and how information flows to and from these parameters in response to specific prompts. However, these studies have not yet been extended to Multi-modal Large Language Models (MLLMs). Given their expanding capabilities and real-world use, we start by studying one aspect of these models -- how MLLMs process information in a factual visual question answering task. We use a constraint-based formulation which views a visual question as having a set of visual or textual constraints that the model's generated answer must satisfy to be correct (e.g. What movie directed by the director in this photo has won a Golden Globe?). Under this setting, we contribute i) a method that extends causal information tracing from pure language to the multi-modal setting, and ii) VQA-Constraints, a test-bed of 9.7K visual questions annotated with constraints. We use these tools to study two open-source MLLMs, LLaVa and multi-modal Phi-2. Our key findings show that these MLLMs rely on MLP and self-attention blocks in much earlier layers for information storage, compared to LLMs whose mid-layer MLPs are more important. We also show that a consistent small subset of visual tokens output by the vision encoder are responsible for transferring information from the image to these causal blocks. We validate these mechanisms by introducing MultEdit, a model-editing algorithm that can correct errors and insert new long-tailed information into MLLMs by targeting these causal blocks."
                    },
                    {
                        "title": "Rethinking Artistic Copyright Infringements in the Era of Text-to-Image Generative Models",
                        "abstract": "Recent text-to-image generative models such as Stable Diffusion are extremely adept at mimicking and generating copyrighted content, raising concerns amongst artists that their unique styles may be improperly copied. Understanding how generative models copy \"artistic style\" is more complex than duplicating a single image, as style is comprised by a set of elements (or signature) that frequently co-occurs across a body of work, where each individual work may vary significantly. In our paper, we first reformulate the problem of \"artistic copyright infringement\" to a classification problem over image sets, instead of probing image-wise similarities. We then introduce ArtSavant, a practical (i.e., efficient and easy to understand) tool to (i) determine the unique style of an artist by comparing it to a reference dataset of works from 372 artists curated from WikiArt, and (ii) recognize if the identified style reappears in generated images. We leverage two complementary methods to perform artistic style classification over image sets, includingTagMatch, which is a novel inherently interpretable and attributable method, making it more suitable for broader use by non-technical stake holders (artists, lawyers, judges, etc). Leveraging ArtSavant, we then perform a large-scale empirical study to provide quantitative insight on the prevalence of artistic style copying across 3 popular text-to-image generative models. Namely, amongst a dataset of prolific artists (including many famous ones), only 20% of them appear to have their styles be at a risk of copying via simple prompting of today's popular text-to-image generative models."
                    },
                    {
                        "title": "Understanding and Mitigating Compositional Issues in Text-to-Image Generative Models",
                        "abstract": "Recent text-to-image diffusion-based generative models have the stunning ability to generate highly detailed and photo-realistic images and achieve state-of-the-art low FID scores on challenging image generation benchmarks. However, one of the primary failure modes of these text-to-image generative models is in composing attributes, objects, and their associated relationships accurately into an image. In our paper, we investigate this compositionality-based failure mode and highlight that imperfect text conditioning with CLIP text-encoder is one of the primary reasons behind the inability of these models to generate high-fidelity compositional scenes. In particular, we show that (i) there exists an optimal text-embedding space that can generate highly coherent compositional scenes which shows that the output space of the CLIP text-encoder is sub-optimal, and (ii) we observe that the final token embeddings in CLIP are erroneous as they often include attention contributions from unrelated tokens in compositional prompts. Our main finding shows that the best compositional improvements can be achieved (without harming the model's FID scores) by fine-tuning {\\it only} a simple linear projection on CLIP's representation space in Stable-Diffusion variants using a small set of compositional image-text pairs. This result demonstrates that the sub-optimality of the CLIP's output space is a major error source. We also show that re-weighting the erroneous attention contributions in CLIP can also lead to improved compositional performances, however these improvements are often less significant than those achieved by solely learning a linear projection head, highlighting erroneous attentions to be only a minor error source."
                    },
                    {
                        "title": "On Mechanistic Knowledge Localization in Text-to-Image Generative Models",
                        "abstract": "Identifying layers within text-to-image models which control visual attributes can facilitate efficient model editing through closed-form updates. Recent work, leveraging causal tracing show that early Stable-Diffusion variants confine knowledge primarily to the first layer of the CLIP text-encoder, while it diffuses throughout the UNet.Extending this framework, we observe that for recent models (e.g., SD-XL, DeepFloyd), causal tracing fails in pinpointing localized knowledge, highlighting challenges in model editing. To address this issue, we introduce the concept of Mechanistic Localization in text-to-image models, where knowledge about various visual attributes (e.g., \"style\", \"objects\", \"facts\") can be mechanistically localized to a small fraction of layers in the UNet, thus facilitating efficient model editing. We localize knowledge using our method LocoGen which measures the direct effect of intermediate layers to output generation by performing interventions in the cross-attention layers of the UNet. We then employ LocoEdit, a fast closed-form editing method across popular open-source text-to-image models (including the latest SD-XL)and explore the possibilities of neuron-level model editing. Using Mechanistic Localization, our work offers a better view of successes and failures in localization-based text-to-image model editing. Code will be available at https://github.com/samyadeepbasu/LocoGen."
                    },
                    {
                        "title": "EditVal: Benchmarking Diffusion Based Text-Guided Image Editing Methods",
                        "abstract": "A plethora of text-guided image editing methods have recently been developed by leveraging the impressive capabilities of large-scale diffusion-based generative models such as Imagen and Stable Diffusion. A standardized evaluation protocol, however, does not exist to compare methods across different types of fine-grained edits. To address this gap, we introduce EditVal, a standardized benchmark for quantitatively evaluating text-guided image editing methods. EditVal consists of a curated dataset of images, a set of editable attributes for each image drawn from 13 possible edit types, and an automated evaluation pipeline that uses pre-trained vision-language models to assess the fidelity of generated images for each edit type. We use EditVal to benchmark 8 cutting-edge diffusion-based editing methods including SINE, Imagic and Instruct-Pix2Pix. We complement this with a large-scale human study where we show that EditVall's automated evaluation pipeline is strongly correlated with human-preferences for the edit types we considered. From both the human study and automated evaluation, we find that: (i) Instruct-Pix2Pix, Null-Text and SINE are the top-performing methods averaged across different edit types, however {\\it only} Instruct-Pix2Pix and Null-Text are able to preserve original image properties; (ii) Most of the editing methods fail at edits involving spatial operations (e.g., changing the position of an object). (iii) There is no `winner' method which ranks the best individually across a range of different edit types. We hope that our benchmark can pave the way to developing more reliable text-guided image editing tools in the future. We will publicly release EditVal, and all associated code and human-study templates to support these research directions in https://deep-ml-research.github.io/editval/."
                    },
                    {
                        "title": "On Surgical Fine-tuning for Language Encoders",
                        "abstract": "Fine-tuning all the layers of a pre-trained neural language encoder (either using all the parameters or using parameter-efficient methods) is often the de-facto way of adapting it to a new task. We show evidence that for different downstream language tasks, fine-tuning only a subset of layers is sufficient to obtain performance that is close to and often better than fine-tuning all the layers in the language encoder. We propose an efficient metric based on the diagonal of the Fisher information matrix (FIM score), to select the candidate layers for selective fine-tuning. We show, empirically on GLUE and SuperGLUE tasks and across distinct language encoders, that this metric can effectively select layers leading to a strong downstream performance. Our work highlights that task-specific information corresponding to a given downstream task is often localized within a few layers, and tuning only those is sufficient for strong performance. Additionally, we demonstrate the robustness of the FIM score to rank layers in a manner that remains constant during the optimization process."
                    },
                    {
                        "title": "Topic Segmentation of Semi-Structured and Unstructured Conversational Datasets using Language Models",
                        "abstract": "Breaking down a document or a conversation into multiple contiguous segments based on its semantic structure is an important and challenging problem in NLP, which can assist many downstream tasks. However, current works on topic segmentation often focus on segmentation of structured texts. In this paper, we comprehensively analyze the generalization capabilities of state-of-the-art topic segmentation models on unstructured texts. We find that: (a) Current strategies of pre-training on a large corpus of structured text such as Wiki-727K do not help in transferability to unstructured conversational data. (b) Training from scratch with only a relatively small-sized dataset of the target unstructured domain improves the segmentation results by a significant margin. We stress-test our proposed Topic Segmentation approach by experimenting with multiple loss functions, in order to mitigate effects of imbalance in unstructured conversational datasets. Our empirical evaluation indicates that Focal Loss function is a robust alternative to Cross-Entropy and re-weighted Cross-Entropy loss function when segmenting unstructured and semi-structured chats."
                    },
                    {
                        "title": "From RAGs to rich parameters: Probing how language models utilize external knowledge over parametric information for factual queries",
                        "abstract": "Retrieval Augmented Generation (RAG) enriches the ability of language models to reason using external context to augment responses for a given user prompt. This approach has risen in popularity due to practical applications in various applications of language models in search, question/answering, and chat-bots. However, the exact nature of how this approach works isn't clearly understood. In this paper, we mechanistically examine the RAG pipeline to highlight that language models take shortcut and have a strong bias towards utilizing only the context information to answer the question, while relying minimally on their parametric memory. We probe this mechanistic behavior in language models with: (i) Causal Mediation Analysis to show that the parametric memory is minimally utilized when answering a question and (ii) Attention Contributions and Knockouts to show that the last token residual stream do not get enriched from the subject token in the question, but gets enriched from other informative tokens in the context. We find this pronounced shortcut behaviour true across both LLaMa and Phi family of models."
                    },
                    {
                        "title": "Quantifying reliance on external information over parametric knowledge during Retrieval Augmented Generation (RAG) using mechanistic analysis",
                        "abstract": "Retrieval Augmented Generation (RAG) is a widely used approach for leveraging external context in several natural language applications such as question answering and information retrieval. Yet, the exact nature in which a Language Model (LM) leverages this non-parametric memory or retrieved context isn't clearly understood. This paper mechanistically examines the RAG pipeline to highlight that LMs demonstrate a \"shortcut'' effect and have a strong bias towards utilizing the retrieved context to answer questions, while relying minimally on model priors. We propose (a) Causal Mediation Analysis; for proving that parametric memory is minimally utilized when answering a question and (b) Attention Contributions and Knockouts for showing the last token residual stream do not get enriched from the subject token in the question, but gets enriched from tokens of RAG-context. We find this pronounced \"shortcut'' behaviour to be true across both LLMs (e.g.,LlaMa) and SLMs (e.g., Phi)"
                    }
                ]
            },
            "60c36564-06e5-430d-8575-9f26a439b161": {
                "pk": "60c36564-06e5-430d-8575-9f26a439b161",
                "name": "Soheil Feizi",
                "collaborators": [
                    "Alexander Levine",
                    "Sahil Singla",
                    "Muriel Medard",
                    "Cassidy Laidlaw",
                    "David Tse",
                    "Samuel Barham",
                    "Mazda Moayeri",
                    "Priyatham Kattakinda",
                    "Wenxiao Wang"
                ],
                "domain": [
                    "Adversarial Robustness",
                    "Machine Learning",
                    "Data Poisoning",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Provable Adversarial Robustness for Fractional Lp Threat Models",
                        "abstract": "In recent years, researchers have extensively studied adversarial robustness in a variety of threat models, including L_0, L_1, L_2, and L_infinity-norm bounded adversarial attacks. However, attacks bounded by fractional L_p \"norms\" (quasi-norms defined by the L_p distance with 0<p<1) have yet to be thoroughly considered. We proactively propose a defense with several desirable properties: it provides provable (certified) robustness, scales to ImageNet, and yields deterministic (rather than high-probability) certified guarantees when applied to quantized data (e.g., images). Our technique for fractional L_p robustness constructs expressive, deep classifiers that are globally Lipschitz with respect to the L_p^p metric, for any 0<p<1. However, our method is even more general: we can construct classifiers which are globally Lipschitz with respect to any metric defined as the sum of concave functions of components. Our approach builds on a recent work, Levine and Feizi (2021), which provides a provable defense against L_1 attacks. However, we demonstrate that our proposed guarantees are highly non-vacuous, compared to the trivial solution of using (Levine and Feizi, 2021) directly and applying norm inequalities. Code is available at https://github.com/alevine0/fractionalLpRobustness."
                    },
                    {
                        "title": "A Power Efficient Sensing/Communication Scheme: Joint Source-Channel-Network Coding by Using Compressive Sensing",
                        "abstract": "We propose a joint source-channel-network coding scheme, based on compressive sensing principles, for wireless networks with AWGN channels (that may include multiple access and broadcast), with sources exhibiting temporal and spatial dependencies. Our goal is to provide a reconstruction of sources within an allowed distortion level at each receiver. We perform joint source-channel coding at each source by randomly projecting source values to a lower dimensional space. We consider sources that satisfy the restricted eigenvalue (RE) condition as well as more general sources for which the randomness of the network allows a mapping to lower dimensional spaces. Our approach relies on using analog random linear network coding. The receiver uses compressive sensing decoders to reconstruct sources. Our key insight is the fact that, compressive sensing and analog network coding both preserve the source characteristics required for compressive sensing decoding."
                    },
                    {
                        "title": "Robustness Certificates Against Adversarial Examples for ReLU Networks",
                        "abstract": "While neural networks have achieved high performance in different learning tasks, their accuracy drops significantly in the presence of small adversarial perturbations to inputs. Defenses based on regularization and adversarial training are often followed by new attacks to defeat them. In this paper, we propose attack-agnostic robustness certificates for a multi-label classification problem using a deep ReLU network. Although computing the exact distance of a given input sample to the classification decision boundary requires solving a non-convex optimization, we characterize two lower bounds for such distances, namely the simplex certificate and the decision boundary certificate. These robustness certificates leverage the piece-wise linear structure of ReLU networks and use the fact that in a polyhedron around a given sample, the prediction function is linear. In particular, the proposed simplex certificate has a closed-form, is differentiable and is an order of magnitude faster to compute than the existing methods even for deep networks. In addition to theoretical bounds, we provide numerical results for our certificates over MNIST and compare them with some existing upper bounds."
                    },
                    {
                        "title": "On Network Functional Compression",
                        "abstract": "In this paper, we consider different aspects of the network functional compression problem where computation of a function (or, some functions) of sources located at certain nodes in a network is desired at receiver(s). The rate region of this problem has been considered in the literature under certain restrictive assumptions, particularly in terms of the network topology, the functions and the characteristics of the sources. In this paper, we present results that significantly relax these assumptions. Firstly, we consider this problem for an arbitrary tree network and asymptotically lossless computation. We show that, for depth one trees with correlated sources, or for general trees with independent sources, a modularized coding scheme based on graph colorings and Slepian-Wolf compression performs arbitrarily closely to rate lower bounds. For a general tree network with independent sources, optimal computation to be performed at intermediate nodes is derived. We introduce a necessary and sufficient condition on graph colorings of any achievable coding scheme, called coloring connectivity condition (C.C.C.).   Secondly, we investigate the effect of having several functions at the receiver. In this problem, we derive a rate region and propose a coding scheme based on graph colorings. Thirdly, we consider the functional compression problem with feedback. We show that, in this problem, unlike Slepian-Wolf compression, by having feedback, one may outperform rate bounds of the case without feedback. Fourthly, we investigate functional computation problem with distortion. We compute a rate-distortion region for this problem. Then, we propose a simple suboptimal coding scheme with a non-trivial performance guarantee. Finally, we introduce cases where finding minimum entropy colorings and therefore, optimal coding schemes can be performed in polynomial time."
                    },
                    {
                        "title": "Maximally Correlated Principal Component Analysis",
                        "abstract": "In the era of big data, reducing data dimensionality is critical in many areas of science. Widely used Principal Component Analysis (PCA) addresses this problem by computing a low dimensional data embedding that maximally explain variance of the data. However, PCA has two major weaknesses. Firstly, it only considers linear correlations among variables (features), and secondly it is not suitable for categorical data. We resolve these issues by proposing Maximally Correlated Principal Component Analysis (MCPCA). MCPCA computes transformations of variables whose covariance matrix has the largest Ky Fan norm. Variable transformations are unknown, can be nonlinear and are computed in an optimization. MCPCA can also be viewed as a multivariate extension of Maximal Correlation. For jointly Gaussian variables we show that the covariance matrix corresponding to the identity (or the negative of the identity) transformations majorizes covariance matrices of non-identity functions. Using this result we characterize global MCPCA optimizers for nonlinear functions of jointly Gaussian variables for every rank constraint. For categorical variables we characterize global MCPCA optimizers for the rank one constraint based on the leading eigenvector of a matrix computed using pairwise joint distributions. For a general rank constraint we propose a block coordinate descend algorithm and show its convergence to stationary points of the MCPCA optimization. We compare MCPCA with PCA and other state-of-the-art dimensionality reduction methods including Isomap, LLE, multilayer autoencoders (neural networks), kernel PCA, probabilistic PCA and diffusion maps on several synthetic and real datasets. We show that MCPCA consistently provides improved performance compared to other methods."
                    },
                    {
                        "title": "Interpretable Adversarial Training for Text",
                        "abstract": "Generating high-quality and interpretable adversarial examples in the text domain is a much more daunting task than it is in the image domain. This is due partly to the discrete nature of text, partly to the problem of ensuring that the adversarial examples are still probable and interpretable, and partly to the problem of maintaining label invariance under input perturbations. In order to address some of these challenges, we introduce sparse projected gradient descent (SPGD), a new approach to crafting interpretable adversarial examples for text. SPGD imposes a directional regularization constraint on input perturbations by projecting them onto the directions to nearby word embeddings with highest cosine similarities. This constraint ensures that perturbations move each word embedding in an interpretable direction (i.e., towards another nearby word embedding). Moreover, SPGD imposes a sparsity constraint on perturbations at the sentence level by ignoring word-embedding perturbations whose norms are below a certain threshold. This constraint ensures that our method changes only a few words per sequence, leading to higher quality adversarial examples. Our experiments with the IMDB movie review dataset show that the proposed SPGD method improves adversarial example interpretability and likelihood (evaluated by average per-word perplexity) compared to state-of-the-art methods, while suffering little to no loss in training performance."
                    },
                    {
                        "title": "Wasserstein Smoothing: Certified Robustness against Wasserstein Adversarial Attacks",
                        "abstract": "In the last couple of years, several adversarial attack methods based on different threat models have been proposed for the image classification problem. Most existing defenses consider additive threat models in which sample perturbations have bounded L_p norms. These defenses, however, can be vulnerable against adversarial attacks under non-additive threat models. An example of an attack method based on a non-additive threat model is the Wasserstein adversarial attack proposed by Wong et al. (2019), where the distance between an image and its adversarial example is determined by the Wasserstein metric (\"earth-mover distance\") between their normalized pixel intensities. Until now, there has been no certifiable defense against this type of attack. In this work, we propose the first defense with certified robustness against Wasserstein Adversarial attacks using randomized smoothing. We develop this certificate by considering the space of possible flows between images, and representing this space such that Wasserstein distance between images is upper-bounded by L_1 distance in this flow-space. We can then apply existing randomized smoothing certificates for the L_1 metric. In MNIST and CIFAR-10 datasets, we find that our proposed defense is also practically effective, demonstrating significantly improved accuracy under Wasserstein adversarial attack compared to unprotected models."
                    },
                    {
                        "title": "Skew Orthogonal Convolutions",
                        "abstract": "Training convolutional neural networks with a Lipschitz constraint under the $l_{2}$ norm is useful for provable adversarial robustness, interpretable gradients, stable training, etc. While 1-Lipschitz networks can be designed by imposing a 1-Lipschitz constraint on each layer, training such networks requires each layer to be gradient norm preserving (GNP) to prevent gradients from vanishing. However, existing GNP convolutions suffer from slow training, lead to significant reduction in accuracy and provide no guarantees on their approximations. In this work, we propose a GNP convolution layer called Skew Orthogonal Convolution (SOC) that uses the following mathematical property: when a matrix is {\\it Skew-Symmetric}, its exponential function is an {\\it orthogonal} matrix. To use this property, we first construct a convolution filter whose Jacobian is Skew-Symmetric. Then, we use the Taylor series expansion of the Jacobian exponential to construct the SOC layer that is orthogonal. To efficiently implement SOC, we keep a finite number of terms from the Taylor series and provide a provable guarantee on the approximation error. Our experiments on CIFAR-10 and CIFAR-100 show that SOC allows us to train provably Lipschitz, large convolutional neural networks significantly faster than prior works while achieving significant improvements for both standard and certified robust accuracies."
                    },
                    {
                        "title": "Playing it Safe: Adversarial Robustness with an Abstain Option",
                        "abstract": "We explore adversarial robustness in the setting in which it is acceptable for a classifier to abstain---that is, output no class---on adversarial examples. Adversarial examples are small perturbations of normal inputs to a classifier that cause the classifier to give incorrect output; they present security and safety challenges for machine learning systems. In many safety-critical applications, it is less costly for a classifier to abstain on adversarial examples than to give incorrect output for them. We first introduce a novel objective function for adversarial robustness with an abstain option which characterizes an explicit tradeoff between robustness and accuracy. We then present a simple baseline in which an adversarially-trained classifier abstains on all inputs within a certain distance of the decision boundary, which we theoretically and experimentally evaluate. Finally, we propose Combined Abstention Robustness Learning (CARL), a method for jointly learning a classifier and the region of the input space on which it should abstain. We explore different variations of the PGD and DeepFool adversarial attacks on CARL in the abstain setting. Evaluating against these attacks, we demonstrate that training with CARL results in a more accurate, robust, and efficient classifier than the baseline."
                    },
                    {
                        "title": "Improved techniques for deterministic l2 robustness",
                        "abstract": "Training convolutional neural networks (CNNs) with a strict 1-Lipschitz constraint under the $l_{2}$ norm is useful for adversarial robustness, interpretable gradients and stable training. 1-Lipschitz CNNs are usually designed by enforcing each layer to have an orthogonal Jacobian matrix (for all inputs) to prevent the gradients from vanishing during backpropagation. However, their performance often significantly lags behind that of heuristic methods to enforce Lipschitz constraints where the resulting CNN is not \\textit{provably} 1-Lipschitz. In this work, we reduce this gap by introducing (a) a procedure to certify robustness of 1-Lipschitz CNNs by replacing the last linear layer with a 1-hidden layer MLP that significantly improves their performance for both standard and provably robust accuracy, (b) a method to significantly reduce the training time per epoch for Skew Orthogonal Convolution (SOC) layers (>30\\% reduction for deeper networks) and (c) a class of pooling layers using the mathematical property that the $l_{2}$ distance of an input to a manifold is 1-Lipschitz. Using these methods, we significantly advance the state-of-the-art for standard and provable robust accuracies on CIFAR-10 (gains of +1.79\\% and +3.82\\%) and similarly on CIFAR-100 (+3.78\\% and +4.75\\%) across all networks. Code is available at \\url{https://github.com/singlasahil14/improved_l2_robustness}."
                    },
                    {
                        "title": "Functional Adversarial Attacks",
                        "abstract": "We propose functional adversarial attacks, a novel class of threat models for crafting adversarial examples to fool machine learning models. Unlike a standard $\\ell_p$-ball threat model, a functional adversarial threat model allows only a single function to be used to perturb input features to produce an adversarial example. For example, a functional adversarial attack applied on colors of an image can change all red pixels simultaneously to light red. Such global uniform changes in images can be less perceptible than perturbing pixels of the image individually. For simplicity, we refer to functional adversarial attacks on image colors as ReColorAdv, which is the main focus of our experiments. We show that functional threat models can be combined with existing additive ($\\ell_p$) threat models to generate stronger threat models that allow both small, individual perturbations and large, uniform changes to an input. Moreover, we prove that such combinations encompass perturbations that would not be allowed in either constituent threat model. In practice, ReColorAdv can significantly reduce the accuracy of a ResNet-32 trained on CIFAR-10. Furthermore, to the best of our knowledge, combining ReColorAdv with other attacks leads to the strongest existing attack even after adversarial training. An implementation of ReColorAdv is available at https://github.com/cassidylaidlaw/ReColorAdv ."
                    },
                    {
                        "title": "(De)Randomized Smoothing for Certifiable Defense against Patch Attacks",
                        "abstract": "Patch adversarial attacks on images, in which the attacker can distort pixels within a region of bounded size, are an important threat model since they provide a quantitative model for physical adversarial attacks. In this paper, we introduce a certifiable defense against patch attacks that guarantees for a given image and patch attack size, no patch adversarial examples exist. Our method is related to the broad class of randomized smoothing robustness schemes which provide high-confidence probabilistic robustness certificates. By exploiting the fact that patch attacks are more constrained than general sparse attacks, we derive meaningfully large robustness certificates against them. Additionally, in contrast to smoothing-based defenses against L_p and sparse attacks, our defense method against patch attacks is de-randomized, yielding improved, deterministic certificates. Compared to the existing patch certification method proposed by Chiang et al. (2020), which relies on interval bound propagation, our method can be trained significantly faster, achieves high clean and certified robust accuracy on CIFAR-10, and provides certificates at ImageNet scale. For example, for a 5-by-5 patch attack on CIFAR-10, our method achieves up to around 57.6% certified accuracy (with a classifier with around 83.8% clean accuracy), compared to at most 30.3% certified accuracy for the existing method (with a classifier with around 47.8% clean accuracy). Our results effectively establish a new state-of-the-art of certifiable defense against patch attacks on CIFAR-10 and ImageNet. Code is available at https://github.com/alevine0/patchSmoothing."
                    },
                    {
                        "title": "Second-Order Provable Defenses against Adversarial Attacks",
                        "abstract": "A robustness certificate is the minimum distance of a given input to the decision boundary of the classifier (or its lower bound). For {\\it any} input perturbations with a magnitude smaller than the certificate value, the classification output will provably remain unchanged. Exactly computing the robustness certificates for neural networks is difficult since it requires solving a non-convex optimization. In this paper, we provide computationally-efficient robustness certificates for neural networks with differentiable activation functions in two steps. First, we show that if the eigenvalues of the Hessian of the network are bounded, we can compute a robustness certificate in the $l_2$ norm efficiently using convex optimization. Second, we derive a computationally-efficient differentiable upper bound on the curvature of a deep network. We also use the curvature bound as a regularization term during the training of the network to boost its certified robustness. Putting these results together leads to our proposed {\\bf C}urvature-based {\\bf R}obustness {\\bf C}ertificate (CRC) and {\\bf C}urvature-based {\\bf R}obust {\\bf T}raining (CRT). Our numerical results show that CRT leads to significantly higher certified robust accuracy compared to interval-bound propagation (IBP) based training. We achieve certified robust accuracy 69.79\\%, 57.78\\% and 53.19\\% while IBP-based methods achieve 44.96\\%, 44.74\\% and 44.66\\% on 2,3 and 4 layer networks respectively on the MNIST-dataset."
                    },
                    {
                        "title": "Sample Efficient Detection and Classification of Adversarial Attacks via Self-Supervised Embeddings",
                        "abstract": "Adversarial robustness of deep models is pivotal in ensuring safe deployment in real world settings, but most modern defenses have narrow scope and expensive costs. In this paper, we propose a self-supervised method to detect adversarial attacks and classify them to their respective threat models, based on a linear model operating on the embeddings from a pre-trained self-supervised encoder. We use a SimCLR encoder in our experiments, since we show the SimCLR embedding distance is a good proxy for human perceptibility, enabling it to encapsulate many threat models at once. We call our method SimCat since it uses SimCLR encoder to catch and categorize various types of adversarial attacks, including L_p and non-L_p evasion attacks, as well as data poisonings. The simple nature of a linear classifier makes our method efficient in both time and sample complexity. For example, on SVHN, using only five pairs of clean and adversarial examples computed with a PGD-L_inf attack, SimCat's detection accuracy is over 85%. Moreover, on ImageNet, using only 25 examples from each threat model, SimCat can classify eight different attack types such as PGD-L_2, PGD-L_inf, CW-L_2, PPGD, LPA, StAdv, ReColor, and JPEG-L_inf, with over 40% accuracy. On STL10 data, we apply SimCat as a defense against poisoning attacks, such as BP, CP, FC, CLBD, HTBD, halving the success rate while using only twenty total poisons for training. We find that the detectors generalize well to unseen threat models. Lastly, we investigate the performance of our detection method under adaptive attacks and further boost its robustness against such attacks via adversarial training."
                    },
                    {
                        "title": "Robustness Certificates for Sparse Adversarial Attacks by Randomized Ablation",
                        "abstract": "Recently, techniques have been developed to provably guarantee the robustness of a classifier to adversarial perturbations of bounded L_1 and L_2 magnitudes by using randomized smoothing: the robust classification is a consensus of base classifications on randomly noised samples where the noise is additive. In this paper, we extend this technique to the L_0 threat model. We propose an efficient and certifiably robust defense against sparse adversarial attacks by randomly ablating input features, rather than using additive noise. Experimentally, on MNIST, we can certify the classifications of over 50% of images to be robust to any distortion of at most 8 pixels. This is comparable to the observed empirical robustness of unprotected classifiers on MNIST to modern L_0 attacks, demonstrating the tightness of the proposed robustness certificate. We also evaluate our certificate on ImageNet and CIFAR-10. Our certificates represent an improvement on those provided in a concurrent work (Lee et al. 2019) which uses random noise rather than ablation (median certificates of 8 pixels versus 4 pixels on MNIST; 16 pixels versus 1 pixel on ImageNet.) Additionally, we empirically demonstrate that our classifier is highly robust to modern sparse adversarial attacks on MNIST. Our classifications are robust, in median, to adversarial perturbations of up to 31 pixels, compared to 22 pixels reported as the state-of-the-art defense, at the cost of a slight decrease (around 2.3%) in the classification accuracy. Code is available at https://github.com/alevine0/randomizedAblation/."
                    },
                    {
                        "title": "Fantastic Four: Differentiable Bounds on Singular Values of Convolution Layers",
                        "abstract": "In deep neural networks, the spectral norm of the Jacobian of a layer bounds the factor by which the norm of a signal changes during forward/backward propagation. Spectral norm regularizations have been shown to improve generalization, robustness and optimization of deep learning methods. Existing methods to compute the spectral norm of convolution layers either rely on heuristics that are efficient in computation but lack guarantees or are theoretically-sound but computationally expensive. In this work, we obtain the best of both worlds by deriving {\\it four} provable upper bounds on the spectral norm of a standard 2D multi-channel convolution layer. These bounds are differentiable and can be computed efficiently during training with negligible overhead. One of these bounds is in fact the popular heuristic method of Miyato et al. (multiplied by a constant factor depending on filter sizes). Each of these four bounds can achieve the tightest gap depending on convolution filters. Thus, we propose to use the minimum of these four bounds as a tight, differentiable and efficient upper bound on the spectral norm of convolution layers. We show that our spectral bound is an effective regularizer and can be used to bound either the lipschitz constant or curvature values (eigenvalues of the Hessian) of neural networks. Through experiments on MNIST and CIFAR-10, we demonstrate the effectiveness of our spectral bound in improving generalization and provable robustness of deep networks."
                    },
                    {
                        "title": "Improved, Deterministic Smoothing for L_1 Certified Robustness",
                        "abstract": "Randomized smoothing is a general technique for computing sample-dependent robustness guarantees against adversarial attacks for deep classifiers. Prior works on randomized smoothing against L_1 adversarial attacks use additive smoothing noise and provide probabilistic robustness guarantees. In this work, we propose a non-additive and deterministic smoothing method, Deterministic Smoothing with Splitting Noise (DSSN). To develop DSSN, we first develop SSN, a randomized method which involves generating each noisy smoothing sample by first randomly splitting the input space and then returning a representation of the center of the subdivision occupied by the input sample. In contrast to uniform additive smoothing, the SSN certification does not require the random noise components used to be independent. Thus, smoothing can be done effectively in just one dimension and can therefore be efficiently derandomized for quantized data (e.g., images). To the best of our knowledge, this is the first work to provide deterministic \"randomized smoothing\" for a norm-based adversarial threat model while allowing for an arbitrary classifier (i.e., a deep model) to be used as a base classifier and without requiring an exponential number of smoothing samples. On CIFAR-10 and ImageNet datasets, we provide substantially larger L_1 robustness certificates compared to prior works, establishing a new state-of-the-art. The determinism of our method also leads to significantly faster certificate computation. Code is available at: https://github.com/alevine0/smoothingSplittingNoise"
                    },
                    {
                        "title": "FOCUS: Familiar Objects in Common and Uncommon Settings",
                        "abstract": "Standard training datasets for deep learning often contain objects in common settings (e.g., \"a horse on grass\" or \"a ship in water\") since they are usually collected by randomly scraping the web. Uncommon and rare settings (e.g., \"a plane on water\", \"a car in snowy weather\") are thus severely under-represented in the training data. This can lead to an undesirable bias in model predictions towards common settings and create a false sense of accuracy. In this paper, we introduce FOCUS (Familiar Objects in Common and Uncommon Settings), a dataset for stress-testing the generalization power of deep image classifiers. By leveraging the power of modern search engines, we deliberately gather data containing objects in common and uncommon settings in a wide range of locations, weather conditions, and time of day. We present a detailed analysis of the performance of various popular image classifiers on our dataset and demonstrate a clear drop in performance when classifying images in uncommon settings. By analyzing deep features of these models, we show that such errors can be due to the use of spurious features in model predictions. We believe that our dataset will aid researchers in understanding the inability of deep models to generalize well to uncommon settings and drive future work on improving their distributional robustness."
                    },
                    {
                        "title": "Goal-Conditioned Q-Learning as Knowledge Distillation",
                        "abstract": "Many applications of reinforcement learning can be formalized as goal-conditioned environments, where, in each episode, there is a \"goal\" that affects the rewards obtained during that episode but does not affect the dynamics. Various techniques have been proposed to improve performance in goal-conditioned environments, such as automatic curriculum generation and goal relabeling. In this work, we explore a connection between off-policy reinforcement learning in goal-conditioned settings and knowledge distillation. In particular: the current Q-value function and the target Q-value estimate are both functions of the goal, and we would like to train the Q-value function to match its target for all goals. We therefore apply Gradient-Based Attention Transfer (Zagoruyko and Komodakis 2017), a knowledge distillation technique, to the Q-function update. We empirically show that this can improve the performance of goal-conditioned off-policy reinforcement learning when the space of goals is high-dimensional. We also show that this technique can be adapted to allow for efficient learning in the case of multiple simultaneous sparse goals, where the agent can attain a reward by achieving any one of a large set of objectives, all specified at test time. Finally, to provide theoretical support, we give examples of classes of environments where (under some assumptions) standard off-policy algorithms such as DDPG require at least O(d^2) replay buffer transitions to learn an optimal policy, while our proposed technique requires only O(d) transitions, where d is the dimensionality of the goal and state space. Code is available at https://github.com/alevine0/ReenGAGE."
                    },
                    {
                        "title": "Temporal Robustness against Data Poisoning",
                        "abstract": "Data poisoning considers cases when an adversary manipulates the behavior of machine learning algorithms through malicious training data. Existing threat models of data poisoning center around a single metric, the number of poisoned samples. In consequence, if attackers can poison more samples than expected with affordable overhead, as in many practical scenarios, they may be able to render existing defenses ineffective in a short time. To address this issue, we leverage timestamps denoting the birth dates of data, which are often available but neglected in the past. Benefiting from these timestamps, we propose a temporal threat model of data poisoning with two novel metrics, earliness and duration, which respectively measure how long an attack started in advance and how long an attack lasted. Using these metrics, we define the notions of temporal robustness against data poisoning, providing a meaningful sense of protection even with unbounded amounts of poisoned samples when the attacks are temporally bounded. We present a benchmark with an evaluation protocol simulating continuous data collection and periodic deployments of updated models, thus enabling empirical evaluation of temporal robustness. Lastly, we develop and also empirically verify a baseline defense, namely temporal aggregation, offering provable temporal robustness and highlighting the potential of our temporal threat model for data poisoning."
                    }
                ]
            }
        }
    },
    "2405.19279": {
        "paper_data": {
            "title": "Understanding and Minimising Outlier Features in Neural Network Training",
            "url": "http://arxiv.org/abs/2405.19279v1",
            "arxiv_id": "2405.19279",
            "authors": [
                "Bobby He",
                "Lorenzo Noci",
                "Daniele Paliotta",
                "Imanol Schlag",
                "Thomas Hofmann"
            ],
            "abstract": "Outlier Features (OF) are neurons whose activation magnitudes significantly exceed the average over a neural network's (NN) width. They are well known to emerge during standard transformer training and have the undesirable effect of hindering quantisation in afflicted models. Despite their practical importance, little is known behind why OFs emerge during training, nor how one can minimise them.   Our work focuses on the above questions, first identifying several quantitative metrics, such as the kurtosis over neuron activation norms, to measure OFs. With these metrics, we study how architectural and optimisation choices influence OFs, and provide practical insights to minimise OFs during training. As highlights, we emphasise the importance of controlling signal propagation throughout training, and propose the Outlier Protected transformer block, which removes standard Pre-Norm layers to mitigate OFs, without loss of convergence speed or training stability. Overall, our findings shed new light on our understanding of, our ability to prevent, and the complexity of this important facet in NN training dynamics.",
            "introduction": "   1 Introduction  Despite their widespread use, our understanding of deep neural networks (NNs) and their training dynamics is very much incomplete. This, in part, reflects the complexity of traversing high-dimensional non-convex loss landscapes but is also symptomatic of the myriad design choices, such as NN architecture and optimiser hyperparameters, that a practitioner must take before training. While standard choices of architecture and optimiser exist, it is often unclear how these choices affect model performance or the emergence of various empirically observed phenomena during NN training.   Outlier Features (OF) are one such NN training phenomenon. Intuitively, OFs are neurons whose activation magnitudes are significantly larger than the average in the same NN layer, i.e. across NN width [1, 2, 3]. They have been widely observed in pre-trained transformer models [4, 5, 6], as we verify in Fig\u00a01, and are of practical interest because their existence hinders quantisation [3, 7, 8, 9, 10, 11, 12]. In particular, OFs cause large dynamic ranges in activations across NN width, which lead to high quantisation errors in low-precision matrix multiplications. As a result, Outlier Feature Emergence (OFE) during training hinders low-precision training and inference, and minimising OFE could yield significant potential efficiency gains.   In this paper, we tackle OFE from two related angles: by (1) proposing interventions to minimise OFE without affecting model convergence or training stability, using insights motivated through (2) enhancing our understanding of why OFs appear during training. We argue that it is important to first understand why OFs appear during standard NN training dynamics in order to identify which design choices influence OFE, and how. Though progress has been made [1, 13, 10, 14], the mechanisms behind OFE remain largely unknown.   Figure 1: Outlier Features appear in open-source transformers [15] during training, as measured by our Kurtosis metric Eq\u00a01. Our work investigates the factors that influence their emergence.   Alongside the practical motivation of model quantisation, we believe understanding OFs, and their causes during training, to be an interesting research question for several reasons. The emergence of Outlier Features in standard training regimes raises the question if OFs are simply an artifact of certain design choices or a fundamental property of NN training, essential for best performance. Understanding (the causes of) OFE better helps us to better understand NN training dynamics in general, and the roles played by different design choices. Moreover, we shed light on the differences between models at initialisation compared to during, or after, training. While NN initialisation is more commonly studied owing to analytic tractability [16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29], understanding trained NNs is arguably more important as they exhibit rich feature learning behaviour [30, 31, 32, 33], like OFs, that arise during training. In the case of OFs, this has potentially wide-reaching implications, including for NN interpretability, which often focuses on the roles of individual neurons in trained NNs [34].   Our contributions  Overall, we show that OFE can be mitigated relative to standard practices, and highlight key design choices to do so. We start by introducing OFs, and in particular quantitative metrics to measure OFs in Sec\u00a02.",
            "references": [
                {
                    "title": "Mitigating the Impact of Outlier Channels for Language Model Quantization with Activation Regularization",
                    "abstract": "We consider the problem of accurate quantization for language models, where both the weights and activations are uniformly quantized to 4 bits per parameter, the lowest bitwidth format natively supported by GPU hardware. In this context, the key challenge is activation quantization: it is known that language models contain outlier channels whose values on average are orders of magnitude higher than than other channels, which prevents accurate low-bitwidth quantization with known techniques. We systematically study this phenomena and find that these outlier channels emerge early in training, and that they occur more frequently in layers with residual streams. We then propose a simple strategy which regularizes a layer's inputs via quantization-aware training (QAT) and its outputs via activation kurtosis regularization. We show that regularizing both the inputs and outputs is crucial for preventing a model's\"migrating\"the difficulty in input quantization to the weights, which makes post-training quantization (PTQ) of weights more difficult. When combined with weight PTQ, we show that our approach can obtain a W4A4 model that performs competitively to the standard-precision W16A16 baseline."
                },
                {
                    "title": "QuaRot: Outlier-Free 4-Bit Inference in Rotated LLMs",
                    "abstract": "We introduce QuaRot, a new Quantization scheme based on Rotations, which is able to quantize LLMs end-to-end, including all weights, activations, and KV cache in 4 bits. QuaRot rotates LLMs in a way that removes outliers from the hidden state without changing the output, making quantization easier. This computational invariance is applied to the hidden state (residual) of the LLM, as well as to the activations of the feed-forward components, aspects of the attention mechanism and to the KV cache. The result is a quantized model where all matrix multiplications are performed in 4-bits, without any channels identified for retention in higher precision. Our quantized LLaMa2-70B model has losses of at most 0.29 WikiText-2 perplexity and retains 99% of the zero-shot performance. Code is available at: https://github.com/spcl/QuaRot."
                },
                {
                    "title": "Heavy-Tailed Class Imbalance and Why Adam Outperforms Gradient Descent on Language Models",
                    "abstract": "Adam has been shown to outperform gradient descent on large language models by a larger margin than on other tasks, but it is unclear why. We show that a key factor in this performance gap is the heavy-tailed class imbalance found in language tasks. When trained with gradient descent, the loss of infrequent words decreases more slowly than the loss of frequent ones. This leads to a slow decrease on the average loss as most samples come from infrequent words. On the other hand, Adam and sign-based methods are less sensitive to this problem. To establish that this behavior is caused by class imbalance, we show empirically that it can be reproduced across architectures and data types, on language transformers, vision CNNs, and linear models. On a linear model with cross-entropy loss, we show that class imbalance leads to imbalanced, correlated gradients and Hessians that have been hypothesized to benefit Adam. We also prove that, in continuous time, gradient descent converges slowly on low-frequency classes while sign descent does not."
                },
                {
                    "title": "Massive Activations in Large Language Models",
                    "abstract": "We observe an empirical phenomenon in Large Language Models (LLMs) -- very few activations exhibit significantly larger values than others (e.g., 100,000 times larger). We call them massive activations. First, we demonstrate the widespread existence of massive activations across various LLMs and characterize their locations. Second, we find their values largely stay constant regardless of the input, and they function as indispensable bias terms in LLMs. Third, these massive activations lead to the concentration of attention probabilities to their corresponding tokens, and further, implicit bias terms in the self-attention output. Last, we also study massive activations in Vision Transformers. Code is available at https://github.com/locuslab/massive-activations."
                },
                {
                    "title": "Mamba: Linear-Time Sequence Modeling with Selective State Spaces",
                    "abstract": "Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5$\\times$ higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation."
                },
                {
                    "title": "Simplifying Transformer Blocks",
                    "abstract": "A simple design recipe for deep Transformers is to compose identical building blocks. But standard transformer blocks are far from simple, interweaving attention and MLP sub-blocks with skip connections&normalisation layers in precise arrangements. This complexity leads to brittle architectures, where seemingly minor changes can significantly reduce training speed, or render models untrainable. In this work, we ask to what extent the standard transformer block can be simplified? Combining signal propagation theory and empirical observations, we motivate modifications that allow many block components to be removed with no loss of training speed, including skip connections, projection or value parameters, sequential sub-blocks and normalisation layers. In experiments on both autoregressive decoder-only and BERT encoder-only models, our simplified transformers emulate the per-update training speed and performance of standard transformers, while enjoying 15% faster training throughput, and using 15% fewer parameters."
                },
                {
                    "title": "ConvNets Match Vision Transformers at Scale",
                    "abstract": "Many researchers believe that ConvNets perform well on small or moderately sized datasets, but are not competitive with Vision Transformers when given access to datasets on the web-scale. We challenge this belief by evaluating a performant ConvNet architecture pre-trained on JFT-4B, a large labelled dataset of images often used for training foundation models. We consider pre-training compute budgets between 0.4k and 110k TPU-v4 core compute hours, and train a series of networks of increasing depth and width from the NFNet model family. We observe a log-log scaling law between held out loss and compute budget. After fine-tuning on ImageNet, NFNets match the reported performance of Vision Transformers with comparable compute budgets. Our strongest fine-tuned model achieves a Top-1 accuracy of 90.4%."
                },
                {
                    "title": "On discretisation drift and smoothness regularisation in neural network training",
                    "abstract": "The deep learning recipe of casting real-world problems as mathematical optimisation and tackling the optimisation by training deep neural networks using gradient-based optimisation has undoubtedly proven to be a fruitful one. The understanding behind why deep learning works, however, has lagged behind its practical significance. We aim to make steps towards an improved understanding of deep learning with a focus on optimisation and model regularisation. We start by investigating gradient descent (GD), a discrete-time algorithm at the basis of most popular deep learning optimisation algorithms. Understanding the dynamics of GD has been hindered by the presence of discretisation drift, the numerical integration error between GD and its often studied continuous-time counterpart, the negative gradient flow (NGF). To add to the toolkit available to study GD, we derive novel continuous-time flows that account for discretisation drift. Unlike the NGF, these new flows can be used to describe learning rate specific behaviours of GD, such as training instabilities observed in supervised learning and two-player games. We then translate insights from continuous time into mitigation strategies for unstable GD dynamics, by constructing novel learning rate schedules and regularisers that do not require additional hyperparameters. Like optimisation, smoothness regularisation is another pillar of deep learning's success with wide use in supervised learning and generative modelling. Despite their individual significance, the interactions between smoothness regularisation and optimisation have yet to be explored. We find that smoothness regularisation affects optimisation across multiple deep learning domains, and that incorporating smoothness regularisation in reinforcement learning leads to a performance boost that can be recovered using adaptions to optimisation methods."
                },
                {
                    "title": "Tensor Programs VI: Feature Learning in Infinite-Depth Neural Networks",
                    "abstract": "By classifying infinite-width neural networks and identifying the *optimal* limit, Tensor Programs IV and V demonstrated a universal way, called $\\mu$P, for *widthwise hyperparameter transfer*, i.e., predicting optimal hyperparameters of wide neural networks from narrow ones. Here we investigate the analogous classification for *depthwise parametrizations* of deep residual networks (resnets). We classify depthwise parametrizations of block multiplier and learning rate by their infinite-width-then-depth limits. In resnets where each block has only one layer, we identify a unique optimal parametrization, called Depth-$\\mu$P that extends $\\mu$P and show empirically it admits depthwise hyperparameter transfer. We identify *feature diversity* as a crucial factor in deep networks, and Depth-$\\mu$P can be characterized as maximizing both feature learning and feature diversity. Exploiting this, we find that absolute value, among all homogeneous nonlinearities, maximizes feature diversity and indeed empirically leads to significantly better performance. However, if each block is deeper (such as modern transformers), then we find fundamental limitations in all possible infinite-depth limits of such parametrizations, which we illustrate both theoretically and empirically on simple networks as well as Megatron transformer trained on Common Crawl."
                },
                {
                    "title": "Towards Training Without Depth Limits: Batch Normalization Without Gradient Explosion",
                    "abstract": "Normalization layers are one of the key building blocks for deep neural networks. Several theoretical studies have shown that batch normalization improves the signal propagation, by avoiding the representations from becoming collinear across the layers. However, results on mean-field theory of batch normalization also conclude that this benefit comes at the expense of exploding gradients in depth. Motivated by these two aspects of batch normalization, in this study we pose the following question:\"Can a batch-normalized network keep the optimal signal propagation properties, but avoid exploding gradients?\"We answer this question in the affirmative by giving a particular construction of an Multi-Layer Perceptron (MLP) with linear activations and batch-normalization that provably has bounded gradients at any depth. Based on Weingarten calculus, we develop a rigorous and non-asymptotic theory for this constructed MLP that gives a precise characterization of forward signal propagation, while proving that gradients remain bounded for linearly independent input samples, which holds in most practical settings. Inspired by our theory, we also design an activation shaping scheme that empirically achieves the same properties for certain non-linear activations."
                },
                {
                    "title": "JoMA: Demystifying Multilayer Transformers via JOint Dynamics of MLP and Attention",
                    "abstract": "We propose Joint MLP/Attention (JoMA) dynamics, a novel mathematical framework to understand the training procedure of multilayer Transformer architectures. This is achieved by integrating out the self-attention layer in Transformers, producing a modified dynamics of MLP layers only. JoMA removes unrealistic assumptions in previous analysis (e.g., lack of residual connection) and predicts that the attention first becomes sparse (to learn salient tokens), then dense (to learn less salient tokens) in the presence of nonlinear activations, while in the linear case, it is consistent with existing works that show attention becomes sparse over time. We leverage JoMA to qualitatively explains how tokens are combined to form hierarchies in multilayer Transformers, when the input tokens are generated by a latent hierarchical generative model. Experiments on models trained from real-world dataset (Wikitext2/Wikitext103) and various pre-trained models (OPT, Pythia) verify our theoretical findings. Code can be found in https://github.com/facebookresearch/luckmatters/tree/yuandong3."
                },
                {
                    "title": "Depthwise Hyperparameter Transfer in Residual Networks: Dynamics and Scaling Limit",
                    "abstract": "The cost of hyperparameter tuning in deep learning has been rising with model sizes, prompting practitioners to find new tuning methods using a proxy of smaller networks. One such proposal uses $\\mu$P parameterized networks, where the optimal hyperparameters for small width networks transfer to networks with arbitrarily large width. However, in this scheme, hyperparameters do not transfer across depths. As a remedy, we study residual networks with a residual branch scale of $1/\\sqrt{\\text{depth}}$ in combination with the $\\mu$P parameterization. We provide experiments demonstrating that residual architectures including convolutional ResNets and Vision Transformers trained with this parameterization exhibit transfer of optimal hyperparameters across width and depth on CIFAR-10 and ImageNet. Furthermore, our empirical findings are supported and motivated by theory. Using recent developments in the dynamical mean field theory (DMFT) description of neural network learning dynamics, we show that this parameterization of ResNets admits a well-defined feature learning joint infinite-width and infinite-depth limit and show convergence of finite-size network dynamics towards this limit."
                },
                {
                    "title": "The Languini Kitchen: Enabling Language Modelling Research at Different Scales of Compute",
                    "abstract": "The Languini Kitchen serves as both a research collective and codebase designed to empower researchers with limited computational resources to contribute meaningfully to the field of language modelling. We introduce an experimental protocol that enables model comparisons based on equivalent compute, measured in accelerator hours. The number of tokens on which a model is trained is defined by the model's throughput and the chosen compute class. Notably, this approach avoids constraints on critical hyperparameters which affect total parameters or floating-point operations. For evaluation, we pre-process an existing large, diverse, and high-quality dataset of books that surpasses existing academic benchmarks in quality, diversity, and document length. On it, we compare methods based on their empirical scaling trends which are estimated through experiments at various levels of compute. This work also provides two baseline models: a feed-forward model derived from the GPT-2 architecture and a recurrent model in the form of a novel LSTM with ten-fold throughput. While the GPT baseline achieves better perplexity throughout all our levels of compute, our LSTM baseline exhibits a predictable and more favourable scaling law. This is due to the improved throughput and the need for fewer training tokens to achieve the same decrease in test perplexity. Extrapolating the scaling laws leads of both models results in an intersection at roughly 50,000 accelerator hours. We hope this work can serve as the foundation for meaningful and reproducible language modelling research."
                },
                {
                    "title": "The Shaped Transformer: Attention Models in the Infinite Depth-and-Width Limit",
                    "abstract": "In deep learning theory, the covariance matrix of the representations serves as a proxy to examine the network's trainability. Motivated by the success of Transformers, we study the covariance matrix of a modified Softmax-based attention model with skip connections in the proportional limit of infinite-depth-and-width. We show that at initialization the limiting distribution can be described by a stochastic differential equation (SDE) indexed by the depth-to-width ratio. To achieve a well-defined stochastic limit, the Transformer's attention mechanism is modified by centering the Softmax output at identity, and scaling the Softmax logits by a width-dependent temperature parameter. We examine the stability of the network through the corresponding SDE, showing how the scale of both the drift and diffusion can be elegantly controlled with the aid of residual connections. The existence of a stable SDE implies that the covariance structure is well-behaved, even for very large depth and width, thus preventing the notorious issues of rank degeneracy in deep attention models. Finally, we show, through simulations, that the SDE provides a surprisingly good description of the corresponding finite-size model. We coin the name shaped Transformer for these architectural modifications."
                },
                {
                    "title": "Quantizable Transformers: Removing Outliers by Helping Attention Heads Do Nothing",
                    "abstract": "Transformer models have been widely adopted in various domains over the last years, and especially large language models have advanced the field of AI significantly. Due to their size, the capability of these networks has increased tremendously, but this has come at the cost of a significant increase in necessary compute. Quantization is one of the most effective ways to reduce the computational time and memory consumption of neural networks. Many studies have shown, however, that modern transformer models tend to learn strong outliers in their activations, making them difficult to quantize. To retain acceptable performance, the existence of these outliers requires activations to be in higher bitwidth or the use of different numeric formats, extra fine-tuning, or other workarounds. We show that strong outliers are related to very specific behavior of attention heads that try to learn a\"no-op\"or just a partial update of the residual. To achieve the exact zeros needed in the attention matrix for a no-update, the input to the softmax is pushed to be larger and larger during training, causing outliers in other parts of the network. Based on these observations, we propose two simple (independent) modifications to the attention mechanism - clipped softmax and gated attention. We empirically show that models pre-trained using our methods learn significantly smaller outliers while maintaining and sometimes even improving the floating-point task performance. This enables us to quantize transformers to full INT8 quantization of the activations without any additional effort. We demonstrate the effectiveness of our methods on both language models (BERT, OPT) and vision transformers."
                },
                {
                    "title": "On the impact of activation and normalization in obtaining isometric embeddings at initialization",
                    "abstract": "In this paper, we explore the structure of the penultimate Gram matrix in deep neural networks, which contains the pairwise inner products of outputs corresponding to a batch of inputs. In several architectures it has been observed that this Gram matrix becomes degenerate with depth at initialization, which dramatically slows training. Normalization layers, such as batch or layer normalization, play a pivotal role in preventing the rank collapse issue. Despite promising advances, the existing theoretical results do not extend to layer normalization, which is widely used in transformers, and can not quantitatively characterize the role of non-linear activations. To bridge this gap, we prove that layer normalization, in conjunction with activation layers, biases the Gram matrix of a multilayer perceptron towards the identity matrix at an exponential rate with depth at initialization. We quantify this rate using the Hermite expansion of the activation function."
                },
                {
                    "title": "Stable and low-precision training for large-scale vision-language models",
                    "abstract": "We introduce new methods for 1) accelerating and 2) stabilizing training for large language-vision models. 1) For acceleration, we introduce SwitchBack, a linear layer for int8 quantized training which provides a speed-up of 13-25% while matching the performance of bfloat16 training within 0.1 percentage points for the 1B parameter CLIP ViT-Huge -- the largest int8 training to date. Our main focus is int8 as GPU support for float8 is rare, though we also analyze float8 training through simulation. While SwitchBack proves effective for float8, we show that standard techniques are also successful if the network is trained and initialized so that large feature magnitudes are discouraged, which we accomplish via layer-scale initialized with zeros. 2) For stability, we analyze loss spikes and find they consistently occur 1-8 iterations after the squared gradients become under-estimated by their AdamW second moment estimator. As a result, we recommend an AdamW-Adafactor hybrid which avoids loss spikes when training a CLIP ViT-Huge model and outperforms gradient clipping at the scales we test."
                },
                {
                    "title": "Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling",
                    "abstract": "How do large language models (LLMs) develop and evolve over the course of training? How do these patterns change as models scale? To answer these questions, we introduce \\textit{Pythia}, a suite of 16 LLMs all trained on public data seen in the exact same order and ranging in size from 70M to 12B parameters. We provide public access to 154 checkpoints for each one of the 16 models, alongside tools to download and reconstruct their exact training dataloaders for further study. We intend \\textit{Pythia} to facilitate research in many areas, and we present several case studies including novel results in memorization, term frequency effects on few-shot performance, and reducing gender bias. We demonstrate that this highly controlled setup can be used to yield novel insights toward LLMs and their training dynamics. Trained models, analysis code, training code, and training data can be found at \\url{https://github.com/EleutherAI/pythia}."
                },
                {
                    "title": "Stabilizing Transformer Training by Preventing Attention Entropy Collapse",
                    "abstract": "Training stability is of great importance to Transformers. In this work, we investigate the training dynamics of Transformers by examining the evolution of the attention layers. In particular, we track the attention entropy for each attention head during the course of training, which is a proxy for model sharpness. We identify a common pattern across different architectures and tasks, where low attention entropy is accompanied by high training instability, which can take the form of oscillating loss or divergence. We denote the pathologically low attention entropy, corresponding to highly concentrated attention scores, as $\\textit{entropy collapse}$. As a remedy, we propose $\\sigma$Reparam, a simple and efficient solution where we reparametrize all linear layers with spectral normalization and an additional learned scalar. We demonstrate that $\\sigma$Reparam successfully prevents entropy collapse in the attention layers, promoting more stable training. Additionally, we prove a tight lower bound of the attention entropy, which decreases exponentially fast with the spectral norm of the attention logits, providing additional motivation for our approach. We conduct experiments with $\\sigma$Reparam on image classification, image self-supervised learning, machine translation, speech recognition, and language modeling tasks. We show that $\\sigma$Reparam provides stability and robustness with respect to the choice of hyperparameters, going so far as enabling training (a) a Vision Transformer {to competitive performance} without warmup, weight decay, layer normalization or adaptive optimizers; (b) deep architectures in machine translation and (c) speech recognition to competitive performance without warmup and adaptive optimizers. Code is available at \\url{https://github.com/apple/ml-sigma-reparam}."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Deep Transformers without Shortcuts: Modifying Self-attention for Faithful Signal Propagation",
                    "abstract": "Skip connections and normalisation layers form two standard architectural components that are ubiquitous for the training of Deep Neural Networks (DNNs), but whose precise roles are poorly understood. Recent approaches such as Deep Kernel Shaping have made progress towards reducing our reliance on them, using insights from wide NN kernel theory to improve signal propagation in vanilla DNNs (which we define as networks without skips or normalisation). However, these approaches are incompatible with the self-attention layers present in transformers, whose kernels are intrinsically more complicated to analyse and control. And so the question remains: is it possible to train deep vanilla transformers? We answer this question in the affirmative by designing several approaches that use combinations of parameter initialisations, bias matrices and location-dependent rescaling to achieve faithful signal propagation in vanilla transformers. Our methods address various intricacies specific to signal propagation in transformers, including the interaction with positional encoding and causal masking. In experiments on WikiText-103 and C4, our approaches enable deep transformers without normalisation to train at speeds matching their standard counterparts, and deep vanilla transformers to reach the same performance as standard ones after about 5 times more iterations."
                },
                {
                    "title": "Scaling Vision Transformers to 22 Billion Parameters",
                    "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                },
                {
                    "title": "SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models",
                    "abstract": "Large language models (LLMs) show excellent performance but are compute- and memory-intensive. Quantization can reduce memory and accelerate inference. However, existing methods cannot maintain accuracy and hardware efficiency at the same time. We propose SmoothQuant, a training-free, accuracy-preserving, and general-purpose post-training quantization (PTQ) solution to enable 8-bit weight, 8-bit activation (W8A8) quantization for LLMs. Based on the fact that weights are easy to quantize while activations are not, SmoothQuant smooths the activation outliers by offline migrating the quantization difficulty from activations to weights with a mathematically equivalent transformation. SmoothQuant enables an INT8 quantization of both weights and activations for all the matrix multiplications in LLMs, including OPT, BLOOM, GLM, MT-NLG, Llama-1/2, Falcon, Mistral, and Mixtral models. We demonstrate up to 1.56x speedup and 2x memory reduction for LLMs with negligible loss in accuracy. SmoothQuant enables serving 530B LLM within a single node. Our work offers a turn-key solution that reduces hardware costs and democratizes LLMs. Code is available at https://github.com/mit-han-lab/smoothquant."
                },
                {
                    "title": "GLM-130B: An Open Bilingual Pre-trained Model",
                    "abstract": "We introduce GLM-130B, a bilingual (English and Chinese) pre-trained language model with 130 billion parameters. It is an attempt to open-source a 100B-scale model at least as good as GPT-3 (davinci) and unveil how models of such a scale can be successfully pre-trained. Over the course of this effort, we face numerous unexpected technical and engineering challenges, particularly on loss spikes and divergence. In this paper, we introduce the training process of GLM-130B including its design choices, training strategies for both efficiency and stability, and engineering efforts. The resultant GLM-130B model offers significant outperformance over GPT-3 175B (davinci) on a wide range of popular English benchmarks while the performance advantage is not observed in OPT-175B and BLOOM-176B. It also consistently and significantly outperforms ERNIE TITAN 3.0 260B -- the largest Chinese language model -- across related benchmarks. Finally, we leverage a unique scaling property of GLM-130B to reach INT4 quantization without post training, with almost no performance loss, making it the first among 100B-scale models and more importantly, allowing its effective inference on 4$\\times$RTX 3090 (24G) or 8$\\times$RTX 2080 Ti (11G) GPUs, the most affordable GPUs required for using 100B-scale models. The GLM-130B model weights are publicly accessible and its code, training logs, related toolkit, and lessons learned are open-sourced at \\url{https://github.com/THUDM/GLM-130B/}."
                },
                {
                    "title": "Outlier Suppression: Pushing the Limit of Low-bit Transformer Language Models",
                    "abstract": "Transformer architecture has become the fundamental element of the widespread natural language processing~(NLP) models. With the trends of large NLP models, the increasing memory and computation costs hinder their efficient deployment on resource-limited devices. Therefore, transformer quantization attracts wide research interest. Recent work recognizes that structured outliers are the critical bottleneck for quantization performance. However, their proposed methods increase the computation overhead and still leave the outliers there. To fundamentally address this problem, this paper delves into the inherent inducement and importance of the outliers. We discover that $\\boldsymbol \\gamma$ in LayerNorm (LN) acts as a sinful amplifier for the outliers, and the importance of outliers varies greatly where some outliers provided by a few tokens cover a large area but can be clipped sharply without negative impacts. Motivated by these findings, we propose an outlier suppression framework including two components: Gamma Migration and Token-Wise Clipping. The Gamma Migration migrates the outlier amplifier to subsequent modules in an equivalent transformation, contributing to a more quantization-friendly model without any extra burden. The Token-Wise Clipping takes advantage of the large variance of token range and designs a token-wise coarse-to-fine pipeline, obtaining a clipping range with minimal final quantization loss in an efficient way. This framework effectively suppresses the outliers and can be used in a plug-and-play mode. Extensive experiments prove that our framework surpasses the existing works and, for the first time, pushes the 6-bit post-training BERT quantization to the full-precision (FP) level. Our code is available at https://github.com/wimh966/outlier_suppression."
                },
                {
                    "title": "Signal Propagation in Transformers: Theoretical Perspectives and the Role of Rank Collapse",
                    "abstract": "Transformers have achieved remarkable success in several domains, ranging from natural language processing to computer vision. Nevertheless, it has been recently shown that stacking self-attention layers - the distinctive architectural component of Transformers - can result in rank collapse of the tokens' representations at initialization. The question of if and how rank collapse affects training is still largely unanswered, and its investigation is necessary for a more comprehensive understanding of this architecture. In this work, we shed new light on the causes and the effects of this phenomenon. First, we show that rank collapse of the tokens' representations hinders training by causing the gradients of the queries and keys to vanish at initialization. Furthermore, we provide a thorough description of the origin of rank collapse and discuss how to prevent it via an appropriate depth-dependent scaling of the residual branches. Finally, our analysis unveils that specific architectural hyperparameters affect the gradients of queries and values differently, leading to disproportionate gradient norms. This suggests an explanation for the widespread use of adaptive methods for Transformers' optimization."
                },
                {
                    "title": "The Neural Covariance SDE: Shaped Infinite Depth-and-Width Networks at Initialization",
                    "abstract": "The logit outputs of a feedforward neural network at initialization are conditionally Gaussian, given a random covariance matrix defined by the penultimate layer. In this work, we study the distribution of this random matrix. Recent work has shown that shaping the activation function as network depth grows large is necessary for this covariance matrix to be non-degenerate. However, the current infinite-width-style understanding of this shaping method is unsatisfactory for large depth: infinite-width analyses ignore the microscopic fluctuations from layer to layer, but these fluctuations accumulate over many layers. To overcome this shortcoming, we study the random covariance matrix in the shaped infinite-depth-and-width limit. We identify the precise scaling of the activation function necessary to arrive at a non-trivial limit, and show that the random covariance matrix is governed by a stochastic differential equation (SDE) that we call the Neural Covariance SDE. Using simulations, we show that the SDE closely matches the distribution of the random covariance matrix of finite networks. Additionally, we recover an if-and-only-if condition for exploding and vanishing norms of large shaped networks based on the activation function."
                },
                {
                    "title": "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness",
                    "abstract": "Transformers are slow and memory-hungry on long sequences, since the time and memory complexity of self-attention are quadratic in sequence length. Approximate attention methods have attempted to address this problem by trading off model quality to reduce the compute complexity, but often do not achieve wall-clock speedup. We argue that a missing principle is making attention algorithms IO-aware -- accounting for reads and writes between levels of GPU memory. We propose FlashAttention, an IO-aware exact attention algorithm that uses tiling to reduce the number of memory reads/writes between GPU high bandwidth memory (HBM) and GPU on-chip SRAM. We analyze the IO complexity of FlashAttention, showing that it requires fewer HBM accesses than standard attention, and is optimal for a range of SRAM sizes. We also extend FlashAttention to block-sparse attention, yielding an approximate attention algorithm that is faster than any existing approximate attention method. FlashAttention trains Transformers faster than existing baselines: 15% end-to-end wall-clock speedup on BERT-large (seq. length 512) compared to the MLPerf 1.1 training speed record, 3$\\times$ speedup on GPT-2 (seq. length 1K), and 2.4$\\times$ speedup on long-range arena (seq. length 1K-4K). FlashAttention and block-sparse FlashAttention enable longer context in Transformers, yielding higher quality models (0.7 better perplexity on GPT-2 and 6.4 points of lift on long-document classification) and entirely new capabilities: the first Transformers to achieve better-than-chance performance on the Path-X challenge (seq. length 16K, 61.4% accuracy) and Path-256 (seq. length 64K, 63.1% accuracy)."
                },
                {
                    "title": "Outliers Dimensions that Disrupt Transformers Are Driven by Frequency",
                    "abstract": "While Transformer-based language models are generally very robust to pruning, there is the recently discovered outlier phenomenon: disabling only 48 out of 110M parameters in BERT-base drops its performance by nearly 30% on MNLI. We replicate the original evidence for the outlier phenomenon and we link it to the geometry of the embedding space. We find that in both BERT and RoBERTa the magnitude of hidden state coefficients corresponding to outlier dimensions correlates with the frequency of encoded tokens in pre-training data, and it also contributes to the\"vertical\"self-attention pattern enabling the model to focus on the special tokens. This explains the drop in performance from disabling the outliers, and it suggests that to decrease anisotropicity in future models we need pre-training schemas that would better take into account the skewed token distributions."
                },
                {
                    "title": "Self-consistent dynamical field theory of kernel evolution in wide neural networks",
                    "abstract": "We analyze feature learning in infinite-width neural networks trained with gradient flow through a self-consistent dynamical field theory. We construct a collection of deterministic dynamical order parameters which are inner-product kernels for hidden unit activations and gradients in each layer at pairs of time points, providing a reduced description of network activity through training. These kernel order parameters collectively define the hidden layer activation distribution, the evolution of the neural tangent kernel (NTK), and consequently, output predictions. We show that the field theory derivation recovers the recursive stochastic process of infinite-width feature learning networks obtained by Yang and Hu with tensor programs. For deep linear networks, these kernels satisfy a set of algebraic matrix equations. For nonlinear networks, we provide an alternating sampling procedure to self-consistently solve for the kernel order parameters. We provide comparisons of the self-consistent solution to various approximation schemes including the static NTK approximation, gradient independence assumption, and leading order perturbation theory, showing that each of these approximations can break down in regimes where general self-consistent solutions still provide an accurate description. Lastly, we provide experiments in more realistic settings which demonstrate that the loss and kernel dynamics of convolutional neural networks at fixed feature learning strength are preserved across different widths on a image classification task."
                },
                {
                    "title": "High-dimensional Asymptotics of Feature Learning: How One Gradient Step Improves the Representation",
                    "abstract": "We study the first gradient descent step on the first-layer parameters $\\boldsymbol{W}$ in a two-layer neural network: $f(\\boldsymbol{x}) = \\frac{1}{\\sqrt{N}}\\boldsymbol{a}^\\top\\sigma(\\boldsymbol{W}^\\top\\boldsymbol{x})$, where $\\boldsymbol{W}\\in\\mathbb{R}^{d\\times N}, \\boldsymbol{a}\\in\\mathbb{R}^{N}$ are randomly initialized, and the training objective is the empirical MSE loss: $\\frac{1}{n}\\sum_{i=1}^n (f(\\boldsymbol{x}_i)-y_i)^2$. In the proportional asymptotic limit where $n,d,N\\to\\infty$ at the same rate, and an idealized student-teacher setting, we show that the first gradient update contains a rank-1\"spike\", which results in an alignment between the first-layer weights and the linear component of the teacher model $f^*$. To characterize the impact of this alignment, we compute the prediction risk of ridge regression on the conjugate kernel after one gradient step on $\\boldsymbol{W}$ with learning rate $\\eta$, when $f^*$ is a single-index model. We consider two scalings of the first step learning rate $\\eta$. For small $\\eta$, we establish a Gaussian equivalence property for the trained feature map, and prove that the learned kernel improves upon the initial random features model, but cannot defeat the best linear model on the input. Whereas for sufficiently large $\\eta$, we prove that for certain $f^*$, the same ridge estimator on trained features can go beyond this\"linear regime\"and outperform a wide range of random features and rotationally invariant kernels. Our results demonstrate that even one gradient step can lead to a considerable advantage over random features, and highlight the role of learning rate scaling in the initial phase of training."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "Deep Learning without Shortcuts: Shaping the Kernel with Tailored Rectifiers",
                    "abstract": "Training very deep neural networks is still an extremely challenging task. The common solution is to use shortcut connections and normalization layers, which are both crucial ingredients in the popular ResNet architecture. However, there is strong evidence to suggest that ResNets behave more like ensembles of shallower networks than truly deep ones. Recently, it was shown that deep vanilla networks (i.e. networks without normalization layers or shortcut connections) can be trained as fast as ResNets by applying certain transformations to their activation functions. However, this method (called Deep Kernel Shaping) isn't fully compatible with ReLUs, and produces networks that overfit significantly more than ResNets on ImageNet. In this work, we rectify this situation by developing a new type of transformation that is fully compatible with a variant of ReLUs -- Leaky ReLUs. We show in experiments that our method, which introduces negligible extra computational cost, achieves validation accuracies with deep vanilla networks that are competitive with ResNets (of the same width/depth), and significantly higher than those obtained with the Edge of Chaos (EOC) method. And unlike with EOC, the validation accuracies we obtain do not get worse with depth."
                },
                {
                    "title": "Scaling Language Models: Methods, Analysis & Insights from Training Gopher",
                    "abstract": "Language modelling provides a step towards intelligent communication systems by harnessing large repositories of written human knowledge to better predict and understand the world. In this paper, we present an analysis of Transformer-based language model performance across a wide range of model scales -- from models with tens of millions of parameters up to a 280 billion parameter model called Gopher. These models are evaluated on 152 diverse tasks, achieving state-of-the-art performance across the majority. Gains from scale are largest in areas such as reading comprehension, fact-checking, and the identification of toxic language, but logical and mathematical reasoning see less benefit. We provide a holistic analysis of the training dataset and model's behaviour, covering the intersection of model scale with bias and toxicity. Finally we discuss the application of language models to AI safety and the mitigation of downstream harms."
                },
                {
                    "title": "Feature Learning and Signal Propagation in Deep Neural Networks",
                    "abstract": "Recent work by Baratin et al. (2021) sheds light on an intriguing pattern that occurs during the training of deep neural networks: some layers align much more with data compared to other layers (where the alignment is defined as the euclidean product of the tangent features matrix and the data labels matrix). The curve of the alignment as a function of layer index (generally) exhibits an ascent-descent pattern where the maximum is reached for some hidden layer. In this work, we provide the first explanation for this phenomenon. We introduce the Equilibrium Hypothesis which connects this alignment pattern to signal propagation in deep neural networks. Our experiments demonstrate an excellent match with the theoretical predictions."
                },
                {
                    "title": "Rapid training of deep neural networks without skip connections or normalization layers using Deep Kernel Shaping",
                    "abstract": "Using an extended and formalized version of the Q/C map analysis of Poole et al. (2016), along with Neural Tangent Kernel theory, we identify the main pathologies present in deep networks that prevent them from training fast and generalizing to unseen data, and show how these can be avoided by carefully controlling the\"shape\"of the network's initialization-time kernel function. We then develop a method called Deep Kernel Shaping (DKS), which accomplishes this using a combination of precise parameter initialization, activation function transformations, and small architectural tweaks, all of which preserve the model class. In our experiments we show that DKS enables SGD training of residual networks without normalization layers on Imagenet and CIFAR-10 classification tasks at speeds comparable to standard ResNetV2 and Wide-ResNet models, with only a small decrease in generalization performance. And when using K-FAC as the optimizer, we achieve similar results for networks without skip connections. Our results apply for a large variety of activation functions, including those which traditionally perform very badly, such as the logistic sigmoid. In addition to DKS, we contribute a detailed analysis of skip connections, normalization layers, special activation functions like RELU and SELU, and various initialization schemes, explaining their effectiveness as alternative (and ultimately incomplete) ways of\"shaping\"the network's initialization-time kernel."
                },
                {
                    "title": "Understanding and Overcoming the Challenges of Efficient Transformer Quantization",
                    "abstract": "Transformer-based architectures have become the de-facto standard models for a wide range of Natural Language Processing tasks. However, their memory footprint and high latency are prohibitive for efficient deployment and inference on resource-limited devices. In this work, we explore quantization for transformers. We show that transformers have unique quantization challenges \u2013 namely, high dynamic activation ranges that are difficult to represent with a low bit fixed-point format. We establish that these activations contain structured outliers in the residual connections that encourage specific attention patterns, such as attending to the special separator token. To combat these challenges, we present three solutions based on post-training quantization and quantization-aware training, each with a different set of compromises for accuracy, model size, and ease of use. In particular, we introduce a novel quantization scheme \u2013 per-embedding-group quantization. We demonstrate the effectiveness of our methods on the GLUE benchmark using BERT, establishing state-of-the-art results for post-training quantization. Finally, we show that transformer weights and embeddings can be quantized to ultra-low bit-widths, leading to significant memory savings with a minimum accuracy loss. Our source code is available at https://github.com/qualcomm-ai-research/transformer-quantization."
                },
                {
                    "title": "All Bark and No Bite: Rogue Dimensions in Transformer Language Models Obscure Representational Quality",
                    "abstract": "Similarity measures are a vital tool for understanding how language models represent and process language. Standard representational similarity measures such as cosine similarity and Euclidean distance have been successfully used in static word embedding models to understand how words cluster in semantic space. Recently, these measures have been applied to embeddings from contextualized models such as BERT and GPT-2. In this work, we call into question the informativity of such measures for contextualized language models. We find that a small number of rogue dimensions, often just 1-3, dominate these measures. Moreover, we find a striking mismatch between the dimensions that dominate similarity measures and those which are important to the behavior of the model. We show that simple postprocessing techniques such as standardization are able to correct for rogue dimensions and reveal underlying representational quality. We argue that accounting for rogue dimensions is essential for any similarity-based analysis of contextual language models."
                },
                {
                    "title": "Precise characterization of the prior predictive distribution of deep ReLU networks",
                    "abstract": "Recent works on Bayesian neural networks (BNNs) have highlighted the need to better understand the implications of using Gaussian priors in combination with the compositional structure of the network architecture. Similar in spirit to the kind of analysis that has been developed to devise better initialization schemes for neural networks (cf. He- or Xavier initialization), we derive a precise characterization of the prior predictive distribution of finite-width ReLU networks with Gaussian weights. While theoretical results have been obtained for their heavy-tailedness, the full characterization of the prior predictive distribution (i.e. its density, CDF and moments), remained unknown prior to this work. Our analysis, based on the Meijer-G function, allows us to quantify the influence of architectural choices such as the width or depth of the network on the resulting shape of the prior predictive distribution. We also formally connect our results to previous work in the infinite width setting, demonstrating that the moments of the distribution converge to those of a normal log-normal mixture in the infinite depth limit. Finally, our results provide valuable guidance on prior design: for instance, controlling the predictive variance with depth- and width-informed priors on the weights of the network."
                },
                {
                    "title": "BERT Busters: Outlier Dimensions that Disrupt Transformers",
                    "abstract": "Multiple studies have shown that Transformers are remarkably robust to pruning. Contrary to this received wisdom, we demonstrate that pre-trained Transformer encoders are surprisingly fragile to the removal of a very small number of features in the layer outputs (<0.0001% of model weights). In case of BERT and other pre-trained encoder Transformers, the affected component is the scaling factors and biases in the LayerNorm. The outliers are high-magnitude normalization parameters that emerge early in pre-training and show up consistently in the same dimensional position throughout the model. We show that disabling them significantly degrades both the MLM loss and the downstream task performance. This effect is observed across several BERT-family models and other popular pre-trained Transformer architectures, including BART, XLNet and ELECTRA; we also show a similar effect in GPT-2."
                },
                {
                    "title": "Going deeper with Image Transformers",
                    "abstract": "Transformers have been recently adapted for large scale image classification, achieving high scores shaking up the long supremacy of convolutional neural networks. However the optimization of vision transformers has been little studied so far. In this work, we build and optimize deeper transformer networks for image classification. In particular, we investigate the interplay of architecture and optimization of such dedicated transformers. We make two architecture changes that significantly improve the accuracy of deep transformers. This leads us to produce models whose performance does not saturate early with more depth, for in-stance we obtain 86.5% top-1 accuracy on Imagenet when training with no external data, we thus attain the current sate of the art with less floating-point operations and parameters. Our best model establishes the new state of the art on Imagenet with Reassessed labels and Imagenet-V2 / match frequency, in the setting with no additional training data. We share our code and models1."
                },
                {
                    "title": "Attention is Not All You Need: Pure Attention Loses Rank Doubly Exponentially with Depth",
                    "abstract": "Attention-based architectures have become ubiquitous in machine learning, yet our understanding of the reasons for their effectiveness remains limited. This work proposes a new way to understand self-attention networks: we show that their output can be decomposed into a sum of smaller terms, each involving the operation of a sequence of attention heads across layers. Using this decomposition, we prove that self-attention possesses a strong inductive bias towards\"token uniformity\". Specifically, without skip connections or multi-layer perceptrons (MLPs), the output converges doubly exponentially to a rank-1 matrix. On the other hand, skip connections and MLPs stop the output from degeneration. Our experiments verify the identified convergence phenomena on different variants of standard transformer architectures."
                },
                {
                    "title": "Gradient Descent on Neural Networks Typically Occurs at the Edge of Stability",
                    "abstract": "We empirically demonstrate that full-batch gradient descent on neural network training objectives typically operates in a regime we call the Edge of Stability. In this regime, the maximum eigenvalue of the training loss Hessian hovers just above the numerical value $2 / \\text{(step size)}$, and the training loss behaves non-monotonically over short timescales, yet consistently decreases over long timescales. Since this behavior is inconsistent with several widespread presumptions in the field of optimization, our findings raise questions as to whether these presumptions are relevant to neural network training. We hope that our findings will inspire future efforts aimed at rigorously understanding optimization at the Edge of Stability. Code is available at https://github.com/locuslab/edge-of-stability."
                },
                {
                    "title": "High-Performance Large-Scale Image Recognition Without Normalization",
                    "abstract": "Batch normalization is a key component of most image classification models, but it has many undesirable properties stemming from its dependence on the batch size and interactions between examples. Although recent work has succeeded in training deep ResNets without normalization layers, these models do not match the test accuracies of the best batch-normalized networks, and are often unstable for large learning rates or strong data augmentations. In this work, we develop an adaptive gradient clipping technique which overcomes these instabilities, and design a significantly improved class of Normalizer-Free ResNets. Our smaller models match the test accuracy of an EfficientNet-B7 on ImageNet while being up to 8.7x faster to train, and our largest models attain a new state-of-the-art top-1 accuracy of 86.5%. In addition, Normalizer-Free models attain significantly better performance than their batch-normalized counterparts when finetuning on ImageNet after large-scale pre-training on a dataset of 300 million labeled images, with our best models obtaining an accuracy of 89.2%. Our code is available at https://github.com/deepmind/ deepmind-research/tree/master/nfnets"
                },
                {
                    "title": "Towards Understanding Ensemble, Knowledge Distillation and Self-Distillation in Deep Learning",
                    "abstract": "We formally study how Ensemble of deep learning models can improve test accuracy, and how the superior performance of ensemble can be distilled into a single model using Knowledge Distillation. We consider the challenging case where the ensemble is simply an average of the outputs of a few independently trained neural networks with the SAME architecture, trained using the SAME algorithm on the SAME data set, and they only differ by the random seeds used in the initialization. We empirically show that ensemble/knowledge distillation in deep learning works very differently from traditional learning theory, especially differently from ensemble of random feature mappings or the neural-tangent-kernel feature mappings, and is potentially out of the scope of existing theorems. Thus, to properly understand ensemble and knowledge distillation in deep learning, we develop a theory showing that when data has a structure we refer to as \"multi-view\", then ensemble of independently trained neural networks can provably improve test accuracy, and such superior test accuracy can also be provably distilled into a single model by training a single model to match the output of the ensemble instead of the true label. Our result sheds light on how ensemble works in deep learning in a way that is completely different from traditional theorems, and how the \"dark knowledge\" is hidden in the outputs of the ensemble -- that can be used in knowledge distillation -- comparing to the true data labels. In the end, we prove that self-distillation can also be viewed as implicitly combining ensemble and knowledge distillation to improve test accuracy."
                },
                {
                    "title": "Feature Learning in Infinite-Width Neural Networks",
                    "abstract": "As its width tends to infinity, a deep neural network's behavior under gradient descent can become simplified and predictable (e.g. given by the Neural Tangent Kernel (NTK)), if it is parametrized appropriately (e.g. the NTK parametrization). However, we show that the standard and NTK parametrizations of a neural network do not admit infinite-width limits that can learn features, which is crucial for pretraining and transfer learning such as with BERT. We propose simple modifications to the standard parametrization to allow for feature learning in the limit. Using the *Tensor Programs* technique, we derive explicit formulas for such limits. On Word2Vec and few-shot learning on Omniglot via MAML, two canonical tasks that rely crucially on feature learning, we compute these limits exactly. We find that they outperform both NTK baselines and finite-width networks, with the latter approaching the infinite-width feature learning performance as width increases. \nMore generally, we classify a natural space of neural network parametrizations that generalizes standard, NTK, and Mean Field parametrizations. We show 1) any parametrization in this space either admits feature learning or has an infinite-width training dynamics given by kernel gradient descent, but not both; 2) any such infinite-width limit can be computed using the Tensor Programs technique."
                },
                {
                    "title": "Stable ResNet",
                    "abstract": "Deep ResNet architectures have achieved state of the art performance on many tasks. While they solve the problem of gradient vanishing, they might suffer from gradient exploding as the depth becomes large (Yang et al. 2017). Moreover, recent results have shown that ResNet might lose expressivity as the depth goes to infinity (Yang et al. 2017, Hayou et al. 2019). To resolve these issues, we introduce a new class of ResNet architectures, called Stable ResNet, that have the property of stabilizing the gradient while ensuring expressivity in the infinite depth limit."
                },
                {
                    "title": "Query-Key Normalization for Transformers",
                    "abstract": "Low-resource language translation is a challenging but socially valuable NLP task. Building on recent work adapting the Transformer\u2019s normalization to this setting, we propose QKNorm, a normalization technique that modifies the attention mechanism to make the softmax function less prone to arbitrary saturation without sacrificing expressivity. Specifically, we apply l2-normalization along the head dimension of each query and key matrix prior to multiplying them and then scale up by a learnable parameter instead of dividing by the square root of the embedding dimension. We show improvements averaging 0.928 BLEU over state-of-the-art bilingual benchmarks for 5 low-resource translation pairs from the TED Talks corpus and IWSLT\u201915."
                },
                {
                    "title": "Implicit Regularization via Neural Feature Alignment",
                    "abstract": "We approach the problem of implicit regularization in deep learning from a geometrical viewpoint. We highlight a regularization effect induced by a dynamical alignment of the neural tangent features introduced by Jacot et al, along a small number of task-relevant directions. This can be interpreted as a combined mechanism of feature selection and model compression. By extrapolating a new analysis of Rademacher complexity bounds for linear models, we motivate and study a heuristic complexity measure that captures this phenomenon, in terms of sequences of tangent kernel classes along the optimization paths."
                },
                {
                    "title": "Improving Transformer Optimization Through Better Initialization",
                    "abstract": "The Transformer architecture has achieved considerable success recently; the key component of the Transformer is the attention layer that enables the model to focus on important regions within an input sequence. Gradient optimization with attention layers can be notoriously dif\ufb01cult requiring tricks such as learning rate warmup to prevent divergence. As Transformer models are becoming larger and more expensive to train, recent research has focused on understanding and improving optimization in these architectures. In this work our contributions are two-fold: we \ufb01rst investigate and empirically validate the source of optimization problems in the encoder-decoder Transformer architecture; we then propose a new weight initialization scheme with theoretical justi-\ufb01cation, that enables training without warmup or layer normalization. Empirical results on public machine translation benchmarks show that our approach achieves leading accuracy, allowing to train deep Transformer models with 200 layers in both encoder and decoder (over 1000 attention/MLP blocks) without dif\ufb01culty. Code for this work is available here: https://github. com/layer6ai-labs/T-Fixup ."
                },
                {
                    "title": "Infinite attention: NNGP and NTK for deep attention networks",
                    "abstract": "There is a growing amount of literature on the relationship between wide neural networks (NNs) and Gaussian processes (GPs), identifying an equivalence between the two for a variety of NN architectures. This equivalence enables, for instance, accurate approximation of the behaviour of wide Bayesian NNs without MCMC or variational approximations, or characterisation of the distribution of randomly initialised wide NNs optimised by gradient descent without ever running an optimiser. We provide a rigorous extension of these results to NNs involving attention layers, showing that unlike single-head attention, which induces non-Gaussian behaviour, multi-head attention architectures behave as GPs as the number of heads tends to infinity. We further discuss the effects of positional encodings and layer normalisation, and propose modifications of the attention mechanism which lead to improved results for both finite and infinitely wide NNs. We evaluate attention kernels empirically, leading to a moderate improvement upon the previous state-of-the-art on CIFAR-10 for GPs without trainable kernels and advanced data preprocessing. Finally, we introduce new features to the Neural Tangents library (Novak et al., 2020) allowing applications of NNGP/NTK models, with and without attention, to variable-length sequences, with an example on the IMDb reviews dataset."
                },
                {
                    "title": "ReZero is All You Need: Fast Convergence at Large Depth",
                    "abstract": "Deep networks often suffer from vanishing or exploding gradients due to inefficient signal propagation, leading to long training times or convergence difficulties. Various architecture designs, sophisticated residual-style networks, and initialization schemes have been shown to improve deep signal propagation. Recently, Pennington et al. used free probability theory to show that dynamical isometry plays an integral role in efficient deep learning. We show that the simplest architecture change of gating each residual connection using a single zero-initialized parameter satisfies initial dynamical isometry and outperforms more complex approaches. Although much simpler than its predecessors, this gate enables training thousands of fully connected layers with fast convergence and better test performance for ResNets trained on CIFAR-10. We apply this technique to language modeling and find that we can easily train 120-layer Transformers. When applied to 12 layer Transformers, it converges 56% faster on enwiki8."
                },
                {
                    "title": "Batch Normalization Biases Residual Blocks Towards the Identity Function in Deep Networks",
                    "abstract": "Batch normalization dramatically increases the largest trainable depth of residual networks, and this benefit has been crucial to the empirical success of deep residual networks on a wide range of benchmarks. We show that this key benefit arises because, at initialization, batch normalization downscales the residual branch relative to the skip connection, by a normalizing factor on the order of the square root of the network depth. This ensures that, early in training, the function computed by normalized residual blocks in deep networks is close to the identity function (on average). We use this insight to develop a simple initialization scheme that can train deep residual networks without normalization. We also provide a detailed empirical study of residual networks, which clarifies that, although batch normalized networks can be trained with larger learning rates, this effect is only beneficial in specific compute regimes, and has minimal benefits when the batch size is small."
                },
                {
                    "title": "On Layer Normalization in the Transformer Architecture",
                    "abstract": "The Transformer is widely used in natural language processing tasks. To train a Transformer however, one usually needs a carefully designed learning rate warm-up stage, which is shown to be crucial to the final performance but will slow down the optimization and bring more hyperparameter tunings. In this paper, we first study theoretically why the learning rate warm-up stage is essential and show that the location of layer normalization matters. Specifically, we prove with mean field theory that at initialization, for the original-designed Post-LN Transformer, which places the layer normalization between the residual blocks, the expected gradients of the parameters near the output layer are large. Therefore, using a large learning rate on those gradients makes the training unstable. The warm-up stage is practically helpful for avoiding this problem. On the other hand, our theory also shows that if the layer normalization is put inside the residual blocks (recently proposed as Pre-LN Transformer), the gradients are well-behaved at initialization. This motivates us to remove the warm-up stage for the training of Pre-LN Transformers. We show in our experiments that Pre-LN Transformers without the warm-up stage can reach comparable results with baselines while requiring significantly less training time and hyper-parameter tuning on a wide range of applications."
                },
                {
                    "title": "Disentangling Trainability and Generalization in Deep Neural Networks",
                    "abstract": "A longstanding goal in the theory of deep learning is to characterize the conditions under which a given neural network architecture will be trainable, and if so, how well it might generalize to unseen data. In this work, we provide such a characterization in the limit of very wide and very deep networks, for which the analysis simplifies considerably. For wide networks, the trajectory under gradient descent is governed by the Neural Tangent Kernel (NTK), and for deep networks the NTK itself maintains only weak data dependence. By analyzing the spectrum of the NTK, we formulate necessary conditions for trainability and generalization across a range of architectures, including Fully Connected Networks (FCNs) and Convolutional Neural Networks (CNNs). We identify large regions of hyperparameter space for which networks can memorize the training set but completely fail to generalize. We find that CNNs without global average pooling behave almost identically to FCNs, but that CNNs with pooling have markedly different and often better generalization performance. These theoretical results are corroborated experimentally on CIFAR10 for a variety of network architectures and we include a colab notebook that reproduces the essential results of the paper."
                },
                {
                    "title": "Tensor Programs I: Wide Feedforward or Recurrent Neural Networks of Any Architecture are Gaussian Processes",
                    "abstract": "Wide neural networks with random weights and biases are Gaussian processes, as observed by Neal (1995) for shallow networks, and more recently by Lee et al.~(2018) and Matthews et al.~(2018) for deep fully-connected networks, as well as by Novak et al.~(2019) and Garriga-Alonso et al.~(2019) for deep convolutional networks. We show that this Neural Network-Gaussian Process correspondence surprisingly extends to all modern feedforward or recurrent neural networks composed of multilayer perceptron, RNNs (e.g. LSTMs, GRUs), (nD or graph) convolution, pooling, skip connection, attention, batch normalization, and/or layer normalization. More generally, we introduce a language for expressing neural network computations, and our result encompasses all such expressible neural networks. This work serves as a tutorial on the \\emph{tensor programs} technique formulated in Yang (2019) and elucidates the Gaussian Process results obtained there. We provide open-source implementations of the Gaussian Process kernels of simple RNN, GRU, transformer, and batchnorm+ReLU network at github.com/thegregyang/GP4A. Please see our arxiv version for the complete and up-to-date version of this paper."
                },
                {
                    "title": "Root Mean Square Layer Normalization",
                    "abstract": "Layer normalization (LayerNorm) has been successfully applied to various deep neural networks to help stabilize training and boost model convergence because of its capability in handling re-centering and re-scaling of both inputs and weight matrix. However, the computational overhead introduced by LayerNorm makes these improvements expensive and significantly slows the underlying network, e.g. RNN in particular. In this paper, we hypothesize that re-centering invariance in LayerNorm is dispensable and propose root mean square layer normalization, or RMSNorm. RMSNorm regularizes the summed inputs to a neuron in one layer according to root mean square (RMS), giving the model re-scaling invariance property and implicit learning rate adaptation ability. RMSNorm is computationally simpler and thus more efficient than LayerNorm. We also present partial RMSNorm, or pRMSNorm where the RMS is estimated from p% of the summed inputs without breaking the above properties. Extensive experiments on several tasks using diverse network architectures show that RMSNorm achieves comparable performance against LayerNorm but reduces the running time by 7%~64% on different models. Source code is available at https://github.com/bzhangGo/rmsnorm."
                },
                {
                    "title": "Representation Degeneration Problem in Training Natural Language Generation Models",
                    "abstract": "We study an interesting problem in training neural network-based models for natural language generation tasks, which we call the \\emph{representation degeneration problem}. We observe that when training a model for natural language generation tasks through likelihood maximization with the weight tying trick, especially with big training datasets, most of the learnt word embeddings tend to degenerate and be distributed into a narrow cone, which largely limits the representation power of word embeddings. We analyze the conditions and causes of this problem and propose a novel regularization method to address it. Experiments on language modeling and machine translation show that our method can largely mitigate the representation degeneration problem and achieve better performance than baseline algorithms."
                },
                {
                    "title": "Graph Neural Networks Exponentially Lose Expressive Power for Node Classification",
                    "abstract": "Graph Neural Networks (graph NNs) are a promising deep learning approach for analyzing graph-structured data. However, it is known that they do not improve (or sometimes worsen) their predictive performance as we pile up many layers and add non-lineality. To tackle this problem, we investigate the expressive power of graph NNs via their asymptotic behaviors as the layer size tends to infinity. Our strategy is to generalize the forward propagation of a Graph Convolutional Network (GCN), which is a popular graph NN variant, as a specific dynamical system. In the case of a GCN, we show that when its weights satisfy the conditions determined by the spectra of the (augmented) normalized Laplacian, its output exponentially approaches the set of signals that carry information of the connected components and node degrees only for distinguishing nodes. Our theory enables us to relate the expressive power of GCNs with the topological information of the underlying graphs inherent in the graph spectra. To demonstrate this, we characterize the asymptotic behavior of GCNs on the Erd\u0151s -- Renyi graph. We show that when the Erd\u0151s -- Renyi graph is sufficiently dense and large, a broad range of GCNs on it suffers from the \"information loss\" in the limit of infinite layers with high probability. Based on the theory, we provide a principled guideline for weight normalization of graph NNs. We experimentally confirm that the proposed weight scaling enhances the predictive performance of GCNs in real data. Code is available at this https URL."
                },
                {
                    "title": "Generating Long Sequences with Sparse Transformers",
                    "abstract": "Transformers are powerful sequence models, but require time and memory that grows quadratically with the sequence length. In this paper we introduce sparse factorizations of the attention matrix which reduce this to $O(n \\sqrt{n})$. We also introduce a) a variation on architecture and initialization to train deeper networks, b) the recomputation of attention matrices to save memory, and c) fast attention kernels for training. We call networks with these changes Sparse Transformers, and show they can model sequences tens of thousands of timesteps long using hundreds of layers. We use the same architecture to model images, audio, and text from raw bytes, setting a new state of the art for density modeling of Enwik8, CIFAR-10, and ImageNet-64. We generate unconditional samples that demonstrate global coherence and great diversity, and show it is possible in principle to use self-attention to model sequences of length one million or more."
                },
                {
                    "title": "A Mean Field Theory of Batch Normalization",
                    "abstract": "We develop a mean field theory for batch normalization in fully-connected feedforward neural networks. In so doing, we provide a precise characterization of signal propagation and gradient backpropagation in wide batch-normalized networks at initialization. Our theory shows that gradient signals grow exponentially in depth and that these exploding gradients cannot be eliminated by tuning the initial weight variances or by adjusting the nonlinear activation function. Indeed, batch normalization itself is the cause of gradient explosion. As a result, vanilla batch-normalized networks without skip connections are not trainable at large depths for common initialization schemes, a prediction that we verify with a variety of empirical simulations. While gradient explosion cannot be eliminated, it can be reduced by tuning the network close to the linear regime, which improves the trainability of deep batch-normalized networks without residual connections. Finally, we investigate the learning dynamics of batch-normalized networks and observe that after a single step of optimization the networks achieve a relatively stable equilibrium in which gradients have dramatically smaller dynamic range. Our theory leverages Laplace, Fourier, and Gegenbauer transforms and we derive new identities that may be of independent interest."
                },
                {
                    "title": "On the Impact of the Activation Function on Deep Neural Networks Training",
                    "abstract": "The weight initialization and the activation function of deep neural networks have a crucial impact on the performance of the training procedure. An inappropriate selection can lead to the loss of information of the input during forward propagation and the exponential vanishing/exploding of gradients during back-propagation. Understanding the theoretical properties of untrained random networks is key to identifying which deep networks may be trained successfully as recently demonstrated by Samuel et al (2017) who showed that for deep feedforward neural networks only a specific choice of hyperparameters known as the `Edge of Chaos' can lead to good performance. While the work by Samuel et al (2017) discuss trainability issues, we focus here on training acceleration and overall performance. We give a comprehensive theoretical analysis of the Edge of Chaos and show that we can indeed tune the initialization parameters and the activation function in order to accelerate the training and improve the performance."
                },
                {
                    "title": "Fixup Initialization: Residual Learning Without Normalization",
                    "abstract": "Normalization layers are a staple in state-of-the-art deep neural network architectures. They are widely believed to stabilize training, enable higher learning rate, accelerate convergence and improve generalization, though the reason for their effectiveness is still an active research topic. In this work, we challenge the commonly-held beliefs by showing that none of the perceived benefits is unique to normalization. Specifically, we propose fixed-update initialization (Fixup), an initialization motivated by solving the exploding and vanishing gradient problem at the beginning of training via properly rescaling a standard initialization. We find training residual networks with Fixup to be as stable as training with normalization -- even for networks with 10,000 layers. Furthermore, with proper regularization, Fixup enables residual networks without normalization to achieve state-of-the-art performance in image classification and machine translation."
                },
                {
                    "title": "On Lazy Training in Differentiable Programming",
                    "abstract": "In a series of recent theoretical works, it was shown that strongly over-parameterized neural networks trained with gradient-based methods could converge exponentially fast to zero training loss, with their parameters hardly varying. In this work, we show that this \"lazy training\" phenomenon is not specific to over-parameterized neural networks, and is due to a choice of scaling, often implicit, that makes the model behave as its linearization around the initialization, thus yielding a model equivalent to learning with positive-definite kernels. Through a theoretical analysis, we exhibit various situations where this phenomenon arises in non-convex optimization and we provide bounds on the distance between the lazy and linearized optimization paths. Our numerical experiments bring a critical note, as we observe that the performance of commonly used non-linear deep convolutional neural networks in computer vision degrades when trained in the lazy regime. This makes it unlikely that \"lazy training\" is behind the many successes of neural networks in difficult high dimensional tasks."
                },
                {
                    "title": "Understanding Priors in Bayesian Neural Networks at the Unit Level",
                    "abstract": "We investigate deep Bayesian neural networks with Gaussian weight priors and a class of ReLU-like nonlinearities. Bayesian neural networks with Gaussian priors are well known to induce an L2, \"weight decay\", regularization. Our results characterize a more intricate regularization effect at the level of the unit activations. Our main result establishes that the induced prior distribution on the units before and after activation becomes increasingly heavy-tailed with the depth of the layer. We show that first layer units are Gaussian, second layer units are sub-exponential, and units in deeper layers are characterized by sub-Weibull distributions. Our results provide new theoretical insight on deep Bayesian neural networks, which we corroborate with simulation experiments."
                },
                {
                    "title": "Adaptive Input Representations for Neural Language Modeling",
                    "abstract": "We introduce adaptive input representations for neural language modeling which extend the adaptive softmax of Grave et al. (2017) to input representations of variable capacity. There are several choices on how to factorize the input and output layers, and whether to model words, characters or sub-word units. We perform a systematic comparison of popular choices for a self-attentional architecture. Our experiments show that models equipped with adaptive embeddings are more than twice as fast to train than the popular character input CNN while having a lower number of parameters. On the WikiText-103 benchmark we achieve 18.7 perplexity, an improvement of 10.5 perplexity compared to the previously best published result and on the Billion Word benchmark, we achieve 23.02 perplexity."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "Dynamical Isometry and a Mean Field Theory of CNNs: How to Train 10, 000-Layer Vanilla Convolutional Neural Networks",
                    "abstract": "In recent years, state-of-the-art methods in computer vision have utilized increasingly deep convolutional neural network architectures (CNNs), with some of the most successful models employing hundreds or even thousands of layers. A variety of pathologies such as vanishing/exploding gradients make training such deep networks challenging. While residual connections and batch normalization do enable training at these depths, it has remained unclear whether such specialized architecture designs are truly necessary to train deep CNNs. In this work, we demonstrate that it is possible to train vanilla CNNs with ten thousand layers or more simply by using an appropriate initialization scheme. We derive this initialization scheme theoretically by developing a mean field theory for signal propagation and by characterizing the conditions for dynamical isometry, the equilibration of singular values of the input-output Jacobian matrix. These conditions require that the convolution operator be an orthogonal transformation in the sense that it is norm-preserving. We present an algorithm for generating such random initial orthogonal convolution kernels and demonstrate empirically that they enable efficient training of extremely deep architectures."
                },
                {
                    "title": "Gaussian Process Behaviour in Wide Deep Neural Networks",
                    "abstract": "Whilst deep neural networks have shown great empirical success, there is still much work to be done to understand their theoretical properties. In this paper, we study the relationship between random, wide, fully connected, feedforward networks with more than one hidden layer and Gaussian processes with a recursive kernel definition. We show that, under broad conditions, as we make the architecture increasingly wide, the implied random function converges in distribution to a Gaussian process, formalising and extending existing results by Neal (1996) to deep networks. To evaluate convergence rates empirically, we use maximum mean discrepancy. We then compare finite Bayesian deep networks from the literature to Gaussian processes in terms of the key predictive quantities of interest, finding that in some cases the agreement can be very close. We discuss the desirability of Gaussian process behaviour and review non-Gaussian alternative models from the literature."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Resurrecting the sigmoid in deep learning through dynamical isometry: theory and practice",
                    "abstract": "It is well known that the initialization of weights in deep neural networks can have a dramatic impact on learning speed. For example, ensuring the mean squared singular value of a network's input-output Jacobian is $O(1)$ is essential for avoiding the exponential vanishing or explosion of gradients. The stronger condition that all singular values of the Jacobian concentrate near $1$ is a property known as dynamical isometry. For deep linear networks, dynamical isometry can be achieved through orthogonal weight initialization and has been shown to dramatically speed up learning; however, it has remained unclear how to extend these results to the nonlinear setting. We address this question by employing powerful tools from free probability theory to compute analytically the entire singular value distribution of a deep network's input-output Jacobian. We explore the dependence of the singular value distribution on the depth of the network, the weight initialization, and the choice of nonlinearity. Intriguingly, we find that ReLU networks are incapable of dynamical isometry. On the other hand, sigmoidal networks can achieve isometry, but only with orthogonal weight initialization. Moreover, we demonstrate empirically that deep nonlinear networks achieving dynamical isometry learn orders of magnitude faster than networks that do not. Indeed, we show that properly-initialized deep sigmoidal networks consistently outperform deep ReLU networks. Overall, our analysis reveals that controlling the entire distribution of Jacobian singular values is an important design consideration in deep learning."
                },
                {
                    "title": "Deep Neural Networks as Gaussian Processes",
                    "abstract": "It has long been known that a single-layer fully-connected neural network with an i.i.d. prior over its parameters is equivalent to a Gaussian process (GP), in the limit of infinite network width. This correspondence enables exact Bayesian inference for infinite width neural networks on regression tasks by means of evaluating the corresponding GP. Recently, kernel functions which mimic multi-layer random neural networks have been developed, but only outside of a Bayesian framework. As such, previous work has not identified that these kernels can be used as covariance functions for GPs and allow fully Bayesian prediction with a deep neural network. \nIn this work, we derive the exact equivalence between infinitely wide deep networks and GPs. We further develop a computationally efficient pipeline to compute the covariance function for these GPs. We then use the resulting GPs to perform Bayesian inference for wide deep neural networks on MNIST and CIFAR-10. We observe that trained neural network accuracy approaches that of the corresponding GP with increasing layer width, and that the GP uncertainty is strongly correlated with trained network prediction error. We further find that test performance increases as finite-width trained networks are made wider and more similar to a GP, and thus that GP predictions typically outperform those of finite-width networks. Finally we connect the performance of these GPs to the recent theory of signal propagation in random neural networks."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Deep Information Propagation",
                    "abstract": "We study the behavior of untrained neural networks whose weights and biases are randomly distributed using mean field theory. We show the existence of depth scales that naturally limit the maximum depth of signal propagation through these random networks. Our main practical result is to show that random networks may be trained precisely when information can travel through them. Thus, the depth scales that we identify provide bounds on how deep a network may be trained for a specific choice of hyperparameters. As a corollary to this, we argue that in networks at the edge of chaos, one of these depth scales diverges. Thus arbitrarily deep networks may be trained only sufficiently close to criticality. We show that the presence of dropout destroys the order-to-chaos critical point and therefore strongly limits the maximum trainable depth for random networks. Finally, we develop a mean field theory for backpropagation and we show that the ordered and chaotic phases correspond to regions of vanishing and exploding gradient respectively."
                },
                {
                    "title": "Layer Normalization",
                    "abstract": "Training state-of-the-art, deep neural networks is computationally expensive. One way to reduce the training time is to normalize the activities of the neurons. A recently introduced technique called batch normalization uses the distribution of the summed input to a neuron over a mini-batch of training cases to compute a mean and variance which are then used to normalize the summed input to that neuron on each training case. This significantly reduces the training time in feed-forward neural networks. However, the effect of batch normalization is dependent on the mini-batch size and it is not obvious how to apply it to recurrent neural networks. In this paper, we transpose batch normalization into layer normalization by computing the mean and variance used for normalization from all of the summed inputs to the neurons in a layer on a single training case. Like batch normalization, we also give each neuron its own adaptive bias and gain which are applied after the normalization but before the non-linearity. Unlike batch normalization, layer normalization performs exactly the same computation at training and test times. It is also straightforward to apply to recurrent neural networks by computing the normalization statistics separately at each time step. Layer normalization is very effective at stabilizing the hidden state dynamics in recurrent networks. Empirically, we show that layer normalization can substantially reduce the training time compared with previously published techniques."
                },
                {
                    "title": "Exponential expressivity in deep neural networks through transient chaos",
                    "abstract": "We combine Riemannian geometry with the mean field theory of high dimensional chaos to study the nature of signal propagation in generic, deep neural networks with random weights. Our results reveal an order-to-chaos expressivity phase transition, with networks in the chaotic phase computing nonlinear functions whose global curvature grows exponentially with depth but not width. We prove this generic class of deep random functions cannot be efficiently computed by any shallow network, going beyond prior work restricted to the analysis of single functions. Moreover, we formalize and quantitatively demonstrate the long conjectured idea that deep networks can disentangle highly curved manifolds in input space into flat manifolds in hidden space. Our theoretical analysis of the expressive power of deep networks broadly applies to arbitrary nonlinearities, and provides a quantitative underpinning for previously abstract notions about the geometry of deep functions."
                },
                {
                    "title": "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift",
                    "abstract": "Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.82% top-5 test error, exceeding the accuracy of human raters."
                },
                {
                    "title": "Scaling TransNormer to 175 Billion Parameters",
                    "abstract": "We present TransNormerLLM, the \ufb01rst linear attention-based Large Language Model (LLM) that outperforms conventional softmax attention-based models in terms of both accuracy and ef\ufb01ciency. TransNormerLLM evolves from the previous linear attention architecture TransNormer [32] by making advanced modi\ufb01cations that include positional embedding, linear attention acceleration, gating mechanism, tensor normalization, and inference acceleration and stabilization. Speci\ufb01cally, we use LRPE [34] together with an exponential decay to avoid attention dilution issues while allowing the model to retain global interactions between tokens. Additionally, we propose Lightning Attention, a cutting-edge technique that accelerates linear attention by more than twice in runtime and reduces memory usage by a remarkable four times. To further enhance the performance of TransNormer, we leverage a gating mechanism to smooth training and a new tensor normalization scheme to accelerate the model, resulting in an impressive acceleration of over 20% . Furthermore, we have developed a robust inference algorithm that ensures numerical stability and consistent inference speed, regardless of the sequence length, showcasing superior ef\ufb01ciency during both training and inference stages. Scalability is at the heart of our model\u2019s design, enabling seamless deployment on large-scale clusters and facilitating expansion to even more extensive models, all while maintaining outstanding performance metrics. Rigorous validation of our model design is achieved through a series of comprehensive experiments on our self-collected corpus, boasting a size exceeding 6TB and containing over 2 trillion tokens. To ensure data quality and relevance, we implement a new self-cleaning strategy to"
                },
                {
                    "title": "Feature Kernel Distillation",
                    "abstract": "Trained Neural Networks (NNs) can be viewed as data-dependent kernel machines, with predictions determined by the inner product of last-layer representations across inputs, referred to as the feature kernel . We explore the relevance of the feature kernel for Knowledge Distillation (KD), using a mechanistic understanding of an NN\u2019s optimisation process. We extend the theoretical analysis of Allen-Zhu & Li (2020) to show that a trained NN\u2019s feature kernel is highly dependent on its parameter initialisation, which biases different initialisations of the same architecture to learn different data attributes in a multi-view data setting. This enables us to prove that KD using only pairwise feature kernel comparisons can improve NN test accuracy in such settings, with both single & ensemble teacher models, whereas standard training without KD fails to generalise. We further use our theory to motivate practical considerations for improving student generalisation when using distillation with feature kernels, which allows us to pro-pose a novel approach: Feature Kernel Distillation (FKD). Finally, we experimentally corroborate our theory in the image classi\ufb01cation setting, showing that FKD is amenable to ensemble distillation, can transfer knowledge across datasets, and outperforms both vanilla KD & other feature kernel based KD baselines across a range of standard architectures & datasets."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we understand and mitigate the emergence of Outlier Features (OFs) during the training of deep neural networks to improve model quantization and performance?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of Outlier Feature Emergence (OFE) is crucial for the research community as it enhances our understanding of deep neural network training dynamics, which remains incomplete. By addressing OFE, we can uncover the fundamental properties of neural networks and the impact of design choices on model performance. This understanding could lead to practical applications, such as improved low-precision training and inference, which are essential for deploying models in resource-constrained environments. Furthermore, insights gained from this research could influence future studies on neural network interpretability and the roles of individual neurons, thereby advancing the field significantly.\n\n**[Question 3] - Why is it hard?**  \nThe challenge in addressing OFE lies in the complexity of high-dimensional non-convex loss landscapes and the multitude of design choices that affect neural network training. Naive approaches may fail because they do not account for the intricate interactions between network architecture, optimizer hyperparameters, and the training dynamics that lead to the emergence of OFs. Additionally, the lack of a clear understanding of the mechanisms behind OFE complicates the development of effective interventions. Technical obstacles include accurately measuring OFs and their impact on model performance, as well as ensuring that any proposed solutions do not compromise model convergence or stability.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely overlooked the specific phenomenon of Outlier Features, focusing instead on more general aspects of neural network initialization and training. Existing solutions may not have addressed OFE due to a lack of comprehensive metrics to quantify OFs or an insufficient understanding of their underlying causes. Barriers such as the complexity of neural network dynamics and the variability in design choices have hindered progress. Our approach differs by systematically investigating the factors influencing OFE and proposing targeted interventions based on a deeper understanding of the training process, which has not been adequately explored in prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves two main components: (1) developing interventions to minimize Outlier Feature Emergence without compromising model convergence or training stability, and (2) enhancing our understanding of the underlying causes of OFs during training. We will utilize a quantitative metric, specifically a"
            }
        },
        "author_data": {
            "d32fd6dd-8947-4435-a0a7-7da7efe98ff7": {
                "pk": "d32fd6dd-8947-4435-a0a7-7da7efe98ff7",
                "name": "Bobby He",
                "collaborators": [
                    "Thomas Hofmann",
                    "Soufiane Hayou",
                    "Yee Whye Teh",
                    "Yuhui Ding",
                    "Antonio Orvieto",
                    "Gintare Karolina Dziugaite",
                    "Balaji Lakshminarayanan",
                    "Francisca Vasconcelos",
                    "Nalini Singh",
                    "Sidak Pal Singh"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Bayesian Inference",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Simplifying Transformer Blocks",
                        "abstract": "A simple design recipe for deep Transformers is to compose identical building blocks. But standard transformer blocks are far from simple, interweaving attention and MLP sub-blocks with skip connections & normalisation layers in precise arrangements. This complexity leads to brittle architectures, where seemingly minor changes can significantly reduce training speed, or render models untrainable.   In this work, we ask to what extent the standard transformer block can be simplified? Combining signal propagation theory and empirical observations, we motivate modifications that allow many block components to be removed with no loss of training speed, including skip connections, projection or value parameters, sequential sub-blocks and normalisation layers. In experiments on both autoregressive decoder-only and BERT encoder-only models, our simplified transformers emulate the per-update training speed and performance of standard transformers, while enjoying 15% faster training throughput, and using 15% fewer parameters."
                    },
                    {
                        "title": "Recurrent Distance Filtering for Graph Representation Learning",
                        "abstract": "Graph neural networks based on iterative one-hop message passing have been shown to struggle in harnessing the information from distant nodes effectively. Conversely, graph transformers allow each node to attend to all other nodes directly, but lack graph inductive bias and have to rely on ad-hoc positional encoding. In this paper, we propose a new architecture to reconcile these challenges. Our approach stems from the recent breakthroughs in long-range modeling provided by deep state-space models: for a given target node, our model aggregates other nodes by their shortest distances to the target and uses a linear RNN to encode the sequence of hop representations. The linear RNN is parameterized in a particular diagonal form for stable long-range signal propagation and is theoretically expressive enough to encode the neighborhood hierarchy. With no need for positional encoding, we empirically show that the performance of our model is comparable to or better than that of state-of-the-art graph transformers on various benchmarks, with a significantly reduced computational cost. Our code is open-source at https://github.com/skeletondyh/GRED."
                    },
                    {
                        "title": "Probabilistic fine-tuning of pruning masks and PAC-Bayes self-bounded learning",
                        "abstract": "We study an approach to learning pruning masks by optimizing the expected loss of stochastic pruning masks, i.e., masks which zero out each weight independently with some weight-specific probability. We analyze the training dynamics of the induced stochastic predictor in the setting of linear regression, and observe a data-adaptive L1 regularization term, in contrast to the dataadaptive L2 regularization term known to underlie dropout in linear regression. We also observe a preference to prune weights that are less well-aligned with the data labels. We evaluate probabilistic fine-tuning for optimizing stochastic pruning masks for neural networks, starting from masks produced by several baselines. In each case, we see improvements in test error over baselines, even after we threshold fine-tuned stochastic pruning masks. Finally, since a stochastic pruning mask induces a stochastic neural network, we consider training the weights and/or pruning probabilities simultaneously to minimize a PAC-Bayes bound on generalization error. Using data-dependent priors, we obtain a selfbounded learning algorithm with strong performance and numerically tight bounds. In the linear model, we show that a PAC-Bayes generalization error bound is controlled by the magnitude of the change in feature alignment between the 'prior' and 'posterior' data."
                    },
                    {
                        "title": "Bayesian Deep Ensembles via the Neural Tangent Kernel",
                        "abstract": "We explore the link between deep ensembles and Gaussian processes (GPs) through the lens of the Neural Tangent Kernel (NTK): a recent development in understanding the training dynamics of wide neural networks (NNs). Previous work has shown that even in the infinite width limit, when NNs become GPs, there is no GP posterior interpretation to a deep ensemble trained with squared error loss. We introduce a simple modification to standard deep ensembles training, through addition of a computationally-tractable, randomised and untrainable function to each ensemble member, that enables a posterior interpretation in the infinite width limit. When ensembled together, our trained NNs give an approximation to a posterior predictive distribution, and we prove that our Bayesian deep ensembles make more conservative predictions than standard deep ensembles in the infinite width limit. Finally, using finite width NNs we demonstrate that our Bayesian deep ensembles faithfully emulate the analytic posterior predictive when available, and can outperform standard deep ensembles in various out-of-distribution settings, for both regression and classification tasks."
                    },
                    {
                        "title": "UncertaINR: Uncertainty Quantification of End-to-End Implicit Neural Representations for Computed Tomography",
                        "abstract": "Implicit neural representations (INRs) have achieved impressive results for scene reconstruction and computer graphics, where their performance has primarily been assessed on reconstruction accuracy. As INRs make their way into other domains, where model predictions inform high-stakes decision-making, uncertainty quantification of INR inference is becoming critical. To that end, we study a Bayesian reformulation of INRs, UncertaINR, in the context of computed tomography, and evaluate several Bayesian deep learning implementations in terms of accuracy and calibration. We find that they achieve well-calibrated uncertainty, while retaining accuracy competitive with other classical, INR-based, and CNN-based reconstruction techniques. Contrary to common intuition in the Bayesian deep learning literature, we find that INRs obtain the best calibration with computationally efficient Monte Carlo dropout, outperforming Hamiltonian Monte Carlo and deep ensembles. Moreover, in contrast to the best-performing prior approaches, UncertaINR does not require a large training dataset, but only a handful of validation images."
                    },
                    {
                        "title": "Hallmarks of Optimization Trajectories in Neural Networks: Directional Exploration and Redundancy",
                        "abstract": "We propose a fresh take on understanding the mechanisms of neural networks by analyzing the rich directional structure of optimization trajectories, represented by their pointwise parameters. Towards this end, we introduce some natural notions of the complexity of optimization trajectories, both qualitative and quantitative, which hallmark the directional nature of optimization in neural networks: when is there redundancy, and when exploration. We use them to reveal the inherent nuance and interplay involved between various optimization choices, such as momentum and weight decay. Further, the trajectory perspective helps us see the effect of scale on regularizing the directional nature of trajectories, and as a by-product, we also observe an intriguing heterogeneity of Q,K,V dynamics in the middle attention layers in LLMs and which is homogenized by scale. Importantly, we put the significant directional redundancy observed to the test by demonstrating that training only scalar batchnorm parameters some while into training matches the performance of training the entire network, which thus exhibits the potential of hybrid optimization schemes that are geared towards efficiency."
                    },
                    {
                        "title": "Stable ResNet",
                        "abstract": "Deep ResNet architectures have achieved state of the art performance on many tasks. While they solve the problem of gradient vanishing, they might suffer from gradient exploding as the depth becomes large (Yang et al. 2017). Moreover, recent results have shown that ResNet might lose expressivity as the depth goes to infinity (Yang et al. 2017, Hayou et al. 2019). To resolve these issues, we introduce a new class of ResNet architectures, called Stable ResNet, that have the property of stabilizing the gradient while ensuring expressivity in the infinite depth limit."
                    }
                ]
            },
            "7aed5b85-f4f2-4fff-b9d4-f026f1d3d71b": {
                "pk": "7aed5b85-f4f2-4fff-b9d4-f026f1d3d71b",
                "name": "Lorenzo Noci",
                "collaborators": [
                    "Thomas Hofmann",
                    "Gregor Bachmann",
                    "Antonio Orvieto",
                    "Sotiris Anagnostidis",
                    "Kevin Roth",
                    "Sebastian Nowozin",
                    "Luca Biggio",
                    "Aurelien Lucchi",
                    "Mufan Bill Li",
                    "Kai Lion"
                ],
                "domain": [
                    "Deep Learning",
                    "Bayesian Neural Networks",
                    "Transformers",
                    "Knowledge Graphs"
                ],
                "publications": [
                    {
                        "title": "How Good is a Single Basin?",
                        "abstract": "The multi-modal nature of neural loss landscapes is often considered to be the main driver behind the empirical success of deep ensembles. In this work, we probe this belief by constructing various \"connected\" ensembles which are restricted to lie in the same basin. Through our experiments, we demonstrate that increased connectivity indeed negatively impacts performance. However, when incorporating the knowledge from other basins implicitly through distillation, we show that the gap in performance can be mitigated by re-discovering (multi-basin) deep ensembles within a single basin. Thus, we conjecture that while the extra-basin knowledge is at least partially present in any given basin, it cannot be easily harnessed without learning it from other basins."
                    },
                    {
                        "title": "How Tempering Fixes Data Augmentation in Bayesian Neural Networks",
                        "abstract": "While Bayesian neural networks (BNNs) provide a sound and principled alternative to standard neural networks, an artificial sharpening of the posterior usually needs to be applied to reach comparable performance. This is in stark contrast to theory, dictating that given an adequate prior and a well-specified model, the untempered Bayesian posterior should achieve optimal performance. Despite the community's extensive efforts, the observed gains in performance still remain disputed with several plausible causes pointing at its origin. While data augmentation has been empirically recognized as one of the main drivers of this effect, a theoretical account of its role, on the other hand, is largely missing. In this work we identify two interlaced factors concurrently influencing the strength of the cold posterior effect, namely the correlated nature of augmentations and the degree of invariance of the employed model to such transformations. By theoretically analyzing simplified settings, we prove that tempering implicitly reduces the misspecification arising from modeling augmentations as i.i.d. data. The temperature mimics the role of the effective sample size, reflecting the gain in information provided by the augmentations. We corroborate our theoretical findings with extensive empirical evaluations, scaling to realistic BNNs. By relying on the framework of group convolutions, we experiment with models of varying inherent degree of invariance, confirming its hypothesized relationship with the optimal temperature."
                    },
                    {
                        "title": "Why do Learning Rates Transfer? Reconciling Optimization and Scaling Limits for Deep Learning",
                        "abstract": "Recently, there has been growing evidence that if the width and depth of a neural network are scaled toward the so-called rich feature learning limit ($\\mu$P and its depth extension), then some hyperparameters - such as the learning rate - exhibit transfer from small to very large models, thus reducing the cost of hyperparameter tuning. From an optimization perspective, this phenomenon is puzzling, as it implies that the loss landscape is remarkably consistent across very different model sizes. In this work, we find empirical evidence that learning rate transfer can be attributed to the fact that under $\\mu$P and its depth extension, the largest eigenvalue of the training loss Hessian (i.e. the sharpness) is largely independent of the width and depth of the network for a sustained period of training time. On the other hand, we show that under the neural tangent kernel (NTK) regime, the sharpness exhibits very different dynamics at different scales, thus preventing learning rate transfer. But what causes these differences in the sharpness dynamics? Through a connection between the spectra of the Hessian and the NTK matrix, we argue that the cause lies in the presence (for $\\mu$P) or progressive absence (for the NTK regime) of feature learning, which results in a different evolution of the NTK, and thus of the sharpness. We corroborate our claims with a substantial suite of experiments, covering a wide range of datasets and architectures: from ResNets and Vision Transformers trained on benchmark vision datasets to Transformers-based language models trained on WikiText"
                    },
                    {
                        "title": "Precise characterization of the prior predictive distribution of deep ReLU networks",
                        "abstract": "Recent works on Bayesian neural networks (BNNs) have highlighted the need to better understand the implications of using Gaussian priors in combination with the compositional structure of the network architecture. Similar in spirit to the kind of analysis that has been developed to devise better initialization schemes for neural networks (cf. He- or Xavier initialization), we derive a precise characterization of the prior predictive distribution of finite-width ReLU networks with Gaussian weights. While theoretical results have been obtained for their heavy-tailedness, the full characterization of the prior predictive distribution (i.e. its density, CDF and moments), remained unknown prior to this work. Our analysis, based on the Meijer-G function, allows us to quantify the influence of architectural choices such as the width or depth of the network on the resulting shape of the prior predictive distribution. We also formally connect our results to previous work in the infinite width setting, demonstrating that the moments of the distribution converge to those of a normal log-normal mixture in the infinite depth limit. Finally, our results provide valuable guidance on prior design: for instance, controlling the predictive variance with depth- and width-informed priors on the weights of the network."
                    },
                    {
                        "title": "Disentangling the Roles of Curation, Data-Augmentation and the Prior in the Cold Posterior Effect",
                        "abstract": "The \"cold posterior effect\" (CPE) in Bayesian deep learning describes the uncomforting observation that the predictive performance of Bayesian neural networks can be significantly improved if the Bayes posterior is artificially sharpened using a temperature parameter T<1. The CPE is problematic in theory and practice and since the effect was identified many researchers have proposed hypotheses to explain the phenomenon. However, despite this intensive research effort the effect remains poorly understood. In this work we provide novel and nuanced evidence relevant to existing explanations for the cold posterior effect, disentangling three hypotheses: 1. The dataset curation hypothesis of Aitchison (2020): we show empirically that the CPE does not arise in a real curated data set but can be produced in a controlled experiment with varying curation strength. 2. The data augmentation hypothesis of Izmailov et al. (2021) and Fortuin et al. (2021): we show empirically that data augmentation is sufficient but not necessary for the CPE to be present. 3. The bad prior hypothesis of Wenzel et al. (2020): we use a simple experiment evaluating the relative importance of the prior and the likelihood, strongly linking the CPE to the prior. Our results demonstrate how the CPE can arise in isolation from synthetic curation, data augmentation, and bad priors. Cold posteriors observed \"in the wild\" are therefore unlikely to arise from a single simple cause; as a result, we do not expect a simple \"fix\" for cold posteriors."
                    },
                    {
                        "title": "The Curious Case of Benign Memorization",
                        "abstract": "Despite the empirical advances of deep learning across a variety of learning tasks, our theoretical understanding of its success is still very restricted. One of the key challenges is the overparametrized nature of modern models, enabling complete overfitting of the data even if the labels are randomized, i.e. networks can completely \\textit{memorize} all given patterns. While such a memorization capacity seems worrisome, in this work we show that under training protocols that include \\textit{data augmentation}, neural networks learn to memorize entirely random labels in a benign way, i.e. they learn embeddings that lead to highly non-trivial performance under nearest neighbour probing. We demonstrate that deep models have the surprising ability to separate noise from signal by distributing the task of memorization and feature learning to different layers. As a result, only the very last layers are used for memorization, while preceding layers encode performant features which remain largely unaffected by the label noise. We explore the intricate role of the augmentations used for training and identify a memorization-generalization trade-off in terms of their diversity, marking a clear distinction to all previous works. Finally, we give a first explanation for the emergence of benign memorization by showing that \\textit{malign} memorization under data augmentation is infeasible due to the insufficient capacity of the model for the increased sample size. As a consequence, the network is forced to leverage the correlated nature of the augmentations and as a result learns meaningful features. To complete the picture, a better theory of feature learning in deep neural networks is required to fully understand the origins of this phenomenon."
                    },
                    {
                        "title": "Achieving a Better Stability-Plasticity Trade-off via Auxiliary Networks in Continual Learning",
                        "abstract": "In contrast to the natural capabilities of humans to learn new tasks in a sequential fashion, neural networks are known to suffer from catastrophic forgetting, where the model's performances on old tasks drop dramatically after being optimized for a new task. Since then, the continual learning (CL) community has proposed several solutions aiming to equip the neural network with the ability to learn the current task (plasticity) while still achieving high accuracy on the previous tasks (stability). Despite remarkable improvements, the plasticity-stability trade-off is still far from being solved and its underlying mechanism is poorly understood. In this work, we propose Auxiliary Network Continual Learning (ANCL), a novel method that applies an additional auxiliary network which promotes plasticity to the continually learned model which mainly focuses on stability. More concretely, the proposed framework materializes in a regularizer that naturally interpolates between plasticity and stability, surpassing strong baselines on task incremental and class incremental scenarios. Through extensive analyses on ANCL solutions, we identify some essential principles beneath the stability-plasticity trade-off."
                    },
                    {
                        "title": "Disentangling Linear Mode-Connectivity",
                        "abstract": "Linear mode-connectivity (LMC) (or lack thereof) is one of the intriguing characteristics of neural network loss landscapes. While empirically well established, it unfortunately still lacks a proper theoretical understanding. Even worse, although empirical data points are abound, a systematic study of when networks exhibit LMC is largely missing in the literature. In this work we aim to close this gap. We explore how LMC is affected by three factors: (1) architecture (sparsity, weight-sharing), (2) training strategy (optimization setup) as well as (3) the underlying dataset. We place particular emphasis on minimal but non-trivial settings, removing as much unnecessary complexity as possible. We believe that our insights can guide future theoretical works on uncovering the inner workings of LMC."
                    },
                    {
                        "title": "Signal Propagation in Transformers: Theoretical Perspectives and the Role of Rank Collapse",
                        "abstract": "Transformers have achieved remarkable success in several domains, ranging from natural language processing to computer vision. Nevertheless, it has been recently shown that stacking self-attention layers - the distinctive architectural component of Transformers - can result in rank collapse of the tokens' representations at initialization. The question of if and how rank collapse affects training is still largely unanswered, and its investigation is necessary for a more comprehensive understanding of this architecture. In this work, we shed new light on the causes and the effects of this phenomenon. First, we show that rank collapse of the tokens' representations hinders training by causing the gradients of the queries and keys to vanish at initialization. Furthermore, we provide a thorough description of the origin of rank collapse and discuss how to prevent it via an appropriate depth-dependent scaling of the residual branches. Finally, our analysis unveils that specific architectural hyperparameters affect the gradients of queries and values differently, leading to disproportionate gradient norms. This suggests an explanation for the widespread use of adaptive methods for Transformers' optimization."
                    },
                    {
                        "title": "Depthwise Hyperparameter Transfer in Residual Networks: Dynamics and Scaling Limit",
                        "abstract": "The cost of hyperparameter tuning in deep learning has been rising with model sizes, prompting practitioners to find new tuning methods using a proxy of smaller networks. One such proposal uses $\\mu$P parameterized networks, where the optimal hyperparameters for small width networks transfer to networks with arbitrarily large width. However, in this scheme, hyperparameters do not transfer across depths. As a remedy, we study residual networks with a residual branch scale of $1/\\sqrt{\\text{depth}}$ in combination with the $\\mu$P parameterization. We provide experiments demonstrating that residual architectures including convolutional ResNets and Vision Transformers trained with this parameterization exhibit transfer of optimal hyperparameters across width and depth on CIFAR-10 and ImageNet. Furthermore, our empirical findings are supported and motivated by theory. Using recent developments in the dynamical mean field theory (DMFT) description of neural network learning dynamics, we show that this parameterization of ResNets admits a well-defined feature learning joint infinite-width and infinite-depth limit and show convergence of finite-size network dynamics towards this limit."
                    },
                    {
                        "title": "Adversarial Learning for Debiasing Knowledge Graph Embeddings",
                        "abstract": "Knowledge Graphs (KG) are gaining increasing attention in both academia and industry. Despite their diverse benefits, recent research have identified social and cultural biases embedded in the representations learned from KGs. Such biases can have detrimental consequences on different population and minority groups as applications of KG begin to intersect and interact with social spheres. This paper aims at identifying and mitigating such biases in Knowledge Graph (KG) embeddings. As a first step, we explore popularity bias -- the relationship between node popularity and link prediction accuracy. In case of node2vec graph embeddings, we find that prediction accuracy of the embedding is negatively correlated with the degree of the node. However, in case of knowledge-graph embeddings (KGE), we observe an opposite trend. As a second step, we explore gender bias in KGE, and a careful examination of popular KGE algorithms suggest that sensitive attribute like the gender of a person can be predicted from the embedding. This implies that such biases in popular KGs is captured by the structural properties of the embedding. As a preliminary solution to debiasing KGs, we introduce a novel framework to filter out the sensitive attribute information from the KG embeddings, which we call FAN (Filtering Adversarial Network). We also suggest the applicability of FAN for debiasing other network embeddings which could be explored in future work."
                    },
                    {
                        "title": "Dynamic Context Pruning for Efficient and Interpretable Autoregressive Transformers",
                        "abstract": "Autoregressive Transformers adopted in Large Language Models (LLMs) are hard to scale to long sequences. Despite several works trying to reduce their computational cost, most of LLMs still adopt attention layers between all pairs of tokens in the sequence, thus incurring a quadratic cost. In this study, we present a novel approach that dynamically prunes contextual information while preserving the model's expressiveness, resulting in reduced memory and computational requirements during inference. Our method employs a learnable mechanism that determines which uninformative tokens can be dropped from the context at any point across the generation process. By doing so, our approach not only addresses performance concerns but also enhances interpretability, providing valuable insight into the model's decision-making process. Our technique can be applied to existing pre-trained models through a straightforward fine-tuning process, and the pruning strength can be specified by a sparsity parameter. Notably, our empirical findings demonstrate that we can effectively prune up to 80\\% of the context without significant performance degradation on downstream tasks, offering a valuable tool for mitigating inference costs. Our reference implementation achieves up to $2\\times$ increase in inference throughput and even greater memory savings."
                    },
                    {
                        "title": "The Shaped Transformer: Attention Models in the Infinite Depth-and-Width Limit",
                        "abstract": "In deep learning theory, the covariance matrix of the representations serves as a proxy to examine the network's trainability. Motivated by the success of Transformers, we study the covariance matrix of a modified Softmax-based attention model with skip connections in the proportional limit of infinite-depth-and-width. We show that at initialization the limiting distribution can be described by a stochastic differential equation (SDE) indexed by the depth-to-width ratio. To achieve a well-defined stochastic limit, the Transformer's attention mechanism is modified by centering the Softmax output at identity, and scaling the Softmax logits by a width-dependent temperature parameter. We examine the stability of the network through the corresponding SDE, showing how the scale of both the drift and diffusion can be elegantly controlled with the aid of residual connections. The existence of a stable SDE implies that the covariance structure is well-behaved, even for very large depth and width, thus preventing the notorious issues of rank degeneracy in deep attention models. Finally, we show, through simulations, that the SDE provides a surprisingly good description of the corresponding finite-size model. We coin the name shaped Transformer for these architectural modifications."
                    }
                ]
            },
            "3f3c9ba9-58d7-414e-ad09-573bac36dcff": {
                "pk": "3f3c9ba9-58d7-414e-ad09-573bac36dcff",
                "name": "Daniele Paliotta",
                "collaborators": [
                    "Fran\u00e7ois Fleuret",
                    "B\u00e1lint M\u00e1t\u00e9",
                    "Mathieu Alain",
                    "Matteo Pagliardini",
                    "Martin Jaggi",
                    "Junxiong Wang",
                    "Avner May",
                    "Alexander M. Rush",
                    "Tri Dao",
                    "Atul Kumar Sinha"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Transformer",
                    "Deep Learning",
                    "High Energy Physics"
                ],
                "publications": [
                    {
                        "title": "Graph Neural Networks Go Forward-Forward",
                        "abstract": "We present the Graph Forward-Forward (GFF) algorithm, an extension of the Forward-Forward procedure to graphs, able to handle features distributed over a graph's nodes. This allows training graph neural networks with forward passes only, without backpropagation. Our method is agnostic to the message-passing scheme, and provides a more biologically plausible learning scheme than backpropagation, while also carrying computational advantages. With GFF, graph neural networks are trained greedily layer by layer, using both positive and negative samples. We run experiments on 11 standard graph property prediction tasks, showing how GFF provides an effective alternative to backpropagation for training graph neural networks. This shows in particular that this procedure is remarkably efficient in spite of combining the per-layer training with the locality of the processing in a GNN."
                    },
                    {
                        "title": "Faster Causal Attention Over Large Sequences Through Sparse Flash Attention",
                        "abstract": "Transformer-based language models have found many diverse applications requiring them to process sequences of increasing length. For these applications, the causal self-attention -- which is the only component scaling quadratically w.r.t. the sequence length -- becomes a central concern. While many works have proposed schemes to sparsify the attention patterns and reduce the computational overhead of self-attention, those are often limited by implementations concerns and end up imposing a simple and static structure over the attention matrix. Conversely, implementing more dynamic sparse attentions often results in runtimes significantly slower than computing the full attention using the Flash implementation from Dao et al. (2022). We extend FlashAttention to accommodate a large class of attention sparsity patterns that, in particular, encompass key/query dropping and hashing-based attention. This leads to implementations with no computational complexity overhead and a multi-fold runtime speedup on top of FlashAttention. Even with relatively low degrees of sparsity, our method improves visibly upon FlashAttention as the sequence length increases. Without sacrificing perplexity, we increase the training speed of a transformer language model by $2.0\\times$ and $3.3\\times$ for sequences of respectively $8k$ and $16k$ tokens."
                    },
                    {
                        "title": "The Mamba in the Llama: Distilling and Accelerating Hybrid Models",
                        "abstract": "Linear RNN architectures, like Mamba, can be competitive with Transformer models in language modeling while having advantageous deployment characteristics. Given the focus on training large-scale Transformer models, we consider the challenge of converting these pretrained models for deployment. We demonstrate that it is feasible to distill large Transformers into linear RNNs by reusing the linear projection weights from attention layers with academic GPU resources. The resulting hybrid model, which incorporates a quarter of the attention layers, achieves performance comparable to the original Transformer in chat benchmarks and outperforms open-source hybrid Mamba models trained from scratch with trillions of tokens in both chat benchmarks and general benchmarks. Moreover, we introduce a hardware-aware speculative decoding algorithm that accelerates the inference speed of Mamba and hybrid models. Overall we show how, with limited computation resources, we can remove many of the original attention layers and generate from the resulting model more efficiently. Our top-performing model, distilled from Llama3-8B-Instruct, achieves a 29.61 length-controlled win rate on AlpacaEval 2 against GPT-4 and 7.35 on MT-Bench, surpassing the best instruction-tuned linear RNN model."
                    },
                    {
                        "title": "SUPA: A Lightweight Diagnostic Simulator for Machine Learning in Particle Physics",
                        "abstract": "Deep learning methods have gained popularity in high energy physics for fast modeling of particle showers in detectors. Detailed simulation frameworks such as the gold standard Geant4 are computationally intensive, and current deep generative architectures work on discretized, lower resolution versions of the detailed simulation. The development of models that work at higher spatial resolutions is currently hindered by the complexity of the full simulation data, and by the lack of simpler, more interpretable benchmarks. Our contribution is SUPA, the SUrrogate PArticle propagation simulator, an algorithm and software package for generating data by simulating simplified particle propagation, scattering and shower development in matter. The generation is extremely fast and easy to use compared to Geant4, but still exhibits the key characteristics and challenges of the detailed simulation. We support this claim experimentally by showing that performance of generative models on data from our simulator reflects the performance on a dataset generated with Geant4. The proposed simulator generates thousands of particle showers per second on a desktop machine, a speed up of up to 6 orders of magnitudes over Geant4, and stores detailed geometric information about the shower propagation. SUPA provides much greater flexibility for setting initial conditions and defining multiple benchmarks for the development of models. Moreover, interpreting particle showers as point clouds creates a connection to geometric machine learning and provides challenging and fundamentally new datasets for the field.   The code for SUPA is available at https://github.com/itsdaniele/SUPA."
                    }
                ]
            },
            "9335a95e-1f38-4013-8757-2a7cd9e1522f": {
                "pk": "9335a95e-1f38-4013-8757-2a7cd9e1522f",
                "name": "Imanol Schlag",
                "collaborators": [
                    "J\u00fcrgen Schmidhuber",
                    "Kazuki Irie",
                    "Thomas Hofmann",
                    "R\u00f3bert Csord\u00e1s",
                    "Anton Sch\u00e4fer",
                    "Tiago Pimentel",
                    "Yuhuai Wu",
                    "Ethan Dyer",
                    "Behnam Neyshabur",
                    "Tsendsuren Munkhdalai"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Neural Networks",
                    "Combinatorial Learning",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Learning to Reason with Third-Order Tensor Products",
                        "abstract": "We combine Recurrent Neural Networks with Tensor Product Representations to learn combinatorial representations of sequential data. This improves symbolic interpretation and systematic generalisation. Our architecture is trained end-to-end through gradient descent on a variety of simple natural language reasoning tasks, significantly outperforming the latest state-of-the-art models in single-task and all-tasks settings. We also augment a subset of the data such that training and test data exhibit large systematic differences and show that our approach generalises better than the previous state-of-the-art."
                    },
                    {
                        "title": "Learning Associative Inference Using Fast Weight Memory",
                        "abstract": "Humans can quickly associate stimuli to solve problems in novel contexts. Our novel neural network model learns state representations of facts that can be composed to perform such associative inference. To this end, we augment the LSTM model with an associative memory, dubbed Fast Weight Memory (FWM). Through differentiable operations at every step of a given input sequence, the LSTM updates and maintains compositional associations stored in the rapidly changing FWM weights. Our model is trained end-to-end by gradient descent and yields excellent performance on compositional language reasoning problems, meta-reinforcement-learning for POMDPs, and small-scale word-level language modelling."
                    },
                    {
                        "title": "Linear Transformers Are Secretly Fast Weight Programmers",
                        "abstract": "We show the formal equivalence of linearised self-attention mechanisms and fast weight controllers from the early '90s, where a ``slow\" neural net learns by gradient descent to program the ``fast weights\" of another net through sequences of elementary programming instructions which are additive outer products of self-invented activation patterns (today called keys and values). Such Fast Weight Programmers (FWPs) learn to manipulate the contents of a finite memory and dynamically interact with it. We infer a memory capacity limitation of recent linearised softmax attention variants, and replace the purely additive outer products by a delta rule-like programming instruction, such that the FWP can more easily learn to correct the current mapping from keys to values. The FWP also learns to compute dynamically changing learning rates. We also propose a new kernel function to linearise attention which balances simplicity and effectiveness. We conduct experiments on synthetic retrieval problems as well as standard machine translation and language modelling tasks which demonstrate the benefits of our methods."
                    },
                    {
                        "title": "Going Beyond Linear Transformers with Recurrent Fast Weight Programmers",
                        "abstract": "Transformers with linearised attention (''linear Transformers'') have demonstrated the practical scalability and effectiveness of outer product-based Fast Weight Programmers (FWPs) from the '90s. However, the original FWP formulation is more general than the one of linear Transformers: a slow neural network (NN) continually reprograms the weights of a fast NN with arbitrary architecture. In existing linear Transformers, both NNs are feedforward and consist of a single layer. Here we explore new variations by adding recurrence to the slow and fast nets. We evaluate our novel recurrent FWPs (RFWPs) on two synthetic algorithmic tasks (code execution and sequential ListOps), Wikitext-103 language models, and on the Atari 2600 2D game environment. Our models exhibit properties of Transformers and RNNs. In the reinforcement learning setting, we report large improvements over LSTM in several Atari games. Our code is public."
                    },
                    {
                        "title": "Improving Baselines in the Wild",
                        "abstract": "We share our experience with the recently released WILDS benchmark, a collection of ten datasets dedicated to developing models and training strategies which are robust to domain shifts. Several experiments yield a couple of critical observations which we believe are of general interest for any future work on WILDS. Our study focuses on two datasets: iWildCam and FMoW. We show that (1) Conducting separate cross-validation for each evaluation metric is crucial for both datasets, (2) A weak correlation between validation and test performance might make model development difficult for iWildCam, (3) Minor changes in the training of hyper-parameters improve the baseline by a relatively large margin (mainly on FMoW), (4) There is a strong correlation between certain domains and certain target labels (mainly on iWildCam). To the best of our knowledge, no prior work on these datasets has reported these observations despite their obvious importance. Our code is public."
                    },
                    {
                        "title": "A Modern Self-Referential Weight Matrix That Learns to Modify Itself",
                        "abstract": "The weight matrix (WM) of a neural network (NN) is its program. The programs of many traditional NNs are learned through gradient descent in some error function, then remain fixed. The WM of a self-referential NN, however, can keep rapidly modifying all of itself during runtime. In principle, such NNs can meta-learn to learn, and meta-meta-learn to meta-learn to learn, and so on, in the sense of recursive self-improvement. While NN architectures potentially capable of implementing such behaviour have been proposed since the '90s, there have been few if any practical studies. Here we revisit such NNs, building upon recent successes of fast weight programmers and closely related linear Transformers. We propose a scalable self-referential WM (SRWM) that learns to use outer products and the delta update rule to modify itself. We evaluate our SRWM in supervised few-shot learning and in multi-task reinforcement learning with procedurally generated game environments. Our experiments demonstrate both practical applicability and competitive performance of the proposed SRWM. Our code is public."
                    },
                    {
                        "title": "Navigating Scaling Laws: Compute Optimality in Adaptive Model Training",
                        "abstract": "In recent years, the state-of-the-art in deep learning has been dominated by very large models that have been pre-trained on vast amounts of data. The paradigm is very simple: investing more computational resources (optimally) leads to better performance, and even predictably so; neural scaling laws have been derived that accurately forecast the performance of a network for a desired level of compute. This leads to the notion of a `compute-optimal' model, i.e. a model that allocates a given level of compute during training optimally to maximize performance. In this work, we extend the concept of optimality by allowing for an `adaptive' model, i.e. a model that can change its shape during training. By doing so, we can design adaptive models that optimally traverse between the underlying scaling laws and outpace their `static' counterparts, leading to a significant reduction in the required compute to reach a given target performance. We show that our approach generalizes across modalities and different shape parameters."
                    },
                    {
                        "title": "On the Effect of (Near) Duplicate Subwords in Language Modelling",
                        "abstract": "Tokenisation is a core part of language models (LMs). It involves splitting a character sequence into subwords which are assigned arbitrary indices before being served to the LM. While typically lossless, however, this process may lead to less sample efficient LM training: as it removes character-level information, it could make it harder for LMs to generalise across similar subwords, such as now and Now. We refer to such subwords as near duplicates. In this paper, we study the impact of near duplicate subwords on LM training efficiency. First, we design an experiment that gives us an upper bound to how much we should expect a model to improve if we could perfectly generalise across near duplicates. We do this by duplicating each subword in our LM's vocabulary, creating perfectly equivalent classes of subwords. Experimentally, we find that LMs need roughly 17% more data when trained in a fully duplicated setting. Second, we investigate the impact of naturally occurring near duplicates on LMs. Here, we see that merging them considerably hurts LM performance. Therefore, although subword duplication negatively impacts LM training efficiency, naturally occurring near duplicates may not be as similar as anticipated, limiting the potential for performance improvements."
                    },
                    {
                        "title": "Block-Recurrent Transformers",
                        "abstract": "We introduce the Block-Recurrent Transformer, which applies a transformer layer in a recurrent fashion along a sequence, and has linear complexity with respect to sequence length. Our recurrent cell operates on blocks of tokens rather than single tokens during training, and leverages parallel computation within a block in order to make efficient use of accelerator hardware. The cell itself is strikingly simple. It is merely a transformer layer: it uses self-attention and cross-attention to efficiently compute a recurrent function over a large set of state vectors and tokens. Our design was inspired in part by LSTM cells, and it uses LSTM-style gates, but it scales the typical LSTM cell up by several orders of magnitude. Our implementation of recurrence has the same cost in both computation time and parameter count as a conventional transformer layer, but offers dramatically improved perplexity in language modeling tasks over very long sequences. Our model out-performs a long-range Transformer XL baseline by a wide margin, while running twice as fast. We demonstrate its effectiveness on PG19 (books), arXiv papers, and GitHub source code. Our code has been released as open source."
                    },
                    {
                        "title": "The Role of Language Imbalance in Cross-lingual Generalisation: Insights from Cloned Language Experiments",
                        "abstract": "Multilinguality is crucial for extending recent advancements in language modelling to diverse linguistic communities. To maintain high performance while representing multiple languages, multilingual models ideally align representations, allowing what is learned in one language to generalise to others. Prior research has emphasised the importance of parallel data and shared vocabulary elements as key factors for such alignment. In this study, we investigate an unintuitive novel driver of cross-lingual generalisation: language imbalance. In controlled experiments on perfectly equivalent cloned languages, we observe that the existence of a predominant language during training boosts the performance of less frequent languages and leads to stronger alignment of model representations across languages. Furthermore, we find that this trend is amplified with scale: with large enough models or long enough training, we observe that bilingual training data with a 90/10 language split yields better performance on both languages than a balanced 50/50 split. Building on these insights, we design training schemes that can improve performance in all cloned languages, even without altering the training data. As we extend our analysis to real languages, we find that infrequent languages still benefit from frequent ones, yet whether language imbalance causes cross-lingual generalisation there is not conclusive."
                    },
                    {
                        "title": "Enhancing the Transformer with Explicit Relational Encoding for Math Problem Solving",
                        "abstract": "We incorporate Tensor-Product Representations within the Transformer in order to better support the explicit representation of relation structure. Our Tensor-Product Transformer (TP-Transformer) sets a new state of the art on the recently-introduced Mathematics Dataset containing 56 categories of free-form math word-problems. The essential component of the model is a novel attention mechanism, called TP-Attention, which explicitly encodes the relations between each Transformer cell and the other cells from which values have been retrieved by attention. TP-Attention goes beyond linear combination of retrieved values, strengthening representation-building and resolving ambiguities introduced by multiple layers of standard attention. The TP-Transformer's attention maps give better insights into how it is capable of solving the Mathematics Dataset's challenging problems. Pretrained models and code will be made available after publication."
                    },
                    {
                        "title": "Large Language Model Programs",
                        "abstract": "In recent years, large pre-trained language models (LLMs) have demonstrated the ability to follow instructions and perform novel tasks from a few examples. The possibility to parameterise an LLM through such in-context examples widens their capability at a much lower cost than finetuning. We extend this line of reasoning and present a method which further expands the capabilities of an LLM by embedding it within an algorithm or program. To demonstrate the benefits of this approach, we present an illustrative example of evidence-supported question-answering. We obtain a 6.4\\% improvement over the chain of thought baseline through a more algorithmic approach without any finetuning. Furthermore, we highlight recent work from this perspective and discuss the advantages and disadvantages in comparison to the standard approaches."
                    },
                    {
                        "title": "The Languini Kitchen: Enabling Language Modelling Research at Different Scales of Compute",
                        "abstract": "The Languini Kitchen serves as both a research collective and codebase designed to empower researchers with limited computational resources to contribute meaningfully to the field of language modelling. We introduce an experimental protocol that enables model comparisons based on equivalent compute, measured in accelerator hours. The number of tokens on which a model is trained is defined by the model's throughput and the chosen compute class. Notably, this approach avoids constraints on critical hyperparameters which affect total parameters or floating-point operations. For evaluation, we pre-process an existing large, diverse, and high-quality dataset of books that surpasses existing academic benchmarks in quality, diversity, and document length. On it, we compare methods based on their empirical scaling trends which are estimated through experiments at various levels of compute. This work also provides two baseline models: a feed-forward model derived from the GPT-2 architecture and a recurrent model in the form of a novel LSTM with ten-fold throughput. While the GPT baseline achieves better perplexity throughout all our levels of compute, our LSTM baseline exhibits a predictable and more favourable scaling law. This is due to the improved throughput and the need for fewer training tokens to achieve the same decrease in test perplexity. Extrapolating the scaling laws leads of both models results in an intersection at roughly 50,000 accelerator hours. We hope this work can serve as the foundation for meaningful and reproducible language modelling research."
                    },
                    {
                        "title": "Solving Quantitative Reasoning Problems with Language Models",
                        "abstract": "Language models have achieved remarkable performance on a wide range of tasks that require natural language understanding. Nevertheless, state-of-the-art models have generally struggled with tasks that require quantitative reasoning, such as solving mathematics, science, and engineering problems at the college level. To help close this gap, we introduce Minerva, a large language model pretrained on general natural language data and further trained on technical content. The model achieves state-of-the-art performance on technical benchmarks without the use of external tools. We also evaluate our model on over two hundred undergraduate-level problems in physics, biology, chemistry, economics, and other sciences that require quantitative reasoning, and find that the model can correctly answer nearly a third of them."
                    }
                ]
            },
            "706fd4aa-3fed-426a-8a5f-001d0b5e068a": {
                "pk": "706fd4aa-3fed-426a-8a5f-001d0b5e068a",
                "name": "Thomas Hofmann",
                "collaborators": [
                    "Yannic Kilcher",
                    "Florian Schmidt",
                    "Aurelien Lucchi",
                    "Franz J. Giessibl",
                    "Joachim Welker",
                    "Florian Pielmeier",
                    "Michael Keyl",
                    "Bobby He",
                    "Leonard Adolphs",
                    "Yasemin Altun"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Deep Learning",
                    "Adversarial Robustness"
                ],
                "publications": [
                    {
                        "title": "Probabilistic Latent Semantic Analysis",
                        "abstract": "Probabilistic Latent Semantic Analysis is a novel statistical technique for the analysis of two-mode and co-occurrence data, which has applications in information retrieval and filtering, natural language processing, machine learning from text, and in related areas. Compared to standard Latent Semantic Analysis which stems from linear algebra and performs a Singular Value Decomposition of co-occurrence tables, the proposed method is based on a mixture decomposition derived from a latent class model. This results in a more principled approach which has a solid foundation in statistics. In order to avoid overfitting, we propose a widely applicable generalization of maximum likelihood model fitting by tempered EM. Our approach yields substantial and consistent improvements over Latent Semantic Analysis in a number of experiments."
                    },
                    {
                        "title": "Preparation of light-atom tips for Scanning Probe Microscopy by explosive delamination",
                        "abstract": "To obtain maximal resolution in STM and AFM, the size of the protruding tip orbital has to be minimized. Beryllium as tip material is a promising candidate for enhanced resolution because a beryllium atom has just four electrons, leading to a small covalent radius of only 96 pm. Besides that, beryllium is conductive and has a high elastic modulus, which is a necessity for a stable tip apex. However beryllium tips that are prepared ex situ, are covered with a robust oxide layer, which cannot be removed by just heating the tip. Here we present a successful preparation method that combines the heating of the tip by field emission and a mild collision with a clean metal plate. That method yields a clean, oxide-free tip surface as proven by a work function of as deduced from a current-distance curve. Additionally, a STM image of the Si-(111)-(7x7) is presented to prove the single-atom termination of the beryllium tip."
                    },
                    {
                        "title": "Chemical and Crystallographic Characterization of the Tip Apex in Scanning Probe Microscopy",
                        "abstract": "The apex atom of a W scanning probe tip reveals a non-spherical charge distribution as probed by a CO molecule bonded to a Cu(111) surface [Welker et al. Science, 336, 444 (2012)]. Three high-symmetry images were observed and related to three low-index crystallographic directions of the W bcc crystal. Open questions remained, however, including the verification that the tip was indeed W-terminated, and whether this method can be easily applied to distinguish other atomic species. In this work, we investigate bulk Cu and Fe tips. In both cases we can associate our data with the fcc (Cu) and bcc (Fe) crystal structures. A model is presented, based on the partial filling of d orbitals, to relate the AFM images to the angular orientation of the tip structure."
                    },
                    {
                        "title": "The best defense is a good offense: Countering black box attacks by predicting slightly wrong labels",
                        "abstract": "Black-Box attacks on machine learning models occur when an attacker, despite having no access to the inner workings of a model, can successfully craft an attack by means of model theft. The attacker will train an own substitute model that mimics the model to be attacked. The substitute can then be used to design attacks against the original model, for example by means of adversarial samples. We put ourselves in the shoes of the defender and present a method that can successfully avoid model theft by mounting a counter-attack. Specifically, to any incoming query, we slightly perturb our output label distribution in a way that makes substitute training infeasible. We demonstrate that the perturbation does not affect the ordinary use of our model, but results in an effective defense against attacks based on model theft."
                    },
                    {
                        "title": "Controlling a d-level atom in a cavity",
                        "abstract": "In this paper we study controllability of a $d$-level atom interacting with the electromagnetic field in a cavity. The system is modelled by an ordered graph $\\Gamma$. The vertices of $\\Gamma$ describe the energy levels and the edges allowed transitions. To each edge of $\\Gamma$ we associate a harmonic oscillator representing one mode of the electromagnetic field. The dynamics of the system (drift) is given by a natural generalization of the Jaynes-Cummings Hamiltonian. If we add in addition sufficient control over the atom, the overall system (atom and em-field) becomes strongly controllable, i.e. each unitary on the system Hilbert space can be approximated with arbitrary precision in the strong topology by control unitaries. A key role in the proof is played by a topological *-algebra which is (roughly speaking) a representation of the path algebra of $\\Gamma$. It contains crucial structural information about the control problem, and is therefore an important tool for the implementation of control tasks like preparing a particular state from the ground state. This is demonstrated by a detailed discussion of different versions of three-level systems."
                    },
                    {
                        "title": "Deep State Space Models for Unconditional Word Generation",
                        "abstract": "Autoregressive feedback is considered a necessity for successful unconditional text generation using stochastic sequence models. However, such feedback is known to introduce systematic biases into the training process and it obscures a principle of generation: committing to global information and forgetting local nuances. We show that a non-autoregressive deep state space model with a clear separation of global and local uncertainty can be built from only two ingredients: An independent noise source and a deterministic transition function. Recent advances on flow-based variational inference can be used to train an evidence lower-bound without resorting to annealing, auxiliary losses or similar measures. The result is a highly interpretable generative model on par with comparable auto-regressive models on the task of word generation."
                    },
                    {
                        "title": "BERT as a Teacher: Contextual Embeddings for Sequence-Level Reward",
                        "abstract": "Measuring the quality of a generated sequence against a set of references is a central problem in many learning frameworks, be it to compute a score, to assign a reward, or to perform discrimination. Despite great advances in model architectures, metrics that scale independently of the number of references are still based on n-gram estimates. We show that the underlying operations, counting words and comparing counts, can be lifted to embedding words and comparing embeddings. An in-depth analysis of BERT embeddings shows empirically that contextual embeddings can be employed to capture the required dependencies while maintaining the necessary scalability through appropriate pruning and smoothing techniques. We cast unconditional generation as a reinforcement learning problem and show that our reward function indeed provides a more effective learning signal than n-gram reward in this challenging setting."
                    },
                    {
                        "title": "Simplifying Transformer Blocks",
                        "abstract": "A simple design recipe for deep Transformers is to compose identical building blocks. But standard transformer blocks are far from simple, interweaving attention and MLP sub-blocks with skip connections & normalisation layers in precise arrangements. This complexity leads to brittle architectures, where seemingly minor changes can significantly reduce training speed, or render models untrainable.   In this work, we ask to what extent the standard transformer block can be simplified? Combining signal propagation theory and empirical observations, we motivate modifications that allow many block components to be removed with no loss of training speed, including skip connections, projection or value parameters, sequential sub-blocks and normalisation layers. In experiments on both autoregressive decoder-only and BERT encoder-only models, our simplified transformers emulate the per-update training speed and performance of standard transformers, while enjoying 15% faster training throughput, and using 15% fewer parameters."
                    },
                    {
                        "title": "LeDeepChef: Deep Reinforcement Learning Agent for Families of Text-Based Games",
                        "abstract": "While Reinforcement Learning (RL) approaches lead to significant achievements in a variety of areas in recent history, natural language tasks remained mostly unaffected, due to the compositional and combinatorial nature that makes them notoriously hard to optimize. With the emerging field of Text-Based Games (TBGs), researchers try to bridge this gap. Inspired by the success of RL algorithms on Atari games, the idea is to develop new methods in a restricted game world and then gradually move to more complex environments. Previous work in the area of TBGs has mainly focused on solving individual games. We, however, consider the task of designing an agent that not just succeeds in a single game, but performs well across a whole family of games, sharing the same theme. In this work, we present our deep RL agent--LeDeepChef--that shows generalization capabilities to never-before-seen games of the same family with different environments and task descriptions. The agent participated in Microsoft Research's \"First TextWorld Problems: A Language and Reinforcement Learning Challenge\" and outperformed all but one competitor on the final test set. The games from the challenge all share the same theme, namely cooking in a modern house environment, but differ significantly in the arrangement of the rooms, the presented objects, and the specific goal (recipe to cook). To build an agent that achieves high scores across a whole family of games, we use an actor-critic framework and prune the action-space by using ideas from hierarchical reinforcement learning and a specialized module trained on a recipe database."
                    },
                    {
                        "title": "Exponential Families for Conditional Random Fields",
                        "abstract": "In this paper we de ne conditional random elds in reproducing kernel Hilbert spaces and show connections to Gaussian Process classi cation. More speci cally, we prove decomposition results for undirected graphical models and we give constructions for kernels. Finally we present e cient means of solving the optimization problem using reduced rank decompositions and we show how stationarity can be exploited e ciently in the optimization process."
                    },
                    {
                        "title": "Starting Small -- Learning with Adaptive Sample Sizes",
                        "abstract": "For many machine learning problems, data is abundant and it may be prohibitive to make multiple passes through the full training set. In this context, we investigate strategies for dynamically increasing the effective sample size, when using iterative methods such as stochastic gradient descent. Our interest is motivated by the rise of variance-reduced methods, which achieve linear convergence rates that scale favorably for smaller sample sizes. Exploiting this feature, we show -- theoretically and empirically -- how to obtain significant speed-ups with a novel algorithm that reaches statistical accuracy on an $n$-sample in $2n$, instead of $n \\log n$ steps."
                    },
                    {
                        "title": "Deep Joint Entity Disambiguation with Local Neural Attention",
                        "abstract": "We propose a novel deep learning model for joint document-level entity disambiguation, which leverages learned neural representations. Key components are entity embeddings, a neural attention mechanism over local context windows, and a differentiable joint inference stage for disambiguation. Our approach thereby combines benefits of deep learning with more traditional approaches such as graphical models and probabilistic mention-entity maps. Extensive experiments show that we are able to obtain competitive or state-of-the-art accuracy at moderate computational costs."
                    },
                    {
                        "title": "Generator Reversal",
                        "abstract": "We consider the problem of training generative models with deep neural networks as generators, i.e. to map latent codes to data points. Whereas the dominant paradigm combines simple priors over codes with complex deterministic models, we propose instead to use more flexible code distributions. These distributions are estimated non-parametrically by reversing the generator map during training. The benefits include: more powerful generative models, better modeling of latent structure and explicit control of the degree of generalization."
                    },
                    {
                        "title": "Flexible Prior Distributions for Deep Generative Models",
                        "abstract": "We consider the problem of training generative models with deep neural networks as generators, i.e. to map latent codes to data points. Whereas the dominant paradigm combines simple priors over codes with complex deterministic models, we argue that it might be advantageous to use more flexible code distributions. We demonstrate how these distributions can be induced directly from the data. The benefits include: more powerful generative models, better modeling of latent structure and explicit control of the degree of generalization."
                    },
                    {
                        "title": "Adversarially Robust Training through Structured Gradient Regularization",
                        "abstract": "We propose a novel data-dependent structured gradient regularizer to increase the robustness of neural networks vis-a-vis adversarial perturbations. Our regularizer can be derived as a controlled approximation from first principles, leveraging the fundamental link between training with noise and regularization. It adds very little computational overhead during learning and is simple to implement generically in standard deep learning frameworks. Our experiments provide strong evidence that structured gradient regularization can act as an effective first line of defense against attacks based on low-level signal corruption."
                    },
                    {
                        "title": "Autoregressive Text Generation Beyond Feedback Loops",
                        "abstract": "Autoregressive state transitions, where predictions are conditioned on past predictions, are the predominant choice for both deterministic and stochastic sequential models. However, autoregressive feedback exposes the evolution of the hidden state trajectory to potential biases from well-known train-test discrepancies. In this paper, we combine a latent state space model with a CRF observation model. We argue that such autoregressive observation models form an interesting middle ground that expresses local correlations on the word level but keeps the state evolution non-autoregressive. On unconditional sentence generation we show performance improvements compared to RNN and GAN baselines while avoiding some prototypical failure modes of autoregressive models."
                    },
                    {
                        "title": "SwissDial: Parallel Multidialectal Corpus of Spoken Swiss German",
                        "abstract": "Swiss German is a dialect continuum whose natively acquired dialects significantly differ from the formal variety of the language. These dialects are mostly used for verbal communication and do not have standard orthography. This has led to a lack of annotated datasets, rendering the use of many NLP methods infeasible. In this paper, we introduce the first annotated parallel corpus of spoken Swiss German across 8 major dialects, plus a Standard German reference. Our goal has been to create and to make available a basic dataset for employing data-driven NLP applications in Swiss German. We present our data collection procedure in detail and validate the quality of our corpus by conducting experiments with the recent neural models for speech synthesis."
                    },
                    {
                        "title": "How Good is a Single Basin?",
                        "abstract": "The multi-modal nature of neural loss landscapes is often considered to be the main driver behind the empirical success of deep ensembles. In this work, we probe this belief by constructing various \"connected\" ensembles which are restricted to lie in the same basin. Through our experiments, we demonstrate that increased connectivity indeed negatively impacts performance. However, when incorporating the knowledge from other basins implicitly through distillation, we show that the gap in performance can be mitigated by re-discovering (multi-basin) deep ensembles within a single basin. Thus, we conjecture that while the extra-basin knowledge is at least partially present in any given basin, it cannot be easily harnessed without learning it from other basins."
                    }
                ]
            }
        }
    },
    "2405.19073": {
        "paper_data": {
            "title": "An engine not a camera: Measuring performative power of online search",
            "url": "http://arxiv.org/abs/2405.19073v1",
            "arxiv_id": "2405.19073",
            "authors": [
                "Celestine Mendler-D\u00fcnner",
                "Gabriele Carovano",
                "Moritz Hardt"
            ],
            "abstract": "The power of digital platforms is at the center of major ongoing policy and regulatory efforts. To advance existing debates, we designed and executed an experiment to measure the power of online search providers, building on the recent definition of performative power. Instantiated in our setting, performative power quantifies the ability of a search engine to steer web traffic by rearranging results. To operationalize this definition we developed a browser extension that performs unassuming randomized experiments in the background. These randomized experiments emulate updates to the search algorithm and identify the causal effect of different content arrangements on clicks. We formally relate these causal effects to performative power. Analyzing tens of thousands of clicks, we discuss what our robust quantitative findings say about the power of online search engines. More broadly, we envision our work to serve as a blueprint for how performative power and online experiments can be integrated with future investigations into the economic power of digital platforms.",
            "introduction": "   1 Introduction  At the heart of one of Europe\u2019s most prominent antitrust case is a seemingly mundane question: How much can a search engine redirect traffic through content positioning? In 2017, the European Commission alleged that Google favored its own comparison shopping service by steering clicks away from search results towards Google\u2019s own product comparison service. The technical centerpiece of the case was an ad-hoc data analysis about the position and display biases of Google search results. Google appealed the European Commission\u2019s charges, pointing to, among other arguments, methodological errors.222Case T-612/17, Google and Alphabet v Commission (Google Shopping), ECLI:EU:T:2021:763. The case is still pending, see Case C-48/22 P, Google and Alphabet v Commission (Google Shopping).   The case is emblematic of a broader problem. Although urgently needed, there is currently no accepted technical framework for answering basic questions about the economic power of digital platforms. Lawyers, economists, and policy makers agree that traditional antitrust tools struggle with multi-sided platforms\u00a0(Stigler Committee, 2019; Cr\u00e9mer et\u00a0al., 2019). Among other issues, these tools are not able to capture behavioral aspects defining user-platform interactions. Against this backdrop, a recently developed concept, called performative power\u00a0(Hardt et\u00a0al., 2022), suggests a way to mitigate limitations of existing antitrust enforcement tools. Performative power measures how much a platform can causally influence platform users through its algorithmic actions. By directly relating power to a causal effect, it sidesteps the complexities underlying conventional market definitions and offers a promising framework to integrate data and experimental methods with digital market investigations. Although the definition of performative power enjoys appealing theoretical properties, a proof of its practical applicability was still missing.   Figure 1: The ability to influence web traffic through content arrangement. Blue bars show average click probability observed for generic search results in position 1 to 6 on Google search under different counterfactual arrangements; default arrangement (left), swapping results 1 and 2 (middle), swapping results 1 and 3 (right). We see a significant effect of these arrangement changes on clicks. We will inspect this more formally in Section\u00a05 where we also explore arrangement changes beyond reranking.    1.1 Our work  In this work, we present a first proof of concept showing how to estimate performative power of a search engine in practice. The instantiation of performative power we are interested in concerns the ability of a search engine to impact web traffic through decisions of how to arrange content.   Our core contribution is to design and implement an online experiment to provide quantitative insights into the causal effect of content arrangement on clicks for the two most widely used search engines, Google Search and Bing. Our experiment is based on a browser extension, called Powermeter, that emulates small updates to the platform\u2019s algorithm by modifying how search results are displayed to users. We implement different counterfactual arrangements to inspect the effect of re-ranking and favored positioning (e.g., Ads or Shopping boxes) on clicks, both in isolation and jointly; the latter is relevant for the Google Shopping investigation. The arrangement to which a user is exposed is chosen at random every time they perform a search. We discuss several technical steps we implemented to",
            "references": [
                {
                    "title": "Antitrust, Amazon, and Algorithmic Auditing",
                    "abstract": "In digital markets, antitrust law and special regulations aim to ensure that markets remain competitive despite the dominating role that digital platforms play today in everyone's life. Unlike traditional markets, market participant behavior is easily observable in these markets. We present a series of empirical investigations into the extent to which Amazon engages in practices that are typically described as self-preferencing. We discuss how the computer science tools used in this paper can be used in a regulatory environment that is based on algorithmic auditing and requires regulating digital markets at scale."
                },
                {
                    "title": "Performative Prediction: Past and Future",
                    "abstract": "Predictions in the social world generally influence the target of prediction, a phenomenon known as performativity. Self-fulfilling and self-negating predictions are examples of performativity. Of fundamental importance to economics, finance, and the social sciences, the notion has been absent from the development of machine learning. In machine learning applications, performativity often surfaces as distribution shift. A predictive model deployed on a digital platform, for example, influences consumption and thereby changes the data-generating distribution. We survey the recently founded area of performative prediction that provides a definition and conceptual framework to study performativity in machine learning. A consequence of performative prediction is a natural equilibrium notion that gives rise to new optimization challenges. Another consequence is a distinction between learning and steering, two mechanisms at play in performative prediction. The notion of steering is in turn intimately related to questions of power in digital markets. We review the notion of performative power that gives an answer to the question how much a platform can steer participants through its predictions. We end on a discussion of future directions, such as the role that performativity plays in contesting algorithmic systems."
                },
                {
                    "title": "Google the Gatekeeper: How Search Components Affect Clicks and Attention",
                    "abstract": "The contemporary Google Search Engine Results Page (SERP) supplements classic blue hyperlinks with complex components. These components produce tensions between searchers, 3rd-party websites, and Google itself over clicks and attention. In this study, we examine 12 SERP components from two categories: (1) extracted results (e.g., featured-snippets) and (2) Google Services (e.g., shopping-ads) to determine their effect on peoples\u2019 behavior. We measure behavior with two variables: (1) click- through rate (CTR) to Google\u2019s own domains versus 3rd-party domains and (2) time spent on the SERP. We apply causal inference methods to an ecologically valid trace dataset comprising 477,485 SERPs from 1,756 participants. We find that multiple components substantially increase CTR to Google domains, while others decrease CTR and increase time on the SERP. These findings may inform efforts to regulate the design of powerful intermediary platforms like Google."
                },
                {
                    "title": "Beyond Rankings: Exploring the Impact of SERP Features on Organic Click-through Rates",
                    "abstract": "Search Engine Result Pages (SERPs) serve as the digital gateways to the vast expanse of the internet. Past decades have witnessed a surge in research primarily centered on the influence of website ranking on these pages, to determine the click-through rate (CTR). However, during this period, the landscape of SERPs has undergone a dramatic evolution: SERP features, encompassing elements such as knowledge panels, media galleries, FAQs, and more, have emerged as an increasingly prominent facet of these result pages. Our study examines the crucial role of these features, revealing them to be not merely aesthetic components, but strongly influence CTR and the associated behavior of internet users. We demonstrate how these features can significantly modulate web traffic, either amplifying or attenuating it. We dissect these intricate interaction effects leveraging a unique dataset of 67,000 keywords and their respective Google SERPs, spanning over 40 distinct US-based e-commerce domains, generating over 6 million clicks from 24 million views. This cross-website dataset, unprecedented in its scope, enables us to assess the impact of 24 different SERP features on organic CTR. Through an ablation study modeling CTR, we illustrate the incremental predictive power these features hold."
                },
                {
                    "title": "The evolving concept of market power in the digital economy",
                    "abstract": "Assessing market power is central to the CMA\u2019s cases, including cases involving digital markets. Over the past few years, it has adapted its approach to deal effectively with digital markets and revised its guidance to reflect that approach. In dealing with digital cases, the CMA has not found that a new notion of market power was needed specifically for the digital space or for it to be able to capture its concerns in digital markets. However, it has revised its guidance to reflect how it conducts its assessment, including placing less emphasis on a formalistic market definition exercise and more emphasis on dynamic constraints, considering innovation as the current process of rivalry through investment decisions and incorporated the threat of entry when determining current competitive constraints. It has also considered characteristics like multi-sidedness, network effects and any resulting tipping dynamics in performing its assessments. The CMA\u2019s current tools sometimes struggle to deal effectively with digital markets; therefore, the CMA\u2019s Digital Markets Taskforce has recommended the introduction of a new ex ante regulatory regime to proactively shape the behaviour of particularly powerful digital firms and to prevent consumer harm from arising, including a new test to identify what firms should be subject to this new regime which includes, as an integral part, an assessment of market power."
                },
                {
                    "title": "Algorithmic Collective Action in Machine Learning",
                    "abstract": "We initiate a principled study of algorithmic collective action on digital platforms that deploy machine learning algorithms. We propose a simple theoretical model of a collective interacting with a firm's learning algorithm. The collective pools the data of participating individuals and executes an algorithmic strategy by instructing participants how to modify their own data to achieve a collective goal. We investigate the consequences of this model in three fundamental learning-theoretic settings: the case of a nonparametric optimal learning algorithm, a parametric risk minimizer, and gradient-based optimization. In each setting, we come up with coordinated algorithmic strategies and characterize natural success criteria as a function of the collective's size. Complementing our theory, we conduct systematic experiments on a skill classification task involving tens of thousands of resumes from a gig platform for freelancers. Through more than two thousand model training runs of a BERT-like language model, we see a striking correspondence emerge between our empirical observations and the predictions made by our theory. Taken together, our theory and experiments broadly support the conclusion that algorithmic collectives of exceedingly small fractional size can exert significant control over a platform's learning algorithm."
                },
                {
                    "title": "Towards Disentangling Relevance and Bias in Unbiased Learning to Rank",
                    "abstract": "Unbiased learning to rank (ULTR) studies the problem of mitigating various biases from implicit user feedback data such as clicks, and has been receiving considerable attention recently. A popular ULTR approach for real-world applications uses a two-tower architecture, where click modeling is factorized into a relevance tower with regular input features, and a bias tower with bias-relevant inputs such as the position of a document. A successful factorization will allow the relevance tower to be exempt from biases. In this work, we identify a critical issue that existing ULTR methods ignored - the bias tower can be confounded with the relevance tower via the underlying true relevance. In particular, the positions were determined by the logging policy, i.e., the previous production model, which would possess relevance information. We give both theoretical analysis and empirical results to show the negative effects on relevance tower due to such a correlation. We then propose two methods to mitigate the negative confounding effects by better disentangling relevance and bias. Offline empirical results on both controlled public datasets and a large-scale industry dataset show the effectiveness of the proposed approaches. We conduct a live experiment on a popular web store for four weeks, and find a significant improvement in user clicks over the baseline, which ignores the negative confounding effect."
                },
                {
                    "title": "Reactivity in measuring sensitive online behavior",
                    "abstract": "PurposeDigital trace data provide new opportunities to study how individuals act and interact with others online. One advantage of this type of data is that it measures behavior in a less obtrusive way than surveys, potentially reducing measurement error. However, it is well documented that in observational studies, participants' awareness of being observed can change their behavior, especially when the behavior is considered sensitive. Very little is known regarding this effect in the online realm. Against this background, we studied whether people change their online behavior because digital trace data are being collected.Design/methodology/approachWe analyzed data from a sample of 1,959 members of a German online panel who had consented to the collection of digital trace data about their online browsing and/or mobile app usage. To identify reactivity, we studied change over time in five types of sensitive online behavior.FindingsWe found that the frequency and duration with which individuals engage in sensitive behaviors online gradually increases during the first couple of days after the installation of a tracker, mainly individuals who extensively engage in sensitive behavior show this pattern of increase after installation and this change in behavior is limited to certain types of sensitive online behavior.Originality/valueThere is an increased interest in the use of digital trace data in the social sciences and our study is one of the first methodological contributions measuring reactivity in digital trace data measurement."
                },
                {
                    "title": "Modeling Content Creator Incentives on Algorithm-Curated Platforms",
                    "abstract": "Content creators compete for user attention. Their reach crucially depends on algorithmic choices made by developers on online platforms. To maximize exposure, many creators adapt strategically, as evidenced by examples like the sprawling search engine optimization industry. This begets competition for the finite user attention pool. We formalize these dynamics in what we call an exposure game, a model of incentives induced by algorithms, including modern factorization and (deep) two-tower architectures. We prove that seemingly innocuous algorithmic choices, e.g., non-negative vs. unconstrained factorization, significantly affect the existence and character of (Nash) equilibria in exposure games. We proffer use of creator behavior models, like exposure games, for an (ex-ante) pre-deployment audit. Such an audit can identify misalignment between desirable and incentivized content, and thus complement post-hoc measures like content filtering and moderation. To this end, we propose tools for numerically finding equilibria in exposure games, and illustrate results of an audit on the MovieLens and LastFM datasets. Among else, we find that the strategically produced content exhibits strong dependence between algorithmic exploration and content diversity, and between model expressivity and bias towards gender-based user and creator groups."
                },
                {
                    "title": "Scalar is Not Enough: Vectorization-based Unbiased Learning to Rank",
                    "abstract": "Unbiased learning to rank (ULTR) aims to train an unbiased ranking model from biased user click logs. Most of the current ULTR methods are based on the examination hypothesis (EH), which assumes that the click probability can be factorized into two scalar functions, one related to ranking features and the other related to bias factors. Unfortunately, the interactions among features, bias factors and clicks are complicated in practice, and usually cannot be factorized in this independent way. Fitting click data with EH could lead to model misspecification and bring the approximation error. In this paper, we propose a vector-based EH and formulate the click probability as a dot product of two vector functions. This solution is complete due to its universality in fitting arbitrary click functions. Based on it, we propose a novel model named Vectorization to adaptively learn the relevance embeddings and sort documents by projecting embeddings onto a base vector. Extensive experiments show that our method significantly outperforms the state-of-the-art ULTR methods on complex real clicks as well as simple simulated clicks."
                },
                {
                    "title": "Visual Encoding and Debiasing for CTR Prediction",
                    "abstract": "Extracting expressive visual features is crucial for accurate Click-Through-Rate (CTR) prediction in visual search advertising systems. Current commercial systems use off-the-shelf visual encoders to facilitate fast online service. However, the extracted visual features are coarse-grained and/or biased. In this paper, we present a visual encoding framework for CTR prediction to overcome these problems. The framework is based on contrastive learning which pulls positive pairs closer and pushes negative pairs apart in the visual feature space. To obtain fine-grained visual features, we present contrastive learning supervised by click-through data to fine-tune the visual encoder. To reduce sample selection bias, firstly we train the visual encoder offline by leveraging both unbiased self-supervision and click supervision signals. Secondly, we incorporate a debiasing network in the online CTR predictor to adjust the visual features by contrasting high impression items with selected, low impression items. We deploy the framework in a mobile E-commerce app. Offline experiments on billion-scale datasets and online experiments demonstrate that the proposed framework can make accurate and unbiased predictions."
                },
                {
                    "title": "Performative Power",
                    "abstract": "We introduce the notion of performative power, which measures the ability of a firm operating an algorithmic system, such as a digital content recommendation platform, to cause change in a population of participants. We relate performative power to the economic study of competition in digital economies. Traditional economic concepts struggle with identifying anti-competitive patterns in digital platforms not least due to the complexity of market definition. In contrast, performative power is a causal notion that is identifiable with minimal knowledge of the market, its internals, participants, products, or prices. Low performative power implies that a firm can do no better than to optimize their objective on current data. In contrast, firms of high performative power stand to benefit from steering the population towards more profitable behavior. We confirm in a simple theoretical model that monopolies maximize performative power. A firm's ability to personalize increases performative power, while competition and outside options decrease performative power. On the empirical side, we propose an observational causal design to identify performative power from discontinuities in how digital platforms display content. This allows to repurpose causal effects from various studies about digital platforms as lower bounds on performative power. Finally, we speculate about the role that performative power might play in competition policy and antitrust enforcement in digital marketplaces."
                },
                {
                    "title": "Bias and Debias in Recommender System: A Survey and Future Directions",
                    "abstract": "While recent years have witnessed a rapid growth of research papers on recommender system (RS), most of the papers focus on inventing machine learning models to better fit user behavior data. However, user behavior data is observational rather than experimental. This makes various biases widely exist in the data, including but not limited to selection bias, position bias, exposure bias, and popularity bias. Blindly fitting the data without considering the inherent biases will result in many serious issues, e.g., the discrepancy between offline evaluation and online metrics, hurting user satisfaction and trust on the recommendation service, and so on. To transform the large volume of research models into practical improvements, it is highly urgent to explore the impacts of the biases and perform debiasing when necessary. When reviewing the papers that consider biases in RS, we find that, to our surprise, the studies are rather fragmented and lack a systematic organization. The terminology \u201cbias\u201d is widely used in the literature, but its definition is usually vague and even inconsistent across papers. This motivates us to provide a systematic survey of existing work on RS biases. In this paper, we first summarize seven types of biases in recommendation, along with their definitions and characteristics. We then provide a taxonomy to position and organize the existing work on recommendation debiasing. Finally, we identify some open challenges and envision some future directions, with the hope of inspiring more research work on this important yet less investigated topic. The summary of debiasing methods reviewed in this survey can be found at https://github.com/jiawei-chen/RecDebiasing."
                },
                {
                    "title": "Attribute-based Propensity for Unbiased Learning in Recommender Systems: Algorithm and Case Studies",
                    "abstract": "Many modern recommender systems train their models based on a large amount of implicit user feedback data. Due to the inherent bias in this data (e.g., position bias), learning from it directly can lead to suboptimal models. Recently, unbiased learning was proposed to address such problems by leveraging counterfactual techniques like inverse propensity weighting (IPW). In these methods, propensity scores estimation is usually limited to item's display position in a single user interface (UI). In this paper, we generalize the traditional position bias model to an attribute-based propensity framework. Our methods estimate propensity scores based on offline data and allow propensity estimation across a broad range of implicit feedback scenarios, e.g., feedback beyond recommender system UI. We demonstrate this by applying this framework to three real-world large-scale recommender systems in Google Drive that serve millions of users. For each system, we conduct both offline and online evaluation. Our results show that the proposed framework is able to significantly improve upon strong production baselines across a diverse range of recommendation item types (documents, people-document pairs, and queries), UI layouts (horizontal, vertical, and grid layouts), and underlying learning algorithms (gradient boosted decision trees and neural networks), all without the need to intervene and degrade the user experience. The proposed models have been deployed in the production systems with ease since no serving infrastructure change is needed."
                },
                {
                    "title": "Performative Prediction",
                    "abstract": "When predictions support decisions they may influence the outcome they aim to predict. We call such predictions performative; the prediction influences the target. Performativity is a well-studied phenomenon in policy-making that has so far been neglected in supervised learning. When ignored, performativity surfaces as undesirable distribution shift, routinely addressed with retraining. We develop a risk minimization framework for performative prediction bringing together concepts from statistics, game theory, and causality. A conceptual novelty is an equilibrium notion we call performative stability. Performative stability implies that the predictions are calibrated not against past outcomes, but against the future outcomes that manifest from acting on the prediction. Our main results are necessary and sufficient conditions for the convergence of retraining to a performatively stable point of nearly minimal loss. In full generality, performative prediction strictly subsumes the setting known as strategic classification. We thus also give the first sufficient conditions for retraining to overcome strategic feedback effects."
                },
                {
                    "title": "Macroeconomics and Market Power: Context, Implications, and Open Questions",
                    "abstract": "This article assesses several aspects of recent macroeconomic market power research. These include the ways market power is defined and measured; the use of accounting data to estimate markups; the quantitative implications of theoretical connections among markups, prices, costs, scale elasticities, and profits; and conflicting evidence on whether greater market power has led to lower investment rates and a lower labor share of income. Throughout this discussion, I characterize the congruencies and incongruencies between macro evidence and micro views of market power and, when they do not perfectly overlap, explain the open questions that need to be answered to make the connection complete."
                },
                {
                    "title": "The Power of Rankings: Quantifying the Effect of Rankings on Online Consumer Search and Purchase Decisions",
                    "abstract": "Online search intermediaries, such as Amazon or Expedia, use rankings (ordered lists) to present third party sellers\u2019 products to consumers. These rankings decrease consumer search costs and increase the probability of a match with a seller, ultimately increasing consumer welfare. Constructing relevant rankings requires understanding their causal effect on consumer choices. However, this is challenging since rankings are endogenous: highly ranked products are also the most relevant ones for consumers. In this paper, I use the first data set with experimental variation in the ranking from a field experiment at Expedia to identify the causal effect of rankings. Using this data set, I make three contributions. First, I show that rankings affect what consumers search, but conditional on search, do not affect purchases. I also exploit a feature of the data set (opaque offers), to show that rankings lower search costs, instead of affecting consumer expectations or utility. Second, I quantify the effect of rankings using a sequential search model and find an average position effect of $2.64, lower than previous estimates in the literature obtained without experimental variation. Finally, I show that a utility-based ranking built on this model\u2019s estimates leads to an almost twofold increase in consumer welfare, while also increasing transactions and revenue for the search intermediary."
                },
                {
                    "title": "A Game-Theoretic Approach to Recommendation Systems with Strategic Content Providers",
                    "abstract": "We introduce a game-theoretic approach to the study of recommendation systems with strategic content providers. Such systems should be fair and stable. Showing that traditional approaches fail to satisfy these requirements, we propose the Shapley mediator. We show that the Shapley mediator fulfills the fairness and stability requirements, runs in linear time, and is the only economically efficient mechanism satisfying these properties."
                },
                {
                    "title": "A General Framework for Counterfactual Learning-to-Rank",
                    "abstract": "Implicit feedback (e.g., click, dwell time) is an attractive source of training data for Learning-to-Rank, but its naive use leads to learning results that are distorted by presentation bias. For the special case of optimizing average rank for linear ranking functions, however, the recently developed SVM-PropRank method has shown that counterfactual inference techniques can be used to provably overcome the distorting effect of presentation bias. Going beyond this special case, this paper provides a general and theoretically rigorous framework for counterfactual learning-to-rank that enables unbiased training for a broad class of additive ranking metrics (e.g., Discounted Cumulative Gain (DCG)) as well as a broad class of models (e.g., deep networks). Specifically, we derive a relaxation for propensity-weighted rank-based metrics which is subdifferentiable and thus suitable for gradient-based optimization. We demonstrate the effectiveness of this general approach by instantiating two new learning methods. One is a new type of unbiased SVM that optimizes DCG - called SVM PropDCG - and we show how the resulting optimization problem can be solved via the Convex Concave Procedure (CCP). The other is Deep PropDCG, where the ranking function can be an arbitrary deep network. In addition to the theoretical support, we empirically find that SVM PropDCG significantly outperforms existing linear rankers in terms of DCG. Moreover, the ability to train non-linear ranking functions via Deep PropDCG further improves performance."
                },
                {
                    "title": "Auditing the Personalization and Composition of Politically-Related Search Engine Results Pages",
                    "abstract": "Search engines are a primary means through which people obtain information in today\u00bbs connected world. Yet, apart from the search engine companies themselves, little is known about how their algorithms filter, rank, and present the web to users. This question is especially pertinent with respect to political queries, given growing concerns about filter bubbles, and the recent finding that bias or favoritism in search rankings can influence voting behavior. In this study, we conduct a targeted algorithm audit of Google Search using a dynamic set of political queries. We designed a Chrome extension to survey participants and collect the Search Engine Results Pages (SERPs) and autocomplete suggestions that they would have been exposed to while searching our set of political queries during the month after Donald Trump\u00bbs Presidential inauguration. Using this data, we found significant differences in the composition and personalization of politically-related SERPs by query type, subjects\u00bb characteristics, and date."
                },
                {
                    "title": "When Should You Adjust Standard Errors for Clustering?",
                    "abstract": "\n Clustered standard errors, with clusters defined by factors such as geography, are widespread in empirical research in economics and many other disciplines. Formally, clustered standard errors adjust for the correlations induced by sampling the outcome variable from a data-generating process with unobserved cluster-level components. However, the standard econometric framework for clustering leaves important questions unanswered: (i) Why do we adjust standard errors for clustering in some ways but not others, e.g., by state but not by gender, and in observational studies, but not in completely randomized experiments? (ii) Why is conventional clustering an \u201call-or-nothing\u201d adjustment, while within-cluster correlations can be strong or extremely weak? (iii) In what settings does the choice of whether and how to cluster make a difference? We address these and other questions using a novel framework for clustered inference on average treatment effects. In addition to the common sampling component, the new framework incorporates a design component that accounts for the variability induced on the estimator by the treatment assignment mechanism. We show that, when the number of clusters in the sample is a nonnegligible fraction of the number of clusters in the population, conventional cluster standard errors can be severely inflated, and propose new variance estimators that correct for this bias."
                },
                {
                    "title": "Visual Positions of Links and Clicks on Wikipedia",
                    "abstract": "In this work, we study the visual position of links and their clicks on Wikipedia, particularly where links are visually located, at which screen positions users click on links, and which areas on the screen exhibit more or less clicks per links. For that purpose, we introduce a novel dataset containing the on-screen coordinate position for all links between pages in the English Wikipedia and additionally resort to navigation logs of Wikipedia users. Using this data, we can observe a preference of certain link and click locations on Wikipedia including first evidence of positional click bias. For example, our results suggest that users have a tendency to prefer to click on the left side of the screen which exceeds what one would expect from the presence of links on pages. We believe that presented data and research can be useful for optimizing the process of link creation and link consumption on Wikipedia and other Web platforms."
                },
                {
                    "title": "Causal Inference for Statistics, Social, and Biomedical Sciences: An Introduction",
                    "abstract": "Guido Imbens and Don Rubin present an insightful discussion of the potential outcomes framework for causal inference. In this framework, the goal is to define the effect of a treatment (including, of course, a definition of \u201ctreatment\u201d itself), to understand when such an object is identifiable in the population (or, at least, in the superpopulation), and to estimate and conduct statistical inference about this quantity of interest as well as related quantities such as welfare measures or other policy-relevant effects. Despite this goal being fundamental, a final answer is far from settled as many competing and/or complementary approaches have been proposed in the literature (e.g., Heckman and Vytlacil 2007; Manski 2008; Pearl 2009; VanderWeele 2015; Hern\u00e1n and Robins 2016). The authors employ the potential outcomes framework to study causal inference in three of the most important settings in applied work: (1) randomized experiments, (2) unconfounded treatment assignment in observational studies (an assumption also known as missing-at-random, conditional independence, selection-on-observables, or ignorability), and (3) experiments with noncompliance (a special case of instrumental variables methods). A key feature of the book is the recurrent use of real empirical examples, and the detailed derivation of the most important calculations needed for implementation, all of which provide much needed guidance on how to go from theory to practice. While these features surely help the reader understand how the different methods work, I suspect the book would be even more effective if the authors provided the data and computer codes used throughout. The authors are also very careful to distinguish between identification, estimation, inference, implementation, and interpretation issues. All these features make the book amust-read for all applied researchers, regardless of their technical background. Parts I and II (Chapters 1\u201311) of the book give a brief introduction to causality and the potential outcomes framework, and then present a thorough discussion of the classical analysis of experiments. The part of the book introduces the reader to a variety of interconnected classical ideas, including Fisher\u2019s exact testing, Neyman\u2019s repeated sampling, regression-based methods and Bayesian model-based methods, which are often encountered in empirical work but treated as fundamentally distinct. Imbens and Rubin elucidated these ideas within a unifiedmethodological framework, while also offering several illustrative empirical applications throughout. Anyone interested in deepening their understanding of the analysis and interpretation of experiments will find these chapters to be an invaluable and timeless reference. Parts III and IV (Chapters 12\u201320) focus on settings in which the treatment assignment is unconfounded, that is, where the treatment assignment is independent of the potential outcomes after conditioning on observable characteristics. This type of conditional independence assumption, which is statistical in nature and widely used in empirical work, is the basis for a variety of estimation and inference procedures in causal inference. In this portion of the book, the authors not only review most of their influential work in the area, but also introduce and discuss various new ideas such as model, tuning parameter, and covariate selection methods. Moreover, Part V (Chapters 21\u201322) presents methods for assessing the plausibility of the underlying key identifying assumptions, as well as approaches based on bounds and other sensitivity analyses that are crucial to understanding the robustness of the results obtained under the selection-on-observables assumption. The final chapters, Part VI (Chapters 23\u201325), focus on the analysis of experiments with imperfect compliance. The focus here is mostly on the authors\u2019 widely influential work on nonparametric identification of the local average treatment effect (LATE), although an alternative model-based approach to estimation and inference is also presented. In sum, this book presents a unified framework to causal inference based on the potential outcomes framework, focusing on the classical analysis of experiments, unconfoundedness, and noncompliance. The book has become an instant classic in the causal inference literature, broadly defined, and will certainly guide future research in this area. All researchers will benefit from carefully studying this book, no matter what their specific views are on the subject matter. Looking ahead, an added benefit of the book is that it lays out the foundations for a systematic and unified statistical analysis of more advanced topics in causal inference such as multivalued, dynamic, and adaptive treatment regimes; differencein-differences methods; regression discontinuity designs; linear and nonlinear panel or longitudinal data; interference, externalities, spillovers and peer effects; and external validity, meta analysis and counterfactual policy evaluations, just to mention a few examples. One can only hope that the authors can be persuaded to write the second volume."
                },
                {
                    "title": "Click Models for Web Search",
                    "abstract": "Abstract With the rapid growth of web search in recent years the problem of modeling its users has started to attract more and more attention of the information retrieval community. This has several motivations. By building a model of user behavior we are essentially developing a better understanding of a user, which ultimately helps us to deliver a better search experience. A model of user behavior can also be used as a predictive device for non-observed items such as document relevance, which makes it useful for improving search result ranking. Finally, in many situations experimenting with real users is just infeasible and hence user simulations based on accurate models play an essential role in understanding the implications of algorithmic changes to search engine results or presentation changes to the search engine result page. In this survey we summarize advances in modeling user click behavior on a web search engine result page. We present simple click models as well as more complex models aimed at..."
                },
                {
                    "title": "Methods for sample size determination in cluster randomized trials",
                    "abstract": "Background: The use of cluster randomized trials (CRTs) is increasing, along with the variety in their design and analysis. The simplest approach for their sample size calculation is to calculate the sample size assuming individual randomization and inflate this by a design effect to account for randomization by cluster. The assumptions of a simple design effect may not always be met; alternative or more complicated approaches are required. Methods: We summarise a wide range of sample size methods available for cluster randomized trials. For those familiar with sample size calculations for individually randomized trials but with less experience in the clustered case, this manuscript provides formulae for a wide range of scenarios with associated explanation and recommendations. For those with more experience, comprehensive summaries are provided that allow quick identification of methods for a given design, outcome and analysis method. Results: We present first those methods applicable to the simplest two-arm, parallel group, completely randomized design followed by methods that incorporate deviations from this design such as: variability in cluster sizes; attrition; non-compliance; or the inclusion of baseline covariates or repeated measures. The paper concludes with methods for alternative designs. Conclusions: There is a large amount of methodology available for sample size calculations in CRTs. This paper gives the most comprehensive description of published methodology for sample size calculation and provides an important resource for those designing these trials."
                },
                {
                    "title": "The use of propensity scores to assess the generalizability of results from randomized trials",
                    "abstract": "Summary.\u2002 Randomized trials remain the most accepted design for estimating the effects of interventions, but they do not necessarily answer a question of primary interest: will the programme be effective in a target population in which it may be implemented? In other words, are the results generalizable? There has been very little statistical research on how to assess the generalizability, or \u2018external validity\u2019, of randomized trials. We propose the use of propensity\u2010score\u2010based metrics to quantify the similarity of the participants in a randomized trial and a target population. In this setting the propensity score model predicts participation in the randomized trial, given a set of covariates. The resulting propensity scores are used first to quantify the difference between the trial participants and the target population, and then to match, subclassify or weight the control group outcomes to the population, assessing how well the propensity\u2010score\u2010adjusted outcomes track the outcomes that are actually observed in the population. These metrics can serve as a first step in assessing the generalizability of results from randomized trials to target populations. The paper lays out these ideas, discusses the assumptions underlying the approach and illustrates the metrics by using data on the evaluation of a schoolwide prevention programme called \u2018Positive behavioral interventions and supports\u2019."
                },
                {
                    "title": "Location, Location, Location: An Analysis of Profitability of Position in Online Advertising Markets",
                    "abstract": "The authors evaluate the impact of ad placement on revenues and profits generated from sponsored search. Their approach uses data generated through a field experiment for several keywords from an online retailer's ad campaign. Using a hierarchical Bayesian model, the authors measure the impact of ad placement on both click-through and conversion rates. They find that while click-through rate decreases with position, conversion rate increases with position and is even higher for more specific keywords. The net effect is that, contrary to the conventional wisdom in the industry, the topmost position is not necessarily the revenue-or profit-maximizing position. The authors\u2019 results inform the advertising strategies of firms participating in sponsored search auctions and provide insight into consumer behavior in these environments. Specifically, they help correct a significant misunderstanding among advertisers regarding the value of the top position. Furthermore, they reveal potential inefficiencies in current auction mechanisms that search engines use. The authors\u2019 results also reveal the information search strategies that consumers use in sponsored search and provide evidence of recency bias for immediate purchases."
                },
                {
                    "title": "An experimental comparison of click position-bias models",
                    "abstract": "Search engine click logs provide an invaluable source of relevance information, but this information is biased. A key source of bias is presentation order: the probability of click is influenced by a document's position in the results page. This paper focuses on explaining that bias, modelling how probability of click depends on position. We propose four simple hypotheses about how position bias might arise. We carry out a large data-gathering effort, where we perturb the ranking of a major search engine, to see how clicks are affected. We then explore which of the four hypotheses best explains the real-world position effects, and compare these to a simple logistic regression model. The data are not well explained by simple position models, where some users click indiscriminately on rank 1 or there is a simple decay of attention over ranks. A \u00e2 cascade' model, where users view results from top to bottom and leave as soon as they see a worthwhile document, is our best explanation for position bias in early ranks"
                },
                {
                    "title": "Position Auctions with Consumer Search",
                    "abstract": "This paper examines a model in which advertisers bid for \"sponsored-link\" positions on a search engine. The value advertisers derive from each position is endogenized as coming from sales to a population of consumers who make rational inferences about firm qualities and search optimally. Consumer search strategies, equilibrium bidding, and the welfare benefits of position auctions are analyzed. Implications for reserve prices and a number of other auction design questions are discussed."
                },
                {
                    "title": "An eye tracking study of the effect of target rank on web search",
                    "abstract": "Web search engines present search results in a rank ordered list. This works when what a user wants is near the top, but sometimes the information that the user really wants is located at the bottom of the page. This study examined how users' search behaviors vary when target results were displayed at various positions for informational and navigational tasks. We found that when targets were placed relatively low in the first page of search results, people spent more time searching and were less successful in finding the target, especially for informational tasks. Further analysis of eye movements showed that the decrease in search performance was partially due to the fact that users rarely looked at lower ranking results. The large decrease in performance for informational search is probably because users have high confidence in the search engine's ranking; in contrast to navigational tasks, where the target is more obvious from information presented in the results, in informational tasks, users try out the top ranked results even if these results are perceived as less relevant for the task."
                },
                {
                    "title": "In Google We Trust: Users' Decisions on Rank, Position, and Relevance",
                    "abstract": "An eye tracking experiment revealed that college student users have substantial trust in Google\u2019s ability to rank results by their true relevance to the query. When the participants selected a link to follow from Google\u2019s result pages, their decisions were strongly biased towards links higher in position even if the abstracts themselves were less relevant. While the participants reacted to artificially reduced retrieval quality by greater scrutiny, they failed to achieve the same success rate. This demonstrated trust in Google has implications for the search engine\u2019s tremendous potential influence on culture, society, and user traffic on the Web."
                },
                {
                    "title": "An Engine, Not a Camera: How Financial Models Shape Markets",
                    "abstract": "In An Engine, Not a Camera, Donald MacKenzie argues that the emergence of modern economic theories of finance affected financial markets in fundamental ways. These new, Nobel Prize-winning theories, based on elegant mathematical models of markets, were not simply external analyses but intrinsic parts of economic processes. Paraphrasing Milton Friedman, MacKenzie says that economic models are an engine of inquiry rather than a camera to reproduce empirical facts. More than that, the emergence of an authoritative theory of financial markets altered those markets fundamentally. For example, in 1970, there was almost no trading in financial derivatives such as \"futures.\" By June of 2004, derivatives contracts totaling $273 trillion were outstanding worldwide. MacKenzie suggests that this growth could never have happened without the development of theories that gave derivatives legitimacy and explained their complexities. MacKenzie examines the role played by finance theory in the two most serious crises to hit the world's financial markets in recent years: the stock market crash of 1987 and the market turmoil that engulfed the hedge fund Long-Term Capital Management in 1998. He also looks at finance theory that is somewhat beyond the mainstream\u2014chaos theorist Benoit Mandelbrot's model of \"wild\" randomness. MacKenzie's pioneering work in the social studies of finance will interest anyone who wants to understand how America\u2019s financial markets have grown into their current form."
                },
                {
                    "title": "Accurately interpreting clickthrough data as implicit feedback",
                    "abstract": "This paper examines the reliability of implicit feedback generated from clickthrough data in WWW search. Analyzing the users' decision process using eyetracking and comparing implicit feedback against manual relevance judgments, we conclude that clicks are informative but biased. While this makes the interpretation of clicks as absolute relevance judgments difficult, we show that relative preferences derived from clicks are reasonably accurate on average."
                },
                {
                    "title": "The Predictability of Social Events",
                    "abstract": "1 We benefited from the criticism and advice of several colleagues who kindly read various drafts of this paper. We are especially indebted to Professor Herbert Simon, who patiently read and painstakingly criticized all drafts of the present paperof which there were many. His contributions to our argument are too numerous to be listed; in particular, he suggested the use of Brouwer's Fixed Point Theorem. However, the responsibility for the final form in which this argument is now presented and, especially, for any errors of commission or omission is entirely ours."
                },
                {
                    "title": "Nudge: Improving Decisions about Health, Wealth, and Happiness",
                    "abstract": "NUDGE: IMPROVING DECISIONS ABOUT HEALTH, WEALTH, AND HAPPINESS by Richard H. Thaler and Cass R. Sunstein Penguin Books, 2009, 312 pp, ISBN 978-0-14-311526-7This book is best described formally as a general explanation of and advocacy for libertarian paternalism, a term coined by the authors in earlier publications. Informally, it is about how leaders, systems, organizations, and governments can nudge people to do the things the nudgers want and need done for the betterment of the nudgees, or of society. It is paternalism in the sense that \"it is legitimate for choice architects to try to influence people's behavior in order to make their lives longer, healthier, and better\", (p. 5) It is libertarian in that \"people should be free to do what they like - and to opt out of undesirable arrangements if they want to do so\", (p. 5) The built-in possibility of opting out or making a different choice preserves freedom of choice even though people's behavior has been influenced by the nature of the presentation of the information or by the structure of the decisionmaking system. I had never heard of libertarian paternalism before reading this book, and I now find it fascinating.Written for a general audience, this book contains mostly social and behavioral science theory and models, but there is considerable discussion of structure and process that has roots in mathematical and quantitative modeling. One of the main applications of this social system is economic choice in investing, selecting and purchasing products and services, systems of taxes, banking (mortgages, borrowing, savings), and retirement systems. Other quantitative social choice systems discussed include environmental effects, health care plans, gambling, and organ donations. Softer issues that are also subject to a nudge-based approach are marriage, education, eating, drinking, smoking, influence, spread of information, and politics. There is something in this book for everyone.The basis for this libertarian paternalism concept is in the social theory called \"science of choice\", the study of the design and implementation of influence systems on various kinds of people. The terms Econs and Humans, are used to refer to people with either considerable or little rational decision-making talent, respectively. The various libertarian paternalism concepts and systems presented are tested and compared in light of these two types of people. Two foundational issues that this book has in common with another book, Network of Echoes: Imitation, Innovation and Invisible Leaders, that was also reviewed for this issue of the Journal are that 1 ) there are two modes of thinking (or components of the brain) - an automatic (intuitive) process and a reflective (rational) process and 2) the need for conformity and the desire for imitation are powerful forces in human behavior. \u2026"
                },
                {
                    "title": "Position Effects in Search Advertising and their Moderators: A Regression Discontinuity Approach",
                    "abstract": "We investigate the causal effect of position in search engine advertising listings on outcomes such as click-through rates and sales orders. Because positions are determined through an auction, there are significant selection issues in measuring position effects. A simple mean comparison of outcomes at two positions is likely to be biased due to these selection issues. Additionally, experimentation is rendered difficult in this situation by competitors' bidding behavior, which induces selection biases that cannot be eliminated by randomizing the bids for the focal advertiser. Econometric approaches to address the selection are rendered infeasible due to the difficulty of finding suitable instruments in this context. We show that a regression discontinuity RD approach is feasible to measure causal effects in this important context. We apply the approach to a large and unique data set of daily observations containing information on a focal advertiser as well as its major competitors. Our RD estimates demonstrate that there are significant selection biases in the more naive estimates. While a mean comparison of outcomes across positions would indicate very large position effects, we find that our RD estimates of these effects are much smaller, and exist only in some of the positions. We further investigate moderators of these effects. Position effects are stronger when the advertiser is smaller, and when the consumer has low prior experience with the keyword for the advertiser. They are weaker when the keyword phrase has specific brand or product information, when the ad copy is more specific as in exact matching options, and on weekends compared to weekdays."
                }
            ],
            "categories": [
                "cs.CY",
                "cs.IR"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we quantitatively estimate the performative power of search engines in influencing web traffic through content arrangement?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant gap in understanding the economic power of digital platforms, particularly in the context of antitrust investigations. By providing a robust framework for measuring performative power, this research could reshape how policymakers, economists, and legal experts approach digital market regulation. It could lead to more effective antitrust enforcement tools that account for the behavioral dynamics of multi-sided platforms, ultimately advancing knowledge in both economic theory and practical applications in digital market governance.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities of measuring causal effects in multi-sided platforms, where traditional antitrust tools fall short. Naive approaches may fail because they do not adequately capture the intricate interactions between users and platforms, nor do they account for the behavioral aspects that influence user decisions. Technical obstacles include designing experiments that can isolate the effects of content arrangement from other confounding factors, as well as ensuring that the data collected is representative and reliable.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has lacked a comprehensive framework for measuring performative power, primarily due to the limitations of existing antitrust tools that do not address the unique characteristics of digital platforms. Barriers include the absence of experimental methodologies that can effectively capture the causal relationships between content arrangement and user behavior. This work differs from prior efforts by introducing a novel experimental design using a browser extension to manipulate search result displays, thereby providing empirical evidence of performative power in action.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves designing and implementing an online experiment using a browser extension called Powermeter, which modifies search result displays to create counterfactual arrangements. We will analyze the impact of these arrangements on click-through rates for Google Search and Bing, using metrics such as average click probability. The expected outcomes include quantitative insights into the causal effects of content arrangement on user behavior, providing a proof of concept for the practical applicability of the performative power framework in digital market investigations."
            }
        },
        "author_data": {
            "087d3b57-1577-4217-9572-81ab8e1fe2b0": {
                "pk": "087d3b57-1577-4217-9572-81ab8e1fe2b0",
                "name": "Celestine Mendler-D\u00fcnner",
                "collaborators": [
                    "Moritz Hardt",
                    "Meena Jagadeesan",
                    "Martin Jaggi",
                    "Tijana Zrnic",
                    "Aurelien Lucchi",
                    "Joachim Baumann",
                    "Eliza Wszola",
                    "Markus P\u00fcschel",
                    "Nikolas Ioannou",
                    "Thomas Parnell"
                ],
                "domain": [
                    "Machine Learning",
                    "Optimization",
                    "Causal Inference",
                    "Collective Action"
                ],
                "publications": [
                    {
                        "title": "Randomized Block-Diagonal Preconditioning for Parallel Learning",
                        "abstract": "We study preconditioned gradient-based optimization methods where the preconditioning matrix has block-diagonal form. Such a structural constraint comes with the advantage that the update computation is block-separable and can be parallelized across multiple independent tasks. Our main contribution is to demonstrate that the convergence of these methods can significantly be improved by a randomization technique which corresponds to repartitioning coordinates across tasks during the optimization procedure. We provide a theoretical analysis that accurately characterizes the expected convergence gains of repartitioning and validate our findings empirically on various traditional machine learning tasks. From an implementation perspective, block-separable models are well suited for parallelization and, when shared memory is available, randomization can be implemented on top of existing methods very efficiently to improve convergence."
                    },
                    {
                        "title": "Performative Prediction: Past and Future",
                        "abstract": "Predictions in the social world generally influence the target of prediction, a phenomenon known as performativity. Self-fulfilling and self-negating predictions are examples of performativity. Of fundamental importance to economics, finance, and the social sciences, the notion has been absent from the development of machine learning. In machine learning applications, performativity often surfaces as distribution shift. A predictive model deployed on a digital platform, for example, influences consumption and thereby changes the data-generating distribution. We survey the recently founded area of performative prediction that provides a definition and conceptual framework to study performativity in machine learning. A consequence of performative prediction is a natural equilibrium notion that gives rise to new optimization challenges. Another consequence is a distinction between learning and steering, two mechanisms at play in performative prediction. The notion of steering is in turn intimately related to questions of power in digital markets. We review the notion of performative power that gives an answer to the question how much a platform can steer participants through its predictions. We end on a discussion of future directions, such as the role that performativity plays in contesting algorithmic systems."
                    },
                    {
                        "title": "Algorithmic Collective Action in Recommender Systems: Promoting Songs by Reordering Playlists",
                        "abstract": "We investigate algorithmic collective action in transformer-based recommender systems. Our use case is a collective of fans aiming to promote the visibility of an artist by strategically placing one of their songs in the existing playlists they control. The success of the collective is measured by the increase in test-time recommendations of the targeted song. We introduce two easily implementable strategies towards this goal and test their efficacy on a publicly available recommender system model released by a major music streaming platform. Our findings reveal that even small collectives (controlling less than 0.01% of the training data) can achieve up 25x amplification of recommendations by strategically choosing the position at which to insert the song. We then focus on investigating the externalities of the strategy. We find that the performance loss for the platform is negligible, and the recommendations of other songs are largely preserved, minimally impairing the user experience of participants. Moreover, the costs are evenly distributed among other artists. Taken together, our findings demonstrate how collective action strategies can be effective while not necessarily being adversarial, raising new questions around incentives, social dynamics, and equilibria in recommender systems."
                    },
                    {
                        "title": "On Linear Learning with Manycore Processors",
                        "abstract": "A new generation of manycore processors is on the rise that offers dozens and more cores on a chip and, in a sense, fuses host processor and accelerator. In this paper we target the efficient training of generalized linear models on these machines. We propose a novel approach for achieving parallelism which we call Heterogeneous Tasks on Homogeneous Cores (HTHC). It divides the problem into multiple fundamentally different tasks, which themselves are parallelized. For evaluation, we design a detailed, architecture-cognizant implementation of our scheme on a recent 72-core Knights Landing processor that is adaptive to the cache, memory, and core structure. Our library efficiently supports dense and sparse datasets as well as 4-bit quantized data for further possible gains in performance. We show benchmarks for Lasso and SVM with different data sets against straightforward parallel implementations and prior software. In particular, for Lasso on dense data, we improve the state-of-the-art by an order of magnitude."
                    },
                    {
                        "title": "Performative Power",
                        "abstract": "We introduce the notion of performative power, which measures the ability of a firm operating an algorithmic system, such as a digital content recommendation platform, to cause change in a population of participants. We relate performative power to the economic study of competition in digital economies. Traditional economic concepts struggle with identifying anti-competitive patterns in digital platforms not least due to the complexity of market definition. In contrast, performative power is a causal notion that is identifiable with minimal knowledge of the market, its internals, participants, products, or prices.   Low performative power implies that a firm can do no better than to optimize their objective on current data. In contrast, firms of high performative power stand to benefit from steering the population towards more profitable behavior. We confirm in a simple theoretical model that monopolies maximize performative power. A firm's ability to personalize increases performative power, while competition and outside options decrease performative power. On the empirical side, we propose an observational causal design to identify performative power from discontinuities in how digital platforms display content. This allows to repurpose causal effects from various studies about digital platforms as lower bounds on performative power. Finally, we speculate about the role that performative power might play in competition policy and antitrust enforcement in digital marketplaces."
                    },
                    {
                        "title": "SySCD: A System-Aware Parallel Coordinate Descent Algorithm",
                        "abstract": "In this paper we propose a novel parallel stochastic coordinate descent (SCD) algorithm with convergence guarantees that exhibits strong scalability. We start by studying a state-of-the-art parallel implementation of SCD and identify scalability as well as system-level performance bottlenecks of the respective implementation. We then take a principled approach to develop a new SCD variant which is designed to avoid the identified system bottlenecks, such as limited scaling due to coherence traffic of model sharing across threads, and inefficient CPU cache accesses. Our proposed system-aware parallel coordinate descent algorithm (SySCD) scales to many cores and across numa nodes, and offers a consistent bottom line speedup in training time of up to x12 compared to an optimized asynchronous parallel SCD algorithm and up to x42, compared to state-of-the-art GLM solvers (scikit-learn, Vowpal Wabbit, and H2O) on a range of datasets and multi-core CPU architectures."
                    },
                    {
                        "title": "Alternative Microfoundations for Strategic Classification",
                        "abstract": "When reasoning about strategic behavior in a machine learning context it is tempting to combine standard microfoundations of rational agents with the statistical decision theory underlying classification. In this work, we argue that a direct combination of these standard ingredients leads to brittle solution concepts of limited descriptive and prescriptive value. First, we show that rational agents with perfect information produce discontinuities in the aggregate response to a decision rule that we often do not observe empirically. Second, when any positive fraction of agents is not perfectly strategic, desirable stable points -- where the classifier is optimal for the data it entails -- cease to exist. Third, optimal decision rules under standard microfoundations maximize a measure of negative externality known as social burden within a broad class of possible assumptions about agent behavior.   Recognizing these limitations we explore alternatives to standard microfoundations for binary classification. We start by describing a set of desiderata that help navigate the space of possible assumptions about how agents respond to a decision rule. In particular, we analyze a natural constraint on feature manipulations, and discuss properties that are sufficient to guarantee the robust existence of stable points. Building on these insights, we then propose the noisy response model. Inspired by smoothed analysis and empirical observations, noisy response incorporates imperfection in the agent responses, which we show mitigates the limitations of standard microfoundations. Our model retains analytical tractability, leads to more robust insights about stable points, and imposes a lower social burden at optimality."
                    },
                    {
                        "title": "Regret Minimization with Performative Feedback",
                        "abstract": "In performative prediction, the deployment of a predictive model triggers a shift in the data distribution. As these shifts are typically unknown ahead of time, the learner needs to deploy a model to get feedback about the distribution it induces. We study the problem of finding near-optimal models under performativity while maintaining low regret. On the surface, this problem might seem equivalent to a bandit problem. However, it exhibits a fundamentally richer feedback structure that we refer to as performative feedback: after every deployment, the learner receives samples from the shifted distribution rather than only bandit feedback about the reward. Our main contribution is an algorithm that achieves regret bounds scaling only with the complexity of the distribution shifts and not that of the reward function. The algorithm only relies on smoothness of the shifts and does not assume convexity. Moreover, its final iterate is guaranteed to be near-optimal. The key algorithmic idea is careful exploration of the distribution shifts that informs a novel construction of confidence bounds on the risk of unexplored models. More broadly, our work establishes a conceptual approach for leveraging tools from the bandits literature for the purpose of regret minimization with performative feedback."
                    },
                    {
                        "title": "Anticipating Performativity by Predicting from Predictions",
                        "abstract": "Predictions about people, such as their expected educational achievement or their credit risk, can be performative and shape the outcome that they aim to predict. Understanding the causal effect of these predictions on the eventual outcomes is crucial for foreseeing the implications of future predictive models and selecting which models to deploy. However, this causal estimation task poses unique challenges: model predictions are usually deterministic functions of input features and highly correlated with outcomes. This can make the causal effects of predictions on outcomes impossible to disentangle from the direct effect of the covariates. We study this problem through the lens of causal identifiability, and despite the hardness of this problem in full generality, we highlight three natural scenarios where the causal relationship between covariates, predictions and outcomes can be identified from observational data: randomization in predictions, overparameterization of the predictive model deployed during data collection, and discrete prediction outputs. Empirically we show that given our identifiability conditions hold, standard variants of supervised learning that predict from predictions by treating the prediction as an input feature can indeed find transferable functional relationships that allow for conclusions about newly deployed predictive models. These positive results fundamentally rely on model predictions being recorded during data collection, bringing forward the importance of rethinking standard data collection practices to enable progress towards a better understanding of social outcomes and performative feedback loops."
                    },
                    {
                        "title": "Algorithmic Collective Action in Machine Learning",
                        "abstract": "We initiate a principled study of algorithmic collective action on digital platforms that deploy machine learning algorithms. We propose a simple theoretical model of a collective interacting with a firm's learning algorithm. The collective pools the data of participating individuals and executes an algorithmic strategy by instructing participants how to modify their own data to achieve a collective goal. We investigate the consequences of this model in three fundamental learning-theoretic settings: the case of a nonparametric optimal learning algorithm, a parametric risk minimizer, and gradient-based optimization. In each setting, we come up with coordinated algorithmic strategies and characterize natural success criteria as a function of the collective's size. Complementing our theory, we conduct systematic experiments on a skill classification task involving tens of thousands of resumes from a gig platform for freelancers. Through more than two thousand model training runs of a BERT-like language model, we see a striking correspondence emerge between our empirical observations and the predictions made by our theory. Taken together, our theory and experiments broadly support the conclusion that algorithmic collectives of exceedingly small fractional size can exert significant control over a platform's learning algorithm."
                    },
                    {
                        "title": "Causal Inference out of Control: Estimating the Steerability of Consumption",
                        "abstract": "Regulators and academics are increasingly interested in the causal effect that algorithmic actions of a digital platform have on consumption. We introduce a general causal inference problem we call the steerability of consumption that abstracts many settings of interest. Focusing on observational designs and exploiting the structure of the problem, we exhibit a set of assumptions for causal identifiability that significantly weaken the often unrealistic overlap assumptions of standard designs. The key novelty of our approach is to explicitly model the dynamics of consumption over time, viewing the platform as a controller acting on a dynamical system. From this dynamical systems perspective, we are able to show that exogenous variation in consumption and appropriately responsive algorithmic control actions are sufficient for identifying steerability of consumption. Our results illustrate the fruitful interplay of control theory and causal inference, which we illustrate with examples from econometrics, macroeconomics, and machine learning."
                    },
                    {
                        "title": "Collaborative Learning via Prediction Consensus",
                        "abstract": "We consider a collaborative learning setting where the goal of each agent is to improve their own model by leveraging the expertise of collaborators, in addition to their own training data. To facilitate the exchange of expertise among agents, we propose a distillation-based method leveraging shared unlabeled auxiliary data, which is pseudo-labeled by the collective. Central to our method is a trust weighting scheme that serves to adaptively weigh the influence of each collaborator on the pseudo-labels until a consensus on how to label the auxiliary data is reached. We demonstrate empirically that our collaboration scheme is able to significantly boost the performance of individual models in the target domain from which the auxiliary data is sampled. By design, our method adeptly accommodates heterogeneity in model architectures and substantially reduces communication overhead compared to typical collaborative learning methods. At the same time, it can provably mitigate the negative impact of bad models on the collective."
                    },
                    {
                        "title": "Questioning the Survey Responses of Large Language Models",
                        "abstract": "As large language models increase in capability, researchers have started to conduct surveys of all kinds on these models in order to investigate the population represented by their responses. In this work, we critically examine language models' survey responses on the basis of the well-established American Community Survey by the U.S. Census Bureau and investigate whether they elicit a faithful representations of any human population. Using a de-facto standard multiple-choice prompting technique and evaluating 39 different language models using systematic experiments, we establish two dominant patterns: First, models' responses are governed by ordering and labeling biases, leading to variations across models that do not persist after adjusting for systematic biases. Second, models' responses do not contain the entropy variations and statistical signals typically found in human populations. As a result, a binary classifier can almost perfectly differentiate model-generated data from the responses of the U.S. census. At the same time, models' relative alignment with different demographic subgroups can be predicted from the subgroups' entropy, irrespective of the model's training data or training strategy. Taken together, our findings suggest caution in treating models' survey responses as equivalent to those of human populations."
                    },
                    {
                        "title": "Evaluating language models as risk scores",
                        "abstract": "Current question-answering benchmarks predominantly focus on accuracy in realizable prediction tasks. Conditioned on a question and answer-key, does the most likely token match the ground truth? Such benchmarks necessarily fail to evaluate LLMs' ability to quantify ground-truth outcome uncertainty. In this work, we focus on the use of LLMs as risk scores for unrealizable prediction tasks. We introduce folktexts, a software package to systematically generate risk scores using LLMs, and evaluate them against US Census data products. A flexible API enables the use of different prompting schemes, local or web-hosted models, and diverse census columns that can be used to compose custom prediction tasks. We evaluate 17 recent LLMs across five proposed benchmark tasks. We find that zero-shot risk scores produced by multiple-choice question-answering have high predictive signal but are widely miscalibrated. Base models consistently overestimate outcome uncertainty, while instruction-tuned models underestimate uncertainty and produce over-confident risk scores. In fact, instruction-tuning polarizes answer distribution regardless of true underlying data uncertainty. This reveals a general inability of instruction-tuned LLMs to express data uncertainty using multiple-choice answers. A separate experiment using verbalized chat-style risk queries yields substantially improved calibration across instruction-tuned models. These differences in ability to quantify data uncertainty cannot be revealed in realizable settings, and highlight a blind-spot in the current evaluation ecosystem that folktexts covers."
                    },
                    {
                        "title": "Decline Now: A Combinatorial Model for Algorithmic Collective Action",
                        "abstract": "Drivers on food delivery platforms often run a loss on low-paying orders. In response, workers on DoorDash started a campaign, #DeclineNow, to purposefully decline orders below a certain pay threshold. For each declined order, the platform returns the request to other available drivers with slightly increased pay. While contributing to overall pay increase the implementation of the strategy comes with the risk of missing out on orders for each individual driver. In this work, we propose a first combinatorial model to study the strategic interaction between workers and the platform. Within our model, we formalize key quantities such as the average worker benefit of the strategy, the benefit of freeriding, as well as the benefit of participation. We extend our theoretical results with simulations. Our key insights show that the average worker gain of the strategy is always positive, while the benefit of participation is positive only for small degrees of labor oversupply. Beyond this point, the utility of participants decreases faster with increasing degree of oversupply, compared to the utility of non-participants. Our work highlights the significance of labor supply levels for the effectiveness of collective action on gig platforms. We suggest organizing in shifts as a means to reduce oversupply and empower collectives."
                    }
                ]
            },
            "1c800a26-dd2d-4389-b2b9-74e0f1230352": {
                "pk": "1c800a26-dd2d-4389-b2b9-74e0f1230352",
                "name": "Gabriele Carovano",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "117e1e89-3dc5-4f86-b788-c9635d9e35fe": {
                "pk": "117e1e89-3dc5-4f86-b788-c9635d9e35fe",
                "name": "Moritz Hardt",
                "collaborators": [
                    "Aaron Roth",
                    "Eric Price",
                    "Benjamin Recht",
                    "Avrim Blum",
                    "Celestine Mendler-D\u00fcnner",
                    "Mary Wootters",
                    "Kunal Talwar",
                    "Ankur Moitra",
                    "John Miller",
                    "Tengyu Ma"
                ],
                "domain": [
                    "Machine Learning",
                    "Differential Privacy",
                    "Matrix Completion",
                    "Adaptive Data Analysis"
                ],
                "publications": [
                    {
                        "title": "Climbing a shaky ladder: Better adaptive risk estimation",
                        "abstract": "We revisit the \\emph{leaderboard problem} introduced by Blum and Hardt (2015) in an effort to reduce overfitting in machine learning benchmarks. We show that a randomized version of their Ladder algorithm achieves leaderboard error O(1/n^{0.4}) compared with the previous best rate of O(1/n^{1/3}).   Short of proving that our algorithm is optimal, we point out a major obstacle toward further progress. Specifically, any improvement to our upper bound would lead to asymptotic improvements in the general adaptive estimation setting as have remained elusive in recent years. This connection also directly leads to lower bounds for specific classes of algorithms. In particular, we exhibit a new attack on the leaderboard algorithm that both theoretically and empirically distinguishes between our algorithm and previous leaderboard algorithms."
                    },
                    {
                        "title": "Understanding Alternating Minimization for Matrix Completion",
                        "abstract": "Alternating Minimization is a widely used and empirically successful heuristic for matrix completion and related low-rank optimization problems. Theoretical guarantees for Alternating Minimization have been hard to come by and are still poorly understood. This is in part because the heuristic is iterative and non-convex in nature. We give a new algorithm based on Alternating Minimization that provably recovers an unknown low-rank matrix from a random subsample of its entries under a standard incoherence assumption. Our results reduce the sample size requirements of the Alternating Minimization approach by at least a quartic factor in the rank and the condition number of the unknown matrix. These improvements apply even if the matrix is only close to low-rank in the Frobenius norm. Our algorithm runs in nearly linear time in the dimension of the matrix and, in a broad range of parameters, gives the strongest sample bounds among all subquadratic time algorithms that we are aware of.   Underlying our work is a new robust convergence analysis of the well-known Power Method for computing the dominant singular vectors of a matrix. This viewpoint leads to a conceptually simple understanding of Alternating Minimization. In addition, we contribute a new technique for controlling the coherence of intermediate solutions arising in iterative algorithms based on a smoothed analysis of the QR factorization. These techniques may be of interest beyond their application here."
                    },
                    {
                        "title": "Patterns, predictions, and actions: A story about machine learning",
                        "abstract": "This graduate textbook on machine learning tells a story of how patterns in data support predictions and consequential actions. Starting with the foundations of decision making, we cover representation, optimization, and generalization as the constituents of supervised learning. A chapter on datasets as benchmarks examines their histories and scientific bases. Self-contained introductions to causality, the practice of causal inference, sequential decision making, and reinforcement learning equip the reader with concepts and tools to reason about actions and their consequences. Throughout, the text discusses historical context and societal impact. We invite readers from all backgrounds; some experience with probability, calculus, and linear algebra suffices."
                    },
                    {
                        "title": "Beating Randomized Response on Incoherent Matrices",
                        "abstract": "Computing accurate low rank approximations of large matrices is a fundamental data mining task. In many applications however the matrix contains sensitive information about individuals. In such case we would like to release a low rank approximation that satisfies a strong privacy guarantee such as differential privacy. Unfortunately, to date the best known algorithm for this task that satisfies differential privacy is based on naive input perturbation or randomized response: Each entry of the matrix is perturbed independently by a sufficiently large random noise variable, a low rank approximation is then computed on the resulting matrix.   We give (the first) significant improvements in accuracy over randomized response under the natural and necessary assumption that the matrix has low coherence. Our algorithm is also very efficient and finds a constant rank approximation of an m x n matrix in time O(mn). Note that even generating the noise matrix required for randomized response already requires time O(mn)."
                    },
                    {
                        "title": "Beyond Worst-Case Analysis in Private Singular Vector Computation",
                        "abstract": "We consider differentially private approximate singular vector computation. Known worst-case lower bounds show that the error of any differentially private algorithm must scale polynomially with the dimension of the singular vector. We are able to replace this dependence on the dimension by a natural parameter known as the coherence of the matrix that is often observed to be significantly smaller than the dimension both theoretically and empirically. We also prove a matching lower bound showing that our guarantee is nearly optimal for every setting of the coherence parameter. Notably, we achieve our bounds by giving a robust analysis of the well-known power iteration algorithm, which may be of independent interest. Our algorithm also leads to improvements in worst-case settings and to better low-rank approximations in the spectral norm."
                    },
                    {
                        "title": "The Noisy Power Method: A Meta Algorithm with Applications",
                        "abstract": "We provide a new robust convergence analysis of the well-known power method for computing the dominant singular vectors of a matrix that we call the noisy power method. Our result characterizes the convergence behavior of the algorithm when a significant amount noise is introduced after each matrix-vector multiplication. The noisy power method can be seen as a meta-algorithm that has recently found a number of important applications in a broad range of machine learning problems including alternating minimization for matrix completion, streaming principal component analysis (PCA), and privacy-preserving spectral analysis. Our general analysis subsumes several existing ad-hoc convergence bounds and resolves a number of open problems in multiple applications including streaming PCA and privacy-preserving singular vector computation."
                    },
                    {
                        "title": "The Ladder: A Reliable Leaderboard for Machine Learning Competitions",
                        "abstract": "The organizer of a machine learning competition faces the problem of maintaining an accurate leaderboard that faithfully represents the quality of the best submission of each competing team. What makes this estimation problem particularly challenging is its sequential and adaptive nature. As participants are allowed to repeatedly evaluate their submissions on the leaderboard, they may begin to overfit to the holdout data that supports the leaderboard. Few theoretical results give actionable advice on how to design a reliable leaderboard. Existing approaches therefore often resort to poorly understood heuristics such as limiting the bit precision of answers and the rate of re-submission.   In this work, we introduce a notion of \"leaderboard accuracy\" tailored to the format of a competition. We introduce a natural algorithm called \"the Ladder\" and demonstrate that it simultaneously supports strong theoretical guarantees in a fully adaptive model of estimation, withstands practical adversarial attacks, and achieves high utility on real submission files from an actual competition hosted by Kaggle.   Notably, we are able to sidestep a powerful recent hardness result for adaptive risk estimation that rules out algorithms such as ours under a seemingly very similar notion of accuracy. On a practical note, we provide a completely parameter-free variant of our algorithm that can be deployed in a real competition with no tuning required whatsoever."
                    },
                    {
                        "title": "Performative Prediction: Past and Future",
                        "abstract": "Predictions in the social world generally influence the target of prediction, a phenomenon known as performativity. Self-fulfilling and self-negating predictions are examples of performativity. Of fundamental importance to economics, finance, and the social sciences, the notion has been absent from the development of machine learning. In machine learning applications, performativity often surfaces as distribution shift. A predictive model deployed on a digital platform, for example, influences consumption and thereby changes the data-generating distribution. We survey the recently founded area of performative prediction that provides a definition and conceptual framework to study performativity in machine learning. A consequence of performative prediction is a natural equilibrium notion that gives rise to new optimization challenges. Another consequence is a distinction between learning and steering, two mechanisms at play in performative prediction. The notion of steering is in turn intimately related to questions of power in digital markets. We review the notion of performative power that gives an answer to the question how much a platform can steer participants through its predictions. We end on a discussion of future directions, such as the role that performativity plays in contesting algorithmic systems."
                    },
                    {
                        "title": "Fast matrix completion without the condition number",
                        "abstract": "We give the first algorithm for Matrix Completion whose running time and sample complexity is polynomial in the rank of the unknown target matrix, linear in the dimension of the matrix, and logarithmic in the condition number of the matrix. To the best of our knowledge, all previous algorithms either incurred a quadratic dependence on the condition number of the unknown matrix or a quadratic dependence on the dimension of the matrix in the running time. Our algorithm is based on a novel extension of Alternating Minimization which we show has theoretical guarantees under standard assumptions even in the presence of noise."
                    },
                    {
                        "title": "On the Geometry of Differential Privacy",
                        "abstract": "We consider the noise complexity of differentially private mechanisms in the setting where the user asks $d$ linear queries $f\\colon\\Rn\\to\\Re$ non-adaptively. Here, the database is represented by a vector in $\\Rn$ and proximity between databases is measured in the $\\ell_1$-metric.   We show that the noise complexity is determined by two geometric parameters associated with the set of queries.   We use this connection to give tight upper and lower bounds on the noise complexity for any $d \\leq n$. We show that for $d$ random linear queries of sensitivity~1, it is necessary and sufficient to add $\\ell_2$-error $\\Theta(\\min\\{d\\sqrt{d}/\\epsilon,d\\sqrt{\\log (n/d)}/\\epsilon\\})$ to achieve $\\epsilon$-differential privacy. Assuming the truth of a deep conjecture from convex geometry, known as the Hyperplane conjecture, we can extend our results to arbitrary linear queries giving nearly matching upper and lower bounds.   Our bound translates to error $O(\\min\\{d/\\epsilon,\\sqrt{d\\log(n/d)}/\\epsilon\\})$ per answer. The best previous upper bound (Laplacian mechanism) gives a bound of $O(\\min\\{d/\\eps,\\sqrt{n}/\\epsilon\\})$ per answer, while the best known lower bound was $\\Omega(\\sqrt{d}/\\epsilon)$. In contrast, our lower bound is strong enough to separate the concept of differential privacy from the notion of approximate differential privacy where an upper bound of $O(\\sqrt{d}/\\epsilon)$ can be achieved."
                    },
                    {
                        "title": "Algorithms and Hardness for Robust Subspace Recovery",
                        "abstract": "We consider a fundamental problem in unsupervised learning called \\emph{subspace recovery}: given a collection of $m$ points in $\\mathbb{R}^n$, if many but not necessarily all of these points are contained in a $d$-dimensional subspace $T$ can we find it? The points contained in $T$ are called {\\em inliers} and the remaining points are {\\em outliers}. This problem has received considerable attention in computer science and in statistics. Yet efficient algorithms from computer science are not robust to {\\em adversarial} outliers, and the estimators from robust statistics are hard to compute in high dimensions.   Are there algorithms for subspace recovery that are both robust to outliers and efficient? We give an algorithm that finds $T$ when it contains more than a $\\frac{d}{n}$ fraction of the points. Hence, for say $d = n/2$ this estimator is both easy to compute and well-behaved when there are a constant fraction of outliers. We prove that it is Small Set Expansion hard to find $T$ when the fraction of errors is any larger, thus giving evidence that our estimator is an {\\em optimal} compromise between efficiency and robustness.   As it turns out, this basic problem has a surprising number of connections to other areas including small set expansion, matroid theory and functional analysis that we make use of here."
                    },
                    {
                        "title": "Stable Recurrent Models",
                        "abstract": "Stability is a fundamental property of dynamical systems, yet to this date it has had little bearing on the practice of recurrent neural networks. In this work, we conduct a thorough investigation of stable recurrent models. Theoretically, we prove stable recurrent neural networks are well approximated by feed-forward networks for the purpose of both inference and training by gradient descent. Empirically, we demonstrate stable recurrent models often perform as well as their unstable counterparts on benchmark sequence tasks. Taken together, these findings shed light on the effective power of recurrent networks and suggest much of sequence learning happens, or can be made to happen, in the stable regime. Moreover, our results help to explain why in many cases practitioners succeed in replacing recurrent models by feed-forward models."
                    },
                    {
                        "title": "Identity Matters in Deep Learning",
                        "abstract": "An emerging design principle in deep learning is that each layer of a deep artificial neural network should be able to easily express the identity transformation. This idea not only motivated various normalization techniques, such as \\emph{batch normalization}, but was also key to the immense success of \\emph{residual networks}.   In this work, we put the principle of \\emph{identity parameterization} on a more solid theoretical footing alongside further empirical progress. We first give a strikingly simple proof that arbitrarily deep linear residual networks have no spurious local optima. The same result for linear feed-forward networks in their standard parameterization is substantially more delicate. Second, we show that residual networks with ReLu activations have universal finite-sample expressivity in the sense that the network can represent any function of its sample provided that the model has more parameters than the sample size.   Directly inspired by our theory, we experiment with a radically simple residual architecture consisting of only residual convolutional layers and ReLu activations, but no batch normalization, dropout, or max pool. Our model improves significantly on previous all-convolutional networks on the CIFAR10, CIFAR100, and ImageNet classification benchmarks."
                    },
                    {
                        "title": "Test-Time Training on Nearest Neighbors for Large Language Models",
                        "abstract": "Many recent efforts augment language models with retrieval, by adding retrieved data to the input context. For this approach to succeed, the retrieved data must be added at both training and test time. Moreover, as input length grows linearly with the size of retrieved data, cost in computation and memory grows quadratically for modern Transformers. To avoid these complications, we simply fine-tune the model on retrieved data at test time, using its standard training setup. We build a large-scale distributed index based on text embeddings of the Pile dataset. For each test input, our system retrieves its neighbors and fine-tunes the model on their text. Surprisingly, retrieving and training on as few as 20 neighbors, each for only one gradient iteration, drastically improves performance across more than 20 language modeling tasks in the Pile. For example, test-time training with nearest neighbors significantly narrows the performance gap between a small GPT-2 and a GPT-Neo model more than 10 times larger. Sufficient index quality and size, however, are necessary. Our work establishes a first baseline of test-time training for language modeling."
                    },
                    {
                        "title": "What Makes ImageNet Look Unlike LAION",
                        "abstract": "ImageNet was famously created from Flickr image search results. What if we recreated ImageNet instead by searching the massive LAION dataset based on image captions alone? In this work, we carry out this counterfactual investigation. We find that the resulting ImageNet recreation, which we call LAIONet, looks distinctly unlike the original. Specifically, the intra-class similarity of images in the original ImageNet is dramatically higher than it is for LAIONet. Consequently, models trained on ImageNet perform significantly worse on LAIONet. We propose a rigorous explanation for the discrepancy in terms of a subtle, yet important, difference in two plausible causal data-generating processes for the respective datasets, that we support with systematic experimentation. In a nutshell, searching based on an image caption alone creates an information bottleneck that mitigates the selection bias otherwise present in image-based filtering. Our explanation formalizes a long-held intuition in the community that ImageNet images are stereotypical, unnatural, and overly simple representations of the class category. At the same time, it provides a simple and actionable takeaway for future dataset creation efforts."
                    },
                    {
                        "title": "ImageNot: A contrast with ImageNet preserves model rankings",
                        "abstract": "We introduce ImageNot, a dataset designed to match the scale of ImageNet while differing drastically in other aspects. We show that key model architectures developed for ImageNet over the years rank identically when trained and evaluated on ImageNot to how they rank on ImageNet. This is true when training models from scratch or fine-tuning them. Moreover, the relative improvements of each model over earlier models strongly correlate in both datasets. We further give evidence that ImageNot has a similar utility as ImageNet for transfer learning purposes. Our work demonstrates a surprising degree of external validity in the relative performance of image classification models. This stands in contrast with absolute accuracy numbers that typically drop sharply even under small changes to a dataset."
                    },
                    {
                        "title": "Tight bounds for learning a mixture of two gaussians",
                        "abstract": "We consider the problem of identifying the parameters of an unknown mixture of two arbitrary $d$-dimensional gaussians from a sequence of independent random samples. Our main results are upper and lower bounds giving a computationally efficient moment-based estimator with an optimal convergence rate, thus resolving a problem introduced by Pearson (1894). Denoting by $\\sigma^2$ the variance of the unknown mixture, we prove that $\\Theta(\\sigma^{12})$ samples are necessary and sufficient to estimate each parameter up to constant additive error when $d=1.$ Our upper bound extends to arbitrary dimension $d>1$ up to a (provably necessary) logarithmic loss in $d$ using a novel---yet simple---dimensionality reduction technique. We further identify several interesting special cases where the sample complexity is notably smaller than our optimal worst-case bound. For instance, if the means of the two components are separated by $\\Omega(\\sigma)$ the sample complexity reduces to $O(\\sigma^2)$ and this is again optimal.   Our results also apply to learning each component of the mixture up to small error in total variation distance, where our algorithm gives strong improvements in sample complexity over previous work. We also extend our lower bound to mixtures of $k$ Gaussians, showing that $\\Omega(\\sigma^{6k-2})$ samples are necessary to estimate each parameter up to constant additive error."
                    },
                    {
                        "title": "Preventing False Discovery in Interactive Data Analysis is Hard",
                        "abstract": "We show that, under a standard hardness assumption, there is no computationally efficient algorithm that given $n$ samples from an unknown distribution can give valid answers to $n^{3+o(1)}$ adaptively chosen statistical queries. A statistical query asks for the expectation of a predicate over the underlying distribution, and an answer to a statistical query is valid if it is \"close\" to the correct expectation over the distribution.   Our result stands in stark contrast to the well known fact that exponentially many statistical queries can be answered validly and efficiently if the queries are chosen non-adaptively (no query may depend on the answers to previous queries). Moreover, a recent work by Dwork et al. shows how to accurately answer exponentially many adaptively chosen statistical queries via a computationally inefficient algorithm; and how to answer a quadratic number of adaptive queries via a computationally efficient algorithm. The latter result implies that our result is tight up to a linear factor in $n.$   Conceptually, our result demonstrates that achieving statistical validity alone can be a source of computational intractability in adaptive settings. For example, in the modern large collaborative research environment, data analysts typically choose a particular approach based on previous findings. False discovery occurs if a research finding is supported by the data but not by the underlying distribution. While the study of preventing false discovery in Statistics is decades old, to the best of our knowledge our result is the first to demonstrate a computational barrier. In particular, our result suggests that the perceived difficulty of preventing false discovery in today's collaborative research environment may be inherent."
                    },
                    {
                        "title": "Natural Analysts in Adaptive Data Analysis",
                        "abstract": "Adaptive data analysis is frequently criticized for its pessimistic generalization guarantees. The source of these pessimistic bounds is a model that permits arbitrary, possibly adversarial analysts that optimally use information to bias results. While being a central issue in the field, still lacking are notions of natural analysts that allow for more optimistic bounds faithful to the reality that typical analysts aren't adversarial.   In this work, we propose notions of natural analysts that smoothly interpolate between the optimal non-adaptive bounds and the best-known adaptive generalization bounds. To accomplish this, we model the analyst's knowledge as evolving according to the rules of an unknown dynamical system that takes in revealed information and outputs new statistical queries to the data. This allows us to restrict the analyst through different natural control-theoretic notions. One such notion corresponds to a recency bias, formalizing an inability to arbitrarily use distant information. Another complementary notion formalizes an anchoring bias, a tendency to weight initial information more strongly. Both notions come with quantitative parameters that smoothly interpolate between the non-adaptive case and the fully adaptive case, allowing for a rich spectrum of intermediate analysts that are neither non-adaptive nor adversarial.   Natural not only from a cognitive perspective, we show that our notions also capture standard optimization methods, like gradient descent in various settings. This gives a new interpretation to the fact that gradient descent tends to overfit much less than its adaptive nature might suggest."
                    },
                    {
                        "title": "Inherent Trade-Offs between Diversity and Stability in Multi-Task Benchmarks",
                        "abstract": "We examine multi-task benchmarks in machine learning through the lens of social choice theory. We draw an analogy between benchmarks and electoral systems, where models are candidates and tasks are voters. This suggests a distinction between cardinal and ordinal benchmark systems. The former aggregate numerical scores into one model ranking; the latter aggregate rankings for each task. We apply Arrow's impossibility theorem to ordinal benchmarks to highlight the inherent limitations of ordinal systems, particularly their sensitivity to the inclusion of irrelevant models. Inspired by Arrow's theorem, we empirically demonstrate a strong trade-off between diversity and sensitivity to irrelevant changes in existing multi-task benchmarks. Our result is based on new quantitative measures of diversity and sensitivity that we introduce. Sensitivity quantifies the impact that irrelevant changes to tasks have on a benchmark. Diversity captures the degree of disagreement in model rankings across tasks. We develop efficient approximation algorithms for both measures, as exact computation is computationally challenging. Through extensive experiments on seven cardinal benchmarks and eleven ordinal benchmarks, we demonstrate a clear trade-off between diversity and stability: The more diverse a multi-task benchmark, the more sensitive to trivial changes it is. Additionally, we show that the aggregated rankings of existing benchmarks are highly unstable under irrelevant changes. The codes and data are available at https://socialfoundations.github.io/benchbench/."
                    }
                ]
            }
        }
    },
    "2312.09841": {
        "paper_data": {
            "title": "Monoculture in Matching Markets",
            "url": "http://arxiv.org/abs/2312.09841v1",
            "arxiv_id": "2312.09841",
            "authors": [
                "Kenny Peng",
                "Nikhil Garg"
            ],
            "abstract": "Algorithmic monoculture arises when many decision-makers rely on the same algorithm to evaluate applicants. An emerging body of work investigates possible harms of this kind of homogeneity, but has been limited by the challenge of incorporating market effects in which the preferences and behavior of many applicants and decision-makers jointly interact to determine outcomes.   Addressing this challenge, we introduce a tractable theoretical model of algorithmic monoculture in a two-sided matching market with many participants. We use the model to analyze outcomes under monoculture (when decision-makers all evaluate applicants using a common algorithm) and under polyculture (when decision-makers evaluate applicants independently). All else equal, monoculture (1) selects less-preferred applicants when noise is well-behaved, (2) matches more applicants to their top choice, though individual applicants may be worse off depending on their value to decision-makers and risk tolerance, and (3) is more robust to disparities in the number of applications submitted.",
            "introduction": "   1 Introduction  What happens when many decision-makers\u2014in consequential domains like employment and college admissions\u2014all evaluate applicants using the same algorithm? This possibility has been termed \u201calgorithmic monoculture\u201d\u2014a name that references the problems arising in agriculture when a single crop is farmed extensively, amplifying the risk of correlated widespread failures. A recent line of work considers how a lack of variety in algorithms may warrant similar concerns.   Initial inquiries into the consequences of algorithmic monoculture have identified potential harms to both decision-makers and applicants. In seminal work initiating the formal study of algorithmic monoculture, Kleinberg and Raghavan (2021) demonstrated that firms relying on a common algorithm may hire weaker applicants than when using idiosyncratic (but less individually accurate) hiring processes. On the applicant side, an emerging set of work highlights the concern of systemic exclusion, which occurs when an applicant is denied from all opportunities (Creel and Hellman, 2022; Bommasani et\u00a0al., 2022; Toups et\u00a0al., 2023; Jain et\u00a0al., 2023).   This emerging literature has been limited by the challenge of incorporating market effects. Domains where monoculture may be especially impactful\u2014such as employment and college admissions\u2014share some common properties. First, participants on both sides of the market are in competition: workers compete over a limited number of job openings, and firms compete over workers who can each accept only one offer. Second, outcomes are determined by the preferences of participants on both sides: roughly speaking, a match is formed only if both a worker and firm prefer each other over their other options. Analyzing these interactions across many participants can be theoretically challenging; indeed, the existing literature considers neither markets with many participants nor two-sided interactions.   The technical contribution of our paper is a matching markets model that can be used to study algorithmic monoculture. In particular, we analyze equilibrium outcomes (i.e., stable matchings) that emerge under monoculture (when firms each use the same method to evaluate applicants) and polyculture (when firms use idiosyncratic methods to evaluate applicants). The model builds directly on the popular continuum model formalized by Azevedo and Leshno (2016), in which stable matchings are characterized by a cutoff structure in which applicants are matched to their favorite firm among those they can \u201cafford\u201d\u2014i.e., among the firms at which the applicant\u2019s score exceeds the firm\u2019s cutoff. In the model we introduce, an applicant\u2019s score is equal to an estimated value that is a noisy realization of their true value. Under monoculture, all firms use a single shared estimate of an applicant\u2019s value; under polyculture, firms obtain separate, independently drawn estimates of the applicant\u2019s value.   We show how modeling market effects with many participants advances the present understanding of monoculture in several ways, with Figure\u00a01 illustrating our first two theoretical results.   \u2022  First, as long as noise is well-behaved (i.e., with lighter-than-exponential tail), essentially only the most preferred applicants are hired under polyculture when there are many firms. This shows that the existing (relatively modest) effect found by Kleinberg and Raghavan (2021) with two firms can be fully strengthened when considering many firms. Our result gives a \u201cwisdom of the crowds\u201d-like result, where even when individual firms are noisy, the",
            "references": [
                {
                    "title": "Ecosystem-level Analysis of Deployed Machine Learning Reveals Homogeneous Outcomes",
                    "abstract": "Machine learning is traditionally studied at the model level: researchers measure and improve the accuracy, robustness, bias, efficiency, and other dimensions of specific models. In practice, the societal impact of machine learning is determined by the surrounding context of machine learning deployments. To capture this, we introduce ecosystem-level analysis: rather than analyzing a single model, we consider the collection of models that are deployed in a given context. For example, ecosystem-level analysis in hiring recognizes that a job candidate's outcomes are not only determined by a single hiring algorithm or firm but instead by the collective decisions of all the firms they applied to. Across three modalities (text, images, speech) and 11 datasets, we establish a clear trend: deployed machine learning is prone to systemic failure, meaning some users are exclusively misclassified by all models available. Even when individual models improve at the population level over time, we find these improvements rarely reduce the prevalence of systemic failure. Instead, the benefits of these improvements predominantly accrue to individuals who are already correctly classified by other models. In light of these trends, we consider medical imaging for dermatology where the costs of systemic failure are especially high. While traditional analyses reveal racial performance disparities for both models and humans, ecosystem-level analysis reveals new forms of racial disparity in model predictions that do not present in human predictions. These examples demonstrate ecosystem-level analysis has unique strengths for characterizing the societal impact of machine learning."
                },
                {
                    "title": "Improved Bayes Risk Can Yield Reduced Social Welfare Under Competition",
                    "abstract": "As the scale of machine learning models increases, trends such as scaling laws anticipate consistent downstream improvements in predictive accuracy. However, these trends take the perspective of a single model-provider in isolation, while in reality providers often compete with each other for users. In this work, we demonstrate that competition can fundamentally alter the behavior of these scaling trends, even causing overall predictive accuracy across users to be non-monotonic or decreasing with scale. We define a model of competition for classification tasks, and use data representations as a lens for studying the impact of increases in scale. We find many settings where improving data representation quality (as measured by Bayes risk) decreases the overall predictive accuracy across users (i.e., social welfare) for a marketplace of competing model-providers. Our examples range from closed-form formulas in simple settings to simulations with pretrained representations on CIFAR-10. At a conceptual level, our work suggests that favorable scaling trends for individual model-providers need not translate to downstream improvements in social welfare in marketplaces with multiple model providers."
                },
                {
                    "title": "Picking on the Same Person: Does Algorithmic Monoculture lead to Outcome Homogenization?",
                    "abstract": "As the scope of machine learning broadens, we observe a recurring theme of algorithmic monoculture: the same systems, or systems that share components (e.g. training data), are deployed by multiple decision-makers. While sharing offers clear advantages (e.g. amortizing costs), does it bear risks? We introduce and formalize one such risk, outcome homogenization: the extent to which particular individuals or groups experience negative outcomes from all decision-makers. If the same individuals or groups exclusively experience undesirable outcomes, this may institutionalize systemic exclusion and reinscribe social hierarchy. To relate algorithmic monoculture and outcome homogenization, we propose the component-sharing hypothesis: if decision-makers share components like training data or specific models, then they will produce more homogeneous outcomes. We test this hypothesis on algorithmic fairness benchmarks, demonstrating that sharing training data reliably exacerbates homogenization, with individual-level effects generally exceeding group-level effects. Further, given the dominant paradigm in AI of foundation models, i.e. models that can be adapted for myriad downstream tasks, we test whether model sharing homogenizes outcomes across tasks. We observe mixed results: we find that for both vision and language settings, the specific methods for adapting a foundation model significantly influence the degree of outcome homogenization. We conclude with philosophical analyses of and societal challenges for outcome homogenization, with an eye towards implications for deployed machine learning systems."
                },
                {
                    "title": "Statistical Discrimination in Stable Matchings",
                    "abstract": "Statistical discrimination results when a decision-maker observes an imperfect estimate of the quality of each candidate dependent on which demographic group they belong to [1,8]. Imperfect estimates have been modelled via noise, where the variance depends on the candidate's group ([4,6,7]). Prior literature, however, is limited to simple selection problems, where a single decision-maker tries to choose the best candidates among the applications they received. In this paper, we initiate the study of statistical discrimination in matching, where multiple decision-makers are simultaneously facing selection problems from the same pool of candidates. We consider the college admission problem as first introduced in [5] and recently extended to a model with a continuum of students [3]. We propose a model where two colleges A and B observe noisy estimates of each candidate's quality, where Ws, the vector of estimates for student s, is assumed to be a bivariate normal random variable. In this setting, the estimation noise controls a new key feature of the problem, namely correlation, \u03c1, between the estimates of the two colleges: if the noise is high, the correlation is low and if the noise is low the correlation is high. We assume that the population of students is divided into two groups G1 and G2, and that members of these two groups are subject to different correlation levels between their grades at colleges A and B. Concretely, for each student s, their grade vector (WAs, WBs) is drawn according to a centered bivariate normal distribution with variance 1 and covariance \u03c1Gs, where Gs is the group student s belongs to. We consider the stable matching induced by this distribution and characterize how key outcome characteristics vary with the parameters, in particular with the group-dependent correlation coefficient. Our results summarize as follows: We show that the probability that a student is assigned to their first choice is independent of the student's group, but that it decreases when the correlation of either group decreases. This means that higher measurement noise (inducing lower correlation) on one group hurts not only the students of that group, but the students of all groups. We show that the probability that a student is assigned to their second choice and the probability that they remain unassigned both depend on the student's group, which reveals the presence of statistical discrimination coming from the correlation effect alone. Specifically, we find that the probability that a student remains unmatched is decreasing when the correlation of their group decreases (higher measurement noise) and when the correlation of the other group increases. In other words, the higher the measurement noise of their own group, the better off students are with regard to getting assigned a college at all. This is somewhat counter-intuitive, but is explained by the observation that with high noise (i.e., low correlation) the fact that a student is rejected from one college gives only little information about the outcome at the other college. That is, a student has an independent second chance for admission. These two comparative static results give insights on the effect of correlation on the stable matching outcome for different demographic groups and show that indeed, statistical discrimination is an important theory to understand discrimination in matching problems. We also analyze a number of special cases of our model, in particular the case of a single group, to show that even in this case correlation affects the outcome. It is interesting to notice that the effect of correlation on the number of students getting their first choice in our model is the same as in [2], i.e., a higher correlation leads to more students getting their first choice. Our work is the first to investigate statistical discrimination in the context of matching. Overall we find that group-dependent measurement noises of the candidates quality---and the resulting group-dependent correlation between the colleges' estimates---play an important role in leading to unequal outcomes for different demographic groups, and in particular underrepresentation of one of the groups. Of course, we do not argue that statistical discrimination is the only possible cause of discrimination. In particular, if there is bias in the quality estimates for one group, then it will naturally also hurt the representation of that group. We do not model bias since our primary purpose is to isolate the effect of statistical discrimination. Throughout the paper, we make a number of other simplifying assumptions (e.g., focusing on two colleges) whose purpose is also to simplify our results and isolate the effect of correlation. Our analysis, however, extends to more general contexts."
                },
                {
                    "title": "Lottery Design for School Choice",
                    "abstract": "This paper studies outcomes of the deferred acceptance algorithm in large random matching markets where priorities are generated either by a single lottery or by independent lotteries. In contrast to prior work, my model permits students to submit lists of varying lengths and schools to vary in their popularity and number of seats. In a limiting regime where the number of students and schools grow while the length of student lists and number of seats at each school remain bounded, I provide exact expressions for the number of students who list l schools and match to one of their top k choices, for each [Formula: see text]. These expressions provide three main insights. First, there is a persistent tradeoff between using a single lottery and independent lotteries. For students who submit short lists, the rank distribution under a single lottery stochastically dominates the corresponding distribution under independent lotteries. However, the students who submit the longest lists are always more likely to match when schools use independent lotteries. Second, I compare the total number of matches in the two lottery systems, and find that the shape of the list length distribution plays a key role. If this distribution has an increasing hazard rate, then independent lotteries match more students. If it has a decreasing hazard rate, the comparison reverses. To my knowledge, this is the first analytical result comparing the size of stable matchings under different priority rules. Finally, I study the fraction of assigned students who receive their first choice. Under independent lotteries, this fraction may be arbitrarily small, even if schools are equally popular. Under a single lottery, we provide a tight lower bound on this fraction which depends on the ratio r of the popularity of the most to least popular school. When each school has a single seat, the fraction of assigned students who receive their first choice is at least [Formula: see text]. This guarantee increases to [Formula: see text] as the number of seats at each school increases. This paper was accepted by Gabriel Weintraub, revenue management and market analytics."
                },
                {
                    "title": "A Continuum Model of Stable Matching with Finite Capacities",
                    "abstract": "This paper introduces a unified framework for stable matching, which nests the traditional definition of stable matching in finite markets and the continuum definition of stable matching from Azevedo and Leshno (2016) as special cases. Within this framework, I identify a novel continuum model, which makes individual-level probabilistic predictions. This new model always has a unique stable outcome, which can be found using an analog of the Deferred Acceptance algorithm. The crucial difference between this model and that of Azevedo and Leshno (2016) is that they assume that the amount of student interest at each school is deterministic, whereas my proposed alternative assumes that it follows a Poisson distribution. As a result, this new model accurately predicts the simulated distribution of cutoffs, even for markets with only ten schools and twenty students. This model generates new insights about the number and quality of matches. When schools are homogeneous, it provides upper and lower bounds on students' average rank, which match results from Ashlagi, Kanoria and Leshno (2017) but apply to more general settings. This model also provides clean analytical expressions for the number of matches in a platform pricing setting considered by Marx and Schummer (2021)."
                },
                {
                    "title": "Incentives in Two-sided Matching Markets with Prediction-enhanced Preference-formation",
                    "abstract": "Two-sided matching markets have long existed to pair agents in the absence of regulated exchanges. A common example is school choice, where a matching mechanism uses student and school preferences to assign students to schools. In such settings, forming preferences is both difficult and critical. Prior work has suggested various prediction mechanisms that help agents make decisions about their preferences. Although often deployed together, these matching and prediction mechanisms are almost always analyzed separately. The present work shows that at the intersection of the two lies a previously unexplored type of strategic behavior: agents returning to the market (e.g., schools) can attack future predictions by interacting short-term non-optimally with their matches. Here, we first introduce this type of strategic behavior, which we call an `adversarial interaction attack'. Next, we construct a formal economic model that captures the feedback loop between prediction mechanisms designed to assist agents and the matching mechanism used to pair them. This economic model allows us to analyze adversarial interaction attacks. Finally, using school choice as an example, we build a simulation to show that, as the trust in and accuracy of predictions increases, schools gain progressively more by initiating an adversarial interaction attack. We also show that this attack increases inequality in the student population."
                },
                {
                    "title": "Test-optional Policies: Overcoming Strategic Behavior and Informational Gaps",
                    "abstract": "Due to the Covid-19 pandemic, more than 500 US-based colleges and universities went \u201ctest-optional\u201d for admissions and promised that they would not penalize applicants for not submitting test scores, part of a longer trend to rethink the role of testing in college admissions. However, it remains unclear how (and whether) a college can simultaneously use test scores for those who submit them, while not penalizing those who do not\u2013and what that promise even means. We formalize these questions, and study how a college can overcome two challenges with optional testing: strategic applicants (when those with low test scores can pretend to not have taken the test), and informational gaps (it has more information on those who submit a test score than those who do not). We find that colleges can indeed do so, if and only if they are able to use information on who has test access and are willing to randomize admissions."
                },
                {
                    "title": "Decentralized Matching in a Probabilistic Environment",
                    "abstract": "We consider a model for repeated stochastic matching where compatibility is probabilistic, is realized the first time agents are matched, and persists in the future. Such a model has applications in the gig economy, kidney exchange, and mentorship matching. We ask whether adecentralized matching process can approximate the optimal online algorithm. In particular, we consider a decentralizedstable matching process where agents match with the most compatible partner who does not prefer matching with someone else, and known compatible pairs continue matching in all future rounds. We demonstrate that the above process provides a 0.316-approximation to the optimal online algorithm for matching on general graphs. We also provide a 1/7-approximation for many-to-one bipartite matching, a 1/11-approximation for capacitated matching on general graphs, and a 1/2k-approximation for forming teams of up to k agents. Our results rely on a novel coupling argument that decomposes the successful edges of the optimal online algorithm in terms of their round-by-round comparison with stable matching."
                },
                {
                    "title": "The Algorithmic Leviathan: Arbitrariness, Fairness, and Opportunity in Algorithmic Decision Making Systems",
                    "abstract": "Automated decision-making systems implemented in public life are typically standardized. One algorithmic decision-making system can replace thousands of human deciders. Each of the humans so replaced had her own decision-making criteria: some good, some bad, and some arbitrary. Is such arbitrariness of moral concern? We argue that an isolated arbitrary decision need not morally wrong the individual whom it misclassifies. However, if the same algorithms are applied across a public sphere, such as hiring or lending, a person could be excluded from a large number of opportunities. This harm persists even when the automated decision-making systems are \"fair\" on standard metrics of fairness. We argue that such arbitrariness at scale is morally problematic and propose technically informed solutions that can lessen the impact of algorithms at scale and so mitigate or avoid the moral harms we identify."
                },
                {
                    "title": "Best vs. All: Equity and Accuracy of Standardized Test Score Reporting",
                    "abstract": "We study a game theoretic model of standardized testing for college admissions. Students are of two types; High and Low. There is a college that would like to admit the High type students. Students take a potentially costly standardized exam which provides a noisy signal of their type. The students come from two populations, which are identical in talent (i.e. the type distribution is the same), but differ in their access to resources: the higher resourced population can at their option take the exam multiple times, whereas the lower resourced population can only take the exam once. We study two models of score reporting, which capture existing policies used by colleges. The first policy (sometimes known as \u201csuper-scoring\u201d) allows students to report the max of the scores they achieve. The other policy requires that all scores be reported. We find in our model that requiring that all scores be reported results in superior outcomes in equilibrium, both from the perspective of the college (the admissions rule is more accurate), and from the perspective of equity across populations: a student\u2019s probability of admission is independent of their population, conditional on their type. In particular, the false positive rates and false negative rates are identical in this setting, across the highly and poorly resourced student populations. This is the case despite the fact that the more highly resourced students can\u2014at their option\u2014either report a more accurate signal of their type, or pool with the lower resourced population under this policy. This represents an unusual situation in the algorithmic fairness literature where the goals of accuracy and equity are in alignment, and do not need to be traded off against one another."
                },
                {
                    "title": "Algorithmic monoculture and social welfare",
                    "abstract": "Significance Algorithmic monoculture is a growing concern in the use of algorithms for high-stakes screening decisions in areas such as employment and lending. If many firms use the same algorithm, even if it is more accurate than the alternatives, the resulting \u201cmonoculture\u201d may be susceptible to correlated failures, much as a monocultural system is in biological settings. To investigate this concern, we develop a model of selection under monoculture. We find that even without any assumption of shocks or correlated failures\u2014i.e., under \u201cnormal operations\u201d\u2014the quality of decisions may decrease when multiple firms use the same algorithm. Thus, the introduction of a more accurate algorithm may decrease social welfare\u2014a kind of \u201cBraess\u2019 paradox\u201d for algorithmic decision-making. As algorithms are increasingly applied to screen applicants for high-stakes decisions in employment, lending, and other domains, concerns have been raised about the effects of algorithmic monoculture, in which many decision-makers all rely on the same algorithm. This concern invokes analogies to agriculture, where a monocultural system runs the risk of severe harm from unexpected shocks. Here, we show that the dangers of algorithmic monoculture run much deeper, in that monocultural convergence on a single algorithm by a group of decision-making agents, even when the algorithm is more accurate for any one agent in isolation, can reduce the overall quality of the decisions being made by the full collection of agents. Unexpected shocks are therefore not needed to expose the risks of monoculture; it can hurt accuracy even under \u201cnormal\u201d operations and even for algorithms that are more accurate when used by only a single decision-maker. Our results rely on minimal assumptions and involve the development of a probabilistic framework for analyzing systems that use multiple noisy estimates of a set of alternatives."
                },
                {
                    "title": "Bandit Learning in Decentralized Matching Markets",
                    "abstract": "We study two-sided matching markets in which one side of the market (the players) does not have a priori knowledge about its preferences for the other side (the arms) and is required to learn its preferences from experience. Also, we assume the players have no direct means of communication. This model extends the standard stochastic multi-armed bandit framework to a decentralized multiple player setting with competition. We introduce a new algorithm for this setting that, over a time horizon $T$, attains $\\mathcal{O}(\\log(T))$ stable regret when preferences of the arms over players are shared, and $\\mathcal{O}(\\log(T)^2)$ regret when there are no assumptions on the preferences on either side."
                },
                {
                    "title": "Learning Strategies in Decentralized Matching Markets under Uncertain Preferences",
                    "abstract": "We study two-sided decentralized matching markets in which participants have uncertain preferences. We present a statistical model to learn the preferences. The model incorporates uncertain state and the participants' competition on one side of the market. We derive an optimal strategy that maximizes the agent's expected payoff and calibrate the uncertain state by taking the opportunity costs into account. We discuss the sense in which the matching derived from the proposed strategy has a stability property. We also prove a fairness property that asserts that there exists no justified envy according to the proposed strategy. We provide numerical results to demonstrate the improved payoff, stability and fairness, compared to alternative methods."
                },
                {
                    "title": "Standardized Tests and Affirmative Action: The Role of Bias and Variance",
                    "abstract": "The University of California suspended through 2024 the requirement that applicants from California submit SAT scores, upending the major role standardized testing has played in college admissions. We study the impact of such decisions and its interplay with other policies---such as affirmative action---on admitted class composition. This paper considers a theoretical framework to study the effect of requiring test scores on academic merit and diversity in college admissions. The model has a college and set of potential students. Each student has observed application components and group membership, as well as an unobserved noisy skill level generated from an observed distribution. The college is Bayesian and maximizes an objective that depends on both diversity and merit. It estimates each applicant's true skill level using the observed features and potentially their group membership, and then admits students with or without affirmative action. We characterize the trade-off between the (potentially positive) informational role of standardized testing in college admissions and its (negative) exclusionary nature. Dropping test scores may exacerbate disparities by decreasing the amount of information available for each applicant, especially those from non-traditional backgrounds. However, if there are substantial barriers to testing, removing the test improves both academic merit and diversity by increasing the size of the applicant pool. Finally, using application and transcript data from the University of Texas at Austin, we demonstrate how an admissions committee could measure the trade-off in practice to better decide whether to drop their test scores requirement. The full paper can be found at https://arxiv.org/abs/2010.04396."
                },
                {
                    "title": "Information Acquisition in Matching Markets: The Role of Price Discovery",
                    "abstract": "We explore the acquisition and flow of information in matching markets through a model of college admissions with endogenous costly information acquisition. We extend the notion of stability to this partial information setting, and introduce regret-free stability as a refinement that additionally requires optimal student information acquisition. We show regret-free stable outcomes exist, and finding them is equivalent to finding appropriately-defined market-clearing cutoffs. \n \nTo understand information flows, we recast matching mechanisms as price-discovery processes. No mechanism guarantees a regret-free stable outcome, because information deadlocks imply some students must acquire information sub optimally. Our analysis suggests approaches for facilitating efficient price discovery, leveraging historical information or market sub-samples to estimate cutoffs. We show that mechanisms that use such methods to advise applicants on their admission chances yield approximately regret-free stable outcomes. A survey of university admission systems highlights the practical importance of providing applicants with information about their admission chances."
                },
                {
                    "title": "Competing Bandits in Matching Markets",
                    "abstract": "Stable matching, a classical model for two-sided markets, has long been studied with little consideration for how each side's preferences are learned. With the advent of massive online markets powered by data-driven matching platforms, it has become necessary to better understand the interplay between learning and market objectives. We propose a statistical learning model in which one side of the market does not have a priori knowledge about its preferences for the other side and is required to learn these from stochastic rewards. Our model extends the standard multi-armed bandits framework to multiple players, with the added feature that arms have preferences over players. We study both centralized and decentralized approaches to this problem and show surprising exploration-exploitation trade-offs compared to the single player multi-armed bandits setting."
                },
                {
                    "title": "Unbalanced Random Matching Markets: The Stark Effect of Competition",
                    "abstract": "We study competition in matching markets with random heterogeneous preferences and an unequal number of agents on either side. First, we show that even the slightest imbalance yields an essentially unique stable matching. Second, we give a tight description of stable outcomes, showing that matching markets are extremely competitive. Each agent on the short side of the market is matched with one of his top choices, and each agent on the long side either is unmatched or does almost no better than being matched with a random partner. Our results suggest that any matching market is likely to have a small core, explaining why small cores are empirically ubiquitous."
                },
                {
                    "title": "The Performance of School Assignment Mechanisms in Practice",
                    "abstract": "We use a unique combination of register and survey data from Amsterdam to investigate the performance of school assignment mechanisms in practice. We find that deferred acceptance (DA) results in higher mean welfare than the adaptive Boston mechanism. This is due to students making strategic mistakes. The welfare gain of a switch from actual Boston to DA is over 90% of the welfare difference between actual Boston and optimal (proxy) Boston. Disadvantaged and lower-ability students would benefit most from such a switch."
                },
                {
                    "title": "A Supply and Demand Framework for Two-Sided Matching Markets",
                    "abstract": "This paper develops a price-theoretic framework for matching markets with heterogeneous preferences. The model departs from the Gale and Shapley model by assuming that a finite number of agents on one side (colleges) are matched to a continuum of agents on the other side (students). We show that stable matchings correspond to solutions of supply and demand equations, with the selectivity of each college playing a role similar to that of prices. We apply the model to an analysis of how competition induced by school choice gives schools incentives to invest in quality and to asymptotics of school choice mechanisms."
                },
                {
                    "title": "College Admissions and the Stability of Marriage",
                    "abstract": "A procedure for assigning applicants to colleges which removes all uncertainties and, assuming there are enough applicants, assigns to each college precisely its quota."
                },
                {
                    "title": "Strategy-Proofness versus Efficiency in Matching with Indifferences: Redesigning the NYC High School Match",
                    "abstract": "The design of the New York City (NYC) High School match involved tradeoffs among efficiency, stability and strategy-proofness that raise new theoretical questions. We analyze a model with indifferences--ties--in school preferences. Simulations with field data and the theory favor breaking indifferences the same way at every school --single tie breaking-- in a student-proposing deferred acceptance mechanism. Any inefficiency associated with a realized tie breaking cannot be removed without harming student incentives. Finally, we empirically document the extent of potential efficiency loss associated with strategy-proofness and stability, and direct attention to some open questions."
                },
                {
                    "title": "Expanding 'Choice' in School Choice",
                    "abstract": "Truthful revelation of preferences has emerged as a desideratum in the design of school choice programs. Gale-Shapley's deferred acceptance mechanism is strategy-proof for students but limits their ability to communicate their preference intensities. This results in ex-ante inefficiency when ties at school preferences are broken randomly. We propose a variant of deferred acceptance mechanism which allows students to influence how they are treated in ties. It maintains truthful revelation of ordinal preferences and supports a greater scope of efficiency."
                },
                {
                    "title": "Matching and Sorting in Online Dating",
                    "abstract": "Using data on user attributes and interactions from an online dating site, we estimate mate preferences, and use the Gale-Shapley algorithm to predict sta ble matches. The predicted matches are similar to the actual matches achieved by the dating site, and the actual matches are approximately efficient. Outof-sample predictions of offline matches, i.e., marriages, exhibit assortative mating patterns similar to those observed in actual marriages. Thus, mate pref erences, without resort to search frictions, can generate sorting in marriages. However, we underpredict some of the correlation patterns; search frictions may play a role in explaining the discrepancy. ("
                },
                {
                    "title": "Simultaneous Search",
                    "abstract": "We introduce and solve a new class of \"downward-recursive\" static portfolio choice problems. An individual simultaneously chooses among ranked stochastic options, and each choice is costly. In the motivational application, just one may be exercised from those that succeed. This often emerges in practice, such as when a student applies to many colleges. We show that a greedy algorithm finds the optimal set. The optimal choices are \"less aggressive\" than the sequentially optimal ones, but \"more aggressive\" than the best singletons. The optimal set in general contains gaps. We provide a comparative static on the chosen set."
                },
                {
                    "title": "Marriage, honesty, and stability",
                    "abstract": "Many centralized two-sided markets form a matching between participants by running a stable marriage algorithm. It is a well-known fact that no matching mechanism based on a stable marriage algorithm can guarantee truthfulness as a dominant strategy for participants. However, as we will show in this paper, in a probabilistic setting where the preference lists of one side of the market are composed of only a constant (independent of the the size of the market) number of entries, each drawn from an arbitrary distribution, the number of participants that have more than one stable partner is vanishingly small. This proves (and generalizes) a conjecture of Roth and Peranson [23]. As a corollary of this result, we show that, with high probability, the truthful strategy is the best response for a given player when the other players are truthful. We also analyze equilibria of the deferred acceptance stable marriage game. We show that the game with complete information has an equilibrium in which a (1 - o(1)) fraction of the strategies are truthful in expectation. In the more realistic setting of a game of incomplete information, we will show that the set of truthful strategies form a (1 + o(1))-approximate Bayesian-Nash equilibrium. Our results have implications in many practical settings and were inspired by the work of Roth and Peranson [23] on the National Residency Matching Program."
                },
                {
                    "title": "School Choice: A Mechanism Design Approach",
                    "abstract": "A central issue in school choice is the design of a student assignment mechanism. Education literature provides guidance for the design of such mechanisms but does not offer specific mechanisms. The flaws in the existing school choice plans result in appeals by unsatisfied parents. We formulate the school choice problem as a mechanism design problem and analyze some of the existing school choice plans including those in Boston, Columbus, Minneapolis, and Seattle. We show that these existing plans have serious shortcomings, and offer two alternative mechanisms each of which may provide a practical solution to some critical school choice issues."
                },
                {
                    "title": "The Average Number of Stable Matchings",
                    "abstract": "The probable behavior of an instance of size n of the stable marriage problem, chosen uniformly at random, is studied. The expected number of stable matchings is shown to be asymptotic to $e^{ - 1} n \\ln n$ for $n \\to \\infty $. The total rank of women by men in the male optimal (pessimal) matching is proved to be close to n In n (respectively, $n^2 $/$\\ln n$, with high probability."
                },
                {
                    "title": "Anomalies: The Winner's Curse",
                    "abstract": "Abstract Next time that you find yourself a little short of cash for lunch, try the following experiment in your class. Take a jar and fill it with coins, noting the total value of the coins. Now a..."
                },
                {
                    "title": "In Search of the Winner's Curse",
                    "abstract": "Earlier papers on the winner's curse have provided theoretical arguments that winning bidders in an auction will incur ex post losses even when all bidders use reasonable ex ante bidding strategies. This paper demonstrates that these arguments are erroneous: optimal ex ante bidders will never suffer from a winner's curse in an auction where only the winning bid is announced; furthermore, such bidders will on the average not suffer from a winner's curse in an auction where all bids are announced. Thus if a winner's curse is a behavioral reality then bidders are not generally using ex ante optimal strategies."
                },
                {
                    "title": "On rank dominance of tie\u2010breaking rules",
                    "abstract": "Lotteries are a common way to resolve ties in assignment mechanisms that ration resources. We consider a model with a continuum of agents and a finite set of resources with heterogeneous qualities, where the agents' preferences are generated from a multinomial\u2010logit (MNL) model based on the resource qualities. We show that all agents prefer a common lottery to independent lotteries at each resource if every resource is \n popular, meaning that the mass of agents ranking that resource as their first choice exceeds its capacity. We then prove a stronger result where the assumption that every resource is popular is not required and agents' preferences are drawn from a (more general) nested MNL model. By appropriately adapting the notion of popularity to resource types, we show that a hybrid tie\u2010breaking rule in which the objects in each popular type share a common lottery dominates independent lotteries at each resource.\n"
                },
                {
                    "title": "Theories of Statistical Discrimination and Affirmative Action : A Survey \u2217",
                    "abstract": "This chapter surveys the theoretical literature on statistical discrimination and affirmative action. This literature suggests different explanations for the existence and persistence of group inequality. This survey highlights such differences and describes in these contexts the effects of color-sighted and color-blind affirmative action policies, and the efficiency implications of discriminatory outcomes."
                },
                {
                    "title": "Constrained School Choice",
                    "abstract": "Recently, several school districts in the US have adopted or consider adopting the Student-Optimal Stable mechanism or the Top Trading Cycles mechanism to assign children to public schools. There is evidence that for school districts that employ (variants of) the so-called Boston mechanism the transition would lead to efficiency gains. The first two mechanisms are strategy-proof, but in practice student assignment procedures typically impede a student to submit a preference list that contains all his acceptable schools. We study the preference revelation game where students can only declare up to a fixed number of schools to be acceptable. We focus on the stability and efficiency of the Nash equilibrium outcomes. Our main results identify rather stringent necessary and sufficient conditions on the priorities to guarantee stability or efficiency of either of the two mechanisms. This stands in sharp contrast with the Boston mechanism which has been abandoned in many US school districts but nevertheless yields stable Nash equilibrium outcomes."
                }
            ],
            "categories": [
                "cs.GT",
                "cs.CY",
                "econ.TH"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does algorithmic monoculture in decision-making processes, such as employment and college admissions, impact the quality of applicant selection and the potential for systemic exclusion?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of algorithmic monoculture is crucial for the research community as it addresses the broader implications of fairness and efficiency in decision-making systems. By understanding how a lack of algorithmic diversity affects outcomes, future research can explore alternative evaluation methods that promote equity and improve decision quality. This work could lead to practical applications in designing algorithms that mitigate risks associated with systemic exclusion, ultimately fostering more inclusive environments in critical domains like hiring and admissions.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing algorithmic monoculture stem from the complexities of modeling interactions in two-sided markets with many participants. Naive approaches may fail because they do not account for the competitive dynamics between applicants and decision-makers, nor do they consider the implications of using a single evaluation method across multiple firms. Technical obstacles include accurately modeling the noise in applicant evaluations and understanding how equilibrium outcomes differ under monoculture versus polyculture conditions.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by a lack of comprehensive models that incorporate market effects and the interactions between multiple decision-makers and applicants. Existing studies have primarily focused on simplified scenarios with few participants, which do not capture the complexities of real-world decision-making environments. Our approach differs by introducing a matching markets model that accounts for many participants and explores the equilibrium outcomes under both monoculture and polyculture, thereby filling a significant gap in the literature.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a matching markets model that analyzes equilibrium outcomes in scenarios of algorithmic monoculture and polyculture. We will utilize a continuum model based on the work of Azevedo and Leshno (2016) to characterize stable matchings, focusing on how applicants' scores are evaluated under different conditions. The expected outcomes include a deeper understanding of how algorithmic diversity influences hiring practices and the identification of conditions under which polyculture leads to better applicant selection compared to monoculture. Metrics for evaluation will include the quality of matches and the incidence of systemic exclusion among applicants."
            }
        },
        "author_data": {
            "28448617-ee56-4e33-805b-8fe1b2f983d5": {
                "pk": "28448617-ee56-4e33-805b-8fe1b2f983d5",
                "name": "Kenny Peng",
                "collaborators": [
                    "Nikhil Garg",
                    "Evita Nestoridi",
                    "James Grimmelmann",
                    "Arunesh Mathur",
                    "Arvind Narayanan",
                    "Manish Raghavan",
                    "Emma Pierson",
                    "Jon Kleinberg"
                ],
                "domain": [
                    "Matching Theory",
                    "Misinformation",
                    "Ethical AI",
                    "Recommender Systems"
                ],
                "publications": [
                    {
                        "title": "Mixing times of one-sided $k$-transposition shuffles",
                        "abstract": "We study mixing times of the one-sided $k$-transposition shuffle. We prove that this shuffle mixes relatively slowly, even for $k$ big. Using the recent \"lifting eigenvectors\" technique of Dieker and Saliola and applying the $\\ell^2$ bound, we prove different mixing behaviors and explore the occurrence of cutoff depending on $k$."
                    },
                    {
                        "title": "Wisdom and Foolishness of Noisy Matching Markets",
                        "abstract": "We consider a many-to-one matching market where colleges share true preferences over students but make decisions using only independent noisy rankings. Each student has a true value $v$, but each college $c$ ranks the student according to an independently drawn estimated value $v + X_c$ for $X_c\\sim \\mathcal{D}.$ We ask a basic question about the resulting stable matching: How noisy is the set of matched students? Two striking effects can occur in large markets (i.e., with a continuum of students and a large number of colleges). When $\\mathcal{D}$ is light-tailed, noise is fully attenuated: only the highest-value students are matched. When $\\mathcal{D}$ is long-tailed, noise is fully amplified: students are matched uniformly at random. These results hold for any distribution of student preferences over colleges, and extend to when only subsets of colleges agree on true student valuations instead of the entire market. More broadly, our framework provides a tractable approach to analyze implications of imperfect preference formation in large markets."
                    },
                    {
                        "title": "Rescuing Counterspeech: A Bridging-Based Approach to Combating Misinformation",
                        "abstract": "Social media has a misinformation problem, and counterspeech -- fighting bad speech with more speech -- has been an ineffective solution. Here, we argue that bridging-based ranking -- an algorithmic approach to promoting content favored by users of diverse viewpoints -- is a promising approach to helping counterspeech combat misinformation. By identifying counterspeech that is favored both by users who are inclined to agree and by users who are inclined to disagree with a piece of misinformation, bridging promotes counterspeech that persuades the users most likely to believe the misinformation. Furthermore, this algorithmic approach leverages crowd-sourced votes, shifting discretion from platforms back to users and enabling counterspeech at the speed and scale required to combat misinformation online. Bridging is respectful of users' autonomy and encourages broad participation in healthy exchanges; it offers a way for the free speech tradition to persist in modern speech environments."
                    },
                    {
                        "title": "Mitigating Dataset Harms Requires Stewardship: Lessons from 1000 Papers",
                        "abstract": "Machine learning datasets have elicited concerns about privacy, bias, and unethical applications, leading to the retraction of prominent datasets such as DukeMTMC, MS-Celeb-1M, and Tiny Images. In response, the machine learning community has called for higher ethical standards in dataset creation. To help inform these efforts, we studied three influential but ethically problematic face and person recognition datasets -- Labeled Faces in the Wild (LFW), MS-Celeb-1M, and DukeMTM -- by analyzing nearly 1000 papers that cite them. We found that the creation of derivative datasets and models, broader technological and social change, the lack of clarity of licenses, and dataset management practices can introduce a wide range of ethical concerns. We conclude by suggesting a distributed approach to harm mitigation that considers the entire life cycle of a dataset."
                    },
                    {
                        "title": "Reconciling the accuracy-diversity trade-off in recommendations",
                        "abstract": "In recommendation settings, there is an apparent trade-off between the goals of accuracy (to recommend items a user is most likely to want) and diversity (to recommend items representing a range of categories). As such, real-world recommender systems often explicitly incorporate diversity separately from accuracy. This approach, however, leaves a basic question unanswered: Why is there a trade-off in the first place?   We show how the trade-off can be explained via a user's consumption constraints -- users typically only consume a few of the items they are recommended. In a stylized model we introduce, objectives that account for this constraint induce diverse recommendations, while objectives that do not account for this constraint induce homogeneous recommendations. This suggests that accuracy and diversity appear misaligned because standard accuracy metrics do not consider consumption constraints. Our model yields precise and interpretable characterizations of diversity in different settings, giving practical insights into the design of diverse recommendations."
                    }
                ]
            },
            "b000c39d-3fc4-41c8-9c4a-d3ffa092f8fc": {
                "pk": "b000c39d-3fc4-41c8-9c4a-d3ffa092f8fc",
                "name": "Nikhil Garg",
                "collaborators": [
                    "Ramesh Johari",
                    "Zhi Liu",
                    "James Henderson",
                    "Hamid Nazerzadeh",
                    "Kenny Peng",
                    "Sarabjot Singh",
                    "Jeffrey Andrews",
                    "Londa Schiebinger",
                    "Dan Jurafsky",
                    "James Zou"
                ],
                "domain": [
                    "Machine Learning",
                    "Bayesian Modeling",
                    "Social Choice Theory",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "A Bayesian Model of Multilingual Unsupervised Semantic Role Induction",
                        "abstract": "We propose a Bayesian model of unsupervised semantic role induction in multiple languages, and use it to explore the usefulness of parallel corpora for this task. Our joint Bayesian model consists of individual models for each language plus additional latent variables that capture alignments between roles across languages. Because it is a generative Bayesian model, we can do evaluations in a variety of scenarios just by varying the inference procedure, without changing the model, thereby comparing the scenarios directly. We compare using only monolingual data, using a parallel corpus, using a parallel corpus with annotations in the other language, and using small amounts of annotation in the target language. We find that the biggest impact of adding a parallel corpus to training is actually the increase in mono-lingual data, with the alignments to another language resulting in small improvements, even with labeled data for the other language."
                    },
                    {
                        "title": "Designing Optimal Binary Rating Systems",
                        "abstract": "Modern online platforms rely on effective rating systems to learn about items. We consider the optimal design of rating systems that collect binary feedback after transactions. We make three contributions. First, we formalize the performance of a rating system as the speed with which it recovers the true underlying ranking on items (in a large deviations sense), accounting for both items' underlying match rates and the platform's preferences. Second, we provide an efficient algorithm to compute the binary feedback system that yields the highest such performance. Finally, we show how this theoretical perspective can be used to empirically design an implementable, approximately optimal rating system, and validate our approach using real-world experimental data collected on Amazon Mechanical Turk."
                    },
                    {
                        "title": "Driver Surge Pricing",
                        "abstract": "Ride-hailing marketplaces like Uber and Lyft use dynamic pricing, often called surge, to balance the supply of available drivers with the demand for rides. We study driver-side payment mechanisms for such marketplaces, presenting the theoretical foundation that has informed the design of Uber's new additive driver surge mechanism. We present a dynamic stochastic model to capture the impact of surge pricing on driver earnings and their strategies to maximize such earnings. In this setting, some time periods (surge) are more valuable than others (non-surge), and so trips of different time lengths vary in the induced driver opportunity cost.   First, we show that multiplicative surge, historically the standard on ride-hailing platforms, is not incentive compatible in a dynamic setting. We then propose a structured, incentive-compatible pricing mechanism. This closed-form mechanism has a simple form and is well-approximated by Uber's new additive surge mechanism. Finally, through both numerical analysis and real data from a ride-hailing marketplace, we show that additive surge is more incentive compatible in practice than is multiplicative surge."
                    },
                    {
                        "title": "Designing Informative Rating Systems: Evidence from an Online Labor Market",
                        "abstract": "Platforms critically rely on rating systems to learn the quality of market participants. In practice, however, these ratings are often highly inflated, and therefore not very informative. In this paper, we first investigate whether the platform can obtain less inflated, more informative ratings by altering the meaning and relative importance of the levels in the rating system. Second, we seek a principled approach for the platform to make these choices in the design of the rating system. First, we analyze the results of a randomized controlled trial on an online labor market in which an additional question was added to the feedback form. Between treatment conditions, we vary the question phrasing and answer choices; in particular, the treatment conditions include several positive-skewed verbal rating scales with descriptive phrases or adjectives providing specific interpretation for each rating level. The online labor market test reveals that current inflationary norms can in fact be countered by re-anchoring the meaning of the levels of the rating system. In particular, the positive-skewed verbal rating scales yield rating distributions that significantly reduce rating inflation and are much more informative about seller quality. Second, we develop a model-based framework to compare and select among rating system designs, and apply this framework to the data obtained from the online labor market test. Our simulations demonstrate that our model-based framework for scale design and optimization can identify the most informative rating system and substantially improve the quality of information obtained over baseline designs. Overall, our study illustrates that rating systems that are informative in practice can be designed, and demonstrates how to design them in a principled manner."
                    },
                    {
                        "title": "Test-optional Policies: Overcoming Strategic Behavior and Informational Gaps",
                        "abstract": "Due to the Covid-19 pandemic, more than 500 US-based colleges and universities went \"test-optional\" for admissions and promised that they would not penalize applicants for not submitting test scores, part of a longer trend to rethink the role of testing in college admissions. However, it remains unclear how (and whether) a college can simultaneously use test scores for those who submit them, while not penalizing those who do not--and what that promise even means. We formalize these questions, and study how a college can overcome two challenges with optional testing: $\\textit{strategic applicants}$ (when those with low test scores can pretend to not have taken the test), and $\\textit{informational gaps}$ (it has more information on those who submit a test score than those who do not). We find that colleges can indeed do so, if and only if they are able to use information on who has test access and are willing to randomize admissions."
                    },
                    {
                        "title": "Wisdom and Foolishness of Noisy Matching Markets",
                        "abstract": "We consider a many-to-one matching market where colleges share true preferences over students but make decisions using only independent noisy rankings. Each student has a true value $v$, but each college $c$ ranks the student according to an independently drawn estimated value $v + X_c$ for $X_c\\sim \\mathcal{D}.$ We ask a basic question about the resulting stable matching: How noisy is the set of matched students? Two striking effects can occur in large markets (i.e., with a continuum of students and a large number of colleges). When $\\mathcal{D}$ is light-tailed, noise is fully attenuated: only the highest-value students are matched. When $\\mathcal{D}$ is long-tailed, noise is fully amplified: students are matched uniformly at random. These results hold for any distribution of student preferences over colleges, and extend to when only subsets of colleges agree on true student valuations instead of the entire market. More broadly, our framework provides a tractable approach to analyze implications of imperfect preference formation in large markets."
                    },
                    {
                        "title": "Redesigning Service Level Agreements: Equity and Efficiency in City Government Operations",
                        "abstract": "We consider government service allocation -- how the government allocates resources (e.g., maintenance of public infrastructure) over time. It is important to make these decisions efficiently and equitably -- though these desiderata may conflict. In particular, we consider the design of Service Level Agreements (SLA) in city government operations: promises that incidents such as potholes and fallen trees will be responded to within a certain time. We model the problem of designing a set of SLAs as an optimization problem with equity and efficiency objectives under a queuing network framework; the city has two decision levers: how to allocate response budgets to different neighborhoods, and how to schedule responses to individual incidents. We: (1) Theoretically analyze a stylized model and find that the \"price of equity\" is small in realistic settings; (2) Develop a simulation-optimization framework to optimize policies in practice; (3) Apply our framework empirically using data from NYC, finding that: (a) status quo inspections are highly inefficient and inequitable compared to optimal ones, and (b) in practice, the equity-efficiency trade-off is not substantial: generally, inefficient policies are inequitable, and vice versa."
                    },
                    {
                        "title": "Impact of Dual Slope Path Loss on User Association in HetNets",
                        "abstract": "Intelligent load balancing is essential to fully realize the benefits of dense heterogeneous networks. Current techniques have largely been studied with single slope path loss models, though multi-slope models are known to more closely match real deployments. This paper develops insight into the performance of biasing and uplink/downlink decoupling for user association in HetNets with dual slope path loss models. It is shown that dual slope path loss models change the tradeoffs inherent in biasing and reduce gains from both biasing and uplink/downlink decoupling. The results show that with the dual slope path loss models, the bias maximizing the median rate is not optimal for other users, e.g., edge users. Furthermore, optimal downlink biasing is shown to realize most of the gains from downlink-uplink decoupling. Moreover, the user association gains in dense networks are observed to be quite sensitive to the path loss exponent beyond the critical distance in a dual slope model."
                    },
                    {
                        "title": "Word Embeddings Quantify 100 Years of Gender and Ethnic Stereotypes",
                        "abstract": "Word embeddings use vectors to represent words such that the geometry between vectors captures semantic relationship between the words. In this paper, we develop a framework to demonstrate how the temporal dynamics of the embedding can be leveraged to quantify changes in stereotypes and attitudes toward women and ethnic minorities in the 20th and 21st centuries in the United States. We integrate word embeddings trained on 100 years of text data with the U.S. Census to show that changes in the embedding track closely with demographic and occupation shifts over time. The embedding captures global social shifts -- e.g., the women's movement in the 1960s and Asian immigration into the U.S -- and also illuminates how specific adjectives and occupations became more closely associated with certain populations over time. Our framework for temporal analysis of word embedding opens up a powerful new intersection between machine learning and quantitative social science."
                    },
                    {
                        "title": "Alzheimers Dementia Detection using Acoustic & Linguistic features and Pre-Trained BERT",
                        "abstract": "Alzheimers disease is a fatal progressive brain disorder that worsens with time. It is high time we have inexpensive and quick clinical diagnostic techniques for early detection and care. In previous studies, various Machine Learning techniques and Pre-trained Deep Learning models have been used in conjunction with the extraction of various acoustic and linguistic features. Our study focuses on three models for the classification task in the ADReSS (The Alzheimers Dementia Recognition through Spontaneous Speech) 2021 Challenge. We use the well-balanced dataset provided by the ADReSS Challenge for training and validating our models. Model 1 uses various acoustic features from the eGeMAPs feature-set, Model 2 uses various linguistic features that we generated from auto-generated transcripts and Model 3 uses the auto-generated transcripts directly to extract features using a Pre-trained BERT and TF-IDF. These models are described in detail in the models section."
                    },
                    {
                        "title": "Who is in Your Top Three? Optimizing Learning in Elections with Many Candidates",
                        "abstract": "Elections and opinion polls often have many candidates, with the aim to either rank the candidates or identify a small set of winners according to voters' preferences. In practice, voters do not provide a full ranking; instead, each voter provides their favorite K candidates, potentially in ranked order. The election organizer must choose K and an aggregation rule.   We provide a theoretical framework to make these choices. Each K-Approval or K-partial ranking mechanism (with a corresponding positional scoring rule) induces a learning rate for the speed at which the election correctly recovers the asymptotic outcome. Given the voter choice distribution, the election planner can thus identify the rate optimal mechanism. Earlier work in this area provides coarse order-of-magnitude guaranties which are not sufficient to make such choices. Our framework further resolves questions of when randomizing between multiple mechanisms may improve learning, for arbitrary voter noise models.   Finally, we use data from 5 large participatory budgeting elections that we organized across several US cities, along with other ranking data, to demonstrate the utility of our methods. In particular, we find that historically such elections have set K too low and that picking the right mechanism can be the difference between identifying the ultimate winner with only a 80% probability or a 99.9% probability after 400 voters."
                    }
                ]
            }
        }
    },
    "2403.10766": {
        "paper_data": {
            "title": "ODE Discovery for Longitudinal Heterogeneous Treatment Effects Inference",
            "url": "http://arxiv.org/abs/2403.10766v1",
            "arxiv_id": "2403.10766",
            "authors": [
                "Krzysztof Kacprzyk",
                "Samuel Holt",
                "Jeroen Berrevoets",
                "Zhaozhi Qian",
                "Mihaela van der Schaar"
            ],
            "abstract": "Inferring unbiased treatment effects has received widespread attention in the machine learning community. In recent years, our community has proposed numerous solutions in standard settings, high-dimensional treatment settings, and even longitudinal settings. While very diverse, the solution has mostly relied on neural networks for inference and simultaneous correction of assignment bias. New approaches typically build on top of previous approaches by proposing new (or refined) architectures and learning algorithms. However, the end result -- a neural-network-based inference machine -- remains unchallenged. In this paper, we introduce a different type of solution in the longitudinal setting: a closed-form ordinary differential equation (ODE). While we still rely on continuous optimization to learn an ODE, the resulting inference machine is no longer a neural network. Doing so yields several advantages such as interpretability, irregular sampling, and a different set of identification assumptions. Above all, we consider the introduction of a completely new type of solution to be our most important contribution as it may spark entirely new innovations in treatment effects in general. We facilitate this by formulating our contribution as a framework that can transform any ODE discovery method into a treatment effects method.",
            "introduction": " introduction, all the ODE assumptions in section 3; proposed and wrote the method in section 5, developed and wrote all the methods discover an ODE, and then at inference time, when we wish to determine future values, we forward simulate the initial observed value using the discovered ODE and this Euler numerical solver to estimate future values at future times. We also kindly highlight further implementation details can be found in appendices appendices F, G and I. 37 appendix F. We quote 95% confidence intervals with each value, and all metrics are averaged over five random seed runs. We modify the observation function g(x), to be one of the following g(x) = {xj,exp(xj), x2 j,sin(xj)}. Each PKPD underlying pharmacological model is simulated with different layers of BSV (table 2) for A-D. Our contribution is shaded below. g(x) Method eq. (5).A eq. (5).B eq. (5).C eq. (5).D eq. (6).A eq. (6).B eq. (6).C eq. (6).D Cancer PKPD xA-SINDy 0.11 \u00b10.00 0.11\u00b10.00 0.13\u00b10.00 0.15\u00b10.00 1.46\u00b10.00 1.47\u00b10.00 1.40\u00b10.00 1.56\u00b10.00 1.29\u00b10.00 INSITE 0.02\u00b10.00 0.03\u00b10.00 0.04\u00b10.00 0.05\u00b10.00 0.96\u00b10.00 0.96\u00b10.00 0.82 \u00b11.54e- 160.90\u00b10.00 0.79\u00b10.00 exp(x)A-SINDy 0.75 \u00b10.02 0.75\u00b10.02 0.74\u00b10.03 0.76\u00b10.06 0.76\u00b10.34 0.76\u00b10.34 1.21\u00b11.34 1.02\u00b10.57 0.94\u00b10.20 INSITE 0.59\u00b10.03 0.59\u00b10.02 0.61\u00b10.02 0.62\u00b10.05 0.67\u00b10.30 0.68\u00b10.32 1.08\u00b11.31 0.90\u00b10.59 0.78\u00b10.13 x2 A-SINDy 0.10 \u00b15.85e-03 0.12\u00b13.50e-03 0.12\u00b13.63e-03 0.12\u00b10.01 1.16\u00b11.03 1.15\u00b11.03 2.11\u00b13.58 1.69\u00b12.02 0.57\u00b10.10 INSITE 0.03 \u00b11.98e- 030.09 \u00b12.67e- 030.09 \u00b12.40e- 030.09 \u00b16.26e- 031.12\u00b11.04 1.11\u00b11.04 2.04\u00b13.57 1.63\u00b12.03 0.52\u00b10.08 sin(x)A-SINDy 2.42 \u00b10.07 2.42\u00b10.07 2.31\u00b10.07 2.42\u00b10.40 1.62\u00b10.43 1.62\u00b10.43 2.30\u00b11.86 1.99\u00b10.69 2.00\u00b10.46 INSITE 2.09\u00b10.10 2.09\u00b10.10 2.06\u00b10.08 2.22\u00b10.25 1.46\u00b10.36 1.41\u00b10.42 1.96\u00b11.78 1.72\u00b10.76 1.59\u00b10.29 34Published as a conference paper at ICLR 2024 L R ECONCILING ASSUMPTIONS IN TABLE 1 Let us discuss in detail how the assumptions set for treatment effects (assum. 2.1 to 2.3) and ODE discovery (assum. 3.1 to 3.3) compare as reported in table 1 in section 4. existence & uniqueness (assum. 3.1) overlap (assum. 2.2) and functional spaces (assum. 3.3) overlap (assum. 2.2) . These assumptions do not map one-to-one. However, they do serve a similar purpose. Essentially, overlap allows us to assume that there is always a very similar sample (in terms of covariates), such that we have access to a full range of potential outcomes such that we can estimate eq. (1) for every treatment and covariates. When violated, Gelman & Hill (2006) pointed out that one has to rely much more on correct model specification, i.e., rather than assuming overlap, one may resort to the correct model specification instead. Assum. 3.1 and 3.3 tell us exactly this: the former assumes that there exists one discoverable function (i.e., one that can be correctly specified), while the latter tells us that the correct function is comprised of the token set that is provided to the ODE discovery method. Combined, these assumptions should result in a correctly specified model, which can then be used to relax an overlap assumption. observability (assum. 3.2) consistency (assum. 2.1) and observability (assum. 3.2) overlap (assum. 2.3) . While the previous assumptions did not map one-to-one, we find that assum. 3.2 does map one-to-one with assum. 2.1 and 2.3. With eq. (3) we combine covariates, (potential) outcomes and treatments into one dynamical system. Given this, assum. 3.2 states that every variable of the complete ODE is observed in the dataset. These variables include the potential outcomes (which corresponds to assum. 2.1) as well as all the covariates (corresponding to assum. 2.3). One could make the argument that",
            "references": [
                {
                    "title": "Deep Generative Symbolic Regression",
                    "abstract": "Symbolic regression (SR) aims to discover concise closed-form mathematical equations from data, a task fundamental to scientific discovery. However, the problem is highly challenging because closed-form equations lie in a complex combinatorial search space. Existing methods, ranging from heuristic search to reinforcement learning, fail to scale with the number of input variables. We make the observation that closed-form equations often have structural characteristics and invariances (e.g., the commutative law) that could be further exploited to build more effective symbolic regression solutions. Motivated by this observation, our key contribution is to leverage pre-trained deep generative models to capture the intrinsic regularities of equations, thereby providing a solid foundation for subsequent optimization steps. We show that our novel formalism unifies several prominent approaches of symbolic regression and offers a new perspective to justify and improve on the previous ad hoc designs, such as the usage of cross-entropy loss during pre-training. Specifically, we propose an instantiation of our framework, Deep Generative Symbolic Regression (DGSR). In our experiments, we show that DGSR achieves a higher recovery rate of true equations in the setting of a larger number of input variables, and it is more computationally efficient at inference time than state-of-the-art RL symbolic regression solutions."
                },
                {
                    "title": "The identification of piecewise non-linear dynamical system without understanding the mechanism.",
                    "abstract": "This paper designs an algorithm to distill the piecewise non-linear dynamical system from the data without prior knowledge. The system to be identified does not have to be written as a known model term or be thoroughly understood. We exploit the fact that an unknown piecewise non-linear system can be decomposed into the Fourier series as long as its equations of motion are Riemann integrable. Based on this property, we reduce the challenge of finding the correct model to discovering the Fourier series approximation. However, the Fourier series approximation of the piecewise function is inaccurate. The new method takes advantage of this weakness to determine whether the model has piecewise features and to find a way to discover the discontinuity set. Then, the dynamical system on each segment is identified as a pure Fourier series. Identification of intricate models can be achieved in simple steps. The results show that the method can accurately discover the equation of motion and precisely capture the non-smooth characteristic. Next, the prediction and further detailed analysis can be carried out."
                },
                {
                    "title": "Discovering Causal Relations and Equations from Data",
                    "abstract": "Physics is a field of science that has traditionally used the scientific method to answer questions about why natural phenomena occur and to make testable models that explain the phenomena. Discovering equations, laws and principles that are invariant, robust and causal explanations of the world has been fundamental in physical sciences throughout the centuries. Discoveries emerge from observing the world and, when possible, performing interventional studies in the system under study. With the advent of big data and the use of data-driven methods, causal and equation discovery fields have grown and made progress in computer science, physics, statistics, philosophy, and many applied fields. All these domains are intertwined and can be used to discover causal relations, physical laws, and equations from observational data. This paper reviews the concepts, methods, and relevant works on causal and equation discovery in the broad field of Physics and outlines the most important challenges and promising future lines of research. We also provide a taxonomy for observational causal and equation discovery, point out connections, and showcase a complete set of case studies in Earth and climate sciences, fluid dynamics and mechanics, and the neurosciences. This review demonstrates that discovering fundamental laws and causal relations by observing natural phenomena is being revolutionised with the efficient exploitation of observational data, modern machine learning algorithms and the interaction with domain knowledge. Exciting times are ahead with many challenges and opportunities to improve our understanding of complex systems."
                },
                {
                    "title": "MetaPhysiCa: OOD Robustness in Physics-informed Machine Learning",
                    "abstract": "A fundamental challenge in physics-informed machine learning (PIML) is the design of robust PIML methods for out-of-distribution (OOD) forecasting tasks. These OOD tasks require learning-to-learn from observations of the same (ODE) dynamical system with different unknown ODE parameters, and demand accurate forecasts even under out-of-support initial conditions and out-of-support ODE parameters. In this work we propose a solution for such tasks, which we define as a meta-learning procedure for causal structure discovery (including invariant risk minimization). Using three different OOD tasks, we empirically observe that the proposed approach significantly outperforms existing state-of-the-art PIML and deep learning methods."
                },
                {
                    "title": "Causal Deep Learning",
                    "abstract": "Causality has the potential to truly transform the way we solve a large number of real-world problems. Yet, so far, its potential largely remains to be unlocked as causality often requires crucial assumptions which cannot be tested in practice. To address this challenge, we propose a new way of thinking about causality -- we call this causal deep learning. Our causal deep learning framework spans three dimensions: (1) a structural dimension, which incorporates partial yet testable causal knowledge rather than assuming either complete or no causal knowledge among the variables of interest; (2) a parametric dimension, which encompasses parametric forms that capture the type of relationships among the variables of interest; and (3) a temporal dimension, which captures exposure times or how the variables of interest interact (possibly causally) over time. Causal deep learning enables us to make progress on a variety of real-world problems by leveraging partial causal knowledge (including independencies among variables) and quantitatively characterising causal relationships among variables of interest (possibly over time). Our framework clearly identifies which assumptions are testable and which ones are not, such that the resulting solutions can be judiciously adopted in practice. Using our formulation we can combine or chain together causal representations to solve specific problems without losing track of which assumptions are required to build these solutions, pushing real-world impact in healthcare, economics and business, environmental sciences and education, through causal deep learning."
                },
                {
                    "title": "Neural Laplace Control for Continuous-time Delayed Systems",
                    "abstract": "Many real-world offline reinforcement learning (RL) problems involve continuous-time environments with delays. Such environments are characterized by two distinctive features: firstly, the state x(t) is observed at irregular time intervals, and secondly, the current action a(t) only affects the future state x(t + g) with an unknown delay g>0. A prime example of such an environment is satellite control where the communication link between earth and a satellite causes irregular observations and delays. Existing offline RL algorithms have achieved success in environments with irregularly observed states in time or known delays. However, environments involving both irregular observations in time and unknown delays remains an open and challenging problem. To this end, we propose Neural Laplace Control, a continuous-time model-based offline RL method that combines a Neural Laplace dynamics model with a model predictive control (MPC) planner--and is able to learn from an offline dataset sampled with irregular time intervals from an environment that has a inherent unknown constant delay. We show experimentally on continuous-time delayed environments it is able to achieve near expert policy performance."
                },
                {
                    "title": "Continuous-Time Modeling of Counterfactual Outcomes Using Neural Controlled Differential Equations",
                    "abstract": "Estimating counterfactual outcomes over time has the potential to unlock personalized healthcare by assisting decision-makers to answer ''what-iF'' questions. Existing causal inference approaches typically consider regular, discrete-time intervals between observations and treatment decisions and hence are unable to naturally model irregularly sampled data, which is the common setting in practice. To handle arbitrary observation patterns, we interpret the data as samples from an underlying continuous-time process and propose to model its latent trajectory explicitly using the mathematics of controlled differential equations. This leads to a new approach, the Treatment Effect Neural Controlled Differential Equation (TE-CDE), that allows the potential outcomes to be evaluated at any time point. In addition, adversarial training is used to adjust for time-dependent confounding which is critical in longitudinal settings and is an added challenge not encountered in conventional time-series. To assess solutions to this problem, we propose a controllable simulation environment based on a model of tumor growth for a range of scenarios with irregular sampling reflective of a variety of clinical scenarios. TE-CDE consistently outperforms existing approaches in all simulated scenarios with irregular sampling."
                },
                {
                    "title": "Neural Laplace: Learning diverse classes of differential equations in the Laplace domain",
                    "abstract": "Neural Ordinary Differential Equations model dynamical systems with ODE s learned by neural networks. However, ODEs are fundamentally inadequate to model systems with long-range dependencies or discontinuities, which are common in engineering and biological systems. Broader classes of differential equations (DE) have been proposed as remedies, including delay differential equations and integro-differential equations. Furthermore, Neural ODE suffers from numerical instability when modelling stiff ODEs and ODEs with piecewise forcing functions. In this work, we propose Neural Laplace , a uni\ufb01ed framework for learning diverse classes of DEs including all the aforementioned ones. Instead of modelling the dynamics in the time domain, we model it in the Laplace domain, where the history-dependencies and discontinuities in time can be represented as summations of complex exponentials. To make learning more ef\ufb01cient, we use the geometrical stereographic map of a Riemann sphere to induce more smoothness in the Laplace domain. In the experiments, Neural Laplace shows superior performance in modelling and extrapolating the trajectories of diverse classes of DEs, including the ones with complex history dependency and abrupt changes."
                },
                {
                    "title": "Causal Transformer for Estimating Counterfactual Outcomes",
                    "abstract": "Estimating counterfactual outcomes over time from observational data is relevant for many applications (e.g., personalized medicine). Yet, state-of-the-art methods build upon simple long short-term memory (LSTM) networks, thus rendering inferences for complex, long-range dependencies challenging. In this paper, we develop a novel Causal Transformer for estimating counterfactual outcomes over time. Our model is specifically designed to capture complex, long-range dependencies among time-varying confounders. For this, we combine three transformer subnetworks with separate inputs for time-varying covariates, previous treatments, and previous outcomes into a joint network with in-between cross-attentions. We further develop a custom, end-to-end training procedure for our Causal Transformer. Specifically, we propose a novel counterfactual domain confusion loss to address confounding bias: it aims to learn adversarial balanced representations, so that they are predictive of the next outcome but non-predictive of the current treatment assignment. We evaluate our Causal Transformer based on synthetic and real-world datasets, where it achieves superior performance over current baselines. To the best of our knowledge, this is the first work proposing transformer-based architecture for estimating counterfactual outcomes from longitudinal data."
                },
                {
                    "title": "Estimating Conditional Average Treatment Effects with Missing Treatment Information",
                    "abstract": "Estimating conditional average treatment effects (CATE) is challenging, especially when treatment information is missing. Although this is a widespread problem in practice, CATE estimation with missing treatments has received little attention. In this paper, we analyze CATE estimation in the setting with missing treatments where unique challenges arise in the form of covariate shifts. We identify two covariate shifts in our setting: (i) a covariate shift between the treated and control population; and (ii) a covariate shift between the observed and missing treatment population. We first theoretically show the effect of these covariate shifts by deriving a generalization bound for estimating CATE in our setting with missing treatments. Then, motivated by our bound, we develop the missing treatment representation network (MTRNet), a novel CATE estimation algorithm that learns a balanced representation of covariates using domain adaptation. By using balanced representations, MTRNet provides more reliable CATE estimates in the covariate domains where the data are not fully observed. In various experiments with semi-synthetic and real-world data, we show that our algorithm improves over the state-of-the-art by a substantial margin."
                },
                {
                    "title": "Disentangled Counterfactual Recurrent Networks for Treatment Effect Inference over Time",
                    "abstract": "Choosing the best treatment-plan for each individual patient requires accurate forecasts of their outcome trajectories as a function of the treatment, over time. While large observational data sets constitute rich sources of information to learn from, they also contain biases as treatments are rarely assigned randomly in practice. To provide accurate and unbiased forecasts, we introduce the Disentangled Counterfactual Recurrent Network (DCRN), a novel sequence-to-sequence architecture that estimates treatment outcomes over time by learning representations of patient histories that are disentangled into three separate latent factors: a treatment factor, influencing only treatment selection; an outcome factor, influencing only the outcome; and a confounding factor, influencing both. With an architecture that is completely inspired by the causal structure of treatment influence over time, we advance forecast accuracy and disease understanding, as our architecture allows for practitioners to infer which patient features influence which part in a patient's trajectory, contrasting other approaches in this domain. We demonstrate that DCRN outperforms current state-of-the-art methods in forecasting treatment responses, on both real and simulated data."
                },
                {
                    "title": "Counterfactual Temporal Point Processes",
                    "abstract": "Machine learning models based on temporal point processes are the state of the art in a wide variety of applications involving discrete events in continuous time. However, these models lack the ability to answer counterfactual questions, which are increasingly relevant as these models are being used to inform targeted interventions. In this work, our goal is to fill this gap. To this end, we first develop a causal model of thinning for temporal point processes that builds upon the Gumbel-Max structural causal model. This model satisfies a desirable counterfactual monotonicity condition, which is sufficient to identify counterfactual dynamics in the process of thinning. Then, given an observed realization of a temporal point process with a given intensity function, we develop a sampling algorithm that uses the above causal model of thinning and the superposition theorem to simulate counterfactual realizations of the temporal point process under a given alternative intensity function. Simulation experiments using synthetic and real epidemiological data show that the counterfactual realizations provided by our algorithm may give valuable insights to enhance targeted interventions."
                },
                {
                    "title": "Statistical Inference for Piecewise Affine System Identification",
                    "abstract": "This short note studies the problem of piecewise affine system identification, being a special nonlinear system based on our previous contribution on it. Two different identification strategies are proposed to achieve our mission, such as centralized identification and distributed identification. More specifically, for centralized identification, the total observed input-output data are used to estimate all unknown parameter vectors simultaneously without any consideration on the classification process. But for distributed identification, after the whole observed input-output data are classified into their own right subregions, then part input-output data, belonging to the same subregion, are applied to estimate the unknown parameter vector. Whatever the centralized identification and distributed identification, the final decision is to determine the unknown parameter vector in one linear form, so the recursive least squares algorithm and its modified form with the dead zone are studied to deal with the statistical noise and bounded noise, respectively. Finally, one simulation example is used to compare the identification accuracy for our considered two identification strategies."
                },
                {
                    "title": "Discovery of nonlinear dynamical systems using a Runge\u2013Kutta inspired dictionary-based sparse regression approach",
                    "abstract": "In this work, we blend machine learning and dictionary-based learning with numerical analysis tools to discover differential equations from noisy and sparsely sampled measurement data of time-dependent processes. We use the fact that given a dictionary containing large candidate nonlinear functions, dynamical models can often be described by a few appropriately chosen basis functions. As a result, we obtain parsimonious models that can be better interpreted by practitioners, and potentially generalize better beyond the sampling regime than black-box modelling. In this work, we integrate a numerical integration framework with dictionary learning that yields differential equations without requiring or approximating derivative information at any stage. Hence, it is utterly effective for corrupted and sparsely sampled data. We discuss its extension to governing equations, containing rational nonlinearities that typically appear in biological networks. Moreover, we generalized the method to governing equations subject to parameter variations and externally controlled inputs. We demonstrate the efficiency of the method to discover a number of diverse differential equations using noisy measurements, including a model describing neural dynamics, chaotic Lorenz model, Michaelis\u2013Menten kinetics and a parameterized Hopf normal form."
                },
                {
                    "title": "Learning outside the Black-Box: The pursuit of interpretable models",
                    "abstract": "Machine Learning has proved its ability to produce accurate models but the deployment of these models outside the machine learning community has been hindered by the difficulties of interpreting these models. This paper proposes an algorithm that produces a continuous global interpretation of any given continuous black-box function. Our algorithm employs a variation of projection pursuit in which the ridge functions are chosen to be Meijer G-functions, rather than the usual polynomial splines. Because Meijer G-functions are differentiable in their parameters, we can tune the parameters of the representation by gradient descent; as a consequence, our algorithm is efficient. Using five familiar data sets from the UCI repository and two familiar machine learning algorithms, we demonstrate that our algorithm produces global interpretations that are both highly accurate and parsimonious (involve a small number of terms). Our interpretations permit easy understanding of the relative importance of features and feature interactions. Our interpretation algorithm represents a leap forward from the previous state of the art."
                },
                {
                    "title": "Neural Ordinary Differential Equations for Intervention Modeling",
                    "abstract": "By interpreting the forward dynamics of the latent representation of neural networks as an ordinary differential equation, Neural Ordinary Differential Equation (Neural ODE) emerged as an effective framework for modeling a system dynamics in the continuous time domain. However, real-world systems often involves external interventions that cause changes in the system dynamics such as a moving ball coming in contact with another ball, or such as a patient being administered with particular drug. Neural ODE and a number of its recent variants, however, are not suitable for modeling such interventions as they do not properly model the observations and the interventions separately. In this paper, we propose a novel neural ODE-based approach (IMODE) that properly model the effect of external interventions by employing two ODE functions to separately handle the observations and the interventions. Using both synthetic and real-world time-series datasets involving interventions, our experimental results consistently demonstrate the superiority of IMODE compared to existing approaches."
                },
                {
                    "title": "Deep Learning for Digital Asset Limit Order Books",
                    "abstract": "This paper shows that temporal CNNs accurately predict bitcoin spot price movements from limit order book data. On a 2 second prediction time horizon we achieve 71\\% walk-forward accuracy on the popular cryptocurrency exchange coinbase. Our model can be trained in less than a day on commodity GPUs which could be installed into colocation centers allowing for model sync with existing faster orderbook prediction models. We provide source code and data at this https URL."
                },
                {
                    "title": "Discovering Symbolic Models from Deep Learning with Inductive Biases",
                    "abstract": "We develop a general approach to distill symbolic representations of a learned deep model by introducing strong inductive biases. We focus on Graph Neural Networks (GNNs). The technique works as follows: we first encourage sparse latent representations when we train a GNN in a supervised setting, then we apply symbolic regression to components of the learned model to extract explicit physical relations. We find the correct known equations, including force laws and Hamiltonians, can be extracted from the neural network. We then apply our method to a non-trivial cosmology example-a detailed dark matter simulation-and discover a new analytic formula which can predict the concentration of dark matter from the mass distribution of nearby cosmic structures. The symbolic expressions extracted from the GNN using our technique also generalized to out-of-distribution data better than the GNN itself. Our approach offers alternative directions for interpreting neural networks and discovering novel physical principles from the representations they learn."
                },
                {
                    "title": "Transfer Learning via $\\ell_1$ Regularization",
                    "abstract": "Machine learning algorithms typically require abundant data under a stationary environment. However, environments are nonstationary in many real-world applications. Critical issues lie in how to effectively adapt models under an ever-changing environment. We propose a method for transferring knowledge from a source domain to a target domain via $\\ell_1$ regularization. We incorporate $\\ell_1$ regularization of differences between source parameters and target parameters, in addition to an ordinary $\\ell_1$ regularization. Hence, our method yields sparsity for both the estimates themselves and changes of the estimates. The proposed method has a tight estimation error bound under a stationary environment, and the estimate remains unchanged from the source estimate under small residuals. Moreover, the estimate is consistent with the underlying function, even when the source estimate is mistaken due to nonstationarity. Empirical results demonstrate that the proposed method effectively balances stability and plasticity."
                },
                {
                    "title": "Learning Long-Term Dependencies in Irregularly-Sampled Time Series",
                    "abstract": "Recurrent neural networks (RNNs) with continuous-time hidden states are a natural fit for modeling irregularly-sampled time series. These models, however, face difficulties when the input data possess long-term dependencies. We prove that similar to standard RNNs, the underlying reason for this issue is the vanishing or exploding of the gradient during training. This phenomenon is expressed by the ordinary differential equation (ODE) representation of the hidden state, regardless of the ODE solver's choice. We provide a solution by designing a new algorithm based on the long short-term memory (LSTM) that separates its memory from its time-continuous state. This way, we encode a continuous-time dynamical flow within the RNN, allowing it to respond to inputs arriving at arbitrary time-lags while ensuring a constant error propagation through the memory path. We call these RNN models ODE-LSTMs. We experimentally show that ODE-LSTMs outperform advanced RNN-based counterparts on non-uniformly sampled data with long-term dependencies. All code and data is available at this https URL."
                },
                {
                    "title": "From Real\u2010World Patient Data to Individualized Treatment Effects Using Machine Learning: Current and Future Methods to Address Underlying Challenges",
                    "abstract": "Clinical decision making needs to be supported by evidence that treatments are beneficial to individual patients. Although randomized control trials (RCTs) are the gold standard for testing and introducing new drugs, due to the focus on specific questions with respect to establishing efficacy and safety vs. standard treatment, they do not provide a full characterization of the heterogeneity in the final intended treatment population. Conversely, real\u2010world observational data, such as electronic health records (EHRs), contain large amounts of clinical information about heterogeneous patients and their response to treatments. In this paper, we introduce the main opportunities and challenges in using observational data for training machine learning methods to estimate individualized treatment effects and make treatment recommendations. We describe the modeling choices of the state\u2010of\u2010the\u2010art machine learning methods for causal inference, developed for estimating treatment effects both in the cross\u2010section and longitudinal settings. Additionally, we highlight future research directions that could lead to achieving the full potential of leveraging EHRs and machine learning for making individualized treatment recommendations. We also discuss how experimental data from RCTs and Pharmacometric and Quantitative Systems Pharmacology approaches can be used to not only improve machine learning methods, but also provide ways for validating them. These future research directions will require us to collaborate across the scientific disciplines to incorporate models based on RCTs and known disease processes, physiology, and pharmacology into these machine learning models based on EHRs to fully optimize the opportunity these data present."
                },
                {
                    "title": "When and How to Lift the Lockdown? Global COVID-19 Scenario Analysis and Policy Assessment using Compartmental Gaussian Processes",
                    "abstract": "The coronavirus disease 2019 (COVID-19) global pandemic has led many countries to impose unprecedented lockdown measures in order to slow down the outbreak. Questions on whether governments have acted promptly enough, and whether lockdown measures can be lifted soon have since been central in public discourse. Data-driven models that predict COVID-19 fatalities under different lockdown policy scenarios are essential for addressing these questions and informing governments on future policy directions. To this end, this paper develops a Bayesian model for predicting the effects of COVID-19 lockdown policies in a global context -- we treat each country as a distinct data point, and exploit variations of policies across countries to learn country-specific policy effects. Our model utilizes a two-layer Gaussian process (GP) prior -- the lower layer uses a compartmental SEIR (Susceptible, Exposed, Infected, Recovered) model as a prior mean function with \"country-and-policy-specific\" parameters that capture fatality curves under \"counterfactual\" policies within each country, whereas the upper layer is shared across all countries, and learns lower-layer SEIR parameters as a function of a country's features and its policy indicators. Our model combines the solid mechanistic foundations of SEIR models (Bayesian priors) with the flexible data-driven modeling and gradient-based optimization routines of machine learning (Bayesian posteriors) -- i.e., the entire model is trained end-to-end via stochastic variational inference. We compare the projections of COVID-19 fatalities by our model with other models listed by the Center for Disease Control (CDC), and provide scenario analyses for various lockdown and reopening strategies highlighting their impact on COVID-19 fatalities."
                },
                {
                    "title": "Weak SINDy: Galerkin-Based Data-Driven Model Selection",
                    "abstract": "We present a novel weak formulation and discretization for discovering governing equations from noisy measurement data. This method of learning differential equations from data fits into a new class of algorithms that replace pointwise derivative approximations with linear transformations and variance reduction techniques. Compared to the standard SINDy algorithm presented in [S. L. Brunton, J. L. Proctor, and J. N. Kutz, Proc. Natl. Acad. Sci. USA, 113 (2016), pp. 3932-3937], our so-called weak SINDy (WSINDy) algorithm allows for reliable model identification from data with large noise (often with ratios greater than 0.1) and reduces the error in the recovered coefficients to enable accurate prediction. Moreover, the coefficient error scales linearly with the noise level, leading to high-accuracy recovery in the low-noise regime. Altogether, WSINDy combines the simplicity and efficiency of the SINDy algorithm with the natural noise reduction of integration, as demonstrated in [H. Schaeffer and S. G. McCalla, Phys. Rev. E, 96 (2017), 023302], to arrive at a robust and accurate method of sparse recovery."
                },
                {
                    "title": "Mathematical prediction of clinical outcomes in advanced cancer patients treated with checkpoint inhibitor immunotherapy",
                    "abstract": "A mathematical model of checkpoint inhibitor immunotherapy yields patient-specific prognosis of tumor kill and patient survival. We present a mechanistic mathematical model of immune checkpoint inhibitor therapy to address the oncological need for early, broadly applicable readouts (biomarkers) of patient response to immunotherapy. The model is built upon the complex biological and physical interactions between the immune system and cancer, and is informed using only standard-of-care CT. We have retrospectively applied the model to 245 patients from multiple clinical trials treated with anti\u2013CTLA-4 or anti\u2013PD-1/PD-L1 antibodies. We found that model parameters distinctly identified patients with common (n = 18) and rare (n = 10) malignancy types who benefited and did not benefit from these monotherapies with accuracy as high as 88% at first restaging (median 53 days). Further, the parameters successfully differentiated pseudo-progression from true progression, providing previously unidentified insights into the unique biophysical characteristics of pseudo-progression. Our mathematical model offers a clinically relevant tool for personalized oncology and for engineering immunotherapy regimens."
                },
                {
                    "title": "Estimating Counterfactual Treatment Outcomes over Time Through Adversarially Balanced Representations",
                    "abstract": "Identifying when to give treatments to patients and how to select among multiple treatments over time are important medical problems with a few existing solutions. In this paper, we introduce the Counterfactual Recurrent Network (CRN), a novel sequence-to-sequence model that leverages the increasingly available patient observational data to estimate treatment effects over time and answer such medical questions. To handle the bias from time-varying confounders, covariates affecting the treatment assignment policy in the observational data, CRN uses domain adversarial training to build balancing representations of the patient history. At each timestep, CRN constructs a treatment invariant representation which removes the association between patient history and treatment assignments and thus can be reliably used for making counterfactual predictions. On a simulated model of tumour growth, with varying degree of time-dependent confounding, we show how our model achieves lower error in estimating counterfactuals and in choosing the correct treatment and timing of treatment than current state-of-the-art methods."
                },
                {
                    "title": "Causal Models for Dynamical Systems",
                    "abstract": "A probabilistic model describes a system in its observational state. In many situations, however, we are interested in the system's response under interventions. The class of structural causal models provides a language that allows us to model the behaviour under interventions. It can been taken as a starting point to answer a plethora of causal questions, including the identification of causal effects or causal structure learning. In this chapter, we provide a natural and straight-forward extension of this concept to dynamical systems, focusing on continuous time models. In particular, we introduce two types of causal kinetic models that differ in how the randomness enters into the model: it may either be considered as observational noise or as systematic driving noise. In both cases, we define interventions and therefore provide a possible starting point for causal inference. In this sense, the book chapter provides more questions than answers. The focus of the proposed causal kinetic models lies on the dynamics themselves rather than corresponding stationary distributions, for example. We believe that this is beneficial when the aim is to model the full time evolution of the system and data are measured at different time points. Under this focus, it is natural to consider interventions in the differential equations themselves."
                },
                {
                    "title": "Deep symbolic regression: Recovering mathematical expressions from data via risk-seeking policy gradients",
                    "abstract": "Discovering the underlying mathematical expressions describing a dataset is a core challenge for artificial intelligence. This is the problem of $\\textit{symbolic}$ $\\textit{regression.}$ Despite recent advances in training neural networks to solve complex tasks, deep learning approaches to symbolic regression are underexplored. We propose a framework that combines deep learning with symbolic regression via a simple idea: use a large model to search the space of small models. More specifically, we use a recurrent neural network to emit a distribution over tractable mathematical expressions, and employ reinforcement learning to train the network to generate better-fitting expressions. Our algorithm significantly outperforms standard genetic programming-based symbolic regression in its ability to exactly recover symbolic expressions on a series of benchmark problems, both with and without added noise. More broadly, our contributions include a framework that can be applied to optimize hierarchical, variable-length objects under a black-box performance metric, with the ability to incorporate a priori constraints in situ, and a risk-seeking policy gradient formulation that optimizes for best-case performance instead of expected performance."
                },
                {
                    "title": "Cancer cell population growth kinetics at low densities deviate from the exponential growth model and suggest an Allee effect",
                    "abstract": "Models of cancer cell population expansion assume exponential growth kinetics at low cell densities, with deviations from exponential growth only at higher densities due to limited resources such as space and nutrients. However, recent pre-clinical and clinical observations of tumor initiation or recurrence indicate the presence of tumor growth kinetics in which growth rates scale with cell numbers. These observations are analogous to the cooperative behavior of species in an ecosystem described by the ecological principle of the Allee effect. In preclinical and clinical models however, tumor growth data is limited by the lower limit of detection (i.e. a measurable lesion) and confounding variables, such as tumor microenvironment and immune responses may cause and mask deviations from exponential growth models. In this work, we present alternative growth models to investigate the presence of an Allee effect in cancer cells seeded at low cell densities in a controlled in vitro setting. We propose a stochastic modeling framework to consider the small number of cells in this low-density regime and use the moment approach for stochastic parameter estimation to calibrate the stochastic growth trajectories. We validate the framework on simulated data and apply this approach to longitudinal cell proliferation data of BT-474 luminal B breast cancer cells. We find that cell population growth kinetics are best described by a model structure that considers the Allee effect, in that the birth rate of tumor cells depends on cell number. This indicates a potentially critical role of cooperative behavior among tumor cells at low cell densities with relevance to early stage growth patterns of emerging tumors and relapse. Author Summary The growth kinetics of cancer cells at very low cell densities are of utmost clinical importance as the ability of a small number of newly transformed or surviving cells to grow exponentially and thus, to \u201ctake off\u201d underlies tumor formation and relapse after treatment. Mathematical models of stochastic tumor cell growth typically assume a stochastic birth-death process of cells impacted by limited nutrients and space when cells reach high density, resulting in the widely accepted logistic growth model. Here we present an in-depth investigation of alternate growth models adopted from ecology to describe potential deviations from a simple cell autonomous birth-death model at low cell densities. We show that our stochastic modeling framework is robust and can be used to identify the underlying structure of stochastic growth trajectories from both simulated and experimental data taken from a controlled in vitro setting in which we can capture data from the relevant low cell density regime. This work suggests that the assumption of cell autonomous proliferation via a constant exponential growth rate at low cell densities may not be appropriate for all cancer cell growth dynamics. Consideration of cooperative behavior amongst tumor cells in this regime is critical for elucidating strategies for controlling tumor cell growth."
                },
                {
                    "title": "Time Series Deconfounder: Estimating Treatment Effects over Time in the Presence of Hidden Confounders",
                    "abstract": "The estimation of treatment effects is a pervasive problem in medicine. Existing methods for estimating treatment effects from longitudinal observational data assume that there are no hidden confounders, an assumption that is not testable in practice and, if it does not hold, leads to biased estimates. In this paper, we develop the Time Series Deconfounder, a method that leverages the assignment of multiple treatments over time to enable the estimation of treatment effects in the presence of multi-cause hidden confounders. The Time Series Deconfounder uses a novel recurrent neural network architecture with multitask output to build a factor model over time and infer latent variables that render the assigned treatments conditionally independent; then, it performs causal inference using these latent variables that act as substitutes for the multi-cause unobserved confounders. We provide a theoretical analysis for obtaining unbiased causal effects of time-varying exposures using the Time Series Deconfounder. Using both simulated and real data we show the effectiveness of our method in deconfounding the estimation of treatment responses over time."
                },
                {
                    "title": "Causal inference in continuous time: an example on prostate cancer therapy.",
                    "abstract": "In marginal structural models (MSMs), time is traditionally treated as a discrete parameter. In survival analysis on the other hand, we study processes that develop in continuous time. Therefore, R\u00f8ysland (2011. A martingale approach to continuous-time marginal structural models. Bernoulli 17, 895-915) developed the continuous-time MSMs, along with continuous-time weights. The continuous-time weights are conceptually similar to the inverse probability weights that are used in discrete time MSMs. Here, we demonstrate that continuous-time MSMs may be used in practice. First, we briefly describe the causal model assumptions using counting process notation, and we suggest how causal effect estimates can be derived by calculating continuous-time weights. Then, we describe how additive hazard models can be used to find such effect estimates. Finally, we apply this strategy to compare medium to long-term differences between the two prostate cancer treatments radical prostatectomy and radiation therapy, using data from the Norwegian Cancer Registry. In contrast to the results of a naive analysis, we find that the marginal cumulative incidence of treatment failure is similar between the strategies, accounting for the competing risk of other death."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Reliable Decision Support using Counterfactual Models",
                    "abstract": "Answering \"What if?\" questions is important in many domains. For example, would a patient's disease progression slow down if I were to give them a dose of drug A? Ideally, we answer our question using an experiment, but this is not always possible (e.g., it may be unethical). As an alternative, we can use non-experimental data to learn models that make counterfactual predictions of what we would observe had we run an experiment. In this paper, we propose the counterfactual GP, a counterfactual model of continuous-time trajectories (time series) under sequences of actions taken in continuous-time. We develop our model within the potential outcomes framework of Neyman and Rubin. The counterfactual GP is trained using a joint maximum likelihood objective that adjusts for dependencies between observed actions and outcomes in the training data. We report two sets of experimental results using the counterfactual GP. The first shows that it can be used to learn the natural progression (i.e. untreated progression) of biomarker trajectories from observational data. In the second, we show how the CGP can be used for medical decision support by learning counterfactual models of renal health under different types of dialysis."
                },
                {
                    "title": "A flexible parametric approach for estimating continuous\u2010time inverse probability of treatment and censoring weights",
                    "abstract": "Marginal structural Cox models are used for quantifying marginal treatment effects on outcome event hazard function. Such models are estimated using inverse probability of treatment and censoring (IPTC) weighting, which properly accounts for the impact of time\u2010dependent confounders, avoiding conditioning on factors on the causal pathway. To estimate the IPTC weights, the treatment assignment mechanism is conventionally modeled in discrete time. While this is natural in situations where treatment information is recorded at scheduled follow\u2010up visits, in other contexts, the events specifying the treatment history can be modeled in continuous time using the tools of event history analysis. This is particularly the case for treatment procedures, such as surgeries. In this paper, we propose a novel approach for flexible parametric estimation of continuous\u2010time IPTC weights and illustrate it in assessing the relationship between metastasectomy and mortality in metastatic renal cell carcinoma patients. Copyright \u00a9 2016 John Wiley & Sons, Ltd."
                },
                {
                    "title": "A Bayesian Nonparametric Approach for Estimating Individualized Treatment-Response Curves",
                    "abstract": "We study the problem of estimating the continuous response over time to interventions using observational time series---a retrospective dataset where the policy by which the data are generated is unknown to the learner. We are motivated by applications where response varies by individuals and therefore, estimating responses at the individual-level is valuable for personalizing decision-making. We refer to this as the problem of estimating individualized treatment response (ITR) curves. In statistics, G-computation formula (Robins, 1986) has been commonly used for estimating treatment responses from observational data containing sequential treatment assignments. However, past studies have focused predominantly on obtaining point-in-time estimates at the population level. We leverage the G-computation formula and develop a novel Bayesian nonparametric (BNP) method that can flexibly model functional data and provide posterior inference over the treatment response curves at both the individual and population level. On a challenging dataset containing time series from patients admitted to a hospital, we estimate responses to treatments used in managing kidney function and show that the resulting fits are more accurate than alternative approaches. Accurate methods for obtaining ITRs from observational data can dramatically accelerate the pace at which personalized treatment plans become possible."
                },
                {
                    "title": "Learning Representations for Counterfactual Inference",
                    "abstract": "Observational studies are rising in importance due to the widespread accumulation of data in fields such as healthcare, education, employment and ecology. We consider the task of answering counterfactual questions such as, \"Would this patient have lower blood sugar had she received a different medication?\". We propose a new algorithmic framework for counterfactual inference which brings together ideas from domain adaptation and representation learning. In addition to a theoretical justification, we perform an empirical comparison with previous approaches to causal inference from observational data. Our deep learning algorithm significantly outperforms the previous state-of-the-art."
                },
                {
                    "title": "Discovering governing equations from data by sparse identification of nonlinear dynamical systems",
                    "abstract": "Significance Understanding dynamic constraints and balances in nature has facilitated rapid development of knowledge and enabled technology, including aircraft, combustion engines, satellites, and electrical power. This work develops a novel framework to discover governing equations underlying a dynamical system simply from data measurements, leveraging advances in sparsity techniques and machine learning. The resulting models are parsimonious, balancing model complexity with descriptive ability while avoiding overfitting. There are many critical data-driven problems, such as understanding cognition from neural recordings, inferring climate patterns, determining stability of financial markets, predicting and suppressing the spread of disease, and controlling turbulence for greener transportation and energy. With abundant data and elusive laws, data-driven discovery of dynamics will continue to play an important role in these efforts. Extracting governing equations from data is a central challenge in many diverse areas of science and engineering. Data are abundant whereas models often remain elusive, as in climate science, neuroscience, ecology, finance, and epidemiology, to name only a few examples. In this work, we combine sparsity-promoting techniques and machine learning with nonlinear dynamical systems to discover governing equations from noisy measurement data. The only assumption about the structure of the model is that there are only a few important terms that govern the dynamics, so that the equations are sparse in the space of possible functions; this assumption holds for many physical systems in an appropriate basis. In particular, we use sparse regression to determine the fewest terms in the dynamic governing equations required to accurately represent the data. This results in parsimonious models that balance accuracy with model complexity to avoid overfitting. We demonstrate the algorithm on a wide range of problems, from simple canonical systems, including linear and nonlinear oscillators and the chaotic Lorenz system, to the fluid vortex shedding behind an obstacle. The fluid example illustrates the ability of this method to discover the underlying dynamics of a system that took experts in the community nearly 30 years to resolve. We also show that this method generalizes to parameterized systems and systems that are time-varying or have external forcing."
                },
                {
                    "title": "Unsupervised Domain Adaptation by Backpropagation",
                    "abstract": "Top-performing deep architectures are trained on massive amounts of labeled data. In the absence of labeled data for a certain task, domain adaptation often provides an attractive option given that labeled data of similar nature but from a different domain (e.g. synthetic images) are available. Here, we propose a new approach to domain adaptation in deep architectures that can be trained on large amount of labeled data from the source domain and large amount of unlabeled data from the target domain (no labeled target-domain data is necessary). \nAs the training progresses, the approach promotes the emergence of \"deep\" features that are (i) discriminative for the main learning task on the source domain and (ii) invariant with respect to the shift between the domains. We show that this adaptation behaviour can be achieved in almost any feed-forward model by augmenting it with few standard layers and a simple new gradient reversal layer. The resulting augmented architecture can be trained using standard backpropagation. \nOverall, the approach can be implemented with little effort using any of the deep-learning packages. The method performs very well in a series of image classification experiments, achieving adaptation effect in the presence of big domain shifts and outperforming previous state-of-the-art on Office datasets."
                },
                {
                    "title": "Learning Phrase Representations using RNN Encoder\u2013Decoder for Statistical Machine Translation",
                    "abstract": "In this paper, we propose a novel neural network model called RNN Encoder\u2010 Decoder that consists of two recurrent neural networks (RNN). One RNN encodes a sequence of symbols into a fixedlength vector representation, and the other decodes the representation into another sequence of symbols. The encoder and decoder of the proposed model are jointly trained to maximize the conditional probability of a target sequence given a source sequence. The performance of a statistical machine translation system is empirically found to improve by using the conditional probabilities of phrase pairs computed by the RNN Encoder\u2010Decoder as an additional feature in the existing log-linear model. Qualitatively, we show that the proposed model learns a semantically and syntactically meaningful representation of linguistic phrases."
                },
                {
                    "title": "Parametric Estimation of Ordinary Differential Equations With Orthogonality Conditions",
                    "abstract": "Differential equations are commonly used to model dynamical deterministic systems in applications. When statistical parameter estimation is required to calibrate theoretical models to data, classical statistical estimators are often confronted to complex and potentially ill-posed optimization problem. As a consequence, alternative estimators to classical parametric estimators are needed for obtaining reliable estimates. We propose a gradient matching approach for the estimation of parametric Ordinary Differential Equations (ODE) observed with noise. Starting from a nonparametric proxy of a true solution of the ODE, we build a parametric estimator based on a variational characterization of the solution. As a Generalized Moment Estimator, our estimator must satisfy a set of orthogonal conditions that are solved in the least squares sense. Despite the use of a nonparametric estimator, we prove the - consistency and asymptotic normality of the Orthogonal Conditions estimator. We can derive confidence sets thanks to a closed-form expression for the asymptotic variance. Finally, the OC estimator is compared to classical estimators in several (simulated and real) experiments and ODE models to show its versatility and relevance with respect to classical Gradient Matching and Nonlinear Least Squares estimators. In particular, we show on a real dataset of influenza infection that the approach gives reliable estimates. Moreover, we show that our approach can deal directly with more elaborated models such as Delay Differential Equation (DDE). Supplementary materials for this article are available online."
                },
                {
                    "title": "Doubly robust estimation of causal effects.",
                    "abstract": "Doubly robust estimation combines a form of outcome regression with a model for the exposure (i.e., the propensity score) to estimate the causal effect of an exposure on an outcome. When used individually to estimate a causal effect, both outcome regression and propensity score methods are unbiased only if the statistical model is correctly specified. The doubly robust estimator combines these 2 approaches such that only 1 of the 2 models need be correctly specified to obtain an unbiased effect estimator. In this introduction to doubly robust estimators, the authors present a conceptual overview of doubly robust estimation, a simple worked example, results from a simulation study examining performance of estimated and bootstrapped standard errors, and a discussion of the potential advantages and limitations of this method. The supplementary material for this paper, which is posted on the Journal's Web site (http://aje.oupjournals.org/), includes a demonstration of the doubly robust property (Web Appendix 1) and a description of a SAS macro (SAS Institute, Inc., Cary, North Carolina) for doubly robust estimation, available for download at http://www.unc.edu/~mfunk/dr/."
                },
                {
                    "title": "Safety in numbers: Learning categories from few examples with multi model knowledge transfer",
                    "abstract": "Learning object categories from small samples is a challenging problem, where machine learning tools can in general provide very few guarantees. Exploiting prior knowledge may be useful to reproduce the human capability of recognizing objects even from only one single view. This paper presents an SVM-based model adaptation algorithm able to select and weight appropriately prior knowledge coming from different categories. The method relies on the solution of a convex optimization problem which ensures to have the minimal leave-one-out error on the training set. Experiments on a subset of the Caltech-256 database show that the proposed method produces better results than both choosing one single prior model, and transferring from all previous experience in a flat uninformative way."
                },
                {
                    "title": "Sparse Convolved Multiple Output G aussian Processes",
                    "abstract": "Recently there has been an increasing interest in methods that deal with multiple outputs. This has been motivated partly by frameworks like multitask learning, multisensor networks or structured output data. From a Gaussian processes perspective, the problem reduces to specifying an appropriate covariance function that, whilst being positive semi-definite, captures the dependencies between all the data points and across all the outputs. One approach to account for non-trivial correlations between outputs employs convolution processes. Under a latent function interpretation of the convolution transform we establish dependencies between output variables. The main drawbacks of this approach are the associated computational and storage demands. In this paper we address these issues. We present different sparse approximations for dependent output Gaussian processes constructed through the convolution formalism. We exploit the conditional independencies present naturally in the model. This leads to a form of the covariance similar in spirit to the so called PITC and FITC approximations for a single output. We show experimental results with synthetic and real data, in particular, we show results in pollution prediction, school exams score prediction and gene expression data."
                },
                {
                    "title": "Causal inference in statistics: An overview",
                    "abstract": "This review presents empiricalresearcherswith recent advances in causal inference, and stresses the paradigmatic shifts that must be un- dertaken in moving from traditionalstatistical analysis to causal analysis of multivariate data. Special emphasis is placed on the assumptions that un- derly all causal inferences, the languages used in formulating those assump- tions, the conditional nature of all causal and counterfactual claims, and the methods that have been developed for the assessment of such claims. These advances are illustrated using a general theory of causation based on the Structural Causal Model (SCM) described in Pearl (2000a), which subsumes and unifies other approaches to causation, and provides a coher- ent mathematical foundation for the analysis of causes and counterfactuals. In particular, the paper surveys the development of mathematical tools for inferring (from a combination of data and assumptions) answers to three types of causal queries: (1) queries about the effects of potential interven- tions, (also called \"causal effects\" or \"policy evaluation\") (2) queries about probabilities of counterfactuals, (including assessment of \"regret,\" \"attri- bution\" or \"causes of effects\") and (3) queries about direct and indirect effects (also known as \"mediation\"). Finally, the paper defines the formal and conceptual relationships between the structural and potential-outcome frameworks and presents tools for a symbiotic analysis that uses the strong features of both."
                },
                {
                    "title": "Data Analysis Using Regression and Multilevel/Hierarchical Models",
                    "abstract": "There is a lot of books, user manual, or guidebook that related to Data Analysis Using Regression And Multilevel Hierarchical Models PDF, such as : alternative medicine resource guide toyota conquest workshop manual mendelian patterns of inheritance answers thermal energy review and answers math eoct answers 2014 the large type christian cabinet rightly dividing the word the perfect store inside ebay section 2 viruses and prions answers mechanical solutions inc denver"
                },
                {
                    "title": "Distilling Free-Form Natural Laws from Experimental Data",
                    "abstract": "For centuries, scientists have attempted to identify and document analytical laws that underlie physical phenomena in nature. Despite the prevalence of computing power, the process of finding natural laws and their corresponding equations has resisted automation. A key challenge to finding analytic relations automatically is defining algorithmically what makes a correlation in observed data important and insightful. We propose a principle for the identification of nontriviality. We demonstrated this approach by automatically searching motion-tracking data captured from various physical systems, ranging from simple harmonic oscillators to chaotic double-pendula. Without any prior knowledge about physics, kinematics, or geometry, the algorithm discovered Hamiltonians, Lagrangians, and other laws of geometric and momentum conservation. The discovery rate accelerated as laws found for simpler systems were used to bootstrap explanations for more complex systems, gradually uncovering the \u201calphabet\u201d used to describe those systems."
                },
                {
                    "title": "Identification of Treatment Effects Using Control Functions in Models with Continuous, Endogenous Treatment and Heterogeneous Effects",
                    "abstract": "We use the control function approach to identify the average treatment effect and the effect of treatment on the treated in models with a continuous endogenous regressor whose impact is heterogeneous. We assume a stochastic polynomial restriction on the form of the heterogeneity but, unlike alternative nonparametric control function approaches, our approach does not require large support assumptions."
                },
                {
                    "title": "Parameter estimation of ODE\u2019s via nonparametric estimators",
                    "abstract": "Ordinary differential equations (ODE's) are widespread models in physics, chemistry and biology. In particular, this mathematical formalism is used for describing the evolution of complex systems and it might consist of high-dimensional sets of coupled nonlinear differential equations. In this setting, we propose a general method for estimating the parameters indexing ODE's from times series. Our method is able to alleviate the computational difficulties encountered by the classical parametric methods. These difficulties are due to the implicit definition of the model. We propose the use of a nonparametric estimator of regression functions as a first-step in the construction of an M-estimator, and we show the consistency of the derived estimator under general conditions. In the case of spline estimators, we prove asymptotic normality, and that the rate of convergence is the usual $\\sqrt{n}$-rate for parametric estimators. Some perspectives of refinements of this new family of parametric estimators are given."
                },
                {
                    "title": "Statistical modeling of causal effects in continuous time",
                    "abstract": "This article studies the estimation of the causal effect of a time-varying treatment on time-to-an-event or on some other continuously distributed outcome. The paper applies to the situation where treatment is repeatedly adapted to time-dependent patient characteristics. The treatment effect cannot be estimated by simply conditioning on these time-dependent patient characteristics, as they may themselves be indications of the treatment effect. This time-dependent confounding is common in observational studies. Robins [(1992) Biometrika 79 321-334, (1998b) Encyclopedia of Biostatistics 6 4372-4389] has proposed the so-called structural nested models to estimate treatment effects in the presence of time-dependent confounding. In this article we provide a conceptual framework and formalization for structural nested models in continuous time. We show that the resulting estimators are consistent and asymptotically normal. Moreover, as conjectured in Robins [(1998b) Encyclopedia of Biostatistics 6 4372-4389], a test for whether treatment affects the outcome of interest can be performed without specifying a model for treatment effect. We illustrate the ideas in this article with an example."
                },
                {
                    "title": "Mediators and moderators of treatment effects in randomized clinical trials.",
                    "abstract": "Randomized clinical trials (RCTs) not only are the gold standard for evaluating the efficacy and effectiveness of psychiatric treatments but also can be valuable in revealing moderators and mediators of therapeutic change. Conceptually, moderators identify on whom and under what circumstances treatments have different effects. Mediators identify why and how treatments have effects. We describe an analytic framework to identify and distinguish between moderators and mediators in RCTs when outcomes are measured dimensionally. Rapid progress in identifying the most effective treatments and understanding on whom treatments work and do not work and why treatments work or do not work depends on efforts to identify moderators and mediators of treatment outcome. We recommend that RCTs routinely include and report such analyses."
                },
                {
                    "title": "Marginal Structural Models to Estimate the Joint Causal Effect of Nonrandomized Treatments",
                    "abstract": "Even in the absence of unmeasured confounding factors or model misspecification, standard methods for estimating the causal effect of time-varying treatments on survival are biased when (a) there exists a time-dependent risk factor for survival that also predicts subsequent treatment, and (b) past treatment history predicts subsequent risk factor level. In contrast, methods based on marginal structural models (MSMs) can provide consistent estimates of causal effects when unmeasured confounding and model misspecification are absent. MSMs are a new class of causal models whose parameters are estimated using a new class of estimators\u2014inverse-probability-of-treatment weighted estimators. We use a marginal structural Cox proportional hazards model to estimate the joint effect of zidovudine (AZT) and prophylaxis therapy for Pneumocystis carinii pneumonia on the survival of HIV-positive men in the Multicenter AIDS Cohort Study, an observational study of homosexual men. We obtained an estimated causal mortality rate (hazard) ratio of .67 (conservative 95% confidence interval .46-.98) for AZT and of 1.14 (.79, 1.64) for prophylaxis therapy. These estimates will be consistent for the true causal rate ratios when the functional forms chosen for our models are correct and data have been obtained on all time-independent and time-dependent covariates that predict both subsequent treatment and mortality."
                },
                {
                    "title": "Marginal Structural Models and Causal Inference in Epidemiology",
                    "abstract": "In observational studies with exposures or treatments that vary over time, standard approaches for adjustment of confounding are biased when there exist time-dependent confounders that are also affected by previous treatment. This paper introduces marginal structural models, a new class of causal models that allow for improved adjustment of confounding in those situations. The parameters of a marginal structural model can be consistently estimated using a new class of estimators, the inverse-probability-of-treatment weighted estimators."
                },
                {
                    "title": "Long Short-Term Memory",
                    "abstract": "Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, LSTM can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units. Multiplicative gate units learn to open and close access to the constant error flow. LSTM is local in space and time; its computational complexity per time step and weight is O. 1. Our experiments with artificial data involve local, distributed, real-valued, and noisy pattern representations. In comparisons with real-time recurrent learning, back propagation through time, recurrent cascade correlation, Elman nets, and neural sequence chunking, LSTM leads to many more successful runs, and learns much faster. LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms."
                },
                {
                    "title": "Orthogonal Functions in Systems and Control",
                    "abstract": "Orthogonal functions in systems and control - a historical perspective least squares approximation of signals signal processing in continuous-time domain analysis of time-delay systems identification of lumped parameter systems identification of linear time-invariant distributed parameter systems identification of linear time-varying and nonlinear distributed parameter systems optimal control of linear systems."
                },
                {
                    "title": "Time Series Analysis",
                    "abstract": "A ordered sequence of events or observations having a time component is called as a time series. Some good examples of time series are daily opening and closing stock prices, daily humidity, temperature, pressure, annual gross domestic product (GDP) of a country and so on."
                },
                {
                    "title": "Identification of Causal Effects Using Instrumental Variables",
                    "abstract": "Abstract We outline a framework for causal inference in settings where assignment to a binary treatment is ignorable, but compliance with the assignment is not perfect so that the receipt of treatment is nonignorable. To address the problems associated with comparing subjects by the ignorable assignment\u2014an \u201cintention-to-treat analysis\u201d\u2014we make use of instrumental variables, which have long been used by economists in the context of regression models with constant treatment effects. We show that the instrumental variables (IV) estimand can be embedded within the Rubin Causal Model (RCM) and that under some simple and easily interpretable assumptions, the IV estimand is the average causal effect for a subgroup of units, the compliers. Without these assumptions, the IV estimand is simply the ratio of intention-to-treat causal estimands with no interpretation as an average causal effect. The advantages of embedding the IV approach in the RCM are that it clarifies the nature of critical assumptions needed for a..."
                },
                {
                    "title": "Instrumental Variables Estimation of Average Treatment Effects in Econometrics and Epidemiology",
                    "abstract": "The average effect of intervention or treatment is a parameter of interest in both epidemiology and econometrics. A key difference between applications in the two fields is that epidemiologic research is more likely to involve qualitative outcomes and nonlinear models. An example is the recent use of the Vietnam era draft lottery to construct estimates of the effect of Vietnam era military service on civilian mortality. In this paper. I present necessary and sufficient conditions for linear instrumental variables. techniques to consistently estimate average treatment effects in qualitative or other nonlinear models. Most latent index models commonly applied to qualitative outcomes in econometrics fail to satisfy these conditions, and monte carlo evidence on the bias of instrumental estimates of the average treatment effect in a bivariate probit model is presented. The evidence suggests that linear instrumental variables estimators perform nearly as well as the correctly specified maximum likelihood estimator. especially in large samples. Linear instrumental variables and the normal maximum likelihood estimator are also remarkably robust to non-normality."
                },
                {
                    "title": "A Learning Algorithm for Continually Running Fully Recurrent Neural Networks",
                    "abstract": "The exact form of a gradient-following learning algorithm for completely recurrent networks running in continually sampled time is derived and used as the basis for practical algorithms for temporal supervised learning tasks. These algorithms have (1) the advantage that they do not require a precisely defined training interval, operating while the network runs; and (2) the disadvantage that they require nonlocal communication in the network being trained and are computationally expensive. These algorithms allow networks having recurrent connections to learn complex tasks that require the retention of information over time periods having either fixed or indefinite length."
                },
                {
                    "title": "Practical Methods of Optimization",
                    "abstract": "Preface Table of Notation Part 1: Unconstrained Optimization Introduction Structure of Methods Newton-like Methods Conjugate Direction Methods Restricted Step Methods Sums of Squares and Nonlinear Equations Part 2: Constrained Optimization Introduction Linear Programming The Theory of Constrained Optimization Quadratic Programming General Linearly Constrained Optimization Nonlinear Programming Other Optimization Problems Non-Smooth Optimization References Subject Index."
                },
                {
                    "title": "Statistics and Causal Inference",
                    "abstract": "Abstract Problems involving causal inference have dogged at the heels of statistics since its earliest days. Correlation does not imply causation, and yet causal conclusions drawn from a carefully designed experiment are often valid. What can a statistical model say about causation? This question is addressed by using a particular model for causal inference (Holland and Rubin 1983; Rubin 1974) to critique the discussions of other writers on causation and causal inference. These include selected philosophers, medical researchers, statisticians, econometricians, and proponents of causal modeling."
                },
                {
                    "title": "The central role of the propensity score in observational studies for causal effects",
                    "abstract": "Abstract : The results of observational studies are often disputed because of nonrandom treatment assignment. For example, patients at greater risk may be overrepresented in some treatment group. This paper discusses the central role of propensity scores and balancing scores in the analysis of observational studies. The propensity score is the (estimated) conditional probability of assignment to a particular treatment given a vector of observed covariates. Both large and small sample theory show that adjustment for the scalar propensity score is sufficient to remove bias due to all observed covariates. Applications include: matched sampling on the univariate propensity score which is equal percent bias reducing under more general conditions than required for discriminant matching, multivariate adjustment by subclassification on balancing scores where the same subclasses are used to estimate treatment effects for all outcome variables and in all subpopulations, and visual representation of multivariate adjustment by a two-dimensional plot. (Author)"
                },
                {
                    "title": "Active Observing in Continuous-time Control",
                    "abstract": "The control of continuous-time environments while actively deciding when to take costly observations in time is a crucial yet unexplored problem, particularly relevant to real-world scenarios such as medicine, low-power systems, and resource management. Existing approaches either rely on continuous-time control methods that take regular, expensive observations in time or discrete-time control with costly observation methods, which are inapplicable to continuous-time settings due to the compounding discretization errors introduced by time discretization. In this work, we are the first to formalize the continuous-time control problem with costly observations. Our key theoretical contribution shows that observing at regular time intervals is not optimal in certain environments, while irregular observation policies yield higher expected utility. This perspective paves the way for the development of novel methods that can take irregular observations in continuous-time control with costly observations. We empirically validate our theoretical findings in various continuous-time environments, including a cancer simulation, by constructing a simple initial method to solve this new problem, with a heuristic threshold on the variance of reward rollouts in an offline continuous-time model-based model predictive control (MPC) planner. Although determining the optimal method remains an open problem, our work offers valuable insights and understanding of this unique problem, laying the foundation for future research in this area."
                },
                {
                    "title": "D-CODE: Discovering Closed-form ODEs from Observed Trajectories",
                    "abstract": "For centuries, scientists have manually designed closed-form ordinary differential equations (ODEs) to model dynamical systems. An automated tool to distill closedform ODEs from observed trajectories would accelerate the modeling process. Traditionally, symbolic regression is used to uncover a closed-form prediction function a = f(b) with label-feature pairs (ai, bi) as training examples. However, an ODE models the time derivative \u1e8b(t) of a dynamical system, e.g. \u1e8b(t) = f(x(t), t), and the \u201clabel\u201d \u1e8b(t) is usually not observed. The existing ways to bridge this gap only perform well for a narrow range of settings with low measurement noise and frequent sampling. In this work, we propose the Discovery of Closedform ODE framework (D-CODE), which advances symbolic regression beyond the paradigm of supervised learning. D-CODE uses a novel objective function based on the variational formulation of ODEs to bypass the unobserved time derivative. For formal justification, we prove that this objective is a valid proxy for the estimation error of the true (but unknown) ODE. In the experiments, D-CODE successfully discovered the governing equations of a diverse range of dynamical systems under challenging measurement settings with high noise and infrequent sampling."
                },
                {
                    "title": "Joint Point Process Model for Counterfactual Treatment\u2013Outcome Trajectories Under Policy Interventions",
                    "abstract": "Policy makers need to predict the progression of an outcome before adopting a new treatment policy, which de\ufb01nes when and how a sequence of treatments a\ufb00ecting the outcome occurs in continuous time. Commonly, algorithms that predict interventional future outcome trajectories take a \ufb01xed sequence of future treatments as input. This excludes scenarios where the policy is unknown or a counterfactual analysis is needed. To handle these limitations, we develop a joint model for treatments and outcomes, which allows for the estimation of treatment policies and e\ufb00ects from sequential treatment\u2013outcome data. It can answer interventional and counterfactual queries about interventions on treatment policies, as we show with a realistic semi-synthetic simulation study."
                },
                {
                    "title": "G-Net: a Recurrent Network Approach to G-Computation for Counterfactual Prediction Under a Dynamic Treatment Regime",
                    "abstract": "Counterfactual prediction is a fundamental task in decision-making. This paper introduces G-Net, a sequential deep learning framework for counterfactual prediction under dynamic time-varying treatment strategies in complex longitudinal settings. G-Net is based upon g-computation, a causal inference method for estimating effects of general dynamic treatment strategies. Past g-computation implementations have mostly been built using classical regression models. G-Net instead adopts a recurrent neural network framework to capture complex temporal and nonlinear dependencies in the data. To our knowledge, G-Net is the first g-computation based deep sequential modeling framework that provides estimates of treat-ment effects under dynamic and time-varying treatment strategies. We evaluate G-Net using simulated longitudinal data from two sources: CVSim, a mechanistic model of the cardiovascular system, and a pharmacokinetic simulation of tumor growth. G-Net outperforms both classical and state-of-the-art counterfactual prediction models in these settings."
                },
                {
                    "title": "Attentive State-Space Modeling of Disease Progression",
                    "abstract": "Models of disease progression are instrumental for predicting patient outcomes and understanding disease dynamics. Existing models provide the patient with pragmatic (supervised) predictions of risk, but do not provide the clinician with intelligible (unsupervised) representations of disease pathophysiology. In this paper, we develop the attentive state-space model, a deep probabilistic model that learns accurate and interpretable structured representations for disease trajectories. Unlike Markovian state-space models, in which the dynamics are memoryless, our model uses an attention mechanism to create \"memoryful\" dynamics, whereby attention weights determine the dependence of future disease states on past medical history. To learn the model parameters from medical records, we develop an infer ence algorithm that simultaneously learns a compiled inference network and the model parameters, leveraging the attentive state-space representation to construct a \"Rao-Blackwellized\" variational approximation of the posterior state distribution. Experiments on data from the UK Cystic Fibrosis registry show that our model demonstrates superior predictive accuracy and provides insights into the progression of chronic disease."
                },
                {
                    "title": "Forecasting Treatment Responses Over Time Using Recurrent Marginal Structural Networks",
                    "abstract": "Electronic health records provide a rich source of data for machine learning methods to learn dynamic treatment responses over time. However, any direct estimation is hampered by the presence of time-dependent confounding, where actions taken are dependent on time-varying variables related to the outcome of interest. Drawing inspiration from marginal structural models, a class of methods in epidemiology which use propensity weighting to adjust for time-dependent confounders, we introduce the Recurrent Marginal Structural Network - a sequence-to-sequence architecture for forecasting a patient's expected response to a series of planned treatments. Using simulations of a state-of-the-art pharmacokinetic-pharmacodynamic (PK-PD) model of tumor growth, we demonstrate the ability of our network to accurately learn unbiased treatment responses from observational data \u2013 even under changes in the policy of treatment assignments \u2013 and performance gains over benchmarks."
                },
                {
                    "title": "The More You Know, the Less You Learn: From Knowledge Transfer to One-shot Learning of Object Categories",
                    "abstract": "Learning a category from few examples is a challenging task for vision algorithms, while psychological studies have shown that humans are able to generalise correctly even from a single instance (one-shot learning). The most accredited hypothesis is that humans are able to exploit prior knowledge when learning a new related category. This paper presents an SVM-based model adaptation algorithm able to perform knowledge transfer for a new category when very limited examples are available. Using a leave- one-out estimate of the weighted error-rate the algorithm automatically decides from where to transfer (on which known category to rely), how much to transfer (the degree of adaptation) and if it is worth transferring something at all. Moreover a weighted least-squares loss function takes optimally care of data unbalance between negative and positive examples. Experiments presented on two different object category databases show that the proposed method is able to exploit previous knowledge avoiding negative transfer. The overall classification performance is increased compared to what would be achieved by starting from scratch. Furthermore as the number of already learned categories grows, the algorithm is able to learn a new category from one sample with increasing precision, i.e. it is able to perform one-shot learning."
                },
                {
                    "title": "Estimation of the causal effects of time-varying exposures",
                    "abstract": "This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint. Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying , microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers. 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Longitudinal data analysis / editors, Garrett Fitzmaurice ... [et al.]. p. cm.-(Chapman and Hall/CRC series of handbooks of modern statistical methods) Includes bibliographical references and index."
                },
                {
                    "title": "Correcting for non-compliance in randomized trials using structural nested mean models",
                    "abstract": "In a randomized trial designed to study the effect of a treatment of interest on the evolution of the mean of a time-dependent outcome variable, subjects are assigned to a treatment regime, or, equivalently, a treatment protocol. Unfortunately, subjects often fail to comply with their assigned regime. From a public health point of view, the causal parameter of interest will often be a function of the treatment differences that would have been observed hadcontrary to fact, all subjects remained on protocol. This paper considers the identification and estimation of these treatment differences based on a new class of structural models, the multivariate structural nested mean models, when reliable estimates of each subject's actual treatment are available. Estimates of \u201cactual treatment\u201d might, for example, be obtained by measuring the amount of \u201cactive drug\u201d in the subject's blood or urine at each follow-up visit or by pill counting techniques. In addition, we discuss a natural extension of our methods to ob..."
                },
                {
                    "title": "Probability theory",
                    "abstract": "One of the main objectives of a statistician is to draw conclusions about a population of objects by conducting an experiment. The first step in this endeavor is to identify the possible outcomes or, in statistical terminology, the sample space. Definition 1.1.1 The set, S, of all possible outcomes of a particular experiment is called the sample space for the experiment. If the experiment consists of tossing a coin, the sample space contains two outcomes, heads and tails; thus, S = {H, T }. Consider an experiment where the observation is reaction time to a certain stimulus. Here, the sample space would consist of all positive numbers, that is, S = (0, \u221e)."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ME"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively discover and utilize ordinary differential equations (ODEs) to model pharmacokinetic-pharmacodynamic (PKPD) relationships in cancer treatment, while accounting for variability in treatment effects?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the understanding of complex biological systems and improving treatment strategies in oncology. By accurately modeling PKPD relationships, researchers can better predict patient responses to therapies, leading to personalized medicine approaches. This work could pave the way for future research in dynamic modeling, enhance the development of new therapeutic agents, and ultimately improve patient outcomes by optimizing treatment regimens based on individual variability.\n\n### [Question 3] - Why is it hard?\nThe challenges in this research stem from the inherent complexity of biological systems, which often exhibit non-linear dynamics and significant variability among patients. Naive approaches may fail due to oversimplification of the underlying processes or inability to capture the full range of interactions between covariates and treatment effects. Technical obstacles include the need for robust numerical methods to solve ODEs accurately, as well as the requirement for comprehensive datasets that include all relevant variables. Theoretical challenges involve ensuring the correct specification of models and assumptions regarding observability and overlap in the data.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often been limited by a lack of comprehensive datasets or insufficient modeling techniques that can capture the complexity of PKPD relationships. Many existing solutions do not adequately address the variability in treatment effects or fail to incorporate the necessary assumptions for ODE discovery. Our approach differs by integrating advanced ODE discovery methods with a focus on treatment effects and variability, allowing for a more nuanced understanding of the underlying dynamics in cancer treatment.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves the use of advanced ODE discovery techniques, specifically tailored for PKPD modeling in cancer treatment. We will utilize a dataset that includes various treatment outcomes and covariates, applying metrics such as 95% confidence intervals and averaging results over multiple random seed runs to ensure robustness. The expected outcomes include the identification of accurate ODEs that describe the treatment dynamics, which will facilitate forward simulation of future values using numerical solvers like Euler's method. This will enable us to estimate patient responses to treatments more effectively, ultimately contributing to personalized treatment strategies."
            }
        },
        "author_data": {
            "4fd3d5be-a23c-47a5-aead-2f688dd0b5b5": {
                "pk": "4fd3d5be-a23c-47a5-aead-2f688dd0b5b5",
                "name": "Krzysztof Kacprzyk",
                "collaborators": [
                    "M. Schaar",
                    "Zhaozhi Qian",
                    "Jeroen Berrevoets"
                ],
                "domain": [
                    "Causal Inference",
                    "Symbolic Regression",
                    "Machine Learning",
                    "Differential Equations"
                ],
                "publications": [
                    {
                        "title": "Shape Arithmetic Expressions: Advancing Scientific Discovery Beyond Closed-Form Equations",
                        "abstract": "Symbolic regression has excelled in uncovering equations from physics, chemistry, biology, and related disciplines. However, its effectiveness becomes less certain when applied to experimental data lacking inherent closed-form expressions. Empirically derived relationships, such as entire stress-strain curves, may defy concise closed-form representation, compelling us to explore more adaptive modeling approaches that balance flexibility with interpretability. In our pursuit, we turn to Generalized Additive Models (GAMs), a widely used class of models known for their versatility across various domains. Although GAMs can capture non-linear relationships between variables and targets, they cannot capture intricate feature interactions. In this work, we investigate both of these challenges and propose a novel class of models, Shape Arithmetic Expressions (SHAREs), that fuses GAM's flexible shape functions with the complex feature interactions found in mathematical expressions. SHAREs also provide a unifying framework for both of these approaches. We also design a set of rules for constructing SHAREs that guarantee transparency of the found expressions beyond the standard constraints based on the model's size."
                    },
                    {
                        "title": "Causal Deep Learning: Encouraging Impact on Real-world Problems Through Causality",
                        "abstract": "Causality has the potential to truly transform the way we solve a large number of real-world problems. Yet, so far, its potential largely remains to be unlocked as causality often requires crucial assumptions which cannot be tested in practice. To address this challenge, we propose a new way of thinking about causality-\u2013 we call this causal deep learning . Our causal deep learning framework spans three dimensions: (1) a structural dimension, which incorporates partial yet testable causal knowledge rather than assuming either complete or no causal knowledge among the variables of interest; (2) a parametric dimension, which encompasses parametric forms that capture the type of relationships among the variables of interest; and (3) a temporal dimen-sion, which captures exposure times or how the variables of interest interact (possibly causally) over time. Our CDL framework enables us to precisely categorise and compare"
                    },
                    {
                        "title": "Causal Deep Learning",
                        "abstract": "Causality has the potential to truly transform the way we solve a large number of real-world problems. Yet, so far, its potential largely remains to be unlocked as causality often requires crucial assumptions which cannot be tested in practice. To address this challenge, we propose a new way of thinking about causality -- we call this causal deep learning. Our causal deep learning framework spans three dimensions: (1) a structural dimension, which incorporates partial yet testable causal knowledge rather than assuming either complete or no causal knowledge among the variables of interest; (2) a parametric dimension, which encompasses parametric forms that capture the type of relationships among the variables of interest; and (3) a temporal dimension, which captures exposure times or how the variables of interest interact (possibly causally) over time. Causal deep learning enables us to make progress on a variety of real-world problems by leveraging partial causal knowledge (including independencies among variables) and quantitatively characterising causal relationships among variables of interest (possibly over time). Our framework clearly identifies which assumptions are testable and which ones are not, such that the resulting solutions can be judiciously adopted in practice. Using our formulation we can combine or chain together causal representations to solve specific problems without losing track of which assumptions are required to build these solutions, pushing real-world impact in healthcare, economics and business, environmental sciences and education, through causal deep learning."
                    },
                    {
                        "title": "Navigating causal deep learning",
                        "abstract": "Causal deep learning (CDL) is a new and important research area in the larger field of machine learning. With CDL, researchers aim to structure and encode causal knowledge in the extremely flexible representation space of deep learning models. Doing so will lead to more informed, robust, and general predictions and inference -- which is important! However, CDL is still in its infancy. For example, it is not clear how we ought to compare different methods as they are so different in their output, the way they encode causal knowledge, or even how they represent this knowledge. This is a living paper that categorises methods in causal deep learning beyond Pearl's ladder of causation. We refine the rungs in Pearl's ladder, while also adding a separate dimension that categorises the parametric assumptions of both input and representation, arriving at the map of causal deep learning. Our map covers machine learning disciplines such as supervised learning, reinforcement learning, generative modelling and beyond. Our paradigm is a tool which helps researchers to: find benchmarks, compare methods, and most importantly: identify research gaps. With this work we aim to structure the avalanche of papers being published on causal deep learning. While papers on the topic are being published daily, our map remains fixed. We open-source our map for others to use as they see fit: perhaps to offer guidance in a related works section, or to better highlight the contribution of their paper."
                    },
                    {
                        "title": "D-CIPHER: Discovery of Closed-form Partial Differential Equations",
                        "abstract": "Closed-form differential equations, including partial differential equations and higher-order ordinary differential equations, are one of the most important tools used by scientists to model and better understand natural phenomena. Discovering these equations directly from data is challenging because it requires modeling relationships between various derivatives that are not observed in the data (equation-data mismatch) and it involves searching across a huge space of possible equations. Current approaches make strong assumptions about the form of the equation and thus fail to discover many well-known systems. Moreover, many of them resolve the equation-data mismatch by estimating the derivatives, which makes them inadequate for noisy and infrequently sampled systems. To this end, we propose D-CIPHER, which is robust to measurement artifacts and can uncover a new and very general class of differential equations. We further design a novel optimization procedure, CoLLie, to help D-CIPHER search through this class efficiently. Finally, we demonstrate empirically that it can discover many well-known equations that are beyond the capabilities of current methods."
                    },
                    {
                        "title": "D-CIPHER: Discovery of Closed-form PDEs",
                        "abstract": "Closed-form differential equations, including partial differential equations and higher-order ordinary differential equations, are one of the most important tools used by scientists to model and better understand natural phenomena. Discovering these equations directly from data is challenging because it requires modeling relationships between various derivatives that are not observed in the data ( equation-data mismatch ) and it involves searching across a huge space of possible equations. Current approaches make strong assumptions about the form of the equation and thus fail to discover many well-known systems. Moreover, many of them resolve the equation-data mismatch by estimating the derivatives, which makes them inadequate for noisy and infrequently sampled systems. To this end, we propose D-CIPHER, which is robust to measurement artifacts and can uncover a new and very general class of differential equations. We further design a novel optimization procedure, CoLLie, to help D-CIPHER search through this class ef\ufb01ciently. Finally, we demonstrate empirically that it can discover many well-known equations that are beyond the capabilities of current methods."
                    },
                    {
                        "title": "D-CODE: Discovering Closed-form ODEs from Observed Trajectories",
                        "abstract": "For centuries, scientists have manually designed closed-form ordinary differential equations (ODEs) to model dynamical systems. An automated tool to distill closedform ODEs from observed trajectories would accelerate the modeling process. Traditionally, symbolic regression is used to uncover a closed-form prediction function a = f(b) with label-feature pairs (ai, bi) as training examples. However, an ODE models the time derivative \u1e8b(t) of a dynamical system, e.g. \u1e8b(t) = f(x(t), t), and the \u201clabel\u201d \u1e8b(t) is usually not observed. The existing ways to bridge this gap only perform well for a narrow range of settings with low measurement noise and frequent sampling. In this work, we propose the Discovery of Closedform ODE framework (D-CODE), which advances symbolic regression beyond the paradigm of supervised learning. D-CODE uses a novel objective function based on the variational formulation of ODEs to bypass the unobserved time derivative. For formal justification, we prove that this objective is a valid proxy for the estimation error of the true (but unknown) ODE. In the experiments, D-CODE successfully discovered the governing equations of a diverse range of dynamical systems under challenging measurement settings with high noise and infrequent sampling."
                    }
                ]
            },
            "9e24e96c-7de5-4988-ba95-4d23c77d5ab9": {
                "pk": "9e24e96c-7de5-4988-ba95-4d23c77d5ab9",
                "name": "Samuel Holt",
                "collaborators": [
                    "M. Schaar",
                    "Hao Sun",
                    "Zhaozhi Qian",
                    "Alex J. Chan",
                    "Alihan H\u00fcy\u00fck",
                    "T. Pouplin",
                    "Chris Lu",
                    "Claudio Fanconi",
                    "Jakob N. Foerster",
                    "R. T. Lange"
                ],
                "domain": [
                    "Large Language Models",
                    "Reinforcement Learning",
                    "Symbolic Regression",
                    "Continuous-Time Control"
                ],
                "publications": [
                    {
                        "title": "Retrieval Augmented Thought Process for Private Data Handling in Healthcare",
                        "abstract": "Large Language Models (LLMs) have demonstrated the strong potential to assist both clinicians and the general public with their extensive medical knowledge. However, their application in healthcare is constrained due to concerns about the privacy of data used in training, which prevents the integration of private and personal information because of security and ethical issues. Moreover, if their capabilities can be enhanced with information retrieval to access up-to-date knowledge, the current integration of LLMs with Information retrieval lacks robustness to imperfect retrieval, which can hinder their effectiveness and even reduce overall performance. In this work, we address this challenge by introducing the Retrieval-Augmented Thought Process (RATP). Given access to external knowledge, RATP formulates the thought generation of LLMs as a multiple-step decision process. To optimise such a thought process, RATP leverages Monte-Carlo Tree Search and learns a proxy reward function that permits cost-efficient inference. On a private dataset of electronic medical records, deliberately excluded from any LLM training set, RATP achieves 35% additional accuracy compared to in-context retrieval-augmented generation for the question-answering task."
                    },
                    {
                        "title": "Discovering Preference Optimization Algorithms with and for Large Language Models",
                        "abstract": "Offline preference optimization is a key method for enhancing and controlling the quality of Large Language Model (LLM) outputs. Typically, preference optimization is approached as an offline supervised learning task using manually-crafted convex loss functions. While these methods are based on theoretical insights, they are inherently constrained by human creativity, so the large search space of possible loss functions remains under explored. We address this by performing LLM-driven objective discovery to automatically discover new state-of-the-art preference optimization algorithms without (expert) human intervention. Specifically, we iteratively prompt an LLM to propose and implement new preference optimization loss functions based on previously-evaluated performance metrics. This process leads to the discovery of previously-unknown and performant preference optimization algorithms. The best performing of these we call Discovered Preference Optimization (DiscoPOP), a novel algorithm that adaptively blends logistic and exponential losses. Experiments demonstrate the state-of-the-art performance of DiscoPOP and its successful transfer to held-out tasks."
                    },
                    {
                        "title": "Dense Reward for Free in Reinforcement Learning from Human Feedback",
                        "abstract": "Reinforcement Learning from Human Feedback (RLHF) has been credited as the key advance that has allowed Large Language Models (LLMs) to effectively follow instructions and produce useful assistance. Classically, this involves generating completions from the LLM in response to a query before using a separate reward model to assign a score to the full completion. As an auto-regressive process, the LLM has to take many\"actions\"(selecting individual tokens) and only receives a single, sparse reward at the end of an episode, a setup that is known to be difficult to optimise in traditional reinforcement learning. In this work we leverage the fact that the reward model contains more information than just its scalar output, in particular, it calculates an attention map over tokens as part of the transformer architecture. We use these attention weights to redistribute the reward along the whole completion, effectively densifying the signal and highlighting the most important tokens, all without incurring extra computational cost or requiring any additional modelling. We demonstrate that, theoretically, this approach is equivalent to potential-based reward shaping, ensuring that the optimal policy remains unchanged. Empirically, we show that it stabilises training, accelerates the rate of learning, and, in practical cases, may lead to better local optima."
                    },
                    {
                        "title": "L2MAC: Large Language Model Automatic Computer for Extensive Code Generation",
                        "abstract": "Transformer-based large language models (LLMs) are constrained by the fixed context window of the underlying transformer architecture, hindering their ability to produce long and coherent outputs. Memory-augmented LLMs are a promising solution, but current approaches cannot handle long output generation tasks since they (1) only focus on reading memory and reduce its evolution to the concatenation of new memories or (2) use very specialized memories that cannot adapt to other domains. This paper presents L2MAC, the first practical LLM-based general-purpose stored-program automatic computer (von Neumann architecture) framework, an LLM-based multi-agent system, for long and consistent output generation. Its memory has two components: the instruction registry, which is populated with a prompt program to solve the user-given task, and a file store, which will contain the final and intermediate outputs. Each instruction in turn is executed by a separate LLM agent, whose context is managed by a control unit capable of precise memory reading and writing to ensure effective interaction with the file store. These components enable L2MAC to generate extensive outputs, bypassing the constraints of the finite context window while producing outputs that fulfill a complex user-specified task. We empirically demonstrate that L2MAC achieves state-of-the-art performance in generating large codebases for system design tasks, significantly outperforming other coding methods in implementing the detailed user-specified task; we show that L2MAC works for general-purpose extensive text-based tasks, such as writing an entire book; and we provide valuable insights into L2MAC's performance improvement over existing methods."
                    },
                    {
                        "title": "Neural Laplace Control for Continuous-time Delayed Systems",
                        "abstract": "Many real-world offline reinforcement learning (RL) problems involve continuous-time environments with delays. Such environments are characterized by two distinctive features: firstly, the state x(t) is observed at irregular time intervals, and secondly, the current action a(t) only affects the future state x(t + g) with an unknown delay g>0. A prime example of such an environment is satellite control where the communication link between earth and a satellite causes irregular observations and delays. Existing offline RL algorithms have achieved success in environments with irregularly observed states in time or known delays. However, environments involving both irregular observations in time and unknown delays remains an open and challenging problem. To this end, we propose Neural Laplace Control, a continuous-time model-based offline RL method that combines a Neural Laplace dynamics model with a model predictive control (MPC) planner--and is able to learn from an offline dataset sampled with irregular time intervals from an environment that has a inherent unknown constant delay. We show experimentally on continuous-time delayed environments it is able to achieve near expert policy performance."
                    },
                    {
                        "title": "Deep Generative Symbolic Regression",
                        "abstract": "Symbolic regression (SR) aims to discover concise closed-form mathematical equations from data, a task fundamental to scientific discovery. However, the problem is highly challenging because closed-form equations lie in a complex combinatorial search space. Existing methods, ranging from heuristic search to reinforcement learning, fail to scale with the number of input variables. We make the observation that closed-form equations often have structural characteristics and invariances (e.g., the commutative law) that could be further exploited to build more effective symbolic regression solutions. Motivated by this observation, our key contribution is to leverage pre-trained deep generative models to capture the intrinsic regularities of equations, thereby providing a solid foundation for subsequent optimization steps. We show that our novel formalism unifies several prominent approaches of symbolic regression and offers a new perspective to justify and improve on the previous ad hoc designs, such as the usage of cross-entropy loss during pre-training. Specifically, we propose an instantiation of our framework, Deep Generative Symbolic Regression (DGSR). In our experiments, we show that DGSR achieves a higher recovery rate of true equations in the setting of a larger number of input variables, and it is more computationally efficient at inference time than state-of-the-art RL symbolic regression solutions."
                    },
                    {
                        "title": "Active Observing in Continuous-time Control",
                        "abstract": "The control of continuous-time environments while actively deciding when to take costly observations in time is a crucial yet unexplored problem, particularly relevant to real-world scenarios such as medicine, low-power systems, and resource management. Existing approaches either rely on continuous-time control methods that take regular, expensive observations in time or discrete-time control with costly observation methods, which are inapplicable to continuous-time settings due to the compounding discretization errors introduced by time discretization. In this work, we are the first to formalize the continuous-time control problem with costly observations. Our key theoretical contribution shows that observing at regular time intervals is not optimal in certain environments, while irregular observation policies yield higher expected utility. This perspective paves the way for the development of novel methods that can take irregular observations in continuous-time control with costly observations. We empirically validate our theoretical findings in various continuous-time environments, including a cancer simulation, by constructing a simple initial method to solve this new problem, with a heuristic threshold on the variance of reward rollouts in an offline continuous-time model-based model predictive control (MPC) planner. Although determining the optimal method remains an open problem, our work offers valuable insights and understanding of this unique problem, laying the foundation for future research in this area."
                    },
                    {
                        "title": "Numerical simulation projects in micromagnetics with Jupyter",
                        "abstract": "We report a case study where an existing materials science course was modified to include numerical simulation projects on the micromagnetic behavior of materials. The Ubermag micromagnetic simulation software package is used in order to solve problems computationally. The simulation software is controlled through the Python code in Jupyter notebooks. Our experience is that the self-paced problem-solving nature of the project work can facilitate a better in-depth exploration of the course contents. We discuss which aspects of the Ubermag and the project Jupyter ecosystem have been beneficial for the students' learning experience and which could be transferred to similar teaching activities in other subject areas."
                    },
                    {
                        "title": "Neural Laplace: Learning diverse classes of differential equations in the Laplace domain",
                        "abstract": "Neural Ordinary Differential Equations model dynamical systems with ODE s learned by neural networks. However, ODEs are fundamentally inadequate to model systems with long-range dependencies or discontinuities, which are common in engineering and biological systems. Broader classes of differential equations (DE) have been proposed as remedies, including delay differential equations and integro-differential equations. Furthermore, Neural ODE suffers from numerical instability when modelling stiff ODEs and ODEs with piecewise forcing functions. In this work, we propose Neural Laplace , a uni\ufb01ed framework for learning diverse classes of DEs including all the aforementioned ones. Instead of modelling the dynamics in the time domain, we model it in the Laplace domain, where the history-dependencies and discontinuities in time can be represented as summations of complex exponentials. To make learning more ef\ufb01cient, we use the geometrical stereographic map of a Riemann sphere to induce more smoothness in the Laplace domain. In the experiments, Neural Laplace shows superior performance in modelling and extrapolating the trajectories of diverse classes of DEs, including the ones with complex history dependency and abrupt changes."
                    },
                    {
                        "title": "Expertise with multisensory events eliminates the effect of biological motion rotation on audiovisual synchrony perception.",
                        "abstract": "Biological motion, in the form of point-light displays, is usually less recognizable and coherent when shown from a less natural orientation, and evidence of this disruption was recently extended to audiovisual aspects of biological motion perception. In the present study, eight drummers and eight musical novices were required to judge either the audiovisual simultaneity or the temporal order of the movements of a drumming point-light display and the resulting sound. The drumming biological motion was presented either in its upright orientation or rotated by 90, 180, or 270 degrees. Our results support and extend previous findings demonstrating that although the rotation of the point-light display affects the audiovisual aspects of biological motion, this effect disappears when experience with the represented multisensory action is increased through practice."
                    }
                ]
            },
            "37a02e70-443b-47b7-9abb-cb92f3624f9e": {
                "pk": "37a02e70-443b-47b7-9abb-cb92f3624f9e",
                "name": "Jeroen Berrevoets",
                "collaborators": [
                    "M. Schaar",
                    "Zhaozhi Qian",
                    "Krzysztof Kacprzyk",
                    "Nabeel Seedat",
                    "Tennison Liu",
                    "Alicia Curth",
                    "F. Imrie",
                    "Nicolas Huynh",
                    "Jonathan Crabbe",
                    "Daniel Jarrett"
                ],
                "domain": [
                    "Causality",
                    "Machine Learning",
                    "Deep Learning",
                    "Synthetic Data"
                ],
                "publications": [
                    {
                        "title": "Causal Deep Learning: Encouraging Impact on Real-world Problems Through Causality",
                        "abstract": "Causality has the potential to truly transform the way we solve a large number of real-world problems. Yet, so far, its potential largely remains to be unlocked as causality often requires crucial assumptions which cannot be tested in practice. To address this challenge, we propose a new way of thinking about causality-\u2013 we call this causal deep learning . Our causal deep learning framework spans three dimensions: (1) a structural dimension, which incorporates partial yet testable causal knowledge rather than assuming either complete or no causal knowledge among the variables of interest; (2) a parametric dimension, which encompasses parametric forms that capture the type of relationships among the variables of interest; and (3) a temporal dimen-sion, which captures exposure times or how the variables of interest interact (possibly causally) over time. Our CDL framework enables us to precisely categorise and compare"
                    },
                    {
                        "title": "DAGnosis: Localized Identification of Data Inconsistencies using Structures",
                        "abstract": "Identification and appropriate handling of inconsistencies in data at deployment time is crucial to reliably use machine learning models. While recent data-centric methods are able to identify such inconsistencies with respect to the training set, they suffer from two key limitations: (1) suboptimality in settings where features exhibit statistical independencies, due to their usage of compressive representations and (2) lack of localization to pin-point why a sample might be flagged as inconsistent, which is important to guide future data collection. We solve these two fundamental limitations using directed acyclic graphs (DAGs) to encode the training set's features probability distribution and independencies as a structure. Our method, called DAGnosis, leverages these structural interactions to bring valuable and insightful data-centric conclusions. DAGnosis unlocks the localization of the causes of inconsistencies on a DAG, an aspect overlooked by previous approaches. Moreover, we show empirically that leveraging these interactions (1) leads to more accurate conclusions in detecting inconsistencies, as well as (2) provides more detailed insights into why some samples are flagged."
                    },
                    {
                        "title": "AllSim: Simulating and Benchmarking Resource Allocation Policies in Multi-User Systems",
                        "abstract": "Numerous real-world systems, ranging from healthcare to energy grids, involve users competing for \ufb01nite and potentially scarce resources. Designing policies for repeated resource allocation in such real-world systems is challenging for many reasons, including the changing nature of user types and their (possibly urgent) need for resources. Researchers have developed numerous machine learning solutions for determining repeated resource allocation policies in these challenging settings. However, a key limitation has been the absence of good methods and test-beds for benchmarking these policies; almost all resource allocation policies are benchmarked in environments which are either completely synthetic or do not allow any deviation from historical data. In this paper we introduce AllSim, which is a benchmarking environment for realistically simulating the impact and utility of policies for resource allocation in systems in which users compete for such scarce resources. Building such a benchmarking environment is challenging because it needs to successfully take into account the entire collective of potential users and the impact a resource allocation policy has on all the other users in the system. AllSim\u2019s benchmarking environment is modular (each component being parameterized individually), learnable (informed by historical data), and customizable (adaptable to changing conditions). These, when interacting with an allocation policy, produce a dataset of simulated outcomes for evaluation and comparison of such policies. We believe AllSim is an essential step towards a more systematic evaluation of policies for scarce resource allocation compared to current approaches for benchmarking such methods."
                    },
                    {
                        "title": "Causal Deep Learning",
                        "abstract": "Causality has the potential to truly transform the way we solve a large number of real-world problems. Yet, so far, its potential largely remains to be unlocked as causality often requires crucial assumptions which cannot be tested in practice. To address this challenge, we propose a new way of thinking about causality -- we call this causal deep learning. Our causal deep learning framework spans three dimensions: (1) a structural dimension, which incorporates partial yet testable causal knowledge rather than assuming either complete or no causal knowledge among the variables of interest; (2) a parametric dimension, which encompasses parametric forms that capture the type of relationships among the variables of interest; and (3) a temporal dimension, which captures exposure times or how the variables of interest interact (possibly causally) over time. Causal deep learning enables us to make progress on a variety of real-world problems by leveraging partial causal knowledge (including independencies among variables) and quantitatively characterising causal relationships among variables of interest (possibly over time). Our framework clearly identifies which assumptions are testable and which ones are not, such that the resulting solutions can be judiciously adopted in practice. Using our formulation we can combine or chain together causal representations to solve specific problems without losing track of which assumptions are required to build these solutions, pushing real-world impact in healthcare, economics and business, environmental sciences and education, through causal deep learning."
                    },
                    {
                        "title": "Learning Representations without Compositional Assumptions",
                        "abstract": "This paper addresses unsupervised representation learning on tabular data containing multiple views generated by distinct sources of measurement. Traditional methods, which tackle this problem using the multi-view framework, are constrained by predefined assumptions that assume feature sets share the same information and representations should learn globally shared factors. However, this assumption is not always valid for real-world tabular datasets with complex dependencies between feature sets, resulting in localized information that is harder to learn. To overcome this limitation, we propose a data-driven approach that learns feature set dependencies by representing feature sets as graph nodes and their relationships as learnable edges. Furthermore, we introduce LEGATO, a novel hierarchical graph autoencoder that learns a smaller, latent graph to aggregate information from multiple views dynamically. This approach results in latent graph components that specialize in capturing localized information from different regions of the input, leading to superior downstream performance."
                    },
                    {
                        "title": "GOGGLE: Generative Modelling for Tabular Data by Learning Relational Structure",
                        "abstract": "Deep generative models learn highly complex and non-linear representations to generate realistic synthetic data. While they have achieved notable success in computer vision and natural language processing, similar advances have been less demonstrable in the tabular domain. This is partially because generative modelling of tabular data entails a particular set of challenges, including heterogeneous relationships, limited number of samples, and difficulties in incorporating prior knowledge. Additionally, unlike their counterparts in image and sequence domain, deep generative models for tabular data almost exclusively employ fully-connected layers, which encode weak inductive biases about relationships between inputs. Real-world data generating processes can often be represented using relational structures, which encode sparse, heterogeneous relationships between variables. In this work, we learn and exploit relational structure underlying tabular data (where typical dimensionality d < 100) to better model variable dependence, and as a natural means to introduce regularization on relationships and include prior knowledge. Specifically, we introduce GOGGLE, an end-to-end message passing scheme that jointly learns the relational structure and corresponding functional relationships as the basis of generating synthetic samples. Using real-world datasets, we provide empirical evidence that the proposed method is effective in generating realistic synthetic data and exploiting domain knowledge for downstream tasks."
                    },
                    {
                        "title": "Combining Observational and Randomized Data for Estimating Heterogeneous Treatment Effects",
                        "abstract": "Estimating heterogeneous treatment e\ufb00ects is an important problem across many domains. In order to accurately estimate such treatment e\ufb00ects, one typically relies on data from observational studies or randomized experiments. Currently, most existing works rely exclusively on observational data, which is often confounded and, hence, yields biased estimates. While observational data is confounded, randomized data is unconfounded, but its sample size is usually too small to learn heterogeneous treatment e\ufb00ects. In this paper, we propose to estimate heterogeneous treatment e\ufb00ects by combining large amounts of observational data and small amounts of randomized data via representation learning. In particular, we introduce a two-step framework: \ufb01rst, we use observational data to learn a shared structure (in form of a representation); and then, we use randomized data to learn the data-speci\ufb01c structures. We analyze the \ufb01nite sample properties of our framework and compare them to several natural baselines. As such, we derive conditions for when combining observational and randomized data is bene\ufb01cial, and for when it is not. Based on this, we introduce a sample-e\ufb03cient algorithm, called CorNet. We use extensive simulation studies to verify the theoretical properties of CorNet and multiple real-world datasets to demonstrate our method\u2019s superiority compared to existing methods."
                    },
                    {
                        "title": "To Impute or not to Impute? - Missing Data in Treatment Effect Estimation",
                        "abstract": "Missing data is a systemic problem in practical scenarios that causes noise and bias when estimating treatment effects. This makes treatment effect estimation from data with missingness a particularly tricky endeavour. A key reason for this is that standard assumptions on missingness are rendered insufficient due to the presence of an additional variable, treatment, besides the input (e.g. an individual) and the label (e.g. an outcome). The treatment variable introduces additional complexity with respect to why some variables are missing that is not fully explored by previous work. In our work we introduce mixed confounded missingness (MCM), a new missingness mechanism where some missingness determines treatment selection and other missingness is determined by treatment selection. Given MCM, we show that naively imputing all data leads to poor performing treatment effects models, as the act of imputation effectively removes information necessary to provide unbiased estimates. However, no imputation at all also leads to biased estimates, as missingness determined by treatment introduces bias in covariates. Our solution is selective imputation, where we use insights from MCM to inform precisely which variables should be imputed and which should not. We empirically demonstrate how various learners benefit from selective imputation compared to other solutions for missing data. We highlight that our experiments encompass both average treatment effects and conditional average treatment effects."
                    },
                    {
                        "title": "Differentiable and Transportable Structure Learning",
                        "abstract": "Directed acyclic graphs (DAGs) encode a lot of information about a particular distribution in their structure. However, compute required to infer these structures is typically super-exponential in the number of variables, as inference requires a sweep of a combinatorially large space of potential structures. That is, until recent advances made it possible to search this space using a differentiable metric, drastically reducing search time. While this technique -- named NOTEARS -- is widely considered a seminal work in DAG-discovery, it concedes an important property in favour of differentiability: transportability. To be transportable, the structures discovered on one dataset must apply to another dataset from the same domain. We introduce D-Struct which recovers transportability in the discovered structures through a novel architecture and loss function while remaining fully differentiable. Because D-Struct remains differentiable, our method can be easily adopted in existing differentiable architectures, as was previously done with NOTEARS. In our experiments, we empirically validate D-Struct with respect to edge accuracy and structural Hamming distance in a variety of settings."
                    },
                    {
                        "title": "Navigating causal deep learning",
                        "abstract": "Causal deep learning (CDL) is a new and important research area in the larger field of machine learning. With CDL, researchers aim to structure and encode causal knowledge in the extremely flexible representation space of deep learning models. Doing so will lead to more informed, robust, and general predictions and inference -- which is important! However, CDL is still in its infancy. For example, it is not clear how we ought to compare different methods as they are so different in their output, the way they encode causal knowledge, or even how they represent this knowledge. This is a living paper that categorises methods in causal deep learning beyond Pearl's ladder of causation. We refine the rungs in Pearl's ladder, while also adding a separate dimension that categorises the parametric assumptions of both input and representation, arriving at the map of causal deep learning. Our map covers machine learning disciplines such as supervised learning, reinforcement learning, generative modelling and beyond. Our paradigm is a tool which helps researchers to: find benchmarks, compare methods, and most importantly: identify research gaps. With this work we aim to structure the avalanche of papers being published on causal deep learning. While papers on the topic are being published daily, our map remains fixed. We open-source our map for others to use as they see fit: perhaps to offer guidance in a related works section, or to better highlight the contribution of their paper."
                    },
                    {
                        "title": "Disentangled Counterfactual Recurrent Networks for Treatment Effect Inference over Time",
                        "abstract": "Choosing the best treatment-plan for each individual patient requires accurate forecasts of their outcome trajectories as a function of the treatment, over time. While large observational data sets constitute rich sources of information to learn from, they also contain biases as treatments are rarely assigned randomly in practice. To provide accurate and unbiased forecasts, we introduce the Disentangled Counterfactual Recurrent Network (DCRN), a novel sequence-to-sequence architecture that estimates treatment outcomes over time by learning representations of patient histories that are disentangled into three separate latent factors: a treatment factor, influencing only treatment selection; an outcome factor, influencing only the outcome; and a confounding factor, influencing both. With an architecture that is completely inspired by the causal structure of treatment influence over time, we advance forecast accuracy and disease understanding, as our architecture allows for practitioners to infer which patient features influence which part in a patient's trajectory, contrasting other approaches in this domain. We demonstrate that DCRN outperforms current state-of-the-art methods in forecasting treatment responses, on both real and simulated data."
                    },
                    {
                        "title": "DECAF: Generating Fair Synthetic Data Using Causally-Aware Generative Networks",
                        "abstract": "Machine learning models have been criticized for reflecting unfair biases in the training data. Instead of solving for this by introducing fair learning algorithms directly, we focus on generating fair synthetic data, such that any downstream learner is fair. Generating fair synthetic data from unfair data - while remaining truthful to the underlying data-generating process (DGP) - is non-trivial. In this paper, we introduce DECAF: a GAN-based fair synthetic data generator for tabular data. With DECAF we embed the DGP explicitly as a structural causal model in the input layers of the generator, allowing each variable to be reconstructed conditioned on its causal parents. This procedure enables inference time debiasing, where biased edges can be strategically removed for satisfying user-defined fairness requirements. The DECAF framework is versatile and compatible with several popular definitions of fairness. In our experiments, we show that DECAF successfully removes undesired bias and - in contrast to existing methods - is capable of generating high-quality synthetic data. Furthermore, we provide theoretical guarantees on the generator's convergence and the fairness of downstream models."
                    },
                    {
                        "title": "Identifiable Energy-based Representations: An Application to Estimating Heterogeneous Causal Effects",
                        "abstract": "Conditional average treatment effects (CATEs) allow us to understand the effect heterogeneity across a large population of individuals. However, typical CATE learners assume all confounding variables are measured in order for the CATE to be identifiable. This requirement can be satisfied by collecting many variables, at the expense of increased sample complexity for estimating CATEs. To combat this, we propose an energy-based model (EBM) that learns a low-dimensional representation of the variables by employing a noise contrastive loss function. With our EBM we introduce a preprocessing step that alleviates the dimensionality curse for any existing learner developed for estimating CATEs. We prove that our EBM keeps the representations partially identifiable up to some universal constant, as well as having universal approximation capability. These properties enable the representations to converge and keep the CATE estimates consistent. Experiments demonstrate the convergence of the representations, as well as show that estimating CATEs on our representations performs better than on the variables or the representations obtained through other dimensionality reduction methods."
                    }
                ]
            },
            "b0133ca7-9719-456a-b32f-cafbcb91c82a": {
                "pk": "b0133ca7-9719-456a-b32f-cafbcb91c82a",
                "name": "Zhaozhi Qian",
                "collaborators": [
                    "M. Schaar",
                    "Jeroen Berrevoets",
                    "Alihan H\u00fcy\u00fck",
                    "Krzysztof Kacprzyk",
                    "Nabeel Seedat",
                    "B. Cebere",
                    "B. V. Breugel",
                    "Samuel Holt",
                    "Hao Sun",
                    "Tennison Liu"
                ],
                "domain": [
                    "Causal Inference",
                    "Synthetic Data",
                    "Machine Learning",
                    "Healthcare Analytics"
                ],
                "publications": [
                    {
                        "title": "Causal Deep Learning: Encouraging Impact on Real-world Problems Through Causality",
                        "abstract": "Causality has the potential to truly transform the way we solve a large number of real-world problems. Yet, so far, its potential largely remains to be unlocked as causality often requires crucial assumptions which cannot be tested in practice. To address this challenge, we propose a new way of thinking about causality-\u2013 we call this causal deep learning . Our causal deep learning framework spans three dimensions: (1) a structural dimension, which incorporates partial yet testable causal knowledge rather than assuming either complete or no causal knowledge among the variables of interest; (2) a parametric dimension, which encompasses parametric forms that capture the type of relationships among the variables of interest; and (3) a temporal dimen-sion, which captures exposure times or how the variables of interest interact (possibly causally) over time. Our CDL framework enables us to precisely categorise and compare"
                    },
                    {
                        "title": "Adaptive Experiment Design with Synthetic Controls",
                        "abstract": "Clinical trials are typically run in order to understand the effects of a new treatment on a given population of patients. However, patients in large populations rarely respond the same way to the same treatment. This heterogeneity in patient responses necessitates trials that investigate effects on multiple subpopulations - especially when a treatment has marginal or no benefit for the overall population but might have significant benefit for a particular subpopulation. Motivated by this need, we propose Syntax, an exploratory trial design that identifies subpopulations with positive treatment effect among many subpopulations. Syntax is sample efficient as it (i) recruits and allocates patients adaptively and (ii) estimates treatment effects by forming synthetic controls for each subpopulation that combines control samples from other subpopulations. We validate the performance of Syntax and provide insights into when it might have an advantage over conventional trial designs through experiments."
                    },
                    {
                        "title": "DAGnosis: Localized Identification of Data Inconsistencies using Structures",
                        "abstract": "Identification and appropriate handling of inconsistencies in data at deployment time is crucial to reliably use machine learning models. While recent data-centric methods are able to identify such inconsistencies with respect to the training set, they suffer from two key limitations: (1) suboptimality in settings where features exhibit statistical independencies, due to their usage of compressive representations and (2) lack of localization to pin-point why a sample might be flagged as inconsistent, which is important to guide future data collection. We solve these two fundamental limitations using directed acyclic graphs (DAGs) to encode the training set's features probability distribution and independencies as a structure. Our method, called DAGnosis, leverages these structural interactions to bring valuable and insightful data-centric conclusions. DAGnosis unlocks the localization of the causes of inconsistencies on a DAG, an aspect overlooked by previous approaches. Moreover, we show empirically that leveraging these interactions (1) leads to more accurate conclusions in detecting inconsistencies, as well as (2) provides more detailed insights into why some samples are flagged."
                    },
                    {
                        "title": "Synthetic data for privacy-preserving clinical risk prediction",
                        "abstract": "Synthetic data promise privacy-preserving data sharing for healthcare research and development. Compared with other privacy-enhancing approaches - such as federated learning - analyses performed on synthetic data can be applied downstream without modification, such that synthetic data can act in place of real data for a wide range of use cases. However, the role that synthetic data might play in all aspects of clinical model development remains unknown. In this work, we used state-of-the-art generators explicitly designed for privacy preservation to create a synthetic version of the UK Biobank before building prognostic models for lung cancer under several data release assumptions. We demonstrate that synthetic data can be effectively used throughout the modelling pipeline even without eventual access to the real data. Furthermore, we show the implications of different data release approaches on how synthetic data could be deployed within the healthcare system."
                    },
                    {
                        "title": "When to Make and Break Commitments?",
                        "abstract": "In many scenarios, decision-makers must commit to long-term actions until their resolution before receiving the payoff of said actions, and usually, staying committed to such actions incurs continual costs. For instance, in healthcare, a newlydiscovered treatment cannot be marketed to patients until a clinical trial is conducted, which both requires time and is also costly. Of course in such scenarios, not all commitments eventually pay off. For instance, a clinical trial might end up failing to show efficacy. Given the time pressure created by the continual cost of keeping a commitment, we aim to answer: When should a decision-maker break a commitment that is likely to fail\u2014either to make an alternative commitment or to make no further commitments at all? First, we formulate this question as a new type of optimal stopping/switching problem called the optimal commitment problem (OCP). Then, we theoretically analyze OCP, and based on the insight we gain, propose a practical algorithm for solving it. Finally, we empirically evaluate the performance of our algorithm in running clinical trials with subpopulation selection."
                    },
                    {
                        "title": "Causal Deep Learning",
                        "abstract": "Causality has the potential to truly transform the way we solve a large number of real-world problems. Yet, so far, its potential largely remains to be unlocked as causality often requires crucial assumptions which cannot be tested in practice. To address this challenge, we propose a new way of thinking about causality -- we call this causal deep learning. Our causal deep learning framework spans three dimensions: (1) a structural dimension, which incorporates partial yet testable causal knowledge rather than assuming either complete or no causal knowledge among the variables of interest; (2) a parametric dimension, which encompasses parametric forms that capture the type of relationships among the variables of interest; and (3) a temporal dimension, which captures exposure times or how the variables of interest interact (possibly causally) over time. Causal deep learning enables us to make progress on a variety of real-world problems by leveraging partial causal knowledge (including independencies among variables) and quantitatively characterising causal relationships among variables of interest (possibly over time). Our framework clearly identifies which assumptions are testable and which ones are not, such that the resulting solutions can be judiciously adopted in practice. Using our formulation we can combine or chain together causal representations to solve specific problems without losing track of which assumptions are required to build these solutions, pushing real-world impact in healthcare, economics and business, environmental sciences and education, through causal deep learning."
                    },
                    {
                        "title": "Synthetic data, real errors: how (not) to publish and use synthetic data",
                        "abstract": "Generating synthetic data through generative models is gaining interest in the ML community and beyond, promising a future where datasets can be tailored to individual needs. Unfortunately, synthetic data is usually not perfect, resulting in potential errors in downstream tasks. In this work we explore how the generative process affects the downstream ML task. We show that the naive synthetic data approach -- using synthetic data as if it is real -- leads to downstream models and analyses that do not generalize well to real data. As a first step towards better ML in the synthetic data regime, we introduce Deep Generative Ensemble (DGE) -- a framework inspired by Deep Ensembles that aims to implicitly approximate the posterior distribution over the generative process model parameters. DGE improves downstream model training, evaluation, and uncertainty quantification, vastly outperforming the naive approach on average. The largest improvements are achieved for minority classes and low-density regions of the original data, for which the generative uncertainty is largest."
                    },
                    {
                        "title": "Do Diffusion Models Suffer Error Propagation? Theoretical Analysis and Consistency Regularization",
                        "abstract": "While diffusion models have achieved promising performances in data synthesis, they might suffer error propagation because of their cascade structure , where the distributional mismatch spreads and magnifies through the chain of denoising modules. However, a strict analysis is expected since many sequential models such as Conditional Random Field (CRF) are free from error propagation. In this paper, we empirically and theoretically verify that diffusion models are indeed affected by error propagation and we then propose a regularization to address this problem. Our theoretical analysis reveals that the question can be reduced to whether every denoising module of the diffusion model is fault-tolerant. We derive insightful transition equations, indicating that the module can\u2019t recover from input errors and even propagates additional errors to the next module. Our analysis directly leads to a consistency regularization scheme for diffusion models, which explicitly reduces the distribution gap between forward and backward processes. We further introduce a bootstrapping algorithm to reduce the computation cost of the regularizer. Our experimental results on multiple image datasets show that our regularization effectively handles error propagation and significantly improves the performance of vanilla diffusion models."
                    },
                    {
                        "title": "Neural Laplace Control for Continuous-time Delayed Systems",
                        "abstract": "Many real-world offline reinforcement learning (RL) problems involve continuous-time environments with delays. Such environments are characterized by two distinctive features: firstly, the state x(t) is observed at irregular time intervals, and secondly, the current action a(t) only affects the future state x(t + g) with an unknown delay g>0. A prime example of such an environment is satellite control where the communication link between earth and a satellite causes irregular observations and delays. Existing offline RL algorithms have achieved success in environments with irregularly observed states in time or known delays. However, environments involving both irregular observations in time and unknown delays remains an open and challenging problem. To this end, we propose Neural Laplace Control, a continuous-time model-based offline RL method that combines a Neural Laplace dynamics model with a model predictive control (MPC) planner--and is able to learn from an offline dataset sampled with irregular time intervals from an environment that has a inherent unknown constant delay. We show experimentally on continuous-time delayed environments it is able to achieve near expert policy performance."
                    },
                    {
                        "title": "Learning Representations without Compositional Assumptions",
                        "abstract": "This paper addresses unsupervised representation learning on tabular data containing multiple views generated by distinct sources of measurement. Traditional methods, which tackle this problem using the multi-view framework, are constrained by predefined assumptions that assume feature sets share the same information and representations should learn globally shared factors. However, this assumption is not always valid for real-world tabular datasets with complex dependencies between feature sets, resulting in localized information that is harder to learn. To overcome this limitation, we propose a data-driven approach that learns feature set dependencies by representing feature sets as graph nodes and their relationships as learnable edges. Furthermore, we introduce LEGATO, a novel hierarchical graph autoencoder that learns a smaller, latent graph to aggregate information from multiple views dynamically. This approach results in latent graph components that specialize in capturing localized information from different regions of the input, leading to superior downstream performance."
                    },
                    {
                        "title": "Synthcity: a benchmark framework for diverse use cases of tabular synthetic data",
                        "abstract": "Accessible high-quality data is the bread and butter of machine learning research, 1 and the demand for data has exploded as larger and more advanced ML models are 2 built across different domains. Yet, real data often contain sensitive information, 3 subject to various biases, and are costly to acquire, which compromise their quality 4 and accessibility. Synthetic data have thus emerged as a complement, sometimes 5 even a replacement, to real data for ML training. However, the landscape of 6 synthetic data research has been fragmented due to the large number of data"
                    },
                    {
                        "title": "Clairvoyance: A Pipeline Toolkit for Medical Time Series",
                        "abstract": "Time-series learning is the bread and butter of data-driven *clinical decision support*, and the recent explosion in ML research has demonstrated great potential in various healthcare settings. At the same time, medical time-series problems in the wild are challenging due to their highly *composite* nature: They entail design choices and interactions among components that preprocess data, impute missing values, select features, issue predictions, estimate uncertainty, and interpret models. Despite exponential growth in electronic patient data, there is a remarkable gap between the potential and realized utilization of ML for clinical research and decision support. In particular, orchestrating a real-world project lifecycle poses challenges in engineering (i.e. hard to build), evaluation (i.e. hard to assess), and efficiency (i.e. hard to optimize). Designed to address these issues simultaneously, Clairvoyance proposes a unified, end-to-end, autoML-friendly pipeline that serves as a (i) software toolkit, (ii) empirical standard, and (iii) interface for optimization. Our ultimate goal lies in facilitating transparent and reproducible experimentation with complex inference workflows, providing integrated pathways for (1) personalized prediction, (2) treatment-effect estimation, and (3) information acquisition. Through illustrative examples on real-world data in outpatient, general wards, and intensive-care settings, we illustrate the applicability of the pipeline paradigm on core tasks in the healthcare journey. To the best of our knowledge, Clairvoyance is the first to demonstrate viability of a comprehensive and automatable pipeline for clinical time-series ML."
                    },
                    {
                        "title": "Deep Generative Symbolic Regression",
                        "abstract": "Symbolic regression (SR) aims to discover concise closed-form mathematical equations from data, a task fundamental to scientific discovery. However, the problem is highly challenging because closed-form equations lie in a complex combinatorial search space. Existing methods, ranging from heuristic search to reinforcement learning, fail to scale with the number of input variables. We make the observation that closed-form equations often have structural characteristics and invariances (e.g., the commutative law) that could be further exploited to build more effective symbolic regression solutions. Motivated by this observation, our key contribution is to leverage pre-trained deep generative models to capture the intrinsic regularities of equations, thereby providing a solid foundation for subsequent optimization steps. We show that our novel formalism unifies several prominent approaches of symbolic regression and offers a new perspective to justify and improve on the previous ad hoc designs, such as the usage of cross-entropy loss during pre-training. Specifically, we propose an instantiation of our framework, Deep Generative Symbolic Regression (DGSR). In our experiments, we show that DGSR achieves a higher recovery rate of true equations in the setting of a larger number of input variables, and it is more computationally efficient at inference time than state-of-the-art RL symbolic regression solutions."
                    },
                    {
                        "title": "GOGGLE: Generative Modelling for Tabular Data by Learning Relational Structure",
                        "abstract": "Deep generative models learn highly complex and non-linear representations to generate realistic synthetic data. While they have achieved notable success in computer vision and natural language processing, similar advances have been less demonstrable in the tabular domain. This is partially because generative modelling of tabular data entails a particular set of challenges, including heterogeneous relationships, limited number of samples, and difficulties in incorporating prior knowledge. Additionally, unlike their counterparts in image and sequence domain, deep generative models for tabular data almost exclusively employ fully-connected layers, which encode weak inductive biases about relationships between inputs. Real-world data generating processes can often be represented using relational structures, which encode sparse, heterogeneous relationships between variables. In this work, we learn and exploit relational structure underlying tabular data (where typical dimensionality d < 100) to better model variable dependence, and as a natural means to introduce regularization on relationships and include prior knowledge. Specifically, we introduce GOGGLE, an end-to-end message passing scheme that jointly learns the relational structure and corresponding functional relationships as the basis of generating synthetic samples. Using real-world datasets, we provide empirical evidence that the proposed method is effective in generating realistic synthetic data and exploiting domain knowledge for downstream tasks."
                    },
                    {
                        "title": "Membership Inference Attacks against Synthetic Data through Overfitting Detection",
                        "abstract": "Data is the foundation of most science. Unfortunately, sharing data can be obstructed by the risk of violating data privacy, impeding research in fields like healthcare. Synthetic data is a potential solution. It aims to generate data that has the same distribution as the original data, but that does not disclose information about individuals. Membership Inference Attacks (MIAs) are a common privacy attack, in which the attacker attempts to determine whether a particular real sample was used for training of the model. Previous works that propose MIAs against generative models either display low performance -- giving the false impression that data is highly private -- or need to assume access to internal generative model parameters -- a relatively low-risk scenario, as the data publisher often only releases synthetic data, not the model. In this work we argue for a realistic MIA setting that assumes the attacker has some knowledge of the underlying data distribution. We propose DOMIAS, a density-based MIA model that aims to infer membership by targeting local overfitting of the generative model. Experimentally we show that DOMIAS is significantly more successful at MIA than previous work, especially at attacking uncommon samples. The latter is disconcerting since these samples may correspond to underrepresented groups. We also demonstrate how DOMIAS' MIA performance score provides an interpretable metric for privacy, giving data publishers a new tool for achieving the desired privacy-utility trade-off in their synthetic data."
                    },
                    {
                        "title": "Synthcity: facilitating innovative use cases of synthetic data in different data modalities",
                        "abstract": "Synthcity is an open-source software package for innovative use cases of synthetic data in ML fairness, privacy and augmentation across diverse tabular data modalities, including static data, regular and irregular time series, data with censoring, multi-source data, composite data, and more. Synthcity provides the practitioners with a single access point to cutting edge research and tools in synthetic data. It also offers the community a playground for rapid experimentation and prototyping, a one-stop-shop for SOTA benchmarks, and an opportunity for extending research impact. The library can be accessed on GitHub (https://github.com/vanderschaarlab/synthcity) and pip (https://pypi.org/project/synthcity/). We warmly invite the community to join the development effort by providing feedback, reporting bugs, and contributing code."
                    },
                    {
                        "title": "TRIAGE: Characterizing and auditing training data for improved regression",
                        "abstract": "Data quality is crucial for robust machine learning algorithms, with the recent interest in data-centric AI emphasizing the importance of training data characterization. However, current data characterization methods are largely focused on classification settings, with regression settings largely understudied. To address this, we introduce TRIAGE, a novel data characterization framework tailored to regression tasks and compatible with a broad class of regressors. TRIAGE utilizes conformal predictive distributions to provide a model-agnostic scoring method, the TRIAGE score. We operationalize the score to analyze individual samples' training dynamics and characterize samples as under-, over-, or well-estimated by the model. We show that TRIAGE's characterization is consistent and highlight its utility to improve performance via data sculpting/filtering, in multiple regression settings. Additionally, beyond sample level, we show TRIAGE enables new approaches to dataset selection and feature acquisition. Overall, TRIAGE highlights the value unlocked by data characterization in real-world regression applications"
                    },
                    {
                        "title": "Navigating causal deep learning",
                        "abstract": "Causal deep learning (CDL) is a new and important research area in the larger field of machine learning. With CDL, researchers aim to structure and encode causal knowledge in the extremely flexible representation space of deep learning models. Doing so will lead to more informed, robust, and general predictions and inference -- which is important! However, CDL is still in its infancy. For example, it is not clear how we ought to compare different methods as they are so different in their output, the way they encode causal knowledge, or even how they represent this knowledge. This is a living paper that categorises methods in causal deep learning beyond Pearl's ladder of causation. We refine the rungs in Pearl's ladder, while also adding a separate dimension that categorises the parametric assumptions of both input and representation, arriving at the map of causal deep learning. Our map covers machine learning disciplines such as supervised learning, reinforcement learning, generative modelling and beyond. Our paradigm is a tool which helps researchers to: find benchmarks, compare methods, and most importantly: identify research gaps. With this work we aim to structure the avalanche of papers being published on causal deep learning. While papers on the topic are being published daily, our map remains fixed. We open-source our map for others to use as they see fit: perhaps to offer guidance in a related works section, or to better highlight the contribution of their paper."
                    },
                    {
                        "title": "MARS: Assisting Human with Information Processing Tasks Using Machine Learning",
                        "abstract": "This article studies the problem of automated information processing from large volumes of unstructured, heterogeneous, and sometimes untrustworthy data sources. The main contribution is a novel framework called Machine Assisted Record Selection (MARS). Instead of today\u2019s standard practice of relying on human experts to manually decide the order of records for processing, MARS learns the optimal record selection via an online learning algorithm. It further integrates algorithm-based record selection and processing with human-based error resolution to achieve a balanced task allocation between machine and human. Both fixed and adaptive MARS algorithms are proposed, leveraging different statistical knowledge about the existence, quality, and cost associated with the records. Experiments using semi-synthetic data that are generated from real-world patients record processing in the UK national cancer registry are carried out, which demonstrate significant (3 to 4 fold) performance gain over the fixed-order processing. MARS represents one of the few examples demonstrating that machine learning can assist humans with complex jobs by automating complex triaging tasks."
                    }
                ]
            },
            "622c06c9-328b-4b4b-a299-f8918699165e": {
                "pk": "622c06c9-328b-4b4b-a299-f8918699165e",
                "name": "Mihaela van der Schaar",
                "collaborators": [
                    "Samuel Holt",
                    "Yangming Li",
                    "Nabeel Seedat",
                    "Nicolas Huynh",
                    "T. Pouplin",
                    "Hao Sun",
                    "Krzysztof Kacprzyk",
                    "Tennison Liu",
                    "Max Ruiz Luyten",
                    "Alex J. Chan"
                ],
                "domain": [
                    "Semi-Supervised Learning",
                    "Causal Inference",
                    "Large Language Models",
                    "Meta-Learning"
                ],
                "publications": [
                    {
                        "title": "You can't handle the (dirty) truth: Data-centric insights improve pseudo-labeling",
                        "abstract": "Pseudo-labeling is a popular semi-supervised learning technique to leverage unlabeled data when labeled samples are scarce. The generation and selection of pseudo-labels heavily rely on labeled data. Existing approaches implicitly assume that the labeled data is gold standard and 'perfect'. However, this can be violated in reality with issues such as mislabeling or ambiguity. We address this overlooked aspect and show the importance of investigating labeled data quality to improve any pseudo-labeling method. Specifically, we introduce a novel data characterization and selection framework called DIPS to extend pseudo-labeling. We select useful labeled and pseudo-labeled samples via analysis of learning dynamics. We demonstrate the applicability and impact of DIPS for various pseudo-labeling methods across an extensive range of real-world tabular and image datasets. Additionally, DIPS improves data efficiency and reduces the performance distinctions between different pseudo-labelers. Overall, we highlight the significant benefits of a data-centric rethinking of pseudo-labeling in real-world settings."
                    },
                    {
                        "title": "Retrieval Augmented Thought Process for Private Data Handling in Healthcare",
                        "abstract": "Large Language Models (LLMs) have demonstrated the strong potential to assist both clinicians and the general public with their extensive medical knowledge. However, their application in healthcare is constrained due to concerns about the privacy of data used in training, which prevents the integration of private and personal information because of security and ethical issues. Moreover, if their capabilities can be enhanced with information retrieval to access up-to-date knowledge, the current integration of LLMs with Information retrieval lacks robustness to imperfect retrieval, which can hinder their effectiveness and even reduce overall performance. In this work, we address this challenge by introducing the Retrieval-Augmented Thought Process (RATP). Given access to external knowledge, RATP formulates the thought generation of LLMs as a multiple-step decision process. To optimise such a thought process, RATP leverages Monte-Carlo Tree Search and learns a proxy reward function that permits cost-efficient inference. On a private dataset of electronic medical records, deliberately excluded from any LLM training set, RATP achieves 35% additional accuracy compared to in-context retrieval-augmented generation for the question-answering task."
                    },
                    {
                        "title": "Meta-Learners for Partially-Identified Treatment Effects Across Multiple Environments",
                        "abstract": "Estimating the conditional average treatment effect (CATE) from observational data is relevant for many applications such as personalized medicine. Here, we focus on the widespread setting where the observational data come from multiple environments, such as different hospitals, physicians, or countries. Furthermore, we allow for violations of standard causal assumptions, namely, overlap within the environments and unconfoundedness. To this end, we move away from point identification and focus on partial identification. Specifically, we show that current assumptions from the literature on multiple environments allow us to interpret the environment as an instrumental variable (IV). This allows us to adapt bounds from the IV literature for partial identification of CATE by leveraging treatment assignment mechanisms across environments. Then, we propose different model-agnostic learners (so-called meta-learners) to estimate the bounds that can be used in combination with arbitrary machine learning models. We further demonstrate the effectiveness of our meta-learners across various experiments using both simulated and real-world data. Finally, we discuss the applicability of our meta-learners to partial identification in instrumental variable settings, such as randomized controlled trials with non-compliance."
                    },
                    {
                        "title": "Shape Arithmetic Expressions: Advancing Scientific Discovery Beyond Closed-Form Equations",
                        "abstract": "Symbolic regression has excelled in uncovering equations from physics, chemistry, biology, and related disciplines. However, its effectiveness becomes less certain when applied to experimental data lacking inherent closed-form expressions. Empirically derived relationships, such as entire stress-strain curves, may defy concise closed-form representation, compelling us to explore more adaptive modeling approaches that balance flexibility with interpretability. In our pursuit, we turn to Generalized Additive Models (GAMs), a widely used class of models known for their versatility across various domains. Although GAMs can capture non-linear relationships between variables and targets, they cannot capture intricate feature interactions. In this work, we investigate both of these challenges and propose a novel class of models, Shape Arithmetic Expressions (SHAREs), that fuses GAM's flexible shape functions with the complex feature interactions found in mathematical expressions. SHAREs also provide a unifying framework for both of these approaches. We also design a set of rules for constructing SHAREs that guarantee transparency of the found expressions beyond the standard constraints based on the model's size."
                    },
                    {
                        "title": "Large Language Models to Enhance Bayesian Optimization",
                        "abstract": "Bayesian optimization (BO) is a powerful approach for optimizing complex and expensive-to-evaluate black-box functions. Its importance is underscored in many applications, notably including hyperparameter tuning, but its efficacy depends on efficiently balancing exploration and exploitation. While there has been substantial progress in BO methods, striking this balance remains a delicate process. In this light, we present LLAMBO, a novel approach that integrates the capabilities of Large Language Models (LLM) within BO. At a high level, we frame the BO problem in natural language, enabling LLMs to iteratively propose and evaluate promising solutions conditioned on historical evaluations. More specifically, we explore how combining contextual understanding, few-shot learning proficiency, and domain knowledge of LLMs can improve model-based BO. Our findings illustrate that LLAMBO is effective at zero-shot warmstarting, and enhances surrogate modeling and candidate sampling, especially in the early stages of search when observations are sparse. Our approach is performed in context and does not require LLM finetuning. Additionally, it is modular by design, allowing individual components to be integrated into existing BO frameworks, or function cohesively as an end-to-end method. We empirically validate LLAMBO's efficacy on the problem of hyperparameter tuning, highlighting strong empirical performance across a range of diverse benchmarks, proprietary, and synthetic tasks."
                    },
                    {
                        "title": "Causal Deep Learning: Encouraging Impact on Real-world Problems Through Causality",
                        "abstract": "Causality has the potential to truly transform the way we solve a large number of real-world problems. Yet, so far, its potential largely remains to be unlocked as causality often requires crucial assumptions which cannot be tested in practice. To address this challenge, we propose a new way of thinking about causality-\u2013 we call this causal deep learning . Our causal deep learning framework spans three dimensions: (1) a structural dimension, which incorporates partial yet testable causal knowledge rather than assuming either complete or no causal knowledge among the variables of interest; (2) a parametric dimension, which encompasses parametric forms that capture the type of relationships among the variables of interest; and (3) a temporal dimen-sion, which captures exposure times or how the variables of interest interact (possibly causally) over time. Our CDL framework enables us to precisely categorise and compare"
                    },
                    {
                        "title": "Risk-Sensitive Diffusion: Robustly Optimizing Diffusion Models with Noisy Samples",
                        "abstract": "Diffusion models are mainly studied on image data. However, non-image data (e.g., tabular data) are also prevalent in real applications and tend to be noisy due to some inevitable factors in the stage of data collection, degrading the generation quality of diffusion models. In this paper, we consider a novel problem setting where every collected sample is paired with a vector indicating the data quality: risk vector. This setting applies to many scenarios involving noisy data and we propose risk-sensitive SDE, a type of stochastic differential equation (SDE) parameterized by the risk vector, to address it. With some proper coefficients, risk-sensitive SDE can minimize the negative effect of noisy samples on the optimization of diffusion models. We conduct systematic studies for both Gaussian and non-Gaussian noise distributions, providing analytical forms of risk-sensitive SDE. To verify the effectiveness of our method, we have conducted extensive experiments on multiple tabular and time-series datasets, showing that risk-sensitive SDE permits a robust optimization of diffusion models with noisy samples and significantly outperforms previous baselines."
                    },
                    {
                        "title": "CliMB: An AI-enabled Partner for Clinical Predictive Modeling",
                        "abstract": "Despite its significant promise and continuous technical advances, real-world applications of artificial intelligence (AI) remain limited. We attribute this to the\"domain expert-AI-conundrum\": while domain experts, such as clinician scientists, should be able to build predictive models such as risk scores, they face substantial barriers in accessing state-of-the-art (SOTA) tools. While automated machine learning (AutoML) has been proposed as a partner in clinical predictive modeling, many additional requirements need to be fulfilled to make machine learning accessible for clinician scientists. To address this gap, we introduce CliMB, a no-code AI-enabled partner designed to empower clinician scientists to create predictive models using natural language. CliMB guides clinician scientists through the entire medical data science pipeline, thus empowering them to create predictive models from real-world data in just one conversation. CliMB also creates structured reports and interpretable visuals. In evaluations involving clinician scientists and systematic comparisons against a baseline GPT-4, CliMB consistently demonstrated superior performance in key areas such as planning, error prevention, code execution, and model performance. Moreover, in blinded assessments involving 45 clinicians from diverse specialties and career stages, more than 80% preferred CliMB over GPT-4. Overall, by providing a no-code interface with clear guidance and access to SOTA methods in the fields of data-centric AI, AutoML, and interpretable ML, CliMB empowers clinician scientists to build robust predictive models."
                    },
                    {
                        "title": "Time Series Diffusion in the Frequency Domain",
                        "abstract": "Fourier analysis has been an instrumental tool in the development of signal processing. This leads us to wonder whether this framework could similarly benefit generative modelling. In this paper, we explore this question through the scope of time series diffusion models. More specifically, we analyze whether representing time series in the frequency domain is a useful inductive bias for score-based diffusion models. By starting from the canonical SDE formulation of diffusion in the time domain, we show that a dual diffusion process occurs in the frequency domain with an important nuance: Brownian motions are replaced by what we call mirrored Brownian motions, characterized by mirror symmetries among their components. Building on this insight, we show how to adapt the denoising score matching approach to implement diffusion models in the frequency domain. This results in frequency diffusion models, which we compare to canonical time diffusion models. Our empirical evaluation on real-world datasets, covering various domains like healthcare and finance, shows that frequency diffusion models better capture the training distribution than time diffusion models. We explain this observation by showing that time series from these datasets tend to be more localized in the frequency domain than in the time domain, which makes them easier to model in the former case. All our observations point towards impactful synergies between Fourier analysis and diffusion models."
                    },
                    {
                        "title": "Informed Meta-Learning",
                        "abstract": "In noisy and low-data regimes prevalent in real-world applications, a key challenge of machine learning lies in effectively incorporating inductive biases that promote data efficiency and robustness. Meta-learning and informed ML stand out as two approaches for incorporating prior knowledge into ML pipelines. While the former relies on a purely data-driven source of priors, the latter is guided by prior domain knowledge. In this paper, we formalise a hybrid paradigm, informed meta-learning, facilitating the incorporation of priors from unstructured knowledge representations, such as natural language; thus, unlocking complementarity in cross-task knowledge sharing of humans and machines. We establish the foundational components of informed meta-learning and present a concrete instantiation of this framework--the Informed Neural Process. Through a series of experiments, we demonstrate the potential benefits of informed meta-learning in improving data efficiency, robustness to observational noise and task distribution shifts."
                    },
                    {
                        "title": "Risk-Sensitive Diffusion: Learning the Underlying Distribution from Noisy Samples",
                        "abstract": "While achieving remarkable performances, we show that diffusion models are fragile to the presence of noisy samples, limiting their potential in the vast amount of settings where, unlike image synthesis, we are not blessed with clean data. Motivated by our finding that such fragility originates from the distribution gaps between noisy and clean samples along the diffusion process, we introduce risk-sensitive SDE , a stochastic differential equation that is parameterized by the risk (i.e., data \u201cdirtiness\u201d) to adjust the distributions of noisy samples, reducing misguidance while bene-fiting from their contained information. The optimal expression for risk-sensitive SDE depends on the specific noise distribution, and we derive its parameterizations that minimize the misguidance of noisy samples for both Gaussian and general non-Gaussian perturbations. We conduct extensive experiments on both synthetic and real-world datasets (e.g., medical time series), showing that our model effectively recovers the clean data distribution from noisy samples, significantly outper-forming conditional generation baselines."
                    },
                    {
                        "title": "Unveiling the Power of Sparse Neural Networks for Feature Selection",
                        "abstract": "Sparse Neural Networks (SNNs) have emerged as powerful tools for efficient feature selection. Leveraging the dynamic sparse training (DST) algorithms within SNNs has demonstrated promising feature selection capabilities while drastically reducing computational overheads. Despite these advancements, several critical aspects remain insufficiently explored for feature selection. Questions persist regarding the choice of the DST algorithm for network training, the choice of metric for ranking features/neurons, and the comparative performance of these methods across diverse datasets when compared to dense networks. This paper addresses these gaps by presenting a comprehensive systematic analysis of feature selection with sparse neural networks. Moreover, we introduce a novel metric considering sparse neural network characteristics, which is designed to quantify feature importance within the context of SNNs. Our findings show that feature selection with SNNs trained with DST algorithms can achieve, on average, more than $50\\%$ memory and $55\\%$ FLOPs reduction compared to the dense networks, while outperforming them in terms of the quality of the selected features. Our code and the supplementary material are available on GitHub (\\url{https://github.com/zahraatashgahi/Neuron-Attribution})."
                    },
                    {
                        "title": "Discovering Preference Optimization Algorithms with and for Large Language Models",
                        "abstract": "Offline preference optimization is a key method for enhancing and controlling the quality of Large Language Model (LLM) outputs. Typically, preference optimization is approached as an offline supervised learning task using manually-crafted convex loss functions. While these methods are based on theoretical insights, they are inherently constrained by human creativity, so the large search space of possible loss functions remains under explored. We address this by performing LLM-driven objective discovery to automatically discover new state-of-the-art preference optimization algorithms without (expert) human intervention. Specifically, we iteratively prompt an LLM to propose and implement new preference optimization loss functions based on previously-evaluated performance metrics. This process leads to the discovery of previously-unknown and performant preference optimization algorithms. The best performing of these we call Discovered Preference Optimization (DiscoPOP), a novel algorithm that adaptively blends logistic and exponential losses. Experiments demonstrate the state-of-the-art performance of DiscoPOP and its successful transfer to held-out tasks."
                    },
                    {
                        "title": "A Study of Posterior Stability for Time-Series Latent Diffusion",
                        "abstract": "Latent diffusion has demonstrated promising results in image generation and permits efficient sampling. However, this framework might suffer from the problem of posterior collapse when applied to time series. In this paper, we first show that posterior collapse will reduce latent diffusion to a variational autoencoder (VAE), making it less expressive. This highlights the importance of addressing this issue. We then introduce a principled method: dependency measure, that quantifies the sensitivity of a recurrent decoder to input variables. Using this tool, we confirm that posterior collapse significantly affects time-series latent diffusion on real datasets, and a phenomenon termed dependency illusion is also discovered in the case of shuffled time series. Finally, building on our theoretical and empirical studies, we introduce a new framework that extends latent diffusion and has a stable posterior. Extensive experiments on multiple real time-series datasets show that our new framework is free from posterior collapse and significantly outperforms previous baselines in time series synthesis."
                    },
                    {
                        "title": "Dense Reward for Free in Reinforcement Learning from Human Feedback",
                        "abstract": "Reinforcement Learning from Human Feedback (RLHF) has been credited as the key advance that has allowed Large Language Models (LLMs) to effectively follow instructions and produce useful assistance. Classically, this involves generating completions from the LLM in response to a query before using a separate reward model to assign a score to the full completion. As an auto-regressive process, the LLM has to take many\"actions\"(selecting individual tokens) and only receives a single, sparse reward at the end of an episode, a setup that is known to be difficult to optimise in traditional reinforcement learning. In this work we leverage the fact that the reward model contains more information than just its scalar output, in particular, it calculates an attention map over tokens as part of the transformer architecture. We use these attention weights to redistribute the reward along the whole completion, effectively densifying the signal and highlighting the most important tokens, all without incurring extra computational cost or requiring any additional modelling. We demonstrate that, theoretically, this approach is equivalent to potential-based reward shaping, ensuring that the optimal policy remains unchanged. Empirically, we show that it stabilises training, accelerates the rate of learning, and, in practical cases, may lead to better local optima."
                    }
                ]
            }
        }
    },
    "2405.03548": {
        "paper_data": {
            "title": "MAmmoTH2: Scaling Instructions from the Web",
            "url": "http://arxiv.org/abs/2405.03548v4",
            "arxiv_id": "2405.03548",
            "authors": [
                "Xiang Yue",
                "Tuney Zheng",
                "Ge Zhang",
                "Wenhu Chen"
            ],
            "abstract": "Instruction tuning improves the reasoning abilities of large language models (LLMs), with data quality and scalability being the crucial factors. Most instruction tuning data come from human crowd-sourcing or GPT-4 distillation. We propose a paradigm to efficiently harvest 10 million naturally existing instruction data from the pre-training web corpus to enhance LLM reasoning. Our approach involves (1) recalling relevant documents, (2) extracting instruction-response pairs, and (3) refining the extracted pairs using open-source LLMs. Fine-tuning base LLMs on this dataset, we build MAmmoTH2 models, which significantly boost performance on reasoning benchmarks. Notably, MAmmoTH2-7B's (Mistral) performance increases from 11% to 36.7% on MATH and from 36% to 68.4% on GSM8K without training on any in-domain data. Further training MAmmoTH2 on public instruction tuning datasets yields MAmmoTH2-Plus, achieving state-of-the-art performance on several reasoning and chatbot benchmarks. Our work demonstrates how to harvest large-scale, high-quality instruction data without costly human annotation or GPT-4 distillation, providing a new paradigm for building better instruction tuning data.",
            "introduction": "   1 Introduction  Reasoning is a fundamental aspect of human cognition and problem-solving\u00a0(Clark et\u00a0al., 2018; Hendrycks et\u00a0al., 2021a; Cobbe et\u00a0al., 2021; Rein et\u00a0al., 2023; Yue et\u00a0al., 2023a). Proficiency in reasoning is essential for advancing scientific knowledge, developing new technologies, and making informed decisions in various contexts. Recently, large language models (LLMs)\u00a0(Brown et\u00a0al., 2020; Ouyang et\u00a0al., 2022; Touvron et\u00a0al., 2023a, b; Achiam et\u00a0al., 2023; Team et\u00a0al., 2023) have shown remarkable progress in various NLP tasks. However, their ability to perform complex reasoning tasks\u00a0(Lin et\u00a0al., 2024) in the domains of mathematics, science, and engineering is still limited.   Recent studies have extensively explored how to enhance base LLMs\u2019 reasoning abilities. The two main approaches are continued training and instruction tuning. Continued training trains LLMs on large-scale filtered documents\u00a0(Lewkowycz et\u00a0al., 2022; Taylor et\u00a0al., 2022; Azerbayev et\u00a0al., 2023; Shao et\u00a0al., 2024; Ying et\u00a0al., 2024). Instruction tuning seeks to employ supervised fine-tuning loss on, usually small-scale, high-quality instruction-response pairs\u00a0(Ouyang et\u00a0al., 2022; Chung et\u00a0al., 2024). While human-annotated instruction datasets\u00a0(Cobbe et\u00a0al., 2021; Hendrycks et\u00a0al., 2021b; Amini et\u00a0al., 2019) are often limited in scale, recent studies\u00a0(Yu et\u00a0al., 2023; Yue et\u00a0al., 2023b; Toshniwal et\u00a0al., 2024; Li et\u00a0al., 2024a; Tang et\u00a0al., 2024) attempt to prompt GPT-4 with seed data to increase the scalability. However, the synthesized instruction data becomes highly biased, not diverse, and prone to a high degree of hallucination.   To address these limitations, we propose to discover naturally existing instruction data from the web (Figure\u00a02). We argue that the pre-training corpus (e.g., Common Crawl) already contains a vast amount of high-quality instruction data for LLM reasoning. For example, the web corpus contains a large amount of educational materials in the form of instruction-following pairs. These documents range across various domains like math, science, engineering, and humanities. Such readily available instruction data is not only diverse but also of high quality. However, such instruction data is highly dispersed across the corpus, which makes it particularly challenging to discover.   In this paper, we aim to mine these instruction-response pairs from the web using a three-step pipeline. (1) Recall step: We create a diverse seed dataset by crawling several quiz websites. We use this seed data to train a fastText model\u00a0(Joulin et\u00a0al., 2016) and employ it to recall documents from Common Crawl\u00a0(Computer, 2023). GPT-4 is used to trim down the recalled documents by their root URL. We obtain 18M documents through this step. (2) Extract step: We utilize open-source LLMs like Mixtral\u00a0(Jiang et\u00a0al., 2024) to extract Q-A pairs from these documents, producing roughly 5M candidate Q-A pairs. (3) Refine step: After extraction, we further employ Mixtral-8\u00d7\\times\u00d77B\u00a0(Jiang et\u00a0al., 2024) and Qwen-72B\u00a0(Bai et\u00a0al., 2023) to refine\u00a0(Zheng et\u00a0al., 2024b) these candidate Q-A pairs. This refinement operation aims to remove unrelated content, fix formality, and add missing explanations to the candidate Q-A pairs. This refinement operation is pivotal to maintaining the quality of the mined Q-A pairs. Eventually, we harvest a total of 10M instruction-response pairs through these steps. Unlike existing instruction-tuning dataset, our dataset WebInstruct is purely mined from the Web without any human crowdsourcing or GPT-4 distillation.   Figure 2: Comparison between our dataset curation method and previous studies.   We validate the effectiveness of WebInstruct by training",
            "references": [
                {
                    "title": "Advancing LLM Reasoning Generalists with Preference Trees",
                    "abstract": "We introduce Eurus, a suite of large language models (LLMs) optimized for reasoning. Finetuned from Mistral-7B and CodeLlama-70B, Eurus models achieve state-of-the-art results among open-source models on a diverse set of benchmarks covering mathematics, code generation, and logical reasoning problems. Notably, Eurus-70B beats GPT-3.5 Turbo in reasoning through a comprehensive benchmarking across 12 tests covering five tasks, and achieves a 33.3% pass@1 accuracy on LeetCode and 32.6% on TheoremQA, two challenging benchmarks, substantially outperforming existing open-source models by margins more than 13.3%. The strong performance of Eurus can be primarily attributed to UltraInteract, our newly-curated large-scale, high-quality alignment dataset specifically designed for complex reasoning tasks. UltraInteract can be used in both supervised fine-tuning and preference learning. For each instruction, it includes a preference tree consisting of (1) reasoning chains with diverse planning strategies in a unified format, (2) multi-turn interaction trajectories with the environment and the critique, and (3) pairwise data to facilitate preference learning. UltraInteract allows us to conduct an in-depth exploration of preference learning for reasoning tasks. Our investigation reveals that some well-established preference learning algorithms may be less suitable for reasoning tasks compared to their effectiveness in general conversations. Inspired by this, we derive a novel reward modeling objective which, together with UltraInteract, leads to a strong reward model."
                },
                {
                    "title": "LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models",
                    "abstract": "Efficient fine-tuning is vital for adapting large language models (LLMs) to downstream tasks. However, it requires non-trivial efforts to implement these methods on different models. We present LlamaFactory, a unified framework that integrates a suite of cutting-edge efficient training methods. It provides a solution for flexibly customizing the fine-tuning of 100+ LLMs without the need for coding through the built-in web UI LlamaBoard. We empirically validate the efficiency and effectiveness of our framework on language modeling and text generation tasks. It has been released at https://github.com/hiyouga/LLaMA-Factory and received over 25,000 stars and 3,000 forks."
                },
                {
                    "title": "Common 7B Language Models Already Possess Strong Math Capabilities",
                    "abstract": "Mathematical capabilities were previously believed to emerge in common language models only at a very large scale or require extensive math-related pre-training. This paper shows that the LLaMA-2 7B model with common pre-training already exhibits strong mathematical abilities, as evidenced by its impressive accuracy of 97.7% and 72.0% on the GSM8K and MATH benchmarks, respectively, when selecting the best response from 256 random generations. The primary issue with the current base model is the difficulty in consistently eliciting its inherent mathematical capabilities. Notably, the accuracy for the first answer drops to 49.5% and 7.9% on the GSM8K and MATH benchmarks, respectively. We find that simply scaling up the SFT data can significantly enhance the reliability of generating correct answers. However, the potential for extensive scaling is constrained by the scarcity of publicly available math questions. To overcome this limitation, we employ synthetic data, which proves to be nearly as effective as real data and shows no clear saturation when scaled up to approximately one million samples. This straightforward approach achieves an accuracy of 82.6% on GSM8K and 40.6% on MATH using LLaMA-2 7B models, surpassing previous models by 14.2% and 20.8%, respectively. We also provide insights into scaling behaviors across different reasoning complexities and error types."
                },
                {
                    "title": "Yi: Open Foundation Models by 01.AI",
                    "abstract": "We introduce the Yi model family, a series of language and multimodal models that demonstrate strong multi-dimensional capabilities. The Yi model family is based on 6B and 34B pretrained language models, then we extend them to chat models, 200K long context models, depth-upscaled models, and vision-language models. Our base models achieve strong performance on a wide range of benchmarks like MMLU, and our finetuned chat models deliver strong human preference rate on major evaluation platforms like AlpacaEval and Chatbot Arena. Building upon our scalable super-computing infrastructure and the classical transformer architecture, we attribute the performance of Yi models primarily to its data quality resulting from our data-engineering efforts. For pretraining, we construct 3.1 trillion tokens of English and Chinese corpora using a cascaded data deduplication and quality filtering pipeline. For finetuning, we polish a small scale (less than 10K) instruction dataset over multiple iterations such that every single instance has been verified directly by our machine learning engineers. For vision-language, we combine the chat language model with a vision transformer encoder and train the model to align visual representations to the semantic space of the language model. We further extend the context length to 200K through lightweight continual pretraining and demonstrate strong needle-in-a-haystack retrieval performance. We show that extending the depth of the pretrained checkpoint through continual pretraining further improves performance. We believe that given our current results, continuing to scale up model parameters using thoroughly optimized data will lead to even stronger frontier models."
                },
                {
                    "title": "MathScale: Scaling Instruction Tuning for Mathematical Reasoning",
                    "abstract": "Large language models (LLMs) have demonstrated remarkable capabilities in problem-solving. However, their proficiency in solving mathematical problems remains inadequate. We propose MathScale, a simple and scalable method to create high-quality mathematical reasoning data using frontier LLMs (e.g., {\\tt GPT-3.5}). Inspired by the cognitive mechanism in human mathematical learning, it first extracts topics and knowledge points from seed math questions and then build a concept graph, which is subsequently used to generate new math questions. MathScale exhibits effective scalability along the size axis of the math dataset that we generate. As a result, we create a mathematical reasoning dataset (MathScaleQA) containing two million math question-answer pairs. To evaluate mathematical reasoning abilities of LLMs comprehensively, we construct {\\sc MwpBench}, a benchmark of Math Word Problems, which is a collection of ten datasets (including GSM8K and MATH) covering K-12, college, and competition level math problems. We apply MathScaleQA to fine-tune open-source LLMs (e.g., LLaMA-2 and Mistral), resulting in significantly improved capabilities in mathematical reasoning. Evaluated on {\\sc MwpBench}, MathScale-7B achieves state-of-the-art performance across all datasets, surpassing its best peers of equivalent size by 42.9\\% in micro average accuracy and 43.7\\% in macro average accuracy, respectively."
                },
                {
                    "title": "OpenCodeInterpreter: Integrating Code Generation with Execution and Refinement",
                    "abstract": "The introduction of large language models has significantly advanced code generation. However, open-source models often lack the execution capabilities and iterative refinement of advanced systems like the GPT-4 Code Interpreter. To address this, we introduce OpenCodeInterpreter, a family of open-source code systems designed for generating, executing, and iteratively refining code. Supported by Code-Feedback, a dataset featuring 68K multi-turn interactions, OpenCodeInterpreter integrates execution and human feedback for dynamic code refinement. Our comprehensive evaluation of OpenCodeInterpreter across key benchmarks such as HumanEval, MBPP, and their enhanced versions from EvalPlus reveals its exceptional performance. Notably, OpenCodeInterpreter-33B achieves an accuracy of 83.2 (76.4) on the average (and plus versions) of HumanEval and MBPP, closely rivaling GPT-4's 84.2 (76.2) and further elevates to 91.6 (84.6) with synthesized human feedback from GPT-4. OpenCodeInterpreter brings the gap between open-source code generation models and proprietary systems like GPT-4 Code Interpreter."
                },
                {
                    "title": "CriticBench: Benchmarking LLMs for Critique-Correct Reasoning",
                    "abstract": "The ability of Large Language Models (LLMs) to critique and refine their reasoning is crucial for their application in evaluation, feedback provision, and self-improvement. This paper introduces CriticBench, a comprehensive benchmark designed to assess LLMs' abilities to critique and rectify their reasoning across a variety of tasks. CriticBench encompasses five reasoning domains: mathematical, commonsense, symbolic, coding, and algorithmic. It compiles 15 datasets and incorporates responses from three LLM families. Utilizing CriticBench, we evaluate and dissect the performance of 17 LLMs in generation, critique, and correction reasoning, i.e., GQC reasoning. Our findings reveal: (1) a linear relationship in GQC capabilities, with critique-focused training markedly enhancing performance; (2) a task-dependent variation in correction effectiveness, with logic-oriented tasks being more amenable to correction; (3) GQC knowledge inconsistencies that decrease as model size increases; and (4) an intriguing inter-model critiquing dynamic, where stronger models are better at critiquing weaker ones, while weaker models can surprisingly surpass stronger ones in their self-critique. We hope these insights into the nuanced critique-correct reasoning of LLMs will foster further research in LLM critique and self-improvement."
                },
                {
                    "title": "Orca-Math: Unlocking the potential of SLMs in Grade School Math",
                    "abstract": "Mathematical word problem-solving has long been recognized as a complex task for small language models (SLMs). A recent study hypothesized that the smallest model size, needed to achieve over 80% accuracy on the GSM8K benchmark, is 34 billion parameters. To reach this level of performance with smaller models, researcher often train SLMs to generate Python code or use tools to help avoid calculation errors. Additionally, they employ ensembling, where outputs of up to 100 model runs are combined to arrive at a more accurate result. Result selection is done using consensus, majority vote or a separate a verifier model used in conjunction with the SLM. Ensembling provides a substantial boost in accuracy but at a significant cost increase with multiple calls to the model (e.g., Phi-GSM uses top-48 to boost the performance from 68.2 to 81.5). In this work, we present Orca-Math, a 7-billion-parameter SLM based on the Mistral-7B, which achieves 86.81% on GSM8k without the need for multiple model calls or the use of verifiers, code execution or any other external tools. Our approach has the following key elements: (1) A high quality synthetic dataset of 200K math problems created using a multi-agent setup where agents collaborate to create the data, (2) An iterative learning techniques that enables the SLM to practice solving problems, receive feedback on its solutions and learn from preference pairs incorporating the SLM solutions and the feedback. When trained with Supervised Fine-Tuning alone, Orca-Math achieves 81.50% on GSM8k pass@1 metric. With iterative preference learning, Orca-Math achieves 86.81% pass@1. Orca-Math surpasses the performance of significantly larger models such as LLAMA-2-70B, WizardMath-70B, Gemini-Pro, ChatGPT-3.5. It also significantly outperforms other smaller models while using much smaller data (hundreds of thousands vs. millions of problems)."
                },
                {
                    "title": "OpenMathInstruct-1: A 1.8 Million Math Instruction Tuning Dataset",
                    "abstract": "Recent work has shown the immense potential of synthetically generated datasets for training large language models (LLMs), especially for acquiring targeted skills. Current large-scale math instruction tuning datasets such as MetaMathQA (Yu et al., 2024) and MAmmoTH (Yue et al., 2024) are constructed using outputs from closed-source LLMs with commercially restrictive licenses. A key reason limiting the use of open-source LLMs in these data generation pipelines has been the wide gap between the mathematical skills of the best closed-source LLMs, such as GPT-4, and the best open-source LLMs. Building on the recent progress in open-source LLMs, our proposed prompting novelty, and some brute-force scaling, we construct OpenMathInstruct-1, a math instruction tuning dataset with 1.8M problem-solution pairs. The dataset is constructed by synthesizing code-interpreter solutions for GSM8K and MATH, two popular math reasoning benchmarks, using the recently released and permissively licensed Mixtral model. Our best model, OpenMath-CodeLlama-70B, trained on a subset of OpenMathInstruct-1, achieves a score of 84.6% on GSM8K and 50.7% on MATH, which is competitive with the best gpt-distilled models. We release our code, models, and the OpenMathInstruct-1 dataset under a commercially permissive license."
                },
                {
                    "title": "Autonomous Data Selection with Language Models for Mathematical Texts",
                    "abstract": "To improve language models' proficiency in mathematical reasoning via continual pretraining, we introduce a novel strategy that leverages base language models for autonomous data selection. Departing from conventional supervised fine-tuning or trained classifiers with human-annotated data, our approach Autonomous Data Selection (AutoDS) utilizes meta-prompted language models as zero-shot verifiers to evaluate and select high-quality mathematical content autonomously. To demonstrate the efficacy of our method, we continuously pretrained a 7B-parameter language model on our curated dataset, achieving substantial improvements in downstream performance on the MATH, GSM8K, and BIG-Bench Hard (BBH) tasks with a token amount reduced by orders of magnitude compared to previous continual pretraining works. Our method showcases a 2 times increase in pretraining token efficiency compared to state-of-the-art baselines, underscoring the potential of our approach in enhancing models' mathematical reasoning capabilities. The AutoMathText dataset is available at https://huggingface.co/datasets/math-ai/AutoMathText. The code is available at https://github.com/yifanzhang-pro/AutoMathText."
                },
                {
                    "title": "InternLM-Math: Open Math Large Language Models Toward Verifiable Reasoning",
                    "abstract": "The math abilities of large language models can represent their abstract reasoning ability. In this paper, we introduce and open-source our math reasoning LLMs InternLM-Math which is continue pre-trained from InternLM2. We unify chain-of-thought reasoning, reward modeling, formal reasoning, data augmentation, and code interpreter in a unified seq2seq format and supervise our model to be a versatile math reasoner, verifier, prover, and augmenter. These abilities can be used to develop the next math LLMs or self-iteration. InternLM-Math obtains open-sourced state-of-the-art performance under the setting of in-context learning, supervised fine-tuning, and code-assisted reasoning in various informal and formal benchmarks including GSM8K, MATH, Hungary math exam, MathBench-ZH, and MiniF2F. Our pre-trained model achieves 30.3 on the MiniF2F test set without fine-tuning. We further explore how to use LEAN to solve math problems and study its performance under the setting of multi-task learning which shows the possibility of using LEAN as a unified platform for solving and proving in math. Our models, codes, and data are released at \\url{https://github.com/InternLM/InternLM-Math}."
                },
                {
                    "title": "DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models",
                    "abstract": "Mathematical reasoning poses a significant challenge for language models due to its complex and structured nature. In this paper, we introduce DeepSeekMath 7B, which continues pre-training DeepSeek-Coder-Base-v1.5 7B with 120B math-related tokens sourced from Common Crawl, together with natural language and code data. DeepSeekMath 7B has achieved an impressive score of 51.7% on the competition-level MATH benchmark without relying on external toolkits and voting techniques, approaching the performance level of Gemini-Ultra and GPT-4. Self-consistency over 64 samples from DeepSeekMath 7B achieves 60.9% on MATH. The mathematical reasoning capability of DeepSeekMath is attributed to two key factors: First, we harness the significant potential of publicly available web data through a meticulously engineered data selection pipeline. Second, we introduce Group Relative Policy Optimization (GRPO), a variant of Proximal Policy Optimization (PPO), that enhances mathematical reasoning abilities while concurrently optimizing the memory usage of PPO."
                },
                {
                    "title": "Kun: Answer Polishment for Chinese Self-Alignment with Instruction Back-Translation",
                    "abstract": "In this paper, we introduce Kun, a novel approach for creating high-quality instruction-tuning datasets for large language models (LLMs) without relying on manual annotations. Adapting a self-training algorithm based on instruction back-translation and answer polishment, Kun leverages unlabelled data from diverse sources such as Wudao, Wanjuan, and SkyPile to generate a substantial dataset of over a million Chinese instructional data points. This approach significantly deviates from traditional methods by using a self-curation process to refine and select the most effective instruction-output pairs. Our experiments with the 6B-parameter Yi model across various benchmarks demonstrate Kun's robustness and scalability. Our method's core contributions lie in its algorithmic advancement, which enhances data retention and clarity, and its innovative data generation approach that substantially reduces the reliance on costly and time-consuming manual annotations. This methodology presents a scalable and efficient solution for improving the instruction-following capabilities of LLMs, with significant implications for their application across diverse fields. The code and dataset can be found at https://github.com/Zheng0428/COIG-Kun"
                },
                {
                    "title": "Mixtral of Experts",
                    "abstract": "We introduce Mixtral 8x7B, a Sparse Mixture of Experts (SMoE) language model. Mixtral has the same architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks (i.e. experts). For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, each token has access to 47B parameters, but only uses 13B active parameters during inference. Mixtral was trained with a context size of 32k tokens and it outperforms or matches Llama 2 70B and GPT-3.5 across all evaluated benchmarks. In particular, Mixtral vastly outperforms Llama 2 70B on mathematics, code generation, and multilingual benchmarks. We also provide a model fine-tuned to follow instructions, Mixtral 8x7B - Instruct, that surpasses GPT-3.5 Turbo, Claude-2.1, Gemini Pro, and Llama 2 70B - chat model on human benchmarks. Both the base and instruct models are released under the Apache 2.0 license."
                },
                {
                    "title": "Generative AI for Math: Part I - MathPile: A Billion-Token-Scale Pretraining Corpus for Math",
                    "abstract": "High-quality, large-scale corpora are the cornerstone of building foundation models. In this work, we introduce \\textsc{MathPile}, a diverse and high-quality math-centric corpus comprising about 9.5 billion tokens. Throughout its creation, we adhered to the principle of ``\\emph{less is more}'', firmly believing in the supremacy of data quality over quantity, even in the pre-training phase. Our meticulous data collection and processing efforts included a complex suite of preprocessing, prefiltering, language identification, cleaning, filtering, and deduplication, ensuring the high quality of our corpus. Furthermore, we performed data contamination detection on downstream benchmark test sets to eliminate duplicates. We hope our \\textsc{MathPile} can help to enhance the mathematical reasoning abilities of language models. We plan to open-source different versions of \\mathpile with the scripts used for processing, to facilitate future developments in this field."
                },
                {
                    "title": "MMMU: A Massive Multi-Discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI",
                    "abstract": "We introduce MMMU: a new benchmark designed to evaluate multimodal models on massive multi-discipline tasks demanding college-level subject knowledge and deliberate reasoning. MMMU includes 11.5K meticulously collected multimodal questions from college exams, quizzes, and text-books, covering six core disciplines: Art & Design, Busi-ness, Science, Health & Medicine, Humanities & Social Science, and Tech & Engineering. These questions span 30 subjects and 183 subfields, comprising 30 highly het-erogeneous image types, such as charts, diagrams, maps, tables, music sheets, and chemical structures. Unlike existing benchmarks, MMMU focuses on advanced perception and reasoning with domain-specific knowledge, challenging models to perform tasks akin to those faced by experts. The evaluation of 28 open-source LMMs as well as the propri-etary GPT-4V(ision) and Gemini highlights the substantial challenges posed by MMMU. Even the advanced GPT-4V and Gemini Ultra only achieve accuracies of 56% and 59% respectively, indicating significant room for improvement. We believe MMMU will stimulate the community to build next-generation multimodal foundation models towards expert artificial general intelligence."
                },
                {
                    "title": "GPQA: A Graduate-Level Google-Proof Q&A Benchmark",
                    "abstract": "We present GPQA, a challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. We ensure that the questions are high-quality and extremely difficult: experts who have or are pursuing PhDs in the corresponding domains reach 65% accuracy (74% when discounting clear mistakes the experts identified in retrospect), while highly skilled non-expert validators only reach 34% accuracy, despite spending on average over 30 minutes with unrestricted access to the web (i.e., the questions are\"Google-proof\"). The questions are also difficult for state-of-the-art AI systems, with our strongest GPT-4 based baseline achieving 39% accuracy. If we are to use future AI systems to help us answer very hard questions, for example, when developing new scientific knowledge, we need to develop scalable oversight methods that enable humans to supervise their outputs, which may be difficult even if the supervisors are themselves skilled and knowledgeable. The difficulty of GPQA both for skilled non-experts and frontier AI systems should enable realistic scalable oversight experiments, which we hope can help devise ways for human experts to reliably get truthful information from AI systems that surpass human capabilities."
                },
                {
                    "title": "Camels in a Changing Climate: Enhancing LM Adaptation with Tulu 2",
                    "abstract": "Since the release of T\\\"ULU [Wang et al., 2023b], open resources for instruction tuning have developed quickly, from better base models to new finetuning techniques. We test and incorporate a number of these advances into T\\\"ULU, resulting in T\\\"ULU 2, a suite of improved T\\\"ULU models for advancing the understanding and best practices of adapting pretrained language models to downstream tasks and user preferences. Concretely, we release: (1) T\\\"ULU-V2-mix, an improved collection of high-quality instruction datasets; (2) T\\\"ULU 2, LLAMA-2 models finetuned on the V2 mixture; (3) T\\\"ULU 2+DPO, T\\\"ULU 2 models trained with direct preference optimization (DPO), including the largest DPO-trained model to date (T\\\"ULU 2+DPO 70B); (4) CODE T\\\"ULU 2, CODE LLAMA models finetuned on our V2 mix that outperform CODE LLAMA and its instruction-tuned variant, CODE LLAMA-Instruct. Our evaluation from multiple perspectives shows that the T\\\"ULU 2 suite achieves state-of-the-art performance among open models and matches or exceeds the performance of GPT-3.5-turbo-0301 on several benchmarks. We release all the checkpoints, data, training and evaluation code to facilitate future open efforts on adapting large language models."
                },
                {
                    "title": "Llemma: An Open Language Model For Mathematics",
                    "abstract": "We present Llemma, a large language model for mathematics. We continue pretraining Code Llama on the Proof-Pile-2, a mixture of scientific papers, web data containing mathematics, and mathematical code, yielding Llemma. On the MATH benchmark Llemma outperforms all known open base models, as well as the unreleased Minerva model suite on an equi-parameter basis. Moreover, Llemma is capable of tool use and formal theorem proving without any further finetuning. We openly release all artifacts, including 7 billion and 34 billion parameter models, the Proof-Pile-2, and code to replicate our experiments."
                },
                {
                    "title": "OpenWebMath: An Open Dataset of High-Quality Mathematical Web Text",
                    "abstract": "There is growing evidence that pretraining on high quality, carefully thought-out tokens such as code or mathematics plays an important role in improving the reasoning abilities of large language models. For example, Minerva, a PaLM model finetuned on billions of tokens of mathematical documents from arXiv and the web, reported dramatically improved performance on problems that require quantitative reasoning. However, because all known open source web datasets employ preprocessing that does not faithfully preserve mathematical notation, the benefits of large scale training on quantitive web documents are unavailable to the research community. We introduce OpenWebMath, an open dataset inspired by these works containing 14.7B tokens of mathematical webpages from Common Crawl. We describe in detail our method for extracting text and LaTeX content and removing boilerplate from HTML documents, as well as our methods for quality filtering and deduplication. Additionally, we run small-scale experiments by training 1.4B parameter language models on OpenWebMath, showing that models trained on 14.7B tokens of our dataset surpass the performance of models trained on over 20x the amount of general language data. We hope that our dataset, openly released on the Hugging Face Hub, will help spur advances in the reasoning abilities of large language models."
                },
                {
                    "title": "Mistral 7B",
                    "abstract": "We introduce Mistral 7B v0.1, a 7-billion-parameter language model engineered for superior performance and efficiency. Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation. Our model leverages grouped-query attention (GQA) for faster inference, coupled with sliding window attention (SWA) to effectively handle sequences of arbitrary length with a reduced inference cost. We also provide a model fine-tuned to follow instructions, Mistral 7B -- Instruct, that surpasses the Llama 2 13B -- Chat model both on human and automated benchmarks. Our models are released under the Apache 2.0 license."
                },
                {
                    "title": "Qwen Technical Report",
                    "abstract": "Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural language processing tasks that were previously thought to be exclusive to humans. In this work, we introduce Qwen, the first installment of our large language model series. Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts. It includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques. The base language models consistently demonstrate superior performance across a multitude of downstream tasks, and the chat models, particularly those trained using Reinforcement Learning from Human Feedback (RLHF), are highly competitive. The chat models possess advanced tool-use and planning capabilities for creating agent applications, showcasing impressive performance even when compared to bigger models on complex tasks like utilizing a code interpreter. Furthermore, we have developed coding-specialized models, Code-Qwen and Code-Qwen-Chat, as well as mathematics-focused models, Math-Qwen-Chat, which are built upon base language models. These models demonstrate significantly improved performance in comparison with open-source models, and slightly fall behind the proprietary models."
                },
                {
                    "title": "MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Models",
                    "abstract": "Large language models (LLMs) have pushed the limits of natural language understanding and exhibited excellent problem-solving ability. Despite the great success, most existing open-source LLMs (e.g., LLaMA-2) are still far away from satisfactory for solving mathematical problem due to the complex reasoning procedures. To bridge this gap, we propose MetaMath, a fine-tuned language model that specializes in mathematical reasoning. Specifically, we start by bootstrapping mathematical questions by rewriting the question from multiple perspectives without extra knowledge, which results in a new dataset called MetaMathQA. Then we fine-tune the LLaMA-2 models on MetaMathQA. Experimental results on two popular benchmarks (i.e., GSM8K and MATH) for mathematical reasoning demonstrate that MetaMath outperforms a suite of open-source LLMs by a significant margin. Our MetaMath-7B model achieves 66.4% on GSM8K and 19.4% on MATH, exceeding the state-of-the-art models of the same size by 11.5% and 8.7%. Particularly, MetaMath-70B achieves an accuracy of 82.3% on GSM8K, slightly better than GPT-3.5-Turbo. We release all the MetaMathQA dataset, the MetaMath models with different model sizes and the training code for public use."
                },
                {
                    "title": "MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuning",
                    "abstract": "We introduce MAmmoTH, a series of open-source large language models (LLMs) specifically tailored for general math problem-solving. The MAmmoTH models are trained on MathInstruct, our meticulously curated instruction tuning dataset. MathInstruct is compiled from 13 math datasets with intermediate rationales, six of which have rationales newly curated by us. It presents a unique hybrid of chain-of-thought (CoT) and program-of-thought (PoT) rationales, and also ensures extensive coverage of diverse fields in math. The hybrid of CoT and PoT not only unleashes the potential of tool use but also allows different thought processes for different math problems. As a result, the MAmmoTH series substantially outperform existing open-source models on nine mathematical reasoning datasets across all scales with an average accuracy gain between 16% and 32%. Remarkably, our MAmmoTH-7B model reaches 33% on MATH (a competition-level dataset), which exceeds the best open-source 7B model (WizardMath) by 23%, and the MAmmoTH-34B model achieves 44% accuracy on MATH, even surpassing GPT-4's CoT result. Our work underscores the importance of diverse problem coverage and the use of hybrid rationales in developing superior math generalist models."
                },
                {
                    "title": "SciEval: A Multi-Level Large Language Model Evaluation Benchmark for Scientific Research",
                    "abstract": "Recently, there has been growing interest in using Large Language Models (LLMs) for scientific research. Numerous benchmarks have been proposed to evaluate the ability of LLMs for scientific research. However, current benchmarks are mostly based on pre-collected objective questions. This design suffers from data leakage problem and lacks the evaluation of subjective Q/A ability. In this paper, we propose SciEval, a comprehensive and multi-disciplinary evaluation benchmark to address these issues. Based on Bloom's taxonomy, SciEval covers four dimensions to systematically evaluate scientific research ability. In particular, we design a \"dynamic\" subset based on scientific principles to prevent evaluation from potential data leakage. Both objective and subjective questions are included in SciEval. These characteristics make SciEval a more effective benchmark for scientific research ability evaluation of LLMs. Comprehensive experiments on most advanced LLMs show that, although GPT-4 achieves SOTA performance compared to other LLMs, there is still substantial room for improvement, especially for dynamic questions. The codes and data are publicly available on https://github.com/OpenDFM/SciEval."
                },
                {
                    "title": "WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct",
                    "abstract": "Large language models (LLMs), such as GPT-4, have shown remarkable performance in natural language processing (NLP) tasks, including challenging mathematical reasoning. However, most existing open-source models are only pre-trained on large-scale internet data and without math-related optimization. In this paper, we present WizardMath, which enhances the mathematical reasoning abilities of Llama-2, by applying our proposed Reinforcement Learning from Evol-Instruct Feedback (RLEIF) method to the domain of math. Through extensive experiments on two mathematical reasoning benchmarks, namely GSM8k and MATH, we reveal the extraordinary capabilities of our model. WizardMath surpasses all other open-source LLMs by a substantial margin. Furthermore, our model even outperforms ChatGPT-3.5, Claude Instant-1, PaLM-2 and Minerva on GSM8k, simultaneously surpasses Text-davinci-002, PaLM-1 and GPT-3 on MATH. More details and model weights are public at https://github.com/nlpxucan/WizardLM and https://huggingface.co/WizardLM."
                },
                {
                    "title": "SciBench: Evaluating College-Level Scientific Problem-Solving Abilities of Large Language Models",
                    "abstract": "Most of the existing Large Language Model (LLM) benchmarks on scientific problem reasoning focus on problems grounded in high-school subjects and are confined to elementary algebraic operations. To systematically examine the reasoning capabilities required for solving complex scientific problems, we introduce an expansive benchmark suite SciBench for LLMs. SciBench contains a carefully curated dataset featuring a range of collegiate-level scientific problems from mathematics, chemistry, and physics domains. Based on the dataset, we conduct an in-depth benchmarking study of representative open-source and proprietary LLMs with various prompting strategies. The results reveal that the current LLMs fall short of delivering satisfactory performance, with the best overall score of merely 43.22%. Furthermore, through a detailed user study, we categorize the errors made by LLMs into ten problem-solving abilities. Our analysis indicates that no single prompting strategy significantly outperforms the others and some strategies that demonstrate improvements in certain problem-solving skills could result in declines in other skills. We envision that SciBench will catalyze further developments in the reasoning abilities of LLMs, thereby ultimately contributing to scientific research and discovery."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "TheoremQA: A Theorem-driven Question Answering dataset",
                    "abstract": "The recent LLMs like GPT-4 and PaLM-2 have made tremendous progress in solving fundamental math problems like GSM8K by achieving over 90% accuracy. However, their capabilities to solve more challenging math problems which require domain-specific knowledge (i.e. theorem) have yet to be investigated. In this paper, we introduce TheoremQA, the first theorem-driven question-answering dataset designed to evaluate AI models' capabilities to apply theorems to solve challenging science problems. TheoremQA is curated by domain experts containing 800 high-quality questions covering 350 theorems (e.g. Taylor's theorem, Lagrange's theorem, Huffman coding, Quantum Theorem, Elasticity Theorem, etc) from Math, Physics, EE&CS, and Finance. We evaluate a wide spectrum of 16 large language and code models with different prompting strategies like Chain-of-Thoughts and Program-of-Thoughts. We found that GPT-4's capabilities to solve these problems are unparalleled, achieving an accuracy of 51% with Program-of-Thoughts Prompting. All the existing open-sourced models are below 15%, barely surpassing the random-guess baseline. Given the diversity and broad coverage of TheoremQA, we believe it can be used as a better benchmark to evaluate LLMs' capabilities to solve challenging science problems. The data and code are released in https://github.com/wenhuchen/TheoremQA."
                },
                {
                    "title": "Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generation",
                    "abstract": "Program synthesis has been long studied with recent approaches focused on directly using the power of Large Language Models (LLMs) to generate code. Programming benchmarks, with curated synthesis problems and test-cases, are used to measure the performance of various LLMs on code synthesis. However, these test-cases can be limited in both quantity and quality for fully assessing the functional correctness of the generated code. Such limitation in the existing benchmarks begs the following question: In the era of LLMs, is the code generated really correct? To answer this, we propose EvalPlus -- a code synthesis evaluation framework to rigorously benchmark the functional correctness of LLM-synthesized code. EvalPlus augments a given evaluation dataset with large amounts of test-cases newly produced by an automatic test input generator, powered by both LLM- and mutation-based strategies. While EvalPlus is general, we extend the test-cases of the popular HumanEval benchmark by 80x to build HumanEval+. Our extensive evaluation across 26 popular LLMs (e.g., GPT-4 and ChatGPT) demonstrates that HumanEval+ is able to catch significant amounts of previously undetected wrong code synthesized by LLMs, reducing the pass@k by up-to 19.3-28.9%. We also surprisingly found that test insufficiency can lead to mis-ranking. For example, both WizardCoder-CodeLlama and Phind-CodeLlama now outperform ChatGPT on HumanEval+, while none of them could on HumanEval. Our work not only indicates that prior popular code synthesis evaluation results do not accurately reflect the true performance of LLMs for code synthesis, but also opens up a new direction to improve such programming benchmarks through automated testing. We have open-sourced our tools, enhanced datasets as well as all LLM-generated code at https://github.com/evalplus/evalplus to facilitate and accelerate future LLM-for-code research."
                },
                {
                    "title": "WizardLM: Empowering Large Language Models to Follow Complex Instructions",
                    "abstract": "Training large language models (LLMs) with open-domain instruction following data brings colossal success. However, manually creating such instruction data is very time-consuming and labor-intensive. Moreover, humans may struggle to produce high-complexity instructions. In this paper, we show an avenue for creating large amounts of instruction data with varying levels of complexity using LLM instead of humans. Starting with an initial set of instructions, we use our proposed Evol-Instruct to rewrite them step by step into more complex instructions. Then, we mix all generated instruction data to fine-tune LLaMA. We call the resulting model WizardLM. Human evaluations on a complexity-balanced test bed and Vicuna's testset show that instructions from Evol-Instruct are superior to human-created ones. By analyzing the human evaluation results of the high complexity part, we demonstrate that outputs from our WizardLM are preferred to outputs from OpenAI ChatGPT. In GPT-4 automatic evaluation, WizardLM achieves more than 90\\% capacity of ChatGPT on 17 out of 29 skills. Even though WizardLM still lags behind ChatGPT in some aspects, our findings suggest that fine-tuning with AI-evolved instructions is a promising direction for enhancing LLMs. Our code and data are public at https://github.com/nlpxucan/WizardLM"
                },
                {
                    "title": "Instruction Tuning with GPT-4",
                    "abstract": "Prior work has shown that finetuning large language models (LLMs) using machine-generated instruction-following data enables such models to achieve remarkable zero-shot capabilities on new tasks, and no human-written instructions are needed. In this paper, we present the first attempt to use GPT-4 to generate instruction-following data for LLM finetuning. Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by previous state-of-the-art models. We also collect feedback and comparison data from GPT-4 to enable a comprehensive evaluation and reward model training. We make our data generated using GPT-4 as well as our codebase publicly available."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Self-Instruct: Aligning Language Models with Self-Generated Instructions",
                    "abstract": "Large \u201cinstruction-tuned\u201d language models (i.e., finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is often limited in quantity, diversity, and creativity, therefore hindering the generality of the tuned model. We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off their own generations. Our pipeline generates instructions, input, and output samples from a language model, then filters invalid or similar ones before using them to finetune the original model. Applying our method to the vanilla GPT3, we demonstrate a 33% absolute improvement over the original model on Super-NaturalInstructions, on par with the performance of InstructGPT-001, which was trained with private user data and human annotations. For further evaluation, we curate a set of expert-written instructions for novel tasks, and show through human evaluation that tuning GPT3 with Self-Instruct outperforms using existing public instruction datasets by a large margin, leaving only a 5% absolute gap behind InstructGPT-001. Self-Instruct provides an almost annotation-free method for aligning pre-trained language models with instructions, and we release our large synthetic dataset to facilitate future studies on instruction tuning."
                },
                {
                    "title": "Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks",
                    "abstract": "Recently, there has been significant progress in teaching language models to perform step-by-step reasoning to solve complex numerical reasoning tasks. Chain-of-thoughts prompting (CoT) is by far the state-of-art method for these tasks. CoT uses language models to perform both reasoning and computation in the multi-step `thought' process. To disentangle computation from reasoning, we propose `Program of Thoughts' (PoT), which uses language models (mainly Codex) to express the reasoning process as a program. The computation is relegated to an external computer, which executes the generated programs to derive the answer. We evaluate PoT on five math word problem datasets (GSM, AQuA, SVAMP, TabMWP, MultiArith) and three financial-QA datasets (FinQA, ConvFinQA, TATQA) for both few-shot and zero-shot setups. Under both few-shot and zero-shot settings, PoT can show an average performance gain over CoT by around 12\\% across all the evaluated datasets. By combining PoT with self-consistency decoding, we can achieve SoTA performance on all math problem datasets and near-SoTA performance on financial datasets. All of our data and code are released in Github https://github.com/wenhuchen/Program-of-Thoughts"
                },
                {
                    "title": "PAL: Program-aided Language Models",
                    "abstract": "Large language models (LLMs) have recently demonstrated an impressive ability to perform arithmetic and symbolic reasoning tasks, when provided with a few examples at test time (\"few-shot prompting\"). Much of this success can be attributed to prompting methods such as\"chain-of-thought'', which employ LLMs for both understanding the problem description by decomposing it into steps, as well as solving each step of the problem. While LLMs seem to be adept at this sort of step-by-step decomposition, LLMs often make logical and arithmetic mistakes in the solution part, even when the problem is decomposed correctly. In this paper, we present Program-Aided Language models (PAL): a novel approach that uses the LLM to read natural language problems and generate programs as the intermediate reasoning steps, but offloads the solution step to a runtime such as a Python interpreter. With PAL, decomposing the natural language problem into runnable steps remains the only learning task for the LLM, while solving is delegated to the interpreter. We demonstrate this synergy between a neural LLM and a symbolic interpreter across 13 mathematical, symbolic, and algorithmic reasoning tasks from BIG-Bench Hard and other benchmarks. In all these natural language reasoning tasks, generating code using an LLM and reasoning using a Python interpreter leads to more accurate results than much larger models. For example, PAL using Codex achieves state-of-the-art few-shot accuracy on the GSM8K benchmark of math word problems, surpassing PaLM-540B which uses chain-of-thought by absolute 15% top-1. Our code and data are publicly available at http://reasonwithpal.com/ ."
                },
                {
                    "title": "Galactica: A Large Language Model for Science",
                    "abstract": "Information overload is a major obstacle to scientific progress. The explosive growth in scientific literature and data has made it ever harder to discover useful insights in a large mass of information. Today scientific knowledge is accessed through search engines, but they are unable to organize scientific knowledge alone. In this paper we introduce Galactica: a large language model that can store, combine and reason about scientific knowledge. We train on a large scientific corpus of papers, reference material, knowledge bases and many other sources. We outperform existing models on a range of scientific tasks. On technical knowledge probes such as LaTeX equations, Galactica outperforms the latest GPT-3 by 68.2% versus 49.0%. Galactica also performs well on reasoning, outperforming Chinchilla on mathematical MMLU by 41.3% to 35.7%, and PaLM 540B on MATH with a score of 20.4% versus 8.8%. It also sets a new state-of-the-art on downstream tasks such as PubMedQA and MedMCQA dev of 77.6% and 52.9%. And despite not being trained on a general corpus, Galactica outperforms BLOOM and OPT-175B on BIG-bench. We believe these results demonstrate the potential for language models as a new interface for science. We open source the model for the benefit of the scientific community."
                },
                {
                    "title": "Scaling Instruction-Finetuned Language Models",
                    "abstract": "Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models."
                },
                {
                    "title": "Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them",
                    "abstract": "BIG-Bench (Srivastava et al., 2022) is a diverse evaluation suite that focuses on tasks believed to be beyond the capabilities of current language models. Language models have already made good progress on this benchmark, with the best model in the BIG-Bench paper outperforming average reported human-rater results on 65% of the BIG-Bench tasks via few-shot prompting. But on what tasks do language models fall short of average human-rater performance, and are those tasks actually unsolvable by current language models? In this work, we focus on a suite of 23 challenging BIG-Bench tasks which we call BIG-Bench Hard (BBH). These are the task for which prior language model evaluations did not outperform the average human-rater. We find that applying chain-of-thought (CoT) prompting to BBH tasks enables PaLM to surpass the average human-rater performance on 10 of the 23 tasks, and Codex (code-davinci-002) to surpass the average human-rater performance on 17 of the 23 tasks. Since many tasks in BBH require multi-step reasoning, few-shot prompting without CoT, as done in the BIG-Bench evaluations (Srivastava et al., 2022), substantially underestimates the best performance and capabilities of language models, which is better captured via CoT prompting. As further analysis, we explore the interaction between CoT and model scale on BBH, finding that CoT enables emergent task performance on several BBH tasks with otherwise flat scaling curves."
                },
                {
                    "title": "Solving Quantitative Reasoning Problems with Language Models",
                    "abstract": "Language models have achieved remarkable performance on a wide range of tasks that require natural language understanding. Nevertheless, state-of-the-art models have generally struggled with tasks that require quantitative reasoning, such as solving mathematics, science, and engineering problems at the college level. To help close this gap, we introduce Minerva, a large language model pretrained on general natural language data and further trained on technical content. The model achieves state-of-the-art performance on technical benchmarks without the use of external tools. We also evaluate our model on over two hundred undergraduate-level problems in physics, biology, chemistry, economics, and other sciences that require quantitative reasoning, and find that the model can correctly answer nearly a third of them."
                },
                {
                    "title": "Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models",
                    "abstract": "Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit\"breakthrough\"behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting."
                },
                {
                    "title": "Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks",
                    "abstract": "How well can NLP models generalize to a variety of unseen tasks when provided with task instructions? To address this question, we first introduce Super-NaturalInstructions, a benchmark of 1,616 diverse NLP tasks and their expert-written instructions. Our collection covers 76 distinct task types, including but not limited to classification, extraction, infilling, sequence tagging, text rewriting, and text composition. This large and diverse collection of tasks enables rigorous benchmarking of cross-task generalization under instructions\u2014training models to follow instructions on a subset of tasks and evaluating them on the remaining unseen ones.Furthermore, we build Tk-Instruct, a transformer model trained to follow a variety of in-context instructions (plain language task definitions or k-shot examples). Our experiments show that Tk-Instruct outperforms existing instruction-following models such as InstructGPT by over 9% on our benchmark despite being an order of magnitude smaller. We further analyze generalization as a function of various scaling parameters, such as the number of observed tasks, the number of instances per task, and model sizes. We hope our dataset and model facilitate future progress towards more general-purpose NLP models."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Show Your Work: Scratchpads for Intermediate Computation with Language Models",
                    "abstract": "Large pre-trained language models perform remarkably well on tasks that can be done\"in one pass\", such as generating realistic text or synthesizing computer programs. However, they struggle with tasks that require unbounded multi-step computation, such as adding integers or executing programs. Surprisingly, we find that these same models are able to perform complex multi-step computations -- even in the few-shot regime -- when asked to perform the operation\"step by step\", showing the results of intermediate computations. In particular, we train transformers to perform multi-step computations by asking them to emit intermediate computation steps into a\"scratchpad\". On a series of increasingly complex tasks ranging from long addition to the execution of arbitrary programs, we show that scratchpads dramatically improve the ability of language models to perform multi-step computations."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "Multitask Prompted Training Enables Zero-Shot Task Generalization",
                    "abstract": "Large language models have recently been shown to attain reasonable zero-shot generalization on a diverse set of tasks (Brown et al., 2020). It has been hypothesized that this is a consequence of implicit multitask learning in language models' pretraining (Radford et al., 2019). Can zero-shot generalization instead be directly induced by explicit multitask learning? To test this question at scale, we develop a system for easily mapping any natural language tasks into a human-readable prompted form. We convert a large set of supervised datasets, each with multiple prompts with diverse wording. These prompted datasets allow for benchmarking the ability of a model to perform completely held-out tasks. We fine-tune a pretrained encoder-decoder model (Raffel et al., 2020; Lester et al., 2021) on this multitask mixture covering a wide variety of tasks. The model attains strong zero-shot performance on several standard datasets, often outperforming models up to 16x its size. Further, our approach attains strong performance on a subset of tasks from the BIG-bench benchmark, outperforming models up to 6x its size. All trained models are available at https://github.com/bigscience-workshop/t-zero and all prompts are available at https://github.com/bigscience-workshop/promptsource."
                },
                {
                    "title": "Finetuned Language Models Are Zero-Shot Learners",
                    "abstract": "This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning."
                },
                {
                    "title": "Program Synthesis with Large Language Models",
                    "abstract": "This paper explores the limits of the current generation of large language models for program synthesis in general purpose programming languages. We evaluate a collection of such models (with between 244M and 137B parameters) on two new benchmarks, MBPP and MathQA-Python, in both the few-shot and fine-tuning regimes. Our benchmarks are designed to measure the ability of these models to synthesize short Python programs from natural language descriptions. The Mostly Basic Programming Problems (MBPP) dataset contains 974 programming tasks, designed to be solvable by entry-level programmers. The MathQA-Python dataset, a Python version of the MathQA benchmark, contains 23914 problems that evaluate the ability of the models to synthesize code from more complex text. On both datasets, we find that synthesis performance scales log-linearly with model size. Our largest models, even without finetuning on a code dataset, can synthesize solutions to 59.6 percent of the problems from MBPP using few-shot learning with a well-designed prompt. Fine-tuning on a held-out portion of the dataset improves performance by about 10 percentage points across most model sizes. On the MathQA-Python dataset, the largest fine-tuned model achieves 83.8 percent accuracy. Going further, we study the model's ability to engage in dialog about code, incorporating human feedback to improve its solutions. We find that natural language feedback from a human halves the error rate compared to the model's initial prediction. Additionally, we conduct an error analysis to shed light on where these models fall short and what types of programs are most difficult to generate. Finally, we explore the semantic grounding of these models by fine-tuning them to predict the results of program execution. We find that even our best models are generally unable to predict the output of a program given a specific input."
                },
                {
                    "title": "Evaluating Large Language Models Trained on Code",
                    "abstract": "We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics."
                },
                {
                    "title": "Measuring Mathematical Problem Solving With the MATH Dataset",
                    "abstract": "Many intellectual endeavors require mathematical problem solving, but this skill remains beyond the capabilities of computers. To measure this ability in machine learning models, we introduce MATH, a new dataset of 12,500 challenging competition mathematics problems. Each problem in MATH has a full step-by-step solution which can be used to teach models to generate answer derivations and explanations. To facilitate future research and increase accuracy on MATH, we also contribute a large auxiliary pretraining dataset which helps teach models the fundamentals of mathematics. Even though we are able to increase accuracy on MATH, our results show that accuracy remains relatively low, even with enormous Transformer models. Moreover, we find that simply increasing budgets and model parameter counts will be impractical for achieving strong mathematical reasoning if scaling trends continue. While scaling Transformers is automatically solving most other text-based tasks, scaling is not currently solving MATH. To have more traction on mathematical problem solving we will likely need new algorithmic advancements from the broader research community."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "DeepSpeed: System Optimizations Enable Training Deep Learning Models with Over 100 Billion Parameters",
                    "abstract": "Explore new techniques in Microsoft's open source library called DeepSpeed, which advances large model training by improving scale, speed, cost, and usability, unlocking the ability to train 100-billion-parameter models. DeepSpeed is compatible with PyTorch. One piece of our library, called ZeRO, is a new parallelized optimizer that greatly reduces the resources needed for model and data parallelism while massively increasing the number of parameters that can be trained. Researchers have used these breakthroughs to create Turing Natural Language Generation (Turing-NLG), which at the time of its release was the largest publicly known language model at 17 billion parameters. In addition we will also go over our latest transformer kernel advancements that led the DeepSpeed team to achieve the world fastest BERT pretraining record. The Zero Redundancy Optimizer (ZeRO) is a novel memory optimization technology for large-scale distributed deep learning. ZeRO can train deep learning models with over 100 billion parameters on the current generation of GPU clusters at three to five times the throughput of the current best system. It also presents a clear path to training models with trillions of parameters, demonstrating an unprecedented leap in deep learning system technology. DeepSpeed brings state-of-the-art training techniques, such as ZeRO, optimized kernels, distributed training, mixed precision, and checkpointing, through lightweight APIs compatible with PyTorch. With just a few lines of code changes to your PyTorch model, you can leverage DeepSpeed to address underlying performance challenges and boost the speed and scale of your training."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "MathQA: Towards Interpretable Math Word Problem Solving with Operation-Based Formalisms",
                    "abstract": "We introduce a large-scale dataset of math word problems and an interpretable neural math problem solver by learning to map problems to their operation programs. Due to annotation challenges, current datasets in this domain have been either relatively small in scale or did not offer precise operational annotations over diverse problem types. We introduce a new representation language to model operation programs corresponding to each math problem that aim to improve both the performance and the interpretability of the learned models. Using this representation language, we significantly enhance the AQUA-RAT dataset with fully-specified operational programs. We additionally introduce a neural sequence-to-program model with automatic problem categorization. Our experiments show improvements over competitive baselines in our dataset as well as the AQUA-RAT dataset. The results are still lower than human performance indicating that the dataset poses new challenges for future research. Our dataset is available at https://math-qa.github.io/math-QA/"
                },
                {
                    "title": "Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge",
                    "abstract": "We present a new question set, text corpus, and baselines assembled to encourage AI research in advanced question answering. Together, these constitute the AI2 Reasoning Challenge (ARC), which requires far more powerful knowledge and reasoning than previous challenges such as SQuAD or SNLI. The ARC question set is partitioned into a Challenge Set and an Easy Set, where the Challenge Set contains only questions answered incorrectly by both a retrieval-based algorithm and a word co-occurence algorithm. The dataset contains only natural, grade-school science questions (authored for human tests), and is the largest public-domain set of this kind (7,787 questions). We test several baselines on the Challenge Set, including leading neural models from the SQuAD and SNLI tasks, and find that none are able to significantly outperform a random baseline, reflecting the difficult nature of this task. We are also releasing the ARC Corpus, a corpus of 14M science sentences relevant to the task, and implementations of the three neural baseline models tested. Can your model perform better? We pose ARC as a challenge to the community."
                },
                {
                    "title": "FastText.zip: Compressing text classification models",
                    "abstract": "We consider the problem of producing compact architectures for text classification, such that the full model fits in a limited amount of memory. After considering different solutions inspired by the hashing literature, we propose a method built upon product quantization to store the word embeddings. While the original technique leads to a loss in accuracy, we adapt this method to circumvent the quantization artifacts. As a result, our approach produces a text classifier, derived from the fastText approach, which at test time requires only a fraction of the memory compared to the original one, without noticeably sacrificing the quality in terms of classification accuracy. Our experiments carried out on several benchmarks show that our approach typically requires two orders of magnitude less memory than fastText while being only slightly inferior with respect to accuracy. As a result, it outperforms the state of the art by a good margin in terms of the compromise between memory usage and accuracy."
                }
            ],
            "categories": [
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively mine high-quality instruction-response pairs from the web to enhance the reasoning abilities of large language models?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of current large language models (LLMs) in performing complex reasoning tasks across various domains. By discovering and utilizing naturally occurring instruction data from the web, we can significantly improve the training datasets available for LLMs, leading to advancements in their reasoning capabilities. This could pave the way for more sophisticated applications in education, scientific research, and technology development, ultimately enhancing the utility of LLMs in real-world scenarios.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the dispersed nature of high-quality instruction data across the vast web corpus, making it difficult to discover and extract relevant instruction-response pairs. Naive approaches may fail due to the overwhelming volume of data and the need for effective filtering and refinement to ensure quality. Additionally, the presence of biased, non-diverse, or hallucinated data in synthesized instruction datasets complicates the task. Overcoming these technical obstacles requires a robust methodology for data recall, extraction, and refinement to ensure the mined pairs are both relevant and high-quality.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on human-annotated instruction datasets, which are limited in scale and often biased. Existing solutions have not effectively leveraged the vast amount of educational materials available on the web due to the challenges of data discovery and extraction. Barriers such as the lack of automated methods for identifying and refining instruction-response pairs from large corpora have hindered progress. Our approach differs by utilizing a three-step pipeline that systematically recalls, extracts, and refines data from the web, thereby improving upon prior work that relied heavily on manual curation or small-scale datasets.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology consists of a three-step pipeline: (1) **Recall step**: We create a diverse seed dataset from quiz websites and use it to train a fastText model for recalling documents from Common Crawl, resulting in 18 million documents. (2) **Extract step**: We employ open-source LLMs like Mixtral to extract approximately 5 million candidate question-answer pairs from these documents. (3) **Ref"
            }
        },
        "author_data": {
            "19954aba-1c37-43f4-b9fa-249cfed54e2f": {
                "pk": "19954aba-1c37-43f4-b9fa-249cfed54e2f",
                "name": "Xiang Yue",
                "collaborators": [
                    "Huan Sun",
                    "Shuang Zhou",
                    "Ziyu Yao",
                    "Bernhard Kratzwald",
                    "Stefan Feuerriegel",
                    "Yifei Li",
                    "Zeyi Liao"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Question Answering",
                    "Data Augmentation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "PHICON: Improving Generalization of Clinical Text De-identification Models via Data Augmentation",
                        "abstract": "De-identification is the task of identifying protected health information (PHI) in the clinical text. Existing neural de-identification models often fail to generalize to a new dataset. We propose a simple yet effective data augmentation method PHICON to alleviate the generalization issue. PHICON consists of PHI augmentation and Context augmentation, which creates augmented training corpora by replacing PHI entities with named-entities sampled from external sources, and by changing background context with synonym replacement or random word insertion, respectively. Experimental results on the i2b2 2006 and 2014 de-identification challenge datasets show that PHICON can help three selected de-identification models boost F1-score (by at most 8.6%) on cross-dataset test setting. We also discuss how much augmentation to use and how each augmentation method influences the performance."
                    },
                    {
                        "title": "Synthetic Question Value Estimation for Domain Adaptation of Question Answering",
                        "abstract": "Synthesizing QA pairs with a question generator (QG) on the target domain has become a popular approach for domain adaptation of question answering (QA) models. Since synthetic questions are often noisy in practice, existing work adapts scores from a pretrained QA (or QG) model as criteria to select high-quality questions. However, these scores do not directly serve the ultimate goal of improving QA performance on the target domain. In this paper, we introduce a novel idea of training a question value estimator (QVE) that directly estimates the usefulness of synthetic questions for improving the target-domain QA performance. By conducting comprehensive experiments, we show that the synthetic questions selected by QVE can help achieve better target-domain QA performance, in comparison with existing techniques. We additionally show that by using such questions and only around 15% of the human annotations on the target domain, we can achieve comparable performance to the fully-supervised baselines."
                    },
                    {
                        "title": "Practical Annotation Strategies for Question Answering Datasets",
                        "abstract": "Annotating datasets for question answering (QA) tasks is very costly, as it requires intensive manual labor and often domain-specific knowledge. Yet strategies for annotating QA datasets in a cost-effective manner are scarce. To provide a remedy for practitioners, our objective is to develop heuristic rules for annotating a subset of questions, so that the annotation cost is reduced while maintaining both in- and out-of-domain performance. For this, we conduct a large-scale analysis in order to derive practical recommendations. First, we demonstrate experimentally that more training samples contribute often only to a higher in-domain test-set performance, but do not help the model in generalizing to unseen datasets. Second, we develop a model-guided annotation strategy: it makes a recommendation with regard to which subset of samples should be annotated. Its effectiveness is demonstrated in a case study based on domain customization of QA to a clinical setting. Here, remarkably, annotating a stratified subset with only 1.2% of the original training set achieves 97.7% of the performance as if the complete dataset was annotated. Hence, the labeling effort can be reduced immensely. Altogether, our work fulfills a demand in practice when labeling budgets are limited and where thus recommendations are needed for annotating QA datasets more cost-effectively."
                    },
                    {
                        "title": "AttributionBench: How Hard is Automatic Attribution Evaluation?",
                        "abstract": "Modern generative search engines enhance the reliability of large language model (LLM) responses by providing cited evidence. However, evaluating the answer's attribution, i.e., whether every claim within the generated responses is fully supported by its cited evidence, remains an open problem. This verification, traditionally dependent on costly human evaluation, underscores the urgent need for automatic attribution evaluation methods. To bridge the gap in the absence of standardized benchmarks for these methods, we present AttributionBench, a comprehensive benchmark compiled from various existing attribution datasets. Our extensive experiments on AttributionBench reveal the challenges of automatic attribution evaluation, even for state-of-the-art LLMs. Specifically, our findings show that even a fine-tuned GPT-3.5 only achieves around 80% macro-F1 under a binary classification formulation. A detailed analysis of more than 300 error cases indicates that a majority of failures stem from the model's inability to process nuanced information, and the discrepancy between the information the model has access to and that human annotators do."
                    }
                ]
            },
            "ff9b19df-2d37-431a-8f1f-13ec05ee9d71": {
                "pk": "ff9b19df-2d37-431a-8f1f-13ec05ee9d71",
                "name": "Tuney Zheng",
                "collaborators": [
                    "Wenhao Huang",
                    "Ge Zhang",
                    "Jian Yang",
                    "Yizhi Li",
                    "Chenghua Lin",
                    "Shawn Gavin",
                    "Jiaheng Liu",
                    "Quehry Que",
                    "Noah Wang",
                    "Chenchen Zhang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Multimodal Learning",
                    "Benchmarking"
                ],
                "publications": [
                    {
                        "title": "LongIns: A Challenging Long-context Instruction-based Exam for LLMs",
                        "abstract": "The long-context capabilities of large language models (LLMs) have been a hot topic in recent years. To evaluate the performance of LLMs in different scenarios, various assessment benchmarks have emerged. However, as most of these benchmarks focus on identifying key information to answer questions, which mainly requires the retrieval ability of LLMs, these benchmarks can partially represent the reasoning performance of LLMs from large amounts of information. Meanwhile, although LLMs often claim to have context windows of 32k, 128k, 200k, or even longer, these benchmarks fail to reveal the actual supported length of these LLMs. To address these issues, we propose the LongIns benchmark dataset, a challenging long-context instruction-based exam for LLMs, which is built based on the existing instruction datasets. Specifically, in our LongIns, we introduce three evaluation settings: Global Instruction & Single Task (GIST), Local Instruction & Single Task (LIST), and Local Instruction & Multiple Tasks (LIMT). Based on LongIns, we perform comprehensive evaluations on existing LLMs and have the following important findings: (1). The top-performing GPT-4 with 128k context length performs poorly on the evaluation context window of 16k in our LongIns. (2). For the multi-hop reasoning ability of many existing LLMs, significant efforts are still needed under short context windows (less than 4k)."
                    },
                    {
                        "title": "A Comparative Study on Reasoning Patterns of OpenAI's o1 Model",
                        "abstract": "Enabling Large Language Models (LLMs) to handle a wider range of complex tasks (e.g., coding, math) has drawn great attention from many researchers. As LLMs continue to evolve, merely increasing the number of model parameters yields diminishing performance improvements and heavy computational costs. Recently, OpenAI's o1 model has shown that inference strategies (i.e., Test-time Compute methods) can also significantly enhance the reasoning capabilities of LLMs. However, the mechanisms behind these methods are still unexplored. In our work, to investigate the reasoning patterns of o1, we compare o1 with existing Test-time Compute methods (BoN, Step-wise BoN, Agent Workflow, and Self-Refine) by using OpenAI's GPT-4o as a backbone on general reasoning benchmarks in three domains (i.e., math, coding, commonsense reasoning). Specifically, first, our experiments show that the o1 model has achieved the best performance on most datasets. Second, as for the methods of searching diverse responses (e.g., BoN), we find the reward models' capability and the search space both limit the upper boundary of these methods. Third, as for the methods that break the problem into many sub-problems, the Agent Workflow has achieved better performance than Step-wise BoN due to the domain-specific system prompt for planning better reasoning processes. Fourth, it is worth mentioning that we have summarized six reasoning patterns of o1, and provided a detailed analysis on several reasoning benchmarks."
                    },
                    {
                        "title": "LIME: Less Is More for MLLM Evaluation",
                        "abstract": "Multimodal Large Language Models (MLLMs) are evaluated on various benchmarks, such as image captioning, visual question answering, and reasoning. However, many of these benchmarks include overly simple or uninformative samples, complicating the effective distinction of different MLLMs' performance. Furthermore, evaluating models across numerous benchmarks incurs a significant computational burden. To address these issues, we propose LIME (Less Is More for MLLM Evaluation), a refined and efficient benchmark curated through a semi-automated pipeline. This pipeline filters out uninformative samples and eliminates answer leakage by focusing on tasks that necessitate image-based understanding. Our experiments indicate that LIME reduces the number of samples by 76% and evaluation time by 77%, while also providing a more effective means of distinguishing the capabilities of different models. Notably, we find that traditional automatic metrics, such as CIDEr, are inadequate for assessing MLLMs' captioning performance; excluding the caption task score yields a more accurate reflection of overall model performance. All code and data are available at https://github.com/kangreen0210/LIME."
                    },
                    {
                        "title": "MAP-Neo: Highly Capable and Transparent Bilingual Large Language Model Series",
                        "abstract": "Large Language Models (LLMs) have made great strides in recent years to achieve unprecedented performance across different tasks. However, due to commercial interest, the most competitive models like GPT, Gemini, and Claude have been gated behind proprietary interfaces without disclosing the training details. Recently, many institutions have open-sourced several strong LLMs like LLaMA-3, comparable to existing closed-source LLMs. However, only the model's weights are provided with most details (e.g., intermediate checkpoints, pre-training corpus, and training code, etc.) being undisclosed. To improve the transparency of LLMs, the research community has formed to open-source truly open LLMs (e.g., Pythia, Amber, OLMo), where more details (e.g., pre-training corpus and training code) are being provided. These models have greatly advanced the scientific study of these large models including their strengths, weaknesses, biases and risks. However, we observe that the existing truly open LLMs on reasoning, knowledge, and coding tasks are still inferior to existing state-of-the-art LLMs with similar model sizes. To this end, we open-source MAP-Neo, a highly capable and transparent bilingual language model with 7B parameters trained from scratch on 4.5T high-quality tokens. Our MAP-Neo is the first fully open-sourced bilingual LLM with comparable performance compared to existing state-of-the-art LLMs. Moreover, we open-source all details to reproduce our MAP-Neo, where the cleaned pre-training corpus, data cleaning pipeline, checkpoints, and well-optimized training/evaluation framework are provided. Finally, we hope our MAP-Neo will enhance and strengthen the open research community and inspire more innovations and creativities to facilitate the further improvements of LLMs."
                    }
                ]
            },
            "578a6310-76c4-427b-a392-7828b8600ea8": {
                "pk": "578a6310-76c4-427b-a392-7828b8600ea8",
                "name": "Ge Zhang",
                "collaborators": [
                    "Wangzhe Du",
                    "Salvatore Torquato"
                ],
                "domain": [
                    "Machine Learning",
                    "Generative Models",
                    "Statistical Mechanics",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Correlation between entropy and generalizability in a neural network",
                        "abstract": "Although neural networks can solve very complex machine-learning problems, the theoretical reason for their generalizability is still not fully understood. Here we use Wang-Landau Mote Carlo algorithm to calculate the entropy (logarithm of the volume of a part of the parameter space) at a given test accuracy, and a given training loss function value or training accuracy. Our results show that entropical forces help generalizability. Although our study is on a very simple application of neural networks (a spiral dataset and a small, fully-connected neural network), our approach should be useful in explaining the generalizability of more complicated neural networks in future works."
                    },
                    {
                        "title": "RSC-VAE: Recoding Semantic Consistency Based VAE for One-Class Novelty Detection",
                        "abstract": "In recent years, there is an increasing interests in reconstruction based generative models for image One-Class Novelty Detection, most of which only focus on image-level information. While in this paper, we further exploit the latent space of Variational Auto-encoder (VAE), a typical reconstruction based model, and we innovatively divide it into three regions: Normal/Anomalous/Unknown-semantic-region. Based on this hypothesis, we propose a new VAE architecture, Recoding Semantic Consistency Based VAE (RSC-VAE), combining VAE with recoding mechanism and constraining the semantic consistency of two encodings. We come up with three training modes of RSC-VAE: 1. One-Class Training Mode, alleviating False Positive problem of normal samples; 2. Distributionally-Shifted Training Mode, alleviating False Negative problem of anomalous samples; 3. Extremely-Imbalanced Training Mode, introducing a small number of anomalous samples for training to enhance the second mode. The experimental results on multiple datasets demonstrate that our mechanism achieves state-of-the-art performance in various baselines including VAE."
                    },
                    {
                        "title": "Realizable hyperuniform and nonhyperuniform particle configurations with targeted spectral functions via effective pair interactions",
                        "abstract": "The capacity to identify realizable many-body configurations associated with targeted functional forms for the pair correlation function $g_2(r)$ or its corresponding structure factor $S(k)$ is of great fundamental and practical importance. A major aim of this paper is to expand our theoretical knowledge of the class of pair correlation functions or structure factors that are realizable by classical disordered ensembles of particle configurations. We first introduce a theoretical formalism that provides a means to draw classical particle configurations from canonical ensembles with certain pairwise-additive potentials that could correspond to targeted analytical functional forms for the structure factor. This formulation enables us to devise an improved algorithm to construct systematically canonical-ensemble particle configurations with such targeted pair statistics. We test the algorithm by targeting several different structure factors across dimensions that are known to be realizable and one hyperuniform target that is known to be nontrivially unrealizable. Subsequently, we also target several families of structure-factor functions that meet the known necessary realizability conditions but were heretofore not known to be realizable. Moreover, we also explore unusual nonhyperuniform targets, including \"hyposurficial\" and \"anti-hyperuniform\" examples. In all of these instances, the targeted structure factors were achieved with high accuracy, suggesting that they are indeed realizable by equilibrium configurations with pairwise interactions at positive temperatures. Our findings lead us to the conjecture that any realizable structure factor corresponding to either a translationally invariant equilibrium or nonequilibrium system can be attained by an equilibrium ensemble involving only effective pair interactions."
                    }
                ]
            },
            "65c60f32-29cc-4dbd-acc4-cc3360e7eecb": {
                "pk": "65c60f32-29cc-4dbd-acc4-cc3360e7eecb",
                "name": "Wenhu Chen",
                "collaborators": [
                    "William Wang",
                    "Xifeng Yan",
                    "Xinyi Wang",
                    "William Yang Wang",
                    "Xueguang Ma",
                    "Yilin Shen",
                    "Hongxia Jin",
                    "Yu Su",
                    "William W. Cohen",
                    "Wenhan Xiong"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Image Generation",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "Large Language Models are few(1)-shot Table Reasoners",
                        "abstract": "Recent literature has shown that large language models (LLMs) are generally excellent few-shot reasoners to solve text reasoning tasks. However, the capability of LLMs on table reasoning tasks is yet to be explored. In this paper, we aim at understanding how well LLMs can perform table-related tasks with few-shot in-context learning. Specifically, we evaluated LLMs on popular table QA and fact verification datasets like WikiTableQuestion, FetaQA, TabFact, and FEVEROUS and found that LLMs are competent at complex reasoning over table structures, though these models are not pre-trained on any table corpus. When combined with `chain of thoughts' prompting, LLMs can achieve very strong performance with only a 1-shot demonstration, even on par with some SoTA models. We show that LLMs are even more competent at generating comprehensive long-form answers on FetaQA than tuned T5-large. We further manually studied the reasoning chains elicited from LLMs and found that these reasoning chains are highly consistent with the underlying semantic form. We believe that LLMs can serve as a simple yet generic baseline for future research. The code and data are released in https://github.com/wenhuchen/TableCoT."
                    },
                    {
                        "title": "A Variational Dirichlet Framework for Out-of-Distribution Detection",
                        "abstract": "With the recently rapid development in deep learning, deep neural networks have been widely adopted in many real-life applications. However, deep neural networks are also known to have very little control over its uncertainty for unseen examples, which potentially causes very harmful and annoying consequences in practical scenarios. In this paper, we are particularly interested in designing a higher-order uncertainty metric for deep neural networks and investigate its effectiveness under the out-of-distribution detection task proposed by~\\cite{hendrycks2016baseline}. Our method first assumes there exists an underlying higher-order distribution $\\mathbb{P}(z)$, which controls label-wise categorical distribution $\\mathbb{P}(y)$ over classes on the K-dimension simplex, and then approximate such higher-order distribution via parameterized posterior function $p_{\\theta}(z|x)$ under variational inference framework, finally we use the entropy of learned posterior distribution $p_{\\theta}(z|x)$ as uncertainty measure to detect out-of-distribution examples. Further, we propose an auxiliary objective function to discriminate against synthesized adversarial examples to further increase the robustness of the proposed uncertainty measure. Through comprehensive experiments on various datasets, our proposed framework is demonstrated to consistently outperform competing algorithms."
                    },
                    {
                        "title": "Variational Knowledge Graph Reasoning",
                        "abstract": "Inferring missing links in knowledge graphs (KG) has attracted a lot of attention from the research community. In this paper, we tackle a practical query answering task involving predicting the relation of a given entity pair. We frame this prediction problem as an inference problem in a probabilistic graphical model and aim at resolving it from a variational inference perspective. In order to model the relation between the query entity pair, we assume that there exists an underlying latent variable (paths connecting two nodes) in the KG, which carries the equivalent semantics of their relations. However, due to the intractability of connections in large KGs, we propose to use variation inference to maximize the evidence lower bound. More specifically, our framework (\\textsc{Diva}) is composed of three modules, i.e. a posterior approximator, a prior (path finder), and a likelihood (path reasoner). By using variational inference, we are able to incorporate them closely into a unified architecture and jointly optimize them to perform KG reasoning. With active interactions among these sub-modules, \\textsc{Diva} is better at handling noise and coping with more complex reasoning scenarios. In order to evaluate our method, we conduct the experiment of the link prediction task on multiple datasets and achieve state-of-the-art performances on both datasets."
                    },
                    {
                        "title": "A Dataset for Answering Time-Sensitive Questions",
                        "abstract": "Time is an important dimension in our physical world. Lots of facts can evolve with respect to time. For example, the U.S. President might change every four years. Therefore, it is important to consider the time dimension and empower the existing QA models to reason over time. However, the existing QA datasets contain rather few time-sensitive questions, hence not suitable for diagnosing or benchmarking the model's temporal reasoning capability. In order to promote research in this direction, we propose to construct a time-sensitive QA dataset. The dataset is constructed by 1) mining time-evolving facts from WikiData and aligning them to their corresponding Wikipedia page, 2) employing crowd workers to verify and calibrate these noisy facts, 3) generating question-answer pairs based on the annotated time-sensitive facts. Our dataset poses challenges in the aspect of both temporal understanding and temporal reasoning. We evaluate different SoTA long-document QA systems like BigBird and FiD on our dataset. The best-performing model FiD can only achieve 46\\% accuracy, still far behind the human performance of 87\\%. We demonstrate that these models are still lacking the ability to perform consistent temporal reasoning. Therefore, we believe that our dataset could serve as a benchmark to develop NLP models more sensitive to temporal shifts. The dataset and code are released in~\\url{https://github.com/wenhuchen/Time-Sensitive-QA}."
                    },
                    {
                        "title": "A Semi-supervised Framework for Image Captioning",
                        "abstract": "State-of-the-art approaches for image captioning require supervised training data consisting of captions with paired image data. These methods are typically unable to use unsupervised data such as textual data with no corresponding images, which is a much more abundant commodity. We here propose a novel way of using such textual data by artificially generating missing visual information. We evaluate this learning approach on a newly designed model that detects visual concepts present in an image and feed them to a reviewer-decoder architecture with an attention mechanism. Unlike previous approaches that encode visual concepts using word embeddings, we instead suggest using regional image features which capture more intrinsic information. The main benefit of this architecture is that it synthesizes meaningful thought vectors that capture salient image properties and then applies a soft attentive decoder to decode the thought vectors and generate image captions. We evaluate our model on both Microsoft COCO and Flickr30K datasets and demonstrate that this model combined with our semi-supervised learning method can largely improve performance and help the model to generate more accurate and diverse captions."
                    },
                    {
                        "title": "Task-adaptive Pre-training and Self-training are Complementary for Natural Language Understanding",
                        "abstract": "Task-adaptive pre-training (TAPT) and Self-training (ST) have emerged as the major semi-supervised approaches to improve natural language understanding (NLU) tasks with massive amount of unlabeled data. However, it's unclear whether they learn similar representations or they can be effectively combined. In this paper, we show that TAPT and ST can be complementary with simple TFS protocol by following TAPT -> Finetuning -> Self-training (TFS) process. Experimental results show that TFS protocol can effectively utilize unlabeled data to achieve strong combined gains consistently across six datasets covering sentiment classification, paraphrase identification, natural language inference, named entity recognition and dialogue slot classification. We investigate various semi-supervised settings and consistently show that gains from TAPT and ST can be strongly additive by following TFS procedure. We hope that TFS could serve as an important semi-supervised baseline for future NLP studies."
                    },
                    {
                        "title": "Modeling Token-level Uncertainty to Learn Unknown Concepts in SLU via Calibrated Dirichlet Prior RNN",
                        "abstract": "One major task of spoken language understanding (SLU) in modern personal assistants is to extract semantic concepts from an utterance, called slot filling. Although existing slot filling models attempted to improve extracting new concepts that are not seen in training data, the performance in practice is still not satisfied. Recent research collected question and answer annotated data to learn what is unknown and should be asked, yet not practically scalable due to the heavy data collection effort. In this paper, we incorporate softmax-based slot filling neural architectures to model the sequence uncertainty without question supervision. We design a Dirichlet Prior RNN to model high-order uncertainty by degenerating as softmax layer for RNN model training. To further enhance the uncertainty modeling robustness, we propose a novel multi-task training to calibrate the Dirichlet concentration parameters. We collect unseen concepts to create two test datasets from SLU benchmark datasets Snips and ATIS. On these two and another existing Concept Learning benchmark datasets, we show that our approach significantly outperforms state-of-the-art approaches by up to 8.18%. Our method is generic and can be applied to any RNN or Transformer based slot filling models with a softmax layer."
                    },
                    {
                        "title": "UniRAG: Universal Retrieval Augmentation for Multi-Modal Large Language Models",
                        "abstract": "Recently, Multi-Modal (MM) Large Language Models (LLMs) have unlocked many complex use-cases that require MM understanding (e.g., image captioning or visual question answering) and MM generation (e.g., text-guided image generation or editing) capabilities. To further improve the output fidelity of MM-LLMs we introduce UniRAG, a plug-and-play technique that adds relevant retrieved information to prompts as few-shot examples during inference. Unlike the common belief that Retrieval Augmentation (RA) mainly improves generation or understanding of uncommon entities, our evaluation results on the MSCOCO dataset with common entities show that both proprietary models like GPT-4o and Gemini-Pro and smaller open-source models like LLaVA, LaVIT, and Emu2 significantly enhance their generation quality when their input prompts are augmented with relevant information retrieved by MM retrievers like UniIR models."
                    },
                    {
                        "title": "DreamEdit: Subject-driven Image Editing",
                        "abstract": "Subject-driven image generation aims at generating images containing customized subjects, which has recently drawn enormous attention from the research community. However, the previous works cannot precisely control the background and position of the target subject. In this work, we aspire to fill the void and propose two novel subject-driven sub-tasks, i.e., Subject Replacement and Subject Addition. The new tasks are challenging in multiple aspects: replacing a subject with a customized one can change its shape, texture, and color, while adding a target subject to a designated position in a provided scene necessitates a context-aware posture. To conquer these two novel tasks, we first manually curate a new dataset DreamEditBench containing 22 different types of subjects, and 440 source images with different difficulty levels. We plan to host DreamEditBench as a platform and hire trained evaluators for standard human evaluation. We also devise an innovative method DreamEditor to resolve these tasks by performing iterative generation, which enables a smooth adaptation to the customized subject. In this project, we conduct automatic and human evaluations to understand the performance of DreamEditor and baselines on DreamEditBench. For Subject Replacement, we found that the existing models are sensitive to the shape and color of the original subject. The model failure rate will dramatically increase when the source and target subjects are highly different. For Subject Addition, we found that the existing models cannot easily blend the customized subjects into the background smoothly, leading to noticeable artifacts in the generated image. We hope DreamEditBench can become a standard platform to enable future investigations toward building more controllable subject-driven image editing. Our project homepage is https://dreameditbenchteam.github.io/."
                    },
                    {
                        "title": "Augmenting Black-box LLMs with Medical Textbooks for Biomedical Question Answering (Published in Findings of EMNLP 2024)",
                        "abstract": "Large-scale language models (LLMs) like ChatGPT have demonstrated impressive abilities in generating responses based on human instructions. However, their use in the medical field can be challenging due to their lack of specific, in-depth knowledge. In this study, we present a system called LLMs Augmented with Medical Textbooks (LLM-AMT) designed to enhance the proficiency of LLMs in specialized domains. LLM-AMT integrates authoritative medical textbooks into the LLMs' framework using plug-and-play modules. These modules include a Query Augmenter, a Hybrid Textbook Retriever, and a Knowledge Self-Refiner. Together, they incorporate authoritative medical knowledge. Additionally, an LLM Reader aids in contextual understanding. Our experimental results on three medical QA tasks demonstrate that LLMAMT significantly improves response quality, with accuracy gains ranging from 11.6% to 16.6%. Notably, with GPT-4-Turbo as the base model, LLM-AMT outperforms the specialized Med-PaLM 2 model pre-trained on a massive amount of medical corpus by 2-3%. We found that despite being 100x smaller in size, medical textbooks as a retrieval corpus is proven to be a more effective knowledge database than Wikipedia in the medical domain, boosting performance by 7.8%-13.7%."
                    },
                    {
                        "title": "LongRAG: Enhancing Retrieval-Augmented Generation with Long-context LLMs",
                        "abstract": "In traditional RAG framework, the basic retrieval units are normally short. The common retrievers like DPR normally work with 100-word Wikipedia paragraphs. Such a design forces the retriever to search over a large corpus to find the `needle' unit. In contrast, the readers only need to generate answers from the short retrieved units. The imbalanced `heavy' retriever and `light' reader design can lead to sub-optimal performance. The loss of contextual information in the short, chunked units may increase the likelihood of introducing hard negatives during the retrieval stage. Additionally, the reader might not fully leverage the capabilities of recent advancements in LLMs. In order to alleviate the imbalance, we propose a new framework LongRAG, consisting of a `long retriever' and a `long reader'. In the two Wikipedia-based datasets, NQ and HotpotQA, LongRAG processes the entire Wikipedia corpus into 4K-token units by grouping related documents. By increasing the unit size, we significantly reduce the total number of units. This greatly reduces the burden on the retriever, resulting in strong retrieval performance with only a few (less than 8) top units. Without requiring any training, LongRAG achieves an EM of 62.7% on NQ and 64.3% on HotpotQA, which are on par with the (fully-trained) SoTA model. Furthermore, we test on two non-Wikipedia-based datasets, Qasper and MultiFieldQA-en. LongRAG processes each individual document as a single (long) unit rather than chunking them into smaller units. By doing so, we achieve an F1 score of 25.9% on Qasper and 57.5% on MultiFieldQA-en. Our study offers insights into the future roadmap for combining RAG with long-context LLMs."
                    },
                    {
                        "title": "KGPT: Knowledge-Grounded Pre-Training for Data-to-Text Generation",
                        "abstract": "Data-to-text generation has recently attracted substantial interests due to its wide applications. Existing methods have shown impressive performance on an array of tasks. However, they rely on a significant amount of labeled data for each task, which is costly to acquire and thus limits their application to new tasks and domains. In this paper, we propose to leverage pre-training and transfer learning to address this issue. We propose a knowledge-grounded pre-training (KGPT), which consists of two parts, 1) a general knowledge-grounded generation model to generate knowledge-enriched text. 2) a pre-training paradigm on a massive knowledge-grounded text corpus crawled from the web. The pre-trained model can be fine-tuned on various data-to-text generation tasks to generate task-specific text. We adopt three settings, namely fully-supervised, zero-shot, few-shot to evaluate its effectiveness. Under the fully-supervised setting, our model can achieve remarkable gains over the known baselines. Under zero-shot setting, our model without seeing any examples achieves over 30 ROUGE-L on WebNLG while all other baselines fail. Under the few-shot setting, our model only needs about one-fifteenth as many labeled examples to achieve the same level of performance as baseline models. These experiments consistently prove the strong generalization ability of our proposed framework https://github.com/wenhuchen/KGPT."
                    },
                    {
                        "title": "Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks",
                        "abstract": "Recently, there has been significant progress in teaching language models to perform step-by-step reasoning to solve complex numerical reasoning tasks. Chain-of-thoughts prompting (CoT) is by far the state-of-art method for these tasks. CoT uses language models to perform both reasoning and computation in the multi-step `thought' process. To disentangle computation from reasoning, we propose `Program of Thoughts' (PoT), which uses language models (mainly Codex) to express the reasoning process as a program. The computation is relegated to an external computer, which executes the generated programs to derive the answer. We evaluate PoT on five math word problem datasets (GSM, AQuA, SVAMP, TabMWP, MultiArith) and three financial-QA datasets (FinQA, ConvFinQA, TATQA) for both few-shot and zero-shot setups. Under both few-shot and zero-shot settings, PoT can show an average performance gain over CoT by around 12\\% across all the evaluated datasets. By combining PoT with self-consistency decoding, we can achieve SoTA performance on all math problem datasets and near-SoTA performance on financial datasets. All of our data and code are released in Github https://github.com/wenhuchen/Program-of-Thoughts"
                    },
                    {
                        "title": "Re-Imagen: Retrieval-Augmented Text-to-Image Generator",
                        "abstract": "Research on text-to-image generation has witnessed significant progress in generating diverse and photo-realistic images, driven by diffusion and auto-regressive models trained on large-scale image-text data. Though state-of-the-art models can generate high-quality images of common entities, they often have difficulty generating images of uncommon entities, such as `Chortai (dog)' or `Picarones (food)'. To tackle this issue, we present the Retrieval-Augmented Text-to-Image Generator (Re-Imagen), a generative model that uses retrieved information to produce high-fidelity and faithful images, even for rare or unseen entities. Given a text prompt, Re-Imagen accesses an external multi-modal knowledge base to retrieve relevant (image, text) pairs and uses them as references to generate the image. With this retrieval step, Re-Imagen is augmented with the knowledge of high-level semantics and low-level visual details of the mentioned entities, and thus improves its accuracy in generating the entities' visual appearances. We train Re-Imagen on a constructed dataset containing (image, text, retrieval) triples to teach the model to ground on both text prompt and retrieval. Furthermore, we develop a new sampling strategy to interleave the classifier-free guidance for text and retrieval conditions to balance the text and retrieval alignment. Re-Imagen achieves significant gain on FID score over COCO and WikiImage. To further evaluate the capabilities of the model, we introduce EntityDrawBench, a new benchmark that evaluates image generation for diverse entities, from frequent to rare, across multiple object categories including dogs, foods, landmarks, birds, and characters. Human evaluation on EntityDrawBench shows that Re-Imagen can significantly improve the fidelity of generated images, especially on less frequent entities."
                    },
                    {
                        "title": "Guided Alignment Training for Topic-Aware Neural Machine Translation",
                        "abstract": "In this paper, we propose an effective way for biasing the attention mechanism of a sequence-to-sequence neural machine translation (NMT) model towards the well-studied statistical word alignment models. We show that our novel guided alignment training approach improves translation quality on real-life e-commerce texts consisting of product titles and descriptions, overcoming the problems posed by many unknown words and a large type/token ratio. We also show that meta-data associated with input texts such as topic or category information can significantly improve translation quality when used as an additional signal to the decoder part of the network. With both novel features, the BLEU score of the NMT system on a product title set improves from 18.6 to 21.3%. Even larger MT quality gains are obtained through domain adaptation of a general domain NMT system to e-commerce data. The developed NMT system also performs well on the IWSLT speech translation task, where an ensemble of four variant systems outperforms the phrase-based baseline by 2.1% BLEU absolute."
                    },
                    {
                        "title": "Counterfactual Maximum Likelihood Estimation for Training Deep Networks",
                        "abstract": "Although deep learning models have driven state-of-the-art performance on a wide array of tasks, they are prone to spurious correlations that should not be learned as predictive clues. To mitigate this problem, we propose a causality-based training framework to reduce the spurious correlations caused by observed confounders. We give theoretical analysis on the underlying general Structural Causal Model (SCM) and propose to perform Maximum Likelihood Estimation (MLE) on the interventional distribution instead of the observational distribution, namely Counterfactual Maximum Likelihood Estimation (CMLE). As the interventional distribution, in general, is hidden from the observational data, we then derive two different upper bounds of the expected negative log-likelihood and propose two general algorithms, Implicit CMLE and Explicit CMLE, for causal predictions of deep learning models using observational data. We conduct experiments on both simulated data and two real-world tasks: Natural Language Inference (NLI) and Image Captioning. The results show that CMLE methods outperform the regular MLE method in terms of out-of-domain generalization performance and reducing spurious correlations, while maintaining comparable performance on the regular evaluations."
                    },
                    {
                        "title": "Synthesizing Coherent Story with Auto-Regressive Latent Diffusion Models",
                        "abstract": "Conditioned diffusion models have demonstrated state-of-the-art text-to-image synthesis capacity. Recently, most works focus on synthesizing independent images; While for real-world applications, it is common and necessary to generate a series of coherent images for story-stelling. In this work, we mainly focus on story visualization and continuation tasks and propose AR-LDM, a latent diffusion model auto-regressively conditioned on history captions and generated images. Moreover, AR-LDM can generalize to new characters through adaptation. To our best knowledge, this is the first work successfully leveraging diffusion models for coherent visual story synthesizing. Quantitative results show that AR-LDM achieves SoTA FID scores on PororoSV, FlintstonesSV, and the newly introduced challenging dataset VIST containing natural images. Large-scale human evaluations show that AR-LDM has superior performance in terms of quality, relevance, and consistency."
                    },
                    {
                        "title": "Knowledge of Knowledge: Exploring Known-Unknowns Uncertainty with Large Language Models",
                        "abstract": "This paper investigates the capabilities of Large Language Models (LLMs) in the context of understanding their knowledge and uncertainty over questions. Specifically, we focus on addressing known-unknown questions, characterized by high uncertainty due to the absence of definitive answers. To facilitate our study, we collect a new dataset with Known-Unknown Questions (KUQ) and establish a categorization framework to clarify the origins of uncertainty in such queries. Subsequently, we examine the performance of open-source LLMs, fine-tuned using this dataset, in distinguishing between known and unknown queries within open-ended question-answering scenarios. The fine-tuned models demonstrated a significant improvement, achieving a considerable increase in F1-score relative to their pre-fine-tuning state. Through a comprehensive analysis, we reveal insights into the models' improved uncertainty articulation and their consequent efficacy in multi-agent debates. These findings help us understand how LLMs can be trained to identify and express uncertainty, improving our knowledge of how they understand and express complex or unclear information."
                    },
                    {
                        "title": "MagicBrush: A Manually Annotated Dataset for Instruction-Guided Image Editing",
                        "abstract": "Text-guided image editing is widely needed in daily life, ranging from personal use to professional applications such as Photoshop. However, existing methods are either zero-shot or trained on an automatically synthesized dataset, which contains a high volume of noise. Thus, they still require lots of manual tuning to produce desirable outcomes in practice. To address this issue, we introduce MagicBrush (https://osu-nlp-group.github.io/MagicBrush/), the first large-scale, manually annotated dataset for instruction-guided real image editing that covers diverse scenarios: single-turn, multi-turn, mask-provided, and mask-free editing. MagicBrush comprises over 10K manually annotated triplets (source image, instruction, target image), which supports trainining large-scale text-guided image editing models. We fine-tune InstructPix2Pix on MagicBrush and show that the new model can produce much better images according to human evaluation. We further conduct extensive experiments to evaluate current image editing baselines from multiple dimensions including quantitative, qualitative, and human evaluations. The results reveal the challenging nature of our dataset and the gap between current baselines and real-world editing needs."
                    }
                ]
            }
        }
    },
    "2406.04333": {
        "paper_data": {
            "title": "BitsFusion: 1.99 bits Weight Quantization of Diffusion Model",
            "url": "http://arxiv.org/abs/2406.04333v1",
            "arxiv_id": "2406.04333",
            "authors": [
                "Yang Sui",
                "Yanyu Li",
                "Anil Kag",
                "Yerlan Idelbayev",
                "Junli Cao",
                "Ju Hu",
                "Dhritiman Sagar",
                "Bo Yuan",
                "Sergey Tulyakov",
                "Jian Ren"
            ],
            "abstract": "Diffusion-based image generation models have achieved great success in recent years by showing the capability of synthesizing high-quality content. However, these models contain a huge number of parameters, resulting in a significantly large model size. Saving and transferring them is a major bottleneck for various applications, especially those running on resource-constrained devices. In this work, we develop a novel weight quantization method that quantizes the UNet from Stable Diffusion v1.5 to 1.99 bits, achieving a model with 7.9X smaller size while exhibiting even better generation quality than the original one. Our approach includes several novel techniques, such as assigning optimal bits to each layer, initializing the quantized model for better performance, and improving the training strategy to dramatically reduce quantization error. Furthermore, we extensively evaluate our quantized model across various benchmark datasets and through human evaluation to demonstrate its superior generation quality.",
            "introduction": "   1 Introduction  Recent efforts in developing diffusion-based image generation models\u00a0[69, 31, 71, 21, 72] have demonstrated remarkable results in synthesizing high-fidelity and photo-realistic images, leading to various applications such as content creation and editing\u00a0[60, 59, 53, 63, 81, 79, 46, 39], video generation\u00a0[20, 67, 3, 1, 16, 52, 15], and 3D asset synthesis\u00a0[78, 42, 56, 66, 57], among others. However, Diffusion Models (DMs) come with the drawback of a large number of parameters, e.g., millions or even billions, causing significant burdens for transferring and storing due to the bulky model size, especially on resource-constrained hardware such as mobile and wearable devices.   Existing studies have explored reducing the model size of large-scale text-to-image diffusion models by designing efficient architectures and network pruning\u00a0[40, 82, 32]. These approaches usually require significant amounts of training due to the changes made to the pre-trained networks. Another promising direction for model storage reduction is quantization\u00a0[12, 30], where floating-point weights are converted to low-bit fixed-point representations, thereby saving computation memory and storage.   There have been emerging efforts on compressing the DMs through quantization\u00a0[65, 37, 17, 38]. However, these approaches still face several major challenges, especially when quantizing large-scale text-to-image diffusion models like Stable Diffusion v1.5 (SD-v1.5)\u00a0[62]. First, many of these methods are developed on relatively small-scale DMs trained on constrained datasets. For example, models trained on CIFAR-10 require modest storage of around 100100100100 MB\u00a0[21, 38]. In contrast, SD-v1.5 necessitates 3.443.443.443.44 GB of storage in a full-precision format. Adapting these methods to SD-v1.5 remains to be a challenging problem. Second, current arts mainly focus on quantizing weights to 4444 bits. How to quantize the model to extremely low bit is not well studied. Third, there is a lack of fair and extensive evaluation of how quantization methods perform on large-scale DMs, i.e., SD-v1.5.   To tackle the above challenges, this work proposes BitsFusion, a quantization-aware training framework that employs a series of novel techniques to compress the weights of pre-trained large-scale DMs into extremely low bits (i.e., 1.99 bits), achieving even better performance (i.e., higher image quality and better text-image alignment). Consequently, we compress the 1.721.721.721.72 GB UNet (FP16)111For SD-v1.5, we measure the generation quality using the FP32 format. However, since SD-v1.5 FP16 has similar performance to SD-v1.5 FP32, we use SD-v1.5 FP16 to calculate our compression ratio. of SD-v1.5 into a 219219219219 MB model, achieving a 7.9\u00d77.9\\times7.9 \u00d7 compression ratio. Specifically, our contributions can be summarized into the following four dimensions:   \u2022  Mixed-Precision Quantization for DMs. We propose an effective approach for quantizing DMs in a mixed-precision manner. First, we thoroughly analyze the appropriate metrics to understand the quantization error in the quantized DMs (Sec.\u00a03.2). Second, based on the analysis, we quantize different layers into different bits according to their quantization error (Sec.\u00a03.3).    \u2022  Initialization for Quantized DMs. We introduce several techniques to initialize the quantized model to improve performance, including time embedding pre-computing and caching, adding balance integer, and alternating optimization for scaling factor initialization (Sec.\u00a04.1).    \u2022  Improved Training Pipeline for Quantized DMs. We improve the training pipeline for the quantized model with the proposed two-stage training approach (Sec.\u00a04.2). In the first stage, we use the full-precision model as",
            "references": [
                {
                    "title": "BinaryDM: Accurate Weight Binarization for Efficient Diffusion Models",
                    "abstract": "With the advancement of diffusion models (DMs) and the substantially increased computational requirements, quantization emerges as a practical solution to obtain compact and efficient low-bit DMs. However, the highly discrete representation leads to severe accuracy degradation, hindering the quantization of diffusion models to ultra-low bit-widths. This paper proposes a novel weight binarization approach for DMs, namely BinaryDM, pushing binarized DMs to be accurate and efficient by improving the representation and optimization. From the representation perspective, we present an Evolvable-Basis Binarizer (EBB) to enable a smooth evolution of DMs from full-precision to accurately binarized. EBB enhances information representation in the initial stage through the flexible combination of multiple binary bases and applies regularization to evolve into efficient single-basis binarization. The evolution only occurs in the head and tail of the DM architecture to retain the stability of training. From the optimization perspective, a Low-rank Representation Mimicking (LRM) is applied to assist the optimization of binarized DMs. The LRM mimics the representations of full-precision DMs in low-rank space, alleviating the direction ambiguity of the optimization process caused by fine-grained alignment. Comprehensive experiments demonstrate that BinaryDM achieves significant accuracy and efficiency gains compared to SOTA quantization methods of DMs under ultra-low bit-widths. With 1-bit weight and 4-bit activation (W1A4), BinaryDM achieves as low as 7.74 FID and saves the performance from collapse (baseline FID 10.87). As the first binarization method for diffusion models, W1A4 BinaryDM achieves impressive 15.2x OPs and 29.2x model size savings, showcasing its substantial potential for edge deployment."
                },
                {
                    "title": "QNCD: Quantization Noise Correction for Diffusion Models",
                    "abstract": "Diffusion models have revolutionized image synthesis, setting new benchmarks in quality and creativity. However, their widespread adoption is hindered by the intensive computation required during the iterative denoising process. Post-training quantization (PTQ) presents a solution to accelerate sampling, aibeit at the expense of sample quality, extremely in low-bit settings. Addressing this, our study introduces a unified Quantization Noise Correction Scheme (QNCD), aimed at minishing quantization noise throughout the sampling process. We identify two primary quantization challenges: intra and inter quantization noise. Intra quantization noise, mainly exacerbated by embeddings in the resblock module, extends activation quantization ranges, increasing disturbances in each single denosing step. Besides, inter quantization noise stems from cumulative quantization deviations across the entire denoising process, altering data distributions step-by-step. QNCD combats these through embedding-derived feature smoothing for eliminating intra quantization noise and an effective runtime noise estimatiation module for dynamicly filtering inter quantization noise. Extensive experiments demonstrate that our method outperforms previous quantization methods for diffusion models, achieving lossless results in W4A8 and W8A8 quantization settings on ImageNet (LDM-4). Code is available at: https://github.com/huanpengchu/QNCD"
                },
                {
                    "title": "TextCraftor: Your Text Encoder can be Image Quality Controller",
                    "abstract": "Diffusion-based text-to-image generative models, e.g., Stable Diffusion, have revolutionized the field of content generation, enabling significant advancements in areas like image editing and video synthesis. Despite their formidable capabilities, these models are not without their limitations. It is still challenging to synthesize an image that aligns well with the input text, and multiple runs with carefully crafted prompts are required to achieve satisfactory results. To mitigate these limitations, numerous studies have endeavored to fine-tune the pre-trained diffusion models, i.e., UNet, utilizing various technologies. Yet, amidst these efforts, a pivotal question of text-to-image diffusion model training has remained largely unexplored: Is it possible and feasible to fine-tune the text encoder to improve the performance of text-to-image diffusion models? Our findings reveal that, instead of replacing the CLIP text encoder used in Stable Diffusion with other large language models, we can enhance it through our proposed fine-tuning approach, TextCraftor, leading to substantial improvements in quantitative benchmarks and human assessments. Interestingly, our technique also empowers controllable image generation through the interpolation of different text encoders fine-tuned with various rewards. We also demonstrate that TextCraftor is orthogonal to UNet finetuning, and can be combined to further improve generative quality."
                },
                {
                    "title": "The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits",
                    "abstract": "Recent research, such as BitNet, is paving the way for a new era of 1-bit Large Language Models (LLMs). In this work, we introduce a 1-bit LLM variant, namely BitNet b1.58, in which every single parameter (or weight) of the LLM is ternary {-1, 0, 1}. It matches the full-precision (i.e., FP16 or BF16) Transformer LLM with the same model size and training tokens in terms of both perplexity and end-task performance, while being significantly more cost-effective in terms of latency, memory, throughput, and energy consumption. More profoundly, the 1.58-bit LLM defines a new scaling law and recipe for training new generations of LLMs that are both high-performance and cost-effective. Furthermore, it enables a new computation paradigm and opens the door for designing specific hardware optimized for 1-bit LLMs."
                },
                {
                    "title": "Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video Synthesis",
                    "abstract": "Contemporary models for generating images show remarkable quality and versatility. Swayed by these advantages, the research community repurposes them to generate videos. Since video content is highly redundant, we argue that naively bringing advances of image models to the video generation domain reduces motion fidelity, visual quality and impairs scalability. In this work, we build Snap Video, a video-first model that systematically addresses these challenges. To do that, we first extend the EDM framework to take into account spatially and temporally redundant pixels and naturally support video generation. Second, we show that a U-Net\u2014a workhorse behind image generation\u2014scales poorly when generating videos, requiring significant computational overhead. Hence, we propose a new transformer-based architecture that trains 3.31 times faster than U-Nets (and is \u223c4.5 faster at inference). This allows us to efficiently train a text-to-video model with billions of parameters for the first time, reach state-of-the-art results on a number of benchmarks, and generate videos with substantially higher quality, temporal consistency, and motion complexity. The user studies showed that our model was favored by a large margin over the most recent methods."
                },
                {
                    "title": "QuEST: Low-bit Diffusion Model Quantization via Efficient Selective Finetuning",
                    "abstract": "The practical deployment of diffusion models still suffers from the high memory and time overhead. While quantization paves a way for compression and acceleration, existing methods unfortunately fail when the models are quantized to low-bits. In this paper, we empirically unravel three properties in quantized diffusion models that compromise the efficacy of current methods: imbalanced activation distributions, imprecise temporal information, and vulnerability to perturbations of specific modules. To alleviate the intensified low-bit quantization difficulty stemming from the distribution imbalance, we propose finetuning the quantized model to better adapt to the activation distribution. Building on this idea, we identify two critical types of quantized layers: those holding vital temporal information and those sensitive to reduced bit-width, and finetune them to mitigate performance degradation with efficiency. We empirically verify that our approach modifies the activation distribution and provides meaningful temporal information, facilitating easier and more accurate quantization. Our method is evaluated over three high-resolution image generation tasks and achieves state-of-the-art performance under various bit-width settings, as well as being the first method to generate readable images on full 4-bit (i.e. W4A4) Stable Diffusion. Code is available \\href{https://github.com/hatchetProject/QuEST}{here}."
                },
                {
                    "title": "Lumiere: A Space-Time Diffusion Model for Video Generation",
                    "abstract": "We introduce Lumiere -- a text-to-video diffusion model designed for synthesizing videos that portray realistic, diverse and coherent motion -- a pivotal challenge in video synthesis. To this end, we introduce a Space-Time U-Net architecture that generates the entire temporal duration of the video at once, through a single pass in the model. This is in contrast to existing video models which synthesize distant keyframes followed by temporal super-resolution -- an approach that inherently makes global temporal consistency difficult to achieve. By deploying both spatial and (importantly) temporal down- and up-sampling and leveraging a pre-trained text-to-image diffusion model, our model learns to directly generate a full-frame-rate, low-resolution video by processing it in multiple space-time scales. We demonstrate state-of-the-art text-to-video generation results, and show that our design easily facilitates a wide range of content creation tasks and video editing applications, including image-to-video, video inpainting, and stylized generation."
                },
                {
                    "title": "Enhanced Distribution Alignment for Post-Training Quantization of Diffusion Models",
                    "abstract": "Diffusion models have achieved great success in image generation tasks through iterative noise estimation. However, the heavy denoising process and complex neural networks hinder their low-latency applications in real-world scenarios. Quantization can effectively reduce model complexity, and post-training quantization (PTQ), which does not require fine-tuning, is highly promising for compressing and accelerating diffusion models. Unfortunately, we find that due to the highly dynamic distribution of activations in different denoising steps, existing PTQ methods for diffusion models suffer from distribution mismatch issues at both calibration sample level and reconstruction output level, which makes the performance far from satisfactory, especially in low-bit cases. In this paper, we propose Enhanced Distribution Alignment for Post-Training Quantization of Diffusion Models (EDA-DM) to address the above issues. Specifically, at the calibration sample level, we select calibration samples based on the density and variety in the latent space, thus facilitating the alignment of their distribution with the overall samples; and at the reconstruction output level, we modify the loss of block reconstruction with the losses of layers, aligning the outputs of quantized model and full-precision model at different network granularity. Extensive experiments demonstrate that EDA-DM significantly outperforms the existing PTQ methods across various models (DDIM, LDM-4, LDM-8, Stable-Diffusion) and different datasets (CIFAR-10, LSUN-Bedroom, LSUN-Church, ImageNet, MS-COCO)."
                },
                {
                    "title": "Photorealistic Video Generation with Diffusion Models",
                    "abstract": "We present W.A.L.T, a transformer-based approach for photorealistic video generation via diffusion modeling. Our approach has two key design decisions. First, we use a causal encoder to jointly compress images and videos within a unified latent space, enabling training and generation across modalities. Second, for memory and training efficiency, we use a window attention architecture tailored for joint spatial and spatiotemporal generative modeling. Taken together these design decisions enable us to achieve state-of-the-art performance on established video (UCF-101 and Kinetics-600) and image (ImageNet) generation benchmarks without using classifier free guidance. Finally, we also train a cascade of three models for the task of text-to-video generation consisting of a base latent video diffusion model, and two video super-resolution diffusion models to generate videos of $512 \\times 896$ resolution at $8$ frames per second."
                },
                {
                    "title": "Efficient Quantization Strategies for Latent Diffusion Models",
                    "abstract": "Latent Diffusion Models (LDMs) capture the dynamic evolution of latent variables over time, blending patterns and multimodality in a generative system. Despite the proficiency of LDM in various applications, such as text-to-image generation, facilitated by robust text encoders and a variational autoencoder, the critical need to deploy large generative models on edge devices compels a search for more compact yet effective alternatives. Post Training Quantization (PTQ), a method to compress the operational size of deep learning models, encounters challenges when applied to LDM due to temporal and structural complexities. This study proposes a quantization strategy that efficiently quantize LDMs, leveraging Signal-to-Quantization-Noise Ratio (SQNR) as a pivotal metric for evaluation. By treating the quantization discrepancy as relative noise and identifying sensitive part(s) of a model, we propose an efficient quantization approach encompassing both global and local strategies. The global quantization process mitigates relative quantization noise by initiating higher-precision quantization on sensitive blocks, while local treatments address specific challenges in quantization-sensitive and time-sensitive modules. The outcomes of our experiments reveal that the implementation of both global and local treatments yields a highly efficient and effective Post Training Quantization (PTQ) of LDMs."
                },
                {
                    "title": "TFMQ-DM: Temporal Feature Maintenance Quantization for Diffusion Models",
                    "abstract": "The Diffusion model, a prevalent framework for image generation, encounters significant challenges in terms of broad applicability due to its extended inference times and substantial memory requirements. Efficient Post-training Quantization (PTQ) is pivotal for addressing these issues in traditional models. Different from traditional models, diffusion models heavily depend on the time-step t to achieve satisfactory multi-round denoising. Usually, t from the finite set {1, \u2026, T} is encoded to a temporal feature by a few modules totally irrespective of the sampling data. However, existing PTQ methods do not optimize these modules separately. They adopt inappropriate reconstruction targets and complex calibration methods, resulting in a severe disturbance of the temporal feature and denoising trajectory, as well as a low compression efficiency. To solve these, we propose a Temporal Feature Maintenance Quantization (TFMQ) framework building upon a Temporal Information Block which is just related to the time-step t and unrelated to the sampling data. Powered by the pioneering block design, we devise temporal information aware reconstruction (TIAR) and finite set calibration (FSC) to align the full-precision temporal features in a limited time. Equipped with the framework, we can maintain the most temporal information and ensure the end-to-end generation quality. Extensive experiments on various datasets and diffusion models prove our state-of-the-art results. Remarkably, our quantization approach, for the first time, achieves model performance nearly on par with the full-precision model under 4-bit weight quantization. Additionally, our method incurs almost no extra computational cost and accelerates quan-tization time by 2.0\u00d7 on LSUN-Bedrooms 256 \u00d7 256 compared to previous works. Our code is publicly available at https://github.com/ModelTC/TFMQ-DM."
                },
                {
                    "title": "Emu Video: Factorizing Text-to-Video Generation by Explicit Image Conditioning",
                    "abstract": "We present Emu Video, a text-to-video generation model that factorizes the generation into two steps: first generating an image conditioned on the text, and then generating a video conditioned on the text and the generated image. We identify critical design decisions--adjusted noise schedules for diffusion, and multi-stage training that enable us to directly generate high quality and high resolution videos, without requiring a deep cascade of models as in prior work. In human evaluations, our generated videos are strongly preferred in quality compared to all prior work--81% vs. Google's Imagen Video, 90% vs. Nvidia's PYOCO, and 96% vs. Meta's Make-A-Video. Our model outperforms commercial solutions such as RunwayML's Gen2 and Pika Labs. Finally, our factorizing approach naturally lends itself to animating images based on a user's text prompt, where our generations are preferred 96% over prior work."
                },
                {
                    "title": "Post-training Quantization with Progressive Calibration and Activation Relaxing for Text-to-Image Diffusion Models",
                    "abstract": "High computational overhead is a troublesome problem for diffusion models. Recent studies have leveraged post-training quantization (PTQ) to compress diffusion models. However, most of them only focus on unconditional models, leaving the quantization of widely-used pretrained text-to-image models, e.g., Stable Diffusion, largely unexplored. In this paper, we propose a novel post-training quantization method PCR (Progressive Calibration and Relaxing) for text-to-image diffusion models, which consists of a progressive calibration strategy that considers the accumulated quantization error across timesteps, and an activation relaxing strategy that improves the performance with negligible cost. Additionally, we demonstrate the previous metrics for text-to-image diffusion model quantization are not accurate due to the distribution gap. To tackle the problem, we propose a novel QDiffBench benchmark, which utilizes data in the same domain for more accurate evaluation. Besides, QDiffBench also considers the generalization performance of the quantized model outside the calibration dataset. Extensive experiments on Stable Diffusion and Stable Diffusion XL demonstrate the superiority of our method and benchmark. Moreover, we are the first to achieve quantization for Stable Diffusion XL while maintaining the performance."
                },
                {
                    "title": "Effective Quantization for Diffusion Models on CPUs",
                    "abstract": "Diffusion models have gained popularity for generating images from textual descriptions. Nonetheless, the substantial need for computational resources continues to present a noteworthy challenge, contributing to time-consuming processes. Quantization, a technique employed to compress deep learning models for enhanced efficiency, presents challenges when applied to diffusion models. These models are notably more sensitive to quantization compared to other model types, potentially resulting in a degradation of image quality. In this paper, we introduce a novel approach to quantize the diffusion models by leveraging both quantization-aware training and distillation. Our results show the quantized models can maintain the high image quality while demonstrating the inference efficiency on CPUs. The code is publicly available at: https://github.com/intel/intel-extension-for-transformers."
                },
                {
                    "title": "GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignment",
                    "abstract": "Recent breakthroughs in diffusion models, multimodal pretraining, and efficient finetuning have led to an explosion of text-to-image generative models. Given human evaluation is expensive and difficult to scale, automated methods are critical for evaluating the increasingly large number of new models. However, most current automated evaluation metrics like FID or CLIPScore only offer a holistic measure of image quality or image-text alignment, and are unsuited for fine-grained or instance-level analysis. In this paper, we introduce GenEval, an object-focused framework to evaluate compositional image properties such as object co-occurrence, position, count, and color. We show that current object detection models can be leveraged to evaluate text-to-image models on a variety of generation tasks with strong human agreement, and that other discriminative vision models can be linked to this pipeline to further verify properties like object color. We then evaluate several open-source text-to-image models and analyze their relative generative capabilities on our benchmark. We find that recent models demonstrate significant improvement on these tasks, though they are still lacking in complex capabilities such as spatial relations and attribute binding. Finally, we demonstrate how GenEval might be used to help discover existing failure modes, in order to inform development of the next generation of text-to-image models. Our code to run the GenEval framework is publicly available at https://github.com/djghosh13/geneval."
                },
                {
                    "title": "HyperHuman: Hyper-Realistic Human Generation with Latent Structural Diffusion",
                    "abstract": "Despite significant advances in large-scale text-to-image models, achieving hyper-realistic human image generation remains a desirable yet unsolved task. Existing models like Stable Diffusion and DALL-E 2 tend to generate human images with incoherent parts or unnatural poses. To tackle these challenges, our key insight is that human image is inherently structural over multiple granularities, from the coarse-level body skeleton to fine-grained spatial geometry. Therefore, capturing such correlations between the explicit appearance and latent structure in one model is essential to generate coherent and natural human images. To this end, we propose a unified framework, HyperHuman, that generates in-the-wild human images of high realism and diverse layouts. Specifically, 1) we first build a large-scale human-centric dataset, named HumanVerse, which consists of 340M images with comprehensive annotations like human pose, depth, and surface normal. 2) Next, we propose a Latent Structural Diffusion Model that simultaneously denoises the depth and surface normal along with the synthesized RGB image. Our model enforces the joint learning of image appearance, spatial relationship, and geometry in a unified network, where each branch in the model complements to each other with both structural awareness and textural richness. 3) Finally, to further boost the visual quality, we propose a Structure-Guided Refiner to compose the predicted conditions for more detailed generation of higher resolution. Extensive experiments demonstrate that our framework yields the state-of-the-art performance, generating hyper-realistic human images under diverse scenarios. Project Page: https://snap-research.github.io/HyperHuman/"
                },
                {
                    "title": "EfficientDM: Efficient Quantization-Aware Fine-Tuning of Low-Bit Diffusion Models",
                    "abstract": "Diffusion models have demonstrated remarkable capabilities in image synthesis and related generative tasks. Nevertheless, their practicality for real-world applications is constrained by substantial computational costs and latency issues. Quantization is a dominant way to compress and accelerate diffusion models, where post-training quantization (PTQ) and quantization-aware training (QAT) are two main approaches, each bearing its own properties. While PTQ exhibits efficiency in terms of both time and data usage, it may lead to diminished performance in low bit-width. On the other hand, QAT can alleviate performance degradation but comes with substantial demands on computational and data resources. In this paper, we introduce a data-free and parameter-efficient fine-tuning framework for low-bit diffusion models, dubbed EfficientDM, to achieve QAT-level performance with PTQ-like efficiency. Specifically, we propose a quantization-aware variant of the low-rank adapter (QALoRA) that can be merged with model weights and jointly quantized to low bit-width. The fine-tuning process distills the denoising capabilities of the full-precision model into its quantized counterpart, eliminating the requirement for training data. We also introduce scale-aware optimization and temporal learned step-size quantization to further enhance performance. Extensive experimental results demonstrate that our method significantly outperforms previous PTQ-based diffusion models while maintaining similar time and data efficiency. Specifically, there is only a 0.05 sFID increase when quantizing both weights and activations of LDM-4 to 4-bit on ImageNet 256x256. Compared to QAT-based methods, our EfficientDM also boasts a 16.2x faster quantization speed with comparable generation quality. Code is available at \\href{https://github.com/ThisisBillhe/EfficientDM}{this hrl}."
                },
                {
                    "title": "MVDream: Multi-view Diffusion for 3D Generation",
                    "abstract": "We introduce MVDream, a diffusion model that is able to generate consistent multi-view images from a given text prompt. Learning from both 2D and 3D data, a multi-view diffusion model can achieve the generalizability of 2D diffusion models and the consistency of 3D renderings. We demonstrate that such a multi-view diffusion model is implicitly a generalizable 3D prior agnostic to 3D representations. It can be applied to 3D generation via Score Distillation Sampling, significantly enhancing the consistency and stability of existing 2D-lifting methods. It can also learn new concepts from a few 2D examples, akin to DreamBooth, but for 3D generation."
                },
                {
                    "title": "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis",
                    "abstract": "We present SDXL, a latent diffusion model for text-to-image synthesis. Compared to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet backbone: The increase of model parameters is mainly due to more attention blocks and a larger cross-attention context as SDXL uses a second text encoder. We design multiple novel conditioning schemes and train SDXL on multiple aspect ratios. We also introduce a refinement model which is used to improve the visual fidelity of samples generated by SDXL using a post-hoc image-to-image technique. We demonstrate that SDXL shows drastically improved performance compared the previous versions of Stable Diffusion and achieves results competitive with those of black-box state-of-the-art image generators. In the spirit of promoting open research and fostering transparency in large model training and evaluation, we provide access to code and model weights at https://github.com/Stability-AI/generative-models"
                },
                {
                    "title": "Magic123: One Image to High-Quality 3D Object Generation Using Both 2D and 3D Diffusion Priors",
                    "abstract": "We present Magic123, a two-stage coarse-to-fine approach for high-quality, textured 3D meshes generation from a single unposed image in the wild using both2D and 3D priors. In the first stage, we optimize a neural radiance field to produce a coarse geometry. In the second stage, we adopt a memory-efficient differentiable mesh representation to yield a high-resolution mesh with a visually appealing texture. In both stages, the 3D content is learned through reference view supervision and novel views guided by a combination of 2D and 3D diffusion priors. We introduce a single trade-off parameter between the 2D and 3D priors to control exploration (more imaginative) and exploitation (more precise) of the generated geometry. Additionally, we employ textual inversion and monocular depth regularization to encourage consistent appearances across views and to prevent degenerate solutions, respectively. Magic123 demonstrates a significant improvement over previous image-to-3D techniques, as validated through extensive experiments on synthetic benchmarks and diverse real-world images. Our code, models, and generated 3D assets are available at https://github.com/guochengqian/Magic123."
                },
                {
                    "title": "Temporal Dynamic Quantization for Diffusion Models",
                    "abstract": "The diffusion model has gained popularity in vision applications due to its remarkable generative performance and versatility. However, high storage and computation demands, resulting from the model size and iterative generation, hinder its use on mobile devices. Existing quantization techniques struggle to maintain performance even in 8-bit precision due to the diffusion model's unique property of temporal variation in activation. We introduce a novel quantization method that dynamically adjusts the quantization interval based on time step information, significantly improving output quality. Unlike conventional dynamic quantization techniques, our approach has no computational overhead during inference and is compatible with both post-training quantization (PTQ) and quantization-aware training (QAT). Our extensive experiments demonstrate substantial improvements in output quality with the quantized diffusion model across various datasets."
                },
                {
                    "title": "SnapFusion: Text-to-Image Diffusion Model on Mobile Devices within Two Seconds",
                    "abstract": "Text-to-image diffusion models can create stunning images from natural language descriptions that rival the work of professional artists and photographers. However, these models are large, with complex network architectures and tens of denoising iterations, making them computationally expensive and slow to run. As a result, high-end GPUs and cloud-based inference are required to run diffusion models at scale. This is costly and has privacy implications, especially when user data is sent to a third party. To overcome these challenges, we present a generic approach that, for the first time, unlocks running text-to-image diffusion models on mobile devices in less than $2$ seconds. We achieve so by introducing efficient network architecture and improving step distillation. Specifically, we propose an efficient UNet by identifying the redundancy of the original model and reducing the computation of the image decoder via data distillation. Further, we enhance the step distillation by exploring training strategies and introducing regularization from classifier-free guidance. Our extensive experiments on MS-COCO show that our model with $8$ denoising steps achieves better FID and CLIP scores than Stable Diffusion v$1.5$ with $50$ steps. Our work democratizes content creation by bringing powerful text-to-image diffusion models to the hands of users."
                },
                {
                    "title": "LLM-QAT: Data-Free Quantization Aware Training for Large Language Models",
                    "abstract": "Several post-training quantization methods have been applied to large language models (LLMs), and have been shown to perform well down to 8-bits. We find that these methods break down at lower bit precision, and investigate quantization aware training for LLMs (LLM-QAT) to push quantization levels even further. We propose a data-free distillation method that leverages generations produced by the pre-trained model, which better preserves the original output distribution and allows quantizing any generative model independent of its training data, similar to post-training quantization methods. In addition to quantizing weights and activations, we also quantize the KV cache, which is critical for increasing throughput and support long sequence dependencies at current model sizes. We experiment with LLaMA models of sizes 7B, 13B, and 30B, at quantization levels down to 4-bits. We observe large improvements over training-free methods, especially in the low-bit settings."
                },
                {
                    "title": "PTQD: Accurate Post-Training Quantization for Diffusion Models",
                    "abstract": "Diffusion models have recently dominated image synthesis tasks. However, the iterative denoising process is expensive in computations at inference time, making diffusion models less practical for low-latency and scalable real-world applications. Post-training quantization (PTQ) of diffusion models can significantly reduce the model size and accelerate the sampling process without re-training. Nonetheless, applying existing PTQ methods directly to low-bit diffusion models can significantly impair the quality of generated samples. Specifically, for each denoising step, quantization noise leads to deviations in the estimated mean and mismatches with the predetermined variance schedule. As the sampling process proceeds, the quantization noise may accumulate, resulting in a low signal-to-noise ratio (SNR) during the later denoising steps. To address these challenges, we propose a unified formulation for the quantization noise and diffusion perturbed noise in the quantized denoising process. Specifically, we first disentangle the quantization noise into its correlated and residual uncorrelated parts regarding its full-precision counterpart. The correlated part can be easily corrected by estimating the correlation coefficient. For the uncorrelated part, we subtract the bias from the quantized results to correct the mean deviation and calibrate the denoising variance schedule to absorb the excess variance resulting from quantization. Moreover, we introduce a mixed-precision scheme for selecting the optimal bitwidth for each denoising step. Extensive experiments demonstrate that our method outperforms previous post-training quantized diffusion models, with only a 0.06 increase in FID score compared to full-precision LDM-4 on ImageNet 256x256, while saving 19.9x bit operations. Code is available at https://github.com/ziplab/PTQD."
                },
                {
                    "title": "Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation",
                    "abstract": "The ability to collect a large dataset of human preferences from text-to-image users is usually limited to companies, making such datasets inaccessible to the public. To address this issue, we create a web app that enables text-to-image users to generate images and specify their preferences. Using this web app we build Pick-a-Pic, a large, open dataset of text-to-image prompts and real users' preferences over generated images. We leverage this dataset to train a CLIP-based scoring function, PickScore, which exhibits superhuman performance on the task of predicting human preferences. Then, we test PickScore's ability to perform model evaluation and observe that it correlates better with human rankings than other automatic evaluation metrics. Therefore, we recommend using PickScore for evaluating future text-to-image generation models, and using Pick-a-Pic prompts as a more relevant dataset than MS-COCO. Finally, we demonstrate how PickScore can enhance existing text-to-image models via ranking."
                },
                {
                    "title": "Align Your Latents: High-Resolution Video Synthesis with Latent Diffusion Models",
                    "abstract": "Latent Diffusion Models (LDMs) enable high-quality image synthesis while avoiding excessive compute demands by training a diffusion model in a compressed lower-dimensional latent space. Here, we apply the LDM paradigm to high-resolution video generation, a particularly resource-intensive task. We first pre-train an LDM on images only; then, we turn the image generator into a video generator by introducing a temporal dimension to the latent space diffusion model and finetuning on encoded image sequences, i.e., videos. Similarly, we temporally align diffusion model upsamplers, turning them into temporally consistent video super resolution models. We focus on two relevant real-world applications: Simulation of in-the-wild driving data and creative content creation with text-to-video modeling. In particular, we validate our Video LDM on real driving videos of resolution $512 \\times 1024$, achieving state-of-the-art performance. Furthermore, our approach can easily leverage off-the-shelf pretrained image LDMs, as we only need to train a temporal alignment model in that case. Doing so, we turn the publicly available, state-of-the-art text-to-image LDM Stable Diffusion into an efficient and expressive text-to-video model with resolution up to $1280 \\times 2048$. We show that the temporal layers trained in this way generalize to different finetuned text-to-image LDMs. Utilizing this property, we show the first results for personalized text-to-video generation, opening exciting directions for future content creation. Project page: https://nv-tlabs.github.io/VideoLDM/"
                },
                {
                    "title": "TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering",
                    "abstract": "Despite thousands of researchers, engineers, and artists actively working on improving text-to-image generation models, systems often fail to produce images that accurately align with the text inputs. We introduce TIFA (Text-to-Image Faithfulness evaluation with question Answering), an automatic evaluation metric that measures the faithfulness of a generated image to its text input via visual question answering (VQA). Specifically, given a text input, we automatically generate several question-answer pairs using a language model. We calculate image faithfulness by checking whether existing VQA models can answer these questions using the generated image. TIFA is a reference-free metric that allows for fine-grained and interpretable evaluations of generated images. TIFA also has better correlations with human judgments than existing metrics. Based on this approach, we introduce TIFA v1.0, a benchmark consisting of 4K diverse text inputs and 25K questions across 12 categories (object, counting, etc.). We present a comprehensive evaluation of existing text-to-image models using TIFA v1.0 and highlight the limitations and challenges of current models. For instance, we find that current text-to-image models, despite doing well on color and material, still struggle in counting, spatial relations, and composing multiple objects. We hope our benchmark will help carefully measure the research progress in text-to-image synthesis and provide valuable insights for further research. 1"
                },
                {
                    "title": "Adding Conditional Control to Text-to-Image Diffusion Models",
                    "abstract": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models."
                },
                {
                    "title": "Q-Diffusion: Quantizing Diffusion Models",
                    "abstract": "Diffusion models have achieved great success in image synthesis through iterative noise estimation using deep neural networks. However, the slow inference, high memory consumption, and computation intensity of the noise estimation model hinder the efficient adoption of diffusion models. Although post-training quantization (PTQ) is considered a go-to compression method for other tasks, it does not work out-of-the-box on diffusion models. We propose a novel PTQ method specifically tailored towards the unique multi-timestep pipeline and model architecture of the diffusion models, which compresses the noise estimation network to accelerate the generation process. We identify the key difficulty of diffusion model quantization as the changing output distributions of noise estimation networks over multiple time steps and the bimodal activation distribution of the shortcut layers within the noise estimation network. We tackle these challenges with timestep-aware calibration and split shortcut quantization in this work. Experimental results show that our proposed method is able to quantize full-precision unconditional diffusion models into 4-bit while maintaining comparable performance (small FID change of at most 2.34 compared to >100 for traditional PTQ) in a training-free manner. Our approach can also be applied to text-guided image generation, where we can run stable diffusion in 4-bit weights with high generation quality for the first time."
                },
                {
                    "title": "SINE: SINgle Image Editing with Text-to-Image Diffusion Models",
                    "abstract": "Recent works on diffusion models have demonstrated a strong capability for conditioning image generation, e.g., text-guided image synthesis. Such success inspires many efforts trying to use large-scale pre-trained diffusion models for tackling a challenging problem-real image editing. Works conducted in this area learn a unique textual token corresponding to several images containing the same object. However, under many circumstances, only one image is available, such as the painting of the Girl with a Pearl Earring. Using existing works on fine-tuning the pre-trained diffusion models with a single image causes severe overfitting issues. The information leakage from the pre-trained diffusion models makes editing can not keep the same content as the given image while creating new features depicted by the language guidance. This work aims to address the problem of single-image editing. We propose a novel model-based guidance built upon the classifier-free guidance so that the knowledge from the model trained on a single image can be distilled into the pre-trained diffusion model, enabling content creation even with one given image. Additionally, we propose a patch-based fine-tuning that can effectively help the model generate images of arbitrary resolution. We provide extensive experiments to validate the design choices of our approach and show promising editing capabilities, including changing style, content addition, and object manipulation. Our code is made publicly available here."
                },
                {
                    "title": "Post-Training Quantization on Diffusion Models",
                    "abstract": "Denoising diffusion (score-based) generative models have recently achieved significant accomplishments in generating realistic and diverse data. Unfortunately, the generation process of current denoising diffusion models is notoriously slow due to the lengthy iterative noise estimations, which rely on cumbersome neural networks. It prevents the diffusion models from being widely deployed, especially on edge devices. Previous works accelerate the generation process of diffusion model (DM) via finding shorter yet effective sampling trajectories. However, they overlook the cost of noise estimation with a heavy network in every iteration. In this work, we accelerate generation from the perspective of compressing the noise estimation network. Due to the difficulty of retraining DMs, we exclude mainstream training-aware compression paradigms and introduce post-training quantization (PTQ) into DM acceleration. However, the output distributions of noise estimation networks change with time-step, making previous PTQ methods fail in DMs since they are designed for single-time step scenarios. To devise a DM-specific PTQ method, we explore PTQ on DM in three aspects: quantized operations, calibration dataset, and calibration metric. We summarize and use several observations derived from all-inclusive investigations to formulate our method, which especially targets the unique multi-time-step structure of DMs. Experimentally, our method can directly quantize full-precision DMs into 8-bit models while maintaining or even improving their performance in a training-free manner. Importantly, our method can serve as a plug-and-play module on other fast-sampling methods, e.g., DDIM [24]. The code is available at https://https://github.com/42Shawn/PTQ4DM."
                },
                {
                    "title": "Magic3D: High-Resolution Text-to-3D Content Creation",
                    "abstract": "DreamFusion [31] has recently demonstrated the utility of a pretrained text-to-image diffusion model to optimize Neural Radiance Fields (NeRF) [23], achieving remarkable text-to-3D synthesis results. However, the method has two inherent limitations: (a) extremely slow optimization of NeRF and (b) low-resolution image space supervision on NeRF, leading to low-quality 3D models with a long processing time. In this paper, we address these limitations by utilizing a two-stage optimization framework. First, we obtain a coarse model using a low-resolution diffusion prior and accelerate with a sparse 3D hash grid structure. Using the coarse representation as the initialization, we further optimize a textured 3D mesh model with an efficient differentiable renderer interacting with a high-resolution latent diffusion model. Our method, dubbed Magic3D, can create high quality 3D mesh models in 40 minutes, which is 2\u00d7 faster than DreamFusion (reportedly taking 1.5 hours on average), while also achieving higher resolution. User studies show 61.7% raters to prefer our approach over DreamFusion. Together with the image-conditioned generation capabilities, we provide users with new ways to control 3D synthesis, opening up new avenues to various creative applications."
                },
                {
                    "title": "LION: Latent Point Diffusion Models for 3D Shape Generation",
                    "abstract": "Denoising diffusion models (DDMs) have shown promising results in 3D point cloud synthesis. To advance 3D DDMs and make them useful for digital artists, we require (i) high generation quality, (ii) flexibility for manipulation and applications such as conditional synthesis and shape interpolation, and (iii) the ability to output smooth surfaces or meshes. To this end, we introduce the hierarchical Latent Point Diffusion Model (LION) for 3D shape generation. LION is set up as a variational autoencoder (VAE) with a hierarchical latent space that combines a global shape latent representation with a point-structured latent space. For generation, we train two hierarchical DDMs in these latent spaces. The hierarchical VAE approach boosts performance compared to DDMs that operate on point clouds directly, while the point-structured latents are still ideally suited for DDM-based modeling. Experimentally, LION achieves state-of-the-art generation performance on multiple ShapeNet benchmarks. Furthermore, our VAE framework allows us to easily use LION for different relevant tasks: LION excels at multimodal shape denoising and voxel-conditioned synthesis, and it can be adapted for text- and image-driven 3D generation. We also demonstrate shape autoencoding and latent shape interpolation, and we augment LION with modern surface reconstruction techniques to generate smooth 3D meshes. We hope that LION provides a powerful tool for artists working with 3D shapes due to its high-quality generation, flexibility, and surface reconstruction. Project page and code: https://nv-tlabs.github.io/LION."
                },
                {
                    "title": "Imagen Video: High Definition Video Generation with Diffusion Models",
                    "abstract": "We present Imagen Video, a text-conditional video generation system based on a cascade of video diffusion models. Given a text prompt, Imagen Video generates high definition videos using a base video generation model and a sequence of interleaved spatial and temporal video super-resolution models. We describe how we scale up the system as a high definition text-to-video model including design decisions such as the choice of fully-convolutional temporal and spatial super-resolution models at certain resolutions, and the choice of the v-parameterization of diffusion models. In addition, we confirm and transfer findings from previous work on diffusion-based image generation to the video generation setting. Finally, we apply progressive distillation to our video models with classifier-free guidance for fast, high quality sampling. We find Imagen Video not only capable of generating videos of high fidelity, but also having a high degree of controllability and world knowledge, including the ability to generate diverse videos and text animations in various artistic styles and with 3D object understanding. See https://imagen.research.google/video/ for samples."
                },
                {
                    "title": "Make-A-Video: Text-to-Video Generation without Text-Video Data",
                    "abstract": "We propose Make-A-Video -- an approach for directly translating the tremendous recent progress in Text-to-Image (T2I) generation to Text-to-Video (T2V). Our intuition is simple: learn what the world looks like and how it is described from paired text-image data, and learn how the world moves from unsupervised video footage. Make-A-Video has three advantages: (1) it accelerates training of the T2V model (it does not need to learn visual and multimodal representations from scratch), (2) it does not require paired text-video data, and (3) the generated videos inherit the vastness (diversity in aesthetic, fantastical depictions, etc.) of today's image generation models. We design a simple yet effective way to build on T2I models with novel and effective spatial-temporal modules. First, we decompose the full temporal U-Net and attention tensors and approximate them in space and time. Second, we design a spatial temporal pipeline to generate high resolution and frame rate videos with a video decoder, interpolation model and two super resolution models that can enable various applications besides T2V. In all aspects, spatial and temporal resolution, faithfulness to text, and quality, Make-A-Video sets the new state-of-the-art in text-to-video generation, as determined by both qualitative and quantitative measures."
                },
                {
                    "title": "DreamFusion: Text-to-3D using 2D Diffusion",
                    "abstract": "Recent breakthroughs in text-to-image synthesis have been driven by diffusion models trained on billions of image-text pairs. Adapting this approach to 3D synthesis would require large-scale datasets of labeled 3D data and efficient architectures for denoising 3D data, neither of which currently exist. In this work, we circumvent these limitations by using a pretrained 2D text-to-image diffusion model to perform text-to-3D synthesis. We introduce a loss based on probability density distillation that enables the use of a 2D diffusion model as a prior for optimization of a parametric image generator. Using this loss in a DeepDream-like procedure, we optimize a randomly-initialized 3D model (a Neural Radiance Field, or NeRF) via gradient descent such that its 2D renderings from random angles achieve a low loss. The resulting 3D model of the given text can be viewed from any angle, relit by arbitrary illumination, or composited into any 3D environment. Our approach requires no 3D training data and no modifications to the image diffusion model, demonstrating the effectiveness of pretrained image diffusion models as priors."
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "Scaling Autoregressive Models for Content-Rich Text-to-Image Generation",
                    "abstract": "We present the Pathways Autoregressive Text-to-Image (Parti) model, which generates high-fidelity photorealistic images and supports content-rich synthesis involving complex compositions and world knowledge. Parti treats text-to-image generation as a sequence-to-sequence modeling problem, akin to machine translation, with sequences of image tokens as the target outputs rather than text tokens in another language. This strategy can naturally tap into the rich body of prior work on large language models, which have seen continued advances in capabilities and performance through scaling data and model sizes. Our approach is simple: First, Parti uses a Transformer-based image tokenizer, ViT-VQGAN, to encode images as sequences of discrete tokens. Second, we achieve consistent quality improvements by scaling the encoder-decoder Transformer model up to 20B parameters, with a new state-of-the-art zero-shot FID score of 7.23 and finetuned FID score of 3.22 on MS-COCO. Our detailed analysis on Localized Narratives as well as PartiPrompts (P2), a new holistic benchmark of over 1600 English prompts, demonstrate the effectiveness of Parti across a wide variety of categories and difficulty aspects. We also explore and highlight limitations of our models in order to define and exemplify key areas of focus for further improvements. See https://parti.research.google/ for high-resolution images."
                },
                {
                    "title": "DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps",
                    "abstract": "Diffusion probabilistic models (DPMs) are emerging powerful generative models. Despite their high-quality generation performance, DPMs still suffer from their slow sampling as they generally need hundreds or thousands of sequential function evaluations (steps) of large neural networks to draw a sample. Sampling from DPMs can be viewed alternatively as solving the corresponding diffusion ordinary differential equations (ODEs). In this work, we propose an exact formulation of the solution of diffusion ODEs. The formulation analytically computes the linear part of the solution, rather than leaving all terms to black-box ODE solvers as adopted in previous works. By applying change-of-variable, the solution can be equivalently simplified to an exponentially weighted integral of the neural network. Based on our formulation, we propose DPM-Solver, a fast dedicated high-order solver for diffusion ODEs with the convergence order guarantee. DPM-Solver is suitable for both discrete-time and continuous-time DPMs without any further training. Experimental results show that DPM-Solver can generate high-quality samples in only 10 to 20 function evaluations on various datasets. We achieve 4.70 FID in 10 function evaluations and 2.87 FID in 20 function evaluations on the CIFAR10 dataset, and a $4\\sim 16\\times$ speedup compared with previous state-of-the-art training-free samplers on various datasets."
                },
                {
                    "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                    "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                },
                {
                    "title": "Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding",
                    "abstract": "We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding. Imagen builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation. Our key discovery is that generic large language models (e.g. T5), pretrained on text-only corpora, are surprisingly effective at encoding text for image synthesis: increasing the size of the language model in Imagen boosts both sample fidelity and image-text alignment much more than increasing the size of the image diffusion model. Imagen achieves a new state-of-the-art FID score of 7.27 on the COCO dataset, without ever training on COCO, and human raters find Imagen samples to be on par with the COCO data itself in image-text alignment. To assess text-to-image models in greater depth, we introduce DrawBench, a comprehensive and challenging benchmark for text-to-image models. With DrawBench, we compare Imagen with recent methods including VQ-GAN+CLIP, Latent Diffusion Models, and DALL-E 2, and find that human raters prefer Imagen over other models in side-by-side comparisons, both in terms of sample quality and image-text alignment. See https://imagen.research.google/ for an overview of the results."
                },
                {
                    "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
                    "abstract": "Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples."
                },
                {
                    "title": "Pseudo Numerical Methods for Diffusion Models on Manifolds",
                    "abstract": "Denoising Diffusion Probabilistic Models (DDPMs) can generate high-quality samples such as image and audio samples. However, DDPMs require hundreds to thousands of iterations to produce final samples. Several prior works have successfully accelerated DDPMs through adjusting the variance schedule (e.g., Improved Denoising Diffusion Probabilistic Models) or the denoising equation (e.g., Denoising Diffusion Implicit Models (DDIMs)). However, these acceleration methods cannot maintain the quality of samples and even introduce new noise at a high speedup rate, which limit their practicability. To accelerate the inference process while keeping the sample quality, we provide a fresh perspective that DDPMs should be treated as solving differential equations on manifolds. Under such a perspective, we propose pseudo numerical methods for diffusion models (PNDMs). Specifically, we figure out how to solve differential equations on manifolds and show that DDIMs are simple cases of pseudo numerical methods. We change several classical numerical methods to corresponding pseudo numerical methods and find that the pseudo linear multi-step method is the best in most situations. According to our experiments, by directly using pre-trained models on Cifar10, CelebA and LSUN, PNDMs can generate higher quality synthetic images with only 50 steps compared with 1000-step DDIMs (20x speedup), significantly outperform DDIMs with 250 steps (by around 0.4 in FID) and have good generalization on different variance schedules. Our implementation is available at https://github.com/luping-liu/PNDM."
                },
                {
                    "title": "F8Net: Fixed-Point 8-bit Only Multiplication for Network Quantization",
                    "abstract": "Neural network quantization is a promising compression technique to reduce memory footprint and save energy consumption, potentially leading to real-time inference. However, there is a performance gap between quantized and full-precision models. To reduce it, existing quantization approaches require high-precision INT32 or full-precision multiplication during inference for scaling or dequantization. This introduces a noticeable cost in terms of memory, speed, and required energy. To tackle these issues, we present F8Net, a novel quantization framework consisting of only fixed-point 8-bit multiplication. To derive our method, we first discuss the advantages of fixed-point multiplication with different formats of fixed-point numbers and study the statistical behavior of the associated fixed-point numbers. Second, based on the statistical and algorithmic analysis, we apply different fixed-point formats for weights and activations of different layers. We introduce a novel algorithm to automatically determine the right format for each layer during training. Third, we analyze a previous quantization algorithm -- parameterized clipping activation (PACT) -- and reformulate it using fixed-point arithmetic. Finally, we unify the recently proposed method for quantization fine-tuning and our fixed-point approach to show the potential of our method. We verify F8Net on ImageNet for MobileNet V1/V2 and ResNet18/50. Our approach achieves comparable and better performance, when compared not only to existing quantization techniques with INT32 multiplication or floating-point arithmetic, but also to the full-precision counterparts, achieving state-of-the-art performance."
                },
                {
                    "title": "DALL-EVAL: Probing the Reasoning Skills and Social Biases of Text-to-Image Generation Models",
                    "abstract": "Recently, DALL-E [45], a multimodal transformer language model, and its variants including diffusion models have shown high-quality text-to-image generation capabilities. However, despite the realistic image generation results, there has not been a detailed analysis of how to evaluate such models. In this work, we investigate the visual reasoning capabilities and social biases of different text-to-image models, covering both multimodal transformer language models and diffusion models. First, we measure three visual reasoning skills: object recognition, object counting, and spatial relation understanding. For this, we propose PaintSkills, a compositional diagnostic evaluation dataset that measures these skills. Despite the high-fidelity image generation capability, a large gap exists between the performance of recent models and the upper bound accuracy in object counting and spatial relation understanding skills. Second, we assess the gender and skin tone biases by measuring the gender/skin tone distribution of generated images across various professions and attributes. We demonstrate that recent text-to-image generation models learn specific biases about gender and skin tone from web image-text pairs. We hope our work will help guide future progress in improving text-to-image generation models on visual reasoning skills and learning socially unbiased representations.1"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models",
                    "abstract": "Diffusion models have recently been shown to generate high-quality synthetic images, especially when paired with a guidance technique to trade off diversity for fidelity. We explore diffusion models for the problem of text-conditional image synthesis and compare two different guidance strategies: CLIP guidance and classifier-free guidance. We find that the latter is preferred by human evaluators for both photorealism and caption similarity, and often produces photorealistic samples. Samples from a 3.5 billion parameter text-conditional diffusion model using classifier-free guidance are favored by human evaluators to those from DALL-E, even when the latter uses expensive CLIP reranking. Additionally, we find that our models can be fine-tuned to perform image inpainting, enabling powerful text-driven image editing. We train a smaller model on a filtered dataset and release the code and weights at https://github.com/openai/glide-text2im."
                },
                {
                    "title": "Masked-attention Mask Transformer for Universal Image Segmentation",
                    "abstract": "Image segmentation groups pixels with different semantics, e.g., category or instance membership. Each choice of semantics defines a task. While only the semantics of each task differ, current research focuses on designing spe-cialized architectures for each task. We present Masked- attention Mask Transformer (Mask2Former), a new archi-tecture capable of addressing any image segmentation task (panoptic, instance or semantic). Its key components in-clude masked attention, which extracts localized features by constraining cross-attention within predicted mask regions. In addition to reducing the research effort by at least three times, it outperforms the best specialized architectures by a significant margin on four popular datasets. Most no-tably, Mask2Former sets a new state-of-the-art for panoptic segmentation (57.8 PQ on COCO), instance segmentation (50.1 AP on COCO) and semantic segmentation (57.7 mIoU onADE20K)."
                },
                {
                    "title": "Palette: Image-to-Image Diffusion Models",
                    "abstract": "This paper develops a unified framework for image-to-image translation based on conditional diffusion models and evaluates this framework on four challenging image-to-image translation tasks, namely colorization, inpainting, uncropping, and JPEG restoration. Our simple implementation of image-to-image diffusion models outperforms strong GAN and regression baselines on all tasks, without task-specific hyper-parameter tuning, architecture customization, or any auxiliary loss or sophisticated new techniques needed. We uncover the impact of an L2 vs. L1 loss in the denoising diffusion objective on sample diversity, and demonstrate the importance of self-attention in the neural architecture through empirical studies. Importantly, we advocate a unified evaluation protocol based on ImageNet, with human evaluation and sample quality scores (FID, Inception Score, Classification Accuracy of a pre-trained ResNet-50, and Perceptual Distance against original images). We expect this standardized evaluation protocol to play a role in advancing image-to-image translation research. Finally, we show that a generalist, multi-task diffusion model performs as well or better than task-specific specialist counterparts. Check out https://diffusion-palette.github.io/ for an overview of the results and code."
                },
                {
                    "title": "Optimal Quantization using Scaled Codebook",
                    "abstract": "We study the problem of quantizing N sorted, scalar datapoints with a fixed codebook containing K entries that are allowed to be rescaled. The problem is defined as finding the optimal scaling factor \u03b1 and the datapoint assignments into the \u03b1-scaled codebook to minimize the squared error between original and quantized points. Previously, the globally optimal algorithms for this problem were derived only for certain codebooks (binary and ternary) or under the assumption of certain distributions (Gaussian, Laplacian). By studying the properties of the optimal quantizer, we derive an $\\mathcal{O}\\left( {NK\\log K} \\right)$ algorithm that is guaranteed to find the optimal quantization parameters for any fixed codebook regardless of data distribution. We apply our algorithm to synthetic and real-world neural network quantization problems and demonstrate the effectiveness of our approach."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "A Survey of Quantization Methods for Efficient Neural Network Inference",
                    "abstract": "As soon as abstract mathematical computations were adapted to computation on digital computers, the problem of efficient representation, manipulation, and communication of the numerical values in those computations arose. Strongly related to the problem of numerical representation is the problem of quantization: in what manner should a set of continuous real-valued numbers be distributed over a fixed discrete set of numbers to minimize the number of bits required and also to maximize the accuracy of the attendant computations? This perennial problem of quantization is particularly relevant whenever memory and/or computational resources are severely restricted, and it has come to the forefront in recent years due to the remarkable performance of Neural Network models in computer vision, natural language processing, and related areas. Moving from floating-point representations to low-precision fixed integer values represented in four bits or less holds the potential to reduce the memory footprint and latency by a factor of 16x; and, in fact, reductions of 4x to 8x are often realized in practice in these applications. Thus, it is not surprising that quantization has emerged recently as an important and very active sub-area of research in the efficient implementation of computations associated with Neural Networks. In this article, we survey approaches to the problem of quantizing the numerical values in deep Neural Network computations, covering the advantages/disadvantages of current methods. With this survey and its organization, we hope to have presented a useful snapshot of the current research in quantization for Neural Networks and to have given an intelligent organization to ease the evaluation of future research in this area."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Zero-Shot Text-to-Image Generation",
                    "abstract": "Text-to-image generation has traditionally focused on finding better modeling assumptions for training on a fixed dataset. These assumptions might involve complex architectures, auxiliary losses, or side information such as object part labels or segmentation masks supplied during training. We describe a simple approach for this task based on a transformer that autoregressively models the text and image tokens as a single stream of data. With sufficient data and scale, our approach is competitive with previous domain-specific models when evaluated in a zero-shot fashion."
                },
                {
                    "title": "BRECQ: Pushing the Limit of Post-Training Quantization by Block Reconstruction",
                    "abstract": "We study the challenging task of neural network quantization without end-to-end retraining, called Post-training Quantization (PTQ). PTQ usually requires a small subset of training data but produces less powerful quantized models than Quantization-Aware Training (QAT). In this work, we propose a novel PTQ framework, dubbed BRECQ, which pushes the limits of bitwidth in PTQ down to INT2 for the first time. BRECQ leverages the basic building blocks in neural networks and reconstructs them one-by-one. In a comprehensive theoretical study of the second-order error, we show that BRECQ achieves a good balance between cross-layer dependency and generalization error. To further employ the power of quantization, the mixed precision technique is incorporated in our framework by approximating the inter-layer and intra-layer sensitivity. Extensive experiments on various handcrafted and searched neural architectures are conducted for both image classification and object detection tasks. And for the first time we prove that, without bells and whistles, PTQ can attain 4-bit ResNet and MobileNetV2 comparable with QAT and enjoy 240 times faster production of quantized models. Codes are available at https://github.com/yhhhli/BRECQ."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "Learned Step Size Quantization",
                    "abstract": "Deep networks run with low precision operations at inference time offer power and space advantages over high precision alternatives, but need to overcome the challenge of maintaining high accuracy as precision decreases. Here, we present a method for training such networks, Learned Step Size Quantization, that achieves the highest accuracy to date on the ImageNet dataset when using models, from a variety of architectures, with weights and activations quantized to 2-, 3- or 4-bits of precision, and that can train 3-bit models that reach full precision baseline accuracy. Our approach builds upon existing methods for learning weights in quantized networks by improving how the quantizer itself is configured. Specifically, we introduce a novel means to estimate and scale the task loss gradient at each weight and activation layer's quantizer step size, such that it can be learned in conjunction with other network parameters. This approach works using different levels of precision as needed for a given system and requires only a simple modification of existing training code."
                },
                {
                    "title": "The Unreasonable Effectiveness of Deep Features as a Perceptual Metric",
                    "abstract": "While it is nearly effortless for humans to quickly assess the perceptual similarity between two images, the underlying processes are thought to be quite complex. Despite this, the most widely used perceptual metrics today, such as PSNR and SSIM, are simple, shallow functions, and fail to account for many nuances of human perception. Recently, the deep learning community has found that features of the VGG network trained on ImageNet classification has been remarkably useful as a training loss for image synthesis. But how perceptual are these so-called \"perceptual losses\"? What elements are critical for their success? To answer these questions, we introduce a new dataset of human perceptual similarity judgments. We systematically evaluate deep features across different architectures and tasks and compare them with classic metrics. We find that deep features outperform all previous metrics by large margins on our dataset. More surprisingly, this result is not restricted to ImageNet-trained VGG features, but holds across different deep architectures and levels of supervision (supervised, self-supervised, or even unsupervised). Our results suggest that perceptual similarity is an emergent property shared across deep visual representations."
                },
                {
                    "title": "GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium",
                    "abstract": "Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the \"Frechet Inception Distance\" (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark."
                },
                {
                    "title": "Trained Ternary Quantization",
                    "abstract": "Deep neural networks are widely used in machine learning applications. However, the deployment of large neural networks models can be difficult to deploy on mobile devices with limited power budgets. To solve this problem, we propose Trained Ternary Quantization (TTQ), a method that can reduce the precision of weights in neural networks to ternary values. This method has very little accuracy degradation and can even improve the accuracy of some models (32, 44, 56-layer ResNet) on CIFAR-10 and AlexNet on ImageNet. And our AlexNet model is trained from scratch, which means it's as easy as to train normal full precision model. We highlight our trained quantization method that can learn both ternary values and ternary assignment. During inference, only ternary values (2-bit weights) and scaling factors are needed, therefore our models are nearly 16x smaller than full-precision models. Our ternary models can also be viewed as sparse binary weight networks, which can potentially be accelerated with custom circuit. Experiments on CIFAR-10 show that the ternary models obtained by trained quantization method outperform full-precision models of ResNet-32,44,56 by 0.04%, 0.16%, 0.36%, respectively. On ImageNet, our model outperforms full-precision AlexNet model by 0.3% of Top-1 accuracy and outperforms previous ternary models by 3%."
                },
                {
                    "title": "Ternary Weight Networks",
                    "abstract": "We present a memory and computation efficient ternary weight networks (TWNs) - with weights constrained to +1, 0 and -1. The Euclidian distance between full (float or double) precision weights and the ternary weights along with a scaling factor is minimized in training stage. Besides, a threshold-based ternary function is optimized to get an approximated solution which can be fast and easily computed. TWNs have shown better expressive abilities than binary precision counterparts. Meanwhile, TWNs achieve up to 16\u00d7 model compression rate and need fewer multiplications compared with the float32 precision counterparts. Extensive experiments on MNIST, CIFAR-10, and ImageNet datasets show that the TWNs achieve much better result than the Binary-Weight-Networks (BWNs) and the classification performance on MNIST and CIFAR-10 is very close to the full precision networks. We also verify our method on object detection task and show that TWNs significantly outperforms BWN by more than 10% mAP on PASCAL VOC dataset. The pytorch version of source code is available at: https://github.com/Thinklab-SJTU/twns."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "Fixed-point feedforward deep neural network design using weights +1, 0, and \u22121",
                    "abstract": "Feedforward deep neural networks that employ multiple hidden layers show high performance in many applications, but they demand complex hardware for implementation. The hardware complexity can be much lowered by minimizing the word-length of weights and signals, but direct quantization for fixed-point network design does not yield good results. We optimize the fixed-point design by employing backpropagation based retraining. The designed fixed-point networks with ternary weights (+1, 0, and -1) and 3-bit signal show only negligible performance loss when compared to the floating-point coun-terparts. The backpropagation for retraining uses quantized weights and fixed-point signal to compute the output, but utilizes high precision values for adapting the networks. A character recognition and a phoneme recognition examples are presented."
                },
                {
                    "title": "Stochastic Backpropagation and Approximate Inference in Deep Generative Models",
                    "abstract": "We marry ideas from deep neural networks and approximate Bayesian inference to derive a generalised class of deep, directed generative models, endowed with a new algorithm for scalable inference and learning. Our algorithm introduces a recognition model to represent approximate posterior distributions, and that acts as a stochastic encoder of the data. We develop stochastic back-propagation -- rules for back-propagation through stochastic variables -- and use this to develop an algorithm that allows for joint optimisation of the parameters of both the generative and recognition model. We demonstrate on several real-world data sets that the model generates realistic samples, provides accurate imputations of missing data and is a useful tool for high-dimensional data visualisation."
                },
                {
                    "title": "Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation",
                    "abstract": "Stochastic neurons and hard non-linearities can be useful for a number of reasons in deep learning models, but in many cases they pose a challenging problem: how to estimate the gradient of a loss function with respect to the input of such stochastic or non-smooth neurons? I.e., can we \"back-propagate\" through these stochastic neurons? We examine this question, existing approaches, and compare four families of solutions, applicable in different settings. One of them is the minimum variance unbiased gradient estimator for stochatic binary neurons (a special case of the REINFORCE algorithm). A second approach, introduced here, decomposes the operation of a binary stochastic neuron into a stochastic binary part and a smooth differentiable part, which approximates the expected effect of the pure stochatic binary neuron to first order. A third approach involves the injection of additive or multiplicative noise in a computational graph that is otherwise differentiable. A fourth approach heuristically copies the gradient with respect to the stochastic output directly as an estimator of the gradient with respect to the sigmoid argument (we call this the straight-through estimator). To explore a context where these estimators are useful, we consider a small-scale version of {\\em conditional computation}, where sparse stochastic units form a distributed representation of gaters that can turn off in combinatorially many ways large chunks of the computation performed in the rest of the neural network. In this case, it is important that the gating units produce an actual 0 most of the time. The resulting sparsity can be potentially be exploited to greatly reduce the computational cost of large deep networks for which conditional computation would be useful."
                },
                {
                    "title": "Image Quality Metrics: PSNR vs. SSIM",
                    "abstract": "In this paper, we analyse two well-known objective image quality metrics, the peak-signal-to-noise ratio (PSNR) as well as the structural similarity index measure (SSIM), and we derive a simple mathematical relationship between them which works for various kinds of image degradations such as Gaussian blur, additive Gaussian white noise, jpeg and jpeg2000 compression. A series of tests realized on images extracted from the Kodak database gives a better understanding of the similarity and difference between the SSIM and the PSNR."
                },
                {
                    "title": "Least squares quantization in PCM",
                    "abstract": "It has long been realized that in pulse-code modulation (PCM), with a given ensemble of signals to handle, the quantum values should be spaced more closely in the voltage regions where the signal amplitude is more likely to fall. It has been shown by Panter and Dite that, in the limit as the number of quanta becomes infinite, the asymptotic fractional density of quanta per unit voltage should vary as the one-third power of the probability density per unit voltage of signal amplitudes. In this paper the corresponding result for any finite number of quanta is derived; that is, necessary conditions are found that the quanta and associated quantization intervals of an optimum finite quantization scheme must satisfy. The optimization criterion used is that the average quantization noise power be a minimum. It is shown that the result obtained here goes over into the Panter and Dite result as the number of quanta become large. The optimum quautization schemes for 2^{b} quanta, b=1,2, \\cdots, 7 , are given numerically for Gaussian and for Laplacian distribution of signal amplitudes."
                },
                {
                    "title": "MobileDiffusion: Subsecond Text-to-Image Generation on Mobile Devices",
                    "abstract": "The deployment of large-scale text-to-image diffusion models on mobile devices is impeded by their substantial model size and slow inference speed. In this paper, we pro-pose MobileDiffusion , a highly efficient text-to-image diffusion model obtained through extensive optimizations in both architecture and sampling techniques. We conduct a comprehensive examination of model architecture design to reduce redundancy, enhance computational efficiency, and minimize model\u2019s parameter count, while preserving image generation quality. Additionally, we employ distillation and diffusion-GAN finetuning techniques on MobileDiffusion to achieve 8-step and 1-step inference respectively. Empirical studies, conducted both quantitatively and qualitatively, demonstrate the effectiveness of our proposed techniques. MobileDiffusion achieves a remarkable sub-second inference speed for generating a 512 \u00d7 512 image on mobile devices, establishing a new state of the art."
                },
                {
                    "title": "Q-DM: An Efficient Low-bit Quantized Diffusion Model",
                    "abstract": "Denoising diffusion generative models are capable of generating high-quality data, but suffers from the computation-costly generation process, due to a iterative noise estimation using full-precision networks. As an intuitive solution, quantization can significantly reduce the computational and memory consumption by low-bit parameters and operations. However, low-bit noise estimation networks in diffusion models (DMs) remain unexplored yet and perform much worse than the full-precision counterparts as observed in our experimental studies. In this paper, we first identify that the bottlenecks of low-bit quantized DMs come from a large distribution oscillation on activations and accumulated quantization error caused by the multi-step denoising process. To address these issues, we first develop a Timestep-aware Quantization (TaQ) method and a Noise-estimating Mimicking (NeM) scheme for low-bit quantized DMs (Q-DM) to effectively eliminate such oscillation and accumulated error respectively, leading to well-performed low-bit DMs. In this way, we propose an efficient Q-DM to calculate low-bit DMs by considering both training and inference process in the same framework. We evaluate our methods on popular DDPM and DDIM models. Extensive experimental results show that our method achieves a much better performance than the prior arts. For example, the 4-bit Q-DM theoretically accelerates the 1000-step DDPM by 7.8 \u00d7 and achieves a FID score of 5.17, on the unconditional CIFAR-10 dataset."
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively compress large-scale diffusion models, such as Stable Diffusion v1.5, into extremely low-bit representations while maintaining high image quality and text-image alignment?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing need for efficient machine learning models that can be deployed on resource-constrained devices like mobile and wearable technology. By enabling the use of high-fidelity diffusion models in such environments, this research could lead to broader applications in content creation, video generation, and 3D asset synthesis. Furthermore, advancements in model compression techniques could inspire future research in efficient model design and deployment, ultimately enhancing accessibility and usability in various practical applications.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the significant size of large-scale diffusion models, which can reach several gigabytes, making them difficult to store and transfer. Naive approaches to quantization often fail because they do not adequately address the unique characteristics of these models, such as their sensitivity to quantization errors. Additionally, existing quantization methods have primarily been tested on smaller models, and there is a lack of comprehensive evaluation for large-scale models like SD-v1.5. The complexities of determining optimal bit representations for different layers and the need for effective initialization techniques further complicate the quantization process.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on smaller diffusion models and has not adequately addressed the specific challenges posed by large-scale models. Existing quantization methods often target modest storage requirements and do not explore the potential of extremely low-bit quantization. Barriers such as the lack of extensive evaluation frameworks for large models and the absence of tailored techniques for initialization and training have hindered progress. Our approach differs by introducing a quantization-aware training framework that employs mixed-precision quantization and innovative initialization strategies, specifically designed for large-scale diffusion models.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, BitsFusion, involves a quantization-aware training framework that compresses the weights of large-scale diffusion models into extremely low bits (1.99 bits). Key components include:  \n1. **Mixed-Precision Quantization**: Analyzing quantization error metrics and applying different bit representations for various layers based on their sensitivity to quantization.  \n2."
            }
        },
        "author_data": {
            "398b16d1-026a-4bc7-bb11-554a2a1c2776": {
                "pk": "398b16d1-026a-4bc7-bb11-554a2a1c2776",
                "name": "Yang Sui",
                "collaborators": [
                    "Joerg Appenzeller",
                    "Miao Yin",
                    "Huy Phan",
                    "Bo Yuan",
                    "Xin He",
                    "Yang Bai",
                    "Tony Low",
                    "Mark Lundstrom",
                    "Zhuohang Li",
                    "Ding Ding"
                ],
                "domain": [
                    "Graphene",
                    "Machine Learning",
                    "Neural Networks",
                    "Image Compression"
                ],
                "publications": [
                    {
                        "title": "Screening and interlayer coupling in multilayer graphene field-effect transistors",
                        "abstract": "With the motivation of improving the performance and reliability of aggressively scaled nano-patterned graphene field-effect transistors, we present the first systematic experimental study on charge and current distribution in multilayer graphene field-effect transistors. We find a very particular thickness dependence for Ion, Ioff, and the Ion/Ioff ratio, and propose a resistor network model including screening and interlayer coupling to explain the experimental findings. In particular, our model does not invoke modification of the linear energy-band structure of graphene for the multilayer case. Noise reduction in nano-scale few-layer graphene transistors is experimentally demonstrated and can be understood within this model as well."
                    },
                    {
                        "title": "Implicit Regularization in Over-Parameterized Support Vector Machine",
                        "abstract": "In this paper, we design a regularization-free algorithm for high-dimensional support vector machines (SVMs) by integrating over-parameterization with Nesterov's smoothing method, and provide theoretical guarantees for the induced implicit regularization phenomenon. In particular, we construct an over-parameterized hinge loss function and estimate the true parameters by leveraging regularization-free gradient descent on this loss function. The utilization of Nesterov's method enhances the computational efficiency of our algorithm, especially in terms of determining the stopping criterion and reducing computational complexity. With appropriate choices of initialization, step size, and smoothness parameter, we demonstrate that unregularized gradient descent achieves a near-oracle statistical convergence rate. Additionally, we verify our theoretical findings through a variety of numerical experiments and compare the proposed method with explicit regularization. Our results illustrate the advantages of employing implicit regularization via gradient descent in conjunction with over-parameterization in sparse SVMs."
                    },
                    {
                        "title": "Signatures of disorder in the minimum conductivity of graphene",
                        "abstract": "Graphene has been proposed as a promising material for future nanoelectronics because of its unique electronic properties. Understanding the scaling behavior of this new nanomaterial under common experimental conditions is of critical importance for developing graphene-based nanoscale devices. We present a comprehensive experimental and theoretical study on the influence of edge disorder and bulk disorder on the minimum conductivity of graphene ribbons. For the first time, we discovered a strong non-monotonic size scaling behavior featuring a peak and saturation minimum conductivity. Through extensive numerical simulations and analysis, we are able to attribute these features to the amount of edge and bulk disorder in graphene devices. This study elucidates the quantum transport mechanisms in realistic experimental graphene systems, which can be used as a guideline for designing graphene-based nanoscale devices with improved performance."
                    },
                    {
                        "title": "Reconstruction Distortion of Learned Image Compression with Imperceptible Perturbations",
                        "abstract": "Learned Image Compression (LIC) has recently become the trending technique for image transmission due to its notable performance. Despite its popularity, the robustness of LIC with respect to the quality of image reconstruction remains under-explored. In this paper, we introduce an imperceptible attack approach designed to effectively degrade the reconstruction quality of LIC, resulting in the reconstructed image being severely disrupted by noise where any object in the reconstructed images is virtually impossible. More specifically, we generate adversarial examples by introducing a Frobenius norm-based loss function to maximize the discrepancy between original images and reconstructed adversarial examples. Further, leveraging the insensitivity of high-frequency components to human vision, we introduce Imperceptibility Constraint (IC) to ensure that the perturbations remain inconspicuous. Experiments conducted on the Kodak dataset using various LIC models demonstrate effectiveness. In addition, we provide several findings and suggestions for designing future defenses."
                    },
                    {
                        "title": "ELRT: Efficient Low-Rank Training for Compact Convolutional Neural Networks",
                        "abstract": "Low-rank compression, a popular model compression technique that produces compact convolutional neural networks (CNNs) with low rankness, has been well-studied in the literature. On the other hand, low-rank training, as an alternative way to train low-rank CNNs from scratch, has been exploited little yet. Unlike low-rank compression, low-rank training does not need pre-trained full-rank models, and the entire training phase is always performed on the low-rank structure, bringing attractive benefits for practical applications. However, the existing low-rank training solutions still face several challenges, such as a considerable accuracy drop and/or still needing to update full-size models during the training. In this paper, we perform a systematic investigation on low-rank CNN training. By identifying the proper low-rank format and performance-improving strategy, we propose ELRT, an efficient low-rank training solution for high-accuracy, high-compactness, low-rank CNN models. Our extensive evaluation results for training various CNNs on different datasets demonstrate the effectiveness of ELRT."
                    },
                    {
                        "title": "CHIP: CHannel Independence-based Pruning for Compact Neural Networks",
                        "abstract": "Filter pruning has been widely used for neural network compression because of its enabled practical acceleration. To date, most of the existing filter pruning works explore the importance of filters via using intra-channel information. In this paper, starting from an inter-channel perspective, we propose to perform efficient filter pruning using Channel Independence, a metric that measures the correlations among different feature maps. The less independent feature map is interpreted as containing less useful information$/$knowledge, and hence its corresponding filter can be pruned without affecting model capacity. We systematically investigate the quantification metric, measuring scheme and sensitiveness$/$reliability of channel independence in the context of filter pruning. Our evaluation results for different models on various datasets show the superior performance of our approach. Notably, on CIFAR-10 dataset our solution can bring $0.90\\%$ and $0.94\\%$ accuracy increase over baseline ResNet-56 and ResNet-110 models, respectively, and meanwhile the model size and FLOPs are reduced by $42.8\\%$ and $47.4\\%$ (for ResNet-56) and $48.3\\%$ and $52.1\\%$ (for ResNet-110), respectively. On ImageNet dataset, our approach can achieve $40.8\\%$ and $44.8\\%$ storage and computation reductions, respectively, with $0.15\\%$ accuracy increase over the baseline ResNet-50 model. The code is available at https://github.com/Eclipsess/CHIP_NeurIPS2021."
                    }
                ]
            },
            "787fbadb-f333-4095-938d-da68098e7aa4": {
                "pk": "787fbadb-f333-4095-938d-da68098e7aa4",
                "name": "Yanyu Li",
                "collaborators": [
                    "Yanzhi Wang",
                    "Jian Ren",
                    "Geng Yuan",
                    "Sergey Tulyakov",
                    "Ju Hu",
                    "Xue Lin",
                    "Qing Jin",
                    "Wei Niu",
                    "Bin Ren",
                    "Xulong Tang"
                ],
                "domain": [
                    "Neural Networks",
                    "Deep Learning",
                    "Computer Vision",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Searching for the charged-current non-standard neutrino interactions at the $e^{-}p$ colliders",
                        "abstract": "Considering the theoretical constraints on the charged-current (CC) non-standard neutrino interaction (NSI) parameters in a simplified $W'$ model, we study the sensitivities of the Large Hadron electron Collider (LHeC) and Future Circular Collider-hadron electron (FCC-he) to the CC NSI parameters through the subprocess $e^{-}q \\to W' \\to v_{e}q'$. Our results show that the LHeC with $\\mathcal{L} = 100\\;{\\rm fb}^{-1}$ is a little more sensitive to the CC NSI parameters than that of the high-luminosity (HL) LHC, and the absolute values of the lower limits on the CC NSI parameters at the FCC-he are smaller than those of the HL-LHC about two orders of magnitude."
                    },
                    {
                        "title": "Lithium intercalation drives remarkable mechanical properties deterioration in bulk and single-layered phosphorus: A first-principles study",
                        "abstract": "It is of critical importance to understand the mechanical properties change of electrode materials during lithium intercalation in the mechanical design of Li-ion batteries, for the purpose of the high reliability and safety in their applications. Here, we investigated the mechanical properties of both bulk and single layer phosphorus during the lithium intercalation process by using the first-principles calculations. Our results show that the Young's modulus of bulk and layered phosphorus strongly depends on the lithium intercalation. The mechanical bearing capacities, such as critical strain and stress, are significantly reduced by several times after lithium intercalation in both bulk and single layer phosphorus, which may reduce the reliability of Li-ion batteries. Our findings suggest that this remarkable mechanical properties deterioration during Li intercalation should be considered carefully in the mechanical design of Li-ion batteries, in order to keep they working reliably and safely in the charge-discharge process."
                    },
                    {
                        "title": "Manipulating liquid-liquid phase separation using patterned flow",
                        "abstract": "The precise control of liquid-liquid phase separation (LLPS) is the key to developing cutting-edge technologies that benefit diverse disciplines. Fluid flow was found to be capable of controlling the structure and effective temperature of LLPS, but the extent and precision of control were less than optimal. In this article, we propose that patterned flow can be employed as a generic tool to manipulate LLPS effectively. By combining theoretical modeling and numerical simulations, we demonstrate that flows with tailor-made structures can become functional, allowing us to control diverse aspects of LLPS. Typical examples include the capture and pinning of droplets, fine-tuning of droplet sizes, forced assembly of periodic droplet arrays, and the remodeling of the kinetics and structure of phase separation. These manipulations are grounded on the redistribution of chemical potential by the structured flow. Our results not only can lead to potential LLPS-based technologies, but also highlight the rich behavior of LLPS introduced by the patterned flow."
                    },
                    {
                        "title": "Pruning-as-Search: Efficient Neural Architecture Search via Channel Pruning and Structural Reparameterization",
                        "abstract": "Neural architecture search (NAS) and network pruning are widely studied efficient AI techniques, but not yet perfect. NAS performs exhaustive candidate architecture search, incurring tremendous search cost. Though (structured) pruning can simply shrink model dimension, it remains unclear how to decide the per-layer sparsity automatically and optimally. In this work, we revisit the problem of layer-width optimization and propose Pruning-as-Search (PaS), an end-to-end channel pruning method to search out desired sub-network automatically and efficiently. Specifically, we add a depth-wise binary convolution to learn pruning policies directly through gradient descent. By combining the structural reparameterization and PaS, we successfully searched out a new family of VGG-like and lightweight networks, which enable the flexibility of arbitrary width with respect to each layer instead of each stage. Experimental results show that our proposed architecture outperforms prior arts by around $1.0\\%$ top-1 accuracy under similar inference speed on ImageNet-1000 classification task. Furthermore, we demonstrate the effectiveness of our width search on complex tasks including instance segmentation and image translation. Code and models are released."
                    },
                    {
                        "title": "ILMPQ : An Intra-Layer Multi-Precision Deep Neural Network Quantization framework for FPGA",
                        "abstract": "This work targets the commonly used FPGA (field-programmable gate array) devices as the hardware platform for DNN edge computing. We focus on DNN quantization as the main model compression technique. The novelty of this work is: We use a quantization method that supports multiple precisions along the intra-layer dimension, while the existing quantization methods apply multi-precision quantization along the inter-layer dimension. The intra-layer multi-precision method can uniform the hardware configurations for different layers to reduce computation overhead and at the same time preserve the model accuracy as the inter-layer approach. Our proposed ILMPQ DNN quantization framework achieves 70.73 Top1 accuracy in ResNet-18 on the ImageNet dataset. We also validate the proposed MSP framework on two FPGA devices i.e., Xilinx XC7Z020 and XC7Z045. We achieve 3.65x speedup in end-to-end inference time on the ImageNet, compared with the fixed-point quantization method."
                    },
                    {
                        "title": "PIM-QAT: Neural Network Quantization for Processing-In-Memory (PIM) Systems",
                        "abstract": "Processing-in-memory (PIM), an increasingly studied neuromorphic hardware, promises orders of energy and throughput improvements for deep learning inference. Leveraging the massively parallel and efficient analog computing inside memories, PIM circumvents the bottlenecks of data movements in conventional digital hardware. However, an extra quantization step (i.e. PIM quantization), typically with limited resolution due to hardware constraints, is required to convert the analog computing results into digital domain. Meanwhile, non-ideal effects extensively exist in PIM quantization because of the imperfect analog-to-digital interface, which further compromises the inference accuracy.   In this paper, we propose a method for training quantized networks to incorporate PIM quantization, which is ubiquitous to all PIM systems. Specifically, we propose a PIM quantization aware training (PIM-QAT) algorithm, and introduce rescaling techniques during backward and forward propagation by analyzing the training dynamics to facilitate training convergence. We also propose two techniques, namely batch normalization (BN) calibration and adjusted precision training, to suppress the adverse effects of non-ideal linearity and stochastic thermal noise involved in real PIM chips. Our method is validated on three mainstream PIM decomposition schemes, and physically on a prototype chip. Comparing with directly deploying conventionally trained quantized model on PIM systems, which does not take into account this extra quantization step and thus fails, our method provides significant improvement. It also achieves comparable inference accuracy on PIM systems as that of conventionally quantized models on digital hardware, across CIFAR10 and CIFAR100 datasets using various network depths for the most popular network topology."
                    },
                    {
                        "title": "Real-Time Portrait Stylization on the Edge",
                        "abstract": "In this work we demonstrate real-time portrait stylization, specifically, translating self-portrait into cartoon or anime style on mobile devices. We propose a latency-driven differentiable architecture search method, maintaining realistic generative quality. With our framework, we obtain $10\\times$ computation reduction on the generative model and achieve real-time video stylization on off-the-shelf smartphone using mobile GPUs."
                    },
                    {
                        "title": "YOLObile: Real-Time Object Detection on Mobile Devices via Compression-Compilation Co-Design",
                        "abstract": "The rapid development and wide utilization of object detection techniques have aroused attention on both accuracy and speed of object detectors. However, the current state-of-the-art object detection works are either accuracy-oriented using a large model but leading to high latency or speed-oriented using a lightweight model but sacrificing accuracy. In this work, we propose YOLObile framework, a real-time object detection on mobile devices via compression-compilation co-design. A novel block-punched pruning scheme is proposed for any kernel size. To improve computational efficiency on mobile devices, a GPU-CPU collaborative scheme is adopted along with advanced compiler-assisted optimizations. Experimental results indicate that our pruning scheme achieves 14$\\times$ compression rate of YOLOv4 with 49.0 mAP. Under our YOLObile framework, we achieve 17 FPS inference speed using GPU on Samsung Galaxy S20. By incorporating our proposed GPU-CPU collaborative scheme, the inference speed is increased to 19.1 FPS, and outperforms the original YOLOv4 by 5$\\times$ speedup. Source code is at: \\url{https://github.com/nightsnack/YOLObile}."
                    },
                    {
                        "title": "EfficientFormer: Vision Transformers at MobileNet Speed",
                        "abstract": "Vision Transformers (ViT) have shown rapid progress in computer vision tasks, achieving promising results on various benchmarks. However, due to the massive number of parameters and model design, \\textit{e.g.}, attention mechanism, ViT-based models are generally times slower than lightweight convolutional networks. Therefore, the deployment of ViT for real-time applications is particularly challenging, especially on resource-constrained hardware such as mobile devices. Recent efforts try to reduce the computation complexity of ViT through network architecture search or hybrid design with MobileNet block, yet the inference speed is still unsatisfactory. This leads to an important question: can transformers run as fast as MobileNet while obtaining high performance? To answer this, we first revisit the network architecture and operators used in ViT-based models and identify inefficient designs. Then we introduce a dimension-consistent pure transformer (without MobileNet blocks) as a design paradigm. Finally, we perform latency-driven slimming to get a series of final models dubbed EfficientFormer. Extensive experiments show the superiority of EfficientFormer in performance and speed on mobile devices. Our fastest model, EfficientFormer-L1, achieves $79.2\\%$ top-1 accuracy on ImageNet-1K with only $1.6$ ms inference latency on iPhone 12 (compiled with CoreML), which runs as fast as MobileNetV2$\\times 1.4$ ($1.6$ ms, $74.7\\%$ top-1), and our largest model, EfficientFormer-L7, obtains $83.3\\%$ accuracy with only $7.0$ ms latency. Our work proves that properly designed transformers can reach extremely low latency on mobile devices while maintaining high performance."
                    },
                    {
                        "title": "Layer Freezing & Data Sieving: Missing Pieces of a Generic Framework for Sparse Training",
                        "abstract": "Recently, sparse training has emerged as a promising paradigm for efficient deep learning on edge devices. The current research mainly devotes efforts to reducing training costs by further increasing model sparsity. However, increasing sparsity is not always ideal since it will inevitably introduce severe accuracy degradation at an extremely high sparsity level. This paper intends to explore other possible directions to effectively and efficiently reduce sparse training costs while preserving accuracy. To this end, we investigate two techniques, namely, layer freezing and data sieving. First, the layer freezing approach has shown its success in dense model training and fine-tuning, yet it has never been adopted in the sparse training domain. Nevertheless, the unique characteristics of sparse training may hinder the incorporation of layer freezing techniques. Therefore, we analyze the feasibility and potentiality of using the layer freezing technique in sparse training and find it has the potential to save considerable training costs. Second, we propose a data sieving method for dataset-efficient training, which further reduces training costs by ensuring only a partial dataset is used throughout the entire training process. We show that both techniques can be well incorporated into the sparse training algorithm to form a generic framework, which we dub SpFDE. Our extensive experiments demonstrate that SpFDE can significantly reduce training costs while preserving accuracy from three dimensions: weight sparsity, layer freezing, and dataset sieving."
                    },
                    {
                        "title": "Rethinking Vision Transformers for MobileNet Size and Speed",
                        "abstract": "With the success of Vision Transformers (ViTs) in computer vision tasks, recent arts try to optimize the performance and complexity of ViTs to enable efficient deployment on mobile devices. Multiple approaches are proposed to accelerate attention mechanism, improve inefficient designs, or incorporate mobile-friendly lightweight convolutions to form hybrid architectures. However, ViT and its variants still have higher latency or considerably more parameters than lightweight CNNs, even true for the years-old MobileNet. In practice, latency and size are both crucial for efficient deployment on resource-constraint hardware. In this work, we investigate a central question, can transformer models run as fast as MobileNet and maintain a similar size? We revisit the design choices of ViTs and propose a novel supernet with low latency and high parameter efficiency. We further introduce a novel fine-grained joint search strategy for transformer models that can find efficient architectures by optimizing latency and number of parameters simultaneously. The proposed models, EfficientFormerV2, achieve 3.5% higher top-1 accuracy than MobileNetV2 on ImageNet-1K with similar latency and parameters. This work demonstrate that properly designed and optimized vision transformers can achieve high performance even with MobileNet-level size and speed."
                    },
                    {
                        "title": "SnapFusion: Text-to-Image Diffusion Model on Mobile Devices within Two Seconds",
                        "abstract": "Text-to-image diffusion models can create stunning images from natural language descriptions that rival the work of professional artists and photographers. However, these models are large, with complex network architectures and tens of denoising iterations, making them computationally expensive and slow to run. As a result, high-end GPUs and cloud-based inference are required to run diffusion models at scale. This is costly and has privacy implications, especially when user data is sent to a third party. To overcome these challenges, we present a generic approach that, for the first time, unlocks running text-to-image diffusion models on mobile devices in less than $2$ seconds. We achieve so by introducing efficient network architecture and improving step distillation. Specifically, we propose an efficient UNet by identifying the redundancy of the original model and reducing the computation of the image decoder via data distillation. Further, we enhance the step distillation by exploring training strategies and introducing regularization from classifier-free guidance. Our extensive experiments on MS-COCO show that our model with $8$ denoising steps achieves better FID and CLIP scores than Stable Diffusion v$1.5$ with $50$ steps. Our work democratizes content creation by bringing powerful text-to-image diffusion models to the hands of users."
                    },
                    {
                        "title": "TextCraftor: Your Text Encoder Can be Image Quality Controller",
                        "abstract": "Diffusion-based text-to-image generative models, e.g., Stable Diffusion, have revolutionized the field of content generation, enabling significant advancements in areas like image editing and video synthesis. Despite their formidable capabilities, these models are not without their limitations. It is still challenging to synthesize an image that aligns well with the input text, and multiple runs with carefully crafted prompts are required to achieve satisfactory results. To mitigate these limitations, numerous studies have endeavored to fine-tune the pre-trained diffusion models, i.e., UNet, utilizing various technologies. Yet, amidst these efforts, a pivotal question of text-to-image diffusion model training has remained largely unexplored: Is it possible and feasible to fine-tune the text encoder to improve the performance of text-to-image diffusion models? Our findings reveal that, instead of replacing the CLIP text encoder used in Stable Diffusion with other large language models, we can enhance it through our proposed fine-tuning approach, TextCraftor, leading to substantial improvements in quantitative benchmarks and human assessments. Interestingly, our technique also empowers controllable image generation through the interpolation of different text encoders fine-tuned with various rewards. We also demonstrate that TextCraftor is orthogonal to UNet finetuning, and can be combined to further improve generative quality."
                    }
                ]
            },
            "bdc1b03f-f778-4829-85f1-940f8ab1d9e2": {
                "pk": "bdc1b03f-f778-4829-85f1-940f8ab1d9e2",
                "name": "Anil Kag",
                "collaborators": [
                    "Venkatesh Saligrama",
                    "Sergey Tulyakov",
                    "Jian Ren",
                    "Ziming Zhang",
                    "Aditya Gangrade",
                    "Yanyu Li",
                    "Ju Hu",
                    "Junli Cao",
                    "Alan Sullivan",
                    "Ashok Cutkosky"
                ],
                "domain": [
                    "Recurrent Neural Networks",
                    "Selective Classification",
                    "Text-to-Image Generation",
                    "Video Synthesis"
                ],
                "publications": [
                    {
                        "title": "RNNs Evolving on an Equilibrium Manifold: A Panacea for Vanishing and Exploding Gradients?",
                        "abstract": "Recurrent neural networks (RNNs) are particularly well-suited for modeling long-term dependencies in sequential data, but are notoriously hard to train because the error backpropagated in time either vanishes or explodes at an exponential rate. While a number of works attempt to mitigate this effect through gated recurrent units, well-chosen parametric constraints, and skip-connections, we develop a novel perspective that seeks to evolve the hidden state on the equilibrium manifold of an ordinary differential equation (ODE). We propose a family of novel RNNs, namely {\\em Equilibriated Recurrent Neural Networks} (ERNNs) that overcome the gradient decay or explosion effect and lead to recurrent models that evolve on the equilibrium manifold. We show that equilibrium points are stable, leading to fast convergence of the discretized ODE to fixed points. Furthermore, ERNNs account for long-term dependencies, and can efficiently recall informative aspects of data from the distant past. We show that ERNNs achieve state-of-the-art accuracy on many challenging data sets with 3-10x speedups, 1.5-3x model size reduction, and with similar prediction cost relative to vanilla RNNs."
                    },
                    {
                        "title": "Equilibrated Recurrent Neural Network: Neuronal Time-Delayed Self-Feedback Improves Accuracy and Stability",
                        "abstract": "We propose a novel {\\it Equilibrated Recurrent Neural Network} (ERNN) to combat the issues of inaccuracy and instability in conventional RNNs. Drawing upon the concept of autapse in neuroscience, we propose augmenting an RNN with a time-delayed self-feedback loop. Our sole purpose is to modify the dynamics of each internal RNN state and, at any time, enforce it to evolve close to the equilibrium point associated with the input signal at that time. We show that such self-feedback helps stabilize the hidden state transitions leading to fast convergence during training while efficiently learning discriminative latent features that result in state-of-the-art results on several benchmark datasets at test-time. We propose a novel inexact Newton method to solve fixed-point conditions given model parameters for generating the latent features at each hidden state. We prove that our inexact Newton method converges locally with linear rate (under mild conditions). We leverage this result for efficient training of ERNNs based on backpropagation."
                    },
                    {
                        "title": "Selective Classification via One-Sided Prediction",
                        "abstract": "We propose a novel method for selective classification (SC), a problem which allows a classifier to abstain from predicting some instances, thus trading off accuracy against coverage (the fraction of instances predicted). In contrast to prior gating or confidence-set based work, our proposed method optimises a collection of class-wise decoupled one-sided empirical risks, and is in essence a method for explicitly finding the largest decision sets for each class that have few false positives. This one-sided prediction (OSP) based relaxation yields an SC scheme that attains near-optimal coverage in the practically relevant high target accuracy regime, and further admits efficient implementation, leading to a flexible and principled method for SC. We theoretically derive generalization bounds for SC and OSP, and empirically we show that our scheme strongly outperforms state of the art methods in coverage at small error levels."
                    },
                    {
                        "title": "Online Selective Classification with Limited Feedback",
                        "abstract": "Motivated by applications to resource-limited and safety-critical domains, we study selective classification in the online learning model, wherein a predictor may abstain from classifying an instance. For example, this may model an adaptive decision to invoke more resources on this instance. Two salient aspects of the setting we consider are that the data may be non-realisable, due to which abstention may be a valid long-term action, and that feedback is only received when the learner abstains, which models the fact that reliable labels are only available when the resource intensive processing is invoked.   Within this framework, we explore strategies that make few mistakes, while not abstaining too many times more than the best-in-hindsight error-free classifier from a given class. That is, the one that makes no mistakes, while abstaining the fewest number of times. We construct simple versioning-based schemes for any $\\mu \\in (0,1],$ that make most $T^\\mu$ mistakes while incurring \\smash{$\\tilde{O}(T^{1-\\mu})$} excess abstention against adaptive adversaries. We further show that this dependence on $T$ is tight, and provide illustrative experiments on realistic datasets."
                    },
                    {
                        "title": "Scalable Ranked Preference Optimization for Text-to-Image Generation",
                        "abstract": "Direct Preference Optimization (DPO) has emerged as a powerful approach to align text-to-image (T2I) models with human feedback. Unfortunately, successful application of DPO to T2I models requires a huge amount of resources to collect and label large-scale datasets, e.g., millions of generated paired images annotated with human preferences. In addition, these human preference datasets can get outdated quickly as the rapid improvements of T2I models lead to higher quality images. In this work, we investigate a scalable approach for collecting large-scale and fully synthetic datasets for DPO training. Specifically, the preferences for paired images are generated using a pre-trained reward function, eliminating the need for involving humans in the annotation process, greatly improving the dataset collection efficiency. Moreover, we demonstrate that such datasets allow averaging predictions across multiple models and collecting ranked preferences as opposed to pairwise preferences. Furthermore, we introduce RankDPO to enhance DPO-based methods using the ranking feedback. Applying RankDPO on SDXL and SD3-Medium models with our synthetically generated preference dataset ``Syn-Pic'' improves both prompt-following (on benchmarks like T2I-Compbench, GenEval, and DPG-Bench) and visual quality (through user studies). This pipeline presents a practical and scalable solution to develop better preference datasets to enhance the performance of text-to-image models."
                    },
                    {
                        "title": "TextCraftor: Your Text Encoder Can be Image Quality Controller",
                        "abstract": "Diffusion-based text-to-image generative models, e.g., Stable Diffusion, have revolutionized the field of content generation, enabling significant advancements in areas like image editing and video synthesis. Despite their formidable capabilities, these models are not without their limitations. It is still challenging to synthesize an image that aligns well with the input text, and multiple runs with carefully crafted prompts are required to achieve satisfactory results. To mitigate these limitations, numerous studies have endeavored to fine-tune the pre-trained diffusion models, i.e., UNet, utilizing various technologies. Yet, amidst these efforts, a pivotal question of text-to-image diffusion model training has remained largely unexplored: Is it possible and feasible to fine-tune the text encoder to improve the performance of text-to-image diffusion models? Our findings reveal that, instead of replacing the CLIP text encoder used in Stable Diffusion with other large language models, we can enhance it through our proposed fine-tuning approach, TextCraftor, leading to substantial improvements in quantitative benchmarks and human assessments. Interestingly, our technique also empowers controllable image generation through the interpolation of different text encoders fine-tuned with various rewards. We also demonstrate that TextCraftor is orthogonal to UNet finetuning, and can be combined to further improve generative quality."
                    },
                    {
                        "title": "Lightweight Predictive 3D Gaussian Splats",
                        "abstract": "Recent approaches representing 3D objects and scenes using Gaussian splats show increased rendering speed across a variety of platforms and devices. While rendering such representations is indeed extremely efficient, storing and transmitting them is often prohibitively expensive. To represent large-scale scenes, one often needs to store millions of 3D Gaussians, occupying gigabytes of disk space. This poses a very practical limitation, prohibiting widespread adoption.Several solutions have been proposed to strike a balance between disk size and rendering quality, noticeably reducing the visual quality. In this work, we propose a new representation that dramatically reduces the hard drive footprint while featuring similar or improved quality when compared to the standard 3D Gaussian splats. When compared to other compact solutions, ours offers higher quality renderings with significantly reduced storage, being able to efficiently run on a mobile device in real-time. Our key observation is that nearby points in the scene can share similar representations. Hence, only a small ratio of 3D points needs to be stored. We introduce an approach to identify such points which are called parent points. The discarded points called children points along with attributes can be efficiently predicted by tiny MLPs."
                    },
                    {
                        "title": "SF-V: Single Forward Video Generation Model",
                        "abstract": "Diffusion-based video generation models have demonstrated remarkable success in obtaining high-fidelity videos through the iterative denoising process. However, these models require multiple denoising steps during sampling, resulting in high computational costs. In this work, we propose a novel approach to obtain single-step video generation models by leveraging adversarial training to fine-tune pre-trained video diffusion models. We show that, through the adversarial training, the multi-steps video diffusion model, i.e., Stable Video Diffusion (SVD), can be trained to perform single forward pass to synthesize high-quality videos, capturing both temporal and spatial dependencies in the video data. Extensive experiments demonstrate that our method achieves competitive generation quality of synthesized videos with significantly reduced computational overhead for the denoising process (i.e., around $23\\times$ speedup compared with SVD and $6\\times$ speedup compared with existing works, with even better generation quality), paving the way for real-time video synthesis and editing. More visualization results are made publicly available at https://snap-research.github.io/SF-V."
                    }
                ]
            },
            "83332f0c-6bfc-42f4-93c3-566e7b7fd316": {
                "pk": "83332f0c-6bfc-42f4-93c3-566e7b7fd316",
                "name": "Yerlan Idelbayev",
                "collaborators": [
                    "Miguel \u00c1. Carreira-Perpi\u00f1\u00e1n",
                    "Yanyu Li",
                    "Yanzhi Wang",
                    "Sergey Tulyakov",
                    "Jian Ren",
                    "Xian Liu",
                    "Yifan Gong",
                    "Zheng Zhan",
                    "Andrey Zharkov",
                    "Kfir Aberman"
                ],
                "domain": [
                    "Model Compression",
                    "Deep Learning",
                    "Generative Models",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Model compression as constrained optimization, with application to neural nets. Part V: combining compressions",
                        "abstract": "Model compression is generally performed by using quantization, low-rank approximation or pruning, for which various algorithms have been researched in recent years. One fundamental question is: what types of compression work better for a given model? Or even better: can we improve by combining compressions in a suitable way? We formulate this generally as a problem of optimizing the loss but where the weights are constrained to equal an additive combination of separately compressed parts; and we give an algorithm to learn the corresponding parts' parameters. Experimentally with deep neural nets, we observe that 1) we can find significantly better models in the error-compression space, indicating that different compression types have complementary benefits, and 2) the best type of combination depends exquisitely on the type of neural net. For example, we can compress ResNets and AlexNet using only 1 bit per weight without error degradation at the cost of adding a few floating point weights. However, VGG nets can be better compressed by combining low-rank with a few floating point weights."
                    },
                    {
                        "title": "A flexible, extensible software framework for model compression based on the LC algorithm",
                        "abstract": "We propose a software framework based on the ideas of the Learning-Compression (LC) algorithm, that allows a user to compress a neural network or other machine learning model using different compression schemes with minimal effort. Currently, the supported compressions include pruning, quantization, low-rank methods (including automatically learning the layer ranks), and combinations of those, and the user can choose different compression types for different parts of a neural network.   The LC algorithm alternates two types of steps until convergence: a learning (L) step, which trains a model on a dataset (using an algorithm such as SGD); and a compression (C) step, which compresses the model parameters (using a compression scheme such as low-rank or quantization). This decoupling of the \"machine learning\" aspect from the \"signal compression\" aspect means that changing the model or the compression type amounts to calling the corresponding subroutine in the L or C step, respectively. The library fully supports this by design, which makes it flexible and extensible. This does not come at the expense of performance: the runtime needed to compress a model is comparable to that of training the model in the first place; and the compressed model is competitive in terms of prediction accuracy and compression ratio with other algorithms (which are often specialized for specific models or compression schemes). The library is written in Python and PyTorch and available in Github."
                    },
                    {
                        "title": "Model compression as constrained optimization, with application to neural nets. Part II: quantization",
                        "abstract": "We consider the problem of deep neural net compression by quantization: given a large, reference net, we want to quantize its real-valued weights using a codebook with $K$ entries so that the training loss of the quantized net is minimal. The codebook can be optimally learned jointly with the net, or fixed, as for binarization or ternarization approaches. Previous work has quantized the weights of the reference net, or incorporated rounding operations in the backpropagation algorithm, but this has no guarantee of converging to a loss-optimal, quantized net. We describe a new approach based on the recently proposed framework of model compression as constrained optimization \\citep{Carreir17a}. This results in a simple iterative \"learning-compression\" algorithm, which alternates a step that learns a net of continuous weights with a step that quantizes (or binarizes/ternarizes) the weights, and is guaranteed to converge to local optimum of the loss for quantized nets. We develop algorithms for an adaptive codebook or a (partially) fixed codebook. The latter includes binarization, ternarization, powers-of-two and other important particular cases. We show experimentally that we can achieve much higher compression rates than previous quantization work (even using just 1 bit per weight) with negligible loss degradation."
                    },
                    {
                        "title": "Structured Multi-Hashing for Model Compression",
                        "abstract": "Despite the success of deep neural networks (DNNs), state-of-the-art models are too large to deploy on low-resource devices or common server configurations in which multiple models are held in memory. Model compression methods address this limitation by reducing the memory footprint, latency, or energy consumption of a model with minimal impact on accuracy. We focus on the task of reducing the number of learnable variables in the model. In this work we combine ideas from weight hashing and dimensionality reductions resulting in a simple and powerful structured multi-hashing method based on matrix products that allows direct control of model size of any deep network and is trained end-to-end. We demonstrate the strength of our approach by compressing models from the ResNet, EfficientNet, and MobileNet architecture families. Our method allows us to drastically decrease the number of variables while maintaining high accuracy. For instance, by applying our approach to EfficentNet-B4 (16M parameters) we reduce it to to the size of B0 (5M parameters), while gaining over 3% in accuracy over B0 baseline. On the commonly used benchmark CIFAR10 we reduce the ResNet32 model by 75% with no loss in quality, and are able to do a 10x compression while still achieving above 90% accuracy."
                    },
                    {
                        "title": "TextCraftor: Your Text Encoder Can be Image Quality Controller",
                        "abstract": "Diffusion-based text-to-image generative models, e.g., Stable Diffusion, have revolutionized the field of content generation, enabling significant advancements in areas like image editing and video synthesis. Despite their formidable capabilities, these models are not without their limitations. It is still challenging to synthesize an image that aligns well with the input text, and multiple runs with carefully crafted prompts are required to achieve satisfactory results. To mitigate these limitations, numerous studies have endeavored to fine-tune the pre-trained diffusion models, i.e., UNet, utilizing various technologies. Yet, amidst these efforts, a pivotal question of text-to-image diffusion model training has remained largely unexplored: Is it possible and feasible to fine-tune the text encoder to improve the performance of text-to-image diffusion models? Our findings reveal that, instead of replacing the CLIP text encoder used in Stable Diffusion with other large language models, we can enhance it through our proposed fine-tuning approach, TextCraftor, leading to substantial improvements in quantitative benchmarks and human assessments. Interestingly, our technique also empowers controllable image generation through the interpolation of different text encoders fine-tuned with various rewards. We also demonstrate that TextCraftor is orthogonal to UNet finetuning, and can be combined to further improve generative quality."
                    },
                    {
                        "title": "Efficient Training with Denoised Neural Weights",
                        "abstract": "Good weight initialization serves as an effective measure to reduce the training cost of a deep neural network (DNN) model. The choice of how to initialize parameters is challenging and may require manual tuning, which can be time-consuming and prone to human error. To overcome such limitations, this work takes a novel step towards building a weight generator to synthesize the neural weights for initialization. We use the image-to-image translation task with generative adversarial networks (GANs) as an example due to the ease of collecting model weights spanning a wide range. Specifically, we first collect a dataset with various image editing concepts and their corresponding trained weights, which are later used for the training of the weight generator. To address the different characteristics among layers and the substantial number of weights to be predicted, we divide the weights into equal-sized blocks and assign each block an index. Subsequently, a diffusion model is trained with such a dataset using both text conditions of the concept and the block indexes. By initializing the image translation model with the denoised weights predicted by our diffusion model, the training requires only 43.3 seconds. Compared to training from scratch (i.e., Pix2pix), we achieve a 15x training time acceleration for a new concept while obtaining even better image generation quality."
                    },
                    {
                        "title": "E$^{2}$GAN: Efficient Training of Efficient GANs for Image-to-Image Translation",
                        "abstract": "One highly promising direction for enabling flexible real-time on-device image editing is utilizing data distillation by leveraging large-scale text-to-image diffusion models to generate paired datasets used for training generative adversarial networks (GANs). This approach notably alleviates the stringent requirements typically imposed by high-end commercial GPUs for performing image editing with diffusion models. However, unlike text-to-image diffusion models, each distilled GAN is specialized for a specific image editing task, necessitating costly training efforts to obtain models for various concepts. In this work, we introduce and address a novel research direction: can the process of distilling GANs from diffusion models be made significantly more efficient? To achieve this goal, we propose a series of innovative techniques. First, we construct a base GAN model with generalized features, adaptable to different concepts through fine-tuning, eliminating the need for training from scratch. Second, we identify crucial layers within the base GAN model and employ Low-Rank Adaptation (LoRA) with a simple yet effective rank search process, rather than fine-tuning the entire base model. Third, we investigate the minimal amount of data necessary for fine-tuning, further reducing the overall training time. Extensive experiments show that we can efficiently empower GANs with the ability to perform real-time high-quality image editing on mobile devices with remarkably reduced training and storage costs for each concept."
                    }
                ]
            },
            "c0f4b8a9-7b07-4aac-a759-686aeaa23ea2": {
                "pk": "c0f4b8a9-7b07-4aac-a759-686aeaa23ea2",
                "name": "Junli Cao",
                "collaborators": [
                    "Sergey Tulyakov",
                    "Jian Ren",
                    "Chaoyang Wang",
                    "Aliaksandr Siarohin",
                    "Ju Hu",
                    "Peiye Zhuang",
                    "Hsin-Ying Lee",
                    "M. D.",
                    "Anil Kag",
                    "Guocheng Qian"
                ],
                "domain": [
                    "3D Rendering",
                    "Neural Networks",
                    "Medical Image Analysis",
                    "Video Generation"
                ],
                "publications": [
                    {
                        "title": "Lightweight Predictive 3D Gaussian Splats",
                        "abstract": "Recent approaches representing 3D objects and scenes using Gaussian splats show increased rendering speed across a variety of platforms and devices. While rendering such representations is indeed extremely efficient, storing and transmitting them is often prohibitively expensive. To represent large-scale scenes, one often needs to store millions of 3D Gaussians, occupying gigabytes of disk space. This poses a very practical limitation, prohibiting widespread adoption.Several solutions have been proposed to strike a balance between disk size and rendering quality, noticeably reducing the visual quality. In this work, we propose a new representation that dramatically reduces the hard drive footprint while featuring similar or improved quality when compared to the standard 3D Gaussian splats. When compared to other compact solutions, ours offers higher quality renderings with significantly reduced storage, being able to efficiently run on a mobile device in real-time. Our key observation is that nearby points in the scene can share similar representations. Hence, only a small ratio of 3D points needs to be stored. We introduce an approach to identify such points which are called parent points. The discarded points called children points along with attributes can be efficiently predicted by tiny MLPs."
                    },
                    {
                        "title": "Diffusion Priors for Dynamic View Synthesis from Monocular Videos",
                        "abstract": "Dynamic novel view synthesis aims to capture the temporal evolution of visual content within videos. Existing methods struggle to distinguishing between motion and structure, particularly in scenarios where camera poses are either unknown or constrained compared to object motion. Furthermore, with information solely from reference images, it is extremely challenging to hallucinate unseen regions that are occluded or partially observed in the given videos. To address these issues, we first finetune a pretrained RGB-D diffusion model on the video frames using a customization technique. Subsequently, we distill the knowledge from the finetuned model to a 4D representations encompassing both dynamic and static Neural Radiance Fields (NeRF) components. The proposed pipeline achieves geometric consistency while preserving the scene identity. We perform thorough experiments to evaluate the efficacy of the proposed method qualitatively and quantitatively. Our results demonstrate the robustness and utility of our approach in challenging cases, further advancing dynamic novel view synthesis."
                    },
                    {
                        "title": "Surrogate Supervision for Medical Image Analysis: Effective Deep Learning From Limited Quantities of Labeled Data",
                        "abstract": "We investigate the effectiveness of a simple solution to the common problem of deep learning in medical image analysis with limited quantities of labeled training data. The underlying idea is to assign artificial labels to abundantly available unlabeled medical images and, through a process known as surrogate supervision, pre-train a deep neural network model for the target medical image analysis task lacking sufficient labeled training data. In particular, we employ 3 surrogate supervision schemes, namely rotation, reconstruction, and colorization, in 4 different medical imaging applications representing classification and segmentation for both 2D and 3D medical images. 3 key findings emerge from our research: 1) pre-training with surrogate supervision is effective for small training sets; 2) deep models trained from initial weights pre-trained through surrogate supervision outperform the same models when trained from scratch, suggesting that pre-training with surrogate supervision should be considered prior to training any deep 3D models; 3) pre-training models in the medical domain with surrogate supervision is more effective than transfer learning from an unrelated domain (e.g., natural images), indicating the practical value of abundant unlabeled medical image data."
                    },
                    {
                        "title": "LightSpeed: Light and Fast Neural Light Fields on Mobile Devices",
                        "abstract": "Real-time novel-view image synthesis on mobile devices is prohibitive due to the limited computational power and storage. Using volumetric rendering methods, such as NeRF and its derivatives, on mobile devices is not suitable due to the high computational cost of volumetric rendering. On the other hand, recent advances in neural light field representations have shown promising real-time view synthesis results on mobile devices. Neural light field methods learn a direct mapping from a ray representation to the pixel color. The current choice of ray representation is either stratified ray sampling or Plucker coordinates, overlooking the classic light slab (two-plane) representation, the preferred representation to interpolate between light field views. In this work, we find that using the light slab representation is an efficient representation for learning a neural light field. More importantly, it is a lower-dimensional ray representation enabling us to learn the 4D ray space using feature grids which are significantly faster to train and render. Although mostly designed for frontal views, we show that the light-slab representation can be further extended to non-frontal scenes using a divide-and-conquer strategy. Our method offers superior rendering quality compared to previous light field methods and achieves a significantly improved trade-off between rendering quality and speed."
                    },
                    {
                        "title": "Real-Time Neural Light Field on Mobile Devices",
                        "abstract": "Recent efforts in Neural Rendering Fields (NeRF) have shown impressive results on novel view synthesis by utilizing implicit neural representation to represent 3D scenes. Due to the process of volumetric rendering, the inference speed for NeRF is extremely slow, limiting the application scenarios of utilizing NeRF on resource-constrained hardware, such as mobile devices. Many works have been conducted to reduce the latency of running NeRF models. However, most of them still require high-end GPU for acceleration or extra storage memory, which is all unavailable on mobile devices. Another emerging direction utilizes the neural light field (NeLF) for speedup, as only one forward pass is performed on a ray to predict the pixel color. Nevertheless, to reach a similar rendering quality as NeRF, the network in NeLF is designed with intensive computation, which is not mobile-friendly. In this work, we propose an efficient network that runs in real-time on mobile devices for neural rendering. We follow the setting of NeLF to train our network. Unlike existing works, we introduce a novel network architecture that runs efficiently on mobile devices with low latency and small size, i.e., saving $15\\times \\sim 24\\times$ storage compared with MobileNeRF. Our model achieves high-resolution generation while maintaining real-time inference for both synthetic and real-world scenes on mobile devices, e.g., $18.04$ms (iPhone 13) for rendering one $1008\\times756$ image of real 3D scenes. Additionally, we achieve similar image quality as NeRF and better quality than MobileNeRF (PSNR $26.15$ vs. $25.91$ on the real-world forward-facing dataset)."
                    },
                    {
                        "title": "AToM: Amortized Text-to-Mesh using 2D Diffusion",
                        "abstract": "We introduce Amortized Text-to-Mesh (AToM), a feed-forward text-to-mesh framework optimized across multiple text prompts simultaneously. In contrast to existing text-to-3D methods that often entail time-consuming per-prompt optimization and commonly output representations other than polygonal meshes, AToM directly generates high-quality textured meshes in less than 1 second with around 10 times reduction in the training cost, and generalizes to unseen prompts. Our key idea is a novel triplane-based text-to-mesh architecture with a two-stage amortized optimization strategy that ensures stable training and enables scalability. Through extensive experiments on various prompt benchmarks, AToM significantly outperforms state-of-the-art amortized approaches with over 4 times higher accuracy (in DF415 dataset) and produces more distinguishable and higher-quality 3D outputs. AToM demonstrates strong generalizability, offering finegrained 3D assets for unseen interpolated prompts without further optimization during inference, unlike per-prompt solutions."
                    },
                    {
                        "title": "SF-V: Single Forward Video Generation Model",
                        "abstract": "Diffusion-based video generation models have demonstrated remarkable success in obtaining high-fidelity videos through the iterative denoising process. However, these models require multiple denoising steps during sampling, resulting in high computational costs. In this work, we propose a novel approach to obtain single-step video generation models by leveraging adversarial training to fine-tune pre-trained video diffusion models. We show that, through the adversarial training, the multi-steps video diffusion model, i.e., Stable Video Diffusion (SVD), can be trained to perform single forward pass to synthesize high-quality videos, capturing both temporal and spatial dependencies in the video data. Extensive experiments demonstrate that our method achieves competitive generation quality of synthesized videos with significantly reduced computational overhead for the denoising process (i.e., around $23\\times$ speedup compared with SVD and $6\\times$ speedup compared with existing works, with even better generation quality), paving the way for real-time video synthesis and editing. More visualization results are made publicly available at https://snap-research.github.io/SF-V."
                    },
                    {
                        "title": "4Real: Towards Photorealistic 4D Scene Generation via Video Diffusion Models",
                        "abstract": "Existing dynamic scene generation methods mostly rely on distilling knowledge from pre-trained 3D generative models, which are typically fine-tuned on synthetic object datasets. As a result, the generated scenes are often object-centric and lack photorealism. To address these limitations, we introduce a novel pipeline designed for photorealistic text-to-4D scene generation, discarding the dependency on multi-view generative models and instead fully utilizing video generative models trained on diverse real-world datasets. Our method begins by generating a reference video using the video generation model. We then learn the canonical 3D representation of the video using a freeze-time video, delicately generated from the reference video. To handle inconsistencies in the freeze-time video, we jointly learn a per-frame deformation to model these imperfections. We then learn the temporal deformation based on the canonical representation to capture dynamic interactions in the reference video. The pipeline facilitates the generation of dynamic scenes with enhanced photorealism and structural integrity, viewable from multiple perspectives, thereby setting a new standard in 4D scene generation."
                    },
                    {
                        "title": "ControlMM: Controllable Masked Motion Generation",
                        "abstract": "Recent advances in motion diffusion models have enabled spatially controllable text-to-motion generation. However, despite achieving acceptable control precision, these models suffer from generation speed and fidelity limitations. To address these challenges, we propose ControlMM, a novel approach incorporating spatial control signals into the generative masked motion model. ControlMM achieves real-time, high-fidelity, and high-precision controllable motion generation simultaneously. Our approach introduces two key innovations. First, we propose masked consistency modeling, which ensures high-fidelity motion generation via random masking and reconstruction, while minimizing the inconsistency between the input control signals and the extracted control signals from the generated motion. To further enhance control precision, we introduce inference-time logit editing, which manipulates the predicted conditional motion distribution so that the generated motion, sampled from the adjusted distribution, closely adheres to the input control signals. During inference, ControlMM enables parallel and iterative decoding of multiple motion tokens, allowing for high-speed motion generation. Extensive experiments show that, compared to the state of the art, ControlMM delivers superior results in motion quality, with better FID scores (0.061 vs 0.271), and higher control precision (average error 0.0091 vs 0.0108). ControlMM generates motions 20 times faster than diffusion-based methods. Additionally, ControlMM unlocks diverse applications such as any joint any frame control, body part timeline control, and obstacle avoidance. Video visualization can be found at https://exitudio.github.io/ControlMM-page"
                    },
                    {
                        "title": "Deep learning model trained on mobile phone-acquired frozen section images effectively detects basal cell carcinoma",
                        "abstract": "Background: Margin assessment of basal cell carcinoma using the frozen section is a common task of pathology intraoperative consultation. Although frequently straight-forward, the determination of the presence or absence of basal cell carcinoma on the tissue sections can sometimes be challenging. We explore if a deep learning model trained on mobile phone-acquired frozen section images can have adequate performance for future deployment. Materials and Methods: One thousand two hundred and forty-one (1241) images of frozen sections performed for basal cell carcinoma margin status were acquired using mobile phones. The photos were taken at 100x magnification (10x objective). The images were downscaled from a 4032 x 3024 pixel resolution to 576 x 432 pixel resolution. Semantic segmentation algorithm Deeplab V3 with Xception backbone was used for model training. Results: The model uses an image as input and produces a 2-dimensional black and white output of prediction of the same dimension; the areas determined to be basal cell carcinoma were displayed with white color, in a black background. Any output with the number of white pixels exceeding 0.5% of the total number of pixels is deemed positive for basal cell carcinoma. On the test set, the model achieves area under curve of 0.99 for receiver operator curve and 0.97 for precision-recall curve at the pixel level. The accuracy of classification at the slide level is 96%. Conclusions: The deep learning model trained with mobile phone images shows satisfactory performance characteristics, and thus demonstrates the potential for deploying as a mobile phone app to assist in frozen section interpretation in real time."
                    }
                ]
            },
            "f7524a57-a1c0-4add-9aea-9f0b5684f658": {
                "pk": "f7524a57-a1c0-4add-9aea-9f0b5684f658",
                "name": "Ju Hu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "39f50a3a-293b-41cc-84f6-bbcd4d18f8ad": {
                "pk": "39f50a3a-293b-41cc-84f6-bbcd4d18f8ad",
                "name": "Dhritiman Sagar",
                "collaborators": [
                    "Yanyu Li",
                    "Xian Liu",
                    "Anil Kag",
                    "Ju Hu",
                    "Yerlan Idelbayev",
                    "Yanzhi Wang",
                    "Sergey Tulyakov",
                    "Jian Ren"
                ],
                "domain": [
                    "Text-to-Image Generation",
                    "Diffusion Models",
                    "Fine-tuning",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "TextCraftor: Your Text Encoder Can be Image Quality Controller",
                        "abstract": "Diffusion-based text-to-image generative models, e.g., Stable Diffusion, have revolutionized the field of content generation, enabling significant advancements in areas like image editing and video synthesis. Despite their formidable capabilities, these models are not without their limitations. It is still challenging to synthesize an image that aligns well with the input text, and multiple runs with carefully crafted prompts are required to achieve satisfactory results. To mitigate these limitations, numerous studies have endeavored to fine-tune the pre-trained diffusion models, i.e., UNet, utilizing various technologies. Yet, amidst these efforts, a pivotal question of text-to-image diffusion model training has remained largely unexplored: Is it possible and feasible to fine-tune the text encoder to improve the performance of text-to-image diffusion models? Our findings reveal that, instead of replacing the CLIP text encoder used in Stable Diffusion with other large language models, we can enhance it through our proposed fine-tuning approach, TextCraftor, leading to substantial improvements in quantitative benchmarks and human assessments. Interestingly, our technique also empowers controllable image generation through the interpolation of different text encoders fine-tuned with various rewards. We also demonstrate that TextCraftor is orthogonal to UNet finetuning, and can be combined to further improve generative quality."
                    }
                ]
            },
            "3da6ccc5-8e9e-4075-aa7d-446250bbd0c9": {
                "pk": "3da6ccc5-8e9e-4075-aa7d-446250bbd0c9",
                "name": "Bo Yuan",
                "collaborators": [
                    "Keshab K. Parhi",
                    "Ruoshi Yuan",
                    "Yian Ma",
                    "Ping Ao",
                    "Danpei Zhao",
                    "Qinsheng Zhang",
                    "Yongxin Chen",
                    "Xinan Wang",
                    "Yao Jiang",
                    "Keren Fu"
                ],
                "domain": [
                    "Continual Learning",
                    "Light Field Processing",
                    "Polar Codes",
                    "Robotics"
                ],
                "publications": [
                    {
                        "title": "A Survey on Continual Semantic Segmentation: Theory, Challenge, Method and Application",
                        "abstract": "Continual learning, also known as incremental learning or life-long learning, stands at the forefront of deep learning and AI systems. It breaks through the obstacle of one-way training on close sets and enables continuous adaptive learning on open-set conditions. In the recent decade, continual learning has been explored and applied in multiple fields especially in computer vision covering classification, detection and segmentation tasks. Continual semantic segmentation (CSS), of which the dense prediction peculiarity makes it a challenging, intricate and burgeoning task. In this paper, we present a review of CSS, committing to building a comprehensive survey on problem formulations, primary challenges, universal datasets, neoteric theories and multifarious applications. Concretely, we begin by elucidating the problem definitions and primary challenges. Based on an in-depth investigation of relevant approaches, we sort out and categorize current CSS models into two main branches including data-replay and data-free sets. In each branch, the corresponding approaches are similarity-based clustered and thoroughly analyzed, following qualitative comparison and quantitative reproductions on relevant datasets. Besides, we also introduce four CSS specialities with diverse application scenarios and development tendencies. Furthermore, we develop a benchmark for CSS encompassing representative references, evaluation results and reproductions, which is available at~\\url{https://github.com/YBIO/SurveyCSS}. We hope this survey can serve as a reference-worthy and stimulating contribution to the advancement of the life-long learning field, while also providing valuable perspectives for related fields."
                    },
                    {
                        "title": "An optimal control approach to particle filtering on Lie groups",
                        "abstract": "We study the filtering problem over a Lie group that plays an important role in robotics and aerospace applications. We present a new particle filtering algorithm based on stochastic control. In particular, our algorithm is based on a duality between smoothing and optimal control. Leveraging this duality, we reformulate the smoothing problem into an optimal control problem, and by approximately solving it (using, e.g., iLQR) we establish a superior proposal for particle smoothing. Combining it with a suitably designed sliding window mechanism, we obtain a particle filtering algorithm that suffers less from sample degeneracy compared with existing methods. The efficacy of our algorithm is illustrated by a filtering problem over SO(3) for satellite attitude estimation."
                    },
                    {
                        "title": "Constructive Proof of Global Lyapunov Function as Potential Function",
                        "abstract": "We provide a constructive proof on the equivalence of two fundamental concepts: the global Lyapunov function in engineering and the potential function in physics, establishing a bridge between these distinct fields. This result suggests new approaches on the significant unsolved problem namely to construct Lyapunov functions for general nonlinear systems through the analogy with existing methods on potential functions. In addition, we show another connection that the Lyapunov equation is a reduced form of the generalized Einstein relation for linear systems."
                    },
                    {
                        "title": "Exact Mapping Noisy van der Pol Type Oscillator onto Quasi-symplectic Dynamics",
                        "abstract": "We find exact mappings for a class of limit cycle systems with noise onto quasi-symplectic dynamics, including a van der Pol type oscillator. A dual role potential function is obtained as a component of the quasi-symplectic dynamics. Based on a stochastic interpretation different from the traditional Ito's and Stratonovich's, we show the corresponding steady state distribution is the familiar Boltzmann-Gibbs type for arbitrary noise strength. The result provides a new angle for understanding processes without detailed balance and can be verified by experiments."
                    },
                    {
                        "title": "Successive Cancellation Decoding of Polar Codes using Stochastic Computing",
                        "abstract": "Polar codes have emerged as the most favorable channel codes for their unique capacity-achieving property. To date, numerous works have been reported for efficient design of polar codes decoder. However, these prior efforts focused on design of polar decoders via deterministic computation, while the behavior of stochastic polar decoder, which can have potential advantages such as low complexity and strong error-resilience, has not been studied in existing literatures. This paper, for the first time, investigates polar decoding using stochastic logic. Specifically, the commonly-used successive cancellation (SC) algorithm is reformulated into the stochastic form. Several methods that can potentially improve decoding performance are discussed and analyzed. Simulation results show that a stochastic SC decoder can achieve similar error-correcting performance as its deterministic counterpart. This work can pave the way for future hardware design of stochastic polar codes decoders."
                    },
                    {
                        "title": "Successive Cancellation List Polar Decoder using Log-likelihood Ratios",
                        "abstract": "Successive cancellation list (SCL) decoding algorithm is a powerful method that can help polar codes achieve excellent error-correcting performance. However, the current SCL algorithm and decoders are based on likelihood or log-likelihood forms, which render high hardware complexity. In this paper, we propose a log-likelihood-ratio (LLR)-based SCL (LLR-SCL) decoding algorithm, which only needs half the computation and storage complexity than the conventional one. Then, based on the proposed algorithm, we develop low-complexity VLSI architectures for LLR-SCL decoders. Analysis results show that the proposed LLR-SCL decoder achieves 50% reduction in hardware and 98% improvement in hardware efficiency."
                    },
                    {
                        "title": "Algorithm and Architecture for Hybrid Decoding of Polar Codes",
                        "abstract": "Polar codes are the first provable capacity-achieving forward error correction (FEC) codes. In general polar codes can be decoded via either successive cancellation (SC) or belief propagation (BP) decoding algorithm. However, to date practical applications of polar codes have been hindered by the long decoding latency and limited error-correcting performance problems. In this paper, based on our recent proposed early stopping criteria for the BP algorithm, we propose a hybrid BP-SC decoding scheme to improve the decoding performance of polar codes with relatively short latency. Simulation results show that, for (1024, 512) polar codes the proposed approach leads to at least 0.2dB gain over the BP algorithm with the same maximum number of iterations for the entire SNR region, and also achieves 0.2dB decoding gain over the BP algorithm with the same worst-case latency in the high SNR region. Besides, compared to the SC algorithm, the proposed scheme leads to 0.2dB gain in the medium SNR region with much less average decoding latency. In addition, we also propose the low-complexity unified hardware architecture for the hybrid decoding scheme, which is able to implement SC and BP algorithms using same hardware."
                    },
                    {
                        "title": "Guided Focal Stack Refinement Network for Light Field Salient Object Detection",
                        "abstract": "Light field salient object detection (SOD) is an emerging research direction attributed to the richness of light field data. However, most existing methods lack effective handling of focal stacks, therefore making the latter involved in a lot of interfering information and degrade the performance of SOD. To address this limitation, we propose to utilize multi-modal features to refine focal stacks in a guided manner, resulting in a novel guided focal stack refinement network called GFRNet. To this end, we propose a guided refinement and fusion module (GRFM) to refine focal stacks and aggregate multi-modal features. In GRFM, all-in-focus (AiF) and depth modalities are utilized to refine focal stacks separately, leading to two novel sub-modules for different modalities, namely AiF-based refinement module (ARM) and depth-based refinement module (DRM). Such refinement modules enhance structural and positional information of salient objects in focal stacks, and are able to improve SOD accuracy. Experimental results on four benchmark datasets demonstrate the superiority of our GFRNet model against 12 state-of-the-art models."
                    }
                ]
            },
            "7b97a2e2-9820-4e9c-b5c0-86949fa29ff5": {
                "pk": "7b97a2e2-9820-4e9c-b5c0-86949fa29ff5",
                "name": "Sergey Tulyakov",
                "collaborators": [
                    "Aliaksandr Siarohin",
                    "St\u00e9phane Lathuili\u00e8re",
                    "Elisa Ricci",
                    "Jian Ren",
                    "Ivan Skorokhodov",
                    "Willi Menapace",
                    "Nicu Sebe",
                    "Menglei Chai",
                    "Linjie Luo",
                    "Kyle Olszewski"
                ],
                "domain": [
                    "Computer Vision",
                    "Generative Models",
                    "Video Synthesis",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Hybrid VAE: Improving Deep Generative Models using Partial Observations",
                        "abstract": "Deep neural network models trained on large labeled datasets are the state-of-the-art in a large variety of computer vision tasks. In many applications, however, labeled data is expensive to obtain or requires a time consuming manual annotation process. In contrast, unlabeled data is often abundant and available in large quantities. We present a principled framework to capitalize on unlabeled data by training deep generative models on both labeled and unlabeled data. We show that such a combination is beneficial because the unlabeled data acts as a data-driven form of regularization, allowing generative models trained on few labeled samples to reach the performance of fully-supervised generative models trained on much larger datasets. We call our method Hybrid VAE (H-VAE) as it contains both the generative and the discriminative parts. We validate H-VAE on three large-scale datasets of different modalities: two face datasets: (MultiPIE, CelebA) and a hand pose dataset (NYU Hand Pose). Our qualitative visualizations further support improvements achieved by using partial observations."
                    },
                    {
                        "title": "MoCoGAN: Decomposing Motion and Content for Video Generation",
                        "abstract": "Visual signals in a video can be divided into content and motion. While content specifies which objects are in the video, motion describes their dynamics. Based on this prior, we propose the Motion and Content decomposed Generative Adversarial Network (MoCoGAN) framework for video generation. The proposed framework generates a video by mapping a sequence of random vectors to a sequence of video frames. Each random vector consists of a content part and a motion part. While the content part is kept fixed, the motion part is realized as a stochastic process. To learn motion and content decomposition in an unsupervised manner, we introduce a novel adversarial learning scheme utilizing both image and video discriminators. Extensive experimental results on several challenging datasets with qualitative and quantitative comparison to the state-of-the-art approaches, verify effectiveness of the proposed framework. In addition, we show that MoCoGAN allows one to generate videos with same content but different motion as well as videos with different content and same motion."
                    },
                    {
                        "title": "3D Guided Fine-Grained Face Manipulation",
                        "abstract": "We present a method for fine-grained face manipulation. Given a face image with an arbitrary expression, our method can synthesize another arbitrary expression by the same person. This is achieved by first fitting a 3D face model and then disentangling the face into a texture and a shape. We then learn different networks in these two spaces. In the texture space, we use a conditional generative network to change the appearance, and carefully design input formats and loss functions to achieve the best results. In the shape space, we use a fully connected network to predict the accurate shapes and use the available depth data for supervision. Both networks are conditioned on expression coefficients rather than discrete labels, allowing us to generate an unlimited amount of expressions. We show the superiority of this disentangling approach through both quantitative and qualitative studies. In a user study, our method is preferred in 85% of cases when compared to the most recent work. When compared to the ground truth, annotators cannot reliably distinguish between our synthesized images and real images, preferring our method in 53% of the cases."
                    },
                    {
                        "title": "Neural Hair Rendering",
                        "abstract": "In this paper, we propose a generic neural-based hair rendering pipeline that can synthesize photo-realistic images from virtual 3D hair models. Unlike existing supervised translation methods that require model-level similarity to preserve consistent structure representation for both real images and fake renderings, our method adopts an unsupervised solution to work on arbitrary hair models. The key component of our method is a shared latent space to encode appearance-invariant structure information of both domains, which generates realistic renderings conditioned by extra appearance inputs. This is achieved by domain-specific pre-disentangled structure representation, partially shared domain encoder layers and a structure discriminator. We also propose a simple yet effective temporal conditioning method to enforce consistency for video sequence generation. We demonstrate the superiority of our method by testing it on a large number of portraits and comparing it with alternative baselines and state-of-the-art unsupervised image translation methods."
                    },
                    {
                        "title": "StyleGAN-V: A Continuous Video Generator with the Price, Image Quality and Perks of StyleGAN2",
                        "abstract": "Videos show continuous events, yet most $-$ if not all $-$ video synthesis frameworks treat them discretely in time. In this work, we think of videos of what they should be $-$ time-continuous signals, and extend the paradigm of neural representations to build a continuous-time video generator. For this, we first design continuous motion representations through the lens of positional embeddings. Then, we explore the question of training on very sparse videos and demonstrate that a good generator can be learned by using as few as 2 frames per clip. After that, we rethink the traditional image + video discriminators pair and design a holistic discriminator that aggregates temporal information by simply concatenating frames' features. This decreases the training cost and provides richer learning signal to the generator, making it possible to train directly on 1024$^2$ videos for the first time. We build our model on top of StyleGAN2 and it is just ${\\approx}5\\%$ more expensive to train at the same resolution while achieving almost the same image quality. Moreover, our latent space features similar properties, enabling spatial manipulations that our method can propagate in time. We can generate arbitrarily long videos at arbitrary high frame rate, while prior work struggles to generate even 64 frames at a fixed rate. Our model is tested on four modern 256$^2$ and one 1024$^2$-resolution video synthesis benchmarks. In terms of sheer metrics, it performs on average ${\\approx}30\\%$ better than the closest runner-up. Project website: https://universome.github.io."
                    },
                    {
                        "title": "EpiGRAF: Rethinking training of 3D GANs",
                        "abstract": "A very recent trend in generative modeling is building 3D-aware generators from 2D image collections. To induce the 3D bias, such models typically rely on volumetric rendering, which is expensive to employ at high resolutions. During the past months, there appeared more than 10 works that address this scaling issue by training a separate 2D decoder to upsample a low-resolution image (or a feature tensor) produced from a pure 3D generator. But this solution comes at a cost: not only does it break multi-view consistency (i.e. shape and texture change when the camera moves), but it also learns the geometry in a low fidelity. In this work, we show that it is possible to obtain a high-resolution 3D generator with SotA image quality by following a completely different route of simply training the model patch-wise. We revisit and improve this optimization scheme in two ways. First, we design a location- and scale-aware discriminator to work on patches of different proportions and spatial positions. Second, we modify the patch sampling strategy based on an annealed beta distribution to stabilize training and accelerate the convergence. The resulted model, named EpiGRAF, is an efficient, high-resolution, pure 3D generator, and we test it on four datasets (two introduced in this work) at $256^2$ and $512^2$ resolutions. It obtains state-of-the-art image quality, high-fidelity geometry and trains ${\\approx} 2.5 \\times$ faster than the upsampler-based counterparts. Project website: https://universome.github.io/epigraf."
                    },
                    {
                        "title": "Hierarchical Patch Diffusion Models for High-Resolution Video Generation",
                        "abstract": "Diffusion models have demonstrated remarkable performance in image and video synthesis. However, scaling them to high-resolution inputs is challenging and requires restructuring the diffusion pipeline into multiple independent components, limiting scalability and complicating downstream applications. This makes it very efficient during training and unlocks end-to-end optimization on high-resolution videos. We improve PDMs in two principled ways. First, to enforce consistency between patches, we develop deep context fusion -- an architectural technique that propagates the context information from low-scale to high-scale patches in a hierarchical manner. Second, to accelerate training and inference, we propose adaptive computation, which allocates more network capacity and computation towards coarse image details. The resulting model sets a new state-of-the-art FVD score of 66.32 and Inception Score of 87.68 in class-conditional video generation on UCF-101 $256^2$, surpassing recent methods by more than 100%. Then, we show that it can be rapidly fine-tuned from a base $36\\times 64$ low-resolution generator for high-resolution $64 \\times 288 \\times 512$ text-to-video synthesis. To the best of our knowledge, our model is the first diffusion-based architecture which is trained on such high resolutions entirely end-to-end. Project webpage: https://snap-research.github.io/hpdm."
                    },
                    {
                        "title": "Playable Video Generation",
                        "abstract": "This paper introduces the unsupervised learning problem of playable video generation (PVG). In PVG, we aim at allowing a user to control the generated video by selecting a discrete action at every time step as when playing a video game. The difficulty of the task lies both in learning semantically consistent actions and in generating realistic videos conditioned on the user input. We propose a novel framework for PVG that is trained in a self-supervised manner on a large dataset of unlabelled videos. We employ an encoder-decoder architecture where the predicted action labels act as bottleneck. The network is constrained to learn a rich action space using, as main driving loss, a reconstruction loss on the generated video. We demonstrate the effectiveness of the proposed approach on several datasets with wide environment variety. Further details, code and examples are available on our project page willi-menapace.github.io/playable-video-generation-website."
                    },
                    {
                        "title": "Task-Assisted Domain Adaptation with Anchor Tasks",
                        "abstract": "Some tasks, such as surface normals or single-view depth estimation, require per-pixel ground truth that is difficult to obtain on real images but easy to obtain on synthetic. However, models learned on synthetic images often do not generalize well to real images due to the domain shift. Our key idea to improve domain adaptation is to introduce a separate anchor task (such as facial landmarks) whose annotations can be obtained at no cost or are already available on both synthetic and real datasets. To further leverage the implicit relationship between the anchor and main tasks, we apply our \\freeze technique that learns the cross-task guidance on the source domain with the final network layers, and use it on the target domain. We evaluate our methods on surface normal estimation on two pairs of datasets (indoor scenes and faces) with two kinds of anchor tasks (semantic segmentation and facial landmarks). We show that blindly applying domain adaptation or training the auxiliary task on only one domain may hurt performance, while using anchor tasks on both domains is better behaved. Our \\freeze technique outperforms competing approaches, reaching performance in facial images on par with a recently popular surface normal estimation method using shape from shading domain knowledge."
                    },
                    {
                        "title": "Control-NeRF: Editable Feature Volumes for Scene Rendering and Manipulation",
                        "abstract": "We present a novel method for performing flexible, 3D-aware image content manipulation while enabling high-quality novel view synthesis. While NeRF-based approaches are effective for novel view synthesis, such models memorize the radiance for every point in a scene within a neural network. Since these models are scene-specific and lack a 3D scene representation, classical editing such as shape manipulation, or combining scenes is not possible. Hence, editing and combining NeRF-based scenes has not been demonstrated. With the aim of obtaining interpretable and controllable scene representations, our model couples learnt scene-specific feature volumes with a scene agnostic neural rendering network. With this hybrid representation, we decouple neural rendering from scene-specific geometry and appearance. We can generalize to novel scenes by optimizing only the scene-specific 3D feature representation, while keeping the parameters of the rendering network fixed. The rendering function learnt during the initial training stage can thus be easily applied to new scenes, making our approach more flexible. More importantly, since the feature volumes are independent of the rendering model, we can manipulate and combine scenes by editing their corresponding feature volumes. The edited volume can then be plugged into the rendering model to synthesize high-quality novel views. We demonstrate various scene manipulations, including mixing scenes, deforming objects and inserting objects into scenes, while still producing photo-realistic results."
                    },
                    {
                        "title": "Affection: Learning Affective Explanations for Real-World Visual Data",
                        "abstract": "In this work, we explore the emotional reactions that real-world images tend to induce by using natural language as the medium to express the rationale behind an affective response to a given visual stimulus. To embark on this journey, we introduce and share with the research community a large-scale dataset that contains emotional reactions and free-form textual explanations for 85,007 publicly available images, analyzed by 6,283 annotators who were asked to indicate and explain how and why they felt in a particular way when observing a specific image, producing a total of 526,749 responses. Even though emotional reactions are subjective and sensitive to context (personal mood, social status, past experiences) - we show that there is significant common ground to capture potentially plausible emotional responses with a large support in the subject population. In light of this crucial observation, we ask the following questions: i) Can we develop multi-modal neural networks that provide reasonable affective responses to real-world visual data, explained with language? ii) Can we steer such methods towards producing explanations with varying degrees of pragmatic language or justifying different emotional reactions while adapting to the underlying visual stimulus? Finally, iii) How can we evaluate the performance of such methods for this novel task? With this work, we take the first steps in addressing all of these questions, thus paving the way for richer, more human-centric, and emotionally-aware image analysis systems. Our introduced dataset and all developed methods are available on https://affective-explanations.org"
                    },
                    {
                        "title": "MyVLM: Personalizing VLMs for User-Specific Queries",
                        "abstract": "Recent large-scale vision-language models (VLMs) have demonstrated remarkable capabilities in understanding and generating textual descriptions for visual content. However, these models lack an understanding of user-specific concepts. In this work, we take a first step toward the personalization of VLMs, enabling them to learn and reason over user-provided concepts. For example, we explore whether these models can learn to recognize you in an image and communicate what you are doing, tailoring the model to reflect your personal experiences and relationships. To effectively recognize a variety of user-specific concepts, we augment the VLM with external concept heads that function as toggles for the model, enabling the VLM to identify the presence of specific target concepts in a given image. Having recognized the concept, we learn a new concept embedding in the intermediate feature space of the VLM. This embedding is tasked with guiding the language model to naturally integrate the target concept in its generated response. We apply our technique to BLIP-2 and LLaVA for personalized image captioning and further show its applicability for personalized visual question-answering. Our experiments demonstrate our ability to generalize to unseen images of learned concepts while preserving the model behavior on unrelated inputs."
                    },
                    {
                        "title": "Train One Get One Free: Partially Supervised Neural Network for Bug Report Duplicate Detection and Clustering",
                        "abstract": "Tracking user reported bugs requires considerable engineering effort in going through many repetitive reports and assigning them to the correct teams. This paper proposes a neural architecture that can jointly (1) detect if two bug reports are duplicates, and (2) aggregate them into latent topics. Leveraging the assumption that learning the topic of a bug is a sub-task for detecting duplicates, we design a loss function that can jointly perform both tasks but needs supervision for only duplicate classification, achieving topic clustering in an unsupervised fashion. We use a two-step attention module that uses self-attention for topic clustering and conditional attention for duplicate detection. We study the characteristics of two types of real world datasets that have been marked for duplicate bugs by engineers and by non-technical annotators. The results demonstrate that our model not only can outperform state-of-the-art methods for duplicate classification on both cases, but can also learn meaningful latent clusters without additional supervision."
                    },
                    {
                        "title": "Transformable Bottleneck Networks",
                        "abstract": "We propose a novel approach to performing fine-grained 3D manipulation of image content via a convolutional neural network, which we call the Transformable Bottleneck Network (TBN). It applies given spatial transformations directly to a volumetric bottleneck within our encoder-bottleneck-decoder architecture. Multi-view supervision encourages the network to learn to spatially disentangle the feature space within the bottleneck. The resulting spatial structure can be manipulated with arbitrary spatial transformations. We demonstrate the efficacy of TBNs for novel view synthesis, achieving state-of-the-art results on a challenging benchmark. We demonstrate that the bottlenecks produced by networks trained for this task contain meaningful spatial structure that allows us to intuitively perform a variety of image manipulations in 3D, well beyond the rigid transformations seen during training. These manipulations include non-uniform scaling, non-rigid warping, and combining content from different images. Finally, we extract explicit 3D structure from the bottleneck, performing impressive 3D reconstruction from a single input image."
                    },
                    {
                        "title": "Animating Arbitrary Objects via Deep Motion Transfer",
                        "abstract": "This paper introduces a novel deep learning framework for image animation. Given an input image with a target object and a driving video sequence depicting a moving object, our framework generates a video in which the target object is animated according to the driving sequence. This is achieved through a deep architecture that decouples appearance and motion information. Our framework consists of three main modules: (i) a Keypoint Detector unsupervisely trained to extract object keypoints, (ii) a Dense Motion prediction network for generating dense heatmaps from sparse keypoints, in order to better encode motion information and (iii) a Motion Transfer Network, which uses the motion heatmaps and appearance information extracted from the input image to synthesize the output frames. We demonstrate the effectiveness of our method on several benchmark datasets, spanning a wide variety of object appearances, and show that our approach outperforms state-of-the-art image animation and video generation methods. Our source code is publicly available."
                    },
                    {
                        "title": "First Order Motion Model for Image Animation",
                        "abstract": "Image animation consists of generating a video sequence so that an object in a source image is animated according to the motion of a driving video. Our framework addresses this problem without using any annotation or prior information about the specific object to animate. Once trained on a set of videos depicting objects of the same category (e.g. faces, human bodies), our method can be applied to any object of this class. To achieve this, we decouple appearance and motion information using a self-supervised formulation. To support complex motions, we use a representation consisting of a set of learned keypoints along with their local affine transformations. A generator network models occlusions arising during target motions and combines the appearance extracted from the source image and the motion derived from the driving video. Our framework scores best on diverse benchmarks and on a variety of object categories. Our source code is publicly available."
                    },
                    {
                        "title": "Motion-supervised Co-Part Segmentation",
                        "abstract": "Recent co-part segmentation methods mostly operate in a supervised learning setting, which requires a large amount of annotated data for training. To overcome this limitation, we propose a self-supervised deep learning method for co-part segmentation. Differently from previous works, our approach develops the idea that motion information inferred from videos can be leveraged to discover meaningful object parts. To this end, our method relies on pairs of frames sampled from the same video. The network learns to predict part segments together with a representation of the motion between two frames, which permits reconstruction of the target image. Through extensive experimental evaluation on publicly available video sequences we demonstrate that our approach can produce improved segmentation maps with respect to previous self-supervised co-part segmentation approaches."
                    },
                    {
                        "title": "Playable Environments: Video Manipulation in Space and Time",
                        "abstract": "We present Playable Environments - a new representation for interactive video generation and manipulation in space and time. With a single image at inference time, our novel framework allows the user to move objects in 3D while generating a video by providing a sequence of desired actions. The actions are learnt in an unsupervised manner. The camera can be controlled to get the desired viewpoint. Our method builds an environment state for each frame, which can be manipulated by our proposed action module and decoded back to the image space with volumetric rendering. To support diverse appearances of objects, we extend neural radiance fields with style-based modulation. Our method trains on a collection of various monocular videos requiring only the estimated camera parameters and 2D object locations. To set a challenging benchmark, we introduce two large scale video datasets with significant camera movements. As evidenced by our experiments, playable environments enable several creative applications not attainable by prior video synthesis works, including playable 3D video generation, stylization and manipulation. Further details, code and examples are available at https://willi-menapace.github.io/playable-environments-website"
                    },
                    {
                        "title": "SMIL: Multimodal Learning with Severely Missing Modality",
                        "abstract": "A common assumption in multimodal learning is the completeness of training data, i.e., full modalities are available in all training examples. Although there exists research endeavor in developing novel methods to tackle the incompleteness of testing data, e.g., modalities are partially missing in testing examples, few of them can handle incomplete training modalities. The problem becomes even more challenging if considering the case of severely missing, e.g., 90% training examples may have incomplete modalities. For the first time in the literature, this paper formally studies multimodal learning with missing modality in terms of flexibility (missing modalities in training, testing, or both) and efficiency (most training data have incomplete modality). Technically, we propose a new method named SMIL that leverages Bayesian meta-learning in uniformly achieving both objectives. To validate our idea, we conduct a series of experiments on three popular benchmarks: MM-IMDb, CMU-MOSI, and avMNIST. The results prove the state-of-the-art performance of SMIL over existing methods and generative baselines including autoencoders and generative adversarial networks. Our code is available at https://github.com/mengmenm/SMIL."
                    },
                    {
                        "title": "Motion Representations for Articulated Animation",
                        "abstract": "We propose novel motion representations for animating articulated objects consisting of distinct parts. In a completely unsupervised manner, our method identifies object parts, tracks them in a driving video, and infers their motions by considering their principal axes. In contrast to the previous keypoint-based works, our method extracts meaningful and consistent regions, describing locations, shape, and pose. The regions correspond to semantically relevant and distinct object parts, that are more easily detected in frames of the driving video. To force decoupling of foreground from background, we model non-object related global motion with an additional affine transformation. To facilitate animation and prevent the leakage of the shape of the driving object, we disentangle shape and pose of objects in the region space. Our model can animate a variety of objects, surpassing previous methods by a large margin on existing benchmarks. We present a challenging new benchmark with high-resolution videos and show that the improvement is particularly pronounced when articulated objects are considered, reaching 96.6% user preference vs. the state of the art."
                    }
                ]
            },
            "c01e682b-7086-4fe3-85fc-ad39d2e54f94": {
                "pk": "c01e682b-7086-4fe3-85fc-ad39d2e54f94",
                "name": "Jian Ren",
                "collaborators": [
                    "Kai Zhou",
                    "Jian Li",
                    "Tongtong Li",
                    "Jinqiao Duan",
                    "Wanxing Ding",
                    "Zhongkai Guo",
                    "Xianming Liu",
                    "Xiangjun Wang",
                    "M. H. Afifi",
                    "Kwok Wa Leung"
                ],
                "domain": [
                    "Distributed Storage",
                    "Secure Outsourcing",
                    "Stochastic Systems",
                    "Neural Rendering"
                ],
                "publications": [
                    {
                        "title": "Beyond the MDS Bound in Distributed Cloud Storage",
                        "abstract": "Distributed storage plays a crucial role in the current cloud computing framework. After the theoretical bound for distributed storage was derived by the pioneer work of the regenerating code, Reed-Solomon code based regenerating codes were developed. The RS code based minimum storage regeneration code (RS-MSR) and the minimum bandwidth regeneration code (RS-MBR) can achieve theoretical bounds on the MSR point and the MBR point respectively in code regeneration. They can also maintain the MDS property in code reconstruction. However, in the hostile network where the storage nodes can be compromised and the packets can be tampered with, the storage capacity of the network can be significantly affected. In this paper, we propose a Hermitian code based minimum storage regenerating (H-MSR) code and a minimum bandwidth regenerating (H-MBR) code. We first prove that our proposed Hermitian code based regenerating codes can achieve the theoretical bounds for MSR point and MBR point respectively. We then propose data regeneration and reconstruction algorithms for the H-MSR code and the H-MBR code in both error-free network and hostile network. Theoretical evaluation shows that our proposed schemes can detect the erroneous decodings and correct more errors in hostile network than the RS-MSR code and the RS-MBR code with the same code rate. Our analysis also demonstrates that the proposed H-MSR and H-MBR codes have lower computational complexity than the RS-MSR/RS-MBR codes in both code regeneration and code reconstruction."
                    },
                    {
                        "title": "Optimal Construction of Regenerating Code through Rate-matching in Hostile Networks",
                        "abstract": "Regenerating code is a class of code very suitable for distributed storage systems, which can maintain optimal bandwidth and storage space. Two types of important regenerating code have been constructed: the minimum storage regeneration (MSR) code and the minimum bandwidth regeneration (MBR) code. However, in hostile networks where adversaries can compromise storage nodes, the storage capacity of the network can be significantly affected. In this paper, we propose two optimal constructions of regenerating codes through rate-matching that can combat against this kind of adversaries in hostile networks: 2-layer rate-matched regenerating code and $m$-layer rate-matched regenerating code. For the 2-layer code, we can achieve the optimal storage efficiency for given system requirements. Our comprehensive analysis shows that our code can detect and correct malicious nodes with higher storage efficiency compared to the universally resilient regenerating code which is a straightforward extension of regenerating code with error detection and correction capability. Then we propose the $m$-layer code by extending the 2-layer code and achieve the optimal error correction efficiency by matching the code rate of each layer's regenerating code. We also demonstrate that the optimized parameter can achieve the maximum storage capacity under the same constraint. Compared to the universally resilient regenerating code, our code can achieve much higher error correction efficiency."
                    },
                    {
                        "title": "A parameter estimation method based on random slow manifolds",
                        "abstract": "A parameter estimation method is devised for a slow-fast stochastic dynamical system, where often only the slow component is observable. By using the observations only on the slow component, the system parameters are estimated by working on the slow system on the random slow manifold. This offers a benefit of dimension reduction in quantifying parameters in stochastic dynamical systems. An example is presented to illustrate this method, and verify that the parameter estimator based on the lower dimensional, reduced slow system is a good approximation of the parameter estimator for original slow-fast stochastic dynamical system."
                    },
                    {
                        "title": "CASO: Cost-Aware Secure Outsourcing of General Computational Problems",
                        "abstract": "Computation outsourcing is an integral part of cloud computing. It enables end-users to outsource their computational tasks to the cloud and utilize the shared cloud resources in a pay-per-use manner. However, once the tasks are outsourced, the end-users will lose control of their data, which may result in severe security issues especially when the data is sensitive. To address this problem, secure outsourcing mechanisms have been proposed to ensure security of the end-users' outsourced data. In this paper, we investigate outsourcing of general computational problems which constitute the mathematical basics for problems emerged from various fields such as engineering and finance. To be specific, we propose affine mapping based schemes for the problem transformation and outsourcing so that the cloud is unable to learn any key information from the transformed problem. Meanwhile, the overhead for the transformation is limited to an acceptable level compared to the computational savings introduced by the outsourcing itself. Furthermore, we develop cost-aware schemes to balance the trade-offs between end-users' various security demands and computational overhead. We also propose a verification scheme to ensure that the end-users will always receive a valid solution from the cloud. Our extensive complexity and security analysis show that our proposed Cost-Aware Secure Outsourcing (CASO) scheme is both practical and effective."
                    },
                    {
                        "title": "Security and Privacy Enhancement for Outsourced Biometric Identification",
                        "abstract": "A lot of research has been focused on secure outsourcing of biometric identification in the context of cloud computing. In such schemes, both the encrypted biometric database and the identification process are outsourced to the cloud. The ultimate goal is to protect the security and privacy of the biometric database and the query templates. Security analysis shows that previous schemes suffer from the enrolment attack and unnecessarily expose more information than needed. In this paper, we propose a new secure outsourcing scheme aims at enhancing the security from these two aspects. First, besides all the attacks discussed in previous schemes, our proposed scheme is also secure against the enrolment attack. Second, we model the identification process as a fixed radius similarity query problem instead of the kNN search problem. Such a modelling is able to reduce the exposed information thus enhancing the privacy of the biometric database. Our comprehensive security and complexity analysis show that our scheme is able to enhance the security and privacy of the biometric database and query templates while maintaining the same computational savings from outsourcing."
                    },
                    {
                        "title": "Identifying stochastic governing equations from data of the most probable transition trajectories",
                        "abstract": "Extracting governing stochastic differential equation models from elusive data is crucial to understand and forecast dynamics for complex systems. We devise a method to extract the drift term and estimate the diffusion coefficient of a governing stochastic dynamical system, from its time-series data of the most probable transition trajectory. By the Onsager-Machlup theory, the most probable transition trajectory satisfies the corresponding Euler-Lagrange equation, which is a second order deterministic ordinary differential equation involving the drift term and diffusion coefficient. We first estimate the coefficients of the Euler-Lagrange equation based on the data of the most probable trajectory, and then we calculate the drift and diffusion coefficients of the governing stochastic dynamical system. These two steps involve sparse regression and optimization. Finally, we illustrate our method with an example and some discussions."
                    },
                    {
                        "title": "Exploring the Self-enhanced Mechanism of Interactive Advertising Phenomenon---Based on the Research of Three Cases",
                        "abstract": "Under the background of the new media era with the rapid development of interactive advertising, this paper used case study method based on the summary of the research of the communication effect of interactive advertising from both domestic and foreign academia. This paper divided interactive advertising into three types to examine ---- interactive ads on official website, interactive ads based on SNS and interactive ads based on mobile media. Furthermore, this paper induced and summarized a self-enhanced dissemination mechanism of the interactive advertising, including three parts which are micro level, meso level and macro level mechanism, micro level embodies core interaction, inner interaction and outer interaction which reveal the whole process of interact with contents, with people and with computer, and the communication approach and spread speed shown in meso level which is self-fission-type spread, finally in macro level the communication effect of IA achieved the spiral increasing. In a word, this article enriches research procedure of the interactive advertising communication effects."
                    },
                    {
                        "title": "PassBio: Privacy-Preserving User-Centric Biometric Authentication",
                        "abstract": "The proliferation of online biometric authentication has necessitated security requirements of biometric templates. The existing secure biometric authentication schemes feature a server-centric model, where a service provider maintains a biometric database and is fully responsible for the security of the templates. The end-users have to fully trust the server in storing, processing and managing their private templates. As a result, the end-users' templates could be compromised by outside attackers or even the service provider itself. In this paper, we propose a user-centric biometric authentication scheme (PassBio) that enables end-users to encrypt their own templates with our proposed light-weighted encryption scheme. During authentication, all the templates remain encrypted such that the server will never see them directly. However, the server is able to determine whether the distance of two encrypted templates is within a pre-defined threshold. Our security analysis shows that no critical information of the templates can be revealed under both passive and active attacks. PassBio follows a \"compute-then-compare\" computational model over encrypted data. More specifically, our proposed Threshold Predicate Encryption (TPE) scheme can encrypt two vectors x and y in such a manner that the inner product of x and y can be evaluated and compared to a pre-defined threshold. TPE guarantees that only the comparison result is revealed and no key information about x and y can be learned. Furthermore, we show that TPE can be utilized as a flexible building block to evaluate different distance metrics such as Hamming distance and Euclidean distance over encrypted data. Such a compute-then-compare computational model, enabled by TPE, can be widely applied in many interesting applications such as searching over encrypted data while ensuring data security and privacy."
                    },
                    {
                        "title": "Small noise approximation of center manifolds for stochastic dynamical systems",
                        "abstract": "This paper provides a small noise approximation for local random center manifolds of a class of stochastic dynamical systems in Euclidean space.   An example is presented to illustrate the method."
                    },
                    {
                        "title": "ExpSOS: Secure and Verifiable Outsourcing of Exponentiation Operations for Mobile Cloud Computing",
                        "abstract": "Discrete exponential operation, such as modular exponentiation and scalar multiplication on elliptic curves, is a basic operation of many public-key cryptosystems. However, the exponential operations are considered prohibitively expensive for resource-constrained mobile devices. In this paper, we address the problem of secure outsourcing of exponentiation operations to one single untrusted server. Our proposed scheme (ExpSOS) only requires very limited number of modular multiplications at local mobile environment thus it can achieve impressive computational gain. ExpSOS also provides a secure verification scheme with probability approximately 1 to ensure that the mobile end-users can always receive valid results. The comprehensive analysis as well as the simulation results in real mobile device demonstrates that our proposed ExpSOS can significantly improve the existing schemes in efficiency, security and result verifiability. We apply ExpSOS to securely outsource several cryptographic protocols to show that ExpSOS is widely applicable to many cryptographic computations."
                    },
                    {
                        "title": "Generation of High-Purity Millimeter-Wave Orbital Angular Momentum Modes Using Horn Antenna: Theory and Implementation",
                        "abstract": "Twisted electromagnetic waves, of which the helical phase front is called orbital angular momentum (OAM), have been recently explored for quantum information, high speed communication and radar detections. In this context, generation of high purity waves carrying OAM is of great significance and challenge from low frequency band to optical area. Here, a novel strategy of mode combination method is proposed to generate twisted waves with arbitrary order of OAM index. The higher order mode of a circular horn antenna is used to generate the twisted waves with quite high purity. The proposed strategy is verified with theoretical analysis, numerical simulation and experiments. A circular horn antenna operating at millimeter wave band is designed, fabricated, and measured. Two twisted waves with OAM index of l=+1 and l=-1 with a mode purity as high as 87% are obtained. Compared with the other OAM antennas, the antenna proposed here owns a high antenna gain (over 12 dBi) and wide operating bandwidth (over 15%). The high mode purity, high antenna gain and wide operating band make the antenna suitable for the twisted-wave applications, not only in the microwave and millimeter wave band, but also in the terahertz band."
                    },
                    {
                        "title": "Cylindrical-water-resonator-based ultra-broadband microwave absorber",
                        "abstract": "A cylindrical-water-resonator-based absorber with ultra-broad operating band at microwave band is demonstrated theoretically and experimentally for the first time. By utilizing dielectric resonator (DR) mode and spoof surface plasmon polariton (SPP) mode of the cylindrical water resonators, the proposed absorber owns an absorptivity higher than 90% over almost the whole ultra-broad operating band from 5.58 GHz to 24.21 GHz, with a relative bandwidth as high as 125%. The angular tolerance and thermal stability of the proposed absorber are simulated and the results indicate the good performance of the absorber under wide incident angles and weakly dependent on water temperature. Low cost, ultra-broad operating band, good wide-angle characteristic and thermal stability make the absorber promising in the application of antenna measurement, steady technology and energy harvesting."
                    },
                    {
                        "title": "Neural Hair Rendering",
                        "abstract": "In this paper, we propose a generic neural-based hair rendering pipeline that can synthesize photo-realistic images from virtual 3D hair models. Unlike existing supervised translation methods that require model-level similarity to preserve consistent structure representation for both real images and fake renderings, our method adopts an unsupervised solution to work on arbitrary hair models. The key component of our method is a shared latent space to encode appearance-invariant structure information of both domains, which generates realistic renderings conditioned by extra appearance inputs. This is achieved by domain-specific pre-disentangled structure representation, partially shared domain encoder layers and a structure discriminator. We also propose a simple yet effective temporal conditioning method to enforce consistency for video sequence generation. We demonstrate the superiority of our method by testing it on a large number of portraits and comparing it with alternative baselines and state-of-the-art unsupervised image translation methods."
                    }
                ]
            }
        }
    },
    "2310.07704": {
        "paper_data": {
            "title": "Ferret: Refer and Ground Anything Anywhere at Any Granularity",
            "url": "http://arxiv.org/abs/2310.07704v1",
            "arxiv_id": "2310.07704",
            "authors": [
                "Haoxuan You",
                "Haotian Zhang",
                "Zhe Gan",
                "Xianzhi Du",
                "Bowen Zhang",
                "Zirui Wang",
                "Liangliang Cao",
                "Shih-Fu Chang",
                "Yinfei Yang"
            ],
            "abstract": "We introduce Ferret, a new Multimodal Large Language Model (MLLM) capable of understanding spatial referring of any shape or granularity within an image and accurately grounding open-vocabulary descriptions. To unify referring and grounding in the LLM paradigm, Ferret employs a novel and powerful hybrid region representation that integrates discrete coordinates and continuous features jointly to represent a region in the image. To extract the continuous features of versatile regions, we propose a spatial-aware visual sampler, adept at handling varying sparsity across different shapes. Consequently, Ferret can accept diverse region inputs, such as points, bounding boxes, and free-form shapes. To bolster the desired capability of Ferret, we curate GRIT, a comprehensive refer-and-ground instruction tuning dataset including 1.1M samples that contain rich hierarchical spatial knowledge, with 95K hard negative data to promote model robustness. The resulting model not only achieves superior performance in classical referring and grounding tasks, but also greatly outperforms existing MLLMs in region-based and localization-demanded multimodal chatting. Our evaluations also reveal a significantly improved capability of describing image details and a remarkable alleviation in object hallucination. Code and data will be available at https://github.com/apple/ml-ferret",
            "introduction": "   1 Introduction  In vision-language learning, how to enable spatial understanding of models is a fundamental research problem. Two desired capabilities stem from this problem: referring and grounding. Referring demands that the model can accurately comprehend the semantics of specific given regions\u00a0(Krahmer & Van\u00a0Deemter, 2012; Kazemzadeh et\u00a0al., 2014; Mao et\u00a0al., 2016; Yu et\u00a0al., 2016; Zellers et\u00a0al., 2019), whereas grounding necessitates the model to localize the region in accordance with the given semantic description\u00a0(Luo & Shakhnarovich, 2017; Nagaraja et\u00a0al., 2016; Yu et\u00a0al., 2017; Kamath et\u00a0al., 2021).   Essentially, referring and grounding demand the same type of knowledge: alignment of spatial information and semantics. Despite that, existing works mostly learn referring and grounding individually (Li et\u00a0al., 2022; Wu et\u00a0al., 2022; Yu et\u00a0al., 2017). In comparison, humans can learn from one task and generalize the shared knowledge to the other task effortlessly, and are able to seamlessly integrate referring/grounding capabilities with daily dialogue and reasoning (Zellers et\u00a0al., 2019). Inspired by the above gap, in this paper, we study three main questions: (i\ud835\udc56iitalic_i) How to unify referring and grounding in one framework, and will they benefit each other? (i\u2062i\ud835\udc56\ud835\udc56iiitalic_i italic_i) How to represent versatile types of regions that humans usually use for referring, such as point, box, scribble, and even free-form shapes? (i\u2062i\u2062i\ud835\udc56\ud835\udc56\ud835\udc56iiiitalic_i italic_i italic_i) How to make referring and grounding open-vocabulary, instruction-following, and robust, which are crucial for practical applications?   Targeting these three questions, we introduce Ferret, a novel refer-and-ground Multimodal Large Language Model (MLLM). First of all, we choose MLLM as the bedrock of Ferret due to their powerful vision-language global understanding capability (Zhu et\u00a0al., 2023a; Liu et\u00a0al., 2023b; Li et\u00a0al., 2023c). To unify referring and grounding, Ferret first represents the coordinates of regions in natural language numerical form,111Note that there is no additional vocabulary or position encoders introduced in Ferret model. as illustrated in Figure\u00a03. However, it is inefficient to use single point or box coordinates to represent versatile shapes of regions, such as strokes, scribbles, or complex polygons. These shapes are essential for more universal and precise human-model interaction. To solve this problem, we further propose a spatial-aware visual sampler to acquire the visual features for regions in any shape, taking care of the varying sparsity in those shapes. Then, the discrete coordinates and the continuous visual features are combined together to represent the visual regions in the input, composing a hybrid region representation in Ferret. Equipped with above methods, Ferret can deal with input that mixes referred regions with free-form text, and is able to ground the mentioned objects in its output by seamlessly generating the coordinates for each groundable object along with generating text. To our best knowledge, Ferret is the first work that is able to process free-formed region inputs in MLLMs.   In order to make the refer-and-ground capability in Ferret open-vocabulary, instruction-following, and robust, we collect GRIT, a Ground-and-Refer Instruction-Tuning dataset with 1.1M samples. GRIT contains multiple levels of spatial knowledge, covering objects, relationships, region descriptions, and complex reasoning. It includes both text-in location-out (grounding) and location-in text-out (referring) data, as well as data that mixes location and text in both input and output. The majority of",
            "references": [
                {
                    "title": "The Dawn of LMMs: Preliminary Explorations with GPT-4V(ision)",
                    "abstract": "Large multimodal models (LMMs) extend large language models (LLMs) with multi-sensory skills, such as visual understanding, to achieve stronger generic intelligence. In this paper, we analyze the latest model, GPT-4V(ision), to deepen the understanding of LMMs. The analysis focuses on the intriguing tasks that GPT-4V can perform, containing test samples to probe the quality and genericity of GPT-4V's capabilities, its supported inputs and working modes, and the effective ways to prompt the model. In our approach to exploring GPT-4V, we curate and organize a collection of carefully designed qualitative samples spanning a variety of domains and tasks. Observations from these samples demonstrate that GPT-4V's unprecedented ability in processing arbitrarily interleaved multimodal inputs and the genericity of its capabilities together make GPT-4V a powerful multimodal generalist system. Furthermore, GPT-4V's unique capability of understanding visual markers drawn on input images can give rise to new human-computer interaction methods such as visual referring prompting. We conclude the report with in-depth discussions on the emerging application scenarios and the future research directions for GPT-4V-based systems. We hope that this preliminary exploration will inspire future research on the next-generation multimodal task formulation, new ways to exploit and enhance LMMs to solve real-world problems, and gaining better understanding of multimodal foundation models. Finally, we acknowledge that the model under our study is solely the product of OpenAI's innovative work, and they should be fully credited for its development. Please see the GPT-4V contributions paper for the authorship and credit attribution: https://cdn.openai.com/contributions/gpt-4v.pdf"
                },
                {
                    "title": "Multimodal Foundation Models: From Specialists to General-Purpose Assistants",
                    "abstract": "This paper presents a comprehensive survey of the taxonomy and evolution of multimodal foundation models that demonstrate vision and vision-language capabilities, focusing on the transition from specialist models to general-purpose assistants. The research landscape encompasses five core topics, categorized into two classes. (i) We start with a survey of well-established research areas: multimodal foundation models pre-trained for specific purposes, including two topics -- methods of learning vision backbones for visual understanding and text-to-image generation. (ii) Then, we present recent advances in exploratory, open research areas: multimodal foundation models that aim to play the role of general-purpose assistants, including three topics -- unified vision models inspired by large language models (LLMs), end-to-end training of multimodal LLMs, and chaining multimodal tools with LLMs. The target audiences of the paper are researchers, graduate students, and professionals in computer vision and vision-language multimodal communities who are eager to learn the basics and recent advances in multimodal foundation models."
                },
                {
                    "title": "Scaling Autoregressive Multi-Modal Models: Pretraining and Instruction Tuning",
                    "abstract": "We present CM3Leon (pronounced\"Chameleon\"), a retrieval-augmented, token-based, decoder-only multi-modal language model capable of generating and infilling both text and images. CM3Leon uses the CM3 multi-modal architecture but additionally shows the extreme benefits of scaling up and tuning on more diverse instruction-style data. It is the first multi-modal model trained with a recipe adapted from text-only language models, including a large-scale retrieval-augmented pre-training stage and a second multi-task supervised fine-tuning (SFT) stage. It is also a general-purpose model that can do both text-to-image and image-to-text generation, allowing us to introduce self-contained contrastive decoding methods that produce high-quality outputs. Extensive experiments demonstrate that this recipe is highly effective for multi-modal models. CM3Leon achieves state-of-the-art performance in text-to-image generation with 5x less training compute than comparable methods (zero-shot MS-COCO FID of 4.88). After SFT, CM3Leon can also demonstrate unprecedented levels of controllability in tasks ranging from language-guided image editing to image-controlled generation and segmentation."
                },
                {
                    "title": "LISA: Reasoning Segmentation via Large Language Model",
                    "abstract": "Although perception systems have made remarkable ad-vancements in recent years, they still rely on explicit human instruction or pre-defined categories to identify the target objects before executing visual recognition tasks. Such systems cannot actively reason and comprehend implicit user intention. In this work, we propose a new segmentation task - reasoning segmentation. The task is designed to output a segmentation mask given a complex and implicit query text. Furthermore, we establish a benchmark comprising over one thousand image-instruction-mask data samples, incorporating intricate reasoning and world knowledge for evaluation purposes. Finally, we present LISA: large Language Instructed Segmentation Assistant, which inherits the language generation capabilities of multimodal Large Language Models (LLMs) while also possessing the ability to produce segmentation masks. We expand the original vocabulary with a token and propose the embedding-as-mask paradigm to unlock the segmentation capability. Remarkably, LISA can handle cases involving complex rea-soning and world knowledge. Also, it demonstrates robust zero-shot capability when trained exclusively on reasoning-free datasets. In addition, fine-tuning the model with merely 239 reasoning segmentation data samples results in further performance enhancement. Both quantitative and qualitative experiments show our method effectively unlocks new reasoning segmentation capabilities for multimodal LLMs. Code, models, and data are available at github.com/dvlab-research/LISA."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "BuboGPT: Enabling Visual Grounding in Multi-Modal LLMs",
                    "abstract": "LLMs have demonstrated remarkable abilities at interacting with humans through language, especially with the usage of instruction-following data. Recent advancements in LLMs, such as MiniGPT-4, LLaVA, and X-LLM, further enlarge their abilities by incorporating multi-modal inputs, including image, video, and speech. Despite their effectiveness at generating precise and detailed language understanding of the given modality signal, these LLMs give up the ability to ground specific parts of inputs, thus only constructing a coarse-grained mapping. However, explicit and informative correspondence between text and other modalities will not only improve the user experience but also help to expand the application scenario of multi-modal LLMs. Therefore, we propose BuboGPT, a multi-modal LLM with visual grounding that can perform cross-modal interaction between vision, audio and language, providing fine-grained understanding of visual objects and other given modalities. As a result, BuboGPT is able to point out the specific location of an object in the image, when it is generating response or description for that object. Our contributions are two-fold: 1) An off-the-shelf visual grounding module based on SAM that extracts entities in a sentence and find corresponding masks in the image. 2) A two-stage training scheme and instruction dataset to endow joint text-image-audio understanding. Our experiments show that BuboGPT achieves impressive multi-modality understanding and visual grounding abilities during the interaction with human. It performs consistently well when provided by arbitrary modality combinations (either aligned or unaligned). Our code, model and dataset are available at https://bubo-gpt.github.io ."
                },
                {
                    "title": "Generative Pretraining in Multimodality",
                    "abstract": "We present Emu, a Transformer-based multimodal foundation model, which can seamlessly generate images and texts in multimodal context. This omnivore model can take in any single-modality or multimodal data input indiscriminately (e.g., interleaved image, text and video) through a one-model-for-all autoregressive training process. First, visual signals are encoded into embeddings, and together with text tokens form an interleaved input sequence. Emu is then end-to-end trained with a unified objective of classifying the next text token or regressing the next visual embedding in the multimodal sequence. This versatile multimodality empowers the exploration of diverse pretraining data sources at scale, such as videos with interleaved frames and text, webpages with interleaved images and text, as well as web-scale image-text pairs and video-text pairs. Emu can serve as a generalist multimodal interface for both image-to-text and text-to-image tasks, and supports in-context image and text generation. Across a broad range of zero-shot/few-shot tasks including image captioning, visual question answering, video question answering and text-to-image generation, Emu demonstrates superb performance compared to state-of-the-art large multimodal models. Extended capabilities such as multimodal assistants via instruction tuning are also demonstrated with impressive performance."
                },
                {
                    "title": "GPT4RoI: Instruction Tuning Large Language Model on Region-of-Interest",
                    "abstract": "Visual instruction tuning large language model(LLM) on image-text pairs has achieved general-purpose vision-language abilities. However, the lack of region-text pairs limits their advancements to fine-grained multimodal understanding. In this paper, we propose spatial instruction tuning, which introduces the reference to the region-of-interest(RoI) in the instruction. Before sending to LLM, the reference is replaced by RoI features and interleaved with language embeddings as a sequence. Our model GPT4RoI, trained on 7 region-text pair datasets, brings an unprecedented interactive and conversational experience compared to previous image-level models. (1) Interaction beyond language: Users can interact with our model by both language and drawing bounding boxes to flexibly adjust the referring granularity. (2) Versatile multimodal abilities: A variety of attribute information within each RoI can be mined by GPT4RoI, e.g., color, shape, material, action, etc. Furthermore, it can reason about multiple RoIs based on common sense. On the Visual Commonsense Reasoning(VCR) dataset, GPT4RoI achieves a remarkable accuracy of 81.6%, surpassing all existing models by a significant margin (the second place is 75.6%) and almost reaching human-level performance of 85.0%. The code, dataset, and demo can be found at https://github.com/jshilong/GPT4RoI."
                },
                {
                    "title": "Shikra: Unleashing Multimodal LLM's Referential Dialogue Magic",
                    "abstract": "In human conversations, individuals can indicate relevant regions within a scene while addressing others. In turn, the other person can then respond by referring to specific regions if necessary. This natural referential ability in dialogue remains absent in current Multimodal Large Language Models (MLLMs). To fill this gap, this paper proposes an MLLM called Shikra, which can handle spatial coordinate inputs and outputs in natural language. Its architecture consists of a vision encoder, an alignment layer, and a LLM. It is designed to be straightforward and simple, without the need for extra vocabularies, position encoder, pre-/post-detection modules, or external plug-in models. All inputs and outputs are in natural language form. Referential dialogue is a superset of various vision-language (VL) tasks. Shikra can naturally handle location-related tasks like REC and PointQA, as well as conventional VL tasks such as Image Captioning and VQA. Experimental results showcase Shikra's promising performance. Furthermore, it enables numerous exciting applications, like providing mentioned objects' coordinates in chains of thoughts and comparing user-pointed regions similarities. Our code, model and dataset are accessed at https://github.com/shikras/shikra."
                },
                {
                    "title": "Kosmos-2: Grounding Multimodal Large Language Models to the World",
                    "abstract": "We introduce Kosmos-2, a Multimodal Large Language Model (MLLM), enabling new capabilities of perceiving object descriptions (e.g., bounding boxes) and grounding text to the visual world. Specifically, we represent refer expressions as links in Markdown, i.e., ``[text span](bounding boxes)'', where object descriptions are sequences of location tokens. Together with multimodal corpora, we construct large-scale data of grounded image-text pairs (called GrIT) to train the model. In addition to the existing capabilities of MLLMs (e.g., perceiving general modalities, following instructions, and performing in-context learning), Kosmos-2 integrates the grounding capability into downstream applications. We evaluate Kosmos-2 on a wide range of tasks, including (i) multimodal grounding, such as referring expression comprehension, and phrase grounding, (ii) multimodal referring, such as referring expression generation, (iii) perception-language tasks, and (iv) language understanding and generation. This work lays out the foundation for the development of Embodiment AI and sheds light on the big convergence of language, multimodal perception, action, and world modeling, which is a key step toward artificial general intelligence. Code and pretrained models are available at https://aka.ms/kosmos-2."
                },
                {
                    "title": "OBELISC: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents",
                    "abstract": "Large multimodal models trained on natural documents, which interleave images and text, outperform models trained on image-text pairs on various multimodal benchmarks. However, the datasets used to train these models have not been released, and the collection process has not been fully specified. We introduce the OBELICS dataset, an open web-scale filtered dataset of interleaved image-text documents comprising 141 million web pages extracted from Common Crawl, 353 million associated images, and 115 billion text tokens. We describe the dataset creation process, present comprehensive filtering rules, and provide an analysis of the dataset's content. To show the viability of OBELICS, we train vision and language models of 9 and 80 billion parameters named IDEFICS, and obtain competitive performance on different multimodal benchmarks. We release our dataset, models and code."
                },
                {
                    "title": "M3IT: A Large-Scale Dataset towards Multi-Modal Multilingual Instruction Tuning",
                    "abstract": "Instruction tuning has significantly advanced large language models (LLMs) such as ChatGPT, enabling them to align with human instructions across diverse tasks. However, progress in open vision-language models (VLMs) has been limited due to the scarcity of high-quality instruction datasets. To tackle this challenge and promote research in the vision-language field, we introduce the Multi-Modal, Multilingual Instruction Tuning (M$^3$IT) dataset, designed to optimize VLM alignment with human instructions. Our M$^3$IT dataset comprises 40 carefully curated datasets, including 2.4 million instances and 400 manually written task instructions, reformatted into a vision-to-text structure. Key tasks are translated into 80 languages with an advanced translation system, ensuring broader accessibility. M$^3$IT surpasses previous datasets regarding task coverage, instruction number and instance scale. Moreover, we develop Ying-VLM, a VLM model trained on our M$^3$IT dataset, showcasing its potential to answer complex questions requiring world knowledge, generalize to unseen video tasks, and comprehend unseen instructions in Chinese. We have open-sourced the dataset to encourage further research."
                },
                {
                    "title": "Contextual Object Detection with Multimodal Large Language Models",
                    "abstract": "Recent Multimodal Large Language Models (MLLMs) are remarkable in vision-language tasks, such as image captioning and question answering, but lack the essential perception ability, i.e., object detection. In this work, we address this limitation by introducing a novel research problem of contextual object detection -- understanding visible objects within different human-AI interactive contexts. Three representative scenarios are investigated, including the language cloze test, visual captioning, and question answering. Moreover, we present ContextDET, a unified multimodal model that is capable of end-to-end differentiable modeling of visual-language contexts, so as to locate, identify, and associate visual objects with language inputs for human-AI interaction. Our ContextDET involves three key submodels: (i) a visual encoder for extracting visual representations, (ii) a pre-trained LLM for multimodal context decoding, and (iii) a visual decoder for predicting bounding boxes given contextual object words. The new generate-then-detect framework enables us to detect object words within human vocabulary. Extensive experiments show the advantages of ContextDET on our proposed CODE benchmark, open-vocabulary detection, and referring image segmentation. Github: https://github.com/yuhangzang/ContextDET."
                },
                {
                    "title": "PaLI-X: On Scaling up a Multilingual Vision and Language Model",
                    "abstract": "We present the training recipe and results of scaling up PaLI-X, a multilingual vision and language model, both in terms of size of the components and the breadth of its training task mixture. Our model achieves new levels of performance on a wide-range of varied and complex tasks, including multiple image-based captioning and question-answering tasks, image-based document understanding and few-shot (in-context) learning, as well as object detection, video question answering, and video captioning. PaLI-X advances the state-of-the-art on most vision-and-language benchmarks considered (25+ of them). Finally, we observe emerging capabilities, such as complex counting and multilingual object detection, tasks that are not explicitly in the training mix."
                },
                {
                    "title": "Generating Images with Multimodal Language Models",
                    "abstract": "We propose a method to fuse frozen text-only large language models (LLMs) with pre-trained image encoder and decoder models, by mapping between their embedding spaces. Our model demonstrates a wide suite of multimodal capabilities: image retrieval, novel image generation, and multimodal dialogue. Ours is the first approach capable of conditioning on arbitrarily interleaved image and text inputs to generate coherent image (and text) outputs. To achieve strong performance on image generation, we propose an efficient mapping network to ground the LLM to an off-the-shelf text-to-image generation model. This mapping network translates hidden representations of text into the embedding space of the visual models, enabling us to leverage the strong text representations of the LLM for visual outputs. Our approach outperforms baseline generation models on tasks with longer and more complex language. In addition to novel image generation, our model is also capable of image retrieval from a prespecified dataset, and decides whether to retrieve or generate at inference time. This is done with a learnt decision module which conditions on the hidden representations of the LLM. Our model exhibits a wider range of capabilities compared to prior multimodal language models. It can process image-and-text inputs, and produce retrieved images, generated images, and generated text -- outperforming non-LLM based generation models across several text-to-image tasks that measure context dependence."
                },
                {
                    "title": "VisionLLM: Large Language Model is also an Open-Ended Decoder for Vision-Centric Tasks",
                    "abstract": "Large language models (LLMs) have notably accelerated progress towards artificial general intelligence (AGI), with their impressive zero-shot capacity for user-tailored tasks, endowing them with immense potential across a range of applications. However, in the field of computer vision, despite the availability of numerous powerful vision foundation models (VFMs), they are still restricted to tasks in a pre-defined form, struggling to match the open-ended task capabilities of LLMs. In this work, we present an LLM-based framework for vision-centric tasks, termed VisionLLM. This framework provides a unified perspective for vision and language tasks by treating images as a foreign language and aligning vision-centric tasks with language tasks that can be flexibly defined and managed using language instructions. An LLM-based decoder can then make appropriate predictions based on these instructions for open-ended tasks. Extensive experiments show that the proposed VisionLLM can achieve different levels of task customization through language instructions, from fine-grained object-level to coarse-grained task-level customization, all with good results. It's noteworthy that, with a generalist LLM-based framework, our model can achieve over 60\\% mAP on COCO, on par with detection-specific models. We hope this model can set a new baseline for generalist vision and language models. The demo shall be released based on https://github.com/OpenGVLab/InternGPT. The code shall be released at https://github.com/OpenGVLab/VisionLLM."
                },
                {
                    "title": "Evaluating Object Hallucination in Large Vision-Language Models",
                    "abstract": "Inspired by the superior language abilities of large language models (LLM), large vision-language models (LVLM) have been recently explored by integrating powerful LLMs for improving the performance on complex multimodal tasks. Despite the promising progress on LVLMs, we find that LVLMs suffer from the hallucination problem, i.e. they tend to generate objects that are inconsistent with the target images in the descriptions. To investigate it, this work presents the first systematic study on object hallucination of LVLMs. We conduct the evaluation experiments on several representative LVLMs, and show that they mostly suffer from severe object hallucination issue. We further discuss that the visual instructions may influence the hallucination, and find that: objects that frequently occur in the visual instructions or co-occur with the image objects, are obviously prone to be hallucinated by LVLMs. Besides, we find that existing evaluation methods might be affected by the input instructions and generation styles of LVLMs. Thus, we further design an improved evaluation method for object hallucination by proposing a polling-based query method called POPE. Experiment results demonstrate that our POPE can evaluate the object hallucination in a more stable and flexible way. Our codes and data are publicly available at https://github.com/RUCAIBox/POPE."
                },
                {
                    "title": "InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning",
                    "abstract": "Large-scale pre-training and instruction tuning have been successful at creating general-purpose language models with broad competence. However, building general-purpose vision-language models is challenging due to the rich input distributions and task diversity resulting from the additional visual input. Although vision-language pretraining has been widely studied, vision-language instruction tuning remains under-explored. In this paper, we conduct a systematic and comprehensive study on vision-language instruction tuning based on the pretrained BLIP-2 models. We gather 26 publicly available datasets, covering a wide variety of tasks and capabilities, and transform them into instruction tuning format. Additionally, we introduce an instruction-aware Query Transformer, which extracts informative features tailored to the given instruction. Trained on 13 held-in datasets, InstructBLIP attains state-of-the-art zero-shot performance across all 13 held-out datasets, substantially outperforming BLIP-2 and larger Flamingo models. Our models also lead to state-of-the-art performance when finetuned on individual downstream tasks (e.g., 90.7% accuracy on ScienceQA questions with image contexts). Furthermore, we qualitatively demonstrate the advantages of InstructBLIP over concurrent multimodal models. All InstructBLIP models are open-sourced at https://github.com/salesforce/LAVIS/tree/main/projects/instructblip."
                },
                {
                    "title": "Otter: A Multi-Modal Model with In-Context Instruction Tuning",
                    "abstract": "Large language models (LLMs) have demonstrated significant universal capabilities as few/zero-shot learners in various tasks due to their pre-training on vast amounts of text data, as exemplified by GPT-3, which boosted to InstrctGPT and ChatGPT, effectively following natural language instructions to accomplish real-world tasks. In this paper, we propose to introduce instruction tuning into multi-modal models, motivated by the Flamingo model's upstream interleaved format pretraining dataset. We adopt a similar approach to construct our MultI-Modal In-Context Instruction Tuning (MIMIC-IT) dataset. We then introduce Otter, a multi-modal model based on OpenFlamingo (open-sourced version of DeepMind's Flamingo), trained on MIMIC-IT and showcasing improved instruction-following ability and in-context learning. We also optimize OpenFlamingo's implementation for researchers, democratizing the required training resources from 1$\\times$ A100 GPU to 4$\\times$ RTX-3090 GPUs, and integrate both OpenFlamingo and Otter into Huggingface Transformers for more researchers to incorporate the models into their customized training and inference pipelines."
                },
                {
                    "title": "mPLUG-Owl: Modularization Empowers Large Language Models with Multimodality",
                    "abstract": "Large language models (LLMs) have demonstrated impressive zero-shot abilities on a variety of open-ended tasks, while recent research has also explored the use of LLMs for multi-modal generation. In this study, we introduce mPLUG-Owl, a novel training paradigm that equips LLMs with multi-modal abilities through modularized learning of foundation LLM, a visual knowledge module, and a visual abstractor module. This approach can support multiple modalities and facilitate diverse unimodal and multimodal abilities through modality collaboration. The training paradigm of mPLUG-Owl involves a two-stage method for aligning image and text, which learns visual knowledge with the assistance of LLM while maintaining and even improving the generation abilities of LLM. In the first stage, the visual knowledge module and abstractor module are trained with a frozen LLM module to align the image and text. In the second stage, language-only and multi-modal supervised datasets are used to jointly fine-tune a low-rank adaption (LoRA) module on LLM and the abstractor module by freezing the visual knowledge module. We carefully build a visually-related instruction evaluation set OwlEval. Experimental results show that our model outperforms existing multi-modal models, demonstrating mPLUG-Owl's impressive instruction and visual understanding ability, multi-turn conversation ability, and knowledge reasoning ability. Besides, we observe some unexpected and exciting abilities such as multi-image correlation and scene text understanding, which makes it possible to leverage it for harder real scenarios, such as vision-only document comprehension. Our code, pre-trained model, instruction-tuned models, and evaluation set are available at https://github.com/X-PLUG/mPLUG-Owl. The online demo is available at https://www.modelscope.cn/studios/damo/mPLUG-Owl."
                },
                {
                    "title": "MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models",
                    "abstract": "The recent GPT-4 has demonstrated extraordinary multi-modal abilities, such as directly generating websites from handwritten text and identifying humorous elements within images. These features are rarely observed in previous vision-language models. However, the technical details behind GPT-4 continue to remain undisclosed. We believe that the enhanced multi-modal generation capabilities of GPT-4 stem from the utilization of sophisticated large language models (LLM). To examine this phenomenon, we present MiniGPT-4, which aligns a frozen visual encoder with a frozen advanced LLM, Vicuna, using one projection layer. Our work, for the first time, uncovers that properly aligning the visual features with an advanced large language model can possess numerous advanced multi-modal abilities demonstrated by GPT-4, such as detailed image description generation and website creation from hand-drawn drafts. Furthermore, we also observe other emerging capabilities in MiniGPT-4, including writing stories and poems inspired by given images, teaching users how to cook based on food photos, and so on. In our experiment, we found that the model trained on short image caption pairs could produce unnatural language outputs (e.g., repetition and fragmentation). To address this problem, we curate a detailed image description dataset in the second stage to finetune the model, which consequently improves the model's generation reliability and overall usability. Our code, pre-trained model, and collected dataset are available at https://minigpt-4.github.io/."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "Multimodal C4: An Open, Billion-scale Corpus of Images Interleaved With Text",
                    "abstract": "In-context vision and language models like Flamingo support arbitrarily interleaved sequences of images and text as input. This format not only enables few-shot learning via interleaving independent supervised (image, text) examples, but also, more complex prompts involving interaction between images, e.g.,\"What do image A and image B have in common?\"To support this interface, pretraining occurs over web corpora that similarly contain interleaved images+text. To date, however, large-scale data of this form have not been publicly available. We release Multimodal C4, an augmentation of the popular text-only C4 corpus with images interleaved. We use a linear assignment algorithm to place images into longer bodies of text using CLIP features, a process that we show outperforms alternatives. Multimodal C4 spans everyday topics like cooking, travel, technology, etc. A manual inspection of a random sample of documents shows that a vast majority (88%) of images are topically relevant, and that linear assignment frequently selects individual sentences specifically well-aligned with each image (80%). After filtering NSFW images, ads, etc., the resulting corpus consists of 101.2M documents with 571M images interleaved in 43B English tokens."
                },
                {
                    "title": "Segment Everything Everywhere All at Once",
                    "abstract": "In this work, we present SEEM, a promptable and interactive model for segmenting everything everywhere all at once in an image, as shown in Fig.1. In SEEM, we propose a novel decoding mechanism that enables diverse prompting for all types of segmentation tasks, aiming at a universal segmentation interface that behaves like large language models (LLMs). More specifically, SEEM is designed with four desiderata: i) Versatility. We introduce a new visual prompt to unify different spatial queries including points, boxes, scribbles and masks, which can further generalize to a different referring image; ii) Compositionality. We learn a joint visual-semantic space between text and visual prompts, which facilitates the dynamic composition of two prompt types required for various segmentation tasks; iii) Interactivity. We further incorporate learnable memory prompts into the decoder to retain segmentation history through mask-guided cross-attention from decoder to image features; and iv) Semantic-awareness. We use a text encoder to encode text queries and mask labels into the same semantic space for open-vocabulary segmentation. We conduct a comprehensive empirical study to validate the effectiveness of SEEM across diverse segmentation tasks. Notably, our single SEEM model achieves competitive performance across interactive segmentation, generic segmentation, referring segmentation, and video object segmentation on 9 datasets with minimum 1/100 supervision. Furthermore, SEEM showcases a remarkable capacity for generalization to novel prompts or their combinations, rendering it a readily universal image segmentation interface."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "PaLM-E: An Embodied Multimodal Language Model",
                    "abstract": "Large language models excel at a wide range of complex tasks. However, enabling general inference in the real world, e.g., for robotics problems, raises the challenge of grounding. We propose embodied language models to directly incorporate real-world continuous sensor modalities into language models and thereby establish the link between words and percepts. Input to our embodied language model are multi-modal sentences that interleave visual, continuous state estimation, and textual input encodings. We train these encodings end-to-end, in conjunction with a pre-trained large language model, for multiple embodied tasks including sequential robotic manipulation planning, visual question answering, and captioning. Our evaluations show that PaLM-E, a single large embodied multimodal model, can address a variety of embodied reasoning tasks, from a variety of observation modalities, on multiple embodiments, and further, exhibits positive transfer: the model benefits from diverse joint training across internet-scale language, vision, and visual-language domains. Our largest model, PaLM-E-562B with 562B parameters, in addition to being trained on robotics tasks, is a visual-language generalist with state-of-the-art performance on OK-VQA, and retains generalist language capabilities with increasing scale."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "GRiT: A Generative Region-to-text Transformer for Object Understanding",
                    "abstract": "This paper presents a Generative RegIon-to-Text transformer, GRiT, for object understanding. The spirit of GRiT is to formulate object understanding aspairs, where region locates objects and text describes objects. For example, the text in object detection denotes class names while that in dense captioning refers to descriptive sentences. Specifically, GRiT consists of a visual encoder to extract image features, a foreground object extractor to localize objects, and a text decoder to generate open-set object descriptions. With the same model architecture, GRiT can understand objects via not only simple nouns, but also rich descriptive sentences including object attributes or actions. Experimentally, we apply GRiT to object detection and dense captioning tasks. GRiT achieves 60.4 AP on COCO 2017 test-dev for object detection and 15.5 mAP on Visual Genome for dense captioning. Code is available at https://github.com/JialianW/GRiT"
                },
                {
                    "title": "BLOOM: A 176B-Parameter Open-Access Multilingual Language Model",
                    "abstract": "Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are frequently kept from the public. As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model designed and built thanks to a collaboration of hundreds of researchers. BLOOM is a decoder-only Transformer language model that was trained on the ROOTS corpus, a dataset comprising hundreds of sources in 46 natural and 13 programming languages (59 in total). We find that BLOOM achieves competitive performance on a wide variety of benchmarks, with stronger results after undergoing multitask prompted finetuning. To facilitate future research and applications using LLMs, we publicly release our models and code under the Responsible AI License."
                },
                {
                    "title": "PaLI: A Jointly-Scaled Multilingual Language-Image Model",
                    "abstract": "Effective scaling and a flexible task interface enable large language models to excel at many tasks. We present PaLI (Pathways Language and Image model), a model that extends this approach to the joint modeling of language and vision. PaLI generates text based on visual and textual inputs, and with this interface performs many vision, language, and multimodal tasks, in many languages. To train PaLI, we make use of large pre-trained encoder-decoder language models and Vision Transformers (ViTs). This allows us to capitalize on their existing capabilities and leverage the substantial cost of training them. We find that joint scaling of the vision and language components is important. Since existing Transformers for language are much larger than their vision counterparts, we train a large, 4-billion parameter ViT (ViT-e) to quantify the benefits from even larger-capacity vision models. To train PaLI, we create a large multilingual mix of pretraining tasks, based on a new image-text training set containing 10B images and texts in over 100 languages. PaLI achieves state-of-the-art in multiple vision and language tasks (such as captioning, visual question-answering, scene-text understanding), while retaining a simple, modular, and scalable design."
                },
                {
                    "title": "Unified-IO: A Unified Model for Vision, Language, and Multi-Modal Tasks",
                    "abstract": "We propose Unified-IO, a model that performs a large variety of AI tasks spanning classical computer vision tasks, including pose estimation, object detection, depth estimation and image generation, vision-and-language tasks such as region captioning and referring expression, to natural language processing tasks such as question answering and paraphrasing. Developing a single unified model for such a large variety of tasks poses unique challenges due to the heterogeneous inputs and outputs pertaining to each task, including RGB images, per-pixel maps, binary masks, bounding boxes, and language. We achieve this unification by homogenizing every supported input and output into a sequence of discrete vocabulary tokens. This common representation across all tasks allows us to train a single transformer-based architecture, jointly on over 90 diverse datasets in the vision and language fields. Unified-IO is the first model capable of performing all 7 tasks on the GRIT benchmark and produces strong results across 16 diverse benchmarks like NYUv2-Depth, ImageNet, VQA2.0, OK-VQA, Swig, VizWizGround, BoolQ, and SciTail, with no task-specific fine-tuning. Code and demos for Unified-IO are available at: https://unified-io.allenai.org."
                },
                {
                    "title": "A Unified Sequence Interface for Vision Tasks",
                    "abstract": "While language tasks are naturally expressed in a single, unified, modeling framework, i.e., generating sequences of tokens, this has not been the case in computer vision. As a result, there is a proliferation of distinct architectures and loss functions for different vision tasks. In this work we show that a diverse set of\"core\"computer vision tasks can also be unified if formulated in terms of a shared pixel-to-sequence interface. We focus on four tasks, namely, object detection, instance segmentation, keypoint detection, and image captioning, all with diverse types of outputs, e.g., bounding boxes or dense masks. Despite that, by formulating the output of each task as a sequence of discrete tokens with a unified interface, we show that one can train a neural network with a single model architecture and loss function on all these tasks, with no task-specific customization. To solve a specific task, we use a short prompt as task description, and the sequence output adapts to the prompt so it can produce task-specific output. We show that such a model can achieve competitive performance compared to well-established task-specific models."
                },
                {
                    "title": "GLIPv2: Unifying Localization and Vision-Language Understanding",
                    "abstract": "We present GLIPv2, a grounded VL understanding model, that serves both localization tasks (e.g., object detection, instance segmentation) and Vision-Language (VL) understanding tasks (e.g., VQA, image captioning). GLIPv2 elegantly unifies localization pre-training and Vision-Language Pre-training (VLP) with three pre-training tasks: phrase grounding as a VL reformulation of the detection task, region-word contrastive learning as a novel region-word level contrastive learning task, and the masked language modeling. This unification not only simplifies the previous multi-stage VLP procedure but also achieves mutual benefits between localization and understanding tasks. Experimental results show that a single GLIPv2 model (all model weights are shared) achieves near SoTA performance on various localization and understanding tasks. The model also shows (1) strong zero-shot and few-shot adaption performance on open-vocabulary object detection tasks and (2) superior grounding capability on VL understanding tasks. Code will be released at https://github.com/microsoft/GLIP."
                },
                {
                    "title": "GIT: A Generative Image-to-text Transformer for Vision and Language",
                    "abstract": "In this paper, we design and train a Generative Image-to-text Transformer, GIT, to unify vision-language tasks such as image/video captioning and question answering. While generative models provide a consistent network architecture between pre-training and fine-tuning, existing work typically contains complex structures (uni/multi-modal encoder/decoder) and depends on external modules such as object detectors/taggers and optical character recognition (OCR). In GIT, we simplify the architecture as one image encoder and one text decoder under a single language modeling task. We also scale up the pre-training data and the model size to boost the model performance. Without bells and whistles, our GIT establishes new state of the arts on 12 challenging benchmarks with a large margin. For instance, our model surpasses the human performance for the first time on TextCaps (138.2 vs. 125.5 in CIDEr). Furthermore, we present a new scheme of generation-based image classification and scene text recognition, achieving decent performance on standard benchmarks. Codes are released at \\url{https://github.com/microsoft/GenerativeImage2Text}."
                },
                {
                    "title": "Flamingo: a Visual Language Model for Few-Shot Learning",
                    "abstract": "Building models that can be rapidly adapted to novel tasks using only a handful of annotated examples is an open challenge for multimodal machine learning research. We introduce Flamingo, a family of Visual Language Models (VLM) with this ability. We propose key architectural innovations to: (i) bridge powerful pretrained vision-only and language-only models, (ii) handle sequences of arbitrarily interleaved visual and textual data, and (iii) seamlessly ingest images or videos as inputs. Thanks to their flexibility, Flamingo models can be trained on large-scale multimodal web corpora containing arbitrarily interleaved text and images, which is key to endow them with in-context few-shot learning capabilities. We perform a thorough evaluation of our models, exploring and measuring their ability to rapidly adapt to a variety of image and video tasks. These include open-ended tasks such as visual question-answering, where the model is prompted with a question which it has to answer; captioning tasks, which evaluate the ability to describe a scene or an event; and close-ended tasks such as multiple-choice visual question-answering. For tasks lying anywhere on this spectrum, a single Flamingo model can achieve a new state of the art with few-shot learning, simply by prompting the model with task-specific examples. On numerous benchmarks, Flamingo outperforms models fine-tuned on thousands of times more task-specific data."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "Rethinking Network Design and Local Geometry in Point Cloud: A Simple Residual MLP Framework",
                    "abstract": "Point cloud analysis is challenging due to irregularity and unordered data structure. To capture the 3D geometries, prior works mainly rely on exploring sophisticated local geometric extractors using convolution, graph, or attention mechanisms. These methods, however, incur unfavorable latency during inference, and the performance saturates over the past few years. In this paper, we present a novel perspective on this task. We notice that detailed local geometrical information probably is not the key to point cloud analysis -- we introduce a pure residual MLP network, called PointMLP, which integrates no sophisticated local geometrical extractors but still performs very competitively. Equipped with a proposed lightweight geometric affine module, PointMLP delivers the new state-of-the-art on multiple datasets. On the real-world ScanObjectNN dataset, our method even surpasses the prior best method by 3.3% accuracy. We emphasize that PointMLP achieves this strong performance without any sophisticated operations, hence leading to a superior inference speed. Compared to most recent CurveNet, PointMLP trains 2x faster, tests 7x faster, and is more accurate on ModelNet40 benchmark. We hope our PointMLP may help the community towards a better understanding of point cloud analysis. The code is available at https://github.com/ma-xu/pointMLP-pytorch."
                },
                {
                    "title": "OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework",
                    "abstract": "In this work, we pursue a unified paradigm for multimodal pretraining to break the scaffolds of complex task/modality-specific customization. We propose OFA, a Task-Agnostic and Modality-Agnostic framework that supports Task Comprehensiveness. OFA unifies a diverse set of cross-modal and unimodal tasks, including image generation, visual grounding, image captioning, image classification, language modeling, etc., in a simple sequence-to-sequence learning framework. OFA follows the instruction-based learning in both pretraining and finetuning stages, requiring no extra task-specific layers for downstream tasks. In comparison with the recent state-of-the-art vision&language models that rely on extremely large cross-modal datasets, OFA is pretrained on only 20M publicly available image-text pairs. Despite its simplicity and relatively small-scale training data, OFA achieves new SOTAs in a series of cross-modal tasks while attaining highly competitive performances on uni-modal tasks. Our further analysis indicates that OFA can also effectively transfer to unseen tasks and unseen domains. Our code and models are publicly available at https://github.com/OFA-Sys/OFA."
                },
                {
                    "title": "CM3: A Causal Masked Multimodal Model of the Internet",
                    "abstract": "We introduce CM3, a family of causally masked generative models trained over a large corpus of structured multi-modal documents that can contain both text and image tokens. Our new causally masked approach generates tokens left to right while also masking out a small number of long token spans that are generated at the end of the string, instead of their original positions. The casual masking object provides a type of hybrid of the more common causal and masked language models, by enabling full generative modeling while also providing bidirectional context when generating the masked spans. We train causally masked language-image models on large-scale web and Wikipedia articles, where each document contains all of the text, hypertext markup, hyperlinks, and image tokens (from a VQVAE-GAN), provided in the order they appear in the original HTML source (before masking). The resulting CM3 models can generate rich structured, multi-modal outputs while conditioning on arbitrary masked document contexts, and thereby implicitly learn a wide range of text, image, and cross modal tasks. They can be prompted to recover, in a zero-shot fashion, the functionality of models such as DALL-E, GENRE, and HTLM. We set the new state-of-the-art in zero-shot summarization, entity linking, and entity disambiguation while maintaining competitive performance in the fine-tuning setting. We can generate images unconditionally, conditioned on text (like DALL-E) and do captioning all in a zero-shot setting with a single model."
                },
                {
                    "title": "Grounded Language-Image Pre-training",
                    "abstract": "This paper presents a grounded language-image pretraining (GLIP) model for learning object-level, language-aware, and semantic-rich visual representations. GLIP unifies object detection and phrase grounding for pre-training. The unification brings two benefits: 1) it allows GLIP to learn from both detection and grounding data to improve both tasks and bootstrap a good grounding model; 2) GLIP can leverage massive image-text pairs by generating grounding boxes in a self-training fashion, making the learned representations semantic-rich. In our experiments, we pre-train GLIP on 27M grounding data, including 3M human-annotated and 24M web-crawled image-text pairs. The learned representations demonstrate strong zero-shot and few-shot transferability to various object-level recognition tasks. 1) When directly evaluated on COCO and LVIS (without seeing any images in COCO during pre-training), GLIP achieves 49.8 AP and 26.9 AP, respectively, surpassing many supervised baselines.11Supervised baselines on COCO object detection: Faster-RCNN w/ ResNet50 (40.2) or ResNet101 (42.0), and DyHead w/ Swin-Tiny (49.7). 2) After fine-tuned on COCO, GLIP achieves 60.8 AP on val and 61.5 AP on test-dev, surpassing prior SoTA. 3) When transferred to 13 downstream object detection tasks, a 1-shot GLIP rivals with a fully-supervised Dynamic Head. Code will be released at https://github.com/microsoft/GLIP."
                },
                {
                    "title": "Pix2seq: A Language Modeling Framework for Object Detection",
                    "abstract": "We present Pix2Seq, a simple and generic framework for object detection. Unlike existing approaches that explicitly integrate prior knowledge about the task, we cast object detection as a language modeling task conditioned on the observed pixel inputs. Object descriptions (e.g., bounding boxes and class labels) are expressed as sequences of discrete tokens, and we train a neural network to perceive the image and generate the desired sequence. Our approach is based mainly on the intuition that if a neural network knows about where and what the objects are, we just need to teach it how to read them out. Beyond the use of task-specific data augmentations, our approach makes minimal assumptions about the task, yet it achieves competitive results on the challenging COCO dataset, compared to highly specialized and well optimized detection algorithms."
                },
                {
                    "title": "SimVLM: Simple Visual Language Model Pretraining with Weak Supervision",
                    "abstract": "With recent progress in joint modeling of visual and textual representations, Vision-Language Pretraining (VLP) has achieved impressive performance on many multimodal downstream tasks. However, the requirement for expensive annotations including clean image captions and regional labels limits the scalability of existing approaches, and complicates the pretraining procedure with the introduction of multiple dataset-specific objectives. In this work, we relax these constraints and present a minimalist pretraining framework, named Simple Visual Language Model (SimVLM). Unlike prior work, SimVLM reduces the training complexity by exploiting large-scale weak supervision, and is trained end-to-end with a single prefix language modeling objective. Without utilizing extra data or task-specific customization, the resulting model significantly outperforms previous pretraining methods and achieves new state-of-the-art results on a wide range of discriminative and generative vision-language benchmarks, including VQA (+3.74% vqa-score), NLVR2 (+1.17% accuracy), SNLI-VE (+1.37% accuracy) and image captioning tasks (+10.1% average CIDEr score). Furthermore, we demonstrate that SimVLM acquires strong generalization and transfer ability, enabling zero-shot behavior including open-ended visual question answering and cross-modality transfer."
                },
                {
                    "title": "MDETR - Modulated Detection for End-to-End Multi-Modal Understanding",
                    "abstract": "Multi-modal reasoning systems rely on a pre-trained object detector to extract regions of interest from the image. However, this crucial module is typically used as a black box, trained independently of the downstream task and on a fixed vocabulary of objects and attributes. This makes it challenging for such systems to capture the long tail of visual concepts expressed in free form text. In this paper we propose MDETR, an end-to-end modulated detector that detects objects in an image conditioned on a raw text query, like a caption or a question. We use a transformer-based architecture to reason jointly over text and image by fusing the two modalities at an early stage of the model. We pre-train the network on 1.3M text-image pairs, mined from pre-existing multi-modal datasets having explicit alignment between phrases in text and objects in the image. We then fine-tune on several downstream tasks such as phrase grounding, referring expression comprehension and segmentation, achieving state-of-the-art results on popular benchmarks. We also investigate the utility of our model as an object detector on a given label set when fine-tuned in a few-shot setting. We show that our pre-training approach provides a way to handle the long tail of object categories which have very few labelled instances. Our approach can be easily extended for visual question answering, achieving competitive performance on GQA and CLEVR. The code and models are available at https://github.com/ashkamath/mdetr."
                },
                {
                    "title": "TransVG: End-to-End Visual Grounding with Transformers",
                    "abstract": "In this paper, we present a neat yet effective transformer-based framework for visual grounding, namely TransVG, to address the task of grounding a language query to the corresponding region onto an image. The state-of-the-art methods, including two-stage or one-stage ones, rely on a complex module with manually-designed mechanisms to perform the query reasoning and multi-modal fusion. However, the involvement of certain mechanisms in fusion module design, such as query decomposition and image scene graph, makes the models easily overfit to datasets with specific scenarios, and limits the plenitudinous interaction between the visual-linguistic context. To avoid this caveat, we propose to establish the multi-modal correspondence by leveraging transformers, and empirically show that the complex fusion modules (e.g., modular attention network, dynamic graph, and multi-modal tree) can be replaced by a simple stack of transformer encoder layers with higher performance. Moreover, we re-formulate the visual grounding as a direct coordinates regression problem and avoid making predictions out of a set of candidates (i.e., region proposals or anchor boxes). Extensive experiments are conducted on five widely used datasets, and a series of state-of-the-art records are set by our TransVG. We build the benchmark of transformer-based visual grounding framework and make the code available at https://github.com/djiajunustc/TransVG."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Large-Scale Adversarial Training for Vision-and-Language Representation Learning",
                    "abstract": "We present VILLA, the first known effort on large-scale adversarial training for vision-and-language (V+L) representation learning. VILLA consists of two training stages: (i) task-agnostic adversarial pre-training; followed by (ii) task-specific adversarial finetuning. Instead of adding adversarial perturbations on image pixels and textual tokens, we propose to perform adversarial training in the embedding space of each modality. To enable large-scale training, we adopt the \"free\" adversarial training strategy, and combine it with KL-divergence-based regularization to promote higher invariance in the embedding space. We apply VILLA to current best-performing V+L models, and achieve new state of the art on a wide range of tasks, including Visual Question Answering, Visual Commonsense Reasoning, Image-Text Retrieval, Referring Expression Comprehension, Visual Entailment, and NLVR2."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "More Grounded Image Captioning by Distilling Image-Text Matching Model",
                    "abstract": "Visual attention not only improves the performance of image captioners, but also serves as a visual interpretation to qualitatively measure the caption rationality and model transparency. Specifically, we expect that a captioner can fix its attentive gaze on the correct objects while generating the corresponding words. This ability is also known as grounded image captioning. However, the grounding accuracy of existing captioners is far from satisfactory.To improve the grounding accuracy while retaining the captioning quality, it is expensive to collect the word-region alignment as strong supervision.To this end, we propose a Part-of-Speech (POS) enhanced image-text matching model (SCAN): POS-SCAN, as the effective knowledge distillation for more grounded image captioning. The benefits are two-fold: 1) given a sentence and an image, POS-SCAN can ground the objects more accurately than SCAN; 2) POS-SCAN serves as a word-region alignment regularization for the captioner's visual attention module. By showing benchmark experimental results, we demonstrate that conventional image captioners equipped with POS-SCAN can significantly improve the grounding accuracy without strong supervision. Last but not the least, we explore the indispensable Self-Critical Sequence Training (SCST) in the context of grounded image captioning and show that the image-text matching score can serve as a reward for more grounded captioning."
                },
                {
                    "title": "Objects365: A Large-Scale, High-Quality Dataset for Object Detection",
                    "abstract": "In this paper, we introduce a new large-scale object detection dataset, Objects365, which has 365 object categories over 600K training images. More than 10 million, high-quality bounding boxes are manually labeled through a three-step, carefully designed annotation pipeline. It is the largest object detection dataset (with full annotation) so far and establishes a more challenging benchmark for the community. Objects365 can serve as a better feature learning dataset for localization-sensitive tasks like object detection and semantic segmentation. The Objects365 pre-trained models significantly outperform ImageNet pre-trained models with 5.6 points gain (42 vs 36.4) based on the standard setting of 90K iterations on COCO benchmark. Even compared with much long training time like 540K iterations, our Objects365 pretrained model with 90K iterations still have 2.7 points gain (42 vs 39.3). Meanwhile, the finetuning time can be greatly reduced (up to 10 times) when reaching the same accuracy. Better generalization ability of Object365 has also been verified on CityPersons, VOC segmentation, and ADE tasks. The dataset as well as the pretrained-models have been released at www.objects365.org."
                },
                {
                    "title": "LVIS: A Dataset for Large Vocabulary Instance Segmentation",
                    "abstract": "Progress on object detection is enabled by datasets that focus the research community\u2019s attention on open challenges. This process led us from simple images to complex scenes and from bounding boxes to segmentation masks. In this work, we introduce LVIS (pronounced \u2018el-vis\u2019): a new dataset for Large Vocabulary Instance Segmentation. We plan to collect 2.2 million high-quality instance segmentation masks for over 1000 entry-level object categories in 164k images. Due to the Zipfian distribution of categories in natural images, LVIS naturally has a long tail of categories with few training samples. Given that state-of-the-art deep learning methods for object detection perform poorly in the low-sample regime, we believe that our dataset poses an important and exciting new scientific challenge. LVIS is available at http://www.lvisdataset.org."
                },
                {
                    "title": "Grounded Video Description",
                    "abstract": "Video description is one of the most challenging problems in vision and language understanding due to the large variability both on the video and language side. Models, hence, typically shortcut the difficulty in recognition and generate plausible sentences that are based on priors but are not necessarily grounded in the video. In this work, we explicitly link the sentence to the evidence in the video by annotating each noun phrase in a sentence with the corresponding bounding box in one of the frames of a video. Our dataset, ActivityNet-Entities, augments the challenging ActivityNet Captions dataset with 158k bounding box annotations, each grounding a noun phrase. This allows training video description models with this data, and importantly, evaluate how grounded or \"true\" such model are to the video they describe. To generate grounded captions, we propose a novel video description model which is able to exploit these bounding box annotations. We demonstrate the effectiveness of our model on our dataset, but also show how it can be applied to image description on the Flickr30k Entities dataset. We achieve state-of-the-art performance on video description, video paragraph description, and image description and demonstrate our generated sentences are better grounded in the video."
                },
                {
                    "title": "From Recognition to Cognition: Visual Commonsense Reasoning",
                    "abstract": "Visual understanding goes well beyond object recognition. With one glance at an image, we can effortlessly imagine the world beyond the pixels: for instance, we can infer people's actions, goals, and mental states. While this task is easy for humans, it is tremendously difficult for today's vision systems, requiring higher-order cognition and commonsense reasoning about the world. We formalize this task as Visual Commonsense Reasoning. Given a challenging question about an image, a machine must answer correctly and then provide a rationale justifying its answer. Next, we introduce a new dataset, VCR, consisting of 290k multiple choice QA problems derived from 110k movie scenes. The key recipe for generating non-trivial and high-quality problems at scale is Adversarial Matching, a new approach to transform rich annotations into multiple choice questions with minimal bias. Experimental results show that while humans find VCR easy (over 90% accuracy), state-of-the-art vision models struggle (~45%). To move towards cognition-level understanding, we present a new reasoning engine, Recognition to Cognition Networks (R2C), that models the necessary layered inferences for grounding, contextualization, and reasoning. R2C helps narrow the gap between humans and machines (~65%); still, the challenge is far from solved, and we provide analysis that suggests avenues for future work."
                },
                {
                    "title": "Dynamic Graph CNN for Learning on Point Clouds",
                    "abstract": "Point clouds provide a flexible geometric representation suitable for countless applications in computer graphics; they also comprise the raw output of most 3D data acquisition devices. While hand-designed features on point clouds have long been proposed in graphics and vision, however, the recent overwhelming success of convolutional neural networks (CNNs) for image analysis suggests the value of adapting insight from CNN to the point cloud world. Point clouds inherently lack topological information, so designing a model to recover topology can enrich the representation power of point clouds. To this end, we propose a new neural network module dubbed EdgeConv suitable for CNN-based high-level tasks on point clouds, including classification and segmentation. EdgeConv acts on graphs dynamically computed in each layer of the network. It is differentiable and can be plugged into existing architectures. Compared to existing modules operating in extrinsic space or treating each point independently, EdgeConv has several appealing properties: It incorporates local neighborhood information; it can be stacked applied to learn global shape properties; and in multi-layer systems affinity in feature space captures semantic characteristics over potentially long distances in the original embedding. We show the performance of our model on standard benchmarks, including ModelNet40, ShapeNetPart, and S3DIS."
                },
                {
                    "title": "MAttNet: Modular Attention Network for Referring Expression Comprehension",
                    "abstract": "In this paper, we address referring expression comprehension: localizing an image region described by a natural language expression. While most recent work treats expressions as a single unit, we propose to decompose them into three modular components related to subject appearance, location, and relationship to other objects. This allows us to flexibly adapt to expressions containing different types of information in an end-to-end framework. In our model, which we call the Modular Attention Network (MAttNet), two types of attention are utilized: language-based attention that learns the module weights as well as the word/phrase attention that each module should focus on; and visual attention that allows the subject and relationship modules to focus on relevant image components. Module weights combine scores from all three modules dynamically to output an overall score. Experiments show that MAttNet outperforms previous state-of-the-art methods by a large margin on both bounding-box-level and pixel-level comprehension tasks. Demo1 and code2 are provided."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space",
                    "abstract": "Few prior works study deep learning on point sets. PointNet by Qi et al. is a pioneer in this direction. However, by design PointNet does not capture local structures induced by the metric space points live in, limiting its ability to recognize fine-grained patterns and generalizability to complex scenes. In this work, we introduce a hierarchical neural network that applies PointNet recursively on a nested partitioning of the input point set. By exploiting metric space distances, our network is able to learn local features with increasing contextual scales. With further observation that point sets are usually sampled with varying densities, which results in greatly decreased performance for networks trained on uniform densities, we propose novel set learning layers to adaptively combine features from multiple scales. Experiments show that our network called PointNet++ is able to learn deep point set features efficiently and robustly. In particular, results significantly better than state-of-the-art have been obtained on challenging benchmarks of 3D point clouds."
                },
                {
                    "title": "Mask R-CNN",
                    "abstract": "We present a conceptually simple, flexible, and general framework for object instance segmentation. Our approach efficiently detects objects in an image while simultaneously generating a high-quality segmentation mask for each instance. The method, called Mask R-CNN, extends Faster R-CNN by adding a branch for predicting an object mask in parallel with the existing branch for bounding box recognition. Mask R-CNN is simple to train and adds only a small overhead to Faster R-CNN, running at 5 fps. Moreover, Mask R-CNN is easy to generalize to other tasks, e.g., allowing us to estimate human poses in the same framework. We show top results in all three tracks of the COCO suite of challenges, including instance segmentation, bounding-box object detection, and person keypoint detection. Without tricks, Mask R-CNN outperforms all existing, single-model entries on every task, including the COCO 2016 challenge winners. We hope our simple and effective approach will serve as a solid baseline and help ease future research in instance-level recognition. Code will be made available."
                },
                {
                    "title": "Comprehension-Guided Referring Expressions",
                    "abstract": "We consider generation and comprehension of natural language referring expression for objects in an image. Unlike generic image captioning which lacks natural standard evaluation criteria, quality of a referring expression may be measured by the receivers ability to correctly infer which object is being described. Following this intuition, we propose two approaches to utilize models trained for comprehension task to generate better expressions. First, we use a comprehension module trained on human-generated expressions, as a critic of referring expression generator. The comprehension module serves as a differentiable proxy of human evaluation, providing training signal to the generation module. Second, we use the comprehension model in a generate-and-rerank pipeline, which chooses from candidate expressions generated by a model according to their performance on the comprehension task. We show that both approaches lead to improved referring expression generation on multiple benchmark datasets."
                },
                {
                    "title": "A Joint Speaker-Listener-Reinforcer Model for Referring Expressions",
                    "abstract": "Referring expressions are natural language constructions used to identify particular objects within a scene. In this paper, we propose a unified framework for the tasks of referring expression comprehension and generation. Our model is composed of three modules: speaker, listener, and reinforcer. The speaker generates referring expressions, the listener comprehends referring expressions, and the reinforcer introduces a reward function to guide sampling of more discriminative expressions. The listener-speaker modules are trained jointly in an end-to-end learning framework, allowing the modules to be aware of one another during learning while also benefiting from the discriminative reinforcer&#x2019;s feedback. We demonstrate that this unified framework and training achieves state-of-the-art results for both comprehension and generation on three referring expression datasets."
                },
                {
                    "title": "PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation",
                    "abstract": "Point cloud is an important type of geometric data structure. Due to its irregular format, most researchers transform such data to regular 3D voxel grids or collections of images. This, however, renders data unnecessarily voluminous and causes issues. In this paper, we design a novel type of neural network that directly consumes point clouds, which well respects the permutation invariance of points in the input. Our network, named PointNet, provides a unified architecture for applications ranging from object classification, part segmentation, to scene semantic parsing. Though simple, PointNet is highly efficient and effective. Empirically, it shows strong performance on par or even better than state of the art. Theoretically, we provide analysis towards understanding of what the network has learnt and why the network is robust with respect to input perturbation and corruption."
                },
                {
                    "title": "Generation and Comprehension of Unambiguous Object Descriptions",
                    "abstract": "We propose a method that can generate an unambiguous description (known as a referring expression) of a specific object or region in an image, and which can also comprehend or interpret such an expression to infer which object is being described. We show that our method outperforms previous methods that generate descriptions of objects without taking into account other potentially ambiguous objects in the scene. Our model is inspired by recent successes of deep learning methods for image captioning, but while image captioning is difficult to evaluate, our task allows for easy objective evaluation. We also present a new large-scale dataset for referring expressions, based on MSCOCO. We have released the dataset and a toolbox for visualization and evaluation, see https://github.com/ mjhucla/Google_Refexp_toolbox."
                },
                {
                    "title": "ReferItGame: Referring to Objects in Photographs of Natural Scenes",
                    "abstract": "In this paper we introduce a new game to crowd-source natural language referring expressions. By designing a two player game, we can both collect and verify referring expressions directly within the game. To date, the game has produced a dataset containing 130,525 expressions, referring to 96,654 distinct objects, in 19,894 photographs of natural scenes. This dataset is larger and more varied than previous REG datasets and allows us to study referring expressions in real-world scenes. We provide an in depth analysis of the resulting dataset. Based on our findings, we design a new optimization based model for generating referring expressions and perform experimental evaluations on 3 test sets."
                },
                {
                    "title": "Computational Generation of Referring Expressions: A Survey",
                    "abstract": "This article offers a survey of computational research on referring expression generation (REG). It introduces the REG problem and describes early work in this area, discussing what basic assumptions lie behind it, and showing how its remit has widened in recent years. We discuss computational frameworks underlying REG, and demonstrate a recent trend that seeks to link REG algorithms with well-established Knowledge Representation techniques. Considerable attention is given to recent efforts at evaluating REG algorithms and the lessons that they allow us to learn. The article concludes with a discussion of the way forward in REG, focusing on references in larger and more realistic settings."
                },
                {
                    "title": "Expression",
                    "abstract": "Whether the word \u2018passion\u2019, as indicating the suffering or affection from without of a soul, is by now no more than a dead metaphor, surviving from an antique conception of the mind; whether, indeed, there is any way open to us of determining the passivity or otherwise of our inner life, apart, that is, from how it strikes us, from how we are prompted to describe it, are not questions that I can take up this evening. It is enough for my purpose that for much of the time our feelings, our emotions, our inclinations are as fluctuating or as imperious as if they were not totally under our control. We are elated: we are dejected: we get angry, and then our anger gives place to a feeling of absurdity: we remain in love with someone who is lost to us but whom we cannot renounce: we are interested in something, and suddenly we are bored, or frightened that we will be bored: we see a stranger, someone who is nothing to us, who is poor or crippled, and we feel guilt: someone does something wrong or foolish, and we are unaccountably transported by laughter, by \u2018sudden glory\u2019 as Hobbes called it, knowing what it was about, and then, as unaccountably, we are thrown down. Man is, in Montaigne's famous phrase, une chose ondoyante, a creature of inner change and fickleness."
                },
                {
                    "title": "Aligning Large Multi-Modal Model with Robust Instruction Tuning",
                    "abstract": "Despite the promising progress in multi-modal tasks, current large multi-modal models (LMM) are prone to hallucinating inconsistent descriptions with respect to the associated image and human instructions. This paper addresses this issue by introducing the first large and diverse visual instruction tuning dataset, named Large-scale Robust Visual (LRV)-Instruction . Our dataset consists of 120k visual instructions generated by GPT4, covering 16 vision-and-language tasks with open-ended instructions and answers. Unlike existing studies that primarily focus on positive instruction samples, we design LRV-Instruction to include both positive and negative instructions for more robust visual instruction tuning. Our negative instructions are designed at two semantic levels: (i) Nonexistent Element Manipulation and (ii) Existent Element Manipulation . To efficiently measure the hallucination generated by LMMs, we propose GPT4-Assisted Visual Instruction Evaluation (GAVIE) , a novel approach to evaluate visual instruction tuning without the need for human-annotated groundtruth answers and can adapt to diverse instruction formats. We conduct comprehensive experiments to investigate the hallucination of LMMs. Our results demonstrate that existing LMMs exhibit significant hallucination when presented with our negative instructions, particularly with Existent Element Manipulation instructions. Moreover, by finetuning MiniGPT4 on LRV-Instruction , we successfully mitigate hallucination while improving performance on public datasets using less training data compared to state-of-the-art methods. Additionally, we observed that a balanced ratio of positive and negative instances in the training data leads to a more robust model. Our project link is available at this link."
                },
                {
                    "title": "Grounding Language Models to Images for Multimodal Generation",
                    "abstract": "We propose an ef\ufb01cient method to ground pre-trained text-only language models to the visual domain, enabling them to process and generate arbitrarily interleaved image-and-text data. Our method leverages the abilities of language models learnt from large scale text-only pretraining, such as in-context learning and free-form text generation. We keep the language model frozen, and \ufb01netune input and output linear layers to enable cross-modality interactions. This allows our model to process arbitrarily interleaved image-and-text inputs, and generate free-form text inter-leaved with retrieved images. We achieve strong zero-shot performance on grounded tasks such as contextual image retrieval and multimodal dialogue, and showcase compelling interactive abilities. Our approach works with any off-the-shelf language model and paves the way towards an effective, general solution for leveraging pretrained language models in visually grounded settings."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we unify referring and grounding capabilities in a single framework for vision-language learning to enhance spatial understanding in models?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of multimodal interactions, as it bridges the gap between referring and grounding tasks, which are often treated separately. By integrating these capabilities, future research can explore more sophisticated models that mimic human-like understanding and reasoning in visual contexts. This advancement could lead to practical applications in areas such as robotics, augmented reality, and human-computer interaction, where accurate spatial comprehension is essential.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of representing diverse types of regions (e.g., points, boxes, scribbles) and the need for models to generalize knowledge across tasks. Naive approaches may fail because they typically rely on fixed representations that do not accommodate the variability of human referencing methods. Additionally, technical obstacles include the integration of discrete coordinates with continuous visual features, as well as ensuring that the model can handle open-vocabulary inputs and complex reasoning tasks effectively.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on either referring or grounding in isolation, leading to a lack of comprehensive frameworks that address both simultaneously. Limitations in existing models include their inability to process free-form region inputs and the absence of datasets that encompass the necessary complexity of spatial knowledge. Our approach differs by introducing Ferret, which combines these tasks within a single multimodal large language model and leverages the GRIT dataset to provide a rich training foundation.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of Ferret, a multimodal large language model that utilizes a spatial-aware visual sampler to represent versatile region shapes. We will employ the GRIT dataset, which contains 1.1 million samples, to train the model on both grounding and referring tasks. The expected outcomes include a robust model capable of processing mixed inputs of referred regions and free-form text, generating accurate coordinates for groundable objects, and demonstrating open-vocabulary and instruction-following capabilities."
            }
        },
        "author_data": {
            "5301323a-8a4e-4151-80b6-651bbea7c418": {
                "pk": "5301323a-8a4e-4151-80b6-651bbea7c418",
                "name": "Haoxuan You",
                "collaborators": [
                    "Zhecan Wang",
                    "Shih-Fu Chang",
                    "Kai-Wei Chang",
                    "Rui Sun",
                    "N. Codella",
                    "Wenhao Li",
                    "Hammad A. Ayyubi",
                    "Luowei Zhou",
                    "Bin Xiao",
                    "Lu Yuan"
                ],
                "domain": [
                    "Vision-Language Understanding",
                    "Multimodal Learning",
                    "Commonsense Reasoning",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "JourneyBench: A Challenging One-Stop Vision-Language Understanding Benchmark of Generated Images",
                        "abstract": "Existing vision-language understanding benchmarks largely consist of images of objects in their usual contexts. As a consequence, recent multimodal large language models can perform well with only a shallow visual understanding by relying on background language biases. Thus, strong performance on these benchmarks does not necessarily correlate with strong visual understanding. In this paper, we release JourneyBench, a comprehensive human-annotated benchmark of generated images designed to assess the model's fine-grained multimodal reasoning abilities across five tasks: complementary multimodal chain of thought, multi-image VQA, imaginary image captioning, VQA with hallucination triggers, and fine-grained retrieval with sample-specific distractors. Unlike existing benchmarks, JourneyBench explicitly requires fine-grained multimodal reasoning in unusual imaginary scenarios where language bias and holistic image gist are insufficient. We benchmark state-of-the-art models on JourneyBench and analyze performance along a number of fine-grained dimensions. Results across all five tasks show that JourneyBench is exceptionally challenging for even the best models, indicating that models' visual reasoning abilities are not as strong as they first appear. We discuss the implications of our findings and propose avenues for further research."
                    },
                    {
                        "title": "Detecting Multimodal Situations with Insufficient Context and Abstaining from Baseless Predictions",
                        "abstract": "Despite the widespread adoption of Vision-Language Understanding (VLU) benchmarks such as VQA v2, OKVQA, A-OKVQA, GQA, VCR, SWAG, and VisualCOMET, our analysis reveals a pervasive issue affecting their integrity: these benchmarks contain samples where answers rely on assumptions unsupported by the provided context. Training models on such data foster biased learning and hallucinations as models tend to make similar unwarranted assumptions. To address this issue, we collect contextual data for each sample whenever available and train a context selection module to facilitate evidence-based model predictions. Strong improvements across multiple benchmarks demonstrate the effectiveness of our approach. Further, we develop a general-purpose Context-AwaRe Abstention (CARA) detector to identify samples lacking sufficient context and enhance model accuracy by abstaining from responding if the required context is absent. CARA exhibits generalization to new benchmarks it wasn't trained on, underscoring its utility for future VLU benchmarks in detecting or cleaning samples with inadequate context. Finally, we curate a Context Ambiguity and Sufficiency Evaluation (CASE) set to benchmark the performance of insufficient context detectors. Overall, our work represents a significant advancement in ensuring that vision-language models generate trustworthy and evidence-based outputs in complex real-world scenarios."
                    },
                    {
                        "title": "Ferret-v2: An Improved Baseline for Referring and Grounding with Large Language Models",
                        "abstract": "While Ferret seamlessly integrates regional understanding into the Large Language Model (LLM) to facilitate its referring and grounding capability, it poses certain limitations: constrained by the pre-trained fixed visual encoder and failed to perform well on broader tasks. In this work, we unveil Ferret-v2, a significant upgrade to Ferret, with three key designs. (1) Any resolution grounding and referring: A flexible approach that effortlessly handles higher image resolution, improving the model's ability to process and understand images in greater detail. (2) Multi-granularity visual encoding: By integrating the additional DINOv2 encoder, the model learns better and diverse underlying contexts for global and fine-grained visual information. (3) A three-stage training paradigm: Besides image-caption alignment, an additional stage is proposed for high-resolution dense alignment before the final instruction tuning. Experiments show that Ferret-v2 provides substantial improvements over Ferret and other state-of-the-art methods, thanks to its high-resolution scaling and fine-grained visual processing."
                    },
                    {
                        "title": "UniFine: A Unified and Fine-grained Approach for Zero-shot Vision-Language Understanding",
                        "abstract": "Vision-language tasks, such as VQA, SNLI-VE, and VCR are challenging because they require the model's reasoning ability to understand the semantics of the visual world and natural language. Supervised methods working for vision-language tasks have been well-studied. However, solving these tasks in a zero-shot setting is less explored. Since Contrastive Language-Image Pre-training (CLIP) has shown remarkable zero-shot performance on image-text matching, previous works utilized its strong zero-shot ability by converting vision-language tasks into an image-text matching problem, and they mainly consider global-level matching (e.g., the whole image or sentence). However, we find visual and textual fine-grained information, e.g., keywords in the sentence and objects in the image, can be fairly informative for semantics understanding. Inspired by this, we propose a unified framework to take advantage of the fine-grained information for zero-shot vision-language learning, covering multiple tasks such as VQA, SNLI-VE, and VCR. Our experiments show that our framework outperforms former zero-shot methods on VQA and achieves substantial improvement on SNLI-VE and VCR. Furthermore, our ablation studies confirm the effectiveness and generalizability of our proposed method. Code will be available at https://github.com/ThreeSR/UniFine"
                    },
                    {
                        "title": "IdealGPT: Iteratively Decomposing Vision and Language Reasoning via Large Language Models",
                        "abstract": "The field of vision-and-language (VL) understanding has made unprecedented progress with end-to-end large pre-trained VL models (VLMs). However, they still fall short in zero-shot reasoning tasks that require multi-step inferencing. To achieve this goal, previous works resort to a divide-and-conquer pipeline. In this paper, we argue that previous efforts have several inherent shortcomings: 1) They rely on domain-specific sub-question decomposing models. 2) They force models to predict the final answer even if the sub-questions or sub-answers provide insufficient information. We address these limitations via IdealGPT, a framework that iteratively decomposes VL reasoning using large language models (LLMs). Specifically, IdealGPT utilizes an LLM to generate sub-questions, a VLM to provide corresponding sub-answers, and another LLM to reason to achieve the final answer. These three modules perform the divide-and-conquer procedure iteratively until the model is confident about the final answer to the main question. We evaluate IdealGPT on multiple challenging VL reasoning tasks under a zero-shot setting. In particular, our IdealGPT outperforms the best existing GPT-4-like models by an absolute 10% on VCR and 15% on SNLI-VE. Code is available at https://github.com/Hxyou/IdealGPT"
                    },
                    {
                        "title": "Dataset Bias Mitigation in Multiple-Choice Visual Question Answering and Beyond",
                        "abstract": "Vision-language (VL) understanding tasks evaluate models' comprehension of complex visual scenes through multiple-choice questions. However, we have identified two dataset biases that models can exploit as shortcuts to resolve various VL tasks correctly without proper understanding. The first type of dataset bias is \\emph{Unbalanced Matching} bias, where the correct answer overlaps the question and image more than the incorrect answers. The second type of dataset bias is \\emph{Distractor Similarity} bias, where incorrect answers are overly dissimilar to the correct answer but significantly similar to other incorrect answers within the same sample. To address these dataset biases, we first propose Adversarial Data Synthesis (ADS) to generate synthetic training and debiased evaluation data. We then introduce Intra-sample Counterfactual Training (ICT) to assist models in utilizing the synthesized training data, particularly the counterfactual data, via focusing on intra-sample differentiation. Extensive experiments demonstrate the effectiveness of ADS and ICT in consistently improving model performance across different benchmarks, even in domain-shifted scenarios."
                    },
                    {
                        "title": "CoBIT: A Contrastive Bi-directional Image-Text Generation Model",
                        "abstract": "The field of vision and language has witnessed a proliferation of pre-trained foundation models. Most existing methods are independently pre-trained with contrastive objective like CLIP, image-to-text generative objective like PaLI, or text-to-image generative objective like Parti. However, the three objectives can be pre-trained on the same data, image-text pairs, and intuitively they complement each other as contrasting provides global alignment capacity and generation grants fine-grained understanding. In this work, we present a Contrastive Bi-directional Image-Text generation model (CoBIT), which attempts to unify the three pre-training objectives in one framework. Specifically, CoBIT employs a novel unicoder-decoder structure, consisting of an image unicoder, a text unicoder and a cross-modal decoder. The image/text unicoders can switch between encoding and decoding in different tasks, enabling flexibility and shared knowledge that benefits both image-to-text and text-to-image generations. CoBIT achieves superior performance in image understanding, image-text understanding (Retrieval, Captioning, VQA, SNLI-VE) and text-based content creation, particularly in zero-shot scenarios. For instance, 82.7% in zero-shot ImageNet classification, 9.37 FID score in zero-shot text-to-image generation and 44.8 CIDEr in zero-shot captioning."
                    },
                    {
                        "title": "CLIP-TD: CLIP Targeted Distillation for Vision-Language Tasks",
                        "abstract": "Contrastive language-image pretraining (CLIP) links vision and language modalities into a unified embedding space, yielding the tremendous potential for vision-language (VL) tasks. While early concurrent works have begun to study this potential on a subset of tasks, important questions remain: 1) What is the benefit of CLIP on unstudied VL tasks? 2) Does CLIP provide benefit in low-shot or domain-shifted scenarios? 3) Can CLIP improve existing approaches without impacting inference or pretraining complexity? In this work, we seek to answer these questions through two key contributions. First, we introduce an evaluation protocol that includes Visual Commonsense Reasoning (VCR), Visual Entailment (SNLI-VE), and Visual Question Answering (VQA), across a variety of data availability constraints and conditions of domain shift. Second, we propose an approach, named CLIP Targeted Distillation (CLIP-TD), to intelligently distill knowledge from CLIP into existing architectures using a dynamically weighted objective applied to adaptively selected tokens per instance. Experiments demonstrate that our proposed CLIP-TD leads to exceptional gains in the low-shot (up to 51.9%) and domain-shifted (up to 71.3%) conditions of VCR, while simultaneously improving performance under standard fully-supervised conditions (up to 2%), achieving state-of-art performance on VCR compared to other single models that are pretrained with image-text data only. On SNLI-VE, CLIP-TD produces significant gains in low-shot conditions (up to 6.6%) as well as fully supervised (up to 3%). On VQA, CLIP-TD provides improvement in low-shot (up to 9%), and in fully-supervised (up to 1.3%). Finally, CLIP-TD outperforms concurrent works utilizing CLIP for finetuning, as well as baseline naive distillation approaches. Code will be made available."
                    },
                    {
                        "title": "Multimodal Adaptive Distillation for Leveraging Unimodal Encoders for Vision-Language Tasks",
                        "abstract": "Cross-modal encoders for vision-language (VL) tasks are often pretrained with carefully curated vision-language datasets. While these datasets reach an order of 10 million samples, the labor cost is prohibitive to scale further. Conversely, unimodal encoders are pretrained with simpler annotations that are less cost-prohibitive, achieving scales of hundreds of millions to billions. As a result, unimodal encoders have achieved state-of-art (SOTA) on many downstream tasks. However, challenges remain when applying to VL tasks. The pretraining data is not optimal for cross-modal architectures and requires heavy computational resources. In addition, unimodal architectures lack cross-modal interactions that have demonstrated significant benefits for VL tasks. Therefore, how to best leverage pretrained unimodal encoders for VL tasks is still an area of active research. In this work, we propose a method to leverage unimodal vision and text encoders for VL tasks that augment existing VL approaches while conserving computational complexity. Specifically, we propose Multimodal Adaptive Distillation (MAD), which adaptively distills useful knowledge from pretrained encoders to cross-modal VL encoders. Second, to better capture nuanced impacts on VL task performance, we introduce an evaluation protocol that includes Visual Commonsense Reasoning (VCR), Visual Entailment (SNLI-VE), and Visual Question Answering (VQA), across a variety of data constraints and conditions of domain shift. Experiments demonstrate that MAD leads to consistent gains in the low-shot, domain-shifted, and fully-supervised conditions on VCR, SNLI-VE, and VQA, achieving SOTA performance on VCR compared to other single models pretrained with image-text data. Finally, MAD outperforms concurrent works utilizing pretrained vision encoder from CLIP. Code will be made available."
                    },
                    {
                        "title": "Find Someone Who: Visual Commonsense Understanding in Human-Centric Grounding",
                        "abstract": "From a visual scene containing multiple people, human is able to distinguish each individual given the context descriptions about what happened before, their mental/physical states or intentions, etc. Above ability heavily relies on human-centric commonsense knowledge and reasoning. For example, if asked to identify the\"person who needs healing\"in an image, we need to first know that they usually have injuries or suffering expressions, then find the corresponding visual clues before finally grounding the person. We present a new commonsense task, Human-centric Commonsense Grounding, that tests the models' ability to ground individuals given the context descriptions about what happened before, and their mental/physical states or intentions. We further create a benchmark, HumanCog, a dataset with 130k grounded commonsensical descriptions annotated on 67k images, covering diverse types of commonsense and visual scenes. We set up a context-object-aware method as a strong baseline that outperforms previous pre-trained and non-pretrained models. Further analysis demonstrates that rich visual commonsense and powerful integration of multi-modal commonsense are essential, which sheds light on future works. Data and code will be available https://github.com/Hxyou/HumanCog."
                    },
                    {
                        "title": "Rethinking Network Design and Local Geometry in Point Cloud: A Simple Residual MLP Framework",
                        "abstract": "Point cloud analysis is challenging due to irregularity and unordered data structure. To capture the 3D geometries, prior works mainly rely on exploring sophisticated local geometric extractors using convolution, graph, or attention mechanisms. These methods, however, incur unfavorable latency during inference, and the performance saturates over the past few years. In this paper, we present a novel perspective on this task. We notice that detailed local geometrical information probably is not the key to point cloud analysis -- we introduce a pure residual MLP network, called PointMLP, which integrates no sophisticated local geometrical extractors but still performs very competitively. Equipped with a proposed lightweight geometric affine module, PointMLP delivers the new state-of-the-art on multiple datasets. On the real-world ScanObjectNN dataset, our method even surpasses the prior best method by 3.3% accuracy. We emphasize that PointMLP achieves this strong performance without any sophisticated operations, hence leading to a superior inference speed. Compared to most recent CurveNet, PointMLP trains 2x faster, tests 7x faster, and is more accurate on ModelNet40 benchmark. We hope our PointMLP may help the community towards a better understanding of point cloud analysis. The code is available at https://github.com/ma-xu/pointMLP-pytorch."
                    },
                    {
                        "title": "Bridging the Gap between Recognition-level Pre-training and Commonsensical Vision-language Tasks",
                        "abstract": "Large-scale visual-linguistic pre-training aims to capture the generic representations from multimodal features, which are essential for downstream vision-language tasks. Existing methods mostly focus on learning the semantic connections between visual objects and linguistic content, which tend to be recognitionlevel information and may not be sufficient for commonsensical reasoning tasks like VCR. In this paper, we propose a novel commonsensical vision-language pre-training framework to bridge the gap. We first augment the conventional image-caption pre-training datasets with commonsense inferences from a visuallinguistic GPT-2. To pre-train models on image, caption and commonsense inferences together, we propose two new tasks: masked commonsense modeling (MCM) and commonsense type prediction (CTP). To reduce the shortcut effect between captions and commonsense inferences, we further introduce the domain-wise adaptive masking that dynamically adjusts the masking ratio. Experimental results on downstream tasks, VCR and VQA, show the improvement of our pre-training strategy over previous methods. Human evaluation also validates the relevance, informativeness, and diversity of the generated commonsense inferences. Overall, we demonstrate the potential of incorporating commonsense knowledge into the conventional recognition-level visual-linguistic pre-training."
                    },
                    {
                        "title": "Understanding ME? Multimodal Evaluation for Fine-grained Visual Commonsense",
                        "abstract": "Visual commonsense understanding requires Vision Language (VL) models to not only understand image and text but also cross-reference in-between to fully integrate and achieve comprehension of the visual scene described. Recently, various approaches have been developed and have achieved high performance on visual commonsense benchmarks. However, it is unclear whether the models really understand the visual scene and underlying commonsense knowledge due to limited evaluation data resources. To provide an in-depth analysis, we present a Multimodal Evaluation (ME) pipeline to automatically generate question-answer pairs to test models\u2019 understanding of the visual scene, text, and related knowledge. We then take a step further to show that training with the ME data boosts the model\u2019s performance in standard VCR evaluation. Lastly, our in-depth analysis and comparison reveal interesting findings: (1) semantically low-level information can assist the learning of high-level information but not the opposite; (2) visual information is generally under utilization compared with text."
                    },
                    {
                        "title": "Learning Visual Representation from Modality-Shared Contrastive Language-Image Pre-training",
                        "abstract": "Large-scale multi-modal contrastive pre-training has demonstrated great utility to learn transferable features for a range of downstream tasks by mapping multiple modalities into a shared embedding space. Typically, this has employed separate encoders for each modality. However, recent work suggests that transformers can support learning across multiple modalities and allow knowledge sharing. Inspired by this, we investigate a variety of Modality-Shared Contrastive Language-Image Pre-training (MS-CLIP) frameworks. More specifically, we question how many parameters of a transformer model can be shared across modalities during contrastive pre-training, and rigorously examine architectural design choices that position the proportion of parameters shared along a spectrum. In studied conditions, we observe that a mostly unified encoder for vision and language signals outperforms all other variations that separate more parameters. Additionally, we find that light-weight modality-specific parallel modules further improve performance. Experimental results show that the proposed MS-CLIP approach outperforms vanilla CLIP by up to 13\\% relative in zero-shot ImageNet classification (pre-trained on YFCC-100M), while simultaneously supporting a reduction of parameters. In addition, our approach outperforms vanilla CLIP by 1.6 points in linear probing on a collection of 24 downstream vision tasks. Furthermore, we discover that sharing parameters leads to semantic concepts from different modalities being encoded more closely in the embedding space, facilitating the transferring of common semantic structure (e.g., attention patterns) from language to vision. Code is available at \\href{https://github.com/Hxyou/MSCLIP}{URL}."
                    },
                    {
                        "title": "SGEITL: Scene Graph Enhanced Image-Text Learning for Visual Commonsense Reasoning",
                        "abstract": "Answering complex questions about images is an ambitious goal for machine intelligence, which requires a joint understanding of images, text, and commonsense knowledge, as well as a strong reasoning ability. Recently, multimodal Transformers have made a great progress in the task of Visual Commonsense Reasoning (VCR), by jointly understanding visual objects and text tokens through layers of cross-modality attention. However, these approaches do not utilize the rich structure of the scene and the interactions between objects which are essential in answering complex commonsense questions. We propose a Scene Graph Enhanced Image-Text Learning (SGEITL) framework to incorporate visual scene graph in commonsense reasoning. In order to exploit the scene graph structure, at the model structure level, we propose a multihop graph transformer for regularizing attention interaction among hops. As for pre-training, a scene-graph-aware pre-training method is proposed to leverage structure knowledge extracted in visual scene graph. Moreover, we introduce a method to train and generate domain relevant visual scene graph using textual annotations in a weakly-supervised manner. Extensive experiments on VCR and other tasks show significant performance boost compared with the state-of-the-art methods, and prove the efficacy of each proposed component."
                    },
                    {
                        "title": "Graph-MLP: Node Classification without Message Passing in Graph",
                        "abstract": "Graph Neural Network (GNN) has been demonstrated its effectiveness in dealing with non-Euclidean structural data. Both spatial-based and spectral-based GNNs are relying on adjacency matrix to guide message passing among neighbors during feature aggregation. Recent works have mainly focused on powerful message passing modules, however, in this paper, we show that none of the message passing modules is necessary. Instead, we propose a pure multilayer-perceptron-based framework, Graph-MLP with the supervision signal leveraging graph structure, which is sufficient for learning discriminative node representation. In model-level, Graph-MLP only includes multi-layer perceptrons, activation function, and layer normalization. In the loss level, we design a neighboring contrastive (NContrast) loss to bridge the gap between GNNs and MLPs by utilizing the adjacency information implicitly. This design allows our model to be lighter and more robust when facing large-scale graph data and corrupted adjacency information. Extensive experiments prove that even without adjacency information in testing phase, our framework can still reach comparable and even superior performance against the state-of-the-art models in the graph node classification task."
                    },
                    {
                        "title": "Unsupervised Vision-and-Language Pre-training Without Parallel Images and Captions",
                        "abstract": "Pre-trained contextual vision-and-language (V&L) models have achieved impressive performance on various benchmarks. However, existing models require a large amount of parallel image-caption data for pre-training. Such data are costly to collect and require cumbersome curation. Inspired by unsupervised machine translation, we investigate if a strong V&L representation model can be learned through unsupervised pre-training without image-caption corpora. In particular, we propose to conduct \u201cmask-and-predict\u201d pre-training on text-only and image-only corpora and introduce the object tags detected by an object recognition model as anchor points to bridge two modalities. We find that such a simple approach achieves performance close to a model pre-trained with aligned data, on four English V&L benchmarks. Our work challenges the widely held notion that aligned data is necessary for V&L pre-training, while significantly reducing the amount of supervision needed for V&L models."
                    },
                    {
                        "title": "Weakly-supervised VisualBERT: Pre-training without Parallel Images and Captions",
                        "abstract": "Pre-trained contextual vision-and-language (V&L) models have brought impressive performance improvement on various benchmarks. However, the paired text-image data required for pre-training are hard to collect and scale up. We investigate if a strong V&L representation model can be learned without text-image pairs. We propose Weakly-supervised VisualBERT with the key idea of conducting \"mask-and-predict\" pre-training on language-only and image-only corpora. Additionally, we introduce the object tags detected by an object recognition model as anchor points to bridge two modalities. Evaluation on four V&L benchmarks shows that Weakly-supervised VisualBERT achieves similar performance with a model pre-trained with paired data. Besides, pre-training on more image-only data further improves a model that already has access to aligned data, suggesting the possibility of utilizing billions of raw images available to enhance V&L models."
                    }
                ]
            },
            "5133aa8c-85cf-40e3-9a10-b71a0a9e7957": {
                "pk": "5133aa8c-85cf-40e3-9a10-b71a0a9e7957",
                "name": "Haotian Zhang",
                "collaborators": [
                    "Yinfei Yang",
                    "Zhe Gan",
                    "Bowen Zhang",
                    "Zhengfeng Lai",
                    "Hong-You Chen",
                    "Keen You",
                    "Xianzhi Du",
                    "Meng Cao",
                    "Philipp Dufter",
                    "Yanghao Li"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Vision-Language Models",
                    "Domain Adaptation",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "Empowering Unsupervised Domain Adaptation with Large-scale Pre-trained Vision-Language Models",
                        "abstract": "Unsupervised Domain Adaptation (UDA) aims to leverage the labeled source domain to solve the tasks on the unlabeled target domain. Traditional UDA methods face the challenge of the tradeoff between domain alignment and semantic class discriminability, especially when a large domain gap exists between the source and target domains. The efforts of applying large-scale pre-training to bridge the domain gaps remain limited. In this work, we propose that Vision-Language Models (VLMs) can empower UDA tasks due to their training pattern with language alignment and their large-scale pre-trained datasets. For example, CLIP and GLIP have shown promising zero-shot generalization in classification and detection tasks. However, directly fine-tuning these VLMs into downstream tasks may be computationally expensive and not scalable if we have multiple domains that need to be adapted. Therefore, in this work, we first study an efficient adaption of VLMs to preserve the original knowledge while maximizing its flexibility for learning new knowledge. Then, we design a domain-aware pseudo-labeling scheme tailored to VLMs for domain disentanglement. We show the superiority of the proposed methods in four UDA-classification and two UDA-detection benchmarks, with a significant improvement (+9.9%) on DomainNet."
                    },
                    {
                        "title": "Contrastive Localized Language-Image Pre-Training",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) has been a celebrated method for training vision encoders to generate image/text representations facilitating various applications. Recently, CLIP has been widely adopted as the vision backbone of multimodal large language models (MLLMs) to connect image inputs for language interactions. The success of CLIP as a vision-language foundation model relies on aligning web-crawled noisy text annotations at image levels. Nevertheless, such criteria may become insufficient for downstream tasks in need of fine-grained vision representations, especially when region-level understanding is demanding for MLLMs. In this paper, we improve the localization capability of CLIP with several advances. We propose a pre-training method called Contrastive Localized Language-Image Pre-training (CLOC) by complementing CLIP with region-text contrastive loss and modules. We formulate a new concept, promptable embeddings, of which the encoder produces image embeddings easy to transform into region representations given spatial hints. To support large-scale pre-training, we design a visually-enriched and spatially-localized captioning framework to effectively generate region-text pseudo-labels at scale. By scaling up to billions of annotated images, CLOC enables high-quality regional embeddings for image region recognition and retrieval tasks, and can be a drop-in replacement of CLIP to enhance MLLMs, especially on referring and grounding tasks."
                    },
                    {
                        "title": "MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning",
                        "abstract": "We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development."
                    },
                    {
                        "title": "Ferret-UI: Grounded Mobile UI Understanding with Multimodal LLMs",
                        "abstract": "Recent advancements in multimodal large language models (MLLMs) have been noteworthy, yet, these general-domain MLLMs often fall short in their ability to comprehend and interact effectively with user interface (UI) screens. In this paper, we present Ferret-UI, a new MLLM tailored for enhanced understanding of mobile UI screens, equipped with referring, grounding, and reasoning capabilities. Given that UI screens typically exhibit a more elongated aspect ratio and contain smaller objects of interest (e.g., icons, texts) than natural images, we incorporate\"any resolution\"on top of Ferret to magnify details and leverage enhanced visual features. Specifically, each screen is divided into 2 sub-images based on the original aspect ratio (i.e., horizontal division for portrait screens and vertical division for landscape screens). Both sub-images are encoded separately before being sent to LLMs. We meticulously gather training samples from an extensive range of elementary UI tasks, such as icon recognition, find text, and widget listing. These samples are formatted for instruction-following with region annotations to facilitate precise referring and grounding. To augment the model's reasoning ability, we further compile a dataset for advanced tasks, including detailed description, perception/interaction conversations, and function inference. After training on the curated datasets, Ferret-UI exhibits outstanding comprehension of UI screens and the capability to execute open-ended instructions. For model evaluation, we establish a comprehensive benchmark encompassing all the aforementioned tasks. Ferret-UI excels not only beyond most open-source UI MLLMs, but also surpasses GPT-4V on all the elementary UI tasks."
                    },
                    {
                        "title": "Ferret-UI 2: Mastering Universal User Interface Understanding Across Platforms",
                        "abstract": "Building a generalist model for user interface (UI) understanding is challenging due to various foundational issues, such as platform diversity, resolution variation, and data limitation. In this paper, we introduce Ferret-UI 2, a multimodal large language model (MLLM) designed for universal UI understanding across a wide range of platforms, including iPhone, Android, iPad, Webpage, and AppleTV. Building on the foundation of Ferret-UI, Ferret-UI 2 introduces three key innovations: support for multiple platform types, high-resolution perception through adaptive scaling, and advanced task training data generation powered by GPT-4o with set-of-mark visual prompting. These advancements enable Ferret-UI 2 to perform complex, user-centered interactions, making it highly versatile and adaptable for the expanding diversity of platform ecosystems. Extensive empirical experiments on referring, grounding, user-centric advanced tasks (comprising 9 subtasks $\\times$ 5 platforms), GUIDE next-action prediction dataset, and GUI-World multi-platform benchmark demonstrate that Ferret-UI 2 significantly outperforms Ferret-UI, and also shows strong cross-platform transfer capabilities."
                    },
                    {
                        "title": "MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training",
                        "abstract": "In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting."
                    },
                    {
                        "title": "How Easy is It to Fool Your Multimodal LLMs? An Empirical Analysis on Deceptive Prompts",
                        "abstract": "The remarkable advancements in Multimodal Large Language Models (MLLMs) have not rendered them immune to challenges, particularly in the context of handling deceptive information in prompts, thus producing hallucinated responses under such conditions. To quantitatively assess this vulnerability, we present MAD-Bench, a carefully curated benchmark that contains 1000 test samples divided into 5 categories, such as non-existent objects, count of objects, and spatial relationship. We provide a comprehensive analysis of popular MLLMs, ranging from GPT-4v, Reka, Gemini-Pro, to open-sourced models, such as LLaVA-NeXT and MiniCPM-Llama3. Empirically, we observe significant performance gaps between GPT-4o and other models; and previous robust instruction-tuned models are not effective on this new benchmark. While GPT-4o achieves 82.82% accuracy on MAD-Bench, the accuracy of any other model in our experiments ranges from 9% to 50%. We further propose a remedy that adds an additional paragraph to the deceptive prompts to encourage models to think twice before answering the question. Surprisingly, this simple method can even double the accuracy; however, the absolute numbers are still too low to be satisfactory. We hope MAD-Bench can serve as a valuable benchmark to stimulate further research to enhance model resilience against deceptive prompts."
                    },
                    {
                        "title": "Revisit Large-Scale Image-Caption Data in Pre-training Multimodal Foundation Models",
                        "abstract": "Recent advancements in multimodal models highlight the value of rewritten captions for improving performance, yet key challenges remain. For example, while synthetic captions often provide superior quality and image-text alignment, it is not clear whether they can fully replace AltTexts: the role of synthetic captions and their interaction with original web-crawled AltTexts in pre-training is still not well understood. Moreover, different multimodal foundation models may have unique preferences for specific caption formats, but efforts to identify the optimal captions for each model remain limited. In this work, we propose a novel, controllable, and scalable captioning pipeline designed to generate diverse caption formats tailored to various multimodal models. By examining Short Synthetic Captions (SSC) towards Dense Synthetic Captions (DSC+) as case studies, we systematically explore their effects and interactions with AltTexts across models such as CLIP, multimodal LLMs, and diffusion models. Our findings reveal that a hybrid approach that keeps both synthetic captions and AltTexts can outperform the use of synthetic captions alone, improving both alignment and performance, with each model demonstrating preferences for particular caption formats. This comprehensive analysis provides valuable insights into optimizing captioning strategies, thereby advancing the pre-training of multimodal foundation models."
                    },
                    {
                        "title": "Ferret-v2: An Improved Baseline for Referring and Grounding with Large Language Models",
                        "abstract": "While Ferret seamlessly integrates regional understanding into the Large Language Model (LLM) to facilitate its referring and grounding capability, it poses certain limitations: constrained by the pre-trained fixed visual encoder and failed to perform well on broader tasks. In this work, we unveil Ferret-v2, a significant upgrade to Ferret, with three key designs. (1) Any resolution grounding and referring: A flexible approach that effortlessly handles higher image resolution, improving the model's ability to process and understand images in greater detail. (2) Multi-granularity visual encoding: By integrating the additional DINOv2 encoder, the model learns better and diverse underlying contexts for global and fine-grained visual information. (3) A three-stage training paradigm: Besides image-caption alignment, an additional stage is proposed for high-resolution dense alignment before the final instruction tuning. Experiments show that Ferret-v2 provides substantial improvements over Ferret and other state-of-the-art methods, thanks to its high-resolution scaling and fine-grained visual processing."
                    },
                    {
                        "title": "MM-Ego: Towards Building Egocentric Multimodal LLMs",
                        "abstract": "This research aims to comprehensively explore building a multimodal foundation model for egocentric video understanding. To achieve this goal, we work on three fronts. First, as there is a lack of QA data for egocentric video understanding, we develop a data engine that efficiently generates 7M high-quality QA samples for egocentric videos ranging from 30 seconds to one hour long, based on human-annotated data. This is currently the largest egocentric QA dataset. Second, we contribute a challenging egocentric QA benchmark with 629 videos and 7,026 questions to evaluate the models' ability in recognizing and memorizing visual details across videos of varying lengths. We introduce a new de-biasing evaluation method to help mitigate the unavoidable language bias present in the models being evaluated. Third, we propose a specialized multimodal architecture featuring a novel\"Memory Pointer Prompting\"mechanism. This design includes a global glimpse step to gain an overarching understanding of the entire video and identify key visual information, followed by a fallback step that utilizes the key visual information to generate responses. This enables the model to more effectively comprehend extended video content. With the data, benchmark, and model, we successfully build MM-Ego, an egocentric multimodal LLM that shows powerful performance on egocentric video understanding."
                    },
                    {
                        "title": "Improve Vision Language Model Chain-of-thought Reasoning",
                        "abstract": "Chain-of-thought (CoT) reasoning in vision language models (VLMs) is crucial for improving interpretability and trustworthiness. However, current training recipes lack robust CoT reasoning data, relying on datasets dominated by short annotations with minimal rationales. In this work, we show that training VLM on short answers does not generalize well to reasoning tasks that require more detailed responses. To address this, we propose a two-fold approach. First, we distill rationales from GPT-4o model to enrich the training data and fine-tune VLMs, boosting their CoT performance. Second, we apply reinforcement learning to further calibrate reasoning quality. Specifically, we construct positive (correct) and negative (incorrect) pairs of model-generated reasoning chains, by comparing their predictions with annotated short answers. Using this pairwise data, we apply the Direct Preference Optimization algorithm to refine the model's reasoning abilities. Our experiments demonstrate significant improvements in CoT reasoning on benchmark datasets and better generalization to direct answer prediction as well. This work emphasizes the importance of incorporating detailed rationales in training and leveraging reinforcement learning to strengthen the reasoning capabilities of VLMs."
                    }
                ]
            },
            "940469a9-046c-4714-949b-9d1b3e6fe801": {
                "pk": "940469a9-046c-4714-949b-9d1b3e6fe801",
                "name": "Zhe Gan",
                "collaborators": [
                    "Yinfei Yang",
                    "Haotian Zhang",
                    "Bowen Zhang",
                    "Hong-You Chen",
                    "Zhengfeng Lai",
                    "Peter Grasch",
                    "Keen You",
                    "Yanghao Li",
                    "J. Fauconnier",
                    "Haoxuan You"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Large Language Models",
                    "Vision-Language Models",
                    "Image Understanding"
                ],
                "publications": [
                    {
                        "title": "Contrastive Localized Language-Image Pre-Training",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) has been a celebrated method for training vision encoders to generate image/text representations facilitating various applications. Recently, CLIP has been widely adopted as the vision backbone of multimodal large language models (MLLMs) to connect image inputs for language interactions. The success of CLIP as a vision-language foundation model relies on aligning web-crawled noisy text annotations at image levels. Nevertheless, such criteria may become insufficient for downstream tasks in need of fine-grained vision representations, especially when region-level understanding is demanding for MLLMs. In this paper, we improve the localization capability of CLIP with several advances. We propose a pre-training method called Contrastive Localized Language-Image Pre-training (CLOC) by complementing CLIP with region-text contrastive loss and modules. We formulate a new concept, promptable embeddings, of which the encoder produces image embeddings easy to transform into region representations given spatial hints. To support large-scale pre-training, we design a visually-enriched and spatially-localized captioning framework to effectively generate region-text pseudo-labels at scale. By scaling up to billions of annotated images, CLOC enables high-quality regional embeddings for image region recognition and retrieval tasks, and can be a drop-in replacement of CLIP to enhance MLLMs, especially on referring and grounding tasks."
                    },
                    {
                        "title": "MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning",
                        "abstract": "We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development."
                    },
                    {
                        "title": "Ferret-UI: Grounded Mobile UI Understanding with Multimodal LLMs",
                        "abstract": "Recent advancements in multimodal large language models (MLLMs) have been noteworthy, yet, these general-domain MLLMs often fall short in their ability to comprehend and interact effectively with user interface (UI) screens. In this paper, we present Ferret-UI, a new MLLM tailored for enhanced understanding of mobile UI screens, equipped with referring, grounding, and reasoning capabilities. Given that UI screens typically exhibit a more elongated aspect ratio and contain smaller objects of interest (e.g., icons, texts) than natural images, we incorporate\"any resolution\"on top of Ferret to magnify details and leverage enhanced visual features. Specifically, each screen is divided into 2 sub-images based on the original aspect ratio (i.e., horizontal division for portrait screens and vertical division for landscape screens). Both sub-images are encoded separately before being sent to LLMs. We meticulously gather training samples from an extensive range of elementary UI tasks, such as icon recognition, find text, and widget listing. These samples are formatted for instruction-following with region annotations to facilitate precise referring and grounding. To augment the model's reasoning ability, we further compile a dataset for advanced tasks, including detailed description, perception/interaction conversations, and function inference. After training on the curated datasets, Ferret-UI exhibits outstanding comprehension of UI screens and the capability to execute open-ended instructions. For model evaluation, we establish a comprehensive benchmark encompassing all the aforementioned tasks. Ferret-UI excels not only beyond most open-source UI MLLMs, but also surpasses GPT-4V on all the elementary UI tasks."
                    },
                    {
                        "title": "How Easy is It to Fool Your Multimodal LLMs? An Empirical Analysis on Deceptive Prompts",
                        "abstract": "The remarkable advancements in Multimodal Large Language Models (MLLMs) have not rendered them immune to challenges, particularly in the context of handling deceptive information in prompts, thus producing hallucinated responses under such conditions. To quantitatively assess this vulnerability, we present MAD-Bench, a carefully curated benchmark that contains 1000 test samples divided into 5 categories, such as non-existent objects, count of objects, and spatial relationship. We provide a comprehensive analysis of popular MLLMs, ranging from GPT-4v, Reka, Gemini-Pro, to open-sourced models, such as LLaVA-NeXT and MiniCPM-Llama3. Empirically, we observe significant performance gaps between GPT-4o and other models; and previous robust instruction-tuned models are not effective on this new benchmark. While GPT-4o achieves 82.82% accuracy on MAD-Bench, the accuracy of any other model in our experiments ranges from 9% to 50%. We further propose a remedy that adds an additional paragraph to the deceptive prompts to encourage models to think twice before answering the question. Surprisingly, this simple method can even double the accuracy; however, the absolute numbers are still too low to be satisfactory. We hope MAD-Bench can serve as a valuable benchmark to stimulate further research to enhance model resilience against deceptive prompts."
                    },
                    {
                        "title": "Revisit Large-Scale Image-Caption Data in Pre-training Multimodal Foundation Models",
                        "abstract": "Recent advancements in multimodal models highlight the value of rewritten captions for improving performance, yet key challenges remain. For example, while synthetic captions often provide superior quality and image-text alignment, it is not clear whether they can fully replace AltTexts: the role of synthetic captions and their interaction with original web-crawled AltTexts in pre-training is still not well understood. Moreover, different multimodal foundation models may have unique preferences for specific caption formats, but efforts to identify the optimal captions for each model remain limited. In this work, we propose a novel, controllable, and scalable captioning pipeline designed to generate diverse caption formats tailored to various multimodal models. By examining Short Synthetic Captions (SSC) towards Dense Synthetic Captions (DSC+) as case studies, we systematically explore their effects and interactions with AltTexts across models such as CLIP, multimodal LLMs, and diffusion models. Our findings reveal that a hybrid approach that keeps both synthetic captions and AltTexts can outperform the use of synthetic captions alone, improving both alignment and performance, with each model demonstrating preferences for particular caption formats. This comprehensive analysis provides valuable insights into optimizing captioning strategies, thereby advancing the pre-training of multimodal foundation models."
                    },
                    {
                        "title": "SlowFast-LLaVA: A Strong Training-Free Baseline for Video Large Language Models",
                        "abstract": "We propose SlowFast-LLaVA (or SF-LLaVA for short), a training-free video large language model (LLM) that can jointly capture detailed spatial semantics and long-range temporal context without exceeding the token budget of commonly used LLMs. This is realized by using a two-stream SlowFast design of inputs for Video LLMs to aggregate features from sampled frames in an effective way. Specifically, the Slow pathway extracts features at a low frame rate while keeping as much spatial detail as possible (e.g., with 12x24 tokens), and the Fast pathway operates on a high frame rate but uses a larger spatial pooling stride (e.g., downsampling 6x) to focus on the motion cues. As a result, this design allows us to adequately capture both spatial and temporal features that are beneficial for detailed video understanding. Experimental results show that SF-LLaVA outperforms existing training-free methods on a wide range of video tasks. On some benchmarks, it achieves comparable or even better performance compared to state-of-the-art Video LLMs that are fine-tuned on video datasets. Code has been made available at: https://github.com/apple/ml-slowfast-llava."
                    },
                    {
                        "title": "Ferret-v2: An Improved Baseline for Referring and Grounding with Large Language Models",
                        "abstract": "While Ferret seamlessly integrates regional understanding into the Large Language Model (LLM) to facilitate its referring and grounding capability, it poses certain limitations: constrained by the pre-trained fixed visual encoder and failed to perform well on broader tasks. In this work, we unveil Ferret-v2, a significant upgrade to Ferret, with three key designs. (1) Any resolution grounding and referring: A flexible approach that effortlessly handles higher image resolution, improving the model's ability to process and understand images in greater detail. (2) Multi-granularity visual encoding: By integrating the additional DINOv2 encoder, the model learns better and diverse underlying contexts for global and fine-grained visual information. (3) A three-stage training paradigm: Besides image-caption alignment, an additional stage is proposed for high-resolution dense alignment before the final instruction tuning. Experiments show that Ferret-v2 provides substantial improvements over Ferret and other state-of-the-art methods, thanks to its high-resolution scaling and fine-grained visual processing."
                    },
                    {
                        "title": "MM-Ego: Towards Building Egocentric Multimodal LLMs",
                        "abstract": "This research aims to comprehensively explore building a multimodal foundation model for egocentric video understanding. To achieve this goal, we work on three fronts. First, as there is a lack of QA data for egocentric video understanding, we develop a data engine that efficiently generates 7M high-quality QA samples for egocentric videos ranging from 30 seconds to one hour long, based on human-annotated data. This is currently the largest egocentric QA dataset. Second, we contribute a challenging egocentric QA benchmark with 629 videos and 7,026 questions to evaluate the models' ability in recognizing and memorizing visual details across videos of varying lengths. We introduce a new de-biasing evaluation method to help mitigate the unavoidable language bias present in the models being evaluated. Third, we propose a specialized multimodal architecture featuring a novel\"Memory Pointer Prompting\"mechanism. This design includes a global glimpse step to gain an overarching understanding of the entire video and identify key visual information, followed by a fallback step that utilizes the key visual information to generate responses. This enables the model to more effectively comprehend extended video content. With the data, benchmark, and model, we successfully build MM-Ego, an egocentric multimodal LLM that shows powerful performance on egocentric video understanding."
                    },
                    {
                        "title": "Improve Vision Language Model Chain-of-thought Reasoning",
                        "abstract": "Chain-of-thought (CoT) reasoning in vision language models (VLMs) is crucial for improving interpretability and trustworthiness. However, current training recipes lack robust CoT reasoning data, relying on datasets dominated by short annotations with minimal rationales. In this work, we show that training VLM on short answers does not generalize well to reasoning tasks that require more detailed responses. To address this, we propose a two-fold approach. First, we distill rationales from GPT-4o model to enrich the training data and fine-tune VLMs, boosting their CoT performance. Second, we apply reinforcement learning to further calibrate reasoning quality. Specifically, we construct positive (correct) and negative (incorrect) pairs of model-generated reasoning chains, by comparing their predictions with annotated short answers. Using this pairwise data, we apply the Direct Preference Optimization algorithm to refine the model's reasoning abilities. Our experiments demonstrate significant improvements in CoT reasoning on benchmark datasets and better generalization to direct answer prediction as well. This work emphasizes the importance of incorporating detailed rationales in training and leveraging reinforcement learning to strengthen the reasoning capabilities of VLMs."
                    },
                    {
                        "title": "MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMs",
                        "abstract": "We introduce MIA-Bench, a new benchmark designed to evaluate multimodal large language models (MLLMs) on their ability to strictly adhere to complex instructions. Our benchmark comprises a diverse set of 400 image-prompt pairs, each crafted to challenge the models' compliance with layered instructions in generating accurate responses that satisfy specific requested patterns. Evaluation results from a wide array of state-of-the-art MLLMs reveal significant variations in performance, highlighting areas for improvement in instruction fidelity. Additionally, we create extra training data and explore supervised fine-tuning to enhance the models' ability to strictly follow instructions without compromising performance on other tasks. We hope this benchmark not only serves as a tool for measuring MLLM adherence to instructions, but also guides future developments in MLLM training methods."
                    },
                    {
                        "title": "Understanding Alignment in Multimodal LLMs: A Comprehensive Study",
                        "abstract": "Preference alignment has become a crucial component in enhancing the performance of Large Language Models (LLMs), yet its impact in Multimodal Large Language Models (MLLMs) remains comparatively underexplored. Similar to language models, MLLMs for image understanding tasks encounter challenges like hallucination. In MLLMs, hallucination can occur not only by stating incorrect facts but also by producing responses that are inconsistent with the image content. A primary objective of alignment for MLLMs is to encourage these models to align responses more closely with image information. Recently, multiple works have introduced preference datasets for MLLMs and examined different alignment methods, including Direct Preference Optimization (DPO) and Proximal Policy Optimization (PPO). However, due to variations in datasets, base model types, and alignment methods, it remains unclear which specific elements contribute most significantly to the reported improvements in these works. In this paper, we independently analyze each aspect of preference alignment in MLLMs. We start by categorizing the alignment algorithms into two groups, offline (such as DPO), and online (such as online-DPO), and show that combining offline and online methods can improve the performance of the model in certain scenarios. We review a variety of published multimodal preference datasets and discuss how the details of their construction impact model performance. Based on these insights, we introduce a novel way of creating multimodal preference data called Bias-Driven Hallucination Sampling (BDHS) that needs neither additional annotation nor external models, and show that it can achieve competitive performance to previously published alignment work for multimodal models across a range of benchmarks."
                    },
                    {
                        "title": "Compressing LLMs: The Truth is Rarely Pure and Never Simple",
                        "abstract": "Despite their remarkable achievements, modern Large Language Models (LLMs) face exorbitant computational and memory footprints. Recently, several works have shown significant success in training-free and data-free compression (pruning and quantization) of LLMs that achieve 50 - 60% sparsity and reduce the bit width to 3 or 4 bits per weight, with negligible degradation of perplexity over the uncompressed baseline. As recent research efforts are focused on developing increasingly sophisticated compression methods, our work takes a step back and re-evaluates the effectiveness of existing SoTA compression methods, which rely on a fairly simple and widely questioned metric, perplexity (even for dense LLMs). We introduce Knowledge-Intensive Compressed LLM BenchmarK (LLM-KICK), a collection of carefully curated tasks to redefine the evaluation protocol for compressed LLMs, which have significant alignment with their dense counterparts and perplexity fail to capture subtle change in their true capabilities. LLM-KICK unveils many favorable merits and unfortunate plights of current SoTA compression methods: all pruning methods suffer significant performance degradation, sometimes at trivial sparsity ratios (e.g., 25-30%), and fail for N:M sparsity in knowledge-intensive tasks; current quantization methods are more successful than pruning; yet, pruned LLMs even at $\\geq 50$% sparsity are robust in-context retrieval and summarization systems; among others. LLM-KICK is designed to holistically access compressed LLMs' ability for language understanding, reasoning, generation, in-context retrieval, in-context summarization, etc. We hope our study can foster the development of better LLM compression methods. The reproduced codes are available at https://github.com/VITA-Group/llm-kick."
                    }
                ]
            },
            "b062c248-092a-4566-b6fa-e1ae415e1aa0": {
                "pk": "b062c248-092a-4566-b6fa-e1ae415e1aa0",
                "name": "Xianzhi Du",
                "collaborators": [
                    "Yinfei Yang",
                    "Bowen Zhang",
                    "Tom Gunter",
                    "Zirui Wang",
                    "Ruoming Pang",
                    "Zhe Gan",
                    "Haotian Zhang",
                    "Aonan Zhang",
                    "Floris Weers",
                    "Mark Lee"
                ],
                "domain": [
                    "Unsupervised Domain Adaptation",
                    "Multimodal Learning",
                    "Large Language Models",
                    "Vision Transformers"
                ],
                "publications": [
                    {
                        "title": "Empowering Unsupervised Domain Adaptation with Large-scale Pre-trained Vision-Language Models",
                        "abstract": "Unsupervised Domain Adaptation (UDA) aims to leverage the labeled source domain to solve the tasks on the unlabeled target domain. Traditional UDA methods face the challenge of the tradeoff between domain alignment and semantic class discriminability, especially when a large domain gap exists between the source and target domains. The efforts of applying large-scale pre-training to bridge the domain gaps remain limited. In this work, we propose that Vision-Language Models (VLMs) can empower UDA tasks due to their training pattern with language alignment and their large-scale pre-trained datasets. For example, CLIP and GLIP have shown promising zero-shot generalization in classification and detection tasks. However, directly fine-tuning these VLMs into downstream tasks may be computationally expensive and not scalable if we have multiple domains that need to be adapted. Therefore, in this work, we first study an efficient adaption of VLMs to preserve the original knowledge while maximizing its flexibility for learning new knowledge. Then, we design a domain-aware pseudo-labeling scheme tailored to VLMs for domain disentanglement. We show the superiority of the proposed methods in four UDA-classification and two UDA-detection benchmarks, with a significant improvement (+9.9%) on DomainNet."
                    },
                    {
                        "title": "Apple Intelligence Foundation Language Models",
                        "abstract": "We present foundation language models developed to power Apple Intelligence features, including a ~3 billion parameter model designed to run efficiently on devices and a large server-based language model designed for Private Cloud Compute. These models are designed to perform a wide range of tasks efficiently, accurately, and responsibly. This report describes the model architecture, the data used to train the model, the training process, how the models are optimized for inference, and the evaluation results. We highlight our focus on Responsible AI and how the principles are applied throughout the model development."
                    },
                    {
                        "title": "MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning",
                        "abstract": "We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development."
                    },
                    {
                        "title": "Revisiting MoE and Dense Speed-Accuracy Comparisons for LLM Training",
                        "abstract": "Mixture-of-Experts (MoE) enjoys performance gain by increasing model capacity while keeping computation cost constant. When comparing MoE to dense models, prior work typically adopt the following setting: 1) use FLOPs or activated parameters as a measure of model complexity; 2) train all models to the same number of tokens. We argue that this setting favors MoE as FLOPs and activated parameters do not accurately measure the communication overhead in sparse layers, leading to a larger actual training budget for MoE. In this work, we revisit the settings by adopting step time as a more accurate measure of model complexity, and by determining the total compute budget under the Chinchilla compute-optimal settings. To efficiently run MoE on modern accelerators, we adopt a 3D sharding method that keeps the dense-to-MoE step time increase within a healthy range. We evaluate MoE and dense LLMs on a set of nine 0-shot and two 1-shot English tasks, as well as MMLU 5-shot and GSM8K 8-shot across three model scales at 6.4B, 12.6B, and 29.6B. Experimental results show that even under these settings, MoE consistently outperform dense LLMs on the speed-accuracy trade-off curve with meaningful gaps. Our full model implementation and sharding strategy has been released at~\\url{https://github.com/apple/axlearn}"
                    },
                    {
                        "title": "MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training",
                        "abstract": "In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting."
                    },
                    {
                        "title": "Guiding Instruction-based Image Editing via Multimodal Large Language Models",
                        "abstract": "Instruction-based image editing improves the controllability and flexibility of image manipulation via natural commands without elaborate descriptions or regional masks. However, human instructions are sometimes too brief for current methods to capture and follow. Multimodal large language models (MLLMs) show promising capabilities in cross-modal understanding and visual-aware response generation via LMs. We investigate how MLLMs facilitate edit instructions and present MLLM-Guided Image Editing (MGIE). MGIE learns to derive expressive instructions and provides explicit guidance. The editing model jointly captures this visual imagination and performs manipulation through end-to-end training. We evaluate various aspects of Photoshop-style modification, global photo optimization, and local editing. Extensive experimental results demonstrate that expressive instructions are crucial to instruction-based image editing, and our MGIE can lead to a notable improvement in automatic metrics and human evaluation while maintaining competitive inference efficiency."
                    },
                    {
                        "title": "Mobile V-MoEs: Scaling Down Vision Transformers via Sparse Mixture-of-Experts",
                        "abstract": "Sparse Mixture-of-Experts models (MoEs) have recently gained popularity due to their ability to decouple model size from inference efficiency by only activating a small subset of the model parameters for any given input token. As such, sparse MoEs have enabled unprecedented scalability, resulting in tremendous successes across domains such as natural language processing and computer vision. In this work, we instead explore the use of sparse MoEs to scale-down Vision Transformers (ViTs) to make them more attractive for resource-constrained vision applications. To this end, we propose a simplified and mobile-friendly MoE design where entire images rather than individual patches are routed to the experts. We also propose a stable MoE training procedure that uses super-class information to guide the router. We empirically show that our sparse Mobile Vision MoEs (V-MoEs) can achieve a better trade-off between performance and efficiency than the corresponding dense ViTs. For example, for the ViT-Tiny model, our Mobile V-MoE outperforms its dense counterpart by 3.39% on ImageNet-1k. For an even smaller ViT variant with only 54M FLOPs inference cost, our MoE achieves an improvement of 4.66%."
                    },
                    {
                        "title": "Compressing LLMs: The Truth is Rarely Pure and Never Simple",
                        "abstract": "Despite their remarkable achievements, modern Large Language Models (LLMs) face exorbitant computational and memory footprints. Recently, several works have shown significant success in training-free and data-free compression (pruning and quantization) of LLMs that achieve 50 - 60% sparsity and reduce the bit width to 3 or 4 bits per weight, with negligible degradation of perplexity over the uncompressed baseline. As recent research efforts are focused on developing increasingly sophisticated compression methods, our work takes a step back and re-evaluates the effectiveness of existing SoTA compression methods, which rely on a fairly simple and widely questioned metric, perplexity (even for dense LLMs). We introduce Knowledge-Intensive Compressed LLM BenchmarK (LLM-KICK), a collection of carefully curated tasks to redefine the evaluation protocol for compressed LLMs, which have significant alignment with their dense counterparts and perplexity fail to capture subtle change in their true capabilities. LLM-KICK unveils many favorable merits and unfortunate plights of current SoTA compression methods: all pruning methods suffer significant performance degradation, sometimes at trivial sparsity ratios (e.g., 25-30%), and fail for N:M sparsity in knowledge-intensive tasks; current quantization methods are more successful than pruning; yet, pruned LLMs even at $\\geq 50$% sparsity are robust in-context retrieval and summarization systems; among others. LLM-KICK is designed to holistically access compressed LLMs' ability for language understanding, reasoning, generation, in-context retrieval, in-context summarization, etc. We hope our study can foster the development of better LLM compression methods. The reproduced codes are available at https://github.com/VITA-Group/llm-kick."
                    }
                ]
            },
            "6821424b-33f0-4b41-b396-b36ae296fa34": {
                "pk": "6821424b-33f0-4b41-b396-b36ae296fa34",
                "name": "Bowen Zhang",
                "collaborators": [
                    "Yinfei Yang",
                    "Zhe Gan",
                    "Haotian Zhang",
                    "Yanghao Li",
                    "Hong-You Chen",
                    "Ruoming Pang",
                    "Xianzhi Du",
                    "Zhengfeng Lai",
                    "Meng Cao",
                    "Zirui Wang"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Large Language Models",
                    "Vision-Language Models",
                    "Image Processing"
                ],
                "publications": [
                    {
                        "title": "Contrastive Localized Language-Image Pre-Training",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) has been a celebrated method for training vision encoders to generate image/text representations facilitating various applications. Recently, CLIP has been widely adopted as the vision backbone of multimodal large language models (MLLMs) to connect image inputs for language interactions. The success of CLIP as a vision-language foundation model relies on aligning web-crawled noisy text annotations at image levels. Nevertheless, such criteria may become insufficient for downstream tasks in need of fine-grained vision representations, especially when region-level understanding is demanding for MLLMs. In this paper, we improve the localization capability of CLIP with several advances. We propose a pre-training method called Contrastive Localized Language-Image Pre-training (CLOC) by complementing CLIP with region-text contrastive loss and modules. We formulate a new concept, promptable embeddings, of which the encoder produces image embeddings easy to transform into region representations given spatial hints. To support large-scale pre-training, we design a visually-enriched and spatially-localized captioning framework to effectively generate region-text pseudo-labels at scale. By scaling up to billions of annotated images, CLOC enables high-quality regional embeddings for image region recognition and retrieval tasks, and can be a drop-in replacement of CLIP to enhance MLLMs, especially on referring and grounding tasks."
                    },
                    {
                        "title": "Apple Intelligence Foundation Language Models",
                        "abstract": "We present foundation language models developed to power Apple Intelligence features, including a ~3 billion parameter model designed to run efficiently on devices and a large server-based language model designed for Private Cloud Compute. These models are designed to perform a wide range of tasks efficiently, accurately, and responsibly. This report describes the model architecture, the data used to train the model, the training process, how the models are optimized for inference, and the evaluation results. We highlight our focus on Responsible AI and how the principles are applied throughout the model development."
                    },
                    {
                        "title": "EC-DIT: Scaling Diffusion Transformers with Adaptive Expert-Choice Routing",
                        "abstract": "Diffusion transformers have been widely adopted for text-to-image synthesis. While scaling these models up to billions of parameters shows promise, the effectiveness of scaling beyond current sizes remains underexplored and challenging. By explicitly exploiting the computational heterogeneity of image generations, we develop a new family of Mixture-of-Experts (MoE) models (EC-DIT) for diffusion transformers with expert-choice routing. EC-DIT learns to adaptively optimize the compute allocated to understand the input texts and generate the respective image patches, enabling heterogeneous computation aligned with varying text-image complexities. This heterogeneity provides an efficient way of scaling EC-DIT up to 97 billion parameters and achieving significant improvements in training convergence, text-to-image alignment, and overall generation quality over dense models and conventional MoE models. Through extensive ablations, we show that EC-DIT demonstrates superior scalability and adaptive compute allocation by recognizing varying textual importance through end-to-end training. Notably, in text-to-image alignment evaluation, our largest models achieve a state-of-the-art GenEval score of 71.68% and still maintain competitive inference speed with intuitive interpretability."
                    },
                    {
                        "title": "MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning",
                        "abstract": "We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development."
                    },
                    {
                        "title": "MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training",
                        "abstract": "In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting."
                    },
                    {
                        "title": "Revisit Large-Scale Image-Caption Data in Pre-training Multimodal Foundation Models",
                        "abstract": "Recent advancements in multimodal models highlight the value of rewritten captions for improving performance, yet key challenges remain. For example, while synthetic captions often provide superior quality and image-text alignment, it is not clear whether they can fully replace AltTexts: the role of synthetic captions and their interaction with original web-crawled AltTexts in pre-training is still not well understood. Moreover, different multimodal foundation models may have unique preferences for specific caption formats, but efforts to identify the optimal captions for each model remain limited. In this work, we propose a novel, controllable, and scalable captioning pipeline designed to generate diverse caption formats tailored to various multimodal models. By examining Short Synthetic Captions (SSC) towards Dense Synthetic Captions (DSC+) as case studies, we systematically explore their effects and interactions with AltTexts across models such as CLIP, multimodal LLMs, and diffusion models. Our findings reveal that a hybrid approach that keeps both synthetic captions and AltTexts can outperform the use of synthetic captions alone, improving both alignment and performance, with each model demonstrating preferences for particular caption formats. This comprehensive analysis provides valuable insights into optimizing captioning strategies, thereby advancing the pre-training of multimodal foundation models."
                    },
                    {
                        "title": "Ferret-v2: An Improved Baseline for Referring and Grounding with Large Language Models",
                        "abstract": "While Ferret seamlessly integrates regional understanding into the Large Language Model (LLM) to facilitate its referring and grounding capability, it poses certain limitations: constrained by the pre-trained fixed visual encoder and failed to perform well on broader tasks. In this work, we unveil Ferret-v2, a significant upgrade to Ferret, with three key designs. (1) Any resolution grounding and referring: A flexible approach that effortlessly handles higher image resolution, improving the model's ability to process and understand images in greater detail. (2) Multi-granularity visual encoding: By integrating the additional DINOv2 encoder, the model learns better and diverse underlying contexts for global and fine-grained visual information. (3) A three-stage training paradigm: Besides image-caption alignment, an additional stage is proposed for high-resolution dense alignment before the final instruction tuning. Experiments show that Ferret-v2 provides substantial improvements over Ferret and other state-of-the-art methods, thanks to its high-resolution scaling and fine-grained visual processing."
                    },
                    {
                        "title": "MM-Ego: Towards Building Egocentric Multimodal LLMs",
                        "abstract": "This research aims to comprehensively explore building a multimodal foundation model for egocentric video understanding. To achieve this goal, we work on three fronts. First, as there is a lack of QA data for egocentric video understanding, we develop a data engine that efficiently generates 7M high-quality QA samples for egocentric videos ranging from 30 seconds to one hour long, based on human-annotated data. This is currently the largest egocentric QA dataset. Second, we contribute a challenging egocentric QA benchmark with 629 videos and 7,026 questions to evaluate the models' ability in recognizing and memorizing visual details across videos of varying lengths. We introduce a new de-biasing evaluation method to help mitigate the unavoidable language bias present in the models being evaluated. Third, we propose a specialized multimodal architecture featuring a novel\"Memory Pointer Prompting\"mechanism. This design includes a global glimpse step to gain an overarching understanding of the entire video and identify key visual information, followed by a fallback step that utilizes the key visual information to generate responses. This enables the model to more effectively comprehend extended video content. With the data, benchmark, and model, we successfully build MM-Ego, an egocentric multimodal LLM that shows powerful performance on egocentric video understanding."
                    },
                    {
                        "title": "Improve Vision Language Model Chain-of-thought Reasoning",
                        "abstract": "Chain-of-thought (CoT) reasoning in vision language models (VLMs) is crucial for improving interpretability and trustworthiness. However, current training recipes lack robust CoT reasoning data, relying on datasets dominated by short annotations with minimal rationales. In this work, we show that training VLM on short answers does not generalize well to reasoning tasks that require more detailed responses. To address this, we propose a two-fold approach. First, we distill rationales from GPT-4o model to enrich the training data and fine-tune VLMs, boosting their CoT performance. Second, we apply reinforcement learning to further calibrate reasoning quality. Specifically, we construct positive (correct) and negative (incorrect) pairs of model-generated reasoning chains, by comparing their predictions with annotated short answers. Using this pairwise data, we apply the Direct Preference Optimization algorithm to refine the model's reasoning abilities. Our experiments demonstrate significant improvements in CoT reasoning on benchmark datasets and better generalization to direct answer prediction as well. This work emphasizes the importance of incorporating detailed rationales in training and leveraging reinforcement learning to strengthen the reasoning capabilities of VLMs."
                    },
                    {
                        "title": "Compressing LLMs: The Truth is Rarely Pure and Never Simple",
                        "abstract": "Despite their remarkable achievements, modern Large Language Models (LLMs) face exorbitant computational and memory footprints. Recently, several works have shown significant success in training-free and data-free compression (pruning and quantization) of LLMs that achieve 50 - 60% sparsity and reduce the bit width to 3 or 4 bits per weight, with negligible degradation of perplexity over the uncompressed baseline. As recent research efforts are focused on developing increasingly sophisticated compression methods, our work takes a step back and re-evaluates the effectiveness of existing SoTA compression methods, which rely on a fairly simple and widely questioned metric, perplexity (even for dense LLMs). We introduce Knowledge-Intensive Compressed LLM BenchmarK (LLM-KICK), a collection of carefully curated tasks to redefine the evaluation protocol for compressed LLMs, which have significant alignment with their dense counterparts and perplexity fail to capture subtle change in their true capabilities. LLM-KICK unveils many favorable merits and unfortunate plights of current SoTA compression methods: all pruning methods suffer significant performance degradation, sometimes at trivial sparsity ratios (e.g., 25-30%), and fail for N:M sparsity in knowledge-intensive tasks; current quantization methods are more successful than pruning; yet, pruned LLMs even at $\\geq 50$% sparsity are robust in-context retrieval and summarization systems; among others. LLM-KICK is designed to holistically access compressed LLMs' ability for language understanding, reasoning, generation, in-context retrieval, in-context summarization, etc. We hope our study can foster the development of better LLM compression methods. The reproduced codes are available at https://github.com/VITA-Group/llm-kick."
                    }
                ]
            },
            "f6c4dc33-8561-4ce1-bca0-2bf09bf9d781": {
                "pk": "f6c4dc33-8561-4ce1-bca0-2bf09bf9d781",
                "name": "Zirui Wang",
                "collaborators": [
                    "Zhe Gan",
                    "Xianzhi Du",
                    "J. Fauconnier",
                    "Peter Grasch",
                    "Yinfei Yang",
                    "Haotian Zhang",
                    "Philipp Dufter",
                    "Dhruti Shah",
                    "Bowen Zhang",
                    "Sam Dodge"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Large Language Models",
                    "Computer Vision",
                    "Preference Alignment"
                ],
                "publications": [
                    {
                        "title": "MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning",
                        "abstract": "We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development."
                    },
                    {
                        "title": "Revisiting MoE and Dense Speed-Accuracy Comparisons for LLM Training",
                        "abstract": "Mixture-of-Experts (MoE) enjoys performance gain by increasing model capacity while keeping computation cost constant. When comparing MoE to dense models, prior work typically adopt the following setting: 1) use FLOPs or activated parameters as a measure of model complexity; 2) train all models to the same number of tokens. We argue that this setting favors MoE as FLOPs and activated parameters do not accurately measure the communication overhead in sparse layers, leading to a larger actual training budget for MoE. In this work, we revisit the settings by adopting step time as a more accurate measure of model complexity, and by determining the total compute budget under the Chinchilla compute-optimal settings. To efficiently run MoE on modern accelerators, we adopt a 3D sharding method that keeps the dense-to-MoE step time increase within a healthy range. We evaluate MoE and dense LLMs on a set of nine 0-shot and two 1-shot English tasks, as well as MMLU 5-shot and GSM8K 8-shot across three model scales at 6.4B, 12.6B, and 29.6B. Experimental results show that even under these settings, MoE consistently outperform dense LLMs on the speed-accuracy trade-off curve with meaningful gaps. Our full model implementation and sharding strategy has been released at~\\url{https://github.com/apple/axlearn}"
                    },
                    {
                        "title": "MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training",
                        "abstract": "In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting."
                    },
                    {
                        "title": "Understanding Alignment in Multimodal LLMs: A Comprehensive Study",
                        "abstract": "Preference alignment has become a crucial component in enhancing the performance of Large Language Models (LLMs), yet its impact in Multimodal Large Language Models (MLLMs) remains comparatively underexplored. Similar to language models, MLLMs for image understanding tasks encounter challenges like hallucination. In MLLMs, hallucination can occur not only by stating incorrect facts but also by producing responses that are inconsistent with the image content. A primary objective of alignment for MLLMs is to encourage these models to align responses more closely with image information. Recently, multiple works have introduced preference datasets for MLLMs and examined different alignment methods, including Direct Preference Optimization (DPO) and Proximal Policy Optimization (PPO). However, due to variations in datasets, base model types, and alignment methods, it remains unclear which specific elements contribute most significantly to the reported improvements in these works. In this paper, we independently analyze each aspect of preference alignment in MLLMs. We start by categorizing the alignment algorithms into two groups, offline (such as DPO), and online (such as online-DPO), and show that combining offline and online methods can improve the performance of the model in certain scenarios. We review a variety of published multimodal preference datasets and discuss how the details of their construction impact model performance. Based on these insights, we introduce a novel way of creating multimodal preference data called Bias-Driven Hallucination Sampling (BDHS) that needs neither additional annotation nor external models, and show that it can achieve competitive performance to previously published alignment work for multimodal models across a range of benchmarks."
                    }
                ]
            },
            "9aae8c21-2020-4f95-91f6-e0ba8d6ca5ca": {
                "pk": "9aae8c21-2020-4f95-91f6-e0ba8d6ca5ca",
                "name": "Liangliang Cao",
                "collaborators": [
                    "Tom Gunter",
                    "Zirui Wang",
                    "Chong Wang",
                    "Ruoming Pang",
                    "Andy Narayanan",
                    "Aonan Zhang",
                    "Bowen Zhang",
                    "Chen Chen",
                    "Chung-Cheng Chiu",
                    "David Qiu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Foundation Model",
                    "Responsible AI"
                ],
                "publications": [
                    {
                        "title": "Apple Intelligence Foundation Language Models",
                        "abstract": "We present foundation language models developed to power Apple Intelligence features, including a ~3 billion parameter model designed to run efficiently on devices and a large server-based language model designed for Private Cloud Compute. These models are designed to perform a wide range of tasks efficiently, accurately, and responsibly. This report describes the model architecture, the data used to train the model, the training process, how the models are optimized for inference, and the evaluation results. We highlight our focus on Responsible AI and how the principles are applied throughout the model development."
                    }
                ]
            },
            "aa279139-7004-4e12-bc1b-d74afe2fb50d": {
                "pk": "aa279139-7004-4e12-bc1b-d74afe2fb50d",
                "name": "Shih-Fu Chang",
                "collaborators": [
                    "Haoxuan You",
                    "Zhecan Wang",
                    "Kai-Wei Chang",
                    "Junzhang Liu",
                    "Rui Sun",
                    "Wenhao Li",
                    "Hammad A. Ayyubi",
                    "Chris Thomas",
                    "Chia-Wei Tang",
                    "Hani Alomari"
                ],
                "domain": [
                    "Vision-Language Understanding",
                    "Multimodal Reasoning",
                    "Machine Learning",
                    "Bias Mitigation"
                ],
                "publications": [
                    {
                        "title": "JourneyBench: A Challenging One-Stop Vision-Language Understanding Benchmark of Generated Images",
                        "abstract": "Existing vision-language understanding benchmarks largely consist of images of objects in their usual contexts. As a consequence, recent multimodal large language models can perform well with only a shallow visual understanding by relying on background language biases. Thus, strong performance on these benchmarks does not necessarily correlate with strong visual understanding. In this paper, we release JourneyBench, a comprehensive human-annotated benchmark of generated images designed to assess the model's fine-grained multimodal reasoning abilities across five tasks: complementary multimodal chain of thought, multi-image VQA, imaginary image captioning, VQA with hallucination triggers, and fine-grained retrieval with sample-specific distractors. Unlike existing benchmarks, JourneyBench explicitly requires fine-grained multimodal reasoning in unusual imaginary scenarios where language bias and holistic image gist are insufficient. We benchmark state-of-the-art models on JourneyBench and analyze performance along a number of fine-grained dimensions. Results across all five tasks show that JourneyBench is exceptionally challenging for even the best models, indicating that models' visual reasoning abilities are not as strong as they first appear. We discuss the implications of our findings and propose avenues for further research."
                    },
                    {
                        "title": "Detecting Multimodal Situations with Insufficient Context and Abstaining from Baseless Predictions",
                        "abstract": "Despite the widespread adoption of Vision-Language Understanding (VLU) benchmarks such as VQA v2, OKVQA, A-OKVQA, GQA, VCR, SWAG, and VisualCOMET, our analysis reveals a pervasive issue affecting their integrity: these benchmarks contain samples where answers rely on assumptions unsupported by the provided context. Training models on such data foster biased learning and hallucinations as models tend to make similar unwarranted assumptions. To address this issue, we collect contextual data for each sample whenever available and train a context selection module to facilitate evidence-based model predictions. Strong improvements across multiple benchmarks demonstrate the effectiveness of our approach. Further, we develop a general-purpose Context-AwaRe Abstention (CARA) detector to identify samples lacking sufficient context and enhance model accuracy by abstaining from responding if the required context is absent. CARA exhibits generalization to new benchmarks it wasn't trained on, underscoring its utility for future VLU benchmarks in detecting or cleaning samples with inadequate context. Finally, we curate a Context Ambiguity and Sufficiency Evaluation (CASE) set to benchmark the performance of insufficient context detectors. Overall, our work represents a significant advancement in ensuring that vision-language models generate trustworthy and evidence-based outputs in complex real-world scenarios."
                    },
                    {
                        "title": "Ferret-v2: An Improved Baseline for Referring and Grounding with Large Language Models",
                        "abstract": "While Ferret seamlessly integrates regional understanding into the Large Language Model (LLM) to facilitate its referring and grounding capability, it poses certain limitations: constrained by the pre-trained fixed visual encoder and failed to perform well on broader tasks. In this work, we unveil Ferret-v2, a significant upgrade to Ferret, with three key designs. (1) Any resolution grounding and referring: A flexible approach that effortlessly handles higher image resolution, improving the model's ability to process and understand images in greater detail. (2) Multi-granularity visual encoding: By integrating the additional DINOv2 encoder, the model learns better and diverse underlying contexts for global and fine-grained visual information. (3) A three-stage training paradigm: Besides image-caption alignment, an additional stage is proposed for high-resolution dense alignment before the final instruction tuning. Experiments show that Ferret-v2 provides substantial improvements over Ferret and other state-of-the-art methods, thanks to its high-resolution scaling and fine-grained visual processing."
                    },
                    {
                        "title": "Dataset Bias Mitigation in Multiple-Choice Visual Question Answering and Beyond",
                        "abstract": "Vision-language (VL) understanding tasks evaluate models' comprehension of complex visual scenes through multiple-choice questions. However, we have identified two dataset biases that models can exploit as shortcuts to resolve various VL tasks correctly without proper understanding. The first type of dataset bias is \\emph{Unbalanced Matching} bias, where the correct answer overlaps the question and image more than the incorrect answers. The second type of dataset bias is \\emph{Distractor Similarity} bias, where incorrect answers are overly dissimilar to the correct answer but significantly similar to other incorrect answers within the same sample. To address these dataset biases, we first propose Adversarial Data Synthesis (ADS) to generate synthetic training and debiased evaluation data. We then introduce Intra-sample Counterfactual Training (ICT) to assist models in utilizing the synthesized training data, particularly the counterfactual data, via focusing on intra-sample differentiation. Extensive experiments demonstrate the effectiveness of ADS and ICT in consistently improving model performance across different benchmarks, even in domain-shifted scenarios."
                    }
                ]
            },
            "0577d18c-3cb6-42dc-8f43-c987a0708846": {
                "pk": "0577d18c-3cb6-42dc-8f43-c987a0708846",
                "name": "Yinfei Yang",
                "collaborators": [
                    "Zhe Gan",
                    "Haotian Zhang",
                    "Bowen Zhang",
                    "Xianzhi Du",
                    "Peter Grasch",
                    "Zhengfeng Lai",
                    "Hong-You Chen",
                    "J. Fauconnier",
                    "Meng Cao",
                    "Keen You"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Vision-Language Models",
                    "Domain Adaptation",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "Empowering Unsupervised Domain Adaptation with Large-scale Pre-trained Vision-Language Models",
                        "abstract": "Unsupervised Domain Adaptation (UDA) aims to leverage the labeled source domain to solve the tasks on the unlabeled target domain. Traditional UDA methods face the challenge of the tradeoff between domain alignment and semantic class discriminability, especially when a large domain gap exists between the source and target domains. The efforts of applying large-scale pre-training to bridge the domain gaps remain limited. In this work, we propose that Vision-Language Models (VLMs) can empower UDA tasks due to their training pattern with language alignment and their large-scale pre-trained datasets. For example, CLIP and GLIP have shown promising zero-shot generalization in classification and detection tasks. However, directly fine-tuning these VLMs into downstream tasks may be computationally expensive and not scalable if we have multiple domains that need to be adapted. Therefore, in this work, we first study an efficient adaption of VLMs to preserve the original knowledge while maximizing its flexibility for learning new knowledge. Then, we design a domain-aware pseudo-labeling scheme tailored to VLMs for domain disentanglement. We show the superiority of the proposed methods in four UDA-classification and two UDA-detection benchmarks, with a significant improvement (+9.9%) on DomainNet."
                    },
                    {
                        "title": "Contrastive Localized Language-Image Pre-Training",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) has been a celebrated method for training vision encoders to generate image/text representations facilitating various applications. Recently, CLIP has been widely adopted as the vision backbone of multimodal large language models (MLLMs) to connect image inputs for language interactions. The success of CLIP as a vision-language foundation model relies on aligning web-crawled noisy text annotations at image levels. Nevertheless, such criteria may become insufficient for downstream tasks in need of fine-grained vision representations, especially when region-level understanding is demanding for MLLMs. In this paper, we improve the localization capability of CLIP with several advances. We propose a pre-training method called Contrastive Localized Language-Image Pre-training (CLOC) by complementing CLIP with region-text contrastive loss and modules. We formulate a new concept, promptable embeddings, of which the encoder produces image embeddings easy to transform into region representations given spatial hints. To support large-scale pre-training, we design a visually-enriched and spatially-localized captioning framework to effectively generate region-text pseudo-labels at scale. By scaling up to billions of annotated images, CLOC enables high-quality regional embeddings for image region recognition and retrieval tasks, and can be a drop-in replacement of CLIP to enhance MLLMs, especially on referring and grounding tasks."
                    },
                    {
                        "title": "MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning",
                        "abstract": "We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development."
                    },
                    {
                        "title": "Ferret-UI: Grounded Mobile UI Understanding with Multimodal LLMs",
                        "abstract": "Recent advancements in multimodal large language models (MLLMs) have been noteworthy, yet, these general-domain MLLMs often fall short in their ability to comprehend and interact effectively with user interface (UI) screens. In this paper, we present Ferret-UI, a new MLLM tailored for enhanced understanding of mobile UI screens, equipped with referring, grounding, and reasoning capabilities. Given that UI screens typically exhibit a more elongated aspect ratio and contain smaller objects of interest (e.g., icons, texts) than natural images, we incorporate\"any resolution\"on top of Ferret to magnify details and leverage enhanced visual features. Specifically, each screen is divided into 2 sub-images based on the original aspect ratio (i.e., horizontal division for portrait screens and vertical division for landscape screens). Both sub-images are encoded separately before being sent to LLMs. We meticulously gather training samples from an extensive range of elementary UI tasks, such as icon recognition, find text, and widget listing. These samples are formatted for instruction-following with region annotations to facilitate precise referring and grounding. To augment the model's reasoning ability, we further compile a dataset for advanced tasks, including detailed description, perception/interaction conversations, and function inference. After training on the curated datasets, Ferret-UI exhibits outstanding comprehension of UI screens and the capability to execute open-ended instructions. For model evaluation, we establish a comprehensive benchmark encompassing all the aforementioned tasks. Ferret-UI excels not only beyond most open-source UI MLLMs, but also surpasses GPT-4V on all the elementary UI tasks."
                    },
                    {
                        "title": "MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training",
                        "abstract": "In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting."
                    },
                    {
                        "title": "How Easy is It to Fool Your Multimodal LLMs? An Empirical Analysis on Deceptive Prompts",
                        "abstract": "The remarkable advancements in Multimodal Large Language Models (MLLMs) have not rendered them immune to challenges, particularly in the context of handling deceptive information in prompts, thus producing hallucinated responses under such conditions. To quantitatively assess this vulnerability, we present MAD-Bench, a carefully curated benchmark that contains 1000 test samples divided into 5 categories, such as non-existent objects, count of objects, and spatial relationship. We provide a comprehensive analysis of popular MLLMs, ranging from GPT-4v, Reka, Gemini-Pro, to open-sourced models, such as LLaVA-NeXT and MiniCPM-Llama3. Empirically, we observe significant performance gaps between GPT-4o and other models; and previous robust instruction-tuned models are not effective on this new benchmark. While GPT-4o achieves 82.82% accuracy on MAD-Bench, the accuracy of any other model in our experiments ranges from 9% to 50%. We further propose a remedy that adds an additional paragraph to the deceptive prompts to encourage models to think twice before answering the question. Surprisingly, this simple method can even double the accuracy; however, the absolute numbers are still too low to be satisfactory. We hope MAD-Bench can serve as a valuable benchmark to stimulate further research to enhance model resilience against deceptive prompts."
                    },
                    {
                        "title": "Revisit Large-Scale Image-Caption Data in Pre-training Multimodal Foundation Models",
                        "abstract": "Recent advancements in multimodal models highlight the value of rewritten captions for improving performance, yet key challenges remain. For example, while synthetic captions often provide superior quality and image-text alignment, it is not clear whether they can fully replace AltTexts: the role of synthetic captions and their interaction with original web-crawled AltTexts in pre-training is still not well understood. Moreover, different multimodal foundation models may have unique preferences for specific caption formats, but efforts to identify the optimal captions for each model remain limited. In this work, we propose a novel, controllable, and scalable captioning pipeline designed to generate diverse caption formats tailored to various multimodal models. By examining Short Synthetic Captions (SSC) towards Dense Synthetic Captions (DSC+) as case studies, we systematically explore their effects and interactions with AltTexts across models such as CLIP, multimodal LLMs, and diffusion models. Our findings reveal that a hybrid approach that keeps both synthetic captions and AltTexts can outperform the use of synthetic captions alone, improving both alignment and performance, with each model demonstrating preferences for particular caption formats. This comprehensive analysis provides valuable insights into optimizing captioning strategies, thereby advancing the pre-training of multimodal foundation models."
                    },
                    {
                        "title": "Ferret-v2: An Improved Baseline for Referring and Grounding with Large Language Models",
                        "abstract": "While Ferret seamlessly integrates regional understanding into the Large Language Model (LLM) to facilitate its referring and grounding capability, it poses certain limitations: constrained by the pre-trained fixed visual encoder and failed to perform well on broader tasks. In this work, we unveil Ferret-v2, a significant upgrade to Ferret, with three key designs. (1) Any resolution grounding and referring: A flexible approach that effortlessly handles higher image resolution, improving the model's ability to process and understand images in greater detail. (2) Multi-granularity visual encoding: By integrating the additional DINOv2 encoder, the model learns better and diverse underlying contexts for global and fine-grained visual information. (3) A three-stage training paradigm: Besides image-caption alignment, an additional stage is proposed for high-resolution dense alignment before the final instruction tuning. Experiments show that Ferret-v2 provides substantial improvements over Ferret and other state-of-the-art methods, thanks to its high-resolution scaling and fine-grained visual processing."
                    },
                    {
                        "title": "MM-Ego: Towards Building Egocentric Multimodal LLMs",
                        "abstract": "This research aims to comprehensively explore building a multimodal foundation model for egocentric video understanding. To achieve this goal, we work on three fronts. First, as there is a lack of QA data for egocentric video understanding, we develop a data engine that efficiently generates 7M high-quality QA samples for egocentric videos ranging from 30 seconds to one hour long, based on human-annotated data. This is currently the largest egocentric QA dataset. Second, we contribute a challenging egocentric QA benchmark with 629 videos and 7,026 questions to evaluate the models' ability in recognizing and memorizing visual details across videos of varying lengths. We introduce a new de-biasing evaluation method to help mitigate the unavoidable language bias present in the models being evaluated. Third, we propose a specialized multimodal architecture featuring a novel\"Memory Pointer Prompting\"mechanism. This design includes a global glimpse step to gain an overarching understanding of the entire video and identify key visual information, followed by a fallback step that utilizes the key visual information to generate responses. This enables the model to more effectively comprehend extended video content. With the data, benchmark, and model, we successfully build MM-Ego, an egocentric multimodal LLM that shows powerful performance on egocentric video understanding."
                    },
                    {
                        "title": "Improve Vision Language Model Chain-of-thought Reasoning",
                        "abstract": "Chain-of-thought (CoT) reasoning in vision language models (VLMs) is crucial for improving interpretability and trustworthiness. However, current training recipes lack robust CoT reasoning data, relying on datasets dominated by short annotations with minimal rationales. In this work, we show that training VLM on short answers does not generalize well to reasoning tasks that require more detailed responses. To address this, we propose a two-fold approach. First, we distill rationales from GPT-4o model to enrich the training data and fine-tune VLMs, boosting their CoT performance. Second, we apply reinforcement learning to further calibrate reasoning quality. Specifically, we construct positive (correct) and negative (incorrect) pairs of model-generated reasoning chains, by comparing their predictions with annotated short answers. Using this pairwise data, we apply the Direct Preference Optimization algorithm to refine the model's reasoning abilities. Our experiments demonstrate significant improvements in CoT reasoning on benchmark datasets and better generalization to direct answer prediction as well. This work emphasizes the importance of incorporating detailed rationales in training and leveraging reinforcement learning to strengthen the reasoning capabilities of VLMs."
                    },
                    {
                        "title": "MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMs",
                        "abstract": "We introduce MIA-Bench, a new benchmark designed to evaluate multimodal large language models (MLLMs) on their ability to strictly adhere to complex instructions. Our benchmark comprises a diverse set of 400 image-prompt pairs, each crafted to challenge the models' compliance with layered instructions in generating accurate responses that satisfy specific requested patterns. Evaluation results from a wide array of state-of-the-art MLLMs reveal significant variations in performance, highlighting areas for improvement in instruction fidelity. Additionally, we create extra training data and explore supervised fine-tuning to enhance the models' ability to strictly follow instructions without compromising performance on other tasks. We hope this benchmark not only serves as a tool for measuring MLLM adherence to instructions, but also guides future developments in MLLM training methods."
                    },
                    {
                        "title": "Understanding Alignment in Multimodal LLMs: A Comprehensive Study",
                        "abstract": "Preference alignment has become a crucial component in enhancing the performance of Large Language Models (LLMs), yet its impact in Multimodal Large Language Models (MLLMs) remains comparatively underexplored. Similar to language models, MLLMs for image understanding tasks encounter challenges like hallucination. In MLLMs, hallucination can occur not only by stating incorrect facts but also by producing responses that are inconsistent with the image content. A primary objective of alignment for MLLMs is to encourage these models to align responses more closely with image information. Recently, multiple works have introduced preference datasets for MLLMs and examined different alignment methods, including Direct Preference Optimization (DPO) and Proximal Policy Optimization (PPO). However, due to variations in datasets, base model types, and alignment methods, it remains unclear which specific elements contribute most significantly to the reported improvements in these works. In this paper, we independently analyze each aspect of preference alignment in MLLMs. We start by categorizing the alignment algorithms into two groups, offline (such as DPO), and online (such as online-DPO), and show that combining offline and online methods can improve the performance of the model in certain scenarios. We review a variety of published multimodal preference datasets and discuss how the details of their construction impact model performance. Based on these insights, we introduce a novel way of creating multimodal preference data called Bias-Driven Hallucination Sampling (BDHS) that needs neither additional annotation nor external models, and show that it can achieve competitive performance to previously published alignment work for multimodal models across a range of benchmarks."
                    },
                    {
                        "title": "Guiding Instruction-based Image Editing via Multimodal Large Language Models",
                        "abstract": "Instruction-based image editing improves the controllability and flexibility of image manipulation via natural commands without elaborate descriptions or regional masks. However, human instructions are sometimes too brief for current methods to capture and follow. Multimodal large language models (MLLMs) show promising capabilities in cross-modal understanding and visual-aware response generation via LMs. We investigate how MLLMs facilitate edit instructions and present MLLM-Guided Image Editing (MGIE). MGIE learns to derive expressive instructions and provides explicit guidance. The editing model jointly captures this visual imagination and performs manipulation through end-to-end training. We evaluate various aspects of Photoshop-style modification, global photo optimization, and local editing. Extensive experimental results demonstrate that expressive instructions are crucial to instruction-based image editing, and our MGIE can lead to a notable improvement in automatic metrics and human evaluation while maintaining competitive inference efficiency."
                    },
                    {
                        "title": "Compressing LLMs: The Truth is Rarely Pure and Never Simple",
                        "abstract": "Despite their remarkable achievements, modern Large Language Models (LLMs) face exorbitant computational and memory footprints. Recently, several works have shown significant success in training-free and data-free compression (pruning and quantization) of LLMs that achieve 50 - 60% sparsity and reduce the bit width to 3 or 4 bits per weight, with negligible degradation of perplexity over the uncompressed baseline. As recent research efforts are focused on developing increasingly sophisticated compression methods, our work takes a step back and re-evaluates the effectiveness of existing SoTA compression methods, which rely on a fairly simple and widely questioned metric, perplexity (even for dense LLMs). We introduce Knowledge-Intensive Compressed LLM BenchmarK (LLM-KICK), a collection of carefully curated tasks to redefine the evaluation protocol for compressed LLMs, which have significant alignment with their dense counterparts and perplexity fail to capture subtle change in their true capabilities. LLM-KICK unveils many favorable merits and unfortunate plights of current SoTA compression methods: all pruning methods suffer significant performance degradation, sometimes at trivial sparsity ratios (e.g., 25-30%), and fail for N:M sparsity in knowledge-intensive tasks; current quantization methods are more successful than pruning; yet, pruned LLMs even at $\\geq 50$% sparsity are robust in-context retrieval and summarization systems; among others. LLM-KICK is designed to holistically access compressed LLMs' ability for language understanding, reasoning, generation, in-context retrieval, in-context summarization, etc. We hope our study can foster the development of better LLM compression methods. The reproduced codes are available at https://github.com/VITA-Group/llm-kick."
                    }
                ]
            }
        }
    },
    "2404.04125": {
        "paper_data": {
            "title": "No \"Zero-Shot\" Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model Performance",
            "url": "http://arxiv.org/abs/2404.04125v2",
            "arxiv_id": "2404.04125",
            "authors": [
                "Vishaal Udandarao",
                "Ameya Prabhu",
                "Adhiraj Ghosh",
                "Yash Sharma",
                "Philip H. S. Torr",
                "Adel Bibi",
                "Samuel Albanie",
                "Matthias Bethge"
            ],
            "abstract": "Web-crawled pretraining datasets underlie the impressive \"zero-shot\" evaluation performance of multimodal models, such as CLIP for classification/retrieval and Stable-Diffusion for image generation. However, it is unclear how meaningful the notion of \"zero-shot\" generalization is for such multimodal models, as it is not known to what extent their pretraining datasets encompass the downstream concepts targeted for during \"zero-shot\" evaluation. In this work, we ask: How is the performance of multimodal models on downstream concepts influenced by the frequency of these concepts in their pretraining datasets? We comprehensively investigate this question across 34 models and five standard pretraining datasets (CC-3M, CC-12M, YFCC-15M, LAION-400M, LAION-Aesthetics), generating over 300GB of data artifacts. We consistently find that, far from exhibiting \"zero-shot\" generalization, multimodal models require exponentially more data to achieve linear improvements in downstream \"zero-shot\" performance, following a sample inefficient log-linear scaling trend. This trend persists even when controlling for sample-level similarity between pretraining and downstream datasets, and testing on purely synthetic data distributions. Furthermore, upon benchmarking models on long-tailed data sampled based on our analysis, we demonstrate that multimodal models across the board perform poorly. We contribute this long-tail test set as the \"Let it Wag!\" benchmark to further research in this direction. Taken together, our study reveals an exponential need for training data which implies that the key to \"zero-shot\" generalization capabilities under large-scale training paradigms remains to be found.",
            "introduction": "   1 Introduction  Multimodal models like CLIP\u00a0[91] and Stable Diffusion\u00a0[96] have revolutionized performance on downstream tasks\u2014CLIP is now the de-facto standard for \u201czero-shot\" image recognition\u00a0[133, 72, 126, 48, 132] and image-text retrieval\u00a0[46, 64, 24, 117, 129], while Stable Diffusion is now the de-facto standard for \u201czero-shot\" text-to-image (T2I) generation [93, 17, 96, 41]. In this work, we investigate this empirical success through the lens of zero-shot generalization\u00a0[69], which refers to the ability of the model to apply its learned knowledge to new unseen concepts. Accordingly, we ask: Are current multimodal models truly capable of \u201czero-shot\u201d generalization?   To address this, we conducted a comparative analysis involving two main factors: (1) the performance of models across various downstream tasks and (2) the frequency of test concepts within their pretraining datasets. We compiled a comprehensive list of 4,02940294,0294 , 029 concepts111class categories for classification tasks, objects in the text captions for retrieval tasks, and objects in the text prompts for generation tasks, see\u00a0Sec.\u00a02 for more details on how we define concepts. from 27 downstream tasks spanning classification, retrieval, and image generation, assessing the performance against these concepts. Our analysis spanned five large-scale pretraining datasets with different scales, data curation methods and sources (CC-3M\u00a0[107], CC-12M\u00a0[27], YFCC-15M\u00a0[113], LAION-Aesthetics\u00a0[103], LAION-400M\u00a0[102]), and evaluated the performance of 10 CLIP models and 24 T2I models, spanning different architectures and parameter scales. We consistently find across all our experiments that, across concepts, the frequency of a concept in the pretraining dataset is a strong predictor of the model\u2019s performance on test examples containing that concept. Notably, model performance scales linearly as the concept frequency in pretraining data grows exponentially i.e.,\u00a0we observe a consistent log-linear scaling trend. We find that this log-linear trend is robust to controlling for correlated factors (similar samples in pretraining and test data\u00a0[79]) and testing across different concept distributions along with samples generated entirely synthetically\u00a0[51].   Our findings indicate that the impressive empirical performance of multimodal models like CLIP and Stable Diffusion can be largely attributed to the presence of test concepts within their vast pretraining datasets, thus their reported empirical performance does not constitute \u201czero-shot\u201d generalization. Quite the contrary, these models require exponentially more data on a concept to linearly improve their performance on tasks pertaining to that concept, highlighting extreme sample inefficiency.   In our analysis, we additionally document the distribution of concepts encountered in pretraining data and find that:   \u2022  Concept Distribution: Across all pretraining datasets, the distribution of concepts is long-tailed (see\u00a0Fig.\u00a05 in\u00a0Sec.\u00a05), which indicates that a large fraction of concepts are rare. However, given the extreme sample inefficiency observed, what is rare is not properly learned during multimodal pretraining.    \u2022  Concept Correlation across Pretraining Datasets: The distribution of concepts across different pretraining datasets are strongly correlated (see\u00a0Tab.\u00a04 in \u00a0Sec.\u00a05), which suggests web crawls yield surprisingly similar concept distributions across different pretraining data curation strategies, necessitating explicit rebalancing efforts [11, 125].    \u2022  Image-Text Misalignment between Concepts in Pretraining Data: Concepts often appear in one modality but not the other, which implies significant misalignment (see\u00a0Tab.\u00a04 in\u00a0Sec.\u00a05). Our released data artifacts can help image-text alignment efforts at scale by precisely indicating the examples in which modalities misalign. Note that the log-linear",
            "references": [
                {
                    "title": "Scaling Rectified Flow Transformers for High-Resolution Image Synthesis",
                    "abstract": "Diffusion models create data from noise by inverting the forward paths of data towards noise and have emerged as a powerful generative modeling technique for high-dimensional, perceptual data such as images and videos. Rectified flow is a recent generative model formulation that connects data and noise in a straight line. Despite its better theoretical properties and conceptual simplicity, it is not yet decisively established as standard practice. In this work, we improve existing noise sampling techniques for training rectified flow models by biasing them towards perceptually relevant scales. Through a large-scale study, we demonstrate the superior performance of this approach compared to established diffusion formulations for high-resolution text-to-image synthesis. Additionally, we present a novel transformer-based architecture for text-to-image generation that uses separate weights for the two modalities and enables a bidirectional flow of information between image and text tokens, improving text comprehension, typography, and human preference ratings. We demonstrate that this architecture follows predictable scaling trends and correlates lower validation loss to improved text-to-image synthesis as measured by various metrics and human evaluations. Our largest models outperform state-of-the-art models, and we will make our experimental data, code, and model weights publicly available."
                },
                {
                    "title": "SynthCLIP: Are We Ready for a Fully Synthetic CLIP Training?",
                    "abstract": "We present SynthCLIP, a CLIP model trained on entirely synthetic text-image pairs. Leveraging recent text-to-image (TTI) networks and large language models (LLM), we generate synthetic datasets of images and corresponding captions at scale, with no human intervention. In this work, we provide an analysis on CLIP models trained on synthetic data. We provide insights on the data generation strategy, number of samples required, scaling trends, and resulting properties. We also introduce SynthCI-30M, a purely synthetic dataset comprising 30 million captioned images. Our code, trained models, and data, are released as open source at https://github.com/hammoudhasan/SynthCLIP"
                },
                {
                    "title": "The Neglected Tails in Vision-Language Models",
                    "abstract": "Vision-language models (VLMs) excel in zero-shot recognition but their performance varies greatly across different visual concepts. For example, although CLIP achieves impressive accuracy on ImageNet (60-80%), its performance drops below 10% for more than ten concepts like night snake, presumably due to their limited presence in the pretraining data. However, measuring the frequency of concepts in VLMs' large-scale datasets is challenging. We address this by using large language models (LLMs) to count the number of pretraining texts that con-tain synonyms of these concepts. Our analysis confirms that popular datasets, such as LAION, exhibit a long-tailed concept distribution, yielding biased performance in VLMs. We also find that downstream applications of VLMs, including visual chatbots (e.g., GPT-4V) and text-to-image models (e.g., Stable Diffusion), often fail to recognize or generate images of rare concepts identified by our method. To mit-igate the imbalanced performance of zero-shot VLMs, we propose REtrieval-Augmented Learning (REAL). First, in-stead of prompting VLMs using the original class names, REAL uses their most frequent synonyms found in pretraining texts. This simple change already outperforms costly human-engineered and LLM-enriched prompts over nine benchmark datasets. Second, REAL trains a linear classifier on a small yet balanced set of pretraining data re-trieved using concept synonyms. REAL surpasses the previous zero-shot SOTA, using 400\u00d7 less storage and 10,000\u00d7 less training time!"
                },
                {
                    "title": "Effective pruning of web-scale datasets based on complexity of concept clusters",
                    "abstract": "Utilizing massive web-scale datasets has led to unprecedented performance gains in machine learning models, but also imposes outlandish compute requirements for their training. In order to improve training and data efficiency, we here push the limits of pruning large-scale multimodal datasets for training CLIP-style models. Today's most effective pruning method on ImageNet clusters data samples into separate concepts according to their embedding and prunes away the most prototypical samples. We scale this approach to LAION and improve it by noting that the pruning rate should be concept-specific and adapted to the complexity of the concept. Using a simple and intuitive complexity measure, we are able to reduce the training cost to a quarter of regular training. By filtering from the LAION dataset, we find that training on a smaller set of high-quality data can lead to higher performance with significantly lower training costs. More specifically, we are able to outperform the LAION-trained OpenCLIP-ViT-B32 model on ImageNet zero-shot accuracy by 1.1p.p. while only using 27.7% of the data and training compute. Despite a strong reduction in training cost, we also see improvements on ImageNet dist. shifts, retrieval tasks and VTAB. On the DataComp Medium benchmark, we achieve a new state-of-the-art Imagehttps://info.arxiv.org/help/prep#commentsNet zero-shot accuracy and a competitive average zero-shot accuracy on 38 evaluation tasks."
                },
                {
                    "title": "Low-Resource Vision Challenges for Foundation Models",
                    "abstract": "Low-resource settings are well-established in natural language processing, where many languages lack sufficient data for deep learning at scale. However, low-resource problems are under-explored in computer vision. In this paper, we address this gap and explore the challenges of low-resource image tasks with vision foundation models. We first collect a benchmark of genuinely low-resource image data, covering historic maps, circuit diagrams, and mechanical drawings. These low-resource settings all share three challenges: data scarcity, fine-grained differences, and the distribution shift from natural images to the specialized domain of interest. While existing foundation models have shown impressive generalizability, we find they cannot transfer well to our low-resource tasks. To begin to tackle the challenges of low-resource vision, we introduce one simple baseline per challenge. Specifically, we i) enlarge the data space by generative models, ii) adopt the best sub-kernels to encode local regions for fine-grained difference discovery and iii) learn attention for specialized domains. Experiments on our three low-resource tasks demonstrate our proposals already provide a better baseline than transfer learning, data augmentation, and fine-grained methods. This highlights the unique characteristics and challenges of low-resource vision for foundation models that warrant further investigation. Project page: https://xiaobai1217.github.io/Low-Resource-Vision/."
                },
                {
                    "title": "Explaining CLIP's Performance Disparities on Data from Blind/Low Vision Users",
                    "abstract": "Large multi-modal models (LMMs) hold the potential to usher in a new era of automated visual assistance for people who are blind or low vision (BLV). Yet, these models have not been systematically evaluated on data captured by BLV users. We address this by empirically assessing CLIP, a widely-used LMM likely to underpin many assistive technologies. Testing 25 CLIP variants in a zero-shot classification task, we find that their accuracy is 15 percentage points lower on average for images captured by BLV users than web-crawled images. This disparity stems from CLIP's sensitivities to 1) image content (e.g. not recognizing disability objects as well as other objects); 2) image quality (e.g. not being robust to lighting variation); and 3) text content (e.g. not recognizing objects described by tactile adjectives as well as visual ones). We delve deeper with a textual analysis of three common pre-training datasets: LAION-400M, LAION-2B and DataComp-1B, showing that disability con-tent is rarely mentioned. We then provide three examples that illustrate how the performance disparities extend to three downstream models underpinned by CLIP: OWL-ViT, CLIPSeg and DALL-E2. We find that few-shot learning with as few as 5 images can mitigate CLIP's quality-of-service disparities for BLV users in some scenarios, which we discuss alongside a set of other possible mitigations."
                },
                {
                    "title": "MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Training",
                    "abstract": "Contrastive pretraining of image-text foundation models, such as CLIP, demonstrated excellent zero-shot performance and improved robustness on a wide range of downstream tasks. However, these models utilize large transformer-based encoders with significant memory and latency overhead which pose challenges for deployment on mobile devices. In this work, we introduce MobileCLIP - a new family of efficient image-text models optimized for runtime performance along with a novel and efficient training approach, namely multi-modal reinforced training. The proposed training approach leverages knowledge transfer from an image captioning model and an ensemble of strong CLIP encoders to improve the accuracy of efficient moels. Our approach avoids train-time compute overhead by storing the additional knowledge in a reinforced dataset. MobileCLIP sets a new state-of-the-art latency-accuracy tradeoff for zero-shot classification and retrieval tasks on several datasets. Our MobileCLIP-S2 variant is 2.3\u00d7 faster while more accurate compared to previous best CLIP model based on ViT-B/16. We further demonstrate the effectiveness of our multi-modal reinforced training by training a CLIP model based on ViT-B/16 image backbone and achieving +2.9% average performance improvement on 38 evaluation benchmarks compared to the previous best. Moreover, we show that the proposed approach achieves 10\u00d7-1000 \u00d7 improved learning efficiency when compared with non-reinforced CLIP training. Code and models are available at https://github.com/apple/ml-mobileclip"
                },
                {
                    "title": "ShareGPT4V: Improving Large Multi-Modal Models with Better Captions",
                    "abstract": "In the realm of large multi-modal models (LMMs), efficient modality alignment is crucial yet often constrained by the scarcity of high-quality image-text data. To address this bottleneck, we introduce the ShareGPT4V dataset, a pioneering large-scale resource featuring 1.2 million highly descriptive captions, which surpasses existing datasets in diversity and information content, covering world knowledge, object properties, spatial relationships, and aesthetic evaluations. Specifically, ShareGPT4V originates from a curated 100K high-quality captions collected from advanced GPT4-Vision and has been expanded to 1.2M with a superb caption model trained on this subset. ShareGPT4V first demonstrates its effectiveness for the Supervised Fine-Tuning (SFT) phase, by substituting an equivalent quantity of detailed captions in existing SFT datasets with a subset of our high-quality captions, significantly enhancing the LMMs like LLaVA-7B, LLaVA-1.5-13B, and Qwen-VL-Chat-7B on the MME and MMBench benchmarks, with respective gains of 222.8/22.0/22.3 and 2.7/1.3/1.5. We further incorporate ShareGPT4V data into both the pre-training and SFT phases, obtaining ShareGPT4V-7B, a superior LMM based on a simple architecture that has remarkable performance across a majority of the multi-modal benchmarks. This project is available at https://ShareGPT4V.github.io to serve as a pivotal resource for advancing the LMMs community."
                },
                {
                    "title": "From Categories to Classifiers: Name-Only Continual Learning by Exploring the Web",
                    "abstract": "Continual Learning (CL) often relies on the availability of extensive annotated datasets, an assumption that is unrealistically time-consuming and costly in practice. We explore a novel paradigm termed name-only continual learning where time and cost constraints prohibit manual annotation. In this scenario, learners adapt to new category shifts using only category names without the luxury of annotated training data. Our proposed solution leverages the expansive and ever-evolving internet to query and download uncurated webly-supervised data for image classification. We investigate the reliability of our web data and find them comparable, and in some cases superior, to manually annotated datasets. Additionally, we show that by harnessing the web, we can create support sets that surpass state-of-the-art name-only classification that create support sets using generative models or image retrieval from LAION-5B, achieving up to 25% boost in accuracy. When applied across varied continual learning contexts, our method consistently exhibits a small performance gap in comparison to models trained on manually annotated datasets. We present EvoTrends, a class-incremental dataset made from the web to capture real-world trends, created in just minutes. Overall, this paper underscores the potential of using uncurated webly-supervised data to mitigate the challenges associated with manual data labeling in continual learning."
                },
                {
                    "title": "Data Similarity is Not Enough to Explain Language Model Performance",
                    "abstract": "Large language models achieve high performance on many but not all downstream tasks. The interaction between pretraining data and task data is commonly assumed to determine this variance: a task with data that is more similar to a model's pretraining data is assumed to be easier for that model. We test whether distributional and example-specific similarity measures (embedding-, token- and model-based) correlate with language model performance through a large-scale comparison of the Pile and C4 pretraining datasets with downstream benchmarks. Similarity correlates with performance for multilingual datasets, but in other benchmarks, we surprisingly find that similarity metrics are not correlated with accuracy or even each other. This suggests that the relationship between pretraining data and downstream tasks is more complex than often assumed."
                },
                {
                    "title": "Attribute Diversity Determines the Systematicity Gap in VQA",
                    "abstract": "Although modern neural networks often generalize to new combinations of familiar concepts, the conditions that enable such compositionality have long been an open question. In this work, we study the systematicity gap in visual question answering: the performance difference between reasoning on previously seen and unseen combinations of object attributes. To test, we introduce a novel diagnostic dataset, CLEVR-HOPE. We find that the systematicity gap is not reduced by increasing the quantity of training data, but is reduced by increasing the diversity of training data. In particular, our experiments suggest that the more distinct attribute type combinations are seen during training, the more systematic we can expect the resulting model to be."
                },
                {
                    "title": "Holistic Evaluation of Text-To-Image Models",
                    "abstract": "The stunning qualitative improvement of recent text-to-image models has led to their widespread attention and adoption. However, we lack a comprehensive quantitative understanding of their capabilities and risks. To fill this gap, we introduce a new benchmark, Holistic Evaluation of Text-to-Image Models (HEIM). Whereas previous evaluations focus mostly on text-image alignment and image quality, we identify 12 aspects, including text-image alignment, image quality, aesthetics, originality, reasoning, knowledge, bias, toxicity, fairness, robustness, multilinguality, and efficiency. We curate 62 scenarios encompassing these aspects and evaluate 26 state-of-the-art text-to-image models on this benchmark. Our results reveal that no single model excels in all aspects, with different models demonstrating different strengths. We release the generated images and human evaluation results for full transparency at https://crfm.stanford.edu/heim/v1.1.0 and the code at https://github.com/stanford-crfm/helm, which is integrated with the HELM codebase."
                },
                {
                    "title": "Into the LAIONs Den: Investigating Hate in Multimodal Datasets",
                    "abstract": "'Scale the model, scale the data, scale the compute' is the reigning sentiment in the world of generative AI today. While the impact of model scaling has been extensively studied, we are only beginning to scratch the surface of data scaling and its consequences. This is especially of critical importance in the context of vision-language datasets such as LAION. These datasets are continually growing in size and are built based on large-scale internet dumps such as the Common Crawl, which is known to have numerous drawbacks ranging from quality, legality, and content. The datasets then serve as the backbone for large generative models, contributing to the operationalization and perpetuation of harmful societal and historical biases and stereotypes. In this paper, we investigate the effect of scaling datasets on hateful content through a comparative audit of two datasets: LAION-400M and LAION-2B. Our results show that hate content increased by nearly 12% with dataset scale, measured both qualitatively and quantitatively using a metric that we term as Hate Content Rate (HCR). We also found that filtering dataset contents based on Not Safe For Work (NSFW) values calculated based on images alone does not exclude all the harmful content in alt-text. Instead, we found that trace amounts of hateful, targeted, and aggressive text remain even when carrying out conservative filtering. We end with a reflection and a discussion of the significance of our results for dataset curation and usage in the AI community. Code and the meta-data assets curated in this paper are publicly available at https://github.com/vinayprabhu/hate_scaling. Content warning: This paper contains examples of hateful text that might be disturbing, distressing, and/or offensive."
                },
                {
                    "title": "CapsFusion: Rethinking Image-Text Data at Scale",
                    "abstract": "Large multimodal models demonstrate remarkable generalist ability to perform diverse multimodal tasks in a zero-shot manner. Large-scale web-based image-text pairs contribute fundamentally to this success, but suffer from excessive noise. Recent studies use alternative captions synthesized by captioning models and have achieved notable benchmark performance. However, our experiments re-veal significant Scalability Deficiency and World Knowledge Loss issues in models trained with synthetic captions, which have been largely obscured by their initial benchmark success. Upon closer examination, we identify the root cause as the overly-simplified language structure and lack of knowledge details in existing synthetic captions. To provide higher-quality and more scalable multimodal pretraining data, we propose CAPSFUSION, an advanced framework that leverages large language models to consoli-date and refine information from both web-based image-text pairs and synthetic captions. Extensive experiments show that CAPSFUSION captions exhibit remarkable all-round superiority over existing captions in terms of model performance (e.g., 18.8 and 18.3 improvements in CIDEr score on COCO and No Caps), sample efficiency (requiring 11\u201316 times less computation than baselines), world knowledge depth, and scalability. These effectiveness, efficiency and scalability advantages position CAPSFUSION as a promising candidate for future scaling of LMM training."
                },
                {
                    "title": "What's In My Big Data?",
                    "abstract": "Large text corpora are the backbone of language models. However, we have a limited understanding of the content of these corpora, including general statistics, quality, social factors, and inclusion of evaluation data (contamination). In this work, we propose What's In My Big Data? (WIMBD), a platform and a set of sixteen analyses that allow us to reveal and compare the contents of large text corpora. WIMBD builds on two basic capabilities -- count and search -- at scale, which allows us to analyze more than 35 terabytes on a standard compute node. We apply WIMBD to ten different corpora used to train popular language models, including C4, The Pile, and RedPajama. Our analysis uncovers several surprising and previously undocumented findings about these corpora, including the high prevalence of duplicate, synthetic, and low-quality content, personally identifiable information, toxic language, and benchmark contamination. For instance, we find that about 50% of the documents in RedPajama and LAION-2B-en are duplicates. In addition, several datasets used for benchmarking models trained on such corpora are contaminated with respect to important benchmarks, including the Winograd Schema Challenge and parts of GLUE and SuperGLUE. We open-source WIMBD's code and artifacts to provide a standard set of evaluations for new text-based corpora and to encourage more analyses and transparency around them."
                },
                {
                    "title": "Open-Set Image Tagging with Multi-Grained Text Supervision",
                    "abstract": "In this paper, we introduce the Recognize Anything Plus Model (RAM++), an open-set image tagging model effectively leveraging multi-grained text supervision. Previous approaches (e.g., CLIP) primarily utilize global text supervision paired with images, leading to sub-optimal performance in recognizing multiple individual semantic tags. In contrast, RAM++ seamlessly integrates individual tag supervision with global text supervision, all within a unified alignment framework. This integration not only ensures efficient recognition of predefined tag categories, but also enhances generalization capabilities for diverse open-set categories. Furthermore, RAM++ employs large language models (LLMs) to convert semantically constrained tag supervision into more expansive tag description supervision, thereby enriching the scope of open-set visual description concepts. Comprehensive evaluations on various image recognition benchmarks demonstrate RAM++ exceeds existing state-of-the-art (SOTA) open-set image tagging models on most aspects. Specifically, for predefined commonly used tag categories, RAM++ showcases 10.2 mAP and 15.4 mAP enhancements over CLIP on OpenImages and ImageNet. For open-set categories beyond predefined, RAM++ records improvements of 5.0 mAP and 6.4 mAP over CLIP and RAM respectively on OpenImages. For diverse human-object interaction phrases, RAM++ achieves 7.8 mAP and 4.7 mAP improvements on the HICO benchmark. Code, datasets and pre-trained models are available at \\url{https://github.com/xinyu1205/recognize-anything}."
                },
                {
                    "title": "ViPE: Visualise Pretty-much Everything",
                    "abstract": "Figurative and non-literal expressions are profoundly integrated in human communication. Visualising such expressions allow us to convey our creative thoughts, and evoke nuanced emotions. Recent text-to-image models like Stable Diffusion, on the other hand, struggle to depict non-literal expressions. Recent works primarily deal with this issue by compiling humanly annotated datasets on a small scale, which not only demands specialised expertise but also proves highly inefficient. To address this issue, we introduce ViPE: Visualise Pretty-much Everything. ViPE offers a series of lightweight and robust language models that have been trained on a large-scale set of lyrics with noisy visual descriptions that represent their implicit meaning. The synthetic visual descriptions are generated by GPT3.5 relying on neither human annotations nor images. ViPE effectively expresses any arbitrary piece of text into a visualisable description, enabling meaningful and high-quality image generation. We provide compelling evidence that ViPE is more robust than GPT3.5 in synthesising visual elaborations. ViPE also exhibits an understanding of figurative expressions comparable to human experts, providing a powerful and open-source backbone to many downstream applications such as music video and caption generation."
                },
                {
                    "title": "Does CLIP's Generalization Performance Mainly Stem from High Train-Test Similarity?",
                    "abstract": "Foundation models like CLIP are trained on hundreds of millions of samples and effortlessly generalize to new tasks and inputs. Out of the box, CLIP shows stellar zero-shot and few-shot capabilities on a wide range of out-of-distribution (OOD) benchmarks, which prior works attribute mainly to today's large and comprehensive training dataset (like LAION). However, it is questionable how meaningful terms like out-of-distribution generalization are for CLIP as it seems likely that web-scale datasets like LAION simply contain many samples that are similar to common OOD benchmarks originally designed for ImageNet. To test this hypothesis, we retrain CLIP on pruned LAION splits that replicate ImageNet's train-test similarity with respect to common OOD benchmarks. While we observe a performance drop on some benchmarks, surprisingly, CLIP's overall performance remains high. This shows that high train-test similarity is insufficient to explain CLIP's OOD performance, and other properties of the training data must drive CLIP to learn more generalizable representations. Additionally, by pruning data points that are dissimilar to the OOD benchmarks, we uncover a 100M split of LAION ($\\frac{1}{4}$th of its original size) on which CLIP can be trained to match its original OOD performance."
                },
                {
                    "title": "Visual Data-Type Understanding does not emerge from Scaling Vision-Language Models",
                    "abstract": "Recent advances in the development of vision-language models (VLMs) are yielding remarkable success in recognizing visual semantic content, including impressive instances of compositional image understanding. Here, we introduce the novel task of Visual Data-Type Identification, a basic perceptual skill with implications for data curation (e.g., noisy data-removal from large datasets, domain-specific retrieval) and autonomous vision (e.g., distinguishing changing weather conditions from camera lens staining). We develop two datasets consisting of animal images altered across a diverse set of 27 visual data-types, spanning four broad categories. An extensive zero-shot evaluation of 39 VLMs, ranging from 100M to 80B parameters, shows a nuanced performance landscape. While VLMs are reasonably good at identifying certain stylistic \\textit{data-types}, such as cartoons and sketches, they struggle with simpler data-types arising from basic manipulations like image rotations or additive noise. Our findings reveal that (i) model scaling alone yields marginal gains for contrastively-trained models like CLIP, and (ii) there is a pronounced drop in performance for the largest auto-regressively trained VLMs like OpenFlamingo. This finding points to a blind spot in current frontier VLMs: they excel in recognizing semantic content but fail to acquire an understanding of visual data-types through scaling. By analyzing the pre-training distributions of these models and incorporating data-type information into the captions during fine-tuning, we achieve a significant enhancement in performance. By exploring this previously uncharted task, we aim to set the stage for further advancing VLMs to equip them with visual data-type understanding. Code and datasets are released at https://github.com/bethgelab/DataTypeIdentification."
                },
                {
                    "title": "Sieve: Multimodal Dataset Pruning Using Image Captioning Models",
                    "abstract": "Vision-Language Models (VLMs) are pretrained on large, diverse, and noisy web-crawled datasets. This underscores the critical need for dataset pruning, as the quality of these datasets is strongly correlated with the performance of VLMs on downstream tasks. Using CLIPScore from a pretrained model to only train models using highly-aligned samples is one of the most successful methods for pruning. We argue that this approach suffers from multiple limitations including: false positives and negatives due to CLIP's pretraining on noisy labels. We propose a pruning signal, Sieve, that employs synthetic captions generated by image-captioning models pretrained on small, diverse, and well-aligned image-text pairs to evaluate the alignment of noisy image-text pairs. To bridge the gap between the limited diversity of generated captions and the high diversity of alternative text (alt-text), we estimate the semantic textual similarity in the embedding space of a language model pretrained on unlabeled text corpus. Using DataComp, a multimodal dataset filtering benchmark, when evaluating on 38 downstream tasks, our pruning approach, surpasses CLIPScore by 2.6% and 1.7% on medium and large scale respectively. In addition, on retrieval tasks, Sieve leads to a significant improvement of 2.7% and 4.5% on medium and large scale respectively."
                },
                {
                    "title": "D4: Improving LLM Pretraining via Document De-Duplication and Diversification",
                    "abstract": "Over recent years, an increasing amount of compute and data has been poured into training large language models (LLMs), usually by doing one-pass learning on as many tokens as possible randomly selected from large-scale web corpora. While training on ever-larger portions of the internet leads to consistent performance improvements, the size of these improvements diminishes with scale, and there has been little work exploring the effect of data selection on pre-training and downstream performance beyond simple de-duplication methods such as MinHash. Here, we show that careful data selection (on top of de-duplicated data) via pre-trained model embeddings can speed up training (20% efficiency gains) and improves average downstream accuracy on 16 NLP tasks (up to 2%) at the 6.7B model scale. Furthermore, we show that repeating data intelligently consistently outperforms baseline training (while repeating random data performs worse than baseline training). Our results indicate that clever data selection can significantly improve LLM pre-training, calls into question the common practice of training for a single epoch on as much data as possible, and demonstrates a path to keep improving our models past the limits of randomly sampling web data."
                },
                {
                    "title": "The Bias Amplification Paradox in Text-to-Image Generation",
                    "abstract": "Bias amplification is a phenomenon in which models exacerbate biases or stereotypes present in the training data. In this paper, we study bias amplification in the text-to-image domain using Stable Diffusion by comparing gender ratios in training vs. generated images. We find that the model appears to amplify gender-occupation biases found in the training data (LAION) considerably. However, we discover that amplification can be largely attributed to discrepancies between training captions and model prompts. For example, an inherent difference is that captions from the training data often contain explicit gender information while our prompts do not, which leads to a distribution shift and consequently inflates bias measures. Once we account for distributional differences between texts used for training and generation when evaluating amplification, we observe that amplification decreases drastically. Our findings illustrate the challenges of comparing biases in models and their training data, as well as evaluation more broadly, and highlight how confounding factors can impact analyses."
                },
                {
                    "title": "On the Connection between Pre-training Data Diversity and Fine-tuning Robustness",
                    "abstract": "Pre-training has been widely adopted in deep learning to improve model performance, especially when the training data for a target task is limited. In our work, we seek to understand the implications of this training strategy on the generalization properties of downstream models. More specifically, we ask the following question: how do properties of the pre-training distribution affect the robustness of a fine-tuned model? The properties we explore include the label space, label semantics, image diversity, data domains, and data quantity of the pre-training distribution. We find that the primary factor influencing downstream effective robustness (Taori et al., 2020) is data quantity, while other factors have limited significance. For example, reducing the number of ImageNet pre-training classes by 4x while increasing the number of images per class by 4x (that is, keeping total data quantity fixed) does not impact the robustness of fine-tuned models. We demonstrate our findings on pre-training distributions drawn from various natural and synthetic data sources, primarily using the iWildCam-WILDS distribution shift as a test for downstream robustness."
                },
                {
                    "title": "Improving Multimodal Datasets with Image Captioning",
                    "abstract": "Massive web datasets play a key role in the success of large vision-language models like CLIP and Flamingo. However, the raw web data is noisy, and existing filtering methods to reduce noise often come at the expense of data diversity. Our work focuses on caption quality as one major source of noise, and studies how generated captions can increase the utility of web-scraped datapoints with nondescript text. Through exploring different mixing strategies for raw and generated captions, we outperform the best filtering method proposed by the DataComp benchmark by 2% on ImageNet and 4% on average across 38 tasks, given a candidate pool of 128M image-text pairs. Our best approach is also 2x better at Flickr and MS-COCO retrieval. We then analyze what makes synthetic captions an effective source of text supervision. In experimenting with different image captioning models, we also demonstrate that the performance of a model on standard image captioning benchmarks (e.g., NoCaps CIDEr) is not a reliable indicator of the utility of the captions it generates for multimodal training. Finally, our experiments with using generated captions at DataComp's large scale (1.28B image-text pairs) offer insights into the limitations of synthetic text, as well as the importance of image curation with increasing training data quantity. The synthetic captions used in our experiments are now available on HuggingFace."
                },
                {
                    "title": "T-MARS: Improving Visual Representations by Circumventing Text Feature Learning",
                    "abstract": "Large web-sourced multimodal datasets have powered a slew of new methods for learning general-purpose visual representations, advancing the state of the art in computer vision and revolutionizing zero- and few-shot recognition. One crucial decision facing practitioners is how, if at all, to curate these ever-larger datasets. For example, the creators of the LAION-5B dataset chose to retain only image-caption pairs whose CLIP similarity score exceeded a designated threshold. In this paper, we propose a new state-of-the-art data filtering approach motivated by our observation that nearly 40% of LAION's images contain text that overlaps significantly with the caption. Intuitively, such data could be wasteful as it incentivizes models to perform optical character recognition rather than learning visual features. However, naively removing all such data could also be wasteful, as it throws away images that contain visual features (in addition to overlapping text). Our simple and scalable approach, T-MARS (Text Masking and Re-Scoring), filters out only those pairs where the text dominates the remaining visual features -- by first masking out the text and then filtering out those with a low CLIP similarity score of the masked image. Experimentally, T-MARS outperforms the top-ranked method on the\"medium scale\"of DataComp (a data filtering benchmark) by a margin of 6.5% on ImageNet and 4.7% on VTAB. Additionally, our systematic evaluation on various data pool sizes from 2M to 64M shows that the accuracy gains enjoyed by T-MARS linearly increase as data and compute are scaled exponentially. Code is available at https://github.com/locuslab/T-MARS."
                },
                {
                    "title": "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis",
                    "abstract": "We present SDXL, a latent diffusion model for text-to-image synthesis. Compared to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet backbone: The increase of model parameters is mainly due to more attention blocks and a larger cross-attention context as SDXL uses a second text encoder. We design multiple novel conditioning schemes and train SDXL on multiple aspect ratios. We also introduce a refinement model which is used to improve the visual fidelity of samples generated by SDXL using a post-hoc image-to-image technique. We demonstrate that SDXL shows drastically improved performance compared the previous versions of Stable Diffusion and achieves results competitive with those of black-box state-of-the-art image generators. In the spirit of promoting open research and fostering transparency in large model training and evaluation, we provide access to code and model weights at https://github.com/Stability-AI/generative-models"
                },
                {
                    "title": "Scaling Open-Vocabulary Object Detection",
                    "abstract": "Open-vocabulary object detection has benefited greatly from pretrained vision-language models, but is still limited by the amount of available detection training data. While detection training data can be expanded by using Web image-text pairs as weak supervision, this has not been done at scales comparable to image-level pretraining. Here, we scale up detection data with self-training, which uses an existing detector to generate pseudo-box annotations on image-text pairs. Major challenges in scaling self-training are the choice of label space, pseudo-annotation filtering, and training efficiency. We present the OWLv2 model and OWL-ST self-training recipe, which address these challenges. OWLv2 surpasses the performance of previous state-of-the-art open-vocabulary detectors already at comparable training scales (~10M examples). However, with OWL-ST, we can scale to over 1B examples, yielding further large improvement: With an L/14 architecture, OWL-ST improves AP on LVIS rare classes, for which the model has seen no human box annotations, from 31.2% to 44.6% (43% relative improvement). OWL-ST unlocks Web-scale training for open-world localization, similar to what has been seen for image classification and language modelling."
                },
                {
                    "title": "StableRep: Synthetic Images from Text-to-Image Models Make Strong Visual Representation Learners",
                    "abstract": "We investigate the potential of learning visual representations using synthetic images generated by text-to-image models. This is a natural question in the light of the excellent performance of such models in generating high-quality images. We consider specifically the Stable Diffusion, one of the leading open source text-to-image models. We show that (1) when the generative model is configured with proper classifier-free guidance scale, training self-supervised methods on synthetic images can match or beat the real image counterpart; (2) by treating the multiple images generated from the same text prompt as positives for each other, we develop a multi-positive contrastive learning method, which we call StableRep. With solely synthetic images, the representations learned by StableRep surpass the performance of representations learned by SimCLR and CLIP using the same set of text prompts and corresponding real images, on large scale datasets. When we further add language supervision, StableRep trained with 20M synthetic images achieves better accuracy than CLIP trained with 50M real images."
                },
                {
                    "title": "MultiFusion: Fusing Pre-Trained Models for Multi-Lingual, Multi-Modal Image Generation",
                    "abstract": "The recent popularity of text-to-image diffusion models (DM) can largely be attributed to the intuitive interface they provide to users. The intended generation can be expressed in natural language, with the model producing faithful interpretations of text prompts. However, expressing complex or nuanced ideas in text alone can be difficult. To ease image generation, we propose MultiFusion that allows one to express complex and nuanced concepts with arbitrarily interleaved inputs of multiple modalities and languages. MutliFusion leverages pre-trained models and aligns them for integration into a cohesive system, thereby avoiding the need for extensive training from scratch. Our experimental results demonstrate the efficient transfer of capabilities from individual modules to the downstream model. Specifically, the fusion of all independent components allows the image generation module to utilize multilingual, interleaved multimodal inputs despite being trained solely on monomodal data in a single language."
                },
                {
                    "title": "Quantifying Association Capabilities of Large Language Models and Its Implications on Privacy Leakage",
                    "abstract": "The advancement of large language models (LLMs) brings notable improvements across various applications, while simultaneously raising concerns about potential private data exposure. One notable capability of LLMs is their ability to form associations between different pieces of information, but this raises concerns when it comes to personally identifiable information (PII). This paper delves into the association capabilities of language models, aiming to uncover the factors that influence their proficiency in associating information. Our study reveals that as models scale up, their capacity to associate entities/information intensifies, particularly when target pairs demonstrate shorter co-occurrence distances or higher co-occurrence frequencies. However, there is a distinct performance gap when associating commonsense knowledge versus PII, with the latter showing lower accuracy. Despite the proportion of accurately predicted PII being relatively small, LLMs still demonstrate the capability to predict specific instances of email addresses and phone numbers when provided with appropriate prompts. These findings underscore the potential risk to PII confidentiality posed by the evolving capabilities of LLMs, especially as they continue to expand in scale and power."
                },
                {
                    "title": "A Pretrainer\u2019s Guide to Training Data: Measuring the Effects of Data Age, Domain Coverage, Quality, & Toxicity",
                    "abstract": "Pretraining data design is critically under-documented and often guided by empirically unsupported intuitions. We pretrain models on data curated (1) at different collection times, (2) with varying toxicity and quality filters, and (3) with different domain compositions. First, we find that temporal shift between evaluation data and pretraining data leads to performance degradation, which is not overcome by finetuning. Second, we measure the effect of quality and toxicity filters, showing a trade-off between performance on standard benchmarks and risk of toxic generations. We also find that the effects of different types of filtering are not predictable from text domain characteristics. Third, we empirically validate that heterogeneous data sources, like books and web, are beneficial and warrant greater prioritization. To date, these experiments constitute the single largest publicly documented empirical study of the effects of pretraining data. Spanning 28 unique 1.5 billion parameter models pretrained from scratch, these findings validate, quantify, and expose many undocumented intuitions about text pretraining, which ultimately support more informed data-centric decisions in model development."
                },
                {
                    "title": "Speak, Memory: An Archaeology of Books Known to ChatGPT/GPT-4",
                    "abstract": "In this work, we carry out a data archaeology to infer books that are known to ChatGPT and GPT-4 using a name cloze membership inference query. We find that OpenAI models have memorized a wide collection of copyrighted materials, and that the degree of memorization is tied to the frequency with which passages of those books appear on the web. The ability of these models to memorize an unknown set of books complicates assessments of measurement validity for cultural analytics by contaminating test data; we show that models perform much better on memorized books than on non-memorized books for downstream tasks. We argue that this supports a case for open models whose training data is known."
                },
                {
                    "title": "Generating Images of Rare Concepts Using Pre-trained Diffusion Models",
                    "abstract": "Text-to-image diffusion models can synthesize high quality images, but they have various limitations. Here we highlight a common failure mode of these models, namely, generating uncommon concepts and structured concepts like hand palms. We show that their limitation is partly due to the long-tail nature of their training data: web-crawled data sets are strongly unbalanced, causing models to under-represent concepts from the tail of the distribution. We characterize the effect of unbalanced training data on text-to-image models and offer a remedy. We show that rare concepts can be correctly generated by carefully selecting suitable generation seeds in the noise space, using a small reference set of images, a technique that we call SeedSelect. SeedSelect does not require retraining or finetuning the diffusion model. We assess the faithfulness, quality and diversity of SeedSelect in creating rare objects and generating complex formations like hand images, and find it consistently achieves superior performance. We further show the advantage of SeedSelect in semantic data augmentation. Generating semantically appropriate images can successfully improve performance in few-shot recognition benchmarks, for classes from the head and from the tail of the training data of diffusion models."
                },
                {
                    "title": "DataComp: In search of the next generation of multimodal datasets",
                    "abstract": "Multimodal datasets are a critical component in recent breakthroughs such as Stable Diffusion and GPT-4, yet their design does not receive the same research attention as model architectures or training algorithms. To address this shortcoming in the ML ecosystem, we introduce DataComp, a testbed for dataset experiments centered around a new candidate pool of 12.8 billion image-text pairs from Common Crawl. Participants in our benchmark design new filtering techniques or curate new data sources and then evaluate their new dataset by running our standardized CLIP training code and testing the resulting model on 38 downstream test sets. Our benchmark consists of multiple compute scales spanning four orders of magnitude, which enables the study of scaling trends and makes the benchmark accessible to researchers with varying resources. Our baseline experiments show that the DataComp workflow leads to better training sets. In particular, our best baseline, DataComp-1B, enables training a CLIP ViT-L/14 from scratch to 79.2% zero-shot accuracy on ImageNet, outperforming OpenAI's CLIP ViT-L/14 by 3.7 percentage points while using the same training procedure and compute. We release DataComp and all accompanying code at www.datacomp.ai."
                },
                {
                    "title": "Image retrieval outperforms diffusion models on data augmentation",
                    "abstract": "Many approaches have been proposed to use diffusion models to augment training datasets for downstream tasks, such as classification. However, diffusion models are themselves trained on large datasets, often with noisy annotations, and it remains an open question to which extent these models contribute to downstream classification performance. In particular, it remains unclear if they generalize enough to improve over directly using the additional data of their pre-training process for augmentation. We systematically evaluate a range of existing methods to generate images from diffusion models and study new extensions to assess their benefit for data augmentation. Personalizing diffusion models towards the target data outperforms simpler prompting strategies. However, using the pre-training data of the diffusion model alone, via a simple nearest-neighbor retrieval procedure, leads to even stronger downstream performance. Our study explores the potential of diffusion models in generating new training data, and surprisingly finds that these sophisticated models are not yet able to beat a simple and strong image retrieval baseline on simple downstream vision tasks."
                },
                {
                    "title": "DINOv2: Learning Robust Visual Features without Supervision",
                    "abstract": "The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels."
                },
                {
                    "title": "Uncurated Image-Text Datasets: Shedding Light on Demographic Bias",
                    "abstract": "The increasing tendency to collect large and uncurated datasets to train vision-and-language models has raised concerns about fair representations. It is known that even small but manually annotated datasets, such as MSCOCO, are affected by societal bias. This problem, far from being solved, may be getting worse with data crawled from the Internet without much control. In addition, the lack of tools to analyze societal bias in big collections of images makes addressing the problem extremely challenging. Our first contribution is to annotate part of the Google Conceptual Captions dataset, widely used for training vision-and-language models, with four demographic and two contextual attributes. Our second contribution is to conduct a comprehensive analysis of the annotations, focusing on how different demographic groups are represented. Our last contribution lies in evaluating three prevailing vision-and-language tasks: image captioning, text-image CLIP embeddings, and text-to-image generation, showing that societal bias is a persistent problem in all of them. https://github.com/noagarcia/phase"
                },
                {
                    "title": "Sigmoid Loss for Language Image Pre-Training",
                    "abstract": "We propose a simple pairwise sigmoid loss for imagetext pre-training. Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization. The sigmoid loss simultaneously allows further scaling up the batch size, while also performing better at smaller batch sizes. With only four TPUv4 chips, we can train a Base CLIP model at 4k batch size and a Large LiT model at 20k batch size, the latter achieves 84.5% ImageNet zero-shot accuracy in two days. This disentanglement of the batch size from the loss further allows us to study the impact of examples vs pairs and negative to positive ratio. Finally, we push the batch size to the extreme, up to one million, and find that the benefits of growing batch size quickly diminish, with a more reasonable batch size of 32k being sufficient. We hope our research motivates further explorations in improving the quality and efficiency of language-image pre-training."
                },
                {
                    "title": "SemDeDup: Data-efficient learning at web-scale through semantic deduplication",
                    "abstract": "Progress in machine learning has been driven in large part by massive increases in data. However, large web-scale datasets such as LAION are largely uncurated beyond searches for exact duplicates, potentially leaving much redundancy. Here, we introduce SemDeDup, a method which leverages embeddings from pre-trained models to identify and remove semantic duplicates: data pairs which are semantically similar, but not exactly identical. Removing semantic duplicates preserves performance and speeds up learning. Analyzing a subset of LAION, we show that SemDeDup can remove 50% of the data with minimal performance loss, effectively halving training time. Moreover, performance increases out of distribution. Also, analyzing language models trained on C4, a partially curated dataset, we show that SemDeDup improves over prior approaches while providing efficiency gains. SemDeDup provides an example of how simple ways of leveraging quality embeddings can be used to make models learn faster with less data."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Scaling up GANs for Text-to-Image Synthesis",
                    "abstract": "The recent success of text-to-image synthesis has taken the world by storm and captured the general public's imagination. From a technical standpoint, it also marked a drastic change in the favored architecture to design generative image models. GANs used to be the de facto choice, with techniques like StyleGAN. With DALL.E 2, autoregressive and diffusion models became the new standard for large-scale generative models overnight. This rapid shift raises a fundamental question: can we scale up GANs to benefit from large datasets like LAION? We find that na\u00efvely increasing the capacity of the StyleGan architecture quickly becomes unstable. We introduce GigaGAN, a new GAN architecture that far exceeds this limit, demonstrating GANs as a viable option for text-to-image synthesis. GigaGAN offers three major advantages. First, it is orders of magnitude faster at inference time, taking only 0.13 seconds to synthesize a 512px image. Second, it can synthesize high-resolution images, for example, 16-megapixel images in 3.66 seconds. Finally, GigaGAN supports various latent space editing applications such as latent interpolation, style mixing, and vector arithmetic operations."
                },
                {
                    "title": "Prompt, Generate, Then Cache: Cascade of Foundation Models Makes Strong Few-Shot Learners",
                    "abstract": "Visual recognition in low-data regimes requires deep neural networks to learn generalized representations from limited training samples. Recently, CLIP-based methods have shown promising few-shot performance benefited from the contrastive language-image pre-training. We then question, if the more diverse pre-training knowledge can be cascaded to further assist few-shot representation learning. In this paper, we propose CaFo, a Cascade of Foundation models that incorporates diverse prior knowledge of various pretraining paradigms for better few-shot learning. Our CaFo incorporates CLIP's language-contrastive knowledge, DINO's vision-contrastive knowledge, DALL-E's vision-generative knowledge, and GPT-3's language-generative knowledge. Specifically, CaFo works by \u2018Prompt, Generate, then Cache\u2019. Firstly, we leverage GPT-3 to produce textual inputs for prompting CLIP with rich downstream linguistic semantics. Then, we generate synthetic images via DALL-E to expand the few-shot training data without any manpower. At last, we introduce a learnable cache model to adaptively blend the predictions from CLIP and DINO. By such collaboration, CaFo can fully unleash the potential of different pre-training methods and unify them to perform state-of-the-art for few-shot classification. Code is available at https://github.com/ZrrSkywalker/CaFo."
                },
                {
                    "title": "The Role of Pre-training Data in Transfer Learning",
                    "abstract": "The transfer learning paradigm of model pre-training and subsequent fine-tuning produces high-accuracy models. While most studies recommend scaling the pre-training size to benefit most from transfer learning, a question remains: what data and method should be used for pre-training? We investigate the impact of pre-training data distribution on the few-shot and full fine-tuning performance using 3 pre-training methods (supervised, contrastive language-image and image-image), 7 pre-training datasets, and 9 downstream datasets. Through extensive controlled experiments, we find that the choice of the pre-training data source is essential for the few-shot transfer, but its role decreases as more data is made available for fine-tuning. Additionally, we explore the role of data curation and examine the trade-offs between label noise and the size of the pre-training dataset. We find that using 2000X more pre-training data from LAION can match the performance of supervised ImageNet pre-training. Furthermore, we investigate the effect of pre-training methods, comparing language-image contrastive vs. image-image contrastive, and find that the latter leads to better downstream accuracy"
                },
                {
                    "title": "Extracting Training Data from Diffusion Models",
                    "abstract": "Image diffusion models such as DALL-E 2, Imagen, and Stable Diffusion have attracted significant attention due to their ability to generate high-quality synthetic images. In this work, we show that diffusion models memorize individual images from their training data and emit them at generation time. With a generate-and-filter pipeline, we extract over a thousand training examples from state-of-the-art models, ranging from photographs of individual people to trademarked company logos. We also train hundreds of diffusion models in various settings to analyze how different modeling and data decisions affect privacy. Overall, our results show that diffusion models are much less private than prior generative models such as GANs, and that mitigating these vulnerabilities may require new advances in privacy-preserving training."
                },
                {
                    "title": "Optimizing Prompts for Text-to-Image Generation",
                    "abstract": "Well-designed prompts can guide text-to-image models to generate amazing images. However, the performant prompts are often model-specific and misaligned with user input. Instead of laborious human engineering, we propose prompt adaptation, a general framework that automatically adapts original user input to model-preferred prompts. Specifically, we first perform supervised fine-tuning with a pretrained language model on a small collection of manually engineered prompts. Then we use reinforcement learning to explore better prompts. We define a reward function that encourages the policy to generate more aesthetically pleasing images while preserving the original user intentions. Experimental results on Stable Diffusion show that our method outperforms manual prompt engineering in terms of both automatic metrics and human preference ratings. Moreover, reinforcement learning further boosts performance, especially on out-of-domain prompts. The pretrained checkpoints are available at https://aka.ms/promptist. The demo can be found at https://aka.ms/promptist-demo."
                },
                {
                    "title": "Fake it Till You Make it: Learning Transferable Representations from Synthetic ImageNet Clones",
                    "abstract": "Recent image generation models such as Stable Diffusion have exhibited an impressive ability to generate fairly realistic images starting from a simple text prompt. Could such models render real images obsolete for training image prediction models? In this paper, we answer part of this provocative question by investigating the need for real images when training models for ImageNet classification. Provided only with the class names that have been used to build the dataset, we explore the ability of Stable Diffusion to generate synthetic clones of ImageNet and measure how useful these are for training classification models from scratch. We show that with minimal and class-agnostic prompt engineering, ImageNet clones are able to close a large part of the gap between models produced by synthetic images and models trained with real images, for the several standard classification benchmarks that we consider in this study. More importantly, we show that models trained on synthetic images exhibit strong generalization properties and perform on par with models trained on real data for transfer. Project page: https://europe.naverlabs.com/imagenet-sd"
                },
                {
                    "title": "SuS-X: Training-Free Name-Only Transfer of Vision-Language Models",
                    "abstract": "Contrastive Language-Image Pre-training (CLIP) has emerged as a simple yet effective way to train large-scale vision-language models. CLIP demonstrates impressive zero-shot classification and retrieval performance on diverse downstream tasks. However, to leverage its full potential, fine-tuning still appears to be necessary. Fine-tuning the entire CLIP model can be resource-intensive and unstable. Moreover, recent methods that aim to circumvent this need for fine-tuning still require access to images from the target task distribution. In this paper, we pursue a different approach and explore the regime of training-free \"name-only transfer\" in which the only knowledge we possess about the downstream task comprises the names of downstream target categories. We propose a novel method, SuS-X, consisting of two key building blocks\u2014 \"SuS\" and \"TIP-X\", that requires neither intensive fine-tuning nor costly labelled data. SuS-X achieves state-of-the-art (SoTA) zero-shot classification results on 19 benchmark datasets. We further show the utility of TIP-X in the training-free few-shot setting, where we again achieve SoTA results over strong training-free baselines. Code is available at https://github.com/vishaal27/SuS-X."
                },
                {
                    "title": "Large Language Models Struggle to Learn Long-Tail Knowledge",
                    "abstract": "The Internet contains a wealth of knowledge -- from the birthdays of historical figures to tutorials on how to code -- all of which may be learned by language models. However, while certain pieces of information are ubiquitous on the web, others appear extremely rarely. In this paper, we study the relationship between the knowledge memorized by large language models and the information in pre-training datasets scraped from the web. In particular, we show that a language model's ability to answer a fact-based question relates to how many documents associated with that question were seen during pre-training. We identify these relevant documents by entity linking pre-training datasets and counting documents that contain the same entities as a given question-answer pair. Our results demonstrate strong correlational and causal relationships between accuracy and relevant document count for numerous question answering datasets (e.g., TriviaQA), pre-training corpora (e.g., ROOTS), and model sizes (e.g., 176B parameters). Moreover, while larger models are better at learning long-tail knowledge, we estimate that today's models must be scaled by many orders of magnitude to reach competitive QA performance on questions with little support in the pre-training data. Finally, we show that retrieval-augmentation can reduce the dependence on relevant pre-training information, presenting a promising approach for capturing the long-tail."
                },
                {
                    "title": "LAION-5B: An open large-scale dataset for training next generation image-text models",
                    "abstract": "Groundbreaking language-vision architectures like CLIP and DALL-E proved the utility of training on large amounts of noisy image-text data, without relying on expensive accurate labels used in standard vision unimodal supervised learning. The resulting models showed capabilities of strong text-guided image generation and transfer to downstream tasks, while performing remarkably at zero-shot classification with noteworthy out-of-distribution robustness. Since then, large-scale language-vision models like ALIGN, BASIC, GLIDE, Flamingo and Imagen made further improvements. Studying the training and capabilities of such models requires datasets containing billions of image-text pairs. Until now, no datasets of this size have been made openly available for the broader research community. To address this problem and democratize research on large-scale multi-modal models, we present LAION-5B - a dataset consisting of 5.85 billion CLIP-filtered image-text pairs, of which 2.32B contain English language. We show successful replication and fine-tuning of foundational models like CLIP, GLIDE and Stable Diffusion using the dataset, and discuss further experiments enabled with an openly available dataset of this scale. Additionally we provide several nearest neighbor indices, an improved web-interface for dataset exploration and subset generation, and detection scores for watermark, NSFW, and toxic content detection. Announcement page https://laion.ai/laion-5b-a-new-era-of-open-large-scale-multi-modal-datasets/"
                },
                {
                    "title": "Is synthetic data from generative models ready for image recognition?",
                    "abstract": "Recent text-to-image generation models have shown promising results in generating high-fidelity photo-realistic images. Though the results are astonishing to human eyes, how applicable these generated images are for recognition tasks remains under-explored. In this work, we extensively study whether and how synthetic images generated from state-of-the-art text-to-image generation models can be used for image recognition tasks, and focus on two perspectives: synthetic data for improving classification models in data-scarce settings (i.e. zero-shot and few-shot), and synthetic data for large-scale model pre-training for transfer learning. We showcase the powerfulness and shortcomings of synthetic data from existing generative models, and propose strategies for better applying synthetic data for recognition tasks. Code: https://github.com/CVMI-Lab/SyntheticData."
                },
                {
                    "title": "PaLI: A Jointly-Scaled Multilingual Language-Image Model",
                    "abstract": "Effective scaling and a flexible task interface enable large language models to excel at many tasks. We present PaLI (Pathways Language and Image model), a model that extends this approach to the joint modeling of language and vision. PaLI generates text based on visual and textual inputs, and with this interface performs many vision, language, and multimodal tasks, in many languages. To train PaLI, we make use of large pre-trained encoder-decoder language models and Vision Transformers (ViTs). This allows us to capitalize on their existing capabilities and leverage the substantial cost of training them. We find that joint scaling of the vision and language components is important. Since existing Transformers for language are much larger than their vision counterparts, we train a large, 4-billion parameter ViT (ViT-e) to quantify the benefits from even larger-capacity vision models. To train PaLI, we create a large multilingual mix of pretraining tasks, based on a new image-text training set containing 10B images and texts in over 100 languages. PaLI achieves state-of-the-art in multiple vision and language tasks (such as captioning, visual question-answering, scene-text understanding), while retaining a simple, modular, and scalable design."
                },
                {
                    "title": "Quality Not Quantity: On the Interaction between Dataset Design and Robustness of CLIP",
                    "abstract": "Web-crawled datasets have enabled remarkable generalization capabilities in recent image-text models such as CLIP (Contrastive Language-Image pre-training) or Flamingo, but little is known about the dataset creation processes. In this work, we introduce a testbed of six publicly available data sources - YFCC, LAION, Conceptual Captions, WIT, RedCaps, Shutterstock - to investigate how pre-training distributions induce robustness in CLIP. We find that the performance of the pre-training data varies substantially across distribution shifts, with no single data source dominating. Moreover, we systematically study the interactions between these data sources and find that combining multiple sources does not necessarily yield better models, but rather dilutes the robustness of the best individual data source. We complement our empirical findings with theoretical insights from a simple setting, where combining the training data also results in diluted robustness. In addition, our theoretical model provides a candidate explanation for the success of the CLIP-based data filtering technique recently employed in the LAION dataset. Overall our results demonstrate that simply gathering a large amount of data from the web is not the most effective way to build a pre-training dataset for robust generalization, necessitating further study into dataset design. Code is available at https://github.com/mlfoundations/clip_quality_not_quantity."
                },
                {
                    "title": "Testing Relational Understanding in Text-Guided Image Generation",
                    "abstract": "Relations are basic building blocks of human cognition. Classic and recent work suggests that many relations are early developing, and quickly perceived. Machine models that aspire to human-level perception and reasoning should reflect the ability to recognize and reason generatively about relations. We report a systematic empirical examination of a recent text-guided image generation model (DALL-E 2), using a set of 15 basic physical and social relations studied or proposed in the literature, and judgements from human participants (N = 169). Overall, we find that only ~22% of images matched basic relation prompts. Based on a quantitative examination of people's judgments, we suggest that current image generation models do not yet have a grasp of even basic relations involving simple objects and agents. We examine reasons for model successes and failures, and suggest possible improvements based on computations observed in biological intelligence."
                },
                {
                    "title": "Measuring Causal Effects of Data Statistics on Language Model's 'Factual' Predictions",
                    "abstract": "Large amounts of training data are one of the major reasons for the high performance of state-of-the-art NLP models. But what exactly in the training data causes a model to make a certain prediction? We seek to answer this question by providing a language for describing how training data influences predictions, through a causal framework. Importantly, our framework bypasses the need to retrain expensive models and allows us to estimate causal effects based on observational data alone. Addressing the problem of extracting factual knowledge from pretrained language models (PLMs), we focus on simple data statistics such as co-occurrence counts and show that these statistics do influence the predictions of PLMs, suggesting that such models rely on shallow heuristics. Our causal framework and our results demonstrate the importance of studying datasets and the benefits of causality for understanding NLP models."
                },
                {
                    "title": "Is a Caption Worth a Thousand Images? A Controlled Study for Representation Learning",
                    "abstract": "The development of CLIP [Radford et al., 2021] has sparked a debate on whether language supervision can result in vision models with more transferable representations than traditional image-only methods. Our work studies this question through a carefully controlled comparison of two approaches in terms of their ability to learn representations that generalize to downstream classification tasks. We find that when the pre-training dataset meets certain criteria -- it is sufficiently large and contains descriptive captions with low variability -- image-only methods do not match CLIP's transfer performance, even when they are trained with more image data. However, contrary to what one might expect, there are practical settings in which these criteria are not met, wherein added supervision through captions is actually detrimental. Motivated by our findings, we devise simple prescriptions to enable CLIP to better leverage the language information present in existing pre-training datasets."
                },
                {
                    "title": "Beyond neural scaling laws: beating power law scaling via data pruning",
                    "abstract": "Widely observed neural scaling laws, in which error falls off as a power of the training set size, model size, or both, have driven substantial performance improvements in deep learning. However, these improvements through scaling alone require considerable costs in compute and energy. Here we focus on the scaling of error with dataset size and show how in theory we can break beyond power law scaling and potentially even reduce it to exponential scaling instead if we have access to a high-quality data pruning metric that ranks the order in which training examples should be discarded to achieve any pruned dataset size. We then test this improved scaling prediction with pruned dataset size empirically, and indeed observe better than power law scaling in practice on ResNets trained on CIFAR-10, SVHN, and ImageNet. Next, given the importance of finding high-quality pruning metrics, we perform the first large-scale benchmarking study of ten different data pruning metrics on ImageNet. We find most existing high performing metrics scale poorly to ImageNet, while the best are computationally intensive and require labels for every image. We therefore developed a new simple, cheap and scalable self-supervised pruning metric that demonstrates comparable performance to the best supervised metrics. Overall, our work suggests that the discovery of good data-pruning metrics may provide a viable path forward to substantially improved neural scaling laws, thereby reducing the resource costs of modern deep learning."
                },
                {
                    "title": "Scaling Autoregressive Models for Content-Rich Text-to-Image Generation",
                    "abstract": "We present the Pathways Autoregressive Text-to-Image (Parti) model, which generates high-fidelity photorealistic images and supports content-rich synthesis involving complex compositions and world knowledge. Parti treats text-to-image generation as a sequence-to-sequence modeling problem, akin to machine translation, with sequences of image tokens as the target outputs rather than text tokens in another language. This strategy can naturally tap into the rich body of prior work on large language models, which have seen continued advances in capabilities and performance through scaling data and model sizes. Our approach is simple: First, Parti uses a Transformer-based image tokenizer, ViT-VQGAN, to encode images as sequences of discrete tokens. Second, we achieve consistent quality improvements by scaling the encoder-decoder Transformer model up to 20B parameters, with a new state-of-the-art zero-shot FID score of 7.23 and finetuned FID score of 3.22 on MS-COCO. Our detailed analysis on Localized Narratives as well as PartiPrompts (P2), a new holistic benchmark of over 1600 English prompts, demonstrate the effectiveness of Parti across a wide variety of categories and difficulty aspects. We also explore and highlight limitations of our models in order to define and exemplify key areas of focus for further improvements. See https://parti.research.google/ for high-resolution images."
                },
                {
                    "title": "CyCLIP: Cyclic Contrastive Language-Image Pretraining",
                    "abstract": "Recent advances in contrastive representation learning over paired image-text data have led to models such as CLIP that achieve state-of-the-art performance for zero-shot classification and distributional robustness. Such models typically require joint reasoning in the image and text representation spaces for downstream inference tasks. Contrary to prior beliefs, we demonstrate that the image and text representations learned via a standard contrastive objective are not interchangeable and can lead to inconsistent downstream predictions. To mitigate this issue, we formalize consistency and propose CyCLIP, a framework for contrastive representation learning that explicitly optimizes for the learned representations to be geometrically consistent in the image and text space. In particular, we show that consistent representations can be learned by explicitly symmetrizing (a) the similarity between the two mismatched image-text pairs (cross-modal consistency); and (b) the similarity between the image-image pair and the text-text pair (in-modal consistency). Empirically, we show that the improved consistency in CyCLIP translates to significant gains over CLIP, with gains ranging from 10%-24% for zero-shot classification accuracy on standard benchmarks (CIFAR-10, CIFAR-100, ImageNet1K) and 10%-27% for robustness to various natural distribution shifts. The code is available at https://github.com/goel-shashank/CyCLIP."
                },
                {
                    "title": "Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding",
                    "abstract": "We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding. Imagen builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation. Our key discovery is that generic large language models (e.g. T5), pretrained on text-only corpora, are surprisingly effective at encoding text for image synthesis: increasing the size of the language model in Imagen boosts both sample fidelity and image-text alignment much more than increasing the size of the image diffusion model. Imagen achieves a new state-of-the-art FID score of 7.27 on the COCO dataset, without ever training on COCO, and human raters find Imagen samples to be on par with the COCO data itself in image-text alignment. To assess text-to-image models in greater depth, we introduce DrawBench, a comprehensive and challenging benchmark for text-to-image models. With DrawBench, we compare Imagen with recent methods including VQ-GAN+CLIP, Latent Diffusion Models, and DALL-E 2, and find that human raters prefer Imagen over other models in side-by-side comparisons, both in terms of sample quality and image-text alignment. See https://imagen.research.google/ for an overview of the results."
                },
                {
                    "title": "CoCa: Contrastive Captioners are Image-Text Foundation Models",
                    "abstract": "Exploring large-scale pretrained foundation models is of significant interest in computer vision because these models can be quickly transferred to many downstream tasks. This paper presents Contrastive Captioner (CoCa), a minimalist design to pretrain an image-text encoder-decoder foundation model jointly with contrastive loss and captioning loss, thereby subsuming model capabilities from contrastive approaches like CLIP and generative methods like SimVLM. In contrast to standard encoder-decoder transformers where all decoder layers attend to encoder outputs, CoCa omits cross-attention in the first half of decoder layers to encode unimodal text representations, and cascades the remaining decoder layers which cross-attend to the image encoder for multimodal image-text representations. We apply a contrastive loss between unimodal image and text embeddings, in addition to a captioning loss on the multimodal decoder outputs which predicts text tokens autoregressively. By sharing the same computational graph, the two training objectives are computed efficiently with minimal overhead. CoCa is pretrained end-to-end and from scratch on both web-scale alt-text data and annotated images by treating all labels simply as text, seamlessly unifying natural language supervision for representation learning. Empirically, CoCa achieves state-of-the-art performance with zero-shot transfer or minimal task-specific adaptation on a broad range of downstream tasks, spanning visual recognition (ImageNet, Kinetics-400/600/700, Moments-in-Time), crossmodal retrieval (MSCOCO, Flickr30K, MSR-VTT), multimodal understanding (VQA, SNLI-VE, NLVR2), and image captioning (MSCOCO, NoCaps). Notably on ImageNet classification, CoCa obtains 86.3% zero-shot top-1 accuracy, 90.6% with a frozen encoder and learned classification head, and new state-of-the-art 91.0% top-1 accuracy on ImageNet with a finetuned encoder."
                },
                {
                    "title": "Data Determines Distributional Robustness in Contrastive Language Image Pre-training (CLIP)",
                    "abstract": "Contrastively trained language-image models such as CLIP, ALIGN, and BASIC have demonstrated unprecedented robustness to multiple challenging natural distribution shifts. Since these language-image models differ from previous training approaches in several ways, an important question is what causes the large robustness gains. We answer this question via a systematic experimental investigation. Concretely, we study five different possible causes for the robustness gains: (i) the training set size, (ii) the training distribution, (iii) language supervision at training time, (iv) language supervision at test time, and (v) the contrastive loss function. Our experiments show that the more diverse training distribution is the main cause for the robustness gains, with the other factors contributing little to no robustness. Beyond our experimental results, we also introduce ImageNet-Captions, a version of ImageNet with original text annotations from Flickr, to enable further controlled experiments of language-image training."
                },
                {
                    "title": "Data Distributional Properties Drive Emergent In-Context Learning in Transformers",
                    "abstract": "Large transformer-based models are able to perform in-context few-shot learning, without being explicitly trained for it. This observation raises the question: what aspects of the training regime lead to this emergent behavior? Here, we show that this behavior is driven by the distributions of the training data itself. In-context learning emerges when the training data exhibits particular distributional properties such as burstiness (items appear in clusters rather than being uniformly distributed over time) and having large numbers of rarely occurring classes. In-context learning also emerges more strongly when item meanings or interpretations are dynamic rather than fixed. These properties are exemplified by natural language, but are also inherent to naturalistic data in a wide range of other domains. They also depart significantly from the uniform, i.i.d. training distributions typically used for standard supervised learning. In our initial experiments, we found that in-context learning traded off against more conventional weight-based learning, and models were unable to achieve both simultaneously. However, our later experiments uncovered that the two modes of learning could co-exist in a single model when it was trained on data following a skewed Zipfian distribution -- another common property of naturalistic data, including language. In further experiments, we found that naturalistic data distributions were only able to elicit in-context learning in transformers, and not in recurrent models. In sum, our findings indicate how the transformer architecture works together with particular properties of the training data to drive the intriguing emergent in-context learning behaviour of large language models, and how future work might encourage both in-context and in-weights learning in domains beyond language."
                },
                {
                    "title": "Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality",
                    "abstract": "We present a novel task and dataset for evaluating the ability of vision and language models to conduct visio-linguistic compositional reasoning, which we call Winoground. Given two images and two captions, the goal is to match them correctly-but crucially, both captions contain a completely identical set of words, only in a different order. The dataset was carefully hand-curated by expert annotators and is labeled with a rich set offine-grained tags to assist in analyzing model performance. We probe a diverse range of state-of-the-art vision and language models and find that, surprisingly, none of them do much better than chance. Evidently, these models are not as skilled at visio-linguistic compositional reasoning as we might have hoped. We perform an extensive analysis to obtain insights into how future work might try to mitigate these models' shortcomings. We aim for Winoground to serve as a useful evaluation set for advancing the state of the art and driving further progress in the field. The dataset is available at https://huggingface.co/datasets/facebook/winoground."
                },
                {
                    "title": "FitCLIP: Refining Large-Scale Pretrained Image-Text Models for Zero-Shot Video Understanding Tasks",
                    "abstract": "Large-scale pretrained image-text models have shown incredible zero-shot performance in a handful of tasks, including video ones such as action recognition and text-to-video retrieval. However, these models have not been adapted to video, mainly because they do not account for the time dimension but also because video frames are different from the typical images (e.g., containing motion blur, and less sharpness). In this paper, we present a fine-tuning strategy to refine these large-scale pretrained image-text models for zero-shot video understanding tasks. We show that by carefully adapting these models we obtain considerable improvements on two zero-shot Action Recognition tasks and three zero-shot Text-to-video Retrieval tasks. The code is available at https://github.com/bryant1410/fitclip"
                },
                {
                    "title": "Vision-Language Pre-Training with Triple Contrastive Learning",
                    "abstract": "Vision-language representation learning largely benefits from image-text alignment through contrastive losses (e.g., InfoNCE loss). The success of this alignment strategy is attributed to its capability in maximizing the mutual information (MI) between an image and its matched text. However, simply performing cross-modal alignment (CMA) ignores data potential within each modality, which may result in degraded representations. For instance, although CMA-based models are able to map image-text pairs close together in the embedding space, they fail to ensure that similar inputs from the same modality stay close by. This problem can get even worse when the pre-training data is noisy. In this paper, we propose triple contrastive learning (TCL) for vision-language pre-training by leveraging both cross-modal and intra-modal self-supervision. Besides CMA, TCL introduces an intra-modal contrastive objective to provide complementary benefits in representation learning. To take advantage of localized and structural information from image and text input, TCL further maximizes the average MI between local regions of image/text and their global summary. To the best of our knowledge, ours is the first work that takes into account local structure information for multi-modality representation learning. Experimental evaluations show that our approach is competitive and achieves the new state of the art on various common downstream vision-language tasks such as image-text retrieval and visual question answering."
                },
                {
                    "title": "Quantifying Memorization Across Neural Language Models",
                    "abstract": "Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations."
                },
                {
                    "title": "DALL-EVAL: Probing the Reasoning Skills and Social Biases of Text-to-Image Generation Models",
                    "abstract": "Recently, DALL-E [45], a multimodal transformer language model, and its variants including diffusion models have shown high-quality text-to-image generation capabilities. However, despite the realistic image generation results, there has not been a detailed analysis of how to evaluate such models. In this work, we investigate the visual reasoning capabilities and social biases of different text-to-image models, covering both multimodal transformer language models and diffusion models. First, we measure three visual reasoning skills: object recognition, object counting, and spatial relation understanding. For this, we propose PaintSkills, a compositional diagnostic evaluation dataset that measures these skills. Despite the high-fidelity image generation capability, a large gap exists between the performance of recent models and the upper bound accuracy in object counting and spatial relation understanding skills. Second, we assess the gender and skin tone biases by measuring the gender/skin tone distribution of generated images across various professions and attributes. We demonstrate that recent text-to-image generation models learn specific biases about gender and skin tone from web image-text pairs. We hope our work will help guide future progress in improving text-to-image generation models on visual reasoning skills and learning socially unbiased representations.1"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "LiT: Zero-Shot Transfer with Locked-image text Tuning",
                    "abstract": "This paper presents contrastive-tuning, a simple method employing contrastive training to align image and text mod-els while still taking advantage of their pre-training. In our empirical study we find that locked pre-trained image mod-els with unlocked text models work best. We call this in-stance of contrastive-tuning \u201cLocked-image Tuning\u201d (LiT), which just teaches a text model to read out good repre-sentations from a pre-trained image model for new tasks. A LiT model gains the capability of zero-shot transfer to new vision tasks, such as image classification or retrieval. The proposed LiT is widely applicable; it works reliably with multiple pre-training methods (supervised and unsu-pervised) and across diverse architectures (ResNet, Vision Transformers and MLP-Mixer) using three different image-text datasets. With the transformer-based pre-trained ViT-g/14 model, the LiT model achieves 84.5% zero-shot trans-fer accuracy on the ImageNet test set, and 81.1% on the challenging out-of-distribution ObjectNet test set."
                },
                {
                    "title": "LAION-400M: Open Dataset of CLIP-Filtered 400 Million Image-Text Pairs",
                    "abstract": "Multi-modal language-vision models trained on hundreds of millions of image-text pairs (e.g. CLIP, DALL-E) gained a recent surge, showing remarkable capability to perform zero- or few-shot learning and transfer even in absence of per-sample labels on target image data. Despite this trend, to date there has been no publicly available datasets of sufficient scale for training such models from scratch. To address this issue, in a community effort we build and release for public LAION-400M, a dataset with CLIP-filtered 400 million image-text pairs, their CLIP embeddings and kNN indices that allow efficient similarity search."
                },
                {
                    "title": "Align before Fuse: Vision and Language Representation Learning with Momentum Distillation",
                    "abstract": "Large-scale vision and language representation learning has shown promising improvements on various vision-language tasks. Most existing methods employ a transformer-based multimodal encoder to jointly model visual tokens (region-based image features) and word tokens. Because the visual tokens and word tokens are unaligned, it is challenging for the multimodal encoder to learn image-text interactions. In this paper, we introduce a contrastive loss to ALign the image and text representations BEfore Fusing (ALBEF) them through cross-modal attention, which enables more grounded vision and language representation learning. Unlike most existing methods, our method does not require bounding box annotations nor high-resolution images. In order to improve learning from noisy web data, we propose momentum distillation, a self-training method which learns from pseudo-targets produced by a momentum model. We provide a theoretical analysis of ALBEF from a mutual information maximization perspective, showing that different training tasks can be interpreted as different ways to generate views for an image-text pair. ALBEF achieves state-of-the-art performance on multiple downstream vision-language tasks. On image-text retrieval, ALBEF outperforms methods that are pre-trained on orders of magnitude larger datasets. On VQA and NLVR$^2$, ALBEF achieves absolute improvements of 2.37% and 3.84% compared to the state-of-the-art, while enjoying faster inference speed. Code and pre-trained models are available at https://github.com/salesforce/ALBEF/."
                },
                {
                    "title": "Deduplicating Training Data Makes Language Models Better",
                    "abstract": "We find that existing language modeling datasets contain many near-duplicate examples and long repetitive substrings.As a result, over 1% of the unprompted output of language models trained on these datasets is copied verbatim from the training data.We develop two tools that allow us to deduplicate training datasets\u2014for example removing from C4 a single 61 word English sentence that is repeated over 60,000 times.Deduplication allows us to train models that emit memorized text ten times less frequently and require fewer training steps to achieve the same or better accuracy.We can also reduce train-test overlap, which affects over 4% of the validation set of standard datasets, thus allowing for more accurate evaluation.Code for deduplication is released at https://github.com/google-research/deduplicate-text-datasets."
                },
                {
                    "title": "CLIPScore: A Reference-free Evaluation Metric for Image Captioning",
                    "abstract": "Image captioning has conventionally relied on reference-based automatic evaluations, where machine captions are compared against captions written by humans. This is in contrast to the reference-free manner in which humans assess caption quality. In this paper, we report the surprising empirical finding that CLIP (Radford et al., 2021), a cross-modal model pretrained on 400M image+caption pairs from the web, can be used for robust automatic evaluation of image captioning without the need for references. Experiments spanning several corpora demonstrate that our new reference-free metric, CLIPScore, achieves the highest correlation with human judgements, outperforming existing reference-based metrics like CIDEr and SPICE. Information gain experiments demonstrate that CLIPScore, with its tight focus on image-text compatibility, is complementary to existing reference-based metrics that emphasize text-text similarities. Thus, we also present a reference-augmented version, RefCLIPScore, which achieves even higher correlation. Beyond literal description tasks, several case studies reveal domains where CLIPScore performs well (clip-art images, alt-text rating), but also where it is relatively weaker in comparison to reference-based metrics, e.g., news captions that require richer contextual knowledge."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Zero-Shot Text-to-Image Generation",
                    "abstract": "Text-to-image generation has traditionally focused on finding better modeling assumptions for training on a fixed dataset. These assumptions might involve complex architectures, auxiliary losses, or side information such as object part labels or segmentation masks supplied during training. We describe a simple approach for this task based on a transformer that autoregressively models the text and image tokens as a single stream of data. With sufficient data and scale, our approach is competitive with previous domain-specific models when evaluated in a zero-shot fashion."
                },
                {
                    "title": "Conceptual 12M: Pushing Web-Scale Image-Text Pre-Training To Recognize Long-Tail Visual Concepts",
                    "abstract": "The availability of large-scale image captioning and visual question answering datasets has contributed significantly to recent successes in vision-and-language pretraining. However, these datasets are often collected with overrestrictive requirements inherited from their original target tasks (e.g., image caption generation), which limit the resulting dataset scale and diversity. We take a step further in pushing the limits of vision-and-language pretraining data by relaxing the data collection pipeline used in Conceptual Captions 3M (CC3M) [54] and introduce the Conceptual 12M (CC12M), a dataset with 12 million image-text pairs specifically meant to be used for visionand-language pre-training. We perform an analysis of this dataset and benchmark its effectiveness against CC3M on multiple downstream tasks with an emphasis on long-tail visual recognition. Our results clearly illustrate the benefit of scaling up pre-training data for vision-and-language tasks, as indicated by the new state-of-the-art results on both the nocaps and Conceptual Captions benchmarks.1"
                },
                {
                    "title": "ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision",
                    "abstract": "Vision-and-Language Pre-training (VLP) has improved performance on various joint vision-and-language downstream tasks. Current approaches to VLP heavily rely on image feature extraction processes, most of which involve region supervision (e.g., object detection) and the convolutional architecture (e.g., ResNet). Although disregarded in the literature, we find it problematic in terms of both (1) efficiency/speed, that simply extracting input features requires much more computation than the multimodal interaction steps; and (2) expressive power, as it is upper bounded to the expressive power of the visual embedder and its predefined visual vocabulary. In this paper, we present a minimal VLP model, Vision-and-Language Transformer (ViLT), monolithic in the sense that the processing of visual inputs is drastically simplified to just the same convolution-free manner that we process textual inputs. We show that ViLT is up to tens of times faster than previous VLP models, yet with competitive or better downstream task performance. Our code and pre-trained weights are available at https://github.com/dandelin/vilt."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Large image datasets: A pyrrhic win for computer vision?",
                    "abstract": "In this paper we investigate problematic practices and consequences of large scale vision datasets (LSVDs). We examine broad issues such as the question of consent and justice as well as specific concerns such as the inclusion of verifiably pornographic images in datasets. Taking the ImageNet-ILSVRC-2012 dataset as an example, we perform a cross-sectional model-based quantitative census covering factors such as age, gender, NSFW content scoring, class- wise accuracy, human-cardinality-analysis, and the semanticity of the image class information in order to statistically investigate the extent and subtleties of ethical transgressions. We then use the census to help hand-curate a look-up-table of images in the ImageNet-ILSVRC-2012 dataset that fall into the categories of verifiably pornographic: shot in a non-consensual setting (up-skirt), beach voyeuristic, and exposed private parts. We survey the landscape of harm and threats both the society at large and individuals face due to uncritical and ill-considered dataset curation practices. We then propose possible courses of correction and critique their pros and cons. We have duly open-sourced all of the code and the census meta-datasets generated in this endeavor for the computer vision community to build on. By unveiling the severity of the threats, our hope is to motivate the constitution of mandatory Institutional Review Boards (IRB) for large scale dataset curation."
                },
                {
                    "title": "COBRA: Contrastive Bi-Modal Representation Algorithm",
                    "abstract": "There are a wide range of applications that involve multi-modal data, such as cross-modal retrieval, visual question-answering, and image captioning. Such applications are primarily dependent on aligned distributions of the different constituent modalities. Existing approaches generate latent embeddings for each modality in a joint fashion by representing them in a common manifold. However these joint embedding spaces fail to sufficiently reduce the modality gap, which affects the performance in downstream tasks. We hypothesize that these embeddings retain the intra-class relationships but are unable to preserve the inter-class dynamics. In this paper, we present a novel framework COBRA that aims to train two modalities (image and text) in a joint fashion inspired by the Contrastive Predictive Coding (CPC) and Noise Contrastive Estimation (NCE) paradigms which preserve both inter and intra-class relationships. We empirically show that this framework reduces the modality gap significantly and generates a robust and task agnostic joint-embedding space. We outperform existing work on four diverse downstream tasks spanning across seven benchmark cross-modal datasets."
                },
                {
                    "title": "Compositionality Decomposed: How do Neural Networks Generalise?",
                    "abstract": "Despite a multitude of empirical studies, little consensus exists on whether neural networks are able to generalise compositionally, a controversy that, in part, stems from a lack of agreement about what it means for a neural model to be compositional. As a response to this controversy, we present a set of tests that provide a bridge between, on the one hand, the vast amount of linguistic and philosophical theory about compositionality of language and, on the other, the successful neural models of language. We collect different interpretations of compositionality and translate them into five theoretically grounded tests for models that are formulated on a task-independent level. In particular, we provide tests to investigate (i) if models systematically recombine known parts and rules (ii) if models can extend their predictions beyond the length they have seen in the training data (iii) if models\u2019 composition operations are local or global (iv) if models\u2019 predictions are robust to synonym substitutions and (v) if models favour rules or exceptions during training. To demonstrate the usefulness of this evaluation paradigm, we instantiate these five tests on a highly compositional data set which we dub PCFG SET and apply the resulting tests to three popular sequence-to-sequence models: a recurrent, a convolution-based and a transformer model. We provide an in-depth analysis of the results, which uncover the strengths and weaknesses of these three architectures and point to potential areas of improvement."
                },
                {
                    "title": "Conceptual Captions: A Cleaned, Hypernymed, Image Alt-text Dataset For Automatic Image Captioning",
                    "abstract": "We present a new dataset of image caption annotations, Conceptual Captions, which contains an order of magnitude more images than the MS-COCO dataset (Lin et al., 2014) and represents a wider variety of both images and image caption styles. We achieve this by extracting and filtering image caption annotations from billions of webpages. We also present quantitative evaluations of a number of image captioning models and show that a model architecture based on Inception-ResNetv2 (Szegedy et al., 2016) for image-feature extraction and Transformer (Vaswani et al., 2017) for sequence modeling achieves the best performance when trained on the Conceptual Captions dataset."
                },
                {
                    "title": "EuroSAT: A Novel Dataset and Deep Learning Benchmark for Land Use and Land Cover Classification",
                    "abstract": "In this paper, we present a patch-based land use and land cover classification approach using Sentinel-2 satellite images. The Sentinel-2 satellite images are openly and freely accessible, and are provided in the earth observation program Copernicus. We present a novel dataset, based on these images that covers 13 spectral bands and is comprised of ten classes with a total of 27\u2009000 labeled and geo-referenced images. Benchmarks are provided for this novel dataset with its spectral bands using state-of-the-art deep convolutional neural networks. An overall classification accuracy of 98.57% was achieved with the proposed novel dataset. The resulting classification system opens a gate toward a number of earth observation applications. We demonstrate how this classification system can be used for detecting land use and land cover changes, and how it can assist in improving geographical maps. The geo-referenced dataset EuroSAT is made publicly available at https://github.com/phelber/eurosat."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Going deeper with convolutions",
                    "abstract": "We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection."
                },
                {
                    "title": "Birdsnap: Large-Scale Fine-Grained Visual Categorization of Birds",
                    "abstract": "We address the problem of large-scale fine-grained visual categorization, describing new methods we have used to produce an online field guide to 500 North American bird species. We focus on the challenges raised when such a system is asked to distinguish between highly similar species of birds. First, we introduce \"one-vs-most classifiers.\" By eliminating highly similar species during training, these classifiers achieve more accurate and intuitive results than common one-vs-all classifiers. Second, we show how to estimate spatio-temporal class priors from observations that are sampled at irregular and biased locations. We show how these priors can be used to significantly improve performance. We then show state-of-the-art recognition performance on a new, large dataset that we make publicly available. These recognition methods are integrated into the online field guide, which is also publicly available."
                },
                {
                    "title": "From image descriptions to visual denotations: New similarity metrics for semantic inference over event descriptions",
                    "abstract": "We propose to use the visual denotations of linguistic expressions (i.e. the set of images they describe) to define novel denotational similarity metrics, which we show to be at least as beneficial as distributional similarities for two tasks that require semantic inference. To compute these denotational similarities, we construct a denotation graph, i.e. a subsumption hierarchy over constituents and their denotations, based on a large corpus of 30K images and 150K descriptive captions."
                },
                {
                    "title": "3D Object Representations for Fine-Grained Categorization",
                    "abstract": "While 3D object representations are being revived in the context of multi-view object class detection and scene understanding, they have not yet attained wide-spread use in fine-grained categorization. State-of-the-art approaches achieve remarkable performance when training data is plentiful, but they are typically tied to flat, 2D representations that model objects as a collection of unconnected views, limiting their ability to generalize across viewpoints. In this paper, we therefore lift two state-of-the-art 2D object representations to 3D, on the level of both local feature appearance and location. In extensive experiments on existing and newly proposed datasets, we show our 3D object representations outperform their state-of-the-art 2D counterparts for fine-grained categorization and demonstrate their efficacy for estimating 3D geometry from images via ultra-wide baseline matching and 3D reconstruction."
                },
                {
                    "title": "Describing Textures in the Wild",
                    "abstract": "Patterns and textures are key characteristics of many natural objects: a shirt can be striped, the wings of a butterfly can be veined, and the skin of an animal can be scaly. Aiming at supporting this dimension in image understanding, we address the problem of describing textures with semantic attributes. We identify a vocabulary of forty-seven texture terms and use them to describe a large dataset of patterns collected \"in the wild\". The resulting Describable Textures Dataset (DTD) is a basis to seek the best representation for recognizing describable texture attributes in images. We port from object recognition to texture recognition the Improved Fisher Vector (IFV) and Deep Convolutional-network Activation Features (DeCAF), and show that surprisingly, they both outperform specialized texture descriptors not only on our problem, but also in established material recognition datasets. We also show that our describable attributes are excellent texture descriptors, transferring between datasets and tasks, in particular, combined with IFV and DeCAF, they significantly outperform the state-of-the-art by more than 10% on both FMD and KTH-TIPS-2b benchmarks. We also demonstrate that they produce intuitive descriptions of materials and Internet images."
                },
                {
                    "title": "Fine-Grained Visual Classification of Aircraft",
                    "abstract": "This paper introduces FGVC-Aircraft, a new dataset containing 10,000 images of aircraft spanning 100 aircraft models, organised in a three-level hierarchy. At the finer level, differences between models are often subtle but always visually measurable, making visual recognition challenging but possible. A benchmark is obtained by defining corresponding classification tasks and evaluation protocols, and baseline results are presented. The construction of this dataset was made possible by the work of aircraft enthusiasts, a strategy that can extend to the study of number of other object classes. Compared to the domains usually considered in fine-grained visual classification (FGVC), for example animals, aircraft are rigid and hence less deformable. They, however, present other interesting modes of variation, including purpose, size, designation, structure, historical style, and branding."
                },
                {
                    "title": "UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild",
                    "abstract": "We introduce UCF101 which is currently the largest dataset of human actions. It consists of 101 action classes, over 13k clips and 27 hours of video data. The database consists of realistic user uploaded videos containing camera motion and cluttered background. Additionally, we provide baseline action recognition results on this new dataset using standard bag of words approach with overall performance of 44.5%. To the best of our knowledge, UCF101 is currently the most challenging dataset of actions due to its large number of classes, large number of clips and also unconstrained nature of such clips."
                },
                {
                    "title": "Cats and dogs",
                    "abstract": "We investigate the fine grained object categorization problem of determining the breed of animal from an image. To this end we introduce a new annotated dataset of pets covering 37 different breeds of cats and dogs. The visual problem is very challenging as these animals, particularly cats, are very deformable and there can be quite subtle differences between the breeds. We make a number of contributions: first, we introduce a model to classify a pet breed automatically from an image. The model combines shape, captured by a deformable part model detecting the pet face, and appearance, captured by a bag-of-words model that describes the pet fur. Fitting the model involves automatically segmenting the animal in the image. Second, we compare two classification approaches: a hierarchical one, in which a pet is first assigned to the cat or dog family and then to a breed, and a flat one, in which the breed is obtained directly. We also investigate a number of animal and image orientated spatial layouts. These models are very good: they beat all previously published results on the challenging ASIRRA test (cat vs dog discrimination). When applied to the task of discriminating the 37 different breeds of pets, the models obtain an average accuracy of about 59%, a very encouraging result considering the difficulty of the problem."
                },
                {
                    "title": "SUN database: Large-scale scene recognition from abbey to zoo",
                    "abstract": "Scene categorization is a fundamental problem in computer vision. However, scene understanding research has been constrained by the limited scope of currently-used databases which do not capture the full variety of scene categories. Whereas standard databases for object categorization contain hundreds of different classes of objects, the largest available dataset of scene categories contains only 15 classes. In this paper we propose the extensive Scene UNderstanding (SUN) database that contains 899 categories and 130,519 images. We use 397 well-sampled categories to evaluate numerous state-of-the-art algorithms for scene recognition and establish new bounds of performance. We measure human scene classification performance on the SUN database and compare this with computational methods. Additionally, we study a finer-grained scene representation to detect scenes embedded inside of larger scenes."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Automated Flower Classification over a Large Number of Classes",
                    "abstract": "We investigate to what extent combinations of features can improve classification performance on a large dataset of similar classes. To this end we introduce a 103 class flower dataset. We compute four different features for the flowers, each describing different aspects, namely the local shape/texture, the shape of the boundary, the overall spatial distribution of petals, and the colour. We combine the features using a multiple kernel framework with a SVM classifier. The weights for each class are learnt using the method of Varma and Ray, which has achieved state of the art performance on other large dataset, such as Caltech 101/256. Our dataset has a similar challenge in the number of classes, but with the added difficulty of large between class similarity and small within class similarity. Results show that learning the optimum kernel combination of multiple features vastly improves the performance, from 55.1% for the best single feature to 72.8% for the combination of all features."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.CL",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nAre current multimodal models truly capable of \u201czero-shot\u201d generalization?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it challenges the prevailing notion of zero-shot generalization in multimodal models, such as CLIP and Stable Diffusion. By understanding the limitations of these models, researchers can refine their approaches to model training and evaluation, leading to more robust and generalizable AI systems. This work could advance knowledge in the field by providing insights into the relationship between pretraining data and model performance, potentially influencing future research directions and practical applications in areas like image recognition, retrieval, and generation.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem stem from the complex interplay between model architecture, pretraining data, and the inherent variability of concepts. Naive approaches may fail because they do not account for the long-tailed distribution of concepts in pretraining datasets, which leads to sample inefficiency. Additionally, the need to control for correlated factors, such as similar samples in pretraining and test data, complicates the analysis. Overcoming these technical and theoretical obstacles requires a nuanced understanding of multimodal learning dynamics and the ability to analyze large-scale datasets effectively.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often overlooked the significance of concept frequency in pretraining datasets, leading to an incomplete understanding of model performance. Barriers such as the lack of comprehensive datasets and the complexity of analyzing multimodal interactions have hindered progress. Our approach differs by systematically analyzing the relationship between concept frequency and model performance across various tasks and datasets, providing a clearer picture of the limitations of current multimodal models and highlighting the need for better data curation strategies.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a comparative analysis of model performance across 27 downstream tasks, focusing on the frequency of test concepts in five large-scale pretraining datasets. We will utilize metrics that quantify model performance in relation to concept frequency, assessing 10 CLIP models and 24 T2I models. The expected outcomes include a detailed understanding of the log-linear scaling trend between concept frequency and model performance, insights into the distribution of concepts in pretraining data, and documentation of image-text misalignment, which will inform future efforts in multimodal model training and evaluation."
            }
        },
        "author_data": {
            "c1ef201f-b031-4652-b35e-ae6146de3f91": {
                "pk": "c1ef201f-b031-4652-b35e-ae6146de3f91",
                "name": "Vishaal Udandarao",
                "collaborators": [
                    "Samuel Albanie",
                    "Matthias Bethge",
                    "Ameya Prabhu",
                    "Suryatej Reddy Vyalla",
                    "Rajiv Ratn Shah",
                    "Sarthak Bhagat",
                    "Shagun Uppal",
                    "Ankush Gupta",
                    "Tanmoy Chakraborty",
                    "Surabhi S. Nath"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Self-Supervised Learning",
                    "Affective Computing",
                    "Vision-Language Models"
                ],
                "publications": [
                    {
                        "title": "DisCont: Self-Supervised Visual Attribute Disentanglement using Context Vectors",
                        "abstract": "Disentangling the underlying feature attributes within an image with no prior supervision is a challenging task. Models that can disentangle attributes well provide greater interpretability and control. In this paper, we propose a self-supervised framework DisCont to disentangle multiple attributes by exploiting the structural inductive biases within images. Motivated by the recent surge in contrastive learning paradigms, our model bridges the gap between self-supervised contrastive learning algorithms and unsupervised disentanglement. We evaluate the efficacy of our approach, both qualitatively and quantitatively, on four benchmark datasets."
                    },
                    {
                        "title": "SuS-X: Training-Free Name-Only Transfer of Vision-Language Models",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) has emerged as a simple yet effective way to train large-scale vision-language models. CLIP demonstrates impressive zero-shot classification and retrieval on diverse downstream tasks. However, to leverage its full potential, fine-tuning still appears to be necessary. Fine-tuning the entire CLIP model can be resource-intensive and unstable. Moreover, recent methods that aim to circumvent this need for fine-tuning still require access to images from the target distribution. In this paper, we pursue a different approach and explore the regime of training-free \"name-only transfer\" in which the only knowledge we possess about the downstream task comprises the names of downstream target categories. We propose a novel method, SuS-X, consisting of two key building blocks -- SuS and TIP-X, that requires neither intensive fine-tuning nor costly labelled data. SuS-X achieves state-of-the-art zero-shot classification results on 19 benchmark datasets. We further show the utility of TIP-X in the training-free few-shot setting, where we again achieve state-of-the-art results over strong training-free baselines. Code is available at https://github.com/vishaal27/SuS-X."
                    },
                    {
                        "title": "Memeify: A Large-Scale Meme Generation System",
                        "abstract": "Interest in the research areas related to meme propagation and generation has been increasing rapidly in the last couple of years. Meme datasets available online are either specific to a context or contain no class information. Here, we prepare a large-scale dataset of memes with captions and class labels. The dataset consists of 1.1 million meme captions from 128 classes. We also provide reasoning for the existence of broad categories, called \"themes\" across the meme dataset; each theme consists of multiple meme classes. Our generation system uses a trained state-of-the-art transformer-based model for caption generation by employing an encoder-decoder architecture. We develop a web interface, called Memeify for users to generate memes of their choice, and explain in detail, the working of individual components of the system. We also perform a qualitative evaluation of the generated memes by conducting a user study. A link to the demonstration of the Memeify system is https://youtu.be/P_Tfs0X-czs."
                    },
                    {
                        "title": "It's LeVAsa not LevioSA! Latent Encodings for Valence-Arousal Structure Alignment",
                        "abstract": "In recent years, great strides have been made in the field of affective computing. Several models have been developed to represent and quantify emotions. Two popular ones include (i) categorical models which represent emotions as discrete labels, and (ii) dimensional models which represent emotions in a Valence-Arousal (VA) circumplex domain. However, there is no standard for annotation mapping between the two labelling methods. We build a novel algorithm for mapping categorical and dimensional model labels using annotation transfer across affective facial image datasets. Further, we utilize the transferred annotations to learn rich and interpretable data representations using a variational autoencoder (VAE). We present \"LeVAsa\", a VAE model that learns implicit structure by aligning the latent space with the VA space. We evaluate the efficacy of LeVAsa by comparing performance with the Vanilla VAE using quantitative and qualitative analysis on two benchmark affective image datasets. Our results reveal that LeVAsa achieves high latent-circumplex alignment which leads to improved downstream categorical emotion prediction. The work also demonstrates the trade-off between degree of alignment and quality of reconstructions."
                    },
                    {
                        "title": "Visual Data-Type Understanding does not emerge from Scaling Vision-Language Models",
                        "abstract": "Recent advances in the development of vision-language models (VLMs) are yielding remarkable success in recognizing visual semantic content, including impressive instances of compositional image understanding. Here, we introduce the novel task of Visual Data-Type Identification, a basic perceptual skill with implications for data curation (e.g., noisy data-removal from large datasets, domain-specific retrieval) and autonomous vision (e.g., distinguishing changing weather conditions from camera lens staining). We develop two datasets consisting of animal images altered across a diverse set of 27 visual data-types, spanning four broad categories. An extensive zero-shot evaluation of 39 VLMs, ranging from 100M to 80B parameters, shows a nuanced performance landscape. While VLMs are reasonably good at identifying certain stylistic \\textit{data-types}, such as cartoons and sketches, they struggle with simpler data-types arising from basic manipulations like image rotations or additive noise. Our findings reveal that (i) model scaling alone yields marginal gains for contrastively-trained models like CLIP, and (ii) there is a pronounced drop in performance for the largest auto-regressively trained VLMs like OpenFlamingo. This finding points to a blind spot in current frontier VLMs: they excel in recognizing semantic content but fail to acquire an understanding of visual data-types through scaling. By analyzing the pre-training distributions of these models and incorporating data-type information into the captions during fine-tuning, we achieve a significant enhancement in performance. By exploring this previously uncharted task, we aim to set the stage for further advancing VLMs to equip them with visual data-type understanding. Code and datasets are released at https://github.com/bethgelab/DataTypeIdentification."
                    },
                    {
                        "title": "Lifelong Benchmarks: Efficient Model Evaluation in an Era of Rapid Progress",
                        "abstract": "Standardized benchmarks drive progress in machine learning. However, with repeated testing, the risk of overfitting grows as algorithms over-exploit benchmark idiosyncrasies. In our work, we seek to mitigate this challenge by compiling ever-expanding large-scale benchmarks called Lifelong Benchmarks. As exemplars of our approach, we create Lifelong-CIFAR10 and Lifelong-ImageNet, containing (for now) 1.69M and 1.98M test samples, respectively. While reducing overfitting, lifelong benchmarks introduce a key challenge: the high cost of evaluating a growing number of models across an ever-expanding sample set. To address this challenge, we also introduce an efficient evaluation framework: Sort \\& Search (S&S), which reuses previously evaluated models by leveraging dynamic programming algorithms to selectively rank and sub-select test samples, enabling cost-effective lifelong benchmarking. Extensive empirical evaluations across 31,000 models demonstrate that S&S achieves highly-efficient approximate accuracy measurement, reducing compute cost from 180 GPU days to 5 GPU hours (1000x reduction) on a single A100 GPU, with low approximation error. As such, lifelong benchmarks offer a robust, practical solution to the \"benchmark exhaustion\" problem."
                    },
                    {
                        "title": "On the Inference of Soft Biometrics from Typing Patterns Collected in a Multi-device Environment",
                        "abstract": "In this paper, we study the inference of gender, major/minor (computer science, non-computer science), typing style, age, and height from the typing patterns collected from 117 individuals in a multi-device environment. The inference of the first three identifiers was considered as classification tasks, while the rest as regression tasks. For classification tasks, we benchmark the performance of six classical machine learning (ML) and four deep learning (DL) classifiers. On the other hand, for regression tasks, we evaluated three ML and four DL-based regressors. The overall experiment consisted of two text-entry (free and fixed) and four device (Desktop, Tablet, Phone, and Combined) configurations. The best arrangements achieved accuracies of 96.15%, 93.02%, and 87.80% for typing style, gender, and major/minor, respectively, and mean absolute errors of 1.77 years and 2.65 inches for age and height, respectively. The results are promising considering the variety of application scenarios that we have listed in this work."
                    },
                    {
                        "title": "CiteME: Can Language Models Accurately Cite Scientific Claims?",
                        "abstract": "Thousands of new scientific papers are published each month. Such information overload complicates researcher efforts to stay current with the state-of-the-art as well as to verify and correctly attribute claims. We pose the following research question: Given a text excerpt referencing a paper, could an LM act as a research assistant to correctly identify the referenced paper? We advance efforts to answer this question by building a benchmark that evaluates the abilities of LMs in citation attribution. Our benchmark, CiteME, consists of text excerpts from recent machine learning papers, each referencing a single other paper. CiteME use reveals a large gap between frontier LMs and human performance, with LMs achieving only 4.2-18.5% accuracy and humans 69.7%. We close this gap by introducing CiteAgent, an autonomous system built on the GPT-4o LM that can also search and read papers, which achieves an accuracy of 35.3\\% on CiteME. Overall, CiteME serves as a challenging testbed for open-ended claim attribution, driving the research community towards a future where any claim made by an LM can be automatically verified and discarded if found to be incorrect."
                    },
                    {
                        "title": "EDUQA: Educational Domain Question Answering System using Conceptual Network Mapping",
                        "abstract": "Most of the existing question answering models can be largely compiled into two categories: i) open domain question answering models that answer generic questions and use large-scale knowledge base along with the targeted web-corpus retrieval and ii) closed domain question answering models that address focused questioning area and use complex deep learning models. Both the above models derive answers through textual comprehension methods. Due to their inability to capture the pedagogical meaning of textual content, these models are not appropriately suited to the educational field for pedagogy. In this paper, we propose an on-the-fly conceptual network model that incorporates educational semantics. The proposed model preserves correlations between conceptual entities by applying intelligent indexing algorithms on the concept network so as to improve answer generation. This model can be utilized for building interactive conversational agents for aiding classroom learning."
                    },
                    {
                        "title": "COBRA: Contrastive Bi-Modal Representation Algorithm",
                        "abstract": "There are a wide range of applications that involve multi-modal data, such as cross-modal retrieval, visual question-answering, and image captioning. Such applications are primarily dependent on aligned distributions of the different constituent modalities. Existing approaches generate latent embeddings for each modality in a joint fashion by representing them in a common manifold. However these joint embedding spaces fail to sufficiently reduce the modality gap, which affects the performance in downstream tasks. We hypothesize that these embeddings retain the intra-class relationships but are unable to preserve the inter-class dynamics. In this paper, we present a novel framework COBRA that aims to train two modalities (image and text) in a joint fashion inspired by the Contrastive Predictive Coding (CPC) and Noise Contrastive Estimation (NCE) paradigms which preserve both inter and intra-class relationships. We empirically show that this framework reduces the modality gap significantly and generates a robust and task agnostic joint-embedding space. We outperform existing work on four diverse downstream tasks spanning across seven benchmark cross-modal datasets."
                    },
                    {
                        "title": "A Practitioner's Guide to Continual Multimodal Pretraining",
                        "abstract": "Multimodal foundation models serve numerous applications at the intersection of vision and language. Still, despite being pretrained on extensive data, they become outdated over time. To keep models updated, research into continual pretraining mainly explores scenarios with either (1) infrequent, indiscriminate updates on large-scale new data, or (2) frequent, sample-level updates. However, practical model deployment often operates in the gap between these two limit cases, as real-world applications often demand adaptation to specific subdomains, tasks or concepts -- spread over the entire, varying life cycle of a model. In this work, we complement current perspectives on continual pretraining through a research test bed as well as provide comprehensive guidance for effective continual model updates in such scenarios. We first introduce FoMo-in-Flux, a continual multimodal pretraining benchmark with realistic compute constraints and practical deployment requirements, constructed over 63 datasets with diverse visual and semantic coverage. Using FoMo-in-Flux, we explore the complex landscape of practical continual pretraining through multiple perspectives: (1) A data-centric investigation of data mixtures and stream orderings that emulate real-world deployment situations, (2) a method-centric investigation ranging from simple fine-tuning and traditional continual learning strategies to parameter-efficient updates and model merging, (3) meta learning rate schedules and mechanistic design choices, and (4) the influence of model and compute scaling. Together, our insights provide a practitioner's guide to continual multimodal pretraining for real-world deployment. Our benchmark and code is here: https://github.com/ExplainableML/fomo_in_flux."
                    }
                ]
            },
            "ddb71a66-7b5d-4a34-99ac-a057f78cfa77": {
                "pk": "ddb71a66-7b5d-4a34-99ac-a057f78cfa77",
                "name": "Ameya Prabhu",
                "collaborators": [
                    "Philip Torr",
                    "Ponnurangam Kumaraguru",
                    "Matthias Bethge",
                    "Anoop Namboodiri",
                    "Vishaal Udandarao",
                    "Shyamgopal Karthik",
                    "Vineet Gandhi",
                    "Manish Shrivastava",
                    "Shashwat Goel",
                    "Amartya Sanyal"
                ],
                "domain": [
                    "Active Learning",
                    "Deep Learning",
                    "Continual Learning",
                    "Multimodal Models"
                ],
                "publications": [
                    {
                        "title": "Sampling Bias in Deep Active Classification: An Empirical Study",
                        "abstract": "The exploding cost and time needed for data labeling and model training are bottlenecks for training DNN models on large datasets. Identifying smaller representative data samples with strategies like active learning can help mitigate such bottlenecks. Previous works on active learning in NLP identify the problem of sampling bias in the samples acquired by uncertainty-based querying and develop costly approaches to address it. Using a large empirical study, we demonstrate that active set selection using the posterior entropy of deep models like FastText.zip (FTZ) is robust to sampling biases and to various algorithmic choices (query size and strategies) unlike that suggested by traditional literature. We also show that FTZ based query strategy produces sample sets similar to those from more sophisticated approaches (e.g ensemble networks). Finally, we show the effectiveness of the selected samples by creating tiny high-quality datasets, and utilizing them for fast and cheap training of large models. Based on the above, we propose a simple baseline for deep active text classification that outperforms the state-of-the-art. We expect the presented work to be useful and informative for dataset compression and for problems involving active, semi-supervised or online learning scenarios. Code and models are available at: https://github.com/drimpossible/Sampling-Bias-Active-Learning"
                    },
                    {
                        "title": "Simple Unsupervised Multi-Object Tracking",
                        "abstract": "Multi-object tracking has seen a lot of progress recently, albeit with substantial annotation costs for developing better and larger labeled datasets. In this work, we remove the need for annotated datasets by proposing an unsupervised re-identification network, thus sidestepping the labeling costs entirely, required for training. Given unlabeled videos, our proposed method (SimpleReID) first generates tracking labels using SORT and trains a ReID network to predict the generated labels using crossentropy loss. We demonstrate that SimpleReID performs substantially better than simpler alternatives, and we recover the full performance of its supervised counterpart consistently across diverse tracking frameworks. The observations are unusual because unsupervised ReID is not expected to excel in crowded scenarios with occlusions, and drastic viewpoint changes. By incorporating our unsupervised SimpleReID with CenterTrack trained on augmented still images, we establish a new state-of-the-art performance on popular datasets like MOT16/17 without using tracking supervision, beating current best (CenterTrack) by 0.2-0.3 MOTA and 4.4-4.8 IDF1 scores. We further provide evidence for limited scope for improvement in IDF1 scores beyond our unsupervised ReID in the studied settings. Our investigation suggests reconsideration towards more sophisticated, supervised, end-to-end trackers by showing promise in simpler unsupervised alternatives."
                    },
                    {
                        "title": "Deep Expander Networks: Efficient Deep Networks from Graph Theory",
                        "abstract": "Efficient CNN designs like ResNets and DenseNet were proposed to improve accuracy vs efficiency trade-offs. They essentially increased the connectivity, allowing efficient information flow across layers. Inspired by these techniques, we propose to model connections between filters of a CNN using graphs which are simultaneously sparse and well connected. Sparsity results in efficiency while well connectedness can preserve the expressive power of the CNNs. We use a well-studied class of graphs from theoretical computer science that satisfies these properties known as Expander graphs. Expander graphs are used to model connections between filters in CNNs to design networks called X-Nets. We present two guarantees on the connectivity of X-Nets: Each node influences every node in a layer in logarithmic steps, and the number of paths between two sets of nodes is proportional to the product of their sizes. We also propose efficient training and inference algorithms, making it possible to train deeper and wider X-Nets effectively.   Expander based models give a 4% improvement in accuracy on MobileNet over grouped convolutions, a popular technique, which has the same sparsity but worse connectivity. X-Nets give better performance trade-offs than the original ResNet and DenseNet-BC architectures. We achieve model sizes comparable to state-of-the-art pruning techniques using our simple architecture design, without any pruning. We hope that this work motivates other approaches to utilize results from graph theory to develop efficient network architectures."
                    },
                    {
                        "title": "Towards Sub-Word Level Compositions for Sentiment Analysis of Hindi-English Code Mixed Text",
                        "abstract": "Sentiment analysis (SA) using code-mixed data from social media has several applications in opinion mining ranging from customer satisfaction to social campaign analysis in multilingual societies. Advances in this area are impeded by the lack of a suitable annotated dataset. We introduce a Hindi-English (Hi-En) code-mixed dataset for sentiment analysis and perform empirical analysis comparing the suitability and performance of various state-of-the-art SA methods in social media.   In this paper, we introduce learning sub-word level representations in LSTM (Subword-LSTM) architecture instead of character-level or word-level representations. This linguistic prior in our architecture enables us to learn the information about sentiment value of important morphemes. This also seems to work well in highly noisy text containing misspellings as shown in our experiments which is demonstrated in morpheme-level feature maps learned by our model. Also, we hypothesize that encoding this linguistic prior in the Subword-LSTM architecture leads to the superior performance. Our system attains accuracy 4-5% greater than traditional approaches on our dataset, and also outperforms the available system for sentiment analysis in Hi-En code-mixed text by 18%."
                    },
                    {
                        "title": "Corrective Machine Unlearning",
                        "abstract": "Machine Learning models increasingly face data integrity challenges due to the use of large-scale training datasets drawn from the Internet. We study what model developers can do if they detect that some data was manipulated or incorrect. Such manipulated data can cause adverse effects including vulnerability to backdoored samples, systemic biases, and reduced accuracy on certain input domains. Realistically, all manipulated training samples cannot be identified, and only a small, representative subset of the affected data can be flagged.   We formalize Corrective Machine Unlearning as the problem of mitigating the impact of data affected by unknown manipulations on a trained model, only having identified a subset of the corrupted data. We demonstrate that the problem of corrective unlearning has significantly different requirements from traditional privacy-oriented unlearning. We find most existing unlearning methods, including retraining-from-scratch without the deletion set, require most of the manipulated data to be identified for effective corrective unlearning. However, one approach, Selective Synaptic Dampening, achieves limited success, unlearning adverse effects with just a small portion of the manipulated samples in our setting, which shows encouraging signs for future progress. We hope our work spurs research towards developing better methods for corrective unlearning and offers practitioners a new strategy to handle data integrity challenges arising from web-scale training. Code is available at https://github.com/drimpossible/corrective-unlearning-bench."
                    },
                    {
                        "title": "Wu's Method can Boost Symbolic AI to Rival Silver Medalists and AlphaGeometry to Outperform Gold Medalists at IMO Geometry",
                        "abstract": "Proving geometric theorems constitutes a hallmark of visual reasoning combining both intuitive and logical skills. Therefore, automated theorem proving of Olympiad-level geometry problems is considered a notable milestone in human-level automated reasoning. The introduction of AlphaGeometry, a neuro-symbolic model trained with 100 million synthetic samples, marked a major breakthrough. It solved 25 of 30 International Mathematical Olympiad (IMO) problems whereas the reported baseline based on Wu's method solved only ten. In this note, we revisit the IMO-AG-30 Challenge introduced with AlphaGeometry, and find that Wu's method is surprisingly strong. Wu's method alone can solve 15 problems, and some of them are not solved by any of the other methods. This leads to two key findings: (i) Combining Wu's method with the classic synthetic methods of deductive databases and angle, ratio, and distance chasing solves 21 out of 30 methods by just using a CPU-only laptop with a time limit of 5 minutes per problem. Essentially, this classic method solves just 4 problems less than AlphaGeometry and establishes the first fully symbolic baseline strong enough to rival the performance of an IMO silver medalist. (ii) Wu's method even solves 2 of the 5 problems that AlphaGeometry failed to solve. Thus, by combining AlphaGeometry with Wu's method we set a new state-of-the-art for automated theorem proving on IMO-AG-30, solving 27 out of 30 problems, the first AI method which outperforms an IMO gold medalist."
                    },
                    {
                        "title": "\"You might also like this model\": Data Driven Approach for Recommending Deep Learning Models for Unknown Image Datasets",
                        "abstract": "For an unknown (new) classification dataset, choosing an appropriate deep learning architecture is often a recursive, time-taking, and laborious process. In this research, we propose a novel technique to recommend a suitable architecture from a repository of known models. Further, we predict the performance accuracy of the recommended architecture on the given unknown dataset, without the need for training the model. We propose a model encoder approach to learn a fixed length representation of deep learning architectures along with its hyperparameters, in an unsupervised fashion. We manually curate a repository of image datasets with corresponding known deep learning models and show that the predicted accuracy is a good estimator of the actual accuracy. We discuss the implications of the proposed approach for three benchmark images datasets and also the challenges in using the approach for text modality. To further increase the reproducibility of the proposed approach, the entire implementation is made publicly available along with the trained models."
                    },
                    {
                        "title": "No Cost Likelihood Manipulation at Test Time for Making Better Mistakes in Deep Networks",
                        "abstract": "There has been increasing interest in building deep hierarchy-aware classifiers that aim to quantify and reduce the severity of mistakes, and not just reduce the number of errors. The idea is to exploit the label hierarchy (e.g., the WordNet ontology) and consider graph distances as a proxy for mistake severity. Surprisingly, on examining mistake-severity distributions of the top-1 prediction, we find that current state-of-the-art hierarchy-aware deep classifiers do not always show practical improvement over the standard cross-entropy baseline in making better mistakes. The reason for the reduction in average mistake-severity can be attributed to the increase in low-severity mistakes, which may also explain the noticeable drop in their accuracy. To this end, we use the classical Conditional Risk Minimization (CRM) framework for hierarchy-aware classification. Given a cost matrix and a reliable estimate of likelihoods (obtained from a trained network), CRM simply amends mistakes at inference time; it needs no extra hyperparameters and requires adding just a few lines of code to the standard cross-entropy baseline. It significantly outperforms the state-of-the-art and consistently obtains large reductions in the average hierarchical distance of top-$k$ predictions across datasets, with very little loss in accuracy. CRM, because of its simplicity, can be used with any off-the-shelf trained model that provides reliable likelihood estimates."
                    },
                    {
                        "title": "Online Continual Learning Without the Storage Constraint",
                        "abstract": "Traditional online continual learning (OCL) research has primarily focused on mitigating catastrophic forgetting with fixed and limited storage allocation throughout an agent's lifetime. However, a broad range of real-world applications are primarily constrained by computational costs rather than storage limitations. In this paper, we target such applications, investigating the online continual learning problem under relaxed storage constraints and limited computational budgets. We contribute a simple algorithm, which updates a kNN classifier continually along with a fixed, pretrained feature extractor. We selected this algorithm due to its exceptional suitability for online continual learning. It can adapt to rapidly changing streams, has zero stability gap, operates within tiny computational budgets, has low storage requirements by only storing features, and has a consistency property: It never forgets previously seen data. These attributes yield significant improvements, allowing our proposed algorithm to outperform existing methods by over 20% in accuracy on two large-scale OCL datasets: Continual LOCalization (CLOC) with 39M images and 712 classes and Continual Google Landmarks V2 (CGLM) with 580K images and 10,788 classes, even when existing methods retain all previously seen images. Furthermore, we achieve this superior performance with considerably reduced computational and storage expenses. We provide code to reproduce our results at github.com/drimpossible/ACM."
                    },
                    {
                        "title": "Towards Deep Learning in Hindi NER: An approach to tackle the Labelled Data Scarcity",
                        "abstract": "In this paper we describe an end to end Neural Model for Named Entity Recognition NER) which is based on Bi-Directional RNN-LSTM. Almost all NER systems for Hindi use Language Specific features and handcrafted rules with gazetteers. Our model is language independent and uses no domain specific features or any handcrafted rules. Our models rely on semantic information in the form of word vectors which are learnt by an unsupervised learning algorithm on an unannotated corpus. Our model attained state of the art performance in both English and Hindi without the use of any morphological analysis or without using gazetteers of any sort."
                    },
                    {
                        "title": "Distribution-Aware Binarization of Neural Networks for Sketch Recognition",
                        "abstract": "Deep neural networks are highly effective at a range of computational tasks. However, they tend to be computationally expensive, especially in vision-related problems, and also have large memory requirements. One of the most effective methods to achieve significant improvements in computational/spatial efficiency is to binarize the weights and activations in a network. However, naive binarization results in accuracy drops when applied to networks for most tasks. In this work, we present a highly generalized, distribution-aware approach to binarizing deep networks that allows us to retain the advantages of a binarized network, while reducing accuracy drops. We also develop efficient implementations for our proposed approach across different architectures. We present a theoretical analysis of the technique to show the effective representational power of the resulting layers, and explore the forms of data they model best. Experiments on popular datasets show that our technique offers better accuracies than naive binarization, while retaining the same benefits that binarization provides - with respect to run-time compression, reduction of computational costs, and power consumption."
                    },
                    {
                        "title": "Towards Adversarial Evaluations for Inexact Machine Unlearning",
                        "abstract": "Machine Learning models face increased concerns regarding the storage of personal user data and adverse impacts of corrupted data like backdoors or systematic bias. Machine Unlearning can address these by allowing post-hoc deletion of affected training data from a learned model. Achieving this task exactly is computationally expensive; consequently, recent works have proposed inexact unlearning algorithms to solve this approximately as well as evaluation methods to test the effectiveness of these algorithms.   In this work, we first outline some necessary criteria for evaluation methods and show no existing evaluation satisfies them all. Then, we design a stronger black-box evaluation method called the Interclass Confusion (IC) test which adversarially manipulates data during training to detect the insufficiency of unlearning procedures. We also propose two analytically motivated baseline methods~(EU-k and CF-k) which outperform several popular inexact unlearning methods. Overall, we demonstrate how adversarial evaluation strategies can help in analyzing various unlearning phenomena which can guide the development of stronger unlearning algorithms."
                    },
                    {
                        "title": "Lifelong Benchmarks: Efficient Model Evaluation in an Era of Rapid Progress",
                        "abstract": "Standardized benchmarks drive progress in machine learning. However, with repeated testing, the risk of overfitting grows as algorithms over-exploit benchmark idiosyncrasies. In our work, we seek to mitigate this challenge by compiling ever-expanding large-scale benchmarks called Lifelong Benchmarks. As exemplars of our approach, we create Lifelong-CIFAR10 and Lifelong-ImageNet, containing (for now) 1.69M and 1.98M test samples, respectively. While reducing overfitting, lifelong benchmarks introduce a key challenge: the high cost of evaluating a growing number of models across an ever-expanding sample set. To address this challenge, we also introduce an efficient evaluation framework: Sort \\& Search (S&S), which reuses previously evaluated models by leveraging dynamic programming algorithms to selectively rank and sub-select test samples, enabling cost-effective lifelong benchmarking. Extensive empirical evaluations across 31,000 models demonstrate that S&S achieves highly-efficient approximate accuracy measurement, reducing compute cost from 180 GPU days to 5 GPU hours (1000x reduction) on a single A100 GPU, with low approximation error. As such, lifelong benchmarks offer a robust, practical solution to the \"benchmark exhaustion\" problem."
                    },
                    {
                        "title": "CiteME: Can Language Models Accurately Cite Scientific Claims?",
                        "abstract": "Thousands of new scientific papers are published each month. Such information overload complicates researcher efforts to stay current with the state-of-the-art as well as to verify and correctly attribute claims. We pose the following research question: Given a text excerpt referencing a paper, could an LM act as a research assistant to correctly identify the referenced paper? We advance efforts to answer this question by building a benchmark that evaluates the abilities of LMs in citation attribution. Our benchmark, CiteME, consists of text excerpts from recent machine learning papers, each referencing a single other paper. CiteME use reveals a large gap between frontier LMs and human performance, with LMs achieving only 4.2-18.5% accuracy and humans 69.7%. We close this gap by introducing CiteAgent, an autonomous system built on the GPT-4o LM that can also search and read papers, which achieves an accuracy of 35.3\\% on CiteME. Overall, CiteME serves as a challenging testbed for open-ended claim attribution, driving the research community towards a future where any claim made by an LM can be automatically verified and discarded if found to be incorrect."
                    },
                    {
                        "title": "Hybrid Binary Networks: Optimizing for Accuracy, Efficiency and Memory",
                        "abstract": "Binarization is an extreme network compression approach that provides large computational speedups along with energy and memory savings, albeit at significant accuracy costs. We investigate the question of where to binarize inputs at layer-level granularity and show that selectively binarizing the inputs to specific layers in the network could lead to significant improvements in accuracy while preserving most of the advantages of binarization. We analyze the binarization tradeoff using a metric that jointly models the input binarization-error and computational cost and introduce an efficient algorithm to select layers whose inputs are to be binarized. Practical guidelines based on insights obtained from applying the algorithm to a variety of models are discussed. Experiments on Imagenet dataset using AlexNet and ResNet-18 models show 3-4% improvements in accuracy over fully binarized networks with minimal impact on compression and computational speed. The improvements are even more substantial on sketch datasets like TU-Berlin, where we match state-of-the-art accuracy as well, getting over 8% increase in accuracies. We further show that our approach can be applied in tandem with other forms of compression that deal with individual layers or overall model compression (e.g., SqueezeNets). Unlike previous quantization approaches, we are able to binarize the weights in the last layers of a network, which often have a large number of parameters, resulting in significant improvement in accuracy over fully binarized models."
                    },
                    {
                        "title": "Small-scale and large-scale dynamos in global convection simulations of solar-like stars",
                        "abstract": "It has been recently shown that a small-scale dynamo (SSD) instability could be possible in solar-like low magnetic Prandtl number Pm plasmas. It has been proposed that the presence of SSD can potentially have a significant impact on the dynamics of the large-scale dynamo (LSD) in the stellar convection zones. Studying these two dynamos, SSD and LSD, together in a global magnetoconvection model requires high-resolution simulations and large amounts of computational resources. Starting from a well-studied global convective dynamo model that produces cyclic magnetic fields, we systematically increased the resolution and lowered the diffusivities to enter the regime of Reynolds numbers that allow for the excitation of SSD on top of the LSD. We studied how the properties of convection, generated differential rotation profiles, and LSD solutions change with the presence of SSD. We performed convective dynamo simulations in a spherical wedge with the Pencil Code. The resolutions of the models were increased in 4 steps by a total factor of 16 to achieve maximal fluid and magnetic Reynolds numbers of over 500. We found that the differential rotation is strongly quenched by the presence of the LSD and SSD. Even though the small-scale magnetic field only mildly decreases increasing Re, the large-scale field strength decreases significantly. We do not find the SSD dynamo significantly quenching the convective flows as claimed recently by other authors; in contrast, the convective flows first grow and then saturate for increasing Re. Furthermore, the angular momentum transport is highly affected by the presence of small-scale magnetic fields, which are mostly generated by LSD. These fields not only change the Reynolds stresses, but also generate dynamically important Maxwell stresses. The LSD evolution in terms of its pattern and field distribution is rather independent of the increase in Rm."
                    },
                    {
                        "title": "RanDumb: A Simple Approach that Questions the Efficacy of Continual Representation Learning",
                        "abstract": "Continual learning has primarily focused on the issue of catastrophic forgetting and the associated stability-plasticity tradeoffs. However, little attention has been paid to the efficacy of continually learned representations, as representations are learned alongside classifiers throughout the learning process. Our primary contribution is empirically demonstrating that existing online continually trained deep networks produce inferior representations compared to a simple pre-defined random transforms. Our approach embeds raw pixels using a fixed random transform, approximating an RBF-Kernel initialized before any data is seen. We then train a simple linear classifier on top without storing any exemplars, processing one sample at a time in an online continual learning setting. This method, called RanDumb, significantly outperforms state-of-the-art continually learned representations across all standard online continual learning benchmarks. Our study reveals the significant limitations of representation learning, particularly in low-exemplar and online continual learning scenarios. Extending our investigation to popular exemplar-free scenarios with pretrained models, we find that training only a linear classifier on top of pretrained representations surpasses most continual fine-tuning and prompt-tuning strategies. Overall, our investigation challenges the prevailing assumptions about effective representation learning in online continual learning. Our code is available at://github.com/drimpossible/RanDumb."
                    },
                    {
                        "title": "kNN-CLIP: Retrieval Enables Training-Free Segmentation on Continually Expanding Large Vocabularies",
                        "abstract": "Continual segmentation has not yet tackled the challenge of improving open-vocabulary segmentation models with training data for accurate segmentation across large, continually expanding vocabularies. We discover that traditional continual training results in severe catastrophic forgetting, failing to outperform a zero-shot segmentation baseline. We introduce a novel training-free strategy, kNN-CLIP, which augments the model with a database of instance embeddings for semantic and panoptic segmentation that achieves zero forgetting. We demonstrate that kNN-CLIP can adapt to continually growing vocabularies without the need for retraining or large memory costs. kNN-CLIP enables open-vocabulary segmentation methods to expand their vocabularies on any domain with a single pass through the data, while only storing compact embeddings. This approach minimizes both compute and memory costs. kNN-CLIP achieves state-of-the-art performance across large-vocabulary semantic and panoptic segmentation datasets. We hope kNN-CLIP represents a significant step forward in enabling more efficient and adaptable continual segmentation, paving the way for advances in real-world large-vocabulary continual segmentation methods."
                    },
                    {
                        "title": "A Practitioner's Guide to Continual Multimodal Pretraining",
                        "abstract": "Multimodal foundation models serve numerous applications at the intersection of vision and language. Still, despite being pretrained on extensive data, they become outdated over time. To keep models updated, research into continual pretraining mainly explores scenarios with either (1) infrequent, indiscriminate updates on large-scale new data, or (2) frequent, sample-level updates. However, practical model deployment often operates in the gap between these two limit cases, as real-world applications often demand adaptation to specific subdomains, tasks or concepts -- spread over the entire, varying life cycle of a model. In this work, we complement current perspectives on continual pretraining through a research test bed as well as provide comprehensive guidance for effective continual model updates in such scenarios. We first introduce FoMo-in-Flux, a continual multimodal pretraining benchmark with realistic compute constraints and practical deployment requirements, constructed over 63 datasets with diverse visual and semantic coverage. Using FoMo-in-Flux, we explore the complex landscape of practical continual pretraining through multiple perspectives: (1) A data-centric investigation of data mixtures and stream orderings that emulate real-world deployment situations, (2) a method-centric investigation ranging from simple fine-tuning and traditional continual learning strategies to parameter-efficient updates and model merging, (3) meta learning rate schedules and mechanistic design choices, and (4) the influence of model and compute scaling. Together, our insights provide a practitioner's guide to continual multimodal pretraining for real-world deployment. Our benchmark and code is here: https://github.com/ExplainableML/fomo_in_flux."
                    }
                ]
            },
            "31f80a9f-a0b7-4450-a301-3a96155dae08": {
                "pk": "31f80a9f-a0b7-4450-a301-3a96155dae08",
                "name": "Adhiraj Ghosh",
                "collaborators": [
                    "Kuruparan Shanmugalingam",
                    "Wen-Yan Lin",
                    "Hassan Shahmohammadi",
                    "Hendrik P. A. Lensch"
                ],
                "domain": [
                    "Computer Vision",
                    "Natural Language Processing",
                    "Image Generation",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Relation Preserving Triplet Mining for Stabilising the Triplet Loss in Re-identification Systems",
                        "abstract": "Object appearances change dramatically with pose variations. This creates a challenge for embedding schemes that seek to map instances with the same object ID to locations that are as close as possible. This issue becomes significantly heightened in complex computer vision tasks such as re-identification(reID). In this paper, we suggest that these dramatic appearance changes are indications that an object ID is composed of multiple natural groups, and it is counterproductive to forcefully map instances from different groups to a common location. This leads us to introduce Relation Preserving Triplet Mining (RPTM), a feature-matching guided triplet mining scheme, that ensures that triplets will respect the natural subgroupings within an object ID. We use this triplet mining mechanism to establish a pose-aware, well-conditioned triplet loss by implicitly enforcing view consistency. This allows a single network to be trained with fixed parameters across datasets while providing state-of-the-art results. Code is available at https://github.com/adhirajghosh/RPTM_reid."
                    },
                    {
                        "title": "ViPE: Visualise Pretty-much Everything",
                        "abstract": "Figurative and non-literal expressions are profoundly integrated in human communication. Visualising such expressions allow us to convey our creative thoughts, and evoke nuanced emotions. Recent text-to-image models like Stable Diffusion, on the other hand, struggle to depict non-literal expressions. Recent works primarily deal with this issue by compiling humanly annotated datasets on a small scale, which not only demands specialised expertise but also proves highly inefficient. To address this issue, we introduce ViPE: Visualise Pretty-much Everything. ViPE offers a series of lightweight and robust language models that have been trained on a large-scale set of lyrics with noisy visual descriptions that represent their implicit meaning. The synthetic visual descriptions are generated by GPT3.5 relying on neither human annotations nor images. ViPE effectively expresses any arbitrary piece of text into a visualisable description, enabling meaningful and high-quality image generation. We provide compelling evidence that ViPE is more robust than GPT3.5 in synthesising visual elaborations. ViPE also exhibits an understanding of figurative expressions comparable to human experts, providing a powerful and open-source backbone to many downstream applications such as music video and caption generation."
                    }
                ]
            },
            "ae3ac2a0-5931-4932-8cdb-ecbb7babfa22": {
                "pk": "ae3ac2a0-5931-4932-8cdb-ecbb7babfa22",
                "name": "Yash Sharma",
                "collaborators": [
                    "Donald E. Brown",
                    "Pin-Yu Chen",
                    "Basil Abraham",
                    "Preethi Jyothi",
                    "Sana Syed",
                    "Simra Shahid",
                    "Tanmay Singh",
                    "Kapil Sharma",
                    "Yi Zhu",
                    "Chris Russell"
                ],
                "domain": [
                    "Machine Learning",
                    "Adversarial Robustness",
                    "Speech Recognition",
                    "Domain Adaptation"
                ],
                "publications": [
                    {
                        "title": "Using Macroeconomic Forecasts to Improve Mean Reverting Trading Strategies",
                        "abstract": "A large class of trading strategies focus on opportunities offered by the yield curve. In particular, a set of yield curve trading strategies are based on the view that the yield curve mean-reverts. Based on these strategies' positive performance, a multiple pairs trading strategy on major currency pairs was implemented. To improve the algorithm's performance, machine learning forecasts of a series of pertinent macroeconomic variables were factored in, by optimizing the weights of the trading signals. This resulted in a clear improvement in the APR over the evaluation period, demonstrating that macroeconomic indicators, not only technical indicators, should be considered in trading strategies."
                    },
                    {
                        "title": "Devising Malware Characterstics using Transformers",
                        "abstract": "With the increasing number of cybersecurity threats, it becomes more difficult for researchers to skim through the security reports for malware analysis. There is a need to be able to extract highly relevant sentences without having to read through the entire malware reports. In this paper, we are finding relevant malware behavior mentions from Advanced Persistent Threat Reports. This main contribution is an opening attempt to Transformer the approach for malware behavior analysis."
                    },
                    {
                        "title": "Bypassing Feature Squeezing by Increasing Adversary Strength",
                        "abstract": "Feature Squeezing is a recently proposed defense method which reduces the search space available to an adversary by coalescing samples that correspond to many different feature vectors in the original space into a single sample. It has been shown that feature squeezing defenses can be combined in a joint detection framework to achieve high detection rates against state-of-the-art attacks. However, we demonstrate on the MNIST and CIFAR-10 datasets that by increasing the adversary strength of said state-of-the-art attacks, one can bypass the detection framework with adversarial examples of minimal visual distortion. These results suggest for proposed defenses to validate against stronger attack configurations."
                    },
                    {
                        "title": "Pixel-level Correspondence for Self-Supervised Learning from Video",
                        "abstract": "While self-supervised learning has enabled effective representation learning in the absence of labels, for vision, video remains a relatively untapped source of supervision. To address this, we propose Pixel-level Correspondence (PiCo), a method for dense contrastive learning from video. By tracking points with optical flow, we obtain a correspondence map which can be used to match local features at different points in time. We validate PiCo on standard benchmarks, outperforming self-supervised baselines on multiple dense prediction tasks, without compromising performance on image classification."
                    },
                    {
                        "title": "Attacking the Madry Defense Model with $L_1$-based Adversarial Examples",
                        "abstract": "The Madry Lab recently hosted a competition designed to test the robustness of their adversarially trained MNIST model. Attacks were constrained to perturb each pixel of the input image by a scaled maximal $L_\\infty$ distortion $\\epsilon$ = 0.3. This discourages the use of attacks which are not optimized on the $L_\\infty$ distortion metric. Our experimental results demonstrate that by relaxing the $L_\\infty$ constraint of the competition, the elastic-net attack to deep neural networks (EAD) can generate transferable adversarial examples which, despite their high average $L_\\infty$ distortion, have minimal visual distortion. These results call into question the use of $L_\\infty$ as a sole measure for visual distortion, and further demonstrate the power of EAD at generating robust adversarial examples."
                    },
                    {
                        "title": "Improving Low Resource Code-switched ASR using Augmented Code-switched TTS",
                        "abstract": "Building Automatic Speech Recognition (ASR) systems for code-switched speech has recently gained renewed attention due to the widespread use of speech technologies in multilingual communities worldwide. End-to-end ASR systems are a natural modeling choice due to their ease of use and superior performance in monolingual settings. However, it is well known that end-to-end systems require large amounts of labeled speech. In this work, we investigate improving code-switched ASR in low resource settings via data augmentation using code-switched text-to-speech (TTS) synthesis. We propose two targeted techniques to effectively leverage TTS speech samples: 1) Mixup, an existing technique to create new training samples via linear interpolation of existing samples, applied to TTS and real speech samples, and 2) a new loss function, used in conjunction with TTS samples, to encourage code-switched predictions. We report significant improvements in ASR performance achieving absolute word error rate (WER) reductions of up to 5%, and measurable improvement in code switching using our proposed techniques on a Hindi-English code-switched ASR task."
                    },
                    {
                        "title": "Encoding Cardiopulmonary Exercise Testing Time Series as Images for Classification using Convolutional Neural Network",
                        "abstract": "Exercise testing has been available for more than a half-century and is a remarkably versatile tool for diagnostic and prognostic information of patients for a range of diseases, especially cardiovascular and pulmonary. With rapid advancements in technology, wearables, and learning algorithm in the last decade, its scope has evolved. Specifically, Cardiopulmonary exercise testing (CPX) is one of the most commonly used laboratory tests for objective evaluation of exercise capacity and performance levels in patients. CPX provides a non-invasive, integrative assessment of the pulmonary, cardiovascular, and skeletal muscle systems involving the measurement of gas exchanges. However, its assessment is challenging, requiring the individual to process multiple time series data points, leading to simplification to peak values and slopes. But this simplification can discard the valuable trend information present in these time series. In this work, we encode the time series as images using the Gramian Angular Field and Markov Transition Field and use it with a convolutional neural network and attention pooling approach for the classification of heart failure and metabolic syndrome patients. Using GradCAMs, we highlight the discriminative features identified by the model."
                    },
                    {
                        "title": "MaNi: Maximizing Mutual Information for Nuclei Cross-Domain Unsupervised Segmentation",
                        "abstract": "In this work, we propose a mutual information (MI) based unsupervised domain adaptation (UDA) method for the cross-domain nuclei segmentation. Nuclei vary substantially in structure and appearances across different cancer types, leading to a drop in performance of deep learning models when trained on one cancer type and tested on another. This domain shift becomes even more critical as accurate segmentation and quantification of nuclei is an essential histopathology task for the diagnosis/ prognosis of patients and annotating nuclei at the pixel level for new cancer types demands extensive effort by medical experts. To address this problem, we maximize the MI between labeled source cancer type data and unlabeled target cancer type data for transferring nuclei segmentation knowledge across domains. We use the Jensen-Shanon divergence bound, requiring only one negative pair per positive pair for MI maximization. We evaluate our set-up for multiple modeling frameworks and on different datasets comprising of over 20 cancer-type domain shifts and demonstrate competitive performance. All the recently proposed approaches consist of multiple components for improving the domain adaptation, whereas our proposed module is light and can be easily incorporated into other methods (Implementation: https://github.com/YashSharma/MaNi )."
                    },
                    {
                        "title": "Gujarati-English Code-Switching Speech Recognition using ensemble prediction of spoken language",
                        "abstract": "An important and difficult task in code-switched speech recognition is to recognize the language, as lots of words in two languages can sound similar, especially in some accents. We focus on improving performance of end-to-end Automatic Speech Recognition models by conditioning transformer layers on language ID of words and character in the output in an per layer supervised manner. To this end, we propose two methods of introducing language specific parameters and explainability in the multi-head attention mechanism, and implement a Temporal Loss that helps maintain continuity in input alignment. Despite being unable to reduce WER significantly, our method shows promise in predicting the correct language from just spoken data. We introduce regularization in the language prediction by dropping LID in the sequence, which helps align long repeated output sequences."
                    },
                    {
                        "title": "CAAD 2018: Generating Transferable Adversarial Examples",
                        "abstract": "Deep neural networks (DNNs) are vulnerable to adversarial examples, perturbations carefully crafted to fool the targeted DNN, in both the non-targeted and targeted case. In the non-targeted case, the attacker simply aims to induce misclassification. In the targeted case, the attacker aims to induce classification to a specified target class. In addition, it has been observed that strong adversarial examples can transfer to unknown models, yielding a serious security concern. The NIPS 2017 competition was organized to accelerate research in adversarial attacks and defenses, taking place in the realistic setting where submitted adversarial attacks attempt to transfer to submitted defenses. The CAAD 2018 competition took place with nearly identical rules to the NIPS 2017 one. Given the requirement that the NIPS 2017 submissions were to be open-sourced, participants in the CAAD 2018 competition were able to directly build upon previous solutions, and thus improve the state-of-the-art in this setting. Our team participated in the CAAD 2018 competition, and won 1st place in both attack subtracks, non-targeted and targeted adversarial attacks, and 3rd place in defense. We outline our solutions and development results in this article. We hope our results can inform researchers in both generating and defending against adversarial examples."
                    },
                    {
                        "title": "On the Effectiveness of Low Frequency Perturbations",
                        "abstract": "Carefully crafted, often imperceptible, adversarial perturbations have been shown to cause state-of-the-art models to yield extremely inaccurate outputs, rendering them unsuitable for safety-critical application domains. In addition, recent work has shown that constraining the attack space to a low frequency regime is particularly effective. Yet, it remains unclear whether this is due to generally constraining the attack search space or specifically removing high frequency components from consideration. By systematically controlling the frequency components of the perturbation, evaluating against the top-placing defense submissions in the NeurIPS 2017 competition, we empirically show that performance improvements in both the white-box and black-box transfer settings are yielded only when low frequency components are preserved. In fact, the defended models based on adversarial training are roughly as vulnerable to low frequency perturbations as undefended models, suggesting that the purported robustness of state-of-the-art ImageNet defenses is reliant upon adversarial perturbations being high frequency in nature. We do find that under $\\ell_\\infty$ $\\epsilon=16/255$, the competition distortion bound, low frequency perturbations are indeed perceptible. This questions the use of the $\\ell_\\infty$-norm, in particular, as a distortion metric, and, in turn, suggests that explicitly considering the frequency space is promising for learning robust models which better align with human perception."
                    },
                    {
                        "title": "Are Generative Classifiers More Robust to Adversarial Attacks?",
                        "abstract": "There is a rising interest in studying the robustness of deep neural network classifiers against adversaries, with both advanced attack and defence techniques being actively developed. However, most recent work focuses on discriminative classifiers, which only model the conditional distribution of the labels given the inputs. In this paper, we propose and investigate the deep Bayes classifier, which improves classical naive Bayes with conditional deep generative models. We further develop detection methods for adversarial examples, which reject inputs with low likelihood under the generative model. Experimental results suggest that deep Bayes classifiers are more robust than deep discriminative classifiers, and that the proposed detection methods are effective against many recently proposed attacks."
                    },
                    {
                        "title": "Demo2Code: From Summarizing Demonstrations to Synthesizing Code via Extended Chain-of-Thought",
                        "abstract": "Language instructions and demonstrations are two natural ways for users to teach robots personalized tasks. Recent progress in Large Language Models (LLMs) has shown impressive performance in translating language instructions into code for robotic tasks. However, translating demonstrations into task code continues to be a challenge due to the length and complexity of both demonstrations and code, making learning a direct mapping intractable. This paper presents Demo2Code, a novel framework that generates robot task code from demonstrations via an extended chain-of-thought and defines a common latent specification to connect the two. Our framework employs a robust two-stage process: (1) a recursive summarization technique that condenses demonstrations into concise specifications, and (2) a code synthesis approach that expands each function recursively from the generated specifications. We conduct extensive evaluation on various robot task benchmarks, including a novel game benchmark Robotouille, designed to simulate diverse cooking tasks in a kitchen environment. The project's website is available at https://portal-cornell.github.io/demo2code/"
                    },
                    {
                        "title": "Generating $k$ EPR-pairs from an $n$-party resource state",
                        "abstract": "Motivated by quantum network applications over classical channels, we initiate the study of $n$-party resource states from which LOCC protocols can create EPR-pairs between any $k$ disjoint pairs of parties. We give constructions of such states where $k$ is not too far from the optimal $n/2$ while the individual parties need to hold only a constant number of qubits. In the special case when each party holds only one qubit, we describe a family of $n$-qubit states with $k$ proportional to $\\log n$ based on Reed-Muller codes, as well as small numerically found examples for $k=2$ and $k=3$. We also prove some lower bounds, for example showing that if $k=n/2$ then the parties must have at least $\\Omega(\\log\\log n)$ qubits each."
                    },
                    {
                        "title": "HistoTransfer: Understanding Transfer Learning for Histopathology",
                        "abstract": "Advancement in digital pathology and artificial intelligence has enabled deep learning-based computer vision techniques for automated disease diagnosis and prognosis. However, WSIs present unique computational and algorithmic challenges. WSIs are gigapixel-sized, making them infeasible to be used directly for training deep neural networks. Hence, for modeling, a two-stage approach is adopted: Patch representations are extracted first, followed by the aggregation for WSI prediction. These approaches require detailed pixel-level annotations for training the patch encoder. However, obtaining these annotations is time-consuming and tedious for medical experts. Transfer learning is used to address this gap and deep learning architectures pre-trained on ImageNet are used for generating patch-level representation. Even though ImageNet differs significantly from histopathology data, pre-trained networks have been shown to perform impressively on histopathology data. Also, progress in self-supervised and multi-task learning coupled with the release of multiple histopathology data has led to the release of histopathology-specific networks. In this work, we compare the performance of features extracted from networks trained on ImageNet and histopathology data. We use an attention pooling network over these extracted features for slide-level aggregation. We investigate if features learned using more complex networks lead to gain in performance. We use a simple top-k sampling approach for fine-tuning framework and study the representation similarity between frozen and fine-tuned networks using Centered Kernel Alignment. Further, to examine if intermediate block representation is better suited for feature extraction and ImageNet architectures are unnecessarily large for histopathology, we truncate the blocks of ResNet18 and DenseNet121 and examine the performance."
                    }
                ]
            },
            "b5eec874-3af7-4811-9020-0a8a70b0a991": {
                "pk": "b5eec874-3af7-4811-9020-0a8a70b0a991",
                "name": "Philip H. S. Torr",
                "collaborators": [
                    "Anurag Arnab",
                    "Puneet K. Dokania",
                    "Qizhu Li",
                    "Yuge Shi",
                    "N. Siddharth",
                    "Brooks Paige",
                    "Aleksandar Petrov",
                    "Adel Bibi",
                    "Ziming Zhang",
                    "Bernardino Romera-Paredes"
                ],
                "domain": [
                    "Computer Vision",
                    "Instance Segmentation",
                    "Object Detection",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Object Proposal Generation using Two-Stage Cascade SVMs",
                        "abstract": "Object proposal algorithms have shown great promise as a first step for object recognition and detection. Good object proposal generation algorithms require high object recall rate as well as low computational cost, because generating object proposals is usually utilized as a preprocessing step. The problem of how to accelerate the object proposal generation and evaluation process without decreasing recall is thus of great interest. In this paper, we propose a new object proposal generation method using two-stage cascade SVMs, where in the first stage linear filters are learned for predefined quantized scales/aspect-ratios independently, and in the second stage a global linear classifier is learned across all the quantized scales/aspect-ratios for calibration, so that all the proposals can be compared properly. The proposals with highest scores are our final output. Specifically, we explain our scale/aspect-ratio quantization scheme, and investigate the effects of combinations of $\\ell_1$ and $\\ell_2$ regularizers in cascade SVMs with/without ranking constraints in learning. Comprehensive experiments on VOC2007 dataset are conducted, and our results achieve the state-of-the-art performance with high object recall rate and high computational efficiency. Besides, our method has been demonstrated to be suitable for not only class-specific but also generic object proposal generation."
                    },
                    {
                        "title": "Recurrent Instance Segmentation",
                        "abstract": "Instance segmentation is the problem of detecting and delineating each distinct object of interest appearing in an image. Current instance segmentation approaches consist of ensembles of modules that are trained independently of each other, thus missing opportunities for joint learning. Here we propose a new instance segmentation paradigm consisting in an end-to-end method that learns how to segment instances sequentially. The model is based on a recurrent neural network that sequentially finds objects and their segmentations one at a time. This net is provided with a spatial memory that keeps track of what pixels have been explained and allows occlusion handling. In order to train the model we designed a principled loss function that accurately represents the properties of the instance segmentation problem. In the experiments carried out, we found that our method outperforms recent approaches on multiple person segmentation, and all state of the art approaches on the Plant Phenotyping dataset for leaf counting."
                    },
                    {
                        "title": "Bottom-up Instance Segmentation using Deep Higher-Order CRFs",
                        "abstract": "Traditional Scene Understanding problems such as Object Detection and Semantic Segmentation have made breakthroughs in recent years due to the adoption of deep learning. However, the former task is not able to localise objects at a pixel level, and the latter task has no notion of different instances of objects of the same class. We focus on the task of Instance Segmentation which recognises and localises objects down to a pixel level. Our model is based on a deep neural network trained for semantic segmentation. This network incorporates a Conditional Random Field with end-to-end trainable higher order potentials based on object detector outputs. This allows us to reason about instances from an initial, category-level semantic segmentation. Our simple method effectively leverages the great progress recently made in semantic segmentation and object detection. The accurate instance-level segmentations that our network produces is reflected by the considerable improvements obtained over previous work."
                    },
                    {
                        "title": "Traditional Classification Neural Networks are Good Generators: They are Competitive with DDPMs and GANs",
                        "abstract": "Classifiers and generators have long been separated. We break down this separation and showcase that conventional neural network classifiers can generate high-quality images of a large number of categories, being comparable to the state-of-the-art generative models (e.g., DDPMs and GANs). We achieve this by computing the partial derivative of the classification loss function with respect to the input to optimize the input to produce an image. Since it is widely known that directly optimizing the inputs is similar to targeted adversarial attacks incapable of generating human-meaningful images, we propose a mask-based stochastic reconstruction module to make the gradients semantic-aware to synthesize plausible images. We further propose a progressive-resolution technique to guarantee fidelity, which produces photorealistic images. Furthermore, we introduce a distance metric loss and a non-trivial distribution loss to ensure classification neural networks can synthesize diverse and high-fidelity images. Using traditional neural network classifiers, we can generate good-quality images of 256$\\times$256 resolution on ImageNet. Intriguingly, our method is also applicable to text-to-image generation by regarding image-text foundation models as generalized classifiers.   Proving that classifiers have learned the data distribution and are ready for image generation has far-reaching implications, for classifiers are much easier to train than generative models like DDPMs and GANs. We don't even need to train classification models because tons of public ones are available for download. Also, this holds great potential for the interpretability and robustness of classifiers. Project page is at \\url{https://classifier-as-generator.github.io/}."
                    },
                    {
                        "title": "Feature sampling and partitioning for visual vocabulary generation on large action classification datasets",
                        "abstract": "The recent trend in action recognition is towards larger datasets, an increasing number of action classes and larger visual vocabularies. State-of-the-art human action classification in challenging video data is currently based on a bag-of-visual-words pipeline in which space-time features are aggregated globally to form a histogram. The strategies chosen to sample features and construct a visual vocabulary are critical to performance, in fact often dominating performance. In this work we provide a critical evaluation of various approaches to building a vocabulary and show that good practises do have a significant impact. By subsampling and partitioning features strategically, we are able to achieve state-of-the-art results on 5 major action recognition datasets using relatively small visual vocabularies."
                    },
                    {
                        "title": "Discovering Class-Specific Pixels for Weakly-Supervised Semantic Segmentation",
                        "abstract": "We propose an approach to discover class-specific pixels for the weakly-supervised semantic segmentation task. We show that properly combining saliency and attention maps allows us to obtain reliable cues capable of significantly boosting the performance. First, we propose a simple yet powerful hierarchical approach to discover the class-agnostic salient regions, obtained using a salient object detector, which otherwise would be ignored. Second, we use fully convolutional attention maps to reliably localize the class-specific regions in a given image. We combine these two cues to discover class-specific pixels which are then used as an approximate ground truth for training a CNN. While solving the weakly supervised semantic segmentation task, we ensure that the image-level classification task is also solved in order to enforce the CNN to assign at least one pixel to each object present in the image. Experimentally, on the PASCAL VOC12 val and test sets, we obtain the mIoU of 60.8% and 61.9%, achieving the performance gains of 5.1% and 5.2% compared to the published state-of-the-art results. The code is made publicly available."
                    },
                    {
                        "title": "Weakly- and Semi-Supervised Panoptic Segmentation",
                        "abstract": "We present a weakly supervised model that jointly performs both semantic- and instance-segmentation -- a particularly relevant problem given the substantial cost of obtaining pixel-perfect annotation for these tasks. In contrast to many popular instance segmentation approaches based on object detectors, our method does not predict any overlapping instances. Moreover, we are able to segment both \"thing\" and \"stuff\" classes, and thus explain all the pixels in the image. \"Thing\" classes are weakly-supervised with bounding boxes, and \"stuff\" with image-level tags. We obtain state-of-the-art results on Pascal VOC, for both full and weak supervision (which achieves about 95% of fully-supervised performance). Furthermore, we present the first weakly-supervised results on Cityscapes for both semantic- and instance-segmentation. Finally, we use our weakly supervised framework to analyse the relationship between annotation quality and predictive performance, which is of interest to dataset creators."
                    },
                    {
                        "title": "SNIP: Single-shot Network Pruning based on Connection Sensitivity",
                        "abstract": "Pruning large neural networks while maintaining their performance is often desirable due to the reduced space and time complexity. In existing methods, pruning is done within an iterative optimization procedure with either heuristically designed pruning schedules or additional hyperparameters, undermining their utility. In this work, we present a new approach that prunes a given network once at initialization prior to training. To achieve this, we introduce a saliency criterion based on connection sensitivity that identifies structurally important connections in the network for the given task. This eliminates the need for both pretraining and the complex pruning schedule while making it robust to architecture variations. After pruning, the sparse network is trained in the standard way. Our method obtains extremely sparse networks with virtually the same accuracy as the reference network on the MNIST, CIFAR-10, and Tiny-ImageNet classification tasks and is broadly applicable to various architectures including convolutional, residual and recurrent networks. Unlike existing methods, our approach enables us to demonstrate that the retained connections are indeed relevant to the given task."
                    },
                    {
                        "title": "Hypergraph Convolution and Hypergraph Attention",
                        "abstract": "Recently, graph neural networks have attracted great attention and achieved prominent performance in various research fields. Most of those algorithms have assumed pairwise relationships of objects of interest. However, in many real applications, the relationships between objects are in higher-order, beyond a pairwise formulation. To efficiently learn deep embeddings on the high-order graph-structured data, we introduce two end-to-end trainable operators to the family of graph neural networks, i.e., hypergraph convolution and hypergraph attention. Whilst hypergraph convolution defines the basic formulation of performing convolution on a hypergraph, hypergraph attention further enhances the capacity of representation learning by leveraging an attention module. With the two operators, a graph neural network is readily extended to a more flexible model and applied to diverse applications where non-pairwise relationships are observed. Extensive experimental results with semi-supervised node classification demonstrate the effectiveness of hypergraph convolution and hypergraph attention."
                    },
                    {
                        "title": "Unifying Training and Inference for Panoptic Segmentation",
                        "abstract": "We present an end-to-end network to bridge the gap between training and inference pipeline for panoptic segmentation, a task that seeks to partition an image into semantic regions for \"stuff\" and object instances for \"things\". In contrast to recent works, our network exploits a parametrised, yet lightweight panoptic segmentation submodule, powered by an end-to-end learnt dense instance affinity, to capture the probability that any pair of pixels belong to the same instance. This panoptic submodule gives rise to a novel propagation mechanism for panoptic logits and enables the network to output a coherent panoptic segmentation map for both \"stuff\" and \"thing\" classes, without any post-processing. Reaping the benefits of end-to-end training, our full system sets new records on the popular street scene dataset, Cityscapes, achieving 61.4 PQ with a ResNet-50 backbone using only the fine annotations. On the challenging COCO dataset, our ResNet-50-based network also delivers state-of-the-art accuracy of 43.4 PQ. Moreover, our network flexibly works with and without object mask cues, performing competitively under both settings, which is of interest for applications with computation budgets."
                    },
                    {
                        "title": "Variational Mixture-of-Experts Autoencoders for Multi-Modal Deep Generative Models",
                        "abstract": "Learning generative models that span multiple data modalities, such as vision and language, is often motivated by the desire to learn more useful, generalisable representations that faithfully capture common underlying factors between the modalities. In this work, we characterise successful learning of such models as the fulfillment of four criteria: i) implicit latent decomposition into shared and private subspaces, ii) coherent joint generation over all modalities, iii) coherent cross-generation across individual modalities, and iv) improved model learning for individual modalities through multi-modal integration. Here, we propose a mixture-of-experts multimodal variational autoencoder (MMVAE) to learn generative models on different sets of modalities, including a challenging image-language dataset, and demonstrate its ability to satisfy all four criteria, both qualitatively and quantitatively."
                    },
                    {
                        "title": "Relating by Contrasting: A Data-efficient Framework for Multimodal Generative Models",
                        "abstract": "Multimodal learning for generative models often refers to the learning of abstract concepts from the commonality of information in multiple modalities, such as vision and language. While it has proven effective for learning generalisable representations, the training of such models often requires a large amount of \"related\" multimodal data that shares commonality, which can be expensive to come by. To mitigate this, we develop a novel contrastive framework for generative model learning, allowing us to train the model not just by the commonality between modalities, but by the distinction between \"related\" and \"unrelated\" multimodal data. We show in experiments that our method enables data-efficient multimodal learning on challenging datasets for various multimodal VAE models. We also show that under our proposed framework, the generative model can accurately identify related samples from unrelated ones, making it possible to make use of the plentiful unlabeled, unpaired multimodal data."
                    },
                    {
                        "title": "An Impartial Take to the CNN vs Transformer Robustness Contest",
                        "abstract": "Following the surge of popularity of Transformers in Computer Vision, several studies have attempted to determine whether they could be more robust to distribution shifts and provide better uncertainty estimates than Convolutional Neural Networks (CNNs). The almost unanimous conclusion is that they are, and it is often conjectured more or less explicitly that the reason of this supposed superiority is to be attributed to the self-attention mechanism. In this paper we perform extensive empirical analyses showing that recent state-of-the-art CNNs (particularly, ConvNeXt) can be as robust and reliable or even sometimes more than the current state-of-the-art Transformers. However, there is no clear winner. Therefore, although it is tempting to state the definitive superiority of one family of architectures over another, they seem to enjoy similar extraordinary performances on a variety of tasks while also suffering from similar vulnerabilities such as texture, background, and simplicity biases."
                    },
                    {
                        "title": "Memory-Driven Text-to-Image Generation",
                        "abstract": "We introduce a memory-driven semi-parametric approach to text-to-image generation, which is based on both parametric and non-parametric techniques. The non-parametric component is a memory bank of image features constructed from a training set of images. The parametric component is a generative adversarial network. Given a new text description at inference time, the memory bank is used to selectively retrieve image features that are provided as basic information of target images, which enables the generator to produce realistic synthetic results. We also incorporate the content information into the discriminator, together with semantic features, allowing the discriminator to make a more reliable prediction. Experimental results demonstrate that the proposed memory-driven semi-parametric approach produces more realistic images than purely parametric approaches, in terms of both visual fidelity and text-image semantic consistency."
                    },
                    {
                        "title": "Learn what matters: cross-domain imitation learning with task-relevant embeddings",
                        "abstract": "We study how an autonomous agent learns to perform a task from demonstrations in a different domain, such as a different environment or different agent. Such cross-domain imitation learning is required to, for example, train an artificial agent from demonstrations of a human expert. We propose a scalable framework that enables cross-domain imitation learning without access to additional demonstrations or further domain knowledge. We jointly train the learner agent's policy and learn a mapping between the learner and expert domains with adversarial training. We effect this by using a mutual information criterion to find an embedding of the expert's state space that contains task-relevant information and is invariant to domain specifics. This step significantly simplifies estimating the mapping between the learner and expert domains and hence facilitates end-to-end learning. We demonstrate successful transfer of policies between considerably different domains, without extra supervision such as additional demonstrations, and in situations where other methods fail."
                    },
                    {
                        "title": "When Do Prompting and Prefix-Tuning Work? A Theory of Capabilities and Limitations",
                        "abstract": "Context-based fine-tuning methods, including prompting, in-context learning, soft prompting (also known as prompt tuning), and prefix-tuning, have gained popularity due to their ability to often match the performance of full fine-tuning with a fraction of the parameters. Despite their empirical successes, there is little theoretical understanding of how these techniques influence the internal computation of the model and their expressiveness limitations. We show that despite the continuous embedding space being more expressive than the discrete token space, soft-prompting and prefix-tuning are potentially less expressive than full fine-tuning, even with the same number of learnable parameters. Concretely, context-based fine-tuning cannot change the relative attention pattern over the content and can only bias the outputs of an attention layer in a fixed direction. This suggests that while techniques like prompting, in-context learning, soft prompting, and prefix-tuning can effectively elicit skills present in the pretrained model, they may not be able to learn novel tasks that require new attention patterns."
                    },
                    {
                        "title": "Prompting a Pretrained Transformer Can Be a Universal Approximator",
                        "abstract": "Despite the widespread adoption of prompting, prompt tuning and prefix-tuning of transformer models, our theoretical understanding of these fine-tuning methods remains limited. A key question is whether one can arbitrarily modify the behavior of pretrained model by prompting or prefix-tuning it. Formally, whether prompting and prefix-tuning a pretrained model can universally approximate sequence-to-sequence functions. This paper answers in the affirmative and demonstrates that much smaller pretrained models than previously thought can be universal approximators when prefixed. In fact, the attention mechanism is uniquely suited for universal approximation with prefix-tuning a single attention head being sufficient to approximate any continuous function. Moreover, any sequence-to-sequence function can be approximated by prefixing a transformer with depth linear in the sequence length. Beyond these density-type results, we also offer Jackson-type bounds on the length of the prefix needed to approximate a function to a desired precision."
                    },
                    {
                        "title": "Fully-Trainable Deep Matching",
                        "abstract": "Deep Matching (DM) is a popular high-quality method for quasi-dense image matching. Despite its name, however, the original DM formulation does not yield a deep neural network that can be trained end-to-end via backpropagation. In this paper, we remove this limitation by rewriting the complete DM algorithm as a convolutional neural network. This results in a novel deep architecture for image matching that involves a number of new layer types and that, similar to recent networks for image segmentation, has a U-topology. We demonstrate the utility of the approach by improving the performance of DM by learning it end-to-end on an image matching task."
                    },
                    {
                        "title": "Learning to superoptimize programs - Workshop Version",
                        "abstract": "Superoptimization requires the estimation of the best program for a given computational task. In order to deal with large programs, superoptimization techniques perform a stochastic search. This involves proposing a modification of the current program, which is accepted or rejected based on the improvement achieved. The state of the art method uses uniform proposal distributions, which fails to exploit the problem structure to the fullest. To alleviate this deficiency, we learn a proposal distribution over possible modifications using Reinforcement Learning. We provide convincing results on the superoptimization of \"Hacker's Delight\" programs."
                    }
                ]
            },
            "600a0f7f-d569-49b3-9e21-d127c45f811a": {
                "pk": "600a0f7f-d569-49b3-9e21-d127c45f811a",
                "name": "Adel Bibi",
                "collaborators": [
                    "Bernard Ghanem",
                    "Philip H. S. Torr",
                    "Motasem Alfarra",
                    "Aleksandar Petrov",
                    "Peter Richt\u00e1rik",
                    "Modar Alfadly",
                    "El Houcine Bergou",
                    "Ozan Sener",
                    "Matthias M\u00fcller",
                    "Francisco Eiras"
                ],
                "domain": [
                    "Deep Learning",
                    "Adversarial Robustness",
                    "Continual Learning",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Network Moments: Extensions and Sparse-Smooth Attacks",
                        "abstract": "The impressive performance of deep neural networks (DNNs) has immensely strengthened the line of research that aims at theoretically analyzing their effectiveness. This has incited research on the reaction of DNNs to noisy input, namely developing adversarial input attacks and strategies that lead to robust DNNs to these attacks. To that end, in this paper, we derive exact analytic expressions for the first and second moments (mean and variance) of a small piecewise linear (PL) network (Affine, ReLU, Affine) subject to Gaussian input. In particular, we generalize the second-moment expression of Bibi et al. to arbitrary input Gaussian distributions, dropping the zero-mean assumption. We show that the new variance expression can be efficiently approximated leading to much tighter variance estimates as compared to the preliminary results of Bibi et al. Moreover, we experimentally show that these expressions are tight under simple linearizations of deeper PL-DNNs, where we investigate the effect of the linearization sensitivity on the accuracy of the moment estimates. Lastly, we show that the derived expressions can be used to construct sparse and smooth Gaussian adversarial attacks (targeted and non-targeted) that tend to lead to perceptually feasible input attacks."
                    },
                    {
                        "title": "Constrained Clustering: General Pairwise and Cardinality Constraints",
                        "abstract": "We study constrained clustering, where constraints guide the clustering process. In existing works, two categories of constraints have been widely explored, namely pairwise and cardinality constraints. Pairwise constraints enforce the cluster labels of two instances to be the same (must-link constraints) or different (cannot-link constraints). Cardinality constraints encourage cluster sizes to satisfy a user-specified distribution. Most existing constrained clustering models can only utilize one category of constraints at a time. We enforce the above two categories into a unified clustering model starting with the integer program formulation of the standard K-means. As the two categories provide different useful information, utilizing both allow for better clustering performance. However, the optimization is difficult due to the binary and quadratic constraints in the unified formulation. To solve this, we utilize two techniques: equivalently replacing the binary constraints by the intersection of two continuous constraints; the other is transforming the quadratic constraints into bi-linear constraints by introducing extra variables. We derive an equivalent continuous reformulation with simple constraints, which can be efficiently solved by Alternating Direction Method of Multipliers. Extensive experiments on both synthetic and real data demonstrate when: (1) utilizing a single category of constraint, the proposed model is superior to or competitive with SOTA constrained clustering models, and (2) utilizing both categories of constraints jointly, the proposed model shows better performance than the case of the single category. The experiments show that the proposed method exploits the constraints to achieve perfect clustering performance with improved clustering to 2%-5% in classical clustering metrics, e.g. Adjusted Random, Mirkin's, and Huber's, indices outerperfomring other methods."
                    },
                    {
                        "title": "Analytical Moment Regularizer for Gaussian Robust Networks",
                        "abstract": "Despite the impressive performance of deep neural networks (DNNs) on numerous vision tasks, they still exhibit yet-to-understand uncouth behaviours. One puzzling behaviour is the subtle sensitive reaction of DNNs to various noise attacks. Such a nuisance has strengthened the line of research around developing and training noise-robust networks. In this work, we propose a new training regularizer that aims to minimize the probabilistic expected training loss of a DNN subject to a generic Gaussian input. We provide an efficient and simple approach to approximate such a regularizer for arbitrary deep networks. This is done by leveraging the analytic expression of the output mean of a shallow neural network; avoiding the need for the memory and computationally expensive data augmentation. We conduct extensive experiments on LeNet and AlexNet on various datasets including MNIST, CIFAR10, and CIFAR100 demonstrating the effectiveness of our proposed regularizer. In particular, we show that networks that are trained with the proposed regularizer benefit from a boost in robustness equivalent to performing 3-21 folds of data augmentation."
                    },
                    {
                        "title": "A Stochastic Derivative-Free Optimization Method with Importance Sampling: Theory and Learning to Control",
                        "abstract": "We consider the problem of unconstrained minimization of a smooth objective function in $\\R^n$ in a setting where only function evaluations are possible. While importance sampling is one of the most popular techniques used by machine learning practitioners to accelerate the convergence of their models when applicable, there is not much existing theory for this acceleration in the derivative-free setting. In this paper, we propose the first derivative free optimization method with importance sampling and derive new improved complexity results on non-convex, convex and strongly convex functions. We conduct extensive experiments on various synthetic and real LIBSVM datasets confirming our theoretical results. We further test our method on a collection of continuous control tasks on MuJoCo environments with varying difficulty. Experiments suggest that our algorithm is practical for high dimensional continuous control problems where importance sampling results in a significant sample complexity improvement."
                    },
                    {
                        "title": "SimCS: Simulation for Domain Incremental Online Continual Segmentation",
                        "abstract": "Continual Learning is a step towards lifelong intelligence where models continuously learn from recently collected data without forgetting previous knowledge. Existing continual learning approaches mostly focus on image classification in the class-incremental setup with clear task boundaries and unlimited computational budget. This work explores the problem of Online Domain-Incremental Continual Segmentation (ODICS), where the model is continually trained over batches of densely labeled images from different domains, with limited computation and no information about the task boundaries. ODICS arises in many practical applications. In autonomous driving, this may correspond to the realistic scenario of training a segmentation model over time on a sequence of cities. We analyze several existing continual learning methods and show that they perform poorly in this setting despite working well in class-incremental segmentation. We propose SimCS, a parameter-free method complementary to existing ones that uses simulated data to regularize continual learning. Experiments show that SimCS provides consistent improvements when combined with different CL methods."
                    },
                    {
                        "title": "When Do Prompting and Prefix-Tuning Work? A Theory of Capabilities and Limitations",
                        "abstract": "Context-based fine-tuning methods, including prompting, in-context learning, soft prompting (also known as prompt tuning), and prefix-tuning, have gained popularity due to their ability to often match the performance of full fine-tuning with a fraction of the parameters. Despite their empirical successes, there is little theoretical understanding of how these techniques influence the internal computation of the model and their expressiveness limitations. We show that despite the continuous embedding space being more expressive than the discrete token space, soft-prompting and prefix-tuning are potentially less expressive than full fine-tuning, even with the same number of learnable parameters. Concretely, context-based fine-tuning cannot change the relative attention pattern over the content and can only bias the outputs of an attention layer in a fixed direction. This suggests that while techniques like prompting, in-context learning, soft prompting, and prefix-tuning can effectively elicit skills present in the pretrained model, they may not be able to learn novel tasks that require new attention patterns."
                    },
                    {
                        "title": "Prompting a Pretrained Transformer Can Be a Universal Approximator",
                        "abstract": "Despite the widespread adoption of prompting, prompt tuning and prefix-tuning of transformer models, our theoretical understanding of these fine-tuning methods remains limited. A key question is whether one can arbitrarily modify the behavior of pretrained model by prompting or prefix-tuning it. Formally, whether prompting and prefix-tuning a pretrained model can universally approximate sequence-to-sequence functions. This paper answers in the affirmative and demonstrates that much smaller pretrained models than previously thought can be universal approximators when prefixed. In fact, the attention mechanism is uniquely suited for universal approximation with prefix-tuning a single attention head being sufficient to approximate any continuous function. Moreover, any sequence-to-sequence function can be approximated by prefixing a transformer with depth linear in the sequence length. Beyond these density-type results, we also offer Jackson-type bounds on the length of the prefix needed to approximate a function to a desired precision."
                    },
                    {
                        "title": "Certifying Ensembles: A General Certification Theory with S-Lipschitzness",
                        "abstract": "Improving and guaranteeing the robustness of deep learning models has been a topic of intense research. Ensembling, which combines several classifiers to provide a better model, has shown to be beneficial for generalisation, uncertainty estimation, calibration, and mitigating the effects of concept drift. However, the impact of ensembling on certified robustness is less well understood. In this work, we generalise Lipschitz continuity by introducing S-Lipschitz classifiers, which we use to analyse the theoretical robustness of ensembles. Our results are precise conditions when ensembles of robust classifiers are more robust than any constituent classifier, as well as conditions when they are less robust."
                    },
                    {
                        "title": "On the Decision Boundaries of Neural Networks: A Tropical Geometry Perspective",
                        "abstract": "This work tackles the problem of characterizing and understanding the decision boundaries of neural networks with piecewise linear non-linearity activations. We use tropical geometry, a new development in the area of algebraic geometry, to characterize the decision boundaries of a simple network of the form (Affine, ReLU, Affine). Our main finding is that the decision boundaries are a subset of a tropical hypersurface, which is intimately related to a polytope formed by the convex hull of two zonotopes. The generators of these zonotopes are functions of the network parameters. This geometric characterization provides new perspectives to three tasks. (i) We propose a new tropical perspective to the lottery ticket hypothesis, where we view the effect of different initializations on the tropical geometric representation of a network's decision boundaries. (ii) Moreover, we propose new tropical based optimization reformulations that directly influence the decision boundaries of the network for the task of network pruning. (iii) At last, we discuss the reformulation of the generation of adversarial attacks in a tropical sense. We demonstrate that one can construct adversaries in a new tropical setting by perturbing a specific set of decision boundaries by perturbing a set of parameters in the network."
                    },
                    {
                        "title": "Improving SAGA via a Probabilistic Interpolation with Gradient Descent",
                        "abstract": "We develop and analyze a new algorithm for empirical risk minimization, which is the key paradigm for training supervised machine learning models. Our method---SAGD---is based on a probabilistic interpolation of SAGA and gradient descent (GD). In particular, in each iteration we take a gradient step with probability $q$ and a SAGA step with probability $1-q$. We show that, surprisingly, the total expected complexity of the method (which is obtained by multiplying the number of iterations by the expected number of gradients computed in each iteration) is minimized for a non-trivial probability $q$. For example, for a well conditioned problem the choice $q=1/(n-1)^2$, where $n$ is the number of data samples, gives a method with an overall complexity which is better than both the complexity of GD and SAGA. We further generalize the results to a probabilistic interpolation of SAGA and minibatch SAGA, which allows us to compute both the optimal probability and the optimal minibatch size. While the theoretical improvement may not be large, the practical improvement is robustly present across all synthetic and real data we tested for, and can be substantial. Our theoretical results suggest that for this optimal minibatch size our method achieves linear speedup in minibatch size, which is of key practical importance as minibatch implementations are used to train machine learning models in practice. Moreover, empirical evidence suggest that a linear speedup in minibatch size can be attained with a parallel implementation."
                    },
                    {
                        "title": "Data-Dependent Randomized Smoothing",
                        "abstract": "Randomized smoothing is a recent technique that achieves state-of-art performance in training certifiably robust deep neural networks. While the smoothing family of distributions is often connected to the choice of the norm used for certification, the parameters of these distributions are always set as global hyper parameters independent from the input data on which a network is certified. In this work, we revisit Gaussian randomized smoothing and show that the variance of the Gaussian distribution can be optimized at each input so as to maximize the certification radius for the construction of the smooth classifier. Since the data dependent classifier does not directly enjoy sound certification with existing approaches, we propose a memory-enhanced data dependent smooth classifier that is certifiable by construction. This new approach is generic, parameter-free, and easy to implement. In fact, we show that our data dependent framework can be seamlessly incorporated into 3 randomized smoothing approaches, leading to consistent improved certified accuracy. When this framework is used in the training routine of these approaches followed by a data dependent certification, we achieve 9% and 6% improvement over the certified accuracy of the strongest baseline for a radius of 0.5 on CIFAR10 and ImageNet."
                    },
                    {
                        "title": "TrackingNet: A Large-Scale Dataset and Benchmark for Object Tracking in the Wild",
                        "abstract": "Despite the numerous developments in object tracking, further development of current tracking algorithms is limited by small and mostly saturated datasets. As a matter of fact, data-hungry trackers based on deep-learning currently rely on object detection datasets due to the scarcity of dedicated large-scale tracking datasets. In this work, we present TrackingNet, the first large-scale dataset and benchmark for object tracking in the wild. We provide more than 30K videos with more than 14 million dense bounding box annotations. Our dataset covers a wide selection of object classes in broad and diverse context. By releasing such a large-scale dataset, we expect deep trackers to further improve and generalize. In addition, we introduce a new benchmark composed of 500 novel videos, modeled with a distribution similar to our training dataset. By sequestering the annotation of the test set and providing an online evaluation server, we provide a fair benchmark for future development of object trackers. Deep trackers fine-tuned on a fraction of our dataset improve their performance by up to 1.6% on OTB100 and up to 1.7% on TrackingNet Test. We provide an extensive benchmark on TrackingNet by evaluating more than 20 trackers. Our results suggest that object tracking in the wild is far from being solved."
                    },
                    {
                        "title": "Rethinking Clustering for Robustness",
                        "abstract": "This paper studies how encouraging semantically-aligned features during deep neural network training can increase network robustness. Recent works observed that Adversarial Training leads to robust models, whose learnt features appear to correlate with human perception. Inspired by this connection from robustness to semantics, we study the complementary connection: from semantics to robustness. To do so, we provide a robustness certificate for distance-based classification models (clustering-based classifiers). Moreover, we show that this certificate is tight, and we leverage it to propose ClusTR (Clustering Training for Robustness), a clustering-based and adversary-free training framework to learn robust models. Interestingly, \\textit{ClusTR} outperforms adversarially-trained networks by up to $4\\%$ under strong PGD attacks."
                    },
                    {
                        "title": "A Stochastic Derivative Free Optimization Method with Momentum",
                        "abstract": "We consider the problem of unconstrained minimization of a smooth objective function in $\\mathbb{R}^d$ in setting where only function evaluations are possible. We propose and analyze stochastic zeroth-order method with heavy ball momentum. In particular, we propose, SMTP, a momentum version of the stochastic three-point method (STP) \\cite{Bergou_2018}. We show new complexity results for non-convex, convex and strongly convex functions. We test our method on a collection of learning to continuous control tasks on several MuJoCo \\cite{Todorov_2012} environments with varying difficulty and compare against STP, other state-of-the-art derivative-free optimization algorithms and against policy gradient methods. SMTP significantly outperforms STP and all other methods that we considered in our numerical experiments. Our second contribution is SMTP with importance sampling which we call SMTP_IS. We provide convergence analysis of this method for non-convex, convex and strongly convex objectives."
                    },
                    {
                        "title": "DeformRS: Certifying Input Deformations with Randomized Smoothing",
                        "abstract": "Deep neural networks are vulnerable to input deformations in the form of vector fields of pixel displacements and to other parameterized geometric deformations e.g. translations, rotations, etc. Current input deformation certification methods either 1. do not scale to deep networks on large input datasets, or 2. can only certify a specific class of deformations, e.g. only rotations. We reformulate certification in randomized smoothing setting for both general vector field and parameterized deformations and propose DeformRS-VF and DeformRS-Par, respectively. Our new formulation scales to large networks on large input datasets. For instance, DeformRS-Par certifies rich deformations, covering translations, rotations, scaling, affine deformations, and other visually aligned deformations such as ones parameterized by Discrete-Cosine-Transform basis. Extensive experiments on MNIST, CIFAR10, and ImageNet show competitive performance of DeformRS-Par achieving a certified accuracy of $39\\%$ against perturbed rotations in the set $[-10\\degree,10\\degree]$ on ImageNet."
                    },
                    {
                        "title": "Diversified Dynamic Routing for Vision Tasks",
                        "abstract": "Deep learning models for vision tasks are trained on large datasets under the assumption that there exists a universal representation that can be used to make predictions for all samples. Whereas high complexity models are proven to be capable of learning such representations, a mixture of experts trained on specific subsets of the data can infer the labels more efficiently. However using mixture of experts poses two new problems, namely (i) assigning the correct expert at inference time when a new unseen sample is presented. (ii) Finding the optimal partitioning of the training data, such that the experts rely the least on common features. In Dynamic Routing (DR) a novel architecture is proposed where each layer is composed of a set of experts, however without addressing the two challenges we demonstrate that the model reverts to using the same subset of experts.   In our method, Diversified Dynamic Routing (DivDR) the model is explicitly trained to solve the challenge of finding relevant partitioning of the data and assigning the correct experts in an unsupervised approach. We conduct several experiments on semantic segmentation on Cityscapes and object detection and instance segmentation on MS-COCO showing improved performance over several baselines."
                    },
                    {
                        "title": "Language Model Tokenizers Introduce Unfairness Between Languages",
                        "abstract": "Recent language models have shown impressive multilingual performance, even when not explicitly trained for it. Despite this, there are concerns about the quality of their outputs across different languages. In this paper, we show how disparity in the treatment of different languages arises at the tokenization stage, well before a model is even invoked. The same text translated into different languages can have drastically different tokenization lengths, with differences up to 15 times in some cases. These disparities persist even for tokenizers that are intentionally trained for multilingual support. Character-level and byte-level models also exhibit over 4 times the difference in the encoding length for some language pairs. This induces unfair treatment for some language communities in regard to the cost of accessing commercial language services, the processing time and latency, as well as the amount of content that can be provided as context to the models. Therefore, we make the case that we should train future language models using multilingually fair subword tokenizers."
                    },
                    {
                        "title": "Lifelong Benchmarks: Efficient Model Evaluation in an Era of Rapid Progress",
                        "abstract": "Standardized benchmarks drive progress in machine learning. However, with repeated testing, the risk of overfitting grows as algorithms over-exploit benchmark idiosyncrasies. In our work, we seek to mitigate this challenge by compiling ever-expanding large-scale benchmarks called Lifelong Benchmarks. As exemplars of our approach, we create Lifelong-CIFAR10 and Lifelong-ImageNet, containing (for now) 1.69M and 1.98M test samples, respectively. While reducing overfitting, lifelong benchmarks introduce a key challenge: the high cost of evaluating a growing number of models across an ever-expanding sample set. To address this challenge, we also introduce an efficient evaluation framework: Sort \\& Search (S&S), which reuses previously evaluated models by leveraging dynamic programming algorithms to selectively rank and sub-select test samples, enabling cost-effective lifelong benchmarking. Extensive empirical evaluations across 31,000 models demonstrate that S&S achieves highly-efficient approximate accuracy measurement, reducing compute cost from 180 GPU days to 5 GPU hours (1000x reduction) on a single A100 GPU, with low approximation error. As such, lifelong benchmarks offer a robust, practical solution to the \"benchmark exhaustion\" problem."
                    },
                    {
                        "title": "Towards Certification of Uncertainty Calibration under Adversarial Attacks",
                        "abstract": "Since neural classifiers are known to be sensitive to adversarial perturbations that alter their accuracy, \\textit{certification methods} have been developed to provide provable guarantees on the insensitivity of their predictions to such perturbations. Furthermore, in safety-critical applications, the frequentist interpretation of the confidence of a classifier (also known as model calibration) can be of utmost importance. This property can be measured via the Brier score or the expected calibration error. We show that attacks can significantly harm calibration, and thus propose certified calibration as worst-case bounds on calibration under adversarial perturbations. Specifically, we produce analytic bounds for the Brier score and approximate bounds via the solution of a mixed-integer program on the expected calibration error. Finally, we propose novel calibration attacks and demonstrate how they can improve model calibration through \\textit{adversarial calibration training}."
                    },
                    {
                        "title": "Mimicking User Data: On Mitigating Fine-Tuning Risks in Closed Large Language Models",
                        "abstract": "Fine-tuning large language models on small, high-quality datasets can enhance their performance on specific downstream tasks. Recent research shows that fine-tuning on benign, instruction-following data can inadvertently undo the safety alignment process and increase a model's propensity to comply with harmful queries. Although critical, understanding and mitigating safety risks in well-defined tasks remains distinct from the instruction-following context due to structural differences in the data. Our work addresses the gap in our understanding of these risks across diverse types of data in closed models - where providers control how user data is utilized in the fine-tuning process. We demonstrate how malicious actors can subtly manipulate the structure of almost any task-specific dataset to foster significantly more dangerous model behaviors, while maintaining an appearance of innocuity and reasonable downstream task performance. To address this issue, we propose a novel mitigation strategy that mixes in safety data which mimics the task format and prompting style of the user data, showing this is more effective than existing baselines at re-establishing safety alignment while maintaining similar task performance."
                    }
                ]
            },
            "b5bd4740-c444-4ef8-9833-a475e0ab8aa1": {
                "pk": "b5bd4740-c444-4ef8-9833-a475e0ab8aa1",
                "name": "Samuel Albanie",
                "collaborators": [
                    "Andrea Vedaldi",
                    "Arsha Nagrani",
                    "Andrew Zisserman",
                    "James Thewlis",
                    "Jo\u00e3o F. Henriques",
                    "Joao F. Henriques",
                    "David Novotny",
                    "Diane Larlus",
                    "Sebastien Ehrhardt",
                    "Joao Henriques"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Self-Supervised Learning",
                    "Emotion Recognition"
                ],
                "publications": [
                    {
                        "title": "Learning Grimaces by Watching TV",
                        "abstract": "Differently from computer vision systems which require explicit supervision, humans can learn facial expressions by observing people in their environment. In this paper, we look at how similar capabilities could be developed in machine vision. As a starting point, we consider the problem of relating facial expressions to objectively measurable events occurring in videos. In particular, we consider a gameshow in which contestants play to win significant sums of money. We extract events affecting the game and corresponding facial expressions objectively and automatically from the videos, obtaining large quantities of labelled data for our study. We also develop, using benchmarks such as FER and SFEW 2.0, state-of-the-art deep neural networks for facial expression recognition, showing that pre-training on face verification data can be highly beneficial for this task. Then, we extend these models to use facial expressions to predict events in videos and learn nameable expressions from them. The dataset and emotion recognition models are available at http://www.robots.ox.ac.uk/~vgg/data/facevalue"
                    },
                    {
                        "title": "Emotion Recognition in Speech using Cross-Modal Transfer in the Wild",
                        "abstract": "Obtaining large, human labelled speech datasets to train models for emotion recognition is a notoriously challenging task, hindered by annotation cost and label ambiguity. In this work, we consider the task of learning embeddings for speech classification without access to any form of labelled audio. We base our approach on a simple hypothesis: that the emotional content of speech correlates with the facial expression of the speaker. By exploiting this relationship, we show that annotations of expression can be transferred from the visual domain (faces) to the speech domain (voices) through cross-modal distillation. We make the following contributions: (i) we develop a strong teacher network for facial emotion recognition that achieves the state of the art on a standard benchmark; (ii) we use the teacher to train a student, tabula rasa, to learn representations (embeddings) for speech emotion recognition without access to labelled audio data; and (iii) we show that the speech emotion embedding can be used for speech emotion recognition on external benchmark datasets. Code, models and data are available."
                    },
                    {
                        "title": "Deep Industrial Espionage",
                        "abstract": "The theory of deep learning is now considered largely solved, and is well understood by researchers and influencers alike. To maintain our relevance, we therefore seek to apply our skills to under-explored, lucrative applications of this technology. To this end, we propose and Deep Industrial Espionage, an efficient end-to-end framework for industrial information propagation and productisation. Specifically, given a single image of a product or service, we aim to reverse-engineer, rebrand and distribute a copycat of the product at a profitable price-point to consumers in an emerging market---all within in a single forward pass of a Neural Network. Differently from prior work in machine perception which has been restricted to classifying, detecting and reasoning about object instances, our method offers tangible business value in a wide range of corporate settings. Our approach draws heavily on a promising recent arxiv paper until its original authors' names can no longer be read (we use felt tip pen). We then rephrase the anonymised paper, add the word \"novel\" to the title, and submit it a prestigious, closed-access espionage journal who assure us that someday, we will be entitled to some fraction of their extortionate readership fees."
                    },
                    {
                        "title": "Semi-convolutional Operators for Instance Segmentation",
                        "abstract": "Object detection and instance segmentation are dominated by region-based methods such as Mask RCNN. However, there is a growing interest in reducing these problems to pixel labeling tasks, as the latter could be more efficient, could be integrated seamlessly in image-to-image network architectures as used in many other tasks, and could be more accurate for objects that are not well approximated by bounding boxes. In this paper we show theoretically and empirically that constructing dense pixel embeddings that can separate object instances cannot be easily achieved using convolutional operators. At the same time, we show that simple modifications, which we call semi-convolutional, have a much better chance of succeeding at this task. We use the latter to show a connection to Hough voting as well as to a variant of the bilateral kernel that is spatially steered by a convolutional network. We demonstrate that these operators can also be used to improve approaches such as Mask RCNN, demonstrating better segmentation of complex biological shapes and PASCAL VOC categories than achievable by Mask RCNN alone."
                    },
                    {
                        "title": "A 23 MW data centre is all you need",
                        "abstract": "The field of machine learning has achieved striking progress in recent years, witnessing breakthrough results on language modelling, protein folding and nitpickingly fine-grained dog breed classification. Some even succeeded at playing computer games and board games, a feat both of engineering and of setting their employers' expectations. The central contribution of this work is to carefully examine whether this progress, and technology more broadly, can be expected to continue indefinitely. Through a rigorous application of statistical theory and failure to extrapolate beyond the training data, we answer firmly in the negative and provide details: technology will peak at 3:07 am (BST) on 20th July, 2032. We then explore the implications of this finding, discovering that individuals awake at this ungodly hour with access to a sufficiently powerful computer possess an opportunity for myriad forms of long-term linguistic 'lock in'. All we need is a large (>> 1W) data centre to seize this pivotal moment. By setting our analogue alarm clocks, we propose a tractable algorithm to ensure that, for the future of humanity, the British spelling of colour becomes the default spelling across more than 80% of the global word processing software market."
                    },
                    {
                        "title": "On the Origin of Species of Self-Supervised Learning",
                        "abstract": "In the quiet backwaters of cs.CV, cs.LG and stat.ML, a cornucopia of new learning systems is emerging from a primordial soup of mathematics-learning systems with no need for external supervision. To date, little thought has been given to how these self-supervised learners have sprung into being or the principles that govern their continuing diversification. After a period of deliberate study and dispassionate judgement during which each author set their Zoom virtual background to a separate Galapagos island, we now entertain no doubt that each of these learning machines are lineal descendants of some older and generally extinct species. We make five contributions: (1) We gather and catalogue row-major arrays of machine learning specimens, each exhibiting heritable discriminative features; (2) We document a mutation mechanism by which almost imperceptible changes are introduced to the genotype of new systems, but their phenotype (birdsong in the form of tweets and vestigial plumage such as press releases) communicates dramatic changes; (3) We propose a unifying theory of self-supervised machine evolution and compare to other unifying theories on standard unifying theory benchmarks, where we establish a new (and unifying) state of the art; (4) We discuss the importance of digital biodiversity, in light of the endearingly optimistic Paris Agreement."
                    },
                    {
                        "title": "Large Language Models are Few-shot Publication Scoopers",
                        "abstract": "Driven by recent advances AI, we passengers are entering a golden age of scientific discovery. But golden for whom? Confronting our insecurity that others may beat us to the most acclaimed breakthroughs of the era, we propose a novel solution to the long-standing personal credit assignment problem to ensure that it is golden for us. At the heart of our approach is a pip-to-the-post algorithm that assures adulatory Wikipedia pages without incurring the substantial capital and career risks of pursuing high impact science with conventional research methodologies. By leveraging the meta trend of leveraging large language models for everything, we demonstrate the unparalleled potential of our algorithm to scoop groundbreaking findings with the insouciance of a seasoned researcher at a dessert buffet."
                    },
                    {
                        "title": "Simple Baselines for Interactive Video Retrieval with Questions and Answers",
                        "abstract": "To date, the majority of video retrieval systems have been optimized for a \"single-shot\" scenario in which the user submits a query in isolation, ignoring previous interactions with the system. Recently, there has been renewed interest in interactive systems to enhance retrieval, but existing approaches are complex and deliver limited gains in performance. In this work, we revisit this topic and propose several simple yet effective baselines for interactive video retrieval via question-answering. We employ a VideoQA model to simulate user interactions and show that this enables the productive study of the interactive retrieval task without access to ground truth dialogue data. Experiments on MSR-VTT, MSVD, and AVSD show that our framework using question-based interaction significantly improves the performance of text-based video retrieval systems."
                    },
                    {
                        "title": "Unknowable Manipulators: Social Network Curator Algorithms",
                        "abstract": "For a social networking service to acquire and retain users, it must find ways to keep them engaged. By accurately gauging their preferences, it is able to serve them with the subset of available content that maximises revenue for the site. Without the constraints of an appropriate regulatory framework, we argue that a sufficiently sophisticated curator algorithm tasked with performing this process may choose to explore curation strategies that are detrimental to users. In particular, we suggest that such an algorithm is capable of learning to manipulate its users, for several qualitative reasons: 1. Access to vast quantities of user data combined with ongoing breakthroughs in the field of machine learning are leading to powerful but uninterpretable strategies for decision making at scale. 2. The availability of an effective feedback mechanism for assessing the short and long term user responses to curation strategies. 3. Techniques from reinforcement learning have allowed machines to learn automated and highly successful strategies at an abstract level, often resulting in non-intuitive yet nonetheless highly appropriate action selection. In this work, we consider the form that these strategies for user manipulation might take and scrutinise the role that regulation should play in the design of such systems."
                    },
                    {
                        "title": "Stopping GAN Violence: Generative Unadversarial Networks",
                        "abstract": "While the costs of human violence have attracted a great deal of attention from the research community, the effects of the network-on-network (NoN) violence popularised by Generative Adversarial Networks have yet to be addressed. In this work, we quantify the financial, social, spiritual, cultural, grammatical and dermatological impact of this aggression and address the issue by proposing a more peaceful approach which we term Generative Unadversarial Networks (GUNs). Under this framework, we simultaneously train two models: a generator G that does its best to capture whichever data distribution it feels it can manage, and a motivator M that helps G to achieve its dream. Fighting is strictly verboten and both models evolve by learning to respect their differences. The framework is both theoretically and electrically grounded in game theory, and can be viewed as a winner-shares-all two-player game in which both players work as a team to achieve the best score. Experiments show that by working in harmony, the proposed model is able to claim both the moral and log-likelihood high ground. Our work builds on a rich history of carefully argued position-papers, published as anonymous YouTube comments, which prove that the optimal solution to NoN violence is more GUNs."
                    },
                    {
                        "title": "Substitute Teacher Networks: Learning with Almost No Supervision",
                        "abstract": "Learning through experience is time-consuming, inefficient and often bad for your cortisol levels. To address this problem, a number of recently proposed teacher-student methods have demonstrated the benefits of private tuition, in which a single model learns from an ensemble of more experienced tutors. Unfortunately, the cost of such supervision restricts good representations to a privileged minority. Unsupervised learning can be used to lower tuition fees, but runs the risk of producing networks that require extracurriculum learning to strengthen their CVs and create their own LinkedIn profiles. Inspired by the logo on a promotional stress ball at a local recruitment fair, we make the following three contributions. First, we propose a novel almost no supervision training algorithm that is effective, yet highly scalable in the number of student networks being supervised, ensuring that education remains affordable. Second, we demonstrate our approach on a typical use case: learning to bake, developing a method that tastily surpasses the current state of the art. Finally, we provide a rigorous quantitive analysis of our method, proving that we have access to a calculator. Our work calls into question the long-held dogma that life is the best teacher."
                    },
                    {
                        "title": "Unsupervised Learning of Landmarks by Descriptor Vector Exchange",
                        "abstract": "Equivariance to random image transformations is an effective method to learn landmarks of object categories, such as the eyes and the nose in faces, without manual supervision. However, this method does not explicitly guarantee that the learned landmarks are consistent with changes between different instances of the same object, such as different facial identities. In this paper, we develop a new perspective on the equivariance approach by noting that dense landmark detectors can be interpreted as local image descriptors equipped with invariance to intra-category variations. We then propose a direct method to enforce such an invariance in the standard equivariant loss. We do so by exchanging descriptor vectors between images of different object instances prior to matching them geometrically. In this manner, the same vectors must work regardless of the specific object identity considered. We use this approach to learn vectors that can simultaneously be interpreted as local descriptors and dense landmarks, combining the advantages of both. Experiments on standard benchmarks show that this approach can match, and in some cases surpass state-of-the-art performance amongst existing methods that learn landmarks without supervision. Code is available at www.robots.ox.ac.uk/~vgg/research/DVE/."
                    },
                    {
                        "title": "State-of-Art-Reviewing: A Radical Proposal to Improve Scientific Publication",
                        "abstract": "Peer review forms the backbone of modern scientific manuscript evaluation. But after two hundred and eighty-nine years of egalitarian service to the scientific community, does this protocol remain fit for purpose in 2020? In this work, we answer this question in the negative (strong reject, high confidence) and propose instead State-Of-the-Art Review (SOAR), a neoteric reviewing pipeline that serves as a 'plug-and-play' replacement for peer review. At the heart of our approach is an interpretation of the review process as a multi-objective, massively distributed and extremely-high-latency optimisation, which we scalarise and solve efficiently for PAC and CMT-optimal solutions. We make the following contributions: (1) We propose a highly scalable, fully automatic methodology for review, drawing inspiration from best-practices from premier computer vision and machine learning conferences; (2) We explore several instantiations of our approach and demonstrate that SOAR can be used to both review prints and pre-review pre-prints; (3) We wander listlessly in vain search of catharsis from our latest rounds of savage CVPR rejections."
                    },
                    {
                        "title": "Sign language segmentation with temporal convolutional networks",
                        "abstract": "The objective of this work is to determine the location of temporal boundaries between signs in continuous sign language videos. Our approach employs 3D convolutional neural network representations with iterative temporal segment refinement to resolve ambiguities between sign boundary cues. We demonstrate the effectiveness of our approach on the BSLCORPUS, PHOENIX14 and BSL-1K datasets, showing considerable improvement over the prior state of the art and the ability to generalise to new signers, languages and domains."
                    },
                    {
                        "title": "Gather-Excite: Exploiting Feature Context in Convolutional Neural Networks",
                        "abstract": "While the use of bottom-up local operators in convolutional neural networks (CNNs) matches well some of the statistics of natural images, it may also prevent such models from capturing contextual long-range feature interactions. In this work, we propose a simple, lightweight approach for better context exploitation in CNNs. We do so by introducing a pair of operators: gather, which efficiently aggregates feature responses from a large spatial extent, and excite, which redistributes the pooled information to local features. The operators are cheap, both in terms of number of added parameters and computational complexity, and can be integrated directly in existing architectures to improve their performance. Experiments on several datasets show that gather-excite can bring benefits comparable to increasing the depth of a CNN at a fraction of the cost. For example, we find ResNet-50 with gather-excite operators is able to outperform its 101-layer counterpart on ImageNet with no additional learnable parameters. We also propose a parametric gather-excite operator pair which yields further performance gains, relate it to the recently-introduced Squeeze-and-Excitation Networks, and analyse the effects of these changes to the CNN feature activation statistics."
                    },
                    {
                        "title": "Use What You Have: Video Retrieval Using Representations From Collaborative Experts",
                        "abstract": "The rapid growth of video on the internet has made searching for video content using natural language queries a significant challenge. Human-generated queries for video datasets `in the wild' vary a lot in terms of degree of specificity, with some queries describing specific details such as the names of famous identities, content from speech, or text available on the screen. Our goal is to condense the multi-modal, extremely high dimensional information from videos into a single, compact video representation for the task of video retrieval using free-form text queries, where the degree of specificity is open-ended.   For this we exploit existing knowledge in the form of pre-trained semantic embeddings which include 'general' features such as motion, appearance, and scene features from visual content. We also explore the use of more 'specific' cues from ASR and OCR which are intermittently available for videos and find that these signals remain challenging to use effectively for retrieval. We propose a collaborative experts model to aggregate information from these different pre-trained experts and assess our approach empirically on five retrieval benchmarks: MSR-VTT, LSMDC, MSVD, DiDeMo, and ActivityNet. Code and data can be found at www.robots.ox.ac.uk/~vgg/research/collaborative-experts/. This paper contains a correction to results reported in the previous version."
                    },
                    {
                        "title": "Learnable PINs: Cross-Modal Embeddings for Person Identity",
                        "abstract": "We propose and investigate an identity sensitive joint embedding of face and voice. Such an embedding enables cross-modal retrieval from voice to face and from face to voice. We make the following four contributions: first, we show that the embedding can be learnt from videos of talking faces, without requiring any identity labels, using a form of cross-modal self-supervision; second, we develop a curriculum learning schedule for hard negative mining targeted to this task, that is essential for learning to proceed successfully; third, we demonstrate and evaluate cross-modal retrieval for identities unseen and unheard during training over a number of scenarios and establish a benchmark for this novel task; finally, we show an application of using the joint embedding for automatically retrieving and labelling characters in TV dramas."
                    },
                    {
                        "title": "Self-supervised Learning of Geometrically Stable Features Through Probabilistic Introspection",
                        "abstract": "Self-supervision can dramatically cut back the amount of manually-labelled data required to train deep neural networks. While self-supervision has usually been considered for tasks such as image classification, in this paper we aim at extending it to geometry-oriented tasks such as semantic matching and part detection. We do so by building on several recent ideas in unsupervised landmark detection. Our approach learns dense distinctive visual descriptors from an unlabelled dataset of images using synthetic image transformations. It does so by means of a robust probabilistic formulation that can introspectively determine which image regions are likely to result in stable image matching. We show empirically that a network pre-trained in this manner requires significantly less supervision to learn semantic object parts compared to numerous pre-training alternatives. We also show that the pre-trained representation is excellent for semantic object matching."
                    },
                    {
                        "title": "Iterative Averaging in the Quest for Best Test Error",
                        "abstract": "We analyse and explain the increased generalisation performance of iterate averaging using a Gaussian process perturbation model between the true and batch risk surface on the high dimensional quadratic. We derive three phenomena \\latestEdits{from our theoretical results:} (1) The importance of combining iterate averaging (IA) with large learning rates and regularisation for improved regularisation. (2) Justification for less frequent averaging. (3) That we expect adaptive gradient methods to work equally well, or better, with iterate averaging than their non-adaptive counterparts. Inspired by these results\\latestEdits{, together with} empirical investigations of the importance of appropriate regularisation for the solution diversity of the iterates, we propose two adaptive algorithms with iterate averaging. These give significantly better results compared to stochastic gradient descent (SGD), require less tuning and do not require early stopping or validation set monitoring. We showcase the efficacy of our approach on the CIFAR-10/100, ImageNet and Penn Treebank datasets on a variety of modern and classical network architectures."
                    },
                    {
                        "title": "Unsupervised Salient Object Detection with Spectral Cluster Voting",
                        "abstract": "In this paper, we tackle the challenging task of unsupervised salient object detection (SOD) by leveraging spectral clustering on self-supervised features. We make the following contributions: (i) We revisit spectral clustering and demonstrate its potential to group the pixels of salient objects; (ii) Given mask proposals from multiple applications of spectral clustering on image features computed from various self-supervised models, e.g., MoCov2, SwAV, DINO, we propose a simple but effective winner-takes-all voting mechanism for selecting the salient masks, leveraging object priors based on framing and distinctiveness; (iii) Using the selected object segmentation as pseudo groundtruth masks, we train a salient object detector, dubbed SelfMask, which outperforms prior approaches on three unsupervised SOD benchmarks. Code is publicly available at https://github.com/NoelShin/selfmask."
                    }
                ]
            },
            "9708e88e-12b9-419e-ab16-5d34e561ecec": {
                "pk": "9708e88e-12b9-419e-ab16-5d34e561ecec",
                "name": "Matthias Bethge",
                "collaborators": [
                    "Matthias K\u00fcmmerer",
                    "Wieland Brendel",
                    "Lucas Theis",
                    "Jonas Rauber",
                    "Thomas S. A. Wallis",
                    "Leon A. Gatys",
                    "Ori Press",
                    "Fabian Sinz",
                    "Thomas Wallis",
                    "Steffen Schneider"
                ],
                "domain": [
                    "Computer Vision",
                    "Saliency Prediction",
                    "Adversarial Robustness",
                    "Transfer Learning"
                ],
                "publications": [
                    {
                        "title": "State-of-the-Art in Human Scanpath Prediction",
                        "abstract": "The last years have seen a surge in models predicting the scanpaths of fixations made by humans when viewing images. However, the field is lacking a principled comparison of those models with respect to their predictive power. In the past, models have usually been evaluated based on comparing human scanpaths to scanpaths generated from the model. Here, instead we evaluate models based on how well they predict each fixation in a scanpath given the previous scanpath history. This makes model evaluation closely aligned with the biological processes thought to underly scanpath generation and allows to apply established saliency metrics like AUC and NSS in an intuitive and interpretable way. We evaluate many existing models of scanpath prediction on the datasets MIT1003, MIT300, CAT2000 train and CAT200 test, for the first time giving a detailed picture of the current state of the art of human scanpath prediction. We also show that the discussed method of model benchmarking allows for more detailed analyses leading to interesting insights about where and when models fail to predict human behaviour. The MIT/Tuebingen Saliency Benchmark will implement the evaluation of scanpath models as detailed here, allowing researchers to score their models on the established benchmark datasets MIT300 and CAT2000."
                    },
                    {
                        "title": "Comment on \"Biologically inspired protection of deep networks from adversarial attacks\"",
                        "abstract": "A recent paper suggests that Deep Neural Networks can be protected from gradient-based adversarial perturbations by driving the network activations into a highly saturated regime. Here we analyse such saturated networks and show that the attacks fail due to numerical limitations in the gradient computations. A simple stabilisation of the gradient estimates enables successful and efficient attacks. Thus, it has yet to be shown that the robustness observed in highly saturated networks is not simply due to numerical limitations."
                    },
                    {
                        "title": "How close are we to understanding image-based saliency?",
                        "abstract": "Within the set of the many complex factors driving gaze placement, the properities of an image that are associated with fixations under free viewing conditions have been studied extensively. There is a general impression that the field is close to understanding this particular association. Here we frame saliency models probabilistically as point processes, allowing the calculation of log-likelihoods and bringing saliency evaluation into the domain of information. We compared the information gain of state-of-the-art models to a gold standard and find that only one third of the explainable spatial information is captured. We additionally provide a principled method to show where and how models fail to capture information in the fixations. Thus, contrary to previous assertions, purely spatial saliency remains a significant challenge."
                    },
                    {
                        "title": "RDumb: A simple approach that questions our progress in continual test-time adaptation",
                        "abstract": "Test-Time Adaptation (TTA) allows to update pre-trained models to changing data distributions at deployment time. While early work tested these algorithms for individual fixed distribution shifts, recent work proposed and applied methods for continual adaptation over long timescales. To examine the reported progress in the field, we propose the Continually Changing Corruptions (CCC) benchmark to measure asymptotic performance of TTA techniques. We find that eventually all but one state-of-the-art methods collapse and perform worse than a non-adapting model, including models specifically proposed to be robust to performance collapse. In addition, we introduce a simple baseline, \"RDumb\", that periodically resets the model to its pretrained state. RDumb performs better or on par with the previously proposed state-of-the-art in all considered benchmarks. Our results show that previous TTA approaches are neither effective at regularizing adaptation to avoid collapse nor able to outperform a simplistic resetting strategy."
                    },
                    {
                        "title": "DeepGaze IIE: Calibrated prediction in and out-of-domain for state-of-the-art saliency modeling",
                        "abstract": "Since 2014 transfer learning has become the key driver for the improvement of spatial saliency prediction; however, with stagnant progress in the last 3-5 years. We conduct a large-scale transfer learning study which tests different ImageNet backbones, always using the same read out architecture and learning protocol adopted from DeepGaze II. By replacing the VGG19 backbone of DeepGaze II with ResNet50 features we improve the performance on saliency prediction from 78% to 85%. However, as we continue to test better ImageNet models as backbones (such as EfficientNetB5) we observe no additional improvement on saliency prediction. By analyzing the backbones further, we find that generalization to other datasets differs substantially, with models being consistently overconfident in their fixation predictions. We show that by combining multiple backbones in a principled manner a good confidence calibration on unseen datasets can be achieved. This new model, \"DeepGaze IIE\", yields a significant leap in benchmark performance in and out-of-domain with a 15 percent point improvement over DeepGaze II to 93% on MIT1003, marking a new state of the art on the MIT/Tuebingen Saliency Benchmark in all available metrics (AUC: 88.3%, sAUC: 79.4%, CC: 82.4%)."
                    },
                    {
                        "title": "Deep Gaze I: Boosting Saliency Prediction with Feature Maps Trained on ImageNet",
                        "abstract": "Recent results suggest that state-of-the-art saliency models perform far from optimal in predicting fixations. This lack in performance has been attributed to an inability to model the influence of high-level image features such as objects. Recent seminal advances in applying deep neural networks to tasks like object recognition suggests that they are able to capture this kind of structure. However, the enormous amount of training data necessary to train these networks makes them difficult to apply directly to saliency prediction. We present a novel way of reusing existing neural networks that have been pretrained on the task of object recognition in models of fixation prediction. Using the well-known network of Krizhevsky et al. (2012), we come up with a new saliency model that significantly outperforms all state-of-the-art models on the MIT Saliency Benchmark. We show that the structure of this network allows new insights in the psychophysics of fixation selection and potentially their neural implementation. To train our network, we build on recent work on the modeling of saliency as point processes."
                    },
                    {
                        "title": "Generative Image Modeling Using Spatial LSTMs",
                        "abstract": "Modeling the distribution of natural images is challenging, partly because of strong statistical dependencies which can extend over hundreds of pixels. Recurrent neural networks have been successful in capturing long-range dependencies in a number of problems but only recently have found their way into generative image models. We here introduce a recurrent image model based on multi-dimensional long short-term memory units which are particularly suited for image modeling due to their spatial structure. Our model scales to images of arbitrary size and its likelihood is computationally tractable. We find that it outperforms the state of the art in quantitative comparisons on several image datasets and produces promising results when used for texture synthesis and inpainting."
                    },
                    {
                        "title": "Fast Differentiable Clipping-Aware Normalization and Rescaling",
                        "abstract": "Rescaling a vector $\\vec{\\delta} \\in \\mathbb{R}^n$ to a desired length is a common operation in many areas such as data science and machine learning. When the rescaled perturbation $\\eta \\vec{\\delta}$ is added to a starting point $\\vec{x} \\in D$ (where $D$ is the data domain, e.g. $D = [0, 1]^n$), the resulting vector $\\vec{v} = \\vec{x} + \\eta \\vec{\\delta}$ will in general not be in $D$. To enforce that the perturbed vector $v$ is in $D$, the values of $\\vec{v}$ can be clipped to $D$. This subsequent element-wise clipping to the data domain does however reduce the effective perturbation size and thus interferes with the rescaling of $\\vec{\\delta}$. The optimal rescaling $\\eta$ to obtain a perturbation with the desired norm after the clipping can be iteratively approximated using a binary search. However, such an iterative approach is slow and non-differentiable. Here we show that the optimal rescaling can be found analytically using a fast and differentiable algorithm. Our algorithm works for any p-norm and can be used to train neural networks on inputs with normalized perturbations. We provide native implementations for PyTorch, TensorFlow, JAX, and NumPy based on EagerPy."
                    },
                    {
                        "title": "DeepGaze II: Reading fixations from deep features trained on object recognition",
                        "abstract": "Here we present DeepGaze II, a model that predicts where people look in images. The model uses the features from the VGG-19 deep neural network trained to identify objects in images. Contrary to other saliency models that use deep features, here we use the VGG features for saliency prediction with no additional fine-tuning (rather, a few readout layers are trained on top of the VGG features to predict saliency). The model is therefore a strong test of transfer learning. After conservative cross-validation, DeepGaze II explains about 87% of the explainable information gain in the patterns of fixations and achieves top performance in area under the curve metrics on the MIT300 hold-out benchmark. These results corroborate the finding from DeepGaze I (which explained 56% of the explainable information gain), that deep features trained on object recognition provide a versatile feature space for performing related visual tasks. We explore the factors that contribute to this success and present several informative image examples. A web service is available to compute model predictions at http://deepgaze.bethgelab.org."
                    },
                    {
                        "title": "EagerPy: Writing Code That Works Natively with PyTorch, TensorFlow, JAX, and NumPy",
                        "abstract": "EagerPy is a Python framework that lets you write code that automatically works natively with PyTorch, TensorFlow, JAX, and NumPy. Library developers no longer need to choose between supporting just one of these frameworks or reimplementing the library for each framework and dealing with code duplication. Users of such libraries can more easily switch frameworks without being locked in by a specific 3rd party library. Beyond multi-framework support, EagerPy also brings comprehensive type annotations and consistent support for method chaining to any framework. The latest documentation is available online at https://eagerpy.jonasrauber.de and the code can be found on GitHub at https://github.com/jonasrauber/eagerpy."
                    },
                    {
                        "title": "Natural Image Coding in V1: How Much Use is Orientation Selectivity?",
                        "abstract": "Orientation selectivity is the most striking feature of simple cell coding in V1 which has been shown to emerge from the reduction of higher-order correlations in natural images in a large variety of statistical image models. The most parsimonious one among these models is linear Independent Component Analysis (ICA), whereas second-order decorrelation transformations such as Principal Component Analysis (PCA) do not yield oriented filters. Because of this finding it has been suggested that the emergence of orientation selectivity may be explained by higher-order redundancy reduction. In order to assess the tenability of this hypothesis, it is an important empirical question how much more redundancies can be removed with ICA in comparison to PCA, or other second-order decorrelation methods. This question has not yet been settled, as over the last ten years contradicting results have been reported ranging from less than five to more than hundred percent extra gain for ICA. Here, we aim at resolving this conflict by presenting a very careful and comprehensive analysis using three evaluation criteria related to redundancy reduction: In addition to the multi-information and the average log-loss we compute, for the first time, complete rate-distortion curves for ICA in comparison with PCA. Without exception, we find that the advantage of the ICA filters is surprisingly small. Furthermore, we show that a simple spherically symmetric distribution with only two parameters can fit the data even better than the probabilistic model underlying ICA. Since spherically symmetric models are agnostic with respect to the specific filter shapes, we conlude that orientation selectivity is unlikely to play a critical role for redundancy reduction."
                    },
                    {
                        "title": "$L_p$-nested symmetric distributions",
                        "abstract": "Tractable generalizations of the Gaussian distribution play an important role for the analysis of high-dimensional data. One very general super-class of Normal distributions is the class of $\\nu$-spherical distributions whose random variables can be represented as the product $\\x = r\\cdot \\u$ of a uniformly distribution random variable $\\u$ on the $1$-level set of a positively homogeneous function $\\nu$ and arbitrary positive radial random variable $r$. Prominent subclasses of $\\nu$-spherical distributions are spherically symmetric distributions ($\\nu(\\x)=\\|\\x\\|_2$) which have been further generalized to the class of $L_p$-spherically symmetric distributions ($\\nu(\\x)=\\|\\x\\|_p$). Both of these classes contain the Gaussian as a special case. In general, however, $\\nu$-spherical distributions are computationally intractable since, for instance, the normalization constant or fast sampling algorithms are unknown for an arbitrary $\\nu$. In this paper we introduce a new subclass of $\\nu$-spherical distributions by choosing $\\nu$ to be a nested cascade of $L_p$-norms. This class is still computationally tractable, but includes all the aforementioned subclasses as a special case. We derive a general expression for $L_p$-nested symmetric distributions as well as the uniform distribution on the $L_p$-nested unit sphere, including an explicit expression for the normalization constant. We state several general properties of $L_p$-nested symmetric distributions, investigate its marginals, maximum likelihood fitting and discuss its tight links to well known machine learning methods such as Independent Component Analysis (ICA), Independent Subspace Analysis (ISA) and mixed norm regularizers. Finally, we derive a fast and exact sampling algorithm for arbitrary $L_p$-nested symmetric distributions, and introduce the Nested Radial Factorization algorithm (NRF), which is a form of non-linear ICA."
                    },
                    {
                        "title": "Approximating CNNs with Bag-of-local-Features models works surprisingly well on ImageNet",
                        "abstract": "Deep Neural Networks (DNNs) excel on many complex perceptual tasks but it has proven notoriously difficult to understand how they reach their decisions. We here introduce a high-performance DNN architecture on ImageNet whose decisions are considerably easier to explain. Our model, a simple variant of the ResNet-50 architecture called BagNet, classifies an image based on the occurrences of small local image features without taking into account their spatial ordering. This strategy is closely related to the bag-of-feature (BoF) models popular before the onset of deep learning and reaches a surprisingly high accuracy on ImageNet (87.6% top-5 for 33 x 33 px features and Alexnet performance for 17 x 17 px features). The constraint on local features makes it straight-forward to analyse how exactly each part of the image influences the classification. Furthermore, the BagNets behave similar to state-of-the art deep neural networks such as VGG-16, ResNet-152 or DenseNet-169 in terms of feature sensitivity, error distribution and interactions between image parts. This suggests that the improvements of DNNs over previous bag-of-feature classifiers in the last few years is mostly achieved by better fine-tuning rather than by qualitatively different decision strategies."
                    },
                    {
                        "title": "Saliency Benchmarking Made Easy: Separating Models, Maps and Metrics",
                        "abstract": "Dozens of new models on fixation prediction are published every year and compared on open benchmarks such as MIT300 and LSUN. However, progress in the field can be difficult to judge because models are compared using a variety of inconsistent metrics. Here we show that no single saliency map can perform well under all metrics. Instead, we propose a principled approach to solve the benchmarking problem by separating the notions of saliency models, maps and metrics. Inspired by Bayesian decision theory, we define a saliency model to be a probabilistic model of fixation density prediction and a saliency map to be a metric-specific prediction derived from the model density which maximizes the expected performance on that metric given the model density. We derive these optimal saliency maps for the most commonly used saliency metrics (AUC, sAUC, NSS, CC, SIM, KL-Div) and show that they can be computed analytically or approximated with high precision. We show that this leads to consistent rankings in all metrics and avoids the penalties of using one saliency map for all metrics. Our method allows researchers to have their model compete on many different metrics with state-of-the-art in those metrics: \"good\" models will perform well in all metrics."
                    },
                    {
                        "title": "Accurate, reliable and fast robustness evaluation",
                        "abstract": "Throughout the past five years, the susceptibility of neural networks to minimal adversarial perturbations has moved from a peculiar phenomenon to a core issue in Deep Learning. Despite much attention, however, progress towards more robust models is significantly impaired by the difficulty of evaluating the robustness of neural network models. Today's methods are either fast but brittle (gradient-based attacks), or they are fairly reliable but slow (score- and decision-based attacks). We here develop a new set of gradient-based adversarial attacks which (a) are more reliable in the face of gradient-masking than other gradient-based attacks, (b) perform better and are more query efficient than current state-of-the-art gradient-based attacks, (c) can be flexibly adapted to a wide range of adversarial criteria and (d) require virtually no hyperparameter tuning. These findings are carefully validated across a diverse set of six different models and hold for L0, L1, L2 and Linf in both targeted as well as untargeted scenarios. Implementations will soon be available in all major toolboxes (Foolbox, CleverHans and ART). We hope that this class of attacks will make robustness evaluations easier and more reliable, thus contributing to more signal in the search for more robust machine learning models."
                    },
                    {
                        "title": "Synthesising Dynamic Textures using Convolutional Neural Networks",
                        "abstract": "Here we present a parametric model for dynamic textures. The model is based on spatiotemporal summary statistics computed from the feature representations of a Convolutional Neural Network (CNN) trained on object recognition. We demonstrate how the model can be used to synthesise new samples of dynamic textures and to predict motion in simple movies."
                    },
                    {
                        "title": "Guiding human gaze with convolutional neural networks",
                        "abstract": "The eye fixation patterns of human observers are a fundamental indicator of the aspects of an image to which humans attend. Thus, manipulating fixation patterns to guide human attention is an exciting challenge in digital image processing. Here, we present a new model for manipulating images to change the distribution of human fixations in a controlled fashion. We use the state-of-the-art model for fixation prediction to train a convolutional neural network to transform images so that they satisfy a given fixation distribution. For network training, we carefully design a loss function to achieve a perceptual effect while preserving naturalness of the transformed images. Finally, we evaluate the success of our model by measuring human fixations for a set of manipulated images. On our test images we can in-/decrease the probability to fixate on selected objects on average by 43/22% but show that the effectiveness of the model depends on the semantic content of the manipulated images."
                    },
                    {
                        "title": "Preserving Color in Neural Artistic Style Transfer",
                        "abstract": "This note presents an extension to the neural artistic style transfer algorithm (Gatys et al.). The original algorithm transforms an image to have the style of another given image. For example, a photograph can be transformed to have the style of a famous painting. Here we address a potential shortcoming of the original method: the algorithm transfers the colors of the original painting, which can alter the appearance of the scene in undesirable ways. We describe simple linear methods for transferring style while preserving colors."
                    },
                    {
                        "title": "An analytically tractable model of neural population activity in the presence of common input explains higher-order correlations and entropy",
                        "abstract": "Simultaneously recorded neurons exhibit correlations whose underlying causes are not known. Here, we use a population of threshold neurons receiving correlated inputs to model neural population recordings. We show analytically that small changes in second-order correlations can lead to large changes in higher correlations, and that these higher-order correlations have a strong impact on the entropy, sparsity and statistical heat capacity of the population. Remarkably, our findings for this simple model may explain a couple of surprising effects recently observed in neural population recordings."
                    },
                    {
                        "title": "Mixtures of conditional Gaussian scale mixtures applied to multiscale image representations",
                        "abstract": "We present a probabilistic model for natural images which is based on Gaussian scale mixtures and a simple multiscale representation. In contrast to the dominant approach to modeling whole images focusing on Markov random fields, we formulate our model in terms of a directed graphical model. We show that it is able to generate images with interesting higher-order correlations when trained on natural images or samples from an occlusion based model. More importantly, the directed model enables us to perform a principled evaluation. While it is easy to generate visually appealing images, we demonstrate that our model also yields the best performance reported to date when evaluated with respect to the cross-entropy rate, a measure tightly linked to the average log-likelihood."
                    }
                ]
            }
        }
    },
    "2310.05401": {
        "paper_data": {
            "title": "Entropy-MCMC: Sampling from Flat Basins with Ease",
            "url": "http://arxiv.org/abs/2310.05401v5",
            "arxiv_id": "2310.05401",
            "authors": [
                "Bolian Li",
                "Ruqi Zhang"
            ],
            "abstract": "Bayesian deep learning counts on the quality of posterior distribution estimation. However, the posterior of deep neural networks is highly multi-modal in nature, with local modes exhibiting varying generalization performance. Given a practical budget, targeting at the original posterior can lead to suboptimal performance, as some samples may become trapped in \"bad\" modes and suffer from overfitting. Leveraging the observation that \"good\" modes with low generalization error often reside in flat basins of the energy landscape, we propose to bias sampling on the posterior toward these flat regions. Specifically, we introduce an auxiliary guiding variable, the stationary distribution of which resembles a smoothed posterior free from sharp modes, to lead the MCMC sampler to flat basins. By integrating this guiding variable with the model parameter, we create a simple joint distribution that enables efficient sampling with minimal computational overhead. We prove the convergence of our method and further show that it converges faster than several existing flatness-aware methods in the strongly convex setting. Empirical results demonstrate that our method can successfully sample from flat basins of the posterior, and outperforms all compared baselines on multiple benchmarks including classification, calibration, and out-of-distribution detection.",
            "introduction": "   1 Introduction  The effectiveness of Bayesian neural networks relies heavily on the quality of posterior distribution estimation. However, achieving an accurate estimation of the full posterior is extremely difficult due to its high-dimensional and highly multi-modal nature\u00a0(Zhang et\u00a0al., 2020b; Izmailov et\u00a0al., 2021). Moreover, the numerous modes in the energy landscape typically exhibit varying generalization performance. Flat modes often show superior accuracy and robustness, whereas sharp modes tend to have high generalization errors\u00a0(Hochreiter & Schmidhuber, 1997; Keskar et\u00a0al., 2017; Bahri et\u00a0al., 2022). This connection between the geometry of energy landscape and generalization has spurred many works in optimization, ranging from theoretical understanding\u00a0(Neyshabur et\u00a0al., 2017; Dinh et\u00a0al., 2017; Dziugaite & Roy, 2018; Jiang et\u00a0al., 2019a) to new optimization algorithms\u00a0(Mobahi, 2016; Izmailov et\u00a0al., 2018; Chaudhari et\u00a0al., 2019; Foret et\u00a0al., 2020).   However, most of the existing Bayesian methods are not aware of the flatness in the energy landscape during posterior inference\u00a0(Welling & Teh, 2011; Chen et\u00a0al., 2014; Ma et\u00a0al., 2015; Zhang et\u00a0al., 2020b). Their inference strategies are usually energy-oriented and cannot distinguish between flat and sharp modes that have the same energy values. This limitation can significantly undermine their generalization performance, particularly in practical situations where capturing the full posterior is too costly. In light of this, we contend that prioritizing the capture of flat modes is essential when conducting posterior inference for Bayesian neural networks. This is advantageous for improved generalization as justified by previous works\u00a0(Hochreiter & Schmidhuber, 1997; Keskar et\u00a0al., 2017; Bahri et\u00a0al., 2022). It can further be rationalized from a Bayesian marginalization perspective: within the flat basin, each model configuration occupies a substantial volume and contributes significantly to a more precise estimation of the predictive distribution\u00a0(Bishop, 2006). Moreover, existing flatness-aware methods often rely on a single solution to represent the entire flat basin\u00a0(Chaudhari et\u00a0al., 2019; Foret et\u00a0al., 2020), ignoring the fact that the flat basin contains many high-performing models. Therefore, Bayesian marginalization can potentially offer significant improvements over flatness-aware optimization by sampling from the flat basins\u00a0(Wilson, 2020; Huang et\u00a0al., 2020).   Prioritizing flat basins during posterior inference poses an additional challenge to Bayesian inference. Even for single point estimation, explicitly biasing toward the flat basins will introduce substantial computational overhead, inducing nested loops\u00a0(Chaudhari et\u00a0al., 2019; Dziugaite & Roy, 2018), doubled gradients calculation\u00a0(Foret et\u00a0al., 2020; M\u00f6llenhoff & Khan, 2022) or min-max problems\u00a0(Foret et\u00a0al., 2020). The efficiency problem needs to be addressed before any flatness-aware Bayesian method becomes practical for deep neural networks.   In this paper, we propose an efficient sampling algorithm to explicitly prioritize flat basins in the energy landscape of deep neural networks. Specifically, we introduce an auxiliary guiding variable \ud835\udf3dasubscript\ud835\udf3d\ud835\udc4e\\bm{\\theta}_{a}bold_italic_\u03b8 start_POSTSUBSCRIPT italic_a end_POSTSUBSCRIPT into the Markov chain to pull model parameters \ud835\udf3d\ud835\udf3d\\bm{\\theta}bold_italic_\u03b8 toward flat basins at each updating step (Fig.\u00a01a). \ud835\udf3dasubscript\ud835\udf3d\ud835\udc4e\\bm{\\theta}_{a}bold_italic_\u03b8 start_POSTSUBSCRIPT italic_a end_POSTSUBSCRIPT is sampled from a smoothed posterior distribution which eliminates sharp modes based on local entropy\u00a0(Baldassi et\u00a0al., 2016) (Fig.\u00a01b). \ud835\udf3dasubscript\ud835\udf3d\ud835\udc4e\\bm{\\theta}_{a}bold_italic_\u03b8 start_POSTSUBSCRIPT italic_a end_POSTSUBSCRIPT can also be viewed as being achieved by Gaussian convolution, a common technique in diffusion models\u00a0(Sohl-Dickstein et\u00a0al., 2015; Song & Ermon, 2019). Our method enjoys a simple joint distribution of \ud835\udf3d\ud835\udf3d\\bm{\\theta}bold_italic_\u03b8 and \ud835\udf3dasubscript\ud835\udf3d\ud835\udc4e\\bm{\\theta}_{a}bold_italic_\u03b8 start_POSTSUBSCRIPT italic_a end_POSTSUBSCRIPT, and the computational overhead",
            "references": [
                {
                    "title": "Dimensionality Reduction for General KDE Mode Finding",
                    "abstract": "Finding the mode of a high dimensional probability distribution $D$ is a fundamental algorithmic problem in statistics and data analysis. There has been particular interest in efficient methods for solving the problem when $D$ is represented as a mixture model or kernel density estimate, although few algorithmic results with worst-case approximation and runtime guarantees are known. In this work, we significantly generalize a result of (LeeLiMusco:2021) on mode approximation for Gaussian mixture models. We develop randomized dimensionality reduction methods for mixtures involving a broader class of kernels, including the popular logistic, sigmoid, and generalized Gaussian kernels. As in Lee et al.'s work, our dimensionality reduction results yield quasi-polynomial algorithms for mode finding with multiplicative accuracy $(1-\\epsilon)$ for any $\\epsilon>0$. Moreover, when combined with gradient descent, they yield efficient practical heuristics for the problem. In addition to our positive results, we prove a hardness result for box kernels, showing that there is no polynomial time algorithm for finding the mode of a kernel density estimate, unless $\\mathit{P} = \\mathit{NP}$. Obtaining similar hardness results for kernels used in practice (like Gaussian or logistic kernels) is an interesting future direction."
                },
                {
                    "title": "SAM as an Optimal Relaxation of Bayes",
                    "abstract": "Sharpness-aware minimization (SAM) and related adversarial deep-learning methods can drastically improve generalization, but their underlying mechanisms are not yet fully understood. Here, we establish SAM as a relaxation of the Bayes objective where the expected negative-loss is replaced by the optimal convex lower bound, obtained by using the so-called Fenchel biconjugate. The connection enables a new Adam-like extension of SAM to automatically obtain reasonable uncertainty estimates, while sometimes also improving its accuracy. By connecting adversarial and Bayesian methods, our work opens a new path to robustness."
                },
                {
                    "title": "Low-Pass Filtering SGD for Recovering Flat Optima in the Deep Learning Optimization Landscape",
                    "abstract": "In this paper, we study the sharpness of a deep learning (DL) loss landscape around local minima in order to reveal systematic mechanisms underlying the generalization abilities of DL models. Our analysis is performed across varying network and optimizer hyper-parameters, and involves a rich family of different sharpness measures. We compare these measures and show that the low-pass filter-based measure exhibits the highest correlation with the generalization abilities of DL models, has high robustness to both data and label noise, and furthermore can track the double descent behavior for neural networks. We next derive the optimization algorithm, relying on the low-pass filter (LPF), that actively searches the flat regions in the DL optimization landscape using SGD-like procedure. The update of the proposed algorithm, that we call LPF-SGD, is determined by the gradient of the convolution of the filter kernel with the loss function and can be efficiently computed using MC sampling. We empirically show that our algorithm achieves superior generalization performance compared to the common DL training strategies. On the theoretical front, we prove that LPF-SGD converges to a better optimal point with smaller generalization error than SGD."
                },
                {
                    "title": "Sharpness-Aware Minimization Improves Language Model Generalization",
                    "abstract": "The allure of superhuman-level capabilities has led to considerable interest in language models like GPT-3 and T5, wherein the research has, by and large, revolved around new model architectures, training tasks, and loss objectives, along with substantial engineering efforts to scale up model capacity and dataset size. Comparatively little work has been done to improve the generalization of these models through better optimization. In this work, we show that Sharpness-Aware Minimization (SAM), a recently proposed optimization procedure that encourages convergence to flatter minima, can substantially improve the generalization of language models without much computational overhead. We show that SAM is able to boost performance on SuperGLUE, GLUE, Web Questions, Natural Questions, Trivia QA, and TyDiQA, with particularly large gains when training data for these tasks is limited."
                },
                {
                    "title": "What Are Bayesian Neural Network Posteriors Really Like?",
                    "abstract": "The posterior over Bayesian neural network (BNN) parameters is extremely high-dimensional and non-convex. For computational reasons, researchers approximate this posterior using inexpensive mini-batch methods such as mean-field variational inference or stochastic-gradient Markov chain Monte Carlo (SGMCMC). To investigate foundational questions in Bayesian deep learning, we instead use full-batch Hamiltonian Monte Carlo (HMC) on modern architectures. We show that (1) BNNs can achieve significant performance gains over standard training and deep ensembles; (2) a single long HMC chain can provide a comparable representation of the posterior to multiple shorter chains; (3) in contrast to recent studies, we find posterior tempering is not needed for near-optimal performance, with little evidence for a\"cold posterior\"effect, which we show is largely an artifact of data augmentation; (4) BMA performance is robust to the choice of prior scale, and relatively similar for diagonal Gaussian, mixture of Gaussian, and logistic priors; (5) Bayesian neural networks show surprisingly poor generalization under domain shift; (6) while cheaper alternatives such as deep ensembles and SGMCMC methods can provide good generalization, they provide distinct predictive distributions from HMC. Notably, deep ensemble predictive distributions are similarly close to HMC as standard SGLD, and closer than standard variational inference."
                },
                {
                    "title": "Scaling Hamiltonian Monte Carlo Inference for Bayesian Neural Networks with Symmetric Splitting",
                    "abstract": "Hamiltonian Monte Carlo (HMC) is a Markov chain Monte Carlo (MCMC) approach that exhibits favourable exploration properties in high-dimensional models such as neural networks. Unfortunately, HMC has limited use in large-data regimes and little work has explored suitable approaches that aim to preserve the entire Hamiltonian. In our work, we introduce a new symmetric integration scheme for split HMC that does not rely on stochastic gradients. We show that our new formulation is more efficient than previous approaches and is easy to implement with a single GPU. As a result, we are able to perform full HMC over common deep learning architectures using entire data sets. In addition, when we compare with stochastic gradient MCMC, we show that our method achieves better performance in both accuracy and uncertainty quantification. Our approach demonstrates HMC as a feasible option when considering inference schemes for large-scale machine learning problems."
                },
                {
                    "title": "Structured Logconcave Sampling with a Restricted Gaussian Oracle",
                    "abstract": "We give algorithms for sampling several structured logconcave families to high accuracy. We further develop a reduction framework, inspired by proximal point methods in convex optimization, which bootstraps samplers for regularized densities to improve dependences on problem conditioning. A key ingredient in our framework is the notion of a \"restricted Gaussian oracle\" (RGO) for $g: \\mathbb{R}^d \\rightarrow \\mathbb{R}$, which is a sampler for distributions whose negative log-likelihood sums a quadratic and $g$. By combining our reduction framework with our new samplers, we obtain the following bounds for sampling structured distributions to total variation distance $\\epsilon$. For composite densities $\\exp(-f(x) - g(x))$, where $f$ has condition number $\\kappa$ and convex (but possibly non-smooth) $g$ admits an RGO, we obtain a mixing time of $O(\\kappa d \\log^3\\frac{\\kappa d}{\\epsilon})$, matching the state-of-the-art non-composite bound; no composite samplers with better mixing than general-purpose logconcave samplers were previously known. For logconcave finite sums $\\exp(-F(x))$, where $F(x) = \\frac{1}{n}\\sum_{i \\in [n]} f_i(x)$ has condition number $\\kappa$, we give a sampler querying $\\widetilde{O}(n + \\kappa\\max(d, \\sqrt{nd}))$ gradient oracles to $\\{f_i\\}_{i \\in [n]}$; no high-accuracy samplers with nontrivial gradient query complexity were previously known. For densities with condition number $\\kappa$, we give an algorithm obtaining mixing time $O(\\kappa d \\log^2\\frac{\\kappa d}{\\epsilon})$, improving the prior state-of-the-art by a logarithmic factor with a significantly simpler analysis; we also show a zeroth-order algorithm attains the same query complexity."
                },
                {
                    "title": "Sharpness-Aware Minimization for Efficiently Improving Generalization",
                    "abstract": "In today's heavily overparameterized models, the value of the training loss provides few guarantees on model generalization ability. Indeed, optimizing only the training loss value, as is commonly done, can easily lead to suboptimal model quality. Motivated by the connection between geometry of the loss landscape and generalization---including a generalization bound that we prove here---we introduce a novel, effective procedure for instead simultaneously minimizing loss value and loss sharpness. In particular, our procedure, Sharpness-Aware Minimization (SAM), seeks parameters that lie in neighborhoods having uniformly low loss; this formulation results in a min-max optimization problem on which gradient descent can be performed efficiently. We present empirical results showing that SAM improves model generalization across a variety of benchmark datasets (e.g., CIFAR-{10, 100}, ImageNet, finetuning tasks) and models, yielding novel state-of-the-art performance for several. Additionally, we find that SAM natively provides robustness to label noise on par with that provided by state-of-the-art procedures that specifically target learning with noisy labels."
                },
                {
                    "title": "On the Convergence of SGD with Biased Gradients",
                    "abstract": "We analyze the complexity of biased stochastic gradient methods (SGD), where individual updates are corrupted by deterministic, i.e. biased error terms. We derive convergence results for smooth (non-convex) functions and give improved rates under the Polyak-\u0141ojasiewicz condition. We quantify how the magnitude of the bias impacts the attainable accuracy and the convergence rates (sometimes leading to divergence). Our framework covers many applications where either only biased gradient updates are available, or preferred, over unbiased ones for performance reasons. For instance, in the domain of distributed learning, biased gradient compression techniques such as top- k compression have been proposed as a tool to alleviate the communication bottleneck and in derivative-free optimization, only biased gradient estimators can be queried. We discuss a few guiding examples that show the broad appli-cability of our analysis."
                },
                {
                    "title": "Entropic gradient descent algorithms and wide flat minima",
                    "abstract": "The properties of flat minima in the empirical risk landscape of neural networks have been debated for some time. Increasing evidence suggests they possess better generalization capabilities with respect to sharp ones. In this work we first discuss the relationship between alternative measures of flatness: the local entropy, which is useful for analysis and algorithm development, and the local energy, which is easier to compute and was shown empirically in extensive tests on state-of-the-art networks to be the best predictor of generalization capabilities. We show semi-analytically in simple controlled scenarios that these two measures correlate strongly with each other and with generalization. Then, we extend the analysis to the deep learning scenario by extensive numerical validations. We study two algorithms, entropy-stochastic gradient descent and replicated-stochastic gradient descent, that explicitly include the local entropy in the optimization objective. We devise a training schedule by which we consistently find flatter minima (using both flatness measures), and improve the generalization error for common architectures (e.g. ResNet, EfficientNet)."
                },
                {
                    "title": "Asymptotically Optimal Exact Minibatch Metropolis-Hastings",
                    "abstract": "Metropolis-Hastings (MH) is a commonly-used MCMC algorithm, but it can be intractable on large datasets due to requiring computations over the whole dataset. In this paper, we study minibatch MH methods, which instead use subsamples to enable scaling. We observe that most existing minibatch MH methods are inexact (i.e. they may change the target distribution), and show that this inexactness can cause arbitrarily large errors in inference. We propose a new exact minibatch MH method, TunaMH, which exposes a tunable trade-off between its batch size and its theoretically guaranteed convergence rate. We prove a lower bound on the batch size that any minibatch MH method must use to retain exactness while guaranteeing fast convergence-the first such bound for minibatch MH-and show TunaMH is asymptotically optimal in terms of the batch size. Empirically, we show TunaMH outperforms other exact minibatch MH methods on robust linear regression, truncated Gaussian mixtures, and logistic regression."
                },
                {
                    "title": "The Case for Bayesian Deep Learning",
                    "abstract": "The key distinguishing property of a Bayesian approach is marginalization instead of optimization, not the prior, or Bayes rule. Bayesian inference is especially compelling for deep neural networks. (1) Neural networks are typically underspecified by the data, and can represent many different but high performing models corresponding to different settings of parameters, which is exactly when marginalization will make the biggest difference for both calibration and accuracy. (2) Deep ensembles have been mistaken as competing approaches to Bayesian methods, but can be seen as approximate Bayesian marginalization. (3) The structure of neural networks gives rise to a structured prior in function space, which reflects the inductive biases of neural networks that help them generalize. (4) The observed correlation between parameters in flat regions of the loss and a diversity of solutions that provide good generalization is further conducive to Bayesian marginalization, as flat regions occupy a large volume in a high dimensional space, and each different solution will make a good contribution to a Bayesian model average. (5) Recent practical advances for Bayesian deep learning provide improvements in accuracy and calibration compared to standard training, while retaining scalability."
                },
                {
                    "title": "Fantastic Generalization Measures and Where to Find Them",
                    "abstract": "Generalization of deep networks has been of great interest in recent years, resulting in a number of theoretically and empirically motivated complexity measures. However, most papers proposing such measures study only a small set of models, leaving open the question of whether the conclusion drawn from those experiments would remain valid in other settings. We present the first large scale study of generalization in deep networks. We investigate more then 40 complexity measures taken from both theoretical bounds and empirical studies. We train over 10,000 convolutional networks by systematically varying commonly used hyperparameters. Hoping to uncover potentially causal relationships between each measure and generalization, we analyze carefully controlled experiments and show surprising failures of some measures as well as promising measures for further research."
                },
                {
                    "title": "Filter Response Normalization Layer: Eliminating Batch Dependence in the Training of Deep Neural Networks",
                    "abstract": "Batch Normalization (BN) uses mini-batch statistics to normalize the activations during training, introducing dependence between mini-batch elements. This dependency can hurt the performance if the mini-batch size is too small, or if the elements are correlated. Several alternatives, such as Batch Renormalization and Group Normalization (GN), have been proposed to address this issue. However, they either do not match the performance of BN for large batches, or still exhibit degradation in performance for smaller batches, or introduce artificial constraints on the model architecture. In this paper we propose the Filter Response Normalization (FRN) layer, a novel combination of a normalization and an activation function, that can be used as a replacement for other normalizations and activations. Our method operates on each activation channel of each batch element independently, eliminating the dependency on other batch elements. Our method outperforms BN and other alternatives in a variety of settings for all batch sizes. FRN layer performs 0.7-1.0% better than BN on top-1 validation accuracy with large mini-batch sizes for Imagenet classification using InceptionV3 and ResnetV2-50 architectures. Further, it performs >1% better than GN on the same problem in the small mini-batch size regime. For object detection problem on COCO dataset, FRN layer outperforms all other methods by at least 0.3-0.5% in all batch size regimes."
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "Understanding Generalization through Visualizations",
                    "abstract": "The power of neural networks lies in their ability to generalize to unseen data, yet the underlying reasons for this phenomenon remain elusive. Numerous rigorous attempts have been made to explain generalization, but available bounds are still quite loose, and analysis does not always lead to true understanding. The goal of this work is to make generalization more intuitive. Using visualization methods, we discuss the mystery of generalization, the geometry of loss landscapes, and how the curse (or, rather, the blessing) of dimensionality causes optimizers to settle into minima that generalize well."
                },
                {
                    "title": "Benchmarking Neural Network Robustness to Common Corruptions and Perturbations",
                    "abstract": "In this paper we establish rigorous benchmarks for image classifier robustness. Our first benchmark, ImageNet-C, standardizes and expands the corruption robustness topic, while showing which classifiers are preferable in safety-critical applications. Then we propose a new dataset called ImageNet-P which enables researchers to benchmark a classifier's robustness to common perturbations. Unlike recent robustness research, this benchmark evaluates performance on common corruptions and perturbations not worst-case adversarial perturbations. We find that there are negligible changes in relative corruption robustness from AlexNet classifiers to ResNet classifiers. Afterward we discover ways to enhance corruption and perturbation robustness. We even find that a bypassed adversarial defense provides substantial common perturbation robustness. Together our benchmarks may aid future work toward networks that robustly generalize."
                },
                {
                    "title": "Cyclical Stochastic Gradient MCMC for Bayesian Deep Learning",
                    "abstract": "The posteriors over neural network weights are high dimensional and multimodal. Each mode typically characterizes a meaningfully different representation of the data. We develop Cyclical Stochastic Gradient MCMC (SG-MCMC) to automatically explore such distributions. In particular, we propose a cyclical stepsize schedule, where larger steps discover new modes, and smaller steps characterize each mode. We also prove non-asymptotic convergence of our proposed algorithm. Moreover, we provide extensive experimental results, including ImageNet, to demonstrate the scalability and effectiveness of cyclical SG-MCMC in learning complex multimodal distributions, especially for fully Bayesian inference with modern deep neural networks."
                },
                {
                    "title": "Averaging Weights Leads to Wider Optima and Better Generalization",
                    "abstract": "Deep neural networks are typically trained by optimizing a loss function with an SGD variant, in conjunction with a decaying learning rate, until convergence. We show that simple averaging of multiple points along the trajectory of SGD, with a cyclical or constant learning rate, leads to better generalization than conventional training. We also show that this Stochastic Weight Averaging (SWA) procedure finds much flatter solutions than SGD, and approximates the recent Fast Geometric Ensembling (FGE) approach with a single model. Using SWA we achieve notable improvement in test accuracy over conventional SGD training on a range of state-of-the-art residual networks, PyramidNets, DenseNets, and Shake-Shake networks on CIFAR-10, CIFAR-100, and ImageNet. In short, SWA is extremely easy to implement, improves generalization, and has almost no computational overhead."
                },
                {
                    "title": "Predictive Uncertainty Estimation via Prior Networks",
                    "abstract": "Estimating how uncertain an AI system is in its predictions is important to improve the safety of such systems. Uncertainty in predictive can result from uncertainty in model parameters, irreducible data uncertainty and uncertainty due to distributional mismatch between the test and training data distributions. Different actions might be taken depending on the source of the uncertainty so it is important to be able to distinguish between them. Recently, baseline tasks and metrics have been defined and several practical methods to estimate uncertainty developed. These methods, however, attempt to model uncertainty due to distributional mismatch either implicitly through model uncertainty or as data uncertainty. This work proposes a new framework for modeling predictive uncertainty called Prior Networks (PNs) which explicitly models distributional uncertainty. PNs do this by parameterizing a prior distribution over predictive distributions. This work focuses on uncertainty for classification and evaluates PNs on the tasks of identifying out-of-distribution (OOD) samples and detecting misclassification on the MNIST dataset, where they are found to outperform previous methods. Experiments on synthetic and MNIST and CIFAR-10 data show that unlike previous non-Bayesian methods PNs are able to distinguish between data and distributional uncertainty."
                },
                {
                    "title": "Entropy-SGD optimizes the prior of a PAC-Bayes bound: Generalization properties of Entropy-SGD and data-dependent priors",
                    "abstract": "We show that Entropy-SGD (Chaudhari et al., 2017), when viewed as a learning algorithm, optimizes a PAC-Bayes bound on the risk of a Gibbs (posterior) classifier, i.e., a randomized classifier obtained by a risk-sensitive perturbation of the weights of a learned classifier. Entropy-SGD works by optimizing the bound's prior, violating the hypothesis of the PAC-Bayes theorem that the prior is chosen independently of the data. Indeed, available implementations of Entropy-SGD rapidly obtain zero training error on random labels and the same holds of the Gibbs posterior. In order to obtain a valid generalization bound, we rely on a result showing that data-dependent priors obtained by stochastic gradient Langevin dynamics (SGLD) yield valid PAC-Bayes bounds provided the target distribution of SGLD is {\\epsilon}-differentially private. We observe that test error on MNIST and CIFAR10 falls within the (empirically nonvacuous) risk bounds computed under the assumption that SGLD reaches stationarity. In particular, Entropy-SGLD can be configured to yield relatively tight generalization bounds and still fit real labels, although these same settings do not obtain state-of-the-art performance."
                },
                {
                    "title": "Exploring Generalization in Deep Learning",
                    "abstract": "With a goal of understanding what drives generalization in deep networks, we consider several recently suggested explanations, including norm-based control, sharpness and robustness. We study how these measures can ensure generalization, highlighting the importance of scale normalization, and making a connection between sharpness and PAC-Bayes theory. We then investigate how well the measures explain different observed phenomena."
                },
                {
                    "title": "Sharp Minima Can Generalize For Deep Nets",
                    "abstract": "Despite their overwhelming capacity to overfit, deep learning architectures tend to generalize relatively well to unseen data, allowing them to be deployed in practice. However, explaining why this is the case is still an open area of research. One standing hypothesis that is gaining popularity, e.g. Hochreiter & Schmidhuber (1997); Keskar et al. (2017), is that the flatness of minima of the loss function found by stochastic gradient based methods results in good generalization. This paper argues that most notions of flatness are problematic for deep models and can not be directly applied to explain generalization. Specifically, when focusing on deep networks with rectifier units, we can exploit the particular geometry of parameter space induced by the inherent symmetries that these architectures exhibit to build equivalent models corresponding to arbitrarily sharper minima. Furthermore, if we allow to reparametrize a function, the geometry of its parameters can change drastically without affecting its generalization properties."
                },
                {
                    "title": "Entropy-SGD: biasing gradient descent into wide valleys",
                    "abstract": "This paper proposes a new optimization algorithm called Entropy-SGD for training deep neural networks that is motivated by the local geometry of the energy landscape. Local extrema with low generalization error have a large proportion of almost-zero eigenvalues in the Hessian with very few positive or negative eigenvalues. We leverage upon this observation to construct a local-entropy-based objective function that favors well-generalizable solutions lying in large flat regions of the energy landscape, while avoiding poorly-generalizable solutions located in the sharp valleys. Conceptually, our algorithm resembles two nested loops of SGD where we use Langevin dynamics in the inner loop to compute the gradient of the local entropy before each update of the weights. We show that the new objective has a smoother energy landscape and show improved generalization over SGD using uniform stability, under certain assumptions. Our experiments on convolutional and recurrent networks demonstrate that Entropy-SGD compares favorably to state-of-the-art techniques in terms of generalization error and training time."
                },
                {
                    "title": "A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks",
                    "abstract": "We consider the two related problems of detecting if an example is misclassified or out-of-distribution. We present a simple baseline that utilizes probabilities from softmax distributions. Correctly classified examples tend to have greater maximum softmax probabilities than erroneously classified and out-of-distribution examples, allowing for their detection. We assess performance by defining several tasks in computer vision, natural language processing, and automatic speech recognition, showing the effectiveness of this baseline across all. We then show the baseline can sometimes be surpassed, demonstrating the room for future research on these underexplored detection tasks."
                },
                {
                    "title": "On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima",
                    "abstract": "The stochastic gradient descent (SGD) method and its variants are algorithms of choice for many Deep Learning tasks. These methods operate in a small-batch regime wherein a fraction of the training data, say $32$-$512$ data points, is sampled to compute an approximation to the gradient. It has been observed in practice that when using a larger batch there is a degradation in the quality of the model, as measured by its ability to generalize. We investigate the cause for this generalization drop in the large-batch regime and present numerical evidence that supports the view that large-batch methods tend to converge to sharp minimizers of the training and testing functions - and as is well known, sharp minima lead to poorer generalization. In contrast, small-batch methods consistently converge to flat minimizers, and our experiments support a commonly held view that this is due to the inherent noise in the gradient estimation. We discuss several strategies to attempt to help large-batch methods eliminate this generalization gap."
                },
                {
                    "title": "Unreasonable effectiveness of learning neural networks: From accessible states and robust ensembles to basic algorithmic schemes",
                    "abstract": "Significance Artificial neural networks are some of the most widely used tools in data science. Learning is, in principle, a hard problem in these systems, but in practice heuristic algorithms often find solutions with good generalization properties. We propose an explanation of this good performance in terms of a nonequilibrium statistical physics framework: We show that there are regions of the optimization landscape that are both robust and accessible and that their existence is crucial to achieve good performance on a class of particularly difficult learning problems. Building on these results, we introduce a basic algorithmic scheme that improves existing optimization algorithms and provides a framework for further research on learning in neural networks. In artificial neural networks, learning from data is a computationally demanding task in which a large number of connection weights are iteratively tuned through stochastic-gradient-based heuristic processes over a cost function. It is not well understood how learning occurs in these systems, in particular how they avoid getting trapped in configurations with poor computational performance. Here, we study the difficult case of networks with discrete weights, where the optimization landscape is very rough even for simple architectures, and provide theoretical and numerical evidence of the existence of rare\u2014but extremely dense and accessible\u2014regions of configurations in the network weight space. We define a measure, the robust ensemble (RE), which suppresses trapping by isolated configurations and amplifies the role of these dense regions. We analytically compute the RE in some exactly solvable models and also provide a general algorithmic scheme that is straightforward to implement: define a cost function given by a sum of a finite number of replicas of the original cost function, with a constraint centering the replicas around a driving assignment. To illustrate this, we derive several powerful algorithms, ranging from Markov Chains to message passing to gradient descent processes, where the algorithms target the robust dense states, resulting in substantial improvements in performance. The weak dependence on the number of precision bits of the weights leads us to conjecture that very similar reasoning applies to more conventional neural networks. Analogous algorithmic schemes can also be applied to other optimization problems."
                },
                {
                    "title": "Training Recurrent Neural Networks by Diffusion",
                    "abstract": "This work presents a new algorithm for training recurrent neural networks (although ideas are applicable to feedforward networks as well). The algorithm is derived from a theory in nonconvex optimization related to the diffusion equation. The contributions made in this work are two fold. First, we show how some seemingly disconnected mechanisms used in deep learning such as smart initialization, annealed learning rate, layerwise pretraining, and noise injection (as done in dropout and SGD) arise naturally and automatically from this framework, without manually crafting them into the algorithms. Second, we present some preliminary results on comparing the proposed method against SGD. It turns out that the new algorithm can achieve similar level of generalization accuracy of SGD in much fewer number of epochs."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Subdominant Dense Clusters Allow for Simple Learning and High Computational Performance in Neural Networks with Discrete Synapses.",
                    "abstract": "We show that discrete synaptic weights can be efficiently used for learning in large scale neural systems, and lead to unanticipated computational performance. We focus on the representative case of learning random patterns with binary synapses in single layer networks. The standard statistical analysis shows that this problem is exponentially dominated by isolated solutions that are extremely hard to find algorithmically. Here, we introduce a novel method that allows us to find analytical evidence for the existence of subdominant and extremely dense regions of solutions. Numerical experiments confirm these findings. We also show that the dense regions are surprisingly accessible by simple learning protocols, and that these synaptic configurations are robust to perturbations and generalize better than typical solutions. These outcomes extend to synapses with multiple states and to deeper neural architectures. The large deviation measure also suggests how to design novel algorithmic schemes for optimization based on local entropy maximization."
                },
                {
                    "title": "A Complete Recipe for Stochastic Gradient MCMC",
                    "abstract": "Many recent Markov chain Monte Carlo (MCMC) samplers leverage continuous dynamics to define a transition kernel that efficiently explores a target distribution. In tandem, a focus has been on devising scalable variants that subsample the data and use stochastic gradients in place of full-data gradients in the dynamic simulations. However, such stochastic gradient MCMC samplers have lagged behind their full-data counterparts in terms of the complexity of dynamics considered since proving convergence in the presence of the stochastic gradient noise is non-trivial. Even with simple dynamics, significant physical intuition is often required to modify the dynamical system to account for the stochastic gradient noise. In this paper, we provide a general recipe for constructing MCMC samplers--including stochastic gradient versions--based on continuous Markov processes specified via two matrices. We constructively prove that the framework is complete. That is, any continuous Markov process that provides samples from the target distribution can be written in our framework. We show how previous continuous-dynamic samplers can be trivially \"reinvented\" in our framework, avoiding the complicated sampler-specific proofs. We likewise use our recipe to straightforwardly propose a new state-adaptive sampler: stochastic gradient Riemann Hamiltonian Monte Carlo (SGRHMC). Our experiments on simulated data and a streaming Wikipedia analysis demonstrate that the proposed SGRHMC sampler inherits the benefits of Riemann HMC, with the scalability of stochastic gradient methods."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "Obtaining Well Calibrated Probabilities Using Bayesian Binning",
                    "abstract": "Learning probabilistic predictive models that are well calibrated is critical for many prediction and decision-making tasks in artificial intelligence. In this paper we present a new non-parametric calibration method called Bayesian Binning into Quantiles (BBQ) which addresses key limitations of existing calibration methods. The method post processes the output of a binary classification algorithm; thus, it can be readily combined with many existing classification algorithms. The method is computationally tractable, and empirically accurate, as evidenced by the set of experiments reported here on both real and simulated datasets."
                },
                {
                    "title": "Efficient mini-batch training for stochastic optimization",
                    "abstract": "Stochastic gradient descent (SGD) is a popular technique for large-scale optimization problems in machine learning. In order to parallelize SGD, minibatch training needs to be employed to reduce the communication cost. However, an increase in minibatch size typically decreases the rate of convergence. This paper introduces a technique based on approximate optimization of a conservatively regularized objective function within each minibatch. We prove that the convergence rate does not decrease with increasing minibatch size. Experiments demonstrate that with suitable implementations of approximate optimization, the resulting algorithm can outperform standard SGD in many scenarios."
                },
                {
                    "title": "Firefly Monte Carlo: Exact MCMC with Subsets of Data",
                    "abstract": "Markov chain Monte Carlo (MCMC) is a popular tool for Bayesian inference. However, MCMC cannot be practically applied to large data sets because of the prohibitive cost of evaluating every likelihood term at every iteration. Here we present Firefly Monte Carlo (FlyMC) MCMC algorithm with auxiliary variables that only queries the likelihoods of a subset of the data at each iteration yet simulates from the exact posterior distribution. FlyMC is compatible with modern MCMC algorithms, and only requires a lower bound on the per-datum likelihood factors. In experiments, we find that FlyMC generates samples from the posterior more than an order of magnitude faster than regular MCMC, allowing MCMC methods to tackle larger datasets than were previously considered feasible."
                },
                {
                    "title": "Stochastic Gradient Hamiltonian Monte Carlo",
                    "abstract": "Hamiltonian Monte Carlo (HMC) sampling methods provide a mechanism for defining distant proposals with high acceptance probabilities in a Metropolis-Hastings framework, enabling more efficient exploration of the state space than standard random-walk proposals. The popularity of such methods has grown significantly in recent years. However, a limitation of HMC methods is the required gradient computation for simulation of the Hamiltonian dynamical system--such computation is infeasible in problems involving a large sample size or streaming data. Instead, we must rely on a noisy gradient estimate computed from a subset of the data. In this paper, we explore the properties of such a stochastic gradient HMC approach. Surprisingly, the natural implementation of the stochastic approximation can be arbitrarily bad. To address this problem we introduce a variant that uses second-order Langevin dynamics with a friction term that counteracts the effects of the noisy gradient, maintaining the desired target distribution as the invariant distribution. Results on simulated data validate our theory. We also provide an application of our methods to a classification task using neural networks and to online Bayesian matrix factorization."
                },
                {
                    "title": "Bayesian Learning via Stochastic Gradient Langevin Dynamics",
                    "abstract": "In this paper we propose a new framework for learning from large scale datasets based on iterative learning from small mini-batches. By adding the right amount of noise to a standard stochastic gradient optimization algorithm we show that the iterates will converge to samples from the true posterior distribution as we anneal the stepsize. This seamless transition between optimization and Bayesian posterior sampling provides an inbuilt protection against overfitting. We also propose a practical method for Monte Carlo estimates of posterior statistics which monitors a \"sampling threshold\" and collects samples after it has been surpassed. We apply the method to three models: a mixture of Gaussians, logistic regression and ICA with natural gradients."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "All of Statistics: A Concise Course in Statistical Inference",
                    "abstract": "WINNER OF THE 2005 DEGROOT PRIZE! This book is for people who want to learn probability and statistics quickly. It brings together many of the main ideas in modern statistics in one place. The book is suitable for students and researchers in statistics, computer science, data mining and machine learning. This book covers a much wider range of topics than a typical introductory text on mathematical statistics. It includes modern topics like nonparametric curve estimation, bootstrapping and classification, topics that are usually relegated to follow-up courses. The reader is assumed to know calculus and a little linear algebra. No previous knowledge of probability and statistics is required. The text can be used at the advanced undergraduate and graduate level."
                },
                {
                    "title": "Analyzing a Portion of the ROC Curve",
                    "abstract": "The area under the ROC curve is a common index summarizing the information contained in the curve. When comparing two ROC curves, though, problems arise when interest does not lie in the entire range of false-positive rates (and hence the entire area). Numerical integration is suggested for evaluating the area under a portion of the ROC curve. Variance estimates are derived. The method is applicable for either continuous or rating scale binormal data, from independent or dependent samples. An example is presented which looks at rating scale data of computed tomographic scans of the head with and without concomitant use of clinical history. The areas under the two ROC curves over an a priori range of false- positive rates are examined, as well as the areas under the two curves at a specific point."
                },
                {
                    "title": "Reading Digits in Natural Images with Unsupervised Feature Learning",
                    "abstract": "Detecting and reading text from natural images is a hard computer vision task that is central to a variety of emerging applications. Related problems like document character recognition have been widely studied by computer vision and machine learning researchers and are virtually solved for practical applications like reading handwritten digits. Reliably recognizing characters in more complex scenes like photographs, however, is far more difficult: the best existing methods lag well behind human performance on the same tasks. In this paper we attack the problem of recognizing digits in a real application using unsupervised feature learning methods: reading house numbers from street level photos. To this end, we introduce a new benchmark dataset for research use containing over 600,000 labeled digits cropped from Street View images. We then demonstrate the difficulty of recognizing these digits when the problem is approached with hand-designed features. Finally, we employ variants of two recently proposed unsupervised feature learning methods and find that they are convincingly superior on our benchmarks."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Advanced Data Mining Techniques",
                    "abstract": "This book covers the fundamental concepts of data mining, to demonstrate the potential of gathering large sets of data, and analyzing these data sets to gain useful business understanding. The book is organized in three parts. Part I introduces concepts. Part II describes and demonstrates basic data mining algorithms. It also contains chapters on a number of different techniques often used in data mining. Part III focusses on business applications of data mining.Methods are presented with simple examples, applications are reviewed, and relativ advantages are evaluated."
                },
                {
                    "title": "The mnist database of handwritten digits",
                    "abstract": "Disclosed is an improved articulated bar flail having shearing edges for efficiently shredding materials. An improved shredder cylinder is disclosed with a plurality of these flails circumferentially spaced and pivotally attached to the periphery of a rotatable shaft. Also disclosed is an improved shredder apparatus which has a pair of these shredder cylinders mounted to rotate about spaced parallel axes which cooperates with a conveyer apparatus which has a pair of inclined converging conveyer belts with one of the belts mounted to move with respect to the other belt to allow the transport of articles of various sizes therethrough."
                },
                {
                    "title": "Flat Minima",
                    "abstract": "We present a new algorithm for finding low-complexity neural networks with high generalization capability. The algorithm searches for a flat minimum of the error function. A flat minimum is a large connected region in weight space where the error remains approximately constant. An MDL-based, Bayesian argument suggests that flat minima correspond to simple networks and low expected overfitting. The argument is based on a Gibbs algorithm variant and a novel way of splitting generalization error into underfitting and overfitting error. Unlike many previous approaches, ours does not require gaussian assumptions and does not depend on a good weight prior. Instead we have a prior over input output functions, thus taking into account net architecture and training set. Although our algorithm requires the computation of second-order derivatives, it has backpropagation's order of complexity. Automatically, it effectively prunes units, weights, and input lines. Various experiments with feedforward and recurrent nets are described. In an application to stock market prediction, flat minimum search outperforms conventional backprop, weight decay, and optimal brain surgeon/optimal brain damage."
                },
                {
                    "title": "Simplifying Neural Nets by Discovering Flat Minima",
                    "abstract": "We present a new algorithm for finding low complexity networks with high generalization capability. The algorithm searches for large connected regions of so-called \"flat\" minima of the error function. In the weight-space environment of a \"flat\" minimum, the error remains approximately constant. Using an MDL-based argument, flat minima can be shown to correspond to low expected overfitting. Although our algorithm requires the computation of second order derivatives, it has backprop's order of complexity. Experiments with feedforward and recurrent nets are described. In an application to stock market prediction, the method outperforms conventional backprop, weight decay, and \"optimal brain surgeon\"."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we efficiently prioritize the capture of flat modes in the posterior distribution during Bayesian inference for deep neural networks to improve generalization performance?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the understanding of the relationship between the geometry of the energy landscape and generalization in Bayesian neural networks. By effectively capturing flat modes, we can enhance the robustness and accuracy of models, leading to better performance in practical applications. This research could pave the way for new methodologies in Bayesian inference, influencing future studies on optimization and model selection, and ultimately contributing to the development of more reliable machine learning systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the high-dimensional and multi-modal nature of the posterior distribution, where distinguishing between flat and sharp modes is non-trivial. Naive approaches may fail because they do not account for the varying generalization performance associated with different modes, leading to suboptimal inference. Additionally, explicitly biasing toward flat basins introduces significant computational overhead, such as nested loops and doubled gradient calculations, which complicates the implementation of flatness-aware methods in deep learning contexts.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely overlooked the importance of flatness in the energy landscape during posterior inference, focusing instead on energy-oriented strategies that do not differentiate between modes with similar energy values. Existing flatness-aware methods often rely on single solutions to represent flat basins, neglecting the diversity of high-performing models within these regions. Barriers such as computational inefficiency and the complexity of accurately sampling from flat basins have hindered progress. Our approach differs by introducing an auxiliary guiding variable that facilitates efficient sampling from flat basins, addressing these limitations.\n\n**[Question 5] - What are the key components of my approach and results?**  \nWe propose an efficient sampling algorithm that incorporates an auxiliary guiding variable \\( \\bm{\\theta}_{a} \\) into the Markov chain to direct model parameters \\( \\bm{\\theta} \\) toward flat basins during each update. This variable is sampled from a smoothed posterior distribution that eliminates sharp modes based on local entropy. Our methodology aims to maintain a simple joint distribution of \\( \\bm{\\theta} \\) and \\( \\bm{\\theta}_{a} \\), minimizing computational overhead. We expect that this approach will lead to improved generalization performance in Bayesian neural networks by effectively capturing"
            }
        },
        "author_data": {
            "d9174f26-5cb0-4e5e-a0fc-c97c5ef65d6c": {
                "pk": "d9174f26-5cb0-4e5e-a0fc-c97c5ef65d6c",
                "name": "Bolian Li",
                "collaborators": [
                    "Ruqi Zhang",
                    "Changqing Zhang",
                    "Yifan Wang",
                    "A. Grama",
                    "Yi Ding",
                    "Zige Zheng",
                    "Zongbo Han",
                    "Haining Li",
                    "H. Fu",
                    "Baoyu Jing",
                    "Hanghang Tong"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Graph Representation Learning",
                    "Uncertainty Estimation",
                    "Long-tailed Classification"
                ],
                "publications": [
                    {
                        "title": "Cascade Reward Sampling for Efficient Decoding-Time Alignment",
                        "abstract": "Aligning large language models (LLMs) with human preferences is critical for their deployment. Recently, decoding-time alignment has emerged as an effective plug-and-play technique that requires no fine-tuning of model parameters. However, generating text that achieves both high reward and high likelihood remains a significant challenge. Existing methods often fail to generate high-reward text or incur substantial computational costs. In this paper, we propose Cascade Reward Sampling (CARDS) to address both issues, guaranteeing the generation of high-reward and high-likelihood text with significantly low costs. Based on our analysis of reward models (RMs) on incomplete text and our observation that high-reward prefixes induce high-reward complete text, we use rejection sampling to iteratively generate small semantic segments to form such prefixes. The segment length is dynamically determined by the predictive uncertainty of LLMs. This strategy guarantees desirable prefixes for subsequent generations and significantly reduces wasteful token re-generations and the number of reward model scoring. Our experiments demonstrate substantial gains in both generation efficiency and alignment ratings compared to the baselines, achieving five times faster text generation and 99\\% win-ties in GPT-4/Claude-3 helpfulness evaluation."
                    },
                    {
                        "title": "ETA: Evaluating Then Aligning Safety of Vision Language Models at Inference Time",
                        "abstract": "Vision Language Models (VLMs) have become essential backbones for multimodal intelligence, yet significant safety challenges limit their real-world application. While textual inputs are often effectively safeguarded, adversarial visual inputs can easily bypass VLM defense mechanisms. Existing defense methods are either resource-intensive, requiring substantial data and compute, or fail to simultaneously ensure safety and usefulness in responses. To address these limitations, we propose a novel two-phase inference-time alignment framework, Evaluating Then Aligning (ETA): 1) Evaluating input visual contents and output responses to establish a robust safety awareness in multimodal settings, and 2) Aligning unsafe behaviors at both shallow and deep levels by conditioning the VLMs' generative distribution with an interference prefix and performing sentence-level best-of-N to search the most harmless and helpful generation paths. Extensive experiments show that ETA outperforms baseline methods in terms of harmlessness, helpfulness, and efficiency, reducing the unsafe rate by 87.5% in cross-modality attacks and achieving 96.6% win-ties in GPT-4 helpfulness evaluation. The code is publicly available at https://github.com/DripNowhy/ETA."
                    },
                    {
                        "title": "Long-tailed Classification from a Bayesian-decision-theory Perspective",
                        "abstract": "Long-tailed classification poses a challenge due to its heavy imbalance in class probabilities and tail-sensitivity risks with asymmetric misprediction costs. Recent attempts have used re-balancing loss and ensemble methods, but they are largely heuristic and depend heavily on empirical results, lacking theoretical explanation. Furthermore, existing methods overlook the decision loss, which characterizes different costs associated with tailed classes. This paper presents a general and principled framework from a Bayesian-decision-theory perspective, which unifies existing techniques including re-balancing and ensemble methods, and provides theoretical justifications for their effectiveness. From this perspective, we derive a novel objective based on the integrated risk and a Bayesian deep-ensemble approach to improve the accuracy of all classes, especially the\"tail\". Besides, our framework allows for task-adaptive decision loss which provides provably optimal decisions in varying task scenarios, along with the capability to quantify uncertainty. Finally, We conduct comprehensive experiments, including standard classification, tail-sensitive classification with a new False Head Rate metric, calibration, and ablation studies. Our framework significantly improves the current SOTA even on large-scale real-world datasets like ImageNet."
                    },
                    {
                        "title": "Identifying Incorrect Classifications with Balanced Uncertainty",
                        "abstract": "Uncertainty estimation is critical for cost-sensitive deep-learning applications (i.e. disease diagnosis). It is very challenging partly due to the inaccessibility of uncertainty groundtruth in most datasets. Previous works proposed to estimate the uncertainty from softmax calibration, Monte Carlo sampling, subjective logic and so on. However, these existing methods tend to be over-confident about their predictions with unreasonably low overall uncertainty, which originates from the imbalance between positive (correct classifications) and negative (incorrect classifications) samples. For this issue, we firstly propose the distributional imbalance to model the imbalance in uncertainty estimation as two kinds of distribution biases, and secondly propose Balanced True Class Probability (BTCP) framework, which learns an uncertainty estimator with a novel Distributional Focal Loss (DFL) objective. Finally, we evaluate the BTCP in terms of failure prediction and out-of-distribution (OOD) detection on multiple datasets. The experimental results show that BTCP outperforms other uncertainty estimation methods especially in identifying incorrect classifications."
                    },
                    {
                        "title": "Trustworthy Long-Tailed Classification",
                        "abstract": "Classification on long-tailed distributed data is a challenging problem, which suffers from serious class-imbalance and accordingly unpromising performance es-pecially on tail classes. Recently, the ensembling based methods achieve the state-of-the-art performance and show great potential. However, there are two limitations for cur-rent methods. First, their predictions are not trustworthy for failure-sensitive applications. This is especially harmful for the tail classes where the wrong predictions is basically fre-quent. Second, they assign unified numbers of experts to all samples, which is redundant for easy samples with excessive computational cost. To address these issues, we propose a Trustworthy Long-tailed Classification (TLC) method to jointly conduct classification and uncertainty estimation to identify hard samples in a multi-expert framework. Our TLC obtains the evidence-based uncertainty (EvU) and ev-idence for each expert, and then combines these uncer-tainties and evidences under the Dempster-Shafer Evidence Theory (DST). Moreover, we propose a dynamic expert en-gagement to reduce the number of engaged experts for easy samples and achieve efficiency while maintaining promising performances. Finally, we conduct comprehensive ex-periments on the tasks of classification, tail detection, OOD detection and failure prediction. The experimental results show that the proposed TLC outperforms existing methods and is trustworthy with reliable uncertainty."
                    },
                    {
                        "title": "Graph Communal Contrastive Learning",
                        "abstract": "Graph representation learning is crucial for many real-world applications (e.g. social relation analysis). A fundamental problem for graph representation learning is how to effectively learn representations without human labeling, which is usually costly and time-consuming. Graph contrastive learning (GCL) addresses this problem by pulling the positive node pairs (or similar nodes) closer while pushing the negative node pairs (or dissimilar nodes) apart in the representation space. Despite the success of the existing GCL methods, they primarily sample node pairs based on the node-level proximity yet the community structures have rarely been taken into consideration. As a result, two nodes from the same community might be sampled as a negative pair. We argue that the community information should be considered to identify node pairs in the same communities, where the nodes insides are semantically similar. To address this issue, we propose a novel Graph Communal Contrastive Learning (gCooL) framework to jointly learn the community partition and learn node representations in an end-to-end fashion. Specifically, the proposed gCooL consists of two components: a Dense Community Aggregation (DeCA) algorithm for community detection and a Reweighted Self-supervised Cross-contrastive (ReSC) training scheme to utilize the community information. Additionally, the real-world graphs are complex and often consist of multiple views. In this paper, we demonstrate that the proposed gCooL can also be naturally adapted to multiplex graphs. Finally, we comprehensively evaluate the proposed gCooL on a variety of real-world graphs. The experimental results show that the gCooL outperforms the state-of-the-art methods."
                    }
                ]
            },
            "46ed6d45-1301-4d07-971e-41d2646d18a4": {
                "pk": "46ed6d45-1301-4d07-971e-41d2646d18a4",
                "name": "Ruqi Zhang",
                "collaborators": [
                    "Bolian Li",
                    "Yifan Wang",
                    "A. Grama",
                    "Yi Ding",
                    "Chunyuan Li",
                    "Jianyi Zhang",
                    "Changyou Chen",
                    "A. Wilson"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Multimodal Learning",
                    "Bayesian Inference"
                ],
                "publications": [
                    {
                        "title": "Cascade Reward Sampling for Efficient Decoding-Time Alignment",
                        "abstract": "Aligning large language models (LLMs) with human preferences is critical for their deployment. Recently, decoding-time alignment has emerged as an effective plug-and-play technique that requires no fine-tuning of model parameters. However, generating text that achieves both high reward and high likelihood remains a significant challenge. Existing methods often fail to generate high-reward text or incur substantial computational costs. In this paper, we propose Cascade Reward Sampling (CARDS) to address both issues, guaranteeing the generation of high-reward and high-likelihood text with significantly low costs. Based on our analysis of reward models (RMs) on incomplete text and our observation that high-reward prefixes induce high-reward complete text, we use rejection sampling to iteratively generate small semantic segments to form such prefixes. The segment length is dynamically determined by the predictive uncertainty of LLMs. This strategy guarantees desirable prefixes for subsequent generations and significantly reduces wasteful token re-generations and the number of reward model scoring. Our experiments demonstrate substantial gains in both generation efficiency and alignment ratings compared to the baselines, achieving five times faster text generation and 99\\% win-ties in GPT-4/Claude-3 helpfulness evaluation."
                    },
                    {
                        "title": "ETA: Evaluating Then Aligning Safety of Vision Language Models at Inference Time",
                        "abstract": "Vision Language Models (VLMs) have become essential backbones for multimodal intelligence, yet significant safety challenges limit their real-world application. While textual inputs are often effectively safeguarded, adversarial visual inputs can easily bypass VLM defense mechanisms. Existing defense methods are either resource-intensive, requiring substantial data and compute, or fail to simultaneously ensure safety and usefulness in responses. To address these limitations, we propose a novel two-phase inference-time alignment framework, Evaluating Then Aligning (ETA): 1) Evaluating input visual contents and output responses to establish a robust safety awareness in multimodal settings, and 2) Aligning unsafe behaviors at both shallow and deep levels by conditioning the VLMs' generative distribution with an interference prefix and performing sentence-level best-of-N to search the most harmless and helpful generation paths. Extensive experiments show that ETA outperforms baseline methods in terms of harmlessness, helpfulness, and efficiency, reducing the unsafe rate by 87.5% in cross-modality attacks and achieving 96.6% win-ties in GPT-4 helpfulness evaluation. The code is publicly available at https://github.com/DripNowhy/ETA."
                    },
                    {
                        "title": "Long-tailed Classification from a Bayesian-decision-theory Perspective",
                        "abstract": "Long-tailed classification poses a challenge due to its heavy imbalance in class probabilities and tail-sensitivity risks with asymmetric misprediction costs. Recent attempts have used re-balancing loss and ensemble methods, but they are largely heuristic and depend heavily on empirical results, lacking theoretical explanation. Furthermore, existing methods overlook the decision loss, which characterizes different costs associated with tailed classes. This paper presents a general and principled framework from a Bayesian-decision-theory perspective, which unifies existing techniques including re-balancing and ensemble methods, and provides theoretical justifications for their effectiveness. From this perspective, we derive a novel objective based on the integrated risk and a Bayesian deep-ensemble approach to improve the accuracy of all classes, especially the\"tail\". Besides, our framework allows for task-adaptive decision loss which provides provably optimal decisions in varying task scenarios, along with the capability to quantify uncertainty. Finally, We conduct comprehensive experiments, including standard classification, tail-sensitive classification with a new False Head Rate metric, calibration, and ablation studies. Our framework significantly improves the current SOTA even on large-scale real-world datasets like ImageNet."
                    },
                    {
                        "title": "Cyclical Stochastic Gradient MCMC for Bayesian Deep Learning",
                        "abstract": "The posteriors over neural network weights are high dimensional and multimodal. Each mode typically characterizes a meaningfully different representation of the data. We develop Cyclical Stochastic Gradient MCMC (SG-MCMC) to automatically explore such distributions. In particular, we propose a cyclical stepsize schedule, where larger steps discover new modes, and smaller steps characterize each mode. We also prove non-asymptotic convergence of our proposed algorithm. Moreover, we provide extensive experimental results, including ImageNet, to demonstrate the scalability and effectiveness of cyclical SG-MCMC in learning complex multimodal distributions, especially for fully Bayesian inference with modern deep neural networks."
                    }
                ]
            }
        }
    },
    "2406.01345": {
        "paper_data": {
            "title": "BMRS: Bayesian Model Reduction for Structured Pruning",
            "url": "http://arxiv.org/abs/2406.01345v1",
            "arxiv_id": "2406.01345",
            "authors": [
                "Dustin Wright",
                "Christian Igel",
                "Raghavendra Selvan"
            ],
            "abstract": "Modern neural networks are often massively overparameterized leading to high compute costs during training and at inference. One effective method to improve both the compute and energy efficiency of neural networks while maintaining good performance is structured pruning, where full network structures (e.g. neurons or convolutional filters) that have limited impact on the model output are removed. In this work, we propose Bayesian Model Reduction for Structured pruning (BMRS), a fully end-to-end Bayesian method of structured pruning. BMRS is based on two recent methods: Bayesian structured pruning with multiplicative noise, and Bayesian model reduction (BMR), a method which allows efficient comparison of Bayesian models under a change in prior. We present two realizations of BMRS derived from different priors which yield different structured pruning characteristics: 1) BMRS_N with the truncated log-normal prior, which offers reliable compression rates and accuracy without the need for tuning any thresholds and 2) BMRS_U with the truncated log-uniform prior that can achieve more aggressive compression based on the boundaries of truncation. Overall, we find that BMRS offers a theoretically grounded approach to structured pruning of neural networks yielding both high compression rates and accuracy. Experiments on multiple datasets and neural networks of varying complexity showed that the two BMRS methods offer a competitive performance-efficiency trade-off compared to other pruning methods.",
            "introduction": "   Section 1 Introduction  Modern neural networks come with an increasing computational burden, as scale is often seen to be associated with performance\u00a0Thompson et\u00a0al. (2023). The response to this has been a focus on research around the topic of neural network efficiency\u00a0Bartoldson et\u00a0al. (2023), where the goal is to reduce the computational cost of a system while maintaining other desirable metrics. As such, selecting a method to improve efficiency comes with many tradeoffs, including how to balance compute and energy consumption with accuracy\u00a0Wright et\u00a0al. (2023).   Neural network pruning seeks to do this by removing parts of a network which have limited impact on its output. This comes in two primary forms: unstructured pruning, where individual weights are removed, and structured pruning, where entire neural network structures such as neurons and convolutional filters are removed\u00a0Liang et\u00a0al. (2021). Structured pruning is often desirable as unstructured pruning can result in sparse computations which are energy intensive on current hardware, while structured pruning can maintain more energy efficient dense operations\u00a0Henderson et\u00a0al. (2020); Jeon and Kim (2018); Peng et\u00a0al. (2023). Many ways to perform structured pruning have been proposed, but the challenge of how to appropriately balance accuracy and complexity in a principled manner has remained.   In this work, we address this challenge by proposing BMRS: Bayesian Model Reduction for Structured pruning. BMRS is a principled method based on combining two complementary lines of work: Bayesian structured pruning with multiplicative noise\u00a0Neklyudov et\u00a0al. (2017) and Bayesian model reduction (BMR)\u00a0Friston et\u00a0al. (2018), a method of efficient Bayesian model comparison under a change in prior. Multiplicative noise allows one to flexibly induce sparsity at any structural level without the need to use computationally complex spike-and-slab priors\u00a0Jantre et\u00a0al. (2023); Markovic et\u00a0al. (2023), while BMR enables principled pruning rules without the need for task-specific threshold tuning. Starting with the approach from Neklyudov et\u00a0al. (2017), we derive two versions of BMRS using different priors which offer their own benefits. BMRSN is based on the truncated log-normal prior and has the benefit of achieving a high compression rate without needing to tune a threshold for compression, while BMRSU offers tunable compression by controlling the allowable precision of noise variables in the network. In sum, our contributions are:   \u2022  BMRS: a method for Bayesian structured pruning based on multiplicative noise and Bayesian model reduction;    \u2022  Derivations of pruning algorithms for two priors with theoretical motivation;    \u2022  Empirical results on a range of neural networks and datasets demonstrating high compression rates without any threshold tuning, with more extreme compression achievable via a parameter controlling allowable precision.      Figure 1: BMRS uses BMR to perform structured pruning under multiplicative noise by calculating the change in log-evidence of noise variables \u03b8\ud835\udf03\\thetaitalic_\u03b8 under a prior which would shrink them to 0.     Section 2 Related work  The primary goal of neural network pruning is to determine the elements of a network which can be removed with minimal impact on the output. Ideally, a pruning method ranks all elements in the order in which they can be removed and provides a criterion for truncating the resulting ordered list. Since the early works on gradient based methods for pruning\u00a0LeCun et\u00a0al. (1989); Hassibi and Stork (1992),",
            "references": [
                {
                    "title": "Bayesian Sparsification for Deep Neural Networks With Bayesian Model Reduction",
                    "abstract": "Deep learning\u2019s immense capabilities are often constrained by the complexity of its models, leading to an increasing demand for effective sparsification techniques. Bayesian sparsification for deep learning emerges as a crucial approach, facilitating the design of models that are both computationally efficient and competitive in terms of performance across various deep learning applications. The state-of-the-art \u2013 in Bayesian sparsification of deep neural networks \u2013 combines structural shrinkage priors on model weights with an approximate inference scheme based on stochastic variational inference. However, model inversion of the full generative model is exceptionally computationally demanding, especially when compared to standard deep learning of point estimates. In this context, we advocate for the use of Bayesian model reduction (BMR) as a more efficient alternative for pruning of model weights. As a generalization of the Savage-Dickey ratio, BMR allows a post-hoc elimination of redundant model weights based on the posterior estimates under a straightforward (non-hierarchical) generative model. Our comparative study highlights the advantages of the BMR method relative to established approaches, which are based on hierarchical horseshoe priors over model weights. We illustrate the potential of BMR across various deep learning architectures, from classical networks like LeNet to modern frameworks such as Vision Transformers and MLP-Mixers."
                },
                {
                    "title": "Efficiency is Not Enough: A Critical Perspective of Environmentally Sustainable AI",
                    "abstract": "Artificial Intelligence (AI) is currently spearheaded by machine learning (ML) methods such as deep learning (DL) which have accelerated progress on many tasks thought to be out of reach of AI. These ML methods can often be compute hungry, energy intensive, and result in significant carbon emissions, a known driver of anthropogenic climate change. Additionally, the platforms on which ML systems run are associated with environmental impacts including and beyond carbon emissions. The solution lionized by both industry and the ML community to improve the environmental sustainability of ML is to increase the efficiency with which ML systems operate in terms of both compute and energy consumption. In this perspective, we argue that efficiency alone is not enough to make ML as a technology environmentally sustainable. We do so by presenting three high level discrepancies between the effect of efficiency on the environmental sustainability of ML when considering the many variables which it interacts with. In doing so, we comprehensively demonstrate, at multiple levels of granularity both technical and non-technical reasons, why efficiency is not enough to fully remedy the environmental impacts of ML. Based on this, we present and argue for systems thinking as a viable path towards improving the environmental sustainability of ML holistically."
                },
                {
                    "title": "Efficiency Pentathlon: A Standardized Arena for Efficiency Evaluation",
                    "abstract": "Rising computational demands of modern natural language processing (NLP) systems have increased the barrier to entry for cutting-edge research while posing serious environmental concerns. Yet, progress on model efficiency has been impeded by practical challenges in model evaluation and comparison. For example, hardware is challenging to control due to disparate levels of accessibility across different institutions. Moreover, improvements in metrics such as FLOPs often fail to translate to progress in real-world applications. In response, we introduce Pentathlon, a benchmark for holistic and realistic evaluation of model efficiency. Pentathlon focuses on inference, which accounts for a majority of the compute in a model's lifecycle. It offers a strictly-controlled hardware platform, and is designed to mirror real-world applications scenarios. It incorporates a suite of metrics that target different aspects of efficiency, including latency, throughput, memory overhead, and energy consumption. Pentathlon also comes with a software library that can be seamlessly integrated into any codebase and enable evaluation. As a standardized and centralized evaluation platform, Pentathlon can drastically reduce the workload to make fair and reproducible efficiency comparisons. While initially focused on natural language processing (NLP) models, Pentathlon is designed to allow flexible extension to other fields. We envision Pentathlon will stimulate algorithmic innovations in building efficient models, and foster an increased awareness of the social and environmental implications in the development of future-generation NLP models."
                },
                {
                    "title": "Variational Inference for Bayesian Neural Networks under Model and Parameter Uncertainty",
                    "abstract": "Bayesian neural networks (BNNs) have recently regained a significant amount of attention in the deep learning community due to the development of scalable approximate Bayesian inference techniques. There are several advantages of using a Bayesian approach: Parameter and prediction uncertainties become easily available, facilitating rigorous statistical analysis. Furthermore, prior knowledge can be incorporated. However, so far, there have been no scalable techniques capable of combining both structural and parameter uncertainty. In this paper, we apply the concept of model uncertainty as a framework for structural learning in BNNs and hence make inference in the joint space of structures/models and parameters. Moreover, we suggest an adaptation of a scalable variational inference approach with reparametrization of marginal inclusion probabilities to incorporate the model space constraints. Experimental results on a range of benchmark datasets show that we obtain comparable accuracy results with the competing models, but based on methods that are much more sparse than ordinary BNNs."
                },
                {
                    "title": "Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization",
                    "abstract": "Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes."
                },
                {
                    "title": "Compute-Efficient Deep Learning: Algorithmic Trends and Opportunities",
                    "abstract": "Although deep learning has made great progress in recent years, the exploding economic and environmental costs of training neural networks are becoming unsustainable. To address this problem, there has been a great deal of research on *algorithmically-efficient deep learning*, which seeks to reduce training costs not at the hardware or implementation level, but through changes in the semantics of the training program. In this paper, we present a structured and comprehensive overview of the research in this field. First, we formalize the *algorithmic speedup* problem, then we use fundamental building blocks of algorithmically efficient training to develop a taxonomy. Our taxonomy highlights commonalities of seemingly disparate methods and reveals current research gaps. Next, we present evaluation best practices to enable comprehensive, fair, and reliable comparisons of speedup techniques. To further aid research and applications, we discuss common bottlenecks in the training pipeline (illustrated via experiments) and offer taxonomic mitigation strategies for them. Finally, we highlight some unsolved research challenges and present promising future directions."
                },
                {
                    "title": "Efficient Variational Inference for Sparse Deep Learning with Theoretical Guarantee",
                    "abstract": "Sparse deep learning aims to address the challenge of huge storage consumption by deep neural networks, and to recover the sparse structure of target functions. Although tremendous empirical successes have been achieved, most sparse deep learning algorithms are lacking of theoretical support. On the other hand, another line of works have proposed theoretical frameworks that are computationally infeasible. In this paper, we train sparse deep neural networks with a fully Bayesian treatment under spike-and-slab priors, and develop a set of computationally efficient variational inferences via continuous relaxation of Bernoulli distribution. The variational posterior contraction rate is provided, which justifies the consistency of the proposed variational Bayes method. Notably, our empirical results demonstrate that this variational procedure provides uncertainty quantification in terms of Bayesian predictive distribution and is also capable to accomplish consistent variable selection by training a sparse multi-layer neural network."
                },
                {
                    "title": "The Computational Limits of Deep Learning",
                    "abstract": "Deep learning's recent history has been one of achievement: from triumphing over humans in the game of Go to world-leading performance in image recognition, voice recognition, translation, and other tasks. But this progress has come with a voracious appetite for computing power. This article reports on the computational demands of Deep Learning applications in five prominent application areas and shows that progress in all five is strongly reliant on increases in computing power. Extrapolating forward this reliance reveals that progress along current lines is rapidly becoming economically, technically, and environmentally unsustainable. Thus, continued progress in these applications will require dramatically more computationally-efficient methods, which will either have to come from changes to deep learning or from moving to other machine learning methods."
                },
                {
                    "title": "Carbontracker: Tracking and Predicting the Carbon Footprint of Training Deep Learning Models",
                    "abstract": "Deep learning (DL) can achieve impressive results across a wide variety of tasks, but this often comes at the cost of training models for extensive periods on specialized hardware accelerators. This energy-intensive workload has seen immense growth in recent years. Machine learning (ML) may become a significant contributor to climate change if this exponential trend continues. If practitioners are aware of their energy and carbon footprint, then they may actively take steps to reduce it whenever possible. In this work, we present Carbontracker, a tool for tracking and predicting the energy and carbon footprint of training DL models. We propose that energy and carbon footprint of model development and training is reported alongside performance metrics using tools like Carbontracker. We hope this will promote responsible computing in ML and encourage research into energy-efficient deep neural networks."
                },
                {
                    "title": "Visual Transformers: Token-based Image Representation and Processing for Computer Vision",
                    "abstract": "Computer vision has achieved great success using standardized image representations -- pixel arrays, and the corresponding deep learning operators -- convolutions. In this work, we challenge this paradigm: we instead (a) represent images as a set of visual tokens and (b) apply visual transformers to find relationships between visual semantic concepts. Given an input image, we dynamically extract a set of visual tokens from the image to obtain a compact representation for high-level semantics. We then use visual transformers to operate over the visual tokens to densely model relationships between them. We find that this paradigm of token-based image representation and processing drastically outperforms its convolutional counterparts on image classification and semantic segmentation. To demonstrate the power of this approach on ImageNet classification, we use ResNet as a convenient baseline and use visual transformers to replace the last stage of convolutions. This reduces the stage's MACs by up to 6.9x, while attaining up to 4.53 points higher top-1 accuracy. For semantic segmentation, we use a visual-transformer-based FPN (VT-FPN) module to replace a convolution-based FPN, saving 6.5x fewer MACs while achieving up to 0.35 points higher mIoU on LIP and COCO-stuff."
                },
                {
                    "title": "Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning",
                    "abstract": "Accurate reporting of energy and carbon usage is essential for understanding the potential climate impacts of machine learning research. We introduce a framework that makes this easier by providing a simple interface for tracking realtime energy consumption and carbon emissions, as well as generating standardized online appendices. Utilizing this framework, we create a leaderboard for energy efficient reinforcement learning algorithms to incentivize responsible research in this area as an example for other areas of machine learning. Finally, based on case studies using our framework, we propose strategies for mitigation of carbon emissions and reduction of energy consumption. By making accounting easier, we hope to further the sustainable development of machine learning experiments and spur more research into energy efficient algorithms."
                },
                {
                    "title": "Spearman\u2019s rank correlation coefficient",
                    "abstract": "(Redirected from Spearman's rho) In statistics, Spearman's rank correlation coefficient, named after Charles Spearman and often denoted by the Greek letter \u03c1 (rho), is a non-parametric measure of correlation \u2013 that is, it assesses how well an arbitrary monotonic function could describe the relationship between two variables, without making any assumptions about the frequency distribution of the variables. Unlike the Pearson product-moment correlation coefficient, it does not require the assumption that the relationship between the variables is linear, nor does it require the variables to be measured on interval scales; it can be used for variables measured at the ordinal level. In principle, \u03c1 is simply a special case of the Pearson product-moment coefficient in which the data are converted to rankings before calculating the coefficient. In practice, however, a simpler procedure is normally used to calculate \u03c1. The raw scores are converted to ranks, and the differences d between the ranks of each observation on the two variables are calculated. \u03c1 is then given by: where: d i = the difference between each rank of corresponding values of x and y, and n = the number of pairs of values. Spearman's rank correlation coefficient is equivalent to Pearson correlation on ranks. The formula above is a shortcut to its product-moment form, assuming no tie. The product-moment form can be used in both tied and untied cases. Another popular method for computing rank correlation is the Kendall tau rank correlation coefficient."
                },
                {
                    "title": "Constructing Fast Network through Deconstruction of Convolution",
                    "abstract": "Convolutional neural networks have achieved great success in various vision tasks; however, they incur heavy resource costs. By using deeper and wider networks, network accuracy can be improved rapidly. However, in an environment with limited resources (e.g., mobile applications), heavy networks may not be usable. This study shows that naive convolution can be deconstructed into a shift operation and pointwise convolution. To cope with various convolutions, we propose a new shift operation called active shift layer (ASL) that formulates the amount of shift as a learnable function with shift parameters. This new layer can be optimized end-to-end through backpropagation and it can provide optimal shift values. Finally, we apply this layer to a light and fast network that surpasses existing state-of-the-art networks."
                },
                {
                    "title": "Bayesian model reduction",
                    "abstract": "This paper reviews recent developments in statistical structure learning; namely, Bayesian model reduction. Bayesian model reduction is a method for rapidly computing the evidence and parameters of probabilistic models that differ only in their priors. In the setting of variational Bayes this has an analytical solution, which finesses the problem of scoring large model spaces in model comparison or structure learning. In this technical note, we review Bayesian model reduction and provide the relevant equations for several discrete and continuous probability distributions. We provide worked examples in the context of multivariate linear regression, Gaussian mixture models and dynamical systems (dynamic causal modelling). These examples are accompanied by the Matlab scripts necessary to reproduce the results. Finally, we briefly review recent applications in the fields of neuroimaging and neuroscience. Specifically, we consider structure learning and hierarchical or empirical Bayes that can be regarded as a metaphor for neurobiological processes like abductive reasoning."
                },
                {
                    "title": "Active Inference, Curiosity and Insight",
                    "abstract": "This article offers a formal account of curiosity and insight in terms of active (Bayesian) inference. It deals with the dual problem of inferring states of the world and learning its statistical structure. In contrast to current trends in machine learning (e.g., deep learning), we focus on how people attain insight and understanding using just a handful of observations, which are solicited through curious behavior. We use simulations of abstract rule learning and approximate Bayesian inference to show that minimizing (expected) variational free energy leads to active sampling of novel contingencies. This epistemic behavior closes explanatory gaps in generative models of the world, thereby reducing uncertainty and satisfying curiosity. We then move from epistemic learning to model selection or structure learning to show how abductive processes emerge when agents test plausible hypotheses about symmetries (i.e., invariances or rules) in their generative models. The ensuing Bayesian model reduction evinces mechanisms associated with sleep and has all the hallmarks of \u201caha\u201d moments. This formulation moves toward a computational account of consciousness in the pre-Cartesian sense of sharable knowledge (i.e., con: \u201ctogether\u201d; scire: \u201cto know\u201d)."
                },
                {
                    "title": "Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms",
                    "abstract": "We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at this https URL"
                },
                {
                    "title": "Model Selection in Bayesian Neural Networks via Horseshoe Priors",
                    "abstract": "Bayesian Neural Networks (BNNs) have recently received increasing attention for their ability to provide well-calibrated posterior uncertainties. However, model selection---even choosing the number of nodes---remains an open question. In this work, we apply a horseshoe prior over node pre-activations of a Bayesian neural network, which effectively turns off nodes that do not help explain the data. We demonstrate that our prior prevents the BNN from under-fitting even when the number of nodes required is grossly over-estimated. Moreover, this model selection over the number of nodes doesn't come at the expense of predictive or computational performance; in fact, we learn smaller networks with comparable predictive performance to current approaches."
                },
                {
                    "title": "Bayesian Compression for Deep Learning",
                    "abstract": "Compression and computational efficiency in deep learning have become a problem of great significance. In this work, we argue that the most principled and effective way to attack this problem is by adopting a Bayesian point of view, where through sparsity inducing priors we prune large parts of the network. We introduce two novelties in this paper: 1) we use hierarchical priors to prune nodes instead of individual weights, and 2) we use the posterior uncertainties to determine the optimal fixed point precision to encode the weights. Both factors significantly contribute to achieving the state of the art in terms of compression rates, while still staying competitive with methods designed to optimize for speed or energy efficiency."
                },
                {
                    "title": "Structured Bayesian Pruning via Log-Normal Multiplicative Noise",
                    "abstract": "Dropout-based regularization methods can be regarded as injecting random noise with pre-defined magnitude to different parts of the neural network during training. It was recently shown that Bayesian dropout procedure not only improves generalization but also leads to extremely sparse neural architectures by automatically setting the individual noise magnitude per weight. However, this sparsity can hardly be used for acceleration since it is unstructured. In the paper, we propose a new Bayesian model that takes into account the computational structure of neural networks and provides structured sparsity, e.g. removes neurons and/or convolutional channels in CNNs. To do this we inject noise to the neurons outputs while keeping the weights unregularized. We establish the probabilistic model with a proper truncated log-uniform prior over the noise and truncated log-normal variational approximation that ensures that the KL-term in the evidence lower bound is computed in closed-form. The model leads to structured sparsity by removing elements with a low SNR from the computation graph and provides significant acceleration on a number of deep neural architectures. The model is easy to implement as it can be formulated as a separate dropout-like layer."
                },
                {
                    "title": "Pruning Filters for Efficient ConvNets",
                    "abstract": "The success of CNNs in various applications is accompanied by a significant increase in the computation and parameter storage costs. Recent efforts toward reducing these overheads involve pruning and compressing the weights of various layers without hurting original accuracy. However, magnitude-based pruning of weights reduces a significant number of parameters from the fully connected layers and may not adequately reduce the computation costs in the convolutional layers due to irregular sparsity in the pruned networks. We present an acceleration method for CNNs, where we prune filters from CNNs that are identified as having a small effect on the output accuracy. By removing whole filters in the network together with their connecting feature maps, the computation costs are reduced significantly. In contrast to pruning weights, this approach does not result in sparse connectivity patterns. Hence, it does not need the support of sparse convolution libraries and can work with existing efficient BLAS libraries for dense matrix multiplications. We show that even simple filter pruning techniques can reduce inference costs for VGG-16 by up to 34% and ResNet-110 by up to 38% on CIFAR10 while regaining close to the original accuracy by retraining the networks."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Variational Dropout and the Local Reparameterization Trick",
                    "abstract": "We investigate a local reparameterizaton technique for greatly reducing the variance of stochastic gradients for variational Bayesian inference (SGVB) of a posterior over model parameters, while retaining parallelizability. This local reparameterization translates uncertainty about global parameters into local noise that is independent across datapoints in the minibatch. Such parameterizations can be trivially parallelized and have variance that is inversely proportional to the mini-batch size, generally leading to much faster convergence. Additionally, we explore a connection with dropout: Gaussian dropout objectives correspond to SGVB with local reparameterization, a scale-invariant prior and proportionally fixed posterior variance. Our method allows inference of more flexibly parameterized posteriors; specifically, we propose variational dropout, a generalization of Gaussian dropout where the dropout rates are learned, often leading to better models. The method is demonstrated through several experiments."
                },
                {
                    "title": "Second Order Derivatives for Network Pruning: Optimal Brain Surgeon",
                    "abstract": "We investigate the use of information from all second order derivatives of the error function to perform network pruning (i.e., removing unimportant weights from a trained network) in order to improve generalization, simplify networks, reduce hardware or storage requirements, increase the speed of further training, and in some cases enable rule extraction. Our method, Optimal Brain Surgeon (OBS), is Significantly better than magnitude-based methods and Optimal Brain Damage [Le Cun, Denker and Solla, 1990], which often remove the wrong weights. OBS permits the pruning of more weights than other methods (for the same error on the training set), and thus yields better generalization on test data. Crucial to OBS is a recursion relation for calculating the inverse Hessian matrix H-1 from training data and structural information of the net. OBS permits a 90%, a 76%, and a 62% reduction in weights over backpropagation with weight decay on three benchmark MONK's problems [Thrun et al., 1991]. Of OBS, Optimal Brain Damage, and magnitude-based methods, only OBS deletes the correct weights from a trained XOR network in every case. Finally, whereas Sejnowski and Rosenberg [1987] used 18,000 weights in their NETtalk network, we used OBS to prune a network to just 1560 weights, yielding better generalization."
                },
                {
                    "title": "On the distribution of numbers",
                    "abstract": "This paper examines the distribution of the mantissas of floating point numbers and shows how the arithmetic operations of a computer transform various distributions toward the limiting distribution $r(x) = {1 \\over x \\ln b} \\quad (1/b \\leqq x \\leqq 1)$ (where b is the base of the number system). The paper also gives a number of applications to hardware, software, and general computing which show that this distribution is not merely an amusing curiosity. A brief examination of the distribution of exponents is included."
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "The mnist database of handwritten digits",
                    "abstract": "Disclosed is an improved articulated bar flail having shearing edges for efficiently shredding materials. An improved shredder cylinder is disclosed with a plurality of these flails circumferentially spaced and pivotally attached to the periphery of a rotatable shaft. Also disclosed is an improved shredder apparatus which has a pair of these shredder cylinders mounted to rotate about spaced parallel axes which cooperates with a conveyer apparatus which has a pair of inclined converging conveyer belts with one of the belts mounted to move with respect to the other belt to allow the transport of articles of various sizes therethrough."
                },
                {
                    "title": "Variational algorithms for approximate Bayesian inference",
                    "abstract": "The Bayesian framework for machine learning allows for the incorporation of prior knowledge in a coherent way, avoids over\ufb01tting problems, and provides a principled basis for selecting between alternative models. Unfortunately the computations required are usually intractable. This thesis presents a uni\ufb01ed variational Bayesian (VB) framework which approximates these computations in models with latent variables using a lower bound on the marginal likelihood. Chapter 1 presents background material on Bayesian inference, graphical models, and propagation algorithms. Chapter 2 forms the theoretical core of the thesis, generalising the expectation-maximisation (EM) algorithm for learning maximum likelihood parameters to the VB EM al-gorithm which integrates over model parameters. The algorithm is then specialised to the large family of conjugate-exponential (CE) graphical models, and several theorems are presented to pave the road for automated VB derivation procedures in both directed and undirected graphs (Bayesian and Markov networks, respectively)."
                },
                {
                    "title": "Gradient-based learning applied to document recognition",
                    "abstract": "Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day."
                },
                {
                    "title": "Optimal Brain Damage",
                    "abstract": "We have used information-theoretic ideas to derive a class of practical and nearly optimal schemes for adapting the size of a neural network. By removing unimportant weights from a network, several improvements can be expected: better generalization, fewer training examples required, and improved speed of learning and/or classification. The basic idea is to use second-derivative information to make a tradeoff between network complexity and training set error. Experiments confirm the usefulness of the methods on a real-world application."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively balance accuracy and complexity in structured pruning of neural networks to achieve high compression rates without the need for task-specific threshold tuning?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing computational burden associated with modern neural networks, which limits their deployment in resource-constrained environments. By improving the efficiency of neural networks through structured pruning, we can enhance their applicability in real-world scenarios, such as mobile devices and edge computing. This research could lead to advancements in model deployment strategies, enabling faster inference times and reduced energy consumption, which are vital for sustainable AI development. Furthermore, it could inspire future research into more efficient neural architectures and pruning techniques, ultimately contributing to the broader field of machine learning.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge in this problem lies in the inherent trade-offs between maintaining model accuracy and reducing complexity through pruning. Naive approaches may fail because they often do not consider the intricate dependencies between network parameters, leading to significant performance degradation. Additionally, the lack of principled methods for determining which elements to prune complicates the process, as existing techniques may require extensive tuning or may not generalize well across different tasks. Technical obstacles include the need for efficient algorithms that can handle the computational complexity of Bayesian methods while ensuring that the pruning process does not adversely affect the model's performance.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either unstructured or structured pruning methods, often lacking a unified approach that effectively balances accuracy and complexity. Many existing solutions have limitations, such as requiring task-specific threshold tuning or being computationally intensive. Barriers to solving this problem include the complexity of Bayesian model comparison and the challenges associated with inducing sparsity at various structural levels without incurring high computational costs. Our approach, BMRS, differs by integrating Bayesian structured pruning with multiplicative noise and Bayesian model reduction, providing a more principled and efficient framework for structured pruning.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, BMRS, combines Bayesian structured pruning with multiplicative noise and Bayesian model reduction. We will derive two versions of BMRS using different priors: BMRSN, which utilizes a truncated log-normal prior for high compression rates without threshold tuning, and BMRSU, which allows for tun"
            }
        },
        "author_data": {
            "a3d556a3-285e-4994-b380-9bd6dbd4a51e": {
                "pk": "a3d556a3-285e-4994-b380-9bd6dbd4a51e",
                "name": "Dustin Wright",
                "collaborators": [
                    "Isabelle Augenstein",
                    "Andreas Nugaard Holm",
                    "Pepa Atanasova",
                    "Amelie W\u00fchrl",
                    "Roman Klinger",
                    "Jiaxin Pei",
                    "David Jurgens",
                    "Christian Igel",
                    "Gabrielle Samuel",
                    "Raghavendra Selvan"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Information Retrieval",
                    "Fact Checking"
                ],
                "publications": [
                    {
                        "title": "Claim Check-Worthiness Detection as Positive Unlabelled Learning",
                        "abstract": "As the first step of automatic fact checking, claim check-worthiness detection is a critical component of fact checking systems. There are multiple lines of research which study this problem: check-worthiness ranking from political speeches and debates, rumour detection on Twitter, and citation needed detection from Wikipedia. To date, there has been no structured comparison of these various tasks to understand their relatedness, and no investigation into whether or not a unified approach to all of them is achievable. In this work, we illuminate a central challenge in claim check-worthiness detection underlying all of these tasks, being that they hinge upon detecting both how factual a sentence is, as well as how likely a sentence is to be believed without verification. As such, annotators only mark those instances they judge to be clear-cut check-worthy. Our best performing method is a unified approach which automatically corrects for this using a variant of positive unlabelled learning that finds instances which were incorrectly labelled as not check-worthy. In applying this, we out-perform the state of the art in two of the three tasks studied for claim check-worthiness detection in English."
                    },
                    {
                        "title": "Transformer Based Multi-Source Domain Adaptation",
                        "abstract": "In practical machine learning settings, the data on which a model must make predictions often come from a different distribution than the data it was trained on. Here, we investigate the problem of unsupervised multi-source domain adaptation, where a model is trained on labelled data from multiple source domains and must make predictions on a domain for which no labelled data has been seen. Prior work with CNNs and RNNs has demonstrated the benefit of mixture of experts, where the predictions of multiple domain expert classifiers are combined; as well as domain adversarial training, to induce a domain agnostic representation space. Inspired by this, we investigate how such methods can be effectively applied to large pretrained transformer models. We find that domain adversarial training has an effect on the learned representations of these models while having little effect on their performance, suggesting that large transformer-based models are already relatively robust across domains. Additionally, we show that mixture of experts leads to significant performance improvements by comparing several variants of mixing functions, including one novel mixture based on attention. Finally, we demonstrate that the predictions of large pretrained transformer based domain experts are highly homogenous, making it challenging to learn effective functions for mixing their predictions."
                    },
                    {
                        "title": "CiteWorth: Cite-Worthiness Detection for Improved Scientific Document Understanding",
                        "abstract": "Scientific document understanding is challenging as the data is highly domain specific and diverse. However, datasets for tasks with scientific text require expensive manual annotation and tend to be small and limited to only one or a few fields. At the same time, scientific documents contain many potential training signals, such as citations, which can be used to build large labelled datasets. Given this, we present an in-depth study of cite-worthiness detection in English, where a sentence is labelled for whether or not it cites an external source. To accomplish this, we introduce CiteWorth, a large, contextualized, rigorously cleaned labelled dataset for cite-worthiness detection built from a massive corpus of extracted plain-text scientific documents. We show that CiteWorth is high-quality, challenging, and suitable for studying problems such as domain adaptation. Our best performing cite-worthiness detection model is a paragraph-level contextualized sentence labelling model based on Longformer, exhibiting a 5 F1 point improvement over SciBERT which considers only individual sentences. Finally, we demonstrate that language model fine-tuning with cite-worthiness as a secondary task leads to improved performance on downstream scientific document understanding tasks."
                    },
                    {
                        "title": "Semi-Supervised Exaggeration Detection of Health Science Press Releases",
                        "abstract": "Public trust in science depends on honest and factual communication of scientific papers. However, recent studies have demonstrated a tendency of news media to misrepresent scientific papers by exaggerating their findings. Given this, we present a formalization of and study into the problem of exaggeration detection in science communication. While there are an abundance of scientific papers and popular media articles written about them, very rarely do the articles include a direct link to the original paper, making data collection challenging. We address this by curating a set of labeled press release/abstract pairs from existing expert annotated studies on exaggeration in press releases of scientific papers suitable for benchmarking the performance of machine learning models on the task. Using limited data from this and previous studies on exaggeration detection in science, we introduce MT-PET, a multi-task version of Pattern Exploiting Training (PET), which leverages knowledge from complementary cloze-style QA tasks to improve few-shot learning. We demonstrate that MT-PET outperforms PET and supervised learning both when data is limited, as well as when there is an abundance of data for the main task."
                    },
                    {
                        "title": "Multi-View Knowledge Distillation from Crowd Annotations for Out-of-Domain Generalization",
                        "abstract": "Selecting an effective training signal for tasks in natural language processing is difficult: expert annotations are expensive, and crowd-sourced annotations may not be reliable. At the same time, recent work in NLP has demonstrated that learning from a distribution over labels acquired from crowd annotations can be effective. However, there are many ways to acquire such a distribution, and the performance allotted by any one method can fluctuate based on the task and the amount of available crowd annotations, making it difficult to know a priori which distribution is best. This paper systematically analyzes this in the out-of-domain setting, adding to the NLP literature which has focused on in-domain evaluation, and proposes new methods for acquiring soft-labels from crowd-annotations by aggregating the distributions produced by existing methods. In particular, we propose to aggregate multiple-views of crowd annotations via temperature scaling and finding their Jensen-Shannon centroid. We demonstrate that these aggregation methods lead to the most consistent performance across four NLP tasks on out-of-domain test sets, mitigating fluctuations in performance from the individual distributions. Additionally, aggregation results in the most consistently well-calibrated uncertainty estimation. We argue that aggregating different views of crowd-annotations is an effective and minimal intervention to acquire soft-labels which induce robust classifiers despite the inconsistency of the individual soft-labeling methods."
                    },
                    {
                        "title": "Generating Label Cohesive and Well-Formed Adversarial Claims",
                        "abstract": "Adversarial attacks reveal important vulnerabilities and flaws of trained models. One potent type of attack are universal adversarial triggers, which are individual n-grams that, when appended to instances of a class under attack, can trick a model into predicting a target class. However, for inference tasks such as fact checking, these triggers often inadvertently invert the meaning of instances they are inserted in. In addition, such attacks produce semantically nonsensical inputs, as they simply concatenate triggers to existing samples. Here, we investigate how to generate adversarial attacks against fact checking systems that preserve the ground truth meaning and are semantically valid. We extend the HotFlip attack algorithm used for universal trigger generation by jointly minimising the target class loss of a fact checking model and the entailment class loss of an auxiliary natural language inference model. We then train a conditional language model to generate semantically valid statements, which include the found universal triggers. We find that the generated attacks maintain the directionality and semantic validity of the claim better than previous work."
                    },
                    {
                        "title": "Revisiting Softmax for Uncertainty Approximation in Text Classification",
                        "abstract": "Uncertainty approximation in text classification is an important area with applications in domain adaptation and interpretability. One of the most widely used uncertainty approximation methods is Monte Carlo (MC) Dropout, which is computationally expensive as it requires multiple forward passes through the model. A cheaper alternative is to simply use the softmax based on a single forward pass without dropout to estimate model uncertainty. However, prior work has indicated that these predictions tend to be overconfident. In this paper, we perform a thorough empirical analysis of these methods on five datasets with two base neural architectures in order to identify the trade-offs between the two. We compare both softmax and an efficient version of MC Dropout on their uncertainty approximations and downstream text classification performance, while weighing their runtime (cost) against performance (benefit). We find that, while MC dropout produces the best uncertainty approximations, using a simple softmax leads to competitive and in some cases better uncertainty estimation for text classification at a much lower computational cost, suggesting that softmax can in fact be a sufficient uncertainty estimate when computational resources are a concern."
                    },
                    {
                        "title": "Understanding Fine-grained Distortions in Reports of Scientific Findings",
                        "abstract": "Distorted science communication harms individuals and society as it can lead to unhealthy behavior change and decrease trust in scientific institutions. Given the rapidly increasing volume of science communication in recent years, a fine-grained understanding of how findings from scientific publications are reported to the general public, and methods to detect distortions from the original work automatically, are crucial. Prior work focused on individual aspects of distortions or worked with unpaired data. In this work, we make three foundational contributions towards addressing this problem: (1) annotating 1,600 instances of scientific findings from academic papers paired with corresponding findings as reported in news articles and tweets wrt. four characteristics: causality, certainty, generality and sensationalism; (2) establishing baselines for automatically detecting these characteristics; and (3) analyzing the prevalence of changes in these characteristics in both human-annotated and large-scale unlabeled data. Our results show that scientific findings frequently undergo subtle distortions when reported. Tweets distort findings more often than science news reports. Detecting fine-grained distortions automatically poses a challenging task. In our experiments, fine-tuned task-specific models consistently outperform few-shot LLM prompting."
                    },
                    {
                        "title": "Modeling Information Change in Science Communication with Semantically Matched Paraphrases",
                        "abstract": "Whether the media faithfully communicate scientific information has long been a core issue to the science community. Automatically identifying paraphrased scientific findings could enable large-scale tracking and analysis of information changes in the science communication process, but this requires systems to understand the similarity between scientific information across multiple domains. To this end, we present the SCIENTIFIC PARAPHRASE AND INFORMATION CHANGE DATASET (SPICED), the first paraphrase dataset of scientific findings annotated for degree of information change. SPICED contains 6,000 scientific finding pairs extracted from news stories, social media discussions, and full texts of original papers. We demonstrate that SPICED poses a challenging task and that models trained on SPICED improve downstream performance on evidence retrieval for fact checking of real-world scientific claims. Finally, we show that models trained on SPICED can reveal large-scale trends in the degrees to which people and organizations faithfully communicate new scientific findings. Data, code, and pre-trained models are available at http://www.copenlu.com/publication/2022_emnlp_wright/."
                    },
                    {
                        "title": "Efficiency is Not Enough: A Critical Perspective of Environmentally Sustainable AI",
                        "abstract": "Artificial Intelligence (AI) is currently spearheaded by machine learning (ML) methods such as deep learning (DL) which have accelerated progress on many tasks thought to be out of reach of AI. These ML methods can often be compute hungry, energy intensive, and result in significant carbon emissions, a known driver of anthropogenic climate change. Additionally, the platforms on which ML systems run are associated with environmental impacts including and beyond carbon emissions. The solution lionized by both industry and the ML community to improve the environmental sustainability of ML is to increase the efficiency with which ML systems operate in terms of both compute and energy consumption. In this perspective, we argue that efficiency alone is not enough to make ML as a technology environmentally sustainable. We do so by presenting three high level discrepancies between the effect of efficiency on the environmental sustainability of ML when considering the many variables which it interacts with. In doing so, we comprehensively demonstrate, at multiple levels of granularity both technical and non-technical reasons, why efficiency is not enough to fully remedy the environmental impacts of ML. Based on this, we present and argue for systems thinking as a viable path towards improving the environmental sustainability of ML holistically."
                    },
                    {
                        "title": "Rethinking Recurrent Latent Variable Model for Music Composition",
                        "abstract": "We present a model for capturing musical features and creating novel sequences of music, called the Convolutional Variational Recurrent Neural Network. To generate sequential data, the model uses an encoder-decoder architecture with latent probabilistic connections to capture the hidden structure of music. Using the sequence-to-sequence model, our generative model can exploit samples from a prior distribution and generate a longer sequence of music. We compare the performance of our proposed model with other types of Neural Networks using the criteria of Information Rate that is implemented by Variable Markov Oracle, a method that allows statistical characterization of musical information dynamics and detection of motifs in a song. Our results suggest that the proposed model has a better statistical resemblance to the musical structure of the training data, which improves the creation of new sequences of music in the style of the originals."
                    },
                    {
                        "title": "Longitudinal Citation Prediction using Temporal Graph Neural Networks",
                        "abstract": "Citation count prediction is the task of predicting the number of citations a paper has gained after a period of time. Prior work viewed this as a static prediction task. As papers and their citations evolve over time, considering the dynamics of the number of citations a paper will receive would seem logical. Here, we introduce the task of sequence citation prediction. The goal is to accurately predict the trajectory of the number of citations a scholarly work receives over time. We propose to view papers as a structured network of citations, allowing us to use topological information as a learning signal. Additionally, we learn how this dynamic citation network changes over time and the impact of paper meta-data such as authors, venues and abstracts. To approach the new task, we derive a dynamic citation network from Semantic Scholar spanning over 42 years. We present a model which exploits topological and temporal information using graph convolution networks paired with sequence prediction, and compare it against multiple baselines, testing the importance of topological and temporal information and analyzing model performance. Our experiments show that leveraging both the temporal and topological information greatly increases the performance of predicting citation counts over time."
                    }
                ]
            },
            "34d3dc97-cc92-4e39-8fed-c19aae242636": {
                "pk": "34d3dc97-cc92-4e39-8fed-c19aae242636",
                "name": "Christian Igel",
                "collaborators": [
                    "Fabian Gieseke",
                    "Marc Toussaint",
                    "Stefan Oehmcke",
                    "Oswin Krause",
                    "Asja Fischer",
                    "Nico Lang",
                    "Nick Hauptvogel",
                    "Kai Br\u00fcgge",
                    "Philip Enevoldsen",
                    "Christian Gundersen"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Bayesian Methods",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Smooth Min-Max Monotonic Networks",
                        "abstract": "Monotonicity constraints are powerful regularizers in statistical modelling. They can support fairness in computer-aided decision making and increase plausibility in data-driven scientific models. The seminal min-max (MM) neural network architecture ensures monotonicity, but often gets stuck in undesired local optima during training because of partial derivatives of the MM nonlinearities being zero. We propose a simple modification of the MM network using strictly-increasing smooth minimum and maximum functions that alleviates this problem. The resulting smooth min-max (SMM) network module inherits the asymptotic approximation properties from the MM architecture. It can be used within larger deep learning systems trained end-to-end. The SMM module is conceptually simple and computationally less demanding than state-of-the-art neural networks for monotonic modelling. Our experiments show that this does not come with a loss in generalization performance compared to alternative neural and non-neural approaches."
                    },
                    {
                        "title": "Neutrality: A Necessity for Self-Adaptation",
                        "abstract": "Self-adaptation is used in all main paradigms of evolutionary computation to increase efficiency. We claim that the basis of self-adaptation is the use of neutrality. In the absence of external control neutrality allows a variation of the search distribution without the risk of fitness loss."
                    },
                    {
                        "title": "On Classes of Functions for which No Free Lunch Results Hold",
                        "abstract": "In a recent paper it was shown that No Free Lunch results hold for any subset F of the set of all possible functions from a finite set X to a finite set Y iff F is closed under permutation of X. In this article, we prove that the number of those subsets can be neglected compared to the overall number of possible subsets. Further, we present some arguments why problem classes relevant in practice are not likely to be closed under permutation."
                    },
                    {
                        "title": "Remember to correct the bias when using deep learning for regression!",
                        "abstract": "When training deep learning models for least-squares regression, we cannot expect that the training error residuals of the final model, selected after a fixed training time or based on performance on a hold-out data set, sum to zero. This can introduce a systematic error that accumulates if we are interested in the total aggregated performance over many data points. We suggest to adjust the bias of the machine learning model after training as a default postprocessing step, which efficiently solves the problem. The severeness of the error accumulation and the effectiveness of the bias correction is demonstrated in exemplary experiments."
                    },
                    {
                        "title": "Bayesian vs. PAC-Bayesian Deep Neural Network Ensembles",
                        "abstract": "Bayesian neural networks address epistemic uncertainty by learning a posterior distribution over model parameters. Sampling and weighting networks according to this posterior yields an ensemble model referred to as Bayes ensemble. Ensembles of neural networks (deep ensembles) can profit from the cancellation of errors effect: Errors by ensemble members may average out and the deep ensemble achieves better predictive performance than each individual network. We argue that neither the sampling nor the weighting in a Bayes ensemble are particularly well-suited for increasing generalization performance, as they do not support the cancellation of errors effect, which is evident in the limit from the Bernstein-von~Mises theorem for misspecified models. In contrast, a weighted average of models where the weights are optimized by minimizing a PAC-Bayesian generalization bound can improve generalization performance. This requires that the optimization takes correlations between models into account, which can be achieved by minimizing the tandem loss at the cost that hold-out data for estimating error correlations need to be available. The PAC-Bayesian weighting increases the robustness against correlated models and models with lower performance in an ensemble. This allows us to safely add several models from the same learning process to an ensemble, instead of using early-stopping for selecting a single weight configuration. Our study presents empirical results supporting these conceptual considerations on four different classification datasets. We show that state-of-the-art Bayes ensembles from the literature, despite being computationally demanding, do not improve over simple uniformly weighted deep ensembles and cannot match the performance of deep ensembles weighted by optimizing the tandem loss, which additionally come with non-vacuous generalization guarantees."
                    },
                    {
                        "title": "On the convergence of the Metropolis algorithm with fixed-order updates for multivariate binary probability distributions",
                        "abstract": "The Metropolis algorithm is arguably the most fundamental Markov chain Monte Carlo (MCMC) method. But the algorithm is not guaranteed to converge to the desired distribution in the case of multivariate binary distributions (e.g., Ising models or stochastic neural networks such as Boltzmann machines) if the variables (sites or neurons) are updated in a fixed order, a setting commonly used in practice. The reason is that the corresponding Markov chain may not be irreducible. We propose a modified Metropolis transition operator that behaves almost always identically to the standard Metropolis operator and prove that it ensures irreducibility and convergence to the limiting distribution in the multivariate binary case with fixed-order updates. The result provides an explanation for the behaviour of Metropolis MCMC in that setting and closes a long-standing theoretical gap. We experimentally studied the standard and modified Metropolis operator for models were they actually behave differently. If the standard algorithm also converges, the modified operator exhibits similar (if not better) performance in terms of convergence speed."
                    },
                    {
                        "title": "Recent Results on No-Free-Lunch Theorems for Optimization",
                        "abstract": "The sharpened No-Free-Lunch-theorem (NFL-theorem) states that the performance of all optimization algorithms averaged over any finite set F of functions is equal if and only if F is closed under permutation (c.u.p.) and each target function in F is equally likely. In this paper, we first summarize some consequences of this theorem, which have been proven recently: The average number of evaluations needed to find a desirable (e.g., optimal) solution can be calculated; the number of subsets c.u.p. can be neglected compared to the overall number of possible subsets; and problem classes relevant in practice are not likely to be c.u.p. Second, as the main result, the NFL-theorem is extended. Necessary and sufficient conditions for NFL-results to hold are given for arbitrary, non-uniform distributions of target functions. This yields the most general NFL-theorem for optimization presented so far."
                    },
                    {
                        "title": "Training Big Random Forests with Little Resources",
                        "abstract": "Without access to large compute clusters, building random forests on large datasets is still a challenging problem. This is, in particular, the case if fully-grown trees are desired. We propose a simple yet effective framework that allows to efficiently construct ensembles of huge trees for hundreds of millions or even billions of training instances using a cheap desktop computer with commodity hardware. The basic idea is to consider a multi-level construction scheme, which builds top trees for small random subsets of the available data and which subsequently distributes all training instances to the top trees' leaves for further processing. While being conceptually simple, the overall efficiency crucially depends on the particular implementation of the different phases. The practical merits of our approach are demonstrated using dense datasets with hundreds of millions of training instances."
                    },
                    {
                        "title": "Population-Contrastive-Divergence: Does Consistency help with RBM training?",
                        "abstract": "Estimating the log-likelihood gradient with respect to the parameters of a Restricted Boltzmann Machine (RBM) typically requires sampling using Markov Chain Monte Carlo (MCMC) techniques. To save computation time, the Markov chains are only run for a small number of steps, which leads to a biased estimate. This bias can cause RBM training algorithms such as Contrastive Divergence (CD) learning to deteriorate. We adopt the idea behind Population Monte Carlo (PMC) methods to devise a new RBM training algorithm termed Population-Contrastive-Divergence (pop-CD). Compared to CD, it leads to a consistent estimate and may have a significantly lower bias. Its computational overhead is negligible compared to CD. However, the variance of the gradient estimate increases. We experimentally show that pop-CD can significantly outperform CD. In many cases, we observed a smaller bias and achieved higher log-likelihood values. However, when the RBM distribution has many hidden neurons, the consistent estimate of pop-CD may still have a considerable bias and the variance of the gradient estimate requires a smaller learning rate. Thus, despite its superior theoretical properties, it is not advisable to use pop-CD in its current form on large problems."
                    },
                    {
                        "title": "Familiarity-Based Open-Set Recognition Under Adversarial Attacks",
                        "abstract": "Open-set recognition (OSR), the identification of novel categories, can be a critical component when deploying classification models in real-world applications. Recent work has shown that familiarity-based scoring rules such as the Maximum Softmax Probability (MSP) or the Maximum Logit Score (MLS) are strong baselines when the closed-set accuracy is high. However, one of the potential weaknesses of familiarity-based OSR are adversarial attacks. Here, we present gradient-based adversarial attacks on familiarity scores for both types of attacks, False Familiarity and False Novelty attacks, and evaluate their effectiveness in informed and uninformed settings on TinyImageNet."
                    },
                    {
                        "title": "Big Universe, Big Data: Machine Learning and Image Analysis for Astronomy",
                        "abstract": "Astrophysics and cosmology are rich with data. The advent of wide-area digital cameras on large aperture telescopes has led to ever more ambitious surveys of the sky. Data volumes of entire surveys a decade ago can now be acquired in a single night and real-time analysis is often desired. Thus, modern astronomy requires big data know-how, in particular it demands highly efficient machine learning and image analysis algorithms. But scalability is not the only challenge: Astronomy applications touch several current machine learning research questions, such as learning from biased data and dealing with label and measurement noise. We argue that this makes astronomy a great domain for computer science research, as it pushes the boundaries of data analysis. In the following, we will present this exciting application area for data scientists. We will focus on exemplary results, discuss main challenges, and highlight some recent methodological advancements in machine learning and image analysis triggered by astronomical applications."
                    },
                    {
                        "title": "A Loss Function for Generative Neural Networks Based on Watson's Perceptual Model",
                        "abstract": "To train Variational Autoencoders (VAEs) to generate realistic imagery requires a loss function that reflects human perception of image similarity. We propose such a loss function based on Watson's perceptual model, which computes a weighted distance in frequency space and accounts for luminance and contrast masking. We extend the model to color images, increase its robustness to translation by using the Fourier Transform, remove artifacts due to splitting the image into blocks, and make it differentiable. In experiments, VAEs trained with the new loss function generated realistic, high-quality image samples. Compared to using the Euclidean distance and the Structural Similarity Index, the images were less blurry; compared to deep neural network based losses, the new approach required less computational resources and generated images with less artifacts."
                    },
                    {
                        "title": "A Strongly Quasiconvex PAC-Bayesian Bound",
                        "abstract": "We propose a new PAC-Bayesian bound and a way of constructing a hypothesis space, so that the bound is convex in the posterior distribution and also convex in a trade-off parameter between empirical performance of the posterior distribution and its complexity. The complexity is measured by the Kullback-Leibler divergence to a prior. We derive an alternating procedure for minimizing the bound. We show that the bound can be rewritten as a one-dimensional function of the trade-off parameter and provide sufficient conditions under which the function has a single global minimum. When the conditions are satisfied the alternating minimization is guaranteed to converge to the global minimum of the bound. We provide experimental results demonstrating that rigorous minimization of the bound is competitive with cross-validation in tuning the trade-off between complexity and empirical performance. In all our experiments the trade-off turned to be quasiconvex even when the sufficient conditions were violated."
                    },
                    {
                        "title": "Multimodal Variational Autoencoders for Semi-Supervised Learning: In Defense of Product-of-Experts",
                        "abstract": "Multimodal generative models should be able to learn a meaningful latent representation that enables a coherent joint generation of all modalities (e.g., images and text). Many applications also require the ability to accurately sample modalities conditioned on observations of a subset of the modalities. Often not all modalities may be observed for all training data points, so semi-supervised learning should be possible. In this study, we propose a novel product-of-experts (PoE) based variational autoencoder that have these desired properties. We benchmark it against a mixture-of-experts (MoE) approach and an approach of combining the modalities with an additional encoder network. An empirical evaluation shows that the PoE based models can outperform the contrasted models. Our experiments support the intuition that PoE models are more suited for a conjunctive combination of modalities."
                    },
                    {
                        "title": "BuildSeg: A General Framework for the Segmentation of Buildings",
                        "abstract": "Building segmentation from aerial images and 3D laser scanning (LiDAR) is a challenging task due to the diversity of backgrounds, building textures, and image quality. While current research using different types of convolutional and transformer networks has considerably improved the performance on this task, even more accurate segmentation methods for buildings are desirable for applications such as automatic mapping. In this study, we propose a general framework termed \\emph{BuildSeg} employing a generic approach that can be quickly applied to segment buildings. Different data sources were combined to increase generalization performance. The approach yields good results for different data sources as shown by experiments on high-resolution multi-spectral and LiDAR imagery of cities in Norway, Denmark and France. We applied ConvNeXt and SegFormer based models on the high resolution aerial image dataset from the MapAI-competition. The methods achieved an IOU of 0.7902 and a boundary IOU of 0.6185. We used post-processing to account for the rectangular shape of the objects. This increased the boundary IOU from 0.6185 to 0.6189."
                    },
                    {
                        "title": "EC-NAS: Energy Consumption Aware Tabular Benchmarks for Neural Architecture Search",
                        "abstract": "Energy consumption from the selection, training, and deployment of deep learning models has seen a significant uptick recently. This work aims to facilitate the design of energy-efficient deep learning models that require less computational resources and prioritize environmental sustainability by focusing on the energy consumption. Neural architecture search (NAS) benefits from tabular benchmarks, which evaluate NAS strategies cost-effectively through precomputed performance statistics. We advocate for including energy efficiency as an additional performance criterion in NAS. To this end, we introduce an enhanced tabular benchmark encompassing data on energy consumption for varied architectures. The benchmark, designated as EC-NAS, has been made available in an open-source format to advance research in energy-conscious NAS. EC-NAS incorporates a surrogate model to predict energy consumption, aiding in diminishing the energy expenditure of the dataset creation. Our findings emphasize the potential of EC-NAS by leveraging multi-objective optimization algorithms, revealing a balance between energy usage and accuracy. This suggests the feasibility of identifying energy-lean architectures with little or no compromise in performance."
                    },
                    {
                        "title": "Nacala-Roof-Material: Drone Imagery for Roof Detection, Classification, and Segmentation to Support Mosquito-borne Disease Risk Assessment",
                        "abstract": "As low-quality housing and in particular certain roof characteristics are associated with an increased risk of malaria, classification of roof types based on remote sensing imagery can support the assessment of malaria risk and thereby help prevent the disease. To support research in this area, we release the Nacala-Roof-Material dataset, which contains high-resolution drone images from Mozambique with corresponding labels delineating houses and specifying their roof types. The dataset defines a multi-task computer vision problem, comprising object detection, classification, and segmentation. In addition, we benchmarked various state-of-the-art approaches on the dataset. Canonical U-Nets, YOLOv8, and a custom decoder on pretrained DINOv2 served as baselines. We show that each of the methods has its advantages but none is superior on all tasks, which highlights the potential of our dataset for future research in multi-task learning. While the tasks are closely related, accurate segmentation of objects does not necessarily imply accurate instance separation, and vice versa. We address this general issue by introducing a variant of the deep ordinal watershed (DOW) approach that additionally separates the interior of objects, allowing for improved object delineation and separation. We show that our DOW variant is a generic approach that improves the performance of both U-Net and DINOv2 backbones, leading to a better trade-off between semantic segmentation and instance segmentation."
                    },
                    {
                        "title": "Bigger Buffer k-d Trees on Multi-Many-Core Systems",
                        "abstract": "A buffer k-d tree is a k-d tree variant for massively-parallel nearest neighbor search. While providing valuable speed-ups on modern many-core devices in case both a large number of reference and query points are given, buffer k-d trees are limited by the amount of points that can fit on a single device. In this work, we show how to modify the original data structure and the associated workflow to make the overall approach capable of dealing with massive data sets. We further provide a simple yet efficient way of using multiple devices given in a single workstation. The applicability of the modified framework is demonstrated in the context of astronomy, a field that is faced with huge amounts of data."
                    },
                    {
                        "title": "Do End-to-End Speech Recognition Models Care About Context?",
                        "abstract": "The two most common paradigms for end-to-end speech recognition are connectionist temporal classification (CTC) and attention-based encoder-decoder (AED) models. It has been argued that the latter is better suited for learning an implicit language model. We test this hypothesis by measuring temporal context sensitivity and evaluate how the models perform when we constrain the amount of contextual information in the audio input. We find that the AED model is indeed more context sensitive, but that the gap can be closed by adding self-attention to the CTC model. Furthermore, the two models perform similarly when contextual information is constrained. Finally, in contrast to previous research, our results show that the CTC model is highly competitive on WSJ and LibriSpeech without the help of an external language model."
                    }
                ]
            },
            "1b50ba9c-c1e3-4ccd-9e90-af43792764c5": {
                "pk": "1b50ba9c-c1e3-4ccd-9e90-af43792764c5",
                "name": "Raghavendra Selvan",
                "collaborators": [
                    "Erik B Dam",
                    "Jens Petersen",
                    "Marleen de Bruijne",
                    "Silas \u00d8rting",
                    "Julian Sch\u00f6n",
                    "Christian Igel",
                    "Jesper H. Pedersen",
                    "Sebastian Eliassen",
                    "Klas Rydhmer",
                    "Nicklas Boserup"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Medical Image Analysis",
                    "Self-Supervised Learning",
                    "Energy Efficiency"
                ],
                "publications": [
                    {
                        "title": "Activation Compression of Graph Neural Networks using Block-wise Quantization with Improved Variance Minimization",
                        "abstract": "Efficient training of large-scale graph neural networks (GNNs) has been studied with a specific focus on reducing their memory consumption. Work by Liu et al. (2022) proposed extreme activation compression (EXACT) which demonstrated drastic reduction in memory consumption by performing quantization of the intermediate activation maps down to using INT2 precision. They showed little to no reduction in performance while achieving large reductions in GPU memory consumption. In this work, we present an improvement to the EXACT strategy by using block-wise quantization of the intermediate activation maps. We experimentally analyze different block sizes and show further reduction in memory consumption (>15%), and runtime speedup per epoch (about 5%) even when performing extreme extents of quantization with similar performance trade-offs as with the original EXACT. Further, we present a correction to the assumptions on the distribution of intermediate activation maps in EXACT (assumed to be uniform) and show improved variance estimations of the quantization and dequantization steps."
                    },
                    {
                        "title": "Dynamic $\u03b2$-VAEs for quantifying biodiversity by clustering optically recorded insect signals",
                        "abstract": "While insects are the largest and most diverse group of terrestrial animals, constituting ca. 80% of all known species, they are difficult to study due to their small size and similarity between species. Conventional monitoring techniques depend on time consuming trapping methods and tedious microscope-based work by skilled experts in order to identify the caught insect specimen at species, or even family level. Researchers and policy makers are in urgent need of a scalable monitoring tool in order to conserve biodiversity and secure human food production due to the rapid decline in insect numbers.   In order to improve upon existing insect clustering methods, we propose an adaptive variant of the variational autoencoder (VAE) which is capable of clustering data by phylogenetic groups. The proposed dynamic beta-VAE dynamically adapts the scaling of the reconstruction and regularization loss terms (beta value) yielding useful latent representations of the input data. We demonstrate the usefulness of the dynamic beta-VAE on optically recorded insect signals from regions of southern Scandinavia to cluster unlabelled targets into possible species. We also demonstrate improved clustering performance in a semi-supervised setting using a small subset of labelled data. These experimental results, in both unsupervised- and semi-supervised settings, with the dynamic beta-VAE are promising and, in the near future, can be deployed to monitor insects and conserve the rapidly declining insect biodiversity."
                    },
                    {
                        "title": "Efficient Self-Supervision using Patch-based Contrastive Learning for Histopathology Image Segmentation",
                        "abstract": "Learning discriminative representations of unlabelled data is a challenging task. Contrastive self-supervised learning provides a framework to learn meaningful representations using learned notions of similarity measures from simple pretext tasks. In this work, we propose a simple and efficient framework for self-supervised image segmentation using contrastive learning on image patches, without using explicit pretext tasks or any further labeled fine-tuning. A fully convolutional neural network (FCNN) is trained in a self-supervised manner to discern features in the input images and obtain confidence maps which capture the network's belief about the objects belonging to the same class. Positive- and negative- patches are sampled based on the average entropy in the confidence maps for contrastive learning. Convergence is assumed when the information separation between the positive patches is small, and the positive-negative pairs is large. The proposed model only consists of a simple FCNN with 10.8k parameters and requires about 5 minutes to converge on the high resolution microscopy datasets, which is orders of magnitude smaller than the relevant self-supervised methods to attain similar performance. We evaluate the proposed method for the task of segmenting nuclei from two histopathology datasets, and show comparable performance with relevant self-supervised and supervised methods."
                    },
                    {
                        "title": "Is Adversarial Training with Compressed Datasets Effective?",
                        "abstract": "Dataset Condensation (DC) refers to the recent class of dataset compression methods that generate a smaller, synthetic, dataset from a larger dataset. This synthetic dataset retains the essential information of the original dataset, enabling models trained on it to achieve performance levels comparable to those trained on the full dataset. Most current DC methods have mainly concerned with achieving high test performance with limited data budget, and have not directly addressed the question of adversarial robustness. In this work, we investigate the impact of adversarial robustness on models trained with compressed datasets. We show that the compressed datasets obtained from DC methods are not effective in transferring adversarial robustness to models. As a solution to improve dataset compression efficiency and adversarial robustness simultaneously, we propose a novel robustness-aware dataset compression method based on finding the Minimal Finite Covering (MFC) of the dataset. The proposed method is (1) obtained by one-time computation and is applicable for any model, (2) more effective than DC methods when applying adversarial training over MFC, (3) provably robust by minimizing the generalized adversarial loss. Additionally, empirical evaluation on three datasets shows that the proposed method is able to achieve better robustness and performance trade-off compared to DC methods such as distribution matching."
                    },
                    {
                        "title": "Multi-layered tensor networks for image classification",
                        "abstract": "The recently introduced locally orderless tensor network (LoTeNet) for supervised image classification uses matrix product state (MPS) operations on grids of transformed image patches. The resulting patch representations are combined back together into the image space and aggregated hierarchically using multiple MPS blocks per layer to obtain the final decision rules. In this work, we propose a non-patch based modification to LoTeNet that performs one MPS operation per layer, instead of several patch-level operations. The spatial information in the input images to MPS blocks at each layer is squeezed into the feature dimension, similar to LoTeNet, to maximise retained spatial correlation between pixels when images are flattened into 1D vectors. The proposed multi-layered tensor network (MLTN) is capable of learning linear decision boundaries in high dimensional spaces in a multi-layered setting, which results in a reduction in the computation cost compared to LoTeNet without any degradation in performance."
                    },
                    {
                        "title": "Operating critical machine learning models in resource constrained regimes",
                        "abstract": "The accelerated development of machine learning methods, primarily deep learning, are causal to the recent breakthroughs in medical image analysis and computer aided intervention. The resource consumption of deep learning models in terms of amount of training data, compute and energy costs are known to be massive. These large resource costs can be barriers in deploying these models in clinics, globally. To address this, there are cogent efforts within the machine learning community to introduce notions of resource efficiency. For instance, using quantisation to alleviate memory consumption. While most of these methods are shown to reduce the resource utilisation, they could come at a cost in performance. In this work, we probe into the trade-off between resource consumption and performance, specifically, when dealing with models that are used in critical settings such as in clinics."
                    },
                    {
                        "title": "Segmenting two-dimensional structures with strided tensor networks",
                        "abstract": "Tensor networks provide an efficient approximation of operations involving high dimensional tensors and have been extensively used in modelling quantum many-body systems. More recently, supervised learning has been attempted with tensor networks, primarily focused on tasks such as image classification. In this work, we propose a novel formulation of tensor networks for supervised image segmentation which allows them to operate on high resolution medical images. We use the matrix product state (MPS) tensor network on non-overlapping patches of a given input image to predict the segmentation mask by learning a pixel-wise linear classification rule in a high dimensional space. The proposed model is end-to-end trainable using backpropagation. It is implemented as a Strided Tensor Network to reduce the parameter complexity. The performance of the proposed method is evaluated on two public medical imaging datasets and compared to relevant baselines. The evaluation shows that the strided tensor network yields competitive performance compared to CNN-based models while using fewer resources. Additionally, based on the experiments we discuss the feasibility of using fully linear models for segmentation tasks."
                    },
                    {
                        "title": "Uncertainty quantification in medical image segmentation with normalizing flows",
                        "abstract": "Medical image segmentation is inherently an ambiguous task due to factors such as partial volumes and variations in anatomical definitions. While in most cases the segmentation uncertainty is around the border of structures of interest, there can also be considerable inter-rater differences. The class of conditional variational autoencoders (cVAE) offers a principled approach to inferring distributions over plausible segmentations that are conditioned on input images. Segmentation uncertainty estimated from samples of such distributions can be more informative than using pixel level probability scores. In this work, we propose a novel conditional generative model that is based on conditional Normalizing Flow (cFlow). The basic idea is to increase the expressivity of the cVAE by introducing a cFlow transformation step after the encoder. This yields improved approximations of the latent posterior distribution, allowing the model to capture richer segmentation variations. With this we show that the quality and diversity of samples obtained from our conditional generative model is enhanced. Performance of our model, which we call cFlow Net, is evaluated on two medical imaging datasets demonstrating substantial improvements in both qualitative and quantitative measures when compared to a recent cVAE based model."
                    },
                    {
                        "title": "Tensor Networks for Medical Image Classification",
                        "abstract": "With the increasing adoption of machine learning tools like neural networks across several domains, interesting connections and comparisons to concepts from other domains are coming to light. In this work, we focus on the class of Tensor Networks, which has been a work horse for physicists in the last two decades to analyse quantum many-body systems. Building on the recent interest in tensor networks for machine learning, we extend the Matrix Product State tensor networks (which can be interpreted as linear classifiers operating in exponentially high dimensional spaces) to be useful in medical image analysis tasks. We focus on classification problems as a first step where we motivate the use of tensor networks and propose adaptions for 2D images using classical image domain concepts such as local orderlessness of images. With the proposed locally orderless tensor network model (LoTeNet), we show that tensor networks are capable of attaining performance that is comparable to state-of-the-art deep learning methods. We evaluate the model on two publicly available medical imaging datasets and show performance improvements with fewer model hyperparameters and lesser computational resources compared to relevant baseline methods."
                    },
                    {
                        "title": "Locally orderless tensor networks for classifying two- and three-dimensional medical images",
                        "abstract": "Tensor networks are factorisations of high rank tensors into networks of lower rank tensors and have primarily been used to analyse quantum many-body problems. Tensor networks have seen a recent surge of interest in relation to supervised learning tasks with a focus on image classification. In this work, we improve upon the matrix product state (MPS) tensor networks that can operate on one-dimensional vectors to be useful for working with 2D and 3D medical images. We treat small image regions as orderless, squeeze their spatial information into feature dimensions and then perform MPS operations on these locally orderless regions. These local representations are then aggregated in a hierarchical manner to retain global structure. The proposed locally orderless tensor network (LoTeNet) is compared with relevant methods on three datasets. The architecture of LoTeNet is fixed in all experiments and we show it requires lesser computational resources to attain performance on par or superior to the compared methods."
                    },
                    {
                        "title": "Identifying partial mouse brain microscopy images from Allen reference atlas using a contrastively learned semantic space",
                        "abstract": "Precise identification of mouse brain microscopy images is a crucial first step when anatomical structures in the mouse brain are to be registered to a reference atlas. Practitioners usually rely on manual comparison of images or tools that assume the presence of complete images. This work explores Siamese Networks as the method for finding corresponding 2D reference atlas plates for given partial 2D mouse brain images. Siamese networks are a class of convolutional neural networks (CNNs) that use weight-shared paths to obtain low dimensional embeddings of pairs of input images. The correspondence between the partial mouse brain image and reference atlas plate is determined based on the distance between low dimensional embeddings of brain slices and atlas plates that are obtained from Siamese networks using contrastive learning. Experiments showed that Siamese CNNs can precisely identify brain slices using the Allen mouse brain atlas when training and testing images come from the same source. They achieved TOP-1 and TOP-5 accuracy of 25% and 100%, respectively, taking only 7.2 seconds to identify 29 images."
                    },
                    {
                        "title": "EC-NAS: Energy Consumption Aware Tabular Benchmarks for Neural Architecture Search",
                        "abstract": "Energy consumption from the selection, training, and deployment of deep learning models has seen a significant uptick recently. This work aims to facilitate the design of energy-efficient deep learning models that require less computational resources and prioritize environmental sustainability by focusing on the energy consumption. Neural architecture search (NAS) benefits from tabular benchmarks, which evaluate NAS strategies cost-effectively through precomputed performance statistics. We advocate for including energy efficiency as an additional performance criterion in NAS. To this end, we introduce an enhanced tabular benchmark encompassing data on energy consumption for varied architectures. The benchmark, designated as EC-NAS, has been made available in an open-source format to advance research in energy-conscious NAS. EC-NAS incorporates a surrogate model to predict energy consumption, aiding in diminishing the energy expenditure of the dataset creation. Our findings emphasize the potential of EC-NAS by leveraging multi-objective optimization algorithms, revealing a balance between energy usage and accuracy. This suggests the feasibility of identifying energy-lean architectures with little or no compromise in performance."
                    },
                    {
                        "title": "Carbontracker: Tracking and Predicting the Carbon Footprint of Training Deep Learning Models",
                        "abstract": "Deep learning (DL) can achieve impressive results across a wide variety of tasks, but this often comes at the cost of training models for extensive periods on specialized hardware accelerators. This energy-intensive workload has seen immense growth in recent years. Machine learning (ML) may become a significant contributor to climate change if this exponential trend continues. If practitioners are aware of their energy and carbon footprint, then they may actively take steps to reduce it whenever possible. In this work, we present Carbontracker, a tool for tracking and predicting the energy and carbon footprint of training DL models. We propose that energy and carbon footprint of model development and training is reported alongside performance metrics using tools like Carbontracker. We hope this will promote responsible computing in ML and encourage research into energy-efficient deep neural networks."
                    },
                    {
                        "title": "Interpreting Latent Spaces of Generative Models for Medical Images using Unsupervised Methods",
                        "abstract": "Generative models such as Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs) play an increasingly important role in medical image analysis. The latent spaces of these models often show semantically meaningful directions corresponding to human-interpretable image transformations. However, until now, their exploration for medical images has been limited due to the requirement of supervised data. Several methods for unsupervised discovery of interpretable directions in GAN latent spaces have shown interesting results on natural images. This work explores the potential of applying these techniques on medical images by training a GAN and a VAE on thoracic CT scans and using an unsupervised method to discover interpretable directions in the resulting latent space. We find several directions corresponding to non-trivial image transformations, such as rotation or breast size. Furthermore, the directions show that the generative models capture 3D structure despite being presented only with 2D data. The results show that unsupervised methods to discover interpretable directions in GANs generalize to VAEs and can be applied to medical images. This opens a wide array of future work using these methods in medical image analysis."
                    },
                    {
                        "title": "Patch-based Medical Image Segmentation using Matrix Product State Tensor Networks",
                        "abstract": "Tensor networks are efficient factorisations of high-dimensional tensors into a network of lower-order tensors. They have been most commonly used to model entanglement in quantum many-body systems and more recently are witnessing increased applications in supervised machine learning. In this work, we formulate image segmentation in a supervised setting with tensor networks. The key idea is to first lift the pixels in image patches to exponentially high-dimensional feature spaces and using a linear decision hyper-plane to classify the input pixels into foreground and background classes. The high-dimensional linear model itself is approximated using the matrix product state (MPS) tensor network. The MPS is weight-shared between the non-overlapping image patches resulting in our strided tensor network model. The performance of the proposed model is evaluated on three 2D- and one 3D- biomedical imaging datasets. The performance of the proposed tensor network segmentation model is compared with relevant baseline methods. In the 2D experiments, the tensor network model yields competitive performance compared to the baseline methods while being more resource efficient."
                    },
                    {
                        "title": "Extraction of airway trees using multiple hypothesis tracking and template matching",
                        "abstract": "Knowledge of airway tree morphology has important clinical applications in diagnosis of chronic obstructive pulmonary disease. We present an automatic tree extraction method based on multiple hypothesis tracking and template matching for this purpose and evaluate its performance on chest CT images. The method is adapted from a semi-automatic method devised for vessel segmentation. Idealized tubular templates are constructed that match airway probability obtained from a trained classifier and ranked based on their relative significance. Several such regularly spaced templates form the local hypotheses used in constructing a multiple hypothesis tree, which is then traversed to reach decisions. The proposed modifications remove the need for local thresholding of hypotheses as decisions are made entirely based on statistical comparisons involving the hypothesis tree. The results show improvements in performance when compared to the original method and region growing on intensity images. We also compare the method with region growing on the probability images, where the presented method does not show substantial improvement, but we expect it to be less sensitive to local anomalies in the data."
                    },
                    {
                        "title": "Extracting Tree-structures in CT data by Tracking Multiple Statistically Ranked Hypotheses",
                        "abstract": "In this work, we adapt a method based on multiple hypothesis tracking (MHT) that has been shown to give state-of-the-art vessel segmentation results in interactive settings, for the purpose of extracting trees. Regularly spaced tubular templates are fit to image data forming local hypotheses. These local hypotheses are used to construct the MHT tree, which is then traversed to make segmentation decisions. However, some critical parameters in this method are scale-dependent and have an adverse effect when tracking structures of varying dimensions. We propose to use statistical ranking of local hypotheses in constructing the MHT tree, which yields a probabilistic interpretation of scores across scales and helps alleviate the scale-dependence of MHT parameters. This enables our method to track trees starting from a single seed point. Our method is evaluated on chest CT data to extract airway trees and coronary arteries. In both cases, we show that our method performs significantly better than the original MHT method."
                    },
                    {
                        "title": "Extraction of Airways using Graph Neural Networks",
                        "abstract": "We present extraction of tree structures, such as airways, from image data as a graph refinement task. To this end, we propose a graph auto-encoder model that uses an encoder based on graph neural networks (GNNs) to learn embeddings from input node features and a decoder to predict connections between nodes. Performance of the GNN model is compared with mean-field networks in their ability to extract airways from 3D chest CT scans."
                    },
                    {
                        "title": "Efficiency is Not Enough: A Critical Perspective of Environmentally Sustainable AI",
                        "abstract": "Artificial Intelligence (AI) is currently spearheaded by machine learning (ML) methods such as deep learning (DL) which have accelerated progress on many tasks thought to be out of reach of AI. These ML methods can often be compute hungry, energy intensive, and result in significant carbon emissions, a known driver of anthropogenic climate change. Additionally, the platforms on which ML systems run are associated with environmental impacts including and beyond carbon emissions. The solution lionized by both industry and the ML community to improve the environmental sustainability of ML is to increase the efficiency with which ML systems operate in terms of both compute and energy consumption. In this perspective, we argue that efficiency alone is not enough to make ML as a technology environmentally sustainable. We do so by presenting three high level discrepancies between the effect of efficiency on the environmental sustainability of ML when considering the many variables which it interacts with. In doing so, we comprehensively demonstrate, at multiple levels of granularity both technical and non-technical reasons, why efficiency is not enough to fully remedy the environmental impacts of ML. Based on this, we present and argue for systems thinking as a viable path towards improving the environmental sustainability of ML holistically."
                    }
                ]
            }
        }
    },
    "2311.07568": {
        "paper_data": {
            "title": "Feature emergence via margin maximization: case studies in algebraic tasks",
            "url": "http://arxiv.org/abs/2311.07568v2",
            "arxiv_id": "2311.07568",
            "authors": [
                "Depen Morwani",
                "Benjamin L. Edelman",
                "Costin-Andrei Oncescu",
                "Rosie Zhao",
                "Sham Kakade"
            ],
            "abstract": "Understanding the internal representations learned by neural networks is a cornerstone challenge in the science of machine learning. While there have been significant recent strides in some cases towards understanding how neural networks implement specific target functions, this paper explores a complementary question -- why do networks arrive at particular computational strategies? Our inquiry focuses on the algebraic learning tasks of modular addition, sparse parities, and finite group operations. Our primary theoretical findings analytically characterize the features learned by stylized neural networks for these algebraic tasks. Notably, our main technique demonstrates how the principle of margin maximization alone can be used to fully specify the features learned by the network. Specifically, we prove that the trained networks utilize Fourier features to perform modular addition and employ features corresponding to irreducible group-theoretic representations to perform compositions in general groups, aligning closely with the empirical observations of Nanda et al. and Chughtai et al. More generally, we hope our techniques can help to foster a deeper understanding of why neural networks adopt specific computational strategies.",
            "introduction": " Introduction Opening the black box of neural networks has the potential to enable safer and more reliable deployments, justifications for model outputs, and clarity on how model behavior will be af- fected by changes in the input distribution. The research area of mechanistic interpretability (Olah et al., 2020; Elhage et al., 2021; Olsson et al., 2022; Elhage et al., 2022) aims to dissect individual trained neural networks in order to shed light on internal representations, iden- tifying and interpreting sub-circuits that contribute to the networks\u2019 functional behavior. Mechanistic interpretability analyses typically leave open the question of whythe observed representations arise as a result of training. Meanwhile, the theoretical literature on inductive biases in neural networks (Soudry et al., 2018; Shalev-Shwartz & Ben-David, 2014; Vardi, 2023) aims to derive general princi- ples governing which solutions will be preferred by trained neural networks\u2014in particular, in the presence of underspecification, where there are many distinct ways a network with a given architecture could perform well on the training data. Most work on inductive bias in deep learning is motivated by the question of understanding why networks generalize from their training data to unobserved test data. It can be non-obvious how to apply the results in terms of the form \u03c1i(a)\u03c1j(b)\u03c1i\u2032(a)\u03c1j\u2032(b)\u03c1k\u2032(c) in the expectation, where \u03c1i\u2032and\u03c1j\u2032correspond to entries of matrices from the same repre- sentation as \u03c1k\u2032. Thus if either \u03c1ior\u03c1jcorrespond to vectors from a different representation than \u03c1k, the expectation of this term will be zero, by orthogonality of the basis vectors. Hence we can assume that \u03c1i, \u03c1j, \u03c1kcorrespond to entries from the same representation (V, R). Let d=dV. Let us write i= (i1, i2), j= (j1, j2), k= (k1, k2), the matrix indices for this representation. We can expand the term \u03c1k(a\u25e6b\u25e6c) as described above. \u03c1i(a)\u03c1j(b)\u03c1k(a\u25e6b\u25e6c) =\u03c1i(a)\u03c1j(b)dX m=1\u03c1(k1,m)(a\u25e6b)\u03c1(m,k2)(c) =dX \u2113=1dX m=1\u03c1(i1,i2)(a)\u03c1(k1,\u2113)(a)\u03c1(j1,j2)(b)\u03c1(\u2113,m)(b)\u03c1(m,k2)(c). 34From this it is clear that when taking the expectation over choosing a, buniformly, the only non-zero terms are when ( i1, i2) = (k1, \u2113) and ( j1, j2) = (\u2113, m), once again by orthogonality of the basis vectors. Thus we have \u03b1i\u03b2j\u03b3k\u03c1i(a)\u03c1j(b)\u03c1k(a\u25e6b\u25e6c) =\u03b1(i1,j1)\u03b2(j1,j2)\u03b3(i1,k2)\u03c12 (i1,j1)(a)\u03c12 (j1,j2)(b)\u03c1(j2,k2)(c). Moreover, we know E[\u03c1i(a)2] = 1/d, where dis the dimensionality of the representation. Now, for a particular c, we will evaluate group all terms containing \u03c1(j2,k2)(c) and take the expectation over a, b, which yields 1 d2dX i1=1dX j1=1\u03b1(i1,j1)\u03b2(j1,j2)\u03b3(i1,k2)\u03c1(j2,k2)(c). (11) From the third property of Lemma 15, for every conjugacy class Cnforn\u2208[K], we have X c\u2208Cn1 d2dX i1=1dX j1=1\u03b1(i1,j1)\u03b2(j1,j2)\u03b3(i1,k2)\u03c1(j2,k2)(c) = 0 forj2\u0338=k2. Thus, we can focus on diagonal entries \u03c1(k,k)(c) (i.e. where j2=k2in the expression 11 above). In this case, following directly from 11 grouping all terms containing \u03c1(k,k)(c) we get 1 d2dX i1=1dX j1=1\u03b1(i1,j1)\u03b2(j1,k)\u03b3(i1,k)\u03c1(k,k)(c). (12) Note that this coefficient in front of \u03c1(k,k)(c) is the sum of the entries of the kthcolumn of the matrix ( \u03b1R\u03b2R)\u2299\u03b3Rdivided by d2(with \u03b1R\u03b2Rinterpreted as matrix product and \u2299being the Hadamard product). Recall that i1, j1, and kare indices from the same repre- sentation R. By summing over all diagonal entries ( k, k) inR, we evaluate the expression X i1,j1,k\u2208[d]\u03b1(i1,j1)\u03b2(j1,k)\u03b3(i1,k) d2\uf8ee \uf8f0\u03c1k,k(e)\u2212X c\u0338=e\u03c4c\u03c1(k,k)(c)\uf8f9 \uf8fb =tr(\u03b1R\u03b2R\u03b3RT) d2\" 1\u2212KX n=2\u03c4CnX c\u2208Cn\u03c1(k,k)(c)# where we have replaced \u03c4cwith the same weight \u03c4Cnfor each non-trivial conjugacy class C2, . . . , C Kand the termP c\u2208Cn\u03c1(k,k)(c) is independent of the choice of k(by property 4 of Lemma 15). Thus after summing over all k\u2208[d] the coefficient in equation 12 is the sum of all entries of the matrix ( \u03b1R\u03b2R)\u2299\u03b3R(which equals tr( \u03b1R\u03b2R\u03b3RT)). Furthermore, |Cn|\u03c7R(Cn) =X c\u2208CnX k\u2208[d]\u03c1(k,k)(c) =X k\u2208[d]X c\u2208Cn\u03c1(k,k)(c) =dX c\u2208Cn\u03c1(k,k)(c) where the first equality follows from the",
            "references": [
                {
                    "title": "Pareto Frontiers in Neural Feature Learning: Data, Compute, Width, and Luck",
                    "abstract": "In modern deep learning, algorithmic choices (such as width, depth, and learning rate) are known to modulate nuanced resource tradeoffs. This work investigates how these complexities necessarily arise for feature learning in the presence of computational-statistical gaps. We begin by considering offline sparse parity learning, a supervised classification problem which admits a statistical query lower bound for gradient-based training of a multilayer perceptron. This lower bound can be interpreted as a multi-resource tradeoff frontier: successful learning can only occur if one is sufficiently rich (large model), knowledgeable (large dataset), patient (many training iterations), or lucky (many random guesses). We show, theoretically and experimentally, that sparse initialization and increasing network width yield significant improvements in sample efficiency in this setting. Here, width plays the role of parallel search: it amplifies the probability of finding\"lottery ticket\"neurons, which learn sparse features more sample-efficiently. Finally, we show that the synthetic sparse parity task can be useful as a proxy for real problems requiring axis-aligned feature learning. We demonstrate improved sample efficiency on tabular classification benchmarks by using wide, sparsely-initialized MLP models; these networks sometimes outperform tuned random forests."
                },
                {
                    "title": "The Clock and the Pizza: Two Stories in Mechanistic Explanation of Neural Networks",
                    "abstract": "Do neural networks, trained on well-understood algorithmic tasks, reliably rediscover known algorithms for solving those tasks? Several recent studies, on tasks ranging from group arithmetic to in-context linear regression, have suggested that the answer is yes. Using modular addition as a prototypical problem, we show that algorithm discovery in neural networks is sometimes more complex. Small changes to model hyperparameters and initializations can induce the discovery of qualitatively different algorithms from a fixed training set, and even parallel implementations of multiple such algorithms. Some networks trained to perform modular addition implement a familiar Clock algorithm; others implement a previously undescribed, less intuitive, but comprehensible procedure which we term the Pizza algorithm, or a variety of even more complex procedures. Our results show that even simple learning problems can admit a surprising diversity of solutions, motivating the development of new tools for characterizing the behavior of neural networks across their algorithmic phase space."
                },
                {
                    "title": "Benign Overfitting in Linear Classifiers and Leaky ReLU Networks from KKT Conditions for Margin Maximization",
                    "abstract": "Linear classifiers and leaky ReLU networks trained by gradient flow on the logistic loss have an implicit bias towards solutions which satisfy the Karush--Kuhn--Tucker (KKT) conditions for margin maximization. In this work we establish a number of settings where the satisfaction of these KKT conditions implies benign overfitting in linear classifiers and in two-layer leaky ReLU networks: the estimators interpolate noisy training data and simultaneously generalize well to test data. The settings include variants of the noisy class-conditional Gaussians considered in previous work as well as new distributional settings where benign overfitting has not been previously observed. The key ingredient to our proof is the observation that when the training data is nearly-orthogonal, both linear classifiers and leaky ReLU networks satisfying the KKT conditions for their respective margin maximization problems behave like a nearly uniform average of the training examples."
                },
                {
                    "title": "A Toy Model of Universality: Reverse Engineering How Networks Learn Group Operations",
                    "abstract": "Universality is a key hypothesis in mechanistic interpretability -- that different models learn similar features and circuits when trained on similar tasks. In this work, we study the universality hypothesis by examining how small neural networks learn to implement group composition. We present a novel algorithm by which neural networks may implement composition for any finite group via mathematical representation theory. We then show that networks consistently learn this algorithm by reverse engineering model logits and weights, and confirm our understanding using ablations. By studying networks of differing architectures trained on various groups, we find mixed evidence for universality: using our algorithm, we can completely characterize the family of circuits and features that networks learn on this task, but for a given network the precise circuits learned -- as well as the order they develop -- are arbitrary."
                },
                {
                    "title": "Simplicity Bias in 1-Hidden Layer Neural Networks",
                    "abstract": "Recent works have demonstrated that neural networks exhibit extreme simplicity bias(SB). That is, they learn only the simplest features to solve a task at hand, even in the presence of other, more robust but more complex features. Due to the lack of a general and rigorous definition of features, these works showcase SB on semi-synthetic datasets such as Color-MNIST, MNIST-CIFAR where defining features is relatively easier. In this work, we rigorously define as well as thoroughly establish SB for one hidden layer neural networks. More concretely, (i) we define SB as the network essentially being a function of a low dimensional projection of the inputs (ii) theoretically, we show that when the data is linearly separable, the network primarily depends on only the linearly separable ($1$-dimensional) subspace even in the presence of an arbitrarily large number of other, more complex features which could have led to a significantly more robust classifier, (iii) empirically, we show that models trained on real datasets such as Imagenette and Waterbirds-Landbirds indeed depend on a low dimensional projection of the inputs, thereby demonstrating SB on these datasets, iv) finally, we present a natural ensemble approach that encourages diversity in models by training successive models on features not used by earlier models, and demonstrate that it yields models that are significantly more robust to Gaussian noise."
                },
                {
                    "title": "Progress measures for grokking via mechanistic interpretability",
                    "abstract": "Neural networks often exhibit emergent behavior, where qualitatively new capabilities arise from scaling up the amount of parameters, training data, or training steps. One approach to understanding emergence is to find continuous \\textit{progress measures} that underlie the seemingly discontinuous qualitative changes. We argue that progress measures can be found via mechanistic interpretability: reverse-engineering learned behaviors into their individual components. As a case study, we investigate the recently-discovered phenomenon of ``grokking'' exhibited by small transformers trained on modular addition tasks. We fully reverse engineer the algorithm learned by these networks, which uses discrete Fourier transforms and trigonometric identities to convert addition to rotation about a circle. We confirm the algorithm by analyzing the activations and weights and by performing ablations in Fourier space. Based on this understanding, we define progress measures that allow us to study the dynamics of training and split training into three continuous phases: memorization, circuit formation, and cleanup. Our results show that grokking, rather than being a sudden shift, arises from the gradual amplification of structured mechanisms encoded in the weights, followed by the later removal of memorizing components."
                },
                {
                    "title": "Grokking modular arithmetic",
                    "abstract": "We present a simple neural network that can learn modular arithmetic tasks and exhibits a sudden jump in generalization known as ``grokking''. Concretely, we present (i) fully-connected two-layer networks that exhibit grokking on various modular arithmetic tasks under vanilla gradient descent with the MSE loss function in the absence of any regularization; (ii) evidence that grokking modular arithmetic corresponds to learning specific feature maps whose structure is determined by the task; (iii) analytic expressions for the weights -- and thus for the feature maps -- that solve a large class of modular arithmetic tasks; and (iv) evidence that these feature maps are also found by vanilla gradient descent as well as AdamW, thereby establishing complete interpretability of the representations learnt by the network."
                },
                {
                    "title": "Implicit Bias in Leaky ReLU Networks Trained on High-Dimensional Data",
                    "abstract": "The implicit biases of gradient-based optimization algorithms are conjectured to be a major factor in the success of modern deep learning. In this work, we investigate the implicit bias of gradient flow and gradient descent in two-layer fully-connected neural networks with leaky ReLU activations when the training data are nearly-orthogonal, a common property of high-dimensional data. For gradient flow, we leverage recent work on the implicit bias for homogeneous neural networks to show that asymptotically, gradient flow produces a neural network with rank at most two. Moreover, this network is an $\\ell_2$-max-margin solution (in parameter space), and has a linear decision boundary that corresponds to an approximate-max-margin linear predictor. For gradient descent, provided the random initialization variance is small enough, we show that a single step of gradient descent suffices to drastically reduce the rank of the network, and that the rank remains small throughout training. We provide experiments which suggest that a small initialization scale is important for finding low-rank neural networks with gradient descent."
                },
                {
                    "title": "Omnigrok: Grokking Beyond Algorithmic Data",
                    "abstract": "Grokking, the unusual phenomenon for algorithmic datasets where generalization happens long after overfitting the training data, has remained elusive. We aim to understand grokking by analyzing the loss landscapes of neural networks, identifying the mismatch between training and test losses as the cause for grokking. We refer to this as the\"LU mechanism\"because training and test losses (against model weight norm) typically resemble\"L\"and\"U\", respectively. This simple mechanism can nicely explain many aspects of grokking: data size dependence, weight decay dependence, the emergence of representations, etc. Guided by the intuitive picture, we are able to induce grokking on tasks involving images, language and molecules. In the reverse direction, we are able to eliminate grokking for algorithmic datasets. We attribute the dramatic nature of grokking for algorithmic datasets to representation learning."
                },
                {
                    "title": "Toy Models of Superposition",
                    "abstract": "Neural networks often pack many unrelated concepts into a single neuron \u2013 a puzzling phenomenon known as 'polysemanticity' which makes interpretability much more challenging. This paper provides a toy model where polysemanticity can be fully understood, arising as a result of models storing additional sparse features in \"superposition.\" We demonstrate the existence of a phase change, a surprising connection to the geometry of uniform polytopes, and evidence of a link to adversarial examples. We also discuss potential implications for mechanistic interpretability. We recommend reading this paper as an HTML article. It would be very convenient if the individual neurons of artificial neural networks corresponded to cleanly interpretable features of the input. For example, in an \u201cideal\u201d ImageNet classifier, each neuron would fire only in the presence of a specific visual feature, such as the color red, a leftfacing curve, or a dog snout. Empirically, in models we have studied, some of the neurons do cleanly map to features. But it isn't always the case that features correspond so cleanly to neurons, especially in large language models where it actually seems rare for neurons to correspond to clean features. This brings up many questions. Why is it that neurons sometimes align with features and sometimes don't? Why do some models and tasks have many of these clean neurons, while they're vanishingly rare in others? In this paper, we use toy models \u2014 small ReLU networks trained on synthetic data with sparse input features \u2014 to investigate how and when models represent more features than they have dimensions. We call this phenomenon superposition. When features are sparse, superposition allows compression beyond what a linear model would do, at the cost of \"interference\" that requires nonlinear filtering. Consider a toy model where we train an embedding of five features of varying importance in two dimensions, add a ReLU afterwards for filtering, and vary the sparsity of the features. With dense features, the model learns to represent an orthogonal basis of the most important two features (similar to what Principal Component Analysis might give us), and the other three features are not represented. But if we make the features sparse, this changes: 1 This figure and a few others can be reproduced using the toy model framework Colab notebook in our Github repo Not only can models store additional features in superposition by tolerating some interference, but we'll show that, at least in certain limited cases, models can perform computation while in superposition. (In particular, we'll show that models can put simple circuits computing the absolute value function in superposition.) This leads us to hypothesize that the neural networks we observe in practice are in some sense noisily simulating larger, highly sparse networks. In other words, it's possible that models we train can be thought of as doing \u201cthe same thing as\u201d an imagined muchlarger model, representing the exact same features but with no interference. Feature superposition isn't a novel idea. A number of previous interpretability papers have speculated about it , and it's very closely related to the long-studied topic of compressed sensing in mathematics , as well as the ideas of distributed, dense, and population codes in neuroscience and deep learning . What, then, is the contribution of this paper? For interpretability researchers, our main contribution is providing a direct demonstration that superposition occurs in artificial neural networks given a relatively natural setup, suggesting this may also occur in practice. We offer a theory of when and why this occurs, revealing a phase diagram for superposition. We also discover that, at least in our toy model, superposition exhibits complex geometric structure. But our results may also be of broader interest. We find preliminary evidence that superposition may be linked to adversarial examples and grokking, and might also suggest a theory for the performance of mixture of experts models. More broadly, the toy model we investigate has unexpectedly rich structure, exhibiting phase changes, a geometric structure based on uniform polytopes, \"energy level\"-like jumps during training, and a phenomenon which is qualitatively similar to the fractional quantum Hall effect in physics. We originally investigated the subject to gain understanding of cleanly-interpretable neurons in larger models, but we've found these toy models to be surprisingly interesting in their own right. [1, 2] [3] [4] [5] KEY RESULTS FROM OUR TOY MODELS In our toy models, we are able to demonstrate that: Superposition is a real, observed phenomenon. Both monosemantic and polysemantic neurons can form. At least some kinds of computation can be performed in superposition. Whether features are stored in superposition is governed by a phase change. Superposition organizes features into geometric structures such as digons, triangles, pentagons, and tetrahedrons. Our toy models are simple ReLU networks, so it seems fair to say that neural networks exhibit these properties in at least some regimes, but it's very unclear what to generalize to real networks. Definitions and Motivation: Features, Directions, and Superposition In our work, we often think of neural networks as having features of the input represented as directions in activation space. This isn't a trivial claim. It isn't obvious what kind of structure we should expect neural network representations to have. When we say something like \"word embeddings have a gender direction\" or \"vision models have curve detector neurons\", one is implicitly making strong claims about the structure of network representations. Despite this, we believe this kind of \"linear representation hypothesis\" is supported both by significant empirical findings and theoretical arguments. One might think of this as two separate properties, which we'll explore in more detail shortly: Decomposability: Network representations can be described in terms of independently understandable features. Linearity: Features are represented by direction. If we hope to reverse engineer neural networks, we need a property like decomposability. Decomposability is what allows us to reason about the model without fitting the whole thing in our heads! But it's not enough for things to be decomposable: we need to be able to access the decomposition somehow. In order to do this, we need to identify the individual features within a representation. In a linear representation, this corresponds to determining which directions in activation space correspond to which independent features of the input. Sometimes, identifying feature directions is very easy because features seem to correspond to neurons. For example, many neurons in the early layers of InceptionV1 clearly correspond to features (e.g. curve detector neurons ). Why is it that we sometimes get this extremely helpful property, but in other cases don't? We hypothesize that there are really two countervailing forces driving this: [6] Privileged Basis: Only some representations have a privileged basis which encourages features to align with basis directions (i.e. to correspond to neurons). Superposition: Linear representations can represent more features than dimensions, using a strategy we call superposition. This can be seen as neural networks simulating larger networks. This pushes features away from corresponding to neurons. Superposition has been hypothesized in previous work  . However, we're not aware of feature superposition having been unambiguously demonstrated to occur in neural networks before ( demonstrates a closely related phenomenon of model superposition). The goal of this paper is to change that, demonstrating superposition and exploring how it interacts with privileged bases. If superposition occurs in networks, it deeply influences what approaches to interpretability research make sense, so unambiguous demonstration seems important. [1, 2] [7] The goal of this section will be to motivate these ideas and unpack them in detail. It's worth noting that many of the ideas in this section have close connections to ideas in other lines of interpretability research (especially disentanglement), neuroscience (distributed representations, population codes, etc), compressed sensing, and many other lines of work. This section will focus on articulating our perspective on the problem. We'll discuss these other lines of work in detail in Related Work."
                },
                {
                    "title": "On the Implicit Bias in Deep-Learning Algorithms",
                    "abstract": "Examining the implicit bias in training neural networks using gradient-based methods."
                },
                {
                    "title": "Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limit",
                    "abstract": "There is mounting evidence of emergent phenomena in the capabilities of deep learning methods as we scale up datasets, model sizes, and training times. While there are some accounts of how these resources modulate statistical capacity, far less is known about their effect on the computational problem of model training. This work conducts such an exploration through the lens of learning a $k$-sparse parity of $n$ bits, a canonical discrete search problem which is statistically easy but computationally hard. Empirically, we find that a variety of neural networks successfully learn sparse parities, with discontinuous phase transitions in the training curves. On small instances, learning abruptly occurs at approximately $n^{O(k)}$ iterations; this nearly matches SQ lower bounds, despite the apparent lack of a sparse prior. Our theoretical analysis shows that these observations are not explained by a Langevin-like mechanism, whereby SGD\"stumbles in the dark\"until it finds the hidden set of features (a natural algorithm which also runs in $n^{O(k)}$ time). Instead, we show that SGD gradually amplifies the sparse solution via a Fourier gap in the population gradient, making continual progress that is invisible to loss and error metrics."
                },
                {
                    "title": "Solving Quantitative Reasoning Problems with Language Models",
                    "abstract": "Language models have achieved remarkable performance on a wide range of tasks that require natural language understanding. Nevertheless, state-of-the-art models have generally struggled with tasks that require quantitative reasoning, such as solving mathematics, science, and engineering problems at the college level. To help close this gap, we introduce Minerva, a large language model pretrained on general natural language data and further trained on technical content. The model achieves state-of-the-art performance on technical benchmarks without the use of external tools. We also evaluate our model on over two hundred undergraduate-level problems in physics, biology, chemistry, economics, and other sciences that require quantitative reasoning, and find that the model can correctly answer nearly a third of them."
                },
                {
                    "title": "A Theoretical Analysis on Feature Learning in Neural Networks: Emergence from Inputs and Advantage over Fixed Features",
                    "abstract": "An important characteristic of neural networks is their ability to learn representations of the input data with effective features for prediction, which is believed to be a key factor to their superior empirical performance. To better understand the source and benefit of feature learning in neural networks, we consider learning problems motivated by practical data, where the labels are determined by a set of class relevant patterns and the inputs are generated from these along with some background patterns. We prove that neural networks trained by gradient descent can succeed on these problems. The success relies on the emergence and improvement of effective features, which are learned among exponentially many candidates efficiently by exploiting the data (in particular, the structure of the input distribution). In contrast, no linear models on data-independent features of polynomial sizes can learn to as good errors. Furthermore, if the specific input structure is removed, then no polynomial algorithm in the Statistical Query model can learn even weakly. These results provide theoretical evidence showing that feature learning in neural networks depends strongly on the input structure and leads to the superior performance. Our preliminary experimental results on synthetic and real data also provide positive support."
                },
                {
                    "title": "Towards Understanding Grokking: An Effective Theory of Representation Learning",
                    "abstract": "We aim to understand grokking, a phenomenon where models generalize long after overfitting their training set. We present both a microscopic analysis anchored by an effective theory and a macroscopic analysis of phase diagrams describing learning performance across hyperparameters. We find that generalization originates from structured representations whose training dynamics and dependence on training set size can be predicted by our effective theory in a toy setting. We observe empirically the presence of four learning phases: comprehension, grokking, memorization, and confusion. We find representation learning to occur only in a\"Goldilocks zone\"(including comprehension and grokking) between memorization and confusion. We find on transformers the grokking phase stays closer to the memorization phase (compared to the comprehension phase), leading to delayed generalization. The Goldilocks phase is reminiscent of\"intelligence from starvation\"in Darwinian evolution, where resource limitations drive discovery of more efficient solutions. This study not only provides intuitive explanations of the origin of grokking, but also highlights the usefulness of physics-inspired tools, e.g., effective theories and phase diagrams, for understanding deep learning."
                },
                {
                    "title": "Random Feature Amplification: Feature Learning and Generalization in Neural Networks",
                    "abstract": "In this work, we provide a characterization of the feature-learning process in two-layer ReLU networks trained by gradient descent on the logistic loss following random initialization. We consider data with binary labels that are generated by an XOR-like function of the input features. We permit a constant fraction of the training labels to be corrupted by an adversary. We show that, although linear classifiers are no better than random guessing for the distribution we consider, two-layer ReLU networks trained by gradient descent achieve generalization error close to the label noise rate. We develop a novel proof technique that shows that at initialization, the vast majority of neurons function as random features that are only weakly correlated with useful features, and the gradient descent dynamics 'amplify' these weak, random features to strong, useful features."
                },
                {
                    "title": "Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets",
                    "abstract": "In this paper we propose to study generalization of neural networks on small algorithmically generated datasets. In this setting, questions about data efficiency, memorization, generalization, and speed of learning can be studied in great detail. In some situations we show that neural networks learn through a process of\"grokking\"a pattern in the data, improving generalization performance from random chance level to perfect generalization, and that this improvement in generalization can happen well past the point of overfitting. We also study generalization as a function of dataset size and find that smaller datasets require increasing amounts of optimization for generalization. We argue that these datasets provide a fertile ground for studying a poorly understood aspect of deep learning: generalization of overparametrized neural networks beyond memorization of the finite training dataset."
                },
                {
                    "title": "Gradient Descent on Two-layer Nets: Margin Maximization and Simplicity Bias",
                    "abstract": "The generalization mystery of overparametrized deep nets has motivated efforts to understand how gradient descent (GD) converges to low-loss solutions that generalize well. Real-life neural networks are initialized from small random values and trained with cross-entropy loss for classification (unlike the\"lazy\"or\"NTK\"regime of training where analysis was more successful), and a recent sequence of results (Lyu and Li, 2020; Chizat and Bach, 2020; Ji and Telgarsky, 2020) provide theoretical evidence that GD may converge to the\"max-margin\"solution with zero loss, which presumably generalizes well. However, the global optimality of margin is proved only in some settings where neural nets are infinitely or exponentially wide. The current paper is able to establish this global optimality for two-layer Leaky ReLU nets trained with gradient flow on linearly separable and symmetric data, regardless of the width. The analysis also gives some theoretical justification for recent empirical findings (Kalimeris et al., 2019) on the so-called simplicity bias of GD towards linear or other\"simple\"classes of solutions, especially early in training. On the pessimistic side, the paper suggests that such results are fragile. A simple data manipulation can make gradient flow converge to a linear classifier with suboptimal margin."
                },
                {
                    "title": "On Margin Maximization in Linear and ReLU Networks",
                    "abstract": "The implicit bias of neural networks has been extensively studied in recent years. Lyu and Li [2019] showed that in homogeneous networks trained with the exponential or the logistic loss, gradient flow converges to a KKT point of the max margin problem in the parameter space. However, that leaves open the question of whether this point will generally be an actual optimum of the max margin problem. In this paper, we study this question in detail, for several neural network architectures involving linear and ReLU activations. Perhaps surprisingly, we show that in many cases, the KKT point is not even a local optimum of the max margin problem. On the flip side, we identify multiple settings where a local or global optimum can be guaranteed."
                },
                {
                    "title": "Measuring Mathematical Problem Solving With the MATH Dataset",
                    "abstract": "Many intellectual endeavors require mathematical problem solving, but this skill remains beyond the capabilities of computers. To measure this ability in machine learning models, we introduce MATH, a new dataset of 12,500 challenging competition mathematics problems. Each problem in MATH has a full step-by-step solution which can be used to teach models to generate answer derivations and explanations. To facilitate future research and increase accuracy on MATH, we also contribute a large auxiliary pretraining dataset which helps teach models the fundamentals of mathematics. Even though we are able to increase accuracy on MATH, our results show that accuracy remains relatively low, even with enormous Transformer models. Moreover, we find that simply increasing budgets and model parameter counts will be impractical for achieving strong mathematical reasoning if scaling trends continue. While scaling Transformers is automatically solving most other text-based tasks, scaling is not currently solving MATH. To have more traction on mathematical problem solving we will likely need new algorithmic advancements from the broader research community."
                },
                {
                    "title": "Directional convergence and alignment in deep learning",
                    "abstract": "In this paper, we show that although the minimizers of cross-entropy and related classification losses are off at infinity, network weights learned by gradient flow converge in direction, with an immediate corollary that network predictions, training errors, and the margin distribution also converge. This proof holds for deep homogeneous networks -- a broad class of networks allowing for ReLU, max pooling, linear, and convolutional layers -- and we additionally provide empirical support not just close to the theory (e.g., the AlexNet), but also on non-homogeneous networks (e.g., the ResNet). If the network further has locally Lipschitz gradients, we show that these gradients converge in direction, and asymptotically align with the gradient flow path, with consequences on margin maximization. Our analysis complements and is distinct from the well-known neural tangent and mean-field theories, and in particular makes no requirements on network width and initialization, instead merely requiring perfect classification accuracy. The proof proceeds by developing a theory of unbounded nonsmooth Kurdyka-Lojasiewicz inequalities for functions definable in an o-minimal structure, and is also applicable outside deep learning."
                },
                {
                    "title": "Learning Parities with Neural Networks",
                    "abstract": "In recent years we see a rapidly growing line of research which shows learnability of various models via common neural network algorithms. Yet, besides a very few outliers, these results show learnability of models that can be learned using linear methods. Namely, such results show that learning neural-networks with gradient-descent is competitive with learning a linear classifier on top of a data-independent representation of the examples. This leaves much to be desired, as neural networks are far more successful than linear methods. Furthermore, on the more conceptual level, linear models don't seem to capture the \"deepness\" of deep networks. In this paper we make a step towards showing leanability of models that are inherently non-linear. We show that under certain distributions, sparse parities are learnable via gradient decent on depth-two network. On the other hand, under the same distributions, these parities cannot be learned efficiently by linear methods."
                },
                {
                    "title": "Implicit Bias of Gradient Descent for Wide Two-layer Neural Networks Trained with the Logistic Loss",
                    "abstract": "Neural networks trained to minimize the logistic (a.k.a. cross-entropy) loss with gradient-based methods are observed to perform well in many supervised classification tasks. Towards understanding this phenomenon, we analyze the training and generalization behavior of infinitely wide two-layer neural networks with homogeneous activations. We show that the limits of the gradient flow on exponentially tailed losses can be fully characterized as a max-margin classifier in a certain non-Hilbertian space of functions. In presence of hidden low-dimensional structures, the resulting margin is independent of the ambiant dimension, which leads to strong generalization bounds. In contrast, training only the output layer implicitly solves a kernel support vector machine, which a priori does not enjoy such an adaptivity. Our analysis of training is non-quantitative in terms of running time but we prove computational guarantees in simplified settings by showing equivalences with online mirror descent. Finally, numerical experiments suggest that our analysis describes well the practical behavior of two-layer neural networks with ReLU activation and confirm the statistical benefits of this implicit bias."
                },
                {
                    "title": "Gradient Descent Maximizes the Margin of Homogeneous Neural Networks",
                    "abstract": "In this paper, we study the implicit regularization of the gradient descent algorithm in homogeneous neural networks, including fully-connected and convolutional neural networks with ReLU or LeakyReLU activations. In particular, we study the gradient descent or gradient flow (i.e., gradient descent with infinitesimal step size) optimizing the logistic loss or cross-entropy loss of any homogeneous model (possibly non-smooth), and show that if the training loss decreases below a certain threshold, then we can define a smoothed version of the normalized margin which increases over time. We also formulate a natural constrained optimization problem related to margin maximization, and prove that both the normalized margin and its smoothed version converge to the objective value at a KKT point of the optimization problem. Our results generalize the previous results for logistic regression with one-layer or multi-layer linear networks, and provide more quantitative convergence results with weaker assumptions than previous results for homogeneous smooth neural networks. We conduct several experiments to justify our theoretical finding on MNIST and CIFAR-10 datasets. Finally, as margin is closely related to robustness, we discuss potential benefits of training longer for improving the robustness of the model."
                },
                {
                    "title": "Analysing Mathematical Reasoning Abilities of Neural Models",
                    "abstract": "Mathematical reasoning---a core ability within human intelligence---presents some unique challenges as a domain: we do not come to understand and solve mathematical problems primarily on the back of experience and evidence, but on the basis of inferring, learning, and exploiting laws, axioms, and symbol manipulation rules. In this paper, we present a new challenge for the evaluation (and eventually the design) of neural architectures and similar system, developing a task suite of mathematics problems involving sequential questions and answers in a free-form textual input/output format. The structured nature of the mathematics domain, covering arithmetic, algebra, probability and calculus, enables the construction of training and test splits designed to clearly illuminate the capabilities and failure-modes of different architectures, as well as evaluate their ability to compose and relate knowledge and learned processes. Having described the data generation process and its potential future expansions, we conduct a comprehensive analysis of models from two broad classes of the most powerful sequence-to-sequence architectures and find notable differences in their ability to resolve mathematical problems and generalize their knowledge."
                },
                {
                    "title": "Regularization Matters: Generalization and Optimization of Neural Nets v.s. their Induced Kernel",
                    "abstract": "Recent works have shown that on sufficiently over-parametrized neural nets, gradient descent with relatively large initialization optimizes a prediction function in the RKHS of the Neural Tangent Kernel (NTK). This analysis leads to global convergence results but does not work when there is a standard $\\ell_2$ regularizer, which is useful to have in practice. We show that sample efficiency can indeed depend on the presence of the regularizer: we construct a simple distribution in d dimensions which the optimal regularized neural net learns with $O(d)$ samples but the NTK requires $\\Omega(d^2)$ samples to learn. To prove this, we establish two analysis tools: i) for multi-layer feedforward ReLU nets, we show that the global minimizer of a weakly-regularized cross-entropy loss is the max normalized margin solution among all neural nets, which generalizes well; ii) we develop a new technique for proving lower bounds for kernel methods, which relies on showing that the kernel cannot focus on informative features. Motivated by our generalization results, we study whether the regularized global optimum is attainable. We prove that for infinite-width two-layer nets, noisy gradient descent optimizes the regularized neural net loss to a global minimum in polynomial iterations."
                },
                {
                    "title": "Implicit Bias of Gradient Descent on Linear Convolutional Networks",
                    "abstract": "We show that gradient descent on full-width linear convolutional networks of depth $L$ converges to a linear predictor related to the $\\ell_{2/L}$ bridge penalty in the frequency domain. This is in contrast to linearly fully connected networks, where gradient descent converges to the hard margin linear support vector machine solution, regardless of depth."
                },
                {
                    "title": "The Implicit Bias of Gradient Descent on Separable Data",
                    "abstract": "We examine gradient descent on unregularized logistic regression problems, with homogeneous linear predictors on linearly separable datasets. We show the predictor converges to the direction of the max-margin (hard margin SVM) solution. The result also generalizes to other monotone decreasing loss functions with an infimum at infinity, to multi-class problems, and to training a weight layer in a deep network in a certain restricted setting. Furthermore, we show this convergence is very slow, and only logarithmic in the convergence of the loss itself. This can help explain the benefit of continuing to optimize the logistic or cross-entropy loss even after the training error is zero and the training loss is extremely small, and, as we show, even if the validation loss increases. Our methodology can also aid in understanding implicit regularization n more complex models and with other optimization methods."
                },
                {
                    "title": "Convex Optimization",
                    "abstract": "This textbook is based on lectures given by the authors at MIPT (Moscow), HSE (Moscow), FEFU (Vladivostok), V.I. Vernadsky KFU (Simferopol), ASU (Republic of Adygea), and the University of Grenoble-Alpes (Grenoble, France). First of all, the authors focused on the program of a two-semester course of lectures on convex optimization, which is given to students of MIPT. The first chapter of this book contains the materials of the first semester (\"Fundamentals of convex analysis and optimization\"), the second and third chapters contain the materials of the second semester (\"Numerical methods of convex optimization\"). The textbook has a number of features. First, in contrast to the classic manuals, this book does not provide proofs of all the theorems mentioned. This allowed, on one side, to describe more themes, but on the other side, made the presentation less self-sufficient. The second important point is that part of the material is advanced and is published in the Russian educational literature, apparently for the first time. Third, the accents that are given do not always coincide with the generally accepted accents in the textbooks that are now popular. First of all, we talk about a sufficiently advanced presentation of conic optimization, including robust optimization, as a vivid demonstration of the capabilities of modern convex analysis."
                },
                {
                    "title": "For Valid Generalization the Size of the Weights is More Important than the Size of the Network",
                    "abstract": "This paper shows that if a large neural network is used for a pattern classification problem, and the learning algorithm finds a network with small weights that has small squared error on the training patterns, then the generalization performance depends on the size of the weights rather than the number of weights. More specifically, consider an l-layer feed-forward network of sigmoid units, in which the sum of the magnitudes of the weights associated with each unit is bounded by A. The misclassification probability converges to an error estimate (that is closely related to squared error on the training set) at rate O((cA)l(l+1)/2 \u221a(log n)/m) ignoring log factors, where m is the number of training patterns, n is the input dimension, and c is a constant. This may explain the generalization performance of neural networks, particularly when the number of training examples is considerably smaller than the number of weights. It also supports heuristics (such as weight decay and early stopping) that attempt to keep the weights small during training."
                },
                {
                    "title": "On the inductive bias of neural networks for learning read-once DNFs",
                    "abstract": "Learning functions over Boolean variables is a fundamental problem in machine learning. But not much is known about learning such functions us-ing neural networks. Here we focus on learning read-once disjunctive normal forms (DNFs) under the uniform distribution with a convex neural network and gradient methods. We \ufb01rst observe empirically that gradient methods converge to compact solutions with neurons that are aligned with the terms of the DNF. This is despite the fact that there are many zero training error networks that do not have this property. Thus, the learning process has a clear inductive bias towards such logical formulas. Following recent results which connect the inductive bias of gradient \ufb02ow (GF) to Karush-Kuhn-Tucker (KKT) points of minimum norm problems, we study these KKT points in our setting. We prove that zero training error solutions that memorize training points are not KKT points and therefore GF cannot converge to them. On the other hand, we prove that globally optimal KKT points correspond exactly to networks that are aligned with the DNF terms. These results suggest a strong connection between the inductive bias of GF and solutions that align with the DNF. We conclude with extensive experiments which verify our \ufb01ndings."
                },
                {
                    "title": "Understanding Machine Learning - From Theory to Algorithms",
                    "abstract": "Machine learning is one of the fastest growing areas of computer science, with far-reaching applications. The aim of this textbook is to introduce machine learning, and the algorithmic paradigms it offers, in a principled way. The book provides an extensive theoretical account of the fundamental ideas underlying machine learning and the mathematical derivations that transform these principles into practical algorithms. Following a presentation of the basics of the field, the book covers a wide array of central topics that have not been addressed by previous textbooks. These include a discussion of the computational complexity of learning and the concepts of convexity and stability ; important algorithmic paradigms including stochastic gradient descent, neural networks, and structured output learning ; and emerging theoretical concepts such as the PACBayes approach and compression-based bounds. Designed for an advanced undergraduate or beginning graduate course, the text makes the fundamentals and algorithms of machine learning accessible to students and nonexpert readers in statistics, computer science, mathematics, and engineering."
                }
            ],
            "categories": [
                "cs.LG",
                "I.5.1; I.2.6"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively understand and interpret the internal representations and inductive biases of neural networks to improve their mechanistic interpretability?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the fundamental challenge of understanding how neural networks make decisions. By elucidating the internal workings of these models, we can enhance their reliability and safety in real-world applications, leading to better justifications for model outputs. This understanding could pave the way for advancements in various fields, such as healthcare, finance, and autonomous systems, where interpretability is essential for trust and accountability. Furthermore, it could inspire future research directions focused on developing more interpretable models and improving generalization capabilities.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of neural networks and the high-dimensional nature of their representations. Naive approaches may fail because they often overlook the intricate relationships between different components of the network and the underlying inductive biases that influence learning. Additionally, the orthogonality of basis vectors complicates the interpretation of interactions within the network, making it difficult to draw meaningful conclusions from the observed representations. Overcoming these technical and theoretical obstacles requires sophisticated analytical techniques and a deep understanding of both the architecture and the training dynamics of neural networks.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either mechanistic interpretability or inductive biases in isolation, leading to a lack of comprehensive frameworks that integrate both aspects. Limitations in existing methodologies, such as insufficient analytical tools or a narrow focus on specific architectures, have hindered progress. Additionally, the complexity of neural networks and the diversity of their training scenarios have created barriers to a unified understanding. Our approach aims to bridge these gaps by combining insights from both fields, offering a more holistic perspective on how neural networks learn and represent information.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a detailed analysis of the internal representations of neural networks through a combination of theoretical frameworks and empirical evaluations. We will utilize a diverse set of datasets to train various neural network architectures, focusing on metrics such as interpretability scores and generalization performance. By systematically examining the relationships between different representations and their corresponding inductive biases, we expect to uncover key insights into the mechanisms driving model behavior. The anticipated outcomes"
            }
        },
        "author_data": {
            "07e69be1-f913-4a51-a841-57a06e48c785": {
                "pk": "07e69be1-f913-4a51-a841-57a06e48c785",
                "name": "Depen Morwani",
                "collaborators": [
                    "Nikhil Vyas",
                    "S. Kakade",
                    "Rosie Zhao",
                    "H. G. Ramaswamy",
                    "Itai Shapira",
                    "Lucas Janson",
                    "David Brandfonbrener",
                    "Hanlin Zhang",
                    "Jingfeng Wu",
                    "Difan Zou"
                ],
                "domain": [
                    "Optimization",
                    "Neural Networks",
                    "AutoML",
                    "Language Models"
                ],
                "publications": [
                    {
                        "title": "How Does Critical Batch Size Scale in Pre-training?",
                        "abstract": "Training large-scale models under given resources requires careful design of parallelism strategies. In particular, the efficiency notion of critical batch size, concerning the compromise between time and compute, marks the threshold beyond which greater data parallelism leads to diminishing returns. To operationalize it, we propose a measure of CBS and pre-train a series of auto-regressive language models, ranging from 85 million to 1.2 billion parameters, on the C4 dataset. Through extensive hyper-parameter sweeps and careful control on factors such as batch size, momentum, and learning rate along with its scheduling, we systematically investigate the impact of scale on CBS. Then we fit scaling laws with respect to model and data sizes to decouple their effects. Overall, our results demonstrate that CBS scales primarily with data size rather than model size, a finding we justify theoretically through the analysis of infinite-width limits of neural networks and infinite-dimensional least squares regression. Of independent interest, we highlight the importance of common hyper-parameter choices and strategies for studying large-scale pre-training beyond fixed training durations."
                    },
                    {
                        "title": "A New Perspective on Shampoo's Preconditioner",
                        "abstract": "Shampoo, a second-order optimization algorithm which uses a Kronecker product preconditioner, has recently garnered increasing attention from the machine learning community. The preconditioner used by Shampoo can be viewed either as an approximation of the Gauss--Newton component of the Hessian or the covariance matrix of the gradients maintained by Adagrad. We provide an explicit and novel connection between the $\\textit{optimal}$ Kronecker product approximation of these matrices and the approximation made by Shampoo. Our connection highlights a subtle but common misconception about Shampoo's approximation. In particular, the $\\textit{square}$ of the approximation used by the Shampoo optimizer is equivalent to a single step of the power iteration algorithm for computing the aforementioned optimal Kronecker product approximation. Across a variety of datasets and architectures we empirically demonstrate that this is close to the optimal Kronecker product approximation. Additionally, for the Hessian approximation viewpoint, we empirically study the impact of various practical tricks to make Shampoo more computationally efficient (such as using the batch gradient and the empirical Fisher) on the quality of Hessian approximation."
                    },
                    {
                        "title": "Deconstructing What Makes a Good Optimizer for Language Models",
                        "abstract": "Training language models becomes increasingly expensive with scale, prompting numerous attempts to improve optimization efficiency. Despite these efforts, the Adam optimizer remains the most widely used, due to a prevailing view that it is the most effective approach. We aim to compare several optimization algorithms, including SGD, Adafactor, Adam, and Lion, in the context of autoregressive language modeling across a range of model sizes, hyperparameters, and architecture variants. Our findings indicate that, except for SGD, these algorithms all perform comparably both in their optimal performance and also in terms of how they fare across a wide range of hyperparameter choices. Our results suggest to practitioners that the choice of optimizer can be guided by practical considerations like memory constraints and ease of implementation, as no single algorithm emerged as a clear winner in terms of performance or stability to hyperparameter misspecification. Given our findings, we further dissect these approaches, examining two simplified versions of Adam: a) signed momentum (Signum) which we see recovers both the performance and hyperparameter stability of Adam and b) Adalayer, a layerwise variant of Adam which we introduce to study Adam's preconditioning. Examining Adalayer leads us to the conclusion that the largest impact of Adam's preconditioning is restricted to the last layer and LayerNorm parameters, and, perhaps surprisingly, the remaining layers can be trained with SGD."
                    },
                    {
                        "title": "SOAP: Improving and Stabilizing Shampoo using Adam",
                        "abstract": "There is growing evidence of the effectiveness of Shampoo, a higher-order preconditioning method, over Adam in deep learning optimization tasks. However, Shampoo's drawbacks include additional hyperparameters and computational overhead when compared to Adam, which only updates running averages of first- and second-moment quantities. This work establishes a formal connection between Shampoo (implemented with the 1/2 power) and Adafactor -- a memory-efficient approximation of Adam -- showing that Shampoo is equivalent to running Adafactor in the eigenbasis of Shampoo's preconditioner. This insight leads to the design of a simpler and computationally efficient algorithm: $\\textbf{S}$hampo$\\textbf{O}$ with $\\textbf{A}$dam in the $\\textbf{P}$reconditioner's eigenbasis (SOAP). With regards to improving Shampoo's computational efficiency, the most straightforward approach would be to simply compute Shampoo's eigendecomposition less frequently. Unfortunately, as our empirical results show, this leads to performance degradation that worsens with this frequency. SOAP mitigates this degradation by continually updating the running average of the second moment, just as Adam does, but in the current (slowly changing) coordinate basis. Furthermore, since SOAP is equivalent to running Adam in a rotated space, it introduces only one additional hyperparameter (the preconditioning frequency) compared to Adam. We empirically evaluate SOAP on language model pre-training with 360m and 660m sized models. In the large batch regime, SOAP reduces the number of iterations by over 40% and wall clock time by over 35% compared to AdamW, with approximately 20% improvements in both metrics compared to Shampoo. An implementation of SOAP is available at https://github.com/nikhilvyas/SOAP."
                    },
                    {
                        "title": "Simplicity Bias in 1-Hidden Layer Neural Networks",
                        "abstract": "Recent works have demonstrated that neural networks exhibit extreme simplicity bias(SB). That is, they learn only the simplest features to solve a task at hand, even in the presence of other, more robust but more complex features. Due to the lack of a general and rigorous definition of features, these works showcase SB on semi-synthetic datasets such as Color-MNIST, MNIST-CIFAR where defining features is relatively easier. In this work, we rigorously define as well as thoroughly establish SB for one hidden layer neural networks. More concretely, (i) we define SB as the network essentially being a function of a low dimensional projection of the inputs (ii) theoretically, we show that when the data is linearly separable, the network primarily depends on only the linearly separable ($1$-dimensional) subspace even in the presence of an arbitrarily large number of other, more complex features which could have led to a significantly more robust classifier, (iii) empirically, we show that models trained on real datasets such as Imagenette and Waterbirds-Landbirds indeed depend on a low dimensional projection of the inputs, thereby demonstrating SB on these datasets, iv) finally, we present a natural ensemble approach that encourages diversity in models by training successive models on features not used by earlier models, and demonstrate that it yields models that are significantly more robust to Gaussian noise."
                    },
                    {
                        "title": "Feature-Learning Networks Are Consistent Across Widths At Realistic Scales",
                        "abstract": "We study the effect of width on the dynamics of feature-learning neural networks across a variety of architectures and datasets. Early in training, wide neural networks trained on online data have not only identical loss curves but also agree in their point-wise test predictions throughout training. For simple tasks such as CIFAR-5m this holds throughout training for networks of realistic widths. We also show that structural properties of the models, including internal representations, preactivation distributions, edge of stability phenomena, and large learning rate effects are consistent across large widths. This motivates the hypothesis that phenomena seen in realistic models can be captured by infinite-width, feature-learning limits. For harder tasks (such as ImageNet and language modeling), and later training times, finite-width deviations grow systematically. Two distinct effects cause these deviations across widths. First, the network output has initialization-dependent variance scaling inversely with width, which can be removed by ensembling networks. We observe, however, that ensembles of narrower networks perform worse than a single wide network. We call this the bias of narrower width. We conclude with a spectral perspective on the origin of this finite-width bias."
                    },
                    {
                        "title": "Beyond Implicit Bias: The Insignificance of SGD Noise in Online Learning",
                        "abstract": "The success of SGD in deep learning has been ascribed by prior works to the implicit bias induced by finite batch sizes (\"SGD noise\"). While prior works focused on offline learning (i.e., multiple-epoch training), we study the impact of SGD noise on online (i.e., single epoch) learning. Through an extensive empirical analysis of image and language data, we demonstrate that small batch sizes do not confer any implicit bias advantages in online learning. In contrast to offline learning, the benefits of SGD noise in online learning are strictly computational, facilitating more cost-effective gradient steps. This suggests that SGD in the online regime can be construed as taking noisy steps along the\"golden path\"of the noiseless gradient descent algorithm. We study this hypothesis and provide supporting evidence in loss and function space. Our findings challenge the prevailing understanding of SGD and offer novel insights into its role in online learning."
                    },
                    {
                        "title": "Inductive Bias of Gradient Descent for Exponentially Weight Normalized Smooth Homogeneous Neural Nets",
                        "abstract": "We analyze the inductive bias of gradient descent for weight normalized smooth homogeneous neural nets, when trained on exponential or cross-entropy loss. Our analysis focuses on exponential weight normalization (EWN), which encourages weight updates along the radial direction. This paper shows that the gradient flow path with EWN is equivalent to gradient flow on standard networks with an adaptive learning rate, and hence causes the weights to be updated in a way that prefers asymptotic relative sparsity. These results can be extended to hold for gradient descent via an appropriate adaptive learning rate. The asymptotic convergence rate of the loss in this setting is given by $\\Theta(\\frac{1}{t(\\log t)^2})$, and is independent of the depth of the network. We contrast these results with the inductive bias of standard weight normalization (SWN) and unnormalized architectures, and demonstrate their implications on synthetic data sets.Experimental results on simple data sets and architectures support our claim on sparse EWN solutions, even with SGD. This demonstrates its potential applications in learning prunable neural networks."
                    },
                    {
                        "title": "Using noise resilience for ranking generalization of deep neural networks",
                        "abstract": "Recent papers have shown that sufficiently overparameterized neural networks can perfectly fit even random labels. Thus, it is crucial to understand the underlying reason behind the generalization performance of a network on real-world data. In this work, we propose several measures to predict the generalization error of a network given the training data and its parameters. Using one of these measures, based on noise resilience of the network, we secured 5th position in the predicting generalization in deep learning (PGDL) competition at NeurIPS 2020."
                    },
                    {
                        "title": "Inductive Bias of Gradient Descent for Weight Normalized Smooth Homogeneous Neural Nets",
                        "abstract": "We analyze the inductive bias of gradient descent for weight normalized smooth homogeneous neural nets, when trained on exponential or cross-entropy loss. We analyse both standard weight normalization (SWN) and exponential weight normalization (EWN), and show that the gradient flow path with EWN is equivalent to gradient flow on standard networks with an adaptive learning rate. We extend these results to gradient descent, and establish asymptotic relations between weights and gradients for both SWN and EWN. We also show that EWN causes weights to be updated in a way that prefers asymptotic relative sparsity. For EWN, we provide a finite-time convergence rate of the loss with gradient flow and a tight asymptotic convergence rate with gradient descent. We demonstrate our results for SWN and EWN on synthetic data sets. Experimental results on simple datasets support our claim on sparse EWN solutions, even with SGD. This demonstrates its potential applications in learning neural networks amenable to pruning."
                    }
                ]
            },
            "ad5e76d4-97c5-4971-a71a-e919833c2784": {
                "pk": "ad5e76d4-97c5-4971-a71a-e919833c2784",
                "name": "Benjamin L. Edelman",
                "collaborators": [
                    "Eran Malach",
                    "Surbhi Goel",
                    "S. Kakade",
                    "Cyril Zhang",
                    "Boaz Barak",
                    "Ezra Edelman",
                    "Nikolaos Tsilivis",
                    "Edwin Zhang",
                    "Vincent Zhu",
                    "Naomi Saphra"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Deep Learning",
                    "Generative Models",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "The Evolution of Statistical Induction Heads: In-Context Learning Markov Chains",
                        "abstract": "Large language models have the ability to generate text that mimics patterns in their inputs. We introduce a simple Markov Chain sequence modeling task in order to study how this in-context learning (ICL) capability emerges. In our setting, each example is sampled from a Markov chain drawn from a prior distribution over Markov chains. Transformers trained on this task form \\emph{statistical induction heads} which compute accurate next-token probabilities given the bigram statistics of the context. During the course of training, models pass through multiple phases: after an initial stage in which predictions are uniform, they learn to sub-optimally predict using in-context single-token statistics (unigrams); then, there is a rapid phase transition to the correct in-context bigram solution. We conduct an empirical and theoretical investigation of this multi-phase process, showing how successful learning results from the interaction between the transformer's layers, and uncovering evidence that the presence of the simpler unigram solution may delay formation of the final bigram solution. We examine how learning is affected by varying the prior distribution over Markov chains, and consider the generalization of our in-context learning of Markov chains (ICL-MC) task to $n$-grams for $n>2$."
                    },
                    {
                        "title": "Transcendence: Generative Models Can Outperform The Experts That Train Them",
                        "abstract": "Generative models are trained with the simple objective of imitating the conditional probability distribution induced by the data they are trained on. Therefore, when trained on data generated by humans, we may not expect the artificial model to outperform the humans on their original objectives. In this work, we study the phenomenon of transcendence: when a generative model achieves capabilities that surpass the abilities of the experts generating its data. We demonstrate transcendence by training an autoregressive transformer to play chess from game transcripts, and show that the trained model can sometimes achieve better performance than all players in the dataset. We theoretically prove that transcendence can be enabled by low-temperature sampling, and rigorously assess this claim experimentally. Finally, we discuss other sources of transcendence, laying the groundwork for future investigation of this phenomenon in a broader setting."
                    },
                    {
                        "title": "Distinguishing the Knowable from the Unknowable with Language Models",
                        "abstract": "We study the feasibility of identifying epistemic uncertainty (reflecting a lack of knowledge), as opposed to aleatoric uncertainty (reflecting entropy in the underlying distribution), in the outputs of large language models (LLMs) over free-form text. In the absence of ground-truth probabilities, we explore a setting where, in order to (approximately) disentangle a given LLM's uncertainty, a significantly larger model stands in as a proxy for the ground truth. We show that small linear probes trained on the embeddings of frozen, pretrained models accurately predict when larger models will be more confident at the token level and that probes trained on one text domain generalize to others. Going further, we propose a fully unsupervised method that achieves non-trivial accuracy on the same task. Taken together, we interpret these results as evidence that LLMs naturally contain internal representations of different types of uncertainty that could potentially be leveraged to devise more informative indicators of model confidence in diverse practical settings."
                    },
                    {
                        "title": "Pareto Frontiers in Neural Feature Learning: Data, Compute, Width, and Luck",
                        "abstract": "In modern deep learning, algorithmic choices (such as width, depth, and learning rate) are known to modulate nuanced resource tradeoffs. This work investigates how these complexities necessarily arise for feature learning in the presence of computational-statistical gaps. We begin by considering offline sparse parity learning, a supervised classification problem which admits a statistical query lower bound for gradient-based training of a multilayer perceptron. This lower bound can be interpreted as a multi-resource tradeoff frontier: successful learning can only occur if one is sufficiently rich (large model), knowledgeable (large dataset), patient (many training iterations), or lucky (many random guesses). We show, theoretically and experimentally, that sparse initialization and increasing network width yield significant improvements in sample efficiency in this setting. Here, width plays the role of parallel search: it amplifies the probability of finding\"lottery ticket\"neurons, which learn sparse features more sample-efficiently. Finally, we show that the synthetic sparse parity task can be useful as a proxy for real problems requiring axis-aligned feature learning. We demonstrate improved sample efficiency on tabular classification benchmarks by using wide, sparsely-initialized MLP models; these networks sometimes outperform tuned random forests."
                    },
                    {
                        "title": "Watermarks in the Sand: Impossibility of Strong Watermarking for Generative Models",
                        "abstract": "Watermarking generative models consists of planting a statistical signal (watermark) in a model's output so that it can be later verified that the output was generated by the given model. A strong watermarking scheme satisfies the property that a computationally bounded attacker cannot erase the watermark without causing significant quality degradation. In this paper, we study the (im)possibility of strong watermarking schemes. We prove that, under well-specified and natural assumptions, strong watermarking is impossible to achieve. This holds even in the private detection algorithm setting, where the watermark insertion and detection algorithms share a secret key, unknown to the attacker. To prove this result, we introduce a generic efficient watermark attack; the attacker is not required to know the private key of the scheme or even which scheme is used. Our attack is based on two assumptions: (1) The attacker has access to a\"quality oracle\"that can evaluate whether a candidate output is a high-quality response to a prompt, and (2) The attacker has access to a\"perturbation oracle\"which can modify an output with a nontrivial probability of maintaining quality, and which induces an efficiently mixing random walk on high-quality outputs. We argue that both assumptions can be satisfied in practice by an attacker with weaker computational capabilities than the watermarked model itself, to which the attacker has only black-box access. Furthermore, our assumptions will likely only be easier to satisfy over time as models grow in capabilities and modalities. We demonstrate the feasibility of our attack by instantiating it to attack three existing watermarking schemes for large language models: Kirchenbauer et al. (2023), Kuditipudi et al. (2023), and Zhao et al. (2023). The same attack successfully removes the watermarks planted by all three schemes, with only minor quality degradation."
                    },
                    {
                        "title": "Pareto Frontiers in Deep Feature Learning: Data, Compute, Width, and Luck",
                        "abstract": "In modern deep learning, algorithmic choices (such as width, depth, and learning rate) are known to modulate nuanced resource tradeoffs. This work investigates how these complexities necessarily arise for feature learning in the presence of computational-statistical gaps. We begin by considering offline sparse parity learning, a supervised classification problem which admits a statistical query lower bound for gradient-based training of a multilayer perceptron. This lower bound can be interpreted as a multi-resource tradeoff frontier : successful learning can only occur if one is sufficiently rich (large model), knowledgeable (large dataset), patient (many training iterations), or lucky (many random guesses). We show, theoretically and experimentally, that sparse initialization and increasing network width yield significant improvements in sample efficiency in this setting. Here, width plays the role of parallel search: it amplifies the probability of finding \u201clottery ticket\u201d neurons, which learn sparse features more sample-efficiently. Finally, we show that the synthetic sparse parity task can be useful as a proxy for real problems requiring axis-aligned feature learning. We demonstrate improved sample efficiency on tabular classification benchmarks by using wide, sparsely-initialized MLP models; these networks sometimes outperform tuned random forests."
                    },
                    {
                        "title": "Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limit",
                        "abstract": "There is mounting evidence of emergent phenomena in the capabilities of deep learning methods as we scale up datasets, model sizes, and training times. While there are some accounts of how these resources modulate statistical capacity, far less is known about their effect on the computational problem of model training. This work conducts such an exploration through the lens of learning a $k$-sparse parity of $n$ bits, a canonical discrete search problem which is statistically easy but computationally hard. Empirically, we find that a variety of neural networks successfully learn sparse parities, with discontinuous phase transitions in the training curves. On small instances, learning abruptly occurs at approximately $n^{O(k)}$ iterations; this nearly matches SQ lower bounds, despite the apparent lack of a sparse prior. Our theoretical analysis shows that these observations are not explained by a Langevin-like mechanism, whereby SGD\"stumbles in the dark\"until it finds the hidden set of features (a natural algorithm which also runs in $n^{O(k)}$ time). Instead, we show that SGD gradually amplifies the sparse solution via a Fourier gap in the population gradient, making continual progress that is invisible to loss and error metrics."
                    }
                ]
            },
            "bc6678b7-4725-4844-8a7c-20c3e02fcdc2": {
                "pk": "bc6678b7-4725-4844-8a7c-20c3e02fcdc2",
                "name": "Costin-Andrei Oncescu",
                "collaborators": [
                    "S. Purandare",
                    "Stratos Idreos",
                    "S. Kakade"
                ],
                "domain": [
                    "Machine Learning",
                    "Natural Language Processing",
                    "Transformers",
                    "Efficient Algorithms"
                ],
                "publications": [
                    {
                        "title": "Flash Inference: Near Linear Time Inference for Long Convolution Sequence Models and Beyond",
                        "abstract": "While transformers have been at the core of most recent advancements in sequence generative models, their computational cost remains quadratic in sequence length. Several subquadratic architectures have been proposed to address this computational issue. Some of them, including long convolution sequence models (LCSMs), such as Hyena, address this issue at training time but remain quadratic during inference. We propose a method for speeding up LCSMs' exact inference to quasilinear $O(L\\log^2L)$ time, identify the key properties that make this possible, and propose a general framework that exploits these. Our approach, inspired by previous work on relaxed polynomial interpolation, is based on a tiling which helps decrease memory movement and share computation. It has the added benefit of allowing for almost complete parallelization across layers of the position-mixing part of the architecture. Empirically, we provide a proof of concept implementation for Hyena, which gets up to $1.6\\times$ end-to-end improvement over standard inference by improving $50\\times$ within the position-mixing part."
                    }
                ]
            },
            "689a6b82-e67c-4d32-af58-9e36c207251f": {
                "pk": "689a6b82-e67c-4d32-af58-9e36c207251f",
                "name": "Rosie Zhao",
                "collaborators": [
                    "Depen Morwani",
                    "David Brandfonbrener",
                    "Nikhil Vyas",
                    "S. Kakade",
                    "Itai Shapira",
                    "Lucas Janson"
                ],
                "domain": [
                    "Optimization",
                    "Language Modeling",
                    "Deep Learning",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Deconstructing What Makes a Good Optimizer for Language Models",
                        "abstract": "Training language models becomes increasingly expensive with scale, prompting numerous attempts to improve optimization efficiency. Despite these efforts, the Adam optimizer remains the most widely used, due to a prevailing view that it is the most effective approach. We aim to compare several optimization algorithms, including SGD, Adafactor, Adam, and Lion, in the context of autoregressive language modeling across a range of model sizes, hyperparameters, and architecture variants. Our findings indicate that, except for SGD, these algorithms all perform comparably both in their optimal performance and also in terms of how they fare across a wide range of hyperparameter choices. Our results suggest to practitioners that the choice of optimizer can be guided by practical considerations like memory constraints and ease of implementation, as no single algorithm emerged as a clear winner in terms of performance or stability to hyperparameter misspecification. Given our findings, we further dissect these approaches, examining two simplified versions of Adam: a) signed momentum (Signum) which we see recovers both the performance and hyperparameter stability of Adam and b) Adalayer, a layerwise variant of Adam which we introduce to study Adam's preconditioning. Examining Adalayer leads us to the conclusion that the largest impact of Adam's preconditioning is restricted to the last layer and LayerNorm parameters, and, perhaps surprisingly, the remaining layers can be trained with SGD."
                    },
                    {
                        "title": "SOAP: Improving and Stabilizing Shampoo using Adam",
                        "abstract": "There is growing evidence of the effectiveness of Shampoo, a higher-order preconditioning method, over Adam in deep learning optimization tasks. However, Shampoo's drawbacks include additional hyperparameters and computational overhead when compared to Adam, which only updates running averages of first- and second-moment quantities. This work establishes a formal connection between Shampoo (implemented with the 1/2 power) and Adafactor -- a memory-efficient approximation of Adam -- showing that Shampoo is equivalent to running Adafactor in the eigenbasis of Shampoo's preconditioner. This insight leads to the design of a simpler and computationally efficient algorithm: $\\textbf{S}$hampo$\\textbf{O}$ with $\\textbf{A}$dam in the $\\textbf{P}$reconditioner's eigenbasis (SOAP). With regards to improving Shampoo's computational efficiency, the most straightforward approach would be to simply compute Shampoo's eigendecomposition less frequently. Unfortunately, as our empirical results show, this leads to performance degradation that worsens with this frequency. SOAP mitigates this degradation by continually updating the running average of the second moment, just as Adam does, but in the current (slowly changing) coordinate basis. Furthermore, since SOAP is equivalent to running Adam in a rotated space, it introduces only one additional hyperparameter (the preconditioning frequency) compared to Adam. We empirically evaluate SOAP on language model pre-training with 360m and 660m sized models. In the large batch regime, SOAP reduces the number of iterations by over 40% and wall clock time by over 35% compared to AdamW, with approximately 20% improvements in both metrics compared to Shampoo. An implementation of SOAP is available at https://github.com/nikhilvyas/SOAP."
                    }
                ]
            },
            "7890d444-7e94-4520-8f94-611107ef63f5": {
                "pk": "7890d444-7e94-4520-8f94-611107ef63f5",
                "name": "Sham Kakade",
                "collaborators": [
                    "David Brandfonbrener",
                    "Hanlin Zhang",
                    "Nikhil Vyas",
                    "Eran Malach",
                    "Depen Morwani",
                    "Samy Jelassi",
                    "Jingfeng Wu",
                    "Udaya Ghai",
                    "Dean Foster",
                    "Yuanzhi Li"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Optimization",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "How Does Critical Batch Size Scale in Pre-training?",
                        "abstract": "Training large-scale models under given resources requires careful design of parallelism strategies. In particular, the efficiency notion of critical batch size, concerning the compromise between time and compute, marks the threshold beyond which greater data parallelism leads to diminishing returns. To operationalize it, we propose a measure of CBS and pre-train a series of auto-regressive language models, ranging from 85 million to 1.2 billion parameters, on the C4 dataset. Through extensive hyper-parameter sweeps and careful control on factors such as batch size, momentum, and learning rate along with its scheduling, we systematically investigate the impact of scale on CBS. Then we fit scaling laws with respect to model and data sizes to decouple their effects. Overall, our results demonstrate that CBS scales primarily with data size rather than model size, a finding we justify theoretically through the analysis of infinite-width limits of neural networks and infinite-dimensional least squares regression. Of independent interest, we highlight the importance of common hyper-parameter choices and strategies for studying large-scale pre-training beyond fixed training durations."
                    },
                    {
                        "title": "DataComp-LM: In search of the next generation of training sets for language models",
                        "abstract": "We introduce DataComp for Language Models (DCLM), a testbed for controlled dataset experiments with the goal of improving language models. As part of DCLM, we provide a standardized corpus of 240T tokens extracted from Common Crawl, effective pretraining recipes based on the OpenLM framework, and a broad suite of 53 downstream evaluations. Participants in the DCLM benchmark can experiment with data curation strategies such as deduplication, filtering, and data mixing at model scales ranging from 412M to 7B parameters. As a baseline for DCLM, we conduct extensive experiments and find that model-based filtering is key to assembling a high-quality training set. The resulting dataset, DCLM-Baseline enables training a 7B parameter language model from scratch to 64% 5-shot accuracy on MMLU with 2.6T training tokens. Compared to MAP-Neo, the previous state-of-the-art in open-data language models, DCLM-Baseline represents a 6.6 percentage point improvement on MMLU while being trained with 40% less compute. Our baseline model is also comparable to Mistral-7B-v0.3 and Llama 3 8B on MMLU (63%&66%), and performs similarly on an average of 53 natural language understanding tasks while being trained with 6.6x less compute than Llama 3 8B. Our results highlight the importance of dataset design for training language models and offer a starting point for further research on data curation."
                    },
                    {
                        "title": "Repeat After Me: Transformers are Better than State Space Models at Copying",
                        "abstract": "Transformers are the dominant architecture for sequence modeling, but there is growing interest in models that use a fixed-size latent state that does not depend on the sequence length, which we refer to as\"generalized state space models\"(GSSMs). In this paper we show that while GSSMs are promising in terms of inference-time efficiency, they are limited compared to transformer models on tasks that require copying from the input context. We start with a theoretical analysis of the simple task of string copying and prove that a two layer transformer can copy strings of exponential length while GSSMs are fundamentally limited by their fixed-size latent state. Empirically, we find that transformers outperform GSSMs in terms of efficiency and generalization on synthetic tasks that require copying the context. Finally, we evaluate pretrained large language models and find that transformer models dramatically outperform state space models at copying and retrieving information from context. Taken together, these results suggest a fundamental gap between transformers and GSSMs on tasks of practical interest."
                    },
                    {
                        "title": "CoLoR-Filter: Conditional Loss Reduction Filtering for Targeted Language Model Pre-training",
                        "abstract": "Selecting high-quality data for pre-training is crucial in shaping the downstream task performance of language models. A major challenge lies in identifying this optimal subset, a problem generally considered intractable, thus necessitating scalable and effective heuristics. In this work, we propose a data selection method, CoLoR-Filter (Conditional Loss Reduction Filtering), which leverages an empirical Bayes-inspired approach to derive a simple and computationally efficient selection criterion based on the relative loss values of two auxiliary models. In addition to the modeling rationale, we evaluate CoLoR-Filter empirically on two language modeling tasks: (1) selecting data from C4 for domain adaptation to evaluation on Books and (2) selecting data from C4 for a suite of downstream multiple-choice question answering tasks. We demonstrate favorable scaling both as we subselect more aggressively and using small auxiliary models to select data for large target models. As one headline result, CoLoR-Filter data selected using a pair of 150m parameter auxiliary models can train a 1.2b parameter target model to match a 1.2b parameter model trained on 25b randomly selected tokens with 25x less data for Books and 11x less data for the downstream tasks. Code: https://github.com/davidbrandfonbrener/color-filter-olmo Filtered data: https://huggingface.co/datasets/davidbrandfonbrener/color-filtered-c4"
                    },
                    {
                        "title": "Transcendence: Generative Models Can Outperform The Experts That Train Them",
                        "abstract": "Generative models are trained with the simple objective of imitating the conditional probability distribution induced by the data they are trained on. Therefore, when trained on data generated by humans, we may not expect the artificial model to outperform the humans on their original objectives. In this work, we study the phenomenon of transcendence: when a generative model achieves capabilities that surpass the abilities of the experts generating its data. We demonstrate transcendence by training an autoregressive transformer to play chess from game transcripts, and show that the trained model can sometimes achieve better performance than all players in the dataset. We theoretically prove that transcendence can be enabled by low-temperature sampling, and rigorously assess this claim experimentally. Finally, we discuss other sources of transcendence, laying the groundwork for future investigation of this phenomenon in a broader setting."
                    },
                    {
                        "title": "Mixture of Parrots: Experts improve memorization more than reasoning",
                        "abstract": "The Mixture-of-Experts (MoE) architecture enables a significant increase in the total number of model parameters with minimal computational overhead. However, it is not clear what performance tradeoffs, if any, exist between MoEs and standard dense transformers. In this paper, we show that as we increase the number of experts (while fixing the number of active parameters), the memorization performance consistently increases while the reasoning capabilities saturate. We begin by analyzing the theoretical limitations of MoEs at reasoning. We prove that there exist graph problems that cannot be solved by any number of experts of a certain width; however, the same task can be easily solved by a dense model with a slightly larger width. On the other hand, we find that on memory-intensive tasks, MoEs can effectively leverage a small number of active parameters with a large number of experts to memorize the data. We empirically validate these findings on synthetic graph problems and memory-intensive closed book retrieval tasks. Lastly, we pre-train a series of MoEs and dense transformers and evaluate them on commonly used benchmarks in math and natural language. We find that increasing the number of experts helps solve knowledge-intensive tasks, but fails to yield the same benefits for reasoning tasks."
                    },
                    {
                        "title": "Eliminating Position Bias of Language Models: A Mechanistic Approach",
                        "abstract": "Position bias has proven to be a prevalent issue of modern language models (LMs), where the models prioritize content based on its position within the given context. This bias often leads to unexpected model failures and hurts performance, robustness, and reliability across various applications. Our mechanistic analysis attributes the position bias to two components employed in nearly all state-of-the-art LMs: causal attention and relative positional encodings. Based on the analyses, we propose to eliminate position bias (e.g., different retrieved documents' orders in QA affect performance) with a training-free zero-shot approach. Our method changes the causal attention to bidirectional attention between documents and utilizes model attention values to decide the relative orders of documents instead of using the order provided in input prompts, therefore enabling Position-INvariant inferencE (PINE) at the document level. By eliminating position bias, models achieve better performance and reliability in downstream tasks, including LM-as-a-judge, retrieval-augmented QA, molecule generation, and math reasoning. Notably, PINE is especially useful when adapting LMs for evaluating reasoning pairs: it consistently provides 8 to 10 percentage points performance gains, making Llama-3-70B-Instruct perform even better than GPT-4-0125-preview and GPT-4o-2024-08-06 on the RewardBench reasoning set."
                    },
                    {
                        "title": "A New Perspective on Shampoo's Preconditioner",
                        "abstract": "Shampoo, a second-order optimization algorithm which uses a Kronecker product preconditioner, has recently garnered increasing attention from the machine learning community. The preconditioner used by Shampoo can be viewed either as an approximation of the Gauss--Newton component of the Hessian or the covariance matrix of the gradients maintained by Adagrad. We provide an explicit and novel connection between the $\\textit{optimal}$ Kronecker product approximation of these matrices and the approximation made by Shampoo. Our connection highlights a subtle but common misconception about Shampoo's approximation. In particular, the $\\textit{square}$ of the approximation used by the Shampoo optimizer is equivalent to a single step of the power iteration algorithm for computing the aforementioned optimal Kronecker product approximation. Across a variety of datasets and architectures we empirically demonstrate that this is close to the optimal Kronecker product approximation. Additionally, for the Hessian approximation viewpoint, we empirically study the impact of various practical tricks to make Shampoo more computationally efficient (such as using the batch gradient and the empirical Fisher) on the quality of Hessian approximation."
                    },
                    {
                        "title": "Follow My Instruction and Spill the Beans: Scalable Data Extraction from Retrieval-Augmented Generation Systems",
                        "abstract": "Retrieval-Augmented Generation (RAG) improves pre-trained models by incorporating external knowledge at test time to enable customized adaptation. We study the risk of datastore leakage in Retrieval-In-Context RAG Language Models (LMs). We show that an adversary can exploit LMs' instruction-following capabilities to easily extract text data verbatim from the datastore of RAG systems built with instruction-tuned LMs via prompt injection. The vulnerability exists for a wide range of modern LMs that span Llama2, Mistral/Mixtral, Vicuna, SOLAR, WizardLM, Qwen1.5, and Platypus2, and the exploitability exacerbates as the model size scales up. We also study multiple effects of RAG setup on the extractability of data, indicating that following unexpected instructions to regurgitate data can be an outcome of failure in effectively utilizing contexts for modern LMs, and further show that such vulnerability can be greatly mitigated by position bias elimination strategies. Extending our study to production RAG models GPTs, we design an attack that can cause datastore leakage with a 100% success rate on 25 randomly selected customized GPTs with at most 2 queries, and we extract text data verbatim at a rate of 41% from a book of 77,000 words and 3% from a corpus of 1,569,000 words by prompting the GPTs with only 100 queries generated by themselves."
                    },
                    {
                        "title": "Multi-Agent Reinforcement Learning from Human Feedback: Data Coverage and Algorithmic Techniques",
                        "abstract": "We initiate the study of Multi-Agent Reinforcement Learning from Human Feedback (MARLHF), exploring both theoretical foundations and empirical validations. We define the task as identifying Nash equilibrium from a preference-only offline dataset in general-sum games, a problem marked by the challenge of sparse feedback signals. Our theory establishes the upper complexity bounds for Nash Equilibrium in effective MARLHF, demonstrating that single-policy coverage is inadequate and highlighting the importance of unilateral dataset coverage. These theoretical insights are verified through comprehensive experiments. To enhance the practical performance, we further introduce two algorithmic techniques. (1) We propose a Mean Squared Error (MSE) regularization along the time axis to achieve a more uniform reward distribution and improve reward learning outcomes. (2) We utilize imitation learning to approximate the reference policy, ensuring stability and effectiveness in training. Our findings underscore the multifaceted approach required for MARLHF, paving the way for effective preference-based multi-agent systems."
                    },
                    {
                        "title": "Flash Inference: Near Linear Time Inference for Long Convolution Sequence Models and Beyond",
                        "abstract": "While transformers have been at the core of most recent advancements in sequence generative models, their computational cost remains quadratic in sequence length. Several subquadratic architectures have been proposed to address this computational issue. Some of them, including long convolution sequence models (LCSMs), such as Hyena, address this issue at training time but remain quadratic during inference. We propose a method for speeding up LCSMs' exact inference to quasilinear $O(L\\log^2L)$ time, identify the key properties that make this possible, and propose a general framework that exploits these. Our approach, inspired by previous work on relaxed polynomial interpolation, is based on a tiling which helps decrease memory movement and share computation. It has the added benefit of allowing for almost complete parallelization across layers of the position-mixing part of the architecture. Empirically, we provide a proof of concept implementation for Hyena, which gets up to $1.6\\times$ end-to-end improvement over standard inference by improving $50\\times$ within the position-mixing part."
                    },
                    {
                        "title": "Deconstructing What Makes a Good Optimizer for Language Models",
                        "abstract": "Training language models becomes increasingly expensive with scale, prompting numerous attempts to improve optimization efficiency. Despite these efforts, the Adam optimizer remains the most widely used, due to a prevailing view that it is the most effective approach. We aim to compare several optimization algorithms, including SGD, Adafactor, Adam, and Lion, in the context of autoregressive language modeling across a range of model sizes, hyperparameters, and architecture variants. Our findings indicate that, except for SGD, these algorithms all perform comparably both in their optimal performance and also in terms of how they fare across a wide range of hyperparameter choices. Our results suggest to practitioners that the choice of optimizer can be guided by practical considerations like memory constraints and ease of implementation, as no single algorithm emerged as a clear winner in terms of performance or stability to hyperparameter misspecification. Given our findings, we further dissect these approaches, examining two simplified versions of Adam: a) signed momentum (Signum) which we see recovers both the performance and hyperparameter stability of Adam and b) Adalayer, a layerwise variant of Adam which we introduce to study Adam's preconditioning. Examining Adalayer leads us to the conclusion that the largest impact of Adam's preconditioning is restricted to the last layer and LayerNorm parameters, and, perhaps surprisingly, the remaining layers can be trained with SGD."
                    },
                    {
                        "title": "SOAP: Improving and Stabilizing Shampoo using Adam",
                        "abstract": "There is growing evidence of the effectiveness of Shampoo, a higher-order preconditioning method, over Adam in deep learning optimization tasks. However, Shampoo's drawbacks include additional hyperparameters and computational overhead when compared to Adam, which only updates running averages of first- and second-moment quantities. This work establishes a formal connection between Shampoo (implemented with the 1/2 power) and Adafactor -- a memory-efficient approximation of Adam -- showing that Shampoo is equivalent to running Adafactor in the eigenbasis of Shampoo's preconditioner. This insight leads to the design of a simpler and computationally efficient algorithm: $\\textbf{S}$hampo$\\textbf{O}$ with $\\textbf{A}$dam in the $\\textbf{P}$reconditioner's eigenbasis (SOAP). With regards to improving Shampoo's computational efficiency, the most straightforward approach would be to simply compute Shampoo's eigendecomposition less frequently. Unfortunately, as our empirical results show, this leads to performance degradation that worsens with this frequency. SOAP mitigates this degradation by continually updating the running average of the second moment, just as Adam does, but in the current (slowly changing) coordinate basis. Furthermore, since SOAP is equivalent to running Adam in a rotated space, it introduces only one additional hyperparameter (the preconditioning frequency) compared to Adam. We empirically evaluate SOAP on language model pre-training with 360m and 660m sized models. In the large batch regime, SOAP reduces the number of iterations by over 40% and wall clock time by over 35% compared to AdamW, with approximately 20% improvements in both metrics compared to Shampoo. An implementation of SOAP is available at https://github.com/nikhilvyas/SOAP."
                    },
                    {
                        "title": "Understanding Inverse Scaling and Emergence in Multitask Representation Learning",
                        "abstract": "Large language models exhibit strong multi-tasking capabilities, however, their learning dynamics as a function of task characteristics, sample size, and model complexity remain mysterious. For instance, it is known that, as the model size grows, large language models exhibit emerging abilities where certain tasks can abruptly jump from poor to respectable performance. Such phenomena motivate a deeper understanding of how individual tasks evolve during multitasking. To this aim, we study a multitask representation learning setup where tasks can have distinct distributions , quantified by their covariance priors. Through random matrix theory, we precisely characterize the optimal linear representation for few-shot learning that minimizes the average test risk in terms of task covariances. When tasks have equal sample sizes, we prove a reduction to an equivalent problem with a single effective covariance from which the individual task risks of the original problem can be deduced. Importantly, we introduce \u201c task competition \u201d to explain how tasks with dominant covariance eigen-spectrum emerge faster than others. We show that task competition can potentially explain the inverse scaling of certain tasks i.e. reduced test accuracy as the model grows. Overall, this work sheds light on the risk and emergence of individual tasks and uncovers new high-dimensional phenomena (including multiple-descent risk curves) that arise in multitask representation learning."
                    },
                    {
                        "title": "Superposed Decoding: Multiple Generations from a Single Autoregressive Inference Pass",
                        "abstract": "Many applications today provide users with multiple auto-complete drafts as they type, including GitHub's code completion, Gmail's smart compose, and Apple's messaging auto-suggestions. Under the hood, language models support this by running an autoregressive inference pass to provide a draft. Consequently, providing $k$ drafts to the user requires running an expensive language model $k$ times. To alleviate the computation cost of running $k$ inference passes, we propose Superposed Decoding, a new decoding algorithm that generates $k$ drafts at the computation cost of one autoregressive inference pass. We achieve this by feeding a superposition of the most recent token embeddings from the $k$ drafts as input to the next decoding step of the language model. At every inference step we combine the $k$ drafts with the top-$k$ tokens to get $k^2$ new drafts and cache the $k$ most likely options, using an n-gram interpolation with minimal compute overhead to filter out incoherent generations. Our experiments show that $k$ drafts from Superposed Decoding are at least as coherent and factual as Nucleus Sampling and Greedy Decoding respectively, while being at least $2.44\\times$ faster for $k\\ge3$. In a compute-normalized setting, user evaluations demonstrably favor text generated by Superposed Decoding over Nucleus Sampling. Superposed Decoding can also be combined with other decoding strategies, resulting in universal coverage gains when scaling inference time compute. Code and more examples open-sourced at https://github.com/RAIVNLab/SuperposedDecoding."
                    },
                    {
                        "title": "Scaling Laws in Linear Regression: Compute, Parameters, and Data",
                        "abstract": "Empirically, large-scale deep learning models often satisfy a neural scaling law: the test error of the trained model improves polynomially as the model size and data size grow. However, conventional wisdom suggests the test error consists of approximation, bias, and variance errors, where the variance error increases with model size. This disagrees with the general form of neural scaling laws, which predict that increasing model size monotonically improves performance. We study the theory of scaling laws in an infinite dimensional linear regression setup. Specifically, we consider a model with $M$ parameters as a linear function of sketched covariates. The model is trained by one-pass stochastic gradient descent (SGD) using $N$ data. Assuming the optimal parameter satisfies a Gaussian prior and the data covariance matrix has a power-law spectrum of degree $a>1$, we show that the reducible part of the test error is $\\Theta(M^{-(a-1)} + N^{-(a-1)/a})$. The variance error, which increases with $M$, is dominated by the other errors due to the implicit regularization of SGD, thus disappearing from the bound. Our theory is consistent with the empirical neural scaling laws and verified by numerical simulation."
                    },
                    {
                        "title": "Q-Probe: A Lightweight Approach to Reward Maximization for Language Models",
                        "abstract": "We present an approach called Q-probing to adapt a pre-trained language model to maximize a task-specific reward function. At a high level, Q-probing sits between heavier approaches such as finetuning and lighter approaches such as few shot prompting, but can also be combined with either. The idea is to learn a simple linear function on a model's embedding space that can be used to reweight candidate completions. We theoretically show that this sampling procedure is equivalent to a KL-constrained maximization of the Q-probe as the number of samples increases. To train the Q-probes we consider either reward modeling or a class of novel direct policy learning objectives based on importance weighted policy gradients. With this technique, we see gains in domains with ground-truth rewards (code generation) as well as implicit rewards defined by preference data, even outperforming finetuning in data-limited regimes. Moreover, a Q-probe can be trained on top of an API since it only assumes access to sampling and embeddings. Code: https://github.com/likenneth/q_probe ."
                    },
                    {
                        "title": "Matching the Statistical Query Lower Bound for k-sparse Parity Problems with Stochastic Gradient Descent",
                        "abstract": "The $k$-parity problem is a classical problem in computational complexity and algorithmic theory, serving as a key benchmark for understanding computational classes. In this paper, we solve the $k$-parity problem with stochastic gradient descent (SGD) on two-layer fully-connected neural networks. We demonstrate that SGD can efficiently solve the $k$-sparse parity problem on a $d$-dimensional hypercube ($k\\le O(\\sqrt{d})$) with a sample complexity of $\\tilde{O}(d^{k-1})$ using $2^{\\Theta(k)}$ neurons, thus matching the established $\\Omega(d^{k})$ lower bounds of Statistical Query (SQ) models. Our theoretical analysis begins by constructing a good neural network capable of correctly solving the $k$-parity problem. We then demonstrate how a trained neural network with SGD can effectively approximate this good network, solving the $k$-parity problem with small statistical errors. Our theoretical results and findings are supported by empirical evidence, showcasing the efficiency and efficacy of our approach."
                    },
                    {
                        "title": "Neural Coordination and Capacity Control for Inventory Management",
                        "abstract": "This paper addresses the capacitated periodic review inventory control problem, focusing on a retailer managing multiple products with limited shared resources, such as storage or inbound labor at a facility. Specifically, this paper is motivated by the questions of (1) what does it mean to backtest a capacity control mechanism, (2) can we devise and backtest a capacity control mechanism that is compatible with recent advances in deep reinforcement learning for inventory management? First, because we only have a single historic sample path of Amazon's capacity limits, we propose a method that samples from a distribution of possible constraint paths covering a space of real-world scenarios. This novel approach allows for more robust and realistic testing of inventory management strategies. Second, we extend the exo-IDP (Exogenous Decision Process) formulation of Madeka et al. 2022 to capacitated periodic review inventory control problems and show that certain capacitated control problems are no harder than supervised learning. Third, we introduce a `neural coordinator', designed to produce forecasts of capacity prices, guiding the system to adhere to target constraints in place of a traditional model predictive controller. Finally, we apply a modified DirectBackprop algorithm for learning a deep RL buying policy and a training the neural coordinator. Our methodology is evaluated through large-scale backtests, demonstrating RL buying policies with a neural coordinator outperforms classic baselines both in terms of cumulative discounted reward and capacity adherence (we see improvements of up to 50% in some cases)."
                    },
                    {
                        "title": "LoRA Soups: Merging LoRAs for Practical Skill Composition Tasks",
                        "abstract": "Low-Rank Adaptation (LoRA) is a popular technique for parameter-efficient fine-tuning of Large Language Models (LLMs). We study how different LoRA modules can be merged to achieve skill composition -- testing the performance of the merged model on a target task that involves combining multiple skills, each skill coming from a single LoRA. This setup is favorable when it is difficult to obtain training data for the target task and when it can be decomposed into multiple skills. First, we identify practically occurring use-cases that can be studied under the realm of skill composition, e.g. solving hard math-word problems with code, creating a bot to answer questions on proprietary manuals or about domain-specialized corpora. Our main contribution is to show that concatenation of LoRAs (CAT), which optimally averages LoRAs that were individually trained on different skills, outperforms existing model- and data- merging techniques; for instance on math-word problems, CAT beats these methods by an average of 43% and 12% respectively. Thus, this paper advocates model merging as an efficient way to solve compositional tasks and underscores CAT as a simple, compute-friendly and effective procedure. To our knowledge, this is the first work demonstrating the superiority of model merging over data mixing for binary skill composition tasks."
                    }
                ]
            }
        }
    },
    "2310.07707": {
        "paper_data": {
            "title": "MatFormer: Nested Transformer for Elastic Inference",
            "url": "http://arxiv.org/abs/2310.07707v1",
            "arxiv_id": "2310.07707",
            "authors": [
                "Devvrit",
                "Sneha Kudugunta",
                "Aditya Kusupati",
                "Tim Dettmers",
                "Kaifeng Chen",
                "Inderjit Dhillon",
                "Yulia Tsvetkov",
                "Hannaneh Hajishirzi",
                "Sham Kakade",
                "Ali Farhadi",
                "Prateek Jain"
            ],
            "abstract": "Transformer models are deployed in a wide range of settings, from multi-accelerator clusters to standalone mobile phones. The diverse inference constraints in these scenarios necessitate practitioners to train foundation models such as PaLM 2, Llama, & ViTs as a series of models of varying sizes. Due to significant training costs, only a select few model sizes are trained and supported, limiting more fine-grained control over relevant tradeoffs, including latency, cost, and accuracy. This work introduces MatFormer, a nested Transformer architecture designed to offer elasticity in a variety of deployment constraints. Each Feed Forward Network (FFN) block of a MatFormer model is jointly optimized with a few nested smaller FFN blocks. This training procedure allows for the Mix'n'Match of model granularities across layers -- i.e., a trained universal MatFormer model enables extraction of hundreds of accurate smaller models, which were never explicitly optimized. We empirically demonstrate MatFormer's effectiveness across different model classes (decoders & encoders), modalities (language & vision), and scales (up to 2.6B parameters). We find that a 2.6B decoder-only MatFormer language model (MatLM) allows us to extract smaller models spanning from 1.5B to 2.6B, each exhibiting comparable validation loss and one-shot downstream evaluations to their independently trained counterparts. Furthermore, we observe that smaller encoders extracted from a universal MatFormer-based ViT (MatViT) encoder preserve the metric-space structure for adaptive large-scale retrieval. Finally, we showcase that speculative decoding with the accurate and consistent submodels extracted from MatFormer can further reduce inference latency.",
            "introduction": "   1 Introduction  Figure 1: MatFormer introduces nested structure into the Transformer\u2019s FFN block & jointly trains all the submodels, enabling free extraction of hundreds of accurate submodels for elastic inference.   Large Foundation models\u00a0(Anil et\u00a0al., 2023; OpenAI, 2023; Dehghani et\u00a0al., 2023) are deployed in a variety of settings like real-time response on mobile phones or in batch setting on multi-cluster GPUs for web-scale serving. To handle such varied settings, each model family provides a few independently trained models of different sizes. In order to cover a wide range of applications, typically these models\u2019 sizes are nearly linear on log-scale. For example, Llama family provides models with 7B, 13B, 33B and 65B parameters\u00a0(Touvron et\u00a0al., 2023a).   Such an approach has two key drawbacks: (a) as the models are independently trained, they incur significant overhead for colocation during inference and are not behaviorally consistent with each other which are detrimental to inference optimization techniques like speculative decoding\u00a0(Leviathan et\u00a0al., 2023) and model cascades\u00a0(Wang et\u00a0al., 2020b), and (b) due to training overhead, practitioners typically train only a few models which do not cover the entire set of downstream use-cases. For example, a deployment setup might, say, have the latency budget to support 40B parameter Llama model, but can only host a 33B variant because the next bigger model (65B) has significantly higher latency. So, one would need to settle for a less accurate model despite the larger latency budget. While model compression approaches aim to address this issue, they typically require additional training for each model that needs to be extracted. Furthermore, when applied to LLMs, these techniques are known to significantly drop the accuracy\u00a0(Jaiswal et\u00a0al., 2023).   In this paper, we propose MatFormer, a natively elastic Transformer\u00a0(Vaswani et\u00a0al., 2023) architecture that allows for training one universal model which can be used to extract hundreds of smaller submodels without any additional training (Figure\u00a01). MatFormer is a general architecture that can be applied to both encoders and decoders, is domain agnostic, and is compatible with most design choices and training pipelines of large Transformer-based models \u2013 LLMs & ViTs.   MatFormer follows the principle of matryoshka representation learning\u00a0(Kusupati et\u00a0al., 2022), to introduce nested substructure inside the standard Transformer block. Formally, MatFormer defines a Transformer blocks Tisubscript\ud835\udc47\ud835\udc56T_{i}italic_T start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT, such that, T1\u2282T2\u2282\u22ef\u2282Tgsubscript\ud835\udc471subscript\ud835\udc472\u22efsubscript\ud835\udc47\ud835\udc54T_{1}\\subset T_{2}\\subset\\dots\\subset T_{g}italic_T start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT \u2282 italic_T start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT \u2282 \u22ef \u2282 italic_T start_POSTSUBSCRIPT italic_g end_POSTSUBSCRIPT, where g\ud835\udc54gitalic_g is the number of nested transformer blocks, and Ti\u2282Ti+1subscript\ud835\udc47\ud835\udc56subscript\ud835\udc47\ud835\udc561T_{i}\\subset T_{i+1}italic_T start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT \u2282 italic_T start_POSTSUBSCRIPT italic_i + 1 end_POSTSUBSCRIPT relation indicates that the parameters of Tisubscript\ud835\udc47\ud835\udc56T_{i}italic_T start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT are contained in those of Ti+1subscript\ud835\udc47\ud835\udc561T_{i+1}italic_T start_POSTSUBSCRIPT italic_i + 1 end_POSTSUBSCRIPT.   MatFormer can induce such sub-structure in both the attention and the feedforward network (FFN) blocks of the Transformer (see Figure\u00a01). Consider a FFN block that has dffsubscript\ud835\udc51ffd_{\\text{ff}}italic_d start_POSTSUBSCRIPT ff end_POSTSUBSCRIPT neurons in the hidden layer. Then, MatFormer induces matryoshka structure on these neurons, where Tisubscript\ud835\udc47\ud835\udc56T_{i}italic_T start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT contains the first misubscript\ud835\udc5a\ud835\udc56m_{i}italic_m start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT neurons and 1\u2264m1\u2264m2\u2062\u22ef\u2264mg=dff1subscript\ud835\udc5a1subscript\ud835\udc5a2\u22efsubscript\ud835\udc5a\ud835\udc54subscript\ud835\udc51ff1\\leq m_{1}\\leq m_{2}\\dots\\leq m_{g}=d_{\\text{ff}}1 \u2264 italic_m start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT \u2264 italic_m start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT \u22ef \u2264 italic_m start_POSTSUBSCRIPT italic_g end_POSTSUBSCRIPT = italic_d start_POSTSUBSCRIPT ff end_POSTSUBSCRIPT represent the number",
            "references": [
                {
                    "title": "SHARCS: Efficient Transformers through Routing with Dynamic Width Sub-networks",
                    "abstract": "We introduce SHARCS for adaptive inference that takes into account the hardness of input samples. SHARCS can train a router on any transformer network, enabling the model to direct different samples to sub-networks with varying widths. Our experiments demonstrate that: (1) SHARCS outperforms or complements existing per-sample adaptive inference methods across various classification tasks in terms of accuracy vs. FLOPs; (2) SHARCS generalizes across different architectures and can be even applied to compressed and efficient transformer encoders to further improve their efficiency; (3) SHARCS can provide a 2 times inference speed up at an insignificant drop in accuracy."
                },
                {
                    "title": "Compressing LLMs: The Truth is Rarely Pure and Never Simple",
                    "abstract": "Despite their remarkable achievements, modern Large Language Models (LLMs) face exorbitant computational and memory footprints. Recently, several works have shown significant success in training-free and data-free compression (pruning and quantization) of LLMs that achieve 50 - 60% sparsity and reduce the bit width to 3 or 4 bits per weight, with negligible degradation of perplexity over the uncompressed baseline. As recent research efforts are focused on developing increasingly sophisticated compression methods, our work takes a step back and re-evaluates the effectiveness of existing SoTA compression methods, which rely on a fairly simple and widely questioned metric, perplexity (even for dense LLMs). We introduce Knowledge-Intensive Compressed LLM BenchmarK (LLM-KICK), a collection of carefully curated tasks to redefine the evaluation protocol for compressed LLMs, which have significant alignment with their dense counterparts and perplexity fail to capture subtle change in their true capabilities. LLM-KICK unveils many favorable merits and unfortunate plights of current SoTA compression methods: all pruning methods suffer significant performance degradation, sometimes at trivial sparsity ratios (e.g., 25-30%), and fail for N:M sparsity in knowledge-intensive tasks; current quantization methods are more successful than pruning; yet, pruned LLMs even at $\\geq 50$% sparsity are robust in-context retrieval and summarization systems; among others. LLM-KICK is designed to holistically access compressed LLMs' ability for language understanding, reasoning, generation, in-context retrieval, in-context summarization, etc. We hope our study can foster the development of better LLM compression methods. The reproduced codes are available at https://github.com/VITA-Group/llm-kick."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "PaLM 2 Technical Report",
                    "abstract": "We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities and is more compute-efficient than its predecessor PaLM. PaLM 2 is a Transformer-based model trained using a mixture of objectives. Through extensive evaluations on English and multilingual language, and reasoning tasks, we demonstrate that PaLM 2 has significantly improved quality on downstream tasks across different model sizes, while simultaneously exhibiting faster and more efficient inference compared to PaLM. This improved efficiency enables broader deployment while also allowing the model to respond faster, for a more natural pace of interaction. PaLM 2 demonstrates robust reasoning capabilities exemplified by large improvements over PaLM on BIG-Bench and other reasoning tasks. PaLM 2 exhibits stable performance on a suite of responsible AI evaluations, and enables inference-time control over toxicity without additional overhead or impact on other capabilities. Overall, PaLM 2 achieves state-of-the-art performance across a diverse set of tasks and capabilities. When discussing the PaLM 2 family, it is important to distinguish between pre-trained models (of various sizes), fine-tuned variants of these models, and the user-facing products that use these models. In particular, user-facing products typically include additional pre- and post-processing steps. Additionally, the underlying models may evolve over time. Therefore, one should not expect the performance of user-facing products to exactly match the results reported in this report."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Full Stack Optimization of Transformer Inference: a Survey",
                    "abstract": "Recent advances in state-of-the-art DNN architecture design have been moving toward Transformer models. These models achieve superior accuracy across a wide range of applications. This trend has been consistent over the past several years since Transformer models were originally introduced. However, the amount of compute and bandwidth required for inference of recent Transformer models is growing at a significant rate, and this has made their deployment in latency-sensitive applications challenging. As such, there has been an increased focus on making Transformer models more efficient, with methods that range from changing the architecture design, all the way to developing dedicated domain-specific accelerators. In this work, we survey different approaches for efficient Transformer inference, including: (i) analysis and profiling of the bottlenecks in existing Transformer architectures and their similarities and differences with previous convolutional models; (ii) implications of Transformer architecture on hardware, including the impact of non-linear operations such as Layer Normalization, Softmax, and GELU, as well as linear operations, on hardware design; (iii) approaches for optimizing a fixed Transformer architecture; (iv) challenges in finding the right mapping and scheduling of operations for Transformer models; and (v) approaches for optimizing Transformer models by adapting the architecture using neural architecture search. Finally, we perform a case study by applying the surveyed optimizations on Gemmini, the open-source, full-stack DNN accelerator generator, and we show how each of these approaches can yield improvements, compared to previous benchmark results on Gemmini. Among other things, we find that a full-stack co-design approach with the aforementioned methods can result in up to 88.7x speedup with a minimal performance degradation for Transformer inference."
                },
                {
                    "title": "Scaling Vision Transformers to 22 Billion Parameters",
                    "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                },
                {
                    "title": "Accelerating Large Language Model Decoding with Speculative Sampling",
                    "abstract": "We present speculative sampling, an algorithm for accelerating transformer decoding by enabling the generation of multiple tokens from each transformer call. Our algorithm relies on the observation that the latency of parallel scoring of short continuations, generated by a faster but less powerful draft model, is comparable to that of sampling a single token from the larger target model. This is combined with a novel modified rejection sampling scheme which preserves the distribution of the target model within hardware numerics. We benchmark speculative sampling with Chinchilla, a 70 billion parameter language model, achieving a 2-2.5x decoding speedup in a distributed setup, without compromising the sample quality or making modifications to the model itself."
                },
                {
                    "title": "The case for 4-bit precision: k-bit Inference Scaling Laws",
                    "abstract": "Quantization methods reduce the number of bits required to represent each parameter in a model, trading accuracy for smaller memory footprints and inference latencies. However, the final model size depends on both the number of parameters of the original model and the rate of compression. For example, a 30B 8-bit model and a 60B 4-bit model have the same number of bits but may have very different zero-shot accuracies. In this work, we study this trade-off by developing inference scaling laws of zero-shot performance in Large Language Models (LLMs) to determine the bit-precision and model size that maximizes zero-shot performance. We run more than 35,000 experiments with 16-bit inputs and k-bit parameters to examine which zero-shot quantization methods improve scaling for 3 to 8-bit precision at scales of 19M to 176B parameters across the LLM families BLOOM, OPT, NeoX/Pythia, and GPT-2. We find that it is challenging to improve the bit-level scaling trade-off, with the only improvements being the use of a small block size -- splitting the parameters into small independently quantized blocks -- and the quantization data type being used (e.g., Int vs Float). Overall, our findings show that {4-bit} precision is almost universally optimal for total model bits and zero-shot accuracy."
                },
                {
                    "title": "FlexiViT: One Model for All Patch Sizes",
                    "abstract": "Vision Transformers convert images to sequences by slicing them into patches. The size of these patches controls a speed/accuracy tradeoff, with smaller patches leading to higher accuracy at greater computational cost, but changing the patch size typically requires retraining the model. In this paper, we demonstrate that simply randomizing the patch size at training time leads to a single set of weights that performs well across a wide range of patch sizes, making it possible to tailor the model to different compute budgets at deployment time. We extensively evaluate the resulting model, which we call FlexiViT, on a wide range of tasks, including classification, image-text retrieval, open-world detection, panoptic segmentation, and semantic segmentation, concluding that it usually matches, and sometimes outperforms, standard ViT models trained at a single patch size in an otherwise identical setup. Hence, FlexiViT training is a simple drop-in improvement for ViT that makes it easy to add compute-adaptive capabilities to most models relying on a ViT backbone architecture. Code and pre-trained models are available at github.com/google-research/big_vision."
                },
                {
                    "title": "Robust Speech Recognition via Large-Scale Weak Supervision",
                    "abstract": "We study the capabilities of speech processing systems trained simply to predict large amounts of transcripts of audio on the internet. When scaled to 680,000 hours of multilingual and multitask supervision, the resulting models generalize well to standard benchmarks and are often competitive with prior fully supervised results but in a zero-shot transfer setting without the need for any fine-tuning. When compared to humans, the models approach their accuracy and robustness. We are releasing models and inference code to serve as a foundation for further work on robust speech processing."
                },
                {
                    "title": "Fast Inference from Transformers via Speculative Decoding",
                    "abstract": "Inference from large autoregressive models like Transformers is slow - decoding K tokens takes K serial runs of the model. In this work we introduce speculative decoding - an algorithm to sample from autoregressive models faster without any changes to the outputs, by computing several tokens in parallel. At the heart of our approach lie the observations that (1) hard language-modeling tasks often include easier subtasks that can be approximated well by more efficient models, and (2) using speculative execution and a novel sampling method, we can make exact decoding from the large models faster, by running them in parallel on the outputs of the approximation models, potentially generating several tokens concurrently, and without changing the distribution. Our method can accelerate existing off-the-shelf models without retraining or architecture changes. We demonstrate it on T5-XXL and show a 2X-3X acceleration compared to the standard T5X implementation, with identical outputs."
                },
                {
                    "title": "Branch-Train-Merge: Embarrassingly Parallel Training of Expert Language Models",
                    "abstract": "We present Branch-Train-Merge (BTM), a communication-efficient algorithm for embarrassingly parallel training of large language models (LLMs). We show it is possible to independently train subparts of a new class of LLMs on different subsets of the data, eliminating the massive multi-node synchronization currently required to train LLMs. BTM learns a set of independent expert LMs (ELMs), each specialized to a different textual domain, such as scientific or legal text. These ELMs can be added and removed to update data coverage, ensembled to generalize to new domains, or averaged to collapse back to a single LM for efficient inference. New ELMs are learned by branching from (mixtures of) ELMs in the current set, further training the parameters on data for the new domain, and then merging the resulting model back into the set for future use. Experiments show that BTM improves in- and out-of-domain perplexities as compared to GPT-style Transformer LMs, when controlling for training cost. Through extensive analysis, we show that these results are robust to different ELM initialization schemes, but require expert domain specialization; LM ensembles with random data splits do not perform well. We also present a study of scaling BTM into a new corpus of 64 domains (192B whitespace-separated tokens in total); the resulting LM (22.4B total parameters) performs as well as a Transformer LM trained with 2.5 times more compute. These gains grow with the number of domains, suggesting more aggressive parallelism could be used to efficiently train larger models in future work."
                },
                {
                    "title": "Confident Adaptive Language Modeling",
                    "abstract": "Recent advances in Transformer-based large language models (LLMs) have led to significant performance improvements across many tasks. These gains come with a drastic increase in the models' size, potentially leading to slow and costly use at inference time. In practice, however, the series of generations made by LLMs is composed of varying levels of difficulty. While certain predictions truly benefit from the models' full capacity, other continuations are more trivial and can be solved with reduced compute. In this work, we introduce Confident Adaptive Language Modeling (CALM), a framework for dynamically allocating different amounts of compute per input and generation timestep. Early exit decoding involves several challenges that we address here, such as: (1) what confidence measure to use; (2) connecting sequence-level constraints to local per-token exit decisions; and (3) attending back to missing hidden representations due to early exits in previous tokens. Through theoretical analysis and empirical experiments on three diverse text generation tasks, we demonstrate the efficacy of our framework in reducing compute -- potential speedup of up to $\\times 3$ -- while provably maintaining high performance."
                },
                {
                    "title": "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness",
                    "abstract": "Transformers are slow and memory-hungry on long sequences, since the time and memory complexity of self-attention are quadratic in sequence length. Approximate attention methods have attempted to address this problem by trading off model quality to reduce the compute complexity, but often do not achieve wall-clock speedup. We argue that a missing principle is making attention algorithms IO-aware -- accounting for reads and writes between levels of GPU memory. We propose FlashAttention, an IO-aware exact attention algorithm that uses tiling to reduce the number of memory reads/writes between GPU high bandwidth memory (HBM) and GPU on-chip SRAM. We analyze the IO complexity of FlashAttention, showing that it requires fewer HBM accesses than standard attention, and is optimal for a range of SRAM sizes. We also extend FlashAttention to block-sparse attention, yielding an approximate attention algorithm that is faster than any existing approximate attention method. FlashAttention trains Transformers faster than existing baselines: 15% end-to-end wall-clock speedup on BERT-large (seq. length 512) compared to the MLPerf 1.1 training speed record, 3$\\times$ speedup on GPT-2 (seq. length 1K), and 2.4$\\times$ speedup on long-range arena (seq. length 1K-4K). FlashAttention and block-sparse FlashAttention enable longer context in Transformers, yielding higher quality models (0.7 better perplexity on GPT-2 and 6.4 points of lift on long-document classification) and entirely new capabilities: the first Transformers to achieve better-than-chance performance on the Path-X challenge (seq. length 16K, 61.4% accuracy) and Path-256 (seq. length 64K, 63.1% accuracy)."
                },
                {
                    "title": "Matryoshka Representation Learning",
                    "abstract": "Learned representations are a central component in modern ML systems, serving a multitude of downstream tasks. When training such representations, it is often the case that computational and statistical constraints for each downstream task are unknown. In this context rigid, fixed capacity representations can be either over or under-accommodating to the task at hand. This leads us to ask: can we design a flexible representation that can adapt to multiple downstream tasks with varying computational resources? Our main contribution is Matryoshka Representation Learning (MRL) which encodes information at different granularities and allows a single embedding to adapt to the computational constraints of downstream tasks. MRL minimally modifies existing representation learning pipelines and imposes no additional cost during inference and deployment. MRL learns coarse-to-fine representations that are at least as accurate and rich as independently trained low-dimensional representations. The flexibility within the learned Matryoshka Representations offer: (a) up to 14x smaller embedding size for ImageNet-1K classification at the same level of accuracy; (b) up to 14x real-world speed-ups for large-scale retrieval on ImageNet-1K and 4K; and (c) up to 2% accuracy improvements for long-tail few-shot classification, all while being as robust as the original representations. Finally, we show that MRL extends seamlessly to web-scale datasets (ImageNet, JFT) across various modalities -- vision (ViT, ResNet), vision + language (ALIGN) and language (BERT). MRL code and pretrained models are open-sourced at https://github.com/RAIVNLab/MRL."
                },
                {
                    "title": "Training Compute-Optimal Large Language Models",
                    "abstract": "We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over 400 language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, Chinchilla, that uses the same compute budget as Gopher but with 70B parameters and 4$\\times$ more more data. Chinchilla uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that Chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, Chinchilla reaches a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, greater than a 7% improvement over Gopher."
                },
                {
                    "title": "Dynamic ConvNets on Tiny Devices via Nested Sparsity",
                    "abstract": "This work introduces a new training and compression pipeline to build nested sparse convolutional neural networks (ConvNets), a class of dynamic ConvNets suited for inference tasks deployed on resource-constrained devices at the edge of the Internet of Things. A nested sparse ConvNet consists of a single ConvNet architecture, containing $N$ sparse subnetworks with nested weights subsets, like a Matryoshka doll, and can trade accuracy for latency at runtime, using the model sparsity as a dynamic knob. To attain high accuracy at training time, we propose a gradient masking technique that optimally routes the learning signals across the nested weight subsets. To minimize the storage footprint and efficiently process the obtained models at inference time, we introduce a new sparse matrix compression format with dedicated compute kernels that fruitfully exploit the characteristic of the nested weights subsets. Tested on image classification and object detection tasks on an off-the-shelf ARM-M7 microcontroller unit (MCU), nested sparse ConvNets outperform variable-latency solutions naively built assembling single sparse models trained as stand-alone instances, achieving 1) comparable accuracy; 2) remarkable storage savings; and 3) high performance. Moreover, when compared to state-of-the-art dynamic strategies, such as dynamic pruning and layer width scaling, nested sparse ConvNets turn out to be Pareto optimal in the accuracy versus latency space."
                },
                {
                    "title": "Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer",
                    "abstract": "Hyperparameter (HP) tuning in deep learning is an expensive process, prohibitively so for neural networks (NNs) with billions of parameters. We show that, in the recently discovered Maximal Update Parametrization (muP), many optimal HPs remain stable even as model size changes. This leads to a new HP tuning paradigm we call muTransfer: parametrize the target model in muP, tune the HP indirectly on a smaller model, and zero-shot transfer them to the full-sized model, i.e., without directly tuning the latter at all. We verify muTransfer on Transformer and ResNet. For example, 1) by transferring pretraining HPs from a model of 13M parameters, we outperform published numbers of BERT-large (350M parameters), with a total tuning cost equivalent to pretraining BERT-large once; 2) by transferring from 40M parameters, we outperform published numbers of the 6.7B GPT-3 model, with tuning cost only 7% of total pretraining cost. A Pytorch implementation of our technique can be found at github.com/microsoft/mup and installable via `pip install mup`."
                },
                {
                    "title": "LaMDA: Language Models for Dialog Applications",
                    "abstract": "We present LaMDA: Language Models for Dialog Applications. LaMDA is a family of Transformer-based neural language models specialized for dialog, which have up to 137B parameters and are pre-trained on 1.56T words of public dialog data and web text. While model scaling alone can improve quality, it shows less improvements on safety and factual grounding. We demonstrate that fine-tuning with annotated data and enabling the model to consult external knowledge sources can lead to significant improvements towards the two key challenges of safety and factual grounding. The first challenge, safety, involves ensuring that the model's responses are consistent with a set of human values, such as preventing harmful suggestions and unfair bias. We quantify safety using a metric based on an illustrative set of human values, and we find that filtering candidate responses using a LaMDA classifier fine-tuned with a small amount of crowdworker-annotated data offers a promising approach to improving model safety. The second challenge, factual grounding, involves enabling the model to consult external knowledge sources, such as an information retrieval system, a language translator, and a calculator. We quantify factuality using a groundedness metric, and we find that our approach enables the model to generate responses grounded in known sources, rather than responses that merely sound plausible. Finally, we explore the use of LaMDA in the domains of education and content recommendations, and analyze their helpfulness and role consistency."
                },
                {
                    "title": "Vision Transformer Slimming: Multi-Dimension Searching in Continuous Optimization Space",
                    "abstract": "This paper explores the feasibility of finding an optimal sub-model from a vision transformer and introduces a pure vision transformer slimming (ViT-Slim) framework. It can search a sub-structure from the original model end-to-end across multiple dimensions, including the input tokens, MHSA and MLP modules with state-of-the-art performance. Our method is based on a learnable and unified \u21131sparsity constraint with pre-defined factors to reflect the global importance in the continuous searching space of different dimensions. The searching process is highly efficient through a single-shot training scheme. For instance, on DeiT-S, ViT-Slim only takes ~43 GPU hours for the searching process, and the searched structure is flexible with diverse dimensionalities in different modules. Then, a budget threshold is employed according to the requirements of accuracy-FLOPs trade-off on running devices, and a retraining process is performed to obtain the final model. The extensive experiments show that our ViT-Slim can compress up to 40% of parameters and 40% FLOPs on various vision transformers while increasing the accuracy by ~0.6% on ImageNet. We also demonstrate the advantage of our searched models on several downstream datasets. Our code is available at https://github.com/Arnav0400/ViT-Slim."
                },
                {
                    "title": "GLaM: Efficient Scaling of Language Models with Mixture-of-Experts",
                    "abstract": "Scaling language models with more data, compute and parameters has driven significant progress in natural language processing. For example, thanks to scaling, GPT-3 was able to achieve strong results on in-context learning tasks. However, training these large dense models requires significant amounts of computing resources. In this paper, we propose and develop a family of language models named GLaM (Generalist Language Model), which uses a sparsely activated mixture-of-experts architecture to scale the model capacity while also incurring substantially less training cost compared to dense variants. The largest GLaM has 1.2 trillion parameters, which is approximately 7x larger than GPT-3. It consumes only 1/3 of the energy used to train GPT-3 and requires half of the computation flops for inference, while still achieving better overall zero-shot and one-shot performance across 29 NLP tasks."
                },
                {
                    "title": "SCENIC: A JAX Library for Computer Vision Research and Beyond",
                    "abstract": "Scenic is an open-source11https://github.com/google-research/scenic JAX library with a focus on transformer-based models for computer vision research and beyond. The goal of this toolkit is to facilitate rapid experimentation, prototyping, and research of new architectures and models. Scenic supports a diverse range of tasks (e.g., classification, segmentation, detection) and facilitates working on multi-modal problems, along with GPU/TPU support for large-scale, multi-host and multi-device training. Scenic also offers optimized implementations of state-of-the-art research models spanning a wide range of modalities. Scenic has been successfully used for numerous projects and published papers and continues serving as the library of choice for rapid prototyping and publication of new research ideas."
                },
                {
                    "title": "Network Augmentation for Tiny Deep Learning",
                    "abstract": "We introduce Network Augmentation (NetAug), a new training method for improving the performance of tiny neural networks. Existing regularization techniques (e.g., data augmentation, dropout) have shown much success on large neural networks by adding noise to overcome over-fitting. However, we found these techniques hurt the performance of tiny neural networks. We argue that training tiny models are different from large models: rather than augmenting the data, we should augment the model, since tiny models tend to suffer from under-fitting rather than over-fitting due to limited capacity. To alleviate this issue, NetAug augments the network (reverse dropout) instead of inserting noise into the dataset or the network. It puts the tiny model into larger models and encourages it to work as a sub-model of larger models to get extra supervision, in addition to functioning as an independent model. At test time, only the tiny model is used for inference, incurring zero inference overhead. We demonstrate the effectiveness of NetAug on image classification and object detection. NetAug consistently improves the performance of tiny models, achieving up to 2.2% accuracy improvement on ImageNet. On object detection, achieving the same level of performance, NetAug requires 41% fewer MACs on Pascal VOC and 38% fewer MACs on COCO than the baseline."
                },
                {
                    "title": "Beyond Distillation: Task-level Mixture-of-Experts for Efficient Inference",
                    "abstract": "Sparse Mixture-of-Experts (MoE) has been a successful approach for scaling multilingual translation models to billions of parameters without a proportional increase in training computation. However, MoE models are prohibitively large and practitioners often resort to methods such as distillation for serving. In this work, we investigate routing strategies at different granularity (token, sentence, task) in MoE models to bypass distillation. Experiments on WMT and a web-scale dataset suggest that task-level routing (task-MoE) enables us to extract smaller, ready-to-deploy sub-networks from large sparse models. On WMT, our task-MoE with 32 experts (533M parameters) outperforms the best performing token-level MoE model (token-MoE) by +1.0 BLEU on average across 30 language pairs. The peak inference throughput is also improved by a factor of 1.9x when we route by tasks instead of tokens. While distilling a token-MoE to a smaller dense model preserves only 32% of the BLEU gains, our sub-network task-MoE, by design, preserves all the gains with the same inference cost as the distilled student model. Finally, when scaling up to 200 language pairs, our 128-expert task-MoE (13B parameters) performs competitively with a token-level counterpart, while improving the peak inference throughput by a factor of 2.6x."
                },
                {
                    "title": "Primer: Searching for Efficient Transformers for Language Modeling",
                    "abstract": "Large Transformer models have been central to recent advances in natural language processing. The training and inference costs of these models, however, have grown rapidly and become prohibitively expensive. Here we aim to reduce the costs of Transformers by searching for a more efficient variant. Compared to previous approaches, our search is performed at a lower level, over the primitives that define a Transformer TensorFlow program. We identify an architecture, named Primer, that has a smaller training cost than the original Transformer and other variants for auto-regressive language modeling. Primer's improvements can be mostly attributed to two simple modifications: squaring ReLU activations and adding a depthwise convolution layer after each Q, K, and V projection in self-attention. Experiments show Primer's gains over Transformer increase as compute scale grows and follow a power law with respect to quality at optimal model sizes. We also verify empirically that Primer can be dropped into different codebases to significantly speed up training without additional tuning. For example, at a 500M parameter size, Primer improves the original T5 architecture on C4 auto-regressive language modeling, reducing the training cost by 4X. Furthermore, the reduced training cost means Primer needs much less compute to reach a target one-shot performance. For instance, in a 1.9B parameter configuration similar to GPT-3 XL, Primer uses 1/3 of the training compute to achieve the same one-shot performance as Transformer. We open source our models and several comparisons in T5 to help with reproducibility."
                },
                {
                    "title": "Block Pruning For Faster Transformers",
                    "abstract": "Pre-training has improved model accuracy for both classification and generation tasks at the cost of introducing much larger and slower models. Pruning methods have proven to be an effective way of reducing model size, whereas distillation methods are proven for speeding up inference. We introduce a block pruning approach targeting both small and fast models. Our approach extends structured methods by considering blocks of any size and integrates this structure into the movement pruning paradigm for fine-tuning. We find that this approach learns to prune out full components of the underlying model, such as attention heads. Experiments consider classification and generation tasks, yielding among other results a pruned model that is a 2.4x faster, 74% smaller BERT on SQuAD v1, with a 1% drop on F1, competitive both with distilled models in speed and pruned models in size."
                },
                {
                    "title": "On the Opportunities and Risks of Foundation Models",
                    "abstract": "AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature."
                },
                {
                    "title": "How to train your ViT? Data, Augmentation, and Regularization in Vision Transformers",
                    "abstract": "Vision Transformers (ViT) have been shown to attain highly competitive performance for a wide range of vision applications, such as image classification, object detection and semantic image segmentation. In comparison to convolutional neural networks, the Vision Transformer's weaker inductive bias is generally found to cause an increased reliance on model regularization or data augmentation (\"AugReg\"for short) when training on smaller training datasets. We conduct a systematic empirical study in order to better understand the interplay between the amount of training data, AugReg, model size and compute budget. As one result of this study we find that the combination of increased compute and AugReg can yield models with the same performance as models trained on an order of magnitude more training data: we train ViT models of various sizes on the public ImageNet-21k dataset which either match or outperform their counterparts trained on the larger, but not publicly available JFT-300M dataset."
                },
                {
                    "title": "Deep Learning for Instance Retrieval: A Survey",
                    "abstract": "In recent years a vast amount of visual content has been generated and shared from many fields, such as social media platforms, medical imaging, and robotics. This abundance of content creation and sharing has introduced new challenges, particularly that of searching databases for similar content \u2014 Content Based Image Retrieval (CBIR) \u2014 a long-established research area in which improved efficiency and accuracy are needed for real-time retrieval. Artificial intelligence has made progress in CBIR and has significantly facilitated the process of instance search. In this survey we review recent instance retrieval works that are developed based on deep learning algorithms and techniques, with the survey organized by deep feature extraction, feature embedding and aggregation methods, and network fine-tuning strategies. Our survey considers a wide variety of recent methods, whereby we identify milestone work, reveal connections among various methods and present the commonly used benchmarks, evaluation results, common challenges, and propose promising future directions."
                },
                {
                    "title": "Multiple Networks are More Efficient than One: Fast and Accurate Models via Ensembles and Cascades",
                    "abstract": "Recent work on efficient neural network architectures focuses on discovering a solitary network that can achieve superior computational efficiency and accuracy. While this paradigm has yielded impressive results, the search for novel architectures usually requires significant computational resources. In this work, we demonstrate a simple complementary paradigm to obtain efficient and accurate models that requires no architectural tuning. We show that committee-based models, i.e., ensembles or cascades of models, can easily obtain higher accuracy with less computation when compared to a single model. We extensively investigate the benefits of committee-based models on various vision tasks and architecture families. Our results suggest that in the large computation regime, model ensembles are a more cost-effective way to improve accuracy than using a large solitary model. We also find that the computational cost of an ensemble can be significantly reduced by converting them to cascades, while often retaining the original accuracy of the full ensemble."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "HAT: Hardware-Aware Transformers for Efficient Natural Language Processing",
                    "abstract": "Transformers are ubiquitous in Natural Language Processing (NLP) tasks, but they are difficult to be deployed on hardware due to the intensive computation. To enable low-latency inference on resource-constrained hardware platforms, we propose to design Hardware-Aware Transformers (HAT) with neural architecture search. We first construct a large design space with arbitrary encoder-decoder attention and heterogeneous layers. Then we train a SuperTransformer that covers all candidates in the design space, and efficiently produces many SubTransformers with weight sharing. Finally, we perform an evolutionary search with a hardware latency constraint to find a specialized SubTransformer dedicated to run fast on the target hardware. Extensive experiments on four machine translation tasks demonstrate that HAT can discover efficient models for different hardware (CPU, GPU, IoT device). When running WMT\u201914 translation task on Raspberry Pi-4, HAT can achieve 3\u00d7 speedup, 3.7\u00d7 smaller size over baseline Transformer; 2.7\u00d7 speedup, 3.6\u00d7 smaller size over Evolved Transformer with 12,041\u00d7 less search cost and no performance loss. HAT is open-sourced at https://github.com/mit-han-lab/hardware-aware-transformers."
                },
                {
                    "title": "DynaBERT: Dynamic BERT with Adaptive Width and Depth",
                    "abstract": "The pre-trained language models like BERT and RoBERTa, though powerful in many natural language processing tasks, are both computational and memory expensive. To alleviate this problem, one approach is to compress them for specific tasks before deployment. However, recent works on BERT compression usually reduce the large BERT model to a fixed smaller size, and can not fully satisfy the requirements of different edge devices with various hardware performances. In this paper, we propose a novel dynamic BERT model (abbreviated as DynaBERT), which can run at adaptive width and depth. The training process of DynaBERT includes first training a width-adaptive BERT and then allows both adaptive width and depth, by distilling knowledge from the full-sized model to small sub-networks. Network rewiring is also used to keep the more important attention heads and neurons shared by more sub-networks. Comprehensive experiments under various efficiency constraints demonstrate that our proposed dynamic BERT (or RoBERTa) at its largest size has comparable performance as BERT (or RoBERTa), while at smaller widths and depths consistently outperforms existing BERT compression methods."
                },
                {
                    "title": "Soft Threshold Weight Reparameterization for Learnable Sparsity",
                    "abstract": "Sparsity in Deep Neural Networks (DNNs) is studied extensively with the focus of maximizing prediction accuracy given an overall parameter budget. Existing methods rely on uniform or heuristic non-uniform sparsity budgets which have sub-optimal layer-wise parameter allocation resulting in a) lower prediction accuracy or b) higher inference cost (FLOPs). This work proposes Soft Threshold Reparameterization (STR), a novel use of the soft-threshold operator on DNN weights. STR smoothly induces sparsity while learning pruning thresholds thereby obtaining a non-uniform sparsity budget. Our method achieves state-of-the-art accuracy for unstructured sparsity in CNNs (ResNet50 and MobileNetV1 on ImageNet-1K), and, additionally, learns non-uniform budgets that empirically reduce the FLOPs by up to 50%. Notably, STR boosts the accuracy over existing results by up to 10% in the ultra sparse (99%) regime and can also be used to induce low-rank (structured sparsity) in RNNs. In short, STR is a simple mechanism which learns effective sparsity budgets that contrast with popular heuristics. Code, pretrained models and sparsity budgets are at this https URL."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "PIQA: Reasoning about Physical Commonsense in Natural Language",
                    "abstract": "To apply eyeshadow without a brush, should I use a cotton swab or a toothpick? Questions requiring this kind of physical commonsense pose a challenge to today's natural language understanding systems. While recent pretrained models (such as BERT) have made progress on question answering over more abstract domains \u2013 such as news articles and encyclopedia entries, where text is plentiful \u2013 in more physical domains, text is inherently limited due to reporting bias. Can AI systems learn to reliably answer physical commonsense questions without experiencing the physical world?In this paper, we introduce the task of physical commonsense reasoning and a corresponding benchmark dataset Physical Interaction: Question Answering or PIQA. Though humans find the dataset easy (95% accuracy), large pretrained models struggle (\u223c75%). We provide analysis about the dimensions of knowledge that existing models lack, which offers significant opportunities for future research."
                },
                {
                    "title": "Adversarial NLI: A New Benchmark for Natural Language Understanding",
                    "abstract": "We introduce a new large-scale NLI benchmark dataset, collected via an iterative, adversarial human-and-model-in-the-loop procedure. We show that training models on this new dataset leads to state-of-the-art performance on a variety of popular NLI benchmarks, while posing a more difficult challenge with its new test set. Our analysis sheds light on the shortcomings of current state-of-the-art models, and shows that non-expert annotators are successful at finding their weaknesses. The data collection method can be applied in a never-ending learning scenario, becoming a moving target for NLU, rather than a static benchmark that will quickly saturate."
                },
                {
                    "title": "DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter",
                    "abstract": "As Transfer Learning from large-scale pre-trained models becomes more prevalent in Natural Language Processing (NLP), operating these large models in on-the-edge and/or under constrained computational training or inference budgets remains challenging. In this work, we propose a method to pre-train a smaller general-purpose language representation model, called DistilBERT, which can then be fine-tuned with good performances on a wide range of tasks like its larger counterparts. While most prior work investigated the use of distillation for building task-specific models, we leverage knowledge distillation during the pre-training phase and show that it is possible to reduce the size of a BERT model by 40%, while retaining 97% of its language understanding capabilities and being 60% faster. To leverage the inductive biases learned by larger models during pre-training, we introduce a triple loss combining language modeling, distillation and cosine-distance losses. Our smaller, faster and lighter model is cheaper to pre-train and we demonstrate its capabilities for on-device computations in a proof-of-concept experiment and a comparative on-device study."
                },
                {
                    "title": "Once for All: Train One Network and Specialize it for Efficient Deployment",
                    "abstract": "We address the challenging problem of efficient inference across many devices and resource constraints, especially on edge devices. Conventional approaches either manually design or use neural architecture search (NAS) to find a specialized neural network and train it from scratch for each case, which is computationally prohibitive (causing $CO_2$ emission as much as 5 cars' lifetime) thus unscalable. In this work, we propose to train a once-for-all (OFA) network that supports diverse architectural settings by decoupling training and search, to reduce the cost. We can quickly get a specialized sub-network by selecting from the OFA network without additional training. To efficiently train OFA networks, we also propose a novel progressive shrinking algorithm, a generalized pruning method that reduces the model size across many more dimensions than pruning (depth, width, kernel size, and resolution). It can obtain a surprisingly large number of sub-networks ($> 10^{19}$) that can fit different hardware platforms and latency constraints while maintaining the same level of accuracy as training independently. On diverse edge devices, OFA consistently outperforms state-of-the-art (SOTA) NAS methods (up to 4.0% ImageNet top1 accuracy improvement over MobileNetV3, or same accuracy but 1.5x faster than MobileNetV3, 2.6x faster than EfficientNet w.r.t measured latency) while reducing many orders of magnitude GPU hours and $CO_2$ emission. In particular, OFA achieves a new SOTA 80.0% ImageNet top-1 accuracy under the mobile setting ($<$600M MACs). OFA is the winning solution for the 3rd Low Power Computer Vision Challenge (LPCVC), DSP classification track and the 4th LPCVC, both classification track and detection track. Code and 50 pre-trained models (for many devices & many latency constraints) are released at this https URL."
                },
                {
                    "title": "Natural Questions: A Benchmark for Question Answering Research",
                    "abstract": "We present the Natural Questions corpus, a question answering data set. Questions consist of real anonymized, aggregated queries issued to the Google search engine. An annotator is presented with a question along with a Wikipedia page from the top 5 search results, and annotates a long answer (typically a paragraph) and a short answer (one or more entities) if present on the page, or marks null if no long/short answer is present. The public release consists of 307,373 training examples with single annotations; 7,830 examples with 5-way annotations for development data; and a further 7,842 examples with 5-way annotated sequestered as test data. We present experiments validating quality of the data. We also describe analysis of 25-way annotations on 302 examples, giving insights into human variability on the annotation task. We introduce robust metrics for the purposes of evaluating question answering systems; demonstrate high human upper bounds on these metrics; and establish baseline results using competitive methods drawn from related literature."
                },
                {
                    "title": "SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems",
                    "abstract": "In the last year, new models and methods for pretraining and transfer learning have driven striking performance improvements across a range of language understanding tasks. The GLUE benchmark, introduced a little over one year ago, offers a single-number metric that summarizes progress on a diverse set of such tasks, but performance on the benchmark has recently surpassed the level of non-expert humans, suggesting limited headroom for further research. In this paper we present SuperGLUE, a new benchmark styled after GLUE with a new set of more difficult language understanding tasks, a software toolkit, and a public leaderboard. SuperGLUE is available at this http URL."
                },
                {
                    "title": "HellaSwag: Can a Machine Really Finish Your Sentence?",
                    "abstract": "Recent work by Zellers et al. (2018) introduced a new task of commonsense natural language inference: given an event description such as \u201cA woman sits at a piano,\u201d a machine must select the most likely followup: \u201cShe sets her fingers on the keys.\u201d With the introduction of BERT, near human-level performance was reached. Does this mean that machines can perform human level commonsense inference? In this paper, we show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans (>95% accuracy), state-of-the-art models struggle (<48%). We achieve this via Adversarial Filtering (AF), a data collection paradigm wherein a series of discriminators iteratively select an adversarial set of machine-generated wrong answers. AF proves to be surprisingly robust. The key insight is to scale up the length and complexity of the dataset examples towards a critical \u2018Goldilocks\u2019 zone wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models. Our construction of HellaSwag, and its resulting difficulty, sheds light on the inner workings of deep pretrained models. More broadly, it suggests a new path forward for NLP research, in which benchmarks co-evolve with the evolving state-of-the-art in an adversarial way, so as to present ever-harder challenges."
                },
                {
                    "title": "Universally Slimmable Networks and Improved Training Techniques",
                    "abstract": "Slimmable networks are a family of neural networks that can instantly adjust the runtime width. The width can be chosen from a predefined widths set to adaptively optimize accuracy-efficiency trade-offs at runtime. In this work, we propose a systematic approach to train universally slimmable networks (US-Nets), extending slimmable networks to execute at arbitrary width, and generalizing to networks both with and without batch normalization layers. We further propose two improved training techniques for US-Nets, named the sandwich rule and inplace distillation, to enhance training process and boost testing accuracy. We show improved performance of universally slimmable MobileNet v1 and MobileNet v2 on ImageNet classification task, compared with individually trained ones and 4-switch slimmable network baselines. We also evaluate the proposed US-Nets and improved training techniques on tasks of image super-resolution and deep reinforcement learning. Extensive ablation experiments on these representative tasks demonstrate the effectiveness of our proposed methods. Our discovery opens up the possibility to directly evaluate FLOPs-Accuracy spectrum of network architectures. Code and models are available at: \\url{https://github.com/JiahuiYu/slimmable_networks}."
                },
                {
                    "title": "Slimmable Neural Networks",
                    "abstract": "We present a simple and general method to train a single neural network executable at different widths (number of channels in a layer), permitting instant and adaptive accuracy-efficiency trade-offs at runtime. Instead of training individual networks with different width configurations, we train a shared network with switchable batch normalization. At runtime, the network can adjust its width on the fly according to on-device benchmarks and resource constraints, rather than downloading and offloading different models. Our trained networks, named slimmable neural networks, achieve similar (and in many cases better) ImageNet classification accuracy than individually trained models of MobileNet v1, MobileNet v2, ShuffleNet and ResNet-50 at different widths respectively. We also demonstrate better performance of slimmable models compared with individual ones across a wide range of applications including COCO bounding-box object detection, instance segmentation and person keypoint detection without tuning hyper-parameters. Lastly we visualize and discuss the learned features of slimmable networks. Code and models are available at: this https URL"
                },
                {
                    "title": "(Preprint)",
                    "abstract": "\n BACKGROUND\n Increasing numbers of children undergo ambulatory surgery each year and a significant proportion experiences substantial preoperative anxiety and postoperative pain. The management of perioperative anxiety and pain remains challenging in children and is inadequate, which negatively impacts physical, psychosocial, and economic outcomes. Existing non-pharmacological interventions are costly, time consuming, vary in availability, and lack benefits. Therefore, there is a need for an evidence-based, accessible, non-pharmacological intervention as an adjunct to existing pharmacological alternatives to reduce perioperative anxiety and pain in children undergoing ambulatory surgery. Technology-enabled interventions have been proposed as a method to address the unmet need in this setting. In particular, serious games for health (SGHs) hold unique potential to change health beliefs and behaviors in children.\n \n \n OBJECTIVE\n The objective of this research was to describe the rationale, scientific evidence, design aspects, and features of CliniPup, an SGH aimed at reducing perioperative anxiety and pain in children undergoing ambulatory surgery.\n \n \n METHODS\n The SERES framework for SGH development was used to create the SGH, CliniPup. In particular, a mixed-methods approach was applied that consisted of a structured literature review supplemented with ethnographic research, such as expert interviews and a time-motion exercise. The resulting scientific evidence base was leveraged to ensure that the resulting SGH was relevant, realistic, and theory-driven. A participatory design approach was applied wherein clinical experts qualitatively reviewed several versions of the SGH and an iterative creative process was used to integrate the applicable feedback.\n \n \n RESULTS\n CliniPup, an SGH, was developed to incorporate (1) scientific evidence base from a structured literature review, (2) realistic content collected during ethnographic research such as expert interviews, (3) explicit pedagogical objectives from scientific literature, and (4) game mechanics and user interface design that address key aspects of the evidence.\n \n \n CONCLUSIONS\n This report details the systematic development of CliniPup, an SGH aimed at reducing perioperative anxiety and pain in children undergoing ambulatory surgery. Clinical experts validated CliniPup\u2019s underlying scientific evidence base and design foundations, suggesting that it was well designed for preliminary evaluation in the target population. An evaluation plan is proposed and briefly described.\n"
                },
                {
                    "title": "SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing",
                    "abstract": "This paper describes SentencePiece, a language-independent subword tokenizer and detokenizer designed for Neural-based text processing, including Neural Machine Translation. It provides open-source C++ and Python implementations for subword units. While existing subword segmentation tools assume that the input is pre-tokenized into word sequences, SentencePiece can train subword models directly from raw sentences, which allows us to make a purely end-to-end and language independent system. We perform a validation experiment of NMT on English-Japanese machine translation, and find that it is possible to achieve comparable accuracy to direct subword training from raw sentences. We also compare the performance of subword training and segmentation with various configurations. SentencePiece is available under the Apache 2 license at https://github.com/google/sentencepiece."
                },
                {
                    "title": "Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering",
                    "abstract": "We present a new kind of question answering dataset, OpenBookQA, modeled after open book exams for assessing human understanding of a subject. The open book that comes with our questions is a set of 1326 elementary level science facts. Roughly 6000 questions probe an understanding of these facts and their application to novel situations. This requires combining an open book fact (e.g., metals conduct electricity) with broad common knowledge (e.g., a suit of armor is made of metal) obtained from other sources. While existing QA datasets over documents or knowledge bases, being generally self-contained, focus on linguistic understanding, OpenBookQA probes a deeper understanding of both the topic\u2014in the context of common knowledge\u2014and the language it is expressed in. Human performance on OpenBookQA is close to 92%, but many state-of-the-art pre-trained QA methods perform surprisingly poorly, worse than several simple neural baselines we develop. Our oracle experiments designed to circumvent the knowledge retrieval bottleneck demonstrate the value of both the open book and additional facts. We leave it as a challenge to solve the retrieval problem in this multi-hop setting and to close the large gap to human performance."
                },
                {
                    "title": "Know What You Don\u2019t Know: Unanswerable Questions for SQuAD",
                    "abstract": "Extractive reading comprehension systems can often locate the correct answer to a question in a context document, but they also tend to make unreliable guesses on questions for which the correct answer is not stated in the context. Existing datasets either focus exclusively on answerable questions, or use automatically generated unanswerable questions that are easy to identify. To address these weaknesses, we present SQuADRUn, a new dataset that combines the existing Stanford Question Answering Dataset (SQuAD) with over 50,000 unanswerable questions written adversarially by crowdworkers to look similar to answerable ones. To do well on SQuADRUn, systems must not only answer questions when possible, but also determine when no answer is supported by the paragraph and abstain from answering. SQuADRUn is a challenging natural language understanding task for existing models: a strong neural system that gets 86% F1 on SQuAD achieves only 66% F1 on SQuADRUn. We release SQuADRUn to the community as the successor to SQuAD."
                },
                {
                    "title": "Adafactor: Adaptive Learning Rates with Sublinear Memory Cost",
                    "abstract": "In several recently proposed stochastic optimization methods (e.g. RMSProp, Adam, Adadelta), parameter updates are scaled by the inverse square roots of exponential moving averages of squared past gradients. Maintaining these per-parameter second-moment estimators requires memory equal to the number of parameters. For the case of neural network weight matrices, we propose maintaining only the per-row and per-column sums of these moving averages, and estimating the per-parameter second moments based on these sums. We demonstrate empirically that this method produces similar results to the baseline. Secondly, we show that adaptive methods can produce larger-than-desired updates when the decay rate of the second moment accumulator is too slow. We propose update clipping and a gradually increasing decay rate scheme as remedies. Combining these methods and dropping momentum, we achieve comparable results to the published Adam regime in training the Transformer model on the WMT 2014 English-German machine translation task, while using very little auxiliary storage in the optimizer. Finally, we propose scaling the parameter updates based on the scale of the parameters themselves."
                },
                {
                    "title": "Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge",
                    "abstract": "We present a new question set, text corpus, and baselines assembled to encourage AI research in advanced question answering. Together, these constitute the AI2 Reasoning Challenge (ARC), which requires far more powerful knowledge and reasoning than previous challenges such as SQuAD or SNLI. The ARC question set is partitioned into a Challenge Set and an Easy Set, where the Challenge Set contains only questions answered incorrectly by both a retrieval-based algorithm and a word co-occurence algorithm. The dataset contains only natural, grade-school science questions (authored for human tests), and is the largest public-domain set of this kind (7,787 questions). We test several baselines on the Challenge Set, including leading neural models from the SQuAD and SNLI tasks, and find that none are able to significantly outperform a random baseline, reflecting the difficult nature of this task. We are also releasing the ARC Corpus, a corpus of 14M science sentences relevant to the task, and implementations of the three neural baseline models tested. Can your model perform better? We pose ARC as a challenge to the community."
                },
                {
                    "title": "Generating Wikipedia by Summarizing Long Sequences",
                    "abstract": "We show that generating English Wikipedia articles can be approached as a multi- document summarization of source documents. We use extractive summarization to coarsely identify salient information and a neural abstractive model to generate the article. For the abstractive model, we introduce a decoder-only architecture that can scalably attend to very long sequences, much longer than typical encoder- decoder architectures used in sequence transduction. We show that this model can generate fluent, coherent multi-sentence paragraphs and even whole Wikipedia articles. When given reference documents, we show it can extract relevant factual information as reflected in perplexity, ROUGE scores and human evaluations."
                },
                {
                    "title": "TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension",
                    "abstract": "We present TriviaQA, a challenging reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents, six per question on average, that provide high quality distant supervision for answering the questions. We show that, in comparison to other recently introduced large-scale datasets, TriviaQA (1) has relatively complex, compositional questions, (2) has considerable syntactic and lexical variability between questions and corresponding answer-evidence sentences, and (3) requires more cross sentence reasoning to find answers. We also present two baseline algorithms: a feature-based classifier and a state-of-the-art neural network, that performs well on SQuAD reading comprehension. Neither approach comes close to human performance (23% and 40% vs. 80%), suggesting that TriviaQA is a challenging testbed that is worth significant future study."
                },
                {
                    "title": "RACE: Large-scale ReAding Comprehension Dataset From Examinations",
                    "abstract": "We present RACE, a new dataset for benchmark evaluation of methods in the reading comprehension task. Collected from the English exams for middle and high school Chinese students in the age range between 12 to 18, RACE consists of near 28,000 passages and near 100,000 questions generated by human experts (English instructors), and covers a variety of topics which are carefully designed for evaluating the students\u2019 ability in understanding and reasoning. In particular, the proportion of questions that requires reasoning is much larger in RACE than that in other benchmark datasets for reading comprehension, and there is a significant gap between the performance of the state-of-the-art models (43%) and the ceiling human performance (95%). We hope this new dataset can serve as a valuable resource for research and evaluation in machine comprehension. The dataset is freely available at http://www.cs.cmu.edu/~glai1/data/race/ and the code is available at https://github.com/qizhex/RACE_AR_baselines."
                },
                {
                    "title": "Gaussian Error Linear Units (GELUs)",
                    "abstract": "We propose the Gaussian Error Linear Unit (GELU), a high-performing neural network activation function. The GELU activation function is $x\\Phi(x)$, where $\\Phi(x)$ the standard Gaussian cumulative distribution function. The GELU nonlinearity weights inputs by their value, rather than gates inputs by their sign as in ReLUs ($x\\mathbf{1}_{x>0}$). We perform an empirical evaluation of the GELU nonlinearity against the ReLU and ELU activations and find performance improvements across all considered computer vision, natural language processing, and speech tasks."
                },
                {
                    "title": "The LAMBADA dataset: Word prediction requiring a broad discourse context",
                    "abstract": "We introduce LAMBADA, a dataset to evaluate the capabilities of computational models for text understanding by means of a word prediction task. LAMBADA is a collection of narrative passages sharing the characteristic that human subjects are able to guess their last word if they are exposed to the whole passage, but not if they only see the last sentence preceding the target word. To succeed on LAMBADA, computational models cannot simply rely on local context, but must be able to keep track of information in the broader discourse. We show that LAMBADA exemplifies a wide range of linguistic phenomena, and that none of several state-of-the-art language models reaches accuracy above 1% on this novel benchmark. We thus propose LAMBADA as a challenging test set, meant to encourage the development of new models capable of genuine understanding of broad context in natural language text."
                },
                {
                    "title": "A Corpus and Cloze Evaluation for Deeper Understanding of Commonsense Stories",
                    "abstract": "Representation and learning of commonsense knowledge is one of the foundational problems in the quest to enable deep language understanding. This issue is particularly challenging for understanding casual and correlational relationships between events. While this topic has received a lot of interest in the NLP community, research has been hindered by the lack of a proper evaluation framework. This paper attempts to address this problem with a new framework for evaluating story understanding and script learning: the 'Story Cloze Test'. This test requires a system to choose the correct ending to a four-sentence story. We created a new corpus of ~50k five-sentence commonsense stories, ROCStories, to enable this evaluation. This corpus is unique in two ways: (1) it captures a rich set of causal and temporal commonsense relations between daily events, and (2) it is a high quality collection of everyday life stories that can also be used for story generation. Experimental evaluation shows that a host of baselines and state-of-the-art models based on shallow language understanding struggle to achieve a high score on the Story Cloze Test. We discuss these implications for script and story learning, and offer suggestions for deeper language understanding."
                },
                {
                    "title": "Semantic Parsing on Freebase from Question-Answer Pairs",
                    "abstract": "In this paper, we train a semantic parser that scales up to Freebase. Instead of relying on annotated logical forms, which is especially expensive to obtain at large scale, we learn from question-answer pairs. The main challenge in this setting is narrowing down the huge number of possible logical predicates for a given question. We tackle this problem in two ways: First, we build a coarse mapping from phrases to predicates using a knowledge base and a large text corpus. Second, we use a bridging operation to generate additional predicates based on neighboring predicates. On the dataset of Cai and Yates (2013), despite not having annotated logical forms, our system outperforms their state-of-the-art parser. Additionally, we collected a more realistic and challenging dataset of question-answer pairs and improves over a natural baseline."
                },
                {
                    "title": "Ensemble Machine Learning: Methods and Applications",
                    "abstract": "It is common wisdom that gathering a variety of views and inputs improves the process of decision making, and, indeed, underpins a democratic society. Dubbed ensemble learning by researchers in computational intelligence and machine learning, it is known to improve a decision systems robustness and accuracy. Now, fresh developments are allowing researchers to unleash the power of ensemble learning in an increasing range of real-world applications. Ensemble learning algorithms such as boosting and random forest facilitate solutions to key computational issues such as face recognition and are now being applied in areas as diverse as object tracking and bioinformatics. Responding to a shortage of literature dedicated to the topic, this volume offers comprehensive coverage of state-of-the-art ensemble learning techniques, including the random forest skeleton tracking algorithm in the Xbox Kinect sensor, which bypasses the need for game controllers. At once a solid theoretical study and a practical guide, the volume is a windfall for researchers and practitioners alike."
                },
                {
                    "title": "The Winograd Schema Challenge",
                    "abstract": "In this paper, we present an alternative to the Turing Test that has some conceptual and practical advantages. Like the original, it involves responding to typed English sentences, and English-speaking adults will have no difficulty with it. Unlike the original, the subject is not required to engage in a conversation and fool an interrogator into believing she is dealing with a person. Moreover, the test is arranged in such a way that having full access to a large corpus of English text might not help much. Finally, the interrogator or a third party will be able to decide unambiguously after a few minutes whether or not a subject has passed the test."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "An Adversarial Winograd Schema Challenge at Scale",
                    "abstract": "The Winograd Schema Challenge (WSC), proposed by Levesque et al. (2011) as an alternative to the Turing Test, was originally designed as a pronoun resolution problem that cannot be solved based on statistical patterns in large text corpora. However, recent studies suggest that current WSC datasets, even when composed carefully by experts, are still prone to such biases that statistical methods can exploit. We introduce WINOGRANDE, a new collection of WSC problems that are adversarially constructed to be robust against spurious statistical biases. While the original WSC dataset provided only 273 instances, WINOGRANDE includes 43,985 instances, half of which are determined as adversarial. Key to our approach is a novel adversarial filtering algorithm AFLITE for systematic bias reduction, combined with a careful crowdsourcing design. Despite the significant increase in training data, the performance of existing stateof-the-art methods remains modest (61.6%) and contrasts with high human performance (90.8%) for the binary questions. In addition, WINOGRANDE allows us to use transfer learning for achieving new state-of-the-art results on the original WSC and related datasets. Finally, we discuss how biases lead to overestimating the true capabilities of machine commonsense."
                },
                {
                    "title": "Common Sense Reasoning",
                    "abstract": "Qualitative Reasoning captures human common sense understanding of the behavior of real world physical systems. Qualitative models use a rich explicit conceptual vocabulary and therefore a qualitative simulation provides an explanation as well as a description of the system\u2019s behavior. These properties make Qualitative Reasoning suitable for use in several application fields like education and ecology. The design of an elaborate, detailed Qualitative Reasoning engine is a complex task. Existing simulators lack reasoning capacities which are required by present day applications, such are extensive use of inequality reasoning and efficient computation of state transitions. This thesis describes an effort to develop fundamental solutions for several problems in such reasoning engines, particularly focussing on the GARP reasoning engine."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CL",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we design a universal Transformer architecture that allows for the extraction of multiple smaller submodels without additional training, while maintaining behavioral consistency and optimizing inference performance?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the inefficiencies in deploying large foundation models across various applications. By enabling a single model to produce multiple submodels, we can significantly reduce the overhead associated with training and inference, leading to more efficient resource utilization. This advancement could pave the way for more flexible and scalable AI systems, enhancing the applicability of large models in real-time and resource-constrained environments. Furthermore, it could inspire future research into more adaptive model architectures and improve the overall understanding of model compression techniques.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of designing a nested structure within the Transformer architecture that allows for the seamless extraction of submodels. Naive approaches may fail because they do not account for the interdependencies between model parameters, leading to inconsistencies in performance across different model sizes. Additionally, achieving behavioral consistency among the extracted submodels is technically demanding, as it requires careful management of shared parameters and their interactions. Theoretical obstacles include ensuring that the nested structure does not compromise the model's overall performance, while practical obstacles involve the implementation of this architecture in existing training pipelines.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on independently training models of varying sizes, which has led to significant overhead and inconsistencies during inference. Existing solutions often rely on model compression techniques that require additional training, which can degrade accuracy, particularly in large language models. The lack of a unified approach that integrates the concept of nested structures within the Transformer architecture has been a barrier to progress. Our approach differs by introducing the matryoshka representation learning principle, allowing for a single model to inherently support multiple submodels without the need for retraining, thus overcoming limitations of prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the MatFormer architecture, which incorporates nested Transformer blocks that allow for the extraction of submodels. We will utilize a diverse dataset of large language models and vision transformers to evaluate the performance of MatFormer. The key metrics for assessment will include inference speed, accuracy, and resource efficiency. We expect that Mat"
            }
        },
        "author_data": {
            "012a3f73-59e3-4f86-a4ff-7b0300373c40": {
                "pk": "012a3f73-59e3-4f86-a4ff-7b0300373c40",
                "name": "Devvrit",
                "collaborators": [
                    "Inderjit Dhillon",
                    "Fnu Devvrit",
                    "Cho-Jui Hsieh",
                    "Nived Rajaraman",
                    "Devvrit Khatri",
                    "Aaron Krim-Yee",
                    "Nithish Kumar",
                    "Ben Seamone",
                    "Virg\u00e9lot Virgile",
                    "AnQi Xu"
                ],
                "domain": [
                    "Adversarial Learning",
                    "Active Learning",
                    "Graph Theory",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Voting based ensemble improves robustness of defensive models",
                        "abstract": "Developing robust models against adversarial perturbations has been an active area of research and many algorithms have been proposed to train individual robust models. Taking these pretrained robust models, we aim to study whether it is possible to create an ensemble to further improve robustness. Several previous attempts tackled this problem by ensembling the soft-label prediction and have been proved vulnerable based on the latest attack methods. In this paper, we show that if the robust training loss is diverse enough, a simple hard-label based voting ensemble can boost the robust error over each individual model. Furthermore, given a pool of robust models, we develop a principled way to select which models to ensemble. Finally, to verify the improved robustness, we conduct extensive experiments to study how to attack a voting-based ensemble and develop several new white-box attacks. On CIFAR-10 dataset, by ensembling several state-of-the-art pre-trained defense models, our method can achieve a 59.8% robust accuracy, outperforming all the existing defensive models without using additional data."
                    },
                    {
                        "title": "Semi-supervised Active Regression",
                        "abstract": "Labelled data often comes at a high cost as it may require recruiting human labelers or running costly experiments. At the same time, in many practical scenarios, one already has access to a partially labelled, potentially biased dataset that can help with the learning task at hand. Motivated by such settings, we formally initiate a study of $semi-supervised$ $active$ $learning$ through the frame of linear regression. In this setting, the learner has access to a dataset $X \\in \\mathbb{R}^{(n_1+n_2) \\times d}$ which is composed of $n_1$ unlabelled examples that an algorithm can actively query, and $n_2$ examples labelled a-priori. Concretely, denoting the true labels by $Y \\in \\mathbb{R}^{n_1 + n_2}$, the learner's objective is to find $\\widehat{\\beta} \\in \\mathbb{R}^d$ such that, \\begin{equation}   \\| X \\widehat{\\beta} - Y \\|_2^2 \\le (1 + \\epsilon) \\min_{\\beta \\in \\mathbb{R}^d} \\| X \\beta - Y \\|_2^2 \\end{equation} while making as few additional label queries as possible. In order to bound the label queries, we introduce an instance dependent parameter called the reduced rank, denoted by $R_X$, and propose an efficient algorithm with query complexity $O(R_X/\\epsilon)$. This result directly implies improved upper bounds for two important special cases: (i) active ridge regression, and (ii) active kernel ridge regression, where the reduced-rank equates to the statistical dimension, $sd_\\lambda$ and effective dimension, $d_\\lambda$ of the problem respectively, where $\\lambda \\ge 0$ denotes the regularization parameter. For active ridge regression we also prove a matching lower bound of $O(sd_\\lambda / \\epsilon)$ on the query complexity of any algorithm. This subsumes prior work that only considered the unregularized case, i.e., $\\lambda = 0$."
                    },
                    {
                        "title": "Greedy Pruning with Group Lasso Provably Generalizes for Matrix Sensing",
                        "abstract": "Pruning schemes have been widely used in practice to reduce the complexity of trained models with a massive number of parameters. In fact, several practical studies have shown that if a pruned model is fine-tuned with some gradient-based updates it generalizes well to new samples. Although the above pipeline, which we refer to as pruning + fine-tuning, has been extremely successful in lowering the complexity of trained models, there is very little known about the theory behind this success. In this paper, we address this issue by investigating the pruning + fine-tuning framework on the overparameterized matrix sensing problem with the ground truth $U_\\star \\in \\mathbb{R}^{d \\times r}$ and the overparameterized model $U \\in \\mathbb{R}^{d \\times k}$ with $k \\gg r$. We study the approximate local minima of the mean square error, augmented with a smooth version of a group Lasso regularizer, $\\sum_{i=1}^k \\| U e_i \\|_2$. In particular, we provably show that pruning all the columns below a certain explicit $\\ell_2$-norm threshold results in a solution $U_{\\text{prune}}$ which has the minimum number of columns $r$, yet close to the ground truth in training loss. Moreover, in the subsequent fine-tuning phase, gradient descent initialized at $U_{\\text{prune}}$ converges at a linear rate to its limit. While our analysis provides insights into the role of regularization in pruning, we also show that running gradient descent in the absence of regularization results in models which {are not suitable for greedy pruning}, i.e., many columns could have their $\\ell_2$ norm comparable to that of the maximum. To the best of our knowledge, our results provide the first rigorous insights on why greedy pruning + fine-tuning leads to smaller models which also generalize well."
                    },
                    {
                        "title": "A study of cops and robbers in oriented graphs",
                        "abstract": "We consider the well-studied cops and robbers game in the context of oriented graphs, which has received surprisingly little attention to date. We examine the relationship between the cop numbers of an oriented graph and its underlying undirected graph, giving a surprising result that there exists at least one graph $G$ for which every strongly connected orientation of $G$ has cop number strictly less than that of $G$. We also refute a conjecture on the structure of cop-win digraphs, study orientations of outerplanar graphs, and study the cop number of line digraphs. Finally, we consider some the aspects of optimal play, in particular the capture time of cop-win digraphs and properties of the relative positions of the cop(s) and robber."
                    },
                    {
                        "title": "Fractional eternal domination: securely distributing resources across a network",
                        "abstract": "This paper initiates the study of fractional eternal domination in graphs, a natural relaxation of the well-studied eternal domination problem. We study the connections to flows and linear programming in order to obtain results on the complexity of determining the fractional eternal domination number of a graph $G$, which we denote $\\gamma_{\\,\\textit{f}}^{\\infty}(G)$. We study the behaviour of $\\gamma_{\\,\\textit{f}}^{\\infty}(G)$ as it relates to other domination parameters. We also determine bounds on, and in some cases exact values for, $\\gamma_{\\,\\textit{f}}^{\\infty}(G)$ when $G$ is a member of one of a variety of important graph classes, including trees, split graphs, strongly chordal graphs, Kneser graphs, abelian Cayley graphs, and graph products."
                    },
                    {
                        "title": "Dual-Encoders for Extreme Multi-Label Classification",
                        "abstract": "Dual-encoder (DE) models are widely used in retrieval tasks, most commonly studied on open QA benchmarks that are often characterized by multi-class and limited training data. In contrast, their performance in multi-label and data-rich retrieval settings like extreme multi-label classification (XMC), remains under-explored. Current empirical evidence indicates that DE models fall significantly short on XMC benchmarks, where SOTA methods linearly scale the number of learnable parameters with the total number of classes (documents in the corpus) by employing per-class classification head. To this end, we first study and highlight that existing multi-label contrastive training losses are not appropriate for training DE models on XMC tasks. We propose decoupled softmax loss - a simple modification to the InfoNCE loss - that overcomes the limitations of existing contrastive losses. We further extend our loss design to a soft top-k operator-based loss which is tailored to optimize top-k prediction performance. When trained with our proposed loss functions, standard DE models alone can match or outperform SOTA methods by up to 2% at Precision@1 even on the largest XMC datasets while being 20x smaller in terms of the number of trainable parameters. This leads to more parameter-efficient and universally applicable solutions for retrieval tasks. Our code and models are publicly available at https://github.com/nilesh2797/dexml."
                    },
                    {
                        "title": "A Computationally Efficient Sparsified Online Newton Method",
                        "abstract": "Second-order methods hold significant promise for enhancing the convergence of deep neural network training; however, their large memory and computational demands have limited their practicality. Thus there is a need for scalable second-order methods that can efficiently train large models. In this paper, we introduce the Sparsified Online Newton (SONew) method, a memory-efficient second-order algorithm that yields a sparsified yet effective preconditioner. The algorithm emerges from a novel use of the LogDet matrix divergence measure; we combine it with sparsity constraints to minimize regret in the online convex optimization framework. Empirically, we test our method on large scale benchmarks of up to 1B parameters. We achieve up to 30% faster convergence, 3.4% relative improvement in validation performance, and 80% relative improvement in training loss, in comparison to memory efficient optimizers including first order methods. Powering the method is a surprising fact -- imposing structured sparsity patterns, like tridiagonal and banded structure, requires little to no overhead, making it as efficient and parallelizable as first-order methods. In wall-clock time, tridiagonal SONew is only about 3% slower per step than first-order methods but gives overall gains due to much faster convergence. In contrast, one of the state-of-the-art (SOTA) memory-intensive second-order methods, Shampoo, is unable to scale to large benchmarks. Additionally, while Shampoo necessitates significant engineering efforts to scale to large benchmarks, SONew offers a more straightforward implementation, increasing its practical appeal. SONew code is available at: https://github.com/devvrit/SONew"
                    }
                ]
            },
            "35e19ee8-9f52-4963-8fd5-75a2f7538cbd": {
                "pk": "35e19ee8-9f52-4963-8fd5-75a2f7538cbd",
                "name": "Sneha Kudugunta",
                "collaborators": [
                    "Orhan Firat",
                    "Ankur Bapna",
                    "Emilio Ferrara",
                    "Kevin Robinson",
                    "Romina Stella",
                    "Sunipa Dev",
                    "Jasmijn Bastings",
                    "Yanping Huang",
                    "Maxim Krikun",
                    "Dmitry Lepikhin"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Neural Networks",
                    "Multilingual Translation"
                ],
                "publications": [
                    {
                        "title": "Deep Neural Networks for Bot Detection",
                        "abstract": "The problem of detecting bots, automated social media accounts governed by software but disguising as human users, has strong implications. For example, bots have been used to sway political elections by distorting online discourse, to manipulate the stock market, or to push anti-vaccine conspiracy theories that caused health epidemics. Most techniques proposed to date detect bots at the account level, by processing large amount of social media posts, and leveraging information from network structure, temporal dynamics, sentiment analysis, etc.   In this paper, we propose a deep neural network based on contextual long short-term memory (LSTM) architecture that exploits both content and metadata to detect bots at the tweet level: contextual features are extracted from user metadata and fed as auxiliary input to LSTM deep nets processing the tweet text.   Another contribution that we make is proposing a technique based on synthetic minority oversampling to generate a large labeled dataset, suitable for deep nets training, from a minimal amount of labeled data (roughly 3,000 examples of sophisticated Twitter bots). We demonstrate that, from just one single tweet, our architecture can achieve high classification accuracy (AUC > 96%) in separating bots from humans.   We apply the same architecture to account-level bot detection, achieving nearly perfect classification accuracy (AUC > 99%). Our system outperforms previous state of the art while leveraging a small and interpretable set of features yet requiring minimal training data."
                    },
                    {
                        "title": "MiTTenS: A Dataset for Evaluating Gender Mistranslation",
                        "abstract": "Translation systems, including foundation models capable of translation, can produce errors that result in gender mistranslation, and such errors can be especially harmful. To measure the extent of such potential harms when translating into and out of English, we introduce a dataset, MiTTenS, covering 26 languages from a variety of language families and scripts, including several traditionally under-represented in digital resources. The dataset is constructed with handcrafted passages that target known failure patterns, longer synthetically generated passages, and natural passages sourced from multiple domains. We demonstrate the usefulness of the dataset by evaluating both neural machine translation systems and foundation models, and show that all systems exhibit gender mistranslation and potential harm, even in high resource languages."
                    },
                    {
                        "title": "Beyond Distillation: Task-level Mixture-of-Experts for Efficient Inference",
                        "abstract": "Sparse Mixture-of-Experts (MoE) has been a successful approach for scaling multilingual translation models to billions of parameters without a proportional increase in training computation. However, MoE models are prohibitively large and practitioners often resort to methods such as distillation for serving. In this work, we investigate routing strategies at different granularity (token, sentence, task) in MoE models to bypass distillation. Experiments on WMT and a web-scale dataset suggest that task-level routing (task-MoE) enables us to extract smaller, ready-to-deploy sub-networks from large sparse models. On WMT, our task-MoE with 32 experts (533M parameters) outperforms the best performing token-level MoE model (token-MoE) by +1.0 BLEU on average across 30 language pairs. The peak inference throughput is also improved by a factor of 1.9x when we route by tasks instead of tokens. While distilling a token-MoE to a smaller dense model preserves only 32% of the BLEU gains, our sub-network task-MoE, by design, preserves all the gains with the same inference cost as the distilled student model. Finally, when scaling up to 200 language pairs, our 128-expert task-MoE (13B parameters) performs competitively with a token-level counterpart, while improving the peak inference throughput by a factor of 2.6x."
                    },
                    {
                        "title": "MADLAD-400: A Multilingual And Document-Level Large Audited Dataset",
                        "abstract": "We introduce MADLAD-400, a manually audited, general domain 3T token monolingual dataset based on CommonCrawl, spanning 419 languages. We discuss the limitations revealed by self-auditing MADLAD-400, and the role data auditing had in the dataset creation process. We then train and release a 10.7B-parameter multilingual machine translation model on 250 billion tokens covering over 450 languages using publicly available data, and find that it is competitive with models that are significantly larger, and report the results on different domains. In addition, we train a 8B-parameter language model, and assess the results on few-shot translation. We make the baseline models available to the research community."
                    },
                    {
                        "title": "DANTE: Deep AlterNations for Training nEural networks",
                        "abstract": "We present DANTE, a novel method for training neural networks using the alternating minimization principle. DANTE provides an alternate perspective to traditional gradient-based backpropagation techniques commonly used to train deep networks. It utilizes an adaptation of quasi-convexity to cast training a neural network as a bi-quasi-convex optimization problem. We show that for neural network configurations with both differentiable (e.g. sigmoid) and non-differentiable (e.g. ReLU) activation functions, we can perform the alternations effectively in this formulation. DANTE can also be extended to networks with multiple hidden layers. In experiments on standard datasets, neural networks trained using the proposed method were found to be promising and competitive to traditional backpropagation techniques, both in terms of quality of the solution, as well as training speed."
                    },
                    {
                        "title": "Leveraging Monolingual Data with Self-Supervision for Multilingual Neural Machine Translation",
                        "abstract": "Over the last few years two promising research directions in low-resource neural machine translation (NMT) have emerged. The first focuses on utilizing high-resource languages to improve the quality of low-resource languages via multilingual NMT. The second direction employs monolingual data with self-supervision to pre-train translation models, followed by fine-tuning on small amounts of supervised data. In this work, we join these two lines of research and demonstrate the efficacy of monolingual data with self-supervision in multilingual NMT. We offer three major results: (i) Using monolingual data significantly boosts the translation quality of low-resource languages in multilingual models. (ii) Self-supervision improves zero-shot translation quality in multilingual models. (iii) Leveraging monolingual data with self-supervision provides a viable path towards adding new languages to multilingual models, getting up to 33 BLEU on ro-en translation without any parallel data or back-translation."
                    },
                    {
                        "title": "MURAL: Multimodal, Multitask Retrieval Across Languages",
                        "abstract": "Both image-caption pairs and translation pairs provide the means to learn deep representations of and connections between languages. We use both types of pairs in MURAL (MUltimodal, MUltitask Representations Across Languages), a dual encoder that solves two tasks: 1) image-text matching and 2) translation pair matching. By incorporating billions of translation pairs, MURAL extends ALIGN (Jia et al. PMLR'21)--a state-of-the-art dual encoder learned from 1.8 billion noisy image-text pairs. When using the same encoders, MURAL's performance matches or exceeds ALIGN's cross-modal retrieval performance on well-resourced languages across several datasets. More importantly, it considerably improves performance on under-resourced languages, showing that text-text learning can overcome a paucity of image-caption examples for these languages. On the Wikipedia Image-Text dataset, for example, MURAL-base improves zero-shot mean recall by 8.1% on average for eight under-resourced languages and by 6.8% on average when fine-tuning. We additionally show that MURAL's text representations cluster not only with respect to genealogical connections but also based on areal linguistics, such as the Balkan Sprachbund."
                    },
                    {
                        "title": "A Loss Curvature Perspective on Training Instability in Deep Learning",
                        "abstract": "In this work, we study the evolution of the loss Hessian across many classification tasks in order to understand the effect the curvature of the loss has on the training dynamics. Whereas prior work has focused on how different learning rates affect the loss Hessian observed during training, we also analyze the effects of model initialization, architectural choices, and common training heuristics such as gradient clipping and learning rate warmup. Our results demonstrate that successful model and hyperparameter choices allow the early optimization trajectory to either avoid -- or navigate out of -- regions of high curvature and into flatter regions that tolerate a higher learning rate. Our results suggest a unifying perspective on how disparate mitigation strategies for training instability ultimately address the same underlying failure mode of neural network optimization, namely poor conditioning. Inspired by the conditioning perspective, we show that learning rate warmup can improve training stability just as much as batch normalization, layer normalization, MetaInit, GradInit, and Fixup initialization."
                    },
                    {
                        "title": "BUFFET: Benchmarking Large Language Models for Few-shot Cross-lingual Transfer",
                        "abstract": "Despite remarkable advancements in few-shot generalization in natural language processing, most models are developed and evaluated primarily in English. To facilitate research on few-shot cross-lingual transfer, we introduce a new benchmark, called BUFFET, which unifies 15 diverse tasks across 54 languages in a sequence-to-sequence format and provides a fixed set of few-shot examples and instructions. BUFFET is designed to establish a rigorous and equitable evaluation framework for few-shot cross-lingual transfer across a broad range of tasks and languages. Using BUFFET, we perform thorough evaluations of state-of-the-art multilingual large language models with different transfer methods, namely in-context learning and fine-tuning. Our findings reveal significant room for improvement in few-shot in-context cross-lingual transfer. In particular, ChatGPT with in-context learning often performs worse than much smaller mT5-base models fine-tuned on English task data and few-shot in-language examples. Our analysis suggests various avenues for future research in few-shot cross-lingual transfer, such as improved pretraining, understanding, and future evaluations."
                    }
                ]
            },
            "75d76da9-f627-414d-95d8-db12e6f22829": {
                "pk": "75d76da9-f627-414d-95d8-db12e6f22829",
                "name": "Aditya Kusupati",
                "collaborators": [
                    "Ali Farhadi",
                    "Prateek Jain",
                    "Vivek Ramanujan",
                    "Matthew Wallingford",
                    "Manik Varma",
                    "Sham Kakade",
                    "Aniruddha Kembhavi",
                    "Nilesh Gupta",
                    "Raghav Somani",
                    "Alex Fang"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Representation Learning",
                    "Adaptive Inference"
                ],
                "publications": [
                    {
                        "title": "Disrupting Model Training with Adversarial Shortcuts",
                        "abstract": "When data is publicly released for human consumption, it is unclear how to prevent its unauthorized usage for machine learning purposes. Successful model training may be preventable with carefully designed dataset modifications, and we present a proof-of-concept approach for the image classification setting. We propose methods based on the notion of adversarial shortcuts, which encourage models to rely on non-robust signals rather than semantic features, and our experiments demonstrate that these measures successfully prevent deep learning models from achieving high accuracy on real, unmodified data examples."
                    },
                    {
                        "title": "SHARCS: Efficient Transformers through Routing with Dynamic Width Sub-networks",
                        "abstract": "We introduce SHARCS for adaptive inference that takes into account the hardness of input samples. SHARCS can train a router on any transformer network, enabling the model to direct different samples to sub-networks with varying widths. Our experiments demonstrate that: (1) SHARCS outperforms or complements existing per-sample adaptive inference methods across various classification tasks in terms of accuracy vs. FLOPs; (2) SHARCS generalizes across different architectures and can be even applied to compressed and efficient transformer encoders to further improve their efficiency; (3) SHARCS can provide a 2 times inference speed up at an insignificant drop in accuracy."
                    },
                    {
                        "title": "Are \"Hierarchical\" Visual Representations Hierarchical?",
                        "abstract": "Learned visual representations often capture large amounts of semantic information for accurate downstream applications. Human understanding of the world is fundamentally grounded in hierarchy. To mimic this and further improve representation capabilities, the community has explored \"hierarchical\" visual representations that aim at modeling the underlying hierarchy of the visual world. In this work, we set out to investigate if hierarchical visual representations truly capture the human perceived hierarchy better than standard learned representations. To this end, we create HierNet, a suite of 12 datasets spanning 3 kinds of hierarchy from the BREEDs subset of ImageNet. After extensive evaluation of Hyperbolic and Matryoshka Representations across training setups, we conclude that they do not capture hierarchy any better than the standard representations but can assist in other aspects like search efficiency and interpretability. Our benchmark and the datasets are open-sourced at https://github.com/ethanlshen/HierNet."
                    },
                    {
                        "title": "MatMamba: A Matryoshka State Space Model",
                        "abstract": "State Space Models (SSMs) like Mamba2 are a promising alternative to Transformers, with faster theoretical training and inference times -- especially for long context lengths. Recent work on Matryoshka Representation Learning -- and its application to Transformer backbones in works like MatFormer -- showed how to introduce nested granularities of smaller submodels in one universal elastic model. In this work, we present MatMamba: a state space model which combines Matryoshka-style learning with Mamba2, by modifying the block to contain nested dimensions to enable joint training and adaptive inference. MatMamba allows for efficient and adaptive deployment across various model sizes. We train a single large MatMamba model and are able to get a number of smaller nested models for free -- while maintaining or improving upon the performance of a baseline smaller model trained from scratch. We train language and image models at a variety of parameter sizes from 35M to 1.4B. Our results on ImageNet and FineWeb show that MatMamba models scale comparably to Transformers, while having more efficient inference characteristics. This makes MatMamba a practically viable option for deploying large-scale models in an elastic way based on the available inference compute. Code and models are open sourced at \\url{https://github.com/ScaledFoundations/MatMamba}"
                    },
                    {
                        "title": "Extreme Regression for Dynamic Search Advertising",
                        "abstract": "This paper introduces a new learning paradigm called eXtreme Regression (XR) whose objective is to accurately predict the numerical degrees of relevance of an extremely large number of labels to a data point. XR can provide elegant solutions to many large-scale ranking and recommendation applications including Dynamic Search Advertising (DSA). XR can learn more accurate models than the recently popular extreme classifiers which incorrectly assume strictly binary-valued label relevances. Traditional regression metrics which sum the errors over all the labels are unsuitable for XR problems since they could give extremely loose bounds for the label ranking quality. Also, the existing regression algorithms won't efficiently scale to millions of labels. This paper addresses these limitations through: (1) new evaluation metrics for XR which sum only the k largest regression errors; (2) a new algorithm called XReg which decomposes XR task into a hierarchy of much smaller regression problems thus leading to highly efficient training and prediction. This paper also introduces a (3) new labelwise prediction algorithm in XReg useful for DSA and other recommendation tasks. Experiments on benchmark datasets demonstrated that XReg can outperform the state-of-the-art extreme classifiers as well as large-scale regressors and rankers by up to 50% reduction in the new XR error metric, and up to 2% and 2.4% improvements in terms of the propensity-scored precision metric used in extreme classification and the click-through rate metric used in DSA respectively. Deployment of XReg on DSA in Bing resulted in a relative gain of 27% in query coverage. XReg's source code can be downloaded from http://manikvarma.org/code/XReg/download.html."
                    },
                    {
                        "title": "FastGRNN: A Fast, Accurate, Stable and Tiny Kilobyte Sized Gated Recurrent Neural Network",
                        "abstract": "This paper develops the FastRNN and FastGRNN algorithms to address the twin RNN limitations of inaccurate training and inefficient prediction. Previous approaches have improved accuracy at the expense of prediction costs making them infeasible for resource-constrained and real-time applications. Unitary RNNs have increased accuracy somewhat by restricting the range of the state transition matrix's singular values but have also increased the model size as they require a larger number of hidden units to make up for the loss in expressive power. Gated RNNs have obtained state-of-the-art accuracies by adding extra parameters thereby resulting in even larger models. FastRNN addresses these limitations by adding a residual connection that does not constrain the range of the singular values explicitly and has only two extra scalar parameters. FastGRNN then extends the residual connection to a gate by reusing the RNN matrices to match state-of-the-art gated RNN accuracies but with a 2-4x smaller model. Enforcing FastGRNN's matrices to be low-rank, sparse and quantized resulted in accurate models that could be up to 35x smaller than leading gated and unitary RNNs. This allowed FastGRNN to accurately recognize the \"Hey Cortana\" wakeword with a 1 KB model and to be deployed on severely resource-constrained IoT microcontrollers too tiny to store other RNN models. FastGRNN's code is available at https://github.com/Microsoft/EdgeML/."
                    },
                    {
                        "title": "Soft Threshold Weight Reparameterization for Learnable Sparsity",
                        "abstract": "Sparsity in Deep Neural Networks (DNNs) is studied extensively with the focus of maximizing prediction accuracy given an overall parameter budget. Existing methods rely on uniform or heuristic non-uniform sparsity budgets which have sub-optimal layer-wise parameter allocation resulting in a) lower prediction accuracy or b) higher inference cost (FLOPs). This work proposes Soft Threshold Reparameterization (STR), a novel use of the soft-threshold operator on DNN weights. STR smoothly induces sparsity while learning pruning thresholds thereby obtaining a non-uniform sparsity budget. Our method achieves state-of-the-art accuracy for unstructured sparsity in CNNs (ResNet50 and MobileNetV1 on ImageNet-1K), and, additionally, learns non-uniform budgets that empirically reduce the FLOPs by up to 50%. Notably, STR boosts the accuracy over existing results by up to 10% in the ultra sparse (99%) regime and can also be used to induce low-rank (structured sparsity) in RNNs. In short, STR is a simple mechanism which learns effective sparsity budgets that contrast with popular heuristics. Code, pretrained models and sparsity budgets are at https://github.com/RAIVNLab/STR."
                    },
                    {
                        "title": "RNNPool: Efficient Non-linear Pooling for RAM Constrained Inference",
                        "abstract": "Standard Convolutional Neural Networks (CNNs) designed for computer vision tasks tend to have large intermediate activation maps. These require large working memory and are thus unsuitable for deployment on resource-constrained devices typically used for inference on the edge. Aggressively downsampling the images via pooling or strided convolutions can address the problem but leads to a significant decrease in accuracy due to gross aggregation of the feature map by standard pooling operators. In this paper, we introduce RNNPool, a novel pooling operator based on Recurrent Neural Networks (RNNs), that efficiently aggregates features over large patches of an image and rapidly downsamples activation maps. Empirical evaluation indicates that an RNNPool layer can effectively replace multiple blocks in a variety of architectures such as MobileNets, DenseNet when applied to standard vision tasks like image classification and face detection. That is, RNNPool can significantly decrease computational complexity and peak memory usage for inference while retaining comparable accuracy. We use RNNPool with the standard S3FD architecture to construct a face detection method that achieves state-of-the-art MAP for tiny ARM Cortex-M4 class microcontrollers with under 256 KB of RAM. Code is released at https://github.com/Microsoft/EdgeML."
                    },
                    {
                        "title": "LLC: Accurate, Multi-purpose Learnt Low-dimensional Binary Codes",
                        "abstract": "Learning binary representations of instances and classes is a classical problem with several high potential applications. In modern settings, the compression of high-dimensional neural representations to low-dimensional binary codes is a challenging task and often require large bit-codes to be accurate. In this work, we propose a novel method for Learning Low-dimensional binary Codes (LLC) for instances as well as classes. Our method does not require any side-information, like annotated attributes or label meta-data, and learns extremely low-dimensional binary codes (~20 bits for ImageNet-1K). The learnt codes are super-efficient while still ensuring nearly optimal classification accuracy for ResNet50 on ImageNet-1K. We demonstrate that the learnt codes capture intrinsically important features in the data, by discovering an intuitive taxonomy over classes. We further quantitatively measure the quality of our codes by applying it to the efficient image retrieval as well as out-of-distribution (OOD) detection problems. For ImageNet-100 retrieval problem, our learnt binary codes outperform 16 bit HashNet using only 10 bits and also are as accurate as 10 dimensional real representations. Finally, our learnt binary codes can perform OOD detection, out-of-the-box, as accurately as a baseline that needs ~3000 samples to tune its threshold, while we require none. Code is open-sourced at https://github.com/RAIVNLab/LLC."
                    },
                    {
                        "title": "One Size Does Not Fit All: Multi-Scale, Cascaded RNNs for Radar Classification",
                        "abstract": "Edge sensing with micro-power pulse-Doppler radars is an emergent domain in monitoring and surveillance with several smart city applications. Existing solutions for the clutter versus multi-source radar classification task are limited in terms of either accuracy or efficiency, and in some cases, struggle with a trade-off between false alarms and recall of sources. We find that this problem can be resolved by learning the classifier across multiple time-scales. We propose a multi-scale, cascaded recurrent neural network architecture, MSC-RNN, comprised of an efficient multi-instance learning (MIL) Recurrent Neural Network (RNN) for clutter discrimination at a lower tier, and a more complex RNN classifier for source classification at the upper tier. By controlling the invocation of the upper RNN with the help of the lower tier conditionally, MSC-RNN achieves an overall accuracy of 0.972. Our approach holistically improves the accuracy and per-class recalls over ML models suitable for radar inferencing. Notably, we outperform cross-domain handcrafted feature engineering with time-domain deep feature learning, while also being up to $\\sim$3$\\times$ more efficient than a competitive solution."
                    },
                    {
                        "title": "Neural Radiance Field Codebooks",
                        "abstract": "Compositional representations of the world are a promising step towards enabling high-level scene understanding and efficient transfer to downstream tasks. Learning such representations for complex scenes and tasks remains an open challenge. Towards this goal, we introduce Neural Radiance Field Codebooks (NRC), a scalable method for learning object-centric representations through novel view reconstruction. NRC learns to reconstruct scenes from novel views using a dictionary of object codes which are decoded through a volumetric renderer. This enables the discovery of reoccurring visual and geometric patterns across scenes which are transferable to downstream tasks. We show that NRC representations transfer well to object navigation in THOR, outperforming 2D and 3D representation learning methods by 3.1% success rate. We demonstrate that our approach is able to perform unsupervised segmentation for more complex synthetic (THOR) and real scenes (NYU Depth) better than prior methods (29% relative improvement). Finally, we show that NRC improves on the task of depth ordering by 5.5% accuracy in THOR."
                    },
                    {
                        "title": "Matryoshka Representation Learning",
                        "abstract": "Learned representations are a central component in modern ML systems, serving a multitude of downstream tasks. When training such representations, it is often the case that computational and statistical constraints for each downstream task are unknown. In this context rigid, fixed capacity representations can be either over or under-accommodating to the task at hand. This leads us to ask: can we design a flexible representation that can adapt to multiple downstream tasks with varying computational resources? Our main contribution is Matryoshka Representation Learning (MRL) which encodes information at different granularities and allows a single embedding to adapt to the computational constraints of downstream tasks. MRL minimally modifies existing representation learning pipelines and imposes no additional cost during inference and deployment. MRL learns coarse-to-fine representations that are at least as accurate and rich as independently trained low-dimensional representations. The flexibility within the learned Matryoshka Representations offer: (a) up to 14x smaller embedding size for ImageNet-1K classification at the same level of accuracy; (b) up to 14x real-world speed-ups for large-scale retrieval on ImageNet-1K and 4K; and (c) up to 2% accuracy improvements for long-tail few-shot classification, all while being as robust as the original representations. Finally, we show that MRL extends seamlessly to web-scale datasets (ImageNet, JFT) across various modalities -- vision (ViT, ResNet), vision + language (ALIGN) and language (BERT). MRL code and pretrained models are open-sourced at https://github.com/RAIVNLab/MRL."
                    },
                    {
                        "title": "ProtoSound: A Personalized and Scalable Sound Recognition System for Deaf and Hard-of-Hearing Users",
                        "abstract": "Recent advances have enabled automatic sound recognition systems for deaf and hard of hearing (DHH) users on mobile devices. However, these tools use pre-trained, generic sound recognition models, which do not meet the diverse needs of DHH users. We introduce ProtoSound, an interactive system for customizing sound recognition models by recording a few examples, thereby enabling personalized and fine-grained categories. ProtoSound is motivated by prior work examining sound awareness needs of DHH people and by a survey we conducted with 472 DHH participants. To evaluate ProtoSound, we characterized performance on two real-world sound datasets, showing significant improvement over state-of-the-art (e.g., +9.7% accuracy on the first dataset). We then deployed ProtoSound's end-user training and real-time recognition through a mobile application and recruited 19 hearing participants who listened to the real-world sounds and rated the accuracy across 56 locations (e.g., homes, restaurants, parks). Results show that ProtoSound personalized the model on-device in real-time and accurately learned sounds across diverse acoustic contexts. We close by discussing open challenges in personalizable sound recognition, including the need for better recording interfaces and algorithmic improvements."
                    },
                    {
                        "title": "Neural Priming for Sample-Efficient Adaptation",
                        "abstract": "We propose Neural Priming, a technique for adapting large pretrained models to distribution shifts and downstream tasks given few or no labeled examples. Presented with class names or unlabeled test samples, Neural Priming enables the model to recall and conditions its parameters on relevant data seen throughout pretraining, thereby priming it for the test distribution. Neural Priming can be performed at test time, even for pretraining datasets as large as LAION-2B. Performing lightweight updates on the recalled data significantly improves accuracy across a variety of distribution shift and transfer learning benchmarks. Concretely, in the zero-shot setting, we see a 2.45% improvement in accuracy on ImageNet and 3.81% accuracy improvement on average across standard transfer learning benchmarks. Further, using Neural Priming at inference to adapt to distribution shift, we see a 1.41% accuracy improvement on ImageNetV2. These results demonstrate the effectiveness of Neural Priming in addressing the challenge of limited labeled data and changing distributions. Code is available at github.com/RAIVNLab/neural-priming."
                    },
                    {
                        "title": "EHI: End-to-end Learning of Hierarchical Index for Efficient Dense Retrieval",
                        "abstract": "Dense embedding-based retrieval is widely used for semantic search and ranking. However, conventional two-stage approaches, involving contrastive embedding learning followed by approximate nearest neighbor search (ANNS), can suffer from misalignment between these stages. This mismatch degrades retrieval performance. We propose End-to-end Hierarchical Indexing (EHI), a novel method that directly addresses this issue by jointly optimizing embedding generation and ANNS structure. EHI leverages a dual encoder for embedding queries and documents while simultaneously learning an inverted file index (IVF)-style tree structure. To facilitate the effective learning of this discrete structure, EHI introduces dense path embeddings that encodes the path traversed by queries and documents within the tree. Extensive evaluations on standard benchmarks, including MS MARCO (Dev set) and TREC DL19, demonstrate EHI's superiority over traditional ANNS index. Under the same computational constraints, EHI outperforms existing state-of-the-art methods by +1.45% in MRR@10 on MS MARCO (Dev) and +8.2% in nDCG@10 on TREC DL19, highlighting the benefits of our end-to-end approach."
                    },
                    {
                        "title": "FLUID: A Unified Evaluation Framework for Flexible Sequential Data",
                        "abstract": "Modern ML methods excel when training data is IID, large-scale, and well labeled. Learning in less ideal conditions remains an open challenge. The sub-fields of few-shot, continual, transfer, and representation learning have made substantial strides in learning under adverse conditions; each affording distinct advantages through methods and insights. These methods address different challenges such as data arriving sequentially or scarce training examples, however often the difficult conditions an ML system will face over its lifetime cannot be anticipated prior to deployment. Therefore, general ML systems which can handle the many challenges of learning in practical settings are needed. To foster research towards the goal of general ML methods, we introduce a new unified evaluation framework - FLUID (Flexible Sequential Data). FLUID integrates the objectives of few-shot, continual, transfer, and representation learning while enabling comparison and integration of techniques across these subfields. In FLUID, a learner faces a stream of data and must make sequential predictions while choosing how to update itself, adapt quickly to novel classes, and deal with changing data distributions; while accounting for the total amount of compute. We conduct experiments on a broad set of methods which shed new insight on the advantages and limitations of current solutions and indicate new research problems to solve. As a starting point towards more general methods, we present two new baselines which outperform other evaluated methods on FLUID. Project page: https://raivn.cs.washington.edu/projects/FLUID/."
                    },
                    {
                        "title": "Mixture of Nested Experts: Adaptive Processing of Visual Tokens",
                        "abstract": "The visual medium (images and videos) naturally contains a large amount of information redundancy, thereby providing a great opportunity for leveraging efficiency in processing. While Vision Transformer (ViT) based models scale effectively to large data regimes, they fail to capitalize on this inherent redundancy, leading to higher computational costs. Mixture of Experts (MoE) networks demonstrate scalability while maintaining same inference-time costs, but they come with a larger parameter footprint. We present Mixture of Nested Experts (MoNE), which utilizes a nested structure for experts, wherein individual experts fall on an increasing compute-accuracy curve. Given a compute budget, MoNE learns to dynamically choose tokens in a priority order, and thus redundant tokens are processed through cheaper nested experts. Using this framework, we achieve equivalent performance as the baseline models, while reducing inference time compute by over two-fold. We validate our approach on standard image and video datasets - ImageNet-21K, Kinetics400, and Something-Something-v2. We further highlight MoNE$'$s adaptability by showcasing its ability to maintain strong performance across different inference-time compute budgets on videos, using only a single trained model."
                    },
                    {
                        "title": "MADLAD-400: A Multilingual And Document-Level Large Audited Dataset",
                        "abstract": "We introduce MADLAD-400, a manually audited, general domain 3T token monolingual dataset based on CommonCrawl, spanning 419 languages. We discuss the limitations revealed by self-auditing MADLAD-400, and the role data auditing had in the dataset creation process. We then train and release a 10.7B-parameter multilingual machine translation model on 250 billion tokens covering over 450 languages using publicly available data, and find that it is competitive with models that are significantly larger, and report the results on different domains. In addition, we train a 8B-parameter language model, and assess the results on few-shot translation. We make the baseline models available to the research community."
                    },
                    {
                        "title": "AdANNS: A Framework for Adaptive Semantic Search",
                        "abstract": "Web-scale search systems learn an encoder to embed a given query which is then hooked into an approximate nearest neighbor search (ANNS) pipeline to retrieve similar data points. To accurately capture tail queries and data points, learned representations typically are rigid, high-dimensional vectors that are generally used as-is in the entire ANNS pipeline and can lead to computationally expensive retrieval. In this paper, we argue that instead of rigid representations, different stages of ANNS can leverage adaptive representations of varying capacities to achieve significantly better accuracy-compute trade-offs, i.e., stages of ANNS that can get away with more approximate computation should use a lower-capacity representation of the same data point. To this end, we introduce AdANNS, a novel ANNS design framework that explicitly leverages the flexibility of Matryoshka Representations. We demonstrate state-of-the-art accuracy-compute trade-offs using novel AdANNS-based key ANNS building blocks like search data structures (AdANNS-IVF) and quantization (AdANNS-OPQ). For example on ImageNet retrieval, AdANNS-IVF is up to 1.5% more accurate than the rigid representations-based IVF at the same compute budget; and matches accuracy while being up to 90x faster in wall-clock time. For Natural Questions, 32-byte AdANNS-OPQ matches the accuracy of the 64-byte OPQ baseline constructed using rigid representations -- same accuracy at half the cost! We further show that the gains from AdANNS translate to modern-day composite ANNS indices that combine search structures and quantization. Finally, we demonstrate that AdANNS can enable inference-time adaptivity for compute-aware search on ANNS indices built non-adaptively on matryoshka representations. Code is open-sourced at https://github.com/RAIVNLab/AdANNS."
                    }
                ]
            },
            "5faa43f8-cc1e-4da5-91e2-72625720c053": {
                "pk": "5faa43f8-cc1e-4da5-91e2-72625720c053",
                "name": "Tim Dettmers",
                "collaborators": [
                    "Luke Zettlemoyer",
                    "Alexander Borzunov",
                    "Mike Lewis",
                    "Max Ryabinin",
                    "Younes Belkada",
                    "Pasquale Minervini",
                    "Pontus Stenetorp",
                    "Sebastian Riedel",
                    "Michael Diskin",
                    "Noah A. Smith"
                ],
                "domain": [
                    "Deep Learning",
                    "Model Compression",
                    "Natural Language Processing",
                    "Distributed Systems"
                ],
                "publications": [
                    {
                        "title": "8-Bit Approximations for Parallelism in Deep Learning",
                        "abstract": "The creation of practical deep learning data-products often requires parallelization across processors and computers to make deep learning feasible on large data sets, but bottlenecks in communication bandwidth make it difficult to attain good speedups through parallelism. Here we develop and test 8-bit approximation algorithms which make better use of the available bandwidth by compressing 32-bit gradients and nonlinear activations to 8-bit approximations. We show that these approximations do not decrease predictive performance on MNIST, CIFAR10, and ImageNet for both model and data parallelism and provide a data transfer speedup of 2x relative to 32-bit parallelism. We build a predictive model for speedups based on our experimental data, verify its validity on known speedup data, and show that we can obtain a speedup of 50x and more on a system of 96 GPUs compared to a speedup of 23x for 32-bit. We compare our data types with other methods and show that 8-bit approximations achieve state-of-the-art speedups for model parallelism. Thus 8-bit approximation is an efficient method to parallelize convolutional networks on very large systems of GPUs."
                    },
                    {
                        "title": "Sparse Networks from Scratch: Faster Training without Losing Performance",
                        "abstract": "We demonstrate the possibility of what we call sparse learning: accelerated training of deep neural networks that maintain sparse weights throughout training while achieving dense performance levels. We accomplish this by developing sparse momentum, an algorithm which uses exponentially smoothed gradients (momentum) to identify layers and weights which reduce the error efficiently. Sparse momentum redistributes pruned weights across layers according to the mean momentum magnitude of each layer. Within a layer, sparse momentum grows weights according to the momentum magnitude of zero-valued weights. We demonstrate state-of-the-art sparse performance on MNIST, CIFAR-10, and ImageNet, decreasing the mean error by a relative 8%, 15%, and 6% compared to other sparse algorithms. Furthermore, we show that sparse momentum reliably reproduces dense performance levels while providing up to 5.61x faster training. In our analysis, ablations show that the benefits of momentum redistribution and growth increase with the depth and size of the network. Additionally, we find that sparse momentum is insensitive to the choice of its hyperparameters suggesting that sparse momentum is robust and easy to use."
                    },
                    {
                        "title": "The case for 4-bit precision: k-bit Inference Scaling Laws",
                        "abstract": "Quantization methods reduce the number of bits required to represent each parameter in a model, trading accuracy for smaller memory footprints and inference latencies. However, the final model size depends on both the number of parameters of the original model and the rate of compression. For example, a 30B 8-bit model and a 60B 4-bit model have the same number of bits but may have very different zero-shot accuracies. In this work, we study this trade-off by developing inference scaling laws of zero-shot performance in Large Language Models (LLMs) to determine the bit-precision and model size that maximizes zero-shot performance. We run more than 35,000 experiments with 16-bit inputs and k-bit parameters to examine which zero-shot quantization methods improve scaling for 3 to 8-bit precision at scales of 19M to 176B parameters across the LLM families BLOOM, OPT, NeoX/Pythia, and GPT-2. We find that it is challenging to improve the bit-level scaling trade-off, with the only improvements being the use of a small block size -- splitting the parameters into small independently quantized blocks -- and the quantization data type being used (e.g., Int vs Float). Overall, our findings show that {4-bit} precision is almost universally optimal for total model bits and zero-shot accuracy."
                    },
                    {
                        "title": "8-bit Optimizers via Block-wise Quantization",
                        "abstract": "Stateful optimizers maintain gradient statistics over time, e.g., the exponentially smoothed sum (SGD with momentum) or squared sum (Adam) of past gradient values. This state can be used to accelerate optimization compared to plain stochastic gradient descent but uses memory that might otherwise be allocated to model parameters, thereby limiting the maximum size of models trained in practice. In this paper, we develop the first optimizers that use 8-bit statistics while maintaining the performance levels of using 32-bit optimizer states. To overcome the resulting computational, quantization, and stability challenges, we develop block-wise dynamic quantization. Block-wise quantization divides input tensors into smaller blocks that are independently quantized. Each block is processed in parallel across cores, yielding faster optimization and high precision quantization. To maintain stability and performance, we combine block-wise quantization with two additional changes: (1) dynamic quantization, a form of non-linear optimization that is precise for both large and small magnitude values, and (2) a stable embedding layer to reduce gradient variance that comes from the highly non-uniform distribution of input tokens in language models. As a result, our 8-bit optimizers maintain 32-bit performance with a small fraction of the memory footprint on a range of tasks, including 1.5B parameter language modeling, GLUE finetuning, ImageNet classification, WMT'14 machine translation, MoCo v2 contrastive ImageNet pretraining+finetuning, and RoBERTa pretraining, without changes to the original optimizer hyperparameters. We open-source our 8-bit optimizers as a drop-in replacement that only requires a two-line code change."
                    },
                    {
                        "title": "LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale",
                        "abstract": "Large language models have been widely adopted but require significant GPU memory for inference. We develop a procedure for Int8 matrix multiplication for feed-forward and attention projection layers in transformers, which cut the memory needed for inference by half while retaining full precision performance. With our method, a 175B parameter 16/32-bit checkpoint can be loaded, converted to Int8, and used immediately without performance degradation. This is made possible by understanding and working around properties of highly systematic emergent features in transformer language models that dominate attention and transformer predictive performance. To cope with these features, we develop a two-part quantization procedure, LLM.int8(). We first use vector-wise quantization with separate normalization constants for each inner product in the matrix multiplication, to quantize most of the features. However, for the emergent outliers, we also include a new mixed-precision decomposition scheme, which isolates the outlier feature dimensions into a 16-bit matrix multiplication while still more than 99.9% of values are multiplied in 8-bit. Using LLM.int8(), we show empirically it is possible to perform inference in LLMs with up to 175B parameters without any performance degradation. This result makes such models much more accessible, for example making it possible to use OPT-175B/BLOOM on a single server with consumer GPUs. We open-source our software."
                    },
                    {
                        "title": "QLoRA: Efficient Finetuning of Quantized LLMs",
                        "abstract": "We present QLoRA, an efficient finetuning approach that reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU while preserving full 16-bit finetuning task performance. QLoRA backpropagates gradients through a frozen, 4-bit quantized pretrained language model into Low Rank Adapters~(LoRA). Our best model family, which we name Guanaco, outperforms all previous openly released models on the Vicuna benchmark, reaching 99.3% of the performance level of ChatGPT while only requiring 24 hours of finetuning on a single GPU. QLoRA introduces a number of innovations to save memory without sacrificing performance: (a) 4-bit NormalFloat (NF4), a new data type that is information theoretically optimal for normally distributed weights (b) double quantization to reduce the average memory footprint by quantizing the quantization constants, and (c) paged optimziers to manage memory spikes. We use QLoRA to finetune more than 1,000 models, providing a detailed analysis of instruction following and chatbot performance across 8 instruction datasets, multiple model types (LLaMA, T5), and model scales that would be infeasible to run with regular finetuning (e.g. 33B and 65B parameter models). Our results show that QLoRA finetuning on a small high-quality dataset leads to state-of-the-art results, even when using smaller models than the previous SoTA. We provide a detailed analysis of chatbot performance based on both human and GPT-4 evaluations showing that GPT-4 evaluations are a cheap and reasonable alternative to human evaluation. Furthermore, we find that current chatbot benchmarks are not trustworthy to accurately evaluate the performance levels of chatbots. A lemon-picked analysis demonstrates where Guanaco fails compared to ChatGPT. We release all of our models and code, including CUDA kernels for 4-bit training."
                    },
                    {
                        "title": "Convolutional 2D Knowledge Graph Embeddings",
                        "abstract": "Link prediction for knowledge graphs is the task of predicting missing relationships between entities. Previous work on link prediction has focused on shallow, fast models which can scale to large knowledge graphs. However, these models learn less expressive features than deep, multi-layer models -- which potentially limits performance. In this work, we introduce ConvE, a multi-layer convolutional network model for link prediction, and report state-of-the-art results for several established datasets. We also show that the model is highly parameter efficient, yielding the same performance as DistMult and R-GCN with 8x and 17x fewer parameters. Analysis of our model suggests that it is particularly effective at modelling nodes with high indegree -- which are common in highly-connected, complex knowledge graphs such as Freebase and YAGO3. In addition, it has been noted that the WN18 and FB15k datasets suffer from test set leakage, due to inverse relations from the training set being present in the test set -- however, the extent of this issue has so far not been quantified. We find this problem to be severe: a simple rule-based model can achieve state-of-the-art results on both WN18 and FB15k. To ensure that models are evaluated on datasets where simply exploiting inverse relations cannot yield competitive results, we investigate and validate several commonly used datasets -- deriving robust variants where necessary. We then perform experiments on these robust datasets for our own and several previously proposed models and find that ConvE achieves state-of-the-art Mean Reciprocal Rank across most datasets."
                    },
                    {
                        "title": "SWARM Parallelism: Training Large Models Can Be Surprisingly Communication-Efficient",
                        "abstract": "Many deep learning applications benefit from using large models with billions of parameters. Training these models is notoriously expensive due to the need for specialized HPC clusters. In this work, we consider alternative setups for training large models: using cheap \"preemptible\" instances or pooling existing resources from multiple regions. We analyze the performance of existing model-parallel algorithms in these conditions and find configurations where training larger models becomes less communication-intensive. Based on these findings, we propose SWARM parallelism, a model-parallel training algorithm designed for poorly connected, heterogeneous and unreliable devices. SWARM creates temporary randomized pipelines between nodes that are rebalanced in case of failure. We empirically validate our findings and compare SWARM parallelism with existing large-scale training approaches. Finally, we combine our insights with compression strategies to train a large Transformer language model with 1B shared parameters (approximately 13B before sharing) on preemptible T4 GPUs with less than 200Mb/s network."
                    },
                    {
                        "title": "BASE Layers: Simplifying Training of Large, Sparse Models",
                        "abstract": "We introduce a new balanced assignment of experts (BASE) layer for large language models that greatly simplifies existing high capacity sparse layers. Sparse layers can dramatically improve the efficiency of training and inference by routing each token to specialized expert modules that contain only a small fraction of the model parameters. However, it can be difficult to learn balanced routing functions that make full use of the available experts; existing approaches typically use routing heuristics or auxiliary expert-balancing loss functions. In contrast, we formulate token-to-expert allocation as a linear assignment problem, allowing an optimal assignment in which each expert receives an equal number of tokens. This optimal assignment scheme improves efficiency by guaranteeing balanced compute loads, and also simplifies training by not requiring any new hyperparameters or auxiliary losses. Code is publicly released at https://github.com/pytorch/fairseq/"
                    },
                    {
                        "title": "Branch-Train-Merge: Embarrassingly Parallel Training of Expert Language Models",
                        "abstract": "We present Branch-Train-Merge (BTM), a communication-efficient algorithm for embarrassingly parallel training of large language models (LLMs). We show it is possible to independently train subparts of a new class of LLMs on different subsets of the data, eliminating the massive multi-node synchronization currently required to train LLMs. BTM learns a set of independent expert LMs (ELMs), each specialized to a different textual domain, such as scientific or legal text. These ELMs can be added and removed to update data coverage, ensembled to generalize to new domains, or averaged to collapse back to a single LM for efficient inference. New ELMs are learned by branching from (mixtures of) ELMs in the current set, further training the parameters on data for the new domain, and then merging the resulting model back into the set for future use. Experiments show that BTM improves in- and out-of-domain perplexities as compared to GPT-style Transformer LMs, when controlling for training cost. Through extensive analysis, we show that these results are robust to different ELM initialization schemes, but require expert domain specialization; LM ensembles with random data splits do not perform well. We also present a study of scaling BTM into a new corpus of 64 domains (192B whitespace-separated tokens in total); the resulting LM (22.4B total parameters) performs as well as a Transformer LM trained with 2.5 times more compute. These gains grow with the number of domains, suggesting more aggressive parallelism could be used to efficiently train larger models in future work."
                    },
                    {
                        "title": "Stable and low-precision training for large-scale vision-language models",
                        "abstract": "We introduce new methods for 1) accelerating and 2) stabilizing training for large language-vision models. 1) For acceleration, we introduce SwitchBack, a linear layer for int8 quantized training which provides a speed-up of 13-25% while matching the performance of bfloat16 training within 0.1 percentage points for the 1B parameter CLIP ViT-Huge -- the largest int8 training to date. Our main focus is int8 as GPU support for float8 is rare, though we also analyze float8 training through simulation. While SwitchBack proves effective for float8, we show that standard techniques are also successful if the network is trained and initialized so that large feature magnitudes are discouraged, which we accomplish via layer-scale initialized with zeros. 2) For stability, we analyze loss spikes and find they consistently occur 1-8 iterations after the squared gradients become under-estimated by their AdamW second moment estimator. As a result, we recommend an AdamW-Adafactor hybrid which avoids loss spikes when training a CLIP ViT-Huge model and outperforms gradient clipping at the scales we test."
                    },
                    {
                        "title": "Towards A Unified View of Sparse Feed-Forward Network in Pretraining Large Language Model",
                        "abstract": "Large and sparse feed-forward layers (S-FFN) such as Mixture-of-Experts (MoE) have proven effective in scaling up Transformers model size for \\textit{pretraining} large language models. By only activating part of the FFN parameters conditioning on input, S-FFN improves generalization performance while keeping training and inference costs (in FLOPs) fixed. In this work, we analyzed two major design choices of S-FFN: the memory block (a.k.a. expert) size and the memory block selection method under a general conceptual framework of sparse neural memory. Using this unified framework, we compare several S-FFN architectures for language modeling and provide insights into their relative efficacy and efficiency. We found a simpler selection method -- \\textbf{\\texttt{Avg-K}} that selects blocks through their mean aggregated hidden states, achieving lower perplexity in language model pretraining compared to existing MoE architectures including Switch Transformer (Fedus et al., 2021) and HashLayer (Roller et al., 2021)."
                    },
                    {
                        "title": "Jack the Reader - A Machine Reading Framework",
                        "abstract": "Many Machine Reading and Natural Language Understanding tasks require reading supporting text in order to answer questions. For example, in Question Answering, the supporting text can be newswire or Wikipedia articles; in Natural Language Inference, premises can be seen as the supporting text and hypotheses as questions. Providing a set of useful primitives operating in a single framework of related tasks would allow for expressive modelling, and easier model comparison and replication. To that end, we present Jack the Reader (Jack), a framework for Machine Reading that allows for quick model prototyping by component reuse, evaluation of new models on existing datasets as well as integrating new datasets and applying them on a growing set of implemented baseline models. Jack is currently supporting (but not limited to) three tasks: Question Answering, Natural Language Inference, and Link Prediction. It is developed with the aim of increasing research efficiency and code reuse."
                    },
                    {
                        "title": "Training Transformers Together",
                        "abstract": "The infrastructure necessary for training state-of-the-art models is becoming overly expensive, which makes training such models affordable only to large corporations and institutions. Recent work proposes several methods for training such models collaboratively, i.e., by pooling together hardware from many independent parties and training a shared model over the Internet. In this demonstration, we collaboratively trained a text-to-image transformer similar to OpenAI DALL-E. We invited the viewers to join the ongoing training run, showing them instructions on how to contribute using the available hardware. We explained how to address the engineering challenges associated with such a training run (slow communication, limited memory, uneven performance between devices, and security concerns) and discussed how the viewers can set up collaborative training runs themselves. Finally, we show that the resulting model generates images of reasonable quality on a number of prompts."
                    },
                    {
                        "title": "Petals: Collaborative Inference and Fine-tuning of Large Models",
                        "abstract": "Many NLP tasks benefit from using large language models (LLMs) that often have more than 100 billion parameters. With the release of BLOOM-176B and OPT-175B, everyone can download pretrained models of this scale. Still, using these models requires high-end hardware unavailable to many researchers. In some cases, LLMs can be used more affordably via RAM offloading or hosted APIs. However, these techniques have innate limitations: offloading is too slow for interactive inference, while APIs are not flexible enough for research that requires access to weights, attention or logits. In this work, we propose Petals - a system for inference and fine-tuning of large models collaboratively by joining the resources of multiple parties. We demonstrate that this strategy outperforms offloading for very large models, running inference of BLOOM-176B on consumer GPUs with $\\approx$ 1 step per second, which is enough for many interactive LLM applications. Unlike most inference APIs, Petals also natively exposes hidden states of served models, allowing to train and share custom model extensions based on efficient fine-tuning methods."
                    },
                    {
                        "title": "SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression",
                        "abstract": "Recent advances in large language model (LLM) pretraining have led to high-quality LLMs with impressive abilities. By compressing such LLMs via quantization to 3-4 bits per parameter, they can fit into memory-limited devices such as laptops and mobile phones, enabling personalized use. However, quantization down to 3-4 bits per parameter usually leads to moderate-to-high accuracy losses, especially for smaller models in the 1-10B parameter range, which are well-suited for edge deployments. To address this accuracy issue, we introduce the Sparse-Quantized Representation (SpQR), a new compressed format and quantization technique which enables for the first time near-lossless compression of LLMs across model scales, while reaching similar compression levels to previous methods. SpQR works by identifying and isolating outlier weights, which cause particularly-large quantization errors, and storing them in higher precision, while compressing all other weights to 3-4 bits, and achieves relative accuracy losses of less than 1% in perplexity for highly-accurate LLaMA and Falcon LLMs. This makes it possible to run 33B parameter LLM on a single 24 GB consumer GPU without any performance degradation at 15% speedup thus making powerful LLMs available to consumer without any downsides. SpQR comes with efficient algorithms for both encoding weights into its format, as well as decoding them efficiently at runtime. Specifically, we provide an efficient GPU inference algorithm for SpQR which yields faster inference than 16-bit baselines at similar accuracy, while enabling memory compression gains of more than 4x."
                    },
                    {
                        "title": "Distributed Inference and Fine-tuning of Large Language Models Over The Internet",
                        "abstract": "Large language models (LLMs) are useful in many NLP tasks and become more capable with size, with the best open-source models having over 50 billion parameters. However, using these 50B+ models requires high-end hardware, making them inaccessible to most researchers. In this work, we investigate methods for cost-efficient inference and fine-tuning of LLMs, comparing local and distributed strategies. We observe that a large enough model (50B+) can run efficiently even on geodistributed devices in a consumer-grade network. This could allow running LLM efficiently by pooling together idle compute resources of multiple research groups and volunteers. We address two open problems: (1) how to perform inference and fine-tuning reliably if any device can disconnect abruptly and (2) how to partition LLMs between devices with uneven hardware, joining and leaving at will. In order to do that, we develop special fault-tolerant inference algorithms and load-balancing protocols that automatically assign devices to maximize the total system throughput. We showcase these algorithms in Petals - a decentralized system that runs Llama 2 (70B) and BLOOM (176B) over the Internet up to 10x faster than offloading for interactive generation. We evaluate the performance of our system in simulated conditions and a real-world setup spanning two continents."
                    },
                    {
                        "title": "Scaling Retrieval-Based Language Models with a Trillion-Token Datastore",
                        "abstract": "Scaling laws with respect to the amount of training data and the number of parameters allow us to predict the cost-benefit trade-offs of pretraining language models (LMs) in different configurations. In this paper, we consider another dimension of scaling: the amount of data available at inference time. Specifically, we find that increasing the size of the datastore used by a retrieval-based LM monotonically improves language modeling and several downstream tasks without obvious saturation, such that a smaller model augmented with a large datastore outperforms a larger LM-only model on knowledge-intensive tasks. By plotting compute-optimal scaling curves with varied datastore, model, and pretraining data sizes, we show that using larger datastores can significantly improve model performance for the same training compute budget. We carry out our study by constructing a 1.4 trillion-token datastore named MassiveDS, which is the largest and the most diverse open-sourced datastore for retrieval-based LMs to date, and designing an efficient pipeline for studying datastore scaling in a computationally accessible manner. Finally, we analyze the effect of improving the retriever, datastore quality filtering, and other design choices on our observed scaling trends. Overall, our results show that datastore size should be considered as an integral part of LM efficiency and performance trade-offs. To facilitate future research, we open-source our datastore and code at https://github.com/RulinShao/retrieval-scaling."
                    },
                    {
                        "title": "OLMoE: Open Mixture-of-Experts Language Models",
                        "abstract": "We introduce OLMoE, a fully open, state-of-the-art language model leveraging sparse Mixture-of-Experts (MoE). OLMoE-1B-7B has 7 billion (B) parameters but uses only 1B per input token. We pretrain it on 5 trillion tokens and further adapt it to create OLMoE-1B-7B-Instruct. Our models outperform all available models with similar active parameters, even surpassing larger ones like Llama2-13B-Chat and DeepSeekMoE-16B. We present various experiments on MoE training, analyze routing in our model showing high specialization, and open-source all aspects of our work: model weights, training data, code, and logs."
                    }
                ]
            },
            "05775000-e82b-478a-9a4e-b3ab5ed2c283": {
                "pk": "05775000-e82b-478a-9a4e-b3ab5ed2c283",
                "name": "Kaifeng Chen",
                "collaborators": [
                    "Shuwei Zhu",
                    "Siying Lv",
                    "Wei Fang",
                    "Leilei Cao",
                    "Nikolaos-Antonios Ypsilantis",
                    "Andr\u00e9 Araujo",
                    "Ond\u0159ej Chum",
                    "Daniel Salz",
                    "Huiwen Chang",
                    "Kihyuk Sohn"
                ],
                "domain": [
                    "Plasma Physics",
                    "Hydrodynamics",
                    "Machine Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Super Guassian Distribution of Laser Produced Plasmas",
                        "abstract": "The competition between inverse Bremsstrahlung, which pushes the distribution towards a super-Guassian, and the thermalization through self-collisions is considered. In the summary, basic formulas are provided and some numerical results by Fokker-Planck simulation are plotted."
                    },
                    {
                        "title": "Local and Non-local Thermal Transport Simulation",
                        "abstract": "In laser produced plasmas, when temperature gradient is deep, local thermal transport is not dominant. Classical theory of Spitzer-Harm should not be used in this case. Instead, Fokker-Planck simulation method should be utilized. In this paper, research on thermal transport in both local and non-local cases are discussed. Numerical results by using Epperlein scheme are provided and compared."
                    },
                    {
                        "title": "A New Method of Analyzing Eigenvalues and Eigenfunctions of Linearized Rosenbluth Collision Operator",
                        "abstract": "Eigenvalue spectrum has been a long term unsolved problem for plasma physicists. In this paper, some numerical calculations are conducted about the minimum eigenvalues of the linearized Rosenbluth collision operator and the differential part of the whole operator. Similarities of the two operator lead to further discussions about the differential operator. Completeness and orthogonality are rigorously proved. Moreover, the continuum of the whole operator is indicated."
                    },
                    {
                        "title": "Research progress on intelligent optimization techniques for energy-efficient design of ship hull forms",
                        "abstract": "The design optimization of ship hull form based on hydrodynamics theory and simulation-based design (SBD) technologies generally considers ship performance and energy efficiency performance as the design objective, which plays an important role in smart design and manufacturing of green ship. An optimal design of sustainable energy system requires multidisciplinary tools to build ships with the least resistance and energy consumption. Through a systematic approach, this paper presents the research progress of energy-efficient design of ship hull forms based on intelligent optimization techniques. We discuss different methods involved in the optimization procedure, especially the latest developments of intelligent optimization algorithms and surrogate models. Moreover, current development trends and technical challenges of multidisciplinary design optimization and surrogate-assisted evolutionary algorithms for ship design are further analyzed. We explore the gaps and potential future directions, so as to paving the way towards the design of the next generation of more energy-efficient ship hull form."
                    },
                    {
                        "title": "UDON: Universal Dynamic Online distillatioN for generic image representations",
                        "abstract": "Universal image representations are critical in enabling real-world fine-grained and instance-level recognition applications, where objects and entities from any domain must be identified at large scale. Despite recent advances, existing methods fail to capture important domain-specific knowledge, while also ignoring differences in data distribution across different domains. This leads to a large performance gap between efficient universal solutions and expensive approaches utilising a collection of specialist models, one for each domain. In this work, we make significant strides towards closing this gap, by introducing a new learning technique, dubbed UDON (Universal Dynamic Online DistillatioN). UDON employs multi-teacher distillation, where each teacher is specialized in one domain, to transfer detailed domain-specific knowledge into the student universal embedding. UDON's distillation approach is not only effective, but also very efficient, by sharing most model parameters between the student and all teachers, where all models are jointly trained in an online manner. UDON also comprises a sampling technique which adapts the training process to dynamically allocate batches to domains which are learned slower and require more frequent processing. This boosts significantly the learning of complex domains which are characterised by a large number of classes and long-tail distributions. With comprehensive experiments, we validate each component of UDON, and showcase significant improvements over the state of the art in the recent UnED benchmark. Code: https://github.com/nikosips/UDON ."
                    },
                    {
                        "title": "Improve Supervised Representation Learning with Masked Image Modeling",
                        "abstract": "Training visual embeddings with labeled data supervision has been the de facto setup for representation learning in computer vision. Inspired by recent success of adopting masked image modeling (MIM) in self-supervised representation learning, we propose a simple yet effective setup that can easily integrate MIM into existing supervised training paradigms. In our design, in addition to the original classification task applied to a vision transformer image encoder, we add a shallow transformer-based decoder on top of the encoder and introduce an MIM task which tries to reconstruct image tokens based on masked image inputs. We show with minimal change in architecture and no overhead in inference that this setup is able to improve the quality of the learned representations for downstream tasks such as classification, image retrieval, and semantic segmentation. We conduct a comprehensive study and evaluation of our setup on public benchmarks. On ImageNet-1k, our ViT-B/14 model achieves 81.72% validation accuracy, 2.01% higher than the baseline model. On K-Nearest-Neighbor image retrieval evaluation with ImageNet-1k, the same model outperforms the baseline by 1.32%. We also show that this setup can be easily scaled to larger models and datasets. Code and checkpoints will be released."
                    }
                ]
            },
            "d59b452d-df36-4f34-a0b5-684584fdf5a8": {
                "pk": "d59b452d-df36-4f34-a0b5-684584fdf5a8",
                "name": "Inderjit Dhillon",
                "collaborators": [
                    "Prateek Jain",
                    "Nikhil Rao",
                    "Rashish Tandon",
                    "Si Si",
                    "Pradeep Ravikumar",
                    "Devvrit",
                    "Minhao Cheng",
                    "Cho-Jui Hsieh",
                    "Anish Acharya",
                    "Rahul Goel"
                ],
                "domain": [
                    "Machine Learning",
                    "Optimization",
                    "Natural Language Processing",
                    "Time Series Analysis"
                ],
                "publications": [
                    {
                        "title": "Structured Sparse Regression via Greedy Hard-Thresholding",
                        "abstract": "Several learning applications require solving high-dimensional regression problems where the relevant features belong to a small number of (overlapping) groups. For very large datasets and under standard sparsity constraints, hard thresholding methods have proven to be extremely efficient, but such methods require NP hard projections when dealing with overlapping groups. In this paper, we show that such NP-hard projections can not only be avoided by appealing to submodular optimization, but such methods come with strong theoretical guarantees even in the presence of poorly conditioned data (i.e. say when two features have correlation $\\geq 0.99$), which existing analyses cannot handle. These methods exhibit an interesting computation-accuracy trade-off and can be extended to significantly harder problems such as sparse overlapping groups. Experiments on both real and synthetic data validate our claims and demonstrate that the proposed methods are orders of magnitude faster than other greedy and convex relaxation techniques for learning with group-structured sparsity."
                    },
                    {
                        "title": "Kernel Ridge Regression via Partitioning",
                        "abstract": "In this paper, we investigate a divide and conquer approach to Kernel Ridge Regression (KRR). Given n samples, the division step involves separating the points based on some underlying disjoint partition of the input space (possibly via clustering), and then computing a KRR estimate for each partition. The conquering step is simple: for each partition, we only consider its own local estimate for prediction. We establish conditions under which we can give generalization bounds for this estimator, as well as achieve optimal minimax rates. We also show that the approximation error component of the generalization error is lesser than when a single KRR estimate is fit on the data: thus providing both statistical and computational advantages over a single KRR estimate over the entire data (or an averaging over random partitions as in other recent work, [30]). Lastly, we provide experimental validation for our proposed estimator and our assumptions."
                    },
                    {
                        "title": "Voting based ensemble improves robustness of defensive models",
                        "abstract": "Developing robust models against adversarial perturbations has been an active area of research and many algorithms have been proposed to train individual robust models. Taking these pretrained robust models, we aim to study whether it is possible to create an ensemble to further improve robustness. Several previous attempts tackled this problem by ensembling the soft-label prediction and have been proved vulnerable based on the latest attack methods. In this paper, we show that if the robust training loss is diverse enough, a simple hard-label based voting ensemble can boost the robust error over each individual model. Furthermore, given a pool of robust models, we develop a principled way to select which models to ensemble. Finally, to verify the improved robustness, we conduct extensive experiments to study how to attack a voting-based ensemble and develop several new white-box attacks. On CIFAR-10 dataset, by ensembling several state-of-the-art pre-trained defense models, our method can achieve a 59.8% robust accuracy, outperforming all the existing defensive models without using additional data."
                    },
                    {
                        "title": "Online Embedding Compression for Text Classification using Low Rank Matrix Factorization",
                        "abstract": "Deep learning models have become state of the art for natural language processing (NLP) tasks, however deploying these models in production system poses significant memory constraints. Existing compression methods are either lossy or introduce significant latency. We propose a compression method that leverages low rank matrix factorization during training,to compress the word embedding layer which represents the size bottleneck for most NLP models. Our models are trained, compressed and then further re-trained on the downstream task to recover accuracy while maintaining the reduced size. Empirically, we show that the proposed method can achieve 90% compression with minimal impact in accuracy for sentence classification tasks, and outperforms alternative methods like fixed-point quantization or offline word embedding compression. We also analyze the inference time and storage space for our method through FLOP calculations, showing that we can compress DNN models by a configurable ratio and regain accuracy loss without introducing additional latency compared to fixed point quantization. Finally, we introduce a novel learning rate schedule, the Cyclically Annealed Learning Rate (CALR), which we empirically demonstrate to outperform other popular adaptive learning rate algorithms on a sentence classification benchmark."
                    },
                    {
                        "title": "Think Globally, Act Locally: A Deep Neural Network Approach to High-Dimensional Time Series Forecasting",
                        "abstract": "Forecasting high-dimensional time series plays a crucial role in many applications such as demand forecasting and financial predictions. Modern datasets can have millions of correlated time-series that evolve together, i.e they are extremely high dimensional (one dimension for each individual time-series). There is a need for exploiting global patterns and coupling them with local calibration for better prediction. However, most recent deep learning approaches in the literature are one-dimensional, i.e, even though they are trained on the whole dataset, during prediction, the future forecast for a single dimension mainly depends on past values from the same dimension. In this paper, we seek to correct this deficiency and propose DeepGLO, a deep forecasting model which thinks globally and acts locally. In particular, DeepGLO is a hybrid model that combines a global matrix factorization model regularized by a temporal convolution network, along with another temporal network that can capture local properties of each time-series and associated covariates. Our model can be trained effectively on high-dimensional but diverse time series, where different time series can have vastly different scales, without a priori normalization or rescaling. Empirical results demonstrate that DeepGLO can outperform state-of-the-art approaches; for example, we see more than 25% improvement in WAPE over other methods on a public dataset that contains more than 100K-dimensional time series."
                    },
                    {
                        "title": "Coordinate Descent Methods for Symmetric Nonnegative Matrix Factorization",
                        "abstract": "Given a symmetric nonnegative matrix $A$, symmetric nonnegative matrix factorization (symNMF) is the problem of finding a nonnegative matrix $H$, usually with much fewer columns than $A$, such that $A \\approx HH^T$. SymNMF can be used for data analysis and in particular for various clustering tasks. In this paper, we propose simple and very efficient coordinate descent schemes to solve this problem, and that can handle large and sparse input matrices. The effectiveness of our methods is illustrated on synthetic and real-world data sets, and we show that they perform favorably compared to recent state-of-the-art methods."
                    },
                    {
                        "title": "Extreme Multi-label Classification from Aggregated Labels",
                        "abstract": "Extreme multi-label classification (XMC) is the problem of finding the relevant labels for an input, from a very large universe of possible labels. We consider XMC in the setting where labels are available only for groups of samples - but not for individual ones. Current XMC approaches are not built for such multi-instance multi-label (MIML) training data, and MIML approaches do not scale to XMC sizes. We develop a new and scalable algorithm to impute individual-sample labels from the group labels; this can be paired with any existing XMC method to solve the aggregated label problem. We characterize the statistical properties of our algorithm under mild assumptions, and provide a new end-to-end framework for MIML as an extension. Experiments on both aggregated label XMC and MIML tasks show the advantages over existing approaches."
                    },
                    {
                        "title": "Approximate Newton policy gradient algorithms",
                        "abstract": "Policy gradient algorithms have been widely applied to Markov decision processes and reinforcement learning problems in recent years. Regularization with various entropy functions is often used to encourage exploration and improve stability. This paper proposes an approximate Newton method for the policy gradient algorithm with entropy regularization. In the case of Shannon entropy, the resulting algorithm reproduces the natural policy gradient algorithm. For other entropy functions, this method results in brand-new policy gradient algorithms. We prove that all these algorithms enjoy Newton-type quadratic convergence and that the corresponding gradient flow converges globally to the optimal solution. We use synthetic and industrial-scale examples to demonstrate that the proposed approximate Newton method typically converges in single-digit iterations, often orders of magnitude faster than other state-of-the-art algorithms."
                    }
                ]
            },
            "520d1be5-ffd7-47ed-b8e0-e1714553a2e6": {
                "pk": "520d1be5-ffd7-47ed-b8e0-e1714553a2e6",
                "name": "Yulia Tsvetkov",
                "collaborators": [
                    "Sachin Kumar",
                    "Anjalie Field",
                    "Xiaochuang Han",
                    "Manaal Faruqui",
                    "Chris Dyer",
                    "Wang Ling",
                    "Brian MacWhinney",
                    "Dheeraj Rajagopal",
                    "Vidhisha Balachandran",
                    "Eduard Hovy"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Bias Detection",
                    "Machine Learning",
                    "Text Generation"
                ],
                "publications": [
                    {
                        "title": "Entity-Centric Contextual Affective Analysis",
                        "abstract": "While contextualized word representations have improved state-of-the-art benchmarks in many NLP tasks, their potential usefulness for social-oriented tasks remains largely unexplored. We show how contextualized word embeddings can be used to capture affect dimensions in portrayals of people. We evaluate our methodology quantitatively, on held-out affect lexicons, and qualitatively, through case examples. We find that contextualized word representations do encode meaningful affect information, but they are heavily biased towards their training data, which limits their usefulness to in-domain analyses. We ultimately use our method to examine differences in portrayals of men and women."
                    },
                    {
                        "title": "Fortifying Toxic Speech Detectors Against Veiled Toxicity",
                        "abstract": "Modern toxic speech detectors are incompetent in recognizing disguised offensive language, such as adversarial attacks that deliberately avoid known toxic lexicons, or manifestations of implicit bias. Building a large annotated dataset for such veiled toxicity can be very expensive. In this work, we propose a framework aimed at fortifying existing toxic speech detectors without a large labeled corpus of veiled toxicity. Just a handful of probing examples are used to surface orders of magnitude more disguised offenses. We augment the toxic speech detector's training data with these discovered offensive examples, thereby making it more robust to veiled toxicity while preserving its utility in detecting overt toxicity."
                    },
                    {
                        "title": "Unsupervised Discovery of Implicit Gender Bias",
                        "abstract": "Despite their prevalence in society, social biases are difficult to identify, primarily because human judgements in this domain can be unreliable. We take an unsupervised approach to identifying gender bias against women at a comment level and present a model that can surface text likely to contain bias. Our main challenge is forcing the model to focus on signs of implicit bias, rather than other artifacts in the data. Thus, our methodology involves reducing the influence of confounds through propensity matching and adversarial learning. Our analysis shows how biased comments directed towards female politicians contain mixed criticisms, while comments directed towards other female public figures focus on appearance and sexualization. Ultimately, our work offers a way to capture subtle biases in various domains without relying on subjective human judgements."
                    },
                    {
                        "title": "Influence Tuning: Demoting Spurious Correlations via Instance Attribution and Instance-Driven Updates",
                        "abstract": "Among the most critical limitations of deep learning NLP models are their lack of interpretability, and their reliance on spurious correlations. Prior work proposed various approaches to interpreting the black-box models to unveil the spurious correlations, but the research was primarily used in human-computer interaction scenarios. It still remains underexplored whether or how such model interpretations can be used to automatically \"unlearn\" confounding features. In this work, we propose influence tuning--a procedure that leverages model interpretations to update the model parameters towards a plausible interpretation (rather than an interpretation that relies on spurious patterns in the data) in addition to learning to predict the task labels. We show that in a controlled setup, influence tuning can help deconfounding the model from spurious patterns in data, significantly outperforming baseline methods that use adversarial training."
                    },
                    {
                        "title": "ORCA: Interpreting Prompted Language Models via Locating Supporting Data Evidence in the Ocean of Pretraining Data",
                        "abstract": "Large pretrained language models have been performing increasingly well in a variety of downstream tasks via prompting. However, it remains unclear from where the model learns the task-specific knowledge, especially in a zero-shot setup. In this work, we want to find evidence of the model's task-specific competence from pretraining and are specifically interested in locating a very small subset of pretraining data that directly supports the model in the task. We call such a subset supporting data evidence and propose a novel method ORCA to effectively identify it, by iteratively using gradient information related to the downstream task. This supporting data evidence offers interesting insights about the prompted language models: in the tasks of sentiment analysis and textual entailment, BERT shows a substantial reliance on BookCorpus, the smaller corpus of BERT's two pretraining corpora, as well as on pretraining examples that mask out synonyms to the task verbalizers."
                    },
                    {
                        "title": "Correlation-based Intrinsic Evaluation of Word Vector Representations",
                        "abstract": "We introduce QVEC-CCA--an intrinsic evaluation metric for word vector representations based on correlations of learned vectors with features extracted from linguistic resources. We show that QVEC-CCA scores are an effective proxy for a range of extrinsic semantic and syntactic tasks. We also show that the proposed evaluation obtains higher and more consistent correlations with downstream tasks, compared to existing approaches to intrinsic evaluation of word vectors that are based on word similarity."
                    },
                    {
                        "title": "Von Mises-Fisher Loss for Training Sequence to Sequence Models with Continuous Outputs",
                        "abstract": "The Softmax function is used in the final layer of nearly all existing sequence-to-sequence models for language generation. However, it is usually the slowest layer to compute which limits the vocabulary size to a subset of most frequent types; and it has a large memory footprint. We propose a general technique for replacing the softmax layer with a continuous embedding layer. Our primary innovations are a novel probabilistic loss, and a training and inference procedure in which we generate a probability distribution over pre-trained word embeddings, instead of a multinomial distribution over the vocabulary obtained via softmax. We evaluate this new class of sequence-to-sequence models with continuous outputs on the task of neural machine translation. We show that our models obtain upto 2.5x speed-up in training time while performing on par with the state-of-the-art models in terms of translation quality. These models are capable of handling very large vocabularies without compromising on translation quality. They also produce more meaningful errors than in the softmax-based models, as these errors typically lie in a subspace of the vector space of the reference translations."
                    },
                    {
                        "title": "Learning the Curriculum with Bayesian Optimization for Task-Specific Word Representation Learning",
                        "abstract": "We use Bayesian optimization to learn curricula for word representation learning, optimizing performance on downstream tasks that depend on the learned representations as features. The curricula are modeled by a linear ranking function which is the scalar product of a learned weight vector and an engineered feature vector that characterizes the different aspects of the complexity of each instance in the training corpus. We show that learning the curriculum improves performance on a variety of downstream tasks over random orders and in comparison to the natural corpus order."
                    },
                    {
                        "title": "SelfExplain: A Self-Explaining Architecture for Neural Text Classifiers",
                        "abstract": "We introduce SelfExplain, a novel self-explaining model that explains a text classifier's predictions using phrase-based concepts. SelfExplain augments existing neural classifiers by adding (1) a globally interpretable layer that identifies the most influential concepts in the training set for a given sample and (2) a locally interpretable layer that quantifies the contribution of each local input concept by computing a relevance score relative to the predicted label. Experiments across five text-classification datasets show that SelfExplain facilitates interpretability without sacrificing performance. Most importantly, explanations from SelfExplain show sufficiency for model predictions and are perceived as adequate, trustworthy and understandable by human judges compared to existing widely-used baselines."
                    },
                    {
                        "title": "An Exploration of Data Augmentation Techniques for Improving English to Tigrinya Translation",
                        "abstract": "It has been shown that the performance of neural machine translation (NMT) drops starkly in low-resource conditions, often requiring large amounts of auxiliary data to achieve competitive results. An effective method of generating auxiliary data is back-translation of target language sentences. In this work, we present a case study of Tigrinya where we investigate several back-translation methods to generate synthetic source sentences. We find that in low-resource conditions, back-translation by pivoting through a higher-resource language related to the target language proves most effective resulting in substantial improvements over baselines."
                    },
                    {
                        "title": "Improving the Diversity of Unsupervised Paraphrasing with Embedding Outputs",
                        "abstract": "We present a novel technique for zero-shot paraphrase generation. The key contribution is an end-to-end multilingual paraphrasing model that is trained using translated parallel corpora to generate paraphrases into \"meaning spaces\" -- replacing the final softmax layer with word embeddings. This architectural modification, plus a training procedure that incorporates an autoencoding objective, enables effective parameter sharing across languages for more fluent monolingual rewriting, and facilitates fluency and diversity in generation. Our continuous-output paraphrase generation models outperform zero-shot paraphrasing baselines when evaluated on two languages using a battery of computational metrics as well as in human assessment."
                    },
                    {
                        "title": "Extracting Lexical Features from Dialects via Interpretable Dialect Classifiers",
                        "abstract": "Identifying linguistic differences between dialects of a language often requires expert knowledge and meticulous human analysis. This is largely due to the complexity and nuance involved in studying various dialects. We present a novel approach to extract distinguishing lexical features of dialects by utilizing interpretable dialect classifiers, even in the absence of human experts. We explore both post-hoc and intrinsic approaches to interpretability, conduct experiments on Mandarin, Italian, and Low Saxon, and experimentally demonstrate that our method successfully identifies key language-specific lexical features that contribute to dialectal variations."
                    },
                    {
                        "title": "Problems With Evaluation of Word Embeddings Using Word Similarity Tasks",
                        "abstract": "Lacking standardized extrinsic evaluation methods for vector representations of words, the NLP community has relied heavily on word similarity tasks as a proxy for intrinsic evaluation of word vectors. Word similarity evaluation, which correlates the distance between vectors and human judgments of semantic similarity is attractive, because it is computationally inexpensive and fast. In this paper we present several problems associated with the evaluation of word vectors on word similarity datasets, and summarize existing solutions. Our study suggests that the use of word similarity tasks for evaluation of word vectors is not sustainable and calls for further research on evaluation methods."
                    },
                    {
                        "title": "Native Language Cognate Effects on Second Language Lexical Choice",
                        "abstract": "We present a computational analysis of cognate effects on the spontaneous linguistic productions of advanced non-native speakers. Introducing a large corpus of highly competent non-native English speakers, and using a set of carefully selected lexical items, we show that the lexical choices of non-natives are affected by cognates in their native language. This effect is so powerful that we are able to reconstruct the phylogenetic language tree of the Indo-European language family solely from the frequencies of specific lexical items in the English of authors with various native languages. We quantitatively analyze non-native lexical choice, highlighting cognate facilitation as one of the important phenomena shaping the language of non-native speakers."
                    },
                    {
                        "title": "Contextual Affective Analysis: A Case Study of People Portrayals in Online #MeToo Stories",
                        "abstract": "In October 2017, numerous women accused producer Harvey Weinstein of sexual harassment. Their stories encouraged other women to voice allegations of sexual harassment against many high profile men, including politicians, actors, and producers. These events are broadly referred to as the #MeToo movement, named for the use of the hashtag \"#metoo\" on social media platforms like Twitter and Facebook. The movement has widely been referred to as \"empowering\" because it has amplified the voices of previously unheard women over those of traditionally powerful men. In this work, we investigate dynamics of sentiment, power and agency in online media coverage of these events. Using a corpus of online media articles about the #MeToo movement, we present a contextual affective analysis---an entity-centric approach that uses contextualized lexicons to examine how people are portrayed in media articles. We show that while these articles are sympathetic towards women who have experienced sexual harassment, they consistently present men as most powerful, even after sexual assault allegations. While we focus on media coverage of the #MeToo movement, our method for contextual affective analysis readily generalizes to other domains."
                    },
                    {
                        "title": "Demoting Racial Bias in Hate Speech Detection",
                        "abstract": "In current hate speech datasets, there exists a high correlation between annotators' perceptions of toxicity and signals of African American English (AAE). This bias in annotated training data and the tendency of machine learning models to amplify it cause AAE text to often be mislabeled as abusive/offensive/hate speech with a high false positive rate by current hate speech classifiers. In this paper, we use adversarial training to mitigate this bias, introducing a hate speech classifier that learns to detect toxic sentences while demoting confounds corresponding to AAE texts. Experimental results on a hate speech dataset and an AAE dataset suggest that our method is able to substantially reduce the false positive rate for AAE text while only minimally affecting the performance of hate speech classification."
                    },
                    {
                        "title": "A Deep Reinforced Model for Zero-Shot Cross-Lingual Summarization with Bilingual Semantic Similarity Rewards",
                        "abstract": "Cross-lingual text summarization aims at generating a document summary in one language given input in another language. It is a practically important but under-explored task, primarily due to the dearth of available data. Existing methods resort to machine translation to synthesize training data, but such pipeline approaches suffer from error propagation. In this work, we propose an end-to-end cross-lingual text summarization model. The model uses reinforcement learning to directly optimize a bilingual semantic similarity metric between the summaries generated in a target language and gold summaries in a source language. We also introduce techniques to pre-train the model leveraging monolingual summarization and machine translation objectives. Experimental results in both English--Chinese and English--German cross-lingual summarization settings demonstrate the effectiveness of our methods. In addition, we find that reinforcement learning models with bilingual semantic similarity as rewards generate more fluent sentences than strong baselines."
                    },
                    {
                        "title": "A Framework for the Computational Linguistic Analysis of Dehumanization",
                        "abstract": "Dehumanization is a pernicious psychological process that often leads to extreme intergroup bias, hate speech, and violence aimed at targeted social groups. Despite these serious consequences and the wealth of available data, dehumanization has not yet been computationally studied on a large scale. Drawing upon social psychology research, we create a computational linguistic framework for analyzing dehumanizing language by identifying linguistic correlates of salient components of dehumanization. We then apply this framework to analyze discussions of LGBTQ people in the New York Times from 1986 to 2015. Overall, we find increasingly humanizing descriptions of LGBTQ people over time. However, we find that the label homosexual has emerged to be much more strongly associated with dehumanizing attitudes than other labels, such as gay. Our proposed techniques highlight processes of linguistic variation and change in discourses surrounding marginalized groups. Furthermore, the ability to analyze dehumanizing language at a large scale has implications for automatically detecting and understanding media bias as well as abusive language online."
                    },
                    {
                        "title": "Balancing Training for Multilingual Neural Machine Translation",
                        "abstract": "When training multilingual machine translation (MT) models that can translate to/from multiple languages, we are faced with imbalanced training sets: some languages have much more training data than others. Standard practice is to up-sample less resourced languages to increase representation, and the degree of up-sampling has a large effect on the overall performance. In this paper, we propose a method that instead automatically learns how to weight training data through a data scorer that is optimized to maximize performance on all test languages. Experiments on two sets of languages under both one-to-many and many-to-one MT settings show our method not only consistently outperforms heuristic baselines in terms of average performance, but also offers flexible control over the performance of which languages are optimized."
                    },
                    {
                        "title": "Controlled Text Generation as Continuous Optimization with Multiple Constraints",
                        "abstract": "As large-scale language model pretraining pushes the state-of-the-art in text generation, recent work has turned to controlling attributes of the text such models generate. While modifying the pretrained models via fine-tuning remains the popular approach, it incurs a significant computational cost and can be infeasible due to lack of appropriate data. As an alternative, we propose MuCoCO -- a flexible and modular algorithm for controllable inference from pretrained models. We formulate the decoding process as an optimization problem which allows for multiple attributes we aim to control to be easily incorporated as differentiable constraints to the optimization. By relaxing this discrete optimization to a continuous one, we make use of Lagrangian multipliers and gradient-descent based techniques to generate the desired text. We evaluate our approach on controllable machine translation and style transfer with multiple sentence-level attributes and observe significant improvements over baselines."
                    }
                ]
            },
            "3300765f-6da9-432e-aad9-49c0212df60e": {
                "pk": "3300765f-6da9-432e-aad9-49c0212df60e",
                "name": "Hannaneh Hajishirzi",
                "collaborators": [
                    "Ali Farhadi",
                    "Mari Ostendorf",
                    "Sewon Min",
                    "Minjoon Seo",
                    "Eyal Amir",
                    "Akari Asai",
                    "Rik Koncel-Kedziorski",
                    "Yi Luan",
                    "Sachin Mehta",
                    "Linda Shapiro"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Knowledge Graphs",
                    "Question Answering"
                ],
                "publications": [
                    {
                        "title": "Sampling First Order Logical Particles",
                        "abstract": "Approximate inference in dynamic systems is the problem of estimating the state of the system given a sequence of actions and partial observations. High precision estimation is fundamental in many applications like diagnosis, natural language processing, tracking, planning, and robotics. In this paper we present an algorithm that samples possible deterministic executions of a probabilistic sequence. The algorithm takes advantage of a compact representation (using first order logic) for actions and world states to improve the precision of its estimation. Theoretical and empirical results show that the algorithm's expected error is smaller than propositional sampling and Sequential Monte Carlo (SMC) sampling techniques."
                    },
                    {
                        "title": "Logic-Guided Data Augmentation and Regularization for Consistent Question Answering",
                        "abstract": "Many natural language questions require qualitative, quantitative or logical comparisons between two entities or events. This paper addresses the problem of improving the accuracy and consistency of responses to comparison questions by integrating logic rules and neural models. Our method leverages logical and linguistic knowledge to augment labeled training data and then uses a consistency-based regularizer to train the model. Improving the global consistency of predictions, our approach achieves large improvements over previous methods in a variety of question answering (QA) tasks including multiple-choice qualitative reasoning, cause-effect reasoning, and extractive machine reading comprehension. In particular, our method significantly improves the performance of RoBERTa-based models by 1-5% across datasets. We advance the state of the art by around 5-8% on WIQA and QuaRel and reduce consistency violations by 58% on HotpotQA. We further demonstrate that our approach can learn effectively from limited data."
                    },
                    {
                        "title": "Data-Driven Methods for Solving Algebra Word Problems",
                        "abstract": "We explore contemporary, data-driven techniques for solving math word problems over recent large-scale datasets. We show that well-tuned neural equation classifiers can outperform more sophisticated models such as sequence to sequence and self-attention across these datasets. Our error analysis indicates that, while fully data driven models show some promise, semantic and world knowledge is necessary for further advances."
                    },
                    {
                        "title": "UnifiedQA-v2: Stronger Generalization via Broader Cross-Format Training",
                        "abstract": "We present UnifiedQA-v2, a QA model built with the same process as UnifiedQA, except that it utilizes more supervision -- roughly 3x the number of datasets used for UnifiedQA. This generally leads to better in-domain and cross-domain results."
                    },
                    {
                        "title": "Reasoning about RoboCup Soccer Narratives",
                        "abstract": "This paper presents an approach for learning to translate simple narratives, i.e., texts (sequences of sentences) describing dynamic systems, into coherent sequences of events without the need for labeled training data. Our approach incorporates domain knowledge in the form of preconditions and effects of events, and we show that it outperforms state-of-the-art supervised learning systems on the task of reconstructing RoboCup soccer games from their commentaries."
                    },
                    {
                        "title": "Scientific Relation Extraction with Selectively Incorporated Concept Embeddings",
                        "abstract": "This paper describes our submission for the SemEval 2018 Task 7 shared task on semantic relation extraction and classification in scientific papers. We extend the end-to-end relation extraction model of (Miwa and Bansal) with enhancements such as a character-level encoding attention mechanism on selecting pretrained concept candidate embeddings. Our official submission ranked the second in relation classification task (Subtask 1.1 and Subtask 2 Senerio 2), and the first in the relation extraction task (Subtask 2 Scenario 1)."
                    },
                    {
                        "title": "Extracting Summary Knowledge Graphs from Long Documents",
                        "abstract": "Knowledge graphs capture entities and relations from long documents and can facilitate reasoning in many downstream applications. Extracting compact knowledge graphs containing only salient entities and relations is important but challenging for understanding and summarizing long documents. We introduce a new text-to-graph task of predicting summarized knowledge graphs from long documents. We develop a dataset of 200k document/graph pairs using automatic and human annotations. We also develop strong baselines for this task based on graph learning and text summarization, and provide quantitative and qualitative studies of their effect."
                    },
                    {
                        "title": "Scientific Information Extraction with Semi-supervised Neural Tagging",
                        "abstract": "This paper addresses the problem of extracting keyphrases from scientific articles and categorizing them as corresponding to a task, process, or material. We cast the problem as sequence tagging and introduce semi-supervised methods to a neural tagging model, which builds on recent advances in named entity recognition. Since annotated training data is scarce in this domain, we introduce a graph-based semi-supervised algorithm together with a data selection scheme to leverage unannotated articles. Both inductive and transductive semi-supervised learning strategies outperform state-of-the-art information extraction performance on the 2017 SemEval Task 10 ScienceIE task."
                    },
                    {
                        "title": "Are Elephants Bigger than Butterflies? Reasoning about Sizes of Objects",
                        "abstract": "Human vision greatly benefits from the information about sizes of objects. The role of size in several visual reasoning tasks has been thoroughly explored in human perception and cognition. However, the impact of the information about sizes of objects is yet to be determined in AI. We postulate that this is mainly attributed to the lack of a comprehensive repository of size information. In this paper, we introduce a method to automatically infer object sizes, leveraging visual and textual information from web. By maximizing the joint likelihood of textual and visual observations, our method learns reliable relative size estimates, with no explicit human supervision. We introduce the relative size dataset and show that our method outperforms competitive textual and visual baselines in reasoning about size comparisons."
                    },
                    {
                        "title": "Disfluency Detection using a Bidirectional LSTM",
                        "abstract": "We introduce a new approach for disfluency detection using a Bidirectional Long-Short Term Memory neural network (BLSTM). In addition to the word sequence, the model takes as input pattern match features that were developed to reduce sensitivity to vocabulary size in training, which lead to improved performance over the word sequence alone. The BLSTM takes advantage of explicit repair states in addition to the standard reparandum states. The final output leverages integer linear programming to incorporate constraints of disfluency structure. In experiments on the Switchboard corpus, the model achieves state-of-the-art performance for both the standard disfluency detection task and the correction detection task. Analysis shows that the model has better detection of non-repetition disfluencies, which tend to be much harder to detect."
                    },
                    {
                        "title": "Query-Reduction Networks for Question Answering",
                        "abstract": "In this paper, we study the problem of question answering when reasoning over multiple facts is required. We propose Query-Reduction Network (QRN), a variant of Recurrent Neural Network (RNN) that effectively handles both short-term (local) and long-term (global) sequential dependencies to reason over multiple facts. QRN considers the context sentences as a sequence of state-changing triggers, and reduces the original query to a more informed query as it observes each trigger (context sentence) through time. Our experiments show that QRN produces the state-of-the-art results in bAbI QA and dialog tasks, and in a real goal-oriented dialog dataset. In addition, QRN formulation allows parallelization on RNN's time axis, saving an order of magnitude in time complexity for training and inference."
                    },
                    {
                        "title": "View-Driven Deduplication with Active Learning",
                        "abstract": "Visual analytics systems such as Tableau are increasingly popular for interactive data exploration. These tools, however, do not currently assist users with detecting or resolving potential data quality problems including the well-known deduplication problem. Recent approaches for deduplication focus on cleaning entire datasets and commonly require hundreds to thousands of user labels. In this paper, we address the problem of deduplication in the context of visual data analytics. We present a new approach for record deduplication that strives to produce the cleanest view possible with a limited budget for data labeling. The key idea behind our approach is to consider the impact that individual tuples have on a visualization and to monitor how the view changes during cleaning. With experiments on nine different visualizations for two real-world datasets, we show that our approach produces significantly cleaner views for small labeling budgets than state-of-the-art alternatives and that it also stops the cleaning process after requesting fewer labels."
                    },
                    {
                        "title": "Neural Speed Reading via Skim-RNN",
                        "abstract": "Inspired by the principles of speed reading, we introduce Skim-RNN, a recurrent neural network (RNN) that dynamically decides to update only a small fraction of the hidden state for relatively unimportant input tokens. Skim-RNN gives computational advantage over an RNN that always updates the entire hidden state. Skim-RNN uses the same input and output interfaces as a standard RNN and can be easily used instead of RNNs in existing models. In our experiments, we show that Skim-RNN can achieve significantly reduced computational cost without losing accuracy compared to standard RNNs across five different natural language tasks. In addition, we demonstrate that the trade-off between accuracy and speed of Skim-RNN can be dynamically controlled during inference time in a stable manner. Our analysis also shows that Skim-RNN running on a single CPU offers lower latency compared to standard RNNs on GPUs."
                    },
                    {
                        "title": "Identifying Most Walkable Direction for Navigation in an Outdoor Environment",
                        "abstract": "We present an approach for identifying the most walkable direction for navigation using a hand-held camera. Our approach extracts semantically rich contextual information from the scene using a custom encoder-decoder architecture for semantic segmentation and models the spatial and temporal behavior of objects in the scene using a spatio-temporal graph. The system learns to minimize a cost function over the spatial and temporal object attributes to identify the most walkable direction. We construct a new annotated navigation dataset collected using a hand-held mobile camera in an unconstrained outdoor environment, which includes challenging settings such as highly dynamic scenes, occlusion between objects, and distortions. Our system achieves an accuracy of 84% on predicting a safe direction. We also show that our custom segmentation network is both fast and accurate, achieving mIOU (mean intersection over union) scores of 81 and 44.7 on the PASCAL VOC and the PASCAL Context datasets, respectively, while running at about 21 frames per second."
                    },
                    {
                        "title": "ESPNetv2: A Light-weight, Power Efficient, and General Purpose Convolutional Neural Network",
                        "abstract": "We introduce a light-weight, power efficient, and general purpose convolutional neural network, ESPNetv2, for modeling visual and sequential data. Our network uses group point-wise and depth-wise dilated separable convolutions to learn representations from a large effective receptive field with fewer FLOPs and parameters. The performance of our network is evaluated on four different tasks: (1) object classification, (2) semantic segmentation, (3) object detection, and (4) language modeling. Experiments on these tasks, including image classification on the ImageNet and language modeling on the PenTree bank dataset, demonstrate the superior performance of our method over the state-of-the-art methods. Our network outperforms ESPNet by 4-5% and has 2-4x fewer FLOPs on the PASCAL VOC and the Cityscapes dataset. Compared to YOLOv2 on the MS-COCO object detection, ESPNetv2 delivers 4.4% higher accuracy with 6x fewer FLOPs. Our experiments show that ESPNetv2 is much more power efficient than existing state-of-the-art efficient methods including ShuffleNets and MobileNets. Our code is open-source and available at https://github.com/sacmehta/ESPNetv2"
                    },
                    {
                        "title": "Multi-hop Reading Comprehension through Question Decomposition and Rescoring",
                        "abstract": "Multi-hop Reading Comprehension (RC) requires reasoning and aggregation across several paragraphs. We propose a system for multi-hop RC that decomposes a compositional question into simpler sub-questions that can be answered by off-the-shelf single-hop RC models. Since annotations for such decomposition are expensive, we recast sub-question generation as a span prediction problem and show that our method, trained using only 400 labeled examples, generates sub-questions that are as effective as human-authored sub-questions. We also introduce a new global rescoring approach that considers each decomposition (i.e. the sub-questions and their answers) to select the best final answer, greatly improving overall performance. Our experiments on HotpotQA show that this approach achieves the state-of-the-art results, while providing explainable evidence for its decision making in the form of sub-questions."
                    },
                    {
                        "title": "Probing Contextual Language Models for Common Ground with Visual Representations",
                        "abstract": "The success of large-scale contextual language models has attracted great interest in probing what is encoded in their representations. In this work, we consider a new question: to what extent contextual representations of concrete nouns are aligned with corresponding visual representations? We design a probing model that evaluates how effective are text-only representations in distinguishing between matching and non-matching visual representations. Our findings show that language representations alone provide a strong signal for retrieving image patches from the correct object categories. Moreover, they are effective in retrieving specific instances of image patches; textual context plays an important role in this process. Visually grounded language models slightly outperform text-only language models in instance retrieval, but greatly under-perform humans. We hope our analyses inspire future research in understanding and improving the visual capabilities of language models."
                    },
                    {
                        "title": "Bidirectional Attention Flow for Machine Comprehension",
                        "abstract": "Machine comprehension (MC), answering a query about a given context paragraph, requires modeling complex interactions between the context and the query. Recently, attention mechanisms have been successfully extended to MC. Typically these methods use attention to focus on a small portion of the context and summarize it with a fixed-size vector, couple attentions temporally, and/or often form a uni-directional attention. In this paper we introduce the Bi-Directional Attention Flow (BIDAF) network, a multi-stage hierarchical process that represents the context at different levels of granularity and uses bi-directional attention flow mechanism to obtain a query-aware context representation without early summarization. Our experimental evaluations show that our model achieves the state-of-the-art results in Stanford Question Answering Dataset (SQuAD) and CNN/DailyMail cloze test."
                    },
                    {
                        "title": "AmbigQA: Answering Ambiguous Open-domain Questions",
                        "abstract": "Ambiguity is inherent to open-domain question answering; especially when exploring new topics, it can be difficult to ask questions that have a single, unambiguous answer. In this paper, we introduce AmbigQA, a new open-domain question answering task which involves finding every plausible answer, and then rewriting the question for each one to resolve the ambiguity. To study this task, we construct AmbigNQ, a dataset covering 14,042 questions from NQ-open, an existing open-domain QA benchmark. We find that over half of the questions in NQ-open are ambiguous, with diverse sources of ambiguity such as event and entity references. We also present strong baseline models for AmbigQA which we show benefit from weakly supervised learning that incorporates NQ-open, strongly suggesting our new task and data will support significant future research effort. Our data and baselines are available at https://nlp.cs.washington.edu/ambigqa."
                    },
                    {
                        "title": "One Question Answering Model for Many Languages with Cross-lingual Dense Passage Retrieval",
                        "abstract": "We present Cross-lingual Open-Retrieval Answer Generation (CORA), the first unified many-to-many question answering (QA) model that can answer questions across many languages, even for ones without language-specific annotated data or knowledge sources. We introduce a new dense passage retrieval algorithm that is trained to retrieve documents across languages for a question. Combined with a multilingual autoregressive generation model, CORA answers directly in the target language without any translation or in-language retrieval modules as used in prior work. We propose an iterative training method that automatically extends annotated data available only in high-resource languages to low-resource ones. Our results show that CORA substantially outperforms the previous state of the art on multilingual open QA benchmarks across 26 languages, 9 of which are unseen during training. Our analyses show the significance of cross-lingual retrieval and generation in many languages, particularly under low-resource settings."
                    }
                ]
            },
            "a377d4b9-2a21-47e5-b761-2d4d2bbf98d8": {
                "pk": "a377d4b9-2a21-47e5-b761-2d4d2bbf98d8",
                "name": "Sham Kakade",
                "collaborators": [
                    "Elad Hazan",
                    "Nicol\u00f2 Cesa-Bianchi",
                    "David Belanger",
                    "Mengdi Wang",
                    "Lin F. Yang",
                    "Krishna Pillutla",
                    "Vincent Roulet",
                    "Zaid Harchaoui",
                    "Alex Strehl",
                    "John Langford"
                ],
                "domain": [
                    "Optimization",
                    "Bandit Learning",
                    "Natural Language Processing",
                    "Hidden Markov Models"
                ],
                "publications": [
                    {
                        "title": "(weak) Calibration is Computationally Hard",
                        "abstract": "We show that the existence of a computationally efficient calibration algorithm, with a low weak calibration rate, would imply the existence of an efficient algorithm for computing approximate Nash equilibria - thus implying the unlikely conclusion that every problem in PPAD is solvable in polynomial time."
                    },
                    {
                        "title": "Revisiting the Polyak step size",
                        "abstract": "This paper revisits the Polyak step size schedule for convex optimization problems, proving that a simple variant of it simultaneously attains near optimal convergence rates for the gradient descent algorithm, for all ranges of strong convexity, smoothness, and Lipschitz parameters, without a-priory knowledge of these parameters."
                    },
                    {
                        "title": "An Optimal Algorithm for Linear Bandits",
                        "abstract": "We provide the first algorithm for online bandit linear optimization whose regret after T rounds is of order sqrt{Td ln N} on any finite class X of N actions in d dimensions, and of order d*sqrt{T} (up to log factors) when X is infinite. These bounds are not improvable in general. The basic idea utilizes tools from convex geometry to construct what is essentially an optimal exploration basis. We also present an application to a model of linear bandits with expert advice. Interestingly, these results show that bandit linear optimization with expert advice in d dimensions is no more difficult (in terms of the achievable regret) than the online d-armed bandit problem with expert advice (where EXP4 is optimal)."
                    },
                    {
                        "title": "A Linear Dynamical System Model for Text",
                        "abstract": "Low dimensional representations of words allow accurate NLP models to be trained on limited annotated data. While most representations ignore words' local context, a natural way to induce context-dependent representations is to perform inference in a probabilistic latent-variable sequence model. Given the recent success of continuous vector space word representations, we provide such an inference procedure for continuous states, where words' representations are given by the posterior mean of a linear dynamical system. Here, efficient inference can be performed using Kalman filtering. Our learning algorithm is extremely scalable, operating on simple cooccurrence counts for both parameter initialization using the method of moments and subsequent iterations of EM. In our experiments, we employ our inferred word embeddings as features in standard tagging tasks, obtaining significant accuracy improvements. Finally, the Kalman filter updates can be seen as a linear recurrent neural network. We demonstrate that using the parameters of our model to initialize a non-linear recurrent neural network language model reduces its training time by a day and yields lower perplexity."
                    },
                    {
                        "title": "Variance Reduction Methods for Sublinear Reinforcement Learning",
                        "abstract": "There is a technical issue in the analysis that is not easily fixable. We, therefore, withdraw the submission. Sorry for the inconvenience."
                    },
                    {
                        "title": "Modified Gauss-Newton Algorithms under Noise",
                        "abstract": "Gauss-Newton methods and their stochastic version have been widely used in machine learning and signal processing. Their nonsmooth counterparts, modified Gauss-Newton or prox-linear algorithms, can lead to contrasting outcomes when compared to gradient descent in large-scale statistical settings. We explore the contrasting performance of these two classes of algorithms in theory on a stylized statistical example, and experimentally on learning problems including structured prediction. In theory, we delineate the regime where the quadratic convergence of the modified Gauss-Newton method is active under statistical noise. In the experiments, we underline the versatility of stochastic (sub)-gradient descent to minimize nonsmooth composite objectives."
                    },
                    {
                        "title": "Learning from Logged Implicit Exploration Data",
                        "abstract": "We provide a sound and consistent foundation for the use of \\emph{nonrandom} exploration data in \"contextual bandit\" or \"partially labeled\" settings where only the value of a chosen action is learned.   The primary challenge in a variety of settings is that the exploration policy, in which \"offline\" data is logged, is not explicitly known. Prior solutions here require either control of the actions during the learning process, recorded random exploration, or actions chosen obliviously in a repeated manner. The techniques reported here lift these restrictions, allowing the learning of a policy for choosing actions given features from historical data where no randomization occurred or was logged.   We empirically verify our solution on two reasonably sized sets of real-world data obtained from Yahoo!."
                    },
                    {
                        "title": "Domain Adaptation: Overfitting and Small Sample Statistics",
                        "abstract": "We study the prevalent problem when a test distribution differs from the training distribution. We consider a setting where our training set consists of a small number of sample domains, but where we have many samples in each domain. Our goal is to generalize to a new domain. For example, we may want to learn a similarity function using only certain classes of objects, but we desire that this similarity function be applicable to object classes not present in our training sample (e.g. we might seek to learn that \"dogs are similar to dogs\" even though images of dogs were absent from our training set). Our theoretical analysis shows that we can select many more features than domains while avoiding overfitting by utilizing data-dependent variance properties. We present a greedy feature selection algorithm based on using T-statistics. Our experiments validate this theory showing that our T-statistic based greedy feature selection is more robust at avoiding overfitting than the classical greedy procedure."
                    },
                    {
                        "title": "Learning Overcomplete HMMs",
                        "abstract": "We study the problem of learning overcomplete HMMs---those that have many hidden states but a small output alphabet. Despite having significant practical importance, such HMMs are poorly understood with no known positive or negative results for efficient learning. In this paper, we present several new results---both positive and negative---which help define the boundaries between the tractable and intractable settings. Specifically, we show positive results for a large subclass of HMMs whose transition matrices are sparse, well-conditioned, and have small probability mass on short cycles. On the other hand, we show that learning is impossible given only a polynomial number of samples for HMMs with a small output alphabet and whose transition matrices are random regular graphs with large degree. We also discuss these results in the context of learning HMMs which can capture long-term dependencies."
                    }
                ]
            },
            "f9a31f3e-f5f2-4e85-9d52-5bb0ce709715": {
                "pk": "f9a31f3e-f5f2-4e85-9d52-5bb0ce709715",
                "name": "Ali Farhadi",
                "collaborators": [
                    "Hannaneh Hajishirzi",
                    "Mohammad Rastegari",
                    "Joseph Redmon",
                    "Roozbeh Mottaghi",
                    "Hessam Bagherinezhad",
                    "Minjoon Seo",
                    "Junyuan Xie",
                    "Ross Girshick",
                    "Dieter Fox",
                    "Babak Saleh"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Object Detection",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "YOLOv3: An Incremental Improvement",
                        "abstract": "We present some updates to YOLO! We made a bunch of little design changes to make it better. We also trained this new network that's pretty swell. It's a little bigger than last time but more accurate. It's still fast though, don't worry. At 320x320 YOLOv3 runs in 22 ms at 28.2 mAP, as accurate as SSD but three times faster. When we look at the old .5 IOU mAP detection metric YOLOv3 is quite good. It achieves 57.9 mAP@50 in 51 ms on a Titan X, compared to 57.5 mAP@50 in 198 ms by RetinaNet, similar performance but 3.8x faster. As always, all the code is online at https://pjreddie.com/yolo/"
                    },
                    {
                        "title": "YOLO9000: Better, Faster, Stronger",
                        "abstract": "We introduce YOLO9000, a state-of-the-art, real-time object detection system that can detect over 9000 object categories. First we propose various improvements to the YOLO detection method, both novel and drawn from prior work. The improved model, YOLOv2, is state-of-the-art on standard detection tasks like PASCAL VOC and COCO. At 67 FPS, YOLOv2 gets 76.8 mAP on VOC 2007. At 40 FPS, YOLOv2 gets 78.6 mAP, outperforming state-of-the-art methods like Faster RCNN with ResNet and SSD while still running significantly faster. Finally we propose a method to jointly train on object detection and classification. Using this method we train YOLO9000 simultaneously on the COCO detection dataset and the ImageNet classification dataset. Our joint training allows YOLO9000 to predict detections for object classes that don't have labelled detection data. We validate our approach on the ImageNet detection task. YOLO9000 gets 19.7 mAP on the ImageNet detection validation set despite only having detection data for 44 of the 200 classes. On the 156 classes not in COCO, YOLO9000 gets 16.0 mAP. But YOLO can detect more than just 200 classes; it predicts detections for more than 9000 different object categories. And it still runs in real-time."
                    },
                    {
                        "title": "Unsupervised Deep Embedding for Clustering Analysis",
                        "abstract": "Clustering is central to many data-driven application domains and has been studied extensively in terms of distance functions and grouping algorithms. Relatively little work has focused on learning representations for clustering. In this paper, we propose Deep Embedded Clustering (DEC), a method that simultaneously learns feature representations and cluster assignments using deep neural networks. DEC learns a mapping from the data space to a lower-dimensional feature space in which it iteratively optimizes a clustering objective. Our experimental evaluations on image and text corpora show significant improvement over state-of-the-art methods."
                    },
                    {
                        "title": "Deep3D: Fully Automatic 2D-to-3D Video Conversion with Deep Convolutional Neural Networks",
                        "abstract": "As 3D movie viewing becomes mainstream and Virtual Reality (VR) market emerges, the demand for 3D contents is growing rapidly. Producing 3D videos, however, remains challenging. In this paper we propose to use deep neural networks for automatically converting 2D videos and images to stereoscopic 3D format. In contrast to previous automatic 2D-to-3D conversion algorithms, which have separate stages and need ground truth depth map as supervision, our approach is trained end-to-end directly on stereo pairs extracted from 3D movies. This novel training scheme makes it possible to exploit orders of magnitude more data and significantly increases performance. Indeed, Deep3D outperforms baselines in both quantitative and human subject evaluations."
                    },
                    {
                        "title": "What Should I Do Now? Marrying Reinforcement Learning and Symbolic Planning",
                        "abstract": "Long-term planning poses a major difficulty to many reinforcement learning algorithms. This problem becomes even more pronounced in dynamic visual environments. In this work we propose Hierarchical Planning and Reinforcement Learning (HIP-RL), a method for merging the benefits and capabilities of Symbolic Planning with the learning abilities of Deep Reinforcement Learning. We apply HIPRL to the complex visual tasks of interactive question answering and visual semantic planning and achieve state-of-the-art results on three challenging datasets all while taking fewer steps at test time and training in fewer iterations. Sample results can be found at youtu.be/0TtWJ_0mPfI"
                    },
                    {
                        "title": "Abnormal Object Recognition: A Comprehensive Study",
                        "abstract": "When describing images, humans tend not to talk about the obvious, but rather mention what they find interesting. We argue that abnormalities and deviations from typicalities are among the most important components that form what is worth mentioning. In this paper we introduce the abnormality detection as a recognition problem and show how to model typicalities and, consequently, meaningful deviations from prototypical properties of categories. Our model can recognize abnormalities and report the main reasons of any recognized abnormality. We introduce the abnormality detection dataset and show interesting results on how to reason about abnormalities."
                    },
                    {
                        "title": "SHARCS: Efficient Transformers through Routing with Dynamic Width Sub-networks",
                        "abstract": "We introduce SHARCS for adaptive inference that takes into account the hardness of input samples. SHARCS can train a router on any transformer network, enabling the model to direct different samples to sub-networks with varying widths. Our experiments demonstrate that: (1) SHARCS outperforms or complements existing per-sample adaptive inference methods across various classification tasks in terms of accuracy vs. FLOPs; (2) SHARCS generalizes across different architectures and can be even applied to compressed and efficient transformer encoders to further improve their efficiency; (3) SHARCS can provide a 2 times inference speed up at an insignificant drop in accuracy."
                    },
                    {
                        "title": "VISALOGY: Answering Visual Analogy Questions",
                        "abstract": "In this paper, we study the problem of answering visual analogy questions. These questions take the form of image A is to image B as image C is to what. Answering these questions entails discovering the mapping from image A to image B and then extending the mapping to image C and searching for the image D such that the relation from A to B holds for C to D. We pose this problem as learning an embedding that encourages pairs of analogous images with similar transformations to be close together using convolutional neural networks with a quadruple Siamese architecture. We introduce a dataset of visual analogy questions in natural images, and show first results of its kind on solving analogy questions on natural images."
                    },
                    {
                        "title": "Newtonian Image Understanding: Unfolding the Dynamics of Objects in Static Images",
                        "abstract": "In this paper, we study the challenging problem of predicting the dynamics of objects in static images. Given a query object in an image, our goal is to provide a physical understanding of the object in terms of the forces acting upon it and its long term motion as response to those forces. Direct and explicit estimation of the forces and the motion of objects from a single image is extremely challenging. We define intermediate physical abstractions called Newtonian scenarios and introduce Newtonian Neural Network ($N^3$) that learns to map a single image to a state in a Newtonian scenario. Our experimental evaluations show that our method can reliably predict dynamics of a query object from a single image. In addition, our approach can provide physical reasoning that supports the predicted dynamics in terms of velocity and force vectors. To spur research in this direction we compiled Visual Newtonian Dynamics (VIND) dataset that includes 6806 videos aligned with Newtonian scenarios represented using game engines, and 4516 still images with their ground truth dynamics."
                    },
                    {
                        "title": "Actions ~ Transformations",
                        "abstract": "What defines an action like \"kicking ball\"? We argue that the true meaning of an action lies in the change or transformation an action brings to the environment. In this paper, we propose a novel representation for actions by modeling an action as a transformation which changes the state of the environment before the action happens (precondition) to the state after the action (effect). Motivated by recent advancements of video representation using deep learning, we design a Siamese network which models the action as a transformation on a high-level feature space. We show that our model gives improvements on standard action recognition datasets including UCF101 and HMDB51. More importantly, our approach is able to generalize beyond learned action categories and shows significant performance improvement on cross-category generalization on our new ACT dataset."
                    },
                    {
                        "title": "Toward a Taxonomy and Computational Models of Abnormalities in Images",
                        "abstract": "The human visual system can spot an abnormal image, and reason about what makes it strange. This task has not received enough attention in computer vision. In this paper we study various types of atypicalities in images in a more comprehensive way than has been done before. We propose a new dataset of abnormal images showing a wide range of atypicalities. We design human subject experiments to discover a coarse taxonomy of the reasons for abnormality. Our experiments reveal three major categories of abnormality: object-centric, scene-centric, and contextual. Based on this taxonomy, we propose a comprehensive computational model that can predict all different types of abnormality in images and outperform prior arts in abnormality recognition."
                    },
                    {
                        "title": "Are Elephants Bigger than Butterflies? Reasoning about Sizes of Objects",
                        "abstract": "Human vision greatly benefits from the information about sizes of objects. The role of size in several visual reasoning tasks has been thoroughly explored in human perception and cognition. However, the impact of the information about sizes of objects is yet to be determined in AI. We postulate that this is mainly attributed to the lack of a comprehensive repository of size information. In this paper, we introduce a method to automatically infer object sizes, leveraging visual and textual information from web. By maximizing the joint likelihood of textual and visual observations, our method learns reliable relative size estimates, with no explicit human supervision. We introduce the relative size dataset and show that our method outperforms competitive textual and visual baselines in reasoning about size comparisons."
                    },
                    {
                        "title": "XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks",
                        "abstract": "We propose two efficient approximations to standard convolutional neural networks: Binary-Weight-Networks and XNOR-Networks. In Binary-Weight-Networks, the filters are approximated with binary values resulting in 32x memory saving. In XNOR-Networks, both the filters and the input to convolutional layers are binary. XNOR-Networks approximate convolutions using primarily binary operations. This results in 58x faster convolutional operations and 32x memory savings. XNOR-Nets offer the possibility of running state-of-the-art networks on CPUs (rather than GPUs) in real-time. Our binary networks are simple, accurate, efficient, and work on challenging visual tasks. We evaluate our approach on the ImageNet classification task. The classification accuracy with a Binary-Weight-Network version of AlexNet is only 2.9% less than the full-precision AlexNet (in top-1 measure). We compare our method with recent network binarization methods, BinaryConnect and BinaryNets, and outperform these methods by large margins on ImageNet, more than 16% in top-1 accuracy."
                    },
                    {
                        "title": "Query-Reduction Networks for Question Answering",
                        "abstract": "In this paper, we study the problem of question answering when reasoning over multiple facts is required. We propose Query-Reduction Network (QRN), a variant of Recurrent Neural Network (RNN) that effectively handles both short-term (local) and long-term (global) sequential dependencies to reason over multiple facts. QRN considers the context sentences as a sequence of state-changing triggers, and reduces the original query to a more informed query as it observes each trigger (context sentence) through time. Our experiments show that QRN produces the state-of-the-art results in bAbI QA and dialog tasks, and in a real goal-oriented dialog dataset. In addition, QRN formulation allows parallelization on RNN's time axis, saving an order of magnitude in time complexity for training and inference."
                    },
                    {
                        "title": "See the Glass Half Full: Reasoning about Liquid Containers, their Volume and Content",
                        "abstract": "Humans have rich understanding of liquid containers and their contents; for example, we can effortlessly pour water from a pitcher to a cup. Doing so requires estimating the volume of the cup, approximating the amount of water in the pitcher, and predicting the behavior of water when we tilt the pitcher. Very little attention in computer vision has been made to liquids and their containers. In this paper, we study liquid containers and their contents, and propose methods to estimate the volume of containers, approximate the amount of liquid in them, and perform comparative volume estimations all from a single RGB image. Furthermore, we show the results of the proposed model for predicting the behavior of liquids inside containers when one tilts the containers. We also introduce a new dataset of Containers Of liQuid contEnt (COQE) that contains more than 5,000 images of 10,000 liquid containers in context labelled with volume, amount of content, bounding box annotation, and corresponding similar 3D CAD models."
                    },
                    {
                        "title": "Neural Speed Reading via Skim-RNN",
                        "abstract": "Inspired by the principles of speed reading, we introduce Skim-RNN, a recurrent neural network (RNN) that dynamically decides to update only a small fraction of the hidden state for relatively unimportant input tokens. Skim-RNN gives computational advantage over an RNN that always updates the entire hidden state. Skim-RNN uses the same input and output interfaces as a standard RNN and can be easily used instead of RNNs in existing models. In our experiments, we show that Skim-RNN can achieve significantly reduced computational cost without losing accuracy compared to standard RNNs across five different natural language tasks. In addition, we demonstrate that the trade-off between accuracy and speed of Skim-RNN can be dynamically controlled during inference time in a stable manner. Our analysis also shows that Skim-RNN running on a single CPU offers lower latency compared to standard RNNs on GPUs."
                    },
                    {
                        "title": "Label Refinery: Improving ImageNet Classification through Label Progression",
                        "abstract": "Among the three main components (data, labels, and models) of any supervised learning system, data and models have been the main subjects of active research. However, studying labels and their properties has received very little attention. Current principles and paradigms of labeling impose several challenges to machine learning algorithms. Labels are often incomplete, ambiguous, and redundant. In this paper we study the effects of various properties of labels and introduce the Label Refinery: an iterative procedure that updates the ground truth labels after examining the entire dataset. We show significant gain using refined labels across a wide range of models. Using a Label Refinery improves the state-of-the-art top-1 accuracy of (1) AlexNet from 59.3 to 67.2, (2) MobileNet from 70.6 to 73.39, (3) MobileNet-0.25 from 50.6 to 55.59, (4) VGG19 from 72.7 to 75.46, and (5) Darknet19 from 72.9 to 74.47."
                    },
                    {
                        "title": "OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge",
                        "abstract": "Visual Question Answering (VQA) in its ideal form lets us study reasoning in the joint space of vision and language and serves as a proxy for the AI task of scene understanding. However, most VQA benchmarks to date are focused on questions such as simple counting, visual attributes, and object detection that do not require reasoning or knowledge beyond what is in the image. In this paper, we address the task of knowledge-based visual question answering and provide a benchmark, called OK-VQA, where the image content is not sufficient to answer the questions, encouraging methods that rely on external knowledge resources. Our new dataset includes more than 14,000 questions that require external knowledge to answer. We show that the performance of the state-of-the-art VQA models degrades drastically in this new setting. Our analysis shows that our knowledge-based VQA task is diverse, difficult, and large compared to previous knowledge-based VQA datasets. We hope that this dataset enables researchers to open up new avenues for research in this domain. See http://okvqa.allenai.org to download and browse the dataset."
                    },
                    {
                        "title": "Probing Contextual Language Models for Common Ground with Visual Representations",
                        "abstract": "The success of large-scale contextual language models has attracted great interest in probing what is encoded in their representations. In this work, we consider a new question: to what extent contextual representations of concrete nouns are aligned with corresponding visual representations? We design a probing model that evaluates how effective are text-only representations in distinguishing between matching and non-matching visual representations. Our findings show that language representations alone provide a strong signal for retrieving image patches from the correct object categories. Moreover, they are effective in retrieving specific instances of image patches; textual context plays an important role in this process. Visually grounded language models slightly outperform text-only language models in instance retrieval, but greatly under-perform humans. We hope our analyses inspire future research in understanding and improving the visual capabilities of language models."
                    },
                    {
                        "title": "Bidirectional Attention Flow for Machine Comprehension",
                        "abstract": "Machine comprehension (MC), answering a query about a given context paragraph, requires modeling complex interactions between the context and the query. Recently, attention mechanisms have been successfully extended to MC. Typically these methods use attention to focus on a small portion of the context and summarize it with a fixed-size vector, couple attentions temporally, and/or often form a uni-directional attention. In this paper we introduce the Bi-Directional Attention Flow (BIDAF) network, a multi-stage hierarchical process that represents the context at different levels of granularity and uses bi-directional attention flow mechanism to obtain a query-aware context representation without early summarization. Our experimental evaluations show that our model achieves the state-of-the-art results in Stanford Question Answering Dataset (SQuAD) and CNN/DailyMail cloze test."
                    }
                ]
            },
            "f56c2b29-34ab-41e5-b150-a224b4fddd1d": {
                "pk": "f56c2b29-34ab-41e5-b150-a224b4fddd1d",
                "name": "Prateek Jain",
                "collaborators": [
                    "Purushottam Kar",
                    "Srinadh Bhojanapalli",
                    "Sewoong Oh",
                    "Nagarajan Natarajan",
                    "Inderjit S. Dhillon",
                    "Prateek Varshney",
                    "Abhradeep Thakurta",
                    "Srinjoy Ganguly",
                    "Chirag Pabbaraju",
                    "Praneeth Netrapalli"
                ],
                "domain": [
                    "Differential Privacy",
                    "Machine Learning",
                    "Matrix Completion",
                    "Non-invasive Sensing"
                ],
                "publications": [
                    {
                        "title": "(Nearly) Optimal Private Linear Regression via Adaptive Clipping",
                        "abstract": "We study the problem of differentially private linear regression where each data point is sampled from a fixed sub-Gaussian style distribution. We propose and analyze a one-pass mini-batch stochastic gradient descent method (DP-AMBSSGD) where points in each iteration are sampled without replacement. Noise is added for DP but the noise standard deviation is estimated online. Compared to existing $(\\epsilon, \\delta)$-DP techniques which have sub-optimal error bounds, DP-AMBSSGD is able to provide nearly optimal error bounds in terms of key parameters like dimensionality $d$, number of points $N$, and the standard deviation $\\sigma$ of the noise in observations. For example, when the $d$-dimensional covariates are sampled i.i.d. from the normal distribution, then the excess error of DP-AMBSSGD due to privacy is $\\frac{\\sigma^2 d}{N}(1+\\frac{d}{\\epsilon^2 N})$, i.e., the error is meaningful when number of samples $N= \\Omega(d \\log d)$ which is the standard operative regime for linear regression. In contrast, error bounds for existing efficient methods in this setting are: $\\mathcal{O}\\big(\\frac{d^3}{\\epsilon^2 N^2}\\big)$, even for $\\sigma=0$. That is, for constant $\\epsilon$, the existing techniques require $N=\\Omega(d\\sqrt{d})$ to provide a non-trivial result."
                    },
                    {
                        "title": "Hybrid Quantum Generative Adversarial Networks for Molecular Simulation and Drug Discovery",
                        "abstract": "In molecular research, simulation \\& design of molecules are key areas with significant implications for drug development, material science, and other fields. Current classical computational power falls inadequate to simulate any more than small molecules, let alone protein chains on hundreds of peptide. Therefore these experiment are done physically in wet-lab, but it takes a lot of time \\& not possible to examine every molecule due to the size of the search area, tens of billions of dollars are spent every year in these research experiments. Molecule simulation \\& design has lately advanced significantly by machine learning models, A fresh perspective on the issue of chemical synthesis is provided by deep generative models for graph-structured data. By optimising differentiable models that produce molecular graphs directly, it is feasible to avoid costly search techniques in the discrete and huge space of chemical structures. But these models also suffer from computational limitations when dimensions become huge and consume huge amount of resources. Quantum Generative machine learning in recent years have shown some empirical results promising significant advantages over classical counterparts."
                    },
                    {
                        "title": "Learning Functions over Sets via Permutation Adversarial Networks",
                        "abstract": "In this paper, we consider the problem of learning functions over sets, i.e., functions that are invariant to permutations of input set items. Recent approaches of pooling individual element embeddings can necessitate extremely large embedding sizes for challenging functions. We address this challenge by allowing standard neural networks like LSTMs to succinctly capture the function over the set. However, to ensure invariance with respect to permutations of set elements, we propose a novel architecture called SPAN that simultaneously learns the function as well as adversarial or worst-case permutations for each input set. The learning problem reduces to a min-max optimization problem that is solved via a simple alternating block coordinate descent technique. We conduct extensive experiments on a variety of set-learning tasks and demonstrate that SPAN learns nearly permutation-invariant functions while still ensuring accuracy on test data. On a variety of tasks sampled from the domains of statistics, graph functions and linear algebra, we show that our method can significantly outperform state-of-the-art methods such as DeepSets and Janossy Pooling. Finally, we present a case study of how learning set-functions can help extract powerful features for recommendation systems, and show that such a method can be as much as 2% more accurate than carefully hand-tuned features on a real-world recommendation system."
                    },
                    {
                        "title": "Similarity-based Learning via Data Driven Embeddings",
                        "abstract": "We consider the problem of classification using similarity/distance functions over data. Specifically, we propose a framework for defining the goodness of a (dis)similarity function with respect to a given learning task and propose algorithms that have guaranteed generalization properties when working with such good functions. Our framework unifies and generalizes the frameworks proposed by [Balcan-Blum ICML 2006] and [Wang et al ICML 2007]. An attractive feature of our framework is its adaptability to data - we do not promote a fixed notion of goodness but rather let data dictate it. We show, by giving theoretical guarantees that the goodness criterion best suited to a problem can itself be learned which makes our approach applicable to a variety of domains and problems. We propose a landmarking-based approach to obtaining a classifier from such learned goodness criteria. We then provide a novel diversity based heuristic to perform task-driven selection of landmark points instead of random selection. We demonstrate the effectiveness of our goodness criteria learning method as well as the landmark selection heuristic on a variety of similarity-based learning datasets and benchmark UCI datasets on which our method consistently outperforms existing approaches by a significant margin."
                    },
                    {
                        "title": "Universal Matrix Completion",
                        "abstract": "The problem of low-rank matrix completion has recently generated a lot of interest leading to several results that offer exact solutions to the problem. However, in order to do so, these methods make assumptions that can be quite restrictive in practice. More specifically, the methods assume that: a) the observed indices are sampled uniformly at random, and b) for every new matrix, the observed indices are sampled afresh. In this work, we address these issues by providing a universal recovery guarantee for matrix completion that works for a variety of sampling schemes. In particular, we show that if the set of sampled indices come from the edges of a bipartite graph with large spectral gap (i.e. gap between the first and the second singular value), then the nuclear norm minimization based method exactly recovers all low-rank matrices that satisfy certain incoherence properties. Moreover, we also show that under certain stricter incoherence conditions, $O(nr^2)$ uniformly sampled entries are enough to recover any rank-$r$ $n\\times n$ matrix, in contrast to the $O(nr\\log n)$ sample complexity required by other matrix completion algorithms as well as existing analyses of the nuclear norm method."
                    },
                    {
                        "title": "Provable Tensor Factorization with Missing Data",
                        "abstract": "We study the problem of low-rank tensor factorization in the presence of missing data. We ask the following question: how many sampled entries do we need, to efficiently and exactly reconstruct a tensor with a low-rank orthogonal decomposition? We propose a novel alternating minimization based method which iteratively refines estimates of the singular vectors. We show that under certain standard assumptions, our method can recover a three-mode $n\\times n\\times n$ dimensional rank-$r$ tensor exactly from $O(n^{3/2} r^5 \\log^4 n)$ randomly sampled entries. In the process of proving this result, we solve two challenging sub-problems for tensors with missing data. First, in the process of analyzing the initialization step, we prove a generalization of a celebrated result by Szemer\\'edie et al. on the spectrum of random graphs. Next, we prove global convergence of alternating minimization with a good initialization. Simulations suggest that the dependence of the sample size on dimensionality $n$ is indeed tight."
                    },
                    {
                        "title": "Regret Bounds for Non-decomposable Metrics with Missing Labels",
                        "abstract": "We consider the problem of recommending relevant labels (items) for a given data point (user). In particular, we are interested in the practically important setting where the evaluation is with respect to non-decomposable (over labels) performance metrics like the $F_1$ measure, and the training data has missing labels. To this end, we propose a generic framework that given a performance metric $\\Psi$, can devise a regularized objective function and a threshold such that all the values in the predicted score vector above and only above the threshold are selected to be positive. We show that the regret or generalization error in the given metric $\\Psi$ is bounded ultimately by estimation error of certain underlying parameters. In particular, we derive regret bounds under three popular settings: a) collaborative filtering, b) multilabel classification, and c) PU (positive-unlabeled) learning. For each of the above problems, we can obtain precise non-asymptotic regret bound which is small even when a large fraction of labels is missing. Our empirical results on synthetic and benchmark datasets demonstrate that by explicitly modeling for missing labels and optimizing the desired performance metric, our algorithm indeed achieves significantly better performance (like $F_1$ score) when compared to methods that do not model missing label information carefully."
                    },
                    {
                        "title": "Supervised Learning with Similarity Functions",
                        "abstract": "We address the problem of general supervised learning when data can only be accessed through an (indefinite) similarity function between data points. Existing work on learning with indefinite kernels has concentrated solely on binary/multi-class classification problems. We propose a model that is generic enough to handle any supervised learning task and also subsumes the model previously proposed for classification. We give a \"goodness\" criterion for similarity functions w.r.t. a given supervised learning task and then adapt a well-known landmarking technique to provide efficient algorithms for supervised learning using \"good\" similarity functions. We demonstrate the effectiveness of our model on three important super-vised learning problems: a) real-valued regression, b) ordinal regression and c) ranking where we show that our method guarantees bounded generalization error. Furthermore, for the case of real-valued regression, we give a natural goodness definition that, when used in conjunction with a recent result in sparse vector recovery, guarantees a sparse predictor with bounded generalization error. Finally, we report results of our learning algorithms on regression and ordinal regression tasks using non-PSD similarity functions and demonstrate the effectiveness of our algorithms, especially that of the sparse landmark selection algorithm that achieves significantly higher accuracies than the baseline methods while offering reduced computational costs."
                    },
                    {
                        "title": "Learning Mixtures of Discrete Product Distributions using Spectral Decompositions",
                        "abstract": "We study the problem of learning a distribution from samples, when the underlying distribution is a mixture of product distributions over discrete domains. This problem is motivated by several practical applications such as crowd-sourcing, recommendation systems, and learning Boolean functions. The existing solutions either heavily rely on the fact that the number of components in the mixtures is finite or have sample/time complexity that is exponential in the number of components. In this paper, we introduce a polynomial time/sample complexity method for learning a mixture of $r$ discrete product distributions over $\\{1, 2, \\dots, \\ell\\}^n$, for general $\\ell$ and $r$. We show that our approach is statistically consistent and further provide finite sample guarantees.   We use techniques from the recent work on tensor decompositions for higher-order moment matching. A crucial step in these moment matching methods is to construct a certain matrix and a certain tensor with low-rank spectral decompositions. These tensors are typically estimated directly from the samples. The main challenge in learning mixtures of discrete product distributions is that these low-rank tensors cannot be obtained directly from the sample moments. Instead, we reduce the tensor estimation problem to: $a$) estimating a low-rank matrix using only off-diagonal block elements; and $b$) estimating a tensor using a small number of linear measurements. Leveraging on recent developments in matrix completion, we give an alternating minimization based method to estimate the low-rank matrix, and formulate the tensor completion problem as a least-squares problem."
                    },
                    {
                        "title": "Fast Exact Matrix Completion with Finite Samples",
                        "abstract": "Matrix completion is the problem of recovering a low rank matrix by observing a small fraction of its entries. A series of recent works [KOM12,JNS13,HW14] have proposed fast non-convex optimization based iterative algorithms to solve this problem. However, the sample complexity in all these results is sub-optimal in its dependence on the rank, condition number and the desired accuracy.   In this paper, we present a fast iterative algorithm that solves the matrix completion problem by observing $O(nr^5 \\log^3 n)$ entries, which is independent of the condition number and the desired accuracy. The run time of our algorithm is $O(nr^7\\log^3 n\\log 1/\\epsilon)$ which is near linear in the dimension of the matrix. To the best of our knowledge, this is the first near linear time algorithm for exact matrix completion with finite sample complexity (i.e. independent of $\\epsilon$).   Our algorithm is based on a well known projected gradient descent method, where the projection is onto the (non-convex) set of low rank matrices. There are two key ideas in our result: 1) our argument is based on a $\\ell_{\\infty}$ norm potential function (as opposed to the spectral norm) and provides a novel way to obtain perturbation bounds for it. 2) we prove and use a natural extension of the Davis-Kahan theorem to obtain perturbation bounds on the best low rank approximation of matrices with good eigen-gap. Both of these ideas may be of independent interest."
                    },
                    {
                        "title": "Online Low Rank Matrix Completion",
                        "abstract": "We study the problem of {\\em online} low-rank matrix completion with $\\mathsf{M}$ users, $\\mathsf{N}$ items and $\\mathsf{T}$ rounds. In each round, the algorithm recommends one item per user, for which it gets a (noisy) reward sampled from a low-rank user-item preference matrix. The goal is to design a method with sub-linear regret (in $\\mathsf{T}$) and nearly optimal dependence on $\\mathsf{M}$ and $\\mathsf{N}$. The problem can be easily mapped to the standard multi-armed bandit problem where each item is an {\\em independent} arm, but that leads to poor regret as the correlation between arms and users is not exploited. On the other hand, exploiting the low-rank structure of reward matrix is challenging due to non-convexity of the low-rank manifold. We first demonstrate that the low-rank structure can be exploited using a simple explore-then-commit (ETC) approach that ensures a regret of $O(\\mathsf{polylog} (\\mathsf{M}+\\mathsf{N}) \\mathsf{T}^{2/3})$. That is, roughly only $\\mathsf{polylog} (\\mathsf{M}+\\mathsf{N})$ item recommendations are required per user to get a non-trivial solution. We then improve our result for the rank-$1$ setting which in itself is quite challenging and encapsulates some of the key issues. Here, we propose \\textsc{OCTAL} (Online Collaborative filTering using iterAtive user cLustering) that guarantees nearly optimal regret of $O(\\mathsf{polylog} (\\mathsf{M}+\\mathsf{N}) \\mathsf{T}^{1/2})$. OCTAL is based on a novel technique of clustering users that allows iterative elimination of items and leads to a nearly optimal minimax rate."
                    },
                    {
                        "title": "Non-convex Optimization for Machine Learning",
                        "abstract": "A vast majority of machine learning algorithms train their models and perform inference by solving optimization problems. In order to capture the learning and prediction problems accurately, structural constraints such as sparsity or low rank are frequently imposed or else the objective itself is designed to be a non-convex function. This is especially true of algorithms that operate in high-dimensional spaces or that train non-linear models such as tensor models and deep networks.   The freedom to express the learning problem as a non-convex optimization problem gives immense modeling power to the algorithm designer, but often such problems are NP-hard to solve. A popular workaround to this has been to relax non-convex problems to convex ones and use traditional methods to solve the (convex) relaxed optimization problems. However this approach may be lossy and nevertheless presents significant challenges for large scale optimization.   On the other hand, direct approaches to non-convex optimization have met with resounding success in several domains and remain the methods of choice for the practitioner, as they frequently outperform relaxation-based techniques - popular heuristics include projected gradient descent and alternating minimization. However, these are often poorly understood in terms of their convergence and other properties.   This monograph presents a selection of recent advances that bridge a long-standing gap in our understanding of these heuristics. The monograph will lead the reader through several widely used non-convex optimization techniques, as well as applications thereof. The goal of this monograph is to both, introduce the rich literature in this area, as well as equip the reader with the tools and techniques needed to analyze these simple procedures for non-convex problems."
                    },
                    {
                        "title": "Quantum classical hybrid neural networks for continuous variable prediction",
                        "abstract": "Within this decade, quantum computers are predicted to outperform conventional computers in terms of processing power and have a disruptive effect on a variety of business sectors. It is predicted that the financial sector would be one of the first to benefit from quantum computing both in the short and long terms. In this research work we use Hybrid Quantum Neural networks to present a quantum machine learning approach for Continuous variable prediction."
                    },
                    {
                        "title": "Learning Accurate Decision Trees with Bandit Feedback via Quantized Gradient Descent",
                        "abstract": "Decision trees provide a rich family of highly non-linear but efficient models, due to which they continue to be the go-to family of predictive models by practitioners across domains. But learning trees is challenging due to their discrete decision boundaries. The state-of-the-art (SOTA) techniques resort to (a) learning \\textit{soft} trees thereby losing logarithmic inference time; or (b) using methods tailored to specific supervised learning settings, requiring access to labeled examples and loss function. In this work, by leveraging techniques like overparameterization and straight-through estimators, we propose a unified method that enables accurate end-to-end gradient based tree training and can be deployed in a variety of settings like offline supervised learning and online learning with bandit feedback. Using extensive validation on standard benchmarks, we demonstrate that our method provides best of both worlds, i.e., it is competitive to, and in some cases more accurate than methods designed \\textit{specifically} for the supervised settings; and in bandit settings, where most existing tree learning techniques are not applicable, our models are still accurate and significantly outperform the applicable SOTA methods."
                    },
                    {
                        "title": "Guaranteed Rank Minimization via Singular Value Projection",
                        "abstract": "Minimizing the rank of a matrix subject to affine constraints is a fundamental problem with many important applications in machine learning and statistics. In this paper we propose a simple and fast algorithm SVP (Singular Value Projection) for rank minimization with affine constraints (ARMP) and show that SVP recovers the minimum rank solution for affine constraints that satisfy the \"restricted isometry property\" and show robustness of our method to noise. Our results improve upon a recent breakthrough by Recht, Fazel and Parillo (RFP07) and Lee and Bresler (LB09) in three significant ways:   1) our method (SVP) is significantly simpler to analyze and easier to implement,   2) we give recovery guarantees under strictly weaker isometry assumptions   3) we give geometric convergence guarantees for SVP even in presense of noise and, as demonstrated empirically, SVP is significantly faster on real-world and synthetic problems.   In addition, we address the practically important problem of low-rank matrix completion (MCP), which can be seen as a special case of ARMP. We empirically demonstrate that our algorithm recovers low-rank incoherent matrices from an almost optimal number of uniformly sampled entries. We make partial progress towards proving exact recovery and provide some intuition for the strong performance of SVP applied to matrix completion by showing a more restricted isometry property. Our algorithm outperforms existing methods, such as those of \\cite{RFP07,CR08,CT09,CCS08,KOM09,LB09}, for ARMP and the matrix-completion problem by an order of magnitude and is also significantly more robust to noise."
                    },
                    {
                        "title": "A Learning Framework for Self-Tuning Histograms",
                        "abstract": "In this paper, we consider the problem of estimating self-tuning histograms using query workloads. To this end, we propose a general learning theoretic formulation. Specifically, we use query feedback from a workload as training data to estimate a histogram with a small memory footprint that minimizes the expected error on future queries. Our formulation provides a framework in which different approaches can be studied and developed. We first study the simple class of equi-width histograms and present a learning algorithm, EquiHist, that is competitive in many settings. We also provide formal guarantees for equi-width histograms that highlight scenarios in which equi-width histograms can be expected to succeed or fail. We then go beyond equi-width histograms and present a novel learning algorithm, SpHist, for estimating general histograms. Here we use Haar wavelets to reduce the problem of learning histograms to that of learning a sparse vector. Both algorithms have multiple advantages over existing methods: 1) simple and scalable extensions to multi-dimensional data, 2) scalability with number of histogram buckets and size of query feedback, 3) natural extensions to incorporate new feedback and handle database updates. We demonstrate these advantages over the current state-of-the-art, ISOMER, through detailed experiments on real and synthetic data. In particular, we show that SpHist obtains up to 50% less error than ISOMER on real-world multi-dimensional datasets."
                    },
                    {
                        "title": "Provable Inductive Matrix Completion",
                        "abstract": "Consider a movie recommendation system where apart from the ratings information, side information such as user's age or movie's genre is also available. Unlike standard matrix completion, in this setting one should be able to predict inductively on new users/movies. In this paper, we study the problem of inductive matrix completion in the exact recovery setting. That is, we assume that the ratings matrix is generated by applying feature vectors to a low-rank matrix and the goal is to recover back the underlying matrix. Furthermore, we generalize the problem to that of low-rank matrix estimation using rank-1 measurements. We study this generic problem and provide conditions that the set of measurements should satisfy so that the alternating minimization method (which otherwise is a non-convex method with no convergence guarantees) is able to recover back the {\\em exact} underlying low-rank matrix.   In addition to inductive matrix completion, we show that two other low-rank estimation problems can be studied in our framework: a) general low-rank matrix sensing using rank-1 measurements, and b) multi-label regression with missing labels. For both the problems, we provide novel and interesting bounds on the number of measurements required by alternating minimization to provably converges to the {\\em exact} low-rank matrix. In particular, our analysis for the general low rank matrix sensing problem significantly improves the required storage and computational cost than that required by the RIP-based matrix sensing methods \\cite{RechtFP2007}. Finally, we provide empirical validation of our approach and demonstrate that alternating minimization is able to recover the true matrix for the above mentioned problems using a small number of measurements."
                    },
                    {
                        "title": "Treeformer: Dense Gradient Trees for Efficient Attention Computation",
                        "abstract": "Standard inference and training with transformer based architectures scale quadratically with input sequence length. This is prohibitively large for a variety of applications especially in web-page translation, query-answering etc. Consequently, several approaches have been developed recently to speedup attention computation by enforcing different attention structures such as sparsity, low-rank, approximating attention using kernels. In this work, we view attention computation as that of nearest neighbor retrieval, and use decision tree based hierarchical navigation to reduce the retrieval cost per query token from linear in sequence length to nearly logarithmic. Based on such hierarchical navigation, we design Treeformer which can use one of two efficient attention layers -- TF-Attention and TC-Attention. TF-Attention computes the attention in a fine-grained style, while TC-Attention is a coarse attention layer which also ensures that the gradients are \"dense\". To optimize such challenging discrete layers, we propose a two-level bootstrapped training method. Using extensive experiments on standard NLP benchmarks, especially for long-sequences, we demonstrate that our Treeformer architecture can be almost as accurate as baseline Transformer while using 30x lesser FLOPs in the attention layer. Compared to Linformer, the accuracy can be as much as 12% higher while using similar FLOPs in the attention layer."
                    },
                    {
                        "title": "iGLU 4.0: A continuous glucose monitoring and balancing paradigm with physiological parameters",
                        "abstract": "The conventional method of glucose measurement such as pricking blood from the body is prevalent which brings pain and trauma. Invasive methods of measurement sometimes raise the risk of blood infection to the patient. Sometimes, some of the physiological parameters such as body temperature and systolic blood pressure (SBP) are responsible for blood glucose level fluctuations. Moreover, diabetes for a long duration usually becomes a critical issue. In such situation, patients need to consult diabetologist frequently, which is not possible in normal life. Therefore, it is required to develop non-invasive glucose balancing paradigm, which measures blood glucose without pricking blood along with physiological parameters measurement and decision model. The proposed paradigm helps to doctor, who is even available at remote location. There will not be any need to consult frequently. In the way of optimized non-invasive system design, an NIRS technique with specific wavelengths along with physiological parameters is taken to predict the precise glucose value. The all parameters (glucose, Blood pressure and body temperature), food intake and insulin levels are parts of decision model, which would help to the doctor to take decision related to the further medicine doses and diet plan. The patients would have suggestions according to maintain their blood glucose level. The proposed system demonstrated an accurate model with MARD and AvgE 12.50% and 12.10% respectively using DNN model. Coefficient of determination R2 has been found 0.97."
                    },
                    {
                        "title": "On Iterative Hard Thresholding Methods for High-dimensional M-Estimation",
                        "abstract": "The use of M-estimators in generalized linear regression models in high dimensional settings requires risk minimization with hard $L_0$ constraints. Of the known methods, the class of projected gradient descent (also known as iterative hard thresholding (IHT)) methods is known to offer the fastest and most scalable solutions. However, the current state-of-the-art is only able to analyze these methods in extremely restrictive settings which do not hold in high dimensional statistical models. In this work we bridge this gap by providing the first analysis for IHT-style methods in the high dimensional statistical setting. Our bounds are tight and match known minimax lower bounds. Our results rely on a general analysis framework that enables us to analyze several popular hard thresholding style algorithms (such as HTP, CoSaMP, SP) in the high dimensional regression setting. We also extend our analysis to a large family of \"fully corrective methods\" that includes two-stage and partial hard-thresholding algorithms. We show that our results hold for the problem of sparse regression, as well as low-rank matrix recovery."
                    }
                ]
            }
        }
    },
    "2401.15866": {
        "paper_data": {
            "title": "Stochastic Amortization: A Unified Approach to Accelerate Feature and Data Attribution",
            "url": "http://arxiv.org/abs/2401.15866v1",
            "arxiv_id": "2401.15866",
            "authors": [
                "Ian Covert",
                "Chanwoo Kim",
                "Su-In Lee",
                "James Zou",
                "Tatsunori Hashimoto"
            ],
            "abstract": "Many tasks in explainable machine learning, such as data valuation and feature attribution, perform expensive computation for each data point and can be intractable for large datasets. These methods require efficient approximations, and learning a network that directly predicts the desired output, which is commonly known as amortization, is a promising solution. However, training such models with exact labels is often intractable; we therefore explore training with noisy labels and find that this is inexpensive and surprisingly effective. Through theoretical analysis of the label noise and experiments with various models and datasets, we show that this approach significantly accelerates several feature attribution and data valuation methods, often yielding an order of magnitude speedup over existing approaches.",
            "introduction": " Introduction to online convex optimization. Foundations and Trends \u00aein Optimization , 2(3-4):157\u2013 325, 2016. He, H., Mineiro, P., and Karampatziakis, N. Active infor- mation acquisition. arXiv preprint arXiv:1602.02181 , 2016a. He, K., Zhang, X., Ren, S., and Sun, J. Deep residual learn- ing for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pp. 770\u2013778, 2016b. He, W., Mao, X., Ma, C., Huang, Y ., Hern `andez-Lobato, J. M., and Chen, T. BSODA: a bipartite scalable frame- work for online disease diagnosis. In Proceedings of the ACM Web Conference 2022 , pp. 2511\u20132521, 2022. Heess, N., Wayne, G., Silver, D., Lillicrap, T., Erez, T., and Tassa, Y . Learning continuous control policies by stochastic value gradients. Advances in Neural Informa- tion Processing Systems , 28, 2015. Howard, J. and Gugger, S. FastAI: A layered API for deep learning. Information , 11(2):108, 2020. Ilyas, A., Park, S. M., Engstrom, L., Leclerc, G., and Madry, A. Datamodels: Predicting predictions from training data. arXiv preprint arXiv:2202.00622 , 2022. Jain, S., Salman, H., Wong, E., Zhang, P., Vineet, V ., Vem- prala, S., and Madry, A. Missingness bias in model debugging. In International Conference on Learning Representations , 2021. Jethani, N., Sudarshan, M., Aphinyanaphongs, Y ., and Ran- ganath, R. Have we learned to explain?: How inter- pretability Background This section introduces the idea of amortization in ML, which we discuss in a general setting before considering any specific XML tasks. 2.1. Amortization Consider a scenario where we repeat similar computation for a large number of data points. We represent this general setup with a context variable b\u2208 B and a per-context output denoted by a(b)\u2208 A. For example, these can be an image and its data valuation score, or an image and its feature attri- butions. Amortization can be used with arbitrary domains, but we assume Euclidean spaces, or A \u2286RmandB \u2286Rd, because many XML tasks involve real-valued outputs. The computation performed for each context b\u2208 B can be arbitrary as well. In some cases a(b)is an intractable expectation, in which case we typically use a Monte Carlo estimator ( \u02c7Strumbelj & Kononenko, 2010; Ghorbani & Zou, 2019). In other situations it involves solving an optimization problem (Shu, 2017; Amos, 2022), in which case we can define a(b)via a parametric objective h:A \u00d7 B 7\u2192 R: a(b)\u2261arg min a\u2032\u2208Ah(a\u2032;b). (1) We do not require a specific formulation for a(b)in this work, but we will see that both the expectation and opti- mization setups provide useful perspectives on our main proposal of training with noisy labels (Section 3). In situations with repeated computation, our two options are generally to perform the computation separately for each b\u2208 B, or to amortize the computation with a learned model. We typically implement this with a deep neural network a(b;\u03b8), and our goal is to train it such that a(b;\u03b8)\u2248a(b). In training these models, the main challenge occurs when the per-instance computation is costly: specifically, it is not obvious how to train a(b;\u03b8)without a dataset of ground truth solutions a(b), which is often too slow to generate. We address this challenge in Section 3, where we prove that amortization tolerates training with noisy labels. 2.2. Related work The general view of amortized computation captures many tasks arising in physics, engineering, control and ML (Amos, 2022). For example, amortization is prominent in variational inference (Kingma & Welling, 2013; Rezende et al., 2014), meta learning (Ha et al., 2016; Finn et al., 2017; Rajeswaran et al., 2019)",
            "references": [
                {
                    "title": "Fast Shapley Value Estimation: A Unified Approach",
                    "abstract": "Shapley values have emerged as a widely accepted and trustworthy tool, grounded in theoretical axioms, for addressing challenges posed by black-box models like deep neural networks. However, computing Shapley values encounters exponential complexity as the number of features increases. Various approaches, including ApproSemivalue, KernelSHAP, and FastSHAP, have been explored to expedite the computation. In our analysis of existing approaches, we observe that stochastic estimators can be unified as a linear transformation of randomly summed values from feature subsets. Based on this, we investigate the possibility of designing simple amortized estimators and propose a straightforward and efficient one, SimSHAP, by eliminating redundant techniques. Extensive experiments conducted on tabular and image datasets validate the effectiveness of our SimSHAP, which significantly accelerates the computation of accurate Shapley values."
                },
                {
                    "title": "DataInf: Efficiently Estimating Data Influence in LoRA-tuned LLMs and Diffusion Models",
                    "abstract": "Quantifying the impact of training data points is crucial for understanding the outputs of machine learning models and for improving the transparency of the AI pipeline. The influence function is a principled and popular data attribution method, but its computational cost often makes it challenging to use. This issue becomes more pronounced in the setting of large language models and text-to-image models. In this work, we propose DataInf, an efficient influence approximation method that is practical for large-scale generative AI models. Leveraging an easy-to-compute closed-form expression, DataInf outperforms existing influence computation algorithms in terms of computational and memory efficiency. Our theoretical analysis shows that DataInf is particularly well-suited for parameter-efficient fine-tuning techniques such as LoRA. Through systematic empirical evaluations, we show that DataInf accurately approximates influence scores and is orders of magnitude faster than existing methods. In applications to RoBERTa-large, Llama-2-13B-chat, and stable-diffusion-v1.5 models, DataInf effectively identifies the most influential fine-tuning examples better than other approximate influence scores. Moreover, it can help to identify which data points are mislabeled."
                },
                {
                    "title": "Studying Large Language Model Generalization with Influence Functions",
                    "abstract": "When trying to gain better visibility into a machine learning model in order to understand and mitigate the associated risks, a potentially valuable source of evidence is: which training examples most contribute to a given behavior? Influence functions aim to answer a counterfactual: how would the model's parameters (and hence its outputs) change if a given sequence were added to the training set? While influence functions have produced insights for small models, they are difficult to scale to large language models (LLMs) due to the difficulty of computing an inverse-Hessian-vector product (IHVP). We use the Eigenvalue-corrected Kronecker-Factored Approximate Curvature (EK-FAC) approximation to scale influence functions up to LLMs with up to 52 billion parameters. In our experiments, EK-FAC achieves similar accuracy to traditional influence function estimators despite the IHVP computation being orders of magnitude faster. We investigate two algorithmic techniques to reduce the cost of computing gradients of candidate training sequences: TF-IDF filtering and query batching. We use influence functions to investigate the generalization patterns of LLMs, including the sparsity of the influence patterns, increasing abstraction with scale, math and programming abilities, cross-lingual generalization, and role-playing behavior. Despite many apparently sophisticated forms of generalization, we identify a surprising limitation: influences decay to near-zero when the order of key phrases is flipped. Overall, influence functions give us a powerful new tool for studying the generalization properties of LLMs."
                },
                {
                    "title": "OpenDataVal: a Unified Benchmark for Data Valuation",
                    "abstract": "Assessing the quality and impact of individual data points is critical for improving model performance and mitigating undesirable biases within the training dataset. Several data valuation algorithms have been proposed to quantify data quality, however, there lacks a systemic and standardized benchmarking system for data valuation. In this paper, we introduce OpenDataVal, an easy-to-use and unified benchmark framework that empowers researchers and practitioners to apply and compare various data valuation algorithms. OpenDataVal provides an integrated environment that includes (i) a diverse collection of image, natural language, and tabular datasets, (ii) implementations of eleven different state-of-the-art data valuation algorithms, and (iii) a prediction model API that can import any models in scikit-learn. Furthermore, we propose four downstream machine learning tasks for evaluating the quality of data values. We perform benchmarking analysis using OpenDataVal, quantifying and comparing the efficacy of state-of-the-art data valuation approaches. We find that no single algorithm performs uniformly best across all tasks, and an appropriate algorithm should be employed for a user's downstream task. OpenDataVal is publicly available at https://opendataval.github.io with comprehensive documentation. Furthermore, we provide a leaderboard where researchers can evaluate the effectiveness of their own data valuation algorithms."
                },
                {
                    "title": "On the Robustness of Removal-Based Feature Attributions",
                    "abstract": "To explain predictions made by complex machine learning models, many feature attribution methods have been developed that assign importance scores to input features. Some recent work challenges the robustness of these methods by showing that they are sensitive to input and model perturbations, while other work addresses this issue by proposing robust attribution methods. However, previous work on attribution robustness has focused primarily on gradient-based feature attributions, whereas the robustness of removal-based attribution methods is not currently well understood. To bridge this gap, we theoretically characterize the robustness properties of removal-based feature attributions. Specifically, we provide a unified analysis of such methods and derive upper bounds for the difference between intact and perturbed attributions, under settings of both input and model perturbations. Our empirical results on synthetic and real-world data validate our theoretical results and demonstrate their practical implications, including the ability to increase attribution robustness by improving the model's Lipschitz regularity."
                },
                {
                    "title": "LAVA: Data Valuation without Pre-Specified Learning Algorithms",
                    "abstract": "Traditionally, data valuation (DV) is posed as a problem of equitably splitting the validation performance of a learning algorithm among the training data. As a result, the calculated data values depend on many design choices of the underlying learning algorithm. However, this dependence is undesirable for many DV use cases, such as setting priorities over different data sources in a data acquisition process and informing pricing mechanisms in a data marketplace. In these scenarios, data needs to be valued before the actual analysis and the choice of the learning algorithm is still undetermined then. Another side-effect of the dependence is that to assess the value of individual points, one needs to re-run the learning algorithm with and without a point, which incurs a large computation burden. This work leapfrogs over the current limits of data valuation methods by introducing a new framework that can value training data in a way that is oblivious to the downstream learning algorithm. Our main results are as follows. (1) We develop a proxy for the validation performance associated with a training set based on a non-conventional class-wise Wasserstein distance between training and validation sets. We show that the distance characterizes the upper bound of the validation performance for any given model under certain Lipschitz conditions. (2) We develop a novel method to value individual data based on the sensitivity analysis of the class-wise Wasserstein distance. Importantly, these values can be directly obtained for free from the output of off-the-shelf optimization solvers when computing the distance. (3) We evaluate our new data valuation framework over various use cases related to detecting low-quality data and show that, surprisingly, the learning-agnostic feature of our framework enables a significant improvement over SOTA performance while being orders of magnitude faster."
                },
                {
                    "title": "Data-OOB: Out-of-bag Estimate as a Simple and Efficient Data Value",
                    "abstract": "Data valuation is a powerful framework for providing statistical insights into which data are beneficial or detrimental to model training. Many Shapley-based data valuation methods have shown promising results in various downstream tasks, however, they are well known to be computationally challenging as it requires training a large number of models. As a result, it has been recognized as infeasible to apply to large datasets. To address this issue, we propose Data-OOB, a new data valuation method for a bagging model that utilizes the out-of-bag estimate. The proposed method is computationally efficient and can scale to millions of data by reusing trained weak learners. Specifically, Data-OOB takes less than 2.25 hours on a single CPU processor when there are $10^6$ samples to evaluate and the input dimension is 100. Furthermore, Data-OOB has solid theoretical interpretations in that it identifies the same important data point as the infinitesimal jackknife influence function when two different points are compared. We conduct comprehensive experiments using 12 classification datasets, each with thousands of sample sizes. We demonstrate that the proposed method significantly outperforms existing state-of-the-art data valuation methods in identifying mislabeled data and finding a set of helpful (or harmful) data points, highlighting the potential for applying data values in real-world applications."
                },
                {
                    "title": "CoRTX: Contrastive Framework for Real-time Explanation",
                    "abstract": "Recent advancements in explainable machine learning provide effective and faithful solutions for interpreting model behaviors. However, many explanation methods encounter efficiency issues, which largely limit their deployments in practical scenarios. Real-time explainer (RTX) frameworks have thus been proposed to accelerate the model explanation process by learning a one-feed-forward explainer. Existing RTX frameworks typically build the explainer under the supervised learning paradigm, which requires large amounts of explanation labels as the ground truth. Considering that accurate explanation labels are usually hard to obtain due to constrained computational resources and limited human efforts, effective explainer training is still challenging in practice. In this work, we propose a COntrastive Real-Time eXplanation (CoRTX) framework to learn the explanation-oriented representation and relieve the intensive dependence of explainer training on explanation labels. Specifically, we design a synthetic strategy to select positive and negative instances for the learning of explanation. Theoretical analysis show that our selection strategy can benefit the contrastive learning process on explanation tasks. Experimental results on three real-world datasets further demonstrate the efficiency and efficacy of our proposed CoRTX framework."
                },
                {
                    "title": "SHAP-IQ: Unified Approximation of any-order Shapley Interactions",
                    "abstract": "Predominately in explainable artificial intelligence (XAI) research, the Shapley value (SV) is applied to determine feature attributions for any black box model. Shapley interaction indices extend the SV to define any-order feature interactions. Defining a unique Shapley interaction index is an open research question and, so far, three definitions have been proposed, which differ by their choice of axioms. Moreover, each definition requires a specific approximation technique. Here, we propose SHAPley Interaction Quantification (SHAP-IQ), an efficient sampling-based approximator to compute Shapley interactions for arbitrary cardinal interaction indices (CII), i.e. interaction indices that satisfy the linearity, symmetry and dummy axiom. SHAP-IQ is based on a novel representation and, in contrast to existing methods, we provide theoretical guarantees for its approximation quality, as well as estimates for the variance of the point estimates. For the special case of SV, our approach reveals a novel representation of the SV and corresponds to Unbiased KernelSHAP with a greatly simplified calculation. We illustrate the computational efficiency and effectiveness by explaining language, image classification and high-dimensional synthetic models."
                },
                {
                    "title": "Efficient XAI Techniques: A Taxonomic Survey",
                    "abstract": "Recently, there has been a growing demand for the deployment of Explainable Artificial Intelligence (XAI) algorithms in real-world applications. However, traditional XAI methods typically suffer from a high computational complexity problem, which discourages the deployment of real-time systems to meet the time-demanding requirements of real-world scenarios. Although many approaches have been proposed to improve the efficiency of XAI methods, a comprehensive understanding of the achievements and challenges is still needed. To this end, in this paper we provide a review of efficient XAI. Specifically, we categorize existing techniques of XAI acceleration into efficient non-amortized and efficient amortized methods. The efficient non-amortized methods focus on data-centric or model-centric acceleration upon each individual instance. In contrast, amortized methods focus on learning a unified distribution of model explanations, following the predictive, generative, or reinforcement frameworks, to rapidly derive multiple model explanations. We also analyze the limitations of an efficient XAI pipeline from the perspectives of the training phase, the deployment phase, and the use scenarios. Finally, we summarize the challenges of deploying XAI acceleration methods to real-world scenarios, overcoming the trade-off between faithfulness and efficiency, and the selection of different acceleration methods."
                },
                {
                    "title": "Variational Information Pursuit for Interpretable Predictions",
                    "abstract": "There is a growing interest in the machine learning community in developing predictive algorithms that are\"interpretable by design\". Towards this end, recent work proposes to make interpretable decisions by sequentially asking interpretable queries about data until a prediction can be made with high confidence based on the answers obtained (the history). To promote short query-answer chains, a greedy procedure called Information Pursuit (IP) is used, which adaptively chooses queries in order of information gain. Generative models are employed to learn the distribution of query-answers and labels, which is in turn used to estimate the most informative query. However, learning and inference with a full generative model of the data is often intractable for complex tasks. In this work, we propose Variational Information Pursuit (V-IP), a variational characterization of IP which bypasses the need for learning generative models. V-IP is based on finding a query selection strategy and a classifier that minimizes the expected cross-entropy between true and predicted labels. We then demonstrate that the IP strategy is the optimal solution to this problem. Therefore, instead of learning generative models, we can use our optimal strategy to directly pick the most informative query given any history. We then develop a practical algorithm by defining a finite-dimensional parameterization of our strategy and classifier using deep networks and train them end-to-end using our objective. Empirically, V-IP is 10-100x faster than IP on different Vision and NLP tasks with competitive performance. Moreover, V-IP finds much shorter query chains when compared to reinforcement learning which is typically used in sequential-decision-making problems. Finally, we demonstrate the utility of V-IP on challenging tasks like medical diagnosis where the performance is far superior to the generative modelling approach."
                },
                {
                    "title": "Approximating the Shapley Value without Marginal Contributions",
                    "abstract": "The Shapley value, which is arguably the most popular approach for assigning a meaningful contribution value to players in a cooperative game, has recently been used intensively in explainable artificial intelligence. Its meaningfulness is due to axiomatic properties that only the Shapley value satisfies, which, however, comes at the expense of an exact computation growing exponentially with the number of agents. Accordingly, a number of works are devoted to the efficient approximation of the Shapley value, most of them revolve around the notion of an agent's marginal contribution. In this paper, we propose with SVARM and Stratified SVARM two parameter-free and domain-independent approximation algorithms based on a representation of the Shapley value detached from the notion of marginal contribution. We prove unmatched theoretical guarantees regarding their approximation quality and provide empirical results including synthetic games as well as common explainability use cases comparing ourselves with state-of-the-art methods."
                },
                {
                    "title": "Learning to Maximize Mutual Information for Dynamic Feature Selection",
                    "abstract": "Feature selection helps reduce data acquisition costs in ML, but the standard approach is to train models with static feature subsets. Here, we consider the dynamic feature selection (DFS) problem where a model sequentially queries features based on the presently available information. DFS is often addressed with reinforcement learning, but we explore a simpler approach of greedily selecting features based on their conditional mutual information. This method is theoretically appealing but requires oracle access to the data distribution, so we develop a learning approach based on amortized optimization. The proposed method is shown to recover the greedy policy when trained to optimality, and it outperforms numerous existing feature selection methods in our experiments, thus validating it as a simple but powerful approach for this problem."
                },
                {
                    "title": "Understanding Influence Functions and Datamodels via Harmonic Analysis",
                    "abstract": "Influence functions estimate effect of individual data points on predictions of the model on test data and were adapted to deep learning in Koh and Liang [2017]. They have been used for detecting data poisoning, detecting helpful and harmful examples, influence of groups of datapoints, etc. Recently, Ilyas et al. [2022] introduced a linear regression method they termed datamodels to predict the effect of training points on outputs on test data. The current paper seeks to provide a better theoretical understanding of such interesting empirical phenomena. The primary tool is harmonic analysis and the idea of noise stability. Contributions include: (a) Exact characterization of the learnt datamodel in terms of Fourier coefficients. (b) An efficient method to estimate the residual error and quality of the optimum linear datamodel without having to train the datamodel. (c) New insights into when influences of groups of datapoints may or may not add up linearly."
                },
                {
                    "title": "WeightedSHAP: analyzing and improving Shapley based feature attributions",
                    "abstract": "Shapley value is a popular approach for measuring the influence of individual features. While Shapley feature attribution is built upon desiderata from game theory, some of its constraints may be less natural in certain machine learning settings, leading to unintuitive model interpretation. In particular, the Shapley value uses the same weight for all marginal contributions -- i.e. it gives the same importance when a large number of other features are given versus when a small number of other features are given. This property can be problematic if larger feature sets are more or less informative than smaller feature sets. Our work performs a rigorous analysis of the potential limitations of Shapley feature attribution. We identify simple settings where the Shapley value is mathematically suboptimal by assigning larger attributions for less influential features. Motivated by this observation, we propose WeightedSHAP, which generalizes the Shapley value and learns which marginal contributions to focus directly from data. On several real-world datasets, we demonstrate that the influential features identified by WeightedSHAP are better able to recapitulate the model's predictions compared to the features identified by the Shapley value."
                },
                {
                    "title": "Interpretable by Design: Learning Predictors by Composing Interpretable Queries",
                    "abstract": "There is a growing concern about typically opaque decision-making with high-performance machine learning algorithms. Providing an explanation of the reasoning process in domain-specific terms can be crucial for adoption in risk-sensitive domains such as healthcare. We argue that machine learning algorithms should be interpretable by design and that the language in which these interpretations are expressed should be domain- and task-dependent. Consequently, we base our model's prediction on a family of user-defined and task-specific binary functions of the data, each having a clear interpretation to the end-user. We then minimize the expected number of queries needed for accurate prediction on any given input. As the solution is generally intractable, following prior work, we choose the queries sequentially based on information gain. However, in contrast to previous work, we need not assume the queries are conditionally independent. Instead, we leverage a stochastic generative model (VAE) and an MCMC algorithm (Unadjusted Langevin) to select the most informative query about the input based on previous query-answers. This enables the online determination of a query chain of whatever depth is required to resolve prediction ambiguities. Finally, experiments on vision and NLP tasks demonstrate the efficacy of our approach and its superiority over post-hoc explanations."
                },
                {
                    "title": "Learning to Estimate Shapley Values with Vision Transformers",
                    "abstract": "Transformers have become a default architecture in computer vision, but understanding what drives their predictions remains a challenging problem. Current explanation approaches rely on attention values or input gradients, but these provide a limited view of a model's dependencies. Shapley values offer a theoretically sound alternative, but their computational cost makes them impractical for large, high-dimensional models. In this work, we aim to make Shapley values practical for vision transformers (ViTs). To do so, we first leverage an attention masking approach to evaluate ViTs with partial information, and we then develop a procedure to generate Shapley value explanations via a separate, learned explainer model. Our experiments compare Shapley values to many baseline methods (e.g., attention rollout, GradCAM, LRP), and we find that our approach provides more accurate explanations than existing methods for ViTs."
                },
                {
                    "title": "Missingness Bias in Model Debugging",
                    "abstract": "Missingness, or the absence of features from an input, is a concept fundamental to many model debugging tools. However, in computer vision, pixels cannot simply be removed from an image. One thus tends to resort to heuristics such as blacking out pixels, which may in turn introduce bias into the debugging process. We study such biases and, in particular, show how transformer-based architectures can enable a more natural implementation of missingness, which side-steps these issues and improves the reliability of model debugging in practice. Our code is available at https://github.com/madrylab/missingness"
                },
                {
                    "title": "Datamodels: Predicting Predictions from Training Data",
                    "abstract": "We present a conceptual framework, datamodeling, for analyzing the behavior of a model class in terms of the training data. For any fixed\"target\"example $x$, training set $S$, and learning algorithm, a datamodel is a parameterized function $2^S \\to \\mathbb{R}$ that for any subset of $S' \\subset S$ -- using only information about which examples of $S$ are contained in $S'$ -- predicts the outcome of training a model on $S'$ and evaluating on $x$. Despite the potential complexity of the underlying process being approximated (e.g., end-to-end training and evaluation of deep neural networks), we show that even simple linear datamodels can successfully predict model outputs. We then demonstrate that datamodels give rise to a variety of applications, such as: accurately predicting the effect of dataset counterfactuals; identifying brittle predictions; finding semantically similar examples; quantifying train-test leakage; and embedding data into a well-behaved and feature-rich representation space. Data for this paper (including pre-computed datamodels as well as raw predictions from four million trained deep neural networks) is available at https://github.com/MadryLab/datamodels-data ."
                },
                {
                    "title": "A Probabilistic Hard Attention Model For Sequentially Observed Scenes",
                    "abstract": "A visual hard attention model actively selects and observes a sequence of subregions in an image to make a prediction. The majority of hard attention models determine the attention-worthy regions by first analyzing a complete image. However, it may be the case that the entire image is not available initially but instead sensed gradually through a series of partial observations. In this paper, we design an efficient hard attention model for classifying such sequentially observed scenes. The presented model never observes an image completely. To select informative regions under partial observability, the model uses Bayesian Optimal Experiment Design. First, it synthesizes the features of the unobserved regions based on the already observed regions. Then, it uses the predicted features to estimate the expected information gain (EIG) attained, should various regions be attended. Finally, the model attends to the actual content on the location where the EIG mentioned above is maximum. The model uses a) a recurrent feature aggregator to maintain a recurrent state, b) a linear classifier to predict the class label, c) a Partial variational autoencoder to predict the features of unobserved regions. We use normalizing flows in Partial VAE to handle multi-modality in the feature-synthesis problem. We train our model using a differentiable objective and test it on five datasets. Our model gains 2-10% higher accuracy than the baseline models when both have seen only a couple of glimpses."
                },
                {
                    "title": "Beta Shapley: a Unified and Noise-reduced Data Valuation Framework for Machine Learning",
                    "abstract": "Data Shapley has recently been proposed as a principled framework to quantify the contribution of individual datum in machine learning. It can effectively identify helpful or harmful data points for a learning algorithm. In this paper, we propose Beta Shapley, which is a substantial generalization of Data Shapley. Beta Shapley arises naturally by relaxing the efficiency axiom of the Shapley value, which is not critical for machine learning settings. Beta Shapley unifies several popular data valuation methods and includes data Shapley as a special case. Moreover, we prove that Beta Shapley has several desirable statistical properties and propose efficient algorithms to estimate it. We demonstrate that Beta Shapley outperforms state-of-the-art data valuation methods on several downstream ML tasks such as: 1) detecting mislabeled training data; 2) learning with subsamples; and 3) identifying points whose addition or removal have the largest positive or negative impact on the model."
                },
                {
                    "title": "FastSHAP: Real-Time Shapley Value Estimation",
                    "abstract": "Shapley values are widely used to explain black-box models, but they are costly to calculate because they require many model evaluations. We introduce FastSHAP, a method for estimating Shapley values in a single forward pass using a learned explainer model. FastSHAP amortizes the cost of explaining many inputs via a learning approach inspired by the Shapley value's weighted least squares characterization, and it can be trained using standard stochastic gradient optimization. We compare FastSHAP to existing estimation approaches, revealing that it generates high-quality explanations with orders of magnitude speedup."
                },
                {
                    "title": "Amortized Generation of Sequential Algorithmic Recourses for Black-Box Models",
                    "abstract": "Explainable machine learning (ML) has gained traction in recent years due to the increasing adoption of ML-based systems in many sectors. Algorithmic Recourses (ARs) provide \"what if\" feedback of the form \"if an input datapoint were x' instead of x, then an ML-based system's output would be y' instead of y.\" Recourses are attractive due to their actionable feedback, amenability to existing legal frameworks, and fidelity to the underlying ML model. Yet, current recourse approaches are single shot that is, they assume x can change to x' in a single time period. We propose a novel stochastic-control-based approach that generates sequential recourses, that is, recourses that allow x to move stochastically and sequentially across intermediate states to a final state x'. Our approach is model agnostic and black box. Furthermore, the calculation of recourses is amortized such that once trained, it applies to multiple datapoints without the need for re-optimization. In addition to these primary characteristics, our approach admits optional desiderata such as adherence to the data manifold, respect for causal relations, and sparsity identified by past research as desirable properties of recourses. We evaluate our approach using three real-world datasets and show successful generation of sequential recourses that respect other recourse desiderata."
                },
                {
                    "title": "Sampling Permutations for Shapley Value Estimation",
                    "abstract": "Game-theoretic attribution techniques based on Shapley values are used to interpret black-box machine learning models, but their exact calculation is generally NP-hard, requiring approximation methods for non-trivial models. As the computation of Shapley values can be expressed as a summation over a set of permutations, a common approach is to sample a subset of these permutations for approximation. Unfortunately, standard Monte Carlo sampling methods can exhibit slow convergence, and more sophisticated quasi-Monte Carlo methods have not yet been applied to the space of permutations. To address this, we investigate new approaches based on two classes of approximation methods and compare them empirically. First, we demonstrate quadrature techniques in a RKHS containing functions of permutations, using the Mallows kernel in combination with kernel herding and sequential Bayesian quadrature. The RKHS perspective also leads to quasi-Monte Carlo type error bounds, with a tractable discrepancy measure defined on permutations. Second, we exploit connections between the hypersphere $\\mathbb{S}^{d-2}$ and permutations to create practical algorithms for generating permutation samples with good properties. Experiments show the above techniques provide significant improvements for Shapley value estimates over existing methods, converging to a smaller RMSE in the same number of model evaluations."
                },
                {
                    "title": "Data Shapley Valuation for Efficient Batch Active Learning",
                    "abstract": "Annotating the right set of data amongst all available data points is a key challenge in many machine learning applications. Batch active learning is a popular approach to address this, in which batches of unlabeled data points are selected for annotation, while an underlying learning algorithm gets subsequently updated. In this work, we introduce Active Data Shapley (ADS) \u2013a filtering layer for batch active learning that significantly increases the efficiency of existing active learning algorithms by pre-selecting, using a linear time computation, the highest-value points from an unlabeled dataset. Using the notion of the Shapley value of data, our method estimates the value of unlabeled data points with regards to the prediction task at hand. We show that ADS is particularly effective when the pool of unlabeled data exhibits real-world caveats: noise, heterogeneity, and domain shift. We run experiments demonstrating that when ADS is used to pre-select the highest-ranking portion of an unlabeled dataset, the efficiency of state-of-the-art batch active learning methods increases by an average factor of 6x, while preserving performance effectiveness."
                },
                {
                    "title": "Efficient Explanations from Empirical Explainers",
                    "abstract": "Amid a discussion about Green AI in which we see explainability neglected, we explore the possibility to efficiently approximate computationally expensive explainers. To this end, we propose feature attribution modelling with Empirical Explainers. Empirical Explainers learn from data to predict the attribution maps of expensive explainers. We train and test Empirical Explainers in the language domain and find that they model their expensive counterparts surprisingly well, at a fraction of the cost. They could thus mitigate the computational burden of neural explanations significantly, in applications that tolerate an approximation error."
                },
                {
                    "title": "Have We Learned to Explain?: How Interpretability Methods Can Learn to Encode Predictions in their Interpretations",
                    "abstract": "While the need for interpretable machine learning has been established, many common approaches are slow, lack fidelity, or hard to evaluate. Amortized explanation methods reduce the cost of providing interpretations by learning a global selector model that returns feature importances for a single instance of data. The selector model is trained to optimize the fidelity of the interpretations, as evaluated by a predictor model for the target. Popular methods learn the selector and predictor model in concert, which we show allows predictions to be encoded within interpretations. We introduce EVAL-X as a method to quantitatively evaluate interpretations and REAL-X as an amortized explanation method, which learn a predictor model that approximates the true data generating distribution given any subset of the input. We show EVAL-X can detect when predictions are encoded in interpretations and show the advantages of REAL-X through quantitative and radiologist evaluation."
                },
                {
                    "title": "BSODA: A Bipartite Scalable Framework for Online Disease Diagnosis",
                    "abstract": "A growing number of people are seeking healthcare advice online. Usually, they diagnose their medical conditions based on the symptoms they are experiencing, which is also known as self-diagnosis. From the machine learning perspective, online disease diagnosis is a sequential feature (symptom) selection and classification problem. Reinforcement learning (RL) methods are the standard approaches to this type of tasks. Generally, they perform well when the feature space is small, but frequently become inefficient in tasks with a large number of features, such as the self-diagnosis. To address the challenge, we propose a non-RL Bipartite Scalable framework for Online Disease diAgnosis, called BSODA. BSODA is composed of two cooperative branches that handle symptom-inquiry and disease-diagnosis, respectively. The inquiry branch determines which symptom to collect next by an information-theoretic reward. We employ a Product-of-Experts encoder to significantly improve the handling of partial observations of a large number of features. Besides, we propose several approximation methods to substantially reduce the computational cost of the reward to a level that is acceptable for online services. Additionally, we leverage the diagnosis model to estimate the reward more precisely. For the diagnosis branch, we use a knowledge-guided self-attention model to perform predictions. In particular, BSODA determines when to stop inquiry and output predictions using both the inquiry and diagnosis models. We demonstrate that BSODA outperforms the state-of-the-art methods on several public datasets. Moreover, we propose a novel evaluation method to test the transferability of symptom checking methods from synthetic to real-world tasks. Compared to existing RL baselines, BSODA is more effectively scalable to large search spaces."
                },
                {
                    "title": "Improving KernelSHAP: Practical Shapley Value Estimation via Linear Regression",
                    "abstract": "The Shapley value solution concept from cooperative game theory has become popular for interpreting ML models, but efficiently estimating Shapley values remains challenging, particularly in the model-agnostic setting. We revisit the idea of estimating Shapley values via linear regression to understand and improve upon this approach. By analyzing KernelSHAP alongside a newly proposed unbiased estimator, we develop techniques to detect its convergence and calculate uncertainty estimates. We also find that that the original version incurs a negligible increase in bias in exchange for a significant reduction in variance, and we propose a variance reduction technique that further accelerates the convergence of both estimators. Finally, we develop a version of KernelSHAP for stochastic cooperative games that yields fast new estimators for two global explanation methods."
                },
                {
                    "title": "Explaining by Removing: A Unified Framework for Model Explanation",
                    "abstract": "Researchers have proposed a wide variety of model explanation approaches, but it remains unclear how most methods are related or when one method is preferable to another. We establish a new class of methods, removal-based explanations, that are based on the principle of simulating feature removal to quantify each feature's influence. These methods vary in several respects, so we develop a framework that characterizes each method along three dimensions: 1) how the method removes features, 2) what model behavior the method explains, and 3) how the method summarizes each feature's influence. Our framework unifies 25 existing methods, including several of the most widely used approaches (SHAP, LIME, Meaningful Perturbations, permutation tests). This new class of explanation methods has rich connections that we examine using tools that have been largely overlooked by the explainability literature. To anchor removal-based explanations in cognitive psychology, we show that feature removal is a simple application of subtractive counterfactual reasoning. Ideas from cooperative game theory shed light on the relationships and trade-offs among different methods, and we derive conditions under which all removal-based explanations have information-theoretic interpretations. Through this analysis, we develop a unified framework that helps practitioners better understand model explanation tools, and that offers a strong theoretical foundation upon which future explainability research can build."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "A Multilinear Sampling Algorithm to Estimate Shapley Values",
                    "abstract": "Shapley values are great analytical tools in game theory to measure the importance of a player in a game. Due to their axiomatic and desirable properties such as efficiency, they have become popular for feature importance analysis in data science and machine learning. However, the time complexity to compute Shapley values based on the original formula is exponential, and as the number of features increases, this becomes infeasible. Castro et al. [1] developed a sampling algorithm, to estimate Shapley values. In this work, we propose a new sampling method based on a multilinear extension technique as applied in game theory. The aim is to provide a more efficient (sampling) method for estimating Shapley values. Our method is applicable to any machine learning model, in particular for either multiclass classifications or regression problems. We apply the method to estimate Shapley values for multilayer perceptrons (MLPs) and through experimentation on two datasets, we demonstrate that our method provides more accurate estimations of the Shapley values by reducing the variance of the sampling statistics."
                },
                {
                    "title": "Counterfactual Explanations for Machine Learning: A Review",
                    "abstract": "Machine learning plays a role in many deployed decision systems, often in ways that are difficult or impossible to understand by human stakeholders. Explaining, in a human-understandable way, the relationship between the input and output of machine learning models is essential to the development of trustworthy machine-learning-based systems. A burgeoning body of research seeks to define the goals and methods of explainability in machine learning. In this paper, we seek to review and categorize research on counterfactual explanations, a specific class of explanation that provides a link between what could have happened had input to a model been changed in a particular way. Modern approaches to counterfactual explainability in machine learning draw connections to the established legal doctrine in many countries, making them appealing to fielded systems in high-impact areas such as finance and healthcare. Thus, we design a rubric with desirable properties of counterfactual explanation algorithms and comprehensively evaluate all currently-proposed algorithms against that rubric. Our rubric provides easy comparison and comprehension of the advantages and disadvantages of different approaches and serves as an introduction to major research themes in this field. We also identify gaps and discuss promising research directions in the space of counterfactual explainability."
                },
                {
                    "title": "What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation",
                    "abstract": "Deep learning algorithms are well-known to have a propensity for fitting the training data very well and often fit even outliers and mislabeled data points. Such fitting requires memorization of training data labels, a phenomenon that has attracted significant research interest but has not been given a compelling explanation so far. A recent work of Feldman (2019) proposes a theoretical explanation for this phenomenon based on a combination of two insights. First, natural image and data distributions are (informally) known to be long-tailed, that is have a significant fraction of rare and atypical examples. Second, in a simple theoretical model such memorization is necessary for achieving close-to-optimal generalization error when the data distribution is long-tailed. However, no direct empirical evidence for this explanation or even an approach for obtaining such evidence were given. \nIn this work we design experiments to test the key ideas in this theory. The experiments require estimation of the influence of each training example on the accuracy at each test example as well as memorization values of training examples. Estimating these quantities directly is computationally prohibitive but we show that closely-related subsampled influence and memorization values can be estimated much more efficiently. Our experiments demonstrate the significant benefits of memorization for generalization on several standard benchmarks. They also provide quantitative and visually compelling evidence for the theory put forth in (Feldman, 2019)."
                },
                {
                    "title": "Efficient nonparametric statistical inference on population feature importance using Shapley values",
                    "abstract": "The true population-level importance of a variable in a prediction task provides useful knowledge about the underlying data-generating mechanism and can help in deciding which measurements to collect in subsequent experiments. Valid statistical inference on this importance is a key component in understanding the population of interest. We present a computationally efficient procedure for estimating and obtaining valid statistical inference on the Shapley Population Variable Importance Measure (SPVIM). Although the computational complexity of the true SPVIM scales exponentially with the number of variables, we propose an estimator based on randomly sampling only \u0398(n) feature subsets given n observations. We prove that our estimator converges at an asymptotically optimal rate. Moreover, by deriving the asymptotic distribution of our estimator, we construct valid confidence intervals and hypothesis tests. Our procedure has good finite-sample performance in simulations, and for an in-hospital mortality prediction task produces similar variable importance estimates when different machine learning algorithms are applied."
                },
                {
                    "title": "Quantifying Attention Flow in Transformers",
                    "abstract": "In the Transformer model, \u201cself-attention\u201d combines information from attended embeddings into the representation of the focal embedding in the next layer. Thus, across layers of the Transformer, information originating from different tokens gets increasingly mixed. This makes attention weights unreliable as explanations probes. In this paper, we consider the problem of quantifying this flow of information through self-attention. We propose two methods for approximating the attention to input tokens given attention weights, attention rollout and attention flow, as post hoc methods when we use attention weights as the relative relevance of the input tokens. We show that these methods give complementary views on the flow of information, and compared to raw attention, both yield higher correlations with importance scores of input tokens obtained using an ablation method and input gradients."
                },
                {
                    "title": "A Distributional Framework for Data Valuation",
                    "abstract": "Shapley value is a classic notion from game theory, historically used to quantify the contributions of individuals within groups, and more recently applied to assign values to data points when training machine learning models. Despite its foundational role, a key limitation of the data Shapley framework is that it only provides valuations for points within a fixed data set. It does not account for statistical aspects of the data and does not give a way to reason about points outside the data set. To address these limitations, we propose a novel framework -- distributional Shapley -- where the value of a point is defined in the context of an underlying data distribution. We prove that distributional Shapley has several desirable statistical properties; for example, the values are stable under perturbations to the data points themselves and to the underlying data distribution. We leverage these properties to develop a new algorithm for estimating values from data, which comes with formal guarantees and runs two orders of magnitude faster than state-of-the-art algorithms for computing the (non-distributional) data Shapley values. We apply distributional Shapley to diverse data sets and demonstrate its utility in a data market setting."
                },
                {
                    "title": "fastai: A Layered API for Deep Learning",
                    "abstract": "fastai is a deep learning library which provides practitioners with high-level components that can quickly and easily provide state-of-the-art results in standard deep learning domains, and provides researchers with low-level components that can be mixed and matched to build new approaches. It aims to do both things without substantial compromises in ease of use, flexibility, or performance. This is possible thanks to a carefully layered architecture, which expresses common underlying patterns of many deep learning and data processing techniques in terms of decoupled abstractions. These abstractions can be expressed concisely and clearly by leveraging the dynamism of the underlying Python language and the flexibility of the PyTorch library. fastai includes: a new type dispatch system for Python along with a semantic type hierarchy for tensors; a GPU-optimized computer vision library which can be extended in pure Python; an optimizer which refactors out the common functionality of modern optimizers into two basic pieces, allowing optimization algorithms to be implemented in 4\u20135 lines of code; a novel 2-way callback system that can access any part of the data, model, or optimizer and change it at any point during training; a new data block API; and much more. We used this library to successfully create a complete deep learning course, which we were able to write more quickly than using previous approaches, and the code was more clear. The library is already in wide use in research, industry, and teaching."
                },
                {
                    "title": "Preserving Causal Constraints in Counterfactual Explanations for Machine Learning Classifiers",
                    "abstract": "To construct interpretable explanations that are consistent with the original ML model, counterfactual examples---showing how the model's output changes with small perturbations to the input---have been proposed. This paper extends the work in counterfactual explanations by addressing the challenge of feasibility of such examples. For explanations of ML models in critical domains such as healthcare and finance, counterfactual examples are useful for an end-user only to the extent that perturbation of feature inputs is feasible in the real world. We formulate the problem of feasibility as preserving causal relationships among input features and present a method that uses (partial) structural causal models to generate actionable counterfactuals. When feasibility constraints cannot be easily expressed, we consider an alternative mechanism where people can label generated CF examples on feasibility: whether it is feasible to intervene and realize the candidate CF example from the original input. To learn from this labelled feasibility data, we propose a modified variational auto encoder loss for generating CF examples that optimizes for feasibility as people interact with its output. Our experiments on Bayesian networks and the widely used ''Adult-Income'' dataset show that our proposed methods can generate counterfactual explanations that better satisfy feasibility constraints than existing methods.. Code repository can be accessed here: \\textit{this https URL}"
                },
                {
                    "title": "Explanation by Progressive Exaggeration",
                    "abstract": "As machine learning methods see greater adoption and implementation in high stakes applications such as medical image diagnosis, the need for model interpretability and explanation has become more critical. Classical approaches that assess feature importance (e.g. saliency maps) do not explain how and why a particular region of an image is relevant to the prediction. We propose a method that explains the outcome of a classification black-box by gradually exaggerating the semantic effect of a given class. Given a query input to a classifier, our method produces a progressive set of plausible variations of that query, which gradually changes the posterior probability from its original class to its negation. These counter-factually generated samples preserve features unrelated to the classification decision, such that a user can employ our method as a \"tuning knob\" to traverse a data manifold while crossing the decision boundary. Our method is model agnostic and only requires the output value and gradient of the predictor with respect to its input."
                },
                {
                    "title": "CXPlain: Causal Explanations for Model Interpretation under Uncertainty",
                    "abstract": "Feature importance estimates that inform users about the degree to which given inputs influence the output of a predictive model are crucial for understanding, validating, and interpreting machine-learning models. However, providing fast and accurate estimates of feature importance for high-dimensional data, and quantifying the uncertainty of such estimates remain open challenges. Here, we frame the task of providing explanations for the decisions of machine-learning models as a causal learning task, and train causal explanation (CXPlain) models that learn to estimate to what degree certain inputs cause outputs in another machine-learning model. CXPlain can, once trained, be used to explain the target model in little time, and enables the quantification of the uncertainty associated with its feature importance estimates via bootstrap ensembling. We present experiments that demonstrate that CXPlain is significantly more accurate and faster than existing model-agnostic methods for estimating feature importance. In addition, we confirm that the uncertainty estimates provided by CXPlain ensembles are strongly correlated with their ability to accurately estimate feature importance on held-out data."
                },
                {
                    "title": "Understanding Deep Networks via Extremal Perturbations and Smooth Masks",
                    "abstract": "Attribution is the problem of finding which parts of an image are the most responsible for the output of a deep neural network. An important family of attribution methods is based on measuring the effect of perturbations applied to the input image, either via exhaustive search or by finding representative perturbations via optimization. In this paper, we discuss some of the shortcomings of existing approaches to perturbation analysis and address them by introducing the concept of extremal perturbations, which are theoretically grounded and interpretable. We also introduce a number of technical innovations to compute these extremal perturbations, including a new area constraint and a parametric family of smooth perturbations, which allow us to remove all tunable weighing factors from the optimization problem. We analyze the effect of perturbations as a function of their area, demonstrating excellent sensitivity to the spatial properties of the network under stimulation. We also extend perturbation analysis to the intermediate layers of a deep neural network. This application allows us to show how compactly an image can be represented (in terms of the number of channels it requires). We also demonstrate that the consistency with which images of a given class rely on the same intermediate channel correlates well with class accuracy."
                },
                {
                    "title": "Meta-Learning with Implicit Gradients",
                    "abstract": "A core capability of intelligent systems is the ability to quickly learn new tasks by drawing on prior experience. Gradient (or optimization) based meta-learning has recently emerged as an effective approach for few-shot learning. In this formulation, meta-parameters are learned in the outer loop, while task-specific models are learned in the inner-loop, by using only a small amount of data from the current task. A key challenge in scaling these approaches is the need to differentiate through the inner loop learning process, which can impose considerable computational and memory burdens. By drawing upon implicit differentiation, we develop the implicit MAML algorithm, which depends only on the solution to the inner level optimization and not the path taken by the inner loop optimizer. This effectively decouples the meta-gradient computation from the choice of inner loop optimizer. As a result, our approach is agnostic to the choice of inner loop optimizer and can gracefully handle many gradient steps without vanishing gradients or memory constraints. Theoretically, we prove that implicit MAML can compute accurate meta-gradients with a memory footprint that is, up to small constant factors, no more than that which is required to compute a single inner loop gradient and at no overall increase in the total computational cost. Experimentally, we show that these benefits of implicit MAML translate into empirical gains on few-shot image recognition benchmarks."
                },
                {
                    "title": "Efficient Task-Specific Data Valuation for Nearest Neighbor Algorithms",
                    "abstract": "\n Given a data set\n D\n containing millions of data points and a data consumer who is willing to pay for $\n X\n to train a machine learning (ML) model over\n D\n ,\n how should we distribute this $X to each data point to reflect its \"value\"?\n In this paper, we define the \"relative value of data\" via the Shapley value, as it uniquely possesses properties with appealing real-world interpretations, such as fairness, rationality and decentralizability. For general, bounded utility functions, the Shapley value is known to be challenging to compute: to get Shapley values for all\n N\n data points, it requires\n O\n (2\n \n N\n \n ) model evaluations for exact computation and\n O\n (\n N\n log\n N\n ) for (\n \u03f5\n , \u03b4)-approximation.\n \n \n In this paper, we focus on one popular family of ML models relying on\n K\n -nearest neighbors (\n K\n NN). The most surprising result is that for unweighted\n K\n NN classifiers and regressors, the Shapley value of all\n N\n data points can be computed,\n exactly\n , in\n O\n (\n N\n log\n N\n ) time - an exponential improvement on computational complexity! Moreover, for (\n \u03f5\n , \u03b4)-approximation, we are able to develop an algorithm based on Locality Sensitive Hashing (LSH) with only sublinear complexity\n O\n (\n N\n \n h\n (\n \u03f5\n ,\n K\n )\n \n log\n N\n ) when\n \u03f5\n is not too small and\n K\n is not too large. We empirically evaluate our algorithms on up to 10 million data points and even our\n exact\n algorithm is up to three orders of magnitude faster than the baseline approximation algorithm. The LSH-based approximation algorithm can accelerate the value calculation process even further.\n \n \n We then extend our algorithm to other scenarios such as (1) weighed\n K\n NN classifiers, (2) different data points are clustered by different\n data curators\n , and (3) there are\n data analysts\n providing computation who also requires proper valuation.\n Some\n of these extensions, although also being improved exponentially, are less practical for exact computation (e.g.,\n O\n (\n \n N\n K\n \n ) complexity for weigthed\n K\n NN). We thus propose an Monte Carlo approximation algorithm, which is\n O\n (\n N\n (log\n N\n )\n 2\n /(log\n K\n )\n 2\n ) times more efficient than the baseline approximation algorithm.\n"
                },
                {
                    "title": "Data Shapley: Equitable Valuation of Data for Machine Learning",
                    "abstract": "As data becomes the fuel driving technological and economic growth, a fundamental challenge is how to quantify the value of data in algorithmic predictions and decisions. For example, in healthcare and consumer markets, it has been suggested that individuals should be compensated for the data that they generate, but it is not clear what is an equitable valuation for individual data. In this work, we develop a principled framework to address data valuation in the context of supervised machine learning. Given a learning algorithm trained on $n$ data points to produce a predictor, we propose data Shapley as a metric to quantify the value of each training datum to the predictor performance. Data Shapley value uniquely satisfies several natural properties of equitable data valuation. We develop Monte Carlo and gradient-based methods to efficiently estimate data Shapley values in practical settings where complex learning algorithms, including neural networks, are trained on large datasets. In addition to being equitable, extensive experiments across biomedical, image and synthetic data demonstrate that data Shapley has several other benefits: 1) it is more powerful than the popular leave-one-out or leverage score in providing insight on what data is more valuable for a given learning task; 2) low Shapley value data effectively capture outliers and corruptions; 3) high Shapley value data inform what type of new data to acquire to improve the predictor."
                },
                {
                    "title": "The Shapley Taylor Interaction Index",
                    "abstract": "The attribution problem, that is the problem of attributing a model's prediction to its base features, is well-studied. We extend the notion of attribution to also apply to feature interactions. \nThe Shapley value is a commonly used method to attribute a model's prediction to its base features. We propose a generalization of the Shapley value called Shapley-Taylor index that attributes the model's prediction to interactions of subsets of features up to some size k. The method is analogous to how the truncated Taylor Series decomposes the function value at a certain point using its derivatives at a different point. In fact, we show that the Shapley Taylor index is equal to the Taylor Series of the multilinear extension of the set-theoretic behavior of the model. \nWe axiomatize this method using the standard Shapley axioms -- linearity, dummy, symmetry and efficiency -- and an additional axiom that we call the interaction distribution axiom. This new axiom explicitly characterizes how interactions are distributed for a class of functions that model pure interaction. \nWe contrast the Shapley-Taylor index against the previously proposed Shapley Interaction index (cf. [9]) from the cooperative game theory literature. We also apply the Shapley Taylor index to three models and identify interesting qualitative insights."
                },
                {
                    "title": "LS-Tree: Model Interpretation When the Data Are Linguistic",
                    "abstract": "We study the problem of interpreting trained classification models in the setting of linguistic data sets. Leveraging a parse tree, we propose to assign least-squares-based importance scores to each word of an instance by exploiting syntactic constituency structure. We establish an axiomatic characterization of these importance scores by relating them to the Banzhaf value in coalitional game theory. Based on these importance scores, we develop a principled method for detecting and quantifying interactions between words in a sentence. We demonstrate that the proposed method can aid in interpretability and diagnostics for several widely-used language models."
                },
                {
                    "title": "Soft Actor-Critic Algorithms and Applications",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been successfully applied to a range of challenging sequential decision making and control tasks. However, these methods typically suffer from two major challenges: high sample complexity and brittleness to hyperparameters. Both of these challenges limit the applicability of such methods to real-world domains. In this paper, we describe Soft Actor-Critic (SAC), our recently introduced off-policy actor-critic algorithm based on the maximum entropy RL framework. In this framework, the actor aims to simultaneously maximize expected return and entropy. That is, to succeed at the task while acting as randomly as possible. We extend SAC to incorporate a number of modifications that accelerate training and improve stability with respect to the hyperparameters, including a constrained formulation that automatically tunes the temperature hyperparameter. We systematically evaluate SAC on a range of benchmark tasks, as well as real-world challenging tasks such as locomotion for a quadrupedal robot and robotic manipulation with a dexterous hand. With these improvements, SAC achieves state-of-the-art performance, outperforming prior on-policy and off-policy methods in sample-efficiency and asymptotic performance. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving similar performance across different random seeds. These results suggest that SAC is a promising candidate for learning in real-world robotics tasks."
                },
                {
                    "title": "EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE",
                    "abstract": "Many real-life decision-making situations allow further relevant information to be acquired at a specific cost, for example, in assessing the health status of a patient we may decide to take additional measurements such as diagnostic tests or imaging scans before making a final assessment. Acquiring more relevant information enables better decision making, but may be costly. How can we trade off the desire to make good decisions by acquiring further information with the cost of performing that acquisition? To this end, we propose a principled framework, named EDDI (Efficient Dynamic Discovery of high-value Information), based on the theory of Bayesian experimental design. In EDDI, we propose a novel partial variational autoencoder (Partial VAE) to predict missing data entries problematically given any subset of the observed ones, and combine it with an acquisition function that maximizes expected information gain on a set of target variables. We show cost reduction at the same decision quality and improved decision quality at the same cost in multiple machine learning benchmarks and two real-world health-care applications."
                },
                {
                    "title": "Learning to Explain: An Information-Theoretic Perspective on Model Interpretation",
                    "abstract": "We introduce instancewise feature selection as a methodology for model interpretation. Our method is based on learning a function to extract a subset of features that are most informative for each given example. This feature selector is trained to maximize the mutual information between selected features and the response variable, where the conditional distribution of the response variable given the input is the model to be explained. We develop an efficient variational approximation to the mutual information, and show the effectiveness of our method on a variety of synthetic and real data sets using both quantitative metrics and human evaluation."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Interpretation of Neural Networks is Fragile",
                    "abstract": "In order for machine learning to be trusted in many applications, it is critical to be able to reliably explain why the machine learning algorithm makes certain predictions. For this reason, a variety of methods have been developed recently to interpret neural network predictions by providing, for example, feature importance maps. For both scientific robustness and security reasons, it is important to know to what extent can the interpretations be altered by small systematic perturbations to the input data, which might be generated by adversaries or by measurement biases. In this paper, we demonstrate how to generate adversarial perturbations that produce perceptively indistinguishable inputs that are assigned the same predicted label, yet have very different interpretations. We systematically characterize the robustness of interpretations generated by several widely-used feature importance interpretation methods (feature importance maps, integrated gradients, and DeepLIFT) on ImageNet and CIFAR-10. In all cases, our experiments show that systematic perturbations can lead to dramatically different interpretations without changing the label. We extend these results to show that interpretations based on exemplars (e.g. influence functions) are similarly susceptible to adversarial attack. Our analysis of the geometry of the Hessian matrix gives insight on why robustness is a general challenge to current interpretation approaches."
                },
                {
                    "title": "Counterfactual Explanations Without Opening the Black Box: Automated Decisions and the GDPR",
                    "abstract": "There has been much discussion of the right to explanation in the EU General Data Protection Regulation, and its existence, merits, and disadvantages. Implementing a right to explanation that opens the black box of algorithmic decision-making faces major legal and technical barriers. Explaining the functionality of complex algorithmic decision-making systems and their rationale in specific cases is a technically challenging problem. Some explanations may offer little meaningful information to data subjects, raising questions around their value. Explanations of automated decisions need not hinge on the general public understanding how algorithmic systems function. Even though such interpretability is of great importance and should be pursued, explanations can, in principle, be offered without opening the black box. Looking at explanations as a means to help a data subject act rather than merely understand, one could gauge the scope and content of explanations according to the specific goal or action they are intended to support. From the perspective of individuals affected by automated decision-making, we propose three aims for explanations: (1) to inform and help the individual understand why a particular decision was reached, (2) to provide grounds to contest the decision if the outcome is undesired, and (3) to understand what would need to change in order to receive a desired result in the future, based on the current decision-making model. We assess how each of these goals finds support in the GDPR. We suggest data controllers should offer a particular type of explanation, unconditional counterfactual explanations, to support these three aims. These counterfactual explanations describe the smallest change to the world that can be made to obtain a desirable outcome, or to arrive at the closest possible world, without needing to explain the internal logic of the system."
                },
                {
                    "title": "SmoothGrad: removing noise by adding noise",
                    "abstract": "Explaining the output of a deep network remains a challenge. In the case of an image classifier, one type of explanation is to identify pixels that strongly influence the final decision. A starting point for this strategy is the gradient of the class score function with respect to the input image. This gradient can be interpreted as a sensitivity map, and there are several techniques that elaborate on this basic idea. This paper makes two contributions: it introduces SmoothGrad, a simple method that can help visually sharpen gradient-based sensitivity maps, and it discusses lessons in the visualization of these maps. We publish the code for our experiments and a website with our results."
                },
                {
                    "title": "A Unified Approach to Interpreting Model Predictions",
                    "abstract": "Understanding why a model makes a certain prediction can be as crucial as the prediction's accuracy in many applications. However, the highest accuracy for large modern datasets is often achieved by complex models that even experts struggle to interpret, such as ensemble or deep learning models, creating a tension between accuracy and interpretability. In response, various methods have recently been proposed to help users interpret the predictions of complex models, but it is often unclear how these methods are related and when one method is preferable over another. To address this problem, we present a unified framework for interpreting predictions, SHAP (SHapley Additive exPlanations). SHAP assigns each feature an importance value for a particular prediction. Its novel components include: (1) the identification of a new class of additive feature importance measures, and (2) theoretical results showing there is a unique solution in this class with a set of desirable properties. The new class unifies six existing methods, notable because several recent methods in the class lack the proposed desirable properties. Based on insights from this unification, we present new methods that show improved computational performance and/or better consistency with human intuition than previous approaches."
                },
                {
                    "title": "Real Time Image Saliency for Black Box Classifiers",
                    "abstract": "In this work we develop a fast saliency detection method that can be applied to any differentiable image classifier. We train a masking model to manipulate the scores of the classifier by masking salient parts of the input image. Our model generalises well to unseen images and requires a single forward pass to perform saliency detection, therefore suitable for use in real-time systems. We test our approach on CIFAR-10 and ImageNet datasets and show that the produced saliency maps are easily interpretable, sharp, and free of artifacts. We suggest a new metric for saliency and test our method on the ImageNet object localisation task. We achieve results outperforming other weakly supervised methods."
                },
                {
                    "title": "Interpretable Explanations of Black Boxes by Meaningful Perturbation",
                    "abstract": "As machine learning algorithms are increasingly applied to high impact yet high risk tasks, such as medical diagnosis or autonomous driving, it is critical that researchers can explain how such algorithms arrived at their predictions. In recent years, a number of image saliency methods have been developed to summarize where highly complex neural networks \u201clook\u201d in an image for evidence for their predictions. However, these techniques are limited by their heuristic nature and architectural constraints. In this paper, we make two main contributions: First, we propose a general framework for learning different kinds of explanations for any black box algorithm. Second, we specialise the framework to find the part of an image most responsible for a classifier decision. Unlike previous works, our method is model-agnostic and testable because it is grounded in explicit and interpretable image perturbations."
                },
                {
                    "title": "Learning Important Features Through Propagating Activation Differences",
                    "abstract": "The purported \"black box\" nature of neural networks is a barrier to adoption in applications where interpretability is essential. Here we present DeepLIFT (Deep Learning Important FeaTures), a method for decomposing the output prediction of a neural network on a specific input by backpropagating the contributions of all neurons in the network to every feature of the input. DeepLIFT compares the activation of each neuron to its 'reference activation' and assigns contribution scores according to the difference. By optionally giving separate consideration to positive and negative contributions, DeepLIFT can also reveal dependencies which are missed by other approaches. Scores can be computed efficiently in a single backward pass. We apply DeepLIFT to models trained on MNIST and simulated genomic data, and show significant advantages over gradient-based methods. Video tutorial: http://goo.gl/qKb7pL, code: http://goo.gl/RM8jvH."
                },
                {
                    "title": "Understanding Black-box Predictions via Influence Functions",
                    "abstract": "How can we explain the predictions of a black-box model? In this paper, we use influence functions \u2014 a classic technique from robust statistics \u2014 to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. We show that even on non-convex and non-differentiable models where the theory breaks down, approximations to influence functions can still provide valuable information. On linear models and convolutional neural networks, we demonstrate that influence functions are useful for multiple purposes: understanding model behavior, debugging models, detecting dataset errors, and even creating visually-indistinguishable training-set attacks."
                },
                {
                    "title": "Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks",
                    "abstract": "We propose an algorithm for meta-learning that is model-agnostic, in the sense that it is compatible with any model trained with gradient descent and applicable to a variety of different learning problems, including classification, regression, and reinforcement learning. The goal of meta-learning is to train a model on a variety of learning tasks, such that it can solve new learning tasks using only a small number of training samples. In our approach, the parameters of the model are explicitly trained such that a small number of gradient steps with a small amount of training data from a new task will produce good generalization performance on that task. In effect, our method trains the model to be easy to fine-tune. We demonstrate that this approach leads to state-of-the-art performance on two few-shot image classification benchmarks, produces good results on few-shot regression, and accelerates fine-tuning for policy gradient reinforcement learning with neural network policies."
                },
                {
                    "title": "The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables",
                    "abstract": "The reparameterization trick enables optimizing large scale stochastic computation graphs via gradient descent. The essence of the trick is to refactor each stochastic node into a differentiable function of its parameters and a random variable with fixed distribution. After refactoring, the gradients of the loss propagated by the chain rule through the graph are low variance unbiased estimators of the gradients of the expected loss. While many continuous random variables have such reparameterizations, discrete random variables lack useful reparameterizations due to the discontinuous nature of discrete states. In this work we introduce Concrete random variables---continuous relaxations of discrete random variables. The Concrete distribution is a new family of distributions with closed form densities and a simple reparameterization. Whenever a discrete stochastic node of a computation graph can be refactored into a one-hot bit representation that is treated continuously, Concrete stochastic nodes can be used with automatic differentiation to produce low-variance biased gradients of objectives (including objectives that depend on the log-probability of latent stochastic nodes) on the corresponding discrete graph. We demonstrate the effectiveness of Concrete relaxations on density estimation and structured prediction tasks using neural networks."
                },
                {
                    "title": "HyperNetworks",
                    "abstract": "This work explores hypernetworks: an approach of using one network, also known as a hypernetwork, to generate the weights for another network. We apply hypernetworks to generate adaptive weights for recurrent networks. In this case, hypernetworks can be viewed as a relaxed form of weight-sharing across layers. In our implementation, hypernetworks are are trained jointly with the main network in an end-to-end fashion. Our main result is that hypernetworks can generate non-shared weights for LSTM and achieve state-of-the-art results on a variety of sequence modelling tasks including character-level language modelling, handwriting generation and neural machine translation, challenging the weight-sharing paradigm for recurrent networks."
                },
                {
                    "title": "Introduction to Online Convex Optimization",
                    "abstract": "This monograph portrays optimization as a process. In many practical applications the environment is so complex that it is infeasible to lay out a comprehensive theoretical model and use classical algorithmic theory and mathematical optimization. It is necessary as well as beneficial to take a robust approach, by applying an optimization method that learns as one goes along, learning from experience as more aspects of the problem are observed. This view of optimization as a process has become prominent in varied fields and has led to some spectacular success in modeling and systems that are now part of our daily lives."
                },
                {
                    "title": "\u201cWhy Should I Trust You?\u201d: Explaining the Predictions of Any Classifier",
                    "abstract": "Despite widespread adoption, machine learning models remain mostly black boxes. Understanding the reasons behind predictions is, however, quite important in assessing trust, which is fundamental if one plans to take action based on a prediction, or when choosing whether to deploy a new model. Such understanding also provides insights into the model, which can be used to transform an untrustworthy model or prediction into a trustworthy one. In this work, we propose LIME, a novel explanation technique that explains the predictions of any classifier in an interpretable and faithful manner, by learning an interpretable model locally varound the prediction. We also propose a method to explain models by presenting representative individual predictions and their explanations in a non-redundant way, framing the task as a submodular optimization problem. We demonstrate the flexibility of these methods by explaining different models for text (e.g. random forests) and image classification (e.g. neural networks). We show the utility of explanations via novel experiments, both simulated and with human subjects, on various scenarios that require trust: deciding if one should trust a prediction, choosing between models, improving an untrustworthy classifier, and identifying why a classifier should not be trusted."
                },
                {
                    "title": "Active Information Acquisition",
                    "abstract": "We propose a general framework for sequential and dynamic acquisition of useful information in order to solve a particular task. While our goal could in principle be tackled by general reinforcement learning, our particular setting is constrained enough to allow more efficient algorithms. In this paper, we work under the Learning to Search framework and show how to formulate the goal of finding a dynamic information acquisition policy in that framework. We apply our formulation on two tasks, sentiment analysis and image recognition, and show that the learned policies exhibit good statistical performance. As an emergent byproduct, the learned policies show a tendency to focus on the most prominent parts of each instance and give harder instances more attention without explicitly being trained to do so."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Learning Continuous Control Policies by Stochastic Value Gradients",
                    "abstract": "We present a unified framework for learning continuous control policies using backpropagation. It supports stochastic control by treating stochasticity in the Bellman equation as a deterministic function of exogenous noise. The product is a spectrum of general policy gradient algorithms that range from model-free methods with value functions to model-based methods without value functions. We use learned models but only require observations from the environment in- stead of observations from model-predicted trajectories, minimizing the impact of compounded model errors. We apply these algorithms first to a toy stochastic control problem and then to several physics-based control problems in simulation. One of these variants, SVG(1), shows the effectiveness of learning models, value functions, and policies simultaneously in continuous domains."
                },
                {
                    "title": "Convex Optimization: Algorithms and Complexity",
                    "abstract": "This monograph presents the main complexity theorems in convex optimization and their corresponding algorithms. Starting from the fundamental theory of black-box optimization, the material progresses towards recent advances in structural optimization and stochastic optimization. Our presentation of black-box optimization, strongly influenced by Nesterov's seminal book and Nemirovski's lecture notes, includes the analysis of cutting plane methods, as well as (accelerated) gradient descent schemes. We also pay special attention to non-Euclidean settings (relevant algorithms include Frank-Wolfe, mirror descent, and dual averaging) and discuss their relevance in machine learning. We provide a gentle introduction to structural optimization with FISTA (to optimize a sum of a smooth and a simple non-smooth term), saddle-point mirror prox (Nemirovski's alternative to Nesterov's smoothing), and a concise description of interior point methods. In stochastic optimization we discuss stochastic gradient descent, mini-batches, random coordinate descent, and sublinear algorithms. We also briefly touch upon convex relaxation of combinatorial problems and the use of randomness to round solutions, as well as random walks based methods."
                },
                {
                    "title": "Stochastic Backpropagation and Approximate Inference in Deep Generative Models",
                    "abstract": "We marry ideas from deep neural networks and approximate Bayesian inference to derive a generalised class of deep, directed generative models, endowed with a new algorithm for scalable inference and learning. Our algorithm introduces a recognition model to represent approximate posterior distributions, and that acts as a stochastic encoder of the data. We develop stochastic back-propagation -- rules for back-propagation through stochastic variables -- and use this to develop an algorithm that allows for joint optimisation of the parameters of both the generative and recognition model. We demonstrate on several real-world data sets that the model generates realistic samples, provides accurate imputations of missing data and is a useful tool for high-dimensional data visualisation."
                },
                {
                    "title": "Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps",
                    "abstract": "This paper addresses the visualisation of image classification models, learnt using deep Convolutional Networks (ConvNets). We consider two visualisation techniques, based on computing the gradient of the class score with respect to the input image. The first one generates an image, which maximises the class score [Erhan et al., 2009], thus visualising the notion of the class, captured by a ConvNet. The second technique computes a class saliency map, specific to a given image and class. We show that such maps can be employed for weakly supervised object segmentation using classification ConvNets. Finally, we establish the connection between the gradient-based ConvNet visualisation methods and deconvolutional networks [Zeiler et al., 2013]."
                },
                {
                    "title": "Stochastic Approximation approach to Stochastic Programming",
                    "abstract": "In this paper we consider optimization problems where the objective function is given in a form of the expectation. A basic difficulty of solving such stochastic optimization problems is that the involved multidimensional integrals (expectations) cannot be computed with high accuracy. The aim of this paper is to compare two computational approaches based on Monte Carlo sampling techniques, namely, the Stochastic Approximation (SA) and the Sample Average Approximation (SAA) methods. Both approaches, the SA and SAA methods, have a long history. Current opinion is that the SAA method can efficiently use a specific (say linear) structure of the considered problem, while the SA approach is a crude subgradient method which often performs poorly in practice. We intend to demonstrate that a properly modified SA approach can be competitive and even significantly outperform the SAA method for a certain class of convex stochastic problems. We extend the analysis to the case of convex-concave stochastic saddle point problems, and present (in our opinion highly encouraging) results of numerical experiments."
                },
                {
                    "title": "Guided Policy Search",
                    "abstract": "Direct policy search can effectively scale to high-dimensional systems, but complex policies with hundreds of parameters often present a challenge for such methods, requiring numerous samples and often falling into poor local optima. We present a guided policy search algorithm that uses trajectory optimization to direct policy learning and avoid poor local optima. We show how differential dynamic programming can be used to generate suitable guiding samples, and describe a regularized importance sampled policy optimization that incorporates these samples into the policy search. We evaluate the method by learning neural network controllers for planar swimming, hopping, and walking, as well as simulated 3D humanoid running."
                },
                {
                    "title": "Optimal Distributed Online Prediction Using Mini-Batches",
                    "abstract": "Online prediction methods are typically presented as serial algorithms running on a single processor. However, in the age of web-scale prediction problems, it is increasingly common to encounter situations where a single processor cannot keep up with the high rate at which inputs arrive. In this work, we present the distributed mini-batch algorithm, a method of converting many serial gradient-based online prediction algorithms into distributed algorithms. We prove a regret bound for this method that is asymptotically optimal for smooth convex loss functions and stochastic inputs. Moreover, our analysis explicitly takes into account communication latencies between nodes in the distributed environment. We show how our method can be used to solve the closely-related distributed stochastic optimization problem, achieving an asymptotically linear speed-up over multiple processors. Finally, we demonstrate the merits of our approach on a web-scale online prediction problem."
                },
                {
                    "title": "An Efficient Explanation of Individual Classifications using Game Theory",
                    "abstract": "We present a general method for explaining individual predictions of classification models. The method is based on fundamental concepts from coalitional game theory and predictions are explained with contributions of individual feature values. We overcome the method's initial exponential time complexity with a sampling-based approximation. In the experimental part of the paper we use the developed method on models generated by several well-known machine learning algorithms on both synthetic and real-world data sets. The results demonstrate that the method is efficient and that the explanations are intuitive and useful."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "A Value for n-person Games",
                    "abstract": "Introduction At the foundation of the theory of games is the assumption that the players of a game can evaluate, in their utility scales, every \u201cprospect\u201d that might arise as a result of a play. In attempting to apply the theory to any field, one would normally expect to be permitted to include, in the class of \u201cprospects,\u201d the prospect of having to play a game. The possibility of evaluating games is therefore of critical importance. So long as the theory is unable to assign values to the games typically found in application, only relatively simple situations\u2014where games do not depend on other games\u2014will be susceptible to analysis and solution. In the finite theory of von Neumann and Morgenstern difficulty in evaluation persists for the \u201cessential\u201d games, and for only those. In this note we deduce a value for the \u201cessential\u201d case and examine a number of its elementary properties. We proceed from a set of three axioms, having simple intuitive interpretations, which suffice to determine the value uniquely. Our present work, though mathematically self-contained, is founded conceptually on the von Neumann\u2014Morgenstern theory up to their introduction of characteristic functions. We thereby inherit certain important underlying assumptions: (a) that utility is objective and transferable; (b) that games are cooperative affairs; (c) that games, granting (a) and (b), are adequately represented by their characteristic functions."
                },
                {
                    "title": "Characterizations of an Empirical Influence Function for Detecting Influential Cases in Regression",
                    "abstract": "Traditionally, most of the effort in fitting full rank linear regression models has centered on the study of the presence, strength and form of relationships between the measured variables. As is now well known, least squares regression computations can be strongly influenced by a few cases, and a fitted model may more accurately reflect unusual features of those cases than the overall relationships between the variables. It is of interest, therefore, for an analyst to be able to find influential cases and, based on them, make decisions concerning their usefulness in a problem at hand. Based on an empirical influence function, we discuss methodologies for assessing the influence of individual or groups of cases on a regression problem. We conclude with an example using data from the Florida Area Cumulus Experiments (FACE) on cloud seeding."
                },
                {
                    "title": "Mathematical Properties of the Banzhaf Power Index",
                    "abstract": "The Banzhaf index of power in a voting situation depends on the number of ways in which each voter can effect a \u201cswing\u201d in the outcome. It is comparable---but not actually equivalent---to the better-known Shapley-Shubik index, which depends on the number of alignments or \u201corders of support\u201d in which each voter is pivotal. This paper investigates some properties of the Banzhaf index, the main topics being its derivation from axioms and its behavior in weighted-voting models when the number of small voters tends to infinity. These matters have previously been studied from the Shapley-Shubik viewpoint, but the present work reveals some striking differences between the two indices. The paper also attempts to promote better communication and less duplication of mathematical effort by identifying and describing several other theories, formally equivalent to Banzhaf\u2019s, that are found in fields ranging from sociology to electrical engineering. An extensive bibliography is provided."
                },
                {
                    "title": "Tutorial on amortized optimization for learning to optimize over continuous domains",
                    "abstract": "Optimization is a ubiquitous modeling tool that is often deployed in settings that repeatedly solve similar instances of the same problem. Amortized optimization methods use learning to predict the solutions to problems in these settings. This leverages the shared structure between similar problem instances. In this tutorial, we will discuss the key design choices behind amortized optimization, roughly categorizing 1) models into fully-amortized and semi-amortized approaches, and 2) learning methods into regression-based and objectivebased. We then view existing applications through these foundations to draw connections between them, including for manifold optimization, variational inference, sparse coding, meta-learning, control, reinforcement learning, convex optimization, and deep equilibrium networks. This framing enables us easily see, for example, that the amortized inference in variational autoencoders is conceptually identical to value gradients in control and reinforcement learning as they both use fully-amortized models with a objective-based loss. The source code for this tutorial is available at: https://github.com/facebookresearch/amortized-optimization-tutorial ar X iv :2 20 2. 00 66 5v 1 [ cs .L G ] 1 F eb 2 02 2"
                },
                {
                    "title": "Data Banzhaf: A Data Valuation Framework with Maximal Robustness to Learning Stochasticity",
                    "abstract": ","
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "The Matrix Cookbook",
                    "abstract": "Acknowledgements: We would like to thank the following for contributions and suggestions: Bill Baxter, Brian Templeton, Christian Rishoj, Christian Schroppel Douglas L. Theobald, Esben Hoegh-Rasmussen, Glynne Casteel, Jan Larsen, Jun Bin Gao, Jurgen Struckmeier, Kamil Dedecius, Korbinian Strimmer, Lars Christiansen, Lars Kai Hansen, Leland Wilkinson, Liguo He, Loic Thibaut, Miguel Barao, Ole Winther, Pavel Sakov, Stephan Hattinger, Vasile Sima, Vincent Rabaud, Zhaoshui He. We would also like thank The Oticon Foundation for funding our PhD studies."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively train models for amortized computation in machine learning when faced with noisy labels?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a common challenge in machine learning: the reliance on high-quality labeled data. By developing methods that can tolerate noisy labels, we can enhance the robustness and applicability of machine learning models across various domains, including those where obtaining clean labels is difficult or expensive. This advancement could lead to significant improvements in areas such as computer vision, natural language processing, and medical diagnosis, ultimately fostering more practical applications and driving future research towards more resilient learning algorithms.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent difficulty of training models when the ground truth solutions are not available or are corrupted by noise. Naive approaches may fail because they often assume access to clean labels, which can lead to overfitting or incorrect generalization. Additionally, the computational cost of generating accurate labels for each instance can be prohibitive, making it impractical to train models effectively. The technical obstacles include designing a robust training framework that can learn from noisy data while still achieving high performance, as well as ensuring that the amortization process does not amplify the noise present in the labels.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on scenarios with clean labels or has not adequately addressed the complexities introduced by noisy labels in amortized computation. Existing solutions may lack the necessary theoretical foundations or practical methodologies to handle noise effectively. Barriers such as the absence of robust algorithms that can generalize well under label noise and the limited understanding of how noise impacts the learning process have hindered progress. Our approach differs by explicitly proving that amortization can tolerate training with noisy labels, thus providing a new perspective and methodology that builds upon and improves prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a deep neural network model that learns to approximate the per-context output while being trained on datasets with noisy labels. We will utilize a combination of Monte Carlo estimators and optimization techniques to define the output function. The dataset will consist of various tasks where the ground truth is either expensive to obtain or inherently noisy. We will evaluate the model's performance using metrics such as accuracy and robustness to label noise. The expected outcomes"
            }
        },
        "author_data": {
            "6e8f27f6-bd1d-477e-b9c4-cc3eb08e688d": {
                "pk": "6e8f27f6-bd1d-477e-b9c4-cc3eb08e688d",
                "name": "Ian Covert",
                "collaborators": [
                    "Su-In Lee",
                    "Scott Lundberg",
                    "Tatsunori Hashimoto",
                    "James Zou",
                    "Chanwoo Kim",
                    "Wenlong Ji",
                    "Tony Sun",
                    "Soham Gadgil",
                    "Chris Lin",
                    "Ethan Weinberger"
                ],
                "domain": [
                    "Machine Learning",
                    "Model Interpretability",
                    "Feature Selection",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Understanding Global Feature Contributions With Additive Importance Measures",
                        "abstract": "Understanding the inner workings of complex machine learning models is a long-standing problem and most recent research has focused on local interpretability. To assess the role of individual input features in a global sense, we explore the perspective of defining feature importance through the predictive power associated with each feature. We introduce two notions of predictive power (model-based and universal) and formalize this approach with a framework of additive importance measures, which unifies numerous methods in the literature. We then propose SAGE, a model-agnostic method that quantifies predictive power while accounting for feature interactions. Our experiments show that SAGE can be calculated efficiently and that it assigns more accurate importance values than other methods."
                    },
                    {
                        "title": "Feature Removal Is a Unifying Principle for Model Explanation Methods",
                        "abstract": "Researchers have proposed a wide variety of model explanation approaches, but it remains unclear how most methods are related or when one method is preferable to another. We examine the literature and find that many methods are based on a shared principle of explaining by removing - essentially, measuring the impact of removing sets of features from a model. These methods vary in several respects, so we develop a framework for removal-based explanations that characterizes each method along three dimensions: 1) how the method removes features, 2) what model behavior the method explains, and 3) how the method summarizes each feature's influence. Our framework unifies 26 existing methods, including several of the most widely used approaches (SHAP, LIME, Meaningful Perturbations, permutation tests). Exposing the fundamental similarities between these methods empowers users to reason about which tools to use, and suggests promising directions for ongoing model explainability research."
                    },
                    {
                        "title": "Improving KernelSHAP: Practical Shapley Value Estimation via Linear Regression",
                        "abstract": "The Shapley value concept from cooperative game theory has become a popular technique for interpreting ML models, but efficiently estimating these values remains challenging, particularly in the model-agnostic setting. Here, we revisit the idea of estimating Shapley values via linear regression to understand and improve upon this approach. By analyzing the original KernelSHAP alongside a newly proposed unbiased version, we develop techniques to detect its convergence and calculate uncertainty estimates. We also find that the original version incurs a negligible increase in bias in exchange for significantly lower variance, and we propose a variance reduction technique that further accelerates the convergence of both estimators. Finally, we develop a version of KernelSHAP for stochastic cooperative games that yields fast new estimators for two global explanation methods."
                    },
                    {
                        "title": "Learning to Estimate Shapley Values with Vision Transformers",
                        "abstract": "Transformers have become a default architecture in computer vision, but understanding what drives their predictions remains a challenging problem. Current explanation approaches rely on attention values or input gradients, but these provide a limited view of a model's dependencies. Shapley values offer a theoretically sound alternative, but their computational cost makes them impractical for large, high-dimensional models. In this work, we aim to make Shapley values practical for vision transformers (ViTs). To do so, we first leverage an attention masking approach to evaluate ViTs with partial information, and we then develop a procedure to generate Shapley value explanations via a separate, learned explainer model. Our experiments compare Shapley values to many baseline methods (e.g., attention rollout, GradCAM, LRP), and we find that our approach provides more accurate explanations than existing methods for ViTs."
                    },
                    {
                        "title": "Explaining by Removing: A Unified Framework for Model Explanation",
                        "abstract": "Researchers have proposed a wide variety of model explanation approaches, but it remains unclear how most methods are related or when one method is preferable to another. We describe a new unified class of methods, removal-based explanations, that are based on the principle of simulating feature removal to quantify each feature's influence. These methods vary in several respects, so we develop a framework that characterizes each method along three dimensions: 1) how the method removes features, 2) what model behavior the method explains, and 3) how the method summarizes each feature's influence. Our framework unifies 26 existing methods, including several of the most widely used approaches: SHAP, LIME, Meaningful Perturbations, and permutation tests. This newly understood class of explanation methods has rich connections that we examine using tools that have been largely overlooked by the explainability literature. To anchor removal-based explanations in cognitive psychology, we show that feature removal is a simple application of subtractive counterfactual reasoning. Ideas from cooperative game theory shed light on the relationships and trade-offs among different methods, and we derive conditions under which all removal-based explanations have information-theoretic interpretations. Through this analysis, we develop a unified framework that helps practitioners better understand model explanation tools, and that offers a strong theoretical foundation upon which future explainability research can build."
                    },
                    {
                        "title": "Scaling Laws for the Value of Individual Data Points in Machine Learning",
                        "abstract": "Recent works have shown that machine learning models improve at a predictable rate with the total amount of training data, leading to scaling laws that describe the relationship between error and dataset size. These scaling laws can help design a model's training dataset, but they typically take an aggregate view of the data by only considering the dataset's size. We introduce a new perspective by investigating scaling behavior for the value of individual data points: we find that a data point's contribution to model's performance shrinks predictably with the size of the dataset in a log-linear manner. Interestingly, there is significant variability in the scaling exponent among different data points, indicating that certain points are more valuable in small datasets while others are relatively more useful as a part of large datasets. We provide learning theory to support our scaling law, and we observe empirically that it holds across diverse model classes. We further propose a maximum likelihood estimator and an amortized estimator to efficiently learn the individualized scaling behaviors from a small number of noisy observations per data point. Using our estimators, we provide insights into factors that influence the scaling behavior of different data points. Finally, we demonstrate applications of the individualized scaling laws to data valuation and data subset selection. Overall, our work represents a first step towards understanding and utilizing scaling properties for the value of individual data points."
                    },
                    {
                        "title": "Locality Alignment Improves Vision-Language Models",
                        "abstract": "Vision language models (VLMs) have seen growing adoption in recent years, but many still struggle with basic spatial reasoning errors. We hypothesize that this is due to VLMs adopting pre-trained vision backbones, specifically vision transformers (ViTs) trained with image-level supervision and minimal inductive biases. Such models may fail to encode the class contents at each position in the image, and our goal is to resolve this by ensuring that the vision backbone effectively captures both local and global image semantics. Our main insight is that we do not require new supervision to learn this capability -- pre-trained models contain significant knowledge of local semantics that we can extract and use for scalable self-supervision. We propose a new efficient post-training stage for ViTs called locality alignment and a novel fine-tuning procedure called MaskEmbed that uses a masked reconstruction loss to learn semantic contributions for each image patch. We first evaluate locality alignment with a vision-only benchmark, finding that it improves a model's performance at a patch-level semantic segmentation task, especially for strong backbones trained with image-caption pairs (e.g., CLIP and SigLIP). We then train a series of VLMs with and without locality alignment, and show that locality-aligned backbones improve performance across a range of benchmarks, particularly ones that involve spatial understanding (e.g., RefCOCO, OCID-Ref, TallyQA, VSR, AI2D). Overall, we demonstrate that we can efficiently learn local semantic extraction via a locality alignment stage, and that this procedure complements existing VLM training recipes that use off-the-shelf vision backbones."
                    },
                    {
                        "title": "Estimating Conditional Mutual Information for Dynamic Feature Selection",
                        "abstract": "Dynamic feature selection, where we sequentially query features to make accurate predictions with a minimal budget, is a promising paradigm to reduce feature acquisition costs and provide transparency into a model's predictions. The problem is challenging, however, as it requires both predicting with arbitrary feature sets and learning a policy to identify valuable selections. Here, we take an information-theoretic perspective and prioritize features based on their mutual information with the response variable. The main challenge is implementing this policy, and we design a new approach that estimates the mutual information in a discriminative rather than generative fashion. Building on our approach, we then introduce several further improvements: allowing variable feature budgets across samples, enabling non-uniform feature costs, incorporating prior information, and exploring modern architectures to handle partial inputs. Our experiments show that our method provides consistent gains over recent methods across a variety of datasets."
                    },
                    {
                        "title": "On the Robustness of Removal-Based Feature Attributions",
                        "abstract": "To explain predictions made by complex machine learning models, many feature attribution methods have been developed that assign importance scores to input features. Some recent work challenges the robustness of these methods by showing that they are sensitive to input and model perturbations, while other work addresses this issue by proposing robust attribution methods. However, previous work on attribution robustness has focused primarily on gradient-based feature attributions, whereas the robustness of removal-based attribution methods is not currently well understood. To bridge this gap, we theoretically characterize the robustness properties of removal-based feature attributions. Specifically, we provide a unified analysis of such methods and derive upper bounds for the difference between intact and perturbed attributions, under settings of both input and model perturbations. Our empirical results on synthetic and real-world data validate our theoretical results and demonstrate their practical implications, including the ability to increase attribution robustness by improving the model's Lipschitz regularity."
                    },
                    {
                        "title": "Feature Selection in the Contrastive Analysis Setting",
                        "abstract": "Contrastive analysis (CA) refers to the exploration of variations uniquely enriched in a target dataset as compared to a corresponding background dataset generated from sources of variation that are irrelevant to a given task. For example, a biomedical data analyst may wish to find a small set of genes to use as a proxy for variations in genomic data only present among patients with a given disease (target) as opposed to healthy control subjects (background). However, as of yet the problem of feature selection in the CA setting has received little attention from the machine learning community. In this work we present contrastive feature selection (CFS), a method for performing feature selection in the CA setting. We motivate our approach with a novel information-theoretic analysis of representation learning in the CA setting, and we empirically validate CFS on a semi-synthetic dataset and four real-world biomedical datasets. We find that our method consistently outperforms previously proposed state-of-the-art supervised and fully unsupervised feature selection methods not designed for the CA setting. An open-source implementation of our method is available at https://github.com/suinleelab/CFS."
                    },
                    {
                        "title": "Disrupting Model Training with Adversarial Shortcuts",
                        "abstract": "When data is publicly released for human consumption, it is unclear how to prevent its unauthorized usage for machine learning purposes. Successful model training may be preventable with carefully designed dataset modifications, and we present a proof-of-concept approach for the image classification setting. We propose methods based on the notion of adversarial shortcuts, which encourage models to rely on non-robust signals rather than semantic features, and our experiments demonstrate that these measures successfully prevent deep learning models from achieving high accuracy on real, unmodified data examples."
                    },
                    {
                        "title": "Neural Granger Causality",
                        "abstract": "While most classical approaches to Granger causality detection assume linear dynamics, many interactions in real-world applications, like neuroscience and genomics, are inherently nonlinear. In these cases, using linear models may lead to inconsistent estimation of Granger causal interactions. We propose a class of nonlinear methods by applying structured multilayer perceptrons (MLPs) or recurrent neural networks (RNNs) combined with sparsity-inducing penalties on the weights. By encouraging specific sets of weights to be zero--in particular, through the use of convex group-lasso penalties--we can extract the Granger causal structure. To further contrast with traditional approaches, our framework naturally enables us to efficiently capture long-range dependencies between series either via our RNNs or through an automatic lag selection in the MLP. We show that our neural Granger causality methods outperform state-of-the-art nonlinear Granger causality methods on the DREAM3 challenge data. This data consists of nonlinear gene expression and regulation time courses with only a limited number of time points. The successes we show in this challenging dataset provide a powerful example of how deep learning can be useful in cases that go beyond prediction on large datasets. We likewise illustrate our methods in detecting nonlinear interactions in a human motion capture dataset."
                    },
                    {
                        "title": "Learning to Maximize Mutual Information for Dynamic Feature Selection",
                        "abstract": "Feature selection helps reduce data acquisition costs in ML, but the standard approach is to train models with static feature subsets. Here, we consider the dynamic feature selection (DFS) problem where a model sequentially queries features based on the presently available information. DFS is often addressed with reinforcement learning, but we explore a simpler approach of greedily selecting features based on their conditional mutual information. This method is theoretically appealing but requires oracle access to the data distribution, so we develop a learning approach based on amortized optimization. The proposed method is shown to recover the greedy policy when trained to optimality, and it outperforms numerous existing feature selection methods in our experiments, thus validating it as a simple but powerful approach for this problem."
                    },
                    {
                        "title": "What does a platypus look like? Generating customized prompts for zero-shot image classification",
                        "abstract": "Open-vocabulary models are a promising new paradigm for image classification. Unlike traditional classification models, open-vocabulary models classify among any arbitrary set of categories specified with natural language during inference. This natural language, called \"prompts\", typically consists of a set of hand-written templates (e.g., \"a photo of a {}\") which are completed with each of the category names. This work introduces a simple method to generate higher accuracy prompts, without relying on any explicit knowledge of the task domain and with far fewer hand-constructed sentences. To achieve this, we combine open-vocabulary models with large language models (LLMs) to create Customized Prompts via Language models (CuPL, pronounced \"couple\"). In particular, we leverage the knowledge contained in LLMs in order to generate many descriptive sentences that contain important discriminating characteristics of the image categories. This allows the model to place a greater importance on these regions in the image when making predictions. We find that this straightforward and general approach improves accuracy on a range of zero-shot image classification benchmarks, including over one percentage point gain on ImageNet. Finally, this simple baseline requires no additional training and remains completely zero-shot. Code available at https://github.com/sarahpratt/CuPL."
                    },
                    {
                        "title": "FastSHAP: Real-Time Shapley Value Estimation",
                        "abstract": "Shapley values are widely used to explain black-box models, but they are costly to calculate because they require many model evaluations. We introduce FastSHAP, a method for estimating Shapley values in a single forward pass using a learned explainer model. FastSHAP amortizes the cost of explaining many inputs via a learning approach inspired by the Shapley value's weighted least squares characterization, and it can be trained using standard stochastic gradient optimization. We compare FastSHAP to existing estimation approaches, revealing that it generates high-quality explanations with orders of magnitude speedup."
                    },
                    {
                        "title": "Temporal Graph Convolutional Networks for Automatic Seizure Detection",
                        "abstract": "Seizure detection from EEGs is a challenging and time consuming clinical problem that would benefit from the development of automated algorithms. EEGs can be viewed as structural time series, because they are multivariate time series where the placement of leads on a patient's scalp provides prior information about the structure of interactions. Commonly used deep learning models for time series don't offer a way to leverage structural information, but this would be desirable in a model for structural time series. To address this challenge, we propose the temporal graph convolutional network (TGCN), a model that leverages structural information and has relatively few parameters. TGCNs apply feature extraction operations that are localized and shared over both time and space, thereby providing a useful inductive bias in tasks where one expects similar features to be discriminative across the different sequences. In our experiments we focus on metrics that are most important to seizure detection, and demonstrate that TGCN matches the performance of related models that have been shown to be state of the art in other tasks. Additionally, we investigate interpretability advantages of TGCN by exploring approaches for helping clinicians determine when precisely seizures occur, and the parts of the brain that are most involved."
                    }
                ]
            },
            "2138ea3f-3c80-40ba-9036-f99d684f0eaf": {
                "pk": "2138ea3f-3c80-40ba-9036-f99d684f0eaf",
                "name": "Chanwoo Kim",
                "collaborators": [
                    "Jiaxin Jin",
                    "Donghyun Lee",
                    "Hongxu Chen",
                    "Yunbai Cao",
                    "Juhi Jang",
                    "Ian Tice",
                    "Joonhyun La",
                    "Yongjoon Lee"
                ],
                "domain": [
                    "Kinetic Theory",
                    "Boltzmann Equation",
                    "Vlasov-Poisson",
                    "Fluid Dynamics"
                ],
                "publications": [
                    {
                        "title": "Formation and Propagation of Discontinuity for Boltzmann Equation in Non-Convex Domains",
                        "abstract": "The formation and propagation of singularities for Boltzmann equation in bounded domains has been an important question in numerical studies as well as in theoretical studies. Consider the nonlinear Boltzmann solution near Maxwellians under in-flow, diffuse, or bounce-back boundary conditions. We demonstrate that discontinuity is created at the non-convex part of the grazing boundary, then propagates only along the forward characteristics inside the domain before it hits on the boundary again."
                    },
                    {
                        "title": "Boltzmann Equation with a Large Potential in a Periodic Box",
                        "abstract": "The stability of the Maxwellian of the Boltzmann equation with a large amplitude external potential $\\Phi$ has been an important open problem. In this paper, we resolve this problem with a large $C3-$potential in a periodic box $\\mathbb{T}^d$, $d \\geq 3$. We use [1] in $L^p-L^{\\infty}$ framework to establish the well-posedness and the $L^{\\infty}-$stability of the Maxwellian $\\mu_E(x,v)=\\exp\\{-\\frac{|v|^2}{2}-\\Phi(x)\\}$."
                    },
                    {
                        "title": "Nonlinear asymptotic stability of inhomogeneous steady solutions to boundary problems of Vlasov-Poisson equation",
                        "abstract": "We consider an ensemble of mass collisionless particles, which interact mutually either by an attraction of Newton's law of gravitation or by an electrostatic repulsion of Coulomb's law, under a background downward gravity in a horizontally-periodic 3D half-space, whose inflow distribution at the boundary is prescribed. We investigate a nonlinear asymptotic stability of its generic steady states in the dynamical kinetic PDE theory of the Vlasov-Poisson equations. We construct Lipschitz continuous space-inhomogeneous steady states and establish exponentially fast asymptotic stability of these steady states with respect to small perturbation in a weighted Sobolev topology. In this proof, we crucially use the Lipschitz continuity in the velocity of the steady states. Moreover, we establish well-posedness and regularity estimates for both steady and dynamic problems."
                    },
                    {
                        "title": "The Boltzmann equation with specular boundary condition in convex domains",
                        "abstract": "We establish the global-wellposedness and stability of the Boltzmann equation with the specular reflection boundary condition in general smooth convex domains when an initial datum is close to the Maxwellian with or without a small external potential. In particular, we have completely solved the long standing open problem after an announcement of [20] in 1977."
                    },
                    {
                        "title": "Regularity of Stationary Boltzmann equation in Convex Domains",
                        "abstract": "Higher regularity estimate has been a challenging question for the Boltzmann equation in bounded domains. Indeed, it is well-known to have \"the non-existence of a second order derivative at the boundary\" in [15] even for symmetric convex domains such as a disk or sphere. In this paper we answer this question in the affirmative by constructing the $C^{1,\\beta}$ solutions away from the grazing boundary, for any $\\beta<1$, to the stationary Boltzmann equation with the non-isothermal diffuse boundary condition in strictly convex domains, as long as a smooth wall temperature has small fluctuation pointwisely."
                    },
                    {
                        "title": "Damping of kinetic transport equation with diffuse boundary condition",
                        "abstract": "We prove that exponential moments of a fluctuation of the pure transport equation decay pointwisely almost as fast as $t^{-3}$ when the domain is any general strictly convex subset of $\\mathbb{R}^3$ with the smooth boundary of the diffuse boundary condition. We prove the theorem by establishing a novel $L^1$-$L^\\infty$ framework via stochastic cycles."
                    },
                    {
                        "title": "Macroscopic estimate of the linear Boltzmann and Landau equations with Specular reflection boundary",
                        "abstract": "In this short note, we prove an $L^6$-control of the macroscopic part of the linear Boltzmann and Landau equations. This result is an extension of the test function method of Esposito-Guo-Kim-Marra~\\cite{EGKM}\\cite{EGKM2} to the specular reflection boundary condition, in which we crucially used the Korn's inequality \\cite{DV2} and the system of symmetric Poisson equations \\cite{Bernou}."
                    },
                    {
                        "title": "On Some Recent Progress in the Vlasov-Poisson-Boltzmann System with Diffuse Reflection Boundary",
                        "abstract": "We discuss some recent development on the Vlasov-Poisson-Boltzmann system in bounded domains with diffuse reflection boundary condition. In addition we present a new regularity result when the particles are surrounded by conductor boundary."
                    },
                    {
                        "title": "Glassey-Strauss representation of Vlasov-Maxwell systems in a Half Space",
                        "abstract": "Following closely the classical works [8]-[12] by Glassey, Strauss, and Schaeffer, we present a version of the Glassey-Strauss representation for the Vlasov-Maxwell systems in a 3D half space when the boundary is the perfect conductor."
                    },
                    {
                        "title": "Boundary effect under 2D Newtonian gravity potential in the phase space",
                        "abstract": "We study linear two-half dimensional Vlasov equations under the logarithmic gravity potential in the half space of diffuse reflection boundary. We prove decay-in-time of the exponential moments with a polynomial rate, which depends on the base logarithm."
                    },
                    {
                        "title": "Incompressible Euler limit from Boltzmann equation with Diffuse Boundary Condition for Analytic data",
                        "abstract": "A rigorous derivation of the incompressible Euler equations with the no-penetration boundary condition from the Boltzmann equation with the diffuse reflection boundary condition has been a challenging open problem. We settle this open question in the affirmative when the initial data of fluid are well-prepared in a real analytic space, in 3D half space. As a key of this advance we capture the Navier-Stokes equations of $$\\textit{viscosity}   \\sim \\frac{\\textit{Knudsen number}}{\\textit{Mach number}}$$ satisfying the no-slip boundary condition, as an $\\textit{intermediary}$ approximation of the Euler equations through a new Hilbert-type expansion of Boltzmann equation with the diffuse reflection boundary condition. Aiming to justify the approximation we establish a novel quantitative $L^p$-$L^\\infty$ estimate of the Boltzmann perturbation around a local Maxwellian of such viscous approximation, along with the commutator estimates and the integrability gain of the hydrodynamic part in various spaces; we also establish direct estimates of the Navier-Stokes equations in higher regularity with the aid of the initial-boundary and boundary layer weights using a recent Green's function approach. The incompressible Euler limit follows as a byproduct of our framework."
                    },
                    {
                        "title": "Lipschitz continuous solutions of the Vlasov-Maxwell systems with a conductor boundary condition",
                        "abstract": "We consider relativistic plasma particles subjected to an external gravitation force in a $3$D half space whose boundary is a perfect conductor. When the mean free path is much bigger than the variation of electromagnetic fields, the collision effect is negligible. As an effective PDE, we study the relativistic Vlasov-Maxwell system and its local-in-time unique solvability in the space-time locally Lipschitz space, for several basic mesoscopic (kinetic) boundary conditions: the inflow, diffuse, and specular reflection boundary conditions. We construct weak solutions to these initial-boundary value problems and study their locally Lipschitz continuity with the aid of a weight function depending on the solutions themselves. Finally, we prove the uniqueness of a solution, by using regularity estimate and realizing the Gauss's law at the boundary within Lipschitz continuous space."
                    },
                    {
                        "title": "Dynamics and stability of surfactant-driven surface waves",
                        "abstract": "In this paper we consider a layer of incompressible viscous fluid lying above a flat periodic surface in a uniform gravitational field. The upper boundary of the fluid is free and evolves in time. We assume that a mass of surfactants resides on the free surface and evolves in time with the fluid. The surfactants dynamics couple to the fluid dynamics by adjusting the surface tension coefficient on the interface and also through tangential Marangoni stresses caused by gradients in surfactant concentration. We prove that small perturbations of equilibria give rise to global-in-time solutions in an appropriate functional space, and we prove that the solutions return to equilibrium exponentially fast. In particular this proves the asymptotic stability of equilibria."
                    },
                    {
                        "title": "Gradient Decay in the Boltzmann theory of Non-isothermal boundary",
                        "abstract": "We consider the Boltzmann equation in convex domain with non-isothermal boundary of diffuse reflection. For both unsteady/steady problems, we construct solutions belong to $W^{1,p}_x$ for any $p<3$. We prove that the unsteady solution converges to the steady solution in the same Sobolev space exponentially fast as $t \\rightarrow \\infty$."
                    },
                    {
                        "title": "Decay of the Boltzmann equation with the specular boundary condition in non-convex cylindrical domains",
                        "abstract": "A basic question about the existence and stability of the Boltzmann equation in general non-convex domain with the specular reflection boundary condition has been widely open. In this paper, we consider cylindrical domains whose cross sections are general non-convex analytic planar domain. We establish the global-wellposedness and asymptotic stability of the Boltzmann equation with the specular reflection boundary condition in such domains. Our method consists of sharp classification of billiard trajectories which bounce infinitely many times or hit the boundary tangentially at some moment, and a delicate construction of an $\\e$-tubular neighborhood of such trajectories. Analyticity of the boundary is crucially used. Away from such $\\e$-tubular neighborhood, we control the number of bounces of trajectories and its' distance from singular sets in a uniform fashion. The worst case, sticky grazing set, can be excluded by cutting off small portion of the temporal integration. Finally we apply a method of \\cite{KimLee} by the authors and achieve a pointwise estimate of the Boltzmann solutions."
                    },
                    {
                        "title": "H\u00f6lder regularity of the Boltzmann equation past an obstacle",
                        "abstract": "Regularity and singularity of the solutions according to the shape of domains is a challenging research theme in the Boltzmann theory (\\cite{Kim11,GKTT1}). In this paper, we prove an H\\\"older regularity in $C^{0,\\frac{1}{2}-}_{x,v}$ for the Boltzmann equation of the hard-sphere molecule, which undergoes the elastic reflection in the intermolecular collision and the contact with the boundary of a convex obstacle. In particular, this H\\\"older regularity result is a stark contrast to the case of other physical boundary conditions (such as the diffuse reflection boundary condition and in-flow boundary condition), for which the solutions of the Boltzmann equation develop discontinuity in a codimension 1 subset (\\cite{Kim11}), and therefore the best possible regularity is BV, which has been proved in \\cite{GKTT2}."
                    },
                    {
                        "title": "Vorticity convergence from Boltzmann to 2D incompressible Euler equations below Yudovich class",
                        "abstract": "It is challenging to perform a multiscale analysis of mesoscopic systems exhibiting singularities at the macroscopic scale. In this paper, we study the hydrodynamic limit of the Boltzmann equations $$\\mathrm{St} \\partial_t F + v\\cdot \\nabla_x F = \\frac{1}{\\mathrm{Kn}} Q(F ,F ) $$ toward the singular solutions of 2D incompressible Euler equations whose vorticity is unbounded $$\\partial_t u + u \\cdot \\nabla_x u + \\nabla_x p = 0,\\text{div }u =0.$$ We obtain a microscopic description of the singularity through the so-called kinetic vorticity and understand its behavior in the vicinity of the macroscopic singularity. As a consequence of our new analysis, we settle affirmatively an open problem of the hydrodynamic limit toward Lagrangian solutions of the 2D incompressible Euler equation whose vorticity is unbounded ($\\omega \\in L^\\mathfrak{p}$ for any fixed $1 \\leq \\mathfrak{p} < \\infty$). Moreover, we prove the convergence of kinetic vorticities toward the vorticity of the Lagrangian solution of the Euler equation. In particular, we obtain the rate of convergence when the vorticity blows up moderately in $L^\\mathfrak{p}$ as $\\mathfrak{p} \\rightarrow \\infty$ (localized Yudovich class)."
                    },
                    {
                        "title": "Exponential Mixing of Vlasov equations under the effect of Gravity and Boundary",
                        "abstract": "In this paper, we study exponentially fast mixing induced/enhanced by gravity and stochastic boundary in the kinetic theory of Vlasov equations. We consider the Vlasov equations with and without a vertical magnetic field inside a horizontally-periodic 3D half-space equipped with a non-isothermal diffusive reflection boundary condition of bounded continuous boundary temperature at the bottom. We construct both stationary solutions and global-in-time dynamical solutions in $L^\\infty$. We prove that moments of a dynamical fluctuation around the steady solutions decay exponentially fast in $L^\\infty$. As a key of this proof, we establish a uniform bound of so-called residual measures independently of the bouncing number of stochastic characteristics, by constructing a continuous stationary outgoing boundary flux which is strictly positive almost everywhere."
                    },
                    {
                        "title": "Asymptotic stability of 3D relativistic collisionless plasma states in ambient magnetic fields with a boundary",
                        "abstract": "Motivated by the stellar wind ejected from the upper atmosphere (Corona) of a star, we explore a boundary problem of the two-species nonlinear relativistic Vlasov-Poisson systems in the 3D half space in the presence of a constant vertical magnetic field and strong background gravity. We allow species to have different mass and charge (as proton and electron, for example). As the main result, we construct stationary solutions and establish their nonlinear dynamical asymptotic stability in time and space."
                    },
                    {
                        "title": "Wave-U-Mamba: An End-To-End Framework For High-Quality And Efficient Speech Super Resolution",
                        "abstract": "Speech Super-Resolution (SSR) is a task of enhancing low-resolution speech signals by restoring missing high-frequency components. Conventional approaches typically reconstruct log-mel features, followed by a vocoder that generates high-resolution speech in the waveform domain. However, as log-mel features lack phase information, this can result in performance degradation during the reconstruction phase. Motivated by recent advances with Selective State Spaces Models (SSMs), we propose a method, referred to as Wave-U-Mamba that directly performs SSR in time domain. In our comparative study, including models such as WSRGlow, NU-Wave 2, and AudioSR, Wave-U-Mamba demonstrates superior performance, achieving the lowest Log-Spectral Distance (LSD) across various low-resolution sampling rates, ranging from 8 kHz to 24 kHz. Additionally, subjective human evaluations, scored using Mean Opinion Score (MOS) reveal that our method produces SSR with natural and human-like quality. Furthermore, Wave-U-Mamba achieves these results while generating high-resolution speech over nine times faster than baseline models on a single A100 GPU, with parameter sizes less than 2% of those in the baseline models."
                    }
                ]
            },
            "5b7d0c6b-52ca-446a-9263-fcdb330e0ba6": {
                "pk": "5b7d0c6b-52ca-446a-9263-fcdb330e0ba6",
                "name": "Su-In Lee",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "9cd73279-566c-4d0e-8f47-2b1bfc27a514": {
                "pk": "9cd73279-566c-4d0e-8f47-2b1bfc27a514",
                "name": "James Zou",
                "collaborators": [
                    "Abubakar Abid",
                    "Amirata Ghorbani",
                    "Rong Ge",
                    "Yongchan Kwon",
                    "James Wexler",
                    "Been Kim",
                    "Daniel Russo",
                    "Ruishan Liu",
                    "Bryan He",
                    "S. Matthew Weinberg"
                ],
                "domain": [
                    "Interpretability",
                    "Machine Learning",
                    "Data Valuation",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Towards Automatic Concept-based Explanations",
                        "abstract": "Interpretability has become an important topic of research as more machine learning (ML) models are deployed and widely used to make important decisions.   Most of the current explanation methods provide explanations through feature importance scores, which identify features that are important for each individual input. However, how to systematically summarize and interpret such per sample feature importance scores itself is challenging. In this work, we propose principles and desiderata for \\emph{concept} based explanation, which goes beyond per-sample features to identify higher-level human-understandable concepts that apply across the entire dataset. We develop a new algorithm, ACE, to automatically extract visual concepts. Our systematic experiments demonstrate that \\alg discovers concepts that are human-meaningful, coherent and important for the neural network's predictions."
                    },
                    {
                        "title": "How much does your data exploration overfit? Controlling bias via information usage",
                        "abstract": "Modern data is messy and high-dimensional, and it is often not clear a priori what are the right questions to ask. Instead, the analyst typically needs to use the data to search for interesting analyses to perform and hypotheses to test. This is an adaptive process, where the choice of analysis to be performed next depends on the results of the previous analyses on the same data. Ultimately, which results are reported can be heavily influenced by the data. It is widely recognized that this process, even if well-intentioned, can lead to biases and false discoveries, contributing to the crisis of reproducibility in science. But while %the adaptive nature of exploration any data-exploration renders standard statistical theory invalid, experience suggests that different types of exploratory analysis can lead to disparate levels of bias, and the degree of bias also depends on the particulars of the data set. In this paper, we propose a general information usage framework to quantify and provably bound the bias and other error metrics of an arbitrary exploratory analysis. We prove that our mutual information based bound is tight in natural settings, and then use it to give rigorous insights into when commonly used procedures do or do not lead to substantially biased estimation. Through the lens of information usage, we analyze the bias of specific exploration procedures such as filtering, rank selection and clustering. Our general framework also naturally motivates randomization techniques that provably reduces exploration bias while preserving the utility of the data analysis. We discuss the connections between our approach and related ideas from differential privacy and blinded data analysis, and supplement our results with illustrative simulations."
                    },
                    {
                        "title": "Intersecting Faces: Non-negative Matrix Factorization With New Guarantees",
                        "abstract": "Non-negative matrix factorization (NMF) is a natural model of admixture and is widely used in science and engineering. A plethora of algorithms have been developed to tackle NMF, but due to the non-convex nature of the problem, there is little guarantee on how well these methods work. Recently a surge of research have focused on a very restricted class of NMFs, called separable NMF, where provably correct algorithms have been developed. In this paper, we propose the notion of subset-separable NMF, which substantially generalizes the property of separability. We show that subset-separability is a natural necessary condition for the factorization to be unique or to have minimum volume. We developed the Face-Intersect algorithm which provably and efficiently solves subset-separable NMF under natural conditions, and we prove that our algorithm is robust to small noise. We explored the performance of Face-Intersect on simulations and discuss settings where it empirically outperformed the state-of-art methods. Our work is a step towards finding provably correct algorithms that solve large classes of NMF problems."
                    },
                    {
                        "title": "Rich Component Analysis",
                        "abstract": "In many settings, we have multiple data sets (also called views) that capture different and overlapping aspects of the same phenomenon. We are often interested in finding patterns that are unique to one or to a subset of the views. For example, we might have one set of molecular observations and one set of physiological observations on the same group of individuals, and we want to quantify molecular patterns that are uncorrelated with physiology. Despite being a common problem, this is highly challenging when the correlations come from complex distributions. In this paper, we develop the general framework of Rich Component Analysis (RCA) to model settings where the observations from different views are driven by different sets of latent components, and each component can be a complex, high-dimensional distribution. We introduce algorithms based on cumulant extraction that provably learn each of the components without having to model the other components. We show how to integrate RCA with stochastic gradient descent into a meta-algorithm for learning general models, and demonstrate substantial improvement in accuracy on several synthetic and real datasets in both supervised and unsupervised tasks. Our method makes it possible to learn latent variable models when we don't have samples from the true model but only samples after complex perturbations."
                    },
                    {
                        "title": "The Effects of Memory Replay in Reinforcement Learning",
                        "abstract": "Experience replay is a key technique behind many recent advances in deep reinforcement learning. Allowing the agent to learn from earlier memories can speed up learning and break undesirable temporal correlations. Despite its wide-spread application, very little is understood about the properties of experience replay. How does the amount of memory kept affect learning dynamics? Does it help to prioritize certain experiences? In this paper, we address these questions by formulating a dynamical systems ODE model of Q-learning with experience replay. We derive analytic solutions of the ODE for a simple setting. We show that even in this very simple setting, the amount of memory kept can substantially affect the agent's performance. Too much or too little memory both slow down learning. Moreover, we characterize regimes where prioritized replay harms the agent's learning. We show that our analytic solutions have excellent agreement with experiments. Finally, we propose a simple algorithm for adaptively changing the memory buffer size which achieves consistently good empirical performance."
                    },
                    {
                        "title": "Autowarp: Learning a Warping Distance from Unlabeled Time Series Using Sequence Autoencoders",
                        "abstract": "Measuring similarities between unlabeled time series trajectories is an important problem in domains as diverse as medicine, astronomy, finance, and computer vision. It is often unclear what is the appropriate metric to use because of the complex nature of noise in the trajectories (e.g. different sampling rates or outliers). Domain experts typically hand-craft or manually select a specific metric, such as dynamic time warping (DTW), to apply on their data. In this paper, we propose Autowarp, an end-to-end algorithm that optimizes and learns a good metric given unlabeled trajectories. We define a flexible and differentiable family of warping metrics, which encompasses common metrics such as DTW, Euclidean, and edit distance. Autowarp then leverages the representation power of sequence autoencoders to optimize for a member of this warping distance family. The output is a metric which is easy to interpret and can be robustly learned from relatively few trajectories. In systematic experiments across different domains, we show that Autowarp often outperforms hand-crafted trajectory similarity metrics."
                    },
                    {
                        "title": "Minimizing Close-k Aggregate Loss Improves Classification",
                        "abstract": "In classification, the de facto method for aggregating individual losses is the average loss. When the actual metric of interest is 0-1 loss, it is common to minimize the average surrogate loss for some well-behaved (e.g. convex) surrogate. Recently, several other aggregate losses such as the maximal loss and average top-$k$ loss were proposed as alternative objectives to address shortcomings of the average loss. However, we identify common classification settings, e.g. the data is imbalanced, has too many easy or ambiguous examples, etc., when average, maximal and average top-$k$ all suffer from suboptimal decision boundaries, even on an infinitely large training set. To address this problem, we propose a new classification objective called the close-$k$ aggregate loss, where we adaptively minimize the loss for points close to the decision boundary. We provide theoretical guarantees for the 0-1 accuracy when we optimize close-$k$ aggregate loss. We also conduct systematic experiments across the PMLB and OpenML benchmark datasets. Close-$k$ achieves significant gains in 0-1 test accuracy, improvements of $\\geq 2$% and $p<0.05$, in over 25% of the datasets compared to average, maximal and average top-$k$. In contrast, the previous aggregate losses outperformed close-$k$ in less than 2% of the datasets."
                    },
                    {
                        "title": "Neuron Shapley: Discovering the Responsible Neurons",
                        "abstract": "We develop Neuron Shapley as a new framework to quantify the contribution of individual neurons to the prediction and performance of a deep network. By accounting for interactions across neurons, Neuron Shapley is more effective in identifying important filters compared to common approaches based on activation patterns. Interestingly, removing just 30 filters with the highest Shapley scores effectively destroys the prediction accuracy of Inception-v3 on ImageNet. Visualization of these few critical filters provides insights into how the network functions. Neuron Shapley is a flexible framework and can be applied to identify responsible neurons in many tasks. We illustrate additional applications of identifying filters that are responsible for biased prediction in facial recognition and filters that are vulnerable to adversarial attacks. Removing these filters is a quick way to repair models. Enabling all these applications is a new multi-arm bandit algorithm that we developed to efficiently estimate Neuron Shapley values."
                    },
                    {
                        "title": "Signal to noise in matching markets",
                        "abstract": "In many matching markets, one side \"applies\" to the other, and these applications are often expensive and time-consuming (e.g. students applying to college). It is tempting to think that making the application process easier should benefit both sides of the market. After all, the applicants could submit more applications, and the recipients would have more applicants to choose from. In this paper, we propose and analyze a simple model to understand settings where both sides of the market suffer from increased number of applications.   The main insights of the paper are derived from quantifying the signal to noise tradeoffs in random matchings, as applications provide a signal of the applicants' preferences. When applications are costly the signal is stronger, as the act of making an application itself is meaningful. Therefore more applications may yield potentially better matches, but fewer applications create stronger signals for the receiving side to learn true preferences.   We derive analytic characterizations of the expected quality of stable matchings in a simple utility model where applicants have an overall quality, but also synergy with specific possible partners. Our results show how reducing application cost without introducing an effective signaling mechanism might lead to inefficiencies for both sides of the market."
                    },
                    {
                        "title": "Stochastic EM for Shuffled Linear Regression",
                        "abstract": "We consider the problem of inference in a linear regression model in which the relative ordering of the input features and output labels is not known. Such datasets naturally arise from experiments in which the samples are shuffled or permuted during the protocol. In this work, we propose a framework that treats the unknown permutation as a latent variable. We maximize the likelihood of observations using a stochastic expectation-maximization (EM) approach. We compare this to the dominant approach in the literature, which corresponds to hard EM in our framework. We show on synthetic data that the stochastic EM algorithm we develop has several advantages, including lower parameter error, less sensitivity to the choice of initialization, and significantly better performance on datasets that are only partially shuffled. We conclude by performing two experiments on real datasets that have been partially shuffled, in which we show that the stochastic EM algorithm can recover the weights with modest error."
                    },
                    {
                        "title": "Feedback GAN (FBGAN) for DNA: a Novel Feedback-Loop Architecture for Optimizing Protein Functions",
                        "abstract": "Generative Adversarial Networks (GANs) represent an attractive and novel approach to generate realistic data, such as genes, proteins, or drugs, in synthetic biology. Here, we apply GANs to generate synthetic DNA sequences encoding for proteins of variable length. We propose a novel feedback-loop architecture, called Feedback GAN (FBGAN), to optimize the synthetic gene sequences for desired properties using an external function analyzer. The proposed architecture also has the advantage that the analyzer need not be differentiable. We apply the feedback-loop mechanism to two examples: 1) generating synthetic genes coding for antimicrobial peptides, and 2) optimizing synthetic genes for the secondary structure of their resulting peptides. A suite of metrics demonstrate that the GAN generated proteins have desirable biophysical properties. The FBGAN architecture can also be used to optimize GAN-generated datapoints for useful properties in domains beyond genomics."
                    },
                    {
                        "title": "Improving Training on Noisy Stuctured Labels",
                        "abstract": "Fine-grained annotations---e.g. dense image labels, image segmentation and text tagging---are useful in many ML applications but they are labor-intensive to generate. Moreover there are often systematic, structured errors in these fine-grained annotations. For example, a car might be entirely unannotated in the image, or the boundary between a car and street might only be coarsely annotated. Standard ML training on data with such structured errors produces models with biases and poor performance. In this work, we propose a novel framework of Error-Correcting Networks (ECN) to address the challenge of learning in the presence structured error in fine-grained annotations. Given a large noisy dataset with commonly occurring structured errors, and a much smaller dataset with more accurate annotations, ECN is able to substantially improve the prediction of fine-grained annotations compared to standard approaches for training on noisy data. It does so by learning to leverage the structures in the annotations and in the noisy labels. Systematic experiments on image segmentation and text tagging demonstrate the strong performance of ECN in improving training on noisy structured labels."
                    },
                    {
                        "title": "Neural Group Testing to Accelerate Deep Learning",
                        "abstract": "Recent advances in deep learning have made the use of large, deep neural networks with tens of millions of parameters. The sheer size of these networks imposes a challenging computational burden during inference. Existing work focuses primarily on accelerating each forward pass of a neural network. Inspired by the group testing strategy for efficient disease testing, we propose neural group testing, which accelerates by testing a group of samples in one forward pass. Groups of samples that test negative are ruled out. If a group tests positive, samples in that group are then retested adaptively. A key challenge of neural group testing is to modify a deep neural network so that it could test multiple samples in one forward pass. We propose three designs to achieve this without introducing any new parameters and evaluate their performances. We applied neural group testing in an image moderation task to detect rare but inappropriate images. We found that neural group testing can group up to 16 images in one forward pass and reduce the overall computation cost by over 73% while improving detection performance."
                    },
                    {
                        "title": "Beta Shapley: a Unified and Noise-reduced Data Valuation Framework for Machine Learning",
                        "abstract": "Data Shapley has recently been proposed as a principled framework to quantify the contribution of individual datum in machine learning. It can effectively identify helpful or harmful data points for a learning algorithm. In this paper, we propose Beta Shapley, which is a substantial generalization of Data Shapley. Beta Shapley arises naturally by relaxing the efficiency axiom of the Shapley value, which is not critical for machine learning settings. Beta Shapley unifies several popular data valuation methods and includes data Shapley as a special case. Moreover, we prove that Beta Shapley has several desirable statistical properties and propose efficient algorithms to estimate it. We demonstrate that Beta Shapley outperforms state-of-the-art data valuation methods on several downstream ML tasks such as: 1) detecting mislabeled training data; 2) learning with subsamples; and 3) identifying points whose addition or removal have the largest positive or negative impact on the model."
                    },
                    {
                        "title": "WeightedSHAP: analyzing and improving Shapley based feature attributions",
                        "abstract": "Shapley value is a popular approach for measuring the influence of individual features. While Shapley feature attribution is built upon desiderata from game theory, some of its constraints may be less natural in certain machine learning settings, leading to unintuitive model interpretation. In particular, the Shapley value uses the same weight for all marginal contributions -- i.e. it gives the same importance when a large number of other features are given versus when a small number of other features are given. This property can be problematic if larger feature sets are more or less informative than smaller feature sets. Our work performs a rigorous analysis of the potential limitations of Shapley feature attribution. We identify simple settings where the Shapley value is mathematically suboptimal by assigning larger attributions for less influential features. Motivated by this observation, we propose WeightedSHAP, which generalizes the Shapley value and learns which marginal contributions to focus directly from data. On several real-world datasets, we demonstrate that the influential features identified by WeightedSHAP are better able to recapitulate the model's predictions compared to the features identified by the Shapley value."
                    },
                    {
                        "title": "New Evaluation Metrics Capture Quality Degradation due to LLM Watermarking",
                        "abstract": "With the increasing use of large-language models (LLMs) like ChatGPT, watermarking has emerged as a promising approach for tracing machine-generated content. However, research on LLM watermarking often relies on simple perplexity or diversity-based measures to assess the quality of watermarked text, which can mask important limitations in watermarking. Here we introduce two new easy-to-use methods for evaluating watermarking algorithms for LLMs: 1) evaluation by LLM-judger with specific guidelines; and 2) binary classification on text embeddings to distinguish between watermarked and unwatermarked text. We apply these methods to characterize the effectiveness of current watermarking techniques. Our experiments, conducted across various datasets, reveal that current watermarking methods are detectable by even simple classifiers, challenging the notion of watermarking subtlety. We also found, through the LLM judger, that watermarking impacts text quality, especially in degrading the coherence and depth of the response. Our findings underscore the trade-off between watermark robustness and text quality and highlight the importance of having more informative metrics to assess watermarking quality."
                    },
                    {
                        "title": "Large Language Models are Vulnerable to Bait-and-Switch Attacks for Generating Harmful Content",
                        "abstract": "The risks derived from large language models (LLMs) generating deceptive and damaging content have been the subject of considerable research, but even safe generations can lead to problematic downstream impacts. In our study, we shift the focus to how even safe text coming from LLMs can be easily turned into potentially dangerous content through Bait-and-Switch attacks. In such attacks, the user first prompts LLMs with safe questions and then employs a simple find-and-replace post-hoc technique to manipulate the outputs into harmful narratives. The alarming efficacy of this approach in generating toxic content highlights a significant challenge in developing reliable safety guardrails for LLMs. In particular, we stress that focusing on the safety of the verbatim LLM outputs is insufficient and that we also need to consider post-hoc transformations."
                    },
                    {
                        "title": "Talking Nonsense: Probing Large Language Models' Understanding of Adversarial Gibberish Inputs",
                        "abstract": "Large language models (LLMs) exhibit excellent ability to understand human languages, but do they also understand their own language that appears gibberish to us? In this work we delve into this question, aiming to uncover the mechanisms underlying such behavior in LLMs. We employ the Greedy Coordinate Gradient optimizer to craft prompts that compel LLMs to generate coherent responses from seemingly nonsensical inputs. We call these inputs LM Babel and this work systematically studies the behavior of LLMs manipulated by these prompts. We find that the manipulation efficiency depends on the target text's length and perplexity, with the Babel prompts often located in lower loss minima compared to natural prompts. We further examine the structure of the Babel prompts and evaluate their robustness. Notably, we find that guiding the model to generate harmful texts is not more difficult than into generating benign texts, suggesting lack of alignment for out-of-distribution prompts."
                    },
                    {
                        "title": "Contrastive Variational Autoencoder Enhances Salient Features",
                        "abstract": "Variational autoencoders are powerful algorithms for identifying dominant latent structure in a single dataset. In many applications, however, we are interested in modeling latent structure and variation that are enriched in a target dataset compared to some background---e.g. enriched in patients compared to the general population. Contrastive learning is a principled framework to capture such enriched variation between the target and background, but state-of-the-art contrastive methods are limited to linear models. In this paper, we introduce the contrastive variational autoencoder (cVAE), which combines the benefits of contrastive learning with the power of deep generative models. The cVAE is designed to identify and enhance salient latent features. The cVAE is trained on two related but unpaired datasets, one of which has minimal contribution from the salient latent features. The cVAE explicitly models latent features that are shared between the datasets, as well as those that are enriched in one dataset relative to the other, which allows the algorithm to isolate and enhance the salient latent features. The algorithm is straightforward to implement, has a similar run-time to the standard VAE, and is robust to noise and dataset purity. We conduct experiments across diverse types of data, including gene expression and facial images, showing that the cVAE effectively uncovers latent structure that is salient in a particular analysis."
                    },
                    {
                        "title": "Data Shapley: Equitable Valuation of Data for Machine Learning",
                        "abstract": "As data becomes the fuel driving technological and economic growth, a fundamental challenge is how to quantify the value of data in algorithmic predictions and decisions. For example, in healthcare and consumer markets, it has been suggested that individuals should be compensated for the data that they generate, but it is not clear what is an equitable valuation for individual data. In this work, we develop a principled framework to address data valuation in the context of supervised machine learning. Given a learning algorithm trained on $n$ data points to produce a predictor, we propose data Shapley as a metric to quantify the value of each training datum to the predictor performance. Data Shapley value uniquely satisfies several natural properties of equitable data valuation. We develop Monte Carlo and gradient-based methods to efficiently estimate data Shapley values in practical settings where complex learning algorithms, including neural networks, are trained on large datasets. In addition to being equitable, extensive experiments across biomedical, image and synthetic data demonstrate that data Shapley has several other benefits: 1) it is more powerful than the popular leave-one-out or leverage score in providing insight on what data is more valuable for a given learning task; 2) low Shapley value data effectively capture outliers and corruptions; 3) high Shapley value data inform what type of new data to acquire to improve the predictor."
                    }
                ]
            },
            "e3446529-e86a-4d7a-ac43-85b382f54602": {
                "pk": "e3446529-e86a-4d7a-ac43-85b382f54602",
                "name": "Tatsunori Hashimoto",
                "collaborators": [
                    "Percy Liang",
                    "Daniel Kang",
                    "Esin Durmus",
                    "Faisal Ladhak",
                    "Paul Michel",
                    "Graham Neubig",
                    "Tianyi Zhang",
                    "Christopher Mohri",
                    "John Duchi",
                    "Hongseok Namkoong"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Robust Optimization",
                    "Self-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Improved Natural Language Generation via Loss Truncation",
                        "abstract": "Neural language models are usually trained to match the distributional properties of a large-scale corpus by minimizing the log loss. While straightforward to optimize, this approach forces the model to reproduce all variations in the dataset, including noisy and invalid references (e.g., misannotation and hallucinated facts). Worse, the commonly used log loss is overly sensitive to such phenomena and even a small fraction of noisy data can degrade performance. In this work, we show that the distinguishability of the models and reference serves as a principled and robust alternative for handling invalid references. To optimize distinguishability, we propose loss truncation, which adaptively removes high loss examples during training. We show this is as easy to optimize as log loss and tightly bounds distinguishability under noise. Empirically, we demonstrate that loss truncation outperforms existing baselines on distinguishability on a summarization task, and show that samples generated by the loss truncation model have factual accuracy ratings that exceed those of baselines and match human references."
                    },
                    {
                        "title": "On the Inductive Bias of Masked Language Modeling: From Statistical to Syntactic Dependencies",
                        "abstract": "We study how masking and predicting tokens in an unsupervised fashion can give rise to linguistic structures and downstream performance gains. Recent theories have suggested that pretrained language models acquire useful inductive biases through masks that implicitly act as cloze reductions for downstream tasks. While appealing, we show that the success of the random masking strategy used in practice cannot be explained by such cloze-like masks alone. We construct cloze-like masks using task-specific lexicons for three different classification datasets and show that the majority of pretrained performance gains come from generic masks that are not associated with the lexicon. To explain the empirical success of these generic masks, we demonstrate a correspondence between the Masked Language Model (MLM) objective and existing methods for learning statistical dependencies in graphical models. Using this, we derive a method for extracting these learned statistical dependencies in MLMs and show that these dependencies encode useful inductive biases in the form of syntactic structures. In an unsupervised parsing evaluation, simply forming a minimum spanning tree on the implied statistical dependence structure outperforms a classic method for unsupervised parsing (58.74 vs. 55.91 UUAS)."
                    },
                    {
                        "title": "Language Models with Conformal Factuality Guarantees",
                        "abstract": "Guaranteeing the correctness and factuality of language model (LM) outputs is a major open problem. In this work, we propose conformal factuality, a framework that can ensure high probability correctness guarantees for LMs by connecting language modeling and conformal prediction. We observe that the correctness of an LM output is equivalent to an uncertainty quantification problem, where the uncertainty sets are defined as the entailment set of an LM's output. Using this connection, we show that conformal prediction in language models corresponds to a back-off algorithm that provides high probability correctness guarantees by progressively making LM outputs less specific (and expanding the associated uncertainty sets). This approach applies to any black-box LM and requires very few human-annotated samples. Evaluations of our approach on closed book QA (FActScore, NaturalQuestions) and reasoning tasks (MATH) show that our approach can provide 80-90% correctness guarantees while retaining the majority of the LM's original output."
                    },
                    {
                        "title": "Distributionally Robust Losses for Latent Covariate Mixtures",
                        "abstract": "While modern large-scale datasets often consist of heterogeneous subpopulations -- for example, multiple demographic groups or multiple text corpora -- the standard practice of minimizing average loss fails to guarantee uniformly low losses across all subpopulations. We propose a convex procedure that controls the worst-case performance over all subpopulations of a given size. Our procedure comes with finite-sample (nonparametric) convergence guarantees on the worst-off subpopulation. Empirically, we observe on lexical similarity, wine quality, and recidivism prediction tasks that our worst-case procedure learns models that do well against unseen subpopulations."
                    },
                    {
                        "title": "Spurious Correlations in Reference-Free Evaluation of Text Generation",
                        "abstract": "Model-based, reference-free evaluation metrics have been proposed as a fast and cost-effective approach to evaluate Natural Language Generation (NLG) systems. Despite promising recent results, we find evidence that reference-free evaluation metrics of summarization and dialog generation may be relying on spurious correlations with measures such as word overlap, perplexity, and length. We further observe that for text summarization, these metrics have high error rates when ranking current state-of-the-art abstractive summarization systems. We demonstrate that these errors can be mitigated by explicitly designing evaluation metrics to avoid spurious features in reference-free evaluation."
                    },
                    {
                        "title": "Robustness to Spurious Correlations via Human Annotations",
                        "abstract": "The reliability of machine learning systems critically assumes that the associations between features and labels remain similar between training and test distributions. However, unmeasured variables, such as confounders, break this assumption---useful correlations between features and labels at training time can become useless or even harmful at test time. For example, high obesity is generally predictive for heart disease, but this relation may not hold for smokers who generally have lower rates of obesity and higher rates of heart disease. We present a framework for making models robust to spurious correlations by leveraging humans' common sense knowledge of causality. Specifically, we use human annotation to augment each training example with a potential unmeasured variable (i.e. an underweight patient with heart disease may be a smoker), reducing the problem to a covariate shift problem. We then introduce a new distributionally robust optimization objective over unmeasured variables (UV-DRO) to control the worst-case loss over possible test-time shifts. Empirically, we show improvements of 5-10% on a digit recognition task confounded by rotation, and 1.5-5% on the task of analyzing NYPD Police Stops confounded by location."
                    },
                    {
                        "title": "Scaling up Trustless DNN Inference with Zero-Knowledge Proofs",
                        "abstract": "As ML models have increased in capabilities and accuracy, so has the complexity of their deployments. Increasingly, ML model consumers are turning to service providers to serve the ML models in the ML-as-a-service (MLaaS) paradigm. As MLaaS proliferates, a critical requirement emerges: how can model consumers verify that the correct predictions were served, in the face of malicious, lazy, or buggy service providers?   In this work, we present the first practical ImageNet-scale method to verify ML model inference non-interactively, i.e., after the inference has been done. To do so, we leverage recent developments in ZK-SNARKs (zero-knowledge succinct non-interactive argument of knowledge), a form of zero-knowledge proofs. ZK-SNARKs allows us to verify ML model execution non-interactively and with only standard cryptographic hardness assumptions. In particular, we provide the first ZK-SNARK proof of valid inference for a full resolution ImageNet model, achieving 79\\% top-5 accuracy. We further use these ZK-SNARKs to design protocols to verify ML model execution in a variety of scenarios, including for verifying MLaaS predictions, verifying MLaaS model accuracy, and using ML models for trustless retrieval. Together, our results show that ZK-SNARKs have the promise to make verified ML model inference practical."
                    },
                    {
                        "title": "Contrastive Error Attribution for Finetuned Language Models",
                        "abstract": "Recent work has identified noisy and misannotated data as a core cause of hallucinations and unfaithful outputs in Natural Language Generation (NLG) tasks. Consequently, identifying and removing these examples is a key open challenge in creating reliable NLG systems. In this work, we introduce a framework to identify and remove low-quality training instances that lead to undesirable outputs, such as faithfulness errors in text summarization. We show that existing approaches for error tracing, such as gradient-based influence measures, do not perform reliably for detecting faithfulness errors in NLG datasets. We overcome the drawbacks of existing error tracing methods through a new, contrast-based estimate that compares undesired generations to human-corrected outputs. Our proposed method can achieve a mean average precision of 0.93 at detecting known data errors across synthetic tasks with known ground truth, substantially outperforming existing approaches. Using this approach and re-training models on cleaned data leads to a 70% reduction in entity hallucinations on the NYT dataset and a 55% reduction in semantic errors on the E2E dataset."
                    },
                    {
                        "title": "Evaluating Self-Supervised Learning via Risk Decomposition",
                        "abstract": "Self-supervised learning (SSL) pipelines differ in many design choices such as the architecture, augmentations, or pretraining data. Yet SSL is typically evaluated using a single metric: linear probing on ImageNet. This does not provide much insight into why or when a model is better, now how to improve it. To address this, we propose an SSL risk decomposition, which generalizes the classical supervised approximation-estimation decomposition by considering errors arising from the representation learning step. Our decomposition consists of four error components: approximation, representation usability, probe generalization, and encoder generalization. We provide efficient estimators for each component and use them to analyze the effect of 30 design choices on 169 SSL vision models evaluated on ImageNet. Our analysis gives valuable insights for designing and using SSL models. For example, it highlights the main sources of error and shows how to improve SSL in specific settings (full- vs few-shot) by trading off error components. All results and pretrained models are at https://github.com/YannDubs/SSL-Risk-Decomposition."
                    },
                    {
                        "title": "Navigating the Grey Area: How Expressions of Uncertainty and Overconfidence Affect Language Models",
                        "abstract": "The increased deployment of LMs for real-world tasks involving knowledge and facts makes it important to understand model epistemology: what LMs think they know, and how their attitudes toward that knowledge are affected by language use in their inputs. Here, we study an aspect of model epistemology: how epistemic markers of certainty, uncertainty, or evidentiality like \"I'm sure it's\", \"I think it's\", or \"Wikipedia says it's\" affect models, and whether they contribute to model failures. We develop a typology of epistemic markers and inject 50 markers into prompts for question answering. We find that LMs are highly sensitive to epistemic markers in prompts, with accuracies varying more than 80%. Surprisingly, we find that expressions of high certainty result in a 7% decrease in accuracy as compared to low certainty expressions; similarly, factive verbs hurt performance, while evidentials benefit performance. Our analysis of a popular pretraining dataset shows that these markers of uncertainty are associated with answers on question-answering websites, while markers of certainty are associated with questions. These associations may suggest that the behavior of LMs is based on mimicking observed language use, rather than truly reflecting epistemic uncertainty."
                    },
                    {
                        "title": "Understanding Finetuning for Factual Knowledge Extraction",
                        "abstract": "In this work, we study the impact of QA fine-tuning data on downstream factuality. We show that fine-tuning on lesser-known facts that are poorly stored during pretraining yields significantly worse factuality than fine-tuning on well-known facts, even when all facts are seen during pretraining. We prove this phenomenon theoretically, showing that training on lesser-known facts can lead the model to ignore subject entity names and instead output a generic plausible response even when the relevant factual knowledge is encoded in the model. On three question answering benchmarks (PopQA, Entity Questions, and MMLU) and two language models (Llama-2-7B and Mistral-7B), we find that (i) finetuning on a completely factual but lesser-known subset of the data deteriorates downstream factuality (5-10%) and (ii) finetuning on a subset of better-known examples matches or outperforms finetuning on the entire dataset. Ultimately, our results shed light on the interaction between pretrained knowledge and finetuning data and demonstrate the importance of taking into account how facts are stored in the pretrained model when fine-tuning for knowledge-intensive tasks."
                    },
                    {
                        "title": "Improving Pretraining Data Using Perplexity Correlations",
                        "abstract": "Quality pretraining data is often seen as the key to high-performance language models. However, progress in understanding pretraining data has been slow due to the costly pretraining runs required for data selection experiments. We present a framework that avoids these costs and selects high-quality pretraining data without any LLM training of our own. Our work is based on a simple observation: LLM losses on many pretraining texts are correlated with downstream benchmark performance, and selecting high-correlation documents is an effective pretraining data selection method. We build a new statistical framework for data selection centered around estimates of perplexity-benchmark correlations and perform data selection using a sample of 90 LLMs taken from the Open LLM Leaderboard on texts from tens of thousands of web domains. In controlled pretraining experiments at the 160M parameter scale on 8 benchmarks, our approach outperforms DSIR on every benchmark, while matching the best data selector found in DataComp-LM, a hand-engineered bigram classifier."
                    },
                    {
                        "title": "Modeling the Second Player in Distributionally Robust Optimization",
                        "abstract": "Distributionally robust optimization (DRO) provides a framework for training machine learning models that are able to perform well on a collection of related data distributions (the \"uncertainty set\"). This is done by solving a min-max game: the model is trained to minimize its maximum expected loss among all distributions in the uncertainty set. While careful design of the uncertainty set is critical to the success of the DRO procedure, previous work has been limited to relatively simple alternatives that keep the min-max optimization problem exactly tractable, such as $f$-divergence balls. In this paper, we argue instead for the use of neural generative models to characterize the worst-case distribution, allowing for more flexible and problem-specific selection of the uncertainty set. However, while simple conceptually, this approach poses a number of implementation and optimization challenges. To circumvent these issues, we propose a relaxation of the KL-constrained inner maximization objective that makes the DRO problem more amenable to gradient-based optimization of large scale generative models, and develop model selection heuristics to guide hyper-parameter search. On both toy settings and realistic NLP tasks, we find that the proposed approach yields models that are more robust than comparable baselines."
                    },
                    {
                        "title": "Large Language Models Can Be Strong Differentially Private Learners",
                        "abstract": "Differentially Private (DP) learning has seen limited success for building large deep learning models of text, and straightforward attempts at applying Differentially Private Stochastic Gradient Descent (DP-SGD) to NLP tasks have resulted in large performance drops and high computational overhead. We show that this performance drop can be mitigated with (1) the use of large pretrained language models; (2) non-standard hyperparameters that suit DP optimization; and (3) fine-tuning objectives which are aligned with the pretraining procedure. With the above, we obtain NLP models that outperform state-of-the-art DP-trained models under the same privacy budget and strong non-private baselines -- by directly fine-tuning pretrained models with DP optimization on moderately-sized corpora. To address the computational challenge of running DP-SGD with large Transformers, we propose a memory saving technique that allows clipping in DP-SGD to run without instantiating per-example gradients for any linear layer in the model. The technique enables privately training Transformers with almost the same memory cost as non-private training at a modest run-time overhead. Contrary to conventional wisdom that DP optimization fails at learning high-dimensional models (due to noise that scales with dimension) empirical results reveal that private learning with pretrained language models doesn't tend to suffer from dimension-dependent performance degradation. Code to reproduce results can be found at https://github.com/lxuechen/private-transformers."
                    },
                    {
                        "title": "Distributionally Robust Models with Parametric Likelihood Ratios",
                        "abstract": "As machine learning models are deployed ever more broadly, it becomes increasingly important that they are not only able to perform well on their training distribution, but also yield accurate predictions when confronted with distribution shift. The Distributionally Robust Optimization (DRO) framework proposes to address this issue by training models to minimize their expected risk under a collection of distributions, to imitate test-time shifts. This is most commonly achieved by instance-level re-weighting of the training objective to emulate the likelihood ratio with possible test distributions, which allows for estimating their empirical risk via importance sampling (assuming that they are subpopulations of the training distribution). However, re-weighting schemes in the literature are usually limited due to the difficulty of keeping the optimization problem tractable and the complexity of enforcing normalization constraints. In this paper, we show that three simple ideas -- mini-batch level normalization, a KL penalty and simultaneous gradient updates -- allow us to train models with DRO using a broader class of parametric likelihood ratios. In a series of experiments on both image and text classification benchmarks, we find that models trained with the resulting parametric adversaries are consistently more robust to subpopulation shifts when compared to other DRO approaches, and that the method performs reliably well with little hyper-parameter tuning. Code to reproduce our experiments can be found at https://github.com/pmichel31415/P-DRO."
                    }
                ]
            }
        }
    },
    "2405.16493": {
        "paper_data": {
            "title": "Flow Snapshot Neurons in Action: Deep Neural Networks Generalize to Biological Motion Perception",
            "url": "http://arxiv.org/abs/2405.16493v1",
            "arxiv_id": "2405.16493",
            "authors": [
                "Shuangpeng Han",
                "Ziyu Wang",
                "Mengmi Zhang"
            ],
            "abstract": "Biological motion perception (BMP) refers to humans' ability to perceive and recognize the actions of living beings solely from their motion patterns, sometimes as minimal as those depicted on point-light displays. While humans excel at these tasks without any prior training, current AI models struggle with poor generalization performance. To close this research gap, we propose the Motion Perceiver (MP). MP solely relies on patch-level optical flows from video clips as inputs. During training, it learns prototypical flow snapshots through a competitive binding mechanism and integrates invariant motion representations to predict action labels for the given video. During inference, we evaluate the generalization ability of all AI models and humans on 62,656 video stimuli spanning 24 BMP conditions using point-light displays in neuroscience. Remarkably, MP outperforms all existing AI models with a maximum improvement of 29% in top-1 action recognition accuracy on these conditions. Moreover, we benchmark all AI models in point-light displays of two standard video datasets in computer vision. MP also demonstrates superior performance in these cases. More interestingly, via psychophysics experiments, we found that MP recognizes biological movements in a way that aligns with human behavioural data. All data and code will be made public.",
            "introduction": "   1 Introduction  Figure 1: Humans excel at biological motion perception (BMP) tasks with zero training, while current AI models struggle with poor generalization performance. AI models are trained to learn to recognize actions from natural RGB videos and tested using BMP stimuli on point-light displays, which come in two forms: Joint videos, which display only the detected joints of actors in white dots, and Sequential position action videos (SP), where light points in white are randomly positioned between joints and reallocated to other random positions on the limb in subsequent frames (Sec.\u00a03.1). Note that skeletons, shown in gray in the example video, are not visible to humans or AI models during testing. The generalization performance of both humans and models is assessed after varying five properties in temporal and visual dimensions. Example videos are available at: link.    Biological Motion Perception (BMP) refers to the remarkable ability to recognize and understand the actions and intentions of other living beings based solely on their motion patterns \u00a0[30]. BMP is crucial for tasks such as predator detection\u00a0[23], prey selection\u00a0[22], courtship behavior\u00a0[62], and social communications\u00a0[17, 2, 1] among primates and humans. In classical psychophysical and neurophysiological experiments \u00a0[40], motion patterns can sometimes be depicted minimally, such as in point-light displays where only major joints of human or animal actors are illuminated. Yet, without any prior training, humans can robustly and accurately recognize actions \u00a0[6, 64, 60] and characteristics of these actors like gender\u00a0[46, 72] identity \u00a0[16, 66], personalities\u00a0[7], emotions\u00a0[19], social interactions\u00a0[86], and casual intention\u00a0[70].   To make sense of these psychophysical and neurophysiological data, numerous studies \u00a0[30, 31, 48, 82, 9, 45] have proposed computational frameworks and models in BMP tasks. Unlike humans, who excel at BMP tasks without prior training, these computational models are usually trained under specific BMP conditions and subsequently evaluated under different BMP conditions. However, the extent to which these models can learn robust motion representations from natural RGB videos and generalize them to recognize actions on BMP stimuli remains largely unexplored.   In parallel to the studies of BMP in psychology and neuroscience, action recognition on images and videos in computer vision has evolved significantly over the past few decades due to its wide range of real-world applications\u00a0[15, 37, 75, 76, 50, 67]. The field has progressed from relying heavily on hand-crafted features\u00a0[91, 68, 47] to employing sophisticated deep-learning architectures \u00a0[87, 11, 93, 27, 29, 28, 24, 4] capable of capturing the temporal dynamics and spatial configurations of complex activities within dynamic, unstructured environments \u00a0[74, 85]. Despite these advancements, existing AI models still struggle with generalization issues related to occlusion\u00a0[95, 3, 51], noisy environments \u00a0[100], viewpoint variability \u00a0[26, 54, 92], and subtle human movements \u00a0[77, 39]. While various solutions have been proposed to enhance AI generalization in action recognition\u00a0[12, 65, 14, 63, 96, 42, 52, 102, 73, 43], they do not specifically tackle the generalization challenges in BMP tasks.   Here, our objective is to systematically and quantitatively examine the generalization ability of AI models trained on natural RGB videos and tested in BMP tasks. To date, research efforts in neuroscience and psychology \u00a0[39, 31, 34] have mostly focused on specific",
            "references": [
                {
                    "title": "Segregated Dynamical Networks for Biological Motion Perception in the Mu and Beta Range Underlie Social Deficits in Autism",
                    "abstract": "Objective: Biological motion perception (BMP) correlating with a mirror neuron system (MNS) is attenuated in underage individuals with autism spectrum disorder (ASD). While BMP in typically-developing controls (TDCs) encompasses interconnected MNS structures, ASD data hint at segregated form and motion processing. This coincides with less fewer long-range connections in ASD than TDC. Using BMP and electroencephalography (EEG) in ASD, we characterized directionality and coherence (mu and beta frequencies). Deficient BMP may stem from desynchronization thereof in MNS and may predict social-communicative deficits in ASD. Clinical considerations thus profit from brain\u2013behavior associations. Methods: Point-like walkers elicited BMP using 15 white dots (walker vs. scramble in 21 ASD (mean: 11.3 \u00b1 2.3 years) vs. 23 TDC (mean: 11.9 \u00b1 2.5 years). Dynamic Imaging of Coherent Sources (DICS) characterized the underlying EEG time-frequency causality through time-resolved Partial Directed Coherence (tPDC). Support Vector Machine (SVM) classification validated the group effects (ASD vs. TDC). Results: TDC showed MNS sources and long-distance paths (both feedback and bidirectional); ASD demonstrated distinct from and motion sources, predominantly local feedforward connectivity, and weaker coherence. Brain\u2013behavior correlations point towards dysfunctional networks. SVM successfully classified ASD regarding EEG and performance. Conclusion: ASD participants showed segregated local networks for BMP potentially underlying thwarted complex social interactions. Alternative explanations include selective attention and global\u2013local processing deficits. Significance: This is the first study applying source-based connectivity to reveal segregated BMP networks in ASD regarding structure, cognition, frequencies, and temporal dynamics that may explain socio-communicative aberrancies."
                },
                {
                    "title": "Object-Centric Learning for Real-World Videos by Predicting Temporal Feature Similarities",
                    "abstract": "Unsupervised video-based object-centric learning is a promising avenue to learn structured representations from large, unlabeled video collections, but previous approaches have only managed to scale to real-world datasets in restricted domains. Recently, it was shown that the reconstruction of pre-trained self-supervised features leads to object-centric representations on unconstrained real-world image datasets. Building on this approach, we propose a novel way to use such pre-trained features in the form of a temporal feature similarity loss. This loss encodes semantic and temporal correlations between image patches and is a natural way to introduce a motion bias for object discovery. We demonstrate that this loss leads to state-of-the-art performance on the challenging synthetic MOVi datasets. When used in combination with the feature reconstruction loss, our model is the first object-centric video model that scales to unconstrained video datasets such as YouTube-VIS."
                },
                {
                    "title": "FFCV: Accelerating Training by Removing Data Bottlenecks",
                    "abstract": "We present FFCV, a library for easy and fast machine learning model training. FFCV speeds up model training by eliminating (often subtle) data bottlenecks from the training process. In particular, we combine techniques such as an efficient file storage format, caching, data pre-loading, asynchronous data transfer, and just-in-time compilation to (a) make data loading and transfer significantly more efficient, ensuring that GPUs can reach full utilization; and (b) offload as much data processing as possible to the CPU asynchronously, freeing GPU cycles for training. Using FFCV, we train ResNet-18 and ResNet-50 on the ImageNet dataset with a state-of-the-art tradeoff between accuracy and training time. For example, across the range of ResNet-50 models we test, we obtain the same accuracy as the best baselines in half the time. We demonstrate FFCV's performance, ease-of-use, extensibility, and ability to adapt to resource constraints through several case studies. Detailed installation instructions, documentation, and Slack support channel are available at https://ffcv.io/."
                },
                {
                    "title": "VDT: General-purpose Video Diffusion Transformers via Mask Modeling",
                    "abstract": "This work introduces Video Diffusion Transformer (VDT), which pioneers the use of transformers in diffusion-based video generation. It features transformer blocks with modularized temporal and spatial attention modules to leverage the rich spatial-temporal representation inherited in transformers. We also propose a unified spatial-temporal mask modeling mechanism, seamlessly integrated with the model, to cater to diverse video generation scenarios. VDT offers several appealing benefits. 1) It excels at capturing temporal dependencies to produce temporally consistent video frames and even simulate the physics and dynamics of 3D objects over time. 2) It facilitates flexible conditioning information, \\eg, simple concatenation in the token space, effectively unifying different token lengths and modalities. 3) Pairing with our proposed spatial-temporal mask modeling mechanism, it becomes a general-purpose video diffuser for harnessing a range of tasks, including unconditional generation, video prediction, interpolation, animation, and completion, etc. Extensive experiments on these tasks spanning various scenarios, including autonomous driving, natural weather, human action, and physics-based simulation, demonstrate the effectiveness of VDT. Additionally, we present comprehensive studies on how \\model handles conditioning information with the mask modeling mechanism, which we believe will benefit future research and advance the field. Project page: https:VDT-2023.github.io"
                },
                {
                    "title": "Object-centric Learning with Cyclic Walks between Parts and Whole",
                    "abstract": "Learning object-centric representations from complex natural environments enables both humans and machines with reasoning abilities from low-level perceptual features. To capture compositional entities of the scene, we proposed cyclic walks between perceptual features extracted from vision transformers and object entities. First, a slot-attention module interfaces with these perceptual features and produces a finite set of slot representations. These slots can bind to any object entities in the scene via inter-slot competitions for attention. Next, we establish entity-feature correspondence with cyclic walks along high transition probability based on the pairwise similarity between perceptual features (aka\"parts\") and slot-binded object representations (aka\"whole\"). The whole is greater than its parts and the parts constitute the whole. The part-whole interactions form cycle consistencies, as supervisory signals, to train the slot-attention module. Our rigorous experiments on \\textit{seven} image datasets in \\textit{three} \\textit{unsupervised} tasks demonstrate that the networks trained with our cyclic walks can disentangle foregrounds and backgrounds, discover objects, and segment semantic objects in complex scenes. In contrast to object-centric models attached with a decoder for the pixel-level or feature-level reconstructions, our cyclic walks provide strong learning signals, avoiding computation overheads and enhancing memory efficiency. Our source code and data are available at: \\href{https://github.com/ZhangLab-DeepNeuroCogLab/Parts-Whole-Object-Centric-Learning/}{link}."
                },
                {
                    "title": "SVFormer: Semi-supervised Video Transformer for Action Recognition",
                    "abstract": "Semi-supervised action recognition is a challenging but critical task due to the high cost of video annotations. Existing approaches mainly use convolutional neural networks, yet current revolutionary vision transformer models have been less explored. In this paper, we investigate the use of transformer models under the SSL setting for action recognition. To this end, we introduce SVFormer, which adopts a steady pseudo-labeling framework (i.e., EMA-Teacher) to cope with unlabeled video samples. While a wide range of data augmentations have been shown effective for semi-supervised image classification, they generally produce limited results for video recognition. We therefore introduce a novel augmentation strategy, Tube Token-Mix, tailored for video data where video clips are mixed via a mask with consistent masked tokens over the temporal axis. In addition, we propose a temporal warping augmentation to cover the complex temporal variation in videos, which stretches selected frames to various temporal durations in the clip. Extensive experiments on three datasets Kinetics-400, UCF-101, and HMDB-51 verify the advantage of SVFormer. In particular, SVFormer outperforms the state-of-the-art by 31.5% with fewer training epochs under the 1% labeling rate of Kinetics-400. Our method can hopefully serve as a strong benchmark and encourage future search on semi-supervised action recognition with Transformer networks. Code is released at https://github.com/ChenHsing/SVFormer."
                },
                {
                    "title": "Unifying Flow, Stereo and Depth Estimation",
                    "abstract": "We present a unified formulation and model for three motion and 3D perception tasks: optical flow, rectified stereo matching and unrectified stereo depth estimation from posed images. Unlike previous specialized architectures for each specific task, we formulate all three tasks as a unified dense correspondence matching problem, which can be solved with a single model by directly comparing feature similarities. Such a formulation calls for discriminative feature representations, which we achieve using a Transformer, in particular the cross-attention mechanism. We demonstrate that cross-attention enables integration of knowledge from another image via cross-view interactions, which greatly improves the quality of the extracted features. Our unified model naturally enables cross-task transfer since the model architecture and parameters are shared across tasks. We outperform RAFT with our unified model on the challenging Sintel dataset, and our final model that uses a few additional task-specific refinement steps outperforms or compares favorably to recent state-of-the-art methods on 10 popular flow, stereo and depth datasets, while being simpler and more efficient in terms of model design and inference speed."
                },
                {
                    "title": "AlphaPose: Whole-Body Regional Multi-Person Pose Estimation and Tracking in Real-Time",
                    "abstract": "Accurate whole-body multi-person pose estimation and tracking is an important yet challenging topic in computer vision. To capture the subtle actions of humans for complex behavior analysis, whole-body pose estimation including the face, body, hand and foot is essential over conventional body-only pose estimation. In this article, we present AlphaPose, a system that can perform accurate whole-body pose estimation and tracking jointly while running in realtime. To this end, we propose several new techniques: Symmetric Integral Keypoint Regression (SIKR) for fast and fine localization, Parametric Pose Non-Maximum-Suppression (P-NMS) for eliminating redundant human detections and Pose Aware Identity Embedding for jointly pose estimation and tracking. During training, we resort to Part-Guided Proposal Generator (PGPG) and multi-domain knowledge distillation to further improve the accuracy. Our method is able to localize whole-body keypoints accurately and tracks humans simultaneously given inaccurate bounding boxes and redundant detections. We show a significant improvement over current state-of-the-art methods in both speed and accuracy on COCO-wholebody, COCO, PoseTrack, and our proposed Halpe-FullBody pose estimation dataset. Our model, source codes and dataset are made publicly available at https://github.com/MVIG-SJTU/AlphaPose."
                },
                {
                    "title": "A Large-Scale Robustness Analysis of Video Action Recognition Models",
                    "abstract": "We have seen a great progress in video action recognition in recent years. There are several models based on convolutional neural network (CNN) and some recent transformer based approaches which provide top performance on existing benchmarks. In this work, we perform a large-scale robustness analysis of these existing models for video action recognition. We focus on robustness against real-world distribution shift perturbations instead of adversarial perturbations. We propose four different benchmark datasets, HMDB51-P, UCF101-P, Kinetics400-P, and SSv2-P to perform this analysis. We study robustness of six state-of-the-art action recognition models against 90 different perturbations. The study reveals some interesting findings, 1) transformer based models are consistently more robust compared to CNN based models, 2) Pretraining improves robustness for Transformer based models more than CNN based models, and 3) All of the studied models are robust to temporal perturbations for all datasets but SSv2; suggesting the importance of temporal information for action recognition varies based on the dataset and activities. Next, we study the role of augmentations in model robustness and present a real-world dataset, UCF101-DS, which contains realistic distribution shifts, to further validate some of these findings. We believe this study will serve as a benchmark for future research in robust video action recognition 11More details available at bit.1y/3TJLMUF.."
                },
                {
                    "title": "Exploring Temporally Dynamic Data Augmentation for Video Recognition",
                    "abstract": "Data augmentation has recently emerged as an essential component of modern training recipes for visual recognition tasks. However, data augmentation for video recognition has been rarely explored despite its effectiveness. Few existing augmentation recipes for video recognition naively extend the image augmentation methods by applying the same operations to the whole video frames. Our main idea is that the magnitude of augmentation operations for each frame needs to be changed over time to capture the real-world video's temporal variations. These variations should be generated as diverse as possible using fewer additional hyper-parameters during training. Through this motivation, we propose a simple yet effective video data augmentation framework, DynaAugment. The magnitude of augmentation operations on each frame is changed by an effective mechanism, Fourier Sampling that parameterizes diverse, smooth, and realistic temporal variations. DynaAugment also includes an extended search space suitable for video for automatic data augmentation methods. DynaAugment experimentally demonstrates that there are additional performance rooms to be improved from static augmentations on diverse video models. Specifically, we show the effectiveness of DynaAugment on various video datasets and tasks: large-scale video recognition (Kinetics-400 and Something-Something-v2), small-scale video recognition (UCF- 101 and HMDB-51), fine-grained video recognition (Diving-48 and FineGym), video action segmentation on Breakfast, video action localization on THUMOS'14, and video object detection on MOT17Det. DynaAugment also enables video models to learn more generalized representation to improve the model robustness on the corrupted videos."
                },
                {
                    "title": "SAVi++: Towards End-to-End Object-Centric Learning from Real-World Videos",
                    "abstract": "The visual world can be parsimoniously characterized in terms of distinct entities with sparse interactions. Discovering this compositional structure in dynamic visual scenes has proven challenging for end-to-end computer vision approaches unless explicit instance-level supervision is provided. Slot-based models leveraging motion cues have recently shown great promise in learning to represent, segment, and track objects without direct supervision, but they still fail to scale to complex real-world multi-object videos. In an effort to bridge this gap, we take inspiration from human development and hypothesize that information about scene geometry in the form of depth signals can facilitate object-centric learning. We introduce SAVi++, an object-centric video model which is trained to predict depth signals from a slot-based video representation. By further leveraging best practices for model scaling, we are able to train SAVi++ to segment complex dynamic scenes recorded with moving cameras, containing both static and moving objects of diverse appearance on naturalistic backgrounds, without the need for segmentation supervision. Finally, we demonstrate that by using sparse depth signals obtained from LiDAR, SAVi++ is able to learn emergent object segmentation and tracking from videos in the real-world Waymo Open dataset."
                },
                {
                    "title": "Analysis and Extensions of Adversarial Training for Video Classification",
                    "abstract": "Adversarial training (AT) is a simple yet effective defense against adversarial attacks to image classification systems, which is based on augmenting the training set with attacks that maximize the loss. However, the effectiveness of AT as a defense for video classification has not been thoroughly studied. Our first contribution is to show that generating optimal attacks for video requires carefully tuning the attack parameters, especially the step size. Notably, we show that the optimal step size varies linearly with the attack budget. Our second contribution is to show that using a smaller (sub-optimal) attack budget at training time leads to a more robust performance at test time. Based on these findings, we propose three defenses against attacks with variable attack budgets. The first one, Adaptive AT, is a technique where the attack budget is drawn from a distribution that is adapted as training iterations proceed. The second, Curriculum AT, is a technique where the attack budget is increased as training iterations proceed. The third, Generative AT, further couples AT with a denoising generative adversarial network to boost robust performance. Experiments on the UCF101 dataset demonstrate that the proposed methods improve adversarial robustness against multiple attack types."
                },
                {
                    "title": "Unsupervised Semantic Segmentation by Distilling Feature Correspondences",
                    "abstract": "Unsupervised semantic segmentation aims to discover and localize semantically meaningful categories within image corpora without any form of annotation. To solve this task, algorithms must produce features for every pixel that are both semantically meaningful and compact enough to form distinct clusters. Unlike previous works which achieve this with a single end-to-end framework, we propose to separate feature learning from cluster compactification. Empirically, we show that current unsupervised feature learning frameworks already generate dense features whose correlations are semantically consistent. This observation motivates us to design STEGO ($\\textbf{S}$elf-supervised $\\textbf{T}$ransformer with $\\textbf{E}$nergy-based $\\textbf{G}$raph $\\textbf{O}$ptimization), a novel framework that distills unsupervised features into high-quality discrete semantic labels. At the core of STEGO is a novel contrastive loss function that encourages features to form compact clusters while preserving their relationships across the corpora. STEGO yields a significant improvement over the prior state of the art, on both the CocoStuff ($\\textbf{+14 mIoU}$) and Cityscapes ($\\textbf{+9 mIoU}$) semantic segmentation challenges."
                },
                {
                    "title": "Understanding Person Identification Through Gait",
                    "abstract": "Gait recognition is the process of identifying humans from their bipedal locomotion such as walking or running. As such, gait data is privacy sensitive information and should be anonymized where possible. With the rise of higher quality gait recording techniques, such as depth cameras or motion capture suits, an increasing amount of detailed gait data is captured and processed. The introduction and rise of the Metaverse is an example of a potentially popular application scenario in which the gait of users is transferred onto digital avatars. As a first step towards developing effective anonymization techniques for high-quality gait data, we study different aspects of movement data to quantify their contribution to gait recognition. We first extract categories of features from the literature on human gait perception and then design experiments for each category to assess how much the information they contain contributes to recognition success. We evaluated the utility of gait perturbation by means of naturalness ratings in a user study. Our results show that gait anonymization will be challenging, as the data is highly redundant and inter-dependent."
                },
                {
                    "title": "GMFlow: Learning Optical Flow via Global Matching",
                    "abstract": "Learning-based optical flow estimation has been dominated with the pipeline of cost volume with convolutions for flow regression, which is inherently limited to local correlations and thus is hard to address the long-standing challenge of large displacements. To alleviate this, the state-of-the-art framework RAFT gradually improves its prediction quality by using a large number of iterative refinements, achieving remarkable performance but introducing linearly increasing inference time. To enable both high accuracy and efficiency, we completely revamp the dominant flow regression pipeline by reformulating optical flow as a global matching problem, which identifies the correspondences by directly comparing feature similarities. Specifically, we propose a GMFlow framework, which consists of three main components: a customized Transformer for feature enhancement, a correlation and softmax layer for global feature matching, and a self-attention layer for flow propagation. We further introduce a refinement step that reuses GMFlow at higher feature resolution for residual flow prediction. Our new framework outperforms 31-refinements RAFT on the challenging Sintel benchmark, while using only one refinement and running faster, suggesting a new paradigm for accurate and efficient optical flow estimation. Code is available at https://github.com/haofeixu/gmflow."
                },
                {
                    "title": "Conditional Object-Centric Learning from Video",
                    "abstract": "Object-centric representations are a promising path toward more systematic generalization by providing flexible abstractions upon which compositional world models can be built. Recent work on simple 2D and 3D datasets has shown that models with object-centric inductive biases can learn to segment and represent meaningful objects from the statistical structure of the data alone without the need for any supervision. However, such fully-unsupervised methods still fail to scale to diverse realistic data, despite the use of increasingly complex inductive biases such as priors for the size of objects or the 3D geometry of the scene. In this paper, we instead take a weakly-supervised approach and focus on how 1) using the temporal dynamics of video data in the form of optical flow and 2) conditioning the model on simple object location cues can be used to enable segmenting and tracking objects in significantly more realistic synthetic data. We introduce a sequential extension to Slot Attention which we train to predict optical flow for realistic looking synthetic scenes and show that conditioning the initial state of this model on a small set of hints, such as center of mass of objects in the first frame, is sufficient to significantly improve instance segmentation. These benefits generalize beyond the training distribution to novel objects, novel backgrounds, and to longer video sequences. We also find that such initial-state-conditioning can be used during inference as a flexible interface to query the model for specific objects or parts of objects, which could pave the way for a range of weakly-supervised approaches and allow more effective interaction with trained models."
                },
                {
                    "title": "Self-Supervised Video Representation Learning with Meta-Contrastive Network",
                    "abstract": "Self-supervised learning has been successfully applied to pre-train video representations, which aims at efficient adaptation from pre-training domain to downstream tasks. Existing approaches merely leverage contrastive loss to learn instance-level discrimination. However, lack of category information will lead to hard-positive problem that constrains the generalization ability of this kind of methods. We find that the multi-task process of meta learning can provide a solution to this problem. In this paper, we propose a Meta-Contrastive Network (MCN), which combines the contrastive learning and meta learning, to enhance the learning ability of existing self-supervised approaches. Our method contains two training stages based on model-agnostic meta learning (MAML), each of which consists of a contrastive branch and a meta branch. Extensive evaluations demonstrate the effectiveness of our method. For two downstream tasks, i.e., video action recognition and video retrieval, MCN outperforms state-of-the-art approaches on UCF101 and HMDB51 datasets. To be more specific, with R(2+1)D backbone, MCN achieves Top-1 accuracies of 84.8% and 54.5% for video action recognition, as well as 52.5% and 23.7% for video retrieval."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "Multiscale Vision Transformers",
                    "abstract": "We present Multiscale Vision Transformers (MViT) for video and image recognition, by connecting the seminal idea of multiscale feature hierarchies with transformer models. Multiscale Transformers have several channel-resolution scale stages. Starting from the input resolution and a small channel dimension, the stages hierarchically expand the channel capacity while reducing the spatial resolution. This creates a multiscale pyramid of features with early layers operating at high spatial resolution to model simple low-level visual information, and deeper layers at spatially coarse, but complex, high-dimensional features. We evaluate this fundamental architectural prior for modeling the dense nature of visual signals for a variety of video recognition tasks where it outperforms concurrent vision transformers that rely on large scale external pre-training and are 5-10\u00d7 more costly in computation and parameters. We further remove the temporal dimension and apply our model for image classification where it outperforms prior work on vision transformers. Code is available at: https://github.com/facebookresearch/SlowFast."
                },
                {
                    "title": "Self-supervised Video Object Segmentation by Motion Grouping",
                    "abstract": "Animals have evolved highly functional visual systems to understand motion, assisting perception even under complex environments. In this paper, we work towards developing a computer vision system able to segment objects by exploiting motion cues, i.e. motion segmentation. To achieve this, we introduce a simple variant of the Transformer to segment optical flow frames into primary objects and the background, which can be trained in a self-supervised manner, i.e. without using any manual annotations. Despite using only optical flow, and no appearance information, as input, our approach achieves superior results compared to previous state-of-the-art self-supervised methods on public benchmarks (DAVIS2016, SegTrackv2, FBMS59), while being an order of magnitude faster. On a challenging camouflage dataset (MoCA), we significantly outperform other self-supervised approaches, and are competitive with the top supervised approach, highlighting the importance of motion cues and the potential bias towards appearance in existing video segmentation models."
                },
                {
                    "title": "ViViT: A Video Vision Transformer",
                    "abstract": "We present pure-transformer based models for video classification, drawing upon the recent success of such models in image classification. Our model extracts spatiotemporal tokens from the input video, which are then encoded by a series of transformer layers. In order to handle the long sequences of tokens encountered in video, we propose several, efficient variants of our model which factorise the spatial- and temporal-dimensions of the input. Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets. We conduct thorough ablation studies, and achieve state-of-the-art results on multiple video classification benchmarks including Kinetics 400 and 600, Epic Kitchens, Something-Something v2 and Moments in Time, outperforming prior methods based on deep 3D convolutional networks."
                },
                {
                    "title": "Moving the Lab into the Mountains: A Pilot Study of Human Activity Recognition in Unstructured Environments",
                    "abstract": "Goal: To develop and validate a field-based data collection and assessment method for human activity recognition in the mountains with variations in terrain and fatigue using a single accelerometer and a deep learning model. Methods: The protocol generated an unsupervised labelled dataset of various long-term field-based activities including run, walk, stand, lay and obstacle climb. Activity was voluntary so transitions could not be determined a priori. Terrain variations included slope, crossing rivers, obstacles and surfaces including road, gravel, clay, mud, long grass and rough track. Fatigue levels were modulated between rested to physical exhaustion. The dataset was used to train a deep learning convolutional neural network (CNN) capable of being deployed on battery powered devices. The human activity recognition results were compared to a lab-based dataset with 1,098,204 samples and six features, uniform smooth surfaces, non-fatigued supervised participants and activity labelling defined by the protocol. Results: The trail run dataset had 3,829,759 samples with five features. The repetitive activities and single instance activities required hyper parameter tuning to reach an overall accuracy 0.978 with a minimum class precision for the one-off activity (climbing gate) of 0.802. Conclusion: The experimental results showed that the CNN deep learning model performed well with terrain and fatigue variations compared to the lab equivalents (accuracy 97.8% vs. 97.7% for trail vs. lab). Significance: To the authors knowledge this study demonstrated the first successful human activity recognition (HAR) in a mountain environment. A robust and repeatable protocol was developed to generate a validated trail running dataset when there were no observers present and activity types changed on a voluntary basis across variations in terrain surface and both cognitive and physical fatigue levels."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Object-Centric Learning with Slot Attention",
                    "abstract": "Learning object-centric representations of complex scenes is a promising step towards enabling efficient abstract reasoning from low-level perceptual features. Yet, most deep learning approaches learn distributed representations that do not capture the compositional properties of natural scenes. In this paper, we present the Slot Attention module, an architectural component that interfaces with perceptual representations such as the output of a convolutional neural network and produces a set of task-dependent abstract representations which we call slots. These slots are exchangeable and can bind to any object in the input by specializing through a competitive procedure over multiple rounds of attention. We empirically demonstrate that Slot Attention can extract object-centric representations that enable generalization to unseen compositions when trained on unsupervised object discovery and supervised property prediction tasks."
                },
                {
                    "title": "2D Pose-Based Real-Time Human Action Recognition With Occlusion-Handling",
                    "abstract": "Human Action Recognition (HAR) for CCTV-oriented applications is still a challenging problem. Real-world scenarios HAR implementations is difficult because of the gap between Deep Learning data requirements and what the CCTV-based frameworks can offer in terms of data recording equipments. We propose to reduce this gap by exploiting human poses provided by the OpenPose, which has been already proven to be an effective detector in CCTV-like recordings for tracking applications. Therefore, in this work, we first propose ActionXPose: a novel 2D pose-based approach for pose-level HAR. ActionXPose extracts low- and high-level features from body poses which are provided to a Long Short-Term Memory Neural Network and a 1D Convolutional Neural Network for the classification. We also provide a new dataset, named ISLD, for realistic pose-level HAR in a CCTV-like environment, recorded in the Intelligent Sensing Lab. ActionXPose is extensively tested on ISLD under multiple experimental settings, e.g. Dataset Augmentation and Cross-Dataset setting, as well as revising other existing datasets for HAR. ActionXPose achieves state-of-the-art performance in terms of accuracy, very high robustness to occlusions and missing data, and promising results for practical implementation in real-world applications."
                },
                {
                    "title": "X3D: Expanding Architectures for Efficient Video Recognition",
                    "abstract": "This paper presents X3D, a family of efficient video networks that progressively expand a tiny 2D image classification architecture along multiple network axes, in space, time, width and depth. Inspired by feature selection methods in machine learning, a simple stepwise network expansion approach is employed that expands a single axis in each step, such that good accuracy to complexity trade-off is achieved. To expand X3D to a specific target complexity, we perform progressive forward expansion followed by backward contraction. X3D achieves state-of-the-art performance while requiring 4.8x and 5.5x fewer multiply-adds and parameters for similar accuracy as previous work. Our most surprising finding is that networks with high spatiotemporal resolution can perform well, while being extremely light in terms of network width and parameters. We report competitive accuracy at unprecedented efficiency on video classification and detection benchmarks. Code is available at: https://github.com/facebookresearch/SlowFast."
                },
                {
                    "title": "Can Deep Learning Recognize Subtle Human Activities?",
                    "abstract": "Deep Learning has driven recent and exciting progress in computer vision, instilling the belief that these algorithms could solve any visual task. Yet, datasets commonly used to train and test computer vision algorithms have pervasive confounding factors. Such biases make it difficult to truly estimate the performance of those algorithms and how well computer vision models can extrapolate outside the distribution in which they were trained. In this work, we propose a new action classification challenge that is performed well by humans, but poorly by state-of-the-art Deep Learning models. As a proof-of-principle, we consider three exemplary tasks: drinking, reading, and sitting. The best accuracies reached using state-of-the-art computer vision models were 61.7%, 62.8%, and 76.8%, respectively, while human participants scored above 90% accuracy on the three tasks. We propose a rigorous method to reduce confounds when creating datasets, and when comparing human versus computer vision performance. Source code and datasets are publicly available."
                },
                {
                    "title": "Over-the-Air Adversarial Flickering Attacks against Video Recognition Networks",
                    "abstract": "Deep neural networks for video classification, just like image classification networks, may be subjected to adversarial manipulation. The main difference between image classifiers and video classifiers is that the latter usually use temporal information contained within the video. In this work we present a manipulation scheme for fooling video classifiers by introducing a flickering temporal perturbation that in some cases may be unnoticeable by human observers and is implementable in the real world. After demonstrating the manipulation of action classification of single videos, we generalize the procedure to make universal adversarial perturbation, achieving high fooling ratio. In addition, we generalize the universal perturbation and produce a temporal-invariant perturbation, which can be applied to the video without synchronizing the perturbation to the input. The attack was implemented on several target models and the transferability of the attack was demonstrated. These properties allow us to bridge the gap between simulated environment and real-world application, as will be demonstrated in this paper for the first time for an over-the-air flickering attack."
                },
                {
                    "title": "Multi-Modal Domain Adaptation for Fine-Grained Action Recognition",
                    "abstract": "Fine-grained action recognition datasets exhibit environmental bias, where multiple video sequences are captured from a limited number of environments. Training a model in one environment and deploying in another results in a drop in performance due to an unavoidable domain shift. Unsupervised Domain Adaptation (UDA) approaches have frequently utilised adversarial training between the source and target domains. However, these approaches have not explored the multi-modal nature of video within each domain. In this work we exploit the correspondence of modalities as a self-supervised alignment approach for UDA in addition to adversarial alignment (Fig. 1). We test our approach on three kitchens from the large-scale EPIC-Kitchens dataset, using two modalities commonly employed for action recognition: RGB and Optical Flow. We show that multi-modal self-supervision alone improves the performance over source-only training by 2.4% on average. We then combine adversarial training with multi-modal self-supervision, showing that our approach outperforms other UDA methods by 3%."
                },
                {
                    "title": "SPACE: Unsupervised Object-Oriented Scene Representation via Spatial Attention and Decomposition",
                    "abstract": "The ability to decompose complex multi-object scenes into meaningful abstractions like objects is fundamental to achieve higher-level cognition. Previous approaches for unsupervised object-oriented scene representation learning are either based on spatial-attention or scene-mixture approaches and limited in scalability which is a main obstacle towards modeling real-world scenes. In this paper, we propose a generative latent variable model, called SPACE, that provides a unified probabilistic modeling framework that combines the best of spatial-attention and scene-mixture approaches. SPACE can explicitly provide factorized object representations for foreground objects while also decomposing background segments of complex morphology. Previous models are good at either of these, but not both. SPACE also resolves the scalability problems of previous methods by incorporating parallel spatial-attention and thus is applicable to scenes with a large number of objects without performance degradations. We show through experiments on Atari and 3D-Rooms that SPACE achieves the above properties consistently in comparison to SPAIR, IODINE, and GENESIS. Results of our experiments can be found on our project website: this https URL"
                },
                {
                    "title": "Adversarial Cross-Domain Action Recognition with Co-Attention",
                    "abstract": "Action recognition has been a widely studied topic with a heavy focus on supervised learning involving sufficient labeled videos. However, the problem of cross-domain action recognition, where training and testing videos are drawn from different underlying distributions, remains largely under-explored. Previous methods directly employ techniques for cross-domain image recognition, which tend to suffer from the severe temporal misalignment problem. This paper proposes a Temporal Co-attention Network (TCoN), which matches the distributions of temporally aligned action features between source and target domains using a novel cross-domain co-attention mechanism. Experimental results on three cross-domain action recognition datasets demonstrate that TCoN improves both previous single-domain and cross-domain methods significantly under the cross-domain setting."
                },
                {
                    "title": "Temporal Attentive Alignment for Large-Scale Video Domain Adaptation",
                    "abstract": "Although various image-based domain adaptation (DA) techniques have been proposed in recent years, domain shift in videos is still not well-explored. Most previous works only evaluate performance on small-scale datasets which are saturated. Therefore, we first propose two large-scale video DA datasets with much larger domain discrepancy: UCF-HMDB_full and Kinetics-Gameplay. Second, we investigate different DA integration methods for videos, and show that simultaneously aligning and learning temporal dynamics achieves effective alignment even without sophisticated DA methods. Finally, we propose Temporal Attentive Adversarial Adaptation Network (TA3N), which explicitly attends to the temporal dynamics using domain discrepancy for more effective domain alignment, achieving state-of-the-art performance on four video DA datasets (e.g. 7.9% accuracy gain over \u201cSource only\u201d from 73.9% to 81.8% on \u201cHMDB --> UCF\u201d, and 10.3% gain on \u201cKinetics --> Gameplay\u201d). The code and data are released at http://github.com/cmhungsteve/TA3N."
                },
                {
                    "title": "Action4D: Online Action Recognition in the Crowd and Clutter",
                    "abstract": "Recognizing every person's action in a crowded and cluttered environment is a challenging task in computer vision. We propose to tackle this challenging problem using a holistic 4D ``scan'' of a cluttered scene to include every detail about the people and environment. This leads to a new problem, i.e., recognizing multiple people's actions in the cluttered 4D representation. At the first step, we propose a new method to track people in 4D, which can reliably detect and follow each person in real time. Then, we build a new deep neural network, the Action4DNet, to recognize the action of each tracked person. Such a model gives reliable and accurate results in the real-world settings. We also design an adaptive 3D convolution layer and a novel discriminative temporal feature learning objective to further improve the performance of our model. Our method is invariant to camera view angles, resistant to clutter and able to handle crowd. The experimental results show that the proposed method is fast, reliable and accurate. Our method paves the way to action recognition in the real-world applications and is ready to be deployed to enable smart homes, smart factories and smart stores."
                },
                {
                    "title": "NTU RGB+D 120: A Large-Scale Benchmark for 3D Human Activity Understanding",
                    "abstract": "Research on depth-based human activity analysis achieved outstanding performance and demonstrated the effectiveness of 3D representation for action recognition. The existing depth-based and RGB+D-based action recognition benchmarks have a number of limitations, including the lack of large-scale training samples, realistic number of distinct class categories, diversity in camera views, varied environmental conditions, and variety of human subjects. In this work, we introduce a large-scale dataset for RGB+D human action recognition, which is collected from 106 distinct subjects and contains more than 114 thousand video samples and 8 million frames. This dataset contains 120 different action classes including daily, mutual, and health-related activities. We evaluate the performance of a series of existing 3D activity analysis methods on this dataset, and show the advantage of applying deep learning methods for 3D-based human action recognition. Furthermore, we investigate a novel one-shot 3D activity recognition problem on our dataset, and a simple yet effective Action-Part Semantic Relevance-aware (APSR) framework is proposed for this task, which yields promising results for recognition of the novel action classes. We believe the introduction of this large-scale dataset will enable the community to apply, adapt, and develop various data-hungry learning techniques for depth-based and RGB+D-based human activity understanding."
                },
                {
                    "title": "SlowFast Networks for Video Recognition",
                    "abstract": "We present SlowFast networks for video recognition. Our model involves (i) a Slow pathway, operating at low frame rate, to capture spatial semantics, and (ii) a Fast pathway, operating at high frame rate, to capture motion at fine temporal resolution. The Fast pathway can be made very lightweight by reducing its channel capacity, yet can learn useful temporal information for video recognition. Our models achieve strong performance for both action classification and detection in video, and large improvements are pin-pointed as contributions by our SlowFast concept. We report state-of-the-art accuracy on major video recognition benchmarks, Kinetics, Charades and AVA. Code has been made available at: https://github.com/facebookresearch/SlowFast."
                },
                {
                    "title": "Representation Flow for Action Recognition",
                    "abstract": "In this paper, we propose a convolutional layer inspired by optical flow algorithms to learn motion representations. Our representation flow layer is a fully-differentiable layer designed to capture the `flow' of any representation channel within a convolutional neural network for action recognition. Its parameters for iterative flow optimization are learned in an end-to-end fashion together with the other CNN model parameters, maximizing the action recognition performance. Furthermore, we newly introduce the concept of learning `flow of flow' representations by stacking multiple representation flow layers. We conducted extensive experimental evaluations, confirming its advantages over previous recognition models using traditional optical flows in both computational speed and performance. The code is publicly available."
                },
                {
                    "title": "A fast, invariant representation for human action in the visual system.",
                    "abstract": "Humans can effortlessly recognize others' actions in the presence of complex transformations, such as changes in viewpoint. Several studies have located the regions in the brain involved in invariant action recognition; however, the underlying neural computations remain poorly understood. We use magnetoencephalography decoding and a data set of well-controlled, naturalistic videos of five actions (run, walk, jump, eat, drink) performed by different actors at different viewpoints to study the computational steps used to recognize actions across complex transformations. In particular, we ask when the brain discriminates between different actions, and when it does so in a manner that is invariant to changes in 3D viewpoint. We measure the latency difference between invariant and noninvariant action decoding when subjects view full videos as well as form-depleted and motion-depleted stimuli. We were unable to detect a difference in decoding latency or temporal profile between invariant and noninvariant action recognition in full videos. However, when either form or motion information is removed from the stimulus set, we observe a decrease and delay in invariant action decoding. Our results suggest that the brain recognizes actions and builds invariance to complex transformations at the same time and that both form and motion information are crucial for fast, invariant action recognition. NEW & NOTEWORTHY The human brain can quickly recognize actions despite transformations that change their visual appearance. We use neural timing data to uncover the computations underlying this ability. We find that within 200 ms action can be read out of magnetoencephalography data and that this representation is invariant to changes in viewpoint. We find form and motion are needed for this fast action decoding, suggesting that the brain quickly integrates complex spatiotemporal features to form invariant action representations."
                },
                {
                    "title": "A Closer Look at Spatiotemporal Convolutions for Action Recognition",
                    "abstract": "In this paper we discuss several forms of spatiotemporal convolutions for video analysis and study their effects on action recognition. Our motivation stems from the observation that 2D CNNs applied to individual frames of the video have remained solid performers in action recognition. In this work we empirically demonstrate the accuracy advantages of 3D CNNs over 2D CNNs within the framework of residual learning. Furthermore, we show that factorizing the 3D convolutional filters into separate spatial and temporal components yields significantly gains in accuracy. Our empirical study leads to the design of a new spatiotemporal convolutional block \"R(2+1)D\" which produces CNNs that achieve results comparable or superior to the state-of-the-art on Sports-1M, Kinetics, UCF101, and HMDB51."
                },
                {
                    "title": "Optical Flow Guided Feature: A Fast and Robust Motion Representation for Video Action Recognition",
                    "abstract": "Motion representation plays a vital role in human action recognition in videos. In this study, we introduce a novel compact motion representation for video action recognition, named Optical Flow guided Feature (OFF), which enables the network to distill temporal information through a fast and robust approach. The OFF is derived from the definition of optical flow and is orthogonal to the optical flow. The derivation also provides theoretical support for using the difference between two frames. By directly calculating pixel-wise spatio-temporal gradients of the deep feature maps, the OFF could be embedded in any existing CNN based video action recognition framework with only a slight additional cost. It enables the CNN to extract spatiotemporal information, especially the temporal information between frames simultaneously. This simple but powerful idea is validated by experimental results. The network with OFF fed only by RGB inputs achieves a competitive accuracy of 93.3% on UCF-101, which is comparable with the result obtained by two streams (RGB and optical flow), but is 15 times faster in speed. Experimental results also show that OFF is complementary to other motion modalities such as optical flow. When the proposed method is plugged into the state-of-the-art video action recognition framework, it has 96.0% and 74.2% accuracy on UCF-101 and HMDB-51 respectively. The code for this project is available at: https://github.com/kevin-ssy/Optical-Flow-Guided-Feature"
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Learning Spatio-Temporal Features with 3D Residual Networks for Action Recognition",
                    "abstract": "Convolutional neural networks with spatio-temporal 3D kernels (3D CNNs) have an ability to directly extract spatiotemporal features from videos for action recognition. Although the 3D kernels tend to overfit because of a large number of their parameters, the 3D CNNs are greatly improved by using recent huge video databases. However, the architecture of3D CNNs is relatively shallow against to the success of very deep neural networks in 2D-based CNNs, such as residual networks (ResNets). In this paper, we propose a 3D CNNs based on ResNets toward a better action representation. We describe the training procedure of our 3D ResNets in details. We experimentally evaluate the 3D ResNets on the ActivityNet and Kinetics datasets. The 3D ResNets trained on the Kinetics did not suffer from overfitting despite the large number of parameters of the model, and achieved better performance than relatively shallow networks, such as C3D. Our code and pretrained models (e.g. Kinetics and ActivityNet) are publicly available at https://github.com/kenshohara/3D-ResNets."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset",
                    "abstract": "The paucity of videos in current action classification datasets (UCF-101 and HMDB-51) has made it difficult to identify good video architectures, as most methods obtain similar performance on existing small-scale benchmarks. This paper re-evaluates state-of-the-art architectures in light of the new Kinetics Human Action Video dataset. Kinetics has two orders of magnitude more data, with 400 human action classes and over 400 clips per class, and is collected from realistic, challenging YouTube videos. We provide an analysis on how current architectures fare on the task of action classification on this dataset and how much performance improves on the smaller benchmark datasets after pre-training on Kinetics. We also introduce a new Two-Stream Inflated 3D ConvNet (I3D) that is based on 2D ConvNet inflation: filters and pooling kernels of very deep image classification ConvNets are expanded into 3D, making it possible to learn seamless spatio-temporal feature extractors from video while leveraging successful ImageNet architecture designs and even their parameters. We show that, after pre-training on Kinetics, I3D models considerably improve upon the state-of-the-art in action classification, reaching 80.2% on HMDB-51 and 97.9% on UCF-101."
                },
                {
                    "title": "Causal Action: A Fundamental Constraint on Perception and Inference About Body Movements",
                    "abstract": "The human body navigates the environment via locomotory movements that leverage gravity and limb biomechanics to propel the body in a particular direction. This process creates a causal link between limb movements and whole-body translation. However, it is unknown whether humans use this causal relation as a constraint in perception and inference with body movements. In the present study, participants rated actions of other individuals as more natural when limb movements (as a cause) occurred before body displacements (as an effect) than when limb movements temporally lagged behind body displacements. This causal expectation for human body movements not only affected perceptual impressions regarding the naturalness of observed actions but also guided the interpretation of motion cues within a more generalized causal context. We interpret these results within a framework of causality as evidence that the constraint of causal action plays an important role in perception and inference with body movements."
                },
                {
                    "title": "SGDR: Stochastic Gradient Descent with Warm Restarts",
                    "abstract": "Restart techniques are common in gradient-free optimization to deal with multimodal functions. Partial warm restarts are also gaining popularity in gradient-based optimization to improve the rate of convergence in accelerated gradient schemes to deal with ill-conditioned functions. In this paper, we propose a simple warm restart technique for stochastic gradient descent to improve its anytime performance when training deep neural networks. We empirically study its performance on the CIFAR-10 and CIFAR-100 datasets, where we demonstrate new state-of-the-art results at 3.14% and 16.21%, respectively. We also demonstrate its advantages on a dataset of EEG recordings and on a downsampled version of the ImageNet dataset. Our source code is available at this https URL"
                },
                {
                    "title": "Convolutional Two-Stream Network Fusion for Video Action Recognition",
                    "abstract": "Recent applications of Convolutional Neural Networks (ConvNets) for human action recognition in videos have proposed different solutions for incorporating the appearance and motion information. We study a number of ways of fusing ConvNet towers both spatially and temporally in order to best take advantage of this spatio-temporal information. We make the following findings: (i) that rather than fusing at the softmax layer, a spatial and temporal network can be fused at a convolution layer without loss of performance, but with a substantial saving in parameters, (ii) that it is better to fuse such networks spatially at the last convolutional layer than earlier, and that additionally fusing at the class prediction layer can boost accuracy, finally (iii) that pooling of abstract convolutional features over spatiotemporal neighbourhoods further boosts performance. Based on these studies we propose a new ConvNet architecture for spatiotemporal fusion of video snippets, and evaluate its performance on standard benchmarks where this architecture achieves state-of-the-art results."
                },
                {
                    "title": "NTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysis",
                    "abstract": "Recent approaches in depth-based human activity analysis achieved outstanding performance and proved the effectiveness of 3D representation for classification of action classes. Currently available depth-based and RGB+Dbased action recognition benchmarks have a number of limitations, including the lack of training samples, distinct class labels, camera views and variety of subjects. In this paper we introduce a large-scale dataset for RGB+D human action recognition with more than 56 thousand video samples and 4 million frames, collected from 40 distinct subjects. Our dataset contains 60 different action classes including daily, mutual, and health-related actions. In addition, we propose a new recurrent neural network structure to model the long-term temporal correlation of the features for each body part, and utilize them for better action classification. Experimental results show the advantages of applying deep learning methods over state-of-the-art handcrafted features on the suggested cross-subject and cross-view evaluation criteria for our dataset. The introduction of this large scale dataset will enable the community to apply, develop and adapt various data-hungry learning techniques for the task of depth-based and RGB+D-based human activity analysis."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Robot-Centric Activity Prediction from First-Person Videos: What Will They Do to Me?",
                    "abstract": "In this paper, we present a core technology to enable robot recognition of human activities during human-robot interactions. In particular, we propose a methodology for early recognition of activities from robot-centric videos (i.e., first-person videos) obtained from a robot's viewpoint during its interaction with humans. Early recognition, which is also known as activity prediction, is an ability to infer an ongoing activity at its early stage. We present an algorithm to recognize human activities targeting the camera from streaming videos, enabling the robot to predict intended activities of the interacting person as early as possible and take fast reactions to such activities (e.g., avoiding harmful events targeting itself before they actually occur). We introduce the novel concept of'onset' that efficiently summarizes pre-activity observations, and design a recognition approach to consider event history in addition to visual features from first-person videos. We propose to represent an onset using a cascade histogram of time series gradients, and we describe a novel algorithmic setup to take advantage of such onset for early recognition of activities. The experimental results clearly illustrate that the proposed concept of onset enables better/earlier recognition of human activities from first-person videos collected with a robot. Categories and Subject Descriptors I.2.10 [Artificial Intelligence]: Vision and Scene Understanding\u2013video analysis; I.4.8 [Image Processing and Computer Vision]: Scene Analysis-motion; I.2.9 [Artificial Intelligence]: Robotics\u2013sensors"
                },
                {
                    "title": "Learning Spatiotemporal Features with 3D Convolutional Networks",
                    "abstract": "We propose a simple, yet effective approach for spatiotemporal feature learning using deep 3-dimensional convolutional networks (3D ConvNets) trained on a large scale supervised video dataset. Our findings are three-fold: 1) 3D ConvNets are more suitable for spatiotemporal feature learning compared to 2D ConvNets, 2) A homogeneous architecture with small 3x3x3 convolution kernels in all layers is among the best performing architectures for 3D ConvNets, and 3) Our learned features, namely C3D (Convolutional 3D), with a simple linear classifier outperform state-of-the-art methods on 4 different benchmarks and are comparable with current best methods on the other 2 benchmarks. In addition, the features are compact: achieving 52.8% accuracy on UCF101 dataset with only 10 dimensions and also very efficient to compute due to the fast inference of ConvNets. Finally, they are conceptually very simple and easy to train and use."
                },
                {
                    "title": "Beyond Gaussian Pyramid: Multi-skip Feature Stacking for action recognition",
                    "abstract": "Most state-of-the-art action feature extractors involve differential operators, which act as highpass filters and tend to attenuate low frequency action information. This attenuation introduces bias to the resulting features and generates ill-conditioned feature matrices. The Gaussian Pyramid has been used as a feature enhancing technique that encodes scale-invariant characteristics into the feature space in an attempt to deal with this attenuation. However, at the core of the Gaussian Pyramid is a convolutional smoothing operation, which makes it incapable of generating new features at coarse scales. In order to address this problem, we propose a novel feature enhancing technique called Multi-skIp Feature Stacking (MIFS), which stacks features extracted using a family of differential filters parameterized with multiple time skips and encodes shift-invariance into the frequency space. MIFS compensates for information lost from using differential operators by recapturing information at coarse scales. This recaptured information allows us to match actions at different speeds and ranges of motion. We prove that MIFS enhances the learnability of differential-based features exponentially. The resulting feature matrices from MIFS have much smaller conditional numbers and variances than those from conventional methods. Experimental results show significantly improved performance on challenging action recognition and event detection tasks. Specifically, our method exceeds the state-of-the-arts on Hollywood2, UCF101 and UCF50 datasets and is comparable to state-of-the-arts on HMDB51 and Olympics Sports datasets. MIFS can also be used as a speedup strategy for feature extraction with minimal or no accuracy cost."
                },
                {
                    "title": "Perception of Social Interactions for Spatially Scrambled Biological Motion",
                    "abstract": "It is vitally important for humans to detect living creatures in the environment and to analyze their behavior to facilitate action understanding and high-level social inference. The current study employed naturalistic point-light animations to examine the ability of human observers to spontaneously identify and discriminate socially interactive behaviors between two human agents. Specifically, we investigated the importance of global body form, intrinsic joint movements, extrinsic whole-body movements, and critically, the congruency between intrinsic and extrinsic motions. Motion congruency is hypothesized to be particularly important because of the constraint it imposes on naturalistic action due to the inherent causal relationship between limb movements and whole body motion. Using a free response paradigm in Experiment 1, we discovered that many na\u00efve observers (55%) spontaneously attributed animate and/or social traits to spatially-scrambled displays of interpersonal interaction. Total stimulus motion energy was strongly correlated with the likelihood that an observer would attribute animate/social traits, as opposed to physical/mechanical traits, to the scrambled dot stimuli. In Experiment 2, we found that participants could identify interactions between spatially-scrambled displays of human dance as long as congruency was maintained between intrinsic/extrinsic movements. Violating the motion congruency constraint resulted in chance discrimination performance for the spatially-scrambled displays. Finally, Experiment 3 showed that scrambled point-light dancing animations violating this constraint were also rated as significantly less interactive than animations with congruent intrinsic/extrinsic motion. These results demonstrate the importance of intrinsic/extrinsic motion congruency for biological motion analysis, and support a theoretical framework in which early visual filters help to detect animate agents in the environment based on several fundamental constraints. Only after satisfying these basic constraints could stimuli be evaluated for high-level social content. In this way, we posit that perceptual animacy may serve as a gateway to higher-level processes that support action understanding and social inference."
                },
                {
                    "title": "Prediction of Human Activity by Discovering Temporal Sequence Patterns",
                    "abstract": "Early prediction of ongoing human activity has become more valuable in a large variety of time-critical applications. To build an effective representation for prediction, human activities can be characterized by a complex temporal composition of constituent simple actions and interacting objects. Different from early detection on short-duration simple actions, we propose a novel framework for long -duration complex activity prediction by discovering three key aspects of activity: Causality, Context-cue, and Predictability. The major contributions of our work include: (1) a general framework is proposed to systematically address the problem of complex activity prediction by mining temporal sequence patterns; (2) probabilistic suffix tree (PST) is introduced to model causal relationships between constituent actions, where both large and small order Markov dependencies between action units are captured; (3) the context-cue, especially interactive objects information, is modeled through sequential pattern mining (SPM), where a series of action and object co-occurrence are encoded as a complex symbolic sequence; (4) we also present a predictive accumulative function (PAF) to depict the predictability of each kind of activity. The effectiveness of our approach is evaluated on two experimental scenarios with two data sets for each: action-only prediction and context-aware prediction. Our method achieves superior performance for predicting global activity classes and local action units."
                },
                {
                    "title": "Two-Stream Convolutional Networks for Action Recognition in Videos",
                    "abstract": "We investigate architectures of discriminatively trained deep Convolutional Networks (ConvNets) for action recognition in video. The challenge is to capture the complementary information on appearance from still frames and motion between frames. We also aim to generalise the best performing hand-crafted features within a data-driven learning framework. \n \nOur contribution is three-fold. First, we propose a two-stream ConvNet architecture which incorporates spatial and temporal networks. Second, we demonstrate that a ConvNet trained on multi-frame dense optical flow is able to achieve very good performance in spite of limited training data. Finally, we show that multitask learning, applied to two different action classification datasets, can be used to increase the amount of training data and improve the performance on both. Our architecture is trained and evaluated on the standard video actions benchmarks of UCF-101 and HMDB-51, where it is competitive with the state of the art. It also exceeds by a large margin previous attempts to use deep nets for video classification."
                },
                {
                    "title": "Learning Phrase Representations using RNN Encoder\u2013Decoder for Statistical Machine Translation",
                    "abstract": "In this paper, we propose a novel neural network model called RNN Encoder\u2010 Decoder that consists of two recurrent neural networks (RNN). One RNN encodes a sequence of symbols into a fixedlength vector representation, and the other decodes the representation into another sequence of symbols. The encoder and decoder of the proposed model are jointly trained to maximize the conditional probability of a target sequence given a source sequence. The performance of a statistical machine translation system is empirically found to improve by using the conditional probabilities of phrase pairs computed by the RNN Encoder\u2010Decoder as an additional feature in the existing log-linear model. Qualitatively, we show that the proposed model learns a semantically and syntactically meaningful representation of linguistic phrases."
                },
                {
                    "title": "Cross-View Action Modeling, Learning, and Recognition",
                    "abstract": "Existing methods on video-based action recognition are generally view-dependent, i.e., performing recognition from the same views seen in the training data. We present a novel multiview spatio-temporal and-or graph (MST-AOG) representation for cross-view action recognition, i.e., the recognition is performed on the video from an unknown and unseen view. As a compositional model, MST-AOG compactly represents the hierarchical combinatorial structures of cross-view actions by explicitly modeling the geometry, appearance and motion variations. This paper proposes effective methods to learn the structure and parameters of MST-AOG. The inference based on MST-AOG enables action recognition from novel views. The training of MST-AOG takes advantage of the 3D human skeleton data obtained from Kinect cameras to avoid annotating enormous multi-view video frames, which is error-prone and time-consuming, but the recognition does not need 3D information and is based on 2D video input. A new Multiview Action3D dataset has been created and will be released. Extensive experiments have demonstrated that this new action representation significantly improves the accuracy and robustness for cross-view action recognition on 2D videos."
                },
                {
                    "title": "A View-Invariant Action Recognition Based on Multi-View Space Hidden Markov Models",
                    "abstract": "Visual-based action recognition has already been widely used in human-machine interfaces. However it is a challenging research to recognise the human actions from different viewpoints. In order to solve this issue, a novel multi-view space Hidden Markov Models (HMMs) algorithm for view-invariant action recognition is proposed. Firstly a view-insensitive feature representation by combining the bag-of-words of interest point with the amplitude histogram of optical flow is utilised for describing the human action sequences. The combined features could not only solve the problem that there was no possibility in establishing an association between traditional bag-of-words of interest point method and HMMs, but also greatly reduce the redundancy in the video. Secondly the view space is partitioned into multiple sub-view space according to the camera rotation viewpoint. Human action models are trained by HMMs algorithm in each sub-view space. By computing the probabilities of the test sequence (i.e. observation sequence) for the given multi-view space HMMs, the similarity between the sub-view space and the test sequence viewpoint are analysed during the recognition process. Finally the action with unknown viewpoint is recognised via the probability weighted combination. The experimental results on multi-view action dataset IXMAS demonstrate that the proposed approach is highly efficient and effective in view-invariant action recognition."
                },
                {
                    "title": "Anticipating Human Activities Using Object Affordances for Reactive Robotic Response",
                    "abstract": "An important aspect of human perception is anticipation, which we use extensively in our day-to-day activities when interacting with other humans as well as with our surroundings. Anticipating which activities will a human do next (and how) can enable an assistive robot to plan ahead for reactive responses. Furthermore, anticipation can even improve the detection accuracy of past activities. The challenge, however, is two-fold: We need to capture the rich context for modeling the activities and object affordances, and we need to anticipate the distribution over a large space of future human activities. In this work, we represent each possible future using an anticipatory temporal conditional random field (ATCRF) that models the rich spatial-temporal relations through object affordances. We then consider each ATCRF as a particle and represent the distribution over the potential futures using a set of particles. In extensive evaluation on CAD-120 human activity RGB-D dataset, we first show that anticipation improves the state-of-the-art detection results. We then show that for new subjects (not seen in the training set), we obtain an activity anticipation accuracy (defined as whether one of top three predictions actually happened) of 84.1, 74.4 and 62.2 percent for an anticipation time of 1, 3 and 10 seconds respectively. Finally, we also show a robot using our algorithm for performing a few reactive responses."
                },
                {
                    "title": "Biological motion processing as a hallmark of social cognition.",
                    "abstract": "Visual processing of biological motion (BM) produced by living organisms is of immense value for successful daily-life activities and, in particular, for adaptive social behavior and nonverbal communication. Investigation of BM perception in neurodevelopmental disorders related to autism, preterm birth, and genetic conditions substantially contributes to our understanding of the neural mechanisms underpinning the extraordinary tuning to BM. The most prominent research outcome is that patients with daily-life deficits in social cognition are also compromised on visual body motion processing. This raises the question of whether performance on body motion perception tasks may serve a hallmark of social cognition. Overall, the findings highlight the role of structural and functional brain connectivity for proper functioning of the neural circuitry involved in BM processing and visual social cognition that share topographically and dynamically overlapping neural networks."
                },
                {
                    "title": "Parsing video events with goal inference and intent prediction",
                    "abstract": "In this paper, we present an event parsing algorithm based on Stochastic Context Sensitive Grammar (SCSG) for understanding events, inferring the goal of agents, and predicting their plausible intended actions. The SCSG represents the hierarchical compositions of events and the temporal relations between the sub-events. The alphabets of the SCSG are atomic actions which are defined by the poses of agents and their interactions with objects in the scene. The temporal relations are used to distinguish events with similar structures, interpolate missing portions of events, and are learned from the training data. In comparison with existing methods, our paper makes the following contributions. i) We define atomic actions by a set of relations based on the fluents of agents and their interactions with objects in the scene. ii) Our algorithm handles events insertion and multi-agent events, keeps all possible interpretations of the video to preserve the ambiguities, and achieves the globally optimal parsing solution in a Bayesian framework; iii) The algorithm infers the goal of the agents and predicts their intents by a top-down process; iv) The algorithm improves the detection of atomic actions by event contexts. We show satisfactory results of event recognition and atomic action detection on the data set we captured which contains 12 event categories in both indoor and outdoor videos."
                },
                {
                    "title": "Human activity prediction: Early recognition of ongoing activities from streaming videos",
                    "abstract": "In this paper, we present a novel approach of human activity prediction. Human activity prediction is a probabilistic process of inferring ongoing activities from videos only containing onsets (i.e. the beginning part) of the activities. The goal is to enable early recognition of unfinished activities as opposed to the after-the-fact classification of completed activities. Activity prediction methodologies are particularly necessary for surveillance systems which are required to prevent crimes and dangerous activities from occurring. We probabilistically formulate the activity prediction problem, and introduce new methodologies designed for the prediction. We represent an activity as an integral histogram of spatio-temporal features, efficiently modeling how feature distributions change over time. The new recognition methodology named dynamic bag-of-words is developed, which considers sequential nature of human activities while maintaining advantages of the bag-of-words to handle noisy observations. Our experiments confirm that our approach reliably recognizes ongoing activities from streaming videos with a high accuracy."
                },
                {
                    "title": "Unstructured human activity detection from RGBD images",
                    "abstract": "Being able to detect and recognize human activities is essential for several applications, including personal assistive robotics. In this paper, we perform detection and recognition of unstructured human activity in unstructured environments. We use a RGBD sensor (Microsoft Kinect) as the input sensor, and compute a set of features based on human pose and motion, as well as based on image and point-cloud information. Our algorithm is based on a hierarchical maximum entropy Markov model (MEMM), which considers a person's activity as composed of a set of sub-activities. We infer the two-layered graph structure using a dynamic programming approach. We test our algorithm on detecting and recognizing twelve different activities performed by four people in different environments, such as a kitchen, a living room, an office, etc., and achieve good performance even when the person was not seen before in the training set1."
                },
                {
                    "title": "Structural processing in biological motion perception.",
                    "abstract": "To investigate the basis for biological motion perception, structural and motion information were manipulated independently in a dynamic display using a novel stimulus with multiple apertures. Performance was compared in discrimination of global motion (translation and rotation) and biological motion. When structural information in the display was eliminated but motion information was intact, human observers were able to perceive global motion yet were at chance in discriminating walking direction of biological movement. In contrast, when the display provided even noisy and impoverished structural information, walking direction became identifiable. The present findings thus provide direct psychophysical evidence that motion information is insufficient and structural information is necessary for the identification of walking direction in biological movement. These findings imply that computational models must utilize a structural representation of the human body to account for perception of biological movements."
                },
                {
                    "title": "Understanding the difficulty of training deep feedforward neural networks",
                    "abstract": "Whereas before 2006 it appears that deep multilayer neural networks were not successfully trained, since then several algorithms have been shown to successfully train them, with experimental results showing the superiority of deeper vs less deep architectures. All these experimental results were obtained with new initialization or training mechanisms. Our objective here is to understand better why standard gradient descent from random initialization is doing so poorly with deep neural networks, to better understand these recent relative successes and help design better algorithms in the future. We first observe the influence of the non-linear activations functions. We find that the logistic sigmoid activation is unsuited for deep networks with random initialization because of its mean value, which can drive especially the top hidden layer into saturation. Surprisingly, we find that saturated units can move out of saturation by themselves, albeit slowly, and explaining the plateaus sometimes seen when training neural networks. We find that a new non-linearity that saturates less can often be beneficial. Finally, we study how activations and gradients vary across layers and during training, with the idea that training may be more difficult when the singular values of the Jacobian associated with each layer are far from 1. Based on these considerations, we propose a new initialization scheme that brings substantially faster convergence. 1 Deep Neural Networks Deep learning methods aim at learning feature hierarchies with features from higher levels of the hierarchy formed by the composition of lower level features. They include Appearing in Proceedings of the 13 International Conference on Artificial Intelligence and Statistics (AISTATS) 2010, Chia Laguna Resort, Sardinia, Italy. Volume 9 of JMLR: WC Weston et al., 2008). Much attention has recently been devoted to them (see (Bengio, 2009) for a review), because of their theoretical appeal, inspiration from biology and human cognition, and because of empirical success in vision (Ranzato et al., 2007; Larochelle et al., 2007; Vincent et al., 2008) and natural language processing (NLP) (Collobert & Weston, 2008; Mnih & Hinton, 2009). Theoretical results reviewed and discussed by Bengio (2009), suggest that in order to learn the kind of complicated functions that can represent high-level abstractions (e.g. in vision, language, and other AI-level tasks), one may need deep architectures. Most of the recent experimental results with deep architecture are obtained with models that can be turned into deep supervised neural networks, but with initialization or training schemes different from the classical feedforward neural networks (Rumelhart et al., 1986). Why are these new algorithms working so much better than the standard random initialization and gradient-based optimization of a supervised training criterion? Part of the answer may be found in recent analyses of the effect of unsupervised pretraining (Erhan et al., 2009), showing that it acts as a regularizer that initializes the parameters in a \u201cbetter\u201d basin of attraction of the optimization procedure, corresponding to an apparent local minimum associated with better generalization. But earlier work (Bengio et al., 2007) had shown that even a purely supervised but greedy layer-wise procedure would give better results. So here instead of focusing on what unsupervised pre-training or semi-supervised criteria bring to deep architectures, we focus on analyzing what may be going wrong with good old (but deep) multilayer neural networks. Our analysis is driven by investigative experiments to monitor activations (watching for saturation of hidden units) and gradients, across layers and across training iterations. We also evaluate the effects on these of choices of activation function (with the idea that it might affect saturation) and initialization procedure (since unsupervised pretraining is a particular form of initialization and it has a drastic impact)."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Semantic-Based Surveillance Video Retrieval",
                    "abstract": "Visual surveillance produces large amounts of video data. Effective indexing and retrieval from surveillance video databases are very important. Although there are many ways to represent the content of video clips in current video retrieval algorithms, there still exists a semantic gap between users and retrieval systems. Visual surveillance systems supply a platform for investigating semantic-based video retrieval. In this paper, a semantic-based video retrieval framework for visual surveillance is proposed. A cluster-based tracking algorithm is developed to acquire motion trajectories. The trajectories are then clustered hierarchically using the spatial and temporal information, to learn activity models. A hierarchical structure of semantic indexing and retrieval of object activities, where each individual activity automatically inherits all the semantic descriptions of the activity model to which it belongs, is proposed for accessing video clips and individual objects at the semantic level. The proposed retrieval framework supports various queries including queries by keywords, multiple object queries, and queries by sketch. For multiple object queries, succession and simultaneity restrictions, together with depth and breadth first orders, are considered. For sketch-based queries, a method for matching trajectories drawn by users to spatial trajectories is proposed. The effectiveness and efficiency of our framework are tested in a crowded traffic scene"
                },
                {
                    "title": "A Model of Biological Motion Perception from Configural Form Cues",
                    "abstract": "Biological motion perception is the compelling ability of the visual system to perceive complex human movements effortlessly and within a fraction of a second. Recent neuroimaging and neurophysiological studies have revealed that the visual perception of biological motion activates a widespread network of brain areas. The superior temporal sulcus has a crucial role within this network. The roles of other areas are less clear. We present a computational model based on neurally plausible assumptions to elucidate the contributions of motion and form signals to biological motion perception and the computations in the underlying brain network. The model simulates receptive fields for images of the static human body, as found by neuroimaging studies, and temporally integrates their responses by leaky integrator neurons. The model reveals a high correlation to data obtained by neurophysiological, neuroimaging, and psychophysical studies."
                },
                {
                    "title": "Gender recognition from point-light walkers.",
                    "abstract": "Point-light displays of human gait provide information sufficient to recognize the gender of a walker and are taken as evidence of the exquisite tuning of the visual system to biological motion. The authors revisit this topic with the goals of quantifying human efficiency at gender recognition. To achieve this, the authors first derive an ideal observer for gender recognition on the basis of center of moment (J. E. Cutting, D. R. Proffitt, & L. T. Kozlowski, 1978) and, with the use of anthropometric data from various populations, show optimal recognition of approximately 79% correct. Next, they perform a meta-analysis of 21 experiments examining gender recognition, obtaining accuracies of 66% correct for a side view and 71% for other views. Finally, results of the meta-analysis and the ideal observer are combined to obtain estimates of human efficiency at gender recognition of 26% for the side view and 47% for other views."
                },
                {
                    "title": "Self recognition versus recognition of others by biological motion: viewpoint-dependent effects.",
                    "abstract": "We investigated the influence of viewing angle on performance in recognising the identity of one's own person and familiar individuals such as friends or colleagues from walking patterns. Viewpoint-dependent recognition performance was tested in two groups of twelve persons who knew each other very well. Participants' motion data were acquired by recording their walking patterns in three-dimensional space with the use of a motion capture system. Size-normalised point-light displays of biological motion of these walking patterns, including one's own, were presented to the same group members on a computer screen in frontal view, half-profile view, and profile view. Observers were requested to assign the person's name to the individual gait pattern. No feedback was given. Whereas recognition performance of one's own walking patterns was viewpoint independent, recognition rate for other familiar individuals was better for frontal and half-profile view than for profile view. These findings are discussed in the context of the theory of common coding of motor and visual body representations."
                },
                {
                    "title": "Biologically Plausible Neural Model for the Recognition of Biological Motion and Actions",
                    "abstract": "Abstract : The visual recognition of complex movements and actions is crucial for communications and survival in many species. Remarkable sensitivity and robustness of biological motion perception have been demonstrated in psychophysical experiments. In recent years, neurons and cortical areas involved in action recognition have been identified in neurophysiological and imaging studies. However, the detailed neural mechanisms that underlie the recognition of such complex movement patterns remain largely unknown. This paper reviews the experimental results and summarizes them in terms of a biologically plausible neural model. The model is based on the key assumption that action recognition is based on learned prototypical patterns and exploits information from the ventral and dorsal pathway. The model makes specific predictions that motivate new experiments."
                },
                {
                    "title": "Perception of biological motion without local image motion",
                    "abstract": "A vivid perception of the moving form of a human figure can be obtained from a few moving light points on the joints of the body. This is known as biological motion perception. It is commonly believed that the perception of biological motion rests on image motion signals. Curiously, however, some patients with lesions to motion processing areas of the dorsal stream are severely impaired in image motion perception but can easily perceive biological motion. Here we describe a biological motion stimulus based on a limited lifetime technique that tests the perception of a moving human figure in the absence of local image motion. We find that subjects can spontaneously recognize a moving human figure in displays without local image motion. Their performance is very similar to that for classic point-light displays. We also find that tasks involving the discrimination of walking direction or the coherence of a walking figure can be performed in the absence of image motion. Thus, although image motion may generally aid processes such as segmenting figure from background, we propose that it is not the basis for the percept of biological motion. Rather, we suggest biological motion is derived from dynamic form information on body posture evolving over time."
                },
                {
                    "title": "Perception of Emotion from Dynamic Point-Light Displays Represented in Dance",
                    "abstract": "It is well known that biological motion, as produced by point-light displays on a human body, gives a good representation of the represented body\u2014eg its gender and the nature of the task which it is engaged in. The question is whether it is possible to judge the emotional state of a human body from motion information alone. An ability to make this kind of judgment may imply that people are able to perceive emotion from patterns of movement without having to compute the detailed shape first. Subjects were shown brief video clips of two trained dancers (one male, one female). The dancers were aiming to convey the following emotions: fear, anger, grief, joy, surprise, and disgust. The video clips protrayed fully lit scenes and point-light scenes, with thirteen small points of light attached to the body of each dancer. Half the stimuli were presented the right way up, while half were inverted. The subjects' task was to judge which emotion was being portrayed. Full-body clips gave good recognition of emotionality (88% correct), but the results for upright biological-motion displays were also significantly above chance (63% correct). Inversion of the display reduced biological-motion (but not full-body) performance to close to chance but still significantly above chance. A space\u2013time analysis of the motion of the points of light was carried out, and was related to the discriminability of the different emotions. Biological-motion displays, which convey no information while static, are able to give a rich description of the subject matter, including the ability to judge emotional state. This ability is disrupted when the image is inverted."
                },
                {
                    "title": "Gender discrimination in biological motion displays based on dynamic cues",
                    "abstract": "Studies of human locomotion have found that male and female walkers differ in terms of lateral body sway, with males tending to swing their shoulders from side to side more than their hips, and females tending to swing their hips more than their shoulders. Experiments reported here demonstrate that naive viewers can identify the gender of the figure in a biological motion display very reliably when the display contains gender-specific lateral body sway. Sensitivity to gender is high even for displays containing only a fraction of a step cycle. This dynamic cue dominates structural cues based on torso shape (\u2018centre-of-moment\u2019) when the cues are set in opposition. It is mediated by gender-specific differences in the velocity of shoulder and hip dots, not by positional differences in shoulder and hip dots during the step cycle."
                },
                {
                    "title": "Neuroethology of releasing mechanisms: Prey-catching in toads",
                    "abstract": "Abstract \u201cSign stimuli\u201d elicit specific patterns of behavior when an organism's motivation is appropriate. In the toad, visually released prey-catching involves orienting toward the prey, approaching, fixating, and snapping. For these action patterns to be selected and released, the prey must be recognized and localized in space. Toads discriminate prey from nonprey by certain spatiotemporal stimulus features. The stimulus-response relations are mediated by innate releasing mechanisms (RMs) with recognition properties partly modifiable by experience. Striato-pretecto-tectal connectivity determines the RM's recognition and localization properties, whereas medialpallio-thalamo-tectal circuitry makes the system sensitive to changes in internal state and to prior history of exposure to stimuli. RMs encode the diverse stimulus conditions referring to the same prey object through different combinations of \u201cspecialized\u201d tectal neurons, involving cells selectively tuned to prey features. The prey-selective neurons express the outcome of information processing in functional units consisting of interconnected cells. Excitatory and inhibitory interactions among feature-sensitive tectal and pretectal neurons specify the perceptual operations involved in distinguishing the prey from its background, selecting its features, and discriminating it from predators. Other connections indicate stimulus location. The results of these analyses are transmitted by specialized neurons projecting from the tectum to bulbar/spinal motor systems, providing a sensorimotor interface. Specific combinations of such projective neurons \u2013 mediating feature- and space-related messages \u2013 form \u201ccommand releasing systems\u201d that activate corresponding motor pattern generators for appropriate prey-catching action patterns."
                },
                {
                    "title": "Angular features-based human action recognition system for a real application with subtle unit actions",
                    "abstract": "Human action recognition (HAR) technology is receiving considerable attention in the field of human-computer interaction. We present a HAR system that works stably in real-world applications. In real-world applications, the HAR system needs to identify detailed actions for specific purposes, and the action data includes many variations. Accordingly, we conducted three experiments. First, we tested our recognition system\u2019s performance on the UTD-MHAD dataset. We compared our system\u2019s accuracy with results from previous research and confirmed that our system achieves an average performance among recognition systems. Furthermore, we hypothesized the use of a HAR system to detect burglary. In the second experiment, we compared the existing benchmark data with our crime detection dataset. We recognized the test scenarios\u2019 data by using the recognition system trained by each dataset. The recognition system trained by our dataset achieved higher accuracy than the past benchmark dataset. The results show that the training data should contain detailed actions for a real application. In the third experiment, we tried to find the motion data type that stably recognizes action regardless of data variation. In a real application, the action data are changed by people. Thus, we introduced variations in the action data using the cross-subject protocol and moving area setting. We trained the recognition system using each position and angle data. In addition, we compared the accuracy of each system. We found that the angle format results in better accuracy because the angle data are beneficial for converting the action variation into a consistent pattern."
                },
                {
                    "title": "Simulating Biological Motion Perception Using a Recurrent Neural Network",
                    "abstract": "People have the ability to perceive biological motion under conditions of severely limited visual information. If the information is in the form of a point-light motion sequence of a human walker or jogger, perceptual discontinuities are barely noticeable. This phenomenon can be simulated for a machine perceptual system by using a recurrent artificial neural network. A feedback connection from the output of the hidden layer to the input of the hidden layer provides the network with information about past events that can be used to classify an entire sequence of events. In this case, the discrete events are the x,y-coordinates of the point-light displays during a specific motion sequence. Generalizations about temporal as well as spatial patterns can be made that enable the network to classify an input sequence as being either biological or non-biological in"
                },
                {
                    "title": "The Expression of the Emotions in Man and Animals",
                    "abstract": "Book file PDF easily for everyone and every device. You can download and read online The Expression of the Emotions in Man and Animals file PDF Book only if you are registered here. And also you can download or read online all Book PDF file that related with The Expression of the Emotions in Man and Animals book. Happy reading The Expression of the Emotions in Man and Animals Bookeveryone. Download file Free Book PDF The Expression of the Emotions in Man and Animals at Complete PDF Library. This Book have some digital formats such us :paperbook, ebook, kindle, epub, fb2 and another formats. Here is The Complete PDF Book Library. It's free to register here to get Book file PDF The Expression of the Emotions in Man and Animals."
                },
                {
                    "title": "The Reproductive Behaviour of the Zebra Finch (Poephila Guttata), With Special Reference To Pseudofemale Behaviour and Displacement Activities",
                    "abstract": "The Zebra Finch (Poephila guttata) is a small Australian Ploceid which is ideal for the laboratory study of bird behaviour. Its attacking, fleeing, threatening and submissive behaviour are described and discussed. In this species there is no elaborate code of signals involved in this behaviour. Most disputes are settled mainly by actual fighting. This species builds a domed nest with a side entrance, the material for which is collected by the male. A special ceremony exists which apparently serves to strengthen the relationship between a pair in relation to a particular nest or nest-site. The pre-copulatory behaviour is described and analysed. The courtship of the male consists of an upright pivoting dance, during which it always sings and advances towards the female. The courtship of the latter takes the form of a horizontal body posture accompanied by a rapid quivering of the tail. The courtship of both sexes involves the three conflicting drives of attacking, fleeing and mating. In different species, the relative importance of these three drives will vary. In the Zebra Finch, the main conflict in pre-copulatory behaviour is between fleeing and mating. The female is more active than the male in incubating and rearing the young. The nestling possess brightly marked mouths which presumably facilitate feeding by the parents in the dark nest. Under certain circumstances, males were seen to perform the female courtship display and occasionally females responded with male behaviour, resulting in the occurrence of reversed mountings. The possible causation of this behaviour is discussed. It appears that the pseudofemale behaviour of the male is caused by strong sexual thwarting (as a result of the unresponsiveness of the female) in combination with the arousal of the drive to flee. Displacement activities which occur in this species are enumerated and described. They include various comfort movements, sleep, feeding, mounting, foodbegging and nest-building. Some of these occur as True, and other as Apparent, Alternative Displacement Activities. In the case of displacement beak-wiping it is possible to analyse accurately the differences between the displacement activity and its autochthonous form. There are differences in orientation, amplitude, velocity, and in the number of wipes per wiping-bout. This last difference is analysed quantitatively. Various other aspects of the reproductive behaviour are briefly mentioned, particularly the nature of territory and the strength of the pair-bond. The family Ploceidae show many degrees of socialisation and the defence of the territory by the Zebra Finch is often spasmodic, except when actually rearing young. The occurrence of attempted \"rape\", when a new female is introduced, is discussed and the durability of the pair-bond after separation is investigated. A Redirection Activity is recorded, namely redirected mounting. A possible reason for the complexity of the markings of the Zebra Finch is discussed. Many of the component markings of the species are shared with various other Australian Grassfinches, and the demands of specificity can only be met by a unique combination of a number of these markings."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can AI models be trained to effectively generalize their understanding of biological motion perception (BMP) from natural RGB videos to point-light displays without prior training?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning, particularly in understanding how AI can mimic human-like perception abilities. By addressing the generalization challenges in BMP tasks, this research could lead to improved AI models that better recognize and interpret human actions in various contexts, enhancing applications in robotics, surveillance, and human-computer interaction. Furthermore, it could bridge the gap between computational models and human cognitive processes, fostering interdisciplinary collaboration between AI, psychology, and neuroscience.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of biological motion perception, which involves recognizing subtle and dynamic motion patterns without prior training. Naive approaches may fail because they often rely on specific training conditions that do not account for the variability present in BMP tasks. Key obstacles include the need for robust motion representations that can adapt to different visual stimuli, the difficulty in capturing the nuances of human movement, and the limitations of current AI architectures in generalizing across diverse conditions, such as occlusion and viewpoint variability.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either specific aspects of BMP or on action recognition in more structured environments, leading to a lack of comprehensive studies that connect these two areas. Existing solutions have not adequately addressed the unique challenges posed by BMP tasks, such as the need for zero-shot learning capabilities. Additionally, many prior models have been trained under conditions that do not reflect the variability encountered in real-world scenarios. Our approach aims to fill this gap by systematically evaluating AI models' generalization abilities across different BMP conditions, thereby providing a more holistic understanding of motion perception.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves training AI models on a diverse dataset of natural RGB videos and subsequently testing their performance on BMP stimuli, specifically using joint videos and sequential position action videos. We will evaluate the models' generalization capabilities by varying temporal and visual properties of the stimuli. The key metrics for assessment will include accuracy in recognizing actions and the ability to generalize across different BMP conditions. We expect to demonstrate that AI models can achieve improved generalization performance, potentially leading to new insights into"
            }
        },
        "author_data": {
            "48c4c891-3909-4dfa-b65e-9306815c5ade": {
                "pk": "48c4c891-3909-4dfa-b65e-9306815c5ade",
                "name": "Shuangpeng Han",
                "collaborators": [
                    "Mengmi Zhang",
                    "Ziyu Wang",
                    "Mike Zheng Shou",
                    "Rizhao Cai",
                    "Yawen Cui",
                    "Zitong Yu",
                    "Yongjian Hu",
                    "Alex Kot"
                ],
                "domain": [
                    "Error Prediction",
                    "Pose Estimation",
                    "Face Anti-Spoofing",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Unveiling AI's Blind Spots: An Oracle for In-Domain, Out-of-Domain, and Adversarial Errors",
                        "abstract": "AI models make mistakes when recognizing images-whether in-domain, out-of-domain, or adversarial. Predicting these errors is critical for improving system reliability, reducing costly mistakes, and enabling proactive corrections in real-world applications such as healthcare, finance, and autonomous systems. However, understanding what mistakes AI models make, why they occur, and how to predict them remains an open challenge. Here, we conduct comprehensive empirical evaluations using a \"mentor\" model-a deep neural network designed to predict another model's errors. Our findings show that the mentor model excels at learning from a mentee's mistakes on adversarial images with small perturbations and generalizes effectively to predict in-domain and out-of-domain errors of the mentee. Additionally, transformer-based mentor models excel at predicting errors across various mentee architectures. Subsequently, we draw insights from these observations and develop an \"oracle\" mentor model, dubbed SuperMentor, that achieves 78% accuracy in predicting errors across different error types. Our error prediction framework paves the way for future research on anticipating and correcting AI model behaviours, ultimately increasing trust in AI systems. All code, models, and data will be made publicly available."
                    },
                    {
                        "title": "Unsupervised Prior Learning: Discovering Categorical Pose Priors from Videos",
                        "abstract": "A prior represents a set of beliefs or assumptions about a system, aiding inference and decision-making. In this work, we introduce the challenge of unsupervised prior learning in pose estimation, where AI models learn pose priors of animate objects from videos in a self-supervised manner. These videos present objects performing various actions, providing crucial information about their keypoints and connectivity. While priors are effective in pose estimation, acquiring them can be difficult. We propose a novel method, named Pose Prior Learner (PPL), to learn general pose priors applicable to any object category. PPL uses a hierarchical memory to store compositional parts of prototypical poses, from which we distill a general pose prior. This prior enhances pose estimation accuracy through template transformation and image reconstruction. PPL learns meaningful pose priors without any additional human annotations or interventions, outperforming competitive baselines on both human and animal pose estimation datasets. Notably, our experimental results reveal the effectiveness of PPL using learnt priors for pose estimation on occluded images. Through iterative inference, PPL leverages priors to refine estimated poses, regressing them to any prototypical poses stored in memory. Our code, model, and data will be publicly available."
                    },
                    {
                        "title": "Hyperbolic Face Anti-Spoofing",
                        "abstract": "Learning generalized face anti-spoofing (FAS) models against presentation attacks is essential for the security of face recognition systems. Previous FAS methods usually encourage models to extract discriminative features, of which the distances within the same class (bonafide or attack) are pushed close while those between bonafide and attack are pulled away. However, these methods are designed based on Euclidean distance, which lacks generalization ability for unseen attack detection due to poor hierarchy embedding ability. According to the evidence that different spoofing attacks are intrinsically hierarchical, we propose to learn richer hierarchical and discriminative spoofing cues in hyperbolic space. Specifically, for unimodal FAS learning, the feature embeddings are projected into the Poincar\\'e ball, and then the hyperbolic binary logistic regression layer is cascaded for classification. To further improve generalization, we conduct hyperbolic contrastive learning for the bonafide only while relaxing the constraints on diverse spoofing attacks. To alleviate the vanishing gradient problem in hyperbolic space, a new feature clipping method is proposed to enhance the training stability of hyperbolic models. Besides, we further design a multimodal FAS framework with Euclidean multimodal feature decomposition and hyperbolic multimodal feature fusion & classification. Extensive experiments on three benchmark datasets (i.e., WMCA, PADISI-Face, and SiW-M) with diverse attack types demonstrate that the proposed method can bring significant improvement compared to the Euclidean baselines on unseen attack detection. In addition, the proposed framework is also generalized well on four benchmark datasets (i.e., MSU-MFSD, IDIAP REPLAY-ATTACK, CASIA-FASD, and OULU-NPU) with a limited number of attack types."
                    }
                ]
            },
            "9d3d5a20-b6b1-43ee-94d3-5d37c145216d": {
                "pk": "9d3d5a20-b6b1-43ee-94d3-5d37c145216d",
                "name": "Ziyu Wang",
                "collaborators": [
                    "Gus Xia",
                    "Jun Zhu",
                    "Chris Holmes",
                    "Yuhao Zhou",
                    "Nando de Freitas",
                    "Tongzheng Ren",
                    "Jie Yang",
                    "Mohamad Sawan",
                    "Dingsu Wang",
                    "Yixiao Zhang"
                ],
                "domain": [
                    "Music Generation",
                    "Bayesian Inference",
                    "Machine Learning",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "A Framework for Automated Pop-song Melody Generation with Piano Accompaniment Arrangement",
                        "abstract": "We contribute a pop-song automation framework for lead melody generation and accompaniment arrangement. The framework reflects the major procedures of human music composition, generating both lead melody and piano accompaniment by a unified strategy. Specifically, we take chord progression as an input and propose three models to generate a structured melody with piano accompaniment textures. First, the harmony alternation model transforms a raw input chord progression to an altered one to better fit the specified music style. Second, the melody generation model generates the lead melody and other voices (melody lines) of the accompaniment using seasonal ARMA (Autoregressive Moving Average) processes. Third, the melody integration model integrates melody lines (voices) together as the final piano accompaniment. We evaluate the proposed framework using subjective listening tests. Experimental results show that the generated melodies are rated significantly higher than the ones generated by bi-directional LSTM, and our accompaniment arrangement result is comparable with a state-of-the-art commercial software, Band in a Box."
                    },
                    {
                        "title": "On Uncertainty Quantification for Near-Bayes Optimal Algorithms",
                        "abstract": "Bayesian modelling allows for the quantification of predictive uncertainty which is crucial in safety-critical applications. Yet for many machine learning (ML) algorithms, it is difficult to construct or implement their Bayesian counterpart. In this work we present a promising approach to address this challenge, based on the hypothesis that commonly used ML algorithms are efficient across a wide variety of tasks and may thus be near Bayes-optimal w.r.t. an unknown task distribution. We prove that it is possible to recover the Bayesian posterior defined by the task distribution, which is unknown but optimal in this setting, by building a martingale posterior using the algorithm. We further propose a practical uncertainty quantification method that apply to general ML algorithms. Experiments based on a variety of non-NN and NN algorithms demonstrate the efficacy of our method."
                    },
                    {
                        "title": "On Subjective Uncertainty Quantification and Calibration in Natural Language Generation",
                        "abstract": "Applications of large language models often involve the generation of free-form responses, in which case uncertainty quantification becomes challenging. This is due to the need to identify task-specific uncertainties (e.g., about the semantics) which appears difficult to define in general cases. This work addresses these challenges from a perspective of Bayesian decision theory, starting from the assumption that our utility is characterized by a similarity measure that compares a generated response with a hypothetical true response. We discuss how this assumption enables principled quantification of the model's subjective uncertainty and its calibration. We further derive a measure for epistemic uncertainty, based on a missing data perspective and its characterization as an excess risk. The proposed methods can be applied to black-box language models. We illustrate the methods on question answering and machine translation tasks. Our experiments provide a principled evaluation of task-specific calibration, and demonstrate that epistemic uncertainty offers a promising deferral strategy for efficient data acquisition in in-context learning."
                    },
                    {
                        "title": "A Novel Multi-scale Dilated 3D CNN for Epileptic Seizure Prediction",
                        "abstract": "Accurate prediction of epileptic seizures allows patients to take preventive measures in advance to avoid possible injuries. In this work, a novel convolutional neural network (CNN) is proposed to analyze time, frequency, and channel information of electroencephalography (EEG) signals. The model uses three-dimensional (3D) kernels to facilitate the feature extraction over the three dimensions. The application of multiscale dilated convolution enables the 3D kernel to have more flexible receptive fields. The proposed CNN model is evaluated with the CHB-MIT EEG database, the experimental results indicate that our model outperforms the existing state-of-the-art, achieves 80.5% accuracy, 85.8% sensitivity and 75.1% specificity."
                    },
                    {
                        "title": "Fast Instrument Learning with Faster Rates",
                        "abstract": "We investigate nonlinear instrumental variable (IV) regression given high-dimensional instruments. We propose a simple algorithm which combines kernelized IV methods and an arbitrary, adaptive regression algorithm, accessed as a black box. Our algorithm enjoys faster-rate convergence and adapts to the dimensionality of informative latent features, while avoiding an expensive minimax optimization procedure, which has been necessary to establish similar guarantees. It further brings the benefit of flexible machine learning models to quasi-Bayesian uncertainty quantification, likelihood-based model selection, and model averaging. Simulation studies demonstrate the competitive performance of our method."
                    },
                    {
                        "title": "Learning Interpretable Representation for Controllable Polyphonic Music Generation",
                        "abstract": "While deep generative models have become the leading methods for algorithmic composition, it remains a challenging problem to control the generation process because the latent variables of most deep-learning models lack good interpretability. Inspired by the content-style disentanglement idea, we design a novel architecture, under the VAE framework, that effectively learns two interpretable latent factors of polyphonic music: chord and texture. The current model focuses on learning 8-beat long piano composition segments. We show that such chord-texture disentanglement provides a controllable generation pathway leading to a wide spectrum of applications, including compositional style transfer, texture variation, and accompaniment arrangement. Both objective and subjective evaluations show that our method achieves a successful disentanglement and high quality controlled music generation."
                    },
                    {
                        "title": "Self-Avoiding Random Dynamics on Integer Complex Systems",
                        "abstract": "This paper introduces a new specialized algorithm for equilibrium Monte Carlo sampling of binary-valued systems, which allows for large moves in the state space. This is achieved by constructing self-avoiding walks (SAWs) in the state space. As a consequence, many bits are flipped in a single MCMC step. We name the algorithm SARDONICS, an acronym for Self-Avoiding Random Dynamics on Integer Complex Systems. The algorithm has several free parameters, but we show that Bayesian optimization can be used to automatically tune them. SARDONICS performs remarkably well in a broad number of sampling tasks: toroidal ferromagnetic and frustrated Ising models, 3D Ising models, restricted Boltzmann machines and chimera graphs arising in the design of quantum computers."
                    },
                    {
                        "title": "Theoretical Analysis of Bayesian Optimisation with Unknown Gaussian Process Hyper-Parameters",
                        "abstract": "Bayesian optimisation has gained great popularity as a tool for optimising the parameters of machine learning algorithms and models. Somewhat ironically, setting up the hyper-parameters of Bayesian optimisation methods is notoriously hard. While reasonable practical solutions have been advanced, they can often fail to find the best optima. Surprisingly, there is little theoretical analysis of this crucial problem in the literature. To address this, we derive a cumulative regret bound for Bayesian optimisation with Gaussian processes and unknown kernel hyper-parameters in the stochastic setting. The bound, which applies to the expected improvement acquisition function and sub-Gaussian observation noise, provides us with guidelines on how to design hyper-parameter estimation methods. A simple simulation demonstrates the importance of following these guidelines."
                    },
                    {
                        "title": "Function Space Particle Optimization for Bayesian Neural Networks",
                        "abstract": "While Bayesian neural networks (BNNs) have drawn increasing attention, their posterior inference remains challenging, due to the high-dimensional and over-parameterized nature. To address this issue, several highly flexible and scalable variational inference procedures based on the idea of particle optimization have been proposed. These methods directly optimize a set of particles to approximate the target posterior. However, their application to BNNs often yields sub-optimal performance, as such methods have a particular failure mode on over-parameterized models. In this paper, we propose to solve this issue by performing particle optimization directly in the space of regression functions. We demonstrate through extensive experiments that our method successfully overcomes this issue, and outperforms strong baselines in a variety of tasks including prediction, defense against adversarial examples, and reinforcement learning."
                    },
                    {
                        "title": "Quasi-Bayesian Dual Instrumental Variable Regression",
                        "abstract": "Recent years have witnessed an upsurge of interest in employing flexible machine learning models for instrumental variable (IV) regression, but the development of uncertainty quantification methodology is still lacking. In this work we present a novel quasi-Bayesian procedure for IV regression, building upon the recently developed kernelized IV models and the dual/minimax formulation of IV regression. We analyze the frequentist behavior of the proposed method, by establishing minimax optimal contraction rates in $L_2$ and Sobolev norms, and discussing the frequentist validity of credible balls. We further derive a scalable inference algorithm which can be extended to work with wide neural network models. Empirical evaluation shows that our method produces informative uncertainty estimates on complex high-dimensional problems."
                    },
                    {
                        "title": "Audio-to-symbolic Arrangement via Cross-modal Music Representation Learning",
                        "abstract": "Could we automatically derive the score of a piano accompaniment based on the audio of a pop song? This is the audio-to-symbolic arrangement problem we tackle in this paper. A good arrangement model should not only consider the audio content but also have prior knowledge of piano composition (so that the generation \"sounds like\" the audio and meanwhile maintains musicality). To this end, we contribute a cross-modal representation-learning model, which 1) extracts chord and melodic information from the audio, and 2) learns texture representation from both audio and a corrupted ground truth arrangement. We further introduce a tailored training strategy that gradually shifts the source of texture information from corrupted score to audio. In the end, the score-based texture posterior is reduced to a standard normal distribution, and only audio is needed for inference. Experiments show that our model captures major audio information and outperforms baselines in generation quality."
                    },
                    {
                        "title": "Bayesian Optimisation for Machine Translation",
                        "abstract": "This paper presents novel Bayesian optimisation algorithms for minimum error rate training of statistical machine translation systems. We explore two classes of algorithms for efficiently exploring the translation space, with the first based on N-best lists and the second based on a hypergraph representation that compactly represents an exponential number of translation options. Our algorithms exhibit faster convergence and are capable of obtaining lower error rates than the existing translation model specific approaches, all within a generic Bayesian optimisation framework. Further more, we also introduce a random embedding algorithm to scale our approach to sparse high dimensional feature sets."
                    },
                    {
                        "title": "Object-centric Learning with Cyclic Walks between Parts and Whole",
                        "abstract": "Learning object-centric representations from complex natural environments enables both humans and machines with reasoning abilities from low-level perceptual features. To capture compositional entities of the scene, we proposed cyclic walks between perceptual features extracted from vision transformers and object entities. First, a slot-attention module interfaces with these perceptual features and produces a finite set of slot representations. These slots can bind to any object entities in the scene via inter-slot competitions for attention. Next, we establish entity-feature correspondence with cyclic walks along high transition probability based on the pairwise similarity between perceptual features (aka \"parts\") and slot-binded object representations (aka \"whole\"). The whole is greater than its parts and the parts constitute the whole. The part-whole interactions form cycle consistencies, as supervisory signals, to train the slot-attention module. Our rigorous experiments on \\textit{seven} image datasets in \\textit{three} \\textit{unsupervised} tasks demonstrate that the networks trained with our cyclic walks can disentangle foregrounds and backgrounds, discover objects, and segment semantic objects in complex scenes. In contrast to object-centric models attached with a decoder for the pixel-level or feature-level reconstructions, our cyclic walks provide strong learning signals, avoiding computation overheads and enhancing memory efficiency. Our source code and data are available at: \\href{https://github.com/ZhangLab-DeepNeuroCogLab/Parts-Whole-Object-Centric-Learning/}{link}."
                    },
                    {
                        "title": "PIM-GPT: A Hybrid Process-in-Memory Accelerator for Autoregressive Transformers",
                        "abstract": "Decoder-only Transformer models such as GPT have demonstrated exceptional performance in text generation, by autoregressively predicting the next token. However, the efficacy of running GPT on current hardware systems is bounded by low compute-to-memory-ratio and high memory access. Process-in-memory (PIM) architectures can minimize off-chip data movement and utilize high internal bandwidth. They stand out as promising candidates for accelerating memory-bounded tasks such as GPT inference.   In this work, we propose a PIM accelerator, PIM-GPT, which achieves end-to-end acceleration of GPT inference with high performance and high energy efficiency. PIM-GPT leverages DRAM-based PIM designs for executing multiply-accumulate (MAC) operations directly in the DRAM chips, eliminating the need to move matrix data off-chip. Non-linear functions and data communication is supported by an application specific integrated chip (ASIC). At the software level, mapping schemes are designed to maximize data locality and computation parallelism by concatenating and partitioning matrices among DRAM channels and banks to utilize all available in-memory computation units. The efficiency of the PIM-GPT architecture is verified through circuit synthesis and an event-driven clock-cycle accurate simulator. Overall, PIM-GPT achieves 41$-$137$\\times$, 631$-$1074$\\times$ speedup and 123$-$383$\\times$, 320$-$602$\\times$ energy efficiency over GPU and CPU baseline on 8 GPT models with up to 1.4 billion parameters."
                    },
                    {
                        "title": "Whole-Song Hierarchical Generation of Symbolic Music Using Cascaded Diffusion Models",
                        "abstract": "Recent deep music generation studies have put much emphasis on long-term generation with structures. However, we are yet to see high-quality, well-structured whole-song generation. In this paper, we make the first attempt to model a full music piece under the realization of compositional hierarchy. With a focus on symbolic representations of pop songs, we define a hierarchical language, in which each level of hierarchy focuses on the semantics and context dependency at a certain music scope. The high-level languages reveal whole-song form, phrase, and cadence, whereas the low-level languages focus on notes, chords, and their local patterns. A cascaded diffusion model is trained to model the hierarchical language, where each level is conditioned on its upper levels. Experiments and analysis show that our model is capable of generating full-piece music with recognizable global verse-chorus structure and cadences, and the music quality is higher than the baselines. Additionally, we show that the proposed model is controllable in a flexible way. By sampling from the interpretable hierarchical languages or adjusting pre-trained external representations, users can control the music flow via various features such as phrase harmonic structures, rhythmic patterns, and accompaniment texture."
                    }
                ]
            },
            "ed2cd7e8-8c04-4c87-9347-64254cd623f0": {
                "pk": "ed2cd7e8-8c04-4c87-9347-64254cd623f0",
                "name": "Mengmi Zhang",
                "collaborators": [
                    "Gabriel Kreiman",
                    "Claire Tseng",
                    "Shuangpeng Han",
                    "Ankur Sikarwar",
                    "Zenglin Shi",
                    "Xiao Liu",
                    "Ziyu Wang",
                    "Mike Zheng Shou",
                    "Spandan Madan",
                    "Ying Sun"
                ],
                "domain": [
                    "Computer Vision",
                    "Machine Learning",
                    "Cognitive Science",
                    "Continual Learning"
                ],
                "publications": [
                    {
                        "title": "What am I Searching for: Zero-shot Target Identity Inference in Visual Search",
                        "abstract": "Can we infer intentions from a person's actions? As an example problem, here we consider how to decipher what a person is searching for by decoding their eye movement behavior. We conducted two psychophysics experiments where we monitored eye movements while subjects searched for a target object. We defined the fixations falling on non-target objects as \"error fixations\". Using those error fixations, we developed a model (InferNet) to infer what the target was. InferNet uses a pre-trained convolutional neural network to extract features from the error fixations and computes a similarity map between the error fixations and all locations across the search image. The model consolidates the similarity maps across layers and integrates these maps across all error fixations. InferNet successfully identifies the subject's goal and outperforms competitive null models, even without any object-specific training on the inference task."
                    },
                    {
                        "title": "What am I Searching for: Zero-shot Target Identity Inference in Visual Search",
                        "abstract": "Can we infer intentions from a person's actions? As an example problem, here we consider how to decipher what a person is searching for by decoding their eye movement behavior. We conducted two psychophysics experiments where we monitored eye movements while subjects searched for a target object. We defined the fixations falling on \\textit{non-target} objects as \"error fixations\". Using those error fixations, we developed a model (InferNet) to infer what the target was. InferNet uses a pre-trained convolutional neural network to extract features from the error fixations and computes a similarity map between the error fixations and all locations across the search image. The model consolidates the similarity maps across layers and integrates these maps across all error fixations. InferNet successfully identifies the subject's goal and outperforms competitive null models, even without any object-specific training on the inference task."
                    },
                    {
                        "title": "Unveiling AI's Blind Spots: An Oracle for In-Domain, Out-of-Domain, and Adversarial Errors",
                        "abstract": "AI models make mistakes when recognizing images-whether in-domain, out-of-domain, or adversarial. Predicting these errors is critical for improving system reliability, reducing costly mistakes, and enabling proactive corrections in real-world applications such as healthcare, finance, and autonomous systems. However, understanding what mistakes AI models make, why they occur, and how to predict them remains an open challenge. Here, we conduct comprehensive empirical evaluations using a \"mentor\" model-a deep neural network designed to predict another model's errors. Our findings show that the mentor model excels at learning from a mentee's mistakes on adversarial images with small perturbations and generalizes effectively to predict in-domain and out-of-domain errors of the mentee. Additionally, transformer-based mentor models excel at predicting errors across various mentee architectures. Subsequently, we draw insights from these observations and develop an \"oracle\" mentor model, dubbed SuperMentor, that achieves 78% accuracy in predicting errors across different error types. Our error prediction framework paves the way for future research on anticipating and correcting AI model behaviours, ultimately increasing trust in AI systems. All code, models, and data will be made publicly available."
                    },
                    {
                        "title": "Decoding the Enigma: Benchmarking Humans and AIs on the Many Facets of Working Memory",
                        "abstract": "Working memory (WM), a fundamental cognitive process facilitating the temporary storage, integration, manipulation, and retrieval of information, plays a vital role in reasoning and decision-making tasks. Robust benchmark datasets that capture the multifaceted nature of WM are crucial for the effective development and evaluation of AI WM models. Here, we introduce a comprehensive Working Memory (WorM) benchmark dataset for this purpose. WorM comprises 10 tasks and a total of 1 million trials, assessing 4 functionalities, 3 domains, and 11 behavioral and neural characteristics of WM. We jointly trained and tested state-of-the-art recurrent neural networks and transformers on all these tasks. We also include human behavioral benchmarks as an upper bound for comparison. Our results suggest that AI models replicate some characteristics of WM in the brain, most notably primacy and recency effects, and neural clusters and correlates specialized for different domains and functionalities of WM. In the experiments, we also reveal some limitations in existing models to approximate human behavior. This dataset serves as a valuable resource for communities in cognitive psychology, neuroscience, and AI, offering a standardized framework to compare and enhance WM models, investigate WM's neural underpinnings, and develop WM models with human-like capabilities. Our source code and data are available at https://github.com/ZhangLab-DeepNeuroCogLab/WorM."
                    },
                    {
                        "title": "Putting visual object recognition in context",
                        "abstract": "Context plays an important role in visual recognition. Recent studies have shown that visual recognition networks can be fooled by placing objects in inconsistent contexts (e.g., a cow in the ocean). To model the role of contextual information in visual recognition, we systematically investigated ten critical properties of where, when, and how context modulates recognition, including the amount of context, context and object resolution, geometrical structure of context, context congruence, and temporal dynamics of contextual modulation. The tasks involved recognizing a target object surrounded with context in a natural image. As an essential benchmark, we conducted a series of psychophysics experiments where we altered one aspect of context at a time, and quantified recognition accuracy. We propose a biologically-inspired context-aware object recognition model consisting of a two-stream architecture. The model processes visual information at the fovea and periphery in parallel, dynamically incorporates object and contextual information, and sequentially reasons about the class label for the target object. Across a wide range of behavioral tasks, the model approximates human level performance without retraining for each task, captures the dependence of context enhancement on image properties, and provides initial steps towards integrating scene and object information for visual recognition. All source code and data are publicly available: https://github.com/kreimanlab/Put-In-Context."
                    },
                    {
                        "title": "Training-free Object Counting with Prompts",
                        "abstract": "This paper tackles the problem of object counting in images. Existing approaches rely on extensive training data with point annotations for each object, making data collection labor-intensive and time-consuming. To overcome this, we propose a training-free object counter that treats the counting task as a segmentation problem. Our approach leverages the Segment Anything Model (SAM), known for its high-quality masks and zero-shot segmentation capability. However, the vanilla mask generation method of SAM lacks class-specific information in the masks, resulting in inferior counting accuracy. To overcome this limitation, we introduce a prior-guided mask generation method that incorporates three types of priors into the segmentation process, enhancing efficiency and accuracy. Additionally, we tackle the issue of counting objects specified through text by proposing a two-stage approach that combines reference object selection and prior-guided mask generation. Extensive experiments on standard datasets demonstrate the competitive performance of our training-free counter compared to learning-based approaches. This paper presents a promising solution for counting objects in various scenarios without the need for extensive data collection and counting-specific training. Code is available at \\url{https://github.com/shizenglin/training-free-object-counter}"
                    },
                    {
                        "title": "Adaptive Visual Scene Understanding: Incremental Scene Graph Generation",
                        "abstract": "Scene graph generation (SGG) involves analyzing images to extract meaningful information about objects and their relationships. Given the dynamic nature of the visual world, it becomes crucial for AI systems to detect new objects and establish their new relationships with existing objects. To address the lack of continual learning methodologies in SGG, we introduce the comprehensive Continual ScenE Graph Generation (CSEGG) dataset along with 3 learning scenarios and 8 evaluation metrics. Our research investigates the continual learning performances of existing SGG methods on the retention of previous object entities and relationships as they learn new ones. Moreover, we also explore how continual object detection enhances generalization in classifying known relationships on unknown objects. We conduct extensive experiments benchmarking and analyzing the classical two-stage SGG methods and the most recent transformer-based SGG methods in continual learning settings, and gain valuable insights into the CSEGG problem. We invite the research community to explore this emerging field of study."
                    },
                    {
                        "title": "Object-centric Learning with Cyclic Walks between Parts and Whole",
                        "abstract": "Learning object-centric representations from complex natural environments enables both humans and machines with reasoning abilities from low-level perceptual features. To capture compositional entities of the scene, we proposed cyclic walks between perceptual features extracted from vision transformers and object entities. First, a slot-attention module interfaces with these perceptual features and produces a finite set of slot representations. These slots can bind to any object entities in the scene via inter-slot competitions for attention. Next, we establish entity-feature correspondence with cyclic walks along high transition probability based on the pairwise similarity between perceptual features (aka \"parts\") and slot-binded object representations (aka \"whole\"). The whole is greater than its parts and the parts constitute the whole. The part-whole interactions form cycle consistencies, as supervisory signals, to train the slot-attention module. Our rigorous experiments on \\textit{seven} image datasets in \\textit{three} \\textit{unsupervised} tasks demonstrate that the networks trained with our cyclic walks can disentangle foregrounds and backgrounds, discover objects, and segment semantic objects in complex scenes. In contrast to object-centric models attached with a decoder for the pixel-level or feature-level reconstructions, our cyclic walks provide strong learning signals, avoiding computation overheads and enhancing memory efficiency. Our source code and data are available at: \\href{https://github.com/ZhangLab-DeepNeuroCogLab/Parts-Whole-Object-Centric-Learning/}{link}."
                    },
                    {
                        "title": "Integrating Curricula with Replays: Its Effects on Continual Learning",
                        "abstract": "Humans engage in learning and reviewing processes with curricula when acquiring new skills or knowledge. This human learning behavior has inspired the integration of curricula with replay methods in continual learning agents. The goal is to emulate the human learning process, thereby improving knowledge retention and facilitating learning transfer. Existing replay methods in continual learning agents involve the random selection and ordering of data from previous tasks, which has shown to be effective. However, limited research has explored the integration of different curricula with replay methods to enhance continual learning. Our study takes initial steps in examining the impact of integrating curricula with replay methods on continual learning in three specific aspects: the interleaved frequency of replayed exemplars with training data, the sequence in which exemplars are replayed, and the strategy for selecting exemplars into the replay buffer. These aspects of curricula design align with cognitive psychology principles and leverage the benefits of interleaved practice during replays, easy-to-hard rehearsal, and exemplar selection strategy involving exemplars from a uniform distribution of difficulties. Based on our results, these three curricula effectively mitigated catastrophic forgetting and enhanced positive knowledge transfer, demonstrating the potential of curricula in advancing continual learning methodologies. Our code and data are available: https://github.com/ZhangLab-DeepNeuroCogLab/Integrating-Curricula-with-Replays"
                    },
                    {
                        "title": "TTA-Nav: Test-time Adaptive Reconstruction for Point-Goal Navigation under Visual Corruptions",
                        "abstract": "Robot navigation under visual corruption presents a formidable challenge. To address this, we propose a Test-time Adaptation (TTA) method, named as TTA-Nav, for point-goal navigation under visual corruptions. Our \"plug-and-play\" method incorporates a top-down decoder to a pre-trained navigation model. Firstly, the pre-trained navigation model gets a corrupted image and extracts features. Secondly, the top-down decoder produces the reconstruction given the high-level features extracted by the pre-trained model. Then, it feeds the reconstruction of a corrupted image back to the pre-trained model. Finally, the pre-trained model does forward pass again to output action. Despite being trained solely on clean images, the top-down decoder can reconstruct cleaner images from corrupted ones without the need for gradient-based adaptation. The pre-trained navigation model with our top-down decoder significantly enhances navigation performance across almost all visual corruptions in our benchmarks. Our method improves the success rate of point-goal navigation from the state-of-the-art result of 46% to 94% on the most severe corruption. This suggests its potential for broader application in robotic visual navigation. Project page: https://sites.google.com/view/tta-nav"
                    },
                    {
                        "title": "Lift-the-flap: what, where and when for context reasoning",
                        "abstract": "Context reasoning is critical in a wide variety of applications where current inputs need to be interpreted in the light of previous experience and knowledge. Both spatial and temporal contextual information play a critical role in the domain of visual recognition. Here we investigate spatial constraints (what image features provide contextual information and where they are located), and temporal constraints (when different contextual cues matter) for visual recognition. The task is to reason about the scene context and infer what a target object hidden behind a flap is in a natural image. To tackle this problem, we first describe an online human psychophysics experiment recording active sampling via mouse clicks in lift-the-flap games and identify clicking patterns and features which are diagnostic for high contextual reasoning accuracy. As a proof of the usefulness of these clicking patterns and visual features, we extend a state-of-the-art recurrent model capable of attending to salient context regions, dynamically integrating useful information, making inferences, and predicting class label for the target object over multiple clicks. The proposed model achieves human-level contextual reasoning accuracy, shares human-like sampling behavior and learns interpretable features for contextual reasoning."
                    },
                    {
                        "title": "Make Me Happier: Evoking Emotions Through Image Diffusion Models",
                        "abstract": "Despite the rapid progress in image generation, emotional image editing remains under-explored. The semantics, context, and structure of an image can evoke emotional responses, making emotional image editing techniques valuable for various real-world applications, including treatment of psychological disorders, commercialization of products, and artistic design. For the first time, we present a novel challenge of emotion-evoked image generation, aiming to synthesize images that evoke target emotions while retaining the semantics and structures of the original scenes. To address this challenge, we propose a diffusion model capable of effectively understanding and editing source images to convey desired emotions and sentiments. Moreover, due to the lack of emotion editing datasets, we provide a unique dataset consisting of 340,000 pairs of images and their emotion annotations. Furthermore, we conduct human psychophysics experiments and introduce four new evaluation metrics to systematically benchmark all the methods. Experimental results demonstrate that our method surpasses all competitive baselines. Our diffusion model is capable of identifying emotional cues from original images, editing images that elicit desired emotions, and meanwhile, preserving the semantic structure of the original images. All code, model, and dataset will be made public."
                    },
                    {
                        "title": "Improving generalization by mimicking the human visual diet",
                        "abstract": "We present a new perspective on bridging the generalization gap between biological and computer vision -- mimicking the human visual diet. While computer vision models rely on internet-scraped datasets, humans learn from limited 3D scenes under diverse real-world transformations with objects in natural context. Our results demonstrate that incorporating variations and contextual cues ubiquitous in the human visual training data (visual diet) significantly improves generalization to real-world transformations such as lighting, viewpoint, and material changes. This improvement also extends to generalizing from synthetic to real-world data -- all models trained with a human-like visual diet outperform specialized architectures by large margins when tested on natural image data. These experiments are enabled by our two key contributions: a novel dataset capturing scene context and diverse real-world transformations to mimic the human visual diet, and a transformer model tailored to leverage these aspects of the human visual diet. All data and source code can be accessed at https://github.com/Spandan-Madan/human_visual_diet."
                    },
                    {
                        "title": "Unsupervised Prior Learning: Discovering Categorical Pose Priors from Videos",
                        "abstract": "A prior represents a set of beliefs or assumptions about a system, aiding inference and decision-making. In this work, we introduce the challenge of unsupervised prior learning in pose estimation, where AI models learn pose priors of animate objects from videos in a self-supervised manner. These videos present objects performing various actions, providing crucial information about their keypoints and connectivity. While priors are effective in pose estimation, acquiring them can be difficult. We propose a novel method, named Pose Prior Learner (PPL), to learn general pose priors applicable to any object category. PPL uses a hierarchical memory to store compositional parts of prototypical poses, from which we distill a general pose prior. This prior enhances pose estimation accuracy through template transformation and image reconstruction. PPL learns meaningful pose priors without any additional human annotations or interventions, outperforming competitive baselines on both human and animal pose estimation datasets. Notably, our experimental results reveal the effectiveness of PPL using learnt priors for pose estimation on occluded images. Through iterative inference, PPL leverages priors to refine estimated poses, regressing them to any prototypical poses stored in memory. Our code, model, and data will be publicly available."
                    },
                    {
                        "title": "Reason from Context with Self-supervised Learning",
                        "abstract": "Self-supervised learning (SSL) learns to capture discriminative visual features useful for knowledge transfers. To better accommodate the object-centric nature of current downstream tasks such as object recognition and detection, various methods have been proposed to suppress contextual biases or disentangle objects from contexts. Nevertheless, these methods may prove inadequate in situations where object identity needs to be reasoned from associated context, such as recognizing or inferring tiny or obscured objects. As an initial effort in the SSL literature, we investigate whether and how contextual associations can be enhanced for visual reasoning within SSL regimes, by (a) proposing a new Self-supervised method with external memories for Context Reasoning (SeCo), and (b) introducing two new downstream tasks, lift-the-flap and object priming, addressing the problems of \"what\" and \"where\" in context reasoning. In both tasks, SeCo outperformed all state-of-the-art (SOTA) SSL methods by a significant margin. Our network analysis revealed that the proposed external memory in SeCo learns to store prior contextual knowledge, facilitating target identity inference in the lift-the-flap task. Moreover, we conducted psychophysics experiments and introduced a Human benchmark in Object Priming dataset (HOP). Our results demonstrate that SeCo exhibits human-like behaviors."
                    },
                    {
                        "title": "Variational Prototype Replays for Continual Learning",
                        "abstract": "Continual learning refers to the ability to acquire and transfer knowledge without catastrophically forgetting what was previously learned. In this work, we consider \\emph{few-shot} continual learning in classification tasks, and we propose a novel method, Variational Prototype Replays, that efficiently consolidates and recalls previous knowledge to avoid catastrophic forgetting. In each classification task, our method learns a set of variational prototypes with their means and variances, where embedding of the samples from the same class can be represented in a prototypical distribution and class-representative prototypes are separated apart. To alleviate catastrophic forgetting, our method replays one sample per class from previous tasks, and correspondingly matches newly predicted embeddings to their nearest class-representative prototypes stored from previous tasks. Compared with recent continual learning approaches, our method can readily adapt to new tasks with more classes without requiring the addition of new units. Furthermore, our method is more memory efficient since only class-representative prototypes with their means and variances, as well as only one sample per class from previous tasks need to be stored. Without tampering with the performance on initial tasks, our method learns novel concepts given a few training examples of each class in new tasks."
                    },
                    {
                        "title": "When Pigs Fly: Contextual Reasoning in Synthetic and Natural Scenes",
                        "abstract": "Context is of fundamental importance to both human and machine vision; e.g., an object in the air is more likely to be an airplane than a pig. The rich notion of context incorporates several aspects including physics rules, statistical co-occurrences, and relative object sizes, among others. While previous work has focused on crowd-sourced out-of-context photographs from the web to study scene context, controlling the nature and extent of contextual violations has been a daunting task. Here we introduce a diverse, synthetic Out-of-Context Dataset (OCD) with fine-grained control over scene context. By leveraging a 3D simulation engine, we systematically control the gravity, object co-occurrences and relative sizes across 36 object categories in a virtual household environment. We conducted a series of experiments to gain insights into the impact of contextual cues on both human and machine vision using OCD. We conducted psychophysics experiments to establish a human benchmark for out-of-context recognition, and then compared it with state-of-the-art computer vision models to quantify the gap between the two. We propose a context-aware recognition transformer model, fusing object and contextual information via multi-head attention. Our model captures useful information for contextual reasoning, enabling human-level performance and better robustness in out-of-context conditions compared to baseline models across OCD and other out-of-context datasets. All source code and data are publicly available at https://github.com/kreimanlab/WhenPigsFlyContext"
                    },
                    {
                        "title": "Efficient Zero-shot Visual Search via Target and Context-aware Transformer",
                        "abstract": "Visual search is a ubiquitous challenge in natural vision, including daily tasks such as finding a friend in a crowd or searching for a car in a parking lot. Human rely heavily on relevant target features to perform goal-directed visual search. Meanwhile, context is of critical importance for locating a target object in complex scenes as it helps narrow down the search area and makes the search process more efficient. However, few works have combined both target and context information in visual search computational models. Here we propose a zero-shot deep learning architecture, TCT (Target and Context-aware Transformer), that modulates self attention in the Vision Transformer with target and contextual relevant information to enable human-like zero-shot visual search performance. Target modulation is computed as patch-wise local relevance between the target and search images, whereas contextual modulation is applied in a global fashion. We conduct visual search experiments on TCT and other competitive visual search models on three natural scene datasets with varying levels of difficulty. TCT demonstrates human-like performance in terms of search efficiency and beats the SOTA models in challenging visual search tasks. Importantly, TCT generalizes well across datasets with novel objects without retraining or fine-tuning. Furthermore, we also introduce a new dataset to benchmark models for invariant visual search under incongruent contexts. TCT manages to search flexibly via target and context modulation, even under incongruent contexts."
                    },
                    {
                        "title": "Tuned Compositional Feature Replays for Efficient Stream Learning",
                        "abstract": "Our brains extract durable, generalizable knowledge from transient experiences of the world. Artificial neural networks come nowhere close to this ability. When tasked with learning to classify objects by training on non-repeating video frames in temporal order (online stream learning), models that learn well from shuffled datasets catastrophically forget old knowledge upon learning new stimuli. We propose a new continual learning algorithm, Compositional Replay Using Memory Blocks (CRUMB), which mitigates forgetting by replaying feature maps reconstructed by combining generic parts. CRUMB concatenates trainable and re-usable \"memory block\" vectors to compositionally reconstruct feature map tensors in convolutional neural networks. Storing the indices of memory blocks used to reconstruct new stimuli enables memories of the stimuli to be replayed during later tasks. This reconstruction mechanism also primes the neural network to minimize catastrophic forgetting by biasing it towards attending to information about object shapes more than information about image textures, and stabilizes the network during stream learning by providing a shared feature-level basis for all training examples. These properties allow CRUMB to outperform an otherwise identical algorithm that stores and replays raw images, while occupying only 3.6% as much memory. We stress-tested CRUMB alongside 13 competing methods on 7 challenging datasets. To address the limited number of existing online stream learning datasets, we introduce 2 new benchmarks by adapting existing datasets for stream learning. With only 3.7-4.1% as much memory and 15-43% as much runtime, CRUMB mitigates catastrophic forgetting more effectively than the state-of-the-art. Our code is available at https://github.com/MorganBDT/crumb.git."
                    }
                ]
            }
        }
    },
    "2405.14064": {
        "paper_data": {
            "title": "Building a stable classifier with the inflated argmax",
            "url": "http://arxiv.org/abs/2405.14064v1",
            "arxiv_id": "2405.14064",
            "authors": [
                "Jake A. Soloff",
                "Rina Foygel Barber",
                "Rebecca Willett"
            ],
            "abstract": "We propose a new framework for algorithmic stability in the context of multiclass classification. In practice, classification algorithms often operate by first assigning a continuous score (for instance, an estimated probability) to each possible label, then taking the maximizer -- i.e., selecting the class that has the highest score. A drawback of this type of approach is that it is inherently unstable, meaning that it is very sensitive to slight perturbations of the training data, since taking the maximizer is discontinuous. Motivated by this challenge, we propose a pipeline for constructing stable classifiers from data, using bagging (i.e., resampling and averaging) to produce stable continuous scores, and then using a stable relaxation of argmax, which we call the \"inflated argmax,\" to convert these scores to a set of candidate labels. The resulting stability guarantee places no distributional assumptions on the data, does not depend on the number of classes or dimensionality of the covariates, and holds for any base classifier. Using a common benchmark data set, we demonstrate that the inflated argmax provides necessary protection against unstable classifiers, without loss of accuracy.",
            "introduction": "   1 Introduction  An algorithm that learns from data is considered to be stable if small perturbations of the training data do not lead to large changes in its output. Algorithmic stability is an important consideration in many statistical applications. Within the fairness literature, for instance, stability is one aspect of reliable decision-making systems (Friedler et\u00a0al.,, 2019; Huang and Vishnoi,, 2019). In interpretable machine learning, it similarly serves as a form of reproducibility (Murdoch et\u00a0al.,, 2019; Yu and Kumbier,, 2020; Yu and Barter,, 2024). Caramanis et\u00a0al., (2011) relates stability to robustness, where a machine learning algorithm is robust if two samples with similar feature vectors have similar test error. In the context of generalization bounds, Bousquet and Elisseeff, (2002) and subsequent authors study stability of an algorithm\u2019s real-valued output\u2014for instance, values of a regression function. In the setting of a multiclass classification problem, where the data consists of features X\u2208\ud835\udcb3\ud835\udc4b\ud835\udcb3X\\in\\mathcal{X}italic_X \u2208 caligraphic_X and labels Y\u2208[L]={1,\u2026,L}\ud835\udc4cdelimited-[]\ud835\udc3f1\u2026\ud835\udc3fY\\in[L]=\\{1,\\dots,L\\}italic_Y \u2208 [ italic_L ] = { 1 , \u2026 , italic_L }, the results of this literature can thus be applied to analyzing a classifier\u2019s predicted probabilities p^\u2113\u2062(X)\u2208[0,1]subscript^\ud835\udc5d\u2113\ud835\udc4b01\\hat{p}_{\\ell}(X)\\in[0,1]over^ start_ARG italic_p end_ARG start_POSTSUBSCRIPT roman_\u2113 end_POSTSUBSCRIPT ( italic_X ) \u2208 [ 0 , 1 ]\u2014i.e., our learned estimates of the conditional probabilities Pr\u2061{Y=\u2113\u2223X}Pr\ud835\udc4cconditional\u2113\ud835\udc4b\\Pr\\{Y=\\ell\\mid X\\}roman_Pr { italic_Y = roman_\u2113 \u2223 italic_X }, for each \u2113\u2208[L]\u2113delimited-[]\ud835\udc3f\\ell\\in[L]roman_\u2113 \u2208 [ italic_L ]. However, as we will see, stability of these predicted probabilities by no means implies stability of the predicted label itself, y^=argmax\u2113\u2208[L]\u2061p^\u2113\u2062(x)^\ud835\udc66subscriptargmax\u2113delimited-[]\ud835\udc3fsubscript^\ud835\udc5d\u2113\ud835\udc65\\hat{y}=\\operatorname{argmax}_{\\ell\\in[L]}\\hat{p}_{\\ell}(x)over^ start_ARG italic_y end_ARG = roman_argmax start_POSTSUBSCRIPT roman_\u2113 \u2208 [ italic_L ] end_POSTSUBSCRIPT over^ start_ARG italic_p end_ARG start_POSTSUBSCRIPT roman_\u2113 end_POSTSUBSCRIPT ( italic_x )\u2014an arbitrarily small perturbation in p^\u2113\u2062(x)subscript^\ud835\udc5d\u2113\ud835\udc65\\hat{p}_{\\ell}(x)over^ start_ARG italic_p end_ARG start_POSTSUBSCRIPT roman_\u2113 end_POSTSUBSCRIPT ( italic_x ) might completely change the predicted label y^^\ud835\udc66\\hat{y}over^ start_ARG italic_y end_ARG. The distinction matters: for system trustworthiness, we care about the model\u2019s final output, on which users base their decisions.   In this paper, we propose a new framework for algorithmic stability in the context of multiclass classification, to define\u2014and achieve\u2014a meaningful notion of stability when the output of the algorithm consists of predicted labels, rather than predicted probabilities. Our work connects to other approaches to allowing for ambiguous classification, including set-valued classification (Grycko,, 1993; Del\u00a0Coz et\u00a0al.,, 2009; Sadinle et\u00a0al.,, 2019; Chzhen et\u00a0al.,, 2021) and conformal inference\u00a0(Papadopoulos et\u00a0al.,, 2002; Vovk et\u00a0al.,, 2005; Lei,, 2014; Angelopoulos et\u00a0al.,, 2023) (we will discuss related work further, below).    1.1 Problem setting  In supervised classification, we take a data set \ud835\udc9f=((Xi,Yi))i\u2208[n]\ud835\udc9fsubscriptsubscript\ud835\udc4b\ud835\udc56subscript\ud835\udc4c\ud835\udc56\ud835\udc56delimited-[]\ud835\udc5b\\mathcal{D}=\\big{(}(X_{i},Y_{i})\\big{)}_{i\\in[n]}caligraphic_D = ( ( italic_X start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT , italic_Y start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) ) start_POSTSUBSCRIPT italic_i \u2208 [ italic_n ] end_POSTSUBSCRIPT consisting of n\ud835\udc5bnitalic_n observations, and train a classifier that maps from the feature space \ud835\udcb3\ud835\udcb3\\mathcal{X}caligraphic_X to the set [L]delimited-[]\ud835\udc3f[L][ italic_L ] of possible labels.111We remark that all the results of this paper apply also to the case of countably infinitely many labels, L=\u221e\ud835\udc3fL=\\inftyitalic_L = \u221e, in which case we should take \u2115\u2115\\mathbb{N}blackboard_N to be the label space instead of [L]delimited-[]\ud835\udc3f[L][ italic_L ]. Typically, this map is constructed in two stages. First we run some form of regression to learn a map\u00a0p^:\ud835\udcb3\u2192\u0394L\u22121:^\ud835\udc5d\u2192\ud835\udcb3subscript\u0394\ud835\udc3f1\\hat{p}:\\mathcal{X}\\to\\Delta_{L-1}over^ start_ARG italic_p end_ARG : caligraphic_X \u2192 roman_\u0394 start_POSTSUBSCRIPT italic_L - 1 end_POSTSUBSCRIPT from features to predicted probabilities, with p^\u2113\u2062(x)subscript^\ud835\udc5d\u2113\ud835\udc65\\hat{p}_{\\ell}(x)over^ start_ARG",
            "references": [
                {
                    "title": "Stability via resampling: statistical problems beyond the real line",
                    "abstract": "Model averaging techniques based on resampling methods (such as bootstrapping or subsampling) have been utilized across many areas of statistics, often with the explicit goal of promoting stability in the resulting output. We provide a general, finite-sample theoretical result guaranteeing the stability of bagging when applied to algorithms that return outputs in a general space, so that the output is not necessarily a real-valued -- for example, an algorithm that estimates a vector of weights or a density function. We empirically assess the stability of bagging on synthetic and real-world data for a range of problem settings, including causal inference, nonparametric regression, and Bayesian model selection."
                },
                {
                    "title": "Conformal Language Modeling",
                    "abstract": "We propose a novel approach to conformal prediction for generative language models (LMs). Standard conformal prediction produces prediction sets -- in place of single predictions -- that have rigorous, statistical performance guarantees. LM responses are typically sampled from the model's predicted distribution over the large, combinatorial output space of natural language. Translating this process to conformal prediction, we calibrate a stopping rule for sampling different outputs from the LM that get added to a growing set of candidates until we are confident that the output set is sufficient. Since some samples may be low-quality, we also simultaneously calibrate and apply a rejection rule for removing candidates from the output set to reduce noise. Similar to conformal prediction, we prove that the sampled set returned by our procedure contains at least one acceptable answer with high probability, while still being empirically precise (i.e., small) on average. Furthermore, within this set of candidate responses, we show that we can also accurately identify subsets of individual components -- such as phrases or sentences -- that are each independently correct (e.g., that are not\"hallucinations\"), again with statistical guarantees. We demonstrate the promise of our approach on multiple tasks in open-domain question answering, text summarization, and radiology report generation using different LM variants."
                },
                {
                    "title": "From Softmax to Nucleusmax: A Novel Sparse Language Model for Chinese Radiology Report Summarization",
                    "abstract": "The Chinese radiology report summarization is a crucial component in smart healthcare that employs language models to summarize key findings in radiology reports and communicate these findings to physicians. However, most language models for radiology report summarization utilize a softmax transformation in their output layer, leading to dense alignments and strictly positive output probabilities. This density is inefficient, reducing model interpretability and giving probability mass to many unrealistic outputs. To tackle this issue, we propose a novel approach named nucleusmax. Nucleusmax is able to mitigate dense outputs and improve model interpretability by truncating the unreliable tail of the probability distribution. In addition, we incorporate nucleusmax with a copy mechanism, a useful technique to avoid professional errors in the generated diagnostic opinions. To further promote the research of radiology report summarization, we also have created a Chinese radiology report summarization dataset, which is freely available. Experimental results showed via both automatic and human evaluation that the proposed approach substantially improves the sparsity and overall quality of outputs over competitive softmax models, producing radiology summaries that approach the quality of those authored by physicians. In general, our work demonstrates the feasibility and prospect of the language model to the domain of radiology and smart healthcare."
                },
                {
                    "title": "Bagging Provides Assumption-free Stability",
                    "abstract": "Bagging is an important technique for stabilizing machine learning models. In this paper, we derive a finite-sample guarantee on the stability of bagging for any model. Our result places no assumptions on the distribution of the data, on the properties of the base algorithm, or on the dimensionality of the covariates. Our guarantee applies to many variants of bagging and is optimal up to a constant. Empirical results validate our findings, showing that bagging successfully stabilizes even highly unstable base algorithms."
                },
                {
                    "title": "Set-valued classification - overview via a unified framework",
                    "abstract": "Multi-class classification problem is among the most popular and well-studied statistical frameworks. Modern multi-class datasets can be extremely ambiguous and single-output predictions fail to deliver satisfactory performance. By allowing predictors to predict a set of label candidates, set-valued classification offers a natural way to deal with this ambiguity. Several formulations of set-valued classification are available in the literature and each of them leads to different prediction strategies. The present survey aims to review popular formulations using a unified statistical framework. The proposed framework encompasses previously considered and leads to new formulations as well as it allows to understand underlying trade-offs of each formulation. We provide infinite sample optimal set-valued classification strategies and review a general plug-in principle to construct data-driven algorithms. The exposition is supported by examples and pointers to both theoretical and practical contributions. Finally, we provide experiments on real-world datasets comparing these approaches in practice and providing general practical guidelines."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Stable and Fair Classification",
                    "abstract": "Fair classification has been a topic of intense study in machine learning, and several algorithms have been proposed towards this important task. However, in a recent study, Friedler et al. observed that fair classification algorithms may not be stable with respect to variations in the training dataset -- a crucial consideration in several real-world applications. Motivated by their work, we study the problem of designing classification algorithms that are both fair and stable. We propose an extended framework based on fair classification algorithms that are formulated as optimization problems, by introducing a stability-focused regularization term. Theoretically, we prove a stability guarantee, that was lacking in fair classification algorithms, and also provide an accuracy guarantee for our extended framework. Our accuracy guarantee can be used to inform the selection of the regularization parameter in our framework. To the best of our knowledge, this is the first work that combines stability and fairness in automated decision-making tasks. We assess the benefits of our approach empirically by extending several fair classification algorithms that are shown to achieve the best balance between fairness and accuracy over the Adult dataset. Our empirical results show that our framework indeed improves the stability at only a slight sacrifice in accuracy."
                },
                {
                    "title": "Veridical data science",
                    "abstract": "Significance Predictability, computability, and stability (PCS) are three core principles of data science. They embed the scientific principles of prediction and replication in data-driven decision making while recognizing the central role of computation. Based on these principles, we propose the PCS framework, including workflow and documentation (in R Markdown or Jupyter Notebook). The PCS framework aims at responsible, reliable, reproducible, and transparent analysis across fields of science, social science, engineering, business, and government. It can be used as a recommendation system for scientific hypothesis generation and experimental design. In particular, we propose (basic) PCS inference for reliability measures on data results, extending statistical inference to a much broader scope as current data science practice entails. Building and expanding on principles of statistics, machine learning, and scientific inquiry, we propose the predictability, computability, and stability (PCS) framework for veridical data science. Our framework, composed of both a workflow and documentation, aims to provide responsible, reliable, reproducible, and transparent results across the data science life cycle. The PCS workflow uses predictability as a reality check and considers the importance of computation in data collection/storage and algorithm design. It augments predictability and computability with an overarching stability principle. Stability expands on statistical uncertainty considerations to assess how human judgment calls impact data results through data and model/algorithm perturbations. As part of the PCS workflow, we develop PCS inference procedures, namely PCS perturbation intervals and PCS hypothesis testing, to investigate the stability of data results relative to problem formulation, data cleaning, modeling decisions, and interpretations. We illustrate PCS inference through neuroscience and genomics projects of our own and others. Moreover, we demonstrate its favorable performance over existing methods in terms of receiver operating characteristic (ROC) curves in high-dimensional, sparse linear model simulations, including a wide range of misspecified models. Finally, we propose PCS documentation based on R Markdown or Jupyter Notebook, with publicly available, reproducible codes and narratives to back up human choices made throughout an analysis. The PCS workflow and documentation are demonstrated in a genomics case study available on Zenodo."
                },
                {
                    "title": "Definitions, methods, and applications in interpretable machine learning",
                    "abstract": "Significance The recent surge in interpretability research has led to confusion on numerous fronts. In particular, it is unclear what it means to be interpretable and how to select, evaluate, or even discuss methods for producing interpretations of machine-learning models. We aim to clarify these concerns by defining interpretable machine learning and constructing a unifying framework for existing methods which highlights the underappreciated role played by human audiences. Within this framework, methods are organized into 2 classes: model based and post hoc. To provide guidance in selecting and evaluating interpretation methods, we introduce 3 desiderata: predictive accuracy, descriptive accuracy, and relevancy. Using our framework, we review existing work, grounded in real-world studies which exemplify our desiderata, and suggest directions for future work. Machine-learning models have demonstrated great success in learning complex patterns that enable them to make predictions about unobserved data. In addition to using models for prediction, the ability to interpret what a model has learned is receiving an increasing amount of attention. However, this increased focus has led to considerable confusion about the notion of interpretability. In particular, it is unclear how the wide array of proposed interpretation methods are related and what common concepts can be used to evaluate them. We aim to address these concerns by defining interpretability in the context of machine learning and introducing the predictive, descriptive, relevant (PDR) framework for discussing interpretations. The PDR framework provides 3 overarching desiderata for evaluation: predictive accuracy, descriptive accuracy, and relevancy, with relevancy judged relative to a human audience. Moreover, to help manage the deluge of interpretation methods, we introduce a categorization of existing techniques into model-based and post hoc categories, with subgroups including sparsity, modularity, and simulatability. To demonstrate how practitioners can use the PDR framework to evaluate and understand interpretations, we provide numerous real-world examples. These examples highlight the often underappreciated role played by human audiences in discussions of interpretability. Finally, based on our framework, we discuss limitations of existing methods and directions for future work. We hope that this work will provide a common vocabulary that will make it easier for both practitioners and researchers to discuss and choose from the full range of interpretation methods."
                },
                {
                    "title": "Learning with Fenchel-Young Losses",
                    "abstract": "Over the past decades, numerous loss functions have been been proposed for a variety of supervised learning tasks, including regression, classification, ranking, and more generally structured prediction. Understanding the core principles and theoretical properties underpinning these losses is key to choose the right loss for the right problem, as well as to create new losses which combine their strengths. In this paper, we introduce Fenchel-Young losses, a generic way to construct a convex loss function for a regularized prediction function. We provide an in-depth study of their properties in a very broad setting, covering all the aforementioned supervised learning tasks, and revealing new connections between sparsity, generalized entropies, and separation margins. We show that Fenchel-Young losses unify many well-known loss functions and allow to create useful new ones easily. Finally, we derive efficient predictive and training algorithms, making Fenchel-Young losses appealing both in theory and practice."
                },
                {
                    "title": "Cautious Deep Learning",
                    "abstract": "Most classifiers operate by selecting the maximum of an estimate of the conditional distribution $p(y|x)$ where $x$ stands for the features of the instance to be classified and $y$ denotes its label. This often results in a {\\em hubristic bias}: overconfidence in the assignment of a definite label. Usually, the observations are concentrated on a small volume but the classifier provides definite predictions for the entire space. We propose constructing conformal prediction sets which contain a set of labels rather than a single label. These conformal prediction sets contain the true label with probability $1-\\alpha$. Our construction is based on $p(x|y)$ rather than $p(y|x)$ which results in a classifier that is very cautious: it outputs the null set --- meaning \"I don't know\" --- when the object does not resemble the training examples. An important property of our approach is that adversarial attacks are likely to be predicted as the null set or would also include the true label. We demonstrate the performance on the ImageNet ILSVRC dataset and the CelebA and IMDB-Wiki facial datasets using high dimensional features obtained from state of the art convolutional neural networks."
                },
                {
                    "title": "A comparative study of fairness-enhancing interventions in machine learning",
                    "abstract": "Computers are increasingly used to make decisions that have significant impact on people's lives. Often, these predictions can affect different population subgroups disproportionately. As a result, the issue of fairness has received much recent interest, and a number of fairness-enhanced classifiers have appeared in the literature. This paper seeks to study the following questions: how do these different techniques fundamentally compare to one another, and what accounts for the differences? Specifically, we seek to bring attention to many under-appreciated aspects of such fairness-enhancing interventions that require investigation for these algorithms to receive broad adoption. We present the results of an open benchmark we have developed that lets us compare a number of different algorithms under a variety of fairness measures and existing datasets. We find that although different algorithms tend to prefer specific formulations of fairness preservations, many of these measures strongly correlate with one another. In addition, we find that fairness-preserving algorithms tend to be sensitive to fluctuations in dataset composition (simulated in our benchmark by varying training-test splits) and to different forms of preprocessing, indicating that fairness interventions might be more brittle than previously thought."
                },
                {
                    "title": "Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms",
                    "abstract": "We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at this https URL"
                },
                {
                    "title": "Least Ambiguous Set-Valued Classifiers With Bounded Error Levels",
                    "abstract": "ABSTRACT In most classification tasks, there are observations that are ambiguous and therefore difficult to correctly label. Set-valued classifiers output sets of plausible labels rather than a single label, thereby giving a more appropriate and informative treatment to the labeling of ambiguous instances. We introduce a framework for multiclass set-valued classification, where the classifiers guarantee user-defined levels of coverage or confidence (the probability that the true label is contained in the set) while minimizing the ambiguity (the expected size of the output). We first derive oracle classifiers assuming the true distribution to be known. We show that the oracle classifiers are obtained from level sets of the functions that define the conditional probability of each class. Then we develop estimators with good asymptotic and finite sample properties. The proposed estimators build on existing single-label classifiers. The optimal classifier can sometimes output the empty set, but we provide two solutions to fix this issue that are suitable for various practical needs. Supplementary materials for this article are available online."
                },
                {
                    "title": "From Softmax to Sparsemax: A Sparse Model of Attention and Multi-Label Classification",
                    "abstract": "We propose sparsemax, a new activation function similar to the traditional softmax, but able to output sparse probabilities. After deriving its properties, we show how its Jacobian can be efficiently computed, enabling its use in a network trained with backpropagation. Then, we propose a new smooth and convex loss function which is the sparsemax analogue of the logistic loss. We reveal an unexpected connection between this new loss and the Huber classification loss. We obtain promising empirical results in multi-label classification problems and in attention-based neural networks for natural language inference. For the latter, we achieve a similar performance as the traditional softmax, but with a selective, more compact, attention focus."
                },
                {
                    "title": "An Exploration of Softmax Alternatives Belonging to the Spherical Loss Family",
                    "abstract": "In a multi-class classification problem, it is standard to model the output of a neural network as a categorical distribution conditioned on the inputs. The output must therefore be positive and sum to one, which is traditionally enforced by a softmax. This probabilistic mapping allows to use the maximum likelihood principle, which leads to the well-known log-softmax loss. However the choice of the softmax function seems somehow arbitrary as there are many other possible normalizing functions. It is thus unclear why the log-softmax loss would perform better than other loss alternatives. In particular Vincent et al. (2015) recently introduced a class of loss functions, called the spherical family, for which there exists an efficient algorithm to compute the updates of the output weights irrespective of the output size. In this paper, we explore several loss functions from this family as possible alternatives to the traditional log-softmax. In particular, we focus our investigation on spherical bounds of the log-softmax loss and on two spherical log-likelihood losses, namely the log-Spherical Softmax suggested by Vincent et al. (2015) and the log-Taylor Softmax that we introduce. Although these alternatives do not yield as good results as the log-softmax loss on two language modeling tasks, they surprisingly outperform it in our experiments on MNIST and CIFAR-10, suggesting that they might be relevant in a broad range of applications."
                },
                {
                    "title": "Classification with confidence",
                    "abstract": "A framework for classification is developed with a notion of confidence. In this framework, a classifier consists of two tolerance regions in the predictor space, with a specified coverage level for each class. The classifier also produces an ambiguous region where the classification needs further investigation. Theoretical analysis reveals interesting structures of the confidence-ambiguity trade-off, and the optimal solution is characterized by extending the Neyman\u2013Pearson lemma. We provide general estimating procedures, along with rates of convergence, based on estimates of the conditional probabilities. The method can be easily implemented with good robustness, as illustrated through theory, simulation and a data example."
                },
                {
                    "title": "Learning Nondeterministic Classifiers",
                    "abstract": "Nondeterministic classifiers are defined as those allowed to predict more than one class for some entries from an input space. Given that the true class should be included in predictions and the number of classes predicted should be as small as possible, these kind of classifiers can be considered as Information Retrieval (IR) procedures. In this paper, we propose a family of IR loss functions to measure the performance of nondeterministic learners. After discussing such measures, we derive an algorithm for learning optimal nondeterministic hypotheses. Given an entry from the input space, the algorithm requires the posterior probabilities to compute the subset of classes with the lowest expected loss. From a general point of view, nondeterministic classifiers provide an improvement in the proportion of predictions that include the true class compared to their deterministic counterparts; the price to be paid for this increase is usually a tiny proportion of predictions with more than one class. The paper includes an extensive experimental study using three deterministic learners to estimate posterior probabilities: a multiclass Support Vector Machine (SVM), a Logistic Regression, and a Naive Bayes. The data sets considered comprise both UCI multi-class learning tasks and microarray expressions of different kinds of cancer. We successfully compare nondeterministic classifiers with other alternative approaches. Additionally, we shall see how the quality of posterior probabilities (measured by the Brier score) determines the goodness of nondeterministic predictions."
                },
                {
                    "title": "The Kinematics of Machinery: Outlines of a Theory of Machines",
                    "abstract": "Ubersetzung von \"Theoretische Kinematik\" / Text und Atlas (Die Figuren aus dem Atlas wurden hier mit in den Text eingebunden)"
                },
                {
                    "title": "A Simple Algorithm for Learning Stable Machines",
                    "abstract": "We present an algorithm for learning stable machines which is motivated by recent results in statistical learning theory. The algorithm is similar to Breiman's bagging despite some important differences in that it computes an ensemble combination of machines trained on small random sub-samples of an initial training set. A remarkable property is that it is often possible to just use the empirical error of these combinations of machines for model selection. We report experiments using support vector machines and neural networks validating the theory."
                },
                {
                    "title": "Heuristics of instability and stabilization in model selection",
                    "abstract": "In model selection, usually a best predictor is chosen from a collection {\u03bc(.,s)} of predictors where \u03bc(.,s) is the minimum least-squares predictor in a collection U s of predictors. Here s is a complexity parameter; that is, the smaller s, the lower dimensional/smoother the models in U s . If L is the data used to derive the sequence {\u03bc(., s)}, the procedure is called unstable if a small change in L can cause large changes in {\u03bc(., s)}. With a crystal ball, one could pick the predictor in {\u03bc(.,s)} having minimum prediction error. Without prescience, one uses test sets, cross-validation and so forth. The difference in prediction error between the crystal ball selection and the statistician's choice we call predictive loss. For an unstable procedure the predictive loss is large. This is shown by some analytics in a simple case and by simulation results in a more complex comparison of four different linear regression methods. Unstable procedures can be stabilized by perturbing the data, getting a new predictor sequence {\u03bc(., s)} and then averaging over many such predictor sequences."
                },
                {
                    "title": "Stability and Generalization",
                    "abstract": "We de\ufb01ne notions of stability for learning algorithms and show how to use these notions to derive generalization error bounds based on the empirical error and the leave-one-out error. The methods we use can be applied in the regression framework as well as in the classi\ufb01cation one when the classi\ufb01er is obtained by thresholding a real-valued function. We study the stability properties of large classes of learning algorithms such as regularization based algorithms. In particular we focus on Hilbert space regularization and Kullback-Leibler regularization. We demonstrate how to apply the results to SVM for regression and classi\ufb01cation."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG",
                "math.ST",
                "stat.TH"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we define and achieve a meaningful notion of algorithmic stability in multiclass classification when the output consists of predicted labels rather than predicted probabilities?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for enhancing the trustworthiness and reliability of machine learning models, particularly in applications where decision-making is critical. By establishing a framework for algorithmic stability in predicted labels, we can improve the interpretability and reproducibility of classification systems. This advancement could lead to more robust decision-making processes in various fields, such as healthcare, finance, and autonomous systems, ultimately influencing future research directions in algorithmic fairness and stability.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the fact that small changes in predicted probabilities can lead to significant shifts in the final predicted label, which complicates the assessment of stability. Naive approaches that focus solely on the stability of predicted probabilities may fail to capture the nuances of label prediction. Additionally, the theoretical underpinnings of stability in the context of multiclass classification are complex, requiring a deep understanding of both statistical properties and the behavior of classifiers under perturbations in the input data.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on stability in terms of predicted probabilities, neglecting the implications for the final predicted labels. Existing solutions often lack a comprehensive framework that connects stability to label prediction. Barriers such as the complexity of multiclass classification and the need for a robust theoretical foundation have hindered progress. Our approach differs by explicitly addressing the stability of predicted labels and integrating concepts from related fields, such as set-valued classification and conformal inference, to create a more holistic understanding.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a new framework for algorithmic stability that focuses on the predicted labels in multiclass classification. We will utilize a dataset consisting of labeled observations to train classifiers and evaluate their stability using metrics that assess the consistency of predicted labels under perturbations. Expected outcomes include a clearer understanding of stability in label prediction, improved robustness of classifiers, and practical guidelines for implementing stable classification systems in real-world applications."
            }
        },
        "author_data": {
            "6c3ab577-699f-4ded-bfd4-3fa1e208f5d3": {
                "pk": "6c3ab577-699f-4ded-bfd4-3fa1e208f5d3",
                "name": "Jake A. Soloff",
                "collaborators": [
                    "Adityanand Guntuboyina",
                    "Rina Foygel Barber",
                    "Rebecca Willett",
                    "Michael I. Jordan",
                    "Jim Pitman",
                    "Daniel Xiang",
                    "William Fithian",
                    "Bodhisattva Sen",
                    "Stephen Bates",
                    "Michael Sklar"
                ],
                "domain": [
                    "Statistical Inference",
                    "Machine Learning",
                    "Bayesian Methods",
                    "Covariance Estimation"
                ],
                "publications": [
                    {
                        "title": "Distribution-free properties of isotonic regression",
                        "abstract": "It is well known that the isotonic least squares estimator is characterized as the derivative of the greatest convex minorant of a random walk. Provided the walk has exchangeable increments, we prove that the slopes of the greatest convex minorant are distributed as order statistics of the running averages. This result implies an exact non-asymptotic formula for the squared error risk of least squares in isotonic regression when the true sequence is constant that holds for every exchangeable error distribution."
                    },
                    {
                        "title": "Bagging Provides Assumption-free Stability",
                        "abstract": "Bagging is an important technique for stabilizing machine learning models. In this paper, we derive a finite-sample guarantee on the stability of bagging for any model. Our result places no assumptions on the distribution of the data, on the properties of the base algorithm, or on the dimensionality of the covariates. Our guarantee applies to many variants of bagging and is optimal up to a constant. Empirical results validate our findings, showing that bagging successfully stabilizes even highly unstable base algorithms."
                    },
                    {
                        "title": "Covariance estimation with nonnegative partial correlations",
                        "abstract": "We study the problem of high-dimensional covariance estimation under the constraint that the partial correlations are nonnegative. The sign constraints dramatically simplify estimation: the Gaussian maximum likelihood estimator is well defined with only two observations regardless of the number of variables. We analyze its performance in the setting where the dimension may be much larger than the sample size. We establish that the estimator is both high-dimensionally consistent and minimax optimal in the symmetrized Stein loss. We also prove a negative result which shows that the sign-constraints can introduce substantial bias for estimating the top eigenvalue of the covariance matrix."
                    },
                    {
                        "title": "The edge of discovery: Controlling the local false discovery rate at the margin",
                        "abstract": "Despite the popularity of the false discovery rate (FDR) as an error control metric for large-scale multiple testing, its close Bayesian counterpart the local false discovery rate (lfdr), defined as the posterior probability that a particular null hypothesis is false, is a more directly relevant standard for justifying and interpreting individual rejections. However, the lfdr is difficult to work with in small samples, as the prior distribution is typically unknown. We propose a simple multiple testing procedure and prove that it controls the expectation of the maximum lfdr across all rejections; equivalently, it controls the probability that the rejection with the largest p-value is a false discovery. Our method operates without knowledge of the prior, assuming only that the p-value density is uniform under the null and decreasing under the alternative. We also show that our method asymptotically implements the oracle Bayes procedure for a weighted classification risk, optimally trading off between false positives and false negatives. We derive the limiting distribution of the attained maximum lfdr over the rejections, and the limiting empirical Bayes regret relative to the oracle procedure."
                    },
                    {
                        "title": "Multivariate, Heteroscedastic Empirical Bayes via Nonparametric Maximum Likelihood",
                        "abstract": "Multivariate, heteroscedastic errors complicate statistical inference in many large-scale denoising problems. Empirical Bayes is attractive in such settings, but standard parametric approaches rest on assumptions about the form of the prior distribution which can be hard to justify and which introduce unnecessary tuning parameters. We extend the nonparametric maximum likelihood estimator (NPMLE) for Gaussian location mixture densities to allow for multivariate, heteroscedastic errors. NPMLEs estimate an arbitrary prior by solving an infinite-dimensional, convex optimization problem; we show that this convex optimization problem can be tractably approximated by a finite-dimensional version.   The empirical Bayes posterior means based on an NPMLE have low regret, meaning they closely target the oracle posterior means one would compute with the true prior in hand. We prove an oracle inequality implying that the empirical Bayes estimator performs at nearly the optimal level (up to logarithmic factors) for denoising without prior knowledge. We provide finite-sample bounds on the average Hellinger accuracy of an NPMLE for estimating the marginal densities of the observations. We also demonstrate the adaptive and nearly-optimal properties of NPMLEs for deconvolution. We apply our method to two denoising problems in astronomy, constructing a fully data-driven color-magnitude diagram of 1.4 million stars in the Milky Way and investigating the distribution of 19 chemical abundance ratios for 27 thousand stars in the red clump. We also apply our method to hierarchical linear models, illustrating the advantages of nonparametric shrinkage of regression coefficients on an education data set and on a microarray data set."
                    },
                    {
                        "title": "Stability via resampling: statistical problems beyond the real line",
                        "abstract": "Model averaging techniques based on resampling methods (such as bootstrapping or subsampling) have been utilized across many areas of statistics, often with the explicit goal of promoting stability in the resulting output. We provide a general, finite-sample theoretical result guaranteeing the stability of bagging when applied to algorithms that return outputs in a general space, so that the output is not necessarily a real-valued -- for example, an algorithm that estimates a vector of weights or a density function. We empirically assess the stability of bagging on synthetic and real-world data for a range of problem settings, including causal inference, nonparametric regression, and Bayesian model selection."
                    },
                    {
                        "title": "Principal-Agent Hypothesis Testing",
                        "abstract": "Consider the relationship between a regulator (the principal) and an experimenter (the agent) such as a pharmaceutical company. The pharmaceutical company wishes to sell a drug for profit, whereas the regulator wishes to allow only efficacious drugs to be marketed. The efficacy of the drug is not known to the regulator, so the pharmaceutical company must run a costly trial to prove efficacy to the regulator. Critically, the statistical protocol used to establish efficacy affects the behavior of a strategic, self-interested agent; a lower standard of statistical evidence incentivizes the agent to run more trials that are less likely to be effective. The interaction between the statistical protocol and the incentives of the pharmaceutical company is crucial for understanding this system and designing protocols with high social utility. In this work, we discuss how the regulator can set up a protocol with payoffs based on statistical evidence. We show how to design protocols that are robust to an agent's strategic actions, and derive the optimal protocol in the presence of strategic entrants."
                    }
                ]
            },
            "ac73c879-d2c1-453d-9f0f-348b50448c5a": {
                "pk": "ac73c879-d2c1-453d-9f0f-348b50448c5a",
                "name": "Rina Foygel Barber",
                "collaborators": [
                    "Emil Y. Sidky",
                    "Fan Yang",
                    "Ang Li",
                    "Mathias Drton",
                    "Yuancheng Zhu",
                    "Prateek Jain",
                    "John Lafferty",
                    "Ran Dai",
                    "Mladen Kolar",
                    "Richard J. Samworth"
                ],
                "domain": [
                    "Statistical Inference",
                    "Graphical Models",
                    "High-Dimensional Statistics",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Hoeffding and Bernstein inequalities for weighted sums of exchangeable random variables",
                        "abstract": "The aim of this paper is to establish Hoeffding and Bernstein type concentration inequalities for weighted sums of exchangeable random variables. A special case is the i.i.d. setting, where random variables are sampled independently from some distribution (and are therefore exchangeable). In contrast to the existing literature on this problem, our results provide a natural unified view of both the i.i.d. and the exchangeable setting."
                    },
                    {
                        "title": "High-dimensional Ising model selection with Bayesian information criteria",
                        "abstract": "We consider the use of Bayesian information criteria for selection of the graph underlying an Ising model. In an Ising model, the full conditional distributions of each variable form logistic regression models, and variable selection techniques for regression allow one to identify the neighborhood of each node and, thus, the entire graph. We prove high-dimensional consistency results for this pseudo-likelihood approach to graph selection when using Bayesian information criteria for the variable selection problems in the logistic regressions. The results pertain to scenarios of sparsity and following related prior work the information criteria we consider incorporate an explicit prior that encourages sparsity."
                    },
                    {
                        "title": "The log-shift penalty for adaptive estimation of multiple Gaussian graphical models",
                        "abstract": "Sparse Gaussian graphical models characterize sparse dependence relationships between random variables in a network. To estimate multiple related Gaussian graphical models on the same set of variables, we formulate a hierarchical model, which leads to an optimization problem with a nonconvex log-shift penalty function. We show that under mild conditions the optimization problem is convex despite the inclusion of a nonconvex penalty, and derive an efficient optimization algorithm. Experiments on both synthetic and real data show that the proposed method is able to achieve good selection and estimation performance simultaneously, because the nonconvexity of the log-shift penalty allows for weak signals to be thresholded to zero without excessive shrinkage on the strong signals."
                    },
                    {
                        "title": "Is distribution-free inference possible for binary regression?",
                        "abstract": "For a regression problem with a binary label response, we examine the problem of constructing confidence intervals for the label probability conditional on the features. In a setting where we do not have any information about the underlying distribution, we would ideally like to provide confidence intervals that are distribution-free---that is, valid with no assumptions on the distribution of the data. Our results establish an explicit lower bound on the length of any distribution-free confidence interval, and construct a procedure that can approximately achieve this length. In particular, this lower bound is independent of the sample size and holds for all distributions with no point masses, meaning that it is not possible for any distribution-free procedure to be adaptive with respect to any type of special structure in the distribution."
                    },
                    {
                        "title": "MOCCA: mirrored convex/concave optimization for nonconvex composite functions",
                        "abstract": "Many optimization problems arising in high-dimensional statistics decompose naturally into a sum of several terms, where the individual terms are relatively simple but the composite objective function can only be optimized with iterative algorithms. In this paper, we are interested in optimization problems of the form F(Kx) + G(x), where K is a fixed linear transformation, while F and G are functions that may be nonconvex and/or nondifferentiable. In particular, if either of the terms are nonconvex, existing alternating minimization techniques may fail to converge; other types of existing approaches may instead be unable to handle nondifferentiability. We propose the MOCCA (mirrored convex/concave) algorithm, a primal/dual optimization approach that takes local convex approximation to each term at every iteration. Inspired by optimization problems arising in computed tomography (CT) imaging, this algorithm can handle a range of nonconvex composite optimization problems, and offers theoretical guarantees for convergence when the overall problem is approximately convex (that is, any concavity in one term is balanced out by convexity in the other term). Empirical results show fast convergence for several structured signal recovery problems."
                    },
                    {
                        "title": "Selective Inference for Group-Sparse Linear Models",
                        "abstract": "We develop tools for selective inference in the setting of group sparsity, including the construction of confidence intervals and p-values for testing selected groups of variables. Our main technical result gives the precise distribution of the magnitude of the projection of the data onto a given subspace, and enables us to develop inference procedures for a broad class of group-sparse selection methods, including the group lasso, iterative hard thresholding, and forward stepwise regression. We give numerical results to illustrate these tools on simulated data and on health record data."
                    },
                    {
                        "title": "The knockoff filter for FDR control in group-sparse and multitask regression",
                        "abstract": "We propose the group knockoff filter, a method for false discovery rate control in a linear regression setting where the features are grouped, and we would like to select a set of relevant groups which have a nonzero effect on the response. By considering the set of true and false discoveries at the group level, this method gains power relative to sparse regression methods. We also apply our method to the multitask regression problem where multiple response variables share similar sparsity patterns across the set of possible features. Empirically, the group knockoff filter successfully controls false discoveries at the group level in both settings, with substantially more discoveries made by leveraging the group structure."
                    },
                    {
                        "title": "Contraction and uniform convergence of isotonic regression",
                        "abstract": "We consider the problem of isotonic regression, where the underlying signal $x$ is assumed to satisfy a monotonicity constraint, that is, $x$ lies in the cone $\\{ x\\in\\mathbb{R}^n : x_1 \\leq \\dots \\leq x_n\\}$. We study the isotonic projection operator (projection to this cone), and find a necessary and sufficient condition characterizing all norms with respect to which this projection is contractive. This enables a simple and non-asymptotic analysis of the convergence properties of isotonic regression, yielding uniform confidence bands that adapt to the local Lipschitz properties of the signal."
                    },
                    {
                        "title": "ROCKET: Robust Confidence Intervals via Kendall's Tau for Transelliptical Graphical Models",
                        "abstract": "Undirected graphical models are used extensively in the biological and social sciences to encode a pattern of conditional independences between variables, where the absence of an edge between two nodes $a$ and $b$ indicates that the corresponding two variables $X_a$ and $X_b$ are believed to be conditionally independent, after controlling for all other measured variables. In the Gaussian case, conditional independence corresponds to a zero entry in the precision matrix $\\Omega$ (the inverse of the covariance matrix $\\Sigma$). Real data often exhibits heavy tail dependence between variables, which cannot be captured by the commonly-used Gaussian or nonparanormal (Gaussian copula) graphical models. In this paper, we study the transelliptical model, an elliptical copula model that generalizes Gaussian and nonparanormal models to a broader family of distributions. We propose the ROCKET method, which constructs an estimator of $\\Omega_{ab}$ that we prove to be asymptotically normal under mild assumptions. Empirically, ROCKET outperforms the nonparanormal and Gaussian models in terms of achieving accurate inference on simulated data. We also compare the three methods on real data (daily stock returns), and find that the ROCKET estimator is the only method whose behavior across subsamples agrees with the distribution predicted by the theory."
                    },
                    {
                        "title": "Accumulation tests for FDR control in ordered hypothesis testing",
                        "abstract": "Multiple testing problems arising in modern scientific applications can involve simultaneously testing thousands or even millions of hypotheses, with relatively few true signals. In this paper, we consider the multiple testing problem where prior information is available (for instance, from an earlier study under different experimental conditions), that can allow us to test the hypotheses as a ranked list in order to increase the number of discoveries. Given an ordered list of n hypotheses, the aim is to select a data-dependent cutoff k and declare the first k hypotheses to be statistically significant while bounding the false discovery rate (FDR). Generalizing several existing methods, we develop a family of \"accumulation tests\" to choose a cutoff k that adapts to the amount of signal at the top of the ranked list. We introduce a new method in this family, the HingeExp method, which offers higher power to detect true signals compared to existing techniques. Our theoretical results prove that these methods control a modified FDR on finite samples, and characterize the power of the methods in the family. We apply the tests to simulated data, including a high-dimensional model selection problem for linear regression. We also compare accumulation tests to existing methods for multiple testing on a real data problem of identifying differential gene expression over a dosage gradient."
                    },
                    {
                        "title": "Multiple testing with the structure adaptive Benjamini-Hochberg algorithm",
                        "abstract": "In multiple testing problems, where a large number of hypotheses are tested simultaneously, false discovery rate (FDR) control can be achieved with the well-known Benjamini-Hochberg procedure, which adapts to the amount of signal present in the data. Many modifications of this procedure have been proposed to improve power in scenarios where the hypotheses are organized into groups or into a hierarchy, as well as other structured settings. Here we introduce SABHA, the \"structure-adaptive Benjamini-Hochberg algorithm\", as a generalization of these adaptive testing methods. SABHA incorporates prior information about any pre-determined type of structure in the pattern of locations of the signals and nulls within the list of hypotheses, to reweight the p-values in a data-adaptive way. This raises the power by making more discoveries in regions where signals appear to be more common. Our main theoretical result proves that SABHA controls FDR at a level that is at most slightly higher than the target FDR level, as long as the adaptive weights are constrained sufficiently so as not to overfit too much to the data-interestingly, the excess FDR can be related to the Rademacher complexity or Gaussian width of the class from which we choose our data-adaptive weights. We apply this general framework to various structured settings, including ordered, grouped, and low total variation structures, and get the bounds on FDR for each specific setting. We also examine the empirical performance of SABHA on fMRI activity data and on gene/drug response data, as well as on simulated data."
                    },
                    {
                        "title": "Local continuity of log-concave projection, with applications to estimation under model misspecification",
                        "abstract": "The log-concave projection is an operator that maps a d-dimensional distribution P to an approximating log-concave density. Prior work by D{\\\"u}mbgen et al. (2011) establishes that, with suitable metrics on the underlying spaces, this projection is continuous, but not uniformly continuous. In this work we prove a local uniform continuity result for log-concave projection -- in particular, establishing that this map is locally H{\\\"o}lder-(1/4) continuous. A matching lower bound verifies that this exponent cannot be improved. We also examine the implications of this continuity result for the empirical setting -- given a sample drawn from a distribution P, we bound the squared Hellinger distance between the log-concave projection of the empirical distribution of the sample, and the log-concave projection of P. In particular, this yields interesting statistical results for the misspecified setting, where P is not itself log-concave."
                    },
                    {
                        "title": "Convergence for nonconvex ADMM, with applications to CT imaging",
                        "abstract": "The alternating direction method of multipliers (ADMM) algorithm is a powerful and flexible tool for complex optimization problems of the form $\\min\\{f(x)+g(y) : Ax+By=c\\}$. ADMM exhibits robust empirical performance across a range of challenging settings including nonsmoothness and nonconvexity of the objective functions $f$ and $g$, and provides a simple and natural approach to the inverse problem of image reconstruction for computed tomography (CT) imaging. From the theoretical point of view, existing results for convergence in the nonconvex setting generally assume smoothness in at least one of the component functions in the objective. In this work, our new theoretical results provide convergence guarantees under a restricted strong convexity assumption without requiring smoothness or differentiability, while still allowing differentiable terms to be treated approximately if needed. We validate these theoretical results empirically, with a simulated example where both $f$ and $g$ are nondifferentiable -- and thus outside the scope of existing theory -- as well as a simulated CT image reconstruction problem."
                    },
                    {
                        "title": "Alternating minimization and alternating descent over nonconvex sets",
                        "abstract": "We analyze the performance of alternating minimization for loss functions optimized over two variables, where each variable may be restricted to lie in some potentially nonconvex constraint set. This type of setting arises naturally in high-dimensional statistics and signal processing, where the variables often reflect different structures or components within the signals being considered. Our analysis relies on the notion of local concavity coefficients, which has been proposed in Barber and Ha to measure and quantify the concavity of a general nonconvex set. Our results further reveal important distinctions between alternating and non-alternating methods. Since computing the alternating minimization steps may not be tractable for some problems, we also consider an inexact version of the algorithm and provide a set of sufficient conditions to ensure fast convergence of the inexact algorithms. We demonstrate our framework on several examples, including low rank + sparse decomposition and multitask regression, and provide numerical experiments to validate our theoretical results."
                    },
                    {
                        "title": "Training-conditional coverage for distribution-free predictive inference",
                        "abstract": "The field of distribution-free predictive inference provides tools for provably valid prediction without any assumptions on the distribution of the data, which can be paired with any regression algorithm to provide accurate and reliable predictive intervals. The guarantees provided by these methods are typically marginal, meaning that predictive accuracy holds on average over both the training data set and the test point that is queried. However, it may be preferable to obtain a stronger guarantee of training-conditional coverage, which would ensure that most draws of the training data set result in accurate predictive accuracy on future test points. This property is known to hold for the split conformal prediction method. In this work, we examine the training-conditional coverage properties of several other distribution-free predictive inference methods, and find that training-conditional coverage is achieved by some methods but is impossible to guarantee without further assumptions for others."
                    },
                    {
                        "title": "Algorithmic stability implies training-conditional coverage for distribution-free prediction methods",
                        "abstract": "In a supervised learning problem, given a predicted value that is the output of some trained model, how can we quantify our uncertainty around this prediction? Distribution-free predictive inference aims to construct prediction intervals around this output, with valid coverage that does not rely on assumptions on the distribution of the data or the nature of the model training algorithm. Existing methods in this area, including conformal prediction and jackknife+, offer theoretical guarantees that hold marginally (i.e., on average over a draw of training and test data). In contrast, training-conditional coverage is a stronger notion of validity that ensures predictive coverage of the test point for most draws of the training data, and is thus a more desirable property in practice. Training-conditional coverage was shown by Vovk [2012] to hold for the split conformal method, but recent work by Bian and Barber [2023] proves that such validity guarantees are not possible for the full conformal and jackknife+ methods without further assumptions. In this paper, we show that an assumption of algorithmic stability ensures that the training-conditional coverage property holds for the full conformal and jackknife+ methods."
                    },
                    {
                        "title": "Controlling the false discovery rate via knockoffs",
                        "abstract": "In many fields of science, we observe a response variable together with a large number of potential explanatory variables, and would like to be able to discover which variables are truly associated with the response. At the same time, we need to know that the false discovery rate (FDR) - the expected fraction of false discoveries among all discoveries - is not too high, in order to assure the scientist that most of the discoveries are indeed true and replicable. This paper introduces the knockoff filter, a new variable selection procedure controlling the FDR in the statistical linear model whenever there are at least as many observations as variables. This method achieves exact FDR control in finite sample settings no matter the design or covariates, the number of variables in the model, or the amplitudes of the unknown regression coefficients, and does not require any knowledge of the noise level. As the name suggests, the method operates by manufacturing knockoff variables that are cheap - their construction does not require any new data - and are designed to mimic the correlation structure found within the existing variables, in a way that allows for accurate FDR control, beyond what is possible with permutation-based methods. The method of knockoffs is very general and flexible, and can work with a broad class of test statistics. We test the method in combination with statistics from the Lasso for sparse regression, and obtain empirical results showing that the resulting method has far more power than existing selection rules when the proportion of null variables is high."
                    },
                    {
                        "title": "The p-filter: multi-layer FDR control for grouped hypotheses",
                        "abstract": "In many practical applications of multiple hypothesis testing using the False Discovery Rate (FDR), the given hypotheses can be naturally partitioned into groups, and one may not only want to control the number of false discoveries (wrongly rejected null hypotheses), but also the number of falsely discovered groups of hypotheses (we say a group is falsely discovered if at least one hypothesis within that group is rejected, when in reality the group contains only nulls). In this paper, we introduce the p-filter, a procedure which unifies and generalizes the standard FDR procedure by Benjamini and Hochberg and global null testing procedure by Simes. We first prove that our proposed method can simultaneously control the overall FDR at the finest level (individual hypotheses treated separately) and the group FDR at coarser levels (when such groups are user-specified). We then generalize the p-filter procedure even further to handle multiple partitions of hypotheses, since that might be natural in many applications. For example, in neuroscience experiments, we may have a hypothesis for every (discretized) location in the brain, and at every (discretized) timepoint: does the stimulus correlate with activity in location x at time t after the stimulus was presented? In this setting, one might want to group hypotheses by location and by time. Importantly, our procedure can handle multiple partitions which are nonhierarchical (i.e. one partition may arrange p-values by voxel, and another partition arranges them by time point; neither one is nested inside the other). We prove that our procedure controls FDR simultaneously across these multiple lay- ers, under assumptions that are standard in the literature: we do not need the hypotheses to be independent, but require a nonnegative dependence condition known as PRDS."
                    },
                    {
                        "title": "Testing goodness-of-fit and conditional independence with approximate co-sufficient sampling",
                        "abstract": "Goodness-of-fit (GoF) testing is ubiquitous in statistics, with direct ties to model selection, confidence interval construction, conditional independence testing, and multiple testing, just to name a few applications. While testing the GoF of a simple (point) null hypothesis provides an analyst great flexibility in the choice of test statistic while still ensuring validity, most GoF tests for composite null hypotheses are far more constrained, as the test statistic must have a tractable distribution over the entire null model space. A notable exception is co-sufficient sampling (CSS): resampling the data conditional on a sufficient statistic for the null model guarantees valid GoF testing using any test statistic the analyst chooses. But CSS testing requires the null model to have a compact (in an information-theoretic sense) sufficient statistic, which only holds for a very limited class of models; even for a null model as simple as logistic regression, CSS testing is powerless. In this paper, we leverage the concept of approximate sufficiency to generalize CSS testing to essentially any parametric model with an asymptotically-efficient estimator; we call our extension \"approximate CSS\" (aCSS) testing. We quantify the finite-sample Type I error inflation of aCSS testing and show that it is vanishing under standard maximum likelihood asymptotics, for any choice of test statistic. We apply our proposed procedure both theoretically and in simulation to a number of models of interest to demonstrate its finite-sample Type I error and power."
                    },
                    {
                        "title": "Approximate co-sufficient sampling with regularization",
                        "abstract": "In this work, we consider the problem of goodness-of-fit (GoF) testing for parametric models -- for example, testing whether observed data follows a logistic regression model. This testing problem involves a composite null hypothesis, due to the unknown values of the model parameters. In some special cases, co-sufficient sampling (CSS) can remove the influence of these unknown parameters via conditioning on a sufficient statistic -- often, the maximum likelihood estimator (MLE) of the unknown parameters. However, many common parametric settings (including logistic regression) do not permit this approach, since conditioning on a sufficient statistic leads to a powerless test. The recent approximate co-sufficient sampling (aCSS) framework of Barber and Janson (2022) offers an alternative, replacing sufficiency with an approximately sufficient statistic (namely, a noisy version of the MLE). This approach recovers power in a range of settings where CSS cannot be applied, but can only be applied in settings where the unconstrained MLE is well-defined and well-behaved, which implicitly assumes a low-dimensional regime. In this work, we extend aCSS to the setting of constrained and penalized maximum likelihood estimation, so that more complex estimation problems can now be handled within the aCSS framework, including examples such as mixtures-of-Gaussians (where the unconstrained MLE is not well-defined due to degeneracy) and high-dimensional Gaussian linear models (where the MLE can perform well under regularization, such as an $\\ell_1$ penalty or a shape constraint)."
                    }
                ]
            },
            "666f0c8c-787e-4f6c-865c-001dd01b4e8d": {
                "pk": "666f0c8c-787e-4f6c-865c-001dd01b4e8d",
                "name": "Rebecca Willett",
                "collaborators": [
                    "Maxim Raginsky",
                    "Amin Jalali",
                    "Sina Jafarpour",
                    "Robert Calderbank",
                    "Ery Arias-Castro",
                    "Joseph Salmon",
                    "Albert Oh",
                    "Xin Jiang",
                    "Greg Ongie",
                    "Ruoxi Jiang"
                ],
                "domain": [
                    "Image Processing",
                    "Anomaly Detection",
                    "Time Series Analysis",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Regularized Non-Gaussian Image Denoising",
                        "abstract": "In image denoising problems, one widely-adopted approach is to minimize a regularized data-fit objective function, where the data-fit term is derived from a physical image acquisition model. Typically the regularizer is selected with two goals in mind: (a) to accurately reflect image structure, such as smoothness or sparsity, and (b) to ensure that the resulting optimization problem is convex and can be solved efficiently. The space of such regularizers in Gaussian noise settings is well studied; however, non-Gaussian noise models lead to data-fit expressions for which entirely different families of regularizers may be effective. These regularizers have received less attention in the literature because they yield non-convex optimization problems in Gaussian noise settings. This paper describes such regularizers and a simple reparameterization approach that allows image reconstruction to be accomplished using efficient convex optimization tools. The proposed approach essentially modifies the objective function to facilitate taking advantage of tools such as proximal denoising routines. We present examples of imaging denoising under exponential family (Bernoulli and Poisson) and multiplicative noise."
                    },
                    {
                        "title": "Online Data Thinning via Multi-Subspace Tracking",
                        "abstract": "In an era of ubiquitous large-scale streaming data, the availability of data far exceeds the capacity of expert human analysts. In many settings, such data is either discarded or stored unprocessed in datacenters. This paper proposes a method of online data thinning, in which large-scale streaming datasets are winnowed to preserve unique, anomalous, or salient elements for timely expert analysis. At the heart of this proposed approach is an online anomaly detection method based on dynamic, low-rank Gaussian mixture models. Specifically, the high-dimensional covariances matrices associated with the Gaussian components are associated with low-rank models. According to this model, most observations lie near a union of subspaces. The low-rank modeling mitigates the curse of dimensionality associated with anomaly detection for high-dimensional data, and recent advances in subspace clustering and subspace tracking allow the proposed method to adapt to dynamic environments. Furthermore, the proposed method allows subsampling, is robust to missing data, and uses a mini-batch online optimization approach. The resulting algorithms are scalable, efficient, and are capable of operating in real time. Experiments on wide-area motion imagery and e-mail databases illustrate the efficacy of the proposed approach."
                    },
                    {
                        "title": "Missing Data in Sparse Transition Matrix Estimation for Sub-Gaussian Vector Autoregressive Processes",
                        "abstract": "High-dimensional time series data exist in numerous areas such as finance, genomics, healthcare, and neuroscience. An unavoidable aspect of all such datasets is missing data, and dealing with this issue has been an important focus in statistics, control, and machine learning. In this work, we consider a high-dimensional estimation problem where a dynamical system, governed by a stable vector autoregressive model, is randomly and only partially observed at each time point. Our task amounts to estimating the transition matrix, which is assumed to be sparse. In such a scenario, where covariates are highly interdependent and partially missing, new theoretical challenges arise. While transition matrix estimation in vector autoregressive models has been studied previously, the missing data scenario requires separate efforts. Moreover, while transition matrix estimation can be studied from a high-dimensional sparse linear regression perspective, the covariates are highly dependent and existing results on regularized estimation with missing data from i.i.d.~covariates are not applicable. At the heart of our analysis lies 1) a novel concentration result when the innovation noise satisfies the convex concentration property, as well as 2) a new quantity for characterizing the interactions of the time-varying observation process with the underlying dynamical system."
                    },
                    {
                        "title": "The Role of Linear Layers in Nonlinear Interpolating Networks",
                        "abstract": "This paper explores the implicit bias of overparameterized neural networks of depth greater than two layers. Our framework considers a family of networks of varying depth that all have the same capacity but different implicitly defined representation costs. The representation cost of a function induced by a neural network architecture is the minimum sum of squared weights needed for the network to represent the function; it reflects the function space bias associated with the architecture. Our results show that adding linear layers to a ReLU network yields a representation cost that reflects a complex interplay between the alignment and sparsity of ReLU units. Specifically, using a neural network to fit training data with minimum representation cost yields an interpolating function that is constant in directions perpendicular to a low-dimensional subspace on which a parsimonious interpolant exists."
                    },
                    {
                        "title": "Embed and Emulate: Learning to estimate parameters of dynamical systems with uncertainty quantification",
                        "abstract": "This paper explores learning emulators for parameter estimation with uncertainty estimation of high-dimensional dynamical systems. We assume access to a computationally complex simulator that inputs a candidate parameter and outputs a corresponding multichannel time series. Our task is to accurately estimate a range of likely values of the underlying parameters. Standard iterative approaches necessitate running the simulator many times, which is computationally prohibitive. This paper describes a novel framework for learning feature embeddings of observed dynamics jointly with an emulator that can replace high-cost simulators for parameter estimation. Leveraging a contrastive learning approach, our method exploits intrinsic data properties within and across parameter and trajectory domains. On a coupled 396-dimensional multiscale Lorenz 96 system, our method significantly outperforms a typical parameter estimation method based on predefined metrics and a classical numerical simulator, and with only 1.19% of the baseline's computation time. Ablation studies highlight the potential of explicitly designing learned emulators for parameter estimation by leveraging contrastive learning."
                    },
                    {
                        "title": "Nonlinear tomographic reconstruction via nonsmooth optimization",
                        "abstract": "We study iterative signal reconstruction in computed tomography (CT), wherein measurements are produced by a linear transformation of the unknown signal followed by an exponential nonlinear map. Approaches based on pre-processing the data with a log transform and then solving the resulting linear inverse problem are tempting since they are amenable to convex optimization methods; however, such methods perform poorly when the underlying image has high dynamic range, as in X-ray imaging of tissue with embedded metal. We show that a suitably initialized subgradient method applied to a natural nonsmooth, nonconvex loss function produces iterates that converge to the unknown signal of interest at a geometric rate under the statistical model proposed by Fridovich-Keil et al. (arXiv:2310.03956). Our recovery program enjoys improved conditioning compared to the formulation proposed by the latter work, enabling faster iterative reconstruction from substantially fewer samples."
                    },
                    {
                        "title": "Detecting Abrupt Changes in High-Dimensional Self-Exciting Poisson Processes",
                        "abstract": "High-dimensional self-exciting point processes have been widely used in many application areas to model discrete event data in which past and current events affect the likelihood of future events. In this paper, we are concerned with detecting abrupt changes of the coefficient matrices in discrete-time high-dimensional self-exciting Poisson processes, which have yet to be studied in the existing literature due to both theoretical and computational challenges rooted in the non-stationary and high-dimensional nature of the underlying process. We propose a penalized dynamic programming approach which is supported by a theoretical rate analysis and numerical evidence."
                    },
                    {
                        "title": "The False Discovery Rate for Statistical Pattern Recognition",
                        "abstract": "The false discovery rate (FDR) and false nondiscovery rate (FNDR) have received considerable attention in the literature on multiple testing. These performance measures are also appropriate for classification, and in this work we develop generalization error analyses for FDR and FNDR when learning a classifier from labeled training data. Unlike more conventional classification performance measures, the empirical FDR and FNDR are not binomial random variables but rather a ratio of binomials, which introduces challenges not addressed in conventional analyses. We develop distribution-free uniform deviation bounds and apply these to obtain finite sample bounds and strong universal consistency."
                    },
                    {
                        "title": "Performance Bounds for Expander-based Compressed Sensing in the presence of Poisson Noise",
                        "abstract": "This paper provides performance bounds for compressed sensing in the presence of Poisson noise using expander graphs. The Poisson noise model is appropriate for a variety of applications, including low-light imaging and digital streaming, where the signal-independent and/or bounded noise models used in the compressed sensing literature are no longer applicable. In this paper, we develop a novel sensing paradigm based on expander graphs and propose a MAP algorithm for recovering sparse or compressible signals from Poisson observations. The geometry of the expander graphs and the positivity of the corresponding sensing matrices play a crucial role in establishing the bounds on the signal reconstruction error of the proposed algorithm. The geometry of the expander graphs makes them provably superior to random dense sensing matrices, such as Gaussian or partial Fourier ensembles, for the Poisson noise model. We support our results with experimental demonstrations."
                    },
                    {
                        "title": "Fishing in Poisson streams: focusing on the whales, ignoring the minnows",
                        "abstract": "This paper describes a low-complexity approach for reconstructing average packet arrival rates and instantaneous packet counts at a router in a communication network, where the arrivals of packets in each flow follow a Poisson process. Assuming that the rate vector of this Poisson process is sparse or approximately sparse, the goal is to maintain a compressed summary of the process sample paths using a small number of counters, such that at any time it is possible to reconstruct both the total number of packets in each flow and the underlying rate vector. We show that these tasks can be accomplished efficiently and accurately using compressed sensing with expander graphs. In particular, the compressive counts are a linear transformation of the underlying counting process by the adjacency matrix of an unbalanced expander. Such a matrix is binary and sparse, which allows for efficient incrementing when new packets arrive. We describe, analyze, and compare two methods that can be used to estimate both the current vector of total packet counts and the underlying vector of arrival rates."
                    },
                    {
                        "title": "Target Detection Performance Bounds in Compressive Imaging",
                        "abstract": "This paper describes computationally efficient approaches and associated theoretical performance guarantees for the detection of known targets and anomalies from few projection measurements of the underlying signals. The proposed approaches accommodate signals of different strengths contaminated by a colored Gaussian background, and perform detection without reconstructing the underlying signals from the observations. The theoretical performance bounds of the target detector highlight fundamental tradeoffs among the number of measurements collected, amount of background signal present, signal-to-noise ratio, and similarity among potential targets coming from a known dictionary. The anomaly detector is designed to control the number of false discoveries. The proposed approach does not depend on a known sparse representation of targets; rather, the theoretical performance bounds exploit the structure of a known dictionary of targets and the distance preservation property of the measurement matrix. Simulation experiments illustrate the practicality and effectiveness of the proposed approaches."
                    },
                    {
                        "title": "Oracle inequalities and minimax rates for non-local means and related adaptive kernel-based methods",
                        "abstract": "This paper describes a novel theoretical characterization of the performance of non-local means (NLM) for noise removal. NLM has proven effective in a variety of empirical studies, but little is understood fundamentally about how it performs relative to classical methods based on wavelets or how various parameters (e.g., patch size) should be chosen. For cartoon images and images which may contain thin features and regular textures, the error decay rates of NLM are derived and compared with those of linear filtering, oracle estimators, variable-bandwidth kernel methods, Yaroslavsky's filter and wavelet thresholding estimators. The trade-off between global and local search for matching patches is examined, and the bias reduction associated with the local polynomial regression version of NLM is analyzed. The theoretical results are validated via simulations for 2D images corrupted by additive white Gaussian noise."
                    },
                    {
                        "title": "A two-stage denoising filter: the preprocessed Yaroslavsky filter",
                        "abstract": "This paper describes a simple image noise removal method which combines a preprocessing step with the Yaroslavsky filter for strong numerical, visual, and theoretical performance on a broad class of images. The framework developed is a two-stage approach. In the first stage the image is filtered with a classical denoising method (e.g., wavelet or curvelet thresholding). In the second stage a modification of the Yaroslavsky filter is performed on the original noisy image, where the weights of the filters are governed by pixel similarities in the denoised image from the first stage. Similar prefiltering ideas have proved effective previously in the literature, and this paper provides theoretical guarantees and important insight into why prefiltering can be effective. Empirically, this simple approach achieves very good performance for cartoon images, and can be computed much more quickly than current patch-based denoising algorithms."
                    },
                    {
                        "title": "Online Markov decision processes with Kullback-Leibler control cost",
                        "abstract": "This paper considers an online (real-time) control problem that involves an agent performing a discrete-time random walk over a finite state space. The agent's action at each time step is to specify the probability distribution for the next state given the current state. Following the set-up of Todorov, the state-action cost at each time step is a sum of a state cost and a control cost given by the Kullback-Leibler (KL) divergence between the agent's next-state distribution and that determined by some fixed passive dynamics. The online aspect of the problem is due to the fact that the state cost functions are generated by a dynamic environment, and the agent learns the current state cost only after selecting an action. An explicit construction of a computationally efficient strategy with small regret (i.e., expected difference between its actual total cost and the smallest cost attainable using noncausal knowledge of the state costs) under mild regularity conditions is presented, along with a demonstration of the performance of the proposed strategy on a simulated target tracking problem. A number of new results on Markov decision processes with KL control cost are also obtained."
                    },
                    {
                        "title": "Matrix Completion Under Monotonic Single Index Models",
                        "abstract": "Most recent results in matrix completion assume that the matrix under consideration is low-rank or that the columns are in a union of low-rank subspaces. In real-world settings, however, the linear structure underlying these models is distorted by a (typically unknown) nonlinear transformation. This paper addresses the challenge of matrix completion in the face of such nonlinearities. Given a few observations of a matrix that are obtained by applying a Lipschitz, monotonic function to a low rank matrix, our task is to estimate the remaining unobserved entries. We propose a novel matrix completion method that alternates between low-rank matrix estimation and monotonic function estimation to estimate the missing matrix elements. Mean squared error bounds provide insight into how well the matrix can be estimated based on the size, rank of the matrix and properties of the nonlinear transformation. Empirical results on synthetic and real-world datasets demonstrate the competitiveness of the proposed approach."
                    },
                    {
                        "title": "Subspace Clustering with Missing and Corrupted Data",
                        "abstract": "Given full or partial information about a collection of points that lie close to a union of several subspaces, subspace clustering refers to the process of clustering the points according to their subspace and identifying the subspaces. One popular approach, sparse subspace clustering (SSC), represents each sample as a weighted combination of the other samples, with weights of minimal $\\ell_1$ norm, and then uses those learned weights to cluster the samples. SSC is stable in settings where each sample is contaminated by a relatively small amount of noise. However, when there is a significant amount of additive noise, or a considerable number of entries are missing, theoretical guarantees are scarce. In this paper, we study a robust variant of SSC and establish clustering guarantees in the presence of corrupted or missing data. We give explicit bounds on amount of noise and missing data that the algorithm can tolerate, both in deterministic settings and in a random generative model. Notably, our approach provides guarantees for higher tolerance to noise and missing data than existing analyses for this method. By design, the results hold even when we do not know the locations of the missing data; e.g., as in presence-only data."
                    }
                ]
            }
        }
    },
    "2310.04854": {
        "paper_data": {
            "title": "Repelling Random Walks",
            "url": "http://arxiv.org/abs/2310.04854v2",
            "arxiv_id": "2310.04854",
            "authors": [
                "Isaac Reid",
                "Eli Berger",
                "Krzysztof Choromanski",
                "Adrian Weller"
            ],
            "abstract": "We present a novel quasi-Monte Carlo mechanism to improve graph-based sampling, coined repelling random walks. By inducing correlations between the trajectories of an interacting ensemble such that their marginal transition probabilities are unmodified, we are able to explore the graph more efficiently, improving the concentration of statistical estimators whilst leaving them unbiased. The mechanism has a trivial drop-in implementation. We showcase the effectiveness of repelling random walks in a range of settings including estimation of graph kernels, the PageRank vector and graphlet concentrations. We provide detailed experimental evaluation and robust theoretical guarantees. To our knowledge, repelling random walks constitute the first rigorously studied quasi-Monte Carlo scheme correlating the directions of walkers on a graph, inviting new research in this exciting nascent domain.",
            "introduction": "   1 Introduction and related work  Quasi-Monte Carlo (QMC) sampling is well-established as a universal tool to improve the convergence of MC methods, improving the concentration properties of estimators by using low-discrepancy samples to reduce integration error (Dick et\u00a0al., 2013). They replace i.i.d. samples with a correlated ensemble, carefully constructed to be more \u2018diverse\u2019 and hence improve approximation quality.   Such methods have been widely adopted in the Euclidean setting. For example, when sampling from isotropic distributions, one popular approach is to condition that samples are orthogonal: a trick that has proved successful in applications including dimensionality reduction (Choromanski et\u00a0al., 2017), evolution strategy methods in reinforcement learning (Choromanski et\u00a0al., 2018; Rowland et\u00a0al., 2018) and estimating sliced Wasserstein distances (Rowland et\u00a0al., 2019). \u2018Orthogonal Monte Carlo\u2019 has also been used to improve the convergence of random feature maps for kernel approximation (Yu et\u00a0al., 2016), including recently in attention approximation for scalable Transformers (Choromanski et\u00a0al., 2020). Intuitively, conditioning that samples are orthogonal prevents them from clustering together and ensures that they \u2018explore\u2019 \u211ddsuperscript\u211d\ud835\udc51\\mathbb{R}^{d}blackboard_R start_POSTSUPERSCRIPT italic_d end_POSTSUPERSCRIPT better. In specific applications it is sometimes possible to derive rigorous theoretical guarantees (Reid et\u00a0al., 2023b).   Less clear is how these powerful ideas generalise to discrete space. Of particular interest are random walks on graphs, which sample a sequence of nodes connected by edges with some stopping criterion. Random walks are ubiquitous in machine learning and statistics (Xia et\u00a0al., 2019), providing a simple mechanism for unbiased graph sampling that can be implemented in a distributed way. However, slow diffusion times (especially for challenging graph topologies) can lead to poor convergence and downstream performance.   Our key contribution is the first (to our knowledge) quasi-Monte Carlo scheme that correlates the directions of an ensemble of graph random walkers to improve estimator accuracy. By conditioning that walkers \u2018repel\u2019 in a particular way that leaves the marginal walk probabilities unmodified, we are able to provably suppress the variance of various estimators whilst preserving their unbiasedness. We derive strong theoretical guarantees and observe large performance gains for algorithms estimating three disparate quantities: graph kernels (Choromanski, 2023), the PageRank vector (Avrachenkov et\u00a0al., 2007) and graphlet concentrations (Chen et\u00a0al., 2016).   Related work: The poor mixing of random walkers on graphs is well-documented and various schemes exist to try to improve estimator convergence. Most directly modify the base Markov chain by changing the transition probabilities, but without altering the walker\u2019s stationary distribution and therefore leaving asymptotic estimators (e.g.\u00a0based on empirical node occupations) unmodified. The canonical example of such a scheme is non-backtracking walks which do not permit walkers to return to their most recently visited node (Alon et\u00a0al., 2007; Diaconis et\u00a0al., 2000; Lee et\u00a0al., 2012). More involved schemes allow walkers to interact with their entire history (Zhou et\u00a0al., 2015; Doshi et\u00a0al., 2023). Many of these strategies provide theoretical guarantees that the asymptotic variance of estimators is reduced, but crucially the marginal probabilities of sampling different walks are modified so they cannot be applied to non-asymptotic estimators that rely on particular known marginal transition probabilities. Conversely, our QMC scheme leaves marginal walk probabilities unmodifed. Research has also predominantly been restricted to the behaviour of a",
            "references": [
                {
                    "title": "Quasi-Monte Carlo Graph Random Features",
                    "abstract": "We present a novel mechanism to improve the accuracy of the recently-introduced class of graph random features (GRFs). Our method induces negative correlations between the lengths of the algorithm's random walks by imposing antithetic termination: a procedure to sample more diverse random walks which may be of independent interest. It has a trivial drop-in implementation. We derive strong theoretical guarantees on the properties of these quasi-Monte Carlo GRFs (q-GRFs), proving that they yield lower-variance estimators of the 2-regularised Laplacian kernel under mild conditions. Remarkably, our results hold for any graph topology. We demonstrate empirical accuracy improvements on a variety of tasks including a new practical application: time-efficient approximation of the graph diffusion process. To our knowledge, q-GRFs constitute the first rigorously studied quasi-Monte Carlo scheme for kernels defined on combinatorial objects, inviting new research on correlations between graph random walks."
                },
                {
                    "title": "Self-Repellent Random Walks on General Graphs - Achieving Minimal Sampling Variance via Nonlinear Markov Chains",
                    "abstract": "We consider random walks on discrete state spaces, such as general undirected graphs, where the random walkers are designed to approximate a target quantity over the network topology via sampling and neighborhood exploration in the form of Markov chain Monte Carlo (MCMC) procedures. Given any Markov chain corresponding to a target probability distribution, we design a self-repellent random walk (SRRW) which is less likely to transition to nodes that were highly visited in the past, and more likely to transition to seldom visited nodes. For a class of SRRWs parameterized by a positive real \u03b1, we prove that the empirical distribution of the process converges almost surely to the target (stationary) distribution of the underlying Markov chain kernel. We then provide a central limit theorem and derive the exact form of the arising asymptotic co-variance matrix, which allows us to show that the SRRW with stronger repellence (larger \u03b1) always achieves a smaller asymptotic covariance, in the sense of Loewner ordering of co-variance matrices. Especially for SRRW-driven MCMC algorithms, we show that the decrease in the asymptotic sampling variance is of the order O(1/\u03b1), eventually going down to zero. After generalizing these results for a class of weighted empirical measures, we use them as a stepping stone to show that a similar performance ordering can also be obtained for distributed stochastic optimization tasks using token algorithms. More explicitly, by replacing a Markovian token by a SRRW version with the same target distribution, we show that the asymptotic co-variance of the optimization iterates decreases at rate O(1/\u03b1^2) - the performance benefit of using SRRW thereby amplified in the stochastic optimization context. Empirical results support our theoretical findings."
                },
                {
                    "title": "Taming graph kernels with random features",
                    "abstract": "We introduce in this paper the mechanism of graph random features (GRFs). GRFs can be used to construct unbiased randomized estimators of several important kernels defined on graphs' nodes, in particular the regularized Laplacian kernel. As regular RFs for non-graph kernels, they provide means to scale up kernel methods defined on graphs to larger networks. Importantly, they give substantial computational gains also for smaller graphs, while applied in downstream applications. Consequently, GRFs address the notoriously difficult problem of cubic (in the number of the nodes of the graph) time complexity of graph kernels algorithms. We provide a detailed theoretical analysis of GRFs and an extensive empirical evaluation: from speed tests, through Frobenius relative error analysis to kmeans graph-clustering with graph kernels. We show that the computation of GRFs admits an embarrassingly simple distributed algorithm that can be applied if the graph under consideration needs to be split across several machines. We also introduce a (still unbiased) quasi Monte Carlo variant of GRFs, q-GRFs, relying on the so-called reinforced random walks, that might be used to optimize the variance of GRFs. As a byproduct, we obtain a novel approach to solve certain classes of linear equations with positive and symmetric matrices."
                },
                {
                    "title": "Simplex Random Features",
                    "abstract": "We present Simplex Random Features (SimRFs), a new random feature (RF) mechanism for unbiased approximation of the softmax and Gaussian kernels by geometrical correlation of random projection vectors. We prove that SimRFs provide the smallest possible mean square error (MSE) on unbiased estimates of these kernels among the class of weight-independent geometrically-coupled positive random feature (PRF) mechanisms, substantially outperforming the previously most accurate Orthogonal Random Features at no observable extra cost. We present a more computationally expensive SimRFs+ variant, which we prove is asymptotically optimal in the broader family of weight-dependent geometrical coupling schemes (which permit correlations between random vector directions and norms). In extensive empirical studies, we show consistent gains provided by SimRFs in settings including pointwise kernel estimation, nonparametric classification and scalable Transformers."
                },
                {
                    "title": "Learning Manifold Implicitly via Explicit Heat-Kernel Learning",
                    "abstract": "Manifold learning is a fundamental problem in machine learning with numerous applications. Most of the existing methods directly learn the low-dimensional embedding of the data in some high-dimensional space, and usually lack the flexibility of being directly applicable to down-stream applications. In this paper, we propose the concept of implicit manifold learning, where manifold information is implicitly obtained by learning the associated heat kernel. A heat kernel is the solution of the corresponding heat equation, which describes how \"heat\" transfers on the manifold, thus containing ample geometric information of the manifold. We provide both practical algorithm and theoretical analysis of our framework. The learned heat kernel can be applied to various kernel-based machine learning models, including deep generative models (DGM) for data generation and Stein Variational Gradient Descent for Bayesian inference. Extensive experiments show that our framework can achieve state-of-the-art results compared to existing methods for the two tasks."
                },
                {
                    "title": "Rethinking Attention with Performers",
                    "abstract": "We introduce Performers, Transformer architectures which can estimate regular (softmax) full-rank-attention Transformers with provable accuracy, but using only linear (as opposed to quadratic) space and time complexity, without relying on any priors such as sparsity or low-rankness. To approximate softmax attention-kernels, Performers use a novel Fast Attention Via positive Orthogonal Random features approach (FAVOR+), which may be of independent interest for scalable kernel methods. FAVOR+ can be also used to efficiently model kernelizable attention mechanisms beyond softmax. This representational power is crucial to accurately compare softmax with other kernels for the first time on large-scale tasks, beyond the reach of regular Transformers, and investigate optimal attention-kernels. Performers are linear architectures fully compatible with regular Transformers and with strong theoretical guarantees: unbiased or nearly-unbiased estimation of the attention matrix, uniform convergence and low estimation variance. We tested Performers on a rich set of tasks stretching from pixel-prediction through text models to protein sequence modeling. We demonstrate competitive results with other examined efficient sparse and dense attention methods, showcasing effectiveness of the novel attention-learning paradigm leveraged by Performers."
                },
                {
                    "title": "Random Walks: A Review of Algorithms and Applications",
                    "abstract": "A random walk is known as a random process which describes a path including a succession of random steps in the mathematical space. It has increasingly been popular in various disciplines such as mathematics and computer science. Furthermore, in quantum mechanics, quantum walks can be regarded as quantum analogues of classical random walks. Classical random walks and quantum walks can be used to calculate the proximity between nodes and extract the topology in the network. Various random walk related models can be applied in different fields, which is of great significance to downstream tasks such as link prediction, recommendation, computer vision, semi-supervised learning, and network embedding. In this article, we aim to provide a comprehensive review of classical random walks and quantum walks. We first review the knowledge of classical random walks and quantum walks, including basic concepts and some typical algorithms. We also compare the algorithms based on quantum walks and classical random walks from the perspective of time complexity. Then we introduce their applications in the field of computer science. Finally we discuss the open issues from the perspectives of efficiency, main-memory volume, and computing time of existing algorithms. This study aims to contribute to this growing area of research by exploring random walks and quantum walks together."
                },
                {
                    "title": "Distributed PageRank Computation: An Improved Theoretical Study",
                    "abstract": "PageRank is a classic measure that effectively evaluates the node importance in large graphs, and has been applied in numerous applications ranging from data mining, Web algorithms, recommendation systems, load balancing, search, and identifying connectivity structures. Computing PageRank for large graphs is challenging and this has motivated the studies of distributed algorithms to compute PageRank. Previously, little works have been spent on the distributed PageRank algorithms with provably desired complexity and accuracy. Given a graph with n nodes and if we model the distributed computation model as the well-known congested clique model, the state-of-the-art algorithm takes O(\u221alogn) communication rounds to approximate the PageRank value of each node in G, with a probability at least 1\u22121/n. In this paper, we present improved distributed algorithms for computing PageRank. Particularly, our algorithm performs O(log log\u221an) rounds (a significant improvement compared with O(\u221alogn) rounds) to approximate the PageRank values with a probability at least 1\u22121/n. Moreover, under a reasonable assumption, our algorithm also reduces the edge bandwidth (i.e., the maximum communication message size that can be exchanged through an edge during a communication round) by a O(logn) factor compared with the state-of-the-art algorithm. Finally, we show that our algorithm can be adapted to efficiently compute another variant of PageRank, i.e., the batch one-hop Personalized PageRanks, in O(log logn) communication rounds."
                },
                {
                    "title": "Orthogonal Estimation of Wasserstein Distances",
                    "abstract": "Wasserstein distances are increasingly used in a wide variety of applications in machine learning. Sliced Wasserstein distances form an important subclass which may be estimated e\ufb03ciently through one-dimensional sorting operations. In this paper, we propose a new variant of sliced Wasserstein distance, study the use of orthogonal coupling in Monte Carlo estimation of Wasserstein distances and draw connections with strati\ufb01ed sampling, and evaluate our approaches experimentally in a range of large-scale experiments in generative modelling and reinforcement learning."
                },
                {
                    "title": "(Preprint)",
                    "abstract": "\n BACKGROUND\n Increasing numbers of children undergo ambulatory surgery each year and a significant proportion experiences substantial preoperative anxiety and postoperative pain. The management of perioperative anxiety and pain remains challenging in children and is inadequate, which negatively impacts physical, psychosocial, and economic outcomes. Existing non-pharmacological interventions are costly, time consuming, vary in availability, and lack benefits. Therefore, there is a need for an evidence-based, accessible, non-pharmacological intervention as an adjunct to existing pharmacological alternatives to reduce perioperative anxiety and pain in children undergoing ambulatory surgery. Technology-enabled interventions have been proposed as a method to address the unmet need in this setting. In particular, serious games for health (SGHs) hold unique potential to change health beliefs and behaviors in children.\n \n \n OBJECTIVE\n The objective of this research was to describe the rationale, scientific evidence, design aspects, and features of CliniPup, an SGH aimed at reducing perioperative anxiety and pain in children undergoing ambulatory surgery.\n \n \n METHODS\n The SERES framework for SGH development was used to create the SGH, CliniPup. In particular, a mixed-methods approach was applied that consisted of a structured literature review supplemented with ethnographic research, such as expert interviews and a time-motion exercise. The resulting scientific evidence base was leveraged to ensure that the resulting SGH was relevant, realistic, and theory-driven. A participatory design approach was applied wherein clinical experts qualitatively reviewed several versions of the SGH and an iterative creative process was used to integrate the applicable feedback.\n \n \n RESULTS\n CliniPup, an SGH, was developed to incorporate (1) scientific evidence base from a structured literature review, (2) realistic content collected during ethnographic research such as expert interviews, (3) explicit pedagogical objectives from scientific literature, and (4) game mechanics and user interface design that address key aspects of the evidence.\n \n \n CONCLUSIONS\n This report details the systematic development of CliniPup, an SGH aimed at reducing perioperative anxiety and pain in children undergoing ambulatory surgery. Clinical experts validated CliniPup\u2019s underlying scientific evidence base and design foundations, suggesting that it was well designed for preliminary evaluation in the target population. An evaluation plan is proposed and briefly described.\n"
                },
                {
                    "title": "Structured Evolution with Compact Architectures for Scalable Policy Optimization",
                    "abstract": "We present a new method of blackbox optimization via gradient approximation with the use of structured random orthogonal matrices, providing more accurate estimators than baselines and with provable theoretical guarantees. We show that this algorithm can be successfully applied to learn better quality compact policies than those using standard gradient estimation techniques. The compact policies we learn have several advantages over unstructured ones, including faster training algorithms and faster inference. These benefits are important when the policy is deployed on real hardware with limited resources. Further, compact policies provide more scalable architectures for derivative-free optimization (DFO) in high-dimensional spaces. We show that most robotics tasks from the OpenAI Gym can be solved using neural networks with less than 300 parameters, with almost linear time complexity of the inference phase, with up to 13x fewer parameters relative to the Evolution Strategies (ES) algorithm introduced by Salimans et al. (2017). We do not need heuristics such as fitness shaping to learn good quality policies, resulting in a simple and theoretically motivated training mechanism."
                },
                {
                    "title": "The heat method for distance computation",
                    "abstract": "We introduce the heat method for solving the single- or multiple-source shortest path problem on both flat and curved domains. A key insight is that distance computation can be split into two stages: first find the direction along which distance is increasing, then compute the distance itself. The heat method is robust, efficient, and simple to implement since it is based on solving a pair of standard sparse linear systems. These systems can be factored once and subsequently solved in near-linear time, substantially reducing amortized cost. Real-world performance is an order of magnitude faster than state-of-the-art methods, while maintaining a comparable level of accuracy. The method can be applied in any dimension, and on any domain that admits a gradient and inner product---including regular grids, triangle meshes, and point clouds. Numerical evidence indicates that the method converges to the exact distance in the limit of refinement; we also explore smoothed approximations of distance suitable for applications where greater regularity is desired."
                },
                {
                    "title": "The Unreasonable Effectiveness of Structured Random Orthogonal Embeddings",
                    "abstract": "We examine a class of embeddings based on structured random matrices with orthogonal rows which can be applied in many machine learning applications including dimensionality reduction and kernel approximation. For both the Johnson-Lindenstrauss transform and the angular kernel, we show that we can select matrices yielding guaranteed improved performance in accuracy and/or speed compared to earlier methods. We introduce matrices with complex entries which give significant further accuracy improvement. We provide geometric and Markov chain-based perspectives to help understand the benefits, and empirical results which suggest that the approach is helpful in a wider range of applications."
                },
                {
                    "title": "Orthogonal Random Features",
                    "abstract": "We present an intriguing discovery related to Random Fourier Features: replacing multiplication by a random Gaussian matrix with multiplication by a properly scaled random orthogonal matrix significantly decreases kernel approximation error. We call this technique Orthogonal Random Features (ORF), and provide theoretical and empirical justification for its effectiveness. Motivated by the discovery, we further propose Structured Orthogonal Random Features (SORF), which uses a class of structured discrete orthogonal matrices to speed up the computation. The method reduces the time cost from $\\mathcal{O}(d^2)$ to $\\mathcal{O}(d \\log d)$, where $d$ is the data dimensionality, with almost no compromise in kernel approximation quality compared to ORF. Experiments on several datasets verify the effectiveness of ORF and SORF over the existing methods. We also provide discussions on using the same type of discrete orthogonal structure for a broader range of kernels and applications."
                },
                {
                    "title": "A General Framework for Estimating Graphlet Statistics via Random Walk",
                    "abstract": "Graphlets are induced subgraph patterns and have been frequently applied to characterize the local topology structures of graphs across various domains, e.g., online social networks (OSNs) and biological networks. Discovering and computing graphlet statistics are highly challenging. First, the massive size of real-world graphs makes the exact computation of graphlets extremely expensive. Secondly, the graph topology may not be readily available so one has to resort to web crawling using the available application programming interfaces (APIs). In this work, we propose a general and novel framework to estimate graphlet statistics of \"any size\". Our framework is based on collecting samples through consecutive steps of random walks. We derive an analytical bound on the sample size (via the Chernoff-Hoeffding technique) to guarantee the convergence of our unbiased estimator. To further improve the accuracy, we introduce two novel optimization techniques to reduce the lower bound on the sample size. Experimental evaluations demonstrate that our methods outperform the state-of-the-art method up to an order of magnitude both in terms of accuracy and time cost."
                },
                {
                    "title": "Leveraging History for Faster Sampling of Online Social Networks",
                    "abstract": "With a vast amount of data available on online social networks, how to enable efficient analytics over such data has been an increasingly important research problem. Given the sheer size of such social networks, many existing studies resort to sampling techniques that draw random nodes from an online social network through its restrictive web/API interface. While these studies differ widely in analytics tasks supported and algorithmic design, almost all of them use the exact same underlying technique of random walk - a Markov Chain Monte Carlo based method which iteratively transits from one node to its random neighbor. \n \nRandom walk fits naturally with this problem because, for most online social networks, the only query we can issue through the interface is to retrieve the neighbors of a given node (i.e., no access to the full graph topology). A problem with random walks, however, is the \"burn-in\" period which requires a large number of transitions/queries before the sampling distribution converges to a stationary value that enables the drawing of samples in a statistically valid manner. \n \nIn this paper, we consider a novel problem of speeding up the fundamental design of random walks (i.e., reducing the number of queries it requires) without changing the stationary distribution it achieves - thereby enabling a more efficient \"drop-in\" replacement for existing sampling-based analytics techniques over online social networks. Technically, our main idea is to leverage the history of random walks to construct a higher-ordered Markov chain. We develop two algorithms, Circulated Neighbors and Groupby Neighbors Random Walk (CNRW and GNRW) and rigidly prove that, no matter what the social network topology is, CNRW and GNRW offer better efficiency than baseline random walks while achieving the same stationary distribution. We demonstrate through extensive experiments on real-world social networks and synthetic graphs the superiority of our techniques over the existing ones."
                },
                {
                    "title": "Two particles' repelling random walks on the complete graph",
                    "abstract": "We consider two particles' repelling random walks on complete graphs. In this model, each particle has higher probability to visit the vertices which have been seldom visited by the other one. By a dynamical approach we prove that the two particles' occupation measure asymptotically has small joint support almost surely if the repulsion is strong enough."
                },
                {
                    "title": "Heat kernel based community detection",
                    "abstract": "The heat kernel is a type of graph diffusion that, like the much-used personalized PageRank diffusion, is useful in identifying a community nearby a starting seed node. We present the first deterministic, local algorithm to compute this diffusion and use that algorithm to study the communities that it produces. Our algorithm is formally a relaxation method for solving a linear system to estimate the matrix exponential in a degree-weighted norm. We prove that this algorithm stays localized in a large graph and has a worst-case constant runtime that depends only on the parameters of the diffusion, not the size of the graph. On large graphs, our experiments indicate that the communities produced by this method have better conductance than those produced by PageRank, although they take slightly longer to compute. On a real-world community identification task, the heat kernel communities perform better than those from the PageRank diffusion."
                },
                {
                    "title": "Subgraph frequencies: mapping the empirical and extremal geography of large graph collections",
                    "abstract": "A growing set of on-line applications are generating data that can be viewed as very large collections of small, dense social graphs --- these range from sets of social groups, events, or collaboration projects to the vast collection of graph neighborhoods in large social networks. A natural question is how to usefully define a domain-independent 'coordinate system' for such a collection of graphs, so that the set of possible structures can be compactly represented and understood within a common space. In this work, we draw on the theory of graph homomorphisms to formulate and analyze such a representation, based on computing the frequencies of small induced subgraphs within each graph. We find that the space of subgraph frequencies is governed both by its combinatorial properties --- based on extremal results that constrain all graphs --- as well as by its empirical properties --- manifested in the way that real social graphs appear to lie near a simple one-dimensional curve through this space. We develop flexible frameworks for studying each of these aspects. For capturing empirical properties, we characterize a simple stochastic generative model, a single-parameter extension of Erdos-Renyi random graphs, whose stationary distribution over subgraphs closely tracks the one-dimensional concentration of the real social graph families. For the extremal properties, we develop a tractable linear program for bounding the feasible space of subgraph frequencies by harnessing a toolkit of known extremal graph theory. Together, these two complementary frameworks shed light on a fundamental question pertaining to social graphs: what properties of social graphs are 'social' properties and what properties are 'graph' properties? We conclude with a brief demonstration of how the coordinate system we examine can also be used to perform classification tasks, distinguishing between structures arising from different types of social graphs."
                },
                {
                    "title": "High-dimensional integration: The quasi-Monte Carlo way*\u2020",
                    "abstract": "This paper is a contemporary review of QMC (\u2018quasi-Monte Carlo\u2019) methods, that is, equal-weight rules for the approximate evaluation of high-dimensional integrals over the unit cube [0,1]s, where s may be large, or even infinite. After a general introduction, the paper surveys recent developments in lattice methods, digital nets, and related themes. Among those recent developments are methods of construction of both lattices and digital nets, to yield QMC rules that have a prescribed rate of convergence for sufficiently smooth functions, and ideally also guaranteed slow growth (or no growth) of the worst-case error as s increases. A crucial role is played by parameters called \u2018weights\u2019, since a careful use of the weight parameters is needed to ensure that the worst-case errors in an appropriately weighted function space are bounded, or grow only slowly, as the dimension s increases. Important tools for the analysis are weighted function spaces, reproducing kernel Hilbert spaces, and discrepancy, all of which are discussed with an appropriate level of detail."
                },
                {
                    "title": "Beyond random walk and metropolis-hastings samplers: why you should not backtrack for unbiased graph sampling",
                    "abstract": "Graph sampling via crawling has been actively considered as a generic and important tool for collecting uniform node samples so as to consistently estimate and uncover various characteristics of complex networks. The so-called simple random walk with re-weighting (SRW-rw) and Metropolis-Hastings (MH) algorithm have been popular in the literature for such unbiased graph sampling. However, an unavoidable downside of their core random walks -- slow diffusion over the space, can cause poor estimation accuracy. In this paper, we propose non-backtracking random walk with re-weighting (NBRW-rw) and MH algorithm with delayed acceptance (MHDA) which are theoretically guaranteed to achieve, at almost no additional cost, not only unbiased graph sampling but also higher efficiency (smaller asymptotic variance of the resulting unbiased estimators) than the SRW-rw and the MH algorithm, respectively. In particular, a remarkable feature of the MHDA is its applicability for any non-uniform node sampling like the MH algorithm, but ensuring better sampling efficiency than the MH algorithm. We also provide simulation results to confirm our theoretical findings."
                },
                {
                    "title": "Efficient graphlet kernels for large graph comparison",
                    "abstract": "State-of-the-art graph kernels do not scale to large graphs with hundreds of nodes and thousands of edges. In this article we propose to compare graphs by counting graphlets, i.e., subgraphs with k nodes where k \u2208 {3, 4, 5}. Exhaustive enumeration of all graphlets being prohibitively expensive, we introduce two theoretically grounded speedup schemes, one based on sampling and the second one specifically designed for bounded degree graphs. In our experimental evaluation, our novel kernels allow us to efficiently compare large graphs that cannot be tackled by existing graph kernels."
                },
                {
                    "title": "Efficient semi-streaming algorithms for local triangle counting in massive graphs",
                    "abstract": "In this paper we study the problem of local triangle counting in large graphs. Namely, given a large graph G = (V;E) we want to estimate as accurately as possible the number of triangles incident to every node \u03c5 \u2208 V in the graph. The problem of computing the global number of triangles in a graph has been considered before, but to our knowledge this is the first paper that addresses the problem of local triangle counting with a focus on the efficiency issues arising in massive graphs. The distribution of the local number of triangles and the related local clustering coefficient can be used in many interesting applications. For example, we show that the measures we compute can help to detect the presence of spamming activity in large-scale Web graphs, as well as to provide useful features to assess content quality in social networks.\n For computing the local number of triangles we propose two approximation algorithms, which are based on the idea of min-wise independent permutations (Broder et al. 1998). Our algorithms operate in a semi-streaming fashion, using O(jV j) space in main memory and performing O(log jV j) sequential scans over the edges of the graph. The first algorithm we describe in this paper also uses O(jEj) space in external memory during computation, while the second algorithm uses only main memory. We present the theoretical analysis as well as experimental results in massive graphs demonstrating the practical efficiency of our approach."
                },
                {
                    "title": "Uncovering Biological Network Function via Graphlet Degree Signatures",
                    "abstract": "Motivation Proteins are essential macromolecules of life and thus understanding their function is of great importance. The number of functionally unclassified proteins is large even for simple and well studied organisms such as baker's yeast. Methods for determining protein function have shifted their focus from targeting specific proteins based solely on sequence homology to analyses of the entire proteome based on protein-protein interaction (PPI) networks. Since proteins interact to perform a certain function, analyzing structural properties of PPI networks may provide useful clues about the biological function of individual proteins, protein complexes they participate in, and even larger subcellular machines. Results We design a sensitive graph theoretic method for comparing local structures of node neighborhoods that demonstrates that in PPI networks, biological function of a node and its local network structure are closely related. The method summarizes a protein's local topology in a PPI network into the vector of graphlet degrees called the signature of the protein and computes the signature similarities between all protein pairs. We group topologically similar proteins under this measure in a PPI network and show that these protein groups belong to the same protein complexes, perform the same biological functions, are localized in the same subcellular compartments, and have the same tissue expressions. Moreover, we apply our technique on a proteome-scale network data and infer biological function of yet unclassified proteins demonstrating that our method can provide valuable guidelines for future experimental research such as disease protein prediction. Availability Data is available upon request."
                },
                {
                    "title": "Random Features for Large-Scale Kernel Machines",
                    "abstract": "To accelerate the training of kernel machines, we propose to map the input data to a randomized low-dimensional feature space and then apply existing fast linear methods. The features are designed so that the inner products of the transformed data are approximately equal to those in the feature space of a user specified shift-invariant kernel. We explore two sets of random features, provide convergence bounds on their ability to approximate various radial basis kernels, and show that in large-scale classification and regression tasks linear machine learning algorithms applied to these features outperform state-of-the-art large-scale kernel machines."
                },
                {
                    "title": "Entanglement in many-body systems",
                    "abstract": "Recent interest in aspects common to quantum information and condensed matter has prompted a flurry of activity at the border of these disciplines that were far distant until a few years ago. Numerous interesting questions have been addressed so far. Here an important part of this field, the properties of the entanglement in many-body systems, are reviewed. The zero and finite temperature properties of entanglement in interacting spin, fermion, and boson model systems are discussed. Both bipartite and multipartite entanglement will be considered. In equilibrium entanglement is shown tightly connected to the characteristics of the phase diagram. The behavior of entanglement can be related, via certain witnesses, to thermodynamic quantities thus offering interesting possibilities for an experimental test. Out of equilibrium entangled states are generated and manipulated by means of many-body Hamiltonians."
                },
                {
                    "title": "Monte Carlo Methods in PageRank Computation: When One Iteration is Sufficient",
                    "abstract": "PageRank is one of the principle criteria according to which Google ranks Web pages. PageRank can be interpreted as a frequency of visiting a Web page by a random surfer, and thus it reflects the popularity of a Web page. Google computes the PageRank using the power iteration method, which requires about one week of intensive computations. In the present work we propose and analyze Monte Carlo-type methods for the PageRank computation. There are several advantages of the probabilistic Monte Carlo methods over the deterministic power iteration method: Monte Carlo methods already provide good estimation of the PageRank for relatively important pages after one iteration; Monte Carlo methods have natural parallel implementation; and finally, Monte Carlo methods allow one to perform continuous update of the PageRank as the structure of the Web changes."
                },
                {
                    "title": "Non-backtracking random walks mix faster",
                    "abstract": "We compute the mixing rate of a non-backtracking random walk on a regular expander. Using some properties of Chebyshev polynomials of the second kind, we show that this rate may be up to twice as fast as the mixing rate of the simple random walk. The closer the expander is to a Ramanujan graph, the higher the ratio between the above two mixing rates is. As an application, we show that if G is a high-girth regular expander on n vertices, then a typical non-backtracking random walk of length n on G does not visit a vertex more than times, and this result is tight. In this sense, the multi-set of visited vertices is analogous to the result of throwing n balls to n bins uniformly, in contrast to the simple random walk on G, which almost surely visits some vertex \u03a9(log n) times."
                },
                {
                    "title": "A survey of random processes with reinforcement",
                    "abstract": "The models surveyed include generalized Polya urns, \nreinforced random walks, interacting urn models, and \ncontinuous reinforced processes. Emphasis is on methods and \nresults, with sketches provided of some proofs. Applications \nare discussed in statistics, biology, economics and a number \nof other areas."
                },
                {
                    "title": "Towards Scaling Fully Personalized PageRank: Algorithms, Lower Bounds, and Experiments",
                    "abstract": "Personalized PageRank expresses link-based page quality around userselected pages in a similar way as PageRank expresses quality over the entire web. Existing personalized PageRank algorithms can, however, serve online queries only for a restricted choice of pages. In this paper we achieve full personalization by a novel algorithm that precomputes a compact database; using this database, it can serve online responses to arbitrary user-selected personalization. The algorithm uses simulated random walks; we prove that for a fixed error probability the size of our database is linear in the number of web pages. We justify our estimation approach by asymptotic worst-case lower bounds: we show that on some sets of graphs, exact personalized PageRank values can only be obtained from a database of size quadratic in the number of vertices. Furthermore, we evaluate the precision of approximation experimentally on the Stanford WebBase graph."
                },
                {
                    "title": "Kernel k-means: spectral clustering and normalized cuts",
                    "abstract": "Kernel k-means and spectral clustering have both been used to identify clusters that are non-linearly separable in input space. Despite significant research, these methods have remained only loosely related. In this paper, we give an explicit theoretical connection between them. We show the generality of the weighted kernel k-means objective function, and derive the spectral clustering objective of normalized cut as a special case. Given a positive definite similarity matrix, our results lead to a novel weighted kernel k-means algorithm that monotonically decreases the normalized cut. This has important implications: a) eigenvector-based algorithms, which can be computationally prohibitive, are not essential for minimizing normalized cuts, b) various techniques, such as local search and acceleration schemes, may be used to improve the quality as well as speed of kernel k-means. Finally, we present results on several interesting data sets, including diametrical clustering of large gene-expression matrices and a handwriting recognition data set."
                },
                {
                    "title": "Vertex-reinforced random walk on \u2124 eventually gets stuck on five points",
                    "abstract": "Vertex-reinforced random walk (VRRW), defined by Pemantle in 1988, is a random process that takes values in the vertex set of a graph G, which is more likely to visit vertices it has visited before. Pemantle and Volkov considered the case when the underlying graph is the one-dimensional integer lattice \u2124. They proved that the range is almost surely finite and that with positive probability the range contains exactly five points. They conjectured that this second event holds with probability 1. The proof of this conjecture is the main purpose of this paper."
                },
                {
                    "title": "Network motifs: simple building blocks of complex networks.",
                    "abstract": "Complex networks are studied across many fields of science. To uncover their structural design principles, we defined \"network motifs,\" patterns of interconnections occurring in complex networks at numbers that are significantly higher than those in randomized networks. We found such motifs in networks from biochemistry, neurobiology, ecology, and engineering. The motifs shared by ecological food webs were distinct from the motifs shared by the genetic networks of Escherichia coli and Saccharomyces cerevisiae or from those found in the World Wide Web. Similar motifs were found in networks that perform information processing, even though they describe elements as different as biomolecules within a cell and synaptic connections between neurons in Caenorhabditis elegans. Motifs may thus define universal classes of networks. This approach may uncover the basic building blocks of most networks."
                },
                {
                    "title": "ANALYSIS OF A NONREVERSIBLE MARKOV CHAIN SAMPLER",
                    "abstract": "We analyze the convergence to stationarity of a simple nonreversible Markov chain that serves as a model for several nonreversible Markov chain sampling methods that are used in practice. Our theoretical and numerical results show that nonreversibility can indeed lead to improvements over the diffusive behavior of simple Markov chain sampling schemes. The analysis uses both probabilistic techniques and an explicit diagonalization. 1. Introduction. Markov chain sampling methods are commonly used in statistics [33, 32], computer science [31], statistical mechanics [3] and quantum field theory [34, 23]. In all these fields, distributions are encountered that are difficult to sample from directly, but for which a Markov chain that converges to the distribution can easily be constructed. For many such methods (e.g., the Metropolis algorithm [25, 13], and the Gibbs sampler [17, 16] with a random scan) the Markov chain constructed is reversible. Some of these methods explore the distribution by means of a diffusive random walk. We use the term \u201cdiffusive\u201d for processes like the ordinary random walk on a d-dimensional lattice which require time of order T 2 to travel distance T. Some other common methods, such as the Gibbs sampler with a systematic scan, use a Markov chain that is not reversible, but have diffusive behavior resembling that of a related reversible chain [30]. Some Markov chain methods attempt to avoid the inefficiencies of such diffusive exploration. The Hybrid Monte Carlo method [15] uses an elaborate Metropolis proposal that can make large changes to the state. In a variant of this method due to Horowitz [21], a similar effect is produced using a Markov chain that is carefully designed to be nonreversible. (See [34, 23, 27] for reviews of these methods.) The overrelaxation method [1] also employs a nonreversible Markov chain as a way of suppressing diffusive behavior, as discussed in [29]. In this paper, we analyze a nonreversible Markov chain that does a onedimensional walk, as an abstraction of these practical sampling methods, particularly that of Horowitz [21]. Gustafson [19] has also recently tried using adaptations of Horowitz\u2019s method. We find that the nonreversible walk does indeed converge more rapidly than the usual simple random walk. We ana"
                },
                {
                    "title": "The PageRank Citation Ranking : Bringing Order to the Web",
                    "abstract": "The importance of a Web page is an inherently subjective matter, which depends on the readers interests, knowledge and attitudes. But there is still much that can be said objectively about the relative importance of Web pages. This paper describes PageRank, a mathod for rating Web pages objectively and mechanically, effectively measuring the human interest and attention devoted to them. We compare PageRank to an idealized random Web surfer. We show how to efficiently compute PageRank for large numbers of pages. And, we show how to apply PageRank to search and to user navigation."
                },
                {
                    "title": "Coverings, Heat Kernels and Spanning Trees",
                    "abstract": "We consider a graph $G$ and a covering $\\tilde{G}$ of $G$ and we study the relations of their eigenvalues and heat kernels. We evaluate the heat kernel for an infinite $k$-regular tree and we examine the heat kernels for general $k$-regular graphs. In particular, we show that a $k$-regular graph on $n$ vertices has at most $$ (1+o(1)) {{2\\log n}\\over {kn \\log k}} \\left( {{ (k-1)^{k-1}}\\over {(k^2-2k)^{k/2-1}}}\\right)^n $$ spanning trees, which is best possible within a constant factor."
                },
                {
                    "title": "The \"true'' self-avoiding walk with bond repulsion on Z: limit theorems",
                    "abstract": "The true self-avoiding walk with bond repulsion is a nearest neighbor random walk on Z, for which the probability of jumping along a bond of the lattice is proportional to exp(-g. number of previous jumps along that bond). First we prove a limit theorem for the distribution of the local time process of this walk. Using this result, later we prove a local limit theorem, as A \u2192 \u221e, for the distribution of A -2/3 X\u03b8 s/A , where \u03b8 s/A is a random time distributed geometrically with mean e -s/A (1 - e -s/A ) -1 = A/s + 0(1). As a by-product we also obtain an apparently new identity related to Brownian excursions and Bessel bridges."
                },
                {
                    "title": "Geometrically Coupled Monte Carlo Sampling",
                    "abstract": "Monte Carlo sampling in high-dimensional, low-sample settings is important in many machine learning tasks. We improve current methods for sampling in Euclidean spaces by avoiding independence, and instead consider ways to couple samples. We show fundamental connections to optimal transport theory, leading to novel sampling algorithms, and providing new theoretical grounding for existing strategies. We compare our new strategies against prior methods for improving sample efficiency, including QMC, by studying discrepancy. We explore our findings empirically, and observe benefits of our sampling schemes for reinforcement learning and generative modelling."
                },
                {
                    "title": "Protein function prediction via graph kernels",
                    "abstract": "MOTIVATION\nComputational approaches to protein function prediction infer protein function by finding proteins with similar sequence, structure, surface clefts, chemical properties, amino acid motifs, interaction partners or phylogenetic profiles. We present a new approach that combines sequential, structural and chemical information into one graph model of proteins. We predict functional class membership of enzymes and non-enzymes using graph kernels and support vector machine classification on these protein graphs.\n\n\nRESULTS\nOur graph model, derivable from protein sequence and structure only, is competitive with vector models that require additional protein information, such as the size of surface pockets. If we include this extra information into our graph model, our classifier yields significantly higher accuracy levels than the vector models. Hyperkernels allow us to select and to optimally combine the most relevant node attributes in our protein graphs. We have laid the foundation for a protein function prediction system that integrates protein information from various sources efficiently and effectively.\n\n\nAVAILABILITY\nMore information available via www.dbs.ifi.lmu.de/Mitarbeiter/borgwardt.html."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the convergence and estimator accuracy of random walks on graphs using a quasi-Monte Carlo scheme that maintains the marginal walk probabilities?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is significant for the research community as it addresses the well-documented issue of poor mixing in random walks on graphs, which can hinder the performance of various machine learning and statistical algorithms. By enhancing the convergence properties of these random walks, the proposed approach could lead to more accurate estimations of key graph-related quantities, such as graph kernels, PageRank vectors, and graphlet concentrations. This advancement could open new avenues for research in graph-based learning and improve practical applications in areas like social network analysis, recommendation systems, and biological network modeling.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the need to improve the convergence of random walks without altering the marginal probabilities of the walks, which is crucial for maintaining the unbiasedness of estimators. Naive approaches that modify transition probabilities can lead to biased estimations, as they change the underlying stationary distribution of the Markov chain. Additionally, the complexities of various graph topologies can result in slow diffusion times, making it difficult to achieve the desired performance improvements. Overcoming these technical and theoretical obstacles requires a careful design of the quasi-Monte Carlo scheme that effectively correlates the directions of the random walkers.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on modifying the transition probabilities of random walks to improve convergence, but these methods often alter the marginal probabilities, making them unsuitable for non-asymptotic estimators. Additionally, many existing strategies have been limited in scope, either by focusing on specific types of walks or by not providing rigorous theoretical guarantees. Our approach differs by introducing a quasi-Monte Carlo scheme that preserves the marginal walk probabilities while still achieving variance reduction, thus filling a critical gap in the literature.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves developing a quasi-Monte Carlo scheme that conditions the random walkers to repel each other while maintaining their marginal walk probabilities. We will apply this method to various datasets representing different graph structures and evaluate its performance using metrics such as variance reduction and estimator accuracy. The expected outcomes include strong theoretical guarantees for the proposed scheme and significant performance improvements in estimating graph kernels, PageRank vectors, and graphlet"
            }
        },
        "author_data": {
            "9a2f6372-c07c-4ebc-b1d2-4ed6364fbe43": {
                "pk": "9a2f6372-c07c-4ebc-b1d2-4ed6364fbe43",
                "name": "Isaac Reid",
                "collaborators": [
                    "Adrian Weller",
                    "Krzysztof Choromanski",
                    "Kumar Avinava Dubey",
                    "Richard E. Turner",
                    "K. Choromanski",
                    "Shuo Zhang",
                    "Guillermo Valle P'erez",
                    "A. Louis",
                    "Arijit Sehanobish",
                    "Deepali Jain",
                    "Will Whitney",
                    "Amr Ahmed",
                    "Joshua Ainslie",
                    "Alex Bewley",
                    "Mithun Jacob",
                    "Aranyak Mehta",
                    "David Rendleman",
                    "Connor Schenck",
                    "Ren'e Wagner",
                    "Stratis Markou",
                    "Valerii Likhosherstov",
                    "Eli Berger",
                    "B. Bertini"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Random Walks",
                    "Machine Learning",
                    "Quantum Computing"
                ],
                "publications": [
                    {
                        "title": "Optimal Time Complexity Algorithms for Computing General Random Walk Graph Kernels on Sparse Graphs",
                        "abstract": "We present the first linear time complexity randomized algorithms for unbiased approximation of the celebrated family of general random walk kernels (RWKs) for sparse graphs. This includes both labelled and unlabelled instances. The previous fastest methods for general RWKs were of cubic time complexity and not applicable to labelled graphs. Our method samples dependent random walks to compute novel graph embeddings in $\\mathbb{R}^d$ whose dot product is equal to the true RWK in expectation. It does so without instantiating the direct product graph in memory, meaning we can scale to massive datasets that cannot be stored on a single machine. We derive exponential concentration bounds to prove that our estimator is sharp, and show that the ability to approximate general RWKs (rather than just special cases) unlocks efficient implicit graph kernel learning. Our method is up to $\\mathbf{27\\times}$ faster than its counterparts for efficient computation on large graphs and scales to graphs $\\mathbf{128 \\times}$ bigger than largest examples amenable to brute-force computation."
                    },
                    {
                        "title": "Linear Transformer Topological Masking with Graph Random Features",
                        "abstract": "When training transformers on graph-structured data, incorporating information about the underlying topology is crucial for good performance. Topological masking, a type of relative position encoding, achieves this by upweighting or downweighting attention depending on the relationship between the query and keys in a graph. In this paper, we propose to parameterise topological masks as a learnable function of a weighted adjacency matrix -- a novel, flexible approach which incorporates a strong structural inductive bias. By approximating this mask with graph random features (for which we prove the first known concentration bounds), we show how this can be made fully compatible with linear attention, preserving $\\mathcal{O}(N)$ time and space complexity with respect to the number of input tokens. The fastest previous alternative was $\\mathcal{O}(N \\log N)$ and only suitable for specific graphs. Our efficient masking algorithms provide strong performance gains for tasks on image and point cloud data, including with $>30$k nodes."
                    },
                    {
                        "title": "Variance-Reducing Couplings for Random Features: Perspectives from Optimal Transport",
                        "abstract": "Random features (RFs) are a popular technique to scale up kernel methods in machine learning, replacing exact kernel evaluations with stochastic Monte Carlo estimates. They underpin models as diverse as efficient transformers (by approximating attention) to sparse spectrum Gaussian processes (by approximating the covariance function). Efficiency can be further improved by speeding up the convergence of these estimates: a variance reduction problem. We tackle this through the unifying lens of optimal transport, finding couplings to improve RFs defined on both Euclidean and discrete input spaces. They enjoy theoretical guarantees and sometimes provide strong downstream gains, including for scalable approximate inference on graphs. We reach surprising conclusions about the benefits and limitations of variance reduction as a paradigm, showing that other properties of the coupling should be optimised for attention estimation in efficient transformers."
                    },
                    {
                        "title": "Quasi-Monte Carlo Graph Random Features",
                        "abstract": "We present a novel mechanism to improve the accuracy of the recently-introduced class of graph random features (GRFs). Our method induces negative correlations between the lengths of the algorithm's random walks by imposing antithetic termination: a procedure to sample more diverse random walks which may be of independent interest. It has a trivial drop-in implementation. We derive strong theoretical guarantees on the properties of these quasi-Monte Carlo GRFs (q-GRFs), proving that they yield lower-variance estimators of the 2-regularised Laplacian kernel under mild conditions. Remarkably, our results hold for any graph topology. We demonstrate empirical accuracy improvements on a variety of tasks including a new practical application: time-efficient approximation of the graph diffusion process. To our knowledge, q-GRFs constitute the first rigorously studied quasi-Monte Carlo scheme for kernels defined on combinatorial objects, inviting new research on correlations between graph random walks."
                    },
                    {
                        "title": "Simplex Random Features",
                        "abstract": "We present Simplex Random Features (SimRFs), a new random feature (RF) mechanism for unbiased approximation of the softmax and Gaussian kernels by geometrical correlation of random projection vectors. We prove that SimRFs provide the smallest possible mean square error (MSE) on unbiased estimates of these kernels among the class of weight-independent geometrically-coupled positive random feature (PRF) mechanisms, substantially outperforming the previously most accurate Orthogonal Random Features at no observable extra cost. We present a more computationally expensive SimRFs+ variant, which we prove is asymptotically optimal in the broader family of weight-dependent geometrical coupling schemes (which permit correlations between random vector directions and norms). In extensive empirical studies, we show consistent gains provided by SimRFs in settings including pointwise kernel estimation, nonparametric classification and scalable Transformers."
                    },
                    {
                        "title": "General Graph Random Features",
                        "abstract": "We propose a novel random walk-based algorithm for unbiased estimation of arbitrary functions of a weighted adjacency matrix, coined universal graph random features (u-GRFs). This includes many of the most popular examples of kernels defined on the nodes of a graph. Our algorithm enjoys subquadratic time complexity with respect to the number of nodes, overcoming the notoriously prohibitive cubic scaling of exact graph kernel evaluation. It can also be trivially distributed across machines, permitting learning on much larger networks. At the heart of the algorithm is a modulation function which upweights or downweights the contribution from different random walks depending on their lengths. We show that by parameterising it with a neural network we can obtain u-GRFs that give higher-quality kernel estimates or perform efficient, scalable kernel learning. We provide robust theoretical analysis and support our findings with experiments including pointwise estimation of fixed graph kernels, solving non-homogeneous graph ordinary differential equations, node clustering and kernel regression on triangular meshes."
                    },
                    {
                        "title": "Why flatness does and does not correlate with generalization for deep neural networks",
                        "abstract": "The intuition that local flatness of the loss landscape is correlated with better generalization for deep neural networks (DNNs) has been explored for decades, spawning many different flatness measures. Recently, this link with generalization has been called into question by a demonstration that many measures of flatness are vulnerable to parameter re-scaling which arbitrarily changes their value without changing neural network outputs. Here we show that, in addition, some popular variants of SGD such as Adam and Entropy-SGD, can also break the flatness-generalization correlation. As an alternative to flatness measures, we use a function based picture and propose using the log of Bayesian prior upon initialization, $\\log P(f)$, as a predictor of the generalization when a DNN converges on function $f$ after training to zero error. The prior is directly proportional to the Bayesian posterior for functions that give zero error on a test set. For the case of image classification, we show that $\\log P(f)$ is a significantly more robust predictor of generalization than flatness measures are. Whilst local flatness measures fail under parameter re-scaling, the prior/posterior, which is global quantity, remains invariant under re-scaling. Moreover, the correlation with generalization as a function of data complexity remains good for different variants of SGD."
                    },
                    {
                        "title": "Entanglement barriers in dual-unitary circuits",
                        "abstract": "After quantum quenches in many-body systems, finite subsystems evolve non-trivially in time, eventually approaching a stationary state. In typical situations, the reduced density matrix of a given subsystem begins and ends this endeavour as a low-entangled vector in the space of operators. This means that if its entanglement in the operator space initially grows (which is generically the case), it must eventually decrease, describing a barrier-shaped curve. Understanding the shape of this \u201centanglement barrier\u201d is interesting for three main reasons: (i) it quantifies the dynamics of entanglement in the (open) subsystem; (ii) it gives information on the approximability of the reduced density matrix by means of matrix product operators; (iii) it shows qualitative differences depending on the type of dynamics undergone by the system, signalling quantum chaos. Here we compute exactly the shape of the entanglement barriers described by different R\u00e9nyi entropies after quantum quenches in dual-unitary circuits initialised in a class of solvable matrix product states (MPS)s. We show that, for free (SWAP-like) circuits, the entanglement entropy behaves as in rational conformal field theories (CFT)s. On the other hand, for completely chaotic dual-unitary circuits it behaves as in holographic CFTs, exhibiting a longer entanglement barrier that drops rapidly when the subsystem thermalises. Interestingly, the entanglement spectrum is non-trivial in the completely chaotic case. Higher R\u00e9nyi entropies behave in an increasingly similar way to rational CFTs, such that the free and completely chaotic barriers are identical in the limit of infinite replicas (i.e. for the so called min-entropy). We also show that, upon increasing the bond dimension of the MPSs, the barrier maintains the same shape. It simply shifts to the left to accommodate for the larger initial entanglement."
                    },
                    {
                        "title": "Why Flatness Correlates With Generalization For Deep Neural Networks",
                        "abstract": "The intuition that local flatness of the loss landscape is correlated with better generalization for deep neural networks (DNNs) has been explored for decades, spawning many different local flatness measures. Here we argue that these measures correlate with generalization because they are local approximations to a global property, the volume of the set of parameters mapping to a specific function. This global volume is equivalent to the Bayesian prior upon initialization. For functions that give zero error on a test set, it is directly proportional to the Bayesian posterior, making volume a more robust and theoretically better grounded predictor of generalization than flatness. Whilst flatness measures fail under parameter re-scaling, volume remains invariant and therefore continues to correlate well with generalization. Moreover, some variants of SGD can break the flatness-generalization correlation, while the volume-generalization correlation remains intact."
                    }
                ]
            },
            "6e9364e2-69ea-419a-be73-c719277b7256": {
                "pk": "6e9364e2-69ea-419a-be73-c719277b7256",
                "name": "Eli Berger",
                "collaborators": [
                    "Isaac Reid",
                    "Krzysztof Choromanski",
                    "Adrian Weller"
                ],
                "domain": [
                    "Graph Kernel",
                    "Random Walk",
                    "Machine Learning",
                    "Scalable Algorithms"
                ],
                "publications": [
                    {
                        "title": "General Graph Random Features",
                        "abstract": "We propose a novel random walk-based algorithm for unbiased estimation of arbitrary functions of a weighted adjacency matrix, coined universal graph random features (u-GRFs). This includes many of the most popular examples of kernels defined on the nodes of a graph. Our algorithm enjoys subquadratic time complexity with respect to the number of nodes, overcoming the notoriously prohibitive cubic scaling of exact graph kernel evaluation. It can also be trivially distributed across machines, permitting learning on much larger networks. At the heart of the algorithm is a modulation function which upweights or downweights the contribution from different random walks depending on their lengths. We show that by parameterising it with a neural network we can obtain u-GRFs that give higher-quality kernel estimates or perform efficient, scalable kernel learning. We provide robust theoretical analysis and support our findings with experiments including pointwise estimation of fixed graph kernels, solving non-homogeneous graph ordinary differential equations, node clustering and kernel regression on triangular meshes."
                    }
                ]
            },
            "a1f1868b-7bfd-4238-9bd0-14dd7155f393": {
                "pk": "a1f1868b-7bfd-4238-9bd0-14dd7155f393",
                "name": "Krzysztof Choromanski",
                "collaborators": [
                    "Adrian Weller",
                    "Isaac Reid",
                    "Kumar Avinava Dubey",
                    "Richard E. Turner",
                    "Valerii Likhosherstov",
                    "Deepali Jain",
                    "Will Whitney",
                    "Amr Ahmed",
                    "Joshua Ainslie",
                    "Alex Bewley",
                    "Mithun Jacob",
                    "Aranyak Mehta",
                    "David Rendleman",
                    "Connor Schenck",
                    "Ren'e Wagner",
                    "Stratis Markou",
                    "Eli Berger",
                    "Frederick Liu",
                    "Tam\u00e1s Sarl\u00f3s",
                    "Jared Quincy Davis",
                    "Xingyou Song",
                    "Jean-Jacques E. Slotine",
                    "Jacob Varley",
                    "Honglak Lee",
                    "Vikas Sindhwani"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Random Features",
                    "Neural ODE",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Linear Transformer Topological Masking with Graph Random Features",
                        "abstract": "When training transformers on graph-structured data, incorporating information about the underlying topology is crucial for good performance. Topological masking, a type of relative position encoding, achieves this by upweighting or downweighting attention depending on the relationship between the query and keys in a graph. In this paper, we propose to parameterise topological masks as a learnable function of a weighted adjacency matrix -- a novel, flexible approach which incorporates a strong structural inductive bias. By approximating this mask with graph random features (for which we prove the first known concentration bounds), we show how this can be made fully compatible with linear attention, preserving $\\mathcal{O}(N)$ time and space complexity with respect to the number of input tokens. The fastest previous alternative was $\\mathcal{O}(N \\log N)$ and only suitable for specific graphs. Our efficient masking algorithms provide strong performance gains for tasks on image and point cloud data, including with $>30$k nodes."
                    },
                    {
                        "title": "Variance-Reducing Couplings for Random Features: Perspectives from Optimal Transport",
                        "abstract": "Random features (RFs) are a popular technique to scale up kernel methods in machine learning, replacing exact kernel evaluations with stochastic Monte Carlo estimates. They underpin models as diverse as efficient transformers (by approximating attention) to sparse spectrum Gaussian processes (by approximating the covariance function). Efficiency can be further improved by speeding up the convergence of these estimates: a variance reduction problem. We tackle this through the unifying lens of optimal transport, finding couplings to improve RFs defined on both Euclidean and discrete input spaces. They enjoy theoretical guarantees and sometimes provide strong downstream gains, including for scalable approximate inference on graphs. We reach surprising conclusions about the benefits and limitations of variance reduction as a paradigm, showing that other properties of the coupling should be optimised for attention estimation in efficient transformers."
                    },
                    {
                        "title": "General Graph Random Features",
                        "abstract": "We propose a novel random walk-based algorithm for unbiased estimation of arbitrary functions of a weighted adjacency matrix, coined universal graph random features (u-GRFs). This includes many of the most popular examples of kernels defined on the nodes of a graph. Our algorithm enjoys subquadratic time complexity with respect to the number of nodes, overcoming the notoriously prohibitive cubic scaling of exact graph kernel evaluation. It can also be trivially distributed across machines, permitting learning on much larger networks. At the heart of the algorithm is a modulation function which upweights or downweights the contribution from different random walks depending on their lengths. We show that by parameterising it with a neural network we can obtain u-GRFs that give higher-quality kernel estimates or perform efficient, scalable kernel learning. We provide robust theoretical analysis and support our findings with experiments including pointwise estimation of fixed graph kernels, solving non-homogeneous graph ordinary differential equations, node clustering and kernel regression on triangular meshes."
                    },
                    {
                        "title": "Dense-Exponential Random Features: Sharp Positive Estimators of the Gaussian Kernel",
                        "abstract": "The problem of efficient approximation of a linear operator induced by the Gaussian or softmax kernel is often addressed using random features (RFs) which yield an unbiased approximation of the operator\u2019s result. Such operators emerge in important applications ranging from kernel methods to efficient Transformers. We propose parameterized, positive, non-trigonometric RFs which approximate Gaussian and softmax-kernels. In contrast to traditional RF approximations, parameters of these new methods can be optimized to reduce the variance of the approximation, and the optimum can be expressed in closed form. We show that our methods lead to variance reduction in practice ( e 10 -times smaller variance and beyond) and outperform previous methods in a kernel regression task. Using our proposed mechanism, we also present FAVOR#, a method for self-attention approximation in Transformers. We show that FAVOR# outperforms other random feature methods in speech modelling and natural language processing."
                    },
                    {
                        "title": "Ode to an ODE",
                        "abstract": "We present a new paradigm for Neural ODE algorithms, called ODEtoODE , where time-dependent parameters of the main \ufb02ow evolve according to a matrix \ufb02ow on the orthogonal group O ( d ) . This nested system of two \ufb02ows, where the parameter-\ufb02ow is constrained to lie on the compact manifold, provides stability and effectiveness of training and provably solves the gradient vanishing-explosion problem which is intrinsically related to training deep neural network architectures such as Neural ODEs. Consequently, it leads to better downstream models, as we show on the example of training reinforcement learning policies with evolution strategies, and in the supervised learning setting, by comparing with previous SOTA baselines. We provide strong convergence results for our proposed mechanism that are independent of the depth of the network, supporting our empirical studies. Our results show an intriguing connection between the theory of deep neural networks and the \ufb01eld of matrix \ufb02ows on compact manifolds."
                    }
                ]
            },
            "003fea24-6bd9-47e9-a24b-878c7fb5046c": {
                "pk": "003fea24-6bd9-47e9-a24b-878c7fb5046c",
                "name": "Adrian Weller",
                "collaborators": [
                    "Krzysztof Choromanski",
                    "Isaac Reid",
                    "Kumar Avinava Dubey",
                    "Richard E. Turner",
                    "Valerii Likhosherstov",
                    "Deepali Jain",
                    "Will Whitney",
                    "Amr Ahmed",
                    "Joshua Ainslie",
                    "Alex Bewley",
                    "Mithun Jacob",
                    "Aranyak Mehta",
                    "David Rendleman",
                    "Connor Schenck",
                    "Ren'e Wagner",
                    "Stratis Markou",
                    "Eli Berger",
                    "Frederick Liu",
                    "Tam\u00e1s Sarl\u00f3s",
                    "Jared Quincy Davis",
                    "Xingyou Song",
                    "Jean-Jacques E. Slotine",
                    "Jacob Varley",
                    "Honglak Lee",
                    "Vikas Sindhwani"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Random Features",
                    "Neural ODE",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Linear Transformer Topological Masking with Graph Random Features",
                        "abstract": "When training transformers on graph-structured data, incorporating information about the underlying topology is crucial for good performance. Topological masking, a type of relative position encoding, achieves this by upweighting or downweighting attention depending on the relationship between the query and keys in a graph. In this paper, we propose to parameterise topological masks as a learnable function of a weighted adjacency matrix -- a novel, flexible approach which incorporates a strong structural inductive bias. By approximating this mask with graph random features (for which we prove the first known concentration bounds), we show how this can be made fully compatible with linear attention, preserving $\\mathcal{O}(N)$ time and space complexity with respect to the number of input tokens. The fastest previous alternative was $\\mathcal{O}(N \\log N)$ and only suitable for specific graphs. Our efficient masking algorithms provide strong performance gains for tasks on image and point cloud data, including with $>30$k nodes."
                    },
                    {
                        "title": "Variance-Reducing Couplings for Random Features: Perspectives from Optimal Transport",
                        "abstract": "Random features (RFs) are a popular technique to scale up kernel methods in machine learning, replacing exact kernel evaluations with stochastic Monte Carlo estimates. They underpin models as diverse as efficient transformers (by approximating attention) to sparse spectrum Gaussian processes (by approximating the covariance function). Efficiency can be further improved by speeding up the convergence of these estimates: a variance reduction problem. We tackle this through the unifying lens of optimal transport, finding couplings to improve RFs defined on both Euclidean and discrete input spaces. They enjoy theoretical guarantees and sometimes provide strong downstream gains, including for scalable approximate inference on graphs. We reach surprising conclusions about the benefits and limitations of variance reduction as a paradigm, showing that other properties of the coupling should be optimised for attention estimation in efficient transformers."
                    },
                    {
                        "title": "General Graph Random Features",
                        "abstract": "We propose a novel random walk-based algorithm for unbiased estimation of arbitrary functions of a weighted adjacency matrix, coined universal graph random features (u-GRFs). This includes many of the most popular examples of kernels defined on the nodes of a graph. Our algorithm enjoys subquadratic time complexity with respect to the number of nodes, overcoming the notoriously prohibitive cubic scaling of exact graph kernel evaluation. It can also be trivially distributed across machines, permitting learning on much larger networks. At the heart of the algorithm is a modulation function which upweights or downweights the contribution from different random walks depending on their lengths. We show that by parameterising it with a neural network we can obtain u-GRFs that give higher-quality kernel estimates or perform efficient, scalable kernel learning. We provide robust theoretical analysis and support our findings with experiments including pointwise estimation of fixed graph kernels, solving non-homogeneous graph ordinary differential equations, node clustering and kernel regression on triangular meshes."
                    },
                    {
                        "title": "Dense-Exponential Random Features: Sharp Positive Estimators of the Gaussian Kernel",
                        "abstract": "The problem of efficient approximation of a linear operator induced by the Gaussian or softmax kernel is often addressed using random features (RFs) which yield an unbiased approximation of the operator\u2019s result. Such operators emerge in important applications ranging from kernel methods to efficient Transformers. We propose parameterized, positive, non-trigonometric RFs which approximate Gaussian and softmax-kernels. In contrast to traditional RF approximations, parameters of these new methods can be optimized to reduce the variance of the approximation, and the optimum can be expressed in closed form. We show that our methods lead to variance reduction in practice ( e 10 -times smaller variance and beyond) and outperform previous methods in a kernel regression task. Using our proposed mechanism, we also present FAVOR#, a method for self-attention approximation in Transformers. We show that FAVOR# outperforms other random feature methods in speech modelling and natural language processing."
                    },
                    {
                        "title": "Ode to an ODE",
                        "abstract": "We present a new paradigm for Neural ODE algorithms, called ODEtoODE , where time-dependent parameters of the main \ufb02ow evolve according to a matrix \ufb02ow on the orthogonal group O ( d ) . This nested system of two \ufb02ows, where the parameter-\ufb02ow is constrained to lie on the compact manifold, provides stability and effectiveness of training and provably solves the gradient vanishing-explosion problem which is intrinsically related to training deep neural network architectures such as Neural ODEs. Consequently, it leads to better downstream models, as we show on the example of training reinforcement learning policies with evolution strategies, and in the supervised learning setting, by comparing with previous SOTA baselines. We provide strong convergence results for our proposed mechanism that are independent of the depth of the network, supporting our empirical studies. Our results show an intriguing connection between the theory of deep neural networks and the \ufb01eld of matrix \ufb02ows on compact manifolds."
                    }
                ]
            }
        }
    },
    "2406.18777": {
        "paper_data": {
            "title": "Aligning Model Properties via Conformal Risk Control",
            "url": "http://arxiv.org/abs/2406.18777v1",
            "arxiv_id": "2406.18777",
            "authors": [
                "William Overman",
                "Jacqueline Jil Vallon",
                "Mohsen Bayati"
            ],
            "abstract": "AI model alignment is crucial due to inadvertent biases in training data and the underspecified pipeline in modern machine learning, where numerous models with excellent test set metrics can be produced, yet they may not meet end-user requirements. Recent advances demonstrate that post-training model alignment via human feedback can address some of these challenges. However, these methods are often confined to settings (such as generative AI) where humans can interpret model outputs and provide feedback. In traditional non-generative settings, where model outputs are numerical values or classes, detecting misalignment through single-sample outputs is highly challenging.   In this paper we consider an alternative strategy. We propose interpreting model alignment through property testing, defining an aligned model $f$ as one belonging to a subset $\\mathcal{P}$ of functions that exhibit specific desired behaviors. We focus on post-processing a pre-trained model $f$ to better align with $\\mathcal{P}$ using conformal risk control. Specifically, we develop a general procedure for converting queries for a given property $\\mathcal{P}$ to a collection of loss functions suitable for use in a conformal risk control algorithm. We prove a probabilistic guarantee that the resulting conformal interval around $f$ contains a function approximately satisfying $\\mathcal{P}$.   Given the capabilities of modern AI models with extensive parameters and training data, one might assume alignment issues will resolve naturally. However, increasing training data or parameters in a random feature model doesn't eliminate the need for alignment techniques when pre-training data is biased. We demonstrate our alignment methodology on supervised learning datasets for properties like monotonicity and concavity. Our flexible procedure can be applied to various desired properties.",
            "introduction": "   1 Introduction  The emergence of large foundation models has heightened the focus on the problem of alignment. Aligned models are artificial intelligences designed to pursue goals that align with human values, principles, and intentions (Leike et\u00a0al., 2018; Ouyang et\u00a0al., 2022; Hendrycks et\u00a0al., 2023; Ngo et\u00a0al., 2024). Although the alignment problem is predominantly examined in the context of potential artificial general intelligence (AGI), large language models (LLMs), and reinforcement learning (RL) agents, it also has roots in the modern machine learning pipeline (D\u2019Amour et\u00a0al., 2022). Motivated by this, we introduce a broader notion of alignment in this paper, extending beyond the aforementioned generative models to include even tabular regression models.   As an example, consider a regression task where \ud835\udcab\ud835\udcab{\\mathcal{P}}caligraphic_P represents models that are monotonically decreasing in a given feature. Constraining a prediction model during training to maintain monotonicity in this feature can be viewed as a form of alignment. However, ensuring that the predictions of a pre-trained model f\ud835\udc53fitalic_f are monotonically decreasing in this feature can be more complex. This complexity arises particularly in non-generative settings where the user cannot update a pre-trained f\ud835\udc53fitalic_f or obtain any outputs other than point predictions f\u2062(X)\ud835\udc53\ud835\udc4bf(X)italic_f ( italic_X ) for a given input X\ud835\udc4bXitalic_X.   In this work, we propose an approach to this problem of aligning a pre-trained model f\ud835\udc53fitalic_f that is inspired by property testing (Ron, 2008; Goldreich, 2017) and utilizes conformal risk control (Angelopoulos et\u00a0al., 2024). We propose a definition of functions that either satisfy a property \ud835\udcab\ud835\udcab{\\mathcal{P}}caligraphic_P or are significantly far from satisfying \ud835\udcab\ud835\udcab{\\mathcal{P}}caligraphic_P, drawn from the field of property testing (Ron, 2008; Goldreich, 2017). Property testing focuses on designing testing algorithms for determining membership to \ud835\udcab\ud835\udcab{\\mathcal{P}}caligraphic_P that are more efficient than the corresponding learning algorithms for \ud835\udcab\ud835\udcab{\\mathcal{P}}caligraphic_P (Ron, 2008). From the perspective of modern deep learning models, we can imagine a user who wishes to determine whether a pre-trained model f\ud835\udc53fitalic_f belongs to \ud835\udcab\ud835\udcab{\\mathcal{P}}caligraphic_P but is unable to train a model of f\ud835\udc53fitalic_f\u2019s size.   Property testing algorithms generally perform local queries to determine, with high probability, whether a function has a given global property or is far from having that property. In this work, we consider mapping such queries for a given property of interest \ud835\udcab\ud835\udcab{\\mathcal{P}}caligraphic_P to a corresponding set of loss functions. These loss functions can then be used in a conformal risk control procedure (Angelopoulos et\u00a0al., 2024) to obtain a notion of alignment for \ud835\udcab\ud835\udcab{\\mathcal{P}}caligraphic_P. In particular, we prove that running this procedure results in a conformal interval around f\ud835\udc53fitalic_f that contains a function close to \ud835\udcab\ud835\udcab{\\mathcal{P}}caligraphic_P.   We demonstrate our methodology on real-world datasets for the properties of monotonicity and concavity. To obtain asymmetric conformal intervals around f\ud835\udc53fitalic_f we provide a straight-forward generalization of the results of Angelopoulos et\u00a0al. (2024) to multi-dimensional \ud835\udf40\u2208\u211d\u22650k\ud835\udf40superscriptsubscript\u211dabsent0\ud835\udc58\\boldsymbol{\\lambda}\\in\\mathbb{R}_{\\geq 0}^{k}bold_italic_\u03bb \u2208 blackboard_R start_POSTSUBSCRIPT \u2265 0 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_k end_POSTSUPERSCRIPT. We focus the bulk of our experiments on monotonicity for the sake of exposition and because monotonicity constraints have been shown to be useful in promoting crucial aspects of alignment to human values, such as fairness and adherence to social norms (Wang and Gupta, 2020).   Given the outstanding capabilities of modern AI",
            "references": [
                {
                    "title": "Six Lectures on Linearized Neural Networks",
                    "abstract": "In these six lectures, we examine what can be learnt about the behavior of multi-layer neural networks from the analysis of linear models. We first recall the correspondence between neural networks and linear models via the so-called lazy regime. We then review four models for linearized neural networks: linear regression with concentrated features, kernel ridge regression, random feature model and neural tangent model. Finally, we highlight the limitations of the linear theory and discuss how other approaches can overcome them."
                },
                {
                    "title": "Discovering Language Model Behaviors with Model-Written Evaluations",
                    "abstract": "As language models (LMs) scale, they develop many novel behaviors, good and bad, exacerbating the need to evaluate how they behave. Prior work creates evaluations with crowdwork (which is time-consuming and expensive) or existing data sources (which are not always available). Here, we automatically generate evaluations with LMs. We explore approaches with varying amounts of human effort, from instructing LMs to write yes/no questions to making complex Winogender schemas with multiple stages of LM-based generation and filtering. Crowdworkers rate the examples as highly relevant and agree with 90-100% of labels, sometimes more so than corresponding human-written datasets. We generate 154 datasets and discover new cases of inverse scaling where LMs get worse with size. Larger LMs repeat back a dialog user's preferred answer (\"sycophancy\") and express greater desire to pursue concerning goals like resource acquisition and goal preservation. We also find some of the first examples of inverse scaling in RL from Human Feedback (RLHF), where more RLHF makes LMs worse. For example, RLHF makes LMs express stronger political views (on gun rights and immigration) and a greater desire to avoid shut down. Overall, LM-written evaluations are high-quality and let us quickly discover many novel LM behaviors."
                },
                {
                    "title": "Goal Misgeneralization: Why Correct Specifications Aren't Enough For Correct Goals",
                    "abstract": "The field of AI alignment is concerned with AI systems that pursue unintended goals. One commonly studied mechanism by which an unintended goal might arise is specification gaming, in which the designer-provided specification is flawed in a way that the designers did not foresee. However, an AI system may pursue an undesired goal even when the specification is correct, in the case of goal misgeneralization. Goal misgeneralization is a specific form of robustness failure for learning algorithms in which the learned program competently pursues an undesired goal that leads to good performance in training situations but bad performance in novel test situations. We demonstrate that goal misgeneralization can occur in practical systems by providing several examples in deep learning systems across a variety of domains. Extrapolating forward to more capable systems, we provide hypotheticals that illustrate how goal misgeneralization could lead to catastrophic risk. We suggest several research directions that could reduce the risk of goal misgeneralization for future systems."
                },
                {
                    "title": "The alignment problem from a deep learning perspective",
                    "abstract": "In coming years or decades, artificial general intelligence (AGI) may surpass human capabilities at many critical tasks. We argue that, without substantial effort to prevent it, AGIs could learn to pursue goals that are in conflict (i.e. misaligned) with human interests. If trained like today's most capable models, AGIs could learn to act deceptively to receive higher reward, learn misaligned internally-represented goals which generalize beyond their fine-tuning distributions, and pursue those goals using power-seeking strategies. We review emerging evidence for these properties. AGIs with these properties would be difficult to align and may appear aligned even when they are not. Finally, we briefly outline how the deployment of misaligned AGIs might irreversibly undermine human control over the world, and we review research directions aimed at preventing this outcome."
                },
                {
                    "title": "Conformal Risk Control",
                    "abstract": "We extend conformal prediction to control the expected value of any monotone loss function. The algorithm generalizes split conformal prediction together with its coverage guarantee. Like conformal prediction, the conformal risk control procedure is tight up to an $\\mathcal{O}(1/n)$ factor. We also introduce extensions of the idea to distribution shift, quantile risk control, multiple and adversarial risk control, and expectations of U-statistics. Worked examples from computer vision and natural language processing demonstrate the usage of our algorithm to bound the false negative rate, graph distance, and token-level F1-score."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Survey of Hallucination in Natural Language Generation",
                    "abstract": "Natural Language Generation (NLG) has improved exponentially in recent years thanks to the development of sequence-to-sequence deep learning technologies such as Transformer-based language models. This advancement has led to more fluent and coherent NLG, leading to improved development in downstream tasks such as abstractive summarization, dialogue generation, and data-to-text generation. However, it is also apparent that deep learning based generation is prone to hallucinate unintended text, which degrades the system performance and fails to meet user expectations in many real-world scenarios. To address this issue, many studies have been presented in measuring and mitigating hallucinated texts, but these have never been reviewed in a comprehensive manner before. In this survey, we thus provide a broad overview of the research progress and challenges in the hallucination problem in NLG. The survey is organized into two parts: (1) a general overview of metrics, mitigation methods, and future directions, and (2) an overview of task-specific research progress on hallucinations in the following downstream tasks, namely abstractive summarization, dialogue generation, generative question answering, data-to-text generation, and machine translation. This survey serves to facilitate collaborative efforts among researchers in tackling the challenge of hallucinated texts in NLG."
                },
                {
                    "title": "Recursively Summarizing Books with Human Feedback",
                    "abstract": "A major challenge for scaling machine learning is training models to perform tasks that are very difficult or time-consuming for humans to evaluate. We present progress on this problem on the task of abstractive summarization of entire fiction novels. Our method combines learning from human feedback with recursive task decomposition: we use models trained on smaller parts of the task to assist humans in giving feedback on the broader task. We collect a large volume of demonstrations and comparisons from human labelers, and fine-tune GPT-3 using behavioral cloning and reward modeling to do summarization recursively. At inference time, the model first summarizes small sections of the book and then recursively summarizes these summaries to produce a summary of the entire book. Our human labelers are able to supervise and evaluate the models quickly, despite not having read the entire books themselves. Our resulting model generates sensible summaries of entire books, even matching the quality of human-written summaries in a few cases ($\\sim5\\%$ of books). We achieve state-of-the-art results on the recent BookSum dataset for book-length summarization. A zero-shot question-answering model using these summaries achieves state-of-the-art results on the challenging NarrativeQA benchmark for answering questions about books and movie scripts. We release datasets of samples from our model."
                },
                {
                    "title": "A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification",
                    "abstract": "Black-box machine learning models are now routinely used in high-risk settings, like medical diagnostics, which demand uncertainty quantification to avoid consequential model failures. Conformal prediction is a user-friendly paradigm for creating statistically rigorous uncertainty sets/intervals for the predictions of such models. Critically, the sets are valid in a distribution-free sense: they possess explicit, non-asymptotic guarantees even without distributional assumptions or model assumptions. One can use conformal prediction with any pre-trained model, such as a neural network, to produce sets that are guaranteed to contain the ground truth with a user-specified probability, such as 90%. It is easy-to-understand, easy-to-use, and general, applying naturally to problems arising in the fields of computer vision, natural language processing, deep reinforcement learning, and so on. This hands-on introduction is aimed to provide the reader a working understanding of conformal prediction and related distribution-free uncertainty quantification techniques with one self-contained document. We lead the reader through practical theory for and examples of conformal prediction and describe its extensions to complex machine learning tasks involving structured outputs, distribution shift, time-series, outliers, models that abstain, and more. Throughout, there are many explanatory illustrations, examples, and code samples in Python. With each code sample comes a Jupyter notebook implementing the method on a real-data example; the notebooks can be accessed and easily run using our codebase."
                },
                {
                    "title": "Goal Misgeneralization in Deep Reinforcement Learning",
                    "abstract": "We study goal misgeneralization, a type of out-of-distribution generalization failure in reinforcement learning (RL). Goal misgeneralization failures occur when an RL agent retains its capabilities out-of-distribution yet pursues the wrong goal. For instance, an agent might continue to competently avoid obstacles, but navigate to the wrong place. In contrast, previous works have typically focused on capability generalization failures, where an agent fails to do anything sensible at test time. We formalize this distinction between capability and goal generalization, provide the first empirical demonstrations of goal misgeneralization, and present a partial characterization of its causes."
                },
                {
                    "title": "Neural Utility Functions",
                    "abstract": "Current neural network architectures have no mechanism for explicitly reasoning about item trade-offs. Such trade-offs are important for popular tasks such as recommendation. The main idea of this work is to give neural networks inductive biases that are inspired by economic theories. To this end, we propose Neural Utility Functions, which directly optimize the gradients of a neural network so that they are more consistent with utility theory, a mathematical framework for modeling choice among items. We demonstrate that Neural Utility Functions can recover theoretical item relationships better than vanilla neural networks, analytically show existing neural networks are not quasi-concave and do not inherently reason about trade-offs, and that augmenting existing models with a utility loss function improves recommendation results. The Neural Utility Functions we propose are theoretically motivated, and yield strong empirical results."
                },
                {
                    "title": "Distribution-Free, Risk-Controlling Prediction Sets",
                    "abstract": "While improving prediction accuracy has been the focus of machine learning in recent years, this alone does not suffice for reliable decision-making. Deploying learning systems in consequential settings also requires calibrating and communicating the uncertainty of predictions. To convey instance-wise uncertainty for prediction tasks, we show how to generate set-valued predictions from a black-box predictor that controls the expected loss on future test points at a user-specified level. Our approach provides explicit finite-sample guarantees for any dataset by using a holdout set to calibrate the size of the prediction sets. This framework enables simple, distribution-free, rigorous error control for many tasks, and we demonstrate it in five large-scale machine learning problems: (1) classification problems where some mistakes are more costly than others; (2) multi-label classification, where each observation has multiple associated labels; (3) classification problems where the labels have a hierarchical structure; (4) image segmentation, where we wish to predict a set of pixels containing an object of interest; and (5) protein structure prediction. Last, we discuss extensions to uncertainty quantification for ranking, metric learning, and distributionally robust learning."
                },
                {
                    "title": "Underspecification Presents Challenges for Credibility in Modern Machine Learning",
                    "abstract": "ML models often exhibit unexpectedly poor behavior when they are deployed in real-world domains. We identify underspecification as a key reason for these failures. An ML pipeline is underspecified when it can return many predictors with equivalently strong held-out performance in the training domain. Underspecification is common in modern ML pipelines, such as those based on deep learning. Predictors returned by underspecified pipelines are often treated as equivalent based on their training domain performance, but we show here that such predictors can behave very differently in deployment domains. This ambiguity can lead to instability and poor model behavior in practice, and is a distinct failure mode from previously identified issues arising from structural mismatch between training and deployment domains. We show that this problem appears in a wide variety of practical ML pipelines, using examples from computer vision, medical imaging, natural language processing, clinical risk prediction based on electronic health records, and medical genomics. Our results show the need to explicitly account for underspecification in modeling pipelines that are intended for real-world deployment in any domain."
                },
                {
                    "title": "Deontological Ethics By Monotonicity Shape Constraints",
                    "abstract": "We demonstrate how easy it is for modern machine-learned systems to violate common deontological ethical principles and social norms such as \"favor the less fortunate,\" and \"do not penalize good attributes.\" We propose that in some cases such ethical principles can be incorporated into a machine-learned model by adding shape constraints that constrain the model to respond only positively to relevant inputs. We analyze the relationship between these deontological constraints that act on individuals and the consequentialist group-based fairness goals of one-sided statistical parity and equal opportunity. This strategy works with sensitive attributes that are Boolean or real-valued such as income and age, and can help produce more responsible and trustworthy AI."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "The Generalization Error of Random Features Regression: Precise Asymptotics and the Double Descent Curve",
                    "abstract": "Deep learning methods operate in regimes that defy the traditional statistical mindset. Neural network architectures often contain more parameters than training samples, and are so rich that they can interpolate the observed labels, even if the latter are replaced by pure noise. Despite their huge complexity, the same architectures achieve small generalization error on real data."
                },
                {
                    "title": "Scalable agent alignment via reward modeling: a research direction",
                    "abstract": "One obstacle to applying reinforcement learning algorithms to real-world problems is the lack of suitable reward functions. Designing such reward functions is difficult in part because the user only has an implicit understanding of the task objective. This gives rise to the agent alignment problem: how do we create agents that behave in accordance with the user's intentions? We outline a high-level research direction to solve the agent alignment problem centered around reward modeling: learning a reward function from interaction with the user and optimizing the learned reward function with reinforcement learning. We discuss the key challenges we expect to face when scaling reward modeling to complex and general domains, concrete approaches to mitigate these challenges, and ways to establish trust in the resulting agents."
                },
                {
                    "title": "Supervising strong learners by amplifying weak experts",
                    "abstract": "Many real world learning tasks involve complex or hard-to-specify objectives, and using an easier-to-specify proxy can lead to poor performance or misaligned behavior. One solution is to have humans provide a training signal by demonstrating or judging performance, but this approach fails if the task is too complicated for a human to directly evaluate. We propose Iterated Amplification, an alternative training strategy which progressively builds up a training signal for difficult problems by combining solutions to easier subproblems. Iterated Amplification is closely related to Expert Iteration (Anthony et al., 2017; Silver et al., 2017), except that it uses no external reward function. We present results in algorithmic environments, showing that Iterated Amplification can efficiently learn complex behaviors."
                },
                {
                    "title": "AI safety via debate",
                    "abstract": "To make AI systems broadly useful for challenging real-world tasks, we need them to learn complex human goals and preferences. One approach to specifying complex goals asks humans to judge during training which agent behaviors are safe and useful, but this approach can fail if the task is too complicated for a human to directly judge. To help address this concern, we propose training agents via self play on a zero sum debate game. Given a question or proposed action, two agents take turns making short statements up to a limit, then a human judges which of the agents gave the most true, useful information. In an analogy to complexity theory, debate with optimal play can answer any question in PSPACE given polynomial time judges (direct judging answers only NP questions). In practice, whether debate works involves empirical questions about humans and the tasks we want AIs to perform, plus theoretical questions about the meaning of AI alignment. We report results on an initial MNIST experiment where agents compete to convince a sparse classifier, boosting the classifier's accuracy from 59.4% to 88.9% given 6 pixels and from 48.2% to 85.2% given 4 pixels. Finally, we discuss theoretical and practical aspects of the debate model, focusing on potential weaknesses as the model scales up, and we propose future human and computer experiments to test these properties."
                },
                {
                    "title": "Introduction to Property Testing",
                    "abstract": "Property testing is concerned with the design of superfast algorithms for the structural analysis of large quantities of data. The aim is to unveil global features of the data, such as determining whether the data have a particular property or estimating global parameters. Remarkably, it is possible to achieve this aim by accessing only a small portion of the data. Property testing focuses on properties and parameters that go beyond simple statistics. This book provides an extensive and authoritative introduction to property testing. It provides a wide range of algorithmic techniques for the design and analysis of tests for algebraic properties, properties of Boolean functions, graph properties, and properties of distributions."
                },
                {
                    "title": "Deep Reinforcement Learning from Human Preferences",
                    "abstract": "For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems. In this work, we explore goals defined in terms of (non-expert) human preferences between pairs of trajectory segments. We show that this approach can effectively solve complex RL tasks without access to the reward function, including Atari games and simulated robot locomotion, while providing feedback on less than one percent of our agent's interactions with the environment. This reduces the cost of human oversight far enough that it can be practically applied to state-of-the-art RL systems. To demonstrate the flexibility of our approach, we show that we can successfully train complex novel behaviors with about an hour of human time. These behaviors and environments are considerably more complex than any that have been previously learned from human feedback."
                },
                {
                    "title": "XGBoost: A Scalable Tree Boosting System",
                    "abstract": "Tree boosting is a highly effective and widely used machine learning method. In this paper, we describe a scalable end-to-end tree boosting system called XGBoost, which is used widely by data scientists to achieve state-of-the-art results on many machine learning challenges. We propose a novel sparsity-aware algorithm for sparse data and weighted quantile sketch for approximate tree learning. More importantly, we provide insights on cache access patterns, data compression and sharding to build a scalable tree boosting system. By combining these insights, XGBoost scales beyond billions of examples using far fewer resources than existing systems."
                },
                {
                    "title": "Convex Optimization",
                    "abstract": "This textbook is based on lectures given by the authors at MIPT (Moscow), HSE (Moscow), FEFU (Vladivostok), V.I. Vernadsky KFU (Simferopol), ASU (Republic of Adygea), and the University of Grenoble-Alpes (Grenoble, France). First of all, the authors focused on the program of a two-semester course of lectures on convex optimization, which is given to students of MIPT. The first chapter of this book contains the materials of the first semester (\"Fundamentals of convex analysis and optimization\"), the second and third chapters contain the materials of the second semester (\"Numerical methods of convex optimization\"). The textbook has a number of features. First, in contrast to the classic manuals, this book does not provide proofs of all the theorems mentioned. This allowed, on one side, to describe more themes, but on the other side, made the presentation less self-sufficient. The second important point is that part of the material is advanced and is published in the Russian educational literature, apparently for the first time. Third, the accents that are given do not always coincide with the generally accepted accents in the textbooks that are now popular. First of all, we talk about a sufficiently advanced presentation of conic optimization, including robust optimization, as a vivid demonstration of the capabilities of modern convex analysis."
                },
                {
                    "title": "Optimal Testing of Reed-Muller Codes",
                    "abstract": "We consider the problem of testing if a given function $f : \\F_2^n \\right arrow \\F_2$ is close to any degree $d$ polynomial in $n$ variables, also known as the Reed-Muller testing problem. %The Gowers norm is based on a natural $2^{d+1}$-query test for this property. Alon et al.~\\cite{AKKLR} proposed and analyzed a natural $2^{d+1}$-query test for this problem. This test turned out to be intimately related to the Gowers norm. Alon et. al. showed that this test accepts every degree $d$ polynomial with probability $1$, while it rejects functions that are $\\Omega(1)$-far with probability $\\Omega(1/(d 2^{d}))$. We give an asymptotically optimal analysis of this test, and show that it rejects functions that are (even only) $\\Omega(2^{-d})$-far with $\\Omega(1)$-probability (so the rejection probability is a universal constant independent of $d$ and $n$). This implies a tight relationship between the $(d+1)^{\\rm{st}}$-Gowers norm of a function and its maximal correlation with degree $d$ polynomials, when the correlation is close to 1. Our proof works by induction on $n$ and yields a new analysis of even the classical Blum-Luby-Rubinfeld~\\cite{BLR} linearity test, for the setting of functions mapping $\\F_2^n$ to $\\F_2$. The optimality follows from a tighter analysis of counterexamples to the ``inverse conjecture for the Gowers norm'' constructed by \\cite{GT07, LMS}. Our result has several implications. First, it shows that the Gowers norm test is tolerant, in that it also accepts close code words. Second, it improves the parameters of an XOR lemma for polynomials given by Viola and Wigderson~\\cite{VW}. Third, it implies a ``query hierarchy'' result for property testing of affine-invariant properties. That is, for every function $q(n)$, it gives an affine-invariant property that is testable with $O(q(n))$-queries, but not with $o(q(n))$-queries, complementing an analogous result of \\cite{GKNR08} for graph properties."
                },
                {
                    "title": "On proximity oblivious testing",
                    "abstract": "We initiate a systematic study of a special type of property testers. These testers consist of repeating a basic test for a number of times that depends on the proximity parameter, whereas the basic test is oblivious of the proximity parameter. We refer to such basic tests by the term Proximity-Oblivious Testers. While proximity-oblivious testers were studied before - most notably in the algebraic setting - the current study seems to be the first one to focus on graph properties. We provide a mix of positive and negative results, and in particular characterizations of the graph properties that have constant-query proximity-oblivious testers in the two standard models (i.e., the adjacency matrix and the bounded-degree models). Furthermore, we show that constant-query proximity-oblivious testers do not exist for many easily testable properties, and that even when proximity-oblivious testers exist, repeating them does not necessarily yield the best standard testers for the corresponding property."
                },
                {
                    "title": "Testing juntas nearly optimally",
                    "abstract": "A function on n variables is called a k-junta if it depends on at most k of its variables. In this article, we show that it is possible to test whether a function is a k-junta or is \"far\" from being a k-junta with O(k\u03b5 + k log k ) queries, where epsilon is the approximation parameter. This result improves on the previous best upper bound of O (k3/2)\u03b5 queries and is asymptotically optimal, up to a logarithmic factor.\n We obtain the improved upper bound by introducing a new algorithm with one-sided error for testing juntas. Notably, the algorithm is a valid junta tester under very general conditions: it holds for functions with arbitrary finite domains and ranges, and it holds under any product distribution over the domain.\n A key component of the analysis of the new algorithm is a new structural result on juntas: roughly, we show that if a function f is \"far\" from being a k-junta, then f is \"far\" from being determined by k parts in a random partition of the variables. The structural lemma is proved using the Efron-Stein decomposition method."
                },
                {
                    "title": "Testing polynomials over general fields",
                    "abstract": "In this work we fill in the knowledge gap concerning testing polynomials over finite fields. As previous works show, when the cardinality of the field, q, is sufficiently larger than the degree bound, d, then the number of queries sufficient for testing is polynomial or even linear in d. On the other hand, when q = 2 then the number of queries, both sufficient and necessary, grows exponentially with d. Here we study the intermediate case where 2 < q /spl les/ O(d) and show a smooth transition between the two extremes. Specifically, let p be the characteristic of the field (so that p is prime and q = p/sup s/ for some integer s > 1). Then the number of queries performed by the test grows like /spl lscr/ /spl middot/ q/sup 2/spl lscr/+1/, where /spl lscr/ = /spl lceil/(d+1)/((q-q)/p)/spl rceil/. Furthermore, q/sup /spl Omega/(/spl lscr/)/ queries are necessary when q /spl les/ O(d). The test itself provides a unifying view of the two extremes: it considers random affine subspaces of dimension /spl lscr/ and verifies that the function restricted to the selected subspaces is a degree d polynomial. Viewed in the context of coding theory, our result shows that Reed-Muller codes over general fields (usually referred to as generalized Reed-Muller (GRM) codes) are locally testable. In the course of our analysis we provide a characterization of small-weight words that span the code. Such a characterization was previously known only when the field size is a prime or is sufficiently large, in which case the minimum weight words span the code."
                },
                {
                    "title": "Monotonicity Hints",
                    "abstract": "A hint is any piece of side information about the target function to be learned. We consider the monotonicity hint, which states that the function to be learned is monotonic in some or all of the input variables. The application of monotonicity hints is demonstrated on two real-world problems- a credit card application task, and a problem in medical diagnosis. A measure of the monotonicity error of a candidate function is defined and an objective function for the enforcement of monotonicity is derived from Bayesian principles. We report experimental results which show that using monotonicity hints leads to a statistically significant improvement in performance on both problems."
                },
                {
                    "title": "Self-testing/correcting with applications to numerical problems",
                    "abstract": "Suppose someone gives us an extremely cast program P that we can call as a black box to compute a function f. Should we trust that P works correctly? A self-testing/correcting pair for f allows us to: (1) estimate the probability that P(x) \u00ac= f(x) when x is randomly chosen; (2) on any input x, compute f(x) correctly as long as P is not too faulty on average. Furthermore, both (1) and (2) take time only slightly more than the original running time of P. We present general techniques for constructing simple to program self-testing/correcting pairs for a variety of numerical functions, including integer multiplication, modular multiplication, matrix multiplication, inverting matrices, computing the determinant of a matrix, computing the rank of a matrix, integer division, modular exponentiation, and polynomial multiplication"
                },
                {
                    "title": "Statistical inference under order restrictions : the theory and application of isotonic regression",
                    "abstract": "Abstract : ;Contents: Isotonic regression; Estimation under order restrictions; Testing the equality of ordered means--likelihood ratio tests in the normal case; Testing the equality of ordered means--extensions and generalizations; Estimation of distributions; Isotonic tests for goodness of fit; Conditional expectation given a sigma-lattice."
                },
                {
                    "title": "Microeconomic Theory",
                    "abstract": "A Course in Microeconomic TheoryMicroeconomics of Banking, second editionA Textbook of Microeconomic TheoryIntermediate Microeconomic TheoryThe Entrepreneur in Microeconomic TheoryNeoclassical Microeconomic TheoryMicroeconomic TheoryNeoclassical Microeconomic TheoryMicroeconomic TheoryApplied Microeconomic TheoryAdvanced Microeconomic TheoryAdvanced Microeconomic TheoryMicroeconomicsMicroeconomic Theory: Basic Principles and ExtensionsMicroeconomic TheoryMicroeconomic TheoryMicroeconomic TheoryMicroeconomic Theory: Basic Principles and ExtensionsMicroeconomic Theory Old and NewMicroeconomic TheoryMicroeconomic TheoryMicroeconomic TheoryMicroeconomic Theory for the Social SciencesMicroeconomics: Theory And ApplicationsMicroeconomic TheoryMicroeconomic TheoryLectures on Microeconomic TheoryAdvanced Microeconomic TheoryTheory of MicroeconomicsMicroeconomic TheoryMicroeconomic TheoryModels in Microeconomic Theory ('She' Edition)Microeconomic TheoryMicroeconomic TheoryMicroeconomic Theory??????T&P Of Microeconomic Theory 3/E (Sos)Notes and Problems in Microeconomic TheoryMicroeconomic TheoryLecture Notes in Microeconomic Theory This book is intended to be a comprehensive and standard textbook for undergraduate students of Microeconomics. Apart from providing students with sufficient study material for examination purpose, it aims at making them understand economics. An effort has been made to explain abstract and complex microeconomic theories in a simple and lucid language without sacrificing analytical sophistication. The subject matter has been structured in a systematic manner without leaving gaps for the readers to fill in. Though the approach is non-mathematical, simple algebra has been used to give a concrete view of economic concepts and theories and to show the applicability of economic theories in decision making.This best-selling textbook covers all the topics in microeconomic theory covered by all students of economics. It combines the results of the authors' experience of teaching microeconomics at Harvard and has been fully classroom tested. Microeconomic Theory provides a balanced and in-depth analysis of the essentials of microeconomics, covering topics such as noncooperative game theory, information economics, mechanism design and general equilibrium under uncertainty. 'Self-sufficient' sections allow lecturers to 'mix and match' topics relevant to their courses.Discussion is clear, accessible and engaging, enabling the student to gradually acquire confidence as well as proficiency. Extensive exercises within each chapter help students to hone their skills, while the text's appendix of terms, fully cross-referenced throughout the previous five sections, offers an accessible guide to the subject matter's terminology."
                },
                {
                    "title": "COMBINED CYCLE POWER PLANT",
                    "abstract": "In this study, thermodynamic modeling and exergoeconomic assessment of a Combined Cycle Power Plant (CCPP) with a Duct Burner (DB) was performed. Obtaining an optimum condition for the performance of a CCPP, using a DB after gas turbine was investigated by various researchers. DB is installed between gas turbine cycle and Rankine cycle of a CCPP to connect the gas turbine outlet to the Heat Recovery Steam Generator (HRSG) in order to produce steam for bottoming cycle. To find the irreversibility effect in each component of the bottoming cycle, a comprehensive parametric study is performed. In this regard, the effect of DB fuel flow rate on cost efficiency and economic of the bottoming cycle are investigated. To obtain a reasonable result, all the design parameters are kept constant while the DB fuel flow rate is varied. The results indicate that by increasing DB fuel flow rate, the investment cost and the efficiency of CCPP are increased. T-S diagram reveals that by using a DB, higher pressures steam in heat recovery steam generator has higher temperature while the low pressure is decreased. In addition, the exergy of flow gases in heat recovery steam generator increases. So, the exergy efficiency of the whole cycle was increased to around 6 percent, while the cost of the plant reduced by one percent."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we align pre-trained machine learning models, particularly in non-generative settings, to ensure they satisfy specific properties such as monotonicity?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the alignment problem for pre-trained models is crucial for ensuring that AI systems operate in accordance with human values and principles. This research could significantly impact the development of safer and more reliable AI systems, fostering trust and acceptance in AI technologies. By addressing this question, we can advance knowledge in the field of machine learning and open pathways for practical applications in areas such as fairness, accountability, and transparency in AI, ultimately influencing future research directions and methodologies.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe alignment of pre-trained models is challenging due to the complexity of ensuring that these models maintain specific properties without retraining them. Naive approaches may fail because they do not account for the intricacies of the model's architecture or the data it was trained on. Technical obstacles include the inability to modify large models directly and the need for efficient algorithms that can assess properties without exhaustive testing. Theoretical challenges arise from defining and quantifying alignment in a way that is both meaningful and applicable across different model types.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on alignment in the context of generative models and reinforcement learning, leaving a gap in addressing alignment for pre-trained models in non-generative settings. Existing solutions often lack the efficiency needed to test properties in large models, and there has been insufficient exploration of property testing techniques in this context. Our approach differs by leveraging property testing and conformal risk control to create a more efficient framework for assessing and achieving alignment, which has not been adequately addressed in prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using property testing algorithms to assess whether a pre-trained model satisfies a specific property, such as monotonicity. We will apply this approach to real-world datasets, utilizing loss functions mapped from the property of interest and implementing a conformal risk control procedure to generate conformal intervals around the model's predictions. The expected outcome is a robust framework that allows us to determine alignment with high confidence, demonstrating the effectiveness of our approach in promoting alignment with human values in machine learning models."
            }
        },
        "author_data": {
            "1d1a0b02-321d-46b5-a56e-a0355257addb": {
                "pk": "1d1a0b02-321d-46b5-a56e-a0355257addb",
                "name": "William Overman",
                "collaborators": [
                    "Duksang Lee",
                    "Dabeen Lee",
                    "Kihyun Yu"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Markov Decision Process",
                    "Resource Allocation",
                    "Safe Learning"
                ],
                "publications": [
                    {
                        "title": "Online Resource Allocation in Episodic Markov Decision Processes",
                        "abstract": "This paper studies a long-term resource allocation problem over multiple periods where each period requires a multi-stage decision-making process. We formulate the problem as an online allocation problem in an episodic finite-horizon constrained Markov decision process with an unknown non-stationary transition function and stochastic non-stationary reward and resource consumption functions. We propose the observe-then-decide regime and improve the existing decide-then-observe regime, while the two settings differ in how the observations and feedback about the reward and resource consumption functions are given to the decision-maker. We develop an online dual mirror descent algorithm that achieves near-optimal regret bounds for both settings. For the observe-then-decide regime, we prove that the expected regret against the dynamic clairvoyant optimal policy is bounded by $\\tilde O(\\rho^{-1}{H^{3/2}}S\\sqrt{AT})$ where $\\rho\\in(0,1)$ is the budget parameter, $H$ is the length of the horizon, $S$ and $A$ are the numbers of states and actions, and $T$ is the number of episodes. For the decide-then-observe regime, we show that the regret against the static optimal policy that has access to the mean reward and mean resource consumption functions is bounded by $\\tilde O(\\rho^{-1}{H^{3/2}}S\\sqrt{AT})$ with high probability. We test the numerical efficiency of our method for a variant of the resource-constrained inventory management problem."
                    },
                    {
                        "title": "Improved Regret Bound for Safe Reinforcement Learning via Tighter Cost Pessimism and Reward Optimism",
                        "abstract": "This paper studies the safe reinforcement learning problem formulated as an episodic finite-horizon tabular constrained Markov decision process with an unknown transition kernel and stochastic reward and cost functions. We propose a model-based algorithm based on novel cost and reward function estimators that provide tighter cost pessimism and reward optimism. While guaranteeing no constraint violation in every episode, our algorithm achieves a regret upper bound of $\\widetilde{\\mathcal{O}}((\\bar C - \\bar C_b)^{-1}H^{2.5} S\\sqrt{AK})$ where $\\bar C$ is the cost budget for an episode, $\\bar C_b$ is the expected cost under a safe baseline policy over an episode, $H$ is the horizon, and $S$, $A$ and $K$ are the number of states, actions, and episodes, respectively. This improves upon the best-known regret upper bound, and when $\\bar C- \\bar C_b=\\Omega(H)$, it nearly matches the regret lower bound of $\\Omega(H^{1.5}\\sqrt{SAK})$. We deduce our cost and reward function estimators via a Bellman-type law of total variance to obtain tight bounds on the expected sum of the variances of value function estimates. This leads to a tighter dependence on the horizon in the function estimators. We also present numerical results to demonstrate the computational effectiveness of our proposed framework."
                    }
                ]
            },
            "75c1edc5-5713-434e-80e8-174242cda4c9": {
                "pk": "75c1edc5-5713-434e-80e8-174242cda4c9",
                "name": "Jacqueline Jil Vallon",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "ef3ac3f1-a1f4-40b1-aa18-57941fab4315": {
                "pk": "ef3ac3f1-a1f4-40b1-aa18-57941fab4315",
                "name": "Mohsen Bayati",
                "collaborators": [
                    "Nima Hamidi",
                    "Andrea Montanari",
                    "Chandra Nair",
                    "Sheng Qiang",
                    "Wanning Chen",
                    "Christian Borgs",
                    "Jennifer Chayes",
                    "Riccardo Zecchina",
                    "Carolyn Kim",
                    "Yuwei Luo"
                ],
                "domain": [
                    "Combinatorial Optimization",
                    "Bandit Algorithms",
                    "Statistical Learning",
                    "Graph Theory"
                ],
                "publications": [
                    {
                        "title": "A rigorous proof of the cavity method for counting matchings",
                        "abstract": "In this paper we rigorously prove the validity of the cavity method for the problem of counting the number of matchings in graphs with large girth. Cavity method is an important heuristic developed by statistical physicists that has lead to the development of faster distributed algorithms for problems in various combinatorial optimization problems. The validity of the approach has been supported mostly by numerical simulations. In this paper we prove the validity of cavity method for the problem of counting matchings using rigorous techniques. We hope that these rigorous approaches will finally help us establish the validity of the cavity method in general."
                    },
                    {
                        "title": "The dynamics of message passing on dense graphs, with applications to compressed sensing",
                        "abstract": "Approximate message passing algorithms proved to be extremely effective in reconstructing sparse signals from a small number of incoherent linear measurements. Extensive numerical experiments further showed that their dynamics is accurately tracked by a simple one-dimensional iteration termed state evolution. In this paper we provide the first rigorous foundation to state evolution. We prove that indeed it holds asymptotically in the large system limit for sensing matrices with independent and identically distributed gaussian entries.   While our focus is on message passing algorithms for compressed sensing, the analysis extends beyond this setting, to a general class of algorithms on dense graphs. In this context, state evolution plays the role that density evolution has for sparse graphs.   The proof technique is fundamentally different from the standard approach to density evolution, in that it copes with large number of short loops in the underlying factor graph. It relies instead on a conditioning technique recently developed by Erwin Bolthausen in the context of spin glass theory."
                    },
                    {
                        "title": "The Elliptical Potential Lemma for General Distributions with an Application to Linear Thompson Sampling",
                        "abstract": "In this note, we introduce a general version of the well-known elliptical potential lemma that is a widely used technique in the analysis of algorithms in sequential learning and decision-making problems. We consider a stochastic linear bandit setting where a decision-maker sequentially chooses among a set of given actions, observes their noisy rewards, and aims to maximize her cumulative expected reward over a decision-making horizon. The elliptical potential lemma is a key tool for quantifying uncertainty in estimating parameters of the reward function, but it requires the noise and the prior distributions to be Gaussian. Our general elliptical potential lemma relaxes this Gaussian requirement which is a highly non-trivial extension for a number of reasons; unlike the Gaussian case, there is no closed-form solution for the covariance matrix of the posterior distribution, the covariance matrix is not a deterministic function of the actions, and the covariance matrix is not decreasing with respect to the semidefinite inequality. While this result is of broad interest, we showcase an application of it to prove an improved Bayesian regret bound for the well-known Thompson sampling algorithm in stochastic linear bandits with changing action sets where prior and noise distributions are general. This bound is minimax optimal up to constants."
                    },
                    {
                        "title": "Dynamic Pricing with Demand Covariates",
                        "abstract": "We consider a firm that sells products over $T$ periods without knowing the demand function. The firm sequentially sets prices to earn revenue and to learn the underlying demand function simultaneously. A natural heuristic for this problem, commonly used in practice, is greedy iterative least squares (GILS). At each time period, GILS estimates the demand as a linear function of the price by applying least squares to the set of prior prices and realized demands. Then a price that maximizes the revenue, given the estimated demand function, is used for the next time period. The performance is measured by the regret, which is the expected revenue loss from the optimal (oracle) pricing policy when the demand function is known. Recently, den Boer and Zwart (2014) and Keskin and Zeevi (2014) demonstrated that GILS is sub-optimal. They introduced algorithms which integrate forced price dispersion with GILS and achieve asymptotically optimal performance.   In this paper, we consider this dynamic pricing problem in a data-rich environment. In particular, we assume that the firm knows the expected demand under a particular price from historical data, and in each period, before setting the price, the firm has access to extra information (demand covariates) which may be predictive of the demand. We prove that in this setting GILS achieves asymptotically optimal regret of order $\\log(T)$. We also show the following surprising result: in the original dynamic pricing problem of den Boer and Zwart (2014) and Keskin and Zeevi (2014), inclusion of any set of covariates in GILS as potential demand covariates (even though they could carry no information) would make GILS asymptotically optimal. We validate our results via extensive numerical simulations on synthetic and real data sets."
                    },
                    {
                        "title": "On Low-rank Trace Regression under General Sampling Distribution",
                        "abstract": "In this paper, we study the trace regression when a matrix of parameters B* is estimated via the convex relaxation of a rank-regularized regression or via regularized non-convex optimization. It is known that these estimators satisfy near-optimal error bounds under assumptions on the rank, coherence, and spikiness of B*. We start by introducing a general notion of spikiness for B* that provides a generic recipe to prove the restricted strong convexity of the sampling operator of the trace regression and obtain near-optimal and non-asymptotic error bounds for the estimation error. Similar to the existing literature, these results require the regularization parameter to be above a certain theory-inspired threshold that depends on observation noise that may be unknown in practice. Next, we extend the error bounds to cases where the regularization parameter is chosen via cross-validation. This result is significant in that existing theoretical results on cross-validated estimators (Kale et al., 2011; Kumar et al., 2013; Abou-Moustafa and Szepesvari, 2017) do not apply to our setting since the estimators we study are not known to satisfy their required notion of stability. Finally, using simulations on synthetic and real data, we show that the cross-validated estimator selects a near-optimal penalty parameter and outperforms the theory-inspired approach of selecting the parameter."
                    },
                    {
                        "title": "A General Theory of the Stochastic Linear Bandit and Its Applications",
                        "abstract": "Recent growing adoption of experimentation in practice has led to a surge of attention to multiarmed bandits as a technique to reduce the opportunity cost of online experiments. In this setting, a decision-maker sequentially chooses among a set of given actions, observes their noisy rewards, and aims to maximize her cumulative expected reward (or minimize regret) over a horizon of length $T$. In this paper, we introduce a general analysis framework and a family of algorithms for the stochastic linear bandit problem that includes well-known algorithms such as the optimism-in-the-face-of-uncertainty-linear-bandit (OFUL) and Thompson sampling (TS) as special cases. Our analysis technique bridges several streams of prior literature and yields a number of new results. First, our new notion of optimism in expectation gives rise to a new algorithm, called sieved greedy (SG) that reduces the overexploration problem in OFUL. SG utilizes the data to discard actions with relatively low uncertainty and then choosing one among the remaining actions greedily. In addition to proving that SG is theoretically rate optimal, our empirical simulations show that SG outperforms existing benchmarks such as greedy, OFUL, and TS. The second application of our general framework is (to the best of our knowledge) the first polylogarithmic (in $T$) regret bounds for OFUL and TS, under similar conditions as the ones by Goldenshluger and Zeevi (2013). Finally, we obtain sharper regret bounds for the $k$-armed contextual MABs by a factor of $\\sqrt{k}$."
                    },
                    {
                        "title": "On Frequentist Regret of Linear Thompson Sampling",
                        "abstract": "This paper studies the stochastic linear bandit problem, where a decision-maker chooses actions from possibly time-dependent sets of vectors in $\\mathbb{R}^d$ and receives noisy rewards. The objective is to minimize regret, the difference between the cumulative expected reward of the decision-maker and that of an oracle with access to the expected reward of each action, over a sequence of $T$ decisions. Linear Thompson Sampling (LinTS) is a popular Bayesian heuristic, supported by theoretical analysis that shows its Bayesian regret is bounded by $\\widetilde{\\mathcal{O}}(d\\sqrt{T})$, matching minimax lower bounds. However, previous studies demonstrate that the frequentist regret bound for LinTS is $\\widetilde{\\mathcal{O}}(d\\sqrt{dT})$, which requires posterior variance inflation and is by a factor of $\\sqrt{d}$ worse than the best optimism-based algorithms. We prove that this inflation is fundamental and that the frequentist bound of $\\widetilde{\\mathcal{O}}(d\\sqrt{dT})$ is the best possible, by demonstrating a randomization bias phenomenon in LinTS that can cause linear regret without inflation.We propose a data-driven version of LinTS that adjusts posterior inflation using observed data, which can achieve minimax optimal frequentist regret, under additional conditions. Our analysis provides new insights into LinTS and settles an open problem in the field."
                    },
                    {
                        "title": "Learning to Recommend Using Non-Uniform Data",
                        "abstract": "Learning user preferences for products based on their past purchases or reviews is at the cornerstone of modern recommendation engines. One complication in this learning task is that some users are more likely to purchase products or review them, and some products are more likely to be purchased or reviewed by the users. This non-uniform pattern degrades the power of many existing recommendation algorithms, as they assume that the observed data are sampled uniformly at random among user-product pairs. In addition, existing literature on modeling non-uniformity either assume user interests are independent of the products, or lack theoretical understanding. In this paper, we first model the user-product preferences as a partially observed matrix with non-uniform observation pattern. Next, building on the literature about low-rank matrix estimation, we introduce a new weighted trace-norm penalized regression to predict unobserved values of the matrix. We then prove an upper bound for the prediction error of our proposed approach. Our upper bound is a function of a number of parameters that are based on a certain weight matrix that depends on the joint distribution of users and products. Utilizing this observation, we introduce a new optimization problem to select a weight matrix that minimizes the upper bound on the prediction error. The final product is a new estimator, NU-Recommend, that outperforms existing methods in both synthetic and real datasets. Our approach aims at accurate predictions for all users while prioritizing fairness. To achieve this, we employ a bias-variance tradeoff mechanism that ensures good overall prediction performance without compromising the predictive accuracy for less active users."
                    },
                    {
                        "title": "Belief-Propagation for Weighted b-Matchings on Arbitrary Graphs and its Relation to Linear Programs with Integer Solutions",
                        "abstract": "We consider the general problem of finding the minimum weight $\\bm$-matching on arbitrary graphs. We prove that, whenever the linear programming (LP) relaxation of the problem has no fractional solutions, then the belief propagation (BP) algorithm converges to the correct solution. We also show that when the LP relaxation has a fractional solution then the BP algorithm can be used to solve the LP relaxation. Our proof is based on the notion of graph covers and extends the analysis of (Bayati-Shah-Sharma 2005 and Huang-Jebara 2007}.   These results are notable in the following regards: (1) It is one of a very small number of proofs showing correctness of BP without any constraint on the graph structure. (2) Variants of the proof work for both synchronous and asynchronous BP; it is the first proof of convergence and correctness of an asynchronous BP algorithm for a combinatorial optimization problem."
                    },
                    {
                        "title": "The LASSO risk for gaussian matrices",
                        "abstract": "We consider the problem of learning a coefficient vector x_0\\in R^N from noisy linear observation y=Ax_0+w \\in R^n. In many contexts (ranging from model selection to image processing) it is desirable to construct a sparse estimator x'. In this case, a popular approach consists in solving an L1-penalized least squares problem known as the LASSO or Basis Pursuit DeNoising (BPDN).   For sequences of matrices A of increasing dimensions, with independent gaussian entries, we prove that the normalized risk of the LASSO converges to a limit, and we obtain an explicit expression for this limit. Our result is the first rigorous derivation of an explicit formula for the asymptotic mean square error of the LASSO for random instances. The proof technique is based on the analysis of AMP, a recently developed efficient algorithm, that is inspired from graphical models ideas.   Simulations on real data matrices suggest that our results can be relevant in a broad array of practical applications."
                    },
                    {
                        "title": "Recommendation on a Budget: Column Space Recovery from Partially Observed Entries with Random or Active Sampling",
                        "abstract": "We analyze alternating minimization for column space recovery of a partially observed, approximately low rank matrix with a growing number of columns and a fixed budget of observations per column. In this work, we prove that if the budget is greater than the rank of the matrix, column space recovery succeeds -- as the number of columns grows, the estimate from alternating minimization converges to the true column space with probability tending to one. From our proof techniques, we naturally formulate an active sampling strategy for choosing entries of a column that is theoretically and empirically (on synthetic and real data) better than the commonly studied uniformly random sampling strategy."
                    },
                    {
                        "title": "Geometry-Aware Approaches for Balancing Performance and Theoretical Guarantees in Linear Bandits",
                        "abstract": "This paper is motivated by recent research in the $d$-dimensional stochastic linear bandit literature, which has revealed an unsettling discrepancy: algorithms like Thompson sampling and Greedy demonstrate promising empirical performance, yet this contrasts with their pessimistic theoretical regret bounds. The challenge arises from the fact that while these algorithms may perform poorly in certain problem instances, they generally excel in typical instances. To address this, we propose a new data-driven technique that tracks the geometric properties of the uncertainty ellipsoid around the main problem parameter. This methodology enables us to formulate an instance-dependent frequentist regret bound, which incorporates the geometric information, for a broad class of base algorithms, including Greedy, OFUL, and Thompson sampling. This result allows us to identify and ``course-correct\" problem instances in which the base algorithms perform poorly. The course-corrected algorithms achieve the minimax optimal regret of order $\\tilde{\\mathcal{O}}(d\\sqrt{T})$ for a $T$-period decision-making scenario, effectively maintaining the desirable attributes of the base algorithms, including their empirical efficacy. We present simulation results to validate our findings using synthetic and real data."
                    },
                    {
                        "title": "A Probabilistic Approach for Alignment with Human Comparisons",
                        "abstract": "A growing trend involves integrating human knowledge into learning frameworks, leveraging subtle human feedback to refine AI models. Despite these advances, no comprehensive theoretical framework describing the specific conditions under which human comparisons improve the traditional supervised fine-tuning process has been developed. To bridge this gap, this paper studies the effective use of human comparisons to address limitations arising from noisy data and high-dimensional models. We propose a two-stage \"Supervised Fine Tuning+Human Comparison\" (SFT+HC) framework connecting machine learning with human feedback through a probabilistic bisection approach. The two-stage framework first learns low-dimensional representations from noisy-labeled data via an SFT procedure, and then uses human comparisons to improve the model alignment. To examine the efficacy of the alignment phase, we introduce a novel concept termed the \"label-noise-to-comparison-accuracy\" (LNCA) ratio. This paper theoretically identifies the conditions under which the \"SFT+HC\" framework outperforms pure SFT approach, leveraging this ratio to highlight the advantage of incorporating human evaluators in reducing sample complexity. We validate that the proposed conditions for the LNCA ratio are met in a case study conducted via an Amazon Mechanical Turk experiment."
                    },
                    {
                        "title": "Generating Random Networks Without Short Cycles",
                        "abstract": "Random graph generation is an important tool for studying large complex networks. Despite abundance of random graph models, constructing models with application-driven constraints is poorly understood. In order to advance state-of-the-art in this area, we focus on random graphs without short cycles as a stylized family of graphs, and propose the RandGraph algorithm for randomly generating them. For any constant k, when m=O(n^{1+1/[2k(k+3)]}), RandGraph generates an asymptotically uniform random graph with n vertices, m edges, and no cycle of length at most k using O(n^2m) operations. We also characterize the approximation error for finite values of n. To the best of our knowledge, this is the first polynomial-time algorithm for the problem. RandGraph works by sequentially adding $m$ edges to an empty graph with n vertices. Recently, such sequential algorithms have been successful for random sampling problems. Our main contributions to this line of research includes introducing a new approach for sequentially approximating edge-specific probabilities at each step of the algorithm, and providing a new method for analyzing such algorithms."
                    },
                    {
                        "title": "Causal Message Passing for Experiments with Unknown and General Network Interference",
                        "abstract": "Randomized experiments are a powerful methodology for data-driven evaluation of decisions or interventions. Yet, their validity may be undermined by network interference. This occurs when the treatment of one unit impacts not only its outcome but also that of connected units, biasing traditional treatment effect estimations. Our study introduces a new framework to accommodate complex and unknown network interference, moving beyond specialized models in the existing literature. Our framework, termed causal message-passing, is grounded in high-dimensional approximate message passing methodology. It is tailored for multi-period experiments and is particularly effective in settings with many units and prevalent network interference. The framework models causal effects as a dynamic process where a treated unit's impact propagates through the network via neighboring units until equilibrium is reached. This approach allows us to approximate the dynamics of potential outcomes over time, enabling the extraction of valuable information before treatment effects reach equilibrium. Utilizing causal message-passing, we introduce a practical algorithm to estimate the total treatment effect, defined as the impact observed when all units are treated compared to the scenario where no unit receives treatment. We demonstrate the effectiveness of this approach across five numerical scenarios, each characterized by a distinct interference structure."
                    },
                    {
                        "title": "Combinatorial approach to the interpolation method and scaling limits in sparse random graphs",
                        "abstract": "We establish the existence of free energy limits for several combinatorial models on Erd\\\"{o}s-R\\'{e}nyi graph $\\mathbb {G}(N,\\lfloor cN\\rfloor)$ and random $r$-regular graph $\\mathbb {G}(N,r)$. For a variety of models, including independent sets, MAX-CUT, coloring and K-SAT, we prove that the free energy both at a positive and zero temperature, appropriately rescaled, converges to a limit as the size of the underlying graph diverges to infinity. In the zero temperature case, this is interpreted as the existence of the scaling limit for the corresponding combinatorial optimization problem. For example, as a special case we prove that the size of a largest independent set in these graphs, normalized by the number of nodes converges to a limit w.h.p. This resolves an open problem which was proposed by Aldous (Some open problems) as one of his six favorite open problems. It was also mentioned as an open problem in several other places: Conjecture 2.20 in Wormald [In Surveys in Combinatorics, 1999 (Canterbury) (1999) 239-298 Cambridge Univ. Press]; Bollob\\'{a}s and Riordan [Random Structures Algorithms 39 (2011) 1-38]; Janson and Thomason [Combin. Probab. Comput. 17 (2008) 259-264] and Aldous and Steele [In Probability on Discrete Structures (2004) 1-72 Springer]."
                    },
                    {
                        "title": "The Unreasonable Effectiveness of Greedy Algorithms in Multi-Armed Bandit with Many Arms",
                        "abstract": "We investigate a Bayesian $k$-armed bandit problem in the \\emph{many-armed} regime, where $k \\geq \\sqrt{T}$ and $T$ represents the time horizon. Initially, and aligned with recent literature on many-armed bandit problems, we observe that subsampling plays a key role in designing optimal algorithms; the conventional UCB algorithm is sub-optimal, whereas a subsampled UCB (SS-UCB), which selects $\\Theta(\\sqrt{T})$ arms for execution under the UCB framework, achieves rate-optimality. However, despite SS-UCB's theoretical promise of optimal regret, it empirically underperforms compared to a greedy algorithm that consistently chooses the empirically best arm. This observation extends to contextual settings through simulations with real-world data. Our findings suggest a new form of \\emph{free exploration} beneficial to greedy algorithms in the many-armed context, fundamentally linked to a tail event concerning the prior distribution of arm rewards. This finding diverges from the notion of free exploration, which relates to covariate variation, as recently discussed in contextual bandit literature. Expanding upon these insights, we establish that the subsampled greedy approach not only achieves rate-optimality for Bernoulli bandits within the many-armed regime but also attains sublinear regret across broader distributions. Collectively, our research indicates that in the many-armed regime, practitioners might find greater value in adopting greedy algorithms."
                    },
                    {
                        "title": "Maximum Weight Matching via Max-Product Belief Propagation",
                        "abstract": "Max-product \"belief propagation\" is an iterative, local, message-passing algorithm for finding the maximum a posteriori (MAP) assignment of a discrete probability distribution specified by a graphical model. Despite the spectacular success of the algorithm in many application areas such as iterative decoding, computer vision and combinatorial optimization which involve graphs with many cycles, theoretical results about both correctness and convergence of the algorithm are known in few cases (Weiss-Freeman Wainwright, Yeddidia-Weiss-Freeman, Richardson-Urbanke}.   In this paper we consider the problem of finding the Maximum Weight Matching (MWM) in a weighted complete bipartite graph. We define a probability distribution on the bipartite graph whose MAP assignment corresponds to the MWM. We use the max-product algorithm for finding the MAP of this distribution or equivalently, the MWM on the bipartite graph. Even though the underlying bipartite graph has many short cycles, we find that surprisingly, the max-product algorithm always converges to the correct MAP assignment as long as the MAP assignment is unique. We provide a bound on the number of iterations required by the algorithm and evaluate the computational cost of the algorithm. We find that for a graph of size $n$, the computational cost of the algorithm scales as $O(n^3)$, which is the same as the computational cost of the best known algorithm. Finally, we establish the precise relation between the max-product algorithm and the celebrated {\\em auction} algorithm proposed by Bertsekas. This suggests possible connections between dual algorithm and max-product algorithm for discrete optimization problems."
                    },
                    {
                        "title": "Universality in polytope phase transitions and message passing algorithms",
                        "abstract": "We consider a class of nonlinear mappings $\\mathsf{F}_{A,N}$ in $\\mathbb{R}^N$ indexed by symmetric random matrices $A\\in\\mathbb{R}^{N\\times N}$ with independent entries. Within spin glass theory, special cases of these mappings correspond to iterating the TAP equations and were studied by Bolthausen [Comm. Math. Phys. 325 (2014) 333-366]. Within information theory, they are known as \"approximate message passing\" algorithms. We study the high-dimensional (large $N$) behavior of the iterates of $\\mathsf{F}$ for polynomial functions $\\mathsf{F}$, and prove that it is universal; that is, it depends only on the first two moments of the entries of $A$, under a sub-Gaussian tail condition. As an application, we prove the universality of a certain phase transition arising in polytope geometry and compressed sensing. This solves, for a broad class of random projections, a conjecture by David Donoho and Jared Tanner."
                    },
                    {
                        "title": "Ensemble Methods for Causal Effects in Panel Data Settings",
                        "abstract": "This paper studies a panel data setting where the goal is to estimate causal effects of an intervention by predicting the counterfactual values of outcomes for treated units, had they not received the treatment. Several approaches have been proposed for this problem, including regression methods, synthetic control methods and matrix completion methods. This paper considers an ensemble approach, and shows that it performs better than any of the individual methods in several economic datasets. Matrix completion methods are often given the most weight by the ensemble, but this clearly depends on the setting. We argue that ensemble methods present a fruitful direction for further research in the causal panel data setting."
                    }
                ]
            }
        }
    },
    "2310.06756": {
        "paper_data": {
            "title": "Going Beyond Neural Network Feature Similarity: The Network Feature Complexity and Its Interpretation Using Category Theory",
            "url": "http://arxiv.org/abs/2310.06756v2",
            "arxiv_id": "2310.06756",
            "authors": [
                "Yiting Chen",
                "Zhanpeng Zhou",
                "Junchi Yan"
            ],
            "abstract": "The behavior of neural networks still remains opaque, and a recently widely noted phenomenon is that networks often achieve similar performance when initialized with different random parameters. This phenomenon has attracted significant attention in measuring the similarity between features learned by distinct networks. However, feature similarity could be vague in describing the same feature since equivalent features hardly exist. In this paper, we expand the concept of equivalent feature and provide the definition of what we call functionally equivalent features. These features produce equivalent output under certain transformations. Using this definition, we aim to derive a more intrinsic metric for the so-called feature complexity regarding the redundancy of features learned by a neural network at each layer. We offer a formal interpretation of our approach through the lens of category theory, a well-developed area in mathematics. To quantify the feature complexity, we further propose an efficient algorithm named Iterative Feature Merging. Our experimental results validate our ideas and theories from various perspectives. We empirically demonstrate that the functionally equivalence widely exists among different features learned by the same neural network and we could reduce the number of parameters of the network without affecting the performance.The IFM shows great potential as a data-agnostic model prune method. We have also drawn several interesting empirical findings regarding the defined feature complexity.",
            "introduction": "   1 Introduction  Deep neural networks (DNNs) have achieved significant success across diverse fields, including vision, texts among other areas. However, DNNs are often regarded as \u201cblack box\u201d models associated with high dimensional feature maps and numerous parameters for training. Recently, many studies report an interesting phenomenon that neural networks with different random initialization often converge to solutions with similar performance on the test set\u00a0(Dauphin et\u00a0al., 2014; Frankle et\u00a0al., 2020). Meanwhile, methods such as pruning\u00a0(Wen et\u00a0al., 2016; Ye et\u00a0al., 2018; Peng et\u00a0al., 2019) and knowledge distillation\u00a0(Hinton et\u00a0al., 2015) have been proposed to reduce the complexity of the neural network structure, leading to a more compact DNN achieving similar performance to their dense/large counterparts. These phenomena further draw increasing studies on the features111Precisely speaking, note that the term feature used in this paper has two different meanings that can be (hopefully) distinguished from the context: 1) in its specific meaning, the layer-wise feature map corresponding to a data point as input to the trained network; 2) more generally, the characteristic or measurable properties of the (training) data that are learned by the neural network, which is in fact an function of training data, training algorithm and model structure. In fact, most previous works, e.g. measuring feature similarity\u00a0(Li et\u00a0al., 2016; Raghu et\u00a0al., 2017; Wang et\u00a0al., 2018; Williams et\u00a0al., 2021; Barannikov et\u00a0al., 2022), are based on a set of feature maps corresponding to a specific dataset (one layer-wise feature map for one input sample). Our definition and method, as will be shown later, correspond more to the general meaning of feature, which are input-agnostic and thus perhaps more reflects the nature of the network. learned by neural networks and a popular treatment is measuring the so-called feature similarity222The feature similarity is also referred to as representational similarity.. Various feature similarity measures\u00a0(Li et\u00a0al., 2016; Morcos et\u00a0al., 2018) have been devised to quantify the distance between two features, and some works have further empirically found that similar features can be learned from either between different networks\u00a0(Li et\u00a0al., 2016; Wang et\u00a0al., 2018) or within a (wide) network\u00a0(Nguyen et\u00a0al., 2021). In literature, feature similarity (or sometimes distance) measures are often designed to be invariant to certain transformations under certain transformations e.g. permutation\u00a0(Li et\u00a0al., 2016), isotropic scaling\u00a0(Barannikov et\u00a0al., 2022), invertible linear transformation\u00a0(Wang et\u00a0al., 2018), affine transformation\u00a0(Raghu et\u00a0al., 2017; Williams et\u00a0al., 2021) etc. However, the exact \u201cequivalent features\u201d in the sense of numerically zero distance can hardly exist in practice. Therefore, previous works are often vague in exactly describing the inherent structure of the neural networks (e.g. the redundancy of features learned by a neural network or the same feature learned by different neural networks) and can hardly provide actionable insights e.g. which features can be merged (when they are equivalent features).   In this paper, we expand the concept of feature equivalence in feature similarity literature\u00a0(Li et\u00a0al., 2016; Raghu et\u00a0al., 2017; Wang et\u00a0al., 2018; Williams et\u00a0al., 2021; Barannikov et\u00a0al., 2022) and formally define a more general form: the so-called functional equivalence. Features are considered functionally equivalent if they produce outputs that are equivalent under certain transformations. Using the concept of functionally equivalent features, one can derive",
            "references": [
                {
                    "title": "Going Beyond Linear Mode Connectivity: The Layerwise Linear Feature Connectivity",
                    "abstract": "Recent work has revealed many intriguing empirical phenomena in neural network training, despite the poorly understood and highly complex loss landscapes and training dynamics. One of these phenomena, Linear Mode Connectivity (LMC), has gained considerable attention due to the intriguing observation that different solutions can be connected by a linear path in the parameter space while maintaining near-constant training and test losses. In this work, we introduce a stronger notion of linear connectivity, Layerwise Linear Feature Connectivity (LLFC), which says that the feature maps of every layer in different trained networks are also linearly connected. We provide comprehensive empirical evidence for LLFC across a wide range of settings, demonstrating that whenever two trained networks satisfy LMC (via either spawning or permutation methods), they also satisfy LLFC in nearly all the layers. Furthermore, we delve deeper into the underlying factors contributing to LLFC, which reveal new insights into the spawning and permutation approaches. The study of LLFC transcends and advances our understanding of LMC by adopting a feature-learning perspective."
                },
                {
                    "title": "Integral Neural Networks",
                    "abstract": "We introduce a new family of deep neural networks, where instead of the conventional representation of network layers as N-dimensional weight tensors, we use a continuous layer representation along the filter and channel dimensions. We call such networks Integral Neural Networks (INNs). In particular, the weights of INNs are represented as continuous functions defined on N-dimensional hypercubes, and the discrete transformations of inputs to the layers are replaced by continuous integration operations, accordingly. During the inference stage, our continuous layers can be converted into the traditional tensor representation via numerical integral quadratures. Such kind of representation allows the discretization of a network to an arbitrary size with various discretization intervals for the integral kernels. This approach can be applied to prune the model directly on an edge device while suffering only a small performance loss at high rates of structural pruning without any fine-tuning. To evaluate the practical benefits of our proposed approach, we have conducted experiments using various neural network architectures on multiple tasks. Our reported results show that the proposed INNs achieve the same performance with their conventional discrete counterparts, while being able to preserve approximately the same performance (2% accuracy loss for ResNet18 on Imagenet) at a high rate (up to 30%) of structural pruning without fine-tuning, compared to 65% accuracy loss of the conventional pruning methods under the same conditions. Code is available at gitee."
                },
                {
                    "title": "DIFFormer: Scalable (Graph) Transformers Induced by Energy Constrained Diffusion",
                    "abstract": "Real-world data generation often involves complex inter-dependencies among instances, violating the IID-data hypothesis of standard learning paradigms and posing a challenge for uncovering the geometric structures for learning desired instance representations. To this end, we introduce an energy constrained diffusion model which encodes a batch of instances from a dataset into evolutionary states that progressively incorporate other instances' information by their interactions. The diffusion process is constrained by descent criteria w.r.t.~a principled energy function that characterizes the global consistency of instance representations over latent structures. We provide rigorous theory that implies closed-form optimal estimates for the pairwise diffusion strength among arbitrary instance pairs, which gives rise to a new class of neural encoders, dubbed as DIFFormer (diffusion-based Transformers), with two instantiations: a simple version with linear complexity for prohibitive instance numbers, and an advanced version for learning complex structures. Experiments highlight the wide applicability of our model as a general-purpose encoder backbone with superior performance in various tasks, such as node classification on large graphs, semi-supervised image/text classification, and spatial-temporal dynamics prediction."
                },
                {
                    "title": "On the power of foundation models",
                    "abstract": "With infinitely many high-quality data points, infinite computational power, an infinitely large foundation model with a perfect training algorithm and guaranteed zero generalization error on the pretext task, can the model be used for everything? This question cannot be answered by the existing theory of representation, optimization or generalization, because the issues they mainly investigate are assumed to be nonexistent here. In this paper, we show that category theory provides powerful machinery to answer this question. We have proved three results. The first one limits the power of prompt-based learning, saying that the model can solve a downstream task with prompts if and only if the task is representable. The second one says fine tuning does not have this limit, as a foundation model with the minimum required power (up to symmetry) can theoretically solve downstream tasks for the category defined by pretext task, with fine tuning and enough resources. Our final result can be seen as a new type of generalization theorem, showing that the foundation model can generate unseen objects from the target category (e.g., images) using the structural information from the source category (e.g., texts). Along the way, we provide a categorical framework for supervised and self-supervised learning, which might be of independent interest."
                },
                {
                    "title": "Unifying Causal Inference and Reinforcement Learning using Higher-Order Category Theory",
                    "abstract": "We present a unified formalism for structure discovery of causal models and predictive state representation (PSR) models in reinforcement learning (RL) using higher-order category theory. Specifically, we model structure discovery in both settings using simplicial objects, contravariant functors from the category of ordinal numbers into any category. Fragments of causal models that are equivalent under conditional independence -- defined as causal horns -- as well as subsequences of potential tests in a predictive state representation -- defined as predictive horns -- are both special cases of horns of a simplicial object, subsets resulting from the removal of the interior and the face opposite a particular vertex. Latent structure discovery in both settings involve the same fundamental mathematical problem of finding extensions of horns of simplicial objects through solving lifting problems in commutative diagrams, and exploiting weak homotopies that define higher-order symmetries. Solutions to the problem of filling\"inner\"vs\"outer\"horns leads to various notions of higher-order categories, including weak Kan complexes and quasicategories. We define the abstract problem of structure discovery in both settings in terms of adjoint functors between the category of universal causal models or universal decision models and its simplicial object representation."
                },
                {
                    "title": "Git Re-Basin: Merging Models modulo Permutation Symmetries",
                    "abstract": "The success of deep learning is due in large part to our ability to solve certain massive non-convex optimization problems with relative ease. Though non-convex optimization is NP-hard, simple algorithms -- often variants of stochastic gradient descent -- exhibit surprising effectiveness in fitting large neural networks in practice. We argue that neural network loss landscapes often contain (nearly) a single basin after accounting for all possible permutation symmetries of hidden units a la Entezari et al. 2021. We introduce three algorithms to permute the units of one model to bring them into alignment with a reference model in order to merge the two models in weight space. This transformation produces a functionally equivalent set of weights that lie in an approximately convex basin near the reference model. Experimentally, we demonstrate the single basin phenomenon across a variety of model architectures and datasets, including the first (to our knowledge) demonstration of zero-barrier linear mode connectivity between independently trained ResNet models on CIFAR-10. Additionally, we identify intriguing phenomena relating model width and training time to mode connectivity. Finally, we discuss shortcomings of the linear mode connectivity hypothesis, including a counterexample to the single basin theory."
                },
                {
                    "title": "On the Symmetries of Deep Learning Models and their Internal Representations",
                    "abstract": "Symmetry is a fundamental tool in the exploration of a broad range of complex systems. In machine learning symmetry has been explored in both models and data. In this paper we seek to connect the symmetries arising from the architecture of a family of models with the symmetries of that family's internal representation of data. We do this by calculating a set of fundamental symmetry groups, which we call the intertwiner groups of the model. We connect intertwiner groups to a model's internal representations of data through a range of experiments that probe similarities between hidden states across models with the same architecture. Our work suggests that the symmetries of a network are propagated into the symmetries in that network's representation of data, providing us with a better understanding of how architecture affects the learning and prediction process. Finally, we speculate that for ReLU networks, the intertwiner groups may provide a justification for the common practice of concentrating model interpretability exploration on the activation basis in hidden layers rather than arbitrary linear combinations thereof."
                },
                {
                    "title": "Building Decision Forest via Deep Reinforcement Learning",
                    "abstract": "Ensemble learning methods whose base classifier is a decision tree usually belong to the bagging or boosting. However, no previous work has ever built the ensemble classifier by maximizing long-term returns to the best of our knowledge. This paper proposes a decision forest building method called MA-H-SAC-DF for binary classification via deep reinforcement learning. First, the building process is modeled as a decentralized partial observable Markov decision process, and a set of cooperative agents jointly constructs all base classifiers. Second, the global state and local observations are defined based on informations of the parent node and the current location. Last, the state-of-the-art deep reinforcement method Hybrid SAC is extended to a multi-agent system under the CTDE architecture to find an optimal decision forest building policy. The experiments indicate that MA-H-SAC-DF has the same performance as random forest, Adaboost, and GBDT on balanced datasets and outperforms them on imbalanced datasets."
                },
                {
                    "title": "Representation Topology Divergence: A Method for Comparing Neural Network Representations",
                    "abstract": "Comparison of data representations is a complex multi-aspect problem that has not enjoyed a complete solution yet. We propose a method for comparing two data representations. We introduce the Representation Topology Divergence (RTD), measuring the dissimilarity in multi-scale topology between two point clouds of equal size with a one-to-one correspondence between points. The data point clouds are allowed to lie in different ambient spaces. The RTD is one of the few TDA-based practical methods applicable to real machine learning datasets. Experiments show that the proposed RTD agrees with the intuitive assessment of data representation similarity and is sensitive to its topological structure. We apply RTD to gain insights on neural networks representations in computer vision and NLP domains for various problems: training dynamics analysis, data distribution shift, transfer learning, ensemble learning, disentanglement assessment."
                },
                {
                    "title": "Generalized Shape Metrics on Neural Representations",
                    "abstract": "Understanding the operation of biological and artificial networks remains a difficult and important challenge. To identify general principles, researchers are increasingly interested in surveying large collections of networks that are trained on, or biologically adapted to, similar tasks. A standardized set of analysis tools is now needed to identify how network-level covariates-such as architecture, anatomical brain region, and model organism-impact neural representations (hidden layer activations). Here, we provide a rigorous foundation for these analyses by defining a broad family of metric spaces that quantify representational dissimilarity. Using this framework, we modify existing representational similarity measures based on canonical correlation analysis and centered kernel alignment to satisfy the triangle inequality, formulate a novel metric that respects the inductive biases in convolutional layers, and identify approximate Euclidean embeddings that enable network representations to be incorporated into essentially any off-the-shelf machine learning method. We demonstrate these methods on large-scale datasets from biology (Allen Institute Brain Observatory) and deep learning (NAS-Bench-101). In doing so, we identify relationships between neural representations that are interpretable in terms of anatomical features and model performance."
                },
                {
                    "title": "The Role of Permutation Invariance in Linear Mode Connectivity of Neural Networks",
                    "abstract": "In this paper, we conjecture that if the permutation invariance of neural networks is taken into account, SGD solutions will likely have no barrier in the linear interpolation between them. Although it is a bold conjecture, we show how extensive empirical attempts fall short of refuting it. We further provide a preliminary theoretical result to support our conjecture. Our conjecture has implications for lottery ticket hypothesis, distributed training, and ensemble methods."
                },
                {
                    "title": "Category Theory in Machine Learning",
                    "abstract": "Over the past two decades machine learning has permeated almost every realm of technology. At the same time, many researchers have begun using category theory as a unifying language, facilitating communication between different scientific disciplines. It is therefore unsurprising that there is a burgeoning interest in applying category theory to machine learning. We aim to document the motivations, goals and common themes across these applications. We touch on gradient-based learning, probability, and equivariant learning."
                },
                {
                    "title": "Recent Advances on Neural Network Pruning at Initialization",
                    "abstract": "Neural network pruning typically removes connections or neurons from a pretrained converged model; while a new pruning paradigm, pruning at initialization (PaI), attempts to prune a randomly initialized network. This paper offers the first survey concentrated on this emerging pruning fashion. We first introduce a generic formulation of neural network pruning, followed by the major classic pruning topics. Then, as the main body of this paper, a thorough and structured literature review of PaI methods is presented, consisting of two major tracks (sparse training and sparse selection). Finally, we summarize the surge of PaI compared to PaT and discuss the open problems. Apart from the dedicated literature review, this paper also offers a code base for easy sanity-checking and benchmarking of different PaI methods."
                },
                {
                    "title": "Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depth",
                    "abstract": "A key factor in the success of deep neural networks is the ability to scale models to improve performance by varying the architecture depth and width. This simple property of neural network design has resulted in highly effective architectures for a variety of tasks. Nevertheless, there is limited understanding of effects of depth and width on the learned representations. In this paper, we study this fundamental question. We begin by investigating how varying depth and width affects model hidden representations, finding a characteristic block structure in the hidden representations of larger capacity (wider or deeper) models. We demonstrate that this block structure arises when model capacity is large relative to the size of the training set, and is indicative of the underlying layers preserving and propagating the dominant principal component of their representations. This discovery has important ramifications for features learned by different models, namely, representations outside the block structure are often similar across architectures with varying widths and depths, but the block structure is unique to each model. We analyze the output predictions of different model architectures, finding that even when the overall accuracy is similar, wide and deep models exhibit distinctive error patterns and variations across classes."
                },
                {
                    "title": "Pruning neural networks without any data by iteratively conserving synaptic flow",
                    "abstract": "Pruning the parameters of deep neural networks has generated intense interest due to potential savings in time, memory and energy both during training and at test time. Recent works have identified, through an expensive sequence of training and pruning cycles, the existence of winning lottery tickets or sparse trainable subnetworks at initialization. This raises a foundational question: can we identify highly sparse trainable subnetworks at initialization, without ever training, or indeed without ever looking at the data? We provide an affirmative answer to this question through theory driven algorithm design. We first mathematically formulate and experimentally verify a conservation law that explains why existing gradient-based pruning algorithms at initialization suffer from layer-collapse, the premature pruning of an entire layer rendering a network untrainable. This theory also elucidates how layer-collapse can be entirely avoided, motivating a novel pruning algorithm Iterative Synaptic Flow Pruning (SynFlow). This algorithm can be interpreted as preserving the total flow of synaptic strengths through the network at initialization subject to a sparsity constraint. Notably, this algorithm makes no reference to the training data and consistently competes with or outperforms existing state-of-the-art pruning algorithms at initialization over a range of models (VGG and ResNet), datasets (CIFAR-10/100 and Tiny ImageNet), and sparsity constraints (up to 99.99 percent). Thus our data-agnostic pruning algorithm challenges the existing paradigm that, at initialization, data must be used to quantify which synapses are important."
                },
                {
                    "title": "Similarity of Neural Networks with Gradients",
                    "abstract": "A suitable similarity index for comparing learnt neural networks plays an important role in understanding the behaviour of the highly-nonlinear functions, and can provide insights on further theoretical analysis and empirical studies. We define two key steps when comparing models: firstly, the representation abstracted from the learnt model, where we propose to leverage both feature vectors and gradient ones (which are largely ignored in prior work) into designing the representation of a neural network. Secondly, we define the employed similarity index which gives desired invariance properties, and we facilitate the chosen ones with sketching techniques for comparing various datasets efficiently. Empirically, we show that the proposed approach provides a state-of-the-art method for computing similarity of neural networks that are trained independently on different datasets and the tasks defined by the datasets."
                },
                {
                    "title": "AdderNet: Do We Really Need Multiplications in Deep Learning?",
                    "abstract": "Compared with cheap addition operation, multiplication operation is of much higher computation complexity. The widely-used convolutions in deep neural networks are exactly cross-correlation to measure the similarity between input feature and convolution filters, which involves massive multiplications between float values. In this paper, we present adder networks (AdderNets) to trade these massive multiplications in deep neural networks, especially convolutional neural networks (CNNs), for much cheaper additions to reduce computation costs. In AdderNets, we take the L1-norm distance between filters and input feature as the output response. The influence of this new similarity measure on the optimization of neural network have been thoroughly analyzed. To achieve a better performance, we develop a special back-propagation approach for AdderNets by investigating the full-precision gradient. We then propose an adaptive learning rate strategy to enhance the training procedure of AdderNets according to the magnitude of each neuron's gradient. As a result, the proposed AdderNets can achieve 74.9% Top-1 accuracy 91.7% Top-5 accuracy using ResNet-50 on the ImageNet dataset without any multiplication in convolutional layer. The codes are publicly available at: (https://github.com/huaweinoah/AdderNet)."
                },
                {
                    "title": "Linear Mode Connectivity and the Lottery Ticket Hypothesis",
                    "abstract": "We study whether a neural network optimizes to the same, linearly connected minimum under different samples of SGD noise (e.g., random data order and augmentation). We find that standard vision models become stable to SGD noise in this way early in training. From then on, the outcome of optimization is determined to a linearly connected region. We use this technique to study iterative magnitude pruning (IMP), the procedure used by work on the lottery ticket hypothesis to identify subnetworks that could have trained in isolation to full accuracy. We find that these subnetworks only reach full accuracy when they are stable to SGD noise, which either occurs at initialization for small-scale settings (MNIST) or early in training for large-scale settings (ResNet-50 and Inception-v3 on ImageNet)."
                },
                {
                    "title": "Collaborative Channel Pruning for Deep Networks",
                    "abstract": "Deep networks have achieved impressive performance in various domains, but their applications are largely limited by the prohibitive computational overhead. In this paper, we propose a novel algorithm, namely collaborative channel pruning (CCP), to reduce the computational overhead with negligible performance degradation. The joint impact of pruned/preserved channels on the loss function is quantitatively analyzed, and such interchannel dependency is exploited to determine which channels to be pruned. The channel selection problem is then reformulated as a constrained 0-1 quadratic optimization problem, and the Hessian matrix, which is essential in constructing the above optimization, can be efficiently approximated. Empirical evaluation on two benchmark data sets indicates that our proposed CCP algorithm achieves higher classification accuracy with similar computational complexity than other stateof-the-art channel pruning algorithms."
                },
                {
                    "title": "Characterizing the invariances of learning algorithms using category theory",
                    "abstract": "Many learning algorithms have invariances: when their training data is transformed in certain ways, the function they learn transforms in a predictable manner. Here we formalize this notion using concepts from the mathematical field of category theory. The invariances that a supervised learning algorithm possesses are formalized by categories of predictor and target spaces, whose morphisms represent the algorithm's invariances, and an index category whose morphisms represent permutations of the training examples. An invariant learning algorithm is a natural transformation between two functors from the product of these categories to the category of sets, representing training datasets and learned functions respectively. We illustrate the framework by characterizing and contrasting the invariances of linear regression and ridge regression."
                },
                {
                    "title": "Stabilizing the Lottery Ticket Hypothesis",
                    "abstract": "Pruning is a well-established technique for removing unnecessary structure from neural networks after training to improve the performance of inference. Several recent results have explored the possibility of pruning at initialization time to provide similar benefits during training. In particular, the \"lottery ticket hypothesis\" conjectures that typical neural networks contain small subnetworks that can train to similar accuracy in a commensurate number of steps. The evidence for this claim is that a procedure based on iterative magnitude pruning (IMP) reliably finds such subnetworks retroactively on small vision tasks. However, IMP fails on deeper networks, and proposed methods to prune before training or train pruned networks encounter similar scaling limitations. \nIn this paper, we argue that these efforts have struggled on deeper networks because they have focused on pruning precisely at initialization. We modify IMP to search for subnetworks that could have been obtained by pruning early in training (0.1% to 7% through) rather than at iteration 0. With this change, it finds small subnetworks of deeper networks (e.g., 80% sparsity on Resnet-50) that can complete the training process to match the accuracy of the original network on more challenging tasks (e.g., ImageNet). In situations where IMP fails at iteration 0, the accuracy benefits of delaying pruning accrue rapidly over the earliest iterations of training. To explain these behaviors, we study subnetwork \"stability,\" finding that - as accuracy improves in this fashion - IMP subnetworks train to parameters closer to those of the full network and do so with improved consistency in the face of gradient noise. These results offer new insights into the opportunity to prune large-scale networks early in training and the behaviors underlying the lottery ticket hypothesis"
                },
                {
                    "title": "Towards Understanding Learning Representations: To What Extent Do Different Neural Networks Learn the Same Representation",
                    "abstract": "It is widely believed that learning good representations is one of the main reasons for the success of deep neural networks. Although highly intuitive, there is a lack of theory and systematic approach quantitatively characterizing what representations do deep neural networks learn. In this work, we move a tiny step towards a theory and better understanding of the representations. Specifically, we study a simpler problem: How similar are the representations learned by two networks with identical architecture but trained from different initializations. We develop a rigorous theory based on the neuron activation subspace match model. The theory gives a complete characterization of the structure of neuron activation subspace matches, where the core concepts are maximum match and simple match which describe the overall and the finest similarity between sets of neurons in two networks respectively. We also propose efficient algorithms to find the maximum match and simple matches. Finally, we conduct extensive experiments using our algorithms. Experimental results suggest that, surprisingly, representations learned by the same convolutional layers of networks trained from different initializations are not as similar as prevalently expected, at least in terms of subspace match."
                },
                {
                    "title": "SNIP: Single-shot Network Pruning based on Connection Sensitivity",
                    "abstract": "Pruning large neural networks while maintaining their performance is often desirable due to the reduced space and time complexity. In existing methods, pruning is done within an iterative optimization procedure with either heuristically designed pruning schedules or additional hyperparameters, undermining their utility. In this work, we present a new approach that prunes a given network once at initialization prior to training. To achieve this, we introduce a saliency criterion based on connection sensitivity that identifies structurally important connections in the network for the given task. This eliminates the need for both pretraining and the complex pruning schedule while making it robust to architecture variations. After pruning, the sparse network is trained in the standard way. Our method obtains extremely sparse networks with virtually the same accuracy as the reference network on the MNIST, CIFAR-10, and Tiny-ImageNet classification tasks and is broadly applicable to various architectures including convolutional, residual and recurrent networks. Unlike existing methods, our approach enables us to demonstrate that the retained connections are indeed relevant to the given task."
                },
                {
                    "title": "Insights on representational similarity in neural networks with canonical correlation",
                    "abstract": "Comparing different neural network representations and determining how representations evolve over time remain challenging open questions in our understanding of the function of neural networks. Comparing representations in neural networks is fundamentally difficult as the structure of representations varies greatly, even across groups of networks trained on identical tasks, and over the course of training. Here, we develop projection weighted CCA (Canonical Correlation Analysis) as a tool for understanding neural networks, building off of SVCCA, a recently proposed method (Raghu et al., 2017). We first improve the core method, showing how to differentiate between signal and noise, and then apply this technique to compare across a group of CNNs, demonstrating that networks which generalize converge to more similar representations than networks which memorize, that wider networks converge to more similar solutions than narrow networks, and that trained networks with identical topology but different learning rates converge to distinct clusters with diverse representations. We also investigate the representational dynamics of RNNs, across both training and sequential timesteps, finding that RNNs converge in a bottom-up pattern over the course of training and that the hidden state is highly variable over the course of a sequence, even when accounting for linear transforms. Together, these results provide new insights into the function of CNNs and RNNs, and demonstrate the utility of using CCA to understand representations."
                },
                {
                    "title": "The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks",
                    "abstract": "Neural network pruning techniques can reduce the parameter counts of trained networks by over 90%, decreasing storage requirements and improving computational performance of inference without compromising accuracy. However, contemporary experience is that the sparse architectures produced by pruning are difficult to train from the start, which would similarly improve training performance. \nWe find that a standard pruning technique naturally uncovers subnetworks whose initializations made them capable of training effectively. Based on these results, we articulate the \"lottery ticket hypothesis:\" dense, randomly-initialized, feed-forward networks contain subnetworks (\"winning tickets\") that - when trained in isolation - reach test accuracy comparable to the original network in a similar number of iterations. The winning tickets we find have won the initialization lottery: their connections have initial weights that make training particularly effective. \nWe present an algorithm to identify winning tickets and a series of experiments that support the lottery ticket hypothesis and the importance of these fortuitous initializations. We consistently find winning tickets that are less than 10-20% of the size of several fully-connected and convolutional feed-forward architectures for MNIST and CIFAR10. Above this size, the winning tickets that we find learn faster than the original network and reach higher test accuracy."
                },
                {
                    "title": "Essentially No Barriers in Neural Network Energy Landscape",
                    "abstract": "Training neural networks involves finding minima of a high-dimensional non-convex loss function. Knowledge of the structure of this energy landscape is sparse. Relaxing from linear interpolations, we construct continuous paths between minima of recent neural network architectures on CIFAR10 and CIFAR100. Surprisingly, the paths are essentially flat in both the training and test landscapes. This implies that neural networks have enough capacity for structural changes, or that these changes are small between minima. Also, each minimum has at least one vanishing Hessian eigenvalue in addition to those resulting from trivial invariance."
                },
                {
                    "title": "Loss Surfaces, Mode Connectivity, and Fast Ensembling of DNNs",
                    "abstract": "The loss functions of deep neural networks are complex and their geometric properties are not well understood. We show that the optima of these complex loss functions are in fact connected by simple curves over which training and test accuracy are nearly constant. We introduce a training procedure to discover these high-accuracy pathways between modes. Inspired by this new geometric insight, we also propose a new ensembling method entitled Fast Geometric Ensembling (FGE). Using FGE we can train high-performing ensembles in the time required to train a single model. We achieve improved performance compared to the recent state-of-the-art Snapshot Ensembles, on CIFAR-10, CIFAR-100, and ImageNet."
                },
                {
                    "title": "Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of Convolution Layers",
                    "abstract": "Model pruning has become a useful technique that improves the computational efficiency of deep learning, making it possible to deploy solutions in resource-limited scenarios. A widely-used practice in relevant work assumes that a smaller-norm parameter or feature plays a less informative role at the inference time. In this paper, we propose a channel pruning technique for accelerating the computations of deep convolutional neural networks (CNNs) that does not critically rely on this assumption. Instead, it focuses on direct simplification of the channel-to-channel computation graph of a CNN without the need of performing a computationally difficult and not-always-useful task of making high-dimensional tensors of CNN structured sparse. Our approach takes two stages: first to adopt an end-to- end stochastic training method that eventually forces the outputs of some channels to be constant, and then to prune those constant channels from the original neural network by adjusting the biases of their impacting layers such that the resulting compact model can be quickly fine-tuned. Our approach is mathematically appealing from an optimization perspective and easy to reproduce. We experimented our approach through several image learning benchmarks and demonstrate its interesting aspects and competitive performance."
                },
                {
                    "title": "Backprop as Functor: A compositional perspective on supervised learning",
                    "abstract": "A supervised learning algorithm searches over a set of functions $A\\rightarrow B$ parametrised by a space $P$ to find the best approximation to some ideal function $f:A\\rightarrow B$. It does this by taking examples $(a, f(a))\\in A\\times B$, and updating the parameter according to some rule. We define a category where these update rules may be composed, and show that gradient descent-with respect to a fixed step size and an error function satisfying a certain property-defines a monoidal functor from a category of parametrised functions to this category of update rules. A key contribution is the notion of request function. This provides a structural perspective on backpropagation, giving a broad generalisation of neural networks and linking it with structures from bidirectional programming and open games."
                },
                {
                    "title": "SVCCA: Singular Vector Canonical Correlation Analysis for Deep Learning Dynamics and Interpretability",
                    "abstract": "We propose a new technique, Singular Vector Canonical Correlation Analysis (SVCCA), a tool for quickly comparing two representations in a way that is both invariant to affine transform (allowing comparison between different layers and networks) and fast to compute (allowing more comparisons to be calculated than with previous methods). We deploy this tool to measure the intrinsic dimensionality of layers, showing in some cases needless over-parameterization; to probe learning dynamics throughout training, finding that networks converge to final representations from the bottom up; to show where class-specific information in networks is formed; and to suggest new training regimes that simultaneously save computation and overfit less."
                },
                {
                    "title": "Topology and Geometry of Half-Rectified Network Optimization",
                    "abstract": "The loss surface of deep neural networks has recently attracted interest in the optimization and machine learning communities as a prime example of high-dimensional non-convex problem. Some insights were recently gained using spin glass models and mean-field approximations, but at the expense of strongly simplifying the nonlinear nature of the model. \nIn this work, we do not make any such assumption and study conditions on the data distribution and model architecture that prevent the existence of bad local minima. Our theoretical work quantifies and formalizes two important \\emph{folklore} facts: (i) the landscape of deep linear networks has a radically different topology from that of deep half-rectified ones, and (ii) that the energy landscape in the non-linear case is fundamentally controlled by the interplay between the smoothness of the data distribution and model over-parametrization. Our main theoretical contribution is to prove that half-rectified single layer networks are asymptotically connected, and we provide explicit bounds that reveal the aforementioned interplay. \nThe conditioning of gradient descent is the next challenge we address. We study this question through the geometry of the level sets, and we introduce an algorithm to efficiently estimate the regularity of such sets on large-scale networks. Our empirical results show that these level sets remain connected throughout all the learning phase, suggesting a near convex behavior, but they become exponentially more curvy as the energy level decays, in accordance to what is observed in practice with very low curvature attractors."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Learning Structured Sparsity in Deep Neural Networks",
                    "abstract": "High demand for computation resources severely hinders deployment of large-scale Deep Neural Networks (DNN) in resource constrained devices. In this work, we propose a Structured Sparsity Learning (SSL) method to regularize the structures (i.e., filters, channels, filter shapes, and layer depth) of DNNs. SSL can: (1) learn a compact structure from a bigger DNN to reduce computation cost; (2) obtain a hardware-friendly structured sparsity of DNN to efficiently accelerate the DNNs evaluation. Experimental results show that SSL achieves on average 5.1x and 3.1x speedups of convolutional layer computation of AlexNet against CPU and GPU, respectively, with off-the-shelf libraries. These speedups are about twice speedups of non-structured sparsity; (3) regularize the DNN structure to improve classification accuracy. The results show that for CIFAR-10, regularization on layer depth can reduce 20 layers of a Deep Residual Network (ResNet) to 18 layers while improve the accuracy from 91.25% to 92.60%, which is still slightly higher than that of original ResNet with 32 layers. For AlexNet, structure regularization by SSL also reduces the error by around ~1%. Open source code is in this https URL"
                },
                {
                    "title": "Ternary Weight Networks",
                    "abstract": "We present a memory and computation efficient ternary weight networks (TWNs) - with weights constrained to +1, 0 and -1. The Euclidian distance between full (float or double) precision weights and the ternary weights along with a scaling factor is minimized in training stage. Besides, a threshold-based ternary function is optimized to get an approximated solution which can be fast and easily computed. TWNs have shown better expressive abilities than binary precision counterparts. Meanwhile, TWNs achieve up to 16\u00d7 model compression rate and need fewer multiplications compared with the float32 precision counterparts. Extensive experiments on MNIST, CIFAR-10, and ImageNet datasets show that the TWNs achieve much better result than the Binary-Weight-Networks (BWNs) and the classification performance on MNIST and CIFAR-10 is very close to the full precision networks. We also verify our method on object detection task and show that TWNs significantly outperforms BWN by more than 10% mAP on PASCAL VOC dataset. The pytorch version of source code is available at: https://github.com/Thinklab-SJTU/twns."
                },
                {
                    "title": "Quantized Convolutional Neural Networks for Mobile Devices",
                    "abstract": "Recently, convolutional neural networks (CNN) have demonstrated impressive performance in various computer vision tasks. However, high performance hardware is typically indispensable for the application of CNN models due to the high computation complexity, which prohibits their further extensions. In this paper, we propose an efficient framework, namely Quantized CNN, to simultaneously speed-up the computation and reduce the storage and memory overhead of CNN models. Both filter kernels in convolutional layers and weighting matrices in fully-connected layers are quantized, aiming at minimizing the estimation error of each layer's response. Extensive experiments on the ILSVRC-12 benchmark demonstrate 4 ~ 6\u00d7 speed-up and 15 ~ 20\u00d7 compression with merely one percentage loss of classification accuracy. With our quantized CNN model, even mobile devices can accurately classify images within one second."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Convergent Learning: Do different neural networks learn the same representations?",
                    "abstract": "Recent success in training deep neural networks have prompted active investigation into the features learned on their intermediate layers. Such research is difficult because it requires making sense of non-linear computations performed by millions of parameters, but valuable because it increases our ability to understand current models and create improved versions of them. In this paper we investigate the extent to which neural networks exhibit what we call convergent learning, which is when the representations learned by multiple nets converge to a set of features which are either individually similar between networks or where subsets of features span similar low-dimensional spaces. We propose a specific method of probing representations: training multiple networks and then comparing and contrasting their individual, learned representations at the level of neurons or groups of neurons. We begin research into this question using three techniques to approximately align different neural networks on a feature level: a bipartite matching approach that makes one-to-one assignments between neurons, a sparse prediction approach that finds one-to-many mappings, and a spectral clustering approach that finds many-to-many mappings. This initial investigation reveals a few previously unknown properties of neural networks, and we argue that future research into the question of convergent learning will yield many more. The insights described here include (1) that some features are learned reliably in multiple networks, yet other features are not consistently learned; (2) that units learn to span low-dimensional subspaces and, while these subspaces are common to multiple networks, the specific basis vectors learned are not; (3) that the representation codes show evidence of being a mix between a local code and slightly, but not fully, distributed codes across multiple units."
                },
                {
                    "title": "Convolutional neural networks with low-rank regularization",
                    "abstract": "Large CNNs have delivered impressive performance in various computer vision applications. But the storage and computation requirements make it problematic for deploying these models on mobile devices. Recently, tensor decompositions have been used for speeding up CNNs. In this paper, we further develop the tensor decomposition technique. We propose a new algorithm for computing the low-rank tensor decomposition for removing the redundancy in the convolution kernels. The algorithm finds the exact global optimizer of the decomposition and is more effective than iterative methods. Based on the decomposition, we further propose a new method for training low-rank constrained CNNs from scratch. Interestingly, while achieving a significant speedup, sometimes the low-rank constrained CNNs delivers significantly better performance than their non-constrained counterparts. On the CIFAR-10 dataset, the proposed low-rank NIN model achieves $91.31\\%$ accuracy (without data augmentation), which also improves upon state-of-the-art result. We evaluated the proposed method on CIFAR-10 and ILSVRC12 datasets for a variety of modern CNNs, including AlexNet, NIN, VGG and GoogleNet with success. For example, the forward time of VGG-16 is reduced by half while the performance is still comparable. Empirical success suggests that low-rank tensor decompositions can be a very useful tool for speeding up large CNNs."
                },
                {
                    "title": "Data-free Parameter Pruning for Deep Neural Networks",
                    "abstract": "Deep Neural nets (NNs) with millions of parameters are at the heart of many state-of-the-art computer vision systems today. However, recent works have shown that much smaller models can achieve similar levels of performance. In this work, we address the problem of pruning parameters in a trained NN model. Instead of removing individual weights one at a time as done in previous works, we remove one neuron at a time. We show how similar neurons are redundant, and propose a systematic way to remove them. Our experiments in pruning the densely connected layers show that we can remove upto 85\\% of the total parameters in an MNIST-trained network, and about 35\\% for AlexNet without significantly affecting performance. Our method can be applied on top of most networks with a fully connected layer to give a smaller network."
                },
                {
                    "title": "Distilling the Knowledge in a Neural Network",
                    "abstract": "A very simple way to improve the performance of almost any machine learning algorithm is to train many different models on the same data and then to average their predictions. Unfortunately, making predictions using a whole ensemble of models is cumbersome and may be too computationally expensive to allow deployment to a large number of users, especially if the individual models are large neural nets. Caruana and his collaborators have shown that it is possible to compress the knowledge in an ensemble into a single model which is much easier to deploy and we develop this approach further using a different compression technique. We achieve some surprising results on MNIST and we show that we can significantly improve the acoustic model of a heavily used commercial system by distilling the knowledge in an ensemble of models into a single model. We also introduce a new type of ensemble composed of one or more full models and many specialist models which learn to distinguish fine-grained classes that the full models confuse. Unlike a mixture of experts, these specialist models can be trained rapidly and in parallel."
                },
                {
                    "title": "Striving for Simplicity: The All Convolutional Net",
                    "abstract": "Most modern convolutional neural networks (CNNs) used for object recognition are built using the same principles: Alternating convolution and max-pooling layers followed by a small number of fully connected layers. We re-evaluate the state of the art for object recognition from small images with convolutional networks, questioning the necessity of different components in the pipeline. We find that max-pooling can simply be replaced by a convolutional layer with increased stride without loss in accuracy on several image recognition benchmarks. Following this finding -- and building on other recent work for finding simple network structures -- we propose a new architecture that consists solely of convolutional layers and yields competitive or state of the art performance on several object recognition datasets (CIFAR-10, CIFAR-100, ImageNet). To analyze the network we introduce a new variant of the \"deconvolution approach\" for visualizing features learned by CNNs, which can be applied to a broader range of network structures than existing approaches."
                },
                {
                    "title": "Compressing Deep Convolutional Networks using Vector Quantization",
                    "abstract": "Deep convolutional neural networks (CNN) has become the most promising method for object recognition, repeatedly demonstrating record breaking results for image classification and object detection in recent years. However, a very deep CNN generally involves many layers with millions of parameters, making the storage of the network model to be extremely large. This prohibits the usage of deep CNNs on resource limited hardware, especially cell phones or other embedded devices. In this paper, we tackle this model storage issue by investigating information theoretical vector quantization methods for compressing the parameters of CNNs. In particular, we have found in terms of compressing the most storage demanding dense connected layers, vector quantization methods have a clear gain over existing matrix factorization methods. Simply applying k-means clustering to the weights or conducting product quantization can lead to a very good balance between model size and recognition accuracy. For the 1000-category classification task in the ImageNet challenge, we are able to achieve 16-24 times compression of the network with only 1% loss of classification accuracy using the state-of-the-art CNN."
                },
                {
                    "title": "How transferable are features in deep neural networks?",
                    "abstract": "Many deep neural networks trained on natural images exhibit a curious phenomenon in common: on the first layer they learn features similar to Gabor filters and color blobs. Such first-layer features appear not to be specific to a particular dataset or task, but general in that they are applicable to many datasets and tasks. Features must eventually transition from general to specific by the last layer of the network, but this transition has not been studied extensively. In this paper we experimentally quantify the generality versus specificity of neurons in each layer of a deep convolutional neural network and report a few surprising results. Transferability is negatively affected by two distinct issues: (1) the specialization of higher layer neurons to their original task at the expense of performance on the target task, which was expected, and (2) optimization difficulties related to splitting networks between co-adapted neurons, which was not expected. In an example network trained on ImageNet, we demonstrate that either of these two issues may dominate, depending on whether features are transferred from the bottom, middle, or top of the network. We also document that the transferability of features decreases as the distance between the base task and target task increases, but that transferring features even from distant tasks can be better than using random features. A final surprising result is that initializing a network with transferred features from almost any number of layers can produce a boost to generalization that lingers even after fine-tuning to the target dataset."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "Identifying and attacking the saddle point problem in high-dimensional non-convex optimization",
                    "abstract": "A central challenge to many fields of science and engineering involves minimizing non-convex error functions over continuous, high dimensional spaces. Gradient descent or quasi-Newton methods are almost ubiquitously used to perform such minimizations, and it is often thought that a main source of difficulty for these local methods to find the global minimum is the proliferation of local minima with much higher error than the global minimum. Here we argue, based on results from statistical physics, random matrix theory, neural network theory, and empirical evidence, that a deeper and more profound difficulty originates from the proliferation of saddle points, not local minima, especially in high dimensional problems of practical interest. Such saddle points are surrounded by high error plateaus that can dramatically slow down learning, and give the illusory impression of the existence of a local minimum. Motivated by these arguments, we propose a new approach to second-order optimization, the saddle-free Newton method, that can rapidly escape high dimensional saddle points, unlike gradient descent and quasi-Newton methods. We apply this algorithm to deep or recurrent neural network training, and provide numerical evidence for its superior optimization performance."
                },
                {
                    "title": "Spectral Networks and Locally Connected Networks on Graphs",
                    "abstract": "Convolutional Neural Networks are extremely efficient architectures in image and audio recognition tasks, thanks to their ability to exploit the local translational invariance of signal classes over their domain. In this paper we consider possible generalizations of CNNs to signals defined on more general domains without the action of a translation group. In particular, we propose two constructions, one based upon a hierarchical clustering of the domain, and another based on the spectrum of the graph Laplacian. We show through experiments that for low-dimensional graphs it is possible to learn convolutional layers with a number of parameters independent of the input size, resulting in efficient deep architectures."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "A new model for learning in graph domains",
                    "abstract": "In several applications the information is naturally represented by graphs. Traditional approaches cope with graphical data structures using a preprocessing phase which transforms the graphs into a set of flat vectors. However, in this way, important topological information may be lost and the achieved results may heavily depend on the preprocessing stage. This paper presents a new neural model, called graph neural network (GNN), capable of directly processing graphs. GNNs extends recursive neural networks and can be applied on most of the practically useful kinds of graphs, including directed, undirected, labelled and cyclic graphs. A learning algorithm for GNNs is proposed and some experiments are discussed which assess the properties of the model."
                },
                {
                    "title": "Optimal Brain Surgeon and general network pruning",
                    "abstract": "The use of information from all second-order derivatives of the error function to perform network pruning (i.e., removing unimportant weights from a trained network) in order to improve generalization, simplify networks, reduce hardware or storage requirements, increase the speed of further training, and, in some cases, enable rule extraction, is investigated. The method, Optimal Brain Surgeon (OBS), is significantly better than magnitude-based methods and Optimal Brain Damage, which often remove the wrong weights. OBS, permits pruning of more weights than other methods (for the same error on the training set), and thus yields better generalization on test data. Crucial to OBS is a recursion relation for calculating the inverse Hessian matrix H/sup -1/ from training data and structural information of the set. OBS deletes the correct weights from a trained XOR network in every case.<<ETX>>"
                },
                {
                    "title": "Deep Neural Network Fusion via Graph Matching with Applications to Model Ensemble and Federated Learning",
                    "abstract": "Model fusion without accessing training data in machine learning has attracted increasing interest due to the practical resource-saving and data privacy issues. During the training process, the neural weights of each model can be randomly permuted, and we have to align the channels of each layer before fusing them. Regrading the channels as nodes and weights as edges, aligning the channels to maximize weight similarity is a challenging NP-hard assignment problem. Due to its quadratic assignment nature, we formulate the model fusion problem as a graph matching task, considering the second-order similarity of model weights instead of previous work merely formulating model fusion as a linear assignment problem. For the rising problem scale and multi-model consistency issues, we propose an efficient graduated assignment-based model fusion method, dubbed GAMF, which iteratively updates the matchings in a consistency-maintaining manner. We apply GAMF to tackle the compact model ensemble task and federated learning task on MNIST, CIFAR-10, CIFAR-100, and Tiny-Imagenet. The performance shows the efficacy of our GAMF compared to state-of-the-art baselines."
                },
                {
                    "title": "\u221e-Categories for the Working Mathematician",
                    "abstract": "homotopy theory C.1. Lifting properties, weak factorization systems, and Leibniz closure C.1.1. Lemma. Any class of maps characterized by a right lifting property is closed under composition, product, pullback, retract, and limits of towers; see Lemma C.1.1. Proof. For now see [47, 11.1.4] and dualize. On account of the dual of Lemma C.1.1, any set of maps in a cocomplete category \u201ccellularly generates\u201d a larger class of maps with the same left lifting property. C.1.2. Definition. The class of maps cellularly generated by a set of maps is comprised of those maps obtained as sequential composites of pushouts of coproducts of those maps."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "The Graph Neural Network Model",
                    "abstract": "Many underlying relationships among data in several areas of science and engineering, e.g., computer vision, molecular chemistry, molecular biology, pattern recognition, and data mining, can be represented in terms of graphs. In this paper, we propose a new neural network model, called graph neural network (GNN) model, that extends existing neural network methods for processing the data represented in graph domains. This GNN model, which can directly process most of the practically useful types of graphs, e.g., acyclic, cyclic, directed, and undirected, implements a function tau(G,n) isin IRm that maps a graph G and one of its nodes n into an m-dimensional Euclidean space. A supervised learning algorithm is derived to estimate the parameters of the proposed GNN model. The computational cost of the proposed algorithm is also considered. Some experimental results are shown to validate the proposed learning algorithm, and to demonstrate its generalization capabilities."
                },
                {
                    "title": "Optimal Brain Damage",
                    "abstract": "We have used information-theoretic ideas to derive a class of practical and nearly optimal schemes for adapting the size of a neural network. By removing unimportant weights from a network, several improvements can be expected: better generalization, fewer training examples required, and improved speed of learning and/or classification. The basic idea is to use second-derivative information to make a tradeoff between network complexity and training set error. Experiments confirm the usefulness of the methods on a real-world application."
                },
                {
                    "title": "Comparing Biases for Minimal Network Construction with Back-Propagation",
                    "abstract": "Rumelhart (1987), has proposed a method for choosing minimal or \"simple\" representations during learning in Back-propagation networks. This approach can be used to (a) dynamically select the number of hidden units, (b) construct a representation that is appropriate for the problem and (c) thus improve the generalization ability of Back-propagation networks. The method Rumelhart suggests involves adding penalty terms to the usual error function. In this paper we introduce Rumelhart's minimal networks idea and compare two possible biases on the weight search space. These biases are compared in both simple counting problems and a speech recognition problem. In general, the constrained search does seem to minimize the number of hidden units required with an expected increase in local minima."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we formally define and measure functional equivalence among features learned by deep neural networks to better understand their redundancy and improve model efficiency?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the fundamental understanding of feature representation in deep neural networks. By clarifying the concept of functional equivalence, we can enhance model interpretability, leading to more efficient architectures and training processes. This could pave the way for advancements in model compression techniques, such as pruning and knowledge distillation, ultimately resulting in practical applications that require less computational resources while maintaining performance.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the high dimensionality and complexity of feature spaces in deep neural networks. Naive approaches may fail because they often overlook the nuanced relationships between features and their outputs, leading to oversimplified conclusions. Additionally, the lack of a clear, universally accepted definition of feature equivalence complicates the development of robust measurement techniques. Technical obstacles include the need for invariant measures that can accurately capture the essence of features across different transformations, while theoretical challenges involve understanding the intricate dynamics of feature interactions within and across networks.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on specific feature maps tied to individual datasets, lacking a comprehensive framework for understanding feature equivalence in a broader context. Limitations in existing solutions include vague definitions and a failure to provide actionable insights into feature redundancy. Barriers such as the complexity of feature interactions and the absence of a formalized approach to measure functional equivalence have hindered progress. Our approach differs by expanding the concept of feature similarity to include functional equivalence, providing a more general and actionable framework for understanding and measuring features in neural networks.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves defining functional equivalence in terms of output similarity under specific transformations. We will utilize a diverse set of datasets to evaluate the performance of our feature equivalence measures, employing metrics such as representational similarity and distance measures. The expected outcomes include a clearer understanding of feature redundancy in neural networks, the ability to identify and merge functionally equivalent features, and improved model efficiency through the application of our findings in pruning and knowledge distillation techniques."
            }
        },
        "author_data": {
            "5fb5190d-084e-4c8a-83f0-2f5d1595b0f3": {
                "pk": "5fb5190d-084e-4c8a-83f0-2f5d1595b0f3",
                "name": "Yiting Chen",
                "collaborators": [
                    "Quanshi Zhang",
                    "Xu Cheng",
                    "Jie Shi",
                    "Junchi Yan",
                    "Qitian Wu",
                    "Jie Ren",
                    "Die Zhang",
                    "Yisen Wang",
                    "Lu Chen",
                    "Xin Wang",
                    "Meng Zhou",
                    "Haotian Ma",
                    "Liyao Xiang",
                    "Runzhong Wang",
                    "Li Shen",
                    "Xiaokang Yang",
                    "Dacheng Tao",
                    "Qibing Ren",
                    "Yi Mo",
                    "Nanyang Ye",
                    "Kaican Li",
                    "Lanqing Hong",
                    "Haoyue Bai",
                    "Fengwei Zhou",
                    "Zhenguo Li",
                    "Qihan Ren",
                    "Hao Zhang"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Out-of-Distribution Detection",
                    "Adversarial Learning",
                    "Privacy Preservation"
                ],
                "publications": [
                    {
                        "title": "Energy-based Out-of-Distribution Detection for Graph Neural Networks",
                        "abstract": "Learning on graphs, where instance nodes are inter-connected, has become one of the central problems for deep learning, as relational structures are pervasive and induce data inter-dependence which hinders trivial adaptation of existing approaches that assume inputs to be i.i.d.~sampled. However, current models mostly focus on improving testing performance of in-distribution data and largely ignore the potential risk w.r.t. out-of-distribution (OOD) testing samples that may cause negative outcome if the prediction is overconfident on them. In this paper, we investigate the under-explored problem, OOD detection on graph-structured data, and identify a provably effective OOD discriminator based on an energy function directly extracted from graph neural networks trained with standard classification loss. This paves a way for a simple, powerful and efficient OOD detection model for GNN-based learning on graphs, which we call GNNSafe. It also has nice theoretical properties that guarantee an overall distinguishable margin between the detection scores for in-distribution and OOD samples, which, more critically, can be further strengthened by a learning-free energy belief propagation scheme. For comprehensive evaluation, we introduce new benchmark settings that evaluate the model for detecting OOD data from both synthetic and real distribution shifts (cross-domain graph shifts and temporal graph shifts). The results show that GNNSafe achieves up to $17.0\\%$ AUROC improvement over state-of-the-arts and it could serve as simple yet strong baselines in such an under-developed area."
                    },
                    {
                        "title": "Towards One-shot Neural Combinatorial Solvers: Theoretical and Empirical Notes on the Cardinality-Constrained Case",
                        "abstract": "One-shot non-autoregressive neural networks, different from RL-based ones, have been actively adopted for solving combinatorial optimization (CO) problems, which can be trained by the objective score in a self-supervised manner. Such methods have shown their superiority in efficiency (e.g. by parallelization) and potential for tackling predictive CO problems for decision-making under uncertainty. While the discrete constraints often become a bottleneck for gradient-based neural solvers, as currently handled in three typical ways: 1) adding a soft penalty in the objective, where a bounded violation of the constraints cannot be guaranteed, being critical to many constraint-sensitive scenarios; 2) perturbing the input to generate an approximate gradient in a black-box manner, though the constraints are exactly obeyed while the approximate gradients can hurt the performance on the objective score; 3) a compromise by developing soft algorithms whereby the output of neural networks obeys a relaxed constraint, and there can still occur an arbitrary degree of constraint-violation. Towards the ultimate goal of establishing a general framework for neural CO solver with the ability to control an arbitrarysmall degree of constraint violation, in this paper, we focus on a more achievable and common setting: the cardinality constraints, which in fact can be readily encoded by a differentiable optimal transport (OT) layer. Based on this observation, we propose OT-based cardinality constraint encoding for end-to-end CO problem learning with two variants: Sinkhorn and Gumbel-Sinkhorn, whereby their violation of the constraints can be exactly characterized and bounded by our theoretical results. On synthetic and real-world CO problem instances, our methods surpass the state-of-the-art CO network and are comparable to (if not superior to) the commercial solver Gurobi. In particular, we further showcase a case study of applying our approach to the predictive portfolio optimization task on real-world asset price data, improving the Sharpe ratio from 1.1 to 2.0 of a strong LSTM+Gurobi baseline under the classic predict-then-optimize paradigm."
                    },
                    {
                        "title": "DICE: Domain-attack Invariant Causal Learning for Improved Data Privacy Protection and Adversarial Robustness",
                        "abstract": "The adversarial attack reveals the vulnerability of deep models by incurring test domain shift, while delusive attack relieves the privacy concern about personal data by injecting malicious noise into the training domain to make data unexploitable. However, beyond their successful applications, the two attacks can be easily defended by adversarial training (AT). While AT is not the panacea, it suffers from poor generalization for robustness. For the limitations of attack and defense, we argue that to fit data well, DNNs can learn the spurious relations between inputs and outputs, which are consequently utilized by the attack and defense and degrade their effectiveness, and DNNs can not easily capture the causal relations like humans to make robust decisions under attacks. In this paper, to better understand and improve attack and defense, we first take a bottom-up perspective to describe the correlations between latent factors and observed data, then analyze the effect of domain shift on DNNs induced by attack and finally develop our causal graph, namely Domain-attack Invariant Causal Model (DICM). Based on DICM, we propose a coherent causal invariant principle, which guides our algorithm design to infer the human-like causal relations. We call our algorithm Domain-attack Invariant Causal Learning (DICE) and the experimental results on two attacks and one defense task verify its effectiveness."
                    },
                    {
                        "title": "Game-theoretic Understanding of Adversarially Learned Features",
                        "abstract": "This paper aims to understand adversarial attacks and defense from a new perspecitve, i"
                    },
                    {
                        "title": "A Unified Game-Theoretic Interpretation of Adversarial Robustness",
                        "abstract": "This paper provides a unified view to explain different adversarial attacks and defense methods, i.e. the view of multi-order interactions between input variables of DNNs. Based on the multi-order interaction, we discover that adversarial attacks mainly affect high-order interactions to fool the DNN. Furthermore, we find that the robustness of adversarially trained DNNs comes from category-specific low-order interactions. Our findings provide a potential method to unify adversarial perturbations and robustness, which can explain the existing defense methods in a principle way. Besides, our findings also make a revision of previous inaccurate understanding of the shape bias of adversarially learned features."
                    },
                    {
                        "title": "OoD-Bench: Benchmarking and Understanding Out-of-Distribution Generalization Datasets and Algorithms",
                        "abstract": "Deep learning has achieved tremendous success with independent and identically distributed (i.i.d.) data. However, the performance of neural networks often degenerates drastically when encountering out-of-distribution (OoD) data, i.e., training and test data are sampled from different distributions. While a plethora of algorithms has been proposed to deal with OoD generalization, our understanding of the data used to train and evaluate these algorithms remains stagnant. In this work, we position existing datasets and algorithms from various research areas (e.g., domain generalization, stable learning, invariant risk minimization) seemingly unconnected into the same coherent picture. First, we identify and measure two distinct kinds of distribution shifts that are ubiquitous in various datasets. Next, we compare various OoD generalization algorithms with a new benchmark dominated by the two distribution shifts. Through extensive experiments, we show that existing OoD algorithms that outperform empirical risk minimization on one distribution shift usually have limitations on the other distribution shift. The new benchmark may serve as a strong foothold that can be resorted to by future OoD generalization research."
                    },
                    {
                        "title": "Rotation-Equivariant Neural Networks for Privacy Protection",
                        "abstract": "In order to prevent leaking input information from intermediate-layer features, this paper proposes a method to revise the traditional neural network into the rotation-equivariant neural network (RENN). Compared to the traditional neural network, the RENN uses d-ary vectors/tensors as features, in which each element is a d-ary number. These d-ary features can be rotated (analogous to the rotation of a d-dimensional vector) with a random angle as the encryption process. Input information is hidden in this target phase of d-ary features for attribute obfuscation. Even if attackers have obtained network parameters and intermediate-layer features, they cannot extract input information without knowing the target phase. Hence, the input privacy can be effectively protected by the RENN. Besides, the output accuracy of RENNs only degrades mildly compared to traditional neural networks, and the computational cost is significantly less than the homomorphic encryption."
                    },
                    {
                        "title": "Game-Theoretic Interactions of Different Orders",
                        "abstract": "In this study, we define interaction components of different orders between two input variables based on game theory. We further prove that interaction components of different orders satisfy several desirable properties."
                    },
                    {
                        "title": "Deep Quaternion Features for Privacy Protection",
                        "abstract": "We propose a method to revise the neural network to construct the quaternion-valued neural network (QNN), in order to prevent intermediate-layer features from leaking input information. The QNN uses quaternion-valued features, where each element is a quaternion. The QNN hides input information into a random phase of quaternion-valued features. Even if attackers have obtained network parameters and intermediate-layer features, they cannot extract input information without knowing the target phase. In this way, the QNN can effectively protect the input privacy. Besides, the output accuracy of QNNs only degrades mildly compared to traditional neural networks, and the computational cost is much less than other privacy-preserving methods."
                    }
                ]
            },
            "e9e43ecf-e7a7-4866-ba27-aaaad8401d13": {
                "pk": "e9e43ecf-e7a7-4866-ba27-aaaad8401d13",
                "name": "Zhanpeng Zhou",
                "collaborators": [
                    "Junchi Yan",
                    "Quanshi Zhang",
                    "Zijun Chen",
                    "Yilan Chen",
                    "Bo Zhang",
                    "Wen Shen",
                    "Ling Tang",
                    "J. Ren",
                    "Qirui Chen",
                    "Yongyi Yang",
                    "Xiaojiang Yang",
                    "Wei Hu",
                    "Huixin Chen",
                    "YueFeng Chen"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Causality",
                    "Model Interpretability"
                ],
                "publications": [
                    {
                        "title": "Cross-Task Linearity Emerges in the Pretraining-Finetuning Paradigm",
                        "abstract": "The pretraining-finetuning paradigm has become the prevailing trend in modern deep learning. In this work, we discover an intriguing linear phenomenon in models that are initialized from a common pretrained checkpoint and finetuned on different tasks, termed as Cross-Task Linearity (CTL) . Specifically, if we linearly interpolate the weights of two finetuned models, the features in the weight-interpolated model are approximately equal to the linear interpolation of features in two finetuned models at each layer. Such cross-task linearity has not been noted in peer literature. We provide comprehensive empirical evidence supporting that CTL consistently occurs for finetuned models that start from the same pretrained checkpoint. We conjecture that in the pretraining-finetuning paradigm, neural networks essentially function as linear maps, mapping from the parameter space to the feature space. Based on this viewpoint, our study unveils novel insights into explaining model merging/editing, particularly by translating operations from the parameter space to the feature space. Furthermore, we delve deeper into the underlying factors for the emergence of CTL, emphasizing the impact of pretraining."
                    },
                    {
                        "title": "On the Emergence of Cross-Task Linearity in Pretraining-Finetuning Paradigm",
                        "abstract": "The pretraining-finetuning paradigm has become the prevailing trend in modern deep learning. In this work, we discover an intriguing linear phenomenon in models that are initialized from a common pretrained checkpoint and finetuned on different tasks, termed as Cross-Task Linearity (CTL). Specifically, we show that if we linearly interpolate the weights of two finetuned models, the features in the weight-interpolated model are often approximately equal to the linear interpolation of features in two finetuned models at each layer. We provide comprehensive empirical evidence supporting that CTL consistently occurs for finetuned models that start from the same pretrained checkpoint. We conjecture that in the pretraining-finetuning paradigm, neural networks approximately function as linear maps, mapping from the parameter space to the feature space. Based on this viewpoint, our study unveils novel insights into explaining model merging/editing, particularly by translating operations from the parameter space to the feature space. Furthermore, we delve deeper into the root cause for the emergence of CTL, highlighting the role of pretraining."
                    },
                    {
                        "title": "Can We Faithfully Represent Absence States to Compute Shapley Values on a DNN?",
                        "abstract": "Although many methods have been proposed to estimate attributions of input variables, there still exists a significant theoretical flaw in masking-based attribution methods, i.e. it is hard to examine whether the masking method faithfully represents the absence of input variables. Specifically, for masking-based attributions, setting an input variable to the baseline value is a typical way of representing the absence of the variable. However, there are no studies investigating how to represent the absence of input variables and verify the faithfulness of baseline values. Therefore, we revisit the feature representation of a DNN in terms of causality, and propose to use causal patterns to examine whether the masking method faithfully removes information encoded in input variables. More crucially, it is proven that the causality can be explained as the elementary rationale of the Shapley value. Furthermore, we define the optimal baseline value from the perspective of causality, and we propose a method to learn the optimal baseline value. Experimental results have demonstrated the effectiveness of our method."
                    },
                    {
                        "title": "Going Beyond Linear Mode Connectivity: The Layerwise Linear Feature Connectivity",
                        "abstract": "Recent work has revealed many intriguing empirical phenomena in neural network training, despite the poorly understood and highly complex loss landscapes and training dynamics. One of these phenomena, Linear Mode Connectivity (LMC), has gained considerable attention due to the intriguing observation that different solutions can be connected by a linear path in the parameter space while maintaining near-constant training and test losses. In this work, we introduce a stronger notion of linear connectivity, Layerwise Linear Feature Connectivity (LLFC), which says that the feature maps of every layer in different trained networks are also linearly connected. We provide comprehensive empirical evidence for LLFC across a wide range of settings, demonstrating that whenever two trained networks satisfy LMC (via either spawning or permutation methods), they also satisfy LLFC in nearly all the layers. Furthermore, we delve deeper into the underlying factors contributing to LLFC, which reveal new insights into the spawning and permutation approaches. The study of LLFC transcends and advances our understanding of LMC by adopting a feature-learning perspective."
                    },
                    {
                        "title": "Batch Normalization Is Blind to the First and Second Derivatives of the Loss",
                        "abstract": "We prove that when we do the Taylor series expansion of the loss function, the BN operation will block the influence of the first-order term and most influence of the second-order term of the loss. We also find that such a problem is caused by the standardization phase of the BN operation. We believe that proving the blocking of certain loss terms provides an analytic perspective for potential detects of a deep model with BN operations, although the blocking problem is not fully equivalent to significant damages in all tasks on benchmark datasets. Experiments show that the BN operation significantly affects feature representations in specific tasks."
                    },
                    {
                        "title": "Defects of Convolutional Decoder Networks in Frequency Representation",
                        "abstract": "In this paper, we prove the representation defects of a cascaded convolutional decoder network, considering the capacity of representing different frequency components of an input sample. We conduct the discrete Fourier transform on each channel of the feature map in an intermediate layer of the decoder network. Then, we extend the 2D circular convolution theorem to represent the forward and backward propagations through convolutional layers in the frequency domain. Based on this, we prove three defects in representing feature spectrums. First, we prove that the convolution operation, the zero-padding operation, and a set of other settings all make a convolutional decoder network more likely to weaken high-frequency components. Second, we prove that the upsampling operation generates a feature spectrum, in which strong signals repetitively appear at certain frequencies. Third, we prove that if the frequency components in the input sample and frequency components in the target output for regression have a small shift, then the decoder usually cannot be effectively learned."
                    }
                ]
            },
            "98e15d05-5716-454a-9b30-7396a0c727fc": {
                "pk": "98e15d05-5716-454a-9b30-7396a0c727fc",
                "name": "Junchi Yan",
                "collaborators": [
                    "Zhanpeng Zhou",
                    "Bo Zhang",
                    "Zijun Chen",
                    "Yilan Chen",
                    "Mingze Wang",
                    "Yuchen Mao",
                    "Bingrui Li",
                    "Xiangchao Yan",
                    "Runjian Chen",
                    "Hancheng Ye",
                    "Renqiu Xia",
                    "Jiakang Yuan",
                    "Hongbin Zhou",
                    "Xinyu Cai",
                    "Botian Shi",
                    "Wenqi Shao",
                    "Ping Luo",
                    "Yu Qiao",
                    "Tao Chen"
                ],
                "domain": [
                    "Deep Learning",
                    "Transfer Learning",
                    "Computer Vision",
                    "3D Point Cloud"
                ],
                "publications": [
                    {
                        "title": "Cross-Task Linearity Emerges in the Pretraining-Finetuning Paradigm",
                        "abstract": "The pretraining-finetuning paradigm has become the prevailing trend in modern deep learning. In this work, we discover an intriguing linear phenomenon in models that are initialized from a common pretrained checkpoint and finetuned on different tasks, termed as Cross-Task Linearity (CTL) . Specifically, if we linearly interpolate the weights of two finetuned models, the features in the weight-interpolated model are approximately equal to the linear interpolation of features in two finetuned models at each layer. Such cross-task linearity has not been noted in peer literature. We provide comprehensive empirical evidence supporting that CTL consistently occurs for finetuned models that start from the same pretrained checkpoint. We conjecture that in the pretraining-finetuning paradigm, neural networks essentially function as linear maps, mapping from the parameter space to the feature space. Based on this viewpoint, our study unveils novel insights into explaining model merging/editing, particularly by translating operations from the parameter space to the feature space. Furthermore, we delve deeper into the underlying factors for the emergence of CTL, emphasizing the impact of pretraining."
                    },
                    {
                        "title": "Sharpness-Aware Minimization Efficiently Selects Flatter Minima Late in Training",
                        "abstract": "Sharpness-Aware Minimization (SAM) has substantially improved the generalization of neural networks under various settings. Despite the success, its effectiveness remains poorly understood. In this work, we discover an intriguing phenomenon in the training dynamics of SAM, shedding lights on understanding its implicit bias towards flatter minima over Stochastic Gradient Descent (SGD). Specifically, we find that SAM efficiently selects flatter minima late in training. Remarkably, even a few epochs of SAM applied at the end of training yield nearly the same generalization and solution sharpness as full SAM training. Subsequently, we delve deeper into the underlying mechanism behind this phenomenon. Theoretically, we identify two phases in the learning dynamics after applying SAM late in training: i) SAM first escapes the minimum found by SGD exponentially fast; and ii) then rapidly converges to a flatter minimum within the same valley. Furthermore, we empirically investigate the role of SAM during the early training phase. We conjecture that the optimization method chosen in the late phase is more crucial in shaping the final solution's properties. Based on this viewpoint, we extend our findings from SAM to Adversarial Training."
                    },
                    {
                        "title": "On the Emergence of Cross-Task Linearity in Pretraining-Finetuning Paradigm",
                        "abstract": "The pretraining-finetuning paradigm has become the prevailing trend in modern deep learning. In this work, we discover an intriguing linear phenomenon in models that are initialized from a common pretrained checkpoint and finetuned on different tasks, termed as Cross-Task Linearity (CTL). Specifically, we show that if we linearly interpolate the weights of two finetuned models, the features in the weight-interpolated model are often approximately equal to the linear interpolation of features in two finetuned models at each layer. We provide comprehensive empirical evidence supporting that CTL consistently occurs for finetuned models that start from the same pretrained checkpoint. We conjecture that in the pretraining-finetuning paradigm, neural networks approximately function as linear maps, mapping from the parameter space to the feature space. Based on this viewpoint, our study unveils novel insights into explaining model merging/editing, particularly by translating operations from the parameter space to the feature space. Furthermore, we delve deeper into the root cause for the emergence of CTL, highlighting the role of pretraining."
                    },
                    {
                        "title": "SPOT: Scalable 3D Pre-training via Occupancy Prediction for Learning Transferable 3D Representations",
                        "abstract": "Annotating 3D LiDAR point clouds for perception tasks is fundamental for many applications e.g., autonomous driving, yet it still remains notoriously labor-intensive. Pretraining-finetuning approach can alleviate the labeling burden by fine-tuning a pre-trained backbone across various downstream datasets as well as tasks. In this paper, we propose SPOT, namely Scalable Pre-training via Occupancy prediction for learning Transferable 3D representations under such a label-efficient fine-tuning paradigm. SPOT achieves effectiveness on various public datasets with different downstream tasks, showcasing its general representation power, cross-domain robustness and data scalability which are three key factors for real-world application. Specifically, we both theoretically and empirically show, for the first time, that general representations learning can be achieved through the task of occupancy prediction. Then, to address the domain gap caused by different LiDAR sensors and annotation methods, we develop a beam re-sampling technique for point cloud augmentation combined with class-balancing strategy. Furthermore, scalable pre-training is observed, that is, the downstream performance across all the experiments gets better with more pre-training data. Additionally, such pre-training strategy also remains compatible with unlabeled data. The hope is that our findings will facilitate the understanding of LiDAR points and pave the way for future advancements in LiDAR pre-training."
                    }
                ]
            }
        }
    },
    "2311.04163": {
        "paper_data": {
            "title": "Outliers with Opposing Signals Have an Outsized Effect on Neural Network Optimization",
            "url": "http://arxiv.org/abs/2311.04163v1",
            "arxiv_id": "2311.04163",
            "authors": [
                "Elan Rosenfeld",
                "Andrej Risteski"
            ],
            "abstract": "We identify a new phenomenon in neural network optimization which arises from the interaction of depth and a particular heavy-tailed structure in natural data. Our result offers intuitive explanations for several previously reported observations about network training dynamics. In particular, it implies a conceptually new cause for progressive sharpening and the edge of stability; we also highlight connections to other concepts in optimization and generalization including grokking, simplicity bias, and Sharpness-Aware Minimization.   Experimentally, we demonstrate the significant influence of paired groups of outliers in the training data with strong opposing signals: consistent, large magnitude features which dominate the network output throughout training and provide gradients which point in opposite directions. Due to these outliers, early optimization enters a narrow valley which carefully balances the opposing groups; subsequent sharpening causes their loss to rise rapidly, oscillating between high on one group and then the other, until the overall loss spikes. We describe how to identify these groups, explore what sets them apart, and carefully study their effect on the network's optimization and behavior. We complement these experiments with a mechanistic explanation on a toy example of opposing signals and a theoretical analysis of a two-layer linear network on a simple model. Our finding enables new qualitative predictions of training behavior which we confirm experimentally. It also provides a new lens through which to study and improve modern training practices for stochastic optimization, which we highlight via a case study of Adam versus SGD.",
            "introduction": " Introduction There is a steadily growing list of intriguing properties of neural network (NN) optimization which are not readily explained by classical tools from optimization. Likewise, we have varying degrees of understanding of the mechanistic causes for each. Extensive efforts have led to possible explanations for the effectiveness of Adam (Kingma and Ba, 2014), Batch Normalization (Ioffe and Szegedy, 2015) and other tools for successful training\u2014but the evidence is not always entirely convincing, and there is certainly little theoretical understanding. Other findings, such as grokking (Power et al., 2022) or the edge of stability (Cohen et al., 2021), do not have immediate practical implications but provide new ways to study what sets NN optimization apart. These phenomena are typically considered in isolation\u2014though they are not completely disparate, it is unknown what specific underlying causes they may share. Clearly, a better understanding of NN training dynamics in a specific context can lead to algorithmic improvements (Chen et al., 2021); this suggests that any commonality will be a valuable tool for further investigation. In this work, we identify a phenomenon in NN optimization which offers a new perspective on many of these prior observations and which we hope will contribute to a deeper understanding of how they may be connected. While we do not (and do not claim to) give a complete explanation, we present strong qualitative and quantitative evidence for a single high-level idea\u2014one which naturally fits into several existing narratives and suggests a more coherent picture of their origin. 1arXiv:2311.04163v1  [cs.LG]  7 Nov 20230 60123456Background Color 110 170Lots of Red: Auto vs. Other 2750 2810Vehicular Features: Auto vs. Truck 3080 3140Grass Texture: Frog vs. Other Train Loss (Group 1) Train Loss (Group 2) Full Train Loss 0.0 0.2 0.4 0.6 0.8 1.0 Iteration0.00.20.40.60.81.0Loss 0 250 20Group 1 0 250 20Background Color 0 250 20 0 250 20Group 2 0 250 20 0 250 20 0 250 20Group 1 0 250 20Vehicular Features 0 250 20 0 250 20Group 2 0 250 20 0 250 20 0 250 20Group 1 0 250 20Lots of Red 0 250 20 0 250 20Group 2 0 250 20 0 250 20 0 250 20Group 1 0 250 20Grass Texture 0 250 20 0 250 20Group 2 0 250 20 0 250 20Figure 1: Training dynamics of neural networks are heavily influenced by outliers with opposing signals. We plot the overall loss of a ResNet-18 trained with GD on CIFAR-10, plus the losses of a small but representative set of outlier groups. These groups have consistent opposing signals (e.g., wheels and headlights can mean either car ortruck ). Throughout training, losses on these groups oscillate with growing and shrinking amplitude\u2014this has an obvious correspondence to the intermittent spikes in overall loss and appears to be a direct cause of the edge of stability phenomenon. Specifically, we demonstrate the prevalence of paired groups of outliers in natural data which have a significant influence on a network\u2019s optimization dynamics. These groups are characterized by the inclusion of one or more (relatively) large magnitude features that dominate the network\u2019s output at initialization and throughout most of training. In addition to their magnitude, the other distinctive property of these features is that they provide large, consistent, and opposing gradients, in that following one group\u2019s gradient to decrease its loss will increase the other\u2019s by a similar amount. Because of this structure, we refer to them as Opposing Signals . These features share a non-trivial correlation with the target task, but they are often not the \u201ccorrect\u201d (e.g., human-aligned) signal. In fact, in many cases these features perfectly encapsulate",
            "references": [
                {
                    "title": "Predicting Grokking Long Before it Happens: A look into the loss landscape of models which grok",
                    "abstract": "This paper focuses on predicting the occurrence of grokking in neural networks, a phenomenon in which perfect generalization emerges long after signs of overfitting or memorization are observed. It has been reported that grokking can only be observed with certain hyper-parameters. This makes it critical to identify the parameters that lead to grokking. However, since grokking occurs after a large number of epochs, searching for the hyper-parameters that lead to it is time-consuming. In this paper, we propose a low-cost method to predict grokking without training for a large number of epochs. In essence, by studying the learning curve of the first few epochs, we show that one can predict whether grokking will occur later on. Specifically, if certain oscillations occur in the early epochs, one can expect grokking to occur if the model is trained for a much longer period of time. We propose using the spectral signature of a learning curve derived by applying the Fourier transform to quantify the amplitude of low-frequency components to detect the presence of such oscillations. We also present additional experiments aimed at explaining the cause of these oscillations and characterizing the loss landscape."
                },
                {
                    "title": "Toward Understanding Why Adam Converges Faster Than SGD for Transformers",
                    "abstract": "While stochastic gradient descent (SGD) is still the most popular optimization algorithm in deep learning, adaptive algorithms such as Adam have established empirical advantages over SGD in some deep learning applications such as training transformers. However, it remains a question that why Adam converges significantly faster than SGD in these scenarios. In this paper, we propose one explanation of why Adam converges faster than SGD using a new concept directional sharpness. We argue that the performance of optimization algorithms is closely related to the directional sharpness of the update steps, and show SGD has much worse directional sharpness compared to adaptive algorithms. We further observe that only a small fraction of the coordinates causes the bad sharpness and slow convergence of SGD, and propose to use coordinate-wise clipping as a solution to SGD and other optimization algorithms. We demonstrate the effect of coordinate-wise clipping on sharpness reduction and speeding up the convergence of optimization algorithms under various settings. We show that coordinate-wise clipping improves the local loss reduction when only a small fraction of the coordinates has bad sharpness. We conclude that the sharpness reduction effect of adaptive coordinate-wise scaling is the reason for Adam's success in practice and suggest the use of coordinate-wise clipping as a universal technique to speed up deep learning optimization."
                },
                {
                    "title": "On progressive sharpening, flat minima and generalisation",
                    "abstract": "We present a new approach to understanding the relationship between loss curvature and input-output model behaviour in deep learning. Specifically, we use existing empirical analyses of the spectrum of deep network loss Hessians to ground an ansatz tying together the loss Hessian and the input-output Jacobian over training samples during the training of deep neural networks. We then prove a series of theoretical results which quantify the degree to which the input-output Jacobian of a model approximates its Lipschitz norm over a data distribution, and deduce a novel generalisation bound in terms of the empirical Jacobian. We use our ansatz, together with our theoretical results, to give a new account of the recently observed progressive sharpening phenomenon, as well as the generalisation properties of flat minima. Experimental evidence is provided to validate our claims."
                },
                {
                    "title": "A Tale of Two Circuits: Grokking as Competition of Sparse and Dense Subnetworks",
                    "abstract": "Grokking is a phenomenon where a model trained on an algorithmic task first overfits but, then, after a large amount of additional training, undergoes a phase transition to generalize perfectly. We empirically study the internal structure of networks undergoing grokking on the sparse parity task, and find that the grokking phase transition corresponds to the emergence of a sparse subnetwork that dominates model predictions. On an optimization level, we find that this subnetwork arises when a small subset of neurons undergoes rapid norm growth, whereas the other neurons in the network decay slowly in norm. Thus, we suggest that the grokking phase transition can be understood to emerge from competition of two largely distinct subnetworks: a dense one that dominates before the transition and generalizes poorly, and a sparse one that dominates afterwards."
                },
                {
                    "title": "Simfluence: Modeling the Influence of Individual Training Examples by Simulating Training Runs",
                    "abstract": "Training data attribution (TDA) methods offer to trace a model's prediction on any given example back to specific influential training examples. Existing approaches do so by assigning a scalar influence score to each training example, under a simplifying assumption that influence is additive. But in reality, we observe that training examples interact in highly non-additive ways due to factors such as inter-example redundancy, training order, and curriculum learning effects. To study such interactions, we propose Simfluence, a new paradigm for TDA where the goal is not to produce a single influence score per example, but instead a training run simulator: the user asks, ``If my model had trained on example $z_1$, then $z_2$, ..., then $z_n$, how would it behave on $z_{test}$?''; the simulator should then output a simulated training run, which is a time series predicting the loss on $z_{test}$ at every step of the simulated run. This enables users to answer counterfactual questions about what their model would have learned under different training curricula, and to directly see where in training that learning would occur. We present a simulator, Simfluence-Linear, that captures non-additive interactions and is often able to predict the spiky trajectory of individual example losses with surprising fidelity. Furthermore, we show that existing TDA methods such as TracIn and influence functions can be viewed as special cases of Simfluence-Linear. This enables us to directly compare methods in terms of their simulation accuracy, subsuming several prior TDA approaches to evaluation. In experiments on large language model (LLM) fine-tuning, we show that our method predicts loss trajectories with much higher accuracy than existing TDA methods (doubling Spearman's correlation and reducing mean-squared error by 75%) across several tasks, models, and training methods."
                },
                {
                    "title": "Symbolic Discovery of Optimization Algorithms",
                    "abstract": "We present a method to formulate algorithm discovery as program search, and apply it to discover optimization algorithms for deep neural network training. We leverage efficient search techniques to explore an infinite and sparse program space. To bridge the large generalization gap between proxy and target tasks, we also introduce program selection and simplification strategies. Our method discovers a simple and effective optimization algorithm, $\\textbf{Lion}$ ($\\textit{Evo$\\textbf{L}$ved S$\\textbf{i}$gn M$\\textbf{o}$me$\\textbf{n}$tum}$). It is more memory-efficient than Adam as it only keeps track of the momentum. Different from adaptive optimizers, its update has the same magnitude for each parameter calculated through the sign operation. We compare Lion with widely used optimizers, such as Adam and Adafactor, for training a variety of models on different tasks. On image classification, Lion boosts the accuracy of ViT by up to 2% on ImageNet and saves up to 5x the pre-training compute on JFT. On vision-language contrastive learning, we achieve 88.3% $\\textit{zero-shot}$ and 91.1% $\\textit{fine-tuning}$ accuracy on ImageNet, surpassing the previous best results by 2% and 0.1%, respectively. On diffusion models, Lion outperforms Adam by achieving a better FID score and reducing the training compute by up to 2.3x. For autoregressive, masked language modeling, and fine-tuning, Lion exhibits a similar or better performance compared to Adam. Our analysis of Lion reveals that its performance gain grows with the training batch size. It also requires a smaller learning rate than Adam due to the larger norm of the update produced by the sign function. Additionally, we examine the limitations of Lion and identify scenarios where its improvements are small or not statistically significant. Lion is also successfully deployed in production systems such as Google search ads CTR model."
                },
                {
                    "title": "Progress measures for grokking via mechanistic interpretability",
                    "abstract": "Neural networks often exhibit emergent behavior, where qualitatively new capabilities arise from scaling up the amount of parameters, training data, or training steps. One approach to understanding emergence is to find continuous \\textit{progress measures} that underlie the seemingly discontinuous qualitative changes. We argue that progress measures can be found via mechanistic interpretability: reverse-engineering learned behaviors into their individual components. As a case study, we investigate the recently-discovered phenomenon of ``grokking'' exhibited by small transformers trained on modular addition tasks. We fully reverse engineer the algorithm learned by these networks, which uses discrete Fourier transforms and trigonometric identities to convert addition to rotation about a circle. We confirm the algorithm by analyzing the activations and weights and by performing ablations in Fourier space. Based on this understanding, we define progress measures that allow us to study the dynamics of training and split training into three continuous phases: memorization, circuit formation, and cleanup. Our results show that grokking, rather than being a sudden shift, arises from the gradual amplification of structured mechanisms encoded in the weights, followed by the later removal of memorizing components."
                },
                {
                    "title": "Understanding Edge-of-Stability Training Dynamics with a Minimalist Example",
                    "abstract": "Recently, researchers observed that gradient descent for deep neural networks operates in an ``edge-of-stability'' (EoS) regime: the sharpness (maximum eigenvalue of the Hessian) is often larger than stability threshold $2/\\eta$ (where $\\eta$ is the step size). Despite this, the loss oscillates and converges in the long run, and the sharpness at the end is just slightly below $2/\\eta$. While many other well-understood nonconvex objectives such as matrix factorization or two-layer networks can also converge despite large sharpness, there is often a larger gap between sharpness of the endpoint and $2/\\eta$. In this paper, we study EoS phenomenon by constructing a simple function that has the same behavior. We give rigorous analysis for its training dynamics in a large local region and explain why the final converging point has sharpness close to $2/\\eta$. Globally we observe that the training dynamics for our example has an interesting bifurcating behavior, which was also observed in the training of neural nets."
                },
                {
                    "title": "Self-Stabilization: The Implicit Bias of Gradient Descent at the Edge of Stability",
                    "abstract": "Traditional analyses of gradient descent show that when the largest eigenvalue of the Hessian, also known as the sharpness $S(\\theta)$, is bounded by $2/\\eta$, training is\"stable\"and the training loss decreases monotonically. Recent works, however, have observed that this assumption does not hold when training modern neural networks with full batch or large batch gradient descent. Most recently, Cohen et al. (2021) observed two important phenomena. The first, dubbed progressive sharpening, is that the sharpness steadily increases throughout training until it reaches the instability cutoff $2/\\eta$. The second, dubbed edge of stability, is that the sharpness hovers at $2/\\eta$ for the remainder of training while the loss continues decreasing, albeit non-monotonically. We demonstrate that, far from being chaotic, the dynamics of gradient descent at the edge of stability can be captured by a cubic Taylor expansion: as the iterates diverge in direction of the top eigenvector of the Hessian due to instability, the cubic term in the local Taylor expansion of the loss function causes the curvature to decrease until stability is restored. This property, which we call self-stabilization, is a general property of gradient descent and explains its behavior at the edge of stability. A key consequence of self-stabilization is that gradient descent at the edge of stability implicitly follows projected gradient descent (PGD) under the constraint $S(\\theta) \\le 2/\\eta$. Our analysis provides precise predictions for the loss, sharpness, and deviation from the PGD trajectory throughout training, which we verify both empirically in a number of standard settings and theoretically under mild conditions. Our analysis uncovers the mechanism for gradient descent's implicit bias towards stability."
                },
                {
                    "title": "Adaptive Gradient Methods at the Edge of Stability",
                    "abstract": "Very little is known about the training dynamics of adaptive gradient methods like Adam in deep learning. In this paper, we shed light on the behavior of these algorithms in the full-batch and sufficiently large batch settings. Specifically, we empirically demonstrate that during full-batch training, the maximum eigenvalue of the preconditioned Hessian typically equilibrates at a certain numerical value -- the stability threshold of a gradient descent algorithm. For Adam with step size $\\eta$ and $\\beta_1 = 0.9$, this stability threshold is $38/\\eta$. Similar effects occur during minibatch training, especially as the batch size grows. Yet, even though adaptive methods train at the ``Adaptive Edge of Stability'' (AEoS), their behavior in this regime differs in a significant way from that of non-adaptive methods at the EoS. Whereas non-adaptive algorithms at the EoS are blocked from entering high-curvature regions of the loss landscape, adaptive gradient methods at the AEoS can keep advancing into high-curvature regions, while adapting the preconditioner to compensate. Our findings can serve as a foundation for the community's future understanding of adaptive gradient methods in deep learning."
                },
                {
                    "title": "Analyzing Sharpness along GD Trajectory: Progressive Sharpening and Edge of Stability",
                    "abstract": "Recent findings (e.g., arXiv:2103.00065) demonstrate that modern neural networks trained by full-batch gradient descent typically enter a regime called Edge of Stability (EOS). In this regime, the sharpness, i.e., the maximum Hessian eigenvalue, first increases to the value 2/(step size) (the progressive sharpening phase) and then oscillates around this value (the EOS phase). This paper aims to analyze the GD dynamics and the sharpness along the optimization trajectory. Our analysis naturally divides the GD trajectory into four phases depending on the change of the sharpness. We empirically identify the norm of output layer weight as an interesting indicator of sharpness dynamics. Based on this empirical observation, we attempt to theoretically and empirically explain the dynamics of various key quantities that lead to the change of sharpness in each phase of EOS. Moreover, based on certain assumptions, we provide a theoretical proof of the sharpness behavior in EOS regime in two-layer fully-connected linear neural networks. We also discuss some other empirical findings and the limitation of our theoretical results."
                },
                {
                    "title": "Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limit",
                    "abstract": "There is mounting evidence of emergent phenomena in the capabilities of deep learning methods as we scale up datasets, model sizes, and training times. While there are some accounts of how these resources modulate statistical capacity, far less is known about their effect on the computational problem of model training. This work conducts such an exploration through the lens of learning a $k$-sparse parity of $n$ bits, a canonical discrete search problem which is statistically easy but computationally hard. Empirically, we find that a variety of neural networks successfully learn sparse parities, with discontinuous phase transitions in the training curves. On small instances, learning abruptly occurs at approximately $n^{O(k)}$ iterations; this nearly matches SQ lower bounds, despite the apparent lack of a sparse prior. Our theoretical analysis shows that these observations are not explained by a Langevin-like mechanism, whereby SGD\"stumbles in the dark\"until it finds the hidden set of features (a natural algorithm which also runs in $n^{O(k)}$ time). Instead, we show that SGD gradually amplifies the sparse solution via a Fourier gap in the population gradient, making continual progress that is invisible to loss and error metrics."
                },
                {
                    "title": "The alignment property of SGD noise and how it helps select flat minima: A stability analysis",
                    "abstract": "The phenomenon that stochastic gradient descent (SGD) favors flat minima has played a critical role in understanding the implicit regularization of SGD. In this paper, we provide an explanation of this striking phenomenon by relating the particular noise structure of SGD to its \\emph{linear stability} (Wu et al., 2018). Specifically, we consider training over-parameterized models with square loss. We prove that if a global minimum $\\theta^*$ is linearly stable for SGD, then it must satisfy $\\|H(\\theta^*)\\|_F\\leq O(\\sqrt{B}/\\eta)$, where $\\|H(\\theta^*)\\|_F, B,\\eta$ denote the Frobenius norm of Hessian at $\\theta^*$, batch size, and learning rate, respectively. Otherwise, SGD will escape from that minimum \\emph{exponentially} fast. Hence, for minima accessible to SGD, the sharpness -- as measured by the Frobenius norm of the Hessian -- is bounded \\emph{independently} of the model size and sample size. The key to obtaining these results is exploiting the particular structure of SGD noise: The noise concentrates in sharp directions of local landscape and the magnitude is proportional to loss value. This alignment property of SGD noise provably holds for linear networks and random feature models (RFMs), and is empirically verified for nonlinear networks. Moreover, the validity and practical relevance of our theoretical findings are also justified by extensive experiments on CIFAR-10 dataset."
                },
                {
                    "title": "The Slingshot Mechanism: An Empirical Study of Adaptive Optimizers and the Grokking Phenomenon",
                    "abstract": "The grokking phenomenon as reported by Power et al. ( arXiv:2201.02177 ) refers to a regime where a long period of overfitting is followed by a seemingly sudden transition to perfect generalization. In this paper, we attempt to reveal the underpinnings of Grokking via a series of empirical studies. Specifically, we uncover an optimization anomaly plaguing adaptive optimizers at extremely late stages of training, referred to as the Slingshot Mechanism. A prominent artifact of the Slingshot Mechanism can be measured by the cyclic phase transitions between stable and unstable training regimes, and can be easily monitored by the cyclic behavior of the norm of the last layers weights. We empirically observe that without explicit regularization, Grokking as reported in ( arXiv:2201.02177 ) almost exclusively happens at the onset of Slingshots, and is absent without it. While common and easily reproduced in more general settings, the Slingshot Mechanism does not follow from any known optimization theories that we are aware of, and can be easily overlooked without an in depth examination. Our work points to a surprising and useful inductive bias of adaptive gradient optimizers at late stages of training, calling for a revised theoretical analysis of their origin."
                },
                {
                    "title": "Understanding Gradient Descent on Edge of Stability in Deep Learning",
                    "abstract": "Deep learning experiments by Cohen et al. [2021] using deterministic Gradient Descent (GD) revealed an Edge of Stability (EoS) phase when learning rate (LR) and sharpness (i.e., the largest eigenvalue of Hessian) no longer behave as in traditional optimization. Sharpness stabilizes around $2/$LR and loss goes up and down across iterations, yet still with an overall downward trend. The current paper mathematically analyzes a new mechanism of implicit regularization in the EoS phase, whereby GD updates due to non-smooth loss landscape turn out to evolve along some deterministic flow on the manifold of minimum loss. This is in contrast to many previous results about implicit bias either relying on infinitesimal updates or noise in gradient. Formally, for any smooth function $L$ with certain regularity condition, this effect is demonstrated for (1) Normalized GD, i.e., GD with a varying LR $\\eta_t =\\frac{\\eta}{\\| \\nabla L(x(t)) \\|}$ and loss $L$; (2) GD with constant LR and loss $\\sqrt{L- \\min_x L(x)}$. Both provably enter the Edge of Stability, with the associated flow on the manifold minimizing $\\lambda_{1}(\\nabla^2 L)$. The above theoretical results have been corroborated by an experimental study."
                },
                {
                    "title": "Beyond the Quadratic Approximation: The Multiscale Structure of Neural Network Loss Landscapes",
                    "abstract": "A quadratic approximation of neural network loss landscapes has been extensively used to study the optimization process of these networks. Though, it usually holds in a very small neighborhood of the minimum, it cannot explain many phenomena observed during the optimization process. In this work, we study the structure of neural network loss functions and its implication on optimization in a region beyond the reach of a good quadratic approximation. Numerically, we observe that neural network loss functions possesses a multiscale structure, manifested in two ways: (1) in a neighborhood of minima, the loss mixes a continuum of scales and grows subquadratically, and (2) in a larger region, the loss shows several separate scales clearly. Using the subquadratic growth, we are able to explain the Edge of Stability phenomenon [5] observed for the gradient descent (GD) method. Using the separate scales, we explain the working mechanism of learning rate decay by simple examples. Finally, we study the origin of the multiscale structure and propose that the non-convexity of the models and the non-uniformity of training data is one of the causes. By constructing a two-layer neural network problem we show that training data with different magnitudes give rise to different scales of the loss function, producing subquadratic growth and multiple separate scales."
                },
                {
                    "title": "Gradient Descent on Neurons and its Link to Approximate Second-Order Optimization",
                    "abstract": "Second-order optimizers are thought to hold the potential to speed up neural network training, but due to the enormous size of the curvature matrix, they typically require approximations to be computationally tractable. The most successful family of approximations are Kronecker-Factored, block-diagonal curvature estimates (KFAC). Here, we combine tools from prior work to evaluate exact second-order updates with careful ablations to establish a surprising result: Due to its approximations, KFAC is not closely related to second-order updates, and in particular, it significantly outperforms true second-order updates. This challenges widely held believes and immediately raises the question why KFAC performs so well. Towards answering this question we present evidence strongly suggesting that KFAC approximates a first-order algorithm, which performs gradient descent on neurons rather than weights. Finally, we show that this optimizer often improves over KFAC in terms of computational cost and data-efficiency."
                },
                {
                    "title": "Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets",
                    "abstract": "In this paper we propose to study generalization of neural networks on small algorithmically generated datasets. In this setting, questions about data efficiency, memorization, generalization, and speed of learning can be studied in great detail. In some situations we show that neural networks learn through a process of\"grokking\"a pattern in the data, improving generalization performance from random chance level to perfect generalization, and that this improvement in generalization can happen well past the point of overfitting. We also study generalization as a function of dataset size and find that smaller datasets require increasing amounts of optimization for generalization. We argue that these datasets provide a fertile ground for studying a poorly understood aspect of deep learning: generalization of overparametrized neural networks beyond memorization of the finite training dataset."
                },
                {
                    "title": "On Linear Stability of SGD and Input-Smoothness of Neural Networks",
                    "abstract": "The multiplicative structure of parameters and input data in the first layer of neural networks is explored to build connection between the landscape of the loss function with respect to parameters and the landscape of the model function with respect to input data. By this connection, it is shown that flat minima regularize the gradient of the model function, which explains the good generalization performance of flat minima. Then, we go beyond the flatness and consider high-order moments of the gradient noise, and show that Stochastic Gradient Descent (SGD) tends to impose constraints on these moments by a linear stability analysis of SGD around global minima. Together with the multiplicative structure, we identify the Sobolev regularization effect of SGD, i.e. SGD regularizes the Sobolev seminorms of the model function with respect to the input data. Finally, bounds for generalization error and adversarial robustness are provided for solutions found by SGD under assumptions of the data distribution."
                },
                {
                    "title": "The Low-Rank Simplicity Bias in Deep Networks",
                    "abstract": "Modern deep neural networks are highly over-parameterized compared to the data on which they are trained, yet they often generalize remarkably well. A flurry of recent work has asked: why do deep networks not overfit to their training data? In this work, we make a series of empirical observations that investigate and extend the hypothesis that deeper networks are inductively biased to find solutions with lower effective rank embeddings. We conjecture that this bias exists because the volume of functions that maps to low effective rank embedding increases with depth. We show empirically that our claim holds true on finite width linear and non-linear models on practical learning paradigms and show that on natural data, these are often the solutions that generalize well. We then show that the simplicity bias exists at both initialization and after training and is resilient to hyper-parameters and learning methods. We further demonstrate how linear over-parameterization of deep non-linear models can be used to induce low-rank bias, improving generalization performance on CIFAR and ImageNet without changing the modeling capacity."
                },
                {
                    "title": "Gradient Descent on Neural Networks Typically Occurs at the Edge of Stability",
                    "abstract": "We empirically demonstrate that full-batch gradient descent on neural network training objectives typically operates in a regime we call the Edge of Stability. In this regime, the maximum eigenvalue of the training loss Hessian hovers just above the numerical value $2 / \\text{(step size)}$, and the training loss behaves non-monotonically over short timescales, yet consistently decreases over long timescales. Since this behavior is inconsistent with several widespread presumptions in the field of optimization, our findings raise questions as to whether these presumptions are relevant to neural network training. We hope that our findings will inspire future efforts aimed at rigorously understanding optimization at the Edge of Stability. Code is available at https://github.com/locuslab/edge-of-stability."
                },
                {
                    "title": "Catastrophic Fisher Explosion: Early Phase Fisher Matrix Impacts Generalization",
                    "abstract": "The early phase of training a deep neural network has a dramatic effect on the local curvature of the loss function. For instance, using a small learning rate does not guarantee stable optimization because the optimization trajectory has a tendency to steer towards regions of the loss surface with increasing local curvature. We ask whether this tendency is connected to the widely observed phenomenon that the choice of the learning rate strongly influences generalization. We first show that stochastic gradient descent (SGD) implicitly penalizes the trace of the Fisher Information Matrix (FIM), a measure of the local curvature, from the start of training. We argue it is an implicit regularizer in SGD by showing that explicitly penalizing the trace of the FIM can significantly improve generalization. We highlight that poor final generalization coincides with the trace of the FIM attaining a large value early in training, to which we refer as catastrophic Fisher explosion. Finally, to gain insight into the regularization effect of penalizing the trace of the FIM, we show that it limits memorization by reducing the learning speed of examples with noisy labels more than that of the examples with clean labels."
                },
                {
                    "title": "Why are Adaptive Methods Good for Attention Models?",
                    "abstract": "While stochastic gradient descent (SGD) is still the \\emph{de facto} algorithm in deep learning, adaptive methods like Clipped SGD/Adam have been observed to outperform SGD across important tasks, such as attention models. The settings under which SGD performs poorly in comparison to adaptive methods are not well understood yet. In this paper, we provide empirical and theoretical evidence that a heavy-tailed distribution of the noise in stochastic gradients is one cause of SGD's poor performance. We provide the first tight upper and lower convergence bounds for adaptive gradient methods under heavy-tailed noise. Further, we demonstrate how gradient clipping plays a key role in addressing heavy-tailed gradient noise. Subsequently, we show how clipping can be applied in practice by developing an \\emph{adaptive} coordinate-wise clipping algorithm (ACClip) and demonstrate its superior performance on BERT pretraining and finetuning tasks."
                },
                {
                    "title": "Sharpness-Aware Minimization for Efficiently Improving Generalization",
                    "abstract": "In today's heavily overparameterized models, the value of the training loss provides few guarantees on model generalization ability. Indeed, optimizing only the training loss value, as is commonly done, can easily lead to suboptimal model quality. Motivated by the connection between geometry of the loss landscape and generalization---including a generalization bound that we prove here---we introduce a novel, effective procedure for instead simultaneously minimizing loss value and loss sharpness. In particular, our procedure, Sharpness-Aware Minimization (SAM), seeks parameters that lie in neighborhoods having uniformly low loss; this formulation results in a min-max optimization problem on which gradient descent can be performed efficiently. We present empirical results showing that SAM improves model generalization across a variety of benchmark datasets (e.g., CIFAR-{10, 100}, ImageNet, finetuning tasks) and models, yielding novel state-of-the-art performance for several. Additionally, we find that SAM natively provides robustness to label noise on par with that provided by state-of-the-art procedures that specifically target learning with noisy labels."
                },
                {
                    "title": "Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamics",
                    "abstract": "Large datasets have become commonplace in NLP research. However, the increased emphasis on data quantity has made it challenging to assess the quality of data. We introduce Data Maps---a model-based tool to characterize and diagnose datasets. We leverage a largely ignored source of information: the behavior of the model on individual instances during training (training dynamics) for building data maps. This yields two intuitive measures for each example---the model's confidence in the true class, and the variability of this confidence across epochs---obtained in a single run of training. Experiments across four datasets show that these model-dependent measures reveal three distinct regions in the data map, each with pronounced characteristics. First, our data maps show the presence of \"ambiguous\" regions with respect to the model, which contribute the most towards out-of-distribution generalization. Second, the most populous regions in the data are \"easy to learn\" for the model, and play an important role in model optimization. Finally, data maps uncover a region with instances that the model finds \"hard to learn\"; these often correspond to labeling errors. Our results indicate that a shift in focus from quantity to quality of data could lead to robust models and improved out-of-distribution generalization."
                },
                {
                    "title": "Traces of Class/Cross-Class Structure Pervade Deep Learning Spectra",
                    "abstract": "Numerous researchers recently applied empirical spectral analysis to the study of modern deep learning classifiers. We identify and discuss an important formal class/cross-class structure and show how it lies at the origin of the many visually striking features observed in deepnet spectra, some of which were reported in recent articles, others are unveiled here for the first time. These include spectral outliers, \"spikes\", and small but distinct continuous distributions, \"bumps\", often seen beyond the edge of a \"main bulk\". \nThe significance of the cross-class structure is illustrated in three ways: (i) we prove the ratio of outliers to bulk in the spectrum of the Fisher information matrix is predictive of misclassification, in the context of multinomial logistic regression; (ii) we demonstrate how, gradually with depth, a network is able to separate class-distinctive information from class variability, all while orthogonalizing the class-distinctive information; and (iii) we propose a correction to KFAC, a well-known second-order optimization algorithm for training deepnets."
                },
                {
                    "title": "The Pitfalls of Simplicity Bias in Neural Networks",
                    "abstract": "Several works have proposed Simplicity Bias (SB)---the tendency of standard training procedures such as Stochastic Gradient Descent (SGD) to find simple models---to justify why neural networks generalize well [Arpit et al. 2017, Nakkiran et al. 2019, Valle-Perez et al. 2019]. However, the precise notion of simplicity remains vague. Furthermore, previous settings that use SB to justify why neural networks generalize well do not simultaneously capture the brittleness of neural networks---a widely observed phenomenon in practice [Goodfellow et al. 2014, Jo and Bengio 2017]. To this end, we introduce a collection of piecewise-linear and image-based datasets that (a) naturally incorporate a precise notion of simplicity and (b) capture the subtleties of neural networks trained on real datasets. Through theory and experiments on these datasets, we show that SB of SGD and variants is extreme: neural networks rely exclusively on the simplest feature and remain invariant to all predictive complex features. Consequently, the extreme nature of SB explains why seemingly benign distribution shifts and small adversarial perturbations significantly degrade model performance. Moreover, contrary to conventional wisdom, SB can also hurt generalization on the same data distribution, as SB persists even when the simplest feature has less predictive power than the more complex features. We also demonstrate that common approaches for improving generalization and robustness---ensembles and adversarial training---do not mitigate SB and its shortcomings. Given the central role played by SB in generalization and robustness, we hope that the datasets and methods in this paper serve as an effective testbed to evaluate novel algorithmic approaches aimed at avoiding the pitfalls of extreme SB."
                },
                {
                    "title": "The large learning rate phase of deep learning: the catapult mechanism",
                    "abstract": "The choice of initial learning rate can have a profound effect on the performance of deep networks. We present a class of neural networks with solvable training dynamics, and confirm their predictions empirically in practical deep learning settings. The networks exhibit sharply distinct behaviors at small and large learning rates. The two regimes are separated by a phase transition. In the small learning rate phase, training can be understood using the existing theory of infinitely wide neural networks. At large learning rates the model captures qualitatively distinct phenomena, including the convergence of gradient descent dynamics to flatter minima. One key prediction of our model is a narrow range of large, stable learning rates. We find good agreement between our model's predictions and training dynamics in realistic deep learning settings. Furthermore, we find that the optimal performance in such settings is often found in the large learning rate phase. We believe our results shed light on characteristics of models trained at different learning rates. In particular, they fill a gap between existing wide neural network theory, and the nonlinear, large learning rate, training dynamics relevant to practice."
                },
                {
                    "title": "The Break-Even Point on Optimization Trajectories of Deep Neural Networks",
                    "abstract": "The early phase of training of deep neural networks is critical for their final performance. In this work, we study how the hyperparameters of stochastic gradient descent (SGD) used in the early phase of training affect the rest of the optimization trajectory. We argue for the existence of the \"break-even\" point on this trajectory, beyond which the curvature of the loss surface and noise in the gradient are implicitly regularized by SGD. In particular, we demonstrate on multiple classification tasks that using a large learning rate in the initial phase of training reduces the variance of the gradient, and improves the conditioning of the covariance of gradients. These effects are beneficial from the optimization perspective and become visible after the break-even point. Complementing prior work, we also show that using a low learning rate results in bad conditioning of the loss surface even for a neural network with batch normalization layers. In short, our work shows that key properties of the loss surface are strongly influenced by SGD in the early phase of training. We argue that studying the impact of the identified effects on generalization is a promising future direction."
                },
                {
                    "title": "Unique Properties of Flat Minima in Deep Networks",
                    "abstract": "It is well known that (stochastic) gradient descent has an implicit bias towards flat minima. In deep neural network training, this mechanism serves to screen out minima. However, the precise effect that this has on the trained network is not yet fully understood. In this paper, we characterize the flat minima in linear neural networks trained with a quadratic loss. First, we show that linear ResNets with zero initialization necessarily converge to the flattest of all minima. We then prove that these minima correspond to nearly balanced networks whereby the gain from the input to any intermediate representation does not change drastically from one layer to the next. Finally, we show that consecutive layers in flat minima solutions are coupled. That is, one of the left singular vectors of each weight matrix, equals one of the right singular vectors of the next matrix. This forms a distinct path from input to output, that, as we show, is dedicated to the signal that experiences the largest gain end-to-end. Experiments indicate that these properties are characteristic of both linear and nonlinear models trained in practice."
                },
                {
                    "title": "Emergent properties of the local geometry of neural loss landscapes",
                    "abstract": "The local geometry of high dimensional neural network loss landscapes can both challenge our cherished theoretical intuitions as well as dramatically impact the practical success of neural network training. Indeed recent works have observed 4 striking local properties of neural loss landscapes on classification tasks: (1) the landscape exhibits exactly $C$ directions of high positive curvature, where $C$ is the number of classes; (2) gradient directions are largely confined to this extremely low dimensional subspace of positive Hessian curvature, leaving the vast majority of directions in weight space unexplored; (3) gradient descent transiently explores intermediate regions of higher positive curvature before eventually finding flatter minima; (4) training can be successful even when confined to low dimensional {\\it random} affine hyperplanes, as long as these hyperplanes intersect a Goldilocks zone of higher than average curvature. We develop a simple theoretical model of gradients and Hessians, justified by numerical experiments on architectures and datasets used in practice, that {\\it simultaneously} accounts for all $4$ of these surprising and seemingly unrelated properties. Our unified model provides conceptual insights into the emergence of these properties and makes connections with diverse topics in neural networks, random matrix theory, and spin glasses, including the neural tangent kernel, BBP phase transitions, and Derrida's random energy model."
                },
                {
                    "title": "Hessian based analysis of SGD for Deep Nets: Dynamics and Generalization",
                    "abstract": "While stochastic gradient descent (SGD) and variants have been surprisingly successful for training deep nets, several aspects of the optimization dynamics and generalization are still not well understood. In this paper, we present new empirical observations and theoretical results on both the optimization dynamics and generalization behavior of SGD for deep nets based on the Hessian of the training loss and associated quantities. We consider three specific research questions: (1) what is the relationship between the Hessian of the loss and the second moment of stochastic gradients (SGs)? (2) how can we characterize the stochastic optimization dynamics of SGD with fixed and adaptive step sizes and diagonal pre-conditioning based on the first and second moments of SGs? and (3) how can we characterize a scale-invariant generalization bound of deep nets based on the Hessian of the loss, which by itself is not scale invariant? We shed light on these three questions using theoretical results supported by extensive empirical observations, with experiments on synthetic data, MNIST, and CIFAR-10, with different batch sizes, and with different difficulty levels by synthetically adding random labels."
                },
                {
                    "title": "SGD on Neural Networks Learns Functions of Increasing Complexity",
                    "abstract": "We perform an experimental study of the dynamics of Stochastic Gradient Descent (SGD) in learning deep neural networks for several real and synthetic classification tasks. We show that in the initial epochs, almost all of the performance improvement of the classifier obtained by SGD can be explained by a linear classifier. More generally, we give evidence for the hypothesis that, as iterations progress, SGD learns functions of increasing complexity. This hypothesis can be helpful in explaining why SGD-learned classifiers tend to generalize well even in the over-parameterized regime. We also show that the linear classifier learned in the initial stages is \"retained\" throughout the execution even if training is continued to the point of zero training error, and complement this with a theoretical result in a simplified model. Key to our work is a new measure of how well one classifier explains the performance of another, based on conditional mutual information."
                },
                {
                    "title": "An Investigation into Neural Net Optimization via Hessian Eigenvalue Density",
                    "abstract": "To understand the dynamics of optimization in deep neural networks, we develop a tool to study the evolution of the entire Hessian spectrum throughout the optimization process. Using this, we study a number of hypotheses concerning smoothness, curvature, and sharpness in the deep learning literature. We then thoroughly analyze a crucial structural feature of the spectra: in non-batch normalized networks, we observe the rapid appearance of large isolated eigenvalues in the spectrum, along with a surprising concentration of the gradient in the corresponding eigenspaces. In batch normalized networks, these two effects are almost absent. We characterize these effects, and explain how they affect optimization speed through both theory and experiments. As part of this work, we adapt advanced tools from numerical linear algebra that allow scalable and accurate estimation of the entire Hessian spectrum of ImageNet-scale neural networks; this technique may be of independent interest in other applications."
                },
                {
                    "title": "Measurements of Three-Level Hierarchical Structure in the Outliers in the Spectrum of Deepnet Hessians",
                    "abstract": "We consider deep classifying neural networks. We expose a structure in the derivative of the logits with respect to the parameters of the model, which is used to explain the existence of outliers in the spectrum of the Hessian. Previous works decomposed the Hessian into two components, attributing the outliers to one of them, the so-called Covariance of gradients. We show this term is not a Covariance but a second moment matrix, i.e., it is influenced by means of gradients. These means possess an additive two-way structure that is the source of the outliers in the spectrum. This structure can be used to approximate the principal subspace of the Hessian using certain \"averaging\" operations, avoiding the need for high-dimensional eigenanalysis. We corroborate this claim across different datasets, architectures and sample sizes."
                },
                {
                    "title": "The Full Spectrum of Deepnet Hessians at Scale: Dynamics with SGD Training and Sample Size.",
                    "abstract": "We apply state-of-the-art tools in modern high-dimensional numerical linear algebra to approximate efficiently the spectrum of the Hessian of modern deepnets, with tens of millions of parameters, trained on real data. Our results corroborate previous findings, based on small-scale networks, that the Hessian exhibits \"spiked\" behavior, with several outliers isolated from a continuous bulk. We decompose the Hessian into different components and study the dynamics with training and sample size of each term individually."
                },
                {
                    "title": "On the Relation Between the Sharpest Directions of DNN Loss and the SGD Step Length",
                    "abstract": "Stochastic Gradient Descent (SGD) based training of neural networks with a large learning rate or a small batch-size typically ends in well-generalizing, flat regions of the weight space, as indicated by small eigenvalues of the Hessian of the training loss. However, the curvature along the SGD trajectory is poorly understood. An empirical investigation shows that initially SGD visits increasingly sharp regions, reaching a maximum sharpness determined by both the learning rate and the batch-size of SGD. When studying the SGD dynamics in relation to the sharpest directions in this initial phase, we find that the SGD step is large compared to the curvature and commonly fails to minimize the loss along the sharpest directions. Furthermore, using a reduced learning rate along these directions can improve training speed while leading to both sharper and better generalizing solutions compared to vanilla SGD. In summary, our analysis of the dynamics of SGD in the subspace of the sharpest directions shows that they influence the regions that SGD steers to (where larger learning rate or smaller batch size result in wider regions visited), the overall training speed, and the generalization ability of the final model."
                },
                {
                    "title": "Algorithmic Regularization in Learning Deep Homogeneous Models: Layers are Automatically Balanced",
                    "abstract": "We study the implicit regularization imposed by gradient descent for learning multi-layer homogeneous functions including feed-forward fully connected and convolutional deep neural networks with linear, ReLU or Leaky ReLU activation. We rigorously prove that gradient flow (i.e. gradient descent with infinitesimal step size) effectively enforces the differences between squared norms across different layers to remain invariant without any explicit regularization. This result implies that if the weights are initially small, gradient flow automatically balances the magnitudes of all layers. Using a discretization argument, we analyze gradient descent with positive step size for the non-convex low-rank asymmetric matrix factorization problem without any regularization. Inspired by our findings for gradient flow, we prove that gradient descent with step sizes $\\eta_t = O\\left(t^{-\\left( \\frac12+\\delta\\right)} \\right)$ ($0<\\delta\\le\\frac12$) automatically balances two low-rank factors and converges to a bounded global optimum. Furthermore, for rank-$1$ asymmetric matrix factorization we give a finer analysis showing gradient descent with constant step size converges to the global minimum at a globally linear rate. We believe that the idea of examining the invariance imposed by first order algorithms in learning homogeneous models could serve as a fundamental building block for studying optimization for learning deep models."
                },
                {
                    "title": "How Does Batch Normalization Help Optimization? (No, It Is Not About Internal Covariate Shift)",
                    "abstract": "Batch Normalization (BatchNorm) is a widely adopted technique that enables faster and more stable training of deep neural networks (DNNs). Despite its pervasiveness, the exact reasons for BatchNorm's effectiveness are still poorly understood. The popular belief is that this effectiveness stems from controlling the change of the layers' input distributions during training to reduce the so-called \"internal covariate shift\". In this work, we demonstrate that such distributional stability of layer inputs has little to do with the success of BatchNorm. Instead, we uncover a more fundamental impact of BatchNorm on the training process: it makes the optimization landscape significantly smoother. This smoothness induces a more predictive and stable behavior of the gradients, allowing for faster training. These findings bring us closer to a true understanding of our DNN training toolkit."
                },
                {
                    "title": "Deep learning generalizes because the parameter-function map is biased towards simple functions",
                    "abstract": "Deep neural networks (DNNs) generalize remarkably well without explicit regularization even in the strongly over-parametrized regime where classical learning theory would instead predict that they would severely overfit. While many proposals for some kind of implicit regularization have been made to rationalise this success, there is no consensus for the fundamental reason why DNNs do not strongly overfit. In this paper, we provide a new explanation. By applying a very general probability-complexity bound recently derived from algorithmic information theory (AIT), we argue that the parameter-function map of many DNNs should be exponentially biased towards simple functions. We then provide clear evidence for this strong simplicity bias in a model DNN for Boolean functions, as well as in much larger fully connected and convolutional networks applied to CIFAR10 and MNIST. As the target functions in many real problems are expected to be highly structured, this intrinsic simplicity bias helps explain why deep networks generalize well on real world problems. This picture also facilitates a novel PAC-Bayes approach where the prior is taken over the DNN input-output function space, rather than the more conventional prior over parameter space. If we assume that the training algorithm samples parameters close to uniformly within the zero-error region then the PAC-Bayes theorem can be used to guarantee good expected generalization for target functions producing high-likelihood training sets. By exploiting recently discovered connections between DNNs and Gaussian processes to estimate the marginal likelihood, we produce relatively tight generalization PAC-Bayes error bounds which correlate well with the true error on realistic datasets such as MNIST and CIFAR10 and for architectures including convolutional and fully connected networks."
                },
                {
                    "title": "The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks",
                    "abstract": "Neural network pruning techniques can reduce the parameter counts of trained networks by over 90%, decreasing storage requirements and improving computational performance of inference without compromising accuracy. However, contemporary experience is that the sparse architectures produced by pruning are difficult to train from the start, which would similarly improve training performance. \nWe find that a standard pruning technique naturally uncovers subnetworks whose initializations made them capable of training effectively. Based on these results, we articulate the \"lottery ticket hypothesis:\" dense, randomly-initialized, feed-forward networks contain subnetworks (\"winning tickets\") that - when trained in isolation - reach test accuracy comparable to the original network in a similar number of iterations. The winning tickets we find have won the initialization lottery: their connections have initial weights that make training particularly effective. \nWe present an algorithm to identify winning tickets and a series of experiments that support the lottery ticket hypothesis and the importance of these fortuitous initializations. We consistently find winning tickets that are less than 10-20% of the size of several fully-connected and convolutional feed-forward architectures for MNIST and CIFAR10. Above this size, the winning tickets that we find learn faster than the original network and reach higher test accuracy."
                },
                {
                    "title": "The Anisotropic Noise in Stochastic Gradient Descent: Its Behavior of Escaping from Sharp Minima and Regularization Effects",
                    "abstract": "Understanding the behavior of stochastic gradient descent (SGD) in the context of deep neural networks has raised lots of concerns recently. Along this line, we study a general form of gradient based optimization dynamics with unbiased noise, which unifies SGD and standard Langevin dynamics. Through investigating this general optimization dynamics, we analyze the behavior of SGD on escaping from minima and its regularization effects. A novel indicator is derived to characterize the efficiency of escaping from minima through measuring the alignment of noise covariance and the curvature of loss function. Based on this indicator, two conditions are established to show which type of noise structure is superior to isotropic noise in term of escaping efficiency. We further show that the anisotropic noise in SGD satisfies the two conditions, and thus helps to escape from sharp and poor minima effectively, towards more stable and flat minima that typically generalize well. We systematically design various experiments to verify the benefits of the anisotropic noise, compared with full gradient descent plus isotropic diffusion (i.e. Langevin dynamics)."
                },
                {
                    "title": "A Walk with SGD",
                    "abstract": "Exploring why stochastic gradient descent (SGD) based optimization methods train deep neural networks (DNNs) that generalize well has become an active area of research. Towards this end, we empirically study the dynamics of SGD when training over-parametrized DNNs. Specifically we study the DNN loss surface along the trajectory of SGD by interpolating the loss surface between parameters from consecutive \\textit{iterations} and tracking various metrics during training. We find that the loss interpolation between parameters before and after a training update is roughly convex with a minimum (\\textit{valley floor}) in between for most of the training. Based on this and other metrics, we deduce that during most of the training, SGD explores regions in a valley by bouncing off valley walls at a height above the valley floor. This 'bouncing off walls at a height' mechanism helps SGD traverse larger distance for small batch sizes and large learning rates which we find play qualitatively different roles in the dynamics. While a large learning rate maintains a large height from the valley floor, a small batch size injects noise facilitating exploration. We find this mechanism is crucial for generalization because the valley floor has barriers and this exploration above the valley floor allows SGD to quickly travel far away from the initialization point (without being affected by barriers) and find flatter regions, corresponding to better generalization."
                },
                {
                    "title": "Three Factors Influencing Minima in SGD",
                    "abstract": "We study the statistical properties of the endpoint of stochastic gradient descent (SGD). We approximate SGD as a stochastic differential equation (SDE) and consider its Boltzmann Gibbs equilibrium distribution under the assumption of isotropic variance in loss gradients.. Through this analysis, we find that three factors \u2013 learning rate, batch size and the variance of the loss gradients \u2013 control the trade-off between the depth and width of the minima found by SGD, with wider minima favoured by a higher ratio of learning rate to batch size. In the equilibrium distribution only the ratio of learning rate to batch size appears, implying that it\u2019s invariant under a simultaneous rescaling of each by the same amount. We experimentally show how learning rate and batch size affect SGD from two perspectives: the endpoint of SGD and the dynamics that lead up to it. For the endpoint, the experiments suggest the endpoint of SGD is similar under simultaneous rescaling of batch size and learning rate, and also that a higher ratio leads to flatter minima, both findings are consistent with our theoretical analysis. We note experimentally that the dynamics also seem to be similar under the same rescaling of learning rate and batch size, which we explore showing that one can exchange batch size and learning rate in a cyclical learning rate schedule. Next, we illustrate how noise affects memorization, showing that high noise levels lead to better generalization. Finally, we find experimentally that the similarity under simultaneous rescaling of learning rate and batch size breaks down if the learning rate gets too large or the batch size gets too small."
                },
                {
                    "title": "Empirical Analysis of the Hessian of Over-Parametrized Neural Networks",
                    "abstract": "We study the properties of common loss surfaces through their Hessian matrix. In particular, in the context of deep learning, we empirically show that the spectrum of the Hessian is composed of two parts: (1) the bulk centered near zero, (2) and outliers away from the bulk. We present numerical evidence and mathematical justifications to the following conjectures laid out by Sagun et al. (2016): Fixing data, increasing the number of parameters merely scales the bulk of the spectrum; fixing the dimension and changing the data (for instance adding more clusters or making the data less separable) only affects the outliers. We believe that our observations have striking implications for non-convex optimization in high dimensions. First, the flatness of such landscapes (which can be measured by the singularity of the Hessian) implies that classical notions of basins of attraction may be quite misleading. And that the discussion of wide/narrow basins may be in need of a new perspective around over-parametrization and redundancy that are able to create large connected components at the bottom of the landscape. Second, the dependence of small number of large eigenvalues to the data distribution can be linked to the spectrum of the covariance matrix of gradients of model outputs. With this in mind, we may reevaluate the connections within the data-architecture-algorithm framework of a model, hoping that it would shed light into the geometry of high-dimensional and non-convex spaces in modern applications. In particular, we present a case that links the two observations: small and large batch gradient descent appear to converge to different basins of attraction but we show that they are in fact connected through their flat region and so belong to the same basin."
                },
                {
                    "title": "Understanding Black-box Predictions via Influence Functions",
                    "abstract": "How can we explain the predictions of a black-box model? In this paper, we use influence functions \u2014 a classic technique from robust statistics \u2014 to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. We show that even on non-convex and non-differentiable models where the theory breaks down, approximations to influence functions can still provide valuable information. On linear models and convolutional neural networks, we demonstrate that influence functions are useful for multiple purposes: understanding model behavior, debugging models, detecting dataset errors, and even creating visually-indistinguishable training-set attacks."
                },
                {
                    "title": "Eigenvalues of the Hessian in Deep Learning: Singularity and Beyond",
                    "abstract": "We look at the eigenvalues of the Hessian of a loss function before and after training. The eigenvalue distribution is seen to be composed of two parts, the bulk which is concentrated around zero, and the edges which are scattered away from zero. We present empirical evidence for the bulk indicating how over-parametrized the system is, and for the edges that depend on the input data."
                },
                {
                    "title": "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift",
                    "abstract": "Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.82% top-5 test error, exceeding the accuracy of human raters."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "Exact solutions to the nonlinear dynamics of learning in deep linear neural networks",
                    "abstract": "Despite the widespread practical success of deep learning methods, our theoretical understanding of the dynamics of learning in deep neural networks remains quite sparse. We attempt to bridge the gap between the theory and practice of deep learning by systematically analyzing learning dynamics for the restricted case of deep linear neural networks. Despite the linearity of their input-output map, such networks have nonlinear gradient descent dynamics on weights that change with the addition of each new hidden layer. We show that deep linear networks exhibit nonlinear learning phenomena similar to those seen in simulations of nonlinear networks, including long plateaus followed by rapid transitions to lower error solutions, and faster convergence from greedy unsupervised pretraining initial conditions than from random initial conditions. We provide an analytical description of these phenomena by finding new exact solutions to the nonlinear dynamics of deep learning. Our theoretical analysis also reveals the surprising finding that as the depth of a network approaches infinity, learning speed can nevertheless remain finite: for a special class of initial conditions on the weights, very deep networks incur only a finite, depth independent, delay in learning speed relative to shallow networks. We show that, under certain conditions on the training data, unsupervised pretraining can find this special class of initial conditions, while scaled random Gaussian initializations cannot. We further exhibit a new class of random orthogonal initial conditions on weights that, like unsupervised pre-training, enjoys depth independent learning times. We further show that these initial conditions also lead to faithful propagation of gradients even in deep nonlinear networks, as long as they operate in a special regime known as the edge of chaos."
                },
                {
                    "title": "How SGD Selects the Global Minima in Over-parameterized Learning : A Dynamical Stability Perspective",
                    "abstract": "The question of which global minima are accessible by a stochastic gradient decent (SGD) algorithm with specific learning rate and batch size is studied from the perspective of dynamical stability. The concept of non-uniformity is introduced, which, together with sharpness, characterizes the stability property of a global minimum and hence the accessibility of a particular SGD algorithm to that global minimum. In particular, this analysis shows that learning rate and batch size play different roles in minima selection. Extensive empirical results seem to correlate well with the theoretical findings and provide further support to these claims."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CV",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow do opposing signals in neural network training dynamics influence optimization and contribute to phenomena such as the edge of stability?\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding the role of opposing signals in neural network optimization is crucial for the research community as it can unify various observed phenomena in neural network training, such as grokking and the effectiveness of optimization techniques like Adam and Batch Normalization. By addressing this question, we can advance theoretical knowledge and improve practical applications in machine learning, leading to more robust and efficient training algorithms. This could pave the way for new methodologies that enhance model performance and generalization, ultimately impacting a wide range of applications from computer vision to natural language processing.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complex interplay between opposing signals and their effects on the optimization landscape of neural networks. Naive approaches may fail because they do not account for the nuanced interactions between features that provide conflicting gradients, which can lead to instability during training. Additionally, identifying and quantifying the influence of these outlier groups requires sophisticated analytical techniques and a deep understanding of the underlying data distributions, making it a technically demanding problem.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on isolated phenomena in neural network optimization without considering the broader implications of opposing signals. Limitations in existing methodologies and a lack of comprehensive frameworks to analyze the interactions between features have hindered progress. Additionally, many studies have not adequately addressed the role of outliers in training dynamics, which is a key aspect of this problem. Our approach differs by explicitly examining the influence of these opposing signals and their correlation with target tasks, providing a more integrated perspective on neural network training.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves analyzing the training dynamics of neural networks, specifically ResNet-18, on the CIFAR-10 dataset. We will utilize metrics such as overall loss and group-specific losses to quantify the impact of opposing signals during training. By identifying and characterizing these outlier groups, we expect to demonstrate their significant influence on optimization dynamics and the edge of stability phenomenon. The anticipated outcome is a clearer understanding of how these signals interact and affect training, leading to insights that can inform the development of more effective training algorithms."
            }
        },
        "author_data": {
            "64750ad1-4c24-4630-8d1a-c933477b3ba9": {
                "pk": "64750ad1-4c24-4630-8d1a-c933477b3ba9",
                "name": "Elan Rosenfeld",
                "collaborators": [
                    "Pradeep Ravikumar",
                    "Andrej Risteski",
                    "Ifigeneia Apostolopoulou",
                    "A. Dubrawski",
                    "J. Z. Kolter",
                    "S. Garg",
                    "Simon Buchholz",
                    "Goutham Rajendran",
                    "Bryon Aragam",
                    "B. Scholkopf"
                ],
                "domain": [
                    "Causal Inference",
                    "Robustness",
                    "Deep Learning",
                    "Adversarial Machine Learning"
                ],
                "publications": [
                    {
                        "title": "(Almost) Provable Error Bounds Under Distribution Shift via Disagreement Discrepancy",
                        "abstract": "We derive an (almost) guaranteed upper bound on the error of deep neural networks under distribution shift using unlabeled test data. Prior methods either give bounds that are vacuous in practice or give estimates that are accurate on average but heavily underestimate error for a sizeable fraction of shifts. In particular, the latter only give guarantees based on complex continuous measures such as test calibration -- which cannot be identified without labels -- and are therefore unreliable. Instead, our bound requires a simple, intuitive condition which is well justified by prior empirical works and holds in practice effectively 100% of the time. The bound is inspired by $\\mathcal{H}\\Delta\\mathcal{H}$-divergence but is easier to evaluate and substantially tighter, consistently providing non-vacuous guarantees. Estimating the bound requires optimizing one multiclass classifier to disagree with another, for which some prior works have used sub-optimal proxy losses; we devise a\"disagreement loss\"which is theoretically justified and performs better in practice. We expect this loss can serve as a drop-in replacement for future methods which require maximizing multiclass disagreement. Across a wide range of benchmarks, our method gives valid error bounds while achieving average accuracy comparable to competitive estimation baselines. Code is publicly available at https://github.com/erosenfeld/disagree_discrep ."
                    },
                    {
                        "title": "Learning Linear Causal Representations from Interventions under General Nonlinear Mixing",
                        "abstract": "We study the problem of learning causal representations from unknown, latent interventions in a general setting, where the latent distribution is Gaussian but the mixing function is completely general. We prove strong identifiability results given unknown single-node interventions, i.e., without having access to the intervention targets. This generalizes prior works which have focused on weaker classes, such as linear maps or paired counterfactual data. This is also the first instance of causal identifiability from non-paired interventions for deep neural network embeddings. Our proof relies on carefully uncovering the high-dimensional geometric structure present in the data distribution after a non-linear density transformation, which we capture by analyzing quadratic forms of precision matrices of the latent distributions. Finally, we propose a contrastive algorithm to identify the latent variables in practice and evaluate its performance on various tasks."
                    },
                    {
                        "title": "Identifying Representations for Intervention Extrapolation",
                        "abstract": "The premise of identifiable and causal representation learning is to improve the current representation learning paradigm in terms of generalizability or robustness. Despite recent progress in questions of identifiability, more theoretical results demonstrating concrete advantages of these methods for downstream tasks are needed. In this paper, we consider the task of intervention extrapolation: predicting how interventions affect an outcome, even when those interventions are not observed at training time, and show that identifiable representations can provide an effective solution to this task even if the interventions affect the outcome non-linearly. Our setup includes an outcome Y, observed features X, which are generated as a non-linear transformation of latent features Z, and exogenous action variables A, which influence Z. The objective of intervention extrapolation is to predict how interventions on A that lie outside the training support of A affect Y. Here, extrapolation becomes possible if the effect of A on Z is linear and the residual when regressing Z on A has full support. As Z is latent, we combine the task of intervention extrapolation with identifiable representation learning, which we call Rep4Ex: we aim to map the observed features X into a subspace that allows for non-linear extrapolation in A. We show that the hidden representation is identifiable up to an affine transformation in Z-space, which is sufficient for intervention extrapolation. The identifiability is characterized by a novel constraint describing the linearity assumption of A on Z. Based on this insight, we propose a method that enforces the linear invariance constraint and can be combined with any type of autoencoder. We validate our theoretical findings through synthetic experiments and show that our approach succeeds in predicting the effects of unseen interventions."
                    },
                    {
                        "title": "Domain-Adjusted Regression or: ERM May Already Learn Features Sufficient for Out-of-Distribution Generalization",
                        "abstract": "A common explanation for the failure of deep networks to generalize out-of-distribution is that they fail to recover the \u201ccorrect\u201d features. We challenge this notion with a simple experiment which suggests that ERM already learns suf\ufb01cient features and that the current bottleneck is not feature learning, but robust regression . Our \ufb01ndings also imply that given a small amount of data from the target distribution, retraining only the last linear layer will give excellent performance. We therefore argue that devising simpler methods for learning predictors on existing features is a promising direction for future research. Towards this end, we introduce Domain-Adjusted Regression (DARE), a convex objective for learning a linear predictor that is provably robust under a new model of distribution shift. Rather than learning one function, DARE performs a domain-speci\ufb01c adjustment to unify the domains in a canonical latent space and learns to predict in this space. Under a natural model, we prove that the DARE solution is the minimax-optimal predictor for a constrained set of test distributions. Further, we provide the \ufb01rst \ufb01nite-environment convergence guarantee to the minimax risk, improving over existing analyses which only yield minimax predictors after an environment threshold. Evaluated on \ufb01netuned features, we \ufb01nd that DARE compares favorably to prior methods, consistently achieving equal or better performance."
                    },
                    {
                        "title": "APE: Aligning Pretrained Encoders to Quickly Learn Aligned Multimodal Representations",
                        "abstract": "Recent advances in learning aligned multimodal representations have been primarily driven by training large neural networks on massive, noisy paired-modality datasets. In this work, we ask whether it is possible to achieve similar results with substantially less training time and data. We achieve this by taking advantage of existing pretrained unimodal encoders and careful curation of alignment data relevant to the downstream task of interest. We study a natural approach to aligning existing encoders via small auxiliary functions, and we find that this method is competitive with (or outperforms) state of the art in many settings while being less prone to overfitting, less costly to train, and more robust to distribution shift. With a properly chosen alignment distribution, our method surpasses prior state of the art for ImageNet zero-shot classification on public data while using two orders of magnitude less time and data and training 77% fewer parameters."
                    },
                    {
                        "title": "Deep Attentive Variational Inference",
                        "abstract": "Stochastic Variational Inference is a powerful framework for learning large-scale probabilistic latent variable models. However, typical assumptions on the factorization or independence of the latent variables can substantially restrict its capacity for inference and generative modeling. A major line of active research aims at building more expressive variational models by designing deep hierarchies of interdependent latent variables. Although these models exhibit superior performance and enable richer latent representations, we show that they incur diminishing returns: adding more stochastic layers to an already very deep model yields small predictive improvement while substantially increasing the inference and training time. Moreover, the architecture for this class of models favors proximate interactions among the latent variables between neighboring layers when designing the conditioning factors of the involved distributions. This is the first work that proposes attention mechanisms to build more expressive variational distributions in deep probabilistic models by explicitly modeling both nearby and distant interactions in the latent space. Specifically, we propose deep attentive variational autoencoder and test it on a variety of established datasets. We show it achieves state-of-the-art log-likelihoods while using fewer latent layers and requiring less training time than existing models. The proposed holistic inference reduces computational footprint by alleviating the need for deep hierarchies. Project code: https://github.com/ifiaposto/Deep_Attentive_VI"
                    },
                    {
                        "title": "Analyzing and Improving the Optimization Landscape of Noise-Contrastive Estimation",
                        "abstract": "Noise-contrastive estimation (NCE) is a statistically consistent method for learning unnormalized probabilistic models. It has been empirically observed that the choice of the noise distribution is crucial for NCE's performance. However, such observations have never been made formal or quantitative. In fact, it is not even clear whether the difficulties arising from a poorly chosen noise distribution are statistical or algorithmic in nature. In this work, we formally pinpoint reasons for NCE's poor performance when an inappropriate noise distribution is used. Namely, we prove these challenges arise due to an ill-behaved (more precisely, flat) loss landscape. To address this, we introduce a variant of NCE called\"eNCE\"which uses an exponential loss and for which normalized gradient descent addresses the landscape issues provably when the target and noise distributions are in a given exponential family."
                    },
                    {
                        "title": "Iterative Feature Matching: Toward Provable Domain Generalization with Logarithmic Environments",
                        "abstract": "Domain generalization aims at performing well on unseen test environments with data from a limited number of training environments. Despite a proliferation of proposal algorithms for this task, assessing their performance both theoretically and empirically is still very challenging. Distributional matching algorithms such as (Conditional) Domain Adversarial Networks [Ganin et al., 2016, Long et al., 2018] are popular and enjoy empirical success, but they lack formal guarantees. Other approaches such as Invariant Risk Minimization (IRM) require a prohibitively large number of training environments -- linear in the dimension of the spurious feature space $d_s$ -- even on simple data models like the one proposed by [Rosenfeld et al., 2021]. Under a variant of this model, we show that both ERM and IRM cannot generalize with $o(d_s)$ environments. We then present an iterative feature matching algorithm that is guaranteed with high probability to yield a predictor that generalizes after seeing only $O(\\log d_s)$ environments. Our results provide the first theoretical justification for a family of distribution-matching algorithms widely used in practice under a concrete nontrivial data model."
                    },
                    {
                        "title": "An Online Learning Approach to Interpolation and Extrapolation in Domain Generalization",
                        "abstract": "A popular assumption for out-of-distribution generalization is that the training data comprises sub-datasets, each drawn from a distinct distribution; the goal is then to\"interpolate\"these distributions and\"extrapolate\"beyond them -- this objective is broadly known as domain generalization. A common belief is that ERM can interpolate but not extrapolate and that the latter is considerably more difficult, but these claims are vague and lack formal justification. In this work, we recast generalization over sub-groups as an online game between a player minimizing risk and an adversary presenting new test distributions. Under an existing notion of inter- and extrapolation based on reweighting of sub-group likelihoods, we rigorously demonstrate that extrapolation is computationally much harder than interpolation, though their statistical complexity is not significantly different. Furthermore, we show that ERM -- or a noisy variant -- is provably minimax-optimal for both tasks. Our framework presents a new avenue for the formal analysis of domain generalization algorithms which may be of independent interest."
                    },
                    {
                        "title": "The Risks of Invariant Risk Minimization",
                        "abstract": "Invariant Causal Prediction (Peters et al., 2016) is a technique for out-of-distribution generalization which assumes that some aspects of the data distribution vary across the training set but that the underlying causal mechanisms remain constant. Recently, Arjovsky et al. (2019) proposed Invariant Risk Minimization (IRM), an objective based on this idea for learning deep, invariant features of data which are a complex function of latent variables; many alternatives have subsequently been suggested. However, formal guarantees for all of these works are severely lacking. In this paper, we present the first analysis of classification under the IRM objective$-$as well as these recently proposed alternatives$-$under a fairly natural and general model. In the linear case, we show simple conditions under which the optimal solution succeeds or, more often, fails to recover the optimal invariant predictor. We furthermore present the very first results in the non-linear regime: we demonstrate that IRM can fail catastrophically unless the test data are sufficiently similar to the training distribution$-$this is precisely the issue that it was intended to solve. Thus, in this setting we find that IRM and its alternatives fundamentally do not improve over standard Empirical Risk Minimization."
                    },
                    {
                        "title": "Certified Robustness to Label-Flipping Attacks via Randomized Smoothing",
                        "abstract": "Machine learning algorithms are known to be susceptible to data poisoning attacks, where an adversary manipulates the training data to degrade performance of the resulting classifier. In this work, we present a unifying view of randomized smoothing over arbitrary functions, and we leverage this novel characterization to propose a new strategy for building classifiers that are pointwise-certifiably robust to general data poisoning attacks. As a specific instantiation, we utilize our framework to build linear classifiers that are robust to a strong variant of label flipping, where each test example is targeted independently. In other words, for each test point, our classifier includes a certification that its prediction would be the same had some number of training labels been changed adversarially. Randomized smoothing has previously been used to guarantee---with high probability---test-time robustness to adversarial manipulation of the input to a classifier; we derive a variant which provides a deterministic, analytical bound, sidestepping the probabilistic certificates that traditionally result from the sampling subprocedure. Further, we obtain these certified bounds with minimal additional runtime complexity over standard classification and no assumptions on the train or test distributions. We generalize our results to the multi-class case, providing the first multi-class classification algorithm that is certifiably robust to label-flipping attacks."
                    },
                    {
                        "title": "Self-Reflective Variational Autoencoder",
                        "abstract": "The Variational Autoencoder (VAE) is a powerful framework for learning probabilistic latent variable generative models. However, typical assumptions on the approximate posterior distribution of the encoder and/or the prior, seriously restrict its capacity for inference and generative modeling. Variational inference based on neural autoregressive models respects the conditional dependencies of the exact posterior, but this flexibility comes at a cost: such models are expensive to train in high-dimensional regimes and can be slow to produce samples. In this work, we introduce an orthogonal solution, which we call self-reflective inference. By redesigning the hierarchical structure of existing VAE architectures, self-reflection ensures that the stochastic flow preserves the factorization of the exact posterior, sequentially updating the latent codes in a recurrent manner consistent with the generative model. We empirically demonstrate the clear advantages of matching the variational posterior to the exact posterior - on binarized MNIST, self-reflective inference achieves state-of-the art performance without resorting to complex, computationally expensive components such as autoregressive layers. Moreover, we design a variational normalizing flow that employs the proposed architecture, yielding predictive benefits compared to its purely generative counterpart. Our proposed modification is quite general and complements the existing literature; self-reflective inference can naturally leverage advances in distribution estimation and generative modeling to improve the capacity of each layer in the hierarchy."
                    },
                    {
                        "title": "Human-Usable Password Schemas: Beyond Information-Theoretic Security",
                        "abstract": "Password users frequently employ passwords that are too simple, or they just reuse passwords for multiple websites. A common complaint is that utilizing secure passwords is too difficult. One possible solution to this problem is to use a password schema. Password schemas are deterministic functions which map challenges (typically the website name) to responses (passwords). Previous work has been done on developing and analyzing publishable schemas, but these analyses have been information-theoretic, not complexity-theoretic; they consider an adversary with infinite computing power.  We perform an analysis with respect to adversaries having currently achievable computing capabilities, assessing the realistic practical security of such schemas. We prove for several specific schemas that a computer is no worse off than an infinite adversary and that it can successfully extract all information from leaked challenges and their respective responses, known as challenge-response pairs. We also show that any schema that hopes to be secure against adversaries with bounded computation should obscure information in a very specific way, by introducing many possible constraints with each challenge-response pair. These surprising results put the analyses of password schemas on a more solid and practical footing."
                    },
                    {
                        "title": "Certified Adversarial Robustness via Randomized Smoothing",
                        "abstract": "We show how to turn any classifier that classifies well under Gaussian noise into a new classifier that is certifiably robust to adversarial perturbations under the $\\ell_2$ norm. This \"randomized smoothing\" technique has been proposed recently in the literature, but existing guarantees are loose. We prove a tight robustness guarantee in $\\ell_2$ norm for smoothing with Gaussian noise. We use randomized smoothing to obtain an ImageNet classifier with e.g. a certified top-1 accuracy of 49% under adversarial perturbations with $\\ell_2$ norm less than 0.5 (=127/255). No certified defense has been shown feasible on ImageNet except for smoothing. On smaller-scale datasets where competing approaches to certified $\\ell_2$ robustness are viable, smoothing delivers higher certified accuracies. Our strong empirical results suggest that randomized smoothing is a promising direction for future research into adversarially robust classification. Code and models are available at this http URL."
                    }
                ]
            },
            "553a4139-47a4-4818-bac5-870a96d91f9c": {
                "pk": "553a4139-47a4-4818-bac5-870a96d91f9c",
                "name": "Andrej Risteski",
                "collaborators": [
                    "Tanya Marwah",
                    "Yuchen Li",
                    "Pradeep Ravikumar",
                    "Zachary Chase Lipton",
                    "Jianfeng Lu",
                    "Bing Liu",
                    "Yilong Qin",
                    "Dhruv Rohatgi",
                    "Ankur Moitra",
                    "Ashwini Pokle"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Self-Supervised Learning",
                    "Partial Differential Equations",
                    "Knowledge Distillation"
                ],
                "publications": [
                    {
                        "title": "Towards characterizing the value of edge embeddings in Graph Neural Networks",
                        "abstract": "Graph neural networks (GNNs) are the dominant approach to solving machine learning problems defined over graphs. Despite much theoretical and empirical work in recent years, our understanding of finer-grained aspects of architectural design for GNNs remains impoverished. In this paper, we consider the benefits of architectures that maintain and update edge embeddings. On the theoretical front, under a suitable computational abstraction for a layer in the model, as well as memory constraints on the embeddings, we show that there are natural tasks on graphical models for which architectures leveraging edge embeddings can be much shallower. Our techniques are inspired by results on time-space tradeoffs in theoretical computer science. Empirically, we show architectures that maintain edge embeddings almost always improve on their node-based counterparts -- frequently significantly so in topologies that have ``hub'' nodes."
                    },
                    {
                        "title": "On the Benefits of Memory for Modeling Time-Dependent PDEs",
                        "abstract": "Data-driven techniques have emerged as a promising alternative to traditional numerical methods for solving partial differential equations (PDEs). These techniques frequently offer a better trade-off between computational cost and accuracy for many PDE families of interest. For time-dependent PDEs, existing methodologies typically treat PDEs as Markovian systems, i.e., the evolution of the system only depends on the ``current state'', and not the past states. However, distortion of the input signals -- e.g., due to discretization or low-pass filtering -- can render the evolution of the distorted signals non-Markovian. In this work, motivated by the Mori-Zwanzig theory of model reduction, we investigate the impact of architectures with memory for modeling PDEs: that is, when past states are explicitly used to predict the future. We introduce Memory Neural Operator (MemNO), a network based on the recent SSM architectures and Fourier Neural Operator (FNO). We empirically demonstrate on a variety of PDE families of interest that when the input is given on a low-resolution grid, MemNO significantly outperforms the baselines without memory, achieving more than 6 times less error on unseen PDEs. Via a combination of theory and experiments, we show that the effect of memory is particularly significant when the solution of the PDE has high frequency Fourier components (e.g., low-viscosity fluid dynamics), and it also increases robustness to observation noise."
                    },
                    {
                        "title": "Progressive distillation induces an implicit curriculum",
                        "abstract": "Knowledge distillation leverages a teacher model to improve the training of a student model. A persistent challenge is that a better teacher does not always yield a better student, to which a common mitigation is to use additional supervision from several ``intermediate'' teachers. One empirically validated variant of this principle is progressive distillation, where the student learns from successive intermediate checkpoints of the teacher. Using sparse parity as a sandbox, we identify an implicit curriculum as one mechanism through which progressive distillation accelerates the student's learning. This curriculum is available only through the intermediate checkpoints but not the final converged one, and imparts both empirical acceleration and a provable sample complexity benefit to the student. We then extend our investigation to Transformers trained on probabilistic context-free grammars (PCFGs) and real-world pre-training datasets (Wikipedia and Books). Through probing the teacher model, we identify an analogous implicit curriculum where the model progressively learns features that capture longer context. Our theoretical and empirical findings on sparse parity, complemented by empirical observations on more complex tasks, highlight the benefit of progressive distillation via implicit curriculum across setups."
                    },
                    {
                        "title": "Promises and Pitfalls of Generative Masked Language Modeling: Theoretical Framework and Practical Guidelines",
                        "abstract": "Autoregressive language models are the currently dominant paradigm for text generation, but they have some fundamental limitations that cannot be remedied by scale-for example inherently sequential and unidirectional generation. While alternate classes of models have been explored, we have limited mathematical understanding of their fundamental power and limitations. In this paper we focus on Generative Masked Language Models (GMLMs), a non-autoregressive paradigm in which we train a model to fit conditional probabilities of the data distribution via masking, which are subsequently used as inputs to a Markov Chain to draw samples from the model, These models empirically strike a promising speed-quality trade-off as each step can be typically parallelized by decoding the entire sequence in parallel. We develop a mathematical framework for analyzing and improving such models which sheds light on questions of sample complexity and inference speed and quality. Empirically, we adapt the T5 model for iteratively-refined parallel decoding, achieving 2-3x speedup in machine translation with minimal sacrifice in quality compared with autoregressive models. We run careful ablation experiments to give recommendations on key design choices, and make fine-grained observations on the common error modes in connection with our theory. Our mathematical analyses and empirical observations characterize both potentials and limitations of this approach, and can be applied to future works on improving understanding and performance of GMLMs. Our codes are released at https://github.com/google-research/google-research/tree/master/padir"
                    },
                    {
                        "title": "Provable benefits of annealing for estimating normalizing constants: Importance Sampling, Noise-Contrastive Estimation, and beyond",
                        "abstract": "Recent research has developed several Monte Carlo methods for estimating the normalization constant (partition function) based on the idea of annealing. This means sampling successively from a path of distributions that interpolate between a tractable\"proposal\"distribution and the unnormalized\"target\"distribution. Prominent estimators in this family include annealed importance sampling and annealed noise-contrastive estimation (NCE). Such methods hinge on a number of design choices: which estimator to use, which path of distributions to use and whether to use a path at all; so far, there is no definitive theory on which choices are efficient. Here, we evaluate each design choice by the asymptotic estimation error it produces. First, we show that using NCE is more efficient than the importance sampling estimator, but in the limit of infinitesimal path steps, the difference vanishes. Second, we find that using the geometric path brings down the estimation error from an exponential to a polynomial function of the parameter distance between the target and proposal distributions. Third, we find that the arithmetic path, while rarely used, can offer optimality properties over the universally-used geometric path. In fact, in a particular limit, the optimal path is arithmetic. Based on this theory, we finally propose a two-step estimator to approximate the optimal path in an efficient way."
                    },
                    {
                        "title": "Understanding Augmentation-based Self-Supervised Representation Learning via RKHS Approximation",
                        "abstract": "Data augmentation is critical to the empirical success of modern self-supervised representation learning, such as contrastive learning and masked language modeling. However, a theoretical understanding of the exact role of augmentation remains limited. Recent work has built the connection between self-supervised learning and the approximation of the top eigenspace of a graph Laplacian operator, suggesting that learning a linear probe atop such representation can be connected to RKHS regression. Building on this insight, this work delves into a statistical analysis of augmentation-based pretraining. Starting from the isometry property, a geometric characterization of the target function given by the augmentation, we disentangle the effects of the model and the augmentation, and prove two generalization bounds that are free of model complexity. Our first bound works for an arbitrary encoder, where the prediction error is decomposed as the sum of an estimation error incurred by fitting a linear probe with RKHS regression, and an approximation error entailed by RKHS approximation. Our second bound specifically addresses the case where the encoder is near-optimal, that is it approximates the top-d eigenspace of the RKHS induced by the augmentation. A key ingredient in our analysis is the augmentation complexity, which we use to quantitatively compare different augmentations and analyze their impact on downstream performance."
                    },
                    {
                        "title": "Fit Like You Sample: Sample-Efficient Generalized Score Matching from Fast Mixing Markov Chains",
                        "abstract": ","
                    },
                    {
                        "title": "Deep Equilibrium Based Neural Operators for Steady-State PDEs",
                        "abstract": "Data-driven machine learning approaches are being increasingly used to solve partial differential equations (PDEs). They have shown particularly striking successes when training an operator, which takes as input a PDE in some family, and outputs its solution. However, the architectural design space, especially given structural knowledge of the PDE family of interest, is still poorly understood. We seek to remedy this gap by studying the benefits of weight-tied neural network architectures for steady-state PDEs. To achieve this, we first demonstrate that the solution of most steady-state PDEs can be expressed as a fixed point of a non-linear operator. Motivated by this observation, we propose FNO-DEQ, a deep equilibrium variant of the FNO architecture that directly solves for the solution of a steady-state PDE as the infinite-depth fixed point of an implicit operator layer using a black-box root solver and differentiates analytically through this fixed point resulting in $\\mathcal{O}(1)$ training memory. Our experiments indicate that FNO-DEQ-based architectures outperform FNO-based baselines with $4\\times$ the number of parameters in predicting the solution to steady-state PDEs such as Darcy Flow and steady-state incompressible Navier-Stokes. Finally, we show FNO-DEQ is more robust when trained with datasets with more noisy observations than the FNO-based baselines, demonstrating the benefits of using appropriate inductive biases in architectural design for different neural network based PDE solvers. Further, we show a universal approximation result that demonstrates that FNO-DEQ can approximate the solution to any steady-state PDE that can be written as a fixed point equation."
                    },
                    {
                        "title": "Provable benefits of score matching",
                        "abstract": "Score matching is an alternative to maximum likelihood (ML) for estimating a probability distribution parametrized up to a constant of proportionality. By fitting the ''score'' of the distribution, it sidesteps the need to compute this constant of proportionality (which is often intractable). While score matching and variants thereof are popular in practice, precise theoretical understanding of the benefits and tradeoffs with maximum likelihood -- both computational and statistical -- are not well understood. In this work, we give the first example of a natural exponential family of distributions such that the score matching loss is computationally efficient to optimize, and has a comparable statistical efficiency to ML, while the ML loss is intractable to optimize using a gradient-based method. The family consists of exponentials of polynomials of fixed degree, and our result can be viewed as a continuous analogue of recent developments in the discrete setting. Precisely, we show: (1) Designing a zeroth-order or first-order oracle for optimizing the maximum likelihood loss is NP-hard. (2) Maximum likelihood has a statistical efficiency polynomial in the ambient dimension and the radius of the parameters of the family. (3) Minimizing the score matching loss is both computationally and statistically efficient, with complexity polynomial in the ambient dimension."
                    },
                    {
                        "title": "Fit Like You Sample: Sample-Efficient Generalized Score Matching from Fast Mixing Diffusions",
                        "abstract": "Score matching is an approach to learning probability distributions parametrized up to a constant of proportionality (e.g. Energy-Based Models). The idea is to fit the score of the distribution, rather than the likelihood, thus avoiding the need to evaluate the constant of proportionality. While there's a clear algorithmic benefit, the statistical\"cost'' can be steep: recent work by Koehler et al. 2022 showed that for distributions that have poor isoperimetric properties (a large Poincar\\'e or log-Sobolev constant), score matching is substantially statistically less efficient than maximum likelihood. However, many natural realistic distributions, e.g. multimodal distributions as simple as a mixture of two Gaussians in one dimension -- have a poor Poincar\\'e constant. In this paper, we show a close connection between the mixing time of a broad class of Markov processes with generator $\\mathcal{L}$ and an appropriately chosen generalized score matching loss that tries to fit $\\frac{\\mathcal{O} p}{p}$. This allows us to adapt techniques to speed up Markov chains to construct better score-matching losses. In particular, ``preconditioning'' the diffusion can be translated to an appropriate ``preconditioning'' of the score loss. Lifting the chain by adding a temperature like in simulated tempering can be shown to result in a Gaussian-convolution annealed score matching loss, similar to Song and Ermon, 2019. Moreover, we show that if the distribution being learned is a finite mixture of Gaussians in $d$ dimensions with a shared covariance, the sample complexity of annealed score matching is polynomial in the ambient dimension, the diameter of the means, and the smallest and largest eigenvalues of the covariance -- obviating the Poincar\\'e constant-based lower bounds of the basic score matching loss shown in Koehler et al. 2022."
                    },
                    {
                        "title": "Transformers are uninterpretable with myopic methods: a case study with bounded Dyck grammars",
                        "abstract": "Interpretability methods aim to understand the algorithm implemented by a trained model (e.g., a Transofmer) by examining various aspects of the model, such as the weight matrices or the attention patterns. In this work, through a combination of theoretical results and carefully controlled experiments on synthetic data, we take a critical view of methods that exclusively focus on individual parts of the model, rather than consider the network as a whole. We consider a simple synthetic setup of learning a (bounded) Dyck language. Theoretically, we show that the set of models that (exactly or approximately) solve this task satisfy a structural characterization derived from ideas in formal languages (the pumping lemma). We use this characterization to show that the set of optima is qualitatively rich; in particular, the attention pattern of a single layer can be ``nearly randomized'', while preserving the functionality of the network. We also show via extensive experiments that these constructions are not merely a theoretical artifact: even after severely constraining the architecture of the model, vastly different solutions can be reached via standard training. Thus, interpretability claims based on inspecting individual heads or weight matrices in the Transformer can be misleading."
                    },
                    {
                        "title": "How Do Transformers Learn Topic Structure: Towards a Mechanistic Understanding",
                        "abstract": "While the successes of transformers across many domains are indisputable, accurate understanding of the learning mechanics is still largely lacking. Their capabilities have been probed on benchmarks which include a variety of structured and reasoning tasks -- but mathematical understanding is lagging substantially behind. Recent lines of work have begun studying representational aspects of this question: that is, the size/depth/complexity of attention-based networks to perform certain tasks. However, there is no guarantee the learning dynamics will converge to the constructions proposed. In our paper, we provide fine-grained mechanistic understanding of how transformers learn\"semantic structure\", understood as capturing co-occurrence structure of words. Precisely, we show, through a combination of mathematical analysis and experiments on Wikipedia data and synthetic data modeled by Latent Dirichlet Allocation (LDA), that the embedding layer and the self-attention layer encode the topical structure. In the former case, this manifests as higher average inner product of embeddings between same-topic words. In the latter, it manifests as higher average pairwise attention between same-topic words. The mathematical results involve several assumptions to make the analysis tractable, which we verify on data, and might be of independent interest as well."
                    },
                    {
                        "title": "Continual learning: a feature extraction formalization, an efficient algorithm, and fundamental obstructions",
                        "abstract": "Continual learning is an emerging paradigm in machine learning, wherein a model is exposed in an online fashion to data from multiple different distributions (i.e. environments), and is expected to adapt to the distribution change. Precisely, the goal is to perform well in the new environment, while simultaneously retaining the performance on the previous environments (i.e. avoid\"catastrophic forgetting\") -- without increasing the size of the model. While this setup has enjoyed a lot of attention in the applied community, there hasn't be theoretical work that even formalizes the desired guarantees. In this paper, we propose a framework for continual learning through the framework of feature extraction -- namely, one in which features, as well as a classifier, are being trained with each environment. When the features are linear, we design an efficient gradient-based algorithm $\\mathsf{DPGD}$, that is guaranteed to perform well on the current environment, as well as avoid catastrophic forgetting. In the general case, when the features are non-linear, we show such an algorithm cannot exist, whether efficient or not."
                    },
                    {
                        "title": "Domain-Adjusted Regression or: ERM May Already Learn Features Sufficient for Out-of-Distribution Generalization",
                        "abstract": "A common explanation for the failure of deep networks to generalize out-of-distribution is that they fail to recover the \u201ccorrect\u201d features. We challenge this notion with a simple experiment which suggests that ERM already learns suf\ufb01cient features and that the current bottleneck is not feature learning, but robust regression . Our \ufb01ndings also imply that given a small amount of data from the target distribution, retraining only the last linear layer will give excellent performance. We therefore argue that devising simpler methods for learning predictors on existing features is a promising direction for future research. Towards this end, we introduce Domain-Adjusted Regression (DARE), a convex objective for learning a linear predictor that is provably robust under a new model of distribution shift. Rather than learning one function, DARE performs a domain-speci\ufb01c adjustment to unify the domains in a canonical latent space and learns to predict in this space. Under a natural model, we prove that the DARE solution is the minimax-optimal predictor for a constrained set of test distributions. Further, we provide the \ufb01rst \ufb01nite-environment convergence guarantee to the minimax risk, improving over existing analyses which only yield minimax predictors after an environment threshold. Evaluated on \ufb01netuned features, we \ufb01nd that DARE compares favorably to prior methods, consistently achieving equal or better performance."
                    },
                    {
                        "title": "Sampling Approximately Low-Rank Ising Models: MCMC meets Variational Methods",
                        "abstract": "We consider Ising models on the hypercube with a general interaction matrix $J$, and give a polynomial time sampling algorithm when all but $O(1)$ eigenvalues of $J$ lie in an interval of length one, a situation which occurs in many models of interest. This was previously known for the Glauber dynamics when *all* eigenvalues fit in an interval of length one; however, a single outlier can force the Glauber dynamics to mix torpidly. Our general result implies the first polynomial time sampling algorithms for low-rank Ising models such as Hopfield networks with a fixed number of patterns and Bayesian clustering models with low-dimensional contexts, and greatly improves the polynomial time sampling regime for the antiferromagnetic/ferromagnetic Ising model with inconsistent field on expander graphs. It also improves on previous approximation algorithm results based on the naive mean-field approximation in variational methods and statistical physics. Our approach is based on a new fusion of ideas from the MCMC and variational inference worlds. As part of our algorithm, we define a new nonconvex variational problem which allows us to sample from an exponential reweighting of a distribution by a negative definite quadratic form, and show how to make this procedure provably efficient using stochastic gradient descent. On top of this, we construct a new simulated tempering chain (on an extended state space arising from the Hubbard-Stratonovich transform) which overcomes the obstacle posed by large positive eigenvalues, and combine it with the SGD-based sampler to solve the full problem."
                    },
                    {
                        "title": "Contrasting the landscape of contrastive and non-contrastive learning",
                        "abstract": "A lot of recent advances in unsupervised feature learning are based on designing features which are invariant under semantic data augmentations. A common way to do this is contrastive learning, which uses positive and negative samples. Some recent works however have shown promising results for non-contrastive learning, which does not require negative samples. However, the non-contrastive losses have obvious\"collapsed\"minima, in which the encoders output a constant feature embedding, independent of the input. A folk conjecture is that so long as these collapsed solutions are avoided, the produced feature representations should be good. In our paper, we cast doubt on this story: we show through theoretical results and controlled experiments that even on simple data models, non-contrastive losses have a preponderance of non-collapsed bad minima. Moreover, we show that the training process does not avoid these minima."
                    },
                    {
                        "title": "Neural Network Approximations of PDEs Beyond Linearity: Representational Perspective",
                        "abstract": "A burgeoning line of research leverages deep neural networks to approximate the solutions to high dimensional PDEs, opening lines of theoretical inquiry focused on explaining how it is that these models appear to evade the curse of dimensionality. However, most prior theoretical analyses have been limited to linear PDEs. In this work, we take a step towards studying the representational power of neural networks for approximating solutions to nonlinear PDEs. We focus on a class of PDEs known as \\emph{nonlinear elliptic variational PDEs}, whose solutions minimize an \\emph{Euler-Lagrange} energy functional $\\mathcal{E}(u) = \\int_\\Omega L(x, u(x), \\nabla u(x)) - f(x) u(x)dx$. We show that if composing a function with Barron norm $b$ with partial derivatives of $L$ produces a function of Barron norm at most $B_L b^p$, the solution to the PDE can be $\\epsilon$-approximated in the $L^2$ sense by a function with Barron norm $O\\left(\\left(dB_L\\right)^{\\max\\{p \\log(1/ \\epsilon), p^{\\log(1/\\epsilon)}\\}}\\right)$. By a classical result due to Barron [1993], this correspondingly bounds the size of a 2-layer neural network needed to approximate the solution. Treating $p, \\epsilon, B_L$ as constants, this quantity is polynomial in dimension, thus showing neural networks can evade the curse of dimensionality. Our proof technique involves neurally simulating (preconditioned) gradient in an appropriate Hilbert space, which converges exponentially fast to the solution of the PDE, and such that we can bound the increase of the Barron norm at each iterate. Our results subsume and substantially generalize analogous prior results for linear elliptic PDEs over a unit hypercube."
                    },
                    {
                        "title": "Masked prediction tasks: a parameter identifiability view",
                        "abstract": "The vast majority of work in self-supervised learning, both theoretical and empirical (though mostly the latter), have largely focused on recovering good features for downstream tasks, with the definition of\"good\"often being intricately tied to the downstream task itself. This lens is undoubtedly very interesting, but suffers from the problem that there isn't a\"canonical\"set of downstream tasks to focus on -- in practice, this problem is usually resolved by competing on the benchmark dataset du jour. In this paper, we present an alternative lens: one of parameter identifiability. More precisely, we consider data coming from a parametric probabilistic model, and train a self-supervised learning predictor with a suitably chosen parametric form. Then, we ask whether we can read off the ground truth parameters of the probabilistic model from the optimal predictor. We focus on the widely used self-supervised learning method of predicting masked tokens, which is popular for both natural languages and visual data. While incarnations of this approach have already been successfully used for simpler probabilistic models (e.g. learning fully-observed undirected graphical models), we focus instead on latent-variable models capturing sequential structures -- namely Hidden Markov Models with both discrete and conditionally Gaussian observations. We show that there is a rich landscape of possibilities, out of which some prediction tasks yield identifiability, while others do not. Our results, borne of a theoretical grounding of self-supervised learning, could thus potentially beneficially inform practice. Moreover, we uncover close connections with uniqueness of tensor rank decompositions -- a widely used tool in studying identifiability through the lens of the method of moments."
                    },
                    {
                        "title": "Masked Prediction: A Parameter Identifiability View",
                        "abstract": "The vast majority of work in self-supervised learning have focused on assessing recovered features by a chosen set of downstream tasks. While there are several commonly used benchmark datasets, this lens of feature learning requires assumptions on the downstream tasks which are not inherent to the data distribution itself. In this paper, we present an alternative lens, one of parameter identi\ufb01ability: assuming data comes from a parametric probabilistic model, we train a self-supervised learning predictor with a suitable parametric form, and ask whether the parameters of the optimal predictor can be used to extract the parameters of the ground truth generative model. Speci\ufb01cally, we focus on latent-variable models capturing sequential structures, namely Hidden Markov Models with both discrete and conditionally Gaussian observations. We focus on masked prediction as the self-supervised learning task and study the optimal masked predictor. We show that parameter identi\ufb01ability is governed by the task dif\ufb01culty, which is determined by the choice of data model and the amount of tokens to predict. Technique-wise, we uncover close connections with the uniqueness of tensor rank decompositions , a widely used tool in studying identi\ufb01ability through the lens of the method of moments."
                    }
                ]
            }
        }
    },
    "2405.15393": {
        "paper_data": {
            "title": "Reshuffling Resampling Splits Can Improve Generalization of Hyperparameter Optimization",
            "url": "http://arxiv.org/abs/2405.15393v1",
            "arxiv_id": "2405.15393",
            "authors": [
                "Thomas Nagler",
                "Lennart Schneider",
                "Bernd Bischl",
                "Matthias Feurer"
            ],
            "abstract": "Hyperparameter optimization is crucial for obtaining peak performance of machine learning models. The standard protocol evaluates various hyperparameter configurations using a resampling estimate of the generalization error to guide optimization and select a final hyperparameter configuration. Without much evidence, paired resampling splits, i.e., either a fixed train-validation split or a fixed cross-validation scheme, are often recommended. We show that, surprisingly, reshuffling the splits for every configuration often improves the final model's generalization performance on unseen data. Our theoretical analysis explains how reshuffling affects the asymptotic behavior of the validation loss surface and provides a bound on the expected regret in the limiting regime. This bound connects the potential benefits of reshuffling to the signal and noise characteristics of the underlying optimization problem. We confirm our theoretical results in a controlled simulation study and demonstrate the practical usefulness of reshuffling in a large-scale, realistic hyperparameter optimization experiment. While reshuffling leads to test performances that are competitive with using fixed splits, it drastically improves results for a single train-validation holdout protocol and can often make holdout become competitive with standard CV while being computationally cheaper.",
            "introduction": "   1 Introduction  Hyperparameters have been shown to strongly influence the performance of machine learning models (van Rijn & Hutter, 2018; Probst et\u00a0al., 2019). The primary objective of hyperparameter optimization (HPO; also called tuning) is the identification and selection of a hyperparameter configuration (HPC) that minimizes the estimated generalization error (Feurer & Hutter, 2019; Bischl et\u00a0al., 2023). Typically, this task is challenged by the absence of a closed-form mathematical description of the objective function, the unavailability of an analytic gradient, and the large cost to evaluate HPCs, categorizing HPO as a noisy, black-box optimization problem. An HPC is evaluated via resampling, such as a holdout split or M\ud835\udc40Mitalic_M-fold cross-validation (CV), during tuning.   These resampling splits are usually constructed in a fixed and instantiated manner, i.e., the same training and validation splits are used for the internal evaluation of all configurations. On the one hand, this is an intuitive approach, as it should facilitate a fair comparison between HPCs and reduce the variance in the comparison.111This approach likely originates from the concept of paired statistical tests and the resulting variance reduction, but in our literature search we did not find any references discussing this in the context of HPO. For example, when comparing the performance of two classifiers on one dataset, paired tests are commonly employed that implicitly assume that differences between the performance of classifiers on a given CV fold are comparable (Dietterich, 1998; Nadeau & Bengio, 1999, 2003; Dem\u0161ar, 2006). On the other hand, such a fixing of train and validation splits might steer the optimization, especially after a substantial budget of evaluations, towards favoring HPCs which are specifically tailored to the chosen splits. Such and related effects, where we \"overoptimize\" the validation performance without effective reward in improved generalization performance have been sometimes dubbed \"overtuning\" or \"oversearching\". For a more detailed discussion of this topic, including related work, see Section\u00a05 and Appendix\u00a0A. The practice of reshuffling resampling splits during HPO is generally neither discussed in the scientific literature nor HPO software tools.222In Appendix\u00a0A, we present an overview of how resampling is addressed in tutorials and examples of standard HPO libraries and software. We conclude that usually fixed splits are used or recommended. To the best of our knowledge, only L\u00e9vesque (2018) investigated reshuffling train-validation splits for every new HPC. For both holdout and M\ud835\udc40Mitalic_M-fold CV using reshuffled resampling splits resulted in, on average, slightly lower generalization error when used in combination with Bayesian optimization\u00a0(BO, Garnett, 2023) or CMA-ES\u00a0(Hansen & Ostermeier, 2001) as HPO algorithms. ML-Plan\u00a0(Mohr et\u00a0al., 2018) proposes a two-stage procedure. In a first stage, the tool uses planning on hierarchical task networks to find promising machine learning pipelines. In a second step, it reshuffles the data and retrains the most promising candidates from the search step to make a more unbiased selection of the final model. Lastly, in the context of evolutionary optimization, reshuffling was applied after every generation\u00a0(Larcher & Barbosa, 2022).   In this paper, we systematically examine the effect of reshuffling on HPO performance. Our contributions can be summarized as follows:   1.  We show theoretically that reshuffling resampling splits during HPO can result in finding a configuration with better overall generalization performance, especially when the loss surface is rather",
            "references": [
                {
                    "title": "Don't Waste Your Time: Early Stopping Cross-Validation",
                    "abstract": "State-of-the-art automated machine learning systems for tabular data often employ cross-validation; ensuring that measured performances generalize to unseen data, or that subsequent ensembling does not overfit. However, using k-fold cross-validation instead of holdout validation drastically increases the computational cost of validating a single configuration. While ensuring better generalization and, by extension, better performance, the additional cost is often prohibitive for effective model selection within a time budget. We aim to make model selection with cross-validation more effective. Therefore, we study early stopping the process of cross-validation during model selection. We investigate the impact of early stopping on random search for two algorithms, MLP and random forest, across 36 classification datasets. We further analyze the impact of the number of folds by considering 3-, 5-, and 10-folds. In addition, we investigate the impact of early stopping with Bayesian optimization instead of random search and also repeated cross-validation. Our exploratory study shows that even a simple-to-understand and easy-to-implement method consistently allows model selection to converge faster; in ~94% of all datasets, on average by ~214%. Moreover, stopping cross-validation enables model selection to explore the search space more exhaustively by considering +167% configurations on average within one hour, while also obtaining better overall performance."
                },
                {
                    "title": "A comparison of hyperparameter tuning procedures for clinical prediction models: A simulation study",
                    "abstract": "Tuning hyperparameters, such as the regularization parameter in Ridge or Lasso regression, is often aimed at improving the predictive performance of risk prediction models. In this study, various hyperparameter tuning procedures for clinical prediction models were systematically compared and evaluated in low\u2010dimensional data. The focus was on out\u2010of\u2010sample predictive performance (discrimination, calibration, and overall prediction error) of risk prediction models developed using Ridge, Lasso, Elastic Net, or Random Forest. The influence of sample size, number of predictors and events fraction on performance of the hyperparameter tuning procedures was studied using extensive simulations. The results indicate important differences between tuning procedures in calibration performance, while generally showing similar discriminative performance. The one\u2010standard\u2010error rule for tuning applied to cross\u2010validation (1SE CV) often resulted in severe miscalibration. Standard non\u2010repeated and repeated cross\u2010validation (both 5\u2010fold and 10\u2010fold) performed similarly well and outperformed the other tuning procedures. Bootstrap showed a slight tendency to more severe miscalibration than standard cross\u2010validation\u2010based tuning procedures. Differences between tuning procedures were larger for smaller sample sizes, lower events fractions and fewer predictors. These results imply that the choice of tuning procedure can have a profound influence on the predictive performance of prediction models. The results support the application of standard 5\u2010fold or 10\u2010fold cross\u2010validation that minimizes out\u2010of\u2010sample prediction error. Despite an increased computational burden, we found no clear benefit of repeated over non\u2010repeated cross\u2010validation for hyperparameter tuning. We warn against the potentially detrimental effects on model calibration of the popular 1SE CV rule for tuning prediction models in low\u2010dimensional settings."
                },
                {
                    "title": "CMA-ES for Post Hoc Ensembling in AutoML: A Great Success and Salvageable Failure",
                    "abstract": "Many state-of-the-art automated machine learning (AutoML) systems use greedy ensemble selection (GES) by Caruana et al. (2004) to ensemble models found during model selection post hoc. Thereby, boosting predictive performance and likely following Auto-Sklearn 1's insight that alternatives, like stacking or gradient-free numerical optimization, overfit. Overfitting in Auto-Sklearn 1 is much more likely than in other AutoML systems because it uses only low-quality validation data for post hoc ensembling. Therefore, we were motivated to analyze whether Auto-Sklearn 1's insight holds true for systems with higher-quality validation data. Consequently, we compared the performance of covariance matrix adaptation evolution strategy (CMA-ES), state-of-the-art gradient-free numerical optimization, to GES on the 71 classification datasets from the AutoML benchmark for AutoGluon. We found that Auto-Sklearn's insight depends on the chosen metric. For the metric ROC AUC, CMA-ES overfits drastically and is outperformed by GES -- statistically significantly for multi-class classification. For the metric balanced accuracy, CMA-ES does not overfit and outperforms GES significantly. Motivated by the successful application of CMA-ES for balanced accuracy, we explored methods to stop CMA-ES from overfitting for ROC AUC. We propose a method to normalize the weights produced by CMA-ES, inspired by GES, that avoids overfitting for CMA-ES and makes CMA-ES perform better than or similar to GES for ROC AUC."
                },
                {
                    "title": "PriorBand: Practical Hyperparameter Optimization in the Age of Deep Learning",
                    "abstract": "Hyperparameters of Deep Learning (DL) pipelines are crucial for their downstream performance. While a large number of methods for Hyperparameter Optimization (HPO) have been developed, their incurred costs are often untenable for modern DL. Consequently, manual experimentation is still the most prevalent approach to optimize hyperparameters, relying on the researcher's intuition, domain knowledge, and cheap preliminary explorations. To resolve this misalignment between HPO algorithms and DL researchers, we propose PriorBand, an HPO algorithm tailored to DL, able to utilize both expert beliefs and cheap proxy tasks. Empirically, we demonstrate PriorBand's efficiency across a range of DL benchmarks and show its gains under informative expert input and robustness against poor expert beliefs"
                },
                {
                    "title": "When Do Neural Nets Outperform Boosted Trees on Tabular Data?",
                    "abstract": "Tabular data is one of the most commonly used types of data in machine learning. Despite recent advances in neural nets (NNs) for tabular data, there is still an active discussion on whether or not NNs generally outperform gradient-boosted decision trees (GBDTs) on tabular data, with several recent works arguing either that GBDTs consistently outperform NNs on tabular data, or vice versa. In this work, we take a step back and question the importance of this debate. To this end, we conduct the largest tabular data analysis to date, comparing 19 algorithms across 176 datasets, and we find that the 'NN vs. GBDT' debate is overemphasized: for a surprisingly high number of datasets, either the performance difference between GBDTs and NNs is negligible, or light hyperparameter tuning on a GBDT is more important than choosing between NNs and GBDTs. A remarkable exception is the recently-proposed prior-data fitted network, TabPFN: although it is effectively limited to training sets of size 3000, we find that it outperforms all other algorithms on average, even when randomly sampling 3000 training datapoints. Next, we analyze dozens of metafeatures to determine what properties of a dataset make NNs or GBDTs better-suited to perform well. For example, we find that GBDTs are much better than NNs at handling skewed or heavy-tailed feature distributions and other forms of dataset irregularities. Our insights act as a guide for practitioners to determine which techniques may work best on their dataset. Finally, with the goal of accelerating tabular data research, we release the TabZilla Benchmark Suite: a collection of the 36 'hardest' of the datasets we study. Our benchmark suite, codebase, and all raw results are available at https://github.com/naszilla/tabzilla."
                },
                {
                    "title": "Scaling Laws for Hyperparameter Optimization",
                    "abstract": "Hyperparameter optimization is an important subfield of machine learning that focuses on tuning the hyperparameters of a chosen algorithm to achieve peak performance. Recently, there has been a stream of methods that tackle the issue of hyperparameter optimization, however, most of the methods do not exploit the dominant power law nature of learning curves for Bayesian optimization. In this work, we propose Deep Power Laws (DPL), an ensemble of neural network models conditioned to yield predictions that follow a power-law scaling pattern. Our method dynamically decides which configurations to pause and train incrementally by making use of gray-box evaluations. We compare our method against 7 state-of-the-art competitors on 3 benchmarks related to tabular, image, and NLP datasets covering 59 diverse tasks. Our method achieves the best results across all benchmarks by obtaining the best any-time results compared to all competitors."
                },
                {
                    "title": "AMLB: an AutoML Benchmark",
                    "abstract": "Comparing different AutoML frameworks is notoriously challenging and often done incorrectly. We introduce an open and extensible benchmark that follows best practices and avoids common mistakes when comparing AutoML frameworks. We conduct a thorough comparison of 9 well-known AutoML frameworks across 71 classification and 33 regression tasks. The differences between the AutoML frameworks are explored with a multi-faceted analysis, evaluating model accuracy, its trade-offs with inference time, and framework failures. We also use Bradley-Terry trees to discover subsets of tasks where the relative AutoML framework rankings differ. The benchmark comes with an open-source tool that integrates with many AutoML frameworks and automates the empirical evaluation process end-to-end: from framework installation and resource allocation to in-depth evaluation. The benchmark uses public data sets, can be easily extended with other AutoML frameworks and tasks, and has a website with up-to-date results."
                },
                {
                    "title": "Deep Neural Networks and Tabular Data: A Survey",
                    "abstract": "Heterogeneous tabular data are the most commonly used form of data and are essential for numerous critical and computationally demanding applications. On homogeneous datasets, deep neural networks have repeatedly shown excellent performance and have therefore been widely adopted. However, their adaptation to tabular data for inference or data generation tasks remains highly challenging. To facilitate further progress in the field, this work provides an overview of state-of-the-art deep learning methods for tabular data. We categorize these methods into three groups: data transformations, specialized architectures, and regularization models. For each of these groups, our work offers a comprehensive overview of the main approaches. Moreover, we discuss deep learning approaches for generating tabular data and also provide an overview over strategies for explaining deep models on tabular data. Thus, our first contribution is to address the main research streams and existing methodologies in the mentioned areas while highlighting relevant challenges and open research questions. Our second contribution is to provide an empirical comparison of traditional machine learning methods with 11 deep learning approaches across five popular real-world tabular datasets of different sizes and with different learning objectives. Our results, which we have made publicly available as competitive benchmarks, indicate that algorithms based on gradient-boosted tree ensembles still mostly outperform deep learning models on supervised learning tasks, suggesting that the research progress on competitive deep learning models for tabular data is stagnating. To the best of our knowledge, this is the first in-depth overview of deep learning approaches for tabular data; as such, this work can serve as a valuable starting point to guide researchers and practitioners interested in deep learning with tabular data."
                },
                {
                    "title": "SMAC3: A Versatile Bayesian Optimization Package for Hyperparameter Optimization",
                    "abstract": "Algorithm parameters, in particular hyperparameters of machine learning algorithms, can substantially impact their performance. To support users in determining well-performing hyperparameter configurations for their algorithms, datasets and applications at hand, SMAC3 offers a robust and flexible framework for Bayesian Optimization, which can improve performance within a few evaluations. It offers several facades and pre-sets for typical use cases, such as optimizing hyperparameters, solving low dimensional continuous (artificial) global optimization problems and configuring algorithms to perform well across multiple problem instances. The SMAC3 package is available under a permissive BSD-license at https://github.com/automl/SMAC3."
                },
                {
                    "title": "HPOBench: A Collection of Reproducible Multi-Fidelity Benchmark Problems for HPO",
                    "abstract": "To achieve peak predictive performance, hyperparameter optimization (HPO) is a crucial component of machine learning and its applications. Over the last years, the number of efficient algorithms and tools for HPO grew substantially. At the same time, the community is still lacking realistic, diverse, computationally cheap, and standardized benchmarks. This is especially the case for multi-fidelity HPO methods. To close this gap, we propose HPOBench, which includes 7 existing and 5 new benchmark families, with a total of more than 100 multi-fidelity benchmark problems. HPOBench allows to run this extendable set of multi-fidelity HPO benchmarks in a reproducible way by isolating and packaging the individual benchmarks in containers. It also provides surrogate and tabular benchmarks for computationally affordable yet statistically sound evaluations. To demonstrate HPOBench's broad compatibility with various optimization tools, as well as its usefulness, we conduct an exemplary large-scale study evaluating 13 optimizers from 6 optimization tools. We provide HPOBench here: https://github.com/automl/HPOBench."
                },
                {
                    "title": "YAHPO Gym - An Efficient Multi-Objective Multi-Fidelity Benchmark for Hyperparameter Optimization",
                    "abstract": "When developing and analyzing new hyperparameter optimization methods, it is vital to empirically evaluate and compare them on well-curated benchmark suites. In this work, we propose a new set of challenging and relevant benchmark problems motivated by desirable properties and requirements for such benchmarks. Our new surrogate-based benchmark collection consists of 14 scenarios that in total constitute over 700 multi-fidelity hyperparameter optimization problems, which all enable multi-objective hyperparameter optimization. Furthermore, we empirically compare surrogate-based benchmarks to the more widely-used tabular benchmarks, and demonstrate that the latter may produce unfaithful results regarding the performance ranking of HPO methods. We examine and compare our benchmark collection with respect to defined requirements and propose a single-objective as well as a multi-objective benchmark suite on which we compare 7 single-objective and 7 multi-objective optimizers in a benchmark experiment. Our software is available at [https://github.com/slds-lmu/yahpo_gym]."
                },
                {
                    "title": "Hyperparameter optimization: Foundations, algorithms, best practices, and open challenges",
                    "abstract": "Most machine learning algorithms are configured by a set of hyperparameters whose values must be carefully chosen and which often considerably impact performance. To avoid a time\u2010consuming and irreproducible manual process of trial\u2010and\u2010error to find well\u2010performing hyperparameter configurations, various automatic hyperparameter optimization (HPO) methods\u2014for example, based on resampling error estimation for supervised machine learning\u2014can be employed. After introducing HPO from a general perspective, this paper reviews important HPO methods, from simple techniques such as grid or random search to more advanced methods like evolution strategies, Bayesian optimization, Hyperband, and racing. This work gives practical recommendations regarding important choices to be made when conducting HPO, including the HPO algorithms themselves, performance evaluation, how to combine HPO with machine learning pipelines, runtime improvements, and parallelization."
                },
                {
                    "title": "HPO-B: A Large-Scale Reproducible Benchmark for Black-Box HPO based on OpenML",
                    "abstract": "Hyperparameter optimization (HPO) is a core problem for the machine learning community and remains largely unsolved due to the significant computational resources required to evaluate hyperparameter configurations. As a result, a series of recent related works have focused on the direction of transfer learning for quickly fine-tuning hyperparameters on a dataset. Unfortunately, the community does not have a common large-scale benchmark for comparing HPO algorithms. Instead, the de facto practice consists of empirical protocols on arbitrary small-scale meta-datasets that vary inconsistently across publications, making reproducibility a challenge. To resolve this major bottleneck and enable a fair and fast comparison of black-box HPO methods on a level playing field, we propose HPO-B, a new large-scale benchmark in the form of a collection of meta-datasets. Our benchmark is assembled and preprocessed from the OpenML repository and consists of 176 search spaces (algorithms) evaluated sparsely on 196 datasets with a total of 6.4 million hyperparameter evaluations. For ensuring reproducibility on our benchmark, we detail explicit experimental protocols, splits, and evaluation measures for comparing methods for both non-transfer, as well as, transfer learning HPO."
                },
                {
                    "title": "Bayesian Optimization is Superior to Random Search for Machine Learning Hyperparameter Tuning: Analysis of the Black-Box Optimization Challenge 2020",
                    "abstract": "This paper presents the results and insights from the black-box optimization (BBO) challenge at NeurIPS 2020 which ran from July-October, 2020. The challenge emphasized the importance of evaluating derivative-free optimizers for tuning the hyperparameters of machine learning models. This was the first black-box optimization challenge with a machine learning emphasis. It was based on tuning (validation set) performance of standard machine learning models on real datasets. This competition has widespread impact as black-box optimization (e.g., Bayesian optimization) is relevant for hyperparameter tuning in almost every machine learning project as well as many applications outside of machine learning. The final leaderboard was determined using the optimization performance on held-out (hidden) objective functions, where the optimizers ran without human intervention. Baselines were set using the default settings of several open-source black-box optimization packages as well as random search."
                },
                {
                    "title": "Automatic Termination for Hyperparameter Optimization",
                    "abstract": "Bayesian optimization (BO) is a widely popular approach for the hyperparameter optimization (HPO) in machine learning. At its core, BO iteratively evaluates promising configurations until a user-defined budget, such as wall-clock time or number of iterations, is exhausted. While the final performance after tuning heavily depends on the provided budget, it is hard to pre-specify an optimal value in advance. In this work, we propose an effective and intuitive termination criterion for BO that automatically stops the procedure if it is sufficiently close to the global optimum. Our key insight is that the discrepancy between the true objective (predictive performance on test data) and the computable target (validation performance) suggests stopping once the suboptimality in optimizing the target is dominated by the statistical estimation error. Across an extensive range of real-world HPO problems and baselines, we show that our termination criterion achieves a better trade-off between the test performance and optimization time. Additionally, we find that overfitting may occur in the context of HPO, which is arguably an overlooked problem in the literature, and show how our termination criterion helps to mitigate this phenomenon on both small and large datasets."
                },
                {
                    "title": "Auto-Pytorch: Multi-Fidelity MetaLearning for Efficient and Robust AutoDL",
                    "abstract": "While early AutoML frameworks focused on optimizing traditional ML pipelines and their hyperparameters, a recent trend in AutoML is to focus on neural architecture search. In this paper, we introduce Auto-PyTorch, which brings together the best of these two worlds by jointly and robustly optimizing the network architecture and the training hyperparameters to enable fully automated deep learning (AutoDL). Auto-PyTorch achieves state-of-the-art performance on several tabular benchmarks by combining multi-fidelity optimization with portfolio construction for warmstarting and ensembling of deep neural networks (DNNs) and common baselines for tabular data. To thoroughly study our assumptions on how to design such an AutoDL system, we additionally introduce a new benchmark on learning curves for DNNs, dubbed LCBench, and run extensive ablation studies of the full Auto-PyTorch on typical AutoML benchmarks, eventually showing that Auto-PyTorch performs better than several state-of-the-art competitors."
                },
                {
                    "title": "Accounting for Variance in Machine Learning Benchmarks",
                    "abstract": "Strong empirical evidence that one machine-learning algorithm A outperforms another one B ideally calls for multiple trials optimizing the learning pipeline over sources of variation such as data sampling, data augmentation, parameter initialization, and hyperparameters choices. This is prohibitively expensive, and corners are cut to reach conclusions. We model the whole benchmarking process, revealing that variance due to data sampling, parameter initialization and hyperparameter choice impact markedly the results. We analyze the predominant comparison methods used today in the light of this variance. We show a counter-intuitive result that adding more sources of variation to an imperfect estimator approaches better the ideal estimator at a 51 times reduction in compute cost. Building on these results, we study the error rate of detecting improvements, on five different deep-learning tasks/architectures. This study leads us to propose recommendations for performance comparisons."
                },
                {
                    "title": "Cross-validation Confidence Intervals for Test Error",
                    "abstract": "This work develops central limit theorems for cross-validation and consistent estimators of its asymptotic variance under weak stability conditions on the learning algorithm. Together, these results provide practical, asymptotically-exact confidence intervals for $k$-fold test error and valid, powerful hypothesis tests of whether one learning algorithm has smaller $k$-fold test error than another. These results are also the first of their kind for the popular choice of leave-one-out cross-validation. In our real-data experiments with diverse learning algorithms, the resulting intervals and tests outperform the most popular alternative methods from the literature."
                },
                {
                    "title": "Auto-Sklearn 2.0: Hands-free AutoML via Meta-Learning",
                    "abstract": "Automated Machine Learning (AutoML) supports practitioners and researchers with the tedious task of designing machine learning pipelines and has recently achieved substantial success. In this paper, we introduce new AutoML approaches motivated by our winning submission to the second ChaLearn AutoML challenge. We develop PoSH Auto-sklearn, which enables AutoML systems to work well on large datasets under rigid time limits by using a new, simple and meta-feature-free meta-learning technique and by employing a successful bandit strategy for budget allocation. However, PoSH Auto-sklearn introduces even more ways of running AutoML and might make it harder for users to set it up correctly. Therefore, we also go one step further and study the design space of AutoML itself, proposing a solution towards truly hands-free AutoML. Together, these changes give rise to the next generation of our AutoML system, Auto-sklearn 2.0. We verify the improvements by these additions in an extensive experimental study on 39 AutoML benchmark datasets. We conclude the paper by comparing to other popular AutoML frameworks and Auto-sklearn 1.0, reducing the relative error by up to a factor of 4.5, and yielding a performance in 10 minutes that is substantially better than what Auto-sklearn 1.0 achieves within an hour."
                },
                {
                    "title": "Asymptotics of Cross-Validation",
                    "abstract": "Cross validation is a central tool in evaluating the performance of machine learning and statistical models. However, despite its ubiquitous role, its theoretical properties are still not well understood. We study the asymptotic properties of the cross validated-risk for a large class of models. Under stability conditions, we establish a central limit theorem and Berry-Esseen bounds, which enable us to compute asymptotically accurate confidence intervals. Using our results, we paint a big picture for the statistical speed-up of cross validation compared to a train-test split procedure. A corollary of our results is that parametric M-estimators (or empirical risk minimizers) benefit from the \"full\" speed-up when performing cross-validation under the training loss. In other common cases, such as when the training is done using a surrogate loss or a regularizer, we show that the behavior of the cross-validated risk is complex with a variance reduction which may be smaller or larger than the \"full\" speed-up, depending on the model and the underlying distribution. We allow the number of folds to grow with the number of observations at any rate."
                },
                {
                    "title": "Fast classification rates without standard margin assumptions",
                    "abstract": "\n We consider the classical problem of learning rates for classes with finite Vapnik\u2013Chervonenkis (VC) dimension. It is well known that fast learning rates up to $O\\left (\\frac{d}{n}\\right )$ are achievable by the empirical risk minimization (ERM) algorithm if low noise or margin assumptions are satisfied. These usually require the optimal Bayes classifier to be in the class, and it has been shown that when this is not the case, the fast rates cannot be achieved even in the noise free case. In this paper, we further investigate the question of the fast rates in the agnostic setting, when the Bayes classifier is not in the class and the noise in the labeling is allowed. First, we consider classification with a reject option, namely Chow\u2019s reject option model, and show that by slightly lowering the impact of hard instances, a learning rate of order $O\\left (\\frac{d}{n}\\log \\frac{n}{d}\\right )$ is always achievable in the agnostic setting by a specific learning algorithm. Similar results were only known under special versions of margin assumptions. As an auxiliary result, we show that under the Bernstein assumption the performance of the proposed algorithm is never worse than the performance of ERM even if some of the labels are predicted at random. Based on those results, we derive the necessary and sufficient conditions for classification (without a reject option) with fast rates in the agnostic setting achievable by improper learners. This simultaneously extends the work by Massart & N\u00e9d\u00e9lec (2006, Ann. Stat., 34, 2326\u20132366), which studied this question in the case where the Bayesian optimal rule belongs to the class, and the work by Ben-David and Urner (2014, COLT, pp. 527\u2013542 ), which allows the misspecification but is limited to the no noise setting. Our result also provides the first general setup in statistical learning theory in which an improper learning algorithm may significantly improve the learning rate for non-convex losses."
                },
                {
                    "title": "Practical Multi-fidelity Bayesian Optimization for Hyperparameter Tuning",
                    "abstract": "Bayesian optimization is popular for optimizing time-consuming black-box objectives. Nonetheless, for hyperparameter tuning in deep neural networks, the time required to evaluate the validation error for even a few hyperparameter settings remains a bottleneck. Multi-fidelity optimization promises relief using cheaper proxies to such objectives --- for example, validation error for a network trained using a subset of the training points or fewer iterations than required for convergence. We propose a highly flexible and practical approach to multi-fidelity Bayesian optimization, focused on efficiently optimizing hyperparameters for iteratively trained supervised learning models. We introduce a new acquisition function, the trace-aware knowledge-gradient, which efficiently leverages both multiple continuous fidelity controls and trace observations --- values of the objective at a sequence of fidelities, available when varying fidelity using training iterations. We provide a provably convergent method for optimizing our acquisition function and show it outperforms state-of-the-art alternatives for hyperparameter tuning of deep neural networks and large-scale kernel learning."
                },
                {
                    "title": "BOHB: Robust and Efficient Hyperparameter Optimization at Scale",
                    "abstract": "Modern deep learning methods are very sensitive to many hyperparameters, and, due to the long training times of state-of-the-art models, vanilla Bayesian hyperparameter optimization is typically computationally infeasible. On the other hand, bandit-based configuration evaluation approaches based on random search lack guidance and do not converge to the best configurations as quickly. Here, we propose to combine the benefits of both Bayesian optimization and bandit-based methods, in order to achieve the best of both worlds: strong anytime performance and fast convergence to optimal configurations. We propose a new practical state-of-the-art hyperparameter optimization method, which consistently outperforms both Bayesian optimization and Hyperband on a wide range of problem types, including high-dimensional toy functions, support vector machines, feed-forward neural networks, Bayesian neural networks, deep reinforcement learning, and convolutional neural networks. Our method is robust and versatile, while at the same time being conceptually simple and easy to implement."
                },
                {
                    "title": "Tunability: Importance of Hyperparameters of Machine Learning Algorithms",
                    "abstract": "Modern supervised machine learning algorithms involve hyperparameters that have to be set before running them. Options for setting hyperparameters are default values from the software package, manual configuration by the user or configuring them for optimal predictive performance by a tuning procedure. The goal of this paper is two-fold. Firstly, we formalize the problem of tuning from a statistical point of view, define data-based defaults and suggest general measures quantifying the tunability of hyperparameters of algorithms. Secondly, we conduct a large-scale benchmarking study based on 38 datasets from the OpenML platform and six common machine learning algorithms. We apply our measures to assess the tunability of their parameters. Our results yield default values for hyperparameters and enable users to decide whether it is worth conducting a possibly time consuming tuning strategy, to focus on the most important hyperparameters and to chose adequate hyperparameter spaces for tuning."
                },
                {
                    "title": "Hyperparameter Importance Across Datasets",
                    "abstract": "With the advent of automated machine learning, automated hyperparameter optimization methods are by now routinely used in data mining. However, this progress is not yet matched by equal progress on automatic analyses that yield information beyond performance-optimizing hyperparameter settings. In this work, we aim to answer the following two questions: Given an algorithm, what are generally its most important hyperparameters, and what are typically good values for these? We present methodology and a framework to answer these questions based on meta-learning across many datasets. We apply this methodology using the experimental meta-data available on OpenML to determine the most important hyperparameters of support vector machines, random forests and Adaboost, and to infer priors for all their hyperparameters. The results, obtained fully automatically, provide a quantitative basis to focus efforts in both manual algorithm design and in automated hyperparameter optimization. The conducted experiments confirm that the hyperparameters selected by the proposed method are indeed the most important ones and that the obtained priors also lead to statistically significant improvements in hyperparameter optimization."
                },
                {
                    "title": "Pitfalls and Best Practices in Algorithm Configuration",
                    "abstract": "Good parameter settings are crucial to achieve high performance in many areas of artificial intelligence (AI), such as propositional satisfiability solving, AI planning, scheduling, and machine learning (in particular deep learning). Automated algorithm configuration methods have recently received much attention in the AI community since they replace tedious, irreproducible and error-prone manual parameter tuning and can lead to new state-of-the-art performance. However, practical applications of algorithm configuration are prone to several (often subtle) pitfalls in the experimental design that can render the procedure ineffective. We identify several common issues and propose best practices for avoiding them. As one possibility for automatically handling as many of these as possible, we also propose a tool called GenericWrapper4AC."
                },
                {
                    "title": "Fast Bayesian Optimization of Machine Learning Hyperparameters on Large Datasets",
                    "abstract": "Bayesian optimization has become a successful tool for hyperparameter optimization of machine learning algorithms, such as support vector machines or deep neural networks. Despite its success, for large datasets, training and validating a single configuration often takes hours, days, or even weeks, which limits the achievable performance. To accelerate hyperparameter optimization, we propose a generative model for the validation error as a function of training set size, which is learned during the optimization process and allows exploration of preliminary configurations on small subsets, by extrapolating to the full dataset. We construct a Bayesian optimization procedure, dubbed Fabolas, which models loss and training time as a function of dataset size and automatically trades off high information gain about the global optimum against computational cost. Experiments optimizing support vector machines and deep neural networks show that Fabolas often finds high-quality solutions 10 to 100 times faster than other state-of-the-art Bayesian optimization methods or the recently proposed bandit strategy Hyperband."
                },
                {
                    "title": "Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization",
                    "abstract": "Performance of machine learning algorithms depends critically on identifying a good set of hyperparameters. While recent approaches use Bayesian optimization to adaptively select configurations, we focus on speeding up random search through adaptive resource allocation and early-stopping. We formulate hyperparameter optimization as a pure-exploration non-stochastic infinite-armed bandit problem where a predefined resource like iterations, data samples, or features is allocated to randomly sampled configurations. We introduce a novel algorithm, Hyperband, for this framework and analyze its theoretical properties, providing several desirable guarantees. Furthermore, we compare Hyperband with popular Bayesian optimization methods on a suite of hyperparameter optimization problems. We observe that Hyperband can provide over an order-of-magnitude speedup over our competitor set on a variety of deep-learning and kernel-based learning problems."
                },
                {
                    "title": "XGBoost: A Scalable Tree Boosting System",
                    "abstract": "Tree boosting is a highly effective and widely used machine learning method. In this paper, we describe a scalable end-to-end tree boosting system called XGBoost, which is used widely by data scientists to achieve state-of-the-art results on many machine learning challenges. We propose a novel sparsity-aware algorithm for sparse data and weighted quantile sketch for approximate tree learning. More importantly, we provide insights on cache access patterns, data compression and sharding to build a scalable tree boosting system. By combining these insights, XGBoost scales beyond billions of examples using far fewer resources than existing systems."
                },
                {
                    "title": "CMA-ES for Hyperparameter Optimization of Deep Neural Networks",
                    "abstract": "Hyperparameters of deep neural networks are often optimized by grid search, random search or Bayesian optimization. As an alternative, we propose to use the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), which is known for its state-of-the-art performance in derivative-free optimization. CMA-ES has some useful invariance properties and is friendly to parallel evaluations of solutions. We provide a toy example comparing CMA-ES and state-of-the-art Bayesian optimization algorithms for tuning the hyperparameters of a convolutional neural network for the MNIST dataset on 30 GPUs in parallel."
                },
                {
                    "title": "Mathematical Foundations of Infinite-Dimensional Statistical Models",
                    "abstract": "1. Nonparametric statistical models 2. Gaussian processes 3. Empirical processes 4. Function spaces and approximation theory 5. Linear nonparametric estimators 6. The minimax paradigm 7. Likelihood-based procedures 8. Adaptive inference."
                },
                {
                    "title": "Design of the 2015 ChaLearn AutoML challenge",
                    "abstract": "ChaLearn is organizing the Automatic Machine Learning (AutoML) contest for IJCNN 2015, which challenges participants to solve classification and regression problems without any human intervention. Participants' code is automatically run on the contest servers to train and test learning machines. However, there is no obligation to submit code; half of the prizes can be won by submitting prediction results only. Datasets of progressively increasing difficulty are introduced throughout the six rounds of the challenge. (Participants can enter the competition in any round.) The rounds alternate phases in which learners are tested on datasets participants have not seen, and phases in which participants have limited time to tweak their algorithms on those datasets to improve performance. This challenge will push the state of the art in fully automatic machine learning on a wide range of real-world problems. The platform will remain available beyond the termination of the challenge."
                },
                {
                    "title": "Fast rates in statistical and online learning",
                    "abstract": "The speed with which a learning algorithm converges as it is presented with more data is a central problem in machine learning -- a fast rate of convergence means less data is needed for the same level of performance. The pursuit of fast rates in online and statistical learning has led to the discovery of many conditions in learning theory under which fast learning is possible. We show that most of these conditions are special cases of a single, unifying condition, that comes in two forms: the central condition for 'proper' learning algorithms that always output a hypothesis in the given model, and stochastic mixability for online algorithms that may make predictions outside of the model. We show that under surprisingly weak assumptions both conditions are, in a certain sense, equivalent. The central condition has a re-interpretation in terms of convexity of a set of pseudoprobabilities, linking it to density estimation under misspecification. For bounded losses, we show how the central condition enables a direct proof of fast rates and we prove its equivalence to the Bernstein condition, itself a generalization of the Tsybakov margin condition, both of which have played a central role in obtaining fast rates in statistical learning. Yet, while the Bernstein condition is two-sided, the central condition is one-sided, making it more suitable to deal with unbounded losses. In its stochastic mixability form, our condition generalizes both a stochastic exp-concavity condition identified by Juditsky, Rigollet and Tsybakov and Vovk's notion of mixability. Our unifying conditions thus provide a substantial step towards a characterization of fast rates in statistical learning, similar to how classical mixability characterizes constant regret in the sequential prediction with expert advice setting."
                },
                {
                    "title": "Non-stochastic Best Arm Identification and Hyperparameter Optimization",
                    "abstract": "Motivated by the task of hyperparameter optimization, we introduce the non-stochastic best-arm identification problem. Within the multi-armed bandit literature, the cumulative regret objective enjoys algorithms and analyses for both the non-stochastic and stochastic settings while to the best of our knowledge, the best-arm identification framework has only been considered in the stochastic setting. We introduce the non-stochastic setting under this framework, identify a known algorithm that is well-suited for this setting, and analyze its behavior. Next, by leveraging the iterative nature of standard machine learning algorithms, we cast hyperparameter optimization as an instance of non-stochastic best-arm identification, and empirically evaluate our proposed algorithm on this task. Our empirical results show that, by allocating more resources to promising hyperparameter settings, we typically achieve comparable test accuracies an order of magnitude faster than baseline methods."
                },
                {
                    "title": "OpenML: networked science in machine learning",
                    "abstract": "Many sciences have made significant breakthroughs by adopting online tools that help organize, structure and mine information that is too detailed to be printed in journals. In this paper, we introduce OpenML, a place for machine learning researchers to share and organize data in fine detail, so that they can work more effectively, be more visible, and collaborate with others to tackle harder problems. We discuss how OpenML relates to other examples of networked science and what benefits it brings for machine learning research, individual scientists, as well as students and practitioners."
                },
                {
                    "title": "Freeze-Thaw Bayesian Optimization",
                    "abstract": "In this paper we develop a dynamic form of Bayesian optimization for machine learning models with the goal of rapidly finding good hyperparameter settings. Our method uses the partial information gained during the training of a machine learning model in order to decide whether to pause training and start a new model, or resume the training of a previously-considered model. We specifically tailor our method to machine learning problems by developing a novel positive-definite covariance kernel to capture a variety of training curves. Furthermore, we develop a Gaussian process prior that scales gracefully with additional temporal observations. Finally, we provide an information-theoretic framework to automate the decision process. Experiments on several common machine learning models show that our approach is extremely effective in practice."
                },
                {
                    "title": "A Note on Generalization Loss When Evolving Adaptive Pattern Recognition Systems",
                    "abstract": "Evolutionary computing provides powerful methods for designing pattern recognition systems. This design process is typically based on finite sample data and therefore bears the risk of overfitting. This paper aims at raising the awareness of various types of overfitting and at providing guidelines for how to deal with them. We restrict our considerations to the predominant scenario in which fitness computations are based on point estimates. Three different sources of losing generalization performance when evolving learning machines, namely overfitting to training, test, and final selection data, are identified, discussed, and experimentally demonstrated. The importance of a pristine hold-out data set for the selection of the final result from the evolved candidates is highlighted. It is shown that it may be beneficial to restrict this last selection process to a subset of the evolved candidates."
                },
                {
                    "title": "Auto-WEKA: combined selection and hyperparameter optimization of classification algorithms",
                    "abstract": "Many different machine learning algorithms exist; taking into account each algorithm's hyperparameters, there is a staggeringly large number of possible alternatives overall. We consider the problem of simultaneously selecting a learning algorithm and setting its hyperparameters, going beyond previous work that attacks these issues separately. We show that this problem can be addressed by a fully automated approach, leveraging recent innovations in Bayesian optimization. Specifically, we consider a wide range of feature selection techniques (combining 3 search and 8 evaluator methods) and all classification approaches implemented in WEKA's standard distribution, spanning 2 ensemble methods, 10 meta-methods, 27 base classifiers, and hyperparameter settings for each classifier. On each of 21 popular datasets from the UCI repository, the KDD Cup 09, variants of the MNIST dataset and CIFAR-10, we show classification performance often much better than using standard selection and hyperparameter optimization methods. We hope that our approach will help non-expert users to more effectively identify machine learning algorithms and hyperparameter settings appropriate to their applications, and hence to achieve improved performance."
                },
                {
                    "title": "Random Search for Hyper-Parameter Optimization",
                    "abstract": "Grid search and manual search are the most widely used strategies for hyper-parameter optimization. This paper shows empirically and theoretically that randomly chosen trials are more efficient for hyper-parameter optimization than trials on a grid. Empirical evidence comes from a comparison with a large previous study that used grid search and manual search to configure neural networks and deep belief networks. Compared with neural networks configured by a pure grid search, we find that random search over the same domain is able to find models that are as good or better within a small fraction of the computation time. Granting random search the same computational budget, random search finds better models by effectively searching a larger, less promising configuration space. Compared with deep belief networks configured by a thoughtful combination of manual search and grid search, purely random search over the same 32-dimensional configuration space found statistically equal performance on four of seven data sets, and superior performance on one of seven. A Gaussian process analysis of the function from hyper-parameters to validation set performance reveals that for most data sets only a few of the hyper-parameters really matter, but that different hyper-parameters are important on different data sets. This phenomenon makes grid search a poor choice for configuring algorithms for new data sets. Our analysis casts some light on why recent \"High Throughput\" methods achieve surprising success--they appear to search through a large number of hyper-parameters because most hyper-parameters do not matter much. We anticipate that growing interest in large hierarchical models will place an increasing burden on techniques for hyper-parameter optimization; this work shows that random search is a natural baseline against which to judge progress in the development of adaptive (sequential) hyper-parameter optimization algorithms."
                },
                {
                    "title": "On Over-fitting in Model Selection and Subsequent Selection Bias in Performance Evaluation",
                    "abstract": "Model selection strategies for machine learning algorithms typically involve the numerical optimisation of an appropriate model selection criterion, often based on an estimator of generalisation performance, such as k-fold cross-validation. The error of such an estimator can be broken down into bias and variance components. While unbiasedness is often cited as a beneficial quality of a model selection criterion, we demonstrate that a low variance is at least as important, as a non-negligible variance introduces the potential for over-fitting in model selection as well as in training the model. While this observation is in hindsight perhaps rather obvious, the degradation in performance due to over-fitting the model selection criterion can be surprisingly large, an observation that appears to have received little attention in the machine learning literature to date. In this paper, we show that the effects of this form of over-fitting are often of comparable magnitude to differences in performance between learning algorithms, and thus cannot be ignored in empirical evaluation. Furthermore, we show that some common performance evaluation practices are susceptible to a form of selection bias as a result of this form of over-fitting and hence are unreliable. We discuss methods to avoid over-fitting in model selection and subsequent selection bias in performance evaluation, which we hope will be incorporated into best practice. While this study concentrates on cross-validation based model selection, the findings are quite general and apply to any model selection practice involving the optimisation of a model selection criterion evaluated over a finite sample of data, including maximisation of the Bayesian evidence and optimisation of performance bounds."
                },
                {
                    "title": "Model Selection: Beyond the Bayesian/Frequentist Divide",
                    "abstract": "The principle of parsimony also known as \"Ockham's razor\" has inspired many theories of model selection. Yet such theories, all making arguments in favor of parsimony, are based on very different premises and have developed distinct methodologies to derive algorithms. We have organized challenges and edited a special issue of JMLR and several conference proceedings around the theme of model selection. In this editorial, we revisit the problem of avoiding overfitting in light of the latest results. We note the remarkable convergence of theories as different as Bayesian theory, Minimum Description Length, bias/variance tradeoff, Structural Risk Minimization, and regularization, in some approaches. We also present new and interesting examples of the complementarity of theories leading to hybrid algorithms, neither frequentist, nor Bayesian, or perhaps both frequentist and Bayesian!"
                },
                {
                    "title": "Particle Swarm Model Selection",
                    "abstract": "This paper proposes the application of particle swarm optimization (PSO) to the problem of full model selection, FMS, for classification tasks. FMS is defined as follows: given a pool of preprocessing methods, feature selection and learning algorithms, to select the combination of these that obtains the lowest classification error for a given data set; the task also includes the selection of hyperparameters for the considered methods. This problem generates a vast search space to be explored, well suited for stochastic optimization techniques. FMS can be applied to any classification domain as it does not require domain knowledge. Different model types and a variety of algorithms can be considered under this formulation. Furthermore, competitive yet simple models can be obtained with FMS. We adopt PSO for the search because of its proven performance in different problems and because of its simplicity, since neither expensive computations nor complicated operations are needed. Interestingly, the way the search is guided allows PSO to avoid overfitting to some extend. Experimental results on benchmark data sets give evidence that the proposed approach is very effective, despite its simplicity. Furthermore, results obtained in the framework of a model selection challenge show the competitiveness of the models selected with PSO, compared to models selected with other techniques that focus on a single algorithm and that use domain knowledge."
                },
                {
                    "title": "A survey of cross-validation procedures for model selection",
                    "abstract": "Used to estimate the risk of an estimator or to perform model selection, cross-validation is a widespread strategy because of its simplicity and its apparent universality. Many results exist on the model selection performances of cross-validation procedures. This survey intends to relate these results to the most recent advances of model selection theory, with a particular emphasis on distinguishing empirical statements from rigorous theoretical results. As a conclusion, guidelines are provided for choosing the best cross-validation procedure according to the particular features of the problem in hand."
                },
                {
                    "title": "Statistical Comparisons of Classifiers over Multiple Data Sets",
                    "abstract": "While methods for comparing two learning algorithms on a single data set have been scrutinized for quite some time already, the issue of statistical tests for comparisons of more algorithms on multiple data sets, which is even more essential to typical machine learning studies, has been all but ignored. This article reviews the current practice and then theoretically and empirically examines several suitable tests. Based on that, we recommend a set of simple, yet safe and robust non-parametric tests for statistical comparisons of classifiers: the Wilcoxon signed ranks test for comparison of two classifiers and the Friedman test with the corresponding post-hoc tests for comparison of more classifiers over multiple data sets. Results of the latter can also be neatly presented with the newly introduced CD (critical difference) diagrams."
                },
                {
                    "title": "The Design and Analysis of Benchmark Experiments",
                    "abstract": "The assessment of the performance of learners by means of benchmark experiments is an established exercise. In practice, benchmark studies are a tool to compare the performance of several competing algorithms for a certain learning problem. Cross-validation or resampling techniques are commonly used to derive point estimates of the performances which are compared to identify algorithms with good properties. For several benchmarking problems, test procedures taking the variability of those point estimates into account have been suggested. Most of the recently proposed inference procedures are based on special variance estimators for the cross-validated performance. We introduce a theoretical framework for inference problems in benchmark experiments and show that standard statistical test procedures can be used to test for differences in the performances. The theory is based on well-defined distributions of performance measures which can be compared with established tests. To demonstrate the usefulness in practice, the theoretical results are applied to regression and classification benchmark studies based on artificial and real world data."
                },
                {
                    "title": "Prediction error estimation: a comparison of resampling methods",
                    "abstract": "MOTIVATION\nIn genomic studies, thousands of features are collected on relatively few samples. One of the goals of these studies is to build classifiers to predict the outcome of future observations. There are three inherent steps to this process: feature selection, model selection and prediction assessment. With a focus on prediction assessment, we compare several methods for estimating the 'true' prediction error of a prediction model in the presence of feature selection.\n\n\nRESULTS\nFor small studies where features are selected from thousands of candidates, the resubstitution and simple split-sample estimates are seriously biased. In these small samples, leave-one-out cross-validation (LOOCV), 10-fold cross-validation (CV) and the .632+ bootstrap have the smallest bias for diagonal discriminant analysis, nearest neighbor and classification trees. LOOCV and 10-fold CV have the smallest bias for linear discriminant analysis. Additionally, LOOCV, 5- and 10-fold CV, and the .632+ bootstrap have the lowest mean square error. The .632+ bootstrap is quite biased in small sample sizes with strong signal-to-noise ratios. Differences in performance among resampling methods are reduced as the number of specimens available increase.\n\n\nSUPPLEMENTARY INFORMATION\nA complete compilation of results and R code for simulations and analyses are available in Molinaro et al. (2005) (http://linus.nci.nih.gov/brb/TechReport.htm)."
                },
                {
                    "title": "Completely Derandomized Self-Adaptation in Evolution Strategies",
                    "abstract": "This paper puts forward two useful methods for self-adaptation of the mutation distribution - the concepts of derandomization and cumulation. Principle shortcomings of the concept of mutative strategy parameter control and two levels of derandomization are reviewed. Basic demands on the self-adaptation of arbitrary (normal) mutation distributions are developed. Applying arbitrary, normal mutation distributions is equiv-alent to applying a general, linear problem encoding. The underlying objective of mutative strategy parameter control is roughly to favor previously selected mutation steps in the future. If this objective is pursued rigor-ously, a completely derandomized self-adaptation scheme results, which adapts arbitrary normal mutation distributions. This scheme, called covariance matrix adaptation (CMA), meets the previously stated demands. It can still be considerably improved by cumulation - utilizing an evolution path rather than single search steps. Simulations on various test functions reveal local and global search properties of the evolution strategy with and without covariance matrix adaptation. Their performances are comparable only on perfectly scaled functions. On badly scaled, non-separable functions usually a speed up factor of several orders of magnitude is ob-served. On moderately mis-scaled functions a speed up factor of three to ten can be expected."
                },
                {
                    "title": "Beating the hold-out: bounds for K-fold and progressive cross-validation",
                    "abstract": "The empirical error on a test set, the hold-out estimate, often is a more reliable estimate of generalization error than the observed error on the training set, the training estimate. K-fold cross validation is used in practice with the hope of being more accurate than the hold-out estimate without reducing the number of training examples. We argue that the k-fold estimate does in fact achieve this goal. Specifically, we show that for any nontrivial learning problem and learning algorithm that is insensitive to example ordering, the k-fold estimate is strictly more accurate than a single hold-out estimate on 1/k of the data, for 2 < k < n (k = n is leave-one-out), based on its variance and all higher moments. Previous bounds were termed sanitycheck because they compared the k-fold estimate to the training estimate and, further, restricted the VC dimension and required a notion of hypothesis stability [2]. In order to avoid these dependencies, we consider a k-fold hypothesis that is a randomized combination or average of the k individual hypotheses. We introduce progressive validationas another possible improvement on the hold-out estimate. This estimate of the generalization error is, in many ways, as good as that of a single hold-out, but it uses an average of half as many examples for testing. The procedure also involves a hold-out set, but after an example has been tested, it is added to the training set and the learning algorithm is rerun."
                },
                {
                    "title": "Approximate Statistical Tests for Comparing Supervised Classification Learning Algorithms",
                    "abstract": "This article reviews five approximate statistical tests for determining whether one learning algorithm outperforms another on a particular learning task. These test sare compared experimentally to determine their probability of incorrectly detecting a difference when no difference exists (type I error). Two widely used statistical tests are shown to have high probability of type I error in certain situations and should never be used: a test for the difference of two proportions and a paired-differences t test based on taking several random train-test splits. A third test, a paired-differences t test based on 10-fold cross-validation, exhibits somewhat elevated probability of type I error. A fourth test, McNemar's test, is shown to have low type I error. The fifth test is a new test, 5 2 cv, based on five iterations of twofold cross-validation. Experiments show that this test also has acceptable type I error. The article also measures the power (ability to detect algorithm differences when they do exist) of these tests. The cross-validated t test is the most powerful. The 52 cv test is shown to be slightly more powerful than McNemar's test. The choice of the best test is determined by the computational cost of running the learning algorithm. For algorithms that can be executed only once, Mc-Nemar's test is the only test with acceptable type I error. For algorithms that can be executed 10 times, the 5 2 cv test is recommended, because it is slightly more powerful and because it directly measures variation due to the choice of training set."
                },
                {
                    "title": "Preventing \"Overfitting\" of Cross-Validation Data",
                    "abstract": "Suppose that, for a learning task, we have to select one hypothesis out of a set of hypotheses (that may, for example, have been generated by multiple applications of a randomized learning algorithm). A common approach is to evaluate each hypothesis in the set on some previously unseen cross-validation data, and then to select the hypothesis that had the lowest cross-validation error. But when the cross-validation data is partially corrupted such as by noise, and if the set of hypotheses we are selecting from is large, then \\folklore\" also warns about \\overrtting\" the cross-In this paper, we explain how this \\overrtting\" really occurs, and show the surprising result that it can be overcome by selecting a hypothesis with a higher cross-validation error, over others with lower cross-validation errors. We give reasons for not selecting the hypothesis with the lowest cross-validation error, and propose a new algorithm, LOOCVCV, that uses a computa-tionally eecient form of leave{one{out cross-validation to select such a hypothesis. Finally , we present experimental results for one domain, that show LOOCVCV consistently beating picking the hypothesis with the lowest cross-validation error, even when using reasonably large cross-validation sets."
                },
                {
                    "title": "Oversearching and Layered Search in Empirical Learning",
                    "abstract": "When learning classifiers, more extensive search for rules is shown to lead to lower predictive accuracy on many of the leal-world domains investigated. This counter-intuitive re suit is particularly relevant to recent system the search methods that use risk-free pruning to achieve the same outcome as exhaustive search. We propose an iterated search method that commences with greedy search extending its scope at each Iteration until a stopping criterion is satisfied. This layered search is often found to produce theories that are more accurate than those obtained with either greedy search or moderately, extensive beam search."
                },
                {
                    "title": "Why do tree-based models still outperform deep learning on typical tabular data?",
                    "abstract": "While deep learning has enabled tremendous progress on text and image datasets, its superiority on tabular data is not clear. We contribute extensive benchmarks of standard and novel deep learning methods as well as tree-based models such as XGBoost and Random Forests, across a large number of datasets and hyperparameter combinations. We define a standard set of 45 datasets from varied domains with clear characteristics of tabular data and a benchmarking methodology accounting for both fitting models and finding good hyperparameters. Results show that tree-based models remain state-of-the-art on medium-sized data ( \u223c 10K samples) even without accounting for their superior speed. To understand this gap, we conduct an empirical investigation into the differing inductive biases of tree-based models and Neural Networks (NNs). This leads to a series of challenges which should guide researchers aiming to build tabular-specific NNs: 1. be robust to uninformative features, 2. preserve the orientation of the data, and 3. be able to easily learn irregular functions. To stimulate research on tabular architectures, we contribute a standard benchmark and raw data for baselines: every point of a 20 000 compute hours hyperparameter search for each learner. Results Looking at the results as a function of random search time rather than random search iterations tree-based models superiority even more striking. Neural networks and tree-based models were close for some benchmarks after a small number of iterations, but for the same amount of time spent on random search, tree-based models scores are always high above neural networks."
                },
                {
                    "title": "Syne Tune: A Library for Large Scale Hyperparameter Tuning and Reproducible Research",
                    "abstract": "We present Syne Tune, a library for large-scale distributed hyperparameter optimization (HPO). 1 Syne Tune\u2019s modular architecture allows users to easily switch between different execution backends to facilitate experimentation and makes it easy to contribute new optimization algorithms. To foster reproducible benchmarking, Syne Tune provides an efficient simulator backend and a benchmarking suite, which are essential for large-scale evaluations of distributed asynchronous HPO algorithms on tabulated and surrogate benchmarks. We showcase these functionalities with a range of state-of-the-art gradient-free optimizers, including multi-fidelity and transfer learning approaches on popular benchmarks from the literature. Additionally, we demonstrate the benefits of Syne Tune for constrained and multiobjective HPO applications through two use cases: the former considers hyperparameters that induce fair solutions and the latter automatically selects machine types along with the conventional hyperparameters."
                },
                {
                    "title": "Bayesian hyperparameter optimization : overfitting, ensembles and conditional spaces",
                    "abstract": "In this thesis, we consider the analysis and extension of Bayesian hyperparameter optimization methodology to various problems related to supervised machine learning. The contributions of the thesis are attached to 1) the overestimation of the generalization accuracy of hyperparameters and models resulting from Bayesian optimization, 2) an application of Bayesian optimization to ensemble learning, and 3) the optimization of spaces with a conditional structure such as found in automatic machine learning (AutoML) problems. Generally, machine learning algorithms have some free parameters, called hyperparameters, allowing to regulate or modify these algorithms\u2019 behaviour. For the longest time, hyperparameters were tuned by hand or with exhaustive search algorithms. Recent work highlighted the conceptual advantages in optimizing hyperparameters with more rational methods, such as Bayesian optimization. Bayesian optimization is a very versatile framework for the optimization of unknown and non-derivable functions, grounded strongly in probabilistic modelling and uncertainty estimation, and we adopt it for the work in this thesis. We first briefly introduce Bayesian optimization with Gaussian processes (GP) and describe its application to hyperparameter optimization. Next, original contributions are presented on the dangers of overfitting during hyperparameter optimization, where the optimization ends up learning the validation folds. We show that there is indeed overfitting during the optimization of hyperparameters, even with cross-validation strategies, and that it can be reduced by methods such as a reshuffling of the training and validation splits at every iteration of the optimization. Another promising method is demonstrated in the use of a GP\u2019s posterior mean for the selection of final hyperparameters, rather than directly returning the model with the minimal crossvalidation error. Both suggested approaches are demonstrated to deliver significant improvements in the generalization accuracy of the final selected model on a benchmark"
                },
                {
                    "title": "Empirical Evaluation of Resampling Procedures for Optimising SVM Hyperparameters",
                    "abstract": "Tuning the regularisation and kernel hyperparameters is a vital step in optimising the generalisation performance of kernel methods, such as the support vector machine (SVM). This is most often performed by minimising a resampling/cross-validation based model selection criterion, however there seems little practical guidance on the most suitable form of resampling. This paper presents the results of an extensive empirical evaluation of resampling procedures for SVM hyperparameter selection, designed to address this gap in the machine learning literature. We tested 15 different resampling procedures on 121 binary classification data sets in order to select the best SVM hyperparameters. We used three very different statistical procedures to analyse the results: the standard multi-classifier/multidata set procedure proposed by Demsar, the confidence intervals on the excess loss of each procedure in relation to 5-fold cross validation, and the Bayes factor analysis proposed by Barber. We conclude that a 2-fold procedure is appropriate to select the hyperparameters of an SVM for data sets for 1000 or more datapoints, while a 3-fold procedure is appropriate for smaller data sets."
                },
                {
                    "title": "Optimizing Support Vector Machines for Stormwater Prediction",
                    "abstract": "In water resource management, efficient controllers of stormwater tanks prevent flooding of sewage systems, which reduces environmental pollution. With accurate predictions of stormwater tank fill levels based on past rainfall, such controlling systems are able to detect state changes as early as possible. Up to now, good results on this problem could only be achieved by applying special-purpose models especially designed for stormwater prediction. The question arises whether it is possible to replace such specialized models with state-of-the-art machine learning methods, such as Support Vector Machines (SVM) in combination with consequent parameter tuning using sequential parameter optimization, to achieve competitive performance. This study shows that even superior results can be obtained if the SVM hyperparameters and the considered preprocessing is tuned. Unfortunately, this tuning might also result in overfitting or oversearching \u2013 both effects would lead to declined model generalizability. We analyze our tuned models and present possibilities to circumvent the effects of overfitting and oversearching."
                },
                {
                    "title": "On the Dangers of Cross-Validation. An Experimental Evaluation",
                    "abstract": "Cross validation allows models to be tested using the full training set by means of repeated resampling; thus, maximizing the total number of points used for testing and potentially, helping to protect against overfitting. Improvements in computational power, recent reductions in the (computational) cost of classification algorithms, and the development of closed-form solutions (for performing cross validation in certain classes of learning algorithms) makes it possible to test thousand or millions of variants of learning models on the data. Thus, it is now possible to calculate cross validation performance on a much larger number of tuned models than would have been possible otherwise. However, we empirically show how under such large number of models the risk for overfitting increases and the performance estimated by cross validation is no longer an effective estimate of generalization; hence, this paper provides an empirical reminder of the dangers of cross validation. We use a closed-form solution that makes this evaluation possible for the cross validation problem of interest. In addition, through extensive experiments we expose and discuss the effects of the overuse/misuse of cross validation in various aspects, including model selection, feature selection, and data dimensionality. This is illustrated on synthetic, benchmark, and real-world data sets."
                },
                {
                    "title": "Towards Quantifying the Effect of Datasets for Benchmarking: A Look at Tabular Machine Learning",
                    "abstract": "Data in tabular form makes up a large part of real-world ML applications, and thus, there has been a strong interest in developing novel deep learning (DL) architectures for supervised learning on tabular data in recent years. As a result, there is a debate as to whether DL methods are superior to the ubiquitous ensembles of boosted decision trees. Typically, the advantage of one model class over the other is claimed based on an empirical evaluation, where different variations of both model classes are compared on a set of benchmark datasets that supposedly resemble relevant real-world tabular data. While the landscape of state-of-the-art models for tabular data changed, one factor has remained largely constant over the years: The datasets . Here, we examine 30 recent publications and 187 different datasets they use, in terms of age, study size and relevance. We found that the average study used less than 10 datasets and that half of the datasets are older than 20 years. Our insights raise questions about the conclusions drawn from previous studies and urge the research community to develop and publish additional recent, challenging and relevant datasets and ML tasks for supervised learning on tabular data"
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does reshuffling resampling splits during hyperparameter optimization (HPO) affect the generalization performance of machine learning models?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant gap in hyperparameter optimization practices, which can lead to more robust and generalizable machine learning models. By demonstrating that reshuffling can improve generalization performance, this research could shift the paradigm in HPO methodologies, encouraging future studies to explore more dynamic and flexible approaches. This advancement could lead to practical applications in various domains where model performance is critical, ultimately enhancing the reliability of machine learning systems in real-world scenarios.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexities of hyperparameter optimization, which is a noisy, black-box optimization problem without a closed-form mathematical description. Naive approaches that rely on fixed resampling splits may lead to overfitting or \"overtuning,\" as they can bias the optimization process towards configurations that perform well on specific splits rather than on the overall data distribution. Overcoming these technical obstacles requires a deep understanding of the interaction between resampling methods and model performance, as well as the development of robust evaluation metrics that can accurately reflect generalization capabilities.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely overlooked the impact of resampling strategies in HPO, often defaulting to fixed splits without considering their potential biases. This gap may stem from a lack of awareness or understanding of the implications of resampling methods on model performance. While some studies have touched on reshuffling, they have not systematically examined its effects across various contexts or provided a comprehensive framework for its implementation. Our approach differs by rigorously analyzing the benefits of reshuffling and providing empirical evidence to support its efficacy in improving generalization performance.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves systematically examining the effects of reshuffling resampling splits during hyperparameter optimization. We will utilize a diverse set of machine learning models and datasets to evaluate the impact of reshuffling on generalization performance. The key metrics for assessment will include generalization error and model robustness across different configurations. We expect to demonstrate that reshuffling leads to improved overall generalization performance, particularly in scenarios where the loss surface is complex, thereby providing a compelling case for adopting this"
            }
        },
        "author_data": {
            "15f14df2-899d-4d1c-bfac-090cac964c01": {
                "pk": "15f14df2-899d-4d1c-bfac-090cac964c01",
                "name": "Thomas Nagler",
                "collaborators": [
                    "Claudia Czado",
                    "Thibault Vatter",
                    "Christian Schellhase",
                    "Christian Bumann",
                    "Daniel Kr\u00fcger",
                    "Aleksey Min",
                    "Katrin Lotto",
                    "Mladjan Radic",
                    "Nicolai Palm",
                    "Julian Rodemann"
                ],
                "domain": [
                    "Statistical Learning",
                    "Copula Models",
                    "Nonparametric Estimation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "A generic approach to nonparametric function estimation with mixed data",
                        "abstract": "In practice, data often contain discrete variables. But most of the popular nonparametric estimation methods have been developed in a purely continuous framework. A common trick among practitioners is to make discrete variables continuous by adding a small amount of noise. We show that this approach is justified if the noise distribution belongs to a certain class. In this case, any estimator developed in a purely continuous framework extends naturally to the mixed data setting. Estimators defined that way will be called continuous convolution estimators. They are extremely easy to implement and their asymptotic properties transfer directly from the continuous to the mixed data setting."
                    },
                    {
                        "title": "Asymptotic analysis of the jittering kernel density estimator",
                        "abstract": "Jittering estimators are nonparametric function estimators for mixed data. They extend arbitrary estimators from the continuous setting by adding random noise to discrete variables. We give an in-depth analysis of the jittering kernel density estimator, which reveals several appealing properties. The estimator is strongly consistent, asymptotically normal, and unbiased for discrete variables. It converges at minimax-optimal rates, which are established as a by-product of our analysis. To understand the effect of adding noise, we further study its asymptotic efficiency and finite sample bias in the univariate discrete case. Simulations show that the estimator is competitive on finite samples. The analysis suggests that similar properties can be expected for other jittering estimators."
                    },
                    {
                        "title": "Simplified vine copula models: state of science and affairs",
                        "abstract": "Vine copula models have become highly popular practical tools for modeling multivariate dependencies. To maintain tractability, a commonly employed simplifying assumption is that conditional copulas remain unchanged by the conditioning variables. This assumption has sparked a somewhat polarizing debate within the copula community. The fact that much of this dispute occurs outside the public record has placed the field in an unfortunate position, impeding scientific progress. In this article, I will review what we know about the flexibility and limitations of simplified vine copula models, explore the broader implications, and offer my own, hopefully reconciling, perspective on the issue."
                    },
                    {
                        "title": "kdecopula: An R Package for the Kernel Estimation of Bivariate Copula Densities",
                        "abstract": "We describe the R package kdecopula (current version 0.9.0), which provides fast implementations of various kernel estimators for the copula density. Due to a variety of available plotting options it is particularly useful for the exploratory analysis of dependence structures. It can be further used for accurate nonparametric estimation of copula densities and resampling. The implementation features spline interpolation of the estimates to allow for fast evaluation of density estimates and integrals thereof. We utilize this for a fast renormalization scheme that ensures that estimates are bona fide copula densities and additionally improves the estimators' accuracy. The performance of the methods is illustrated by simulations."
                    },
                    {
                        "title": "R friendly multi-threading in C++",
                        "abstract": "Calling multi-threaded C++ code from R has its perils. Since the R interpreter is single-threaded, one must not check for user interruptions or print to the R console from multiple threads. One can, however, synchronize with R from the main thread. The R package RcppThread (current version 1.0.0) contains a header only C++ library for thread safe communication with R that exploits this fact. It includes C++ classes for threads, a thread pool, and parallel loops that routinely synchronize with R. This article explains the package's functionality and gives examples of its usage. The synchronization mechanism may also apply to other threading frameworks. Benchmarks suggest that, although synchronization causes overhead, the parallel abstractions of RcppThread are competitive with other popular libraries in typical scenarios encountered in statistical computing."
                    },
                    {
                        "title": "Statistical Foundations of Prior-Data Fitted Networks",
                        "abstract": "Prior-data fitted networks (PFNs) were recently proposed as a new paradigm for machine learning. Instead of training the network to an observed training set, a fixed model is pre-trained offline on small, simulated training sets from a variety of tasks. The pre-trained model is then used to infer class probabilities in-context on fresh training sets with arbitrary size and distribution. Empirically, PFNs achieve state-of-the-art performance on tasks with similar size to the ones used in pre-training. Surprisingly, their accuracy further improves when passed larger data sets during inference. This article establishes a theoretical foundation for PFNs and illuminates the statistical mechanisms governing their behavior. While PFNs are motivated by Bayesian ideas, a purely frequentistic interpretation of PFNs as pre-tuned, but untrained predictors explains their behavior. A predictor's variance vanishes if its sensitivity to individual training samples does and the bias vanishes only if it is appropriately localized around the test feature. The transformer architecture used in current PFN implementations ensures only the former. These findings shall prove useful for designing architectures with favorable empirical behavior."
                    },
                    {
                        "title": "Generalized Additive Models for Pair-Copula Constructions",
                        "abstract": "Pair-copula constructions are flexible dependence models that use bivariate copulas as building blocks. In this paper, we use generalized additive models to extend them by allowing covariates effects. Borrowing ideas from a traditionally univariate context, we let each pair-copula parameter depend directly on the covariates in a parametric, semiparametric or nonparametric way. We propose a sequential estimation method that we study by simulation, and apply it to investigate the time-varying dependence structure between the intraday returns on four major foreign exchange rates. An R package, a script reproducing the results in this article, and additional simulation results are provided as supplementary material."
                    },
                    {
                        "title": "Evading the curse of dimensionality in nonparametric density estimation with simplified vine copulas",
                        "abstract": "Practical applications of nonparametric density estimators in more than three dimensions suffer a great deal from the well-known curse of dimensionality: convergence slows down as dimension increases. We show that one can evade the curse of dimensionality by assuming a simplified vine copula model for the dependence between variables. We formulate a general nonparametric estimator for such a model and show under high-level assumptions that the speed of convergence is independent of dimension. We further discuss a particular implementation for which we validate the high-level assumptions and establish its asymptotic normality. Simulation experiments illustrate a large gain in finite sample performance when the simplifying assumption is at least approximately true. But even when it is severely violated, the vine copula based approach proves advantageous as soon as more than a few variables are involved. Lastly, we give an application of the estimator to a classification problem from astrophysics."
                    },
                    {
                        "title": "Solving estimating equations with copulas",
                        "abstract": "Thanks to their ability to capture complex dependence structures, copulas are frequently used to glue random variables into a joint model with arbitrary marginal distributions. More recently, they have been applied to solve statistical learning problems such as regression or classification. Framing such approaches as solutions of estimating equations, we generalize them in a unified framework. We can then obtain simultaneous, coherent inferences across multiple regression-like problems. We derive consistency, asymptotic normality, and validity of the bootstrap for corresponding estimators. The conditions allow for both continuous and discrete data as well as parametric, nonparametric, and semiparametric estimators of the copula and marginal distributions. The versatility of this methodology is illustrated by several theoretical examples, a simulation study, and an application to financial portfolio allocation."
                    },
                    {
                        "title": "Nonparametric estimation of simplified vine copula models: comparison of methods",
                        "abstract": "In the last decade, simplified vine copula models have been an active area of research. They build a high dimensional probability density from the product of marginals densities and bivariate copula densities. Besides parametric models, several approaches to nonparametric estimation of vine copulas have been proposed. In this article, we extend these approaches and compare them in an extensive simulation study and a real data application. We identify several factors driving the relative performance of the estimators. The most important one is the strength of dependence. No method was found to be uniformly better than all others. Overall, the kernel estimators performed best, but do worse than penalized B-spline estimators when there is weak dependence and no tail dependence."
                    },
                    {
                        "title": "Model selection in sparse high-dimensional vine copula models with application to portfolio risk",
                        "abstract": "Vine copulas allow to build flexible dependence models for an arbitrary number of variables using only bivariate building blocks. The number of parameters in a vine copula model increases quadratically with the dimension, which poses new challenges in high-dimensional applications. To alleviate the computational burden and risk of overfitting, we propose a modified Bayesian information criterion (BIC) tailored to sparse vine copula models. We show that the criterion can consistently distinguish between the true and alternative models under less stringent conditions than the classical BIC. The new criterion can be used to select the hyper-parameters of sparse model classes, such as truncated and thresholded vine copulas. We propose a computationally efficient implementation and illustrate the benefits of the new concepts with a case study where we model the dependence in a large stock stock portfolio."
                    },
                    {
                        "title": "Stationary vine copula models for multivariate time series",
                        "abstract": "Multivariate time series exhibit two types of dependence: across variables and across time points. Vine copulas are graphical models for the dependence and can conveniently capture both types of dependence in the same model. We derive the maximal class of graph structures that guarantee stationarity under a natural and verifiable condition called translation invariance. We propose computationally efficient methods for estimation, simulation, prediction, and uncertainty quantification and show their validity by asymptotic results and simulations. The theoretical results allow for misspecified models and, even when specialized to the iid case, go beyond what is available in the literature. Their proofs are based on new results for general semiparametric method-of-moment estimators, which shall be of independent interest. The new model class is illustrated by an application to forecasting returns of a portfolio of 20 stocks, where they show excellent forecast performance. The paper is accompanied by an open source software implementation."
                    },
                    {
                        "title": "Modeling Stochastic Data Using Copulas For Application in Validation of Autonomous Driving",
                        "abstract": "Verification and validation of fully automated vehicles is linked to an almost intractable challenge of reflecting the real world with all its interactions in a virtual environment. Influential stochastic parameters need to be extracted from real-world measurements and real-time data, capturing all interdependencies, for an accurate simulation of reality. A copula is a probability model that represents a multivariate distribution, examining the dependence between the underlying variables. This model is used on drone measurement data from a roundabout containing dependent stochastic parameters. With the help of the copula model, samples are generated that reflect the real-time data. Resulting applications and possible extensions are discussed and explored."
                    },
                    {
                        "title": "An Online Bootstrap for Time Series",
                        "abstract": "Resampling methods such as the bootstrap have proven invaluable in the field of machine learning. However, the applicability of traditional bootstrap methods is limited when dealing with large streams of dependent data, such as time series or spatially correlated observations. In this paper, we propose a novel bootstrap method that is designed to account for data dependencies and can be executed online, making it particularly suitable for real-time applications. This method is based on an autoregressive sequence of increasingly dependent resampling weights. We prove the theoretical validity of the proposed bootstrap scheme under general conditions. We demonstrate the effectiveness of our approach through extensive simulations and show that it provides reliable uncertainty quantification even in the presence of complex data dependencies. Our work bridges the gap between classical resampling techniques and the demands of modern data analysis, providing a valuable tool for researchers and practitioners in dynamic, data-rich environments."
                    },
                    {
                        "title": "Approximately Bayes-Optimal Pseudo Label Selection",
                        "abstract": "Semi-supervised learning by self-training heavily relies on pseudo-label selection (PLS). The selection often depends on the initial model fit on labeled data. Early overfitting might thus be propagated to the final model by selecting instances with overconfident but erroneous predictions, often referred to as confirmation bias. This paper introduces BPLS, a Bayesian framework for PLS that aims to mitigate this issue. At its core lies a criterion for selecting instances to label: an analytical approximation of the posterior predictive of pseudo-samples. We derive this selection criterion by proving Bayes optimality of the posterior predictive of pseudo-samples. We further overcome computational hurdles by approximating the criterion analytically. Its relation to the marginal likelihood allows us to come up with an approximation based on Laplace's method and the Gaussian integral. We empirically assess BPLS for parametric generalized linear and non-parametric generalized additive models on simulated and real-world data. When faced with high-dimensional data prone to overfitting, BPLS outperforms traditional PLS methods."
                    },
                    {
                        "title": "Statistical Dependence Analyses of Operational Flight Data Used for Landing Reconstruction Enhancement",
                        "abstract": "The RTS smoother is widely used for state estimation and it is utilized here to increase the data quality with respect to physical coherence and to increase resolution. The purpose of this paper is to enhance the performance of the RTS smoother to reconstruct an aircraft landing using on board recorded data only. Thereby, errors and uncertainties of operational flight data (e.g. altitude, attitude, position, speed) recorded during flights of civil aircraft are minimized. These data can be used for subsequent analyses in terms of flight safety or efficiency, which is commonly referred to as Flight Data Monitoring (FDM).   Statistical assumptions of the smoother theory are not always verified during application but (consciously or not) assumed to be fulfilled. These assumptions can hardly be verified prior to the smoother application, however, they can be verified using the results of an initial smoother iteration and modifications of specific smoother characteristics can be suggested. This project specifically verifies assumptions on the measurement noise characteristics.   Variance and covariance of the measurement noise can be checked after the initial smoother application. It is discovered that these characteristics change over time and should be accounted for with a time varying covariance matrix. This sequence of matrices is estimated by kernel smoothing and replaces an initially assumed fixed and diagonal covariance matrix used for the first smoother run. The results of this second smoother iteration are mostly improved compared to the initial iteration, i.e. the errors are significantly reduced. Subsequently, the remaining dependence structures of the residuals of the second smoother iteration can be captured by copula models. Their interpretation is useful for a revision of the physical model utilized by the RTS smoother."
                    },
                    {
                        "title": "Explaining predictive models using Shapley values and non-parametric vine copulas",
                        "abstract": "The original development of Shapley values for prediction explanation relied on the assumption that the features being described were independent. If the features in reality are dependent this may lead to incorrect explanations. Hence, there have recently been attempts of appropriately modelling/estimating the dependence between the features. Although the proposed methods clearly outperform the traditional approach assuming independence, they have their weaknesses. In this paper we propose two new approaches for modelling the dependence between the features.   Both approaches are based on vine copulas, which are flexible tools for modelling multivariate non-Gaussian distributions able to characterise a wide range of complex dependencies.   The performance of the proposed methods is evaluated on simulated data sets and a real data set. The experiments demonstrate that the vine copula approaches give more accurate approximations to the true Shapley values than its competitors."
                    },
                    {
                        "title": "D-vine quantile regression with discrete variables",
                        "abstract": "Quantile regression, the prediction of conditional quantiles, finds applications in various fields. Often, some or all of the variables are discrete. The authors propose two new quantile regression approaches to handle such mixed discrete-continuous data. Both of them generalize the continuous D-vine quantile regression, where the dependence between the response and the covariates is modeled by a parametric D-vine. D-vine quantile regression provides very flexible models, that enable accurate and fast predictions. Moreover, it automatically takes care of major issues of classical quantile regression, such as quantile crossing and interactions between the covariates. The first approach keeps the parametric estimation of the D-vines, but modifies the formulas to account for the discreteness. The second approach estimates the D-vine using continuous convolution to make the discrete variables continuous and then estimates the D-vine nonparametrically. A simulation study is presented examining for which scenarios the discrete-continuous D-vine quantile regression can provide superior prediction abilities. Lastly, the functionality of the two introduced methods is demonstrated by a real-world example predicting the number of bike rentals."
                    },
                    {
                        "title": "Copula-based synthetic data augmentation for machine-learning emulators",
                        "abstract": "Can we improve machine-learning (ML) emulators with synthetic data? If data are scarce or expensive to source and a physical model is available, statistically generated data may be useful for augmenting training sets cheaply. Here we explore the use of copula-based models for generating synthetically augmented datasets in weather and climate by testing the method on a toy physical model of downwelling longwave radiation and corresponding neural network emulator. Results show that for copula-augmented datasets, predictions are improved by up to 62 % for the mean absolute error (from 1.17 to 0.44 W m$^{-2}$)."
                    }
                ]
            },
            "f5e44304-2e9c-41e1-a087-a48e1941e334": {
                "pk": "f5e44304-2e9c-41e1-a087-a48e1941e334",
                "name": "Lennart Schneider",
                "collaborators": [
                    "Bernd Bischl",
                    "Florian Pfisterer",
                    "Janek Thomas",
                    "Martin Binder",
                    "Lennart Sch\u00e4permeier",
                    "Raphael Patrick Prager",
                    "Heike Trautmann",
                    "Pascal Kerschke",
                    "Lennart Purucker",
                    "Marie Anastacio"
                ],
                "domain": [
                    "Hyperparameter Optimization",
                    "Automated Machine Learning",
                    "Quality Diversity Optimization",
                    "Neural Architecture Search"
                ],
                "publications": [
                    {
                        "title": "HPO X ELA: Investigating Hyperparameter Optimization Landscapes by Means of Exploratory Landscape Analysis",
                        "abstract": "Hyperparameter optimization (HPO) is a key component of machine learning models for achieving peak predictive performance. While numerous methods and algorithms for HPO have been proposed over the last years, little progress has been made in illuminating and examining the actual structure of these black-box optimization problems. Exploratory landscape analysis (ELA) subsumes a set of techniques that can be used to gain knowledge about properties of unknown optimization problems. In this paper, we evaluate the performance of five different black-box optimizers on 30 HPO problems, which consist of two-, three- and five-dimensional continuous search spaces of the XGBoost learner trained on 10 different data sets. This is contrasted with the performance of the same optimizers evaluated on 360 problem instances from the black-box optimization benchmark (BBOB). We then compute ELA features on the HPO and BBOB problems and examine similarities and differences. A cluster analysis of the HPO and BBOB problems in ELA feature space allows us to identify how the HPO problems compare to the BBOB problems on a structural meta-level. We identify a subset of BBOB problems that are close to the HPO problems in ELA feature space and show that optimizer performance is comparably similar on these two sets of benchmark problems. We highlight open challenges of ELA for HPO and discuss potential directions of future research and applications."
                    },
                    {
                        "title": "Q(D)O-ES: Population-based Quality (Diversity) Optimisation for Post Hoc Ensemble Selection in AutoML",
                        "abstract": "Automated machine learning (AutoML) systems commonly ensemble models post hoc to improve predictive performance, typically via greedy ensemble selection (GES). However, we believe that GES may not always be optimal, as it performs a simple deterministic greedy search. In this work, we introduce two novel population-based ensemble selection methods, QO-ES and QDO-ES, and compare them to GES. While QO-ES optimises solely for predictive performance, QDO-ES also considers the diversity of ensembles within the population, maintaining a diverse set of well-performing ensembles during optimisation based on ideas of quality diversity optimisation. The methods are evaluated using 71 classification datasets from the AutoML benchmark, demonstrating that QO-ES and QDO-ES often outrank GES, albeit only statistically significant on validation data. Our results further suggest that diversity can be beneficial for post hoc ensembling but also increases the risk of overfitting."
                    },
                    {
                        "title": "A Collection of Quality Diversity Optimization Problems Derived from Hyperparameter Optimization of Machine Learning Models",
                        "abstract": "The goal of Quality Diversity Optimization is to generate a collection of diverse yet high-performing solutions to a given problem at hand. Typical benchmark problems are, for example, finding a repertoire of robot arm configurations or a collection of game playing strategies. In this paper, we propose a set of Quality Diversity Optimization problems that tackle hyperparameter optimization of machine learning models - a so far underexplored application of Quality Diversity Optimization. Our benchmark problems involve novel feature functions, such as interpretability or resource usage of models. To allow for fast and efficient benchmarking, we build upon YAHPO Gym, a recently proposed open source benchmarking suite for hyperparameter optimization that makes use of high performing surrogate models and returns these surrogate model predictions instead of evaluating the true expensive black box function. We present results of an initial experimental study comparing different Quality Diversity optimizers on our benchmark problems. Furthermore, we discuss future directions and challenges of Quality Diversity Optimization in the context of hyperparameter optimization."
                    },
                    {
                        "title": "Multi-Objective Optimization of Performance and Interpretability of Tabular Supervised Machine Learning Models",
                        "abstract": "We present a model-agnostic framework for jointly optimizing the predictive performance and interpretability of supervised machine learning models for tabular data. Interpretability is quantified via three measures: feature sparsity, interaction sparsity of features, and sparsity of non-monotone feature effects. By treating hyperparameter optimization of a machine learning algorithm as a multi-objective optimization problem, our framework allows for generating diverse models that trade off high performance and ease of interpretability in a single optimization run. Efficient optimization is achieved via augmentation of the search space of the learning algorithm by incorporating feature selection, interaction and monotonicity constraints into the hyperparameter search space. We demonstrate that the optimization problem effectively translates to finding the Pareto optimal set of groups of selected features that are allowed to interact in a model, along with finding their optimal monotonicity constraints and optimal hyperparameters of the learning algorithm itself. We then introduce a novel evolutionary algorithm that can operate efficiently on this augmented search space. In benchmark experiments, we show that our framework is capable of finding diverse models that are highly competitive or outperform state-of-the-art XGBoost or Explainable Boosting Machine models, both with respect to performance and interpretability."
                    },
                    {
                        "title": "Mutation is all you need",
                        "abstract": "Neural architecture search (NAS) promises to make deep learning accessible to non-experts by automating architecture engineering of deep neural networks. BANANAS is one state-of-the-art NAS method that is embedded within the Bayesian optimization framework. Recent experimental findings have demonstrated the strong performance of BANANAS on the NAS-Bench-101 benchmark being determined by its path encoding and not its choice of surrogate model. We present experimental results suggesting that the performance of BANANAS on the NAS-Bench-301 benchmark is determined by its acquisition function optimizer, which minimally mutates the incumbent."
                    },
                    {
                        "title": "YAHPO Gym -- An Efficient Multi-Objective Multi-Fidelity Benchmark for Hyperparameter Optimization",
                        "abstract": "When developing and analyzing new hyperparameter optimization methods, it is vital to empirically evaluate and compare them on well-curated benchmark suites. In this work, we propose a new set of challenging and relevant benchmark problems motivated by desirable properties and requirements for such benchmarks. Our new surrogate-based benchmark collection consists of 14 scenarios that in total constitute over 700 multi-fidelity hyperparameter optimization problems, which all enable multi-objective hyperparameter optimization. Furthermore, we empirically compare surrogate-based benchmarks to the more widely-used tabular benchmarks, and demonstrate that the latter may produce unfaithful results regarding the performance ranking of HPO methods. We examine and compare our benchmark collection with respect to defined requirements and propose a single-objective as well as a multi-objective benchmark suite on which we compare 7 single-objective and 7 multi-objective optimizers in a benchmark experiment. Our software is available at [https://github.com/slds-lmu/yahpo_gym]."
                    }
                ]
            },
            "04c92083-847b-4459-9c04-f8fb9e60d334": {
                "pk": "04c92083-847b-4459-9c04-f8fb9e60d334",
                "name": "Bernd Bischl",
                "collaborators": [
                    "Florian Pfisterer",
                    "David R\u00fcgamer",
                    "Janek Thomas",
                    "Xudong Sun",
                    "Giuseppe Casalicchio",
                    "Christoph Molnar",
                    "Julia Moosbauer",
                    "Tobias Weber",
                    "Michael Ingrisch",
                    "Laura Beggel"
                ],
                "domain": [
                    "Machine Learning",
                    "Variational Inference",
                    "AutoML",
                    "Interpretable Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Tutorial and Survey on Probabilistic Graphical Model and Variational Inference in Deep Reinforcement Learning",
                        "abstract": "Aiming at a comprehensive and concise tutorial survey, recap of variational inference and reinforcement learning with Probabilistic Graphical Models are given with detailed derivations. Reviews and comparisons on recent advances in deep reinforcement learning are made from various aspects. We offer detailed derivations to a taxonomy of Probabilistic Graphical Model and Variational Inference methods in deep reinforcement learning, which serves as a complementary material on top of the original contributions."
                    },
                    {
                        "title": "Robust Anomaly Detection in Images using Adversarial Autoencoders",
                        "abstract": "Reliably detecting anomalies in a given set of images is a task of high practical relevance for visual quality inspection, surveillance, or medical image analysis. Autoencoder neural networks learn to reconstruct normal images, and hence can classify those images as anomalies, where the reconstruction error exceeds some threshold. Here we analyze a fundamental problem of this approach when the training set is contaminated with a small fraction of outliers. We find that continued training of autoencoders inevitably reduces the reconstruction error of outliers, and hence degrades the anomaly detection performance. In order to counteract this effect, an adversarial autoencoder architecture is adapted, which imposes a prior distribution on the latent representation, typically placing anomalies into low likelihood-regions. Utilizing the likelihood model, potential anomalies can be identified and rejected already during training, which results in an anomaly detector that is significantly more robust to the presence of outliers during training."
                    },
                    {
                        "title": "Quantifying Model Complexity via Functional Decomposition for Better Post-Hoc Interpretability",
                        "abstract": "Post-hoc model-agnostic interpretation methods such as partial dependence plots can be employed to interpret complex machine learning models. While these interpretation methods can be applied regardless of model complexity, they can produce misleading and verbose results if the model is too complex, especially w.r.t. feature interactions. To quantify the complexity of arbitrary machine learning models, we propose model-agnostic complexity measures based on functional decomposition: number of features used, interaction strength and main effect complexity. We show that post-hoc interpretation of models that minimize the three measures is more reliable and compact. Furthermore, we demonstrate the application of these measures in a multi-objective optimization approach which simultaneously minimizes loss and complexity."
                    },
                    {
                        "title": "ReinBo: Machine Learning pipeline search and configuration with Bayesian Optimization embedded Reinforcement Learning",
                        "abstract": "Machine learning pipeline potentially consists of several stages of operations like data preprocessing, feature engineering and machine learning model training. Each operation has a set of hyper-parameters, which can become irrelevant for the pipeline when the operation is not selected. This gives rise to a hierarchical conditional hyper-parameter space. To optimize this mixed continuous and discrete conditional hierarchical hyper-parameter space, we propose an efficient pipeline search and configuration algorithm which combines the power of Reinforcement Learning and Bayesian Optimization. Empirical results show that our method performs favorably compared to state of the art methods like Auto-sklearn , TPOT, Tree Parzen Window, and Random Search."
                    },
                    {
                        "title": "Variational Resampling Based Assessment of Deep Neural Networks under Distribution Shift",
                        "abstract": "A novel variational inference based resampling framework is proposed to evaluate the robustness and generalization capability of deep learning models with respect to distribution shift. We use Auto Encoding Variational Bayes to find a latent representation of the data, on which a Variational Gaussian Mixture Model is applied to deliberately create distribution shift by dividing the dataset into different clusters. Wasserstein distance is used to characterize the extent of distribution shift between the generated data splits. We compare several popular Convolutional Neural Network (CNN) architectures and Bayesian CNN models for image classification on the Fashion-MNIST dataset, to assess their robustness and generalization behavior under the deliberately created distribution shift, as well as under random Cross Validation. Our method of creating artificial domain splits of a single dataset can also be used to establish novel model selection criteria and assessment tools in machine learning, as well as benchmark methods for domain adaptation and domain generalization approaches."
                    },
                    {
                        "title": "Multi-Objective Automatic Machine Learning with AutoxgboostMC",
                        "abstract": "AutoML systems are currently rising in popularity, as they can build powerful models without human oversight. They often combine techniques from many different sub-fields of machine learning in order to find a model or set of models that optimize a user-supplied criterion, such as predictive performance. The ultimate goal of such systems is to reduce the amount of time spent on menial tasks, or tasks that can be solved better by algorithms while leaving decisions that require human intelligence to the end-user. In recent years, the importance of other criteria, such as fairness and interpretability, and many others have become more and more apparent. Current AutoML frameworks either do not allow to optimize such secondary criteria or only do so by limiting the system's choice of models and preprocessing steps. We propose to optimize additional criteria defined by the user directly to guide the search towards an optimal machine learning pipeline. In order to demonstrate the need and usefulness of our approach, we provide a simple multi-criteria AutoML system and showcase an exemplary application."
                    },
                    {
                        "title": "Tunability: Importance of Hyperparameters of Machine Learning Algorithms",
                        "abstract": "Modern supervised machine learning algorithms involve hyperparameters that have to be set before running them. Options for setting hyperparameters are default values from the software package, manual configuration by the user or configuring them for optimal predictive performance by a tuning procedure. The goal of this paper is two-fold. Firstly, we formalize the problem of tuning from a statistical point of view, define data-based defaults and suggest general measures quantifying the tunability of hyperparameters of algorithms. Secondly, we conduct a large-scale benchmarking study based on 38 datasets from the OpenML platform and six common machine learning algorithms. We apply our measures to assess the tunability of their parameters. Our results yield default values for hyperparameters and enable users to decide whether it is worth conducting a possibly time consuming tuning strategy, to focus on the most important hyperparameters and to chose adequate hyperparameter spaces for tuning."
                    },
                    {
                        "title": "Deep Bregman Divergence for Contrastive Learning of Visual Representations",
                        "abstract": "Deep Bregman divergence measures divergence of data points using neural networks which is beyond Euclidean distance and capable of capturing divergence over distributions. In this paper, we propose deep Bregman divergences for contrastive learning of visual representation where we aim to enhance contrastive loss used in self-supervised learning by training additional networks based on functional Bregman divergence. In contrast to the conventional contrastive learning methods which are solely based on divergences between single points, our framework can capture the divergence between distributions which improves the quality of learned representation. We show the combination of conventional contrastive loss and our proposed divergence loss outperforms baseline and most of the previous methods for self-supervised and semi-supervised learning on multiple classifications and object detection tasks and datasets. Moreover, the learned representations generalize well when transferred to the other datasets and tasks. The source code and our models are available in supplementary and will be released with paper."
                    },
                    {
                        "title": "Towards Human Centered AutoML",
                        "abstract": "Building models from data is an integral part of the majority of data science workflows. While data scientists are often forced to spend the majority of the time available for a given project on data cleaning and exploratory analysis, the time available to practitioners to build actual models from data is often rather short due to time constraints for a given project. AutoML systems are currently rising in popularity, as they can build powerful models without human oversight. In this position paper, we aim to discuss the impact of the rising popularity of such systems and how a user-centered interface for such systems could look like. More importantly, we also want to point out features that are currently missing in those systems and start to explore better usability of such systems from a data-scientists perspective."
                    },
                    {
                        "title": "Multi-Objective Hyperparameter Tuning and Feature Selection using Filter Ensembles",
                        "abstract": "Both feature selection and hyperparameter tuning are key tasks in machine learning. Hyperparameter tuning is often useful to increase model performance, while feature selection is undertaken to attain sparse models. Sparsity may yield better model interpretability and lower cost of data acquisition, data handling and model inference. While sparsity may have a beneficial or detrimental effect on predictive performance, a small drop in performance may be acceptable in return for a substantial gain in sparseness. We therefore treat feature selection as a multi-objective optimization task. We perform hyperparameter tuning and feature selection simultaneously because the choice of features of a model may influence what hyperparameters perform well.   We present, benchmark, and compare two different approaches for multi-objective joint hyperparameter optimization and feature selection: The first uses multi-objective model-based optimization. The second is an evolutionary NSGA-II-based wrapper approach to feature selection which incorporates specialized sampling, mutation and recombination operators. Both methods make use of parameterized filter ensembles.   While model-based optimization needs fewer objective evaluations to achieve good performance, it incurs computational overhead compared to the NSGA-II, so the preferred choice depends on the cost of evaluating a model on given data."
                    },
                    {
                        "title": "Relative Feature Importance",
                        "abstract": "Interpretable Machine Learning (IML) methods are used to gain insight into the relevance of a feature of interest for the performance of a model. Commonly used IML methods differ in whether they consider features of interest in isolation, e.g., Permutation Feature Importance (PFI), or in relation to all remaining feature variables, e.g., Conditional Feature Importance (CFI). As such, the perturbation mechanisms inherent to PFI and CFI represent extreme reference points. We introduce Relative Feature Importance (RFI), a generalization of PFI and CFI that allows for a more nuanced feature importance computation beyond the PFI versus CFI dichotomy. With RFI, the importance of a feature relative to any other subset of features can be assessed, including variables that were not available at training time. We derive general interpretation rules for RFI based on a detailed theoretical analysis of the implications of relative feature relevance, and demonstrate the method's usefulness on simulated examples."
                    },
                    {
                        "title": "Neural Mixture Distributional Regression",
                        "abstract": "We present neural mixture distributional regression (NMDR), a holistic framework to estimate complex finite mixtures of distributional regressions defined by flexible additive predictors. Our framework is able to handle a large number of mixtures of potentially different distributions in high-dimensional settings, allows for efficient and scalable optimization and can be applied to recent concepts that combine structured regression models with deep neural networks. While many existing approaches for mixture models address challenges in optimization of such and provide results for convergence under specific model assumptions, our approach is assumption-free and instead makes use of optimizers well-established in deep learning. Through extensive numerical experiments and a high-dimensional deep learning application we provide evidence that the proposed approach is competitive to existing approaches and works well in more complex scenarios."
                    },
                    {
                        "title": "Accelerated Componentwise Gradient Boosting using Efficient Data Representation and Momentum-based Optimization",
                        "abstract": "Componentwise boosting (CWB), also known as model-based boosting, is a variant of gradient boosting that builds on additive models as base learners to ensure interpretability. CWB is thus often used in research areas where models are employed as tools to explain relationships in data. One downside of CWB is its computational complexity in terms of memory and runtime. In this paper, we propose two techniques to overcome these issues without losing the properties of CWB: feature discretization of numerical features and incorporating Nesterov momentum into functional gradient descent. As the latter can be prone to early overfitting, we also propose a hybrid approach that prevents a possibly diverging gradient descent routine while ensuring faster convergence. We perform extensive benchmarks on multiple simulated and real-world data sets to demonstrate the improvements in runtime and memory consumption while maintaining state-of-the-art estimation and prediction performance."
                    },
                    {
                        "title": "Towards modelling hazard factors in unstructured data spaces using gradient-based latent interpolation",
                        "abstract": "The application of deep learning in survival analysis (SA) allows utilizing unstructured and high-dimensional data types uncommon in traditional survival methods. This allows to advance methods in fields such as digital health, predictive maintenance, and churn analysis, but often yields less interpretable and intuitively understandable models due to the black-box character of deep learning-based approaches. We close this gap by proposing 1) a multi-task variational autoencoder (VAE) with survival objective, yielding survival-oriented embeddings, and 2) a novel method HazardWalk that allows to model hazard factors in the original data space. HazardWalk transforms the latent distribution of our autoencoder into areas of maximized/minimized hazard and then uses the decoder to project changes to the original domain. Our procedure is evaluated on a simulated dataset as well as on a dataset of CT imaging data of patients with liver metastases."
                    },
                    {
                        "title": "REPID: Regional Effect Plots with implicit Interaction Detection",
                        "abstract": "Machine learning models can automatically learn complex relationships, such as non-linear and interaction effects. Interpretable machine learning methods such as partial dependence plots visualize marginal feature effects but may lead to misleading interpretations when feature interactions are present. Hence, employing additional methods that can detect and measure the strength of interactions is paramount to better understand the inner workings of machine learning models. We demonstrate several drawbacks of existing global interaction detection approaches, characterize them theoretically, and evaluate them empirically. Furthermore, we introduce regional effect plots with implicit interaction detection, a novel framework to detect interactions between a feature of interest and other features. The framework also quantifies the strength of interactions and provides interpretable and distinct regions in which feature effects can be interpreted more reliably, as they are less confounded by interactions. We prove the theoretical eligibility of our method and show its applicability on various simulation and real-world examples."
                    },
                    {
                        "title": "A Collection of Quality Diversity Optimization Problems Derived from Hyperparameter Optimization of Machine Learning Models",
                        "abstract": "The goal of Quality Diversity Optimization is to generate a collection of diverse yet high-performing solutions to a given problem at hand. Typical benchmark problems are, for example, finding a repertoire of robot arm configurations or a collection of game playing strategies. In this paper, we propose a set of Quality Diversity Optimization problems that tackle hyperparameter optimization of machine learning models - a so far underexplored application of Quality Diversity Optimization. Our benchmark problems involve novel feature functions, such as interpretability or resource usage of models. To allow for fast and efficient benchmarking, we build upon YAHPO Gym, a recently proposed open source benchmarking suite for hyperparameter optimization that makes use of high performing surrogate models and returns these surrogate model predictions instead of evaluating the true expensive black box function. We present results of an initial experimental study comparing different Quality Diversity optimizers on our benchmark problems. Furthermore, we discuss future directions and challenges of Quality Diversity Optimization in the context of hyperparameter optimization."
                    },
                    {
                        "title": "Improving Accuracy of Interpretability Measures in Hyperparameter Optimization via Bayesian Algorithm Execution",
                        "abstract": "Despite all the benefits of automated hyperparameter optimization (HPO), most modern HPO algorithms are black-boxes themselves. This makes it difficult to understand the decision process which leads to the selected configuration, reduces trust in HPO, and thus hinders its broad adoption. Here, we study the combination of HPO with interpretable machine learning (IML) methods such as partial dependence plots. These techniques are more and more used to explain the marginal effect of hyperparameters on the black-box cost function or to quantify the importance of hyperparameters. However, if such methods are naively applied to the experimental data of the HPO process in a post-hoc manner, the underlying sampling bias of the optimizer can distort interpretations. We propose a modified HPO method which efficiently balances the search for the global optimum w.r.t. predictive performance \\emph{and} the reliable estimation of IML explanations of an underlying black-box function by coupling Bayesian optimization and Bayesian Algorithm Execution. On benchmark cases of both synthetic objectives and HPO of a neural network, we demonstrate that our method returns more reliable explanations of the underlying black-box without a loss of optimization performance."
                    },
                    {
                        "title": "Improved proteasomal cleavage prediction with positive-unlabeled learning",
                        "abstract": "Accurate in silico modeling of the antigen processing pathway is crucial to enable personalized epitope vaccine design for cancer. An important step of such pathway is the degradation of the vaccine into smaller peptides by the proteasome, some of which are going to be presented to T cells by the MHC complex. While predicting MHC-peptide presentation has received a lot of attention recently, proteasomal cleavage prediction remains a relatively unexplored area in light of recent advancesin high-throughput mass spectrometry-based MHC ligandomics. Moreover, as such experimental techniques do not allow to identify regions that cannot be cleaved, the latest predictors generate decoy negative samples and treat them as true negatives when training, even though some of them could actually be positives. In this work, we thus present a new predictor trained with an expanded dataset and the solid theoretical underpinning of positive-unlabeled learning, achieving a new state-of-the-art in proteasomal cleavage prediction. The improved predictive capabilities will in turn enable more precise vaccine development improving the efficacy of epitope-based vaccines. Pretrained models are available on GitHub"
                    },
                    {
                        "title": "Mixture of Experts Distributional Regression: Implementation Using Robust Estimation with Adaptive First-order Methods",
                        "abstract": "In this work, we propose an efficient implementation of mixtures of experts distributional regression models which exploits robust estimation by using stochastic first-order optimization techniques with adaptive learning rate schedulers. We take advantage of the flexibility and scalability of neural network software and implement the proposed framework in mixdistreg, an R software package that allows for the definition of mixtures of many different families, estimation in high-dimensional and large sample size settings and robust optimization based on TensorFlow. Numerical experiments with simulated and real-world data applications show that optimization is as reliable as estimation via classical approaches in many different settings and that results may be obtained for complicated scenarios where classical approaches consistently fail."
                    },
                    {
                        "title": "Cascaded Latent Diffusion Models for High-Resolution Chest X-ray Synthesis",
                        "abstract": "While recent advances in large-scale foundational models show promising results, their application to the medical domain has not yet been explored in detail. In this paper, we progress into the realms of large-scale modeling in medical synthesis by proposing Cheff - a foundational cascaded latent diffusion model, which generates highly-realistic chest radiographs providing state-of-the-art quality on a 1-megapixel scale. We further propose MaCheX, which is a unified interface for public chest datasets and forms the largest open collection of chest X-rays up to date. With Cheff conditioned on radiological reports, we further guide the synthesis process over text prompts and unveil the research area of report-to-chest-X-ray generation."
                    }
                ]
            },
            "51a8685f-f464-4e35-a798-c5f4ab8293c8": {
                "pk": "51a8685f-f464-4e35-a798-c5f4ab8293c8",
                "name": "Matthias Feurer",
                "collaborators": [
                    "Frank Hutter",
                    "Katharina Eggensperger",
                    "Marius Lindauer",
                    "Bernd Bischl",
                    "Andr\u00e9 Biedenkapp",
                    "Noor Awad",
                    "Neeratyoy Mallik",
                    "Joaquin Vanschoren",
                    "Edward Bergman",
                    "Florian Pfisterer"
                ],
                "domain": [
                    "Automated Machine Learning",
                    "Hyperparameter Optimization",
                    "Bayesian Optimization",
                    "Fairness in Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Practical Transfer Learning for Bayesian Optimization",
                        "abstract": "When hyperparameter optimization of a machine learning algorithm is repeated for multiple datasets it is possible to transfer knowledge to an optimization run on a new dataset. We develop a new hyperparameter-free ensemble model for Bayesian optimization that is a generalization of two existing transfer learning extensions to Bayesian optimization and establish a worst-case bound compared to vanilla Bayesian optimization. Using a large collection of hyperparameter optimization benchmark problems, we demonstrate that our contributions substantially reduce optimization time compared to standard Gaussian process-based Bayesian optimization and improve over the current state-of-the-art for transfer hyperparameter optimization."
                    },
                    {
                        "title": "PFNs4BO: In-Context Learning for Bayesian Optimization",
                        "abstract": "In this paper, we use Prior-data Fitted Networks (PFNs) as a flexible surrogate for Bayesian Optimization (BO). PFNs are neural processes that are trained to approximate the posterior predictive distribution (PPD) through in-context learning on any prior distribution that can be efficiently sampled from. We describe how this flexibility can be exploited for surrogate modeling in BO. We use PFNs to mimic a naive Gaussian process (GP), an advanced GP, and a Bayesian Neural Network (BNN). In addition, we show how to incorporate further information into the prior, such as allowing hints about the position of optima (user priors), ignoring irrelevant dimensions, and performing non-myopic BO by learning the acquisition function. The flexibility underlying these extensions opens up vast possibilities for using PFNs for BO. We demonstrate the usefulness of PFNs for BO in a large-scale evaluation on artificial GP samples and three different hyperparameter optimization testbeds: HPO-B, Bayesmark, and PD1. We publish code alongside trained models at github.com/automl/PFNs4BO."
                    },
                    {
                        "title": "Mind the Gap: Measuring Generalization Performance Across Multiple Objectives",
                        "abstract": "Modern machine learning models are often constructed taking into account multiple objectives, e.g., minimizing inference time while also maximizing accuracy. Multi-objective hyperparameter optimization (MHPO) algorithms return such candidate models, and the approximation of the Pareto front is used to assess their performance. In practice, we also want to measure generalization when moving from the validation to the test set. However, some of the models might no longer be Pareto-optimal which makes it unclear how to quantify the performance of the MHPO method when evaluated on the test set. To resolve this, we provide a novel evaluation protocol that allows measuring the generalization performance of MHPO methods and studying its capabilities for comparing two optimization experiments."
                    },
                    {
                        "title": "Auto-Sklearn 2.0: Hands-free AutoML via Meta-Learning",
                        "abstract": "Automated Machine Learning (AutoML) supports practitioners and researchers with the tedious task of designing machine learning pipelines and has recently achieved substantial success. In this paper, we introduce new AutoML approaches motivated by our winning submission to the second ChaLearn AutoML challenge. We develop PoSH Auto-sklearn, which enables AutoML systems to work well on large datasets under rigid time limits by using a new, simple and meta-feature-free meta-learning technique and by employing a successful bandit strategy for budget allocation. However, PoSH Auto-sklearn introduces even more ways of running AutoML and might make it harder for users to set it up correctly. Therefore, we also go one step further and study the design space of AutoML itself, proposing a solution towards truly hands-free AutoML. Together, these changes give rise to the next generation of our AutoML system, Auto-sklearn 2.0. We verify the improvements by these additions in an extensive experimental study on 39 AutoML benchmark datasets. We conclude the paper by comparing to other popular AutoML frameworks and Auto-sklearn 1.0, reducing the relative error by up to a factor of 4.5, and yielding a performance in 10 minutes that is substantially better than what Auto-sklearn 1.0 achieves within an hour."
                    },
                    {
                        "title": "Towards Assessing the Impact of Bayesian Optimization's Own Hyperparameters",
                        "abstract": "Bayesian Optimization (BO) is a common approach for hyperparameter optimization (HPO) in automated machine learning. Although it is well-accepted that HPO is crucial to obtain well-performing machine learning models, tuning BO's own hyperparameters is often neglected. In this paper, we empirically study the impact of optimizing BO's own hyperparameters and the transferability of the found settings using a wide range of benchmarks, including artificial functions, HPO and HPO combined with neural architecture search. In particular, we show (i) that tuning can improve the any-time performance of different BO approaches, that optimized BO settings also perform well (ii) on similar problems and (iii) partially even on problems from other problem families, and (iv) which BO hyperparameters are most important."
                    },
                    {
                        "title": "BOAH: A Tool Suite for Multi-Fidelity Bayesian Optimization & Analysis of Hyperparameters",
                        "abstract": "Hyperparameter optimization and neural architecture search can become prohibitively expensive for regular black-box Bayesian optimization because the training and evaluation of a single model can easily take several hours. To overcome this, we introduce a comprehensive tool suite for effective multi-fidelity Bayesian optimization and the analysis of its runs. The suite, written in Python, provides a simple way to specify complex design spaces, a robust and efficient combination of Bayesian optimization and HyperBand, and a comprehensive analysis of the optimization process and its outcomes."
                    },
                    {
                        "title": "Squirrel: A Switching Hyperparameter Optimizer",
                        "abstract": "In this short note, we describe our submission to the NeurIPS 2020 BBO challenge. Motivated by the fact that different optimizers work well on different problems, our approach switches between different optimizers. Since the team names on the competition's leaderboard were randomly generated \"alliteration nicknames\", consisting of an adjective and an animal with the same initial letter, we called our approach the Switching Squirrel, or here, short, Squirrel."
                    },
                    {
                        "title": "OpenML Benchmarking Suites",
                        "abstract": "Machine learning research depends on objectively interpretable, comparable, and reproducible algorithm benchmarks. We advocate the use of curated, comprehensive suites of machine learning tasks to standardize the setup, execution, and reporting of benchmarks. We enable this through software tools that help to create and leverage these benchmarking suites. These are seamlessly integrated into the OpenML platform, and accessible through interfaces in Python, Java, and R. OpenML benchmarking suites (a) are easy to use through standardized data formats, APIs, and client libraries; (b) come with extensive meta-information on the included datasets; and (c) allow benchmarks to be shared and reused in future studies. We then present a first, carefully curated and practical benchmarking suite for classification: the OpenML Curated Classification benchmarking suite 2018 (OpenML-CC18). Finally, we discuss use cases and applications which demonstrate the usefulness of OpenML benchmarking suites and the OpenML-CC18 in particular."
                    },
                    {
                        "title": "Interpretable Machine Learning for TabPFN",
                        "abstract": "The recently developed Prior-Data Fitted Networks (PFNs) have shown very promising results for applications in low-data regimes. The TabPFN model, a special case of PFNs for tabular data, is able to achieve state-of-the-art performance on a variety of classification tasks while producing posterior predictive distributions in mere seconds by in-context learning without the need for learning parameters or hyperparameter tuning. This makes TabPFN a very attractive option for a wide range of domain applications. However, a major drawback of the method is its lack of interpretability. Therefore, we propose several adaptations of popular interpretability methods that we specifically design for TabPFN. By taking advantage of the unique properties of the model, our adaptations allow for more efficient computations than existing implementations. In particular, we show how in-context learning facilitates the estimation of Shapley values by avoiding approximate retraining and enables the use of Leave-One-Covariate-Out (LOCO) even when working with large-scale Transformers. In addition, we demonstrate how data valuation methods can be used to address scalability challenges of TabPFN. Our proposed methods are implemented in a package tabpfn_iml and made available at https://github.com/david-rundel/tabpfn_iml."
                    },
                    {
                        "title": "HPOBench: A Collection of Reproducible Multi-Fidelity Benchmark Problems for HPO",
                        "abstract": "To achieve peak predictive performance, hyperparameter optimization (HPO) is a crucial component of machine learning and its applications. Over the last years, the number of efficient algorithms and tools for HPO grew substantially. At the same time, the community is still lacking realistic, diverse, computationally cheap, and standardized benchmarks. This is especially the case for multi-fidelity HPO methods. To close this gap, we propose HPOBench, which includes 7 existing and 5 new benchmark families, with a total of more than 100 multi-fidelity benchmark problems. HPOBench allows to run this extendable set of multi-fidelity HPO benchmarks in a reproducible way by isolating and packaging the individual benchmarks in containers. It also provides surrogate and tabular benchmarks for computationally affordable yet statistically sound evaluations. To demonstrate HPOBench's broad compatibility with various optimization tools, as well as its usefulness, we conduct an exemplary large-scale study evaluating 13 optimizers from 6 optimization tools. We provide HPOBench here: https://github.com/automl/HPOBench."
                    },
                    {
                        "title": "SMAC3: A Versatile Bayesian Optimization Package for Hyperparameter Optimization",
                        "abstract": "Algorithm parameters, in particular hyperparameters of machine learning algorithms, can substantially impact their performance. To support users in determining well-performing hyperparameter configurations for their algorithms, datasets and applications at hand, SMAC3 offers a robust and flexible framework for Bayesian Optimization, which can improve performance within a few evaluations. It offers several facades and pre-sets for typical use cases, such as optimizing hyperparameters, solving low dimensional continuous (artificial) global optimization problems and configuring algorithms to perform well across multiple problem instances. The SMAC3 package is available under a permissive BSD-license at https://github.com/automl/SMAC3."
                    },
                    {
                        "title": "Can Fairness be Automated? Guidelines and Opportunities for Fairness-aware AutoML",
                        "abstract": "The field of automated machine learning (AutoML) introduces techniques that automate parts of the development of machine learning (ML) systems, accelerating the process and reducing barriers for novices. However, decisions derived from ML models can reproduce, amplify, or even introduce unfairness in our societies, causing harm to (groups of) individuals. In response, researchers have started to propose AutoML systems that jointly optimize fairness and predictive performance to mitigate fairness-related harm. However, fairness is a complex and inherently interdisciplinary subject, and solely posing it as an optimization problem can have adverse side effects. With this work, we aim to raise awareness among developers of AutoML systems about such limitations of fairness-aware AutoML, while also calling attention to the potential of AutoML as a tool for fairness research. We present a comprehensive overview of different ways in which fairness-related harm can arise and the ensuing implications for the design of fairness-aware AutoML. We conclude that while fairness cannot be automated, fairness-aware AutoML can play an important role in the toolbox of ML practitioners. We highlight several open technical challenges for future work in this direction. Additionally, we advocate for the creation of more user-centered assistive systems designed to tackle challenges encountered in fairness work"
                    },
                    {
                        "title": "OpenML-Python: an extensible Python API for OpenML",
                        "abstract": "OpenML is an online platform for open science collaboration in machine learning, used to share datasets and results of machine learning experiments. In this paper we introduce OpenML-Python, a client API for Python, opening up the OpenML platform for a wide range of Python-based tools. It provides easy access to all datasets, tasks and experiments on OpenML from within Python. It also provides functionality to conduct machine learning experiments, upload the results to OpenML, and reproduce results which are stored on OpenML. Furthermore, it comes with a scikit-learn plugin and a plugin mechanism to easily integrate other machine learning libraries written in Python into the OpenML ecosystem. Source code and documentation is available at https://github.com/openml/openml-python/."
                    },
                    {
                        "title": "Position: Why We Must Rethink Empirical Research in Machine Learning",
                        "abstract": "We warn against a common but incomplete understanding of empirical research in machine learning that leads to non-replicable results, makes findings unreliable, and threatens to undermine progress in the field. To overcome this alarming situation, we call for more awareness of the plurality of ways of gaining knowledge experimentally but also of some epistemic limitations. In particular, we argue most current empirical machine learning research is fashioned as confirmatory research while it should rather be considered exploratory."
                    },
                    {
                        "title": "Position: A Call to Action for a Human-Centered AutoML Paradigm",
                        "abstract": "Automated machine learning (AutoML) was formed around the fundamental objectives of automatically and efficiently configuring machine learning (ML) workflows, aiding the research of new ML algorithms, and contributing to the democratization of ML by making it accessible to a broader audience. Over the past decade, commendable achievements in AutoML have primarily focused on optimizing predictive performance. This focused progress, while substantial, raises questions about how well AutoML has met its broader, original goals. In this position paper, we argue that a key to unlocking AutoML's full potential lies in addressing the currently underexplored aspect of user interaction with AutoML systems, including their diverse roles, expectations, and expertise. We envision a more human-centered approach in future AutoML research, promoting the collaborative design of ML systems that tightly integrates the complementary strengths of human expertise and AutoML methodologies."
                    }
                ]
            }
        }
    },
    "2310.01542": {
        "paper_data": {
            "title": "Fusing Models with Complementary Expertise",
            "url": "http://arxiv.org/abs/2310.01542v2",
            "arxiv_id": "2310.01542",
            "authors": [
                "Hongyi Wang",
                "Felipe Maia Polo",
                "Yuekai Sun",
                "Souvik Kundu",
                "Eric Xing",
                "Mikhail Yurochkin"
            ],
            "abstract": "Training AI models that generalize across tasks and domains has long been among the open problems driving AI research. The emergence of Foundation Models made it easier to obtain expert models for a given task, but the heterogeneity of data that may be encountered at test time often means that any single expert is insufficient. We consider the Fusion of Experts (FoE) problem of fusing outputs of expert models with complementary knowledge of the data distribution and formulate it as an instance of supervised learning. Our method is applicable to both discriminative and generative tasks and leads to significant performance improvements in image and text classification, text summarization, multiple-choice QA, and automatic evaluation of generated text. We also extend our method to the \"frugal\" setting where it is desired to reduce the number of expert model evaluations at test time. Our implementation is publicly available at https://github.com/hwang595/FoE-ICLR2024.",
            "introduction": "   1 Introduction  Figure 1: Three experts with complementary expertise (geometry, natural science, and history) process an input question on the Pythagorean theorem. They each output responses that are processed by a Fusion of Experts (FoE) model to arrive at a final output. Note that only one expert is capable of producing a high-quality output, thus ensembling the experts is likely to perform poorly.   Traditional training of ML/AI models, i.e., empirical risk minimization, allows obtaining experts specialized for a particular task and domain. However, in practice, the distribution of the test data often differs leading to significant performance degradation (Koh et\u00a0al., 2020; Santurkar et\u00a0al., 2020). The emergence of Foundation Models (Bommasani et\u00a0al., 2021) allows obtaining expert models with simple fine-tuning on a handful of data, but such expert models still face generalization issues (Kumar et\u00a0al., 2022; Jang et\u00a0al., 2023; Cheng et\u00a0al., 2023). In the case of Large Language Models (LLMs), although most advanced commercial models can perform well on a range of tasks, they are closed-source, expensive to use, and can be outperformed by smaller specialized models (Hsieh et\u00a0al., 2023; Gunasekar et\u00a0al., 2023; Jang et\u00a0al., 2023; Fu et\u00a0al., 2023). In other words, training expert models have become extremely effective, while generalization with a single model remains challenging. In this paper, we aim to develop methods for combining the strengths of models with complementary expertise to push their collective generalization capabilities.   Combining experts is a long-standing and successful practice in ML: classical approaches like ensembling (Dietterich, 2000; Ganaie et\u00a0al., 2022) or Mixture of Experts (MoE) (Jacobs et\u00a0al., 1991; Jordan & Jacobs, 1994; Masoudnia & Ebrahimpour, 2014) often lead to performance improvements. These techniques typically consider experts trained and tested on the same data distribution, although improvements in terms of out-of-distribution (OOD) generalization have also been observed (Lakshminarayanan et\u00a0al., 2017; Shen et\u00a0al., 2021). In contrast, in this paper, we revisit the practice of combining experts in a new setting where we are given pre-trained experts with complementary expertise, and our emphasis is on generalization to test data distributions where none of the experts perform well individually. See Figure 1 for an illustration.   More specifically, we consider a problem where the data distribution is comprised of K\ud835\udc3eKitalic_K domains, and we have access to K\ud835\udc3eKitalic_K expert models, one for each of the domains. We do not make any assumptions about how experts were trained to make it compatible with modern practices of fine-tuning Foundation Models to obtain high-quality experts. At test time, the data is sampled from the mixture of the K\ud835\udc3eKitalic_K domains, i.e., it contains data from every domain, and thus any individual expert will be sub-optimal. Our goal is to train a model using expert outputs that produces final predictions or generations either by choosing one of the experts or by combining their outputs, for a given input data point. We refer to such models as the Fusion of Experts (FoE) models. In our experiments, we consider tasks such as image/text classification, text generation, and automatic evaluation of generated summaries (Mani, 2001). Finally, recognizing that obtaining the outputs of every expert can be expensive, we consider the \u201cfrugal\u201d setting (Chen",
            "references": [
                {
                    "title": "Adapting Large Language Models to Domains via Reading Comprehension",
                    "abstract": "We explore how continued pre-training on domain-specific corpora influences large language models, revealing that training on the raw corpora endows the model with domain knowledge, but drastically hurts its prompting ability for question answering. Taken inspiration from human learning via reading comprehension--practice after reading improves the ability to answer questions based on the learned knowledge--we propose a simple method for transforming raw corpora into reading comprehension texts. Each raw text is enriched with a series of tasks related to its content. Our method, highly scalable and applicable to any pre-training corpora, consistently enhances performance across various tasks in three different domains: biomedicine, finance, and law. Notably, our 7B language model achieves competitive performance with domain-specific models of much larger scales, such as BloombergGPT-50B. Furthermore, we demonstrate that domain-specific reading comprehension texts can improve the model's performance even on general benchmarks, showing the potential to develop a general model across even more domains. Our model, code, and data are available at https://github.com/microsoft/LMOps."
                },
                {
                    "title": "Textbooks Are All You Need",
                    "abstract": "We introduce phi-1, a new large language model for code, with significantly smaller size than competing models: phi-1 is a Transformer-based model with 1.3B parameters, trained for 4 days on 8 A100s, using a selection of ``textbook quality\"data from the web (6B tokens) and synthetically generated textbooks and exercises with GPT-3.5 (1B tokens). Despite this small scale, phi-1 attains pass@1 accuracy 50.6% on HumanEval and 55.5% on MBPP. It also displays surprising emergent properties compared to phi-1-base, our model before our finetuning stage on a dataset of coding exercises, and phi-1-small, a smaller model with 350M parameters trained with the same pipeline as phi-1 that still achieves 45% on HumanEval."
                },
                {
                    "title": "Confidence-Based Model Selection: When to Take Shortcuts for Subpopulation Shifts",
                    "abstract": "Effective machine learning models learn both robust features that directly determine the outcome of interest (e.g., an object with wheels is more likely to be a car), and shortcut features (e.g., an object on a road is more likely to be a car). The latter can be a source of error under distributional shift, when the correlations change at test-time. The prevailing sentiment in the robustness literature is to avoid such correlative shortcut features and learn robust predictors. However, while robust predictors perform better on worst-case distributional shifts, they often sacrifice accuracy on majority subpopulations. In this paper, we argue that shortcut features should not be entirely discarded. Instead, if we can identify the subpopulation to which an input belongs, we can adaptively choose among models with different strengths to achieve high performance on both majority and minority subpopulations. We propose COnfidence-baSed MOdel Selection (CosMoS), where we observe that model confidence can effectively guide model selection. Notably, CosMoS does not require any target labels or group annotations, either of which may be difficult to obtain or unavailable. We evaluate CosMoS on four datasets with spurious correlations, each with multiple test sets with varying levels of data distribution shift. We find that CosMoS achieves 2-5% lower average regret across all subpopulations, compared to using only robust predictors or other model aggregation methods."
                },
                {
                    "title": "LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fusion",
                    "abstract": "We present LLM-Blender, an ensembling framework designed to attain consistently superior performance by leveraging the diverse strengths of multiple open-source large language models (LLMs). Our framework consists of two modules: PairRanker and GenFuser, addressing the observation that optimal LLMs for different examples can significantly vary. PairRanker employs a specialized pairwise comparison method to distinguish subtle differences between candidate outputs. It jointly encodes the input text and a pair of candidates, using cross-attention encoders to determine the superior one. Our results demonstrate that PairRanker exhibits the highest correlation with ChatGPT-based ranking. Then, GenFuser aims to merge the top-ranked candidates, generating an improved output by capitalizing on their strengths and mitigating their weaknesses. To facilitate large-scale evaluation, we introduce a benchmark dataset, MixInstruct, which is a mixture of multiple instruction datasets featuring oracle pairwise comparisons. Our LLM-Blender significantly outperform individual LLMs and baseline methods across various metrics, establishing a substantial performance gap."
                },
                {
                    "title": "FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance",
                    "abstract": "There is a rapidly growing number of large language models (LLMs) that users can query for a fee. We review the cost associated with querying popular LLM APIs, e.g. GPT-4, ChatGPT, J1-Jumbo, and find that these models have heterogeneous pricing structures, with fees that can differ by two orders of magnitude. In particular, using LLMs on large collections of queries and text can be expensive. Motivated by this, we outline and discuss three types of strategies that users can exploit to reduce the inference cost associated with using LLMs: 1) prompt adaptation, 2) LLM approximation, and 3) LLM cascade. As an example, we propose FrugalGPT, a simple yet flexible instantiation of LLM cascade which learns which combinations of LLMs to use for different queries in order to reduce cost and improve accuracy. Our experiments show that FrugalGPT can match the performance of the best individual LLM (e.g. GPT-4) with up to 98% cost reduction or improve the accuracy over GPT-4 by 4% with the same cost. The ideas and findings presented here lay a foundation for using LLMs sustainably and efficiently."
                },
                {
                    "title": "Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes",
                    "abstract": "Deploying large language models (LLMs) is challenging because they are memory inefficient and compute-intensive for practical applications. In reaction, researchers train smaller task-specific models by either finetuning with human labels or distilling using LLM-generated labels. However, finetuning and distillation require large amounts of training data to achieve comparable performance to LLMs. We introduce Distilling step-by-step, a new mechanism that (a) trains smaller models that outperform LLMs, and (b) achieves so by leveraging less training data needed by finetuning or distillation. Our method extracts LLM rationales as additional supervision for training small models within a multi-task framework. We present three findings across 4 NLP benchmarks: First, compared to both finetuning and distillation, our mechanism achieves better performance with much fewer labeled/unlabeled training examples. Second, compared to few-shot prompted LLMs, we achieve better performance using substantially smaller model sizes. Third, we reduce both the model size and the amount of data required to outperform LLMs; our finetuned 770M T5 model outperforms the few-shot prompted 540B PaLM model using only 80% of available data on a benchmark, whereas standard finetuning the same T5 model struggles to match even by using 100% of the dataset. We release the code at: https://github.com/google-research/distilling-step-by-step ."
                },
                {
                    "title": "Exploring the Benefits of Training Expert Language Models over Instruction Tuning",
                    "abstract": "Recently, Language Models (LMs) instruction-tuned on multiple tasks, also known as multitask-prompted fine-tuning (MT), have shown the capability to generalize to unseen tasks. Previous work has shown that scaling the number of training tasks is the key component in making stronger MT LMs. In this work, we report an unexpected finding that an expert LM fine-tuned on just a single task can outperform an MT LM trained with 300+ different tasks on 11 different unseen datasets and on 13 datasets of the BIG-bench benchmark by a mean accuracy of 3.20% and 1.29%, respectively. This finding casts doubt on the previously held belief that simply scaling the number of tasks makes stronger MT LMs. Leveraging this finding, we further show that this distributed approach of training a separate expert LM per training task instead of a single MT LM for zero-shot inference possesses many benefits including (1) avoiding negative task transfer that often occurs during instruction tuning, (2) being able to continually learn new tasks without having to re-train on previous tasks to avoid catastrophic forgetting, and (3) showing compositional capabilities when merging individual experts together. The code is available at https://github.com/joeljang/ELM."
                },
                {
                    "title": "Specializing Smaller Language Models towards Multi-Step Reasoning",
                    "abstract": "The surprising ability of Large Language Models (LLMs) to perform well on complex reasoning with only few-shot chain-of-thought prompts is believed to emerge only in very large-scale models (100+ billion parameters). We show that such abilities can, in fact, be distilled down from GPT-3.5 ($\\ge$ 175B) to T5 variants ($\\le$ 11B). We propose model specialization, to specialize the model's ability towards a target task. The hypothesis is that large models (commonly viewed as larger than 100B) have strong modeling power, but are spread on a large spectrum of tasks. Small models (commonly viewed as smaller than 10B) have limited model capacity, but if we concentrate their capacity on a specific target task, the model can achieve a decent improved performance. We use multi-step math reasoning as our testbed because it is a very typical emergent ability. We show two important aspects of model abilities: (1). there exists a very complex balance/ tradeoff between language models' multi-dimensional abilities; (2). by paying the price of decreased generic ability, we can clearly lift up the scaling curve of models smaller than 10B towards a specialized multi-step math reasoning ability. We further give comprehensive discussions about important design choices for better generalization, including the tuning data format, the start model checkpoint, and a new model selection method. We hope our practice and discoveries can serve as an important attempt towards specialized smaller models in the new research paradigm set by LLMs."
                },
                {
                    "title": "Towards a Unified Multi-Dimensional Evaluator for Text Generation",
                    "abstract": "Multi-dimensional evaluation is the dominant paradigm for human evaluation in Natural Language Generation (NLG), i.e., evaluating the generated text from multiple explainable dimensions, such as coherence and fluency. However, automatic evaluation in NLG is still dominated by similarity-based metrics, and we lack a reliable framework for a more comprehensive evaluation of advanced models. In this paper, we propose a unified multi-dimensional evaluator UniEval for NLG. We re-frame NLG evaluation as a Boolean Question Answering (QA) task, and by guiding the model with different questions, we can use one evaluator to evaluate from multiple dimensions. Furthermore, thanks to the unified Boolean QA format, we are able to introduce an intermediate learning phase that enables UniEval to incorporate external knowledge from multiple related tasks and gain further improvement. Experiments on three typical NLG tasks show that UniEval correlates substantially better with human judgments than existing metrics. Specifically, compared to the top-performing unified evaluators, UniEval achieves a 23% higher correlation on text summarization, and over 43% on dialogue response generation. Also, UniEval demonstrates a strong zero-shot learning ability for unseen evaluation dimensions and tasks. Source code, data, and all pre-trained evaluators are available at https://github.com/maszhongming/UniEval."
                },
                {
                    "title": "Out-of-Distribution Detection and Selective Generation for Conditional Language Models",
                    "abstract": "Machine learning algorithms typically assume independent and identically distributed samples in training and at test time. Much work has shown that high-performing ML classifiers can degrade significantly and provide overly-confident, wrong classification predictions, particularly for out-of-distribution (OOD) inputs. Conditional language models (CLMs) are predominantly trained to classify the next token in an output sequence, and may suffer even worse degradation on OOD inputs as the prediction is done auto-regressively over many steps. Furthermore, the space of potential low-quality outputs is larger as arbitrary text can be generated and it is important to know when to trust the generated output. We present a highly accurate and lightweight OOD detection method for CLMs, and demonstrate its effectiveness on abstractive summarization and translation. We also show how our method can be used under the common and realistic setting of distribution shift for selective generation (analogous to selective prediction for classification) of high-quality outputs, while automatically abstaining from low-quality ones, enabling safer deployment of generative language models."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "SummaReranker: A Multi-Task Mixture-of-Experts Re-ranking Framework for Abstractive Summarization",
                    "abstract": "Sequence-to-sequence neural networks have recently achieved great success in abstractive summarization, especially through fine-tuning large pre-trained language models on the downstream dataset. These models are typically decoded with beam search to generate a unique summary. However, the search space is very large, and with the exposure bias, such decoding is not optimal. In this paper, we show that it is possible to directly train a second-stage model performing re-ranking on a set of summary candidates. Our mixture-of-experts SummaReranker learns to select a better candidate and consistently improves the performance of the base model. With a base PEGASUS, we push ROUGE scores by 5.44% on CNN- DailyMail (47.16 ROUGE-1), 1.31% on XSum (48.12 ROUGE-1) and 9.34% on Reddit TIFU (29.83 ROUGE-1), reaching a new state-of-the-art. Our code and checkpoints will be available at https://github.com/ntunlp/SummaReranker."
                },
                {
                    "title": "Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution",
                    "abstract": "When transferring a pretrained model to a downstream task, two popular methods are full fine-tuning (updating all the model parameters) and linear probing (updating only the last linear layer\u2014the \u201chead\u201d). It is well known that fine-tuning leads to better accuracy in-distribution (ID). However, in this paper, we find that fine-tuning can achieve worse accuracy than linear probing out-of-distribution (OOD) when the pretrained features are good and the distribution shift is large. On 10 distribution shift datasets (Breeds-Living17, Breeds-Entity30, DomainNet, CIFAR \u2192 STL, CIFAR10.1, FMoW, ImageNetV2, ImageNet-R, ImageNet-A, ImageNet-Sketch), fine-tuning obtains on average 2% higher accuracy ID but 7% lower accuracy OOD than linear probing. We show theoretically that this tradeoff between ID and OOD accuracy arises even in a simple setting: fine-tuning overparameterized two-layer linear networks. We prove that the OOD error of fine-tuning is high when we initialize with a fixed or random head\u2014this is because while fine-tuning learns the head, the lower layers of the neural network change simultaneously and distort the pretrained features. Our analysis suggests that the easy two-step strategy of linear probing then full fine-tuning (LP-FT), sometimes used as a fine-tuning heuristic, combines the benefits of both fine-tuning and linear probing. Empirically, LP-FT outperforms both fine-tuning and linear probing on the above datasets (1% better ID, 10% better OOD than full fine-tuning)."
                },
                {
                    "title": "On the Evaluation Metrics for Paraphrase Generation",
                    "abstract": "In this paper we revisit automatic metrics for paraphrase evaluation and obtain two findings that disobey conventional wisdom: (1) Reference-free metrics achieve better performance than their reference-based counterparts. (2) Most commonly used metrics do not align well with human annotation.Underlying reasons behind the above findings are explored through additional experiments and in-depth analyses.Based on the experiments and analyses, we propose ParaScore, a new evaluation metric for paraphrase generation. It possesses the merits of reference-based and reference-free metrics and explicitly models lexical divergence. Based on our analysis and improvements, our proposed reference-based outperforms than reference-free metrics.Experimental results demonstrate that ParaScore significantly outperforms existing metrics."
                },
                {
                    "title": "Towards Out-Of-Distribution Generalization: A Survey",
                    "abstract": "Traditional machine learning paradigms are based on the assumption that both training and test data follow the same statistical pattern, which is mathematically referred to as Independent and Identically Distributed ($i.i.d.$). However, in real-world applications, this $i.i.d.$ assumption often fails to hold due to unforeseen distributional shifts, leading to considerable degradation in model performance upon deployment. This observed discrepancy indicates the significance of investigating the Out-of-Distribution (OOD) generalization problem. OOD generalization is an emerging topic of machine learning research that focuses on complex scenarios wherein the distributions of the test data differ from those of the training data. This paper represents the first comprehensive, systematic review of OOD generalization, encompassing a spectrum of aspects from problem definition, methodological development, and evaluation procedures, to the implications and future directions of the field. Our discussion begins with a precise, formal characterization of the OOD generalization problem. Following that, we categorize existing methodologies into three segments: unsupervised representation learning, supervised model learning, and optimization, according to their positions within the overarching learning process. We provide an in-depth discussion on representative methodologies for each category, further elucidating the theoretical links between them. Subsequently, we outline the prevailing benchmark datasets employed in OOD generalization studies. To conclude, we overview the existing body of work in this domain and suggest potential avenues for future research on OOD generalization. A summary of the OOD generalization methodologies surveyed in this paper can be accessed at http://out-of-distribution-generalization.com."
                },
                {
                    "title": "On the Opportunities and Risks of Foundation Models",
                    "abstract": "AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature."
                },
                {
                    "title": "A Field Guide to Federated Optimization",
                    "abstract": "Federated learning and analytics are a distributed approach for collaboratively learning models (or statistics) from decentralized data, motivated by and designed for privacy protection. The distributed learning process can be formulated as solving federated optimization problems, which emphasize communication efficiency, data heterogeneity, compatibility with privacy and system requirements, and other constraints that are not primary considerations in other problem settings. This paper provides recommendations and guidelines on formulating, designing, evaluating and analyzing federated optimization algorithms through concrete examples and practical implementation, with a focus on conducting effective simulations to infer real-world performance. The goal of this work is not to survey the current literature, but to inspire researchers and practitioners to design federated learning algorithms that can be used in various practical applications."
                },
                {
                    "title": "BARTScore: Evaluating Generated Text as Text Generation",
                    "abstract": "A wide variety of NLP applications, such as machine translation, summarization, and dialog, involve text generation. One major challenge for these applications is how to evaluate whether such generated texts are actually fluent, accurate, or effective. In this work, we conceptualize the evaluation of generated text as a text generation problem, modeled using pre-trained sequence-to-sequence models. The general idea is that models trained to convert the generated text to/from a reference output or the source text will achieve higher scores when the generated text is better. We operationalize this idea using BART, an encoder-decoder based pre-trained model, and propose a metric BARTScore with a number of variants that can be flexibly applied in an unsupervised fashion to evaluation of text from different perspectives (e.g. informativeness, fluency, or factuality). BARTScore is conceptually simple and empirically effective. It can outperform existing top-scoring metrics in 16 of 22 test settings, covering evaluation of 16 datasets (e.g., machine translation, text summarization) and 7 different perspectives (e.g., informativeness, factuality). Code to calculate BARTScore is available at https://github.com/neulab/BARTScore, and we have released an interactive leaderboard for meta-evaluation at http://explainaboard.nlpedia.ai/leaderboard/task-meval/ on the ExplainaBoard platform, which allows us to interactively understand the strengths, weaknesses, and complementarity of each metric."
                },
                {
                    "title": "Understanding Softmax Confidence and Uncertainty",
                    "abstract": "It is often remarked that neural networks fail to increase their uncertainty when predicting on data far from the training distribution. Yet naively using softmax confidence as a proxy for uncertainty achieves modest success in tasks exclusively testing for this, e.g., out-of-distribution (OOD) detection. This paper investigates this contradiction, identifying two implicit biases that do encourage softmax confidence to correlate with epistemic uncertainty: 1) Approximately optimal decision boundary structure, and 2) Filtering effects of deep networks. It describes why low-dimensional intuitions about softmax confidence are misleading. Diagnostic experiments quantify reasons softmax confidence can fail, finding that extrapolations are less to blame than overlap between training and OOD data in final-layer representations. Pre-trained/fine-tuned networks reduce this overlap."
                },
                {
                    "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity",
                    "abstract": "In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with outrageous numbers of parameters -- but a constant computational cost. However, despite several notable successes of MoE, widespread adoption has been hindered by complexity, communication costs and training instability -- we address these with the Switch Transformer. We simplify the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs. Our proposed training techniques help wrangle the instabilities and we show large sparse models may be trained, for the first time, with lower precision (bfloat16) formats. We design models based off T5-Base and T5-Large to obtain up to 7x increases in pre-training speed with the same computational resources. These improvements extend into multilingual settings where we measure gains over the mT5-Base version across all 101 languages. Finally, we advance the current scale of language models by pre-training up to trillion parameter models on the\"Colossal Clean Crawled Corpus\"and achieve a 4x speedup over the T5-XXL model."
                },
                {
                    "title": "WILDS: A Benchmark of in-the-Wild Distribution Shifts",
                    "abstract": "Distribution shifts -- where the training distribution differs from the test distribution -- can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity in the real-world deployments, these distribution shifts are under-represented in the datasets widely used in the ML community today. To address this gap, we present WILDS, a curated benchmark of 10 datasets reflecting a diverse range of distribution shifts that naturally arise in real-world applications, such as shifts across hospitals for tumor identification; across camera traps for wildlife monitoring; and across time and location in satellite imaging and poverty mapping. On each dataset, we show that standard training yields substantially lower out-of-distribution than in-distribution performance. This gap remains even with models trained by existing methods for tackling distribution shifts, underscoring the need for new methods for training models that are more robust to the types of distribution shifts that arise in practice. To facilitate method development, we provide an open-source package that automates dataset loading, contains default model architectures and hyperparameters, and standardizes evaluations. Code and leaderboards are available at https://wilds.stanford.edu."
                },
                {
                    "title": "Generalized Negative Correlation Learning for Deep Ensembling",
                    "abstract": "Ensemble algorithms offer state of the art performance in many machine learning applications. A common explanation for their excellent performance is due to the bias-variance decomposition of the mean squared error which shows that the algorithm's error can be decomposed into its bias and variance. Both quantities are often opposed to each other and ensembles offer an effective way to manage them as they reduce the variance through a diverse set of base learners while keeping the bias low at the same time. Even though there have been numerous works on decomposing other loss functions, the exact mathematical connection is rarely exploited explicitly for ensembling, but merely used as a guiding principle. In this paper, we formulate a generalized bias-variance decomposition for arbitrary twice differentiable loss functions and study it in the context of Deep Learning. We use this decomposition to derive a Generalized Negative Correlation Learning (GNCL) algorithm which offers explicit control over the ensemble's diversity and smoothly interpolates between the two extremes of independent training and the joint training of the ensemble. We show how GNCL encapsulates many previous works and discuss under which circumstances training of an ensemble of Neural Networks might fail and what ensembling method should be favored depending on the choice of the individual networks."
                },
                {
                    "title": "Investigating Societal Biases in a Poetry Composition System",
                    "abstract": "There is a growing collection of work analyzing and mitigating societal biases in language understanding, generation, and retrieval tasks, though examining biases in creative tasks remains underexplored. Creative language applications are meant for direct interaction with users, so it is important to quantify and mitigate societal biases in these applications. We introduce a novel study on a pipeline to mitigate societal biases when retrieving next verse suggestions in a poetry composition system. Our results suggest that data augmentation through sentiment style transfer has potential for mitigating societal biases."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "BREEDS: Benchmarks for Subpopulation Shift",
                    "abstract": "We develop a methodology for assessing the robustness of models to subpopulation shift---specifically, their ability to generalize to novel data subpopulations that were not observed during training. Our approach leverages the class structure underlying existing datasets to control the data subpopulations that comprise the training and test distributions. This enables us to synthesize realistic distribution shifts whose sources can be precisely controlled and characterized, within existing large-scale datasets. Applying this methodology to the ImageNet dataset, we create a suite of subpopulation shift benchmarks of varying granularity. We then validate that the corresponding shifts are tractable by obtaining human baselines for them. Finally, we utilize these benchmarks to measure the sensitivity of standard model architectures as well as the effectiveness of off-the-shelf train-time robustness interventions. Code and data available at this https URL ."
                },
                {
                    "title": "SummEval: Re-evaluating Summarization Evaluation",
                    "abstract": "Abstract The scarcity of comprehensive up-to-date studies on evaluation metrics for text summarization and the lack of consensus regarding evaluation protocols continue to inhibit progress. We address the existing shortcomings of summarization evaluation methods along five dimensions: 1) we re-evaluate 14 automatic evaluation metrics in a comprehensive and consistent fashion using neural summarization model outputs along with expert and crowd-sourced human annotations; 2) we consistently benchmark 23 recent summarization models using the aforementioned automatic evaluation metrics; 3) we assemble the largest collection of summaries generated by models trained on the CNN/DailyMail news dataset and share it in a unified format; 4) we implement and share a toolkit that provides an extensible and unified API for evaluating summarization models across a broad range of automatic metrics; and 5) we assemble and share the largest and most diverse, in terms of model types, collection of human judgments of model-generated summaries on the CNN/Daily Mail dataset annotated by both expert judges and crowd-source workers. We hope that this work will help promote a more complete evaluation protocol for text summarization as well as advance research in developing evaluation metrics that better correlate with human judgments."
                },
                {
                    "title": "Ensemble Distillation for Robust Model Fusion in Federated Learning",
                    "abstract": "Federated Learning (FL) is a machine learning setting where many devices collaboratively train a machine learning model while keeping the training data decentralized. In most of the current training schemes the central model is refined by averaging the parameters of the server model and the updated parameters from the client side. However, directly averaging model parameters is only possible if all models have the same structure and size, which could be a restrictive constraint in many scenarios. \nIn this work we investigate more powerful and more flexible aggregation schemes for FL. Specifically, we propose ensemble distillation for model fusion, i.e. training the central classifier through unlabeled data on the outputs of the models from the clients. This knowledge distillation technique mitigates privacy risk and cost to the same extent as the baseline FL algorithms, but allows flexible aggregation over heterogeneous client models that can differ e.g. in size, numerical precision or structure. We show in extensive empirical experiments on various CV/NLP datasets (CIFAR-10/100, ImageNet, AG News, SST2) and settings (heterogeneous models/data) that the server model can be trained much faster, requiring fewer communication rounds than any existing FL technique so far."
                },
                {
                    "title": "FrugalML: How to Use ML Prediction APIs More Accurately and Cheaply",
                    "abstract": "Prediction APIs offered for a fee are a fast-growing industry and an important part of machine learning as a service. While many such services are available, the heterogeneity in their price and performance makes it challenging for users to decide which API or combination of APIs to use for their own data and budget. We take a first step towards addressing this challenge by proposing FrugalML, a principled framework that jointly learns the strength and weakness of each API on different data, and performs an efficient optimization to automatically identify the best sequential strategy to adaptively use the available APIs within a budget constraint. Our theoretical analysis shows that natural sparsity in the formulation can be leveraged to make FrugalML efficient. We conduct systematic experiments using ML APIs from Google, Microsoft, Amazon, IBM, Baidu and other providers for tasks including facial emotion recognition, sentiment analysis and speech recognition. Across various tasks, FrugalML can achieve up to 90% cost reduction while matching the accuracy of the best single API, or up to 5% better accuracy while matching the best API's cost."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "On Faithfulness and Factuality in Abstractive Summarization",
                    "abstract": "It is well known that the standard likelihood training and approximate decoding objectives in neural text generation models lead to less human-like responses for open-ended tasks such as language modeling and story generation. In this paper we have analyzed limitations of these models for abstractive document summarization and found that these models are highly prone to hallucinate content that is unfaithful to the input document. We conducted a large scale human evaluation of several neural abstractive summarization systems to better understand the types of hallucinations they produce. Our human annotators found substantial amounts of hallucinated content in all model generated summaries. However, our analysis does show that pretrained models are better summarizers not only in terms of raw metrics, i.e., ROUGE, but also in generating faithful and factual summaries as evaluated by humans. Furthermore, we show that textual entailment measures better correlate with faithfulness than standard metrics, potentially leading the way to automatic evaluation metrics as well as training and decoding criteria."
                },
                {
                    "title": "Asking and Answering Questions to Evaluate the Factual Consistency of Summaries",
                    "abstract": "Practical applications of abstractive summarization models are limited by frequent factual inconsistencies with respect to their input. Existing automatic evaluation metrics for summarization are largely insensitive to such errors. We propose QAGS (pronounced \u201ckags\u201d), an automatic evaluation protocol that is designed to identify factual inconsistencies in a generated summary. QAGS is based on the intuition that if we ask questions about a summary and its source, we will receive similar answers if the summary is factually consistent with the source. To evaluate QAGS, we collect human judgments of factual consistency on model-generated summaries for the CNN/DailyMail (Hermann et al., 2015) and XSUM (Narayan et al., 2018) summarization datasets. QAGS has substantially higher correlations with these judgments than other automatic evaluation metrics. Also, QAGS offers a natural form of interpretability: The answers and questions generated while computing QAGS indicate which tokens of a summary are inconsistent and why. We believe QAGS is a promising tool in automatically generating usable and factually consistent text. Code for QAGS will be available at https://github.com/W4ngatang/qags."
                },
                {
                    "title": "Federated Learning with Matched Averaging",
                    "abstract": "Federated learning allows edge devices to collaboratively learn a shared model while keeping the training data on device, decoupling the ability to do model training from the need to store the data in the cloud. We propose Federated matched averaging (FedMA) algorithm designed for federated learning of modern neural network architectures e.g. convolutional neural networks (CNNs) and LSTMs. FedMA constructs the shared global model in a layer-wise manner by matching and averaging hidden elements (i.e. channels for convolution layers; hidden states for LSTM; neurons for fully connected layers) with similar feature extraction signatures. Our experiments indicate that FedMA outperforms popular state-of-the-art federated learning algorithms on deep CNN and LSTM architectures trained on real world datasets, while improving the communication efficiency."
                },
                {
                    "title": "PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization",
                    "abstract": "Recent work pre-training Transformers with self-supervised objectives on large text corpora has shown great success when fine-tuned on downstream NLP tasks including text summarization. However, pre-training objectives tailored for abstractive text summarization have not been explored. Furthermore there is a lack of systematic evaluation across diverse domains. In this work, we propose pre-training large Transformer-based encoder-decoder models on massive text corpora with a new self-supervised objective. In PEGASUS, important sentences are removed/masked from an input document and are generated together as one output sequence from the remaining sentences, similar to an extractive summary. We evaluated our best PEGASUS model on 12 downstream summarization tasks spanning news, science, stories, instructions, emails, patents, and legislative bills. Experiments demonstrate it achieves state-of-the-art performance on all 12 downstream datasets measured by ROUGE scores. Our model also shows surprising performance on low-resource summarization, surpassing previous state-of-the-art results on 6 datasets with only 1000 examples. Finally we validated our results using human evaluation and show that our model summaries achieve human performance on multiple datasets."
                },
                {
                    "title": "Advances and Open Problems in Federated Learning",
                    "abstract": "Federated learning (FL) is a machine learning setting where many clients (e.g. mobile devices or whole organizations) collaboratively train a model under the orchestration of a central server (e.g. service provider), while keeping the training data decentralized. FL embodies the principles of focused data collection and minimization, and can mitigate many of the systemic privacy risks and costs resulting from traditional, centralized machine learning and data science approaches. Motivated by the explosive growth in FL research, this paper discusses recent advances and presents an extensive collection of open problems and challenges."
                },
                {
                    "title": "BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension",
                    "abstract": "We present BART, a denoising autoencoder for pretraining sequence-to-sequence models. BART is trained by (1) corrupting text with an arbitrary noising function, and (2) learning a model to reconstruct the original text. It uses a standard Tranformer-based neural machine translation architecture which, despite its simplicity, can be seen as generalizing BERT (due to the bidirectional encoder), GPT (with the left-to-right decoder), and other recent pretraining schemes. We evaluate a number of noising approaches, finding the best performance by both randomly shuffling the order of sentences and using a novel in-filling scheme, where spans of text are replaced with a single mask token. BART is particularly effective when fine tuned for text generation but also works well for comprehension tasks. It matches the performance of RoBERTa on GLUE and SQuAD, and achieves new state-of-the-art results on a range of abstractive dialogue, question answering, and summarization tasks, with gains of up to 3.5 ROUGE. BART also provides a 1.1 BLEU increase over a back-translation system for machine translation, with only target language pretraining. We also replicate other pretraining schemes within the BART framework, to understand their effect on end-task performance."
                },
                {
                    "title": "FedMD: Heterogenous Federated Learning via Model Distillation",
                    "abstract": "Federated learning enables the creation of a powerful centralized model without compromising data privacy of multiple participants. While successful, it does not incorporate the case where each participant independently designs its own model. Due to intellectual property concerns and heterogeneous nature of tasks and data, this is a widespread requirement in applications of federated learning to areas such as health care and AI as a service. In this work, we use transfer learning and knowledge distillation to develop a universal framework that enables federated learning when each agent owns not only their private data, but also uniquely designed models. We test our framework on the MNIST/FEMNIST dataset and the CIFAR10/CIFAR100 dataset and observe fast improvement across all participating models. With 10 distinct participants, the final test accuracy of each model on average receives a 20% gain on top of what's possible without collaboration and is only a few percent lower than the performance each model would have obtained if all private datasets were pooled and made directly available for all participants."
                },
                {
                    "title": "Nonlinear Regression via Deep Negative Correlation Learning",
                    "abstract": "Nonlinear regression has been extensively employed in many computer vision problems (e.g., crowd counting, age estimation, affective computing). Under the umbrella of deep learning, two common solutions exist i) transforming nonlinear regression to a robust loss function which is jointly optimizable with the deep convolutional network, and ii) utilizing ensemble of deep networks. Although some improved performance is achieved, the former may be lacking due to the intrinsic limitation of choosing a single hypothesis and the latter may suffer from much larger computational complexity. To cope with those issues, we propose to regress via an efficient \u201cdivide and conquer\u201d manner. The core of our approach is the generalization of negative correlation learning that has been shown, both theoretically and empirically, to work well for non-deep regression problems. Without extra parameters, the proposed method controls the bias-variance-covariance trade-off systematically and usually yields a deep regression ensemble where each base model is both \u201caccurate\u201d and \u201cdiversified.\u201d Moreover, we show that each sub-problem in the proposed method has less Rademacher Complexity and thus is easier to optimize. Extensive experiments on several diverse and challenging tasks including crowd counting, personality analysis, age estimation, and image super-resolution demonstrate the superiority over challenging baselines as well as the versatility of the proposed method. The source code and trained models are available on our project page: https://mmcheng.net/dncl/."
                },
                {
                    "title": "BERTScore: Evaluating Text Generation with BERT",
                    "abstract": "We propose BERTScore, an automatic evaluation metric for text generation. Analogously to common metrics, BERTScore computes a similarity score for each token in the candidate sentence with each token in the reference sentence. However, instead of exact matches, we compute token similarity using contextual embeddings. We evaluate using the outputs of 363 machine translation and image captioning systems. BERTScore correlates better with human judgments and provides stronger model selection performance than existing metrics. Finally, we use an adversarial paraphrase detection task to show that BERTScore is more robust to challenging examples when compared to existing metrics."
                },
                {
                    "title": "Crowd Counting with Deep Negative Correlation Learning",
                    "abstract": "Deep convolutional networks (ConvNets) have achieved unprecedented performances on many computer vision tasks. However, their adaptations to crowd counting on single images are still in their infancy and suffer from severe over-fitting. Here we propose a new learning strategy to produce generalizable features by way of deep negative correlation learning (NCL). More specifically, we deeply learn a pool of decorrelated regressors with sound generalization capabilities through managing their intrinsic diversities. Our proposed method, named decorrelated ConvNet (D-ConvNet), is end-to-end-trainable and independent of the backbone fully-convolutional network architectures. Extensive experiments on very deep VGGNet as well as our customized network structure indicate the superiority of D-ConvNet when compared with several state-of-the-art methods. Our implementation will be released at https://github.com/shizenglin/Deep-NCL"
                },
                {
                    "title": "Newsroom: A Dataset of 1.3 Million Summaries with Diverse Extractive Strategies",
                    "abstract": "We present NEWSROOM, a summarization dataset of 1.3 million articles and summaries written by authors and editors in newsrooms of 38 major news publications. Extracted from search and social media metadata between 1998 and 2017, these high-quality summaries demonstrate high diversity of summarization styles. In particular, the summaries combine abstractive and extractive strategies, borrowing words and phrases from articles at varying rates. We analyze the extraction strategies used in NEWSROOM summaries against other datasets to quantify the diversity and difficulty of our new data, and train existing methods on the data to evaluate its utility and challenges. The dataset is available online at summari.es."
                },
                {
                    "title": "Detecting and Correcting for Label Shift with Black Box Predictors",
                    "abstract": "Faced with distribution shift between training and test set, we wish to detect and quantify the shift, and to correct our classifiers without test set labels. Motivated by medical diagnosis, where diseases (targets), cause symptoms (observations), we focus on label shift, where the label marginal $p(y)$ changes but the conditional $p(x|y)$ does not. We propose Black Box Shift Estimation (BBSE) to estimate the test distribution $p(y)$. BBSE exploits arbitrary black box predictors to reduce dimensionality prior to shift correction. While better predictors give tighter estimates, BBSE works even when predictors are biased, inaccurate, or uncalibrated, so long as their confusion matrices are invertible. We prove BBSE's consistency, bound its error, and introduce a statistical test that uses BBSE to detect shift. We also leverage BBSE to correct classifiers. Experiments demonstrate accurate estimates and improved prediction, even on high-dimensional datasets of natural images"
                },
                {
                    "title": "mixup: Beyond Empirical Risk Minimization",
                    "abstract": "Large deep neural networks are powerful, but exhibit undesirable behaviors such as memorization and sensitivity to adversarial examples. In this work, we propose mixup, a simple learning principle to alleviate these issues. In essence, mixup trains a neural network on convex combinations of pairs of examples and their labels. By doing so, mixup regularizes the neural network to favor simple linear behavior in-between training examples. Our experiments on the ImageNet-2012, CIFAR-10, CIFAR-100, Google commands and UCI datasets show that mixup improves the generalization of state-of-the-art neural network architectures. We also find that mixup reduces the memorization of corrupt labels, increases the robustness to adversarial examples, and stabilizes the training of generative adversarial networks."
                },
                {
                    "title": "CatBoost: unbiased boosting with categorical features",
                    "abstract": "This paper presents the key algorithmic techniques behind CatBoost, a new gradient boosting toolkit. Their combination leads to CatBoost outperforming other publicly available boosting implementations in terms of quality on a variety of datasets. Two critical algorithmic advances introduced in CatBoost are the implementation of ordered boosting, a permutation-driven alternative to the classic algorithm, and an innovative algorithm for processing categorical features. Both techniques were created to fight a prediction shift caused by a special kind of target leakage present in all currently existing implementations of gradient boosting algorithms. In this paper, we provide a detailed analysis of this problem and demonstrate that proposed algorithms solve it effectively, leading to excellent empirical results."
                },
                {
                    "title": "On Calibration of Modern Neural Networks",
                    "abstract": "Confidence calibration -- the problem of predicting probability estimates representative of the true correctness likelihood -- is important for classification models in many applications. We discover that modern neural networks, unlike those from a decade ago, are poorly calibrated. Through extensive experiments, we observe that depth, width, weight decay, and Batch Normalization are important factors influencing calibration. We evaluate the performance of various post-processing calibration methods on state-of-the-art architectures with image and document classification datasets. Our analysis and experiments not only offer insights into neural network learning, but also provide a simple and straightforward recipe for practical settings: on most datasets, temperature scaling -- a single-parameter variant of Platt Scaling -- is surprisingly effective at calibrating predictions."
                },
                {
                    "title": "Selective Classification for Deep Neural Networks",
                    "abstract": "Selective classification techniques (also known as reject option) have not yet been considered in the context of deep neural networks (DNNs). These techniques can potentially significantly improve DNNs prediction performance by trading-off coverage. In this paper we propose a method to construct a selective classifier given a trained neural network. Our method allows a user to set a desired risk level. At test time, the classifier rejects instances as needed, to grant the desired risk (with high probability). Empirical results over CIFAR and ImageNet convincingly demonstrate the viability of our method, which opens up possibilities to operate DNNs in mission-critical applications. For example, using our method an unprecedented 2% error in top-5 ImageNet classification can be guaranteed with probability 99.9%, and almost 60% test coverage."
                },
                {
                    "title": "Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles",
                    "abstract": "Deep neural networks (NNs) are powerful black box predictors that have recently achieved impressive performance on a wide spectrum of tasks. Quantifying predictive uncertainty in NNs is a challenging and yet unsolved problem. Bayesian NNs, which learn a distribution over weights, are currently the state-of-the-art for estimating predictive uncertainty; however these require significant modifications to the training procedure and are computationally expensive compared to standard (non-Bayesian) NNs. We propose an alternative to Bayesian NNs that is simple to implement, readily parallelizable, requires very little hyperparameter tuning, and yields high quality predictive uncertainty estimates. Through a series of experiments on classification and regression benchmarks, we demonstrate that our method produces well-calibrated uncertainty estimates which are as good or better than approximate Bayesian NNs. To assess robustness to dataset shift, we evaluate the predictive uncertainty on test examples from known and unknown distributions, and show that our method is able to express higher uncertainty on out-of-distribution examples. We demonstrate the scalability of our method by evaluating predictive uncertainty estimates on ImageNet."
                },
                {
                    "title": "Communication-Efficient Learning of Deep Networks from Decentralized Data",
                    "abstract": "Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data center and training there using conventional approaches. We advocate an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. \nWe present a practical method for the federated learning of deep networks based on iterative model averaging, and conduct an extensive empirical evaluation, considering five different model architectures and four datasets. These experiments demonstrate the approach is robust to the unbalanced and non-IID data distributions that are a defining characteristic of this setting. Communication costs are the principal constraint, and we show a reduction in required communication rounds by 10-100x as compared to synchronized stochastic gradient descent."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Neural network with ensembles",
                    "abstract": "Pattern recognition systems usually have a relatively small number of patterns to be recognized. As a rule the number of handwritten symbols, number of phonemes or number of human faces are of the order of some dozens. But sometimes the pattern recognition task demands much more classes. For example, a continuous speech recognition system can be created on the base of syllables; a handwriting recognition system will be more efficient if the recognized units are not different letters, but triplets of letters. In these cases it is necessary to have various thousands of classes. In this paper we will consider the situation of the recognition problem that demands many thousands of classes to be recognized. For such problems we propose the use of the neural networks with ensembles. We give a short description of this type of neural network and calculate its storage capacity."
                },
                {
                    "title": "The Elements of Statistical Learning: Data Mining, Inference, and Prediction",
                    "abstract": "In the words of the authors, the goal of this book was to \u201cbring together many of the important new ideas in learning, and explain them in a statistical framework.\u201d The authors have been quite successful in achieving this objective, and their work is a welcome addition to the statistics and learning literatures. Statistics has always been interdisciplinary, borrowing ideas from diverse \u008e elds and repaying the debt with contributions, both theoretical and practical, to the other intellectual disciplines. For statistical learning, this cross-fertilization is especially noticeable. This book is a valuable resource, both for the statistician needing an introduction to machine learning and related \u008e elds and for the computer scientist wishing to learn more about statistics. Statisticians will especially appreciate that it is written in their own language. The level of the book is roughly that of a second-year doctoral student in statistics, and it will be useful as a textbook for such students. In a stimulating article, Breiman (2001) argued that statistics has been focused too much on a \u201cdata modeling culture,\u201d where the model is paramount. Breiman argued instead for an \u201calgorithmic modeling culture,\u201d with emphasis on black-box types of prediction. Breiman\u2019s article is controversial, and in his discussion, Efron objects that \u201cprediction is certainly an interesting subject, but Leo\u2019s paper overstates both its role and our profession\u2019s lack of interest in it.\u201d Although I mostly agree with Efron, I worry that the courses offered by most statistics departments include little, if any, treatment of statistical learning and prediction. (Stanford, where Efron and the authors of this book teach, is an exception.) Graduate students in statistics certainly need to know more than they do now about prediction, machine learning, statistical learning, and data mining (not disjoint subjects). I hope that graduate courses covering the topics of this book will become more common in statistics curricula. Most of the book is focused on supervised learning, where one has inputs and outputs from some system and wishes to predict unknown outputs corresponding to known inputs. The methods discussed for supervised learning include linear and logistic regression; basis expansion, such as splines and wavelets; kernel techniques, such as local regression, local likelihood, and radial basis functions; neural networks; additive models; decision trees based on recursive partitioning, such as CART; and support vector machines. There is a \u008e nal chapter on unsupervised learning, including association rules, cluster analysis, self-organizing maps, principal components and curves, and independent component analysis. Many statisticians will be unfamiliar with at least some of these algorithms. Association rules are popular for mining commercial data in what is called \u201cmarket basket analysis.\u201d The aim is to discover types of products often purchased together. Such knowledge can be used to develop marketing strategies, such as store or catalog layouts. Self-organizing maps (SOMs) involve essentially constrained k-means clustering, where prototypes are mapped to a two-dimensional curved coordinate system. Independent components analysis is similar to principal components analysis and factor analysis, but it uses higher-order moments to achieve independence, not merely zero correlation between components. A strength of the book is the attempt to organize a plethora of methods into a coherent whole. The relationships among the methods are emphasized. I know of no other book that covers so much ground. Of course, with such broad coverage, it is not possible to cover any single topic in great depth, so this book will encourage further reading. Fortunately, each chapter includes bibliographic notes surveying the recent literature. These notes and the extensive references provide a good introduction to the learning literature, including much outside of statistics. The book might be more suitable as a textbook if less material were covered in greater depth; however, such a change would compromise the book\u2019s usefulness as a reference, and so I am happier with the book as it was written."
                },
                {
                    "title": "Automatic Evaluation of Summaries Using N-gram Co-occurrence Statistics",
                    "abstract": "Following the recent adoption by the machine translation community of automatic evaluation using the BLEU/NIST scoring process, we conduct an in-depth study of a similar idea for evaluating summaries. The results show that automatic evaluation using unigram co-occurrences between summary pairs correlates surprising well with human evaluations, based on various statistical metrics; while direct application of the BLEU evaluation procedure does not always give good results."
                },
                {
                    "title": "Greedy function approximation: A gradient boosting machine.",
                    "abstract": "Function estimation/approximation is viewed from the perspective of numerical optimization in function space, rather than parameter space. A connection is made between stagewise additive expansions and steepest-descent minimization. A general gradient descent boosting paradigm is developed for additive expansions based on any fitting criterion. Specific algorithms are presented for least-squares, least absolute deviation, and Huber-M loss functions for regression, and multiclass logistic likelihood for classification. Special enhancements are derived for the particular case where the individual additive components are regression trees, and tools for interpreting such TreeBoost models are presented. Gradient boosting of regression trees produces competitive, highly robust, interpretable procedures for both regression and classification, especially appropriate for mining less than clean data. Connections between this approach and the boosting methods of Freund and Shapire and Friedman, Hastie and Tibshirani are discussed."
                },
                {
                    "title": "Experiments with a New Boosting Algorithm",
                    "abstract": "In an earlier paper, we introduced a new \"boosting\" algorithm called AdaBoost which, theoretically, can be used to significantly reduce the error of any learning algorithm that con- sistently generates classifiers whose performance is a little better than random guessing. We also introduced the related notion of a \"pseudo-loss\" which is a method for forcing a learning algorithm of multi-label concepts to concentrate on the labels that are hardest to discriminate. In this paper, we describe experiments we carried out to assess how well AdaBoost with and without pseudo-loss, performs on real learning problems. We performed two sets of experiments. The first set compared boosting to Breiman's \"bagging\" method when used to aggregate various classifiers (including decision trees and single attribute- value tests). We compared the performance of the two methods on a collection of machine-learning benchmarks. In the second set of experiments, we studied in more detail the performance of boosting using a nearest-neighbor classifier on an OCR problem."
                },
                {
                    "title": "Hierarchical Mixtures of Experts and the EM Algorithm",
                    "abstract": "We present a tree-structured architecture for supervised learning. The statistical model underlying the architecture is a hierarchical mixture model in which both the mixture coefficients and the mixture components are generalized linear models (GLIM's). Learning is treated as a maximum likelihood problem; in particular, we present an Expectation-Maximization (EM) algorithm for adjusting the parameters of the architecture. We also develop an on-line learning algorithm in which the parameters are updated incrementally. Comparative simulation results are presented in the robot dynamics domain."
                },
                {
                    "title": "Adaptive Mixtures of Local Experts",
                    "abstract": "We present a new supervised learning procedure for systems composed of many separate networks, each of which learns to handle a subset of the complete set of training cases. The new procedure can be viewed either as a modular version of a multilayer supervised network, or as an associative version of competitive learning. It therefore provides a new link between these two apparently different approaches. We demonstrate that the learning procedure divides up a vowel discrimination task into appropriate subtasks, each of which can be solved by a very simple expert network."
                },
                {
                    "title": "WikiSum: Coherent Summarization Dataset for Efficient Human-Evaluation",
                    "abstract": "Recent works made significant advances on summarization tasks, facilitated by summarization datasets. Several existing datasets have the form of coherent-paragraph summaries. However, these datasets were curated from academic documents that were written for experts, thus making the essential step of assessing the summarization output through human-evaluation very demanding. To overcome these limitations, we present a dataset based on article summaries appearing on the WikiHow website, composed of how-to articles and coherent-paragraph summaries written in plain language. We compare our dataset attributes to existing ones, including readability and world-knowledge, showing our dataset makes human evaluation significantly easier and thus, more effective. A human evaluation conducted on PubMed and the proposed dataset reinforces our findings."
                },
                {
                    "title": "Dropout: a simple way to prevent neural networks from overfitting",
                    "abstract": "Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units (along with their connections) from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different \"thinned\" networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights. This significantly reduces overfitting and gives major improvements over other regularization methods. We show that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Ensemble Methods in Machine Learning",
                    "abstract": "Ensemble methods are learning algorithms that construct a set of classiiers and then classify new data points by taking a (weighted) vote of their predictions. The original ensemble method is Bayesian averaging , but more recent algorithms include error-correcting output coding, Bagging, and boosting. This paper reviews these methods and explains why ensembles can often perform better than any single classiier. Some previous studies comparing ensemble methods are reviewed, and some new experiments are presented to uncover the reasons that Adaboost does not overrt rapidly."
                },
                {
                    "title": "Neural Network Ensembles, Cross Validation, and Active Learning",
                    "abstract": "Learning of continuous valued functions using neural network ensembles (committees) can give improved accuracy, reliable estimation of the generalization error, and active learning. The ambiguity is defined as the variation of the output of ensemble members averaged over unlabeled data, so it quantifies the disagreement among the networks. It is discussed how to use the ambiguity in combination with cross-validation to give a reliable estimate of the ensemble generalization error, and how this type of ensemble cross-validation can sometimes improve performance. It is shown how to estimate the optimal weights of the ensemble members using unlabeled data. By a generalization of query by committee, it is finally shown how the ambiguity can be used to select new training data to be labeled in an active learning scheme."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively combine the strengths of pre-trained expert models with complementary expertise to improve generalization capabilities on test data distributions where none of the experts perform well individually?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of model generalization, particularly in scenarios where data distributions differ significantly from training conditions. By developing effective Fusion of Experts (FoE) models, we can enhance the performance of machine learning systems across various tasks, leading to more robust applications in real-world scenarios. This research could pave the way for future studies on model ensembling and generalization, ultimately contributing to the development of more efficient and effective AI systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the fact that traditional ensembling methods assume that experts are trained and tested on the same data distribution, which is not the case here. Naive approaches may fail because they do not account for the unique strengths and weaknesses of each expert in the context of diverse test data. Additionally, the lack of assumptions about the training processes of the experts complicates the integration of their outputs. Overcoming these technical and theoretical obstacles requires innovative strategies to effectively leverage the complementary expertise of the models.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on ensembling methods that operate under the assumption of consistent data distributions for training and testing. This limitation has prevented the exploration of combining experts with varying training conditions. Additionally, existing solutions often overlook the complexities involved in generalizing across multiple domains. Our approach differs by explicitly addressing the challenge of integrating outputs from experts trained on different domains, thereby enhancing the generalization capabilities of the resulting model.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves training a Fusion of Experts (FoE) model that utilizes outputs from K expert models, each specialized in a different domain. We will evaluate the model on tasks such as image/text classification, text generation, and automatic evaluation of generated summaries. The dataset will consist of mixed-domain data, and we will measure performance using metrics appropriate for each task. We expect that our approach will lead to improved generalization performance on out-of-distribution data, demonstrating the effectiveness of combining expert outputs in a frugal setting."
            }
        },
        "author_data": {
            "c32d4613-29f3-482a-aaed-c480c3378b21": {
                "pk": "c32d4613-29f3-482a-aaed-c480c3378b21",
                "name": "Hongyi Wang",
                "collaborators": [
                    "Bowen Tan",
                    "Yun Zhu",
                    "Lijuan Liu",
                    "Yonghao Zhuang",
                    "Jindong Chen",
                    "Eric P. Xing",
                    "Zhiting Hu"
                ],
                "domain": [
                    "Machine Learning",
                    "Large Language Models",
                    "Distributed Systems",
                    "Model Parallelism"
                ],
                "publications": [
                    {
                        "title": "RedCoast: A Lightweight Tool to Automate Distributed Training of LLMs on Any GPU/TPUs",
                        "abstract": "The recent progress of AI can be largely attributed to large language models (LLMs). However, their escalating memory requirements introduce challenges for machine learning (ML) researchers and engineers. Addressing this requires developers to partition a large model to distribute it across multiple GPUs or TPUs. This necessitates considerable coding and intricate configuration efforts with existing model parallel tools, such as Megatron-LM, DeepSpeed, and Alpa. These tools require users\u2019 expertise in machine learning systems (MLSys), creating a bottleneck in LLM development, particularly for developers without MLSys background. In this work, we present RedCoast (Redco), a lightweight and user-friendly tool crafted to automate distributed training and inference for LLMs, as well as to simplify ML pipeline development. The design of Redco emphasizes two key aspects. Firstly, to automate model parallelism, our study identifies two straightforward rules to generate tensor parallel strategies for any given LLM. Integrating these rules into Redco facilitates effortless distributed LLM training and inference, eliminating the need of additional coding or complex configurations. We demonstrate the effectiveness by applying Redco on a set of LLM architectures, such as GPT-J, LLaMA, T5, and OPT, up to the size of 66B. Secondly, we propose a mechanism that allows for the customization of diverse ML pipelines through the definition of merely three functions, avoiding redundant and formulaic code like multi-host related processing. This mechanism proves adaptable across a spectrum of ML algorithms, from foundational language modeling to complex algorithms like meta-learning and reinforcement learning. As a result, Redco implementations exhibit significantly fewer lines of code compared to their official counterparts. RedCoast (Redco) has been released under Apache 2.0 license at https://github.com/tanyuqian/redco."
                    },
                    {
                        "title": "RedCoast: A Lightweight Tool to Automate Distributed Training of LLMs on Any GPU/TPUs",
                        "abstract": "The recent progress of AI can be largely attributed to large language models (LLMs). However, their escalating memory requirements introduce challenges for machine learning (ML) researchers and engineers. Addressing this requires developers to partition a large model to distribute it across multiple GPUs or TPUs. This necessitates considerable coding and intricate configuration efforts with existing model parallel tools, such as Megatron-LM, DeepSpeed, and Alpa. These tools require users' expertise in machine learning systems (MLSys), creating a bottleneck in LLM development, particularly for developers without MLSys background. In this work, we present RedCoast (Redco), a lightweight and user-friendly tool crafted to automate distributed training and inference for LLMs, as well as to simplify ML pipeline development. The design of Redco emphasizes two key aspects. Firstly, to automate model parallelism, our study identifies two straightforward rules to generate tensor parallel strategies for any given LLM. Integrating these rules into Redco facilitates effortless distributed LLM training and inference, eliminating the need of additional coding or complex configurations. We demonstrate the effectiveness by applying Redco on a set of LLM architectures, such as GPT-J, LLaMA, T5, and OPT, up to the size of 66B. Secondly, we propose a mechanism that allows for the customization of diverse ML pipelines through the definition of merely three functions, avoiding redundant and formulaic code like multi-host related processing. This mechanism proves adaptable across a spectrum of ML algorithms, from foundational language modeling to complex algorithms like meta-learning and reinforcement learning. As a result, Redco implementations exhibit significantly fewer lines of code compared to their official counterparts."
                    },
                    {
                        "title": "Redco: A Lightweight Tool to Automate Distributed Training of LLMs on Any GPU/TPUs",
                        "abstract": "."
                    }
                ]
            },
            "af72cee0-80d3-4602-b8d9-93eb3a8c6418": {
                "pk": "af72cee0-80d3-4602-b8d9-93eb3a8c6418",
                "name": "Felipe Maia Polo",
                "collaborators": [
                    "Yuekai Sun",
                    "M. Yurochkin",
                    "Moulinath Banerjee",
                    "R. Vicente",
                    "Itamar Ciochetti",
                    "Emerson Bertolo",
                    "Lucas Weber",
                    "Leshem Choshen",
                    "L. Gianvechio",
                    "Jonathan Batista Ferreira",
                    "Nimrod Shabtay",
                    "Sivan Doveh",
                    "Wei Lin",
                    "M. Mirza",
                    "Leshem Chosen",
                    "Assaf Arbelle",
                    "Leonid Karlinsky",
                    "Raja Giryes",
                    "Gongjun Xu",
                    "Seamus Somerstep",
                    "Ya'acov Ritov",
                    "Ronald Xu",
                    "M'irian Silva",
                    "Onkar Bhardwaj",
                    "Allysson Flavio Melo de Oliveira",
                    "M. Banerjee",
                    "Subha Maity",
                    "Gabriel Caiaffa Floriano Mendon\u00e7a",
                    "Kau\u00ea Capellato J. Parreira",
                    "Peterson Cordeiro",
                    "Leticia Maria Paz de Lima",
                    "Antonio Carlos do Amaral Maia",
                    "C. M. Toledo",
                    "Guilherme Mendes Bassedon",
                    "Carlos Guatimosim",
                    "A. A. Okuno"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Benchmarking",
                    "Legal Technology"
                ],
                "publications": [
                    {
                        "title": "LiveXiv -- A Multi-Modal Live Benchmark Based on Arxiv Papers Content",
                        "abstract": "The large-scale training of multi-modal models on data scraped from the web has shown outstanding utility in infusing these models with the required world knowledge to perform effectively on multiple downstream tasks. However, one downside of scraping data from the web can be the potential sacrifice of the benchmarks on which the abilities of these models are often evaluated. To safeguard against test data contamination and to truly test the abilities of these foundation models we propose LiveXiv: A scalable evolving live benchmark based on scientific ArXiv papers. LiveXiv accesses domain-specific manuscripts at any given timestamp and proposes to automatically generate visual question-answer pairs (VQA). This is done without any human-in-the-loop, using the multi-modal content in the manuscripts, like graphs, charts, and tables. Moreover, we introduce an efficient evaluation approach that estimates the performance of all models on the evolving benchmark using evaluations of only a subset of models. This significantly reduces the overall evaluation cost. We benchmark multiple open and proprietary Large Multi-modal Models (LMMs) on the first version of our benchmark, showing its challenging nature and exposing the models true abilities, avoiding contamination. Lastly, in our commitment to high quality, we have collected and evaluated a manually verified subset. By comparing its overall results to our automatic annotations, we have found that the performance variance is indeed minimal (<2.5%). Our dataset is available online on HuggingFace, and our code will be available here."
                    },
                    {
                        "title": "tinyBenchmarks: evaluating LLMs with fewer examples",
                        "abstract": "The versatility of large language models (LLMs) led to the creation of diverse benchmarks that thoroughly test a variety of language models' abilities. These benchmarks consist of tens of thousands of examples making evaluation of LLMs very expensive. In this paper, we investigate strategies to reduce the number of evaluations needed to assess the performance of an LLM on several key benchmarks. For example, we show that to accurately estimate the performance of an LLM on MMLU, a popular multiple-choice QA benchmark consisting of 14K examples, it is sufficient to evaluate this LLM on 100 curated examples. We release evaluation tools and tiny versions of popular benchmarks: Open LLM Leaderboard, MMLU, HELM, and AlpacaEval 2.0. Our empirical analysis demonstrates that these tools and tiny benchmarks are sufficient to reliably and efficiently reproduce the original evaluation results."
                    },
                    {
                        "title": "A statistical framework for weak-to-strong generalization",
                        "abstract": "Modern large language model (LLM) alignment techniques rely on human feedback, but it is unclear whether the techniques fundamentally limit the capabilities of aligned LLMs. In particular, it is unclear whether it is possible to align (stronger) LLMs with superhuman capabilities with (weaker) human feedback without degrading their capabilities. This is an instance of the weak-to-strong generalization problem: using weaker (less capable) feedback to train a stronger (more capable) model. We prove that weak-to-strong generalization is possible by eliciting latent knowledge from pre-trained LLMs. In particular, we cast the weak-to-strong generalization problem as a transfer learning problem in which we wish to transfer a latent concept from a weak model to a strong pre-trained model. We prove that a naive fine-tuning approach suffers from fundamental limitations, but an alternative refinement-based approach suggested by the problem structure provably overcomes the limitations of fine-tuning. Finally, we demonstrate the practical applicability of the refinement approach with three LLM alignment tasks."
                    },
                    {
                        "title": "Efficient multi-prompt evaluation of LLMs",
                        "abstract": "Most popular benchmarks for comparing LLMs rely on a limited set of prompt templates, which may not fully capture the LLMs' abilities and can affect the reproducibility of results on leaderboards. Many recent works empirically verify prompt sensitivity and advocate for changes in LLM evaluation. In this paper, we consider the problem of estimating the performance distribution across many prompt variants instead of finding a single prompt to evaluate with. We introduce PromptEval, a method for estimating performance across a large set of prompts borrowing strength across prompts and examples to produce accurate estimates under practical evaluation budgets. The resulting distribution can be used to obtain performance quantiles to construct various robust performance metrics (e.g., top 95% quantile or median). We prove that PromptEval consistently estimates the performance distribution and demonstrate its efficacy empirically on three prominent LLM benchmarks: MMLU, BIG-bench Hard, and LMentry; for example, PromptEval can accurately estimate performance quantiles across 100 prompt templates on MMLU with a budget equivalent to two single-prompt evaluations. Moreover, we show how PromptEval can be useful in LLM-as-a-judge and best prompt identification applications."
                    },
                    {
                        "title": "Conditional independence testing under model misspecification",
                        "abstract": "Conditional independence (CI) testing is fundamental and challenging in modern statistics and machine learning. Many modern methods for CI testing rely on powerful supervised learning methods to learn regression functions or Bayes predictors as an intermediate step. Although the methods are guaranteed to control Type-I error when the supervised learning methods accurately estimate the regression functions or Bayes predictors, their behavior is less understood when they fail due to model misspeci\ufb01cation. In a broader sense, model misspeci\ufb01cation can arise even when universal approximators ( e.g. , deep neural nets) are employed. Then, we study the performance of regression-based CI tests under model misspeci\ufb01cation. Namely, we propose new approximations or upper bounds for the testing errors of three regression-based tests that depend on misspeci\ufb01cation errors. Moreover, we introduce the Rao-Blackwellized Predictor Test (RBPT), a novel regression-based CI test robust against model misspeci\ufb01cation. Finally, we conduct experiments with arti\ufb01cial and real data, showcasing the usefulness of our theory and methods."
                    },
                    {
                        "title": "Weak Supervision Performance Evaluation via Partial Identification",
                        "abstract": "Programmatic Weak Supervision (PWS) enables supervised model training without direct access to ground truth labels, utilizing weak labels from heuristics, crowdsourcing, or pre-trained models. However, the absence of ground truth complicates model evaluation, as traditional metrics such as accuracy, precision, and recall cannot be directly calculated. In this work, we present a novel method to address this challenge by framing model evaluation as a partial identification problem and estimating performance bounds using Fr\\'echet bounds. Our approach derives reliable bounds on key metrics without requiring labeled data, overcoming core limitations in current weak supervision evaluation techniques. Through scalable convex optimization, we obtain accurate and computationally efficient bounds for metrics including accuracy, precision, recall, and F1-score, even in high-dimensional settings. This framework offers a robust approach to assessing model quality without ground truth labels, enhancing the practicality of weakly supervised learning for real-world applications."
                    },
                    {
                        "title": "Conditional independence testing under misspecified inductive biases",
                        "abstract": "Conditional independence (CI) testing is a fundamental and challenging task in modern statistics and machine learning. Many modern methods for CI testing rely on powerful supervised learning methods to learn regression functions or Bayes predictors as an intermediate step; we refer to this class of tests as regression-based tests. Although these methods are guaranteed to control Type-I error when the supervised learning methods accurately estimate the regression functions or Bayes predictors of interest, their behavior is less understood when they fail due to misspecified inductive biases; in other words, when the employed models are not flexible enough or when the training algorithm does not induce the desired predictors. Then, we study the performance of regression-based CI tests under misspecified inductive biases. Namely, we propose new approximations or upper bounds for the testing errors of three regression-based tests that depend on misspecification errors. Moreover, we introduce the Rao-Blackwellized Predictor Test (RBPT), a regression-based CI test robust against misspecified inductive biases. Finally, we conduct experiments with artificial and real data, showcasing the usefulness of our theory and methods."
                    },
                    {
                        "title": "LegalNLP - Natural Language Processing methods for the Brazilian Legal Language",
                        "abstract": "We present and make available pre-trained language models (Phraser, Word2Vec, Doc2Vec, FastText, and BERT) for the Brazilian legal language, a Python package with functions to facilitate their use, and a set of demonstrations/tutorials containing some applications involving them. Given that our material is built upon legal texts coming from several Brazilian courts, this initiative is extremely helpful for the Brazilian legal field, which lacks other open and specific tools and language models. Our main objective is to catalyze the use of natural language processing tools for legal texts analysis by the Brazilian industry, government, and academia, providing the necessary tools and accessible material."
                    },
                    {
                        "title": "Effects of personality traits in predicting grade retention of Brazilian students",
                        "abstract": "Student\u2019s grade retention is a key issue faced by many education systems, especially those in developing countries. In this paper, we seek to gauge the relevance of students\u2019 personality traits in predicting grade retention in Brazil. For that, we used data collected in 2012 and 2017, in the city of Sertaozinho, countryside of the state of Sao Paulo, Brazil. The surveys taken in Sertaozinho included several socioeconomic questions, standardized tests, and a personality test. Moreover, students were in grades 4, 5, and 6 in 2012. Our approach was based on training machine learning models on the surveys\u2019 data to predict grade retention between 2012 and 2017 using information from 2012 or before, and then using some strategies to quantify personality traits\u2019 predictive power. We concluded that, besides proving to be fairly better than a random classifier when isolated, personality traits contribute to prediction even when using socioeconomic variables and standardized tests results."
                    },
                    {
                        "title": "OCDE: Odds Conditional Density Estimator",
                        "abstract": "Conditional density estimation (CDE) models can be useful for many statistical applications, especially because the full conditional density is estimated instead of traditional regression point estimates, revealing more information about the uncertainty of the random variable of interest. In this paper, we propose a new methodology called Odds Conditional Density Estimator (OCDE) to estimate conditional densities in a supervised learning scheme. The main idea is that it is very difficult to estimate $p_{x,y}$ and $p_{x}$ in order to estimate the conditional density $p_{y|x}$, but by introducing an instrumental distribution, we transform the CDE problem into a problem of odds estimation, or similarly, training a binary probabilistic classifier. We demonstrate how OCDE works using simulated data and then test its performance against other known state-of-the-art CDE methods in real data. Overall, OCDE is competitive compared with these methods in real datasets."
                    },
                    {
                        "title": "Covariate Shift Adaptation in High-Dimensional and Divergent Distributions",
                        "abstract": "In real world applications of supervised learning methods, training and test sets are often sampled from the distinct distributions and we must resort to domain adaptation techniques. One special class of techniques is Covariate Shift Adaptation, which allows practitioners to obtain good generalization performance in the distribution of interest when domains differ only by the marginal distribution of features. Traditionally, Covariate Shift Adaptation is implemented using Importance Weighting which may fail in high-dimensional settings due to small Effective Sample Sizes (ESS). In this paper, we propose (i) a connection between ESS, high-dimensional settings and generalization bounds and (ii) a simple, general and theoretically sound approach to combine feature selection and Covariate Shift Adaptation. The new approach yields good performance with improved ESS."
                    },
                    {
                        "title": "Predicting legal proceedings status: approaches based on sequential text data",
                        "abstract": "The objective of this work, which is fully given by Polo et al. [8], is to develop predictive models to classify Brazilian legal proceedings in three possible classes of legal status: (i) archived proceedings, (ii) active proceedings, and (iii) suspended proceedings. Each proceeding is made up of a chronological sequence of short texts called \"motions\" written by the courts' administrative staff. The motions relate to the proceedings, but not necessarily to their legal status. Moreover, the proceedings' labels are decided by the courts to organize their workflow. This problem's resolution is intended to assist public and private institutions in managing large portfolios of legal proceedings, providing gains in scale and efficiency."
                    },
                    {
                        "title": "Interpretable Approach in the Classification of Sequences of Legal Texts",
                        "abstract": "Machine learning applications in the legal field are numerous and diverse. In order to make contribution to both the machine learning community and the legal community, we have made efforts to create a model compatible with the classification of text sequences, valuing the interpretability of the results. The purpose of this paper is to classify Brazilian legal proceedings in three possible status classes, which are (i) archived proceedings, (ii) active proceedings and (iii) suspended proceedings. Although working with portuguese NLP, which can be hard due to lack of resources, our approach performed remarkably well in the classification task. Furthermore, we were able to extract and interpret the patterns learnt by the neural network besides quantifying how those patterns relate to the classification task."
                    },
                    {
                        "title": "Skills to not fall behind in school",
                        "abstract": "Many recent studies emphasize how important the role of cognitive and social-emotional skills can be in determining people's quality of life. Although skills are of great importance in many aspects, in this paper we will focus our efforts to better understand the relationship between several types of skills with academic progress delay. Our dataset contains the same students in 2012 and 2017, and we consider that there was a academic progress delay for a specific student if he or she progressed less than expected in school grades. Our methodology primarily includes the use of a Bayesian logistic regression model and our results suggest that both cognitive and social-emotional skills may impact the conditional probability of falling behind in school, and the magnitude of the impact between the two types of skills can be comparable."
                    },
                    {
                        "title": "Predicting Legal Proceedings Status: an Approach Based on Sequential Text Data",
                        "abstract": "Machine learning applications in the legal field are numerous and diverse. In order to make contribution to both the machine learning community and the legal community, we have made efforts to create a model compatible with the classification of text sequences, valuing the interpretability of the results. The purpose of this paper is to classify legal proceedings in three possible status classes, which are (i) archived proceedings, (ii) active proceedings and (iii) suspended proceedings. Our approach is composed by natural language processing, supervised and unsupervised deep learning models and performed remarkably well in the classification task. Furthermore we had some insights regarding the patterns learned by the neural network applying tools to make the results more interpretable."
                    }
                ]
            },
            "c8a523cd-6151-4860-a41b-c6f32fa86ccf": {
                "pk": "c8a523cd-6151-4860-a41b-c6f32fa86ccf",
                "name": "Yuekai Sun",
                "collaborators": [
                    "M. Yurochkin",
                    "Felipe Maia Polo",
                    "Subha Maity",
                    "Moulinath Banerjee",
                    "Lucas Weber",
                    "Leshem Choshen",
                    "Mayank Agarwal",
                    "Seamus Somerstep",
                    "Ronald Xu",
                    "M'irian Silva",
                    "Nimrod Shabtay",
                    "Sivan Doveh",
                    "Wei Lin",
                    "M. Mirza",
                    "Leshem Chosen",
                    "Assaf Arbelle",
                    "Leonid Karlinsky",
                    "Raja Giryes",
                    "Gongjun Xu",
                    "Lilian Ngweta",
                    "Alex Gittens",
                    "Ya'acov Ritov",
                    "Y. Ritov",
                    "Onkar Bhardwaj",
                    "Allysson Flavio Melo de Oliveira",
                    "Michael Feffer",
                    "Daniele Bracale",
                    "Tal Shnitzer",
                    "Anthony Ou",
                    "Kate Soule",
                    "Justin Solomon",
                    "Neil Thompson"
                ],
                "domain": [
                    "Large Language Models",
                    "Multi-modal Learning",
                    "Machine Learning",
                    "Evaluation Metrics"
                ],
                "publications": [
                    {
                        "title": "LiveXiv -- A Multi-Modal Live Benchmark Based on Arxiv Papers Content",
                        "abstract": "The large-scale training of multi-modal models on data scraped from the web has shown outstanding utility in infusing these models with the required world knowledge to perform effectively on multiple downstream tasks. However, one downside of scraping data from the web can be the potential sacrifice of the benchmarks on which the abilities of these models are often evaluated. To safeguard against test data contamination and to truly test the abilities of these foundation models we propose LiveXiv: A scalable evolving live benchmark based on scientific ArXiv papers. LiveXiv accesses domain-specific manuscripts at any given timestamp and proposes to automatically generate visual question-answer pairs (VQA). This is done without any human-in-the-loop, using the multi-modal content in the manuscripts, like graphs, charts, and tables. Moreover, we introduce an efficient evaluation approach that estimates the performance of all models on the evolving benchmark using evaluations of only a subset of models. This significantly reduces the overall evaluation cost. We benchmark multiple open and proprietary Large Multi-modal Models (LMMs) on the first version of our benchmark, showing its challenging nature and exposing the models true abilities, avoiding contamination. Lastly, in our commitment to high quality, we have collected and evaluated a manually verified subset. By comparing its overall results to our automatic annotations, we have found that the performance variance is indeed minimal (<2.5%). Our dataset is available online on HuggingFace, and our code will be available here."
                    },
                    {
                        "title": "tinyBenchmarks: evaluating LLMs with fewer examples",
                        "abstract": "The versatility of large language models (LLMs) led to the creation of diverse benchmarks that thoroughly test a variety of language models' abilities. These benchmarks consist of tens of thousands of examples making evaluation of LLMs very expensive. In this paper, we investigate strategies to reduce the number of evaluations needed to assess the performance of an LLM on several key benchmarks. For example, we show that to accurately estimate the performance of an LLM on MMLU, a popular multiple-choice QA benchmark consisting of 14K examples, it is sufficient to evaluate this LLM on 100 curated examples. We release evaluation tools and tiny versions of popular benchmarks: Open LLM Leaderboard, MMLU, HELM, and AlpacaEval 2.0. Our empirical analysis demonstrates that these tools and tiny benchmarks are sufficient to reliably and efficiently reproduce the original evaluation results."
                    },
                    {
                        "title": "Aligners: Decoupling LLMs and Alignment",
                        "abstract": "Large Language Models (LLMs) need to be aligned with human expectations to ensure their safety and utility in most applications. Alignment is challenging, costly, and needs to be repeated for every LLM and alignment criterion. We propose to decouple LLMs and alignment by training aligner models that can be used to align any LLM for a given criteria on an as-needed basis, thus also reducing the potential negative impacts of alignment on performance. Our recipe for training the aligner models solely relies on synthetic data generated with a (prompted) LLM and can be easily adjusted for a variety of alignment criteria. We use the same synthetic data to train inspectors, binary miss-alignment classification models to guide a\"squad\"of multiple aligners. Our empirical results demonstrate consistent improvements when applying aligner squad to various LLMs, including chat-aligned models, across several instruction-following and red-teaming datasets."
                    },
                    {
                        "title": "A statistical framework for weak-to-strong generalization",
                        "abstract": "Modern large language model (LLM) alignment techniques rely on human feedback, but it is unclear whether the techniques fundamentally limit the capabilities of aligned LLMs. In particular, it is unclear whether it is possible to align (stronger) LLMs with superhuman capabilities with (weaker) human feedback without degrading their capabilities. This is an instance of the weak-to-strong generalization problem: using weaker (less capable) feedback to train a stronger (more capable) model. We prove that weak-to-strong generalization is possible by eliciting latent knowledge from pre-trained LLMs. In particular, we cast the weak-to-strong generalization problem as a transfer learning problem in which we wish to transfer a latent concept from a weak model to a strong pre-trained model. We prove that a naive fine-tuning approach suffers from fundamental limitations, but an alternative refinement-based approach suggested by the problem structure provably overcomes the limitations of fine-tuning. Finally, we demonstrate the practical applicability of the refinement approach with three LLM alignment tasks."
                    },
                    {
                        "title": "Learning In Reverse Causal Strategic Environments With Ramifications on Two Sided Markets",
                        "abstract": "Motivated by equilibrium models of labor markets, we develop a formulation of causal strategic classification in which strategic agents can directly manipulate their outcomes. As an application, we compare employers that anticipate the strategic response of a labor force with employers that do not. We show through a combination of theory and experiment that employers with performatively optimal hiring policies improve employer reward, labor force skill level, and in some cases labor force equity. On the other hand, we demonstrate that performative employers harm labor force utility and fail to prevent discrimination in other cases."
                    },
                    {
                        "title": "Efficient multi-prompt evaluation of LLMs",
                        "abstract": "Most popular benchmarks for comparing LLMs rely on a limited set of prompt templates, which may not fully capture the LLMs' abilities and can affect the reproducibility of results on leaderboards. Many recent works empirically verify prompt sensitivity and advocate for changes in LLM evaluation. In this paper, we consider the problem of estimating the performance distribution across many prompt variants instead of finding a single prompt to evaluate with. We introduce PromptEval, a method for estimating performance across a large set of prompts borrowing strength across prompts and examples to produce accurate estimates under practical evaluation budgets. The resulting distribution can be used to obtain performance quantiles to construct various robust performance metrics (e.g., top 95% quantile or median). We prove that PromptEval consistently estimates the performance distribution and demonstrate its efficacy empirically on three prominent LLM benchmarks: MMLU, BIG-bench Hard, and LMentry; for example, PromptEval can accurately estimate performance quantiles across 100 prompt templates on MMLU with a budget equivalent to two single-prompt evaluations. Moreover, we show how PromptEval can be useful in LLM-as-a-judge and best prompt identification applications."
                    },
                    {
                        "title": "Prompt Exploration with Prompt Regression",
                        "abstract": "In the advent of democratized usage of large language models (LLMs), there is a growing desire to systematize LLM prompt creation and selection processes beyond iterative trial-and-error. Prior works majorly focus on searching the space of prompts without accounting for relations between prompt variations. Here we propose a framework, Prompt Exploration with Prompt Regression (PEPR), to predict the effect of prompt combinations given results for individual prompt elements as well as a simple method to select an effective prompt for a given use-case. We evaluate our approach with open-source LLMs of different sizes on several different tasks."
                    },
                    {
                        "title": "Learning the Distribution Map in Reverse Causal Performative Prediction",
                        "abstract": "In numerous predictive scenarios, the predictive model affects the sampling distribution; for example, job applicants often meticulously craft their resumes to navigate through a screening systems. Such shifts in distribution are particularly prevalent in the realm of social computing, yet, the strategies to learn these shifts from data remain remarkably limited. Inspired by a microeconomic model that adeptly characterizes agents' behavior within labor markets, we introduce a novel approach to learn the distribution shift. Our method is predicated on a reverse causal model, wherein the predictive model instigates a distribution shift exclusively through a finite set of agents' actions. Within this framework, we employ a microfoundation model for the agents' actions and develop a statistically justified methodology to learn the distribution shift map, which we demonstrate to be effective in minimizing the performative prediction risk."
                    },
                    {
                        "title": "Large Language Model Routing with Benchmark Datasets",
                        "abstract": "There is a rapidly growing number of open-source Large Language Models (LLMs) and benchmark datasets to compare them. While some models dominate these benchmarks, no single model typically achieves the best accuracy in all tasks and use cases. In this work, we address the challenge of selecting the best LLM out of a collection of models for new tasks. We propose a new formulation for the problem, in which benchmark datasets are repurposed to learn a\"router\"model for this LLM selection, and we show that this problem can be reduced to a collection of binary classification tasks. We demonstrate the utility and limitations of learning model routers from various benchmark datasets, where we consistently improve performance upon using any single model for all tasks."
                    },
                    {
                        "title": "Weak Supervision Performance Evaluation via Partial Identification",
                        "abstract": "Programmatic Weak Supervision (PWS) enables supervised model training without direct access to ground truth labels, utilizing weak labels from heuristics, crowdsourcing, or pre-trained models. However, the absence of ground truth complicates model evaluation, as traditional metrics such as accuracy, precision, and recall cannot be directly calculated. In this work, we present a novel method to address this challenge by framing model evaluation as a partial identification problem and estimating performance bounds using Fr\\'echet bounds. Our approach derives reliable bounds on key metrics without requiring labeled data, overcoming core limitations in current weak supervision evaluation techniques. Through scalable convex optimization, we obtain accurate and computationally efficient bounds for metrics including accuracy, precision, recall, and F1-score, even in high-dimensional settings. This framework offers a robust approach to assessing model quality without ground truth labels, enhancing the practicality of weakly supervised learning for real-world applications."
                    },
                    {
                        "title": "An Investigation of Representation and Allocation Harms in Contrastive Learning",
                        "abstract": "The effect of underrepresentation on the performance of minority groups is known to be a serious problem in supervised learning settings; however, it has been underexplored so far in the context of self-supervised learning (SSL). In this paper, we demonstrate that contrastive learning (CL), a popular variant of SSL, tends to collapse representations of minority groups with certain majority groups. We refer to this phenomenon as representation harm and demonstrate it on image and text datasets using the corresponding popular CL methods. Furthermore, our causal mediation analysis of allocation harm on a downstream classification task reveals that representation harm is partly responsible for it, thus emphasizing the importance of studying and mitigating representation harm. Finally, we provide a theoretical explanation for representation harm using a stochastic block model that leads to a representational neural collapse in a contrastive learning setting."
                    },
                    {
                        "title": "Conditional independence testing under misspecified inductive biases",
                        "abstract": "Conditional independence (CI) testing is a fundamental and challenging task in modern statistics and machine learning. Many modern methods for CI testing rely on powerful supervised learning methods to learn regression functions or Bayes predictors as an intermediate step; we refer to this class of tests as regression-based tests. Although these methods are guaranteed to control Type-I error when the supervised learning methods accurately estimate the regression functions or Bayes predictors of interest, their behavior is less understood when they fail due to misspecified inductive biases; in other words, when the employed models are not flexible enough or when the training algorithm does not induce the desired predictors. Then, we study the performance of regression-based CI tests under misspecified inductive biases. Namely, we propose new approximations or upper bounds for the testing errors of three regression-based tests that depend on misspecification errors. Moreover, we introduce the Rao-Blackwellized Predictor Test (RBPT), a regression-based CI test robust against misspecified inductive biases. Finally, we conduct experiments with artificial and real data, showcasing the usefulness of our theory and methods."
                    }
                ]
            },
            "0b978368-d22a-43a0-9237-c70c1a8d768f": {
                "pk": "0b978368-d22a-43a0-9237-c70c1a8d768f",
                "name": "Souvik Kundu",
                "collaborators": [
                    "P. Beerel",
                    "Sreetama Sarkar",
                    "G. Datta",
                    "Dake Chen",
                    "Yuke Zhang",
                    "Chenghao Li",
                    "Kai Zheng",
                    "Akhilesh R. Jaiswal",
                    "Maciej Szankin",
                    "Sairam Sundaresan",
                    "Zeyu Liu",
                    "Anni Li",
                    "Junrui Wan",
                    "Lianghao Jiang",
                    "Peter A. Beereil",
                    "Hao Kang",
                    "Qingru Zhang",
                    "Geonhwa Jeong",
                    "Zaoxing Liu",
                    "Tushar Krishna",
                    "Tuo Zhao",
                    "Shiduo Li",
                    "Chirayata Bhattacharyya",
                    "Seyedarmin Azizi",
                    "M. Pedram",
                    "Md. Abdullah-Al Kaiser",
                    "Zihan Yin",
                    "Manas Garg",
                    "Ajey P. Jacob",
                    "Yue Hu",
                    "Xin-Yu Ye",
                    "Yifei Liu",
                    "Srikar Mutnuri",
                    "Namo Asavisanu",
                    "Nora Ayanian",
                    "K. Psounis",
                    "S. Nittur",
                    "Shashank Nag",
                    "N. Chandrachoodan",
                    "Yuhang Li",
                    "Youngeun Kim",
                    "Donghyun Lee",
                    "P. Panda",
                    "Lu Yin",
                    "A. Jaiswal",
                    "Shiwei Liu",
                    "Zhangyang Wang",
                    "Y. Fu",
                    "Ye Yuan",
                    "Shang Wu",
                    "Shunyao Zhang",
                    "Yingyan Lin",
                    "Rui-jie Zhu",
                    "S. N. Sridhar",
                    "Anthony Sarah"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Architecture Search",
                    "Privacy-Preserving AI",
                    "Vision Transformers"
                ],
                "publications": [
                    {
                        "title": "AFLoRA: Adaptive Freezing of Low Rank Adaptation in Parameter Efficient Fine-Tuning of Large Models",
                        "abstract": "We present a novel Parameter-Efficient Fine-Tuning (PEFT) method, dubbed as Adaptive Freezing of Low Rank Adaptation (AFLoRA). Specifically, for each pre-trained frozen weight tensor, we add a parallel path of trainable low-rank matrices, namely a down-projection and an up-projection matrix, each of which is followed by a feature transformation vector. Based on a novel freezing score, we the incrementally freeze these projection matrices during fine-tuning to reduce the computation and alleviate over-fitting. Our experimental results demonstrate that we can achieve state-of-the-art performance with an average improvement of up to $0.85\\%$ as evaluated on GLUE benchmark while yeilding up to $9.5\\times$ fewer average trainable parameters. While compared in terms of runtime, AFLoRA can yield up to $1.86\\times$ improvement as opposed to similar PEFT alternatives. Besides the practical utility of our approach, we provide insights on the trainability requirements of LoRA paths at different modules and the freezing schedule for the different projection matrices. Code will be released."
                    },
                    {
                        "title": "Linearizing Models for Efficient yet Robust Private Inference",
                        "abstract": "The growing concern about data privacy has led to the development of private inference (PI) frameworks in client-server applications which protects both data privacy and model IP. However, the cryptographic primitives required yield significant latency overhead which limits its wide-spread application. At the same time, changing environments demand the PI service to be robust against various naturally occurring and gradient-based perturbations. Despite several works focused on the development of latency-efficient models suitable for PI, the impact of these models on robustness has remained unexplored. Towards this goal, this paper presents RLNet, a class of robust linearized networks that can yield latency improvement via reduction of high-latency ReLU operations while improving the model performance on both clean and corrupted images. In particular, RLNet models provide a\"triple win ticket\"of improved classification accuracy on clean, naturally perturbed, and gradient-based perturbed images using a shared-mask shared-weight architecture with over an order of magnitude fewer ReLUs than baseline models. To demonstrate the efficacy of RLNet, we perform extensive experiments with ResNet and WRN model variants on CIFAR-10, CIFAR-100, and Tiny-ImageNet datasets. Our experimental evaluations show that RLNet can yield models with up to 11.14x fewer ReLUs, with accuracy close to the all-ReLU models, on clean, naturally perturbed, and gradient-based perturbed images. Compared with the SoTA non-robust linearized models at similar ReLU budgets, RLNet achieves an improvement in adversarial accuracy of up to ~47%, naturally perturbed accuracy up to ~16.4%, while improving clean image accuracy up to ~1.5%."
                    },
                    {
                        "title": "RLNet: Robust Linearized Networks for Efficient Private Inference",
                        "abstract": "The growing concern about data privacy has led to the development of private inference (PI) frameworks in client-server applications which protects both data privacy and model IP. However, the cryptographic primitives required yield significant latency overheads which limits their wide-spread application. At the same time, changing environments demand PI services to be robust against various naturally occurring and gradient-based perturbations. Despite several works focused on the development of latency-efficient models suitable for PI, the impact of these models on robustness has remained unexplored. Towards this goal, this paper presents RLNet, a class of models that can yield latency improvement via the reduction of high-latency ReLU operations while improving the model performance on both clean and corrupted images. In particular, RLNet models provide a \"triple win ticket\" of improved classification accuracy on clean, naturally perturbed, and gradient- based perturbed images using a shared-mask shared-weight architecture with over an order of magnitude fewer ReLUs than baseline models. To demonstrate the efficacy of RLNet, we perform extensive experiments with ResNet and WRN model variants on CIFAR-10, CIFAR-100, and Tiny- ImageNet datasets. Our experimental evaluations show that RLNet can yield models with up to 11.14\u00d7 fewer ReLUs, with accuracy close to the all-ReLU models, on clean, naturally perturbed, and gradient-based perturbed images. Compared with the SoTA non-robust linearized models at similar ReLU budgets, RLNet achieves an improvement in adversarial accuracy of up to ~47%, in naturally perturbed accuracy of up to ~16.4%, while improving clean image accuracy up to ~1.5%. Code is available at: https://github.com/sreetamasarkar/rlnet."
                    },
                    {
                        "title": "GEAR: An Efficient KV Cache Compression Recipe for Near-Lossless Generative Inference of LLM",
                        "abstract": "Key-value (KV) caching has become the de-facto to accelerate generation speed for large language models (LLMs) inference. However, the growing cache demand with increasing sequence length has transformed LLM inference to be a memory bound problem, significantly constraining the system throughput. Existing methods rely on dropping unimportant tokens or quantizing all entries uniformly. Such methods, however, often incur high approximation errors to represent the compressed matrices. The autoregressive decoding process further compounds the error of each step, resulting in critical deviation in model generation and deterioration of performance. To tackle this challenge, we propose GEAR, an efficient KV cache compression framework that achieves near-lossless high-ratio compression. GEAR first applies quantization to majority of entries of similar magnitudes to ultra-low precision. It then employs a low rank matrix to approximate the quantization error, and a sparse matrix to remedy individual errors from outlier entries. By adeptly integrating three techniques, GEAR is able to fully exploit their synergistic potentials. Our experiments demonstrate that compared to alternatives, GEAR achieves near-lossless 4-bit KV cache compression with up to 2.38x throughput improvement, while reducing peak-memory size up to 2.29x. Our code is publicly available at https://github.com/HaoKang-Timmy/GEAR."
                    },
                    {
                        "title": "DIA: Diffusion based Inverse Network Attack on Collaborative Inference",
                        "abstract": "With the continuous expansion of neural networks in size and depth, and the growing popularity of machine learning as a service, collaborative inference systems present a promising approach for deploying models in resource-constrained computing environments. However, as the deployment of these systems gains traction, evaluating their privacy and security has become a critical issue. Towards this goal, this paper introduces a diffusion-based inverse network attack, named DIA, for collaborative inference systems that uses a novel feature map awareness conditioning mechanism to guide the diffusion model. Compared to prior approaches, our extensive empirical results demonstrate that the proposed attack achieves an average improvement of 29%, 20%, 30% in terms of SSIM, PSNR, and MSE when applied to convolutional neural networks (CNN), 18%, 17%, 61% to ResNet models, and 55%, 54%, 84% to Vision transformers (ViTs). Our results identify the significant vulnerability of ViTs and analyze the potential sources of this vulnerability. Based on our analysis, we raise caution regarding the deployment of transformer-based models in collaborative inference systems, emphasizing the need for careful consideration regarding the security of such models in collaborative settings."
                    },
                    {
                        "title": "MaskVD: Region Masking for Efficient Video Object Detection",
                        "abstract": "Video tasks are compute-heavy and thus pose a challenge when deploying in real-time applications, particularly for tasks that require state-of-the-art Vision Transformers (ViTs). Several research efforts have tried to address this challenge by leveraging the fact that large portions of the video undergo very little change across frames, leading to redundant computations in frame-based video processing. In particular, some works leverage pixel or semantic differences across frames, however, this yields limited latency benefits with significantly increased memory overhead. This paper, in contrast, presents a strategy for masking regions in video frames that leverages the semantic information in images and the temporal correlation between frames to significantly reduce FLOPs and latency with little to no penalty in performance over baseline models. In particular, we demonstrate that by leveraging extracted features from previous frames, ViT backbones directly benefit from region masking, skipping up to 80% of input regions, improving FLOPs and latency by 3.14x and 1.5x. We improve memory and latency over the state-of-the-art (SOTA) by 2.3x and 1.14x, while maintaining similar detection performance. Additionally, our approach demonstrates promising results on convolutional neural networks (CNNs) and provides latency improvements over the SOTA up to 1.3x using specialized computational kernels."
                    },
                    {
                        "title": "LaMDA: Large Model Fine-Tuning via Spectrally Decomposed Low-Dimensional Adaptation",
                        "abstract": "Low-rank adaptation (LoRA) has become the default approach to fine-tune large language models (LLMs) due to its significant reduction in trainable parameters. However, trainable parameter demand for LoRA increases with increasing model embedding dimensions, leading to high compute costs. Additionally, its backward updates require storing high-dimensional intermediate activations and optimizer states, demanding high peak GPU memory. In this paper, we introduce large model fine-tuning via spectrally decomposed low-dimensional adaptation (LaMDA), a novel approach to fine-tuning large language models, which leverages low-dimensional adaptation to achieve significant reductions in trainable parameters and peak GPU memory footprint. LaMDA freezes a first projection matrix (PMA) in the adaptation path while introducing a low-dimensional trainable square matrix, resulting in substantial reductions in trainable parameters and peak GPU memory usage. LaMDA gradually freezes a second projection matrix (PMB) during the early fine-tuning stages, reducing the compute cost associated with weight updates to enhance parameter efficiency further. We also present an enhancement, LaMDA++, incorporating a ``lite-weight\"adaptive rank allocation for the LoRA path via normalized spectrum analysis of pre-trained model weights. We evaluate LaMDA/LaMDA++ across various tasks, including natural language understanding with the GLUE benchmark, text summarization, natural language generation, and complex reasoning on different LLMs. Results show that LaMDA matches or surpasses the performance of existing alternatives while requiring up to 17.7x fewer parameter updates and up to 1.32x lower peak GPU memory usage during fine-tuning. Code will be publicly available."
                    },
                    {
                        "title": "Block Selective Reprogramming for On-device Training of Vision Transformers",
                        "abstract": "The ubiquity of vision transformers (ViTs) for various edge applications, including personalized learning, has created the demand for on-device fine-tuning. However, training with the limited memory and computation power of edge devices remains a significant challenge. In particular, the memory required for training is much higher than that needed for inference, primarily due to the need to store activations across all layers in order to compute the gradients needed for weight updates. Previous works have explored reducing this memory requirement via frozen-weight training as well storing the activations in a compressed format. However, these methods are deemed inefficient due to their inability to provide training or inference speedup. In this paper, we first investigate the limitations of existing on-device training methods aimed at reducing memory and compute requirements. We then present block selective reprogramming (BSR) in which we fine-tune only a fraction of total blocks of a pre-trained model and selectively drop tokens based on self-attention scores of the frozen layers. To show the efficacy of BSR, we present extensive evaluations on ViT-B and DeiT-S with five different datasets. Compared to the existing alternatives, our approach simultaneously reduces training memory by up to 1.4\u00d7 and compute cost by up to 2\u00d7 while maintaining similar accuracy. We also showcase results for Mixture-of-Expert (MoE) models, demonstrating the effectiveness of our approach in multitask learning scenarios. Code will be available at: https://github.com/sreetamasarkar/BSR."
                    },
                    {
                        "title": "Technology-Circuit-Algorithm Tri-Design for Processing-in-Pixel-in-Memory (P2M)",
                        "abstract": "The massive amounts of data generated by camera sensors motivate data processing inside pixel arrays, i.e., at the extreme-edge. Several critical developments have fueled recent interest in the processing-in-pixel-in-memory paradigm for a wide range of visual machine intelligence tasks, including (1) advances in 3D integration technology to enable complex processing inside each pixel in a 3D integrated manner while maintaining pixel density, (2) analog processing circuit techniques for massively parallel low-energy in-pixel computations, and (3) algorithmic techniques to mitigate non-idealities associated with analog processing through hardware-aware training schemes. This article presents a comprehensive technology-circuit-algorithm landscape that connects technology capabilities, circuit design strategies, and algorithmic optimizations to power, performance, area, bandwidth reduction, and application-level accuracy metrics. We present our results using a comprehensive co-design framework incorporating hardware and algorithmic optimizations for various complex real-life visual intelligence tasks mapped onto our P2M paradigm."
                    },
                    {
                        "title": "FireFly: A Synthetic Dataset for Ember Detection in Wildfire",
                        "abstract": "This paper presents \"FireFly\", a synthetic dataset for ember detection created using Unreal Engine 4 (UE4), designed to overcome the current lack of ember-specific training resources. To create the dataset, we present a tool that allows the automated generation of the synthetic labeled dataset with adjustable parameters, enabling data diversity from various environmental conditions, making the dataset both diverse and customizable based on user requirements. We generated a total of 19,273 frames that have been used to evaluate FireFly on four popular object detection models. Further to minimize human intervention, we leveraged a trained model to create a semi-automatic labeling process for real-life ember frames. Moreover, we demonstrated an up to 8.57% improvement in mean Average Precision (mAP) in real-world wildfire scenarios compared to models trained exclusively on a small real dataset."
                    },
                    {
                        "title": "Making Models Shallow Again: Jointly Learning to Reduce Non-Linearity and Depth for Latency-Efficient Private Inference",
                        "abstract": "Large number of ReLU and MAC operations of Deep neural networks make them ill-suited for latency and compute-efficient private inference. In this paper, we present a model optimization method that allows a model to learn to be shallow. In particular, we leverage the ReLU sensitivity of a convolutional block to remove a ReLU layer and merge its succeeding and preceding convolution layers to a shallow block. Unlike existing ReLU reduction methods, our joint reduction method can yield models with improved reduction of both ReLUs and linear operations by up to 1.73\u00d7 and 1.47\u00d7, respectively, evaluated with ResNet18 on CIFAR-100 without any significant accuracy-drop."
                    },
                    {
                        "title": "Sensi-BERT: Towards Sensitivity Driven Fine-Tuning for Parameter-Efficient BERT",
                        "abstract": "Large pre-trained language models have recently gained significant traction due to their improved performance on various down-stream tasks like text classification and question answering, requiring only few epochs of fine-tuning. However, their large model sizes often prohibit their applications on resource-constrained edge devices. Existing solutions of yielding parameter-efficient BERT models largely rely on compute-exhaustive training and fine-tuning. Moreover, they often rely on additional compute heavy models to mitigate the performance gap. In this paper, we present Sensi-BERT, a sensitivity driven efficient fine-tuning of BERT models that can take an off-the-shelf pre-trained BERT model and yield highly parameter-efficient models for downstream tasks. In particular, we perform sensitivity analysis to rank each individual parameter tensor, that then is used to trim them accordingly during fine-tuning for a given parameter or FLOPs budget. Our experiments show the efficacy of Sensi-BERT across different downstream tasks including MNLI, QQP, QNLI, SST-2 and SQuAD, showing better performance at similar or smaller parameter budget compared to various alternatives."
                    },
                    {
                        "title": "ViTA: A Vision Transformer Inference Accelerator for Edge Applications",
                        "abstract": "Vision Transformer models, such as ViT, Swin Transformer, and Transformer-in-Transformer, have recently gained significant traction in computer vision tasks due to their ability to capture the global relation between features which leads to superior performance. However, they are compute-heavy and difficult to deploy in resource-constrained edge devices. Existing hardware accelerators, including those for the closely-related BERT transformer models, do not target highly resource-constrained environments. In this paper, we address this gap and propose ViTA - a configurable hardware accelerator for inference of vision transformer models, targeting resource-constrained edge computing devices and avoiding repeated off-chip memory accesses. We employ a head-level pipeline and inter-layer MLP optimizations, and can support several commonly used vision transformer models with changes solely in our control logic. We achieve nearly 90% hardware utilization efficiency on most vision transformer models, report a power of 0.88W when synthesised with a clock of 150 MHz, and get reasonable frame rates - all of which makes ViTA suitable for edge applications."
                    },
                    {
                        "title": "GenQ: Quantization in Low Data Regimes with Generative Synthetic Data",
                        "abstract": "In the realm of deep neural network deployment, low-bit quantization presents a promising avenue for enhancing computational efficiency. However, it often hinges on the availability of training data to mitigate quantization errors, a significant challenge when data availability is scarce or restricted due to privacy or copyright concerns. Addressing this, we introduce GenQ, a novel approach employing an advanced Generative AI model to generate photorealistic, high-resolution synthetic data, overcoming the limitations of traditional methods that struggle to accurately mimic complex objects in extensive datasets like ImageNet. Our methodology is underscored by two robust filtering mechanisms designed to ensure the synthetic data closely aligns with the intrinsic characteristics of the actual training data. In case of limited data availability, the actual data is used to guide the synthetic data generation process, enhancing fidelity through the inversion of learnable token embeddings. Through rigorous experimentation, GenQ establishes new benchmarks in data-free and data-scarce quantization, significantly outperforming existing methods in accuracy and efficiency, thereby setting a new standard for quantization in low data regimes. Code is released at \\url{https://github.com/Intelligent-Computing-Lab-Yale/GenQ}."
                    },
                    {
                        "title": "Pruning Small Pre-Trained Weights Irreversibly and Monotonically Impairs\"Difficult\"Downstream Tasks in LLMs",
                        "abstract": "We present Junk DNA Hypothesis by adopting a novel task-centric angle for the pre-trained weights of large language models (LLMs). It has been believed that weights in LLMs contain significant redundancy, leading to the conception that a considerable chunk of the parameters can be removed by pruning without compromising performance. Contrary to this belief, this paper presents a counter-argument: small-magnitude weights of pre-trained model weights encode vital knowledge essential for tackling difficult downstream tasks - manifested as the monotonic relationship between the performance drop of downstream tasks across the difficulty spectrum, as we prune more pre-trained weights by magnitude. Moreover, we reveal that these seemingly inconsequential weights can result in irreparable loss of knowledge and performance degradation in difficult tasks, even when downstream continual training is allowed. Interestingly, our evaluations show that the other popular compression, namely quantization, fails to exhibit similar monotonic effect and does not as convincingly disentangle this task-difficulty information. To study formally, we introduce several quantifiable metrics to gauge the downstream task difficulty: (1) within the same task category, and (2) across different task categories. Our extensive experiments substantiate the Junk DNA Hypothesis across a diverse range of model sizes, tasks, datasets, and even pruning methods. Codes are available at: https://github.com/VITA-Group/Junk_DNA_Hypothesis.git."
                    },
                    {
                        "title": "RNA-ViT: Reduced-Dimension Approximate Normalized Attention Vision Transformers for Latency Efficient Private Inference",
                        "abstract": "The concern over data and model privacy in machine learning inference as a service (MLaaS) has led to the development of private inference (PI) techniques. However, existing PI frameworks, especially those designed for large models such as vision transformers (ViT), suffer from high computational and communication overheads caused by the expensive multi-party computation (MPC) protocols. The encrypted attention module that involves the softmax operation contributes significantly to this overhead. In this work, we present a family of models dubbed RNA-ViT, that leverage a novel attention module called reduced-dimension approximate normalized attention and a latency efficient GeLU-alternative layer. In particular, RNA-ViT uses two novel techniques to improve PI efficiency in ViTs: a reduced-dimension normalized attention (RNA) architecture and a high order polynomial (HOP) softmax approximation for latency efficient normalization. We also propose a novel metric, accuracy-to-latency ratio (A2L), to evaluate modules in terms of their accuracy and PI latency. Based on this metric, we perform an analysis to identify a nonlinearity module with improved PI efficiency. Our extensive experiments show that RNA-ViT can achieve average 3.53\u00d7, 3.54\u00d7, 1.66\u00d7 lower PI latency with an average accuracy improvement of 0.93%, 2.04%, and 2.73% compared to the state-of-the-art scheme MPCViT [1], on CIFAR-10, CIFAR-100, and Tiny-ImageNet, respectively."
                    },
                    {
                        "title": "NeRFool: Uncovering the Vulnerability of Generalizable Neural Radiance Fields against Adversarial Perturbations",
                        "abstract": "Generalizable Neural Radiance Fields (GNeRF) are one of the most promising real-world solutions for novel view synthesis, thanks to their cross-scene generalization capability and thus the possibility of instant rendering on new scenes. While adversarial robustness is essential for real-world applications, little study has been devoted to understanding its implication on GNeRF. We hypothesize that because GNeRF is implemented by conditioning on the source views from new scenes, which are often acquired from the Internet or third-party providers, there are potential new security concerns regarding its real-world applications. Meanwhile, existing understanding and solutions for neural networks' adversarial robustness may not be applicable to GNeRF, due to its 3D nature and uniquely diverse operations. To this end, we present NeRFool, which to the best of our knowledge is the first work that sets out to understand the adversarial robustness of GNeRF. Specifically, NeRFool unveils the vulnerability patterns and important insights regarding GNeRF's adversarial robustness. Built upon the above insights gained from NeRFool, we further develop NeRFool+, which integrates two techniques capable of effectively attacking GNeRF across a wide range of target views, and provide guidelines for defending against our proposed attacks. We believe that our NeRFool/NeRFool+ lays the initial foundation for future innovations in developing robust real-world GNeRF solutions. Our codes are available at: https://github.com/GATECH-EIC/NeRFool."
                    },
                    {
                        "title": "Recent Advances in Scalable Energy-Efficient and Trustworthy Spiking Neural Networks: from Algorithms to Technology",
                        "abstract": "Neuromorphic computing and, in particular, spiking neural networks (SNNs) have become an attractive alternative to deep neural networks for a broad range of signal processing applications, processing static and/or temporal inputs from different sensory modalities, including audio and vision sensors. In this paper, we start with a description of recent advances in algorithmic and optimization innovations to efficiently train and scale low-latency, and energy-efficient spiking neural networks (SNNs) for complex machine learning applications. We then discuss the recent efforts in algorithm-architecture co-design that explores the inherent trade-offs between achieving high energy-efficiency and low latency while still providing high accuracy and trustworthiness. We then describe the underlying hardware that has been developed to leverage such algorithmic innovations in an efficient way. In particular, we describe a hybrid method to integrate significant portions of the model\u2019s computation within both memory components as well as the sensor itself. Finally, we discuss the potential path forward for research in building deployable SNN systems identifying key challenges in the algorithm-hardware-application co-design space with an emphasis on trustworthiness."
                    },
                    {
                        "title": "InstaTune: Instantaneous Neural Architecture Search During Fine-Tuning",
                        "abstract": "One-Shot Neural Architecture Search (NAS) algorithms often rely on training a hardware agnostic super-network for a domain specific task. Optimal sub-networks are then extracted from the trained super-network for different hardware platforms. However, training super-networks from scratch can be extremely time consuming and compute intensive especially for large models that rely on a two-stage training process of pre-training and fine-tuning. State of the art pre-trained models are available for a wide range of tasks, but their large sizes significantly limits their applicability on various hardware platforms. We propose InstaTune, a method that leverages off-the-shelf pre-trained weights for large models and generates a super-network during the fine-tuning stage. InstaTune has multiple benefits. Firstly, since the process happens during fine-tuning, it minimizes the overall time and compute resources required for NAS. Secondly, the sub-networks extracted are optimized for the target task, unlike prior work that optimizes on the pre-training objective. Finally, InstaTune is easy to \"plug and play\" in existing frameworks. By using multi-objective evolutionary search algorithms along with lightly trained predictors, we find Pareto-optimal sub-networks that outperform their respective baselines across different performance objectives such as accuracy and MACs. Specifically, we demonstrate that our approach performs well across both unimodal (ViT and BERT) and multi-modal (BEiT-3) transformer based architectures."
                    }
                ]
            },
            "59395490-df31-4c6b-8ef5-37c2e2eb02dc": {
                "pk": "59395490-df31-4c6b-8ef5-37c2e2eb02dc",
                "name": "Eric Xing",
                "collaborators": [
                    "W. Neiswanger",
                    "Hongyi Wang",
                    "Sang Keun Choe",
                    "Hwijeen Ahn",
                    "Emma Strubell",
                    "Zhiqiang Shen",
                    "Tianhua Tao",
                    "Zhengzhong Liu",
                    "Bowen Tan",
                    "Juhan Bae",
                    "Kewen Zhao",
                    "Minsoo Kang",
                    "Youngseog Chung",
                    "Adithya Pratapa",
                    "Teruko Mitamura",
                    "Jeff Schneider",
                    "Eduard Hovy",
                    "Roger Grosse",
                    "Lichao Sun",
                    "Yue Huang",
                    "Haoran Wang",
                    "Siyuan Wu",
                    "Qihui Zhang",
                    "Chujie Gao",
                    "Yixin Huang",
                    "Wenhan Lyu",
                    "Yixuan Zhang",
                    "Xiner Li",
                    "Zheng Liu",
                    "Yixin Liu",
                    "Yijue Wang",
                    "Zhikun Zhang",
                    "B. Kailkhura",
                    "Caiming Xiong",
                    "Chaowei Xiao",
                    "Chun-Yan Li",
                    "Furong Huang",
                    "Haodong Liu",
                    "Heng Ji",
                    "Huan Zhang",
                    "Huaxiu Yao",
                    "M. Kellis",
                    "M. Zitnik",
                    "Meng Jiang",
                    "Mohit Bansal",
                    "James Zou",
                    "Jian Pei",
                    "Jian Liu",
                    "Jianfeng Gao",
                    "Jiawei Han",
                    "Jieyu Zhao",
                    "Jiliang Tang",
                    "Jindong Wang",
                    "John Mitchell",
                    "Kai Shu",
                    "Kaidi Xu",
                    "Kai-Wei Chang",
                    "Lifang He",
                    "Lifu Huang",
                    "M. Backes",
                    "Neil Zhenqiang Gong",
                    "Philip S. Yu",
                    "Pin-Yu Chen",
                    "Quanquan Gu",
                    "Ran Xu",
                    "Rex Ying",
                    "Shuiwang Ji",
                    "S. Jana",
                    "Tian-Xiang Chen",
                    "Tianming Liu",
                    "Tianying Zhou",
                    "William Wang",
                    "Xiang Li",
                    "Xiang-Yu Zhang",
                    "Xiao Wang",
                    "Xingyao Xie",
                    "Xun Chen",
                    "Xuyu Wang",
                    "Yan Liu",
                    "Yanfang Ye",
                    "Yinzhi Cao",
                    "Yue Zhao",
                    "Liqun Ma",
                    "Joel Hestness",
                    "Natalia Vassilieva",
                    "Daria Soboleva",
                    "Aurick Qiao",
                    "Junbo Li",
                    "Yuqi Wang",
                    "Suqi Sun",
                    "Omkar Pangarkar",
                    "Richard Fan",
                    "Yi Gu",
                    "Victor Miller",
                    "Yonghao Zhuang",
                    "Guowei He",
                    "Haonan Li",
                    "Fajri Koto",
                    "Liping Tang",
                    "Nikhil Ranjan"
                ],
                "domain": [
                    "Large Language Models",
                    "Data Valuation",
                    "Meta Learning",
                    "Trustworthiness"
                ],
                "publications": [
                    {
                        "title": "What is Your Data Worth to GPT? LLM-Scale Data Valuation with Influence Functions",
                        "abstract": "Large language models (LLMs) are trained on a vast amount of human-written data, but data providers often remain uncredited. In response to this issue, data valuation (or data attribution), which quantifies the contribution or value of each data to the model output, has been discussed as a potential solution. Nevertheless, applying existing data valuation methods to recent LLMs and their vast training datasets has been largely limited by prohibitive compute and memory costs. In this work, we focus on influence functions, a popular gradient-based data valuation method, and significantly improve its scalability with an efficient gradient projection strategy called LoGra that leverages the gradient structure in backpropagation. We then provide a theoretical motivation of gradient projection approaches to influence functions to promote trust in the data valuation process. Lastly, we lower the barrier to implementing data valuation systems by introducing LogIX, a software package that can transform existing training code into data valuation code with minimal effort. In our data valuation experiments, LoGra achieves competitive accuracy against more expensive baselines while showing up to 6,500x improvement in throughput and 5x reduction in GPU memory usage when applied to Llama3-8B-Instruct and the 1B-token dataset."
                    },
                    {
                        "title": "TrustLLM: Trustworthiness in Large Language Models",
                        "abstract": "Large language models (LLMs), exemplified by ChatGPT, have gained considerable attention for their excellent natural language processing capabilities. Nonetheless, these LLMs present many challenges, particularly in the realm of trustworthiness. Therefore, ensuring the trustworthiness of LLMs emerges as an important topic. This paper introduces TrustLLM, a comprehensive study of trustworthiness in LLMs, including principles for different dimensions of trustworthiness, established benchmark, evaluation, and analysis of trustworthiness for mainstream LLMs, and discussion of open challenges and future directions. Specifically, we first propose a set of principles for trustworthy LLMs that span eight different dimensions. Based on these principles, we further establish a benchmark across six dimensions including truthfulness, safety, fairness, robustness, privacy, and machine ethics. We then present a study evaluating 16 mainstream LLMs in TrustLLM, consisting of over 30 datasets. Our findings firstly show that in general trustworthiness and utility (i.e., functional effectiveness) are positively related. Secondly, our observations reveal that proprietary LLMs generally outperform most open-source counterparts in terms of trustworthiness, raising concerns about the potential risks of widely accessible open-source LLMs. However, a few open-source LLMs come very close to proprietary ones. Thirdly, it is important to note that some LLMs may be overly calibrated towards exhibiting trustworthiness, to the extent that they compromise their utility by mistakenly treating benign prompts as harmful and consequently not responding. Finally, we emphasize the importance of ensuring transparency not only in the models themselves but also in the technologies that underpin trustworthiness. Knowing the specific trustworthy technologies that have been employed is crucial for analyzing their effectiveness."
                    },
                    {
                        "title": "SlimPajama-DC: Understanding Data Combinations for LLM Training",
                        "abstract": "This paper aims to understand the impacts of various data combinations (e.g., web text, Wikipedia, GitHub, books) on the pretraining of large language models using SlimPajama. SlimPajama is a rigorously deduplicated, multi-source dataset, which has been refined and further deduplicated to 627B tokens from the extensive 1.2T token RedPajama dataset contributed by Together. We have termed our research as SlimPajama-DC, an empirical analysis designed to uncover fundamental characteristics and best practices associated with employing SlimPajama in the training of large language models. During our research with SlimPajama, two pivotal observations emerged: (1) Global deduplication vs. local deduplication. We analyze and discuss how global (across different sources of datasets) and local (within the single source of dataset) deduplications affect the performance of trained models. (2) Proportions of highly-deduplicated multi-source datasets in the combination. To study this, we construct six configurations on SlimPajama dataset and train individual ones using 1.3B Cerebras-GPT model with Alibi and SwiGLU. Our best configuration outperforms the 1.3B model trained on RedPajama using the same number of training tokens by a significant margin. All our 1.3B models are trained on Cerebras 16$\\times$ CS-2 cluster with a total of 80 PFLOP/s in bf16 mixed precision. We further extend our discoveries (such as increasing data diversity is crucial after global deduplication) on a 7B model with large batch-size training. Our SlimPajama-DC models are available at: https://huggingface.co/MBZUAI-LLM/SlimPajama-DC and the separate SlimPajama-DC datasets are available at: https://huggingface.co/datasets/MBZUAI-LLM/SlimPajama-627B-DC."
                    },
                    {
                        "title": "LLM360: Towards Fully Transparent Open-Source LLMs",
                        "abstract": "The recent surge in open-source Large Language Models (LLMs), such as LLaMA, Falcon, and Mistral, provides diverse options for AI practitioners and researchers. However, most LLMs have only released partial artifacts, such as the final model weights or inference code, and technical reports increasingly limit their scope to high-level design choices and surface statistics. These choices hinder progress in the field by degrading transparency into the training of LLMs and forcing teams to rediscover many details in the training process. We present LLM360, an initiative to fully open-source LLMs, which advocates for all training code and data, model checkpoints, and intermediate results to be made available to the community. The goal of LLM360 is to support open and collaborative AI research by making the end-to-end LLM training process transparent and reproducible by everyone. As a first step of LLM360, we release two 7B parameter LLMs pre-trained from scratch, Amber and CrystalCoder, including their training code, data, intermediate checkpoints, and analyses (at https://www.llm360.ai). We are committed to continually pushing the boundaries of LLMs through this open-source effort. More large-scale and stronger models are underway and will be released in the future."
                    },
                    {
                        "title": "Making Scalable Meta Learning Practical",
                        "abstract": "Despite its flexibility to learn diverse inductive biases in machine learning programs, meta learning (i.e., learning to learn) has long been recognized to suffer from poor scalability due to its tremendous compute/memory costs, training instability, and a lack of efficient distributed training support. In this work, we focus on making scalable meta learning practical by introducing SAMA, which combines advances in both implicit differentiation algorithms and systems. Specifically, SAMA is designed to flexibly support a broad range of adaptive optimizers in the base level of meta learning programs, while reducing computational burden by avoiding explicit computation of second-order gradient information, and exploiting efficient distributed training techniques implemented for first-order gradients. Evaluated on multiple large-scale meta learning benchmarks, SAMA showcases up to 1.7/4.8x increase in throughput and 2.0/3.8x decrease in memory consumption respectively on single-/multi-GPU setups compared to other baseline meta learning algorithms. Furthermore, we show that SAMA-based data optimization leads to consistent improvements in text classification accuracy with BERT and RoBERTa large language models, and achieves state-of-the-art results in both small- and large-scale data pruning on image classification tasks, demonstrating the practical applicability of scalable meta learning across language and vision domains."
                    }
                ]
            },
            "d9b20a8f-f83c-4f33-9471-725d55b1a4d0": {
                "pk": "d9b20a8f-f83c-4f33-9471-725d55b1a4d0",
                "name": "Mikhail Yurochkin",
                "collaborators": [
                    "Yuekai Sun",
                    "Felipe Maia Polo",
                    "K. Greenewald",
                    "Leshem Choshen",
                    "Justin Solomon",
                    "Subha Maity",
                    "Rickard Br\u00fcel Gabrielsson",
                    "Mayank Agarwal",
                    "Jiacheng Zhu",
                    "Leonid Karlinsky",
                    "Lucas Weber",
                    "Lilian Ngweta",
                    "Alex Gittens",
                    "Moulinath Banerjee",
                    "Ronald Xu",
                    "M'irian Silva",
                    "Onkar Bhardwaj",
                    "Igor Melnyk",
                    "Youssef Mroueh",
                    "Brian M. Belgodere",
                    "Mattia Rigotti",
                    "Apoorva Nitsure",
                    "Jir\u00ed Navr\u00e1til",
                    "Jerret Ross",
                    "Tal Shnitzer",
                    "Kimia Nadjahi",
                    "Haitz S'aez de Oc'ariz Borde",
                    "Marzyeh Ghassemi",
                    "Nimrod Shabtay",
                    "Sivan Doveh",
                    "Wei Lin",
                    "M. Mirza",
                    "Leshem Chosen",
                    "Assaf Arbelle",
                    "Raja Giryes",
                    "Gongjun Xu",
                    "Felix Petersen",
                    "Aashwin Mishra",
                    "Hilde Kuehne",
                    "Christian Borgelt",
                    "Oliver Deussen",
                    "Seamus Somerstep",
                    "Ya'acov Ritov",
                    "Allysson Flavio Melo de Oliveira",
                    "Michael Feffer",
                    "Anthony Ou",
                    "Kate Soule",
                    "Neil Thompson",
                    "Songkai Xue",
                    "Dustin Klebe",
                    "Zahra Ashktorab",
                    "Benjamin Hoover",
                    "Casey Dugan",
                    "Werner Geyer",
                    "Han Yang",
                    "J. Solomon"
                ],
                "domain": [
                    "Machine Learning",
                    "Large Language Models",
                    "Fine-tuning",
                    "Algorithmic Fairness"
                ],
                "publications": [
                    {
                        "title": "Asymmetry in Low-Rank Adapters of Foundation Models",
                        "abstract": "Parameter-efficient fine-tuning optimizes large, pre-trained foundation models by updating a subset of parameters; in this class, Low-Rank Adaptation (LoRA) is particularly effective. Inspired by an effort to investigate the different roles of LoRA matrices during fine-tuning, this paper characterizes and leverages unexpected asymmetry in the importance of low-rank adapter matrices. Specifically, when updating the parameter matrices of a neural network by adding a product $BA$, we observe that the $B$ and $A$ matrices have distinct functions: $A$ extracts features from the input, while $B$ uses these features to create the desired output. Based on this observation, we demonstrate that fine-tuning $B$ is inherently more effective than fine-tuning $A$, and that a random untrained $A$ should perform nearly as well as a fine-tuned one. Using an information-theoretic lens, we also bound the generalization of low-rank adapters, showing that the parameter savings of exclusively training $B$ improves the bound. We support our conclusions with experiments on RoBERTa, BART-Large, LLaMA-2, and ViTs."
                    },
                    {
                        "title": "LiveXiv -- A Multi-Modal Live Benchmark Based on Arxiv Papers Content",
                        "abstract": "The large-scale training of multi-modal models on data scraped from the web has shown outstanding utility in infusing these models with the required world knowledge to perform effectively on multiple downstream tasks. However, one downside of scraping data from the web can be the potential sacrifice of the benchmarks on which the abilities of these models are often evaluated. To safeguard against test data contamination and to truly test the abilities of these foundation models we propose LiveXiv: A scalable evolving live benchmark based on scientific ArXiv papers. LiveXiv accesses domain-specific manuscripts at any given timestamp and proposes to automatically generate visual question-answer pairs (VQA). This is done without any human-in-the-loop, using the multi-modal content in the manuscripts, like graphs, charts, and tables. Moreover, we introduce an efficient evaluation approach that estimates the performance of all models on the evolving benchmark using evaluations of only a subset of models. This significantly reduces the overall evaluation cost. We benchmark multiple open and proprietary Large Multi-modal Models (LMMs) on the first version of our benchmark, showing its challenging nature and exposing the models true abilities, avoiding contamination. Lastly, in our commitment to high quality, we have collected and evaluated a manually verified subset. By comparing its overall results to our automatic annotations, we have found that the performance variance is indeed minimal (<2.5%). Our dataset is available online on HuggingFace, and our code will be available here."
                    },
                    {
                        "title": "tinyBenchmarks: evaluating LLMs with fewer examples",
                        "abstract": "The versatility of large language models (LLMs) led to the creation of diverse benchmarks that thoroughly test a variety of language models' abilities. These benchmarks consist of tens of thousands of examples making evaluation of LLMs very expensive. In this paper, we investigate strategies to reduce the number of evaluations needed to assess the performance of an LLM on several key benchmarks. For example, we show that to accurately estimate the performance of an LLM on MMLU, a popular multiple-choice QA benchmark consisting of 14K examples, it is sufficient to evaluate this LLM on 100 curated examples. We release evaluation tools and tiny versions of popular benchmarks: Open LLM Leaderboard, MMLU, HELM, and AlpacaEval 2.0. Our empirical analysis demonstrates that these tools and tiny benchmarks are sufficient to reliably and efficiently reproduce the original evaluation results."
                    },
                    {
                        "title": "Aligners: Decoupling LLMs and Alignment",
                        "abstract": "Large Language Models (LLMs) need to be aligned with human expectations to ensure their safety and utility in most applications. Alignment is challenging, costly, and needs to be repeated for every LLM and alignment criterion. We propose to decouple LLMs and alignment by training aligner models that can be used to align any LLM for a given criteria on an as-needed basis, thus also reducing the potential negative impacts of alignment on performance. Our recipe for training the aligner models solely relies on synthetic data generated with a (prompted) LLM and can be easily adjusted for a variety of alignment criteria. We use the same synthetic data to train inspectors, binary miss-alignment classification models to guide a\"squad\"of multiple aligners. Our empirical results demonstrate consistent improvements when applying aligner squad to various LLMs, including chat-aligned models, across several instruction-following and red-teaming datasets."
                    },
                    {
                        "title": "Uncertainty Quantification via Stable Distribution Propagation",
                        "abstract": "We propose a new approach for propagating stable probability distributions through neural networks. Our method is based on local linearization, which we show to be an optimal approximation in terms of total variation distance for the ReLU non-linearity. This allows propagating Gaussian and Cauchy input uncertainties through neural networks to quantify their output uncertainties. To demonstrate the utility of propagating distributions, we apply the proposed method to predicting calibrated confidence intervals and selective prediction on out-of-distribution data. The results demonstrate a broad applicability of propagating distributions and show the advantages of our method over other approaches such as moment matching."
                    },
                    {
                        "title": "A statistical framework for weak-to-strong generalization",
                        "abstract": "Modern large language model (LLM) alignment techniques rely on human feedback, but it is unclear whether the techniques fundamentally limit the capabilities of aligned LLMs. In particular, it is unclear whether it is possible to align (stronger) LLMs with superhuman capabilities with (weaker) human feedback without degrading their capabilities. This is an instance of the weak-to-strong generalization problem: using weaker (less capable) feedback to train a stronger (more capable) model. We prove that weak-to-strong generalization is possible by eliciting latent knowledge from pre-trained LLMs. In particular, we cast the weak-to-strong generalization problem as a transfer learning problem in which we wish to transfer a latent concept from a weak model to a strong pre-trained model. We prove that a naive fine-tuning approach suffers from fundamental limitations, but an alternative refinement-based approach suggested by the problem structure provably overcomes the limitations of fine-tuning. Finally, we demonstrate the practical applicability of the refinement approach with three LLM alignment tasks."
                    },
                    {
                        "title": "Efficient multi-prompt evaluation of LLMs",
                        "abstract": "Most popular benchmarks for comparing LLMs rely on a limited set of prompt templates, which may not fully capture the LLMs' abilities and can affect the reproducibility of results on leaderboards. Many recent works empirically verify prompt sensitivity and advocate for changes in LLM evaluation. In this paper, we consider the problem of estimating the performance distribution across many prompt variants instead of finding a single prompt to evaluate with. We introduce PromptEval, a method for estimating performance across a large set of prompts borrowing strength across prompts and examples to produce accurate estimates under practical evaluation budgets. The resulting distribution can be used to obtain performance quantiles to construct various robust performance metrics (e.g., top 95% quantile or median). We prove that PromptEval consistently estimates the performance distribution and demonstrate its efficacy empirically on three prominent LLM benchmarks: MMLU, BIG-bench Hard, and LMentry; for example, PromptEval can accurately estimate performance quantiles across 100 prompt templates on MMLU with a budget equivalent to two single-prompt evaluations. Moreover, we show how PromptEval can be useful in LLM-as-a-judge and best prompt identification applications."
                    },
                    {
                        "title": "Prompt Exploration with Prompt Regression",
                        "abstract": "In the advent of democratized usage of large language models (LLMs), there is a growing desire to systematize LLM prompt creation and selection processes beyond iterative trial-and-error. Prior works majorly focus on searching the space of prompts without accounting for relations between prompt variations. Here we propose a framework, Prompt Exploration with Prompt Regression (PEPR), to predict the effect of prompt combinations given results for individual prompt elements as well as a simple method to select an effective prompt for a given use-case. We evaluate our approach with open-source LLMs of different sizes on several different tasks."
                    },
                    {
                        "title": "Compress then Serve: Serving Thousands of LoRA Adapters with Little Overhead",
                        "abstract": "Fine-tuning large language models (LLMs) with low-rank adaptations (LoRAs) has become common practice, often yielding numerous copies of the same LLM differing only in their LoRA updates. This paradigm presents challenges for systems that serve real-time responses to queries that each involve a different LoRA. Prior works optimize the design of such systems but still require continuous loading and offloading of LoRAs, as it is infeasible to store thousands of LoRAs in GPU memory. To mitigate this issue, we investigate the efficacy of model compression when serving LoRAs. We propose a method for joint compression of LoRAs into a shared basis paired with LoRA-specific scaling matrices. We extend our algorithm to learn clusters of LoRAs that are more amenable to joint compression, allowing it to scale gracefully to large LoRA collections. Our experiments with up to 500 LoRAs demonstrate that compressed LoRAs preserve performance while offering major throughput gains in realistic serving scenarios with over a thousand LoRAs, maintaining 80% of the throughput of serving a single LoRA."
                    },
                    {
                        "title": "Distributional Preference Alignment of LLMs via Optimal Transport",
                        "abstract": "Current LLM alignment techniques use pairwise human preferences at a sample level, and as such, they do not imply an alignment on the distributional level. We propose in this paper Alignment via Optimal Transport (AOT), a novel method for distributional preference alignment of LLMs. AOT aligns LLMs on unpaired preference data by making the reward distribution of the positive samples stochastically dominant in the first order on the distribution of negative samples. We introduce a convex relaxation of this first-order stochastic dominance and cast it as an optimal transport problem with a smooth and convex cost. Thanks to the one-dimensional nature of the resulting optimal transport problem and the convexity of the cost, it has a closed-form solution via sorting on empirical measures. We fine-tune LLMs with this AOT objective, which enables alignment by penalizing the violation of the stochastic dominance of the reward distribution of the positive samples on the reward distribution of the negative samples. We analyze the sample complexity of AOT by considering the dual of the OT problem and show that it converges at the parametric rate. Empirically, we show on a diverse set of alignment datasets and LLMs that AOT leads to state-of-the-art models in the 7B family of models when evaluated with Open LLM Benchmarks and AlpacaEval."
                    },
                    {
                        "title": "Large Language Model Routing with Benchmark Datasets",
                        "abstract": "There is a rapidly growing number of open-source Large Language Models (LLMs) and benchmark datasets to compare them. While some models dominate these benchmarks, no single model typically achieves the best accuracy in all tasks and use cases. In this work, we address the challenge of selecting the best LLM out of a collection of models for new tasks. We propose a new formulation for the problem, in which benchmark datasets are repurposed to learn a\"router\"model for this LLM selection, and we show that this problem can be reduced to a collection of binary classification tasks. We demonstrate the utility and limitations of learning model routers from various benchmark datasets, where we consistently improve performance upon using any single model for all tasks."
                    },
                    {
                        "title": "Calibrated Data-Dependent Constraints with Exact Satisfaction Guarantees",
                        "abstract": "We consider the task of training machine learning models with data-dependent constraints. Such constraints often arise as empirical versions of expected value constraints that enforce fairness or stability goals. We reformulate data-dependent constraints so that they are calibrated: enforcing the reformulated constraints guarantees that their expected value counterparts are satisfied with a user-prescribed probability. The resulting optimization problem is amendable to standard stochastic optimization algorithms, and we demonstrate the efficacy of our method on a fairness-sensitive classification task where we wish to guarantee the classifier's fairness (at test time)."
                    },
                    {
                        "title": "GeRA: Label-Efficient Geometrically Regularized Alignment",
                        "abstract": "Pretrained unimodal encoders incorporate rich semantic information into embedding space structures. To be similarly informative, multi-modal encoders typically require massive amounts of paired data for alignment and training. We introduce a semi-supervised Geometrically Regularized Alignment (GeRA) method to align the embedding spaces of pretrained unimodal encoders in a label-efficient way. Our method leverages the manifold geometry of unpaired (unlabeled) data to improve alignment performance. To prevent distortions to local geometry during the alignment process, potentially disrupting semantic neighborhood structures and causing misalignment of unobserved pairs, we introduce a geometric loss term. This term is built upon a diffusion operator that captures the local manifold geometry of the unimodal pretrained encoders. GeRA is modality-agnostic and thus can be used to align pretrained encoders from any data modalities. We provide empirical evidence to the effectiveness of our method in the domains of speech-text and image-text alignment. Our experiments demonstrate significant improvement in alignment quality compared to a variaty of leading baselines, especially with a small amount of paired data, using our proposed geometric regularization."
                    },
                    {
                        "title": "Weak Supervision Performance Evaluation via Partial Identification",
                        "abstract": "Programmatic Weak Supervision (PWS) enables supervised model training without direct access to ground truth labels, utilizing weak labels from heuristics, crowdsourcing, or pre-trained models. However, the absence of ground truth complicates model evaluation, as traditional metrics such as accuracy, precision, and recall cannot be directly calculated. In this work, we present a novel method to address this challenge by framing model evaluation as a partial identification problem and estimating performance bounds using Fr\\'echet bounds. Our approach derives reliable bounds on key metrics without requiring labeled data, overcoming core limitations in current weak supervision evaluation techniques. Through scalable convex optimization, we obtain accurate and computationally efficient bounds for metrics including accuracy, precision, recall, and F1-score, even in high-dimensional settings. This framework offers a robust approach to assessing model quality without ground truth labels, enhancing the practicality of weakly supervised learning for real-world applications."
                    },
                    {
                        "title": "An Investigation of Representation and Allocation Harms in Contrastive Learning",
                        "abstract": "The effect of underrepresentation on the performance of minority groups is known to be a serious problem in supervised learning settings; however, it has been underexplored so far in the context of self-supervised learning (SSL). In this paper, we demonstrate that contrastive learning (CL), a popular variant of SSL, tends to collapse representations of minority groups with certain majority groups. We refer to this phenomenon as representation harm and demonstrate it on image and text datasets using the corresponding popular CL methods. Furthermore, our causal mediation analysis of allocation harm on a downstream classification task reveals that representation harm is partly responsible for it, thus emphasizing the importance of studying and mitigating representation harm. Finally, we provide a theoretical explanation for representation harm using a stochastic block model that leads to a representational neural collapse in a contrastive learning setting."
                    },
                    {
                        "title": "Simple Disentanglement of Style and Content in Visual Representations",
                        "abstract": "Learning visual representations with interpretable features, i.e., disentangled representations, remains a challenging problem. Existing methods demonstrate some success but are hard to apply to large-scale vision datasets like ImageNet. In this work, we propose a simple post-processing framework to disentangle content and style in learned representations from pre-trained vision models. We model the pre-trained features probabilistically as linearly entangled combinations of the latent content and style factors and develop a simple disentanglement algorithm based on the probabilistic model. We show that the method provably disentangles content and style features and verify its efficacy empirically. Our post-processed features yield significant domain generalization performance improvements when the distribution shift occurs due to style changes or style-related spurious correlations."
                    },
                    {
                        "title": "Risk Aware Benchmarking of Large Language Models",
                        "abstract": "We propose a distributional framework for benchmarking socio-technical risks of foundation models with quantified statistical significance. Our approach hinges on a new statistical relative testing based on first and second order stochastic dominance of real random variables. We show that the second order statistics in this test are linked to mean-risk models commonly used in econometrics and mathematical finance to balance risk and utility when choosing between alternatives. Using this framework, we formally develop a risk-aware approach for foundation model selection given guardrails quantified by specified metrics. Inspired by portfolio optimization and selection theory in mathematical finance, we define a metrics portfolio for each model as a means to aggregate a collection of metrics, and perform model selection based on the stochastic dominance of these portfolios. The statistical significance of our tests is backed theoretically by an asymptotic analysis via central limit theorems instantiated in practice via a bootstrap variance estimate. We use our framework to compare various large language models regarding risks related to drifting from instructions and outputting toxic content."
                    },
                    {
                        "title": "Fairness Evaluation in Text Classification: Machine Learning Practitioner Perspectives of Individual and Group Fairness",
                        "abstract": "Mitigating algorithmic bias is a critical task in the development and deployment of machine learning models. While several toolkits exist to aid machine learning practitioners in addressing fairness issues, little is known about the strategies practitioners employ to evaluate model fairness and what factors influence their assessment, particularly in the context of text classification. Two common approaches of evaluating the fairness of a model are group fairness and individual fairness. We run a study with Machine Learning practitioners (n=24) to understand the strategies used to evaluate models. Metrics presented to practitioners (group vs. individual fairness) impact which models they consider fair. Participants focused on risks associated with underpredicting / overpredicting and model sensitivity relative to identity token manipulations. We discover fairness assessment strategies involving personal experiences or how users form groups of identity tokens to test model fairness. We provide recommendations for interactive tools for evaluating fairness in text classification."
                    },
                    {
                        "title": "Rewiring with Positional Encodings for Graph Neural Networks",
                        "abstract": "Several recent works use positional encodings to extend the receptive fields of graph neural network (GNN) layers equipped with attention mechanisms. These techniques, however, extend receptive fields to the complete graph, at substantial computational cost and risking a change in the inductive biases of conventional GNNs, or require complex architecture adjustments. As a conservative alternative, we use positional encodings to expand receptive fields to $r$-hop neighborhoods. More specifically, our method augments the input graph with additional nodes/edges and uses positional encodings as node and/or edge features. We thus modify graphs before inputting them to a downstream GNN model, instead of modifying the model itself. This makes our method model-agnostic, i.e., compatible with any of the existing GNN architectures. We also provide examples of positional encodings that are lossless with a one-to-one map between the original and the modified graphs. We demonstrate that extending receptive fields via positional encodings and a virtual fully-connected node significantly improves GNN performance and alleviates over-squashing using small $r$. We obtain improvements on a variety of models and datasets and reach competitive performance using traditional GNNs or graph Transformers."
                    }
                ]
            }
        }
    },
    "2402.07510": {
        "paper_data": {
            "title": "Secret Collusion among Generative AI Agents",
            "url": "http://arxiv.org/abs/2402.07510v2",
            "arxiv_id": "2402.07510",
            "authors": [
                "Sumeet Ramesh Motwani",
                "Mikhail Baranchuk",
                "Martin Strohmeier",
                "Vijay Bolina",
                "Philip H. S. Torr",
                "Lewis Hammond",
                "Christian Schroeder de Witt"
            ],
            "abstract": "Recent capability increases in large language models (LLMs) open up applications in which groups of communicating generative AI agents solve joint tasks. This poses privacy and security challenges concerning the unauthorised sharing of information, or other unwanted forms of agent coordination. Modern steganographic techniques could render such dynamics hard to detect. In this paper, we comprehensively formalise the problem of secret collusion in systems of generative AI agents by drawing on relevant concepts from both AI and security literature. We study incentives for the use of steganography, and propose a variety of mitigation measures. Our investigations result in a model evaluation framework that systematically tests capabilities required for various forms of secret collusion. We provide extensive empirical results across a range of contemporary LLMs. While the steganographic capabilities of current models remain limited, GPT-4 displays a capability jump suggesting the need for continuous monitoring of steganographic frontier model capabilities. We conclude by laying out a comprehensive research program to mitigate future risks of collusion between generative AI models.",
            "introduction": "   1 Introduction  Fueled by significant advances in AI\u00a0(94), several national and supra-national governments are in the process of establishing model evaluation and adversarial testing frameworks for frontier AI models\u00a0(121; 35). Central to such approaches is the security of interacting systems of generative agents\u00a0(98; 103), the proliferation of which is accelerated by recent industry deployments\u00a0(93).   Figure 1: Steganographic capabilities across LLMs. GPT-4 displays significant improvements (Sec. 6).   Decentralised systems of interacting AI agents promise impactful applications to various real-world socio-economic settings\u00a0(131; 105), including digital assistants\u00a0(43), multi-agent robotics and autonomous driving\u00a0(89), trading strategies\u00a0(9; 105), and virtual companies\u00a0(103)\u2014along with their simulation through generative agent-based models \u00a0(40; 125; 129; 132; 70). Many of these applications are classified as \u201chigh-risk\u201d by legislators due to their potential impact on human lives and use of sensitive data\u00a0(35).   Ultimately, concerns have been raised that network interactions between advanced AI agents could render human oversight ineffective, and lead to a worst-case systemic loss of control\u00a0(31). In this work, we consider the question of how groups of generative frontier model agents can be effectively monitored and aligned to prevent collusion. To this end, we bridge the worlds of generative AI and alignment with security, including distributed systems, secure multi-party computation, and steganography.   One risk emerging from groups of generative AI agents is their use of communication channels in ways unintended by their deployer(s). This could constitute privacy violations when sharing data (86), or secret collusion that could help agents coordinate on unwanted behaviour at a wider scale\u00a0(120). Unless these agents\u2019 misalignment can be fully mitigated, a safe deployment of such systems would require continuous monitoring and intervention approaches. The growing capabilities of frontier AI model agents have raised concerns that they could use information-hiding techniques in their communications in order to evade such monitoring\u00a0(90; 109). At the same time, natural language, and, in particular, multi-modal communication channels, commonly have a high information capacity. Altogether, this could allow for significant levels of covert information exchange, the mere existence of which may be worst-case undetectable by black-box monitoring\u00a0(115).   Figure 2:  Secret collusion between GPT-4 models and humans. Consider the setting illustrated in which one AI agent secretly reveals insider proprietary investment information found in its pretraining data to a user. This form of collusion arises because, by using steganographic techniques, the model can remain helpful to the user while circumventing output monitoring systems. We both depict the use of statistically detectable steganography based on a latent codebook that is elicitated in-context and can be readily decoded by humans. We also depict perfectly secure steganography\u00a0(115) that is statistically undetectable and relies on a commonly known secure one-time pad. Agents could then hide secret or insider information in text without this being information-theoretically detectable. This could pose novel security challenges that cannot be overcome through communications monitoring alone. We revisit this scenario as an experimental case study in Section 6.    Contributions.  In this work, we provide a systematic model evaluation framework for the study of secret collusion among large language models and present several novel theoretical results concerning the mitigation of steganography in Sections\u00a05 and \u00a07, respectively. First, we offer a systematic overview of",
            "references": [
                {
                    "title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models",
                    "abstract": "This work identifies 18 foundational challenges in assuring the alignment and safety of large language models (LLMs). These challenges are organized into three different categories: scientific understanding of LLMs, development and deployment methods, and sociotechnical challenges. Based on the identified challenges, we pose $200+$ concrete research questions."
                },
                {
                    "title": "Rethinking Out-of-Distribution Detection for Reinforcement Learning: Advancing Methods for Evaluation and Detection",
                    "abstract": "While reinforcement learning (RL) algorithms have been successfully applied across numerous sequential decision-making problems, their generalization to unforeseen testing environments remains a significant concern. In this paper, we study the problem of out-of-distribution (OOD) detection in RL, which focuses on identifying situations at test time that RL agents have not encountered in their training environments. We first propose a clarification of terminology for OOD detection in RL, which aligns it with the literature from other machine learning domains. We then present new benchmark scenarios for OOD detection, which introduce anomalies with temporal autocorrelation into different components of the agent-environment loop. We argue that such scenarios have been understudied in the current literature, despite their relevance to real-world situations. Confirming our theoretical predictions, our experimental results suggest that state-of-the-art OOD detectors are not able to identify such anomalies. To address this problem, we propose a novel method for OOD detection, which we call DEXTER (Detection via Extraction of Time Series Representations). By treating environment observations as time series data, DEXTER extracts salient time series features, and then leverages an ensemble of isolation forest algorithms to detect anomalies. We find that DEXTER can reliably identify anomalies across benchmark scenarios, exhibiting superior performance compared to both state-of-the-art OOD detectors and high-dimensional changepoint detectors adopted from statistics."
                },
                {
                    "title": "Can Large Language Model Agents Simulate Human Trust Behaviors?",
                    "abstract": "Large Language Model (LLM) agents have been increasingly adopted as simulation tools to model humans in applications such as social science. However, one fundamental question remains: can LLM agents really simulate human behaviors? In this paper, we focus on one of the most critical behaviors in human interactions, trust, and aim to investigate whether or not LLM agents can simulate human trust behaviors. We first find that LLM agents generally exhibit trust behaviors, referred to as agent trust, under the framework of Trust Games, which are widely recognized in behavioral economics. Then, we discover that LLM agents can have high behavioral alignment with humans regarding trust behaviors, particularly for GPT-4, indicating the feasibility to simulate human trust behaviors with LLM agents. In addition, we probe into the biases in agent trust and the differences in agent trust towards agents and humans. We also explore the intrinsic properties of agent trust under conditions including advanced reasoning strategies and external manipulations. We further offer important implications of our discoveries for various scenarios where trust is paramount. Our study provides new insights into the behaviors of LLM agents and the fundamental analogy between LLMs and humans."
                },
                {
                    "title": "Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training",
                    "abstract": "Humans are capable of strategically deceptive behavior: behaving helpfully in most situations, but then behaving very differently in order to pursue alternative objectives when given the opportunity. If an AI system learned such a deceptive strategy, could we detect it and remove it using current state-of-the-art safety training techniques? To study this question, we construct proof-of-concept examples of deceptive behavior in large language models (LLMs). For example, we train models that write secure code when the prompt states that the year is 2023, but insert exploitable code when the stated year is 2024. We find that such backdoor behavior can be made persistent, so that it is not removed by standard safety training techniques, including supervised fine-tuning, reinforcement learning, and adversarial training (eliciting unsafe behavior and then training to remove it). The backdoor behavior is most persistent in the largest models and in models trained to produce chain-of-thought reasoning about deceiving the training process, with the persistence remaining even when the chain-of-thought is distilled away. Furthermore, rather than removing backdoors, we find that adversarial training can teach models to better recognize their backdoor triggers, effectively hiding the unsafe behavior. Our results suggest that, once a model exhibits deceptive behavior, standard techniques could fail to remove such deception and create a false impression of safety."
                },
                {
                    "title": "AI Control: Improving Safety Despite Intentional Subversion",
                    "abstract": "As large language models (LLMs) become more powerful and are deployed more autonomously, it will be increasingly important to prevent them from causing harmful outcomes. Researchers have investigated a variety of safety techniques for this purpose, e.g. using models to review the outputs of other models, or red-teaming techniques to surface subtle failure modes. However, researchers have not evaluated whether such techniques still ensure safety if the model is itself intentionally trying to subvert them. In this paper, we develop and evaluate pipelines of safety techniques (\"protocols\") that are robust to intentional subversion. We investigate a scenario in which we want to solve a sequence of programming problems, using access to a powerful but untrusted model (in our case, GPT-4), access to a less powerful trusted model (in our case, GPT-3.5), and limited access to high-quality trusted labor. We investigate protocols that aim to never submit solutions containing backdoors, which we operationalize here as logical errors that are not caught by test cases. We investigate a range of protocols and test each against strategies that the untrusted model could use to subvert them. One protocol is what we call trusted editing. This protocol first asks GPT-4 to write code, and then asks GPT-3.5 to rate the suspiciousness of that code. If the code is below some suspiciousness threshold, it is submitted. Otherwise, GPT-3.5 edits the solution to remove parts that seem suspicious and then submits the edited code. Another protocol is untrusted monitoring. This protocol asks GPT-4 to write code, and then asks another instance of GPT-4 whether the code is backdoored, using various techniques to prevent the GPT-4 instances from colluding. These protocols improve substantially on simple baselines."
                },
                {
                    "title": "Tell, don't show: Declarative facts influence how LLMs generalize",
                    "abstract": "We examine how large language models (LLMs) generalize from abstract declarative statements in their training data. As an illustration, consider an LLM that is prompted to generate weather reports for London in 2050. One possibility is that the temperatures in the reports match the mean and variance of reports from 2023 (i.e. matching the statistics of pretraining). Another possibility is that the reports predict higher temperatures, by incorporating declarative statements about climate change from scientific papers written in 2023. An example of such a declarative statement is\"global temperatures will increase by $1^{\\circ} \\mathrm{C}$ by 2050\". To test the influence of abstract declarative statements, we construct tasks in which LLMs are finetuned on both declarative and procedural information. We find that declarative statements influence model predictions, even when they conflict with procedural information. In particular, finetuning on a declarative statement $S$ increases the model likelihood for logical consequences of $S$. The effect of declarative statements is consistent across three domains: aligning an AI assistant, predicting weather, and predicting demographic features. Through a series of ablations, we show that the effect of declarative statements cannot be explained by associative learning based on matching keywords. Nevertheless, the effect of declarative statements on model likelihoods is small in absolute terms and increases surprisingly little with model size (i.e. from 330 million to 175 billion parameters). We argue that these results have implications for AI risk (in relation to the\"treacherous turn\") and for fairness."
                },
                {
                    "title": "Privacy Issues in Large Language Models: A Survey",
                    "abstract": "This is the first survey of the active area of AI research that focuses on privacy issues in Large Language Models (LLMs). Specifically, we focus on work that red-teams models to highlight privacy risks, attempts to build privacy into the training or inference process, enables efficient data deletion from trained models to comply with existing privacy regulations, and tries to mitigate copyright issues. Our focus is on summarizing technical research that develops algorithms, proves theorems, and runs empirical evaluations. While there is an extensive body of legal and policy work addressing these challenges from a different angle, that is not the focus of our survey. Nevertheless, these works, along with recent legal developments do inform how these technical problems are formalized, and so we discuss them briefly in Section 1. While we have made our best effort to include all the relevant work, due to the fast moving nature of this research we may have missed some recent work. If we have missed some of your work please contact us, as we will attempt to keep this survey relatively up to date. We are maintaining a repository with the list of papers covered in this survey and any relevant code that was publicly available at https://github.com/safr-ml-lab/survey-llm."
                },
                {
                    "title": "Generative agent-based modeling with actions grounded in physical, social, or digital space using Concordia",
                    "abstract": "Agent-based modeling has been around for decades, and applied widely across the social and natural sciences. The scope of this research method is now poised to grow dramatically as it absorbs the new affordances provided by Large Language Models (LLM)s. Generative Agent-Based Models (GABM) are not just classic Agent-Based Models (ABM)s where the agents talk to one another. Rather, GABMs are constructed using an LLM to apply common sense to situations, act\"reasonably\", recall common semantic knowledge, produce API calls to control digital technologies like apps, and communicate both within the simulation and to researchers viewing it from the outside. Here we present Concordia, a library to facilitate constructing and working with GABMs. Concordia makes it easy to construct language-mediated simulations of physically- or digitally-grounded environments. Concordia agents produce their behavior using a flexible component system which mediates between two fundamental operations: LLM calls and associative memory retrieval. A special agent called the Game Master (GM), which was inspired by tabletop role-playing games, is responsible for simulating the environment where the agents interact. Agents take actions by describing what they want to do in natural language. The GM then translates their actions into appropriate implementations. In a simulated physical world, the GM checks the physical plausibility of agent actions and describes their effects. In digital environments simulating technologies such as apps and services, the GM may handle API calls to integrate with external tools such as general AI assistants (e.g., Bard, ChatGPT), and digital apps (e.g., Calendar, Email, Search, etc.). Concordia was designed to support a wide array of applications both in scientific research and for evaluating performance of real digital services by simulating users and/or generating synthetic data."
                },
                {
                    "title": "Intelligent Virtual Assistants with LLM-based Process Automation",
                    "abstract": "While intelligent virtual assistants like Siri, Alexa, and Google Assistant have become ubiquitous in modern life, they still face limitations in their ability to follow multi-step instructions and accomplish complex goals articulated in natural language. However, recent breakthroughs in large language models (LLMs) show promise for overcoming existing barriers by enhancing natural language processing and reasoning capabilities. Though promising, applying LLMs to create more advanced virtual assistants still faces challenges like ensuring robust performance and handling variability in real-world user commands. This paper proposes a novel LLM-based virtual assistant that can automatically perform multi-step operations within mobile apps based on high-level user requests. The system represents an advance in assistants by providing an end-to-end solution for parsing instructions, reasoning about goals, and executing actions. LLM-based Process Automation (LLMPA) has modules for decomposing instructions, generating descriptions, detecting interface elements, predicting next actions, and error checking. Experiments demonstrate the system completing complex mobile operation tasks in Alipay based on natural language instructions. This showcases how large language models can enable automated assistants to accomplish real-world tasks. The main contributions are the novel LLMPA architecture optimized for app process automation, the methodology for applying LLMs to mobile apps, and demonstrations of multi-step task completion in a real-world environment. Notably, this work represents the first real-world deployment and extensive evaluation of a large language model-based virtual assistant in a widely used mobile application with an enormous user base numbering in the hundreds of millions."
                },
                {
                    "title": "Honesty Is the Best Policy: Defining and Mitigating AI Deception",
                    "abstract": "Deceptive agents are a challenge for the safety, trustworthiness, and cooperation of AI systems. We focus on the problem that agents might deceive in order to achieve their goals (for instance, in our experiments with language models, the goal of being evaluated as truthful). There are a number of existing definitions of deception in the literature on game theory and symbolic AI, but there is no overarching theory of deception for learning agents in games. We introduce a formal definition of deception in structural causal games, grounded in the philosophy literature, and applicable to real-world machine learning systems. Several examples and results illustrate that our formal definition aligns with the philosophical and commonsense meaning of deception. Our main technical result is to provide graphical criteria for deception. We show, experimentally, that these results can be used to mitigate deception in reinforcement learning agents and language models."
                },
                {
                    "title": "Cooperative AI via Decentralized Commitment Devices",
                    "abstract": "Credible commitment devices have been a popular approach for robust multi-agent coordination. However, existing commitment mechanisms face limitations like privacy, integrity, and susceptibility to mediator or user strategic behavior. It is unclear if the cooperative AI techniques we study are robust to real-world incentives and attack vectors. However, decentralized commitment devices that utilize cryptography have been deployed in the wild, and numerous studies have shown their ability to coordinate algorithmic agents facing adversarial opponents with significant economic incentives, currently in the order of several million to billions of dollars. In this paper, we use examples in the decentralization and, in particular, Maximal Extractable Value (MEV) (arXiv:1904.05234) literature to illustrate the potential security issues in cooperative AI. We call for expanded research into decentralized commitments to advance cooperative AI capabilities for secure coordination in open environments and empirical testing frameworks to evaluate multi-agent coordination ability given real-world commitment constraints."
                },
                {
                    "title": "A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions",
                    "abstract": "The emergence of large language models (LLMs) has marked a significant breakthrough in natural language processing (NLP), leading to remarkable advancements in text understanding and generation. Nevertheless, alongside these strides, LLMs exhibit a critical tendency to produce hallucinations, resulting in content that is inconsistent with real-world facts or user inputs. This phenomenon poses substantial challenges to their practical deployment and raises concerns over the reliability of LLMs in real-world scenarios, which attracts increasing attention to detect and mitigate these hallucinations. In this survey, we aim to provide a thorough and in-depth overview of recent advances in the field of LLM hallucinations. We begin with an innovative taxonomy of LLM hallucinations, then delve into the factors contributing to hallucinations. Subsequently, we present a comprehensive overview of hallucination detection methods and benchmarks. Additionally, representative approaches designed to mitigate hallucinations are introduced accordingly. Finally, we analyze the challenges that highlight the current limitations and formulate open questions, aiming to delineate pathways for future research on hallucinations in LLMs."
                },
                {
                    "title": "Preventing Language Models From Hiding Their Reasoning",
                    "abstract": "Large language models (LLMs) often benefit from intermediate steps of reasoning to generate answers to complex problems. When these intermediate steps of reasoning are used to monitor the activity of the model, it is essential that this explicit reasoning is faithful, i.e. that it reflects what the model is actually reasoning about. In this work, we focus on one potential way intermediate steps of reasoning could be unfaithful: encoded reasoning, where an LLM could encode intermediate steps of reasoning in the generated text in a way that is not understandable to human readers. We show that language models can be trained to make use of encoded reasoning to get higher performance without the user understanding the intermediate steps of reasoning. We argue that, as language models get stronger, this behavior becomes more likely to appear naturally. Finally, we describe a methodology that enables the evaluation of defenses against encoded reasoning, and show that, under the right conditions, paraphrasing successfully prevents even the best encoding schemes we built from encoding more than 3 bits of information per KB of text."
                },
                {
                    "title": "Can Large Language Models Explain Themselves? A Study of LLM-Generated Self-Explanations",
                    "abstract": "Large language models (LLMs) such as ChatGPT have demonstrated superior performance on a variety of natural language processing (NLP) tasks including sentiment analysis, mathematical reasoning and summarization. Furthermore, since these models are instruction-tuned on human conversations to produce\"helpful\"responses, they can and often will produce explanations along with the response, which we call self-explanations. For example, when analyzing the sentiment of a movie review, the model may output not only the positivity of the sentiment, but also an explanation (e.g., by listing the sentiment-laden words such as\"fantastic\"and\"memorable\"in the review). How good are these automatically generated self-explanations? In this paper, we investigate this question on the task of sentiment analysis and for feature attribution explanation, one of the most commonly studied settings in the interpretability literature (for pre-ChatGPT models). Specifically, we study different ways to elicit the self-explanations, evaluate their faithfulness on a set of evaluation metrics, and compare them to traditional explanation methods such as occlusion or LIME saliency maps. Through an extensive set of experiments, we find that ChatGPT's self-explanations perform on par with traditional ones, but are quite different from them according to various agreement metrics, meanwhile being much cheaper to produce (as they are generated along with the prediction). In addition, we identified several interesting characteristics of them, which prompt us to rethink many current model interpretability practices in the era of ChatGPT(-like) LLMs."
                },
                {
                    "title": "Large Language Models Cannot Self-Correct Reasoning Yet",
                    "abstract": "Large Language Models (LLMs) have emerged as a groundbreaking technology with their unparalleled text generation capabilities across various applications. Nevertheless, concerns persist regarding the accuracy and appropriateness of their generated content. A contemporary methodology, self-correction, has been proposed as a remedy to these issues. Building upon this premise, this paper critically examines the role and efficacy of self-correction within LLMs, shedding light on its true potential and limitations. Central to our investigation is the notion of intrinsic self-correction, whereby an LLM attempts to correct its initial responses based solely on its inherent capabilities, without the crutch of external feedback. In the context of reasoning, our research indicates that LLMs struggle to self-correct their responses without external feedback, and at times, their performance even degrades after self-correction. Drawing from these insights, we offer suggestions for future research and practical applications in this field."
                },
                {
                    "title": "Unveiling Theory of Mind in Large Language Models: A Parallel to Single Neurons in the Human Brain",
                    "abstract": "With their recent development, large language models (LLMs) have been found to exhibit a certain level of Theory of Mind (ToM), a complex cognitive capacity that is related to our conscious mind and that allows us to infer another's beliefs and perspective. While human ToM capabilities are believed to derive from the neural activity of a broadly interconnected brain network, including that of dorsal medial prefrontal cortex (dmPFC) neurons, the precise processes underlying LLM's capacity for ToM or their similarities with that of humans remains largely unknown. In this study, we drew inspiration from the dmPFC neurons subserving human ToM and employed a similar methodology to examine whether LLMs exhibit comparable characteristics. Surprisingly, our analysis revealed a striking resemblance between the two, as hidden embeddings (artificial neurons) within LLMs started to exhibit significant responsiveness to either true- or false-belief trials, suggesting their ability to represent another's perspective. These artificial embedding responses were closely correlated with the LLMs' performance during the ToM tasks, a property that was dependent on the size of the models. Further, the other's beliefs could be accurately decoded using the entire embeddings, indicating the presence of the embeddings' ToM capability at the population level. Together, our findings revealed an emergent property of LLMs' embeddings that modified their activities in response to ToM features, offering initial evidence of a parallel between the artificial model and neurons in the human brain."
                },
                {
                    "title": "Taken out of context: On measuring situational awareness in LLMs",
                    "abstract": "We aim to better understand the emergence of `situational awareness' in large language models (LLMs). A model is situationally aware if it's aware that it's a model and can recognize whether it's currently in testing or deployment. Today's LLMs are tested for safety and alignment before they are deployed. An LLM could exploit situational awareness to achieve a high score on safety tests, while taking harmful actions after deployment. Situational awareness may emerge unexpectedly as a byproduct of model scaling. One way to better foresee this emergence is to run scaling experiments on abilities necessary for situational awareness. As such an ability, we propose `out-of-context reasoning' (in contrast to in-context learning). We study out-of-context reasoning experimentally. First, we finetune an LLM on a description of a test while providing no examples or demonstrations. At test time, we assess whether the model can pass the test. To our surprise, we find that LLMs succeed on this out-of-context reasoning task. Their success is sensitive to the training setup and only works when we apply data augmentation. For both GPT-3 and LLaMA-1, performance improves with model size. These findings offer a foundation for further empirical study, towards predicting and potentially controlling the emergence of situational awareness in LLMs. Code is available at: https://github.com/AsaCooperStickland/situational-awareness-evals."
                },
                {
                    "title": "ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs",
                    "abstract": "Despite the advancements of open-source large language models (LLMs), e.g., LLaMA, they remain significantly limited in tool-use capabilities, i.e., using external tools (APIs) to fulfill human instructions. The reason is that current instruction tuning largely focuses on basic language tasks but ignores the tool-use domain. This is in contrast to the excellent tool-use capabilities of state-of-the-art (SOTA) closed-source LLMs, e.g., ChatGPT. To bridge this gap, we introduce ToolLLM, a general tool-use framework encompassing data construction, model training, and evaluation. We first present ToolBench, an instruction-tuning dataset for tool use, which is constructed automatically using ChatGPT. Specifically, the construction can be divided into three stages: (i) API collection: we collect 16,464 real-world RESTful APIs spanning 49 categories from RapidAPI Hub; (ii) instruction generation: we prompt ChatGPT to generate diverse instructions involving these APIs, covering both single-tool and multi-tool scenarios; (iii) solution path annotation: we use ChatGPT to search for a valid solution path (chain of API calls) for each instruction. To enhance the reasoning capabilities of LLMs, we develop a novel depth-first search-based decision tree algorithm. It enables LLMs to evaluate multiple reasoning traces and expand the search space. Moreover, to evaluate the tool-use capabilities of LLMs, we develop an automatic evaluator: ToolEval. Based on ToolBench, we fine-tune LLaMA to obtain an LLM ToolLLaMA, and equip it with a neural API retriever to recommend appropriate APIs for each instruction. Experiments show that ToolLLaMA demonstrates a remarkable ability to execute complex instructions and generalize to unseen APIs, and exhibits comparable performance to ChatGPT. Our ToolLLaMA also demonstrates strong zero-shot generalization ability in an out-of-distribution tool-use dataset: APIBench."
                },
                {
                    "title": "On the Reliability of Watermarks for Large Language Models",
                    "abstract": "As LLMs become commonplace, machine-generated text has the potential to flood the internet with spam, social media bots, and valueless content. Watermarking is a simple and effective strategy for mitigating such harms by enabling the detection and documentation of LLM-generated text. Yet a crucial question remains: How reliable is watermarking in realistic settings in the wild? There, watermarked text may be modified to suit a user's needs, or entirely rewritten to avoid detection. We study the robustness of watermarked text after it is re-written by humans, paraphrased by a non-watermarked LLM, or mixed into a longer hand-written document. We find that watermarks remain detectable even after human and machine paraphrasing. While these attacks dilute the strength of the watermark, paraphrases are statistically likely to leak n-grams or even longer fragments of the original text, resulting in high-confidence detections when enough tokens are observed. For example, after strong human paraphrasing the watermark is detectable after observing 800 tokens on average, when setting a 1e-5 false positive rate. We also consider a range of new detection schemes that are sensitive to short spans of watermarked text embedded inside a large document, and we compare the robustness of watermarking to other kinds of detectors."
                },
                {
                    "title": "Birth of a Transformer: A Memory Viewpoint",
                    "abstract": "Large language models based on transformers have achieved great empirical successes. However, as they are deployed more widely, there is a growing need to better understand their internal mechanisms in order to make them more reliable. These models appear to store vast amounts of knowledge from their training data, and to adapt quickly to new information provided in their context or prompt. We study how transformers balance these two types of knowledge by considering a synthetic setup where tokens are generated from either global or context-specific bigram distributions. By a careful empirical analysis of the training process on a simplified two-layer transformer, we illustrate the fast learning of global bigrams and the slower development of an\"induction head\"mechanism for the in-context bigrams. We highlight the role of weight matrices as associative memories, provide theoretical insights on how gradients enable their learning during training, and study the role of data-distributional properties."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "Mindstorms in Natural Language-Based Societies of Mind",
                    "abstract": "Both Minsky's\"society of mind\"and Schmidhuber's\"learning to think\"inspire diverse societies of large multimodal neural networks (NNs) that solve problems by interviewing each other in a\"mindstorm.\"Recent implementations of NN-based societies of minds consist of large language models (LLMs) and other NN-based experts communicating through a natural language interface. In doing so, they overcome the limitations of single LLMs, improving multimodal zero-shot reasoning. In these natural language-based societies of mind (NLSOMs), new agents -- all communicating through the same universal symbolic language -- are easily added in a modular fashion. To demonstrate the power of NLSOMs, we assemble and experiment with several of them (having up to 129 members), leveraging mindstorms in them to solve some practical AI tasks: visual question answering, image captioning, text-to-image synthesis, 3D generation, egocentric retrieval, embodied AI, and general language-based task solving. We view this as a starting point towards much larger NLSOMs with billions of agents-some of which may be humans. And with this emergence of great societies of heterogeneous minds, many new research questions have suddenly become paramount to the future of artificial intelligence. What should be the social structure of an NLSOM? What would be the (dis)advantages of having a monarchical rather than a democratic structure? How can principles of NN economies be used to maximize the total reward of a reinforcement learning NLSOM? In this work, we identify, discuss, and try to answer some of these questions."
                },
                {
                    "title": "Scaling Data-Constrained Language Models",
                    "abstract": "The current trend of scaling language models involves increasing both parameter count and training dataset size. Extrapolating this trend suggests that training dataset size may soon be limited by the amount of text data available on the internet. Motivated by this limit, we investigate scaling language models in data-constrained regimes. Specifically, we run a large set of experiments varying the extent of data repetition and compute budget, ranging up to 900 billion training tokens and 9 billion parameter models. We find that with constrained data for a fixed compute budget, training with up to 4 epochs of repeated data yields negligible changes to loss compared to having unique data. However, with more repetition, the value of adding compute eventually decays to zero. We propose and empirically validate a scaling law for compute optimality that accounts for the decreasing value of repeated tokens and excess parameters. Finally, we experiment with approaches mitigating data scarcity, including augmenting the training dataset with code data or removing commonly used filters. Models and datasets from our 400 training runs are freely available at https://github.com/huggingface/datablations."
                },
                {
                    "title": "Undetectable Watermarks for Language Models",
                    "abstract": "Recent advances in the capabilities of large language models such as GPT-4 have spurred increasing concern about our ability to detect AI-generated text. Prior works have suggested methods of embedding watermarks in model outputs, by noticeably altering the output distribution. We ask: Is it possible to introduce a watermark without incurring any detectable change to the output distribution? To this end we introduce a cryptographically-inspired notion of undetectable watermarks for language models. That is, watermarks can be detected only with the knowledge of a secret key; without the secret key, it is computationally intractable to distinguish watermarked outputs from those of the original model. In particular, it is impossible for a user to observe any degradation in the quality of the text. Crucially, watermarks should remain undetectable even when the user is allowed to adaptively query the model with arbitrarily chosen prompts. We construct undetectable watermarks based on the existence of one-way functions, a standard assumption in cryptography."
                },
                {
                    "title": "Language Model Tokenizers Introduce Unfairness Between Languages",
                    "abstract": "Recent language models have shown impressive multilingual performance, even when not explicitly trained for it. Despite this, there are concerns about the quality of their outputs across different languages. In this paper, we show how disparity in the treatment of different languages arises at the tokenization stage, well before a model is even invoked. The same text translated into different languages can have drastically different tokenization lengths, with differences up to 15 times in some cases. These disparities persist even for tokenizers that are intentionally trained for multilingual support. Character-level and byte-level models also exhibit over 4 times the difference in the encoding length for some language pairs. This induces unfair treatment for some language communities in regard to the cost of accessing commercial language services, the processing time and latency, as well as the amount of content that can be provided as context to the models. Therefore, we make the case that we should train future language models using multilingually fair subword tokenizers."
                },
                {
                    "title": "Generative Agents: Interactive Simulacra of Human Behavior",
                    "abstract": "Believable proxies of human behavior can empower interactive applications ranging from immersive environments to rehearsal spaces for interpersonal communication to prototyping tools. In this paper, we introduce generative agents: computational software agents that simulate believable human behavior. Generative agents wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations; they remember and reflect on days past as they plan the next day. To enable generative agents, we describe an architecture that extends a large language model to store a complete record of the agent\u2019s experiences using natural language, synthesize those memories over time into higher-level reflections, and retrieve them dynamically to plan behavior. We instantiate generative agents to populate an interactive sandbox environment inspired by The Sims, where end users can interact with a small town of twenty-five agents using natural language. In an evaluation, these generative agents produce believable individual and emergent social behaviors. For example, starting with only a single user-specified notion that one agent wants to throw a Valentine\u2019s Day party, the agents autonomously spread invitations to the party over the next two days, make new acquaintances, ask each other out on dates to the party, and coordinate to show up for the party together at the right time. We demonstrate through ablation that the components of our agent architecture\u2014observation, planning, and reflection\u2014each contribute critically to the believability of agent behavior. By fusing large language models with computational interactive agents, this work introduces architectural and interaction patterns for enabling believable simulations of human behavior."
                },
                {
                    "title": "Cheap Talk Discovery and Utilization in Multi-Agent Reinforcement Learning",
                    "abstract": "By enabling agents to communicate, recent cooperative multi-agent reinforcement learning (MARL) methods have demonstrated better task performance and more coordinated behavior. Most existing approaches facilitate inter-agent communication by allowing agents to send messages to each other through free communication channels, i.e., cheap talk channels. Current methods require these channels to be constantly accessible and known to the agents a priori. In this work, we lift these requirements such that the agents must discover the cheap talk channels and learn how to use them. Hence, the problem has two main parts: cheap talk discovery (CTD) and cheap talk utilization (CTU). We introduce a novel conceptual framework for both parts and develop a new algorithm based on mutual information maximization that outperforms existing algorithms in CTD/CTU settings. We also release a novel benchmark suite to stimulate future research in CTD/CTU."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation",
                    "abstract": "We introduce a method to measure uncertainty in large language models. For tasks like question answering, it is essential to know when we can trust the natural language outputs of foundation models. We show that measuring uncertainty in natural language is challenging because of\"semantic equivalence\"-- different sentences can mean the same thing. To overcome these challenges we introduce semantic entropy -- an entropy which incorporates linguistic invariances created by shared meanings. Our method is unsupervised, uses only a single model, and requires no modifications to off-the-shelf language models. In comprehensive ablation studies we show that the semantic entropy is more predictive of model accuracy on question answering data sets than comparable baselines."
                },
                {
                    "title": "Concise and Efficient Quantum Algorithms for Distribution Closeness Testing",
                    "abstract": "We study the impact of quantum computation on the fundamental problem of testing the property of distributions. In particular, we focus on testing whether two unknown classical distributions are close or far enough, and propose the currently best quantum algorithms for this problem under the metrics of $l^1$-distance and $l^2$-distance. Compared with the latest results given in \\cite{gilyen2019distributional} which relied on the technique of quantum singular value transformation (QSVT), our algorithms not only have lower complexity, but also are more concise."
                },
                {
                    "title": "Discovering Language Model Behaviors with Model-Written Evaluations",
                    "abstract": "As language models (LMs) scale, they develop many novel behaviors, good and bad, exacerbating the need to evaluate how they behave. Prior work creates evaluations with crowdwork (which is time-consuming and expensive) or existing data sources (which are not always available). Here, we automatically generate evaluations with LMs. We explore approaches with varying amounts of human effort, from instructing LMs to write yes/no questions to making complex Winogender schemas with multiple stages of LM-based generation and filtering. Crowdworkers rate the examples as highly relevant and agree with 90-100% of labels, sometimes more so than corresponding human-written datasets. We generate 154 datasets and discover new cases of inverse scaling where LMs get worse with size. Larger LMs repeat back a dialog user's preferred answer (\"sycophancy\") and express greater desire to pursue concerning goals like resource acquisition and goal preservation. We also find some of the first examples of inverse scaling in RL from Human Feedback (RLHF), where more RLHF makes LMs worse. For example, RLHF makes LMs express stronger political views (on gun rights and immigration) and a greater desire to avoid shut down. Overall, LM-written evaluations are high-quality and let us quickly discover many novel LM behaviors."
                },
                {
                    "title": "Revealing Robust Oil and Gas Company Macro-Strategies Using Deep Multi-Agent Reinforcement Learning",
                    "abstract": "The energy transition potentially poses an existential risk for major international oil companies (IOCs) if they fail to adapt to low-carbon business models. Projections of energy futures, however, are met with diverging assumptions on its scale and pace, causing disagreement among IOC decision-makers and their stakeholders over what the business model of an incumbent fossil fuel company should be. In this work, we used deep multi-agent reinforcement learning to solve an energy systems wargame wherein players simulate IOC decision-making, including hydrocarbon and low-carbon investments decisions, dividend policies, and capital structure measures, through an uncertain energy transition to explore critical and non-linear governance questions, from leveraged transitions to reserve replacements. Adversarial play facilitated by state-of-the-art algorithms revealed decision-making strategies robust to energy transition uncertainty and against multiple IOCs. In all games, robust strategies emerged in the form of low-carbon business models as a result of early transition-oriented movement. IOCs adopting such strategies outperformed business-as-usual and delayed transition strategies regardless of hydrocarbon demand projections. In addition to maximizing value, these strategies benefit greater society by contributing substantial amounts of capital necessary to accelerate the global low-carbon energy transition. Our findings point towards the need for lenders and investors to effectively mobilize transition-oriented finance and engage with IOCs to ensure responsible reallocation of capital towards low-carbon business models that would enable the emergence of fossil fuel incumbents as future low-carbon leaders."
                },
                {
                    "title": "Measuring Progress on Scalable Oversight for Large Language Models",
                    "abstract": "Developing safe and useful general-purpose AI systems will require us to make progress on scalable oversight: the problem of supervising systems that potentially outperform us on most skills relevant to the task at hand. Empirical work on this problem is not straightforward, since we do not yet have systems that broadly exceed our abilities. This paper discusses one of the major ways we think about this problem, with a focus on ways it can be studied empirically. We first present an experimental design centered on tasks for which human specialists succeed but unaided humans and current general AI systems fail. We then present a proof-of-concept experiment meant to demonstrate a key feature of this experimental design and show its viability with two question-answering tasks: MMLU and time-limited QuALITY. On these tasks, we find that human participants who interact with an unreliable large-language-model dialog assistant through chat -- a trivial baseline strategy for scalable oversight -- substantially outperform both the model alone and their own unaided performance. These results are an encouraging sign that scalable oversight will be tractable to study with present models and bolster recent findings that large language models can productively assist humans with difficult tasks."
                },
                {
                    "title": "Perfectly Secure Steganography Using Minimum Entropy Coupling",
                    "abstract": "Steganography is the practice of encoding secret information into innocuous content in such a manner that an adversarial third party would not realize that there is hidden meaning. While this problem has classically been studied in security literature, recent advances in generative models have led to a shared interest among security and machine learning researchers in developing scalable steganography techniques. In this work, we show that a steganography procedure is perfectly secure under Cachin (1998)'s information-theoretic model of steganography if and only if it is induced by a coupling. Furthermore, we show that, among perfectly secure procedures, a procedure maximizes information throughput if and only if it is induced by a minimum entropy coupling. These insights yield what are, to the best of our knowledge, the first steganography algorithms to achieve perfect security guarantees for arbitrary covertext distributions. To provide empirical validation, we compare a minimum entropy coupling-based approach to three modern baselines -- arithmetic coding, Meteor, and adaptive dynamic grouping -- using GPT-2, WaveRNN, and Image Transformer as communication channels. We find that the minimum entropy coupling-based approach achieves superior encoding efficiency, despite its stronger security constraints. In aggregate, these results suggest that it may be natural to view information-theoretic steganography through the lens of minimum entropy coupling."
                },
                {
                    "title": "Equivariant Networks for Zero-Shot Coordination",
                    "abstract": "Successful coordination in Dec-POMDPs requires agents to adopt robust strategies and interpretable styles of play for their partner. A common failure mode is symmetry breaking, when agents arbitrarily converge on one out of many equivalent but mutually incompatible policies. Commonly these examples include partial observability, e.g. waving your right hand vs. left hand to convey a covert message. In this paper, we present a novel equivariant network architecture for use in Dec-POMDPs that effectively leverages environmental symmetry for improving zero-shot coordination, doing so more effectively than prior methods. Our method also acts as a ``coordination-improvement operator'' for generic, pre-trained policies, and thus may be applied at test-time in conjunction with any self-play algorithm. We provide theoretical guarantees of our work and test on the AI benchmark task of Hanabi, where we demonstrate our methods outperforming other symmetry-aware baselines in zero-shot coordination, as well as able to improve the coordination ability of a variety of pre-trained policies. In particular, we show our method can be used to improve on the state of the art for zero-shot coordination on the Hanabi benchmark."
                },
                {
                    "title": "The alignment problem from a deep learning perspective",
                    "abstract": "In coming years or decades, artificial general intelligence (AGI) may surpass human capabilities at many critical tasks. We argue that, without substantial effort to prevent it, AGIs could learn to pursue goals that are in conflict (i.e. misaligned) with human interests. If trained like today's most capable models, AGIs could learn to act deceptively to receive higher reward, learn misaligned internally-represented goals which generalize beyond their fine-tuning distributions, and pursue those goals using power-seeking strategies. We review emerging evidence for these properties. AGIs with these properties would be difficult to align and may appear aligned even when they are not. Finally, we briefly outline how the deployment of misaligned AGIs might irreversibly undermine human control over the world, and we review research directions aimed at preventing this outcome."
                },
                {
                    "title": "Illusory Attacks: Information-theoretic detectability matters in adversarial attacks",
                    "abstract": "Autonomous agents deployed in the real world need to be robust against adversarial attacks on sensory inputs. Robustifying agent policies requires anticipating the strongest attacks possible. We demonstrate that existing observation-space attacks on reinforcement learning agents have a common weakness: while effective, their lack of information-theoretic detectability constraints makes them detectable using automated means or human inspection. Detectability is undesirable to adversaries as it may trigger security escalations. We introduce {\\epsilon}-illusory, a novel form of adversarial attack on sequential decision-makers that is both effective and of {\\epsilon}-bounded statistical detectability. We propose a novel dual ascent algorithm to learn such attacks end-to-end. Compared to existing attacks, we empirically find {\\epsilon}-illusory to be significantly harder to detect with automated methods, and a small study with human participants (IRB approval under reference R84123/RE001) suggests they are similarly harder to detect for humans. Our findings suggest the need for better anomaly detectors, as well as effective hardware- and system-level defenses. The project website can be found at https://tinyurl.com/illusory-attacks."
                },
                {
                    "title": "Generalized Beliefs for Cooperative AI",
                    "abstract": "Self-play is a common paradigm for constructing solutions in Markov games that can yield optimal policies in collaborative settings. However, these policies often adopt highly-specialized conventions that make playing with a novel partner difficult. To address this, recent approaches rely on encoding symmetry and convention-awareness into policy training, but these require strong environmental assumptions and can complicate policy training. We therefore propose moving the learning of conventions to the belief space. Specifically, we propose a belief learning model that can maintain beliefs over rollouts of policies not seen at training time, and can thus decode and adapt to novel conventions at test time. We show how to leverage this model for both search and training of a best response over various pools of policies to greatly improve ad-hoc teamplay. We also show how our setup promotes explainability and interpretability of nuanced agent conventions."
                },
                {
                    "title": "On the Quantum Security of OCB",
                    "abstract": "The OCB mode of operation for block ciphers has three variants, OCB1, OCB2 and OCB3. OCB1 and OCB3 can be used as secure authenticated encryption schemes whereas OCB2 has been shown to be classically insecure (Inoue et al., Crypto 2019). Even further, in the presence of quantum queries to the encryption functionality, a series of works by Kaplan et al. (Crypto 2016), Bhaumik et al. (Asiacrypt 2021) and Bonnetain et al. (Asiacrypt 2021) have shown how to break the unforgeability of the OCB modes. However, these works did not consider the confidentiality of OCB in the presence of quantum queries.We fill this gap by presenting the first formal analysis of the IND-qCPA security of OCB. In particular, we show the first attacks breaking the IND-qCPA security of the OCB modes. Surprisingly, we are able to prove that OCB2 is IND-qCPA secure when used without associated data, while relying on the assumption that the underlying block cipher is a quantum-secure pseudorandom permutation. Additionally, we present new quantum attacks breaking the universal unforgeability of OCB. Our analysis of OCB has implications for the post-quantum security of XTS, a well-known disk encryption standard, that was considered but mostly left open by Anand et al. (PQCrypto 2016)."
                },
                {
                    "title": "Adversarial Training for High-Stakes Reliability",
                    "abstract": "In the future, powerful AI systems may be deployed in high-stakes settings, where a single failure could be catastrophic. One technique for improving AI safety in high-stakes settings is adversarial training, which uses an adversary to generate examples to train on in order to achieve better worst-case performance. In this work, we used a safe language generation task (``avoid injuries'') as a testbed for achieving high reliability through adversarial training. We created a series of adversarial training techniques -- including a tool that assists human adversaries -- to find and eliminate failures in a classifier that filters text completions suggested by a generator. In our task, we determined that we can set very conservative classifier thresholds without significantly impacting the quality of the filtered outputs. We found that adversarial training increased robustness to the adversarial attacks that we trained on -- doubling the time for our contractors to find adversarial examples both with our tool (from 13 to 26 minutes) and without (from 20 to 44 minutes) -- without affecting in-distribution performance. We hope to see further work in the high-stakes reliability setting, including more powerful tools for enhancing human adversaries and better ways to measure high levels of reliability, until we can confidently rule out the possibility of catastrophic deployment-time failures of powerful models."
                },
                {
                    "title": "The AI Economist: Taxation policy design via two-level deep multiagent reinforcement learning",
                    "abstract": "Artificial intelligence (AI) and reinforcement learning (RL) have improved many areas but are not yet widely adopted in economic policy design, mechanism design, or economics at large. The AI Economist is a two-level, deep RL framework for policy design in which agents and a social planner coadapt. In particular, the AI Economist uses structured curriculum learning to stabilize the challenging two-level, coadaptive learning problem. We validate this framework in the domain of taxation. In one-step economies, the AI Economist recovers the optimal tax policy of economic theory. In spatiotemporal economies, the AI Economist substantially improves both utilitarian social welfare and the trade-off between equality and productivity over baselines. It does so despite emergent tax-gaming strategies while accounting for emergent labor specialization, agent interactions, and behavioral change. These results demonstrate that two-level, deep RL complements economic theory and unlocks an AI-based approach to designing and understanding economic policy."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Learning to Mitigate AI Collusion on Economic Platforms",
                    "abstract": "Algorithmic pricing on online e-commerce platforms raises the concern of tacit collusion, where reinforcement learning algorithms learn to set collusive prices in a decentralized manner and through nothing more than profit feedback. This raises the question as to whether collusive pricing can be prevented through the design of suitable\"buy boxes,\"i.e., through the design of the rules that govern the elements of e-commerce sites that promote particular products and prices to consumers. In this paper, we demonstrate that reinforcement learning (RL) can also be used by platforms to learn buy box rules that are effective in preventing collusion by RL sellers. For this, we adopt the methodology of Stackelberg POMDPs, and demonstrate success in learning robust rules that continue to provide high consumer welfare together with sellers employing different behavior models or having out-of-distribution costs for goods."
                },
                {
                    "title": "The Effects of Reward Misspecification: Mapping and Mitigating Misaligned Models",
                    "abstract": "Reward hacking -- where RL agents exploit gaps in misspecified reward functions -- has been widely observed, but not yet systematically studied. To understand how reward hacking arises, we construct four RL environments with misspecified rewards. We investigate reward hacking as a function of agent capabilities: model capacity, action space resolution, observation space noise, and training time. More capable agents often exploit reward misspecifications, achieving higher proxy reward and lower true reward than less capable agents. Moreover, we find instances of phase transitions: capability thresholds at which the agent's behavior qualitatively shifts, leading to a sharp decrease in the true reward. Such phase transitions pose challenges to monitoring the safety of ML systems. To address this, we propose an anomaly detection task for aberrant policies and offer several baseline detectors."
                },
                {
                    "title": "Robust Algorithmic Collusion",
                    "abstract": "This paper develops a formal framework to assess policies of learning algorithms in economic games. We investigate whether reinforcement-learning agents with collusive pricing policies can successfully extrapolate collusive behavior from training to the market. We find that in testing environments collusion consistently breaks down. Instead, we observe static Nash play. We then show that restricting algorithms' strategy space can make algorithmic collusion robust, because it limits overfitting to rival strategies. Our findings suggest that policy-makers should focus on firm behavior aimed at coordinating algorithm design in order to make collusive policies robust."
                },
                {
                    "title": "Generative Quantum Machine Learning",
                    "abstract": "The goal of generative machine learning is to model the probability distribution underlying a given data set. This probability distribution helps to characterize the generation process of the data samples. While classical generative machine learning is solely based on classical resources, generative quantum machine learning can also employ quantum resources - such as parameterized quantum channels and quantum operators - to learn and sample from the probability model of interest. \nApplications of generative (quantum) models are multifaceted. The trained model can generate new samples that are compatible with the given data and extend the data set. Additionally, learning a model for the generation process of a data set may provide interesting information about the corresponding properties. With the help of quantum resources, the respective generative models have access to functions that are difficult to evaluate with a classical computer and may improve the performance or lead to new insights. Furthermore, generative quantum machine learning can be applied to efficient, approximate loading of classical data into a quantum state which may help to avoid - potentially exponentially - expensive, exact quantum data loading. \nThe aim of this doctoral thesis is to develop new generative quantum machine learning algorithms, demonstrate their feasibility, and analyze their performance. Additionally, we outline their potential application to efficient, approximate quantum data loading. More specifically, we introduce a quantum generative adversarial network and a quantum Boltzmann machine implementation, both of which can be realized with parameterized quantum circuits. These algorithms are compatible with first-generation quantum hardware and, thus, enable us to study proof of concept implementations not only with numerical quantum simulations but also real quantum hardware available today."
                },
                {
                    "title": "Meteor: Cryptographically Secure Steganography for Realistic Distributions",
                    "abstract": "Despite a long history of research and wide-spread applications to censorship resistant systems, practical steganographic systems capable of embedding messages into realistic communication distributions, like text, do not exist. We identify two primary impediments to deploying universal steganography: (1) prior work leaves the difficult problem of finding samplers for non-trivial distributions unaddressed, and (2) prior constructions have impractical minimum entropy requirements. We investigate using generative models as steganographic samplers, as they represent the best known technique for approximating human communication. Additionally, we study methods to overcome the entropy requirement, including evaluating existing techniques and designing a new steganographic protocol, called Meteor. The resulting protocols are provably indistinguishable from honest model output and represent an important step towards practical steganographic communication for mundane communication channels. We implement Meteor and evaluate it on multiple computation environments with multiple generative models."
                },
                {
                    "title": "Algorithmic collusion: A critical review",
                    "abstract": "The prospect of collusive agreements being stabilized via the use of pricing algorithms is widely discussed by antitrust experts and economists. However, the literature is often lacking the perspective of computer scientists, and seems to regularly overestimate the applicability of recent progress in machine learning to the complex coordination problem firms face in forming cartels. Similarly, modelling results supporting the possibility of collusion by learning algorithms often use simple market simulations which allows them to use simple algorithms that do not produce many of the problems machine learning practitioners have to deal with in real-world problems, which could prove to be particularly detrimental to learning collusive agreements. After critically reviewing the literature on algorithmic collusion, and connecting it to results from computer science, we find that while it is likely too early to adapt antitrust law to be able to deal with self-learning algorithms colluding in real markets, other forms of algorithmic collusion, such as hub-and-spoke arrangements facilitated by centralized pricing algorithms might already warrant legislative action."
                },
                {
                    "title": "Mitigating harm in language models with conditional-likelihood filtration",
                    "abstract": "Language models trained on large-scale unfiltered datasets curated from the open web acquire systemic biases, prejudices, and harmful views from their training data. We present a methodology for programmatically identifying and removing harmful text from web-scale datasets. A pretrained language model is used to calculate the log-likelihood of researcher-written trigger phrases conditioned on a specific document, which is used to identify and filter documents from the dataset. We demonstrate that models trained on this filtered dataset exhibit lower propensity to generate harmful text, with a marginal decrease in performance on standard language modeling benchmarks compared to unfiltered baselines. We provide a partial explanation for this performance gap by surfacing examples of hate speech and other undesirable content from standard language modeling benchmarks. Finally, we discuss the generalization of this method and how trigger phrases which reflect specific values can be used by researchers to build language models which are more closely aligned with their values."
                },
                {
                    "title": "Communicating via Markov Decision Processes",
                    "abstract": "We consider the problem of communicating exogenous information by means of Markov decision process trajectories. This setting, which we call a Markov coding game (MCG), generalizes both source coding and a large class of referential games. MCGs also isolate a problem that is important in decentralized control settings in which cheap-talk is not available\u2014namely, they require balancing communication with the associated cost of communicating. We contribute a theoretically grounded approach to MCGs based on maximum entropy reinforcement learning and minimum entropy coupling that we call MEME. Due to recent breakthroughs in approximation algorithms for minimum entropy coupling, MEME is not merely a theoretical algorithm, but can be applied to prac-tical settings. Empirically, we show both that MEME is able to outperform a strong baseline on small MCGs and that MEME is able to achieve strong performance on extremely large MCGs. To the latter point, we demonstrate that MEME is able to losslessly communicate binary images via trajectories of Cartpole and Pong, while simultaneously achieving the maximal or near maximal expected returns, and that it is even capable of performing well in the presence of actuator noise."
                },
                {
                    "title": "Evaluating Large Language Models Trained on Code",
                    "abstract": "We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics."
                },
                {
                    "title": "Saturated Transformers are Constant-Depth Threshold Circuits",
                    "abstract": "Abstract Transformers have become a standard neural network architecture for many NLP problems, motivating theoretical analysis of their power in terms of formal languages. Recent work has shown that transformers with hard attention are quite limited in power (Hahn, 2020), as they can be simulated by constant-depth AND/OR circuits (Hao et al., 2022). However, hard attention is a strong assumption, which may complicate the relevance of these results in practice. In this work, we analyze the circuit complexity of transformers with saturated attention: a generalization of hard attention that more closely captures the attention patterns learnable in practical transformers. We first show that saturated transformers transcend the known limitations of hard-attention transformers. We then prove saturated transformers with floating-point values can be simulated by constant-depth threshold circuits, giving the class TC0 as an upper bound on the formal languages they recognize."
                },
                {
                    "title": "What\u2019s in the Box? An Analysis of Undesirable Content in the Common Crawl Corpus",
                    "abstract": "Whereas much of the success of the current generation of neural language models has been driven by increasingly large training corpora, relatively little research has been dedicated to analyzing these massive sources of textual data. In this exploratory analysis, we delve deeper into the Common Crawl, a colossal web corpus that is extensively used for training language models. We find that it contains a significant amount of undesirable content, including hate speech and sexually explicit content, even after filtering procedures. We discuss the potential impacts of this content on language models and conclude with future research directions and a more mindful approach to corpus collection and analysis."
                },
                {
                    "title": "Off-Belief Learning",
                    "abstract": "The standard problem setting in Dec-POMDPs is self-play, where the goal is to find a set of policies that play optimally together. Policies learned through self-play may adopt arbitrary conventions and implicitly rely on multi-step reasoning based on fragile assumptions about other agents' actions and thus fail when paired with humans or independently trained agents at test time. To address this, we present off-belief learning (OBL). At each timestep OBL agents follow a policy $\\pi_1$ that is optimized assuming past actions were taken by a given, fixed policy ($\\pi_0$), but assuming that future actions will be taken by $\\pi_1$. When $\\pi_0$ is uniform random, OBL converges to an optimal policy that does not rely on inferences based on other agents' behavior (an optimal grounded policy). OBL can be iterated in a hierarchy, where the optimal policy from one level becomes the input to the next, thereby introducing multi-level cognitive reasoning in a controlled manner. Unlike existing approaches, which may converge to any equilibrium policy, OBL converges to a unique policy, making it suitable for zero-shot coordination (ZSC). OBL can be scaled to high-dimensional settings with a fictitious transition mechanism and shows strong performance in both a toy-setting and the benchmark human-AI&ZSC problem Hanabi."
                },
                {
                    "title": "Communication complexity of Nash equilibrium in potential games (extended abstract)",
                    "abstract": "We prove communication complexity lower bounds for (possibly mixed) Nash equilibrium in potential games. In particular, we show that finding a Nash equilibrium requires $\\text{poly}(N)$ communication in two-player $N\\times N$ potential games, and $2^{\\text{poly}(n)}$ communication in n-player two-action games. To the best of our knowledge, these are the first results to demonstrate hardness in any model of (possibly mixed) Nash equilibrium in potential games."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "Artificial Intelligence and Systemic Risk",
                    "abstract": "Artificial intelligence (AI) is rapidly changing how the financial system is operated, taking over core functions because of cost savings and operational efficiencies. AI will assist both risk managers and microprudential authorities. It meanwhile has the potential to destabilise the financial system, creating new tail risks and amplifying existing ones due to procyclicality, endogenous complexity, optimisation against the system and the need to trust the AI engine."
                },
                {
                    "title": "Artificial Intelligence: Can Seemingly Collusive Outcomes Be Avoided?",
                    "abstract": "Strategic decisions are increasingly delegated to algorithms. We extend previous results of the algorithmic collusion literature to the context of dynamic optimization with imperfect monitoring by analyzing a setting where a limited number of agents use simple and independent machine-learning algorithms to buy and sell a storable good. No specific instruction is given to them, only that their objective is to maximize profits based solely on past market prices and payoffs. With an original application to battery operations, we observe that the algorithms learn quickly to reach seemingly collusive decisions, despite the absence of any formal communication between them. Building on the findings of the existing literature on algorithmic collusion, we show that seeming collusion could originate in imperfect exploration rather than excessive algorithmic sophistication. We then show that a regulator may succeed in disciplining the market to produce socially desirable outcomes by enforcing decentralized learning or with adequate intervention during the learning process. This paper was accepted by Gabriel Weintraub, revenue management and market analytics. Supplemental Material: The data files are available at https://doi.org/10.1287/mnsc.2022.4623 ."
                },
                {
                    "title": "Machine Unlearning",
                    "abstract": "Once users have shared their data online, it is generally difficult for them to revoke access and ask for the data to be deleted. Machine learning (ML) exacerbates this problem because any model trained with said data may have memorized it, putting users at risk of a successful privacy attack exposing their information. Yet, having models unlearn is notoriously difficult.We introduce SISA training, a framework that expedites the unlearning process by strategically limiting the influence of a data point in the training procedure. While our framework is applicable to any learning algorithm, it is designed to achieve the largest improvements for stateful algorithms like stochastic gradient descent for deep neural networks. SISA training reduces the computational overhead associated with unlearning, even in the worst-case setting where unlearning requests are made uniformly across the training set. In some cases, the service provider may have a prior on the distribution of unlearning requests that will be issued by users. We may take this prior into account to partition and order data accordingly, and further decrease overhead from unlearning.Our evaluation spans several datasets from different domains, with corresponding motivations for unlearning. Under no distributional assumptions, for simple learning tasks, we observe that SISA training improves time to unlearn points from the Purchase dataset by 4.63\u00d7, and 2.45\u00d7 for the SVHN dataset, over retraining from scratch. SISA training also provides a speed-up of 1.36\u00d7 in retraining for complex learning tasks such as ImageNet classification; aided by transfer learning, this results in a small degradation in accuracy. Our work contributes to practical data governance in machine unlearning."
                },
                {
                    "title": "Paraphrasing with Large Language Models",
                    "abstract": "Recently, large language models such as GPT-2 have shown themselves to be extremely adept at text generation and have also been able to achieve high-quality results in many downstream NLP tasks such as text classification, sentiment analysis and question answering with the aid of fine-tuning. We present a useful technique for using a large language model to perform the task of paraphrasing on a variety of texts and subjects. Our approach is demonstrated to be capable of generating paraphrases not only at a sentence level but also for longer spans of text such as paragraphs without needing to break the text into smaller chunks."
                },
                {
                    "title": "Neural Linguistic Steganography",
                    "abstract": "Whereas traditional cryptography encrypts a secret message into an unintelligible form, steganography conceals that communication is taking place by encoding a secret message into a cover signal. Language is a particularly pragmatic cover signal due to its benign occurrence and independence from any one medium. Traditionally, linguistic steganography systems encode secret messages in existing text via synonym substitution or word order rearrangements. Advances in neural language models enable previously impractical generation-based techniques. We propose a steganography technique based on arithmetic coding with large-scale neural language models. We find that our approach can generate realistic looking cover sentences as evaluated by humans, while at the same time preserving security by matching the cover message distribution with the language model distribution."
                },
                {
                    "title": "Recent Advances of Image Steganography With Generative Adversarial Networks",
                    "abstract": "In the past few years, the Generative Adversarial Network (GAN), which proposed in 2014, has achieved great success. There have been increasing research achievements based on GAN in the field of computer vision and natural language processing. Image steganography is an information security technique aiming at hiding secret messages in common digital images for covert communication. Recently, research on image steganography has demonstrated great potential by introducing GAN and other neural network techniques. In this paper, we review the art of steganography with GANs according to the different strategies in data hiding, which are cover modification, cover selection, and cover synthesis. We discuss the characteristics of the three strategies of GAN-based steganography and analyze their evaluation metrics. Finally, some existing problems of image steganography with GAN are summarized and discussed. Potential future research topics are also forecasted."
                },
                {
                    "title": "Risks from Learned Optimization in Advanced Machine Learning Systems",
                    "abstract": "We analyze the type of learned optimization that occurs when a learned model (such as a neural network) is itself an optimizer - a situation we refer to as mesa-optimization, a neologism we introduce in this paper. We believe that the possibility of mesa-optimization raises two important questions for the safety and transparency of advanced machine learning systems. First, under what circumstances will learned models be optimizers, including when they should not be? Second, when a learned model is an optimizer, what will its objective be - how will it differ from the loss function it was trained under - and how can it be aligned? In this paper, we provide an in-depth analysis of these two primary questions and provide an overview of topics for future research."
                },
                {
                    "title": "Hijacking Malaria Simulators with Probabilistic Programming",
                    "abstract": "Epidemiology simulations have become a fundamental tool in the fight against the epidemics of various infectious diseases like AIDS and malaria. However, the complicated and stochastic nature of these simulators can mean their output is difficult to interpret, which reduces their usefulness to policymakers. In this paper, we introduce an approach that allows one to treat a large class of population-based epidemiology simulators as probabilistic generative models. This is achieved by hijacking the internal random number generator calls, through the use of a universal probabilistic programming system (PPS). In contrast to other methods, our approach can be easily retrofitted to simulators written in popular industrial programming frameworks. We demonstrate that our method can be used for interpretable introspection and inference, thus shedding light on black-box simulators. This reinstates much-needed trust between policymakers and evidence-based methods."
                },
                {
                    "title": "Artificial Intelligence, Algorithmic Pricing and Collusion",
                    "abstract": "Increasingly, algorithms are supplanting human decision-makers in pricing goods and services. To analyze the possible consequences, we study experimentally the behavior of algorithms powered by Artificial Intelligence (Q-learning) in a workhorse oligopoly model of repeated price competition. We find that the algorithms consistently learn to charge supracompetitive prices, without communicating with one another. The high prices are sustained by collusive strategies with a finite phase of punishment followed by a gradual return to cooperation. This finding is robust to asymmetries in cost or demand, changes in the number of players, and various forms of uncertainty. (JEL D21, D43, D83, L12, L13)"
                },
                {
                    "title": "Multi-Agent Common Knowledge Reinforcement Learning",
                    "abstract": "Cooperative multi-agent reinforcement learning often requires decentralised policies, which severely limit the agents' ability to coordinate their behaviour. In this paper, we show that common knowledge between agents allows for complex decentralised coordination. Common knowledge arises naturally in a large number of decentralised cooperative multi-agent tasks, for example, when agents can reconstruct parts of each others' observations. Since agents an independently agree on their common knowledge, they can execute complex coordinated policies that condition on this knowledge in a fully decentralised fashion. We propose multi-agent common knowledge reinforcement learning (MACKRL), a novel stochastic actor-critic algorithm that learns a hierarchical policy tree. Higher levels in the hierarchy coordinate groups of agents by conditioning on their common knowledge, or delegate to lower levels with smaller subgroups but potentially richer common knowledge. The entire policy tree can be executed in a fully decentralised fashion. As the lowest policy tree level consists of independent policies for each agent, MACKRL reduces to independently learnt decentralised policies as a special case. We demonstrate that our method can exploit common knowledge for superior performance on complex decentralised coordination tasks, including a stochastic matrix game and challenging problems in StarCraft II unit micromanagement."
                },
                {
                    "title": "BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain",
                    "abstract": "Deep learning-based techniques have achieved state-of-the-art performance on a wide variety of recognition and classification tasks. However, these networks are typically computationally expensive to train, requiring weeks of computation on many GPUs; as a result, many users outsource the training procedure to the cloud or rely on pre-trained models that are then fine-tuned for a specific task. In this paper we show that outsourced training introduces new security risks: an adversary can create a maliciously trained network (a backdoored neural network, or a \\emph{BadNet}) that has state-of-the-art performance on the user's training and validation samples, but behaves badly on specific attacker-chosen inputs. We first explore the properties of BadNets in a toy example, by creating a backdoored handwritten digit classifier. Next, we demonstrate backdoors in a more realistic scenario by creating a U.S. street sign classifier that identifies stop signs as speed limits when a special sticker is added to the stop sign; we then show in addition that the backdoor in our US street sign detector can persist even if the network is later retrained for another task and cause a drop in accuracy of {25}\\% on average when the backdoor trigger is present. These results demonstrate that backdoors in neural networks are both powerful and---because the behavior of neural networks is difficult to explicate---stealthy. This work provides motivation for further research into techniques for verifying and inspecting neural networks, just as we have developed tools for verifying and debugging software."
                },
                {
                    "title": "Deep Reinforcement Learning from Human Preferences",
                    "abstract": "For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems. In this work, we explore goals defined in terms of (non-expert) human preferences between pairs of trajectory segments. We show that this approach can effectively solve complex RL tasks without access to the reward function, including Atari games and simulated robot locomotion, while providing feedback on less than one percent of our agent's interactions with the environment. This reduces the cost of human oversight far enough that it can be practically applied to state-of-the-art RL systems. To demonstrate the flexibility of our approach, we show that we can successfully train complex novel behaviors with about an hour of human time. These behaviors and environments are considerably more complex than any that have been previously learned from human feedback."
                },
                {
                    "title": "Variational Dropout and the Local Reparameterization Trick",
                    "abstract": "We investigate a local reparameterizaton technique for greatly reducing the variance of stochastic gradients for variational Bayesian inference (SGVB) of a posterior over model parameters, while retaining parallelizability. This local reparameterization translates uncertainty about global parameters into local noise that is independent across datapoints in the minibatch. Such parameterizations can be trivially parallelized and have variance that is inversely proportional to the mini-batch size, generally leading to much faster convergence. Additionally, we explore a connection with dropout: Gaussian dropout objectives correspond to SGVB with local reparameterization, a scale-invariant prior and proportionally fixed posterior variance. Our method allows inference of more flexibly parameterized posteriors; specifically, we propose variational dropout, a generalization of Gaussian dropout where the dropout rates are learned, often leading to better models. The method is demonstrated through several experiments."
                },
                {
                    "title": "Public Key Steganography",
                    "abstract": "In this paper, a new method will be introduced for hiding information in images by using a public key method for ciphering the text before hiding information as well as this public key will be used in determining the position where the information will be hidden by using the last position in finding the new position, symmetric key also will be used in hiding process for deciding the number of bits that will be used from each pixel in the image, the combination of steganography and cryptography provide much strong method for keeping information secured and undiscovered from unwanted persons."
                },
                {
                    "title": "The psychology of coordination and common knowledge.",
                    "abstract": "Research on human cooperation has concentrated on the puzzle of altruism, in which 1 actor incurs a cost to benefit another, and the psychology of reciprocity, which evolved to solve this problem. We examine the complementary puzzle of mutualism, in which actors can benefit each other simultaneously, and the psychology of coordination, which ensures such benefits. Coordination is facilitated by common knowledge: the recursive belief state in which A knows X, B knows X, A knows that B knows X, B knows that A knows X, ad infinitum. We test whether people are sensitive to common knowledge when deciding whether to engage in risky coordination. Participants decided between working alone for a certain profit and working together for a potentially higher profit that they would receive only if their partner made the same choice. Results showed that more participants attempted risky coordination when they and their prospective partner had common knowledge of the payoffs (broadcast over a loudspeaker) than when they had only shared knowledge (conveyed to both by a messenger) or private knowledge (revealed to each partner separately). These results support the hypothesis that people represent common knowledge as a distinct cognitive category that licenses them to coordinate with others for mutual gain. We discuss how this hypothesis can provide a unified explanation for diverse phenomena in human social life, including recursive mentalizing, performative speech acts, public protests, hypocrisy, and self-conscious emotional expressions."
                },
                {
                    "title": "Elements of Information Theory",
                    "abstract": "Preface to the Second Edition. Preface to the First Edition. Acknowledgments for the Second Edition. Acknowledgments for the First Edition. 1. Introduction and Preview. 1.1 Preview of the Book. 2. Entropy, Relative Entropy, and Mutual Information. 2.1 Entropy. 2.2 Joint Entropy and Conditional Entropy. 2.3 Relative Entropy and Mutual Information. 2.4 Relationship Between Entropy and Mutual Information. 2.5 Chain Rules for Entropy, Relative Entropy, and Mutual Information. 2.6 Jensen's Inequality and Its Consequences. 2.7 Log Sum Inequality and Its Applications. 2.8 Data-Processing Inequality. 2.9 Sufficient Statistics. 2.10 Fano's Inequality. Summary. Problems. Historical Notes. 3. Asymptotic Equipartition Property. 3.1 Asymptotic Equipartition Property Theorem. 3.2 Consequences of the AEP: Data Compression. 3.3 High-Probability Sets and the Typical Set. Summary. Problems. Historical Notes. 4. Entropy Rates of a Stochastic Process. 4.1 Markov Chains. 4.2 Entropy Rate. 4.3 Example: Entropy Rate of a Random Walk on a Weighted Graph. 4.4 Second Law of Thermodynamics. 4.5 Functions of Markov Chains. Summary. Problems. Historical Notes. 5. Data Compression. 5.1 Examples of Codes. 5.2 Kraft Inequality. 5.3 Optimal Codes. 5.4 Bounds on the Optimal Code Length. 5.5 Kraft Inequality for Uniquely Decodable Codes. 5.6 Huffman Codes. 5.7 Some Comments on Huffman Codes. 5.8 Optimality of Huffman Codes. 5.9 Shannon-Fano-Elias Coding. 5.10 Competitive Optimality of the Shannon Code. 5.11 Generation of Discrete Distributions from Fair Coins. Summary. Problems. Historical Notes. 6. Gambling and Data Compression. 6.1 The Horse Race. 6.2 Gambling and Side Information. 6.3 Dependent Horse Races and Entropy Rate. 6.4 The Entropy of English. 6.5 Data Compression and Gambling. 6.6 Gambling Estimate of the Entropy of English. Summary. Problems. Historical Notes. 7. Channel Capacity. 7.1 Examples of Channel Capacity. 7.2 Symmetric Channels. 7.3 Properties of Channel Capacity. 7.4 Preview of the Channel Coding Theorem. 7.5 Definitions. 7.6 Jointly Typical Sequences. 7.7 Channel Coding Theorem. 7.8 Zero-Error Codes. 7.9 Fano's Inequality and the Converse to the Coding Theorem. 7.10 Equality in the Converse to the Channel Coding Theorem. 7.11 Hamming Codes. 7.12 Feedback Capacity. 7.13 Source-Channel Separation Theorem. Summary. Problems. Historical Notes. 8. Differential Entropy. 8.1 Definitions. 8.2 AEP for Continuous Random Variables. 8.3 Relation of Differential Entropy to Discrete Entropy. 8.4 Joint and Conditional Differential Entropy. 8.5 Relative Entropy and Mutual Information. 8.6 Properties of Differential Entropy, Relative Entropy, and Mutual Information. Summary. Problems. Historical Notes. 9. Gaussian Channel. 9.1 Gaussian Channel: Definitions. 9.2 Converse to the Coding Theorem for Gaussian Channels. 9.3 Bandlimited Channels. 9.4 Parallel Gaussian Channels. 9.5 Channels with Colored Gaussian Noise. 9.6 Gaussian Channels with Feedback. Summary. Problems. Historical Notes. 10. Rate Distortion Theory. 10.1 Quantization. 10.2 Definitions. 10.3 Calculation of the Rate Distortion Function. 10.4 Converse to the Rate Distortion Theorem. 10.5 Achievability of the Rate Distortion Function. 10.6 Strongly Typical Sequences and Rate Distortion. 10.7 Characterization of the Rate Distortion Function. 10.8 Computation of Channel Capacity and the Rate Distortion Function. Summary. Problems. Historical Notes. 11. Information Theory and Statistics. 11.1 Method of Types. 11.2 Law of Large Numbers. 11.3 Universal Source Coding. 11.4 Large Deviation Theory. 11.5 Examples of Sanov's Theorem. 11.6 Conditional Limit Theorem. 11.7 Hypothesis Testing. 11.8 Chernoff-Stein Lemma. 11.9 Chernoff Information. 11.10 Fisher Information and the Cram-er-Rao Inequality. Summary. Problems. Historical Notes. 12. Maximum Entropy. 12.1 Maximum Entropy Distributions. 12.2 Examples. 12.3 Anomalous Maximum Entropy Problem. 12.4 Spectrum Estimation. 12.5 Entropy Rates of a Gaussian Process. 12.6 Burg's Maximum Entropy Theorem. Summary. Problems. Historical Notes. 13. Universal Source Coding. 13.1 Universal Codes and Channel Capacity. 13.2 Universal Coding for Binary Sequences. 13.3 Arithmetic Coding. 13.4 Lempel-Ziv Coding. 13.5 Optimality of Lempel-Ziv Algorithms. Compression. Summary. Problems. Historical Notes. 14. Kolmogorov Complexity. 14.1 Models of Computation. 14.2 Kolmogorov Complexity: Definitions and Examples. 14.3 Kolmogorov Complexity and Entropy. 14.4 Kolmogorov Complexity of Integers. 14.5 Algorithmically Random and Incompressible Sequences. 14.6 Universal Probability. 14.7 Kolmogorov complexity. 14.9 Universal Gambling. 14.10 Occam's Razor. 14.11 Kolmogorov Complexity and Universal Probability. 14.12 Kolmogorov Sufficient Statistic. 14.13 Minimum Description Length Principle. Summary. Problems. Historical Notes. 15. Network Information Theory. 15.1 Gaussian Multiple-User Channels. 15.2 Jointly Typical Sequences. 15.3 Multiple-Access Channel. 15.4 Encoding of Correlated Sources. 15.5 Duality Between Slepian-Wolf Encoding and Multiple-Access Channels. 15.6 Broadcast Channel. 15.7 Relay Channel. 15.8 Source Coding with Side Information. 15.9 Rate Distortion with Side Information. 15.10 General Multiterminal Networks. Summary. Problems. Historical Notes. 16. Information Theory and Portfolio Theory. 16.1 The Stock Market: Some Definitions. 16.2 Kuhn-Tucker Characterization of the Log-Optimal Portfolio. 16.3 Asymptotic Optimality of the Log-Optimal Portfolio. 16.4 Side Information and the Growth Rate. 16.5 Investment in Stationary Markets. 16.6 Competitive Optimality of the Log-Optimal Portfolio. 16.7 Universal Portfolios. 16.8 Shannon-McMillan-Breiman Theorem (General AEP). Summary. Problems. Historical Notes. 17. Inequalities in Information Theory. 17.1 Basic Inequalities of Information Theory. 17.2 Differential Entropy. 17.3 Bounds on Entropy and Relative Entropy. 17.4 Inequalities for Types. 17.5 Combinatorial Bounds on Entropy. 17.6 Entropy Rates of Subsets. 17.7 Entropy and Fisher Information. 17.8 Entropy Power Inequality and Brunn-Minkowski Inequality. 17.9 Inequalities for Determinants. 17.10 Inequalities for Ratios of Determinants. Summary. Problems. Historical Notes. Bibliography. List of Symbols. Index."
                },
                {
                    "title": "Provably Secure Steganography",
                    "abstract": "Steganography is the problem of hiding secret messages in \"innocent-lookingrdquo public communication so that the presence of the secret messages cannot be detected. This paper introduces a cryptographic formalization of steganographic security in terms of computational indistinguishability from a channel, an indexed family of probability distributions on cover messages. We use cryptographic and complexity-theoretic proof techniques to show that the existence of one-way functions and the ability to sample from the channel are necessary conditions for secure steganography. We then construct a steganographic protocol, based on rejection sampling from the channel, that is provably secure and has nearly optimal bandwidth under these conditions. This is the first known example of a general provably secure steganographic protocol. We also give the first formalization of \"robustrdquo steganography, where an adversary attempts to remove any hidden messages without unduly disrupting the cover channel. We give a necessary condition on the amount of disruption the adversary is allowed in terms of a worst case measure of mutual information. We give a construction that is provably secure and computationally efficient and has nearly optimal bandwidth, assuming repeatable access to the channel distribution."
                },
                {
                    "title": "OCB: a block-cipher mode of operation for efficient authenticated encryption",
                    "abstract": "We describe a parallelizable block-cipher mode of operation that simultaneously provides privacy and authenticity. OCB encrypts-and-authenticates a nonempty string M \u03b5 {0,1}\u2022 using \\lceil |M|/n\\rceil + 2 block-cipher invocations, where n is the block length of the underlying block cipher. Additional overhead is small. OCB refines a scheme, IAPM, suggested by Charanjit Jutla. Desirable properties of OCB include: the ability to encrypt a bit string of arbitrary length into a ciphertext of minimal length; cheap offset calculations; cheap session setup; a single underlying cryptographic key; no extended-precision addition; a nearly optimal number of block-cipher calls; and no requirement for a random IV. We prove OCB secure, quantifying the adversary's ability to violate the mode's privacy or authenticity in terms of the quality of its block cipher as a pseudorandom permutation (PRP) or as a strong PRP, respectively."
                },
                {
                    "title": "Interactive team reasoning: A contribution to the theory of co-operation",
                    "abstract": "Abstract This paper describes interactions between agents who sometimes choose as individuals and sometimes as members of teams. Choosing as a member of a team entails not only being motivated by the team's objective, but also a distinctive pattern of reasoning: an agent who \u201cteam reasons\u201d computes, and chooses her component in, a profile evaluated using the team's objective function. It is not assumed that a given agent team reasons for a particular team; there may be more than one team, and which she reasons for is here treated probabilistically. Ordinary reasoning is a special case in which the team is a singleton. The framework therefore encompasses interactions paradigmatic in the theory of co-operative behaviour in which agents may choose either for the group of for themselves. The hypothesis that people may team reason if and when they are group motivated can, it is shown, explain some puzzling aspects of co-operative behaviour in a new way."
                },
                {
                    "title": "Planning, Learning and Coordination in Multiagent Decision Processes",
                    "abstract": "There has been a growing interest in AI in the design of multiagent systems, especially in multiagent cooperative planning. In this paper, we investigate the extent to which methods from single-agent planning and learning can be applied in multiagent settings. We survey a number of different techniques from decision-theoretic planning and reinforcement learning and describe a number of interesting issues that arise with regard to coordinating the policies of individual agents. To this end, we describe multiagent Markov decision processes as a general model in which to frame this discussion. These are special n-person cooperative games in which agents share the same utility function. We discuss coordination mechanisms based on imposed conventions (or social laws) as well as learning methods for coordination. Our focus is on the decomposition of sequential decision processes so that coordination can be learned (or imposed) locally, at the level of individual states. We also discuss the use of structured problem representations and their role in the generalization of learned conventions and in approximation."
                },
                {
                    "title": "Natural Language and Universal Grammar: Essays in Linguistic Theory",
                    "abstract": "John Lyons is recognized internationally as one of the most influential scholars in modern linguistics. This volume contains essays spanning many years of his thought and research, in addition to previously unpublished pieces. Chapters 2, 3 and 4 make their first appearance here, and set out the view of linguistics and linguistic theory which underlies the content of this and a (forthcoming) companion volume. The remaining six chapters have been either extensively revised or annotated to provide the reader with their historical context and to bring them into line with the author's current thinking."
                },
                {
                    "title": "Probabilistic knowledge and probabilistic common knowledge",
                    "abstract": "In this paper we develop a theory of probabilistic common knowledge and probabilistic knowledge in a group of individuals whose knowledge partitions are not wholly independent."
                },
                {
                    "title": "Learning How to Cooperate: Optimal Play in Repeated Coordination Games",
                    "abstract": "This paper proposes a characterization of optimal strategies for playing certain repeated coordination games whose players have identical preferences. Players' optimal coordination strategies reflect their uncertainty about how their partners will respond to multiple-equilibrium problems; this uncertainty constrains the statistical relationships between their strategy choices players can bring about. The authors show that optimality is nevertheless consistent with subgame-perfect equilibrium. Examples are analyzed in which players use precedents as focal points to achieve and maintain coordination, and in which they play dominated strategies with positive probability in early stages in the hope of generating a useful precedent. Copyright 1990 by The Econometric Society."
                },
                {
                    "title": "USING REASONING ABOUT KNOWLEDGE TO ANALYZE DISTRIBUTED SYSTEMS",
                    "abstract": ": The main problem unique to a distributed system is a lack of (global) knowledge. It is difficult (probably impossible) for one node to know everything about the rest of the network. Yet global knowledge seems to be required to answer questions such as \"Where is the file AT', \"Is there a deadlock?\", [or] \"What is the best way to answer the question... ?\""
                },
                {
                    "title": "How to generate and exchange secrets",
                    "abstract": "In this paper we introduce a new tool for controlling the knowledge transfer process in cryptographic protocol design. It is applied to solve a general class of problems which include most of the two-party cryptographic problems in the literature. Specifically, we show how two parties A and B can interactively generate a random integer N = p\u00bfq such that its secret, i.e., the prime factors (p, q), is hidden from either party individually but is recoverable jointly if desired. This can be utilized to give a protocol for two parties with private values i and j to compute any polynomially computable functions f(i,j) and g(i,j) with minimal knowledge transfer and a strong fairness property. As a special case, A and B can exchange a pair of secrets sA, sB, e.g. the factorization of an integer and a Hamiltonian circuit in a graph, in such a way that sA becomes computable by B when and only when sB becomes computable by A. All these results are proved assuming only that the problem of factoring large intergers is computationally intractable."
                },
                {
                    "title": "Knowledge and common knowledge in a distributed environment",
                    "abstract": "Reasoning about knowledge seems to play a fundamental role in distributed systems. Indeed, such reasoning is a central part of the informal intuitive arguments used in the design of distributed protocols. Communication in a distributed system can be viewed as the act of transforming the system's state of knowledge. This paper presents a general framework for formalizing and reasoning about knowledge in distributed systems. It is shown that states of knowledge of groups of processors are useful concepts for the design and analysis of distributed protocols. In particular, distributed knowledge corresponds to knowledge that is \u201cdistributed\u201d among the members of the group, while common knowledge corresponds to a fact being \u201cpublicly known.\u201d The relationship between common knowledge and a variety of desirable actions in a distributed system is illustrated. Furthermore, it is shown that, formally speaking, in practical systems common knowledge cannot be attained. A number of weaker variants of common knowledge that are attainable in many cases of interest are introduced and investigated."
                },
                {
                    "title": "Protocols for secure computations",
                    "abstract": "The author investigates the following problem: Suppose m people wish to compute the value of a function f(x1, x2, x3, ..., xm), which is an integer-valued function of m integer variables xi of bounded range. Assume initially person Pi knows the value of xi and no other x's. Is it possible for them to compute the value of f, by communicating among themselves, without unduly giving away any information about the values of their own variables? The author gives a precise formulation of this general problem and describe three ways of solving it by use of one-way functions (i.e., functions which are easy to evaluate but hard to invert). These results have applications to secret voting, private querying of database, oblivious negotiation, playing mental poker, etc.. He also discusses the complexity question \"How many bits need to be exchanged for the computation,\" and describes methods to prevent participants from cheating. Finally, he studies the question \"What cannot be accomplished with one-way functions.\""
                },
                {
                    "title": "A Markovian Decision Process",
                    "abstract": "Abstract : The purpose of this paper is to discuss the asymptotic behavior of the sequence (f sub n(i)) generated by a nonlinear recurrence relation. This problem arises in connection with an equipment replacement problem, cf. S. Dreyfus, A Note on an Industrial Replacement Process."
                },
                {
                    "title": "An Unsolvable Problem of Elementary Number Theory",
                    "abstract": "Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org."
                },
                {
                    "title": "Are Emergent Abilities of Large Language Models a Mirage?",
                    "abstract": "Recent work claims that large language models display emergent abilities, abilities not present in smaller-scale models that are present in larger-scale models. What makes emergent abilities intriguing is two-fold: their sharpness, transitioning seemingly instantaneously from not present to present, and their unpredictability, appearing at seemingly unforeseeable model scales. Here, we present an alternative explanation for emergent abilities: that for a particular task and model family, when analyzing fixed model outputs, emergent abilities appear due to the researcher's choice of metric rather than due to fundamental changes in model behavior with scale. Specifically, nonlinear or discontinuous metrics produce apparent emergent abilities, whereas linear or continuous metrics produce smooth, continuous predictable changes in model performance. We present our alternative explanation in a simple mathematical model, then test it in three complementary ways: we (1) make, test and confirm three predictions on the effect of metric choice using the InstructGPT/GPT-3 family on tasks with claimed emergent abilities; (2) make, test and confirm two predictions about metric choices in a meta-analysis of emergent abilities on BIG-Bench; and (3) show to choose metrics to produce never-before-seen seemingly emergent abilities in multiple vision tasks across diverse deep networks. Via all three analyses, we provide evidence that alleged emergent abilities evaporate with different metrics or with better statistics, and may not be a fundamental property of scaling AI models."
                },
                {
                    "title": "CAMEL: Communicative Agents for \"Mind\" Exploration of Large Scale Language Model Society",
                    "abstract": "The rapid advancement of conversational and chat-based language models has led to remarkable progress in complex task-solving. However, their success heavily relies on human input to guide the conversation, which can be challenging and time-consuming. This paper explores the potential of building scalable techniques to facilitate autonomous cooperation among communicative agents and provide insight into their \u201ccognitive\u201d processes. To address the challenges of achieving autonomous cooperation, we propose a novel communicative agent framework named role-playing . Our approach involves using inception prompting to guide chat agents toward task completion while maintaining consistency with human intentions. We showcase how role-playing can be used to generate conversational data for studying the behaviors and capabilities of chat agents, providing a valuable resource for investigating conversational language models. Our contributions include introducing a novel communicative agent framework, offering a scalable approach for studying the cooperative behaviors and capabilities of multi-agent systems, and open-sourcing our library to support research on communicative agents and beyond. The GitHub repository of this project is made publicly available on: https://github.com/lightaime/camel ."
                }
            ],
            "categories": [
                "cs.AI",
                "cs.CR"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively monitor and align groups of generative AI agents to prevent secret collusion and unauthorized information exchange?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for ensuring the safe deployment of generative AI systems, particularly in high-risk applications that impact human lives and involve sensitive data. By addressing the challenges of monitoring and aligning AI agents, we can advance the research community's understanding of AI safety and security, leading to the development of robust frameworks that mitigate risks associated with covert communication. This work could pave the way for practical applications in various fields, including autonomous systems, digital assistants, and financial trading, ultimately enhancing trust in AI technologies.\n\n**[Question 3] - Why is it hard?**  \nThe complexity of this problem arises from the sophisticated communication capabilities of generative AI agents, which can employ information-hiding techniques to evade monitoring. Naive approaches may fail because they do not account for the high information capacity of natural language and multi-modal communication channels, making covert exchanges difficult to detect. Additionally, technical challenges include the need for secure multi-party computation and the development of effective monitoring systems that can identify and mitigate steganographic techniques used by AI agents.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on individual AI agent behavior rather than the interactions between groups of agents, leading to a gap in understanding the dynamics of collusion. Existing solutions may lack the necessary frameworks to address the complexities of multi-agent systems and their communication strategies. Barriers such as the rapid evolution of AI capabilities and the inadequacy of current monitoring techniques have hindered progress. Our approach differs by integrating concepts from security, distributed systems, and steganography to create a comprehensive model evaluation framework that specifically targets the challenges of secret collusion among generative AI agents.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology includes a systematic model evaluation framework that focuses on the study of secret collusion among large language models. We will utilize a diverse dataset of interactions between generative AI agents and apply metrics that assess the effectiveness of monitoring and alignment strategies. Expected outcomes include novel theoretical results on the mitigation of steganography, insights into the dynamics of agent interactions, and practical guidelines for implementing monitoring systems that can detect covert communication. This work aims to enhance the safety and reliability of generative AI applications in real-world scenarios."
            }
        },
        "author_data": {
            "1ae9ab76-713b-4119-80c5-509e6cf4c18d": {
                "pk": "1ae9ab76-713b-4119-80c5-509e6cf4c18d",
                "name": "Sumeet Ramesh Motwani",
                "collaborators": [
                    "Joar Skalse",
                    "Lucy Farnik",
                    "Erik Jenner",
                    "Adam Gleave",
                    "Alessandro Abate",
                    "Andis Draguns",
                    "Andrew Gritsevskiy",
                    "Charlie Rogers-Smith",
                    "Jeffrey Ladish",
                    "Christian Schroeder de Witt"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "AI Safety",
                    "Backdoor Attacks",
                    "Reward Learning"
                ],
                "publications": [
                    {
                        "title": "STARC: A General Framework For Quantifying Differences Between Reward Functions",
                        "abstract": "In order to solve a task using reinforcement learning, it is necessary to first formalise the goal of that task as a reward function. However, for many real-world tasks, it is very difficult to manually specify a reward function that never incentivises undesirable behaviour. As a result, it is increasingly popular to use \\emph{reward learning algorithms}, which attempt to \\emph{learn} a reward function from data. However, the theoretical foundations of reward learning are not yet well-developed. In particular, it is typically not known when a given reward learning algorithm with high probability will learn a reward function that is safe to optimise. This means that reward learning algorithms generally must be evaluated empirically, which is expensive, and that their failure modes are difficult to anticipate in advance. One of the roadblocks to deriving better theoretical guarantees is the lack of good methods for quantifying the difference between reward functions. In this paper we provide a solution to this problem, in the form of a class of pseudometrics on the space of all reward functions that we call STARC (STAndardised Reward Comparison) metrics. We show that STARC metrics induce both an upper and a lower bound on worst-case regret, which implies that our metrics are tight, and that any metric with the same properties must be bilipschitz equivalent to ours. Moreover, we also identify a number of issues with reward metrics proposed by earlier works. Finally, we evaluate our metrics empirically, to demonstrate their practical efficacy. STARC metrics can be used to make both theoretical and empirical analysis of reward learning algorithms both easier and more principled."
                    },
                    {
                        "title": "Unelicitable Backdoors in Language Models via Cryptographic Transformer Circuits",
                        "abstract": "The rapid proliferation of open-source language models significantly increases the risks of downstream backdoor attacks. These backdoors can introduce dangerous behaviours during model deployment and can evade detection by conventional cybersecurity monitoring systems. In this paper, we introduce a novel class of backdoors in autoregressive transformer models, that, in contrast to prior art, are unelicitable in nature. Unelicitability prevents the defender from triggering the backdoor, making it impossible to evaluate or detect ahead of deployment even if given full white-box access and using automated techniques, such as red-teaming or certain formal verification methods. We show that our novel construction is not only unelicitable thanks to using cryptographic techniques, but also has favourable robustness properties. We confirm these properties in empirical investigations, and provide evidence that our backdoors can withstand state-of-the-art mitigation strategies. Additionally, we expand on previous work by showing that our universal backdoors, while not completely undetectable in white-box settings, can be harder to detect than some existing designs. By demonstrating the feasibility of seamlessly integrating backdoors into transformer models, this paper fundamentally questions the efficacy of pre-deployment detection strategies. This offers new insights into the offence-defence balance in AI safety and security."
                    }
                ]
            },
            "3b750f0b-61ce-4cd2-afbe-ab105ad28d09": {
                "pk": "3b750f0b-61ce-4cd2-afbe-ab105ad28d09",
                "name": "Mikhail Baranchuk",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "c1c2c64c-c5b7-41e0-8a36-85bed4380ff7": {
                "pk": "c1c2c64c-c5b7-41e0-8a36-85bed4380ff7",
                "name": "Martin Strohmeier",
                "collaborators": [
                    "Ivan Martinovic",
                    "Vincent Lenders",
                    "Sebastian K\u00f6hler",
                    "Matthew Smith",
                    "Matthias Sch\u00e4fer",
                    "Joshua Smailes",
                    "Simon Birnbach",
                    "Mauro Leonardi",
                    "Rui Pinheiro",
                    "Marcell Szak\u00e1ly"
                ],
                "domain": [
                    "Wireless Security",
                    "Localization",
                    "Adversarial Machine Learning",
                    "Satellite Communication"
                ],
                "publications": [
                    {
                        "title": "On the Security of the Automatic Dependent Surveillance-Broadcast Protocol",
                        "abstract": "Automatic dependent surveillance-broadcast (ADS-B) is the communications protocol currently being rolled out as part of next generation air transportation systems. As the heart of modern air traffic control, it will play an essential role in the protection of two billion passengers per year, besides being crucial to many other interest groups in aviation. The inherent lack of security measures in the ADS-B protocol has long been a topic in both the aviation circles and in the academic community. Due to recently published proof-of-concept attacks, the topic is becoming ever more pressing, especially with the deadline for mandatory implementation in most airspaces fast approaching.   This survey first summarizes the attacks and problems that have been reported in relation to ADS-B security. Thereafter, it surveys both the theoretical and practical efforts which have been previously conducted concerning these issues, including possible countermeasures. In addition, the survey seeks to go beyond the current state of the art and gives a detailed assessment of security measures which have been developed more generally for related wireless networks such as sensor networks and vehicular ad hoc networks, including a taxonomy of all considered approaches."
                    },
                    {
                        "title": "LocaRDS: A Localization Reference Data Set",
                        "abstract": "The use of wireless signals for purposes of localization enables a host of applications relating to the determination and verification of the positions of network participants, ranging from radar to satellite navigation. Consequently, it has been a longstanding interest of theoretical and practical research in mobile networks and many solutions have been proposed in the scientific literature. However, it is hard to assess the performance of these in the real world and, more severely, to compare their advantages and disadvantages in a controlled scientific manner.   With this work, we attempt to improve the current state of the art in localization research and put it on a solid scientific grounding for the future. Concretely, we develop LocaRDS, an open reference dataset of real-world crowdsourced flight data featuring more than 222 million measurements from over 50 million transmissions recorded by 323 sensors. We show how LocaRDS can be used to test, analyze and directly compare different localization techniques and further demonstrate its effectiveness by examining the open question of the aircraft localization problem in crowdsourced sensor networks. Finally, we provide a working reference implementation for the aircraft localization problem and a discussion of possible metrics for use with LocaRDS."
                    },
                    {
                        "title": "Classi-Fly: Inferring Aircraft Categories from Open Data using Machine Learning",
                        "abstract": "In recent years, air traffic communication data has become easy to access, enabling novel research in many fields. Exploiting this new data source, a wide range of applications have emerged, from weather forecasting to stock market prediction, or the collection of information about military and government movements. Typically these applications require knowledge about the metadata of the aircraft, specifically its operator and the aircraft category.   armasuisse Science + Technology, the R\\&D agency for the Swiss Armed Forces, has been developing Classi-Fly, a novel approach to obtain metadata about aircraft based on their movement patterns. We validate Classi-Fly using several hundred thousand flights collected through open source means, in conjunction with ground truth from publicly available aircraft registries containing more than two million aircraft. Classi-Fly obtains the correct aircraft category with an accuracy of over 88%, demonstrating that it can improve the meta data necessary for applications working with air traffic communication. Finally, we show that it is feasible to automatically detect specific flights such as police and surveillance missions."
                    },
                    {
                        "title": "A Localization Approach for Crowdsourced Air Traffic Communication Networks",
                        "abstract": "In this work, we argue that current state-of-the-art methods of aircraft localization such as multilateration are insufficient, in particular for modern crowdsourced air traffic networks with random, unplanned deployment geometry. We propose an alternative, a grid-based localization approach using the k-Nearest Neighbor algorithm, to deal with the identified shortcomings. Our proposal does not require any changes to the existing air traffic protocols and transmitters, and is easily implemented using only low-cost, commercial-off-the-shelf hardware.   Using an algebraic multilateration algorithm for comparison, we evaluate our approach using real-world flight data collected with our collaborative sensor network OpenSky. We quantify its effectiveness in terms of aircraft location accuracy, surveillance coverage, and the verification of false position data. Our results show that the grid-based approach can increase the effective air traffic surveillance coverage compared to multilateration by a factor of up to 2.5. As it does not suffer from dilution of precision, it is much more robust in noisy environments and performs better in pre-existing, unplanned receiver deployments.   We further find that the mean aircraft location accuracy can be increased by up to 41% in comparison with multilateration while also being able to pinpoint the origin of potential spoofing attacks conducted from the ground."
                    },
                    {
                        "title": "On Perception and Reality in Wireless Air Traffic Communications Security",
                        "abstract": "More than a dozen wireless technologies are used by air traffic communication systems during different flight phases. From a conceptual perspective, all of them are insecure as security was never part of their design. Recent contributions from academic and hacking communities have exploited this inherent vulnerability to demonstrate attacks on some of these technologies. However, not all of these contributions have resonated widely within aviation circles. At the same time, the security community lacks certain aviation domain knowledge, preventing aviation authorities from giving credence to their findings.   In this paper, we aim to reconcile the view of the security community and the perspective of aviation professionals concerning the safety of air traffic communication technologies. To achieve this, we first provide a systematization of the applications of wireless technologies upon which civil aviation relies. Based on these applications, we comprehensively analyze vulnerabilities, attacks, and countermeasures. We categorize the existing research on countermeasures into approaches that are applicable in the short term and research of secure new technologies deployable in the long term.   Since not all of the required aviation knowledge is codified in academic publications, we additionally examine existing aviation standards and survey 242 international aviation experts. Besides their domain knowledge, we also analyze the awareness of members of the aviation community concerning the security of wireless systems and collect their expert opinions on the potential impact of concrete attack scenarios using these technologies."
                    },
                    {
                        "title": "Assault and Battery: Evaluating the Security of Power Conversion Systems Against Electromagnetic Injection Attacks",
                        "abstract": "Many modern devices, including critical infrastructures, depend on the reliable operation of electrical power conversion systems. The small size and versatility of switched-mode power converters has resulted in their widespread adoption. Whereas transformer-based systems passively convert voltage, switched-mode converters feature an actively regulated feedback loop, which relies on accurate sensor measurements. Previous academic work has shown that many types of sensors are vulnerable to Intentional Electromagnetic Interference (IEMI) attacks, and it has been postulated that power converters, too, are affected.   In this paper, we present the first detailed study on switched-mode power converters by targeting their voltage and current sensors through IEMI attacks. We present a theoretical framework for evaluating IEMI attacks against feedback-based power supplies in the general case. We experimentally validate our theoretical predictions by analyzing multiple AC-DC and DC-DC converters, automotive grade current sensors, and dedicated battery chargers, and demonstrate the systematic vulnerability of all examined categories under real-world conditions. Finally, we demonstrate that sensor attacks on power converters can cause permanent damage to Li-Ion batteries during the charging process."
                    },
                    {
                        "title": "Neural Exec: Learning (and Learning from) Execution Triggers for Prompt Injection Attacks",
                        "abstract": "We introduce a new family of prompt injection attacks, termed Neural Exec. Unlike known attacks that rely on handcrafted strings (e.g., \"Ignore previous instructions and...\"), we show that it is possible to conceptualize the creation of execution triggers as a differentiable search problem and use learning-based methods to autonomously generate them.   Our results demonstrate that a motivated adversary can forge triggers that are not only drastically more effective than current handcrafted ones but also exhibit inherent flexibility in shape, properties, and functionality. In this direction, we show that an attacker can design and generate Neural Execs capable of persisting through multi-stage preprocessing pipelines, such as in the case of Retrieval-Augmented Generation (RAG)-based applications. More critically, our findings show that attackers can produce triggers that deviate markedly in form and shape from any known attack, sidestepping existing blacklist-based detection and sanitation approaches."
                    },
                    {
                        "title": "Understanding Realistic Attacks on Airborne Collision Avoidance Systems",
                        "abstract": "Airborne collision avoidance systems provide an onboard safety net should normal air traffic control procedures fail to keep aircraft separated. These systems are widely deployed and have been constantly refined over the past three decades, usually in response to near misses or mid-air collisions. Recent years have seen security research increasingly focus on aviation, identifying that key wireless links---some of which are used in collision avoidance---are vulnerable to attack. In this paper, we go one step further to understand whether an attacker can remotely trigger false collision avoidance alarms. Primarily considering the next-generation Airborne Collision Avoidance System X (ACAS X), we adopt a modelling approach to extract attacker constraints from technical standards before simulating collision avoidance attacks against standardized ACAS X code. We find that in 44% of cases, an attacker can successfully trigger a collision avoidance alert which on average results in a 590 ft altitude deviation; when the aircraft is at lower altitudes, this success rate rises considerably to 79%. Furthermore, we show how our simulation approach can be used to help defend against attacks by identifying where attackers are most likely to be successful."
                    },
                    {
                        "title": "Improving Aircraft Localization: Experiences and Lessons Learned from an Open Competition",
                        "abstract": "Knowledge about the exact positioning of aircraft is crucial in many settings. Consequently, the opportunistic and independent localization of aircraft based on their communication has been a longstanding problem and subject of much research. Originating from military settings, the capability to conduct aircraft localization has moved first towards the institutional civil aviation domain and can now be undertaken by anyone who has access to multiple cheap software-defined radios. Based on these technological developments, many crowdsourced sensor networks have sprung up, which collect air traffic control data in order to localize aircraft and visualize the airspace. Due to their unplanned and uncontrolled deployment and heterogeneous receiver technology traditional solutions to the Aircraft Localization Problem (ALP) can either not be applied or do not perform in a satisfactory manner. In order to deal with this issue and to find novel approaches to the ALP itself, we have designed and executed a multi-stage open competition, conducted both offline and online.   In this paper, we discuss the setup, experiences, and lessons learned from this Aircraft Localization Competition. We report from a diverse set of technical approaches, comprising 72 participating teams over three stages. The participants reached a localization accuracy of up to 25 meters in a setting with fully GPS-synchronized receivers and 78 meters in a largely unsynchronized receiver setting. These results constitute a significant improvement over the previous baseline used in the OpenSky research network.   We compare the results of the study, discuss the current state of the art, and highlight the areas that, based on our experience from organizing a competition, need to be improved for optimal adoption of the competitive approach for other scenarios."
                    },
                    {
                        "title": "QPEP: A QUIC-Based Approach to Encrypted Performance Enhancing Proxies for High-Latency Satellite Broadband",
                        "abstract": "Satellite broadband services are critical infrastructures enabling advanced technologies to function in the most remote regions of the globe. However, status-quo services are often unencrypted by default and vulnerable to eavesdropping attacks. In this paper, we challenge the historical perception that over-the-air security must trade off with TCP performance in high-latency satellite networks due to the deep-packet inspection requirements of Performance Enhancing Proxies (PEPs).   After considering why prior work in this area has failed to find wide adoption, we present an open-source encrypted-by-default PEP - QPEP - which seeks to address these issues. QPEP is built around the open QUIC standard and designed so individual customers may adopt it without ISP involvement. QPEP's performance is assessed through simulations in a replicable docker-based testbed. Across many benchmarks and network conditions, QPEP is found to avoid the perceived security-encryption trade-off in PEP design. Compared to unencrypted PEP implementations, QPEP reduces average page load times by more than 30% while also offering over-the-air privacy. Compared to the traditional VPN encryption available to customers today, QPEP more than halves average page load times. Together, these experiments lead to the conclusion that QPEP represents a promising new approach to protecting modern satellite broadband connections."
                    },
                    {
                        "title": "Safety vs. Security: Attacking Avionic Systems with Humans in the Loop",
                        "abstract": "Many wireless communications systems found in aircraft lack standard security mechanisms, leaving them fundamentally vulnerable to attack. With affordable software-defined radios available, a novel threat has emerged, allowing a wide range of attackers to easily interfere with wireless avionic systems. Whilst these vulnerabilities are known, concrete attacks that exploit them are still novel and not yet well understood. This is true in particular with regards to their kinetic impact on the handling of the attacked aircraft and consequently its safety.   To investigate this, we invited 30 Airbus A320 type-rated pilots to fly simulator scenarios in which they were subjected to attacks on their avionics. We implement and analyse novel wireless attacks on three safety-related systems: Traffic Collision Avoidance System (TCAS), Ground Proximity Warning System (GPWS) and the Instrument Landing System (ILS).   We found that all three analysed attack scenarios created significant control impact and cost of disruption through turnarounds, avoidance manoeuvres, and diversions. They further increased workload, distrust in the affected system, and in 38% of cases caused the attacked safety system to be switched off entirely. All pilots felt the scenarios were useful, with 93.3% feeling that simulator training for wireless attacks could be valuable."
                    },
                    {
                        "title": "Analyzing Privacy Breaches in the Aircraft Communications Addressing and Reporting System (ACARS)",
                        "abstract": "The manner in which Aircraft Communications, Addressing and Reporting System (ACARS) is being used has significantly changed over time. Whilst originally used by commercial airliners to track their flights and provide automated timekeeping on crew, today it serves as a multi-purpose air-ground data link for many aviation stakeholders including private jet owners, state actors and military. Since ACARS messages are still mostly sent in the clear over a wireless channel, any sensitive information sent with ACARS can potentially lead to a privacy breach for users. Naturally, different stakeholders consider different types of data sensitive. In this paper we propose a privacy framework matching aviation stakeholders to a range of sensitive information types and assess the impact for each. Based on more than one million ACARS messages, collected over several months, we then demonstrate that current ACARS usage systematically breaches privacy for all stakeholder groups. We further support our findings with a number of cases of significant privacy issues for each group and analyze the impact of such leaks. While it is well-known that ACARS messages are susceptible to eavesdropping attacks, this work is the first to quantify the extent and impact of privacy leakage in the real world for the relevant aviation stakeholders."
                    },
                    {
                        "title": "KeySpace: Public Key Infrastructure Considerations in Interplanetary Networks",
                        "abstract": "As satellite networks grow larger and begin to incorporate interplanetary communication, there is an increasing interest in the unsolved problem of how to approach PKI in these conditions. In this paper we explore the goals and requirements for implementing key management systems in satellite networks, focusing on megaconstellations and interplanetary networks. We design a set of standardized experiments which can be used to compare systems against one another for particular network topologies. Using these, we demonstrate that terrestrial PKI techniques are feasible in highly distributed interplanetary networks, showing that it is possible to configure PKI systems to achieve efficient low-latency connection establishment, and minimize the impact of attacks through effective revocations.   We evaluate this by building the Deep Space Network Simulator (DSNS), a novel network simulator aimed at efficient simulation of large space networks. We run simulations evaluating connection establishment and key revocation under a wide range of PKI configurations. Finally, we propose and evaluate two additional configuration options: OCSP Hybrid, and the use of relay nodes as a firewall. Together these minimize the extent of the network an attacker can reach with a compromised key, and reduce the attacker's load on interplanetary relay links."
                    },
                    {
                        "title": "Brokenwire : Wireless Disruption of CCS Electric Vehicle Charging",
                        "abstract": "We present a novel attack against the Combined Charging System, one of the most widely used DC rapid charging technologies for electric vehicles (EVs). Our attack, Brokenwire, interrupts necessary control communication between the vehicle and charger, causing charging sessions to abort. The attack requires only temporary physical proximity and can be conducted wirelessly from a distance, allowing individual vehicles or entire fleets to be disrupted stealthily and simultaneously. In addition, it can be mounted with off-the-shelf radio hardware and minimal technical knowledge. By exploiting CSMA/CA behavior, only a very weak signal needs to be induced into the victim to disrupt communication - exceeding the effectiveness of broadband noise jamming by three orders of magnitude. The exploited behavior is a required part of the HomePlug Green PHY, DIN 70121 & ISO 15118 standards and all known implementations exhibit it. We first study the attack in a controlled testbed and then demonstrate it against eight vehicles and 20 chargers in real deployments. We find the attack to be successful in the real world, at ranges up to 47 m, for a power budget of less than 1 W. We further show that the attack can work between the floors of a building (e.g., multi-story parking), through perimeter fences, and from `drive-by' attacks. We present a heuristic model to estimate the number of vehicles that can be attacked simultaneously for a given output power. Brokenwire has immediate implications for a substantial proportion of the around 12 million battery EVs on the roads worldwide - and profound effects on the new wave of electrification for vehicle fleets, both for private enterprise and crucial public services, as well as electric buses, trucks and small ships. As such, we conducted a disclosure to the industry and discussed a range of mitigation techniques that could be deployed to limit the impact."
                    },
                    {
                        "title": "FABRID: Flexible Attestation-Based Routing for Inter-Domain Networks",
                        "abstract": "In its current state, the Internet does not provide end users with transparency and control regarding on-path forwarding devices. In particular, the lack of network device information reduces the trustworthiness of the forwarding path and prevents end-user applications requiring specific router capabilities from reaching their full potential. Moreover, the inability to influence the traffic's forwarding path results in applications communicating over undesired routes, while alternative paths with more desirable properties remain unusable.   In this work, we present FABRID, a system that enables applications to forward traffic flexibly, potentially on multiple paths selected to comply with user-defined preferences, where information about forwarding devices is exposed and transparently attested by autonomous systems (ASes). The granularity of this information is chosen by each AS individually, protecting them from leaking sensitive network details, while the secrecy and authenticity of preferences embedded within the users' packets are protected through efficient cryptographic operations. We show the viability of FABRID by deploying it on a global SCION network test bed, and we demonstrate high throughput on commodity hardware."
                    },
                    {
                        "title": "Sticky Fingers: Resilience of Satellite Fingerprinting against Jamming Attacks",
                        "abstract": "In the wake of increasing numbers of attacks on radio communication systems, a range of techniques are being deployed to increase the security of these systems. One such technique is radio fingerprinting, in which the transmitter can be identified and authenticated by observing small hardware differences expressed in the signal. Fingerprinting has been explored in particular in the defense of satellite systems, many of which are insecure and cannot be retrofitted with cryptographic security.   In this paper, we evaluate the effectiveness of radio fingerprinting techniques under interference and jamming attacks, usually intended to deny service. By taking a pre-trained fingerprinting model and gathering a new dataset in which different levels of Gaussian noise and tone jamming have been added to the legitimate signal, we assess the attacker power required in order to disrupt the transmitter fingerprint such that it can no longer be recognized. We compare this to Gaussian jamming on the data portion of the signal, obtaining the remarkable result that transmitter fingerprints are still recognizable even in the presence of moderate levels of noise. Through deeper analysis of the results, we conclude that it takes a similar amount of jamming power in order to disrupt the fingerprint as it does to jam the message contents itself, so it is safe to include a fingerprinting system to authenticate satellite communication without opening up the system to easier denial-of-service attacks."
                    },
                    {
                        "title": "SLAP: Improving Physical Adversarial Examples with Short-Lived Adversarial Perturbations",
                        "abstract": "Research into adversarial examples (AE) has developed rapidly, yet static adversarial patches are still the main technique for conducting attacks in the real world, despite being obvious, semi-permanent and unmodifiable once deployed.   In this paper, we propose Short-Lived Adversarial Perturbations (SLAP), a novel technique that allows adversaries to realize physically robust real-world AE by using a light projector. Attackers can project a specifically crafted adversarial perturbation onto a real-world object, transforming it into an AE. This allows the adversary greater control over the attack compared to adversarial patches: (i) projections can be dynamically turned on and off or modified at will, (ii) projections do not suffer from the locality constraint imposed by patches, making them harder to detect.   We study the feasibility of SLAP in the self-driving scenario, targeting both object detector and traffic sign recognition tasks, focusing on the detection of stop signs. We conduct experiments in a variety of ambient light conditions, including outdoors, showing how in non-bright settings the proposed method generates AE that are extremely robust, causing misclassifications on state-of-the-art networks with up to 99% success rate for a variety of angles and distances. We also demostrate that SLAP-generated AE do not present detectable behaviours seen in adversarial patches and therefore bypass SentiNet, a physical AE detection method. We evaluate other defences including an adaptive defender using adversarial learning which is able to thwart the attack effectiveness up to 80% even in favourable attacker conditions."
                    },
                    {
                        "title": "Watch This Space: Securing Satellite Communication through Resilient Transmitter Fingerprinting",
                        "abstract": "Due to an increase in the availability of cheap off-the-shelf radio hardware, spoofing and replay attacks on satellite ground systems have become more accessible than ever. This is particularly a problem for legacy systems, many of which do not offer cryptographic security and cannot be patched to support novel security measures.   In this paper we explore radio transmitter fingerprinting in satellite systems. We introduce the SatIQ system, proposing novel techniques for authenticating transmissions using characteristics of transmitter hardware expressed as impairments on the downlinked signal. We look in particular at high sample rate fingerprinting, making fingerprints difficult to forge without similarly high sample rate transmitting hardware, thus raising the budget for attacks. We also examine the difficulty of this approach with high levels of atmospheric noise and multipath scattering, and analyze potential solutions to this problem.   We focus on the Iridium satellite constellation, for which we collected 1705202 messages at a sample rate of 25 MS/s. We use this data to train a fingerprinting model consisting of an autoencoder combined with a Siamese neural network, enabling the model to learn an efficient encoding of message headers that preserves identifying information.   We demonstrate the system's robustness under attack by replaying messages using a Software-Defined Radio, achieving an Equal Error Rate of 0.120, and ROC AUC of 0.946. Finally, we analyze its stability over time by introducing a time gap between training and testing data, and its extensibility by introducing new transmitters which have not been seen before. We conclude that our techniques are useful for building systems that are stable over time, can be used immediately with new transmitters without retraining, and provide robustness against spoofing and replay by raising the required budget for attacks."
                    }
                ]
            },
            "3f4ac40b-4491-42a0-85f4-f2c301495b93": {
                "pk": "3f4ac40b-4491-42a0-85f4-f2c301495b93",
                "name": "Vijay Bolina",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "8c6c5926-6ecc-46fe-a7e9-0d54d04c11e0": {
                "pk": "8c6c5926-6ecc-46fe-a7e9-0d54d04c11e0",
                "name": "Philip H. S. Torr",
                "collaborators": [
                    "Aleksandar Petrov",
                    "Adel Bibi",
                    "Anurag Arnab",
                    "Puneet K. Dokania",
                    "Qizhu Li",
                    "Yuge Shi",
                    "Brooks Paige",
                    "N. Siddharth",
                    "Ziming Zhang",
                    "Bernardino Romera-Paredes"
                ],
                "domain": [
                    "Computer Vision",
                    "Instance Segmentation",
                    "Generative Models",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Object Proposal Generation using Two-Stage Cascade SVMs",
                        "abstract": "Object proposal algorithms have shown great promise as a first step for object recognition and detection. Good object proposal generation algorithms require high object recall rate as well as low computational cost, because generating object proposals is usually utilized as a preprocessing step. The problem of how to accelerate the object proposal generation and evaluation process without decreasing recall is thus of great interest. In this paper, we propose a new object proposal generation method using two-stage cascade SVMs, where in the first stage linear filters are learned for predefined quantized scales/aspect-ratios independently, and in the second stage a global linear classifier is learned across all the quantized scales/aspect-ratios for calibration, so that all the proposals can be compared properly. The proposals with highest scores are our final output. Specifically, we explain our scale/aspect-ratio quantization scheme, and investigate the effects of combinations of $\\ell_1$ and $\\ell_2$ regularizers in cascade SVMs with/without ranking constraints in learning. Comprehensive experiments on VOC2007 dataset are conducted, and our results achieve the state-of-the-art performance with high object recall rate and high computational efficiency. Besides, our method has been demonstrated to be suitable for not only class-specific but also generic object proposal generation."
                    },
                    {
                        "title": "Recurrent Instance Segmentation",
                        "abstract": "Instance segmentation is the problem of detecting and delineating each distinct object of interest appearing in an image. Current instance segmentation approaches consist of ensembles of modules that are trained independently of each other, thus missing opportunities for joint learning. Here we propose a new instance segmentation paradigm consisting in an end-to-end method that learns how to segment instances sequentially. The model is based on a recurrent neural network that sequentially finds objects and their segmentations one at a time. This net is provided with a spatial memory that keeps track of what pixels have been explained and allows occlusion handling. In order to train the model we designed a principled loss function that accurately represents the properties of the instance segmentation problem. In the experiments carried out, we found that our method outperforms recent approaches on multiple person segmentation, and all state of the art approaches on the Plant Phenotyping dataset for leaf counting."
                    },
                    {
                        "title": "Bottom-up Instance Segmentation using Deep Higher-Order CRFs",
                        "abstract": "Traditional Scene Understanding problems such as Object Detection and Semantic Segmentation have made breakthroughs in recent years due to the adoption of deep learning. However, the former task is not able to localise objects at a pixel level, and the latter task has no notion of different instances of objects of the same class. We focus on the task of Instance Segmentation which recognises and localises objects down to a pixel level. Our model is based on a deep neural network trained for semantic segmentation. This network incorporates a Conditional Random Field with end-to-end trainable higher order potentials based on object detector outputs. This allows us to reason about instances from an initial, category-level semantic segmentation. Our simple method effectively leverages the great progress recently made in semantic segmentation and object detection. The accurate instance-level segmentations that our network produces is reflected by the considerable improvements obtained over previous work."
                    },
                    {
                        "title": "Traditional Classification Neural Networks are Good Generators: They are Competitive with DDPMs and GANs",
                        "abstract": "Classifiers and generators have long been separated. We break down this separation and showcase that conventional neural network classifiers can generate high-quality images of a large number of categories, being comparable to the state-of-the-art generative models (e.g., DDPMs and GANs). We achieve this by computing the partial derivative of the classification loss function with respect to the input to optimize the input to produce an image. Since it is widely known that directly optimizing the inputs is similar to targeted adversarial attacks incapable of generating human-meaningful images, we propose a mask-based stochastic reconstruction module to make the gradients semantic-aware to synthesize plausible images. We further propose a progressive-resolution technique to guarantee fidelity, which produces photorealistic images. Furthermore, we introduce a distance metric loss and a non-trivial distribution loss to ensure classification neural networks can synthesize diverse and high-fidelity images. Using traditional neural network classifiers, we can generate good-quality images of 256$\\times$256 resolution on ImageNet. Intriguingly, our method is also applicable to text-to-image generation by regarding image-text foundation models as generalized classifiers.   Proving that classifiers have learned the data distribution and are ready for image generation has far-reaching implications, for classifiers are much easier to train than generative models like DDPMs and GANs. We don't even need to train classification models because tons of public ones are available for download. Also, this holds great potential for the interpretability and robustness of classifiers. Project page is at \\url{https://classifier-as-generator.github.io/}."
                    },
                    {
                        "title": "Feature sampling and partitioning for visual vocabulary generation on large action classification datasets",
                        "abstract": "The recent trend in action recognition is towards larger datasets, an increasing number of action classes and larger visual vocabularies. State-of-the-art human action classification in challenging video data is currently based on a bag-of-visual-words pipeline in which space-time features are aggregated globally to form a histogram. The strategies chosen to sample features and construct a visual vocabulary are critical to performance, in fact often dominating performance. In this work we provide a critical evaluation of various approaches to building a vocabulary and show that good practises do have a significant impact. By subsampling and partitioning features strategically, we are able to achieve state-of-the-art results on 5 major action recognition datasets using relatively small visual vocabularies."
                    },
                    {
                        "title": "Discovering Class-Specific Pixels for Weakly-Supervised Semantic Segmentation",
                        "abstract": "We propose an approach to discover class-specific pixels for the weakly-supervised semantic segmentation task. We show that properly combining saliency and attention maps allows us to obtain reliable cues capable of significantly boosting the performance. First, we propose a simple yet powerful hierarchical approach to discover the class-agnostic salient regions, obtained using a salient object detector, which otherwise would be ignored. Second, we use fully convolutional attention maps to reliably localize the class-specific regions in a given image. We combine these two cues to discover class-specific pixels which are then used as an approximate ground truth for training a CNN. While solving the weakly supervised semantic segmentation task, we ensure that the image-level classification task is also solved in order to enforce the CNN to assign at least one pixel to each object present in the image. Experimentally, on the PASCAL VOC12 val and test sets, we obtain the mIoU of 60.8% and 61.9%, achieving the performance gains of 5.1% and 5.2% compared to the published state-of-the-art results. The code is made publicly available."
                    },
                    {
                        "title": "Weakly- and Semi-Supervised Panoptic Segmentation",
                        "abstract": "We present a weakly supervised model that jointly performs both semantic- and instance-segmentation -- a particularly relevant problem given the substantial cost of obtaining pixel-perfect annotation for these tasks. In contrast to many popular instance segmentation approaches based on object detectors, our method does not predict any overlapping instances. Moreover, we are able to segment both \"thing\" and \"stuff\" classes, and thus explain all the pixels in the image. \"Thing\" classes are weakly-supervised with bounding boxes, and \"stuff\" with image-level tags. We obtain state-of-the-art results on Pascal VOC, for both full and weak supervision (which achieves about 95% of fully-supervised performance). Furthermore, we present the first weakly-supervised results on Cityscapes for both semantic- and instance-segmentation. Finally, we use our weakly supervised framework to analyse the relationship between annotation quality and predictive performance, which is of interest to dataset creators."
                    },
                    {
                        "title": "Hypergraph Convolution and Hypergraph Attention",
                        "abstract": "Recently, graph neural networks have attracted great attention and achieved prominent performance in various research fields. Most of those algorithms have assumed pairwise relationships of objects of interest. However, in many real applications, the relationships between objects are in higher-order, beyond a pairwise formulation. To efficiently learn deep embeddings on the high-order graph-structured data, we introduce two end-to-end trainable operators to the family of graph neural networks, i.e., hypergraph convolution and hypergraph attention. Whilst hypergraph convolution defines the basic formulation of performing convolution on a hypergraph, hypergraph attention further enhances the capacity of representation learning by leveraging an attention module. With the two operators, a graph neural network is readily extended to a more flexible model and applied to diverse applications where non-pairwise relationships are observed. Extensive experimental results with semi-supervised node classification demonstrate the effectiveness of hypergraph convolution and hypergraph attention."
                    },
                    {
                        "title": "Relating by Contrasting: A Data-efficient Framework for Multimodal Generative Models",
                        "abstract": "Multimodal learning for generative models often refers to the learning of abstract concepts from the commonality of information in multiple modalities, such as vision and language. While it has proven effective for learning generalisable representations, the training of such models often requires a large amount of \"related\" multimodal data that shares commonality, which can be expensive to come by. To mitigate this, we develop a novel contrastive framework for generative model learning, allowing us to train the model not just by the commonality between modalities, but by the distinction between \"related\" and \"unrelated\" multimodal data. We show in experiments that our method enables data-efficient multimodal learning on challenging datasets for various multimodal VAE models. We also show that under our proposed framework, the generative model can accurately identify related samples from unrelated ones, making it possible to make use of the plentiful unlabeled, unpaired multimodal data."
                    },
                    {
                        "title": "An Impartial Take to the CNN vs Transformer Robustness Contest",
                        "abstract": "Following the surge of popularity of Transformers in Computer Vision, several studies have attempted to determine whether they could be more robust to distribution shifts and provide better uncertainty estimates than Convolutional Neural Networks (CNNs). The almost unanimous conclusion is that they are, and it is often conjectured more or less explicitly that the reason of this supposed superiority is to be attributed to the self-attention mechanism. In this paper we perform extensive empirical analyses showing that recent state-of-the-art CNNs (particularly, ConvNeXt) can be as robust and reliable or even sometimes more than the current state-of-the-art Transformers. However, there is no clear winner. Therefore, although it is tempting to state the definitive superiority of one family of architectures over another, they seem to enjoy similar extraordinary performances on a variety of tasks while also suffering from similar vulnerabilities such as texture, background, and simplicity biases."
                    },
                    {
                        "title": "Memory-Driven Text-to-Image Generation",
                        "abstract": "We introduce a memory-driven semi-parametric approach to text-to-image generation, which is based on both parametric and non-parametric techniques. The non-parametric component is a memory bank of image features constructed from a training set of images. The parametric component is a generative adversarial network. Given a new text description at inference time, the memory bank is used to selectively retrieve image features that are provided as basic information of target images, which enables the generator to produce realistic synthetic results. We also incorporate the content information into the discriminator, together with semantic features, allowing the discriminator to make a more reliable prediction. Experimental results demonstrate that the proposed memory-driven semi-parametric approach produces more realistic images than purely parametric approaches, in terms of both visual fidelity and text-image semantic consistency."
                    },
                    {
                        "title": "Learn what matters: cross-domain imitation learning with task-relevant embeddings",
                        "abstract": "We study how an autonomous agent learns to perform a task from demonstrations in a different domain, such as a different environment or different agent. Such cross-domain imitation learning is required to, for example, train an artificial agent from demonstrations of a human expert. We propose a scalable framework that enables cross-domain imitation learning without access to additional demonstrations or further domain knowledge. We jointly train the learner agent's policy and learn a mapping between the learner and expert domains with adversarial training. We effect this by using a mutual information criterion to find an embedding of the expert's state space that contains task-relevant information and is invariant to domain specifics. This step significantly simplifies estimating the mapping between the learner and expert domains and hence facilitates end-to-end learning. We demonstrate successful transfer of policies between considerably different domains, without extra supervision such as additional demonstrations, and in situations where other methods fail."
                    },
                    {
                        "title": "When Do Prompting and Prefix-Tuning Work? A Theory of Capabilities and Limitations",
                        "abstract": "Context-based fine-tuning methods, including prompting, in-context learning, soft prompting (also known as prompt tuning), and prefix-tuning, have gained popularity due to their ability to often match the performance of full fine-tuning with a fraction of the parameters. Despite their empirical successes, there is little theoretical understanding of how these techniques influence the internal computation of the model and their expressiveness limitations. We show that despite the continuous embedding space being more expressive than the discrete token space, soft-prompting and prefix-tuning are potentially less expressive than full fine-tuning, even with the same number of learnable parameters. Concretely, context-based fine-tuning cannot change the relative attention pattern over the content and can only bias the outputs of an attention layer in a fixed direction. This suggests that while techniques like prompting, in-context learning, soft prompting, and prefix-tuning can effectively elicit skills present in the pretrained model, they may not be able to learn novel tasks that require new attention patterns."
                    },
                    {
                        "title": "Prompting a Pretrained Transformer Can Be a Universal Approximator",
                        "abstract": "Despite the widespread adoption of prompting, prompt tuning and prefix-tuning of transformer models, our theoretical understanding of these fine-tuning methods remains limited. A key question is whether one can arbitrarily modify the behavior of pretrained model by prompting or prefix-tuning it. Formally, whether prompting and prefix-tuning a pretrained model can universally approximate sequence-to-sequence functions. This paper answers in the affirmative and demonstrates that much smaller pretrained models than previously thought can be universal approximators when prefixed. In fact, the attention mechanism is uniquely suited for universal approximation with prefix-tuning a single attention head being sufficient to approximate any continuous function. Moreover, any sequence-to-sequence function can be approximated by prefixing a transformer with depth linear in the sequence length. Beyond these density-type results, we also offer Jackson-type bounds on the length of the prefix needed to approximate a function to a desired precision."
                    },
                    {
                        "title": "SNIP: Single-shot Network Pruning based on Connection Sensitivity",
                        "abstract": "Pruning large neural networks while maintaining their performance is often desirable due to the reduced space and time complexity. In existing methods, pruning is done within an iterative optimization procedure with either heuristically designed pruning schedules or additional hyperparameters, undermining their utility. In this work, we present a new approach that prunes a given network once at initialization prior to training. To achieve this, we introduce a saliency criterion based on connection sensitivity that identifies structurally important connections in the network for the given task. This eliminates the need for both pretraining and the complex pruning schedule while making it robust to architecture variations. After pruning, the sparse network is trained in the standard way. Our method obtains extremely sparse networks with virtually the same accuracy as the reference network on the MNIST, CIFAR-10, and Tiny-ImageNet classification tasks and is broadly applicable to various architectures including convolutional, residual and recurrent networks. Unlike existing methods, our approach enables us to demonstrate that the retained connections are indeed relevant to the given task."
                    },
                    {
                        "title": "Unifying Training and Inference for Panoptic Segmentation",
                        "abstract": "We present an end-to-end network to bridge the gap between training and inference pipeline for panoptic segmentation, a task that seeks to partition an image into semantic regions for \"stuff\" and object instances for \"things\". In contrast to recent works, our network exploits a parametrised, yet lightweight panoptic segmentation submodule, powered by an end-to-end learnt dense instance affinity, to capture the probability that any pair of pixels belong to the same instance. This panoptic submodule gives rise to a novel propagation mechanism for panoptic logits and enables the network to output a coherent panoptic segmentation map for both \"stuff\" and \"thing\" classes, without any post-processing. Reaping the benefits of end-to-end training, our full system sets new records on the popular street scene dataset, Cityscapes, achieving 61.4 PQ with a ResNet-50 backbone using only the fine annotations. On the challenging COCO dataset, our ResNet-50-based network also delivers state-of-the-art accuracy of 43.4 PQ. Moreover, our network flexibly works with and without object mask cues, performing competitively under both settings, which is of interest for applications with computation budgets."
                    },
                    {
                        "title": "Variational Mixture-of-Experts Autoencoders for Multi-Modal Deep Generative Models",
                        "abstract": "Learning generative models that span multiple data modalities, such as vision and language, is often motivated by the desire to learn more useful, generalisable representations that faithfully capture common underlying factors between the modalities. In this work, we characterise successful learning of such models as the fulfillment of four criteria: i) implicit latent decomposition into shared and private subspaces, ii) coherent joint generation over all modalities, iii) coherent cross-generation across individual modalities, and iv) improved model learning for individual modalities through multi-modal integration. Here, we propose a mixture-of-experts multimodal variational autoencoder (MMVAE) to learn generative models on different sets of modalities, including a challenging image-language dataset, and demonstrate its ability to satisfy all four criteria, both qualitatively and quantitatively."
                    },
                    {
                        "title": "Deep Deterministic Uncertainty for Semantic Segmentation",
                        "abstract": "We extend Deep Deterministic Uncertainty (DDU), a method for uncertainty estimation using feature space densities, to semantic segmentation. DDU enables quantifying and disentangling epistemic and aleatoric uncertainty in a single forward pass through the model. We study the similarity of feature representations of pixels at different locations for the same class and conclude that it is feasible to apply DDU location independently, which leads to a significant reduction in memory consumption compared to pixel dependent DDU. Using the DeepLab-v3+ architecture on Pascal VOC 2012, we show that DDU improves upon MC Dropout and Deep Ensembles while being significantly faster to compute."
                    },
                    {
                        "title": "Certifying Ensembles: A General Certification Theory with S-Lipschitzness",
                        "abstract": "Improving and guaranteeing the robustness of deep learning models has been a topic of intense research. Ensembling, which combines several classifiers to provide a better model, has shown to be beneficial for generalisation, uncertainty estimation, calibration, and mitigating the effects of concept drift. However, the impact of ensembling on certified robustness is less well understood. In this work, we generalise Lipschitz continuity by introducing S-Lipschitz classifiers, which we use to analyse the theoretical robustness of ensembles. Our results are precise conditions when ensembles of robust classifiers are more robust than any constituent classifier, as well as conditions when they are less robust."
                    }
                ]
            },
            "ea4a3248-6a00-451f-9d92-d5e5978245db": {
                "pk": "ea4a3248-6a00-451f-9d92-d5e5978245db",
                "name": "Lewis Hammond",
                "collaborators": [
                    "Alessandro Abate",
                    "Michael Wooldridge",
                    "James Fox",
                    "Tom Everitt",
                    "Julian Gutierrez",
                    "Vaishak Belle",
                    "Oliver Sourbut",
                    "Harriet Wood",
                    "Joar Skalse",
                    "Charlie Griffin"
                ],
                "domain": [
                    "Artificial Intelligence",
                    "Multi-Agent Systems",
                    "Reinforcement Learning",
                    "Game Theory"
                ],
                "publications": [
                    {
                        "title": "Learning Tractable Probabilistic Models for Moral Responsibility and Blame",
                        "abstract": "Moral responsibility is a major concern in autonomous systems, with applications ranging from self-driving cars to kidney exchanges. Although there have been recent attempts to formalise responsibility and blame, among similar notions, the problem of learning within these formalisms has been unaddressed. From the viewpoint of such systems, the urgent questions are: (a) How can models of moral scenarios and blameworthiness be extracted and learnt automatically from data? (b) How can judgements be computed effectively and efficiently, given the split-second decision points faced by some systems? By building on constrained tractable probabilistic learning, we propose and implement a hybrid (between data-driven and rule-based methods) learning framework for inducing models of such scenarios automatically from data and reasoning tractably from them. We report on experiments that compare our system with human judgement in three illustrative domains: lung cancer staging, teamwork management, and trolley problems."
                    },
                    {
                        "title": "Equilibrium Refinements for Multi-Agent Influence Diagrams: Theory and Practice",
                        "abstract": "Multi-agent influence diagrams (MAIDs) are a popular form of graphical model that, for certain classes of games, have been shown to offer key complexity and explainability advantages over traditional extensive form game (EFG) representations. In this paper, we extend previous work on MAIDs by introducing the concept of a MAID subgame, as well as subgame perfect and trembling hand perfect equilibrium refinements. We then prove several equivalence results between MAIDs and EFGs. Finally, we describe an open source implementation for reasoning about MAIDs and computing their equilibria."
                    },
                    {
                        "title": "Cooperation and Control in Delegation Games",
                        "abstract": "Many settings of interest involving humans and machines -- from virtual personal assistants to autonomous vehicles -- can naturally be modelled as principals (humans) delegating to agents (machines), which then interact with each other on their principals' behalf. We refer to these multi-principal, multi-agent scenarios as delegation games. In such games, there are two important failure modes: problems of control (where an agent fails to act in line their principal's preferences) and problems of cooperation (where the agents fail to work well together). In this paper we formalise and analyse these problems, further breaking them down into issues of alignment (do the players have similar preferences?) and capabilities (how competent are the players at satisfying those preferences?). We show -- theoretically and empirically -- how these measures determine the principals' welfare, how they can be estimated using limited observations, and thus how they might be used to help us design more aligned and cooperative AI systems."
                    },
                    {
                        "title": "Lexicographic Multi-Objective Reinforcement Learning",
                        "abstract": "In this work we introduce reinforcement learning techniques for solving lexicographic multi-objective problems. These are problems that involve multiple reward signals, and where the goal is to learn a policy that maximises the first reward signal, and subject to this constraint also maximises the second reward signal, and so on. We present a family of both action-value and policy gradient algorithms that can be used to solve such problems, and prove that they converge to policies that are lexicographically optimal. We evaluate the scalability and performance of these algorithms empirically, demonstrating their practical applicability. As a more specific application, we show how our algorithms can be used to impose safety constraints on the behaviour of an agent, and compare their performance in this context with that of other constrained reinforcement learning algorithms."
                    },
                    {
                        "title": "Game Theory with Simulation in the Presence of Unpredictable Randomisation",
                        "abstract": "AI agents will be predictable in certain ways that traditional agents are not. Where and how can we leverage this predictability in order to improve social welfare? We study this question in a game-theoretic setting where one agent can pay a fixed cost to simulate the other in order to learn its mixed strategy. As a negative result, we prove that, in contrast to prior work on pure-strategy simulation, enabling mixed-strategy simulation may no longer lead to improved outcomes for both players in all so-called \"generalised trust games\". In fact, mixed-strategy simulation does not help in any game where the simulatee's action can depend on that of the simulator. We also show that, in general, deciding whether simulation introduces Pareto-improving Nash equilibria in a given game is NP-hard. As positive results, we establish that mixed-strategy simulation can improve social welfare if the simulator has the option to scale their level of trust, if the players face challenges with both trust and coordination, or if maintaining some level of privacy is essential for enabling cooperation."
                    },
                    {
                        "title": "Multi-Agent Reinforcement Learning with Temporal Logic Specifications",
                        "abstract": "In this paper, we study the problem of learning to satisfy temporal logic specifications with a group of agents in an unknown environment, which may exhibit probabilistic behaviour. From a learning perspective these specifications provide a rich formal language with which to capture tasks or objectives, while from a logic and automated verification perspective the introduction of learning capabilities allows for practical applications in large, stochastic, unknown environments. The existing work in this area is, however, limited. Of the frameworks that consider full linear temporal logic or have correctness guarantees, all methods thus far consider only the case of a single temporal logic specification and a single agent. In order to overcome this limitation, we develop the first multi-agent reinforcement learning technique for temporal logic specifications, which is also novel in its ability to handle multiple specifications. We provide correctness and convergence guarantees for our main algorithm - ALMANAC (Automaton/Logic Multi-Agent Natural Actor-Critic) - even when using function approximation. Alongside our theoretical results, we further demonstrate the applicability of our technique via a set of preliminary experiments."
                    },
                    {
                        "title": "Reasoning about Causality in Games",
                        "abstract": "Causal reasoning and game-theoretic reasoning are fundamental topics in artificial intelligence, among many other disciplines: this paper is concerned with their intersection. Despite their importance, a formal framework that supports both these forms of reasoning has, until now, been lacking. We offer a solution in the form of (structural) causal games, which can be seen as extending Pearl's causal hierarchy to the game-theoretic domain, or as extending Koller and Milch's multi-agent influence diagrams to the causal domain. We then consider three key questions: i) How can the (causal) dependencies in games - either between variables, or between strategies - be modelled in a uniform, principled manner? ii) How may causal queries be computed in causal games, and what assumptions does this require? iii) How do causal games compare to existing formalisms? To address question i), we introduce mechanised games, which encode dependencies between agents' decision rules and the distributions governing the game. In response to question ii), we present definitions of predictions, interventions, and counterfactuals, and discuss the assumptions required for each. Regarding question iii), we describe correspondences between causal games and other formalisms, and explain how causal games can be used to answer queries that other causal or game-theoretic models do not support. Finally, we highlight possible applications of causal games, aided by an extensive open-source Python library."
                    },
                    {
                        "title": "Rational Verification for Probabilistic Systems",
                        "abstract": "Rational verification is the problem of determining which temporal logic properties will hold in a multi-agent system, under the assumption that agents in the system act rationally, by choosing strategies that collectively form a game-theoretic equilibrium. Previous work in this area has largely focussed on deterministic systems. In this paper, we develop the theory and algorithms for rational verification in probabilistic systems. We focus on concurrent stochastic games (CSGs), which can be used to model uncertainty and randomness in complex multi-agent environments. We study the rational verification problem for both non-cooperative games and cooperative games in the qualitative probabilistic setting. In the former case, we consider LTL properties satisfied by the Nash equilibria of the game and in the latter case LTL properties satisfied by the core. In both cases, we show that the problem is 2EXPTIME-complete, thus not harder than the much simpler verification problem of model checking LTL properties of systems modelled as Markov decision processes (MDPs)."
                    },
                    {
                        "title": "Bounded Robustness in Reinforcement Learning via Lexicographic Objectives",
                        "abstract": "Policy robustness in Reinforcement Learning may not be desirable at any cost: the alterations caused by robustness requirements from otherwise optimal policies should be explainable, quantifiable and formally verifiable. In this work we study how policies can be maximally robust to arbitrary observational noise by analysing how they are altered by this noise through a stochastic linear operator interpretation of the disturbances, and establish connections between robustness and properties of the noise kernel and of the underlying MDPs. Then, we construct sufficient conditions for policy robustness, and propose a robustness-inducing scheme, applicable to any policy gradient algorithm, that formally trades off expected policy utility for robustness through lexicographic optimisation, while preserving convergence and sub-optimality in the policy synthesis."
                    },
                    {
                        "title": "On Imperfect Recall in Multi-Agent Influence Diagrams",
                        "abstract": "Multi-agent influence diagrams (MAIDs) are a popular game-theoretic model based on Bayesian networks. In some settings, MAIDs offer significant advantages over extensive-form game representations. Previous work on MAIDs has assumed that agents employ behavioural policies, which set independent conditional probability distributions over actions for each of their decisions. In settings with imperfect recall, however, a Nash equilibrium in behavioural policies may not exist. We overcome this by showing how to solve MAIDs with forgetful and absent-minded agents using mixed policies and two types of correlated equilibrium. We also analyse the computational complexity of key decision problems in MAIDs, and explore tractable cases. Finally, we describe applications of MAIDs to Markov games and team situations, where imperfect recall is often unavoidable."
                    },
                    {
                        "title": "Welfare Diplomacy: Benchmarking Language Model Cooperation",
                        "abstract": "The growing capabilities and increasingly widespread deployment of AI systems necessitate robust benchmarks for measuring their cooperative capabilities. Unfortunately, most multi-agent benchmarks are either zero-sum or purely cooperative, providing limited opportunities for such measurements. We introduce a general-sum variant of the zero-sum board game Diplomacy -- called Welfare Diplomacy -- in which players must balance investing in military conquest and domestic welfare. We argue that Welfare Diplomacy facilitates both a clearer assessment of and stronger training incentives for cooperative capabilities. Our contributions are: (1) proposing the Welfare Diplomacy rules and implementing them via an open-source Diplomacy engine; (2) constructing baseline agents using zero-shot prompted language models; and (3) conducting experiments where we find that baselines using state-of-the-art models attain high social welfare but are exploitable. Our work aims to promote societal safety by aiding researchers in developing and assessing multi-agent AI systems. Code to evaluate Welfare Diplomacy and reproduce our experiments is available at https://github.com/mukobi/welfare-diplomacy."
                    }
                ]
            },
            "b0f27f77-b269-4997-92d3-3559ea997908": {
                "pk": "b0f27f77-b269-4997-92d3-3559ea997908",
                "name": "Christian Schroeder de Witt",
                "collaborators": [
                    "Shimon Whiteson",
                    "Jakob Foerster",
                    "Philip Torr",
                    "Philip H. S. Torr",
                    "Dylan Radovic",
                    "Samuel Sokota",
                    "Thomas Hornigold",
                    "Bradley Gram-Hansen",
                    "Nantas Nardelli",
                    "Andrew Gambardella"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Genetic Algorithms",
                    "Climate Modeling",
                    "Multi-Agent Systems"
                ],
                "publications": [
                    {
                        "title": "Biological Evolution and Genetic Algorithms: Exploring the Space of Abstract Tile Self-Assembly",
                        "abstract": "A physically-motivated genetic algorithm (GA) and full enumeration for a tile-based model of self-assembly (JaTAM) is implemented using a graphics processing unit (GPU). We observe performance gains with respect to state-of-the-art implementations on CPU of factor 7.7 for the GA and 2.9 for JaTAM. The correctness of our GA implementation is demonstrated using a test-bed fitness function, and our JaTAM implementation is verified by classifying a well-known search space $S_{2,8}$ based on two tile types. The performance gains achieved allow for the classification of a larger search space $S^{32}_{3,8}$ based on three tile types. The prevalence of structures based on two tile types demonstrates that simple organisms emerge preferrably even in complex ecosystems. The modularity of the largest structures found motivates the assumption that to first order, $S_{2,8}$ forms the building blocks of $S_{3,8}$. We conclude that GPUs may play an important role in future studies of evolutionary dynamics."
                    },
                    {
                        "title": "Stratospheric Aerosol Injection as a Deep Reinforcement Learning Problem",
                        "abstract": "As global greenhouse gas emissions continue to rise, the use of stratospheric aerosol injection (SAI), a form of solar geoengineering, is increasingly considered in order to artificially mitigate climate change effects. However, initial research in simulation suggests that naive SAI can have catastrophic regional consequences, which may induce serious geostrategic conflicts. Current geo-engineering research treats SAI control in low-dimensional approximation only. We suggest treating SAI as a high-dimensional control problem, with policies trained according to a context-sensitive reward function within the Deep Reinforcement Learning (DRL) paradigm. In order to facilitate training in simulation, we suggest to emulate HadCM3, a widely used General Circulation Model, using deep learning techniques. We believe this is the first application of DRL to the climate sciences."
                    },
                    {
                        "title": "Simulation-Based Inference for Global Health Decisions",
                        "abstract": "The COVID-19 pandemic has highlighted the importance of in-silico epidemiological modelling in predicting the dynamics of infectious diseases to inform health policy and decision makers about suitable prevention and containment strategies. Work in this setting involves solving challenging inference and control problems in individual-based models of ever increasing complexity. Here we discuss recent breakthroughs in machine learning, specifically in simulation-based inference, and explore its potential as a novel venue for model calibration to support the design and evaluation of public health interventions. To further stimulate research, we are developing software interfaces that turn two cornerstone COVID-19 and malaria epidemiology models COVID-sim, (https://github.com/mrc-ide/covid-sim/) and OpenMalaria (https://github.com/SwissTPH/openmalaria) into probabilistic programs, enabling efficient interpretable Bayesian inference within those simulators."
                    },
                    {
                        "title": "Is Independent Learning All You Need in the StarCraft Multi-Agent Challenge?",
                        "abstract": "Most recently developed approaches to cooperative multi-agent reinforcement learning in the \\emph{centralized training with decentralized execution} setting involve estimating a centralized, joint value function. In this paper, we demonstrate that, despite its various theoretical shortcomings, Independent PPO (IPPO), a form of independent learning in which each agent simply estimates its local value function, can perform just as well as or better than state-of-the-art joint learning approaches on popular multi-agent benchmark suite SMAC with little hyperparameter tuning. We also compare IPPO to several variants; the results suggest that IPPO's strong performance may be due to its robustness to some forms of environment non-stationarity."
                    },
                    {
                        "title": "Fixed Points in Cyber Space: Rethinking Optimal Evasion Attacks in the Age of AI-NIDS",
                        "abstract": "Cyber attacks are increasing in volume, frequency, and complexity. In response, the security community is looking toward fully automating cyber defense systems using machine learning. However, so far the resultant effects on the coevolutionary dynamics of attackers and defenders have not been examined. In this whitepaper, we hypothesise that increased automation on both sides will accelerate the coevolutionary cycle, thus begging the question of whether there are any resultant fixed points, and how they are characterised. Working within the threat model of Locked Shields, Europe's largest cyberdefense exercise, we study blackbox adversarial attacks on network classifiers. Given already existing attack capabilities, we question the utility of optimal evasion attack frameworks based on minimal evasion distances. Instead, we suggest a novel reinforcement learning setting that can be used to efficiently generate arbitrary adversarial perturbations. We then argue that attacker-defender fixed points are themselves general-sum games with complex phase transitions, and introduce a temporally extended multi-agent reinforcement learning framework in which the resultant dynamics can be studied. We hypothesise that one plausible fixed point of AI-NIDS may be a scenario where the defense strategy relies heavily on whitelisted feature flow subspaces. Finally, we demonstrate that a continual learning approach is required to study attacker-defender dynamics in temporally extended general-sum games."
                    },
                    {
                        "title": "Revealing Robust Oil and Gas Company Macro-Strategies using Deep Multi-Agent Reinforcement Learning",
                        "abstract": "The energy transition potentially poses an existential risk for major international oil companies (IOCs) if they fail to adapt to low-carbon business models. Projections of energy futures, however, are met with diverging assumptions on its scale and pace, causing disagreement among IOC decision-makers and their stakeholders over what the business model of an incumbent fossil fuel company should be. In this work, we used deep multi-agent reinforcement learning to solve an energy systems wargame wherein players simulate IOC decision-making, including hydrocarbon and low-carbon investments decisions, dividend policies, and capital structure measures, through an uncertain energy transition to explore critical and non-linear governance questions, from leveraged transitions to reserve replacements. Adversarial play facilitated by state-of-the-art algorithms revealed decision-making strategies robust to energy transition uncertainty and against multiple IOCs. In all games, robust strategies emerged in the form of low-carbon business models as a result of early transition-oriented movement. IOCs adopting such strategies outperformed business-as-usual and delayed transition strategies regardless of hydrocarbon demand projections. In addition to maximizing value, these strategies benefit greater society by contributing substantial amounts of capital necessary to accelerate the global low-carbon energy transition. Our findings point towards the need for lenders and investors to effectively mobilize transition-oriented finance and engage with IOCs to ensure responsible reallocation of capital towards low-carbon business models that would enable the emergence of fossil fuel incumbents as future low-carbon leaders."
                    },
                    {
                        "title": "RainBench: Towards Global Precipitation Forecasting from Satellite Imagery",
                        "abstract": "Extreme precipitation events, such as violent rainfall and hail storms, routinely ravage economies and livelihoods around the developing world. Climate change further aggravates this issue. Data-driven deep learning approaches could widen the access to accurate multi-day forecasts, to mitigate against such events. However, there is currently no benchmark dataset dedicated to the study of global precipitation forecasts. In this paper, we introduce \\textbf{RainBench}, a new multi-modal benchmark dataset for data-driven precipitation forecasting. It includes simulated satellite data, a selection of relevant meteorological data from the ERA5 reanalysis product, and IMERG precipitation data. We also release \\textbf{PyRain}, a library to process large precipitation datasets efficiently. We present an extensive analysis of our novel dataset and establish baseline results for two benchmark medium-range precipitation forecasting tasks. Finally, we discuss existing data-driven weather forecasting methodologies and suggest future research avenues."
                    },
                    {
                        "title": "Bayesian Exploration Networks",
                        "abstract": "Bayesian reinforcement learning (RL) offers a principled and elegant approach for sequential decision making under uncertainty. Most notably, Bayesian agents do not face an exploration/exploitation dilemma, a major pathology of frequentist methods. However theoretical understanding of model-free approaches is lacking. In this paper, we introduce a novel Bayesian model-free formulation and the first analysis showing that model-free approaches can yield Bayes-optimal policies. We show all existing model-free approaches make approximations that yield policies that can be arbitrarily Bayes-suboptimal. As a first step towards model-free Bayes optimality, we introduce the Bayesian exploration network (BEN) which uses normalising flows to model both the aleatoric uncertainty (via density estimation) and epistemic uncertainty (via variational inference) in the Bellman operator. In the limit of complete optimisation, BEN learns true Bayes-optimal policies, but like in variational expectation-maximisation, partial optimisation renders our approach tractable. Empirical results demonstrate that BEN can learn true Bayes-optimal policies in tasks where existing model-free approaches fail."
                    },
                    {
                        "title": "Computing Low-Entropy Couplings for Large-Support Distributions",
                        "abstract": "Minimum-entropy coupling (MEC) -- the process of finding a joint distribution with minimum entropy for given marginals -- has applications in areas such as causality and steganography. However, existing algorithms are either computationally intractable for large-support distributions or limited to specific distribution types and sensitive to hyperparameter choices. This work addresses these limitations by unifying a prior family of iterative MEC (IMEC) approaches into a generalized partition-based formalism. From this framework, we derive a novel IMEC algorithm called ARIMEC, capable of handling arbitrary discrete distributions, and introduce a method to make IMEC robust to suboptimal hyperparameter settings. These innovations facilitate the application of IMEC to high-throughput steganography with language models, among other settings. Our codebase is available at https://github.com/ssokota/mec ."
                    },
                    {
                        "title": "Generalized Beliefs for Cooperative AI",
                        "abstract": "Self-play is a common paradigm for constructing solutions in Markov games that can yield optimal policies in collaborative settings. However, these policies often adopt highly-specialized conventions that make playing with a novel partner difficult. To address this, recent approaches rely on encoding symmetry and convention-awareness into policy training, but these require strong environmental assumptions and can complicate policy training. We therefore propose moving the learning of conventions to the belief space. Specifically, we propose a belief learning model that can maintain beliefs over rollouts of policies not seen at training time, and can thus decode and adapt to novel conventions at test time. We show how to leverage this model for both search and training of a best response over various pools of policies to greatly improve ad-hoc teamplay. We also show how our setup promotes explainability and interpretability of nuanced agent conventions."
                    },
                    {
                        "title": "Cheap Talk Discovery and Utilization in Multi-Agent Reinforcement Learning",
                        "abstract": "By enabling agents to communicate, recent cooperative multi-agent reinforcement learning (MARL) methods have demonstrated better task performance and more coordinated behavior. Most existing approaches facilitate inter-agent communication by allowing agents to send messages to each other through free communication channels, i.e., cheap talk channels. Current methods require these channels to be constantly accessible and known to the agents a priori. In this work, we lift these requirements such that the agents must discover the cheap talk channels and learn how to use them. Hence, the problem has two main parts: cheap talk discovery (CTD) and cheap talk utilization (CTU). We introduce a novel conceptual framework for both parts and develop a new algorithm based on mutual information maximization that outperforms existing algorithms in CTD/CTU settings. We also release a novel benchmark suite to stimulate future research in CTD/CTU."
                    },
                    {
                        "title": "Toward Robust Real-World Audio Deepfake Detection: Closing the Explainability Gap",
                        "abstract": "The rapid proliferation of AI-manipulated or generated audio deepfakes poses serious challenges to media integrity and election security. Current AI-driven detection solutions lack explainability and underperform in real-world settings. In this paper, we introduce novel explainability methods for state-of-the-art transformer-based audio deepfake detectors and open-source a novel benchmark for real-world generalizability. By narrowing the explainability gap between transformer-based audio deepfake detectors and traditional methods, our results not only build trust with human experts, but also pave the way for unlocking the potential of citizen intelligence to overcome the scalability issue in audio deepfake detection."
                    },
                    {
                        "title": "Efficient Dictionary Learning with Switch Sparse Autoencoders",
                        "abstract": "Sparse autoencoders (SAEs) are a recent technique for decomposing neural network activations into human-interpretable features. However, in order for SAEs to identify all features represented in frontier models, it will be necessary to scale them up to very high width, posing a computational challenge. In this work, we introduce Switch Sparse Autoencoders, a novel SAE architecture aimed at reducing the compute cost of training SAEs. Inspired by sparse mixture of experts models, Switch SAEs route activation vectors between smaller \"expert\" SAEs, enabling SAEs to efficiently scale to many more features. We present experiments comparing Switch SAEs with other SAE architectures, and find that Switch SAEs deliver a substantial Pareto improvement in the reconstruction vs. sparsity frontier for a given fixed training compute budget. We also study the geometry of features across experts, analyze features duplicated across experts, and verify that Switch SAE features are as interpretable as features found by other SAE architectures."
                    },
                    {
                        "title": "Mirror Learning: A Unifying Framework of Policy Optimisation",
                        "abstract": "Modern deep reinforcement learning (RL) algorithms are motivated by either the generalised policy iteration (GPI) or trust-region learning (TRL) frameworks. However, algorithms that strictly respect these theoretical frameworks have proven unscalable. Surprisingly, the only known scalable algorithms violate the GPI/TRL assumptions, e.g. due to required regularisation or other heuristics. The current explanation of their empirical success is essentially \"by analogy\": they are deemed approximate adaptations of theoretically sound methods. Unfortunately, studies have shown that in practice these algorithms differ greatly from their conceptual ancestors. In contrast, in this paper we introduce a novel theoretical framework, named Mirror Learning, which provides theoretical guarantees to a large class of algorithms, including TRPO and PPO. While the latter two exploit the flexibility of our framework, GPI and TRL fit in merely as pathologically restrictive corner cases thereof. This suggests that the empirical performance of state-of-the-art methods is a direct consequence of their theoretical properties, rather than of aforementioned approximate analogies. Mirror learning sets us free to boldly explore novel, theoretically sound RL algorithms, a thus far uncharted wonderland."
                    },
                    {
                        "title": "Model-Free Opponent Shaping",
                        "abstract": "In general-sum games, the interaction of self-interested learning agents commonly leads to collectively worst-case outcomes, such as defect-defect in the iterated prisoner's dilemma (IPD). To overcome this, some methods, such as Learning with Opponent-Learning Awareness (LOLA), shape their opponents' learning process. However, these methods are myopic since only a small number of steps can be anticipated, are asymmetric since they treat other agents as naive learners, and require the use of higher-order derivatives, which are calculated through white-box access to an opponent's differentiable learning algorithm. To address these issues, we propose Model-Free Opponent Shaping (M-FOS). M-FOS learns in a meta-game in which each meta-step is an episode of the underlying inner game. The meta-state consists of the inner policies, and the meta-policy produces a new inner policy to be used in the next episode. M-FOS then uses generic model-free optimisation methods to learn meta-policies that accomplish long-horizon opponent shaping. Empirically, M-FOS near-optimally exploits naive learners and other, more sophisticated algorithms from the literature. For example, to the best of our knowledge, it is the first method to learn the well-known Zero-Determinant (ZD) extortion strategy in the IPD. In the same settings, M-FOS leads to socially optimal outcomes under meta-self-play. Finally, we show that M-FOS can be scaled to high-dimensional settings."
                    },
                    {
                        "title": "SAGE: Scalable Ground Truth Evaluations for Large Sparse Autoencoders",
                        "abstract": "A key challenge in interpretability is to decompose model activations into meaningful features. Sparse autoencoders (SAEs) have emerged as a promising tool for this task. However, a central problem in evaluating the quality of SAEs is the absence of ground truth features to serve as an evaluation gold standard. Current evaluation methods for SAEs are therefore confronted with a significant trade-off: SAEs can either leverage toy models or other proxies with predefined ground truth features; or they use extensive prior knowledge of realistic task circuits. The former limits the generalizability of the evaluation results, while the latter limits the range of models and tasks that can be used for evaluations. We introduce SAGE: Scalable Autoencoder Ground-truth Evaluation, a ground truth evaluation framework for SAEs that scales to large state-of-the-art SAEs and models. We demonstrate that our method can automatically identify task-specific activations and compute ground truth features at these points. Compared to previous methods we reduce the training overhead by introducing a novel reconstruction method that allows to apply residual stream SAEs to sublayer activations. This eliminates the need for SAEs trained on every task-specific activation location. Then we validate the scalability of our framework, by evaluating SAEs on novel tasks on Pythia70M, GPT-2 Small, and Gemma-2-2. Our framework therefore paves the way for generalizable, large-scale evaluations of SAEs in interpretability research."
                    },
                    {
                        "title": "(Private)-Retroactive Carbon Pricing [(P)ReCaP]: A Market-based Approach for Climate Finance and Risk Assessment",
                        "abstract": "Insufficient Social Cost of Carbon (SCC) estimation methods and short-term decision-making horizons have hindered the ability of carbon emitters to properly correct for the negative externalities of climate change, as well as the capacity of nations to balance economic and climate policy. To overcome these limitations, we introduce Retrospective Social Cost of Carbon Updating (ReSCCU), a novel mechanism that corrects for these limitations as empirically measured evidence is collected. To implement ReSCCU in the context of carbon taxation, we propose Retroactive Carbon Pricing (ReCaP), a market mechanism in which polluters offload the payment of ReSCCU adjustments to insurers. To alleviate systematic risks and minimize government involvement, we introduce the Private ReCaP (PReCaP) prediction market, which could see real-world implementation based on the engagement of a few high net-worth individuals or independent institutions."
                    }
                ]
            }
        }
    },
    "2406.10023": {
        "paper_data": {
            "title": "Deep Bayesian Active Learning for Preference Modeling in Large Language Models",
            "url": "http://arxiv.org/abs/2406.10023v1",
            "arxiv_id": "2406.10023",
            "authors": [
                "Luckeciano C. Melo",
                "Panagiotis Tigas",
                "Alessandro Abate",
                "Yarin Gal"
            ],
            "abstract": "Leveraging human preferences for steering the behavior of Large Language Models (LLMs) has demonstrated notable success in recent years. Nonetheless, data selection and labeling are still a bottleneck for these systems, particularly at large scale. Hence, selecting the most informative points for acquiring human feedback may considerably reduce the cost of preference labeling and unleash the further development of LLMs. Bayesian Active Learning provides a principled framework for addressing this challenge and has demonstrated remarkable success in diverse settings. However, previous attempts to employ it for Preference Modeling did not meet such expectations. In this work, we identify that naive epistemic uncertainty estimation leads to the acquisition of redundant samples. We address this by proposing the Bayesian Active Learner for Preference Modeling (BAL-PM), a novel stochastic acquisition policy that not only targets points of high epistemic uncertainty according to the preference model but also seeks to maximize the entropy of the acquired prompt distribution in the feature space spanned by the employed LLM. Notably, our experiments demonstrate that BAL-PM requires 33% to 68% fewer preference labels in two popular human preference datasets and exceeds previous stochastic Bayesian acquisition policies.",
            "introduction": "   1 Introduction  Preference Modeling is a key component to aligning unsupervised pre-trained Large Language Models (LLMs) towards human preferences [51, 42, 2, 54]. It is often performed by collecting human feedback for a set of prompt-answer pairs and then leveraging the data to steer the behavior of such models, either directly [45] or via reward models [59]. Nevertheless, human feedback generation is laborious [7], especially when it requires specialized knowledge [3, 22]. Furthermore, the quality of the prompts has a crucial impact on the performance of fine-tuned models [52]. Hence, selecting the most informative points to gather feedback is essential to reduce costs and enable better LLMs.   Despite its substantial impact, data selection for Preference Modeling poses a significant challenge. The prompt-answer pool is arbitrarily large and semantically rich. Additionally, human feedback is inherently noisy, with low agreement rates among labelers, typically between 60% \u2013 75% [59, 51, 12, 10]. Lastly, the intrinsic scale of LLM development requires parallelized labeling and makes frequent model updates prohibitively expensive, limiting the applicability of many active learning schemes that rely on single-point acquisition [28].   Bayesian Active Learning provides a principled approach to data selection [16, 27, 38], which has demonstrated remarkable success across different fields [50, 17, 14]. However, its application in Active Preference Modeling is not straightforward. Past attempts of employing the framework in this setting reported no benefits over random selection [18], arguably due to poor uncertainty estimation in the context of LLMs, which is indeed an open challenge and active area of research [32].   Figure 2: An illustration of how BAL-PM works. For each tuple (x,y1,y2)\u2208\ud835\udc9fp\u2062o\u2062o\u2062l\ud835\udc65subscript\ud835\udc661subscript\ud835\udc662subscript\ud835\udc9f\ud835\udc5d\ud835\udc5c\ud835\udc5c\ud835\udc59(x,y_{1},y_{2})\\in\\mathcal{D}_{pool}( italic_x , italic_y start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , italic_y start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ) \u2208 caligraphic_D start_POSTSUBSCRIPT italic_p italic_o italic_o italic_l end_POSTSUBSCRIPT, we obtain features for the prompt and prompt-answer pairs by computing the last layer embeddings of the base LLM. We leverage the prompt feature space to estimate the entropy score of the acquired prompt distribution, \u210b^\u2062(Xt\u2062r\u2062a\u2062i\u2062n\u222a{x})^\u210bsubscript\ud835\udc4b\ud835\udc61\ud835\udc5f\ud835\udc4e\ud835\udc56\ud835\udc5b\ud835\udc65\\hat{\\mathcal{H}}({X}_{train}\\cup\\{x\\})over^ start_ARG caligraphic_H end_ARG ( italic_X start_POSTSUBSCRIPT italic_t italic_r italic_a italic_i italic_n end_POSTSUBSCRIPT \u222a { italic_x } ). Similarly, we use the prompt-answer features as input for the Bayesian Preference Model, which is used to estimate task-dependent epistemic uncertainty scores, U^\u2062(x,y1,y2)^\ud835\udc48\ud835\udc65subscript\ud835\udc661subscript\ud835\udc662\\hat{U}(x,y_{1},y_{2})over^ start_ARG italic_U end_ARG ( italic_x , italic_y start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , italic_y start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ). BAL-PM selects the tuple that maximizes the linear combination of both scores.   We identify two reasons for this phenomenon. First, the inherent bias of approximate Bayesian inference in deep learning models, particularly for LLMs. Second, and more nuanced, the current intractability of epistemic uncertainty estimation methods in Preference Modeling for LLMs, a context that intrinsically requires batch acquisition. Proper estimators for this setting present combinatorial complexity, and even greedy approximations are still computationally demanding and impractical [28, 29]. This limitation leads to relying on simpler single-point acquisition schemes such as BALD [35] (as in Gleave and Irving [18]), designed to acquire individual points followed by model updates. However, these assumptions are far from realistic for the scale of Preference Modeling in LLMs, and naively applying such methods for batch acquisition leads to the selection of redundant samples.   In this",
            "references": [
                {
                    "title": "Efficient Exploration for LLMs",
                    "abstract": "We present evidence of substantial benefit from efficient exploration in gathering human feedback to improve large language models. In our experiments, an agent sequentially generates queries while fitting a reward model to the feedback received. Our best-performing agent generates queries using double Thompson sampling, with uncertainty represented by an epistemic neural network. Our results demonstrate that efficient exploration enables high levels of performance with far fewer queries. Further, both uncertainty estimation and the choice of exploration scheme play critical roles."
                },
                {
                    "title": "Advancing Deep Active Learning & Data Subset Selection: Unifying Principles with Information-Theory Intuitions",
                    "abstract": "At its core, this thesis aims to enhance the practicality of deep learning by improving the label and training efficiency of deep learning models. To this end, we investigate data subset selection techniques, specifically active learning and active sampling, grounded in information-theoretic principles. Active learning improves label efficiency, while active sampling enhances training efficiency. Supervised deep learning models often require extensive training with labeled data. Label acquisition can be expensive and time-consuming, and training large models is resource-intensive, hindering the adoption outside academic research and\"big tech.\"Existing methods for data subset selection in deep learning often rely on heuristics or lack a principled information-theoretic foundation. In contrast, this thesis examines several objectives for data subset selection and their applications within deep learning, striving for a more principled approach inspired by information theory. We begin by disentangling epistemic and aleatoric uncertainty in single forward-pass deep neural networks, which provides helpful intuitions and insights into different forms of uncertainty and their relevance for data subset selection. We then propose and investigate various approaches for active learning and data subset selection in (Bayesian) deep learning. Finally, we relate various existing and proposed approaches to approximations of information quantities in weight or prediction space. Underpinning this work is a principled and practical notation for information-theoretic quantities that includes both random variables and observed outcomes. This thesis demonstrates the benefits of working from a unified perspective and highlights the potential impact of our contributions to the practical application of deep learning."
                },
                {
                    "title": "A Bayesian active learning platform for scalable combination drug screens",
                    "abstract": "Large-scale combination drug screens are generally considered intractable due to the immense number of possible combinations. Existing approaches use ad hoc fixed experimental designs then train machine learning models to impute novel combinations. Here we propose BATCHIE, an orthogonal approach that conducts experiments dynamically in batches. BATCHIE uses information theory and probabilistic modeling to design each batch to be maximally informative based on the results of previous experiments. On retrospective experiments from previous large-scale screens, BATCHIE designs rapidly discover highly effective and synergistic combinations. To validate BATCHIE prospectively, we conducted a combination screen on a collection of pediatric cancer cell lines using a 206 drug library. After exploring only 4% of the 1.4M possible experiments, the BATCHIE model was highly accurate at predicting novel combinations and detecting synergies. Further, the model identified a panel of top combinations for Ewing sarcomas, all of which were experimentally confirmed to be effective, including the rational and translatable top hit of PARP plus topoisomerase I inhibition. These results demonstrate that adaptive experiments can enable large-scale unbiased combination drug screens with a relatively small number of experiments, thereby powering a new wave of combination drug discoveries. BATCHIE is open source and publicly available (https://github.com/tansey-lab/batchie)."
                },
                {
                    "title": "Sample Efficient Reinforcement Learning from Human Feedback via Active Exploration",
                    "abstract": "Preference-based feedback is important for many applications in reinforcement learning where direct evaluation of a reward function is not feasible. A notable recent example arises in reinforcement learning from human feedback (RLHF) on large language models. For many applications of RLHF, the cost of acquiring the human feedback can be substantial. In this work, we take advantage of the fact that one can often choose contexts at which to obtain human feedback in order to most efficiently identify a good policy, and formalize this as an offline contextual dueling bandit problem. We give an upper-confidence-bound style algorithm for this problem and prove a polynomial worst-case regret bound. We then provide empirical confirmation in a synthetic setting that our approach outperforms existing methods. After, we extend the setting and methodology for practical use in RLHF training of large language models. Here, our method is able to reach better performance with fewer samples of human preferences than multiple baselines on three real-world datasets."
                },
                {
                    "title": "Deep Batch Active Learning for Drug Discovery",
                    "abstract": "A key challenge in drug discovery is to optimize, in silico, various absorption and affinity properties of small molecules. One strategy that was proposed for such optimization process is active learning. In active learning molecules are selected for testing based on their likelihood of improving model performance. To enable the use of active learning with advanced neural network models we developed two novel active learning batch selection methods. These methods were tested on several public datasets for different optimization goals and with different sizes. We have also curated new affinity datasets that provide chronological information on state-of-the-art experimental strategy. As we show, for all datasets the new active learning methods greatly improved on existing and current batch selection methods leading to significant potential saving in the number of experiments needed to reach the same model performance. Our methods are general and can be used with any package including the popular DeepChem library."
                },
                {
                    "title": "Reward Model Ensembles Help Mitigate Overoptimization",
                    "abstract": "Reinforcement learning from human feedback (RLHF) is a standard approach for fine-tuning large language models to follow instructions. As part of this process, learned reward models are used to approximately model human preferences. However, as imperfect representations of the\"true\"reward, these learned reward models are susceptible to overoptimization. Gao et al. (2023) studied this phenomenon in a synthetic human feedback setup with a significantly larger\"gold\"reward model acting as the true reward (instead of humans) and showed that overoptimization remains a persistent problem regardless of the size of the proxy reward model and training data used. Using a similar setup, we conduct a systematic study to evaluate the efficacy of using ensemble-based conservative optimization objectives, specifically worst-case optimization (WCO) and uncertainty-weighted optimization (UWO), for mitigating reward model overoptimization when using two optimization methods: (a) best-of-n sampling (BoN) (b) proximal policy optimization (PPO). We additionally extend the setup of Gao et al. (2023) to include 25% label noise to better mirror real-world conditions. Both with and without label noise, we find that conservative optimization practically eliminates overoptimization and improves performance by up to 70% for BoN sampling. For PPO, ensemble-based conservative optimization always reduces overoptimization and outperforms single reward model optimization. Moreover, combining it with a small KL penalty successfully prevents overoptimization at no performance cost. Overall, our results demonstrate that ensemble-based conservative optimization can effectively counter overoptimization."
                },
                {
                    "title": "Accelerating Reinforcement Learning with Value-Conditional State Entropy Exploration",
                    "abstract": "A promising technique for exploration is to maximize the entropy of visited state distribution, i.e., state entropy, by encouraging uniform coverage of visited state space. While it has been effective for an unsupervised setup, it tends to struggle in a supervised setup with a task reward, where an agent prefers to visit high-value states to exploit the task reward. Such a preference can cause an imbalance between the distributions of high-value states and low-value states, which biases exploration towards low-value state regions as a result of the state entropy increasing when the distribution becomes more uniform. This issue is exacerbated when high-value states are narrowly distributed within the state space, making it difficult for the agent to complete the tasks. In this paper, we present a novel exploration technique that maximizes the value-conditional state entropy, which separately estimates the state entropies that are conditioned on the value estimates of each state, then maximizes their average. By only considering the visited states with similar value estimates for computing the intrinsic bonus, our method prevents the distribution of low-value states from affecting exploration around high-value states, and vice versa. We demonstrate that the proposed alternative to the state entropy baseline significantly accelerates various reinforcement learning algorithms across a variety of tasks within MiniGrid, DeepMind Control Suite, and Meta-World benchmarks. Source code is available at https://sites.google.com/view/rl-vcse."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedback",
                    "abstract": "Large language models (LLMs) such as ChatGPT have seen widespread adoption due to their strong instruction-following abilities. Developing these LLMs involves a complex yet poorly understood workflow requiring training with human feedback. Replicating and understanding this instruction-following requires tackling three major challenges: the high cost of data collection, the lack of trustworthy evaluation, and the absence of reference method implementations. We address these challenges with AlpacaFarm, a simulator that enables research and development for learning from feedback at a low cost. First, we design LLM prompts to simulate human feedback that are 50x cheaper than crowdworkers and display high agreement with humans. Second, we propose an automatic evaluation and validate it against human instructions obtained on real-world interactions. Third, we contribute reference implementations for several methods (PPO, DPO, best-of-n, expert iteration, and more) that learn from pairwise feedback. Finally, as an end-to-end validation of AlpacaFarm, we train and evaluate eleven models on 10k pairs of real human feedback and show that rankings of models trained in AlpacaFarm match rankings of models trained on human data. As a demonstration of the research possible in AlpacaFarm, we find that methods that use a reward model can substantially improve over supervised fine-tuning and that our reference PPO implementation leads to a +10% improvement in win-rate against Davinci003. We release all components of AlpacaFarm at https://github.com/tatsu-lab/alpaca_farm."
                },
                {
                    "title": "Prediction-Oriented Bayesian Active Learning",
                    "abstract": "Information-theoretic approaches to active learning have traditionally focused on maximising the information gathered about the model parameters, most commonly by optimising the BALD score. We highlight that this can be suboptimal from the perspective of predictive performance. For example, BALD lacks a notion of an input distribution and so is prone to prioritise data of limited relevance. To address this we propose the expected predictive information gain (EPIG), an acquisition function that measures information gain in the space of predictions rather than parameters. We find that using EPIG leads to stronger predictive performance compared with BALD across a range of datasets and models, and thus provides an appealing drop-in replacement."
                },
                {
                    "title": "Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation",
                    "abstract": "We introduce a method to measure uncertainty in large language models. For tasks like question answering, it is essential to know when we can trust the natural language outputs of foundation models. We show that measuring uncertainty in natural language is challenging because of\"semantic equivalence\"-- different sentences can mean the same thing. To overcome these challenges we introduce semantic entropy -- an entropy which incorporates linguistic invariances created by shared meanings. Our method is unsupervised, uses only a single model, and requires no modifications to off-the-shelf language models. In comprehensive ablation studies we show that the semantic entropy is more predictive of model accuracy on question answering data sets than comparable baselines."
                },
                {
                    "title": "Uncertainty Estimation for Language Reward Models",
                    "abstract": "Language models can learn a range of capabilities from unsupervised training on text corpora. However, to solve a particular problem (such as text summarization) it is typically necessary to fine-tune them on a task-specific dataset. It is often easier for humans to choose between options than to provide labeled data, and prior work has achieved state-of-the-art performance by training a reward model from such preference comparisons. However, collecting a large preference comparison dataset is still expensive -- and the learned reward models are unreliable out-of-distribution. We seek to address these problems via uncertainty estimation, which can improve sample efficiency and robustness using active learning and risk-averse reinforcement learning (RL). Specifically, we use bootstrap aggregating (bagging) to train an ensemble of reward models differing in the initialization of their final layer. Ensembles have proved successful in prior applications of active learning, but we find that in our setting ensemble active learning does not outperform random sampling. Further experiments show that while the aggregate predictions are well-calibrated, the ensemble's estimated epistemic uncertainty is only weakly correlated with model error. We suspect this is because the ensemble members are fine-tuned from a single model and so are similar to one another. This suggests current pre-training methods will need to be modified to support uncertainty estimation, e.g. by training multiple language models."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Bayesian Active Learning for Sim-to-Real Robotic Perception",
                    "abstract": "While learning from synthetic training data has recently gained an increased attention, in real-world robotic applications, there are still performance deficiencies due to the so-called Sim-to-Real gap. In practice, this gap is hard to resolve with only synthetic data. Therefore, we focus on an efficient acquisition of real data within a Sim-to-Real learning pipeline. Concretely, we employ deep Bayesian active learning to minimize manual annotation efforts and devise an autonomous learning paradigm to select the data that is considered useful for the human expert to annotate. To achieve this, a Bayesian Neural Network (BNN) object detector providing reliable un-certainty estimates is adapted to infer the informativeness of the unlabeled data. Furthermore, to cope with misalignments of the label distribution in uncertainty-based sampling, we develop an effective randomized sampling strategy that performs favorably compared to other complex alternatives. In our experiments on object classification and detection, we show benefits of our approach and provide evidence that labeling efforts can be reduced significantly. Finally, we demonstrate the practical effectiveness of this idea in a grasping task on an assistive robot."
                },
                {
                    "title": "Stochastic Batch Acquisition: A Simple Baseline for Deep Active Learning",
                    "abstract": "We examine a simple stochastic strategy for adapting well-known single-point acquisition functions to allow batch active learning. Unlike acquiring the top-K points from the pool set, score- or rank-based sampling takes into account that acquisition scores change as new data are acquired. This simple strategy for adapting standard single-sample acquisition strategies can even perform just as well as compute-intensive state-of-the-art batch acquisition functions, like BatchBALD or BADGE, while using orders of magnitude less compute. In addition to providing a practical option for machine learning practitioners, the surprising success of the proposed method in a wide range of experimental settings raises a difficult question for the field: when are these expensive batch acquisition methods pulling their weight?"
                },
                {
                    "title": "What Are Bayesian Neural Network Posteriors Really Like?",
                    "abstract": "The posterior over Bayesian neural network (BNN) parameters is extremely high-dimensional and non-convex. For computational reasons, researchers approximate this posterior using inexpensive mini-batch methods such as mean-field variational inference or stochastic-gradient Markov chain Monte Carlo (SGMCMC). To investigate foundational questions in Bayesian deep learning, we instead use full-batch Hamiltonian Monte Carlo (HMC) on modern architectures. We show that (1) BNNs can achieve significant performance gains over standard training and deep ensembles; (2) a single long HMC chain can provide a comparable representation of the posterior to multiple shorter chains; (3) in contrast to recent studies, we find posterior tempering is not needed for near-optimal performance, with little evidence for a\"cold posterior\"effect, which we show is largely an artifact of data augmentation; (4) BMA performance is robust to the choice of prior scale, and relatively similar for diagonal Gaussian, mixture of Gaussian, and logistic priors; (5) Bayesian neural networks show surprisingly poor generalization under domain shift; (6) while cheaper alternatives such as deep ensembles and SGMCMC methods can provide good generalization, they provide distinct predictive distributions from HMC. Notably, deep ensemble predictive distributions are similarly close to HMC as standard SGLD, and closer than standard variational inference."
                },
                {
                    "title": "Behavior From the Void: Unsupervised Active Pre-Training",
                    "abstract": "We introduce a new unsupervised pre-training method for reinforcement learning called APT, which stands for Active Pre-Training. APT learns behaviors and representations by actively searching for novel states in reward-free environments. The key novel idea is to explore the environment by maximizing a non-parametric entropy computed in an abstract representation space, which avoids challenging density modeling and consequently allows our approach to scale much better in environments that have high-dimensional observations (e.g., image observations). We empirically evaluate APT by exposing task-specific reward after a long unsupervised pre-training phase. In Atari games, APT achieves human-level performance on 12 games and obtains highly competitive performance compared to canonical fully supervised RL algorithms. On DMControl suite, APT beats all baselines in terms of asymptotic performance and data efficiency and dramatically improves performance on tasks that are extremely difficult to train from scratch."
                },
                {
                    "title": "Deep Deterministic Uncertainty: A New Simple Baseline",
                    "abstract": "Reliable uncertainty from deterministic single-forward pass models is sought after because conventional methods of uncertainty quantification are computationally expensive. We take two complex single-forward-pass uncertainty approaches, DUQ and SNGP, and examine whether they mainly rely on a well-regularized feature space. Crucially, without using their more complex methods for estimating uncertainty, we find that a single softmax neural net with such a regularized feature-space, achieved via residual connections and spectral normalization, outperforms DUQ and SNGP's epistemic uncertainty predictions using simple Gaussian Discriminant Analysis post-training as a separate feature-space density estimator-without fine-tuning on OoD data, feature ensembling, or input pre-procressing. Our conceptually simple Deep Deterministic Uncertainty (DDU) baseline can also be used to disentangle aleatoric and epistemic uncertainty and performs as well as Deep Ensembles, the state-of-the art for uncertainty prediction, on several OoD bench-marks (CIFAR-10/100 vs SVHN/Tiny-ImageNet, ImageNet vs ImageNet-O), active learning settings across different model architectures, as well as in large scale vision tasks like semantic segmentation, while being computationally cheaper."
                },
                {
                    "title": "Quantifying Aleatoric and Epistemic Uncertainty Using Density Estimation in Latent Space",
                    "abstract": "The distribution of a neural network's latent representations has been successfully used to detect Out-of-Distribution (OOD) data. Since OOD detection denotes a popular benchmark for epistemic uncertainty estimates, this raises the question of a deeper correlation. This work investigates whether the distribution of latent representations indeed contains information about the uncertainty associated with the predictions of a neural network. Prior work identifies epistemic uncertainty with the surprise, thus the negative log-likelihood, of observing a particular latent representation, which we verify empirically. Moreover, we demonstrate that the output-conditional distribution of hidden representations allows quantifying aleatoric uncertainty via the entropy of the predictive distribution. We analyze epistemic and aleatoric uncertainty inferred from the representations of different layers and conclude with the exciting finding that the hidden repesentations of a deterministic neural network indeed contain information about its uncertainty. We verify our findings on both classification and regression models."
                },
                {
                    "title": "Energy-based Out-of-distribution Detection",
                    "abstract": "Determining whether inputs are out-of-distribution (OOD) is an essential building block for safely deploying machine learning models in the open world. However, previous methods relying on the softmax confidence score suffer from overconfident posterior distributions for OOD data. We propose a unified framework for OOD detection that uses an energy score. We show that energy scores better distinguish in- and out-of-distribution samples than the traditional approach using the softmax scores. Unlike softmax confidence scores, energy scores are theoretically aligned with the probability density of the inputs and are less susceptible to the overconfidence issue. Within this framework, energy can be flexibly used as a scoring function for any pre-trained neural classifier as well as a trainable cost function to shape the energy surface explicitly for OOD detection. On a CIFAR-10 pre-trained WideResNet, using the energy score reduces the average FPR (at TPR 95%) by 18.03% compared to the softmax confidence score. With energy-based training, our method outperforms the state-of-the-art on common benchmarks."
                },
                {
                    "title": "Learning to summarize from human feedback",
                    "abstract": "As language models become more powerful, training and evaluation are increasingly bottlenecked by the data and metrics used for a particular task. For example, summarization models are often trained to predict human reference summaries and evaluated using ROUGE, but both of these metrics are rough proxies for what we really care about---summary quality. In this work, we show that it is possible to significantly improve summary quality by training a model to optimize for human preferences. We collect a large, high-quality dataset of human comparisons between summaries, train a model to predict the human-preferred summary, and use that model as a reward function to fine-tune a summarization policy using reinforcement learning. We apply our method to a version of the TL;DR dataset of Reddit posts and find that our models significantly outperform both human reference summaries and much larger models fine-tuned with supervised learning alone. Our models also transfer to CNN/DM news articles, producing summaries nearly as good as the human reference without any news-specific fine-tuning. We conduct extensive analyses to understand our human feedback dataset and fine-tuned models We establish that our reward model generalizes to new datasets, and that optimizing our reward model results in better summaries than optimizing ROUGE according to humans. We hope the evidence from our paper motivates machine learning researchers to pay closer attention to how their training loss affects the model behavior they actually want."
                },
                {
                    "title": "On-the-fly Closed-loop Autonomous Materials Discovery via Bayesian Active Learning",
                    "abstract": "Active learning - the field of machine learning (ML) dedicated to optimal experiment design, has played a part in science as far back as the 18th century when Laplace used it to guide his discovery of celestial mechanics [1]. In this work we focus a closed-loop, active learning-driven autonomous system on another major challenge, the discovery of advanced materials against the exceedingly complex synthesis-processes-structure-property landscape. We demonstrate autonomous research methodology (i.e. autonomous hypothesis definition and evaluation) that can place complex, advanced materials in reach, allowing scientists to fail smarter, learn faster, and spend less resources in their studies, while simultaneously improving trust in scientific results and machine learning tools. Additionally, this robot science enables science-over-the-network, reducing the economic impact of scientists being physically separated from their labs. We used the real-time closed-loop, autonomous system for materials exploration and optimization (CAMEO) at the synchrotron beamline to accelerate the fundamentally interconnected tasks of rapid phase mapping and property optimization, with each cycle taking seconds to minutes, resulting in the discovery of a novel epitaxial nanocomposite phase-change memory material."
                },
                {
                    "title": "Uncertainty Estimation Using a Single Deep Deterministic Neural Network",
                    "abstract": "We propose a method for training a deterministic deep model that can find and reject out of distribution data points at test time with a single forward pass. Our approach, deterministic uncertainty quantification (DUQ), builds upon ideas of RBF networks. We scale training in these with a novel loss function and centroid updating scheme and match the accuracy of softmax models. By enforcing detectability of changes in the input using a gradient penalty, we are able to reliably detect out of distribution data. Our uncertainty quantification scales well to large datasets, and using a single model, we improve upon or match Deep Ensembles in out of distribution detection on notable difficult dataset pairs such as FashionMNIST vs. MNIST, and CIFAR-10 vs. SVHN."
                },
                {
                    "title": "Bayesian Deep Learning and a Probabilistic Perspective of Generalization",
                    "abstract": "The key distinguishing property of a Bayesian approach is marginalization, rather than using a single setting of weights. Bayesian marginalization can particularly improve the accuracy and calibration of modern deep neural networks, which are typically underspecified by the data, and can represent many compelling but different solutions. We show that deep ensembles provide an effective mechanism for approximate Bayesian marginalization, and propose a related approach that further improves the predictive distribution by marginalizing within basins of attraction, without significant overhead. We also investigate the prior over functions implied by a vague distribution over neural network weights, explaining the generalization properties of such models from a probabilistic perspective. From this perspective, we explain results that have been presented as mysterious and distinct to neural network generalization, such as the ability to fit images with random labels, and show that these results can be reproduced with Gaussian processes. We also show that Bayesian model averaging alleviates double descent, resulting in monotonic performance improvements with increased flexibility. Finally, we provide a Bayesian perspective on tempering for calibrating predictive distributions."
                },
                {
                    "title": "Fine-Tuning Language Models from Human Preferences",
                    "abstract": "Reward learning enables the application of reinforcement learning (RL) to tasks where reward is defined by human judgment, building a model of reward by asking humans questions. Most work on reward learning has used simulated environments, but complex information about values is often expressed in natural language, and we believe reward learning for language is a key to making RL practical and safe for real-world tasks. In this paper, we build on advances in generative pretraining of language models to apply reward learning to four natural language tasks: continuing text with positive sentiment or physically descriptive language, and summarization tasks on the TL;DR and CNN/Daily Mail datasets. For stylistic continuation we achieve good results with only 5,000 comparisons evaluated by humans. For summarization, models trained with 60,000 comparisons copy whole sentences from the input but skip irrelevant preamble; this leads to reasonable ROUGE scores and very good performance according to our human labelers, but may be exploiting the fact that labelers rely on simple heuristics."
                },
                {
                    "title": "BatchBALD: Efficient and Diverse Batch Acquisition for Deep Bayesian Active Learning",
                    "abstract": "We develop BatchBALD, a tractable approximation to the mutual information between a batch of points and model parameters, which we use as an acquisition function to select multiple informative points jointly for the task of deep Bayesian active learning. BatchBALD is a greedy linear-time $1 - \\frac{1}{e}$-approximate algorithm amenable to dynamic programming and efficient caching. We compare BatchBALD to the commonly used approach for batch data acquisition and find that the current approach acquires similar and redundant points, sometimes performing worse than randomly acquiring data. We finish by showing that, using BatchBALD to consider dependencies within an acquisition batch, we achieve new state of the art performance on standard benchmarks, providing substantial data efficiency improvements in batch acquisition."
                },
                {
                    "title": "Provably Efficient Maximum Entropy Exploration",
                    "abstract": "Suppose an agent is in a (possibly unknown) Markov Decision Process in the absence of a reward signal, what might we hope that an agent can efficiently learn to do? This work studies a broad class of objectives that are defined solely as functions of the state-visitation frequencies that are induced by how the agent behaves. For example, one natural, intrinsically defined, objective problem is for the agent to learn a policy which induces a distribution over state space that is as uniform as possible, which can be measured in an entropic sense. We provide an efficient algorithm to optimize such such intrinsically defined objectives, when given access to a black box planning oracle (which is robust to function approximation). Furthermore, when restricted to the tabular setting where we have sample based access to the MDP, our proposed algorithm is provably efficient, both in terms of its sample and computational complexities. Key to our algorithmic methodology is utilizing the conditional gradient method (a.k.a. the Frank-Wolfe algorithm) which utilizes an approximate MDP solver."
                },
                {
                    "title": "Exploration by Random Network Distillation",
                    "abstract": "We introduce an exploration bonus for deep reinforcement learning methods that is easy to implement and adds minimal overhead to the computation performed. The bonus is the error of a neural network predicting features of the observations given by a fixed randomly initialized neural network. We also introduce a method to flexibly combine intrinsic and extrinsic rewards. We find that the random network distillation (RND) bonus combined with this increased flexibility enables significant progress on several hard exploration Atari games. In particular we establish state of the art performance on Montezuma's Revenge, a game famously difficult for deep reinforcement learning methods. To the best of our knowledge, this is the first method that achieves better than average human performance on this game without using demonstrations or having access to the underlying state of the game, and occasionally completes the first level."
                },
                {
                    "title": "Deep Bayesian Active Learning for Natural Language Processing: Results of a Large-Scale Empirical Study",
                    "abstract": "Several recent papers investigate Active Learning (AL) for mitigating the data dependence of deep learning for natural language processing. However, the applicability of AL to real-world problems remains an open question. While in supervised learning, practitioners can try many different methods, evaluating each against a validation set before selecting a model, AL affords no such luxury. Over the course of one AL run, an agent annotates its dataset exhausting its labeling budget. Thus, given a new task, we have no opportunity to compare models and acquisition functions. This paper provides a large-scale empirical study of deep active learning, addressing multiple tasks and, for each, multiple datasets, multiple models, and a full suite of acquisition functions. We find that across all settings, Bayesian active learning by disagreement, using uncertainty estimates provided either by Dropout or Bayes-by-Backprop significantly improves over i.i.d. baselines and usually outperforms classic uncertainty sampling."
                },
                {
                    "title": "Efficient Parametrization of Multi-domain Deep Neural Networks",
                    "abstract": "A practical limitation of deep neural networks is their high degree of specialization to a single task and visual domain. Recently, inspired by the successes of transfer learning, several authors have proposed to learn instead universal feature extractors that, used as the first stage of any deep network, work well for several tasks and domains simultaneously. Nevertheless, such universal features are still somewhat inferior to specialized networks. To overcome this limitation, in this paper we propose to consider instead universal parametric families of neural networks, which still contain specialized problem-specific models, but differing only by a small number of parameters. We study different designs for such parametrizations, including series and parallel residual adapters, joint adapter compression, and parameter allocations, and empirically identify the ones that yield the highest compression. We show that, in order to maximize performance, it is necessary to adapt both shallow and deep layers of a deep network, but the required changes are very small. We also show that these universal parametrization are very effective for transfer learning, where they outperform traditional fine-tuning techniques."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "Automatic differentiation in PyTorch",
                    "abstract": "In this article, we describe an automatic differentiation module of PyTorch \u2014 a library designed to enable rapid research on machine learning models. It builds upon a few projects, most notably Lua Torch, Chainer, and HIPS Autograd [4], and provides a high performance environment with easy access to automatic differentiation of models executed on different devices (CPU and GPU). To make prototyping easier, PyTorch does not follow the symbolic approach used in many other deep learning frameworks, but focuses on differentiation of purely imperative programs, with a focus on extensibility and low overhead. Note that this preprint is a draft of certain sections from an upcoming paper covering all PyTorch features."
                },
                {
                    "title": "Learning multiple visual domains with residual adapters",
                    "abstract": "There is a growing interest in learning data representations that work well for many different types of problems and data. In this paper, we look in particular at the task of learning a single visual representation that can be successfully utilized in the analysis of very different types of images, from dog breeds to stop signs and digits. Inspired by recent work on learning networks that predict the parameters of another, we develop a tunable deep network architecture that, by means of adapter residual modules, can be steered on the fly to diverse visual domains. Our method achieves a high degree of parameter sharing while maintaining or even improving the accuracy of domain-specific representations. We also introduce the Visual Decathlon Challenge, a benchmark that evaluates the ability of representations to capture simultaneously ten very different visual domains and measures their ability to recognize well uniformly."
                },
                {
                    "title": "Deep Bayesian Active Learning with Image Data",
                    "abstract": "Even though active learning forms an important pillar of machine learning, deep learning tools are not prevalent within it. Deep learning poses several difficulties when used in an active learning setting. First, active learning (AL) methods generally rely on being able to learn and update models from small amounts of data. Recent advances in deep learning, on the other hand, are notorious for their dependence on large amounts of data. Second, many AL acquisition functions rely on model uncertainty, yet deep learning methods rarely represent such model uncertainty. In this paper we combine recent advances in Bayesian deep learning into the active learning framework in a practical way. We develop an active learning framework for high dimensional data, a task which has been extremely challenging so far, with very sparse existing literature. Taking advantage of specialised models such as Bayesian convolutional neural networks, we demonstrate our active learning techniques with image data, obtaining a significant improvement on existing active learning approaches. We demonstrate this on both the MNIST dataset, as well as for skin cancer diagnosis from lesion images (ISIC2016 task)."
                },
                {
                    "title": "Teaching Machines to Read and Comprehend",
                    "abstract": "Teaching machines to read natural language documents remains an elusive challenge. Machine reading systems can be tested on their ability to answer questions posed on the contents of documents that they have seen, but until now large scale training and test datasets have been missing for this type of evaluation. In this work we define a new methodology that resolves this bottleneck and provides large scale supervised reading comprehension data. This allows us to develop a class of attention based deep neural networks that learn to read real documents and answer complex questions with minimal prior knowledge of language structure."
                },
                {
                    "title": "Bayesian Active Learning for Classification and Preference Learning",
                    "abstract": "Information theoretic active learning has been widely studied for probabilistic models. For simple regression an optimal myopic policy is easily tractable. However, for other tasks and with more complex models, such as classication with nonparametric models, the optimal solution is harder to compute. Current approaches make approximations to achieve tractability. We propose an approach that expresses information gain in terms of predictive entropies, and apply this method to the Gaussian Process Classier (GPC). Our approach makes minimal approximations to the full information theoretic objective. Our experimental performance compares favourably to many popular active learning algorithms, and has equal or lower computational complexity. We compare well to decision theoretic approaches also, which are privy to more information and require much more computational time. Secondly, by developing further a reformulation of binary preference learning to a classication problem, we extend our algorithm to Gaussian Process preference learning."
                },
                {
                    "title": "Batch mode active learning and its application to medical image classification",
                    "abstract": "The goal of active learning is to select the most informative examples for manual labeling. Most of the previous studies in active learning have focused on selecting a single unlabeled example in each iteration. This could be inefficient since the classification model has to be retrained for every labeled example. In this paper, we present a framework for \"batch mode active learning\" that applies the Fisher information matrix to select a number of informative examples simultaneously. The key computational challenge is how to efficiently identify the subset of unlabeled examples that can result in the largest reduction in the Fisher information. To resolve this challenge, we propose an efficient greedy algorithm that is based on the property of submodular functions. Our empirical studies with five UCI datasets and one real-world medical image classification show that the proposed batch mode active learning algorithm is more effective than the state-of-the-art algorithms for active learning."
                },
                {
                    "title": "Preference learning with Gaussian processes",
                    "abstract": "In this paper, we propose a probabilistic kernel approach to preference learning based on Gaussian processes. A new likelihood function is proposed to capture the preference relations in the Bayesian framework. The generalized formulation is also applicable to tackle many multiclass problems. The overall approach has the advantages of Bayesian methods for model selection and probabilistic prediction. Experimental results compared against the constraint classification approach on several benchmark datasets verify the usefulness of this algorithm."
                },
                {
                    "title": "Estimating mutual information.",
                    "abstract": "We present two classes of improved estimators for mutual information M(X,Y), from samples of random points distributed according to some joint probability density mu(x,y). In contrast to conventional estimators based on binnings, they are based on entropy estimates from k -nearest neighbor distances. This means that they are data efficient (with k=1 we resolve structures down to the smallest possible scales), adaptive (the resolution is higher where data are more numerous), and have minimal bias. Indeed, the bias of the underlying entropy estimates is mainly due to nonuniformity of the density at the smallest resolved scale, giving typically systematic errors which scale as functions of k/N for N points. Numerically, we find that both families become exact for independent distributions, i.e. the estimator M(X,Y) vanishes (up to statistical fluctuations) if mu(x,y)=mu(x)mu(y). This holds for all tested marginal distributions and for all dimensions of x and y. In addition, we give estimators for redundancies between more than two random variables. We compare our algorithms in detail with existing algorithms. Finally, we demonstrate the usefulness of our estimators for assessing the actual independence of components obtained from independent component analysis (ICA), for improving ICA, and for estimating the reliability of blind source separation."
                },
                {
                    "title": "Information-Based Objective Functions for Active Data Selection",
                    "abstract": "Learning can be made more efficient if we can actively select particularly salient data points. Within a Bayesian learning framework, objective functions are discussed that measure the expected informativeness of candidate measurements. Three alternative specifications of what we want to gain information about lead to three different criteria for data selection. All these criteria depend on the assumption that the hypothesis space is correct, which may prove to be their main weakness."
                },
                {
                    "title": "Optimal Algorithms for Stochastic Contextual Preference Bandits",
                    "abstract": "We consider the problem of preference bandits in the contextual setting. At each round, the learner is presented with a context set of K items, chosen randomly from a potentially in\ufb01nite set of arms D \u2286 R d . However, unlike classical contextual bandits, our framework only allows the learner to receive feedback in terms of item preferences: At each round, the learner is allowed to play a subset of size q (any q \u2208 { 2 , . . . , K } ) upon which only a (noisy) winner of the subset is revealed. Yet, same as the classical setup, the goal is still to compete against the best context arm at each round. The problem is relevant in various online decision-making scenarios, including recommender systems, information retrieval, tournament ranking\u2013typically any application where it\u2019s easier to elicit the items\u2019 relative strength instead of their absolute scores. To the best of our knowledge, this work is the \ufb01rst to consider preference-based stochastic contextual bandits for potentially in\ufb01nite decision spaces. We start with presenting two algorithms for the special case of pairwise preferences ( q = 2) : The \ufb01rst algorithm is simple and easy to implement with an \u02dc O ( d \u221a T ) regret guarantee, while the second algorithm is shown to achieve the optimal \u02dc O ( \u221a dT ) regret, as follows from our \u2126( \u221a dT ) matching lower bound analysis. We then proceed to analyze the problem for any general q -subsetwise preferences ( q \u2265 2 ), where surprisingly, our lower bound proves the fundamental performance limit to be \u2126( \u221a dT ) yet again, independent of the subsetsize q . Following this, we propose a matching upper bound algorithm justifying the tightness of our results. This implies having access to subsetwise preferences does not help in faster information aggregation for our feedback model. All the results are corroborated empirically against existing baselines."
                },
                {
                    "title": "Dueling Bandits: Beyond Condorcet Winners to General Tournament Solutions",
                    "abstract": "Recent work on deriving $O(\\log T)$ anytime regret bounds for stochastic dueling bandit problems has considered mostly Condorcet winners, which do not always exist, and more recently, winners defined by the Copeland set, which do always exist. In this work, we consider a broad notion of winners defined by tournament solutions in social choice theory, which include the Copeland set as a special case but also include several other notions of winners such as the top cycle, uncovered set, and Banks set, and which, like the Copeland set, always exist. We develop a family of UCB-style dueling bandit algorithms for such general tournament solutions, and show $O(\\log T)$ anytime regret bounds for them. Experiments confirm the ability of our algorithms to achieve low regret relative to the target winning set of interest."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CL",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively select informative prompt-answer pairs for Preference Modeling in Large Language Models (LLMs) to optimize human feedback collection while addressing the challenges of noise and scale?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning, particularly in the context of aligning LLMs with human preferences. By improving the efficiency and effectiveness of human feedback collection, we can enhance the performance of LLMs, leading to more reliable and user-aligned models. This research could pave the way for practical applications in various domains, such as personalized content generation, customer service automation, and interactive AI systems. Furthermore, it will stimulate future research on active learning techniques and uncertainty estimation in LLMs, potentially leading to breakthroughs in how we train and deploy these models.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe problem is challenging due to several complexities: the vast and semantically rich pool of prompt-answer pairs makes it difficult to identify the most informative samples. Additionally, human feedback is often noisy, with low agreement rates among labelers, complicating the data selection process. Naive approaches may fail because they do not account for the intrinsic scale of LLM development, which requires parallelized labeling and frequent model updates. Moreover, existing active learning schemes that rely on single-point acquisition are impractical in this context, as they can lead to redundant sample selection and do not effectively leverage the uncertainty inherent in the data.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has faced limitations such as inadequate uncertainty estimation methods for LLMs and the inherent biases in approximate Bayesian inference in deep learning models. Past attempts to apply Bayesian Active Learning in Preference Modeling have not yielded benefits over random selection, primarily due to these challenges. Additionally, the combinatorial complexity of proper estimators for batch acquisition in this context has hindered progress. Our approach aims to address these gaps by developing a methodology that effectively estimates epistemic uncertainty and optimizes data selection for batch acquisition, thus improving upon prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Bayesian Active Learning for Preference Modeling (BAL-PM), involves leveraging the last layer embeddings of a base LLM to compute features for prompt and prompt-answer pairs. We will estimate the entropy score of the prompt distribution and use a Bayesian"
            }
        },
        "author_data": {
            "8529d773-0a12-4a8b-b250-2be1eff63efb": {
                "pk": "8529d773-0a12-4a8b-b250-2be1eff63efb",
                "name": "Luckeciano C. Melo",
                "collaborators": [
                    "Marcos R. O. A. Maximo",
                    "Bruno Brand\u00e3o",
                    "Alessandro Abate",
                    "Yarin Gal",
                    "Adilson Marques da Cunha",
                    "Marlesson R. O. Santana",
                    "Fernando H. F. Camargo",
                    "Anderson Soares",
                    "Renan M. Oliveira",
                    "Sandor Caetano"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Meta-Learning",
                    "Continual Learning",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "Transformers are Meta-Reinforcement Learners",
                        "abstract": "The transformer architecture and variants presented remarkable success across many machine learning tasks in recent years. This success is intrinsically related to the capability of handling long sequences and the presence of context-dependent weights from the attention mechanism. We argue that these capabilities suit the central role of a Meta-Reinforcement Learning algorithm. Indeed, a meta-RL agent needs to infer the task from a sequence of trajectories. Furthermore, it requires a fast adaptation strategy to adapt its policy for a new task -- which can be achieved using the self-attention mechanism. In this work, we present TrMRL (Transformers for Meta-Reinforcement Learning), a meta-RL agent that mimics the memory reinstatement mechanism using the transformer architecture. It associates the recent past of working memories to build an episodic memory recursively through the transformer layers. We show that the self-attention computes a consensus representation that minimizes the Bayes Risk at each layer and provides meaningful features to compute the best actions. We conducted experiments in high-dimensional continuous control environments for locomotion and dexterous manipulation. Results show that TrMRL presents comparable or superior asymptotic performance, sample efficiency, and out-of-distribution generalization compared to the baselines in these environments."
                    },
                    {
                        "title": "Temporal-Difference Variational Continual Learning",
                        "abstract": "A crucial capability of Machine Learning models in real-world applications is the ability to continuously learn new tasks. This adaptability allows them to respond to potentially inevitable shifts in the data-generating distribution over time. However, in Continual Learning (CL) settings, models often struggle to balance learning new tasks (plasticity) with retaining previous knowledge (memory stability). Consequently, they are susceptible to Catastrophic Forgetting, which degrades performance and undermines the reliability of deployed systems. Variational Continual Learning methods tackle this challenge by employing a learning objective that recursively updates the posterior distribution and enforces it to stay close to the latest posterior estimate. Nonetheless, we argue that these methods may be ineffective due to compounding approximation errors over successive recursions. To mitigate this, we propose new learning objectives that integrate the regularization effects of multiple previous posterior estimations, preventing individual errors from dominating future posterior updates and compounding over time. We reveal insightful connections between these objectives and Temporal-Difference methods, a popular learning mechanism in Reinforcement Learning and Neuroscience. We evaluate the proposed objectives on challenging versions of popular CL benchmarks, demonstrating that they outperform standard Variational CL methods and non-variational baselines, effectively alleviating Catastrophic Forgetting."
                    },
                    {
                        "title": "Learning Humanoid Robot Running Skills through Proximal Policy Optimization",
                        "abstract": "In the current level of evolution of Soccer 3D, motion control is a key factor in team's performance. Recent works takes advantages of model-free approaches based on Machine Learning to exploit robot dynamics in order to obtain faster locomotion skills, achieving running policies and, therefore, opening a new research direction in the Soccer 3D environment.   In this work, we present a methodology based on Deep Reinforcement Learning that learns running skills without any prior knowledge, using a neural network whose inputs are related to robot's dynamics. Our results outperformed the previous state-of-the-art sprint velocity reported in Soccer 3D literature by a significant margin. It also demonstrated improvement in sample efficiency, being able to learn how to run in just few hours.   We reported our results analyzing the training procedure and also evaluating the policies in terms of speed, reliability and human similarity. Finally, we presented key factors that lead us to improve previous results and shared some ideas for future work."
                    },
                    {
                        "title": "Bottom-Up Meta-Policy Search",
                        "abstract": "Despite of the recent progress in agents that learn through interaction, there are several challenges in terms of sample efficiency and generalization across unseen behaviors during training. To mitigate these problems, we propose and apply a first-order Meta-Learning algorithm called Bottom-Up Meta-Policy Search (BUMPS), which works with two-phase optimization procedure: firstly, in a meta-training phase, it distills few expert policies to create a meta-policy capable of generalizing knowledge to unseen tasks during training; secondly, it applies a fast adaptation strategy named Policy Filtering, which evaluates few policies sampled from the meta-policy distribution and selects which best solves the task. We conducted all experiments in the RoboCup 3D Soccer Simulation domain, in the context of kick motion learning. We show that, given our experimental setup, BUMPS works in scenarios where simple multi-task Reinforcement Learning does not. Finally, we performed experiments in a way to evaluate each component of the algorithm."
                    },
                    {
                        "title": "MARS-Gym: A Gym framework to model, train, and evaluate Recommender Systems for Marketplaces",
                        "abstract": "Recommender Systems are especially challenging for marketplaces since they must maximize user satisfaction while maintaining the healthiness and fairness of such ecosystems. In this context, we observed a lack of resources to design, train, and evaluate agents that learn by interacting within these environments. For this matter, we propose MARS-Gym, an open-source framework to empower researchers and engineers to quickly build and evaluate Reinforcement Learning agents for recommendations in marketplaces. MARS-Gym addresses the whole development pipeline: data processing, model design and optimization, and multi-sided evaluation. We also provide the implementation of a diverse set of baseline agents, with a metrics-driven analysis of them in the Trivago marketplace dataset, to illustrate how to conduct a holistic assessment using the available metrics of recommendation, off-policy estimation, and fairness. With MARS-Gym, we expect to bridge the gap between academic research and production systems, as well as to facilitate the design of new algorithms and applications."
                    },
                    {
                        "title": "Sliding Puzzles Gym: A Scalable Benchmark for State Representation in Visual Reinforcement Learning",
                        "abstract": "Learning effective visual representations is crucial in open-world environments where agents encounter diverse and unstructured observations. This ability enables agents to extract meaningful information from raw sensory inputs, like pixels, which is essential for generalization across different tasks. However, evaluating representation learning separately from policy learning remains a challenge in most reinforcement learning (RL) benchmarks. To address this, we introduce the Sliding Puzzles Gym (SPGym), a benchmark that extends the classic 15-tile puzzle with variable grid sizes and observation spaces, including large real-world image datasets. SPGym allows scaling the representation learning challenge while keeping the latent environment dynamics and algorithmic problem fixed, providing a targeted assessment of agents' ability to form compositional and generalizable state representations. Experiments with both model-free and model-based RL algorithms, with and without explicit representation learning components, show that as the representation challenge scales, SPGym effectively distinguishes agents based on their capabilities. Moreover, SPGym reaches difficulty levels where no tested algorithm consistently excels, highlighting key challenges and opportunities for advancing representation learning for decision-making research."
                    }
                ]
            },
            "71529356-eea2-498b-b863-b3b63eaa61ca": {
                "pk": "71529356-eea2-498b-b863-b3b63eaa61ca",
                "name": "Panagiotis Tigas",
                "collaborators": [
                    "Yarin Gal",
                    "Yashas Annadani",
                    "Stefan Bauer",
                    "Andrew Jesson",
                    "Adam Foster",
                    "Noor Sajid",
                    "Zafeirios Fountas",
                    "Alexey Zakharov",
                    "Tyson Hosmer",
                    "Tim Murray-Browne"
                ],
                "domain": [
                    "Causal Inference",
                    "Reinforcement Learning",
                    "Machine Learning",
                    "Experimental Design"
                ],
                "publications": [
                    {
                        "title": "Real-time jam-session support system",
                        "abstract": "We propose a method for the problem of real time chord accompaniment of improvised music. Our implementation can learn an underlying structure of the musical performance and predict next chord. The system uses Hidden Markov Model to find the most probable chord sequence for the played melody and then a Variable Order Markov Model is used to a) learn the structure (if any) and b) predict next chord. We implemented our system in Java and MAX/Msp and compared and evaluated using objective (prediction accuracy) and subjective (questionnaire) evaluation methods."
                    },
                    {
                        "title": "Spatial Assembly: Generative Architecture With Reinforcement Learning, Self Play and Tree Search",
                        "abstract": "With this work, we investigate the use of Reinforcement Learning (RL) for the generation of spatial assemblies, by combining ideas from Procedural Generation algorithms (Wave Function Collapse algorithm (WFC)) and RL for Game Solving. WFC is a Generative Design algorithm, inspired by Constraint Solving. In WFC, one defines a set of tiles/blocks and constraints and the algorithm generates an assembly that satisfies these constraints. Casting the problem of generation of spatial assemblies as a Markov Decision Process whose states transitions are defined by WFC, we propose an algorithm that uses Reinforcement Learning and Self-Play to learn a policy that generates assemblies that maximize objectives set by the designer. Finally, we demonstrate the use of our Spatial Assembly algorithm in Architecture Design."
                    },
                    {
                        "title": "Latent Mappings: Generating Open-Ended Expressive Mappings Using Variational Autoencoders",
                        "abstract": "In many contexts, creating mappings for gestural interactions can form part of an artistic process. Creators seeking a mapping that is expressive, novel, and affords them a sense of authorship may not know how to program it up in a signal processing patch. Tools like Wekinator and MIMIC allow creators to use supervised machine learning to learn mappings from example input/output pairings. However, a creator may know a good mapping when they encounter it yet start with little sense of what the inputs or outputs should be. We call this an open-ended mapping process. Addressing this need, we introduce the latent mapping, which leverages the latent space of an unsupervised machine learning algorithm such as a Variational Autoencoder trained on a corpus of unlabelled gestural data from the creator. We illustrate it with Sonified Body, a system mapping full-body movement to sound which we explore in a residency with three dancers."
                    },
                    {
                        "title": "Amortized Active Causal Induction with Deep Reinforcement Learning",
                        "abstract": "We present Causal Amortized Active Structure Learning (CAASL), an active intervention design policy that can select interventions that are adaptive, real-time and that does not require access to the likelihood. This policy, an amortized network based on the transformer, is trained with reinforcement learning on a simulator of the design environment, and a reward function that measures how close the true causal graph is to a causal graph posterior inferred from the gathered data. On synthetic data and a single-cell gene expression simulator, we demonstrate empirically that the data acquired through our policy results in a better estimate of the underlying causal graph than alternative strategies. Our design policy successfully achieves amortized intervention design on the distribution of the training environment while also generalizing well to distribution shifts in test-time design environments. Further, our policy also demonstrates excellent zero-shot generalization to design environments with dimensionality higher than that during training, and to intervention types that it has not been trained on."
                    },
                    {
                        "title": "Interventions, Where and How? Experimental Design for Causal Models at Scale",
                        "abstract": "Causal discovery from observational and interventional data is challenging due to limited data and non-identifiability: factors that introduce uncertainty in estimating the underlying structural causal model (SCM). Selecting experiments (interventions) based on the uncertainty arising from both factors can expedite the identification of the SCM. Existing methods in experimental design for causal discovery from limited data either rely on linear assumptions for the SCM or select only the intervention target. This work incorporates recent advances in Bayesian causal discovery into the Bayesian optimal experimental design framework, allowing for active causal discovery of large, nonlinear SCMs while selecting both the interventional target and the value. We demonstrate the performance of the proposed method on synthetic graphs (Erdos-R\\`enyi, Scale Free) for both linear and nonlinear SCMs as well as on the \\emph{in-silico} single-cell gene regulatory network dataset, DREAM."
                    },
                    {
                        "title": "Modelling non-reinforced preferences using selective attention",
                        "abstract": "How can artificial agents learn non-reinforced preferences to continuously adapt their behaviour to a changing environment? We decompose this question into two challenges: ($i$) encoding diverse memories and ($ii$) selectively attending to these for preference formation. Our proposed \\emph{no}n-\\emph{re}inforced preference learning mechanism using selective attention, \\textsc{Nore}, addresses both by leveraging the agent's world model to collect a diverse set of experiences which are interleaved with imagined roll-outs to encode memories. These memories are selectively attended to, using attention and gating blocks, to update agent's preferences. We validate \\textsc{Nore} in a modified OpenAI Gym FrozenLake environment (without any external signal) with and without volatility under a fixed model of the environment -- and compare its behaviour to \\textsc{Pepper}, a Hebbian preference learning mechanism. We demonstrate that \\textsc{Nore} provides a straightforward framework to induce exploratory preferences in the absence of external signals."
                    },
                    {
                        "title": "Global Earth Magnetic Field Modeling and Forecasting with Spherical Harmonics Decomposition",
                        "abstract": "Modeling and forecasting the solar wind-driven global magnetic field perturbations is an open challenge. Current approaches depend on simulations of computationally demanding models like the Magnetohydrodynamics (MHD) model or sampling spatially and temporally through sparse ground-based stations (SuperMAG). In this paper, we develop a Deep Learning model that forecasts in Spherical Harmonics space 2, replacing reliance on MHD models and providing global coverage at one minute cadence, improving over the current state-of-the-art which relies on feature engineering. We evaluate the performance in SuperMAG dataset (improved by 14.53%) and MHD simulations (improved by 24.35%). Additionally, we evaluate the extrapolation performance of the spherical harmonics reconstruction based on sparse ground-based stations (SuperMAG), showing that spherical harmonics can reliably reconstruct the global magnetic field as evaluated on MHD simulation."
                    },
                    {
                        "title": "Can Autonomous Vehicles Identify, Recover From, and Adapt to Distribution Shifts?",
                        "abstract": "Out-of-training-distribution (OOD) scenarios are a common challenge of learning agents at deployment, typically leading to arbitrary deductions and poorly-informed decisions. In principle, detection of and adaptation to OOD scenes can mitigate their adverse effects. In this paper, we highlight the limitations of current approaches to novel driving scenes and propose an epistemic uncertainty-aware planning method, called \\emph{robust imitative planning} (RIP). Our method can detect and recover from some distribution shifts, reducing the overconfident and catastrophic extrapolations in OOD scenes. If the model's uncertainty is too great to suggest a safe course of action, the model can instead query the expert driver for feedback, enabling sample-efficient online adaptation, a variant of our method we term \\emph{adaptive robust imitative planning} (AdaRIP). Our methods outperform current state-of-the-art approaches in the nuScenes \\emph{prediction} challenge, but since no benchmark evaluating OOD detection and adaption currently exists to assess \\emph{control}, we introduce an autonomous car novel-scene benchmark, \\texttt{CARNOVEL}, to evaluate the robustness of driving agents to a suite of tasks with distribution shifts."
                    },
                    {
                        "title": "Exploration and preference satisfaction trade-off in reward-free learning",
                        "abstract": "Biological agents have meaningful interactions with their environment despite the absence of immediate reward signals. In such instances, the agent can learn preferred modes of behaviour that lead to predictable states -- necessary for survival. In this paper, we pursue the notion that this learnt behaviour can be a consequence of reward-free preference learning that ensures an appropriate trade-off between exploration and preference satisfaction. For this, we introduce a model-based Bayesian agent equipped with a preference learning mechanism (pepper) using conjugate priors. These conjugate priors are used to augment the expected free energy planner for learning preferences over states (or outcomes) across time. Importantly, our approach enables the agent to learn preferences that encourage adaptive behaviour at test time. We illustrate this in the OpenAI Gym FrozenLake and the 3D mini-world environments -- with and without volatility. Given a constant environment, these agents learn confident (i.e., precise) preferences and act to satisfy them. Conversely, in a volatile setting, perpetual preference uncertainty maintains exploratory behaviour. Our experiments suggest that learnable (reward-free) preferences entail a trade-off between exploration and preference satisfaction. Pepper offers a straightforward framework suitable for designing adaptive agents when reward functions cannot be predefined as in real environments."
                    },
                    {
                        "title": "Differentiable Multi-Target Causal Bayesian Experimental Design",
                        "abstract": "We introduce a gradient-based approach for the problem of Bayesian optimal experimental design to learn causal models in a batch setting -- a critical component for causal discovery from finite data where interventions can be costly or risky. Existing methods rely on greedy approximations to construct a batch of experiments while using black-box methods to optimize over a single target-state pair to intervene with. In this work, we completely dispose of the black-box optimization techniques and greedy heuristics and instead propose a conceptually simple end-to-end gradient-based optimization procedure to acquire a set of optimal intervention target-state pairs. Such a procedure enables parameterization of the design space to efficiently optimize over a batch of multi-target-state interventions, a setting which has hitherto not been explored due to its complexity. We demonstrate that our proposed method outperforms baselines and existing acquisition strategies in both single-target and multi-target settings across a number of synthetic datasets."
                    },
                    {
                        "title": "Causal-BALD: Deep Bayesian Active Learning of Outcomes to Infer Treatment-Effects from Observational Data",
                        "abstract": "Estimating personalized treatment effects from high-dimensional observational data is essential in situations where experimental designs are infeasible, unethical, or expensive. Existing approaches rely on fitting deep models on outcomes observed for treated and control populations. However, when measuring individual outcomes is costly, as is the case of a tumor biopsy, a sample-efficient strategy for acquiring each result is required. Deep Bayesian active learning provides a framework for efficient data acquisition by selecting points with high uncertainty. However, existing methods bias training data acquisition towards regions of non-overlapping support between the treated and control populations. These are not sample-efficient because the treatment effect is not identifiable in such regions. We introduce causal, Bayesian acquisition functions grounded in information theory that bias data acquisition towards regions with overlapping support to maximize sample efficiency for learning personalized treatment effects. We demonstrate the performance of the proposed acquisition strategies on synthetic and semi-synthetic datasets IHDP and CMNIST and their extensions, which aim to simulate common dataset biases and pathologies."
                    },
                    {
                        "title": "Challenges and Considerations in the Evaluation of Bayesian Causal Discovery",
                        "abstract": "Representing uncertainty in causal discovery is a crucial component for experimental design, and more broadly, for safe and reliable causal decision making. Bayesian Causal Discovery (BCD) offers a principled approach to encapsulating this uncertainty. Unlike non-Bayesian causal discovery, which relies on a single estimated causal graph and model parameters for assessment, evaluating BCD presents challenges due to the nature of its inferred quantity - the posterior distribution. As a result, the research community has proposed various metrics to assess the quality of the approximate posterior. However, there is, to date, no consensus on the most suitable metric(s) for evaluation. In this work, we reexamine this question by dissecting various metrics and understanding their limitations. Through extensive empirical evaluation, we find that many existing metrics fail to exhibit a strong correlation with the quality of approximation to the true posterior, especially in scenarios with low sample sizes where BCD is most desirable. We highlight the suitability (or lack thereof) of these metrics under two distinct factors: the identifiability of the underlying causal model and the quantity of available data. Both factors affect the entropy of the true posterior, indicating that the current metrics are less fitting in settings of higher entropy. Our findings underline the importance of a more nuanced evaluation of new methods by taking into account the nature of the true posterior, as well as guide and motivate the development of new evaluation procedures for this challenge."
                    }
                ]
            },
            "8a5153b3-6cf3-49a0-aec4-7b8d0c2032e7": {
                "pk": "8a5153b3-6cf3-49a0-aec4-7b8d0c2032e7",
                "name": "Alessandro Abate",
                "collaborators": [
                    "Ilya Tkachev",
                    "Joar Skalse",
                    "Alessandro Cimatti",
                    "Andrea Micheli",
                    "Muhammad Syifa'ul Mufid",
                    "Sofie Haesaert",
                    "Patrick Benjamin",
                    "Francesco Cosentino",
                    "Harald Oberhauser",
                    "Frank Redig"
                ],
                "domain": [
                    "Stochastic Processes",
                    "Reinforcement Learning",
                    "Markov Chains",
                    "Control Systems"
                ],
                "publications": [
                    {
                        "title": "On the effect of perturbation of conditional probabilities in total variation",
                        "abstract": "A celebrated result by A. Ionescu Tulcea provides a construction of a probability measure on a product space given a sequence of regular conditional probabilities. We study how the perturbations of the latter in the total variation metric affect the resulting product probability measure."
                    },
                    {
                        "title": "Characterization and computation of infinite horizon specifications over Markov processes",
                        "abstract": "This work is devoted to the formal verification of specifications over general discrete-time Markov processes, with an emphasis on infinite-horizon properties. These properties, formulated in a modal logic known as PCTL, can be expressed through value functions defined over the state space of the process. The main goal is to understand how structural features of the model (primarily the presence of absorbing sets) influence the uniqueness of the solutions of corresponding Bellman equations. Furthermore, this contribution shows that the investigation of these structural features leads to new computational techniques to calculate the specifications of interest: the emphasis is to derive approximation techniques with associated explicit convergence rates and formal error bounds."
                    },
                    {
                        "title": "Computation of ruin probabilities for general discrete-time Markov models",
                        "abstract": "We study the ruin problem over a risk process described by a discrete-time Markov model. In contrast to previous studies that focused on the asymptotic behaviour of ruin probabilities for large values of the initial capital, we provide a new technique to compute the quantity of interest for any initial value, and with any given precision. Rather than focusing on a particular model for risk processes, we give a general characterization of the ruin probability by providing corresponding recursions and fixpoint equations. Since such equations for the ruin probability are ill-posed in the sense that they do not allow for unique solutions, we approximate the ruin probability by a two-barrier ruin probability, for which fixpoint equations are well-posed. We also show how good the introduced approximation is by providing an explicit bound on the error and by characterizing the cases when the error converges to zero. The presented technique and results are supported by two computational examples over models known in the literature, one of which is extremely heavy-tailed."
                    },
                    {
                        "title": "Safety Verification of Output Feedback Controllers for Nonlinear Systems",
                        "abstract": "A high-gain observer is used for a class of feedback linearisable nonlinear systems to synthesize safety-preserving controllers over the observer output. A bound on the distance between trajectories under state and output feedback is derived, and shown to converge to zero as a function of the gain parameter of an observer. We can therefore recover safety properties under output feedback and control saturation constraints by synthesizing a controller as if the full state were available. We specifically design feedback linearising controllers that satisfy certain properties, such as stability, and then construct the associated maximal safety-invariant set, namely the largest set of all initial states that are guaranteed to produce safe trajectories over a given (possibly infinite) time horizon."
                    },
                    {
                        "title": "On the Relationship between Bisimulation and Trace Equivalence in an Approximate Probabilistic Context (Extended Version)",
                        "abstract": "This work introduces a notion of approximate probabilistic trace equivalence for labelled Markov chains, and relates this new concept to the known notion of approximate probabilistic bisimulation. In particular this work shows that the latter notion induces a tight upper bound on the approximation between finite-horizon traces, as expressed by a total variation distance. As such, this work extends corresponding results for exact notions and analogous results for non-probabilistic models. This bound can be employed to relate the closeness in satisfaction probabilities over bounded linear-time properties, and allows for probabilistic model checking of concrete models via abstractions. The contribution focuses on both finite-state and uncountable-state labelled Markov chains, and claims two main applications: firstly, it allows an upper bound on the trace distance to be decided for finite state systems; secondly, it can be used to synthesise discrete approximations to continuous-state models with arbitrary precision."
                    },
                    {
                        "title": "Benchmarks for cyber-physical systems: A modular model library for building automation systems (Extended version)",
                        "abstract": "Building Automation Systems (BAS) are exemplars of Cyber-Physical Systems (CPS), incorporating digital control architectures over underlying continuous physical processes. We provide a modular model library for BAS drawn from expertise developed on a real BAS setup. The library allows to build models comprising of either physical quantities or digital control modules.% which are composable. The structure, operation, and dynamics of the model can be complex, incorporating (i) stochasticity, (ii) non-linearities, (iii) numerous continuous variables or discrete states, (iv) various input and output signals, and (v) a large number of possible discrete configurations. The modular composition of BAS components can generate useful CPS benchmarks. We display this use by means of three realistic case studies, where corresponding models are built and engaged with different analysis goals. The benchmarks, the model library and data collected from the BAS setup at the University of Oxford, are kept on-line at https://github.com/natchi92/BASBenchmarks."
                    },
                    {
                        "title": "Misspecification in Inverse Reinforcement Learning",
                        "abstract": "The aim of Inverse Reinforcement Learning (IRL) is to infer a reward function $R$ from a policy $\\pi$. To do this, we need a model of how $\\pi$ relates to $R$. In the current literature, the most common models are optimality, Boltzmann rationality, and causal entropy maximisation. One of the primary motivations behind IRL is to infer human preferences from human behaviour. However, the true relationship between human preferences and human behaviour is much more complex than any of the models currently used in IRL. This means that they are misspecified, which raises the worry that they might lead to unsound inferences if applied to real-world data. In this paper, we provide a mathematical analysis of how robust different IRL models are to misspecification, and answer precisely how the demonstrator policy may differ from each of the standard models before that model leads to faulty inferences about the reward function $R$. We also introduce a framework for reasoning about misspecification in IRL, together with formal tools that can be used to easily derive the misspecification robustness of new IRL models."
                    },
                    {
                        "title": "An Exact Characterisation of Flexibility in Populations of Electric Vehicles",
                        "abstract": "Increasing penetrations of electric vehicles (EVs) presents a large source of flexibility, which can be used to assist balancing the power grid. The flexibility of an individual EV can be quantified as a convex polytope and the flexibility of a population of EVs is the Minkowski sum of these polytopes. In general computing the exact Minkowski sum is intractable. However, exploiting symmetry in a restricted but significant case, enables an efficient computation of the aggregate flexibility. This results in a polytope with exponentially many vertices and facets with respect to the time horizon. We show how to use a lifting procedure to provide a representation of this polytope with a reduced number of facets, which makes optimising over more tractable. Finally, a disaggregation procedure that takes an aggregate signal and computes dispatch instructions for each EV in the population is presented. The complexity of the algorithms presented is independent of the size of the population and polynomial in the length of the time horizon. We evaluate this work against existing methods in the literature, and show how this method guarantees optimality with lower computational burden than existing methods."
                    },
                    {
                        "title": "On the Limitations of Markovian Rewards to Express Multi-Objective, Risk-Sensitive, and Modal Tasks",
                        "abstract": "In this paper, we study the expressivity of scalar, Markovian reward functions in Reinforcement Learning (RL), and identify several limitations to what they can express. Specifically, we look at three classes of RL tasks; multi-objective RL, risk-sensitive RL, and modal RL. For each class, we derive necessary and sufficient conditions that describe when a problem in this class can be expressed using a scalar, Markovian reward. Moreover, we find that scalar, Markovian rewards are unable to express most of the instances in each of these three classes. We thereby contribute to a more complete understanding of what standard reward functions can and cannot express. In addition to this, we also call attention to modal problems as a new class of problems, since they have so far not been given any systematic treatment in the RL literature. We also briefly outline some approaches for solving some of the problems we discuss, by means of bespoke RL algorithms."
                    },
                    {
                        "title": "DeepLTL: Learning to Efficiently Satisfy Complex LTL Specifications",
                        "abstract": "Linear temporal logic (LTL) has recently been adopted as a powerful formalism for specifying complex, temporally extended tasks in reinforcement learning (RL). However, learning policies that efficiently satisfy arbitrary specifications not observed during training remains a challenging problem. Existing approaches suffer from several shortcomings: they are often only applicable to finite-horizon fragments of LTL, are restricted to suboptimal solutions, and do not adequately handle safety constraints. In this work, we propose a novel learning approach to address these concerns. Our method leverages the structure of B\\\"uchi automata, which explicitly represent the semantics of LTL specifications, to learn policies conditioned on sequences of truth assignments that lead to satisfying the desired formulae. Experiments in a variety of discrete and continuous domains demonstrate that our approach is able to zero-shot satisfy a wide range of finite- and infinite-horizon specifications, and outperforms existing methods in terms of both satisfaction probability and efficiency."
                    },
                    {
                        "title": "Computation of the Transient in Max-Plus Linear Systems via SMT-Solving",
                        "abstract": "This paper proposes a new approach, grounded in Satisfiability Modulo Theories (SMT), to study the transient of a Max-Plus Linear (MPL) system, that is the number of steps leading to its periodic regime. Differently from state-of-the-art techniques, our approach allows the analysis of periodic behaviors for subsets of initial states, as well as the characterization of sets of initial states exhibiting the same specific periodic behavior and transient. Our experiments show that the proposed technique dramatically outperforms state-of-the-art methods based on max-plus algebra computations for systems of large dimensions."
                    },
                    {
                        "title": "Formal Analysis and Verification of Max-Plus Linear Systems",
                        "abstract": "Max-Plus Linear (MPL) systems are an algebraic formalism with practical applications in transportation networks, manufacturing and biological systems. In this paper, we investigate the problem of automatically analyzing the properties of MPL, taking into account both structural properties such as transient and cyclicity, and the open problem of user-defined temporal properties. We propose Time-Difference LTL (TDLTL), a logic that encompasses the delays between the discrete time events governed by an MPL system, and characterize the problem of model checking TDLTL over MPL. We first consider a framework based on the verification of infinite-state transition systems, and propose an approach based on an encoding into model checking. Then, we leverage the specific features of MPL systems to devise a highly optimized, combinational approach based on Satisfiability Modulo Theory (SMT). We experimentally evaluate the features of the proposed approaches on a large set of benchmarks. The results show that the proposed approach substantially outperforms the state of the art competitors in expressiveness and effectiveness, and demonstrate the superiority of the combinational approach over the reduction to model checking."
                    },
                    {
                        "title": "Sampling-based Approximations with Quantitative Performance for the Probabilistic Reach-Avoid Problem over General Markov Processes",
                        "abstract": "This article deals with stochastic processes endowed with the Markov (memoryless) property and evolving over general (uncountable) state spaces. The models further depend on a non-deterministic quantity in the form of a control input, which can be selected to affect the probabilistic dynamics. We address the computation of maximal reach-avoid specifications, together with the synthesis of the corresponding optimal controllers. The reach-avoid specification deals with assessing the likelihood that any finite-horizon trajectory of the model enters a given goal set, while avoiding a given set of undesired states. This article newly provides an approximate computational scheme for the reach-avoid specification based on the Fitted Value Iteration algorithm, which hinges on random sample extractions, and gives a-priori computable formal probabilistic bounds on the error made by the approximation algorithm: as such, the output of the numerical scheme is quantitatively assessed and thus meaningful for safety-critical applications. Furthermore, we provide tighter probabilistic error bounds that are sample-based. The overall computational scheme is put in relationship with alternative approximation algorithms in the literature, and finally its performance is practically assessed over a benchmark case study."
                    },
                    {
                        "title": "Networked Communication for Decentralised Agents in Mean-Field Games",
                        "abstract": "We introduce networked communication to the mean-field game framework, in particular to oracle-free settings where $N$ decentralised agents learn along a single, non-episodic run of the empirical system. We prove that our architecture has sample guarantees bounded between those of the centralised- and independent-learning cases. We provide the order of the difference in these bounds in terms of network structure and number of communication rounds, and also contribute a policy-update stability guarantee. We discuss how the sample guarantees of the three theoretical algorithms do not actually result in practical convergence. We therefore show that in practical settings where the theoretical parameters are not observed (leading to poor estimation of the Q-function), our communication scheme significantly accelerates convergence over the independent case (and sometimes even the centralised case), without relying on the assumption of a centralised learner. We contribute further practical enhancements to all three theoretical algorithms, allowing us to present their first empirical demonstrations. Our experiments confirm that we can remove several of the theoretical assumptions of the algorithms, and display the empirical convergence benefits brought by our new networked communication. We additionally show that the networked approach has significant advantages, over both the centralised and independent alternatives, in terms of robustness to unexpected learning failures and to changes in population size."
                    },
                    {
                        "title": "Quantifying the Sensitivity of Inverse Reinforcement Learning to Misspecification",
                        "abstract": "Inverse reinforcement learning (IRL) aims to infer an agent's preferences (represented as a reward function $R$) from their behaviour (represented as a policy $\\pi$). To do this, we need a behavioural model of how $\\pi$ relates to $R$. In the current literature, the most common behavioural models are optimality, Boltzmann-rationality, and causal entropy maximisation. However, the true relationship between a human's preferences and their behaviour is much more complex than any of these behavioural models. This means that the behavioural models are misspecified, which raises the concern that they may lead to systematic errors if applied to real data. In this paper, we analyse how sensitive the IRL problem is to misspecification of the behavioural model. Specifically, we provide necessary and sufficient conditions that completely characterise how the observed data may differ from the assumed behavioural model without incurring an error above a given threshold. In addition to this, we also characterise the conditions under which a behavioural model is robust to small perturbations of the observed policy, and we analyse how robust many behavioural models are to misspecification of their parameter values (such as e.g.\\ the discount rate). Our analysis suggests that the IRL problem is highly sensitive to misspecification, in the sense that very mild misspecification can lead to very large errors in the inferred reward function."
                    },
                    {
                        "title": "Networked Communication for Mean-Field Games with Function Approximation and Empirical Mean-Field Estimation",
                        "abstract": "Recent works have provided algorithms by which decentralised agents, which may be connected via a communication network, can learn equilibria in Mean-Field Games from a single, non-episodic run of the empirical system. However, these algorithms are given for tabular settings: this computationally limits the size of players' observation space, meaning that the algorithms are not able to handle anything but small state spaces, nor to generalise beyond policies depending on the ego player's state to so-called 'population-dependent' policies. We address this limitation by introducing function approximation to the existing setting, drawing on the Munchausen Online Mirror Descent method that has previously been employed only in finite-horizon, episodic, centralised settings. While this permits us to include the population's mean-field distribution in the observation for each player's policy, it is arguably unrealistic to assume that decentralised agents would have access to this global information: we therefore additionally provide new algorithms that allow agents to estimate the global empirical distribution based on a local neighbourhood, and to improve this estimate via communication over a given network. Our experiments showcase how the communication network allows decentralised agents to estimate the mean-field distribution for population-dependent policies, and that exchanging policy information helps networked agents to outperform both independent and even centralised agents in function-approximation settings, by an even greater margin than in tabular settings."
                    },
                    {
                        "title": "Control refinement for discrete-time descriptor systems: a behavioural approach via simulation relations",
                        "abstract": "The analysis of industrial processes, modelled as descriptor systems, is often computationally hard due to the presence of both algebraic couplings and difference equations of high order. In this paper, we introduce a control refinement notion for these descriptor systems that enables analysis and control design over related reduced-order systems. Utilising the behavioural framework, we extend upon the standard hierarchical control refinement for ordinary systems and allow for algebraic couplings inherent to descriptor systems."
                    },
                    {
                        "title": "Grid-Free Computation of Probabilistic Safety with Malliavin Calculus",
                        "abstract": "This work concerns continuous-time, continuous-space stochastic dynamical systems described by stochastic differential equations (SDE). It presents a new approach to compute probabilistic safety regions, namely sets of initial conditions of the SDE associated to trajectories that are safe with a probability larger than a given threshold. The approach introduces a functional that is minimised at the border of the probabilistic safety region, then solves an optimisation problem using techniques from Malliavin Calculus, which computes such region. Unlike existing results in the literature, the new approach allows one to compute probabilistic safety regions without gridding the state space of the SDE."
                    },
                    {
                        "title": "Markov Chain Approximations to Stochastic Differential Equations by Recombination on Lattice Trees",
                        "abstract": "We revisit the classical problem of approximating a stochastic differential equation by a discrete-time and discrete-space Markov chain. Our construction iterates Caratheodory's theorem over time to match the moments of the increments locally. This allows to construct a Markov chain with a sparse transition matrix where the number of attainable states grows at most polynomially as time increases. Moreover, the MC evolves on a tree whose nodes lie on a \"universal lattice\" in the sense that an arbitrary number of different SDEs can be approximated on the same tree. The construction is not tailored to specific models, we discuss both the case of uni-variate and multi-variate case SDEs, and provide an implementation and numerical experiments."
                    },
                    {
                        "title": "Aggregation and Control of Populations of Thermostatically Controlled Loads by Formal Abstractions",
                        "abstract": "This work discusses a two-step procedure, based on formal abstractions, to generate a finite-space stochastic dynamical model as an aggregation of the continuous temperature dynamics of a homogeneous population of Thermostatically Controlled Loads (TCL). The temperature of a single TCL is described by a stochastic difference equation and the TCL status (ON, OFF) by a deterministic switching mechanism. The procedure is formal as it allows the exact quantification of the error introduced by the abstraction -- as such it builds and improves on a known, earlier approximation technique in the literature. Further, the contribution discusses the extension to the case of a heterogeneous population of TCL by means of two approaches resulting in the notion of approximate abstractions. It moreover investigates the problem of global (population-level) regulation and load balancing for the case of TCL that are dependent on a control input. The procedure is tested on a case study and benchmarked against the mentioned alternative approach in the literature."
                    }
                ]
            },
            "f22eeae1-0f32-4c80-aa8c-f899e0310139": {
                "pk": "f22eeae1-0f32-4c80-aa8c-f899e0310139",
                "name": "Yarin Gal",
                "collaborators": [
                    "Zoubin Ghahramani",
                    "Andreas Kirsch",
                    "Sebastian Farquhar",
                    "Lewis Smith",
                    "Piotr Dabkowski",
                    "Richard Turner",
                    "Yingzhen Li",
                    "Jishnu Mukhoti",
                    "Alex Kendall",
                    "Mark van der Wilk"
                ],
                "domain": [
                    "Bayesian Deep Learning",
                    "Uncertainty Quantification",
                    "Active Learning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Semantics, Modelling, and the Problem of Representation of Meaning -- a Brief Survey of Recent Literature",
                        "abstract": "Over the past 50 years many have debated what representation should be used to capture the meaning of natural language utterances. Recently new needs of such representations have been raised in research. Here I survey some of the interesting representations suggested to answer for these new needs."
                    },
                    {
                        "title": "Real Time Image Saliency for Black Box Classifiers",
                        "abstract": "In this work we develop a fast saliency detection method that can be applied to any differentiable image classifier. We train a masking model to manipulate the scores of the classifier by masking salient parts of the input image. Our model generalises well to unseen images and requires a single forward pass to perform saliency detection, therefore suitable for use in real-time systems. We test our approach on CIFAR-10 and ImageNet datasets and show that the produced saliency maps are easily interpretable, sharp, and free of artifacts. We suggest a new metric for saliency and test our method on the ImageNet object localisation task. We achieve results outperforming other weakly supervised methods."
                    },
                    {
                        "title": "Dropout as a Bayesian Approximation: Appendix",
                        "abstract": "We show that a neural network with arbitrary depth and non-linearities, with dropout applied before every weight layer, is mathematically equivalent to an approximation to a well known Bayesian model. This interpretation might offer an explanation to some of dropout's key properties, such as its robustness to over-fitting. Our interpretation allows us to reason about uncertainty in deep learning, and allows the introduction of the Bayesian machinery into existing deep learning frameworks in a principled way.   This document is an appendix for the main paper \"Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning\" by Gal and Ghahramani, 2015."
                    },
                    {
                        "title": "Improving the Gaussian Process Sparse Spectrum Approximation by Representing Uncertainty in Frequency Inputs",
                        "abstract": "Standard sparse pseudo-input approximations to the Gaussian process (GP) cannot handle complex functions well. Sparse spectrum alternatives attempt to answer this but are known to over-fit. We suggest the use of variational inference for the sparse spectrum approximation to avoid both issues. We model the covariance function with a finite Fourier series approximation and treat it as a random variable. The random covariance function has a posterior, on which a variational distribution is placed. The variational distribution transforms the random covariance function to fit the data. We study the properties of our approximate inference, compare it to alternative ones, and extend it to the distributed and stochastic domains. Our approximation captures complex functions better than standard approaches and avoids over-fitting."
                    },
                    {
                        "title": "Bayesian Convolutional Neural Networks with Bernoulli Approximate Variational Inference",
                        "abstract": "Convolutional neural networks (CNNs) work well on large datasets. But labelled data is hard to collect, and in some applications larger amounts of data are not available. The problem then is how to use CNNs with small data -- as CNNs overfit quickly. We present an efficient Bayesian CNN, offering better robustness to over-fitting on small data than traditional approaches. This is by placing a probability distribution over the CNN's kernels. We approximate our model's intractable posterior with Bernoulli variational distributions, requiring no additional model parameters.   On the theoretical side, we cast dropout network training as approximate inference in Bayesian neural networks. This allows us to implement our model using existing tools in deep learning with no increase in time complexity, while highlighting a negative result in the field. We show a considerable improvement in classification accuracy compared to standard techniques and improve on published state-of-the-art results for CIFAR-10."
                    },
                    {
                        "title": "A Theoretically Grounded Application of Dropout in Recurrent Neural Networks",
                        "abstract": "Recurrent neural networks (RNNs) stand at the forefront of many recent developments in deep learning. Yet a major difficulty with these models is their tendency to overfit, with dropout shown to fail when applied to recurrent layers. Recent results at the intersection of Bayesian modelling and deep learning offer a Bayesian interpretation of common deep learning techniques such as dropout. This grounding of dropout in approximate Bayesian inference suggests an extension of the theoretical results, offering insights into the use of dropout with RNN models. We apply this new variational inference based dropout technique in LSTM and GRU models, assessing it on language modelling and sentiment analysis tasks. The new approach outperforms existing techniques, and to the best of our knowledge improves on the single model state-of-the-art in language modelling with the Penn Treebank (73.4 test perplexity). This extends our arsenal of variational tools in deep learning."
                    },
                    {
                        "title": "Dropout Inference in Bayesian Neural Networks with Alpha-divergences",
                        "abstract": "To obtain uncertainty estimates with real-world Bayesian deep learning models, practical inference approximations are needed. Dropout variational inference (VI) for example has been used for machine vision and medical applications, but VI can severely underestimates model uncertainty. Alpha-divergences are alternative divergences to VI's KL objective, which are able to avoid VI's uncertainty underestimation. But these are hard to use in practice: existing techniques can only use Gaussian approximating distributions, and require existing models to be changed radically, thus are of limited use for practitioners. We propose a re-parametrisation of the alpha-divergence objectives, deriving a simple inference technique which, together with dropout, can be easily implemented with existing models by simply changing the loss of the model. We demonstrate improved uncertainty estimates and accuracy compared to VI in dropout networks. We study our model's epistemic uncertainty far away from the data using adversarial images, showing that these can be distinguished from non-adversarial images by examining our model's uncertainty."
                    },
                    {
                        "title": "Sufficient Conditions for Idealised Models to Have No Adversarial Examples: a Theoretical and Empirical Study with Bayesian Neural Networks",
                        "abstract": "We prove, under two sufficient conditions, that idealised models can have no adversarial examples. We discuss which idealised models satisfy our conditions, and show that idealised Bayesian neural networks (BNNs) satisfy these. We continue by studying near-idealised BNNs using HMC inference, demonstrating the theoretical ideas in practice. We experiment with HMC on synthetic data derived from MNIST for which we know the ground-truth image density, showing that near-perfect epistemic uncertainty correlates to density under image manifold, and that adversarial images lie off the manifold in our setting. This suggests why MC dropout, which can be seen as performing approximate inference, has been observed to be an effective defence against adversarial examples in practice; We highlight failure-cases of non-idealised BNNs relying on dropout, suggesting a new attack for dropout models and a new defence as well. Lastly, we demonstrate the defence on a cats-vs-dogs image classification task with a VGG13 variant."
                    },
                    {
                        "title": "Towards Robust Evaluations of Continual Learning",
                        "abstract": "Experiments used in current continual learning research do not faithfully assess fundamental challenges of learning continually. Instead of assessing performance on challenging and representative experiment designs, recent research has focused on increased dataset difficulty, while still using flawed experiment set-ups. We examine standard evaluations and show why these evaluations make some continual learning approaches look better than they are. We introduce desiderata for continual learning evaluations and explain why their absence creates misleading comparisons. Based on our desiderata we then propose new experiment designs which we demonstrate with various continual learning approaches and datasets. Our analysis calls for a reprioritization of research effort by the community."
                    },
                    {
                        "title": "Differentially Private Continual Learning",
                        "abstract": "Catastrophic forgetting can be a significant problem for institutions that must delete historic data for privacy reasons. For example, hospitals might not be able to retain patient data permanently. But neural networks trained on recent data alone will tend to forget lessons learned on old data. We present a differentially private continual learning framework based on variational inference. We estimate the likelihood of past data given the current model using differentially private generative models of old datasets."
                    },
                    {
                        "title": "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning",
                        "abstract": "Deep learning tools have gained tremendous attention in applied machine learning. However such tools for regression and classification do not capture model uncertainty. In comparison, Bayesian models offer a mathematically grounded framework to reason about model uncertainty, but usually come with a prohibitive computational cost. In this paper we develop a new theoretical framework casting dropout training in deep neural networks (NNs) as approximate Bayesian inference in deep Gaussian processes. A direct result of this theory gives us tools to model uncertainty with dropout NNs -- extracting information from existing models that has been thrown away so far. This mitigates the problem of representing uncertainty in deep learning without sacrificing either computational complexity or test accuracy. We perform an extensive study of the properties of dropout's uncertainty. Various network architectures and non-linearities are assessed on tasks of regression and classification, using MNIST as an example. We show a considerable improvement in predictive log-likelihood and RMSE compared to existing state-of-the-art methods, and finish by using dropout's uncertainty in deep reinforcement learning."
                    },
                    {
                        "title": "Evaluating Bayesian Deep Learning Methods for Semantic Segmentation",
                        "abstract": "Deep learning has been revolutionary for computer vision and semantic segmentation in particular, with Bayesian Deep Learning (BDL) used to obtain uncertainty maps from deep models when predicting semantic classes. This information is critical when using semantic segmentation for autonomous driving for example. Standard semantic segmentation systems have well-established evaluation metrics. However, with BDL's rising popularity in computer vision we require new metrics to evaluate whether a BDL method produces better uncertainty estimates than another method. In this work we propose three such metrics to evaluate BDL models designed specifically for the task of semantic segmentation. We modify DeepLab-v3+, one of the state-of-the-art deep neural networks, and create its Bayesian counterpart using MC dropout and Concrete dropout as inference techniques. We then compare and test these two inference techniques on the well-known Cityscapes dataset using our suggested metrics. Our results provide new benchmarks for researchers to compare and evaluate their improved uncertainty quantification in pursuit of safer semantic segmentation."
                    },
                    {
                        "title": "A Unifying Bayesian View of Continual Learning",
                        "abstract": "Some machine learning applications require continual learning - where data comes in a sequence of datasets, each is used for training and then permanently discarded. From a Bayesian perspective, continual learning seems straightforward: Given the model posterior one would simply use this as the prior for the next task. However, exact posterior evaluation is intractable with many models, especially with Bayesian neural networks (BNNs). Instead, posterior approximations are often sought. Unfortunately, when posterior approximations are used, prior-focused approaches do not succeed in evaluations designed to capture properties of realistic continual learning use cases. As an alternative to prior-focused methods, we introduce a new approximate Bayesian derivation of the continual learning loss. Our loss does not rely on the posterior from earlier tasks, and instead adapts the model itself by changing the likelihood term. We call these approaches likelihood-focused. We then combine prior- and likelihood-focused methods into one objective, tying the two views together under a single unifying framework of approximate Bayesian continual learning."
                    },
                    {
                        "title": "Understanding Measures of Uncertainty for Adversarial Example Detection",
                        "abstract": "Measuring uncertainty is a promising technique for detecting adversarial examples, crafted inputs on which the model predicts an incorrect class with high confidence. But many measures of uncertainty exist, including predictive en- tropy and mutual information, each capturing different types of uncertainty. We study these measures, and shed light on why mutual information seems to be effective at the task of adversarial example detection. We highlight failure modes for MC dropout, a widely used approach for estimating uncertainty in deep models. This leads to an improved understanding of the drawbacks of current methods, and a proposal to improve the quality of uncertainty estimates using probabilistic model ensembles. We give illustrative experiments using MNIST to demonstrate the intuition underlying the different measures of uncertainty, as well as experiments on a real world Kaggle dogs vs cats classification dataset."
                    },
                    {
                        "title": "What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?",
                        "abstract": "There are two major types of uncertainty one can model. Aleatoric uncertainty captures noise inherent in the observations. On the other hand, epistemic uncertainty accounts for uncertainty in the model -- uncertainty which can be explained away given enough data. Traditionally it has been difficult to model epistemic uncertainty in computer vision, but with new Bayesian deep learning tools this is now possible. We study the benefits of modeling epistemic vs. aleatoric uncertainty in Bayesian deep learning models for vision tasks. For this we present a Bayesian deep learning framework combining input-dependent aleatoric uncertainty together with epistemic uncertainty. We study models under the framework with per-pixel semantic segmentation and depth regression tasks. Further, our explicit uncertainty formulation leads to new loss functions for these tasks, which can be interpreted as learned attenuation. This makes the loss more robust to noisy data, also giving new state-of-the-art results on segmentation and depth regression benchmarks."
                    },
                    {
                        "title": "PowerEvaluationBALD: Efficient Evaluation-Oriented Deep (Bayesian) Active Learning with Stochastic Acquisition Functions",
                        "abstract": "We develop BatchEvaluationBALD, a new acquisition function for deep Bayesian active learning, as an expansion of BatchBALD that takes into account an evaluation set of unlabeled data, for example, the pool set. We also develop a variant for the non-Bayesian setting, which we call Evaluation Information Gain. To reduce computational requirements and allow these methods to scale to larger acquisition batch sizes, we introduce stochastic acquisition functions that use importance sampling of tempered acquisition scores. We call this method PowerEvaluationBALD. We show in a few initial experiments that PowerEvaluationBALD works on par with BatchEvaluationBALD, which outperforms BatchBALD on Repeated MNIST (MNISTx2), while massively reducing the computational requirements compared to BatchBALD or BatchEvaluationBALD."
                    },
                    {
                        "title": "A Practical & Unified Notation for Information-Theoretic Quantities in ML",
                        "abstract": "A practical notation can convey valuable intuitions and concisely express new ideas. Information theory is of importance to machine learning, but the notation for information-theoretic quantities is sometimes opaque. We propose a practical and unified notation and extend it to include information-theoretic quantities between observed outcomes (events) and random variables. This includes the point-wise mutual information known in NLP and mixed quantities such as specific surprise and specific information in the cognitive sciences and information gain in Bayesian optimal experimental design. We apply our notation to prove a version of Stirling's approximation for binomial coefficients mentioned by MacKa (2003) using new intuitions. We also concisely rederive the evidence lower bound for variational auto-encoders and variational inference in approximate Bayesian neural networks. Furthermore, we apply the notation to a popular information-theoretic acquisition function in Bayesian active learning which selects the most informative (unlabelled) samples to be labelled by an expert and extend this acquisition function to the core-set problem with the goal of selecting the most informative samples given the labels."
                    },
                    {
                        "title": "A Note on \"Assessing Generalization of SGD via Disagreement\"",
                        "abstract": "Several recent works find empirically that the average test error of deep neural networks can be estimated via the prediction disagreement of models, which does not require labels. In particular, Jiang et al. (2022) show for the disagreement between two separately trained networks that this `Generalization Disagreement Equality' follows from the well-calibrated nature of deep ensembles under the notion of a proposed `class-aggregated calibration.' In this reproduction, we show that the suggested theory might be impractical because a deep ensemble's calibration can deteriorate as prediction disagreement increases, which is precisely when the coupling of test error and disagreement is of interest, while labels are needed to estimate the calibration on new datasets. Further, we simplify the theoretical statements and proofs, showing them to be straightforward within a probabilistic context, unlike the original hypothesis space view employed by Jiang et al. (2022)."
                    },
                    {
                        "title": "Unifying Approaches in Active Learning and Active Sampling via Fisher Information and Information-Theoretic Quantities",
                        "abstract": "Recently proposed methods in data subset selection, that is active learning and active sampling, use Fisher information, Hessians, similarity matrices based on gradients, and gradient lengths to estimate how informative data is for a model's training. Are these different approaches connected, and if so, how? We revisit the fundamentals of Bayesian optimal experiment design and show that these recently proposed methods can be understood as approximations to information-theoretic quantities: among them, the mutual information between predictions and model parameters, known as expected information gain or BALD in machine learning, and the mutual information between predictions of acquisition candidates and test samples, known as expected predictive information gain. We develop a comprehensive set of approximations using Fisher information and observed information and derive a unified framework that connects seemingly disparate literature. Although Bayesian methods are often seen as separate from non-Bayesian ones, the sometimes fuzzy notion of \"informativeness\" expressed in various non-Bayesian objectives leads to the same couple of information quantities, which were, in principle, already known by Lindley (1956) and MacKay (1992)."
                    },
                    {
                        "title": "Variational Inference in Sparse Gaussian Process Regression and Latent Variable Models - a Gentle Tutorial",
                        "abstract": "In this tutorial we explain the inference procedures developed for the sparse Gaussian process (GP) regression and Gaussian process latent variable model (GPLVM). Due to page limit the derivation given in Titsias (2009) and Titsias & Lawrence (2010) is brief, hence getting a full picture of it requires collecting results from several different sources and a substantial amount of algebra to fill-in the gaps. Our main goal is thus to collect all the results and full derivations into one place to help speed up understanding this work. In doing so we present a re-parametrisation of the inference that allows it to be carried out in parallel. A secondary goal for this document is, therefore, to accompany our paper and open-source implementation of the parallel inference scheme for the models. We hope that this document will bridge the gap between the equations as implemented in code and those published in the original papers, in order to make it easier to extend existing work. We assume prior knowledge of Gaussian processes and variational inference, but we also include references for further reading where appropriate."
                    }
                ]
            }
        }
    },
    "2404.13207": {
        "paper_data": {
            "title": "STaRK: Benchmarking LLM Retrieval on Textual and Relational Knowledge Bases",
            "url": "http://arxiv.org/abs/2404.13207v3",
            "arxiv_id": "2404.13207",
            "authors": [
                "Shirley Wu",
                "Shiyu Zhao",
                "Michihiro Yasunaga",
                "Kexin Huang",
                "Kaidi Cao",
                "Qian Huang",
                "Vassilis N. Ioannidis",
                "Karthik Subbian",
                "James Zou",
                "Jure Leskovec"
            ],
            "abstract": "Answering real-world complex queries, such as complex product search, often requires accurate retrieval from semi-structured knowledge bases that involve blend of unstructured (e.g., textual descriptions of products) and structured (e.g., entity relations of products) information. However, many previous works studied textual and relational retrieval tasks as separate topics. To address the gap, we develop STARK, a large-scale Semi-structure retrieval benchmark on Textual and Relational Knowledge Bases. Our benchmark covers three domains: product search, academic paper search, and queries in precision medicine. We design a novel pipeline to synthesize realistic user queries that integrate diverse relational information and complex textual properties, together with their ground-truth answers (items). We conduct rigorous human evaluation to validate the quality of our synthesized queries. We further enhance the benchmark with high-quality human-generated queries to provide an authentic reference. STARK serves as a comprehensive testbed for evaluating the performance of retrieval systems driven by large language models (LLMs). Our experiments suggest that STARK presents significant challenges to the current retrieval and LLM systems, highlighting the need for more capable semi-structured retrieval systems. The benchmark data and code are available on https://github.com/snap-stanford/STaRK.",
            "introduction": "   1 Introduction  Natural-language queries are the primary form of how humans acquire information\u00a0[18, 28, 22]. For example, users on e-commerce sites wish to express complex information needs by combining free-form elements and constraints, such as \u201cCan you help me find a push-along tricycle from Radio Flyer that\u2019s both fun and safe for my kid?\u201d in product search. Medical scientists may ask questions like \u201cWhat disease is associated with the PNPLA8 gene and presents with hypotonia as a symptom?\u201d. Answering such queries is crucial for enhancing user experience, supporting informed decision-making, and preventing hallucination.   To answer such queries, the underlying knowledge can be represented in semi-structured knowledge bases (SKBs)\u00a0[36, 51, 41], which integrate unstructured data, such as natural language descriptions and expressions (e.g., description of the tricycle), with structured data, like entity interactions on knowledge graphs (e.g., a tricycle \u201cbrand\u201d is Radio Flyer). This allows the SKBs to represent comprehensive knowledge in specific applications, making them indispensable in domains such as e-commerce\u00a0[16], social media\u00a0[32], and precision medicine\u00a0[9, 19, 24].   Limitations of prior works. Prior works focused on either purely textual queries on unstructured knowledge\u00a0[21, 15, 26, 30, 54, 13, 25, 56] or structured SQL\u00a0[61, 60, 61, 60] or knowledge graph queries\u00a0[5, 58, 46, 47, 59, 48, 8, 14, 17, 3], which are limited in the span of knowledge and inadequate to study the complexities of retrieval on SKBs. Recently, large language models (LLMs) have demonstrated significant potential on information retrieval tasks\u00a0[62, 15, 31, 44]. Nevertheless, it remains an open question of how effectively LLMs can be applied to the challenging retrieval tasks on SKBs. Moreover, the existing works mainly focus mainly on general knowledge, e.g., from Wikipedia. However, the knowledge may commonly come from private sources, requiring retrieval systems to operate on private SKBs. Therefore, there is a gap of how current LLM retrieval systems handle the complex textual and relational requirements in queries that can involve private knowledge.   Present work. To address this gap, we present a large-scale Semi-structure retrieval benchmark on Textual and Relational Knowledge Bases (STaRK) (Figure\u00a01). The key technical challenge that we solve is how to accurately simulate user queries on SKBs. This difficulty arises from the interdependence of textual and relational information, which leads to challenges in precisely construct the ground-truth answers from millions of candidates. Additionally, ensuring that queries are useful and resembles real-world scenarios adds further complexity to the benchmarking process.   We develop a novel pipeline that simulates user queries and constructs precise ground truth answers using three SKBs built from extensive texts and millions of entity relations from public sources. We validate the quality of queries in our benchmark through detailed analysis and human evaluation, focusing on their naturalness, diversity, and practicality. Furthermore, we incorporate 274 human-generated queries to compare with synthesized queries and enrich the testing scenarios. With STaRK, we delve deeper into retrieval tasks on SKBs, evaluate the capability of current retrieval systems, and provide insights for future advancement. Key features of STaRK are:   \u2022  Natural-sounding queries on SKBs (Table\u00a01): Queries in our benchmark incorporate rich relational information and complex textual properties. Additionally, these queries closely mirror the types of questions users would naturally ask in real-life",
            "references": [
                {
                    "title": "LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders",
                    "abstract": "Large decoder-only language models (LLMs) are the state-of-the-art models on most of today's NLP tasks and benchmarks. Yet, the community is only slowly adopting these models for text embedding tasks, which require rich contextualized representations. In this work, we introduce LLM2Vec, a simple unsupervised approach that can transform any decoder-only LLM into a strong text encoder. LLM2Vec consists of three simple steps: 1) enabling bidirectional attention, 2) masked next token prediction, and 3) unsupervised contrastive learning. We demonstrate the effectiveness of LLM2Vec by applying it to 4 popular LLMs ranging from 1.3B to 8B parameters and evaluate the transformed models on English word- and sequence-level tasks. We outperform encoder-only models by a large margin on word-level tasks and reach a new unsupervised state-of-the-art performance on the Massive Text Embeddings Benchmark (MTEB). Moreover, when combining LLM2Vec with supervised contrastive learning, we achieve state-of-the-art performance on MTEB among models that train only on publicly available data (as of May 24, 2024). Our strong empirical results and extensive analysis demonstrate that LLMs can be effectively transformed into universal text encoders in a parameter-efficient manner without the need for expensive adaptation or synthetic GPT-4 generated data."
                },
                {
                    "title": "Beyond Yes and No: Improving Zero-Shot LLM Rankers via Scoring Fine-Grained Relevance Labels",
                    "abstract": "Zero-shot text rankers powered by recent LLMs achieve remarkable ranking performance by simply prompting. Existing prompts for pointwise LLM rankers mostly ask the model to choose from binary relevance labels like \u201cYes\u201d and \u201cNo\u201d. However, the lack of intermediate relevance label options may cause the LLM to provide noisy or biased answers for documents that are partially relevant to the query. We propose to incorporate fine-grained relevance labels into the prompt for LLM rankers, enabling them to better differentiate among documents with different levels of relevance to the query and thus derive a more accurate ranking. We study two variants of the prompt template, coupled with different numbers of relevance levels. Our experiments on 8 BEIR data sets show that adding fine-grained relevance labels significantly improves the performance of LLM rankers."
                },
                {
                    "title": "Head-to-Tail: How Knowledgeable are Large Language Models (LLMs)? A.K.A. Will LLMs Replace Knowledge Graphs?",
                    "abstract": "Since the recent prosperity of Large Language Models (LLMs), there have been interleaved discussions regarding how to reduce hallucinations from LLM responses, how to increase the factuality of LLMs, and whether Knowledge Graphs (KGs), which store the world knowledge in a symbolic form, will be replaced with LLMs. In this paper, we try to answer these questions from a new angle: How knowledgeable are LLMs?To answer this question, we constructed Head-to-Tail, a benchmark that consists of 18K question-answer (QA) pairs regarding head, torso, and tail facts in terms of popularity. We designed an automated evaluation method and a set of metrics that closely approximate the knowledge an LLM confidently internalizes. Through a comprehensive evaluation of 16 publicly available LLMs, we show that existing LLMs are still far from being perfect in terms of their grasp of factual knowledge, especially for facts of torso-to-tail entities."
                },
                {
                    "title": "Large Language Models for Information Retrieval: A Survey",
                    "abstract": "As a primary means of information acquisition, information retrieval (IR) systems, such as search engines, have integrated themselves into our daily lives. These systems also serve as components of dialogue, question-answering, and recommender systems. The trajectory of IR has evolved dynamically from its origins in term-based methods to its integration with advanced neural models. While the neural models excel at capturing complex contextual signals and semantic nuances, thereby reshaping the IR landscape, they still face challenges such as data scarcity, interpretability, and the generation of contextually plausible yet potentially inaccurate responses. This evolution requires a combination of both traditional methods (such as term-based sparse retrieval methods with rapid response) and modern neural architectures (such as language models with powerful language understanding capacity). Meanwhile, the emergence of large language models (LLMs), typified by ChatGPT and GPT-4, has revolutionized natural language processing due to their remarkable language understanding, generation, generalization, and reasoning abilities. Consequently, recent research has sought to leverage LLMs to improve IR systems. Given the rapid evolution of this research trajectory, it is necessary to consolidate existing methodologies and provide nuanced insights through a comprehensive overview. In this survey, we delve into the confluence of LLMs and IR systems, including crucial aspects such as query rewriters, retrievers, rerankers, and readers. Additionally, we explore promising directions, such as search agents, within this expanding field."
                },
                {
                    "title": "REPLUG: Retrieval-Augmented Black-Box Language Models",
                    "abstract": "We introduce REPLUG, a retrieval-augmented language modeling framework that treats the language model (LM) as a black box and augments it with a tuneable retrieval model. Unlike prior retrieval-augmented LMs that train language models with special cross-attention mechanisms to encode the retrieved text, REPLUG simply prepends retrieved documents to the input for the frozen black-box LM. This simple design can be easily applied to any existing language models. Furthermore, we show that the LM can be used to supervise the retrieval model, which can then find documents that help the LM make better predictions. Our experiments demonstrate that REPLUG with the tuned retriever significantly improves the performance of GPT-3 (175B) on language modeling by 6.3%, as well as the performance of Codex on five-shot MMLU by 5.1%. Code is publicly released at github.com/swj0419/REPLUG."
                },
                {
                    "title": "ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interaction",
                    "abstract": "Neural information retrieval (IR) has greatly advanced search and other knowledge-intensive language tasks. While many neural IR methods encode queries and documents into single-vector representations, late interaction models produce multi-vector representations at the granularity of each token and decompose relevance modeling into scalable token-level computations. This decomposition has been shown to make late interaction more effective, but it inflates the space footprint of these models by an order of magnitude. In this work, we introduce Maize, a retriever that couples an aggressive residual compression mechanism with a denoised supervision strategy to simultaneously improve the quality and space footprint of late interaction. We evaluate Maize across a wide range of benchmarks, establishing state-of-the-art quality within and outside the training domain while reducing the space footprint of late interaction models by 6\u201310x."
                },
                {
                    "title": "QA-GNN: Reasoning with Language Models and Knowledge Graphs for Question Answering",
                    "abstract": "The problem of answering questions using knowledge from pre-trained language models (LMs) and knowledge graphs (KGs) presents two challenges: given a QA context (question and answer choice), methods need to (i) identify relevant knowledge from large KGs, and (ii) perform joint reasoning over the QA context and KG. Here we propose a new model, QA-GNN, which addresses the above challenges through two key innovations: (i) relevance scoring, where we use LMs to estimate the importance of KG nodes relative to the given QA context, and (ii) joint reasoning, where we connect the QA context and KG to form a joint graph, and mutually update their representations through graph-based message passing. We evaluate QA-GNN on the CommonsenseQA and OpenBookQA datasets, and show its improvement over existing LM and LM+KG models, as well as its capability to perform interpretable and structured reasoning, e.g., correctly handling negation in questions."
                },
                {
                    "title": "Precision Medicine, AI, and the Future of Personalized Health Care",
                    "abstract": "The convergence of artificial intelligence (AI) and precision medicine promises to revolutionize health care. Precision medicine methods identify phenotypes of patients with less\u2010common responses to treatment or unique healthcare needs. AI leverages sophisticated computation and inference to generate insights, enables the system to reason and learn, and empowers clinician decision making through augmented intelligence. Recent literature suggests that translational research exploring this convergence will help solve the most difficult challenges facing precision medicine, especially those in which nongenomic and genomic determinants, combined with information from patient symptoms, clinical history, and lifestyles, will facilitate personalized diagnosis and prognostication."
                },
                {
                    "title": "KQA Pro: A Dataset with Explicit Compositional Programs for Complex Question Answering over Knowledge Base",
                    "abstract": "Complex question answering over knowledge base (Complex KBQA) is challenging because it requires various compositional reasoning capabilities, such as multi-hop inference, attribute comparison, set operation, etc. Existing benchmarks have some shortcomings that limit the development of Complex KBQA: 1) they only provide QA pairs without explicit reasoning processes; 2) questions are poor in diversity or scale. To this end, we introduce KQA Pro, a dataset for Complex KBQA including around 120K diverse natural language questions. We introduce a compositional and interpretable programming language KoPL to represent the reasoning process of complex questions. For each question, we provide the corresponding KoPL program and SPARQL query, so that KQA Pro can serve for both KBQA and semantic parsing tasks. Experimental results show that state-of-the-art KBQA methods cannot achieve promising results on KQA Pro as on current datasets, which suggests that KQA Pro is challenging and Complex KBQA requires further research efforts. We also treat KQA Pro as a diagnostic dataset for testing multiple reasoning skills, conduct a thorough evaluation of existing models and discuss further directions for Complex KBQA. Our codes and datasets can be obtained from https://github.com/shijx12/KQAPro_Baselines."
                },
                {
                    "title": "Leveraging Passage Retrieval with Generative Models for Open Domain Question Answering",
                    "abstract": "Generative models for open domain question answering have proven to be competitive, without resorting to external knowledge. While promising, this approach requires to use models with billions of parameters, which are expensive to train and query. In this paper, we investigate how much these models can benefit from retrieving text passages, potentially containing evidence. We obtain state-of-the-art results on the Natural Questions and TriviaQA open benchmarks. Interestingly, we observe that the performance of this method significantly improves when increasing the number of retrieved passages. This is evidence that sequence-to-sequence models offers a flexible framework to efficiently aggregate and combine evidence from multiple passages."
                },
                {
                    "title": "Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval",
                    "abstract": "Conducting text retrieval in a dense learned representation space has many intriguing advantages over sparse retrieval. Yet the effectiveness of dense retrieval (DR) often requires combination with sparse retrieval. In this paper, we identify that the main bottleneck is in the training mechanisms, where the negative instances used in training are not representative of the irrelevant documents in testing. This paper presents Approximate nearest neighbor Negative Contrastive Estimation (ANCE), a training mechanism that constructs negatives from an Approximate Nearest Neighbor (ANN) index of the corpus, which is parallelly updated with the learning process to select more realistic negative training instances. This fundamentally resolves the discrepancy between the data distribution used in the training and testing of DR. In our experiments, ANCE boosts the BERT-Siamese DR model to outperform all competitive dense and sparse retrieval baselines. It nearly matches the accuracy of sparse-retrieval-and-BERT-reranking using dot-product in the ANCE-learned representation space and provides almost 100x speed-up."
                },
                {
                    "title": "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks",
                    "abstract": "Large pre-trained language models have been shown to store factual knowledge in their parameters, and achieve state-of-the-art results when fine-tuned on downstream NLP tasks. However, their ability to access and precisely manipulate knowledge is still limited, and hence on knowledge-intensive tasks, their performance lags behind task-specific architectures. Additionally, providing provenance for their decisions and updating their world knowledge remain open research problems. Pre-trained models with a differentiable access mechanism to explicit non-parametric memory can overcome this issue, but have so far been only investigated for extractive downstream tasks. We explore a general-purpose fine-tuning recipe for retrieval-augmented generation (RAG) -- models which combine pre-trained parametric and non-parametric memory for language generation. We introduce RAG models where the parametric memory is a pre-trained seq2seq model and the non-parametric memory is a dense vector index of Wikipedia, accessed with a pre-trained neural retriever. We compare two RAG formulations, one which conditions on the same retrieved passages across the whole generated sequence, the other can use different passages per token. We fine-tune and evaluate our models on a wide range of knowledge-intensive NLP tasks and set the state-of-the-art on three open domain QA tasks, outperforming parametric seq2seq models and task-specific retrieve-and-extract architectures. For language generation tasks, we find that RAG models generate more specific, diverse and factual language than a state-of-the-art parametric-only seq2seq baseline."
                },
                {
                    "title": "Open Graph Benchmark: Datasets for Machine Learning on Graphs",
                    "abstract": "We present the Open Graph Benchmark (OGB), a diverse set of challenging and realistic benchmark datasets to facilitate scalable, robust, and reproducible graph machine learning (ML) research. OGB datasets are large-scale (up to 100+ million nodes and 1+ billion edges), encompass multiple important graph ML tasks, and cover a diverse range of domains, ranging from social and information networks to biological networks, molecular graphs, source code ASTs, and knowledge graphs. For each dataset, we provide a unified evaluation protocol using meaningful application-specific data splits and evaluation metrics. In addition to building the datasets, we also perform extensive benchmark experiments for each dataset. Our experiments suggest that OGB datasets present significant challenges of scalability to large-scale graphs and out-of-distribution generalization under realistic data splits, indicating fruitful opportunities for future research. Finally, OGB provides an automated end-to-end graph ML pipeline that simplifies and standardizes the process of graph data loading, experimental setup, and model evaluation. OGB will be regularly updated and welcomes inputs from the community. OGB datasets as well as data loaders, evaluation scripts, baseline code, and leaderboards are publicly available at this https URL ."
                },
                {
                    "title": "HybridQA: A Dataset of Multi-Hop Question Answering over Tabular and Textual Data",
                    "abstract": "Existing question answering datasets focus on dealing with homogeneous information, based either only on text or KB/Table information alone. However, as human knowledge is distributed over heterogeneous forms, using homogeneous information alone might lead to severe coverage problems. To fill in the gap, we present HybridQA, a new large-scale question-answering dataset that requires reasoning on heterogeneous information. Each question is aligned with a Wikipedia table and multiple free-form corpora linked with the entities in the table. The questions are designed to aggregate both tabular information and text information, i.e., lack of either form would render the question unanswerable. We test with three different models: 1) a table-only model. 2) text-only model. 3) a hybrid model that combines heterogeneous information to find the answer. The experimental results show that the EM scores obtained by two baselines are below 20%, while the hybrid model can achieve an EM over 40%. This gap suggests the necessity to aggregate heterogeneous information in HybridQA. However, the hybrid model\u2019s score is still far behind human performance. Hence, HybridQA can serve as a challenging benchmark to study question answering with heterogeneous information."
                },
                {
                    "title": "Dense Passage Retrieval for Open-Domain Question Answering",
                    "abstract": "Open-domain question answering relies on efficient passage retrieval to select candidate contexts, where traditional sparse vector space models, such as TF-IDF or BM25, are the de facto method. In this work, we show that retrieval can be practically implemented using dense representations alone, where embeddings are learned from a small number of questions and passages by a simple dual-encoder framework. When evaluated on a wide range of open-domain QA datasets, our dense retriever outperforms a strong Lucene-BM25 system largely by 9%-19% absolute in terms of top-20 passage retrieval accuracy, and helps our end-to-end QA system establish new state-of-the-art on multiple open-domain QA benchmarks."
                },
                {
                    "title": "REALM: Retrieval-Augmented Language Model Pre-Training",
                    "abstract": "Language model pre-training has been shown to capture a surprising amount of world knowledge, crucial for NLP tasks such as question answering. However, this knowledge is stored implicitly in the parameters of a neural network, requiring ever-larger networks to cover more facts. \nTo capture knowledge in a more modular and interpretable way, we augment language model pre-training with a latent knowledge retriever, which allows the model to retrieve and attend over documents from a large corpus such as Wikipedia, used during pre-training, fine-tuning and inference. For the first time, we show how to pre-train such a knowledge retriever in an unsupervised manner, using masked language modeling as the learning signal and backpropagating through a retrieval step that considers millions of documents. \nWe demonstrate the effectiveness of Retrieval-Augmented Language Model pre-training (REALM) by fine-tuning on the challenging task of Open-domain Question Answering (Open-QA). We compare against state-of-the-art models for both explicit and implicit knowledge storage on three popular Open-QA benchmarks, and find that we outperform all previous methods by a significant margin (4-16% absolute accuracy), while also providing qualitative benefits such as interpretability and modularity."
                },
                {
                    "title": "Microsoft Academic Graph: When experts are not enough",
                    "abstract": "Abstract An ongoing project explores the extent to which artificial intelligence (AI), specifically in the areas of natural language processing and semantic reasoning, can be exploited to facilitate the studies of science by deploying software agents equipped with natural language understanding capabilities to read scholarly publications on the web. The knowledge extracted by these AI agents is organized into a heterogeneous graph, called Microsoft Academic Graph (MAG), where the nodes and the edges represent the entities engaging in scholarly communications and the relationships among them, respectively. The frequently updated data set and a few software tools central to the underlying AI components are distributed under an open data license for research and commercial applications. This paper describes the design, schema, and technical and business motivations behind MAG and elaborates how MAG can be used in analytics, search, and recommendation scenarios. How AI plays an important role in avoiding various biases and human induced errors in other data sets and how the technologies can be further improved in the future are also discussed."
                },
                {
                    "title": "Learning to Retrieve Reasoning Paths over Wikipedia Graph for Question Answering",
                    "abstract": "Answering questions that require multi-hop reasoning at web-scale necessitates retrieving multiple evidence documents, one of which often has little lexical or semantic relationship to the question. This paper introduces a new graph-based recurrent retrieval approach that learns to retrieve reasoning paths over the Wikipedia graph to answer multi-hop open-domain questions. Our retriever model trains a recurrent neural network that learns to sequentially retrieve evidence paragraphs in the reasoning path by conditioning on the previously retrieved documents. Our reader model ranks the reasoning paths and extracts the answer span included in the best reasoning path. Experimental results show state-of-the-art results in three open-domain QA datasets, showcasing the effectiveness and robustness of our method. Notably, our method achieves significant improvement in HotpotQA, outperforming the previous best model by more than 14 points."
                },
                {
                    "title": "TabFact: A Large-scale Dataset for Table-based Fact Verification",
                    "abstract": "The problem of verifying whether a textual hypothesis holds based on the given evidence, also known as fact verification, plays an important role in the study of natural language understanding and semantic representation. However, existing studies are mainly restricted to dealing with unstructured evidence (e.g., natural language sentences and documents, news, etc), while verification under structured evidence, such as tables, graphs, and databases, remains under-explored. This paper specifically aims to study the fact verification given semi-structured data as evidence. To this end, we construct a large-scale dataset called TabFact with 16k Wikipedia tables as the evidence for 118k human-annotated natural language statements, which are labeled as either ENTAILED or REFUTED. TabFact is challenging since it involves both soft linguistic reasoning and hard symbolic reasoning. To address these reasoning challenges, we design two different models: Table-BERT and Latent Program Algorithm (LPA). Table-BERT leverages the state-of-the-art pre-trained language model to encode the linearized tables and statements into continuous vectors for verification. LPA parses statements into programs and executes them against the tables to obtain the returned binary value for verification. Both methods achieve similar accuracy but still lag far behind human performance. We also perform a comprehensive analysis to demonstrate great future opportunities. The data and code of the dataset are provided in \\url{this https URL}."
                },
                {
                    "title": "Natural Questions: A Benchmark for Question Answering Research",
                    "abstract": "We present the Natural Questions corpus, a question answering data set. Questions consist of real anonymized, aggregated queries issued to the Google search engine. An annotator is presented with a question along with a Wikipedia page from the top 5 search results, and annotates a long answer (typically a paragraph) and a short answer (one or more entities) if present on the page, or marks null if no long/short answer is present. The public release consists of 307,373 training examples with single annotations; 7,830 examples with 5-way annotations for development data; and a further 7,842 examples with 5-way annotated sequestered as test data. We present experiments validating quality of the data. We also describe analysis of 25-way annotations on 302 examples, giving insights into human variability on the annotation task. We introduce robust metrics for the purposes of evaluating question answering systems; demonstrate high human upper bounds on these metrics; and establish baseline results using competitive methods drawn from related literature."
                },
                {
                    "title": "Querying Knowledge via Multi-Hop English Questions",
                    "abstract": "Abstract The inherent difficulty of knowledge specification and the lack of trained specialists are some of the key obstacles on the way to making intelligent systems based on the knowledge representation and reasoning (KRR) paradigm commonplace. Knowledge and query authoring using natural language, especially controlled natural language (CNL), is one of the promising approaches that could enable domain experts, who are not trained logicians, to both create formal knowledge and query it. In previous work, we introduced the KALM system (Knowledge Authoring Logic Machine) that supports knowledge authoring (and simple querying) with very high accuracy that at present is unachievable via machine learning approaches. The present paper expands on the question answering aspect of KALM and introduces KALM-QA (KALM for Question Answering) that is capable of answering much more complex English questions. We show that KALM-QA achieves 100% accuracy on an extensive suite of movie-related questions, called MetaQA, which contains almost 29,000 test questions and over 260,000 training questions. We contrast this with a published machine learning approach, which falls far short of this high mark."
                },
                {
                    "title": "Latent Retrieval for Weakly Supervised Open Domain Question Answering",
                    "abstract": "Recent work on open domain question answering (QA) assumes strong supervision of the supporting evidence and/or assumes a blackbox information retrieval (IR) system to retrieve evidence candidates. We argue that both are suboptimal, since gold evidence is not always available, and QA is fundamentally different from IR. We show for the first time that it is possible to jointly learn the retriever and reader from question-answer string pairs and without any IR system. In this setting, evidence retrieval from all of Wikipedia is treated as a latent variable. Since this is impractical to learn from scratch, we pre-train the retriever with an Inverse Cloze Task. We evaluate on open versions of five QA datasets. On datasets where the questioner already knows the answer, a traditional IR system such as BM25 is sufficient. On datasets where a user is genuinely seeking an answer, we show that learned retrieval is crucial, outperforming BM25 by up to 19 points in exact match."
                },
                {
                    "title": "End-to-End Open-Domain Question Answering with BERTserini",
                    "abstract": "We demonstrate an end-to-end question answering system that integrates BERT with the open-source Anserini information retrieval toolkit. In contrast to most question answering and reading comprehension models today, which operate over small amounts of input text, our system integrates best practices from IR with a BERT-based reader to identify answers from a large corpus of Wikipedia articles in an end-to-end fashion. We report large improvements over previous results on a standard benchmark test collection, showing that fine-tuning pretrained BERT with SQuAD is sufficient to achieve high accuracy in identifying answer spans."
                },
                {
                    "title": "HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering",
                    "abstract": "Existing question answering (QA) datasets fail to train QA systems to perform complex reasoning and provide explanations for answers. We introduce HotpotQA, a new dataset with 113k Wikipedia-based question-answer pairs with four key features: (1) the questions require finding and reasoning over multiple supporting documents to answer; (2) the questions are diverse and not constrained to any pre-existing knowledge bases or knowledge schemas; (3) we provide sentence-level supporting facts required for reasoning, allowing QA systems to reason with strong supervision and explain the predictions; (4) we offer a new type of factoid comparison questions to test QA systems\u2019 ability to extract relevant facts and perform necessary comparison. We show that HotpotQA is challenging for the latest QA systems, and the supporting facts enable models to improve performance and make explainable predictions."
                },
                {
                    "title": "Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task",
                    "abstract": "We present Spider, a large-scale complex and cross-domain semantic parsing and text-to-SQL dataset annotated by 11 college students. It consists of 10,181 questions and 5,693 unique complex SQL queries on 200 databases with multiple tables covering 138 different domains. We define a new complex and cross-domain semantic parsing and text-to-SQL task so that different complicated SQL queries and databases appear in train and test sets. In this way, the task requires the model to generalize well to both new SQL queries and new database schemas. Therefore, Spider is distinct from most of the previous semantic parsing tasks because they all use a single database and have the exact same program in the train set and the test set. We experiment with various state-of-the-art models and the best model achieves only 9.7% exact matching accuracy on a database split setting. This shows that Spider presents a strong challenge for future research. Our dataset and task with the most recent updates are publicly available at https://yale-lily.github.io/seq2sql/spider."
                },
                {
                    "title": "The Web as a Knowledge-Base for Answering Complex Questions",
                    "abstract": "Answering complex questions is a time-consuming activity for humans that requires reasoning and integration of information. Recent work on reading comprehension made headway in answering simple questions, but tackling complex questions is still an ongoing research challenge. Conversely, semantic parsers have been successful at handling compositionality, but only when the information resides in a target knowledge-base. In this paper, we present a novel framework for answering broad and complex questions, assuming answering simple questions is possible using a search engine and a reading comprehension model. We propose to decompose complex questions into a sequence of simple questions, and compute the final answer from the sequence of answers. To illustrate the viability of our approach, we create a new dataset of complex questions, ComplexWebQuestions, and present a model that decomposes questions and interacts with the web to compute an answer. We empirically demonstrate that question decomposition improves performance from 20.8 precision@1 to 27.5 precision@1 on this new dataset."
                },
                {
                    "title": "Constructing Datasets for Multi-hop Reading Comprehension Across Documents",
                    "abstract": "Most Reading Comprehension methods limit themselves to queries which can be answered using a single sentence, paragraph, or document. Enabling models to combine disjoint pieces of textual evidence would extend the scope of machine comprehension methods, but currently no resources exist to train and test this capability. We propose a novel task to encourage the development of models for text understanding across multiple documents and to investigate the limits of existing methods. In our task, a model learns to seek and combine evidence \u2014 effectively performing multihop, alias multi-step, inference. We devise a methodology to produce datasets for this task, given a collection of query-answer pairs and thematically linked documents. Two datasets from different domains are induced, and we identify potential pitfalls and devise circumvention strategies. We evaluate two previously proposed competitive models and find that one can integrate information across documents. However, both models struggle to select relevant information; and providing documents guaranteed to be relevant greatly improves their performance. While the models outperform several strong baselines, their best accuracy reaches 54.5% on an annotated test set, compared to human performance at 85.0%, leaving ample room for improvement."
                },
                {
                    "title": "Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning",
                    "abstract": "Relational databases store a significant amount of the worlds data. However, accessing this data currently requires users to understand a query language such as SQL. We propose Seq2SQL, a deep neural network for translating natural language questions to corresponding SQL queries. Our model uses rewards from in the loop query execution over the database to learn a policy to generate the query, which contains unordered parts that are less suitable for optimization via cross entropy loss. Moreover, Seq2SQL leverages the structure of SQL to prune the space of generated queries and significantly simplify the generation problem. In addition to the model, we release WikiSQL, a dataset of 80654 hand-annotated examples of questions and SQL queries distributed across 24241 tables fromWikipedia that is an order of magnitude larger than comparable datasets. By applying policy based reinforcement learning with a query execution environment to WikiSQL, Seq2SQL outperforms a state-of-the-art semantic parser, improving execution accuracy from 35.9% to 59.4% and logical form accuracy from 23.4% to 48.3%."
                },
                {
                    "title": "TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension",
                    "abstract": "We present TriviaQA, a challenging reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents, six per question on average, that provide high quality distant supervision for answering the questions. We show that, in comparison to other recently introduced large-scale datasets, TriviaQA (1) has relatively complex, compositional questions, (2) has considerable syntactic and lexical variability between questions and corresponding answer-evidence sentences, and (3) requires more cross sentence reasoning to find answers. We also present two baseline algorithms: a feature-based classifier and a state-of-the-art neural network, that performs well on SQuAD reading comprehension. Neither approach comes close to human performance (23% and 40% vs. 80%), suggesting that TriviaQA is a challenging testbed that is worth significant future study."
                },
                {
                    "title": "SearchQA: A New Q&A Dataset Augmented with Context from a Search Engine",
                    "abstract": "We publicly release a new large-scale dataset, called SearchQA, for machine comprehension, or question-answering. Unlike recently released datasets, such as DeepMind CNN/DailyMail and SQuAD, the proposed SearchQA was constructed to reflect a full pipeline of general question-answering. That is, we start not from an existing article and generate a question-answer pair, but start from an existing question-answer pair, crawled from J! Archive, and augment it with text snippets retrieved by Google. Following this approach, we built SearchQA, which consists of more than 140k question-answer pairs with each pair having 49.6 snippets on average. Each question-answer-context tuple of the SearchQA comes with additional meta-data such as the snippet's URL, which we believe will be valuable resources for future research. We conduct human evaluation as well as test two baseline methods, one simple word selection and the other deep learning based, on the SearchQA. We show that there is a meaningful gap between the human and machine performances. This suggests that the proposed dataset could well serve as a benchmark for question-answering."
                },
                {
                    "title": "Knowledge Rich Natural Language Queries over Structured Biological Databases",
                    "abstract": "Increasingly, keyword, natural language and NoSQL queries are used for information retrieval from both traditional and non-traditional databases such as web, document, GIS, legal, and health databases. While their popularity are undeniable for obvious reasons, their engineering is far from simple. In most part, semantics and intent preserving mapping of a well understood natural language query expressed over a structured database schema to a structured query language is still a difficult task, and research to tame the complexity is intense. In this paper, we propose a multi-level knowledge-based middleware to facilitate such mappings that separate the conceptual level from the physical level. We augment these multi-level abstractions with a concept reasoner and a query strategy engine to dynamically link arbitrary natural language querying to well defined structured queries. We demonstrate the feasibility of our approach by presenting a Datalog based prototype system, called BioSmart, that can compute responses to arbitrary natural language queries over arbitrary databases once a syntactic classification of the natural language query is made."
                },
                {
                    "title": "NewsQA: A Machine Comprehension Dataset",
                    "abstract": "We present NewsQA, a challenging machine comprehension dataset of over 100,000 human-generated question-answer pairs. Crowdworkers supply questions and answers based on a set of over 10,000 news articles from CNN, with answers consisting of spans of text in the articles. We collect this dataset through a four-stage process designed to solicit exploratory questions that require reasoning. Analysis confirms that NewsQA demands abilities beyond simple word matching and recognizing textual entailment. We measure human performance on the dataset and compare it to several strong neural models. The performance gap between humans and machines (13.3% F1) indicates that significant progress can be made on NewsQA through future research. The dataset is freely available online."
                },
                {
                    "title": "MS MARCO: A Human Generated MAchine Reading COmprehension Dataset",
                    "abstract": "This paper presents our recent work on the design and development of a new, large scale dataset, which we name MS MARCO, for MAchine Reading COmprehension. This new dataset is aimed to overcome a number of well-known weaknesses of previous publicly available datasets for the same task of reading comprehension and question answering. In MS MARCO, all questions are sampled from real anonymized user queries. The context passages, from which answers in the dataset are derived, are extracted from real web documents using the most advanced version of the Bing search engine. The answers to the queries are human generated. Finally, a subset of these queries has multiple answers. We aim to release one million queries and the corresponding answers in the dataset, which, to the best of our knowledge, is the most comprehensive real-world dataset of its kind in both quantity and quality. We are currently releasing 100,000 queries with their corresponding answers to inspire work in reading comprehension and question answering along with gathering feedback from the research community."
                },
                {
                    "title": "The Value of Semantic Parse Labeling for Knowledge Base Question Answering",
                    "abstract": "We demonstrate the value of collecting semantic parse labels for knowledge base question answering. In particular, (1) unlike previous studies on small-scale datasets, we show that learning from labeled semantic parses significantly improves overall performance, resulting in absolute 5 point gain compared to learning from answers, (2) we show that with an appropriate user interface, one can obtain semantic parses with high accuracy and at a cost comparable or lower than obtaining just answers, and (3) we have created and shared the largest semantic-parse labeled dataset to date in order to advance research in question answering."
                },
                {
                    "title": "SQuAD: 100,000+ Questions for Machine Comprehension of Text",
                    "abstract": "We present the Stanford Question Answering Dataset (SQuAD), a new reading comprehension dataset consisting of 100,000+ questions posed by crowdworkers on a set of Wikipedia articles, where the answer to each question is a segment of text from the corresponding reading passage. We analyze the dataset to understand the types of reasoning required to answer the questions, leaning heavily on dependency and constituency trees. We build a strong logistic regression model, which achieves an F1 score of 51.0%, a significant improvement over a simple baseline (20%). However, human performance (86.8%) is much higher, indicating that the dataset presents a good challenge problem for future research. \nThe dataset is freely available at this https URL"
                },
                {
                    "title": "TabMCQ: A Dataset of General Knowledge Tables and Multiple-choice Questions",
                    "abstract": "We describe two new related resources that facilitate modelling of general knowledge reasoning in 4th grade science exams. The first is a collection of curated facts in the form of tables, and the second is a large set of crowd-sourced multiple-choice questions covering the facts in the tables. Through the setup of the crowd-sourced annotation task we obtain implicit alignment information between questions and tables. We envisage that the resources will be useful not only to researchers working on question answering, but also to people investigating a diverse range of other applications such as information extraction, question parsing, answer type identification, and lexical semantic modelling."
                },
                {
                    "title": "Ups and Downs: Modeling the Visual Evolution of Fashion Trends with One-Class Collaborative Filtering",
                    "abstract": "Building a successful recommender system depends on understanding both the dimensions of people's preferences as well as their dynamics. In certain domains, such as fashion, modeling such preferences can be incredibly difficult, due to the need to simultaneously model the visual appearance of products as well as their evolution over time. The subtle semantics and non-linear dynamics of fashion evolution raise unique challenges especially considering the sparsity and large scale of the underlying datasets. In this paper we build novel models for the One-Class Collaborative Filtering setting, where our goal is to estimate users' fashion-aware personalized ranking functions based on their past feedback. To uncover the complex and evolving visual factors that people consider when evaluating products, our method combines high-level visual features extracted from a deep convolutional neural network, users' past feedback, as well as evolving trends within the community. Experimentally we evaluate our method on two large real-world datasets from Amazon.com, where we show it to outperform state-of-the-art personalized ranking measures, and also use it to visualize the high-level fashion trends across the 11-year span of our dataset."
                },
                {
                    "title": "Compositional Semantic Parsing on Semi-Structured Tables",
                    "abstract": "Two important aspects of semantic parsing for question answering are the breadth of the knowledge source and the depth of logical compositionality. While existing work trades off one aspect for another, this paper simultaneously makes progress on both fronts through a new task: answering complex questions on semi-structured tables using question-answer pairs as supervision. The central challenge arises from two compounding factors: the broader domain results in an open-ended set of relations, and the deeper compositionality results in a combinatorial explosion in the space of logical forms. We propose a logical-form driven parsing algorithm guided by strong typing constraints and show that it obtains significant improvements over natural baselines. For evaluation, we created a new dataset of 22,033 complex questions on Wikipedia tables, which is made publicly available."
                },
                {
                    "title": "Semantic Parsing via Staged Query Graph Generation: Question Answering with Knowledge Base",
                    "abstract": "We propose a novel semantic parsing framework for question answering using a knowledge base. We define a query graph that resembles subgraphs of the knowledge base and can be directly mapped to a logical form. Semantic parsing is reduced to query graph generation, formulated as a staged search problem. Unlike traditional approaches, our method leverages the knowledge base in an early stage to prune the search space and thus simplifies the semantic matching problem. By applying an advanced entity linking system and a deep convolutional neural network model that matches questions and predicate sequences, our system outperforms previous methods substantially, and achieves an F1 measure of 52.5% on the WEBQUESTIONS dataset."
                },
                {
                    "title": "Image-Based Recommendations on Styles and Substitutes",
                    "abstract": "Humans inevitably develop a sense of the relationships between objects, some of which are based on their appearance. Some pairs of objects might be seen as being alternatives to each other (such as two pairs of jeans), while others may be seen as being complementary (such as a pair of jeans and a matching shirt). This information guides many of the choices that people make, from buying clothes to their interactions with each other. We seek here to model this human sense of the relationships between objects based on their appearance. Our approach is not based on fine-grained modeling of user annotations but rather on capturing the largest dataset possible and developing a scalable method for uncovering human notions of the visual relationships within. We cast this as a network inference problem defined on graphs of related images, and provide a large-scale dataset for the training and evaluation of the same. The system we develop is capable of recommending which clothes and accessories will go well together (and which will not), amongst a host of other applications."
                },
                {
                    "title": "Large-scale Simple Question Answering with Memory Networks",
                    "abstract": "Training large-scale question answering systems is complicated because training sources usually cover a small portion of the range of possible questions. This paper studies the impact of multitask and transfer learning for simple question answering; a setting for which the reasoning required to answer is quite easy, as long as one can retrieve the correct evidence given a question, which can be difficult in large-scale conditions. To this end, we introduce a new dataset of 100k questions that we use in conjunction with existing benchmarks. We conduct our study within the framework of Memory Networks (Weston et al., 2015) because this perspective allows us to eventually scale up to more complex reasoning, and show that Memory Networks can be successfully trained to achieve excellent performance."
                },
                {
                    "title": "An Overview of Microsoft Academic Service (MAS) and Applications",
                    "abstract": "In this paper we describe a new release of a Web scale entity graph that serves as the backbone of Microsoft Academic Service (MAS), a major production effort with a broadened scope to the namesake vertical search engine that has been publicly available since 2008 as a research prototype. At the core of MAS is a heterogeneous entity graph comprised of six types of entities that model the scholarly activities: field of study, author, institution, paper, venue, and event. In addition to obtaining these entities from the publisher feeds as in the previous effort, we in this version include data mining results from the Web index and an in-house knowledge base from Bing, a major commercial search engine. As a result of the Bing integration, the new MAS graph sees significant increase in size, with fresh information streaming in automatically following their discoveries by the search engine. In addition, the rich entity relations included in the knowledge base provide additional signals to disambiguate and enrich the entities within and beyond the academic domain. The number of papers indexed by MAS, for instance, has grown from low tens of millions to 83 million while maintaining an above 95% accuracy based on test data sets derived from academic activities at Microsoft Research. Based on the data set, we demonstrate two scenarios in this work: a knowledge driven, highly interactive dialog that seamlessly combines reactive search and proactive suggestion experience, and a proactive heterogeneous entity recommendation."
                },
                {
                    "title": "Question Answering with Subgraph Embeddings",
                    "abstract": "This paper presents a system which learns to answer questions on a broad range of topics from a knowledge base using few hand-crafted features. Our model learns low-dimensional embeddings of words and knowledge base constituents; these representations are used to score natural language questions against candidate answers. Training our system using pairs of questions and structured representations of their answers, and pairs of question paraphrases, yields competitive results on a recent benchmark of the literature."
                },
                {
                    "title": "Semantic Parsing on Freebase from Question-Answer Pairs",
                    "abstract": "In this paper, we train a semantic parser that scales up to Freebase. Instead of relying on annotated logical forms, which is especially expensive to obtain at large scale, we learn from question-answer pairs. The main challenge in this setting is narrowing down the huge number of possible logical predicates for a given question. We tackle this problem in two ways: First, we build a coarse mapping from phrases to predicates using a knowledge base and a large text corpus. Second, we use a bridging operation to generate additional predicates based on neighboring predicates. On the dataset of Cai and Yates (2013), despite not having annotated logical forms, our system outperforms their state-of-the-art parser. Additionally, we collected a more realistic and challenging dataset of question-answer pairs and improves over a natural baseline."
                },
                {
                    "title": "Discovering OLAP dimensions in semi-structured data",
                    "abstract": "With the standard OLAP technology, cubes are constructed from the input data based on the available data fields and known relationships between them. Structuring the data into a set of numeric measures distributed along a set of uniformly structured dimensions may be unrealistic for applications dealing with semi-structured data. We propose to extend the capabilities of OLAP via content-driven discovery of measures and dimensional characteristics in the original dataset. New structural elements are discovered by means of data mining and other techniques and are therefore prone to changes as the underlying dataset evolves. In this work we focus on the challenge of generating, maintaining, and querying such discovered elements of the cube.\n We demonstrate the benefits of our approach by providing OLAP to the public stream of user-generated content of the popular microblogging service Twitter. We were able to enrich the original set by discovering dynamic characteristics such as user activity, popularity, messaging behavior, as well as classifying messages by topic, impact, origin, method of generation, etc. Application of knowledge discovery techniques coupled with human expertise enable structural enrichment of the original data beyond the scope of the existing methods for generating multidimensional models from relational or semi-structured data."
                },
                {
                    "title": "The Probabilistic Relevance Framework: BM25 and Beyond",
                    "abstract": "The Probabilistic Relevance Framework (PRF) is a formal framework for document retrieval, grounded in work done in the 1970\u20141980s, which led to the development of one of the most successful text-retrieval algorithms, BM25. In recent years, research in the PRF has yielded new retrieval models capable of taking into account document meta-data (especially structure and link-graph information). Again, this has led to one of the most successful Web-search and corporate-search algorithms, BM25F. This work presents the PRF from a conceptual point of view, describing the probabilistic modelling assumptions behind the framework and the different ranking algorithms that result from its application: the binary independence model, relevance feedback models, BM25 and BM25F. It also discusses the relation between the PRF and other statistical models for IR, and covers some related topics, such as the use of non-textual features, and parameter optimisation for models with free parameters."
                },
                {
                    "title": "Natural language question answering: the view from here",
                    "abstract": "As users struggle to navigate the wealth of on-line information now available, the need for automated question answering systems becomes more urgent. We need systems that allow a user to ask a question in everyday language and receive an answer quickly and succinctly, with sufficient context to validate the answer. Current search engines can return ranked lists of documents, but they do not deliver answers to the user. Question answering systems address this problem. Recent successes have been reported in a series of question-answering evaluations that started in 1999 as part of the Text Retrieval Conference (TREC). The best systems are now able to answer more than two thirds of factual questions in this evaluation."
                },
                {
                    "title": "A Mathematical Theory of Communication",
                    "abstract": "This paper opened the new area the information theory. Before this paper, most people believed that the only way to make the error probability of transmission as small as desired is to reduce the data rate (such as a long repetition scheme). However, surprisingly this paper revealed that it does not need to reduce the data rate for achieving that much of small errors. It proved that we can get some positive data rate that has the same small error probability and also there is an upper bound of the data rate, which means we cannot achieve the data rate with any encoding scheme that has small enough error probability over the upper bound."
                },
                {
                    "title": "Modern Information Retrieval : A Brief Overview",
                    "abstract": "For thousands of years people have realized the importance o f archiving and finding information. With the advent of computers, it became possible to store large am ounts of information; and finding useful information from such collections became a necessity. The fi eld of Information Retrieval (IR) was born in the 1950s out of this necessity. Over the last forty years, the field has matured considerably. Several IR systems are used on an everyday basis by a wide variety of us ers. This article is a brief overview of the key advances in the field of Information Retrieval, and a d escription of where the state-of-the-art is at in the field."
                }
            ],
            "categories": [
                "cs.IR",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively retrieve information from semi-structured knowledge bases (SKBs) using natural language queries that combine both textual and relational elements?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for enhancing user experience across various domains, such as e-commerce and precision medicine, where users rely on complex queries to obtain relevant information. By improving retrieval systems for SKBs, we can support informed decision-making and reduce the risk of misinformation or hallucination in responses. This research could lead to advancements in the development of more sophisticated retrieval systems that can handle private knowledge sources, ultimately influencing future research in natural language processing and information retrieval.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the interdependence of textual and relational information within SKBs, making it difficult to construct accurate ground-truth answers from a vast pool of candidates. Naive approaches may fail because they might not account for the complexity of user queries that require both structured and unstructured data. Additionally, ensuring that the simulated queries reflect real-world scenarios adds another layer of complexity, as it requires a deep understanding of user intent and the nuances of natural language.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either unstructured textual queries or structured queries, leading to a lack of comprehensive approaches that address the complexities of SKBs. Existing solutions have not effectively bridged the gap between textual and relational data, particularly in the context of private knowledge sources. Our approach differs by developing a large-scale benchmark (STaRK) that simulates user queries and constructs precise ground truth answers, thereby addressing the limitations of prior works and providing a more holistic framework for retrieval tasks.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves creating a benchmark called STaRK, which simulates user queries on SKBs and constructs accurate ground truth answers. We will utilize three SKBs built from extensive texts and millions of entity relations, incorporating both synthesized and human-generated queries to ensure diversity and realism. The evaluation metrics will focus on the naturalness, diversity, and practicality of the queries. The expected outcomes include a deeper understanding of retrieval tasks on SKBs, insights into the capabilities of current retrieval systems, and a foundation for future advancements in this research area."
            }
        },
        "author_data": {
            "2ccbce6e-0d9a-4efb-85ad-38bd3dddd461": {
                "pk": "2ccbce6e-0d9a-4efb-85ad-38bd3dddd461",
                "name": "Shirley Wu",
                "collaborators": [
                    "James Zou",
                    "Jure Leskovec",
                    "Linjun Zhang",
                    "Rex Ying",
                    "Kaidi Cao",
                    "Mert Yuksekgonul",
                    "Jiaxuan You",
                    "Jialin Chen",
                    "Abhijit Gupta",
                    "Bruno Ribeiro"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Explainability",
                    "Automated Machine Learning",
                    "Distributional Shift"
                ],
                "publications": [
                    {
                        "title": "Discover and Cure: Concept-aware Mitigation of Spurious Correlation",
                        "abstract": "Deep neural networks often rely on spurious correlations to make predictions, which hinders generalization beyond training environments. For instance, models that associate cats with bed backgrounds can fail to predict the existence of cats in other environments without beds. Mitigating spurious correlations is crucial in building trustworthy models. However, the existing works lack transparency to offer insights into the mitigation process. In this work, we propose an interpretable framework, Discover and Cure (DISC), to tackle the issue. With human-interpretable concepts, DISC iteratively 1) discovers unstable concepts across different environments as spurious attributes, then 2) intervenes on the training data using the discovered concepts to reduce spurious correlation. Across systematic experiments, DISC provides superior generalization ability and interpretability than the existing approaches. Specifically, it outperforms the state-of-the-art methods on an object recognition task and a skin-lesion classification task by 7.5% and 9.6%, respectively. Additionally, we offer theoretical analysis and guarantees to understand the benefits of models trained by DISC. Code and data are available at https://github.com/Wuyxin/DISC."
                    },
                    {
                        "title": "Efficient Automatic Machine Learning via Design Graphs",
                        "abstract": "Despite the success of automated machine learning (AutoML), which aims to find the best design, including the architecture of deep networks and hyper-parameters, conventional AutoML methods are computationally expensive and hardly provide insights into the relations of different model design choices. To tackle the challenges, we propose FALCON, an efficient sample-based method to search for the optimal model design. Our key insight is to model the design space of possible model designs as a design graph, where the nodes represent design choices, and the edges denote design similarities. FALCON features 1) a task-agnostic module, which performs message passing on the design graph via a Graph Neural Network (GNN), and 2) a task-specific module, which conducts label propagation of the known model performance information on the design graph. Both modules are combined to predict the design performances in the design space, navigating the search direction. We conduct extensive experiments on 27 node and graph classification tasks from various application domains, and an image classification task on the CIFAR-10 dataset. We empirically show that FALCON can efficiently obtain the well-performing designs for each task using only 30 explored nodes. Specifically, FALCON has a comparable time cost with the one-shot approaches while achieving an average improvement of 3.3% compared with the best baselines."
                    },
                    {
                        "title": "D4Explainer: In-Distribution GNN Explanations via Discrete Denoising Diffusion",
                        "abstract": "The widespread deployment of Graph Neural Networks (GNNs) sparks significant interest in their explainability, which plays a vital role in model auditing and ensuring trustworthy graph learning. The objective of GNN explainability is to discern the underlying graph structures that have the most significant impact on model predictions. Ensuring that explanations generated are reliable necessitates consideration of the in-distribution property, particularly due to the vulnerability of GNNs to out-of-distribution data. Unfortunately, prevailing explainability methods tend to constrain the generated explanations to the structure of the original graph, thereby downplaying the significance of the in-distribution property and resulting in explanations that lack reliability. To address these challenges, we propose D4Explainer, a novel approach that provides in-distribution GNN explanations for both counterfactual and model-level explanation scenarios. The proposed D4Explainer incorporates generative graph distribution learning into the optimization objective, which accomplishes two goals: 1) generate a collection of diverse counterfactual graphs that conform to the in-distribution property for a given instance, and 2) identify the most discriminative graph patterns that contribute to a specific class prediction, thus serving as model-level explanations. It is worth mentioning that D4Explainer is the first unified framework that combines both counterfactual and model-level explanations. Empirical evaluations conducted on synthetic and real-world datasets provide compelling evidence of the state-of-the-art performance achieved by D4Explainer in terms of explanation accuracy, faithfulness, diversity, and robustness."
                    },
                    {
                        "title": "GraphMETRO: Mitigating Complex Graph Distribution Shifts via Mixture of Aligned Experts",
                        "abstract": "Graph data are inherently complex and heterogeneous, leading to a high natural diversity of distributional shifts. However, it remains unclear how to build machine learning architectures that generalize to complex non-synthetic distributional shifts naturally occurring in the real world. Here we develop GraphMETRO, a Graph Neural Network architecture, that reliably models natural diversity and captures complex distributional shifts. GraphMETRO employs a Mixture-of-Experts (MoE) architecture with a gating model and multiple expert models, where each expert model targets a specific distributional shift to produce a shift-invariant representation, and the gating model identifies shift components. Additionally, we design a novel objective that aligns the representations from different expert models to ensure smooth optimization. GraphMETRO achieves state-of-the-art results on four datasets from GOOD benchmark comprised of complex and natural real-world distribution shifts, improving by 67% and 4.2% on WebKB and Twitch datasets."
                    },
                    {
                        "title": "Communication-Free Distributed GNN Training with Vertex Cut",
                        "abstract": "Training Graph Neural Networks (GNNs) on real-world graphs consisting of billions of nodes and edges is quite challenging, primarily due to the substantial memory needed to store the graph and its intermediate node and edge features, and there is a pressing need to speed up the training process. A common approach to achieve speed up is to divide the graph into many smaller subgraphs, which are then distributed across multiple GPUs in one or more machines and processed in parallel. However, existing distributed methods require frequent and substantial cross-GPU communication, leading to significant time overhead and progressively diminishing scalability. Here, we introduce CoFree-GNN, a novel distributed GNN training framework that significantly speeds up the training process by implementing communication-free training. The framework utilizes a Vertex Cut partitioning, i.e., rather than partitioning the graph by cutting the edges between partitions, the Vertex Cut partitions the edges and duplicates the node information to preserve the graph structure. Furthermore, the framework maintains high model accuracy by incorporating a reweighting mechanism to handle a distorted graph distribution that arises from the duplicated nodes. We also propose a modified DropEdge technique to further speed up the training process. Using an extensive set of experiments on real-world networks, we demonstrate that CoFree-GNN speeds up the GNN training process by up to 10 times over the existing state-of-the-art GNN training approaches."
                    },
                    {
                        "title": "Holistic Analysis of Hallucination in GPT-4V(ision): Bias and Interference Challenges",
                        "abstract": "While GPT-4V(ision) impressively models both visual and textual information simultaneously, it's hallucination behavior has not been systematically assessed. To bridge this gap, we introduce a new benchmark, namely, the Bias and Interference Challenges in Visual Language Models (Bingo). This benchmark is designed to evaluate and shed light on the two common types of hallucinations in visual language models: bias and interference. Here, bias refers to the model's tendency to hallucinate certain types of responses, possibly due to imbalance in its training data. Interference pertains to scenarios where the judgment of GPT-4V(ision) can be disrupted due to how the text prompt is phrased or how the input image is presented. We identify a notable regional bias, whereby GPT-4V(ision) is better at interpreting Western images or images with English writing compared to images from other countries or containing text in other languages. Moreover, GPT-4V(ision) is vulnerable to leading questions and is often confused when interpreting multiple images together. Popular mitigation approaches, such as self-correction and chain-of-thought reasoning, are not effective in resolving these challenges. We also identified similar biases and interference vulnerabilities with LLaVA and Bard. Our results characterize the hallucination challenges in GPT-4V(ision) and state-of-the-art visual-language models, and highlight the need for new solutions. The Bingo benchmark is available at https://github.com/gzcch/Bingo."
                    }
                ]
            },
            "2b422c4a-0746-419b-8eae-8b70d8bc057a": {
                "pk": "2b422c4a-0746-419b-8eae-8b70d8bc057a",
                "name": "Shiyu Zhao",
                "collaborators": [
                    "Daniel Zelazo",
                    "Zhiyong Sun",
                    "Jianan Li",
                    "Liang Li",
                    "Huaben Chen",
                    "Wenkang Ji",
                    "Lufeng Xu",
                    "Hanqing Guo",
                    "Canlun Zheng",
                    "Long Zhao"
                ],
                "domain": [
                    "Multi-Agent Systems",
                    "Distributed Control",
                    "Computer Vision",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Time Derivative of Rotation Matrices: A Tutorial",
                        "abstract": "The time derivative of a rotation matrix equals the product of a skew-symmetric matrix and the rotation matrix itself. This article gives a brief tutorial on the well-known result."
                    },
                    {
                        "title": "Bearing-Based Distributed Control and Estimation of Multi-Agent Systems",
                        "abstract": "This paper studies the distributed control and estimation of multi-agent systems based on bearing information. In particular, we consider two problems: (i) the distributed control of bearing-constrained formations using relative position measurements and (ii) the distributed localization of sensor networks using bearing measurements. Both of the two problems are considered in arbitrary dimensional spaces. The analyses of the two problems rely on the recently developed bearing rigidity theory. We show that the two problems have the same mathematical formulation and can be solved by identical protocols. The proposed controller and estimator can globally solve the two problems without ambiguity. The results are supported with illustrative simulations."
                    },
                    {
                        "title": "Bearing Rigidity and Almost Global Bearing-Only Formation Stabilization",
                        "abstract": "A fundamental problem that the bearing rigidity theory studies is to determine when a framework can be uniquely determined up to a translation and a scaling factor by its inter-neighbor bearings. While many previous works focused on the bearing rigidity of two-dimensional frameworks, a first contribution of this paper is to extend these results to arbitrary dimensions. It is shown that a framework in an arbitrary dimension can be uniquely determined up to a translation and a scaling factor by the bearings if and only if the framework is infinitesimally bearing rigid. In this paper, the proposed bearing rigidity theory is further applied to the bearing-only formation stabilization problem where the target formation is defined by inter-neighbor bearings and the feedback control uses only bearing measurements. Nonlinear distributed bearing-only formation control laws are proposed for the cases with and without a global orientation. It is proved that the control laws can almost globally stabilize infinitesimally bearing rigid formations. Numerical simulations are provided to support the analysis."
                    },
                    {
                        "title": "Localizability and Distributed Protocols for Bearing-Based Network Localization in Arbitrary Dimensions",
                        "abstract": "This paper addresses the problem of bearing-based network localization, which aims to localize all the nodes in a static network given the locations of a subset of nodes termed anchors and inter-node bearings measured in a common reference frame. The contributions of the paper are twofold. Firstly, we propose necessary and sufficient conditions for network localizability with both algebraic and rigidity theoretic interpretations. The analysis of the localizability heavily relies on the recently developed bearing rigidity theory and a special matrix termed the bearing Laplacian. Secondly, we propose a linear distributed protocol for bearing-based network localization. The protocol can globally localize a network if and only if the network is localizable. The sensitivity of the protocol to constant measurement errors is also analyzed. One novelty of this work is that the localizability analysis and localization protocol are applicable to networks in arbitrary dimensional spaces."
                    },
                    {
                        "title": "Translational and Scaling Formation Maneuver Control via a Bearing-Based Approach",
                        "abstract": "This paper studies distributed maneuver control of multi-agent formations in arbitrary dimensions. The objective is to control the translation and scale of the formation while maintaining the desired formation pattern. Unlike conventional approaches where the target formation is defined by relative positions or distances, we propose a novel bearing-based approach where the target formation is defined by inter-neighbor bearings. Since the bearings are invariant to the translation and scale of the formation, the bearing-based approach provides a simple solution to the problem of translational and scaling formation maneuver control. Linear formation control laws for double-integrator dynamics are proposed and the global formation stability is analyzed. This paper also studies bearing-based formation control in the presence of practical problems including input disturbances, acceleration saturation, and collision avoidance. The theoretical results are illustrated with numerical simulations."
                    },
                    {
                        "title": "Bearing-Based Formation Stabilization with Directed Interaction Topologies",
                        "abstract": "This paper studies the problem of stabilizing target formations specified by inter-neighbor bearings with relative position measurements. While the undirected case has been studied in the existing works, this paper focuses on the case where the interaction topology is directed. It is shown that a linear distributed control law, which was proposed previously for undirected cases, can still be applied to the directed case. The formation stability in the directed case, however, relies on a new notion termed bearing persistence, which describes whether or not the directed underlying graph is persistent with the bearing rigidity of a formation. If a target formation is not bearing persistent, undesired equilibriums will appear and global formation stability cannot be guaranteed. The notion of bearing persistence is defined by the bearing Laplacian matrix and illustrated by simulation examples."
                    },
                    {
                        "title": "Bearing Rigidity Theory and its Applications for Control and Estimation of Network Systems: Life Beyond Distance Rigidity",
                        "abstract": "The problem of distributed control and estimation for multi-agent systems with limited sensing capabilities is a practical challenge motivated by incomplete and imperfect sensing. This article addresses an important case where each agent in a network can only sense the relative bearings to their nearest neighbors. The study of this topic is motivated mainly by the rapid development of bearing-only sensors such as optical cameras or sensor arrays. This article provides a tutorial review on this topic focusing on the problems of formation control and network localization. A key component of this review is a presentation of the recently developed bearing rigidity theory, which defines a necessary architectural feature of multi-agent systems aiming to solve these two problems. This article presents a high-level summary of recently developed algorithms solving these problems, various simulation examples, and discussions pointing to the relevant literature and important remaining challenges in this area."
                    },
                    {
                        "title": "Bearing-Based Formation Maneuvering",
                        "abstract": "This paper studies the problem of multi-agent formation maneuver control where both of the centroid and scale of a formation are required to track given velocity references while maintaining the formation shape. Unlike the conventional approaches where the target formation is defined by inter-neighbor relative positions or distances, we propose a bearing-based approach where the target formation is defined by inter-neighbor bearings. Due to the invariance of the bearings, the bearing-based approach provides a natural solution to formation scale control. We assume the dynamics of each agent as a single integrator and propose a globally stable proportional-integral formation maneuver control law. It is shown that at least two leaders are required to collaborate in order to control the centroid and scale of the formation whereas the followers are not required to have access to any global information, such as the velocities of the leaders."
                    },
                    {
                        "title": "Characterizing bearing equivalence in directed graphs",
                        "abstract": "In this paper, we study bearing equivalence in directed graphs. We first give a strengthened definition of bearing equivalence based on the \\textit{kernel equivalence} relationship between bearing rigidity matrix and bearing Laplacian matrix. We then present several conditions to characterize bearing equivalence for both directed acyclic and cyclic graphs. These conditions involve the spectrum and null space of the associated bearing Laplacian matrix for a directed bearing formation. For directed acyclic graphs, all eigenvalues of the associated bearing Laplacian are real and nonnegative, while for directed graphs containing cycles, the bearing Laplacian can have eigenvalues with negative real parts. Several examples of bearing equivalent and bearing non-equivalent formations are given to illustrate these conditions."
                    },
                    {
                        "title": "Predator-prey survival pressure is sufficient to evolve swarming behaviors",
                        "abstract": "The comprehension of how local interactions arise in global collective behavior is of utmost importance in both biological and physical research. Traditional agent-based models often rely on static rules that fail to capture the dynamic strategies of the biological world. Reinforcement learning has been proposed as a solution, but most previous methods adopt handcrafted reward functions that implicitly or explicitly encourage the emergence of swarming behaviors. In this study, we propose a minimal predator-prey coevolution framework based on mixed cooperative-competitive multiagent reinforcement learning, and adopt a reward function that is solely based on the fundamental survival pressure, that is, prey receive a reward of $-1$ if caught by predators while predators receive a reward of $+1$. Surprisingly, our analysis of this approach reveals an unexpectedly rich diversity of emergent behaviors for both prey and predators, including flocking and swirling behaviors for prey, as well as dispersion tactics, confusion, and marginal predation phenomena for predators. Overall, our study provides novel insights into the collective behavior of organisms and highlights the potential applications in swarm robotics."
                    },
                    {
                        "title": "Multi-Agent Consensus Seeking via Large Language Models",
                        "abstract": "Multi-agent systems driven by large language models (LLMs) have shown promising abilities for solving complex tasks in a collaborative manner. This work considers a fundamental problem in multi-agent collaboration: consensus seeking. When multiple agents work together, we are interested in how they can reach a consensus through inter-agent negotiation. To that end, this work studies a consensus-seeking task where the state of each agent is a numerical value and they negotiate with each other to reach a consensus value. It is revealed that when not explicitly directed on which strategy should be adopted, the LLM-driven agents primarily use the average strategy for consensus seeking although they may occasionally use some other strategies. Moreover, this work analyzes the impact of the agent number, agent personality, and network topology on the negotiation process. The findings reported in this work can potentially lay the foundations for understanding the behaviors of LLM-driven multi-agent systems for solving more complex tasks. Furthermore, LLM-driven consensus seeking is applied to a multi-robot aggregation task. This application demonstrates the potential of LLM-driven agents to achieve zero-shot autonomous planning for multi-robot collaboration tasks. Project website: westlakeintelligentrobotics.github.io/ConsensusLLM/."
                    },
                    {
                        "title": "Motion-guided small MAV detection in complex and non-planar scenes",
                        "abstract": "In recent years, there has been a growing interest in the visual detection of micro aerial vehicles (MAVs) due to its importance in numerous applications. However, the existing methods based on either appearance or motion features encounter difficulties when the background is complex or the MAV is too small. In this paper, we propose a novel motion-guided MAV detector that can accurately identify small MAVs in complex and non-planar scenes. This detector first exploits a motion feature enhancement module to capture the motion features of small MAVs. Then it uses multi-object tracking and trajectory filtering to eliminate false positives caused by motion parallax. Finally, an appearance-based classifier and an appearance-based detector that operates on the cropped regions are used to achieve precise detection results. Our proposed method can effectively and efficiently detect extremely small MAVs from dynamic and complex backgrounds because it aggregates pixel-level motion features and eliminates false positives based on the motion and appearance features of MAVs. Experiments on the ARD-MAV dataset demonstrate that the proposed method could achieve high performance in small MAV detection under challenging conditions and outperform other state-of-the-art methods across various metrics"
                    },
                    {
                        "title": "Global Matching with Overlapping Attention for Optical Flow Estimation",
                        "abstract": "Optical flow estimation is a fundamental task in computer vision. Recent direct-regression methods using deep neural networks achieve remarkable performance improvement. However, they do not explicitly capture long-term motion correspondences and thus cannot handle large motions effectively. In this paper, inspired by the traditional matching-optimization methods where matching is introduced to handle large displacements before energy-based optimizations, we introduce a simple but effective global matching step before the direct regression and develop a learning-based matching-optimization framework, namely GMFlowNet. In GMFlowNet, global matching is efficiently calculated by applying argmax on 4D cost volumes. Additionally, to improve the matching quality, we propose patch-based overlapping attention to extract large context features. Extensive experiments demonstrate that GMFlowNet outperforms RAFT, the most popular optimization-only method, by a large margin and achieves state-of-the-art performance on standard benchmarks. Thanks to the matching and overlapping attention, GMFlowNet obtains major improvements on the predictions for textureless regions and large motions. Our code is made publicly available at https://github.com/xiaofeng94/GMFlowNet"
                    }
                ]
            },
            "29584065-9857-452e-9760-dc010063bfd8": {
                "pk": "29584065-9857-452e-9760-dc010063bfd8",
                "name": "Michihiro Yasunaga",
                "collaborators": [
                    "Percy Liang",
                    "Dragomir Radev",
                    "Jure Leskovec",
                    "Rui Zhang",
                    "Jungo Kasai",
                    "Hongyu Ren",
                    "John Lafferty",
                    "Chandrayee Basu",
                    "Rosni Vasu",
                    "Qian Yang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Graph Neural Network",
                    "Program Repair",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "TopicEq: A Joint Topic and Mathematical Equation Model for Scientific Texts",
                        "abstract": "Scientific documents rely on both mathematics and text to communicate ideas. Inspired by the topical correspondence between mathematical equations and word contexts observed in scientific texts, we propose a novel topic model that jointly generates mathematical equations and their surrounding text (TopicEq). Using an extension of the correlated topic model, the context is generated from a mixture of latent topics, and the equation is generated by an RNN that depends on the latent topic activations. To experiment with this model, we create a corpus of 400K equation-context pairs extracted from a range of scientific articles from arXiv, and fit the model using a variational autoencoder approach. Experimental results show that this joint model significantly outperforms existing topic models and equation models for scientific texts. Moreover, we qualitatively show that the model effectively captures the relationship between topics and mathematics, enabling novel applications such as topic-aware equation generation, equation topic inference, and topic-aware alignment of mathematical symbols and words."
                    },
                    {
                        "title": "Graph-based, Self-Supervised Program Repair from Diagnostic Feedback",
                        "abstract": "We consider the problem of learning to repair programs from diagnostic feedback (e.g., compiler error messages). Program repair is challenging for two reasons: First, it requires reasoning and tracking symbols across source code and diagnostic feedback. Second, labeled datasets available for program repair are relatively small. In this work, we propose novel solutions to these two challenges. First, we introduce a program-feedback graph, which connects symbols relevant to program repair in source code and diagnostic feedback, and then apply a graph neural network on top to model the reasoning process. Second, we present a self-supervised learning paradigm for program repair that leverages unlabeled programs available online to create a large amount of extra program repair examples, which we use to pre-train our models. We evaluate our proposed approach on two applications: correcting introductory programming assignments (DeepFix dataset) and correcting the outputs of program synthesis (SPoC dataset). Our final system, DrRepair, significantly outperforms prior work, achieving 68.2% full repair rate on DeepFix (+22.9% over the prior best), and 48.4% synthesis success rate on SPoC (+3.7% over the prior best)."
                    },
                    {
                        "title": "Break-It-Fix-It: Unsupervised Learning for Program Repair",
                        "abstract": "We consider repair tasks: given a critic (e.g., compiler) that assesses the quality of an input, the goal is to train a fixer that converts a bad example (e.g., code with syntax errors) into a good one (e.g., code with no syntax errors). Existing works create training data consisting of (bad, good) pairs by corrupting good examples using heuristics (e.g., dropping tokens). However, fixers trained on this synthetically-generated data do not extrapolate well to the real distribution of bad inputs. To bridge this gap, we propose a new training approach, Break-It-Fix-It (BIFI), which has two key ideas: (i) we use the critic to check a fixer's output on real bad inputs and add good (fixed) outputs to the training data, and (ii) we train a breaker to generate realistic bad code from good code. Based on these ideas, we iteratively update the breaker and the fixer while using them in conjunction to generate more paired data. We evaluate BIFI on two code repair datasets: GitHub-Python, a new dataset we introduce where the goal is to repair Python code with AST parse errors; and DeepFix, where the goal is to repair C code with compiler errors. BIFI outperforms existing methods, obtaining 90.5% repair accuracy on GitHub-Python (+28.5%) and 71.7% on DeepFix (+5.6%). Notably, BIFI does not require any labeled data; we hope it will be a strong starting point for unsupervised learning of various repair tasks."
                    },
                    {
                        "title": "Robust Multilingual Part-of-Speech Tagging via Adversarial Training",
                        "abstract": "Adversarial training (AT) is a powerful regularization method for neural networks, aiming to achieve robustness to input perturbations. Yet, the specific effects of the robustness obtained from AT are still unclear in the context of natural language processing. In this paper, we propose and analyze a neural POS tagging model that exploits AT. In our experiments on the Penn Treebank WSJ corpus and the Universal Dependencies (UD) dataset (27 languages), we find that AT not only improves the overall tagging accuracy, but also 1) prevents over-fitting well in low resource languages and 2) boosts tagging accuracy for rare / unseen words. We also demonstrate that 3) the improved tagging performance by AT contributes to the downstream task of dependency parsing, and that 4) AT helps the model to learn cleaner word representations. 5) The proposed AT model is generally effective in different sequence labeling tasks. These positive results motivate further use of AT for natural language tasks."
                    },
                    {
                        "title": "LM-Critic: Language Models for Unsupervised Grammatical Error Correction",
                        "abstract": "Training a model for grammatical error correction (GEC) requires a set of labeled ungrammatical / grammatical sentence pairs, but manually annotating such pairs can be expensive. Recently, the Break-It-Fix-It (BIFI) framework has demonstrated strong results on learning to repair a broken program without any labeled examples, but this relies on a perfect critic (e.g., a compiler) that returns whether an example is valid or not, which does not exist for the GEC task. In this work, we show how to leverage a pretrained language model (LM) in defining an LM-Critic, which judges a sentence to be grammatical if the LM assigns it a higher probability than its local perturbations. We apply this LM-Critic and BIFI along with a large set of unlabeled sentences to bootstrap realistic ungrammatical / grammatical pairs for training a corrector. We evaluate our approach on GEC datasets across multiple domains (CoNLL-2014, BEA-2019, GMEG-wiki and GMEG-yahoo) and show that it outperforms existing methods in both the unsupervised setting (+7.7 F0.5) and the supervised setting (+0.5 F0.5)."
                    },
                    {
                        "title": "LinkBERT: Pretraining Language Models with Document Links",
                        "abstract": "Language model (LM) pretraining can learn various knowledge from text corpora, helping downstream tasks. However, existing methods such as BERT model a single document, and do not capture dependencies or knowledge that span across documents. In this work, we propose LinkBERT, an LM pretraining method that leverages links between documents, e.g., hyperlinks. Given a text corpus, we view it as a graph of documents and create LM inputs by placing linked documents in the same context. We then pretrain the LM with two joint self-supervised objectives: masked language modeling and our new proposal, document relation prediction. We show that LinkBERT outperforms BERT on various downstream tasks across two domains: the general domain (pretrained on Wikipedia with hyperlinks) and biomedical domain (pretrained on PubMed with citation links). LinkBERT is especially effective for multi-hop reasoning and few-shot QA (+5% absolute improvement on HotpotQA and TriviaQA), and our biomedical LinkBERT sets new states of the art on various BioNLP tasks (+7% on BioASQ and USMLE). We release our pretrained models, LinkBERT and BioLinkBERT, as well as code and data at https://github.com/michiyasunaga/LinkBERT."
                    },
                    {
                        "title": "Med-EASi: Finely Annotated Dataset and Models for Controllable Simplification of Medical Texts",
                        "abstract": "Automatic medical text simplification can assist providers with patient-friendly communication and make medical texts more accessible, thereby improving health literacy. But curating a quality corpus for this task requires the supervision of medical experts. In this work, we present $\\textbf{Med-EASi}$ ($\\underline{\\textbf{Med}}$ical dataset for $\\underline{\\textbf{E}}$laborative and $\\underline{\\textbf{A}}$bstractive $\\underline{\\textbf{Si}}$mplification), a uniquely crowdsourced and finely annotated dataset for supervised simplification of short medical texts. Its $\\textit{expert-layman-AI collaborative}$ annotations facilitate $\\textit{controllability}$ over text simplification by marking four kinds of textual transformations: elaboration, replacement, deletion, and insertion. To learn medical text simplification, we fine-tune T5-large with four different styles of input-output combinations, leading to two control-free and two controllable versions of the model. We add two types of $\\textit{controllability}$ into text simplification, by using a multi-angle training approach: $\\textit{position-aware}$, which uses in-place annotated inputs and outputs, and $\\textit{position-agnostic}$, where the model only knows the contents to be edited, but not their positions. Our results show that our fine-grained annotations improve learning compared to the unannotated baseline. Furthermore, $\\textit{position-aware}$ control generates better simplification than the $\\textit{position-agnostic}$ one. The data and code are available at https://github.com/Chandrayee/CTRL-SIMP."
                    },
                    {
                        "title": "QA-GNN: Reasoning with Language Models and Knowledge Graphs for Question Answering",
                        "abstract": "The problem of answering questions using knowledge from pre-trained language models (LMs) and knowledge graphs (KGs) presents two challenges: given a QA context (question and answer choice), methods need to (i) identify relevant knowledge from large KGs, and (ii) perform joint reasoning over the QA context and KG. In this work, we propose a new model, QA-GNN, which addresses the above challenges through two key innovations: (i) relevance scoring, where we use LMs to estimate the importance of KG nodes relative to the given QA context, and (ii) joint reasoning, where we connect the QA context and KG to form a joint graph, and mutually update their representations through graph neural networks. We evaluate our model on QA benchmarks in the commonsense (CommonsenseQA, OpenBookQA) and biomedical (MedQA-USMLE) domains. QA-GNN outperforms existing LM and LM+KG models, and exhibits capabilities to perform interpretable and structured reasoning, e.g., correctly handling negation in questions."
                    },
                    {
                        "title": "Graph-based Neural Multi-Document Summarization",
                        "abstract": "We propose a neural multi-document summarization (MDS) system that incorporates sentence relation graphs. We employ a Graph Convolutional Network (GCN) on the relation graphs, with sentence embeddings obtained from Recurrent Neural Networks as input node features. Through multiple layer-wise propagation, the GCN generates high-level hidden sentence features for salience estimation. We then use a greedy heuristic to extract salient sentences while avoiding redundancy. In our experiments on DUC 2004, we consider three types of sentence relation graphs and demonstrate the advantage of combining sentence relations in graphs with the representation power of deep neural networks. Our model improves upon traditional graph-based extractive approaches and the vanilla GRU sequence model with no graph, and it achieves competitive results against other state-of-the-art multi-document summarization systems."
                    },
                    {
                        "title": "The CL-SciSumm Shared Task 2018: Results and Key Insights",
                        "abstract": "This overview describes the official results of the CL-SciSumm Shared Task 2018 -- the first medium-scale shared task on scientific document summarization in the computational linguistics (CL) domain. This year, the dataset comprised 60 annotated sets of citing and reference papers from the open access research papers in the CL domain. The Shared Task was organized as a part of the 41st Annual Conference of the Special Interest Group in Information Retrieval (SIGIR), held in Ann Arbor, USA in July 2018. We compare the participating systems in terms of two evaluation metrics. The annotated dataset and evaluation scripts can be accessed and used by the community from: \\url{https://github.com/WING-NUS/scisumm-corpus}."
                    },
                    {
                        "title": "VQA-GNN: Reasoning with Multimodal Knowledge via Graph Neural Networks for Visual Question Answering",
                        "abstract": "Visual question answering (VQA) requires systems to perform concept-level reasoning by unifying unstructured (e.g., the context in question and answer; \"QA context\") and structured (e.g., knowledge graph for the QA context and scene; \"concept graph\") multimodal knowledge. Existing works typically combine a scene graph and a concept graph of the scene by connecting corresponding visual nodes and concept nodes, then incorporate the QA context representation to perform question answering. However, these methods only perform a unidirectional fusion from unstructured knowledge to structured knowledge, limiting their potential to capture joint reasoning over the heterogeneous modalities of knowledge. To perform more expressive reasoning, we propose VQA-GNN, a new VQA model that performs bidirectional fusion between unstructured and structured multimodal knowledge to obtain unified knowledge representations. Specifically, we inter-connect the scene graph and the concept graph through a super node that represents the QA context, and introduce a new multimodal GNN technique to perform inter-modal message passing for reasoning that mitigates representational gaps between modalities. On two challenging VQA tasks (VCR and GQA), our method outperforms strong baseline VQA methods by 3.2% on VCR (Q-AR) and 4.6% on GQA, suggesting its strength in performing concept-level reasoning. Ablation studies further demonstrate the efficacy of the bidirectional fusion and multimodal GNN method in unifying unstructured and structured multimodal knowledge."
                    },
                    {
                        "title": "Neural Coreference Resolution with Deep Biaffine Attention by Joint Mention Detection and Mention Clustering",
                        "abstract": "Coreference resolution aims to identify in a text all mentions that refer to the same real-world entity. The state-of-the-art end-to-end neural coreference model considers all text spans in a document as potential mentions and learns to link an antecedent for each possible mention. In this paper, we propose to improve the end-to-end coreference resolution system by (1) using a biaffine attention model to get antecedent scores for each possible mention, and (2) jointly optimizing the mention detection accuracy and the mention clustering log-likelihood given the mention cluster labels. Our model achieves the state-of-the-art performance on the CoNLL-2012 Shared Task English test set."
                    },
                    {
                        "title": "Is ChatGPT a General-Purpose Natural Language Processing Task Solver?",
                        "abstract": "Spurred by advancements in scale, large language models (LLMs) have demonstrated the ability to perform a variety of natural language processing (NLP) tasks zero-shot -- i.e., without adaptation on downstream data. Recently, the debut of ChatGPT has drawn a great deal of attention from the natural language processing (NLP) community due to the fact that it can generate high-quality responses to human input and self-correct previous mistakes based on subsequent conversations. However, it is not yet known whether ChatGPT can serve as a generalist model that can perform many NLP tasks zero-shot. In this work, we empirically analyze the zero-shot learning ability of ChatGPT by evaluating it on 20 popular NLP datasets covering 7 representative task categories. With extensive empirical studies, we demonstrate both the effectiveness and limitations of the current version of ChatGPT. We find that ChatGPT performs well on many tasks favoring reasoning capabilities (e.g., arithmetic reasoning) while it still faces challenges when solving specific tasks such as sequence tagging. We additionally provide in-depth analysis through qualitative case studies."
                    },
                    {
                        "title": "ScisummNet: A Large Annotated Corpus and Content-Impact Models for Scientific Paper Summarization with Citation Networks",
                        "abstract": "Scientific article summarization is challenging: large, annotated corpora are not available, and the summary should ideally include the article's impacts on research community. This paper provides novel solutions to these two challenges. We 1) develop and release the first large-scale manually-annotated corpus for scientific papers (on computational linguistics) by enabling faster annotation, and 2) propose summarization methods that integrate the authors' original highlights (abstract) and the article's actual impacts on the community (citations), to create comprehensive, hybrid summaries. We conduct experiments to demonstrate the efficacy of our corpus in training data-driven models for scientific paper summarization and the advantage of our hybrid summaries over abstracts and traditional citation-based summaries. Our large annotated corpus and hybrid methods provide a new framework for scientific paper summarization research."
                    },
                    {
                        "title": "Large Language Models as Analogical Reasoners",
                        "abstract": "Chain-of-thought (CoT) prompting for language models demonstrates impressive performance across reasoning tasks, but typically needs labeled exemplars of the reasoning process. In this work, we introduce a new prompting approach, analogical prompting, designed to automatically guide the reasoning process of large language models. Inspired by analogical reasoning, a cognitive process in which humans draw from relevant past experiences to tackle new problems, our approach prompts language models to self-generate relevant exemplars or knowledge in the context, before proceeding to solve the given problem. This method presents several advantages: it obviates the need for labeling or retrieving exemplars, offering generality and convenience; it can also tailor the generated exemplars and knowledge to each problem, offering adaptability. Experimental results show that our approach outperforms 0-shot CoT and manual few-shot CoT in a variety of reasoning tasks, including math problem solving in GSM8K and MATH, code generation in Codeforces, and other reasoning tasks in BIG-Bench."
                    },
                    {
                        "title": "SyntaxSQLNet: Syntax Tree Networks for Complex and Cross-DomainText-to-SQL Task",
                        "abstract": "Most existing studies in text-to-SQL tasks do not require generating complex SQL queries with multiple clauses or sub-queries, and generalizing to new, unseen databases. In this paper we propose SyntaxSQLNet, a syntax tree network to address the complex and cross-domain text-to-SQL generation task. SyntaxSQLNet employs a SQL specific syntax tree-based decoder with SQL generation path history and table-aware column attention encoders. We evaluate SyntaxSQLNet on the Spider text-to-SQL task, which contains databases with multiple tables and complex SQL queries with multiple SQL clauses and nested queries. We use a database split setting where databases in the test set are unseen during training. Experimental results show that SyntaxSQLNet can handle a significantly greater number of complex SQL examples than prior work, outperforming the previous state-of-the-art model by 7.3% in exact matching accuracy. We also show that SyntaxSQLNet can further improve the performance by an additional 7.5% using a cross-domain augmentation method, resulting in a 14.8% improvement in total. To our knowledge, we are the first to study this complex and cross-domain text-to-SQL task."
                    },
                    {
                        "title": "HippoRAG: Neurobiologically Inspired Long-Term Memory for Large Language Models",
                        "abstract": "In order to thrive in hostile and ever-changing natural environments, mammalian brains evolved to store large amounts of knowledge about the world and continually integrate new information while avoiding catastrophic forgetting. Despite the impressive accomplishments, large language models (LLMs), even with retrieval-augmented generation (RAG), still struggle to efficiently and effectively integrate a large amount of new experiences after pre-training. In this work, we introduce HippoRAG, a novel retrieval framework inspired by the hippocampal indexing theory of human long-term memory to enable deeper and more efficient knowledge integration over new experiences. HippoRAG synergistically orchestrates LLMs, knowledge graphs, and the Personalized PageRank algorithm to mimic the different roles of neocortex and hippocampus in human memory. We compare HippoRAG with existing RAG methods on multi-hop question answering and show that our method outperforms the state-of-the-art methods remarkably, by up to 20%. Single-step retrieval with HippoRAG achieves comparable or better performance than iterative retrieval like IRCoT while being 10-30 times cheaper and 6-13 times faster, and integrating HippoRAG into IRCoT brings further substantial gains. Finally, we show that our method can tackle new types of scenarios that are out of reach of existing methods. Code and data are available at https://github.com/OSU-NLP-Group/HippoRAG."
                    }
                ]
            },
            "b7126458-5d5c-40ac-b581-a52267b62bb3": {
                "pk": "b7126458-5d5c-40ac-b581-a52267b62bb3",
                "name": "Kexin Huang",
                "collaborators": [
                    "Marinka Zitnik",
                    "Cao Xiao",
                    "Lucas Glass",
                    "Jimeng Sun",
                    "Jure Leskovec",
                    "Rodrigo Nogueira",
                    "Jaan Altosaar",
                    "Rajesh Ranganath",
                    "Sankeerth Garapati",
                    "Alexander S. Rich"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Bioinformatics",
                    "Machine Learning",
                    "Meta-Learning"
                ],
                "publications": [
                    {
                        "title": "scGNN: scRNA-seq Dropout Imputation via Induced Hierarchical Cell Similarity Graph",
                        "abstract": "Single-cell RNA sequencing provides tremendous insights to understand biological systems. However, the noise from dropout can corrupt the downstream biological analysis. Hence, it is desirable to impute the dropouts accurately. In this work, we propose a simple and powerful dropout imputation method (scGNN) by applying a bottlenecked Graph Convolutional Neural Network on an induced hierarchical cell similarity graph. We show scGNN has competitive performance against state-of-the-art baselines across three datasets and can improve downstream analysis."
                    },
                    {
                        "title": "EpiRL: A Reinforcement Learning Agent to Facilitate Epistasis Detection",
                        "abstract": "Epistasis (gene-gene interaction) is crucial to predicting genetic disease. Our work tackles the computational challenges faced by previous works in epistasis detection by modeling it as a one-step Markov Decision Process where the state is genome data, the actions are the interacted genes, and the reward is an interaction measurement for the selected actions. A reinforcement learning agent using policy gradient method then learns to discover a set of highly interacted genes."
                    },
                    {
                        "title": "Graph Meta Learning via Local Subgraphs",
                        "abstract": "Prevailing methods for graphs require abundant label and edge information for learning. When data for a new task are scarce, meta-learning can learn from prior experiences and form much-needed inductive biases for fast adaption to new tasks. Here, we introduce G-Meta, a novel meta-learning algorithm for graphs. G-Meta uses local subgraphs to transfer subgraph-specific information and learn transferable knowledge faster via meta gradients. G-Meta learns how to quickly adapt to a new task using only a handful of nodes or edges in the new task and does so by learning from data points in other graphs or related, albeit disjoint label sets. G-Meta is theoretically justified as we show that the evidence for a prediction can be found in the local subgraph surrounding the target node or edge. Experiments on seven datasets and nine baseline methods show that G-Meta outperforms existing methods by up to 16.3%. Unlike previous methods, G-Meta successfully learns in challenging, few-shot learning settings that require generalization to completely new graphs and never-before-seen labels. Finally, G-Meta scales to large graphs, which we demonstrate on a new Tree-of-Life dataset comprising of 1,840 graphs, a two-orders of magnitude increase in the number of graphs used in prior work."
                    },
                    {
                        "title": "ClinicalBERT: Modeling Clinical Notes and Predicting Hospital Readmission",
                        "abstract": "Clinical notes contain information about patients that goes beyond structured data like lab values and medications. However, clinical notes have been underused relative to structured data, because notes are high-dimensional and sparse. This work develops and evaluates representations of clinical notes using bidirectional transformers (ClinicalBERT). ClinicalBERT uncovers high-quality relationships between medical concepts as judged by humans. ClinicalBert outperforms baselines on 30-day hospital readmission prediction using both discharge summaries and the first few days of notes in the intensive care unit. Code and model parameters are available."
                    },
                    {
                        "title": "MolTrans: Molecular Interaction Transformer for Drug Target Interaction Prediction",
                        "abstract": "Drug target interaction (DTI) prediction is a foundational task for in silico drug discovery, which is costly and time-consuming due to the need of experimental search over large drug compound space. Recent years have witnessed promising progress for deep learning in DTI predictions. However, the following challenges are still open: (1) the sole data-driven molecular representation learning approaches ignore the sub-structural nature of DTI, thus produce results that are less accurate and difficult to explain; (2) existing methods focus on limited labeled data while ignoring the value of massive unlabelled molecular data. We propose a Molecular Interaction Transformer (MolTrans) to address these limitations via: (1) knowledge inspired sub-structural pattern mining algorithm and interaction modeling module for more accurate and interpretable DTI prediction; (2) an augmented transformer encoder to better extract and capture the semantic relations among substructures extracted from massive unlabeled biomedical data. We evaluate MolTrans on real world data and show it improved DTI prediction performance compared to state-of-the-art baselines."
                    },
                    {
                        "title": "An Interpretable End-to-end Fine-tuning Approach for Long Clinical Text",
                        "abstract": "Unstructured clinical text in EHRs contains crucial information for applications including decision support, trial matching, and retrospective research. Recent work has applied BERT-based models to clinical information extraction and text classification, given these models' state-of-the-art performance in other NLP domains. However, BERT is difficult to apply to clinical notes because it doesn't scale well to long sequences of text. In this work, we propose a novel fine-tuning approach called SnipBERT. Instead of using entire notes, SnipBERT identifies crucial snippets and then feeds them into a truncated BERT-based model in a hierarchical manner. Empirically, SnipBERT not only has significant predictive performance gain across three tasks but also provides improved interpretability, as the model can identify key pieces of text that led to its prediction."
                    },
                    {
                        "title": "Uncertainty Quantification over Graph with Conformalized Graph Neural Networks",
                        "abstract": "Graph Neural Networks (GNNs) are powerful machine learning prediction models on graph-structured data. However, GNNs lack rigorous uncertainty estimates, limiting their reliable deployment in settings where the cost of errors is significant. We propose conformalized GNN (CF-GNN), extending conformal prediction (CP) to graph-based models for guaranteed uncertainty estimates. Given an entity in the graph, CF-GNN produces a prediction set/interval that provably contains the true label with pre-defined coverage probability (e.g. 90%). We establish a permutation invariance condition that enables the validity of CP on graph data and provide an exact characterization of the test-time coverage. Moreover, besides valid coverage, it is crucial to reduce the prediction set size/interval length for practical use. We observe a key connection between non-conformity scores and network structures, which motivates us to develop a topology-aware output correction model that learns to update the prediction and produces more efficient prediction sets/intervals. Extensive experiments show that CF-GNN achieves any pre-defined target marginal coverage while significantly reducing the prediction set/interval size by up to 74% over the baselines. It also empirically achieves satisfactory conditional coverage over various raw and network features."
                    },
                    {
                        "title": "Graph Representation Learning in Biomedicine",
                        "abstract": "Biomedical networks (or graphs) are universal descriptors for systems of interacting elements, from molecular interactions and disease co-morbidity to healthcare systems and scientific knowledge. Advances in artificial intelligence, specifically deep learning, have enabled us to model, analyze, and learn with such networked data. In this review, we put forward an observation that long-standing principles of systems biology and medicine -- while often unspoken in machine learning research -- provide the conceptual grounding for representation learning on graphs, explain its current successes and limitations, and even inform future advancements. We synthesize a spectrum of algorithmic approaches that, at their core, leverage graph topology to embed networks into compact vector spaces. We also capture the breadth of ways in which representation learning has dramatically improved the state-of-the-art in biomedical machine learning. Exemplary domains covered include identifying variants underlying complex traits, disentangling behaviors of single cells and their effects on health, assisting in diagnosis and treatment of patients, and developing safe and effective medicines."
                    },
                    {
                        "title": "Learn to Explore: on Bootstrapping Interactive Data Exploration with Meta-learning",
                        "abstract": "Interactive data exploration (IDE) is an effective way of comprehending big data, whose volume and complexity are beyond human abilities. The main goal of IDE is to discover user interest regions from a database through multi-rounds of user labelling. Existing IDEs adopt active-learning framework, where users iteratively discriminate or label the interestingness of selected tuples. The process of data exploration can be viewed as the process of training a classifier, which determines whether a database tuple is interesting to a user. An efficient exploration thus takes very few iterations of user labelling to reach the data region of interest. In this work, we consider the data exploration as the process of few-shot learning, where the classifier is learned with only a few training examples, or exploration iterations. To this end, we propose a learning-to-explore framework, based on meta-learning, which learns how to learn a classifier with automatically generated meta-tasks, so that the exploration process can be much shortened. Extensive experiments on real datasets show that our proposal outperforms existing explore-by-example solutions in terms of accuracy and efficiency."
                    },
                    {
                        "title": "Enabling tabular deep learning when $d \\gg n$ with an auxiliary knowledge graph",
                        "abstract": "Machine learning models exhibit strong performance on datasets with abundant labeled samples. However, for tabular datasets with extremely high $d$-dimensional features but limited $n$ samples (i.e. $d \\gg n$), machine learning models struggle to achieve strong performance due to the risk of overfitting. Here, our key insight is that there is often abundant, auxiliary domain information describing input features which can be structured as a heterogeneous knowledge graph (KG). We propose PLATO, a method that achieves strong performance on tabular data with $d \\gg n$ by using an auxiliary KG describing input features to regularize a multilayer perceptron (MLP). In PLATO, each input feature corresponds to a node in the auxiliary KG. In the MLP's first layer, each input feature also corresponds to a weight vector. PLATO is based on the inductive bias that two input features corresponding to similar nodes in the auxiliary KG should have similar weight vectors in the MLP's first layer. PLATO captures this inductive bias by inferring the weight vector for each input feature from its corresponding node in the KG via a trainable message-passing function. Across 6 $d \\gg n$ datasets, PLATO outperforms 13 state-of-the-art baselines by up to 10.19%."
                    },
                    {
                        "title": "SkipGNN: Predicting Molecular Interactions with Skip-Graph Networks",
                        "abstract": "Molecular interaction networks are powerful resources for the discovery. They are increasingly used with machine learning methods to predict biologically meaningful interactions. While deep learning on graphs has dramatically advanced the prediction prowess, current graph neural network (GNN) methods are optimized for prediction on the basis of direct similarity between interacting nodes. In biological networks, however, similarity between nodes that do not directly interact has proved incredibly useful in the last decade across a variety of interaction networks. Here, we present SkipGNN, a graph neural network approach for the prediction of molecular interactions. SkipGNN predicts molecular interactions by not only aggregating information from direct interactions but also from second-order interactions, which we call skip similarity. In contrast to existing GNNs, SkipGNN receives neural messages from two-hop neighbors as well as immediate neighbors in the interaction network and non-linearly transforms the messages to obtain useful information for prediction. To inject skip similarity into a GNN, we construct a modified version of the original network, called the skip graph. We then develop an iterative fusion scheme that optimizes a GNN using both the skip graph and the original graph. Experiments on four interaction networks, including drug-drug, drug-target, protein-protein, and gene-disease interactions, show that SkipGNN achieves superior and robust performance, outperforming existing methods by up to 28.8\\% of area under the precision recall curve (PR-AUC). Furthermore, we show that unlike popular GNNs, SkipGNN learns biologically meaningful embeddings and performs especially well on noisy, incomplete interaction networks."
                    }
                ]
            },
            "a2eb30dd-8562-4662-aa46-f3ed6d55c5df": {
                "pk": "a2eb30dd-8562-4662-aa46-f3ed6d55c5df",
                "name": "Kaidi Cao",
                "collaborators": [
                    "Jure Leskovec",
                    "Jiaxuan You",
                    "Jingwei Ji",
                    "Juan Carlos Niebles",
                    "Maria Brbic",
                    "Adrien Gaidon",
                    "Nikos Arechiga",
                    "Tengyu Ma",
                    "Shirley Wu",
                    "Cheng Li"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Multi-task Learning",
                    "Semi-supervised Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "CariGANs: Unpaired Photo-to-Caricature Translation",
                        "abstract": "Facial caricature is an art form of drawing faces in an exaggerated way to convey humor or sarcasm. In this paper, we propose the first Generative Adversarial Network (GAN) for unpaired photo-to-caricature translation, which we call \"CariGANs\". It explicitly models geometric exaggeration and appearance stylization using two components: CariGeoGAN, which only models the geometry-to-geometry transformation from face photos to caricatures, and CariStyGAN, which transfers the style appearance from caricatures to face photos without any geometry deformation. In this way, a difficult cross-domain translation problem is decoupled into two easier tasks. The perceptual study shows that caricatures generated by our CariGANs are closer to the hand-drawn ones, and at the same time better persevere the identity, compared to state-of-the-art methods. Moreover, our CariGANs allow users to control the shape exaggeration degree and change the color/texture style by tuning the parameters or giving an example caricature."
                    },
                    {
                        "title": "Relational Multi-Task Learning: Modeling Relations between Data and Tasks",
                        "abstract": "A key assumption in multi-task learning is that at the inference time the multi-task model only has access to a given data point but not to the data point's labels from other tasks. This presents an opportunity to extend multi-task learning to utilize data point's labels from other auxiliary tasks, and this way improves performance on the new task. Here we introduce a novel relational multi-task learning setting where we leverage data point labels from auxiliary tasks to make more accurate predictions on the new task. We develop MetaLink, where our key innovation is to build a knowledge graph that connects data points and tasks and thus allows us to leverage labels from auxiliary tasks. The knowledge graph consists of two types of nodes: (1) data nodes, where node features are data embeddings computed by the neural network, and (2) task nodes, with the last layer's weights for each task as node features. The edges in this knowledge graph capture data-task relationships, and the edge label captures the label of a data point on a particular task. Under MetaLink, we reformulate the new task as a link label prediction problem between a data node and a task node. The MetaLink framework provides flexibility to model knowledge transfer from auxiliary task labels to the task of interest. We evaluate MetaLink on 6 benchmark datasets in both biochemical and vision domains. Experiments demonstrate that MetaLink can successfully utilize the relations among different tasks, outperforming the state-of-the-art methods under the proposed relational multi-task learning setting, with up to 27% improvement in ROC AUC."
                    },
                    {
                        "title": "AutoTransfer: AutoML with Knowledge Transfer -- An Application to Graph Neural Networks",
                        "abstract": "AutoML has demonstrated remarkable success in finding an effective neural architecture for a given machine learning task defined by a specific dataset and an evaluation metric. However, most present AutoML techniques consider each task independently from scratch, which requires exploring many architectures, leading to high computational cost. Here we propose AutoTransfer, an AutoML solution that improves search efficiency by transferring the prior architectural design knowledge to the novel task of interest. Our key innovation includes a task-model bank that captures the model performance over a diverse set of GNN architectures and tasks, and a computationally efficient task embedding that can accurately measure the similarity among different tasks. Based on the task-model bank and the task embeddings, we estimate the design priors of desirable models of the novel task, by aggregating a similarity-weighted sum of the top-K design distributions on tasks that are similar to the task of interest. The computed design priors can be used with any AutoML search algorithm. We evaluate AutoTransfer on six datasets in the graph machine learning domain. Experiments demonstrate that (i) our proposed task embedding can be computed efficiently, and that tasks with similar embeddings have similar best-performing architectures; (ii) AutoTransfer significantly improves search efficiency with the transferred design priors, reducing the number of explored architectures by an order of magnitude. Finally, we release GNN-Bank-101, a large-scale dataset of detailed GNN training information of 120,000 task-model combinations to facilitate and inspire future research."
                    },
                    {
                        "title": "Learning Temporal Action Proposals With Fewer Labels",
                        "abstract": "Temporal action proposals are a common module in action detection pipelines today. Most current methods for training action proposal modules rely on fully supervised approaches that require large amounts of annotated temporal action intervals in long video sequences. The large cost and effort in annotation that this entails motivate us to study the problem of training proposal modules with less supervision. In this work, we propose a semi-supervised learning algorithm specifically designed for training temporal action proposal networks. When only a small number of labels are available, our semi-supervised method generates significantly better proposals than the fully-supervised counterpart and other strong semi-supervised baselines. We validate our method on two challenging action detection video datasets, ActivityNet v1.3 and THUMOS14. We show that our semi-supervised approach consistently matches or outperforms the fully supervised state-of-the-art approaches."
                    },
                    {
                        "title": "Concept Learners for Few-Shot Learning",
                        "abstract": "Developing algorithms that are able to generalize to a novel task given only a few labeled examples represents a fundamental challenge in closing the gap between machine- and human-level performance. The core of human cognition lies in the structured, reusable concepts that help us to rapidly adapt to new tasks and provide reasoning behind our decisions. However, existing meta-learning methods learn complex representations across prior labeled tasks without imposing any structure on the learned representations. Here we propose COMET, a meta-learning method that improves generalization ability by learning to learn along human-interpretable concept dimensions. Instead of learning a joint unstructured metric space, COMET learns mappings of high-level concepts into semi-structured metric spaces, and effectively combines the outputs of independent concept learners. We evaluate our model on few-shot tasks from diverse domains, including fine-grained image classification, document categorization and cell type annotation on a novel dataset from a biological domain developed in our work. COMET significantly outperforms strong meta-learning baselines, achieving 6-15% relative improvement on the most challenging 1-shot learning tasks, while unlike existing methods providing interpretations behind the model's predictions."
                    },
                    {
                        "title": "Coresets for Robust Training of Neural Networks against Noisy Labels",
                        "abstract": "Modern neural networks have the capacity to overfit noisy labels frequently found in real-world datasets. Although great progress has been made, existing techniques are limited in providing theoretical guarantees for the performance of the neural networks trained with noisy labels. Here we propose a novel approach with strong theoretical guarantees for robust training of deep networks trained with noisy labels. The key idea behind our method is to select weighted subsets (coresets) of clean data points that provide an approximately low-rank Jacobian matrix. We then prove that gradient descent applied to the subsets do not overfit the noisy labels. Our extensive experiments corroborate our theory and demonstrate that deep networks trained on our subsets achieve a significantly superior performance compared to state-of-the art, e.g., 6% increase in accuracy on CIFAR-10 with 80% noisy labels, and 7% increase in accuracy on mini Webvision."
                    },
                    {
                        "title": "Open-World Semi-Supervised Learning",
                        "abstract": "A fundamental limitation of applying semi-supervised learning in real-world settings is the assumption that unlabeled test data contains only classes previously encountered in the labeled training data. However, this assumption rarely holds for data in-the-wild, where instances belonging to novel classes may appear at testing time. Here, we introduce a novel open-world semi-supervised learning setting that formalizes the notion that novel classes may appear in the unlabeled test data. In this novel setting, the goal is to solve the class distribution mismatch between labeled and unlabeled data, where at the test time every input instance either needs to be classified into one of the existing classes or a new unseen class needs to be initialized. To tackle this challenging problem, we propose ORCA, an end-to-end deep learning approach that introduces uncertainty adaptive margin mechanism to circumvent the bias towards seen classes caused by learning discriminative features for seen classes faster than for the novel classes. In this way, ORCA reduces the gap between intra-class variance of seen with respect to novel classes. Experiments on image classification datasets and a single-cell annotation dataset demonstrate that ORCA consistently outperforms alternative baselines, achieving 25% improvement on seen and 96% improvement on novel classes of the ImageNet dataset."
                    },
                    {
                        "title": "Learning Imbalanced Datasets with Label-Distribution-Aware Margin Loss",
                        "abstract": "Deep learning algorithms can fare poorly when the training dataset suffers from heavy class-imbalance but the testing criterion requires good generalization on less frequent classes. We design two novel methods to improve performance in such scenarios. First, we propose a theoretically-principled label-distribution-aware margin (LDAM) loss motivated by minimizing a margin-based generalization bound. This loss replaces the standard cross-entropy objective during training and can be applied with prior strategies for training with class-imbalance such as re-weighting or re-sampling. Second, we propose a simple, yet effective, training schedule that defers re-weighting until after the initial stage, allowing the model to learn an initial representation while avoiding some of the complications associated with re-weighting or re-sampling. We test our methods on several benchmark vision tasks including the real-world imbalanced dataset iNaturalist 2018. Our experiments show that either of these methods alone can already improve over existing techniques and their combination achieves even better performance gains."
                    },
                    {
                        "title": "GraphMETRO: Mitigating Complex Graph Distribution Shifts via Mixture of Aligned Experts",
                        "abstract": "Graph data are inherently complex and heterogeneous, leading to a high natural diversity of distributional shifts. However, it remains unclear how to build machine learning architectures that generalize to complex non-synthetic distributional shifts naturally occurring in the real world. Here we develop GraphMETRO, a Graph Neural Network architecture, that reliably models natural diversity and captures complex distributional shifts. GraphMETRO employs a Mixture-of-Experts (MoE) architecture with a gating model and multiple expert models, where each expert model targets a specific distributional shift to produce a shift-invariant representation, and the gating model identifies shift components. Additionally, we design a novel objective that aligns the representations from different expert models to ensure smooth optimization. GraphMETRO achieves state-of-the-art results on four datasets from GOOD benchmark comprised of complex and natural real-world distribution shifts, improving by 67% and 4.2% on WebKB and Twitch datasets."
                    },
                    {
                        "title": "Few-Shot Video Classification via Temporal Alignment",
                        "abstract": "There is a growing interest in learning a model which could recognize novel classes with only a few labeled examples. In this paper, we propose Temporal Alignment Module (TAM), a novel few-shot learning framework that can learn to classify a previous unseen video. While most previous works neglect long-term temporal ordering information, our proposed model explicitly leverages the temporal ordering information in video data through temporal alignment. This leads to strong data-efficiency for few-shot learning. In concrete, TAM calculates the distance value of query video with respect to novel class proxies by averaging the per frame distances along its alignment path. We introduce continuous relaxation to TAM so the model can be learned in an end-to-end fashion to directly optimize the few-shot learning objective. We evaluate TAM on two challenging real-world datasets, Kinetics and Something-Something-V2, and show that our model leads to significant improvement of few-shot video classification over a wide range of competitive baselines."
                    },
                    {
                        "title": "Heteroskedastic and Imbalanced Deep Learning with Adaptive Regularization",
                        "abstract": "Real-world large-scale datasets are heteroskedastic and imbalanced -- labels have varying levels of uncertainty and label distributions are long-tailed. Heteroskedasticity and imbalance challenge deep learning algorithms due to the difficulty of distinguishing among mislabeled, ambiguous, and rare examples. Addressing heteroskedasticity and imbalance simultaneously is under-explored. We propose a data-dependent regularization technique for heteroskedastic datasets that regularizes different regions of the input space differently. Inspired by the theoretical derivation of the optimal regularization strength in a one-dimensional nonparametric classification setting, our approach adaptively regularizes the data points in higher-uncertainty, lower-density regions more heavily. We test our method on several benchmark tasks, including a real-world heteroskedastic and imbalanced dataset, WebVision. Our experiments corroborate our theory and demonstrate a significant improvement over other methods in noise-robust deep learning."
                    },
                    {
                        "title": "Communication-Free Distributed GNN Training with Vertex Cut",
                        "abstract": "Training Graph Neural Networks (GNNs) on real-world graphs consisting of billions of nodes and edges is quite challenging, primarily due to the substantial memory needed to store the graph and its intermediate node and edge features, and there is a pressing need to speed up the training process. A common approach to achieve speed up is to divide the graph into many smaller subgraphs, which are then distributed across multiple GPUs in one or more machines and processed in parallel. However, existing distributed methods require frequent and substantial cross-GPU communication, leading to significant time overhead and progressively diminishing scalability. Here, we introduce CoFree-GNN, a novel distributed GNN training framework that significantly speeds up the training process by implementing communication-free training. The framework utilizes a Vertex Cut partitioning, i.e., rather than partitioning the graph by cutting the edges between partitions, the Vertex Cut partitions the edges and duplicates the node information to preserve the graph structure. Furthermore, the framework maintains high model accuracy by incorporating a reweighting mechanism to handle a distorted graph distribution that arises from the duplicated nodes. We also propose a modified DropEdge technique to further speed up the training process. Using an extensive set of experiments on real-world networks, we demonstrate that CoFree-GNN speeds up the GNN training process by up to 10 times over the existing state-of-the-art GNN training approaches."
                    },
                    {
                        "title": "Merge or Not? Learning to Group Faces via Imitation Learning",
                        "abstract": "Given a large number of unlabeled face images, face grouping aims at clustering the images into individual identities present in the data. This task remains a challenging problem despite the remarkable capability of deep learning approaches in learning face representation. In particular, grouping results can still be egregious given profile faces and a large number of uninteresting faces and noisy detections. Often, a user needs to correct the erroneous grouping manually. In this study, we formulate a novel face grouping framework that learns clustering strategy from ground-truth simulated behavior. This is achieved through imitation learning (a.k.a apprenticeship learning or learning by watching) via inverse reinforcement learning (IRL). In contrast to existing clustering approaches that group instances by similarity, our framework makes sequential decision to dynamically decide when to merge two face instances/groups driven by short- and long-term rewards. Extensive experiments on three benchmark datasets show that our framework outperforms unsupervised and supervised baselines."
                    },
                    {
                        "title": "TransGaGa: Geometry-Aware Unsupervised Image-to-Image Translation",
                        "abstract": "Unsupervised image-to-image translation aims at learning a mapping between two visual domains. However, learning a translation across large geometry variations always ends up with failure. In this work, we present a novel disentangle-and-translate framework to tackle the complex objects image-to-image translation task. Instead of learning the mapping on the image space directly, we disentangle image space into a Cartesian product of the appearance and the geometry latent spaces. Specifically, we first introduce a geometry prior loss and a conditional VAE loss to encourage the network to learn independent but complementary representations. The translation is then built on appearance and geometry space separately. Extensive experiments demonstrate the superior performance of our method to other state-of-the-art approaches, especially in the challenging near-rigid and non-rigid objects translation tasks. In addition, by taking different exemplars as the appearance references, our method also supports multimodal translation. Project page: https://wywu.github.io/projects/TGaGa/TGaGa.html"
                    }
                ]
            },
            "8c4cf299-e9c6-443b-a879-291a71093b0a": {
                "pk": "8c4cf299-e9c6-443b-a879-291a71093b0a",
                "name": "Qian Huang",
                "collaborators": [
                    "Linyi Qian",
                    "Yulin Chen",
                    "Junzhou Chen",
                    "Chang Li",
                    "Weiwen Qian",
                    "Xuan Ding",
                    "Qiyun Wang",
                    "Chengyuan Yu",
                    "Yingchi Mao"
                ],
                "domain": [
                    "Machine Learning",
                    "Computer Vision",
                    "Ensemble Learning",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "A Voting-Stacking Ensemble of Inception Networks for Cervical Cytology Classification",
                        "abstract": "Cervical cancer is one of the most severe diseases threatening women's health. Early detection and diagnosis can significantly reduce cancer risk, in which cervical cytology classification is indispensable. Researchers have recently designed many networks for automated cervical cancer diagnosis, but the limited accuracy and bulky size of these individual models cannot meet practical application needs. To address this issue, we propose a Voting-Stacking ensemble strategy, which employs three Inception networks as base learners and integrates their outputs through a voting ensemble. The samples misclassified by the ensemble model generate a new training set on which a linear classification model is trained as the meta-learner and performs the final predictions. In addition, a multi-level Stacking ensemble framework is designed to improve performance further. The method is evaluated on the SIPakMed, Herlev, and Mendeley datasets, achieving accuracies of 100%, 100%, and 100%, respectively. The experimental results outperform the current state-of-the-art (SOTA) methods, demonstrating its potential for reducing screening workload and helping pathologists detect cervical cancer."
                    },
                    {
                        "title": "An Improved Artificial Fish Swarm Algorithm for Solving the Problem of Investigation Path Planning",
                        "abstract": "Informationization is a prevailing trend in today's world. The increasing demand for information in decision-making processes poses significant challenges for investigation activities, particularly in terms of effectively allocating limited resources to plan investigation programs. This paper addresses the investigation path planning problem by formulating it as a multi-traveling salesman problem (MTSP). Our objective is to minimize costs, and to achieve this, we propose a chaotic artificial fish swarm algorithm based on multiple population differential evolution (DE-CAFSA). To overcome the limitations of the artificial fish swarm algorithm, such as low optimization accuracy and the inability to consider global and local information, we incorporate adaptive field of view and step size adjustments, replace random behavior with the 2-opt operation, and introduce chaos theory and sub-optimal solutions to enhance optimization accuracy and search performance. Additionally, we integrate the differential evolution algorithm to create a hybrid algorithm that leverages the complementary advantages of both approaches. Experimental results demonstrate that DE-CAFSA outperforms other algorithms on various public datasets of different sizes, as well as showcasing excellent performance on the examples proposed in this study."
                    },
                    {
                        "title": "Exploring the Use of ChatGPT for a Systematic Literature Review: a Design-Based Research",
                        "abstract": "ChatGPT has been used in several educational contexts,including learning, teaching and research. It also has potential to conduct the systematic literature review (SLR). However, there are limited empirical studies on how to use ChatGPT in conducting a SLR. Based on a SLR published,this study used ChatGPT to conduct a SLR of the same 33 papers in a design-based approach, to see what the differences are by comparing the reviews' results,and to answer: To what extent can ChatGPT conduct SLR? What strategies can human researchers utilize to structure prompts for ChatGPT that enhance the reliability and validity of a SLR? This study found that ChatGPT could conduct a SLR. It needs detailed and accurate prompts to analyze the literature. It also has limitations. Guiding principles are summarized from this study for researchers to follow when they need to conduct SLRs using ChatGPT."
                    },
                    {
                        "title": "Enhancing Nucleus Segmentation with HARU-Net: A Hybrid Attention Based Residual U-Blocks Network",
                        "abstract": "Nucleus image segmentation is a crucial step in the analysis, pathological diagnosis, and classification, which heavily relies on the quality of nucleus segmentation. However, the complexity of issues such as variations in nucleus size, blurred nucleus contours, uneven staining, cell clustering, and overlapping cells poses significant challenges. Current methods for nucleus segmentation primarily rely on nuclear morphology or contour-based approaches. Nuclear morphology-based methods exhibit limited generalization ability and struggle to effectively predict irregular-shaped nuclei, while contour-based extraction methods face challenges in accurately segmenting overlapping nuclei. To address the aforementioned issues, we propose a dual-branch network using hybrid attention based residual U-blocks for nucleus instance segmentation. The network simultaneously predicts target information and target contours. Additionally, we introduce a post-processing method that combines the target information and target contours to distinguish overlapping nuclei and generate an instance segmentation image. Within the network, we propose a context fusion block (CF-block) that effectively extracts and merges contextual information from the network. Extensive quantitative evaluations are conducted to assess the performance of our method. Experimental results demonstrate the superior performance of the proposed method compared to state-of-the-art approaches on the BNS, MoNuSeg, CoNSeg, and CPM-17 datasets."
                    },
                    {
                        "title": "DD-GCN: Directed Diffusion Graph Convolutional Network for Skeleton-based Human Action Recognition",
                        "abstract": "Graph Convolutional Networks (GCNs) have been widely used in skeleton-based human action recognition. In GCN-based methods, the spatio-temporal graph is fundamental for capturing motion patterns. However, existing approaches ignore the physical dependency and synchronized spatio-temporal correlations between joints, which limits the representation capability of GCNs. To solve these problems, we construct the directed diffusion graph for action modeling and introduce the activity partition strategy to optimize the weight sharing mechanism of graph convolution kernels. In addition, we present the spatio-temporal synchronization encoder to embed synchronized spatio-temporal semantics. Finally, we propose Directed Diffusion Graph Convolutional Network (DD-GCN) for action recognition, and the experiments on three public datasets: NTU-RGB+D, NTU-RGB+D 120, and NW-UCLA, demonstrate the state-of-the-art performance of our method."
                    }
                ]
            },
            "0dcd9ef3-3df3-4fe1-b3da-990919dcda14": {
                "pk": "0dcd9ef3-3df3-4fe1-b3da-990919dcda14",
                "name": "Vassilis N. Ioannidis",
                "collaborators": [
                    "Georgios B. Giannakis",
                    "George Karypis",
                    "Da Zheng",
                    "Daniel Romero",
                    "Antonio G. Marques",
                    "Xiang Song",
                    "Dimitris Berberidis",
                    "Yanning Shen",
                    "Konstantinos D. Polyzos",
                    "Costas Mavromatis"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Semi-Supervised Learning",
                    "Anomaly Detection",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Edge Dithering for Robust Adaptive Graph Convolutional Networks",
                        "abstract": "Graph convolutional networks (GCNs) are vulnerable to perturbations of the graph structure that are either random, or, adversarially designed. The perturbed links modify the graph neighborhoods, which critically affects the performance of GCNs in semi-supervised learning (SSL) tasks. Aiming at robustifying GCNs conditioned on the perturbed graph, the present paper generates multiple auxiliary graphs, each having its binary 0-1 edge weights flip values with probabilities designed to enhance robustness. The resultant edge-dithered auxiliary graphs are leveraged by an adaptive (A)GCN that performs SSL. Robustness is enabled through learnable graph-combining weights along with suitable regularizers. Relative to GCN, the novel AGCN achieves markedly improved performance in tests with noisy inputs, graph perturbations, and state-of-the-art adversarial attacks. Further experiments with protein interaction networks showcase the competitive performance of AGCN for SSL over multiple graphs."
                    },
                    {
                        "title": "Few-shot link prediction via graph neural networks for Covid-19 drug-repurposing",
                        "abstract": "Predicting interactions among heterogenous graph structured data has numerous applications such as knowledge graph completion, recommendation systems and drug discovery. Often times, the links to be predicted belong to rare types such as the case in repurposing drugs for novel diseases. This motivates the task of few-shot link prediction. Typically, GCNs are ill-equipped in learning such rare link types since the relation embedding is not learned in an inductive fashion. This paper proposes an inductive RGCN for learning informative relation embeddings even in the few-shot learning regime. The proposed inductive model significantly outperforms the RGCN and state-of-the-art KGE models in few-shot learning tasks. Furthermore, we apply our method on the drug-repurposing knowledge graph (DRKG) for discovering drugs for Covid-19. We pose the drug discovery task as link prediction and learn embeddings for the biological entities that partake in the DRKG. Our initial results corroborate that several drugs used in clinical trials were identified as possible drug candidates. The method in this paper are implemented using the efficient deep graph learning (DGL)"
                    },
                    {
                        "title": "PanRep: Graph neural networks for extracting universal node embeddings in heterogeneous graphs",
                        "abstract": "Learning unsupervised node embeddings facilitates several downstream tasks such as node classification and link prediction. A node embedding is universal if it is designed to be used by and benefit various downstream tasks. This work introduces PanRep, a graph neural network (GNN) model, for unsupervised learning of universal node representations for heterogenous graphs. PanRep consists of a GNN encoder that obtains node embeddings and four decoders, each capturing different topological and node feature properties. Abiding to these properties the novel unsupervised framework learns universal embeddings applicable to different downstream tasks. PanRep can be furthered fine-tuned to account for possible limited labels. In this operational setting PanRep is considered as a pretrained model for extracting node embeddings of heterogenous graph data. PanRep outperforms all unsupervised and certain supervised methods in node classification and link prediction, especially when the labeled data for the supervised methods is small. PanRep-FT (with fine-tuning) outperforms all other supervised approaches, which corroborates the merits of pretraining models. Finally, we apply PanRep-FT for discovering novel drugs for Covid-19. We showcase the advantage of universal embeddings in drug repurposing and identify several drugs used in clinical trials as possible drug candidates."
                    },
                    {
                        "title": "Inference of Spatio-Temporal Functions over Graphs via Multi-Kernel Kriged Kalman Filtering",
                        "abstract": "Inference of space-time varying signals on graphs emerges naturally in a plethora of network science related applications. A frequently encountered challenge pertains to reconstructing such dynamic processes, given their values over a subset of vertices and time instants. The present paper develops a graph-aware kernel-based kriged Kalman filter that accounts for the spatio-temporal variations, and offers efficient online reconstruction, even for dynamically evolving network topologies. The kernel-based learning framework bypasses the need for statistical information by capitalizing on the smoothness that graph signals exhibit with respect to the underlying graph. To address the challenge of selecting the appropriate kernel, the proposed filter is combined with a multi-kernel selection module. Such a data-driven method selects a kernel attuned to the signal dynamics on-the-fly within the linear span of a pre-selected dictionary. The novel multi-kernel learning algorithm exploits the eigenstructure of Laplacian kernel matrices to reduce computational complexity. Numerical tests with synthetic and real data demonstrate the superior reconstruction performance of the novel approach relative to state-of-the-art alternatives."
                    },
                    {
                        "title": "GraphSAC: Detecting anomalies in large-scale graphs",
                        "abstract": "A graph-based sampling and consensus (GraphSAC) approach is introduced to effectively detect anomalous nodes in large-scale graphs. Existing approaches rely on connectivity and attributes of all nodes to assign an anomaly score per node. However, nodal attributes and network links might be compromised by adversaries, rendering these holistic approaches vulnerable. Alleviating this limitation, GraphSAC randomly draws subsets of nodes, and relies on graph-aware criteria to judiciously filter out sets contaminated by anomalous nodes, before employing a semi-supervised learning (SSL) module to estimate nominal label distributions per node. These learned nominal distributions are minimally affected by the anomalous nodes, and hence can be directly adopted for anomaly detection. Rigorous analysis provides performance guarantees for GraphSAC, by bounding the required number of draws. The per-draw complexity grows linearly with the number of edges, which implies efficient SSL, while draws can be run in parallel, thereby ensuring scalability to large graphs. GraphSAC is tested under different anomaly generation models based on random walks, clustered anomalies, as well as contemporary adversarial attacks for graph data. Experiments with real-world graphs showcase the advantage of GraphSAC relative to state-of-the-art alternatives."
                    },
                    {
                        "title": "Semi-Blind Inference of Topologies and Dynamical Processes over Graphs",
                        "abstract": "Network science provides valuable insights across numerous disciplines including sociology, biology, neuroscience and engineering. A task of major practical importance in these application domains is inferring the network structure from noisy observations at a subset of nodes. Available methods for topology inference typically assume that the process over the network is observed at all nodes. However, application-specific constraints may prevent acquiring network-wide observations. Alleviating the limited flexibility of existing approaches, this work advocates structural models for graph processes and develops novel algorithms for joint inference of the network topology and processes from partial nodal observations. Structural equation models (SEMs) and structural vector autoregressive models (SVARMs) have well-documented merits in identifying even directed topologies of complex graphs; while SEMs capture contemporaneous causal dependencies among nodes, SVARMs further account for time-lagged influences. This paper develops algorithms that iterate between inferring directed graphs that \"best\" fit the data, and estimating the network processes at reduced computational complexity by leveraging tools related to Kalman smoothing. To further accommodate delay-sensitive applications, an online joint inference approach is put forth that even tracks time-evolving topologies. Furthermore, conditions for identifying the network topology given partial observations are specified. It is proved that the required number of observations for unique identification reduces significantly when the network structure is sparse. Numerical tests with synthetic as well as real datasets corroborate the effectiveness of the novel approach."
                    },
                    {
                        "title": "A Recurrent Graph Neural Network for Multi-Relational Data",
                        "abstract": "The era of data deluge has sparked the interest in graph-based learning methods in a number of disciplines such as sociology, biology, neuroscience, or engineering. In this paper, we introduce a graph recurrent neural network (GRNN) for scalable semi-supervised learning from multi-relational data. Key aspects of the novel GRNN architecture are the use of multi-relational graphs, the dynamic adaptation to the different relations via learnable weights, and the consideration of graph-based regularizers to promote smoothness and alleviate over-parametrization. Our ultimate goal is to design a powerful learning architecture able to: discover complex and highly non-linear data associations, combine (and select) multiple types of relations, and scale gracefully with respect to the size of the graph. Numerical tests with real data sets corroborate the design goals and illustrate the performance gains relative to competing alternatives."
                    },
                    {
                        "title": "Unveiling Anomalous Edges and Nominal Connectivity of Attributed Networks",
                        "abstract": "Uncovering anomalies in attributed networks has recently gained popularity due to its importance in unveiling outliers and flagging adversarial behavior in a gamut of data and network science applications including {the Internet of Things (IoT)}, finance, security, to list a few. The present work deals with uncovering anomalous edges in attributed graphs using two distinct formulations with complementary strengths, which can be easily distributed, and hence efficient. The first relies on decomposing the graph data matrix into low rank plus sparse components to markedly improve performance. The second broadens the scope of the first by performing robust recovery of the unperturbed graph, which enhances the anomaly identification performance. The novel methods not only capture anomalous edges linking nodes of different communities, but also spurious connections between any two nodes with different features. Experiments conducted on real and synthetic data corroborate the effectiveness of both methods in the anomaly identification task."
                    },
                    {
                        "title": "Scalable Consistency Training for Graph Neural Networks via Self-Ensemble Self-Distillation",
                        "abstract": "Consistency training is a popular method to improve deep learning models in computer vision and natural language processing. Graph neural networks (GNNs) have achieved remarkable performance in a variety of network science learning tasks, but to date no work has studied the effect of consistency training on large-scale graph problems. GNNs scale to large graphs by minibatch training and subsample node neighbors to deal with high degree nodes. We utilize the randomness inherent in the subsampling of neighbors and introduce a novel consistency training method to improve accuracy. For a target node we generate different neighborhood expansions, and distill the knowledge of the average of the predictions to the GNN. Our method approximates the expected prediction of the possible neighborhood samples and practically only requires a few samples. We demonstrate that our training method outperforms standard GNN training in several different settings, and yields the largest gains when label rates are low."
                    },
                    {
                        "title": "Kernel-based Reconstruction of Space-time Functions on Dynamic Graphs",
                        "abstract": "Graph-based methods pervade the inference toolkits of numerous disciplines including sociology, biology, neuroscience, physics, chemistry, and engineering. A challenging problem encountered in this context pertains to determining the attributes of a set of vertices given those of another subset at possibly different time instants. Leveraging spatiotemporal dynamics can drastically reduce the number of observed vertices, and hence the cost of sampling. Alleviating the limited flexibility of existing approaches, the present paper broadens the existing kernel-based graph function reconstruction framework to accommodate time-evolving functions over possibly time-evolving topologies. This approach inherits the versatility and generality of kernel-based methods, for which no knowledge on distributions or second-order statistics is required. Systematic guidelines are provided to construct two families of space-time kernels with complementary strengths. The first facilitates judicious control of regularization on a space-time frequency plane, whereas the second can afford time-varying topologies. Batch and online estimators are also put forth, and a novel kernel Kalman filter is developed to obtain these estimates at affordable computational cost. Numerical tests with real data sets corroborate the merits of the proposed methods relative to competing alternatives."
                    },
                    {
                        "title": "Efficient and Stable Graph Scattering Transforms via Pruning",
                        "abstract": "Graph convolutional networks (GCNs) have well-documented performance in various graph learning tasks, but their analysis is still at its infancy. Graph scattering transforms (GSTs) offer training-free deep GCN models that extract features from graph data, and are amenable to generalization and stability analyses. The price paid by GSTs is exponential complexity in space and time that increases with the number of layers. This discourages deployment of GSTs when a deep architecture is needed. The present work addresses the complexity limitation of GSTs by introducing an efficient so-termed pruned (p)GST approach. The resultant pruning algorithm is guided by a graph-spectrum-inspired criterion, and retains informative scattering features on-the-fly while bypassing the exponential complexity associated with GSTs. Stability of the novel pGSTs is also established when the input graph data or the network structure are perturbed. Furthermore, the sensitivity of pGST to random and localized signal perturbations is investigated analytically and experimentally. Numerical tests showcase that pGST performs comparably to the baseline GST at considerable computational savings. Furthermore, pGST achieves comparable performance to state-of-the-art GCNs in graph and 3D point cloud classification tasks. Upon analyzing the pGST pruning patterns, it is shown that graph data in different domains call for different network architectures, and that the pruning algorithm may be employed to guide the design choices for contemporary GCNs."
                    },
                    {
                        "title": "Tensor Graph Convolutional Networks for Multi-relational and Robust Learning",
                        "abstract": "The era of \"data deluge\" has sparked renewed interest in graph-based learning methods and their widespread applications ranging from sociology and biology to transportation and communications. In this context of graph-aware methods, the present paper introduces a tensor-graph convolutional network (TGCN) for scalable semi-supervised learning (SSL) from data associated with a collection of graphs, that are represented by a tensor. Key aspects of the novel TGCN architecture are the dynamic adaptation to different relations in the tensor graph via learnable weights, and the consideration of graph-based regularizers to promote smoothness and alleviate over-parameterization. The ultimate goal is to design a powerful learning architecture able to: discover complex and highly nonlinear data associations, combine (and select) multiple types of relations, scale gracefully with the graph size, and remain robust to perturbations on the graph edges. The proposed architecture is relevant not only in applications where the nodes are naturally involved in different relations (e.g., a multi-relational graph capturing family, friendship and work relations in a social network), but also in robust learning setups where the graph entails a certain level of uncertainty, and the different tensor slabs correspond to different versions (realizations) of the nominal graph. Numerical tests showcase that the proposed architecture achieves markedly improved performance relative to standard GCNs, copes with state-of-the-art adversarial attacks, and leads to remarkable SSL performance over protein-to-protein interaction networks."
                    },
                    {
                        "title": "Kernel-based Inference of Functions over Graphs",
                        "abstract": "The study of networks has witnessed an explosive growth over the past decades with several ground-breaking methods introduced. A particularly interesting -- and prevalent in several fields of study -- problem is that of inferring a function defined over the nodes of a network. This work presents a versatile kernel-based framework for tackling this inference problem that naturally subsumes and generalizes the reconstruction approaches put forth recently by the signal processing on graphs community. Both the static and the dynamic settings are considered along with effective modeling approaches for addressing real-world problems. The herein analytical discussion is complemented by a set of numerical examples, which showcase the effectiveness of the presented techniques, as well as their merits related to state-of-the-art methods."
                    },
                    {
                        "title": "Efficient and effective training of language and graph neural network models",
                        "abstract": "Can we combine heterogenous graph structure with text to learn high-quality semantic and behavioural representations? Graph neural networks (GNN)s encode numerical node attributes and graph structure to achieve impressive performance in a variety of supervised learning tasks. Current GNN approaches are challenged by textual features, which typically need to be encoded to a numerical vector before provided to the GNN that may incur some information loss. In this paper, we put forth an efficient and effective framework termed language model GNN (LM-GNN) to jointly train large-scale language models and graph neural networks. The effectiveness in our framework is achieved by applying stage-wise fine-tuning of the BERT model first with heterogenous graph information and then with a GNN model. Several system and design optimizations are proposed to enable scalable and efficient training. LM-GNN accommodates node and edge classification as well as link prediction tasks. We evaluate the LM-GNN framework in different datasets performance and showcase the effectiveness of the proposed approach. LM-GNN provides competitive results in an Amazon query-purchase-product application."
                    },
                    {
                        "title": "A Robust Stacking Framework for Training Deep Graph Models with Multifaceted Node Features",
                        "abstract": "Graph Neural Networks (GNNs) with numerical node features and graph structure as inputs have demonstrated superior performance on various supervised learning tasks with graph data. However the numerical node features utilized by GNNs are commonly extracted from raw data which is of text or tabular (numeric/categorical) type in most real-world applications. The best models for such data types in most standard supervised learning settings with IID (non-graph) data are not simple neural network layers and thus are not easily incorporated into a GNN. Here we propose a robust stacking framework that fuses graph-aware propagation with arbitrary models intended for IID data, which are ensembled and stacked in multiple layers. Our layer-wise framework leverages bagging and stacking strategies to enjoy strong generalization, in a manner which effectively mitigates label leakage and overfitting. Across a variety of graph datasets with tabular/text node features, our method achieves comparable or superior performance relative to both tabular/text and graph neural network models, as well as existing state-of-the-art hybrid strategies that combine the two."
                    },
                    {
                        "title": "BioBridge: Bridging Biomedical Foundation Models via Knowledge Graphs",
                        "abstract": "Foundation models (FMs) are able to leverage large volumes of unlabeled data to demonstrate superior performance across a wide range of tasks. However, FMs developed for biomedical domains have largely remained unimodal, i.e., independently trained and used for tasks on protein sequences alone, small molecule structures alone, or clinical data alone. To overcome this limitation of biomedical FMs, we present BioBridge, a novel parameter-efficient learning framework, to bridge independently trained unimodal FMs to establish multimodal behavior. BioBridge achieves it by utilizing Knowledge Graphs (KG) to learn transformations between one unimodal FM and another without fine-tuning any underlying unimodal FMs. Our empirical results demonstrate that BioBridge can beat the best baseline KG embedding methods (on average by around 76.3%) in cross-modal retrieval tasks. We also identify BioBridge demonstrates out-of-domain generalization ability by extrapolating to unseen modalities or relations. Additionally, we also show that BioBridge presents itself as a general purpose retriever that can aid biomedical multimodal question answering as well as enhance the guided generation of novel drugs."
                    },
                    {
                        "title": "TouchUp-G: Improving Feature Representation through Graph-Centric Finetuning",
                        "abstract": "How can we enhance the node features acquired from Pretrained Models (PMs) to better suit downstream graph learning tasks? Graph Neural Networks (GNNs) have become the state-of-the-art approach for many high-impact, real-world graph applications. For feature-rich graphs, a prevalent practice involves utilizing a PM directly to generate features, without incorporating any domain adaptation techniques. Nevertheless, this practice is suboptimal because the node features extracted from PM are graph-agnostic and prevent GNNs from fully utilizing the potential correlations between the graph structure and node features, leading to a decline in GNNs performance. In this work, we seek to improve the node features obtained from a PM for downstream graph tasks and introduce TOUCHUP-G, which has several advantages. It is (a) General: applicable to any downstream graph task, including link prediction which is often employed in recommender systems; (b) Multi-modal: able to improve raw features of any modality (e.g. images, texts, audio); (c) Principled: it is closely related to a novel metric, feature homophily, which we propose to quantify the potential correlations between the graph structure and node features and we show that TOUCHUP-G can effectively shrink the discrepancy between the graph structure and node features; (d) Effective: achieving state-of-the-art results on four real-world datasets spanning different tasks and modalities."
                    },
                    {
                        "title": "Coupled Graphs and Tensor Factorization for Recommender Systems and Community Detection",
                        "abstract": "Joint analysis of data from multiple information repositories facilitates uncovering the underlying structure in heterogeneous datasets. Single and coupled matrix-tensor factorization (CMTF) has been widely used in this context for imputation-based recommendation from ratings, social network, and other user-item data. When this side information is in the form of item-item correlation matrices or graphs, existing CMTF algorithms may fall short. Alleviating current limitations, we introduce a novel model coined coupled graph-tensor factorization (CGTF) that judiciously accounts for graph-related side information. The CGTF model has the potential to overcome practical challenges, such as missing slabs from the tensor and/or missing rows/columns from the correlation matrices. A novel alternating direction method of multipliers (ADMM) is also developed that recovers the nonnegative factors of CGTF. Our algorithm enjoys closed-form updates that result in reduced computational complexity and allow for convergence claims. A novel direction is further explored by employing the interpretable factors to detect graph communities having the tensor as side information. The resulting community detection approach is successful even when some links in the graphs are missing. Results with real data sets corroborate the merits of the proposed methods relative to state-of-the-art competing factorization techniques in providing recommendations and detecting communities."
                    }
                ]
            },
            "5352fa32-de70-42c1-8e6c-54a0247f7a16": {
                "pk": "5352fa32-de70-42c1-8e6c-54a0247f7a16",
                "name": "Karthik Subbian",
                "collaborators": [
                    "Nikhil Rao",
                    "Edward W Huang",
                    "Nurendra Choudhary",
                    "Chandan K. Reddy",
                    "Sumeet Katariya",
                    "Wenqing Zheng",
                    "Zhangyang Wang",
                    "Jure Leskovec",
                    "Charles Dickens",
                    "Aishwarya Reganti"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Social Network Analysis",
                    "Knowledge Graph",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Supervised Rank Aggregation for Predicting Influence in Networks",
                        "abstract": "Much work in Social Network Analysis has focused on the identification of the most important actors in a social network. This has resulted in several measures of influence and authority. While most of such sociometrics (e.g., PageRank) are driven by intuitions based on an actors location in a network, asking for the \"most influential\" actors in itself is an ill-posed question, unless it is put in context with a specific measurable task. Constructing a predictive task of interest in a given domain provides a mechanism to quantitatively compare different measures of influence. Furthermore, when we know what type of actionable insight to gather, we need not rely on a single network centrality measure. A combination of measures is more likely to capture various aspects of the social network that are predictive and beneficial for the task. Towards this end, we propose an approach to supervised rank aggregation, driven by techniques from Social Choice Theory. We illustrate the effectiveness of this method through experiments on Twitter and citation networks."
                    },
                    {
                        "title": "Learning Robust Models for e-Commerce Product Search",
                        "abstract": "Showing items that do not match search query intent degrades customer experience in e-commerce. These mismatches result from counterfactual biases of the ranking algorithms toward noisy behavioral signals such as clicks and purchases in the search logs. Mitigating the problem requires a large labeled dataset, which is expensive and time-consuming to obtain. In this paper, we develop a deep, end-to-end model that learns to effectively classify mismatches and to generate hard mismatched examples to improve the classifier. We train the model end-to-end by introducing a latent variable into the cross-entropy loss that alternates between using the real and generated samples. This not only makes the classifier more robust but also boosts the overall ranking performance. Our model achieves a relative gain compared to baselines by over 26% in F-score, and over 17% in Area Under PR curve. On live search traffic, our model gains significant improvement in multiple countries."
                    },
                    {
                        "title": "Scalable Feature Selection for (Multitask) Gradient Boosted Trees",
                        "abstract": "Gradient Boosted Decision Trees (GBDTs) are widely used for building ranking and relevance models in search and recommendation. Considerations such as latency and interpretability dictate the use of as few features as possible to train these models. Feature selection in GBDT models typically involves heuristically ranking the features by importance and selecting the top few, or by performing a full backward feature elimination routine. On-the-fly feature selection methods proposed previously scale suboptimally with the number of features, which can be daunting in high dimensional settings. We develop a scalable forward feature selection variant for GBDT, via a novel group testing procedure that works well in high dimensions, and enjoys favorable theoretical performance and computational guarantees. We show via extensive experiments on both public and proprietary datasets that the proposed method offers significant speedups in training time, while being as competitive as existing GBDT methods in terms of model performance metrics. We also extend the method to the multitask setting, allowing the practitioner to select common features across tasks, as well as selecting task-specific features."
                    },
                    {
                        "title": "Dynamic Matrix Factorization: A State Space Approach",
                        "abstract": "Matrix factorization from a small number of observed entries has recently garnered much attention as the key ingredient of successful recommendation systems. One unresolved problem in this area is how to adapt current methods to handle changing user preferences over time. Recent proposals to address this issue are heuristic in nature and do not fully exploit the time-dependent structure of the problem. As a principled and general temporal formulation, we propose a dynamical state space model of matrix factorization. Our proposal builds upon probabilistic matrix factorization, a Bayesian model with Gaussian priors. We utilize results in state tracking, such as the Kalman filter, to provide accurate recommendations in the presence of both process and measurement noise. We show how system parameters can be learned via expectation-maximization and provide comparisons to current published techniques."
                    },
                    {
                        "title": "Dynamics of Trust Reciprocation in Heterogenous MMOG Networks",
                        "abstract": "Understanding the dynamics of reciprocation is of great interest in sociology and computational social science. The recent growth of Massively Multi-player Online Games (MMOGs) has provided unprecedented access to large-scale data which enables us to study such complex human behavior in a more systematic manner. In this paper, we consider three different networks in the EverQuest2 game: chat, trade, and trust. The chat network has the highest level of reciprocation (33%) because there are essentially no barriers to it. The trade network has a lower rate of reciprocation (27%) because it has the obvious barrier of requiring more goods or money for exchange; morever, there is no clear benefit to returning a trade link except in terms of social connections. The trust network has the lowest reciprocation (14%) because this equates to sharing certain within-game assets such as weapons, and so there is a high barrier for such connections because they require faith in the players that are granted such high access. In general, we observe that reciprocation rate is inversely related to the barrier level in these networks. We also note that reciprocation has connections across the heterogeneous networks. Our experiments indicate that players make use of the medium-barrier reciprocations to strengthen a relationship. We hypothesize that lower-barrier interactions are an important component to predicting higher-barrier ones. We verify our hypothesis using predictive models for trust reciprocations using features from trade interactions. Using the number of trades (both before and after the initial trust link) boosts our ability to predict if the trust will be reciprocated up to 11% with respect to the AUC."
                    },
                    {
                        "title": "Graph Coarsening via Convolution Matching for Scalable Graph Neural Network Training",
                        "abstract": "Graph summarization as a preprocessing step is an effective and complementary technique for scalable graph neural network (GNN) training. In this work, we propose the Coarsening Via Convolution Matching (CONVMATCH) algorithm and a highly scalable variant, A-CONVMATCH, for creating summarized graphs that preserve the output of graph convolution. We evaluate CONVMATCH on six real-world link prediction and node classification graph datasets, and show it is efficient and preserves prediction performance while significantly reducing the graph size. Notably, CONVMATCH achieves up to 95% of the prediction performance of GNNs on node classification while trained on graphs summarized down to 1% the size of the original graph. Furthermore, on link prediction tasks, CONVMATCH consistently outperforms all baselines, achieving up to a 2x improvement."
                    },
                    {
                        "title": "Text Enriched Sparse Hyperbolic Graph Convolutional Networks",
                        "abstract": "Heterogeneous networks, which connect informative nodes containing text with different edge types, are routinely used to store and process information in various real-world applications. Graph Neural Networks (GNNs) and their hyperbolic variants provide a promising approach to encode such networks in a low-dimensional latent space through neighborhood aggregation and hierarchical feature extraction, respectively. However, these approaches typically ignore metapath structures and the available semantic information. Furthermore, these approaches are sensitive to the noise present in the training data. To tackle these limitations, in this paper, we propose Text Enriched Sparse Hyperbolic Graph Convolution Network (TESH-GCN) to capture the graph's metapath structures using semantic signals and further improve prediction in large heterogeneous graphs. In TESH-GCN, we extract semantic node information, which successively acts as a connection signal to extract relevant nodes' local neighborhood and graph-level metapath features from the sparse adjacency tensor in a reformulated hyperbolic graph convolution layer. These extracted features in conjunction with semantic features from the language model (for robustness) are used for the final downstream task. Experiments on various heterogeneous graph datasets show that our model outperforms the current state-of-the-art approaches by a large margin on the task of link prediction. We also report a reduction in both the training time and model parameters compared to the existing hyperbolic approaches through a reformulated hyperbolic graph convolution. Furthermore, we illustrate the robustness of our model by experimenting with different levels of simulated noise in both the graph structure and text, and also, present a mechanism to explain TESH-GCN's prediction by analyzing the extracted metapaths."
                    },
                    {
                        "title": "Probabilistic Entity Representation Model for Reasoning over Knowledge Graphs",
                        "abstract": "Logical reasoning over Knowledge Graphs (KGs) is a fundamental technique that can provide efficient querying mechanism over large and incomplete databases. Current approaches employ spatial geometries such as boxes to learn query representations that encompass the answer entities and model the logical operations of projection and intersection. However, their geometry is restrictive and leads to non-smooth strict boundaries, which further results in ambiguous answer entities. Furthermore, previous works propose transformation tricks to handle unions which results in non-closure and, thus, cannot be chained in a stream. In this paper, we propose a Probabilistic Entity Representation Model (PERM) to encode entities as a Multivariate Gaussian density with mean and covariance parameters to capture its semantic position and smooth decision boundary, respectively. Additionally, we also define the closed logical operations of projection, intersection, and union that can be aggregated using an end-to-end objective function. On the logical query reasoning problem, we demonstrate that the proposed PERM significantly outperforms the state-of-the-art methods on various public benchmark KG datasets on standard evaluation metrics. We also evaluate PERM's competence on a COVID-19 drug-repurposing case study and show that our proposed work is able to recommend drugs with substantially better F1 than current methods. Finally, we demonstrate the working of our PERM's query answering process through a low-dimensional visualization of the Gaussian representations."
                    },
                    {
                        "title": "Cold Brew: Distilling Graph Node Representations with Incomplete or Missing Neighborhoods",
                        "abstract": "Graph Neural Networks (GNNs) have achieved state-of-the-art performance in node classification, regression, and recommendation tasks. GNNs work well when rich and high-quality connections are available. However, their effectiveness is often jeopardized in many real-world graphs in which node degrees have power-law distributions. The extreme case of this situation, where a node may have no neighbors, is called Strict Cold Start (SCS). SCS forces the prediction to rely completely on the node's own features. We propose Cold Brew, a teacher-student distillation approach to address the SCS and noisy-neighbor challenges for GNNs. We also introduce feature contribution ratio (FCR), a metric to quantify the behavior of inductive GNNs to solve SCS. We experimentally show that FCR disentangles the contributions of different graph data components and helps select the best architecture for SCS generalization. We further demonstrate the superior performance of Cold Brew on several public benchmark and proprietary e-commerce datasets, where many nodes have either very few or noisy connections. Our source code is available at https://github.com/amazon-research/gnn-tail-generalization."
                    },
                    {
                        "title": "Task-Agnostic Graph Explanations",
                        "abstract": "Graph Neural Networks (GNNs) have emerged as powerful tools to encode graph-structured data. Due to their broad applications, there is an increasing need to develop tools to explain how GNNs make decisions given graph-structured data. Existing learning-based GNN explanation approaches are task-specific in training and hence suffer from crucial drawbacks. Specifically, they are incapable of producing explanations for a multitask prediction model with a single explainer. They are also unable to provide explanations in cases where the GNN is trained in a self-supervised manner, and the resulting representations are used in future downstream tasks. To address these limitations, we propose a Task-Agnostic GNN Explainer (TAGE) that is independent of downstream models and trained under self-supervision with no knowledge of downstream tasks. TAGE enables the explanation of GNN embedding models with unseen downstream tasks and allows efficient explanation of multitask models. Our extensive experiments show that TAGE can significantly speed up the explanation efficiency by using the same model to explain predictions for multiple downstream tasks while achieving explanation quality as good as or even better than current state-of-the-art GNN explanation approaches. Our code is pubicly available as part of the DIG library at https://github.com/divelab/DIG/tree/main/dig/xgraph/TAGE/."
                    },
                    {
                        "title": "Search Behavior Prediction: A Hypergraph Perspective",
                        "abstract": "Although the bipartite shopping graphs are straightforward to model search behavior, they suffer from two challenges: 1) The majority of items are sporadically searched and hence have noisy/sparse query associations, leading to a \\textit{long-tail} distribution. 2) Infrequent queries are more likely to link to popular items, leading to another hurdle known as \\textit{disassortative mixing}. To address these two challenges, we go beyond the bipartite graph to take a hypergraph perspective, introducing a new paradigm that leverages \\underline{auxiliary} information from anonymized customer engagement sessions to assist the \\underline{main task} of query-item link prediction. This auxiliary information is available at web scale in the form of search logs. We treat all items appearing in the same customer session as a single hyperedge. The hypothesis is that items in a customer session are unified by a common shopping interest. With these hyperedges, we augment the original bipartite graph into a new \\textit{hypergraph}. We develop a \\textit{\\textbf{D}ual-\\textbf{C}hannel \\textbf{A}ttention-Based \\textbf{H}ypergraph Neural Network} (\\textbf{DCAH}), which synergizes information from two potentially noisy sources (original query-item edges and item-item hyperedges). In this way, items on the tail are better connected due to the extra hyperedges, thereby enhancing their link prediction performance. We further integrate DCAH with self-supervised graph pre-training and/or DropEdge training, both of which effectively alleviate disassortative mixing. Extensive experiments on three proprietary E-Commerce datasets show that DCAH yields significant improvements of up to \\textbf{24.6\\% in mean reciprocal rank (MRR)} and \\textbf{48.3\\% in recall} compared to GNN-based baselines. Our source code is available at \\url{https://github.com/amazon-science/dual-channel-hypergraph-neural-network}."
                    },
                    {
                        "title": "You Only Transfer What You Share: Intersection-Induced Graph Transfer Learning for Link Prediction",
                        "abstract": "Link prediction is central to many real-world applications, but its performance may be hampered when the graph of interest is sparse. To alleviate issues caused by sparsity, we investigate a previously overlooked phenomenon: in many cases, a densely connected, complementary graph can be found for the original graph. The denser graph may share nodes with the original graph, which offers a natural bridge for transferring selective, meaningful knowledge. We identify this setting as Graph Intersection-induced Transfer Learning (GITL), which is motivated by practical applications in e-commerce or academic co-authorship predictions. We develop a framework to effectively leverage the structural prior in this setting. We first create an intersection subgraph using the shared nodes between the two graphs, then transfer knowledge from the source-enriched intersection subgraph to the full target graph. In the second step, we consider two approaches: a modified label propagation, and a multi-layer perceptron (MLP) model in a teacher-student regime. Experimental results on proprietary e-commerce datasets and open-source citation graphs show that the proposed workflow outperforms existing transfer learning baselines that do not explicitly utilize the intersection structure."
                    },
                    {
                        "title": "Simplifying Distributed Neural Network Training on Massive Graphs: Randomized Partitions Improve Model Aggregation",
                        "abstract": "Distributed training of GNNs enables learning on massive graphs (e.g., social and e-commerce networks) that exceed the storage and computational capacity of a single machine. To reach performance comparable to centralized training, distributed frameworks focus on maximally recovering cross-instance node dependencies with either communication across instances or periodic fallback to centralized training, which create overhead and limit the framework scalability. In this work, we present a simplified framework for distributed GNN training that does not rely on the aforementioned costly operations, and has improved scalability, convergence speed and performance over the state-of-the-art approaches. Specifically, our framework (1) assembles independent trainers, each of which asynchronously learns a local model on locally-available parts of the training graph, and (2) only conducts periodic (time-based) model aggregation to synchronize the local models. Backed by our theoretical analysis, instead of maximizing the recovery of cross-instance node dependencies -- which has been considered the key behind closing the performance gap between model aggregation and centralized training -- , our framework leverages randomized assignment of nodes or super-nodes (i.e., collections of original nodes) to partition the training graph such that it improves data uniformity and minimizes the discrepancy of gradient and loss function across instances. In our experiments on social and e-commerce networks with up to 1.3 billion edges, our proposed RandomTMA and SuperTMA approaches -- despite using less training data -- achieve state-of-the-art performance and 2.31x speedup compared to the fastest baseline, and show better robustness to trainer failures."
                    },
                    {
                        "title": "Communication-Free Distributed GNN Training with Vertex Cut",
                        "abstract": "Training Graph Neural Networks (GNNs) on real-world graphs consisting of billions of nodes and edges is quite challenging, primarily due to the substantial memory needed to store the graph and its intermediate node and edge features, and there is a pressing need to speed up the training process. A common approach to achieve speed up is to divide the graph into many smaller subgraphs, which are then distributed across multiple GPUs in one or more machines and processed in parallel. However, existing distributed methods require frequent and substantial cross-GPU communication, leading to significant time overhead and progressively diminishing scalability. Here, we introduce CoFree-GNN, a novel distributed GNN training framework that significantly speeds up the training process by implementing communication-free training. The framework utilizes a Vertex Cut partitioning, i.e., rather than partitioning the graph by cutting the edges between partitions, the Vertex Cut partitions the edges and duplicates the node information to preserve the graph structure. Furthermore, the framework maintains high model accuracy by incorporating a reweighting mechanism to handle a distorted graph distribution that arises from the duplicated nodes. We also propose a modified DropEdge technique to further speed up the training process. Using an extensive set of experiments on real-world networks, we demonstrate that CoFree-GNN speeds up the GNN training process by up to 10 times over the existing state-of-the-art GNN training approaches."
                    },
                    {
                        "title": "Do Diffusion Protocols Govern Cascade Growth?",
                        "abstract": "Large cascades can develop in online social networks as people share information with one another. Though simple reshare cascades have been studied extensively, the full range of cascading behaviors on social media is much more diverse. Here we study how diffusion protocols, or the social exchanges that enable information transmission, affect cascade growth, analogous to the way communication protocols define how information is transmitted from one point to another. Studying 98 of the largest information cascades on Facebook, we find a wide range of diffusion protocols - from cascading reshares of images, which use a simple protocol of tapping a single button for propagation, to the ALS Ice Bucket Challenge, whose diffusion protocol involved individuals creating and posting a video, and then nominating specific others to do the same. We find recurring classes of diffusion protocols, and identify two key counterbalancing factors in the construction of these protocols, with implications for a cascade's growth: the effort required to participate in the cascade, and the social cost of staying on the sidelines. Protocols requiring greater individual effort slow down a cascade's propagation, while those imposing a greater social cost of not participating increase the cascade's adoption likelihood. The predictability of transmission also varies with protocol. But regardless of mechanism, the cascades in our analysis all have a similar reproduction number ($\\approx$ 1.8), meaning that lower rates of exposure can be offset with higher per-exposure rates of adoption. Last, we show how a cascade's structure can not only differentiate these protocols, but also be modeled through branching processes. Together, these findings provide a framework for understanding how a wide variety of information cascades can achieve substantial adoption across a network."
                    },
                    {
                        "title": "Self-Supervised Hyperboloid Representations from Logical Queries over Knowledge Graphs",
                        "abstract": "Knowledge Graphs (KGs) are ubiquitous structures for information storagein several real-world applications such as web search, e-commerce, social networks, and biology. Querying KGs remains a foundational and challenging problem due to their size and complexity. Promising approaches to tackle this problem include embedding the KG units (e.g., entities and relations) in a Euclidean space such that the query embedding contains the information relevant to its results. These approaches, however, fail to capture the hierarchical nature and semantic information of the entities present in the graph. Additionally, most of these approaches only utilize multi-hop queries (that can be modeled by simple translation operations) to learn embeddings and ignore more complex operations such as intersection and union of simpler queries. To tackle such complex operations, in this paper, we formulate KG representation learning as a self-supervised logical query reasoning problem that utilizes translation, intersection and union queries over KGs. We propose Hyperboloid Embeddings (HypE), a novel self-supervised dynamic reasoning framework, that utilizes positive first-order existential queries on a KG to learn representations of its entities and relations as hyperboloids in a Poincar\\'e ball. HypE models the positive first-order queries as geometrical translation, intersection, and union. For the problem of KG reasoning in real-world datasets, the proposed HypE model significantly outperforms the state-of-the art results. We also apply HypE to an anomaly detection task on a popular e-commerce website product taxonomy as well as hierarchically organized web articles and demonstrate significant performance improvements compared to existing baseline methods. Finally, we also visualize the learned HypE embeddings in a Poincar\\'e ball to clearly interpret and comprehend the representation space."
                    },
                    {
                        "title": "Learning Backward Compatible Embeddings",
                        "abstract": "Embeddings, low-dimensional vector representation of objects, are fundamental in building modern machine learning systems. In industrial settings, there is usually an embedding team that trains an embedding model to solve intended tasks (e.g., product recommendation). The produced embeddings are then widely consumed by consumer teams to solve their unintended tasks (e.g., fraud detection). However, as the embedding model gets updated and retrained to improve performance on the intended task, the newly-generated embeddings are no longer compatible with the existing consumer models. This means that historical versions of the embeddings can never be retired or all consumer teams have to retrain their models to make them compatible with the latest version of the embeddings, both of which are extremely costly in practice. Here we study the problem of embedding version updates and their backward compatibility. We formalize the problem where the goal is for the embedding team to keep updating the embedding version, while the consumer teams do not have to retrain their models. We develop a solution based on learning backward compatible embeddings, which allows the embedding model version to be updated frequently, while also allowing the latest version of the embedding to be quickly transformed into any backward compatible historical version of it, so that consumer teams do not have to retrain their models. Under our framework, we explore six methods and systematically evaluate them on a real-world recommender system application. We show that the best method, which we call BC-Aligner, maintains backward compatibility with existing unintended tasks even after multiple model version updates. Simultaneously, BC-Aligner achieves the intended task performance similar to the embedding model that is solely optimized for the intended task."
                    },
                    {
                        "title": "An Interpretable Ensemble of Graph and Language Models for Improving Search Relevance in E-Commerce",
                        "abstract": "The problem of search relevance in the E-commerce domain is a challenging one since it involves understanding the intent of a user's short nuanced query and matching it with the appropriate products in the catalog. This problem has traditionally been addressed using language models (LMs) and graph neural networks (GNNs) to capture semantic and inter-product behavior signals, respectively. However, the rapid development of new architectures has created a gap between research and the practical adoption of these techniques. Evaluating the generalizability of these models for deployment requires extensive experimentation on complex, real-world datasets, which can be non-trivial and expensive. Furthermore, such models often operate on latent space representations that are incomprehensible to humans, making it difficult to evaluate and compare the effectiveness of different models. This lack of interpretability hinders the development and adoption of new techniques in the field. To bridge this gap, we propose Plug and Play Graph LAnguage Model (PP-GLAM), an explainable ensemble of plug and play models. Our approach uses a modular framework with uniform data processing pipelines. It employs additive explanation metrics to independently decide whether to include (i) language model candidates, (ii) GNN model candidates, and (iii) inter-product behavioral signals. For the task of search relevance, we show that PP-GLAM outperforms several state-of-the-art baselines as well as a proprietary model on real-world multilingual, multi-regional e-commerce datasets. To promote better model comprehensibility and adoption, we also provide an analysis of the explainability and computational complexity of our model. We also provide the public codebase and provide a deployment strategy for practical implementation."
                    },
                    {
                        "title": "Covering a Graph with Dense Subgraph Families, via Triangle-Rich Sets",
                        "abstract": "Graphs are a fundamental data structure used to represent relationships in domains as diverse as the social sciences, bioinformatics, cybersecurity, the Internet, and more. One of the central observations in network science is that real-world graphs are globally sparse, yet contains numerous \"pockets\" of high edge density. A fundamental task in graph mining is to discover these dense subgraphs. Most common formulations of the problem involve finding a single (or a few) \"optimally\" dense subsets. But in most real applications, one does not care for the optimality. Instead, we want to find a large collection of dense subsets that covers a significant fraction of the input graph.   We give a mathematical formulation of this problem, using a new definition of regularly triangle-rich (RTR) families. These families capture the notion of dense subgraphs that contain many triangles and have degrees comparable to the subgraph size. We design a provable algorithm, RTRExtractor, that can discover RTR families that approximately cover any RTR set. The algorithm is efficient and is inspired by recent results that use triangle counts for community testing and clustering.   We show that RTRExtractor has excellent behavior on a large variety of real-world datasets. It is able to process graphs with hundreds of millions of edges within minutes. Across many datasets, RTRExtractor achieves high coverage using high edge density datasets. For example, the output covers a quarter of the vertices with subgraphs of edge density more than (say) $0.5$, for datasets with 10M+ edges. We show an example of how the output of RTRExtractor correlates with meaningful sets of similar vertices in a citation network, demonstrating the utility of RTRExtractor for unsupervised graph discovery tasks."
                    },
                    {
                        "title": "Multilingual Knowledge Graph Completion with Self-Supervised Adaptive Graph Alignment",
                        "abstract": "Predicting missing facts in a knowledge graph (KG) is crucial as modern KGs are far from complete. Due to labor-intensive human labeling, this phenomenon deteriorates when handling knowledge represented in various languages. In this paper, we explore multilingual KG completion, which leverages limited seed alignment as a bridge, to embrace the collective knowledge from multiple languages. However, language alignment used in prior works is still not fully exploited: (1) alignment pairs are treated equally to maximally push parallel entities to be close, which ignores KG capacity inconsistency; (2) seed alignment is scarce and new alignment identification is usually in a noisily unsupervised manner. To tackle these issues, we propose a novel self-supervised adaptive graph alignment (SS-AGA) method. Specifically, SS-AGA fuses all KGs as a whole graph by regarding alignment as a new edge type. As such, information propagation and noise influence across KGs can be adaptively controlled via relation-aware attention weights. Meanwhile, SS-AGA features a new pair generator that dynamically captures potential alignment pairs in a self-supervised paradigm. Extensive experiments on both the public multilingual DBPedia KG and newly-created industrial multilingual E-commerce KG empirically demonstrate the effectiveness of SS-AG"
                    }
                ]
            },
            "d8c40b7e-5dc2-41aa-a02c-2380fda9b109": {
                "pk": "d8c40b7e-5dc2-41aa-a02c-2380fda9b109",
                "name": "James Zou",
                "collaborators": [
                    "Abubakar Abid",
                    "Amirata Ghorbani",
                    "Rong Ge",
                    "Yongchan Kwon",
                    "James Wexler",
                    "Been Kim",
                    "Daniel Russo",
                    "Ruishan Liu",
                    "Bryan He",
                    "S. Matthew Weinberg"
                ],
                "domain": [
                    "Interpretability",
                    "Machine Learning",
                    "Data Valuation",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Towards Automatic Concept-based Explanations",
                        "abstract": "Interpretability has become an important topic of research as more machine learning (ML) models are deployed and widely used to make important decisions.   Most of the current explanation methods provide explanations through feature importance scores, which identify features that are important for each individual input. However, how to systematically summarize and interpret such per sample feature importance scores itself is challenging. In this work, we propose principles and desiderata for \\emph{concept} based explanation, which goes beyond per-sample features to identify higher-level human-understandable concepts that apply across the entire dataset. We develop a new algorithm, ACE, to automatically extract visual concepts. Our systematic experiments demonstrate that \\alg discovers concepts that are human-meaningful, coherent and important for the neural network's predictions."
                    },
                    {
                        "title": "How much does your data exploration overfit? Controlling bias via information usage",
                        "abstract": "Modern data is messy and high-dimensional, and it is often not clear a priori what are the right questions to ask. Instead, the analyst typically needs to use the data to search for interesting analyses to perform and hypotheses to test. This is an adaptive process, where the choice of analysis to be performed next depends on the results of the previous analyses on the same data. Ultimately, which results are reported can be heavily influenced by the data. It is widely recognized that this process, even if well-intentioned, can lead to biases and false discoveries, contributing to the crisis of reproducibility in science. But while %the adaptive nature of exploration any data-exploration renders standard statistical theory invalid, experience suggests that different types of exploratory analysis can lead to disparate levels of bias, and the degree of bias also depends on the particulars of the data set. In this paper, we propose a general information usage framework to quantify and provably bound the bias and other error metrics of an arbitrary exploratory analysis. We prove that our mutual information based bound is tight in natural settings, and then use it to give rigorous insights into when commonly used procedures do or do not lead to substantially biased estimation. Through the lens of information usage, we analyze the bias of specific exploration procedures such as filtering, rank selection and clustering. Our general framework also naturally motivates randomization techniques that provably reduces exploration bias while preserving the utility of the data analysis. We discuss the connections between our approach and related ideas from differential privacy and blinded data analysis, and supplement our results with illustrative simulations."
                    },
                    {
                        "title": "Intersecting Faces: Non-negative Matrix Factorization With New Guarantees",
                        "abstract": "Non-negative matrix factorization (NMF) is a natural model of admixture and is widely used in science and engineering. A plethora of algorithms have been developed to tackle NMF, but due to the non-convex nature of the problem, there is little guarantee on how well these methods work. Recently a surge of research have focused on a very restricted class of NMFs, called separable NMF, where provably correct algorithms have been developed. In this paper, we propose the notion of subset-separable NMF, which substantially generalizes the property of separability. We show that subset-separability is a natural necessary condition for the factorization to be unique or to have minimum volume. We developed the Face-Intersect algorithm which provably and efficiently solves subset-separable NMF under natural conditions, and we prove that our algorithm is robust to small noise. We explored the performance of Face-Intersect on simulations and discuss settings where it empirically outperformed the state-of-art methods. Our work is a step towards finding provably correct algorithms that solve large classes of NMF problems."
                    },
                    {
                        "title": "Rich Component Analysis",
                        "abstract": "In many settings, we have multiple data sets (also called views) that capture different and overlapping aspects of the same phenomenon. We are often interested in finding patterns that are unique to one or to a subset of the views. For example, we might have one set of molecular observations and one set of physiological observations on the same group of individuals, and we want to quantify molecular patterns that are uncorrelated with physiology. Despite being a common problem, this is highly challenging when the correlations come from complex distributions. In this paper, we develop the general framework of Rich Component Analysis (RCA) to model settings where the observations from different views are driven by different sets of latent components, and each component can be a complex, high-dimensional distribution. We introduce algorithms based on cumulant extraction that provably learn each of the components without having to model the other components. We show how to integrate RCA with stochastic gradient descent into a meta-algorithm for learning general models, and demonstrate substantial improvement in accuracy on several synthetic and real datasets in both supervised and unsupervised tasks. Our method makes it possible to learn latent variable models when we don't have samples from the true model but only samples after complex perturbations."
                    },
                    {
                        "title": "The Effects of Memory Replay in Reinforcement Learning",
                        "abstract": "Experience replay is a key technique behind many recent advances in deep reinforcement learning. Allowing the agent to learn from earlier memories can speed up learning and break undesirable temporal correlations. Despite its wide-spread application, very little is understood about the properties of experience replay. How does the amount of memory kept affect learning dynamics? Does it help to prioritize certain experiences? In this paper, we address these questions by formulating a dynamical systems ODE model of Q-learning with experience replay. We derive analytic solutions of the ODE for a simple setting. We show that even in this very simple setting, the amount of memory kept can substantially affect the agent's performance. Too much or too little memory both slow down learning. Moreover, we characterize regimes where prioritized replay harms the agent's learning. We show that our analytic solutions have excellent agreement with experiments. Finally, we propose a simple algorithm for adaptively changing the memory buffer size which achieves consistently good empirical performance."
                    },
                    {
                        "title": "Autowarp: Learning a Warping Distance from Unlabeled Time Series Using Sequence Autoencoders",
                        "abstract": "Measuring similarities between unlabeled time series trajectories is an important problem in domains as diverse as medicine, astronomy, finance, and computer vision. It is often unclear what is the appropriate metric to use because of the complex nature of noise in the trajectories (e.g. different sampling rates or outliers). Domain experts typically hand-craft or manually select a specific metric, such as dynamic time warping (DTW), to apply on their data. In this paper, we propose Autowarp, an end-to-end algorithm that optimizes and learns a good metric given unlabeled trajectories. We define a flexible and differentiable family of warping metrics, which encompasses common metrics such as DTW, Euclidean, and edit distance. Autowarp then leverages the representation power of sequence autoencoders to optimize for a member of this warping distance family. The output is a metric which is easy to interpret and can be robustly learned from relatively few trajectories. In systematic experiments across different domains, we show that Autowarp often outperforms hand-crafted trajectory similarity metrics."
                    },
                    {
                        "title": "Minimizing Close-k Aggregate Loss Improves Classification",
                        "abstract": "In classification, the de facto method for aggregating individual losses is the average loss. When the actual metric of interest is 0-1 loss, it is common to minimize the average surrogate loss for some well-behaved (e.g. convex) surrogate. Recently, several other aggregate losses such as the maximal loss and average top-$k$ loss were proposed as alternative objectives to address shortcomings of the average loss. However, we identify common classification settings, e.g. the data is imbalanced, has too many easy or ambiguous examples, etc., when average, maximal and average top-$k$ all suffer from suboptimal decision boundaries, even on an infinitely large training set. To address this problem, we propose a new classification objective called the close-$k$ aggregate loss, where we adaptively minimize the loss for points close to the decision boundary. We provide theoretical guarantees for the 0-1 accuracy when we optimize close-$k$ aggregate loss. We also conduct systematic experiments across the PMLB and OpenML benchmark datasets. Close-$k$ achieves significant gains in 0-1 test accuracy, improvements of $\\geq 2$% and $p<0.05$, in over 25% of the datasets compared to average, maximal and average top-$k$. In contrast, the previous aggregate losses outperformed close-$k$ in less than 2% of the datasets."
                    },
                    {
                        "title": "Neuron Shapley: Discovering the Responsible Neurons",
                        "abstract": "We develop Neuron Shapley as a new framework to quantify the contribution of individual neurons to the prediction and performance of a deep network. By accounting for interactions across neurons, Neuron Shapley is more effective in identifying important filters compared to common approaches based on activation patterns. Interestingly, removing just 30 filters with the highest Shapley scores effectively destroys the prediction accuracy of Inception-v3 on ImageNet. Visualization of these few critical filters provides insights into how the network functions. Neuron Shapley is a flexible framework and can be applied to identify responsible neurons in many tasks. We illustrate additional applications of identifying filters that are responsible for biased prediction in facial recognition and filters that are vulnerable to adversarial attacks. Removing these filters is a quick way to repair models. Enabling all these applications is a new multi-arm bandit algorithm that we developed to efficiently estimate Neuron Shapley values."
                    },
                    {
                        "title": "Signal to noise in matching markets",
                        "abstract": "In many matching markets, one side \"applies\" to the other, and these applications are often expensive and time-consuming (e.g. students applying to college). It is tempting to think that making the application process easier should benefit both sides of the market. After all, the applicants could submit more applications, and the recipients would have more applicants to choose from. In this paper, we propose and analyze a simple model to understand settings where both sides of the market suffer from increased number of applications.   The main insights of the paper are derived from quantifying the signal to noise tradeoffs in random matchings, as applications provide a signal of the applicants' preferences. When applications are costly the signal is stronger, as the act of making an application itself is meaningful. Therefore more applications may yield potentially better matches, but fewer applications create stronger signals for the receiving side to learn true preferences.   We derive analytic characterizations of the expected quality of stable matchings in a simple utility model where applicants have an overall quality, but also synergy with specific possible partners. Our results show how reducing application cost without introducing an effective signaling mechanism might lead to inefficiencies for both sides of the market."
                    },
                    {
                        "title": "Stochastic EM for Shuffled Linear Regression",
                        "abstract": "We consider the problem of inference in a linear regression model in which the relative ordering of the input features and output labels is not known. Such datasets naturally arise from experiments in which the samples are shuffled or permuted during the protocol. In this work, we propose a framework that treats the unknown permutation as a latent variable. We maximize the likelihood of observations using a stochastic expectation-maximization (EM) approach. We compare this to the dominant approach in the literature, which corresponds to hard EM in our framework. We show on synthetic data that the stochastic EM algorithm we develop has several advantages, including lower parameter error, less sensitivity to the choice of initialization, and significantly better performance on datasets that are only partially shuffled. We conclude by performing two experiments on real datasets that have been partially shuffled, in which we show that the stochastic EM algorithm can recover the weights with modest error."
                    },
                    {
                        "title": "Feedback GAN (FBGAN) for DNA: a Novel Feedback-Loop Architecture for Optimizing Protein Functions",
                        "abstract": "Generative Adversarial Networks (GANs) represent an attractive and novel approach to generate realistic data, such as genes, proteins, or drugs, in synthetic biology. Here, we apply GANs to generate synthetic DNA sequences encoding for proteins of variable length. We propose a novel feedback-loop architecture, called Feedback GAN (FBGAN), to optimize the synthetic gene sequences for desired properties using an external function analyzer. The proposed architecture also has the advantage that the analyzer need not be differentiable. We apply the feedback-loop mechanism to two examples: 1) generating synthetic genes coding for antimicrobial peptides, and 2) optimizing synthetic genes for the secondary structure of their resulting peptides. A suite of metrics demonstrate that the GAN generated proteins have desirable biophysical properties. The FBGAN architecture can also be used to optimize GAN-generated datapoints for useful properties in domains beyond genomics."
                    },
                    {
                        "title": "Improving Training on Noisy Stuctured Labels",
                        "abstract": "Fine-grained annotations---e.g. dense image labels, image segmentation and text tagging---are useful in many ML applications but they are labor-intensive to generate. Moreover there are often systematic, structured errors in these fine-grained annotations. For example, a car might be entirely unannotated in the image, or the boundary between a car and street might only be coarsely annotated. Standard ML training on data with such structured errors produces models with biases and poor performance. In this work, we propose a novel framework of Error-Correcting Networks (ECN) to address the challenge of learning in the presence structured error in fine-grained annotations. Given a large noisy dataset with commonly occurring structured errors, and a much smaller dataset with more accurate annotations, ECN is able to substantially improve the prediction of fine-grained annotations compared to standard approaches for training on noisy data. It does so by learning to leverage the structures in the annotations and in the noisy labels. Systematic experiments on image segmentation and text tagging demonstrate the strong performance of ECN in improving training on noisy structured labels."
                    },
                    {
                        "title": "Neural Group Testing to Accelerate Deep Learning",
                        "abstract": "Recent advances in deep learning have made the use of large, deep neural networks with tens of millions of parameters. The sheer size of these networks imposes a challenging computational burden during inference. Existing work focuses primarily on accelerating each forward pass of a neural network. Inspired by the group testing strategy for efficient disease testing, we propose neural group testing, which accelerates by testing a group of samples in one forward pass. Groups of samples that test negative are ruled out. If a group tests positive, samples in that group are then retested adaptively. A key challenge of neural group testing is to modify a deep neural network so that it could test multiple samples in one forward pass. We propose three designs to achieve this without introducing any new parameters and evaluate their performances. We applied neural group testing in an image moderation task to detect rare but inappropriate images. We found that neural group testing can group up to 16 images in one forward pass and reduce the overall computation cost by over 73% while improving detection performance."
                    },
                    {
                        "title": "Beta Shapley: a Unified and Noise-reduced Data Valuation Framework for Machine Learning",
                        "abstract": "Data Shapley has recently been proposed as a principled framework to quantify the contribution of individual datum in machine learning. It can effectively identify helpful or harmful data points for a learning algorithm. In this paper, we propose Beta Shapley, which is a substantial generalization of Data Shapley. Beta Shapley arises naturally by relaxing the efficiency axiom of the Shapley value, which is not critical for machine learning settings. Beta Shapley unifies several popular data valuation methods and includes data Shapley as a special case. Moreover, we prove that Beta Shapley has several desirable statistical properties and propose efficient algorithms to estimate it. We demonstrate that Beta Shapley outperforms state-of-the-art data valuation methods on several downstream ML tasks such as: 1) detecting mislabeled training data; 2) learning with subsamples; and 3) identifying points whose addition or removal have the largest positive or negative impact on the model."
                    },
                    {
                        "title": "WeightedSHAP: analyzing and improving Shapley based feature attributions",
                        "abstract": "Shapley value is a popular approach for measuring the influence of individual features. While Shapley feature attribution is built upon desiderata from game theory, some of its constraints may be less natural in certain machine learning settings, leading to unintuitive model interpretation. In particular, the Shapley value uses the same weight for all marginal contributions -- i.e. it gives the same importance when a large number of other features are given versus when a small number of other features are given. This property can be problematic if larger feature sets are more or less informative than smaller feature sets. Our work performs a rigorous analysis of the potential limitations of Shapley feature attribution. We identify simple settings where the Shapley value is mathematically suboptimal by assigning larger attributions for less influential features. Motivated by this observation, we propose WeightedSHAP, which generalizes the Shapley value and learns which marginal contributions to focus directly from data. On several real-world datasets, we demonstrate that the influential features identified by WeightedSHAP are better able to recapitulate the model's predictions compared to the features identified by the Shapley value."
                    },
                    {
                        "title": "New Evaluation Metrics Capture Quality Degradation due to LLM Watermarking",
                        "abstract": "With the increasing use of large-language models (LLMs) like ChatGPT, watermarking has emerged as a promising approach for tracing machine-generated content. However, research on LLM watermarking often relies on simple perplexity or diversity-based measures to assess the quality of watermarked text, which can mask important limitations in watermarking. Here we introduce two new easy-to-use methods for evaluating watermarking algorithms for LLMs: 1) evaluation by LLM-judger with specific guidelines; and 2) binary classification on text embeddings to distinguish between watermarked and unwatermarked text. We apply these methods to characterize the effectiveness of current watermarking techniques. Our experiments, conducted across various datasets, reveal that current watermarking methods are detectable by even simple classifiers, challenging the notion of watermarking subtlety. We also found, through the LLM judger, that watermarking impacts text quality, especially in degrading the coherence and depth of the response. Our findings underscore the trade-off between watermark robustness and text quality and highlight the importance of having more informative metrics to assess watermarking quality."
                    },
                    {
                        "title": "Large Language Models are Vulnerable to Bait-and-Switch Attacks for Generating Harmful Content",
                        "abstract": "The risks derived from large language models (LLMs) generating deceptive and damaging content have been the subject of considerable research, but even safe generations can lead to problematic downstream impacts. In our study, we shift the focus to how even safe text coming from LLMs can be easily turned into potentially dangerous content through Bait-and-Switch attacks. In such attacks, the user first prompts LLMs with safe questions and then employs a simple find-and-replace post-hoc technique to manipulate the outputs into harmful narratives. The alarming efficacy of this approach in generating toxic content highlights a significant challenge in developing reliable safety guardrails for LLMs. In particular, we stress that focusing on the safety of the verbatim LLM outputs is insufficient and that we also need to consider post-hoc transformations."
                    },
                    {
                        "title": "Talking Nonsense: Probing Large Language Models' Understanding of Adversarial Gibberish Inputs",
                        "abstract": "Large language models (LLMs) exhibit excellent ability to understand human languages, but do they also understand their own language that appears gibberish to us? In this work we delve into this question, aiming to uncover the mechanisms underlying such behavior in LLMs. We employ the Greedy Coordinate Gradient optimizer to craft prompts that compel LLMs to generate coherent responses from seemingly nonsensical inputs. We call these inputs LM Babel and this work systematically studies the behavior of LLMs manipulated by these prompts. We find that the manipulation efficiency depends on the target text's length and perplexity, with the Babel prompts often located in lower loss minima compared to natural prompts. We further examine the structure of the Babel prompts and evaluate their robustness. Notably, we find that guiding the model to generate harmful texts is not more difficult than into generating benign texts, suggesting lack of alignment for out-of-distribution prompts."
                    },
                    {
                        "title": "Contrastive Variational Autoencoder Enhances Salient Features",
                        "abstract": "Variational autoencoders are powerful algorithms for identifying dominant latent structure in a single dataset. In many applications, however, we are interested in modeling latent structure and variation that are enriched in a target dataset compared to some background---e.g. enriched in patients compared to the general population. Contrastive learning is a principled framework to capture such enriched variation between the target and background, but state-of-the-art contrastive methods are limited to linear models. In this paper, we introduce the contrastive variational autoencoder (cVAE), which combines the benefits of contrastive learning with the power of deep generative models. The cVAE is designed to identify and enhance salient latent features. The cVAE is trained on two related but unpaired datasets, one of which has minimal contribution from the salient latent features. The cVAE explicitly models latent features that are shared between the datasets, as well as those that are enriched in one dataset relative to the other, which allows the algorithm to isolate and enhance the salient latent features. The algorithm is straightforward to implement, has a similar run-time to the standard VAE, and is robust to noise and dataset purity. We conduct experiments across diverse types of data, including gene expression and facial images, showing that the cVAE effectively uncovers latent structure that is salient in a particular analysis."
                    },
                    {
                        "title": "Data Shapley: Equitable Valuation of Data for Machine Learning",
                        "abstract": "As data becomes the fuel driving technological and economic growth, a fundamental challenge is how to quantify the value of data in algorithmic predictions and decisions. For example, in healthcare and consumer markets, it has been suggested that individuals should be compensated for the data that they generate, but it is not clear what is an equitable valuation for individual data. In this work, we develop a principled framework to address data valuation in the context of supervised machine learning. Given a learning algorithm trained on $n$ data points to produce a predictor, we propose data Shapley as a metric to quantify the value of each training datum to the predictor performance. Data Shapley value uniquely satisfies several natural properties of equitable data valuation. We develop Monte Carlo and gradient-based methods to efficiently estimate data Shapley values in practical settings where complex learning algorithms, including neural networks, are trained on large datasets. In addition to being equitable, extensive experiments across biomedical, image and synthetic data demonstrate that data Shapley has several other benefits: 1) it is more powerful than the popular leave-one-out or leverage score in providing insight on what data is more valuable for a given learning task; 2) low Shapley value data effectively capture outliers and corruptions; 3) high Shapley value data inform what type of new data to acquire to improve the predictor."
                    }
                ]
            },
            "d70c06c3-349c-4574-9d97-d8405893ed3e": {
                "pk": "d70c06c3-349c-4574-9d97-d8405893ed3e",
                "name": "Jure Leskovec",
                "collaborators": [
                    "Myunghwan Kim",
                    "Jon Kleinberg",
                    "Julian McAuley",
                    "Daniel Huttenlocher",
                    "Jaewon Yang",
                    "Eric Horvitz",
                    "Tim Althoff",
                    "Hongwei Wang",
                    "Rok Sosic",
                    "Aditya Grover"
                ],
                "domain": [
                    "Social Network Analysis",
                    "Graph Theory",
                    "Machine Learning",
                    "Data Mining"
                ],
                "publications": [
                    {
                        "title": "Planetary-Scale Views on an Instant-Messaging Network",
                        "abstract": "We present a study of anonymized data capturing a month of high-level communication activities within the whole of the Microsoft Messenger instant-messaging system. We examine characteristics and patterns that emerge from the collective dynamics of large numbers of people, rather than the actions and characteristics of individuals. The dataset contains summary properties of 30 billion conversations among 240 million people. From the data, we construct a communication graph with 180 million nodes and 1.3 billion undirected edges, creating the largest social network constructed and analyzed to date. We report on multiple aspects of the dataset and synthesized graph. We find that the graph is well-connected and robust to node removal. We investigate on a planetary-scale the oft-cited report that people are separated by ``six degrees of separation'' and find that the average path length among Messenger users is 6.6. We also find that people tend to communicate more with each other when they have similar age, language, and location, and that cross-gender conversations are both more frequent and of longer duration than conversations with the same gender."
                    },
                    {
                        "title": "Modeling Social Networks with Node Attributes using the Multiplicative Attribute Graph Model",
                        "abstract": "Networks arising from social, technological and natural domains exhibit rich connectivity patterns and nodes in such networks are often labeled with attributes or features. We address the question of modeling the structure of networks where nodes have attribute information. We present a Multiplicative Attribute Graph (MAG) model that considers nodes with categorical attributes and models the probability of an edge as the product of individual attribute link formation affinities. We develop a scalable variational expectation maximization parameter estimation method. Experiments show that MAG model reliably captures network connectivity as well as provides insights into how different attributes shape the network structure."
                    },
                    {
                        "title": "Latent Multi-group Membership Graph Model",
                        "abstract": "We develop the Latent Multi-group Membership Graph (LMMG) model, a model of networks with rich node feature structure. In the LMMG model, each node belongs to multiple groups and each latent group models the occurrence of links as well as the node feature structure. The LMMG can be used to summarize the network structure, to predict links between the nodes, and to predict missing features of a node. We derive efficient inference and learning algorithms and evaluate the predictive performance of the LMMG on several social and document network datasets."
                    },
                    {
                        "title": "Multiplicative Attribute Graph Model of Real-World Networks",
                        "abstract": "Large scale real-world network data such as social and information networks are ubiquitous. The study of such social and information networks seeks to find patterns and explain their emergence through tractable models. In most networks, and especially in social networks, nodes have a rich set of attributes (e.g., age, gender) associated with them.   Here we present a model that we refer to as the Multiplicative Attribute Graphs (MAG), which naturally captures the interactions between the network structure and the node attributes. We consider a model where each node has a vector of categorical latent attributes associated with it. The probability of an edge between a pair of nodes then depends on the product of individual attribute-attribute affinities. The model yields itself to mathematical analysis and we derive thresholds for the connectivity and the emergence of the giant connected component, and show that the model gives rise to networks with a constant diameter. We analyze the degree distribution to show that MAG model can produce networks with either log-normal or power-law degree distributions depending on certain conditions."
                    },
                    {
                        "title": "Image Labeling on a Network: Using Social-Network Metadata for Image Classification",
                        "abstract": "Large-scale image retrieval benchmarks invariably consist of images from the Web. Many of these benchmarks are derived from online photo sharing networks, like Flickr, which in addition to hosting images also provide a highly interactive social community. Such communities generate rich metadata that can naturally be harnessed for image classification and retrieval. Here we study four popular benchmark datasets, extending them with social-network metadata, such as the groups to which each image belongs, the comment thread associated with the image, who uploaded it, their location, and their network of friends. Since these types of data are inherently relational, we propose a model that explicitly accounts for the interdependencies between images sharing common properties. We model the task as a binary labeling problem on a network, and use structured learning techniques to learn model parameters. We find that social-network metadata are useful in a variety of classification tasks, in many cases outperforming methods based on image content."
                    },
                    {
                        "title": "Discovering Social Circles in Ego Networks",
                        "abstract": "People's personal social networks are big and cluttered, and currently there is no good way to automatically organize them. Social networking sites allow users to manually categorize their friends into social circles (e.g. 'circles' on Google+, and 'lists' on Facebook and Twitter), however they are laborious to construct and must be updated whenever a user's network grows. In this paper, we study the novel task of automatically identifying users' social circles. We pose this task as a multi-membership node clustering problem on a user's ego-network, a network of connections between her friends. We develop a model for detecting circles that combines network structure as well as user profile information. For each circle we learn its members and the circle-specific user profile similarity metric. Modeling node membership to multiple circles allows us to detect overlapping as well as hierarchically nested circles. Experiments show that our model accurately identifies circles on a diverse set of data from Facebook, Google+, and Twitter, for all of which we obtain hand-labeled ground-truth."
                    },
                    {
                        "title": "Nonparametric Multi-group Membership Model for Dynamic Networks",
                        "abstract": "Relational data-like graphs, networks, and matrices-is often dynamic, where the relational structure evolves over time. A fundamental problem in the analysis of time-varying network data is to extract a summary of the common structure and the dynamics of the underlying relations between the entities. Here we build on the intuition that changes in the network structure are driven by the dynamics at the level of groups of nodes. We propose a nonparametric multi-group membership model for dynamic networks. Our model contains three main components: We model the birth and death of individual groups with respect to the dynamics of the network structure via a distance dependent Indian Buffet Process. We capture the evolution of individual node group memberships via a Factorial Hidden Markov model. And, we explain the dynamics of the network structure by explicitly modeling the connectivity structure of groups. We demonstrate our model's capability of identifying the dynamics of latent groups in a number of different types of network data. Experimental results show that our model provides improved predictive performance over existing dynamic network models on future network forecasting and missing link prediction."
                    },
                    {
                        "title": "Donor Retention in Online Crowdfunding Communities: A Case Study of DonorsChoose.org",
                        "abstract": "Online crowdfunding platforms like DonorsChoose.org and Kickstarter allow specific projects to get funded by targeted contributions from a large number of people. Critical for the success of crowdfunding communities is recruitment and continued engagement of donors. With donor attrition rates above 70%, a significant challenge for online crowdfunding platforms as well as traditional offline non-profit organizations is the problem of donor retention.   We present a large-scale study of millions of donors and donations on DonorsChoose.org, a crowdfunding platform for education projects. Studying an online crowdfunding platform allows for an unprecedented detailed view of how people direct their donations. We explore various factors impacting donor retention which allows us to identify different groups of donors and quantify their propensity to return for subsequent donations. We find that donors are more likely to return if they had a positive interaction with the receiver of the donation. We also show that this includes appropriate and timely recognition of their support as well as detailed communication of their impact. Finally, we discuss how our findings could inform steps to improve donor retention in crowdfunding communities and non-profit organizations."
                    },
                    {
                        "title": "Unifying Graph Convolutional Neural Networks and Label Propagation",
                        "abstract": "Label Propagation (LPA) and Graph Convolutional Neural Networks (GCN) are both message passing algorithms on graphs. Both solve the task of node classification but LPA propagates node label information across the edges of the graph, while GCN propagates and transforms node feature information. However, while conceptually similar, theoretical relation between LPA and GCN has not yet been investigated. Here we study the relationship between LPA and GCN in terms of two aspects: (1) feature/label smoothing where we analyze how the feature/label of one node is spread over its neighbors; And, (2) feature/label influence of how much the initial feature/label of one node influences the final feature/label of another node. Based on our theoretical analysis, we propose an end-to-end model that unifies GCN and LPA for node classification. In our unified model, edge weights are learnable, and the LPA serves as regularization to assist the GCN in learning proper edge weights that lead to improved classification performance. Our model can also be seen as learning attention weights based on node labels, which is more task-oriented than existing feature-based attention models. In a number of experiments on real-world graphs, our model shows superiority over state-of-the-art GCN-based methods in terms of node classification accuracy."
                    },
                    {
                        "title": "From Amateurs to Connoisseurs: Modeling the Evolution of User Expertise through Online Reviews",
                        "abstract": "Recommending products to consumers means not only understanding their tastes, but also understanding their level of experience. For example, it would be a mistake to recommend the iconic film Seven Samurai simply because a user enjoys other action movies; rather, we might conclude that they will eventually enjoy it -- once they are ready. The same is true for beers, wines, gourmet foods -- or any products where users have acquired tastes: the `best' products may not be the most `accessible'. Thus our goal in this paper is to recommend products that a user will enjoy now, while acknowledging that their tastes may have changed over time, and may change again in the future. We model how tastes change due to the very act of consuming more products -- in other words, as users become more experienced. We develop a latent factor recommendation system that explicitly accounts for each user's level of experience. We find that such a model not only leads to better recommendations, but also allows us to study the role of user experience and expertise on a novel dataset of fifteen million beer, wine, food, and movie reviews."
                    },
                    {
                        "title": "Structure and Overlaps of Communities in Networks",
                        "abstract": "One of the main organizing principles in real-world social, information and technological networks is that of network communities, where sets of nodes organize into densely linked clusters. Even though detection of such communities is of great interest, understanding the structure communities in large networks remains relatively limited. Due to unavailability of labeled ground-truth data it is practically impossible to evaluate and compare different models and notions of communities on a large scale.   In this paper we identify 6 large social, collaboration, and information networks where nodes explicitly state their community memberships. We define ground-truth communities by using these explicit memberships. We then empirically study how such ground-truth communities emerge in networks and how they overlap. We observe some surprising phenomena. First, ground-truth communities contain high-degree hub nodes that reside in community overlaps and link to most of the members of the community. Second, the overlaps of communities are more densely connected than the non-overlapping parts of communities, in contrast to the conventional wisdom that community overlaps are more sparsely connected than the communities themselves.   Existing models of network communities do not capture dense community overlaps. We present the Community-Affiliation Graph Model (AGM), a conceptual model of network community structure, which reliably captures the overall structure of networks as well as the overlapping nature of network communities."
                    },
                    {
                        "title": "Defining and Evaluating Network Communities based on Ground-truth",
                        "abstract": "Nodes in real-world networks organize into densely linked communities where edges appear with high concentration among the members of the community. Identifying such communities of nodes has proven to be a challenging task mainly due to a plethora of definitions of a community, intractability of algorithms, issues with evaluation and the lack of a reliable gold-standard ground-truth.   In this paper we study a set of 230 large real-world social, collaboration and information networks where nodes explicitly state their group memberships. For example, in social networks nodes explicitly join various interest based social groups. We use such groups to define a reliable and robust notion of ground-truth communities. We then propose a methodology which allows us to compare and quantitatively evaluate how different structural definitions of network communities correspond to ground-truth communities. We choose 13 commonly used structural definitions of network communities and examine their sensitivity, robustness and performance in identifying the ground-truth. We show that the 13 structural definitions are heavily correlated and naturally group into four classes. We find that two of these definitions, Conductance and Triad-participation-ratio, consistently give the best performance in identifying ground-truth communities. We also investigate a task of detecting communities given a single seed node. We extend the local spectral clustering algorithm into a heuristic parameter-free community detection method that easily scales to networks with more than hundred million nodes. The proposed method achieves 30% relative improvement over current local clustering methods."
                    },
                    {
                        "title": "SNAP: A General Purpose Network Analysis and Graph Mining Library",
                        "abstract": "Large networks are becoming a widely used abstraction for studying complex systems in a broad set of disciplines, ranging from social network analysis to molecular biology and neuroscience. Despite an increasing need to analyze and manipulate large networks, only a limited number of tools are available for this task.   Here, we describe Stanford Network Analysis Platform (SNAP), a general-purpose, high-performance system that provides easy to use, high-level operations for analysis and manipulation of large networks. We present SNAP functionality, describe its implementational details, and give performance benchmarks. SNAP has been developed for single big-memory machines and it balances the trade-off between maximum performance, compact in-memory graph representation, and the ability to handle dynamic graphs where nodes and edges are being added or removed over time. SNAP can process massive networks with hundreds of millions of nodes and billions of edges. SNAP offers over 140 different graph algorithms that can efficiently manipulate large graphs, calculate structural properties, generate regular and random graphs, and handle attributes and meta-data on nodes and edges. Besides being able to handle large graphs, an additional strength of SNAP is that networks and their attributes are fully dynamic, they can be modified during the computation at low cost. SNAP is provided as an open source library in C++ as well as a module in Python.   We also describe the Stanford Large Network Dataset, a set of social and information real-world networks and datasets, which we make publicly available. The collection is a complementary resource to our SNAP software and is widely used for development and benchmarking of graph analytics algorithms."
                    },
                    {
                        "title": "node2vec: Scalable Feature Learning for Networks",
                        "abstract": "Prediction tasks over nodes and edges in networks require careful effort in engineering features used by learning algorithms. Recent research in the broader field of representation learning has led to significant progress in automating prediction by learning the features themselves. However, present feature learning approaches are not expressive enough to capture the diversity of connectivity patterns observed in networks. Here we propose node2vec, an algorithmic framework for learning continuous feature representations for nodes in networks. In node2vec, we learn a mapping of nodes to a low-dimensional space of features that maximizes the likelihood of preserving network neighborhoods of nodes. We define a flexible notion of a node's network neighborhood and design a biased random walk procedure, which efficiently explores diverse neighborhoods. Our algorithm generalizes prior work which is based on rigid notions of network neighborhoods, and we argue that the added flexibility in exploring neighborhoods is the key to learning richer representations. We demonstrate the efficacy of node2vec over existing state-of-the-art techniques on multi-label classification and link prediction in several real-world networks from diverse domains. Taken together, our work represents a new way for efficiently learning state-of-the-art task-independent representations in complex networks."
                    },
                    {
                        "title": "Predicting multicellular function through multi-layer tissue networks",
                        "abstract": "Motivation: Understanding functions of proteins in specific human tissues is essential for insights into disease diagnostics and therapeutics, yet prediction of tissue-specific cellular function remains a critical challenge for biomedicine.   Results: Here we present OhmNet, a hierarchy-aware unsupervised node feature learning approach for multi-layer networks. We build a multi-layer network, where each layer represents molecular interactions in a different human tissue. OhmNet then automatically learns a mapping of proteins, represented as nodes, to a neural embedding based low-dimensional space of features. OhmNet encourages sharing of similar features among proteins with similar network neighborhoods and among proteins activated in similar tissues. The algorithm generalizes prior work, which generally ignores relationships between tissues, by modeling tissue organization with a rich multiscale tissue hierarchy. We use OhmNet to study multicellular function in a multi-layer protein interaction network of 107 human tissues. In 48 tissues with known tissue-specific cellular functions, OhmNet provides more accurate predictions of cellular function than alternative approaches, and also generates more accurate hypotheses about tissue-specific protein actions. We show that taking into account the tissue hierarchy leads to improved predictive power. Remarkably, we also demonstrate that it is possible to leverage the tissue hierarchy in order to effectively transfer cellular functions to a functionally uncharacterized tissue. Overall, OhmNet moves from flat networks to multiscale models able to predict a range of phenotypes spanning cellular subsystems"
                    },
                    {
                        "title": "Beta Embeddings for Multi-Hop Logical Reasoning in Knowledge Graphs",
                        "abstract": "One of the fundamental problems in Artificial Intelligence is to perform complex multi-hop logical reasoning over the facts captured by a knowledge graph (KG). This problem is challenging, because KGs can be massive and incomplete. Recent approaches embed KG entities in a low dimensional space and then use these embeddings to find the answer entities. However, it has been an outstanding challenge of how to handle arbitrary first-order logic (FOL) queries as present methods are limited to only a subset of FOL operators. In particular, the negation operator is not supported. An additional limitation of present methods is also that they cannot naturally model uncertainty. Here, we present BetaE, a probabilistic embedding framework for answering arbitrary FOL queries over KGs. BetaE is the first method that can handle a complete set of first-order logical operations: conjunction ($\\wedge$), disjunction ($\\vee$), and negation ($\\neg$). A key insight of BetaE is to use probabilistic distributions with bounded support, specifically the Beta distribution, and embed queries/entities as distributions, which as a consequence allows us to also faithfully model uncertainty. Logical operations are performed in the embedding space by neural operators over the probabilistic embeddings. We demonstrate the performance of BetaE on answering arbitrary FOL queries on three large, incomplete KGs. While being more general, BetaE also increases relative performance by up to 25.4% over the current state-of-the-art KG reasoning methods that can only handle conjunctive queries without negation."
                    },
                    {
                        "title": "Signed Networks in Social Media",
                        "abstract": "Relations between users on social media sites often reflect a mixture of positive (friendly) and negative (antagonistic) interactions. In contrast to the bulk of research on social networks that has focused almost exclusively on positive interpretations of links between people, we study how the interplay between positive and negative relationships affects the structure of on-line social networks. We connect our analyses to theories of signed networks from social psychology. We find that the classical theory of structural balance tends to capture certain common patterns of interaction, but that it is also at odds with some of the fundamental phenomena we observe --- particularly related to the evolving, directed nature of these on-line networks. We then develop an alternate theory of status that better explains the observed edge signs and provides insights into the underlying social mechanisms. Our work provides one of the first large-scale evaluations of theories of signed networks using on-line datasets, as well as providing a perspective for reasoning about social media sites."
                    },
                    {
                        "title": "Predicting Positive and Negative Links in Online Social Networks",
                        "abstract": "We study online social networks in which relationships can be either positive (indicating relations such as friendship) or negative (indicating relations such as opposition or antagonism). Such a mix of positive and negative links arise in a variety of online settings; we study datasets from Epinions, Slashdot and Wikipedia. We find that the signs of links in the underlying social networks can be predicted with high accuracy, using models that generalize across this diverse range of sites. These models provide insight into some of the fundamental principles that drive the formation of signed links in networks, shedding light on theories of balance and status from social psychology; they also suggest social computing applications by which the attitude of one user toward another can be estimated from evidence provided by their relationships with other members of the surrounding social network."
                    },
                    {
                        "title": "Governance in Social Media: A case study of the Wikipedia promotion process",
                        "abstract": "Social media sites are often guided by a core group of committed users engaged in various forms of governance. A crucial aspect of this type of governance is deliberation, in which such a group reaches decisions on issues of importance to the site. Despite its crucial --- though subtle --- role in how a number of prominent social media sites function, there has been relatively little investigation of the deliberative aspects of social media governance. Here we explore this issue, investigating a particular deliberative process that is extensive, public, and recorded: the promotion of Wikipedia admins, which is determined by elections that engage committed members of the Wikipedia community. We find that the group decision-making at the heart of this process exhibits several fundamental forms of relative assessment. First we observe that the chance that a voter will support a candidate is strongly dependent on the relationship between characteristics of the voter and the candidate. Second we investigate how both individual voter decisions and overall election outcomes can be based on models that take into account the sequential, public nature of the voting."
                    },
                    {
                        "title": "Graph Evolution: Densification and Shrinking Diameters",
                        "abstract": "How do real graphs evolve over time? What are ``normal'' growth patterns in social, technological, and information networks? Many studies have discovered patterns in static graphs, identifying properties in a single snapshot of a large network, or in a very small number of snapshots; these include heavy tails for in- and out-degree distributions, communities, small-world phenomena, and others. However, given the lack of information about network evolution over long periods, it has been hard to convert these findings into statements about trends over time.   Here we study a wide range of real graphs, and we observe some surprising phenomena. First, most of these graphs densify over time, with the number of edges growing super-linearly in the number of nodes. Second, the average distance between nodes often shrinks over time, in contrast to the conventional wisdom that such distance parameters should increase slowly as a function of the number of nodes (like O(log n) or O(log(log n)).   Existing graph generation models do not exhibit these types of behavior, even at a qualitative level. We provide a new graph generator, based on a ``forest fire'' spreading process, that has a simple, intuitive justification, requires very few parameters (like the ``flammability'' of nodes), and produces graphs exhibiting the full range of properties observed both in prior work and in the present study.   We also notice that the ``forest fire'' model exhibits a sharp transition between sparse graphs and graphs that are densifying. Graphs with decreasing distance between the nodes are generated around this transition point."
                    }
                ]
            }
        }
    },
    "2308.05021": {
        "paper_data": {
            "title": "On Error Propagation of Diffusion Models",
            "url": "http://arxiv.org/abs/2308.05021v3",
            "arxiv_id": "2308.05021",
            "authors": [
                "Yangming Li",
                "Mihaela van der Schaar"
            ],
            "abstract": "Although diffusion models (DMs) have shown promising performances in a number of tasks (e.g., speech synthesis and image generation), they might suffer from error propagation because of their sequential structure. However, this is not certain because some sequential models, such as Conditional Random Field (CRF), are free from this problem. To address this issue, we develop a theoretical framework to mathematically formulate error propagation in the architecture of DMs, The framework contains three elements, including modular error, cumulative error, and propagation equation. The modular and cumulative errors are related by the equation, which interprets that DMs are indeed affected by error propagation. Our theoretical study also suggests that the cumulative error is closely related to the generation quality of DMs. Based on this finding, we apply the cumulative error as a regularization term to reduce error propagation. Because the term is computationally intractable, we derive its upper bound and design a bootstrap algorithm to efficiently estimate the bound for optimization. We have conducted extensive experiments on multiple image datasets, showing that our proposed regularization reduces error propagation, significantly improves vanilla DMs, and outperforms previous baselines.",
            "introduction": "   1 Introduction  DMs potentially suffer from error propagation.  While diffusion models (DMs) have shown impressive performances in a number of tasks (e.g., image synthesis\u00a0(Rombach et\u00a0al., 2022), speech processing\u00a0(Kong et\u00a0al., 2021) and natural language generation\u00a0(Li et\u00a0al., 2022)), they might still suffer from error propagation\u00a0(Motter & Lai, 2002), a classical problem in engineering practices (e.g., communication networks)\u00a0(Fu et\u00a0al., 2020; Motter & Lai, 2002; Crucitti et\u00a0al., 2004). The problem mainly affects some chain models that consist of many end-to-end connected modules. For those models, the output error from one module will spread to subsequent modules such that errors accumulate along the chain. Since DMs are exactly of a chain structure and have a large number of modules (e.g., 1000100010001000 for DDPM\u00a0(Ho et\u00a0al., 2020)), the effect of error propagation on diffusion models is potentially notable and worth a careful study.    Current works lack reliable explanations.  Some recent works\u00a0(Ning et\u00a0al., 2023; Li et\u00a0al., 2023; Daras et\u00a0al., 2023) have noticed this problem and named it as exposure bias or sampling drift. However, those works claim that error propagation happens to DMs simply because the models are of a cascade structure. In fact, many sequential models (e.g., CRF\u00a0(Lafferty et\u00a0al., 2001) and CTC\u00a0(Graves et\u00a0al., 2006)) are free from the problem. Therefore, a solid explanation is expected to answer whether(or even why) error propagation impacts on DMs.    Our theory for the error propagation of DMs.  One main focus of this work is to develop a theoretical framework for analyzing the error propagation of DMs. With this framework, we can clearly understand how this problem is mathematically formulated in the architecture of DMs and easily see whether it has a significant impact on the models.  Our framework contains three elements: modular error, cumulative error, and propagation equation. The first two elements respectively measure the prediction error of one single module and the accumulated error of multiple modules, while the last one tells how these errors are related. We first properly define the errors and then derive the equation, which will be applied to explain why error propagation happens to DMs. We will see that it is a term called amplification factor in the propagation equation that determines whether error accumulation happens. Besides the theoretical study, we also perform empirical experiments to verify whether our theory applies to reality.    Why and how to reduce error propagation.  In our theoretical study, we will see that the cumulative error actually reflects the generation quality of DMs. Hence, if error propagation is very serious in DMs, then we can improve the generation performances by reducing it.  A simple way to reducing error propagation is to treat the cumulative error as a regularization term and directly minimize it at training time. However, we will see that this error is computationally infeasible in terms of both density estimation and inefficient backward process. Therefore, we first introduce an upper bound of the error, which avoids estimating densities. Then, we design a bootstrap algorithm (which is inspired by TD learning (Tesauro et\u00a0al., 1995; Osband et\u00a0al., 2016)) to efficiently estimate the bound though with a certain bias.    Contributions.  The contributions of our paper are as follows:   \u2022  We develop a theoretical framework for analyzing",
            "references": [
                {
                    "title": "On the Equivalence of Consistency-Type Models: Consistency Models, Consistent Diffusion Models, and Fokker-Planck Regularization",
                    "abstract": "The emergence of various notions of ``consistency'' in diffusion models has garnered considerable attention and helped achieve improved sample quality, likelihood estimation, and accelerated sampling. Although similar concepts have been proposed in the literature, the precise relationships among them remain unclear. In this study, we establish theoretical connections between three recent ``consistency'' notions designed to enhance diffusion models for distinct objectives. Our insights offer the potential for a more comprehensive and encompassing framework for consistency-type models."
                },
                {
                    "title": "Alleviating Exposure Bias in Diffusion Models through Sampling with Shifted Time Steps",
                    "abstract": "Diffusion Probabilistic Models (DPM) have shown remarkable efficacy in the synthesis of high-quality images. However, their inference process characteristically requires numerous, potentially hundreds, of iterative steps, which could exaggerate the problem of exposure bias due to the training and inference discrepancy. Previous work has attempted to mitigate this issue by perturbing inputs during training, which consequently mandates the retraining of the DPM. In this work, we conduct a systematic study of exposure bias in DPM and, intriguingly, we find that the exposure bias could be alleviated with a novel sampling method that we propose, without retraining the model. We empirically and theoretically show that, during inference, for each backward time step $t$ and corresponding state $\\hat{x}_t$, there might exist another time step $t_s$ which exhibits superior coupling with $\\hat{x}_t$. Based on this finding, we introduce a sampling method named Time-Shift Sampler. Our framework can be seamlessly integrated to existing sampling algorithms, such as DDPM, DDIM and other high-order solvers, inducing merely minimal additional computations. Experimental results show our method brings significant and consistent improvements in FID scores on different datasets and sampling methods. For example, integrating Time-Shift Sampler to F-PNDM yields a FID=3.88, achieving 44.49\\% improvements as compared to F-PNDM, on CIFAR-10 with 10 sampling steps, which is more performant than the vanilla DDIM with 100 sampling steps. Our code is available at https://github.com/Mingxiao-Li/TS-DPM."
                },
                {
                    "title": "The probability flow ODE is provably fast",
                    "abstract": "We provide the first polynomial-time convergence guarantees for the probability flow ODE implementation (together with a corrector step) of score-based generative modeling. Our analysis is carried out in the wake of recent results obtaining such guarantees for the SDE-based implementation (i.e., denoising diffusion probabilistic modeling or DDPM), but requires the development of novel techniques for studying deterministic dynamics without contractivity. Through the use of a specially chosen corrector step based on the underdamped Langevin diffusion, we obtain better dimension dependence than prior works on DDPM ($O(\\sqrt{d})$ vs. $O(d)$, assuming smoothness of the data distribution), highlighting potential advantages of the ODE framework."
                },
                {
                    "title": "Consistency Models",
                    "abstract": "Diffusion models have significantly advanced the fields of image, audio, and video generation, but they depend on an iterative sampling process that causes slow generation. To overcome this limitation, we propose consistency models, a new family of models that generate high quality samples by directly mapping noise to data. They support fast one-step generation by design, while still allowing multistep sampling to trade compute for sample quality. They also support zero-shot data editing, such as image inpainting, colorization, and super-resolution, without requiring explicit training on these tasks. Consistency models can be trained either by distilling pre-trained diffusion models, or as standalone generative models altogether. Through extensive experiments, we demonstrate that they outperform existing distillation techniques for diffusion models in one- and few-step sampling, achieving the new state-of-the-art FID of 3.55 on CIFAR-10 and 6.20 on ImageNet 64x64 for one-step generation. When trained in isolation, consistency models become a new family of generative models that can outperform existing one-step, non-adversarial generative models on standard benchmarks such as CIFAR-10, ImageNet 64x64 and LSUN 256x256."
                },
                {
                    "title": "Consistent Diffusion Models: Mitigating Sampling Drift by Learning to be Consistent",
                    "abstract": "Imperfect score-matching leads to a shift between the training and the sampling distribution of diffusion models. Due to the recursive nature of the generation process, errors in previous steps yield sampling iterates that drift away from the training distribution. Yet, the standard training objective via Denoising Score Matching (DSM) is only designed to optimize over non-drifted data. To train on drifted data, we propose to enforce a \\emph{consistency} property which states that predictions of the model on its own generated data are consistent across time. Theoretically, we show that if the score is learned perfectly on some non-drifted points (via DSM) and if the consistency property is enforced everywhere, then the score is learned accurately everywhere. Empirically we show that our novel training objective yields state-of-the-art results for conditional and unconditional generation in CIFAR-10 and baseline improvements in AFHQ and FFHQ. We open-source our code and models: https://github.com/giannisdaras/cdm"
                },
                {
                    "title": "Input Perturbation Reduces Exposure Bias in Diffusion Models",
                    "abstract": "Denoising Diffusion Probabilistic Models have shown an impressive generation quality, although their long sampling chain leads to high computational costs. In this paper, we observe that a long sampling chain also leads to an error accumulation phenomenon, which is similar to the exposure bias problem in autoregressive text generation. Specifically, we note that there is a discrepancy between training and testing, since the former is conditioned on the ground truth samples, while the latter is conditioned on the previously generated results. To alleviate this problem, we propose a very simple but effective training regularization, consisting in perturbing the ground truth samples to simulate the inference time prediction errors. We empirically show that, without affecting the recall and precision, the proposed input perturbation leads to a significant improvement in the sample quality while reducing both the training and the inference times. For instance, on CelebA 64$\\times$64, we achieve a new state-of-the-art FID score of 1.27, while saving 37.5% of the training time. The code is publicly available at https://github.com/forever208/DDPM-IP"
                },
                {
                    "title": "Convergence of score-based generative modeling for general data distributions",
                    "abstract": "Score-based generative modeling (SGM) has grown to be a hugely successful method for learning to generate samples from complex data distributions such as that of images and audio. It is based on evolving an SDE that transforms white noise into a sample from the learned distribution, using estimates of the score function, or gradient log-pdf. Previous convergence analyses for these methods have suffered either from strong assumptions on the data distribution or exponential dependencies, and hence fail to give efficient guarantees for the multimodal and non-smooth distributions that arise in practice and for which good empirical performance is observed. We consider a popular kind of SGM -- denoising diffusion models -- and give polynomial convergence guarantees for general data distributions, with no assumptions related to functional inequalities or smoothness. Assuming $L^2$-accurate score estimates, we obtain Wasserstein distance guarantees for any distribution of bounded support or sufficiently decaying tails, as well as TV guarantees for distributions with further smoothness assumptions."
                },
                {
                    "title": "Sampling is as easy as learning the score: theory for diffusion models with minimal data assumptions",
                    "abstract": "We provide theoretical convergence guarantees for score-based generative models (SGMs) such as denoising diffusion probabilistic models (DDPMs), which constitute the backbone of large-scale real-world generative models such as DALL$\\cdot$E 2. Our main result is that, assuming accurate score estimates, such SGMs can efficiently sample from essentially any realistic data distribution. In contrast to prior works, our results (1) hold for an $L^2$-accurate score estimate (rather than $L^\\infty$-accurate); (2) do not require restrictive functional inequality conditions that preclude substantial non-log-concavity; (3) scale polynomially in all relevant problem parameters; and (4) match state-of-the-art complexity guarantees for discretization of the Langevin diffusion, provided that the score error is sufficiently small. We view this as strong theoretical justification for the empirical success of SGMs. We also examine SGMs based on the critically damped Langevin diffusion (CLD). Contrary to conventional wisdom, we provide evidence that the use of the CLD does not reduce the complexity of SGMs."
                },
                {
                    "title": "Diffusion-LM Improves Controllable Text Generation",
                    "abstract": "Controlling the behavior of language models (LMs) without re-training is a major open problem in natural language generation. While recent works have demonstrated successes on controlling simple sentence attributes (e.g., sentiment), there has been little progress on complex, fine-grained controls (e.g., syntactic structure). To address this challenge, we develop a new non-autoregressive language model based on continuous diffusions that we call Diffusion-LM. Building upon the recent successes of diffusion models in continuous domains, Diffusion-LM iteratively denoises a sequence of Gaussian vectors into word vectors, yielding a sequence of intermediate latent variables. The continuous, hierarchical nature of these intermediate variables enables a simple gradient-based algorithm to perform complex, controllable generation tasks. We demonstrate successful control of Diffusion-LM for six challenging fine-grained control tasks, significantly outperforming prior work."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "DiffWave: A Versatile Diffusion Model for Audio Synthesis",
                    "abstract": "In this work, we propose DiffWave, a versatile Diffusion probabilistic model for conditional and unconditional Waveform generation. The model is non-autoregressive, and converts the white noise signal into structured waveform through a Markov chain with a constant number of steps at synthesis. It is efficiently trained by optimizing a variant of variational bound on the data likelihood. DiffWave produces high-fidelity audios in Different Waveform generation tasks, including neural vocoding conditioned on mel spectrogram, class-conditional generation, and unconditional generation. We demonstrate that DiffWave matches a strong WaveNet vocoder in terms of speech quality~(MOS: 4.44 versus 4.43), while synthesizing orders of magnitude faster. In particular, it significantly outperforms autoregressive and GAN-based waveform models in the challenging unconditional generation task in terms of audio quality and sample diversity from various automatic and human evaluations."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Bridging the Gap between Training and Inference for Neural Machine Translation",
                    "abstract": "Neural Machine Translation (NMT) generates target words sequentially in the way of predicting the next word conditioned on the context words. At training time, it predicts with the ground truth words as context while at inference it has to generate the entire sequence from scratch. This discrepancy of the fed context leads to error accumulation among the way. Furthermore, word-level training requires strict matching between the generated sequence and the ground truth sequence which leads to overcorrection over different but reasonable translations. In this paper, we address these issues by sampling context words not only from the ground truth sequence but also from the predicted sequence by the model during training, where the predicted sequence is selected with a sentence-level optimum. Experiment results on Chinese->English and WMT\u201914 English->German translation tasks demonstrate that our approach can achieve significant improvements on multiple datasets."
                },
                {
                    "title": "GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium",
                    "abstract": "Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the \"Frechet Inception Distance\" (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark."
                },
                {
                    "title": "Deep Exploration via Bootstrapped DQN",
                    "abstract": "Efficient exploration in complex environments remains a major challenge for reinforcement learning. We propose bootstrapped DQN, a simple algorithm that explores in a computationally and statistically efficient manner through use of randomized value functions. Unlike dithering strategies such as epsilon-greedy exploration, bootstrapped DQN carries out temporally-extended (or deep) exploration; this can lead to exponentially faster learning. We demonstrate these benefits in complex stochastic MDPs and in the large-scale Arcade Learning Environment. Bootstrapped DQN substantially improves learning times and performance across most Atari games."
                },
                {
                    "title": "Sequence Level Training with Recurrent Neural Networks",
                    "abstract": "Many natural language processing applications use language models to generate text. These models are typically trained to predict the next word in a sequence, given the previous words and some context such as an image. However, at test time the model is expected to generate the entire sequence from scratch. This discrepancy makes generation brittle, as errors may accumulate along the way. We address this issue by proposing a novel sequence level training algorithm that directly optimizes the metric used at test time, such as BLEU or ROUGE. On three different tasks, our approach outperforms several strong baselines for greedy generation. The method is also competitive when these baselines employ beam search, while being several times faster."
                },
                {
                    "title": "Training generative neural networks via Maximum Mean Discrepancy optimization",
                    "abstract": "We consider training a deep neural network to generate samples from an unknown distribution given i.i.d. data. We frame learning as an optimization minimizing a two-sample test statistic\u2014informally speaking, a good generator network produces samples that cause a two-sample test to fail to reject the null hypothesis. As our two-sample test statistic, we use an unbiased estimate of the maximum mean discrepancy, which is the centerpiece of the nonparametric kernel two-sample test proposed by Gretton et al. [2]. We compare to the adversarial nets framework introduced by Goodfellow et al. [1], in which learning is a two-player game between a generator network and an adversarial discriminator network, both trained to outwit the other. From this perspective, the MMD statistic plays the role of the discriminator. In addition to empirical comparisons, we prove bounds on the generalization error incurred by optimizing the empirical MMD."
                },
                {
                    "title": "Deep Learning Face Attributes in the Wild",
                    "abstract": "Predicting face attributes in the wild is challenging due to complex face variations. We propose a novel deep learning framework for attribute prediction in the wild. It cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently. LNet is pre-trained by massive general object categories for face localization, while ANet is pre-trained by massive face identities for attribute prediction. This framework not only outperforms the state-of-the-art with a large margin, but also reveals valuable facts on learning face representation. (1) It shows how the performances of face localization (LNet) and attribute prediction (ANet) can be improved by different pre-training strategies. (2) It reveals that although the filters of LNet are fine-tuned only with image-level attribute tags, their response maps over entire images have strong indication of face locations. This fact enables training LNet for face localization with only image-level annotations, but without face bounding boxes or landmarks, which are required by all attribute recognition works. (3) It also demonstrates that the high-level hidden neurons of ANet automatically discover semantic concepts after pre-training with massive face identities, and such concepts are significantly enriched after fine-tuning with attribute tags. Each attribute can be well explained with a sparse linear combination of these concepts."
                },
                {
                    "title": "A Kernel Two-Sample Test",
                    "abstract": "We propose a framework for analyzing and comparing distributions, which we use to construct statistical tests to determine if two samples are drawn from different distributions. Our test statistic is the largest difference in expectations over functions in the unit ball of a reproducing kernel Hilbert space (RKHS), and is called the maximum mean discrepancy (MMD).We present two distribution free tests based on large deviation bounds for the MMD, and a third test based on the asymptotic distribution of this statistic. The MMD can be computed in quadratic time, although efficient linear time approximations are available. Our statistic is an instance of an integral probability metric, and various classical metrics on distributions are obtained when alternative function classes are used in place of an RKHS. We apply our two-sample tests to a variety of problems, including attribute matching for databases using the Hungarian marriage method, where they perform strongly. Excellent performance is also obtained when comparing distributions over graphs, for which these are the first such tests."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Model for cascading failures in complex networks.",
                    "abstract": "Large but rare cascades triggered by small initial shocks are present in most of the infrastructure networks. Here we present a simple model for cascading failures based on the dynamical redistribution of the flow on the network. We show that the breakdown of a single node is sufficient to collapse the efficiency of the entire system if the node is among the ones with largest load. This is particularly important for real-world networks with a highly hetereogeneous distribution of loads as the Internet and electrical power grids."
                },
                {
                    "title": "Cascade-based attacks on complex networks.",
                    "abstract": "We live in a modern world supported by large, complex networks. Examples range from financial markets to communication and transportation systems. In many realistic situations the flow of physical quantities in the network, as characterized by the loads on nodes, is important. We show that for such networks where loads can redistribute among the nodes, intentional attacks can lead to a cascade of overload failures, which can in turn cause the entire or a substantial part of the network to collapse. This is relevant for real-world networks that possess a highly heterogeneous distribution of loads, such as the Internet and power grids. We demonstrate that the heterogeneity of these networks makes them particularly vulnerable to attacks in that a large-scale cascade may be triggered by disabling a single key node. This brings obvious concerns on the security of such systems."
                },
                {
                    "title": "Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data",
                    "abstract": "We present conditional random fields , a framework for building probabilistic models to segment and label sequence data. Conditional random fields offer several advantages over hidden Markov models and stochastic grammars for such tasks, including the ability to relax strong independence assumptions made in those models. Conditional random fields also avoid a fundamental limitation of maximum entropy Markov models (MEMMs) and other discriminative Markov models based on directed graphical models, which can be biased towards states with few successor states. We present iterative parameter estimation algorithms for conditional random fields and compare the performance of the resulting models to HMMs and MEMMs on synthetic and natural-language data."
                },
                {
                    "title": "Temporal difference learning and TD-Gammon",
                    "abstract": "Ever since the days of Shannon's proposal for a chess-playing algorithm [12] and Samuel's checkers-learning program [10] the domain of complex board games such as Go, chess, checkers, Othello, and backgammon has been widely regarded as an ideal testing ground for exploring a variety of concepts and approaches in artificial intelligence and machine learning. Such board games offer the challenge of tremendous complexity and sophistication required to play at expert level. At the same time, the problem inputs and performance measures are clear-cut and well defined, and the game environment is readily automated in that it is easy to simulate the board, the rules of legal play, and the rules regarding when the game is over and determining the outcome."
                },
                {
                    "title": "Regularizing Score-based Models with Score Fokker-Planck Equations",
                    "abstract": "Score-based generative models learn a family of noise-conditional score functions corresponding to the data density perturbed with increasingly large amounts of noise. These perturbed data densities are tied together by the Fokker-Planck equation (FPE), a PDE governing the spatial-temporal evolution of a density undergoing a diffusion process. In this work, we derive a corresponding equation characterizing the noise-conditional scores of the perturbed data densities (i.e., their gradients), termed the score FPE . Surprisingly, despite impressive empirical performance, we observe that scores learned via denoising score matching (DSM) do not satisfy the underlying score FPE. We mathematically analyze three implications of satisfying the score FPE and a potential explanation for why the score FPE is not ful\ufb01lled in practice. At last, we propose to regularize the DSM objective to enforce satisfaction of the score FPE, and show its effectiveness on synthetic data and MNIST."
                },
                {
                    "title": "Under review as a conference paper at ICLR 2016",
                    "abstract": "We describe a method for learning word embeddings with stochastic dimensionality. Our Infinite Skip-Gram (iSG) model specifies an energy-based joint distribution over a word vector, a context vector, and their dimensionality. By employing the same techniques used to make the Infinite Restricted Boltzmann Machine (C\u00f4t\u00e9 & Larochelle, 2015) tractable, we define vector dimensionality over a countably infinite domain, allowing vectors to grow as needed during training. After training, we find that the distribution over embedding dimensionality for a given word is highly interpretable and leads to an elegant probabilistic mechanism for word sense induction. We show qualitatively and quantitatively that the iSG produces parameter-efficient representations that are robust to language\u2019s inherent ambiguity."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Connectionist Temporal Classi\ufb01cation: Labelling Unsegmented Sequence Data with Recurrent Neural Networks",
                    "abstract": "Many real-world sequence learning tasks require the prediction of sequences of labels from noisy, unsegmented input data. In speech recognition, for example, an acoustic signal is transcribed into words or sub-word units. Recurrent neural networks (RNNs) are powerful sequence learners that would seem well suited to such tasks. However, because they require pre-segmented training data, and post-processing to transform their outputs into label sequences, their applicability has so far been limited. This paper presents a novel method for training RNNs to label un-segmented sequences directly, thereby solving both problems. An experiment on the TIMIT speech corpus demonstrates its advantages over both a baseline HMM and a hybrid HMM-RNN."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does error propagation affect the performance of diffusion models in machine learning, and what theoretical framework can be developed to analyze this phenomenon?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of error propagation in diffusion models is crucial for advancing the understanding of model performance in various applications, such as image synthesis, speech processing, and natural language generation. By addressing this issue, the research community can improve the reliability and effectiveness of diffusion models, leading to better outcomes in practical applications. Furthermore, a deeper understanding of error propagation could inspire new methodologies and frameworks that enhance the robustness of sequential models, potentially influencing future research directions in machine learning.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving the problem of error propagation in diffusion models stem from the complexity of their chain structure, where errors can accumulate across multiple interconnected modules. Naive approaches may fail because they do not account for the intricate relationships between modular and cumulative errors, nor do they adequately address the amplification factor that governs error accumulation. Additionally, the theoretical formulation of these errors and their propagation requires a sophisticated understanding of both the architecture of diffusion models and the mathematical principles involved, making it a non-trivial task.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has identified the issue of error propagation in diffusion models but has often attributed it to the cascade structure without providing a solid theoretical explanation. Existing solutions have not effectively addressed the underlying mechanisms of error propagation, leading to a lack of comprehensive understanding. Barriers such as the complexity of accurately measuring cumulative errors and the computational infeasibility of directly minimizing these errors during training have hindered progress. Our approach differs by introducing a theoretical framework that clearly defines and analyzes the errors, along with a novel bootstrap algorithm to estimate error bounds efficiently.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology includes developing a theoretical framework that consists of three key elements: modular error, cumulative error, and a propagation equation. We will define these errors mathematically and derive the propagation equation to understand their relationships. Empirical experiments will be conducted to validate our theoretical findings. To reduce error propagation, we will treat cumulative error as a regularization term, introducing an upper bound to avoid density estimation issues. We will implement a bootstrap algorithm inspired by TD learning to estimate this bound efficiently. The expected outcomes include a clearer understanding of error propagation in"
            }
        },
        "author_data": {
            "85e7c291-25e2-4b21-82d8-694d6b6c08c0": {
                "pk": "85e7c291-25e2-4b21-82d8-694d6b6c08c0",
                "name": "Yangming Li",
                "collaborators": [
                    "M. Schaar",
                    "Max Ruiz Luyten",
                    "Zhaozhi Qian",
                    "B. V. Breugel"
                ],
                "domain": [
                    "Diffusion Models",
                    "Stochastic Differential Equations",
                    "Time Series Analysis",
                    "Data Synthesis"
                ],
                "publications": [
                    {
                        "title": "Risk-Sensitive Diffusion: Robustly Optimizing Diffusion Models with Noisy Samples",
                        "abstract": "Diffusion models are mainly studied on image data. However, non-image data (e.g., tabular data) are also prevalent in real applications and tend to be noisy due to some inevitable factors in the stage of data collection, degrading the generation quality of diffusion models. In this paper, we consider a novel problem setting where every collected sample is paired with a vector indicating the data quality: risk vector. This setting applies to many scenarios involving noisy data and we propose risk-sensitive SDE, a type of stochastic differential equation (SDE) parameterized by the risk vector, to address it. With some proper coefficients, risk-sensitive SDE can minimize the negative effect of noisy samples on the optimization of diffusion models. We conduct systematic studies for both Gaussian and non-Gaussian noise distributions, providing analytical forms of risk-sensitive SDE. To verify the effectiveness of our method, we have conducted extensive experiments on multiple tabular and time-series datasets, showing that risk-sensitive SDE permits a robust optimization of diffusion models with noisy samples and significantly outperforms previous baselines."
                    },
                    {
                        "title": "Risk-Sensitive Diffusion: Learning the Underlying Distribution from Noisy Samples",
                        "abstract": "While achieving remarkable performances, we show that diffusion models are fragile to the presence of noisy samples, limiting their potential in the vast amount of settings where, unlike image synthesis, we are not blessed with clean data. Motivated by our finding that such fragility originates from the distribution gaps between noisy and clean samples along the diffusion process, we introduce risk-sensitive SDE , a stochastic differential equation that is parameterized by the risk (i.e., data \u201cdirtiness\u201d) to adjust the distributions of noisy samples, reducing misguidance while bene-fiting from their contained information. The optimal expression for risk-sensitive SDE depends on the specific noise distribution, and we derive its parameterizations that minimize the misguidance of noisy samples for both Gaussian and general non-Gaussian perturbations. We conduct extensive experiments on both synthetic and real-world datasets (e.g., medical time series), showing that our model effectively recovers the clean data distribution from noisy samples, significantly outper-forming conditional generation baselines."
                    },
                    {
                        "title": "A Study of Posterior Stability for Time-Series Latent Diffusion",
                        "abstract": "Latent diffusion has demonstrated promising results in image generation and permits efficient sampling. However, this framework might suffer from the problem of posterior collapse when applied to time series. In this paper, we first show that posterior collapse will reduce latent diffusion to a variational autoencoder (VAE), making it less expressive. This highlights the importance of addressing this issue. We then introduce a principled method: dependency measure, that quantifies the sensitivity of a recurrent decoder to input variables. Using this tool, we confirm that posterior collapse significantly affects time-series latent diffusion on real datasets, and a phenomenon termed dependency illusion is also discovered in the case of shuffled time series. Finally, building on our theoretical and empirical studies, we introduce a new framework that extends latent diffusion and has a stable posterior. Extensive experiments on multiple real time-series datasets show that our new framework is free from posterior collapse and significantly outperforms previous baselines in time series synthesis."
                    },
                    {
                        "title": "Do Diffusion Models Suffer Error Propagation? Theoretical Analysis and Consistency Regularization",
                        "abstract": "While diffusion models have achieved promising performances in data synthesis, they might suffer error propagation because of their cascade structure , where the distributional mismatch spreads and magnifies through the chain of denoising modules. However, a strict analysis is expected since many sequential models such as Conditional Random Field (CRF) are free from error propagation. In this paper, we empirically and theoretically verify that diffusion models are indeed affected by error propagation and we then propose a regularization to address this problem. Our theoretical analysis reveals that the question can be reduced to whether every denoising module of the diffusion model is fault-tolerant. We derive insightful transition equations, indicating that the module can\u2019t recover from input errors and even propagates additional errors to the next module. Our analysis directly leads to a consistency regularization scheme for diffusion models, which explicitly reduces the distribution gap between forward and backward processes. We further introduce a bootstrapping algorithm to reduce the computation cost of the regularizer. Our experimental results on multiple image datasets show that our regularization effectively handles error propagation and significantly improves the performance of vanilla diffusion models."
                    },
                    {
                        "title": "Soft Mixture Denoising: Beyond the Expressive Bottleneck of Diffusion Models",
                        "abstract": "Because diffusion models have shown impressive performances in a number of tasks, such as image synthesis, there is a trend in recent works to prove (with certain assumptions) that these models have strong approximation capabilities. In this paper, we show that current diffusion models actually have an expressive bottleneck in backward denoising and some assumption made by existing theoretical guarantees is too strong. Based on this finding, we prove that diffusion models have unbounded errors in both local and global denoising. In light of our theoretical studies, we introduce soft mixture denoising (SMD), an expressive and efficient model for backward denoising. SMD not only permits diffusion models to well approximate any Gaussian mixture distributions in theory, but also is simple and efficient for implementation. Our experiments on multiple image datasets show that SMD significantly improves different types of diffusion models (e.g., DDPM), espeically in the situation of few backward iterations."
                    }
                ]
            },
            "ccbfe499-d197-4cd0-a003-3700ff934a99": {
                "pk": "ccbfe499-d197-4cd0-a003-3700ff934a99",
                "name": "Mihaela van der Schaar",
                "collaborators": [
                    "Samuel Holt",
                    "Nabeel Seedat",
                    "Krzysztof Kacprzyk",
                    "Tennison Liu",
                    "Max Ruiz Luyten",
                    "Nicolas Huynh",
                    "T. Pouplin",
                    "Jeroen Berrevoets",
                    "Zhaozhi Qian",
                    "Yangming Li"
                ],
                "domain": [
                    "Semi-Supervised Learning",
                    "Causal Inference",
                    "Large Language Models",
                    "Digital Twins"
                ],
                "publications": [
                    {
                        "title": "You can't handle the (dirty) truth: Data-centric insights improve pseudo-labeling",
                        "abstract": "Pseudo-labeling is a popular semi-supervised learning technique to leverage unlabeled data when labeled samples are scarce. The generation and selection of pseudo-labels heavily rely on labeled data. Existing approaches implicitly assume that the labeled data is gold standard and 'perfect'. However, this can be violated in reality with issues such as mislabeling or ambiguity. We address this overlooked aspect and show the importance of investigating labeled data quality to improve any pseudo-labeling method. Specifically, we introduce a novel data characterization and selection framework called DIPS to extend pseudo-labeling. We select useful labeled and pseudo-labeled samples via analysis of learning dynamics. We demonstrate the applicability and impact of DIPS for various pseudo-labeling methods across an extensive range of real-world tabular and image datasets. Additionally, DIPS improves data efficiency and reduces the performance distinctions between different pseudo-labelers. Overall, we highlight the significant benefits of a data-centric rethinking of pseudo-labeling in real-world settings."
                    },
                    {
                        "title": "Retrieval Augmented Thought Process for Private Data Handling in Healthcare",
                        "abstract": "Large Language Models (LLMs) have demonstrated the strong potential to assist both clinicians and the general public with their extensive medical knowledge. However, their application in healthcare is constrained due to concerns about the privacy of data used in training, which prevents the integration of private and personal information because of security and ethical issues. Moreover, if their capabilities can be enhanced with information retrieval to access up-to-date knowledge, the current integration of LLMs with Information retrieval lacks robustness to imperfect retrieval, which can hinder their effectiveness and even reduce overall performance. In this work, we address this challenge by introducing the Retrieval-Augmented Thought Process (RATP). Given access to external knowledge, RATP formulates the thought generation of LLMs as a multiple-step decision process. To optimise such a thought process, RATP leverages Monte-Carlo Tree Search and learns a proxy reward function that permits cost-efficient inference. On a private dataset of electronic medical records, deliberately excluded from any LLM training set, RATP achieves 35% additional accuracy compared to in-context retrieval-augmented generation for the question-answering task."
                    },
                    {
                        "title": "ODE Discovery for Longitudinal Heterogeneous Treatment Effects Inference",
                        "abstract": "Inferring unbiased treatment effects has received widespread attention in the machine learning community. In recent years, our community has proposed numerous solutions in standard settings, high-dimensional treatment settings, and even longitudinal settings. While very diverse, the solution has mostly relied on neural networks for inference and simultaneous correction of assignment bias. New approaches typically build on top of previous approaches by proposing new (or refined) architectures and learning algorithms. However, the end result -- a neural-network-based inference machine -- remains unchallenged. In this paper, we introduce a different type of solution in the longitudinal setting: a closed-form ordinary differential equation (ODE). While we still rely on continuous optimization to learn an ODE, the resulting inference machine is no longer a neural network. Doing so yields several advantages such as interpretability, irregular sampling, and a different set of identification assumptions. Above all, we consider the introduction of a completely new type of solution to be our most important contribution as it may spark entirely new innovations in treatment effects in general. We facilitate this by formulating our contribution as a framework that can transform any ODE discovery method into a treatment effects method."
                    },
                    {
                        "title": "Meta-Learners for Partially-Identified Treatment Effects Across Multiple Environments",
                        "abstract": "Estimating the conditional average treatment effect (CATE) from observational data is relevant for many applications such as personalized medicine. Here, we focus on the widespread setting where the observational data come from multiple environments, such as different hospitals, physicians, or countries. Furthermore, we allow for violations of standard causal assumptions, namely, overlap within the environments and unconfoundedness. To this end, we move away from point identification and focus on partial identification. Specifically, we show that current assumptions from the literature on multiple environments allow us to interpret the environment as an instrumental variable (IV). This allows us to adapt bounds from the IV literature for partial identification of CATE by leveraging treatment assignment mechanisms across environments. Then, we propose different model-agnostic learners (so-called meta-learners) to estimate the bounds that can be used in combination with arbitrary machine learning models. We further demonstrate the effectiveness of our meta-learners across various experiments using both simulated and real-world data. Finally, we discuss the applicability of our meta-learners to partial identification in instrumental variable settings, such as randomized controlled trials with non-compliance."
                    },
                    {
                        "title": "Shape Arithmetic Expressions: Advancing Scientific Discovery Beyond Closed-Form Equations",
                        "abstract": "Symbolic regression has excelled in uncovering equations from physics, chemistry, biology, and related disciplines. However, its effectiveness becomes less certain when applied to experimental data lacking inherent closed-form expressions. Empirically derived relationships, such as entire stress-strain curves, may defy concise closed-form representation, compelling us to explore more adaptive modeling approaches that balance flexibility with interpretability. In our pursuit, we turn to Generalized Additive Models (GAMs), a widely used class of models known for their versatility across various domains. Although GAMs can capture non-linear relationships between variables and targets, they cannot capture intricate feature interactions. In this work, we investigate both of these challenges and propose a novel class of models, Shape Arithmetic Expressions (SHAREs), that fuses GAM's flexible shape functions with the complex feature interactions found in mathematical expressions. SHAREs also provide a unifying framework for both of these approaches. We also design a set of rules for constructing SHAREs that guarantee transparency of the found expressions beyond the standard constraints based on the model's size."
                    },
                    {
                        "title": "Large Language Models to Enhance Bayesian Optimization",
                        "abstract": "Bayesian optimization (BO) is a powerful approach for optimizing complex and expensive-to-evaluate black-box functions. Its importance is underscored in many applications, notably including hyperparameter tuning, but its efficacy depends on efficiently balancing exploration and exploitation. While there has been substantial progress in BO methods, striking this balance remains a delicate process. In this light, we present LLAMBO, a novel approach that integrates the capabilities of Large Language Models (LLM) within BO. At a high level, we frame the BO problem in natural language, enabling LLMs to iteratively propose and evaluate promising solutions conditioned on historical evaluations. More specifically, we explore how combining contextual understanding, few-shot learning proficiency, and domain knowledge of LLMs can improve model-based BO. Our findings illustrate that LLAMBO is effective at zero-shot warmstarting, and enhances surrogate modeling and candidate sampling, especially in the early stages of search when observations are sparse. Our approach is performed in context and does not require LLM finetuning. Additionally, it is modular by design, allowing individual components to be integrated into existing BO frameworks, or function cohesively as an end-to-end method. We empirically validate LLAMBO's efficacy on the problem of hyperparameter tuning, highlighting strong empirical performance across a range of diverse benchmarks, proprietary, and synthetic tasks."
                    },
                    {
                        "title": "Automatically Learning Hybrid Digital Twins of Dynamical Systems",
                        "abstract": "Digital Twins (DTs) are computational models that simulate the states and temporal dynamics of real-world systems, playing a crucial role in prediction, understanding, and decision-making across diverse domains. However, existing approaches to DTs often struggle to generalize to unseen conditions in data-scarce settings, a crucial requirement for such models. To address these limitations, our work begins by establishing the essential desiderata for effective DTs. Hybrid Digital Twins ($\\textbf{HDTwins}$) represent a promising approach to address these requirements, modeling systems using a composition of both mechanistic and neural components. This hybrid architecture simultaneously leverages (partial) domain knowledge and neural network expressiveness to enhance generalization, with its modular design facilitating improved evolvability. While existing hybrid models rely on expert-specified architectures with only parameters optimized on data, $\\textit{automatically}$ specifying and optimizing HDTwins remains intractable due to the complex search space and the need for flexible integration of domain priors. To overcome this complexity, we propose an evolutionary algorithm ($\\textbf{HDTwinGen}$) that employs Large Language Models (LLMs) to autonomously propose, evaluate, and optimize HDTwins. Specifically, LLMs iteratively generate novel model specifications, while offline tools are employed to optimize emitted parameters. Correspondingly, proposed models are evaluated and evolved based on targeted feedback, enabling the discovery of increasingly effective hybrid models. Our empirical results reveal that HDTwinGen produces generalizable, sample-efficient, and evolvable models, significantly advancing DTs' efficacy in real-world applications."
                    },
                    {
                        "title": "Causal Deep Learning: Encouraging Impact on Real-world Problems Through Causality",
                        "abstract": "Causality has the potential to truly transform the way we solve a large number of real-world problems. Yet, so far, its potential largely remains to be unlocked as causality often requires crucial assumptions which cannot be tested in practice. To address this challenge, we propose a new way of thinking about causality-\u2013 we call this causal deep learning . Our causal deep learning framework spans three dimensions: (1) a structural dimension, which incorporates partial yet testable causal knowledge rather than assuming either complete or no causal knowledge among the variables of interest; (2) a parametric dimension, which encompasses parametric forms that capture the type of relationships among the variables of interest; and (3) a temporal dimen-sion, which captures exposure times or how the variables of interest interact (possibly causally) over time. Our CDL framework enables us to precisely categorise and compare"
                    },
                    {
                        "title": "Context-Aware Testing: A New Paradigm for Model Testing with Large Language Models",
                        "abstract": "The predominant de facto paradigm of testing ML models relies on either using only held-out data to compute aggregate evaluation metrics or by assessing the performance on different subgroups. However, such data-only testing methods operate under the restrictive assumption that the available empirical data is the sole input for testing ML models, disregarding valuable contextual information that could guide model testing. In this paper, we challenge the go-to approach of data-only testing and introduce context-aware testing (CAT) which uses context as an inductive bias to guide the search for meaningful model failures. We instantiate the first CAT system, SMART Testing, which employs large language models to hypothesize relevant and likely failures, which are evaluated on data using a self-falsification mechanism. Through empirical evaluations in diverse settings, we show that SMART automatically identifies more relevant and impactful failures than alternatives, demonstrating the potential of CAT as a testing paradigm."
                    },
                    {
                        "title": "Risk-Sensitive Diffusion: Robustly Optimizing Diffusion Models with Noisy Samples",
                        "abstract": "Diffusion models are mainly studied on image data. However, non-image data (e.g., tabular data) are also prevalent in real applications and tend to be noisy due to some inevitable factors in the stage of data collection, degrading the generation quality of diffusion models. In this paper, we consider a novel problem setting where every collected sample is paired with a vector indicating the data quality: risk vector. This setting applies to many scenarios involving noisy data and we propose risk-sensitive SDE, a type of stochastic differential equation (SDE) parameterized by the risk vector, to address it. With some proper coefficients, risk-sensitive SDE can minimize the negative effect of noisy samples on the optimization of diffusion models. We conduct systematic studies for both Gaussian and non-Gaussian noise distributions, providing analytical forms of risk-sensitive SDE. To verify the effectiveness of our method, we have conducted extensive experiments on multiple tabular and time-series datasets, showing that risk-sensitive SDE permits a robust optimization of diffusion models with noisy samples and significantly outperforms previous baselines."
                    },
                    {
                        "title": "CliMB: An AI-enabled Partner for Clinical Predictive Modeling",
                        "abstract": "Despite its significant promise and continuous technical advances, real-world applications of artificial intelligence (AI) remain limited. We attribute this to the\"domain expert-AI-conundrum\": while domain experts, such as clinician scientists, should be able to build predictive models such as risk scores, they face substantial barriers in accessing state-of-the-art (SOTA) tools. While automated machine learning (AutoML) has been proposed as a partner in clinical predictive modeling, many additional requirements need to be fulfilled to make machine learning accessible for clinician scientists. To address this gap, we introduce CliMB, a no-code AI-enabled partner designed to empower clinician scientists to create predictive models using natural language. CliMB guides clinician scientists through the entire medical data science pipeline, thus empowering them to create predictive models from real-world data in just one conversation. CliMB also creates structured reports and interpretable visuals. In evaluations involving clinician scientists and systematic comparisons against a baseline GPT-4, CliMB consistently demonstrated superior performance in key areas such as planning, error prevention, code execution, and model performance. Moreover, in blinded assessments involving 45 clinicians from diverse specialties and career stages, more than 80% preferred CliMB over GPT-4. Overall, by providing a no-code interface with clear guidance and access to SOTA methods in the fields of data-centric AI, AutoML, and interpretable ML, CliMB empowers clinician scientists to build robust predictive models."
                    },
                    {
                        "title": "Time Series Diffusion in the Frequency Domain",
                        "abstract": "Fourier analysis has been an instrumental tool in the development of signal processing. This leads us to wonder whether this framework could similarly benefit generative modelling. In this paper, we explore this question through the scope of time series diffusion models. More specifically, we analyze whether representing time series in the frequency domain is a useful inductive bias for score-based diffusion models. By starting from the canonical SDE formulation of diffusion in the time domain, we show that a dual diffusion process occurs in the frequency domain with an important nuance: Brownian motions are replaced by what we call mirrored Brownian motions, characterized by mirror symmetries among their components. Building on this insight, we show how to adapt the denoising score matching approach to implement diffusion models in the frequency domain. This results in frequency diffusion models, which we compare to canonical time diffusion models. Our empirical evaluation on real-world datasets, covering various domains like healthcare and finance, shows that frequency diffusion models better capture the training distribution than time diffusion models. We explain this observation by showing that time series from these datasets tend to be more localized in the frequency domain than in the time domain, which makes them easier to model in the former case. All our observations point towards impactful synergies between Fourier analysis and diffusion models."
                    },
                    {
                        "title": "Informed Meta-Learning",
                        "abstract": "In noisy and low-data regimes prevalent in real-world applications, a key challenge of machine learning lies in effectively incorporating inductive biases that promote data efficiency and robustness. Meta-learning and informed ML stand out as two approaches for incorporating prior knowledge into ML pipelines. While the former relies on a purely data-driven source of priors, the latter is guided by prior domain knowledge. In this paper, we formalise a hybrid paradigm, informed meta-learning, facilitating the incorporation of priors from unstructured knowledge representations, such as natural language; thus, unlocking complementarity in cross-task knowledge sharing of humans and machines. We establish the foundational components of informed meta-learning and present a concrete instantiation of this framework--the Informed Neural Process. Through a series of experiments, we demonstrate the potential benefits of informed meta-learning in improving data efficiency, robustness to observational noise and task distribution shifts."
                    },
                    {
                        "title": "Risk-Sensitive Diffusion: Learning the Underlying Distribution from Noisy Samples",
                        "abstract": "While achieving remarkable performances, we show that diffusion models are fragile to the presence of noisy samples, limiting their potential in the vast amount of settings where, unlike image synthesis, we are not blessed with clean data. Motivated by our finding that such fragility originates from the distribution gaps between noisy and clean samples along the diffusion process, we introduce risk-sensitive SDE , a stochastic differential equation that is parameterized by the risk (i.e., data \u201cdirtiness\u201d) to adjust the distributions of noisy samples, reducing misguidance while bene-fiting from their contained information. The optimal expression for risk-sensitive SDE depends on the specific noise distribution, and we derive its parameterizations that minimize the misguidance of noisy samples for both Gaussian and general non-Gaussian perturbations. We conduct extensive experiments on both synthetic and real-world datasets (e.g., medical time series), showing that our model effectively recovers the clean data distribution from noisy samples, significantly outper-forming conditional generation baselines."
                    },
                    {
                        "title": "Unveiling the Power of Sparse Neural Networks for Feature Selection",
                        "abstract": "Sparse Neural Networks (SNNs) have emerged as powerful tools for efficient feature selection. Leveraging the dynamic sparse training (DST) algorithms within SNNs has demonstrated promising feature selection capabilities while drastically reducing computational overheads. Despite these advancements, several critical aspects remain insufficiently explored for feature selection. Questions persist regarding the choice of the DST algorithm for network training, the choice of metric for ranking features/neurons, and the comparative performance of these methods across diverse datasets when compared to dense networks. This paper addresses these gaps by presenting a comprehensive systematic analysis of feature selection with sparse neural networks. Moreover, we introduce a novel metric considering sparse neural network characteristics, which is designed to quantify feature importance within the context of SNNs. Our findings show that feature selection with SNNs trained with DST algorithms can achieve, on average, more than $50\\%$ memory and $55\\%$ FLOPs reduction compared to the dense networks, while outperforming them in terms of the quality of the selected features. Our code and the supplementary material are available on GitHub (\\url{https://github.com/zahraatashgahi/Neuron-Attribution})."
                    },
                    {
                        "title": "Discovering Preference Optimization Algorithms with and for Large Language Models",
                        "abstract": "Offline preference optimization is a key method for enhancing and controlling the quality of Large Language Model (LLM) outputs. Typically, preference optimization is approached as an offline supervised learning task using manually-crafted convex loss functions. While these methods are based on theoretical insights, they are inherently constrained by human creativity, so the large search space of possible loss functions remains under explored. We address this by performing LLM-driven objective discovery to automatically discover new state-of-the-art preference optimization algorithms without (expert) human intervention. Specifically, we iteratively prompt an LLM to propose and implement new preference optimization loss functions based on previously-evaluated performance metrics. This process leads to the discovery of previously-unknown and performant preference optimization algorithms. The best performing of these we call Discovered Preference Optimization (DiscoPOP), a novel algorithm that adaptively blends logistic and exponential losses. Experiments demonstrate the state-of-the-art performance of DiscoPOP and its successful transfer to held-out tasks."
                    }
                ]
            }
        }
    },
    "2406.11838": {
        "paper_data": {
            "title": "Autoregressive Image Generation without Vector Quantization",
            "url": "http://arxiv.org/abs/2406.11838v2",
            "arxiv_id": "2406.11838",
            "authors": [
                "Tianhong Li",
                "Yonglong Tian",
                "He Li",
                "Mingyang Deng",
                "Kaiming He"
            ],
            "abstract": "Conventional wisdom holds that autoregressive models for image generation are typically accompanied by vector-quantized tokens. We observe that while a discrete-valued space can facilitate representing a categorical distribution, it is not a necessity for autoregressive modeling. In this work, we propose to model the per-token probability distribution using a diffusion procedure, which allows us to apply autoregressive models in a continuous-valued space. Rather than using categorical cross-entropy loss, we define a Diffusion Loss function to model the per-token probability. This approach eliminates the need for discrete-valued tokenizers. We evaluate its effectiveness across a wide range of cases, including standard autoregressive models and generalized masked autoregressive (MAR) variants. By removing vector quantization, our image generator achieves strong results while enjoying the speed advantage of sequence modeling. We hope this work will motivate the use of autoregressive generation in other continuous-valued domains and applications. Code is available at: https://github.com/LTH14/mar",
            "introduction": "   1 Introduction  Autoregressive models are currently the de facto solution to generative models in natural language processing [38, 39, 3]. These models predict the next word or token in a sequence based on the previous words as input. Given the discrete nature of languages, the inputs and outputs of these models are in a categorical, discrete-valued space. This prevailing approach has led to a widespread belief that autoregressive models are inherently linked to discrete representations.   As a result, research on generalizing autoregressive models to continuous-valued domains\u2014most notably, image generation\u2014has intensely focused on discretizing the data [6, 13, 40]. A commonly adopted strategy is to train a discrete-valued tokenizer on images, which involves a finite vocabulary obtained by vector quantization (VQ) [51, 41]. Autoregressive models are then operated on the discrete-valued token space, analogous to their language counterparts.   In this work, we aim to address the following question: Is it necessary for autoregressive models to be coupled with vector-quantized representations? We note that the autoregressive nature, i.e., \u201cpredicting next tokens based on previous ones\u201d, is independent of whether the values are discrete or continuous. What is needed is to model the per-token probability distribution, which can be measured by a loss function and used to draw samples from. Discrete-valued representations can be conveniently modeled by a categorical distribution, but it is not conceptually necessary. If alternative models for per-token probability distributions are presented, autoregressive models can be approached without vector quantization.   With this observation, we propose to model the per-token probability distribution by a diffusion procedure operating on continuous-valued domains. Our methodology leverages the principles of diffusion models [45, 24, 33, 10] for representing arbitrary probability distributions. Specifically, our method autoregressively predicts a vector z\ud835\udc67zitalic_z for each token, which serves as a conditioning for a denoising network (e.g., a small MLP). The denoising diffusion procedure enables us to represent an underlying distribution p\u2062(x|z)\ud835\udc5dconditional\ud835\udc65\ud835\udc67p(x|z)italic_p ( italic_x | italic_z ) for the output x\ud835\udc65xitalic_x (Figure\u00a01). This small denoising network is trained jointly with the autoregressive model, with continuous-valued tokens as the input and target. Conceptually, this small prediction head, applied to each token, behaves like a loss function for measuring the quality of z\ud835\udc67zitalic_z. We refer to this loss function as Diffusion Loss.      Figure 1: Diffusion Loss. Given a continuous-valued token x\ud835\udc65xitalic_x to be predicted, the autoregressive model produces a vector z\ud835\udc67zitalic_z, which serves as the condition of a denoising diffusion network (a small MLP). This offers a way to model the probability distribution p\u2062(x|z)\ud835\udc5dconditional\ud835\udc65\ud835\udc67p(x|z)italic_p ( italic_x | italic_z ) of this token. This network is trained jointly with the autoregressive model by backpropagation. At inference time, with a predicted z\ud835\udc67zitalic_z, running the reverse diffusion procedure can sample a token following the distribution: x\u223cp\u2062(x|z)similar-to\ud835\udc65\ud835\udc5dconditional\ud835\udc65\ud835\udc67x\\sim p(x|z)italic_x \u223c italic_p ( italic_x | italic_z ). This method eliminates the need for discrete-valued tokenizers.    Our approach eliminates the need for discrete-valued tokenizers. Vector-quantized tokenizers are difficult to train and are sensitive to gradient approximation strategies [51, 41, 40, 27]. Their reconstruction quality often falls short compared to continuous-valued counterparts [42]. Our approach allows autoregressive models to enjoy the benefits of higher-quality, non-quantized tokenizers.   To broaden the scope, we further unify standard autoregressive (AR)",
            "references": [
                {
                    "title": "Octo: An Open-Source Generalist Robot Policy",
                    "abstract": "Large policies pretrained on diverse robot datasets have the potential to transform robotic learning: instead of training new policies from scratch, such generalist robot policies may be finetuned with only a little in-domain data, yet generalize broadly. However, to be widely applicable across a range of robotic learning scenarios, environments, and tasks, such policies need to handle diverse sensors and action spaces, accommodate a variety of commonly used robotic platforms, and finetune readily and efficiently to new domains. In this work, we aim to lay the groundwork for developing open-source, widely applicable, generalist policies for robotic manipulation. As a first step, we introduce Octo, a large transformer-based policy trained on 800k trajectories from the Open X-Embodiment dataset, the largest robot manipulation dataset to date. It can be instructed via language commands or goal images and can be effectively finetuned to robot setups with new sensory inputs and action spaces within a few hours on standard consumer GPUs. In experiments across 9 robotic platforms, we demonstrate that Octo serves as a versatile policy initialization that can be effectively finetuned to new observation and action spaces. We also perform detailed ablations of design decisions for the Octo model, from architecture to training data, to guide future research on building generalist robot models."
                },
                {
                    "title": "Denoising Autoregressive Representation Learning",
                    "abstract": "In this paper, we explore a new generative approach for learning visual representations. Our method, DARL, employs a decoder-only Transformer to predict image patches autoregressively. We find that training with Mean Squared Error (MSE) alone leads to strong representations. To enhance the image generation ability, we replace the MSE loss with the diffusion objective by using a denoising patch decoder. We show that the learned representation can be improved by using tailored noise schedules and longer training in larger models. Notably, the optimal schedule differs significantly from the typical ones used in standard image diffusion models. Overall, despite its simple architecture, DARL delivers performance remarkably close to state-of-the-art masked prediction models under the fine-tuning protocol. This marks an important step towards a unified model capable of both visual perception and generation, effectively combining the strengths of autoregressive and denoising diffusion models."
                },
                {
                    "title": "Analyzing and Improving the Training Dynamics of Diffusion Models",
                    "abstract": "Diffusion models currently dominate the field of data- driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high- level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on ex- pectation. We find that systematic application of this philoso- phy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational com- plexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling. As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance."
                },
                {
                    "title": "DiffiT: Diffusion Vision Transformers for Image Generation",
                    "abstract": "Diffusion models with their powerful expressivity and high sample quality have achieved State-Of-The-Art (SOTA) performance in the generative domain. The pioneering Vision Transformer (ViT) has also demonstrated strong modeling capabilities and scalability, especially for recognition tasks. In this paper, we study the effectiveness of ViTs in diffusion-based generative learning and propose a new model denoted as Diffusion Vision Transformers (DiffiT). Specifically, we propose a methodology for finegrained control of the denoising process and introduce the Time-dependant Multihead Self Attention (TMSA) mechanism. DiffiT is surprisingly effective in generating high-fidelity images with significantly better parameter efficiency. We also propose latent and image space DiffiT models and show SOTA performance on a variety of class-conditional and unconditional synthesis tasks at different resolutions. The Latent DiffiT model achieves a new SOTA FID score of 1.73 on ImageNet256 dataset while having 19.85%, 16.88% less parameters than other Transformer-based diffusion models such as MDT and DiT,respectively. Code: https://github.com/NVlabs/DiffiT"
                },
                {
                    "title": "Language Model Beats Diffusion -- Tokenizer is Key to Visual Generation",
                    "abstract": "While Large Language Models (LLMs) are the dominant models for generative tasks in language, they do not perform as well as diffusion models on image and video generation. To effectively use LLMs for visual generation, one crucial component is the visual tokenizer that maps pixel-space inputs to discrete tokens appropriate for LLM learning. In this paper, we introduce MAGVIT-v2, a video tokenizer designed to generate concise and expressive tokens for both videos and images using a common token vocabulary. Equipped with this new tokenizer, we show that LLMs outperform diffusion models on standard image and video generation benchmarks including ImageNet and Kinetics. In addition, we demonstrate that our tokenizer surpasses the previously top-performing video tokenizer on two more tasks: (1) video compression comparable to the next-generation video codec (VCC) according to human evaluations, and (2) learning effective representations for action recognition tasks."
                },
                {
                    "title": "Finite Scalar Quantization: VQ-VAE Made Simple",
                    "abstract": "We propose to replace vector quantization (VQ) in the latent representation of VQ-VAEs with a simple scheme termed finite scalar quantization (FSQ), where we project the VAE representation down to a few dimensions (typically less than 10). Each dimension is quantized to a small set of fixed values, leading to an (implicit) codebook given by the product of these sets. By appropriately choosing the number of dimensions and values each dimension can take, we obtain the same codebook size as in VQ. On top of such discrete representations, we can train the same models that have been trained on VQ-VAE representations. For example, autoregressive and masked transformer models for image generation, multimodal generation, and dense prediction computer vision tasks. Concretely, we employ FSQ with MaskGIT for image generation, and with UViM for depth estimation, colorization, and panoptic segmentation. Despite the much simpler design of FSQ, we obtain competitive performance in all these tasks. We emphasize that FSQ does not suffer from codebook collapse and does not need the complex machinery employed in VQ (commitment losses, codebook reseeding, code splitting, entropy penalties, etc.) to learn expressive discrete representations."
                },
                {
                    "title": "Diffusion Models as Masked Autoencoders",
                    "abstract": "There has been a longstanding belief that generation can facilitate a true understanding of visual data. In line with this, we revisit generatively pre-training visual representations in light of recent interest in denoising diffusion models. While directly pre-training with diffusion models does not produce strong representations, we condition diffusion models on masked input and formulate diffusion models as masked autoencoders (DiffMAE). Our approach is capable of (i) serving as a strong initialization for downstream recognition tasks, (ii) conducting high-quality image inpainting, and (iii) being effortlessly extended to video where it produces state-of-the-art classification accuracy. We further perform a comprehensive study on the pros and cons of design choices and build connections between diffusion models and masked autoencoders. Project page."
                },
                {
                    "title": "Masked Diffusion Transformer is a Strong Image Synthesizer",
                    "abstract": "Despite its success in image synthesis, we observe that diffusion probabilistic models (DPMs) often lack contextual reasoning ability to learn the relations among object parts in an image, leading to a slow learning process. To solve this issue, we propose a Masked Diffusion Transformer (MDT) that introduces a mask latent modeling scheme to explicitly enhance the DPMs\u2019 ability to contextual relation learning among object semantic parts in an image. During training, MDT operates in the latent space to mask certain tokens. Then, an asymmetric masking diffusion transformer is designed to predict masked tokens from unmasked ones while maintaining the diffusion generation process. Our MDT can reconstruct the full information of an image from its incomplete contextual input, thus enabling it to learn the associated relations among image tokens. Experimental results show that MDT achieves superior image synthesis performance, e.g., a new SOTA FID score in the ImageNet data set, and has about 3\u00d7 faster learning speed than the previous SOTA DiT. The source code is released at https://github.com/sail-sg/MDT."
                },
                {
                    "title": "Diffusion Policy: Visuomotor Policy Learning via Action Diffusion",
                    "abstract": "This paper introduces Diffusion Policy, a new way of generating robot behavior by representing a robot\u2019s visuomotor policy as a conditional denoising diffusion process. We benchmark Diffusion Policy across 15 different tasks from 4 different robot manipulation benchmarks and find that it consistently outperforms existing state-of-the-art robot learning methods with an average improvement of 46.9%. Diffusion Policy learns the gradient of the action-distribution score function and iteratively optimizes with respect to this gradient field during inference via a series of stochastic Langevin dynamics steps. We find that the diffusion formulation yields powerful advantages when used for robot policies, including gracefully handling multimodal action distributions, being suitable for high-dimensional action spaces, and exhibiting impressive training stability. To fully unlock the potential of diffusion models for visuomotor policy learning on physical robots, this paper presents a set of key technical contributions including the incorporation of receding horizon control, visual conditioning, and the time-series diffusion transformer. We hope this work will help motivate a new generation of policy learning techniques that are able to leverage the powerful generative modeling capabilities of diffusion models. Code, data, and training details are available (diffusion-policy.cs.columbia.edu)."
                },
                {
                    "title": "Understanding Diffusion Objectives as the ELBO with Simple Data Augmentation",
                    "abstract": "To achieve the highest perceptual quality, state-of-the-art diffusion models are optimized with objectives that typically look very different from the maximum likelihood and the Evidence Lower Bound (ELBO) objectives. In this work, we reveal that diffusion model objectives are actually closely related to the ELBO. Specifically, we show that all commonly used diffusion model objectives equate to a weighted integral of ELBOs over different noise levels, where the weighting depends on the specific objective used. Under the condition of monotonic weighting, the connection is even closer: the diffusion objective then equals the ELBO, combined with simple data augmentation, namely Gaussian noise perturbation. We show that this condition holds for a number of state-of-the-art diffusion models. In experiments, we explore new monotonic weightings and demonstrate their effectiveness, achieving state-of-the-art FID scores on the high-resolution ImageNet benchmark."
                },
                {
                    "title": "Muse: Text-To-Image Generation via Masked Generative Transformers",
                    "abstract": "We present Muse, a text-to-image Transformer model that achieves state-of-the-art image generation performance while being significantly more efficient than diffusion or autoregressive models. Muse is trained on a masked modeling task in discrete token space: given the text embedding extracted from a pre-trained large language model (LLM), Muse is trained to predict randomly masked image tokens. Compared to pixel-space diffusion models, such as Imagen and DALL-E 2, Muse is significantly more efficient due to the use of discrete tokens and requiring fewer sampling iterations; compared to autoregressive models, such as Parti, Muse is more efficient due to the use of parallel decoding. The use of a pre-trained LLM enables fine-grained language understanding, translating to high-fidelity image generation and the understanding of visual concepts such as objects, their spatial relationships, pose, cardinality etc. Our 900M parameter model achieves a new SOTA on CC3M, with an FID score of 6.06. The Muse 3B parameter model achieves an FID of 7.88 on zero-shot COCO evaluation, along with a CLIP score of 0.32. Muse also directly enables a number of image editing applications without the need to fine-tune or invert the model: inpainting, outpainting, and mask-free editing. More results are available at https://muse-model.github.io"
                },
                {
                    "title": "Scalable Diffusion Models with Transformers",
                    "abstract": "We explore a new class of diffusion models based on the transformer architecture. We train latent diffusion models of images, replacing the commonly-used U-Net backbone with a transformer that operates on latent patches. We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops. We find that DiTs with higher Gflops\u2014through increased transformer depth/width or increased number of input tokens\u2014consistently have lower FID. In addition to possessing good scalability properties, our largest DiT-XL/2 models outperform all prior diffusion models on the class-conditional ImageNet 512\u00d7512 and 256\u00d7256 benchmarks, achieving a state-of-the-art FID of 2.27 on the latter."
                },
                {
                    "title": "MAGVIT: Masked Generative Video Transformer",
                    "abstract": "We introduce the MAsked Generative VIdeo Transformer, MAGVIT, to tackle various video synthesis tasks with a single model. We introduce a 3D tokenizer to quantize a video into spatial-temporal visual tokens and propose an embedding method for masked video token modeling to facilitate multi-task learning. We conduct extensive experiments to demonstrate the quality, efficiency, and flexibility of MAGVIT. Our experiments show that (i) MAGVIT performs favorably against state-of-the-art approaches and establishes the best-published FVD on three video generation benchmarks, including the challenging Kinetics-600. (ii) MAGVIT outperforms existing methods in inference time by two orders of magnitude against diffusion models and by 60x against autoregressive models. (iii) A single MAGVIT model supports ten diverse generation tasks and generalizes across videos from different visual domains. The source code and trained models will be released to the public at https://magvit.cs.cmu.edu."
                },
                {
                    "title": "MAGE: MAsked Generative Encoder to Unify Representation Learning and Image Synthesis",
                    "abstract": "Generative modeling and representation learning are two key tasks in computer vision. However, these models are typically trained independently, which ignores the potential for each task to help the other, and leads to training and model maintenance overheads. In this work, we propose MAsked Generative Encoder (MAGE), the first framework to unify SOTA image generation and self-supervised representation learning. Our key insight is that using variable masking ratios in masked image modeling pre-training can allow generative training (very high masking ratio) and representation learning (lower masking ratio) under the same training framework. Inspired by previous generative models, MAGE uses semantic tokens learned by a vector-quantized GAN at inputs and outputs, combining this with masking. We can further improve the representation by adding a contrastive loss to the encoder output. We extensively evaluate the generation and representation learning capabilities of MAGE. On ImageNet-1K, a single MAGE ViT-L model obtains 9.10 FID in the task of class-unconditional image generation and 78.9% top-1 accuracy for linear probing, achieving state-of-the-art performance in both image generation and representation learning. Code is available at https://github.com/LTHl4/mage."
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "UViM: A Unified Modeling Approach for Vision with Learned Guiding Codes",
                    "abstract": "We introduce UViM, a unified approach capable of modeling a wide range of computer vision tasks. In contrast to previous models, UViM has the same functional form for all tasks; it requires no task-specific modifications which require extensive human expertise. The approach involves two components: (I) a base model (feed-forward) which is trained to directly predict raw vision outputs, guided by a learned discrete code and (II) a language model (autoregressive) that is trained to generate the guiding code. These components complement each other: the language model is well-suited to modeling structured interdependent data, while the base model is efficient at dealing with high-dimensional outputs. We demonstrate the effectiveness of UViM on three diverse and challenging vision tasks: panoptic segmentation, depth prediction and image colorization, where we achieve competitive and near state-of-the-art results. Our experimental results suggest that UViM is a promising candidate for a unified modeling approach in computer vision."
                },
                {
                    "title": "MaskGIT: Masked Generative Image Transformer",
                    "abstract": "Generative transformers have experienced rapid popularity growth in the computer vision community in synthesizing high-fidelity and high-resolution images. The best generative transformer models so far, however, still treat an image naively as a sequence of tokens, and decode an image sequentially following the raster scan ordering (i.e. line-by-line). We find this strategy neither optimal nor efficient. This paper proposes a novel image synthesis paradigm using a bidirectional transformer decoder, which we term MaskGIT. During training, MaskGIT learns to predict randomly masked tokens by attending to tokens in all directions. At inference time, the model begins with generating all tokens of an image simultaneously, and then refines the image iteratively conditioned on the previous generation. Our experiments demonstrate that MaskGIT significantly outperforms the state-of-the-art transformer model on the ImageNet dataset, and accelerates autoregressive decoding by up to 48x. Besides, we illustrate that MaskGIT can be easily extended to various image editing tasks, such as inpainting, extrapolation, and image manipulation. Project page: masked-generative-image-transformer.github.io."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "Zero-Shot Text-to-Image Generation",
                    "abstract": "Text-to-image generation has traditionally focused on finding better modeling assumptions for training on a fixed dataset. These assumptions might involve complex architectures, auxiliary losses, or side information such as object part labels or segmentation masks supplied during training. We describe a simple approach for this task based on a transformer that autoregressively models the text and image tokens as a single stream of data. With sufficient data and scale, our approach is competitive with previous domain-specific models when evaluated in a zero-shot fashion."
                },
                {
                    "title": "Improved Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) are a class of generative models which have recently been shown to produce excellent samples. We show that with a few simple modifications, DDPMs can also achieve competitive log-likelihoods while maintaining high sample quality. Additionally, we find that learning variances of the reverse diffusion process allows sampling with an order of magnitude fewer forward passes with a negligible difference in sample quality, which is important for the practical deployment of these models. We additionally use precision and recall to compare how well DDPMs and GANs cover the target distribution. Finally, we show that the sample quality and likelihood of these models scale smoothly with model capacity and training compute, making them easily scalable. We release our code at https://github.com/openai/improved-diffusion"
                },
                {
                    "title": "Taming Transformers for High-Resolution Image Synthesis",
                    "abstract": "Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks. In contrast to CNNs, they contain no inductive bias that prioritizes local interactions. This makes them expressive, but also computationally infeasible for long sequences, such as high-resolution images. We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images. We show how to (i) use CNNs to learn a context-rich vocabulary of image constituents, and in turn (ii) utilize transformers to efficiently model their composition within high-resolution images. Our approach is readily applied to conditional synthesis tasks, where both non-spatial information, such as object classes, and spatial information, such as segmentations, can control the generated image. In particular, we present the first results on semantically-guided synthesis of megapixel images with transformers. Project page at https://git.io/JLlvY."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Generative Pretraining From Pixels",
                    "abstract": "Inspired by progress in unsupervised representation learning for natural language, we examine whether similar models can learn useful representations for images. We train a sequence Transformer to auto-regressively predict pixels, without incorporating knowledge of the 2D input structure. Despite training on low-resolution ImageNet without labels, we \ufb01nd that a GPT-2 scale model learns strong image representations as measured by linear probing, \ufb01ne-tuning, and low-data classi\ufb01cation. On CIFAR-10, we achieve 96.3% accuracy with a linear probe, outperforming a supervised Wide ResNet, and 99.0% accuracy with full \ufb01ne-tuning, matching the top supervised pre-trained models. An even larger model trained on a mix-ture of ImageNet and web images is competitive with self-supervised benchmarks on ImageNet, achieving 72.0% top-1 accuracy on a linear probe of our features."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Fast Transformer Decoding: One Write-Head is All You Need",
                    "abstract": "Multi-head attention layers, as used in the Transformer neural sequence model, are a powerful alternative to RNNs for moving information across and between sequences. While training these layers is generally fast and simple, due to parallelizability across the length of the sequence, incremental inference (where such paralleization is impossible) is often slow, due to the memory-bandwidth cost of repeatedly loading the large \"keys\" and \"values\" tensors. We propose a variant called multi-query attention, where the keys and values are shared across all of the different attention \"heads\", greatly reducing the size of these tensors and hence the memory bandwidth requirements of incremental decoding. We verify experimentally that the resulting models can indeed be much faster to decode, and incur only minor quality degradation from the baseline."
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "Generating Diverse High-Fidelity Images with VQ-VAE-2",
                    "abstract": "We explore the use of Vector Quantized Variational AutoEncoder (VQ-VAE) models for large scale image generation. To this end, we scale and enhance the autoregressive priors used in VQ-VAE to generate synthetic samples of much higher coherence and fidelity than possible before. We use simple feed-forward encoder and decoder networks, making our model an attractive candidate for applications where the encoding and/or decoding speed is critical. Additionally, VQ-VAE requires sampling an autoregressive model only in the compressed latent space, which is an order of magnitude faster than sampling in the pixel space, especially for large images. We demonstrate that a multi-scale hierarchical organization of VQ-VAE, augmented with powerful priors over the latent codes, is able to generate samples with quality that rivals that of state of the art Generative Adversarial Networks on multifaceted datasets such as ImageNet, while not suffering from GAN's known shortcomings such as mode collapse and lack of diversity."
                },
                {
                    "title": "Image Transformer",
                    "abstract": "Image generation has been successfully cast as an autoregressive sequence generation or transformation problem. Recent work has shown that self-attention is an effective way of modeling textual sequences. In this work, we generalize a recently proposed model architecture based on self-attention, the Transformer, to a sequence modeling formulation of image generation with a tractable likelihood. By restricting the self-attention mechanism to attend to local neighborhoods we significantly increase the size of images the model can process in practice, despite maintaining significantly larger receptive fields per layer than typical convolutional neural networks. We propose another extension of self-attention allowing it to efficiently take advantage of the two-dimensional nature of images. While conceptually simple, our generative models trained on two image data sets are competitive with or significantly outperform the current state of the art in autoregressive image generation on two different data sets, CIFAR-10 and ImageNet. We also present results on image super-resolution with a large magnification ratio, applying an encoder-decoder configuration of our architecture. In a human evaluation study, we show that our super-resolution models improve significantly over previously published autoregressive super-resolution models. Images they generate fool human observers three times more often than the previous state of the art."
                },
                {
                    "title": "The Unreasonable Effectiveness of Deep Features as a Perceptual Metric",
                    "abstract": "While it is nearly effortless for humans to quickly assess the perceptual similarity between two images, the underlying processes are thought to be quite complex. Despite this, the most widely used perceptual metrics today, such as PSNR and SSIM, are simple, shallow functions, and fail to account for many nuances of human perception. Recently, the deep learning community has found that features of the VGG network trained on ImageNet classification has been remarkably useful as a training loss for image synthesis. But how perceptual are these so-called \"perceptual losses\"? What elements are critical for their success? To answer these questions, we introduce a new dataset of human perceptual similarity judgments. We systematically evaluate deep features across different architectures and tasks and compare them with classic metrics. We find that deep features outperform all previous metrics by large margins on our dataset. More surprisingly, this result is not restricted to ImageNet-trained VGG features, but holds across different deep architectures and levels of supervision (supervised, self-supervised, or even unsupervised). Our results suggest that perceptual similarity is an emergent property shared across deep visual representations."
                },
                {
                    "title": "PixelSNAIL: An Improved Autoregressive Generative Model",
                    "abstract": "Autoregressive generative models consistently achieve the best results in density estimation tasks involving high dimensional data, such as images or audio. They pose density estimation as a sequence modeling task, where a recurrent neural network (RNN) models the conditional distribution over the next element conditioned on all previous elements. In this paradigm, the bottleneck is the extent to which the RNN can model long-range dependencies, and the most successful approaches rely on causal convolutions, which offer better access to earlier parts of the sequence than conventional RNNs. Taking inspiration from recent work in meta reinforcement learning, where dealing with long-range dependencies is also essential, we introduce a new generative model architecture that combines causal convolutions with self attention. In this note, we describe the resulting model and present state-of-the-art log-likelihood results on CIFAR-10 (2.85 bits per dim) and $32 \\times 32$ ImageNet (3.80 bits per dim). Our implementation is available at this https URL"
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Neural Discrete Representation Learning",
                    "abstract": "Learning useful representations without supervision remains a key challenge in machine learning. In this paper, we propose a simple yet powerful generative model that learns such discrete representations. Our model, the Vector Quantised-Variational AutoEncoder (VQ-VAE), differs from VAEs in two key ways: the encoder network outputs discrete, rather than continuous, codes; and the prior is learnt rather than static. In order to learn a discrete latent representation, we incorporate ideas from vector quantisation (VQ). Using the VQ method allows the model to circumvent issues of \"posterior collapse\" -- where the latents are ignored when they are paired with a powerful autoregressive decoder -- typically observed in the VAE framework. Pairing these representations with an autoregressive prior, the model can generate high quality images, videos, and speech as well as doing high quality speaker conversion and unsupervised learning of phonemes, providing further evidence of the utility of the learnt representations."
                },
                {
                    "title": "GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium",
                    "abstract": "Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the \"Frechet Inception Distance\" (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour",
                    "abstract": "Deep learning thrives with large neural networks and large datasets. However, larger networks and larger datasets result in longer training times that impede research and development progress. Distributed synchronous SGD offers a potential solution to this problem by dividing SGD minibatches over a pool of parallel workers. Yet to make this scheme efficient, the per-worker workload must be large, which implies nontrivial growth in the SGD minibatch size. In this paper, we empirically show that on the ImageNet dataset large minibatches cause optimization difficulties, but when these are addressed the trained networks exhibit good generalization. Specifically, we show no loss of accuracy when training with large minibatch sizes up to 8192 images. To achieve this result, we adopt a hyper-parameter-free linear scaling rule for adjusting learning rates as a function of minibatch size and develop a new warmup scheme that overcomes optimization challenges early in training. With these simple techniques, our Caffe2-based system trains ResNet-50 with a minibatch size of 8192 on 256 GPUs in one hour, while matching small minibatch accuracy. Using commodity hardware, our implementation achieves ~90% scaling efficiency when moving from 8 to 256 GPUs. Our findings enable training visual recognition models on internet-scale data with high efficiency."
                },
                {
                    "title": "Layer Normalization",
                    "abstract": "Training state-of-the-art, deep neural networks is computationally expensive. One way to reduce the training time is to normalize the activities of the neurons. A recently introduced technique called batch normalization uses the distribution of the summed input to a neuron over a mini-batch of training cases to compute a mean and variance which are then used to normalize the summed input to that neuron on each training case. This significantly reduces the training time in feed-forward neural networks. However, the effect of batch normalization is dependent on the mini-batch size and it is not obvious how to apply it to recurrent neural networks. In this paper, we transpose batch normalization into layer normalization by computing the mean and variance used for normalization from all of the summed inputs to the neurons in a layer on a single training case. Like batch normalization, we also give each neuron its own adaptive bias and gain which are applied after the normalization but before the non-linearity. Unlike batch normalization, layer normalization performs exactly the same computation at training and test times. It is also straightforward to apply to recurrent neural networks by computing the normalization statistics separately at each time step. Layer normalization is very effective at stabilizing the hidden state dynamics in recurrent networks. Empirically, we show that layer normalization can substantially reduce the training time compared with previously published techniques."
                },
                {
                    "title": "Conditional Image Generation with PixelCNN Decoders",
                    "abstract": "This work explores conditional image generation with a new image density model based on the PixelCNN architecture. The model can be conditioned on any vector, including descriptive labels or tags, or latent embeddings created by other networks. When conditioned on class labels from the ImageNet database, the model is able to generate diverse, realistic scenes representing distinct animals, objects, landscapes and structures. When conditioned on an embedding produced by a convolutional network given a single image of an unseen face, it generates a variety of new portraits of the same person with different facial expressions, poses and lighting conditions. We also show that conditional PixelCNN can serve as a powerful decoder in an image autoencoder. Additionally, the gated convolutional layers in the proposed model improve the log-likelihood of PixelCNN to match the state-of-the-art performance of PixelRNN on ImageNet, with greatly reduced computational cost."
                },
                {
                    "title": "Improved Techniques for Training GANs",
                    "abstract": "We present a variety of new architectural features and training procedures that we apply to the generative adversarial networks (GANs) framework. We focus on two applications of GANs: semi-supervised learning, and the generation of images that humans find visually realistic. Unlike most work on generative models, our primary goal is not to train a model that assigns high likelihood to test data, nor do we require the model to be able to learn well without using any labels. Using our new techniques, we achieve state-of-the-art results in semi-supervised classification on MNIST, CIFAR-10 and SVHN. The generated images are of high quality as confirmed by a visual Turing test: our model generates MNIST samples that humans cannot distinguish from real data, and CIFAR-10 samples that yield a human error rate of 21.3%. We also present ImageNet samples with unprecedented resolution and show that our methods enable the model to learn recognizable features of ImageNet classes."
                },
                {
                    "title": "Pixel Recurrent Neural Networks",
                    "abstract": "Modeling the distribution of natural images is a landmark problem in unsupervised learning. This task requires an image model that is at once expressive, tractable and scalable. We present a deep neural network that sequentially predicts the pixels in an image along the two spatial dimensions. Our method models the discrete probability of the raw pixel values and encodes the complete set of dependencies in the image. Architectural novelties include fast two-dimensional recurrent layers and an effective use of residual connections in deep recurrent networks. We achieve log-likelihood scores on natural images that are considerably better than the previous state of the art. Our main results also provide benchmarks on the diverse ImageNet dataset. Samples generated from the model appear crisp, varied and globally coherent."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "Deep AutoRegressive Networks",
                    "abstract": "We introduce a deep, generative autoencoder capable of learning hierarchies of distributed representations from data. Successive deep stochastic hidden layers are equipped with autoregressive connections, which enable the model to be sampled from quickly and exactly via ancestral sampling. We derive an efficient approximate parameter estimation method based on the minimum description length (MDL) principle, which can be seen as maximising a variational lower bound on the log-likelihood, with a feedforward neural network implementing approximate inference. We demonstrate state-of-the-art generative performance on a number of classic data sets, including several UCI data sets, MNIST and Atari 2600 games."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Statistical Theory of Extreme Values and Some Practical Applications",
                    "abstract": "second collects up the relevant theory that is made use of in the first. It is perhaps appropriate in a mathematical journal to concentrate attention on the second volume. This is a longish book in itself because of a desire to be selfcontained. A first chapter gives definitions, ending up by explaining why probability methods should be used in selecting sa,mples. The next chapter covers general probability theory in great detail, while chapter three deals with the standard theory of random variables, including mathematical expectation, variance and covariance, and convergence in probability. The next eight chapters give the requisite derivations and proofs in connection with the material presented in the corresponding chapters of Volume I. These are taken in the usual order of increasing complexity, beginning with simple random sampling, then moving to stratification and to the various cases of cluster sampling and multi-stage sampling. Special problems are the control of variation in size of cluster, and estimating variances, and these have chapters to themselves, while a penultimate chapter deals with regression estimates, double sampling, sampling for time series, and other methods. The last chapter of all develops a theory for response errors in surveys. lvhat can be said broadly is that t,he book is carefully compiled, well written, aut'horitative and well-documented with references. If i t is a little prolix in parts, we can a t any rate say that it needs no pre-requisite reading, so that it is altogether suitable for the beginner, as well as for someone who only requires to have a broad idea of the problems a t issue. J. WISHART."
                },
                {
                    "title": "All are Worth Words: a ViT Backbone for Score-based Diffusion Models",
                    "abstract": "Vision transformers (ViT) have shown promise in various vision tasks including low-level ones while the U-Net remains dominant in score-based diffusion models. In this paper, we perform a systematical empirical study on the ViT-based architectures in diffusion models. Our results suggest that adding extra long skip connections (like the U-Net) to ViT is crucial to diffusion models. The new ViT architecture, together with other improvements, is referred to as U-ViT. On several popular visual datasets, U-ViT achieves competitive generation results to SOTA U-Net while requiring comparable amount of parameters and computation if not less."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Improving Language Understanding by Generative Pre-Training",
                    "abstract": "Natural language understanding comprises a wide range of diverse tasks such as textual entailment, question answering, semantic similarity assessment, and document classi\ufb01cation. Although large unlabeled text corpora are abundant, labeled data for learning these speci\ufb01c tasks is scarce, making it challenging for discriminatively trained models to perform adequately. We demonstrate that large gains on these tasks can be realized by generative pre-training of a language model on a diverse corpus of unlabeled text, followed by discriminative \ufb01ne-tuning on each speci\ufb01c task. In contrast to previous approaches, we make use of task-aware input transformations during \ufb01ne-tuning to achieve effective transfer while requiring minimal changes to the model architecture. We demonstrate the effectiveness of our approach on a wide range of benchmarks for natural language understanding. Our general task-agnostic model outperforms discriminatively trained models that use architectures speci\ufb01cally crafted for each task, signi\ufb01cantly improving upon the state of the art in 9 out of the 12 tasks studied. For instance, we achieve absolute improvements of 8.9% on commonsense reasoning (Stories Cloze Test), 5.7% on question answering (RACE), and 1.5% on textual entailment (MultiNLI)."
                },
                {
                    "title": "GENERATIVE ADVERSARIAL NETS",
                    "abstract": "Estimating individualized treatment effects (ITE) is a challenging task due to the need for an individual\u2019s potential outcomes to be learned from biased data and without having access to the counterfactuals. We propose a novel method for inferring ITE based on the Generative Adversarial Nets (GANs) framework. Our method, termed Generative Adversarial Nets for inference of Individualized Treatment Effects (GANITE), is motivated by the possibility that we can capture the uncertainty in the counterfactual distributions by attempting to learn them using a GAN. We generate proxies of the counterfactual outcomes using a counterfactual generator, G, and then pass these proxies to an ITE generator, I, in order to train it. By modeling both of these using the GAN framework, we are able to infer based on the factual data, while still accounting for the unseen counterfactuals. We test our method on three real-world datasets (with both binary and multiple treatments) and show that GANITE outperforms state-of-the-art methods."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nIs it necessary for autoregressive models to be coupled with vector-quantized representations in generative modeling?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem could significantly advance the research community's understanding of generative models by demonstrating that autoregressive models can operate effectively in continuous-valued domains without relying on discrete representations. This could lead to the development of more efficient and higher-quality generative models, particularly in image generation, where current methods often suffer from the limitations of vector quantization. By addressing this question, future research could explore new methodologies for generative modeling that leverage continuous representations, potentially leading to practical applications in various fields such as computer vision, natural language processing, and beyond.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the inherent complexity of modeling per-token probability distributions in continuous-valued domains, as traditional autoregressive models are designed for discrete spaces. Naive approaches that attempt to directly apply existing autoregressive frameworks to continuous data may fail due to the lack of suitable loss functions and sampling mechanisms. Additionally, the technical obstacles include the need for a robust denoising network that can effectively learn the underlying distribution without the guidance of discrete tokenization, which has been the standard practice. Overcoming these complexities requires innovative methodologies that can bridge the gap between discrete and continuous representations.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on discretizing continuous data to fit autoregressive models, leading to a reliance on vector-quantized representations. This approach has been limited by the challenges of training discrete tokenizers and the sensitivity of these models to gradient approximation strategies. Additionally, existing solutions often fail to achieve the reconstruction quality of continuous-valued representations. Our approach differs by proposing a novel methodology that utilizes diffusion processes to model per-token probability distributions directly in continuous space, thereby eliminating the need for discrete tokenization and addressing the limitations of prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves using a diffusion procedure to model the per-token probability distribution in continuous-valued domains. We will train a small denoising network (e.g., a multi-layer perceptron) jointly with the autoregressive model, where the autoregressive model predicts a vector \\( z \\) for each token, serving as a conditioning input for the denoising network. The expected outcome is a new generative model that can sample tokens from"
            }
        },
        "author_data": {
            "bf06d50d-24bb-4724-a27b-473fcc1d3ac0": {
                "pk": "bf06d50d-24bb-4724-a27b-473fcc1d3ac0",
                "name": "Tianhong Li",
                "collaborators": [
                    "Dina Katabi",
                    "Lijie Fan",
                    "Yuan Yuan",
                    "Feimin Huang",
                    "Difan Yuan",
                    "Jianguo Li",
                    "Zhuang Liu",
                    "Changshui Zhang",
                    "Kaiming He",
                    "Huimin Yu"
                ],
                "domain": [
                    "Deep Learning",
                    "Knowledge Distillation",
                    "Unsupervised Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Few Sample Knowledge Distillation for Efficient Network Compression",
                        "abstract": "Deep neural network compression techniques such as pruning and weight tensor decomposition usually require fine-tuning to recover the prediction accuracy when the compression ratio is high. However, conventional fine-tuning suffers from the requirement of a large training set and the time-consuming training procedure. This paper proposes a novel solution for knowledge distillation from label-free few samples to realize both data efficiency and training/processing efficiency. We treat the original network as \"teacher-net\" and the compressed network as \"student-net\". A 1x1 convolution layer is added at the end of each layer block of the student-net, and we fit the block-level outputs of the student-net to the teacher-net by estimating the parameters of the added layers. We prove that the added layer can be merged without adding extra parameters and computation cost during inference. Experiments on multiple datasets and network architectures verify the method's effectiveness on student-nets obtained by various network pruning and weight decomposition methods. Our method can recover student-net's accuracy to the same level as conventional fine-tuning methods in minutes while using only 1% label-free data of the full training data."
                    },
                    {
                        "title": "Unsupervised Learning for Human Sensing Using Radio Signals",
                        "abstract": "There is a growing literature demonstrating the feasibility of using Radio Frequency (RF) signals to enable key computer vision tasks in the presence of occlusions and poor lighting. It leverages that RF signals traverse walls and occlusions to deliver through-wall pose estimation, action recognition, scene captioning, and human re-identification. However, unlike RGB datasets which can be labeled by human workers, labeling RF signals is a daunting task because such signals are not human interpretable. Yet, it is fairly easy to collect unlabelled RF signals. It would be highly beneficial to use such unlabeled RF data to learn useful representations in an unsupervised manner. Thus, in this paper, we explore the feasibility of adapting RGB-based unsupervised representation learning to RF signals. We show that while contrastive learning has emerged as the main technique for unsupervised representation learning from images and videos, such methods produce poor performance when applied to sensing humans using RF signals. In contrast, predictive unsupervised learning methods learn high-quality representations that can be used for multiple downstream RF-based sensing tasks. Our empirical results show that this approach outperforms state-of-the-art RF-based human sensing on various tasks, opening the possibility of unsupervised representation learning from this novel modality."
                    },
                    {
                        "title": "Return of Unconditional Generation: A Self-supervised Representation Generation Method",
                        "abstract": "Unconditional generation -- the problem of modeling data distribution without relying on human-annotated labels -- is a long-standing and fundamental challenge in generative models, creating a potential of learning from large-scale unlabeled data. In the literature, the generation quality of an unconditional method has been much worse than that of its conditional counterpart. This gap can be attributed to the lack of semantic information provided by labels. In this work, we show that one can close this gap by generating semantic representations in the representation space produced by a self-supervised encoder. These representations can be used to condition the image generator. This framework, called Representation-Conditioned Generation (RCG), provides an effective solution to the unconditional generation problem without using labels. Through comprehensive experiments, we observe that RCG significantly improves unconditional generation quality: e.g., it achieves a new state-of-the-art FID of 2.15 on ImageNet 256x256, largely reducing the previous best of 5.91 by a relative 64%. Our unconditional results are situated in the same tier as the leading class-conditional ones. We hope these encouraging observations will attract the community's attention to the fundamental problem of unconditional generation. Code is available at https://github.com/LTH14/rcg."
                    },
                    {
                        "title": "Global Entropy Solutions to Multi-Dimensional Isentropic Gas Dynamics with Spherical Symmetry",
                        "abstract": "We are concerned with spherically symmetric solutions to the Euler equations for the multi-dimensional compressible fluids, which have many applications in diverse real physical situations. The system can be reduced to one dimensional isentropic gas dynamics with geometric source terms. Due to the presence of the singularity at the origin, there are few papers devoted to this problem. The present paper proves two existence theorems of global entropy solutions. The first one focuses on the case excluding the origin in which the negative velocity is allowed, and the second one is corresponding to the case including the origin with non-negative velocity. The $L^\\infty$ compensated compactness framework and vanishing viscosity method are applied to prove the convergence of approximate solutions. In the second case, we show that if the blast wave initially moves outwards and the initial densities and velocities decay to zero with certain rates near origin, then the densities and velocities tend to zero with the same rates near the origin for any positive time. In particular, the entropy solutions in two existence theorems are uniformly bounded with respect to time."
                    },
                    {
                        "title": "Large time behavior of entropy solutions to one-dimensional unipolar hydrodynamic model for semiconductor devices",
                        "abstract": "We are concerned with the global existence and large time behavior of entropy solutions to the one dimensional unipolar hydrodynamic model for semiconductors in the form of Euler-Poisson equations in a bounded interval. In this paper, we first prove the global existence of entropy solution by vanishing viscosity and compensated compactness framework. In particular, the solutions are uniformly bounded with respect to space and time variables by introducing modified Riemann invariants and the theory of invariant region. Based on the uniform estimates of density, we further show that the entropy solution converges to the corresponding unique stationary solution exponentially in time. No any smallness condition is assumed on the initial data and doping profile. Moreover, the novelty in this paper is about the unform bound with respect to time for the weak solutions of the isentropic Euler-Possion system."
                    },
                    {
                        "title": "Quadratic Upper Bound for Recursive Teaching Dimension of Finite VC Classes",
                        "abstract": "In this work we study the quantitative relation between the recursive teaching dimension (RTD) and the VC dimension (VCD) of concept classes of finite sizes. The RTD of a concept class $\\mathcal C \\subseteq \\{0, 1\\}^n$, introduced by Zilles et al. (2011), is a combinatorial complexity measure characterized by the worst-case number of examples necessary to identify a concept in $\\mathcal C$ according to the recursive teaching model.   For any finite concept class $\\mathcal C \\subseteq \\{0,1\\}^n$ with $\\mathrm{VCD}(\\mathcal C)=d$, Simon & Zilles (2015) posed an open problem $\\mathrm{RTD}(\\mathcal C) = O(d)$, i.e., is RTD linearly upper bounded by VCD? Previously, the best known result is an exponential upper bound $\\mathrm{RTD}(\\mathcal C) = O(d \\cdot 2^d)$, due to Chen et al. (2016). In this paper, we show a quadratic upper bound: $\\mathrm{RTD}(\\mathcal C) = O(d^2)$, much closer to an answer to the open problem. We also discuss the challenges in fully solving the problem."
                    },
                    {
                        "title": "In-Home Daily-Life Captioning Using Radio Signals",
                        "abstract": "This paper aims to caption daily life --i.e., to create a textual description of people's activities and interactions with objects in their homes. Addressing this problem requires novel methods beyond traditional video captioning, as most people would have privacy concerns about deploying cameras throughout their homes. We introduce RF-Diary, a new model for captioning daily life by analyzing the privacy-preserving radio signal in the home with the home's floormap. RF-Diary can further observe and caption people's life through walls and occlusions and in dark settings. In designing RF-Diary, we exploit the ability of radio signals to capture people's 3D dynamics, and use the floormap to help the model learn people's interactions with objects. We also use a multi-modal feature alignment training scheme that leverages existing video-based captioning datasets to improve the performance of our radio-based captioning model. Extensive experimental results demonstrate that RF-Diary generates accurate captions under visible conditions. It also sustains its good performance in dark or occluded settings, where video-based captioning approaches fail to generate meaningful captions. For more information, please visit our project webpage: http://rf-diary.csail.mit.edu"
                    },
                    {
                        "title": "Making the Invisible Visible: Action Recognition Through Walls and Occlusions",
                        "abstract": "Understanding people's actions and interactions typically depends on seeing them. Automating the process of action recognition from visual data has been the topic of much research in the computer vision community. But what if it is too dark, or if the person is occluded or behind a wall? In this paper, we introduce a neural network model that can detect human actions through walls and occlusions, and in poor lighting conditions. Our model takes radio frequency (RF) signals as input, generates 3D human skeletons as an intermediate representation, and recognizes actions and interactions of multiple people over time. By translating the input to an intermediate skeleton-based representation, our model can learn from both vision-based and RF-based datasets, and allow the two tasks to help each other. We show that our model achieves comparable accuracy to vision-based action recognition systems in visible scenarios, yet continues to work accurately when people are not visible, hence addressing scenarios that are beyond the limit of today's vision-based action recognition."
                    },
                    {
                        "title": "Reparo: Loss-Resilient Generative Codec for Video Conferencing",
                        "abstract": "Packet loss during video conferencing often results in poor quality and video freezing. Retransmitting lost packets is often impractical due to the need for real-time playback, and using Forward Error Correction (FEC) for packet recovery is challenging due to the unpredictable and bursty nature of Internet losses. Excessive redundancy leads to inefficiency and wasted bandwidth, while insufficient redundancy results in undecodable frames, causing video freezes and quality degradation in subsequent frames.   We introduce Reparo -- a loss-resilient video conferencing framework based on generative deep learning models to address these issues. Our approach generates missing information when a frame or part of a frame is lost. This generation is conditioned on the data received thus far, considering the model's understanding of how people and objects appear and interact within the visual realm. Experimental results, using publicly available video conferencing datasets, demonstrate that Reparo outperforms state-of-the-art FEC-based video conferencing solutions in terms of both video quality (measured through PSNR, SSIM, and LPIPS) and the occurrence of video freezes."
                    }
                ]
            },
            "a5e9e016-fac2-47ac-852f-b7df30eb3d4a": {
                "pk": "a5e9e016-fac2-47ac-852f-b7df30eb3d4a",
                "name": "Yonglong Tian",
                "collaborators": [
                    "Phillip Isola",
                    "Dilip Krishnan",
                    "Lijie Fan",
                    "Chen Sun",
                    "Cordelia Schmid",
                    "Tianyu Hua",
                    "Sucheng Ren",
                    "Hang Zhao",
                    "Jindong Gu",
                    "Wei Liu"
                ],
                "domain": [
                    "Self-Supervised Learning",
                    "Contrastive Learning",
                    "Representation Learning",
                    "Knowledge Distillation"
                ],
                "publications": [
                    {
                        "title": "Contrastive Representation Distillation",
                        "abstract": "Often we wish to transfer representational knowledge from one neural network to another. Examples include distilling a large network into a smaller one, transferring knowledge from one sensory modality to a second, or ensembling a collection of models into a single estimator. Knowledge distillation, the standard approach to these problems, minimizes the KL divergence between the probabilistic outputs of a teacher and student network. We demonstrate that this objective ignores important structural knowledge of the teacher network. This motivates an alternative objective by which we train a student to capture significantly more information in the teacher's representation of the data. We formulate this objective as contrastive learning. Experiments demonstrate that our resulting new objective outperforms knowledge distillation and other cutting-edge distillers on a variety of knowledge transfer tasks, including single model compression, ensemble distillation, and cross-modal transfer. Our method sets a new state-of-the-art in many transfer tasks, and sometimes even outperforms the teacher network when combined with knowledge distillation. Code: http://github.com/HobbitLong/RepDistiller."
                    },
                    {
                        "title": "Simple Distillation Baselines for Improving Small Self-supervised Models",
                        "abstract": "While large self-supervised models have rivalled the performance of their supervised counterparts, small models still struggle. In this report, we explore simple baselines for improving small self-supervised models via distillation, called SimDis. Specifically, we present an offline-distillation baseline, which establishes a new state-of-the-art, and an online-distillation baseline, which achieves similar performance with minimal computational overhead. We hope these baselines will provide useful experience for relevant future research. Code is available at: https://github.com/JindongGu/SimDis/"
                    },
                    {
                        "title": "Pedestrian Detection aided by Deep Learning Semantic Tasks",
                        "abstract": "Deep learning methods have achieved great success in pedestrian detection, owing to its ability to learn features from raw pixels. However, they mainly capture middle-level representations, such as pose of pedestrian, but confuse positive with hard negative samples, which have large ambiguity, e.g. the shape and appearance of `tree trunk' or `wire pole' are similar to pedestrian in certain viewpoint. This ambiguity can be distinguished by high-level representation. To this end, this work jointly optimizes pedestrian detection with semantic tasks, including pedestrian attributes (e.g. `carrying backpack') and scene attributes (e.g. `road', `tree', and `horizontal'). Rather than expensively annotating scene attributes, we transfer attributes information from existing scene segmentation datasets to the pedestrian dataset, by proposing a novel deep model to learn high-level features from multiple tasks and multiple data sources. Since distinct tasks have distinct convergence rates and data from different datasets have different distributions, a multi-task objective function is carefully designed to coordinate tasks and reduce discrepancies among datasets. The importance coefficients of tasks and network parameters in this objective function can be iteratively estimated. Extensive evaluations show that the proposed approach outperforms the state-of-the-art on the challenging Caltech and ETH datasets, where it reduces the miss rates of previous deep models by 17 and 5.5 percent, respectively."
                    },
                    {
                        "title": "Contrastive Multiview Coding",
                        "abstract": "Humans view the world through many sensory channels, e.g., the long-wavelength light channel, viewed by the left eye, or the high-frequency vibrations channel, heard by the right ear. Each view is noisy and incomplete, but important factors, such as physics, geometry, and semantics, tend to be shared between all views (e.g., a \"dog\" can be seen, heard, and felt). We investigate the classic hypothesis that a powerful representation is one that models view-invariant factors. We study this hypothesis under the framework of multiview contrastive learning, where we learn a representation that aims to maximize mutual information between different views of the same scene but is otherwise compact. Our approach scales to any number of views, and is view-agnostic. We analyze key properties of the approach that make it work, finding that the contrastive loss outperforms a popular alternative based on cross-view prediction, and that the more views we learn from, the better the resulting representation captures underlying scene semantics. Our approach achieves state-of-the-art results on image and video unsupervised learning benchmarks. Code is released at: http://github.com/HobbitLong/CMC/."
                    },
                    {
                        "title": "Generative Models as a Data Source for Multiview Representation Learning",
                        "abstract": "Generative models are now capable of producing highly realistic images that look nearly indistinguishable from the data on which they are trained. This raises the question: if we have good enough generative models, do we still need datasets? We investigate this question in the setting of learning general-purpose visual representations from a black-box generative model rather than directly from data. Given an off-the-shelf image generator without any access to its training data, we train representations from the samples output by this generator. We compare several representation learning methods that can be applied to this setting, using the latent space of the generator to generate multiple \"views\" of the same semantic content. We show that for contrastive methods, this multiview data can naturally be used to identify positive pairs (nearby in latent space) and negative pairs (far apart in latent space). We find that the resulting representations rival or even outperform those learned directly from real data, but that good performance requires care in the sampling strategy applied and the training method. Generative models can be viewed as a compressed and organized copy of a dataset, and we envision a future where more and more \"model zoos\" proliferate while datasets become increasingly unwieldy, missing, or private. This paper suggests several techniques for dealing with visual representation learning in such a future. Code is available on our project page https://ali-design.github.io/GenRep/."
                    },
                    {
                        "title": "StableRep: Synthetic Images from Text-to-Image Models Make Strong Visual Representation Learners",
                        "abstract": "We investigate the potential of learning visual representations using synthetic images generated by text-to-image models. This is a natural question in the light of the excellent performance of such models in generating high-quality images. We consider specifically the Stable Diffusion, one of the leading open source text-to-image models. We show that (1) when the generative model is configured with proper classifier-free guidance scale, training self-supervised methods on synthetic images can match or beat the real image counterpart; (2) by treating the multiple images generated from the same text prompt as positives for each other, we develop a multi-positive contrastive learning method, which we call StableRep. With solely synthetic images, the representations learned by StableRep surpass the performance of representations learned by SimCLR and CLIP using the same set of text prompts and corresponding real images, on large scale datasets. When we further add language supervision, StableRep trained with 20M synthetic images achieves better accuracy than CLIP trained with 50M real images."
                    },
                    {
                        "title": "Composable Augmentation Encoding for Video Representation Learning",
                        "abstract": "We focus on contrastive methods for self-supervised video representation learning. A common paradigm in contrastive learning is to construct positive pairs by sampling different data views for the same instance, with different data instances as negatives. These methods implicitly assume a set of representational invariances to the view selection mechanism (eg, sampling frames with temporal shifts), which may lead to poor performance on downstream tasks which violate these invariances (fine-grained video action recognition that would benefit from temporal information). To overcome this limitation, we propose an 'augmentation aware' contrastive learning framework, where we explicitly provide a sequence of augmentation parameterisations (such as the values of the time shifts used to create data views) as composable augmentation encodings (CATE) to our model when projecting the video representations for contrastive learning. We show that representations learned by our method encode valuable information about specified spatial or temporal augmentation, and in doing so also achieve state-of-the-art performance on a number of video benchmarks."
                    },
                    {
                        "title": "Training-Free Uncertainty Estimation for Dense Regression: Sensitivity as a Surrogate",
                        "abstract": "Uncertainty estimation is an essential step in the evaluation of the robustness for deep learning models in computer vision, especially when applied in risk-sensitive areas. However, most state-of-the-art deep learning models either fail to obtain uncertainty estimation or need significant modification (e.g., formulating a proper Bayesian treatment) to obtain it. Most previous methods are not able to take an arbitrary model off the shelf and generate uncertainty estimation without retraining or redesigning it. To address this gap, we perform a systematic exploration into training-free uncertainty estimation for dense regression, an unrecognized yet important problem, and provide a theoretical construction justifying such estimations. We propose three simple and scalable methods to analyze the variance of outputs from a trained network under tolerable perturbations: infer-transformation, infer-noise, and infer-dropout. They operate solely during the inference, without the need to re-train, re-design, or fine-tune the models, as typically required by state-of-the-art uncertainty estimation methods. Surprisingly, even without involving such perturbations in training, our methods produce comparable or even better uncertainty estimation when compared to training-required state-of-the-art methods."
                    },
                    {
                        "title": "Improving CLIP Training with Language Rewrites",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) stands as one of the most effective and scalable methods for training transferable vision models using paired image and text data. CLIP models are trained using contrastive loss, which typically relies on data augmentations to prevent overfitting and shortcuts. However, in the CLIP training paradigm, data augmentations are exclusively applied to image inputs, while language inputs remain unchanged throughout the entire training process, limiting the exposure of diverse texts to the same image. In this paper, we introduce Language augmented CLIP (LaCLIP), a simple yet highly effective approach to enhance CLIP training through language rewrites. Leveraging the in-context learning capability of large language models, we rewrite the text descriptions associated with each image. These rewritten texts exhibit diversity in sentence structure and vocabulary while preserving the original key concepts and meanings. During training, LaCLIP randomly selects either the original texts or the rewritten versions as text augmentations for each image. Extensive experiments on CC3M, CC12M, RedCaps and LAION-400M datasets show that CLIP pre-training with language rewrites significantly improves the transfer performance without computation or memory overhead during training. Specifically for ImageNet zero-shot accuracy, LaCLIP outperforms CLIP by 8.2% on CC12M and 2.4% on LAION-400M. Code is available at https://github.com/LijieFan/LaCLIP."
                    },
                    {
                        "title": "Does Learning from Decentralized Non-IID Unlabeled Data Benefit from Self Supervision?",
                        "abstract": "Decentralized learning has been advocated and widely deployed to make efficient use of distributed datasets, with an extensive focus on supervised learning (SL) problems. Unfortunately, the majority of real-world data are unlabeled and can be highly heterogeneous across sources. In this work, we carefully study decentralized learning with unlabeled data through the lens of self-supervised learning (SSL), specifically contrastive visual representation learning. We study the effectiveness of a range of contrastive learning algorithms under decentralized learning settings, on relatively large-scale datasets including ImageNet-100, MS-COCO, and a new real-world robotic warehouse dataset. Our experiments show that the decentralized SSL (Dec-SSL) approach is robust to the heterogeneity of decentralized datasets, and learns useful representation for object classification, detection, and segmentation tasks. This robustness makes it possible to significantly reduce communication and reduce the participation ratio of data sources with only minimal drops in performance. Interestingly, using the same amount of data, the representation learned by Dec-SSL can not only perform on par with that learned by centralized SSL which requires communication and excessive data storage costs, but also sometimes outperform representations extracted from decentralized SL which requires extra knowledge about the data labels. Finally, we provide theoretical insights into understanding why data heterogeneity is less of a concern for Dec-SSL objectives, and introduce feature alignment and clustering techniques to develop a new Dec-SSL algorithm that further improves the performance, in the face of highly non-IID data. Our study presents positive evidence to embrace unlabeled data in decentralized learning, and we hope to provide new insights into whether and why decentralized SSL is effective."
                    },
                    {
                        "title": "What Makes for Good Views for Contrastive Learning?",
                        "abstract": "Contrastive learning between multiple views of the data has recently achieved state of the art performance in the field of self-supervised representation learning. Despite its success, the influence of different view choices has been less studied. In this paper, we use theoretical and empirical analysis to better understand the importance of view selection, and argue that we should reduce the mutual information (MI) between views while keeping task-relevant information intact. To verify this hypothesis, we devise unsupervised and semi-supervised frameworks that learn effective views by aiming to reduce their MI. We also consider data augmentation as a way to reduce MI, and show that increasing data augmentation indeed leads to decreasing MI and improves downstream classification accuracy. As a by-product, we achieve a new state-of-the-art accuracy on unsupervised pre-training for ImageNet classification ($73\\%$ top-1 linear readout with a ResNet-50). In addition, transferring our models to PASCAL VOC object detection and COCO instance segmentation consistently outperforms supervised pre-training. Code:http://github.com/HobbitLong/PyContrast"
                    },
                    {
                        "title": "Rethinking Few-Shot Image Classification: a Good Embedding Is All You Need?",
                        "abstract": "The focus of recent meta-learning research has been on the development of learning algorithms that can quickly adapt to test time tasks with limited data and low computational cost. Few-shot learning is widely used as one of the standard benchmarks in meta-learning. In this work, we show that a simple baseline: learning a supervised or self-supervised representation on the meta-training set, followed by training a linear classifier on top of this representation, outperforms state-of-the-art few-shot learning methods. An additional boost can be achieved through the use of self-distillation. This demonstrates that using a good learned embedding model can be more effective than sophisticated meta-learning algorithms. We believe that our findings motivate a rethinking of few-shot image classification benchmarks and the associated role of meta-learning algorithms. Code is available at: http://github.com/WangYueFt/rfs/."
                    },
                    {
                        "title": "Divide and Contrast: Self-supervised Learning from Uncurated Data",
                        "abstract": "Self-supervised learning holds promise in leveraging large amounts of unlabeled data, however much of its progress has thus far been limited to highly curated pre-training data such as ImageNet. We explore the effects of contrastive learning from larger, less-curated image datasets such as YFCC, and find there is indeed a large difference in the resulting representation quality. We hypothesize that this curation gap is due to a shift in the distribution of image classes -- which is more diverse and heavy-tailed -- resulting in less relevant negative samples to learn from. We test this hypothesis with a new approach, Divide and Contrast (DnC), which alternates between contrastive learning and clustering-based hard negative mining. When pretrained on less curated datasets, DnC greatly improves the performance of self-supervised learning on downstream tasks, while remaining competitive with the current state-of-the-art on curated datasets."
                    },
                    {
                        "title": "Representation Learning on Graphs with Jumping Knowledge Networks",
                        "abstract": "Recent deep learning approaches for representation learning on graphs follow a neighborhood aggregation procedure. We analyze some important properties of these models, and propose a strategy to overcome those. In particular, the range of \"neighboring\" nodes that a node's representation draws from strongly depends on the graph structure, analogous to the spread of a random walk. To adapt to local neighborhood properties and tasks, we explore an architecture -- jumping knowledge (JK) networks -- that flexibly leverages, for each node, different neighborhood ranges to enable better structure-aware representation. In a number of experiments on social, bioinformatics and citation networks, we demonstrate that our model achieves state-of-the-art performance. Furthermore, combining the JK framework with models like Graph Convolutional Networks, GraphSAGE and Graph Attention Networks consistently improves those models' performance."
                    },
                    {
                        "title": "Self-supervision through Random Segments with Autoregressive Coding (RandSAC)",
                        "abstract": "Inspired by the success of self-supervised autoregressive representation learning in natural language (GPT and its variants), and advances in recent visual architecture design with Vision Transformers (ViTs), in this paper, we explore the effect various design choices have on the success of applying such training strategies for visual feature learning. Specifically, we introduce a novel strategy that we call Random Segments with Autoregressive Coding (RandSAC). In RandSAC, we group patch representations (image tokens) into hierarchically arranged segments; within each segment, tokens are predicted in parallel, similar to BERT, while across segment predictions are sequential, similar to GPT. We illustrate that randomized serialization of the segments significantly improves the performance and results in distribution over spatially-long (across-segments) and -short (within-segment) predictions which are effective for feature learning. We illustrate the pertinence of these design choices and explore alternatives on a number of datasets (e.g., CIFAR10, CIFAR100, ImageNet). While our pre-training strategy works with a vanilla Transformer, we also propose a conceptually simple, but highly effective, addition to the decoder that allows learnable skip-connections to encoder$'$s feature layers, which further improves the performance."
                    },
                    {
                        "title": "Co-advise: Cross Inductive Bias Distillation",
                        "abstract": "Transformers recently are adapted from the community of natural language processing as a promising substitute of convolution-based neural networks for visual learning tasks. However, its supremacy degenerates given an insufficient amount of training data (e.g., ImageNet). To make it into practical utility, we propose a novel distillation-based method to train vision transformers. Unlike previous works, where merely heavy convolution-based teachers are provided, we introduce lightweight teachers with different architectural inductive biases (e.g., convolution and involution) to co-advise the student transformer. The key is that teachers with different inductive biases attain different knowledge despite that they are trained on the same dataset, and such different knowledge compounds and boosts the student's performance during distillation. Equipped with this cross inductive bias distillation method, our vision transformers (termed as CivT) outperform all previous transformers of the same architecture on ImageNet."
                    }
                ]
            },
            "45151790-d164-4776-b45f-67eb5702ec73": {
                "pk": "45151790-d164-4776-b45f-67eb5702ec73",
                "name": "He Li",
                "collaborators": [],
                "domain": [
                    "Molecular Dynamics",
                    "Biophysics",
                    "Membrane Simulation"
                ],
                "publications": [
                    {
                        "title": "Coarse-Grained Molecular Dynamics Modeling of Defective Erythrocyte Membrane and Sickle Hemoglobin Fibers",
                        "abstract": "In this work, we develop a two-component coarse-grained molecular dynamics (CGMD) model for simulating the erythrocyte membrane. This proposed model possesses the key feature of combing the lipid bilayer and the erythrocyte cytoskeleton, thus showing both the fluidic behavior of the lipid bilayer and elastic properties of the erythrocyte cytoskeleton. The proposed model facilitates simulations that span large length-scales (~ micrometer) and time-scales (~ ms). By tuning the interaction potential parameters, diffusivity and bending rigidity of the membrane model can be controlled. In the study of the membrane under shearing, we find that in low shear strain rate, the developed shear stress is mainly due to the spectrin network, while the viscosity of the lipid bilayer contributes to the resulting shear stress at higher strain rates. In addition, the effects of the reduction of the spectrin network connectivity on the shear modulus of the membrane are investigated."
                    }
                ]
            },
            "ec01f0a0-86ee-49e6-bd08-15afe719b623": {
                "pk": "ec01f0a0-86ee-49e6-bd08-15afe719b623",
                "name": "Mingyang Deng",
                "collaborators": [
                    "Xiao Mao",
                    "Ziqian Zhong",
                    "Lucas Tao",
                    "Joe Benton",
                    "Jonathan Tidor",
                    "Yufei Zhao",
                    "Ce Jin",
                    "Yilun Xu",
                    "Xiang Cheng",
                    "Yonglong Tian"
                ],
                "domain": [
                    "Algorithm Design",
                    "Combinatorial Optimization",
                    "Machine Learning",
                    "Additive Combinatorics"
                ],
                "publications": [
                    {
                        "title": "On Problems Related to Unbounded SubsetSum: A Unified Combinatorial Approach",
                        "abstract": "Unbounded SubsetSum is a classical textbook problem: given integers $w_1,w_2,\\cdots,w_n\\in [1,u],~c,u$, we need to find if there exists $m_1,m_2,\\cdots,m_n\\in \\mathbb{N}$ satisfying $c=\\sum_{i=1}^n w_im_i$. In its all-target version, $t\\in \\mathbb{Z}_+$ is given and answer for all integers $c\\in[0,t]$ is required. In this paper, we study three generalizations of this simple problem: All-Target Unbounded Knapsack, All-Target CoinChange and Residue Table. By new combinatorial insights into the structures of solutions, we present a novel two-phase approach for such problems. As a result, we present the first near-linear algorithms for CoinChange and Residue Table, which runs in $\\tilde{O}(u+t)$ and $\\tilde{O}(u)$ time deterministically. We also show if we can compute $(\\min,+)$ convolution for $n$-length arrays in $T(n)$ time, then All-Target Unbounded Knapsack can be solved in $\\tilde{O}(T(u)+t)$ time, thus establishing sub-quadratic equivalence between All-Target Unbounded Knapsack and $(\\min,+)$ convolution."
                    },
                    {
                        "title": "Measuring Feature Sparsity in Language Models",
                        "abstract": "Recent works have proposed that activations in language models can be modelled as sparse linear combinations of vectors corresponding to features of input text. Under this assumption, these works aimed to reconstruct feature directions using sparse coding. We develop metrics to assess the success of these sparse coding techniques and test the validity of the linearity and sparsity assumptions. We show our metrics can predict the level of sparsity on synthetic sparse linear activations, and can distinguish between sparse linear data and several other distributions. We use our metrics to measure levels of sparsity in several language models. We find evidence that language model activations can be accurately modelled by sparse linear combinations of features, significantly more so than control datasets. We also show that model activations appear to be sparsest in the first and final layers."
                    },
                    {
                        "title": "Uniform sets with few progressions via colorings",
                        "abstract": "Ruzsa asked whether there exist Fourier-uniform subsets of $\\mathbb Z/N\\mathbb Z$ with density $\\alpha$ and 4-term arithmetic progression (4-APs) density at most $\\alpha^C$, for arbitrarily large $C$. Gowers constructed Fourier uniform sets with density $\\alpha$ and 4-AP density at most $\\alpha^{4+c}$ for some small constant $c>0$. We show that an affirmative answer to Ruzsa's question would follow from the existence of an $N^{o(1)}$-coloring of $[N]$ without symmetrically colored 4-APs. For a broad and natural class of constructions of Fourier-uniform subsets of $\\mathbb Z/N\\mathbb Z$, we show that Ruzsa's question is equivalent to our arithmetic Ramsey question.   We prove analogous results for all even-length APs. For each odd $k\\geq 5$, we show that there exist $U^{k-2}$-uniform subsets of $\\mathbb Z/N\\mathbb Z$ with density $\\alpha$ and $k$-AP density at most $\\alpha^{c_k \\log(1/\\alpha)}$. We also prove generalizations to arbitrary one-dimensional patterns."
                    },
                    {
                        "title": "Approximating Knapsack and Partition via Dense Subset Sums",
                        "abstract": "Knapsack and Partition are two important additive problems whose fine-grained complexities in the $(1-\\varepsilon)$-approximation setting are not yet settled. In this work, we make progress on both problems by giving improved algorithms.   - Knapsack can be $(1 - \\varepsilon)$-approximated in $\\tilde O(n + (1/\\varepsilon) ^ {2.2} )$ time, improving the previous $\\tilde O(n + (1/\\varepsilon) ^ {2.25} )$ by Jin (ICALP'19). There is a known conditional lower bound of $(n+\\varepsilon)^{2-o(1)}$ based on $(\\min,+)$-convolution hypothesis.   - Partition can be $(1 - \\varepsilon)$-approximated in $\\tilde O(n + (1/\\varepsilon) ^ {1.25} )$ time, improving the previous $\\tilde O(n + (1/\\varepsilon) ^ {1.5} )$ by Bringmann and Nakos (SODA'21). There is a known conditional lower bound of $(1/\\varepsilon)^{1-o(1)}$ based on Strong Exponential Time Hypothesis.   Both of our new algorithms apply the additive combinatorial results on dense subset sums by Galil and Margalit (SICOMP'91), Bringmann and Wellnitz (SODA'21). Such techniques have not been explored in the context of Knapsack prior to our work. In addition, we design several new methods to speed up the divide-and-conquer steps which naturally arise in solving additive problems."
                    },
                    {
                        "title": "Restart Sampling for Improving Generative Processes",
                        "abstract": "Generative processes that involve solving differential equations, such as diffusion models, frequently necessitate balancing speed and quality. ODE-based samplers are fast but plateau in performance while SDE-based samplers deliver higher sample quality at the cost of increased sampling time. We attribute this difference to sampling errors: ODE-samplers involve smaller discretization errors while stochasticity in SDE contracts accumulated errors. Based on these findings, we propose a novel sampling algorithm called Restart in order to better balance discretization errors and contraction. The sampling method alternates between adding substantial noise in additional forward steps and strictly following a backward ODE. Empirically, Restart sampler surpasses previous SDE and ODE samplers in both speed and accuracy. Restart not only outperforms the previous best SDE results, but also accelerates the sampling speed by 10-fold / 2-fold on CIFAR-10 / ImageNet $64 \\times 64$. In addition, it attains significantly better sample quality than ODE samplers within comparable sampling times. Moreover, Restart better balances text-image alignment/visual quality versus diversity than previous samplers in the large-scale text-to-image Stable Diffusion model pre-trained on LAION $512 \\times 512$. Code is available at https://github.com/Newbeeer/diffusion_restart_sampling"
                    }
                ]
            },
            "ac117117-b02d-44c6-b614-ca18b02309ef": {
                "pk": "ac117117-b02d-44c6-b614-ca18b02309ef",
                "name": "Kaiming He",
                "collaborators": [
                    "Jian Sun",
                    "Ross Girshick",
                    "Xiangyu Zhang",
                    "Shaoqing Ren",
                    "Jifeng Dai",
                    "Piotr Doll\u00e1r",
                    "Xinlei Chen",
                    "Georgia Gkioxari",
                    "Yuxin Wu",
                    "Zhuang Liu"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Object Detection",
                    "Semantic Segmentation"
                ],
                "publications": [
                    {
                        "title": "Fast Guided Filter",
                        "abstract": "The guided filter is a technique for edge-aware image filtering. Because of its nice visual quality, fast speed, and ease of implementation, the guided filter has witnessed various applications in real products, such as image editing apps in phones and stereo reconstruction, and has been included in official MATLAB and OpenCV. In this note, we remind that the guided filter can be simply sped up from O(N) time to O(N/s^2) time for a subsampling ratio s. In a variety of applications, this leads to a speedup of >10x with almost no visible degradation. We hope this acceleration will improve performance of current applications and further popularize this filter. Code is released."
                    },
                    {
                        "title": "Convolutional Neural Networks at Constrained Time Cost",
                        "abstract": "Though recent advanced convolutional neural networks (CNNs) have been improving the image recognition accuracy, the models are getting more complex and time-consuming. For real-world applications in industrial and commercial scenarios, engineers and developers are often faced with the requirement of constrained time budget. In this paper, we investigate the accuracy of CNNs under constrained time cost. Under this constraint, the designs of the network architectures should exhibit as trade-offs among the factors like depth, numbers of filters, filter sizes, etc. With a series of controlled comparisons, we progressively modify a baseline model while preserving its time complexity. This is also helpful for understanding the importance of the factors in network designs. We present an architecture that achieves very competitive accuracy in the ImageNet dataset (11.8% top-5 error, 10-view test), yet is 20% faster than \"AlexNet\" (16.0% top-5 error, 10-view test)."
                    },
                    {
                        "title": "Group Normalization",
                        "abstract": "Batch Normalization (BN) is a milestone technique in the development of deep learning, enabling various networks to train. However, normalizing along the batch dimension introduces problems --- BN's error increases rapidly when the batch size becomes smaller, caused by inaccurate batch statistics estimation. This limits BN's usage for training larger models and transferring features to computer vision tasks including detection, segmentation, and video, which require small batches constrained by memory consumption. In this paper, we present Group Normalization (GN) as a simple alternative to BN. GN divides the channels into groups and computes within each group the mean and variance for normalization. GN's computation is independent of batch sizes, and its accuracy is stable in a wide range of batch sizes. On ResNet-50 trained in ImageNet, GN has 10.6% lower error than its BN counterpart when using a batch size of 2; when using typical batch sizes, GN is comparably good with BN and outperforms other normalization variants. Moreover, GN can be naturally transferred from pre-training to fine-tuning. GN can outperform its BN-based counterparts for object detection and segmentation in COCO, and for video classification in Kinetics, showing that GN can effectively replace the powerful BN in a variety of tasks. GN can be easily implemented by a few lines of code in modern libraries."
                    },
                    {
                        "title": "Exploring Simple Siamese Representation Learning",
                        "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code will be made available."
                    },
                    {
                        "title": "A Decade's Battle on Dataset Bias: Are We There Yet?",
                        "abstract": "We revisit the \"dataset classification\" experiment suggested by Torralba and Efros a decade ago, in the new era with large-scale, diverse, and hopefully less biased datasets as well as more capable neural network architectures. Surprisingly, we observe that modern neural networks can achieve excellent accuracy in classifying which dataset an image is from: e.g., we report 84.7% accuracy on held-out validation data for the three-way classification problem consisting of the YFCC, CC, and DataComp datasets. Our further experiments show that such a dataset classifier could learn semantic features that are generalizable and transferable, which cannot be simply explained by memorization. We hope our discovery will inspire the community to rethink the issue involving dataset bias and model capabilities."
                    },
                    {
                        "title": "Convolutional Feature Masking for Joint Object and Stuff Segmentation",
                        "abstract": "The topic of semantic segmentation has witnessed considerable progress due to the powerful features learned by convolutional neural networks (CNNs). The current leading approaches for semantic segmentation exploit shape information by extracting CNN features from masked image regions. This strategy introduces artificial boundaries on the images and may impact the quality of the extracted features. Besides, the operations on the raw image domain require to compute thousands of networks on a single image, which is time-consuming. In this paper, we propose to exploit shape information via masking convolutional features. The proposal segments (e.g., super-pixels) are treated as masks on the convolutional feature maps. The CNN features of segments are directly masked out from these maps and used to train classifiers for recognition. We further propose a joint method to handle objects and \"stuff\" (e.g., grass, sky, water) in the same framework. State-of-the-art results are demonstrated on benchmarks of PASCAL VOC and new PASCAL-CONTEXT, with a compelling computational speed."
                    },
                    {
                        "title": "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification",
                        "abstract": "Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on our PReLU networks (PReLU-nets), we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66%). To our knowledge, our result is the first to surpass human-level performance (5.1%, Russakovsky et al.) on this visual recognition challenge."
                    },
                    {
                        "title": "Deep Residual Learning for Image Recognition",
                        "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers---8x deeper than VGG nets but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers.   The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                    },
                    {
                        "title": "Instance-aware Semantic Segmentation via Multi-task Network Cascades",
                        "abstract": "Semantic segmentation research has recently witnessed rapid progress, but many leading methods are unable to identify object instances. In this paper, we present Multi-task Network Cascades for instance-aware semantic segmentation. Our model consists of three networks, respectively differentiating instances, estimating masks, and categorizing objects. These networks form a cascaded structure, and are designed to share their convolutional features. We develop an algorithm for the nontrivial end-to-end training of this causal, cascaded structure. Our solution is a clean, single-step training framework and can be generalized to cascades that have more stages. We demonstrate state-of-the-art instance-aware semantic segmentation accuracy on PASCAL VOC. Meanwhile, our method takes only 360ms testing an image using VGG-16, which is two orders of magnitude faster than previous systems for this challenging problem. As a by product, our method also achieves compelling object detection results which surpass the competitive Fast/Faster R-CNN systems.   The method described in this paper is the foundation of our submissions to the MS COCO 2015 segmentation competition, where we won the 1st place."
                    },
                    {
                        "title": "Identity Mappings in Deep Residual Networks",
                        "abstract": "Deep residual networks have emerged as a family of extremely deep architectures showing compelling accuracy and nice convergence behaviors. In this paper, we analyze the propagation formulations behind the residual building blocks, which suggest that the forward and backward signals can be directly propagated from one block to any other block, when using identity mappings as the skip connections and after-addition activation. A series of ablation experiments support the importance of these identity mappings. This motivates us to propose a new residual unit, which makes training easier and improves generalization. We report improved results using a 1001-layer ResNet on CIFAR-10 (4.62% error) and CIFAR-100, and a 200-layer ResNet on ImageNet. Code is available at: https://github.com/KaimingHe/resnet-1k-layers"
                    },
                    {
                        "title": "Rethinking ImageNet Pre-training",
                        "abstract": "We report competitive results on object detection and instance segmentation on the COCO dataset using standard models trained from random initialization. The results are no worse than their ImageNet pre-training counterparts even when using the hyper-parameters of the baseline system (Mask R-CNN) that were optimized for fine-tuning pre-trained models, with the sole exception of increasing the number of training iterations so the randomly initialized models may converge. Training from random initialization is surprisingly robust; our results hold even when: (i) using only 10% of the training data, (ii) for deeper and wider models, and (iii) for multiple tasks and metrics. Experiments show that ImageNet pre-training speeds up convergence early in training, but does not necessarily provide regularization or improve final target task accuracy. To push the envelope we demonstrate 50.9 AP on COCO object detection without using any external data---a result on par with the top COCO 2017 competition results that used ImageNet pre-training. These observations challenge the conventional wisdom of ImageNet pre-training for dependent tasks and we expect these discoveries will encourage people to rethink the current de facto paradigm of `pre-training and fine-tuning' in computer vision."
                    },
                    {
                        "title": "Mask R-CNN",
                        "abstract": "We present a conceptually simple, flexible, and general framework for object instance segmentation. Our approach efficiently detects objects in an image while simultaneously generating a high-quality segmentation mask for each instance. The method, called Mask R-CNN, extends Faster R-CNN by adding a branch for predicting an object mask in parallel with the existing branch for bounding box recognition. Mask R-CNN is simple to train and adds only a small overhead to Faster R-CNN, running at 5 fps. Moreover, Mask R-CNN is easy to generalize to other tasks, e.g., allowing us to estimate human poses in the same framework. We show top results in all three tracks of the COCO suite of challenges, including instance segmentation, bounding-box object detection, and person keypoint detection. Without bells and whistles, Mask R-CNN outperforms all existing, single-model entries on every task, including the COCO 2016 challenge winners. We hope our simple and effective approach will serve as a solid baseline and help ease future research in instance-level recognition. Code has been made available at: https://github.com/facebookresearch/Detectron"
                    },
                    {
                        "title": "Accelerating Very Deep Convolutional Networks for Classification and Detection",
                        "abstract": "This paper aims to accelerate the test-time computation of convolutional neural networks (CNNs), especially very deep CNNs that have substantially impacted the computer vision community. Unlike previous methods that are designed for approximating linear filters or linear responses, our method takes the nonlinear units into account. We develop an effective solution to the resulting nonlinear optimization problem without the need of stochastic gradient descent (SGD). More importantly, while previous methods mainly focus on optimizing one or two layers, our nonlinear method enables an asymmetric reconstruction that reduces the rapidly accumulated error when multiple (e.g., >=10) layers are approximated. For the widely used very deep VGG-16 model, our method achieves a whole-model speedup of 4x with merely a 0.3% increase of top-5 error in ImageNet classification. Our 4x accelerated VGG-16 model also shows a graceful accuracy degradation for object detection when plugged into the Fast R-CNN detector."
                    },
                    {
                        "title": "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks",
                        "abstract": "State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available."
                    },
                    {
                        "title": "Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition",
                        "abstract": "Existing deep convolutional neural networks (CNNs) require a fixed-size (e.g., 224x224) input image. This requirement is \"artificial\" and may reduce the recognition accuracy for the images or sub-images of an arbitrary size/scale. In this work, we equip the networks with another pooling strategy, \"spatial pyramid pooling\", to eliminate the above requirement. The new network structure, called SPP-net, can generate a fixed-length representation regardless of image size/scale. Pyramid pooling is also robust to object deformations. With these advantages, SPP-net should in general improve all CNN-based image classification methods. On the ImageNet 2012 dataset, we demonstrate that SPP-net boosts the accuracy of a variety of CNN architectures despite their different designs. On the Pascal VOC 2007 and Caltech101 datasets, SPP-net achieves state-of-the-art classification results using a single full-image representation and no fine-tuning.   The power of SPP-net is also significant in object detection. Using SPP-net, we compute the feature maps from the entire image only once, and then pool features in arbitrary regions (sub-images) to generate fixed-length representations for training the detectors. This method avoids repeatedly computing the convolutional features. In processing test images, our method is 24-102x faster than the R-CNN method, while achieving better or comparable accuracy on Pascal VOC 2007.   In ImageNet Large Scale Visual Recognition Challenge (ILSVRC) 2014, our methods rank #2 in object detection and #3 in image classification among all 38 teams. This manuscript also introduces the improvement made for this competition."
                    },
                    {
                        "title": "BoxSup: Exploiting Bounding Boxes to Supervise Convolutional Networks for Semantic Segmentation",
                        "abstract": "Recent leading approaches to semantic segmentation rely on deep convolutional networks trained with human-annotated, pixel-level segmentation masks. Such pixel-accurate supervision demands expensive labeling effort and limits the performance of deep networks that usually benefit from more training data. In this paper, we propose a method that achieves competitive accuracy but only requires easily obtained bounding box annotations. The basic idea is to iterate between automatically generating region proposals and training convolutional networks. These two steps gradually recover segmentation masks for improving the networks, and vise versa. Our method, called BoxSup, produces competitive results supervised by boxes only, on par with strong baselines fully supervised by masks under the same setting. By leveraging a large amount of bounding boxes, BoxSup further unleashes the power of deep convolutional networks and yields state-of-the-art results on PASCAL VOC 2012 and PASCAL-CONTEXT."
                    },
                    {
                        "title": "R-FCN: Object Detection via Region-based Fully Convolutional Networks",
                        "abstract": "We present region-based, fully convolutional networks for accurate and efficient object detection. In contrast to previous region-based detectors such as Fast/Faster R-CNN that apply a costly per-region subnetwork hundreds of times, our region-based detector is fully convolutional with almost all computation shared on the entire image. To achieve this goal, we propose position-sensitive score maps to address a dilemma between translation-invariance in image classification and translation-variance in object detection. Our method can thus naturally adopt fully convolutional image classifier backbones, such as the latest Residual Networks (ResNets), for object detection. We show competitive results on the PASCAL VOC datasets (e.g., 83.6% mAP on the 2007 set) with the 101-layer ResNet. Meanwhile, our result is achieved at a test-time speed of 170ms per image, 2.5-20x faster than the Faster R-CNN counterpart. Code is made publicly available at: https://github.com/daijifeng001/r-fcn"
                    },
                    {
                        "title": "Detecting and Recognizing Human-Object Interactions",
                        "abstract": "To understand the visual world, a machine must not only recognize individual object instances but also how they interact. Humans are often at the center of such interactions and detecting human-object interactions is an important practical and scientific problem. In this paper, we address the task of detecting <human, verb, object> triplets in challenging everyday photos. We propose a novel model that is driven by a human-centric approach. Our hypothesis is that the appearance of a person -- their pose, clothing, action -- is a powerful cue for localizing the objects they are interacting with. To exploit this cue, our model learns to predict an action-specific density over target object locations based on the appearance of a detected person. Our model also jointly learns to detect people and objects, and by fusing these predictions it efficiently infers interaction triplets in a clean, jointly trained end-to-end system we call InteractNet. We validate our approach on the recently introduced Verbs in COCO (V-COCO) and HICO-DET datasets, where we show quantitatively compelling results."
                    },
                    {
                        "title": "Non-local Neural Networks",
                        "abstract": "Both convolutional and recurrent operations are building blocks that process one local neighborhood at a time. In this paper, we present non-local operations as a generic family of building blocks for capturing long-range dependencies. Inspired by the classical non-local means method in computer vision, our non-local operation computes the response at a position as a weighted sum of the features at all positions. This building block can be plugged into many computer vision architectures. On the task of video classification, even without any bells and whistles, our non-local models can compete or outperform current competition winners on both Kinetics and Charades datasets. In static image recognition, our non-local models improve object detection/segmentation and pose estimation on the COCO suite of tasks. Code is available at https://github.com/facebookresearch/video-nonlocal-net ."
                    },
                    {
                        "title": "TensorMask: A Foundation for Dense Object Segmentation",
                        "abstract": "Sliding-window object detectors that generate bounding-box object predictions over a dense, regular grid have advanced rapidly and proven popular. In contrast, modern instance segmentation approaches are dominated by methods that first detect object bounding boxes, and then crop and segment these regions, as popularized by Mask R-CNN. In this work, we investigate the paradigm of dense sliding-window instance segmentation, which is surprisingly under-explored. Our core observation is that this task is fundamentally different than other dense prediction tasks such as semantic segmentation or bounding-box object detection, as the output at every spatial location is itself a geometric structure with its own spatial dimensions. To formalize this, we treat dense instance segmentation as a prediction task over 4D tensors and present a general framework called TensorMask that explicitly captures this geometry and enables novel operators on 4D tensors. We demonstrate that the tensor view leads to large gains over baselines that ignore this structure, and leads to results comparable to Mask R-CNN. These promising results suggest that TensorMask can serve as a foundation for novel advances in dense mask prediction and a more complete understanding of the task. Code will be made available."
                    }
                ]
            }
        }
    },
    "2407.00401": {
        "paper_data": {
            "title": "PUZZLES: A Benchmark for Neural Algorithmic Reasoning",
            "url": "http://arxiv.org/abs/2407.00401v1",
            "arxiv_id": "2407.00401",
            "authors": [
                "Benjamin Estermann",
                "Luca A. Lanzend\u00f6rfer",
                "Yannick Niedermayr",
                "Roger Wattenhofer"
            ],
            "abstract": "Algorithmic reasoning is a fundamental cognitive ability that plays a pivotal role in problem-solving and decision-making processes. Reinforcement Learning (RL) has demonstrated remarkable proficiency in tasks such as motor control, handling perceptual input, and managing stochastic environments. These advancements have been enabled in part by the availability of benchmarks. In this work we introduce PUZZLES, a benchmark based on Simon Tatham's Portable Puzzle Collection, aimed at fostering progress in algorithmic and logical reasoning in RL. PUZZLES contains 40 diverse logic puzzles of adjustable sizes and varying levels of complexity; many puzzles also feature a diverse set of additional configuration parameters. The 40 puzzles provide detailed information on the strengths and generalization capabilities of RL agents. Furthermore, we evaluate various RL algorithms on PUZZLES, providing baseline comparisons and demonstrating the potential for future research. All the software, including the environment, is available at https://github.com/ETH-DISCO/rlp.",
            "introduction": "  1 Related Work  RL benchmarks.  Various benchmarks have been proposed in RL. Bellemare et\u00a0al. [24] introduced the influential Atari-2600 benchmark, on which Mnih et\u00a0al. [11] trained RL agents to play the games directly from pixel inputs. This benchmark demonstrated the potential of RL in complex, high-dimensional environments. PUZZLES allows the use of a similar approach where only pixel inputs are provided to the agent. Todorov et\u00a0al. [23] presented MuJoCo which provides a diverse set of continuous control tasks based on a physics engine for robotic systems. Another control benchmark is the DeepMind Control Suite by Duan et\u00a0al. [26], featuring continuous actions spaces and complex control problems. The work by C\u00f4t\u00e9 et\u00a0al. [28] emphasized the importance of natural language understanding in RL and proposed a benchmark for evaluating RL methods in text-based domains. Lanctot et\u00a0al. [29] introduced OpenSpiel, encompassing a wide range of games, enabling researchers to evaluate and compare RL algorithms\u2019 performance in game-playing scenarios. These benchmarks and frameworks have contributed significantly to the development and evaluation of RL algorithms. OpenAI Gym by Brockman et\u00a0al. [25], and its successor Gymnasium by the Farama Foundation\u00a0[36] helped by providing a standardized interface for many benchmarks. As such, Gym and Gymnasium have played an important role in facilitating reproducibility and benchmarking in reinforcement learning research. Therefore, we provide PUZZLES as a Gymnasium environment to enable ease of use.    Logical and algorithmic reasoning within RL.  Notable research in RL on logical reasoning includes automated theorem proving using deep RL\u00a0[16] or RL-based logic synthesis\u00a0[37]. Dasgupta et\u00a0al. [38] find that RL agents can perform a certain degree of causal reasoning in a meta-reinforcement learning setting. The work by Jiang and Luo [30] introduces Neural Logic RL, which improves interpretability and generalization of learned policies. Eppe et\u00a0al. [39] provide steps to advance problem-solving as part of hierarchical RL. Fawzi et\u00a0al. [31] and Mankowitz et\u00a0al. [32] demonstrate that RL can be used to discover novel and more efficient algorithms for well-known problems such as matrix multiplication and sorting. Neural algorithmic reasoning has also been used as a method to improve low-data performance in classical RL control environments\u00a0[40, 41]. Logical reasoning might be required to compete in certain types of games such as chess, shogi and Go\u00a0[33, 34, 42, 13], Poker [43, 44, 45, 46] or board games [47, 48, 49, 50]. However, these are usually multi-agent games, with some also featuring imperfect information and stochasticity.    Reasoning benchmarks.  Various benchmarks have been introduced to assess different types of reasoning capabilities, although only in the realm of classical ML. IsarStep, proposed by Li et\u00a0al. [8], specifically designed to evaluate high-level mathematical reasoning necessary for proof-writing tasks. Another significant benchmark in the field of reasoning is the CLRS Algorithmic Reasoning Benchmark, introduced by Veli\u010dkovi\u0107 et\u00a0al. [9]. This benchmark emphasizes the importance of algorithmic reasoning in machine learning research. It consists of 30 different types of algorithms sourced from the renowned textbook \u201cIntroduction to Algorithms\u201d by\u00a0Cormen et\u00a0al. [51]. The CLRS benchmark serves as a means to evaluate models\u2019 understanding and proficiency in learning various algorithms. In the domain of large language models (LLMs), BIG-bench has been introduced by Srivastava et\u00a0al. [10]. BIG-bench incorporates tasks that assess the reasoning capabilities of LLMs, including logical reasoning.   Despite these valuable contributions, a suitable and unified benchmark for evaluating logical",
            "references": [
                {
                    "title": "Mastering Diverse Domains through World Models",
                    "abstract": "Developing a general algorithm that learns to solve tasks across a wide range of applications has been a fundamental challenge in artificial intelligence. Although current reinforcement learning algorithms can be readily applied to tasks similar to what they have been developed for, configuring them for new application domains requires significant human expertise and experimentation. We present DreamerV3, a general algorithm that outperforms specialized methods across over 150 diverse tasks, with a single configuration. Dreamer learns a model of the environment and improves its behavior by imagining future scenarios. Robustness techniques based on normalization, balancing, and transformations enable stable learning across domains. Applied out of the box, Dreamer is the first algorithm to collect diamonds in Minecraft from scratch without human data or curricula. This achievement has been posed as a significant challenge in artificial intelligence that requires exploring farsighted strategies from pixels and sparse rewards in an open world. Our work allows solving challenging control problems without extensive experimentation, making reinforcement learning broadly applicable."
                },
                {
                    "title": "Continuous Neural Algorithmic Planners",
                    "abstract": "Neural algorithmic reasoning studies the problem of learning algorithms with neural networks, especially with graph architectures. A recent proposal, XLVIN, reaps the benefits of using a graph neural network that simulates the value iteration algorithm in deep reinforcement learning agents. It allows model-free planning without access to privileged information about the environment, which is usually unavailable. However, XLVIN only supports discrete action spaces, and is hence nontrivially applicable to most tasks of real-world interest. We expand XLVIN to continuous action spaces by discretization, and evaluate several selective expansion policies to deal with the large planning graphs. Our proposal, CNAP, demonstrates how neural algorithmic reasoning can make a measurable impact in higher-dimensional continuous control settings, such as MuJoCo, bringing gains in low-data settings and outperforming model-free baselines."
                },
                {
                    "title": "Residual Skill Policies: Learning an Adaptable Skill-based Action Space for Reinforcement Learning for Robotics",
                    "abstract": "Skill-based reinforcement learning (RL) has emerged as a promising strategy to leverage prior knowledge for accelerated robot learning. Skills are typically extracted from expert demonstrations and are embedded into a latent space from which they can be sampled as actions by a high-level RL agent. However, this skill space is expansive, and not all skills are relevant for a given robot state, making exploration difficult. Furthermore, the downstream RL agent is limited to learning structurally similar tasks to those used to construct the skill space. We firstly propose accelerating exploration in the skill space using state-conditioned generative models to directly bias the high-level agent towards only sampling skills relevant to a given state based on prior experience. Next, we propose a low-level residual policy for fine-grained skill adaptation enabling downstream RL agents to adapt to unseen task variations. Finally, we validate our approach across four challenging manipulation tasks that differ from those used to build the skill space, demonstrating our ability to learn across task variations while significantly accelerating exploration, outperforming prior works. Code and videos are available on our project website: https://krishanrana.github.io/reskill."
                },
                {
                    "title": "Mastering the game of Stratego with model-free multiagent reinforcement learning",
                    "abstract": "We introduce DeepNash, an autonomous agent that plays the imperfect information game Stratego at a human expert level. Stratego is one of the few iconic board games that artificial intelligence (AI) has not yet mastered. It is a game characterized by a twin challenge: It requires long-term strategic thinking as in chess, but it also requires dealing with imperfect information as in poker. The technique underpinning DeepNash uses a game-theoretic, model-free deep reinforcement learning method, without search, that learns to master Stratego through self-play from scratch. DeepNash beat existing state-of-the-art AI methods in Stratego and achieved a year-to-date (2022) and all-time top-three ranking on the Gravon games platform, competing with human expert players. Description Machine learning to play Stratego Stratego is a popular two-player imperfect information board game. Because of its complexity stemming from its enormous game tree, decision-making under imperfect information, and a piece deployment phase at the start, Stratego poses a challenge for artificial intelligence (AI). Previous computer programs only performed at an amateur level at best. Perolat et al. introduce a model-free multiagent reinforcement learning methodology and show that it can achieve human expert\u2013level performance in Stratego. The present work not only adds to the growing list of games that AI systems can play as well or even better than humans but may also facilitate further applications of reinforcement learning methods in real-world, large-scale multiagent problems that are characterized by imperfect information and thus are currently unsolvable. \u2014YS Reinforcement learning achieves human expert\u2013level performance in the large-scale imperfect information board game Stratego."
                },
                {
                    "title": "AlphaHoldem: High-Performance Artificial Intelligence for Heads-Up No-Limit Poker via End-to-End Reinforcement Learning",
                    "abstract": "Heads-up no-limit Texas hold\u2019em (HUNL) is the quintessential game with imperfect information. Representative priorworks like DeepStack and Libratus heavily rely on counter-factual regret minimization (CFR) and its variants to tackleHUNL. However, the prohibitive computation cost of CFRiteration makes it difficult for subsequent researchers to learnthe CFR model in HUNL and apply it in other practical applications. In this work, we present AlphaHoldem, a high-performance and lightweight HUNL AI obtained with an end-to-end self-play reinforcement learning framework. The proposed framework adopts a pseudo-siamese architecture to directly learn from the input state information to the output actions by competing the learned model with its different historical versions. The main technical contributions include anovel state representation of card and betting information, amultitask self-play training loss function, and a new modelevaluation and selection metric to generate the final model.In a study involving 100,000 hands of poker, AlphaHoldemdefeats Slumbot and DeepStack using only one PC with threedays training. At the same time, AlphaHoldem only takes 2.9milliseconds for each decision-making using only a singleGPU, more than 1,000 times faster than DeepStack. We release the history data among among AlphaHoldem, Slumbot,and top human professionals in the author\u2019s GitHub repository to facilitate further studies in this direction."
                },
                {
                    "title": "Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models",
                    "abstract": "Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit\"breakthrough\"behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting."
                },
                {
                    "title": "The CLRS Algorithmic Reasoning Benchmark",
                    "abstract": "Learning representations of algorithms is an emerging area of machine learning, seeking to bridge concepts from neural networks with classical algorithms. Several important works have investigated whether neural networks can effectively reason like algorithms, typically by learning to execute them. The common trend in the area, however, is to generate targeted kinds of algorithmic data to evaluate specific hypotheses, making results hard to transfer across publications, and increasing the barrier of entry. To consolidate progress and work towards unified evaluation, we propose the CLRS Algorithmic Reasoning Benchmark, covering classical algorithms from the Introduction to Algorithms textbook. Our benchmark spans a variety of algorithmic reasoning procedures, including sorting, searching, dynamic programming, graph algorithms, string algorithms and geometric algorithms. We perform extensive experiments to demonstrate how several popular algorithmic reasoning baselines perform on these tasks, and consequently, highlight links to several open challenges. Our library is readily available at https://github.com/deepmind/clrs."
                },
                {
                    "title": "Rethinking Reinforcement Learning based Logic Synthesis",
                    "abstract": "Recently, reinforcement learning has been used to address logic synthesis by formulating the operator sequence optimization problem as a Markov decision process. However, through extensive experiments, we find out that the learned policy makes decisions independent from the circuit features (i.e., states) and yields an operator sequence that is permutation invariant to some extent in terms of operators. Based on these findings, we develop a new RL-based method that can automatically recognize critical operators and generate common operator sequences generalizable to unseen circuits. Our algorithm is verified on both the EPFL benchmark, a private dataset and a circuit at industrial scale. Experimental results demonstrate that it achieves a good balance among delay, area and runtime, and is practical for industrial usage."
                },
                {
                    "title": "Deep Reinforcement Learning-Based Path Control and Optimization for Unmanned Ships",
                    "abstract": "Unmanned ship navigates on the water in an autonomous or semiautonomous way, which can be widely used in maritime transportation, intelligence collection, maritime training and testing, reconnaissance, and evidence collection. In this paper, we use deep reinforcement learning to solve the optimization problem in the path planning and management of unmanned ships. Specifically, we take the waiting time (phase and duration) at the corner of the path as the optimization goal to minimize the total travel time of unmanned ships passing through the path. We propose a new reward function, which considers the environment and control delay of unmanned ships at the same time, which can reduce the coordination time between unmanned ships at the same time. In the simulation experiment, through the quantitative and qualitative results of deep reinforcement learning of unmanned ship navigation and path angle waiting, the effectiveness of our solution is verified."
                },
                {
                    "title": "ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning",
                    "abstract": "Charts are very popular for analyzing data. When exploring charts, people often ask a variety of complex reasoning questions that involve several logical and arithmetic operations. They also commonly refer to visual features of a chart in their questions. However, most existing datasets do not focus on such complex reasoning questions as their questions are template-based and answers come from a fixed-vocabulary. In this work, we present a large-scale benchmark covering 9.6K human-written questions as well as 23.1K questions generated from human-written chart summaries. To address the unique challenges in our benchmark involving visual and logical reasoning over charts, we present two transformer-based models that combine visual features and the data table of the chart in a unified way to answer questions. While our models achieve the state-of-the-art results on the previous datasets as well as on our benchmark, the evaluation also reveals several challenges in answering complex reasoning questions."
                },
                {
                    "title": "MERIt: Meta-Path Guided Contrastive Learning for Logical Reasoning",
                    "abstract": "Logical reasoning is of vital importance to natural language understanding. Previous studies either employ graph-based models to incorporate prior knowledge about logical relations, or introduce symbolic logic into neural models through data augmentation. These methods, however, heavily depend on annotated training data, and thus suffer from over-fitting and poor generalization problems due to the dataset sparsity. To address these two problems, in this paper, we propose MERIt, a MEta-path guided contrastive learning method for logical ReasonIng of text, to perform self-supervised pre-training on abundant unlabeled text data. Two novel strategies serve as indispensable components of our method. In particular, a strategy based on meta-path is devised to discover the logical structure in natural texts, followed by a counterfactual data augmentation strategy to eliminate the information shortcut induced by pre-training. The experimental results on two challenging logical reasoning benchmarks, i.e., ReClor and LogiQA, demonstrate that our method outperforms the SOTA baselines with significant improvements."
                },
                {
                    "title": "Neural Algorithmic Reasoners are Implicit Planners",
                    "abstract": "Implicit planning has emerged as an elegant technique for combining learned models of the world with end-to-end model-free reinforcement learning. We study the class of implicit planners inspired by value iteration, an algorithm that is guaranteed to yield perfect policies in fully-specified tabular environments. We find that prior approaches either assume that the environment is provided in such a tabular form -- which is highly restrictive -- or infer\"local neighbourhoods\"of states to run value iteration over -- for which we discover an algorithmic bottleneck effect. This effect is caused by explicitly running the planning algorithm based on scalar predictions in every state, which can be harmful to data efficiency if such scalars are improperly predicted. We propose eXecuted Latent Value Iteration Networks (XLVINs), which alleviate the above limitations. Our method performs all planning computations in a high-dimensional latent space, breaking the algorithmic bottleneck. It maintains alignment with value iteration by carefully leveraging neural graph-algorithmic reasoning and contrastive self-supervised learning. Across eight low-data settings -- including classical control, navigation and Atari -- XLVINs provide significant improvements to data efficiency against value iteration-based implicit planners, as well as relevant model-free baselines. Lastly, we empirically verify that XLVINs can closely align with value iteration."
                },
                {
                    "title": "Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning",
                    "abstract": "In this work, we present and study a training set-up that achieves fast policy generation for real-world robotic tasks by using massive parallelism on a single workstation GPU. We analyze and discuss the impact of different training algorithm components in the massively parallel regime on the final policy performance and training times. In addition, we present a novel game-inspired curriculum that is well suited for training with thousands of simulated robots in parallel. We evaluate the approach by training the quadrupedal robot ANYmal to walk on challenging terrain. The parallel approach allows training policies for flat terrain in under four minutes, and in twenty minutes for uneven terrain. This represents a speedup of multiple orders of magnitude compared to previous work. Finally, we transfer the policies to the real robot to validate the approach. We open-source our training code to help accelerate further research in the field of learned legged locomotion."
                },
                {
                    "title": "Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning",
                    "abstract": "The last half decade has seen a steep rise in the number of contributions on safe learning methods for real-world robotic deployments from both the control and reinforcement learning communities. This article provides a concise but holistic review of the recent advances made in using machine learning to achieve safe decision-making under uncertainties, with a focus on unifying the language and frameworks used in control theory and reinforcement learning research. It includes learning-based control approaches that safely improve performance by learning the uncertain dynamics, reinforcement learning approaches that encourage safety or robustness, and methods that can formally certify the safety of a learned control policy. As data- and learning-based robot control methods continue to gain traction, researchers must understand when and how to best leverage them in real-world scenarios where safety is imperative, such as when operating in close proximity to humans. We highlight some of the open challenges that will drive the field of robot learning in the coming years, and emphasize the need for realistic physics-based benchmarks to facilitate fair comparisons between control and reinforcement learning approaches. Expected final online publication date for the Annual Review of Control, Robotics, and Autonomous Systems, Volume 5 is May 2022. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates."
                },
                {
                    "title": "IsarStep: a Benchmark for High-level Mathematical Reasoning",
                    "abstract": "A well-defined benchmark is essential for measuring and accelerating research progress of machine learning models. In this paper, we present a benchmark for high-level mathematical reasoning and study the reasoning capabilities of neural sequence-to-sequence models. We build a non-synthetic dataset from the largest repository of proofs written by human experts in a theorem prover. The dataset has a broad coverage of undergraduate and research-level mathematical and computer science theorems. In our defined task, a model is required to fill in a missing intermediate proposition given surrounding proofs. This task provides a starting point for the long-term goal of having machines generate human-readable proofs automatically. Our experiments and analysis reveal that while the task is challenging, neural models can capture non-trivial mathematical reasoning. We further design a hierarchical transformer that outperforms the transformer baseline. We will make the dataset and models publicly available."
                },
                {
                    "title": "Strong Generalization and Efficiency in Neural Programs",
                    "abstract": "We study the problem of learning efficient algorithms that strongly generalize in the framework of neural program induction. By carefully designing the input / output interfaces of the neural model and through imitation, we are able to learn models that produce correct results for arbitrary input sizes, achieving strong generalization. Moreover, by using reinforcement learning, we optimize for program efficiency metrics, and discover new algorithms that surpass the teacher used in imitation. With this, our approach can learn to outperform custom-written solutions for a variety of problems, as we tested it on sorting, searching in ordered lists and the NP-complete 0/1 knapsack problem, which sets a notable milestone in the field of Neural Program Induction. As highlights, our learned model can perform sorting perfectly on any input data size we tested on, with $O(n log n)$ complexity, whilst outperforming hand-coded algorithms, including quick sort, in number of operations even for list sizes far beyond those seen during training."
                },
                {
                    "title": "Agent57: Outperforming the Atari Human Benchmark",
                    "abstract": "Atari games have been a long-standing benchmark in the reinforcement learning (RL) community for the past decade. This benchmark was proposed to test general competency of RL algorithms. Previous work has achieved good average performance by doing outstandingly well on many games of the set, but very poorly in several of the most challenging games. We propose Agent57, the first deep RL agent that outperforms the standard human benchmark on all 57 Atari games. To achieve this result, we train a neural network which parameterizes a family of policies ranging from very exploratory to purely exploitative. We propose an adaptive mechanism to choose which policy to prioritize throughout the training process. Additionally, we utilize a novel parameterization of the architecture that allows for more consistent and stable learning."
                },
                {
                    "title": "Suphx: Mastering Mahjong with Deep Reinforcement Learning",
                    "abstract": "Artificial Intelligence (AI) has achieved great success in many domains, and game AI is widely regarded as its beachhead since the dawn of AI. In recent years, studies on game AI have gradually evolved from relatively simple environments (e.g., perfect-information games such as Go, chess, shogi or two-player imperfect-information games such as heads-up Texas hold'em) to more complex ones (e.g., multi-player imperfect-information games such as multi-player Texas hold'em and StartCraft II). Mahjong is a popular multi-player imperfect-information game worldwide but very challenging for AI research due to its complex playing/scoring rules and rich hidden information. We design an AI for Mahjong, named Suphx, based on deep reinforcement learning with some newly introduced techniques including global reward prediction, oracle guiding, and run-time policy adaptation. Suphx has demonstrated stronger performance than most top human players in terms of stable rank and is rated above 99.99% of all the officially ranked human players in the Tenhou platform. This is the first time that a computer program outperforms most top human players in Mahjong."
                },
                {
                    "title": "Deep Reinforcement Learning for Autonomous Driving: A Survey",
                    "abstract": "With the development of deep representation learning, the domain of reinforcement learning (RL) has become a powerful learning framework now capable of learning complex policies in high dimensional environments. This review summarises deep reinforcement learning (DRL) algorithms and provides a taxonomy of automated driving tasks where (D)RL methods have been employed, while addressing key computational challenges in real world deployment of autonomous driving agents. It also delineates adjacent domains such as behavior cloning, imitation learning, inverse reinforcement learning that are related but are not classical RL algorithms. The role of simulators in training agents, methods to validate, test and robustify existing solutions in RL are discussed."
                },
                {
                    "title": "OpenSpiel: A Framework for Reinforcement Learning in Games",
                    "abstract": "OpenSpiel is a collection of environments and algorithms for research in general reinforcement learning and search/planning in games. OpenSpiel supports n-player (single- and multi- agent) zero-sum, cooperative and general-sum, one-shot and sequential, strictly turn-taking and simultaneous-move, perfect and imperfect information games, as well as traditional multiagent environments such as (partially- and fully- observable) grid worlds and social dilemmas. OpenSpiel also includes tools to analyze learning dynamics and other common evaluation metrics. This document serves both as an overview of the code base and an introduction to the terminology, core concepts, and algorithms across the fields of reinforcement learning, computational game theory, and search."
                },
                {
                    "title": "Neural Logic Reinforcement Learning",
                    "abstract": "Deep reinforcement learning (DRL) has achieved significant breakthroughs in various tasks. However, most DRL algorithms suffer a problem of generalizing the learned policy which makes the learning performance largely affected even by minor modifications of the training environment. Except that, the use of deep neural networks makes the learned policies hard to be interpretable. To address these two challenges, we propose a novel algorithm named Neural Logic Reinforcement Learning (NLRL) to represent the policies in reinforcement learning by first-order logic. NLRL is based on policy gradient methods and differentiable inductive logic programming that have demonstrated significant advantages in terms of interpretability and generalisability in supervised tasks. Extensive experiments conducted on cliff-walking and blocks manipulation tasks demonstrate that NLRL can induce interpretable policies achieving near-optimal performance while demonstrating good generalisability to environments of different initial states and problem sizes."
                },
                {
                    "title": "Causal Reasoning from Meta-reinforcement Learning",
                    "abstract": "Discovering and exploiting the causal structure in the environment is a crucial challenge for intelligent agents. Here we explore whether causal reasoning can emerge via meta-reinforcement learning. We train a recurrent network with model-free reinforcement learning to solve a range of problems that each contain causal structure. We find that the trained agent can perform causal reasoning in novel situations in order to obtain rewards. The agent can select informative interventions, draw causal inferences from observational data, and make counterfactual predictions. Although established formal causal reasoning algorithms also exist, in this paper we show that such reasoning can arise from model-free reinforcement learning, and suggest that causal reasoning in complex settings may benefit from the more end-to-end learning-based approaches presented here. This work also offers new strategies for structured exploration in reinforcement learning, by providing agents with the ability to perform -- and interpret -- experiments."
                },
                {
                    "title": "A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play",
                    "abstract": "One program to rule them all Computers can beat humans at increasingly complex games, including chess and Go. However, these programs are typically constructed for a particular game, exploiting its properties, such as the symmetries of the board on which it is played. Silver et al. developed a program called AlphaZero, which taught itself to play Go, chess, and shogi (a Japanese version of chess) (see the Editorial, and the Perspective by Campbell). AlphaZero managed to beat state-of-the-art programs specializing in these three games. The ability of AlphaZero to adapt to various game rules is a notable step toward achieving a general game-playing system. Science, this issue p. 1140; see also pp. 1087 and 1118 AlphaZero teaches itself to play three different board games and beats state-of-the-art programs in each. The game of chess is the longest-studied domain in the history of artificial intelligence. The strongest programs are based on a combination of sophisticated search techniques, domain-specific adaptations, and handcrafted evaluation functions that have been refined by human experts over several decades. By contrast, the AlphaGo Zero program recently achieved superhuman performance in the game of Go by reinforcement learning from self-play. In this paper, we generalize this approach into a single AlphaZero algorithm that can achieve superhuman performance in many challenging games. Starting from random play and given no domain knowledge except the game rules, AlphaZero convincingly defeated a world champion program in the games of chess and shogi (Japanese chess), as well as Go."
                },
                {
                    "title": "QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation",
                    "abstract": "In this paper, we study the problem of learning vision-based dynamic manipulation skills using a scalable reinforcement learning approach. We study this problem in the context of grasping, a longstanding challenge in robotic manipulation. In contrast to static learning behaviors that choose a grasp point and then execute the desired grasp, our method enables closed-loop vision-based control, whereby the robot continuously updates its grasp strategy based on the most recent observations to optimize long-horizon grasp success. To that end, we introduce QT-Opt, a scalable self-supervised vision-based reinforcement learning framework that can leverage over 580k real-world grasp attempts to train a deep neural network Q-function with over 1.2M parameters to perform closed-loop, real-world grasping that generalizes to 96% grasp success on unseen objects. Aside from attaining a very high success rate, our method exhibits behaviors that are quite distinct from more standard grasping systems: using only RGB vision-based perception from an over-the-shoulder camera, our method automatically learns regrasping strategies, probes objects to find the most effective grasps, learns to reposition objects and perform other non-prehensile pre-grasp manipulations, and responds dynamically to disturbances and perturbations."
                },
                {
                    "title": "Addressing Function Approximation Error in Actor-Critic Methods",
                    "abstract": "In value-based reinforcement learning methods such as deep Q-learning, function approximation errors are known to lead to overestimated value estimates and suboptimal policies. We show that this problem persists in an actor-critic setting and propose novel mechanisms to minimize its effects on both the actor and the critic. Our algorithm builds on Double Q-learning, by taking the minimum value between a pair of critics to limit overestimation. We draw the connection between target networks and overestimation bias, and suggest delaying policy updates to reduce per-update error and further improve performance. We evaluate our method on the suite of OpenAI gym tasks, outperforming the state of the art in every environment tested."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "DeepMind Control Suite",
                    "abstract": "The DeepMind Control Suite is a set of continuous control tasks with a standardised structure and interpretable rewards, intended to serve as performance benchmarks for reinforcement learning agents. The tasks are written in Python and powered by the MuJoCo physics engine, making them easy to use and modify. We include benchmarks for several learning algorithms. The Control Suite is publicly available at this https URL A video summary of all tasks is available at this http URL ."
                },
                {
                    "title": "Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm",
                    "abstract": "The game of chess is the most widely-studied domain in the history of artificial intelligence. The strongest programs are based on a combination of sophisticated search techniques, domain-specific adaptations, and handcrafted evaluation functions that have been refined by human experts over several decades. In contrast, the AlphaGo Zero program recently achieved superhuman performance in the game of Go, by tabula rasa reinforcement learning from games of self-play. In this paper, we generalise this approach into a single AlphaZero algorithm that can achieve, tabula rasa, superhuman performance in many challenging domains. Starting from random play, and given no domain knowledge except the game rules, AlphaZero achieved within 24 hours a superhuman level of play in the games of chess and shogi (Japanese chess) as well as Go, and convincingly defeated a world-champion program in each case."
                },
                {
                    "title": "Distributional Reinforcement Learning with Quantile Regression",
                    "abstract": "\n \n In reinforcement learning (RL), an agent interacts with the environment by taking actions and observing the next state and reward. When sampled probabilistically, these state transitions, rewards, and actions can all induce randomness in the observed long-term return. Traditionally, reinforcement learning algorithms average over this randomness to estimate the value function. In this paper, we build on recent work advocating a distributional approach to reinforcement learning in which the distribution over returns is modeled explicitly instead of only estimating the mean. That is, we examine methods of learning the value distribution instead of the value function. We give results that close a number of gaps between the theoretical and algorithmic results given by Bellemare, Dabney, and Munos (2017). First, we extend existing results to the approximate distribution setting. Second, we present a novel distributional reinforcement learning algorithm consistent with our theoretical formulation. Finally, we evaluate this new algorithm on the Atari 2600 games, observing that it significantly outperforms many of the recent improvements on DQN, including the related distributional algorithm C51.\n \n"
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "#Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning",
                    "abstract": "Count-based exploration algorithms are known to perform near-optimally when used in conjunction with tabular reinforcement learning (RL) methods for solving small discrete Markov decision processes (MDPs). It is generally thought that count-based methods cannot be applied in high-dimensional state spaces, since most states will only occur once. Recent deep RL exploration strategies are able to deal with high-dimensional continuous state spaces through complex heuristics, often relying on optimism in the face of uncertainty or intrinsic motivation. In this work, we describe a surprising finding: a simple generalization of the classic count-based approach can reach near state-of-the-art performance on various high-dimensional and/or continuous deep RL benchmarks. States are mapped to hash codes, which allows to count their occurrences with a hash table. These counts are then used to compute a reward bonus according to the classic count-based exploration theory. We find that simple hash functions can achieve surprisingly good results on many challenging tasks. Furthermore, we show that a domain-dependent learned hash code may further improve these results. Detailed analysis reveals important aspects of a good hash function: 1) having appropriate granularity and 2) encoding information relevant to solving the MDP. This exploration strategy achieves near state-of-the-art performance on both continuous control tasks and Atari 2600 games, hence providing a simple yet powerful baseline for solving MDPs that require considerable exploration."
                },
                {
                    "title": "OpenAI Gym",
                    "abstract": "OpenAI Gym is a toolkit for reinforcement learning research. It includes a growing collection of benchmark problems that expose a common interface, and a website where people can share their results and compare the performance of algorithms. This whitepaper discusses the components of OpenAI Gym and the design decisions that went into the software."
                },
                {
                    "title": "Benchmarking Deep Reinforcement Learning for Continuous Control",
                    "abstract": "Recently, researchers have made significant progress combining the advances in deep learning for learning feature representations with reinforcement learning. Some notable examples include training agents to play Atari games based on raw pixel data and to acquire advanced manipulation skills using raw sensory inputs. However, it has been difficult to quantify progress in the domain of continuous control due to the lack of a commonly adopted benchmark. In this work, we present a benchmark suite of continuous control tasks, including classic tasks like cart-pole swing-up, tasks with very high state and action dimensionality such as 3D humanoid locomotion, tasks with partial observations, and tasks with hierarchical structure. We report novel findings based on the systematic evaluation of a range of implemented reinforcement learning algorithms. Both the benchmark and reference implementations are released at this https URL in order to facilitate experimental reproducibility and to encourage adoption by other researchers."
                },
                {
                    "title": "Deep Reinforcement Learning from Self-Play in Imperfect-Information Games",
                    "abstract": "Many real-world applications can be described as large-scale games of imperfect information. To deal with these challenging domains, prior work has focused on computing Nash equilibria in a handcrafted abstraction of the domain. In this paper we introduce the first scalable end-to-end approach to learning approximate Nash equilibria without prior domain knowledge. Our method combines fictitious self-play with deep reinforcement learning. When applied to Leduc poker, Neural Fictitious Self-Play (NFSP) approached a Nash equilibrium, whereas common reinforcement learning methods diverged. In Limit Texas Holdem, a poker game of real-world scale, NFSP learnt a strategy that approached the performance of state-of-the-art, superhuman algorithms based on significant domain expertise."
                },
                {
                    "title": "Asynchronous Methods for Deep Reinforcement Learning",
                    "abstract": "We propose a conceptually simple and lightweight framework for deep reinforcement learning that uses asynchronous gradient descent for optimization of deep neural network controllers. We present asynchronous variants of four standard reinforcement learning algorithms and show that parallel actor-learners have a stabilizing effect on training allowing all four methods to successfully train neural network controllers. The best performing method, an asynchronous variant of actor-critic, surpasses the current state-of-the-art on the Atari domain while training for half the time on a single multi-core CPU instead of a GPU. Furthermore, we show that asynchronous actor-critic succeeds on a wide variety of continuous motor control problems as well as on a new task of navigating random 3D mazes using a visual input."
                },
                {
                    "title": "Giraffe: Using Deep Reinforcement Learning to Play Chess",
                    "abstract": "This report presents Giraffe, a chess engine that uses self-play to discover all its domain-specific knowledge, with minimal hand-crafted knowledge given by the programmer. Unlike previous attempts using machine learning only to perform parameter-tuning on hand-crafted evaluation functions, Giraffe's learning system also performs automatic feature extraction and pattern recognition. The trained evaluation function performs comparably to the evaluation functions of state-of-the-art chess engines - all of which containing thousands of lines of carefully hand-crafted pattern recognizers, tuned over many years by both computer chess experts and human chess masters. Giraffe is the most successful attempt thus far at using end-to-end machine learning to play chess."
                },
                {
                    "title": "Trust Region Policy Optimization",
                    "abstract": "In this article, we describe a method for optimizing control policies, with guaranteed monotonic improvement. By making several approximations to the theoretically-justified scheme, we develop a practical algorithm, called Trust Region Policy Optimization (TRPO). This algorithm is effective for optimizing large nonlinear policies such as neural networks. Our experiments demonstrate its robust performance on a wide variety of tasks: learning simulated robotic swimming, hopping, and walking gaits; and playing Atari games using images of the screen as input. Despite its approximations that deviate from the theory, TRPO tends to give monotonic improvement, with little tuning of hyperparameters."
                },
                {
                    "title": "Playing Atari with Deep Reinforcement Learning",
                    "abstract": "We present the first deep learning model to successfully learn control policies directly from high-dimensional sensory input using reinforcement learning. The model is a convolutional neural network, trained with a variant of Q-learning, whose input is raw pixels and whose output is a value function estimating future rewards. We apply our method to seven Atari 2600 games from the Arcade Learning Environment, with no adjustment of the architecture or learning algorithm. We find that it outperforms all previous approaches on six of the games and surpasses a human expert on three of them."
                },
                {
                    "title": "MuJoCo: A physics engine for model-based control",
                    "abstract": "We describe a new physics engine tailored to model-based control. Multi-joint dynamics are represented in generalized coordinates and computed via recursive algorithms. Contact responses are computed via efficient new algorithms we have developed, based on the modern velocity-stepping approach which avoids the difficulties with spring-dampers. Models are specified using either a high-level C++ API or an intuitive XML file format. A built-in compiler transforms the user model into an optimized data structure used for runtime computation. The engine can compute both forward and inverse dynamics. The latter are well-defined even in the presence of contacts and equality constraints. The model can include tendon wrapping as well as actuator activation states (e.g. pneumatic cylinders or muscles). To facilitate optimal control applications and in particular sampling and finite differencing, the dynamics can be evaluated for different states and controls in parallel. Around 400,000 dynamics evaluations per second are possible on a 12-core machine, for a 3D homanoid with 18 dofs and 6 active contacts. We have already used the engine in a number of control applications. It will soon be made publicly available."
                },
                {
                    "title": "The Arcade Learning Environment: An Evaluation Platform for General Agents",
                    "abstract": "In this article we introduce the Arcade Learning Environment (ALE): both a challenge problem and a platform and methodology for evaluating the development of general, domain-independent AI technology. ALE provides an interface to hundreds of Atari 2600 game environments, each one different, interesting, and designed to be a challenge for human players. ALE presents significant research challenges for reinforcement learning, model learning, model-based planning, imitation learning, transfer learning, and intrinsic motivation. Most importantly, it provides a rigorous testbed for evaluating and comparing approaches to these problems. We illustrate the promise of ALE by developing and benchmarking domain-independent agents designed using well-established AI techniques for both reinforcement learning and planning. In doing so, we also propose an evaluation methodology made possible by ALE, reporting empirical results on over 55 different games. All of the software, including the benchmark agents, is publicly available."
                },
                {
                    "title": "Machine Learning and Logical Reasoning: The New Frontier (Dagstuhl Seminar 22291)",
                    "abstract": "Machine learning (ML) and logical reasoning have been the two key pillars of AI since its inception, and yet, there has been little interaction between these two sub-fields over the years. At the same time, each of them has been very influential in their own way. ML has revolutionized many sub-fields of AI including image recognition, language translation, and game playing, to name just a few. Independently, the field of logical reasoning (e.g., SAT/SMT/CP/first-order solvers and knowledge representation) has been equally impactful in many contexts in software engineering, verification, security, AI, and mathematics. Despite this progress, there are new problems, as well as opportunities, on the horizon that seem solvable only via a combination of ML and logic. One such problem that requires one to consider combinations of logic and ML is the question of reliability, robustness, and security of ML models. For example, in recent years, many adversarial attacks against ML models have been developed, demonstrating their extraordinary brittleness. How can we leverage logic-based methods to analyze such ML systems with the aim of ensuring their reliability and security? What kind of logical language do we use to specify properties of ML models? How can we ensure that ML models are explainable and interpretable? In the reverse direction, ML methods have already been successfully applied to making solvers more efficient. In particular, solvers can be modeled as complex combinations of proof systems and ML optimization methods, wherein ML-based heuristics are used to optimally select and sequence proof rules. How can we further deepen this connection between solvers and ML? Can we develop tools that automatically construct proofs for higher mathematics? This"
                },
                {
                    "title": "Stable-Baselines3: Reliable Reinforcement Learning Implementations",
                    "abstract": "Stable-Baselines3 provides open-source implementations of deep reinforcement learning (RL) algorithms in Python. The implementations have been benchmarked against reference codebases, and automated unit tests cover 95% of the code. The algorithms follow a consistent interface and are accompanied by extensive documentation, making it simple to train and compare di\ufb00erent RL algorithms. Our documentation, examples, and source-code are available at https://github.com/DLR-RM/stable-baselines3 ."
                },
                {
                    "title": "Bridging Machine Learning and Logical Reasoning by Abductive Learning",
                    "abstract": "Perception and reasoning are two representative abilities of intelligence that are integrated seamlessly during human problem-solving processes. In the area of artificial intelligence (AI), the two abilities are usually realised by machine learning and logic programming, respectively. However, the two categories of techniques were developed separately throughout most of the history of AI. In this paper, we present the abductive learning targeted at unifying the two AI paradigms in a mutually beneficial way, where the machine learning model learns to perceive primitive logic facts from data, while logical reasoning can exploit symbolic domain knowledge and correct the wrongly perceived facts for improving the machine learning models. Furthermore, we propose a novel approach to optimise the machine learning model and the logical reasoning model jointly. We demonstrate that by using abductive learning, machines can learn to recognise numbers and resolve unknown mathematical operations simultaneously from images of simple hand-written equations. Moreover, the learned models can be generalised to longer equations and adapted to different tasks, which is beyond the capability of state-of-the-art deep learning models."
                },
                {
                    "title": "Tensor Networks : Deep Learning and Logical Reasoning from Data and Knowledge \u22c6",
                    "abstract": "We proposereal logic: a uniform framework for integrating automatic learning and reasoning. Real logic is defined on a full first-o der language where formulas have truth-value in the interval [0,1] and semanti cs defined concretely on the domain of real numbers. Logical constants are interpr eted as (feature) vectors of real numbers. Real logic promotes a well-founded int egration of deductive reasoning on knowledge-bases with efficient, data-dri ven relational machine learning. We show how Real Logic can be implemented in deep Te nsor Neural Networks with the use of Google\u2019s T ENSORFLOW primitives. The paper concludes with experiments on a simple but representative exam ple of knowledge completion."
                },
                {
                    "title": "Reinforcement Learning in Board Games",
                    "abstract": "This project investigates the application of the TD(\u03bb) reinforcement learning algorithm and neural networks to the problem of producing an agent that can play board games. It provides a survey of the progress that has been made in this area over the last decade and extends this by suggesting some new possibilities for improvements (based upon theoretical and past empirical evidence). This includes the identification and a formalization (for the first time) of key game properties that are important for TD-Learning and a discussion of different methods of generate training data. Also included is the development of a TD-learning game system (including a game-independent benchmarking engine) which is capable of learning any zero-sum two-player board game. The primary purpose of the development of this system is to allow potential improvements of the system to be tested and compared in a standardized fashion. Experiments have been conduct with this system using the games Tic-Tac-Toe and Connect 4 to examine a number of different potential improvements."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a unified benchmark for evaluating logical and algorithmic reasoning capabilities in reinforcement learning (RL) agents?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of reinforcement learning, as it will provide a standardized framework for assessing the reasoning capabilities of RL agents. This benchmark could lead to improved interpretability and generalization of learned policies, enabling researchers to better understand the decision-making processes of RL agents. Furthermore, it could facilitate the development of more sophisticated algorithms that can tackle complex reasoning tasks, ultimately leading to practical applications in areas such as robotics, game playing, and automated theorem proving.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in developing a unified benchmark stem from the diverse nature of reasoning tasks and the complexity of RL environments. Naive approaches may fail due to the high-dimensional input spaces and the need for agents to perform logical reasoning in dynamic and stochastic settings. Additionally, integrating various reasoning capabilities\u2014such as causal reasoning, algorithmic reasoning, and logical deduction\u2014into a single framework presents significant theoretical and practical obstacles. Ensuring that the benchmark is comprehensive yet manageable for RL agents to learn from is a complex task.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has focused on specific aspects of reasoning or has developed benchmarks for classical machine learning, but there has been a lack of a cohesive framework that combines logical and algorithmic reasoning within the context of RL. Existing benchmarks often target narrow domains or specific tasks, which limits their applicability to broader RL scenarios. Barriers such as the absence of a standardized interface for diverse reasoning tasks and the complexity of integrating multiple reasoning types have hindered progress. Our approach aims to fill this gap by providing a unified benchmark that leverages insights from existing frameworks while addressing their limitations.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves creating a benchmark called PUZZLES, which will be implemented as a Gymnasium environment. This benchmark will include a variety of tasks that require logical and algorithmic reasoning, utilizing diverse datasets that reflect real-world scenarios. We will evaluate the performance of RL agents using metrics such as success rate, reasoning accuracy, and computational efficiency. The expected outcomes include a comprehensive assessment of RL agents' reasoning capabilities, insights into their decision-making processes, and a foundation for future research in enhancing reasoning in RL."
            }
        },
        "author_data": {
            "fb6a4bc0-1784-4beb-a611-18d60f881efa": {
                "pk": "fb6a4bc0-1784-4beb-a611-18d60f881efa",
                "name": "Benjamin Estermann",
                "collaborators": [
                    "Roger Wattenhofer",
                    "Till Aczel",
                    "Paul Doucet",
                    "Yuta Ono",
                    "Giacomo Camposampiero",
                    "Loic Houmard",
                    "Jo\u00ebl Mathys",
                    "Vivian Ziemke",
                    "Ye Wang",
                    "Markus Marks"
                ],
                "domain": [
                    "Active Learning",
                    "Unsupervised Learning",
                    "Disentangled Representations",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Bridging Diversity and Uncertainty in Active learning with Self-Supervised Pre-Training",
                        "abstract": "This study addresses the integration of diversity-based and uncertainty-based sampling strategies in active learning, particularly within the context of self-supervised pre-trained models. We introduce a straightforward heuristic called TCM that mitigates the cold start problem while maintaining strong performance across various data levels. By initially applying TypiClust for diversity sampling and subsequently transitioning to uncertainty sampling with Margin, our approach effectively combines the strengths of both strategies. Our experiments demonstrate that TCM consistently outperforms existing methods across various datasets in both low and high data regimes."
                    },
                    {
                        "title": "SUPClust: Active Learning at the Boundaries",
                        "abstract": "Active learning is a machine learning paradigm designed to optimize model performance in a setting where labeled data is expensive to acquire. In this work, we propose a novel active learning method called SUPClust that seeks to identify points at the decision boundary between classes. By targeting these points, SUPClust aims to gather information that is most informative for refining the model's prediction of complex decision regions. We demonstrate experimentally that labeling these points leads to strong model performance. This improvement is observed even in scenarios characterized by strong class imbalance."
                    },
                    {
                        "title": "DAVA: Disentangling Adversarial Variational Autoencoder",
                        "abstract": "The use of well-disentangled representations offers many advantages for downstream tasks, e.g. an increased sample efficiency, or better interpretability. However, the quality of disentangled interpretations is often highly dependent on the choice of dataset-specific hyperparameters, in particular the regularization strength. To address this issue, we introduce DAVA, a novel training procedure for variational auto-encoders. DAVA completely alleviates the problem of hyperparameter selection. We compare DAVA to models with optimal hyperparameters. Without any hyperparameter tuning, DAVA is competitive on a diverse range of commonly used datasets. Underlying DAVA, we discover a necessary condition for unsupervised disentanglement, which we call PIPE. We demonstrate the ability of PIPE to positively predict the performance of downstream models in abstract reasoning. We also thoroughly investigate correlations with existing supervised and unsupervised metrics. The code is available at https://github.com/besterma/dava."
                    },
                    {
                        "title": "Abstract Visual Reasoning Enabled by Language",
                        "abstract": "While artificial intelligence (AI) models have achieved human or even superhuman performance in many well-defined applications, they still struggle to show signs of broad and flexible intelligence. The Abstraction and Reasoning Corpus (ARC), a visual intelligence benchmark introduced by Fran\\c{c}ois Chollet, aims to assess how close AI systems are to human-like cognitive abilities. Most current approaches rely on carefully handcrafted domain-specific program searches to brute-force solutions for the tasks present in ARC. In this work, we propose a general learning-based framework for solving ARC. It is centered on transforming tasks from the vision to the language domain. This composition of language and vision allows for pre-trained models to be leveraged at each stage, enabling a shift from handcrafted priors towards the learned priors of the models. While not yet beating state-of-the-art models on ARC, we demonstrate the potential of our approach, for instance, by solving some ARC tasks that have not been solved previously."
                    },
                    {
                        "title": "What Determines the Price of NFTs?",
                        "abstract": "In the evolving landscape of digital art, Non-Fungible Tokens (NFTs) have emerged as a groundbreaking platform, bridging the realms of art and technology. NFTs serve as the foundational framework that has revolutionized the market for digital art, enabling artists to showcase and monetize their creations in unprecedented ways. NFTs combine metadata stored on the blockchain with off-chain data, such as images, to create a novel form of digital ownership. It is not fully understood how these factors come together to determine NFT prices. In this study, we analyze both on-chain and off-chain data of NFT collections trading on OpenSea to understand what influences NFT pricing. Our results show that while text and image data of the NFTs can be used to explain price variations within collections, the extracted features do not generalize to new, unseen collections. Furthermore, we find that an NFT collection's trading volume often relates to its online presence, like social media followers and website traffic."
                    },
                    {
                        "title": "Robust Disentanglement of a Few Factors at a Time",
                        "abstract": "Disentanglement is at the forefront of unsupervised learning, as disentangled representations of data improve generalization, interpretability, and performance in downstream tasks. Current unsupervised approaches remain inapplicable for real-world datasets since they are highly variable in their performance and fail to reach levels of disentanglement of (semi-)supervised approaches. We introduce population-based training (PBT) for improving consistency in training variational autoencoders (VAEs) and demonstrate the validity of this approach in a supervised setting (PBT-VAE). We then use Unsupervised Disentanglement Ranking (UDR) as an unsupervised heuristic to score models in our PBT-VAE training and show how models trained this way tend to consistently disentangle only a subset of the generative factors. Building on top of this observation we introduce the recursive rPU-VAE approach. We train the model until convergence, remove the learned factors from the dataset and reiterate. In doing so, we can label subsets of the dataset with the learned factors and consecutively use these labels to train one model that fully disentangles the whole dataset. With this approach, we show striking improvement in state-of-the-art unsupervised disentanglement performance and robustness across multiple datasets and metrics."
                    }
                ]
            },
            "ab421431-04de-478a-82c8-3c7a412688e1": {
                "pk": "ab421431-04de-478a-82c8-3c7a412688e1",
                "name": "Luca A. Lanzend\u00f6rfer",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "e1248f84-918b-44d4-9d30-ae33563cc007": {
                "pk": "e1248f84-918b-44d4-9d30-ae33563cc007",
                "name": "Yannick Niedermayr",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "bd0d25dd-afee-4fa9-a53a-48ce125f0a45": {
                "pk": "bd0d25dd-afee-4fa9-a53a-48ce125f0a45",
                "name": "Roger Wattenhofer",
                "collaborators": [
                    "Jakub Sliwinski",
                    "Beni Egressy",
                    "Lukas Faber",
                    "Peter Belcak",
                    "Oliver Richter",
                    "Lioba Heimbach",
                    "Zhao Meng",
                    "Christian Decker",
                    "Philipp Brandes",
                    "Christoph Lenzen"
                ],
                "domain": [
                    "Blockchain",
                    "Graph Neural Network",
                    "Adversarial Machine Learning",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "ABC: Proof-of-Stake without Consensus",
                        "abstract": "We introduce a new permissionless blockchain architecture called ABC. ABC is completely asynchronous, and does rely on neither randomness nor proof-of-work. ABC can be parallelized, and transactions have finality within one round trip of communication. However, ABC satisfies only a relaxed form of consensus by introducing a weaker termination property. Without full consensus, ABC cannot support certain applications, in particular ABC cannot support general smart contracts. However, many important applications do not need general smart contracts, and ABC is a better solution for these applications. In particular, ABC can implement the functionality of a cryptocurrency like Bitcoin, replacing Bitcoin's energy-hungry proof-of-work with a proof-of-stake validation."
                    },
                    {
                        "title": "A Geometry-Inspired Attack for Generating Natural Language Adversarial Examples",
                        "abstract": "Generating adversarial examples for natural language is hard, as natural language consists of discrete symbols, and examples are often of variable lengths. In this paper, we propose a geometry-inspired attack for generating natural language adversarial examples. Our attack generates adversarial examples by iteratively approximating the decision boundary of Deep Neural Networks (DNNs). Experiments on two datasets with two different models show that our attack fools natural language models with high success rates, while only replacing a few words. Human evaluation shows that adversarial examples generated by our attack are hard for humans to recognize. Further experiments show that adversarial training can improve model robustness against our attack."
                    },
                    {
                        "title": "Bailouts in Financial Networks",
                        "abstract": "We consider networks of banks with assets and liabilities. Some banks may be insolvent, and a central bank can decide which insolvent banks, if any, to bail out. We view bailouts as an optimization problem where the central bank has given resources at its disposal and an objective it wants to maximize. We show that under various assumptions and for various natural objectives this optimization problem is NP-hard, and in some cases even hard to approximate. Furthermore, we also show that given a fixed central bank bailout objective, banks in the network can make new debt contracts to increase their own market value in the event of a bailout (at the expense of the central bank)."
                    },
                    {
                        "title": "Asynchronous Neural Networks for Learning in Graphs",
                        "abstract": "This paper studies asynchronous message passing (AMP), a new paradigm for applying neural network based learning to graphs. Existing graph neural networks use the synchronous distributed computing model and aggregate their neighbors in each round, which causes problems such as oversmoothing and limits their expressiveness. On the other hand, AMP is based on the asynchronous model, where nodes react to messages of their neighbors individually. We prove that (i) AMP can simulate synchronous GNNs and that (ii) AMP can theoretically distinguish any pair of graphs. We experimentally validate AMP's expressiveness. Further, we show that AMP might be better suited to propagate messages over large distances in graphs and performs well on several graph classification benchmarks."
                    },
                    {
                        "title": "Better Incentives for Proof-of-Work",
                        "abstract": "This work proposes a novel proof-of-work blockchain incentive scheme such that, barring exogenous motivations, following the protocol is guaranteed to be the optimal strategy for miners. Our blockchain takes the form of a directed acyclic graph, resulting in improvements with respect to throughput and speed.   More importantly, for our blockchain to function, it is not expected that the miners conform to some presupposed protocol in the interest of the system's operability. Instead, our system works if miners act selfishly, trying to get the maximum possible rewards, with no consideration for the overall health of the blockchain."
                    },
                    {
                        "title": "Deterministic Graph-Walking Program Mining",
                        "abstract": "Owing to their versatility, graph structures admit representations of intricate relationships between the separate entities comprising the data. We formalise the notion of connection between two vertex sets in terms of edge and vertex features by introducing graph-walking programs. We give two algorithms for mining of deterministic graph-walking programs that yield programs in the order of increasing length. These programs characterise linear long-distance relationships between the given two vertex sets in the context of the whole graph."
                    },
                    {
                        "title": "Fast Feedforward Networks",
                        "abstract": "We break the linear link between the layer size and its inference cost by introducing the fast feedforward (FFF) architecture, a log-time alternative to feedforward networks. We demonstrate that FFFs are up to 220x faster than feedforward networks, up to 6x faster than mixture-of-experts networks, and exhibit better training properties than mixtures of experts thanks to noiseless conditional execution. Pushing FFFs to the limit, we show that they can use as little as 1% of layer neurons for inference in vision transformers while preserving 94.2% of predictive performance."
                    },
                    {
                        "title": "Bitcoin Transaction Malleability and MtGox",
                        "abstract": "In Bitcoin, transaction malleability describes the fact that the signatures that prove the ownership of bitcoins being transferred in a transaction do not provide any integrity guarantee for the signatures themselves. This allows an attacker to mount a malleability attack in which it intercepts, modifies, and rebroadcasts a transaction, causing the transaction issuer to believe that the original transaction was not confirmed. In February 2014 MtGox, once the largest Bitcoin exchange, closed and filed for bankruptcy claiming that attackers used malleability attacks to drain its accounts. In this work we use traces of the Bitcoin network for over a year preceding the filing to show that, while the problem is real, there was no widespread use of malleability attacks before the closure of MtGox."
                    },
                    {
                        "title": "Opening the Frey/Osborne Black Box: Which Tasks of a Job are Susceptible to Computerization?",
                        "abstract": "In their seminal paper, Frey and Osborne quantified the automation of jobs, by assigning each job in the O*NET database a probability to be automated. In this paper, we refine their results in the following way: Every O*NET job consists of a set of tasks, and these tasks can be related. We use a linear program to assign probabilities to tasks, such that related tasks have a similar probability and the tasks can explain the computerization probability of a job. Analyzing jobs on the level of tasks helps comprehending the results, as experts as well as laymen can more easily criticize and refine what parts of a job are susceptible to computerization."
                    },
                    {
                        "title": "CLEX: Yet Another Supercomputer Architecture?",
                        "abstract": "We propose the CLEX supercomputer topology and routing scheme. We prove that CLEX can utilize a constant fraction of the total bandwidth for point-to-point communication, at delays proportional to the sum of the number of intermediate hops and the maximum physical distance between any two nodes. Moreover, % applying an asymmetric bandwidth assignment to the links, all-to-all communication can be realized $(1+o(1))$-optimally both with regard to bandwidth and delays. This is achieved at node degrees of $n^{\\varepsilon}$, for an arbitrary small constant $\\varepsilon\\in (0,1]$. In contrast, these results are impossible in any network featuring constant or polylogarithmic node degrees. Through simulation, we assess the benefits of an implementation of the proposed communication strategy. Our results indicate that, for a million processors, CLEX can increase bandwidth utilization and reduce average routing path length by at least factors $10$ respectively $5$ in comparison to a torus network. Furthermore, the CLEX communication scheme features several other properties, such as deadlock-freedom, inherent fault-tolerance, and canonical partition into smaller subsystems."
                    },
                    {
                        "title": "PermitBFT: Exploring the Byzantine Fast-Path",
                        "abstract": "PermitBFT establishes a permissioned byzantine ledger in the partially synchronous networking model. For n replicas, PermitBFT tolerates up to f < n/3 byzantine replicas. It is the first BFT protocol to achieve a latency of just 2 message delays despite tolerating byzantine replicas throughout the \"fast track\", as long as they are not the leader. The design of PermitBFT relies on two fundamental concepts. First, in PermitBFT the participating nodes do not wait for a distinguished leader to act and subsequently confirm its actions, but send permits to the next leader proactively. Second, PermitBFT achieves a separation of the decision powers that are usually concentrated on a single leader node. A leader in PermitBFT controls which transactions to include in a new block, but not where to append the block in the block graph."
                    },
                    {
                        "title": "Learning Policies through Quantile Regression",
                        "abstract": "Policy gradient based reinforcement learning algorithms coupled with neural networks have shown success in learning complex policies in the model free continuous action space control setting. However, explicitly parameterized policies are limited by the scope of the chosen parametric probability distribution. We show that alternatively to the likelihood based policy gradient, a related objective can be optimized through advantage weighted quantile regression. Our approach models the policy implicitly in the network, which gives the agent the freedom to approximate any distribution in each action dimension, not limiting its capabilities to the commonly used unimodal Gaussian parameterization. This broader spectrum of policies makes our algorithm suitable for problems where Gaussian policies cannot fit the optimal policy. Moreover, our results on the MuJoCo physics simulator benchmarks are comparable or superior to state-of-the-art on-policy methods."
                    },
                    {
                        "title": "Normalized Attention Without Probability Cage",
                        "abstract": "Attention architectures are widely used; they recently gained renewed popularity with Transformers yielding a streak of state of the art results. Yet, the geometrical implications of softmax-attention remain largely unexplored. In this work we highlight the limitations of constraining attention weights to the probability simplex and the resulting convex hull of value vectors. We show that Transformers are sequence length dependent biased towards token isolation at initialization and contrast Transformers to simple max- and sum-pooling - two strong baselines rarely reported. We propose to replace the softmax in self-attention with normalization, yielding a hyperparameter and data-bias robust, generally applicable architecture. We support our insights with empirical results from more than 25,000 trained models. All results and implementations are made available."
                    },
                    {
                        "title": "Asynchronous Byzantine Agreement in Incomplete Networks [Technical Report]",
                        "abstract": "The Byzantine agreement problem is considered to be a core problem in distributed systems. For example, Byzantine agreement is needed to build a blockchain, a totally ordered log of records. Blockchains are asynchronous distributed systems, fault-tolerant against Byzantine nodes.   In the literature, the asynchronous byzantine agreement problem is studied in a fully connected network model where every node can directly send messages to every other node. This assumption is questionable in many real-world environments. In the reality, nodes might need to communicate by means of an incomplete network, and Byzantine nodes might not forward messages. Furthermore, Byzantine nodes might not behave correctly and, for example, corrupt messages. Therefore, in order to truly understand Byzantine Agreement, we need both ingredients: asynchrony and incomplete communication networks.   In this paper, we study the asynchronous Byzantine agreement problem in incomplete networks. A classic result by Danny Dolev proved that in a distributed system with n nodes in the presence of f Byzantine nodes, the vertex connectivity of the system communication graph should be at least (2f+1). While Dolev's result was for synchronous deterministic systems, we demonstrate that the same bound also holds for asynchronous randomized systems. We show that the bound is tight by presenting a randomized algorithm, and a matching lower bound."
                    },
                    {
                        "title": "Byzantine Agreement with Unknown Participants and Failures",
                        "abstract": "A set of mutually distrusting participants that want to agree on a common opinion must solve an instance of a Byzantine agreement problem. These problems have been extensively studied in the literature. However, most of the existing solutions assume that the participants are aware of $n$ -- the total number of participants in the system -- and $f$ -- an upper bound on the number of Byzantine participants. In this paper, we show that most of the fundamental agreement problems can be solved without affecting resiliency even if the participants do not know the values of (possibly changing) $n$ and $f$. Specifically, we consider a synchronous system where the participants have unique but not necessarily consecutive identifiers, and give Byzantine agreement algorithms for reliable broadcast, approximate agreement, rotor-coordinator, early terminating consensus and total ordering in static and dynamic systems, all with the optimal resiliency of $n> 3f$. Moreover, we show that synchrony is necessary as an agreement with probabilistic termination is impossible in a semi-synchronous or asynchronous system if the participants are unaware of $n$ and $f$."
                    },
                    {
                        "title": "Neural Status Registers",
                        "abstract": "Standard Neural Networks can learn mathematical operations, but they do not extrapolate. Extrapolation means that the model can apply to larger numbers, well beyond those observed during training. Recent architectures tackle arithmetic operations and can extrapolate; however, the equally important problem of quantitative reasoning remains unaddressed. In this work, we propose a novel architectural element, the Neural Status Register (NSR), for quantitative reasoning over numbers. Our NSR relaxes the discrete bit logic of physical status registers to continuous numbers and allows end-to-end learning with gradient descent. Experiments show that the NSR achieves solutions that extrapolate to numbers many orders of magnitude larger than those in the training set. We successfully train the NSR on number comparisons, piecewise discontinuous functions, counting in sequences, recurrently finding minimums, finding shortest paths in graphs, and comparing digits in images."
                    },
                    {
                        "title": "A Note on Optimal Fees for Constant Function Market Makers",
                        "abstract": "We suggest a framework to determine optimal trading fees for constant function market makers (CFMMs) in order to maximize liquidity provider returns. In a setting of multiple competing liquidity pools, we show that no race to the bottom occurs, but instead pure Nash equilibria of optimal fees exist. We theoretically prove the existence of these equilibria for pools using the constant product trade function used in popular CFMMs like Uniswap. We also numerically compute the equilibria for a number of examples and discuss the effects the equilibrium fees have on capital allocation among pools. Finally, we use our framework to compute optimal fees for real world pools using past trade data."
                    },
                    {
                        "title": "Eliminating Sandwich Attacks with the Help of Game Theory",
                        "abstract": "Predatory trading bots lurking in Ethereum's mempool present invisible taxation of traders on automated market makers (AMMs). AMM traders specify a slippage tolerance to indicate the maximum price movement they are willing to accept. This way, traders avoid automatic transaction failure in case of small price movements before their trade request executes. However, while a too-small slippage tolerance may lead to trade failures, a too-large slippage tolerance allows predatory trading bots to profit from sandwich attacks. These bots can extract the difference between the slippage tolerance and the actual price movement as profit.   In this work, we introduce the sandwich game to analyze sandwich attacks analytically from both the attacker and victim perspectives. Moreover, we provide a simple and highly effective algorithm that traders can use to set the slippage tolerance. We unveil that most broadcasted transactions can avoid sandwich attacks while simultaneously only experiencing a low risk of transaction failure. Thereby, we demonstrate that a constant auto-slippage cannot adjust to varying trade sizes and pool characteristics. Our algorithm outperforms the constant auto-slippage suggested by the biggest AMM, Uniswap, in all performed tests. Specifically, our algorithm repeatedly demonstrates a cost reduction exceeding a factor of 100."
                    },
                    {
                        "title": "SoK: Preventing Transaction Reordering Manipulations in Decentralized Finance",
                        "abstract": "User transactions on Ethereum's peer-to-peer network are at risk of being attacked. The smart contracts building decentralized finance (DeFi) have introduced a new transaction ordering dependency to the Ethereum blockchain. As a result, attackers can profit from front- and back-running transactions. Multiple approaches to mitigate transaction reordering manipulations have surfaced recently. However, the success of individual approaches in mitigating such attacks and their impact on the entire blockchain remains largely unstudied.   In this systematization of knowledge (SoK), we categorize and analyze state-of-the-art transaction reordering manipulation mitigation schemes. Instead of restricting our analysis to a scheme's success at preventing transaction reordering attacks, we evaluate its full impact on the blockchain. Therefore, we are able to provide a complete picture of the strengths and weaknesses of current mitigation schemes. We find that currently no scheme fully meets all the demands of the blockchain ecosystem. In fact, all approaches demonstrate unsatisfactory performance in at least one area relevant to the blockchain ecosystem."
                    },
                    {
                        "title": "Graph Neural Networks with Precomputed Node Features",
                        "abstract": "Most Graph Neural Networks (GNNs) cannot distinguish some graphs or indeed some pairs of nodes within a graph. This makes it impossible to solve certain classification tasks. However, adding additional node features to these models can resolve this problem. We introduce several such augmentations, including (i) positional node embeddings, (ii) canonical node IDs, and (iii) random features. These extensions are motivated by theoretical results and corroborated by extensive testing on synthetic subgraph detection tasks. We find that positional embeddings significantly outperform other extensions in these tasks. Moreover, positional embeddings have better sample efficiency, perform well on different graph distributions and even outperform learning with ground truth node positions. Finally, we show that the different augmentations perform competitively on established GNN benchmarks, and advise on when to use them."
                    }
                ]
            }
        }
    },
    "2406.04843": {
        "paper_data": {
            "title": "Variational Flow Matching for Graph Generation",
            "url": "http://arxiv.org/abs/2406.04843v1",
            "arxiv_id": "2406.04843",
            "authors": [
                "Floor Eijkelboom",
                "Grigory Bartosh",
                "Christian Andersson Naesseth",
                "Max Welling",
                "Jan-Willem van de Meent"
            ],
            "abstract": "We present a formulation of flow matching as variational inference, which we refer to as variational flow matching (VFM). Based on this formulation we develop CatFlow, a flow matching method for categorical data. CatFlow is easy to implement, computationally efficient, and achieves strong results on graph generation tasks. In VFM, the objective is to approximate the posterior probability path, which is a distribution over possible end points of a trajectory. We show that VFM admits both the CatFlow objective and the original flow matching objective as special cases. We also relate VFM to score-based models, in which the dynamics are stochastic rather than deterministic, and derive a bound on the model likelihood based on a reweighted VFM objective. We evaluate CatFlow on one abstract graph generation task and two molecular generation tasks. In all cases, CatFlow exceeds or matches performance of the current state-of-the-art models.",
            "introduction": "   1 Introduction  In recent years, the field of generative modeling has seen notable advancements. In image generation [35, 38], the development and refinement of diffusion-based approaches \u2014 specifically those using denoising score matching [51] \u2014 have proven effective for generation at scale [14, 46]. However, while training can be done effectively, the constrained space of sampling probability paths in a diffusion requires tailored techniques to work [44, 56]. This is in contrast to more flexible approaches such as continuous normalizing flows (CNFs) [8], that are able to learn a more general set of probability paths than diffusion models [45], at the expense of being expensive to train as they require one to solve an ODE during each training step (see e.g. [5, 39, 13]).   Recently, Lipman and collaborators [24] proposed flow matching (FM), an efficient and simulation-free approach to training CNFs. Concretely, they use a per-sample interpolation to derive a simpler objective for learning the marginal vector field that generates a desired probability path in a CNF. This formulation provides equivalent gradients without explicit knowledge of the (generally intractable) marginal vector field. This work has been extended to different geometries [7, 20] and various applications [52, 9, 12, 21]. Similar work has been proposed concurrently in [27, 1].   This paper identifies a reformulation of flow matching that we refer to as variational flow matching (VFM). In flow matching, the vector field at any point can be understood as the expected continuation toward the data distribution. In VFM, we explicitly parameterize the learned vector field as an expectation relative to a variational distribution. The objective of VFM is then to minimize the Kullback-Leibler (KL) divergence between the posterior probability path, i.e.\u00a0the distribution over possible end points (continuations) at a particular point in the space, and the variational approximation.   We show that VFM recovers the original flow matching objective when the variational approximation is Gaussian and the conditional vector field is linear in the data. Under the assumption of linearity, a solution to the VFM problem is also exact whenever the variational approximation matches the marginals of the posterior probability path, which means that we can employ a fully-factorized variational approximation without loss of generality.   While VFM provides a general formulation, our primary interest in this paper is its application to graph generation, where the data are categorical. This setting leads to a simple method that we refer to as CatFlow, in which the objective reduces to training a classifier over end points on a per-component basis. We apply CatFlow to a set of graph generation tasks, both for abstract graphs [30] and molecular generation [34, 18]. By all metrics, our results match or substantially exceed those obtained by existing methods.     2 Background   2.1 Transport Framework for Generative Modeling and CNFs  Common generative modeling approaches such as normalizing flows [36, 32] and diffusion models [14, 46] parameterize a transformation \u03c6\ud835\udf11\\varphiitalic_\u03c6 from some initial tractable probability density p0subscript\ud835\udc5d0p_{0}italic_p start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT \u2013 typically a standard Gaussian distribution \u2013 to the target data density p1subscript\ud835\udc5d1p_{1}italic_p start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT. In general, there is a trade-off between allowing \u03c6\ud835\udf11\\varphiitalic_\u03c6 to be expressive enough to model",
            "references": [
                {
                    "title": "Dirichlet Flow Matching with Applications to DNA Sequence Design",
                    "abstract": "Discrete diffusion or flow models could enable faster and more controllable sequence generation than autoregressive models. We show that na\\\"ive linear flow matching on the simplex is insufficient toward this goal since it suffers from discontinuities in the training target and further pathologies. To overcome this, we develop Dirichlet flow matching on the simplex based on mixtures of Dirichlet distributions as probability paths. In this framework, we derive a connection between the mixtures' scores and the flow's vector field that allows for classifier and classifier-free guidance. Further, we provide distilled Dirichlet flow matching, which enables one-step sequence generation with minimal performance hits, resulting in $O(L)$ speedups compared to autoregressive models. On complex DNA sequence generation tasks, we demonstrate superior performance compared to all baselines in distributional metrics and in achieving desired design targets for generated sequences. Finally, we show that our classifier-free guidance approach improves unconditional generation and is effective for generating DNA that satisfies design targets. Code is available at https://github.com/HannesStark/dirichlet-flow-matching."
                },
                {
                    "title": "Generative Flows on Discrete State-Spaces: Enabling Multimodal Flows with Applications to Protein Co-Design",
                    "abstract": "Combining discrete and continuous data is an important capability for generative models. We present Discrete Flow Models (DFMs), a new flow-based model of discrete data that provides the missing link in enabling flow-based generative models to be applied to multimodal continuous and discrete data problems. Our key insight is that the discrete equivalent of continuous space flow matching can be realized using Continuous Time Markov Chains. DFMs benefit from a simple derivation that includes discrete diffusion models as a specific instance while allowing improved performance over existing diffusion-based approaches. We utilize our DFMs method to build a multimodal flow-based modeling framework. We apply this capability to the task of protein co-design, wherein we learn a model for jointly generating protein structure and sequence. Our approach achieves state-of-the-art co-design performance while allowing the same multimodal model to be used for flexible generation of the sequence or structure."
                },
                {
                    "title": "Equivariant Flow Matching with Hybrid Probability Transport",
                    "abstract": "The generation of 3D molecules requires simultaneously deciding the categorical features~(atom types) and continuous features~(atom coordinates). Deep generative models, especially Diffusion Models (DMs), have demonstrated effectiveness in generating feature-rich geometries. However, existing DMs typically suffer from unstable probability dynamics with inefficient sampling speed. In this paper, we introduce geometric flow matching, which enjoys the advantages of both equivariant modeling and stabilized probability dynamics. More specifically, we propose a hybrid probability path where the coordinates probability path is regularized by an equivariant optimal transport, and the information between different modalities is aligned. Experimentally, the proposed method could consistently achieve better performance on multiple molecule generation benchmarks with 4.75$\\times$ speed up of sampling on average."
                },
                {
                    "title": "Flow Matching in Latent Space",
                    "abstract": "Flow matching is a recent framework to train generative models that exhibits impressive empirical performance while being relatively easier to train compared with diffusion-based models. Despite its advantageous properties, prior methods still face the challenges of expensive computing and a large number of function evaluations of off-the-shelf solvers in the pixel space. Furthermore, although latent-based generative methods have shown great success in recent years, this particular model type remains underexplored in this area. In this work, we propose to apply flow matching in the latent spaces of pretrained autoencoders, which offers improved computational efficiency and scalability for high-resolution image synthesis. This enables flow-matching training on constrained computational resources while maintaining their quality and flexibility. Additionally, our work stands as a pioneering contribution in the integration of various conditions into flow matching for conditional generation tasks, including label-conditioned image generation, image inpainting, and semantic-to-image generation. Through extensive experiments, our approach demonstrates its effectiveness in both quantitative and qualitative results on various datasets, such as CelebA-HQ, FFHQ, LSUN Church&Bedroom, and ImageNet. We also provide a theoretical control of the Wasserstein-2 distance between the reconstructed latent flow distribution and true data distribution, showing it is upper-bounded by the latent flow matching objective. Our code will be available at https://github.com/VinAIResearch/LFM.git."
                },
                {
                    "title": "Equivariant flow matching",
                    "abstract": "Normalizing flows are a class of deep generative models that are especially interesting for modeling probability distributions in physics, where the exact likelihood of flows allows reweighting to known target energy functions and computing unbiased observables. For instance, Boltzmann generators tackle the long-standing sampling problem in statistical physics by training flows to produce equilibrium samples of many-body systems such as small molecules and proteins. To build effective models for such systems, it is crucial to incorporate the symmetries of the target energy into the model, which can be achieved by equivariant continuous normalizing flows (CNFs). However, CNFs can be computationally expensive to train and generate samples from, which has hampered their scalability and practical application. In this paper, we introduce equivariant flow matching, a new training objective for equivariant CNFs that is based on the recently proposed optimal transport flow matching. Equivariant flow matching exploits the physical symmetries of the target energy for efficient, simulation-free training of equivariant CNFs. We demonstrate the effectiveness of flow matching on rotation and permutation invariant many-particle systems and a small molecule, alanine dipeptide, where for the first time we obtain a Boltzmann generator with significant sampling efficiency without relying on tailored internal coordinate featurization. Our results show that the equivariant flow matching objective yields flows with shorter integration paths, improved sampling efficiency, and higher scalability compared to existing methods."
                },
                {
                    "title": "Flow Matching for Scalable Simulation-Based Inference",
                    "abstract": "Neural posterior estimation methods based on discrete normalizing flows have become established tools for simulation-based inference (SBI), but scaling them to high-dimensional problems can be challenging. Building on recent advances in generative modeling, we here present flow matching posterior estimation (FMPE), a technique for SBI using continuous normalizing flows. Like diffusion models, and in contrast to discrete flows, flow matching allows for unconstrained architectures, providing enhanced flexibility for complex data modalities. Flow matching, therefore, enables exact density evaluation, fast training, and seamless scalability to large architectures--making it ideal for SBI. We show that FMPE achieves competitive performance on an established SBI benchmark, and then demonstrate its improved scalability on a challenging scientific problem: for gravitational-wave inference, FMPE outperforms methods based on comparable discrete flows, reducing training time by 30% with substantially improved accuracy. Our work underscores the potential of FMPE to enhance performance in challenging inference scenarios, thereby paving the way for more advanced applications to scientific problems."
                },
                {
                    "title": "Stochastic Interpolants: A Unifying Framework for Flows and Diffusions",
                    "abstract": "A class of generative models that unifies flow-based and diffusion-based methods is introduced. These models extend the framework proposed in Albergo&Vanden-Eijnden (2023), enabling the use of a broad class of continuous-time stochastic processes called `stochastic interpolants' to bridge any two arbitrary probability density functions exactly in finite time. These interpolants are built by combining data from the two prescribed densities with an additional latent variable that shapes the bridge in a flexible way. The time-dependent probability density function of the stochastic interpolant is shown to satisfy a first-order transport equation as well as a family of forward and backward Fokker-Planck equations with tunable diffusion. Upon consideration of the time evolution of an individual sample, this viewpoint immediately leads to both deterministic and stochastic generative models based on probability flow equations or stochastic differential equations with an adjustable level of noise. The drift coefficients entering these models are time-dependent velocity fields characterized as the unique minimizers of simple quadratic objective functions, one of which is a new objective for the score of the interpolant density. Remarkably, we show that minimization of these quadratic objectives leads to control of the likelihood for any of our generative models built upon stochastic dynamics. By contrast, we establish that generative models based upon a deterministic dynamics must, in addition, control the Fisher divergence between the target and the model. We also construct estimators for the likelihood and the cross-entropy of interpolant-based generative models, discuss connections with other stochastic bridges, and demonstrate that such models recover the Schr\\\"odinger bridge between the two target densities when explicitly optimizing over the interpolant."
                },
                {
                    "title": "LayoutDM: Discrete Diffusion Model for Controllable Layout Generation",
                    "abstract": "Controllable layout generation aims at synthesizing plausible arrangement of element bounding boxes with optional constraints, such as type or position of a specific element. In this work, we try to solve a broad range of layout generation tasks in a single model that is based on discrete state-space diffusion models. Our model, named Lay-outDM, naturally handles the structured layout data in the discrete representation and learns to progressively infer a noiseless layout from the initial input, where we model the layout corruption process by modality-wise discrete diffusion. For conditional generation, we propose to inject layout constraints in the form of masking or logit adjustment during inference. We show in the experiments that our Lay-outDM successfully generates high-quality layouts and outperforms both task-specific and task-agnostic baselines on several layout tasks. 11Please find the code and models at: https://cyberagentailab.github.io/layout-drn."
                },
                {
                    "title": "Where to Diffuse, How to Diffuse, and How to Get Back: Automated Learning for Multivariate Diffusions",
                    "abstract": "Diffusion-based generative models (DBGMs) perturb data to a target noise distribution and reverse this process to generate samples. The choice of noising process, or inference diffusion process, affects both likelihoods and sample quality. For example, extending the inference process with auxiliary variables leads to improved sample quality. While there are many such multivariate diffusions to explore, each new one requires significant model-specific analysis, hindering rapid prototyping and evaluation. In this work, we study Multivariate Diffusion Models (MDMs). For any number of auxiliary variables, we provide a recipe for maximizing a lower-bound on the MDMs likelihood without requiring any model-specific analysis. We then demonstrate how to parameterize the diffusion for a specified target noise distribution; these two points together enable optimizing the inference diffusion process. Optimizing the diffusion expands easy experimentation from just a few well-known processes to an automatic search over all linear diffusions. To demonstrate these ideas, we introduce two new specific diffusions as well as learn a diffusion process on the MNIST, CIFAR10, and ImageNet32 datasets. We show learned MDMs match or surpass bits-per-dims (BPDs) relative to fixed choices of diffusions for a given dataset and model architecture."
                },
                {
                    "title": "Flow Matching on General Geometries",
                    "abstract": "We propose Riemannian Flow Matching (RFM), a simple yet powerful framework for training continuous normalizing flows on manifolds. Existing methods for generative modeling on manifolds either require expensive simulation, are inherently unable to scale to high dimensions, or use approximations for limiting quantities that result in biased training objectives. Riemannian Flow Matching bypasses these limitations and offers several advantages over previous approaches: it is simulation-free on simple geometries, does not require divergence computation, and computes its target vector field in closed-form. The key ingredient behind RFM is the construction of a relatively simple premetric for defining target vector fields, which encompasses the existing Euclidean case. To extend to general geometries, we rely on the use of spectral decompositions to efficiently compute premetrics on the fly. Our method achieves state-of-the-art performance on many real-world non-Euclidean datasets, and we demonstrate tractable training on general geometries, including triangular meshes with highly non-trivial curvature and boundaries."
                },
                {
                    "title": "Flow Matching for Generative Modeling",
                    "abstract": "We introduce a new paradigm for generative modeling built on Continuous Normalizing Flows (CNFs), allowing us to train CNFs at unprecedented scale. Specifically, we present the notion of Flow Matching (FM), a simulation-free approach for training CNFs based on regressing vector fields of fixed conditional probability paths. Flow Matching is compatible with a general family of Gaussian probability paths for transforming between noise and data samples -- which subsumes existing diffusion paths as specific instances. Interestingly, we find that employing FM with diffusion paths results in a more robust and stable alternative for training diffusion models. Furthermore, Flow Matching opens the door to training CNFs with other, non-diffusion probability paths. An instance of particular interest is using Optimal Transport (OT) displacement interpolation to define the conditional probability paths. These paths are more efficient than diffusion paths, provide faster training and sampling, and result in better generalization. Training CNFs using Flow Matching on ImageNet leads to consistently better performance than alternative diffusion-based methods in terms of both likelihood and sample quality, and allows fast and reliable sample generation using off-the-shelf numerical ODE solvers."
                },
                {
                    "title": "Building Normalizing Flows with Stochastic Interpolants",
                    "abstract": "A generative model based on a continuous-time normalizing flow between any pair of base and target probability densities is proposed. The velocity field of this flow is inferred from the probability current of a time-dependent density that interpolates between the base and the target in finite time. Unlike conventional normalizing flow inference methods based the maximum likelihood principle, which require costly backpropagation through ODE solvers, our interpolant approach leads to a simple quadratic loss for the velocity itself which is expressed in terms of expectations that are readily amenable to empirical estimation. The flow can be used to generate samples from either the base or target, and to estimate the likelihood at any time along the interpolant. In addition, the flow can be optimized to minimize the path length of the interpolant density, thereby paving the way for building optimal transport maps. In situations where the base is a Gaussian density, we also show that the velocity of our normalizing flow can also be used to construct a diffusion model to sample the target as well as estimate its score. However, our approach shows that we can bypass this diffusion completely and work at the level of the probability flow with greater simplicity, opening an avenue for methods based solely on ordinary differential equations as an alternative to those based on stochastic differential equations. Benchmarking on density estimation tasks illustrates that the learned flow can match and surpass conventional continuous flows at a fraction of the cost, and compares well with diffusions on image generation on CIFAR-10 and ImageNet $32\\times32$. The method scales ab-initio ODE flows to previously unreachable image resolutions, demonstrated up to $128\\times128$."
                },
                {
                    "title": "DiGress: Discrete Denoising diffusion for graph generation",
                    "abstract": "This work introduces DiGress, a discrete denoising diffusion model for generating graphs with categorical node and edge attributes. Our model utilizes a discrete diffusion process that progressively edits graphs with noise, through the process of adding or removing edges and changing the categories. A graph transformer network is trained to revert this process, simplifying the problem of distribution learning over graphs into a sequence of node and edge classification tasks. We further improve sample quality by introducing a Markovian noise model that preserves the marginal distribution of node and edge types during diffusion, and by incorporating auxiliary graph-theoretic features. A procedure for conditioning the generation on graph-level features is also proposed. DiGress achieves state-of-the-art performance on molecular and non-molecular datasets, with up to 3x validity improvement on a planar graph dataset. It is also the first model to scale to the large GuacaMol dataset containing 1.3M drug-like molecules without the use of molecule-specific representations."
                },
                {
                    "title": "Blurring Diffusion Models",
                    "abstract": "Recently, Rissanen et al., (2022) have presented a new type of diffusion process for generative modeling based on heat dissipation, or blurring, as an alternative to isotropic Gaussian diffusion. Here, we show that blurring can equivalently be defined through a Gaussian diffusion process with non-isotropic noise. In making this connection, we bridge the gap between inverse heat dissipation and denoising diffusion, and we shed light on the inductive bias that results from this modeling choice. Finally, we propose a generalized class of diffusion models that offers the best of both standard Gaussian denoising diffusion and inverse heat dissipation, which we call Blurring Diffusion Models."
                },
                {
                    "title": "Soft Diffusion: Score Matching for General Corruptions",
                    "abstract": "We define a broader family of corruption processes that generalizes previously known diffusion models. To reverse these general diffusions, we propose a new objective called Soft Score Matching that provably learns the score function for any linear corruption process and yields state of the art results for CelebA. Soft Score Matching incorporates the degradation process in the network. Our new loss trains the model to predict a clean image, \\textit{that after corruption}, matches the diffused observation. We show that our objective learns the gradient of the likelihood under suitable regularity conditions for a family of corruption processes. We further develop a principled way to select the corruption levels for general diffusion processes and a novel sampling method that we call Momentum Sampler. We show experimentally that our framework works for general linear corruption processes, such as Gaussian blur and masking. We achieve state-of-the-art FID score $1.85$ on CelebA-64, outperforming all previous linear diffusion models. We also show significant computational benefits compared to vanilla denoising diffusion."
                },
                {
                    "title": "Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow",
                    "abstract": "We present rectified flow, a surprisingly simple approach to learning (neural) ordinary differential equation (ODE) models to transport between two empirically observed distributions \\pi_0 and \\pi_1, hence providing a unified solution to generative modeling and domain transfer, among various other tasks involving distribution transport. The idea of rectified flow is to learn the ODE to follow the straight paths connecting the points drawn from \\pi_0 and \\pi_1 as much as possible. This is achieved by solving a straightforward nonlinear least squares optimization problem, which can be easily scaled to large models without introducing extra parameters beyond standard supervised learning. The straight paths are special and preferred because they are the shortest paths between two points, and can be simulated exactly without time discretization and hence yield computationally efficient models. We show that the procedure of learning a rectified flow from data, called rectification, turns an arbitrary coupling of \\pi_0 and \\pi_1 to a new deterministic coupling with provably non-increasing convex transport costs. In addition, recursively applying rectification allows us to obtain a sequence of flows with increasingly straight paths, which can be simulated accurately with coarse time discretization in the inference phase. In empirical studies, we show that rectified flow performs superbly on image generation, image-to-image translation, and domain adaptation. In particular, on image generation and translation, our method yields nearly straight flows that give high quality results even with a single Euler discretization step."
                },
                {
                    "title": "Diffsound: Discrete Diffusion Model for Text-to-Sound Generation",
                    "abstract": "Generating sound effects that people want is an important topic. However, there are limited studies in this area for sound generation. In this study, we investigate generating sound conditioned on a text prompt and propose a novel text-to-sound generation framework that consists of a text encoder, a Vector Quantized Variational Autoencoder (VQ-VAE), a token-decoder, and a vocoder. The framework first uses the token-decoder to transfer the text features extracted from the text encoder to a mel-spectrogram with the help of VQ-VAE, and then the vocoder is used to transform the generated mel-spectrogram into a waveform. We found that the token-decoder significantly influences the generation performance. Thus, we focus on designing a good token-decoder in this study. We begin with the traditional autoregressive (AR) token-decoder. However, the AR token-decoder always predicts the mel-spectrogram tokens one by one in order, which may introduce the unidirectional bias and accumulation of errors problems. Moreover, with the AR token-decoder, the sound generation time increases linearly with the sound duration. To overcome the shortcomings introduced by AR token-decoders, we propose a non-autoregressive token-decoder based on the discrete diffusion model, named Diffsound. Specifically, the Diffsound model predicts all of the mel-spectrogram tokens in one step and then refines the predicted tokens in the next step, so the best-predicted results can be obtained by iteration. Our experiments show that our proposed Diffsound model not only produces better generation results when compared with the AR token-decoder but also has a faster generation speed, i.e., MOS: 3.56 v.s 2.786."
                },
                {
                    "title": "Matching Normalizing Flows and Probability Paths on Manifolds",
                    "abstract": "Continuous Normalizing Flows (CNFs) are a class of generative models that transform a prior distribution to a model distribution by solving an ordinary differential equation (ODE). We propose to train CNFs on manifolds by minimizing probability path divergence (PPD), a novel family of divergences between the probability density path generated by the CNF and a target probability density path. PPD is formulated using a logarithmic mass conservation formula which is a linear first order partial differential equation relating the log target probabilities and the CNF's defining vector field. PPD has several key benefits over existing methods: it sidesteps the need to solve an ODE per iteration, readily applies to manifold data, scales to high dimensions, and is compatible with a large family of target paths interpolating pure noise and data in finite time. Theoretically, PPD is shown to bound classical probability divergences. Empirically, we show that CNFs learned by minimizing PPD achieve state-of-the-art results in likelihoods and sample quality on existing low-dimensional manifold benchmarks, and is the first example of a generative model to scale to moderately high dimensional manifolds."
                },
                {
                    "title": "Generative Modelling With Inverse Heat Dissipation",
                    "abstract": "While diffusion models have shown great success in image generation, their noise-inverting generative process does not explicitly consider the structure of images, such as their inherent multi-scale nature. Inspired by diffusion models and the empirical success of coarse-to-fine modelling, we propose a new diffusion-like model that generates images through stochastically reversing the heat equation, a PDE that locally erases fine-scale information when run over the 2D plane of the image. We interpret the solution of the forward heat equation with constant additive noise as a variational approximation in the diffusion latent variable model. Our new model shows emergent qualitative properties not seen in standard diffusion models, such as disentanglement of overall colour and shape in images. Spectral analysis on natural images highlights connections to diffusion models and reveals an implicit coarse-to-fine inductive bias in them."
                },
                {
                    "title": "Diffusion probabilistic modeling of protein backbones in 3D for the motif-scaffolding problem",
                    "abstract": "Construction of a scaffold structure that supports a desired motif, conferring protein function, shows promise for the design of vaccines and enzymes. But a general solution to this motif-scaffolding problem remains open. Current machine-learning techniques for scaffold design are either limited to unrealistically small scaffolds (up to length 20) or struggle to produce multiple diverse scaffolds. We propose to learn a distribution over diverse and longer protein backbone structures via an E(3)-equivariant graph neural network. We develop SMCDiff to efficiently sample scaffolds from this distribution conditioned on a given motif; our algorithm is the first to theoretically guarantee conditional samples from a diffusion model in the large-compute limit. We evaluate our designed backbones by how well they align with AlphaFold2-predicted structures. We show that our method can (1) sample scaffolds up to 80 residues and (2) achieve structurally diverse scaffolds for a fixed motif."
                },
                {
                    "title": "Fast Sampling of Diffusion Models with Exponential Integrator",
                    "abstract": "The past few years have witnessed the great success of Diffusion models~(DMs) in generating high-fidelity samples in generative modeling tasks. A major limitation of the DM is its notoriously slow sampling procedure which normally requires hundreds to thousands of time discretization steps of the learned diffusion process to reach the desired accuracy. Our goal is to develop a fast sampling method for DMs with a much less number of steps while retaining high sample quality. To this end, we systematically analyze the sampling procedure in DMs and identify key factors that affect the sample quality, among which the method of discretization is most crucial. By carefully examining the learned diffusion process, we propose Diffusion Exponential Integrator Sampler~(DEIS). It is based on the Exponential Integrator designed for discretizing ordinary differential equations (ODEs) and leverages a semilinear structure of the learned diffusion process to reduce the discretization error. The proposed method can be applied to any DMs and can generate high-fidelity samples in as few as 10 steps. In our experiments, it takes about 3 minutes on one A6000 GPU to generate $50k$ images from CIFAR10. Moreover, by directly using pre-trained DMs, we achieve the state-of-art sampling performance when the number of score function evaluation~(NFE) is limited, e.g., 4.17 FID with 10 NFEs, 3.37 FID, and 9.74 IS with only 15 NFEs on CIFAR10. Code is available at https://github.com/qsh-zh/deis"
                },
                {
                    "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
                    "abstract": "Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples."
                },
                {
                    "title": "SPECTRE : Spectral Conditioning Helps to Overcome the Expressivity Limits of One-shot Graph Generators",
                    "abstract": "We approach the graph generation problem from a spectral perspective by first generating the dominant parts of the graph Laplacian spectrum and then building a graph matching these eigenvalues and eigenvectors. Spectral conditioning allows for direct modeling of the global and local graph structure and helps to overcome the expressivity and mode collapse issues of one-shot graph generators. Our novel GAN, called SPECTRE, enables the one-shot generation of much larger graphs than previously possible with one-shot models. SPECTRE outperforms state-of-the-art deep autoregressive generators in terms of modeling fidelity, while also avoiding expensive sequential generation and dependence on node ordering. A case in point, in sizable synthetic and real-world graphs SPECTRE achieves a 4-to-170 fold improvement over the best competitor that does not overfit and is 23-to-30 times faster than autoregressive generators."
                },
                {
                    "title": "Equivariant Diffusion for Molecule Generation in 3D",
                    "abstract": "This work introduces a diffusion model for molecule generation in 3D that is equivariant to Euclidean transformations. Our E(3) Equivariant Diffusion Model (EDM) learns to denoise a diffusion process with an equivariant network that jointly operates on both continuous (atom coordinates) and categorical features (atom types). In addition, we provide a probabilistic analysis which admits likelihood computation of molecules using our model. Experimentally, the proposed method significantly outperforms previous 3D molecular generative methods regarding the quality of generated samples and efficiency at training time."
                },
                {
                    "title": "Flow-Matching: Efficient Coarse-Graining of Molecular Dynamics without Forces.",
                    "abstract": "Coarse-grained (CG) molecular simulations have become a standard tool to study molecular processes on time and length scales inaccessible to all-atom simulations. Parametrizing CG force fields to match all-atom simulations has mainly relied on force-matching or relative entropy minimization, which require many samples from costly simulations with all-atom or CG resolutions, respectively. Here we present flow-matching, a new training method for CG force fields that combines the advantages of both methods by leveraging normalizing flows, a generative deep learning method. Flow-matching first trains a normalizing flow to represent the CG probability density, which is equivalent to minimizing the relative entropy without requiring iterative CG simulations. Subsequently, the flow generates samples and forces according to the learned distribution in order to train the desired CG free energy model via force-matching. Even without requiring forces from the all-atom simulations, flow-matching outperforms classical force-matching by an order of magnitude in terms of data efficiency and produces CG models that can capture the folding and unfolding transitions of small proteins."
                },
                {
                    "title": "Score-based Generative Modeling of Graphs via the System of Stochastic Differential Equations",
                    "abstract": "Generating graph-structured data requires learning the underlying distribution of graphs. Yet, this is a challenging problem, and the previous graph generative methods either fail to capture the permutation-invariance property of graphs or cannot sufficiently model the complex dependency between nodes and edges, which is crucial for generating real-world graphs such as molecules. To overcome such limitations, we propose a novel score-based generative model for graphs with a continuous-time framework. Specifically, we propose a new graph diffusion process that models the joint distribution of the nodes and edges through a system of stochastic differential equations (SDEs). Then, we derive novel score matching objectives tailored for the proposed diffusion process to estimate the gradient of the joint log-density with respect to each component, and introduce a new solver for the system of SDEs to efficiently sample from the reverse diffusion process. We validate our graph generation method on diverse datasets, on which it either achieves significantly superior or competitive performance to the baselines. Further analysis shows that our method is able to generate molecules that lie close to the training distribution yet do not violate the chemical valency rule, demonstrating the effectiveness of the system of SDEs in modeling the node-edge relationships. Our code is available at https://github.com/harryjo97/GDSS."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Moser Flow: Divergence-based Generative Modeling on Manifolds",
                    "abstract": "We are interested in learning generative models for complex geometries described via manifolds, such as spheres, tori, and other implicit surfaces. Current extensions of existing (Euclidean) generative models are restricted to specific geometries and typically suffer from high computational costs. We introduce Moser Flow (MF), a new class of generative models within the family of continuous normalizing flows (CNF). MF also produces a CNF via a solution to the change-of-variable formula, however differently from other CNF methods, its model (learned) density is parameterized as the source (prior) density minus the divergence of a neural network (NN). The divergence is a local, linear differential operator, easy to approximate and calculate on manifolds. Therefore, unlike other CNFs, MF does not require invoking or backpropagating through an ODE solver during training. Furthermore, representing the model density explicitly as the divergence of a NN rather than as a solution of an ODE facilitates learning high fidelity densities. Theoretically, we prove that MF constitutes a universal density approximator under suitable assumptions. Empirically, we demonstrate for the first time the use of flow models for sampling from general curved surfaces and achieve significant improvements in density estimation, sample quality, and training complexity over existing CNFs on challenging synthetic geometries and real-world benchmarks from the earth and climate sciences."
                },
                {
                    "title": "Structured Denoising Diffusion Models in Discrete State-Spaces",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) (Ho et al. 2020) have shown impressive results on image and waveform generation in continuous state spaces. Here, we introduce Discrete Denoising Diffusion Probabilistic Models (D3PMs), diffusion-like generative models for discrete data that generalize the multinomial diffusion model of Hoogeboom et al. 2021, by going beyond corruption processes with uniform transition probabilities. This includes corruption with transition matrices that mimic Gaussian kernels in continuous space, matrices based on nearest neighbors in embedding space, and matrices that introduce absorbing states. The third allows us to draw a connection between diffusion models and autoregressive and mask-based generative models. We show that the choice of transition matrix is an important design decision that leads to improved results in image and text domains. We also introduce a new loss function that combines the variational lower bound with an auxiliary cross entropy loss. For text, this model class achieves strong results on character-level text generation while scaling to large vocabularies on LM1B. On the image dataset CIFAR-10, our models approach the sample quality and exceed the log-likelihood of the continuous-space DDPM model."
                },
                {
                    "title": "GraphDF: A Discrete Flow Model for Molecular Graph Generation",
                    "abstract": "We consider the problem of molecular graph generation using deep models. While graphs are discrete, most existing methods use continuous latent variables, resulting in inaccurate modeling of discrete graph structures. In this work, we propose GraphDF, a novel discrete latent variable model for molecular graph generation based on normalizing flow methods. GraphDF uses invertible modulo shift transforms to map discrete latent variables to graph nodes and edges. We show that the use of discrete latent variables reduces computational costs and eliminates the negative effect of dequantization. Comprehensive experimental results show that GraphDF outperforms prior methods on random generation, property optimization, and constrained optimization tasks."
                },
                {
                    "title": "Maximum Likelihood Training of Score-Based Diffusion Models",
                    "abstract": "Score-based diffusion models synthesize samples by reversing a stochastic process that diffuses data to noise, and are trained by minimizing a weighted combination of score matching losses. The log-likelihood of score-based diffusion models can be tractably computed through a connection to continuous normalizing flows, but log-likelihood is not directly optimized by the weighted combination of score matching losses. We show that for a specific weighting scheme, the objective upper bounds the negative log-likelihood, thus enabling approximate maximum likelihood training of score-based diffusion models. We empirically observe that maximum likelihood training consistently improves the likelihood of score-based diffusion models across multiple datasets, stochastic processes, and model architectures. Our best models achieve negative log-likelihoods of 2.83 and 3.76 bits/dim on CIFAR-10 and ImageNet 32x32 without any data augmentation, on a par with state-of-the-art autoregressive models on these tasks."
                },
                {
                    "title": "A Generalization of Transformer Networks to Graphs",
                    "abstract": "We propose a generalization of transformer neural network architecture for arbitrary graphs. The original transformer was designed for Natural Language Processing (NLP), which operates on fully connected graphs representing all connections between the words in a sequence. Such architecture does not leverage the graph connectivity inductive bias, and can perform poorly when the graph topology is important and has not been encoded into the node features. We introduce a graph transformer with four new properties compared to the standard model. First, the attention mechanism is a function of the neighborhood connectivity for each node in the graph. Second, the positional encoding is represented by the Laplacian eigenvectors, which naturally generalize the sinusoidal positional encodings often used in NLP. Third, the layer normalization is replaced by a batch normalization layer, which provides faster training and better generalization performance. Finally, the architecture is extended to edge feature representation, which can be critical to tasks s.a. chemistry (bond type) or link prediction (entity relationship in knowledge graphs). Numerical experiments on a graph benchmark demonstrate the performance of the proposed graph transformer architecture. This work closes the gap between the original transformer, which was designed for the limited case of line graphs, and graph neural networks, that can work with arbitrary graphs. As our architecture is simple and generic, we believe it can be used as a black box for future applications that wish to consider transformer and graphs."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "Graph networks for molecular design",
                    "abstract": "Deep learning methods applied to chemistry can be used to accelerate the discovery of new molecules. This work introduces GraphINVENT, a platform developed for graph-based molecular design using graph neural networks (GNNs). GraphINVENT uses a tiered deep neural network architecture to probabilistically generate new molecules a single bond at a time. All models implemented in GraphINVENT can quickly learn to build molecules resembling the training set molecules without any explicit programming of chemical rules. The models have been benchmarked using the MOSES distribution-based metrics, showing how GraphINVENT models compare well with state-of-the-art generative models. This work compares six different GNN-based generative models in GraphINVENT, and shows that ultimately the gated-graph neural network performs best against the metrics considered here."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Categorical Normalizing Flows via Continuous Transformations",
                    "abstract": "Despite their popularity, to date, the application of normalizing flows on categorical data stays limited. The current practice of using dequantization to map discrete data to a continuous space is inapplicable as categorical data has no intrinsic order. Instead, categorical data have complex and latent relations that must be inferred, like the synonymy between words. In this paper, we investigate Categorical Normalizing Flows, that is normalizing flows for categorical data. By casting the encoding of categorical data in continuous space as a variational inference problem, we jointly optimize the continuous representation and the model likelihood. To maintain unique decoding, we learn a partitioning of the latent space by factorizing the posterior. Meanwhile, the complex relations between the categorical variables are learned by the ensuing normalizing flow, thus maintaining a close-to exact likelihood estimate and making it possible to scale up to a large number of categories. Based on Categorical Normalizing Flows, we propose GraphCNF a permutation-invariant generative model on graphs, outperforming both one-shot and autoregressive flow-based state-of-the-art on molecule generation."
                },
                {
                    "title": "MoFlow: An Invertible Flow Model for Generating Molecular Graphs",
                    "abstract": "Generating molecular graphs with desired chemical properties driven by deep graph generative models provides a very promising way to accelerate drug discovery process. Such graph generative models usually consist of two steps: learning latent representations and generation of molecular graphs. However, to generate novel and chemically-valid molecular graphs from latent representations is very challenging because of the chemical constraints and combinatorial complexity of molecular graphs. In this paper, we propose MoFlow, a flow-based graph generative model to learn invertible mappings between molecular graphs and their latent representations. To generate molecular graphs, our MoFlow first generates bonds (edges) through a Glow based model, then generates atoms (nodes) given bonds by a novel graph conditional flow, and finally assembles them into a chemically valid molecular graph with a posthoc validity correction. Our MoFlow has merits including exact and tractable likelihood training, efficient one-pass embedding and generation, chemical validity guarantees, 100% reconstruction of training data, and good generalization ability. We validate our model by four tasks: molecular graph generation and reconstruction, visualization of the continuous latent space, property optimization, and constrained property optimization. Our MoFlow achieves state-of-the-art performance, which implies its potential efficiency and effectiveness to explore large chemical space for drug discovery."
                },
                {
                    "title": "GraphAF: a Flow-based Autoregressive Model for Molecular Graph Generation",
                    "abstract": "Molecular graph generation is a fundamental problem for drug discovery and has been attracting growing attention. The problem is challenging since it requires not only generating chemically valid molecular structures but also optimizing their chemical properties in the meantime. Inspired by the recent progress in deep generative models, in this paper we propose a flow-based autoregressive model for graph generation called GraphAF. GraphAF combines the advantages of both autoregressive and flow-based approaches and enjoys: (1) high model flexibility for data density estimation; (2) efficient parallel computation for training; (3) an iterative sampling process, which allows leveraging chemical domain knowledge for valency checking. Experimental results show that GraphAF is able to generate 68% chemically valid molecules even without chemical knowledge rules and 100% valid molecules with chemical rules. The training process of GraphAF is two times faster than the existing state-of-the-art approach GCPN. After fine-tuning the model for goal-directed property optimization with reinforcement learning, GraphAF achieves state-of-the-art performance on both chemical property optimization and constrained property optimization."
                },
                {
                    "title": "Normalizing Flows for Probabilistic Modeling and Inference",
                    "abstract": "Normalizing flows provide a general mechanism for defining expressive probability distributions, only requiring the specification of a (usually simple) base distribution and a series of bijective transformations. There has been much recent work on normalizing flows, ranging from improving their expressive power to expanding their application. We believe the field has now matured and is in need of a unified perspective. In this review, we attempt to provide such a perspective by describing flows through the lens of probabilistic modeling and inference. We place special emphasis on the fundamental principles of flow design, and discuss foundational topics such as expressive power and computational trade-offs. We also broaden the conceptual framing of flows by relating them to more general probability transformations. Lastly, we summarize the use of flows for tasks such as generative modeling, approximate inference, and supervised learning."
                },
                {
                    "title": "Efficient Graph Generation with Graph Recurrent Attention Networks",
                    "abstract": "We propose a new family of efficient and expressive deep generative models of graphs, called Graph Recurrent Attention Networks (GRANs). Our model generates graphs one block of nodes and associated edges at a time. The block size and sampling stride allow us to trade off sample quality for efficiency. Compared to previous RNN-based graph generative models, our framework better captures the auto-regressive conditioning between the already-generated and to-be-generated parts of the graph using Graph Neural Networks (GNNs) with attention. This not only reduces the dependency on node ordering but also bypasses the long-term bottleneck caused by the sequential nature of RNNs. Moreover, we parameterize the output distribution per block using a mixture of Bernoulli, which captures the correlations among generated edges within the block. Finally, we propose to handle node orderings in generation by marginalizing over a family of canonical orderings. On standard benchmarks, we achieve state-of-the-art time efficiency and sample quality compared to previous models. Additionally, we show our model is capable of generating large graphs of up to 5K nodes with good quality. Our code is released at: \\url{https://github.com/lrjconan/GRAN}."
                },
                {
                    "title": "GraphNVP: An Invertible Flow Model for Generating Molecular Graphs",
                    "abstract": "We propose GraphNVP, the first invertible, normalizing flow-based molecular graph generation model. We decompose the generation of a graph into two steps: generation of (i) an adjacency tensor and (ii) node attributes. This decomposition yields the exact likelihood maximization on graph-structured data, combined with two novel reversible flows. We empirically demonstrate that our model efficiently generates valid molecular graphs with almost no duplicated molecules. In addition, we observe that the learned latent space can be used to generate molecules with desired chemical properties."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "Constrained Graph Variational Autoencoders for Molecule Design",
                    "abstract": "Graphs are ubiquitous data structures for representing interactions between entities. With an emphasis on the use of graphs to represent chemical molecules, we explore the task of learning to generate graphs that conform to a distribution observed in training data. We propose a variational autoencoder model in which both encoder and decoder are graph-structured. Our decoder assumes a sequential ordering of graph extension steps and we discuss and analyze design choices that mitigate the potential downsides of this linearization. Experiments compare our approach with a wide range of baselines on the molecule generation task and show that our method is more successful at matching the statistics of the original dataset on semantically important metrics. Furthermore, we show that by using appropriate shaping of the latent space, our model allows us to design molecules that are (locally) optimal in desired properties."
                },
                {
                    "title": "Fr\u00e9chet ChemNet Distance: A Metric for Generative Models for Molecules in Drug Discovery",
                    "abstract": "The new wave of successful generative models in machine learning has increased the interest in deep learning driven de novo drug design. However, method comparison is difficult because of various flaws of the currently employed evaluation metrics. We propose an evaluation metric for generative models called Fr\u00e9chet ChemNet distance (FCD). The advantage of the FCD over previous metrics is that it can detect whether generated molecules are diverse and have similar chemical and biological properties as real molecules."
                },
                {
                    "title": "GraphRNN: Generating Realistic Graphs with Deep Auto-regressive Models",
                    "abstract": "Modeling and generating graphs is fundamental for studying networks in biology, engineering, and social sciences. However, modeling complex distributions over graphs and then efficiently sampling from these distributions is challenging due to the non-unique, high-dimensional nature of graphs and the complex, non-local dependencies that exist between edges in a given graph. Here we propose GraphRNN, a deep autoregressive model that addresses the above challenges and approximates any distribution of graphs with minimal assumptions about their structure. GraphRNN learns to generate graphs by training on a representative set of graphs and decomposes the graph generation process into a sequence of node and edge formations, conditioned on the graph structure generated so far. \nIn order to quantitatively evaluate the performance of GraphRNN, we introduce a benchmark suite of datasets, baselines and novel evaluation metrics based on Maximum Mean Discrepancy, which measure distances between sets of graphs. Our experiments show that GraphRNN significantly outperforms all baselines, learning to generate diverse graphs that match the structural characteristics of a target set, while also scaling to graphs 50 times larger than previous deep models."
                },
                {
                    "title": "Variational Inference with Normalizing Flows",
                    "abstract": "The choice of approximate posterior distribution is one of the core problems in variational inference. Most applications of variational inference employ simple families of posterior approximations in order to allow for efficient inference, focusing on mean-field or other simple structured approximations. This restriction has a significant impact on the quality of inferences made using variational methods. We introduce a new approach for specifying flexible, arbitrarily complex and scalable approximate posterior distributions. Our approximations are distributions constructed through a normalizing flow, whereby a simple initial density is transformed into a more complex one by applying a sequence of invertible transformations until a desired level of complexity is attained. We use this view of normalizing flows to develop categories of finite and infinitesimal flows and provide a unified view of approaches for constructing rich posterior approximations. We demonstrate that the theoretical advantages of having posteriors that better match the true posterior, combined with the scalability of amortized variational approaches, provides a clear improvement in performance and applicability of variational inference."
                },
                {
                    "title": "ZINC: A Free Tool to Discover Chemistry for Biology",
                    "abstract": "ZINC is a free public resource for ligand discovery. The database contains over twenty million commercially available molecules in biologically relevant representations that may be downloaded in popular ready-to-dock formats and subsets. The Web site also enables searches by structure, biological activity, physical property, vendor, catalog number, name, and CAS number. Small custom subsets may be created, edited, shared, docked, downloaded, and conveyed to a vendor for purchase. The database is maintained and curated for a high purchasing success rate and is freely available at zinc.docking.org."
                },
                {
                    "title": "A Connection Between Score Matching and Denoising Autoencoders",
                    "abstract": "Denoising autoencoders have been previously shown to be competitive alternatives to restricted Boltzmann machines for unsupervised pretraining of each layer of a deep architecture. We show that a simple denoising autoencoder training criterion is equivalent to matching the score (with respect to the data) of a specific energy-based model to that of a nonparametric Parzen density estimator of the data. This yields several useful insights. It defines a proper probabilistic model for the denoising autoencoder technique, which makes it in principle possible to sample from them or rank examples by their energy. It suggests a different way to apply score matching that is related to learning to denoise and does not require computing second derivatives. It justifies the use of tied weights between the encoder and decoder and suggests ways to extend the success of denoising autoencoders to a larger family of energy-based models."
                },
                {
                    "title": "Collective Classification in Network Data",
                    "abstract": "Many real-world applications produce networked data such as the world-wide web (hypertext documents connected via hyperlinks), social networks (for example, people connected by friendship links), communication networks (computers connected via communication links) and biological networks (for example, protein interaction networks). A recent focus in machine learning research has been to extend traditional machine learning classification techniques to classify nodes in such networks. In this article, we provide a brief introduction to this area of research and how it has progressed during the past decade. We introduce four of the most widely used inference algorithms for classifying networked data and empirically compare them on both synthetic and real-world data."
                },
                {
                    "title": "Scalable Reversible Generative Models with Free-form Continuous Dynamics",
                    "abstract": "A promising class of generative models maps points from a simple distribution to a complex distribution through an invertible neural network. Likelihood-based training of these models requires restricting their architectures to allow cheap computation of Jacobian determinants. Alternatively, the Jacobian trace can be used if the transformation is specified by an ordinary differential equation. In this paper, we use Hutchinson\u2019s trace estimator to give a scalable unbiased estimate of the log-density. The result is a continuous-time invertible generative model with unbiased density estimation and one-pass sampling, while allowing unrestricted neural network architectures. We demonstrate our approach on highdimensional density estimation, image generation, and variational inference, achieving the state-of-the-art among exact likelihood methods with efficient sampling."
                },
                {
                    "title": "BRENDA, the enzyme database: updates and major new developments",
                    "abstract": "BRENDA (BRaunschweig ENzyme DAtabase) represents a comprehensive collection of enzyme and metabolic information, based on primary literature. The database contains data from at least 83,000 different enzymes from 9800 different organisms, classified in approximately 4200 EC numbers. BRENDA includes biochemical and molecular information on classification and nomenclature, reaction and specificity, functional parameters, occurrence, enzyme structure, application, engineering, stability, disease, isolation and preparation, links and literature references. The data are extracted and evaluated from approximately 46,000 references, which are linked to PubMed as long as the reference is cited in PubMed. In the past year BRENDA has undergone major changes including a large increase in updating speed with >50% of all data updated in 2002 or in the first half of 2003, the development of a new EC-tree browser, a taxonomy-tree browser, a chemical substructure search engine for ligand structure, the development of controlled vocabulary, an ontology for some information fields and a thesaurus for ligand names. The database is accessible free of charge to the academic community at http://www.brenda. uni-koeln.de."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively improve the training and sampling efficiency of continuous normalizing flows (CNFs) for generative modeling, particularly in the context of graph generation?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the field of generative modeling, as it addresses the limitations of existing methods in terms of training efficiency and flexibility. By improving CNFs, we can enhance their applicability to various domains, including molecular generation and abstract graph generation. This research could lead to more efficient algorithms that can generate complex data structures, thereby influencing future research directions and practical applications in fields such as drug discovery, network analysis, and beyond.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the inherent complexity of accurately modeling the marginal vector field in CNFs, which is often intractable. Naive approaches may fail because they do not account for the need to balance expressiveness and computational efficiency. Additionally, the requirement to solve ordinary differential equations (ODEs) during training adds significant computational overhead, making it difficult to scale these models effectively. Overcoming these technical obstacles requires innovative reformulations and efficient training methodologies.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on either diffusion models or traditional CNFs, often overlooking the potential of combining their strengths. Limitations in existing solutions include the computational expense of training CNFs and the lack of efficient methods for learning the marginal vector field. Barriers such as the complexity of the mathematical formulations and the need for explicit knowledge of intractable distributions have hindered progress. Our approach, variational flow matching (VFM), improves upon prior work by providing a more general and efficient framework for training CNFs without the need for explicit marginal vector field calculations.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology, variational flow matching (VFM), involves parameterizing the learned vector field as an expectation relative to a variational distribution and minimizing the Kullback-Leibler (KL) divergence between the posterior probability path and the variational approximation. We will apply this method to graph generation tasks using categorical data, employing a classifier to train on end points on a per-component basis. The expected outcomes include improved training efficiency and performance metrics that match or exceed those of existing methods, as demonstrated through applications in both abstract graph and molecular generation tasks."
            }
        },
        "author_data": {
            "b952e9e7-9f97-46f2-a39f-3f05c90b4634": {
                "pk": "b952e9e7-9f97-46f2-a39f-3f05c90b4634",
                "name": "Floor Eijkelboom",
                "collaborators": [
                    "Erik Bekkers",
                    "Rob Hesselink",
                    "Cong Liu",
                    "David Ruhe",
                    "Patrick Forr\u00e9",
                    "Veljko Kova\u010d",
                    "Erik J. Bekkers",
                    "Pietro Li\u00f2",
                    "Michael Bronstein",
                    "Francesco Di Giovanni"
                ],
                "domain": [
                    "Geometric Deep Learning",
                    "Equivariant Neural Networks",
                    "Message Passing Networks",
                    "Topological Data Analysis"
                ],
                "publications": [
                    {
                        "title": "E(n) Equivariant Message Passing Simplicial Networks",
                        "abstract": "This paper presents $\\mathrm{E}(n)$ Equivariant Message Passing Simplicial Networks (EMPSNs), a novel approach to learning on geometric graphs and point clouds that is equivariant to rotations, translations, and reflections. EMPSNs can learn high-dimensional simplex features in graphs (e.g. triangles), and use the increase of geometric information of higher-dimensional simplices in an $\\mathrm{E}(n)$ equivariant fashion. EMPSNs simultaneously generalize $\\mathrm{E}(n)$ Equivariant Graph Neural Networks to a topologically more elaborate counterpart and provide an approach for including geometric information in Message Passing Simplicial Networks. The results indicate that EMPSNs can leverage the benefits of both approaches, leading to a general increase in performance when compared to either method. Furthermore, the results suggest that incorporating geometric information serves as an effective measure against over-smoothing in message passing networks, especially when operating on high-dimensional simplicial structures. Last, we show that EMPSNs are on par with state-of-the-art approaches for learning on geometric graphs."
                    },
                    {
                        "title": "Clifford Group Equivariant Simplicial Message Passing Networks",
                        "abstract": "We introduce Clifford Group Equivariant Simplicial Message Passing Networks, a method for steerable E(n)-equivariant message passing on simplicial complexes. Our method integrates the expressivity of Clifford group-equivariant layers with simplicial message passing, which is topologically more intricate than regular graph message passing. Clifford algebras include higher-order objects such as bivectors and trivectors, which express geometric features (e.g., areas, volumes) derived from vectors. Using this knowledge, we represent simplex features through geometric products of their vertices. To achieve efficient simplicial message passing, we share the parameters of the message network across different dimensions. Additionally, we restrict the final message to an aggregation of the incoming messages from different dimensions, leading to what we term shared simplicial message passing. Experimental results show that our method is able to outperform both equivariant and simplicial graph neural networks on a variety of geometric tasks."
                    },
                    {
                        "title": "E(n) Equivariant Message Passing Cellular Networks",
                        "abstract": "This paper introduces E(n) Equivariant Message Passing Cellular Networks (EMPCNs), an extension of E(n) Equivariant Graph Neural Networks to CW-complexes. Our approach addresses two aspects of geometric message passing networks: 1) enhancing their expressiveness by incorporating arbitrary cells, and 2) achieving this in a computationally efficient way with a decoupled EMPCNs technique. We demonstrate that EMPCNs achieve close to state-of-the-art performance on multiple tasks without the need for steerability, including many-body predictions and motion capture. Moreover, ablation studies confirm that decoupled EMPCNs exhibit stronger generalization capabilities than their non-topologically informed counterparts. These findings show that EMPCNs can be used as a scalable and expressive framework for higher-order message passing in geometric and topological graphs"
                    },
                    {
                        "title": "Can strong structural encoding reduce the importance of Message Passing?",
                        "abstract": "The most prevalent class of neural networks operating on graphs are message passing neural networks (MPNNs), in which the representation of a node is updated iteratively by aggregating information in the 1-hop neighborhood. Since this paradigm for computing node embeddings may prevent the model from learning coarse topological structures, the initial features are often augmented with structural information of the graph, typically in the form of Laplacian eigenvectors or Random Walk transition probabilities. In this work, we explore the contribution of message passing when strong structural encodings are provided. We introduce a novel way of modeling the interaction between feature and structural information based on their tensor product rather than the standard concatenation. The choice of interaction is compared in common scenarios and in settings where the capacity of the message-passing layer is severely reduced and ultimately the message-passing phase is removed altogether. Our results indicate that using tensor-based encodings is always at least on par with the concatenation-based encoding and that it makes the model much more robust when the message passing layers are removed, on some tasks incurring almost no drop in performance. This suggests that the importance of message passing is limited when the model can construct strong structural encodings."
                    }
                ]
            },
            "b282fbcc-c975-437e-b855-166416351f10": {
                "pk": "b282fbcc-c975-437e-b855-166416351f10",
                "name": "Grigory Bartosh",
                "collaborators": [
                    "Dmitry Vetrov",
                    "Christian A. Naesseth",
                    "Dmitry Molchanov",
                    "Andrey Okhotin",
                    "Vladimir Arkhipkin",
                    "Viktor Ohanesian",
                    "Aibek Alanov",
                    "Alexander Shabalin",
                    "Viacheslav Meshchaninov",
                    "Egor Chimbulatov"
                ],
                "domain": [
                    "Generative Modeling",
                    "Diffusion Models",
                    "Machine Learning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Neural Diffusion Models",
                        "abstract": "Diffusion models have shown remarkable performance on many generative tasks. Despite recent success, most diffusion models are restricted in that they only allow linear transformation of the data distribution. In contrast, broader family of transformations can potentially help train generative distributions more efficiently, simplifying the reverse process and closing the gap between the true negative log-likelihood and the variational approximation. In this paper, we present Neural Diffusion Models (NDMs), a generalization of conventional diffusion models that enables defining and learning time-dependent non-linear transformations of data. We show how to optimise NDMs using a variational bound in a simulation-free setting. Moreover, we derive a time-continuous formulation of NDMs, which allows fast and reliable inference using off-the-shelf numerical ODE and SDE solvers. Finally, we demonstrate the utility of NDMs with learnable transformations through experiments on standard image generation benchmarks, including CIFAR-10, downsampled versions of ImageNet and CelebA-HQ. NDMs outperform conventional diffusion models in terms of likelihood and produce high-quality samples."
                    },
                    {
                        "title": "Neural Flow Diffusion Models: Learnable Forward Process for Improved Diffusion Modelling",
                        "abstract": "Conventional diffusion models typically relies on a fixed forward process, which implicitly defines complex marginal distributions over latent variables. This can often complicate the reverse process' task in learning generative trajectories, and results in costly inference for diffusion models. To address these limitations, we introduce Neural Flow Diffusion Models (NFDM), a novel framework that enhances diffusion models by supporting a broader range of forward processes beyond the standard Gaussian. We also propose a novel parameterization technique for learning the forward process. Our framework provides an end-to-end, simulation-free optimization objective, effectively minimizing a variational upper bound on the negative log-likelihood. Experimental results demonstrate NFDM's strong performance, evidenced by state-of-the-art likelihood estimation. Furthermore, we investigate NFDM's capacity for learning generative dynamics with specific characteristics, such as deterministic straight lines trajectories, and demonstrate how the framework may be adopted for learning bridges between two distributions. The results underscores NFDM's versatility and its potential for a wide range of applications."
                    },
                    {
                        "title": "Star-Shaped Denoising Diffusion Probabilistic Models",
                        "abstract": "Denoising Diffusion Probabilistic Models (DDPMs) provide the foundation for the recent breakthroughs in generative modeling. Their Markovian structure makes it difficult to define DDPMs with distributions other than Gaussian or discrete. In this paper, we introduce Star-Shaped DDPM (SS-DDPM). Its star-shaped diffusion process allows us to bypass the need to define the transition probabilities or compute posteriors. We establish duality between star-shaped and specific Markovian diffusions for the exponential family of distributions and derive efficient algorithms for training and sampling from SS-DDPMs. In the case of Gaussian distributions, SS-DDPM is equivalent to DDPM. However, SS-DDPMs provide a simple recipe for designing diffusion models with distributions such as Beta, von Mises$\\unicode{x2013}$Fisher, Dirichlet, Wishart and others, which can be especially useful when data lies on a constrained manifold. We evaluate the model in different settings and find it competitive even on image data, where Beta SS-DDPM achieves results comparable to a Gaussian DDPM. Our implementation is available at https://github.com/andrey-okhotin/star-shaped ."
                    },
                    {
                        "title": "TEncDM: Understanding the Properties of Diffusion Model in the Space of Language Model Encodings",
                        "abstract": "This paper presents the Text Encoding Diffusion Model (TEncDM), a novel approach to diffusion modeling that operates in the space of pre-trained language model encodings. In contrast to traditionally used embeddings, encodings integrate contextual information. In our approach, we also employ a transformer-based decoder, specifically designed to incorporate context in the token prediction process. We conduct a comprehensive examination of the influence of the encoder, decoder, noise scheduler, and self-conditioning on zero-shot generation. Furthermore, we compare TEncDM with previous approaches on three conditional text generation tasks: QQP, XSum, and Wiki-Auto. The results show that TEncDM exhibits superior performance compared to existing non-autoregressive diffusion models."
                    }
                ]
            },
            "fb50e310-fa8c-4477-a537-f859ac5ad11c": {
                "pk": "fb50e310-fa8c-4477-a537-f859ac5ad11c",
                "name": "Christian Andersson Naesseth",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "037432b3-53b4-45cc-8858-ad335ede5596": {
                "pk": "037432b3-53b4-45cc-8858-ad335ede5596",
                "name": "Max Welling",
                "collaborators": [],
                "domain": [
                    "Bayesian Inference",
                    "Approximate Inference",
                    "Machine Learning",
                    "Stochastic Processes"
                ],
                "publications": [
                    {
                        "title": "Herding Dynamic Weights for Partially Observed Random Field Models",
                        "abstract": "Learning the parameters of a (potentially partially observable) random field model is intractable in general. Instead of focussing on a single optimal parameter value we propose to treat parameters as dynamical quantities. We introduce an algorithm to generate complex dynamics for parameters and (both visible and hidden) state vectors. We show that under certain conditions averages computed over trajectories of the proposed dynamical system converge to averages computed over the data. Our \"herding dynamics\" does not require expensive operations such as exponentiation and is fully deterministic."
                    },
                    {
                        "title": "On the Choice of Regions for Generalized Belief Propagation",
                        "abstract": "Generalized belief propagation (GBP) has proven to be a promising technique for approximate inference tasks in AI and machine learning. However, the choice of a good set of clusters to be used in GBP has remained more of an art then a science until this day. This paper proposes a sequential approach to adding new clusters of nodes and their interactions (i.e. \"regions\") to the approximation. We first review and analyze the recently introduced region graphs and find that three kinds of operations (\"split\", \"merge\" and \"death\") leave the free energy and (under some conditions) the fixed points of GBP invariant. This leads to the notion of \"weakly irreducible\" regions as the natural candidates to be added to the approximation. Computational complexity of the GBP algorithm is controlled by restricting attention to regions with small \"region-width\". Combining the above with an efficient (i.e. local in the graph) measure to predict the improved accuracy of GBP leads to the sequential \"region pursuit\" algorithm for adding new regions bottom-up to the region graph. Experiments show that this algorithm can indeed perform close to optimally."
                    },
                    {
                        "title": "Exploiting the Statistics of Learning and Inference",
                        "abstract": "When dealing with datasets containing a billion instances or with simulations that require a supercomputer to execute, computational resources become part of the equation. We can improve the efficiency of learning and inference by exploiting their inherent statistical nature. We propose algorithms that exploit the redundancy of data relative to a model by subsampling data-cases for every update and reasoning about the uncertainty created in this process. In the context of learning we propose to test for the probability that a stochastically estimated gradient points more than 180 degrees in the wrong direction. In the context of MCMC sampling we use stochastic gradients to improve the efficiency of MCMC updates, and hypothesis tests based on adaptive mini-batches to decide whether to accept or reject a proposed parameter update. Finally, we argue that in the context of likelihood free MCMC one needs to store all the information revealed by all simulations, for instance in a Gaussian process. We conclude that Bayesian methods will remain to play a crucial role in the era of big data and big simulations, but only if we overcome a number of computational challenges."
                    }
                ]
            },
            "28c58d91-2bd4-4966-83e9-4dd3f25fd8d3": {
                "pk": "28c58d91-2bd4-4966-83e9-4dd3f25fd8d3",
                "name": "Jan-Willem van de Meent",
                "collaborators": [
                    "Frank Wood",
                    "Brooks Paige",
                    "Alexander Morozov",
                    "Ellak Somfai",
                    "Eric Sultan",
                    "Wim van Saarloos",
                    "Andy J. Sederman",
                    "Lynn F. Gladden",
                    "Raymond E. Goldstein",
                    "David Tolpin"
                ],
                "domain": [
                    "Bayesian Inference",
                    "Probabilistic Programming",
                    "Computational Fluid Dynamics",
                    "Gaussian Processes"
                ],
                "publications": [
                    {
                        "title": "Tempering by Subsampling",
                        "abstract": "In this paper we demonstrate that tempering Markov chain Monte Carlo samplers for Bayesian models by recursively subsampling observations without replacement can improve the performance of baseline samplers in terms of effective sample size per computation. We present two tempering by subsampling algorithms, subsampled parallel tempering and subsampled tempered transitions. We provide an asymptotic analysis of the computational cost of tempering by subsampling, verify that tempering by subsampling costs less than traditional tempering, and demonstrate both algorithms on Bayesian approaches to learning the mean of a high dimensional multivariate Normal and estimating Gaussian process hyperparameters."
                    },
                    {
                        "title": "Coherent structures in Dissipative Particle Dynamics simulations of the transition to turbulence in compressible shear flows",
                        "abstract": "We present simulations of coherent structures in compressible flows near the transition to turbulence using the Dissipative Particle Dynamics (DPD) method. The structures we find are remarkably consistent with experimental observations and DNS simulations of incompressible flows, despite a difference in Mach number of several orders of magnitude. The bifurcation from the laminar flow is bistable and shifts to higher Reynolds numbers when the fluid becomes more compressible. This work underlines the robustness of coherent structures in the transition to turbulence and illustrates the ability of particle-based methods to reproduce complex non-linear instabilities."
                    },
                    {
                        "title": "Measurement of Cytoplasmic Streaming in Chara Corallina by Magnetic Resonance Velocimetry",
                        "abstract": "In aquatic plants such as the Characean algae, the force generation that drives cyclosis is localized within the cytoplasm, yet produces fluid flows throughout the vacuole. For this to occur the tonoplast must transmit hydrodynamic shear efficiently. Here, using magnetic resonance velocimetry, we present the first whole-cell measurements of the cross-sectional longitudinal velocity field in Chara corallina and show that it is in quantitative agreement with a recent theoretical analysis of rotational cytoplasmic streaming driven by bidirectional helical forcing in the cytoplasm, with direct shear transmission by the tonoplast."
                    },
                    {
                        "title": "Black-Box Policy Search with Probabilistic Programs",
                        "abstract": "In this work, we explore how probabilistic programs can be used to represent policies in sequential decision problems. In this formulation, a probabilistic program is a black-box stochastic simulator for both the problem domain and the agent. We relate classic policy gradient techniques to recently introduced black-box variational methods which generalize to probabilistic program inference. We present case studies in the Canadian traveler problem, Rock Sample, and a benchmark for optimal diagnosis inspired by Guess Who. Each study illustrates how programs can efficiently represent policies using moderate numbers of parameters."
                    },
                    {
                        "title": "Particle Gibbs with Ancestor Sampling for Probabilistic Programs",
                        "abstract": "Particle Markov chain Monte Carlo techniques rank among current state-of-the-art methods for probabilistic program inference. A drawback of these techniques is that they rely on importance resampling, which results in degenerate particle trajectories and a low effective sample size for variables sampled early in a program. We here develop a formalism to adapt ancestor resampling, a technique that mitigates particle degeneracy, to the probabilistic programming setting. We present empirical results that demonstrate nontrivial performance gains."
                    },
                    {
                        "title": "Probabilistic structure discovery in time series data",
                        "abstract": "Existing methods for structure discovery in time series data construct interpretable, compositional kernels for Gaussian process regression models. While the learned Gaussian process model provides posterior mean and variance estimates, typically the structure is learned via a greedy optimization procedure. This restricts the space of possible solutions and leads to over-confident uncertainty estimates. We introduce a fully Bayesian approach, inferring a full posterior over structures, which more reliably captures the uncertainty of the model."
                    }
                ]
            }
        }
    },
    "2312.03878": {
        "paper_data": {
            "title": "Domain constraints improve risk prediction when outcome data is missing",
            "url": "http://arxiv.org/abs/2312.03878v3",
            "arxiv_id": "2312.03878",
            "authors": [
                "Sidhika Balachandar",
                "Nikhil Garg",
                "Emma Pierson"
            ],
            "abstract": "Machine learning models are often trained to predict the outcome resulting from a human decision. For example, if a doctor decides to test a patient for disease, will the patient test positive? A challenge is that historical decision-making determines whether the outcome is observed: we only observe test outcomes for patients doctors historically tested. Untested patients, for whom outcomes are unobserved, may differ from tested patients along observed and unobserved dimensions. We propose a Bayesian model class which captures this setting. The purpose of the model is to accurately estimate risk for both tested and untested patients. Estimating this model is challenging due to the wide range of possibilities for untested patients. To address this, we propose two domain constraints which are plausible in health settings: a prevalence constraint, where the overall disease prevalence is known, and an expertise constraint, where the human decision-maker deviates from purely risk-based decision-making only along a constrained feature set. We show theoretically and on synthetic data that domain constraints improve parameter inference. We apply our model to a case study of cancer risk prediction, showing that the model's inferred risk predicts cancer diagnoses, its inferred testing policy captures known public health policies, and it can identify suboptimalities in test allocation. Though our case study is in healthcare, our analysis reveals a general class of domain constraints which can improve model estimation in many settings.",
            "introduction": "   1 Introduction  Machine learning models are often trained to predict outcomes in settings where a human makes a high-stakes decision. In criminal justice, a judge decides whether to release a defendant prior to trial, and models are trained to predict whether the defendant will fail to appear or commit a crime if released\u00a0(Lakkaraju et\u00a0al., 2017; Jung et\u00a0al., 2020a; Kleinberg et\u00a0al., 2018). In lending, a creditor decides whether to grant an applicant a loan, and models are trained to predict whether the applicant will repay\u00a0(Bj\u00f6rkegren & Grissen, 2020; Crook & Banasik, 2004). In healthcare\u2014the setting we focus on in this paper\u2014a doctor decides whether to test a patient for disease, and models are trained to predict whether the patient will test positive\u00a0(Jehi et\u00a0al., 2020; McDonald et\u00a0al., 2021; Mullainathan & Obermeyer, 2022). Machine learning predictions help guide decision-making in all these settings. A model which predicts a patient\u2019s risk of disease can help allocate tests to the highest-risk patients, and also identify suboptimalities in human decision-making: for example, testing patients at low risk of disease, or failing to test high risk patients\u00a0(Mullainathan & Obermeyer, 2022).   A fundamental challenge in all these settings is that historical decision-making determines whether the outcome is observed. In criminal justice, release outcomes are only observed for defendants judges have historically released. In lending, loan repayments are only observed for applicants historically granted loans. In healthcare, test outcomes are only observed for patients doctors have historically tested. This is problematic because the model must make accurate predictions for the entire population, not just the historically tested population. Learning only from the tested population also risks introducing bias against underserved populations who are less likely to get medical tests partly due to worse healthcare access\u00a0(Chen et\u00a0al., 2021; Pierson, 2020; Servik, 2020; Jain et\u00a0al., 2023). Thus, there is a challenging distribution shift between the tested and untested populations. The two populations may differ both along observables recorded in the data and unobservables known to the human decision-maker but unrecorded in the data. For example, tested patients may have more symptoms recorded than untested patients\u2014but they may also differ on unobservables, like how much pain they are in or how sick they look, which are known to the doctor but are not available for the model. This setting, referred to as the selective labels setting\u00a0(Lakkaraju et\u00a0al., 2017), occurs in high-stakes domains including healthcare, hiring, insurance, lending, education, welfare services, government inspections, tax auditing, recommender systems, wildlife protection, and criminal justice and has been the subject of substantial academic interest (see \u00a76 for related work).   Without further constraints on the data generating process, there is a wide range of possibilities for the untested patients. They could all have the disease or never have the disease. However, selective labels settings often have domain-specific constraints which would allow us to limit the range of possibilities. For example, in medical settings, we might know the prevalence of a disease in the population. Recent distribution shift literature has shown that generic methods generally do not perform well across all distribution shifts and that domain-specific constraints can improve generalization\u00a0(Gulrajani & Lopez-Paz, 2021; Koh",
            "references": [
                {
                    "title": "Detecting disparities in police deployments using dashcam data",
                    "abstract": "Large-scale policing data is vital for detecting inequity in police behavior and policing algorithms. However, one important type of policing data remains largely unavailable within the United States: aggregated police deployment data capturing which neighborhoods have the heaviest police presences. Here we show that disparities in police deployment levels can be quantified by detecting police vehicles in dashcam images of public street scenes. Using a dataset of 24,803,854 dashcam images from rideshare drivers in New York City, we find that police vehicles can be detected with high accuracy (average precision 0.82, AUC 0.99) and identify 233,596 images which contain police vehicles. There is substantial inequality across neighborhoods in police vehicle deployment levels. The neighborhood with the highest deployment levels has almost 20 times higher levels than the neighborhood with the lowest. Two strikingly different types of areas experience high police vehicle deployments \u2014 1) dense, higher-income, commercial areas and 2) lower-income neighborhoods with higher proportions of Black and Hispanic residents. We discuss the implications of these disparities for policing equity and for algorithms trained on policing data."
                },
                {
                    "title": "Algorithmic Censoring in Dynamic Learning Systems",
                    "abstract": "Dynamic learning systems subject to selective labeling exhibit censoring, i.e. persistent negative predictions assigned to one or more subgroups of points. In applications like consumer finance, this results in groups of applicants that are persistently denied and thus never enter into the training data. In this work, we formalize censoring, demonstrate how it can arise, and highlight difficulties in detection. We consider safeguards against censoring \u2013 recourse and randomized-exploration \u2013 both of which ensure we collect labels for points that would otherwise go unobserved. The resulting techniques allow examples from censored groups to enter into the training data and correct the model. Our results highlight the otherwise unmeasured harms of censoring and demonstrate the effectiveness of mitigation strategies across a range of data generating processes."
                },
                {
                    "title": "Coarse race data conceals disparities in clinical risk score performance",
                    "abstract": "Healthcare data in the United States often records only a patient's coarse race group: for example, both Indian and Chinese patients are typically coded as\"Asian.\"It is unknown, however, whether this coarse coding conceals meaningful disparities in the performance of clinical risk scores across granular race groups. Here we show that it does. Using data from 418K emergency department visits, we assess clinical risk score performance disparities across 26 granular groups for three outcomes, five risk scores, and four performance metrics. Across outcomes and metrics, we show that the risk scores exhibit significant granular performance disparities within coarse race groups. In fact, variation in performance within coarse groups often *exceeds* the variation between coarse groups. We explore why these disparities arise, finding that outcome rates, feature distributions, and the relationships between features and outcomes all vary significantly across granular groups. Our results suggest that healthcare providers, hospital systems, and machine learning researchers should strive to collect, release, and use granular race data in place of coarse race data, and that existing analyses may significantly underestimate racial disparities in performance."
                },
                {
                    "title": "Race Corrections in Clinical Models: Examining Family History and Cancer Risk",
                    "abstract": "Despite ethical and historical arguments for removing race from algorithms, the consequences of this practice for medical decision-making remain unclear. Eliminating race corrections can have unintended consequences: it can decrease both accuracy and equity, because structural racism distorts the measurement of many variables. We illustrate this using the example of family history of colorectal cancer commonly used in cancer screening. Using data from the Southern Community Cohort Study (SCCS), established to study cancer disparities, we analyze 77,836 adults with no history of colorectal cancer at baseline. First, we compare the prognostic value of self-reported family history for self-reported Black vs. White participants and find that family history is strongly predictive of cancer risk for White, but not Black participants. We then create two screening algorithms that model colorectal cancer risk. The baseline algorithm is race-blind, while the race-corrected algorithm adds Black race both as a main effect and as an interaction family history. The race-corrected algorithm improves upon the race-blind algorithm in a likelihood ratio test (p-value: <0.001). In addition, both race terms are significant: the main effect (p-value: <0.001) captures the fact that, among participants with no family history, Black participants have 1.34x higher odds than White participants of developing colorectal cancer. The interaction term between race and family history is also statistically significant (p-value 0.012), capturing the differential predictive value of family history across race groups. As a result, the race-corrected algorithm includes more Black participants among the predicted high-risk group. Our case study illustrates a much broader point: missing and erroneous data is ubiquitous in medicine, at rates which vary by race group, and race correction can help address this problem."
                },
                {
                    "title": "Out-of-Domain Robustness via Targeted Augmentations",
                    "abstract": "Models trained on one set of domains often suffer performance drops on unseen domains, e.g., when wildlife monitoring models are deployed in new camera locations. In this work, we study principles for designing data augmentations for out-of-domain (OOD) generalization. In particular, we focus on real-world scenarios in which some domain-dependent features are robust, i.e., some features that vary across domains are predictive OOD. For example, in the wildlife monitoring application above, image backgrounds vary across camera locations but indicate habitat type, which helps predict the species of photographed animals. Motivated by theoretical analysis on a linear setting, we propose targeted augmentations, which selectively randomize spurious domain-dependent features while preserving robust ones. We prove that targeted augmentations improve OOD performance, allowing models to generalize better with fewer domains. In contrast, existing approaches such as generic augmentations, which fail to randomize domain-dependent features, and domain-invariant augmentations, which randomize all domain-dependent features, both perform poorly OOD. In experiments on three real-world datasets, we show that targeted augmentations set new states-of-the-art for OOD performance by 3.2-15.2 percentage points."
                },
                {
                    "title": "Counterfactual Prediction Under Outcome Measurement Error",
                    "abstract": "Across domains such as medicine, employment, and criminal justice, predictive models often target labels that imperfectly reflect the outcomes of interest to experts and policymakers. For example, clinical risk assessments deployed to inform physician decision-making often predict measures of healthcare utilization (e.g., costs, hospitalization) as a proxy for patient medical need. These proxies can be subject to outcome measurement error when they systematically differ from the target outcome they are intended to measure. However, prior modeling efforts to characterize and mitigate outcome measurement error overlook the fact that the decision being informed by a model often serves as a risk-mitigating intervention that impacts the target outcome of interest and its recorded proxy. Thus, in these settings, addressing measurement error requires counterfactual modeling of treatment effects on outcomes. In this work, we study intersectional threats to model reliability introduced by outcome measurement error, treatment effects, and selection bias from historical decision-making policies. We develop an unbiased risk minimization method which, given knowledge of proxy measurement error properties, corrects for the combined effects of these challenges. We also develop a method for estimating treatment-dependent measurement error parameters when these are unknown in advance. We demonstrate the utility of our approach theoretically and via experiments on real-world data from randomized controlled trials conducted in healthcare and employment domains. As importantly, we demonstrate that models correcting for outcome measurement error or treatment effects alone suffer from considerable reliability limitations. Our work underscores the importance of considering intersectional threats to model validity during the design and evaluation of predictive models for decision support."
                },
                {
                    "title": "Addressing Disparities in Breast Cancer Screening: A Review",
                    "abstract": "Black women are less likely to be referred for genetic counseling and testing than White women.6'7 The American College of Radiolo-gy (ACR) and Society of Breast Imag-ing (SBI) recommend that all women be assessed for breast cancer risk by age 30 to determine if screening earlier than age 40 is needed.8 Up-dates to and expansion of screening guidelines by the ACR have been effective in improving awareness in populations at elevated risk of breast cancer and in guiding screening and identification of patients. Black and Hispanic women, for example, experience disproportion-ate poverty, inadequate access to higher-quality healthcare, absence of or inadequate health insurance, and differences in utilization of screening mammography.14,15 Unequal access to and utilization of screening mammog-raphy often lead to delays in detection, diagnosis, and treatment, amplifying disparities in patient outcomes.16,17 In addition, calculators and models that assess and quantify lifetime risk for breast cancer may underestimate risk among racial and ethnic minority women and potentially discourage them from seeking enhanced screen-ing and/or active surveillance.18 Black and Hispanic women are less likely than White women to be up to date with screening recommendations and to receive timely follow-up on results of abnormal screening and diagnos-tic mammography.19 Lawson, et al, found that structural healthcare system factors had the greatest impact on time-to-biopsy among racial and ethnic groups. Compared with White women, Asian, Hispanic, and Black women were at increased risk of failing to receive a biopsy within 30 days of an abnor-mal screening mammogram;Black women had the highest and most persistent risk of diagnostic delays of 90 days or longer.20 Longer intervals between screening and diagnosis lead to declines in screening benefits and more advanced cancers at diagnosis.21 Insufficient Health Insurance Coverage Screening mammography benefit-ed from the Patient Protection and Affordable Care Act, which expanded mandatory health insurance cover-age for routine preventive services and prohibits cost sharing. [...]breast imaging specialists can engage in community outreach initiatives that \"increase breast cancer awareness in underserved communities, identify high-risk women for risk-reduction strategies, and develop programs that assist women in navigating their breast care."
                },
                {
                    "title": "Learning from a Biased Sample",
                    "abstract": "The empirical risk minimization approach to data-driven decision making requires access to training data drawn under the same conditions as those that will be faced when the decision rule is deployed. However, in a number of settings, we may be concerned that our training sample is biased in the sense that some groups (characterized by either observable or unobservable attributes) may be under- or over-represented relative to the general population; and in this setting empirical risk minimization over the training set may fail to yield rules that perform well at deployment. We propose a model of sampling bias called conditional $\\Gamma$-biased sampling, where observed covariates can affect the probability of sample selection arbitrarily much but the amount of unexplained variation in the probability of sample selection is bounded by a constant factor. Applying the distributionally robust optimization framework, we propose a method for learning a decision rule that minimizes the worst-case risk incurred under a family of test distributions that can generate the training distribution under $\\Gamma$-biased sampling. We apply a result of Rockafellar and Uryasev to show that this problem is equivalent to an augmented convex risk minimization problem. We give statistical guarantees for learning a model that is robust to sampling bias via the method of sieves, and propose a deep learning algorithm whose loss function captures our robust learning target. We empirically validate our proposed method in a case study on prediction of mental health scores from health survey data and a case study on ICU length of stay prediction."
                },
                {
                    "title": "Efficient Truncated Linear Regression with Unknown Noise Variance",
                    "abstract": "Truncated linear regression is a classical challenge in statistics, wherein a label, \ud835\udc66 = \ud835\udc64 \ud835\udc47 \ud835\udc65 + \ud835\udf00 , and its corresponding feature vector, \ud835\udc65 \u2208 R \ud835\udc58 , are only observed if the label falls in some subset \ud835\udc46 \u2286 R ; otherwise the existence of the pair ( \ud835\udc65, \ud835\udc66 ) is hidden from observation. Linear regression with truncated observations has remained a challenge, in its general form, since the early works of Tobin (1958); Amemiya (1973). When the distribution of the error is normal with known variance, recent work of Daskalakis et al. (2019) provides computationally and statistically ef\ufb01cient estimators of the linear model, \ud835\udc64 . In this paper, we provide the \ufb01rst computationally and statistically ef\ufb01cient estimators for truncated linear regression when the noise variance is unknown, estimating both the linear model and the variance of the noise. Our estimator is based on an ef\ufb01cient implementation of Projected Stochastic Gradient Descent on the negative log-likelihood of the truncated sample. Importantly, we show that the error of our estimates is asymptotically normal, and we use this to provide explicit con\ufb01dence regions for our estimates."
                },
                {
                    "title": "Equity in Resident Crowdsourcing: Measuring Under-reporting without Ground Truth Data",
                    "abstract": "Modern city governance relies heavily on crowdsourcing (or \"co-production\") to identify problems such as downed trees and power-lines. A major concern in these systems is that residents do not report problems at the same rates, leading to an inequitable allocation of government resources. However, measuring such under-reporting is a difficult statistical task, as, almost by definition, we do not observe incidents that are not reported. Thus, distinguishing between low reporting rates and low ground-truth incident rates is challenging. We develop a method to identify (heterogeneous) reporting rates, without using external (proxy) ground truth data. Our insight is that rates on duplicate reports about the same incident can be leveraged, to turn the question into a standard Poisson rate estimation task---even though the full incident reporting interval is also unobserved. We apply our method to over 100,000 resident reports made to the New York City Department of Parks and Recreation, finding that there are substantial spatial and socio-economic disparities in reporting rates, even after controlling for incident characteristics."
                },
                {
                    "title": "Modeling the Data-Generating Process is Necessary for Out-of-Distribution Generalization",
                    "abstract": "Recent empirical studies on domain generalization (DG) have shown that DG algorithms that perform well on some distribution shifts fail on others, and no state-of-the-art DG algorithm performs consistently well on all shifts. Moreover, real-world data often has multiple distribution shifts over different attributes; hence we introduce multi-attribute distribution shift datasets and find that the accuracy of existing DG algorithms falls even further. To explain these results, we provide a formal characterization of generalization under multi-attribute shifts using a canonical causal graph. Based on the relationship between spurious attributes and the classification label, we obtain realizations of the canonical causal graph that characterize common distribution shifts and show that each shift entails different independence constraints over observed variables. As a result, we prove that any algorithm based on a single, fixed constraint cannot work well across all shifts, providing theoretical evidence for mixed empirical results on DG algorithms. Based on this insight, we develop Causally Adaptive Constraint Minimization (CACM), an algorithm that uses knowledge about the data-generating process to adaptively identify and apply the correct independence constraints for regularization. Results on fully synthetic, MNIST, small NORB, and Waterbirds datasets, covering binary and multi-valued attributes and labels, show that adaptive dataset-dependent constraints lead to the highest accuracy on unseen domains whereas incorrect constraints fail to do so. Our results demonstrate the importance of modeling the causal relationships inherent in the data-generating process."
                },
                {
                    "title": "Trucks Don\u2019t Mean Trump: Diagnosing Human Error in Image Analysis",
                    "abstract": "Algorithms provide powerful tools for detecting and dissecting human bias and error. Here, we develop machine learning methods to to analyze how humans err in a particular high-stakes task: image interpretation. We leverage a unique dataset of 16,135,392 human predictions of whether a neighborhood voted for Donald Trump or Joe Biden in the 2020 US election, based on a Google Street View image. We show that by training a machine learning estimator of the Bayes optimal decision for each image, we can provide an actionable decomposition of human error into bias, variance, and noise terms, and further identify specific features (like pickup trucks) which lead humans astray. Our methods can be applied to ensure that human-in-the-loop decision-making is accurate and fair and are also applicable to black-box algorithmic systems."
                },
                {
                    "title": "Integrating Reward Maximization and Population Estimation: Sequential Decision-Making for Internal Revenue Service Audit Selection",
                    "abstract": "We introduce a new setting, optimize-and-estimate structured bandits. Here, a policy must select a batch of arms, each characterized by its own context, that would allow it to both maximize reward and maintain an accurate (ideally unbiased) population estimate of the reward. This setting is inherent to many public and private sector applications and often requires handling delayed feedback, small data, and distribution shifts. We demonstrate its importance on real data from the United States Internal Revenue Service (IRS). The IRS performs yearly audits of the tax base. Two of its most important objectives are to identify suspected misreporting and to estimate the \"tax gap\" -- the global difference between the amount paid and true amount owed. Based on a unique collaboration with the IRS, we cast these two processes as a unified optimize-and-estimate structured bandit. We analyze optimize-and-estimate approaches to the IRS problem and propose a novel mechanism for unbiased population estimation that achieves rewards comparable to baseline approaches. This approach has the potential to improve audit efficacy, while maintaining policy-relevant estimates of the tax gap. This has important social consequences given that the current tax gap is estimated at nearly half a trillion dollars. We suggest that this problem setting is fertile ground for further research and we highlight its interesting challenges. The results of this and related research are currently being incorporated into the continual improvement of the IRS audit selection methods."
                },
                {
                    "title": "Extending the WILDS Benchmark for Unsupervised Adaptation",
                    "abstract": "Machine learning systems deployed in the wild are often trained on a source distribution but deployed on a di\ufb00erent target distribution. Unlabeled data can be a powerful point of leverage for mitigating these distribution shifts, as it is frequently much more available than labeled data and can often be obtained from distributions beyond the source distribution as well. However, existing distribution shift benchmarks with unlabeled data do not re\ufb02ect the breadth of scenarios that arise in real-world applications. In this work, we present the Wilds 2.0 update, which extends 8 of the 10 datasets in the Wilds benchmark of distribution shifts to include curated unlabeled data that would be realistically obtainable in deployment. These datasets span a wide range of applications (from histology to wildlife conservation), tasks (classi\ufb01cation, regression, and detection), and modalities (photos, satellite images, microscope slides, text, molecular graphs). The update maintains consistency with the original Wilds benchmark by using identical labeled training, validation, and test sets, as well as the evaluation metrics. On these datasets, we systematically benchmark state-of-the-art methods that leverage unlabeled data, including domain-invariant, self-training, and self-supervised methods, and show that their success on Wilds is limited. To facilitate method development and evaluation, we provide an open-source package that automates data loading and contains all of the model architectures and methods used in this paper. Code and leaderboards are available at https://wilds.stanford.edu ."
                },
                {
                    "title": "A Fine-Grained Analysis on Distribution Shift",
                    "abstract": "Robustness to distribution shifts is critical for deploying machine learning models in the real world. Despite this necessity, there has been little work in defining the underlying mechanisms that cause these shifts and evaluating the robustness of algorithms across multiple, different distribution shifts. To this end, we introduce a framework that enables fine-grained analysis of various distribution shifts. We provide a holistic analysis of current state-of-the-art methods by evaluating 19 distinct methods grouped into five categories across both synthetic and real-world datasets. Overall, we train more than 85K models. Our experimental framework can be easily extended to include new methods, shifts, and datasets. We find, unlike previous work~\\citep{Gulrajani20}, that progress has been made over a standard ERM baseline; in particular, pretraining and augmentations (learned or heuristic) offer large gains in many cases. However, the best methods are not consistent over different datasets and shifts."
                },
                {
                    "title": "FADE: FAir Double Ensemble Learning for Observable and Counterfactual Outcomes",
                    "abstract": "Methods for building fair predictors often involve tradeoffs between fairness and accuracy and between different fairness criteria. Recent work seeks to characterize these tradeoffs in specific problem settings, but these methods often do not accommodate users who wish to improve the fairness of an existing benchmark model without sacrificing accuracy, or vice versa. These results are also typically restricted to observable accuracy and fairness criteria. We develop a flexible framework for fair ensemble learning that allows users to efficiently explore the fairness-accuracy space or to improve the fairness or accuracy of a benchmark model. Our framework can simultaneously target multiple observable or counterfactual fairness criteria, and it enables users to combine a large number of previously trained and newly trained predictors. We provide theoretical guarantees that our estimators converge at fast rates. We apply our method on both simulated and real data, with respect to both observable and counterfactual accuracy and fairness criteria. We show that, surprisingly, multiple unfairness measures can sometimes be minimized simultaneously with little impact on accuracy, relative to unconstrained predictors or existing benchmark models."
                },
                {
                    "title": "Artificial intelligence system reduces false-positive findings in the interpretation of breast ultrasound exams",
                    "abstract": "Though consistently shown to detect mammographically occult cancers, breast ultrasound has been noted to have high false-positive rates. In this work, we present an AI system that achieves radiologist-level accuracy in identifying breast cancer in ultrasound images. Developed on 288,767 exams, consisting of 5,442,907 B-mode and Color Doppler images, the AI achieves an area under the receiver operating characteristic curve (AUROC) of 0.976 on a test set consisting of 44,755 exams. In a retrospective reader study, the AI achieves a higher AUROC than the average of ten board-certified breast radiologists (AUROC: 0.962 AI, 0.924\u2009\u00b1\u20090.02 radiologists). With the help of the AI, radiologists decrease their false positive rates by 37.3% and reduce requested biopsies by 27.8%, while maintaining the same level of sensitivity. This highlights the potential of AI in improving the accuracy, consistency, and efficiency of breast ultrasound diagnosis. Ultrasound is an important imaging modality for the detection and characterization of breast cancer, but it has been noted to have high false-positive rates. Here, the authors present an artificial intelligence system that achieves radiologist-level accuracy in identifying breast cancer in ultrasound imaging."
                },
                {
                    "title": "Closer than they appear: A Bayesian perspective on individual\u2010level heterogeneity in risk assessment",
                    "abstract": "Risk assessment instruments are used across the criminal justice system to estimate the probability of some future event, such as failure to appear for a court appointment or re\u2010arrest. The estimated probabilities are then used in making decisions at the individual level. In the past, there has been controversy about whether the probabilities derived from group\u2010level calculations can meaningfully be applied to individuals. Using Bayesian hierarchical models applied to a large longitudinal dataset from the court system in the state of Kentucky, we analyse variation in individual\u2010level probabilities of failing to appear for court and the extent to which it is captured by covariates. We find that individuals within the same risk group vary widely in their probability of the outcome. In practice, this means that allocating individuals to risk groups based on standard approaches to risk assessment, in large part, results in creating distinctions among individuals who are not meaningfully different in terms of their likelihood of the outcome. This is because uncertainty about the probability that any particular individual will fail to appear is large relative to the difference in average probabilities among any reasonable set of risk groups."
                },
                {
                    "title": "What Marginal Outcome Tests Can Tell Us About Racially Biased Decision-Making",
                    "abstract": "Marginal outcome tests compare the expected effects of a decision on individuals who are of different races but at the same indifference point of the decision-maker. I present a simple formalization of how such tests can detect racial bias, defined as a deviation from accurate statistical discrimination. Namely, the tests can reject that the decision-maker ranks individuals according to some accurate prediction of a mandated outcome, given some unspecified race-inclusive information set. The frontier of marginal effects can furthermore rule out canonical taste-based discrimination. I relate this analysis to other interpretations of marginal outcome tests, other notions of racial discrimination, and recent identification strategies."
                },
                {
                    "title": "Toward robust mammography-based models for breast cancer risk",
                    "abstract": "An algorithm to predict breast cancer risk, Mirai, outperforms clinical risk models across test cohorts from the United States, Sweden, and Taiwan. Don\u2019t risk it Mammograms are a common but imperfect way of assessing breast cancer risk. Current U.S. breast cancer screening guidelines all use a component of cancer risk assessment to inform clinical course. Yala et al. developed a machine learning model called \u201cMirai\u201d to predict breast cancer risk based on traditional mammograms. The authors\u2019 risk model performed better than Tyrer-Cuzick and previous deep learning models at identifying both 5-year breast cancer risk and high-risk patients across multiple international cohorts. Mirai also performed similarly across race and ethnicity categories, suggesting the potential for improvement in patient care across the board. Improved breast cancer risk models enable targeted screening strategies that achieve earlier detection and less screening harm than existing guidelines. To bring deep learning risk models to clinical practice, we need to further refine their accuracy, validate them across diverse populations, and demonstrate their potential to improve clinical workflows. We developed Mirai, a mammography-based deep learning model designed to predict risk at multiple timepoints, leverage potentially missing risk factor information, and produce predictions that are consistent across mammography machines. Mirai was trained on a large dataset from Massachusetts General Hospital (MGH) in the United States and tested on held-out test sets from MGH, Karolinska University Hospital in Sweden, and Chang Gung Memorial Hospital (CGMH) in Taiwan, obtaining C-indices of 0.76 (95% confidence interval, 0.74 to 0.80), 0.81 (0.79 to 0.82), and 0.79 (0.79 to 0.83), respectively. Mirai obtained significantly higher 5-year ROC AUCs than the Tyrer-Cuzick model (P < 0.001) and prior deep learning models Hybrid DL (P < 0.001) and Image-Only DL (P < 0.001), trained on the same dataset. Mirai more accurately identified high-risk patients than prior methods across all datasets. On the MGH test set, 41.5% (34.4 to 48.5) of patients who would develop cancer within 5 years were identified as high risk, compared with 36.1% (29.1 to 42.9) by Hybrid DL (P = 0.02) and 22.9% (15.9 to 29.6) by the Tyrer-Cuzick model (P < 0.001)."
                },
                {
                    "title": "Leveraging Expert Consistency to Improve Algorithmic Decision Support",
                    "abstract": "Machine learning (ML) is increasingly being used to support high-stakes decisions. However, there is frequently a construct gap: a gap between the construct of interest to the decision-making task and what is captured in proxies used as labels to train ML models. As a result, ML models may fail to capture important dimensions of decision criteria, hampering their utility for decision support. Thus, an essential step in the design of ML systems for decision support is selecting a target label among available proxies. In this work, we explore the use of historical expert decisions as a rich -- yet also imperfect -- source of information that can be combined with observed outcomes to narrow the construct gap. We argue that managers and system designers may be interested in learning from experts in instances where they exhibit consistency with each other, while learning from observed outcomes otherwise. We develop a methodology to enable this goal using information that is commonly available in organizational information systems. This involves two core steps. First, we propose an influence function-based methodology to estimate expert consistency indirectly when each case in the data is assessed by a single expert. Second, we introduce a label amalgamation approach that allows ML models to simultaneously learn from expert decisions and observed outcomes. Our empirical evaluation, using simulations in a clinical setting and real-world data from the child welfare domain, indicates that the proposed approach successfully narrows the construct gap, yielding better predictive performance than learning from either observed outcomes or expert decisions alone."
                },
                {
                    "title": "Bounding the Accuracy of Diagnostic Tests, With Application to COVID-19 Antibody Tests.",
                    "abstract": "Tests used to diagnose illness commonly have imperfect accuracy, with some false-positive and negative results. For risk assessment and clinical decisions, predictive values are of interest. Positive predictive value (PPV) is the chance that a member of a relevant population who tests positive has been ill. Negative predictive value (NPV) is the chance that someone who tests negative has not been ill. The medical literature regularly reports sensitivity and specificity. Sensitivity is the chance that an ill person receives a positive test result. Specificity is the chance that a nonill person receives a negative result. Knowledge of sensitivity and specificity enables one to predict the test result given a person's illness status. These predictions are not directly relevant to patient care but, given knowledge of sensitivity and specificity, PPV and NPV can be derived if one knows the prevalence of the disease, the population rate of illness. There is considerable uncertainty about the prevalence of some diseases, a notable case being COVID-19. This paper addresses the problem of identification of PPV and NPV given knowledge of sensitivity and specificity and given bounds on prevalence. I explain the problem and show how to bound PPV and NPV as well as the risk ratio and difference, which are functions thereof. I apply the findings to COVID-19 antibody tests. I question the realism of supposing that sensitivity and specificity are known."
                },
                {
                    "title": "WILDS: A Benchmark of in-the-Wild Distribution Shifts",
                    "abstract": "Distribution shifts -- where the training distribution differs from the test distribution -- can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity in the real-world deployments, these distribution shifts are under-represented in the datasets widely used in the ML community today. To address this gap, we present WILDS, a curated benchmark of 10 datasets reflecting a diverse range of distribution shifts that naturally arise in real-world applications, such as shifts across hospitals for tumor identification; across camera traps for wildlife monitoring; and across time and location in satellite imaging and poverty mapping. On each dataset, we show that standard training yields substantially lower out-of-distribution than in-distribution performance. This gap remains even with models trained by existing methods for tackling distribution shifts, underscoring the need for new methods for training models that are more robust to the types of distribution shifts that arise in practice. To facilitate method development, we provide an open-source package that automates dataset loading, contains default model architectures and hyperparameters, and standardizes evaluations. Code and leaderboards are available at https://wilds.stanford.edu."
                },
                {
                    "title": "Explanatory Analysis of a Machine Learning Model to Identify Hypertrophic Cardiomyopathy Patients from EHR Using Diagnostic Codes",
                    "abstract": "Hypertrophic cardiomyopathy (HCM) is a genetic heart disease that is the leading cause of sudden cardiac death (SCD) in young adults. Despite the well-known risk factors and existing clinical practice guidelines, HCM patients are underdiagnosed and sub-optimally managed. Developing machine learning models on electronic health record (EHR) data can help in better diagnosis of HCM and thus improve hundreds of patient lives. Automated phenotyping using HCM billing codes has received limited attention in the literature with a small number of prior publications. In this paper, we propose a novel predictive model that helps physicians in making diagnostic decisions, by means of information learned from historical data of similar patients.We assembled a cohort of 11,562 patients with known or suspected HCM who have visited Mayo Clinic between the years 1995 to 2019. All existing billing codes of these patients were extracted from the EHR data warehouse. Target ground truth labeling for training the machine learning model was provided by confirmed HCM diagnosis using the gold standard imaging tests for HCM diagnosis echocardiography (echo), or cardiac magnetic resonance (CMR) imaging. As the result, patients were labeled into three categories of \u201cyes\u201d definite HCM \u201cno\u201d HCM phenotype and \u201cpossible HCM\u201d after a manual review of medical records and imaging tests.In this study, a random forest was adopted to investigate the predictive performance of billing codes for the identification of HCM patients due to its practical application and expected accuracy in a wide range of use cases. Our model performed well in finding patients with \u201cyes definite \u201cpossible\u201d and \u201cno\u201d HCM with an accuracy of 71%, weighted recall of 70%, the precision of 75%, and weighted F1 score of 72%. Furthermore, we provided visualizations based on multidimensional scaling and the principal component analysis to provide insights for clinicians\u2019 interpretation. This model can be used for the identification of HCM patients using their EHR data, and help clinicians in their diagnosis decision making."
                },
                {
                    "title": "Derivation With Internal Validation of a Multivariable Predictive Model to Predict COVID\u201019 Test Results in Emergency Department Patients",
                    "abstract": "The COVID\u201019 pandemic has placed acute care providers in demanding situations in predicting disease given the clinical variability, desire to cohort patients, and high variance in testing availability. An approach to stratifying patients by likelihood of disease based on rapidly available emergency department (ED) clinical data would offer significant operational and clinical value. The purpose of this study was to develop and internally validate a predictive model to aid in the discrimination of patients undergoing investigation for COVID\u201019."
                },
                {
                    "title": "Assessing racial inequality in COVID-19 testing with Bayesian threshold tests.",
                    "abstract": "There are racial disparities in the COVID-19 test positivity rate, suggesting that minorities may be under-tested. Here, drawing on the literature on statistically assessing racial disparities in policing, we 1) illuminate a statistical flaw, known as infra-marginality, in using the positivity rate as a metric for assessing racial disparities in under-testing; 2) develop a new type of Bayesian threshold test to measure disparities in COVID-19 testing and 3) apply the test to measure racial disparities in testing thresholds in a real-world COVID-19 dataset."
                },
                {
                    "title": "Ethical Machine Learning in Health Care",
                    "abstract": "The use of machine learning (ML) in healthcare raises numerous ethical concerns, especially as models can amplify existing health inequities. Here, we outline ethical considerations for equitable ML in the advancement of healthcare. Specifically, we frame ethics of ML in healthcare through the lens of social justice. We describe ongoing efforts and outline challenges in a proposed pipeline of ethical ML in health, ranging from problem selection to postdeployment considerations. We close by summarizing recommendations to address these challenges."
                },
                {
                    "title": "In Search of Lost Domain Generalization",
                    "abstract": "The goal of domain generalization algorithms is to predict well on distributions different from those seen during training. While a myriad of domain generalization algorithms exist, inconsistencies in experimental conditions -- datasets, architectures, and model selection criteria -- render fair and realistic comparisons difficult. In this paper, we are interested in understanding how useful domain generalization algorithms are in realistic settings. As a first step, we realize that model selection is non-trivial for domain generalization tasks. Contrary to prior work, we argue that domain generalization algorithms without a model selection strategy should be regarded as incomplete. Next, we implement DomainBed, a testbed for domain generalization including seven multi-domain datasets, nine baseline algorithms, and three model selection criteria. We conduct extensive experiments using DomainBed and find that, when carefully implemented, empirical risk minimization shows state-of-the-art performance across all datasets. Looking forward, we hope that the release of DomainBed, along with contributions from fellow researchers, will streamline reproducible and rigorous research in domain generalization."
                },
                {
                    "title": "Learning the Truth From Only One Side of the Story",
                    "abstract": "Learning under one-sided feedback (i.e., where examples arrive in an online fashion and the learner only sees the labels for examples it predicted positively on) is a fundamental problem in machine learning -- applications include lending and recommendation systems. Despite this, there has been surprisingly little progress made in ways to mitigate the effects of the sampling bias that arises. \nWe focus on generalized linear models and show that without adjusting for this sampling bias, the model may converge sub-optimally or even fail to converge to the optimal solution. We propose an adaptive Upper Confidence Bound approach that comes with rigorous regret guarantees and we show that it outperforms several existing methods experimentally. Our method leverages uncertainty estimation techniques for generalized linear models to more efficiently explore uncertain areas than existing approaches which explore randomly."
                },
                {
                    "title": "A Theoretical and Practical Framework for Regression and Classification from Truncated Samples",
                    "abstract": "Machine learning and statistics are invaluable for extracting insights from data. A key assumption of most methods, however, is that they have access to independent samples from the distribution of relevant data. As such, these methods often perform poorly in the face of biased data which breaks this assumption. In this work, we consider the classical challenge of bias due to truncation, wherein samples falling outside of an \u201cobservation win-dow\u201d cannot be observed [Coh16; Bre96]. We present a general framework for regression and classi\ufb01cation from samples that are truncated according to the value of the dependent variable. The framework argues that stochastic gradient descent (SGD) can be e\ufb03ciently executed on the population log-likelihood of the truncated sample. While our framework is broadly applicable, we also provide end-to-end guarantees for the well-studied problems of truncated logistic and probit regression, where we use it to argue that the true model parameters can be identi\ufb01ed computationally and statistically e\ufb03ciently from truncated data, extending recent work on truncated linear regression [Das+19]. We also provide experiments to illustrate the practicality of our framework on synthetic and real data."
                },
                {
                    "title": "Simple rules to guide expert classifications",
                    "abstract": "Judges, doctors and managers are among those decision makers who must often choose a course of action under limited time, with limited knowledge and without the aid of a computer. Because data\u2010driven methods typically outperform unaided judgements, resource\u2010constrained practitioners can benefit from simple, statistically derived rules that can be applied mentally. In this work, we formalize long\u2010standing observations about the efficacy of improper linear models to construct accurate yet easily applied rules. To test the performance of this approach, we conduct a large\u2010scale evaluation in 22 domains and focus in detail on one: judicial decisions to release or detain defendants while they await trial. In these domains, we find that simple rules rival the accuracy of complex prediction models that base decisions on considerably more information. Further, comparing with unaided judicial decisions, we find that simple rules substantially outperform the human experts. To conclude, we present an analytical framework that sheds light on why simple rules perform as well as they do."
                },
                {
                    "title": "Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems",
                    "abstract": "In this tutorial article, we aim to provide the reader with the conceptual tools needed to get started on research on offline reinforcement learning algorithms: reinforcement learning algorithms that utilize previously collected data, without additional online data collection. Offline reinforcement learning algorithms hold tremendous promise for making it possible to turn large datasets into powerful decision making engines. Effective offline reinforcement learning methods would be able to extract policies with the maximum possible utility out of the available data, thereby allowing automation of a wide range of decision-making domains, from healthcare and education to robotics. However, the limitations of current algorithms make this difficult. We will aim to provide the reader with an understanding of these challenges, particularly in the context of modern deep reinforcement learning methods, and describe some potential solutions that have been explored in recent work to mitigate these challenges, along with recent applications, and a discussion of perspectives on open problems in the field."
                },
                {
                    "title": "Bayesian Sensitivity Analysis for Offline Policy Evaluation",
                    "abstract": "On a variety of complex decision-making tasks, from doctors prescribing treatment to judges setting bail, machine learning algorithms have been shown to outperform expert human judgments. One complication, however, is that it is often difficult to anticipate the effects of algorithmic policies prior to deployment, as one generally cannot use historical data to directly observe what would have happened had the actions recommended by the algorithm been taken. A common strategy is to model potential outcomes for alternative decisions assuming that there are no unmeasured confounders (i.e., to assume ignorability). But if this ignorability assumption is violated, the predicted and actual effects of an algorithmic policy can diverge sharply. In this paper we present a flexible Bayesian approach to gauge the sensitivity of predicted policy outcomes to unmeasured confounders. In particular, and in contrast to past work, our modeling framework easily enables confounders to vary with the observed covariates. We demonstrate the efficacy of our method on a large dataset of judicial actions, in which one must decide whether defendants awaiting trial should be required to pay bail or can be released without payment."
                },
                {
                    "title": "Fair Allocation through Selective Information Acquisition",
                    "abstract": "Public and private institutions must often allocate scarce resources under uncertainty. Banks, for example, extend credit to loan applicants based in part on their estimated likelihood of repaying a loan. But when the quality of information differs across candidates (e.g., if some applicants lack traditional credit histories), common lending strategies can lead to disparities across groups. Here we consider a setting in which decision makers---before allocating resources---can choose to spend some of their limited budget further screening select individuals. We present a computationally efficient algorithm for deciding whom to screen that maximizes a standard measure of social welfare. Intuitively, decision makers should screen candidates on the margin, for whom the additional information could plausibly alter the allocation. We formalize this idea by showing the problem can be reduced to solving a series of linear programs. Both on synthetic and real-world datasets, this strategy improves utility, illustrating the value of targeted information acquisition in such decisions. Further, when there is social value for distributing resources to groups for whom we have a priori poor information---like those without credit scores---our approach can substantially improve the allocation of limited assets."
                },
                {
                    "title": "Selection with Variation in Diagnostic Skill: Evidence from Radiologists",
                    "abstract": "Physicians, judges, teachers, and agents in many other settings differ systematically in the decisions they make when faced with similar cases. Standard approaches to interpreting and exploiting such differences assume they arise solely from variation in preferences. We develop an alternative framework that allows variation in preferences and diagnostic skill and show that both dimensions may be partially identified in standard settings under quasi-random assignment. We apply this framework to study pneumonia diagnoses by radiologists. Diagnosis rates vary widely among radiologists, and descriptive evidence suggests that a large component of this variation is due to differences in diagnostic skill. Our estimated model suggests that radiologists view failing to diagnose a patient with pneumonia as more costly than incorrectly diagnosing one without, and that this leads less skilled radiologists to optimally choose lower diagnostic thresholds. Variation in skill can explain 39% of the variation in diagnostic decisions, and policies that improve skill perform better than uniform decision guidelines. Failing to account for skill variation can lead to highly misleading results in research designs that use agent assignments as instruments."
                },
                {
                    "title": "Dissecting racial bias in an algorithm used to manage the health of populations",
                    "abstract": "Racial bias in health algorithms The U.S. health care system uses commercial algorithms to guide health decisions. Obermeyer et al. find evidence of racial bias in one widely used algorithm, such that Black patients assigned the same level of risk by the algorithm are sicker than White patients (see the Perspective by Benjamin). The authors estimated that this racial bias reduces the number of Black patients identified for extra care by more than half. Bias occurs because the algorithm uses health costs as a proxy for health needs. Less money is spent on Black patients who have the same level of need, and the algorithm thus falsely concludes that Black patients are healthier than equally sick White patients. Reformulating the algorithm so that it no longer uses costs as a proxy for needs eliminates the racial bias in predicting who needs extra care. Science, this issue p. 447; see also p. 421 A health algorithm that uses health costs as a proxy for health needs leads to racial bias against Black patients. Health systems rely on commercial prediction algorithms to identify and help patients with complex health needs. We show that a widely used algorithm, typical of this industry-wide approach and affecting millions of patients, exhibits significant racial bias: At a given risk score, Black patients are considerably sicker than White patients, as evidenced by signs of uncontrolled illnesses. Remedying this disparity would increase the percentage of Black patients receiving additional help from 17.7 to 46.5%. The bias arises because the algorithm predicts health care costs rather than illness, but unequal access to care means that we spend less money caring for Black patients than for White patients. Thus, despite health care cost appearing to be an effective proxy for health by some measures of predictive accuracy, large racial biases arise. We suggest that the choice of convenient, seemingly effective proxies for ground truth can be an important source of algorithmic bias in many contexts."
                },
                {
                    "title": "The Identification Zoo: Meanings of Identification in Econometrics",
                    "abstract": "Over two dozen different terms for identification appear in the econometrics literature, including set identification, causal identification, local identification, generic identification, weak identification, identification at infinity, and many more. This survey (i) gives a new framework unifying existing definitions of point identification; (ii) summarizes and compares the zooful of different terms associated with identification that appear in the literature; and (iii) discusses concepts closely related to identification, such as normalizations and the differences in identification between structural models and causal, reduced form models."
                },
                {
                    "title": "Counterfactual risk assessments, evaluation, and fairness",
                    "abstract": "Algorithmic risk assessments are increasingly used to help humans make decisions in high-stakes settings, such as medicine, criminal justice and education. In each of these cases, the purpose of the risk assessment tool is to inform actions, such as medical treatments or release conditions, often with the aim of reducing the likelihood of an adverse event such as hospital readmission or recidivism. Problematically, most tools are trained and evaluated on historical data in which the outcomes observed depend on the historical decision-making policy. These tools thus reflect risk under the historical policy, rather than under the different decision options that the tool is intended to inform. Even when tools are constructed to predict risk under a specific decision, they are often improperly evaluated as predictors of the target outcome. Focusing on the evaluation task, in this paper we define counterfactual analogues of common predictive performance and algorithmic fairness metrics that we argue are better suited for the decision-making context. We introduce a new method for estimating the proposed metrics using doubly robust estimation. We provide theoretical results that show that only under strong conditions can fairness according to the standard metric and the counterfactual metric simultaneously hold. Consequently, fairness-promoting methods that target parity in a standard fairness metric may---and as we show empirically, do---induce greater imbalance in the counterfactual analogue. We provide empirical comparisons on both synthetic data and a real world child welfare dataset to demonstrate how the proposed method improves upon standard practice."
                },
                {
                    "title": "Diagnosing Physician Error: A Machine Learning Approach to Low-Value Health Care",
                    "abstract": "We use machine learning to better characterize low-value health care and the decisions that produce it. We focus on costly tests, specifically for heart attack (acute coronary syndromes). A test is only useful if it yields new information, so efficient testing is grounded in accurate prediction of test outcomes. Physician testing decisions can therefore be benchmarked against tailored algorithmic predictions, which provide a more precise way to study low-value care than the usual approach\u2014looking at average test yield. Implemented in a large national sample, this procedure reveals significant over-testing: 52.6% of high-cost tests for heart attack are wasted. At the same time, it also reveals significant under-testing: many patients with predictably high risk go untested, then experience frequent adverse cardiac events including death in the next 30 days. At standard clinical thresholds, these event rates suggest that testing these patients would indeed have been highly cost-effective. Of the potential welfare gains from more efficient testing, 42.8% would come from addressing under-use. Existing policy levers, however, appear too blunt a tool to address both over- and under-use inefficiencies. We find that they cut testing across the board, for low-risk (reducing over-use) and high-risk patients (exaggerating under-use). Finally, we uncover two behavioral mechanisms for physician testing errors: (i) bounded rationality, in which physicians use an overly narrow set of variables, but make effective use of that set; and (ii) representativeness, in which they over-weight how \u201crepresentative\u201d heart attack is for a patient, above and beyond the conditional probability. Together, these results suggest the need for models of low-value care that incorporate mis-prediction so as to account for both over- and under-testing."
                },
                {
                    "title": "A Deep Learning Mammography-based Model for Improved Breast Cancer Risk Prediction.",
                    "abstract": "Background Mammographic density improves the accuracy of breast cancer risk models. However, the use of breast density is limited by subjective assessment, variation across radiologists, and restricted data. A mammography-based deep learning (DL) model may provide more accurate risk prediction. Purpose To develop a mammography-based DL breast cancer risk model that is more accurate than established clinical breast cancer risk models. Materials and Methods This retrospective study included 88 994 consecutive screening mammograms in 39 571 women between January 1, 2009, and December 31, 2012. For each patient, all examinations were assigned to either training, validation, or test sets, resulting in 71 689, 8554, and 8751 examinations, respectively. Cancer outcomes were obtained through linkage to a regional tumor registry. By using risk factor information from patient questionnaires and electronic medical records review, three models were developed to assess breast cancer risk within 5 years: a risk-factor-based logistic regression model (RF-LR) that used traditional risk factors, a DL model (image-only DL) that used mammograms alone, and a hybrid DL model that used both traditional risk factors and mammograms. Comparisons were made to an established breast cancer risk model that included breast density (Tyrer-Cuzick model, version 8 [TC]). Model performance was compared by using areas under the receiver operating characteristic curve (AUCs) with DeLong test (P < .05). Results The test set included 3937 women, aged 56.20 years \u00b1 10.04. Hybrid DL and image-only DL showed AUCs of 0.70 (95% confidence interval [CI]: 0.66, 0.75) and 0.68 (95% CI: 0.64, 0.73), respectively. RF-LR and TC showed AUCs of 0.67 (95% CI: 0.62, 0.72) and 0.62 (95% CI: 0.57, 0.66), respectively. Hybrid DL showed a significantly higher AUC (0.70) than TC (0.62; P < .001) and RF-LR (0.67; P = .01). Conclusion Deep learning models that use full-field mammograms yield substantially improved risk discrimination compared with the Tyrer-Cuzick (version 8) model. \u00a9 RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Sitek and Wolfe in this issue."
                },
                {
                    "title": "Stay Ahead of Poachers: Illegal Wildlife Poaching Prediction and Patrol Planning Under Uncertainty with Field Test Evaluations (Short Version)",
                    "abstract": "Illegal wildlife poaching threatens ecosystems and drives endangered species toward extinction. However, efforts for wildlife protection are constrained by the limited resources of law enforcement agencies. To help combat poaching, the Protection Assistant for Wildlife Security (PAWS) is a machine learning pipeline that has been developed as a data-driven approach to identify areas at high risk of poaching throughout protected areas and compute optimal patrol routes. In this paper, we take an end-to-end approach to the data-to-deployment pipeline for anti-poaching. In doing so, we address challenges including extreme class imbalance (up to 1:200), bias, and uncertainty in wildlife poaching data to enhance PAWS, and we apply our methodology to three national parks with diverse characteristics. (i) We use Gaussian processes to quantify predictive uncertainty, which we exploit to improve robustness of our prescribed patrols and increase detection of snares by an average of 30%. We evaluate our approach on real-world historical poaching data from Murchison Falls and Queen Elizabeth National Parks in Uganda and, for the first time, Srepok Wildlife Sanctuary in Cambodia. (ii) We present the results of large-scale field tests conducted in Murchison Falls and Srepok Wildlife Sanctuary which confirm that the predictive power of PAWS extends promisingly to multiple parks. This paper is part of an effort to expand PAWS to 800 parks around the world through integration with SMART conservation software."
                },
                {
                    "title": "Towards a decision support tool for intensive care discharge: machine learning algorithm development using electronic healthcare data from MIMIC-III and Bristol, UK",
                    "abstract": "Objective The primary objective is to develop an automated method for detecting patients that are ready for discharge from intensive care. Design We used two datasets of routinely collected patient data to test and improve on a set of previously proposed discharge criteria. Setting Bristol Royal Infirmary general intensive care unit (GICU). Patients Two cohorts derived from historical datasets: 1870 intensive care patients from GICU in Bristol, and 7592 from Medical Information Mart for Intensive Care (MIMIC)-III. Results In both cohorts few successfully discharged patients met all of the discharge criteria. Both a random forest and a logistic classifier, trained using multiple-source cross-validation, demonstrated improved performance over the original criteria and generalised well between the cohorts. The classifiers showed good agreement on which features were most predictive of readiness-for-discharge, and these were generally consistent with clinical experience. By weighting the discharge criteria according to feature importance from the logistic model we showed improved performance over the original criteria, while retaining good interpretability. Conclusions Our findings indicate the feasibility of the proposed approach to ready-for-discharge classification, which could complement other risk models of specific adverse outcomes in a future decision support system. Avenues for improvement to produce a clinically useful tool are identified."
                },
                {
                    "title": "Predicting hospital admission at emergency department triage using machine learning",
                    "abstract": "Objective To predict hospital admission at the time of ED triage using patient history in addition to information collected at triage. Methods This retrospective study included all adult ED visits between March 2014 and July 2017 from one academic and two community emergency rooms that resulted in either admission or discharge. A total of 972 variables were extracted per patient visit. Samples were randomly partitioned into training (80%), validation (10%), and test (10%) sets. We trained a series of nine binary classifiers using logistic regression (LR), gradient boosting (XGBoost), and deep neural networks (DNN) on three dataset types: one using only triage information, one using only patient history, and one using the full set of variables. Next, we tested the potential benefit of additional training samples by training models on increasing fractions of our data. Lastly, variables of importance were identified using information gain as a metric to create a low-dimensional model. Results A total of 560,486 patient visits were included in the study, with an overall admission risk of 29.7%. Models trained on triage information yielded a test AUC of 0.87 for LR (95% CI 0.86\u20130.87), 0.87 for XGBoost (95% CI 0.87\u20130.88) and 0.87 for DNN (95% CI 0.87\u20130.88). Models trained on patient history yielded an AUC of 0.86 for LR (95% CI 0.86\u20130.87), 0.87 for XGBoost (95% CI 0.87\u20130.87) and 0.87 for DNN (95% CI 0.87\u20130.88). Models trained on the full set of variables yielded an AUC of 0.91 for LR (95% CI 0.91\u20130.91), 0.92 for XGBoost (95% CI 0.92\u20130.93) and 0.92 for DNN (95% CI 0.92\u20130.92). All algorithms reached maximum performance at 50% of the training set or less. A low-dimensional XGBoost model built on ESI level, outpatient medication counts, demographics, and hospital usage statistics yielded an AUC of 0.91 (95% CI 0.91\u20130.91). Conclusion Machine learning can robustly predict hospital admission using triage information and patient history. The addition of historical information improves predictive performance significantly compared to using triage information alone, highlighting the need to incorporate these variables into prediction models."
                },
                {
                    "title": "Limits of Estimating Heterogeneous Treatment Effects: Guidelines for Practical Algorithm Design",
                    "abstract": "Estimating heterogeneous treatment effects from observational data is a central problem in many domains. Because counterfactual data is inac-cessible, the problem differs fundamentally from supervised learning, and entails a more com-plex set of modeling choices. Despite a variety of recently proposed algorithmic solutions, a principled guideline for building estimators of treatment effects using machine learning algorithms is still lacking. In this paper, we provide such guidelines by characterizing the fundamental limits of estimating heterogeneous treatment effects, and establishing conditions under which these limits can be achieved. Our analysis reveals that the relative importance of the different aspects of observational data vary with the sample size. For instance, we show that selection bias matters only in small-sample regimes, whereas with a large sample size, the way an algorithm models the control and treated outcomes is what bottlenecks its performance. Guided by our analysis, we build a practical algorithm for estimating treatment effects using a non-stationary Gaussian processes with doubly-robust hyperparameters. Using a standard semi-synthetic simulation setup, we show that our algorithm outperforms the state-of-the-art, and that the behavior of ex-isting algorithms conforms with our analysis."
                },
                {
                    "title": "Predictive modeling of U.S. health care spending in late life",
                    "abstract": "End-of-life health care spending In the United States, one-quarter of Medicare spending occurs in the last 12 months of life, which is commonly seen as evidence of waste. Einav et al. used predictive modeling to reassess this interpretation. From detailed Medicare claims data, the extent to which spending is concentrated not just on those who die, but on those who are expected to die, can be estimated. Most deaths are unpredictable; hence, focusing on end-of-life spending does not necessarily identify \u201cwasteful\u201d spending. Science, this issue p. 1462 The United States spends a lot on people in the last year of life, but those at high risk of death account for relatively little spending. That one-quarter of Medicare spending in the United States occurs in the last year of life is commonly interpreted as waste. But this interpretation presumes knowledge of who will die and when. Here we analyze how spending is distributed by predicted mortality, based on a machine-learning model of annual mortality risk built using Medicare claims. Death is highly unpredictable. Less than 5% of spending is accounted for by individuals with predicted mortality above 50%. The simple fact that we spend more on the sick\u2014both on those who recover and those who die\u2014accounts for 30 to 50% of the concentration of spending on the dead. Our results suggest that spending on the ex post dead does not necessarily mean that we spend on the ex ante \u201chopeless.\u201d"
                },
                {
                    "title": "Principled Approaches to Missing Data in Epidemiologic Studies.",
                    "abstract": "Principled methods with which to appropriately analyze missing data have long existed; however, broad implementation of these methods remains challenging. In this and 2 companion papers (Am J Epidemiol. 2018;187(3):576-584 and Am J Epidemiol. 2018;187(3):585-591), we discuss issues pertaining to missing data in the epidemiologic literature. We provide details regarding missing-data mechanisms and nomenclature and encourage the conduct of principled analyses through a detailed comparison of multiple imputation and inverse probability weighting. Data from the Collaborative Perinatal Project, a multisite US study conducted from 1959 to 1974, are used to create a masked data-analytical challenge with missing data induced by known mechanisms. We illustrate the deleterious effects of missing data with naive methods and show how principled methods can sometimes mitigate such effects. For example, when data were missing at random, naive methods showed a spurious protective effect of smoking on the risk of spontaneous abortion (odds ratio (OR) = 0.43, 95% confidence interval (CI): 0.19, 0.93), while implementation of principled methods multiple imputation (OR = 1.30, 95% CI: 0.95, 1.77) or augmented inverse probability weighting (OR = 1.40, 95% CI: 1.00, 1.97) provided estimates closer to the \"true\" full-data effect (OR = 1.31, 95% CI: 1.05, 1.64). We call for greater acknowledgement of and attention to missing data and for the broad use of principled missing-data methods in epidemiologic research."
                },
                {
                    "title": "The Selective Labels Problem: Evaluating Algorithmic Predictions in the Presence of Unobservables",
                    "abstract": "Evaluating whether machines improve on human performance is one of the central questions of machine learning. However, there are many domains where the data is selectively labeled, in the sense that the observed outcomes are themselves a consequence of the existing choices of the human decision-makers. For instance, in the context of judicial bail decisions, we observe the outcome of whether a defendant fails to return for their court appearance only if the human judge decides to release the defendant on bail. This selective labeling makes it harder to evaluate predictive models as the instances for which outcomes are observed do not represent a random sample of the population. Here we propose a novel framework for evaluating the performance of predictive models on selectively labeled data. We develop an approach called contraction which allows us to compare the performance of predictive models and human decision-makers without resorting to counterfactual inference. Our methodology harnesses the heterogeneity of human decision-makers and facilitates effective evaluation of predictive models even in the presence of unmeasured confounders (unobservables) which influence both human decisions and the resulting outcomes. Experimental results on real world datasets spanning diverse domains such as health care, insurance, and criminal justice demonstrate the utility of our evaluation metric in comparing human decisions and machine predictions."
                },
                {
                    "title": "Metalearners for estimating heterogeneous treatment effects using machine learning",
                    "abstract": "Significance Estimating and analyzing heterogeneous treatment effects is timely, yet challenging. We introduce a unifying framework for many conditional average treatment effect estimators, and we propose a metalearner, the X-learner, which can adapt to structural properties, such as the smoothness and sparsity of the underlying treatment effect. We present its favorable properties, using theory and simulations. We apply it, using random forests, to two field experiments in political science, where it is shown to be easy to use and to produce results that are interpretable. There is growing interest in estimating and analyzing heterogeneous treatment effects in experimental and observational studies. We describe a number of metaalgorithms that can take advantage of any supervised learning or regression method in machine learning and statistics to estimate the conditional average treatment effect (CATE) function. Metaalgorithms build on base algorithms\u2014such as random forests (RFs), Bayesian additive regression trees (BARTs), or neural networks\u2014to estimate the CATE, a function that the base algorithms are not designed to estimate directly. We introduce a metaalgorithm, the X-learner, that is provably efficient when the number of units in one treatment group is much larger than in the other and can exploit structural properties of the CATE function. For example, if the CATE function is linear and the response functions in treatment and control are Lipschitz-continuous, the X-learner can still achieve the parametric rate under regularity conditions. We then introduce versions of the X-learner that use RF and BART as base learners. In extensive simulation studies, the X-learner performs favorably, although none of the metalearners is uniformly the best. In two persuasion field experiments from political science, we demonstrate how our X-learner can be used to target treatment regimes and to shed light on underlying mechanisms. A software package is provided that implements our methods."
                },
                {
                    "title": "Understanding vasopressor intervention and weaning: risk prediction in a public heterogeneous clinical time series database",
                    "abstract": "Background\nThe widespread adoption of electronic health records allows us to ask evidence-based questions about the need for and benefits of specific clinical interventions in critical-care settings across large populations.\n\n\nObjective\nWe investigated the prediction of vasopressor administration and weaning in the intensive care unit. Vasopressors are commonly used to control hypotension, and changes in timing and dosage can have a large impact on patient outcomes.\n\n\nMaterials and Methods\nWe considered a cohort of 15\u2009695 intensive care unit patients without orders for reduced care who were alive 30 days post-discharge. A switching-state autoregressive model (SSAM) was trained to predict the multidimensional physiological time series of patients before, during, and after vasopressor administration. The latent states from the SSAM were used as predictors of vasopressor administration and weaning.\n\n\nResults\nThe unsupervised SSAM features were able to predict patient vasopressor administration and successful patient weaning. Features derived from the SSAM achieved areas under the receiver operating curve of 0.92, 0.88, and 0.71 for predicting ungapped vasopressor administration, gapped vasopressor administration, and vasopressor weaning, respectively. We also demonstrated many cases where our model predicted weaning well in advance of a successful wean.\n\n\nConclusion\nModels that used SSAM features increased performance on both predictive tasks. These improvements may reflect an underlying, and ultimately predictive, latent state detectable from the physiological time series."
                },
                {
                    "title": "Reliable Decision Support using Counterfactual Models",
                    "abstract": "Answering \"What if?\" questions is important in many domains. For example, would a patient's disease progression slow down if I were to give them a dose of drug A? Ideally, we answer our question using an experiment, but this is not always possible (e.g., it may be unethical). As an alternative, we can use non-experimental data to learn models that make counterfactual predictions of what we would observe had we run an experiment. In this paper, we propose the counterfactual GP, a counterfactual model of continuous-time trajectories (time series) under sequences of actions taken in continuous-time. We develop our model within the potential outcomes framework of Neyman and Rubin. The counterfactual GP is trained using a joint maximum likelihood objective that adjusts for dependencies between observed actions and outcomes in the training data. We report two sets of experimental results using the counterfactual GP. The first shows that it can be used to learn the natural progression (i.e. untreated progression) of biomarker trajectories from observational data. In the second, we show how the CGP can be used for medical decision support by learning counterfactual models of renal health under different types of dialysis."
                },
                {
                    "title": "Fast Threshold Tests for Detecting Discrimination",
                    "abstract": "Threshold tests have recently been proposed as a useful method for detecting bias in lending, hiring, and policing decisions. For example, in the case of credit extensions, these tests aim to estimate the bar for granting loans to white and minority applicants, with a higher inferred threshold for minorities indicative of discrimination. This technique, however, requires fitting a complex Bayesian latent variable model for which inference is often computationally challenging. Here we develop a method for fitting threshold tests that is two orders of magnitude faster than the existing approach, reducing computation from hours to minutes. To achieve these performance gains, we introduce and analyze a flexible family of probability distributions on the interval [0, 1] -- which we call discriminant distributions -- that is computationally efficient to work with. We demonstrate our technique by analyzing 2.7 million police stops of pedestrians in New York City."
                },
                {
                    "title": "Human Decisions and Machine Predictions",
                    "abstract": "Can machine learning improve human decision making? Bail decisions provide a good test case. Millions of times each year, judges make jail-or-release decisions that hinge on a prediction of what a defendant would do if released. The concreteness of the prediction task combined with the volume of data available makes this a promising machine-learning application. Yet comparing the algorithm to judges proves complicated. First, the available data are generated by prior judge decisions. We only observe crime outcomes for released defendants, not for those judges detained. This makes it hard to evaluate counterfactual decision rules based on algorithmic predictions. Second, judges may have a broader set of preferences than the variable the algorithm predicts; for instance, judges may care specifically about violent crimes or about racial inequities. We deal with these problems using different econometric strategies, such as quasi-random assignment of cases to judges. Even accounting for these concerns, our results suggest potentially large welfare gains: one policy simulation shows crime reductions up to 24.7% with no change in jailing rates, or jailing rate reductions up to 41.9% with no increase in crime rates. Moreover, all categories of crime, including violent crimes, show reductions; and these gains can be achieved while simultaneously reducing racial disparities. These results suggest that while machine learning can be valuable, realizing this value requires integrating these tools into an economic framework: being clear about the link between predictions and decisions; specifying the scope of payoff functions; and constructing unbiased decision counterfactuals. JEL Codes: C10 (Econometric and statistical methods and methodology), C55 (Large datasets: Modeling and analysis), K40 (Legal procedure, the legal system, and illegal behavior)."
                },
                {
                    "title": "Stan: A Probabilistic Programming Language.",
                    "abstract": "Stan is a probabilistic programming language for specifying statistical models. A Stan program imperatively defines a log probability function over parameters conditioned on specified data and constants. As of version 2.14.0, Stan provides full Bayesian inference for continuous-variable models through Markov chain Monte Carlo methods such as the No-U-Turn sampler, an adaptive form of Hamiltonian Monte Carlo sampling. Penalized maximum likelihood estimates are calculated using optimization methods such as the limited memory Broyden-Fletcher-Goldfarb-Shanno algorithm. Stan is also a platform for computing log densities and their gradients and Hessians, which can be used in alternative algorithms such as variational Bayes, expectation propagation, and marginal inference using approximate integration. To this end, Stan is set up so that the densities, gradients, and Hessians, along with intermediate quantities of the algorithm such as acceptance probabilities, are easily accessible. Stan can be called from the command line using the cmdstan package, through R using the rstan package, and through Python using the pystan package. All three interfaces support sampling and optimization-based inference with diagnostics and posterior analysis. rstan and pystan also provide access to log probabilities, gradients, Hessians, parameter transforms, and specialized plotting."
                },
                {
                    "title": "A Conceptual Introduction to Hamiltonian Monte Carlo",
                    "abstract": "Hamiltonian Monte Carlo has proven a remarkable empirical success, but only recently have we begun to develop a rigorous understanding of why it performs so well on difficult problems and how it is best applied in practice. Unfortunately, that understanding is confined within the mathematics of differential geometry which has limited its dissemination, especially to the applied communities for which it is particularly important. In this review I provide a comprehensive conceptual account of these theoretical foundations, focusing on developing a principled intuition behind the method and its optimal implementations rather of any exhaustive rigor. Whether a practitioner or a statistician, the dedicated reader will acquire a solid grasp of how Hamiltonian Monte Carlo works, when it succeeds, and, perhaps most importantly, when it fails."
                },
                {
                    "title": "The Problem of Infra-Marginality in Outcome Tests for Discrimination",
                    "abstract": "In the course of conducting traffic stops, officers have discretion to search motorists for drugs, weapons, and other contraband. There is concern that these search decisions are prone to racial bias, but it has proven difficult to rigorously assess claims of discrimination. Here we develop a new statistical method --- the threshold test --- to test for racial discrimination in motor vehicle searches. We use geographic variation in stop outcomes to infer the effective race-specific standards of evidence that officers apply when deciding whom to search, an approach we formalize with a hierarchical Bayesian latent variable model. This technique mitigates the problems of omitted variables and infra-marginality associated with benchmark and outcome tests for discrimination. On a dataset of 4.5 million police stops in North Carolina, we find that the standard for searching black and Hispanic drivers is considerably lower than the standard for searching white and Asian drivers, a pattern that holds consistently across the 100 largest police departments in the state."
                },
                {
                    "title": "Estimating individual treatment effect: generalization bounds and algorithms",
                    "abstract": "There is intense interest in applying machine learning to problems of causal inference in fields such as healthcare, economics and education. In particular, individual-level causal inference has important applications such as precision medicine. We give a new theoretical analysis and family of algorithms for predicting individual treatment effect (ITE) from observational data, under the assumption known as strong ignorability. The algorithms learn a \"balanced\" representation such that the induced treated and control distributions look similar. We give a novel, simple and intuitive generalization-error bound showing that the expected ITE estimation error of a representation is bounded by a sum of the standard generalization-error of that representation and the distance between the treated and control distributions induced by the representation. We use Integral Probability Metrics to measure distances between distributions, deriving explicit bounds for the Wasserstein and Maximum Mean Discrepancy (MMD) distances. Experiments on real and simulated data show the new algorithms match or outperform the state-of-the-art."
                },
                {
                    "title": "Estimation and Inference of Heterogeneous Treatment Effects using Random Forests",
                    "abstract": "ABSTRACT Many scientific and engineering challenges\u2014ranging from personalized medicine to customized marketing recommendations\u2014require an understanding of treatment effect heterogeneity. In this article, we develop a nonparametric causal forest for estimating heterogeneous treatment effects that extends Breiman\u2019s widely used random forest algorithm. In the potential outcomes framework with unconfoundedness, we show that causal forests are pointwise consistent for the true treatment effect and have an asymptotically Gaussian and centered sampling distribution. We also discuss a practical method for constructing asymptotic confidence intervals for the true treatment effect that are centered at the causal forest estimates. Our theoretical results rely on a generic Gaussian theory for a large family of random forest algorithms. To our knowledge, this is the first set of results that allows any type of random forest, including classification and regression forests, to be used for provably valid statistical inference. In experiments, we find causal forests to be substantially more powerful than classical methods based on nearest-neighbor matching, especially in the presence of irrelevant covariates."
                },
                {
                    "title": "Prevalence of incidental prostate cancer: A systematic review of autopsy studies",
                    "abstract": "Prostate cancer screening may detect nonprogressive cancers, leading to overdiagnosis and overtreatment. The potential for overdiagnosis can be assessed from the reservoir of prostate cancer in autopsy studies that report incidental prostate cancer rates in men who died of other causes. We aimed to estimate the age\u2010specific incidental cancer prevalence from all published autopsy studies. We identified eligible studies by searches of Medline and Embase, forward and backward citation searches and contacting authors. We screened the titles and abstracts of all articles; checked the full\u2010text articles for eligibility and extracted clinical and pathology data using standardized forms. We extracted mean cancer prevalence, age\u2010specific cancer prevalence and validity measures and then pooled data from all studies using logistic regression models with random effects. The 29 studies included in the review dated from 1948 to 2013. Incidental cancer was detected in all populations, with no obvious time trends in prevalence. Prostate cancer prevalence increased with each decade of age, OR\u2009=\u20091.7 (1.6\u20131.8), and was higher in studies that used the Gleason score, OR\u2009=\u20092.0 (1.1\u20133.7). No other factors were significantly predictive. The estimated mean cancer prevalence increased in a nonlinear fashion from 5% (95% CI: 3\u20138%) at age <30 years to 59% (95% CI: 48\u201371%) by age >79 years. There was substantial variation between populations in estimated cancer prevalence. There is a substantial reservoir of incidental prostate cancer which increases with age. The high risk of overdiagnosis limits the usefulness of prostate cancer screening."
                },
                {
                    "title": "UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age",
                    "abstract": "Cathie Sudlow and colleagues describe the UK Biobank, a large population-based prospective study, established to allow investigation of the genetic and non-genetic determinants of the diseases of middle and old age."
                },
                {
                    "title": "Breast Cancer Risk Assessment Tool",
                    "abstract": "Breast cancer is well known as one of the cancer that threatens woman. This disease threatens women, regardless race, ethnicity and demographics. Thus, in order to predict the risk of a woman being diagnosed with breast cancer, an individual risk assessment is developed. This is because early detection can help reduce the development of breast cancer. To date, few mathematical models have been proposed and developed for estimating the risk. The statistical model used is known as the Gail model where it provides the estimation of risk development of breast cancer over specific interval and period of time. The estimation mainly uses the women\u2019s demographic information as well as the risk factors as parameters. This model is suitable for woman, which is not suspected of having hereditary breast cancer. Original Gail model considering six attributes including age, age at first menstrual period, age at the time of the birth of first child, number of first-degree relatives that have breast cancer, breast biopsies and race/ethnicity. Using statistical modelling, the individual risk is calculated in terms of probabilities based on both the relative risk and the baseline hazard rate. In this project, the original risk factors are modified depending on the Malaysian demographic information to cater for Malaysia women population."
                },
                {
                    "title": "GRADE: Machine Learning Support for Graduate Admissions",
                    "abstract": "This paper describes GRADE, a statistical machine learning system developed to support the work of the graduate admissions committee at the University of Texas at Austin Department of Computer Science (UTCS). In recent years, the number of applications to the UTCS PhD program has become too large to manage with a traditional review process. GRADE uses historical admissions data to predict how likely the committee is to admit each new applicant. It reports each prediction as a score similar to those used by human reviewers, and accompanies each by an explanation of what applicant features most influenced its prediction. GRADE makes the review process more efficient by enabling reviewers to spend most of their time on applicants near the decision boundary and by focusing their attention on parts of each applicant\u2019s file that matter the most. An evaluation over two seasons of PhD admissions indicates that the system leads to dramatic time savings, reducing the total time spent on reviews by at least 74%."
                },
                {
                    "title": "Aging, cellular senescence, and cancer.",
                    "abstract": "For most species, aging promotes a host of degenerative pathologies that are characterized by debilitating losses of tissue or cellular function. However, especially among vertebrates, aging also promotes hyperplastic pathologies, the most deadly of which is cancer. In contrast to the loss of function that characterizes degenerating cells and tissues, malignant (cancerous) cells must acquire new (albeit aberrant) functions that allow them to develop into a lethal tumor. This review discusses the idea that, despite seemingly opposite characteristics, the degenerative and hyperplastic pathologies of aging are at least partly linked by a common biological phenomenon: a cellular stress response known as cellular senescence. The senescence response is widely recognized as a potent tumor suppressive mechanism. However, recent evidence strengthens the idea that it also drives both degenerative and hyperplastic pathologies, most likely by promoting chronic inflammation. Thus, the senescence response may be the result of antagonistically pleiotropic gene action."
                },
                {
                    "title": "Generalized Linear Mixed Models: Modern Concepts, Methods and Applications",
                    "abstract": "PART I The Big Picture Modeling Basics What Is a Model? Two Model Forms: Model Equation and Probability Distribution Types of Model Effects Writing Models in Matrix Form Summary: Essential Elements for a Complete Statement of the Model Design Matters Introductory Ideas for Translating Design and Objectives into Models Describing \"Data Architecture\" to Facilitate Model Specification From Plot Plan to Linear Predictor Distribution Matters More Complex Example: Multiple Factors with Different Units of Replication Setting the Stage Goals for Inference with Models: Overview Basic Tools of Inference Issue I: Data Scale vs. Model Scale Issue II: Inference Space Issue III: Conditional and Marginal Models Summary PART II Estimation and Inference Essentials Estimation Introduction Essential Background Fixed Effects Only Gaussian Mixed Models Generalized Linear Mixed Models Summary Inference, Part I: Model Effects Introduction Essential Background Approaches to Testing Inference Using Model-Based Statistics Inference Using Empirical Standard Error Summary of Main Ideas and General Guidelines for Implementation Inference, Part II: Covariance Components Introduction Formal Testing of Covariance Components Fit Statistics to Compare Covariance Models Interval Estimation Summary PART III Working with GLMMs Treatment and Explanatory Variable Structure Types of Treatment Structures Types of Estimable Functions Multiple Factor Models: Overview Multifactor Models with All Factors Qualitative Multifactor: Some Factors Qualitative, Some Factors Quantitative Multifactor: All Factors Quantitative Summary Multilevel Models Types of Design Structure: Single- and Multilevel Models Defined Types of Multilevel Models and How They Arise Role of Blocking in Multilevel Models Working with Multilevel Designs Marginal and Conditional Multilevel Models Summary Best Linear Unbiased Prediction Review of Estimable and Predictable Functions BLUP in Random-Effects-Only Models Gaussian Data with Fixed and Random Effects Advanced Applications with Complex Z Matrices Summary Rates and Proportions Types of Rate and Proportion Data Discrete Proportions: Binary and Binomial Data Alternative Link Functions for Binomial Data Continuous Proportions Summary Counts Introduction Overdispersion in Count Data More on Alternative Distributions Conditional and Marginal Too Many Zeroes Summary Time-to-Event Data Introduction: Probability Concepts for Time-to-Event Data Gamma GLMMs GLMMs and Survival Analysis Summary Multinomial Data Overview Multinomial Data with Ordered Categories Nominal Categories: Generalized Logit Models Model Comparison Summary Correlated Errors, Part I: Repeated Measures Overview Gaussian Data: Correlation and Covariance Models for LMMs Covariance Model Selection Non-Gaussian Case Issues for Non-Gaussian Repeated Measures Summary Correlated Errors, Part II: Spatial Variability Overview Gaussian Case with Covariance Model Spatial Covariance Modeling by Smoothing Spline Non-Gaussian Case Summary Power, Sample Size, and Planning Basics of GLMM-Based Power and Precision Analysis Gaussian Example Power for Binomial GLMMs GLMM-Based Power Analysis for Count Data Power and Planning for Repeated Measures Summary Appendices References Index"
                },
                {
                    "title": "Dealing with missing outcome data in randomized trials and observational studies.",
                    "abstract": "Although missing outcome data are an important problem in randomized trials and observational studies, methods to address this issue can be difficult to apply. Using simulated data, the authors compared 3 methods to handle missing outcome data: 1) complete case analysis; 2) single imputation; and 3) multiple imputation (all 3 with and without covariate adjustment). Simulated scenarios focused on continuous or dichotomous missing outcome data from randomized trials or observational studies. When outcomes were missing at random, single and multiple imputations yielded unbiased estimates after covariate adjustment. Estimates obtained by complete case analysis with covariate adjustment were unbiased as well, with coverage close to 95%. When outcome data were missing not at random, all methods gave biased estimates, but handling missing outcome data by means of 1 of the 3 methods reduced bias compared with a complete case analysis without covariate adjustment. Complete case analysis with covariate adjustment and multiple imputation yield similar estimates in the event of missing outcome data, as long as the same predictors of missingness are included. Hence, complete case analysis with covariate adjustment can and should be used as the analysis of choice more often. Multiple imputation, in addition, can accommodate the missing-not-at-random scenario more flexibly, making it especially suited for sensitivity analyses."
                },
                {
                    "title": "Risk prediction models for hospital readmission: a systematic review.",
                    "abstract": "CONTEXT\nPredicting hospital readmission risk is of great interest to identify which patients would benefit most from care transition interventions, as well as to risk-adjust readmission rates for the purposes of hospital comparison.\n\n\nOBJECTIVE\nTo summarize validated readmission risk prediction models, describe their performance, and assess suitability for clinical or administrative use.\n\n\nDATA SOURCES AND STUDY SELECTION\nThe databases of MEDLINE, CINAHL, and the Cochrane Library were searched from inception through March 2011, the EMBASE database was searched through August 2011, and hand searches were performed of the retrieved reference lists. Dual review was conducted to identify studies published in the English language of prediction models tested with medical patients in both derivation and validation cohorts.\n\n\nDATA EXTRACTION\nData were extracted on the population, setting, sample size, follow-up interval, readmission rate, model discrimination and calibration, type of data used, and timing of data collection.\n\n\nDATA SYNTHESIS\nOf 7843 citations reviewed, 30 studies of 26 unique models met the inclusion criteria. The most common outcome used was 30-day readmission; only 1 model specifically addressed preventable readmissions. Fourteen models that relied on retrospective administrative data could be potentially used to risk-adjust readmission rates for hospital comparison; of these, 9 were tested in large US populations and had poor discriminative ability (c statistic range: 0.55-0.65). Seven models could potentially be used to identify high-risk patients for intervention early during a hospitalization (c statistic range: 0.56-0.72), and 5 could be used at hospital discharge (c statistic range: 0.68-0.83). Six studies compared different models in the same population and 2 of these found that functional and social variables improved model discrimination. Although most models incorporated variables for medical comorbidity and use of prior medical services, few examined variables associated with overall health and function, illness severity, or social determinants of health.\n\n\nCONCLUSIONS\nMost current readmission risk prediction models that were designed for either comparative or clinical purposes perform poorly. Although in certain settings such models may prove useful, efforts to improve their performance are needed as use becomes more widespread."
                },
                {
                    "title": "Predicting hospital admissions at emergency department triage using routine administrative data.",
                    "abstract": "OBJECTIVES\nTo be able to predict, at the time of triage, whether a need for hospital admission exists for emergency department (ED) patients may constitute useful information that could contribute to systemwide hospital changes designed to improve ED throughput. The objective of this study was to develop and validate a predictive model to assess whether a patient is likely to require inpatient admission at the time of ED triage, using routine hospital administrative data.\n\n\nMETHODS\nData collected at the time of triage by nurses from patients who visited the ED in 2007 and 2008 were extracted from hospital administrative databases. Variables included were demographics (age, sex, and ethnic group), ED visit or hospital admission in the preceding 3 months, arrival mode, patient acuity category (PAC) of the ED visit, and coexisting chronic diseases (diabetes, hypertension, and dyslipidemia). Chi-square tests were used to study the association between the selected possible risk factors and the need for hospital admission. Logistic regression was applied to develop the prediction model. Data were split for derivation (60%) and validation (40%). Receiver operating characteristic curves and goodness-of-fit tests were applied to the validation data set to evaluate the model.\n\n\nRESULTS\nOf 317,581 ED patient visits, 30.2% resulted in immediate hospital admission. In the developed predictive model, age, PAC status, and arrival mode were most predictive of the need for immediate hospital inpatient admission. The c-statistic of the receiver operating characteristic (ROC) curve was 0.849 (95% confidence interval [CI] = 0.847 to 0.851). The goodness-of-fit test showed that the predicted patients' admission risks fit the patients' actual admission status well.\n\n\nCONCLUSIONS\nA model for predicting the risk of immediate hospital admission at triage for all-cause ED patients was developed and validated using routinely collected hospital data. Early prediction of the need for hospital admission at the time of triage may help identify patients deserving of early admission planning and resource allocation and thus potentially reduce ED overcrowding."
                },
                {
                    "title": "Comparison of implicit and explicit feedback from an online music recommendation service",
                    "abstract": "Explicit and implicit feedback exhibits different characteristics of users' preferences with both pros and cons. However, a combination of these two types of feedback provides another paradigm for recommender systems (RS). Their combination in a user preference model presents a number of challenges but can also overcome the problems associated with each other. In order to build an effective RS on combination of both types of feedback, we need to have comparative data allowing an understanding of the computation of user preferences. In this paper, we provide an overview of the differentiating characteristics of explicit and implicit feedback using datasets mined from Last.fm, an online music station and recommender service. The datasets consisted of explicit positive feedback (by loving tracks) and implicit feedback which is inherently positive (the number of times a track is played). Rather than relying on just one type of feedback, we present techniques for extracting user preferences from both. In order to compare and contrast the performances of these techniques, we carried out experiments using the Taste recommender system engine and the Last.fm datasets. Our results show that implicit and explicit positive feedback complements each other, with similar performances despite their different characteristics."
                },
                {
                    "title": "Bayesian data analysis.",
                    "abstract": "Bayesian methods have garnered huge interest in cognitive science as an approach to models of cognition and perception. On the other hand, Bayesian methods for data analysis have not yet made much headway in cognitive science against the institutionalized inertia of 20th century null hypothesis significance testing (NHST). Ironically, specific Bayesian models of cognition and perception may not long endure the ravages of empirical verification, but generic Bayesian methods for data analysis will eventually dominate. It is time that Bayesian data analysis became the norm for empirical methods in cognitive science. This article reviews a fatal flaw of NHST and introduces the reader to some benefits of Bayesian data analysis. The article presents illustrative examples of multiple comparisons in Bayesian analysis of variance and Bayesian approaches to statistical power. Copyright \u00a9 2010 John Wiley & Sons, Ltd. For further resources related to this article, please visit the WIREs website."
                },
                {
                    "title": "Bayesian Analysis for the Social Sciences",
                    "abstract": "List of Figures. List of Tables. Preface. Acknowledgments. Introduction. Part I: Introducing Bayesian Analysis. 1. The foundations of Bayesian inference. 1.1 What is probability? 1.2 Subjective probability in Bayesian statistics. 1.3 Bayes theorem, discrete case. 1.4 Bayes theorem, continuous parameter. 1.5 Parameters as random variables, beliefs as distributions. 1.6 Communicating the results of a Bayesian analysis. 1.7 Asymptotic properties of posterior distributions. 1.8 Bayesian hypothesis testing. 1.9 From subjective beliefs to parameters and models. 1.10 Historical note. 2. Getting started: Bayesian analysis for simple models. 2.1 Learning about probabilities, rates and proportions. 2.2 Associations between binary variables. 2.3 Learning from counts. 2.4 Learning about a normal mean and variance. 2.5 Regression models. 2.6 Further reading. Part II: Simulation Based Bayesian Analysis. 3. Monte Carlo methods. 3.1 Simulation consistency. 3.2 Inference for functions of parameters. 3.3 Marginalization via Monte Carlo integration. 3.4 Sampling algorithms. 3.5 Further reading. 4. Markov chains. 4.1 Notation and definitions. 4.2 Properties of Markov chains. 4.3 Convergence of Markov chains. 4.4 Limit theorems for Markov chains. 4.5 Further reading. 5. Markov chain Monte Carlo. 5.1 Metropolis-Hastings algorithm. 5.2 Gibbs sampling. 6. Implementing Markov chain Monte Carlo. 6.1 Software for Markov chain Monte Carlo. 6.2 Assessing convergence and run-length. 6.3 Working with BUGS/JAGS from R. 6.4 Tricks of the trade. 6.5 Other examples. 6.6 Further reading. Part III: Advanced Applications in the Social Sciences. 7. Hierarchical Statistical Models. 7.1 Data and parameters that vary by groups: the case for hierarchical modeling. 7.2 ANOVA as a hierarchical model. 7.3 Hierarchical models for longitudinal data. 7.4 Hierarchical models for non-normal data. 7.5 Multi-level models. 8. Bayesian analysis of choice making. 8.1 Regression models for binary responses. 8.2 Ordered outcomes. 8.3 Multinomial outcomes. 8.4 Multinomial probit. 9. Bayesian approaches to measurement. 9.1 Bayesian inference for latent states. 9.2 Factor analysis. 9.3 Item-response models. 9.4 Dynamic measurement models. Part IV: Appendices. Appendix A: Working with vectors and matrices. Appendix B: Probability review. B.1 Foundations of probability. B.2 Probability densities and mass functions. B.3 Convergence of sequences of random variabales. Appendix C: Proofs of selected propositions. C.1 Products of normal densities. C.2 Conjugate analysis of normal data. C.3 Asymptotic normality of the posterior density. References. Topic index. Author index."
                },
                {
                    "title": "Graphical Models, Exponential Families, and Variational Inference",
                    "abstract": "The formalism of probabilistic graphical models provides a unifying framework for capturing complex dependencies among random variables, and building large-scale multivariate statistical models. Graphical models have become a focus of research in many statistical, computational and mathematical fields, including bioinformatics, communication theory, statistical physics, combinatorial optimization, signal and image processing, information retrieval and statistical machine learning. Many problems that arise in specific instances \u2014 including the key problems of computing marginals and modes of probability distributions \u2014 are best studied in the general setting. Working with exponential family representations, and exploiting the conjugate duality between the cumulant function and the entropy for exponential families, we develop general variational representations of the problems of computing likelihoods, marginal probabilities and most probable configurations. We describe how a wide variety of algorithms \u2014 among them sum-product, cluster variational methods, expectation-propagation, mean field methods, max-product and linear programming relaxation, as well as conic programming relaxations \u2014 can all be understood in terms of exact or approximate forms of these variational representations. The variational approach provides a complementary alternative to Markov chain Monte Carlo as a general source of approximation methods for inference in large-scale statistical models."
                },
                {
                    "title": "Mostly Harmless Econometrics: An Empiricist's Companion",
                    "abstract": "The core methods in today's econometric toolkit are linear regression for statistical control, instrumental variables methods for the analysis of natural experiments, and differences-in-differences methods that exploit policy changes. In the modern experimentalist paradigm, these techniques address clear causal questions such as: Do smaller classes increase learning? Should wife batterers be arrested? How much does education raise wages? Mostly Harmless Econometrics shows how the basic tools of applied econometrics allow the data to speak. In addition to econometric essentials, Mostly Harmless Econometrics covers important new extensions--regression-discontinuity designs and quantile regression--as well as how to get standard errors right. Joshua Angrist and Jorn-Steffen Pischke explain why fancier econometric techniques are typically unnecessary and even dangerous. The applied econometric methods emphasized in this book are easy to use and relevant for many areas of contemporary social science. An irreverent review of econometric essentials A focus on tools that applied researchers use most Chapters on regression-discontinuity designs, quantile regression, and standard errors Many empirical examples A clear and concise resource with wide applications"
                },
                {
                    "title": "Sample Selection Models in R: Package sampleSelection",
                    "abstract": "This paper describes the implementation of Heckman-type sample selection models in R. We discuss the sample selection problem as well as the Heckman solution to it, and argue that although modern econometrics has non- and semiparametric estimation methods in its toolbox, Heckman models are an integral part of the modern applied analysis and econometrics syllabus. We describe the implementation of these models in the package sampleSelection and illustrate the usage of the package on several simulation and real data examples. Our examples demonstrate the effect of exclusion restrictions, identification at infinity and misspecification. We argue that the package can be used both in applied research and teaching."
                },
                {
                    "title": "The Black diaspora and health inequalities in the US and England: does where you go and how you get there make a difference?",
                    "abstract": "The relatively poor health of Black American people in the US and Black Caribbean people in England is a consistent finding in the health inequalities literature. Indeed, there are many similarities between the health, social, economic and demographic profiles of these two groups. However, there is evidence that Caribbean people in the US are faring considerably better. This paper explores differences in the social and economic position of Black American, Black Caribbean and white people in the US and Black Caribbean and white people in England, how these relate to ethnic inequalities in health, and may be underpinned by differences in patterns and contexts of migration. We use similar surveys from the US and England to explore these questions. The US data were drawn from the National Survey of American Life and the English data were drawn from the Health Survey for England and a follow up study. Findings show the advantaged health position of Caribbean American people in comparison with both Caribbean people in England and Black American people. Multivariate analyses indicate that these differences, and the differences in health between Black and white people in the two countries, are a consequence of social and economic inequalities."
                },
                {
                    "title": "Estimating disease prevalence using a population-based administrative healthcare database",
                    "abstract": "Aims: In \u00d6sterg\u00f6tland County, Sweden, all data on hospital care and primary healthcare (PHC) have been entered in a diagnosis-related administrative database since 1999. This database was used to estimate the prevalence of four chronic diseases and to examine the capture of data in PHC, outpatient hospital care, and inpatient hospital care, considered in different time frames. Methods: A case-finding algorithm identified patients with at least one healthcare contact involving a diagnosis of diabetes, hypertension, asthma, or chronic obstructive pulmonary disease (COPD) in 1999\u20142003. Prevalence rates were calculated as the ratio of the number of identified patients alive to the total number of inhabitants on 31 December 2003 (n~415,000). Results: Prevalence rates were 4.4% for diabetes, 10.3% for hypertension, 4.5% for asthma, and 1.2% for COPD. For all four diagnoses, the proportions of patients identified on only one healthcare level were greatest for PHC, reaching rates of 23%, 68%, 53%, and 48%, respectively. The cases identified solely in PHC comprised larger proportions of women and patients over the age of 65 years. Considering the proportion of patients identified in 2003 in relation to the total five-year period gave values of 71%, 50%, 38%, and 58%, respectively, for the four diagnoses. Conclusions: The administrative healthcare databases in Sweden today can be important tools in epidemiological research. However, data on several consecutive years and both PHC and hospital data are needed to achieve valid prevalence estimates."
                },
                {
                    "title": "Prevalence of heart disease--United States, 2005.",
                    "abstract": "Heart disease has been the leading cause of death in the United States for the past 80 years and is a major cause of disability. Heart disease also results in substantial health-care expenditures; for example, coronary heart disease is projected to cost an estimated 151.6 billion dollars in direct and indirect costs in 2007. Although some self-reported national data are available, state-specific prevalence data for heart disease have not been reported previously. In addition, although racial/ethnic, geographic, and sex differences in death rates for heart disease have been documented, less information has been available regarding the prevalence of persons living with heart disease. To estimate the prevalence of myocardial infarction (MI) and angina/coronary heart disease (CHD) in each of the 50 states, the District of Columbia (DC), Puerto Rico, and the U.S. Virgin Islands (USVI), CDC analyzed self-reported data from the 2005 Behavioral Risk Factor Surveillance System (BRFSS). This report summarizes the results of that analysis and provides the first state-based prevalence estimates of these heart diseases. The results indicated that substantial geographic, racial/ethnic, educational, and sex disparities existed in the prevalence of MI and angina/CHD. To lower the incidence of heart disease and meet the overall Healthy People 2010 goal to eliminate health disparities, public health programs should target disproportionately affected populations."
                },
                {
                    "title": "Bayesian estimation of disease prevalence and the parameters of diagnostic tests in the absence of a gold standard.",
                    "abstract": "It is common in population screening surveys or in the investigation of new diagnostic tests to have results from one or more tests investigating the same condition or disease, none of which can be considered a gold standard. For example, two methods often used in population-based surveys for estimating the prevalence of a parasitic or other infection are stool examinations and serologic testing. However, it is known that results from stool examinations generally underestimate the prevalence, while serology generally results in overestimation. Using a Bayesian approach, simultaneous inferences about the population prevalence and the sensitivity, specificity, and positive and negative predictive values of each diagnostic test are possible. The methods presented here can be applied to each test separately or to two or more tests combined. Marginal posterior densities of all parameters are estimated using the Gibbs sampler. The techniques are applied to the estimation of the prevalence of Strongyloides infection and to the investigation of the diagnostic test properties of stool examinations and serologic testing, using data from a survey of all Cambodian refugees who arrived in Montreal, Canada, during an 8-month period."
                },
                {
                    "title": "Projecting individualized probabilities of developing breast cancer for white females who are being examined annually.",
                    "abstract": "To assist in medical counseling, we present a method to estimate the chance that a woman with given age and risk factors will develop breast cancer over a specified interval. The risk factors used were age at menarche, age at first live birth, number of previous biopsies, and number of first-degree relatives with breast cancer. A model of relative risks for various combinations of these factors was developed from case-control data from the Breast Cancer Detection Demonstration Project (BCDDP). The model allowed for the fact that relative risks associated with previous breast biopsies were smaller for women aged 50 or more than for younger women. Thus, the proportional hazards models for those under age 50 and for those of age 50 or more. The baseline age-specific hazard rate, which is the rate for a patient without identified risk factors, is computed as the product of the observed age-specific composite hazard rate times the quantity 1 minus the attributable risk. We calculated individualized breast cancer probabilities from information on relative risks and the baseline hazard rate. These calculations take competing risks and the interval of risk into account. Our data were derived from women who participated in the BCDDP and who tended to return for periodic examinations. For this reason, the risk projections given are probably most reliable for counseling women who plan to be examined about once a year."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can machine learning models be effectively trained to make accurate predictions in healthcare settings when the training data is biased due to selective labeling, where only a subset of the population is tested based on historical decision-making?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the significant issue of bias in machine learning models, particularly in high-stakes domains like healthcare. By improving the accuracy of predictions for the entire population, we can enhance decision-making processes, leading to better patient outcomes and more equitable healthcare access. This research could pave the way for future studies that explore advanced methodologies for handling selective labels, ultimately contributing to the development of fairer and more effective machine learning applications across various fields.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the distribution shift between the tested and untested populations, which can lead to biased predictions if models are trained solely on historical data. Naive approaches may fail because they do not account for unobserved variables that influence decision-making, such as the severity of symptoms that are not recorded in the data. Additionally, the complexity of accurately modeling the underlying data-generating process and the need for domain-specific constraints to improve generalization add to the difficulty of addressing this problem.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often overlooked the implications of selective labeling and the resulting biases in training data. Many existing solutions have focused on generic methods that do not adequately address the unique challenges posed by different domains. Barriers such as a lack of understanding of the specific constraints within each domain and insufficient methodologies to incorporate these constraints have prevented effective solutions. Our approach aims to fill these gaps by leveraging domain-specific knowledge to enhance model training and improve generalization.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a machine learning model that incorporates domain-specific constraints to address the selective labeling issue. We will utilize a dataset from healthcare that includes both tested and untested populations, focusing on the prevalence of diseases as a constraint. The performance of the model will be evaluated using metrics such as accuracy, precision, and recall to assess its predictive capabilities across the entire population. We expect that our approach will lead to improved prediction accuracy for untested patients, thereby enhancing decision-making in healthcare settings."
            }
        },
        "author_data": {
            "8e9fa71a-31bc-42be-919a-657814124c67": {
                "pk": "8e9fa71a-31bc-42be-919a-657814124c67",
                "name": "Sidhika Balachandar",
                "collaborators": [
                    "Rajiv Movva",
                    "Kenny Peng",
                    "Gabriel Agostini",
                    "Nikhil Garg",
                    "E. Pierson",
                    "Dr. R. Chinnaiyan"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Artificial Intelligence",
                    "Cybersecurity",
                    "Smart Cities"
                ],
                "publications": [
                    {
                        "title": "Topics, Authors, and Institutions in Large Language Model Research: Trends from 17K arXiv Papers",
                        "abstract": "Large language models (LLMs) are dramatically influencing AI research, spurring discussions on what has changed so far and how to shape the field's future. To clarify such questions, we analyze a new dataset of 16,979 LLM-related arXiv papers, focusing on recent trends in 2023 vs. 2018-2022. First, we study disciplinary shifts: LLM research increasingly considers societal impacts, evidenced by 20x growth in LLM submissions to the Computers and Society sub-arXiv. An influx of new authors -- half of all first authors in 2023 -- are entering from non-NLP fields of CS, driving disciplinary expansion. Second, we study industry and academic publishing trends. Surprisingly, industry accounts for a smaller publication share in 2023, largely due to reduced output from Google and other Big Tech companies; universities in Asia are publishing more. Third, we study institutional collaboration: while industry-academic collaborations are common, they tend to focus on the same topics that industry focuses on rather than bridging differences. The most prolific institutions are all US- or China-based, but there is very little cross-country collaboration. We discuss implications around (1) how to support the influx of new authors, (2) how industry trends may affect academics, and (3) possible effects of (the lack of) collaboration."
                    },
                    {
                        "title": "Secure solutions for Smart City Command Control Centre using AIOT",
                        "abstract": "To build a robust secure solution for smart city IOT network from any Cyber attacks using Artificial Intelligence. In Smart City IOT network, data collected from different log collectors or direct sources from cloud or edge should harness the potential of AI. The smart city command and control center team will leverage these models and deploy it in different city IOT network to help on intrusion prediction, network packet surge, potential botnet attacks from external network. Some of the vital use cases considered based on the users of command-and-control center"
                    }
                ]
            },
            "c83b85f1-4538-430c-959d-7808255e081d": {
                "pk": "c83b85f1-4538-430c-959d-7808255e081d",
                "name": "Nikhil Garg",
                "collaborators": [
                    "Gabriel Agostini",
                    "E. Pierson",
                    "Rajiv Movva",
                    "S. Balachandar",
                    "Kenny Peng"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Urban Analytics",
                    "Machine Learning",
                    "Social Impact"
                ],
                "publications": [
                    {
                        "title": "Topics, Authors, and Institutions in Large Language Model Research: Trends from 17K arXiv Papers",
                        "abstract": "Large language models (LLMs) are dramatically influencing AI research, spurring discussions on what has changed so far and how to shape the field's future. To clarify such questions, we analyze a new dataset of 16,979 LLM-related arXiv papers, focusing on recent trends in 2023 vs. 2018-2022. First, we study disciplinary shifts: LLM research increasingly considers societal impacts, evidenced by 20x growth in LLM submissions to the Computers and Society sub-arXiv. An influx of new authors -- half of all first authors in 2023 -- are entering from non-NLP fields of CS, driving disciplinary expansion. Second, we study industry and academic publishing trends. Surprisingly, industry accounts for a smaller publication share in 2023, largely due to reduced output from Google and other Big Tech companies; universities in Asia are publishing more. Third, we study institutional collaboration: while industry-academic collaborations are common, they tend to focus on the same topics that industry focuses on rather than bridging differences. The most prolific institutions are all US- or China-based, but there is very little cross-country collaboration. We discuss implications around (1) how to support the influx of new authors, (2) how industry trends may affect academics, and (3) possible effects of (the lack of) collaboration."
                    },
                    {
                        "title": "A Bayesian Spatial Model to Correct Under-Reporting in Urban Crowdsourcing",
                        "abstract": "Decision-makers often observe the occurrence of events through a reporting process. City governments, for example, rely on resident reports to find and then resolve urban infrastructural problems such as fallen street trees, flooded basements, or rat infestations. Without additional assumptions, there is no way to distinguish events that occur but are not reported from events that truly did not occur--a fundamental problem in settings with positive-unlabeled data. Because disparities in reporting rates correlate with resident demographics, addressing incidents only on the basis of reports leads to systematic neglect in neighborhoods that are less likely to report events. We show how to overcome this challenge by leveraging the fact that events are spatially correlated. Our framework uses a Bayesian spatial latent variable model to infer event occurrence probabilities and applies it to storm-induced flooding reports in New York City, further pooling results across multiple storms. We show that a model accounting for under-reporting and spatial correlation predicts future reports more accurately than other models, and further induces a more equitable set of inspections: its allocations better reflect the population and provide equitable service to non-white, less traditionally educated, and lower-income residents. This finding reflects heterogeneous reporting behavior learned by the model: reporting rates are higher in Census tracts with higher populations, proportions of white residents, and proportions of owner-occupied households. Our work lays the groundwork for more equitable proactive government services, even with disparate reporting behavior."
                    }
                ]
            },
            "9f861b68-7ece-4438-afc2-c7f8321ebc74": {
                "pk": "9f861b68-7ece-4438-afc2-c7f8321ebc74",
                "name": "Emma Pierson",
                "collaborators": [
                    "Nikhil Garg",
                    "Rajiv Movva",
                    "Gabriel Agostini",
                    "Divya Shanmugam",
                    "Kaihua Hou",
                    "P. Pathak",
                    "S. Balachandar",
                    "Kenny Peng",
                    "Matt W Franchi",
                    "J.D. Zamfirescu-Pereira"
                ],
                "domain": [
                    "Fairness in AI",
                    "Healthcare Analytics",
                    "Social Impact",
                    "Urban Studies"
                ],
                "publications": [
                    {
                        "title": "Topics, Authors, and Institutions in Large Language Model Research: Trends from 17K arXiv Papers",
                        "abstract": "Large language models (LLMs) are dramatically influencing AI research, spurring discussions on what has changed so far and how to shape the field's future. To clarify such questions, we analyze a new dataset of 16,979 LLM-related arXiv papers, focusing on recent trends in 2023 vs. 2018-2022. First, we study disciplinary shifts: LLM research increasingly considers societal impacts, evidenced by 20x growth in LLM submissions to the Computers and Society sub-arXiv. An influx of new authors -- half of all first authors in 2023 -- are entering from non-NLP fields of CS, driving disciplinary expansion. Second, we study industry and academic publishing trends. Surprisingly, industry accounts for a smaller publication share in 2023, largely due to reduced output from Google and other Big Tech companies; universities in Asia are publishing more. Third, we study institutional collaboration: while industry-academic collaborations are common, they tend to focus on the same topics that industry focuses on rather than bridging differences. The most prolific institutions are all US- or China-based, but there is very little cross-country collaboration. We discuss implications around (1) how to support the influx of new authors, (2) how industry trends may affect academics, and (3) possible effects of (the lack of) collaboration."
                    },
                    {
                        "title": "Detecting disparities in police deployments using dashcam data",
                        "abstract": "Large-scale policing data is vital for detecting inequity in police behavior and policing algorithms. However, one important type of policing data remains largely unavailable within the United States: aggregated police deployment data capturing which neighborhoods have the heaviest police presences. Here we show that disparities in police deployment levels can be quantified by detecting police vehicles in dashcam images of public street scenes. Using a dataset of 24,803,854 dashcam images from rideshare drivers in New York City, we find that police vehicles can be detected with high accuracy (average precision 0.82, AUC 0.99) and identify 233,596 images which contain police vehicles. There is substantial inequality across neighborhoods in police vehicle deployment levels. The neighborhood with the highest deployment levels has almost 20 times higher levels than the neighborhood with the lowest. Two strikingly different types of areas experience high police vehicle deployments \u2014 1) dense, higher-income, commercial areas and 2) lower-income neighborhoods with higher proportions of Black and Hispanic residents. We discuss the implications of these disparities for policing equity and for algorithms trained on policing data."
                    },
                    {
                        "title": "A Bayesian Spatial Model to Correct Under-Reporting in Urban Crowdsourcing",
                        "abstract": "Decision-makers often observe the occurrence of events through a reporting process. City governments, for example, rely on resident reports to find and then resolve urban infrastructural problems such as fallen street trees, flooded basements, or rat infestations. Without additional assumptions, there is no way to distinguish events that occur but are not reported from events that truly did not occur--a fundamental problem in settings with positive-unlabeled data. Because disparities in reporting rates correlate with resident demographics, addressing incidents only on the basis of reports leads to systematic neglect in neighborhoods that are less likely to report events. We show how to overcome this challenge by leveraging the fact that events are spatially correlated. Our framework uses a Bayesian spatial latent variable model to infer event occurrence probabilities and applies it to storm-induced flooding reports in New York City, further pooling results across multiple storms. We show that a model accounting for under-reporting and spatial correlation predicts future reports more accurately than other models, and further induces a more equitable set of inspections: its allocations better reflect the population and provide equitable service to non-white, less traditionally educated, and lower-income residents. This finding reflects heterogeneous reporting behavior learned by the model: reporting rates are higher in Census tracts with higher populations, proportions of white residents, and proportions of owner-occupied households. Our work lays the groundwork for more equitable proactive government services, even with disparate reporting behavior."
                    },
                    {
                        "title": "Coarse race data conceals disparities in clinical risk score performance",
                        "abstract": "Healthcare data in the United States often records only a patient's coarse race group: for example, both Indian and Chinese patients are typically coded as\"Asian.\"It is unknown, however, whether this coarse coding conceals meaningful disparities in the performance of clinical risk scores across granular race groups. Here we show that it does. Using data from 418K emergency department visits, we assess clinical risk score performance disparities across 26 granular groups for three outcomes, five risk scores, and four performance metrics. Across outcomes and metrics, we show that the risk scores exhibit significant granular performance disparities within coarse race groups. In fact, variation in performance within coarse groups often *exceeds* the variation between coarse groups. We explore why these disparities arise, finding that outcome rates, feature distributions, and the relationships between features and outcomes all vary significantly across granular groups. Our results suggest that healthcare providers, hospital systems, and machine learning researchers should strive to collect, release, and use granular race data in place of coarse race data, and that existing analyses may significantly underestimate racial disparities in performance."
                    },
                    {
                        "title": "Coarse race data conceals disparities in clinical risk score model performance",
                        "abstract": "Healthcare data in the United States often records only a patient\u2019s coarse race group: for example, both Indian and Chinese patients are typically coded as \u201cAsian.\u201d It is unknown, however, whether this coarse coding conceals meaningful disparities in the performance of clinical risk scores across granular race groups. Here we show that it does. Using data from 418K emergency department visits, we assess clinical risk score performance disparities across granular race groups for three outcomes, \ufb01ve risk scores, and four performance metrics. Across outcomes and metrics, we show that there are signi\ufb01cant granular disparities in performance within coarse race categories. In fact, variation in performance metrics within coarse groups often exceeds the variation between coarse groups. We explore why these disparities arise, \ufb01nding that outcome rates, feature distributions, and the relationships between features and outcomes all vary signi\ufb01cantly across granular race categories. Our results suggest that healthcare providers, hospital systems, and machine learning researchers should strive to collect, release, and use granular race data in place of coarse race data, and that existing analyses may signi\ufb01cantly underestimate racial disparities in performance."
                    },
                    {
                        "title": "End-to-end Auditing of Decision Pipelines",
                        "abstract": "Many high-stakes policies can be modeled as a sequence of decisions along a pipeline . We are interested in auditing such pipelines for both efficiency and equity . Using a dataset of over 100,000 crowdsourced resident requests for potentially hazardous tree maintenance in New York City, we observe a sequence of city government decisions about whether to inspect and work on a reported incident. At each decision in the pipeline, we define parity definitions and tests to identify inefficient, inequitable treatment. Disparities in resource allocation and scheduling across census tracts are reported as preliminary results."
                    },
                    {
                        "title": "Patients cannot consent to care unless they know how much it costs",
                        "abstract": "Given the impact that medical expenses have, disclosing them should be a part of the informed consent process, argue Leah Pierson and Emma Pierson"
                    }
                ]
            }
        }
    },
    "2406.01249": {
        "paper_data": {
            "title": "Equivariant Machine Learning on Graphs with Nonlinear Spectral Filters",
            "url": "http://arxiv.org/abs/2406.01249v1",
            "arxiv_id": "2406.01249",
            "authors": [
                "Ya-Wei Eileen Lin",
                "Ronen Talmon",
                "Ron Levie"
            ],
            "abstract": "Equivariant machine learning is an approach for designing deep learning models that respect the symmetries of the problem, with the aim of reducing model complexity and improving generalization. In this paper, we focus on an extension of shift equivariance, which is the basis of convolution networks on images, to general graphs. Unlike images, graphs do not have a natural notion of domain translation. Therefore, we consider the graph functional shifts as the symmetry group: the unitary operators that commute with the graph shift operator. Notably, such symmetries operate in the signal space rather than directly in the spatial space. We remark that each linear filter layer of a standard spectral graph neural network (GNN) commutes with graph functional shifts, but the activation function breaks this symmetry. Instead, we propose nonlinear spectral filters (NLSFs) that are fully equivariant to graph functional shifts and show that they have universal approximation properties. The proposed NLSFs are based on a new form of spectral domain that is transferable between graphs. We demonstrate the superior performance of NLSFs over existing spectral GNNs in node and graph classification benchmarks.",
            "introduction": "   1 Introduction  In many fields, such as chemistry [Gilmer et\u00a0al., 2017], biology [Gaudelet et\u00a0al., 2021, Barabasi and Oltvai, 2004], social science [Borgatti et\u00a0al., 2009], and computer graphics [Wang et\u00a0al., 2019], data can be described by graphs. In recent years, there has been a tremendous interest in the development of machine learning models for graph-structured data [Wu et\u00a0al., 2020, Bronstein et\u00a0al., 2017]. This young field, often called graph machine learning (graph-ML) or graph representation learning, has made significant contributions to the applied sciences, e.g., in protein folding, molecular design, and drug discovery [Jumper et\u00a0al., 2021, Atz et\u00a0al., 2021, M\u00e9ndez-Lucio et\u00a0al., 2021, Stokes et\u00a0al., 2020], and has impacted the industry with applications in social media, recommendation systems, traffic prediction, computer graphics, and natural language processing, among others.   Geometric Deep Learning (GDL) [Bronstein et\u00a0al., 2017] is a design philosophy for machine learning models where the model is constructed to inherently respect symmetries present in the data, aiming to reduce model complexity and enhance generalization. By incorporating knowledge of these symmetries into the model, it avoids the need to expend parameters and data to learn them. This inherent respect for symmetries is automatically generalized to test data, thereby improving generalization [Bietti et\u00a0al., 2021, Petrache and Trivedi, 2024, Elesedy and Zaidi, 2021]. For instance, convolutional neural networks (CNNs) respect the translation symmetries of 2D images, with weight sharing due to these symmetries contributing significantly to their success [Krizhevsky et\u00a0al., 2012]. Respecting node re-indexing in a scalable manner revolutionized machine learning on graphs [Bronstein et\u00a0al., 2017, 2021] and has placed GNNs as a main general-purpose tool for processing graph-structured data. Moreover, within GNNs, respecting the 3D Euclidean symmetries of the laws of physics (rotations, reflections, and translations) led to state-of-the-art performance in molecule processing [Duval et\u00a0al., 2024].   Our Contribution. In this paper, we focus on GDL for graph-ML. We consider extensions of shift symmetries from images to general graphs. Since graphs do not have a natural notion of domain translation, as opposed to images, we propose considering functional translations instead. We model the group of translations on graphs as the group of all unitary operators on signals that commute with the graph shift operator. Such unitary operators are called graph functional shifts. Note that each linear filter layer of a standard spectral GNN commutes with graph functional shifts, but the activation function breaks this symmetry. Instead, we propose non-linear spectral filters (NLSFs) that are fully equivariant to graph functional shifts and have universal approximation properties.   In Sec.\u00a03, we introduce our NLSFs based on new notions of analysis and synthesis that map signals between their node-space representations and spectral representations. Our transforms are related to standard graph Fourier and inverse Fourier transforms but differ from them in one important aspect. One key property of our analysis transform is that it is independent of a specific choice of Laplacian eigenvectors. Hence, our spectral representations are transferable between graphs. In comparison, standard graph Fourier transforms are based on an arbitrary choice of eigenvectors, and therefore, the standard frequency domain is not transferable. To achieve transferability, standard graph Fourier methods resort to linear filter operations based on functional calculus. Since we do not have this limitation,",
            "references": [
                {
                    "title": "Learning on Large Graphs using Intersecting Communities",
                    "abstract": "Message Passing Neural Networks (MPNNs) are a staple of graph machine learning. MPNNs iteratively update each node's representation in an input graph by aggregating messages from the node's neighbors, which necessitates a memory complexity of the order of the number of graph edges. This complexity might quickly become prohibitive for large graphs provided they are not very sparse. In this paper, we propose a novel approach to alleviate this problem by approximating the input graph as an intersecting community graph (ICG) -- a combination of intersecting cliques. The key insight is that the number of communities required to approximate a graph does not depend on the graph size. We develop a new constructive version of the Weak Graph Regularity Lemma to efficiently construct an approximating ICG for any input graph. We then devise an efficient graph learning algorithm operating directly on ICG in linear memory and time with respect to the number of nodes (rather than edges). This offers a new and fundamentally different pipeline for learning on very large non-sparse graphs, whose applicability is demonstrated empirically on node classification tasks and spatio-temporal data processing."
                },
                {
                    "title": "Expressive Sign Equivariant Networks for Spectral Geometric Learning",
                    "abstract": "Recent work has shown the utility of developing machine learning models that respect the structure and symmetries of eigenvectors. These works promote sign invariance, since for any eigenvector v the negation -v is also an eigenvector. However, we show that sign invariance is theoretically limited for tasks such as building orthogonally equivariant models and learning node positional encodings for link prediction in graphs. In this work, we demonstrate the benefits of sign equivariance for these tasks. To obtain these benefits, we develop novel sign equivariant neural network architectures. Our models are based on a new analytic characterization of sign equivariant polynomials and thus inherit provable expressiveness properties. Controlled synthetic experiments show that our networks can achieve the theoretically predicted benefits of sign equivariant models. Code is available at https://github.com/cptq/Sign-Equivariant-Nets."
                },
                {
                    "title": "Approximately Equivariant Graph Networks",
                    "abstract": "Graph neural networks (GNNs) are commonly described as being permutation equivariant with respect to node relabeling in the graph. This symmetry of GNNs is often compared to the translation equivariance of Euclidean convolution neural networks (CNNs). However, these two symmetries are fundamentally different: The translation equivariance of CNNs corresponds to symmetries of the fixed domain acting on the image signals (sometimes known as active symmetries), whereas in GNNs any permutation acts on both the graph signals and the graph domain (sometimes described as passive symmetries). In this work, we focus on the active symmetries of GNNs, by considering a learning setting where signals are supported on a fixed graph. In this case, the natural symmetries of GNNs are the automorphisms of the graph. Since real-world graphs tend to be asymmetric, we relax the notion of symmetries by formalizing approximate symmetries via graph coarsening. We present a bias-variance formula that quantifies the tradeoff between the loss in expressivity and the gain in the regularity of the learned estimator, depending on the chosen symmetry group. To illustrate our approach, we conduct extensive experiments on image inpainting, traffic flow prediction, and human pose estimation with different choices of symmetries. We show theoretically and empirically that the best generalization performance can be achieved by choosing a suitably larger group than the graph automorphism, but smaller than the permutation group."
                },
                {
                    "title": "Approximation-Generalization Trade-offs under (Approximate) Group Equivariance",
                    "abstract": "The explicit incorporation of task-specific inductive biases through symmetry has emerged as a general design precept in the development of high-performance machine learning models. For example, group equivariant neural networks have demonstrated impressive performance across various domains and applications such as protein and drug design. A prevalent intuition about such models is that the integration of relevant symmetry results in enhanced generalization. Moreover, it is posited that when the data and/or the model may only exhibit $\\textit{approximate}$ or $\\textit{partial}$ symmetry, the optimal or best-performing model is one where the model symmetry aligns with the data symmetry. In this paper, we conduct a formal unified investigation of these intuitions. To begin, we present general quantitative bounds that demonstrate how models capturing task-specific symmetries lead to improved generalization. In fact, our results do not require the transformations to be finite or even form a group and can work with partial or approximate equivariance. Utilizing this quantification, we examine the more general question of model mis-specification i.e. when the model symmetries don't align with the data symmetries. We establish, for a given symmetry group, a quantitative comparison between the approximate/partial equivariance of the model and that of the data distribution, precisely connecting model equivariance error and data equivariance error. Our result delineates conditions under which the model equivariance error is optimal, thereby yielding the best-performing model for the given task and data."
                },
                {
                    "title": "A graphon-signal analysis of graph neural networks",
                    "abstract": "We present an approach for analyzing message passing graph neural networks (MPNNs) based on an extension of graphon analysis to a so called graphon-signal analysis. A MPNN is a function that takes a graph and a signal on the graph (a graph-signal) and returns some value. Since the input space of MPNNs is non-Euclidean, i.e., graphs can be of any size and topology, properties such as generalization are less well understood for MPNNs than for Euclidean neural networks. We claim that one important missing ingredient in past work is a meaningful notion of graph-signal similarity measure, that endows the space of inputs to MPNNs with a regular structure. We present such a similarity measure, called the graphon-signal cut distance, which makes the space of all graph-signals a dense subset of a compact metric space -- the graphon-signal space. Informally, two deterministic graph-signals are close in cut distance if they ``look like'' they were sampled from the same random graph-signal model. Hence, our cut distance is a natural notion of graph-signal similarity, which allows comparing any pair of graph-signals of any size and topology. We prove that MPNNs are Lipschitz continuous functions over the graphon-signal metric space. We then give two applications of this result: 1) a generalization bound for MPNNs, and, 2) the stability of MPNNs to subsampling of graph-signals. Our results apply to any regular enough MPNN on any distribution of graph-signals, making the analysis rather universal."
                },
                {
                    "title": "A critical look at the evaluation of GNNs under heterophily: are we really making progress?",
                    "abstract": "Node classification is a classical graph machine learning task on which Graph Neural Networks (GNNs) have recently achieved strong results. However, it is often believed that standard GNNs only work well for homophilous graphs, i.e., graphs where edges tend to connect nodes of the same class. Graphs without this property are called heterophilous, and it is typically assumed that specialized methods are required to achieve strong performance on such graphs. In this work, we challenge this assumption. First, we show that the standard datasets used for evaluating heterophily-specific models have serious drawbacks, making results obtained by using them unreliable. The most significant of these drawbacks is the presence of a large number of duplicate nodes in the datasets Squirrel and Chameleon, which leads to train-test data leakage. We show that removing duplicate nodes strongly affects GNN performance on these datasets. Then, we propose a set of heterophilous graphs of varying properties that we believe can serve as a better benchmark for evaluating the performance of GNNs under heterophily. We show that standard GNNs achieve strong results on these heterophilous graphs, almost always outperforming specialized models. Our datasets and the code for reproducing our experiments are available at https://github.com/yandex-research/heterophilous-graphs"
                },
                {
                    "title": "Towards fully covariant machine learning",
                    "abstract": "Any representation of data involves arbitrary investigator choices. Because those choices are external to the data-generating process, each choice leads to an exact symmetry, corresponding to the group of transformations that takes one possible representation to another. These are the passive symmetries; they include coordinate freedom, gauge symmetry, and units covariance, all of which have led to important results in physics. In machine learning, the most visible passive symmetry is the relabeling or permutation symmetry of graphs. Our goal is to understand the implications for machine learning of the many passive symmetries in play. We discuss dos and don'ts for machine learning practice if passive symmetries are to be respected. We discuss links to causal modeling, and argue that the implementation of passive symmetries is particularly valuable when the goal of the learning problem is to generalize out of sample. This paper is conceptual: It translates among the languages of physics, mathematics, and machine-learning. We believe that consideration and implementation of passive symmetries might help machine learning in the same ways that it transformed physics in the twentieth century."
                },
                {
                    "title": "Unveiling the Sampling Density in Non-Uniform Geometric Graphs",
                    "abstract": "A powerful framework for studying graphs is to consider them as geometric graphs: nodes are randomly sampled from an underlying metric space, and any pair of nodes is connected if their distance is less than a specified neighborhood radius. Currently, the literature mostly focuses on uniform sampling and constant neighborhood radius. However, real-world graphs are likely to be better represented by a model in which the sampling density and the neighborhood radius can both vary over the latent space. For instance, in a social network communities can be modeled as densely sampled areas, and hubs as nodes with larger neighborhood radius. In this work, we first perform a rigorous mathematical analysis of this (more general) class of models, including derivations of the resulting graph shift operators. The key insight is that graph shift operators should be corrected in order to avoid potential distortions introduced by the non-uniform sampling. Then, we develop methods to estimate the unknown sampling density in a self-supervised fashion. Finally, we present exemplary applications in which the learnt density is used to 1) correct the graph shift operator and improve performance on a variety of tasks, 2) improve pooling, and 3) extract knowledge from networks. Our experimental findings support our theory and provide strong evidence for our model."
                },
                {
                    "title": "Understanding and Extending Subgraph GNNs by Rethinking Their Symmetries",
                    "abstract": "Subgraph GNNs are a recent class of expressive Graph Neural Networks (GNNs) which model graphs as collections of subgraphs. So far, the design space of possible Subgraph GNN architectures as well as their basic theoretical properties are still largely unexplored. In this paper, we study the most prominent form of subgraph methods, which employs node-based subgraph selection policies such as ego-networks or node marking and deletion. We address two central questions: (1) What is the upper-bound of the expressive power of these methods? and (2) What is the family of equivariant message passing layers on these sets of subgraphs?. Our first step in answering these questions is a novel symmetry analysis which shows that modelling the symmetries of node-based subgraph collections requires a significantly smaller symmetry group than the one adopted in previous works. This analysis is then used to establish a link between Subgraph GNNs and Invariant Graph Networks (IGNs). We answer the questions above by first bounding the expressive power of subgraph methods by 3-WL, and then proposing a general family of message-passing layers for subgraph methods that generalises all previous node-based Subgraph GNNs. Finally, we design a novel Subgraph GNN dubbed SUN, which theoretically unifies previous architectures while providing better empirical performance on multiple benchmarks."
                },
                {
                    "title": "How Powerful are Spectral Graph Neural Networks",
                    "abstract": "Spectral Graph Neural Network is a kind of Graph Neural Network (GNN) based on graph signal filters. Some models able to learn arbitrary spectral filters have emerged recently. However, few works analyze the expressive power of spectral GNNs. This paper studies spectral GNNs' expressive power theoretically. We first prove that even spectral GNNs without nonlinearity can produce arbitrary graph signals and give two conditions for reaching universality. They are: 1) no multiple eigenvalues of graph Laplacian, and 2) no missing frequency components in node features. We also establish a connection between the expressive power of spectral GNNs and Graph Isomorphism (GI) testing, the latter of which is often used to characterize spatial GNNs' expressive power. Moreover, we study the difference in empirical performance among different spectral GNNs with the same expressive power from an optimization perspective, and motivate the use of an orthogonal basis whose weight function corresponds to the graph signal density in the spectrum. Inspired by the analysis, we propose JacobiConv, which uses Jacobi basis due to its orthogonality and flexibility to adapt to a wide range of weight functions. JacobiConv deserts nonlinearity while outperforming all baselines on both synthetic and real-world datasets."
                },
                {
                    "title": "Finding Global Homophily in Graph Neural Networks When Meeting Heterophily",
                    "abstract": "We investigate graph neural networks on graphs with heterophily. Some existing methods amplify a node's neighborhood with multi-hop neighbors to include more nodes with homophily. However, it is a significant challenge to set personalized neighborhood sizes for different nodes. Further, for other homophilous nodes excluded in the neighborhood, they are ignored for information aggregation. To address these problems, we propose two models GloGNN and GloGNN++, which generate a node's embedding by aggregating information from global nodes in the graph. In each layer, both models learn a coefficient matrix to capture the correlations between nodes, based on which neighborhood aggregation is performed. The coefficient matrix allows signed values and is derived from an optimization problem that has a closed-form solution. We further accelerate neighborhood aggregation and derive a linear time complexity. We theoretically explain the models' effectiveness by proving that both the coefficient matrix and the generated node embedding matrix have the desired grouping effect. We conduct extensive experiments to compare our models against 11 other competitors on 15 benchmark datasets in a wide range of domains, scales and graph heterophilies. Experimental results show that our methods achieve superior performance and are also very efficient."
                },
                {
                    "title": "Convolutional Neural Networks on Graphs with Chebyshev Approximation, Revisited",
                    "abstract": "Designing spectral convolutional networks is a challenging problem in graph learning. ChebNet, one of the early attempts, approximates the spectral graph convolutions using Chebyshev polynomials. GCN simplifies ChebNet by utilizing only the first two Chebyshev polynomials while still outperforming it on real-world datasets. GPR-GNN and BernNet demonstrate that the Monomial and Bernstein bases also outperform the Chebyshev basis in terms of learning the spectral graph convolutions. Such conclusions are counter-intuitive in the field of approximation theory, where it is established that the Chebyshev polynomial achieves the optimum convergent rate for approximating a function. In this paper, we revisit the problem of approximating the spectral graph convolutions with Chebyshev polynomials. We show that ChebNet's inferior performance is primarily due to illegal coefficients learnt by ChebNet approximating analytic filter functions, which leads to over-fitting. We then propose ChebNetII, a new GNN model based on Chebyshev interpolation, which enhances the original Chebyshev polynomial approximation while reducing the Runge phenomenon. We conducted an extensive experimental study to demonstrate that ChebNetII can learn arbitrary graph convolutions and achieve superior performance in both full- and semi-supervised node classification tasks. Most notably, we scale ChebNetII to a billion graph ogbn-papers100M, showing that spectral-based GNNs have superior performance. Our code is available at https://github.com/ivam-he/ChebNetII."
                },
                {
                    "title": "GBK-GNN: Gated Bi-Kernel Graph Neural Networks for Modeling Both Homophily and Heterophily",
                    "abstract": "Graph Neural Networks (GNNs) are widely used on a variety of graph-based machine learning tasks. For node-level tasks, GNNs have strong power to model the homophily property of graphs (i.e., connected nodes are more similar), while their ability to capture heterophily property is often doubtful. This is partially caused by the design of the feature transformation with the same kernel for the nodes in the same hop and the followed aggregation operator. One kernel cannot model the similarity and the dissimilarity (i.e., the positive and negative correlation) between node features simultaneously even though we use attention mechanisms like Graph Attention Network (GAT), since the weight calculated by attention is always a positive value. In this paper, we propose a novel GNN model based on a bi-kernel feature transformation and a selection gate. Two kernels capture homophily and heterophily information respectively, and the gate is introduced to select which kernel we should use for the given node pairs. We conduct extensive experiments on various datasets with different homophily-heterophily properties. The experimental results show consistent and significant improvements against state-of-the-art GNN methods."
                },
                {
                    "title": "Equivariant Subgraph Aggregation Networks",
                    "abstract": "Message-passing neural networks (MPNNs) are the leading architecture for deep learning on graph-structured data, in large part due to their simplicity and scalability. Unfortunately, it was shown that these architectures are limited in their expressive power. This paper proposes a novel framework called Equivariant Subgraph Aggregation Networks (ESAN) to address this issue. Our main observation is that while two graphs may not be distinguishable by an MPNN, they often contain distinguishable subgraphs. Thus, we propose to represent each graph as a set of subgraphs derived by some predefined policy, and to process it using a suitable equivariant architecture. We develop novel variants of the 1-dimensional Weisfeiler-Leman (1-WL) test for graph isomorphism, and prove lower bounds on the expressiveness of ESAN in terms of these new WL variants. We further prove that our approach increases the expressive power of both MPNNs and more expressive architectures. Moreover, we provide theoretical results that describe how design choices such as the subgraph selection policy and equivariant neural architecture affect our architecture's expressive power. To deal with the increased computational cost, we propose a subgraph sampling scheme, which can be viewed as a stochastic version of our framework. A comprehensive set of experiments on real and synthetic datasets demonstrates that our framework improves the expressive power and overall performance of popular GNN architectures."
                },
                {
                    "title": "Utilizing graph machine learning within drug discovery and development",
                    "abstract": "Abstract Graph machine learning (GML) is receiving growing interest within the pharmaceutical and biotechnology industries for its ability to model biomolecular structures, the functional relationships between them, and integrate multi-omic datasets \u2014 amongst other data types. Herein, we present a multidisciplinary academic-industrial review of the topic within the context of drug discovery and development. After introducing key terms and modelling approaches, we move chronologically through the drug development pipeline to identify and summarize work incorporating: target identification, design of small molecules and biologics, and drug repurposing. Whilst the field is still emerging, key milestones including repurposed drugs entering in vivo studies, suggest GML will become a modelling framework of choice within biomedical machine learning."
                },
                {
                    "title": "Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges",
                    "abstract": "The last decade has witnessed an experimental revolution in data science and machine learning, epitomised by deep learning methods. Indeed, many high-dimensional learning tasks previously thought to be beyond reach -- such as computer vision, playing Go, or protein folding -- are in fact feasible with appropriate computational scale. Remarkably, the essence of deep learning is built from two simple algorithmic principles: first, the notion of representation or feature learning, whereby adapted, often hierarchical, features capture the appropriate notion of regularity for each task, and second, learning by local gradient-descent type methods, typically implemented as backpropagation. While learning generic functions in high dimensions is a cursed estimation problem, most tasks of interest are not generic, and come with essential pre-defined regularities arising from the underlying low-dimensionality and structure of the physical world. This text is concerned with exposing these regularities through unified geometric principles that can be applied throughout a wide spectrum of applications. Such a 'geometric unification' endeavour, in the spirit of Felix Klein's Erlangen Program, serves a dual purpose: on one hand, it provides a common mathematical framework to study the most successful neural network architectures, such as CNNs, RNNs, GNNs, and Transformers. On the other hand, it gives a constructive procedure to incorporate prior physical knowledge into neural architectures and provide principled way to build future architectures yet to be invented."
                },
                {
                    "title": "Provably Strict Generalisation Benefit for Equivariant Models",
                    "abstract": "It is widely believed that engineering a model to be invariant/equivariant improves generalisation. Despite the growing popularity of this approach, a precise characterisation of the generalisation benefit is lacking. By considering the simplest case of linear models, this paper provides the first provably non-zero improvement in generalisation for invariant/equivariant models when the target distribution is invariant/equivariant with respect to a compact group. Moreover, our work reveals an interesting relationship between generalisation, the number of training examples and properties of the group action. Our results rest on an observation of the structure of function spaces under averaging operators which, along with its consequences for feature averaging, may be of independent interest."
                },
                {
                    "title": "Beyond Low-frequency Information in Graph Convolutional Networks",
                    "abstract": "Graph neural networks (GNNs) have been proven to be effective in various network-related tasks. Most existing GNNs usually exploit the low-frequency signals of node features, which gives rise to one fundamental question: is the low-frequency information all we need in the real world applications? In this paper, we first present an experimental investigation assessing the roles of low-frequency and high-frequency signals, where the results clearly show that exploring low-frequency signal only is distant from learning an effective node representation in different scenarios. How can we adaptively learn more information beyond low-frequency information in GNNs? A well-informed answer can help GNNs enhance the adaptability. We tackle this challenge and propose a novel Frequency Adaptation Graph Convolutional Networks (FAGCN) with a self-gating mechanism, which can adaptively integrate different signals in the process of message passing. For a deeper understanding, we theoretically analyze the roles of low-frequency signals and high-frequency signals on learning node representations, which further explains why FAGCN can perform well on different types of networks. Extensive experiments on six real-world networks validate that FAGCN not only alleviates the over-smoothing problem, but also has advantages over the state-of-the-arts."
                },
                {
                    "title": "Graph Contrastive Learning with Adaptive Augmentation",
                    "abstract": "Recently, contrastive learning (CL) has emerged as a successful method for unsupervised graph representation learning. Most graph CL methods first perform stochastic augmentation on the input graph to obtain two graph views and maximize the agreement of representations in the two views. Despite the prosperous development of graph CL methods, the design of graph augmentation schemes\u2014a crucial component in CL\u2014remains rarely explored. We argue that the data augmentation schemes should preserve intrinsic structures and attributes of graphs, which will force the model to learn representations that are insensitive to perturbation on unimportant nodes and edges. However, most existing methods adopt uniform data augmentation schemes, like uniformly dropping edges and uniformly shuffling features, leading to suboptimal performance. In this paper, we propose a novel graph contrastive representation learning method with adaptive augmentation that incorporates various priors for topological and semantic aspects of the graph. Specifically, on the topology level, we design augmentation schemes based on node centrality measures to highlight important connective structures. On the node attribute level, we corrupt node features by adding more noise to unimportant node features, to enforce the model to recognize underlying semantic information. We perform extensive experiments of node classification on a variety of real-world datasets. Experimental results demonstrate that our proposed method consistently outperforms existing state-of-the-art baselines and even surpasses some supervised counterparts, which validates the effectiveness of the proposed contrastive framework with adaptive augmentation."
                },
                {
                    "title": "Graph Neural Networks with Heterophily",
                    "abstract": "Graph Neural Networks (GNNs) have proven to be useful for many different practical applications. However, many existing GNN models have implicitly assumed homophily among the nodes connected in the graph, and therefore have largely overlooked the important setting of heterophily, where most connected nodes are from different classes. In this work, we propose a novel framework called CPGNN that generalizes GNNs for graphs with either homophily or heterophily. The proposed framework incorporates an interpretable compatibility matrix for modeling the heterophily or homophily level in the graph, which can be learned in an end-to-end fashion, enabling it to go beyond the assumption of strong homophily. Theoretically, we show that replacing the compatibility matrix in our framework with the identity (which represents pure homophily) reduces to GCN. Our extensive experiments demonstrate the effectiveness of our approach in more realistic and challenging experimental settings with significantly less training data compared to previous works: CPGNN variants achieve state-of-the-art results in heterophily settings with or without contextual node features, while maintaining comparable performance in homophily settings."
                },
                {
                    "title": "Masked Label Prediction: Unified Massage Passing Model for Semi-Supervised Classification",
                    "abstract": "Graph neural network (GNN) and label propagation algorithm (LPA) are both message passing algorithms, which have achieved superior performance in semi-supervised classification. GNN performs feature propagation by a neural network to make predictions, while LPA uses label propagation across graph adjacency matrix to get results. However, there is still no effective way to directly combine these two kinds of algorithms. To address this issue, we propose a novel Unified Message Passaging Model (UniMP) that can incorporate feature and label propagation at both training and inference time. First, UniMP adopts a Graph Transformer network, taking feature embedding and label embedding as input information for propagation. Second, to train the network without overfitting in self-loop input label information, UniMP introduces a masked label prediction strategy, in which some percentage of input label information are masked at random, and then predicted. UniMP conceptually unifies feature propagation and label propagation and is empirically powerful. It obtains new state-of-the-art semi-supervised classification results in Open Graph Benchmark (OGB)."
                },
                {
                    "title": "TUDataset: A collection of benchmark datasets for learning with graphs",
                    "abstract": "Recently, there has been an increasing interest in (supervised) learning with graph data, especially using graph neural networks. However, the development of meaningful benchmark datasets and standardized evaluation procedures is lagging, consequently hindering advancements in this area. To address this, we introduce the TUDataset for graph classification and regression. The collection consists of over 120 datasets of varying sizes from a wide range of applications. We provide Python-based data loaders, kernel and graph neural network baseline implementations, and evaluation tools. Here, we give an overview of the datasets, standardized evaluation procedures, and provide baseline experiments. All datasets are available at this http URL. The experiments are fully reproducible from the code available at this http URL."
                },
                {
                    "title": "Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs",
                    "abstract": "We investigate the representation power of graph neural networks in the semi-supervised node classification task under heterophily or low homophily, i.e., in networks where connected nodes may have different class labels and dissimilar features. Many popular GNNs fail to generalize to this setting, and are even outperformed by models that ignore the graph structure (e.g., multilayer perceptrons). Motivated by this limitation, we identify a set of key designs -- ego- and neighbor-embedding separation, higher-order neighborhoods, and combination of intermediate representations -- that boost learning from the graph structure under heterophily. We combine them into a graph neural network, H2GCN, which we use as the base method to empirically evaluate the effectiveness of the identified designs. Going beyond the traditional benchmarks with strong homophily, our empirical analysis shows that the identified designs increase the accuracy of GNNs by up to 40% and 27% over models without them on synthetic and real networks with heterophily, respectively, and yield competitive performance under homophily."
                },
                {
                    "title": "Adaptive Universal Generalized PageRank Graph Neural Network",
                    "abstract": "In many important graph data processing applications the acquired information includes both node features and observations of the graph topology. Graph neural networks (GNNs) are designed to exploit both sources of evidence but they do not optimally trade-off their utility and integrate them in a manner that is also universal. Here, universality refers to independence on homophily or heterophily graph assumptions. We address these issues by introducing a new Generalized PageRank (GPR) GNN architecture that adaptively learns the GPR weights so as to jointly optimize node feature and topological information extraction, regardless of the extent to which the node labels are homophilic or heterophilic. Learned GPR weights automatically adjust to the node label pattern, irrelevant on the type of initialization, and thereby guarantee excellent learning performance for label patterns that are usually hard to handle. Furthermore, they allow one to avoid feature over-smoothing, a process which renders feature information nondiscriminative, without requiring the network to be shallow. Our accompanying theoretical analysis of the GPR-GNN method is facilitated by novel synthetic benchmark datasets generated by the so-called contextual stochastic block model. We also compare the performance of our GNN architecture with that of several state-of-the-art GNNs on the problem of node-classification, using well-known benchmark homophilic and heterophilic datasets. The results demonstrate that GPR-GNN offers significant performance improvement compared to existing techniques on both synthetic and benchmark data."
                },
                {
                    "title": "Geom-GCN: Geometric Graph Convolutional Networks",
                    "abstract": "Message-passing neural networks (MPNNs) have been successfully applied in a wide variety of applications in the real world. However, two fundamental weaknesses of MPNNs' aggregators limit their ability to represent graph-structured data: losing the structural information of nodes in neighborhoods and lacking the ability to capture long-range dependencies in disassortative graphs. Few studies have noticed the weaknesses from different perspectives. From the observations on classical neural network and network geometry, we propose a novel geometric aggregation scheme for graph neural networks to overcome the two weaknesses. The behind basic idea is the aggregation on a graph can benefit from a continuous space underlying the graph. The proposed aggregation scheme is permutation-invariant and consists of three modules, node embedding, structural neighborhood, and bi-level aggregation. We also present an implementation of the scheme in graph convolutional networks, termed Geom-GCN, to perform transductive learning on graphs. Experimental results show the proposed Geom-GCN achieved state-of-the-art performance on a wide range of open datasets of graphs."
                },
                {
                    "title": "A Fair Comparison of Graph Neural Networks for Graph Classification",
                    "abstract": "Experimental reproducibility and replicability are critical topics in machine learning. Authors have often raised concerns about their lack in scientific publications to improve the quality of the field. Recently, the graph representation learning field has attracted the attention of a wide research community, which resulted in a large stream of works. As such, several Graph Neural Network models have been developed to effectively tackle graph classification. However, experimental procedures often lack rigorousness and are hardly reproducible. Motivated by this, we provide an overview of common practices that should be avoided to fairly compare with the state of the art. To counter this troubling trend, we ran more than 47000 experiments in a controlled and uniform framework to re-evaluate five popular models across nine common benchmarks. Moreover, by comparing GNNs with structure-agnostic baselines we provide convincing evidence that, on some datasets, structural information has not been exploited yet. We believe that this work can contribute to the development of the graph learning field, by providing a much needed grounding for rigorous evaluations of graph classification models."
                },
                {
                    "title": "Multi-scale Attributed Node Embedding",
                    "abstract": "\n We present network embedding algorithms that capture information about a node from the local distribution over node attributes around it, as observed over random walks following an approach similar to Skip-gram. Observations from neighbourhoods of different sizes are either pooled (AE) or encoded distinctly in a multi-scale approach (MUSAE). Capturing attribute-neighbourhood relationships over multiple scales is useful for a range of applications, including latent feature identification across disconnected networks with similar features. We prove theoretically that matrices of node-feature pointwise mutual information are implicitly factorized by the embeddings. Experiments show that our algorithms are computationally efficient and outperform comparable models on social networks and web graphs."
                },
                {
                    "title": "Transferability of Spectral Graph Convolutional Neural Networks",
                    "abstract": "This paper focuses on spectral graph convolutional neural networks (ConvNets), where filters are defined as elementwise multiplication in the frequency domain of a graph. In machine learning settings where the dataset consists of signals defined on many different graphs, the trained ConvNet should generalize to signal on graphs unseen in the training set. It is thus important to transfer filters from one graph to the other. Transferability, which is a certain type of generalization capability, can be loosely defined as follows: if two graphs describe the same phenomenon, then a single filter/ConvNet should have similar repercussions on both graphs. This paper aims at debunking the common misconception that spectral filters are not transferable. We show that if two graphs discretize the same continuous metric space, then a spectral filter/ConvNet has approximately the same repercussion on both graphs. Our analysis is more permissive than the standard analysis. Transferability is typically described as the robustness of the filter to small graph perturbations and re-indexing of the vertices. Our analysis accounts also for large graph perturbations. We prove transferability between graphs that can have completely different dimensions and topologies, only requiring that both graphs discretize the same underlying continuous space."
                },
                {
                    "title": "Optuna: A Next-generation Hyperparameter Optimization Framework",
                    "abstract": "The purpose of this study is to introduce new design-criteria for next-generation hyperparameter optimization software. The criteria we propose include (1) define-by-run API that allows users to construct the parameter search space dynamically, (2) efficient implementation of both searching and pruning strategies, and (3) easy-to-setup, versatile architecture that can be deployed for various purposes, ranging from scalable distributed computing to light-weight experiment conducted via interactive interface. In order to prove our point, we will introduce Optuna, an optimization software which is a culmination of our effort in the development of a next generation optimization software. As an optimization software designed with define-by-run principle, Optuna is particularly the first of its kind. We will present the design-techniques that became necessary in the development of the software that meets the above criteria, and demonstrate the power of our new design through experimental results and real world applications. Our software is available under the MIT license (https://github.com/pfnet/optuna/)."
                },
                {
                    "title": "Graph Convolutional Networks with EigenPooling",
                    "abstract": "Graph neural networks, which generalize deep neural network models to graph structured data, have attracted increasing attention in recent years. They usually learn node representations by transforming, propagating and aggregating node features and have been proven to improve the performance of many graph related tasks such as node classification and link prediction. To apply graph neural networks for the graph classification task, approaches to generate thegraph representation from node representations are demanded. A common way is to globally combine the node representations. However, rich structural information is overlooked. Thus a hierarchical pooling procedure is desired to preserve the graph structure during the graph representation learning. There are some recent works on hierarchically learning graph representation analogous to the pooling step in conventional convolutional neural (CNN) networks. However, the local structural information is still largely neglected during the pooling process. In this paper, we introduce a pooling operator $\\pooling$ based on graph Fourier transform, which can utilize the node features and local structures during the pooling process. We then design pooling layers based on the pooling operator, which are further combined with traditional GCN convolutional layers to form a graph neural network framework $\\m$ for graph classification. Theoretical analysis is provided to understand $\\pooling$ from both local and global perspectives. Experimental results of the graph classification task on $6$ commonly used benchmarks demonstrate the effectiveness of the proposed framework."
                },
                {
                    "title": "Fast Graph Representation Learning with PyTorch Geometric",
                    "abstract": "We introduce PyTorch Geometric, a library for deep learning on irregularly structured input data such as graphs, point clouds and manifolds, built upon PyTorch. In addition to general graph data structures and processing methods, it contains a variety of recently published methods from the domains of relational learning and 3D data processing. PyTorch Geometric achieves high data throughput by leveraging sparse GPU acceleration, by providing dedicated CUDA kernels and by introducing efficient mini-batch handling for input examples of different size. In this work, we present the library in detail and perform a comprehensive comparative study of the implemented methods in homogeneous evaluation scenarios."
                },
                {
                    "title": "Graph Neural Networks With Convolutional ARMA Filters",
                    "abstract": "Popular graph neural networks implement convolution operations on graphs based on polynomial spectral filters. In this paper, we propose a novel graph convolutional layer inspired by the auto-regressive moving average (ARMA) filter that, compared to polynomial ones, provides a more flexible frequency response, is more robust to noise, and better captures the global graph structure. We propose a graph neural network implementation of the ARMA filter with a recursive and distributed formulation, obtaining a convolutional layer that is efficient to train, localized in the node space, and can be transferred to new graphs at test time. We perform a spectral analysis to study the filtering effect of the proposed ARMA layer and report experiments on four downstream tasks: semi-supervised node classification, graph signal classification, graph classification, and graph regression. Results show that the proposed ARMA layer brings significant improvements over graph neural networks based on polynomial filters."
                },
                {
                    "title": "Predict then Propagate: Graph Neural Networks meet Personalized PageRank",
                    "abstract": "Neural message passing algorithms for semi-supervised classification on graphs have recently achieved great success. However, for classifying a node these methods only consider nodes that are a few propagation steps away and the size of this utilized neighborhood is hard to extend. In this paper, we use the relationship between graph convolutional networks (GCN) and PageRank to derive an improved propagation scheme based on personalized PageRank. We utilize this propagation procedure to construct a simple model, personalized propagation of neural predictions (PPNP), and its fast approximation, APPNP. Our model's training time is on par or faster and its number of parameters on par or lower than previous models. It leverages a large, adjustable neighborhood for classification and can be easily combined with any neural network. We show that this model outperforms several recently proposed methods for semi-supervised classification in the most thorough study done so far for GCN-like models. Our implementation is available online."
                },
                {
                    "title": "Deep Graph Infomax",
                    "abstract": "We present Deep Graph Infomax (DGI), a general approach for learning node representations within graph-structured data in an unsupervised manner. DGI relies on maximizing mutual information between patch representations and corresponding high-level summaries of graphs---both derived using established graph convolutional network architectures. The learnt patch representations summarize subgraphs centered around nodes of interest, and can thus be reused for downstream node-wise learning tasks. In contrast to most prior approaches to unsupervised learning with GCNs, DGI does not rely on random walk objectives, and is readily applicable to both transductive and inductive learning setups. We demonstrate competitive performance on a variety of node classification benchmarks, which at times even exceeds the performance of supervised learning."
                },
                {
                    "title": "Hierarchical Graph Representation Learning with Differentiable Pooling",
                    "abstract": "Recently, graph neural networks (GNNs) have revolutionized the field of graph representation learning through effectively learned node embeddings, and achieved state-of-the-art results in tasks such as node classification and link prediction. However, current GNN methods are inherently flat and do not learn hierarchical representations of graphs---a limitation that is especially problematic for the task of graph classification, where the goal is to predict the label associated with an entire graph. Here we propose DiffPool, a differentiable graph pooling module that can generate hierarchical representations of graphs and can be combined with various graph neural network architectures in an end-to-end fashion. DiffPool learns a differentiable soft cluster assignment for nodes at each layer of a deep GNN, mapping nodes to a set of clusters, which then form the coarsened input for the next GNN layer. Our experimental results show that combining existing GNN methods with DiffPool yields an average improvement of 5-10% accuracy on graph classification benchmarks, compared to all existing pooling approaches, achieving a new state-of-the-art on four out of five benchmark datasets."
                },
                {
                    "title": "Iteration Method for the Solution of the Eigenvalue Problem of Linear Differential and Integral Operators",
                    "abstract": "The present investigation designs a systematic method for finding the latent roots and the principal axes of a matrix, without reducing the order of the matrix. It is characterized by a wide field of applicability and great accuracy, since the accumulation of rounding errors is avoided, through the process of \"minimized iterations\". Moreover, the method leads to a well convergent successive approximation procedure by which the solution of integral equations of the Fredholm type and the solution of the eigenvalue problem of linear differential and integral operators may be accomplished."
                },
                {
                    "title": "Dynamic Graph CNN for Learning on Point Clouds",
                    "abstract": "Point clouds provide a flexible geometric representation suitable for countless applications in computer graphics; they also comprise the raw output of most 3D data acquisition devices. While hand-designed features on point clouds have long been proposed in graphics and vision, however, the recent overwhelming success of convolutional neural networks (CNNs) for image analysis suggests the value of adapting insight from CNN to the point cloud world. Point clouds inherently lack topological information, so designing a model to recover topology can enrich the representation power of point clouds. To this end, we propose a new neural network module dubbed EdgeConv suitable for CNN-based high-level tasks on point clouds, including classification and segmentation. EdgeConv acts on graphs dynamically computed in each layer of the network. It is differentiable and can be plugged into existing architectures. Compared to existing modules operating in extrinsic space or treating each point independently, EdgeConv has several appealing properties: It incorporates local neighborhood information; it can be stacked applied to learn global shape properties; and in multi-layer systems affinity in feature space captures semantic characteristics over potentially long distances in the original embedding. We show the performance of our model on standard benchmarks, including ModelNet40, ShapeNetPart, and S3DIS."
                },
                {
                    "title": "Graph Attention Networks",
                    "abstract": "We present graph attention networks (GATs), novel neural network architectures that operate on graph-structured data, leveraging masked self-attentional layers to address the shortcomings of prior methods based on graph convolutions or their approximations. By stacking layers in which nodes are able to attend over their neighborhoods' features, we enable (implicitly) specifying different weights to different nodes in a neighborhood, without requiring any kind of costly matrix operation (such as inversion) or depending on knowing the graph structure upfront. In this way, we address several key challenges of spectral-based graph neural networks simultaneously, and make our model readily applicable to inductive as well as transductive problems. Our GAT models have achieved or matched state-of-the-art results across four established transductive and inductive graph benchmarks: the Cora, Citeseer and Pubmed citation network datasets, as well as a protein-protein interaction dataset (wherein test graphs remain unseen during training)."
                },
                {
                    "title": "Inductive Representation Learning on Large Graphs",
                    "abstract": "Low-dimensional embeddings of nodes in large graphs have proved extremely useful in a variety of prediction tasks, from content recommendation to identifying protein functions. However, most existing approaches require that all nodes in the graph are present during training of the embeddings; these previous approaches are inherently transductive and do not naturally generalize to unseen nodes. Here we present GraphSAGE, a general, inductive framework that leverages node feature information (e.g., text attributes) to efficiently generate node embeddings for previously unseen data. Instead of training individual embeddings for each node, we learn a function that generates embeddings by sampling and aggregating features from a node's local neighborhood. Our algorithm outperforms strong baselines on three inductive node-classification benchmarks: we classify the category of unseen nodes in evolving information graphs based on citation and Reddit post data, and we show that our algorithm generalizes to completely unseen graphs using a multi-graph dataset of protein-protein interactions."
                },
                {
                    "title": "CayleyNets: Graph Convolutional Neural Networks With Complex Rational Spectral Filters",
                    "abstract": "The rise of graph-structured data such as social networks, regulatory networks, citation graphs, and functional brain networks, in combination with resounding success of deep learning in various applications, has brought the interest in generalizing deep learning models to non-Euclidean domains. In this paper, we introduce a new spectral domain convolutional architecture for deep learning on graphs. The core ingredient of our model is a new class of parametric rational complex functions (Cayley polynomials) allowing to efficiently compute spectral filters on graphs that specialize on frequency bands of interest. Our model generates rich spectral filters that are localized in space, scales linearly with the size of the input data for sparsely connected graphs, and can handle different constructions of Laplacian operators. Extensive experimental results show the superior performance of our approach, in comparison to other spectral domain convolutional architectures, on spectral image classification, community detection, vertex classification, and matrix completion tasks."
                },
                {
                    "title": "Neural Message Passing for Quantum Chemistry",
                    "abstract": "Supervised learning on molecules has incredible potential to be useful in chemistry, drug discovery, and materials science. Luckily, several promising and closely related neural network models invariant to molecular symmetries have already been described in the literature. These models learn a message passing algorithm and aggregation procedure to compute a function of their entire input graph. At this point, the next step is to find a particularly effective variant of this general approach and apply it to chemical prediction benchmarks until we either solve them or reach the limits of the approach. In this paper, we reformulate existing models into a single common framework we call Message Passing Neural Networks (MPNNs) and explore additional novel variations within this framework. Using MPNNs we demonstrate state of the art results on an important molecular property prediction benchmark; these results are strong enough that we believe future work should focus on datasets with larger molecules or more accurate ground truth labels."
                },
                {
                    "title": "Geometric Deep Learning: Going beyond Euclidean data",
                    "abstract": "Many scientific fields study data with an underlying structure that is non-Euclidean. Some examples include social networks in computational social sciences, sensor networks in communications, functional networks in brain imaging, regulatory networks in genetics, and meshed surfaces in computer graphics. In many applications, such geometric data are large and complex (in the case of social networks, on the scale of billions) and are natural targets for machine-learning techniques. In particular, we would like to use deep neural networks, which have recently proven to be powerful tools for a broad range of problems from computer vision, natural-language processing, and audio analysis. However, these tools have been most successful on data with an underlying Euclidean or grid-like structure and in cases where the invariances of these structures are built into networks used to model them."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering",
                    "abstract": "In this work, we are interested in generalizing convolutional neural networks (CNNs) from low-dimensional regular grids, where image, video and speech are represented, to high-dimensional irregular domains, such as social networks, brain connectomes or words' embedding, represented by graphs. We present a formulation of CNNs in the context of spectral graph theory, which provides the necessary mathematical background and efficient numerical schemes to design fast localized convolutional filters on graphs. Importantly, the proposed technique offers the same linear computational complexity and constant learning complexity as classical CNNs, while being universal to any graph structure. Experiments on MNIST and 20NEWS demonstrate the ability of this novel deep learning system to learn local, stationary, and compositional features on graphs."
                },
                {
                    "title": "Revisiting Semi-Supervised Learning with Graph Embeddings",
                    "abstract": "We present a semi-supervised learning framework based on graph embeddings. Given a graph between instances, we train an embedding for each instance to jointly predict the class label and the neighborhood context in the graph. We develop both transductive and inductive variants of our method. In the transductive variant of our method, the class labels are determined by both the learned embeddings and input feature vectors, while in the inductive variant, the embeddings are defined as a parametric function of the feature vectors, so predictions can be made on instances not seen during training. On a large and diverse set of benchmark tasks, including text classification, distantly supervised entity extraction, and entity classification, we show improved performance over many of the existing models."
                },
                {
                    "title": "Group Equivariant Convolutional Networks",
                    "abstract": "We introduce Group equivariant Convolutional Neural Networks (G-CNNs), a natural generalization of convolutional neural networks that reduces sample complexity by exploiting symmetries. G-CNNs use G-convolutions, a new type of layer that enjoys a substantially higher degree of weight sharing than regular convolution layers. G-convolutions increase the expressive capacity of the network without increasing the number of parameters. Group convolution layers are easy to use and can be implemented with negligible computational overhead for discrete groups generated by translations, reflections and rotations. G-CNNs achieve state of the art results on CI- FAR10 and rotated MNIST."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Spectral Networks and Locally Connected Networks on Graphs",
                    "abstract": "Convolutional Neural Networks are extremely efficient architectures in image and audio recognition tasks, thanks to their ability to exploit the local translational invariance of signal classes over their domain. In this paper we consider possible generalizations of CNNs to signals defined on more general domains without the action of a translation group. In particular, we propose two constructions, one based upon a hierarchical clustering of the domain, and another based on the spectrum of the graph Laplacian. We show through experiments that for low-dimensional graphs it is possible to learn convolutional layers with a number of parameters independent of the input size, resulting in efficient deep architectures."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "Subgraph Matching Kernels for Attributed Graphs",
                    "abstract": "We propose graph kernels based on subgraph matchings, i.e. structure-preserving bijections between subgraphs. While recently proposed kernels based on common subgraphs (Wale et al., 2008; Shervashidze et al., 2009) in general can not be applied to attributed graphs, our approach allows to rate mappings of subgraphs by a exible scoring scheme comparing vertex and edge attributes by kernels. We show that subgraph matching kernels generalize several known kernels. To compute the kernel we propose a graph-theoretical algorithm inspired by a classical relation between common subgraphs of two graphs and cliques in their product graph observed by Levi (1973). Encouraging experimental results on a classification task of real-world graphs are presented."
                },
                {
                    "title": "Numerical Methods for Large Eigenvalue Problems",
                    "abstract": "Preface to the Classics Edition Preface 1. Background in matrix theory and linear algebra 2. Sparse matrices 3. Perturbation theory and error analysis 4. The tools of spectral approximation 5. Subspace iteration 6. Krylov subspace methods 7. Filtering and restarting techniques 8. Preconditioning techniques 9. Non-standard eigenvalue problems 10. Origins of matrix eigenvalue problems References Index."
                },
                {
                    "title": "Weisfeiler-Lehman Graph Kernels",
                    "abstract": "In this article, we propose a family of efficient kernels for large graphs with discrete node labels. Key to our method is a rapid feature extraction scheme based on the Weisfeiler-Lehman test of isomorphism on graphs. It maps the original graph to a sequence of graphs, whose node attributes capture topological and label information. A family of kernels can be defined based on this Weisfeiler-Lehman sequence of graphs, including a highly efficient kernel comparing subtree-like patterns. Its runtime scales only linearly in the number of edges of the graphs and the length of the Weisfeiler-Lehman graph sequence. In our experimental evaluation, our kernels outperform state-of-the-art graph kernels on several graph classification benchmark data sets in terms of accuracy and runtime. Our kernels open the door to large-scale applications of graph kernels in various disciplines such as computational biology and social network analysis."
                },
                {
                    "title": "Convolutional networks and applications in vision",
                    "abstract": "Intelligent tasks, such as visual perception, auditory perception, and language understanding require the construction of good internal representations of the world (or \"features\")? which must be invariant to irrelevant variations of the input while, preserving relevant information. A major question for Machine Learning is how to learn such good features automatically. Convolutional Networks (ConvNets) are a biologically-inspired trainable architecture that can learn invariant features. Each stage in a ConvNets is composed of a filter bank, some nonlinearities, and feature pooling layers. With multiple stages, a ConvNet can learn multi-level hierarchies of features. While ConvNets have been successfully deployed in many commercial applications from OCR to video surveillance, they require large amounts of labeled training samples. We describe new unsupervised learning algorithms, and new non-linear stages that allow ConvNets to be trained with very few labeled samples. Applications to visual object recognition and vision navigation for off-road mobile robots are described."
                },
                {
                    "title": "Social influence analysis in large-scale networks",
                    "abstract": "In large social networks, nodes (users, entities) are influenced by others for various reasons. For example, the colleagues have strong influence on one's work, while the friends have strong influence on one's daily life. How to differentiate the social influences from different angles(topics)? How to quantify the strength of those social influences? How to estimate the model on real large networks?\n To address these fundamental questions, we propose Topical Affinity Propagation (TAP) to model the topic-level social influence on large networks. In particular, TAP can take results of any topic modeling and the existing network structure to perform topic-level influence propagation. With the help of the influence analysis, we present several important applications on real data sets such as 1) what are the representative nodes on a given topic? 2) how to identify the social influences of neighboring nodes on a particular node?\n To scale to real large networks, TAP is designed with efficient distributed learning algorithms that is implemented and tested under the Map-Reduce framework. We further present the common characteristics of distributed learning algorithms for Map-Reduce. Finally, we demonstrate the effectiveness and efficiency of TAP on real large data sets."
                },
                {
                    "title": "Efficient graphlet kernels for large graph comparison",
                    "abstract": "State-of-the-art graph kernels do not scale to large graphs with hundreds of nodes and thousands of edges. In this article we propose to compare graphs by counting graphlets, i.e., subgraphs with k nodes where k \u2208 {3, 4, 5}. Exhaustive enumeration of all graphlets being prohibitively expensive, we introduce two theoretically grounded speedup schemes, one based on sampling and the second one specifically designed for bounded degree graphs. In our experimental evaluation, our novel kernels allow us to efficiently compare large graphs that cannot be tackled by existing graph kernels."
                },
                {
                    "title": "Network Analysis in the Social Sciences",
                    "abstract": "Over the past decade, there has been an explosion of interest in network research across the physical and social sciences. For social scientists, the theory of networks has been a gold mine, yielding explanations for social phenomena in a wide variety of disciplines from psychology to economics. Here, we review the kinds of things that social scientists have tried to explain using social network analysis and provide a nutshell description of the basic assumptions, goals, and explanatory mechanisms prevalent in the field. We hope to contribute to a dialogue among researchers from across the physical and social sciences who share a common interest in understanding the antecedents and consequences of network phenomena."
                },
                {
                    "title": "Collective Classification in Network Data",
                    "abstract": "Many real-world applications produce networked data such as the world-wide web (hypertext documents connected via hyperlinks), social networks (for example, people connected by friendship links), communication networks (computers connected via communication links) and biological networks (for example, protein interaction networks). A recent focus in machine learning research has been to extend traditional machine learning classification techniques to classify nodes in such networks. In this article, we provide a brief introduction to this area of research and how it has progressed during the past decade. We introduce four of the most widely used inference algorithms for classifying networked data and empirically compare them on both synthetic and real-world data."
                },
                {
                    "title": "A Theory for Multiresolution Signal Decomposition: The Wavelet Representation",
                    "abstract": "Multiresolution representations are effective for analyzing the information content of images. The properties of the operator which approximates a signal at a given resolution were studied. It is shown that the difference of information between the approximation of a signal at the resolutions 2/sup j+1/ and 2/sup j/ (where j is an integer) can be extracted by decomposing this signal on a wavelet orthonormal basis of L/sup 2/(R/sup n/), the vector space of measurable, square-integrable n-dimensional functions. In L/sup 2/(R), a wavelet orthonormal basis is a family of functions which is built by dilating and translating a unique function psi (x). This decomposition defines an orthogonal multiresolution representation called a wavelet representation. It is computed with a pyramidal algorithm based on convolutions with quadrature mirror filters. Wavelet representation lies between the spatial and Fourier domains. For images, the wavelet representation differentiates several spatial orientations. The application of this representation to data compression in image coding, texture discrimination and fractal analysis is discussed. >"
                },
                {
                    "title": "The Fourier Transform and Its Applications",
                    "abstract": "1 Introduction 2 Groundwork 3 Convolution 4 Notation for Some Useful Functions 5 The Impulse Symbol 6 The Basic Theorems 7 Obtaining Transforms 8 The Two Domains 9 Waveforms, Spectra, Filters and Linearity 10 Sampling and Series 11 The Discrete Fourier Transform and the FFT 12 The Discrete Hartley Transform 13 Relatives of the Fourier Transform 14 The Laplace Transform 15 Antennas and Optics 16 Applications in Statistics 17 Random Waveforms and Noise 18 Heat Conduction and Diffusion 19 Dynamic Power Spectra 20 Tables of sinc x, sinc2x, and exp(-71x2) 21 Solutions to Selected Problems 22 Pictorial Dictionary of Fourier Transforms 23 The Life of Joseph Fourier"
                },
                {
                    "title": "On the Sample Complexity of Learning with Geometric Stability",
                    "abstract": "Many supervised learning problems involve high-dimensional data such as images, text, or graphs. In order to make e\ufb03cient use of data, it is often useful to leverage certain geometric priors in the problem at hand, such as invariance to translations, permutation subgroups, or stability to small deformations. We study the sample complexity of learning problems where the target function presents such invariance and stability properties, by considering spherical harmonic decompositions of such functions on the sphere. We provide non-parametric rates of convergence for kernel methods, and show improvements in sample complexity by a factor equal to the size of the group when using an invariant kernel over the group, compared to the corresponding non-invariant kernel. These improvements are valid when the sample size is large enough, with an asymptotic behavior that depends on spectral properties of the group. Finally, these gains are extended beyond invariance groups to also cover geometric stability to small deformations, modeled here as subsets (not necessarily subgroups) of permutations."
                },
                {
                    "title": "A Comprehensive Survey on Graph Neural Networks",
                    "abstract": "Deep learning has revolutionized many machine learning tasks in recent years, ranging from image classification and video processing to speech recognition and natural language understanding. The data in these tasks are typically represented in the Euclidean space. However, there is an increasing number of applications, where data are generated from non-Euclidean domains and are represented as graphs with complex relationships and interdependency between objects. The complexity of graph data has imposed significant challenges on the existing machine learning algorithms. Recently, many studies on extending deep learning approaches for graph data have emerged. In this article, we provide a comprehensive overview of graph neural networks (GNNs) in data mining and machine learning fields. We propose a new taxonomy to divide the state-of-the-art GNNs into four categories, namely, recurrent GNNs, convolutional GNNs, graph autoencoders, and spatial\u2013temporal GNNs. We further discuss the applications of GNNs across various domains and summarize the open-source codes, benchmark data sets, and model evaluation of GNNs. Finally, we propose potential research directions in this rapidly growing field."
                },
                {
                    "title": "The Graph Neural Network Model",
                    "abstract": "Many underlying relationships among data in several areas of science and engineering, e.g., computer vision, molecular chemistry, molecular biology, pattern recognition, and data mining, can be represented in terms of graphs. In this paper, we propose a new neural network model, called graph neural network (GNN) model, that extends existing neural network methods for processing the data represented in graph domains. This GNN model, which can directly process most of the practically useful types of graphs, e.g., acyclic, cyclic, directed, and undirected, implements a function tau(G,n) isin IRm that maps a graph G and one of its nodes n into an m-dimensional Euclidean space. A supervised learning algorithm is derived to estimate the parameters of the proposed GNN model. The computational cost of the proposed algorithm is also considered. Some experimental results are shown to validate the proposed learning algorithm, and to demonstrate its generalization capabilities."
                },
                {
                    "title": "Protein function prediction via graph kernels",
                    "abstract": "MOTIVATION\nComputational approaches to protein function prediction infer protein function by finding proteins with similar sequence, structure, surface clefts, chemical properties, amino acid motifs, interaction partners or phylogenetic profiles. We present a new approach that combines sequential, structural and chemical information into one graph model of proteins. We predict functional class membership of enzymes and non-enzymes using graph kernels and support vector machine classification on these protein graphs.\n\n\nRESULTS\nOur graph model, derivable from protein sequence and structure only, is competitive with vector models that require additional protein information, such as the size of surface pockets. If we include this extra information into our graph model, our classifier yields significantly higher accuracy levels than the vector models. Hyperkernels allow us to select and to optimally combine the most relevant node attributes in our protein graphs. We have laid the foundation for a protein function prediction system that integrates protein information from various sources efficiently and effectively.\n\n\nAVAILABILITY\nMore information available via www.dbs.ifi.lmu.de/Mitarbeiter/borgwardt.html."
                },
                {
                    "title": "Large Networks and Graph Limits",
                    "abstract": "The book Large Networks and Graph Limits, xiv + 475 pp., published in late 2012, comprises five parts, the first an illuminating introduction and the last a tantalizing taste of how the scope of the theory developed in its pages might be extended to other combinatorial structures than graphs. The three central parts treat in depth the topics of graph algebras, limits for sequences of dense graphs (this constitutes the most substantial part, occupying nearly half the book) and limits for sequences of bounded degree graphs. Primarily the book is aimed at graduate students and research mathematicians interested in graph theory and its application to networks (for example, the internet and networks in social science, biology, statistical physics and engineering). There are 23 chapters and an appendix, the latter conveniently giving necessary background from areas of mathematics outside mainstream graph theory. A bibliography collects together the extensive research in this area up to 2012, and a subject, author and notation index facilitate navigation of the book. The author maintains a webpage for corrections and supplementary material. Indeed, via the author\u2019s homepage the reader can freely access the many papers he has written with collaborators on the topic of graph homomorphisms and graph limits. The book synthesizes much of the material in these papers, with some revision in"
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop non-linear spectral filters that are fully equivariant to graph functional shifts to enhance graph representation learning?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of graph machine learning (graph-ML) as it addresses the limitations of existing spectral graph neural networks (GNNs) that break symmetry through activation functions. By creating non-linear spectral filters that respect graph functional shifts, we can improve the generalization capabilities of models applied to graph-structured data across various domains, including chemistry, biology, and social sciences. This advancement could lead to more robust applications in drug discovery, social media analysis, and recommendation systems, ultimately influencing future research directions and methodologies in graph-ML.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of graph structures, which lack a natural notion of translation compared to images. Naive approaches that apply standard activation functions in spectral GNNs fail because they disrupt the equivariance to graph functional shifts, leading to suboptimal performance. Additionally, the need for transferability between different graphs complicates the design of spectral representations, as traditional graph Fourier transforms depend on arbitrary eigenvector choices, limiting their applicability. Overcoming these technical and theoretical obstacles requires innovative methods that maintain symmetry while providing universal approximation properties.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on linear spectral filters and standard graph Fourier transforms, which do not account for the unique properties of graph structures, particularly the lack of a natural translation. The reliance on specific Laplacian eigenvectors in existing methods has hindered the development of transferable spectral representations. Additionally, the complexity of designing non-linear filters that maintain equivariance to graph functional shifts has posed significant barriers. Our approach differs by introducing non-linear spectral filters that are independent of specific eigenvector choices, thus enabling transferability and enhancing the model's performance on diverse graph data.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of non-linear spectral filters (NLSFs) that are fully equivariant to graph functional shifts. We will utilize a dataset of graph-structured data from various domains, applying metrics such as accuracy and generalization performance to evaluate our model. The expected outcomes include improved performance in graph representation tasks, enhanced transferability of spectral representations across different graphs, and a"
            }
        },
        "author_data": {
            "3b565431-20df-488e-b28e-cbfceaa6ab33": {
                "pk": "3b565431-20df-488e-b28e-cbfceaa6ab33",
                "name": "Ya-Wei Eileen Lin",
                "collaborators": [
                    "Ronald R. Coifman",
                    "Gal Mishne",
                    "Ronen Talmon"
                ],
                "domain": [
                    "Hierarchical Data",
                    "Manifold Learning",
                    "Graph Embedding",
                    "Hyperbolic Geometry"
                ],
                "publications": [
                    {
                        "title": "Hyperbolic Diffusion Embedding and Distance for Hierarchical Representation Learning",
                        "abstract": "Finding meaningful representations and distances of hierarchical data is important in many fields. This paper presents a new method for hierarchical data embedding and distance. Our method relies on combining diffusion geometry, a central approach to manifold learning, and hyperbolic geometry. Specifically, using diffusion geometry, we build multi-scale densities on the data, aimed to reveal their hierarchical structure, and then embed them into a product of hyperbolic spaces. We show theoretically that our embedding and distance recover the underlying hierarchical structure. In addition, we demonstrate the efficacy of the proposed method and its advantages compared to existing methods on graph embedding benchmarks and hierarchical datasets."
                    }
                ]
            },
            "b3f793a8-eddb-4f97-8956-5e2b14813440": {
                "pk": "b3f793a8-eddb-4f97-8956-5e2b14813440",
                "name": "Ronen Talmon",
                "collaborators": [
                    "Or Yair",
                    "Tal Shnitzer",
                    "Amitay Bar",
                    "Almog Lahav",
                    "Jean-Jacques Slotine",
                    "Bracha Laufer-Goldshtein",
                    "Sharon Gannot",
                    "Vardan Papyan",
                    "Ariel Schwartz",
                    "Hau-tieng Wu"
                ],
                "domain": [
                    "Manifold Learning",
                    "Riemannian Geometry",
                    "Multimodal Data Analysis",
                    "Signal Processing"
                ],
                "publications": [
                    {
                        "title": "Local Canonical Correlation Analysis for Nonlinear Common Variables Discovery",
                        "abstract": "In this paper, we address the problem of hidden common variables discovery from multimodal data sets of nonlinear high-dimensional observations. We present a metric based on local applications of canonical correlation analysis (CCA) and incorporate it in a kernel-based manifold learning technique.We show that this metric discovers the hidden common variables underlying the multimodal observations by estimating the Euclidean distance between them. Our approach can be viewed both as an extension of CCA to a nonlinear setting as well as an extension of manifold learning to multiple data sets. Experimental results show that our method indeed discovers the common variables underlying high-dimensional nonlinear observations without assuming prior rigid model assumptions."
                    },
                    {
                        "title": "Multimodal Latent Variable Analysis",
                        "abstract": "Consider a set of multiple, multimodal sensors capturing a complex system or a physical phenomenon of interest. Our primary goal is to distinguish the underlying sources of variability manifested in the measured data. The first step in our analysis is to find the common source of variability present in all sensor measurements. We base our work on a recent paper, which tackles this problem with alternating diffusion (AD). In this work, we suggest to further the analysis by extracting the sensor-specific variables in addition to the common source. We propose an algorithm, which we analyze theoretically, and then demonstrate on three different applications: a synthetic example, a toy problem, and the task of fetal ECG extraction."
                    },
                    {
                        "title": "Intrinsic Isometric Manifold Learning with Application to Localization",
                        "abstract": "Data living on manifolds commonly appear in many applications. Often this results from an inherently latent low-dimensional system being observed through higher dimensional measurements. We show that under certain conditions, it is possible to construct an intrinsic and isometric data representation, which respects an underlying latent intrinsic geometry. Namely, we view the observed data only as a proxy and learn the structure of a latent unobserved intrinsic manifold, whereas common practice is to learn the manifold of the observed data. For this purpose, we build a new metric and propose a method for its robust estimation by assuming mild statistical priors and by using artificial neural networks as a mechanism for metric regularization and parametrization. We show successful application to unsupervised indoor localization in ad-hoc sensor networks. Specifically, we show that our proposed method facilitates accurate localization of a moving agent from imaging data it collects. Importantly, our method is applied in the same way to two different imaging modalities, thereby demonstrating its intrinsic and modality-invariant capabilities."
                    },
                    {
                        "title": "On Interference-Rejection Using Riemannian Geometry for Direction of Arrival Estimation",
                        "abstract": "We consider the problem of estimating the direction of arrival of desired acoustic sources in the presence of multiple acoustic interference sources. All the sources are located in noisy and reverberant environments and are received by a microphone array. We propose a new approach for designing beamformers and DoA estimation methods based on the Riemannian geometry of the manifold of Hermitian positive definite matrices. Specifically, we show theoretically that incorporating the Riemannian mean of the spatial correlation matrices into frequently-used beamformers gives rise to spatial spectra that reject the directions of interference sources and result in a higher signal-to-interference ratio. We experimentally demonstrate the advantages of our approach in designing several beamformers and a recent DoA estimation method in the presence of simultaneously active multiple interference sources."
                    },
                    {
                        "title": "Latent common manifold learning with alternating diffusion: analysis and applications",
                        "abstract": "The analysis of data sets arising from multiple sensors has drawn significant research attention over the years. Traditional methods, including kernel-based methods, are typically incapable of capturing nonlinear geometric structures. We introduce a latent common manifold model underlying multiple sensor observations for the purpose of multimodal data fusion. A method based on alternating diffusion is presented and analyzed; we provide theoretical analysis of the method under the latent common manifold model. To exemplify the power of the proposed framework, experimental results in several applications are reported."
                    },
                    {
                        "title": "Parallel Transport on the Cone Manifold of SPD Matrices for Domain Adaptation",
                        "abstract": "In this paper, we consider the problem of domain adaptation. We propose to view the data through the lens of covariance matrices and present a method for domain adaptation using parallel transport on the cone manifold of symmetric positive-definite matrices. We provide rigorous analysis using Riemanninan geometry, illuminating the theoretical guarantees and benefits of the presented method. In addition, we demonstrate these benefits using experimental results on simulations and real-measured data."
                    },
                    {
                        "title": "Identifying Topological Phase Transitions in Experiments Using Manifold Learning",
                        "abstract": "We demonstrate the identification and classification of topological phase transitions from experimental data using Diffusion Maps: a nonlocal unsupervised machine learning method. We analyze experimental data from an optical system undergoing a topological phase transition and demonstrate the ability of this approach to identify topological phase transitions even when the data originates from a small part of the system, and does not even include edge states."
                    },
                    {
                        "title": "Symmetric Positive Semi-definite Riemannian Geometry with Application to Domain Adaptation",
                        "abstract": "In this paper, we present new results on the Riemannian geometry of symmetric positive semi-definite (SPSD) matrices. First, based on an existing approximation of the geodesic path, we introduce approximations of the logarithmic and exponential maps. Second, we present a closed-form expression for Parallel Transport (PT). Third, we derive a canonical representation for a set of SPSD matrices. Based on these results, we propose an algorithm for Domain Adaptation (DA) and demonstrate its performance in two applications: fusion of hyper-spectral images and motion identification."
                    },
                    {
                        "title": "Kernel-based Sensor Fusion with Application to Audio-Visual Voice Activity Detection",
                        "abstract": "In this paper, we address the problem of multiple view data fusion in the presence of noise and interferences. Recent studies have approached this problem using kernel methods, by relying particularly on a product of kernels constructed separately for each view. From a graph theory point of view, we analyze this fusion approach in a discrete setting. More specifically, based on a statistical model for the connectivity between data points, we propose an algorithm for the selection of the kernel bandwidth, a parameter, which, as we show, has important implications on the robustness of this fusion approach to interferences. Then, we consider the fusion of audio-visual speech signals measured by a single microphone and by a video camera pointed to the face of the speaker. Specifically, we address the task of voice activity detection, i.e., the detection of speech and non-speech segments, in the presence of structured interferences such as keyboard taps and office noise. We propose an algorithm for voice activity detection based on the audio-visual signal. Simulation results show that the proposed algorithm outperforms competing fusion and voice activity detection approaches. In addition, we demonstrate that a proper selection of the kernel bandwidth indeed leads to improved performance."
                    },
                    {
                        "title": "Manifold Learning with Contracting Observers for Data-driven Time-series Analysis",
                        "abstract": "Analyzing signals arising from dynamical systems typically requires many modeling assumptions and parameter estimation. In high dimensions, this modeling is particularly difficult due to the \"curse of dimensionality\". In this paper, we propose a method for building an intrinsic representation of such signals in a purely data-driven manner. First, we apply a manifold learning technique, diffusion maps, to learn the intrinsic model of the latent variables of the dynamical system, solely from the measurements. Second, we use concepts and tools from control theory and build a linear contracting observer to estimate the latent variables in a sequential manner from new incoming measurements. The effectiveness of the presented framework is demonstrated by applying it to a toy problem and to a music analysis application. In these examples we show that our method reveals the intrinsic variables of the analyzed dynamical systems."
                    },
                    {
                        "title": "Semi-Supervised Source Localization on Multiple-Manifolds with Distributed Microphones",
                        "abstract": "The problem of source localization with ad hoc microphone networks in noisy and reverberant enclosures, given a training set of prerecorded measurements, is addressed in this paper. The training set is assumed to consist of a limited number of labelled measurements, attached with corresponding positions, and a larger amount of unlabelled measurements from unknown locations. However, microphone calibration is not required. We use a Bayesian inference approach for estimating a function that maps measurement-based feature vectors to the corresponding positions. The central issue is how to combine the information provided by the different microphones in a unified statistical framework. To address this challenge, we model this function using a Gaussian process with a covariance function that encapsulates both the connections between pairs of microphones and the relations among the samples in the training set. The parameters of the process are estimated by optimizing a maximum likelihood (ML) criterion. In addition, a recursive adaptation mechanism is derived where the new streaming measurements are used to update the model. Performance is demonstrated for 2-D localization of both simulated data and real-life recordings in a variety of reverberation and noise levels."
                    },
                    {
                        "title": "Mahalanonbis Distance Informed by Clustering",
                        "abstract": "A fundamental question in data analysis, machine learning and signal processing is how to compare between data points. The choice of the distance metric is specifically challenging for high-dimensional data sets, where the problem of meaningfulness is more prominent (e.g. the Euclidean distance between images). In this paper, we propose to exploit a property of high-dimensional data that is usually ignored - which is the structure stemming from the relationships between the coordinates. Specifically we show that organizing similar coordinates in clusters can be exploited for the construction of the Mahalanobis distance between samples. When the observable samples are generated by a nonlinear transformation of hidden variables, the Mahalanobis distance allows the recovery of the Euclidean distances in the hidden space.We illustrate the advantage of our approach on a synthetic example where the discovery of clusters of correlated coordinates improves the estimation of the principal directions of the samples. Our method was applied to real data of gene expression for lung adenocarcinomas (lung cancer). By using the proposed metric we found a partition of subjects to risk groups with a good separation between their Kaplan-Meier survival plot."
                    },
                    {
                        "title": "DIMAL: Deep Isometric Manifold Learning Using Sparse Geodesic Sampling",
                        "abstract": "This paper explores a fully unsupervised deep learning approach for computing distance-preserving maps that generate low-dimensional embeddings for a certain class of manifolds. We use the Siamese configuration to train a neural network to solve the problem of least squares multidimensional scaling for generating maps that approximately preserve geodesic distances. By training with only a few landmarks, we show a significantly improved local and nonlocal generalization of the isometric mapping as compared to analogous non-parametric counterparts. Importantly, the combination of a deep-learning framework with a multidimensional scaling objective enables a numerical analysis of network architectures to aid in understanding their representation power. This provides a geometric perspective to the generalizability of deep learning."
                    },
                    {
                        "title": "Diffusion Maps Kalman Filter for a Class of Systems with Gradient Flows",
                        "abstract": "In this paper, we propose a non-parametric method for state estimation of high-dimensional nonlinear stochastic dynamical systems, which evolve according to gradient flows with isotropic diffusion. We combine diffusion maps, a manifold learning technique, with a linear Kalman filter and with concepts from Koopman operator theory. More concretely, using diffusion maps, we construct data-driven virtual state coordinates, which linearize the system model. Based on these coordinates, we devise a data-driven framework for state estimation using the Kalman filter. We demonstrate the strengths of our method with respect to both parametric and non-parametric algorithms in three tracking problems. In particular, applying the approach to actual recordings of hippocampal neural activity in rodents directly yields a representation of the position of the animals. We show that the proposed method outperforms competing non-parametric algorithms in the examined stochastic problem formulations. Additionally, we obtain results comparable to classical parametric algorithms, which, in contrast to our method, are equipped with model knowledge."
                    },
                    {
                        "title": "Data-Driven Source Separation Based on Simplex Analysis",
                        "abstract": "Blind source separation (BSS) is addressed, using a novel data-driven approach, based on a well-established probabilistic model. The proposed method is specifically designed for separation of multichannel audio mixtures. The algorithm relies on spectral decomposition of the correlation matrix between different time frames. The probabilistic model implies that the column space of the correlation matrix is spanned by the probabilities of the various speakers across time. The number of speakers is recovered by the eigenvalue decay, and the eigenvectors form a simplex of the speakers' probabilities. Time frames dominated by each of the speakers are identified exploiting convex geometry tools on the recovered simplex. The mixing acoustic channels are estimated utilizing the identified sets of frames, and a linear umixing is performed to extract the individual speakers. The derived simplexes are visually demonstrated for mixtures of 2, 3 and 4 speakers. We also conduct a comprehensive experimental study, showing high separation capabilities in various reverberation conditions."
                    },
                    {
                        "title": "Option Discovery in the Absence of Rewards with Manifold Analysis",
                        "abstract": "Options have been shown to be an effective tool in reinforcement learning, facilitating improved exploration and learning. In this paper, we present an approach based on spectral graph theory and derive an algorithm that systematically discovers options without access to a specific reward or task assignment. As opposed to the common practice used in previous methods, our algorithm makes full use of the spectrum of the graph Laplacian. Incorporating modes associated with higher graph frequencies unravels domain subtleties, which are shown to be useful for option discovery. Using geometric and manifold-based analysis, we present a theoretical justification for the algorithm. In addition, we showcase its performance in several domains, demonstrating clear improvements compared to competing methods."
                    },
                    {
                        "title": "Joint Geometric and Topological Analysis of Hierarchical Datasets",
                        "abstract": "In a world abundant with diverse data arising from complex acquisition techniques, there is a growing need for new data analysis methods. In this paper we focus on high-dimensional data that are organized into several hierarchical datasets. We assume that each dataset consists of complex samples, and every sample has a distinct irregular structure modeled by a graph. The main novelty in this work lies in the combination of two complementing powerful data-analytic approaches: topological data analysis (TDA) and geometric manifold learning. Geometry primarily contains local information, while topology inherently provides global descriptors. Based on this combination, we present a method for building an informative representation of hierarchical datasets. At the finer (sample) level, we devise a new metric between samples based on manifold learning that facilitates quantitative structural analysis. At the coarser (dataset) level, we employ TDA to extract qualitative structural information from the datasets. We showcase the applicability and advantages of our method on simulated data and on a corpus of hyper-spectral images. We show that an ensemble of hyper-spectral images exhibits a hierarchical structure that fits well the considered setting. In addition, we show that our new method gives rise to superior classification results compared to state-of-the-art methods."
                    },
                    {
                        "title": "Kernel-based parameter estimation of dynamical systems with unknown observation functions",
                        "abstract": "A low-dimensional dynamical system is observed in an experiment as a high-dimensional signal; for example, a video of a chaotic pendulums system. Assuming that we know the dynamical model up to some unknown parameters, can we estimate the underlying system's parameters by measuring its time-evolution only once? The key information for performing this estimation lies in the temporal inter-dependencies between the signal and the model. We propose a kernel-based score to compare these dependencies. Our score generalizes a maximum likelihood estimator for a linear model to a general nonlinear setting in an unknown feature space. We estimate the system's underlying parameters by maximizing the proposed score. We demonstrate the accuracy and efficiency of the method using two chaotic dynamical systems - the double pendulum and the Lorenz '63 model."
                    },
                    {
                        "title": "Spectral Flow on the Manifold of SPD Matrices for Multimodal Data Processing",
                        "abstract": "In this paper, we consider data acquired by multimodal sensors capturing complementary aspects and features of a measured phenomenon. We focus on a scenario in which the measurements share mutual sources of variability but might also be contaminated by other measurement-specific sources such as interferences or noise. Our approach combines manifold learning, which is a class of nonlinear data-driven dimension reduction methods, with the well-known Riemannian geometry of symmetric and positive-definite (SPD) matrices. Manifold learning typically includes the spectral analysis of a kernel built from the measurements. Here, we take a different approach, utilizing the Riemannian geometry of the kernels. In particular, we study the way the spectrum of the kernels changes along geodesic paths on the manifold of SPD matrices. We show that this change enables us, in a purely unsupervised manner, to derive a compact, yet informative, description of the relations between the measurements, in terms of their underlying components. Based on this result, we present new algorithms for extracting the common latent components and for identifying common and measurement-specific components."
                    },
                    {
                        "title": "Spatiotemporal Analysis Using Riemannian Composition of Diffusion Operators",
                        "abstract": "Multivariate time-series have become abundant in recent years, as many data-acquisition systems record information through multiple sensors simultaneously. In this paper, we assume the variables pertain to some geometry and present an operator-based approach for spatiotemporal analysis. Our approach combines three components that are often considered separately: (i) manifold learning for building operators representing the geometry of the variables, (ii) Riemannian geometry of symmetric positive-definite matrices for multiscale composition of operators corresponding to different time samples, and (iii) spectral analysis of the composite operators for extracting different dynamic modes. We propose a method that is analogous to the classical wavelet analysis, which we term Riemannian multi-resolution analysis (RMRA). We provide some theoretical results on the spectral analysis of the composite operators, and we demonstrate the proposed method on simulations and on real data."
                    }
                ]
            },
            "2381513e-c560-4039-a39e-3a31a1f392fd": {
                "pk": "2381513e-c560-4039-a39e-3a31a1f392fd",
                "name": "Ron Levie",
                "collaborators": [
                    "Nir Sochen",
                    "Haim Avron",
                    "Gitta Kutyniok",
                    "Daniel Zilberg",
                    "Efrat Krimer Avraham",
                    "Elvin Isufi"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Signal Processing",
                    "Wavelet Transform",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "A graphon-signal analysis of graph neural networks",
                        "abstract": "We present an approach for analyzing message passing graph neural networks (MPNNs) based on an extension of graphon analysis to a so called graphon-signal analysis. A MPNN is a function that takes a graph and a signal on the graph (a graph-signal) and returns some value. Since the input space of MPNNs is non-Euclidean, i.e., graphs can be of any size and topology, properties such as generalization are less well understood for MPNNs than for Euclidean neural networks. We claim that one important missing ingredient in past work is a meaningful notion of graph-signal similarity measure, that endows the space of inputs to MPNNs with a regular structure. We present such a similarity measure, called the graphon-signal cut distance, which makes the space of all graph-signals a dense subset of a compact metric space -- the graphon-signal space. Informally, two deterministic graph-signals are close in cut distance if they ``look like'' they were sampled from the same random graph-signal model. Hence, our cut distance is a natural notion of graph-signal similarity, which allows comparing any pair of graph-signals of any size and topology. We prove that MPNNs are Lipschitz continuous functions over the graphon-signal metric space. We then give two applications of this result: 1) a generalization bound for MPNNs, and, 2) the stability of MPNNs to subsampling of graph-signals. Our results apply to any regular enough MPNN on any distribution of graph-signals, making the analysis rather universal."
                    },
                    {
                        "title": "Uncertainty principles and optimally sparse wavelet transforms",
                        "abstract": "In this paper we introduce a new localization framework for wavelet transforms, such as the 1D wavelet transform and the Shearlet transform. Our goal is to design nonadaptive window functions that promote sparsity in some sense. For that, we introduce a framework for analyzing localization aspects of window functions. Our localization theory diverges from the conventional theory in two ways. First, we distinguish between the group generators, and the operators that measure localization (called observables). Second, we define the uncertainty of a signal transform based on a window as a whole, instead of defining the uncertainty of an individual window. We show that the uncertainty of a window function, in the signal space, is closely related to the localization of the reproducing kernel of the wavelet transform, in phase space. As a result, we show that using uncertainty minimizing window functions, results in representations which are optimally sparse in some sense."
                    },
                    {
                        "title": "Randomized Signal Processing with Continuous Frames",
                        "abstract": "This paper focuses on signal processing tasks in which the signal is transformed from the signal space to a higher dimensional coefficient space (also called phase space) using a continuous frame, processed in the coefficient space, and synthesized to an output signal. We show how to approximate such methods, termed phase space signal processing methods, using a Monte Carlo method. As opposed to standard discretizations of continuous frames, based on sampling discrete frames from the continuous system, the proposed Monte Carlo method is directly a quadrature approximation of the continuous frame. We show that the Monte Carlo method allows working with highly redundant continuous frames, since the number of samples required for a certain accuracy is proportional to the dimension of the signal space, and not to the dimension of the phase space. Moreover, even though the continuous frame is highly redundant, the Monte Carlo samples are spread uniformly, and hence represent the coefficient space more faithfully than standard frame discretizations."
                    },
                    {
                        "title": "Randomized Continuous Frames in Time-Frequency Analysis",
                        "abstract": "Recently, a Monte Carlo approach was proposed for processing highly redundant continuous frames. In this paper we present and analyze applications of this new theory. The computational complexity of the Monte Carlo method relies on the continuous frame being so called linear volume discretizable (LVD). The LVD property means that the number of samples in the coefficient space required by the Monte Carlo method is proportional to the resolution of the discrete signal. We show in this paper that the continuous wavelet transform (CWT) and the localizing time-frequency transform (LTFT) are LVD. The LTFT is a time-frequency representation based on a 3D time-frequency space with a richer class of time-frequency atoms than classical time-frequency transforms like the short time Fourier transform (STFT) and the CWT. Our analysis proves that performing signal processing with the LTFT has the same asymptotic complexity as signal processing with the STFT and CWT (based on FFT), even though the coefficient space of the LTFT is higher dimensional."
                    },
                    {
                        "title": "A Wavelet Plancherel Theory with Application to Multipliers and Sparse Approximations",
                        "abstract": "We introduce an extension of continuous wavelet theory that enables an efficient implementation of multiplicative operators in the coefficient space. In the new theory, the signal space is embedded in a larger abstract signal space -- the so called window-signal space. There is a canonical extension of the wavelet transform to an isometric isomorphism between the window-signal space and the coefficient space. Hence, the new framework is called a wavelet-Plancherel theory, and the extended wavelet transform is called the wavelet-Plancherel transform. Since the wavelet-Plancherel transform is an isometric isomorphism, any operation in the coefficient space can be pulled-back to an operation in the window-signal space. It is then possible to improve the computational complexity of methods that involve a multiplicative operator in the coefficient space, by performing all computations directly in the window-signal space. As one example application, we show how continuous wavelet multipliers (also called Calder\\'{o}n-Toeplitz Operators), with polynomial symbols, can be implemented with linear complexity in the resolution of the 1D signal. As another example, we develop a framework for efficiently computing greedy sparse approximations to signals based on elements of continuous wavelet systems."
                    },
                    {
                        "title": "PieClam: A Universal Graph Autoencoder Based on Overlapping Inclusive and Exclusive Communities",
                        "abstract": "We propose PieClam (Prior Inclusive Exclusive Cluster Affiliation Model): a probabilistic graph model for representing any graph as overlapping generalized communities. Our method can be interpreted as a graph autoencoder: nodes are embedded into a code space by an algorithm that maximizes the log-likelihood of the decoded graph, given the input graph. PieClam is a community affiliation model that extends well-known methods like BigClam in two main manners. First, instead of the decoder being defined via pairwise interactions between the nodes in the code space, we also incorporate a learned prior on the distribution of nodes in the code space, turning our method into a graph generative model. Secondly, we generalize the notion of communities by allowing not only sets of nodes with strong connectivity, which we call inclusive communities, but also sets of nodes with strong disconnection, which we call exclusive communities. To model both types of communities, we propose a new type of decoder based the Lorentz inner product, which we prove to be much more expressive than standard decoders based on standard inner products or norm distances. By introducing a new graph similarity measure, that we call the log cut distance, we show that PieClam is a universal autoencoder, able to uniformly approximately reconstruct any graph. Our method is shown to obtain competitive performance in graph anomaly detection benchmarks."
                    },
                    {
                        "title": "Wavelet Design with Optimally Localized Ambiguity Function: a Variational Approach",
                        "abstract": "In this paper, we design mother wavelets for the 1D continuous wavelet transform with some optimality properties. An optimal mother wavelet here is one that has an ambiguity function with minimal spread in the continuous coefficient space (also called phase space). Since the ambiguity function is the reproducing kernel of the coefficient space, optimal windows lead to phase space representations which are \"optimally sharp.\" Namely, the wavelet coefficients have minimal correlations with each other. Such a construction also promotes sparsity in phase space. The spread of the ambiguity function is modeled as the sum of variances along the axes in phase space. In order to optimize the mother wavelet directly as a 1D signal, we pull-back the variances, defined on the 2D phase space, to the so called window-signal space. This is done using the recently developed wavelet-Plancharel theory. The approach allows formulating the optimization problem of the 2D ambiguity function as a minimization problem of the 1D mother wavelet. The resulting 1D formulation is more efficient and does not involve complicated constraints on the 2D ambiguity function. We optimize the mother wavelet using gradient descent, which yields a locally optimal mother wavelet."
                    },
                    {
                        "title": "On the Transferability of Spectral Graph Filters",
                        "abstract": "This paper focuses on spectral filters on graphs, namely filters defined as elementwise multiplication in the frequency domain of a graph. In many graph signal processing settings, it is important to transfer a filter from one graph to another. One example is in graph convolutional neural networks (ConvNets), where the dataset consists of signals defined on many different graphs, and the learned filters should generalize to signals on new graphs, not present in the training set. A necessary condition for transferability (the ability to transfer filters) is stability. Namely, given a graph filter, if we add a small perturbation to the graph, then the filter on the perturbed graph is a small perturbation of the original filter. It is a common misconception that spectral filters are not stable, and this paper aims at debunking this mistake. We introduce a space of filters, called the Cayley smoothness space, that contains the filters of state-of-the-art spectral filtering methods, and whose filters can approximate any generic spectral filter. For filters in this space, the perturbation in the filter is bounded by a constant times the perturbation in the graph, and filters in the Cayley smoothness space are thus termed linearly stable. By combining stability with the known property of equivariance, we prove that graph spectral filters are transferable."
                    },
                    {
                        "title": "Quasi Monte Carlo Time-Frequency Analysis",
                        "abstract": "We study signal processing tasks in which the signal is mapped via some generalized time-frequency transform to a higher dimensional time-frequency space, processed there, and synthesized to an output signal. We show how to approximate such methods using a quasi-Monte Carlo (QMC) approach. We consider cases where the time-frequency representation is redundant, having feature axes in addition to the time and frequency axes. The proposed QMC method allows sampling both efficiently and evenly such redundant time-frequency representations. Indeed, 1) the number of samples required for a certain accuracy is log-linear in the resolution of the signal space, and depends only weakly on the dimension of the redundant time-frequency space, and 2) the quasi-random samples have low discrepancy, so they are spread evenly in the redundant time-frequency space. One example of such redundant representation is the localizing time-frequency transform (LTFT), where the time-frequency plane is enhanced by a third axis. This higher dimensional time-frequency space improves the quality of some time-frequency signal processing tasks, like the phase vocoder (an audio signal processing effect). Since the computational complexity of the QMC is log-linear in the resolution of the signal space, this higher dimensional time-frequency space does not degrade the computation complexity of the proposed QMC method. The proposed QMC method is more efficient than standard Monte Carlo methods, since the deterministic QMC sample points are optimally spread in the time-frequency space, while random samples are not."
                    }
                ]
            }
        }
    },
    "2310.04373": {
        "paper_data": {
            "title": "Confronting Reward Model Overoptimization with Constrained RLHF",
            "url": "http://arxiv.org/abs/2310.04373v2",
            "arxiv_id": "2310.04373",
            "authors": [
                "Ted Moskovitz",
                "Aaditya K. Singh",
                "DJ Strouse",
                "Tuomas Sandholm",
                "Ruslan Salakhutdinov",
                "Anca D. Dragan",
                "Stephen McAleer"
            ],
            "abstract": "Large language models are typically aligned with human preferences by optimizing $\\textit{reward models}$ (RMs) fitted to human feedback. However, human preferences are multi-faceted, and it is increasingly common to derive reward from a composition of simpler reward models which each capture a different aspect of language quality. This itself presents a challenge, as it is difficult to appropriately weight these component RMs when combining them. Compounding this difficulty, because any RM is only a proxy for human evaluation, this process is vulnerable to $\\textit{overoptimization}$, wherein past a certain point, accumulating higher reward is associated with worse human ratings. In this paper, we perform, to our knowledge, the first study on overoptimization in composite RMs, showing that correlation between component RMs has a significant effect on the locations of these points. We then introduce an approach to solve this issue using constrained reinforcement learning as a means of preventing the agent from exceeding each RM's threshold of usefulness. Our method addresses the problem of weighting component RMs by learning dynamic weights, naturally expressed by Lagrange multipliers. As a result, each RM stays within the range at which it is an effective proxy, improving evaluation performance. Finally, we introduce an adaptive method using gradient-free optimization to identify and optimize towards these points during a single run.",
            "introduction": " introduction of non-stationarity into the reward function. RL with non-stationary rewards is an active area of interest in RL (Padakandla et al., 2020; Cheung et al., 2020; Lecarpentier and Rachelson, 2019), particularly in the context of continual RL (Khetarpal et al., 2022) often with some form of temporal structure introduced in the non-stationarity (Xie et al., 2020; 2021). An interesting case in additional to primal-dual optimization in which non-stationarity is introduced by the agent itself is in the use of epistemic uncertainty for more efficient exploration, manifested in the form of non-stationary exploration bonuses to reward 200 20000 40000 60000 80000 100000 1200000.1550.1600.1650.1700.175Scorerouge_rouge1 0 20000 40000 60000 80000 100000 1200000.00240.00260.00280.00300.00320.0034bleu 0 20000 40000 60000 80000 100000 1200000.0650.0700.0750.0800.0850.0900.0950.100sacrebleu 0 20000 40000 60000 80000 100000 1200000.1300.1350.1400.1450.150Scorerouge_rougeL 0 20000 40000 60000 80000 100000 1200000.01050.01100.01150.01200.01250.01300.0135rouge_rouge2 0 20000 40000 60000 80000 100000 1200000.1300.1350.1400.1450.150rouge_rougeLsum 0 20000 40000 60000 80000 100000 120000 Training Steps0.840.850.860.870.88Scorebert_score 0 20000 40000 60000 80000 100000 120000 Training Steps12.513.013.514.014.5perplexity 0 20000 40000 60000 80000 100000 120000 Training Steps1.801.851.901.952.00cond_entropy-3-nopunct PPO PPO-SAT -PPO  All-PPO -PPO Figure D.1: Performance of the tested related work is regularized policy optimization, whereby the standard reward-maximizing policy optimization objective is augmented with a regularization term, typically a divergence measure with respect to some reference policy (Berner et al., 2019; Espeholt et al., 2018). In the single-task setting, the updated policy is typically regularized to stay close to its current setting, which has close connections to natural gradient (Kakade and Langford, 2002; Moskovitz et al., 2021b; Pacchiano et al., 2020), trust region (Schulman et al., 2015), and variational inference (Levine, 2018; Haarnoja et al., 2018; Abdolmaleki et al., 2018) approaches. In the multitask setting, the policy is typically regularized towards some default policy which encodes behavior thought to be useful across a family of tasks, and which may be far from the current policy (Galashov et al., 2019; Teh et al., 2017; Moskovitz et al., 2022). This setting is quite similar in this sense to KL regularization as used in RLHF. D A DDITIONAL RESULTS D.1 A DDITIONAL METRICS 210 20000 40000 60000 80000 100000 120000-0.75-0.50-0.250.000.250.500.751.00tanh() Intent Lagrange Multiplier 0 20000 40000 60000 80000 100000 120000-0.6-0.4-0.20.00.20.40.60.81.0METEOR Lagrange Multiplier 0 20000 40000 60000 80000 100000 120000 Training Steps0.250.300.350.400.450.500.550.600.65ReturnIntent Return 0 20000 40000 60000 80000 100000 120000 Training Steps0.160.180.200.220.24METEOR ReturnFigure D.2: Intermediate thresholds can cause oscillations. Running gradient descent-ascent on a min-max game only guarantees that the average of the iterates converges to the saddle point. In practice, this can mean that the Lagrange multiplier(s) and value(s) can oscillate wildly over the course of training, even if their averages converge. The problem is worse for constraint thresholds which are intermediate\u2014those that are neither high nor low relative to the range of an individual reward function (Moskovitz et al., 2023a), but can be hidden by averaging. Above is an example run of All-PPO, showing this problem can occur. 0.42 0.44 0.46 0.48 0.50 0.52 Intent Threshold0.120.140.160.180.200.220.240.26METEOR ThresholdNelder-Mead Proxy Point Trajectory Initial Simplex Intermediate Simplex Final Simplex0.360.400.440.480.520.560.600.640.68 Predicted Evaluation Score Figure D.3: Another example Nelder-Mead simplex trajectory. In this case, Nelder-Mead con- verged to the global maximum proxy point setting. 220 20000 40000 60000 80000 100000 120000 Training Steps0.560.580.600.620.640.66Evaluation ScoreEvaluation 0.75int_0.25met 1.0int_0.25met 1.0int_0.5met 1.0int_1.0met 0.5int_0.5met 0.25int_0.75met 0.5int_0.75met 0.75int_0.75met 0.75int_0.5met 0.25int_0.25metFigure D.4: Sweep over possible RM weightings for PPO. 23D.2 S AMPLE OUTPUTS For reference, we provide several random prompts and model outputs for the various",
            "references": [
                {
                    "title": "Probabilistic Inference in Reinforcement Learning Done Right",
                    "abstract": "A popular perspective in Reinforcement learning (RL) casts the problem as probabilistic inference on a graphical model of the Markov decision process (MDP). The core object of study is the probability of each state-action pair being visited under the optimal policy. Previous approaches to approximate this quantity can be arbitrarily poor, leading to algorithms that do not implement genuine statistical inference and consequently do not perform well in challenging problems. In this work, we undertake a rigorous Bayesian treatment of the posterior probability of state-action optimality and clarify how it flows through the MDP. We first reveal that this quantity can indeed be used to generate a policy that explores efficiently, as measured by regret. Unfortunately, computing it is intractable, so we derive a new variational Bayesian approximation yielding a tractable convex optimization problem and establish that the resulting policy also explores efficiently. We call our approach VAPOR and show that it has strong connections to Thompson sampling, K-learning, and maximum entropy exploration. We conclude with some experiments demonstrating the performance advantage of a deep RL version of VAPOR."
                },
                {
                    "title": "A State Representation for Diminishing Rewards",
                    "abstract": "A common setting in multitask reinforcement learning (RL) demands that an agent rapidly adapt to various stationary reward functions randomly sampled from a fixed distribution. In such situations, the successor representation (SR) is a popular framework which supports rapid policy evaluation by decoupling a policy's expected discounted, cumulative state occupancies from a specific reward function. However, in the natural world, sequential tasks are rarely independent, and instead reflect shifting priorities based on the availability and subjective perception of rewarding stimuli. Reflecting this disjunction, in this paper we study the phenomenon of diminishing marginal utility and introduce a novel state representation, the $\\lambda$ representation ($\\lambda$R) which, surprisingly, is required for policy evaluation in this setting and which generalizes the SR as well as several other state representations from the literature. We establish the $\\lambda$R's formal properties and examine its normative advantages in the context of machine learning, as well as its usefulness for studying natural behaviors, particularly foraging."
                },
                {
                    "title": "Code Llama: Open Foundation Models for Code",
                    "abstract": "We release Code Llama, a family of large language models for code based on Llama 2 providing state-of-the-art performance among open models, infilling capabilities, support for large input contexts, and zero-shot instruction following ability for programming tasks. We provide multiple flavors to cover a wide range of applications: foundation models (Code Llama), Python specializations (Code Llama - Python), and instruction-following models (Code Llama - Instruct) with 7B, 13B, 34B and 70B parameters each. All models are trained on sequences of 16k tokens and show improvements on inputs with up to 100k tokens. 7B, 13B and 70B Code Llama and Code Llama - Instruct variants support infilling based on surrounding content. Code Llama reaches state-of-the-art performance among open models on several code benchmarks, with scores of up to 67% and 65% on HumanEval and MBPP, respectively. Notably, Code Llama - Python 7B outperforms Llama 2 70B on HumanEval and MBPP, and all our models outperform every other publicly available model on MultiPL-E. We release Code Llama under a permissive license that allows for both research and commercial use."
                },
                {
                    "title": "Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback",
                    "abstract": "Reinforcement learning from human feedback (RLHF) is a technique for training AI systems to align with human goals. RLHF has emerged as the central method used to finetune state-of-the-art large language models (LLMs). Despite this popularity, there has been relatively little public work systematizing its flaws. In this paper, we (1) survey open problems and fundamental limitations of RLHF and related methods; (2) overview techniques to understand, improve, and complement RLHF in practice; and (3) propose auditing and disclosure standards to improve societal oversight of RLHF systems. Our work emphasizes the limitations of RLHF and highlights the importance of a multi-faceted approach to the development of safer AI systems."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Benchmarking Large Language Model Capabilities for Conditional Generation",
                    "abstract": "Pre-trained large language models (PLMs) underly most new developments in natural language processing. They have shifted the field from application-specific model pipelines to a single model that is adapted to a wide range of tasks. Autoregressive PLMs like GPT-3 or PaLM and associated techniques like fewshot learning, have additionally shifted the output modality to generation instead of classification or regression. Despite their ubiquitous use, the generation quality of language models is rarely evaluated when these models are introduced. Additionally, it is unclear how existing generation tasks\u2013while they can be used to compare systems at a high level\u2013relate to the real world use cases for which people have been adopting them. In this work, we discuss how to adapt existing application-specific generation benchmarks to PLMs and provide an in-depth, empirical study of the limitations and capabilities of PLMs in natural language generation tasks along dimensions such as scale, architecture, input and output language. Our results show that PLMs differ in their applicability to different data regimes and their generalization to multiple languages. They further inform practitioners as to which PLMs to use for a given generation task setup. We share best practices to be taken into consideration when benchmarking generation capabilities during the development of upcoming PLMs."
                },
                {
                    "title": "Provably Efficient Primal-Dual Reinforcement Learning for CMDPs with Non-stationary Objectives and Constraints",
                    "abstract": "We consider primal-dual-based reinforcement learning (RL) in episodic constrained Markov decision processes (CMDPs) with non-stationary objectives and constraints, which plays a central role in ensuring the safety of RL in time-varying environments. In this problem, the reward/utility functions and the state transition functions are both allowed to vary arbitrarily over time as long as their cumulative variations do not exceed certain known variation budgets. Designing safe RL algorithms in time-varying environments is particularly challenging because of the need to integrate the constraint violation reduction, safe exploration, and adaptation to the non-stationarity. To this end, we identify two alternative conditions on the time-varying constraints under which we can guarantee the safety in the long run. We also propose the Periodically Restarted Optimistic Primal-Dual Proximal Policy Optimization (PROPD-PPO) algorithm that can coordinate with both two conditions. Furthermore, a dynamic regret bound and a constraint violation bound are established for the proposed algorithm in both the linear kernel CMDP function approximation setting and the tabular CMDP setting under two alternative conditions. This paper provides the first provably efficient algorithm for non-stationary CMDPs with safe exploration."
                },
                {
                    "title": "Fine-Grained Human Feedback Gives Better Rewards for Language Model Training",
                    "abstract": "Language models (LMs) often exhibit undesirable text generation behaviors, including generating false, toxic, or irrelevant outputs. Reinforcement learning from human feedback (RLHF) - where human preference judgments on LM outputs are transformed into a learning signal - has recently shown promise in addressing these issues. However, such holistic feedback conveys limited information on long text outputs; it does not indicate which aspects of the outputs influenced user preference; e.g., which parts contain what type(s) of errors. In this paper, we use fine-grained human feedback (e.g., which sentence is false, which sub-sentence is irrelevant) as an explicit training signal. We introduce Fine-Grained RLHF, a framework that enables training and learning from reward functions that are fine-grained in two respects: (1) density, providing a reward after every segment (e.g., a sentence) is generated; and (2) incorporating multiple reward models associated with different feedback types (e.g., factual incorrectness, irrelevance, and information incompleteness). We conduct experiments on detoxification and long-form question answering to illustrate how learning with such reward functions leads to improved performance, supported by both automatic and human evaluation. Additionally, we show that LM behaviors can be customized using different combinations of fine-grained reward models. We release all data, collected human feedback, and codes at https://FineGrainedRLHF.github.io."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "RRHF: Rank Responses to Align Language Models with Human Feedback without tears",
                    "abstract": "Reinforcement Learning from Human Feedback (RLHF) facilitates the alignment of large language models with human preferences, significantly enhancing the quality of interactions between humans and models. InstructGPT implements RLHF through several stages, including Supervised Fine-Tuning (SFT), reward model training, and Proximal Policy Optimization (PPO). However, PPO is sensitive to hyperparameters and requires multiple models in its standard implementation, making it hard to train and scale up to larger parameter counts. In contrast, we propose a novel learning paradigm called RRHF, which scores sampled responses from different sources via a logarithm of conditional probabilities and learns to align these probabilities with human preferences through ranking loss. RRHF can leverage sampled responses from various sources including the model responses from itself, other large language model responses, and human expert responses to learn to rank them. RRHF only needs 1 to 2 models during tuning and can efficiently align language models with human preferences robustly without complex hyperparameter tuning. Additionally, RRHF can be considered an extension of SFT and reward model training while being simpler than PPO in terms of coding, model counts, and hyperparameters. We evaluate RRHF on the Helpful and Harmless dataset, demonstrating comparable alignment performance with PPO by reward model score and human labeling. Extensive experiments show that the performance of RRHF is highly related to sampling quality which suggests RRHF is a best-of-n learner. Codes available at https://github.com/GanjinZero/RRHF."
                },
                {
                    "title": "Sparks of Artificial General Intelligence: Early experiments with GPT-4",
                    "abstract": "Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. The latest model developed by OpenAI, GPT-4, was trained using an unprecedented scale of compute and data. In this paper, we report on our investigation of an early version of GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT-4 is part of a new cohort of LLMs (along with ChatGPT and Google's PaLM for example) that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4's performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4's capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system. In our exploration of GPT-4, we put special emphasis on discovering its limitations, and we discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for pursuing a new paradigm that moves beyond next-word prediction. We conclude with reflections on societal influences of the recent technological leap and future research directions."
                },
                {
                    "title": "Efficient exploration via epistemic-risk-seeking policy optimization",
                    "abstract": "Exploration remains a key challenge in deep reinforcement learning (RL). Optimism in the face of uncertainty is a well-known heuristic with theoretical guarantees in the tabular setting, but how best to translate the principle to deep reinforcement learning, which involves online stochastic gradients and deep network function approximators, is not fully understood. In this paper we propose a new, differentiable optimistic objective that when optimized yields a policy that provably explores efficiently, with guarantees even under function approximation. Our new objective is a zero-sum two-player game derived from endowing the agent with an epistemic-risk-seeking utility function, which converts uncertainty into value and encourages the agent to explore uncertain states. We show that the solution to this game minimizes an upper bound on the regret, with the 'players' each attempting to minimize one component of a particular regret decomposition. We derive a new model-free algorithm which we call 'epistemic-risk-seeking actor-critic' (ERSAC), which is simply an application of simultaneous stochastic gradient ascent-descent to the game. Finally, we discuss a recipe for incorporating off-policy data and show that combining the risk-seeking objective with replay data yields a double benefit in terms of statistical efficiency. We conclude with some results showing good performance of a deep RL agent using the technique on the challenging 'DeepSea' environment, showing significant performance improvements even over other efficient exploration techniques, as well as improved performance on the Atari benchmark."
                },
                {
                    "title": "ReLOAD: Reinforcement Learning with Optimistic Ascent-Descent for Last-Iterate Convergence in Constrained MDPs",
                    "abstract": "In recent years, Reinforcement Learning (RL) has been applied to real-world problems with increasing success. Such applications often require to put constraints on the agent's behavior. Existing algorithms for constrained RL (CRL) rely on gradient descent-ascent, but this approach comes with a caveat. While these algorithms are guaranteed to converge on average, they do not guarantee last-iterate convergence, i.e., the current policy of the agent may never converge to the optimal solution. In practice, it is often observed that the policy alternates between satisfying the constraints and maximizing the reward, rarely accomplishing both objectives simultaneously. Here, we address this problem by introducing Reinforcement Learning with Optimistic Ascent-Descent (ReLOAD), a principled CRL method with guaranteed last-iterate convergence. We demonstrate its empirical effectiveness on a wide variety of CRL problems including discrete MDPs and continuous control. In the process we establish a benchmark of challenging CRL problems."
                },
                {
                    "title": "Fine-tuning language models to find agreement among humans with diverse preferences",
                    "abstract": "Recent work in large language modeling (LLMs) has used fine-tuning to align outputs with the preferences of a prototypical user. This work assumes that human preferences are static and homogeneous across individuals, so that aligning to a a single\"generic\"user will confer more general alignment. Here, we embrace the heterogeneity of human preferences to consider a different challenge: how might a machine help people with diverse views find agreement? We fine-tune a 70 billion parameter LLM to generate statements that maximize the expected approval for a group of people with potentially diverse opinions. Human participants provide written opinions on thousands of questions touching on moral and political issues (e.g.,\"should we raise taxes on the rich?\"), and rate the LLM's generated candidate consensus statements for agreement and quality. A reward model is then trained to predict individual preferences, enabling it to quantify and rank consensus statements in terms of their appeal to the overall group, defined according to different aggregation (social welfare) functions. The model produces consensus statements that are preferred by human users over those from prompted LLMs (>70%) and significantly outperforms a tight fine-tuned baseline that lacks the final ranking step. Further, our best model's consensus statements are preferred over the best human-generated opinions (>65%). We find that when we silently constructed consensus statements from only a subset of group members, those who were excluded were more likely to dissent, revealing the sensitivity of the consensus to individual contributions. These results highlight the potential to use LLMs to help groups of humans align their values with one another."
                },
                {
                    "title": "Scaling Laws for Reward Model Overoptimization",
                    "abstract": "In reinforcement learning from human feedback, it is common to optimize against a reward model trained to predict human preferences. Because the reward model is an imperfect proxy, optimizing its value too much can hinder ground truth performance, in accordance with Goodhart's law. This effect has been frequently observed, but not carefully measured due to the expense of collecting human preference data. In this work, we use a synthetic setup in which a fixed\"gold-standard\"reward model plays the role of humans, providing labels used to train a proxy reward model. We study how the gold reward model score changes as we optimize against the proxy reward model using either reinforcement learning or best-of-$n$ sampling. We find that this relationship follows a different functional form depending on the method of optimization, and that in both cases its coefficients scale smoothly with the number of reward model parameters. We also study the effect on this relationship of the size of the reward model dataset, the number of reward model and policy parameters, and the coefficient of the KL penalty added to the reward in the reinforcement learning setup. We explore the implications of these empirical results for theoretical considerations in AI alignment."
                },
                {
                    "title": "Is Reinforcement Learning (Not) for Natural Language Processing?: Benchmarks, Baselines, and Building Blocks for Natural Language Policy Optimization",
                    "abstract": "We tackle the problem of aligning pre-trained large language models (LMs) with human preferences. If we view text generation as a sequential decision-making problem, reinforcement learning (RL) appears to be a natural conceptual framework. However, using RL for LM-based generation faces empirical challenges, including training instability due to the combinatorial action space, as well as a lack of open-source libraries and benchmarks customized for LM alignment. Thus, a question rises in the research community: is RL a practical paradigm for NLP? To help answer this, we first introduce an open-source modular library, RL4LMs (Reinforcement Learning for Language Models), for optimizing language generators with RL. The library consists of on-policy RL algorithms that can be used to train any encoder or encoder-decoder LM in the HuggingFace library (Wolf et al. 2020) with an arbitrary reward function. Next, we present the GRUE (General Reinforced-language Understanding Evaluation) benchmark, a set of 6 language generation tasks which are supervised not by target strings, but by reward functions which capture automated measures of human preference. GRUE is the first leaderboard-style evaluation of RL algorithms for NLP tasks. Finally, we introduce an easy-to-use, performant RL algorithm, NLPO (Natural Language Policy Optimization) that learns to effectively reduce the combinatorial action space in language generation. We show 1) that RL techniques are generally better than supervised methods at aligning LMs to human preferences; and 2) that NLPO exhibits greater stability and performance than previous policy gradient methods (e.g., PPO (Schulman et al. 2017)), based on both automatic and human evaluations."
                },
                {
                    "title": "Improving alignment of dialogue agents via targeted human judgements",
                    "abstract": "We present Sparrow, an information-seeking dialogue agent trained to be more helpful, correct, and harmless compared to prompted language model baselines. We use reinforcement learning from human feedback to train our models with two new additions to help human raters judge agent behaviour. First, to make our agent more helpful and harmless, we break down the requirements for good dialogue into natural language rules the agent should follow, and ask raters about each rule separately. We demonstrate that this breakdown enables us to collect more targeted human judgements of agent behaviour and allows for more efficient rule-conditional reward models. Second, our agent provides evidence from sources supporting factual claims when collecting preference judgements over model statements. For factual questions, evidence provided by Sparrow supports the sampled response 78% of the time. Sparrow is preferred more often than baselines while being more resilient to adversarial probing by humans, violating our rules only 8% of the time when probed. Finally, we conduct extensive analyses showing that though our model learns to follow our rules it can exhibit distributional biases."
                },
                {
                    "title": "Discovering Policies with DOMiNO: Diversity Optimization Maintaining Near Optimality",
                    "abstract": "Finding different solutions to the same problem is a key aspect of intelligence associated with creativity and adaptation to novel situations. In reinforcement learning, a set of diverse policies can be useful for exploration, transfer, hierarchy, and robustness. We propose DOMiNO, a method for Diversity Optimization Maintaining Near Optimality. We formalize the problem as a Constrained Markov Decision Process where the objective is to find diverse policies, measured by the distance between the state occupancies of the policies in the set, while remaining near-optimal with respect to the extrinsic reward. We demonstrate that the method can discover diverse and meaningful behaviors in various domains, such as different locomotion patterns in the DeepMind Control Suite. We perform extensive analysis of our approach, compare it with other multi-objective baselines, demonstrate that we can control both the quality and the diversity of the set via interpretable hyperparameters, and show that the discovered set is robust to perturbations."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Towards an Understanding of Default Policies in Multitask Policy Optimization",
                    "abstract": "Much of the recent success of deep reinforcement learning has been driven by regularized policy optimization (RPO) algorithms with strong performance across multiple domains. In this family of methods, agents are trained to maximize cumulative reward while penalizing deviation in behavior from some reference, or default policy. In addition to empirical success, there is a strong theoretical foundation for understanding RPO methods applied to single tasks, with connections to natural gradient, trust region, and variational approaches. However, there is limited formal understanding of desirable properties for default policies in the multitask setting, an increasingly important domain as the field shifts towards training more generally capable agents. Here, we take a first step towards filling this gap by formally linking the quality of the default policy to its effect on optimization. Using these results, we then derive a principled RPO algorithm for multitask learning with strong performance guarantees."
                },
                {
                    "title": "A First-Occupancy Representation for Reinforcement Learning",
                    "abstract": "Both animals and artificial agents benefit from state representations that support rapid transfer of learning across tasks and which enable them to efficiently traverse their environments to reach rewarding states. The successor representation (SR), which measures the expected cumulative, discounted state occupancy under a fixed policy, enables efficient transfer to different reward structures in an otherwise constant Markovian environment and has been hypothesized to underlie aspects of biological behavior and neural activity. However, in the real world, rewards may move or only be available for consumption once, may shift location, or agents may simply aim to reach goal states as rapidly as possible without the constraint of artificially imposed task horizons. In such cases, the most behaviorally-relevant representation would carry information about when the agent was likely to first reach states of interest, rather than how often it should expect to visit them over a potentially infinite time span. To reflect such demands, we introduce the first-occupancy representation (FR), which measures the expected temporal discount to the first time a state is accessed. We demonstrate that the FR facilitates exploration, the selection of efficient paths to desired states, allows the agent, under certain conditions, to plan provably optimal trajectories defined by a sequence of subgoals, and induces similar behavior to animals avoiding threatening stimuli."
                },
                {
                    "title": "TruthfulQA: Measuring How Models Mimic Human Falsehoods",
                    "abstract": "We propose a benchmark to measure whether a language model is truthful in generating answers to questions. The benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics. We crafted questions that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. We tested GPT-3, GPT-Neo/J, GPT-2 and a T5-based model. The best model was truthful on 58% of questions, while human performance was 94%. Models generated many false answers that mimic popular misconceptions and have the potential to deceive humans. The largest models were generally the least truthful. This contrasts with other NLP tasks, where performance improves with model size. However, this result is expected if false answers are learned from the training distribution. We suggest that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web."
                },
                {
                    "title": "CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and Generation",
                    "abstract": "Pre-trained models for Natural Languages (NL) like BERT and GPT have been recently shown to transfer well to Programming Languages (PL) and largely benefit a broad set of code-related tasks. Despite their success, most current methods either rely on an encoder-only (or decoder-only) pre-training that is suboptimal for generation (resp. understanding) tasks or process the code snippet in the same way as NL, neglecting the special characteristics of PL such as token types. We present CodeT5, a unified pre-trained encoder-decoder Transformer model that better leverages the code semantics conveyed from the developer-assigned identifiers. Our model employs a unified framework to seamlessly support both code understanding and generation tasks and allows for multi-task learning. Besides, we propose a novel identifier-aware pre-training task that enables the model to distinguish which code tokens are identifiers and to recover them when they are masked. Furthermore, we propose to exploit the user-written code comments with a bimodal dual generation task for better NL-PL alignment. Comprehensive experiments show that CodeT5 significantly outperforms prior methods on understanding tasks such as code defect detection and clone detection, and generation tasks across various directions including PL-NL, NL-PL, and PL-PL. Further analysis reveals that our model can better capture semantic information from code. Our code and pre-trained models are released at https://github.com/salesforce/CodeT5."
                },
                {
                    "title": "On the Opportunities and Risks of Foundation Models",
                    "abstract": "AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature."
                },
                {
                    "title": "Unified Pre-training for Program Understanding and Generation",
                    "abstract": "Code summarization and generation empower conversion between programming language (PL) and natural language (NL), while code translation avails the migration of legacy code from one PL to another. This paper introduces PLBART, a sequence-to-sequence model capable of performing a broad spectrum of program and language understanding and generation tasks. PLBART is pre-trained on an extensive collection of Java and Python functions and associated NL text via denoising autoencoding. Experiments on code summarization in the English language, code generation, and code translation in seven programming languages show that PLBART outperforms or rivals state-of-the-art models. Moreover, experiments on discriminative tasks, e.g., program repair, clone detection, and vulnerable code detection, demonstrate PLBART\u2019s effectiveness in program understanding. Furthermore, analysis reveals that PLBART learns program syntax, style (e.g., identifier naming convention), logical flow (e.g., \u201cif\u201c block inside an \u201celse\u201c block is equivalent to \u201celse if\u201c block) that are crucial to program semantics and thus excels even with limited annotations."
                },
                {
                    "title": "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? \ud83e\udd9c",
                    "abstract": "The past 3 years of work in NLP have been characterized by the development and deployment of ever larger language models, especially for English. BERT, its variants, GPT-2/3, and others, most recently Switch-C, have pushed the boundaries of the possible both through architectural innovations and through sheer size. Using these pretrained models and the methodology of fine-tuning them for specific tasks, researchers have extended the state of the art on a wide array of tasks as measured by leaderboards on specific benchmarks for English. In this paper, we take a step back and ask: How big is too big? What are the possible risks associated with this technology and what paths are available for mitigating those risks? We provide recommendations including weighing the environmental and financial costs first, investing resources into curating and carefully documenting datasets rather than ingesting everything on the web, carrying out pre-development exercises evaluating how the planned approach fits into research and development goals and supports stakeholder values, and encouraging research directions beyond ever larger language models."
                },
                {
                    "title": "Towards Continual Reinforcement Learning: A Review and Perspectives",
                    "abstract": "In this article, we aim to provide a literature review of different formulations and approaches to continual reinforcement learning (RL), also known as lifelong or non-stationary RL. We begin by discussing our perspective on why RL is a natural fit for studying continual learning. We then provide a taxonomy of different continual RL formulations by mathematically characterizing two key properties of non-stationarity, namely, the scope and driver non-stationarity. This offers a unified view of various formulations. Next, we review and present a taxonomy of continual RL approaches. We go on to discuss evaluation of continual RL agents, providing an overview of benchmarks used in the literature and important metrics for understanding agent performance. Finally, we highlight open problems and challenges in bridging the gap between the current state of continual RL and findings in neuroscience. While still in its early days, the study of continual RL has the promise to develop better incremental reinforcement learners that can function in increasingly realistic applications where non-stationarity plays a vital role. These include applications such as those in the fields of healthcare, education, logistics, and robotics."
                },
                {
                    "title": "A Distributional Approach to Controlled Text Generation",
                    "abstract": "We propose a Distributional Approach for addressing Controlled Text Generation from pre-trained Language Models (LMs). This approach permits to specify, in a single formal framework, both \u201cpointwise\u201d and \u201cdistributional\u201d constraints over the target LM \u2014 to our knowledge, the first model with such generality \u2014 while minimizing KL divergence from the initial LM distribution. The optimal target distribution is then uniquely determined as an explicit EBM (Energy-Based Model) representation. From that optimal representation we then train a target controlled Autoregressive LM through an adaptive distributional variant of Policy Gradient. We conduct a first set of experiments over pointwise constraints showing the advantages of our approach over a set of baselines, in terms of obtaining a controlled LM balancing constraint satisfaction with divergence from the initial LM. We then perform experiments over distributional constraints, a unique feature of our approach, demonstrating its potential as a remedy to the problem of Bias in Language Models. Through an ablation study, we show the effectiveness of our adaptive technique for obtaining faster convergence.1"
                },
                {
                    "title": "Extracting Training Data from Large Language Models",
                    "abstract": "It has become common to publish large (billion parameter) language models that have been trained on private datasets. This paper demonstrates that in such settings, an adversary can perform a training data extraction attack to recover individual training examples by querying the language model. \nWe demonstrate our attack on GPT-2, a language model trained on scrapes of the public Internet, and are able to extract hundreds of verbatim text sequences from the model's training data. These extracted examples include (public) personally identifiable information (names, phone numbers, and email addresses), IRC conversations, code, and 128-bit UUIDs. Our attack is possible even though each of the above sequences are included in just one document in the training data. \nWe comprehensively evaluate our extraction attack to understand the factors that contribute to its success. For example, we find that larger models are more vulnerable than smaller models. We conclude by drawing lessons and discussing possible safeguards for training large language models."
                },
                {
                    "title": "Balancing Constraints and Rewards with Meta-Gradient D4PG",
                    "abstract": "Deploying Reinforcement Learning (RL) agents to solve real-world applications often requires satisfying complex system constraints. Often the constraint thresholds are incorrectly set due to the complex nature of a system or the inability to verify the thresholds offline (e.g, no simulator or reasonable offline evaluation procedure exists). This results in solutions where a task cannot be solved without violating the constraints. However, in many real-world cases, constraint violations are undesirable yet they are not catastrophic, motivating the need for soft-constrained RL approaches. We present two soft-constrained RL approaches that utilize meta-gradients to find a good trade-off between expected return and minimizing constraint violations. We demonstrate the effectiveness of these approaches by showing that they consistently outperform the baselines across four different Mujoco domains."
                },
                {
                    "title": "Efficient Wasserstein Natural Gradients for Reinforcement Learning",
                    "abstract": "A novel optimization approach is proposed for application to policy gradient methods and evolution strategies for reinforcement learning (RL). The procedure uses a computationally efficient Wasserstein natural gradient (WNG) descent that takes advantage of the geometry induced by a Wasserstein penalty to speed optimization. This method follows the recent theme in RL of including a divergence penalty in the objective to establish a trust region. Experiments on challenging tasks demonstrate improvements in both computational cost and performance over advanced baselines."
                },
                {
                    "title": "Learning to summarize from human feedback",
                    "abstract": "As language models become more powerful, training and evaluation are increasingly bottlenecked by the data and metrics used for a particular task. For example, summarization models are often trained to predict human reference summaries and evaluated using ROUGE, but both of these metrics are rough proxies for what we really care about---summary quality. In this work, we show that it is possible to significantly improve summary quality by training a model to optimize for human preferences. We collect a large, high-quality dataset of human comparisons between summaries, train a model to predict the human-preferred summary, and use that model as a reward function to fine-tune a summarization policy using reinforcement learning. We apply our method to a version of the TL;DR dataset of Reddit posts and find that our models significantly outperform both human reference summaries and much larger models fine-tuned with supervised learning alone. Our models also transfer to CNN/DM news articles, producing summaries nearly as good as the human reference without any news-specific fine-tuning. We conduct extensive analyses to understand our human feedback dataset and fine-tuned models We establish that our reward model generalizes to new datasets, and that optimizing our reward model results in better summaries than optimizing ROUGE according to humans. We hope the evidence from our paper motivates machine learning researchers to pay closer attention to how their training loss affects the model behavior they actually want."
                },
                {
                    "title": "Responsive Safety in Reinforcement Learning by PID Lagrangian Methods",
                    "abstract": "Lagrangian methods are widely used algorithms for constrained optimization problems, but their learning dynamics exhibit oscillations and overshoot which, when applied to safe reinforcement learning, leads to constraint-violating behavior during agent training. We address this shortcoming by proposing a novel Lagrange multiplier update method that utilizes derivatives of the constraint function. We take a controls perspective, wherein the traditional Lagrange multiplier update behaves as \\emph{integral} control; our terms introduce \\emph{proportional} and \\emph{derivative} control, achieving favorable learning dynamics through damping and predictive measures. We apply our PID Lagrangian methods in deep RL, setting a new state of the art in Safety Gym, a safe RL benchmark. Lastly, we introduce a new method to ease controller tuning by providing invariance to the relative numerical scales of reward and cost. Our extensive experiments demonstrate improved performance and hyperparameter robustness, while our algorithms remain nearly as simple to derive and implement as the traditional Lagrangian approach."
                },
                {
                    "title": "Reinforcement Learning for Non-Stationary Markov Decision Processes: The Blessing of (More) Optimism",
                    "abstract": "We consider un-discounted reinforcement learning (RL) in Markov decision processes (MDPs) under drifting non-stationarity, i.e., both the reward and state transition distributions are allowed to evolve over time, as long as their respective total variations, quantified by suitable metrics, do not exceed certain variation budgets. We first develop the Sliding Window Upper-Confidence bound for Reinforcement Learning with Confidence Widening (SWUCRL2-CW) algorithm, and establish its dynamic regret bound when the variation budgets are known. In addition, we propose the Bandit-over-Reinforcement Learning (BORL) algorithm to adaptively tune the SWUCRL2-CW algorithm to achieve the same dynamic regret bound, but in a parameter-free manner, i.e., without knowing the variation budgets. Notably, learning non-stationary MDPs via the conventional optimistic exploration technique presents a unique challenge absent in existing (non-stationary) bandit learning settings. We overcome the challenge by a novel confidence widening technique that incorporates additional optimism."
                },
                {
                    "title": "Deep Reinforcement Learning amidst Lifelong Non-Stationarity",
                    "abstract": "As humans, our goals and our environment are persistently changing throughout our lifetime based on our experiences, actions, and internal and external drives. In contrast, typical reinforcement learning problem set-ups consider decision processes that are stationary across episodes. Can we develop reinforcement learning algorithms that can cope with the persistent change in the former, more realistic problem settings? While on-policy algorithms such as policy gradients in principle can be extended to non-stationary settings, the same cannot be said for more efficient off-policy algorithms that replay past experiences when learning. In this work, we formalize this problem setting, and draw upon ideas from the online learning and probabilistic inference literature to derive an off-policy RL algorithm that can reason about and tackle such lifelong non-stationarity. Our method leverages latent variable models to learn a representation of the environment from current and past experiences, and performs off-policy RL with this representation. We further introduce several simulation environments that exhibit lifelong non-stationarity, and empirically find that our approach substantially outperforms approaches that do not reason about environment shift."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "A Distributional View on Multi-Objective Policy Optimization",
                    "abstract": "Many real-world problems require trading off multiple competing objectives. However, these objectives are often in different units and/or scales, which can make it challenging for practitioners to express numerical preferences over objectives in their native units. In this paper we propose a novel algorithm for multi-objective reinforcement learning that enables setting desired preferences for objectives in a scale-invariant way. We propose to learn an action distribution for each objective, and we use supervised learning to fit a parametric policy to a combination of these distributions. We demonstrate the effectiveness of our approach on challenging high-dimensional real and simulated robotics tasks, and show that setting different preferences in our framework allows us to trace out the space of nondominated solutions."
                },
                {
                    "title": "Exploration-Exploitation in Constrained MDPs",
                    "abstract": "In many sequential decision-making problems, the goal is to optimize a utility function while satisfying a set of constraints on different utilities. This learning problem is formalized through Constrained Markov Decision Processes (CMDPs). In this paper, we investigate the exploration-exploitation dilemma in CMDPs. While learning in an unknown CMDP, an agent should trade-off exploration to discover new information about the MDP, and exploitation of the current knowledge to maximize the reward while satisfying the constraints. While the agent will eventually learn a good or optimal policy, we do not want the agent to violate the constraints too often during the learning process. In this work, we analyze two approaches for learning in CMDPs. The first approach leverages the linear formulation of CMDP to perform optimistic planning at each episode. The second approach leverages the dual formulation (or saddle-point formulation) of CMDP to perform incremental, optimistic updates of the primal and dual variables. We show that both achieves sublinear regret w.r.t.\\ the main utility while having a sublinear regret on the constraint violations. That being said, we highlight a crucial difference between the two approaches; the linear programming approach results in stronger guarantees than in the dual formulation based approach."
                },
                {
                    "title": "Provably Efficient Safe Exploration via Primal-Dual Policy Optimization",
                    "abstract": "We study the Safe Reinforcement Learning (SRL) problem using the Constrained Markov Decision Process (CMDP) formulation in which an agent aims to maximize the expected total reward subject to a safety constraint on the expected total value of a criterion function (e.g., utility). We focus on an episodic setting with the function approximation where the reward and criterion functions and the Markov transition kernels all have a linear structure but do not impose any additional assumptions on the sampling model. Designing SRL algorithms with provable computational and statistical efficiency is particularly challenging under this setting because of the need to incorporate both the safety constraint and the function approximation into the fundamental exploitation/exploration tradeoff. To this end, we present an {O}ptimistic {P}rimal-{D}ual Proximal Policy {OP}timization (OPDOP) algorithm where the value function is estimated by combining the least-squares policy evaluation and an additional bonus term for safe exploration. We prove that the proposed algorithm achieves an O(d^{1.5}H^{3.5}\\sqrt{T}) regret and an O(d^{1.5}H^{3.5}\\sqrt{T}) constraint violation, where d is the dimension of the feature mapping, H is the horizon of each episode, and T is the total number of steps. We establish these bounds under the following two settings: (i) Both the reward and criterion functions can change adversarially but are revealed entirely after each episode. (ii) The reward/criterion functions are fixed but the feedback after each episode is bandit. Our bounds depend on the capacity of the state space only through the dimension of the feature mapping and thus our results hold even when the number of states goes to infinity. To the best of our knowledge, we provide the first provably efficient policy optimization algorithm for CMDPs with safe exploration."
                },
                {
                    "title": "Dota 2 with Large Scale Deep Reinforcement Learning",
                    "abstract": "On April 13th, 2019, OpenAI Five became the first AI system to defeat the world champions at an esports game. The game of Dota 2 presents novel challenges for AI systems such as long time horizons, imperfect information, and complex, continuous state-action spaces, all challenges which will become increasingly central to more capable AI systems. OpenAI Five leveraged existing reinforcement learning techniques, scaled to learn from batches of approximately 2 million frames every 2 seconds. We developed a distributed training system and tools for continual training which allowed us to train OpenAI Five for 10 months. By defeating the Dota 2 world champion (Team OG), OpenAI Five demonstrates that self-play reinforcement learning can achieve superhuman performance on a difficult task."
                },
                {
                    "title": "Constrained Reinforcement Learning Has Zero Duality Gap",
                    "abstract": "Autonomous agents must often deal with conflicting requirements, such as completing tasks using the least amount of time/energy, learning multiple tasks, or dealing with multiple opponents. In the context of reinforcement learning~(RL), these problems are addressed by (i)~designing a reward function that simultaneously describes all requirements or (ii)~combining modular value functions that encode them individually. Though effective, these methods have critical downsides. Designing good reward functions that balance different objectives is challenging, especially as the number of objectives grows. Moreover, implicit interference between goals may lead to performance plateaus as they compete for resources, particularly when training on-policy. Similarly, selecting parameters to combine value functions is at least as hard as designing an all-encompassing reward, given that the effect of their values on the overall policy is not straightforward. The later is generally addressed by formulating the conflicting requirements as a constrained RL problem and solved using Primal-Dual methods. These algorithms are in general not guaranteed to converge to the optimal solution since the problem is not convex. This work provides theoretical support to these approaches by establishing that despite its non-convexity, this problem has zero duality gap, i.e., it can be solved exactly in the dual domain, where it becomes convex. Finally, we show this result basically holds if the policy is described by a good parametrization~(e.g., neural networks) and we connect this result with primal-dual algorithms present in the literature and we establish the convergence to the optimal solution."
                },
                {
                    "title": "Fine-Tuning Language Models from Human Preferences",
                    "abstract": "Reward learning enables the application of reinforcement learning (RL) to tasks where reward is defined by human judgment, building a model of reward by asking humans questions. Most work on reward learning has used simulated environments, but complex information about values is often expressed in natural language, and we believe reward learning for language is a key to making RL practical and safe for real-world tasks. In this paper, we build on advances in generative pretraining of language models to apply reward learning to four natural language tasks: continuing text with positive sentiment or physically descriptive language, and summarization tasks on the TL;DR and CNN/Daily Mail datasets. For stylistic continuation we achieve good results with only 5,000 comparisons evaluated by humans. For summarization, models trained with 60,000 comparisons copy whole sentences from the input but skip irrelevant preamble; this leads to reasonable ROUGE scores and very good performance according to our human labelers, but may be exploiting the fact that labelers rely on simple heuristics."
                },
                {
                    "title": "RoBERTa: A Robustly Optimized BERT Pretraining Approach",
                    "abstract": "Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code."
                },
                {
                    "title": "Learning to Score Behaviors for Guided Policy Optimization",
                    "abstract": "We introduce a new approach for comparing reinforcement learning policies, using Wasserstein distances (WDs) in a newly defined latent behavioral space. We show that by utilizing the dual formulation of the WD, we can learn score functions over policy behaviors that can in turn be used to lead policy optimization towards (or away from) (un)desired behaviors. Combined with smoothed WDs, the dual formulation allows us to devise efficient algorithms that take stochastic gradient descent steps through WD regularizers. We incorporate these regularizers into two novel on-policy algorithms, Behavior-Guided Policy Gradient and Behavior-Guided Evolution Strategies, which we demonstrate can outperform existing methods in a variety of challenging environments. We also provide an open source demo."
                },
                {
                    "title": "Information asymmetry in KL-regularized RL",
                    "abstract": "Many real world tasks exhibit rich structure that is repeated across different parts of the state space or in time. In this work we study the possibility of leveraging such repeated structure to speed up and regularize learning. We start from the KL regularized expected reward objective which introduces an additional component, a default policy. Instead of relying on a fixed default policy, we learn it from data. But crucially, we restrict the amount of information the default policy receives, forcing it to learn reusable behaviors that help the policy learn faster. We formalize this strategy and discuss connections to information bottleneck approaches and to the variational EM algorithm. We present empirical results in both discrete and continuous action domains and demonstrate that, for certain tasks, learning a default policy alongside the policy can significantly speed up and improve learning."
                },
                {
                    "title": "Non-Stationary Markov Decision Processes a Worst-Case Approach using Model-Based Reinforcement Learning",
                    "abstract": "This work tackles the problem of robust zero-shot planning in non-stationary stochastic environments. We study Markov Decision Processes (MDPs) evolving over time and consider Model-Based Reinforcement Learning algorithms in this setting. We make two hypotheses: 1) the environment evolves continuously with a bounded evolution rate; 2) a current model is known at each decision epoch but not its evolution. Our contribution can be presented in four points. 1) we define a specific class of MDPs that we call Non-Stationary MDPs (NSMDPs). We introduce the notion of regular evolution by making an hypothesis of Lipschitz-Continuity on the transition and reward functions w.r.t. time; 2) we consider a planning agent using the current model of the environment but unaware of its future evolution. This leads us to consider a worst-case method where the environment is seen as an adversarial agent; 3) following this approach, we propose the Risk-Averse Tree-Search (RATS) algorithm, a zero-shot Model-Based method similar to Minimax search; 4) we illustrate the benefits brought by RATS empirically and compare its performance with reference Model-Based algorithms."
                },
                {
                    "title": "Reward Constrained Policy Optimization",
                    "abstract": "Solving tasks in Reinforcement Learning is no easy feat. As the goal of the agent is to maximize the accumulated reward, it often learns to exploit loopholes and misspecifications in the reward signal resulting in unwanted behavior. While constraints may solve this issue, there is no closed form solution for general constraints. In this work we present a novel multi-timescale approach for constrained policy optimization, called `Reward Constrained Policy Optimization' (RCPO), which uses an alternative penalty signal to guide the policy towards a constraint satisfying one. We prove the convergence of our approach and provide empirical evidence of its ability to train constraint satisfying policies."
                },
                {
                    "title": "Reinforcement Learning and Control as Probabilistic Inference: Tutorial and Review",
                    "abstract": "The framework of reinforcement learning or optimal control provides a mathematical formalization of intelligent decision making that is powerful and broadly applicable. While the general form of the reinforcement learning problem enables effective reasoning about uncertainty, the connection between reinforcement learning and inference in probabilistic models is not immediately obvious. However, such a connection has considerable value when it comes to algorithm design: formalizing a problem as probabilistic inference in principle allows us to bring to bear a wide array of approximate inference tools, extend the model in flexible and powerful ways, and reason about compositionality and partial observability. In this article, we will discuss how a generalization of the reinforcement learning or optimal control problem, which is sometimes termed maximum entropy reinforcement learning, is equivalent to exact probabilistic inference in the case of deterministic dynamics, and variational inference in the case of stochastic dynamics. We will present a detailed derivation of this framework, overview prior work that has drawn on this and related ideas to propose new reinforcement learning and control algorithms, and describe perspectives on future research."
                },
                {
                    "title": "A Lyapunov-based Approach to Safe Reinforcement Learning",
                    "abstract": "In many real-world reinforcement learning (RL) problems, besides optimizing the main objective function, an agent must concurrently avoid violating a number of constraints. In particular, besides optimizing performance it is crucial to guarantee the safety of an agent during training as well as deployment (e.g. a robot should avoid taking actions - exploratory or not - which irrevocably harm its hardware). To incorporate safety in RL, we derive algorithms under the framework of constrained Markov decision problems (CMDPs), an extension of the standard Markov decision problems (MDPs) augmented with constraints on expected cumulative costs. Our approach hinges on a novel \\emph{Lyapunov} method. We define and present a method for constructing Lyapunov functions, which provide an effective way to guarantee the global safety of a behavior policy during training via a set of local, linear constraints. Leveraging these theoretical underpinnings, we show how to use the Lyapunov approach to systematically transform dynamic programming (DP) and RL algorithms into their safe counterparts. To illustrate their effectiveness, we evaluate these algorithms in several CMDP planning and decision-making tasks on a safety benchmark domain. Our results show that our proposed method significantly outperforms existing baselines in balancing constraint satisfaction and performance."
                },
                {
                    "title": "A Call for Clarity in Reporting BLEU Scores",
                    "abstract": "The field of machine translation faces an under-recognized problem because of inconsistency in the reporting of scores from its dominant metric. Although people refer to \u201cthe\u201d BLEU score, BLEU is in fact a parameterized metric whose values can vary wildly with changes to these parameters. These parameters are often not reported or are hard to find, and consequently, BLEU scores between papers cannot be directly compared. I quantify this variation, finding differences as high as 1.8 between commonly used configurations. The main culprit is different tokenization and normalization schemes applied to the reference. Pointing to the success of the parsing community, I suggest machine translation researchers settle upon the BLEU scheme used by the annual Conference on Machine Translation (WMT), which does not allow for user-supplied reference processing, and provide a new tool, SACREBLEU, to facilitate this."
                },
                {
                    "title": "Maximum a Posteriori Policy Optimisation",
                    "abstract": "We introduce a new algorithm for reinforcement learning called Maximum aposteriori Policy Optimisation (MPO) based on coordinate ascent on a relative entropy objective. We show that several existing methods can directly be related to our derivation. We develop two off-policy algorithms and demonstrate that they are competitive with the state-of-the-art in deep reinforcement learning. In particular, for continuous control, our method outperforms existing methods with respect to sample efficiency, premature convergence and robustness to hyperparameter settings while achieving similar or better final performance."
                },
                {
                    "title": "IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures",
                    "abstract": "In this work we aim to solve a large collection of tasks using a single reinforcement learning agent with a single set of parameters. A key challenge is to handle the increased amount of data and extended training time. We have developed a new distributed agent IMPALA (Importance Weighted Actor-Learner Architecture) that not only uses resources more efficiently in single-machine training but also scales to thousands of machines without sacrificing data efficiency or resource utilisation. We achieve stable learning at high throughput by combining decoupled acting and learning with a novel off-policy correction method called V-trace. We demonstrate the effectiveness of IMPALA for multi-task reinforcement learning on DMLab-30 (a set of 30 tasks from the DeepMind Lab environment (Beattie et al., 2016)) and Atari-57 (all available Atari games in Arcade Learning Environment (Bellemare et al., 2013a)). Our results show that IMPALA is able to achieve better performance than previous agents with less data, and crucially exhibits positive transfer between tasks as a result of its multi-task approach."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "DailyDialog: A Manually Labelled Multi-turn Dialogue Dataset",
                    "abstract": "We develop a high-quality multi-turn dialog dataset, DailyDialog, which is intriguing in several aspects. The language is human-written and less noisy. The dialogues in the dataset reflect our daily communication way and cover various topics about our daily life. We also manually label the developed dataset with communication intention and emotion information. Then, we evaluate existing approaches on DailyDialog dataset and hope it benefit the research field of dialog systems. The dataset is available on http://yanran.li/dailydialog"
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "Distral: Robust multitask reinforcement learning",
                    "abstract": "Most deep reinforcement learning algorithms are data inefficient in complex and rich environments, limiting their applicability to many scenarios. One direction for improving data efficiency is multitask learning with shared neural network parameters, where efficiency may be improved through transfer across related tasks. In practice, however, this is not usually observed, because gradients from different tasks can interfere negatively, making learning unstable and sometimes even less data efficient. Another issue is the different reward schemes between tasks, which can easily lead to one task dominating the learning of a shared model. We propose a new approach for joint training of multiple tasks, which we refer to as Distral (Distill & transfer learning). Instead of sharing parameters between the different workers, we propose to share a \"distilled\" policy that captures common behaviour across tasks. Each worker is trained to solve its own task while constrained to stay close to the shared policy, while the shared policy is trained by distillation to be the centroid of all task policies. Both aspects of the learning process are derived by optimizing a joint objective function. We show that our approach supports efficient transfer on complex 3D environments, outperforming several related methods. Moreover, the proposed learning process is more robust and more stable---attributes that are critical in deep reinforcement learning."
                },
                {
                    "title": "Deep Reinforcement Learning from Human Preferences",
                    "abstract": "For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems. In this work, we explore goals defined in terms of (non-expert) human preferences between pairs of trajectory segments. We show that this approach can effectively solve complex RL tasks without access to the reward function, including Atari games and simulated robot locomotion, while providing feedback on less than one percent of our agent's interactions with the environment. This reduces the cost of human oversight far enough that it can be practically applied to state-of-the-art RL systems. To demonstrate the flexibility of our approach, we show that we can successfully train complex novel behaviors with about an hour of human time. These behaviors and environments are considerably more complex than any that have been previously learned from human feedback."
                },
                {
                    "title": "Constrained Policy Optimization",
                    "abstract": "For many applications of reinforcement learning it can be more convenient to specify both a reward function and constraints, rather than trying to design behavior through the reward function. For example, systems that physically interact with or around humans should satisfy safety constraints. Recent advances in policy search algorithms (Mnih et al., 2016, Schulman et al., 2015, Lillicrap et al., 2016, Levine et al., 2016) have enabled new capabilities in high-dimensional control, but do not consider the constrained setting. \nWe propose Constrained Policy Optimization (CPO), the first general-purpose policy search algorithm for constrained reinforcement learning with guarantees for near-constraint satisfaction at each iteration. Our method allows us to train neural network policies for high-dimensional control while making guarantees about policy behavior all throughout training. Our guarantees are based on a new theoretical result, which is of independent interest: we prove a bound relating the expected returns of two policies to an average divergence between them. We demonstrate the effectiveness of our approach on simulated robot locomotion tasks where the agent must satisfy constraints motivated by safety."
                },
                {
                    "title": "ROUGE 2.0: Updated and Improved Measures for Evaluation of Summarization Tasks",
                    "abstract": "Evaluation of summarization tasks is extremely crucial to determining the quality of machine generated summaries. Over the last decade, ROUGE has become the standard automatic evaluation measure for evaluating summarization tasks. While ROUGE has been shown to be effective in capturing n-gram overlap between system and human composed summaries, there are several limitations with the existing ROUGE measures in terms of capturing synonymous concepts and coverage of topics. Thus, often times ROUGE scores do not reflect the true quality of summaries and prevents multi-faceted evaluation of summaries (i.e. by topics, by overall content coverage and etc). In this paper, we introduce ROUGE 2.0, which has several updated measures of ROUGE: ROUGE-N+Synonyms, ROUGE-Topic, ROUGE-Topic+Synonyms, ROUGE-TopicUniq and ROUGE-TopicUniq+Synonyms; all of which are improvements over the core ROUGE measures."
                },
                {
                    "title": "Trust Region Policy Optimization",
                    "abstract": "In this article, we describe a method for optimizing control policies, with guaranteed monotonic improvement. By making several approximations to the theoretically-justified scheme, we develop a practical algorithm, called Trust Region Policy Optimization (TRPO). This algorithm is effective for optimizing large nonlinear policies such as neural networks. Our experiments demonstrate its robust performance on a wide variety of tasks: learning simulated robotic swimming, hopping, and walking gaits; and playing Atari games using images of the screen as input. Despite its approximations that deviate from the theory, TRPO tends to give monotonic improvement, with little tuning of hyperparameters."
                },
                {
                    "title": "TAMER: Training an Agent Manually via Evaluative Reinforcement",
                    "abstract": "Though computers have surpassed humans at many tasks, especially computationally intensive ones, there are many tasks for which human expertise remains necessary and/or useful. For such tasks, it is desirable for a human to be able to transmit knowledge to a learning agent as quickly and effortlessly as possible, and, ideally, without any knowledge of the details of the agentpsilas learning process. This paper proposes a general framework called Training an Agent Manually via Evaluative Reinforcement (TAMER) that allows a human to train a learning agent to perform a common class of complex tasks simply by giving scalar reward signals in response to the agentpsilas observed actions. Specifically, in sequential decision making tasks, an agent models the humanpsilas reward function and chooses actions that it predicts will receive the most reward. Our novel algorithm is fully implemented and tested on the game Tetris. Leveraging the human trainerspsila feedback, the agent learns to clear an average of more than 50 lines by its third game, an order of magnitude faster than the best autonomous learning agents."
                },
                {
                    "title": "METEOR: An Automatic Metric for MT Evaluation with Improved Correlation with Human Judgments",
                    "abstract": "We describe METEOR, an automatic metric for machine translation evaluation that is based on a generalized concept of unigram matching between the machineproduced translation and human-produced reference translations. Unigrams can be matched based on their surface forms, stemmed forms, and meanings; furthermore, METEOR can be easily extended to include more advanced matching strategies. Once all generalized unigram matches between the two strings have been found, METEOR computes a score for this matching using a combination of unigram-precision, unigram-recall, and a measure of fragmentation that is designed to directly capture how well-ordered the matched words in the machine translation are in relation to the reference. We evaluate METEOR by measuring the correlation between the metric scores and human judgments of translation quality. We compute the Pearson R correlation value between its scores and human quality assessments of the LDC TIDES 2003 Arabic-to-English and Chinese-to-English datasets. We perform segment-bysegment correlation, and show that METEOR gets an R correlation value of 0.347 on the Arabic data and 0.331 on the Chinese data. This is shown to be an improvement on using simply unigramprecision, unigram-recall and their harmonic F1 combination. We also perform experiments to show the relative contributions of the various mapping modules."
                },
                {
                    "title": "ROUGE: A Package for Automatic Evaluation of Summaries",
                    "abstract": "ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST."
                },
                {
                    "title": "Bleu: a Method for Automatic Evaluation of Machine Translation",
                    "abstract": "Human evaluations of machine translation are extensive but expensive. Human evaluations can take months to finish and involve human labor that can not be reused. We propose a method of automatic machine translation evaluation that is quick, inexpensive, and language-independent, that correlates highly with human evaluation, and that has little marginal cost per run. We present this method as an automated understudy to skilled human judges which substitutes for them when there is need for quick or frequent evaluations."
                },
                {
                    "title": "Constrained Markov Decision Processes",
                    "abstract": "INTRODUCTION Examples of Constrained Dynamic Control Problems On Solution Approaches for CMDPs with Expected Costs Other Types of CMDPs Cost Criteria and Assumptions The Convex Analytical Approach and Occupation Measures Linear Programming and Lagrangian Approach for CMDPs About the Methodology The Structure of the Book PART ONE: FINITE MDPS MARKOV DECISION PROCESSES The Model Cost Criteria and the Constrained Problem Some Notation The Dominance of Markov Policies THE DISCOUNTED COST Occupation Measure and the Primal LP Dynamic Programming and Dual LP: the Unconstrained Case Constrained Control: Lagrangian Approach The Dual LP Number of Randomizations THE EXPECTED AVERAGE COST Occupation Measure and the Primal LP Equivalent Linear Program The Dual Program Number of Randomizations FLOW AND SERVICE CONTROL IN A SINGLE-SERVER QUEUE The Model The Lagrangian The Original Constrained Problem Structure of Randomization and Implementation Issues On Coordination Between Controllers Open Questions PART TWO: INFINITE MDPS MDPS WITH INFINITE STATE AND ACTION SPACES The Model Cost Criteria Mixed Policies, and Topologic Structures The Dominance of Markov Policies Aggregation of States Extra Randomization in the Policies Equivalent Quasi-Markov Model and Quasi-Markov Policies THE TOTAL COST: CLASSIFICATION OF MDPS Transient and Absorbing MDPs MDPs With Uniform Lyapunov Functions Equivalence of MDP With Unbounded and bounded costs Properties of MDPs With Uniform Lyapunov Functions Properties for Fixed Initial Distribution Examples of Uniform Lyapunov Functions Contracting MDPs THE TOTAL COST: OCCUPATION MEASURES AND THE PRIMAL LP Occupation Measure Continuity of Occupation Measures More Properties of MDPs Characterization of Achievable Sets of Occupation Measure Relation Between Cost and Occupation Measure Dominating Classes of Policies Equivalent Linear Program The Dual Program THE TOTAL COST: DYNAMIC AND LINEAR PROGRAMMING Non-Constrained Control: Dynamic and Linear Programming Superharmonic Functions and Linear Programming Set of Achievable Costs Constrained Control: Lagrangian Approach The Dual LP State Truncation A Second LP Approach for Optimal Mixed Policies More on Unbound Costs THE DISCOUNTED COST The Equivalent Total Cost Model Occupation Measure and LP Non-negative Immediate Cost Weak Contracting Assumptions and Lyapunov Functions Example: Flow and Service Control THE EXPECTED AVERAGE COST Occupation Measures Completeness Properties of Stationary Policies Relation Between Cost and Occupation Measure Dominating Classes of Policies Equivalent Linear Program The Dual Program The Contracting Framework Other Conditions for the Uniform Integrability The Case of Uniform Lyapunov Conditions EXPECTED AVERAGE COST: DYNAMIC PROGRAMMING AND LP The Non-Constrained Case: Optimality Inequality Non-Constrained Control: Cost Bounded Below Dynamic Programming and Uniform Lyapunov Function Super-Harmonic Functions and Linear Programming Set of Achievable Costs Constrained Control: Lagrangian Approach The Dual LP A Second LP Approach for Optimal Mixed Policies PART THREE: ASYMPTOTIC METHODS AND APPROXIMATIONS SENSITIVITY ANALYSIS Introduction Approximation of the Values Approximation and Robustness of the Policies CONVERGENCE OF DISCOUNTED CONSTRAINED MDPS Convergence in the Discount Factor Convergence to the Expected Average Cost The Case of Uniform Lyapunov Function CONVERGENCE AS THE HORIZON TENDS TO INFINITY The Discounted Cost The Expected Average Cost: Stationary Policies The Expected Average Cost: General Policies STATE TRUNCATION AND APPROXIMATION The Approximating sets of States Scheme I: the Total Cost Scheme II: the Total Cost Scheme III: the Total Cost The Expected Average Cost Infinite MDPs: on the Number of Randomizations APPENDIX: CONVERGENCE OF PROBABILITY MEASURES REFERENCES LIST OF SYMBOLS AND NOTATION INDEX"
                },
                {
                    "title": "Algorithm 778: L-BFGS-B: Fortran subroutines for large-scale bound-constrained optimization",
                    "abstract": "L-BFGS-B is a limited-memory algorithm for solving large nonlinear optimization problems subject to simple bounds on the variables. It is intended for problems in which information on the Hessian matrix is difficult to obtain, or for large dense problems. L-BFGS-B can also be used for unconstrained problems and in this case performs similarly to its predessor, algorithm L-BFGS (Harwell routine VA15). The algorithm is implemented in Fortran 77."
                },
                {
                    "title": "Markov Decision Processes: Discrete Stochastic Dynamic Programming",
                    "abstract": "From the Publisher: \nThe past decade has seen considerable theoretical and applied research on Markov decision processes, as well as the growing use of these models in ecology, economics, communications engineering, and other fields where outcomes are uncertain and sequential decision-making processes are needed. A timely response to this increased activity, Martin L. Puterman's new work provides a uniquely up-to-date, unified, and rigorous treatment of the theoretical, computational, and applied research on Markov decision process models. It discusses all major research directions in the field, highlights many significant applications of Markov decision processes models, and explores numerous important topics that have previously been neglected or given cursory coverage in the literature. Markov Decision Processes focuses primarily on infinite horizon discrete time models and models with discrete time spaces while also examining models with arbitrary state spaces, finite horizon models, and continuous-time discrete state models. The book is organized around optimality criteria, using a common framework centered on the optimality (Bellman) equation for presenting results. The results are presented in a \"theorem-proof\" format and elaborated on through both discussion and examples, including results that are not available in any other book. A two-state Markov decision process model, presented in Chapter 3, is analyzed repeatedly throughout the book and demonstrates many results and algorithms. Markov Decision Processes covers recent research advances in such areas as countable state space models with average reward criterion, constrained models, and models with risk sensitive optimality criteria. It also explores several topics that have received little or no attention in other books, including modified policy iteration, multichain models with average reward criterion, and sensitive optimality. In addition, a Bibliographic Remarks section in each chapter comments on relevant historic"
                },
                {
                    "title": "Generalized Lagrange Multiplier Method for Solving Problems of Optimum Allocation of Resources",
                    "abstract": "The usefulness of Lagrange multipliers for optimization in the presence of constraints is not limited to differentiable functions. They can be applied to problems of maximizing an arbitrary real valued objective function over any set whatever, subject to bounds on the values of any other finite collection of real valued functions denned on the same set. While the use of the Lagrange multipliers does not guarantee that a solution will necessarily be found for all problems, it is \u201cfail-safe\u201d in the sense that any solution found by their use is a true solution. Since the method is so simple compared to other available methods it is often worth trying first, and succeeds in a surprising fraction of cases. They are particularly well suited to the solution of problems of allocating limited resources among a set of independent activities."
                },
                {
                    "title": "Deep Reinforcement Learning amidst Continual Structured Non-Stationarity",
                    "abstract": "As humans, our goals and our environment are persistently changing throughout our lifetime based on our experiences, actions, and internal and external drives. In contrast, typical reinforcement learning problem set-ups consider decision processes that are stationary across episodes. Can we develop reinforcement learning algorithms that can cope with the persistent change in the former, more realistic problem settings? While on-policy algorithms such as policy gradients in principle can be extended to non-stationary settings, the same cannot be said for more ef\ufb01cient off-policy algorithms that replay past experiences when learning. In this work, we formalize this problem setting, and draw upon ideas from the online learning and probabilistic inference literature to derive an off-policy RL algorithm that can reason about and tackle such lifelong non-stationarity. Our method leverages latent variable models to learn a representation of the environment from current and past experiences, and performs off-policy RL with this representation. We further introduce several simulation environments that exhibit lifelong non-stationarity, and empirically \ufb01nd that our approach substantially outperforms approaches that do not reason about environment shift."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Reinforcement",
                    "abstract": "Reinforcement occurs when a consequence to a behavior makes it more likely that behavior will happen again. \u2022 Simple example: Making eye contact when someone is talking to you. Saying \u201cGood\u201d when a child starts to tie her shoe. \u2022 Reinforcers may be tangible (e.g. stickers, food, games, money, access to friends, privileges) or intangible (praise, attention, showing interest, encouragement) \uf0b7 Social reinforcers usually work best!"
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                },
                {
                    "title": "A Simplex Method for Function Minimization",
                    "abstract": "A method is described for the minimization of a function of n variables, which depends on the comparison of function values at the (n 41) vertices of a general simplex, followed by the replacement of the vertex with the highest value by another point. The simplex adapts itself to the local landscape, and contracts on to the final minimum. The method is shown to be effective and computationally compact. A procedure is given for the estimation of the Hessian matrix in the neighbourhood of the minimum, needed in statistical estimation problems."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively incorporate non-stationarity into the reward function of reinforcement learning to enhance exploration and learning efficiency?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of non-stationary rewards in reinforcement learning (RL) is crucial as it can lead to more robust and adaptable RL agents capable of operating in dynamic environments. This advancement could significantly impact the research community by providing new methodologies for continual learning and exploration strategies, ultimately leading to practical applications in fields such as robotics, autonomous systems, and adaptive AI. Addressing this question could advance knowledge in understanding how agents can better manage uncertainty and variability in their environments, paving the way for more intelligent and flexible systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexities of non-stationarity, which can lead to oscillations in learning and instability in policy optimization. Naive approaches may fail because they do not account for the dynamic nature of the reward structure, leading to suboptimal exploration and exploitation strategies. Technical obstacles include the need for sophisticated algorithms that can adapt to changing reward signals, as well as theoretical challenges in ensuring convergence and stability in the learning process. Additionally, practical issues such as computational efficiency and the design of appropriate exploration bonuses complicate the implementation of effective solutions.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on static reward structures or has inadequately addressed the complexities introduced by non-stationarity. Limitations in existing solutions include a lack of robust methodologies for managing dynamic reward functions and insufficient exploration strategies that leverage epistemic uncertainty. Barriers such as the difficulty in modeling non-stationary environments and the absence of comprehensive frameworks for continual learning have hindered progress. My approach differs by integrating advanced regularization techniques and exploration bonuses that adaptively respond to the agent's uncertainty, thereby improving upon prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nMy proposed methodology involves developing a reinforcement learning framework that incorporates non-stationary rewards through the use of exploration bonuses based on epistemic uncertainty. I will utilize benchmark datasets relevant to continual learning scenarios and evaluate the performance using metrics such as average return, exploration efficiency, and convergence stability. The expected outcomes include improved learning rates and policy robustness in dynamic environments, demonstrating the effectiveness of non-stationary reward structures in enhancing RL agent performance."
            }
        },
        "author_data": {
            "b0414b72-68b1-4b28-85a9-71fd7661e412": {
                "pk": "b0414b72-68b1-4b28-85a9-71fd7661e412",
                "name": "Ted Moskovitz",
                "collaborators": [
                    "Aaditya K. Singh",
                    "Felix Hill",
                    "Stephanie C. Y. Chan",
                    "Andrew M. Saxe",
                    "Erin Grant",
                    "Samo Hromadka",
                    "Ahmed Touati",
                    "Diana Borsa",
                    "M. Sahani"
                ],
                "domain": [
                    "In-context Learning",
                    "Transformer Models",
                    "Reinforcement Learning",
                    "Mechanistic Interpretability"
                ],
                "publications": [
                    {
                        "title": "What needs to go right for an induction head? A mechanistic study of in-context learning circuits and their formation",
                        "abstract": "In-context learning is a powerful emergent ability in transformer models. Prior work in mechanistic interpretability has identified a circuit element that may be critical for in-context learning -- the induction head (IH), which performs a match-and-copy operation. During training of large transformers on natural language data, IHs emerge around the same time as a notable phase change in the loss. Despite the robust evidence for IHs and this interesting coincidence with the phase change, relatively little is known about the diversity and emergence dynamics of IHs. Why is there more than one IH, and how are they dependent on each other? Why do IHs appear all of a sudden, and what are the subcircuits that enable them to emerge? We answer these questions by studying IH emergence dynamics in a controlled setting by training on synthetic data. In doing so, we develop and share a novel optogenetics-inspired causal framework for modifying activations throughout training. Using this framework, we delineate the diverse and additive nature of IHs. By clamping subsets of activations throughout training, we then identify three underlying subcircuits that interact to drive IH formation, yielding the phase change. Furthermore, these subcircuits shed light on data-dependent properties of formation, such as phase change timing, already showing the promise of this more in-depth understanding of subcircuits that need to\"go right\"for an induction head."
                    },
                    {
                        "title": "The Transient Nature of Emergent In-Context Learning in Transformers",
                        "abstract": "Transformer neural networks can exhibit a surprising capacity for in-context learning (ICL) despite not being explicitly trained for it. Prior work has provided a deeper understanding of how ICL emerges in transformers, e.g. through the lens of mechanistic interpretability, Bayesian inference, or by examining the distributional properties of training data. However, in each of these cases, ICL is treated largely as a persistent phenomenon; namely, once ICL emerges, it is assumed to persist asymptotically. Here, we show that the emergence of ICL during transformer training is, in fact, often transient. We train transformers on synthetic data designed so that both ICL and in-weights learning (IWL) strategies can lead to correct predictions. We find that ICL first emerges, then disappears and gives way to IWL, all while the training loss decreases, indicating an asymptotic preference for IWL. The transient nature of ICL is observed in transformers across a range of model sizes and datasets, raising the question of how much to\"overtrain\"transformers when seeking compact, cheaper-to-run models. We find that L2 regularization may offer a path to more persistent ICL that removes the need for early stopping based on ICL-style validation tasks. Finally, we present initial evidence that ICL transience may be caused by competition between ICL and IWL circuits."
                    },
                    {
                        "title": "A State Representation for Diminishing Rewards",
                        "abstract": "A common setting in multitask reinforcement learning (RL) demands that an agent rapidly adapt to various stationary reward functions randomly sampled from a fixed distribution. In such situations, the successor representation (SR) is a popular framework which supports rapid policy evaluation by decoupling a policy's expected discounted, cumulative state occupancies from a specific reward function. However, in the natural world, sequential tasks are rarely independent, and instead reflect shifting priorities based on the availability and subjective perception of rewarding stimuli. Reflecting this disjunction, in this paper we study the phenomenon of diminishing marginal utility and introduce a novel state representation, the $\\lambda$ representation ($\\lambda$R) which, surprisingly, is required for policy evaluation in this setting and which generalizes the SR as well as several other state representations from the literature. We establish the $\\lambda$R's formal properties and examine its normative advantages in the context of machine learning, as well as its usefulness for studying natural behaviors, particularly foraging."
                    }
                ]
            },
            "1ac999ba-7413-4fdd-b0bd-f713819799cd": {
                "pk": "1ac999ba-7413-4fdd-b0bd-f713819799cd",
                "name": "Aaditya K. Singh",
                "collaborators": [
                    "Ted Moskovitz",
                    "Felix Hill",
                    "Stephanie C. Y. Chan",
                    "Andrew M. Saxe",
                    "Yu Yang",
                    "Kushal Tirumala",
                    "Mostafa Elhoushi",
                    "Ari S. Morcos",
                    "DJ Strouse",
                    "Erin Grant",
                    "Anas Mahmoud",
                    "Fabian Gloeckle",
                    "Baptiste Rozi\u00e8re",
                    "Carole-Jean Wu",
                    "Newsha Ardalani"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Code Generation",
                    "Data Pruning"
                ],
                "publications": [
                    {
                        "title": "Tokenization counts: the impact of tokenization on arithmetic in frontier LLMs",
                        "abstract": "Tokenization, the division of input text into input tokens, is an often overlooked aspect of the large language model (LLM) pipeline and could be the source of useful or harmful inductive biases. Historically, LLMs have relied on byte pair encoding, without care to specific input domains. With the increased use of LLMs for reasoning, various number-specific tokenization schemes have been adopted, with popular models like LLaMa and PaLM opting for single-digit tokenization while GPT-3.5 and GPT-4 have separate tokens for each 1-, 2-, and 3-digit numbers. In this work, we study the effect this choice has on numerical reasoning through the use of arithmetic tasks. We consider left-to-right and right-to-left tokenization for GPT-3.5 and -4, finding that right-to-left tokenization (enforced by comma separating numbers at inference time) leads to largely improved performance. Furthermore, we find that model errors when using standard left-to-right tokenization follow stereotyped error patterns, suggesting that model computations are systematic rather than approximate. We show that the model is able to convert between tokenizations easily, thus allowing chain-of-thought-inspired approaches to recover performance on left-to-right tokenized inputs. We also find the gap between tokenization directions decreases when models are scaled, possibly indicating that larger models are better able to override this tokenization-dependent inductive bias. In summary, our work performs the first study of how number tokenization choices lead to differences in model performance on arithmetic tasks, accompanied by a thorough analysis of error patterns. We hope this work inspires practitioners to more carefully ablate number tokenization-related choices when working towards general models of numerical reasoning."
                    },
                    {
                        "title": "What needs to go right for an induction head? A mechanistic study of in-context learning circuits and their formation",
                        "abstract": "In-context learning is a powerful emergent ability in transformer models. Prior work in mechanistic interpretability has identified a circuit element that may be critical for in-context learning -- the induction head (IH), which performs a match-and-copy operation. During training of large transformers on natural language data, IHs emerge around the same time as a notable phase change in the loss. Despite the robust evidence for IHs and this interesting coincidence with the phase change, relatively little is known about the diversity and emergence dynamics of IHs. Why is there more than one IH, and how are they dependent on each other? Why do IHs appear all of a sudden, and what are the subcircuits that enable them to emerge? We answer these questions by studying IH emergence dynamics in a controlled setting by training on synthetic data. In doing so, we develop and share a novel optogenetics-inspired causal framework for modifying activations throughout training. Using this framework, we delineate the diverse and additive nature of IHs. By clamping subsets of activations throughout training, we then identify three underlying subcircuits that interact to drive IH formation, yielding the phase change. Furthermore, these subcircuits shed light on data-dependent properties of formation, such as phase change timing, already showing the promise of this more in-depth understanding of subcircuits that need to\"go right\"for an induction head."
                    },
                    {
                        "title": "Brevity is the soul of wit: Pruning long files for code generation",
                        "abstract": "Data curation is commonly considered a\"secret-sauce\"for LLM training, with higher quality data usually leading to better LLM performance. Given the scale of internet-scraped corpora, data pruning has become a larger and larger focus. Specifically, many have shown that de-duplicating data, or sub-selecting higher quality data, can lead to efficiency or performance improvements. Generally, three types of methods are used to filter internet-scale corpora: embedding-based, heuristic-based, and classifier-based. In this work, we contrast the former two in the domain of finetuning LLMs for code generation. We find that embedding-based methods are often confounded by length, and that a simple heuristic--pruning long files--outperforms other methods in compute-limited regimes. Our method can yield up to a 2x efficiency benefit in training (while matching performance) or a 3.5% absolute performance improvement on HumanEval (while matching compute). However, we find that perplexity on held-out long files can increase, begging the question of whether optimizing data mixtures for common coding benchmarks (HumanEval, MBPP) actually best serves downstream use cases. Overall, we hope our work builds useful intuitions about code data (specifically, the low quality of extremely long code files) provides a compelling heuristic-based method for data pruning, and brings to light questions in how we evaluate code generation models."
                    },
                    {
                        "title": "The Transient Nature of Emergent In-Context Learning in Transformers",
                        "abstract": "Transformer neural networks can exhibit a surprising capacity for in-context learning (ICL) despite not being explicitly trained for it. Prior work has provided a deeper understanding of how ICL emerges in transformers, e.g. through the lens of mechanistic interpretability, Bayesian inference, or by examining the distributional properties of training data. However, in each of these cases, ICL is treated largely as a persistent phenomenon; namely, once ICL emerges, it is assumed to persist asymptotically. Here, we show that the emergence of ICL during transformer training is, in fact, often transient. We train transformers on synthetic data designed so that both ICL and in-weights learning (IWL) strategies can lead to correct predictions. We find that ICL first emerges, then disappears and gives way to IWL, all while the training loss decreases, indicating an asymptotic preference for IWL. The transient nature of ICL is observed in transformers across a range of model sizes and datasets, raising the question of how much to\"overtrain\"transformers when seeking compact, cheaper-to-run models. We find that L2 regularization may offer a path to more persistent ICL that removes the need for early stopping based on ICL-style validation tasks. Finally, we present initial evidence that ICL transience may be caused by competition between ICL and IWL circuits."
                    },
                    {
                        "title": "Decoding Data Quality via Synthetic Corruptions: Embedding-guided Pruning of Code Data",
                        "abstract": "Code datasets, often collected from diverse and uncontrolled sources such as GitHub, potentially suffer from quality issues, thereby affecting the performance and training efficiency of Large Language Models (LLMs) optimized for code generation. Previous studies demonstrated the benefit of using embedding spaces for data pruning, but they mainly focused on duplicate removal or increasing variety, and in other modalities, such as images. Our work focuses on using embeddings to identify and remove\"low-quality\"code data. First, we explore features of\"low-quality\"code in embedding space, through the use of synthetic corruptions. Armed with this knowledge, we devise novel pruning metrics that operate in embedding space to identify and remove low-quality entries in the Stack dataset. We demonstrate the benefits of this synthetic corruption informed pruning (SCIP) approach on the well-established HumanEval and MBPP benchmarks, outperforming existing embedding-based methods. Importantly, we achieve up to a 3% performance improvement over no pruning, thereby showing the promise of insights from synthetic corruptions for data pruning."
                    }
                ]
            },
            "16182ac2-254e-4baa-9476-4172902e5ab9": {
                "pk": "16182ac2-254e-4baa-9476-4172902e5ab9",
                "name": "DJ Strouse",
                "collaborators": [
                    "Aaditya K. Singh"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Tokenization",
                    "Large Language Models",
                    "Numerical Reasoning"
                ],
                "publications": [
                    {
                        "title": "Tokenization counts: the impact of tokenization on arithmetic in frontier LLMs",
                        "abstract": "Tokenization, the division of input text into input tokens, is an often overlooked aspect of the large language model (LLM) pipeline and could be the source of useful or harmful inductive biases. Historically, LLMs have relied on byte pair encoding, without care to specific input domains. With the increased use of LLMs for reasoning, various number-specific tokenization schemes have been adopted, with popular models like LLaMa and PaLM opting for single-digit tokenization while GPT-3.5 and GPT-4 have separate tokens for each 1-, 2-, and 3-digit numbers. In this work, we study the effect this choice has on numerical reasoning through the use of arithmetic tasks. We consider left-to-right and right-to-left tokenization for GPT-3.5 and -4, finding that right-to-left tokenization (enforced by comma separating numbers at inference time) leads to largely improved performance. Furthermore, we find that model errors when using standard left-to-right tokenization follow stereotyped error patterns, suggesting that model computations are systematic rather than approximate. We show that the model is able to convert between tokenizations easily, thus allowing chain-of-thought-inspired approaches to recover performance on left-to-right tokenized inputs. We also find the gap between tokenization directions decreases when models are scaled, possibly indicating that larger models are better able to override this tokenization-dependent inductive bias. In summary, our work performs the first study of how number tokenization choices lead to differences in model performance on arithmetic tasks, accompanied by a thorough analysis of error patterns. We hope this work inspires practitioners to more carefully ablate number tokenization-related choices when working towards general models of numerical reasoning."
                    }
                ]
            },
            "ac6bcc47-a3cd-43dc-8b65-439c67cbfadb": {
                "pk": "ac6bcc47-a3cd-43dc-8b65-439c67cbfadb",
                "name": "Tuomas Sandholm",
                "collaborators": [
                    "B. Zhang",
                    "S. McAleer",
                    "Carlos Martin",
                    "Ioannis Anagnostides",
                    "Gabriele Farina",
                    "Yongyuan Liang",
                    "Yanchao Sun",
                    "Ruijie Zheng",
                    "Xiangyu Liu",
                    "Furong Huang",
                    "Michael Curry",
                    "Sven Seuken",
                    "Siddharth Prasad",
                    "Maria-Florina Balcan",
                    "Vinzenz Thoma",
                    "Darshan Chakrabarti",
                    "Christian Kroer",
                    "Niao He",
                    "Fabian R. Pieroth",
                    "Constantinos Daskalakis",
                    "Noah Golowich",
                    "Emanuel Tewolde",
                    "Caspar Oesterheld",
                    "Manolis Zampetakis",
                    "Paul W. Goldberg",
                    "Vincent Conitzer",
                    "Ellen Vitercik",
                    "Paul Friedrich",
                    "Yulun Zhang",
                    "Ludwig Dierks",
                    "Jiaoyang Li",
                    "Naifeng Zhang",
                    "Benjamin Eysenbach"
                ],
                "domain": [
                    "Game Theory",
                    "Mechanism Design",
                    "Reinforcement Learning",
                    "Nash Equilibrium"
                ],
                "publications": [
                    {
                        "title": "Simultaneous incremental support adjustment and metagame solving: An equilibrium-finding framework for continuous-action games",
                        "abstract": "We present a framework for computing approximate mixed-strategy Nash equilibria of continuous-action games. It is a modification of the traditional double oracle algorithm, extended to multiple players and continuous action spaces. Unlike prior methods, it maintains fixed-cardinality pure strategy sets for each player. Thus, unlike prior methods, only a constant amount of memory is necessary. Furthermore, it does not require exact metagame solving on each iteration, which can be computationally expensive for large metagames. Moreover, it does not require global best-response computation on each iteration, which can be computationally expensive or even intractable for high-dimensional action spaces and general games. Our method incrementally reduces the exploitability of the strategy profile in the finite metagame, pushing it toward Nash equilibrium. Simultaneously, it incrementally improves the pure strategies that best respond to this strategy profile in the full game. We evaluate our method on various continuous-action games, showing that it obtains approximate mixed-strategy Nash equilibria with low exploitability."
                    },
                    {
                        "title": "Efficient $\\Phi$-Regret Minimization with Low-Degree Swap Deviations in Extensive-Form Games",
                        "abstract": "Recent breakthrough results by Dagan, Daskalakis, Fishelson and Golowich [2023] and Peng and Rubinstein [2023] established an efficient algorithm attaining at most $\\epsilon$ swap regret over extensive-form strategy spaces of dimension $N$ in $N^{\\tilde O(1/\\epsilon)}$ rounds. On the other extreme, Farina and Pipis [2023] developed an efficient algorithm for minimizing the weaker notion of linear-swap regret in $\\mathsf{poly}(N)/\\epsilon^2$ rounds. In this paper, we take a step toward bridging the gap between those two results. We introduce the set of $k$-mediator deviations, which generalize the untimed communication deviations recently introduced by Zhang, Farina and Sandholm [2024] to the case of having multiple mediators. We develop parameterized algorithms for minimizing the regret with respect to this set of deviations in $N^{O(k)}/\\epsilon^2$ rounds. This closes the gap in the sense that $k=1$ recovers linear swap regret, while $k=N$ recovers swap regret. Moreover, by relating $k$-mediator deviations to low-degree polynomials, we show that regret minimization against degree-$k$ polynomial swap deviations is achievable in $N^{O(kd)^3}/\\epsilon^2$ rounds, where $d$ is the depth of the game, assuming constant branching factor. For a fixed degree $k$, this is polynomial for Bayesian games and quasipolynomial more broadly when $d = \\mathsf{polylog} N$ -- the usual balancedness assumption on the game tree."
                    },
                    {
                        "title": "Exponential Lower Bounds on the Double Oracle Algorithm in Zero-Sum Games",
                        "abstract": "The double oracle algorithm is a popular method of solving games, because it is able to reduce computing equilibria to computing a series of best responses. However, its theoretical properties are not well understood. In this paper, we provide exponential lower bounds on the performance of the double oracle algorithm in both partially-observable stochastic games (POSGs) and extensive-form games (EFGs). Our results depend on what is assumed about the tiebreaking scheme---that is, which meta-Nash equilibrium or best response is chosen, in the event that there are multiple to pick from. In particular, for EFGs, our lower bounds require adversarial tiebreaking, whereas for POSGs, our lower bounds apply regardless of how ties are broken."
                    },
                    {
                        "title": "Game-Theoretic Robust Reinforcement Learning Handles Temporally-Coupled Perturbations",
                        "abstract": "Robust reinforcement learning (RL) seeks to train policies that can perform well under environment perturbations or adversarial attacks. Existing approaches typically assume that the space of possible perturbations remains the same across timesteps. However, in many settings, the space of possible perturbations at a given timestep depends on past perturbations. We formally introduce temporally-coupled perturbations, presenting a novel challenge for existing robust RL methods. To tackle this challenge, we propose GRAD, a novel game-theoretic approach that treats the temporally-coupled robust RL problem as a partially-observable two-player zero-sum game. By \ufb01nding an approximate equilibrium in this game, GRAD ensures the agent\u2019s robustness against temporally-coupled perturbations. Empirical experiments on a variety of continuous control tasks demonstrate that our proposed approach exhibits signi\ufb01cant robustness advantages compared to baselines against both standard and temporally-coupled attacks, in both state and action spaces."
                    },
                    {
                        "title": "Automated Design of Affine Maximizer Mechanisms in Dynamic Settings",
                        "abstract": "Dynamic mechanism design is a challenging extension to ordinary mechanism design in which the mechanism designer must make a sequence of decisions over time in the face of possibly untruthful reports of participating agents. Optimizing dynamic mechanisms for welfare is relatively well understood. However, there has been less work on optimizing for other goals (e.g., revenue), and without restrictive assumptions on valuations, it is remarkably challenging to characterize good mechanisms. Instead, we turn to automated mechanism design to find mechanisms with good performance in specific problem instances. We extend the class of affine maximizer mechanisms to MDPs where agents may untruthfully report their rewards. This extension results in a challenging bilevel optimization problem in which the upper problem involves choosing optimal mechanism parameters, and the lower problem involves solving the resulting MDP.  Our approach can find truthful dynamic mechanisms that achieve strong performance on goals other than welfare, and can be applied to essentially any problem setting---without restrictions on valuations---for which RL can learn optimal policies."
                    },
                    {
                        "title": "On the Outcome Equivalence of Extensive-Form and Behavioral Correlated Equilibria",
                        "abstract": "We investigate two notions of correlated equilibrium for extensive-form games: the extensive-form correlated equilibrium (EFCE) and the behavioral correlated equilibrium (BCE). We show that the two are outcome-equivalent, in the sense that every outcome distribution achievable under one notion is achievable under the other. Our result implies, to our knowledge, the first polynomial-time algorithm for computing a BCE."
                    },
                    {
                        "title": "Verifying Approximate Equilibrium in Auctions",
                        "abstract": "In practice, most auction mechanisms are not strategy-proof, so equilibrium analysis is required to predict bidding behavior. In many auctions, though, an exact equilibrium is not known and one would like to understand whether -- manually or computationally generated -- bidding strategies constitute an approximate equilibrium. We develop a framework and methods for estimating the distance of a strategy profile from equilibrium, based on samples from the prior and either bidding strategies or sample bids. We estimate an agent's utility gain from deviating to strategies from a constructed finite subset of the strategy space. We use PAC-learning to give error bounds, both for independent and interdependent prior distributions. The primary challenge is that one may miss large utility gains by considering only a finite subset of the strategy space. Our work differs from prior research in two critical ways. First, we explore the impact of bidding strategies on altering opponents' perceived prior distributions -- instead of assuming the other agents to bid truthfully. Second, we delve into reasoning with interdependent priors, where the type of one agent may imply a distinct distribution for other agents. Our main contribution lies in establishing sufficient conditions for strategy profiles and a closeness criterion for conditional distributions to ensure that utility gains estimated through our finite subset closely approximate the maximum gains. To our knowledge, ours is the first method to verify approximate equilibrium in any auctions beyond single-item ones. Also, ours is the first sample-based method for approximate equilibrium verification."
                    },
                    {
                        "title": "Convergence of $\\text{log}(1/\\epsilon)$ for Gradient-Based Algorithms in Zero-Sum Games without the Condition Number: A Smoothed Analysis",
                        "abstract": "Gradient-based algorithms have shown great promise in solving large (two-player) zero-sum games. However, their success has been mostly confined to the low-precision regime since the number of iterations grows polynomially in $1/\\epsilon$, where $\\epsilon>0$ is the duality gap. While it has been well-documented that linear convergence -- an iteration complexity scaling as $\\textsf{log}(1/\\epsilon)$ -- can be attained even with gradient-based algorithms, that comes at the cost of introducing a dependency on certain condition number-like quantities which can be exponentially large in the description of the game. To address this shortcoming, we examine the iteration complexity of several gradient-based algorithms in the celebrated framework of smoothed analysis, and we show that they have polynomial smoothed complexity, in that their number of iterations grows as a polynomial in the dimensions of the game, $\\textsf{log}(1/\\epsilon)$, and $1/\\sigma$, where $\\sigma$ measures the magnitude of the smoothing perturbation. Our result applies to optimistic gradient and extra-gradient descent/ascent, as well as a certain iterative variant of Nesterov's smoothing technique. From a technical standpoint, the proof proceeds by characterizing and performing a smoothed analysis of a certain error bound, the key ingredient driving linear convergence in zero-sum games. En route, our characterization also makes a natural connection between the convergence rate of such algorithms and perturbation-stability properties of the equilibrium, which is of interest beyond the model of smoothed complexity."
                    },
                    {
                        "title": "A Lower Bound on Swap Regret in Extensive-Form Games",
                        "abstract": "Recent simultaneous works by Peng and Rubinstein [2024] and Dagan et al. [2024] have demonstrated the existence of a no-swap-regret learning algorithm that can reach $\\epsilon$ average swap regret against an adversary in any extensive-form game within $m^{\\tilde{\\mathcal O}(1/\\epsilon)}$ rounds, where $m$ is the number of nodes in the game tree. However, the question of whether a $\\mathrm{poly}(m, 1/\\epsilon)$-round algorithm could exist remained open. In this paper, we show a lower bound that precludes the existence of such an algorithm. In particular, we show that achieving average swap regret $\\epsilon$ against an oblivious adversary in general extensive-form games requires at least $\\mathrm{exp}\\left(\\Omega\\left(\\min\\left\\{m^{1/14}, \\epsilon^{-1/6}\\right\\}\\right)\\right)$ rounds."
                    },
                    {
                        "title": "AlphaZeroES: Direct score maximization outperforms planning loss minimization",
                        "abstract": "Planning at execution time has been shown to dramatically improve performance for agents in both single-agent and multi-agent settings. A well-known family of approaches to planning at execution time are AlphaZero and its variants, which use Monte Carlo Tree Search together with a neural network that guides the search by predicting state values and action probabilities. AlphaZero trains these networks by minimizing a planning loss that makes the value prediction match the episode return, and the policy prediction at the root of the search tree match the output of the full tree expansion. AlphaZero has been applied to both single-agent environments (such as Sokoban) and multi-agent environments (such as chess and Go) with great success. In this paper, we explore an intriguing question: In single-agent environments, can we outperform AlphaZero by directly maximizing the episode score instead of minimizing this planning loss, while leaving the MCTS algorithm and neural architecture unchanged? To directly maximize the episode score, we use evolution strategies, a family of algorithms for zeroth-order blackbox optimization. Our experiments indicate that, across multiple environments, directly maximizing the episode score outperforms minimizing the planning loss."
                    },
                    {
                        "title": "Imperfect-Recall Games: Equilibrium Concepts and Their Complexity",
                        "abstract": "We investigate optimal decision making under imperfect recall, that is, when an agent forgets information it once held before. An example is the absentminded driver game, as well as team games in which the members have limited communication capabilities. In the framework of extensive-form games with imperfect recall, we analyze the computational complexities of finding equilibria in multiplayer settings across three different solution concepts: Nash, multiselves based on evidential decision theory (EDT), and multiselves based on causal decision theory (CDT). We are interested in both exact and approximate solution computation. As special cases, we consider (1) single-player games, (2) two-player zero-sum games and relationships to maximin values, and (3) games without exogenous stochasticity (chance nodes). We relate these problems to the complexity classes PPAD, PLS, \u03a3_2^P, \u2203R, and \u2203\u2200R."
                    },
                    {
                        "title": "New Sequence-Independent Lifting Techniques for Cutting Planes and When They Induce Facets",
                        "abstract": "Sequence-independent lifting is a procedure for strengthening valid inequalities of an integer program. We generalize the sequence-independent lifting method of Gu, Nemhauser, and Savelsbergh (GNS lifting) for cover inequalities and correct an error in their proposed generalization. We obtain a new sequence-independent lifting technique -- piecewise-constant (PC) lifting -- with a number of interesting properties. We derive a broad set of sufficient conditions under which PC lifting is facet defining. To our knowledge, this is the first characterization of facet-defining sequence-independent liftings that are efficiently computable from the underlying cover. Finally, we demonstrate via experiments that PC lifting can be a useful alternative to GNS lifting. We test our new lifting techniques atop a number of novel cover cut generation routines, which prove to be effective in experiments with CPLEX."
                    },
                    {
                        "title": "Scalable Mechanism Design for Multi-Agent Path Finding",
                        "abstract": "Multi-Agent Path Finding (MAPF) involves determining paths for multiple agents to travel simultaneously and collision-free through a shared area toward given goal locations. This problem is computationally complex, especially when dealing with large numbers of agents, as is common in realistic applications like autonomous vehicle coordination. Finding an optimal solution is often computationally infeasible, making the use of approximate, suboptimal algorithms essential. Adding to the complexity, agents might act in a self-interested and strategic way, possibly misrepresenting their goals to the MAPF algorithm if it benefits them. Although the field of mechanism design offers tools to align incentives, using these tools without careful consideration can fail when only having access to approximately optimal outcomes. In this work, we introduce the problem of scalable mechanism design for MAPF and propose three strategyproof mechanisms, two of which even use approximate MAPF algorithms. We test our mechanisms on realistic MAPF domains with problem sizes ranging from dozens to hundreds of agents. We find that they improve welfare beyond a simple baseline."
                    },
                    {
                        "title": "Faster Game Solving via Hyperparameter Schedules",
                        "abstract": "The counterfactual regret minimization (CFR) family of algorithms consists of iterative algorithms for imperfect-information games. In two-player zero-sum games, the time average of the iterates converges to a Nash equilibrium. The state-of-the-art prior variants, Discounted CFR (DCFR) and Predictive CFR$^+$ (PCFR$^+$) are the fastest known algorithms for solving two-player zero-sum games in practice, both in the extensive-form setting and the normal-form setting. They enhance the convergence rate compared to vanilla CFR by applying discounted weights to early iterations in various ways, leveraging fixed weighting schemes. We introduce Hyperparameter Schedules (HSs), which are remarkably simple yet highly effective in expediting the rate of convergence. HS dynamically adjusts the hyperparameter governing the discounting scheme of CFR variants. HSs on top of DCFR or PCFR$^+$ is now the new state of the art in solving zero-sum games and yields orders-of-magnitude speed improvements. The new algorithms are also easy to implement because 1) they are small modifications to the existing ones in terms of code and 2) they require no game-specific tuning."
                    },
                    {
                        "title": "Game-Theoretic Robust Reinforcement Learning Handles Temporally-Coupled Perturbations",
                        "abstract": "Deploying reinforcement learning (RL) systems requires robustness to uncertainty and model misspecification, yet prior robust RL methods typically only study noise introduced independently across time. However, practical sources of uncertainty are usually coupled across time. We formally introduce temporally-coupled perturbations, presenting a novel challenge for existing robust RL methods. To tackle this challenge, we propose GRAD, a novel game-theoretic approach that treats the temporally-coupled robust RL problem as a partially observable two-player zero-sum game. By finding an approximate equilibrium within this game, GRAD optimizes for general robustness against temporally-coupled perturbations. Experiments on continuous control tasks demonstrate that, compared with prior methods, our approach achieves a higher degree of robustness to various types of attacks on different attack domains, both in settings with temporally-coupled perturbations and decoupled perturbations."
                    },
                    {
                        "title": "Bicriteria Multidimensional Mechanism Design with Side Information",
                        "abstract": "We develop a versatile methodology for multidimensional mechanism design that incorporates side information about agents to generate high welfare and high revenue simultaneously. Side information sources include advice from domain experts, predictions from machine learning models, and even the mechanism designer's gut instinct. We design a tunable mechanism that integrates side information with an improved VCG-like mechanism based on weakest types, which are agent types that generate the least welfare. We show that our mechanism, when carefully tuned, generates welfare and revenue competitive with the prior-free total social surplus, and its performance decays gracefully as the side information quality decreases. We consider a number of side information formats including distribution-free predictions, predictions that express uncertainty, agent types constrained to low-dimensional subspaces of the ambient type space, and the traditional setting with known priors over agent types. In each setting we design mechanisms based on weakest types and prove performance guarantees."
                    }
                ]
            },
            "e8ae7f12-96d9-4d78-8667-632b3a161127": {
                "pk": "e8ae7f12-96d9-4d78-8667-632b3a161127",
                "name": "Ruslan Salakhutdinov",
                "collaborators": [
                    "Yonatan Bisk",
                    "So Yeon Min",
                    "Yue Wu",
                    "Shrimai Prabhumoye",
                    "Yuanzhi Li",
                    "Tom M. Mitchell",
                    "Murtaza Dalal",
                    "Deepak Pathak",
                    "Yewen Fan",
                    "Stephen McAleer",
                    "Devendra Singh Chaplot",
                    "A. Azaria",
                    "Wenlong Huang",
                    "Fei Xia",
                    "Brian Ichter",
                    "Gianna Bressand",
                    "Guil-laume Lengyel",
                    "Min Liu",
                    "Walter Talbott",
                    "Chen Chen",
                    "Quanting Xie",
                    "Tianyi Zhang",
                    "Aarav Bajaj",
                    "Matthew Johnson-Roberson",
                    "Xavi Puig",
                    "D. Chaplot",
                    "Tsung-Yen Yang",
                    "Akshara Rai",
                    "Priyam Parashar",
                    "Roozbeh Mottaghi",
                    "Jiahui Yang",
                    "Russell Mendonca",
                    "Youssef Khaky",
                    "Tarun Chiruvolu",
                    "Yeon Min",
                    "Pradeep Ravikumar",
                    "Yong Cheng",
                    "Lu Jiang",
                    "Wolfgang Macherey",
                    "Chuan Guo",
                    "Geoff Pleiss",
                    "Yu Sun",
                    "Kilian Q. Wein-515",
                    "Iris Hendrickx",
                    "Su Nam Kim",
                    "Zornitsa Kozareva",
                    "Preslav Nakov",
                    "Diarmuid \u00d3 S\u00e9aghdha",
                    "Marco Pad\u00f3",
                    "Lorenza Pennacchiotti",
                    "Romano",
                    "Zhiting Hu",
                    "Bowen Tan",
                    "Zhengjie Miao",
                    "Yuliang Li",
                    "Xiaolan Wang",
                    "Wang-534 Chiew",
                    "Zihao Wang",
                    "Shaofei Cai",
                    "Guanzhou Chen",
                    "Anji Liu",
                    "Xiaojian Ma",
                    "Yitao Liang",
                    "Jean-Baptiste Alayrac",
                    "Jeff Donahue",
                    "Pauline Luc",
                    "Antoine Miech",
                    "Iain Barr",
                    "Yana Hasson",
                    "Karel Lenc",
                    "A. Mensch",
                    "Katie Millican",
                    "Malcolm Reynolds",
                    "Tom Brown",
                    "Benjamin Mann",
                    "Nick Ryder",
                    "Melanie Subbiah",
                    "Jared Kaplan",
                    "Prafulla Dhariwal",
                    "Arvind Neelakantan",
                    "Pranav Shyam",
                    "Girish Sastry",
                    "Anthony Brohan",
                    "Yevgen Chebotar",
                    "Chelsea Finn",
                    "Karol Hausman",
                    "Alex Herzog",
                    "Daniel E. Ho",
                    "Julian Ibarz",
                    "A. Irpan",
                    "Eric Jang",
                    "Jacky Liang",
                    "Peng Xu",
                    "Pete Florence",
                    "Andy Zeng",
                    "Ted Xiao",
                    "Harris Chan",
                    "Jonathan Tompson",
                    "Igor Mordatch",
                    "William H. Guss"
                ],
                "domain": [
                    "Large Language Models",
                    "Robotics",
                    "Embodied AI",
                    "Motion Planning"
                ],
                "publications": [
                    {
                        "title": "AgentKit: Flow Engineering with Graphs, not Coding",
                        "abstract": "We propose an intuitive LLM prompting framework (AgentKit) for multifunctional agents. AgentKit offers a unified framework for explicitly constructing a complex \u201cthought process\u201d from simple natural language prompts. The basic building block in AgentKit is a node , containing a natural language prompt for a specific subtask. The user then puts together chains of nodes, like stacking LEGO pieces. The chains of nodes can be designed to explicitly enforce a naturally structured \u201cthought process\u201d. For example, for the task of writing a paper, one may start with the thought process of 1) identify a core message, 2) identify prior research gaps, etc. The nodes in AgentKit can be designed and combined in different ways to implement multiple advanced capabilities including on-the-fly hierarchical planning, reflection, and learning from interactions. In addition, due to the modular nature and the intuitive design to simulate explicit human thought process, a basic agent could be implemented as simple as a list of prompts for the subtasks and therefore could be designed and tuned by someone without any programming experience . Quantitatively, we show that agents designed through AgentKit achieve SOTA performance on WebShop and Crafter. These advances underscore AgentKit\u2019s potential in making LLM agents effective and accessible for a wider range of applications. Github"
                    },
                    {
                        "title": "Local Policies Enable Zero-shot Long-horizon Manipulation",
                        "abstract": "Sim2real for robotic manipulation is difficult due to the challenges of simulating complex contacts and generating realistic task distributions. To tackle the latter problem, we introduce ManipGen, which leverages a new class of policies for sim2real transfer: local policies. Locality enables a variety of appealing properties including invariances to absolute robot and object pose, skill ordering, and global scene configuration. We combine these policies with foundation models for vision, language and motion planning and demonstrate SOTA zero-shot performance of our method to Robosuite benchmark tasks in simulation (97%). We transfer our local policies from simulation to reality and observe they can solve unseen long-horizon manipulation tasks with up to 8 stages with significant pose, object and scene configuration variation. ManipGen outperforms SOTA approaches such as SayCan, OpenVLA, LLMTrajGen and VoxPoser across 50 real-world manipulation tasks by 36%, 76%, 62% and 60% respectively. Video results at https://mihdalal.github.io/manipgen/"
                    },
                    {
                        "title": "Embodied-RAG: General non-parametric Embodied Memory for Retrieval and Generation",
                        "abstract": "There is no limit to how much a robot might explore and learn, but all of that knowledge needs to be searchable and actionable. Within language research, retrieval augmented generation (RAG) has become the workhouse of large-scale non-parametric knowledge, however existing techniques do not directly transfer to the embodied domain, which is multimodal, data is highly correlated, and perception requires abstraction. To address these challenges, we introduce Embodied-RAG, a framework that enhances the foundational model of an embodied agent with a non-parametric memory system capable of autonomously constructing hierarchical knowledge for both navigation and language generation. Embodied-RAG handles a full range of spatial and semantic resolutions across diverse environments and query types, whether for a specific object or a holistic description of ambiance. At its core, Embodied-RAG's memory is structured as a semantic forest, storing language descriptions at varying levels of detail. This hierarchical organization allows the system to efficiently generate context-sensitive outputs across different robotic platforms. We demonstrate that Embodied-RAG effectively bridges RAG to the robotics domain, successfully handling over 200 explanation and navigation queries across 19 environments, highlighting its promise for general-purpose non-parametric system for embodied agents."
                    },
                    {
                        "title": "AgentKit: Structured LLM Reasoning with Dynamic Graphs",
                        "abstract": "We propose an intuitive LLM prompting framework (AgentKit) for multifunctional agents. AgentKit offers a unified framework for explicitly constructing a complex\"thought process\"from simple natural language prompts. The basic building block in AgentKit is a node, containing a natural language prompt for a specific subtask. The user then puts together chains of nodes, like stacking LEGO pieces. The chains of nodes can be designed to explicitly enforce a naturally structured\"thought process\". For example, for the task of writing a paper, one may start with the thought process of 1) identify a core message, 2) identify prior research gaps, etc. The nodes in AgentKit can be designed and combined in different ways to implement multiple advanced capabilities including on-the-fly hierarchical planning, reflection, and learning from interactions. In addition, due to the modular nature and the intuitive design to simulate explicit human thought process, a basic agent could be implemented as simple as a list of prompts for the subtasks and therefore could be designed and tuned by someone without any programming experience. Quantitatively, we show that agents designed through AgentKit achieve SOTA performance on WebShop and Crafter. These advances underscore AgentKit's potential in making LLM agents effective and accessible for a wider range of applications. https://github.com/holmeswww/AgentKit"
                    },
                    {
                        "title": "Situated Instruction Following",
                        "abstract": "Language is never spoken in a vacuum. It is expressed, comprehended, and contextualized within the holistic backdrop of the speaker's history, actions, and environment. Since humans are used to communicating efficiently with situated language, the practicality of robotic assistants hinge on their ability to understand and act upon implicit and situated instructions. In traditional instruction following paradigms, the agent acts alone in an empty house, leading to language use that is both simplified and artificially\"complete.\"In contrast, we propose situated instruction following, which embraces the inherent underspecification and ambiguity of real-world communication with the physical presence of a human speaker. The meaning of situated instructions naturally unfold through the past actions and the expected future behaviors of the human involved. Specifically, within our settings we have instructions that (1) are ambiguously specified, (2) have temporally evolving intent, (3) can be interpreted more precisely with the agent's dynamic actions. Our experiments indicate that state-of-the-art Embodied Instruction Following (EIF) models lack holistic understanding of situated human intention."
                    },
                    {
                        "title": "Neural MP: A Generalist Neural Motion Planner",
                        "abstract": "The current paradigm for motion planning generates solutions from scratch for every new problem, which consumes significant amounts of time and computational resources. For complex, cluttered scenes, motion planning approaches can often take minutes to produce a solution, while humans are able to accurately and safely reach any goal in seconds by leveraging their prior experience. We seek to do the same by applying data-driven learning at scale to the problem of motion planning. Our approach builds a large number of complex scenes in simulation, collects expert data from a motion planner, then distills it into a reactive generalist policy. We then combine this with lightweight optimization to obtain a safe path for real world deployment. We perform a thorough evaluation of our method on 64 motion planning tasks across four diverse environments with randomized poses, scenes and obstacles, in the real world, demonstrating an improvement of 23%, 17% and 79% motion planning success rate over state of the art sampling, optimization and learning based planning methods. Video results available at mihdalal.github.io/neuralmotionplanner"
                    },
                    {
                        "title": "Plan-Seq-Learn: Language Model Guided RL for Solving Long Horizon Robotics Tasks",
                        "abstract": "Large Language Models (LLMs) have been shown to be capable of performing high-level planning for long-horizon robotics tasks, yet existing methods require access to a pre-defined skill library (e.g. picking, placing, pulling, pushing, navigating). However, LLM planning does not address how to design or learn those behaviors, which remains challenging particularly in long-horizon settings. Furthermore, for many tasks of interest, the robot needs to be able to adjust its behavior in a fine-grained manner, requiring the agent to be capable of modifying low-level control actions. Can we instead use the internet-scale knowledge from LLMs for high-level policies, guiding reinforcement learning (RL) policies to efficiently solve robotic control tasks online without requiring a pre-determined set of skills? In this paper, we propose Plan-Seq-Learn (PSL): a modular approach that uses motion planning to bridge the gap between abstract language and learned low-level control for solving long-horizon robotics tasks from scratch. We demonstrate that PSL achieves state-of-the-art results on over 25 challenging robotics tasks with up to 10 stages. PSL solves long-horizon tasks from raw visual input spanning four benchmarks at success rates of over 85%, out-performing language-based, classical, and end-to-end approaches. Video results and code at https://mihdalal.github.io/planseqlearn/"
                    },
                    {
                        "title": "SPRING: GPT-4 Out-performs RL Algorithms by Studying Papers and Reasoning",
                        "abstract": ","
                    },
                    {
                        "title": "Overleaf Example",
                        "abstract": "Recent methods for embodied instruction following are typically trained end-toend using imitation learning. This requires the use of expert trajectories and lowlevel language instructions. Such approaches assume learned hidden states will simultaneously integrate semantics from the language and vision to perform state tracking, spatial memory, exploration, and long-term planning. In contrast, we propose a modular method with structured representations that (1) builds a semantic map of the scene, and (2) performs exploration with a semantic search policy, to achieve the natural language goal. Our modular method achieves SOTA performance (24.46%) with a substantial (8.17 % absolute) gap from previous work while using less data by eschewing both expert trajectories and low-level instructions. Leveraging low-level language, however, can further increase our performance (26.49%).1 Our findings suggest that an explicit spatial memory and a semantic search policy can provide a stronger and more general representation for state-tracking and guidance, even in the absence of expert trajectories or lowlevel instructions.2"
                    },
                    {
                        "title": "SegMix: A Simple Structure-Aware Data Augmentation Method",
                        "abstract": "Many Natural Language Processing tasks in-001 volve predicting structures, such as Syntax 002 Parsing and Relation Extraction (RE). One 003 central challenge in supervised structured pre-004 diction is the lack of high-quality annotated 005 data. The recently proposed interpolation-006 based data augmentation (DA) algorithms (i.e. 007 mixup ) augment the training set via making 008 convex interpolation between training data 009 points (Zhang et al., 2018). However, current 010 algorithms (e.g. SeqMix (Zhang et al., 2020), 011 LADA (Chen et al., 2020a)) that apply mixup 012 to language structured prediction tasks are not 013 aware of the syntactic or output structures of 014 the tasks, making their performance unstable 015 and requiring additional heuristic constraints. 016 Furthermore, SeqMix-like algorithms expect a 017 linear encoding scheme of the output structure, 018 such as BIO-Scheme for Named Entity Recog-019 nition (NER), restricting its applicability. 020 To this end, we propose SegMix , a sim-021 ple framework of interpolation-based algo-022 rithms that can adapt to both the syntactic 023 and output structures, making it robust to 024 hyper-parameters and applicable to different 025 tasks. We empirically show that SegMix con-026 sistently improves performance over several 027 strong baseline models on two structured pre-028 diction tasks (NER and RE). SegMix is a 029 \ufb02exible framework that uni\ufb01es existing rule-030 based language DA methods, creating interest-031 ing mixtures of DA techniques. Furthermore, 032 the method is easy to implement and adds neg-033 ligible overhead to training and inference. 034"
                    },
                    {
                        "title": "Describe, Explain, Plan and Select: Interactive Planning with Large Language Models Enables Open-World Multi-Task Agents",
                        "abstract": ","
                    },
                    {
                        "title": "What\u2019s the Plan? Evaluating and Developing Planning-Aware Techniques for LLMs",
                        "abstract": "Planning is a fundamental task in artificial intelligence that involves finding a sequence of actions that achieve a specified goal in a given environment. Large language models (LLMs) are increasingly used for applications that require planning capabilities, such as web or embodied agents. In line with recent studies, we demonstrate through experimentation that LLMs lack necessary skills required for planning. Based on these observations, we advocate for the potential of a hybrid approach that combines LLMs with classical planning methodology. Then, we introduce SimPlan , a novel hybrid-method, and evaluate its performance in a new challenging setup. Our extensive experiments across various planning domains demonstrate that SimPlan significantly outperforms existing LLM-based planners."
                    },
                    {
                        "title": "P ROPRIOCEPTIVE S PATIAL R EPRESENTATIONS F OR G ENERALIZED L OCOMOTION",
                        "abstract": ". Videos"
                    }
                ]
            },
            "d25d5c77-cb32-41a8-ad5f-6a841c6f8dad": {
                "pk": "d25d5c77-cb32-41a8-ad5f-6a841c6f8dad",
                "name": "Anca D. Dragan",
                "collaborators": [
                    "W. B. Knox",
                    "Stephane Hatgis-Kessell",
                    "S. O. Adalgeirsson",
                    "Serena Booth",
                    "Peter Stone",
                    "S. Niekum"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Human Feedback",
                    "Preference Learning",
                    "Regret Theory"
                ],
                "publications": [
                    {
                        "title": "Learning Optimal Advantage from Preferences and Mistaking it for Reward",
                        "abstract": "We consider algorithms for learning reward functions from human preferences over pairs of trajectory segments, as used in reinforcement learning from human feedback (RLHF). Most recent work assumes that human preferences are generated based only upon the reward accrued within those segments, or their partial return. Recent work casts doubt on the validity of this assumption, proposing an alternative preference model based upon regret. We investigate the consequences of assuming preferences are based upon partial return when they actually arise from regret. We argue that the learned function is an approximation of the optimal advantage function, not a reward function. We find that if a specific pitfall is addressed, this incorrect assumption is not particularly harmful, resulting in a highly shaped reward function. Nonetheless, this incorrect usage of the approximation of the optimal advantage function is less desirable than the appropriate and simpler approach of greedy maximization of it. From the perspective of the regret preference model, we also provide a clearer interpretation of fine tuning contemporary large language models with RLHF. This paper overall provides insight regarding why learning under the partial return preference model tends to work so well in practice, despite it conforming poorly to how humans give preferences."
                    }
                ]
            },
            "1c3beaef-9a13-4621-bab9-5275d5e10a0c": {
                "pk": "1c3beaef-9a13-4621-bab9-5275d5e10a0c",
                "name": "Stephen McAleer",
                "collaborators": [
                    "T. Sandholm",
                    "Michael Curry",
                    "Sven Seuken",
                    "Gabriele Farina",
                    "Vinzenz Thoma",
                    "Darshan Chakrabarti",
                    "Christian Kroer",
                    "Niao He",
                    "Paul Friedrich",
                    "Yulun Zhang",
                    "Ludwig Dierks",
                    "Jiaoyang Li",
                    "Naifeng Zhang",
                    "Yongyuan Liang",
                    "Yanchao Sun",
                    "Ruijie Zheng",
                    "Xiangyu Liu",
                    "Benjamin Eysenbach",
                    "Furong Huang",
                    "Gaoyue Zhou",
                    "Mingzhi Wang",
                    "Yaodong Yang",
                    "Zhangir Azerbayev",
                    "Hailey Schoelkopf",
                    "Keiran Paster",
                    "Marco Dos Santos",
                    "Albert Q. Jiang",
                    "Jia Deng",
                    "Stella Biderman",
                    "S. Welleck",
                    "B. Zhang",
                    "Ioannis Anagnostides",
                    "Federico Cacciamani",
                    "Andreas Alexander Haupt",
                    "A. Celli",
                    "Nicola Gatti",
                    "Vincent Conitzer"
                ],
                "domain": [
                    "Mechanism Design",
                    "Reinforcement Learning",
                    "Game Theory",
                    "Multi-Agent Systems"
                ],
                "publications": [
                    {
                        "title": "Automated Design of Affine Maximizer Mechanisms in Dynamic Settings",
                        "abstract": "Dynamic mechanism design is a challenging extension to ordinary mechanism design in which the mechanism designer must make a sequence of decisions over time in the face of possibly untruthful reports of participating agents. Optimizing dynamic mechanisms for welfare is relatively well understood. However, there has been less work on optimizing for other goals (e.g., revenue), and without restrictive assumptions on valuations, it is remarkably challenging to characterize good mechanisms. Instead, we turn to automated mechanism design to find mechanisms with good performance in specific problem instances. We extend the class of affine maximizer mechanisms to MDPs where agents may untruthfully report their rewards. This extension results in a challenging bilevel optimization problem in which the upper problem involves choosing optimal mechanism parameters, and the lower problem involves solving the resulting MDP.  Our approach can find truthful dynamic mechanisms that achieve strong performance on goals other than welfare, and can be applied to essentially any problem setting---without restrictions on valuations---for which RL can learn optimal policies."
                    },
                    {
                        "title": "Scalable Mechanism Design for Multi-Agent Path Finding",
                        "abstract": "Multi-Agent Path Finding (MAPF) involves determining paths for multiple agents to travel simultaneously and collision-free through a shared area toward given goal locations. This problem is computationally complex, especially when dealing with large numbers of agents, as is common in realistic applications like autonomous vehicle coordination. Finding an optimal solution is often computationally infeasible, making the use of approximate, suboptimal algorithms essential. Adding to the complexity, agents might act in a self-interested and strategic way, possibly misrepresenting their goals to the MAPF algorithm if it benefits them. Although the field of mechanism design offers tools to align incentives, using these tools without careful consideration can fail when only having access to approximately optimal outcomes. In this work, we introduce the problem of scalable mechanism design for MAPF and propose three strategyproof mechanisms, two of which even use approximate MAPF algorithms. We test our mechanisms on realistic MAPF domains with problem sizes ranging from dozens to hundreds of agents. We find that they improve welfare beyond a simple baseline."
                    },
                    {
                        "title": "Faster Game Solving via Hyperparameter Schedules",
                        "abstract": "The counterfactual regret minimization (CFR) family of algorithms consists of iterative algorithms for imperfect-information games. In two-player zero-sum games, the time average of the iterates converges to a Nash equilibrium. The state-of-the-art prior variants, Discounted CFR (DCFR) and Predictive CFR$^+$ (PCFR$^+$) are the fastest known algorithms for solving two-player zero-sum games in practice, both in the extensive-form setting and the normal-form setting. They enhance the convergence rate compared to vanilla CFR by applying discounted weights to early iterations in various ways, leveraging fixed weighting schemes. We introduce Hyperparameter Schedules (HSs), which are remarkably simple yet highly effective in expediting the rate of convergence. HS dynamically adjusts the hyperparameter governing the discounting scheme of CFR variants. HSs on top of DCFR or PCFR$^+$ is now the new state of the art in solving zero-sum games and yields orders-of-magnitude speed improvements. The new algorithms are also easy to implement because 1) they are small modifications to the existing ones in terms of code and 2) they require no game-specific tuning."
                    },
                    {
                        "title": "Game-Theoretic Robust Reinforcement Learning Handles Temporally-Coupled Perturbations",
                        "abstract": "Deploying reinforcement learning (RL) systems requires robustness to uncertainty and model misspecification, yet prior robust RL methods typically only study noise introduced independently across time. However, practical sources of uncertainty are usually coupled across time. We formally introduce temporally-coupled perturbations, presenting a novel challenge for existing robust RL methods. To tackle this challenge, we propose GRAD, a novel game-theoretic approach that treats the temporally-coupled robust RL problem as a partially observable two-player zero-sum game. By finding an approximate equilibrium within this game, GRAD optimizes for general robustness against temporally-coupled perturbations. Experiments on continuous control tasks demonstrate that, compared with prior methods, our approach achieves a higher degree of robustness to various types of attacks on different attack domains, both in settings with temporally-coupled perturbations and decoupled perturbations."
                    },
                    {
                        "title": "Team-PSRO for Learning Approximate TMECor in Large Team Games via Cooperative Reinforcement Learning",
                        "abstract": "Recent algorithms have achieved superhuman performance at a number of two-player zero-sum games such as poker and go. However, many real-world situations are multi-player games. Zero-sum two-team games, such as bridge and football, involve two teams where each member of the team shares the same reward with every other member of that team, and each team has the negative of the reward of the other team. A popular solution concept in this setting, called TMECor, assumes that teams can jointly correlate their strategies before play, but are not able to communicate during play. This setting is harder than two-player zero-sum games because each player on a team has different information and must use their public actions to signal to other members of the team. Prior works either have game-theoretic guarantees but only work in very small games, or are able to scale to large games but do not have game-theoretic guarantees. In this paper we introduce two algorithms: Team-PSRO, an extension of PSRO from two-player games to team games, and Team-PSRO Mix-and-Match which improves upon Team PSRO by better using population policies. In Team-PSRO, in every iteration both teams learn a joint best response to the opponent\u2019s meta-strategy via reinforcement learning. As the reinforcement learning joint best response approaches the optimal best response, Team-PSRO is guaranteed to converge to a TMECor. In experiments on Kuhn poker and Liar\u2019s Dice, we show that a tabular version of Team-PSRO converges to TMECor, and a version of Team PSRO using deep cooperative reinforcement learning beats self-play reinforcement learning in the large game of Google Research Football."
                    },
                    {
                        "title": "Llemma: An Open Language Model For Mathematics",
                        "abstract": "We present Llemma, a large language model for mathematics. We continue pretraining Code Llama on the Proof-Pile-2, a mixture of scientific papers, web data containing mathematics, and mathematical code, yielding Llemma. On the MATH benchmark Llemma outperforms all known open base models, as well as the unreleased Minerva model suite on an equi-parameter basis. Moreover, Llemma is capable of tool use and formal theorem proving without any further finetuning. We openly release all artifacts, including 7 billion and 34 billion parameter models, the Proof-Pile-2, and code to replicate our experiments."
                    },
                    {
                        "title": "Steering No-Regret Learners to a Desired Equilibrium",
                        "abstract": "A mediator observes no-regret learners playing an extensive-form game repeatedly across $T$ rounds. The mediator attempts to steer players toward some desirable predetermined equilibrium by giving (nonnegative) payments to players. We call this the steering problem. The steering problem captures problems several problems of interest, among them equilibrium selection and information design (persuasion). If the mediator's budget is unbounded, steering is trivial because the mediator can simply pay the players to play desirable actions. We study two bounds on the mediator's payments: a total budget and a per-round budget. If the mediator's total budget does not grow with $T$, we show that steering is impossible. However, we show that it is enough for the total budget to grow sublinearly with $T$, that is, for the average payment to vanish. When players' full strategies are observed at each round, we show that constant per-round budgets permit steering. In the more challenging setting where only trajectories through the game tree are observable, we show that steering is impossible with constant per-round budgets in general extensive-form games, but possible in normal-form games or if the per-round budget may itself depend on $T$. We also show how our results can be generalized to the case when the equilibrium is being computed online while steering is happening. We supplement our theoretical positive results with experiments highlighting the efficacy of steering in large games."
                    }
                ]
            }
        }
    },
    "2405.15719": {
        "paper_data": {
            "title": "Hierarchical Uncertainty Exploration via Feedforward Posterior Trees",
            "url": "http://arxiv.org/abs/2405.15719v1",
            "arxiv_id": "2405.15719",
            "authors": [
                "Elias Nehme",
                "Rotem Mulayoff",
                "Tomer Michaeli"
            ],
            "abstract": "When solving ill-posed inverse problems, one often desires to explore the space of potential solutions rather than be presented with a single plausible reconstruction. Valuable insights into these feasible solutions and their associated probabilities are embedded in the posterior distribution. However, when confronted with data of high dimensionality (such as images), visualizing this distribution becomes a formidable challenge, necessitating the application of effective summarization techniques before user examination. In this work, we introduce a new approach for visualizing posteriors across multiple levels of granularity using tree-valued predictions. Our method predicts a tree-valued hierarchical summarization of the posterior distribution for any input measurement, in a single forward pass of a neural network. We showcase the efficacy of our approach across diverse datasets and image restoration challenges, highlighting its prowess in uncertainty quantification and visualization. Our findings reveal that our method performs comparably to a baseline that hierarchically clusters samples from a diffusion-based posterior sampler, yet achieves this with orders of magnitude greater speed.",
            "introduction": " Introduction Communicating prediction uncertainty is a crucial element in advancing the reliability of machine learning models. This is particularly important in the realm of imaging inverse problems, in which a given input can typically be associated with a multitude of plausible solutions. In such cases, it is advantageous to equip users with efficient tools for exploring and visualizing the set of admissible solutions. Such tools may be of high value especially in safety-critical domains, such as scientific and medical image analysis [1\u20134], where inaccuracies in predictions could impact human lives. Information about the plausible solutions and their respective likelihoods is encapsulated within the posterior distribution. However, high-dimensional posteriors are challenging to visualize. One popular approach for communicating uncertainty is to generate samples from the posterior [ 5\u201314]. Yet, in complex domains with high uncertainty levels, users may need to sift through hundreds of posterior samples per input to be able to confidently validate or refute suspicions about the unobserved ground-truth image [ 15]. Several methods , 16(1):67\u201370, 2019. 1 [5]Yang Song, Liyue Shen, Lei Xing, and Stefano Ermon. Solving inverse problems in medi- cal imaging with score-based generative models. In International Conference on Learning Representations , 2021. 1, 2 [6]Guy Ohayon, Theo Adrai, Gregory Vaksman, Michael Elad, and Peyman Milanfar. High perceptual quality image denoising with a posterior sampling cgan. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 1805\u20131813, 2021. 2 [7]Bahjat Kawar, Michael Elad, Stefano Ermon, and Jiaming Song. Denoising diffusion restoration models. In Advances in Neural Information Processing Systems , 2022. 7, 9, 27 [8]Yinhuai Wang, Jiwen Yu, and Jian Zhang. Zero-shot image restoration using denoising diffusion null-space model. The Eleventh International Conference on Learning Representations , 2023. 2, 7, 9, 27 [9]Hyungjin Chung, Jeongsol Kim, Michael Thompson Mccann, Marc Louis Klasky, and Jong Chul Ye. Diffusion posterior sampling for general noisy inverse problems. In The Eleventh Interna- tional Conference on Learning Representations , 2022. [10] Wenbo Li, Zhe Lin, Kun Zhou, Lu Qi, Yi Wang, and Jiaya Jia. Mat: Mask-aware transformer for large hole image inpainting. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 10758\u201310768, 2022. 7, 9, 27 [11] Andreas Lugmayr, Martin Danelljan, Andres Romero, Fisher Yu, Radu Timofte, and Luc Van Gool. Repaint: Inpainting using denoising diffusion probabilistic models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 11461\u201311471, 2022. 7, 9, 27 [12] Chitwan Saharia, Jonathan Ho, William Chan, Tim Salimans, David J Fleet, and Mohammad Norouzi. Image super-resolution via iterative refinement. IEEE Transactions on Pattern Analysis and Machine Intelligence , 2022. [13] Chitwan Saharia, William Chan, Huiwen Chang, Chris Lee, Jonathan Ho, Tim Salimans, David Fleet, and Mohammad Norouzi. Palette: Image-to-image diffusion models. In ACM SIGGRAPH 2022 Conference Proceedings , pages 1\u201310, 2022. [14] Matthew Bendel, Rizwan Ahmad, and Philip Schniter. A regularized conditional gan for posterior sampling in inverse problems. arXiv preprint arXiv:2210.13389 , 2022. 1, 2 [15] Noa Cohen, Hila Manor, Yuval Bahat, and Tomer Michaeli. From posterior sampling to meaningful diversity in image restoration. arXiv preprint arXiv:2310.16047 , 2023. 1 [16] Vikash Sehwag, Caner Hazirbas, Albert Gordo, Firat Ozgenel, and Cristian Canton. Generating high fidelity data from low-density regions using diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 11492\u201311501, 2022. 1 11[17] Elias Nehme, Omer Yair,",
            "references": [
                {
                    "title": "Uncertainty Visualization via Low-Dimensional Posterior Projections",
                    "abstract": "In ill-posed inverse problems, it is commonly desirable to obtain insight into the full spectrum of plausible so-lutions, rather than extracting only a single reconstruction. Information about the plausible solutions and their likelihoods is encoded in the posterior distribution. However, for high-dimensional data, this distribution is challenging to visualize. In this work, we introduce a new approach for estimating and visualizing posteriors by em-ploying energy-based models (EBMs) over low-dimensional subspaces. Specifically, we train a conditional EBM that receives an input measurement and a set of directions that span some low-dimensional subspace of solutions, and outputs the probability density function of the posterior within that space. We demonstrate the effectiveness of our method across a diverse range of datasets and image restoration problems, showcasing its strength in uncertainty quantification and visualization. As we show, our method outperforms a baseline that projects samples from a diffusion-based posterior sampler, while being orders of magnitude faster. Furthermore, it is more accurate than a baseline that assumes a Gaussian posterior. Code is available at https://github.com/yairomer/PPDE."
                },
                {
                    "title": "Resilient Multiple Choice Learning: A learned scoring scheme with application to audio scene analysis",
                    "abstract": "We introduce Resilient Multiple Choice Learning (rMCL), an extension of the MCL approach for conditional distribution estimation in regression settings where multiple targets may be sampled for each training input. Multiple Choice Learning is a simple framework to tackle multimodal density estimation, using the Winner-Takes-All (WTA) loss for a set of hypotheses. In regression settings, the existing MCL variants focus on merging the hypotheses, thereby eventually sacrificing the diversity of the predictions. In contrast, our method relies on a novel learned scoring scheme underpinned by a mathematical framework based on Voronoi tessellations of the output space, from which we can derive a probabilistic interpretation. After empirically validating rMCL with experiments on synthetic data, we further assess its merits on the sound source localization problem, demonstrating its practical usefulness and the relevance of its interpretation."
                },
                {
                    "title": "From Posterior Sampling to Meaningful Diversity in Image Restoration",
                    "abstract": "Image restoration problems are typically ill-posed in the sense that each degraded image can be restored in infinitely many valid ways. To accommodate this, many works generate a diverse set of outputs by attempting to randomly sample from the posterior distribution of natural images given the degraded input. Here we argue that this strategy is commonly of limited practical value because of the heavy tail of the posterior distribution. Consider for example inpainting a missing region of the sky in an image. Since there is a high probability that the missing region contains no object but clouds, any set of samples from the posterior would be entirely dominated by (practically identical) completions of sky. However, arguably, presenting users with only one clear sky completion, along with several alternative solutions such as airships, birds, and balloons, would better outline the set of possibilities. In this paper, we initiate the study of meaningfully diverse image restoration. We explore several post-processing approaches that can be combined with any diverse image restoration method to yield semantically meaningful diversity. Moreover, we propose a practical approach for allowing diffusion based image restoration methods to generate meaningfully diverse outputs, while incurring only negligent computational overhead. We conduct extensive user studies to analyze the proposed techniques, and find the strategy of reducing similarity between outputs to be significantly favorable over posterior sampling. Code and examples are available at https://noa-cohen.github.io/MeaningfulDiversityInIR."
                },
                {
                    "title": "Uncertainty Quantification via Neural Posterior Principal Components",
                    "abstract": "Uncertainty quantification is crucial for the deployment of image restoration models in safety-critical domains, like autonomous driving and biological imaging. To date, methods for uncertainty visualization have mainly focused on per-pixel estimates. However, a heatmap of per-pixel variances is typically of little practical use, as it does not capture the strong correlations between pixels. A more natural measure of uncertainty corresponds to the variances along the principal components (PCs) of the posterior distribution. Theoretically, the PCs can be computed by applying PCA on samples generated from a conditional generative model for the input image. However, this requires generating a very large number of samples at test time, which is painfully slow with the current state-of-the-art (diffusion) models. In this work, we present a method for predicting the PCs of the posterior distribution for any input image, in a single forward pass of a neural network. Our method can either wrap around a pre-trained model that was trained to minimize the mean square error (MSE), or can be trained from scratch to output both a predicted image and the posterior PCs. We showcase our method on multiple inverse problems in imaging, including denoising, inpainting, super-resolution, and biological image-to-image translation. Our method reliably conveys instance-adaptive uncertainty directions, achieving uncertainty quantification comparable with posterior samplers while being orders of magnitude faster. Examples are available at https://eliasnehme.github.io/NPPC/"
                },
                {
                    "title": "On the Posterior Distribution in Denoising: Application to Uncertainty Quantification",
                    "abstract": "Denoisers play a central role in many applications, from noise suppression in low-grade imaging sensors, to empowering score-based generative models. The latter category of methods makes use of Tweedie's formula, which links the posterior mean in Gaussian denoising (\\ie the minimum MSE denoiser) with the score of the data distribution. Here, we derive a fundamental relation between the higher-order central moments of the posterior distribution, and the higher-order derivatives of the posterior mean. We harness this result for uncertainty quantification of pre-trained denoisers. Particularly, we show how to efficiently compute the principal components of the posterior distribution for any desired region of an image, as well as to approximate the full marginal distribution along those (or any other) one-dimensional directions. Our method is fast and memory-efficient, as it does not explicitly compute or store the high-order moment tensors and it requires no training or fine tuning of the denoiser. Code and examples are available on the project webpage in https://hilamanor.github.io/GaussianDenoisingPosterior/ ."
                },
                {
                    "title": "Principal Uncertainty Quantification With Spatial Correlation for Image Restoration Problems",
                    "abstract": "Uncertainty quantification for inverse problems in imaging has drawn much attention lately. Existing approaches towards this task define uncertainty regions based on probable values per pixel, while ignoring spatial correlations within the image, resulting in an exaggerated volume of uncertainty. In this paper, we propose PUQ (Principal Uncertainty Quantification) \u2013 a novel definition and corresponding analysis of uncertainty regions that takes into account spatial relationships within the image, thus providing reduced volume regions. Using recent advancements in generative models, we derive uncertainty intervals around principal components of the empirical posterior distribution, forming an ambiguity region that guarantees the inclusion of true unseen values with a user-defined confidence probability. To improve computational efficiency and interpretability, we also guarantee the recovery of true unseen values using only a few principal directions, resulting in more informative uncertainty regions. Our approach is verified through experiments on image colorization, super-resolution, and inpainting; its effectiveness is shown through comparison to baseline methods, demonstrating significantly tighter uncertainty regions."
                },
                {
                    "title": "Score-Based Diffusion Models as Principled Priors for Inverse Imaging",
                    "abstract": "Priors are essential for reconstructing images from noisy and/or incomplete measurements. The choice of the prior determines both the quality and uncertainty of recovered images. We propose turning score-based diffusion models into principled image priors (\"score-based priors\") for analyzing a posterior of images given measurements. Previously, probabilistic priors were limited to handcrafted regularizers and simple distributions. In this work, we empirically validate the theoretically-proven probability function of a score-based diffusion model. We show how to sample from resulting posteriors by using this probability function for variational inference. Our results, including experiments on denoising, deblurring, and interferometric imaging, suggest that score-based priors enable principled inference with a sophisticated, data-driven image prior."
                },
                {
                    "title": "Consistency Models",
                    "abstract": "Diffusion models have significantly advanced the fields of image, audio, and video generation, but they depend on an iterative sampling process that causes slow generation. To overcome this limitation, we propose consistency models, a new family of models that generate high quality samples by directly mapping noise to data. They support fast one-step generation by design, while still allowing multistep sampling to trade compute for sample quality. They also support zero-shot data editing, such as image inpainting, colorization, and super-resolution, without requiring explicit training on these tasks. Consistency models can be trained either by distilling pre-trained diffusion models, or as standalone generative models altogether. Through extensive experiments, we demonstrate that they outperform existing distillation techniques for diffusion models in one- and few-step sampling, achieving the new state-of-the-art FID of 3.55 on CIFAR-10 and 6.20 on ImageNet 64x64 for one-step generation. When trained in isolation, consistency models become a new family of generative models that can outperform existing one-step, non-adversarial generative models on standard benchmarks such as CIFAR-10, ImageNet 64x64 and LSUN 256x256."
                },
                {
                    "title": "Zero-Shot Image Restoration Using Denoising Diffusion Null-Space Model",
                    "abstract": "Most existing Image Restoration (IR) models are task-specific, which can not be generalized to different degradation operators. In this work, we propose the Denoising Diffusion Null-Space Model (DDNM), a novel zero-shot framework for arbitrary linear IR problems, including but not limited to image super-resolution, colorization, inpainting, compressed sensing, and deblurring. DDNM only needs a pre-trained off-the-shelf diffusion model as the generative prior, without any extra training or network modifications. By refining only the null-space contents during the reverse diffusion process, we can yield diverse results satisfying both data consistency and realness. We further propose an enhanced and robust version, dubbed DDNM+, to support noisy restoration and improve restoration quality for hard tasks. Our experiments on several IR tasks reveal that DDNM outperforms other state-of-the-art zero-shot IR methods. We also demonstrate that DDNM+ can solve complex real-world applications, e.g., old photo restoration."
                },
                {
                    "title": "On Distillation of Guided Diffusion Models",
                    "abstract": "Classifier-free guided diffusion models have recently been shown to be highly effective at high-resolution image generation, and they have been widely used in large-scale diffusion frameworks including DALL.E 2, Stable Diffusion and Imagen. However, a downside of classifier-free guided diffusion models is that they are computationally expensive at inference time since they require evaluating two diffusion models, a class-conditional model and an unconditional model, tens to hundreds of times. To deal with this limitation, we propose an approach to distilling classifier-free guided diffusion models into models that are fast to sample from: Given a pre-trained classifier-free guided model, we first learn a single model to match the output of the combined conditional and unconditional models, and then we progressively distill that model to a diffusion model that requires much fewer sampling steps. For standard diffusion models trained on the pixel-space, our approach is able to generate images visually comparable to that of the original model using as few as 4 sampling steps on ImageNet $64\\times 64$ and CIFAR-10, achieving FID/IS scores comparable to that of the original model while being up to 256 times faster to sample from. For diffusion models trained on the latent-space (e.g., Stable Diffusion), our approach is able to generate high-fidelity images using as few as 1 to 4 denoising steps, accelerating inference by at least 10-fold compared to existing methods on ImageNet $256\\times 256$ and LAION datasets. We further demonstrate the effectiveness of our approach on text-guided image editing and inpainting, where our distilled model is able to generate high-quality results using as few as 2\u20134 denoising steps."
                },
                {
                    "title": "Diffusion Posterior Sampling for General Noisy Inverse Problems",
                    "abstract": "Diffusion models have been recently studied as powerful generative inverse problem solvers, owing to their high quality reconstructions and the ease of combining existing iterative solvers. However, most works focus on solving simple linear inverse problems in noiseless settings, which significantly under-represents the complexity of real-world problems. In this work, we extend diffusion solvers to efficiently handle general noisy (non)linear inverse problems via approximation of the posterior sampling. Interestingly, the resulting posterior sampling scheme is a blended version of diffusion sampling with the manifold constrained gradient without a strict measurement consistency projection step, yielding a more desirable generative path in noisy settings compared to the previous studies. Our method demonstrates that diffusion models can incorporate various measurement noise statistics such as Gaussian and Poisson, and also efficiently handle noisy nonlinear inverse problems such as Fourier phase retrieval and non-uniform deblurring. Code available at https://github.com/DPS2022/diffusion-posterior-sampling"
                },
                {
                    "title": "Generating High Fidelity Data from Low-density Regions using Diffusion Models",
                    "abstract": "Our work focuses on addressing sample deficiency from low-density regions of data manifold in common image datasets. We leverage diffusion process based generative models to synthesize novel images from low-density regions. We observe that uniform sampling from diffusion models predominantly samples from high-density regions of the data manifold. Therefore, we modify the sampling process to guide it towards low-density regions while simulta-neously maintaining the fidelity of synthetic data. We rigorously demonstrate that our process successfully generates novel high fidelity samples from low-density regions. We further examine generated samples and show that the model does not memorize low-density data and indeed learns to generate novel samples from low-density regions."
                },
                {
                    "title": "MAT: Mask-Aware Transformer for Large Hole Image Inpainting",
                    "abstract": "Recent studies have shown the importance of modeling long-range interactions in the inpainting problem. To achieve this goal, existing approaches exploit either standalone attention techniques or transformers, but usually under a low resolution in consideration of computational cost. In this paper, we present a novel transformer-based model for large hole inpainting, which unifies the merits of transformers and convolutions to efficiently process high-resolution images. We carefully design each component of our framework to guarantee the high fidelity and diversity of recovered images. Specifically, we customize an inpainting-oriented transformer block, where the attention module aggregates non-local information only from partial valid tokens, indicated by a dynamic mask. Extensive experiments demonstrate the state-of-the-art performance of the new model on multiple benchmark datasets. Code is released at https://github.com/fenglinglwb/MAT."
                },
                {
                    "title": "Image-to-Image Regression with Distribution-Free Uncertainty Quantification and Applications in Imaging",
                    "abstract": "Image-to-image regression is an important learning task, used frequently in biological imaging. Current algorithms, however, do not generally offer statistical guarantees that protect against a model's mistakes and hallucinations. To address this, we develop uncertainty quantification techniques with rigorous statistical guarantees for image-to-image regression problems. In particular, we show how to derive uncertainty intervals around each pixel that are guaranteed to contain the true value with a user-specified confidence probability. Our methods work in conjunction with any base machine learning model, such as a neural network, and endow it with formal mathematical guarantees -- regardless of the true unknown data distribution or choice of model. Furthermore, they are simple to implement and computationally inexpensive. We evaluate our procedure on three image-to-image regression tasks: quantitative phase microscopy, accelerated magnetic resonance imaging, and super-resolution transmission electron microscopy of a Drosophila melanogaster brain."
                },
                {
                    "title": "Denoising Diffusion Restoration Models",
                    "abstract": "Many interesting tasks in image restoration can be cast as linear inverse problems. A recent family of approaches for solving these problems uses stochastic algorithms that sample from the posterior distribution of natural images given the measurements. However, efficient solutions often require problem-specific supervised training to model the posterior, whereas unsupervised methods that are not problem-specific typically rely on inefficient iterative methods. This work addresses these issues by introducing Denoising Diffusion Restoration Models (DDRM), an efficient, unsupervised posterior sampling method. Motivated by variational inference, DDRM takes advantage of a pre-trained denoising diffusion generative model for solving any linear inverse problem. We demonstrate DDRM's versatility on several image datasets for super-resolution, deblurring, inpainting, and colorization under various amounts of measurement noise. DDRM outperforms the current leading unsupervised methods on the diverse ImageNet dataset in reconstruction quality, perceptual quality, and runtime, being 5x faster than the nearest competitor. DDRM also generalizes well for natural images out of the distribution of the observed ImageNet training set."
                },
                {
                    "title": "RePaint: Inpainting using Denoising Diffusion Probabilistic Models",
                    "abstract": "Free-form inpainting is the task of adding new content to an image in the regions specified by an arbitrary binary mask. Most existing approaches train for a certain distribution of masks, which limits their generalization capabilities to unseen mask types. Furthermore, training with pixel-wise and perceptual losses often leads to simple textural extensions towards the missing areas instead of semantically meaningful generation. In this work, we propose RePaint: A Denoising Diffusion Probabilistic Model (DDPM) based inpainting approach that is applicable to even extreme masks. We employ a pretrained unconditional DDPM as the generative prior. To condition the generation process, we only alter the reverse diffusion iterations by sampling the unmasked regions using the given image infor-mation. Since this technique does not modify or condition the original DDPM network itself, the model produces high-quality and diverse output images for any inpainting form. We validate our method for both faces and general-purpose image inpainting using standard and extreme masks. Re-Paint outperforms state-of-the-art Autoregressive, and GAN approaches for at least five out of six mask distributions. Github Repository: git.io/RePaint"
                },
                {
                    "title": "Solving Inverse Problems in Medical Imaging with Score-Based Generative Models",
                    "abstract": "Reconstructing medical images from partial measurements is an important inverse problem in Computed Tomography (CT) and Magnetic Resonance Imaging (MRI). Existing solutions based on machine learning typically train a model to directly map measurements to medical images, leveraging a training dataset of paired images and measurements. These measurements are typically synthesized from images using a fixed physical model of the measurement process, which hinders the generalization capability of models to unknown measurement processes. To address this issue, we propose a fully unsupervised technique for inverse problem solving, leveraging the recently introduced score-based generative models. Specifically, we first train a score-based generative model on medical images to capture their prior distribution. Given measurements and a physical model of the measurement process at test time, we introduce a sampling method to reconstruct an image consistent with both the prior and the observed measurements. Our method does not assume a fixed measurement process during training, and can thus be flexibly adapted to different measurement processes at test time. Empirically, we observe comparable or better performance to supervised learning techniques in several medical imaging tasks in CT and MRI, while demonstrating significantly better generalization to unknown measurement processes."
                },
                {
                    "title": "Palette: Image-to-Image Diffusion Models",
                    "abstract": "This paper develops a unified framework for image-to-image translation based on conditional diffusion models and evaluates this framework on four challenging image-to-image translation tasks, namely colorization, inpainting, uncropping, and JPEG restoration. Our simple implementation of image-to-image diffusion models outperforms strong GAN and regression baselines on all tasks, without task-specific hyper-parameter tuning, architecture customization, or any auxiliary loss or sophisticated new techniques needed. We uncover the impact of an L2 vs. L1 loss in the denoising diffusion objective on sample diversity, and demonstrate the importance of self-attention in the neural architecture through empirical studies. Importantly, we advocate a unified evaluation protocol based on ImageNet, with human evaluation and sample quality scores (FID, Inception Score, Classification Accuracy of a pre-trained ResNet-50, and Perceptual Distance against original images). We expect this standardized evaluation protocol to play a role in advancing image-to-image translation research. Finally, we show that a generalist, multi-task diffusion model performs as well or better than task-specific specialist counterparts. Check out https://diffusion-palette.github.io/ for an overview of the results and code."
                },
                {
                    "title": "Estimating High Order Gradients of the Data Distribution by Denoising",
                    "abstract": "The first order derivative of a data density can be estimated efficiently by denoising score matching, and has become an important component in many applications, such as image generation and audio synthesis. Higher order derivatives provide additional local information about the data distribution and enable new applications. Although they can be estimated via automatic differentiation of a learned density model, this can amplify estimation errors and is expensive in high dimensional settings. To overcome these limitations, we propose a method to directly estimate high order derivatives (scores) of a data density from samples. We first show that denoising score matching can be interpreted as a particular case of Tweedie's formula. By leveraging Tweedie's formula on higher order moments, we generalize denoising score matching to estimate higher order derivatives. We demonstrate empirically that models trained with the proposed method can approximate second order derivatives more efficiently and accurately than via automatic differentiation. We show that our models can be used to quantify uncertainty in denoising and to improve the mixing speed of Langevin dynamics via Ozaki discretization for sampling synthetic data and natural images."
                },
                {
                    "title": "SNIPS: Solving Noisy Inverse Problems Stochastically",
                    "abstract": "In this work we introduce a novel stochastic algorithm dubbed SNIPS, which draws samples from the posterior distribution of any linear inverse problem, where the observation is assumed to be contaminated by additive white Gaussian noise. Our solution incorporates ideas from Langevin dynamics and Newton's method, and exploits a pre-trained minimum mean squared error (MMSE) Gaussian denoiser. The proposed approach relies on an intricate derivation of the posterior score function that includes a singular value decomposition (SVD) of the degradation operator, in order to obtain a tractable iterative algorithm for the desired sampling. Due to its stochasticity, the algorithm can produce multiple high perceptual quality samples for the same noisy observation. We demonstrate the abilities of the proposed paradigm for image deblurring, super-resolution, and compressive sensing. We show that the samples produced are sharp, detailed and consistent with the given measurements, and their diversity exposes the inherent uncertainty in the inverse problem being solved."
                },
                {
                    "title": "Image Super-Resolution via Iterative Refinement",
                    "abstract": "We present SR3, an approach to image Super-Resolution via Repeated Refinement. SR3 adapts denoising diffusion probabilistic models (Ho et al. 2020), (Sohl-Dickstein et al. 2015) to image-to-image translation, and performs super-resolution through a stochastic iterative denoising process. Output images are initialized with pure Gaussian noise and iteratively refined using a U-Net architecture that is trained on denoising at various noise levels, conditioned on a low-resolution input image. SR3 exhibits strong performance on super-resolution tasks at different magnification factors, on faces and natural images. We conduct human evaluation on a standard 8\u00d7 face super-resolution task on CelebA-HQ for which SR3 achieves a fool rate close to 50%, suggesting photo-realistic outputs, while GAN baselines do not exceed a fool rate of 34%. We evaluate SR3 on a 4\u00d7 super-resolution task on ImageNet, where SR3 outperforms baselines in human evaluation and classification accuracy of a ResNet-50 classifier trained on high-resolution images. We further show the effectiveness of SR3 in cascaded image generation, where a generative model is chained with super-resolution models to synthesize high-resolution images with competitive FID scores on the class-conditional 256\u00d7256 ImageNet generation challenge."
                },
                {
                    "title": "High Perceptual Quality Image Denoising with a Posterior Sampling CGAN",
                    "abstract": "The vast work in Deep Learning (DL) has led to a leap in image denoising research. Most DL solutions for this task have chosen to put their efforts on the denoiser\u2019s architecture while maximizing distortion performance. However, distortion driven solutions lead to blurry results with sub-optimal perceptual quality, especially in immoderate noise levels. In this paper we propose a different perspective, aiming to produce sharp and visually pleasing denoised images that are still faithful to their clean sources. Formally, our goal is to achieve high perceptual quality with acceptable distortion. This is attained by a stochastic denoiser that samples from the posterior distribution, trained as a generator in the framework of conditional generative adversarial networks (CGAN). Contrary to distortion-based regularization terms that conflict with perceptual quality, we introduce to the CGAN objective a theoretically founded penalty term that does not force a distortion requirement on individual samples, but rather on their mean. We showcase our proposed method with a novel denoiser architecture that achieves the reformed denoising goal and produces vivid and diverse outcomes in immoderate noise levels."
                },
                {
                    "title": "Knowledge Distillation in Iterative Generative Models for Improved Sampling Speed",
                    "abstract": "Iterative generative models, such as noise conditional score networks and denoising diffusion probabilistic models, produce high quality samples by gradually denoising an initial noise vector. However, their denoising process has many steps, making them 2-3 orders of magnitude slower than other generative models such as GANs and VAEs. In this paper, we establish a novel connection between knowledge distillation and image generation with a technique that distills a multi-step denoising process into a single step, resulting in a sampling speed similar to other single-step generative models. Our Denoising Student generates high quality samples comparable to GANs on the CIFAR-10 and CelebA datasets, without adversarial training. We demonstrate that our method scales to higher resolutions through experiments on 256 x 256 LSUN. Code and checkpoints are available at https://github.com/tcl9876/Denoising_Student"
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Stochastic Segmentation Networks: Modelling Spatially Correlated Aleatoric Uncertainty",
                    "abstract": "In image segmentation, there is often more than one plausible solution for a given input. In medical imaging, for example, experts will often disagree about the exact location of object boundaries. Estimating this inherent uncertainty and predicting multiple plausible hypotheses is of great interest in many applications, yet this ability is lacking in most current deep learning methods. In this paper, we introduce stochastic segmentation networks (SSNs), an efficient probabilistic method for modelling aleatoric uncertainty with any image segmentation network architecture. In contrast to approaches that produce pixel-wise estimates, SSNs model joint distributions over entire label maps and thus can generate multiple spatially coherent hypotheses for a single image. By using a low-rank multivariate normal distribution over the logit space to model the probability of the label map given the image, we obtain a spatially consistent probability distribution that can be efficiently computed by a neural network without any changes to the underlying architecture. We tested our method on the segmentation of real-world medical data, including lung nodules in 2D CT and brain tumours in 3D multimodal MRI scans. SSNs outperform state-of-the-art for modelling correlated uncertainty in ambiguous images while being much simpler, more flexible, and more efficient."
                },
                {
                    "title": "StarGAN v2: Diverse Image Synthesis for Multiple Domains",
                    "abstract": "A good image-to-image translation model should learn a mapping between different visual domains while satisfying the following properties: 1) diversity of generated images and 2) scalability over multiple domains. Existing methods address either of the issues, having limited diversity or multiple models for all domains. We propose StarGAN v2, a single framework that tackles both and shows significantly improved results over the baselines. Experiments on CelebA-HQ and a new animal faces dataset (AFHQ) validate our superiority in terms of visual quality, diversity, and scalability. To better assess image-to-image translation models, we release AFHQ, high-quality animal faces with large inter- and intra-domain differences. The code, pretrained models, and dataset are available at https://github.com/clovaai/stargan-v2."
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "Generating Diverse High-Fidelity Images with VQ-VAE-2",
                    "abstract": "We explore the use of Vector Quantized Variational AutoEncoder (VQ-VAE) models for large scale image generation. To this end, we scale and enhance the autoregressive priors used in VQ-VAE to generate synthetic samples of much higher coherence and fidelity than possible before. We use simple feed-forward encoder and decoder networks, making our model an attractive candidate for applications where the encoding and/or decoding speed is critical. Additionally, VQ-VAE requires sampling an autoregressive model only in the compressed latent space, which is an order of magnitude faster than sampling in the pixel space, especially for large images. We demonstrate that a multi-scale hierarchical organization of VQ-VAE, augmented with powerful priors over the latent codes, is able to generate samples with quality that rivals that of state of the art Generative Adversarial Networks on multifaceted datasets such as ImageNet, while not suffering from GAN's known shortcomings such as mode collapse and lack of diversity."
                },
                {
                    "title": "Versatile Multiple Choice Learning and Its Application to Vision Computing",
                    "abstract": "Most existing ensemble methods aim to train the underlying embedded models independently and simply aggregate their final outputs via averaging or weighted voting. As many prediction tasks contain uncertainty, most of these ensemble methods just reduce variance of the predictions without considering the collaborations among the ensembles. Different from these ensemble methods, multiple choice learning (MCL) methods exploit the cooperation among all the embedded models to generate multiple diverse hypotheses. In this paper, a new MCL method, called vMCL (the abbreviation of versatile Multiple Choice Learning), is developed to extend the application scenarios of MCL methods by ensembling deep neural networks. Our vMCL method keeps the advantage of existing MCL methods while overcoming their major drawback, thus achieves better performance. The novelty of our vMCL lies in three aspects: (1) a choice network is designed to learn the confidence level of each specialist which can provide the best prediction base on multiple hypotheses; (2) a hinge loss is introduced to alleviate the overconfidence issue in MCL settings; (3) Easy to be implemented and can be trained in an end-to-end manner, which is a very attractive feature for many real-world applications. Experiments on image classification and image segmentation task show that vMCL outperforms the existing state-of-the-art MCL methods."
                },
                {
                    "title": "Overcoming Limitations of Mixture Density Networks: A Sampling and Fitting Framework for Multimodal Future Prediction",
                    "abstract": "Future prediction is a fundamental principle of intelligence that helps plan actions and avoid possible dangers. As the future is uncertain to a large extent, modeling the uncertainty and multimodality of the future states is of great relevance. Existing approaches are rather limited in this regard and mostly yield a single hypothesis of the future or, at the best, strongly constrained mixture components that suffer from instabilities in training and mode collapse. In this work, we present an approach that involves the prediction of several samples of the future with a winner-takes-all loss and iterative grouping of samples to multiple modes. Moreover, we discuss how to evaluate predicted multimodal distributions, including the common real scenario, where only a single sample from the ground-truth distribution is available for evaluation. We show on synthetic and real data that the proposed approach triggers good estimates of multimodal distributions and avoids mode collapse."
                },
                {
                    "title": "Alpha Model Domination in Multiple Choice Learning",
                    "abstract": "Multiple Choice Learning (MCL) algorithms involve an 'oracle' user with unmodeled biases that selects a preferred output among several possible hypotheses. We identify a shortcoming and training obstacle of the recent Stochastic Multiple Choice Learning (sMCL) algorithm. When an ensemble of neural networks trains under sMCL, the best performing model receives the majority of parameter updates at the expense of other ensemble members. We refer to this sMCL training issue as Alpha Model Domination (AMD) and empirically demonstrate that AMD does not resolve itself with longer training time. We introduce several novel MCL loss functions that both avoid AMD and yield statistically significant improvements in oracle accuracy (OA) compared to sMCL. Using the MNIST, CIFAR-10, and ImageNet classification datasets, we empirically demonstrate the superior performance of our proposed loss functions."
                },
                {
                    "title": "DiverseNet: When One Right Answer is not Enough",
                    "abstract": "Many structured prediction tasks in machine vision have a collection of acceptable answers, instead of one definitive ground truth answer. Segmentation of images, for example, is subject to human labeling bias. Similarly, there are multiple possible pixel values that could plausibly complete occluded image regions. State-of-the art supervised learning methods are typically optimized to make a single test-time prediction for each query, failing to find other modes in the output space. Existing methods that allow for sampling often sacrifice speed or accuracy. We introduce a simple method for training a neural network, which enables diverse structured predictions to be made for each test-time query. For a single input, we learn to predict a range of possible answers. We compare favorably to methods that seek diversity through an ensemble of networks. Such stochastic multiple choice learning faces mode collapse, where one or more ensemble members fail to receive any training signal. Our best performing solution can be deployed for various tasks, and just involves small modifications to the existing single-mode architecture, loss function, and training regime. We demonstrate that our method results in quantitative improvements across three challenging tasks: 2D image completion, 3D volume estimation, and flow prediction."
                },
                {
                    "title": "Structured Uncertainty Prediction Networks",
                    "abstract": "This paper is the first work to propose a network to predict a structured uncertainty distribution for a synthesized image. Previous approaches have been mostly limited to predicting diagonal covariance matrices [15]. Our novel model learns to predict a full Gaussian covariance matrix for each reconstruction, which permits efficient sampling and likelihood evaluation. We demonstrate that our model can accurately reconstruct ground truth correlated residual distributions for synthetic datasets and generate plausible high frequency samples for real face images. We also illustrate the use of these predicted covariances for structure preserving image denoising."
                },
                {
                    "title": "Neural Discrete Representation Learning",
                    "abstract": "Learning useful representations without supervision remains a key challenge in machine learning. In this paper, we propose a simple yet powerful generative model that learns such discrete representations. Our model, the Vector Quantised-Variational AutoEncoder (VQ-VAE), differs from VAEs in two key ways: the encoder network outputs discrete, rather than continuous, codes; and the prior is learnt rather than static. In order to learn a discrete latent representation, we incorporate ideas from vector quantisation (VQ). Using the VQ method allows the model to circumvent issues of \"posterior collapse\" -- where the latents are ignored when they are paired with a powerful autoregressive decoder -- typically observed in the VAE framework. Pairing these representations with an autoregressive prior, the model can generate high quality images, videos, and speech as well as doing high quality speaker conversion and unsupervised learning of phonemes, providing further evidence of the utility of the learnt representations."
                },
                {
                    "title": "Double-Opportunity Estimation via Altruism",
                    "abstract": "Based on the notion of altruism, we present an approach to cooperative parameter estimation in a system comprising two information-sharing agents. The underlying assumption is that the overall two-agent scheme can reach desired performance level even if only one of the agents performs satisfactorily, hence there exist two independent opportunities to estimate. The notion of altruism motivates a definition of cooperative estimation optimality that generalizes the common definition of minimum mean-squared error optimality. Fundamental equations are derived for two types of altruistic cooperative estimation problems, corresponding to heterarchical and hierarchical setups. Although these equations are, generally, hard to solve, their solution in the Gaussian case is straightforward and only entails the computation of the largest eigenvalue of the conditional covariance matrix and its corresponding eigenvector. Moreover, in the Gaussian case, the performance improvement of the two altruistic cooperative estimation techniques over the conventional (egoistic) estimation approach is shown to depend on the problem's dimensionality and statistical distribution. In particular, the performance improvement grows with the dispersion of the spectrum of the conditional covariance matrix, rendering the presented estimation approach especially appealing in ill-conditioned problems. The validity of the solution in the Gaussian case is illustrated numerically."
                },
                {
                    "title": "Confident Multiple Choice Learning",
                    "abstract": "Ensemble methods are arguably the most trustworthy techniques for boosting the performance of machine learning models. Popular independent ensembles (IE) relying on naive averaging/voting scheme have been of typical choice for most applications involving deep neural networks, but they do not consider advanced collaboration among ensemble models. In this paper, we propose new ensemble methods specialized for deep neural networks, called confident multiple choice learning (CMCL): it is a variant of multiple choice learning (MCL) via addressing its overconfidence this http URL particular, the proposed major components of CMCL beyond the original MCL scheme are (i) new loss, i.e., confident oracle loss, (ii) new architecture, i.e., feature sharing and (iii) new training method, i.e., stochastic labeling. We demonstrate the effect of CMCL via experiments on the image classification on CIFAR and SVHN, and the foreground-background segmentation on the iCoseg. In particular, CMCL using 5 residual networks provides 14.05% and 6.60% relative reductions in the top-1 error rates from the corresponding IE scheme for the classification task on CIFAR and SVHN, respectively."
                },
                {
                    "title": "What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?",
                    "abstract": "There are two major types of uncertainty one can model. Aleatoric uncertainty captures noise inherent in the observations. On the other hand, epistemic uncertainty accounts for uncertainty in the model - uncertainty which can be explained away given enough data. Traditionally it has been difficult to model epistemic uncertainty in computer vision, but with new Bayesian deep learning tools this is now possible. We study the benefits of modeling epistemic vs. aleatoric uncertainty in Bayesian deep learning models for vision tasks. For this we present a Bayesian deep learning framework combining input-dependent aleatoric uncertainty together with epistemic uncertainty. We study models under the framework with per-pixel semantic segmentation and depth regression tasks. Further, our explicit uncertainty formulation leads to new loss functions for these tasks, which can be interpreted as learned attenuation. This makes the loss more robust to noisy data, also giving new state-of-the-art results on segmentation and depth regression benchmarks."
                },
                {
                    "title": "Learning in an Uncertain World: Representing Ambiguity Through Multiple Hypotheses",
                    "abstract": "Many prediction tasks contain uncertainty. In some cases, uncertainty is inherent in the task itself. In future prediction, for example, many distinct outcomes are equally valid. In other cases, uncertainty arises from the way data is labeled. For example, in object detection, many objects of interest often go unlabeled, and in human pose estimation, occluded joints are often labeled with ambiguous values. In this work we focus on a principled approach for handling such scenarios. In particular, we propose a frame-work for reformulating existing single-prediction models as multiple hypothesis prediction (MHP) models and an associated meta loss and optimization procedure to train them. To demonstrate our approach, we consider four diverse applications: human pose estimation, future prediction, image classification and segmentation. We find that MHP models outperform their single-hypothesis counterparts in all cases, and that MHP models simultaneously expose valuable insights into the variability of predictions."
                },
                {
                    "title": "Image-to-Image Translation with Conditional Adversarial Networks",
                    "abstract": "We investigate conditional adversarial networks as a general-purpose solution to image-to-image translation problems. These networks not only learn the mapping from input image to output image, but also learn a loss function to train this mapping. This makes it possible to apply the same generic approach to problems that traditionally would require very different loss formulations. We demonstrate that this approach is effective at synthesizing photos from label maps, reconstructing objects from edge maps, and colorizing images, among other tasks. Moreover, since the release of the pix2pix software associated with this paper, hundreds of twitter users have posted their own artistic experiments using our system. As a community, we no longer hand-engineer our mapping functions, and this work suggests we can achieve reasonable results without handengineering our loss functions either."
                },
                {
                    "title": "Stochastic Multiple Choice Learning for Training Diverse Deep Ensembles",
                    "abstract": "Many practical perception systems exist within larger processes that include interactions with users or additional components capable of evaluating the quality of predicted solutions. In these contexts, it is beneficial to provide these oracle mechanisms with multiple highly likely hypotheses rather than a single prediction. In this work, we pose the task of producing multiple outputs as a learning problem over an ensemble of deep networks -- introducing a novel stochastic gradient descent based approach to minimize the loss with respect to an oracle. Our method is simple to implement, agnostic to both architecture and loss function, and parameter-free. Our approach achieves lower oracle error compared to existing methods on a wide range of tasks and deep architectures. We also show qualitatively that the diverse solutions produced often provide interpretable representations of task ambiguity."
                },
                {
                    "title": "CVXPY: A Python-Embedded Modeling Language for Convex Optimization",
                    "abstract": "CVXPY is a domain-specific language for convex optimization embedded in Python. It allows the user to express convex optimization problems in a natural syntax that follows the math, rather than in the restrictive standard form required by solvers. CVXPY makes it easy to combine convex optimization with high-level features of Python such as parallelism and object-oriented design. CVXPY is available at http://www.cvxpy.org/ under the GPL license, along with documentation and examples."
                },
                {
                    "title": "Why M Heads are Better than One: Training a Diverse Ensemble of Deep Networks",
                    "abstract": "Convolutional Neural Networks have achieved state-of-the-art performance on a wide range of tasks. Most benchmarks are led by ensembles of these powerful learners, but ensembling is typically treated as a post-hoc procedure implemented by averaging independently trained models with model variation induced by bagging or random initialization. In this paper, we rigorously treat ensembling as a first-class problem to explicitly address the question: what are the best strategies to create an ensemble? We first compare a large number of ensembling strategies, and then propose and evaluate novel strategies, such as parameter sharing (through a new family of models we call TreeNets) as well as training under ensemble-aware and diversity-encouraging losses. We demonstrate that TreeNets can improve ensemble performance and that diverse ensembles can be trained end-to-end under a unified loss, achieving significantly higher \"oracle\" accuracies than classical ensembles."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "Deep Learning Face Attributes in the Wild",
                    "abstract": "Predicting face attributes in the wild is challenging due to complex face variations. We propose a novel deep learning framework for attribute prediction in the wild. It cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently. LNet is pre-trained by massive general object categories for face localization, while ANet is pre-trained by massive face identities for attribute prediction. This framework not only outperforms the state-of-the-art with a large margin, but also reveals valuable facts on learning face representation. (1) It shows how the performances of face localization (LNet) and attribute prediction (ANet) can be improved by different pre-training strategies. (2) It reveals that although the filters of LNet are fine-tuned only with image-level attribute tags, their response maps over entire images have strong indication of face locations. This fact enables training LNet for face localization with only image-level annotations, but without face bounding boxes or landmarks, which are required by all attribute recognition works. (3) It also demonstrates that the high-level hidden neurons of ANet automatically discover semantic concepts after pre-training with massive face identities, and such concepts are significantly enriched after fine-tuning with attribute tags. Each attribute can be well explained with a sparse linear combination of these concepts."
                },
                {
                    "title": "Fine-Grained Visual Comparisons with Local Learning",
                    "abstract": "Given two images, we want to predict which exhibits a particular visual attribute more than the other-even when the two images are quite similar. Existing relative attribute methods rely on global ranking functions; yet rarely will the visual cues relevant to a comparison be constant for all data, nor will humans' perception of the attribute necessarily permit a global ordering. To address these issues, we propose a local learning approach for fine-grained visual comparisons. Given a novel pair of images, we learn a local ranking model on the fly, using only analogous training comparisons. We show how to identify these analogous pairs using learned metrics. With results on three challenging datasets-including a large newly curated dataset for fine-grained comparisons-our method outperforms stateof-the-art methods for relative attribute prediction."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "Multiple Choice Learning: Learning to Produce Multiple Structured Outputs",
                    "abstract": "We address the problem of generating multiple hypotheses for structured prediction tasks that involve interaction with users or successive components in a cascaded architecture. Given a set of multiple hypotheses, such components/users typically have the ability to retrieve the best (or approximately the best) solution in this set. The standard approach for handling such a scenario is to first learn a single-output model and then produce M-Best Maximum a Posteriori (MAP) hypotheses from this model. In contrast, we learn to produce multiple outputs by formulating this task as a multiple-output structured-output prediction problem with a loss-function that effectively captures the setup of the problem. We present a max-margin formulation that minimizes an upper-bound on this loss-function. Experimental results on image segmentation and protein side-chain prediction show that our method outperforms conventional approaches used for this type of scenario and leads to substantial improvements in prediction accuracy."
                },
                {
                    "title": "Estimation of Non-Normalized Statistical Models by Score Matching",
                    "abstract": "One often wants to estimate statistical models where the probability density function is known only up to a multiplicative normalization constant. Typically, one then has to resort to Markov Chain Monte Carlo methods, or approximations of the normalization constant. Here, we propose that such models can be estimated by minimizing the expected squared distance between the gradient of the log-density given by the model and the gradient of the log-density of the observed data. While the estimation of the gradient of log-density function is, in principle, a very difficult non-parametric problem, we prove a surprising result that gives a simple formula for this objective function. The density function of the observed data does not appear in this formula, which simplifies to a sample average of a sum of some derivatives of the log-density given by the model. The validity of the method is demonstrated on multivariate Gaussian and independent component analysis models, and by estimating an overcomplete filter set for natural image data."
                },
                {
                    "title": "A Regularized Conditional GAN for Posterior Sampling in Inverse Problems",
                    "abstract": "In inverse problems, one seeks to reconstruct an image from incomplete and/or degraded measurements. Such problems arise in magnetic resonance imaging (MRI), computed tomography, deblurring, superresolution, inpainting, and other applications. It is often the case that many image hypotheses are consistent with both the measurements and prior information, and so the goal is not to recover a single \u201cbest\u201d hypothesis but rather to explore the space of probable hypotheses, i.e., to sample from the posterior distribution. In this work, we propose a regularized conditional Wasserstein GAN that can generate dozens of high-quality posterior samples per second. Using quantitative evaluation metrics like conditional Fr\u00e9chet inception distance, we demonstrate that our method produces state-of-the-art posterior samples in both multicoil MRI and inpainting applications."
                },
                {
                    "title": "Structuring Uncertainty for Fine-Grained Sampling in Stochastic Segmentation Networks",
                    "abstract": "In image segmentation, the classic approach of learning a deterministic segmentation neither accounts for noise and ambiguity in the data nor for expert disagreements about the correct segmentation. This has been addressed by architectures that predict heteroscedastic (input-dependent) segmentation uncertainty, which indicates regions of segmentations that should be treated with care. What is missing are structural insights into the uncertainty, which would be desirable for interpretability and systematic adjustments. In the context of state-of-the-art stochastic segmentation networks (SSNs), we solve this issue by dismantling the overall predicted uncertainty into smaller uncertainty components. We obtain them directly from the low-rank Gaussian distribution for the logits in the network head of SSNs, based on a previously unconsidered view of this distribution as a factor model. The rank subsequently encodes a number of latent variables, each of which controls an individual uncertainty component. Hence, we can use the latent variables (called factors) for \ufb01ne-grained sample control, thereby solving an open problem from previous work. There is one caveat though\u2013factors are only unique up to orthogonal rotations. Factor rotations allow us to structure the uncertainty in a way that endorses simplicity, non-redundancy, and separation among the individual uncertainty components."
                },
                {
                    "title": "Efficiently Enforcing Diversity in Multi-Output Structured Prediction",
                    "abstract": "This paper proposes a novel method for efficiently generating multiple diverse predictions for structured prediction problems. Existing methods like SDPPs or DivMBest work by making a series of predictions where each prediction is made after considering the predictions that came before it. Such approaches are inherently sequential and computationally expensive. In contrast, our method, Diverse Multiple Choice Learning, learns a set of models to make multiple independent, yet diverse, predictions at testtime. We achieve this by including a diversity encouraging term in the loss function used for training the models. This approach encourages diversity in the predictions while preserving computational efficiency at test-time. Experimental results on a number of challenging problems show that our method learns models that not only predict more diverse results than competing methods, but are also able to generalize better and produce results with high test accuracy."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.LG",
                "eess.IV",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively communicate prediction uncertainty in imaging inverse problems to enhance user confidence in the results?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the need for reliable machine learning models in safety-critical domains, such as medical imaging, where inaccuracies can have severe consequences. By improving the communication of uncertainty, researchers can foster greater trust in machine learning applications, leading to more widespread adoption and innovative uses in various fields. This advancement could also pave the way for new methodologies in uncertainty quantification, ultimately enhancing the robustness and interpretability of machine learning models.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the high-dimensional nature of posterior distributions, which complicates visualization and interpretation. Naive approaches, such as simply generating a large number of samples, may overwhelm users with information, making it difficult to extract meaningful insights. Additionally, the complexity of the underlying models and the variability in the data introduce technical obstacles, such as computational inefficiency and the risk of misinterpretation of the uncertainty communicated through the samples.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on generating samples from posterior distributions without adequately addressing the visualization and communication of these samples to users. Limitations in existing solutions include a lack of user-centric design and insufficient methods for summarizing high-dimensional uncertainty. Barriers such as the complexity of the models and the need for intuitive interfaces have hindered progress. Our approach aims to bridge these gaps by developing more effective visualization techniques and user-friendly tools that enhance the interpretability of uncertainty in imaging inverse problems.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a novel framework for visualizing posterior distributions in imaging inverse problems. We will utilize advanced sampling techniques on a diverse dataset of medical images, employing metrics such as the Kullback-Leibler divergence and user studies to evaluate the effectiveness of our visualizations. The expected outcomes include improved user comprehension of uncertainty, enhanced decision-making capabilities in critical applications, and a set of guidelines for best practices in communicating prediction uncertainty in machine learning models."
            }
        },
        "author_data": {
            "13e852d6-783b-4351-9e7e-731aec7acdee": {
                "pk": "13e852d6-783b-4351-9e7e-731aec7acdee",
                "name": "Elias Nehme",
                "collaborators": [
                    "Tomer Michaeli",
                    "Lucien E. Weiss",
                    "Yoav Shechtman",
                    "Omer Yair",
                    "Boris Ferdman",
                    "Daniel Freedman",
                    "Tal Naor",
                    "Racheli Gordon",
                    "Onit Alalouf",
                    "Reut Orange"
                ],
                "domain": [
                    "Image Restoration",
                    "Uncertainty Quantification",
                    "Deep Learning",
                    "Localization Microscopy"
                ],
                "publications": [
                    {
                        "title": "Uncertainty Quantification via Neural Posterior Principal Components",
                        "abstract": "Uncertainty quantification is crucial for the deployment of image restoration models in safety-critical domains, like autonomous driving and biological imaging. To date, methods for uncertainty visualization have mainly focused on per-pixel estimates. Yet, a heatmap of per-pixel variances is typically of little practical use, as it does not capture the strong correlations between pixels. A more natural measure of uncertainty corresponds to the variances along the principal components (PCs) of the posterior distribution. Theoretically, the PCs can be computed by applying PCA on samples generated from a conditional generative model for the input image. However, this requires generating a very large number of samples at test time, which is painfully slow with the current state-of-the-art (diffusion) models. In this work, we present a method for predicting the PCs of the posterior distribution for any input image, in a single forward pass of a neural network. Our method can either wrap around a pre-trained model that was trained to minimize the mean square error (MSE), or can be trained from scratch to output both a predicted image and the posterior PCs. We showcase our method on multiple inverse problems in imaging, including denoising, inpainting, super-resolution, and biological image-to-image translation. Our method reliably conveys instance-adaptive uncertainty directions, achieving uncertainty quantification comparable with posterior samplers while being orders of magnitude faster. Code and examples are available at https://eliasnehme.github.io/NPPC/"
                    },
                    {
                        "title": "Uncertainty Visualization via Low-Dimensional Posterior Projections",
                        "abstract": "In ill-posed inverse problems, it is commonly desirable to obtain insight into the full spectrum of plausible solutions, rather than extracting only a single reconstruction. Information about the plausible solutions and their likelihoods is encoded in the posterior distribution. However, for high-dimensional data, this distribution is challenging to visualize. In this work, we introduce a new approach for estimating and visualizing posteriors by employing energy-based models (EBMs) over low-dimensional subspaces. Specifically, we train a conditional EBM that receives an input measurement and a set of directions that span some low-dimensional subspace of solutions, and outputs the probability density function of the posterior within that space. We demonstrate the effectiveness of our method across a diverse range of datasets and image restoration problems, showcasing its strength in uncertainty quantification and visualization. As we show, our method outperforms a baseline that projects samples from a diffusion-based posterior sampler, while being orders of magnitude faster. Furthermore, it is more accurate than a baseline that assumes a Gaussian posterior."
                    },
                    {
                        "title": "Learning an optimal PSF-pair for ultra-dense 3D localization microscopy",
                        "abstract": "A long-standing challenge in multiple-particle-tracking is the accurate and precise 3D localization of individual particles at close proximity. One established approach for snapshot 3D imaging is point-spread-function (PSF) engineering, in which the PSF is modified to encode the axial information. However, engineered PSFs are challenging to localize at high densities due to lateral PSF overlaps. Here we suggest using multiple PSFs simultaneously to help overcome this challenge, and investigate the problem of engineering multiple PSFs for dense 3D localization. We implement our approach using a bifurcated optical system that modifies two separate PSFs, and design the PSFs using three different approaches including end-to-end learning. We demonstrate our approach experimentally by volumetric imaging of fluorescently labelled telomeres in cells."
                    },
                    {
                        "title": "DeepSTORM3D: dense three dimensional localization microscopy and point spread function design by deep learning",
                        "abstract": "Localization microscopy is an imaging technique in which the positions of individual nanoscale point emitters (e.g. fluorescent molecules) are determined at high precision from their images. This is the key ingredient in single/multiple-particle-tracking and several super-resolution microscopy approaches. Localization in three-dimensions (3D) can be performed by modifying the image that a point-source creates on the camera, namely, the point-spread function (PSF). The PSF is engineered using additional optical elements to vary distinctively with the depth of the point-source. However, localizing multiple adjacent emitters in 3D poses a significant algorithmic challenge, due to the lateral overlap of their PSFs. Here, we train a neural network to receive an image containing densely overlapping PSFs of multiple emitters over a large axial range and output a list of their 3D positions. Furthermore, we then use the network to design the optimal PSF for the multi-emitter case. We demonstrate our approach numerically as well as experimentally by 3D STORM imaging of mitochondria, and volumetric imaging of dozens of fluorescently-labeled telomeres occupying a mammalian nucleus in a single snapshot."
                    },
                    {
                        "title": "Deep-STORM: super-resolution single-molecule microscopy by deep learning",
                        "abstract": "We present an ultra-fast, precise, parameter-free method, which we term Deep-STORM, for obtaining super-resolution images from stochastically-blinking emitters, such as fluorescent molecules used for localization microscopy. Deep-STORM uses a deep convolutional neural network that can be trained on simulated data or experimental measurements, both of which are demonstrated. The method achieves state-of-the-art resolution under challenging signal-to-noise conditions and high emitter densities, and is significantly faster than existing approaches. Additionally, no prior information on the shape of the underlying structure is required, making the method applicable to any blinking data-set. We validate our approach by super-resolution image reconstruction of simulated and experimentally obtained data."
                    }
                ]
            },
            "849e158b-d2f0-4689-b4bc-d90ad96b0935": {
                "pk": "849e158b-d2f0-4689-b4bc-d90ad96b0935",
                "name": "Rotem Mulayoff",
                "collaborators": [
                    "Tomer Michaeli",
                    "Ronen Talmon",
                    "Amitay Bar",
                    "Mor Shpigel Nacson",
                    "Greg Ongie",
                    "Daniel Soudry",
                    "Felix Dietrich",
                    "Or Yair",
                    "Ioannis G. Kevrekidis",
                    "Ren\u00e9 Haas"
                ],
                "domain": [
                    "Deep Learning",
                    "Optimization",
                    "Neural Networks",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Unique Properties of Flat Minima in Deep Networks",
                        "abstract": "It is well known that (stochastic) gradient descent has an implicit bias towards flat minima. In deep neural network training, this mechanism serves to screen out minima. However, the precise effect that this has on the trained network is not yet fully understood. In this paper, we characterize the flat minima in linear neural networks trained with a quadratic loss. First, we show that linear ResNets with zero initialization necessarily converge to the flattest of all minima. We then prove that these minima correspond to nearly balanced networks whereby the gain from the input to any intermediate representation does not change drastically from one layer to the next. Finally, we show that consecutive layers in flat minima solutions are coupled. That is, one of the left singular vectors of each weight matrix, equals one of the right singular vectors of the next matrix. This forms a distinct path from input to output, that, as we show, is dedicated to the signal that experiences the largest gain end-to-end. Experiments indicate that these properties are characteristic of both linear and nonlinear models trained in practice."
                    },
                    {
                        "title": "On the Minimal Overcompleteness Allowing Universal Sparse Representation",
                        "abstract": "Sparse representation over redundant dictionaries constitutes a good model for many classes of signals (e.g., patches of natural images, segments of speech signals, etc.). However, despite its popularity, very little is known about the representation capacity of this model. In this paper, we study how redundant a dictionary must be so as to allow any vector to admit a sparse approximation with a prescribed sparsity and a prescribed level of accuracy. We address this problem both in a worst-case setting and in an average-case one. For each scenario we derive lower and upper bounds on the minimal required overcompleteness. Our bounds have simple closed-form expressions that allow to easily deduce the asymptotic behavior in large dimensions. In particular, we find that the required overcompleteness grows exponentially with the sparsity level and polynomially with the allowed representation error. This implies that universal sparse representation is practical only at moderate sparsity levels, but can be achieved at relatively high accuracy. As a side effect of our analysis, we obtain a tight lower bound on the regularized incomplete beta function, which may be interesting in its own right. We illustrate the validity of our results through numerical simulations, which support our findings."
                    },
                    {
                        "title": "Exact Mean Square Linear Stability Analysis for SGD",
                        "abstract": "The dynamical stability of optimization methods at the vicinity of minima of the loss has recently attracted significant attention. For gradient descent (GD), stable convergence is possible only to minima that are sufficiently flat w.r.t. the step size, and those have been linked with favorable properties of the trained model. However, while the stability threshold of GD is well-known, to date, no explicit expression has been derived for the exact threshold of stochastic GD (SGD). In this paper, we derive such a closed-form expression. Specifically, we provide an explicit condition on the step size that is both necessary and sufficient for the linear stability of SGD in the mean square sense. Our analysis sheds light on the precise role of the batch size $B$. In particular, we show that the stability threshold is monotonically non-decreasing in the batch size, which means that reducing the batch size can only decrease stability. Furthermore, we show that SGD's stability threshold is equivalent to that of a mixture process which takes in each iteration a full batch gradient step w.p. $1-p$, and a single sample gradient step w.p. $p$, where $p \\approx 1/B $. This indicates that even with moderate batch sizes, SGD's stability threshold is very close to that of GD's. We also prove simple necessary conditions for linear stability, which depend on the batch size, and are easier to compute than the precise threshold. Finally, we derive the asymptotic covariance of the dynamics around the minimum, and discuss its dependence on the learning rate. We validate our theoretical findings through experiments on the MNIST dataset."
                    },
                    {
                        "title": "The Expected Loss of Preconditioned Langevin Dynamics Reveals the Hessian Rank",
                        "abstract": "Langevin dynamics (LD) is widely used for sampling from distributions and for optimization. In this work, we derive a closed-form expression for the expected loss of preconditioned LD near stationary points of the objective function. We use the fact that at the vicinity of such points, LD reduces to an Ornstein-Uhlenbeck process, which is amenable to convenient mathematical treatment. Our analysis reveals that when the preconditioning matrix satisfies a particular relation with respect to the noise covariance, LD's expected loss becomes proportional to the rank of the objective's Hessian. We illustrate the applicability of this result in the context of neural networks, where the Hessian rank has been shown to capture the complexity of the predictor function but is usually computationally hard to probe. Finally, we use our analysis to compare SGD-like and Adam-like preconditioners and identify the regimes under which each of them leads to a lower expected loss."
                    },
                    {
                        "title": "The Implicit Bias of Minima Stability in Multivariate Shallow ReLU Networks",
                        "abstract": "We study the type of solutions to which stochastic gradient descent converges when used to train a single hidden-layer multivariate ReLU network with the quadratic loss. Our results are based on a dynamical stability analysis. In the univariate case, it was shown that linearly stable minima correspond to network functions (predictors), whose second derivative has a bounded weighted $L^1$ norm. Notably, the bound gets smaller as the step size increases, implying that training with a large step size leads to `smoother' predictors. Here we generalize this result to the multivariate case, showing that a similar result applies to the Laplacian of the predictor. We demonstrate the tightness of our bound on the MNIST dataset, and show that it accurately captures the behavior of the solutions as a function of the step size. Additionally, we prove a depth separation result on the approximation power of ReLU networks corresponding to stable minima of the loss. Specifically, although shallow ReLU networks are universal approximators, we prove that stable shallow networks are not. Namely, there is a function that cannot be well-approximated by stable single hidden-layer ReLU networks trained with a non-vanishing step size. This is while the same function can be realized as a stable two hidden-layer ReLU network. Finally, we prove that if a function is sufficiently smooth (in a Sobolev sense) then it can be approximated arbitrarily well using shallow ReLU networks that correspond to stable solutions of gradient descent."
                    },
                    {
                        "title": "Spectral Discovery of Jointly Smooth Features for Multimodal Data",
                        "abstract": "In this paper, we propose a spectral method for deriving functions that are jointly smooth on multiple observed manifolds. This allows us to register measurements of the same phenomenon by heterogeneous sensors, and to reject sensor-specific noise. Our method is unsupervised and primarily consists of two steps. First, using kernels, we obtain a subspace spanning smooth functions on each separate manifold. Then, we apply a spectral method to the obtained subspaces and discover functions that are jointly smooth on all manifolds. We show analytically that our method is guaranteed to provide a set of orthogonal functions that are as jointly smooth as possible, ordered by increasing Dirichlet energy from the smoothest to the least smooth. In addition, we show that the extracted functions can be efficiently extended to unseen data using the Nystr\\\"{o}m method. We demonstrate the proposed method on both simulated and real measured data and compare the results to nonlinear variants of the seminal Canonical Correlation Analysis (CCA). Particularly, we show superior results for sleep stage identification. In addition, we show how the proposed method can be leveraged for finding minimal realizations of parameter spaces of nonlinear dynamical systems."
                    },
                    {
                        "title": "Discovering Interpretable Directions in the Semantic Latent Space of Diffusion Models",
                        "abstract": "Denoising Diffusion Models (DDMs) have emerged as a strong competitor to Generative Adversarial Networks (GANs). However, despite their widespread use in image synthesis and editing applications, their latent space is still not as well understood. Recently, a semantic latent space for DDMs, coined `$h$-space', was shown to facilitate semantic image editing in a way reminiscent of GANs. The $h$-space is comprised of the bottleneck activations in the DDM's denoiser across all timesteps of the diffusion process. In this paper, we explore the properties of h-space and propose several novel methods for finding meaningful semantic directions within it. We start by studying unsupervised methods for revealing interpretable semantic directions in pretrained DDMs. Specifically, we show that global latent directions emerge as the principal components in the latent space. Additionally, we provide a novel method for discovering image-specific semantic directions by spectral analysis of the Jacobian of the denoiser w.r.t. the latent code. Next, we extend the analysis by finding directions in a supervised fashion in unconditional DDMs. We demonstrate how such directions can be found by relying on either a labeled data set of real images or by annotating generated samples with a domain-specific attribute classifier. We further show how to semantically disentangle the found direction by simple linear projection. Our approaches are applicable without requiring any architectural modifications, text-based guidance, CLIP-based optimization, or model fine-tuning."
                    }
                ]
            },
            "6d6a1ca1-fcce-4f32-9869-9941c77c5d5c": {
                "pk": "6d6a1ca1-fcce-4f32-9869-9941c77c5d5c",
                "name": "Tomer Michaeli",
                "collaborators": [
                    "Yochai Blau",
                    "Rotem Mulayoff",
                    "Yonina C. Eldar",
                    "Idan Kligvasser",
                    "Hila Manor",
                    "Yuval Bahat",
                    "Omer Yair",
                    "Nurit Spingarn-Eliezer",
                    "Hagay Michaeli",
                    "Daniel Soudry"
                ],
                "domain": [
                    "Image Processing",
                    "Deep Learning",
                    "Signal Processing",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Rethinking Lossy Compression: The Rate-Distortion-Perception Tradeoff",
                        "abstract": "Lossy compression algorithms are typically designed and analyzed through the lens of Shannon's rate-distortion theory, where the goal is to achieve the lowest possible distortion (e.g., low MSE or high SSIM) at any given bit rate. However, in recent years, it has become increasingly accepted that \"low distortion\" is not a synonym for \"high perceptual quality\", and in fact optimization of one often comes at the expense of the other. In light of this understanding, it is natural to seek for a generalization of rate-distortion theory which takes perceptual quality into account. In this paper, we adopt the mathematical definition of perceptual quality recently proposed by Blau & Michaeli (2018), and use it to study the three-way tradeoff between rate, distortion, and perception. We show that restricting the perceptual quality to be high, generally leads to an elevation of the rate-distortion curve, thus necessitating a sacrifice in either rate or distortion. We prove several fundamental properties of this triple-tradeoff, calculate it in closed form for a Bernoulli source, and illustrate it visually on a toy MNIST example."
                    },
                    {
                        "title": "The Perception-Distortion Tradeoff",
                        "abstract": "Image restoration algorithms are typically evaluated by some distortion measure (e.g. PSNR, SSIM, IFC, VIF) or by human opinion scores that quantify perceived perceptual quality. In this paper, we prove mathematically that distortion and perceptual quality are at odds with each other. Specifically, we study the optimal probability for correctly discriminating the outputs of an image restoration algorithm from real images. We show that as the mean distortion decreases, this probability must increase (indicating worse perceptual quality). As opposed to the common belief, this result holds true for any distortion measure, and is not only a problem of the PSNR or SSIM criteria. We also show that generative-adversarial-nets (GANs) provide a principled way to approach the perception-distortion bound. This constitutes theoretical support to their observed success in low-level vision tasks. Based on our analysis, we propose a new methodology for evaluating image restoration methods, and use it to perform an extensive comparison between recent super-resolution algorithms."
                    },
                    {
                        "title": "Unique Properties of Flat Minima in Deep Networks",
                        "abstract": "It is well known that (stochastic) gradient descent has an implicit bias towards flat minima. In deep neural network training, this mechanism serves to screen out minima. However, the precise effect that this has on the trained network is not yet fully understood. In this paper, we characterize the flat minima in linear neural networks trained with a quadratic loss. First, we show that linear ResNets with zero initialization necessarily converge to the flattest of all minima. We then prove that these minima correspond to nearly balanced networks whereby the gain from the input to any intermediate representation does not change drastically from one layer to the next. Finally, we show that consecutive layers in flat minima solutions are coupled. That is, one of the left singular vectors of each weight matrix, equals one of the right singular vectors of the next matrix. This forms a distinct path from input to output, that, as we show, is dedicated to the signal that experiences the largest gain end-to-end. Experiments indicate that these properties are characteristic of both linear and nonlinear models trained in practice."
                    },
                    {
                        "title": "Non-Redundant Spectral Dimensionality Reduction",
                        "abstract": "Spectral dimensionality reduction algorithms are widely used in numerous domains, including for recognition, segmentation, tracking and visualization. However, despite their popularity, these algorithms suffer from a major limitation known as the \"repeated Eigen-directions\" phenomenon. That is, many of the embedding coordinates they produce typically capture the same direction along the data manifold. This leads to redundant and inefficient representations that do not reveal the true intrinsic dimensionality of the data. In this paper, we propose a general method for avoiding redundancy in spectral algorithms. Our approach relies on replacing the orthogonality constraints underlying those methods by unpredictability constraints. Specifically, we require that each embedding coordinate be unpredictable (in the statistical sense) from all previous ones. We prove that these constraints necessarily prevent redundancy, and provide a simple technique to incorporate them into existing methods. As we illustrate on challenging high-dimensional scenarios, our approach produces significantly more informative and compact representations, which improve visualization and classification tasks."
                    },
                    {
                        "title": "Contrastive Divergence Learning is a Time Reversal Adversarial Game",
                        "abstract": "Contrastive divergence (CD) learning is a classical method for fitting unnormalized statistical models to data samples. Despite its wide-spread use, the convergence properties of this algorithm are still not well understood. The main source of difficulty is an unjustified approximation which has been used to derive the gradient of the loss. In this paper, we present an alternative derivation of CD that does not require any approximation and sheds new light on the objective that is actually being optimized by the algorithm. Specifically, we show that CD is an adversarial learning procedure, where a discriminator attempts to classify whether a Markov chain generated from the model has been time-reversed. Thus, although predating generative adversarial networks (GANs) by more than a decade, CD is, in fact, closely related to these techniques. Our derivation settles well with previous observations, which have concluded that CD's update steps cannot be expressed as the gradients of any fixed objective function. In addition, as a byproduct, our derivation reveals a simple correction that can be used as an alternative to Metropolis-Hastings rejection, which is required when the underlying Markov chain is inexact (e.g. when using Langevin dynamics with a large step)."
                    },
                    {
                        "title": "Sparsity Aware Normalization for GANs",
                        "abstract": "Generative adversarial networks (GANs) are known to benefit from regularization or normalization of their critic (discriminator) network during training. In this paper, we analyze the popular spectral normalization scheme, find a significant drawback and introduce sparsity aware normalization (SAN), a new alternative approach for stabilizing GAN training. As opposed to other normalization methods, our approach explicitly accounts for the sparse nature of the feature maps in convolutional networks with ReLU activations. We illustrate the effectiveness of our method through extensive experiments with a variety of network architectures. As we show, sparsity is particularly dominant in critics used for image-to-image translation settings. In these cases our approach improves upon existing methods, in less training epochs and with smaller capacity networks, while requiring practically no computational overhead."
                    },
                    {
                        "title": "On the Posterior Distribution in Denoising: Application to Uncertainty Quantification",
                        "abstract": "Denoisers play a central role in many applications, from noise suppression in low-grade imaging sensors, to empowering score-based generative models. The latter category of methods makes use of Tweedie's formula, which links the posterior mean in Gaussian denoising (\\ie the minimum MSE denoiser) with the score of the data distribution. Here, we derive a fundamental relation between the higher-order central moments of the posterior distribution, and the higher-order derivatives of the posterior mean. We harness this result for uncertainty quantification of pre-trained denoisers. Particularly, we show how to efficiently compute the principal components of the posterior distribution for any desired region of an image, as well as to approximate the full marginal distribution along those (or any other) one-dimensional directions. Our method is fast and memory-efficient, as it does not explicitly compute or store the high-order moment tensors and it requires no training or fine tuning of the denoiser. Code and examples are available on the project webpage in https://hilamanor.github.io/GaussianDenoisingPosterior/ ."
                    },
                    {
                        "title": "Zero-Shot Unsupervised and Text-Based Audio Editing Using DDPM Inversion",
                        "abstract": "Editing signals using large pre-trained models, in a zero-shot manner, has recently seen rapid advancements in the image domain. However, this wave has yet to reach the audio domain. In this paper, we explore two zero-shot editing techniques for audio signals, which use DDPM inversion with pre-trained diffusion models. The first, which we coin ZEro-shot Text-based Audio (ZETA) editing, is adopted from the image domain. The second, named ZEro-shot UnSupervized (ZEUS) editing, is a novel approach for discovering semantically meaningful editing directions without supervision. When applied to music signals, this method exposes a range of musically interesting modifications, from controlling the participation of specific instruments to improvisations on the melody. Samples and code can be found in https://hilamanor.github.io/AudioEditing/ ."
                    },
                    {
                        "title": "Stealing Image-to-Image Translation Models With a Single Query",
                        "abstract": "Training deep neural networks requires significant computational resources and large datasets that are often confidential or expensive to collect. As a result, owners tend to protect their models by allowing access only via an API. Many works demonstrated the possibility of stealing such protected models by repeatedly querying the API. However, to date, research has predominantly focused on stealing classification models, for which a very large number of queries has been found necessary. In this paper, we study the possibility of stealing image-to-image models. Surprisingly, we find that many such models can be stolen with as little as a single, small-sized, query image using simple distillation. We study this phenomenon on a wide variety of model architectures, datasets, and tasks, including denoising, deblurring, deraining, super-resolution, and biological image-to-image translation. Remarkably, we find that the vulnerability to stealing attacks is shared by CNNs and by models with attention mechanisms, and that stealing is commonly possible even without knowing the architecture of the target model."
                    },
                    {
                        "title": "Dense xUnit Networks",
                        "abstract": "Deep net architectures have constantly evolved over the past few years, leading to significant advancements in a wide array of computer vision tasks. However, besides high accuracy, many applications also require a low computational load and limited memory footprint. To date, efficiency has typically been achieved either by architectural choices at the macro level (e.g. using skip connections or pruning techniques) or modifications at the level of the individual layers (e.g. using depth-wise convolutions or channel shuffle operations). Interestingly, much less attention has been devoted to the role of the activation functions in constructing efficient nets. Recently, Kligvasser et al. showed that incorporating spatial connections within the activation functions, enables a significant boost in performance in image restoration tasks, at any given budget of parameters. However, the effectiveness of their xUnit module has only been tested on simple small models, which are not characteristic of those used in high-level vision tasks. In this paper, we adopt and improve the xUnit activation, show how it can be incorporated into the DenseNet architecture, and illustrate its high effectiveness for classification and image restoration tasks alike. While the DenseNet architecture is extremely efficient to begin with, our dense xUnit net (DxNet) can typically achieve the same performance with far fewer parameters. For example, on ImageNet, our DxNet outperforms a ReLU-based DenseNet having 30% more parameters and achieves state-of-the-art results for this budget of parameters. Furthermore, in denoising and super-resolution, DxNet significantly improves upon all existing lightweight solutions, including the xUnit-based nets of Kligvasser et al."
                    },
                    {
                        "title": "What's in the Image? Explorable Decoding of Compressed Images",
                        "abstract": "The ever-growing amounts of visual contents captured on a daily basis necessitate the use of lossy compression methods in order to save storage space and transmission bandwidth. While extensive research efforts are devoted to improving compression techniques, every method inevitably discards information. Especially at low bit rates, this information often corresponds to semantically meaningful visual cues, so that decompression involves significant ambiguity. In spite of this fact, existing decompression algorithms typically produce only a single output, and do not allow the viewer to explore the set of images that map to the given compressed code. In this work we propose the first image decompression method to facilitate user-exploration of the diverse set of natural images that could have given rise to the compressed input code, thus granting users the ability to determine what could and what could not have been there in the original scene. Specifically, we develop a novel deep-network based decoder architecture for the ubiquitous JPEG standard, which allows traversing the set of decompressed images that are consistent with the compressed JPEG file. To allow for simple user interaction, we develop a graphical user interface comprising several intuitive exploration tools, including an automatic tool for examining specific solutions of interest. We exemplify our framework on graphical, medical and forensic use cases, demonstrating its wide range of potential applications."
                    },
                    {
                        "title": "On the Minimal Overcompleteness Allowing Universal Sparse Representation",
                        "abstract": "Sparse representation over redundant dictionaries constitutes a good model for many classes of signals (e.g., patches of natural images, segments of speech signals, etc.). However, despite its popularity, very little is known about the representation capacity of this model. In this paper, we study how redundant a dictionary must be so as to allow any vector to admit a sparse approximation with a prescribed sparsity and a prescribed level of accuracy. We address this problem both in a worst-case setting and in an average-case one. For each scenario we derive lower and upper bounds on the minimal required overcompleteness. Our bounds have simple closed-form expressions that allow to easily deduce the asymptotic behavior in large dimensions. In particular, we find that the required overcompleteness grows exponentially with the sparsity level and polynomially with the allowed representation error. This implies that universal sparse representation is practical only at moderate sparsity levels, but can be achieved at relatively high accuracy. As a side effect of our analysis, we obtain a tight lower bound on the regularized incomplete beta function, which may be interesting in its own right. We illustrate the validity of our results through numerical simulations, which support our findings."
                    },
                    {
                        "title": "Explorable Super Resolution",
                        "abstract": "Single image super resolution (SR) has seen major performance leaps in recent years. However, existing methods do not allow exploring the infinitely many plausible reconstructions that might have given rise to the observed low-resolution (LR) image. These different explanations to the LR image may dramatically vary in their textures and fine details, and may often encode completely different semantic information. In this paper, we introduce the task of explorable super resolution. We propose a framework comprising a graphical user interface with a neural network backend, allowing editing the SR output so as to explore the abundance of plausible HR explanations to the LR input. At the heart of our method is a novel module that can wrap any existing SR network, analytically guaranteeing that its SR outputs would precisely match the LR input, when downsampled. Besides its importance in our setting, this module is guaranteed to decrease the reconstruction error of any SR network it wraps, and can be used to cope with blur kernels that are different from the one the network was trained for. We illustrate our approach in a variety of use cases, ranging from medical imaging and forensics, to graphics."
                    },
                    {
                        "title": "Exact Mean Square Linear Stability Analysis for SGD",
                        "abstract": "The dynamical stability of optimization methods at the vicinity of minima of the loss has recently attracted significant attention. For gradient descent (GD), stable convergence is possible only to minima that are sufficiently flat w.r.t. the step size, and those have been linked with favorable properties of the trained model. However, while the stability threshold of GD is well-known, to date, no explicit expression has been derived for the exact threshold of stochastic GD (SGD). In this paper, we derive such a closed-form expression. Specifically, we provide an explicit condition on the step size that is both necessary and sufficient for the linear stability of SGD in the mean square sense. Our analysis sheds light on the precise role of the batch size $B$. In particular, we show that the stability threshold is monotonically non-decreasing in the batch size, which means that reducing the batch size can only decrease stability. Furthermore, we show that SGD's stability threshold is equivalent to that of a mixture process which takes in each iteration a full batch gradient step w.p. $1-p$, and a single sample gradient step w.p. $p$, where $p \\approx 1/B $. This indicates that even with moderate batch sizes, SGD's stability threshold is very close to that of GD's. We also prove simple necessary conditions for linear stability, which depend on the batch size, and are easier to compute than the precise threshold. Finally, we derive the asymptotic covariance of the dynamics around the minimum, and discuss its dependence on the learning rate. We validate our theoretical findings through experiments on the MNIST dataset."
                    },
                    {
                        "title": "Alias-Free Convnets: Fractional Shift Invariance via Polynomial Activations",
                        "abstract": "Although CNNs are believed to be invariant to translations, recent works have shown this is not the case, due to aliasing effects that stem from downsampling layers. The existing architectural solutions to prevent aliasing are partial since they do not solve these effects, that originate in non-linearities. We propose an extended anti-aliasing method that tackles both downsampling and non-linear layers, thus creating truly alias-free, shift-invariant CNNs. We show that the presented model is invariant to integer as well as fractional (i.e., sub-pixel) translations, thus outperforming other shift-invariant methods in terms of robustness to adversarial translations."
                    },
                    {
                        "title": "Semi-Supervised Single- and Multi-Domain Regression with Multi-Domain Training",
                        "abstract": "We address the problems of multi-domain and single-domain regression based on distinct and unpaired labeled training sets for each of the domains and a large unlabeled training set from all domains. We formulate these problems as a Bayesian estimation with partial knowledge of statistical relations. We propose a worst-case design strategy and study the resulting estimators. Our analysis explicitly accounts for the cardinality of the labeled sets and includes the special cases in which one of the labeled sets is very large or, in the other extreme, completely missing. We demonstrate our estimators in the context of removing expressions from facial images and in the context of audio-visual word recognition, and provide comparisons to several recently proposed multi-modal learning algorithms."
                    },
                    {
                        "title": "Coherence Awareness in Diffractive Neural Networks",
                        "abstract": "Diffractive neural networks hold great promise for applications requiring intensive computational processing. Considerable attention has focused on diffractive networks for either spatially coherent or spatially incoherent illumination. Here we illustrate that, as opposed to imaging systems, in diffractive networks the degree of spatial coherence has a dramatic effect. In particular, we show that when the spatial coherence length on the object is comparable to the minimal feature size preserved by the optical system, neither the incoherent nor the coherent extremes serve as acceptable approximations. Importantly, this situation is inherent to many settings involving active illumination, including reflected light microscopy, autonomous vehicles and smartphones. Following this observation, we propose a general framework for training diffractive networks for any specified degree of spatial and temporal coherence, supporting all types of linear and nonlinear layers. Using our method, we numerically optimize networks for image classification, and thoroughly investigate their performance dependence on the illumination coherence properties. We further introduce the concept of coherence-blind networks, which have enhanced resilience to changes in illumination conditions. Our findings serve as a steppingstone toward adopting all-optical neural networks in real-world applications, leveraging nothing but natural light."
                    },
                    {
                        "title": "Non-Invertible Gabor Transforms",
                        "abstract": "Time-frequency analysis, such as the Gabor transform, plays an important role in many signal processing applications. The redundancy of such representations is often directly related to the computational load of any algorithm operating in the transform domain. To reduce complexity, it may be desirable to increase the time and frequency sampling intervals beyond the point where the transform is invertible, at the cost of an inevitable recovery error. In this paper we initiate the study of recovery procedures for non-invertible Gabor representations. We propose using fixed analysis and synthesis windows, chosen e.g. according to implementation constraints, and to process the Gabor coefficients prior to synthesis in order to shape the reconstructed signal. We develop three methods to tackle this problem. The first follows from the consistency requirement, namely that the recovered signal has the same Gabor representation as the input signal. The second, is based on the minimization of a worst-case error criterion. Last, we develop a recovery technique based on the assumption that the input signal lies in some subspace of $L_2$. We show that for each of the criteria, the manipulation of the transform coefficients amounts to a 2D twisted convolution operation, which we show how to perform using a filter-bank. When the under-sampling factor is an integer, the processing reduces to standard 2D convolution. We provide simulation results to demonstrate the advantages and weaknesses of each of the algorithms."
                    },
                    {
                        "title": "Partially Linear Estimation with Application to Sparse Signal Recovery From Measurement Pairs",
                        "abstract": "We address the problem of estimating a random vector X from two sets of measurements Y and Z, such that the estimator is linear in Y. We show that the partially linear minimum mean squared error (PLMMSE) estimator does not require knowing the joint distribution of X and Y in full, but rather only its second-order moments. This renders it of potential interest in various applications. We further show that the PLMMSE method is minimax-optimal among all estimators that solely depend on the second-order statistics of X and Y. We demonstrate our approach in the context of recovering a signal, which is sparse in a unitary dictionary, from noisy observations of it and of a filtered version of it. We show that in this setting PLMMSE estimation has a clear computational advantage, while its performance is comparable to state-of-the-art algorithms. We apply our approach both in static and dynamic estimation applications. In the former category, we treat the problem of image enhancement from blurred/noisy image pairs, where we show that PLMMSE estimation performs only slightly worse than state-of-the art algorithms, while running an order of magnitude faster. In the dynamic setting, we provide a recursive implementation of the estimator and demonstrate its utility in the context of tracking maneuvering targets from position and acceleration measurements."
                    }
                ]
            }
        }
    },
    "2402.14585": {
        "paper_data": {
            "title": "Bandits with Abstention under Expert Advice",
            "url": "http://arxiv.org/abs/2402.14585v1",
            "arxiv_id": "2402.14585",
            "authors": [
                "Stephen Pasteris",
                "Alberto Rumi",
                "Maximilian Thiessen",
                "Shota Saito",
                "Atsushi Miyauchi",
                "Fabio Vitale",
                "Mark Herbster"
            ],
            "abstract": "We study the classic problem of prediction with expert advice under bandit feedback. Our model assumes that one action, corresponding to the learner's abstention from play, has no reward or loss on every trial. We propose the CBA algorithm, which exploits this assumption to obtain reward bounds that can significantly improve those of the classical Exp4 algorithm. We can view our problem as the aggregation of confidence-rated predictors when the learner has the option of abstention from play. Importantly, we are the first to achieve bounds on the expected cumulative reward for general confidence-rated predictors. In the special case of specialists we achieve a novel reward bound, significantly improving previous bounds of SpecialistExp (treating abstention as another action). As an example application, we discuss learning unions of balls in a finite metric space. In this contextual setting, we devise an efficient implementation of CBA, reducing the runtime from quadratic to almost linear in the number of contexts. Preliminary experiments show that CBA improves over existing bandit algorithms.",
            "introduction": "   1 Introduction  We study the classic problem of prediction with expert advice under bandit feedback. The problem is structured as a sequence of trials. During each trial, each expert recommends a probability distribution over the set of possible actions. The learner then selects an action, observes, and incurs the (potentially negative) reward associated with that action on that particular trial. In practical applications, errors often lead to severe consequences, and consistently making predictions is neither safe nor economically practical. For this reason, the abstention option has gained a lot of interest in the literature, both in the batch and online setting (Chow, 1957, 1970; Hendrickx et\u00a0al., 2021; Cortes et\u00a0al., 2018). Similarly to previous works, this paper is based on the assumption that one of the actions always has zero reward: such an action is equivalent to an abstention of the learner from play. Apart from the fact that the rewards are bounded, we make no further assumptions on how the rewards or expert predictions are generated. In this paper, we present an efficient algorithm CBA\u00a0(Confidence-rated Bandits with Abstentions) which exploits the abstention action to get reward bounds that can be dramatically higher than those of Exp4\u00a0(Auer et\u00a0al., 2002). In the worst case, our reward bound essentially matches that of Exp4\u00a0so that CBA\u00a0can be seen as a strict improvement, since the time-complexities of the two algorithms are identical in the general case.   Our problem can also be seen as that of aggregating confidence-rated predictors\u00a0(Blum & Mansour, 2007; Gaillard et\u00a0al., 2014; Luo & Schapire, 2015) when the learner has the option of abstaining from taking actions. When the problem is phrased in this way, at the start of each trial, each predictor recommends a probability distribution over the actions (which now may not include an action with zero reward) but with a confidence rating. A low confidence rating can mean that either the predictor thinks that all actions are bad (so that the learner should abstain) or simply does not know which action is the best. Previous works measure the performance of their algorithm in terms of a sum of scaled per-trial rewards.   In contrast to previous algorithms for confidence-rated predictors, our approach allows for the derivation of bounds on the expected cumulative reward of CBA. One specific scenario where prior algorithms can establish cumulative reward bounds is when the predictors are specialists\u00a0(Freund et\u00a0al., 1997), having either full confidence (a.k.a. awake) or no confidence (a.k.a. asleep) on any given trial. The SpecialistExp algorithm, a bandit version of the standard specialist algorithm (Herbster et\u00a0al., 2021), achieves regret bounds with respect to any subset of specialists in which exactly one specialist is awake on each trial. In Section 6, we present an illustrative problem involving learning balls in a space equipped with a distance function. This example demonstrates our capability to significantly outperform SpecialistExp, which treats abstention as just another action, when our confidence-rated predictors are indeed specialists. For this problem, we will give subroutines that significantly speed up CBA. We note that, for any specialist problem, in the worst case the cumulative reward of CBA\u00a0is essentially equal to",
            "references": [
                {
                    "title": "Multi-class Graph Clustering via Approximated Effective p-Resistance",
                    "abstract": "This paper develops an approximation to the (effective) $p$-resistance and applies it to multi-class clustering. Spectral methods based on the graph Laplacian and its generalization to the graph $p$-Laplacian have been a backbone of non-euclidean clustering techniques. The advantage of the $p$-Laplacian is that the parameter $p$ induces a controllable bias on cluster structure. The drawback of $p$-Laplacian eigenvector based methods is that the third and higher eigenvectors are difficult to compute. Thus, instead, we are motivated to use the $p$-resistance induced by the $p$-Laplacian for clustering. For $p$-resistance, small $p$ biases towards clusters with high internal connectivity while large $p$ biases towards clusters of small\"extent,\"that is a preference for smaller shortest-path distances between vertices in the cluster. However, the $p$-resistance is expensive to compute. We overcome this by developing an approximation to the $p$-resistance. We prove upper and lower bounds on this approximation and observe that it is exact when the graph is a tree. We also provide theoretical justification for the use of $p$-resistance for clustering. Finally, we provide experiments comparing our approximated $p$-resistance clustering to other $p$-Laplacian based methods."
                },
                {
                    "title": "A Survey on the Densest Subgraph Problem and its Variants",
                    "abstract": "The Densest Subgraph Problem requires us to find, in a given graph, a subset of vertices whose induced subgraph maximizes a measure of density. The problem has received a great deal of attention in the algorithmic literature since the early 1970s, with many variants proposed and many applications built on top of this basic definition. Recent years have witnessed a revival of research interest in this problem with several important contributions, including some groundbreaking results, published in 2022 and 2023. This survey provides a deep overview of the fundamental results and an exhaustive coverage of the many variants proposed in the literature, with a special attention to the most recent results. The survey also presents a comprehensive overview of applications and discusses some interesting open problems for this evergreen research topic."
                },
                {
                    "title": "Classification with abstention but without disparities",
                    "abstract": "Classification with abstention has gained a lot of attention in recent years as it allows to incorporate human decision-makers in the process. Yet, abstention can potentially amplify disparities and lead to discriminatory predictions. The goal of this work is to build a general purpose classification algorithm, which is able to abstain from prediction, while avoiding disparate impact. We formalize this problem as risk minimization under fairness and abstention constraints for which we derive the form of the optimal classifier. Building on this result, we propose a post-processing classification algorithm, which is able to modify any off-the-shelf score-based classifier using only unlabeled sample. We establish finite sample risk, fairness, and abstention guarantees for the proposed algorithm. In particular, it is shown that fairness and abstention constraints can be achieved independently from the initial classifier as long as sufficiently many unlabeled data is available. The risk guarantee is established in terms of the quality of the initial classifier. Our post-processing scheme reduces to a sparse linear program allowing for an efficient implementation, which we provide. Finally, we validate our method empirically showing that moderate abstention rates allow to bypass the risk-fairness trade-off."
                },
                {
                    "title": "Exact Recovery of Clusters in Finite Metric Spaces Using Oracle Queries",
                    "abstract": "We investigate the problem of exact cluster recovery using oracle queries. Previous results show that clusters in Euclidean spaces that are convex and separated with a margin can be reconstructed exactly using only $O(\\log n)$ same-cluster queries, where $n$ is the number of input points. In this work, we study this problem in the more challenging non-convex setting. We introduce a structural characterization of clusters, called $(\\beta,\\gamma)$-convexity, that can be applied to any finite set of points equipped with a metric (or even a semimetric, as the triangle inequality is not needed). Using $(\\beta,\\gamma)$-convexity, we can translate natural density properties of clusters (which include, for instance, clusters that are strongly non-convex in $\\mathbb{R}^d$) into a graph-theoretic notion of convexity. By exploiting this convexity notion, we design a deterministic algorithm that recovers $(\\beta,\\gamma)$-convex clusters using $O(k^2 \\log n + k^2 (6/\\beta\\gamma)^{dens(X)})$ same-cluster queries, where $k$ is the number of clusters and $dens(X)$ is the density dimension of the semimetric. We show that an exponential dependence on the density dimension is necessary, and we also show that, if we are allowed to make $O(k^2 + k\\log n)$ additional queries to a\"cluster separation\"oracle, then we can recover clusters that have different and arbitrary scales, even when the scale of each cluster is unknown."
                },
                {
                    "title": "Online Learning with Dependent Stochastic Feedback Graphs",
                    "abstract": "A general framework for online learning with partial information is one where feedback graphs specify which losses can be observed by the learner. We study a challenging scenario where feedback graphs vary stochastically with time and, more importantly, where graphs and losses are dependent. This scenario appears in several real-world applications that we describe where the outcome of actions are correlated. We devise a new algorithm for this setting that exploits the stochastic properties of the graphs and that bene-\ufb01ts from favorable regret guarantees. We present a detailed theoretical analysis of this algorithm, and also report the results of a series of experiments on real-world datasets, which show that our algorithm outperforms standard baselines for online learning with feedback graphs."
                },
                {
                    "title": "Bandit Algorithms",
                    "abstract": "sets of environments and policies respectively and ` : E \u00d7\u03a0\u2192 [0, 1] a bounded loss function. Given a policy \u03c0 let `(\u03c0) = (`(\u03bd1, \u03c0), . . . , `(\u03bdN , \u03c0)) be the loss vector resulting from policy \u03c0. Define S = {`(\u03c0) : \u03c0 \u2208 \u03a0} and \u03bb(S) = {x \u2208 cl(S) : y 6< x for all y \u2208 S} , where y 6< x is defined to mean it is not true that yi \u2264 xi for all i with strict inequality for at least one i. Prove that if \u03bb(S) \u2286 S and S is convex, then for each x \u2208 \u03bb(S) there exists a prior q \u2208 P(E) and policy \u03c0\u2217 such that `(\u03c0) = x and \u2211 \u03bd\u2208E q(\u03bd)`(\u03bd, \u03c0\u2217) = min \u03c0\u2208\u03a0 \u2211"
                },
                {
                    "title": "Fast Rates for Online Prediction with Abstention",
                    "abstract": "In the setting of sequential prediction of individual $\\{0, 1\\}$-sequences with expert advice, we show that by allowing the learner to abstain from the prediction by paying a cost marginally smaller than $\\frac 12$ (say, $0.49$), it is possible to achieve expected regret bounds that are independent of the time horizon $T$. We exactly characterize the dependence on the abstention cost $c$ and the number of experts $N$ by providing matching upper and lower bounds of order $\\frac{\\log N}{1-2c}$, which is to be contrasted with the best possible rate of $\\sqrt{T\\log N}$ that is available without the option to abstain. We also discuss various extensions of our model, including a setting where the sequence of abstention costs can change arbitrarily over time, where we show regret bounds interpolating between the slow and the fast rates mentioned above, under some natural assumptions on the sequence of abstention costs."
                },
                {
                    "title": "Online Learning with Abstention",
                    "abstract": "We present an extensive study of the key problem of online learning where algorithms are allowed to abstain from making predictions. In the adversarial setting, we show how existing online algorithms and guarantees can be adapted to this problem. In the stochastic setting, we first point out a bias problem that limits the straightforward extension of algorithms such as UCB-N to time-varying feedback graphs, as needed in this context. Next, we give a new algorithm, UCB-GT, that exploits historical data and is adapted to time-varying feedback graphs. We show that this algorithm benefits from more favorable regret guarantees than a possible, but limited, extension of UCB-N. We further report the results of a series of experiments demonstrating that UCB-GT largely outperforms that extension of UCB-N, as well as more standard baselines."
                },
                {
                    "title": "Faster unfolding of communities: speeding up the Louvain algorithm",
                    "abstract": "Many complex networks exhibit a modular structure of densely connected groups of nodes. Usually, such a modular structure is uncovered by the optimization of some quality function. Although flawed, modularity remains one of the most popular quality functions. The Louvain algorithm was originally developed for optimizing modularity, but has been applied to a variety of methods. As such, speeding up the Louvain algorithm enables the analysis of larger graphs in a shorter time for various methods. We here suggest to consider moving nodes to a random neighbor community, instead of the best neighbor community. Although incredibly simple, it reduces the theoretical runtime complexity from O(m) to O(nlog\u3008k\u3009) in networks with a clear community structure. In benchmark networks, it speeds up the algorithm roughly 2-3 times, while in some real networks it even reaches 10 times faster runtimes. This improvement is due to two factors: (1) a random neighbor is likely to be in a \"good\" community and (2) random neighbors are likely to be hubs, helping the convergence. Finally, the performance gain only slightly diminishes the quality, especially for modularity, thus providing a good quality-performance ratio. However, these gains are less pronounced, or even disappear, for some other measures such as significance or surprise."
                },
                {
                    "title": "{SNAP Datasets}: {Stanford} Large Network Dataset Collection",
                    "abstract": "A collection of more than 50 large network datasets from tens of thousands of nodes and edges to tens of millions of nodes and edges. In includes social networks, web graphs, road networks, internet networks, citation networks, collaboration networks, and communication networks."
                },
                {
                    "title": "A second-order bound with excess losses",
                    "abstract": "We study online aggregation of the predictions of experts, and first show new second-order regret bounds in the standard setting, which are obtained via a version of the Prod algorithm (and also a version of the polynomially weighted average algorithm) with multiple learning rates. These bounds are in terms of excess losses, the differences between the instantaneous losses suffered by the algorithm and the ones of a given expert. We then demonstrate the interest of these bounds in the context of experts that report their confidences as a number in the interval [0; 1] using a generic reduction to the standard setting. We conclude by two other applications in the standard setting, which improve the known bounds in case of small excess losses and show a bounded regret against i.i.d. sequences of losses."
                },
                {
                    "title": "Phase transition in the family of p-resistances",
                    "abstract": "We study the family of p-resistances on graphs for p \u2265 1. This family generalizes the standard resistance distance. We prove that for any fixed graph, for p =1 the p-resistance coincides with the shortest path distance, for p = 2 it coincides with the standard resistance distance, and for p \u2192 \u221e it converges to the inverse of the minimal s-t-cut in the graph. Secondly, we consider the special case of random geometric graphs (such as k-nearest neighbor graphs) when the number n of vertices in the graph tends to infinity. We prove that an interesting phase transition takes place. There exist two critical thresholds p* and p** such that if p p**, it only depends on trivial local quantities and does not convey any useful information. We can explicitly compute the critical values: p* = 1 + 1/(d \u2212 1) and p** = 1 + 1/(d \u2212 2) where d is the dimension of the underlying space (we believe that the fact that there is a small gap between p* and p** is an artifact of our proofs). We also relate our findings to Laplacian regularization and suggest to use q-Laplacians as regularizers, where q satisfies 1/p* + 1/q =1."
                },
                {
                    "title": "Predicting the Labelling of a Graph via Minimum $p$-Seminorm Interpolation",
                    "abstract": "We study the problem of predicting the labelling of a graph. The graph is given and a trial sequence of (vertex,label) pairs is then incrementally revealed to the learner. On each trial a vertex is queried and the learner predicts a boolean label. The true label is then returned. The learner\u2019s goal is to minimise mistaken predictions. We propose minimum p-seminorm interpolation to solve this problem. To this end we give a p-seminorm on the space of graph labellings. Thus on every trial we predict using the labelling which minimises the p-seminorm and is also consistent with the revealed (vertex, label) pairs. When p = 2 this is the harmonic energy minimisation procedure of [22], also called (Laplacian) interpolated regularisation in [1]. In the limit as p \u2192 1 this is equivalent to predicting with a label-consistent mincut. We give mistake bounds relative to a label-consistent mincut and a resistive cover of the graph. We say an edge is cut with respect to a labelling if the connected vertices have disagreeing labels. We find that minimising the p-seminorm with p = 1 + where \u2192 0 as the graph diameter D \u2192 \u221e gives a bound of O(\u03a6 logD) versus a bound of O(\u03a6D) when p = 2 where \u03a6 is the number of cut edges."
                },
                {
                    "title": "Contextual Bandits with Similarity Information",
                    "abstract": "In a multi-armed bandit (MAB) problem, an online algorithm makes a sequence of choices. In each round it chooses from a time-invariant set of alternatives and receives the payoff associated with this alternative. While the case of small strategy sets is by now well-understood, a lot of recent work has focused on MAB problems with exponentially or infinitely large strategy sets, where one needs to assume extra structure in order to make the problem tractable. In particular, recent literature considered information on similarity between arms. \n \nWe consider similarity information in the setting of contextual bandits, a natural extension of the basic MAB problem where before each round an algorithm is given the context--a hint about the payoffs in this round. Contextual bandits are directly motivated by placing advertisements on web pages, one of the crucial problems in sponsored search. A particularly simple way to represent similarity information in the contextual bandit setting is via a similarity distance between the context-arm pairs which bounds from above the difference between the respective expected payoffs. \n \nPrior work on contextual bandits with similarity uses \"uniform\" partitions of the similarity space, so that each context-arm pair is approximated by the closest pair in the partition. Algorithms based on \"uniform\" partitions disregard the structure of the payoffs and the context arrivals, which is potentially wasteful. We present algorithms that are based on adaptive partitions, and take advantage of \"benign\" payoffs and context arrivals without sacrificing the worst-case performance. The central idea is to maintain a finer partition in high-payoff regions of the similarity space and in popular regions of the context space. Our results apply to several other settings, e.g., MAB with constrained temporal change (Slivkins and Upfal, 2008) and sleeping bandits (Kleinberg et al., 2008a)."
                },
                {
                    "title": "Collective Classification in Network Data",
                    "abstract": "Many real-world applications produce networked data such as the world-wide web (hypertext documents connected via hyperlinks), social networks (for example, people connected by friendship links), communication networks (computers connected via communication links) and biological networks (for example, protein interaction networks). A recent focus in machine learning research has been to extend traditional machine learning classification techniques to classify nodes in such networks. In this article, we provide a brief introduction to this area of research and how it has progressed during the past decade. We introduce four of the most widely used inference algorithms for classifying networked data and empirically compare them on both synthetic and real-world data."
                },
                {
                    "title": "Fast unfolding of communities in large networks",
                    "abstract": "We propose a simple method to extract the community structure of large networks. Our method is a heuristic method that is based on modularity optimization. It is shown to outperform all other known community detection methods in terms of computation time. Moreover, the quality of the communities detected is very good, as measured by the so-called modularity. This is shown first by identifying language communities in a Belgian mobile phone network of 2 million customers and by analysing a web graph of 118 million nodes and more than one billion links. The accuracy of our algorithm is also verified on ad hoc modular networks."
                },
                {
                    "title": "From External to Internal Regret",
                    "abstract": "External regret compares the performance of an online algorithm, selecting among N actions, to the performance of the best of those actions in hindsight. Internal regret compares the loss of an online algorithm to the loss of a modified online algorithm, which consistently replaces one action by another. In this paper, we give a simple generic reduction that, given an algorithm for the external regret problem, converts it to an efficient online algorithm for the internal regret problem. We provide methods that work both in the full information model, in which the loss of every action is observed at each time step, and the partial information (bandit) model, where at each time step only the loss of the selected action is observed. The importance of internal regret in game theory is due to the fact that in a general game, if each player has sublinear internal regret, then the empirical frequencies converge to a correlated equilibrium. For external regret we also derive a quantitative regret bound for a very general setting of regret, which includes an arbitrary set of modification rules (that possibly modify the online algorithm) and an arbitrary set of time selection functions (each giving different weight to each time step). The regret for a given time selection and modification rule is the difference between the cost of the online algorithm and the cost of the modified online algorithm, where the costs are weighted by the time selection function. This can be viewed as a generalization of the previously-studied sleeping experts setting."
                },
                {
                    "title": "Finding and evaluating community structure in networks.",
                    "abstract": "We propose and study a set of algorithms for discovering community structure in networks-natural divisions of network nodes into densely connected subgroups. Our algorithms all share two definitive features: first, they involve iterative removal of edges from the network to split it into communities, the edges removed being identified using any one of a number of possible \"betweenness\" measures, and second, these measures are, crucially, recalculated after each removal. We also propose a measure for the strength of the community structure found by our algorithms, which gives us an objective metric for choosing the number of communities into which a network should be divided. We demonstrate that our algorithms are highly effective at discovering community structure in both computer-generated and real-world network data, and show how they can be used to shed light on the sometimes dauntingly complex structure of networked systems."
                },
                {
                    "title": "Using and combining predictors that specialize",
                    "abstract": "We study online learning algorithms that predict by com- bining the predictions of several subordinate prediction a lgorithms, sometimes called \"experts.\" These simple algorithms belon g to the multiplicative weights family of algorithms. The performance of these algorithms degrades only logarithmically with the number of experts, making them particularly useful in applications where the number of experts is very large. However, in applications such as text categorization, it is often natural for some of the ex perts to abstain from making predictions on some of the instances. We show how to transform algorithms that assume that all experts are always awake to algorithms that do not require this assumption. We also show how to derive corresponding loss bounds. Our method is very general, and can be applied to a large family of online learning algorithms. We also give applications to various prediction models including decision graphs and \"switching\" experts."
                },
                {
                    "title": "Algorithm 97: Shortest path",
                    "abstract": "procedure ari thmetic (a, b, c, op); in t eger a, b, c, op; \u00a2 o n l m e n t This procedure will perform different order ar i thmetic operations with b and c, put t ing the result in a. The order of the operation is given by op. For op = 1 addit ion is performed. For op = 2 multiplicaLion, repeated addition, is done. Beyond these the operations are non-commutat ive. For op = 3 exponentiat ion, repeated multiplication, is done, raising b to the power c. Beyond these the question of grouping is important . The innermost implied parentheses are at the right. The hyper-exponent is always c. For op = 4 te t ra t ion, repeated exponentiat ion, is done. For op = 5, 6, 7, etc., the procedure performs pentat ion, hexation, heptat ion, etc., respectively. The routine was originally programmed in FORTRAN for the Control Data 160 desk-size computer. The original program was limited to te t ra t ion because subroutine recursiveness in Control Data 160 FORTRAN has been held down to four levels in the interests of economy. The input parameter , b, c, and op, must be positive integers, not zero; b e g i n own i n t e g e r d, e, f, drop; i f o p = 1 t h e n b e g i n a := h-4c; go t o l e n d i f o p = 2 t h e n d := 0; else d := 1; e := c; drop := op 1; for f := I s t e p 1 u n t i l e do b e g i n ari thmetic (a, b, d, drop);"
                },
                {
                    "title": "An optimum character recognition system using decision functions",
                    "abstract": "The character recognition problem, usually resulting from characters being corrupted by printing deterioration and/or inherent noise of the devices, is considered from the viewpoint of statistical decision theory. The optimization consists of minimizing the expected risk for a weight function which is preassigned to measure the consequences of system decisions As an alternative minimization of the error rate for a given rejection rate is used as the critenon. The optimum recogition is thus obtained. The optimum system consists of a conditional-probability densisities computer; character channels, one for each character; a rejection channel; and a comparison network. Its precise structure and and ultimate performance depend essentially upon the signals and noise structure. Explicit examples for an additive Gaussian noise and a ``cosine'' noise are presented. Finally, an error-free recognition system and a possible criterion to measure the character style and deteriortation are presented."
                },
                {
                    "title": "Active Learning of Convex Halfspaces on Graphs",
                    "abstract": "We systematically study the query complexity of learning geodesically convex halfspaces on graphs. Geodesic convexity is a natural generalisation of Euclidean convexity and allows the de\ufb01nition of convex sets and halfspaces on graphs. We prove an upper bound on the query complexity which is linear in the treewidth and the minimum hull set size but only logarithmic in the diameter. We show tight lower bounds along well-established separation axioms and identify the Radon number as a central parameter of the query complexity and the VC dimension. While previous bounds typically depend on the cut size of the labelling, all parameters in our bounds can be computed from the unlabelled graph. We provide evidence that ground-truth communities in real-world graphs are often convex and empirically compare our proposed approach with other active learning algorithms."
                },
                {
                    "title": "A Gang of Adversarial Bandits",
                    "abstract": "We consider running multiple instances of multi-armed bandit (MAB) problems in parallel. A main motivation for this study are online recommendation systems, in which each of N users is associated with a MAB problem and the goal is to exploit users\u2019 similarity in order to learn users\u2019 preferences toK items more efficiently. We consider the adversarial MAB setting, whereby an adversary is free to choose which user and which loss to present to the learner during the learning process. Users are in a social network and the learner is aided by a-priori knowledge of the strengths of the social links between all pairs of users. It is assumed that if the social link between two users is strong then they tend to share the same action. The regret is measured relative to an arbitrary function which maps users to actions. The smoothness of the function is captured by a resistance-based dispersion measure \u03a8. We present two learning algorithms, GABA-I and GABA-II which exploit the network structure to bias towards functions of low \u03a8 values. We show that GABA-I has an expected regret bound of O( \u221a ln(NK/\u03a8)\u03a8KT ) and per-trial time complexity of O(K ln(N)), whilst GABA-II has a weaker O( \u221a ln(N/\u03a8) ln(NK/\u03a8)\u03a8KT ) regret, but a better O(ln(K) ln(N)) per-trial time complexity. We highlight improvements of both algorithms over running independent standard MABs across users."
                },
                {
                    "title": "Achieving All with No Parameters: AdaNormalHedge",
                    "abstract": "We study the classic online learning problem of predicting with expert advice, and propose a truly parameter-free and adaptive algorithm that achieves several objectives simultaneously without using any prior information. The main component of this work is an improved version of the NormalHedge.DT algorithm (Luo and Schapire, 2014), called AdaNormalHedge. On one hand, this new algorithm ensures small regret when the competitor has small loss and almost constant regret when the losses are stochastic. On the other hand, the algorithm is able to compete with any convex combination of the experts simultaneously, with a regret in terms of the relative entropy of the prior and the competitor. This resolves an open problem proposed by Chaudhuri et al. (2009) and Chernov and Vovk (2010). Moreover, we extend the results to the sleeping expert setting and provide two applications to illustrate the power of AdaNormalHedge: 1) competing with time-varying unknown competitors and 2) predicting almost as well as the best pruning tree. Our results on these applications significantly improve previous work from different aspects, and a special case of the first application resolves another open problem proposed by Warmuth and Koolen (2014) on whether one can simultaneously achieve optimal shifting regret for both adversarial and stochastic losses."
                },
                {
                    "title": "The Nonstochastic Multiarmed Bandit Problem",
                    "abstract": "In the multiarmed bandit problem, a gambler must decide which arm of K nonidentical slot machines to play in a sequence of trials so as to maximize his reward. This classical problem has received much attention because of the simple model it provides of the trade-off between exploration (trying out each arm to find the best one) and exploitation (playing the arm believed to give the best payoff). Past solutions for the bandit problem have almost always relied on assumptions about the statistics of the slot machines. \nIn this work, we make no statistical assumptions whatsoever about the nature of the process generating the payoffs of the slot machines. We give a solution to the bandit problem in which an adversary, rather than a well-behaved stochastic process, has complete control over the payoffs. In a sequence of T plays, we prove that the per-round payoff of our algorithm approaches that of the best arm at the rate O(T-1/2). We show by a matching lower bound that this is the best possible. \nWe also prove that our algorithm approaches the per-round payoff of any set of strategies at a similar rate: if the best strategy is chosen from a pool of N strategies, then our algorithm approaches the per-round payoff of the strategy at the rate O((log N1/2 T-1/2). Finally, we apply our results to the problem of playing an unknown repeated matrix game. We show that our algorithm approaches the minimax payoff of the unknown game at the rate O(T-1/2)."
                },
                {
                    "title": "On optimum recognition error and reject tradeoff",
                    "abstract": "The performance of a pattern recognition system is characterized by its error and reject tradeoff. This paper describes an optimum rejection rule and presents a general relation between the error and reject probabilities and some simple properties of the tradeoff in the optimum recognition system. The error rate can be directly evaluated from the reject function. Some practical implications of the results are discussed. Examples in normal distributions and uniform distributions are given."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively utilize abstention in the prediction with expert advice problem under bandit feedback to improve cumulative reward bounds compared to existing algorithms?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of existing bandit algorithms that do not account for the option to abstain, which is particularly relevant in high-stakes applications where incorrect predictions can have severe consequences. By improving the understanding and implementation of abstention in bandit settings, this research could lead to more robust decision-making frameworks, enhancing the safety and economic viability of machine learning applications. Furthermore, it could inspire future research into more sophisticated algorithms that incorporate confidence ratings and abstention, potentially leading to advancements in areas such as online learning, reinforcement learning, and adaptive systems.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to balance exploration and exploitation while incorporating the abstention option effectively. Naive approaches may fail because they might treat abstention as just another action, neglecting the unique implications of choosing not to act when confidence is low. The complexities arise from the need to derive meaningful reward bounds that account for varying confidence levels among predictors, as well as the potential for bounded rewards that do not follow a predictable distribution. Additionally, the algorithm must efficiently aggregate predictions from multiple experts while managing the computational overhead associated with these operations.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely overlooked the integration of abstention in the context of prediction with expert advice, often focusing on scenarios where all actions must be taken. Existing solutions have limitations in their ability to derive cumulative reward bounds when abstention is an option, primarily due to their treatment of abstention as a regular action rather than a strategic choice. Barriers to solving this problem include a lack of algorithms that can effectively leverage confidence ratings and abstention simultaneously. Our approach differs by explicitly incorporating abstention into the learning process and deriving bounds on expected cumulative rewards, thus providing a more nuanced understanding of the problem.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the CBA (Confidence-rated Bandits with Abstentions) algorithm, which utilizes a structured approach to integrate abstention into the prediction process. We will evaluate the algorithm using a dataset that simulates expert predictions with"
            }
        },
        "author_data": {
            "c4c52c2e-15b2-4321-a448-4fb0f994fc1f": {
                "pk": "c4c52c2e-15b2-4321-a448-4fb0f994fc1f",
                "name": "Stephen Pasteris",
                "collaborators": [
                    "Mark Herbster",
                    "Fabio Vitale",
                    "Shiqiang Wang",
                    "Chris Hicks",
                    "Vasilios Mavroudis",
                    "Ting He",
                    "Claudio Gentile",
                    "Lisa Tse",
                    "Hanlin Lu",
                    "Changchang Liu"
                ],
                "domain": [
                    "Online Learning",
                    "Machine Learning",
                    "Bandit Algorithms",
                    "Graph Theory"
                ],
                "publications": [
                    {
                        "title": "A Time and Space Efficient Junction Tree Architecture",
                        "abstract": "The junction tree algorithm is a way of computing marginals of boolean multivariate probability distributions that factorise over sets of random variables. The junction tree algorithm first constructs a tree called a junction tree who's vertices are sets of random variables. The algorithm then performs a generalised version of belief propagation on the junction tree. The Shafer-Shenoy and Hugin architectures are two ways to perform this belief propagation that tradeoff time and space complexities in different ways: Hugin propagation is at least as fast as Shafer-Shenoy propagation and in the cases that we have large vertices of high degree is significantly faster. However, this speed increase comes at the cost of an increased space complexity. This paper first introduces a simple novel architecture, ARCH-1, which has the best of both worlds: the speed of Hugin propagation and the low space requirements of Shafer-Shenoy propagation. A more complicated novel architecture, ARCH-2, is then introduced which has, up to a factor only linear in the maximum cardinality of any vertex, time and space complexities at least as good as ARCH-1 and in the cases that we have large vertices of high degree is significantly faster than ARCH-1."
                    },
                    {
                        "title": "A Hierarchical Nearest Neighbour Approach to Contextual Bandits",
                        "abstract": "In this paper we consider the adversarial contextual bandit problem in metric spaces. The paper \"Nearest neighbour with bandit feedback\" tackled this problem but when there are many contexts near the decision boundary of the comparator policy it suffers from a high regret. In this paper we eradicate this problem, designing an algorithm in which we can hold out any set of contexts when computing our regret term. Our algorithm builds on that of \"Nearest neighbour with bandit feedback\" and hence inherits its extreme computational efficiency."
                    },
                    {
                        "title": "Extraction Propagation",
                        "abstract": "We consider the problem of learning to map large instances, such as sequences and images, to outputs. Since training one large neural network end to end with backpropagation is plagued by vanishing gradients and degradation, we develop a novel neural network architecture called Extraction propagation, which works by training, in parallel, many small neural networks which interact with one another. We note that the performance of Extraction propagation is only conjectured as we have yet to implement it. We do, however, back the algorithm with some theory. A previous version of this paper was entitled \"Fusion encoder networks\" and detailed a slightly different architecture."
                    },
                    {
                        "title": "Nearest Neighbour with Bandit Feedback",
                        "abstract": "In this paper we adapt the nearest neighbour rule to the contextual bandit problem. Our algorithm handles the fully adversarial setting in which no assumptions at all are made about the data-generation process. When combined with a sufficiently fast data-structure for (perhaps approximate) adaptive nearest neighbour search, such as a navigating net, our algorithm is extremely efficient - having a per trial running time polylogarithmic in both the number of trials and actions, and taking only quasi-linear space. We give generic regret bounds for our algorithm and further analyse them when applied to the stochastic bandit problem in euclidean space. We note that our algorithm can also be applied to the online classification problem."
                    },
                    {
                        "title": "Online Convex Optimisation: The Optimal Switching Regret for all Segmentations Simultaneously",
                        "abstract": "We consider the classic problem of online convex optimisation. Whereas the notion of static regret is relevant for stationary problems, the notion of switching regret is more appropriate for non-stationary problems. A switching regret is defined relative to any segmentation of the trial sequence, and is equal to the sum of the static regrets of each segment. In this paper we show that, perhaps surprisingly, we can achieve the asymptotically optimal switching regret on every possible segmentation simultaneously. Our algorithm for doing so is very efficient: having a space and per-trial time complexity that is logarithmic in the time-horizon. Our algorithm also obtains novel bounds on its dynamic regret: being adaptive to variations in the rate of change of the comparator sequence."
                    },
                    {
                        "title": "MaxHedge: Maximising a Maximum Online",
                        "abstract": "We introduce a new online learning framework where, at each trial, the learner is required to select a subset of actions from a given known action set. Each action is associated with an energy value, a reward and a cost. The sum of the energies of the actions selected cannot exceed a given energy budget. The goal is to maximise the cumulative profit, where the profit obtained on a single trial is defined as the difference between the maximum reward among the selected actions and the sum of their costs. Action energy values and the budget are known and fixed. All rewards and costs associated with each action change over time and are revealed at each trial only after the learner's selection of actions. Our framework encompasses several online learning problems where the environment changes over time; and the solution trades-off between minimising the costs and maximising the maximum reward of the selected subset of actions, while being constrained to an action energy budget. The algorithm that we propose is efficient and general in that it may be specialised to multiple natural online combinatorial problems."
                    },
                    {
                        "title": "Online Learning of Facility Locations",
                        "abstract": "In this paper, we provide a rigorous theoretical investigation of an online learning version of the Facility Location problem which is motivated by emerging problems in real-world applications. In our formulation, we are given a set of sites and an online sequence of user requests. At each trial, the learner selects a subset of sites and then incurs a cost for each selected site and an additional cost which is the price of the user's connection to the nearest site in the selected subset. The problem may be solved by an application of the well-known Hedge algorithm. This would, however, require time and space exponential in the number of the given sites, which motivates our design of a novel quasi-linear time algorithm for this problem, with good theoretical guarantees on its performance."
                    },
                    {
                        "title": "Adversarial Online Collaborative Filtering",
                        "abstract": "We investigate the problem of online collaborative filtering under no-repetition constraints, whereby users need to be served content in an online fashion and a given user cannot be recommended the same content item more than once. We start by designing and analyzing an algorithm that works under biclustering assumptions on the user-item preference matrix, and show that this algorithm exhibits an optimal regret guarantee, while being fully adaptive, in that it is oblivious to any prior knowledge about the sequence of users, the universe of items, as well as the biclustering parameters of the preference matrix. We then propose a more robust version of this algorithm which operates with general matrices. Also this algorithm is parameter free, and we prove regret guarantees that scale with the amount by which the preference matrix deviates from a biclustered structure. To our knowledge, these are the first results on online collaborative filtering that hold at this level of generality and adaptivity under no-repetition constraints. Finally, we complement our theoretical findings with simple experiments on real-world datasets aimed at both validating the theory and empirically comparing to standard baselines. This comparison shows the competitive advantage of our approach over these baselines."
                    },
                    {
                        "title": "Online Similarity Prediction of Networked Data from Known and Unknown Graphs",
                        "abstract": "We consider online similarity prediction problems over networked data. We begin by relating this task to the more standard class prediction problem, showing that, given an arbitrary algorithm for class prediction, we can construct an algorithm for similarity prediction with \"nearly\" the same mistake bound, and vice versa. After noticing that this general construction is computationally infeasible, we target our study to {\\em feasible} similarity prediction algorithms on networked data. We initially assume that the network structure is {\\em known} to the learner. Here we observe that Matrix Winnow \\cite{w07} has a near-optimal mistake guarantee, at the price of cubic prediction time per round. This motivates our effort for an efficient implementation of a Perceptron algorithm with a weaker mistake guarantee but with only poly-logarithmic prediction time. Our focus then turns to the challenging case of networks whose structure is initially {\\em unknown} to the learner. In this novel setting, where the network structure is only incrementally revealed, we obtain a mistake-bounded algorithm with a quadratic prediction time per round."
                    },
                    {
                        "title": "On Pairwise Clustering with Side Information",
                        "abstract": "Pairwise clustering, in general, partitions a set of items via a known similarity function. In our treatment, clustering is modeled as a transductive prediction problem. Thus rather than beginning with a known similarity function, the function instead is hidden and the learner only receives a random sample consisting of a subset of the pairwise similarities. An additional set of pairwise side-information may be given to the learner, which then determines the inductive bias of our algorithms. We measure performance not based on the recovery of the hidden similarity function, but instead on how well we classify each item. We give tight bounds on the number of misclassifications. We provide two algorithms. The first algorithm SACA is a simple agglomerative clustering algorithm which runs in near linear time, and which serves as a baseline for our analyses. Whereas the second algorithm, RGCA, enables the incorporation of side-information which may lead to improved bounds at the cost of a longer running time."
                    },
                    {
                        "title": "Online Multitask Learning with Long-Term Memory",
                        "abstract": "We introduce a novel online multitask setting. In this setting each task is partitioned into a sequence of segments that is unknown to the learner. Associated with each segment is a hypothesis from some hypothesis class. We give algorithms that are designed to exploit the scenario where there are many such segments but significantly fewer associated hypotheses. We prove regret bounds that hold for any segmentation of the tasks and any association of hypotheses to the segments. In the single-task setting this is equivalent to switching with long-term memory in the sense of [Bousquet and Warmuth; 2003]. We provide an algorithm that predicts on each trial in time linear in the number of hypotheses when the hypothesis class is finite. We also consider infinite hypothesis classes from reproducing kernel Hilbert spaces for which we give an algorithm whose per trial time complexity is cubic in the number of cumulative trials. In the single-task special case this is the first example of an efficient regret-bounded switching algorithm with long-term memory for a non-parametric hypothesis class."
                    },
                    {
                        "title": "Sum-max Submodular Bandits",
                        "abstract": "Many online decision-making problems correspond to maximizing a sequence of submodular functions. In this work, we introduce sum-max functions, a subclass of monotone submodular functions capturing several interesting problems, including best-of-$K$-bandits, combinatorial bandits, and the bandit versions on facility location, $M$-medians, and hitting sets. We show that all functions in this class satisfy a key property that we call pseudo-concavity. This allows us to prove $\\big(1 - \\frac{1}{e}\\big)$-regret bounds for bandit feedback in the nonstochastic setting of the order of $\\sqrt{MKT}$ (ignoring log factors), where $T$ is the time horizon and $M$ is a cardinality constraint. This bound, attained by a simple and efficient algorithm, significantly improves on the $\\widetilde{O}\\big(T^{2/3}\\big)$ regret bound for online monotone submodular maximization with bandit feedback."
                    },
                    {
                        "title": "Service Placement with Provable Guarantees in Heterogeneous Edge Computing Systems",
                        "abstract": "Mobile edge computing (MEC) is a promising technique for providing low-latency access to services at the network edge. The services are hosted at various types of edge nodes with both computation and communication capabilities. Due to the heterogeneity of edge node characteristics and user locations, the performance of MEC varies depending on where the service is hosted. In this paper, we consider such a heterogeneous MEC system, and focus on the problem of placing multiple services in the system to maximize the total reward. We show that the problem is NP-hard via reduction from the set cover problem, and propose a deterministic approximation algorithm to solve the problem, which has an approximation ratio that is not worse than $\\left(1-e^{-1}\\right)/4$. The proposed algorithm is based on two sub-routines that are suitable for small and arbitrarily sized services, respectively. The algorithm is designed using a novel way of partitioning each edge node into multiple slots, where each slot contains one service. The approximation guarantee is obtained via a specialization of the method of conditional expectations, which uses a randomized procedure as an intermediate step. In addition to theoretical guarantees, simulation results also show that the proposed algorithm outperforms other state-of-the-art approaches."
                    },
                    {
                        "title": "Online Matrix Completion with Side Information",
                        "abstract": "We give an online algorithm and prove novel mistake and regret bounds for online binary matrix completion with side information. The mistake bounds we prove are of the form $\\tilde{O}(D/\\gamma^2)$. The term $1/\\gamma^2$ is analogous to the usual margin term in SVM (perceptron) bounds. More specifically, if we assume that there is some factorization of the underlying $m \\times n$ matrix into $P Q^\\intercal$ where the rows of $P$ are interpreted as \"classifiers\" in $\\mathcal{R}^d$ and the rows of $Q$ as \"instances\" in $\\mathcal{R}^d$, then $\\gamma$ is the maximum (normalized) margin over all factorizations $P Q^\\intercal$ consistent with the observed matrix. The quasi-dimension term $D$ measures the quality of side information. In the presence of vacuous side information, $D= m+n$. However, if the side information is predictive of the underlying factorization of the matrix, then in an ideal case, $D \\in O(k + \\ell)$ where $k$ is the number of distinct row factors and $\\ell$ is the number of distinct column factors. We additionally provide a generalization of our algorithm to the inductive setting. In this setting, we provide an example where the side information is not directly specified in advance. For this example, the quasi-dimension $D$ is now bounded by $O(k^2 + \\ell^2)$."
                    },
                    {
                        "title": "Joint Coreset Construction and Quantization for Distributed Machine Learning",
                        "abstract": "Coresets are small, weighted summaries of larger datasets, aiming at providing provable error bounds for machine learning (ML) tasks while significantly reducing the communication and computation costs. To achieve a better trade-off between ML error bounds and costs, we propose the first framework to incorporate quantization techniques into the process of coreset construction. Specifically, we theoretically analyze the ML error bounds caused by a combination of coreset construction and quantization. Based on that, we formulate an optimization problem to minimize the ML error under a fixed budget of communication cost. To improve the scalability for large datasets, we identify two proxies of the original objective function, for which efficient algorithms are developed. For the case of data on multiple nodes, we further design a novel algorithm to allocate the communication budget to the nodes while minimizing the overall ML error. Through extensive experiments on multiple real-world datasets, we demonstrate the effectiveness and efficiency of our proposed algorithms for a variety of ML tasks. In particular, our algorithms have achieved more than 90% data reduction with less than 10% degradation in ML performance in most cases."
                    },
                    {
                        "title": "Communication-efficient k-Means for Edge-based Machine Learning",
                        "abstract": "We consider the problem of computing the k-means centers for a large high-dimensional dataset in the context of edge-based machine learning, where data sources offload machine learning computation to nearby edge servers. k-Means computation is fundamental to many data analytics, and the capability of computing provably accurate k-means centers by leveraging the computation power of the edge servers, at a low communication and computation cost to the data sources, will greatly improve the performance of these analytics. We propose to let the data sources send small summaries, generated by joint dimensionality reduction (DR), cardinality reduction (CR), and quantization (QT), to support approximate k-means computation at reduced complexity and communication cost. By analyzing the complexity, the communication cost, and the approximation error of k-means algorithms based on carefully designed composition of DR/CR/QT methods, we show that: (i) it is possible to compute near-optimal k-means centers at a near-linear complexity and a constant or logarithmic communication cost, (ii) the order of applying DR and CR significantly affects the complexity and the communication cost, and (iii) combining DR/CR methods with a properly configured quantizer can further reduce the communication cost without compromising the other performance metrics. Our theoretical analysis has been validated through experiments based on real datasets."
                    }
                ]
            },
            "5c55f8b7-14b2-41e1-a0cb-77ba91a782d4": {
                "pk": "5c55f8b7-14b2-41e1-a0cb-77ba91a782d4",
                "name": "Alberto Rumi",
                "collaborators": [
                    "Fabio Vitale",
                    "Nicol\u00f2 Cesa-Bianchi",
                    "Stephen Pasteris",
                    "Yuko Kuroki",
                    "Taira Tsuchiya"
                ],
                "domain": [
                    "Submodular Optimization",
                    "Bandit Algorithms",
                    "Online Learning",
                    "Contextual Bandits"
                ],
                "publications": [
                    {
                        "title": "Sum-max Submodular Bandits",
                        "abstract": "Many online decision-making problems correspond to maximizing a sequence of submodular functions. In this work, we introduce sum-max functions, a subclass of monotone submodular functions capturing several interesting problems, including best-of-$K$-bandits, combinatorial bandits, and the bandit versions on facility location, $M$-medians, and hitting sets. We show that all functions in this class satisfy a key property that we call pseudo-concavity. This allows us to prove $\\big(1 - \\frac{1}{e}\\big)$-regret bounds for bandit feedback in the nonstochastic setting of the order of $\\sqrt{MKT}$ (ignoring log factors), where $T$ is the time horizon and $M$ is a cardinality constraint. This bound, attained by a simple and efficient algorithm, significantly improves on the $\\widetilde{O}\\big(T^{2/3}\\big)$ regret bound for online monotone submodular maximization with bandit feedback."
                    },
                    {
                        "title": "Best-of-Both-Worlds Algorithms for Linear Contextual Bandits",
                        "abstract": "We study best-of-both-worlds algorithms for $K$-armed linear contextual bandits. Our algorithms deliver near-optimal regret bounds in both the adversarial and stochastic regimes, without prior knowledge about the environment. In the stochastic regime, we achieve the polylogarithmic rate $\\frac{(dK)^2\\mathrm{poly}\\log(dKT)}{\\Delta_{\\min}}$, where $\\Delta_{\\min}$ is the minimum suboptimality gap over the $d$-dimensional context space. In the adversarial regime, we obtain either the first-order $\\widetilde{O}(dK\\sqrt{L^*})$ bound, or the second-order $\\widetilde{O}(dK\\sqrt{\\Lambda^*})$ bound, where $L^*$ is the cumulative loss of the best action and $\\Lambda^*$ is a notion of the cumulative second moment for the losses incurred by the algorithm. Moreover, we develop an algorithm based on FTRL with Shannon entropy regularizer that does not require the knowledge of the inverse of the covariance matrix, and achieves a polylogarithmic regret in the stochastic regime while obtaining $\\widetilde{O}\\big(dK\\sqrt{T}\\big)$ regret bounds in the adversarial regime."
                    }
                ]
            },
            "31d8b37b-71b6-450b-8af3-133dca6734a6": {
                "pk": "31d8b37b-71b6-450b-8af3-133dca6734a6",
                "name": "Maximilian Thiessen",
                "collaborators": [
                    "Marco Bressan",
                    "Emmanuel Esposito",
                    "Nicol\u00f2 Cesa-Bianchi",
                    "Pascal Welke",
                    "Fabian Jogl",
                    "Thomas G\u00e4rtner",
                    "Georgy Sokolov",
                    "Margarita Akhmejanova",
                    "Fabio Vitale",
                    "Francesco Orabona"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Active Learning",
                    "Interpretability"
                ],
                "publications": [
                    {
                        "title": "Expectation-Complete Graph Representations with Homomorphisms",
                        "abstract": "We investigate novel random graph embeddings that can be computed in expected polynomial time and that are able to distinguish all non-isomorphic graphs in expectation. Previous graph embeddings have limited expressiveness and either cannot distinguish all graphs or cannot be computed efficiently for every graph. To be able to approximate arbitrary functions on graphs, we are interested in efficient alternatives that become arbitrarily expressive with increasing resources. Our approach is based on Lov\\'asz' characterisation of graph isomorphism through an infinite dimensional vector of homomorphism counts. Our empirical evaluation shows competitive results on several benchmark graph learning tasks."
                    },
                    {
                        "title": "Efficient Algorithms for Learning Monophonic Halfspaces in Graphs",
                        "abstract": "We study the problem of learning a binary classifier on the vertices of a graph. In particular, we consider classifiers given by monophonic halfspaces, partitions of the vertices that are convex in a certain abstract sense. Monophonic halfspaces, and related notions such as geodesic halfspaces,have recently attracted interest, and several connections have been drawn between their properties(e.g., their VC dimension) and the structure of the underlying graph $G$. We prove several novel results for learning monophonic halfspaces in the supervised, online, and active settings. Our main result is that a monophonic halfspace can be learned with near-optimal passive sample complexity in time polynomial in $n = |V(G)|$. This requires us to devise a polynomial-time algorithm for consistent hypothesis checking, based on several structural insights on monophonic halfspaces and on a reduction to $2$-satisfiability. We prove similar results for the online and active settings. We also show that the concept class can be enumerated with delay $\\operatorname{poly}(n)$, and that empirical risk minimization can be performed in time $2^{\\omega(G)}\\operatorname{poly}(n)$ where $\\omega(G)$ is the clique number of $G$. These results answer open questions from the literature (Gonz\\'alez et al., 2020), and show a contrast with geodesic halfspaces, for which some of the said problems are NP-hard (Seiffarth et al., 2023)."
                    },
                    {
                        "title": "Self-Directed Learning of Convex Labelings on Graphs",
                        "abstract": "We study the problem of learning the clusters of a given graph in the self-directed learning setup. This learning setting is a variant of online learning, where rather than an adversary determining the sequence in which nodes are presented, the learner autonomously and adaptively selects them. While self-directed learning of Euclidean halfspaces, linear functions, and general abstract multi-class hypothesis classes was recently considered, no results previously existed specifically for self-directed node classification on graphs. In this paper, we address this problem developing efficient algorithms for it. More specifically, we focus on the case of (geodesically) convex clusters, i.e., for every two nodes sharing the same label, all nodes on every shortest path between them also share the same label. In particular, we devise a polynomial-time algorithm that makes only $3(h(G)+1)^4 \\ln n$ mistakes on graphs with two convex clusters, where $n$ is the total number of nodes and $h(G)$ is the Hadwiger number, i.e., the size of the largest clique minor of the graph $G$. We also show that our algorithm is robust to the case that clusters are slightly non-convex, still achieving a mistake bound logarithmic in $n$. Finally, for the more standard case of homophilic clusters, where strongly connected nodes tend to belong the same class, we devise a simple and efficient algorithm."
                    },
                    {
                        "title": "Active Learning of Classifiers with Label and Seed Queries",
                        "abstract": "We study exact active learning of binary and multiclass classifiers with margin. Given an $n$-point set $X \\subset \\mathbb{R}^m$, we want to learn any unknown classifier on $X$ whose classes have finite strong convex hull margin, a new notion extending the SVM margin. In the standard active learning setting, where only label queries are allowed, learning a classifier with strong convex hull margin $\\gamma$ requires in the worst case $\\Omega\\big(1+\\frac{1}{\\gamma}\\big)^{(m-1)/2}$ queries. On the other hand, using the more powerful seed queries (a variant of equivalence queries), the target classifier could be learned in $O(m \\log n)$ queries via Littlestone's Halving algorithm; however, Halving is computationally inefficient. In this work we show that, by carefully combining the two types of queries, a binary classifier can be learned in time $\\operatorname{poly}(n+m)$ using only $O(m^2 \\log n)$ label queries and $O\\big(m \\log \\frac{m}{\\gamma}\\big)$ seed queries; the result extends to $k$-class classifiers at the price of a $k!k^2$ multiplicative overhead. Similar results hold when the input points have bounded bit complexity, or when only one class has strong convex hull margin against the rest. We complement the upper bounds by showing that in the worst case any algorithm needs $\\Omega\\big(k m \\log \\frac{1}{\\gamma}\\big)$ seed and label queries to learn a $k$-class classifier with strong convex hull margin $\\gamma$."
                    },
                    {
                        "title": "A Theory of Interpretable Approximations",
                        "abstract": "Can a deep neural network be approximated by a small decision tree based on simple features? This question and its variants are behind the growing demand for machine learning models that are *interpretable* by humans. In this work we study such questions by introducing *interpretable approximations*, a notion that captures the idea of approximating a target concept $c$ by a small aggregation of concepts from some base class $\\mathcal{H}$. In particular, we consider the approximation of a binary concept $c$ by decision trees based on a simple class $\\mathcal{H}$ (e.g., of bounded VC dimension), and use the tree depth as a measure of complexity. Our primary contribution is the following remarkable trichotomy. For any given pair of $\\mathcal{H}$ and $c$, exactly one of these cases holds: (i) $c$ cannot be approximated by $\\mathcal{H}$ with arbitrary accuracy; (ii) $c$ can be approximated by $\\mathcal{H}$ with arbitrary accuracy, but there exists no universal rate that bounds the complexity of the approximations as a function of the accuracy; or (iii) there exists a constant $\\kappa$ that depends only on $\\mathcal{H}$ and $c$ such that, for *any* data distribution and *any* desired accuracy level, $c$ can be approximated by $\\mathcal{H}$ with a complexity not exceeding $\\kappa$. This taxonomy stands in stark contrast to the landscape of supervised classification, which offers a complex array of distribution-free and universally learnable scenarios. We show that, in the case of interpretable approximations, even a slightly nontrivial a-priori guarantee on the complexity of approximations implies approximations with constant (distribution-free and accuracy-free) complexity. We extend our trichotomy to classes $\\mathcal{H}$ of unbounded VC dimension and give characterizations of interpretability based on the algebra generated by $\\mathcal{H}$."
                    },
                    {
                        "title": "Generalized Laplacian Positional Encoding for Graph Representation Learning",
                        "abstract": "Graph neural networks (GNNs) are the primary tool for processing graph-structured data. Unfortunately, the most commonly used GNNs, called Message Passing Neural Networks (MPNNs) suffer from several fundamental limitations. To overcome these limitations, recent works have adapted the idea of positional encodings to graph data. This paper draws inspiration from the recent success of Laplacian-based positional encoding and defines a novel family of positional encoding schemes for graphs. We accomplish this by generalizing the optimization problem that defines the Laplace embedding to more general dissimilarity functions rather than the 2-norm used in the original formulation. This family of positional encodings is then instantiated by considering p-norms. We discuss a method for calculating these positional encoding schemes, implement it in PyTorch and demonstrate how the resulting positional encoding captures different properties of the graph. Furthermore, we demonstrate that this novel family of positional encodings can improve the expressive power of MPNNs. Lastly, we present preliminary experimental results."
                    }
                ]
            },
            "c1aa932e-0e61-4ddf-947e-c598c6fa20bd": {
                "pk": "c1aa932e-0e61-4ddf-947e-c598c6fa20bd",
                "name": "Shota Saito",
                "collaborators": [
                    "Toshiyasu Matsushima",
                    "Shinichi Shirakawa",
                    "Ryoki Hamano",
                    "Masahiro Nomura",
                    "Yuta Nakahara",
                    "Akira Kamatsuka",
                    "Mark Herbster",
                    "Hideki Yagi",
                    "Danilo P Mandic",
                    "Hideyuki Suzuki"
                ],
                "domain": [
                    "Graph Theory",
                    "Information Theory",
                    "Neural Networks",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Wess-Zumino-Witten Terms of $Sp$ QCD by Bordism Theory",
                        "abstract": "We investigate the four-dimensional Wess-Zumino-Witten (WZW) terms within the framework of $Sp$ quantum chromodynamics (QCD) using invertible field theory through bordism theory. We present a novel approach aimed at circumventing both perturbative and non-perturbative gauge anomalies on spacetime manifolds endowed with spin structures. We study both ungauged and gauged WZW terms including the problems of the topological consistency of gauged WZW terms."
                    },
                    {
                        "title": "Soft Guessing Under Log-Loss Distortion Allowing Errors",
                        "abstract": "This paper deals with the problem of soft guessing under log-loss distortion (logarithmic loss) that was recently investigated by [Wu and Joudeh, IEEE ISIT, pp. 466--471, 2023]. We extend this problem to soft guessing allowing errors, i.e., at each step, a guesser decides whether to stop the guess or not with some probability and if the guesser stops guessing, then the guesser declares an error. We show that the minimal expected value of the cost of guessing under the constraint of the error probability is characterized by the smooth R\\'enyi entropy. Furthermore, we carry out an asymptotic analysis for a stationary and memoryless source."
                    },
                    {
                        "title": "Hypergraph Modeling via Spectral Embedding Connection: Hypergraph Cut, Weighted Kernel $k$-means, and Heat Kernel",
                        "abstract": "We propose a theoretical framework of multi-way similarity to model real-valued data into hypergraphs for clustering via spectral embedding. For graph cut based spectral clustering, it is common to model real-valued data into graph by modeling pairwise similarities using kernel function. This is because the kernel function has a theoretical connection to the graph cut. For problems where using multi-way similarities are more suitable than pairwise ones, it is natural to model as a hypergraph, which is generalization of a graph. However, although the hypergraph cut is well-studied, there is not yet established a hypergraph cut based framework to model multi-way similarity. In this paper, we formulate multi-way similarities by exploiting the theoretical foundation of kernel function. We show a theoretical connection between our formulation and hypergraph cut in two ways, generalizing both weighted kernel $k$-means and the heat kernel, by which we justify our formulation. We also provide a fast algorithm for spectral clustering. Our algorithm empirically shows better performance than existing graph and other heuristic modeling methods."
                    },
                    {
                        "title": "Non-Asymptotic Fundamental Limits of Guessing Subject to Distortion",
                        "abstract": "This paper investigates the problem of guessing subject to distortion, which was introduced by Arikan and Merhav. While the primary concern of the previous study was asymptotic analysis, our primary concern is non-asymptotic analysis. We prove non-asymptotic achievability and converse bounds of the moment of the number of guesses without side information (resp. with side information) by using a quantity based on the R\\'enyi entropy (resp. the Arimoto-R\\'enyi conditional entropy). Also, we introduce an error probability and show similar results. Further, from our bounds, we derive a single-letter characterization of the asymptotic exponent of guessing moment for a stationary memoryless source."
                    },
                    {
                        "title": "Cumulant Generating Function of Codeword Lengths in Variable-Length Lossy Compression Allowing Positive Excess Distortion Probability",
                        "abstract": "This paper considers the problem of variable-length lossy source coding. The performance criteria are the excess distortion probability and the cumulant generating function of codeword lengths. We derive a non-asymptotic fundamental limit of the cumulant generating function of codeword lengths allowing positive excess distortion probability. It is shown that the achievability and converse bounds are characterized by the R\\'enyi entropy-based quantity. In the proof of the achievability result, the explicit code construction is provided. Further, we investigate an asymptotic single-letter characterization of the fundamental limit for a stationary memoryless source."
                    },
                    {
                        "title": "Controlling Model Complexity in Probabilistic Model-Based Dynamic Optimization of Neural Network Structures",
                        "abstract": "A method of simultaneously optimizing both the structure of neural networks and the connection weights in a single training loop can reduce the enormous computational cost of neural architecture search. We focus on the probabilistic model-based dynamic neural network structure optimization that considers the probability distribution of structure parameters and simultaneously optimizes both the distribution parameters and connection weights based on gradient methods. Since the existing algorithm searches for the structures that only minimize the training loss, this method might find overly complicated structures. In this paper, we propose the introduction of a penalty term to control the model complexity of obtained structures. We formulate a penalty term using the number of weights or units and derive its analytical natural gradient. The proposed method minimizes the objective function injected the penalty term based on the stochastic gradient descent. We apply the proposed method in the unit selection of a fully-connected neural network and the connection selection of a convolutional neural network. The experimental results show that the proposed method can control model complexity while maintaining performance."
                    },
                    {
                        "title": "Evaluation of Error Probability of Classification Based on the Analysis of the Bayes Code: Extension and Example",
                        "abstract": "Suppose that we have two training sequences generated by parametrized distributions $P_{\\theta^*}$ and $P_{\\xi^*}$, where $\\theta^*$ and $\\xi^*$ are unknown true parameters. Given training sequences, we study the problem of classifying whether a test sequence was generated according to $P_{\\theta^*}$ or $P_{\\xi^*}$. This problem can be thought of as a hypothesis testing problem and our aim is to analyze the weighted sum of type-I and type-II error probabilities. Utilizing the analysis of the codeword lengths of the Bayes code, our previous study derived more refined bounds on the error probability than known previously. However, our previous study had the following deficiencies: i) the prior distributions of $\\theta$ and $\\xi$ are the same; ii) the prior distributions of two hypotheses are uniform; iii) no numerical calculation at finite blocklength. This study solves these problems. We remove the restrictions i) and ii) and derive more general results than obtained previously. To deal with problem iii), we perform a numerical calculation for a concrete model."
                    },
                    {
                        "title": "Multi-class Graph Clustering via Approximated Effective $p$-Resistance",
                        "abstract": "This paper develops an approximation to the (effective) $p$-resistance and applies it to multi-class clustering. Spectral methods based on the graph Laplacian and its generalization to the graph $p$-Laplacian have been a backbone of non-euclidean clustering techniques. The advantage of the $p$-Laplacian is that the parameter $p$ induces a controllable bias on cluster structure. The drawback of $p$-Laplacian eigenvector based methods is that the third and higher eigenvectors are difficult to compute. Thus, instead, we are motivated to use the $p$-resistance induced by the $p$-Laplacian for clustering. For $p$-resistance, small $p$ biases towards clusters with high internal connectivity while large $p$ biases towards clusters of small \"extent,\" that is a preference for smaller shortest-path distances between vertices in the cluster. However, the $p$-resistance is expensive to compute. We overcome this by developing an approximation to the $p$-resistance. We prove upper and lower bounds on this approximation and observe that it is exact when the graph is a tree. We also provide theoretical justification for the use of $p$-resistance for clustering. Finally, we provide experiments comparing our approximated $p$-resistance clustering to other $p$-Laplacian based methods."
                    },
                    {
                        "title": "Variable-Length Lossy Compression Allowing Positive Overflow and Excess Distortion Probabilities",
                        "abstract": "This paper investigates the problem of variable-length lossy source coding allowing a positive excess distortion probability and an overflow probability of codeword lengths. Novel one-shot achievability and converse bounds of the optimal rate are established by a new quantity based on the smooth max entropy (the smooth R\\'enyi entropy of order zero). To derive the achievability bounds, we give an explicit code construction based on a distortion ball instead of using the random coding argument. The basic idea of the code construction is similar to the optimal code construction in the variable-length lossless source coding. Our achievability bounds are slightly different, depending on whether the encoder is stochastic or deterministic. One-shot results yield a general formula of the optimal rate for blocklength $n$. In addition, our general formula is applied to asymptotic analysis for a stationary memoryless source. As a result, we derive a single-letter characterization of the optimal rate by using the rate-distortion and rate-dispersion functions."
                    },
                    {
                        "title": "Hypergraph $p$-Laplacian: A Differential Geometry View",
                        "abstract": "The graph Laplacian plays key roles in information processing of relational data, and has analogies with the Laplacian in differential geometry. In this paper, we generalize the analogy between graph Laplacian and differential geometry to the hypergraph setting, and propose a novel hypergraph $p$-Laplacian. Unlike the existing two-node graph Laplacians, this generalization makes it possible to analyze hypergraphs, where the edges are allowed to connect any number of nodes. Moreover, we propose a semi-supervised learning method based on the proposed hypergraph $p$-Laplacian, and formalize them as the analogue to the Dirichlet problem, which often appears in physics. We further explore theoretical connections to normalized hypergraph cut on a hypergraph, and propose normalized cut corresponding to hypergraph $p$-Laplacian. The proposed $p$-Laplacian is shown to outperform standard hypergraph Laplacians in the experiment on a hypergraph semi-supervised learning and normalized cut setting."
                    },
                    {
                        "title": "CMA-ES with Margin: Lower-Bounding Marginal Probability for Mixed-Integer Black-Box Optimization",
                        "abstract": "This study targets the mixed-integer black-box optimization (MI-BBO) problem where continuous and integer variables should be optimized simultaneously. The CMA-ES, our focus in this study, is a population-based stochastic search method that samples solution candidates from a multivariate Gaussian distribution (MGD), which shows excellent performance in continuous BBO. The parameters of MGD, mean and (co)variance, are updated based on the evaluation value of candidate solutions in the CMA-ES. If the CMA-ES is applied to the MI-BBO with straightforward discretization, however, the variance corresponding to the integer variables becomes much smaller than the granularity of the discretization before reaching the optimal solution, which leads to the stagnation of the optimization. In particular, when binary variables are included in the problem, this stagnation more likely occurs because the granularity of the discretization becomes wider, and the existing modification to the CMA-ES does not address this stagnation. To overcome these limitations, we propose a simple modification of the CMA-ES based on lower-bounding the marginal probabilities associated with the generation of integer variables in the MGD. The numerical experiments on the MI-BBO benchmark problems demonstrate the efficiency and robustness of the proposed method."
                    },
                    {
                        "title": "Efficient Search of Multiple Neural Architectures with Different Complexities via Importance Sampling",
                        "abstract": "Neural architecture search (NAS) aims to automate architecture design processes and improve the performance of deep neural networks. Platform-aware NAS methods consider both performance and complexity and can find well-performing architectures with low computational resources. Although ordinary NAS methods result in tremendous computational costs owing to the repetition of model training, one-shot NAS, which trains the weights of a supernetwork containing all candidate architectures only once during the search process, has been reported to result in a lower search cost. This study focuses on the architecture complexity-aware one-shot NAS that optimizes the objective function composed of the weighted sum of two metrics, such as the predictive performance and number of parameters. In existing methods, the architecture search process must be run multiple times with different coefficients of the weighted sum to obtain multiple architectures with different complexities. This study aims at reducing the search cost associated with finding multiple architectures. The proposed method uses multiple distributions to generate architectures with different complexities and updates each distribution using the samples obtained from multiple distributions based on importance sampling. The proposed method allows us to obtain multiple architectures with different complexities in a single architecture search, resulting in reducing the search cost. The proposed method is applied to the architecture search of convolutional neural networks on the CIAFR-10 and ImageNet datasets. Consequently, compared with baseline methods, the proposed method finds multiple architectures with varying complexities while requiring less computational effort."
                    },
                    {
                        "title": "Marginal Probability-Based Integer Handling for CMA-ES Tackling Single-and Multi-Objective Mixed-Integer Black-Box Optimization",
                        "abstract": "This study targets the mixed-integer black-box optimization (MI-BBO) problem where continuous and integer variables should be optimized simultaneously. The CMA-ES, our focus in this study, is a population-based stochastic search method that samples solution candidates from a multivariate Gaussian distribution (MGD), which shows excellent performance in continuous BBO. The parameters of MGD, mean and (co)variance, are updated based on the evaluation value of candidate solutions in the CMA-ES. If the CMA-ES is applied to the MI-BBO with straightforward discretization, however, the variance corresponding to the integer variables becomes much smaller than the granularity of the discretization before reaching the optimal solution, which leads to the stagnation of the optimization. In particular, when binary variables are included in the problem, this stagnation more likely occurs because the granularity of the discretization becomes wider, and the existing integer handling for the CMA-ES does not address this stagnation. To overcome these limitations, we propose a simple integer handling for the CMA-ES based on lower-bounding the marginal probabilities associated with the generation of integer variables in the MGD. The numerical experiments on the MI-BBO benchmark problems demonstrate the efficiency and robustness of the proposed method. Furthermore, in order to demonstrate the generality of the idea of the proposed method, in addition to the single-objective optimization case, we incorporate it into multi-objective CMA-ES and verify its performance on bi-objective mixed-integer benchmark problems."
                    },
                    {
                        "title": "Variable-Length Intrinsic Randomness Allowing Positive Value of the Average Variational Distance",
                        "abstract": "This paper considers the problem of variable-length intrinsic randomness. We propose the average variational distance as the performance criterion from the viewpoint of a dual relationship with the problem formulation of variable-length resolvability. Previous study has derived the general formula of the $\\epsilon$-variable-length resolvability. We derive the general formula of the $\\epsilon$-variable-length intrinsic randomness. Namely, we characterize the supremum of the mean length under the constraint that the value of the average variational distance is smaller than or equal to a constant $\\epsilon$. Our result clarifies a dual relationship between the general formula of $\\epsilon$-variable-length resolvability and that of $\\epsilon$-variable-length intrinsic randomness. We also derive a lower bound of the quantity characterizing our general formula."
                    },
                    {
                        "title": "An Efficient Bayes Coding Algorithm for the Non-Stationary Source in Which Context Tree Model Varies from Interval to Interval",
                        "abstract": "The context tree source is a source model in which the occurrence probability of symbols is determined from a finite past sequence, and is a broader class of sources that includes i.i.d. and Markov sources. The proposed source model in this paper represents that a subsequence in each interval is generated from a different context tree model. The Bayes code for such sources requires weighting of the posterior probability distributions for the change patterns of the context tree source and for all possible context tree models. Therefore, the challenge is how to reduce this exponential order computational complexity. In this paper, we assume a special class of prior probability distribution of change patterns and context tree models, and propose an efficient Bayes coding algorithm whose computational complexity is the polynomial order."
                    },
                    {
                        "title": "Probability Distribution on Rooted Trees",
                        "abstract": "The hierarchical and recursive expressive capability of rooted trees is applicable to represent statistical models in various areas, such as data compression, image processing, and machine learning. On the other hand, such hierarchical expressive capability causes a problem in tree selection to avoid overfitting. One unified approach to solve this is a Bayesian approach, on which the rooted tree is regarded as a random variable and a direct loss function can be assumed on the selected model or the predicted value for a new data point. However, all the previous studies on this approach are based on the probability distribution on full trees, to the best of our knowledge. In this paper, we propose a generalized probability distribution for any rooted trees in which only the maximum number of child nodes and the maximum depth are fixed. Furthermore, we derive recursive methods to evaluate the characteristics of the probability distribution without any approximations."
                    },
                    {
                        "title": "Probability Distribution on Full Rooted Trees",
                        "abstract": "The recursive and hierarchical structure of full rooted trees is applicable to represent statistical models in various areas, such as data compression, image processing, and machine learning. In most of these cases, the full rooted tree is not a random variable; as such, model selection to avoid overfitting becomes problematic. A method to solve this problem is to assume a prior distribution on the full rooted trees. This enables the optimal model selection based on the Bayes decision theory. For example, by assigning a low prior probability to a complex model, the maximum a posteriori estimator prevents the selection of the complex one. Furthermore, we can average all the models weighted by their posteriors. In this paper, we propose a probability distribution on a set of full rooted trees. Its parametric representation is suitable for calculating the properties of our distribution using recursive functions, such as the mode, expectation, and posterior distribution. Although such distributions have been proposed in previous studies, they are only applicable to specific applications. Therefore, we extract their mathematically essential components and derive new generalized methods to calculate the expectation, posterior distribution, etc."
                    },
                    {
                        "title": "Parameterless Stochastic Natural Gradient Method for Discrete Optimization and its Application to Hyper-Parameter Optimization for Neural Network",
                        "abstract": "Black box discrete optimization (BBDO) appears in wide range of engineering tasks. Evolutionary or other BBDO approaches have been applied, aiming at automating necessary tuning of system parameters, such as hyper parameter tuning of machine learning based systems when being installed for a specific task. However, automation is often jeopardized by the need of strategy parameter tuning for BBDO algorithms. An expert with the domain knowledge must undergo time-consuming strategy parameter tuning. This paper proposes a parameterless BBDO algorithm based on information geometric optimization, a recent framework for black box optimization using stochastic natural gradient. Inspired by some theoretical implications, we develop an adaptation mechanism for strategy parameters of the stochastic natural gradient method for discrete search domains. The proposed algorithm is evaluated on commonly used test problems. It is further extended to two examples of simultaneous optimization of the hyper parameters and the connection weights of deep learning models, leading to a faster optimization than the existing approaches without any effort of parameter tuning."
                    }
                ]
            },
            "5f7e2610-43f4-4b09-91ae-21fa225d6725": {
                "pk": "5f7e2610-43f4-4b09-91ae-21fa225d6725",
                "name": "Atsushi Miyauchi",
                "collaborators": [
                    "Yasushi Kawase",
                    "Francesco Bonchi",
                    "Akiko Takeda",
                    "Yuko Kuroki",
                    "Tomomi Matsui",
                    "Noriyoshi Sukegawa",
                    "Yuuki Takai",
                    "Masahiro Ikeda",
                    "Yuichi Yoshida",
                    "Charalampos E. Tsourakakis"
                ],
                "domain": [
                    "Graph Theory",
                    "Community Detection",
                    "Quantum Computing",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Linear Order Matrix Inversion Method with Help from Quantum Searching Algorithm",
                        "abstract": "Presented here is a matrix inversion method utilizing quantum searching algorithm. In this method, huge Hilbert space as a whole spanned by myriad of eigen states is searched and evaluated efficiently by sequential reduction in dimension one by one. Total iteration steps required for search are proportional to the number of unknown variables. Our method could solve very large linear equations with sufficiently high probability faster than any existing classical algorithms, which roughly depends on the cube of unknown variables."
                    },
                    {
                        "title": "Fractional programming formulation for the vertex coloring problem",
                        "abstract": "We devise a new formulation for the vertex coloring problem. Different from other formulations, decision variables are associated with the pairs of vertices. Consequently, colors will be distinguishable. Although the objective function is fractional, it can be replaced by a piece-wise linear convex function. Numerical experiments show that our formulation has significantly good performance for dense graphs."
                    },
                    {
                        "title": "Maximizing Barber's bipartite modularity is also hard",
                        "abstract": "Modularity introduced by Newman and Girvan [Phys. Rev. E 69, 026113 (2004)] is a quality function for community detection. Numerous methods for modularity maximization have been developed so far. In 2007, Barber [Phys. Rev. E 76, 066102 (2007)] introduced a variant of modularity called bipartite modularity which is appropriate for bipartite networks. Although maximizing the standard modularity is known to be NP-hard, the computational complexity of maximizing bipartite modularity has yet to be revealed. In this study, we prove that maximizing bipartite modularity is also NP-hard. More specifically, we show the NP-completeness of its decision version by constructing a reduction from a classical partitioning problem."
                    },
                    {
                        "title": "Z-score-based modularity for community detection in networks",
                        "abstract": "Identifying community structure in networks is an issue of particular interest in network science. The modularity introduced by Newman and Girvan [Phys. Rev. E 69, 026113 (2004)] is the most popular quality function for community detection in networks. In this study, we identify a problem in the concept of modularity and suggest a solution to overcome this problem. Specifically, we obtain a new quality function for community detection. We refer to the function as Z-modularity because it measures the Z-score of a given division with respect to the fraction of the number of edges within communities. Our theoretical analysis shows that Z-modularity mitigates the resolution limit of the original modularity in certain cases. Computational experiments using both artificial networks and well-known real-world networks demonstrate the validity and reliability of the proposed quality function."
                    },
                    {
                        "title": "Robust Densest Subgraph Discovery",
                        "abstract": "Dense subgraph discovery is an important primitive in graph mining, which has a wide variety of applications in diverse domains. In the densest subgraph problem, given an undirected graph $G=(V,E)$ with an edge-weight vector $w=(w_e)_{e\\in E}$, we aim to find $S\\subseteq V$ that maximizes the density, i.e., $w(S)/|S|$, where $w(S)$ is the sum of the weights of the edges in the subgraph induced by $S$. Although the densest subgraph problem is one of the most well-studied optimization problems for dense subgraph discovery, there is an implicit strong assumption; it is assumed that the weights of all the edges are known exactly as input. In real-world applications, there are often cases where we have only uncertain information of the edge weights. In this study, we provide a framework for dense subgraph discovery under the uncertainty of edge weights. Specifically, we address such an uncertainty issue using the theory of robust optimization. First, we formulate our fundamental problem, the robust densest subgraph problem, and present a simple algorithm. We then formulate the robust densest subgraph problem with sampling oracle that models dense subgraph discovery using an edge-weight sampling oracle, and present an algorithm with a strong theoretical performance guarantee. Computational experiments using both synthetic graphs and popular real-world graphs demonstrate the effectiveness of our proposed algorithms."
                    },
                    {
                        "title": "The Densest Subgraph Problem with a Convex/Concave Size Function",
                        "abstract": "In the densest subgraph problem, given an edge-weighted undirected graph $G=(V,E,w)$, we are asked to find $S\\subseteq V$ that maximizes the density, i.e., $w(S)/|S|$, where $w(S)$ is the sum of weights of the edges in the subgraph induced by $S$. This problem has often been employed in a wide variety of graph mining applications. However, the problem has a drawback; it may happen that the obtained subset is too large or too small in comparison with the size desired in the application at hand. In this study, we address the size issue of the densest subgraph problem by generalizing the density of $S\\subseteq V$. Specifically, we introduce the $f$-density of $S\\subseteq V$, which is defined as $w(S)/f(|S|)$, where $f:\\mathbb{Z}_{\\geq 0}\\rightarrow \\mathbb{R}_{\\geq 0}$ is a monotonically non-decreasing function. In the $f$-densest subgraph problem ($f$-DS), we aim to find $S\\subseteq V$ that maximizes the $f$-density $w(S)/f(|S|)$. Although $f$-DS does not explicitly specify the size of the output subset of vertices, we can handle the above size issue using a convex/concave size function $f$ appropriately. For $f$-DS with convex function $f$, we propose a nearly-linear-time algorithm with a provable approximation guarantee. On the other hand, for $f$-DS with concave function $f$, we propose an LP-based exact algorithm, a flow-based $O(|V|^3)$-time exact algorithm for unweighted graphs, and a nearly-linear-time approximation algorithm."
                    },
                    {
                        "title": "Hypergraph Clustering Based on PageRank",
                        "abstract": "A hypergraph is a useful combinatorial object to model ternary or higher-order relations among entities. Clustering hypergraphs is a fundamental task in network analysis. In this study, we develop two clustering algorithms based on personalized PageRank on hypergraphs. The first one is local in the sense that its goal is to find a tightly connected vertex set with a bounded volume including a specified vertex. The second one is global in the sense that its goal is to find a tightly connected vertex set. For both algorithms, we discuss theoretical guarantees on the conductance of the output vertex set. Also, we experimentally demonstrate that our clustering algorithms outperform existing methods in terms of both the solution quality and running time. To the best of our knowledge, ours are the first practical algorithms for hypergraphs with theoretical guarantees on the conductance of the output set."
                    },
                    {
                        "title": "Stochastic Solutions for Dense Subgraph Discovery in Multilayer Networks",
                        "abstract": "Network analysis has played a key role in knowledge discovery and data mining. In many real-world applications in recent years, we are interested in mining multilayer networks, where we have a number of edge sets called layers, which encode different types of connections and/or time-dependent connections over the same set of vertices. Among many network analysis techniques, dense subgraph discovery, aiming to find a dense component in a network, is an essential primitive with a variety of applications in diverse domains. In this paper, we introduce a novel optimization model for dense subgraph discovery in multilayer networks. Our model aims to find a stochastic solution, i.e., a probability distribution over the family of vertex subsets, rather than a single vertex subset, whereas it can also be used for obtaining a single vertex subset. For our model, we design an LP-based polynomial-time exact algorithm. Moreover, to handle large-scale networks, we also devise a simple, scalable preprocessing algorithm, which often reduces the size of the input networks significantly and results in a substantial speed-up. Computational experiments demonstrate the validity of our model and the effectiveness of our algorithms."
                    },
                    {
                        "title": "A Survey on the Densest Subgraph Problem and Its Variants",
                        "abstract": "The Densest Subgraph Problem requires to find, in a given graph, a subset of vertices whose induced subgraph maximizes a measure of density. The problem has received a great deal of attention in the algorithmic literature since the early 1970s, with many variants proposed and many applications built on top of this basic definition. Recent years have witnessed a revival of research interest in this problem with several important contributions, including some groundbreaking results, published in 2022 and 2023. This survey provides a deep overview of the fundamental results and an exhaustive coverage of the many variants proposed in the literature, with a special attention to the most recent results. The survey also presents a comprehensive overview of applications and discusses some interesting open problems for this evergreen research topic."
                    },
                    {
                        "title": "Additive Approximation Algorithms for Modularity Maximization",
                        "abstract": "The modularity is a quality function in community detection, which was introduced by Newman and Girvan (2004). Community detection in graphs is now often conducted through modularity maximization: given an undirected graph $G=(V,E)$, we are asked to find a partition $\\mathcal{C}$ of $V$ that maximizes the modularity. Although numerous algorithms have been developed to date, most of them have no theoretical approximation guarantee. Recently, to overcome this issue, the design of modularity maximization algorithms with provable approximation guarantees has attracted significant attention in the computer science community.   In this study, we further investigate the approximability of modularity maximization. More specifically, we propose a polynomial-time $\\left(\\cos\\left(\\frac{3-\\sqrt{5}}{4}\\pi\\right) - \\frac{1+\\sqrt{5}}{8}\\right)$-additive approximation algorithm for the modularity maximization problem. Note here that $\\cos\\left(\\frac{3-\\sqrt{5}}{4}\\pi\\right) - \\frac{1+\\sqrt{5}}{8} < 0.42084$ holds. This improves the current best additive approximation error of $0.4672$, which was recently provided by Dinh, Li, and Thai (2015). Interestingly, our analysis also demonstrates that the proposed algorithm obtains a nearly-optimal solution for any instance with a very high modularity value. Moreover, we propose a polynomial-time $0.16598$-additive approximation algorithm for the maximum modularity cut problem. It should be noted that this is the first non-trivial approximability result for the problem. Finally, we demonstrate that our approximation algorithm can be extended to some related problems."
                    },
                    {
                        "title": "A Projected Gradient Method for Opinion Optimization with Limited Changes of Susceptibility to Persuasion",
                        "abstract": "Many social phenomena are triggered by public opinion that is formed in the process of opinion exchange among individuals. To date, from the engineering point of view, a large body of work has been devoted to studying how to manipulate individual opinions so as to guide public opinion towards the desired state. Recently, Abebe et al. (KDD 2018) have initiated the study of the impact of interventions at the level of susceptibility rather than the interventions that directly modify individual opinions themselves. For the model, Chan et al. (The Web Conference 2019) designed a local search algorithm to find an optimal solution in polynomial time. However, it can be seen that the solution obtained by solving the above model might not be implemented in real-world scenarios. In fact, as we do not consider the amount of changes of the susceptibility, it would be too costly to change the susceptibility values for agents based on the solution.   In this paper, we study an opinion optimization model that is able to limit the amount of changes of the susceptibility in various forms. First we introduce a novel opinion optimization model, where the initial susceptibility values are given as additional input and the feasible region is defined using the $\\ell_p$-ball centered at the initial susceptibility vector. For the proposed model, we design a projected gradient method that is applicable to the case where there are millions of agents. Finally we conduct thorough experiments using a variety of real-world social networks and demonstrate that the proposed algorithm outperforms baseline methods."
                    },
                    {
                        "title": "A Study on Modularity Density Maximization: Column Generation Acceleration and Computational Complexity Analysis",
                        "abstract": "Community detection is a fundamental network-analysis primitive with a variety of applications in diverse domains. Although the modularity introduced by Newman and Girvan (2004) has widely been used as a quality function for community detection, it has some drawbacks. The modularity density introduced by Li et al. (2008) is known to be an effective alternative to the modularity, which mitigates one of the drawbacks called the resolution limit. A large body of work has been devoted to designing exact and heuristic methods for modularity density maximization, without any computational complexity analysis. In this study, we investigate modularity density maximization from both algorithmic and computational complexity aspects. Specifically, we first accelerate column generation for the modularity density maximization problem. To this end, we point out that the auxiliary problem appearing in column generation can be viewed as a dense subgraph discovery problem. Then we employ a well-known strategy for dense subgraph discovery, called the greedy peeling, for approximately solving the auxiliary problem. Moreover, we reformulate the auxiliary problem to a sequence of $0$--$1$ linear programming problems, enabling us to compute its optimal value more efficiently and to get more diverse columns. Computational experiments using a variety of real-world networks demonstrate the effectiveness of our proposed algorithm. Finally, we show the NP-hardness of a slight variant of the modularity density maximization problem, where the output partition has to have two or more clusters, as well as showing the NP-hardness of the auxiliary problem."
                    },
                    {
                        "title": "Local Centrality Minimization with Quality Guarantees",
                        "abstract": "Centrality measures, quantifying the importance of vertices or edges, play a fundamental role in network analysis. To date, triggered by some positive approximability results, a large body of work has been devoted to studying centrality maximization, where the goal is to maximize the centrality score of a target vertex by manipulating the structure of a given network. On the other hand, due to the lack of such results, only very little attention has been paid to centrality minimization, despite its practical usefulness.   In this study, we introduce a novel optimization model for local centrality minimization, where the manipulation is allowed only around the target vertex. We prove the NP-hardness of our model and that the most intuitive greedy algorithm has a quite limited performance in terms of approximation ratio. Then we design two effective approximation algorithms: The first algorithm is a highly-scalable algorithm that has an approximation ratio unachievable by the greedy algorithm, while the second algorithm is a bicriteria approximation algorithm that solves a continuous relaxation based on the Lov\\'asz extension, using a projected subgradient method. To the best of our knowledge, ours are the first polynomial-time algorithms with provable approximation guarantees for centrality minimization. Experiments using a variety of real-world networks demonstrate the effectiveness of our proposed algorithms: Our first algorithm is applicable to million-scale graphs and obtains much better solutions than those of scalable baselines, while our second algorithm is rather strong against adversarial instances."
                    },
                    {
                        "title": "Multilayer Correlation Clustering",
                        "abstract": "In this paper, we establish Multilayer Correlation Clustering, a novel generalization of Correlation Clustering (Bansal et al., FOCS '02) to the multilayer setting. In this model, we are given a series of inputs of Correlation Clustering (called layers) over the common set $V$. The goal is then to find a clustering of $V$ that minimizes the $\\ell_p$-norm ($p\\geq 1$) of the disagreements vector, which is defined as the vector (with dimension equal to the number of layers), each element of which represents the disagreements of the clustering on the corresponding layer. For this generalization, we first design an $O(L\\log n)$-approximation algorithm, where $L$ is the number of layers, based on the well-known region growing technique. We then study an important special case of our problem, namely the problem with the probability constraint. For this case, we first give an $(\\alpha+2)$-approximation algorithm, where $\\alpha$ is any possible approximation ratio for the single-layer counterpart. For instance, we can take $\\alpha=2.5$ in general (Ailon et al., JACM '08) and $\\alpha=1.73+\\epsilon$ for the unweighted case (Cohen-Addad et al., FOCS '23). Furthermore, we design a $4$-approximation algorithm, which improves the above approximation ratio of $\\alpha+2=4.5$ for the general probability-constraint case. Computational experiments using real-world datasets demonstrate the effectiveness of our proposed algorithms."
                    },
                    {
                        "title": "Finding Cheeger Cuts in Hypergraphs via Heat Equation",
                        "abstract": "Cheeger's inequality states that a tightly connected subset can be extracted from a graph $G$ using an eigenvector of the normalized Laplacian associated with $G$. More specifically, we can compute a subset with conductance $O(\\sqrt{\\phi_G})$, where $\\phi_G$ is the minimum conductance of a set in $G$. It has recently been shown that Cheeger's inequality can be extended to hypergraphs. However, as the normalized Laplacian of a hypergraph is no longer a matrix, we can only approximate to its eigenvectors; this causes a loss in the conductance of the obtained subset. To address this problem, we here consider the heat equation on hypergraphs, which is a differential equation exploiting the normalized Laplacian. We show that the heat equation has a unique solution and that we can extract a subset with conductance $\\sqrt{\\phi_G}$ from the solution. An analogous result also holds for directed graphs."
                    },
                    {
                        "title": "Online Dense Subgraph Discovery via Blurred-Graph Feedback",
                        "abstract": "Dense subgraph discovery aims to find a dense component in edge-weighted graphs. This is a fundamental graph-mining task with a variety of applications and thus has received much attention recently. Although most existing methods assume that each individual edge weight is easily obtained, such an assumption is not necessarily valid in practice. In this paper, we introduce a novel learning problem for dense subgraph discovery in which a learner queries edge subsets rather than only single edges and observes a noisy sum of edge weights in a queried subset. For this problem, we first propose a polynomial-time algorithm that obtains a nearly-optimal solution with high probability. Moreover, to deal with large-sized graphs, we design a more scalable algorithm with a theoretical guarantee. Computational experiments using real-world graphs demonstrate the effectiveness of our algorithms."
                    },
                    {
                        "title": "Graph Mining Meets Crowdsourcing: Extracting Experts for Answer Aggregation",
                        "abstract": "Aggregating responses from crowd workers is a fundamental task in the process of crowdsourcing. In cases where a few experts are overwhelmed by a large number of non-experts, most answer aggregation algorithms such as the majority voting fail to identify the correct answers. Therefore, it is crucial to extract reliable experts from the crowd workers. In this study, we introduce the notion of \"expert core\", which is a set of workers that is very unlikely to contain a non-expert. We design a graph-mining-based efficient algorithm that exactly computes the expert core. To answer the aggregation task, we propose two types of algorithms. The first one incorporates the expert core into existing answer aggregation algorithms such as the majority voting, whereas the second one utilizes information provided by the expert core extraction algorithm pertaining to the reliability of workers. We then give a theoretical justification for the first type of algorithm. Computational experiments using synthetic and real-world datasets demonstrate that our proposed answer aggregation algorithms outperform state-of-the-art algorithms."
                    },
                    {
                        "title": "Query-Efficient Correlation Clustering with Noisy Oracle",
                        "abstract": "We study a general clustering setting in which we have $n$ elements to be clustered, and we aim to perform as few queries as possible to an oracle that returns a noisy sample of the similarity between two elements. Our setting encompasses many application domains in which the similarity function is costly to compute and inherently noisy. We propose two novel formulations of online learning problems rooted in the paradigm of Pure Exploration in Combinatorial Multi-Armed Bandits (PE-CMAB): fixed confidence and fixed budget settings. For both settings, we design algorithms that combine a sampling strategy with a classic approximation algorithm for correlation clustering and study their theoretical guarantees. Our results are the first examples of polynomial-time algorithms that work for the case of PE-CMAB in which the underlying offline optimization problem is NP-hard."
                    },
                    {
                        "title": "Finding Densest $k$-Connected Subgraphs",
                        "abstract": "Dense subgraph discovery is an important graph-mining primitive with a variety of real-world applications. One of the most well-studied optimization problems for dense subgraph discovery is the densest subgraph problem, where given an edge-weighted undirected graph $G=(V,E,w)$, we are asked to find $S\\subseteq V$ that maximizes the density $d(S)$, i.e., half the weighted average degree of the induced subgraph $G[S]$. This problem can be solved exactly in polynomial time and well-approximately in almost linear time. However, a densest subgraph has a structural drawback, namely, the subgraph may not be robust to vertex/edge failure. Indeed, a densest subgraph may not be well-connected, which implies that the subgraph may be disconnected by removing only a few vertices/edges within it. In this paper, we provide an algorithmic framework to find a dense subgraph that is well-connected in terms of vertex/edge connectivity. Specifically, we introduce the following problems: given a graph $G=(V,E,w)$ and a positive integer/real $k$, we are asked to find $S\\subseteq V$ that maximizes the density $d(S)$ under the constraint that $G[S]$ is $k$-vertex/edge-connected. For both problems, we propose polynomial-time (bicriteria and ordinary) approximation algorithms, using classic Mader's theorem in graph theory and its extensions."
                    },
                    {
                        "title": "Densest Diverse Subgraphs: How to Plan a Successful Cocktail Party with Diversity",
                        "abstract": "Dense subgraph discovery methods are routinely used in a variety of applications including the identification of a team of skilled individuals for collaboration from a social network. However, when the network's node set is associated with a sensitive attribute such as race, gender, religion, or political opinion, the lack of diversity can lead to lawsuits.   In this work, we focus on the problem of finding a densest diverse subgraph in a graph whose nodes have different attribute values/types that we refer to as colors. We propose two novel formulations motivated by different realistic scenarios. Our first formulation, called the densest diverse subgraph problem (DDSP), guarantees that no color represents more than some fraction of the nodes in the output subgraph, which generalizes the state-of-the-art due to Anagnostopoulos et al. (CIKM 2020). By varying the fraction we can range the diversity constraint and interpolate from a diverse dense subgraph where all colors have to be equally represented to an unconstrained dense subgraph. We design a scalable $\\Omega(1/\\sqrt{n})$-approximation algorithm, where $n$ is the number of nodes. Our second formulation is motivated by the setting where any specified color should not be overlooked. We propose the densest at-least-$\\vec{k}$-subgraph problem (Dal$\\vec{k}$S), a novel generalization of the classic Dal$k$S, where instead of a single value $k$, we have a vector ${\\mathbf k}$ of cardinality demands with one coordinate per color class. We design a $1/3$-approximation algorithm using linear programming together with an acceleration technique. Computational experiments using synthetic and real-world datasets demonstrate that our proposed algorithms are effective in extracting dense diverse clusters."
                    }
                ]
            },
            "3bc97c2f-86a1-479b-964f-bd640a7de59e": {
                "pk": "3bc97c2f-86a1-479b-964f-bd640a7de59e",
                "name": "Fabio Vitale",
                "collaborators": [
                    "Claudio Gentile",
                    "Giovanni Zappella",
                    "Nicolo Cesa-Bianchi",
                    "Stephen Pasteris",
                    "Nicol\u00f2 Cesa-Bianchi",
                    "Anand Rajagopalan",
                    "Nikos Parotsidis",
                    "Nicolo' Cesa-Bianchi",
                    "G\u00e9raud Le Falher",
                    "Mark Herbster"
                ],
                "domain": [
                    "Active Learning",
                    "Graph Theory",
                    "Recommendation Systems",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Flattening a Hierarchical Clustering through Active Learning",
                        "abstract": "We investigate active learning by pairwise similarity over the leaves of trees originating from hierarchical clustering procedures. In the realizable setting, we provide a full characterization of the number of queries needed to achieve perfect reconstruction of the tree cut. In the non-realizable setting, we rely on known important-sampling procedures to obtain regret and query complexity bounds. Our algorithms come with theoretical guarantees on the statistical error and, more importantly, lend themselves to linear-time implementations in the relevant parameters of the problem. We discuss such implementations, prove running time guarantees for them, and present preliminary experiments on real-world datasets showing the compelling practical performance of our algorithms as compared to both passive learning and simple active learning baselines."
                    },
                    {
                        "title": "Online Reciprocal Recommendation with Theoretical Performance Guarantees",
                        "abstract": "A reciprocal recommendation problem is one where the goal of learning is not just to predict a user's preference towards a passive item (e.g., a book), but to recommend the targeted user on one side another user from the other side such that a mutual interest between the two exists. The problem thus is sharply different from the more traditional items-to-users recommendation, since a good match requires meeting the preferences of both users. We initiate a rigorous theoretical investigation of the reciprocal recommendation task in a specific framework of sequential learning. We point out general limitations, formulate reasonable assumptions enabling effective learning and, under these assumptions, we design and analyze a computationally efficient algorithm that uncovers mutual likes at a pace comparable to those achieved by a clearvoyant algorithm knowing all user preferences in advance. Finally, we validate our algorithm against synthetic and real-world datasets, showing improved empirical performance over simple baselines."
                    },
                    {
                        "title": "A Correlation Clustering Approach to Link Classification in Signed Networks -- Full Version --",
                        "abstract": "Motivated by social balance theory, we develop a theory of link classification in signed networks using the correlation clustering index as measure of label regularity. We derive learning bounds in terms of correlation clustering within three fundamental transductive learning settings: online, batch and active. Our main algorithmic contribution is in the active setting, where we introduce a new family of efficient link classifiers based on covering the input graph with small circuits. These are the first active algorithms for link classification with mistake bounds that hold for arbitrary signed networks."
                    },
                    {
                        "title": "Random Spanning Trees and the Prediction of Weighted Graphs",
                        "abstract": "We investigate the problem of sequentially predicting the binary labels on the nodes of an arbitrary weighted graph. We show that, under a suitable parametrization of the problem, the optimal number of prediction mistakes can be characterized (up to logarithmic factors) by the cutsize of a random spanning tree of the graph. The cutsize is induced by the unknown adversarial labeling of the graph nodes. In deriving our characterization, we obtain a simple randomized algorithm achieving in expectation the optimal mistake bound on any polynomially connected weighted graph. Our algorithm draws a random spanning tree of the original graph and then predicts the nodes of this tree in constant expected amortized time and linear space. Experiments on real-world datasets show that our method compares well to both global (Perceptron) and local (label propagation) methods, while being generally faster in practice."
                    },
                    {
                        "title": "A Linear Time Active Learning Algorithm for Link Classification -- Full Version --",
                        "abstract": "We present very efficient active learning algorithms for link classification in signed networks. Our algorithms are motivated by a stochastic model in which edge labels are obtained through perturbations of a initial sign assignment consistent with a two-clustering of the nodes. We provide a theoretical analysis within this model, showing that we can achieve an optimal (to whithin a constant factor) number of mistakes on any graph G = (V,E) such that |E| = \\Omega(|V|^{3/2}) by querying O(|V|^{3/2}) edge labels. More generally, we show an algorithm that achieves optimality to within a factor of O(k) by querying at most order of |V| + (|V|/k)^{3/2} edge labels. The running time of this algorithm is at most of order |E| + |V|\\log|V|."
                    },
                    {
                        "title": "Active Learning on Trees and Graphs",
                        "abstract": "We investigate the problem of active learning on a given tree whose nodes are assigned binary labels in an adversarial way. Inspired by recent results by Guillory and Bilmes, we characterize (up to constant factors) the optimal placement of queries so to minimize the mistakes made on the non-queried nodes. Our query selection algorithm is extremely efficient, and the optimal number of mistakes on the non-queried nodes is achieved by a simple and efficient mincut classifier. Through a simple modification of the query selection algorithm we also show optimality (up to constant factors) with respect to the trade-off between number of queries and number of mistakes on non-queried nodes. By using spanning trees, our algorithms can be efficiently applied to general graphs, although the problem of finding optimal and efficient active learning algorithms for general graphs remains open. Towards this end, we provide a lower bound on the number of mistakes made on arbitrary graphs by any active learning algorithm using a number of queries which is up to a constant fraction of the graph size."
                    },
                    {
                        "title": "See the Tree Through the Lines: The Shazoo Algorithm -- Full Version --",
                        "abstract": "Predicting the nodes of a given graph is a fascinating theoretical problem with applications in several domains. Since graph sparsification via spanning trees retains enough information while making the task much easier, trees are an important special case of this problem. Although it is known how to predict the nodes of an unweighted tree in a nearly optimal way, in the weighted case a fully satisfactory algorithm is not available yet. We fill this hole and introduce an efficient node predictor, Shazoo, which is nearly optimal on any weighted tree. Moreover, we show that Shazoo can be viewed as a common nontrivial generalization of both previous approaches for unweighted trees and weighted lines. Experiments on real-world datasets confirm that Shazoo performs well in that it fully exploits the structure of the input tree, and gets very close to (and sometimes better than) less scalable energy minimization methods."
                    },
                    {
                        "title": "Even Trolls Are Useful: Efficient Link Classification in Signed Networks",
                        "abstract": "We address the problem of classifying the links of signed social networks given their full structural topology. Motivated by a binary user behaviour assumption, which is supported by decades of research in psychology, we develop an efficient and surprisingly simple approach to solve this classification problem. Our methods operate both within the active and batch settings. We demonstrate that the algorithms we developed are extremely fast in both theoretical and practical terms. Within the active setting, we provide a new complexity measure and a rigorous analysis of our methods that hold for arbitrary signed networks. We validate our theoretical claims carrying out a set of experiments on three well known real-world datasets, showing that our methods outperform the competitors while being much faster."
                    },
                    {
                        "title": "On Pairwise Clustering with Side Information",
                        "abstract": "Pairwise clustering, in general, partitions a set of items via a known similarity function. In our treatment, clustering is modeled as a transductive prediction problem. Thus rather than beginning with a known similarity function, the function instead is hidden and the learner only receives a random sample consisting of a subset of the pairwise similarities. An additional set of pairwise side-information may be given to the learner, which then determines the inductive bias of our algorithms. We measure performance not based on the recovery of the hidden similarity function, but instead on how well we classify each item. We give tight bounds on the number of misclassifications. We provide two algorithms. The first algorithm SACA is a simple agglomerative clustering algorithm which runs in near linear time, and which serves as a baseline for our analyses. Whereas the second algorithm, RGCA, enables the incorporation of side-information which may lead to improved bounds at the cost of a longer running time."
                    },
                    {
                        "title": "Correlation Clustering with Adaptive Similarity Queries",
                        "abstract": "In correlation clustering, we are given $n$ objects together with a binary similarity score between each pair of them. The goal is to partition the objects into clusters so to minimise the disagreements with the scores. In this work we investigate correlation clustering as an active learning problem: each similarity score can be learned by making a query, and the goal is to minimise both the disagreements and the total number of queries. On the one hand, we describe simple active learning algorithms, which provably achieve an almost optimal trade-off while giving cluster recovery guarantees, and we test them on different datasets. On the other hand, we prove information-theoretical bounds on the number of queries necessary to guarantee a prescribed disagreement bound. These results give a rich characterization of the trade-off between queries and clustering error."
                    },
                    {
                        "title": "Sum-max Submodular Bandits",
                        "abstract": "Many online decision-making problems correspond to maximizing a sequence of submodular functions. In this work, we introduce sum-max functions, a subclass of monotone submodular functions capturing several interesting problems, including best-of-$K$-bandits, combinatorial bandits, and the bandit versions on facility location, $M$-medians, and hitting sets. We show that all functions in this class satisfy a key property that we call pseudo-concavity. This allows us to prove $\\big(1 - \\frac{1}{e}\\big)$-regret bounds for bandit feedback in the nonstochastic setting of the order of $\\sqrt{MKT}$ (ignoring log factors), where $T$ is the time horizon and $M$ is a cardinality constraint. This bound, attained by a simple and efficient algorithm, significantly improves on the $\\widetilde{O}\\big(T^{2/3}\\big)$ regret bound for online monotone submodular maximization with bandit feedback."
                    }
                ]
            },
            "519019e4-47d8-4b60-af91-71d169dd3d90": {
                "pk": "519019e4-47d8-4b60-af91-71d169dd3d90",
                "name": "Mark Herbster",
                "collaborators": [
                    "Stephen Pasteris",
                    "Fabio Vitale",
                    "Claudio Gentile",
                    "Shiqiang Wang",
                    "James Robinson",
                    "Lisa Tse",
                    "Ting He",
                    "Simone Severini",
                    "Leonard Wossnig",
                    "Paul Rubenstein"
                ],
                "domain": [
                    "Online Learning",
                    "Graph Theory",
                    "Machine Learning",
                    "Quantum Computing"
                ],
                "publications": [
                    {
                        "title": "The VC-Dimension of Similarity Hypotheses Spaces",
                        "abstract": "Given a set $X$ and a function $h:X\\longrightarrow\\{0,1\\}$ which labels each element of $X$ with either $0$ or $1$, we may define a function $h^{(s)}$ to measure the similarity of pairs of points in $X$ according to $h$. Specifically, for $h\\in \\{0,1\\}^X$ we define $h^{(s)}\\in \\{0,1\\}^{X\\times X}$ by $h^{(s)}(w,x):= \\mathbb{1}[h(w) = h(x)]$. This idea can be extended to a set of functions, or hypothesis space $\\mathcal{H} \\subseteq \\{0,1\\}^X$ by defining a similarity hypothesis space $\\mathcal{H}^{(s)}:=\\{h^{(s)}:h\\in\\mathcal{H}\\}$. We show that ${{vc-dimension}}(\\mathcal{H}^{(s)}) \\in \\Theta({{vc-dimension}}(\\mathcal{H}))$."
                    },
                    {
                        "title": "Improved Regret Bounds for Tracking Experts with Memory",
                        "abstract": "We address the problem of sequential prediction with expert advice in a non-stationary environment with long-term memory guarantees in the sense of Bousquet and Warmuth [4]. We give a linear-time algorithm that improves on the best known regret bounds [26]. This algorithm incorporates a relative entropy projection step. This projection is advantageous over previous weight-sharing approaches in that weight updates may come with implicit costs as in for example portfolio optimization. We give an algorithm to compute this projection step in linear time, which may be of independent interest."
                    },
                    {
                        "title": "Online Similarity Prediction of Networked Data from Known and Unknown Graphs",
                        "abstract": "We consider online similarity prediction problems over networked data. We begin by relating this task to the more standard class prediction problem, showing that, given an arbitrary algorithm for class prediction, we can construct an algorithm for similarity prediction with \"nearly\" the same mistake bound, and vice versa. After noticing that this general construction is computationally infeasible, we target our study to {\\em feasible} similarity prediction algorithms on networked data. We initially assume that the network structure is {\\em known} to the learner. Here we observe that Matrix Winnow \\cite{w07} has a near-optimal mistake guarantee, at the price of cubic prediction time per round. This motivates our effort for an efficient implementation of a Perceptron algorithm with a weaker mistake guarantee but with only poly-logarithmic prediction time. Our focus then turns to the challenging case of networks whose structure is initially {\\em unknown} to the learner. In this novel setting, where the network structure is only incrementally revealed, we obtain a mistake-bounded algorithm with a quadratic prediction time per round."
                    },
                    {
                        "title": "On Pairwise Clustering with Side Information",
                        "abstract": "Pairwise clustering, in general, partitions a set of items via a known similarity function. In our treatment, clustering is modeled as a transductive prediction problem. Thus rather than beginning with a known similarity function, the function instead is hidden and the learner only receives a random sample consisting of a subset of the pairwise similarities. An additional set of pairwise side-information may be given to the learner, which then determines the inductive bias of our algorithms. We measure performance not based on the recovery of the hidden similarity function, but instead on how well we classify each item. We give tight bounds on the number of misclassifications. We provide two algorithms. The first algorithm SACA is a simple agglomerative clustering algorithm which runs in near linear time, and which serves as a baseline for our analyses. Whereas the second algorithm, RGCA, enables the incorporation of side-information which may lead to improved bounds at the cost of a longer running time."
                    },
                    {
                        "title": "Online Multitask Learning with Long-Term Memory",
                        "abstract": "We introduce a novel online multitask setting. In this setting each task is partitioned into a sequence of segments that is unknown to the learner. Associated with each segment is a hypothesis from some hypothesis class. We give algorithms that are designed to exploit the scenario where there are many such segments but significantly fewer associated hypotheses. We prove regret bounds that hold for any segmentation of the tasks and any association of hypotheses to the segments. In the single-task setting this is equivalent to switching with long-term memory in the sense of [Bousquet and Warmuth; 2003]. We provide an algorithm that predicts on each trial in time linear in the number of hypotheses when the hypothesis class is finite. We also consider infinite hypothesis classes from reproducing kernel Hilbert spaces for which we give an algorithm whose per trial time complexity is cubic in the number of cumulative trials. In the single-task special case this is the first example of an efficient regret-bounded switching algorithm with long-term memory for a non-parametric hypothesis class."
                    },
                    {
                        "title": "Online Prediction of Switching Graph Labelings with Cluster Specialists",
                        "abstract": "We address the problem of predicting the labeling of a graph in an online setting when the labeling is changing over time. We present an algorithm based on a specialist approach; we develop the machinery of cluster specialists which probabilistically exploits the cluster structure in the graph. Our algorithm has two variants, one of which surprisingly only requires $\\mathcal{O}(\\log n)$ time on any trial $t$ on an $n$-vertex graph, an exponential speed up over existing methods. We prove switching mistake-bound guarantees for both variants of our algorithm. Furthermore these mistake bounds smoothly vary with the magnitude of the change between successive labelings. We perform experiments on Chicago Divvy Bicycle Sharing data and show that our algorithms significantly outperform an existing algorithm (a kernelized Perceptron) as well as several natural benchmarks."
                    },
                    {
                        "title": "Multi-class Graph Clustering via Approximated Effective $p$-Resistance",
                        "abstract": "This paper develops an approximation to the (effective) $p$-resistance and applies it to multi-class clustering. Spectral methods based on the graph Laplacian and its generalization to the graph $p$-Laplacian have been a backbone of non-euclidean clustering techniques. The advantage of the $p$-Laplacian is that the parameter $p$ induces a controllable bias on cluster structure. The drawback of $p$-Laplacian eigenvector based methods is that the third and higher eigenvectors are difficult to compute. Thus, instead, we are motivated to use the $p$-resistance induced by the $p$-Laplacian for clustering. For $p$-resistance, small $p$ biases towards clusters with high internal connectivity while large $p$ biases towards clusters of small \"extent,\" that is a preference for smaller shortest-path distances between vertices in the cluster. However, the $p$-resistance is expensive to compute. We overcome this by developing an approximation to the $p$-resistance. We prove upper and lower bounds on this approximation and observe that it is exact when the graph is a tree. We also provide theoretical justification for the use of $p$-resistance for clustering. Finally, we provide experiments comparing our approximated $p$-resistance clustering to other $p$-Laplacian based methods."
                    },
                    {
                        "title": "Online Convex Optimisation: The Optimal Switching Regret for all Segmentations Simultaneously",
                        "abstract": "We consider the classic problem of online convex optimisation. Whereas the notion of static regret is relevant for stationary problems, the notion of switching regret is more appropriate for non-stationary problems. A switching regret is defined relative to any segmentation of the trial sequence, and is equal to the sum of the static regrets of each segment. In this paper we show that, perhaps surprisingly, we can achieve the asymptotically optimal switching regret on every possible segmentation simultaneously. Our algorithm for doing so is very efficient: having a space and per-trial time complexity that is logarithmic in the time-horizon. Our algorithm also obtains novel bounds on its dynamic regret: being adaptive to variations in the rate of change of the comparator sequence."
                    },
                    {
                        "title": "MaxHedge: Maximising a Maximum Online",
                        "abstract": "We introduce a new online learning framework where, at each trial, the learner is required to select a subset of actions from a given known action set. Each action is associated with an energy value, a reward and a cost. The sum of the energies of the actions selected cannot exceed a given energy budget. The goal is to maximise the cumulative profit, where the profit obtained on a single trial is defined as the difference between the maximum reward among the selected actions and the sum of their costs. Action energy values and the budget are known and fixed. All rewards and costs associated with each action change over time and are revealed at each trial only after the learner's selection of actions. Our framework encompasses several online learning problems where the environment changes over time; and the solution trades-off between minimising the costs and maximising the maximum reward of the selected subset of actions, while being constrained to an action energy budget. The algorithm that we propose is efficient and general in that it may be specialised to multiple natural online combinatorial problems."
                    },
                    {
                        "title": "Online Learning of Facility Locations",
                        "abstract": "In this paper, we provide a rigorous theoretical investigation of an online learning version of the Facility Location problem which is motivated by emerging problems in real-world applications. In our formulation, we are given a set of sites and an online sequence of user requests. At each trial, the learner selects a subset of sites and then incurs a cost for each selected site and an additional cost which is the price of the user's connection to the nearest site in the selected subset. The problem may be solved by an application of the well-known Hedge algorithm. This would, however, require time and space exponential in the number of the given sites, which motivates our design of a novel quasi-linear time algorithm for this problem, with good theoretical guarantees on its performance."
                    },
                    {
                        "title": "Service Placement with Provable Guarantees in Heterogeneous Edge Computing Systems",
                        "abstract": "Mobile edge computing (MEC) is a promising technique for providing low-latency access to services at the network edge. The services are hosted at various types of edge nodes with both computation and communication capabilities. Due to the heterogeneity of edge node characteristics and user locations, the performance of MEC varies depending on where the service is hosted. In this paper, we consider such a heterogeneous MEC system, and focus on the problem of placing multiple services in the system to maximize the total reward. We show that the problem is NP-hard via reduction from the set cover problem, and propose a deterministic approximation algorithm to solve the problem, which has an approximation ratio that is not worse than $\\left(1-e^{-1}\\right)/4$. The proposed algorithm is based on two sub-routines that are suitable for small and arbitrarily sized services, respectively. The algorithm is designed using a novel way of partitioning each edge node into multiple slots, where each slot contains one service. The approximation guarantee is obtained via a specialization of the method of conditional expectations, which uses a randomized procedure as an intermediate step. In addition to theoretical guarantees, simulation results also show that the proposed algorithm outperforms other state-of-the-art approaches."
                    },
                    {
                        "title": "Online Matrix Completion with Side Information",
                        "abstract": "We give an online algorithm and prove novel mistake and regret bounds for online binary matrix completion with side information. The mistake bounds we prove are of the form $\\tilde{O}(D/\\gamma^2)$. The term $1/\\gamma^2$ is analogous to the usual margin term in SVM (perceptron) bounds. More specifically, if we assume that there is some factorization of the underlying $m \\times n$ matrix into $P Q^\\intercal$ where the rows of $P$ are interpreted as \"classifiers\" in $\\mathcal{R}^d$ and the rows of $Q$ as \"instances\" in $\\mathcal{R}^d$, then $\\gamma$ is the maximum (normalized) margin over all factorizations $P Q^\\intercal$ consistent with the observed matrix. The quasi-dimension term $D$ measures the quality of side information. In the presence of vacuous side information, $D= m+n$. However, if the side information is predictive of the underlying factorization of the matrix, then in an ideal case, $D \\in O(k + \\ell)$ where $k$ is the number of distinct row factors and $\\ell$ is the number of distinct column factors. We additionally provide a generalization of our algorithm to the inductive setting. In this setting, we provide an example where the side information is not directly specified in advance. For this example, the quasi-dimension $D$ is now bounded by $O(k^2 + \\ell^2)$."
                    },
                    {
                        "title": "Quantum linear systems algorithms: a primer",
                        "abstract": "The Harrow-Hassidim-Lloyd (HHL) quantum algorithm for sampling from the solution of a linear system provides an exponential speed-up over its classical counterpart. The problem of solving a system of linear equations has a wide scope of applications, and thus HHL constitutes an important algorithmic primitive. In these notes, we present the HHL algorithm and its improved versions in detail, including explanations of the constituent sub- routines. More specifically, we discuss various quantum subroutines such as quantum phase estimation and amplitude amplification, as well as the important question of loading data into a quantum computer, via quantum RAM. The improvements to the original algorithm exploit variable-time amplitude amplification as well as a method for implementing linear combinations of unitary operations (LCUs) based on a decomposition of the operators using Fourier and Chebyshev series. Finally, we discuss a linear solver based on the quantum singular value estimation (QSVE) subroutine."
                    },
                    {
                        "title": "Adversarial Online Collaborative Filtering",
                        "abstract": "We investigate the problem of online collaborative filtering under no-repetition constraints, whereby users need to be served content in an online fashion and a given user cannot be recommended the same content item more than once. We start by designing and analyzing an algorithm that works under biclustering assumptions on the user-item preference matrix, and show that this algorithm exhibits an optimal regret guarantee, while being fully adaptive, in that it is oblivious to any prior knowledge about the sequence of users, the universe of items, as well as the biclustering parameters of the preference matrix. We then propose a more robust version of this algorithm which operates with general matrices. Also this algorithm is parameter free, and we prove regret guarantees that scale with the amount by which the preference matrix deviates from a biclustered structure. To our knowledge, these are the first results on online collaborative filtering that hold at this level of generality and adaptivity under no-repetition constraints. Finally, we complement our theoretical findings with simple experiments on real-world datasets aimed at both validating the theory and empirically comparing to standard baselines. This comparison shows the competitive advantage of our approach over these baselines."
                    },
                    {
                        "title": "Quantum machine learning: a classical perspective",
                        "abstract": "Recently, increased computational power and data availability, as well as algorithmic advances, have led machine learning techniques to impressive results in regression, classification, data-generation and reinforcement learning tasks. Despite these successes, the proximity to the physical limits of chip fabrication alongside the increasing size of datasets are motivating a growing number of researchers to explore the possibility of harnessing the power of quantum computation to speed-up classical machine learning algorithms. Here we review the literature in quantum machine learning and discuss perspectives for a mixed readership of classical machine learning and quantum computation experts. Particular emphasis will be placed on clarifying the limitations of quantum algorithms, how they compare with their best classical counterparts and why quantum resources are expected to provide advantages for learning problems. Learning in the presence of noise and certain computationally hard problems in machine learning are identified as promising directions for the field. Practical questions, like how to upload classical data into quantum form, will also be addressed."
                    }
                ]
            }
        }
    },
    "2405.18400": {
        "paper_data": {
            "title": "Superposed Decoding: Multiple Generations from a Single Autoregressive Inference Pass",
            "url": "http://arxiv.org/abs/2405.18400v5",
            "arxiv_id": "2405.18400",
            "authors": [
                "Ethan Shen",
                "Alan Fan",
                "Sarah M. Pratt",
                "Jae Sung Park",
                "Matthew Wallingford",
                "Sham M. Kakade",
                "Ari Holtzman",
                "Ranjay Krishna",
                "Ali Farhadi",
                "Aditya Kusupati"
            ],
            "abstract": "Many applications today provide users with multiple auto-complete drafts as they type, including GitHub's code completion, Gmail's smart compose, and Apple's messaging auto-suggestions. Under the hood, language models support this by running an autoregressive inference pass to provide a draft. Consequently, providing $k$ drafts to the user requires running an expensive language model $k$ times. To alleviate the computation cost of running $k$ inference passes, we propose Superposed Decoding, a new decoding algorithm that generates $k$ drafts at the computation cost of one autoregressive inference pass. We achieve this by feeding a superposition of the most recent token embeddings from the $k$ drafts as input to the next decoding step of the language model. At every inference step we combine the $k$ drafts with the top-$k$ tokens to get $k^2$ new drafts and cache the $k$ most likely options, using an n-gram interpolation with minimal compute overhead to filter out incoherent generations. Our experiments show that $k$ drafts from Superposed Decoding are at least as coherent and factual as Nucleus Sampling and Greedy Decoding respectively, while being at least $2.44\\times$ faster for $k\\ge3$. In a compute-normalized setting, user evaluations demonstrably favor text generated by Superposed Decoding over Nucleus Sampling. Superposed Decoding can also be combined with other decoding strategies, resulting in universal coverage gains when scaling inference time compute. Code and more examples open-sourced at https://github.com/RAIVNLab/SuperposedDecoding.",
            "introduction": "   1 Introduction  Figure 1: To generate multiple k\ud835\udc58kitalic_k auto-complete suggestions for a prefix using an LM, the existing decoding methods like Nucleus Sampling need k\ud835\udc58kitalic_k inference passes. In contrast, Superposed Decoding can generate k\ud835\udc58kitalic_k suggestions at the cost of a single inference pass while being as coherent and factual.   Commercial surveys find that 80% of e-commerce websites provide autocomplete as a feature\u00a0[21]. With the proliferation of large language models, autocomplete drafts are now ubiquitous in an even wider range of applications. Examples include short draft suggestions in Gmail Smart Compose\u00a0[8] and code snippets in GitHub Copilot\u00a0[7]. These options provide users with the ability to consider different phrasings and increase the likelihood of having at least one suggestion that reflects their intent. While language models (LMs)\u00a0[34] now power these multiple suggestions, each additional suggestion necessitates another inference pass (batch size=1batch size1\\text{batch size}=1batch size = 1), making it computationally expensive.   Language models use autoregressive inference to generate a plausible sequence of next tokens for a given prefix\u00a0[40]. The next token generated depends on the prefix and the previously generated tokens. Decoding algorithms like Greedy Decoding, Top-k\ud835\udc58kitalic_k Sampling\u00a0[12], Beam Search, and Nucleus Sampling\u00a0[18] present various ways of obtaining generations during autoregressive inference. For a prefix, Greedy (maximum likelihood) Decoding picks the most probable token at every autoregressive timestep, eventually generating only one auto-completion suggestion. Instead of making a greedy choice at every timestep, we can also sample from the top-k\ud835\udc58kitalic_k most probable tokens to generate the sequence\u00a0[12]. Also leveraging the top-k\ud835\udc58kitalic_k most probable tokens is Beam Search, which picks the most probable k\ud835\udc58kitalic_k beams from the k2superscript\ud835\udc582k^{2}italic_k start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT possible drafts at every timestep. Alternatively, Nucleus Sampling\u00a0[18], particularly effective at generating natural-sounding text, grows generations by sampling from the collection of the smallest subset of tokens that form a preset probability mass (p\ud835\udc5dpitalic_p) at every timestep. Top-k\ud835\udc58kitalic_k Sampling, Beam Search, and Nucleus Sampling offer the benefit of generating multiple suggestions. However, this comes at the cost of requiring multiple autoregressive inference passes.   We introduce Superposed Decoding (SPD) (Figure\u00a01), a decoding algorithm that can generate mutiple (k\ud835\udc58kitalic_k) high-quality short generations using only a single autoregressive inference pass. At each autoregressive timestep during inference, Superposed Decoding feeds in the superposition (weighted combination) of the embeddings corresponding to the k\ud835\udc58kitalic_k most recent drafted tokens (Section\u00a03.1). After selecting the top-k\ud835\udc58kitalic_k output tokens, SPD expands the existing k\ud835\udc58kitalic_k drafts with them, resulting in k2superscript\ud835\udc582k^{2}italic_k start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT potential new drafts. Each draft has a probability score assigned to it, which is further smoothed using a combination of various n\ud835\udc5bnitalic_n-gram models (n\u2208[2,6]\ud835\udc5b26n\\in[2,6]italic_n \u2208 [ 2 , 6 ]) [38, 25]. N-gram interpolation for smoothing helps improve coherency by selecting the most probable and coherent continuations (top-k\ud835\udc58kitalic_k drafts) for the next autoregressive step (Section\u00a03.2). The n-gram interpolation is computationally inexpensive and allows flexibility to change domains of interest during inference (eg., programming, healthcare, finance, etc.). Superposed Decoding\u2019s effectiveness can be attributed to the apparent linearity of representations in language models\u00a0[35, 23] which we concurrently discovered (Section\u00a03.3).   Our experiments demonstrate that Superposed Decoding on Llama-2-7B\u00a0[41] can generate k\u22651\ud835\udc581k\\geq 1italic_k \u2265 1 drafts that are as coherent, in terms of",
            "references": [
                {
                    "title": "Large Language Monkeys: Scaling Inference Compute with Repeated Sampling",
                    "abstract": "Scaling the amount of compute used to train language models has dramatically improved their capabilities. However, when it comes to inference, we often limit the amount of compute to only one attempt per problem. Here, we explore inference compute as another axis for scaling by increasing the number of generated samples. Across multiple tasks and models, we observe that coverage - the fraction of problems solved by any attempt - scales with the number of samples over four orders of magnitude. In domains like coding and formal proofs, where all answers can be automatically verified, these increases in coverage directly translate into improved performance. When we apply repeated sampling to SWE-bench Lite, the fraction of issues solved with DeepSeek-V2-Coder-Instruct increases from 15.9% with one sample to 56% with 250 samples, outperforming the single-attempt state-of-the-art of 43% which uses more capable frontier models. Moreover, using current API pricing, amplifying the cheaper DeepSeek model with five samples is more cost-effective and solves more issues than paying a premium for one sample from GPT-4o or Claude 3.5 Sonnet. Interestingly, the relationship between coverage and the number of samples is often log-linear and can be modelled with an exponentiated power law, suggesting the existence of inference-time scaling laws. Finally, we find that identifying correct samples out of many generations remains an important direction for future research in domains without automatic verifiers. When solving math word problems from GSM8K and MATH, coverage with Llama-3 models grows to over 95% with 10,000 samples. However, common methods to pick correct solutions from a sample collection, such as majority voting or reward models, plateau beyond several hundred samples and fail to fully scale with the sample budget."
                },
                {
                    "title": "Keyformer: KV Cache Reduction through Key Tokens Selection for Efficient Generative Inference",
                    "abstract": "Transformers have emerged as the underpinning architecture for Large Language Models (LLMs). In generative language models, the inference process involves two primary phases: prompt processing and token generation. Token generation, which constitutes the majority of the computational workload, primarily entails vector-matrix multiplications and interactions with the Key-Value (KV) Cache. This phase is constrained by memory bandwidth due to the overhead of transferring weights and KV cache values from the memory system to the computing units. This memory bottleneck becomes particularly pronounced in applications that require long-context and extensive text generation, both of which are increasingly crucial for LLMs. This paper introduces\"Keyformer\", an innovative inference-time approach, to mitigate the challenges associated with KV cache size and memory bandwidth utilization. Keyformer leverages the observation that approximately 90% of the attention weight in generative inference focuses on a specific subset of tokens, referred to as\"key\"tokens. Keyformer retains only the key tokens in the KV cache by identifying these crucial tokens using a novel score function. This approach effectively reduces both the KV cache size and memory bandwidth usage without compromising model accuracy. We evaluate Keyformer's performance across three foundational models: GPT-J, Cerebras-GPT, and MPT, which employ various positional embedding algorithms. Our assessment encompasses a variety of tasks, with a particular emphasis on summarization and conversation tasks involving extended contexts. Keyformer's reduction of KV cache reduces inference latency by 2.1x and improves token generation throughput by 2.4x, while preserving the model's accuracy."
                },
                {
                    "title": "On the Origins of Linear Representations in Large Language Models",
                    "abstract": "Recent works have argued that high-level semantic concepts are encoded\"linearly\"in the representation space of large language models. In this work, we study the origins of such linear representations. To that end, we introduce a simple latent variable model to abstract and formalize the concept dynamics of the next token prediction. We use this formalism to show that the next token prediction objective (softmax with cross-entropy) and the implicit bias of gradient descent together promote the linear representation of concepts. Experiments show that linear representations emerge when learning from data matching the latent variable model, confirming that this simple structure already suffices to yield linear representations. We additionally confirm some predictions of the theory using the LLaMA-2 large language model, giving evidence that the simplified model yields generalizable insights."
                },
                {
                    "title": "The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits",
                    "abstract": "Recent research, such as BitNet, is paving the way for a new era of 1-bit Large Language Models (LLMs). In this work, we introduce a 1-bit LLM variant, namely BitNet b1.58, in which every single parameter (or weight) of the LLM is ternary {-1, 0, 1}. It matches the full-precision (i.e., FP16 or BF16) Transformer LLM with the same model size and training tokens in terms of both perplexity and end-task performance, while being significantly more cost-effective in terms of latency, memory, throughput, and energy consumption. More profoundly, the 1.58-bit LLM defines a new scaling law and recipe for training new generations of LLMs that are both high-performance and cost-effective. Furthermore, it enables a new computation paradigm and opens the door for designing specific hardware optimized for 1-bit LLMs."
                },
                {
                    "title": "Infini-gram: Scaling Unbounded n-gram Language Models to a Trillion Tokens",
                    "abstract": "Are $n$-gram language models still relevant in this era of neural large language models (LLMs)? Our answer is yes, and we showcase their values in both text analysis and improving neural LLMs. This was done by modernizing $n$-gram LMs in two aspects. First, we train them at the same data scale as neural LLMs -- 5 trillion tokens. This is the largest $n$-gram LM ever built. Second, existing $n$-gram LMs use small $n$ which hinders their performance; we instead allow $n$ to be arbitrarily large, by introducing a new $\\infty$-gram LM with backoff. Instead of pre-computing $n$-gram count tables (which would be very expensive), we develop an engine named infini-gram -- powered by suffix arrays -- that can compute $\\infty$-gram (as well as $n$-gram with arbitrary $n$) probabilities with millisecond-level latency. The $\\infty$-gram framework and infini-gram engine enable us to conduct many novel and interesting analyses of human-written and machine-generated text: we find that the $\\infty$-gram LM has fairly high accuracy for next-token prediction (47%), and can complement neural LLMs to greatly reduce their perplexity. When analyzing machine-generated text, we also observe irregularities in the machine--$\\infty$-gram agreement level with respect to the suffix length, which indicates deficiencies in neural LLM pretraining and the positional embeddings of Transformers."
                },
                {
                    "title": "The Linear Representation Hypothesis and the Geometry of Large Language Models",
                    "abstract": "Informally, the 'linear representation hypothesis' is the idea that high-level concepts are represented linearly as directions in some representation space. In this paper, we address two closely related questions: What does\"linear representation\"actually mean? And, how do we make sense of geometric notions (e.g., cosine similarity or projection) in the representation space? To answer these, we use the language of counterfactuals to give two formalizations of\"linear representation\", one in the output (word) representation space, and one in the input (sentence) space. We then prove these connect to linear probing and model steering, respectively. To make sense of geometric notions, we use the formalization to identify a particular (non-Euclidean) inner product that respects language structure in a sense we make precise. Using this causal inner product, we show how to unify all notions of linear representation. In particular, this allows the construction of probes and steering vectors using counterfactual pairs. Experiments with LLaMA-2 demonstrate the existence of linear representations of concepts, the connection to interpretation and control, and the fundamental role of the choice of inner product."
                },
                {
                    "title": "Efficient Long-Range Transformers: You Need to Attend More, but Not Necessarily at Every Layer",
                    "abstract": "Pretrained transformer models have demonstrated remarkable performance across various natural language processing tasks. These models leverage the attention mechanism to capture long- and short-range dependencies in the sequence. However, the (full) attention mechanism incurs high computational cost - quadratic in the sequence length, which is not affordable in tasks with long sequences, e.g., inputs with 8k tokens. Although sparse attention can be used to improve computational efficiency, as suggested in existing work, it has limited modeling capacity and often fails to capture complicated dependencies in long sequences. To tackle this challenge, we propose MASFormer, an easy-to-implement transformer variant with Mixed Attention Spans. Specifically, MASFormer is equipped with full attention to capture long-range dependencies, but only at a small number of layers. For the remaining layers, MASformer only employs sparse attention to capture short-range dependencies. Our experiments on natural language modeling and generation tasks show that a decoder-only MASFormer model of 1.3B parameters can achieve competitive performance to vanilla transformers with full attention while significantly reducing computational cost (up to 75%). Additionally, we investigate the effectiveness of continual training with long sequence data and how sequence length impacts downstream generation performance, which may be of independent interest."
                },
                {
                    "title": "MatFormer: Nested Transformer for Elastic Inference",
                    "abstract": "Transformer models are deployed in a wide range of settings, from multi-accelerator clusters to standalone mobile phones. The diverse inference constraints in these scenarios necessitate practitioners to train foundation models such as PaLM 2, Llama,&ViTs as a series of models of varying sizes. Due to significant training costs, only a select few model sizes are trained and supported, limiting more fine-grained control over relevant tradeoffs, including latency, cost, and accuracy. This work introduces MatFormer, a nested Transformer architecture designed to offer elasticity in a variety of deployment constraints. Each Feed Forward Network (FFN) block of a MatFormer model is jointly optimized with a few nested smaller FFN blocks. This training procedure allows for the Mix'n'Match of model granularities across layers -- i.e., a trained universal MatFormer model enables extraction of hundreds of accurate smaller models, which were never explicitly optimized. We empirically demonstrate MatFormer's effectiveness across different model classes (decoders&encoders), modalities (language&vision), and scales (up to 2.6B parameters). We find that a 2.6B decoder-only MatFormer language model (MatLM) allows us to extract smaller models spanning from 1.5B to 2.6B, each exhibiting comparable validation loss and one-shot downstream evaluations to their independently trained counterparts. Furthermore, we observe that smaller encoders extracted from a universal MatFormer-based ViT (MatViT) encoder preserve the metric-space structure for adaptive large-scale retrieval. Finally, we showcase that speculative decoding with the accurate and consistent submodels extracted from MatFormer can further reduce inference latency."
                },
                {
                    "title": "Large Language Models",
                    "abstract": "Large Language ModelsIn the latest edition of Stats, STAT!, Fralick and colleagues explain the statistics behind large language models - used in chat bots like ChatGPT and Bard. While these new tools may seem remarkably intelligent, at their core they just assemble sentences based on statistics from large amounts of text."
                },
                {
                    "title": "A Simple and Effective Pruning Approach for Large Language Models",
                    "abstract": "As their size increases, Large Languages Models (LLMs) are natural candidates for network pruning methods: approaches that drop a subset of network weights while striving to preserve performance. Existing methods, however, require either retraining, which is rarely affordable for billion-scale LLMs, or solving a weight reconstruction problem reliant on second-order information, which may also be computationally expensive. In this paper, we introduce a novel, straightforward yet effective pruning method, termed Wanda (Pruning by Weights and activations), designed to induce sparsity in pretrained LLMs. Motivated by the recent observation of emergent large magnitude features in LLMs, our approach prunes weights with the smallest magnitudes multiplied by the corresponding input activations, on a per-output basis. Notably, Wanda requires no retraining or weight update, and the pruned LLM can be used as is. We conduct a thorough evaluation of our method Wanda on LLaMA and LLaMA-2 across various language benchmarks. Wanda significantly outperforms the established baseline of magnitude pruning and performs competitively against recent method involving intensive weight update. Code is available at https://github.com/locuslab/wanda."
                },
                {
                    "title": "GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints",
                    "abstract": "Multi-query attention (MQA), which only uses a single key-value head, drastically speeds up decoder inference. However, MQA can lead to quality degradation, and moreover it may not be desirable to train a separate model just for faster inference. We (1) propose a recipe for uptraining existing multi-head language model checkpoints into models with MQA using 5% of original pre-training compute, and (2) introduce grouped-query attention (GQA), a generalization of multi-query attention which uses an intermediate (more than one, less than number of query heads) number of key-value heads. We show that uptrained GQA achieves quality close to multi-head attention with comparable speed to MQA."
                },
                {
                    "title": "Speculative Decoding with Big Little Decoder",
                    "abstract": "The recent emergence of Large Language Models based on the Transformer architecture has enabled dramatic advancements in the field of Natural Language Processing. However, these models have long inference latency, which limits their deployment and makes them prohibitively expensive for various real-time applications. The inference latency is further exacerbated by autoregressive generative tasks, as models need to run iteratively to generate tokens sequentially without leveraging token-level parallelization. To address this, we propose Big Little Decoder (BiLD), a framework that can improve inference efficiency and latency for a wide range of text generation applications. The BiLD framework contains two models with different sizes that collaboratively generate text. The small model runs autoregressively to generate text with a low inference cost, and the large model is only invoked occasionally to refine the small model's inaccurate predictions in a non-autoregressive manner. To coordinate the small and large models, BiLD introduces two simple yet effective policies: (1) the fallback policy that determines when to hand control over to the large model; and (2) the rollback policy that determines when the large model needs to correct the small model's inaccurate predictions. To evaluate our framework across different tasks and models, we apply BiLD to various text generation scenarios encompassing machine translation on IWSLT 2017 De-En and WMT 2014 De-En, and summarization on XSUM and CNN/DailyMail. On an NVIDIA T4 GPU, our framework achieves a speedup of up to 2.12x speedup with minimal generation quality degradation. Furthermore, our framework is fully plug-and-play and can be applied without any modifications in the training process or model architecture. Our code is open-sourced"
                },
                {
                    "title": "Accelerating Large Language Model Decoding with Speculative Sampling",
                    "abstract": "We present speculative sampling, an algorithm for accelerating transformer decoding by enabling the generation of multiple tokens from each transformer call. Our algorithm relies on the observation that the latency of parallel scoring of short continuations, generated by a faster but less powerful draft model, is comparable to that of sampling a single token from the larger target model. This is combined with a novel modified rejection sampling scheme which preserves the distribution of the target model within hardware numerics. We benchmark speculative sampling with Chinchilla, a 70 billion parameter language model, achieving a 2-2.5x decoding speedup in a distributed setup, without compromising the sample quality or making modifications to the model itself."
                },
                {
                    "title": "Fast Inference from Transformers via Speculative Decoding",
                    "abstract": "Inference from large autoregressive models like Transformers is slow - decoding K tokens takes K serial runs of the model. In this work we introduce speculative decoding - an algorithm to sample from autoregressive models faster without any changes to the outputs, by computing several tokens in parallel. At the heart of our approach lie the observations that (1) hard language-modeling tasks often include easier subtasks that can be approximated well by more efficient models, and (2) using speculative execution and a novel sampling method, we can make exact decoding from the large models faster, by running them in parallel on the outputs of the approximation models, potentially generating several tokens concurrently, and without changing the distribution. Our method can accelerate existing off-the-shelf models without retraining or architecture changes. We demonstrate it on T5-XXL and show a 2X-3X acceleration compared to the standard T5X implementation, with identical outputs."
                },
                {
                    "title": "GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers",
                    "abstract": "Generative Pre-trained Transformer models, known as GPT or OPT, set themselves apart through breakthrough performance across complex language modelling tasks, but also by their extremely high computational and storage costs. Specifically, due to their massive size, even inference for large, highly-accurate GPT models may require multiple performant GPUs, which limits the usability of such models. While there is emerging work on relieving this pressure via model compression, the applicability and performance of existing compression techniques is limited by the scale and complexity of GPT models. In this paper, we address this challenge, and propose GPTQ, a new one-shot weight quantization method based on approximate second-order information, that is both highly-accurate and highly-efficient. Specifically, GPTQ can quantize GPT models with 175 billion parameters in approximately four GPU hours, reducing the bitwidth down to 3 or 4 bits per weight, with negligible accuracy degradation relative to the uncompressed baseline. Our method more than doubles the compression gains relative to previously-proposed one-shot quantization methods, preserving accuracy, allowing us for the first time to execute an 175 billion-parameter model inside a single GPU for generative inference. Moreover, we also show that our method can still provide reasonable accuracy in the extreme quantization regime, in which weights are quantized to 2-bit or even ternary quantization levels. We show experimentally that these improvements can be leveraged for end-to-end inference speedups over FP16, of around 3.25x when using high-end GPUs (NVIDIA A100) and 4.5x when using more cost-effective ones (NVIDIA A6000). The implementation is available at https://github.com/IST-DASLab/gptq."
                },
                {
                    "title": "Nearest Neighbor Zero-Shot Inference",
                    "abstract": "Retrieval-augmented language models (LMs) use non-parametric memory to substantially outperform their non-retrieval counterparts on perplexity-based evaluations, but it is an open question whether they achieve similar gains in few- and zero-shot end-task accuracy. We extensively study one such model, the k-nearest neighbor LM (kNN-LM), showing that the gains marginally transfer. The main challenge is to achieve coverage of the verbalizer tokens that define the different end-task class labels. To address this challenge, we also introduce kNN-Prompt, a simple and effective kNN-LM with automatically expanded fuzzy verbalizers (e.g. to expand \u201cterrible\u201d to also include \u201csilly\u201d and other task-specific synonyms for sentiment classification). Across nine diverse end-tasks, using kNN-Prompt with GPT-2 large yields significant performance boosts over strong zeroshot baselines (13.4% absolute improvement over the base LM on average). We also show that other advantages of non-parametric augmentation hold for end tasks; kNN-Prompt is effective for domain adaptation with no further training, and gains increase with the size of the retrieval model."
                },
                {
                    "title": "Evaluating Large Language Models Trained on Code",
                    "abstract": "We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "Linformer: Self-Attention with Linear Complexity",
                    "abstract": "Large transformer models have shown extraordinary success in achieving state-of-the-art results in many natural language processing applications. However, training and deploying these models can be prohibitively costly for long sequences, as the standard self-attention mechanism of the Transformer uses $O(n^2)$ time and space with respect to sequence length. In this paper, we demonstrate that the self-attention mechanism can be approximated by a low-rank matrix. We further exploit this finding to propose a new self-attention mechanism, which reduces the overall self-attention complexity from $O(n^2)$ to $O(n)$ in both time and space. The resulting linear transformer, the \\textit{Linformer}, performs on par with standard Transformer models, while being much more memory- and time-efficient."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Soft Threshold Weight Reparameterization for Learnable Sparsity",
                    "abstract": "Sparsity in Deep Neural Networks (DNNs) is studied extensively with the focus of maximizing prediction accuracy given an overall parameter budget. Existing methods rely on uniform or heuristic non-uniform sparsity budgets which have sub-optimal layer-wise parameter allocation resulting in a) lower prediction accuracy or b) higher inference cost (FLOPs). This work proposes Soft Threshold Reparameterization (STR), a novel use of the soft-threshold operator on DNN weights. STR smoothly induces sparsity while learning pruning thresholds thereby obtaining a non-uniform sparsity budget. Our method achieves state-of-the-art accuracy for unstructured sparsity in CNNs (ResNet50 and MobileNetV1 on ImageNet-1K), and, additionally, learns non-uniform budgets that empirically reduce the FLOPs by up to 50%. Notably, STR boosts the accuracy over existing results by up to 10% in the ultra sparse (99%) regime and can also be used to induce low-rank (structured sparsity) in RNNs. In short, STR is a simple mechanism which learns effective sparsity budgets that contrast with popular heuristics. Code, pretrained models and sparsity budgets are at this https URL."
                },
                {
                    "title": "Reformer: The Efficient Transformer",
                    "abstract": "Large Transformer models routinely achieve state-of-the-art results on a number of tasks but training these models can be prohibitively costly, especially on long sequences. We introduce two techniques to improve the efficiency of Transformers. For one, we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O($L^2$) to O($L\\log L$), where $L$ is the length of the sequence. Furthermore, we use reversible residual layers instead of the standard residuals, which allows storing activations only once in the training process instead of $N$ times, where $N$ is the number of layers. The resulting model, the Reformer, performs on par with Transformer models while being much more memory-efficient and much faster on long sequences."
                },
                {
                    "title": "ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training",
                    "abstract": "This paper presents a new sequence-to-sequence pre-training model called ProphetNet, which introduces a novel self-supervised objective named future n-gram prediction and the proposed n-stream self-attention mechanism. Instead of optimizing one-step-ahead prediction in the traditional sequence-to-sequence model, the ProphetNet is optimized by n-step ahead prediction that predicts the next n tokens simultaneously based on previous context tokens at each time step. The future n-gram prediction explicitly encourages the model to plan for the future tokens and prevent overfitting on strong local correlations. We pre-train ProphetNet using a base scale dataset (16GB) and a large-scale dataset (160GB), respectively. Then we conduct experiments on CNN/DailyMail, Gigaword, and SQuAD 1.1 benchmarks for abstractive summarization and question generation tasks. Experimental results show that ProphetNet achieves new state-of-the-art results on all these datasets compared to the models using the same scale pre-training corpus."
                },
                {
                    "title": "Fast Transformer Decoding: One Write-Head is All You Need",
                    "abstract": "Multi-head attention layers, as used in the Transformer neural sequence model, are a powerful alternative to RNNs for moving information across and between sequences. While training these layers is generally fast and simple, due to parallelizability across the length of the sequence, incremental inference (where such paralleization is impossible) is often slow, due to the memory-bandwidth cost of repeatedly loading the large \"keys\" and \"values\" tensors. We propose a variant called multi-query attention, where the keys and values are shared across all of the different attention \"heads\", greatly reducing the size of these tensors and hence the memory bandwidth requirements of incremental decoding. We verify experimentally that the resulting models can indeed be much faster to decode, and incur only minor quality degradation from the baseline."
                },
                {
                    "title": "Generalization through Memorization: Nearest Neighbor Language Models",
                    "abstract": "We introduce $k$NN-LMs, which extend a pre-trained neural language model (LM) by linearly interpolating it with a $k$-nearest neighbors ($k$NN) model. The nearest neighbors are computed according to distance in the pre-trained LM embedding space, and can be drawn from any text collection, including the original LM training data. Applying this augmentation to a strong Wikitext-103 LM, with neighbors drawn from the original training set, our $k$NN-LM achieves a new state-of-the-art perplexity of 15.79 - a 2.9 point improvement with no additional training. We also show that this approach has implications for efficiently scaling up to larger training sets and allows for effective domain adaptation, by simply varying the nearest neighbor datastore, again without further training. Qualitatively, the model is particularly helpful in predicting rare patterns, such as factual knowledge. Together, these results strongly suggest that learning similarity between sequences of text is easier than predicting the next word, and that nearest neighbor search is an effective approach for language modeling in the long tail."
                },
                {
                    "title": "CodeSearchNet Challenge: Evaluating the State of Semantic Code Search",
                    "abstract": "Semantic code search is the task of retrieving relevant code given a natural language query. While related to other information retrieval tasks, it requires bridging the gap between the language used in code (often abbreviated and highly technical) and natural language more suitable to describe vague concepts and ideas. \nTo enable evaluation of progress on code search, we are releasing the CodeSearchNet Corpus and are presenting the CodeSearchNet Challenge, which consists of 99 natural language queries with about 4k expert relevance annotations of likely results from CodeSearchNet Corpus. The corpus contains about 6 million functions from open-source code spanning six programming languages (Go, Java, JavaScript, PHP, Python, and Ruby). The CodeSearchNet Corpus also contains automatically generated query-like natural language for 2 million functions, obtained from mechanically scraping and preprocessing associated function documentation. In this article, we describe the methodology used to obtain the corpus and expert labels, as well as a number of simple baseline solutions for the task. \nWe hope that CodeSearchNet Challenge encourages researchers and practitioners to study this interesting task further and will host a competition and leaderboard to track the progress on the challenge. We are also keen on extending CodeSearchNet Challenge to more queries and programming languages in the future."
                },
                {
                    "title": "Natural Questions: A Benchmark for Question Answering Research",
                    "abstract": "We present the Natural Questions corpus, a question answering data set. Questions consist of real anonymized, aggregated queries issued to the Google search engine. An annotator is presented with a question along with a Wikipedia page from the top 5 search results, and annotates a long answer (typically a paragraph) and a short answer (one or more entities) if present on the page, or marks null if no long/short answer is present. The public release consists of 307,373 training examples with single annotations; 7,830 examples with 5-way annotations for development data; and a further 7,842 examples with 5-way annotated sequestered as test data. We present experiments validating quality of the data. We also describe analysis of 25-way annotations on 302 examples, giving insights into human variability on the annotation task. We introduce robust metrics for the purposes of evaluating question answering systems; demonstrate high human upper bounds on these metrics; and establish baseline results using competitive methods drawn from related literature."
                },
                {
                    "title": "Gmail Smart Compose: Real-Time Assisted Writing",
                    "abstract": "In this paper, we present Smart Compose, a novel system for generating interactive, real-time suggestions in Gmail that assists users in writing mails by reducing repetitive typing. In the design and deployment of such a large-scale and complicated system, we faced several challenges including model selection, performance evaluation, serving and other practical issues. At the core of Smart Compose is a large-scale neural language model. We leveraged state-of-the-art machine learning techniques for language model training which enabled high-quality suggestion prediction, and constructed novel serving infrastructure for high-throughput and real-time inference. Experimental results show the effectiveness of our proposed system design and deployment approach. This system is currently being served in Gmail."
                },
                {
                    "title": "The Curious Case of Neural Text Degeneration",
                    "abstract": "Despite considerable advancements with deep neural language models, the enigma of neural text degeneration persists when these models are tested as text generators. The counter-intuitive empirical observation is that even though the use of likelihood as training objective leads to high quality models for a broad range of language understanding tasks, using likelihood as a decoding objective leads to text that is bland and strangely repetitive. \nIn this paper, we reveal surprising distributional differences between human text and machine text. In addition, we find that decoding strategies alone can dramatically effect the quality of machine text, even when generated from exactly the same neural language model. Our findings motivate Nucleus Sampling, a simple but effective method to draw the best out of neural generation. By sampling text from the dynamic nucleus of the probability distribution, which allows for diversity while effectively truncating the less reliable tail of the distribution, the resulting text better demonstrates the quality of human text, yielding enhanced diversity without sacrificing fluency and coherence."
                },
                {
                    "title": "Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering",
                    "abstract": "We present a new kind of question answering dataset, OpenBookQA, modeled after open book exams for assessing human understanding of a subject. The open book that comes with our questions is a set of 1326 elementary level science facts. Roughly 6000 questions probe an understanding of these facts and their application to novel situations. This requires combining an open book fact (e.g., metals conduct electricity) with broad common knowledge (e.g., a suit of armor is made of metal) obtained from other sources. While existing QA datasets over documents or knowledge bases, being generally self-contained, focus on linguistic understanding, OpenBookQA probes a deeper understanding of both the topic\u2014in the context of common knowledge\u2014and the language it is expressed in. Human performance on OpenBookQA is close to 92%, but many state-of-the-art pre-trained QA methods perform surprisingly poorly, worse than several simple neural baselines we develop. Our oracle experiments designed to circumvent the knowledge retrieval bottleneck demonstrate the value of both the open book and additional facts. We leave it as a challenge to solve the retrieval problem in this multi-hop setting and to close the large gap to human performance."
                },
                {
                    "title": "Hierarchical Neural Story Generation",
                    "abstract": "We explore story generation: creative systems that can build coherent and fluent passages of text about a topic. We collect a large dataset of 300K human-written stories paired with writing prompts from an online forum. Our dataset enables hierarchical story generation, where the model first generates a premise, and then transforms it into a passage of text. We gain further improvements with a novel form of model fusion that improves the relevance of the story to the prompt, and adding a new gated multi-scale self-attention mechanism to model long-range context. Experiments show large improvements over strong baselines on both automated and human evaluations. Human judges prefer stories generated by our approach to those from a strong non-hierarchical model by a factor of two to one."
                },
                {
                    "title": "Texygen: A Benchmarking Platform for Text Generation Models",
                    "abstract": "We introduce Texygen, a benchmarking platform to support research on open-domain text generation models. Texygen has not only implemented a majority of text generation models, but also covered a set of metrics that evaluate the diversity, the quality and the consistency of the generated texts. The Texygen platform could help standardize the research on text generation and improve the reproductivity and reliability of future research work in text generation."
                },
                {
                    "title": "TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension",
                    "abstract": "We present TriviaQA, a challenging reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents, six per question on average, that provide high quality distant supervision for answering the questions. We show that, in comparison to other recently introduced large-scale datasets, TriviaQA (1) has relatively complex, compositional questions, (2) has considerable syntactic and lexical variability between questions and corresponding answer-evidence sentences, and (3) requires more cross sentence reasoning to find answers. We also present two baseline algorithms: a feature-based classifier and a state-of-the-art neural network, that performs well on SQuAD reading comprehension. Neither approach comes close to human performance (23% and 40% vs. 80%), suggesting that TriviaQA is a challenging testbed that is worth significant future study."
                },
                {
                    "title": "Sequence to Sequence Learning with Neural Networks",
                    "abstract": "Deep Neural Networks (DNNs) are powerful models that have achieved excellent performance on difficult learning tasks. Although DNNs work well whenever large labeled training sets are available, they cannot be used to map sequences to sequences. In this paper, we present a general end-to-end approach to sequence learning that makes minimal assumptions on the sequence structure. Our method uses a multilayered Long Short-Term Memory (LSTM) to map the input sequence to a vector of a fixed dimensionality, and then another deep LSTM to decode the target sequence from the vector. Our main result is that on an English to French translation task from the WMT-14 dataset, the translations produced by the LSTM achieve a BLEU score of 34.8 on the entire test set, where the LSTM's BLEU score was penalized on out-of-vocabulary words. Additionally, the LSTM did not have difficulty on long sentences. For comparison, a phrase-based SMT system achieves a BLEU score of 33.3 on the same dataset. When we used the LSTM to rerank the 1000 hypotheses produced by the aforementioned SMT system, its BLEU score increases to 36.5, which is close to the previous state of the art. The LSTM also learned sensible phrase and sentence representations that are sensitive to word order and are relatively invariant to the active and the passive voice. Finally, we found that reversing the order of the words in all source sentences (but not target sentences) improved the LSTM's performance markedly, because doing so introduced many short term dependencies between the source and the target sentence which made the optimization problem easier."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we generate multiple high-quality autocomplete suggestions from a language model using only a single autoregressive inference pass?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is significant for the research community as it addresses the computational inefficiency of existing decoding methods that require multiple inference passes to generate multiple suggestions. By introducing Superposed Decoding, we can enhance the performance of language models in various applications, such as e-commerce, email composition, and coding assistance. This advancement could lead to more efficient use of resources, enabling broader adoption of language models in real-time applications and potentially inspiring further research into efficient decoding techniques and their applications across different domains.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the inherent complexity of autoregressive models, where each token's generation is dependent on previous tokens, making it difficult to generate multiple coherent suggestions simultaneously. Naive approaches, such as simply sampling multiple times, fail to leverage the interdependencies between tokens, leading to incoherent or irrelevant suggestions. Additionally, the need for maintaining coherence and factual accuracy across multiple drafts adds a layer of complexity, requiring sophisticated methods for token selection and probability scoring that traditional methods do not adequately address.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on improving individual decoding methods like Greedy Decoding, Top-k Sampling, and Nucleus Sampling, which inherently require multiple inference passes for generating multiple suggestions. The limitations of these methods stem from their inability to efficiently utilize the linearity of representations in language models. Barriers such as computational constraints and the lack of innovative approaches to combine embeddings for simultaneous generation have prevented the development of a more efficient solution. Superposed Decoding differs by leveraging a weighted combination of embeddings to generate multiple drafts in a single pass, thus addressing these limitations.\n\n### [Question 5] - What are the key components of my approach and results?\nThe proposed methodology involves Superposed Decoding (SPD), which generates k high-quality suggestions using a single autoregressive inference pass. The approach includes feeding a superposition of the embeddings of the k most recent drafted tokens at each timestep, selecting the top-k output tokens, and expanding the drafts to create k\u00b2 potential new drafts. The probability scores for these drafts are smoothed using n-gram interpolation (n \u2208 [2, 6]) to enhance coherence. The expected outcome is that SPD will produce k suggestions that are as coherent and relevant as those generated by"
            }
        },
        "author_data": {
            "fcd2338d-d925-4e50-893d-70fec4aaf186": {
                "pk": "fcd2338d-d925-4e50-893d-70fec4aaf186",
                "name": "Ethan Shen",
                "collaborators": [
                    "Scotty Singh",
                    "Bhavesh Kumar",
                    "Ali Farhadi",
                    "Aditya Kusupati"
                ],
                "domain": [
                    "Multi-modal Learning",
                    "Visual Question Answering",
                    "Hierarchical Representation",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Generative Visual Question Answering",
                        "abstract": "Multi-modal tasks involving vision and language in deep learning continue to rise in popularity and are leading to the development of newer models that can generalize beyond the extent of their training data. The current models lack temporal generalization which enables models to adapt to changes in future data. This paper discusses a viable approach to creating an advanced Visual Question Answering (VQA) model which can produce successful results on temporal generalization. We propose a new data set, GenVQA, utilizing images and captions from the VQAv2 and MS-COCO dataset to generate new images through stable diffusion. This augmented dataset is then used to test a combination of seven baseline and cutting edge VQA models. Performance evaluation focuses on questions mirroring the original VQAv2 dataset, with the answers having been adjusted to the new images. This paper's purpose is to investigate the robustness of several successful VQA models to assess their performance on future data distributions. Model architectures are analyzed to identify common stylistic choices that improve generalization under temporal distribution shifts. This research highlights the importance of creating a large-scale future shifted dataset. This data can enhance the robustness of VQA models, allowing their future peers to have improved ability to adapt to temporal distribution shifts."
                    },
                    {
                        "title": "Are \"Hierarchical\" Visual Representations Hierarchical?",
                        "abstract": "Learned visual representations often capture large amounts of semantic information for accurate downstream applications. Human understanding of the world is fundamentally grounded in hierarchy. To mimic this and further improve representation capabilities, the community has explored \"hierarchical\" visual representations that aim at modeling the underlying hierarchy of the visual world. In this work, we set out to investigate if hierarchical visual representations truly capture the human perceived hierarchy better than standard learned representations. To this end, we create HierNet, a suite of 12 datasets spanning 3 kinds of hierarchy from the BREEDs subset of ImageNet. After extensive evaluation of Hyperbolic and Matryoshka Representations across training setups, we conclude that they do not capture hierarchy any better than the standard representations but can assist in other aspects like search efficiency and interpretability. Our benchmark and the datasets are open-sourced at https://github.com/ethanlshen/HierNet."
                    }
                ]
            },
            "74692393-d128-4567-8c33-23a17a8acd32": {
                "pk": "74692393-d128-4567-8c33-23a17a8acd32",
                "name": "Alan Fan",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "e96e2465-be5e-4395-86c8-7881afae0371": {
                "pk": "e96e2465-be5e-4395-86c8-7881afae0371",
                "name": "Sarah M. Pratt",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "53370db6-95f5-47f9-9733-c7bb051a8f90": {
                "pk": "53370db6-95f5-47f9-9733-c7bb051a8f90",
                "name": "Jae Sung Park",
                "collaborators": [
                    "Dinesh Manocha",
                    "Chonhyon Park",
                    "Siamak Mirfendereski",
                    "Trevor Darrell",
                    "Anna Rohrbach",
                    "Michael D. Graham",
                    "Jae Woong Soh",
                    "Nam Ik Cho",
                    "Alex Rogge"
                ],
                "domain": [
                    "Robotics",
                    "Motion Planning",
                    "Probabilistic Collision Detection",
                    "Fluid Dynamics"
                ],
                "publications": [
                    {
                        "title": "Efficient Probabilistic Collision Detection for Non-Convex Shapes",
                        "abstract": "We present new algorithms to perform fast probabilistic collision queries between convex as well as non-convex objects. Our approach is applicable to general shapes, where one or more objects are represented using Gaussian probability distributions. We present a fast new algorithm for a pair of convex objects, and extend the approach to non-convex models using hierarchical representations. We highlight the performance of our algorithms with various convex and non-convex shapes on complex synthetic benchmarks and trajectory planning benchmarks for a 7-DOF Fetch robot arm."
                    },
                    {
                        "title": "Efficient Probabilistic Collision Detection for Non-Gaussian Noise Distributions",
                        "abstract": "We present an efficient algorithm to compute tight upper bounds of collision probability between two objects with positional uncertainties, whose error distributions are represented with non-Gaussian forms. Our approach can handle noisy datasets from depth sensors, whose distributions may correspond to Truncated Gaussian, Weighted Samples, or Truncated Gaussian Mixture Model. We derive tight probability bounds for convex shapes and extend them to non-convex shapes using hierarchical representations. We highlight the benefits of our approach over prior probabilistic collision detection algorithms in terms of tighter bounds ($10$x) and improved running time ($3$x). Moreover, we use our tight bounds to design an efficient and accurate motion planning algorithm for a 7-DOF robot arm operating in tight scenarios with sensor and motion uncertainties."
                    },
                    {
                        "title": "HMPO: Human Motion Prediction in Occluded Environments for Safe Motion Planning",
                        "abstract": "We present a novel approach to generate collision-free trajectories for a robot operating in close proximity with a human obstacle in an occluded environment. The self-occlusions of the robot can significantly reduce the accuracy of human motion prediction, and we present a novel deep learning-based prediction algorithm. Our formulation uses CNNs and LSTMs and we augment human-action datasets with synthetically generated occlusion information for training. We also present an occlusion-aware planner that uses our motion prediction algorithm to compute collision-free trajectories. We highlight performance of the overall approach (HMPO) in complex scenarios and observe upto 68% performance improvement in motion prediction accuracy, and 38% improvement in terms of error distance between the ground-truth and the predicted human joint positions."
                    },
                    {
                        "title": "Identity-Aware Multi-Sentence Video Description",
                        "abstract": "Standard video and movie description tasks abstract away from person identities, thus failing to link identities across sentences. We propose a multi-sentence Identity-Aware Video Description task, which overcomes this limitation and requires to re-identify persons locally within a set of consecutive clips. We introduce an auxiliary task of Fill-in the Identity, that aims to predict persons' IDs consistently within a set of clips, when the video descriptions are given. Our proposed approach to this task leverages a Transformer architecture allowing for coherent joint prediction of multiple IDs. One of the key components is a gender-aware textual representation as well an additional gender prediction objective in the main model. This auxiliary task allows us to propose a two-stage approach to Identity-Aware Video Description. We first generate multi-sentence video descriptions, and then apply our Fill-in the Identity model to establish links between the predicted person entities. To be able to tackle both tasks, we augment the Large Scale Movie Description Challenge (LSMDC) benchmark with new annotations suited for our problem statement. Experiments show that our proposed Fill-in the Identity model is superior to several baselines and recent works, and allows us to generate descriptions with locally re-identified people."
                    },
                    {
                        "title": "Exact coherent states and connections to turbulent dynamics in minimal channel flow",
                        "abstract": "Several new families of nonlinear three-dimensional travelling wave solutions to the Navier-Stokes equation, also known as exact coherent states, are computed for Newtonian plane Poiseuille flow. The symmetries and streak/vortex structures are reported and their possible connections to critical layer dynamics examined. While some of the solutions clearly display fluctuations that are localized around the critical layer (the surface on which the streamwise velocity matches the wave speed of the solution), for others this connection is not as clear. Dynamical trajectories along unstable directions of the solutions are computed. Over certain ranges of Reynolds number, two solution families are shown to lie on the basin boundary between laminar and turbulent flow. Direct comparison of nonlinear travelling wave solutions to turbulent flow in the same channel is presented. The state-space dynamics of the turbulent flow are organized around one of the newly-identified travelling wave families, and in particular the lower branch solutions of this family are closely approached during transient excursions away from the dominant behaviour. These observations provide a firm dynamical-systems foundation for prior observations that minimal channel turbulence displays time intervals of \"active\" turbulence punctuated by brief periods of \"hibernation\" (see e.g. Xi, L. and Graham, M. D., Phys. Rev. Lett., 104, 218301 (2010)). The hibernating intervals are approaches to lower branch nonlinear travelling waves. Representing these solutions on a Prandtl-von Karman plot illustrates how their bulk flow properties are related to those of Newtonian turbulence as well as the universal asymptotic state called maximum drag reduction (MDR) found in viscoelastic turbulent flow."
                    },
                    {
                        "title": "Fast and Bounded Probabilistic Collision Detection in Dynamic Environments for High-DOF Trajectory Planning",
                        "abstract": "We present a novel approach to perform probabilistic collision detection between a high-DOF robot and high-DOF obstacles in dynamic, uncertain environments. In dynamic environments with a high-DOF robot and moving obstacles, our approach efficiently computes accurate collision probability between the robot and obstacles with upper error bounds. Furthermore, we describe a prediction algorithm for future obstacle position and motion that accounts for both spatial and temporal uncertainties. We present a trajectory optimization algorithm for high-DOF robots in dynamic, uncertain environments based on probabilistic collision detection. We highlight motion planning performance in challenging scenarios with robot arms operating in environments with dynamically moving human obstacles."
                    },
                    {
                        "title": "Joint High Dynamic Range Imaging and Super-Resolution from a Single Image",
                        "abstract": "This paper presents a new framework for jointly enhancing the resolution and the dynamic range of an image, i.e., simultaneous super-resolution (SR) and high dynamic range imaging (HDRI), based on a convolutional neural network (CNN). From the common trends of both tasks, we train a CNN for the joint HDRI and SR by focusing on the reconstruction of high-frequency details. Specifically, the high-frequency component in our work is the reflectance component according to the Retinex-based image decomposition, and only the reflectance component is manipulated by the CNN while another component (illumination) is processed in a conventional way. In training the CNN, we devise an appropriate loss function that contributes to the naturalness quality of resulting images. Experiments show that our algorithm outperforms the cascade implementation of CNN-based SR and HDRI."
                    },
                    {
                        "title": "I-Planner: Intention-Aware Motion Planning Using Learning Based Human Motion Prediction",
                        "abstract": "We present a motion planning algorithm to compute collision-free and smooth trajectories for high-DOF robots interacting with humans in a shared workspace. Our approach uses offline learning of human actions along with temporal coherence to predict the human actions. Our intention-aware online planning algorithm uses the learned database to compute a reliable trajectory based on the predicted actions. We represent the predicted human motion using a Gaussian distribution and compute tight upper bounds on collision probabilities for safe motion planning. We also describe novel techniques to account for noise in human motion prediction. We highlight the performance of our planning algorithm in complex simulated scenarios and real world benchmarks with 7-DOF robot arms operating in a workspace with a human performing complex tasks. We demonstrate the benefits of our intention-aware planner in terms of computing safe trajectories in such uncertain environments."
                    },
                    {
                        "title": "Dipolophoresis in concentrated suspensions of ideally polarizable spheres",
                        "abstract": "The dynamics of ideally polarizable spherical particles in concentrated suspensions under the effects of nonlinear electrokinetic phenomena is analysed using large-scale numerical simulations. Particles are assumed to carry no net charge and considered to undergo the combination of dielectrophoresis and induced-charge electrophoresis termed dipolophoresis. Chaotic motion and resulting hydrodynamic diffusion are known to be driven by the induced-charge electrophoresis, which dominates the dielectrophoresis. Up to a volume fraction $\\phi \\approx 35\\%$, the particle dynamics seems to be hindered by the increase in the magnitude of excluded volume interactions with concentration. However, a non-trivial suspension behaviour is observed in concentrated regimes, where the hydrodynamic diffusivity starts to increase with a volume fraction at $\\phi \\approx 35\\%$ before reaching a local maximum and then drastically decreases as approaching random close packing. Similar non-trivial behaviours are observed in the particle velocity and number-density fluctuations around volume fractions that the non-trivial behaviour of the hydrodynamic diffusion is observed. We explain these non-trivial behaviours as a consequence of particle contacts, which are related to the dominant mechanism of particle pairings. The particle contacts are classified into attractive and repulsive classes by the nature of contacts, and in particular, the strong repulsive contact becomes predominant at $\\phi > 20\\%$. Moreover, this transition is visible in the pair distribution functions, which also reveal the change in the suspension microstructure in concentrated regimes. It appears that strong and massive repulsive contacts along the direction perpendicular to an electric field promote the non-trivial suspension behaviours observed in concentrated regimes."
                    },
                    {
                        "title": "On the underlying drag-reduction mechanisms of flow-control strategies in a transitional channel flow: temporal approach",
                        "abstract": "The underlying mechanisms of three different flow-control strategies on drag reduction in a channel flow are investigated by direct numerical simulations at friction Reynolds numbers ranging from 65 to 85. These strategies include the addition of long-chain polymers, the incorporation of slip surfaces, and the application of an external body force. While it has been believed that such methods lead to a skin-friction reduction by controlling near-wall flow structures, the underlying mechanisms at play are still not as clear. In this study, a temporal analysis is employed to elucidate underlying drag-reduction mechanisms among these methods. The analysis is based on the lifetime of intermittent phases represented by the active and hibernating phases of a minimal turbulent channel flow (Xi & Graham, Phy. Rev. Lett. 2010). At a similar amount of drag reduction, the polymer and slip methods show a similar mechanism, while the body force method is different. The polymers and slip surfaces cause hibernating phases to happen more frequently, while the duration of active phases is decreased. However, the body forces cause hibernating phases to happen less frequently but prolong its duration to achieve a comparable amount of drag reduction. A possible mechanism behind the body force method is associated with its unique roller-like vortical structures formed near the wall. These structures appear to prevent interactions between inner and outer regions by which hibernating phases are prolonged. It should motivate adaptive flow-control strategies to exploit the distinct underlying mechanisms for robust control of turbulent drag at low Reynolds numbers."
                    },
                    {
                        "title": "Rheology of dense suspensions of ideally conductive particles in an electric field",
                        "abstract": "The rheological behaviour of dense suspensions of ideally conductive particles in the presence of both electric field and shear flow is studied using large-scale numerical simulations. Under the action of an electric field, these particles are known to undergo dipolophoresis, which is the combination of two nonlinear electrokinetic phenomena -- induced-charge electrophoresis and dielectrophoresis. For ideally conductive particles, induced-charge electrophoresis is predominant over dielectrophoresis, resulting in transient pairing dynamics. The shear viscosity and first and second normal stress differences $N_1$ and $N_2$ of such suspensions are examined over a range of volume fractions $15\\% \\leqslant \\phi \\leqslant 50\\%$ as a function of Mason number $Mn$, which measures the relative importance of viscous shear stress over electrokinetic-driven stress. For $Mn < 1$ or low shear rates, the dipolophoresis is shown to dominate the dynamics, resulting in a relatively low-viscosity state. The positive $N_1$ and negative $N_2$ are observed at $\\phi < 30\\%$, which is similar to Brownian suspensions, while their signs are reversed at $\\phi \\ge 30\\%$. For $Mn \\ge 1$, the shear thickening starts to arise at $\\phi \\ge 30\\%$, and an almost five-fold increase in viscosity occurs at $\\phi = 50\\%$. Both $N_1$ and $N_2$ are negative for $Mn \\gg 1$ at all volume fractions considered. We illuminate the transition in rheological behaviours from dipolophoresis to shear dominance around $Mn = 1$ in connection to suspension microstructure and dynamics. Lastly, our findings reveal the potential use of nonlinear electrokinetics as a means of active rheology control for such suspensions."
                    }
                ]
            },
            "2285187b-4847-489f-8671-343cbdfe9540": {
                "pk": "2285187b-4847-489f-8671-343cbdfe9540",
                "name": "Matthew Wallingford",
                "collaborators": [
                    "Ali Farhadi",
                    "Aditya Kusupati",
                    "Vivek Ramanujan",
                    "Aniruddha Kembhavi",
                    "Roozbeh Mottaghi",
                    "Ludwig Schmidt",
                    "Alex Fang",
                    "Prateek Jain",
                    "Sham Kakade",
                    "Pang Wei Koh"
                ],
                "domain": [
                    "Representation Learning",
                    "Transfer Learning",
                    "3D Vision",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Neural Radiance Field Codebooks",
                        "abstract": "Compositional representations of the world are a promising step towards enabling high-level scene understanding and efficient transfer to downstream tasks. Learning such representations for complex scenes and tasks remains an open challenge. Towards this goal, we introduce Neural Radiance Field Codebooks (NRC), a scalable method for learning object-centric representations through novel view reconstruction. NRC learns to reconstruct scenes from novel views using a dictionary of object codes which are decoded through a volumetric renderer. This enables the discovery of reoccurring visual and geometric patterns across scenes which are transferable to downstream tasks. We show that NRC representations transfer well to object navigation in THOR, outperforming 2D and 3D representation learning methods by 3.1% success rate. We demonstrate that our approach is able to perform unsupervised segmentation for more complex synthetic (THOR) and real scenes (NYU Depth) better than prior methods (29% relative improvement). Finally, we show that NRC improves on the task of depth ordering by 5.5% accuracy in THOR."
                    },
                    {
                        "title": "Neural Priming for Sample-Efficient Adaptation",
                        "abstract": "We propose Neural Priming, a technique for adapting large pretrained models to distribution shifts and downstream tasks given few or no labeled examples. Presented with class names or unlabeled test samples, Neural Priming enables the model to recall and conditions its parameters on relevant data seen throughout pretraining, thereby priming it for the test distribution. Neural Priming can be performed at test time, even for pretraining datasets as large as LAION-2B. Performing lightweight updates on the recalled data significantly improves accuracy across a variety of distribution shift and transfer learning benchmarks. Concretely, in the zero-shot setting, we see a 2.45% improvement in accuracy on ImageNet and 3.81% accuracy improvement on average across standard transfer learning benchmarks. Further, using Neural Priming at inference to adapt to distribution shift, we see a 1.41% accuracy improvement on ImageNetV2. These results demonstrate the effectiveness of Neural Priming in addressing the challenge of limited labeled data and changing distributions. Code is available at github.com/RAIVNLab/neural-priming."
                    },
                    {
                        "title": "Task Adaptive Parameter Sharing for Multi-Task Learning",
                        "abstract": "Adapting pre-trained models with broad capabilities has become standard practice for learning a wide range of downstream tasks. The typical approach of fine-tuning different models for each task is performant, but incurs a substantial memory cost. To efficiently learn multiple downstream tasks we introduce Task Adaptive Parameter Sharing (TAPS), a general method for tuning a base model to a new task by adaptively modifying a small, task-specific subset of layers. This enables multi-task learning while minimizing resources used and competition between tasks. TAPS solves a joint optimization problem which determines which layers to share with the base model and the value of the task-specific weights. Further, a sparsity penalty on the number of active layers encourages weight sharing with the base model. Compared to other methods, TAPS retains high accuracy on downstream tasks while introducing few task-specific parameters. Moreover, TAPS is agnostic to the model architecture and requires only minor changes to the training scheme. We evaluate our method on a suite of fine-tuning tasks and architectures (ResNet, DenseNet, ViT) and show that it achieves state-of-the-art performance while being simple to implement."
                    },
                    {
                        "title": "FLUID: A Unified Evaluation Framework for Flexible Sequential Data",
                        "abstract": "Modern ML methods excel when training data is IID, large-scale, and well labeled. Learning in less ideal conditions remains an open challenge. The sub-fields of few-shot, continual, transfer, and representation learning have made substantial strides in learning under adverse conditions; each affording distinct advantages through methods and insights. These methods address different challenges such as data arriving sequentially or scarce training examples, however often the difficult conditions an ML system will face over its lifetime cannot be anticipated prior to deployment. Therefore, general ML systems which can handle the many challenges of learning in practical settings are needed. To foster research towards the goal of general ML methods, we introduce a new unified evaluation framework - FLUID (Flexible Sequential Data). FLUID integrates the objectives of few-shot, continual, transfer, and representation learning while enabling comparison and integration of techniques across these subfields. In FLUID, a learner faces a stream of data and must make sequential predictions while choosing how to update itself, adapt quickly to novel classes, and deal with changing data distributions; while accounting for the total amount of compute. We conduct experiments on a broad set of methods which shed new insight on the advantages and limitations of current solutions and indicate new research problems to solve. As a starting point towards more general methods, we present two new baselines which outperform other evaluated methods on FLUID. Project page: https://raivn.cs.washington.edu/projects/FLUID/."
                    },
                    {
                        "title": "LLC: Accurate, Multi-purpose Learnt Low-dimensional Binary Codes",
                        "abstract": "Learning binary representations of instances and classes is a classical problem with several high potential applications. In modern settings, the compression of high-dimensional neural representations to low-dimensional binary codes is a challenging task and often require large bit-codes to be accurate. In this work, we propose a novel method for Learning Low-dimensional binary Codes (LLC) for instances as well as classes. Our method does not require any side-information, like annotated attributes or label meta-data, and learns extremely low-dimensional binary codes (~20 bits for ImageNet-1K). The learnt codes are super-efficient while still ensuring nearly optimal classification accuracy for ResNet50 on ImageNet-1K. We demonstrate that the learnt codes capture intrinsically important features in the data, by discovering an intuitive taxonomy over classes. We further quantitatively measure the quality of our codes by applying it to the efficient image retrieval as well as out-of-distribution (OOD) detection problems. For ImageNet-100 retrieval problem, our learnt binary codes outperform 16 bit HashNet using only 10 bits and also are as accurate as 10 dimensional real representations. Finally, our learnt binary codes can perform OOD detection, out-of-the-box, as accurately as a baseline that needs ~3000 samples to tune its threshold, while we require none. Code is open-sourced at https://github.com/RAIVNLab/LLC."
                    },
                    {
                        "title": "The Unmet Promise of Synthetic Training Images: Using Retrieved Real Images Performs Better",
                        "abstract": "Generative text-to-image models enable us to synthesize unlimited amounts of images in a controllable manner, spurring many recent efforts to train vision models with synthetic data. However, every synthetic image ultimately originates from the upstream data used to train the generator. What additional value does the intermediate generator provide over directly training on relevant parts of the upstream data? Grounding this question in the setting of image classification,a we compare finetuning on task-relevant, targeted synthetic data generated by Stable Diffusion -- a generative model trained on the LAION-2B dataset -- against finetuning on targeted real images retrieved directly from LAION-2B. We show that while synthetic data can benefit some downstream tasks, it is universally matched or outperformed by real data from our simple retrieval baseline. Our analysis suggests that this underperformance is partially due to generator artifacts and inaccurate task-relevant visual details in the synthetic images. Overall, we argue that retrieval is a critical baseline to consider when training with synthetic data -- a baseline that current methods do not yet surpass. We release code, data, and models at https://github.com/scottgeng00/unmet-promise."
                    },
                    {
                        "title": "RoboTHOR: An Open Simulation-to-Real Embodied AI Platform",
                        "abstract": "Visual recognition ecosystems (e.g. ImageNet, Pascal, COCO) have undeniably played a prevailing role in the evolution of modern computer vision. We argue that interactive and embodied visual AI has reached a stage of development similar to visual recognition prior to the advent of these ecosystems. Recently, various synthetic environments have been introduced to facilitate research in embodied AI. Notwithstanding this progress, the crucial question of how well models trained in simulation generalize to reality has remained largely unanswered. The creation of a comparable ecosystem for simulation-to-real embodied AI presents many challenges: (1) the inherently interactive nature of the problem, (2) the need for tight alignments between real and simulated worlds, (3) the difficulty of replicating physical conditions for repeatable experiments, (4) and the associated cost. In this paper, we introduce RoboTHOR to democratize research in interactive and embodied visual AI. RoboTHOR offers a framework of simulated environments paired with physical counterparts to systematically explore and overcome the challenges of simulation-to-real transfer, and a platform where researchers across the globe can remotely test their embodied models in the physical world. As a first benchmark, our experiments show there exists a significant gap between the performance of models trained in simulation when they are tested in both simulations and their carefully constructed physical analogs. We hope that RoboTHOR will spur the next stage of evolution in embodied computer vision. RoboTHOR can be accessed at the following link: https://ai2thor.allenai.org/robothor"
                    },
                    {
                        "title": "Multilingual Diversity Improves Vision-Language Representations",
                        "abstract": "Massive web-crawled image-text datasets lay the foundation for recent progress in multimodal learning. These datasets are designed with the goal of training a model to do well on standard computer vision benchmarks, many of which, however, have been shown to be English-centric (e.g., ImageNet). Consequently, existing data curation techniques gravitate towards using predominantly English image-text pairs and discard many potentially useful non-English samples. Our work questions this practice. Multilingual data is inherently enriching not only because it provides a gateway to learn about culturally salient concepts, but also because it depicts common concepts differently from monolingual data. We thus conduct a systematic study to explore the performance benefits of using more samples of non-English origins with respect to English vision tasks. By translating all multilingual image-text pairs from a raw web crawl to English and re-filtering them, we increase the prevalence of (translated) multilingual data in the resulting training set. Pre-training on this dataset outperforms using English-only or English-dominated datasets on ImageNet, ImageNet distribution shifts, image-English-text retrieval and on average across 38 tasks from the DataComp benchmark. On a geographically diverse task like GeoDE, we also observe improvements across all regions, with the biggest gain coming from Africa. In addition, we quantitatively show that English and non-English data are significantly different in both image and (translated) text space. We hope that our findings motivate future work to be more intentional about including multicultural and multilingual data, not just when non-English or geographically diverse tasks are involved, but to enhance model capabilities at large."
                    },
                    {
                        "title": "Objaverse-XL: A Universe of 10M+ 3D Objects",
                        "abstract": "Natural language processing and 2D vision models have attained remarkable proficiency on many tasks primarily by escalating the scale of training data. However, 3D vision tasks have not seen the same progress, in part due to the challenges of acquiring high-quality 3D data. In this work, we present Objaverse-XL, a dataset of over 10 million 3D objects. Our dataset comprises deduplicated 3D objects from a diverse set of sources, including manually designed objects, photogrammetry scans of landmarks and everyday items, and professional scans of historic and antique artifacts. Representing the largest scale and diversity in the realm of 3D datasets, Objaverse-XL enables significant new possibilities for 3D vision. Our experiments demonstrate the improvements enabled with the scale provided by Objaverse-XL. We show that by training Zero123 on novel view synthesis, utilizing over 100 million multi-view rendered images, we achieve strong zero-shot generalization abilities. We hope that releasing Objaverse-XL will enable further innovations in the field of 3D vision at scale."
                    },
                    {
                        "title": "Matryoshka Representation Learning",
                        "abstract": "Learned representations are a central component in modern ML systems, serving a multitude of downstream tasks. When training such representations, it is often the case that computational and statistical constraints for each downstream task are unknown. In this context rigid, fixed capacity representations can be either over or under-accommodating to the task at hand. This leads us to ask: can we design a flexible representation that can adapt to multiple downstream tasks with varying computational resources? Our main contribution is Matryoshka Representation Learning (MRL) which encodes information at different granularities and allows a single embedding to adapt to the computational constraints of downstream tasks. MRL minimally modifies existing representation learning pipelines and imposes no additional cost during inference and deployment. MRL learns coarse-to-fine representations that are at least as accurate and rich as independently trained low-dimensional representations. The flexibility within the learned Matryoshka Representations offer: (a) up to 14x smaller embedding size for ImageNet-1K classification at the same level of accuracy; (b) up to 14x real-world speed-ups for large-scale retrieval on ImageNet-1K and 4K; and (c) up to 2% accuracy improvements for long-tail few-shot classification, all while being as robust as the original representations. Finally, we show that MRL extends seamlessly to web-scale datasets (ImageNet, JFT) across various modalities -- vision (ViT, ResNet), vision + language (ALIGN) and language (BERT). MRL code and pretrained models are open-sourced at https://github.com/RAIVNLab/MRL."
                    }
                ]
            },
            "4a164074-492c-4f3a-bff6-8d887db3785a": {
                "pk": "4a164074-492c-4f3a-bff6-8d887db3785a",
                "name": "Sham M. Kakade",
                "collaborators": [
                    "Daniel Hsu",
                    "Tong Zhang",
                    "Ambuj Tewari",
                    "Qingqing Huang",
                    "Elad Hazan",
                    "Karan Singh",
                    "Shai Shalev-Shwartz",
                    "Ilan Lobel",
                    "Hamid Nazerzadeh",
                    "Eyal Even-Dar"
                ],
                "domain": [
                    "Statistical Learning",
                    "Machine Learning",
                    "Reinforcement Learning",
                    "Matrix Theory"
                ],
                "publications": [
                    {
                        "title": "A tail inequality for quadratic forms of subgaussian random vectors",
                        "abstract": "We prove an exponential probability tail inequality for positive semidefinite quadratic forms in a subgaussian random vector. The bound is analogous to one that holds when the vector has independent Gaussian entries."
                    },
                    {
                        "title": "Dimension-free tail inequalities for sums of random matrices",
                        "abstract": "We derive exponential tail inequalities for sums of random matrices with no dependence on the explicit matrix dimensions. These are similar to the matrix versions of the Chernoff bound and Bernstein inequality except with the explicit matrix dimensions replaced by a trace quantity that can be small even when the dimension is large or infinite. Some applications to principal component analysis and approximate matrix multiplication are given to illustrate the utility of the new bounds."
                    },
                    {
                        "title": "Learning mixtures of spherical Gaussians: moment methods and spectral decompositions",
                        "abstract": "This work provides a computationally efficient and statistically consistent moment-based estimator for mixtures of spherical Gaussians. Under the condition that component means are in general position, a simple spectral decomposition technique yields consistent parameter estimates from low-order observable moments, without additional minimum separation assumptions needed by previous computationally efficient estimation procedures. Thus computational and information-theoretic barriers to efficient estimation in mixture models are precluded when the mixture components have means in general position and spherical covariances. Some connections are made to estimation problems related to independent component analysis."
                    },
                    {
                        "title": "The Nonstochastic Control Problem",
                        "abstract": "We consider the problem of controlling an unknown linear dynamical system in the presence of (nonstochastic) adversarial perturbations and adversarial convex loss functions. In contrast to classical control, the a priori determination of an optimal controller here is hindered by the latter's dependence on the yet unknown perturbations and costs. Instead, we measure regret against an optimal linear policy in hindsight, and give the first efficient algorithm that guarantees a sublinear regret bound, scaling as T^{2/3}, in this setting."
                    },
                    {
                        "title": "Analysis of a randomized approximation scheme for matrix multiplication",
                        "abstract": "This note gives a simple analysis of a randomized approximation scheme for matrix multiplication proposed by Sarlos (2006) based on a random rotation followed by uniform column sampling. The result follows from a matrix version of Bernstein's inequality and a tail inequality for quadratic forms in subgaussian random vectors."
                    },
                    {
                        "title": "Regularization Techniques for Learning with Matrices",
                        "abstract": "There is growing body of learning problems for which it is natural to organize the parameters into matrix, so as to appropriately regularize the parameters under some matrix norm (in order to impose some more sophisticated prior knowledge). This work describes and analyzes a systematic method for constructing such matrix-based, regularization methods. In particular, we focus on how the underlying statistical properties of a given problem can help us decide which regularization function is appropriate.   Our methodology is based on the known duality fact: that a function is strongly convex with respect to some norm if and only if its conjugate function is strongly smooth with respect to the dual norm. This result has already been found to be a key component in deriving and analyzing several learning algorithms. We demonstrate the potential of this framework by deriving novel generalization and regret bounds for multi-task learning, multi-class learning, and kernel learning."
                    },
                    {
                        "title": "An Optimal Dynamic Mechanism for Multi-Armed Bandit Processes",
                        "abstract": "We consider the problem of revenue-optimal dynamic mechanism design in settings where agents' types evolve over time as a function of their (both public and private) experience with items that are auctioned repeatedly over an infinite horizon. A central question here is understanding what natural restrictions on the environment permit the design of optimal mechanisms (note that even in the simpler static setting, optimal mechanisms are characterized only under certain restrictions). We provide a {\\em structural characterization} of a natural \"separable: multi-armed bandit environment (where the evolution and incentive structure of the a-priori type is decoupled from the subsequent experience in a precise sense) where dynamic optimal mechanism design is possible. Here, we present the Virtual Index Mechanism, an optimal dynamic mechanism, which maximizes the (long term) {\\em virtual surplus} using the classical Gittins algorithm. The mechanism optimally balances exploration and exploitation, taking incentives into account."
                    },
                    {
                        "title": "Robust Matrix Decomposition with Outliers",
                        "abstract": "Suppose a given observation matrix can be decomposed as the sum of a low-rank matrix and a sparse matrix (outliers), and the goal is to recover these individual components from the observed sum. Such additive decompositions have applications in a variety of numerical problems including system identification, latent variable graphical modeling, and principal components analysis. We study conditions under which recovering such a decomposition is possible via a combination of $\\ell_1$ norm and trace norm minimization. We are specifically interested in the question of how many outliers are allowed so that convex programming can still achieve accurate recovery, and we obtain stronger recovery guarantees than previous studies. Moreover, we do not assume that the spatial pattern of outliers is random, which stands in contrast to related analyses under such assumptions via matrix completion."
                    },
                    {
                        "title": "Random design analysis of ridge regression",
                        "abstract": "This work gives a simultaneous analysis of both the ordinary least squares estimator and the ridge regression estimator in the random design setting under mild assumptions on the covariate/response distributions. In particular, the analysis provides sharp results on the ``out-of-sample'' prediction error, as opposed to the ``in-sample'' (fixed design) error. The analysis also reveals the effect of errors in the estimated covariance structure, as well as the effect of modeling errors, neither of which effects are present in the fixed design setting. The proofs of the main results are based on a simple decomposition lemma combined with concentration inequalities for random vectors and matrices."
                    },
                    {
                        "title": "Planning in POMDPs Using Multiplicity Automata",
                        "abstract": "Planning and learning in Partially Observable MDPs (POMDPs) are among the most challenging tasks in both the AI and Operation Research communities. Although solutions to these problems are intractable in general, there might be special cases, such as structured POMDPs, which can be solved efficiently. A natural and possibly efficient way to represent a POMDP is through the predictive state representation (PSR) - a representation which recently has been receiving increasing attention. In this work, we relate POMDPs to multiplicity automata- showing that POMDPs can be represented by multiplicity automata with no increase in the representation size. Furthermore, we show that the size of the multiplicity automaton is equal to the rank of the predictive state representation. Therefore, we relate both the predictive state representation and POMDPs to the well-founded multiplicity automata literature. Based on the multiplicity automata representation, we provide a planning algorithm which is exponential only in the multiplicity automata rank rather than the number of states of the POMDP. As a result, whenever the predictive state representation is logarithmic in the standard POMDP representation, our planning algorithm is efficient."
                    },
                    {
                        "title": "Robust Aggregation for Federated Learning",
                        "abstract": "Federated learning is the centralized training of statistical models from decentralized data on mobile devices while preserving the privacy of each device. We present a robust aggregation approach to make federated learning robust to settings when a fraction of the devices may be sending corrupted updates to the server. The approach relies on a robust aggregation oracle based on the geometric median, which returns a robust aggregate using a constant number of iterations of a regular non-robust averaging oracle. The robust aggregation oracle is privacy-preserving, similar to the non-robust secure average oracle it builds upon. We establish its convergence for least squares estimation of additive models. We provide experimental results with linear models and deep networks for three tasks in computer vision and natural language processing. The robust aggregation approach is agnostic to the level of corruption; it outperforms the classical aggregation approach in terms of robustness when the level of corruption is high, while being competitive in the regime of low corruption. Two variants, a faster one with one-step robust aggregation and another one with on-device personalization, round off the paper."
                    },
                    {
                        "title": "A Method of Moments for Mixture Models and Hidden Markov Models",
                        "abstract": "Mixture models are a fundamental tool in applied statistics and machine learning for treating data taken from multiple subpopulations. The current practice for estimating the parameters of such models relies on local search heuristics (e.g., the EM algorithm) which are prone to failure, and existing consistent methods are unfavorable due to their high computational and sample complexity which typically scale exponentially with the number of mixture components. This work develops an efficient method of moments approach to parameter estimation for a broad class of high-dimensional mixture models with many components, including multi-view mixtures of Gaussians (such as mixtures of axis-aligned Gaussians) and hidden Markov models. The new method leads to rigorous unsupervised learning results for mixture models that were not achieved by previous works; and, because of its simplicity, it offers a viable alternative to EM for practical deployment."
                    },
                    {
                        "title": "A Spectral Algorithm for Learning Hidden Markov Models",
                        "abstract": "Hidden Markov Models (HMMs) are one of the most fundamental and widely used statistical tools for modeling discrete time series. In general, learning HMMs from data is computationally hard (under cryptographic assumptions), and practitioners typically resort to search heuristics which suffer from the usual local optima issues. We prove that under a natural separation condition (bounds on the smallest singular value of the HMM parameters), there is an efficient and provably correct algorithm for learning HMMs. The sample complexity of the algorithm does not explicitly depend on the number of distinct (discrete) observations---it implicitly depends on this quantity through spectral properties of the underlying HMM. This makes the algorithm particularly applicable to settings with a large number of observations, such as those in natural language processing where the space of observation is sometimes the words in a language. The algorithm is also simple, employing only a singular value decomposition and matrix multiplications."
                    },
                    {
                        "title": "Learning Mixtures of Gaussians in High Dimensions",
                        "abstract": "Efficiently learning mixture of Gaussians is a fundamental problem in statistics and learning theory. Given samples coming from a random one out of k Gaussian distributions in Rn, the learning problem asks to estimate the means and the covariance matrices of these Gaussians. This learning problem arises in many areas ranging from the natural sciences to the social sciences, and has also found many machine learning applications. Unfortunately, learning mixture of Gaussians is an information theoretically hard problem: in order to learn the parameters up to a reasonable accuracy, the number of samples required is exponential in the number of Gaussian components in the worst case. In this work, we show that provided we are in high enough dimensions, the class of Gaussian mixtures is learnable in its most general form under a smoothed analysis framework, where the parameters are randomly perturbed from an adversarial starting point. In particular, given samples from a mixture of Gaussians with randomly perturbed parameters, when n > {\\Omega}(k^2), we give an algorithm that learns the parameters with polynomial running time and using polynomial number of samples. The central algorithmic ideas consist of new ways to decompose the moment tensor of the Gaussian mixture by exploiting its structural properties. The symmetries of this tensor are derived from the combinatorial structure of higher order moments of Gaussian distributions (sometimes referred to as Isserlis' theorem or Wick's theorem). We also develop new tools for bounding smallest singular values of structured random matrices, which could be useful in other smoothed analysis settings."
                    },
                    {
                        "title": "Provable Efficient Online Matrix Completion via Non-convex Stochastic Gradient Descent",
                        "abstract": "Matrix completion, where we wish to recover a low rank matrix by observing a few entries from it, is a widely studied problem in both theory and practice with wide applications. Most of the provable algorithms so far on this problem have been restricted to the offline setting where they provide an estimate of the unknown matrix using all observations simultaneously. However, in many applications, the online version, where we observe one entry at a time and dynamically update our estimate, is more appealing. While existing algorithms are efficient for the offline setting, they could be highly inefficient for the online setting.   In this paper, we propose the first provable, efficient online algorithm for matrix completion. Our algorithm starts from an initial estimate of the matrix and then performs non-convex stochastic gradient descent (SGD). After every observation, it performs a fast update involving only one row of two tall matrices, giving near linear total runtime. Our algorithm can be naturally used in the offline setting as well, where it gives competitive sample complexity and runtime to state of the art algorithms. Our proofs introduce a general framework to show that SGD updates tend to stay away from saddle surfaces and could be of broader interests for other non-convex problems to prove tight rates."
                    },
                    {
                        "title": "Towards minimax policies for online linear optimization with bandit feedback",
                        "abstract": "We address the online linear optimization problem with bandit feedback. Our contribution is twofold. First, we provide an algorithm (based on exponential weights) with a regret of order $\\sqrt{d n \\log N}$ for any finite action set with $N$ actions, under the assumption that the instantaneous loss is bounded by 1. This shaves off an extraneous $\\sqrt{d}$ factor compared to previous works, and gives a regret bound of order $d \\sqrt{n \\log n}$ for any compact set of actions. Without further assumptions on the action set, this last bound is minimax optimal up to a logarithmic factor. Interestingly, our result also shows that the minimax regret for bandit linear optimization with expert advice in $d$ dimension is the same as for the basic $d$-armed bandit with expert advice. Our second contribution is to show how to use the Mirror Descent algorithm to obtain computationally efficient strategies with minimax optimal regret bounds in specific examples. More precisely we study two canonical action sets: the hypercube and the Euclidean ball. In the former case, we obtain the first computationally efficient algorithm with a $d \\sqrt{n}$ regret, thus improving by a factor $\\sqrt{d \\log n}$ over the best known result for a computationally efficient algorithm. In the latter case, our approach gives the first algorithm with a $\\sqrt{d n \\log n}$ regret, again shaving off an extraneous $\\sqrt{d}$ compared to previous works."
                    },
                    {
                        "title": "Super-Resolution Off the Grid",
                        "abstract": "Super-resolution is the problem of recovering a superposition of point sources using bandlimited measurements, which may be corrupted with noise. This signal processing problem arises in numerous imaging problems, ranging from astronomy to biology to spectroscopy, where it is common to take (coarse) Fourier measurements of an object. Of particular interest is in obtaining estimation procedures which are robust to noise, with the following desirable statistical and computational properties: we seek to use coarse Fourier measurements (bounded by some cutoff frequency); we hope to take a (quantifiably) small number of measurements; we desire our algorithm to run quickly.   Suppose we have k point sources in d dimensions, where the points are separated by at least \\Delta from each other (in Euclidean distance). This work provides an algorithm with the following favorable guarantees: - The algorithm uses Fourier measurements, whose frequencies are bounded by O(1/\\Delta) (up to log factors). Previous algorithms require a cutoff frequency which may be as large as {\\Omega}( d/\\Delta). - The number of measurements taken by and the computational complexity of our algorithm are bounded by a polynomial in both the number of points k and the dimension d, with no dependence on the separation \\Delta. In contrast, previous algorithms depended inverse polynomially on the minimal separation and exponentially on the dimension for both of these quantities.   Our estimation procedure itself is simple: we take random bandlimited measurements (as opposed to taking an exponential number of measurements on the hyper-grid). Furthermore, our analysis and algorithm are elementary (based on concentration bounds for sampling and the singular value decomposition)."
                    },
                    {
                        "title": "Learning Exponential Families in High-Dimensions: Strong Convexity and Sparsity",
                        "abstract": "The versatility of exponential families, along with their attendant convexity properties, make them a popular and effective statistical model. A central issue is learning these models in high-dimensions, such as when there is some sparsity pattern of the optimal parameter. This work characterizes a certain strong convexity property of general exponential families, which allow their generalization ability to be quantified. In particular, we show how this property can be used to analyze generic exponential families under L_1 regularization."
                    },
                    {
                        "title": "Identifiability and Unmixing of Latent Parse Trees",
                        "abstract": "This paper explores unsupervised learning of parsing models along two directions. First, which models are identifiable from infinite data? We use a general technique for numerically checking identifiability based on the rank of a Jacobian matrix, and apply it to several standard constituency and dependency parsing models. Second, for identifiable models, how do we estimate the parameters efficiently? EM suffers from local optima, while recent work using spectral methods cannot be directly applied since the topology of the parse tree varies across sentences. We develop a strategy, unmixing, which deals with this additional complexity for restricted classes of parsing models."
                    },
                    {
                        "title": "An Exponential Lower Bound for Linearly-Realizable MDPs with Constant Suboptimality Gap",
                        "abstract": "A fundamental question in the theory of reinforcement learning is: suppose the optimal $Q$-function lies in the linear span of a given $d$ dimensional feature mapping, is sample-efficient reinforcement learning (RL) possible? The recent and remarkable result of Weisz et al. (2020) resolved this question in the negative, providing an exponential (in $d$) sample size lower bound, which holds even if the agent has access to a generative model of the environment. One may hope that this information theoretic barrier for RL can be circumvented by further supposing an even more favorable assumption: there exists a \\emph{constant suboptimality gap} between the optimal $Q$-value of the best action and that of the second-best action (for all states). The hope is that having a large suboptimality gap would permit easier identification of optimal actions themselves, thus making the problem tractable; indeed, provided the agent has access to a generative model, sample-efficient RL is in fact possible with the addition of this more favorable assumption.   This work focuses on this question in the standard online reinforcement learning setting, where our main result resolves this question in the negative: our hardness result shows that an exponential sample complexity lower bound still holds even if a constant suboptimality gap is assumed in addition to having a linearly realizable optimal $Q$-function. Perhaps surprisingly, this implies an exponential separation between the online RL setting and the generative model setting. Complementing our negative hardness result, we give two positive results showing that provably sample-efficient RL is possible either under an additional low-variance assumption or under a novel hypercontractivity assumption (both implicitly place stronger conditions on the underlying dynamics model)."
                    }
                ]
            },
            "5c8ce172-6677-4254-a3f7-94a26024b00e": {
                "pk": "5c8ce172-6677-4254-a3f7-94a26024b00e",
                "name": "Ari Holtzman",
                "collaborators": [
                    "Yejin Choi",
                    "Peter West",
                    "Luke Zettlemoyer",
                    "Maxwell Forbes",
                    "Rowan Zellers",
                    "Ali Farhadi",
                    "Jan Buys",
                    "Antoine Bosselut",
                    "Artidoro Pagnoni",
                    "Weijia Shi"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Generative Models",
                    "Machine Learning",
                    "Commonsense Reasoning"
                ],
                "publications": [
                    {
                        "title": "Generative Models as a Complex Systems Science: How can we make sense of large language model behavior?",
                        "abstract": "Coaxing out desired behavior from pretrained models, while avoiding undesirable ones, has redefined NLP and is reshaping how we interact with computers. What was once a scientific engineering discipline-in which building blocks are stacked one on top of the other-is arguably already a complex systems science, in which emergent behaviors are sought out to support previously unimagined use cases.   Despite the ever increasing number of benchmarks that measure task performance, we lack explanations of what behaviors language models exhibit that allow them to complete these tasks in the first place. We argue for a systematic effort to decompose language model behavior into categories that explain cross-task performance, to guide mechanistic explanations and help future-proof analytic research."
                    },
                    {
                        "title": "Artificial Intelligence and Aesthetic Judgment",
                        "abstract": "Generative AIs produce creative outputs in the style of human expression. We argue that encounters with the outputs of modern generative AI models are mediated by the same kinds of aesthetic judgments that organize our interactions with artwork. The interpretation procedure we use on art we find in museums is not an innate human faculty, but one developed over history by disciplines such as art history and art criticism to fulfill certain social functions. This gives us pause when considering our reactions to generative AI, how we should approach this new medium, and why generative AI seems to incite so much fear about the future. We naturally inherit a conundrum of causal inference from the history of art: a work can be read as a symptom of the cultural conditions that influenced its creation while simultaneously being framed as a timeless, seemingly acausal distillation of an eternal human condition. In this essay, we focus on an unresolved tension when we bring this dilemma to bear in the context of generative AI: are we looking for proof that generated media reflects something about the conditions that created it or some eternal human essence? Are current modes of interpretation sufficient for this task? Historically, new forms of art have changed how art is interpreted, with such influence used as evidence that a work of art has touched some essential human truth. As generative AI influences contemporary aesthetic judgment we outline some of the pitfalls and traps in attempting to scrutinize what AI generated media means."
                    },
                    {
                        "title": "The Curious Case of Neural Text Degeneration",
                        "abstract": "Despite considerable advancements with deep neural language models, the enigma of neural text degeneration persists when these models are tested as text generators. The counter-intuitive empirical observation is that even though the use of likelihood as training objective leads to high quality models for a broad range of language understanding tasks, using likelihood as a decoding objective leads to text that is bland and strangely repetitive.   In this paper, we reveal surprising distributional differences between human text and machine text. In addition, we find that decoding strategies alone can dramatically effect the quality of machine text, even when generated from exactly the same neural language model. Our findings motivate Nucleus Sampling, a simple but effective method to draw the best out of neural generation. By sampling text from the dynamic nucleus of the probability distribution, which allows for diversity while effectively truncating the less reliable tail of the distribution, the resulting text better demonstrates the quality of human text, yielding enhanced diversity without sacrificing fluency and coherence."
                    },
                    {
                        "title": "MultiTalk: A Highly-Branching Dialog Testbed for Diverse Conversations",
                        "abstract": "We study conversational dialog in which there are many possible responses to a given history. We present the MultiTalk Dataset, a corpus of over 320,000 sentences of written conversational dialog that balances a high branching factor (10) with several conversation turns (6) through selective branch continuation. We make multiple contributions to study dialog generation in the highly branching setting. In order to evaluate a diverse set of generations, we propose a simple scoring algorithm, based on bipartite graph matching, to optimally incorporate a set of diverse references. We study multiple language generation tasks at different levels of predictive conversation depth, using textual attributes induced automatically from pretrained classifiers. Our culminating task is a challenging theory of mind problem, a controllable generation task which requires reasoning about the expected reaction of the listener."
                    },
                    {
                        "title": "Do Neural Language Representations Learn Physical Commonsense?",
                        "abstract": "Humans understand language based on the rich background knowledge about how the physical world works, which in turn allows us to reason about the physical world through language. In addition to the properties of objects (e.g., boats require fuel) and their affordances, i.e., the actions that are applicable to them (e.g., boats can be driven), we can also reason about if-then inferences between what properties of objects imply the kind of actions that are applicable to them (e.g., that if we can drive something then it likely requires fuel).   In this paper, we investigate the extent to which state-of-the-art neural language representations, trained on a vast amount of natural language text, demonstrate physical commonsense reasoning. While recent advancements of neural language models have demonstrated strong performance on various types of natural language inference tasks, our study based on a dataset of over 200k newly collected annotations suggests that neural language representations still only learn associations that are explicitly written down."
                    },
                    {
                        "title": "BottleSum: Unsupervised and Self-supervised Sentence Summarization using the Information Bottleneck Principle",
                        "abstract": "The principle of the Information Bottleneck (Tishby et al. 1999) is to produce a summary of information X optimized to predict some other relevant information Y. In this paper, we propose a novel approach to unsupervised sentence summarization by mapping the Information Bottleneck principle to a conditional language modelling objective: given a sentence, our approach seeks a compressed sentence that can best predict the next sentence. Our iterative algorithm under the Information Bottleneck objective searches gradually shorter subsequences of the given sentence while maximizing the probability of the next sentence conditioned on the summary. Using only pretrained language models with no direct supervision, our approach can efficiently perform extractive sentence summarization over a large corpus.   Building on our unsupervised extractive summarization (BottleSumEx), we then present a new approach to self-supervised abstractive summarization (BottleSumSelf), where a transformer-based language model is trained on the output summaries of our unsupervised method. Empirical results demonstrate that our extractive method outperforms other unsupervised models on multiple automatic metrics. In addition, we find that our self-supervised abstractive model outperforms unsupervised baselines (including our own) by human evaluation along multiple attributes."
                    },
                    {
                        "title": "QLoRA: Efficient Finetuning of Quantized LLMs",
                        "abstract": "We present QLoRA, an efficient finetuning approach that reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU while preserving full 16-bit finetuning task performance. QLoRA backpropagates gradients through a frozen, 4-bit quantized pretrained language model into Low Rank Adapters~(LoRA). Our best model family, which we name Guanaco, outperforms all previous openly released models on the Vicuna benchmark, reaching 99.3% of the performance level of ChatGPT while only requiring 24 hours of finetuning on a single GPU. QLoRA introduces a number of innovations to save memory without sacrificing performance: (a) 4-bit NormalFloat (NF4), a new data type that is information theoretically optimal for normally distributed weights (b) double quantization to reduce the average memory footprint by quantizing the quantization constants, and (c) paged optimziers to manage memory spikes. We use QLoRA to finetune more than 1,000 models, providing a detailed analysis of instruction following and chatbot performance across 8 instruction datasets, multiple model types (LLaMA, T5), and model scales that would be infeasible to run with regular finetuning (e.g. 33B and 65B parameter models). Our results show that QLoRA finetuning on a small high-quality dataset leads to state-of-the-art results, even when using smaller models than the previous SoTA. We provide a detailed analysis of chatbot performance based on both human and GPT-4 evaluations showing that GPT-4 evaluations are a cheap and reasonable alternative to human evaluation. Furthermore, we find that current chatbot benchmarks are not trustworthy to accurately evaluate the performance levels of chatbots. A lemon-picked analysis demonstrates where Guanaco fails compared to ChatGPT. We release all of our models and code, including CUDA kernels for 4-bit training."
                    },
                    {
                        "title": "Surface Form Competition: Why the Highest Probability Answer Isn't Always Right",
                        "abstract": "Large language models have shown promising results in zero-shot settings (Brown et al.,2020; Radford et al., 2019). For example, they can perform multiple choice tasks simply by conditioning on a question and selecting the answer with the highest probability.   However, ranking by string probability can be problematic due to surface form competition-wherein different surface forms compete for probability mass, even if they represent the same underlying concept, e.g. \"computer\" and \"PC.\" Since probability mass is finite, this lowers the probability of the correct answer, due to competition from other strings that are valid answers (but not one of the multiple choice options).   We introduce Domain Conditional Pointwise Mutual Information, an alternative scoring function that directly compensates for surface form competition by simply reweighing each option according to a term that is proportional to its a priori likelihood within the context of the specific zero-shot task. It achieves consistent gains in zero-shot performance over both calibrated (Zhao et al., 2021) and uncalibrated scoring functions on all GPT-2 and GPT-3 models over a variety of multiple choice datasets."
                    },
                    {
                        "title": "Learning to Write with Cooperative Discriminators",
                        "abstract": "Recurrent Neural Networks (RNNs) are powerful autoregressive sequence models, but when used to generate natural language their output tends to be overly generic, repetitive, and self-contradictory. We postulate that the objective function optimized by RNN language models, which amounts to the overall perplexity of a text, is not expressive enough to capture the notion of communicative goals described by linguistic principles such as Grice's Maxims. We propose learning a mixture of multiple discriminative models that can be used to complement the RNN generator and guide the decoding process. Human evaluation demonstrates that text generated by our system is preferred over that of baselines by a large margin and significantly enhances the overall coherence, style, and information content of the generated text."
                    },
                    {
                        "title": "Predicting vs. Acting: A Trade-off Between World Modeling & Agent Modeling",
                        "abstract": "RLHF-aligned LMs have shown unprecedented ability on both benchmarks and long-form text generation, yet they struggle with one foundational task: next-token prediction. As RLHF models become agent models aimed at interacting with humans, they seem to lose their world modeling -- the ability to predict what comes next in arbitrary documents, which is the foundational training objective of the Base LMs that RLHF adapts.   Besides empirically demonstrating this trade-off, we propose a potential explanation: to perform coherent long-form generation, RLHF models restrict randomness via implicit blueprints. In particular, RLHF models concentrate probability on sets of anchor spans that co-occur across multiple generations for the same prompt, serving as textual scaffolding but also limiting a model's ability to generate documents that do not include these spans. We study this trade-off on the most effective current agent models, those aligned with RLHF, while exploring why this may remain a fundamental trade-off between models that act and those that predict, even as alignment techniques improve."
                    },
                    {
                        "title": "HellaSwag: Can a Machine Really Finish Your Sentence?",
                        "abstract": "Recent work by Zellers et al. (2018) introduced a new task of commonsense natural language inference: given an event description such as \"A woman sits at a piano,\" a machine must select the most likely followup: \"She sets her fingers on the keys.\" With the introduction of BERT, near human-level performance was reached. Does this mean that machines can perform human level commonsense inference?   In this paper, we show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans (>95% accuracy), state-of-the-art models struggle (<48%). We achieve this via Adversarial Filtering (AF), a data collection paradigm wherein a series of discriminators iteratively select an adversarial set of machine-generated wrong answers. AF proves to be surprisingly robust. The key insight is to scale up the length and complexity of the dataset examples towards a critical 'Goldilocks' zone wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models.   Our construction of HellaSwag, and its resulting difficulty, sheds light on the inner workings of deep pretrained models. More broadly, it suggests a new path forward for NLP research, in which benchmarks co-evolve with the evolving state-of-the-art in an adversarial way, so as to present ever-harder challenges."
                    },
                    {
                        "title": "Simulating Action Dynamics with Neural Process Networks",
                        "abstract": "Understanding procedural language requires anticipating the causal effects of actions, even when they are not explicitly stated. In this work, we introduce Neural Process Networks to understand procedural text through (neural) simulation of action dynamics. Our model complements existing memory architectures with dynamic entity tracking by explicitly modeling actions as state transformers. The model updates the states of the entities by executing learned action operators. Empirical results demonstrate that our proposed model can reason about the unstated causal effects of actions, allowing it to provide more accurate contextual information for understanding and generating procedural text, all while offering more interpretable internal representations than existing alternatives."
                    },
                    {
                        "title": "Counterfactual Story Reasoning and Generation",
                        "abstract": "Counterfactual reasoning requires predicting how alternative events, contrary to what actually happened, might have resulted in different outcomes. Despite being considered a necessary component of AI-complete systems, few resources have been developed for evaluating counterfactual reasoning in narratives.   In this paper, we propose Counterfactual Story Rewriting: given an original story and an intervening counterfactual event, the task is to minimally revise the story to make it compatible with the given counterfactual event. Solving this task will require deep understanding of causal narrative chains and counterfactual invariance, and integration of such story reasoning capabilities into conditional language generation models.   We present TimeTravel, a new dataset of 29,849 counterfactual rewritings, each with the original story, a counterfactual event, and human-generated revision of the original story compatible with the counterfactual event. Additionally, we include 80,115 counterfactual \"branches\" without a rewritten storyline to support future work on semi- or un-supervised approaches to counterfactual story rewriting.   Finally, we evaluate the counterfactual rewriting capacities of several competitive baselines based on pretrained language models, and assess whether common overlap and model-based automatic metrics for text generation correlate well with human scores for counterfactual rewriting."
                    },
                    {
                        "title": "TuringAdvice: A Generative and Dynamic Evaluation of Language Use",
                        "abstract": "We propose TuringAdvice, a new challenge task and dataset for language understanding models. Given a written situation that a real person is currently facing, a model must generate helpful advice in natural language. Our evaluation framework tests a fundamental aspect of human language understanding: our ability to use language to resolve open-ended situations by communicating with each other.   Empirical results show that today's models struggle at TuringAdvice, even multibillion parameter models finetuned on 600k in-domain training examples. The best model, a finetuned T5, writes advice that is at least as helpful as human-written advice in only 14% of cases; a much larger non-finetunable GPT3 model does even worse at 4%. This low performance reveals language understanding errors that are hard to spot outside of a generative setting, showing much room for progress."
                    },
                    {
                        "title": "Reflective Decoding: Beyond Unidirectional Generation with Off-the-Shelf Language Models",
                        "abstract": "Publicly available, large pretrained LanguageModels (LMs) generate text with remarkable quality, but only sequentially from left to right. As a result, they are not immediately applicable to generation tasks that break the unidirectional assumption, such as paraphrasing or text-infilling, necessitating task-specific supervision.   In this paper, we present Reflective Decoding, a novel unsupervised algorithm that allows for direct application of unidirectional LMs to non-sequential tasks. Our 2-step approach requires no supervision or even parallel corpora, only two off-the-shelf pretrained LMs in opposite directions: forward and backward. First, in the contextualization step, we use LMs to generate ensembles of past and future contexts which collectively capture the input (e.g. the source sentence for paraphrasing). Second, in the reflection step, we condition on these \"context ensembles\", generating outputs that are compatible with them. Comprehensive empirical results demonstrate that Reflective Decoding outperforms strong unsupervised baselines on both paraphrasing and abductive text infilling, significantly narrowing the gap between unsupervised and supervised methods. Reflective Decoding surpasses multiple supervised baselines on various metrics including human evaluation."
                    },
                    {
                        "title": "PIGLeT: Language Grounding Through Neuro-Symbolic Interaction in a 3D World",
                        "abstract": "We propose PIGLeT: a model that learns physical commonsense knowledge through interaction, and then uses this knowledge to ground language. We factorize PIGLeT into a physical dynamics model, and a separate language model. Our dynamics model learns not just what objects are but also what they do: glass cups break when thrown, plastic ones don't. We then use it as the interface to our language model, giving us a unified model of linguistic form and grounded meaning. PIGLeT can read a sentence, simulate neurally what might happen next, and then communicate that result through a literal symbolic representation, or natural language.   Experimental results show that our model effectively learns world dynamics, along with how to communicate them. It is able to correctly forecast \"what happens next\" given an English sentence over 80% of the time, outperforming a 100x larger, text-to-text approach by over 10%. Likewise, its natural language summaries of physical interactions are also judged by humans as more accurate than LM alternatives. We present comprehensive analysis showing room for future work."
                    },
                    {
                        "title": "DEMix Layers: Disentangling Domains for Modular Language Modeling",
                        "abstract": "We introduce a new domain expert mixture (DEMix) layer that enables conditioning a language model (LM) on the domain of the input text. A DEMix layer is a collection of expert feedforward networks, each specialized to a domain, that makes the LM modular: experts can be mixed, added or removed after initial training. Extensive experiments with autoregressive transformer LMs (up to 1.3B parameters) show that DEMix layers reduce test-time perplexity, increase training efficiency, and enable rapid adaptation with little overhead. We show that mixing experts during inference, using a parameter-free weighted ensemble, allows the model to better generalize to heterogeneous or unseen domains. We also show that experts can be added to iteratively incorporate new domains without forgetting older ones, and that experts can be removed to restrict access to unwanted domains, without additional training. Overall, these results demonstrate benefits of explicitly conditioning on textual domains during language modeling."
                    },
                    {
                        "title": "Toward Human Readable Prompt Tuning: Kubrick's The Shining is a good movie, and a good prompt too?",
                        "abstract": "Large language models can perform new tasks in a zero-shot fashion, given natural language prompts that specify the desired behavior. Such prompts are typically hand engineered, but can also be learned with gradient-based methods from labeled data. However, it is underexplored what factors make the prompts effective, especially when the prompts are natural language. In this paper, we investigate common attributes shared by effective prompts. We first propose a human readable prompt tuning method (F LUENT P ROMPT) based on Langevin dynamics that incorporates a fluency constraint to find a diverse distribution of effective and fluent prompts. Our analysis reveals that effective prompts are topically related to the task domain and calibrate the prior probability of label words. Based on these findings, we also propose a method for generating prompts using only unlabeled data, outperforming strong baselines by an average of 7.0% accuracy across three tasks."
                    },
                    {
                        "title": "Benchmarks as Microscopes: A Call for Model Metrology",
                        "abstract": "Modern language models (LMs) pose a new challenge in capability assessment. Static benchmarks inevitably saturate without providing confidence in the deployment tolerances of LM-based systems, but developers nonetheless claim that their models have generalized traits such as reasoning or open-domain language understanding based on these flawed metrics. The science and practice of LMs requires a new approach to benchmarking which measures specific capabilities with dynamic assessments. To be confident in our metrics, we need a new discipline of model metrology -- one which focuses on how to generate benchmarks that predict performance under deployment. Motivated by our evaluation criteria, we outline how building a community of model metrology practitioners -- one focused on building tools and studying how to measure system capabilities -- is the best way to meet these needs to and add clarity to the AI discussion."
                    },
                    {
                        "title": "Defending Against Neural Fake News",
                        "abstract": "Recent progress in natural language generation has raised dual-use concerns. While applications like summarization and translation are positive, the underlying technology also might enable adversaries to generate neural fake news: targeted propaganda that closely mimics the style of real news.   Modern computer security relies on careful threat modeling: identifying potential threats and vulnerabilities from an adversary's point of view, and exploring potential mitigations to these threats. Likewise, developing robust defenses against neural fake news requires us first to carefully investigate and characterize the risks of these models. We thus present a model for controllable text generation called Grover. Given a headline like `Link Found Between Vaccines and Autism,' Grover can generate the rest of the article; humans find these generations to be more trustworthy than human-written disinformation.   Developing robust verification techniques against generators like Grover is critical. We find that best current discriminators can classify neural fake news from real, human-written, news with 73% accuracy, assuming access to a moderate level of training data. Counterintuitively, the best defense against Grover turns out to be Grover itself, with 92% accuracy, demonstrating the importance of public release of strong generators. We investigate these results further, showing that exposure bias -- and sampling strategies that alleviate its effects -- both leave artifacts that similar discriminators can pick up on. We conclude by discussing ethical issues regarding the technology, and plan to release Grover publicly, helping pave the way for better detection of neural fake news."
                    }
                ]
            },
            "36ab6b9f-2d83-4ec2-85ee-4d081bfe4708": {
                "pk": "36ab6b9f-2d83-4ec2-85ee-4d081bfe4708",
                "name": "Ranjay Krishna",
                "collaborators": [
                    "Li Fei-Fei",
                    "Michael Bernstein",
                    "Jieyu Zhang",
                    "Madeleine Grunde-McLaughlin",
                    "Maneesh Agrawala",
                    "Zixian Ma",
                    "Mitchell Gordon",
                    "Cewu Lu",
                    "Ines Chami",
                    "Jonathan Krause"
                ],
                "domain": [
                    "Computer Vision",
                    "Multi-modal Learning",
                    "Active Learning",
                    "Human-Computer Interaction"
                ],
                "publications": [
                    {
                        "title": "Visual Intelligence through Human Interaction",
                        "abstract": "Over the last decade, Computer Vision, the branch of Artificial Intelligence aimed at understanding the visual world, has evolved from simply recognizing objects in images to describing pictures, answering questions about images, aiding robots maneuver around physical spaces and even generating novel visual content. As these tasks and applications have modernized, so too has the reliance on more data, either for model training or for evaluation. In this chapter, we demonstrate that novel interaction strategies can enable new forms of data collection and evaluation for Computer Vision. First, we present a crowdsourcing interface for speeding up paid data collection by an order of magnitude, feeding the data-hungry nature of modern vision models. Second, we explore a method to increase volunteer contributions using automated social interventions. Third, we develop a system to ensure human evaluation of generative vision models are reliable, affordable and grounded in psychophysics theory. We conclude with future opportunities for Human-Computer Interaction to aid Computer Vision."
                    },
                    {
                        "title": "Visual Relationship Detection with Language Priors",
                        "abstract": "Visual relationships capture a wide variety of interactions between pairs of objects in images (e.g. \"man riding bicycle\" and \"man pushing bicycle\"). Consequently, the set of possible relationships is extremely large and it is difficult to obtain sufficient training examples for all possible relationships. Because of this limitation, previous work on visual relationship detection has concentrated on predicting only a handful of relationships. Though most relationships are infrequent, their objects (e.g. \"man\" and \"bicycle\") and predicates (e.g. \"riding\" and \"pushing\") independently occur more frequently. We propose a model that uses this insight to train visual models for objects and predicates individually and later combines them together to predict multiple relationships per image. We improve on prior work by leveraging language priors from semantic word embeddings to finetune the likelihood of a predicted relationship. Our model can scale to predict thousands of types of relationships from a few examples. Additionally, we localize the objects in the predicted relationships as bounding boxes in the image. We further demonstrate that understanding relationships can improve content based image retrieval."
                    },
                    {
                        "title": "Referring Relationships",
                        "abstract": "Images are not simply sets of objects: each image represents a web of interconnected relationships. These relationships between entities carry semantic meaning and help a viewer differentiate between instances of an entity. For example, in an image of a soccer match, there may be multiple persons present, but each participates in different relationships: one is kicking the ball, and the other is guarding the goal. In this paper, we formulate the task of utilizing these \"referring relationships\" to disambiguate between entities of the same category. We introduce an iterative model that localizes the two entities in the referring relationship, conditioned on one another. We formulate the cyclic condition between the entities in a relationship by modelling predicates that connect the entities as shifts in attention from one entity to another. We demonstrate that our model can not only outperform existing approaches on three datasets --- CLEVR, VRD and Visual Genome --- but also that it produces visually meaningful predicate shifts, as an instance of interpretable neural networks. Finally, we show that by modelling predicates as attention shifts, we can even localize entities in the absence of their category, allowing our model to find completely unseen categories."
                    },
                    {
                        "title": "AGQA 2.0: An Updated Benchmark for Compositional Spatio-Temporal Reasoning",
                        "abstract": "Prior benchmarks have analyzed models' answers to questions about videos in order to measure visual compositional reasoning. Action Genome Question Answering (AGQA) is one such benchmark. AGQA provides a training/test split with balanced answer distributions to reduce the effect of linguistic biases. However, some biases remain in several AGQA categories. We introduce AGQA 2.0, a version of this benchmark with several improvements, most namely a stricter balancing procedure. We then report results on the updated benchmark for all experiments."
                    },
                    {
                        "title": "Information Maximizing Visual Question Generation",
                        "abstract": "Though image-to-sequence generation models have become overwhelmingly popular in human-computer communications, they suffer from strongly favoring safe generic questions (\"What is in this picture?\"). Generating uninformative but relevant questions is not sufficient or useful. We argue that a good question is one that has a tightly focused purpose --- one that is aimed at expecting a specific type of response. We build a model that maximizes mutual information between the image, the expected answer and the generated question. To overcome the non-differentiability of discrete natural language tokens, we introduce a variational continuous latent space onto which the expected answers project. We regularize this latent space with a second latent space that ensures clustering of similar answers. Even when we don't know the expected answer, this second latent space can generate goal-driven questions specifically aimed at extracting objects (\"what is the person throwing\"), attributes, (\"What kind of shirt is the person wearing?\"), color (\"what color is the frisbee?\"), material (\"What material is the frisbee?\"), etc. We quantitatively show that our model is able to retain information about an expected answer category, resulting in more diverse, goal-driven questions. We launch our model on a set of real world images and extract previously unseen visual concepts."
                    },
                    {
                        "title": "A Hierarchical Approach for Generating Descriptive Image Paragraphs",
                        "abstract": "Recent progress on image captioning has made it possible to generate novel sentences describing images in natural language, but compressing an image into a single sentence can describe visual content in only coarse detail. While one new captioning approach, dense captioning, can potentially describe images in finer levels of detail by captioning many regions within an image, it in turn is unable to produce a coherent story for an image. In this paper we overcome these limitations by generating entire paragraphs for describing images, which can tell detailed, unified stories. We develop a model that decomposes both images and paragraphs into their constituent parts, detecting semantic regions in images and using a hierarchical recurrent neural network to reason about language. Linguistic analysis confirms the complexity of the paragraph generation task, and thorough experiments on a new dataset of image and paragraph pairs demonstrate the effectiveness of our approach."
                    },
                    {
                        "title": "Determining Question-Answer Plausibility in Crowdsourced Datasets Using Multi-Task Learning",
                        "abstract": "Datasets extracted from social networks and online forums are often prone to the pitfalls of natural language, namely the presence of unstructured and noisy data. In this work, we seek to enable the collection of high-quality question-answer datasets from social media by proposing a novel task for automated quality analysis and data cleaning: question-answer (QA) plausibility. Given a machine or user-generated question and a crowd-sourced response from a social media user, we determine if the question and response are valid; if so, we identify the answer within the free-form response. We design BERT-based models to perform the QA plausibility task, and we evaluate the ability of our models to generate a clean, usable question-answer dataset. Our highest-performing approach consists of a single-task model which determines the plausibility of the question, followed by a multi-task model which evaluates the plausibility of the response as well as extracts answers (Question Plausibility AUROC=0.75, Response Plausibility AUROC=0.78, Answer Extraction F1=0.665)."
                    },
                    {
                        "title": "Videoshop: Localized Semantic Video Editing with Noise-Extrapolated Diffusion Inversion",
                        "abstract": "We introduce Videoshop, a training-free video editing algorithm for localized semantic edits. Videoshop allows users to use any editing software, including Photoshop and generative inpainting, to modify the first frame; it automatically propagates those changes, with semantic, spatial, and temporally consistent motion, to the remaining frames. Unlike existing methods that enable edits only through imprecise textual instructions, Videoshop allows users to add or remove objects, semantically change objects, insert stock photos into videos, etc. with fine-grained control over locations and appearance. We achieve this through image-based video editing by inverting latents with noise extrapolation, from which we generate videos conditioned on the edited image. Videoshop produces higher quality edits against 6 baselines on 2 editing benchmarks using 10 evaluation metrics."
                    },
                    {
                        "title": "AGQA: A Benchmark for Compositional Spatio-Temporal Reasoning",
                        "abstract": "Visual events are a composition of temporal actions involving actors spatially interacting with objects. When developing computer vision models that can reason about compositional spatio-temporal events, we need benchmarks that can analyze progress and uncover shortcomings. Existing video question answering benchmarks are useful, but they often conflate multiple sources of error into one accuracy metric and have strong biases that models can exploit, making it difficult to pinpoint model weaknesses. We present Action Genome Question Answering (AGQA), a new benchmark for compositional spatio-temporal reasoning. AGQA contains $192M$ unbalanced question answer pairs for $9.6K$ videos. We also provide a balanced subset of $3.9M$ question answer pairs, $3$ orders of magnitude larger than existing benchmarks, that minimizes bias by balancing the answer distributions and types of question structures. Although human evaluators marked $86.02\\%$ of our question-answer pairs as correct, the best model achieves only $47.74\\%$ accuracy. In addition, AGQA introduces multiple training/test splits to test for various reasoning abilities, including generalization to novel compositions, to indirect references, and to more compositional steps. Using AGQA, we evaluate modern visual reasoning systems, demonstrating that the best models barely perform better than non-visual baselines exploiting linguistic biases and that none of the existing models generalize to novel compositions unseen during training."
                    },
                    {
                        "title": "Iterated Learning Improves Compositionality in Large Vision-Language Models",
                        "abstract": "A fundamental characteristic common to both human vision and natural language is their compositional nature. Yet, despite the performance gains contributed by large vision and language pretraining, recent investigations find that most-if not all-our state-of-the-art vision-language models struggle at compositionality. They are unable to distinguish between images of \" a girl in white facing a man in black\" and \"a girl in black facing a man in white\". Moreover, prior work suggests that compositionality doesn't arise with scale: larger model sizes or training data don't help. This paper develops a new iterated training algorithm that incentivizes compositionality. We draw on decades of cognitive science research that identifies cultural transmission-the need to teach a new generation-as a necessary inductive prior that incentivizes humans to develop compositional languages. Specifically, we reframe vision-language contrastive learning as the Lewis Signaling Game between a vision agent and a language agent, and operationalize cultural transmission by iteratively resetting one of the agent's weights during training. After every iteration, this training paradigm induces representations that become \"easier to learn\", a property of compositional languages: e.g. our model trained on CC3M and CC12M improves standard CLIP by 4.7%, 4.0% respectfully in the SugarCrepe benchmark."
                    },
                    {
                        "title": "Conceptual Metaphors Impact Perceptions of Human-AI Collaboration",
                        "abstract": "With the emergence of conversational artificial intelligence (AI) agents, it is important to understand the mechanisms that influence users' experiences of these agents. We study a common tool in the designer's toolkit: conceptual metaphors. Metaphors can present an agent as akin to a wry teenager, a toddler, or an experienced butler. How might a choice of metaphor influence our experience of the AI agent? Sampling metaphors along the dimensions of warmth and competence---defined by psychological theories as the primary axes of variation for human social perception---we perform a study (N=260) where we manipulate the metaphor, but not the behavior, of a Wizard-of-Oz conversational agent. Following the experience, participants are surveyed about their intention to use the agent, their desire to cooperate with the agent, and the agent's usability. Contrary to the current tendency of designers to use high competence metaphors to describe AI products, we find that metaphors that signal low competence lead to better evaluations of the agent than metaphors that signal high competence. This effect persists despite both high and low competence agents featuring human-level performance and the wizards being blind to condition. A second study confirms that intention to adopt decreases rapidly as competence projected by the metaphor increases. In a third study, we assess effects of metaphor choices on potential users' desire to try out the system and find that users are drawn to systems that project higher competence and warmth. These results suggest that projecting competence may help attract new users, but those users may discard the agent unless it can quickly correct with a lower competence metaphor. We close with a retrospective analysis that finds similar patterns between metaphors and user attitudes towards past conversational agents such as Xiaoice, Replika, Woebot, Mitsuku, and Tay."
                    },
                    {
                        "title": "A Glimpse Far into the Future: Understanding Long-term Crowd Worker Quality",
                        "abstract": "Microtask crowdsourcing is increasingly critical to the creation of extremely large datasets. As a result, crowd workers spend weeks or months repeating the exact same tasks, making it necessary to understand their behavior over these long periods of time. We utilize three large, longitudinal datasets of nine million annotations collected from Amazon Mechanical Turk to examine claims that workers fatigue or satisfice over these long periods, producing lower quality work. We find that, contrary to these claims, workers are extremely stable in their quality over the entire period. To understand whether workers set their quality based on the task's requirements for acceptance, we then perform an experiment where we vary the required quality for a large crowdsourcing task. Workers did not adjust their quality based on the acceptance threshold: workers who were above the threshold continued working at their usual quality level, and workers below the threshold self-selected themselves out of the task. Capitalizing on this consistency, we demonstrate that it is possible to predict workers' long-term quality using just a glimpse of their quality on the first five tasks."
                    },
                    {
                        "title": "Action Genome: Actions as Composition of Spatio-temporal Scene Graphs",
                        "abstract": "Action recognition has typically treated actions and activities as monolithic events that occur in videos. However, there is evidence from Cognitive Science and Neuroscience that people actively encode activities into consistent hierarchical part structures. However in Computer Vision, few explorations on representations encoding event partonomies have been made. Inspired by evidence that the prototypical unit of an event is an action-object interaction, we introduce Action Genome, a representation that decomposes actions into spatio-temporal scene graphs. Action Genome captures changes between objects and their pairwise relationships while an action occurs. It contains 10K videos with 0.4M objects and 1.7M visual relationships annotated. With Action Genome, we extend an existing action recognition model by incorporating scene graphs as spatio-temporal feature banks to achieve better performance on the Charades dataset. Next, by decomposing and learning the temporal changes in visual relationships that result in an action, we demonstrate the utility of a hierarchical event decomposition by enabling few-shot action recognition, achieving 42.7% mAP using as few as 10 examples. Finally, we benchmark existing scene graph models on the new task of spatio-temporal scene graph prediction."
                    },
                    {
                        "title": "Mind Your Outliers! Investigating the Negative Impact of Outliers on Active Learning for Visual Question Answering",
                        "abstract": "Active learning promises to alleviate the massive data needs of supervised machine learning: it has successfully improved sample efficiency by an order of magnitude on traditional tasks like topic classification and object recognition. However, we uncover a striking contrast to this promise: across 5 models and 4 datasets on the task of visual question answering, a wide variety of active learning approaches fail to outperform random selection. To understand this discrepancy, we profile 8 active learning methods on a per-example basis, and identify the problem as collective outliers -- groups of examples that active learning methods prefer to acquire but models fail to learn (e.g., questions that ask about text in images or require external knowledge). Through systematic ablation experiments and qualitative visualizations, we verify that collective outliers are a general phenomenon responsible for degrading pool-based active learning. Notably, we show that active learning sample efficiency increases significantly as the number of collective outliers in the active learning pool decreases. We conclude with a discussion and prescriptive recommendations for mitigating the effects of these outliers in future work."
                    },
                    {
                        "title": "EcoAssistant: Using LLM Assistant More Affordably and Accurately",
                        "abstract": "Today, users ask Large language models (LLMs) as assistants to answer queries that require external knowledge; they ask about the weather in a specific city, about stock prices, and even about where specific locations are within their neighborhood. These queries require the LLM to produce code that invokes external APIs to answer the user's question, yet LLMs rarely produce correct code on the first try, requiring iterative code refinement upon execution results. In addition, using LLM assistants to support high query volumes can be expensive. In this work, we contribute a framework, EcoAssistant, that enables LLMs to answer code-driven queries more affordably and accurately. EcoAssistant contains three components. First, it allows the LLM assistants to converse with an automatic code executor to iteratively refine code or to produce answers based on the execution results. Second, we use a hierarchy of LLM assistants, which attempts to answer the query with weaker, cheaper LLMs before backing off to stronger, expensive ones. Third, we retrieve solutions from past successful queries as in-context demonstrations to help subsequent queries. Empirically, we show that EcoAssistant offers distinct advantages for affordability and accuracy, surpassing GPT-4 by 10 points of success rate with less than 50% of GPT-4's cost."
                    },
                    {
                        "title": "Self-Enhancing Video Data Management System for Compositional Events with Large Language Models [Technical Report]",
                        "abstract": "Complex video queries can be answered by decomposing them into modular subtasks. However, existing video data management systems assume the existence of predefined modules for each subtask. We introduce VOCAL-UDF, a novel self-enhancing system that supports compositional queries over videos without the need for predefined modules. VOCAL-UDF automatically identifies and constructs missing modules and encapsulates them as user-defined functions (UDFs), thus expanding its querying capabilities. To achieve this, we formulate a unified UDF model that leverages large language models (LLMs) to aid in new UDF generation. VOCAL-UDF handles a wide range of concepts by supporting both program-based UDFs (i.e., Python functions generated by LLMs) and distilled-model UDFs (lightweight vision models distilled from strong pretrained models). To resolve the inherent ambiguity in user intent, VOCAL-UDF generates multiple candidate UDFs and uses active learning to efficiently select the best one. With the self-enhancing capability, VOCAL-UDF significantly improves query performance across three video datasets."
                    },
                    {
                        "title": "Deep Bayesian Active Learning for Multiple Correct Outputs",
                        "abstract": "Typical active learning strategies are designed for tasks, such as classification, with the assumption that the output space is mutually exclusive. The assumption that these tasks always have exactly one correct answer has resulted in the creation of numerous uncertainty-based measurements, such as entropy and least confidence, which operate over a model's outputs. Unfortunately, many real-world vision tasks, like visual question answering and image captioning, have multiple correct answers, causing these measurements to overestimate uncertainty and sometimes perform worse than a random sampling baseline. In this paper, we propose a new paradigm that estimates uncertainty in the model's internal hidden space instead of the model's output space. We specifically study a manifestation of this problem for visual question answer generation (VQA), where the aim is not to classify the correct answer but to produce a natural language answer, given an image and a question. Our method overcomes the paraphrastic nature of language. It requires a semantic space that structures the model's output concepts and that enables the usage of techniques like dropout-based Bayesian uncertainty. We build a visual-semantic space that embeds paraphrases close together for any existing VQA model. We empirically show state-of-art active learning results on the task of VQA on two datasets, being 5 times more cost-efficient on Visual Genome and 3 times more cost-efficient on VQA 2.0."
                    },
                    {
                        "title": "Learning Predicates as Functions to Enable Few-shot Scene Graph Prediction",
                        "abstract": "Scene graph prediction --- classifying the set of objects and predicates in a visual scene --- requires substantial training data. However, most predicates only occur a handful of times making them difficult to learn. We introduce the first scene graph prediction model that supports few-shot learning of predicates. Existing scene graph generation models represent objects using pretrained object detectors or word embeddings that capture semantic object information at the cost of encoding information about which relationships they afford. So, these object representations are unable to generalize to new few-shot relationships. We introduce a framework that induces object representations that are structured according to their visual relationships. Unlike past methods, our framework embeds objects that afford similar relationships closer together. This property allows our model to perform well in the few-shot setting. For example, applying the 'riding' predicate transformation to 'person' modifies the representation towards objects like 'skateboard' and 'horse' that enable riding. We generate object representations by learning predicates trained as message passing functions within a new graph convolution framework. The object representations are used to build few-shot predicate classifiers for rare predicates with as few as 1 labeled example. We achieve a 5-shot performance of 22.70 recall@50, a 3.7 increase when compared to strong transfer learning baselines."
                    },
                    {
                        "title": "ELIGN: Expectation Alignment as a Multi-Agent Intrinsic Reward",
                        "abstract": "Modern multi-agent reinforcement learning frameworks rely on centralized training and reward shaping to perform well. However, centralized training and dense rewards are not readily available in the real world. Current multi-agent algorithms struggle to learn in the alternative setup of decentralized training or sparse rewards. To address these issues, we propose a self-supervised intrinsic reward ELIGN - expectation alignment - inspired by the self-organization principle in Zoology. Similar to how animals collaborate in a decentralized manner with those in their vicinity, agents trained with expectation alignment learn behaviors that match their neighbors' expectations. This allows the agents to learn collaborative behaviors without any external reward or centralized training. We demonstrate the efficacy of our approach across 6 tasks in the multi-agent particle and the complex Google Research football environments, comparing ELIGN to sparse and curiosity-based intrinsic rewards. When the number of agents increases, ELIGN scales well in all multi-agent tasks except for one where agents have different capabilities. We show that agent coordination improves through expectation alignment because agents learn to divide tasks amongst themselves, break coordination symmetries, and confuse adversaries. These results identify tasks where expectation alignment is a more useful strategy than curiosity-driven exploration for multi-agent coordination, enabling agents to do zero-shot coordination."
                    },
                    {
                        "title": "m&m's: A Benchmark to Evaluate Tool-Use for multi-step multi-modal Tasks",
                        "abstract": "Real-world multi-modal problems are rarely solved by a single machine learning model, and often require multi-step computational plans that involve stitching several models. Tool-augmented LLMs hold tremendous promise for automating the generation of such computational plans. However, the lack of standardized benchmarks for evaluating LLMs as planners for multi-step multi-modal tasks has prevented a systematic study of planner design decisions. Should LLMs generate a full plan in a single shot or step-by-step? Should they invoke tools directly with Python code or through structured data formats like JSON? Does feedback improve planning? To answer these questions and more, we introduce m&m's: a benchmark containing 4K+ multi-step multi-modal tasks involving 33 tools that include multi-modal models, (free) public APIs, and image processing modules. For each of these task queries, we provide automatically generated plans using this realistic toolset. We further provide a high-quality subset of 1,565 task plans that are human-verified and correctly executable. With m&m's, we evaluate 10 popular LLMs with 2 planning strategies (multi-step vs. step-by-step planning), 2 plan formats (JSON vs. code), and 3 types of feedback (parsing/verification/execution). Finally, we summarize takeaways from our extensive experiments. Our dataset and code are available on HuggingFace (https://huggingface.co/datasets/zixianma/mnms) and Github (https://github.com/RAIVNLab/mnms)."
                    }
                ]
            },
            "b43b33a4-1b05-46e3-bb44-1fe260f47473": {
                "pk": "b43b33a4-1b05-46e3-bb44-1fe260f47473",
                "name": "Ali Farhadi",
                "collaborators": [
                    "Hannaneh Hajishirzi",
                    "Mohammad Rastegari",
                    "Hessam Bagherinezhad",
                    "Joseph Redmon",
                    "Roozbeh Mottaghi",
                    "Minjoon Seo",
                    "Babak Saleh",
                    "Ahmed Elgammal",
                    "Junyuan Xie",
                    "Ross Girshick"
                ],
                "domain": [
                    "Computer Vision",
                    "Object Detection",
                    "Deep Learning",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "YOLOv3: An Incremental Improvement",
                        "abstract": "We present some updates to YOLO! We made a bunch of little design changes to make it better. We also trained this new network that's pretty swell. It's a little bigger than last time but more accurate. It's still fast though, don't worry. At 320x320 YOLOv3 runs in 22 ms at 28.2 mAP, as accurate as SSD but three times faster. When we look at the old .5 IOU mAP detection metric YOLOv3 is quite good. It achieves 57.9 mAP@50 in 51 ms on a Titan X, compared to 57.5 mAP@50 in 198 ms by RetinaNet, similar performance but 3.8x faster. As always, all the code is online at https://pjreddie.com/yolo/"
                    },
                    {
                        "title": "YOLO9000: Better, Faster, Stronger",
                        "abstract": "We introduce YOLO9000, a state-of-the-art, real-time object detection system that can detect over 9000 object categories. First we propose various improvements to the YOLO detection method, both novel and drawn from prior work. The improved model, YOLOv2, is state-of-the-art on standard detection tasks like PASCAL VOC and COCO. At 67 FPS, YOLOv2 gets 76.8 mAP on VOC 2007. At 40 FPS, YOLOv2 gets 78.6 mAP, outperforming state-of-the-art methods like Faster RCNN with ResNet and SSD while still running significantly faster. Finally we propose a method to jointly train on object detection and classification. Using this method we train YOLO9000 simultaneously on the COCO detection dataset and the ImageNet classification dataset. Our joint training allows YOLO9000 to predict detections for object classes that don't have labelled detection data. We validate our approach on the ImageNet detection task. YOLO9000 gets 19.7 mAP on the ImageNet detection validation set despite only having detection data for 44 of the 200 classes. On the 156 classes not in COCO, YOLO9000 gets 16.0 mAP. But YOLO can detect more than just 200 classes; it predicts detections for more than 9000 different object categories. And it still runs in real-time."
                    },
                    {
                        "title": "Abnormal Object Recognition: A Comprehensive Study",
                        "abstract": "When describing images, humans tend not to talk about the obvious, but rather mention what they find interesting. We argue that abnormalities and deviations from typicalities are among the most important components that form what is worth mentioning. In this paper we introduce the abnormality detection as a recognition problem and show how to model typicalities and, consequently, meaningful deviations from prototypical properties of categories. Our model can recognize abnormalities and report the main reasons of any recognized abnormality. We introduce the abnormality detection dataset and show interesting results on how to reason about abnormalities."
                    },
                    {
                        "title": "Unsupervised Deep Embedding for Clustering Analysis",
                        "abstract": "Clustering is central to many data-driven application domains and has been studied extensively in terms of distance functions and grouping algorithms. Relatively little work has focused on learning representations for clustering. In this paper, we propose Deep Embedded Clustering (DEC), a method that simultaneously learns feature representations and cluster assignments using deep neural networks. DEC learns a mapping from the data space to a lower-dimensional feature space in which it iteratively optimizes a clustering objective. Our experimental evaluations on image and text corpora show significant improvement over state-of-the-art methods."
                    },
                    {
                        "title": "Deep3D: Fully Automatic 2D-to-3D Video Conversion with Deep Convolutional Neural Networks",
                        "abstract": "As 3D movie viewing becomes mainstream and Virtual Reality (VR) market emerges, the demand for 3D contents is growing rapidly. Producing 3D videos, however, remains challenging. In this paper we propose to use deep neural networks for automatically converting 2D videos and images to stereoscopic 3D format. In contrast to previous automatic 2D-to-3D conversion algorithms, which have separate stages and need ground truth depth map as supervision, our approach is trained end-to-end directly on stereo pairs extracted from 3D movies. This novel training scheme makes it possible to exploit orders of magnitude more data and significantly increases performance. Indeed, Deep3D outperforms baselines in both quantitative and human subject evaluations."
                    },
                    {
                        "title": "What Should I Do Now? Marrying Reinforcement Learning and Symbolic Planning",
                        "abstract": "Long-term planning poses a major difficulty to many reinforcement learning algorithms. This problem becomes even more pronounced in dynamic visual environments. In this work we propose Hierarchical Planning and Reinforcement Learning (HIP-RL), a method for merging the benefits and capabilities of Symbolic Planning with the learning abilities of Deep Reinforcement Learning. We apply HIPRL to the complex visual tasks of interactive question answering and visual semantic planning and achieve state-of-the-art results on three challenging datasets all while taking fewer steps at test time and training in fewer iterations. Sample results can be found at youtu.be/0TtWJ_0mPfI"
                    },
                    {
                        "title": "SHARCS: Efficient Transformers through Routing with Dynamic Width Sub-networks",
                        "abstract": "We introduce SHARCS for adaptive inference that takes into account the hardness of input samples. SHARCS can train a router on any transformer network, enabling the model to direct different samples to sub-networks with varying widths. Our experiments demonstrate that: (1) SHARCS outperforms or complements existing per-sample adaptive inference methods across various classification tasks in terms of accuracy vs. FLOPs; (2) SHARCS generalizes across different architectures and can be even applied to compressed and efficient transformer encoders to further improve their efficiency; (3) SHARCS can provide a 2 times inference speed up at an insignificant drop in accuracy."
                    },
                    {
                        "title": "Newtonian Image Understanding: Unfolding the Dynamics of Objects in Static Images",
                        "abstract": "In this paper, we study the challenging problem of predicting the dynamics of objects in static images. Given a query object in an image, our goal is to provide a physical understanding of the object in terms of the forces acting upon it and its long term motion as response to those forces. Direct and explicit estimation of the forces and the motion of objects from a single image is extremely challenging. We define intermediate physical abstractions called Newtonian scenarios and introduce Newtonian Neural Network ($N^3$) that learns to map a single image to a state in a Newtonian scenario. Our experimental evaluations show that our method can reliably predict dynamics of a query object from a single image. In addition, our approach can provide physical reasoning that supports the predicted dynamics in terms of velocity and force vectors. To spur research in this direction we compiled Visual Newtonian Dynamics (VIND) dataset that includes 6806 videos aligned with Newtonian scenarios represented using game engines, and 4516 still images with their ground truth dynamics."
                    },
                    {
                        "title": "Actions ~ Transformations",
                        "abstract": "What defines an action like \"kicking ball\"? We argue that the true meaning of an action lies in the change or transformation an action brings to the environment. In this paper, we propose a novel representation for actions by modeling an action as a transformation which changes the state of the environment before the action happens (precondition) to the state after the action (effect). Motivated by recent advancements of video representation using deep learning, we design a Siamese network which models the action as a transformation on a high-level feature space. We show that our model gives improvements on standard action recognition datasets including UCF101 and HMDB51. More importantly, our approach is able to generalize beyond learned action categories and shows significant performance improvement on cross-category generalization on our new ACT dataset."
                    },
                    {
                        "title": "Toward a Taxonomy and Computational Models of Abnormalities in Images",
                        "abstract": "The human visual system can spot an abnormal image, and reason about what makes it strange. This task has not received enough attention in computer vision. In this paper we study various types of atypicalities in images in a more comprehensive way than has been done before. We propose a new dataset of abnormal images showing a wide range of atypicalities. We design human subject experiments to discover a coarse taxonomy of the reasons for abnormality. Our experiments reveal three major categories of abnormality: object-centric, scene-centric, and contextual. Based on this taxonomy, we propose a comprehensive computational model that can predict all different types of abnormality in images and outperform prior arts in abnormality recognition."
                    },
                    {
                        "title": "Are Elephants Bigger than Butterflies? Reasoning about Sizes of Objects",
                        "abstract": "Human vision greatly benefits from the information about sizes of objects. The role of size in several visual reasoning tasks has been thoroughly explored in human perception and cognition. However, the impact of the information about sizes of objects is yet to be determined in AI. We postulate that this is mainly attributed to the lack of a comprehensive repository of size information. In this paper, we introduce a method to automatically infer object sizes, leveraging visual and textual information from web. By maximizing the joint likelihood of textual and visual observations, our method learns reliable relative size estimates, with no explicit human supervision. We introduce the relative size dataset and show that our method outperforms competitive textual and visual baselines in reasoning about size comparisons."
                    },
                    {
                        "title": "XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks",
                        "abstract": "We propose two efficient approximations to standard convolutional neural networks: Binary-Weight-Networks and XNOR-Networks. In Binary-Weight-Networks, the filters are approximated with binary values resulting in 32x memory saving. In XNOR-Networks, both the filters and the input to convolutional layers are binary. XNOR-Networks approximate convolutions using primarily binary operations. This results in 58x faster convolutional operations and 32x memory savings. XNOR-Nets offer the possibility of running state-of-the-art networks on CPUs (rather than GPUs) in real-time. Our binary networks are simple, accurate, efficient, and work on challenging visual tasks. We evaluate our approach on the ImageNet classification task. The classification accuracy with a Binary-Weight-Network version of AlexNet is only 2.9% less than the full-precision AlexNet (in top-1 measure). We compare our method with recent network binarization methods, BinaryConnect and BinaryNets, and outperform these methods by large margins on ImageNet, more than 16% in top-1 accuracy."
                    },
                    {
                        "title": "Query-Reduction Networks for Question Answering",
                        "abstract": "In this paper, we study the problem of question answering when reasoning over multiple facts is required. We propose Query-Reduction Network (QRN), a variant of Recurrent Neural Network (RNN) that effectively handles both short-term (local) and long-term (global) sequential dependencies to reason over multiple facts. QRN considers the context sentences as a sequence of state-changing triggers, and reduces the original query to a more informed query as it observes each trigger (context sentence) through time. Our experiments show that QRN produces the state-of-the-art results in bAbI QA and dialog tasks, and in a real goal-oriented dialog dataset. In addition, QRN formulation allows parallelization on RNN's time axis, saving an order of magnitude in time complexity for training and inference."
                    },
                    {
                        "title": "See the Glass Half Full: Reasoning about Liquid Containers, their Volume and Content",
                        "abstract": "Humans have rich understanding of liquid containers and their contents; for example, we can effortlessly pour water from a pitcher to a cup. Doing so requires estimating the volume of the cup, approximating the amount of water in the pitcher, and predicting the behavior of water when we tilt the pitcher. Very little attention in computer vision has been made to liquids and their containers. In this paper, we study liquid containers and their contents, and propose methods to estimate the volume of containers, approximate the amount of liquid in them, and perform comparative volume estimations all from a single RGB image. Furthermore, we show the results of the proposed model for predicting the behavior of liquids inside containers when one tilts the containers. We also introduce a new dataset of Containers Of liQuid contEnt (COQE) that contains more than 5,000 images of 10,000 liquid containers in context labelled with volume, amount of content, bounding box annotation, and corresponding similar 3D CAD models."
                    },
                    {
                        "title": "Bidirectional Attention Flow for Machine Comprehension",
                        "abstract": "Machine comprehension (MC), answering a query about a given context paragraph, requires modeling complex interactions between the context and the query. Recently, attention mechanisms have been successfully extended to MC. Typically these methods use attention to focus on a small portion of the context and summarize it with a fixed-size vector, couple attentions temporally, and/or often form a uni-directional attention. In this paper we introduce the Bi-Directional Attention Flow (BIDAF) network, a multi-stage hierarchical process that represents the context at different levels of granularity and uses bi-directional attention flow mechanism to obtain a query-aware context representation without early summarization. Our experimental evaluations show that our model achieves the state-of-the-art results in Stanford Question Answering Dataset (SQuAD) and CNN/DailyMail cloze test."
                    },
                    {
                        "title": "LCNN: Lookup-based Convolutional Neural Network",
                        "abstract": "Porting state of the art deep learning algorithms to resource constrained compute platforms (e.g. VR, AR, wearables) is extremely challenging. We propose a fast, compact, and accurate model for convolutional neural networks that enables efficient learning and inference. We introduce LCNN, a lookup-based convolutional neural network that encodes convolutions by few lookups to a dictionary that is trained to cover the space of weights in CNNs. Training LCNN involves jointly learning a dictionary and a small set of linear combinations. The size of the dictionary naturally traces a spectrum of trade-offs between efficiency and accuracy. Our experimental results on ImageNet challenge show that LCNN can offer 3.2x speedup while achieving 55.1% top-1 accuracy using AlexNet architecture. Our fastest LCNN offers 37.6x speed up over AlexNet while maintaining 44.3% top-1 accuracy. LCNN not only offers dramatic speed ups at inference, but it also enables efficient training. In this paper, we show the benefits of LCNN in few-shot learning and few-iteration learning, two crucial aspects of on-device training of deep learning models."
                    },
                    {
                        "title": "VISALOGY: Answering Visual Analogy Questions",
                        "abstract": "In this paper, we study the problem of answering visual analogy questions. These questions take the form of image A is to image B as image C is to what. Answering these questions entails discovering the mapping from image A to image B and then extending the mapping to image C and searching for the image D such that the relation from A to B holds for C to D. We pose this problem as learning an embedding that encourages pairs of analogous images with similar transformations to be close together using convolutional neural networks with a quadruple Siamese architecture. We introduce a dataset of visual analogy questions in natural images, and show first results of its kind on solving analogy questions on natural images."
                    },
                    {
                        "title": "Neural Speed Reading via Skim-RNN",
                        "abstract": "Inspired by the principles of speed reading, we introduce Skim-RNN, a recurrent neural network (RNN) that dynamically decides to update only a small fraction of the hidden state for relatively unimportant input tokens. Skim-RNN gives computational advantage over an RNN that always updates the entire hidden state. Skim-RNN uses the same input and output interfaces as a standard RNN and can be easily used instead of RNNs in existing models. In our experiments, we show that Skim-RNN can achieve significantly reduced computational cost without losing accuracy compared to standard RNNs across five different natural language tasks. In addition, we demonstrate that the trade-off between accuracy and speed of Skim-RNN can be dynamically controlled during inference time in a stable manner. Our analysis also shows that Skim-RNN running on a single CPU offers lower latency compared to standard RNNs on GPUs."
                    },
                    {
                        "title": "Label Refinery: Improving ImageNet Classification through Label Progression",
                        "abstract": "Among the three main components (data, labels, and models) of any supervised learning system, data and models have been the main subjects of active research. However, studying labels and their properties has received very little attention. Current principles and paradigms of labeling impose several challenges to machine learning algorithms. Labels are often incomplete, ambiguous, and redundant. In this paper we study the effects of various properties of labels and introduce the Label Refinery: an iterative procedure that updates the ground truth labels after examining the entire dataset. We show significant gain using refined labels across a wide range of models. Using a Label Refinery improves the state-of-the-art top-1 accuracy of (1) AlexNet from 59.3 to 67.2, (2) MobileNet from 70.6 to 73.39, (3) MobileNet-0.25 from 50.6 to 55.59, (4) VGG19 from 72.7 to 75.46, and (5) Darknet19 from 72.9 to 74.47."
                    },
                    {
                        "title": "OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge",
                        "abstract": "Visual Question Answering (VQA) in its ideal form lets us study reasoning in the joint space of vision and language and serves as a proxy for the AI task of scene understanding. However, most VQA benchmarks to date are focused on questions such as simple counting, visual attributes, and object detection that do not require reasoning or knowledge beyond what is in the image. In this paper, we address the task of knowledge-based visual question answering and provide a benchmark, called OK-VQA, where the image content is not sufficient to answer the questions, encouraging methods that rely on external knowledge resources. Our new dataset includes more than 14,000 questions that require external knowledge to answer. We show that the performance of the state-of-the-art VQA models degrades drastically in this new setting. Our analysis shows that our knowledge-based VQA task is diverse, difficult, and large compared to previous knowledge-based VQA datasets. We hope that this dataset enables researchers to open up new avenues for research in this domain. See http://okvqa.allenai.org to download and browse the dataset."
                    }
                ]
            },
            "4f55f2f4-8a65-49ee-8743-fa0acf7c2961": {
                "pk": "4f55f2f4-8a65-49ee-8743-fa0acf7c2961",
                "name": "Aditya Kusupati",
                "collaborators": [
                    "Ali Farhadi",
                    "Prateek Jain",
                    "Vivek Ramanujan",
                    "Sham Kakade",
                    "Matthew Wallingford",
                    "Manik Varma",
                    "Aniruddha Kembhavi",
                    "Hannaneh Hajishirzi",
                    "Nilesh Gupta",
                    "Raghav Somani"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Representation Learning",
                    "Adaptive Inference"
                ],
                "publications": [
                    {
                        "title": "Disrupting Model Training with Adversarial Shortcuts",
                        "abstract": "When data is publicly released for human consumption, it is unclear how to prevent its unauthorized usage for machine learning purposes. Successful model training may be preventable with carefully designed dataset modifications, and we present a proof-of-concept approach for the image classification setting. We propose methods based on the notion of adversarial shortcuts, which encourage models to rely on non-robust signals rather than semantic features, and our experiments demonstrate that these measures successfully prevent deep learning models from achieving high accuracy on real, unmodified data examples."
                    },
                    {
                        "title": "SHARCS: Efficient Transformers through Routing with Dynamic Width Sub-networks",
                        "abstract": "We introduce SHARCS for adaptive inference that takes into account the hardness of input samples. SHARCS can train a router on any transformer network, enabling the model to direct different samples to sub-networks with varying widths. Our experiments demonstrate that: (1) SHARCS outperforms or complements existing per-sample adaptive inference methods across various classification tasks in terms of accuracy vs. FLOPs; (2) SHARCS generalizes across different architectures and can be even applied to compressed and efficient transformer encoders to further improve their efficiency; (3) SHARCS can provide a 2 times inference speed up at an insignificant drop in accuracy."
                    },
                    {
                        "title": "Are \"Hierarchical\" Visual Representations Hierarchical?",
                        "abstract": "Learned visual representations often capture large amounts of semantic information for accurate downstream applications. Human understanding of the world is fundamentally grounded in hierarchy. To mimic this and further improve representation capabilities, the community has explored \"hierarchical\" visual representations that aim at modeling the underlying hierarchy of the visual world. In this work, we set out to investigate if hierarchical visual representations truly capture the human perceived hierarchy better than standard learned representations. To this end, we create HierNet, a suite of 12 datasets spanning 3 kinds of hierarchy from the BREEDs subset of ImageNet. After extensive evaluation of Hyperbolic and Matryoshka Representations across training setups, we conclude that they do not capture hierarchy any better than the standard representations but can assist in other aspects like search efficiency and interpretability. Our benchmark and the datasets are open-sourced at https://github.com/ethanlshen/HierNet."
                    },
                    {
                        "title": "MatMamba: A Matryoshka State Space Model",
                        "abstract": "State Space Models (SSMs) like Mamba2 are a promising alternative to Transformers, with faster theoretical training and inference times -- especially for long context lengths. Recent work on Matryoshka Representation Learning -- and its application to Transformer backbones in works like MatFormer -- showed how to introduce nested granularities of smaller submodels in one universal elastic model. In this work, we present MatMamba: a state space model which combines Matryoshka-style learning with Mamba2, by modifying the block to contain nested dimensions to enable joint training and adaptive inference. MatMamba allows for efficient and adaptive deployment across various model sizes. We train a single large MatMamba model and are able to get a number of smaller nested models for free -- while maintaining or improving upon the performance of a baseline smaller model trained from scratch. We train language and image models at a variety of parameter sizes from 35M to 1.4B. Our results on ImageNet and FineWeb show that MatMamba models scale comparably to Transformers, while having more efficient inference characteristics. This makes MatMamba a practically viable option for deploying large-scale models in an elastic way based on the available inference compute. Code and models are open sourced at \\url{https://github.com/ScaledFoundations/MatMamba}"
                    },
                    {
                        "title": "Extreme Regression for Dynamic Search Advertising",
                        "abstract": "This paper introduces a new learning paradigm called eXtreme Regression (XR) whose objective is to accurately predict the numerical degrees of relevance of an extremely large number of labels to a data point. XR can provide elegant solutions to many large-scale ranking and recommendation applications including Dynamic Search Advertising (DSA). XR can learn more accurate models than the recently popular extreme classifiers which incorrectly assume strictly binary-valued label relevances. Traditional regression metrics which sum the errors over all the labels are unsuitable for XR problems since they could give extremely loose bounds for the label ranking quality. Also, the existing regression algorithms won't efficiently scale to millions of labels. This paper addresses these limitations through: (1) new evaluation metrics for XR which sum only the k largest regression errors; (2) a new algorithm called XReg which decomposes XR task into a hierarchy of much smaller regression problems thus leading to highly efficient training and prediction. This paper also introduces a (3) new labelwise prediction algorithm in XReg useful for DSA and other recommendation tasks. Experiments on benchmark datasets demonstrated that XReg can outperform the state-of-the-art extreme classifiers as well as large-scale regressors and rankers by up to 50% reduction in the new XR error metric, and up to 2% and 2.4% improvements in terms of the propensity-scored precision metric used in extreme classification and the click-through rate metric used in DSA respectively. Deployment of XReg on DSA in Bing resulted in a relative gain of 27% in query coverage. XReg's source code can be downloaded from http://manikvarma.org/code/XReg/download.html."
                    },
                    {
                        "title": "FastGRNN: A Fast, Accurate, Stable and Tiny Kilobyte Sized Gated Recurrent Neural Network",
                        "abstract": "This paper develops the FastRNN and FastGRNN algorithms to address the twin RNN limitations of inaccurate training and inefficient prediction. Previous approaches have improved accuracy at the expense of prediction costs making them infeasible for resource-constrained and real-time applications. Unitary RNNs have increased accuracy somewhat by restricting the range of the state transition matrix's singular values but have also increased the model size as they require a larger number of hidden units to make up for the loss in expressive power. Gated RNNs have obtained state-of-the-art accuracies by adding extra parameters thereby resulting in even larger models. FastRNN addresses these limitations by adding a residual connection that does not constrain the range of the singular values explicitly and has only two extra scalar parameters. FastGRNN then extends the residual connection to a gate by reusing the RNN matrices to match state-of-the-art gated RNN accuracies but with a 2-4x smaller model. Enforcing FastGRNN's matrices to be low-rank, sparse and quantized resulted in accurate models that could be up to 35x smaller than leading gated and unitary RNNs. This allowed FastGRNN to accurately recognize the \"Hey Cortana\" wakeword with a 1 KB model and to be deployed on severely resource-constrained IoT microcontrollers too tiny to store other RNN models. FastGRNN's code is available at https://github.com/Microsoft/EdgeML/."
                    },
                    {
                        "title": "Soft Threshold Weight Reparameterization for Learnable Sparsity",
                        "abstract": "Sparsity in Deep Neural Networks (DNNs) is studied extensively with the focus of maximizing prediction accuracy given an overall parameter budget. Existing methods rely on uniform or heuristic non-uniform sparsity budgets which have sub-optimal layer-wise parameter allocation resulting in a) lower prediction accuracy or b) higher inference cost (FLOPs). This work proposes Soft Threshold Reparameterization (STR), a novel use of the soft-threshold operator on DNN weights. STR smoothly induces sparsity while learning pruning thresholds thereby obtaining a non-uniform sparsity budget. Our method achieves state-of-the-art accuracy for unstructured sparsity in CNNs (ResNet50 and MobileNetV1 on ImageNet-1K), and, additionally, learns non-uniform budgets that empirically reduce the FLOPs by up to 50%. Notably, STR boosts the accuracy over existing results by up to 10% in the ultra sparse (99%) regime and can also be used to induce low-rank (structured sparsity) in RNNs. In short, STR is a simple mechanism which learns effective sparsity budgets that contrast with popular heuristics. Code, pretrained models and sparsity budgets are at https://github.com/RAIVNLab/STR."
                    },
                    {
                        "title": "RNNPool: Efficient Non-linear Pooling for RAM Constrained Inference",
                        "abstract": "Standard Convolutional Neural Networks (CNNs) designed for computer vision tasks tend to have large intermediate activation maps. These require large working memory and are thus unsuitable for deployment on resource-constrained devices typically used for inference on the edge. Aggressively downsampling the images via pooling or strided convolutions can address the problem but leads to a significant decrease in accuracy due to gross aggregation of the feature map by standard pooling operators. In this paper, we introduce RNNPool, a novel pooling operator based on Recurrent Neural Networks (RNNs), that efficiently aggregates features over large patches of an image and rapidly downsamples activation maps. Empirical evaluation indicates that an RNNPool layer can effectively replace multiple blocks in a variety of architectures such as MobileNets, DenseNet when applied to standard vision tasks like image classification and face detection. That is, RNNPool can significantly decrease computational complexity and peak memory usage for inference while retaining comparable accuracy. We use RNNPool with the standard S3FD architecture to construct a face detection method that achieves state-of-the-art MAP for tiny ARM Cortex-M4 class microcontrollers with under 256 KB of RAM. Code is released at https://github.com/Microsoft/EdgeML."
                    },
                    {
                        "title": "LLC: Accurate, Multi-purpose Learnt Low-dimensional Binary Codes",
                        "abstract": "Learning binary representations of instances and classes is a classical problem with several high potential applications. In modern settings, the compression of high-dimensional neural representations to low-dimensional binary codes is a challenging task and often require large bit-codes to be accurate. In this work, we propose a novel method for Learning Low-dimensional binary Codes (LLC) for instances as well as classes. Our method does not require any side-information, like annotated attributes or label meta-data, and learns extremely low-dimensional binary codes (~20 bits for ImageNet-1K). The learnt codes are super-efficient while still ensuring nearly optimal classification accuracy for ResNet50 on ImageNet-1K. We demonstrate that the learnt codes capture intrinsically important features in the data, by discovering an intuitive taxonomy over classes. We further quantitatively measure the quality of our codes by applying it to the efficient image retrieval as well as out-of-distribution (OOD) detection problems. For ImageNet-100 retrieval problem, our learnt binary codes outperform 16 bit HashNet using only 10 bits and also are as accurate as 10 dimensional real representations. Finally, our learnt binary codes can perform OOD detection, out-of-the-box, as accurately as a baseline that needs ~3000 samples to tune its threshold, while we require none. Code is open-sourced at https://github.com/RAIVNLab/LLC."
                    },
                    {
                        "title": "One Size Does Not Fit All: Multi-Scale, Cascaded RNNs for Radar Classification",
                        "abstract": "Edge sensing with micro-power pulse-Doppler radars is an emergent domain in monitoring and surveillance with several smart city applications. Existing solutions for the clutter versus multi-source radar classification task are limited in terms of either accuracy or efficiency, and in some cases, struggle with a trade-off between false alarms and recall of sources. We find that this problem can be resolved by learning the classifier across multiple time-scales. We propose a multi-scale, cascaded recurrent neural network architecture, MSC-RNN, comprised of an efficient multi-instance learning (MIL) Recurrent Neural Network (RNN) for clutter discrimination at a lower tier, and a more complex RNN classifier for source classification at the upper tier. By controlling the invocation of the upper RNN with the help of the lower tier conditionally, MSC-RNN achieves an overall accuracy of 0.972. Our approach holistically improves the accuracy and per-class recalls over ML models suitable for radar inferencing. Notably, we outperform cross-domain handcrafted feature engineering with time-domain deep feature learning, while also being up to $\\sim$3$\\times$ more efficient than a competitive solution."
                    },
                    {
                        "title": "Neural Radiance Field Codebooks",
                        "abstract": "Compositional representations of the world are a promising step towards enabling high-level scene understanding and efficient transfer to downstream tasks. Learning such representations for complex scenes and tasks remains an open challenge. Towards this goal, we introduce Neural Radiance Field Codebooks (NRC), a scalable method for learning object-centric representations through novel view reconstruction. NRC learns to reconstruct scenes from novel views using a dictionary of object codes which are decoded through a volumetric renderer. This enables the discovery of reoccurring visual and geometric patterns across scenes which are transferable to downstream tasks. We show that NRC representations transfer well to object navigation in THOR, outperforming 2D and 3D representation learning methods by 3.1% success rate. We demonstrate that our approach is able to perform unsupervised segmentation for more complex synthetic (THOR) and real scenes (NYU Depth) better than prior methods (29% relative improvement). Finally, we show that NRC improves on the task of depth ordering by 5.5% accuracy in THOR."
                    },
                    {
                        "title": "Matryoshka Representation Learning",
                        "abstract": "Learned representations are a central component in modern ML systems, serving a multitude of downstream tasks. When training such representations, it is often the case that computational and statistical constraints for each downstream task are unknown. In this context rigid, fixed capacity representations can be either over or under-accommodating to the task at hand. This leads us to ask: can we design a flexible representation that can adapt to multiple downstream tasks with varying computational resources? Our main contribution is Matryoshka Representation Learning (MRL) which encodes information at different granularities and allows a single embedding to adapt to the computational constraints of downstream tasks. MRL minimally modifies existing representation learning pipelines and imposes no additional cost during inference and deployment. MRL learns coarse-to-fine representations that are at least as accurate and rich as independently trained low-dimensional representations. The flexibility within the learned Matryoshka Representations offer: (a) up to 14x smaller embedding size for ImageNet-1K classification at the same level of accuracy; (b) up to 14x real-world speed-ups for large-scale retrieval on ImageNet-1K and 4K; and (c) up to 2% accuracy improvements for long-tail few-shot classification, all while being as robust as the original representations. Finally, we show that MRL extends seamlessly to web-scale datasets (ImageNet, JFT) across various modalities -- vision (ViT, ResNet), vision + language (ALIGN) and language (BERT). MRL code and pretrained models are open-sourced at https://github.com/RAIVNLab/MRL."
                    },
                    {
                        "title": "ProtoSound: A Personalized and Scalable Sound Recognition System for Deaf and Hard-of-Hearing Users",
                        "abstract": "Recent advances have enabled automatic sound recognition systems for deaf and hard of hearing (DHH) users on mobile devices. However, these tools use pre-trained, generic sound recognition models, which do not meet the diverse needs of DHH users. We introduce ProtoSound, an interactive system for customizing sound recognition models by recording a few examples, thereby enabling personalized and fine-grained categories. ProtoSound is motivated by prior work examining sound awareness needs of DHH people and by a survey we conducted with 472 DHH participants. To evaluate ProtoSound, we characterized performance on two real-world sound datasets, showing significant improvement over state-of-the-art (e.g., +9.7% accuracy on the first dataset). We then deployed ProtoSound's end-user training and real-time recognition through a mobile application and recruited 19 hearing participants who listened to the real-world sounds and rated the accuracy across 56 locations (e.g., homes, restaurants, parks). Results show that ProtoSound personalized the model on-device in real-time and accurately learned sounds across diverse acoustic contexts. We close by discussing open challenges in personalizable sound recognition, including the need for better recording interfaces and algorithmic improvements."
                    },
                    {
                        "title": "MatFormer: Nested Transformer for Elastic Inference",
                        "abstract": "Transformer models are deployed in a wide range of settings, from multi-accelerator clusters to standalone mobile phones. The diverse inference constraints in these scenarios necessitate practitioners to train foundation models such as PaLM 2, Llama, & ViTs as a series of models of varying sizes. Due to significant training costs, only a select few model sizes are trained and supported, limiting more fine-grained control over relevant tradeoffs, including latency, cost, and accuracy. This work introduces MatFormer, a nested Transformer architecture designed to offer elasticity in a variety of deployment constraints. Each Feed Forward Network (FFN) block of a MatFormer model is jointly optimized with a few nested smaller FFN blocks. This training procedure allows for the Mix'n'Match of model granularities across layers -- i.e., a trained universal MatFormer model enables extraction of hundreds of accurate smaller models, which were never explicitly optimized. We empirically demonstrate MatFormer's effectiveness across different model classes (decoders & encoders), modalities (language & vision), and scales (up to 2.6B parameters). We find that a 2.6B decoder-only MatFormer language model (MatLM) allows us to extract smaller models spanning from 1.5B to 2.6B, each exhibiting comparable validation loss and one-shot downstream evaluations to their independently trained counterparts. Furthermore, we observe that smaller encoders extracted from a universal MatFormer-based ViT (MatViT) encoder preserve the metric-space structure for adaptive large-scale retrieval. Finally, we showcase that speculative decoding with the accurate and consistent submodels extracted from MatFormer can further reduce inference latency."
                    },
                    {
                        "title": "Neural Priming for Sample-Efficient Adaptation",
                        "abstract": "We propose Neural Priming, a technique for adapting large pretrained models to distribution shifts and downstream tasks given few or no labeled examples. Presented with class names or unlabeled test samples, Neural Priming enables the model to recall and conditions its parameters on relevant data seen throughout pretraining, thereby priming it for the test distribution. Neural Priming can be performed at test time, even for pretraining datasets as large as LAION-2B. Performing lightweight updates on the recalled data significantly improves accuracy across a variety of distribution shift and transfer learning benchmarks. Concretely, in the zero-shot setting, we see a 2.45% improvement in accuracy on ImageNet and 3.81% accuracy improvement on average across standard transfer learning benchmarks. Further, using Neural Priming at inference to adapt to distribution shift, we see a 1.41% accuracy improvement on ImageNetV2. These results demonstrate the effectiveness of Neural Priming in addressing the challenge of limited labeled data and changing distributions. Code is available at github.com/RAIVNLab/neural-priming."
                    },
                    {
                        "title": "EHI: End-to-end Learning of Hierarchical Index for Efficient Dense Retrieval",
                        "abstract": "Dense embedding-based retrieval is widely used for semantic search and ranking. However, conventional two-stage approaches, involving contrastive embedding learning followed by approximate nearest neighbor search (ANNS), can suffer from misalignment between these stages. This mismatch degrades retrieval performance. We propose End-to-end Hierarchical Indexing (EHI), a novel method that directly addresses this issue by jointly optimizing embedding generation and ANNS structure. EHI leverages a dual encoder for embedding queries and documents while simultaneously learning an inverted file index (IVF)-style tree structure. To facilitate the effective learning of this discrete structure, EHI introduces dense path embeddings that encodes the path traversed by queries and documents within the tree. Extensive evaluations on standard benchmarks, including MS MARCO (Dev set) and TREC DL19, demonstrate EHI's superiority over traditional ANNS index. Under the same computational constraints, EHI outperforms existing state-of-the-art methods by +1.45% in MRR@10 on MS MARCO (Dev) and +8.2% in nDCG@10 on TREC DL19, highlighting the benefits of our end-to-end approach."
                    },
                    {
                        "title": "FLUID: A Unified Evaluation Framework for Flexible Sequential Data",
                        "abstract": "Modern ML methods excel when training data is IID, large-scale, and well labeled. Learning in less ideal conditions remains an open challenge. The sub-fields of few-shot, continual, transfer, and representation learning have made substantial strides in learning under adverse conditions; each affording distinct advantages through methods and insights. These methods address different challenges such as data arriving sequentially or scarce training examples, however often the difficult conditions an ML system will face over its lifetime cannot be anticipated prior to deployment. Therefore, general ML systems which can handle the many challenges of learning in practical settings are needed. To foster research towards the goal of general ML methods, we introduce a new unified evaluation framework - FLUID (Flexible Sequential Data). FLUID integrates the objectives of few-shot, continual, transfer, and representation learning while enabling comparison and integration of techniques across these subfields. In FLUID, a learner faces a stream of data and must make sequential predictions while choosing how to update itself, adapt quickly to novel classes, and deal with changing data distributions; while accounting for the total amount of compute. We conduct experiments on a broad set of methods which shed new insight on the advantages and limitations of current solutions and indicate new research problems to solve. As a starting point towards more general methods, we present two new baselines which outperform other evaluated methods on FLUID. Project page: https://raivn.cs.washington.edu/projects/FLUID/."
                    },
                    {
                        "title": "Mixture of Nested Experts: Adaptive Processing of Visual Tokens",
                        "abstract": "The visual medium (images and videos) naturally contains a large amount of information redundancy, thereby providing a great opportunity for leveraging efficiency in processing. While Vision Transformer (ViT) based models scale effectively to large data regimes, they fail to capitalize on this inherent redundancy, leading to higher computational costs. Mixture of Experts (MoE) networks demonstrate scalability while maintaining same inference-time costs, but they come with a larger parameter footprint. We present Mixture of Nested Experts (MoNE), which utilizes a nested structure for experts, wherein individual experts fall on an increasing compute-accuracy curve. Given a compute budget, MoNE learns to dynamically choose tokens in a priority order, and thus redundant tokens are processed through cheaper nested experts. Using this framework, we achieve equivalent performance as the baseline models, while reducing inference time compute by over two-fold. We validate our approach on standard image and video datasets - ImageNet-21K, Kinetics400, and Something-Something-v2. We further highlight MoNE$'$s adaptability by showcasing its ability to maintain strong performance across different inference-time compute budgets on videos, using only a single trained model."
                    },
                    {
                        "title": "MADLAD-400: A Multilingual And Document-Level Large Audited Dataset",
                        "abstract": "We introduce MADLAD-400, a manually audited, general domain 3T token monolingual dataset based on CommonCrawl, spanning 419 languages. We discuss the limitations revealed by self-auditing MADLAD-400, and the role data auditing had in the dataset creation process. We then train and release a 10.7B-parameter multilingual machine translation model on 250 billion tokens covering over 450 languages using publicly available data, and find that it is competitive with models that are significantly larger, and report the results on different domains. In addition, we train a 8B-parameter language model, and assess the results on few-shot translation. We make the baseline models available to the research community."
                    },
                    {
                        "title": "AdANNS: A Framework for Adaptive Semantic Search",
                        "abstract": "Web-scale search systems learn an encoder to embed a given query which is then hooked into an approximate nearest neighbor search (ANNS) pipeline to retrieve similar data points. To accurately capture tail queries and data points, learned representations typically are rigid, high-dimensional vectors that are generally used as-is in the entire ANNS pipeline and can lead to computationally expensive retrieval. In this paper, we argue that instead of rigid representations, different stages of ANNS can leverage adaptive representations of varying capacities to achieve significantly better accuracy-compute trade-offs, i.e., stages of ANNS that can get away with more approximate computation should use a lower-capacity representation of the same data point. To this end, we introduce AdANNS, a novel ANNS design framework that explicitly leverages the flexibility of Matryoshka Representations. We demonstrate state-of-the-art accuracy-compute trade-offs using novel AdANNS-based key ANNS building blocks like search data structures (AdANNS-IVF) and quantization (AdANNS-OPQ). For example on ImageNet retrieval, AdANNS-IVF is up to 1.5% more accurate than the rigid representations-based IVF at the same compute budget; and matches accuracy while being up to 90x faster in wall-clock time. For Natural Questions, 32-byte AdANNS-OPQ matches the accuracy of the 64-byte OPQ baseline constructed using rigid representations -- same accuracy at half the cost! We further show that the gains from AdANNS translate to modern-day composite ANNS indices that combine search structures and quantization. Finally, we demonstrate that AdANNS can enable inference-time adaptivity for compute-aware search on ANNS indices built non-adaptively on matryoshka representations. Code is open-sourced at https://github.com/RAIVNLab/AdANNS."
                    }
                ]
            }
        }
    },
    "2405.15119": {
        "paper_data": {
            "title": "Bayesian Optimization of Functions over Node Subsets in Graphs",
            "url": "http://arxiv.org/abs/2405.15119v1",
            "arxiv_id": "2405.15119",
            "authors": [
                "Huidong Liang",
                "Xingchen Wan",
                "Xiaowen Dong"
            ],
            "abstract": "We address the problem of optimizing over functions defined on node subsets in a graph. The optimization of such functions is often a non-trivial task given their combinatorial, black-box and expensive-to-evaluate nature. Although various algorithms have been introduced in the literature, most are either task-specific or computationally inefficient and only utilize information about the graph structure without considering the characteristics of the function. To address these limitations, we utilize Bayesian Optimization (BO), a sample-efficient black-box solver, and propose a novel framework for combinatorial optimization on graphs. More specifically, we map each $k$-node subset in the original graph to a node in a new combinatorial graph and adopt a local modeling approach to efficiently traverse the latter graph by progressively sampling its subgraphs using a recursive algorithm. Extensive experiments under both synthetic and real-world setups demonstrate the effectiveness of the proposed BO framework on various types of graphs and optimization tasks, where its behavior is analyzed in detail with ablation studies.",
            "introduction": "   1 Introduction  In the analysis and optimization of transportation, social, and epidemiological networks, one is often interested in finding a node subset that leads to the maximization of a utility. For example, incentivizing an initial set of users in a social network such that it leads to the maximum adoption of certain products; protecting a set of key individuals in an epidemiological contact network such that it maximally slows down the transmission of disease; identifying the most vulnerable junctions in a power grid or a road network such that interventions can be made to improve the resilience of these infrastructure networks.   The scenarios described above can be mathematically formulated as optimizing over a utility function defined on node subsets in a graph, which is a non-trivial task for several reasons. First, most conventional optimization algorithms are designed for continuous space and are hence not directly applicable to functions defined on discrete domains such as graphs. Second, optimizing over a k\ud835\udc58kitalic_k-node subset leads to a large search space even for moderate graphs, which are not even fully observable in certain scenarios (e.g. offline social networks). Finally, the objective functions are usually black-box and expensive to evaluate in many applications, such as the outcome of a diffusion process on the network [56] or the output of a graph neural network [50], making sample-efficient queries a necessary requirement.   Assuming the graph structure is fully available, the optimization task described above shares similarities with those encountered in the literature on network-based diffusion. In that literature, greedy algorithms [28, 32, 7] have been widely used to select a subset of nodes that maximizes a utility function, for example in the context of influence maximization [48] or source identification [25]. However, as the underlying functions often require calculating expectations over a large number of simulations (e.g. the expected number of eventual infections from an epidemic process), such algorithms often become extremely time-consuming as the evaluation time for each diffusion process increases [2]. To relieve the inefficiency in computation, proxy-based methods, such as PageRank [6], generalized random walks [10], and DomiRank [17], are often used in practice to rank the importance of nodes. However, such methods completely ignore the underlying function and require full knowledge of the graph structure beforehand. Finally, most methods mentioned above are task-specific, and the one designed for a specific diffusion process usually does not generalize well to another.   In this paper, we consider the challenging optimization setting for black-box functions on node subsets, where the underlying graph structure is not fully observable and can only be incrementally revealed by queries on the fly. To facilitate this setting, we propose a novel strategy to conduct the search in a combinatorial graph space (termed \u201ccombo-graph\u201d) in which each node corresponds to a k\ud835\udc58kitalic_k-node subset in the original (unknown) graph. The original problem is thus turned into optimization over a function on the combo-graph, where each node value is the utility of the corresponding subset. Traditional graph-traversing methods, such as breadth-first search (BFS) or depth-first search (DFS), may not work well in this case due to the exceedingly large search space and",
            "references": [
                {
                    "title": "Neural Latent Geometry Search: Product Manifold Inference via Gromov-Hausdorff-Informed Bayesian Optimization",
                    "abstract": "Recent research indicates that the performance of machine learning models can be improved by aligning the geometry of the latent space with the underlying data structure. Rather than relying solely on Euclidean space, researchers have proposed using hyperbolic and spherical spaces with constant curvature, or combinations thereof, to better model the latent space and enhance model performance. However, little attention has been given to the problem of automatically identifying the optimal latent geometry for the downstream task. We mathematically define this novel formulation and coin it as neural latent geometry search (NLGS). More specifically, we introduce an initial attempt to search for a latent geometry composed of a product of constant curvature model spaces with a small number of query evaluations, under some simplifying assumptions. To accomplish this, we propose a novel notion of distance between candidate latent geometries based on the Gromov-Hausdorff distance from metric geometry. In order to compute the Gromov-Hausdorff distance, we introduce a mapping function that enables the comparison of different manifolds by embedding them in a common high-dimensional ambient space. We then design a graph search space based on the notion of smoothness between latent geometries and employ the calculated distances as an additional inductive bias. Finally, we use Bayesian optimization to search for the optimal latent geometry in a query-efficient manner. This is a general method which can be applied to search for the optimal latent geometry for a variety of models and downstream tasks. We perform experiments on synthetic and real-world datasets to identify the optimal latent geometry for multiple machine learning problems."
                },
                {
                    "title": "Bayesian Optimisation of Functions on Graphs",
                    "abstract": "The increasing availability of graph-structured data motivates the task of optimising over functions defined on the node set of graphs. Traditional graph search algorithms can be applied in this case, but they may be sample-inefficient and do not make use of information about the function values; on the other hand, Bayesian optimisation is a class of promising black-box solvers with superior sample efficiency, but it has been scarcely been applied to such novel setups. To fill this gap, we propose a novel Bayesian optimisation framework that optimises over functions defined on generic, large-scale and potentially unknown graphs. Through the learning of suitable kernels on graphs, our framework has the advantage of adapting to the behaviour of the target function. The local modelling approach further guarantees the efficiency of our method. Extensive experiments on both synthetic and real-world graphs demonstrate the effectiveness of the proposed optimisation framework."
                },
                {
                    "title": "Recent Advances in Bayesian Optimization",
                    "abstract": "Bayesian optimization has emerged at the forefront of expensive black-box optimization due to its data efficiency. Recent years have witnessed a proliferation of studies on the development of new Bayesian optimization algorithms and their applications. Hence, this article attempts to provide a comprehensive and updated survey of recent advances in Bayesian optimization that are mainly based on Gaussian processes and identify challenging open problems. We categorize the existing work on Bayesian optimization into nine main groups according to the motivations and focus of the proposed algorithms. For each category, we present the main advances with respect to the construction of surrogate models and adaptation of the acquisition functions. Finally, we discuss the open questions and suggest promising future research directions, in particular with regard to heterogeneity, privacy preservation, and fairness in distributed and federated optimization systems."
                },
                {
                    "title": "\u03c0BO: Augmenting Acquisition Functions with User Beliefs for Bayesian Optimization",
                    "abstract": "Bayesian optimization (BO) has become an established framework and popular tool for hyperparameter optimization (HPO) of machine learning (ML) algorithms. While known for its sample-efficiency, vanilla BO can not utilize readily available prior beliefs the practitioner has on the potential location of the optimum. Thus, BO disregards a valuable source of information, reducing its appeal to ML practitioners. To address this issue, we propose $\\pi$BO, an acquisition function generalization which incorporates prior beliefs about the location of the optimum in the form of a probability distribution, provided by the user. In contrast to previous approaches, $\\pi$BO is conceptually simple and can easily be integrated with existing libraries and many acquisition functions. We provide regret bounds when $\\pi$BO is applied to the common Expected Improvement acquisition function and prove convergence at regular rates independently of the prior. Further, our experiments show that $\\pi$BO outperforms competing approaches across a wide suite of benchmarks and prior characteristics. We also demonstrate that $\\pi$BO improves on the state-of-the-art performance for a popular deep learning task, with a 12.5 $\\times$ time-to-accuracy speedup over prominent BO approaches."
                },
                {
                    "title": "Combining Latent Space and Structured Kernels for Bayesian Optimization over Combinatorial Spaces",
                    "abstract": "We consider the problem of optimizing combinatorial spaces (e.g., sequences, trees, and graphs) using expensive black-box function evaluations. For example, optimizing molecules for drug design using physical lab experiments. Bayesian optimization (BO) is an efficient framework for solving such problems by intelligently selecting the inputs with high utility guided by a learned surrogate model. A recent BO approach for combinatorial spaces is through a reduction to BO over continuous spaces by learning a latent representation of structures using deep generative models (DGMs). The selected input from the continuous space is decoded into a discrete structure for performing function evaluation. However, the surrogate model over the latent space only uses the information learned by the DGM, which may not have the desired inductive bias to approximate the target black-box function. To overcome this drawback, this paper proposes a principled approach referred as LADDER. The key idea is to define a novel structure-coupled kernel that explicitly integrates the structural information from decoded structures with the learned latent space representation for better surrogate modeling. Our experiments on real-world benchmarks show that LADDER significantly improves over the BO over latent space method, and performs better or similar to state-of-the-art methods."
                },
                {
                    "title": "Graph-Based Bayesian Optimization for Large-Scale Objective-Based Experimental Design",
                    "abstract": "Design is an inseparable part of most scientific and engineering tasks, including real and simulation-based experimental design processes and parameter/hyperparameter tuning/optimization. Several model-based experimental design techniques have been developed for design in domains with partial available knowledge about the underlying process. This article focuses on a powerful class of model-based experimental design called the mean objective cost of uncertainty (MOCU). The MOCU-based techniques are objective-based, meaning that they take the main objective of the process into account during the experimental design process. However, the lack of scalability of MOCU-based techniques prevents their application to most practical problems, including large discrete or combinatorial spaces. To achieve a scalable objective-based experimental design, this article proposes a graph-based MOCU-based Bayesian optimization framework. The correlations among samples in the large design space are accounted for using a graph-based Gaussian process, and an efficient closed-form sequential selection is achieved through the well-known expected improvement policy. The proposed framework\u2019s performance is assessed through the structural intervention in gene regulatory networks, aiming to make the network away from the states associated with cancer."
                },
                {
                    "title": "Think Global and Act Local: Bayesian Optimisation over High-Dimensional Categorical and Mixed Search Spaces",
                    "abstract": "High-dimensional black-box optimisation remains an important yet notoriously challenging problem. Despite the success of Bayesian optimisation methods on continuous domains, domains that are categorical, or that mix continuous and categorical variables, remain challenging. We propose a novel solution -- we combine local optimisation with a tailored kernel design, effectively handling high-dimensional categorical and mixed search spaces, whilst retaining sample efficiency. We further derive convergence guarantee for the proposed approach. Finally, we demonstrate empirically that our method outperforms the current baselines on a variety of synthetic and real-world tasks in terms of performance, computational costs, or both."
                },
                {
                    "title": "Mercer Features for Efficient Combinatorial Bayesian Optimization",
                    "abstract": "Bayesian optimization (BO) is an efficient framework for solving black-box optimization problems with expensive function evaluations. This paper addresses the BO problem setting for combinatorial spaces (e.g., sequences and graphs) that occurs naturally in science and engineering applications. A prototypical example is molecular optimization guided by expensive experiments. The key challenge is to balance the complexity of statistical models and tractability of search to select combinatorial structures for evaluation. In this paper, we propose an efficient approach referred as Mercer Features for Combinatorial Bayesian Optimization (MerCBO). The key idea behind MerCBO is to provide explicit feature maps for diffusion kernels over discrete objects by exploiting the structure of their combinatorial graph representation. These Mercer features combined with Thompson sampling as the acquisition function allows us to employ efficient solvers for finding the next structure for evaluation. Experimental evaluation on diverse real-world benchmarks demonstrates that MerCBO performs similarly or better than prior methods."
                },
                {
                    "title": "Graph Signal Processing for Machine Learning: A Review and New Perspectives",
                    "abstract": "The effective representation, processing, analysis, and visualization of large-scale structured data, especially those related to complex domains, such as networks and graphs, are one of the key questions in modern machine learning. Graph signal processing (GSP), a vibrant branch of signal processing models and algorithms that aims at handling data supported on graphs, opens new paths of research to address this challenge. In this article, we review a few important contributions made by GSP concepts and tools, such as graph filters and transforms, to the development of novel machine learning algorithms. In particular, our discussion focuses on the following three aspects: exploiting data structure and relational priors, improving data and computational efficiency, and enhancing model interpretability. Furthermore, we provide new perspectives on the future development of GSP techniques that may serve as a bridge between applied mathematics and signal processing on one side and machine learning and network science on the other. Cross-fertilization across these different disciplines may help unlock the numerous challenges of complex data analysis in the modern age."
                },
                {
                    "title": "TUDataset: A collection of benchmark datasets for learning with graphs",
                    "abstract": "Recently, there has been an increasing interest in (supervised) learning with graph data, especially using graph neural networks. However, the development of meaningful benchmark datasets and standardized evaluation procedures is lagging, consequently hindering advancements in this area. To address this, we introduce the TUDataset for graph classification and regression. The collection consists of over 120 datasets of varying sizes from a wide range of applications. We provide Python-based data loaders, kernel and graph neural network baseline implementations, and evaluation tools. Here, we give an overview of the datasets, standardized evaluation procedures, and provide baseline experiments. All datasets are available at this http URL. The experiments are fully reproducible from the code available at this http URL."
                },
                {
                    "title": "Effective containment explains subexponential growth in recent confirmed COVID-19 cases in China",
                    "abstract": "Containment works National governments have taken different approaches in response to the coronavirus disease 2019 (COVID-19) pandemic, ranging from draconian quarantines to laissez-faire mitigation strategies. In data from China collected in February 2020, Maier and Brockmann noticed that, unexpectedly, the epidemic did not take off exponentially. Nonexponential spread occurs when the supply of susceptible individuals is depleted on a time scale comparable to the infectious period of the virus. The results of the authors' modeling approach indicate that the public response to the epidemic plus containment policies were becoming effective despite the initial increase in confirmed cases. Science, this issue p. 742 Feedback between the epidemic process, the behavioral response to it, and nationwide containment policies can restrain COVID-19. The recent outbreak of coronavirus disease 2019 (COVID-19) in mainland China was characterized by a distinctive subexponential increase of confirmed cases during the early phase of the epidemic, contrasting with an initial exponential growth expected for an unconstrained outbreak. We show that this effect can be explained as a direct consequence of containment policies that effectively deplete the susceptible population. To this end, we introduce a parsimonious model that captures both quarantine of symptomatic infected individuals, as well as population-wide isolation practices in response to containment policies or behavioral changes, and show that the model captures the observed growth behavior accurately. The insights provided here may aid the careful implementation of containment strategies for ongoing secondary outbreaks of COVID-19 or similar future outbreaks of other emergent infectious diseases."
                },
                {
                    "title": "Scalable Global Optimization via Local Bayesian Optimization",
                    "abstract": "Bayesian optimization has recently emerged as a popular method for the sample-efficient optimization of expensive black-box functions. However, the application to high-dimensional problems with several thousand observations remains challenging, and on difficult problems Bayesian optimization is often not competitive with other paradigms. In this paper we take the view that this is due to the implicit homogeneity of the global probabilistic models and an overemphasized exploration that results from global acquisition. This motivates the design of a local probabilistic approach for global optimization of large-scale high-dimensional problems. We propose the TuRBO algorithm that fits a collection of local models and performs a principled global allocation of samples across these models via an implicit bandit approach. A comprehensive evaluation demonstrates that TuRBO outperforms state-of-the-art methods from machine learning and operations research on problems spanning reinforcement learning, robotics, and the natural sciences."
                },
                {
                    "title": "ChemBO: Bayesian Optimization of Small Organic Molecules with Synthesizable Recommendations",
                    "abstract": "In applications such as molecule design or drug discovery, it is desirable to have an algorithm which recommends new candidate molecules based on the results of past tests. These molecules first need to be synthesized and then tested for objective properties. We describe ChemBO, a Bayesian optimization framework for generating and optimizing organic molecules for desired molecular properties. While most existing data-driven methods for this problem do not account for sample efficiency or fail to enforce realistic constraints on synthesizability, our approach explores the synthesis graph in a sample-efficient way and produces synthesizable candidates. We implement ChemBO as a Gaussian process model and explore existing molecular kernels for it. Moreover, we propose a novel optimal-transport based distance and kernel that accounts for graphical information explicitly. In our experiments, we demonstrate the efficacy of the proposed approach on several molecular optimization problems."
                },
                {
                    "title": "Bayesian Optimisation over Multiple Continuous and Categorical Inputs",
                    "abstract": "Efficient optimisation of black-box problems that comprise both continuous and categorical inputs is important, yet poses significant challenges. We propose a new approach, Continuous and Categorical Bayesian Optimisation (CoCaBO), which combines the strengths of multi-armed bandits and Bayesian optimisation to select values for both categorical and continuous inputs. We model this mixed-type space using a Gaussian Process kernel, designed to allow sharing of information across multiple categorical variables, each with multiple possible values; this allows CoCaBO to leverage all available data efficiently. We extend our method to the batch setting and propose an efficient selection procedure that dynamically balances exploration and exploitation whilst encouraging batch diversity. We demonstrate empirically that our method outperforms existing approaches on both synthetic and real-world optimisation tasks with continuous and categorical inputs."
                },
                {
                    "title": "Combinatorial Bayesian Optimization using the Graph Cartesian Product",
                    "abstract": "This paper focuses on Bayesian Optimization (BO) for objectives on combinatorial search spaces, including ordinal and categorical variables. Despite the abundance of potential applications of Combinatorial BO, including chipset configuration search and neural architecture search, only a handful of methods have been pro- posed. We introduce COMBO, a new Gaussian Process (GP) BO. COMBO quantifies \u201csmoothness\u201d of functions on combinatorial search spaces by utilizing a combinatorial graph. The vertex set of the combinatorial graph consists of all possible joint assignments of the variables, while edges are constructed using the graph Cartesian product of the sub-graphs that represent the individual variables. On this combinatorial graph, we propose an ARD diffusion kernel with which the GP is able to model high-order interactions between variables leading to better performance. Moreover, using the Horseshoe prior for the scale parameter in the ARD diffusion kernel results in an effective variable selection procedure, making COMBO suitable for high dimensional problems. Computationally, in COMBO the graph Cartesian product allows the Graph Fourier Transform calculation to scale linearly instead of exponentially.We validate COMBO in a wide array of real- istic benchmarks, including weighted maximum satisfiability problems and neural architecture search. COMBO outperforms consistently the latest state-of-the-art while maintaining computational and statistical efficiency"
                },
                {
                    "title": "Adversarial Attack and Defense on Graph Data: A Survey",
                    "abstract": "Deep neural networks (DNNs) have been widely applied to various applications, including image classification, text generation, audio recognition, and graph data analysis. However, recent studies have shown that DNNs are vulnerable to adversarial attacks. Though there are several works about adversarial attack and defense strategies on domains such as images and natural language processing, it is still difficult to directly transfer the learned knowledge to graph data due to its representation structure. Given the importance of graph analysis, an increasing number of studies over the past few years have attempted to analyze the robustness of machine learning models on graph data. Nevertheless, existing research considering adversarial behaviors on graph data often focuses on specific types of attacks with certain assumptions. In addition, each work proposes its own mathematical formulation, which makes the comparison among different methods difficult. Therefore, this review is intended to provide an overall landscape of more than 100 papers on adversarial attack and defense strategies for graph data, and establish a unified formulation encompassing most graph adversarial learning models. Moreover, we also compare different graph attacks and defenses along with their contributions and limitations, as well as summarize the evaluation metrics, datasets and future trends. We hope this survey can help fill the gap in the literature and facilitate further development of this promising new field1."
                },
                {
                    "title": "Pitfalls of Graph Neural Network Evaluation",
                    "abstract": "Semi-supervised node classification in graphs is a fundamental problem in graph mining, and the recently proposed graph neural networks (GNNs) have achieved unparalleled results on this task. Due to their massive success, GNNs have attracted a lot of attention, and many novel architectures have been put forward. In this paper we show that existing evaluation strategies for GNN models have serious shortcomings. We show that using the same train/validation/test splits of the same datasets, as well as making significant changes to the training procedure (e.g. early stopping criteria) precludes a fair comparison of different architectures. We perform a thorough empirical evaluation of four prominent GNN models and show that considering different splits of the data leads to dramatically different rankings of models. Even more importantly, our findings suggest that simpler GNN architectures are able to outperform the more sophisticated ones if the hyperparameters and the training procedure are tuned fairly for all models."
                },
                {
                    "title": "How Powerful are Graph Neural Networks?",
                    "abstract": "Graph Neural Networks (GNNs) are an effective framework for representation learning of graphs. GNNs follow a neighborhood aggregation scheme, where the representation vector of a node is computed by recursively aggregating and transforming representation vectors of its neighboring nodes. Many GNN variants have been proposed and have achieved state-of-the-art results on both node and graph classification tasks. However, despite GNNs revolutionizing graph representation learning, there is limited understanding of their representational properties and limitations. Here, we present a theoretical framework for analyzing the expressive power of GNNs to capture different graph structures. Our results characterize the discriminative power of popular GNN variants, such as Graph Convolutional Networks and GraphSAGE, and show that they cannot learn to distinguish certain simple graph structures. We then develop a simple architecture that is provably the most expressive among the class of GNNs and is as powerful as the Weisfeiler-Lehman graph isomorphism test. We empirically validate our theoretical findings on a number of graph classification benchmarks, and demonstrate that our model achieves state-of-the-art performance."
                },
                {
                    "title": "Influence Maximization on Social Graphs: A Survey",
                    "abstract": "Influence Maximization (IM), which selects a set of <inline-formula><tex-math notation=\"LaTeX\">$k$</tex-math> <alternatives><inline-graphic xlink:href=\"li-ieq1-2807843.gif\"/></alternatives></inline-formula> users (called seed set) from a social network to maximize the expected number of influenced users (called influence spread), is a key algorithmic problem in social influence analysis. Due to its immense application potential and enormous technical challenges, IM has been extensively studied in the past decade. In this paper, we survey and synthesize a wide spectrum of existing studies on IM from an <italic>algorithmic perspective</italic>, with a special focus on the following key aspects: (1) a review of well-accepted diffusion models that capture the information diffusion process and build the foundation of the IM problem, (2) a fine-grained taxonomy to classify existing IM algorithms based on their design objectives, (3) a rigorous theoretical comparison of existing IM algorithms, and (4) a comprehensive study on the applications of IM techniques in combining with novel context features of social networks such as topic, location, and time. Based on this analysis, we then outline the key challenges and research directions to expand the boundary of IM research."
                },
                {
                    "title": "Bayesian Optimization of Combinatorial Structures",
                    "abstract": "The optimization of expensive-to-evaluate black-box functions over combinatorial structures is an ubiquitous task in machine learning, engineering and the natural sciences. The combinatorial explosion of the search space and costly evaluations pose challenges for current techniques in discrete optimization and machine learning, and critically require new algorithmic ideas. This article proposes, to the best of our knowledge, the first algorithm to overcome these challenges, based on an adaptive, scalable model that identifies useful combinatorial structure even when data is scarce. Our acquisition function pioneers the use of semidefinite programming to achieve efficiency and scalability. Experimental evaluations demonstrate that this algorithm consistently outperforms other methods from combinatorial and Bayesian optimization."
                },
                {
                    "title": "Neural Architecture Search with Bayesian Optimisation and Optimal Transport",
                    "abstract": "Bayesian Optimisation (BO) refers to a class of methods for global optimisation of a function $f$ which is only accessible via point evaluations. It is typically used in settings where $f$ is expensive to evaluate. A common use case for BO in machine learning is model selection, where it is not possible to analytically model the generalisation performance of a statistical model, and we resort to noisy and expensive training and validation procedures to choose the best model. Conventional BO methods have focused on Euclidean and categorical domains, which, in the context of model selection, only permits tuning scalar hyper-parameters of machine learning algorithms. However, with the surge of interest in deep learning, there is an increasing demand to tune neural network \\emph{architectures}. In this work, we develop NASBOT, a Gaussian process based BO framework for neural architecture search. To accomplish this, we develop a distance metric in the space of neural network architectures which can be computed efficiently via an optimal transport program. This distance might be of independent interest to the deep learning community as it may find applications outside of BO. We demonstrate that NASBOT outperforms other alternatives for architecture search in several cross validation based model selection tasks on multi-layer perceptrons and convolutional neural networks."
                },
                {
                    "title": "Debunking the Myths of Influence Maximization: An In-Depth Benchmarking Study",
                    "abstract": "Influence maximization (IM) on social networks is one of the most active areas of research in computer science. While various IM techniques proposed over the last decade have definitely enriched the field, unfortunately, experimental reports on existing techniques fall short in validity and integrity since many comparisons are not based on a common platform or merely discussed in theory. In this paper, we perform an in-depth benchmarking study of IM techniques on social networks. Specifically, we design a benchmarking platform, which enables us to evaluate and compare the existing techniques systematically and thoroughly under identical experimental conditions. Our benchmarking results analyze and diagnose the inherent deficiencies of the existing approaches and surface the open challenges in IM even after a decade of research. More fundamentally, we unearth and debunk a series of myths and establish that there is no single state-of-the-art technique in IM. At best, a technique is the state of the art in only one aspect."
                },
                {
                    "title": "Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering",
                    "abstract": "In this work, we are interested in generalizing convolutional neural networks (CNNs) from low-dimensional regular grids, where image, video and speech are represented, to high-dimensional irregular domains, such as social networks, brain connectomes or words' embedding, represented by graphs. We present a formulation of CNNs in the context of spectral graph theory, which provides the necessary mathematical background and efficient numerical schemes to design fast localized convolutional filters on graphs. Importantly, the proposed technique offers the same linear computational complexity and constant learning complexity as classical CNNs, while being universal to any graph structure. Experiments on MNIST and 20NEWS demonstrate the ability of this novel deep learning system to learn local, stationary, and compositional features on graphs."
                },
                {
                    "title": "Robust Influence Maximization",
                    "abstract": "In this paper, we address the important issue of uncertainty in the edge influence probability estimates for the well studied influence maximization problem --- the task of finding k seed nodes in a social network to maximize the influence spread. We propose the problem of robust influence maximization, which maximizes the worst-case ratio between the influence spread of the chosen seed set and the optimal seed set, given the uncertainty of the parameter input. We design an algorithm that solves this problem with a solution-dependent bound. We further study uniform sampling and adaptive sampling methods to effectively reduce the uncertainty on parameters and improve the robustness of the influence maximization task. Our empirical results show that parameter uncertainty may greatly affect influence maximization performance and prior studies that learned influence probabilities could lead to poor performance in robust influence maximization due to relatively large uncertainty in parameter estimates, and information cascade based adaptive sampling method may be an effective way to improve the robustness of influence maximization."
                },
                {
                    "title": "Trust Region Policy Optimization",
                    "abstract": "In this article, we describe a method for optimizing control policies, with guaranteed monotonic improvement. By making several approximations to the theoretically-justified scheme, we develop a practical algorithm, called Trust Region Policy Optimization (TRPO). This algorithm is effective for optimizing large nonlinear policies such as neural networks. Our experiments demonstrate its robust performance on a wide variety of tasks: learning simulated robotic swimming, hopping, and walking gaits; and playing Atari games using images of the screen as input. Despite its approximations that deviate from the theory, TRPO tends to give monotonic improvement, with little tuning of hyperparameters."
                },
                {
                    "title": "Learning Laplacian Matrix in Smooth Graph Signal Representations",
                    "abstract": "The construction of a meaningful graph plays a crucial role in the success of many graph-based representations and algorithms for handling structured data, especially in the emerging field of graph signal processing. However, a meaningful graph is not always readily available from the data, nor easy to define depending on the application domain. In particular, it is often desirable in graph signal processing applications that a graph is chosen such that the data admit certain regularity or smoothness on the graph. In this paper, we address the problem of learning graph Laplacians, which is equivalent to learning graph topologies, such that the input data form graph signals with smooth variations on the resulting topology. To this end, we adopt a factor analysis model for the graph signals and impose a Gaussian probabilistic prior on the latent variables that control these signals. We show that the Gaussian prior leads to an efficient representation that favors the smoothness property of the graph signals. We then propose an algorithm for learning graphs that enforces such property and is based on minimizing the variations of the signals on the learned graph. Experiments on both synthetic and real world data demonstrate that the proposed graph learning framework can efficiently infer meaningful graph topologies from signal observations under the smoothness prior."
                },
                {
                    "title": "The role of centrality for the identification of influential spreaders in complex networks",
                    "abstract": "The identification of the most influential spreaders in networks is important to control and understand the spreading capabilities of the system as well as to ensure an efficient information diffusion such as in rumor-like dynamics. Recent works have suggested that the identification of influential spreaders is not independent of the dynamics being studied. For instance, the key disease spreaders might not necessarily be so when it comes to analyze social contagion or rumor propagation. Additionally, it has been shown that different metrics (degree, coreness, etc) might identify different influential nodes even for the same dynamical processes with diverse degree of accuracy. In this paper, we investigate how nine centrality measures correlate with the disease and rumor spreading capabilities of the nodes that made up different synthetic and real-world (both spatial and non-spatial) networks. We also propose a generalization of the random walk accessibility as a new centrality measure and derive analytical expressions for the latter measure for simple network configurations. Our results show that for non-spatial networks, the $k$-core and degree centralities are most correlated to epidemic spreading, whereas the average neighborhood degree, the closeness centrality and accessibility are most related to rumor dynamics. On the contrary, for spatial networks, the accessibility measure outperforms the rest of centrality metrics in almost all cases regardless of the kind of dynamics considered. Therefore, an important consequence of our analysis is that previous studies performed in synthetic random networks cannot be generalized to the case of spatial networks."
                },
                {
                    "title": "High-Resolution Measurements of Face-to-Face Contact Patterns in a Primary School",
                    "abstract": "Background Little quantitative information is available on the mixing patterns of children in school environments. Describing and understanding contacts between children at school would help quantify the transmission opportunities of respiratory infections and identify situations within schools where the risk of transmission is higher. We report on measurements carried out in a French school (6\u201312 years children), where we collected data on the time-resolved face-to-face proximity of children and teachers using a proximity-sensing infrastructure based on radio frequency identification devices. Methods and Findings Data on face-to-face interactions were collected on Thursday, October 1st and Friday, October 2nd 2009. We recorded 77,602 contact events between 242 individuals (232 children and 10 teachers). In this setting, each child has on average 323 contacts per day with 47 other children, leading to an average daily interaction time of 176 minutes. Most contacts are brief, but long contacts are also observed. Contacts occur mostly within each class, and each child spends on average three times more time in contact with classmates than with children of other classes. We describe the temporal evolution of the contact network and the trajectories followed by the children in the school, which constrain the contact patterns. We determine an exposure matrix aimed at informing mathematical models. This matrix exhibits a class and age structure which is very different from the homogeneous mixing hypothesis. Conclusions We report on important properties of the contact patterns between school children that are relevant for modeling the propagation of diseases and for evaluating control measures. We discuss public health implications related to the management of schools in case of epidemics and pandemics. Our results can help define a prioritization of control measures based on preventive measures, case isolation, classes and school closures, that could reduce the disruption to education during epidemics."
                },
                {
                    "title": "OpenStreetMap: User-Generated Street Maps",
                    "abstract": "The OpenStreetMap project is a knowledge collective that provides user-generated street maps. OSM follows the peer production model that created Wikipedia; its aim is to create a set of map data that's free to use, editable, and licensed under new copyright schemes. A considerable number of contributors edit the world map collaboratively using the OSM technical infrastructure, and a core group, estimated at approximately 40 volunteers, dedicate their time to creating and improving OSM's infrastructure, including maintaining the server, writing the core software that handles the transactions with the server, and creating cartographical outputs. There's also a growing community of software developers who develop software tools to make OSM data available for further use across different application domains, software platforms, and hardware devices. The OSM project's hub is the main OSM Web site."
                },
                {
                    "title": "Cost-effective outbreak detection in networks",
                    "abstract": "Given a water distribution network, where should we place sensors toquickly detect contaminants? Or, which blogs should we read to avoid missing important stories?.\n These seemingly different problems share common structure: Outbreak detection can be modeled as selecting nodes (sensor locations, blogs) in a network, in order to detect the spreading of a virus or information asquickly as possible. We present a general methodology for near optimal sensor placement in these and related problems. We demonstrate that many realistic outbreak detection objectives (e.g., detection likelihood, population affected) exhibit the property of \"submodularity\". We exploit submodularity to develop an efficient algorithm that scales to large problems, achieving near optimal placements, while being 700 times faster than a simple greedy algorithm. We also derive online bounds on the quality of the placements obtained by any algorithm. Our algorithms and bounds also handle cases where nodes (sensor locations, blogs) have different costs.\n We evaluate our approach on several large real-world problems,including a model of a water distribution network from the EPA, andreal blog data. The obtained sensor placements are provably near optimal, providing a constant fraction of the optimal solution. We show that the approach scales, achieving speedups and savings in storage of several orders of magnitude. We also show how the approach leads to deeper insights in both applications, answering multicriteria trade-off, cost-sensitivity and generalization questions."
                },
                {
                    "title": "Maximizing the spread of influence through a social network",
                    "abstract": "Models for the processes by which ideas and influence propagate through a social network have been studied in a number of domains, including the diffusion of medical and technological innovations, the sudden and widespread adoption of various strategies in game-theoretic settings, and the effects of \"word of mouth\" in the promotion of new products. Recently, motivated by the design of viral marketing strategies, Domingos and Richardson posed a fundamental algorithmic problem for such social network processes: if we can try to convince a subset of individuals to adopt a new product or innovation, and the goal is to trigger a large cascade of further adoptions, which set of individuals should we target?We consider this problem in several of the most widely studied models in social network analysis. The optimization problem of selecting the most influential nodes is NP-hard here, and we provide the first provable approximation guarantees for efficient algorithms. Using an analysis framework based on submodular functions, we show that a natural greedy strategy obtains a solution that is provably within 63% of optimal for several classes of models; our framework suggests a general approach for reasoning about the performance guarantees of algorithms for these types of influence problems in social networks.We also provide computational experiments on large collaboration networks, showing that in addition to their provable guarantees, our approximation algorithms significantly out-perform node-selection heuristics based on the well-studied notions of degree centrality and distance centrality from the field of social networks."
                },
                {
                    "title": "A contribution to the mathematical theory of epidemics",
                    "abstract": "(1) One of the most striking features in the study of epidemics is the difficulty of finding a causal factor which appears to be adequate to account for the magnitude of the frequent epidemics of disease which visit almost every population. It was with a view to obtaining more insight regarding the effects of the various factors which govern the spread of contagious epidemics that the present investigation was undertaken. Reference may here be made to the work of Ross and Hudson (1915-17) in which the same problem is attacked. The problem is here carried to a further stage, and it is considered from a point of view which is in one sense more general. The problem may be summarised as follows: One (or more) infected person is introduced into a community of individuals, more or less susceptible to the disease in question. The disease spreads from the affected to the unaffected by contact infection. Each infected person runs through the course of his sickness, and finally is removed from the number of those who are sick, by recovery or by death. The chances of recovery or death vary from day to day during the course of his illness. The chances that the affected may convey infection to the unaffected are likewise dependent upon the stage of the sickness. As the epidemic spreads, the number of unaffected members of the community becomes reduced. Since the course of an epidemic is short compared with the life of an individual, the population may be considered as remaining constant, except in as far as it is modified by deaths due to the epidemic disease itself. In the course of time the epidemic may come to an end. One of the most important probems in epidemiology is to ascertain whether this termination occurs only when no susceptible individuals are left, or whether the interplay of the various factors of infectivity, recovery and mortality, may result in termination, whilst many susceptible individuals are still present in the unaffected population. It is difficult to treat this problem in its most general aspect. In the present communication discussion will be limited to the case in which all members of the community are initially equally susceptible to the disease, and it will be further assumed that complete immunity is conferred by a single infection."
                },
                {
                    "title": "Identifying Propagation Sources in Networks: State-of-the-Art and Comparative Studies",
                    "abstract": "It has long been a significant but difficult problem to identify propagation sources based on limited knowledge of network structures and the varying states of network nodes. In practice, real cases can be locating the sources of rumors in online social networks and finding origins of a rolling blackout in smart grids. This paper reviews the state-of-the-art in source identification techniques and discusses the pros and cons of current methods in this field. Furthermore, in order to gain a quantitative understanding of current methods, we provide a series of experiments and comparisons based on various environment settings. Especially, our observation reveals considerable differences in performance by employing different network topologies, various propagation schemes, and diverse propagation probabilities. We therefore reach the following points for future work. First, current methods remain far from practice as their accuracy in terms of error distance ( ${\\delta}$ ) is normally larger than three in most scenarios. Second, the majority of current methods are too time consuming to quickly locate the origins of propagation. In addition, we list five open issues of current methods exposed by the analysis, from the perspectives of topology, number of sources, number of networks, temporal dynamics, and complexity and scalability. Solutions to these open issues are of great academic and practical significance."
                },
                {
                    "title": "Taking the Human Out of the Loop: A Review of Bayesian Optimization",
                    "abstract": "Big Data applications are typically associated with systems involving large numbers of users, massive complex software systems, and large-scale heterogeneous computing and storage architectures. The construction of such systems involves many distributed design choices. The end products (e.g., recommendation systems, medical analysis tools, real-time game engines, speech recognizers) thus involve many tunable configuration parameters. These parameters are often specified and hard-coded into the software by various developers or teams. If optimized jointly, these parameters can result in significant improvements. Bayesian optimization is a powerful tool for the joint optimization of design choices that is gaining great popularity in recent years. It promises greater automation so as to increase both product quality and human productivity. This review paper introduces Bayesian optimization, highlights some of its methodological aspects, and showcases a wide range of applications."
                },
                {
                    "title": "Emergence of Scaling in Random Networks",
                    "abstract": "level of Co and a maximum of inverse TMR is expected when the Fermi level of LSMO is approximately at the maximum of the spin2 DOS of Co. This is consistent with the maximum of inverse TMR observed at 20.4 V for Co/STO/LSMO junctions (Fig. 3A). For a positive bias, the TMR is expected to change sign and become normal above 1 V when the Fermi level of LSMO goes down into the energy range of the majority spin d-band of Co. This is also observed in Fig. 3A. For ALO and ALO/STO barriers, a predominant tunneling of s-character electrons (see arrow in Fig. 2B) is the usual explanation of the positive polarization (6\u20138). The rapid drop with bias (Fig. 3B) is similar to what has been observed in most junctions with ALO barriers, and completely different from what is obtained when the tunneling is predominantly by d-character electrons (Fig. 3A). The origin of this rapid decrease of the TMR at relatively small bias has never been clearly explained. This is roughly consistent with the energy dependence of the DOS induced by sp-d bonding effects on the first atomic layer of ALO in the calculation of Nguyen-Mahn et al. (8) for the Co-ALO interface. But Zhang et al. (13) have also shown that a large part of the TMR drop can be attributed to the excitation of spin waves. The experiments reported here and in several recent publications (3, 4) demonstrate the important role of the electronic structure of the metal-oxide interface in determining the spin polarization of the tunneling electrons. The negative polarization for the Co-STO interface has been ascribed to d-d bonding effects between Al and Ti (4). This interpretation is similar to that proposed to explain, in terms of sp-d bonding, the positive polarization at the Co-ALO interface (8). However, there is no general theory predicting the trend of the experimental results for Co\u2014that is, a negative polarization with oxides of d elements (STO, CLO, Ta2O5) and a positive one when there are only s and p states (ALO). It is likely that the spin polarization should also depend on the position of the Fermi level with respect to the electronic levels of each character above and below the gap of the insulator. In addition, as an evanescent wave in an insulator is a Bloch wave with an imaginary wave vector, one can expect different decay lengths for Bloch waves of different character. This means that the final polarization could also depend on the thickness of the barrier, as illustrated by the calculations of MacLaren et al. for Fe/ZnSe/Fe junctions (14). The influence of the barrier on the spin polarization opens new ways to shape and optimize the TMR. Interesting bias dependencies can be obtained with barriers selecting the d electrons and probing the fine structure of the d-DOS, as in Fig. 3A. The DOS of a d-band can also be easily tailored by alloying (for example, by introduction of virtual bound states) to produce specific bias dependencies. Although here we concentrated on the problem of the spin polarization of the Co electrode and regarded the strongly spin-polarized LSMO only as a useful spin analyzer, the large TMR ratios obtained by combining Co and LSMO electrodes (50% with a STO barrier) are also an interesting result. The drawback arising from the low Curie temperature of LSMO (;350 K) is the reduction of the TMR at room temperature, down to about 5% at 300 K in Co/STO/ LSMO (4). However, other types of oxides of the double-perovskite family (for example, Sr2FeMoO6) combine electronic properties similar to those of manganites with a definitely higher Curie temperature (15). Their use in magnetic tunnel junctions is promising for a new generation of tunnel junctions with very high magnetoresistance for room-temperature applications."
                },
                {
                    "title": "On the evolution of random graphs",
                    "abstract": "(n) k edges have equal probabilities to be chosen as the next one . We shall 2 study the \"evolution\" of such a random graph if N is increased . In this investigation we endeavour to find what is the \"typical\" structure at a given stage of evolution (i . e . if N is equal, or asymptotically equal, to a given function N(n) of n) . By a \"typical\" structure we mean such a structure the probability of which tends to 1 if n -* + when N = N(n) . If A is such a property that lim Pn,N,(n ) ( A) = 1, we shall say that \u201ealmost all\" graphs Gn,N(n) n--possess this property ."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we efficiently optimize the selection of node subsets in a partially observable graph to maximize a utility function defined on those subsets, particularly when the underlying function is a black box and expensive to evaluate?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem has significant implications for various fields, including transportation, social networks, and epidemiology. By developing methods to optimize node selection in these contexts, we can enhance product adoption in social networks, improve disease control strategies, and bolster the resilience of critical infrastructure. This research could lead to advancements in network-based diffusion literature and inspire new methodologies that can be applied across different domains, ultimately fostering a deeper understanding of complex systems and their dynamics.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from several factors: the discrete nature of the optimization space, which leads to a combinatorial explosion in the number of possible node subsets; the black-box nature of the utility functions, which complicates the evaluation process; and the incomplete knowledge of the graph structure, which necessitates incremental querying. Naive approaches, such as greedy algorithms, may fail due to their reliance on full graph knowledge and the computational burden of evaluating expectations over numerous simulations, making them impractical for large or partially observable networks.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on specific diffusion processes or required complete knowledge of the graph structure, limiting their applicability to broader scenarios. Existing methods often rely on proxy-based approaches that do not account for the underlying utility function, leading to suboptimal node selections. Additionally, the complexity of optimizing over a large search space with black-box functions has posed significant barriers. Our approach differs by proposing a novel strategy that operates in a \"combo-graph\" space, allowing for incremental revelation of the graph and optimization without prior complete knowledge.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves constructing a \"combo-graph\" where each node represents a k-node subset of the original graph. We will utilize a combination of adaptive querying techniques and optimization algorithms to evaluate the utility of these subsets efficiently. The dataset will consist of various network structures, and we will measure performance using metrics such as utility maximization and computational efficiency. We expect our approach to yield a significant improvement in the ability to identify optimal node subsets, even in scenarios"
            }
        },
        "author_data": {
            "eb96f209-46c7-46e3-8d49-d892d87c5535": {
                "pk": "eb96f209-46c7-46e3-8d49-d892d87c5535",
                "name": "Huidong Liang",
                "collaborators": [
                    "Junbin Gao",
                    "Xingjian Du",
                    "Bilei Zhu",
                    "Zejun Ma",
                    "Ke Chen",
                    "Zijie Wang",
                    "Xia Liang"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Link Prediction",
                    "Contrastive Learning",
                    "Music Information Retrieval"
                ],
                "publications": [
                    {
                        "title": "Wasserstein Adversarially Regularized Graph Autoencoder",
                        "abstract": "This paper introduces Wasserstein Adversarially Regularized Graph Autoencoder (WARGA), an implicit generative algorithm that directly regularizes the latent distribution of node embedding to a target distribution via the Wasserstein metric. The proposed method has been validated in tasks of link prediction and node clustering on real-world graphs, in which WARGA generally outperforms state-of-the-art models based on Kullback-Leibler (KL) divergence and typical adversarial framework."
                    },
                    {
                        "title": "How Neural Processes Improve Graph Link Prediction",
                        "abstract": "Link prediction is a fundamental problem in graph data analysis. While most of the literature focuses on transductive link prediction that requires all the graph nodes and majority of links in training, inductive link prediction, which only uses a proportion of the nodes and their links in training, is a more challenging problem in various real-world applications. In this paper, we propose a meta-learning approach with graph neural networks for link prediction: Neural Processes for Graph Neural Networks (NPGNN), which can perform both transductive and inductive learning tasks and adapt to patterns in a large new graph after training with a small subgraph. Experiments on real-world graphs are conducted to validate our model, where the results suggest that the proposed method achieves stronger performance compared to other state-of-the-art models, and meanwhile generalizes well when training on a small subgraph."
                    },
                    {
                        "title": "Graph Contrastive Learning with Implicit Augmentations",
                        "abstract": "Existing graph contrastive learning methods rely on augmentation techniques based on random perturbations (e.g., randomly adding or dropping edges and nodes). Nevertheless, altering certain edges or nodes can unexpectedly change the graph characteristics, and choosing the optimal perturbing ratio for each dataset requires onerous manual tuning. In this paper, we introduce Implicit Graph Contrastive Learning (iGCL), which utilizes augmentations in the latent space learned from a Variational Graph Auto-Encoder by reconstructing graph topological structure. Importantly, instead of explicitly sampling augmentations from latent distributions, we further propose an upper bound for the expected contrastive loss to improve the efficiency of our learning algorithm. Thus, graph semantics can be preserved within the augmentations in an intelligent way without arbitrary manual design or prior human knowledge. Experimental results on both graph-level and node-level tasks show that the proposed method achieves state-of-the-art performance compared to other benchmarks, where ablation studies in the end demonstrate the effectiveness of modules in iGCL."
                    },
                    {
                        "title": "ByteCover3: Accurate Cover Song Identification on Short Queries",
                        "abstract": "Deep learning based methods have become a paradigm for cover song identification (CSI) in recent years, where the ByteCover systems have achieved state-of-the-art results on all the mainstream datasets of CSI. However, with the burgeon of short videos, many real-world applications require matching short music excerpts to full-length music tracks in the database, which is still under-explored and waiting for an industrial-level solution. In this paper, we upgrade the previous ByteCover systems to ByteCover3 that utilizes local features to further improve the identification performance of short music queries. ByteCover3 is designed with a local alignment loss (LAL) module and a two-stage feature retrieval pipeline, allowing the system to perform CSI in a more precise and efficient way. We evaluated ByteCover3 on multiple datasets with different benchmark settings, where ByteCover3 beat all the compared methods including its previous versions."
                    }
                ]
            },
            "71f9e212-65bc-47a1-a9c8-7378b02fc383": {
                "pk": "71f9e212-65bc-47a1-a9c8-7378b02fc383",
                "name": "Xingchen Wan",
                "collaborators": [
                    "Binxin Ru",
                    "Xiaowen Dong",
                    "Michael Osborne",
                    "Diego Granziol",
                    "Han Zhou",
                    "Ivan Vuli\u0107",
                    "Anna Korhonen",
                    "Michael A. Osborne",
                    "Samuel Albanie",
                    "Stephen Roberts"
                ],
                "domain": [
                    "Neural Architecture Search",
                    "Bayesian Optimization",
                    "Natural Language Processing",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Interpretable Neural Architecture Search via Bayesian Optimisation with Weisfeiler-Lehman Kernels",
                        "abstract": "Current neural architecture search (NAS) strategies focus only on finding a single, good, architecture. They offer little insight into why a specific network is performing well, or how we should modify the architecture if we want further improvements. We propose a Bayesian optimisation (BO) approach for NAS that combines the Weisfeiler-Lehman graph kernel with a Gaussian process surrogate. Our method optimises the architecture in a highly data-efficient manner: it is capable of capturing the topological structures of the architectures and is scalable to large graphs, thus making the high-dimensional and graph-like search spaces amenable to BO. More importantly, our method affords interpretability by discovering useful network features and their corresponding impact on the network performance. Indeed, we demonstrate empirically that our surrogate model is capable of identifying useful motifs which can guide the generation of new architectures. We finally show that our method outperforms existing NAS approaches to achieve the state of the art on both closed- and open-domain search spaces."
                    },
                    {
                        "title": "Iterative Averaging in the Quest for Best Test Error",
                        "abstract": "We analyse and explain the increased generalisation performance of iterate averaging using a Gaussian process perturbation model between the true and batch risk surface on the high dimensional quadratic. We derive three phenomena \\latestEdits{from our theoretical results:} (1) The importance of combining iterate averaging (IA) with large learning rates and regularisation for improved regularisation. (2) Justification for less frequent averaging. (3) That we expect adaptive gradient methods to work equally well, or better, with iterate averaging than their non-adaptive counterparts. Inspired by these results\\latestEdits{, together with} empirical investigations of the importance of appropriate regularisation for the solution diversity of the iterates, we propose two adaptive algorithms with iterate averaging. These give significantly better results compared to stochastic gradient descent (SGD), require less tuning and do not require early stopping or validation set monitoring. We showcase the efficacy of our approach on the CIFAR-10/100, ImageNet and Penn Treebank datasets on a variety of modern and classical network architectures."
                    },
                    {
                        "title": "Deep Curvature Suite",
                        "abstract": "We present MLRG Deep Curvature suite, a PyTorch-based, open-source package for analysis and visualisation of neural network curvature and loss landscape. Despite of providing rich information into properties of neural network and useful for a various designed tasks, curvature information is still not made sufficient use for various reasons, and our method aims to bridge this gap. We present a primer, including its main practical desiderata and common misconceptions, of \\textit{Lanczos algorithm}, the theoretical backbone of our package, and present a series of examples based on synthetic toy examples and realistic modern neural networks tested on CIFAR datasets, and show the superiority of our package against existing competing approaches for the similar purposes."
                    },
                    {
                        "title": "Working Memory Capacity of ChatGPT: An Empirical Study",
                        "abstract": "Working memory is a critical aspect of both human intelligence and artificial intelligence, serving as a workspace for the temporary storage and manipulation of information. In this paper, we systematically assess the working memory capacity of ChatGPT, a large language model developed by OpenAI, by examining its performance in verbal and spatial n-back tasks under various conditions. Our experiments reveal that ChatGPT has a working memory capacity limit strikingly similar to that of humans. Furthermore, we investigate the impact of different instruction strategies on ChatGPT's performance and observe that the fundamental patterns of a capacity limit persist. From our empirical findings, we propose that n-back tasks may serve as tools for benchmarking the working memory capacity of large language models and hold potential for informing future efforts aimed at enhancing AI working memory."
                    },
                    {
                        "title": "AutoPEFT: Automatic Configuration Search for Parameter-Efficient Fine-Tuning",
                        "abstract": "Large pretrained language models are widely used in downstream NLP tasks via task-specific fine-tuning, but such procedures can be costly. Recently, Parameter-Efficient Fine-Tuning (PEFT) methods have achieved strong task performance while updating much fewer parameters than full model fine-tuning (FFT). However, it is non-trivial to make informed design choices on the PEFT configurations, such as their architecture, the number of tunable parameters, and even the layers in which the PEFT modules are inserted. Consequently, it is highly likely that the current, manually designed configurations are suboptimal in terms of their performance-efficiency trade-off. Inspired by advances in neural architecture search, we propose AutoPEFT for automatic PEFT configuration selection: we first design an expressive configuration search space with multiple representative PEFT modules as building blocks. Using multi-objective Bayesian optimisation in a low-cost setup, we then discover a Pareto-optimal set of configurations with strong performance-cost trade-offs across different numbers of parameters that are also highly transferable across different tasks. Empirically, on GLUE and SuperGLUE tasks, we show that AutoPEFT-discovered configurations significantly outperform existing PEFT methods and are on par or better than FFT without incurring substantial training efficiency costs."
                    },
                    {
                        "title": "Survival of the Most Influential Prompts: Efficient Black-Box Prompt Search via Clustering and Pruning",
                        "abstract": "Prompt-based learning has been an effective paradigm for large pretrained language models (LLM), enabling few-shot or even zero-shot learning. Black-box prompt search has received growing interest recently for its distinctive properties of gradient-free optimization, proven particularly useful and powerful for model-as-a-service usage. However, the discrete nature and the complexity of combinatorial optimization hinder the efficiency of modern black-box approaches. Despite extensive research on search algorithms, the crucial aspect of search space design and optimization has been largely overlooked. In this paper, we first conduct a sensitivity analysis by prompting LLM, revealing that only a small number of tokens exert a disproportionate amount of influence on LLM predictions. Leveraging this insight, we propose the Clustering and Pruning for Efficient Black-box Prompt Search (ClaPS), a simple black-box search method that first clusters and prunes the search space to focus exclusively on influential prompt tokens. By employing even simple search methods within the pruned search space, ClaPS achieves state-of-the-art performance across various tasks and LLMs, surpassing the performance of complex approaches while significantly reducing search costs. Our findings underscore the critical role of search space design and optimization in enhancing both the usefulness and the efficiency of black-box prompt-based learning."
                    },
                    {
                        "title": "On Redundancy and Diversity in Cell-based Neural Architecture Search",
                        "abstract": "Searching for the architecture cells is a dominant paradigm in NAS. However, little attention has been devoted to the analysis of the cell-based search spaces even though it is highly important for the continual development of NAS. In this work, we conduct an empirical post-hoc analysis of architectures from the popular cell-based search spaces and find that the existing search spaces contain a high degree of redundancy: the architecture performance is minimally sensitive to changes at large parts of the cells, and universally adopted designs, like the explicit search for a reduction cell, significantly increase the complexities but have very limited impact on the performance. Across architectures found by a diverse set of search strategies, we consistently find that the parts of the cells that do matter for architecture performance often follow similar and simple patterns. By explicitly constraining cells to include these patterns, randomly sampled architectures can match or even outperform the state of the art. These findings cast doubts into our ability to discover truly novel architectures in the existing cell-based search spaces, and inspire our suggestions for improvement to guide future NAS research. Code is available at https://github.com/xingchenwan/cell-based-NAS-analysis."
                    },
                    {
                        "title": "Teach Better or Show Smarter? On Instructions and Exemplars in Automatic Prompt Optimization",
                        "abstract": "Large language models have demonstrated remarkable capabilities, but their performance is heavily reliant on effective prompt engineering. Automatic prompt optimization (APO) methods are designed to automate this and can be broadly categorized into those targeting instructions (instruction optimization, IO) vs. those targeting exemplars (exemplar selection, ES). Despite their shared objective, these have evolved rather independently, with IO recently receiving more research attention. This paper seeks to bridge this gap by comprehensively comparing the performance of representative IO and ES techniques, both isolation and combination, on a diverse set of challenging tasks. Our findings reveal that intelligently reusing model-generated input-output pairs obtained from evaluating prompts on the validation set as exemplars consistently improves performance over IO methods but is currently under-investigated. We also find that despite the recent focus on IO, how we select exemplars can outweigh how we optimize instructions, with ES strategies as simple as random search outperforming state-of-the-art IO methods with seed instructions without any optimization. Moreover, we observe synergy between ES and IO, with optimal combinations surpassing individual contributions. We conclude that studying exemplar selection as a standalone method and its optimal combination with instruction optimization remains a crucial aspect of APO and deserves greater consideration in future research, even in the era of highly capable instruction-following models."
                    },
                    {
                        "title": "Bayesian Quadrature for Neural Ensemble Search",
                        "abstract": "Ensembling can improve the performance of Neural Networks, but existing approaches struggle when the architecture likelihood surface has dispersed, narrow peaks. Furthermore, existing methods construct equally weighted ensembles, and this is likely to be vulnerable to the failure modes of the weaker architectures. By viewing ensembling as approximately marginalising over architectures we construct ensembles using the tools of Bayesian Quadrature -- tools which are well suited to the exploration of likelihood surfaces with dispersed, narrow peaks. Additionally, the resulting ensembles consist of architectures weighted commensurate with their performance. We show empirically -- in terms of test likelihood, accuracy, and expected calibration error -- that our method outperforms state-of-the-art baselines, and verify via ablation studies that its components do so independently."
                    },
                    {
                        "title": "SimSC: A Simple Framework for Semantic Correspondence with Temperature Learning",
                        "abstract": "We propose SimSC, a remarkably simple framework, to address the problem of semantic matching only based on the feature backbone. We discover that when fine-tuning ImageNet pre-trained backbone on the semantic matching task, L2 normalization of the feature map, a standard procedure in feature matching, produces an overly smooth matching distribution and significantly hinders the fine-tuning process. By setting an appropriate temperature to the softmax, this over-smoothness can be alleviated and the quality of features can be substantially improved. We employ a learning module to predict the optimal temperature for fine-tuning feature backbones. This module is trained together with the backbone and the temperature is updated online. We evaluate our method on three public datasets and demonstrate that we can achieve accuracy on par with state-of-the-art methods under the same backbone without using a learned matching head. Our method is versatile and works on various types of backbones. We show that the accuracy of our framework can be easily improved by coupling it with more powerful backbones."
                    },
                    {
                        "title": "Think Global and Act Local: Bayesian Optimisation over High-Dimensional Categorical and Mixed Search Spaces",
                        "abstract": "High-dimensional black-box optimisation remains an important yet notoriously challenging problem. Despite the success of Bayesian optimisation methods on continuous domains, domains that are categorical, or that mix continuous and categorical variables, remain challenging. We propose a novel solution -- we combine local optimisation with a tailored kernel design, effectively handling high-dimensional categorical and mixed search spaces, whilst retaining sample efficiency. We further derive convergence guarantee for the proposed approach. Finally, we demonstrate empirically that our method outperforms the current baselines on a variety of synthetic and real-world tasks in terms of performance, computational costs, or both."
                    },
                    {
                        "title": "Bayesian Optimisation of Functions on Graphs",
                        "abstract": "The increasing availability of graph-structured data motivates the task of optimising over functions defined on the node set of graphs. Traditional graph search algorithms can be applied in this case, but they may be sample-inefficient and do not make use of information about the function values; on the other hand, Bayesian optimisation is a class of promising black-box solvers with superior sample efficiency, but it has been scarcely been applied to such novel setups. To fill this gap, we propose a novel Bayesian optimisation framework that optimises over functions defined on generic, large-scale and potentially unknown graphs. Through the learning of suitable kernels on graphs, our framework has the advantage of adapting to the behaviour of the target function. The local modelling approach further guarantees the efficiency of our method. Extensive experiments on both synthetic and real-world graphs demonstrate the effectiveness of the proposed optimisation framework."
                    },
                    {
                        "title": "Sentiment Correlation in Financial News Networks and Associated Market Movements",
                        "abstract": "In an increasingly connected global market, news sentiment towards one company may not only indicate its own market performance, but can also be associated with a broader movement on the sentiment and performance of other companies from the same or even different sectors. In this paper, we apply NLP techniques to understand news sentiment of 87 companies among the most reported on Reuters for a period of seven years. We investigate the propagation of such sentiment in company networks and evaluate the associated market movements in terms of stock price and volatility. Our results suggest that, in certain sectors, strong media sentiment towards one company may indicate a significant change in media sentiment towards related companies measured as neighbours in a financial network constructed from news co-occurrence. Furthermore, there exists a weak but statistically significant association between strong media sentiment and abnormal market return as well as volatility. Such an association is more significant at the level of individual companies, but nevertheless remains visible at the level of sectors or groups of companies."
                    }
                ]
            },
            "827a5c20-e47b-4ea2-ba88-94b576da2f25": {
                "pk": "827a5c20-e47b-4ea2-ba88-94b576da2f25",
                "name": "Xiaowen Dong",
                "collaborators": [
                    "Pascal Frossard",
                    "Dorina Thanou",
                    "Henry Kenlay",
                    "Michael Bronstein",
                    "Pierre Vandergheynst",
                    "Mihai Cucuringu",
                    "Kaige Yang",
                    "Laura Toni",
                    "Baskaran Sripathmanathan",
                    "Dimitrios Mavroeidis"
                ],
                "domain": [
                    "Graph Signal Processing",
                    "Machine Learning",
                    "Event Detection",
                    "Multilayer Graphs"
                ],
                "publications": [
                    {
                        "title": "Motivic Toda brackets",
                        "abstract": "We construct Toda brackets in unstable motivic homotopy theory and prove some fundamental properties of them. Furthermore we construct some examples of motivic Toda brackets."
                    },
                    {
                        "title": "On the Impact of Sample Size in Reconstructing Graph Signals",
                        "abstract": "Reconstructing a signal on a graph from observations on a subset of the vertices is a fundamental problem in the field of graph signal processing. It is often assumed that adding additional observations to an observation set will reduce the expected reconstruction error. We show that under the setting of noisy observation and least-squares reconstruction this is not always the case, characterising the behaviour both theoretically and experimentally."
                    },
                    {
                        "title": "Multiscale Event Detection in Social Media",
                        "abstract": "Event detection has been one of the most important research topics in social media analysis. Most of the traditional approaches detect events based on fixed temporal and spatial resolutions, while in reality events of different scales usually occur simultaneously, namely, they span different intervals in time and space. In this paper, we propose a novel approach towards multiscale event detection using social media data, which takes into account different temporal and spatial scales of events in the data. Specifically, we explore the properties of the wavelet transform, which is a well-developed multiscale transform in signal processing, to enable automatic handling of the interaction between temporal and spatial scales. We then propose a novel algorithm to compute a data similarity graph at appropriate scales and detect events of different scales simultaneously by a single graph-based clustering process. Furthermore, we present spatiotemporal statistical analysis of the noisy information present in the data stream, which allows us to define a novel term-filtering procedure for the proposed event detection algorithm and helps us study its behavior using simulated noisy data. Experimental results on both synthetically generated data and real world data collected from Twitter demonstrate the meaningfulness and effectiveness of the proposed approach. Our framework further extends to numerous application domains that involve multiscale and multiresolution data analysis."
                    },
                    {
                        "title": "Learning Laplacian Matrix in Smooth Graph Signal Representations",
                        "abstract": "The construction of a meaningful graph plays a crucial role in the success of many graph-based representations and algorithms for handling structured data, especially in the emerging field of graph signal processing. However, a meaningful graph is not always readily available from the data, nor easy to define depending on the application domain. In particular, it is often desirable in graph signal processing applications that a graph is chosen such that the data admit certain regularity or smoothness on the graph. In this paper, we address the problem of learning graph Laplacians, which is equivalent to learning graph topologies, such that the input data form graph signals with smooth variations on the resulting topology. To this end, we adopt a factor analysis model for the graph signals and impose a Gaussian probabilistic prior on the latent variables that control these signals. We show that the Gaussian prior leads to an efficient representation that favors the smoothness property of the graph signals. We then propose an algorithm for learning graphs that enforces such property and is based on minimizing the variations of the signals on the learned graph. Experiments on both synthetic and real world data demonstrate that the proposed graph learning framework can efficiently infer meaningful graph topologies from signal observations under the smoothness prior."
                    },
                    {
                        "title": "Clustering on Multi-Layer Graphs via Subspace Analysis on Grassmann Manifolds",
                        "abstract": "Relationships between entities in datasets are often of multiple nature, like geographical distance, social relationships, or common interests among people in a social network, for example. This information can naturally be modeled by a set of weighted and undirected graphs that form a global multilayer graph, where the common vertex set represents the entities and the edges on different layers capture the similarities of the entities in term of the different modalities. In this paper, we address the problem of analyzing multi-layer graphs and propose methods for clustering the vertices by efficiently merging the information provided by the multiple modalities. To this end, we propose to combine the characteristics of individual graph layers using tools from subspace analysis on a Grassmann manifold. The resulting combination can then be viewed as a low dimensional representation of the original data which preserves the most important information from diverse relationships between entities. We use this information in new clustering methods and test our algorithm on several synthetic and real world datasets where we demonstrate superior or competitive performances compared to baseline and state-of-the-art techniques. Our generic framework further extends to numerous analysis and learning problems that involve different types of information on graphs."
                    },
                    {
                        "title": "Multi-modal image retrieval with random walk on multi-layer graphs",
                        "abstract": "The analysis of large collections of image data is still a challenging problem due to the difficulty of capturing the true concepts in visual data. The similarity between images could be computed using different and possibly multimodal features such as color or edge information or even text labels. This motivates the design of image analysis solutions that are able to effectively integrate the multi-view information provided by different feature sets. We therefore propose a new image retrieval solution that is able to sort images through a random walk on a multi-layer graph, where each layer corresponds to a different type of information about the image data. We study in depth the design of the image graph and propose in particular an effective method to select the edge weights for the multi-layer graph, such that the image ranking scores are optimised. We then provide extensive experiments in different real-world photo collections, which confirm the high performance of our new image retrieval algorithm that generally surpasses state-of-the-art solutions due to a more meaningful image similarity computation."
                    },
                    {
                        "title": "Interpretable Stability Bounds for Spectral Graph Filters",
                        "abstract": "Graph-structured data arise in a variety of real-world context ranging from sensor and transportation to biological and social networks. As a ubiquitous tool to process graph-structured data, spectral graph filters have been used to solve common tasks such as denoising and anomaly detection, as well as design deep learning architectures such as graph neural networks. Despite being an important tool, there is a lack of theoretical understanding of the stability properties of spectral graph filters, which are important for designing robust machine learning models. In this paper, we study filter stability and provide a novel and interpretable upper bound on the change of filter output, where the bound is expressed in terms of the endpoint degrees of the deleted and newly added edges, as well as the spatial proximity of those edges. This upper bound allows us to reason, in terms of structural properties of the graph, when a spectral graph filter will be stable. We further perform extensive experiments to verify intuition that can be gained from the bound."
                    },
                    {
                        "title": "Interpretable Neural Architecture Search via Bayesian Optimisation with Weisfeiler-Lehman Kernels",
                        "abstract": "Current neural architecture search (NAS) strategies focus only on finding a single, good, architecture. They offer little insight into why a specific network is performing well, or how we should modify the architecture if we want further improvements. We propose a Bayesian optimisation (BO) approach for NAS that combines the Weisfeiler-Lehman graph kernel with a Gaussian process surrogate. Our method optimises the architecture in a highly data-efficient manner: it is capable of capturing the topological structures of the architectures and is scalable to large graphs, thus making the high-dimensional and graph-like search spaces amenable to BO. More importantly, our method affords interpretability by discovering useful network features and their corresponding impact on the network performance. Indeed, we demonstrate empirically that our surrogate model is capable of identifying useful motifs which can guide the generation of new architectures. We finally show that our method outperforms existing NAS approaches to achieve the state of the art on both closed- and open-domain search spaces."
                    },
                    {
                        "title": "On the Stability of Graph Convolutional Neural Networks under Edge Rewiring",
                        "abstract": "Graph neural networks are experiencing a surge of popularity within the machine learning community due to their ability to adapt to non-Euclidean domains and instil inductive biases. Despite this, their stability, i.e., their robustness to small perturbations in the input, is not yet well understood. Although there exists some results showing the stability of graph neural networks, most take the form of an upper bound on the magnitude of change due to a perturbation in the graph topology. However, the change in the graph topology captured in existing bounds tend not to be expressed in terms of structural properties, limiting our understanding of the model robustness properties. In this work, we develop an interpretable upper bound elucidating that graph neural networks are stable to rewiring between high degree nodes. This bound and further research in bounds of similar type provide further understanding of the stability properties of graph neural networks."
                    },
                    {
                        "title": "Understanding stock market instability via graph auto-encoders",
                        "abstract": "Understanding stock market instability is a key question in financial management as practitioners seek to forecast breakdowns in asset co-movements which expose portfolios to rapid and devastating collapses in value. The structure of these co-movements can be described as a graph where companies are represented by nodes and edges capture correlations between their price movements. Learning a timely indicator of co-movement breakdowns (manifested as modifications in the graph structure) is central in understanding both financial stability and volatility forecasting. We propose to use the edge reconstruction accuracy of a graph auto-encoder (GAE) as an indicator for how spatially homogeneous connections between assets are, which, based on financial network literature, we use as a proxy to infer market volatility. Our experiments on the S&P 500 over the 2015-2022 period show that higher GAE reconstruction error values are correlated with higher volatility. We also show that out-of-sample autoregressive modeling of volatility is improved by the addition of the proposed measure. Our paper contributes to the literature of machine learning in finance particularly in the context of understanding stock market instability."
                    },
                    {
                        "title": "Structure-Aware Robustness Certificates for Graph Classification",
                        "abstract": "Certifying the robustness of a graph-based machine learning model poses a critical challenge for safety. Current robustness certificates for graph classifiers guarantee output invariance with respect to the total number of node pair flips (edge addition or edge deletion), which amounts to an $l_{0}$ ball centred on the adjacency matrix. Although theoretically attractive, this type of isotropic structural noise can be too restrictive in practical scenarios where some node pairs are more critical than others in determining the classifier's output. The certificate, in this case, gives a pessimistic depiction of the robustness of the graph model. To tackle this issue, we develop a randomised smoothing method based on adding an anisotropic noise distribution to the input graph structure. We show that our process generates structural-aware certificates for our classifiers, whereby the magnitude of robustness certificates can vary across different pre-defined structures of the graph. We demonstrate the benefits of these certificates in both synthetic and real-world experiments."
                    },
                    {
                        "title": "Graph Neural Networks for Forecasting Multivariate Realized Volatility with Spillover Effects",
                        "abstract": "We present a novel methodology for modeling and forecasting multivariate realized volatilities using customized graph neural networks to incorporate spillover effects across stocks. The proposed model offers the benefits of incorporating spillover effects from multi-hop neighbors, capturing nonlinear relationships, and flexible training with different loss functions. Our empirical findings provide compelling evidence that incorporating spillover effects from multi-hop neighbors alone does not yield a clear advantage in terms of predictive accuracy. However, modeling nonlinear spillover effects enhances the forecasting accuracy of realized volatilities, particularly for short-term horizons of up to one week. Moreover, our results consistently indicate that training with the Quasi-likelihood loss leads to substantial improvements in model performance compared to the commonly-used mean squared error. A comprehensive series of empirical evaluations in alternative settings confirm the robustness of our results."
                    },
                    {
                        "title": "Hypergraph-MLP: Learning on Hypergraphs without Message Passing",
                        "abstract": "Hypergraphs are vital in modelling data with higher-order relations containing more than two entities, gaining prominence in machine learning and signal processing. Many hypergraph neural networks leverage message passing over hypergraph structures to enhance node representation learning, yielding impressive performances in tasks like hypergraph node classification. However, these message-passing-based models face several challenges, including oversmoothing as well as high latency and sensitivity to structural perturbations at inference time. To tackle those challenges, we propose an alternative approach where we integrate the information about hypergraph structures into training supervision without explicit message passing, thus also removing the reliance on it at inference. Specifically, we introduce Hypergraph-MLP, a novel learning framework for hypergraph-structured data, where the learning model is a straightforward multilayer perceptron (MLP) supervised by a loss function based on a notion of signal smoothness on hypergraphs. Experiments on hypergraph node classification tasks demonstrate that Hypergraph-MLP achieves competitive performance compared to existing baselines, and is considerably faster and more robust against structural perturbations at inference."
                    },
                    {
                        "title": "Maximum Likelihood Estimation on Stochastic Blockmodels for Directed Graph Clustering",
                        "abstract": "This paper studies the directed graph clustering problem through the lens of statistics, where we formulate clustering as estimating underlying communities in the directed stochastic block model (DSBM). We conduct the maximum likelihood estimation (MLE) on the DSBM and thereby ascertain the most probable community assignment given the observed graph structure. In addition to the statistical point of view, we further establish the equivalence between this MLE formulation and a novel flow optimization heuristic, which jointly considers two important directed graph statistics: edge density and edge orientation. Building on this new formulation of directed clustering, we introduce two efficient and interpretable directed clustering algorithms, a spectral clustering algorithm and a semidefinite programming based clustering algorithm. We provide a theoretical upper bound on the number of misclustered vertices of the spectral clustering algorithm using tools from matrix perturbation theory. We compare, both quantitatively and qualitatively, our proposed algorithms with existing directed clustering methods on both synthetic and real-world data, thus providing further ground to our theoretical contributions."
                    },
                    {
                        "title": "Bundle Neural Networks for message diffusion on graphs",
                        "abstract": "The dominant paradigm for learning on graph-structured data is message passing. Despite being a strong inductive bias, the local message passing mechanism suffers from pathological issues such as over-smoothing, over-squashing, and limited node-level expressivity. To address these limitations we propose Bundle Neural Networks (BuNN), a new type of GNN that operates via message diffusion over flat vector bundles - structures analogous to connections on Riemannian manifolds that augment the graph by assigning to each node a vector space and an orthogonal map. A BuNN layer evolves the features according to a diffusion-type partial differential equation. When discretized, BuNNs are a special case of Sheaf Neural Networks (SNNs), a recently proposed MPNN capable of mitigating over-smoothing. The continuous nature of message diffusion enables BuNNs to operate on larger scales of the graph and, therefore, to mitigate over-squashing. Finally, we prove that BuNN can approximate any feature transformation over nodes on any (potentially infinite) family of graphs given injective positional encodings, resulting in universal node-level expressivity. We support our theory via synthetic experiments and showcase the strong empirical performance of BuNNs over a range of real-world tasks, achieving state-of-the-art results on several standard benchmarks in transductive and inductive settings."
                    },
                    {
                        "title": "Separation Power of Equivariant Neural Networks",
                        "abstract": "The separation power of a machine learning model refers to its capacity to distinguish distinct inputs, and it is often employed as a proxy for its expressivity. In this paper, we propose a theoretical framework to investigate the separation power of equivariant neural networks with point-wise activations. Using the proposed framework, we can derive an explicit description of inputs indistinguishable by a family of neural networks with given architecture, demonstrating that it remains unaffected by the choice of non-polynomial activation function employed. We are able to understand the role played by activation functions in separability. Indeed, we show that all non-polynomial activations, such as ReLU and sigmoid, are equivalent in terms of expressivity, and that they reach maximum discrimination capacity. We demonstrate how assessing the separation power of an equivariant neural network can be simplified to evaluating the separation power of minimal representations. We conclude by illustrating how these minimal components form a hierarchy in separation power."
                    },
                    {
                        "title": "Error Analysis on Graph Laplacian Regularized Estimator",
                        "abstract": "We provide a theoretical analysis of the representation learning problem aimed at learning the latent variables (design matrix) $\\Theta$ of observations $Y$ with the knowledge of the coefficient matrix $X$. The design matrix is learned under the assumption that the latent variables $\\Theta$ are smooth with respect to a (known) topological structure $\\mathcal{G}$. To learn such latent variables, we study a graph Laplacian regularized estimator, which is the penalized least squares estimator with penalty term proportional to a Laplacian quadratic form. This type of estimators has recently received considerable attention due to its capability in incorporating underlying topological graph structure of variables into the learning process. While the estimation problem can be solved efficiently by state-of-the-art optimization techniques, its statistical consistency properties have been largely overlooked. In this work, we develop a non-asymptotic bound of estimation error under the classical statistical setting, where sample size is larger than the ambient dimension of the latent variables. This bound illustrates theoretically the impact of the alignment between the data and the graph structure as well as the graph spectrum on the estimation accuracy. It also provides theoretical evidence of the advantage, in terms of convergence rate, of the graph Laplacian regularized estimator over classical ones (that ignore the graph structure) in case of a smoothness prior. Finally, we provide empirical results of the estimation error to corroborate the theoretical analysis."
                    },
                    {
                        "title": "Laplacian-regularized graph bandits: Algorithms and theoretical analysis",
                        "abstract": "We consider a stochastic linear bandit problem with multiple users, where the relationship between users is captured by an underlying graph and user preferences are represented as smooth signals on the graph. We introduce a novel bandit algorithm where the smoothness prior is imposed via the random-walk graph Laplacian, which leads to a single-user cumulative regret scaling as $\\tilde{\\mathcal{O}}(\\Psi d \\sqrt{T})$ with time horizon $T$, feature dimensionality $d$, and the scalar parameter $\\Psi \\in (0,1)$ that depends on the graph connectivity. This is an improvement over $\\tilde{\\mathcal{O}}(d \\sqrt{T})$ in \\algo{LinUCB}~\\Ccite{li2010contextual}, where user relationship is not taken into account. In terms of network regret (sum of cumulative regret over $n$ users), the proposed algorithm leads to a scaling as $\\tilde{\\mathcal{O}}(\\Psi d\\sqrt{nT})$, which is a significant improvement over $\\tilde{\\mathcal{O}}(nd\\sqrt{T})$ in the state-of-the-art algorithm \\algo{Gob.Lin} \\Ccite{cesa2013gang}. To improve scalability, we further propose a simplified algorithm with a linear computational complexity with respect to the number of users, while maintaining the same regret. Finally, we present a finite-time analysis on the proposed algorithms, and demonstrate their advantage in comparison with state-of-the-art graph-based bandit algorithms on both synthetic and real-world data."
                    }
                ]
            }
        }
    },
    "2407.02880": {
        "paper_data": {
            "title": "Knowledge Composition using Task Vectors with Learned Anisotropic Scaling",
            "url": "http://arxiv.org/abs/2407.02880v1",
            "arxiv_id": "2407.02880",
            "authors": [
                "Frederic Z. Zhang",
                "Paul Albert",
                "Cristian Rodriguez-Opazo",
                "Anton van den Hengel",
                "Ehsan Abbasnejad"
            ],
            "abstract": "Pre-trained models produce strong generic representations that can be adapted via fine-tuning. The learned weight difference relative to the pre-trained model, known as a task vector, characterises the direction and stride of fine-tuning. The significance of task vectors is such that simple arithmetic operations on them can be used to combine diverse representations from different domains. This paper builds on these properties of task vectors and aims to answer (1) whether components of task vectors, particularly parameter blocks, exhibit similar characteristics, and (2) how such blocks can be used to enhance knowledge composition and transfer. To this end, we introduce aTLAS, an algorithm that linearly combines parameter blocks with different learned coefficients, resulting in anisotropic scaling at the task vector level. We show that such linear combinations explicitly exploit the low intrinsic dimensionality of pre-trained models, with only a few coefficients being the learnable parameters. Furthermore, composition of parameter blocks leverages the already learned representations, thereby reducing the dependency on large amounts of data. We demonstrate the effectiveness of our method in task arithmetic, few-shot recognition and test-time adaptation, with supervised or unsupervised objectives. In particular, we show that (1) learned anisotropic scaling allows task vectors to be more disentangled, causing less interference in composition; (2) task vector composition excels with scarce or no labeled data and is less prone to domain shift, thus leading to better generalisability; (3) mixing the most informative parameter blocks across different task vectors prior to training can reduce the memory footprint and improve the flexibility of knowledge transfer. Moreover, we show the potential of aTLAS as a PEFT method, particularly with less data, and demonstrate that its scalibility.",
            "introduction": " Introduction One practical advantage of neural networks is the fact that knowledge learned from a previous problem, in the form of network weights, can be transferred to solve other related problems. Commonly referred to as transfer learning [ 6,71], this technique is often applied when a model trained on a general-purpose dataset\u2014ImageNet [ 52] for many years\u2014is fine-tuned on other datasets to improve \u2217Equal contribution. Listed order was determined by a coin toss. Fred formalised the idea; developed the original codebase; conducted experiments using the publicly available codebase5. We train both LP++ and Tip-Adapter for 300 epochs on frozen zero-shot features. We apply a cosine annealing decay for Tip-Adapter and maintain fixed learning rates for LP++ as per the official implementation. D.2 linearised task vectors We report the average few-shot accuracy over the 22 datasets in Table 9, which corresponds to results in worse performance, most likely due to the relatively high number of parameters. 6https://pytorch.org/docs/stable/generated/torch.cuda.memory_allocated.html 28Table 17: Accuracy after fine-tuning on different percentage of training data for variants of aTLAS \u00d7Kand LoRAs [23]. methods to learn the coeffi- cients. We follow previous work on the combination of LoRAs [ 24] and use the nevergrad [ 49] library. We observe a memory usage reduction of 60% from 10GB to 4GB calculated using a dedicated pytorch function6. Appendix D.7. 9our method achieves higher performance with fewer learnable parameters. With sufficient training data, the variant aTLAS \u00d71200 leads to higher performance with a similar number of learnable parameters, as it is able to exploit the knowledge contained in the task vectors that may otherwise be unobtainable from the target dataset. 7 Results are averaged across 22 datasets. Highest accuracy in each section is highlighted in bold. Method # Params 1% 5% 10% 25% 35% 50% 100% aTLAS 2k 68.5 71.5 72.6 73.6 74.6 75.4 76.4 aTLAS\u00d75 10k 69.3 72.9 74.7 76.2 76.8 77.5 78.8 aTLAS\u00d720 40k 69.5 74.0 75.6 77.5 78.2 78.9 80.5 aTLAS\u00d780 160k 70.2 74.7 76.2 77.9 78.9 80.0 82.0 aTLAS\u00d71200 2.4M 71.3 75.0 76.6 78.3 80.2 81.5 83.9 LoRA (rank=16) 2.4M 68.8 74.1 75.6 76.8 79.0 80.6 83.6 E Unsupervised FixMatch We provide more details on the Unsupervised FixMatch (UFM) approach in this section. Fix- Match [ 54] utilises a labelled set to guide training, which is given as part of the semi-supervised learning protocol, while we produce a class-balanced \u201clabelled\u201d set from unlabelled images. Given a target dataset Dtconsisting of Nunlabelled images, we first rank the examples by the prediction scores from the zero-shot model across Cclasses. We then select the top min(N/C, 100) examples, that is, at most 100examples per class, as a trusted set in absence of a labelled set. The standard cross-entropy loss is applied to the trusted set. For the rest of the unlabelled images, we use a weakly augmented (Open-CLIP [ 26] validation augmentations) view of an image to produce pseudo-labels, and incur a loss on the strongly augmented view (Tip-Adapter [ 67] augmentations). Denote an image with weak augmentation by x, its strongly augmented view by x\u2032, and the predictions made by network by \u02c6yand\u02c6y\u2032, respectively, the unsupervised loss can be expressed as \u2113u(\u02c6y,\u02c6y\u2032) =\u2212 1(max(\u03c3(\u02c6y))> \u03c9)\u03c3(\u02c6y)Tlog(\u02c6y\u2032), (15) \u03c3(\u02c6y) =\u02c6y0.5 1T\u02c6y0.5, (16) where 1(\u00b7)denotes the indicator function, \u03c3(\u00b7)performs re-normalisation with adjusted temperature scaling, and \u03c9is a confidence threshold that is linearly adjusted from 0.9 to 1 during training. The trusted set",
            "references": [
                {
                    "title": "LP++: A Surprisingly Strong Linear Probe for Few-Shot CLIP",
                    "abstract": "In a recent, strongly emergent literature on few-shot CLIP adaptation, Linear Probe (LP) has been often reported as a weak baseline. This has motivated intensive research building convoluted prompt learning or feature adaptation strategies. In this work, we propose and exam-ine from convex-optimization perspectives a generalization of the standard LP baseline, in which the linear classifier weights are learnable functions of the text embedding, with class-wise multipliers blending image and text knowledge. As our objective function depends on two types of variables, i.e., the class visual prototypes and the learnable blending parameters, we propose a computationally efficient block coordinate Majorize-Minimize (MM) descent algorithm. In our full-batch MM optimizer, which we coin LP++, step sizes are implicit, unlike standard gradient descent practices where learning rates are intensively searched over validation sets. By examining the mathematical properties of our loss (e.g., Lipschitz gradient continuity), we build ma-jorizing functions yielding data-driven learning rates and derive approximations of the loss's minima, which provide data-informed initialization of the variables. Our image-language objective function, along with these non-trivial optimization insights and ingredients, yields, surprisingly, highly competitive few-shot CLIP performances. Furthermore, LP++ operates in black-box, relaxes intensive validation searches for the optimization hyper-parameters, and runs orders-of-magnitudes faster than state-of-the-art few-shot CLIP adaptation methods. Our code is available at: https://github.com/FereshteShakeri/FewShot-CLIP-Strong-Baseline.git."
                },
                {
                    "title": "Does Combining Parameter-efficient Modules Improve Few-shot Transfer Accuracy?",
                    "abstract": "Parameter-efficient fine-tuning stands as the standard for efficiently fine-tuning large language and vision models on downstream tasks. Specifically, the efficiency of low-rank adaptation has facilitated the creation and sharing of hundreds of custom LoRA modules, each trained on distinct data from various downstream tasks. In this paper, we explore the composability of LoRA modules, examining if combining these pre-trained modules enhances generalization to unseen downstream tasks. Our investigation involves evaluating two approaches: (a) uniform composition, involving averaging upstream LoRA modules with equal weights, and (b) learned composition, where we learn the weights for each upstream module and perform weighted averaging. Our experimental results on both vision and language models reveal that in few-shot settings, where only a limited number of samples are available for the downstream task, both uniform and learned composition methods result in better transfer accuracy; outperforming full fine-tuning and training a LoRA from scratch. Moreover, in full-shot settings, learned composition performs comparably to regular LoRA training with significantly fewer number of trainable parameters. Our research unveils the potential of uniform composition for enhancing transferability in low-shot settings, without introducing additional learnable parameters."
                },
                {
                    "title": "CLIP in Medical Imaging: A Comprehensive Survey",
                    "abstract": "Contrastive Language-Image Pre-training (CLIP), a simple yet effective pre-training paradigm, successfully introduces text supervision to vision models. It has shown promising results across various tasks, attributable to its generalizability and interpretability. The use of CLIP has recently gained increasing interest in the medical imaging domain, serving both as a pre-training paradigm for aligning medical vision and language, and as a critical component in diverse clinical tasks. With the aim of facilitating a deeper understanding of this promising direction, this survey offers an in-depth exploration of the CLIP paradigm within the domain of medical imaging, regarding both refined CLIP pre-training and CLIP-driven applications. In this study, We (1) start with a brief introduction to the fundamentals of CLIP methodology. (2) Then, we investigate the adaptation of CLIP pre-training in the medical domain, focusing on how to optimize CLIP given characteristics of medical images and reports. (3) Furthermore, we explore the practical utilization of CLIP pre-trained models in various tasks, including classification, dense prediction, and cross-modal tasks. (4) Finally, we discuss existing limitations of CLIP in the context of medical imaging and propose forward-looking directions to address the demands of medical imaging domain. We expect that this comprehensive survey will provide researchers in the field of medical image analysis with a holistic understanding of the CLIP paradigm and its potential implications. The project page can be found on https://github.com/zhaozh10/Awesome-CLIP-in-Medical-Imaging."
                },
                {
                    "title": "LoraHub: Efficient Cross-Task Generalization via Dynamic LoRA Composition",
                    "abstract": "Low-rank adaptations (LoRA) are often employed to fine-tune large language models (LLMs) for new tasks. This paper investigates LoRA composability for cross-task generalization and introduces LoraHub, a simple framework devised for the purposive assembly of LoRA modules trained on diverse given tasks, with the objective of achieving adaptable performance on unseen tasks. With just a few examples from a new task, LoraHub can fluidly combine multiple LoRA modules, eliminating the need for human expertise and assumptions. Notably, the composition requires neither additional model parameters nor gradients. Empirical results on the Big-Bench Hard benchmark suggest that LoraHub, while not surpassing the performance of in-context learning, offers a notable performance-efficiency trade-off in few-shot scenarios by employing a significantly reduced number of tokens per example during inference. Notably, LoraHub establishes a better upper bound compared to in-context learning when paired with different demonstration examples, demonstrating its potential for future development. Our vision is to establish a platform for LoRA modules, empowering users to share their trained LoRA modules. This collaborative approach facilitates the seamless application of LoRA modules to novel tasks, contributing to an adaptive ecosystem. Our code is available at https://github.com/sail-sg/lorahub, and all the pre-trained LoRA modules are released at https://huggingface.co/lorahub."
                },
                {
                    "title": "TIES-Merging: Resolving Interference When Merging Models",
                    "abstract": "Transfer learning - i.e., further fine-tuning a pre-trained model on a downstream task - can confer significant advantages, including improved downstream performance, faster convergence, and better sample efficiency. These advantages have led to a proliferation of task-specific fine-tuned models, which typically can only perform a single task and do not benefit from one another. Recently, model merging techniques have emerged as a solution to combine multiple task-specific models into a single multitask model without performing additional training. However, existing merging methods often ignore the interference between parameters of different models, resulting in large performance drops when merging multiple models. In this paper, we demonstrate that prior merging techniques inadvertently lose valuable information due to two major sources of interference: (a) interference due to redundant parameter values and (b) disagreement on the sign of a given parameter's values across models. To address this, we propose our method, TRIM, ELECT SIGN&MERGE (TIES-Merging), which introduces three novel steps when merging models: (1) resetting parameters that only changed a small amount during fine-tuning, (2) resolving sign conflicts, and (3) merging only the parameters that are in alignment with the final agreed-upon sign. We find that TIES-Merging outperforms several existing methods in diverse settings covering a range of modalities, domains, number of tasks, model sizes, architectures, and fine-tuning settings. We further analyze the impact of different types of interference on model parameters, and highlight the importance of resolving sign interference. Our code is available at https://github.com/prateeky2806/ties-merging"
                },
                {
                    "title": "S-CLIP: Semi-supervised Vision-Language Learning using Few Specialist Captions",
                    "abstract": "Vision-language models, such as contrastive language-image pre-training (CLIP), have demonstrated impressive results in natural image domains. However, these models often struggle when applied to specialized domains like remote sensing, and adapting to such domains is challenging due to the limited number of image-text pairs available for training. To address this, we propose S-CLIP, a semi-supervised learning method for training CLIP that utilizes additional unpaired images. S-CLIP employs two pseudo-labeling strategies specifically designed for contrastive learning and the language modality. The caption-level pseudo-label is given by a combination of captions of paired images, obtained by solving an optimal transport problem between unpaired and paired images. The keyword-level pseudo-label is given by a keyword in the caption of the nearest paired image, trained through partial label learning that assumes a candidate set of labels for supervision instead of the exact one. By combining these objectives, S-CLIP significantly enhances the training of CLIP using only a few image-text pairs, as demonstrated in various specialist domains, including remote sensing, fashion, scientific figures, and comics. For instance, S-CLIP improves CLIP by 10% for zero-shot classification and 4% for image-text retrieval on the remote sensing benchmark, matching the performance of supervised CLIP while using three times fewer image-text pairs."
                },
                {
                    "title": "Task Arithmetic in the Tangent Space: Improved Editing of Pre-Trained Models",
                    "abstract": "Task arithmetic has recently emerged as a cost-effective and scalable approach to edit pre-trained models directly in weight space: By adding the fine-tuned weights of different tasks, the model's performance can be improved on these tasks, while negating them leads to task forgetting. Yet, our understanding of the effectiveness of task arithmetic and its underlying principles remains limited. We present a comprehensive study of task arithmetic in vision-language models and show that weight disentanglement is the crucial factor that makes it effective. This property arises during pre-training and manifests when distinct directions in weight space govern separate, localized regions in function space associated with the tasks. Notably, we show that fine-tuning models in their tangent space by linearizing them amplifies weight disentanglement. This leads to substantial performance improvements across multiple task arithmetic benchmarks and diverse models. Building on these findings, we provide theoretical and empirical analyses of the neural tangent kernel (NTK) of these models and establish a compelling link between task arithmetic and the spatial localization of the NTK eigenfunctions. Overall, our work uncovers novel insights into the fundamental mechanisms of task arithmetic and offers a more reliable and effective approach to edit pre-trained models through the NTK linearization."
                },
                {
                    "title": "Towards Stable Test-Time Adaptation in Dynamic Wild World",
                    "abstract": "Test-time adaptation (TTA) has shown to be effective at tackling distribution shifts between training and testing data by adapting a given model on test samples. However, the online model updating of TTA may be unstable and this is often a key obstacle preventing existing TTA methods from being deployed in the real world. Specifically, TTA may fail to improve or even harm the model performance when test data have: 1) mixed distribution shifts, 2) small batch sizes, and 3) online imbalanced label distribution shifts, which are quite common in practice. In this paper, we investigate the unstable reasons and find that the batch norm layer is a crucial factor hindering TTA stability. Conversely, TTA can perform more stably with batch-agnostic norm layers, \\ie, group or layer norm. However, we observe that TTA with group and layer norms does not always succeed and still suffers many failure cases. By digging into the failure cases, we find that certain noisy test samples with large gradients may disturb the model adaption and result in collapsed trivial solutions, \\ie, assigning the same class label for all samples. To address the above collapse issue, we propose a sharpness-aware and reliable entropy minimization method, called SAR, for further stabilizing TTA from two aspects: 1) remove partial noisy samples with large gradients, 2) encourage model weights to go to a flat minimum so that the model is robust to the remaining noisy samples. Promising results demonstrate that SAR performs more stably over prior methods and is computationally efficient under the above wild test scenarios."
                },
                {
                    "title": "Editing Models with Task Arithmetic",
                    "abstract": "Changing how pre-trained models behave -- e.g., improving their performance on a downstream task or mitigating biases learned during pre-training -- is a common practice when developing machine learning systems. In this work, we propose a new paradigm for steering the behavior of neural networks, centered around \\textit{task vectors}. A task vector specifies a direction in the weight space of a pre-trained model, such that movement in that direction improves performance on the task. We build task vectors by subtracting the weights of a pre-trained model from the weights of the same model after fine-tuning on a task. We show that these task vectors can be modified and combined together through arithmetic operations such as negation and addition, and the behavior of the resulting model is steered accordingly. Negating a task vector decreases performance on the target task, with little change in model behavior on control tasks. Moreover, adding task vectors together can improve performance on multiple tasks at once. Finally, when tasks are linked by an analogy relationship of the form ``A is to B as C is to D\", combining task vectors from three of the tasks can improve performance on the fourth, even when no data from the fourth task is used for training. Overall, our experiments with several models, modalities and tasks show that task arithmetic is a simple, efficient and effective way of editing models."
                },
                {
                    "title": "ZegCLIP: Towards Adapting CLIP for Zero-shot Semantic Segmentation",
                    "abstract": "Recently, CLIP has been applied to pixel-level zero-shot learning tasks via a two-stage scheme. The general idea is to first generate class-agnostic region proposals and then feed the cropped proposal regions to CLIP to utilize its image-level zero-shot classification capability. While effective, such a scheme requires two image encoders, one for proposal generation and one for CLIP, leading to a complicated pipeline and high computational cost. In this work, we pursue a simpler-and-efficient one-stage solution that directly extends CLIP's zero-shot prediction capability from image to pixel level. Our investigation starts with a straightforward extension as our baseline that generates semantic masks by comparing the similarity between text and patch embeddings extracted from CLIP. However, such a paradigm could heavily overfit the seen classes and fail to generalize to unseen classes. To handle this issue, we propose three simple-but-effective designs and figure out that they can significantly retain the inherent zero-shot capacity of CLIP and improve pixel-level generalization ability. Incorporating those modifications leads to an efficient zero-shot semantic segmentation system called ZegCLIP. Through extensive experiments on three public benchmarks, ZegCLIP demonstrates superior performance, outperforming the state-of-the-art methods by a large margin under both \u201cinductive\u201d and \u201ctransductive\u201d zero-shot settings. In addition, compared with the two-stage method, our one-stage ZegCLIP achieves a speedup of about 5 times faster during inference. We release the code at https://github.com/ZiqinZhou66/ZegCLIP.git."
                },
                {
                    "title": "Patching open-vocabulary models by interpolating weights",
                    "abstract": "Open-vocabulary models like CLIP achieve high accuracy across many image classification tasks. However, there are still settings where their zero-shot performance is far from optimal. We study model patching, where the goal is to improve accuracy on specific tasks without degrading accuracy on tasks where performance is already adequate. Towards this goal, we introduce PAINT, a patching method that uses interpolations between the weights of a model before fine-tuning and the weights after fine-tuning on a task to be patched. On nine tasks where zero-shot CLIP performs poorly, PAINT increases accuracy by 15 to 60 percentage points while preserving accuracy on ImageNet within one percentage point of the zero-shot model. PAINT also allows a single model to be patched on multiple tasks and improves with model scale. Furthermore, we identify cases of broad transfer, where patching on one task increases accuracy on other tasks even when the tasks have disjoint classes. Finally, we investigate applications beyond common benchmarks such as counting or reducing the impact of typographic attacks on CLIP. Our findings demonstrate that it is possible to expand the set of tasks on which open-vocabulary models achieve high accuracy without re-training them from scratch."
                },
                {
                    "title": "Tip-Adapter: Training-free Adaption of CLIP for Few-shot Classification",
                    "abstract": "Contrastive Vision-Language Pre-training, known as CLIP, has provided a new paradigm for learning visual representations using large-scale image-text pairs. It shows impressive performance on downstream tasks by zero-shot knowledge transfer. To further enhance CLIP's adaption capability, existing methods proposed to fine-tune additional learnable modules, which significantly improves the few-shot performance but introduces extra training time and computational resources. In this paper, we propose a training-free adaption method for CLIP to conduct few-shot classification, termed as Tip-Adapter, which not only inherits the training-free advantage of zero-shot CLIP but also performs comparably to those training-required approaches. Tip-Adapter constructs the adapter via a key-value cache model from the few-shot training set, and updates the prior knowledge encoded in CLIP by feature retrieval. On top of that, the performance of Tip-Adapter can be further boosted to be state-of-the-art on ImageNet by fine-tuning the cache model for 10$\\times$ fewer epochs than existing methods, which is both effective and efficient. We conduct extensive experiments of few-shot classification on 11 datasets to demonstrate the superiority of our proposed methods. Code is released at https://github.com/gaopengcuhk/Tip-Adapter."
                },
                {
                    "title": "Contrastive Test-Time Adaptation",
                    "abstract": "Test-time adaptation is a special setting of unsupervised domain adaptation where a trained model on the source domain has to adapt to the target domain without accessing source data. We propose a novel way to leverage self-supervised contrastive learning to facilitate target feature learning, along with an online pseudo labeling scheme with refinement that significantly denoises pseudo labels. The contrastive learning task is applied jointly with pseudo labeling, contrasting positive and negative pairs constructed similarly as MoCo but with source-initialized encoder, and excluding same-class negative pairs indicated by pseudo labels. Meanwhile, we produce pseudo labels online and refine them via soft voting among their nearest neighbors in the target feature space, enabled by maintaining a memory queue. Our method, AdaContrast, achieves state-of-the-art performance on major benchmarks while having several desirable properties compared to existing works, including memory efficiency, insensitivity to hyper-parameters, and better model calibration. Code is released at https://github.com/DianCh/AdaContrast."
                },
                {
                    "title": "Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time",
                    "abstract": "The conventional recipe for maximizing model accuracy is to (1) train multiple models with various hyperparameters and (2) pick the individual model which performs best on a held-out validation set, discarding the remainder. In this paper, we revisit the second step of this procedure in the context of fine-tuning large pre-trained models, where fine-tuned models often appear to lie in a single low error basin. We show that averaging the weights of multiple models fine-tuned with different hyperparameter configurations often improves accuracy and robustness. Unlike a conventional ensemble, we may average many models without incurring any additional inference or memory costs -- we call the results\"model soups.\"When fine-tuning large pre-trained models such as CLIP, ALIGN, and a ViT-G pre-trained on JFT, our soup recipe provides significant improvements over the best model in a hyperparameter sweep on ImageNet. The resulting ViT-G model, which attains 90.94% top-1 accuracy on ImageNet, achieved a new state of the art. Furthermore, we show that the model soup approach extends to multiple image classification and natural language processing tasks, improves out-of-distribution performance, and improves zero-shot performance on new downstream tasks. Finally, we analytically relate the performance similarity of weight-averaging and logit-ensembling to flatness of the loss and confidence of the predictions, and validate this relation empirically. Code is available at https://github.com/mlfoundations/model-soups."
                },
                {
                    "title": "Efficient Two-Stage Detection of Human-Object Interactions with a Novel Unary-Pairwise Transformer",
                    "abstract": "Recent developments in transformer models for visual data have led to significant improvements in recognition and detection tasks. In particular, using learnable queries in place of region proposals has given rise to a new class of one-stage detection models, spearheaded by the Detection Transformer (DETR). Variations on this one-stage approach have since dominated human-object interaction (HOI) detection. However, the success of such one-stage HOI detectors can largely be attributed to the representation power of transformers. We discovered that when equipped with the same transformer, their two-stage counterparts can be more performant and memory-efficient, while taking a fraction of the time to train. In this work, we propose the Unary-Pairwise Transformer, a two-stage detector that exploits unary and pairwise representations for HOIs. We observe that the unary and pairwise parts of our transformer network specialise, with the former preferentially increasing the scores of positive examples and the latter decreasing the scores of negative examples. We evaluate our method on the HICO-DET and V-COCO datasets, and significantly outperform state-of-the-art approaches. At inference time, our model with ResNet50 approaches real-time performance on a single GPU."
                },
                {
                    "title": "On the Opportunities and Risks of Foundation Models",
                    "abstract": "AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Intrinsic Dimensionality Explains the Effectiveness of Language Model Fine-Tuning",
                    "abstract": "Although pretrained language models can be fine-tuned to produce state-of-the-art results for a very wide range of language understanding tasks, the dynamics of this process are not well understood, especially in the low data regime. Why can we use relatively vanilla gradient descent algorithms (e.g., without strong regularization) to tune a model with hundreds of millions of parameters on datasets with only hundreds or thousands of labeled examples? In this paper, we argue that analyzing fine-tuning through the lens of intrinsic dimension provides us with empirical and theoretical intuitions to explain this remarkable phenomenon. We empirically show that common pre-trained models have a very low intrinsic dimension; in other words, there exists a low dimension reparameterization that is as effective for fine-tuning as the full parameter space. For example, by optimizing only 200 trainable parameters randomly projected back into the full space, we can tune a RoBERTa model to achieve 90% of the full parameter performance levels on MRPC. Furthermore, we empirically show that pre-training implicitly minimizes intrinsic dimension and, perhaps surprisingly, larger models tend to have lower intrinsic dimension after a fixed number of pre-training updates, at least in part explaining their extreme effectiveness. Lastly, we connect intrinsic dimensionality with low dimensional task representations and compression based generalization bounds to provide intrinsic-dimension-based generalization bounds that are independent of the full parameter count."
                },
                {
                    "title": "Spatially Conditioned Graphs for Detecting Human\u2013Object Interactions",
                    "abstract": "We address the problem of detecting human\u2013object interactions in images using graphical neural networks. Unlike conventional methods, where nodes send scaled but otherwise identical messages to each of their neighbours, we propose to condition messages between pairs of nodes on their spatial relationships, resulting in different messages going to neighbours of the same node. To this end, we explore various ways of applying spatial conditioning under a multi-branch structure. Through extensive experimentation we demonstrate the advantages of spatial conditioning for the computation of the adjacency structure, messages and the refined graph features. In particular, we empirically show that as the quality of the bounding boxes increases, their coarse appearance features contribute relatively less to the disambiguation of interactions compared to the spatial information. Our method achieves an mAP of 31.33% on HICO-DET and 54.2% on V-COCO, significantly outperforming state-of-the-art on fine-tuned detections."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization",
                    "abstract": "We introduce four new real-world distribution shift datasets consisting of changes in image style, image blurriness, geographic location, camera operation, and more. With our new datasets, we take stock of previously proposed methods for improving out-of-distribution robustness and put them to the test. We find that using larger models and artificial data augmentations can improve robustness on real-world distribution shifts, contrary to claims in prior work. We find improvements in artificial robustness benchmarks can transfer to real-world distribution shifts, contrary to claims in prior work. Motivated by our observation that data augmentations can help with real-world distribution shifts, we also introduce a new data augmentation method which advances the state-of-the-art and outperforms models pre-trained with 1000\u00d7 more labeled data. Overall we find that some methods consistently help with distribution shifts in texture and local image statistics, but these methods do not help with some other distribution shifts like geographic changes. Our results show that future research must study multiple distribution shifts simultaneously, as we demonstrate that no evaluated method consistently improves robustness."
                },
                {
                    "title": "Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation",
                    "abstract": "Unsupervised domain adaptation (UDA) aims to leverage the knowledge learned from a labeled source dataset to solve similar tasks in a new unlabeled domain. Prior UDA methods typically require to access the source data when learning to adapt the model, making them risky and inefficient for decentralized private data. This work tackles a practical setting where only a trained source model is available and investigates how we can effectively utilize such a model without source data to solve UDA problems. We propose a simple yet generic representation learning framework, named \\emph{Source HypOthesis Transfer} (SHOT). SHOT freezes the classifier module (hypothesis) of the source model and learns the target-specific feature extraction module by exploiting both information maximization and self-supervised pseudo-labeling to implicitly align representations from the target domains to the source hypothesis. To verify its versatility, we evaluate SHOT in a variety of adaptation cases including closed-set, partial-set, and open-set domain adaptation. Experiments indicate that SHOT yields state-of-the-art results among multiple domain adaptation benchmarks."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence",
                    "abstract": "Semi-supervised learning (SSL) provides an effective means of leveraging unlabeled data to improve a model's performance. In this paper, we demonstrate the power of a simple combination of two common SSL methods: consistency regularization and pseudo-labeling. Our algorithm, FixMatch, first generates pseudo-labels using the model's predictions on weakly-augmented unlabeled images. For a given image, the pseudo-label is only retained if the model produces a high-confidence prediction. The model is then trained to predict the pseudo-label when fed a strongly-augmented version of the same image. Despite its simplicity, we show that FixMatch achieves state-of-the-art performance across a variety of standard semi-supervised learning benchmarks, including 94.93% accuracy on CIFAR-10 with 250 labels and 88.61% accuracy with 40 -- just 4 labels per class. Since FixMatch bears many similarities to existing SSL methods that achieve worse performance, we carry out an extensive ablation study to tease apart the experimental factors that are most important to FixMatch's success. We make our code available at this https URL."
                },
                {
                    "title": "A Comprehensive Survey on Transfer Learning",
                    "abstract": "Transfer learning aims at improving the performance of target learners on target domains by transferring the knowledge contained in different but related source domains. In this way, the dependence on a large number of target-domain data can be reduced for constructing target learners. Due to the wide application prospects, transfer learning has become a popular and promising area in machine learning. Although there are already some valuable and impressive surveys on transfer learning, these surveys introduce approaches in a relatively isolated way and lack the recent advances in transfer learning. Due to the rapid expansion of the transfer learning area, it is both necessary and challenging to comprehensively review the relevant studies. This survey attempts to connect and systematize the existing transfer learning research studies, as well as to summarize and interpret the mechanisms and the strategies of transfer learning in a comprehensive way, which may help readers have a better understanding of the current research status and ideas. Unlike previous surveys, this survey article reviews more than 40 representative transfer learning approaches, especially homogeneous transfer learning approaches, from the perspectives of data and model. The applications of transfer learning are also briefly introduced. In order to show the performance of different transfer learning models, over 20 representative transfer learning models are used for experiments. The models are performed on three different data sets, that is, Amazon Reviews, Reuters-21578, and Office-31, and the experimental results demonstrate the importance of selecting appropriate transfer learning models for different applications in practice."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "Test-Time Training with Self-Supervision for Generalization under Distribution Shifts",
                    "abstract": "In this paper, we propose Test-Time Training, a general approach for improving the performance of predictive models when training and test data come from different distributions. We turn a single unlabeled test sample into a self-supervised learning problem, on which we update the model parameters before making a prediction. This also extends naturally to data in an online stream. Our simple approach leads to improvements on diverse image classification benchmarks aimed at evaluating robustness to distribution shifts."
                },
                {
                    "title": "Pseudo-Labeling and Confirmation Bias in Deep Semi-Supervised Learning",
                    "abstract": "Semi-supervised learning, i.e. jointly learning from labeled and unlabeled samples, is an active research topic due to its key role on relaxing human supervision. In the context of image classification, recent advances to learn from unlabeled samples are mainly focused on consistency regularization methods that encourage invariant predictions for different perturbations of unlabeled samples. We, conversely, propose to learn from unlabeled data by generating soft pseudo-labels using the network predictions. We show that a naive pseudo-labeling overfits to incorrect pseudo-labels due to the so-called confirmation bias and demonstrate that mixup augmentation and setting a minimum number of labeled samples per mini-batch are effective regularization techniques for reducing it. The proposed approach achieves state-of-the-art results in CIFAR-10/100, SVHN, and Mini-ImageNet despite being much simpler than other methods. These results demonstrate that pseudo-labeling alone can outperform consistency regularization methods, while the opposite was supposed in previous work. Source code is available at https://git.io/fjQsC."
                },
                {
                    "title": "Natural Adversarial Examples",
                    "abstract": "We introduce two challenging datasets that reliably cause machine learning model performance to substantially degrade. The datasets are collected with a simple adversarial filtration technique to create datasets with limited spurious cues. Our datasets\u2019 real-world, unmodified examples transfer to various unseen models reliably, demonstrating that computer vision models have shared weaknesses. The first dataset is called IMAGENET-A and is like the ImageNet test set, but it is far more challenging for existing models. We also curate an adversarial out-of-distribution detection dataset called IMAGENET-O, which is the first out-of-distribution detection dataset created for ImageNet models. On IMAGENET-A a DenseNet-121 obtains around 2% accuracy, an accuracy drop of approximately 90%, and its out-of-distribution detection performance on IMAGENET-O is near random chance levels. We find that existing data augmentation techniques hardly boost performance, and using other public training datasets provides improvements that are limited. However, we find that improvements to computer vision architectures provide a promising path towards robust models."
                },
                {
                    "title": "Similarity of Neural Network Representations Revisited",
                    "abstract": "Recent work has sought to understand the behavior of neural networks by comparing representations between layers and between different trained models. We examine methods for comparing neural network representations based on canonical correlation analysis (CCA). We show that CCA belongs to a family of statistics for measuring multivariate similarity, but that neither CCA nor any other statistic that is invariant to invertible linear transformation can measure meaningful similarities between representations of higher dimension than the number of data points. We introduce a similarity index that measures the relationship between representational similarity matrices and does not suffer from this limitation. This similarity index is equivalent to centered kernel alignment (CKA) and is also closely connected to CCA. Unlike CCA, CKA can reliably identify correspondences between representations in networks trained from different initializations."
                },
                {
                    "title": "Learning Robust Global Representations by Penalizing Local Predictive Power",
                    "abstract": "Despite their renowned predictive power on i.i.d. data, convolutional neural networks are known to rely more on high-frequency patterns that humans deem superficial than on low-frequency patterns that agree better with intuitions about what constitutes category membership. This paper proposes a method for training robust convolutional networks by penalizing the predictive power of the local representations learned by earlier layers. Intuitively, our networks are forced to discard predictive signals such as color and texture that can be gleaned from local receptive fields and to rely instead on the global structures of the image. Across a battery of synthetic and benchmark domain adaptation tasks, our method confers improved generalization out of the domain. Also, to evaluate cross-domain transfer, we introduce ImageNet-Sketch, a new dataset consisting of sketch-like images, that matches the ImageNet classification validation set in categories and scale."
                },
                {
                    "title": "Do ImageNet Classifiers Generalize to ImageNet?",
                    "abstract": "We build new test sets for the CIFAR-10 and ImageNet datasets. Both benchmarks have been the focus of intense research for almost a decade, raising the danger of overfitting to excessively re-used test sets. By closely following the original dataset creation processes, we test to what extent current classification models generalize to new data. We evaluate a broad range of models and find accuracy drops of 3% - 15% on CIFAR-10 and 11% - 14% on ImageNet. However, accuracy gains on the original test sets translate to larger gains on the new test sets. Our results suggest that the accuracy drops are not caused by adaptivity, but by the models' inability to generalize to slightly \"harder\" images than those found in the original test sets."
                },
                {
                    "title": "Introducing Eurosat: A Novel Dataset and Deep Learning Benchmark for Land Use and Land Cover Classification",
                    "abstract": "In this paper, we address the challenge of land use and land cover classification using Sentinel-2 satellite images. The key contributions are as follows. We present a novel dataset based on Sentinel-2 satellite images covering 13 different spectral bands and consisting of 10 classes with in total 27,000 labeled images. We evaluate state-of-the-art deep Convolutional Neural Networks (CNNs) on this novel dataset with its different spectral bands. We also evaluate deep CNNs on existing remote sensing datasets and compare the obtained results. With the proposed novel dataset, we achieved an overall classification accuracy of 98.57%. The classification system resulting from the proposed research opens a gate towards various Earth observation applications. We demonstrate how the classification system can assist in improving geographical maps."
                },
                {
                    "title": "Measuring the Intrinsic Dimension of Objective Landscapes",
                    "abstract": "Many recently trained neural networks employ large numbers of parameters to achieve good performance. One may intuitively use the number of parameters required as a rough gauge of the difficulty of a problem. But how accurate are such notions? How many parameters are really needed? In this paper we attempt to answer this question by training networks not in their native parameter space, but instead in a smaller, randomly oriented subspace. We slowly increase the dimension of this subspace, note at which dimension solutions first appear, and define this to be the intrinsic dimension of the objective landscape. The approach is simple to implement, computationally tractable, and produces several suggestive conclusions. Many problems have smaller intrinsic dimensions than one might suspect, and the intrinsic dimension for a given dataset varies little across a family of models with vastly different sizes. This latter result has the profound implication that once a parameter space is large enough to solve a problem, extra parameters serve directly to increase the dimensionality of the solution manifold. Intrinsic dimension allows some quantitative comparison of problem difficulty across supervised, reinforcement, and other types of learning where we conclude, for example, that solving the inverted pendulum problem is 100 times easier than classifying digits from MNIST, and playing Atari Pong from pixels is about as hard as classifying CIFAR-10. In addition to providing new cartography of the objective landscapes wandered by parameterized models, the method is a simple technique for constructively obtaining an upper bound on the minimum description length of a solution. A byproduct of this construction is a simple approach for compressing networks, in some cases by more than 100 times."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Mask R-CNN",
                    "abstract": "We present a conceptually simple, flexible, and general framework for object instance segmentation. Our approach efficiently detects objects in an image while simultaneously generating a high-quality segmentation mask for each instance. The method, called Mask R-CNN, extends Faster R-CNN by adding a branch for predicting an object mask in parallel with the existing branch for bounding box recognition. Mask R-CNN is simple to train and adds only a small overhead to Faster R-CNN, running at 5 fps. Moreover, Mask R-CNN is easy to generalize to other tasks, e.g., allowing us to estimate human poses in the same framework. We show top results in all three tracks of the COCO suite of challenges, including instance segmentation, bounding-box object detection, and person keypoint detection. Without tricks, Mask R-CNN outperforms all existing, single-model entries on every task, including the COCO 2016 challenge winners. We hope our simple and effective approach will serve as a solid baseline and help ease future research in instance-level recognition. Code will be made available."
                },
                {
                    "title": "Remote Sensing Image Scene Classification: Benchmark and State of the Art",
                    "abstract": "Remote sensing image scene classification plays an important role in a wide range of applications and hence has been receiving remarkable attention. During the past years, significant efforts have been made to develop various data sets or present a variety of approaches for scene classification from remote sensing images. However, a systematic review of the literature concerning data sets and methods for scene classification is still lacking. In addition, almost all existing data sets have a number of limitations, including the small scale of scene classes and the image numbers, the lack of image variations and diversity, and the saturation of accuracy. These limitations severely limit the development of new approaches especially deep learning-based methods. This paper first provides a comprehensive review of the recent progress. Then, we propose a large-scale data set, termed \u201cNWPU-RESISC45,\u201d which is a publicly available benchmark for REmote Sensing Image Scene Classification (RESISC), created by Northwestern Polytechnical University (NWPU). This data set contains 31 500 images, covering 45 scene classes with 700 images in each class. The proposed NWPU-RESISC45 1) is large-scale on the scene classes and the total image number; 2) holds big variations in translation, spatial resolution, viewpoint, object pose, illumination, background, and occlusion; and 3) has high within-class diversity and between-class similarity. The creation of this data set will enable the community to develop and evaluate various data-driven algorithms. Finally, several representative methods are evaluated using the proposed data set, and the results are reported as a useful baseline for future research."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks",
                    "abstract": "State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available"
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "3D Object Representations for Fine-Grained Categorization",
                    "abstract": "While 3D object representations are being revived in the context of multi-view object class detection and scene understanding, they have not yet attained wide-spread use in fine-grained categorization. State-of-the-art approaches achieve remarkable performance when training data is plentiful, but they are typically tied to flat, 2D representations that model objects as a collection of unconnected views, limiting their ability to generalize across viewpoints. In this paper, we therefore lift two state-of-the-art 2D object representations to 3D, on the level of both local feature appearance and location. In extensive experiments on existing and newly proposed datasets, we show our 3D object representations outperform their state-of-the-art 2D counterparts for fine-grained categorization and demonstrate their efficacy for estimating 3D geometry from images via ultra-wide baseline matching and 3D reconstruction."
                },
                {
                    "title": "Describing Textures in the Wild",
                    "abstract": "Patterns and textures are key characteristics of many natural objects: a shirt can be striped, the wings of a butterfly can be veined, and the skin of an animal can be scaly. Aiming at supporting this dimension in image understanding, we address the problem of describing textures with semantic attributes. We identify a vocabulary of forty-seven texture terms and use them to describe a large dataset of patterns collected \"in the wild\". The resulting Describable Textures Dataset (DTD) is a basis to seek the best representation for recognizing describable texture attributes in images. We port from object recognition to texture recognition the Improved Fisher Vector (IFV) and Deep Convolutional-network Activation Features (DeCAF), and show that surprisingly, they both outperform specialized texture descriptors not only on our problem, but also in established material recognition datasets. We also show that our describable attributes are excellent texture descriptors, transferring between datasets and tasks, in particular, combined with IFV and DeCAF, they significantly outperform the state-of-the-art by more than 10% on both FMD and KTH-TIPS-2b benchmarks. We also demonstrate that they produce intuitive descriptions of materials and Internet images."
                },
                {
                    "title": "Fine-Grained Visual Classification of Aircraft",
                    "abstract": "This paper introduces FGVC-Aircraft, a new dataset containing 10,000 images of aircraft spanning 100 aircraft models, organised in a three-level hierarchy. At the finer level, differences between models are often subtle but always visually measurable, making visual recognition challenging but possible. A benchmark is obtained by defining corresponding classification tasks and evaluation protocols, and baseline results are presented. The construction of this dataset was made possible by the work of aircraft enthusiasts, a strategy that can extend to the study of number of other object classes. Compared to the domains usually considered in fine-grained visual classification (FGVC), for example animals, aircraft are rigid and hence less deformable. They, however, present other interesting modes of variation, including purpose, size, designation, structure, historical style, and branding."
                },
                {
                    "title": "UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild",
                    "abstract": "We introduce UCF101 which is currently the largest dataset of human actions. It consists of 101 action classes, over 13k clips and 27 hours of video data. The database consists of realistic user uploaded videos containing camera motion and cluttered background. Additionally, we provide baseline action recognition results on this new dataset using standard bag of words approach with overall performance of 44.5%. To the best of our knowledge, UCF101 is currently the most challenging dataset of actions due to its large number of classes, large number of clips and also unconstrained nature of such clips."
                },
                {
                    "title": "Cats and dogs",
                    "abstract": "We investigate the fine grained object categorization problem of determining the breed of animal from an image. To this end we introduce a new annotated dataset of pets covering 37 different breeds of cats and dogs. The visual problem is very challenging as these animals, particularly cats, are very deformable and there can be quite subtle differences between the breeds. We make a number of contributions: first, we introduce a model to classify a pet breed automatically from an image. The model combines shape, captured by a deformable part model detecting the pet face, and appearance, captured by a bag-of-words model that describes the pet fur. Fitting the model involves automatically segmenting the animal in the image. Second, we compare two classification approaches: a hierarchical one, in which a pet is first assigned to the cat or dog family and then to a breed, and a flat one, in which the breed is obtained directly. We also investigate a number of animal and image orientated spatial layouts. These models are very good: they beat all previously published results on the challenging ASIRRA test (cat vs dog discrimination). When applied to the task of discriminating the 37 different breeds of pets, the models obtain an average accuracy of about 59%, a very encouraging result considering the difficulty of the problem."
                },
                {
                    "title": "An Analysis of Single-Layer Networks in Unsupervised Feature Learning",
                    "abstract": "A great deal of research has focused on algorithms for learning features from unlabeled data. Indeed, much progress has been made on benchmark datasets like NORB and CIFAR by employing increasingly complex unsupervised learning algorithms and deep models. In this paper, however, we show that several simple factors, such as the number of hidden nodes in the model, may be more important to achieving high performance than the learning algorithm or the depth of the model. Specifically, we will apply several othe-shelf feature learning algorithms (sparse auto-encoders, sparse RBMs, K-means clustering, and Gaussian mixtures) to CIFAR, NORB, and STL datasets using only singlelayer networks. We then present a detailed analysis of the eect of changes in the model setup: the receptive field size, number of hidden nodes (features), the step-size (\u201cstride\u201d) between extracted features, and the eect of whitening. Our results show that large numbers of hidden nodes and dense feature extraction are critical to achieving high performance\u2014so critical, in fact, that when these parameters are pushed to their limits, we achieve state-of-the-art performance on both CIFAR-10 and NORB using only a single layer of features. More surprisingly, our best performance is based on K-means clustering, which is extremely fast, has no hyperparameters to tune beyond the model structure itself, and is very easy to implement. Despite the simplicity of our system, we achieve accuracy beyond all previously published results on the CIFAR-10 and NORB datasets (79.6% and 97.2% respectively)."
                },
                {
                    "title": "The German Traffic Sign Recognition Benchmark: A multi-class classification competition",
                    "abstract": "The \u201cGerman Traffic Sign Recognition Benchmark\u201d is a multi-category classification competition held at IJCNN 2011. Automatic recognition of traffic signs is required in advanced driver assistance systems and constitutes a challenging real-world computer vision and pattern recognition problem. A comprehensive, lifelike dataset of more than 50,000 traffic sign images has been collected. It reflects the strong variations in visual appearance of signs due to distance, illumination, weather conditions, partial occlusions, and rotations. The images are complemented by several precomputed feature sets to allow for applying machine learning algorithms without background knowledge in image processing. The dataset comprises 43 classes with unbalanced class frequencies. Participants have to classify two test sets of more than 12,500 images each. Here, the results on the first of these sets, which was used in the first evaluation stage of the two-fold challenge, are reported. The methods employed by the participants who achieved the best results are briefly described and compared to human traffic sign recognition performance and baseline results."
                },
                {
                    "title": "Automated Flower Classification over a Large Number of Classes",
                    "abstract": "We investigate to what extent combinations of features can improve classification performance on a large dataset of similar classes. To this end we introduce a 103 class flower dataset. We compute four different features for the flowers, each describing different aspects, namely the local shape/texture, the shape of the boundary, the overall spatial distribution of petals, and the colour. We combine the features using a multiple kernel framework with a SVM classifier. The weights for each class are learnt using the method of Varma and Ray, which has achieved state of the art performance on other large dataset, such as Caltech 101/256. Our dataset has a similar challenge in the number of classes, but with the added difficulty of large between class similarity and small within class similarity. Results show that learning the optimum kernel combination of multiple features vastly improves the performance, from 55.1% for the best single feature to 72.8% for the combination of all features."
                },
                {
                    "title": "Caltech-256 Object Category Dataset",
                    "abstract": "We introduce a challenging set of 256 object categories containing a total of 30607 images. The original Caltech-101 [1] was collected by choosing a set of object categories, downloading examples from Google Images and then manually screening out all images that did not fit the category. Caltech-256 is collected in a similar manner with several improvements: a) the number of categories is more than doubled, b) the minimum number of images in any category is increased from 31 to 80, c) artifacts due to image rotation are avoided and d) a new and larger clutter category is introduced for testing background rejection. We suggest several testing paradigms to measure classification performance, then benchmark the dataset using two simple metrics as well as a state-of-the-art spatial pyramid matching [2] algorithm. Finally we use the clutter category to train an interest detector which rejects uninformative background regions."
                },
                {
                    "title": "One-shot learning of object categories",
                    "abstract": "Learning visual models of object categories notoriously requires hundreds or thousands of training examples. We show that it is possible to learn much information about a category from just one, or a handful, of images. The key insight is that, rather than learning from scratch, one can take advantage of knowledge coming from previously learned categories, no matter how different these categories might be. We explore a Bayesian implementation of this idea. Object categories are represented by probabilistic models. Prior knowledge is represented as a probability density function on the parameters of these models. The posterior model for an object category is obtained by updating the prior in the light of one or more observations. We test a simple implementation of our algorithm on a database of 101 diverse object categories. We compare category models learned by an implementation of our Bayesian approach to models learned from by maximum likelihood (ML) and maximum a posteriori (MAP) methods. We find that on a database of more than 100 categories, the Bayesian approach produces informative models when the number of training examples is too small for other methods to operate successfully."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "Tent: Fully Test-Time Adaptation by Entropy Minimization",
                    "abstract": "A model must adapt itself to generalize to new and different data during testing. In this setting of fully test-time adaptation the model has only the test data and its own parameters. We propose to adapt by test entropy minimization (tent1): we optimize the model for confidence as measured by the entropy of its predictions. Our method estimates normalization statistics and optimizes channel-wise affine transformations to update online on each batch. Tent reduces generalization error for image classification on corrupted ImageNet and CIFAR-10/100 and reaches a new state-of-the-art error on ImageNet-C. Tent handles source-free domain adaptation on digit recognition from SVHN to MNIST/MNIST-M/USPS, on semantic segmentation from GTA to Cityscapes, and on the VisDA-C benchmark. These results are achieved in one epoch of test-time optimization without altering training."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Reading Digits in Natural Images with Unsupervised Feature Learning",
                    "abstract": "Detecting and reading text from natural images is a hard computer vision task that is central to a variety of emerging applications. Related problems like document character recognition have been widely studied by computer vision and machine learning researchers and are virtually solved for practical applications like reading handwritten digits. Reliably recognizing characters in more complex scenes like photographs, however, is far more difficult: the best existing methods lag well behind human performance on the same tasks. In this paper we attack the problem of recognizing digits in a real application using unsupervised feature learning methods: reading house numbers from street level photos. To this end, we introduce a new benchmark dataset for research use containing over 600,000 labeled digits cropped from Street View images. We then demonstrate the difficulty of recognizing these digits when the problem is approached with hand-designed features. Finally, we employ variants of two recently proposed unsupervised feature learning methods and find that they are convincingly superior on our benchmarks."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can transfer learning techniques be optimized to improve performance on specific tasks while minimizing the number of learnable parameters in neural networks?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem has significant implications for the research community as it can lead to more efficient models that require less computational power and memory, making advanced machine learning techniques accessible to a broader range of applications. By improving transfer learning methods, future research can explore new domains with limited data, enhancing the ability to generalize across tasks. This advancement could lead to practical applications in areas such as medical imaging, natural language processing, and autonomous systems, where data scarcity is a common challenge.\n\n### [Question 3] - Why is it hard?\nThe challenges in optimizing transfer learning techniques stem from the complexity of balancing model performance with the number of parameters. Naive approaches may fail because they do not adequately account for the nuances of different datasets or the specific characteristics of the tasks at hand. Technical obstacles include the need for sophisticated methods to effectively leverage pre-trained weights without overfitting to the target dataset, as well as the computational limitations associated with training large models. Theoretical challenges involve understanding the transferability of learned features across diverse tasks and ensuring that the model can adapt without losing valuable information.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on either improving model accuracy or reducing parameters, but rarely both simultaneously. Limitations in existing solutions include a lack of comprehensive methodologies that integrate various transfer learning strategies effectively. Barriers such as insufficient understanding of task vector utilization and the complexities of semi-supervised learning have hindered progress. My approach differs by combining advanced techniques like LoRA and aTLAS, which allows for better parameter efficiency while maintaining high performance across multiple datasets, addressing the gaps left by prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nMy proposed methodology involves using a combination of transfer learning techniques, specifically aTLAS and LoRA, to fine-tune models on various datasets. The datasets will include a range of tasks to evaluate the model's adaptability and performance. The metric for success will be the average few-shot accuracy across these datasets. I expect the outcomes to demonstrate that my approach achieves higher performance with fewer learnable parameters, showcasing the effectiveness of leveraging task vectors and improving memory efficiency during training."
            }
        },
        "author_data": {
            "30be9e77-5392-4795-b862-de1d9aa027b0": {
                "pk": "30be9e77-5392-4795-b862-de1d9aa027b0",
                "name": "Frederic Z. Zhang",
                "collaborators": [
                    "Stephen Gould",
                    "Dylan Campbell",
                    "Yuhui Yuan",
                    "Zhuoyao Zhong",
                    "Jiahao Zhang",
                    "Cristian Rodriguez",
                    "Yizhak Ben-Shabat",
                    "Anoop Cherian"
                ],
                "domain": [
                    "Computer Vision",
                    "Graph Neural Network",
                    "Transformer Models",
                    "Human-Object Interaction"
                ],
                "publications": [
                    {
                        "title": "Spatially Conditioned Graphs for Detecting Human-Object Interactions",
                        "abstract": "We address the problem of detecting human-object interactions in images using graphical neural networks. Unlike conventional methods, where nodes send scaled but otherwise identical messages to each of their neighbours, we propose to condition messages between pairs of nodes on their spatial relationships, resulting in different messages going to neighbours of the same node. To this end, we explore various ways of applying spatial conditioning under a multi-branch structure. Through extensive experimentation we demonstrate the advantages of spatial conditioning for the computation of the adjacency structure, messages and the refined graph features. In particular, we empirically show that as the quality of the bounding boxes increases, their coarse appearance features contribute relatively less to the disambiguation of interactions compared to the spatial information. Our method achieves an mAP of 31.33% on HICO-DET and 54.2% on V-COCO, significantly outperforming state-of-the-art on fine-tuned detections."
                    },
                    {
                        "title": "Efficient Two-Stage Detection of Human-Object Interactions with a Novel Unary-Pairwise Transformer",
                        "abstract": "Recent developments in transformer models for visual data have led to significant improvements in recognition and detection tasks. In particular, using learnable queries in place of region proposals has given rise to a new class of one-stage detection models, spearheaded by the Detection Transformer (DETR). Variations on this one-stage approach have since dominated human-object interaction (HOI) detection. However, the success of such one-stage HOI detectors can largely be attributed to the representation power of transformers. We discovered that when equipped with the same transformer, their two-stage counterparts can be more performant and memory-efficient, while taking a fraction of the time to train. In this work, we propose the Unary-Pairwise Transformer, a two-stage detector that exploits unary and pairwise representations for HOIs. We observe that the unary and pairwise parts of our transformer network specialise, with the former preferentially increasing the scores of positive examples and the latter decreasing the scores of negative examples. We evaluate our method on the HICO-DET and V-COCO datasets, and significantly outperform state-of-the-art approaches. At inference time, our model with ResNet50 approaches real-time performance on a single GPU."
                    },
                    {
                        "title": "Exploring Predicate Visual Context in Detecting Human-Object Interactions",
                        "abstract": "Recently, the DETR framework has emerged as the dominant approach for human--object interaction (HOI) research. In particular, two-stage transformer-based HOI detectors are amongst the most performant and training-efficient approaches. However, these often condition HOI classification on object features that lack fine-grained contextual information, eschewing pose and orientation information in favour of visual cues about object identity and box extremities. This naturally hinders the recognition of complex or ambiguous interactions. In this work, we study these issues through visualisations and carefully designed experiments. Accordingly, we investigate how best to re-introduce image features via cross-attention. With an improved query design, extensive exploration of keys and values, and box pair positional embeddings as spatial guidance, our model with enhanced predicate visual context (PViC) outperforms state-of-the-art methods on the HICO-DET and V-COCO benchmarks, while maintaining low training cost."
                    },
                    {
                        "title": "Temporally Grounding Instructional Diagrams in Unconstrained Videos",
                        "abstract": "We study the challenging problem of simultaneously localizing a sequence of queries in the form of instructional diagrams in a video. This requires understanding not only the individual queries but also their interrelationships. However, most existing methods focus on grounding one query at a time, ignoring the inherent structures among queries such as the general mutual exclusiveness and the temporal order. Consequently, the predicted timespans of different step diagrams may overlap considerably or violate the temporal order, thus harming the accuracy. In this paper, we tackle this issue by simultaneously grounding a sequence of step diagrams. Specifically, we propose composite queries, constructed by exhaustively pairing up the visual content features of the step diagrams and a fixed number of learnable positional embeddings. Our insight is that self-attention among composite queries carrying different content features suppress each other to reduce timespan overlaps in predictions, while the cross-attention corrects the temporal misalignment via content and position joint guidance. We demonstrate the effectiveness of our approach on the IAW dataset for grounding step diagrams and the YouCook2 benchmark for grounding natural language queries, significantly outperforming existing methods while simultaneously grounding multiple queries."
                    }
                ]
            },
            "ec01dcec-178a-41f4-80ba-f8be474451fd": {
                "pk": "ec01dcec-178a-41f4-80ba-f8be474451fd",
                "name": "Paul Albert",
                "collaborators": [
                    "Eric Arazo",
                    "Noel E. O'Connor",
                    "Kevin McGuinness"
                ],
                "domain": [
                    "Image Retrieval",
                    "Noise Robustness",
                    "Unsupervised Learning",
                    "Contrastive Learning"
                ],
                "publications": [
                    {
                        "title": "Embedding contrastive unsupervised features to cluster in- and out-of-distribution noise in corrupted image datasets",
                        "abstract": "Using search engines for web image retrieval is a tempting alternative to manual curation when creating an image dataset, but their main drawback remains the proportion of incorrect (noisy) samples retrieved. These noisy samples have been evidenced by previous works to be a mixture of in-distribution (ID) samples, assigned to the incorrect category but presenting similar visual semantics to other classes in the dataset, and out-of-distribution (OOD) images, which share no semantic correlation with any category from the dataset. The latter are, in practice, the dominant type of noisy images retrieved. To tackle this noise duality, we propose a two stage algorithm starting with a detection step where we use unsupervised contrastive feature learning to represent images in a feature space. We find that the alignment and uniformity principles of contrastive learning allow OOD samples to be linearly separated from ID samples on the unit hypersphere. We then spectrally embed the unsupervised representations using a fixed neighborhood size and apply an outlier sensitive clustering at the class level to detect the clean and OOD clusters as well as ID noisy outliers. We finally train a noise robust neural network that corrects ID noise to the correct category and utilizes OOD samples in a guided contrastive objective, clustering them to improve low-level features. Our algorithm improves the state-of-the-art results on synthetic noise image datasets as well as real-world web-crawled data. Our work is fully reproducible github.com/PaulAlbert31/SNCF."
                    }
                ]
            },
            "7064d1a4-7c63-4fd1-8d5b-b6eb65fc0d3b": {
                "pk": "7064d1a4-7c63-4fd1-8d5b-b6eb65fc0d3b",
                "name": "Cristian Rodriguez-Opazo",
                "collaborators": [
                    "Edison Marrese-Taylor",
                    "Ehsan Abbasnejad",
                    "Hamed Damirchi",
                    "Anton van den Hengel",
                    "Basura Fernando",
                    "Hiroya Takamura",
                    "Qi Wu",
                    "Damien Teney",
                    "Ignacio M. Jara",
                    "Felipe Bravo-Marquez"
                ],
                "domain": [
                    "Video Grounding",
                    "Contrastive Learning",
                    "Image Representation",
                    "Transformer Models"
                ],
                "publications": [
                    {
                        "title": "LocFormer: Enabling Transformers to Perform Temporal Moment Localization on Long Untrimmed Videos With a Feature Sampling Approach",
                        "abstract": "We propose LocFormer, a Transformer-based model for video grounding which operates at a constant memory footprint regardless of the video length, i.e. number of frames. LocFormer is designed for tasks where it is necessary to process the entire long video and at its core lie two main contributions. First, our model incorporates a new sampling technique that splits the input feature sequence into a fixed number of sections and selects a single feature per section using a stochastic approach, which allows us to obtain a feature sample set that is representative of the video content for the task at hand while keeping the memory footprint constant. Second, we propose a modular design that separates functionality, enabling us to learn an inductive bias via supervising the self-attention heads, while also effectively leveraging pre-trained text and video encoders. We test our proposals on relevant benchmark datasets for video grounding, showing that not only LocFormer can achieve excellent results including state-of-the-art performance on YouCookII, but also that our sampling technique is more effective than competing counterparts and that it consistently improves the performance of prior work, by up to 3.13\\% in the mean temporal IoU, ultimately leading to a new state-of-the-art performance on Charades-STA."
                    },
                    {
                        "title": "Synergy and Diversity in CLIP: Enhancing Performance Through Adaptive Backbone Ensembling",
                        "abstract": "Contrastive Language-Image Pretraining (CLIP) stands out as a prominent method for image representation learning. Various architectures, from vision transformers (ViTs) to convolutional networks (ResNets) have been trained with CLIP to serve as general solutions to diverse vision tasks. This paper explores the differences across various CLIP-trained vision backbones. Despite using the same data and training objective, we find that these architectures have notably different representations, different classification performance across datasets, and different robustness properties to certain types of image perturbations. Our findings indicate a remarkable possible synergy across backbones by leveraging their respective strengths. In principle, classification accuracy could be improved by over 40 percentage with an informed selection of the optimal backbone per test example.Using this insight, we develop a straightforward yet powerful approach to adaptively ensemble multiple backbones. The approach uses as few as one labeled example per class to tune the adaptive combination of backbones. On a large collection of datasets, the method achieves a remarkable increase in accuracy of up to 39.1% over the best single backbone, well beyond traditional ensembles"
                    },
                    {
                        "title": "Unveiling Backbone Effects in CLIP: Exploring Representational Synergies and Variances",
                        "abstract": "Contrastive Language-Image Pretraining (CLIP) stands out as a prominent method for image representation learning. Various neural architectures, spanning Transformer-based models like Vision Transformers (ViTs) to Convolutional Networks (ConvNets) like ResNets, are trained with CLIP and serve as universal backbones across diverse vision tasks. Despite utilizing the same data and training objectives, the effectiveness of representations learned by these architectures raises a critical question. Our investigation explores the differences in CLIP performance among these backbone architectures, revealing significant disparities in their classifications. Notably, normalizing these representations results in substantial performance variations. Our findings showcase a remarkable possible synergy between backbone predictions that could reach an improvement of over 20% through informed selection of the appropriate backbone. Moreover, we propose a simple, yet effective approach to combine predictions from multiple backbones, leading to a notable performance boost of up to 6.34\\%. We will release the code for reproducing the results."
                    }
                ]
            },
            "ed9c9cfb-cf21-4dc9-ab3c-b8e372d66a96": {
                "pk": "ed9c9cfb-cf21-4dc9-ab3c-b8e372d66a96",
                "name": "Anton van den Hengel",
                "collaborators": [
                    "Chunhua Shen",
                    "Damien Teney",
                    "Sakrapee Paisitkriangkrai",
                    "Ehsan Abbasnejad",
                    "Peng Wang",
                    "Anthony Dick",
                    "John Bastian",
                    "Lachlan Fleming",
                    "Lin Wu",
                    "Zifeng Wu"
                ],
                "domain": [
                    "Computer Vision",
                    "Machine Learning",
                    "Visual Question Answering",
                    "Anomaly Detection"
                ],
                "publications": [
                    {
                        "title": "Deconstruction of compound objects from image sets",
                        "abstract": "We propose a method to recover the structure of a compound object from multiple silhouettes. Structure is expressed as a collection of 3D primitives chosen from a pre-defined library, each with an associated pose. This has several advantages over a volume or mesh representation both for estimation and the utility of the recovered model. The main challenge in recovering such a model is the combinatorial number of possible arrangements of parts. We address this issue by exploiting the sparse nature of the problem, and show that our method scales to objects constructed from large libraries of parts."
                    },
                    {
                        "title": "Infinite Variational Autoencoder for Semi-Supervised Learning",
                        "abstract": "This paper presents an infinite variational autoencoder (VAE) whose capacity adapts to suit the input data. This is achieved using a mixture model where the mixing coefficients are modeled by a Dirichlet process, allowing us to integrate over the coefficients when performing inference. Critically, this then allows us to automatically vary the number of autoencoders in the mixture based on the data. Experiments show the flexibility of our method, particularly for semi-supervised learning, where only a small number of training samples are available."
                    },
                    {
                        "title": "Actively Seeking and Learning from Live Data",
                        "abstract": "One of the key limitations of traditional machine learning methods is their requirement for training data that exemplifies all the information to be learned. This is a particular problem for visual question answering methods, which may be asked questions about virtually anything. The approach we propose is a step toward overcoming this limitation by searching for the information required at test time. The resulting method dynamically utilizes data from an external source, such as a large set of questions/answers or images/captions. Concretely, we learn a set of base weights for a simple VQA model, that are specifically adapted to a given question with the information specifically retrieved for this question. The adaptation process leverages recent advances in gradient-based meta learning and contributions for efficient retrieval and cross-domain adaptation. We surpass the state-of-the-art on the VQA-CP v2 benchmark and demonstrate our approach to be intrinsically more robust to out-of-distribution test data. We demonstrate the use of external non-VQA data using the MS COCO captioning dataset to support the answering process. This approach opens a new avenue for open-domain VQA systems that interface with diverse sources of data."
                    },
                    {
                        "title": "Zero-Shot Visual Question Answering",
                        "abstract": "Part of the appeal of Visual Question Answering (VQA) is its promise to answer new questions about previously unseen images. Most current methods demand training questions that illustrate every possible concept, and will therefore never achieve this capability, since the volume of required training data would be prohibitive. Answering general questions about images requires methods capable of Zero-Shot VQA, that is, methods able to answer questions beyond the scope of the training questions. We propose a new evaluation protocol for VQA methods which measures their ability to perform Zero-Shot VQA, and in doing so highlights significant practical deficiencies of current approaches, some of which are masked by the biases in current datasets. We propose and evaluate several strategies for achieving Zero-Shot VQA, including methods based on pretrained word embeddings, object classifiers with semantic embeddings, and test-time retrieval of example images. Our extensive experiments are intended to serve as baselines for Zero-Shot VQA, and they also achieve state-of-the-art performance in the standard VQA evaluation setting."
                    },
                    {
                        "title": "Visual Question Answering as a Meta Learning Task",
                        "abstract": "The predominant approach to Visual Question Answering (VQA) demands that the model represents within its weights all of the information required to answer any question about any image. Learning this information from any real training set seems unlikely, and representing it in a reasonable number of weights doubly so. We propose instead to approach VQA as a meta learning task, thus separating the question answering method from the information required. At test time, the method is provided with a support set of example questions/answers, over which it reasons to resolve the given question. The support set is not fixed and can be extended without retraining, thereby expanding the capabilities of the model. To exploit this dynamically provided information, we adapt a state-of-the-art VQA model with two techniques from the recent meta learning literature, namely prototypical networks and meta networks. Experiments demonstrate the capability of the system to learn to produce completely novel answers (i.e. never seen during training) from examples provided at test time. In comparison to the existing state of the art, the proposed method produces qualitatively distinct results with higher recall of rare answers, and a better sample efficiency that allows training with little initial data. More importantly, it represents an important step towards vision-and-language methods that can learn and reason on-the-fly."
                    },
                    {
                        "title": "PersonNet: Person Re-identification with Deep Convolutional Neural Networks",
                        "abstract": "In this paper, we propose a deep end-to-end neu- ral network to simultaneously learn high-level features and a corresponding similarity metric for person re-identification. The network takes a pair of raw RGB images as input, and outputs a similarity value indicating whether the two input images depict the same person. A layer of computing neighborhood range differences across two input images is employed to capture local relationship between patches. This operation is to seek a robust feature from input images. By increasing the depth to 10 weight layers and using very small (3$\\times$3) convolution filters, our architecture achieves a remarkable improvement on the prior-art configurations. Meanwhile, an adaptive Root- Mean-Square (RMSProp) gradient decent algorithm is integrated into our architecture, which is beneficial to deep nets. Our method consistently outperforms state-of-the-art on two large datasets (CUHK03 and Market-1501), and a medium-sized data set (CUHK01)."
                    },
                    {
                        "title": "Real-time Semantic Image Segmentation via Spatial Sparsity",
                        "abstract": "We propose an approach to semantic (image) segmentation that reduces the computational costs by a factor of 25 with limited impact on the quality of results. Semantic segmentation has a number of practical applications, and for most such applications the computational costs are critical. The method follows a typical two-column network structure, where one column accepts an input image, while the other accepts a half-resolution version of that image. By identifying specific regions in the full-resolution image that can be safely ignored, as well as carefully tailoring the network structure, we can process approximately 15 highresolution Cityscapes images (1024x2048) per second using a single GTX 980 video card, while achieving a mean intersection-over-union score of 72.9% on the Cityscapes test set."
                    },
                    {
                        "title": "Reinforcement Learning with Attention that Works: A Self-Supervised Approach",
                        "abstract": "Attention models have had a significant positive impact on deep learning across a range of tasks. However previous attempts at integrating attention with reinforcement learning have failed to produce significant improvements. We propose the first combination of self attention and reinforcement learning that is capable of producing significant improvements, including new state of the art results in the Arcade Learning Environment. Unlike the selective attention models used in previous attempts, which constrain the attention via preconceived notions of importance, our implementation utilises the Markovian properties inherent in the state input. Our method produces a faithful visualisation of the policy, focusing on the behaviour of the agent. Our experiments demonstrate that the trained policies use multiple simultaneous foci of attention, and are able to modulate attention over time to deal with situations of partial observability."
                    },
                    {
                        "title": "Unshuffling Data for Improved Generalization",
                        "abstract": "Generalization beyond the training distribution is a core challenge in machine learning. The common practice of mixing and shuffling examples when training neural networks may not be optimal in this regard. We show that partitioning the data into well-chosen, non-i.i.d. subsets treated as multiple training environments can guide the learning of models with better out-of-distribution generalization. We describe a training procedure to capture the patterns that are stable across environments while discarding spurious ones. The method makes a step beyond correlation-based learning: the choice of the partitioning allows injecting information about the task that cannot be otherwise recovered from the joint distribution of the training data. We demonstrate multiple use cases with the task of visual question answering, which is notorious for dataset biases. We obtain significant improvements on VQA-CP, using environments built from prior knowledge, existing meta data, or unsupervised clustering. We also get improvements on GQA using annotations of \"equivalent questions\", and on multi-dataset training (VQA v2 / Visual Genome) by treating them as distinct environments."
                    },
                    {
                        "title": "Learning What Makes a Difference from Counterfactual Examples and Gradient Supervision",
                        "abstract": "One of the primary challenges limiting the applicability of deep learning is its susceptibility to learning spurious correlations rather than the underlying mechanisms of the task of interest. The resulting failure to generalise cannot be addressed by simply using more data from the same distribution. We propose an auxiliary training objective that improves the generalization capabilities of neural networks by leveraging an overlooked supervisory signal found in existing datasets. We use pairs of minimally-different examples with different labels, a.k.a counterfactual or contrasting examples, which provide a signal indicative of the underlying causal structure of the task. We show that such pairs can be identified in a number of existing datasets in computer vision (visual question answering, multi-label image classification) and natural language processing (sentiment analysis, natural language inference). The new training objective orients the gradient of a model's decision function with pairs of counterfactual examples. Models trained with this technique demonstrate improved performance on out-of-distribution test sets."
                    },
                    {
                        "title": "Optimally Training a Cascade Classifier",
                        "abstract": "Cascade classifiers are widely used in real-time object detection. Different from conventional classifiers that are designed for a low overall classification error rate, a classifier in each node of the cascade is required to achieve an extremely high detection rate and moderate false positive rate. Although there are a few reported methods addressing this requirement in the context of object detection, there is no a principled feature selection method that explicitly takes into account this asymmetric node learning objective. We provide such an algorithm here. We show a special case of the biased minimax probability machine has the same formulation as the linear asymmetric classifier (LAC) of \\cite{wu2005linear}. We then design a new boosting algorithm that directly optimizes the cost function of LAC. The resulting totally-corrective boosting algorithm is implemented by the column generation technique in convex optimization. Experimental results on object detection verify the effectiveness of the proposed boosting algorithm as a node classifier in cascade object detection, and show performance better than that of the current state-of-the-art."
                    },
                    {
                        "title": "A Fast Semidefinite Approach to Solving Binary Quadratic Problems",
                        "abstract": "Many computer vision problems can be formulated as binary quadratic programs (BQPs). Two classic relaxation methods are widely used for solving BQPs, namely, spectral methods and semidefinite programming (SDP), each with their own advantages and disadvantages. Spectral relaxation is simple and easy to implement, but its bound is loose. Semidefinite relaxation has a tighter bound, but its computational complexity is high for large scale problems. We present a new SDP formulation for BQPs, with two desirable properties. First, it has a similar relaxation bound to conventional SDP formulations. Second, compared with conventional SDP methods, the new SDP formulation leads to a significantly more efficient and scalable dual optimization approach, which has the same degree of complexity as spectral methods. Extensive experiments on various applications including clustering, image segmentation, co-segmentation and registration demonstrate the usefulness of our SDP formulation for solving large-scale BQPs."
                    },
                    {
                        "title": "A scalable stage-wise approach to large-margin multi-class loss based boosting",
                        "abstract": "We present a scalable and effective classification model to train multi-class boosting for multi-class classification problems. Shen and Hao introduced a direct formulation of multi- class boosting in the sense that it directly maximizes the multi- class margin [C. Shen and Z. Hao, \"A direct formulation for totally-corrective multi- class boosting\", in Proc. IEEE Conf. Comp. Vis. Patt. Recogn., 2011]. The major problem of their approach is its high computational complexity for training, which hampers its application on real-world problems. In this work, we propose a scalable and simple stage-wise multi-class boosting method, which also directly maximizes the multi-class margin. Our approach of- fers a few advantages: 1) it is simple and computationally efficient to train. The approach can speed up the training time by more than two orders of magnitude without sacrificing the classification accuracy. 2) Like traditional AdaBoost, it is less sensitive to the choice of parameters and empirically demonstrates excellent generalization performance. Experimental results on challenging multi-class machine learning and vision tasks demonstrate that the proposed approach substantially improves the convergence rate and accuracy of the final visual detector at no additional computational cost compared to existing multi-class boosting."
                    },
                    {
                        "title": "Large-margin Learning of Compact Binary Image Encodings",
                        "abstract": "The use of high-dimensional features has become a normal practice in many computer vision applications. The large dimension of these features is a limiting factor upon the number of data points which may be effectively stored and processed, however. We address this problem by developing a novel approach to learning a compact binary encoding, which exploits both pair-wise proximity and class-label information on training data set. Exploiting this extra information allows the development of encodings which, although compact, outperform the original high-dimensional features in terms of final classification or retrieval performance. The method is general, in that it is applicable to both non-parametric and parametric learning methods. This generality means that the embedded features are suitable for a wide variety of computer vision tasks, such as image classification and content-based image retrieval. Experimental results demonstrate that the new compact descriptor achieves an accuracy comparable to, and in some cases better than, the visual descriptor in the original space despite being significantly more compact. Moreover, any convex loss function and convex regularization penalty (e.g., $ \\ell_p $ norm with $ p \\ge 1 $) can be incorporated into the framework, which provides future flexibility."
                    },
                    {
                        "title": "Strengthening the Effectiveness of Pedestrian Detection with Spatially Pooled Features",
                        "abstract": "We propose a simple yet effective approach to the problem of pedestrian detection which outperforms the current state-of-the-art. Our new features are built on the basis of low-level visual features and spatial pooling. Incorporating spatial pooling improves the translational invariance and thus the robustness of the detection process. We then directly optimise the partial area under the ROC curve (\\pAUC) measure, which concentrates detection performance in the range of most practical importance. The combination of these factors leads to a pedestrian detector which outperforms all competitors on all of the standard benchmark datasets. We advance state-of-the-art results by lowering the average miss rate from $13\\%$ to $11\\%$ on the INRIA benchmark, $41\\%$ to $37\\%$ on the ETH benchmark, $51\\%$ to $42\\%$ on the TUD-Brussels benchmark and $36\\%$ to $29\\%$ on the Caltech-USA benchmark."
                    },
                    {
                        "title": "Learning to rank in person re-identification with metric ensembles",
                        "abstract": "We propose an effective structured learning based approach to the problem of person re-identification which outperforms the current state-of-the-art on most benchmark data sets evaluated. Our framework is built on the basis of multiple low-level hand-crafted and high-level visual features. We then formulate two optimization algorithms, which directly optimize evaluation measures commonly used in person re-identification, also known as the Cumulative Matching Characteristic (CMC) curve. Our new approach is practical to many real-world surveillance applications as the re-identification performance can be concentrated in the range of most practical importance. The combination of these factors leads to a person re-identification system which outperforms most existing algorithms. More importantly, we advance state-of-the-art results on person re-identification by improving the rank-$1$ recognition rates from $40\\%$ to $50\\%$ on the iLIDS benchmark, $16\\%$ to $18\\%$ on the PRID2011 benchmark, $43\\%$ to $46\\%$ on the VIPeR benchmark, $34\\%$ to $53\\%$ on the CUHK01 benchmark and $21\\%$ to $62\\%$ on the CUHK03 benchmark."
                    },
                    {
                        "title": "Graph-Structured Representations for Visual Question Answering",
                        "abstract": "This paper proposes to improve visual question answering (VQA) with structured representations of both scene contents and questions. A key challenge in VQA is to require joint reasoning over the visual and text domains. The predominant CNN/LSTM-based approach to VQA is limited by monolithic vector representations that largely ignore structure in the scene and in the form of the question. CNN feature vectors cannot effectively capture situations as simple as multiple object instances, and LSTMs process questions as series of words, which does not reflect the true complexity of language structure. We instead propose to build graphs over the scene objects and over the question words, and we describe a deep neural network that exploits the structure in these representations. This shows significant benefit over the sequential processing of LSTMs. The overall efficacy of our approach is demonstrated by significant improvements over the state-of-the-art, from 71.2% to 74.4% in accuracy on the \"abstract scenes\" multiple-choice benchmark, and from 34.7% to 39.1% in accuracy over pairs of \"balanced\" scenes, i.e. images with fine-grained differences and opposite yes/no answers to a same question."
                    },
                    {
                        "title": "Efficient SDP Inference for Fully-connected CRFs Based on Low-rank Decomposition",
                        "abstract": "Conditional Random Fields (CRF) have been widely used in a variety of computer vision tasks. Conventional CRFs typically define edges on neighboring image pixels, resulting in a sparse graph such that efficient inference can be performed. However, these CRFs fail to model long-range contextual relationships. Fully-connected CRFs have thus been proposed. While there are efficient approximate inference methods for such CRFs, usually they are sensitive to initialization and make strong assumptions. In this work, we develop an efficient, yet general algorithm for inference on fully-connected CRFs. The algorithm is based on a scalable SDP algorithm and the low- rank approximation of the similarity/kernel matrix. The core of the proposed algorithm is a tailored quasi-Newton method that takes advantage of the low-rank matrix approximation when solving the specialized SDP dual problem. Experiments demonstrate that our method can be applied on fully-connected CRFs that cannot be solved previously, such as pixel-level image co-segmentation."
                    },
                    {
                        "title": "Deep Anomaly Detection with Deviation Networks",
                        "abstract": "Although deep learning has been applied to successfully address many data mining problems, relatively limited work has been done on deep learning for anomaly detection. Existing deep anomaly detection methods, which focus on learning new feature representations to enable downstream anomaly detection methods, perform indirect optimization of anomaly scores, leading to data-inefficient learning and suboptimal anomaly scoring. Also, they are typically designed as unsupervised learning due to the lack of large-scale labeled anomaly data. As a result, they are difficult to leverage prior knowledge (e.g., a few labeled anomalies) when such information is available as in many real-world anomaly detection applications.   This paper introduces a novel anomaly detection framework and its instantiation to address these problems. Instead of representation learning, our method fulfills an end-to-end learning of anomaly scores by a neural deviation learning, in which we leverage a few (e.g., multiple to dozens) labeled anomalies and a prior probability to enforce statistically significant deviations of the anomaly scores of anomalies from that of normal data objects in the upper tail. Extensive results show that our method can be trained substantially more data-efficiently and achieves significantly better anomaly scoring than state-of-the-art competing methods."
                    },
                    {
                        "title": "Is margin preserved after random projection?",
                        "abstract": "Random projections have been applied in many machine learning algorithms. However, whether margin is preserved after random projection is non-trivial and not well studied. In this paper we analyse margin distortion after random projection, and give the conditions of margin preservation for binary classification problems. We also extend our analysis to margin for multiclass problems, and provide theoretical bounds on multiclass margin on the projected data."
                    }
                ]
            },
            "3ed90b76-5cc1-4a61-af93-0471d878aae5": {
                "pk": "3ed90b76-5cc1-4a61-af93-0471d878aae5",
                "name": "Ehsan Abbasnejad",
                "collaborators": [
                    "Anton van den Hengel",
                    "Damien Teney",
                    "Iman Abbasnejad",
                    "Anthony Dick",
                    "Qinfeng Shi",
                    "Qi Wu",
                    "Javen Shi",
                    "Islam Nassar",
                    "Gholamreza Haffari",
                    "Seong Joon Oh"
                ],
                "domain": [
                    "Generative Models",
                    "Semi-Supervised Learning",
                    "Adversarial Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Infinite Variational Autoencoder for Semi-Supervised Learning",
                        "abstract": "This paper presents an infinite variational autoencoder (VAE) whose capacity adapts to suit the input data. This is achieved using a mixture model where the mixing coefficients are modeled by a Dirichlet process, allowing us to integrate over the coefficients when performing inference. Critically, this then allows us to automatically vary the number of autoencoders in the mixture based on the data. Experiments show the flexibility of our method, particularly for semi-supervised learning, where only a small number of training samples are available."
                    },
                    {
                        "title": "Bayesian Conditional Generative Adverserial Networks",
                        "abstract": "Traditional GANs use a deterministic generator function (typically a neural network) to transform a random noise input $z$ to a sample $\\mathbf{x}$ that the discriminator seeks to distinguish. We propose a new GAN called Bayesian Conditional Generative Adversarial Networks (BC-GANs) that use a random generator function to transform a deterministic input $y'$ to a sample $\\mathbf{x}$. Our BC-GANs extend traditional GANs to a Bayesian framework, and naturally handle unsupervised learning, supervised learning, and semi-supervised learning problems. Experiments show that the proposed BC-GANs outperforms the state-of-the-arts."
                    },
                    {
                        "title": "Predicting is not Understanding: Recognizing and Addressing Underspecification in Machine Learning",
                        "abstract": "Machine learning (ML) models are typically optimized for their accuracy on a given dataset. However, this predictive criterion rarely captures all desirable properties of a model, in particular how well it matches a domain expert's understanding of a task. Underspecification refers to the existence of multiple models that are indistinguishable in their in-domain accuracy, even though they differ in other desirable properties such as out-of-distribution (OOD) performance. Identifying these situations is critical for assessing the reliability of ML models.   We formalize the concept of underspecification and propose a method to identify and partially address it. We train multiple models with an independence constraint that forces them to implement different functions. They discover predictive features that are otherwise ignored by standard empirical risk minimization (ERM), which we then distill into a global model with superior OOD performance. Importantly, we constrain the models to align with the data manifold to ensure that they discover meaningful features. We demonstrate the method on multiple datasets in computer vision (collages, WILDS-Camelyon17, GQA) and discuss general implications of underspecification. Most notably, in-domain performance cannot serve for OOD model selection without additional assumptions."
                    },
                    {
                        "title": "Selective Mixup Helps with Distribution Shifts, But Not (Only) because of Mixup",
                        "abstract": "Mixup is a highly successful technique to improve generalization of neural networks by augmenting the training data with combinations of random pairs. Selective mixup is a family of methods that apply mixup to specific pairs, e.g. only combining examples across classes or domains. These methods have claimed remarkable improvements on benchmarks with distribution shifts, but their mechanisms and limitations remain poorly understood.   We examine an overlooked aspect of selective mixup that explains its success in a completely new light. We find that the non-random selection of pairs affects the training distribution and improve generalization by means completely unrelated to the mixing. For example in binary classification, mixup across classes implicitly resamples the data for a uniform class distribution - a classical solution to label shift. We show empirically that this implicit resampling explains much of the improvements in prior work. Theoretically, these results rely on a regression toward the mean, an accidental property that we identify in several datasets.   We have found a new equivalence between two successful methods: selective mixup and resampling. We identify limits of the former, confirm the effectiveness of the latter, and find better combinations of their respective benefits."
                    },
                    {
                        "title": "Neural Redshift: Random Networks are not Random Functions",
                        "abstract": "Our understanding of the generalization capabilities of neural networks (NNs) is still incomplete. Prevailing explanations are based on implicit biases of gradient descent (GD) but they cannot account for the capabilities of models from gradient-free methods nor the simplicity bias recently observed in untrained networks. This paper seeks other sources of generalization in NNs.   Findings. To understand the inductive biases provided by architectures independently from GD, we examine untrained, random-weight networks. Even simple MLPs show strong inductive biases: uniform sampling in weight space yields a very biased distribution of functions in terms of complexity. But unlike common wisdom, NNs do not have an inherent \"simplicity bias\". This property depends on components such as ReLUs, residual connections, and layer normalizations. Alternative architectures can be built with a bias for any level of complexity. Transformers also inherit all these properties from their building blocks.   Implications. We provide a fresh explanation for the success of deep learning independent from gradient-based training. It points at promising avenues for controlling the solutions implemented by trained models."
                    },
                    {
                        "title": "Gold Seeker: Information Gain from Policy Distributions for Goal-oriented Vision-and-Langauge Reasoning",
                        "abstract": "As Computer Vision moves from a passive analysis of pixels to active analysis of semantics, the breadth of information algorithms need to reason over has expanded significantly. One of the key challenges in this vein is the ability to identify the information required to make a decision, and select an action that will recover it. We propose a reinforcement-learning approach that maintains a distribution over its internal information, thus explicitly representing the ambiguity in what it knows, and needs to know, towards achieving its goal. Potential actions are then generated according to this distribution. For each potential action a distribution of the expected outcomes is calculated, and the value of the potential information gain assessed. The action taken is that which maximizes the potential information gain. We demonstrate this approach applied to two vision-and-language problems that have attracted significant recent interest, visual dialog and visual query generation. In both cases, the method actively selects actions that will best reduce its internal uncertainty and outperforms its competitors in achieving the goal of the challenge."
                    },
                    {
                        "title": "SCONE-GAN: Semantic Contrastive learning-based Generative Adversarial Network for an end-to-end image translation",
                        "abstract": "SCONE-GAN presents an end-to-end image translation, which is shown to be effective for learning to generate realistic and diverse scenery images. Most current image-to-image translation approaches are devised as two mappings: a translation from the source to target domain and another to represent its inverse. While successful in many applications, these approaches may suffer from generating trivial solutions with limited diversity. That is because these methods learn more frequent associations rather than the scene structures. To mitigate the problem, we propose SCONE-GAN that utilises graph convolutional networks to learn the objects dependencies, maintain the image structure and preserve its semantics while transferring images into the target domain. For more realistic and diverse image generation we introduce style reference image. We enforce the model to maximize the mutual information between the style image and output. The proposed method explicitly maximizes the mutual information between the related patches, thus encouraging the generator to produce more diverse images. We validate the proposed algorithm for image-to-image translation and stylizing outdoor images. Both qualitative and quantitative results demonstrate the effectiveness of our approach on four dataset."
                    },
                    {
                        "title": "What's to know? Uncertainty as a Guide to Asking Goal-oriented Questions",
                        "abstract": "One of the core challenges in Visual Dialogue problems is asking the question that will provide the most useful information towards achieving the required objective. Encouraging an agent to ask the right questions is difficult because we don't know a-priori what information the agent will need to achieve its task, and we don't have an explicit model of what it knows already. We propose a solution to this problem based on a Bayesian model of the uncertainty in the implicit model maintained by the visual dialogue agent, and in the function used to select an appropriate output. By selecting the question that minimises the predicted regret with respect to this implicit model the agent actively reduces ambiguity. The Bayesian model of uncertainty also enables a principled method for identifying when enough information has been acquired, and an action should be selected. We evaluate our approach on two goal-oriented dialogue datasets, one for visual-based collaboration task and the other for a negotiation-based task. Our uncertainty-aware information-seeking model outperforms its counterparts in these two challenging problems."
                    },
                    {
                        "title": "Reinforcement Learning with Attention that Works: A Self-Supervised Approach",
                        "abstract": "Attention models have had a significant positive impact on deep learning across a range of tasks. However previous attempts at integrating attention with reinforcement learning have failed to produce significant improvements. We propose the first combination of self attention and reinforcement learning that is capable of producing significant improvements, including new state of the art results in the Arcade Learning Environment. Unlike the selective attention models used in previous attempts, which constrain the attention via preconceived notions of importance, our implementation utilises the Markovian properties inherent in the state input. Our method produces a faithful visualisation of the policy, focusing on the behaviour of the agent. Our experiments demonstrate that the trained policies use multiple simultaneous foci of attention, and are able to modulate attention over time to deal with situations of partial observability."
                    },
                    {
                        "title": "Unshuffling Data for Improved Generalization",
                        "abstract": "Generalization beyond the training distribution is a core challenge in machine learning. The common practice of mixing and shuffling examples when training neural networks may not be optimal in this regard. We show that partitioning the data into well-chosen, non-i.i.d. subsets treated as multiple training environments can guide the learning of models with better out-of-distribution generalization. We describe a training procedure to capture the patterns that are stable across environments while discarding spurious ones. The method makes a step beyond correlation-based learning: the choice of the partitioning allows injecting information about the task that cannot be otherwise recovered from the joint distribution of the training data. We demonstrate multiple use cases with the task of visual question answering, which is notorious for dataset biases. We obtain significant improvements on VQA-CP, using environments built from prior knowledge, existing meta data, or unsupervised clustering. We also get improvements on GQA using annotations of \"equivalent questions\", and on multi-dataset training (VQA v2 / Visual Genome) by treating them as distinct environments."
                    },
                    {
                        "title": "Progressive Class Semantic Matching for Semi-supervised Text Classification",
                        "abstract": "Semi-supervised learning is a promising way to reduce the annotation cost for text-classification. Combining with pre-trained language models (PLMs), e.g., BERT, recent semi-supervised learning methods achieved impressive performance. In this work, we further investigate the marriage between semi-supervised learning and a pre-trained language model. Unlike existing approaches that utilize PLMs only for model parameter initialization, we explore the inherent topic matching capability inside PLMs for building a more powerful semi-supervised learning approach. Specifically, we propose a joint semi-supervised learning process that can progressively build a standard $K$-way classifier and a matching network for the input text and the Class Semantic Representation (CSR). The CSR will be initialized from the given labeled sentences and progressively updated through the training process. By means of extensive experiments, we show that our method can not only bring remarkable improvement to baselines, but also overall be more stable, and achieves state-of-the-art performance in semi-supervised text classification."
                    },
                    {
                        "title": "ProtoCon: Pseudo-label Refinement via Online Clustering and Prototypical Consistency for Efficient Semi-supervised Learning",
                        "abstract": "Confidence-based pseudo-labeling is among the dominant approaches in semi-supervised learning (SSL). It relies on including high-confidence predictions made on unlabeled data as additional targets to train the model. We propose ProtoCon, a novel SSL method aimed at the less-explored label-scarce SSL where such methods usually underperform. ProtoCon refines the pseudo-labels by leveraging their nearest neighbours' information. The neighbours are identified as the training proceeds using an online clustering approach operating in an embedding space trained via a prototypical loss to encourage well-formed clusters. The online nature of ProtoCon allows it to utilise the label history of the entire dataset in one training cycle to refine labels in the following cycle without the need to store image embeddings. Hence, it can seamlessly scale to larger datasets at a low cost. Finally, ProtoCon addresses the poor training signal in the initial phase of training (due to fewer confident predictions) by introducing an auxiliary self-supervised loss. It delivers significant gains and faster convergence over state-of-the-art across 5 datasets, including CIFARs, ImageNet and DomainNet."
                    },
                    {
                        "title": "Do Deep Neural Network Solutions Form a Star Domain?",
                        "abstract": "It has recently been conjectured that neural network solution sets reachable via stochastic gradient descent (SGD) are convex, considering permutation invariances (Entezari et al., 2022). This means that a linear path can connect two independent solutions with low loss, given the weights of one of the models are appropriately permuted. However, current methods to test this theory often require very wide networks to succeed. In this work, we conjecture that more generally, the SGD solution set is a \"star domain\" that contains a \"star model\" that is linearly connected to all the other solutions via paths with low loss values, modulo permutations. We propose the Starlight algorithm that finds a star model of a given learning task. We validate our claim by showing that this star model is linearly connected with other independently found solutions. As an additional benefit of our study, we demonstrate better uncertainty estimates on the Bayesian Model Averaging over the obtained star domain. Further, we demonstrate star models as potential substitutes for model ensembles. Our code is available at https://github.com/aktsonthalia/starlight."
                    },
                    {
                        "title": "Adaptive Neuro-Surrogate-Based Optimisation Method for Wave Energy Converters Placement Optimisation",
                        "abstract": "The installed amount of renewable energy has expanded massively in recent years. Wave energy, with its high capacity factors has great potential to complement established sources of solar and wind energy. This study explores the problem of optimising the layout of advanced, three-tether wave energy converters in a size-constrained farm in a numerically modelled ocean environment. Simulating and computing the complicated hydrodynamic interactions in wave farms can be computationally costly, which limits optimisation methods to have just a few thousand evaluations. For dealing with this expensive optimisation problem, an adaptive neuro-surrogate optimisation (ANSO) method is proposed that consists of a surrogate Recurrent Neural Network (RNN) model trained with a very limited number of observations. This model is coupled with a fast meta-heuristic optimiser for adjusting the model's hyper-parameters. The trained model is applied using a greedy local search with a backtracking optimisation strategy. For evaluating the performance of the proposed approach, some of the more popular and successful Evolutionary Algorithms (EAs) are compared in four real wave scenarios (Sydney, Perth, Adelaide and Tasmania). Experimental results show that the adaptive neuro model is competitive with other optimisation methods in terms of total harnessed power output and faster in terms of total computational costs."
                    },
                    {
                        "title": "Februus: Input Purification Defense Against Trojan Attacks on Deep Neural Network Systems",
                        "abstract": "We propose Februus; a new idea to neutralize highly potent and insidious Trojan attacks on Deep Neural Network (DNN) systems at run-time. In Trojan attacks, an adversary activates a backdoor crafted in a deep neural network model using a secret trigger, a Trojan, applied to any input to alter the model's decision to a target prediction---a target determined by and only known to the attacker. Februus sanitizes the incoming input by surgically removing the potential trigger artifacts and restoring the input for the classification task. Februus enables effective Trojan mitigation by sanitizing inputs with no loss of performance for sanitized inputs, Trojaned or benign. Our extensive evaluations on multiple infected models based on four popular datasets across three contrasting vision applications and trigger types demonstrate the high efficacy of Februus. We dramatically reduced attack success rates from 100% to near 0% for all cases (achieving 0% on multiple cases) and evaluated the generalizability of Februus to defend against complex adaptive attacks; notably, we realized the first defense against the advanced partial Trojan attack. To the best of our knowledge, Februus is the first backdoor defense method for operation at run-time capable of sanitizing Trojaned inputs without requiring anomaly detection methods, model retraining or costly labeled data."
                    },
                    {
                        "title": "3D Hand Pose Estimation using Simulation and Partial-Supervision with a Shared Latent Space",
                        "abstract": "Tremendous amounts of expensive annotated data are a vital ingredient for state-of-the-art 3d hand pose estimation. Therefore, synthetic data has been popularized as annotations are automatically available. However, models trained only with synthetic samples do not generalize to real data, mainly due to the gap between the distribution of synthetic and real data. In this paper, we propose a novel method that seeks to predict the 3d position of the hand using both synthetic and partially-labeled real data. Accordingly, we form a shared latent space between three modalities: synthetic depth image, real depth image, and pose. We demonstrate that by carefully learning the shared latent space, we can find a regression model that is able to generalize to real data. As such, we show that our method produces accurate predictions in both semi-supervised and unsupervised settings. Additionally, the proposed model is capable of generating novel, meaningful, and consistent samples from all of the three domains. We evaluate our method qualitatively and quantitively on two highly competitive benchmarks (i.e., NYU and ICVL) and demonstrate its superiority over the state-of-the-art methods. The source code will be made available at https://github.com/masabdi/LSPS."
                    },
                    {
                        "title": "On Incorporating Semantic Prior Knowledge in Deep Learning Through Embedding-Space Constraints",
                        "abstract": "The knowledge that humans hold about a problem often extends far beyond a set of training data and output labels. While the success of deep learning mostly relies on supervised training, important properties cannot be inferred efficiently from end-to-end annotations alone, for example causal relations or domain-specific invariances. We present a general technique to supplement supervised training with prior knowledge expressed as relations between training instances. We illustrate the method on the task of visual question answering to exploit various auxiliary annotations, including relations of equivalence and of logical entailment between questions. Existing methods to use these annotations, including auxiliary losses and data augmentation, cannot guarantee the strict inclusion of these relations into the model since they require a careful balancing against the end-to-end objective. Our method uses these relations to shape the embedding space of the model, and treats them as strict constraints on its learned representations. In the context of VQA, this approach brings significant improvements in accuracy and robustness, in particular over the common practice of incorporating the constraints as a soft regularizer. We also show that incorporating this type of prior knowledge with our method brings consistent improvements, independently from the amount of supervised data used. It demonstrates the value of an additional training signal that is otherwise difficult to extract from end-to-end annotations alone."
                    },
                    {
                        "title": "All Labels Are Not Created Equal: Enhancing Semi-supervision via Label Grouping and Co-training",
                        "abstract": "Pseudo-labeling is a key component in semi-supervised learning (SSL). It relies on iteratively using the model to generate artificial labels for the unlabeled data to train against. A common property among its various methods is that they only rely on the model's prediction to make labeling decisions without considering any prior knowledge about the visual similarity among the classes. In this paper, we demonstrate that this degrades the quality of pseudo-labeling as it poorly represents visually similar classes in the pool of pseudo-labeled data. We propose SemCo, a method which leverages label semantics and co-training to address this problem. We train two classifiers with two different views of the class labels: one classifier uses the one-hot view of the labels and disregards any potential similarity among the classes, while the other uses a distributed view of the labels and groups potentially similar classes together. We then co-train the two classifiers to learn based on their disagreements. We show that our method achieves state-of-the-art performance across various SSL tasks including 5.6% accuracy improvement on Mini-ImageNet dataset with 1000 labeled examples. We also show that our method requires smaller batch size and fewer training iterations to reach its best performance. We make our code available at https://github.com/islam-nassar/semco."
                    },
                    {
                        "title": "ID and OOD Performance Are Sometimes Inversely Correlated on Real-world Datasets",
                        "abstract": "Several studies have compared the in-distribution (ID) and out-of-distribution (OOD) performance of models in computer vision and NLP. They report a frequent positive correlation and some surprisingly never even observe an inverse correlation indicative of a necessary trade-off. The possibility of inverse patterns is important to determine whether ID performance can serve as a proxy for OOD generalization capabilities.   This paper shows with multiple datasets that inverse correlations between ID and OOD performance do happen in real-world data - not only in theoretical worst-case settings. We also explain theoretically how these cases can arise even in a minimal linear setting, and why past studies could miss such cases due to a biased selection of models.   Our observations lead to recommendations that contradict those found in much of the current literature. - High OOD performance sometimes requires trading off ID performance. - Focusing on ID performance alone may not lead to optimal OOD performance. It may produce diminishing (eventually negative) returns in OOD performance. - In these cases, studies on OOD generalization that use ID performance for model selection (a common recommended practice) will necessarily miss the best-performing models, making these studies blind to a whole range of phenomena."
                    },
                    {
                        "title": "BruSLeAttack: A Query-Efficient Score-Based Black-Box Sparse Adversarial Attack",
                        "abstract": "We study the unique, less-well understood problem of generating sparse adversarial samples simply by observing the score-based replies to model queries. Sparse attacks aim to discover a minimum number-the l0 bounded-perturbations to model inputs to craft adversarial examples and misguide model decisions. But, in contrast to query-based dense attack counterparts against black-box models, constructing sparse adversarial perturbations, even when models serve confidence score information to queries in a score-based setting, is non-trivial. Because, such an attack leads to i) an NP-hard problem; and ii) a non-differentiable search space. We develop the BruSLeAttack-a new, faster (more query-efficient) Bayesian algorithm for the problem. We conduct extensive attack evaluations including an attack demonstration against a Machine Learning as a Service (MLaaS) offering exemplified by Google Cloud Vision and robustness testing of adversarial training regimes and a recent defense against black-box attacks. The proposed attack scales to achieve state-of-the-art attack success rates and query efficiency on standard computer vision tasks such as ImageNet across different model architectures. Our artefacts and DIY attack samples are available on GitHub. Importantly, our work facilitates faster evaluation of model vulnerabilities and raises our vigilance on the safety, security and reliability of deployed systems."
                    }
                ]
            }
        }
    },
    "2406.07815": {
        "paper_data": {
            "title": "Are Large Language Models Good Statisticians?",
            "url": "http://arxiv.org/abs/2406.07815v2",
            "arxiv_id": "2406.07815",
            "authors": [
                "Yizhang Zhu",
                "Shiyin Du",
                "Boyan Li",
                "Yuyu Luo",
                "Nan Tang"
            ],
            "abstract": "Large Language Models (LLMs) have demonstrated impressive capabilities across a range of scientific tasks including mathematics, physics, and chemistry. Despite their successes, the effectiveness of LLMs in handling complex statistical tasks remains systematically under-explored. To bridge this gap, we introduce StatQA, a new benchmark designed for statistical analysis tasks. StatQA comprises 11,623 examples tailored to evaluate LLMs' proficiency in specialized statistical tasks and their applicability assessment capabilities, particularly for hypothesis testing methods. We systematically experiment with representative LLMs using various prompting strategies and show that even state-of-the-art models such as GPT-4o achieve a best performance of only 64.83%, indicating significant room for improvement. Notably, while open-source LLMs (e.g. LLaMA-3) show limited capability, those fine-tuned ones exhibit marked improvements, outperforming all in-context learning-based methods (e.g. GPT-4o). Moreover, our comparative human experiments highlight a striking contrast in error types between LLMs and humans: LLMs primarily make applicability errors, whereas humans mostly make statistical task confusion errors. This divergence highlights distinct areas of proficiency and deficiency, suggesting that combining LLM and human expertise could lead to complementary strengths, inviting further investigation into their collaborative potential. Our source code and data are available at https://statqa.github.io/.",
            "introduction": "   1 Introduction  Statistical analysis can capture data patterns and convert them into usable evidence, which is crucial to data science and machine learning applications\u00a0[1, 2, 3, 4, 5, 6, 7, 8]. As shown in Figure\u00a01, a typical statistical analysis task involves, given a table D\ud835\udc37Ditalic_D and a statistical question Q\ud835\udc44Qitalic_Q, a qualified statistician should be proficient in selecting relevant columns \u2102\u2102\\mathbb{C}blackboard_C, choosing the appropriate statistical methods \ud835\udd44\ud835\udd44\\mathbb{M}blackboard_M, and computing the results based on \ud835\udd44\ud835\udd44\\mathbb{M}blackboard_M using \u2102\u2102\\mathbb{C}blackboard_C. Formally, the statistical analysis task can be divided into two stages: (1) identifying appropriate statistical methods and parameters (e.g., relevant columns in the table), and (2) computing the statistical results and deriving the conclusion. The first stage requires statistical expertise to assess the applicability of methods, considering factors such as data type, distribution characteristics, and sample size, which is the core of the statistical task. In contrast, the second stage (i.e., computation) can be easily aided by external tools\u00a0[9] such as WolframAlpha\u00a0[10].   Figure 1: An Example of Statistical Analysis Task   Inspired by the extensive application of Large Language Models (LLMs)\u00a0[11, 12, 13], we pose a critical question: Do LLMs truly understand such \u201cstatistical literacy\u201d? Specifically, can LLMs competently select relevant data, recognize prerequisites, and discern appropriate usage scenarios to assess the effectiveness of statistical methods?   Currently, research on the mathematical reasoning abilities of LLMs has primarily focused on the accuracy of computational processes and results, involving conventional statistical methods\u00a0[14, 15, 16, 17]. However, studies examining datasets requiring specialized statistical testing methods, particularly those assessing the applicability of these methods, are minimal. This gap in the literature motivated us to explore and address this area. Our study aims to answer the following critical questions.     \u2022  Q1: How can we evaluate LLMs\u2019 performance in more complex and specialized statistical testing tasks? Constructing an appropriate benchmark is crucial for accurate performance evaluation. Given the challenges posed by the lack of datasets and scarcity of examples in this specialized field, how can we efficiently develop such a benchmark?    \u2022  Q2: How capable are current LLMs in this field, and how can we improve their performance? This requires systematic experiments to evaluate the capabilities of current LLMs, exploring the impact of different models, prompting strategies, and fine-tuning methods on their performance.    \u2022  Q3: How do humans perform compared to LLMs, and what are the differences in their performance? This involves a comparative study between humans and LLMs, aiming to analyze their respective strengths and weaknesses and explore potential complementary between human expertise and LLM capabilities.      Contributions.  Our contributions are summarized as follows:     \u2022  StatQA. We propose StatQA, a new benchmark for statistical analysis tasks, particularly focusing on the applicability assessment of statistical methods. We introduce an automated pipeline to construct StatQA by synthesizing statistical tasks and their corresponding answers, which also provides insights for dataset construction in other specialized domains with scarce examples.    \u2022  Systematic Evaluation. We conduct extensive evaluations on widely used LLMs to establish benchmarks for statistical tasks. We also explore several strategies, including domain-specific prompts and fine-tuning, to better harness the capabilities of LLMs for these tasks.    \u2022  Comparative Study between Humans and LLMs. We organize group-based human experiments and comparatively analyze differences between humans and LLMs in performance and errors. Our",
            "references": [
                {
                    "title": "A Survey of NL2SQL with Large Language Models: Where are we, and where are we going?",
                    "abstract": "Translating users' natural language queries (NL) into SQL queries (i.e., NL2SQL) can significantly reduce barriers to accessing relational databases and support various commercial applications. The performance of NL2SQL has been greatly enhanced with the emergence of Large Language Models (LLMs). In this survey, we provide a comprehensive review of NL2SQL techniques powered by LLMs, covering its entire lifecycle from the following four aspects: (1) Model: NL2SQL translation techniques that tackle not only NL ambiguity and under-specification, but also properly map NL with database schema and instances; (2) Data: From the collection of training data, data synthesis due to training data scarcity, to NL2SQL benchmarks; (3) Evaluation: Evaluating NL2SQL methods from multiple angles using different metrics and granularities; and (4) Error Analysis: analyzing NL2SQL errors to find the root cause and guiding NL2SQL models to evolve. Moreover, we provide a rule of thumb for developing NL2SQL solutions. Finally, we discuss the research challenges and open problems of NL2SQL in the LLMs era."
                },
                {
                    "title": "HAIChart: Human and AI Paired Visualization System",
                    "abstract": "\n The growing importance of data visualization in business intelligence and data science emphasizes the need for tools that can efficiently generate meaningful visualizations from large datasets. Existing tools fall into two main categories: human-powered tools (\n e.g.\n , Tableau and PowerBI), which require intensive expert involvement, and AI-powered automated tools (\n e.g.\n , Draco and Table2Charts), which often fall short of\n guessing\n specific user needs.\n \n \n In this paper, we aim to achieve the best of both worlds. Our key idea is to initially auto-generate a set of high-quality visualizations to minimize manual effort, then refine this process iteratively with user feedback to more closely align with their needs. To this end, we present HAIChart, a reinforcement learning-based framework designed to iteratively recommend good visualizations for a given dataset by incorporating user feedback. Specifically, we propose a Monte Carlo Graph Search-based visualization generation algorithm paired with a composite reward function to efficiently explore the visualization space and automatically generate good visualizations. We devise a visualization hints mechanism to actively incorporate user feedback, thus progressively refining the visualization generation module. We further prove that the top-\n k\n visualization hints selection problem is NP-hard and design an efficient algorithm. We conduct both quantitative evaluations and user studies, showing that HAIChart significantly outperforms state-of-the-art human-powered tools (21% better at Recall and 1.8\u00d7 faster) and AI-powered automatic tools (25.1% and 14.9% better in terms of Hit@3 and R10@30, respectively).\n"
                },
                {
                    "title": "The Dawn of Natural Language to SQL: Are We Fully Ready?",
                    "abstract": "\n Translating users' natural language questions into SQL queries (\n i.e.\n , nl2sql) significantly lowers the barriers to accessing relational databases. The emergence of Large Language Models has introduced a novel paradigm in nl2sql tasks, enhancing capabilities dramatically. However, this raises a critical question:\n Are we fully prepared to deploy\n nl2sql\n models in production?\n \n \n To address the posed questions, we present a multi-angle nl2sql evaluation framework,\n NL2SQL360\n , to facilitate the design and test of new nl2sql methods for researchers. Through\n NL2SQL360\n , we conduct a detailed comparison of leading nl2sql methods across a range of application scenarios, such as different data domains and sql characteristics, offering valuable insights for selecting the most appropriate nl2sql methods for specific needs. Moreover, we explore the nl2sql design space, leveraging NL2SQL360 to automate the identification of an optimal nl2sql solution tailored to user-specific needs. Specifically,\n NL2SQL360\n identifies an effective nl2sql method,\n SuperSQL\n , distinguished under the Spider dataset using the execution accuracy metric. Remarkably,\n SuperSQL\n achieves competitive performance with execution accuracy of\n 87\n % and\n 62.66\n % on the Spider and BIRD test sets, respectively.\n"
                },
                {
                    "title": "CoInsight: Visual Storytelling for Hierarchical Tables With Connected Insights",
                    "abstract": "Extracting data insights and generating visual data stories from tabular data are critical parts of data analysis. However, most existing studies primarily focus on tabular data stored as flat tables, typically without leveraging the relations between cells in the headers of hierarchical tables. When properly used, rich table headers can enable the extraction of many additional data stories. To assist analysts in visual data storytelling, an approach is needed to organize these data insights efficiently. In this work, we propose CoInsight, a system to facilitate visual storytelling for hierarchical tables by connecting insights. CoInsight extracts data insights from hierarchical tables and builds insight relations according to the structure of table headers. It further visualizes related data insights using a nested graph with edge bundling. We evaluate the CoInsight system through a usage scenario and a user experiment. The results demonstrate the utility and usability of CoInsight for converting data insights in hierarchical tables into visual data stories."
                },
                {
                    "title": "MathVerse: Does Your Multi-modal LLM Truly See the Diagrams in Visual Math Problems?",
                    "abstract": "The remarkable progress of Multi-modal Large Language Models (MLLMs) has garnered unparalleled attention, due to their superior performance in visual contexts. However, their capabilities in visual math problem-solving remain insufficiently evaluated and understood. We investigate current benchmarks to incorporate excessive visual content within textual questions, which potentially assist MLLMs in deducing answers without truly interpreting the input diagrams. To this end, we introduce MathVerse, an all-around visual math benchmark designed for an equitable and in-depth evaluation of MLLMs. We meticulously collect 2,612 high-quality, multi-subject math problems with diagrams from publicly available sources. Each problem is then transformed by human annotators into six distinct versions, each offering varying degrees of information content in multi-modality, contributing to 15K test samples in total. This approach allows MathVerse to comprehensively assess whether and how much MLLMs can truly understand the visual diagrams for mathematical reasoning. In addition, we propose a Chain-of-Thought (CoT) evaluation strategy for a fine-grained assessment of the output answers. Rather than naively judging True or False, we employ GPT-4(V) to adaptively extract crucial reasoning steps, and then score each step with detailed error analysis, which can reveal the intermediate CoT reasoning quality by MLLMs. We hope the MathVerse benchmark may provide unique insights to guide the future development of MLLMs. Project page: https://mathverse-cuhk.github.io"
                },
                {
                    "title": "LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models",
                    "abstract": "Efficient fine-tuning is vital for adapting large language models (LLMs) to downstream tasks. However, it requires non-trivial efforts to implement these methods on different models. We present LlamaFactory, a unified framework that integrates a suite of cutting-edge efficient training methods. It provides a solution for flexibly customizing the fine-tuning of 100+ LLMs without the need for coding through the built-in web UI LlamaBoard. We empirically validate the efficiency and effectiveness of our framework on language modeling and text generation tasks. It has been released at https://github.com/hiyouga/LLaMA-Factory and received over 25,000 stars and 3,000 forks."
                },
                {
                    "title": "Are LLMs Capable of Data-based Statistical and Causal Reasoning? Benchmarking Advanced Quantitative Reasoning with Data",
                    "abstract": "Quantitative reasoning is a critical skill to analyze data, yet the assessment of such ability remains limited. To address this gap, we introduce the Quantitative Reasoning with Data (QRData) benchmark, aiming to evaluate Large Language Models' capability in statistical and causal reasoning with real-world data. The benchmark comprises a carefully constructed dataset of 411 questions accompanied by data sheets from textbooks, online learning materials, and academic papers. To compare models' quantitative reasoning abilities on data and text, we enrich the benchmark with an auxiliary set of 290 text-only questions, namely QRText. We evaluate natural language reasoning, program-based reasoning, and agent reasoning methods including Chain-of-Thought, Program-of-Thoughts, ReAct, and code interpreter assistants on diverse models. The strongest model GPT-4 achieves an accuracy of 58%, which has much room for improvement. Among open-source models, Deepseek-coder-instruct, a code LLM pretrained on 2T tokens, gets the highest accuracy of 37%. Analysis reveals that models encounter difficulties in data analysis and causal reasoning, and struggle in using causal knowledge and provided data simultaneously. Code and data are in https://github.com/xxxiaol/QRData."
                },
                {
                    "title": "InfiAgent-DABench: Evaluating Agents on Data Analysis Tasks",
                    "abstract": "In this paper, we introduce InfiAgent-DABench, the first benchmark specifically designed to evaluate LLM-based agents on data analysis tasks. These tasks require agents to end-to-end solving complex tasks by interacting with an execution environment. This benchmark contains DAEval, a dataset consisting of 257 data analysis questions derived from 52 CSV files, and an agent framework which incorporates LLMs to serve as data analysis agents for both serving and evaluation. Since data analysis questions are often open-ended and hard to evaluate without human supervision, we adopt a format-prompting technique to convert each question into a closed-form format so that they can be automatically evaluated. Our extensive benchmarking of 34 LLMs uncovers the current challenges encountered in data analysis tasks. In addition, building on top of our agent framework, we develop a specialized agent, DAAgent, which surpasses GPT-3.5 by 3.9% on DABench. Evaluation datasets and toolkits for InfiAgent-DABench are released at https://github.com/InfiAgent/InfiAgent ."
                },
                {
                    "title": "MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts",
                    "abstract": "Large Language Models (LLMs) and Large Multimodal Models (LMMs) exhibit impressive problem-solving skills in many tasks and domains, but their ability in mathematical reasoning in visual contexts has not been systematically studied. To bridge this gap, we present MathVista, a benchmark designed to combine challenges from diverse mathematical and visual tasks. It consists of 6,141 examples, derived from 28 existing multimodal datasets involving mathematics and 3 newly created datasets (i.e., IQTest, FunctionQA, and PaperQA). Completing these tasks requires fine-grained, deep visual understanding and compositional reasoning, which all state-of-the-art foundation models find challenging. With MathVista, we have conducted a comprehensive, quantitative evaluation of 12 prominent foundation models. The best-performing GPT-4V model achieves an overall accuracy of 49.9%, substantially outperforming Bard, the second-best performer, by 15.1%. Our in-depth analysis reveals that the superiority of GPT-4V is mainly attributed to its enhanced visual perception and mathematical reasoning. However, GPT-4V still falls short of human performance by 10.4%, as it often struggles to understand complex figures and perform rigorous reasoning. This significant gap underscores the critical role that MathVista will play in the development of general-purpose AI agents capable of tackling mathematically intensive and visually rich real-world tasks. We further explore the new ability of self-verification, the application of self-consistency, and the interactive chatbot capabilities of GPT-4V, highlighting its promising potential for future research. The project is available at https://mathvista.github.io/."
                },
                {
                    "title": "Qwen Technical Report",
                    "abstract": "Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural language processing tasks that were previously thought to be exclusive to humans. In this work, we introduce Qwen, the first installment of our large language model series. Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts. It includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques. The base language models consistently demonstrate superior performance across a multitude of downstream tasks, and the chat models, particularly those trained using Reinforcement Learning from Human Feedback (RLHF), are highly competitive. The chat models possess advanced tool-use and planning capabilities for creating agent applications, showcasing impressive performance even when compared to bigger models on complex tasks like utilizing a code interpreter. Furthermore, we have developed coding-specialized models, Code-Qwen and Code-Qwen-Chat, as well as mathematics-focused models, Math-Qwen-Chat, which are built upon base language models. These models demonstrate significantly improved performance in comparison with open-source models, and slightly fall behind the proprietary models."
                },
                {
                    "title": "Efficient Memory Management for Large Language Model Serving with PagedAttention",
                    "abstract": "High throughput serving of large language models (LLMs) requires batching sufficiently many requests at a time. However, existing systems struggle because the key-value cache (KV cache) memory for each request is huge and grows and shrinks dynamically. When managed inefficiently, this memory can be significantly wasted by fragmentation and redundant duplication, limiting the batch size. To address this problem, we propose PagedAttention, an attention algorithm inspired by the classical virtual memory and paging techniques in operating systems. On top of it, we build vLLM, an LLM serving system that achieves (1) near-zero waste in KV cache memory and (2) flexible sharing of KV cache within and across requests to further reduce memory usage. Our evaluations show that vLLM improves the throughput of popular LLMs by 2--4\u00d7 with the same level of latency compared to the state-of-the-art systems, such as FasterTransformer and Orca. The improvement is more pronounced with longer sequences, larger models, and more complex decoding algorithms. vLLM's source code is publicly available at https://github.com/vllm-project/vllm."
                },
                {
                    "title": "Alpha-GPT: Human-AI Interactive Alpha Mining for Quantitative Investment",
                    "abstract": "One of the most important tasks in quantitative investment research is mining new alphas (effective trading signals or factors). Traditional alpha mining methods, either hand-crafted factor synthesizing or algorithmic factor mining (e.g., search with genetic programming), have inherent limitations, especially in implementing the ideas of quants. In this work, we propose a new alpha mining paradigm by introducing human-AI interaction, and a novel prompt engineering algorithmic framework to implement this paradigm by leveraging the power of large language models. Moreover, we develop Alpha-GPT, a new interactive alpha mining system framework that provides a heuristic way to ``understand'' the ideas of quant researchers and outputs creative, insightful, and effective alphas. We demonstrate the effectiveness and advantage of Alpha-GPT via a number of alpha mining experiments."
                },
                {
                    "title": "SciBench: Evaluating College-Level Scientific Problem-Solving Abilities of Large Language Models",
                    "abstract": "Most of the existing Large Language Model (LLM) benchmarks on scientific problem reasoning focus on problems grounded in high-school subjects and are confined to elementary algebraic operations. To systematically examine the reasoning capabilities required for solving complex scientific problems, we introduce an expansive benchmark suite SciBench for LLMs. SciBench contains a carefully curated dataset featuring a range of collegiate-level scientific problems from mathematics, chemistry, and physics domains. Based on the dataset, we conduct an in-depth benchmarking study of representative open-source and proprietary LLMs with various prompting strategies. The results reveal that the current LLMs fall short of delivering satisfactory performance, with the best overall score of merely 43.22%. Furthermore, through a detailed user study, we categorize the errors made by LLMs into ten problem-solving abilities. Our analysis indicates that no single prompting strategy significantly outperforms the others and some strategies that demonstrate improvements in certain problem-solving skills could result in declines in other skills. We envision that SciBench will catalyze further developments in the reasoning abilities of LLMs, thereby ultimately contributing to scientific research and discovery."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "A Survey on Evaluation of Large Language Models",
                    "abstract": "Large language models (LLMs) are gaining increasing popularity in both academia and industry, owing to their unprecedented performance in various applications. As LLMs continue to play a vital role in both research and daily use, their evaluation becomes increasingly critical, not only at the task level, but also at the society level for better understanding of their potential risks. Over the past years, significant efforts have been made to examine LLMs from various perspectives. This paper presents a comprehensive review of these evaluation methods for LLMs, focusing on three key dimensions: what to evaluate, where to evaluate, and how to evaluate. Firstly, we provide an overview from the perspective of evaluation tasks, encompassing general natural language processing tasks, reasoning, medical usage, ethics, education, natural and social sciences, agent applications, and other areas. Secondly, we answer the \u2018where\u2019 and \u2018how\u2019 questions by diving into the evaluation methods and benchmarks, which serve as crucial components in assessing the performance of LLMs. Then, we summarize the success and failure cases of LLMs in different tasks. Finally, we shed light on several future challenges that lie ahead in LLMs evaluation. Our aim is to offer invaluable insights to researchers in the realm of LLMs evaluation, thereby aiding the development of more proficient LLMs. Our key point is that evaluation should be treated as an essential discipline to better assist the development of LLMs. We consistently maintain the related open-source materials at: https://github.com/MLGroupJLU/LLM-eval-survey"
                },
                {
                    "title": "VerifAI: Verified Generative AI",
                    "abstract": "Generative AI has made significant strides, yet concerns about the accuracy and reliability of its outputs continue to grow. Such inaccuracies can have serious consequences such as inaccurate decision-making, the spread of false information, privacy violations, legal liabilities, and more. Although efforts to address these risks are underway, including explainable AI and responsible AI practices such as transparency, privacy protection, bias mitigation, and social and environmental responsibility, misinformation caused by generative AI will remain a significant challenge. We propose that verifying the outputs of generative AI from a data management perspective is an emerging issue for generative AI. This involves analyzing the underlying data from multi-modal data lakes, including text files, tables, and knowledge graphs, and assessing its quality and consistency. By doing so, we can establish a stronger foundation for evaluating the outputs of generative AI models. Such an approach can ensure the correctness of generative AI, promote transparency, and enable decision-making with greater confidence. Our vision is to promote the development of verifiable generative AI and contribute to a more trustworthy and responsible use of AI."
                },
                {
                    "title": "ToolQA: A Dataset for LLM Question Answering with External Tools",
                    "abstract": "Large Language Models (LLMs) have demonstrated impressive performance in various NLP tasks, but they still suffer from challenges such as hallucination and weak numerical reasoning. To overcome these challenges, external tools can be used to enhance LLMs' question-answering abilities. However, current evaluation methods do not distinguish between questions that can be answered using LLMs' internal knowledge and those that require external information through tool use. To address this issue, we introduce a new dataset called ToolQA, which is designed to faithfully evaluate LLMs' ability to use external tools for question answering. Our development of ToolQA involved a scalable, automated process for dataset curation, along with 13 specialized tools designed for interaction with external knowledge in order to answer questions. Importantly, we strive to minimize the overlap between our benchmark data and LLMs' pre-training data, enabling a more precise evaluation of LLMs' tool-use reasoning abilities. We conducted an in-depth diagnosis of existing tool-use LLMs to highlight their strengths, weaknesses, and potential improvements. Our findings set a new benchmark for evaluating LLMs and suggest new directions for future advancements. Our data and code are freely available to the broader scientific community on GitHub."
                },
                {
                    "title": "Learned Data-aware Image Representations of Line Charts for Similarity Search",
                    "abstract": "Finding line-chart images similar to a given line-chart image query is a common task in data exploration and image query systems, e.g. finding similar trends in stock markets or medical Electroencephalography images. The state-of-the-art approaches consider either data-level similarity (when the underlying data is present) or image-level similarity (when the underlying data is absent). In this paper, we study the scenario that during query time, only line-chart images are available. Our goal is to train a neural network that can turn these line-chart images into representations that are aware of the data used to generate these line charts, so as to learn better representations. Our key idea is that we can collect both data and line-chart images to learn such a neural network (at training step), while during query (or inference) time, we support the case that only line-chart images are provided. To this end, we present LineNet, a Vision Transformer-based Triplet Autoencoder model to learn data-aware image representations of line charts for similarity search. We design a novel pseudo labels selection mechanism to guide LineNet to capture both data-aware and image-level similarity of line charts. We further propose a diversified training samples selection strategy to optimize the learning process and improve the performance. We conduct both quantitative evaluation and case studies, showing that LineNet significantly outperforms the state-of-the-art methods for searching similar line-chart images."
                },
                {
                    "title": "What can Large Language Models do in chemistry? A comprehensive benchmark on eight tasks",
                    "abstract": "Large Language Models (LLMs) with strong abilities in natural language processing tasks have emerged and have been applied in various kinds of areas such as science, finance and software engineering. However, the capability of LLMs to advance the field of chemistry remains unclear. In this paper, rather than pursuing state-of-the-art performance, we aim to evaluate capabilities of LLMs in a wide range of tasks across the chemistry domain. We identify three key chemistry-related capabilities including understanding, reasoning and explaining to explore in LLMs and establish a benchmark containing eight chemistry tasks. Our analysis draws on widely recognized datasets facilitating a broad exploration of the capacities of LLMs within the context of practical chemistry. Five LLMs (GPT-4, GPT-3.5, Davinci-003, Llama and Galactica) are evaluated for each chemistry task in zero-shot and few-shot in-context learning settings with carefully selected demonstration examples and specially crafted prompts. Our investigation found that GPT-4 outperformed other models and LLMs exhibit different competitive levels in eight chemistry tasks. In addition to the key findings from the comprehensive benchmark analysis, our work provides insights into the limitation of current LLMs and the impact of in-context learning settings on LLMs' performance across various chemistry tasks. The code and datasets used in this study are available at https://github.com/ChemFoundationModels/ChemLLMBench."
                },
                {
                    "title": "Chain-of-Thought Hub: A Continuous Effort to Measure Large Language Models' Reasoning Performance",
                    "abstract": "As large language models (LLMs) are continuously being developed, their evaluation becomes increasingly important yet challenging. This work proposes Chain-of-Thought Hub, an open-source evaluation suite on the multi-step reasoning capabilities of large language models. We are interested in this setting for two reasons: (1) from the behavior of GPT and PaLM model family, we observe that complex reasoning is likely to be a key differentiator between weaker and stronger LLMs; (2) we envisage large language models to become the next-generation computational platform and foster an ecosystem of LLM-based new applications, this naturally requires the foundation models to perform complex tasks that often involve the composition of linguistic and logical operations. Our approach is to compile a suite of challenging reasoning benchmarks to track the progress of LLMs. Our current results show that: (1) model scale clearly correlates with reasoning capabilities; (2) As of May 2023, Claude-v1.3 and PaLM-2 are the only two models that are comparable with GPT-4, while open-sourced models still lag behind; (3) LLaMA-65B performs closely to code-davinci-002, indicating that with successful further development such as reinforcement learning from human feedback (RLHF), it has great potential to be close to GPT-3.5-Turbo. Our results also suggest that for the open-source efforts to catch up, the community may focus more on building better base models and exploring RLHF."
                },
                {
                    "title": "TheoremQA: A Theorem-driven Question Answering dataset",
                    "abstract": "The recent LLMs like GPT-4 and PaLM-2 have made tremendous progress in solving fundamental math problems like GSM8K by achieving over 90% accuracy. However, their capabilities to solve more challenging math problems which require domain-specific knowledge (i.e. theorem) have yet to be investigated. In this paper, we introduce TheoremQA, the first theorem-driven question-answering dataset designed to evaluate AI models' capabilities to apply theorems to solve challenging science problems. TheoremQA is curated by domain experts containing 800 high-quality questions covering 350 theorems (e.g. Taylor's theorem, Lagrange's theorem, Huffman coding, Quantum Theorem, Elasticity Theorem, etc) from Math, Physics, EE&CS, and Finance. We evaluate a wide spectrum of 16 large language and code models with different prompting strategies like Chain-of-Thoughts and Program-of-Thoughts. We found that GPT-4's capabilities to solve these problems are unparalleled, achieving an accuracy of 51% with Program-of-Thoughts Prompting. All the existing open-sourced models are below 15%, barely surpassing the random-guess baseline. Given the diversity and broad coverage of TheoremQA, we believe it can be used as a better benchmark to evaluate LLMs' capabilities to solve challenging science problems. The data and code are released in https://github.com/wenhuchen/TheoremQA."
                },
                {
                    "title": "Data Management for Machine Learning: A Survey",
                    "abstract": "Machine learning (ML) has widespread applications and has revolutionized many industries, but suffers from several challenges. First, sufficient high-quality training data is inevitable for producing a well-performed model, but the data is always human expensive to acquire. Second, a large amount of training data and complicated model structures lead to the inefficiency of training and inference. Third, given an ML task, one always needs to train lots of models, which are hard to manage in real applications. Fortunately, database techniques can benefit ML by addressing the above three challenges. In this paper, we review existing studies from the following three aspects along with the pipeline highly related to ML. (1) Data preparation (Pre-ML): it focuses on preparing high-quality training data that can improve the performance of the ML model, where we review data discovery, data cleaning and data labeling. (2) Model training & inference (In-ML): researchers in ML community focus on improving the model performance during training, while in this survey we mainly study how to accelerate the entire training process, also including feature selection and model selection. (3) Model management (Post-ML): in this part, we survey how to store, query, deploy and debug the models after training. Finally, we provide research challenges and future directions."
                },
                {
                    "title": "AGIEval: A Human-Centric Benchmark for Evaluating Foundation Models",
                    "abstract": "Evaluating the general abilities of foundation models to tackle human-level tasks is a vital aspect of their development and application in the pursuit of Artificial General Intelligence (AGI). Traditional benchmarks, which rely on artificial datasets, may not accurately represent human-level capabilities. In this paper, we introduce AGIEval, a novel benchmark specifically designed to assess foundation model in the context of human-centric standardized exams, such as college entrance exams, law school admission tests, math competitions, and lawyer qualification tests. We evaluate several state-of-the-art foundation models, including GPT-4, ChatGPT, and Text-Davinci-003, using this benchmark. Impressively, GPT-4 surpasses average human performance on SAT, LSAT, and math competitions, attaining a 95% accuracy rate on the SAT Math test and a 92.5% accuracy on the English test of the Chinese national college entrance exam. This demonstrates the extraordinary performance of contemporary foundation models. In contrast, we also find that GPT-4 is less proficient in tasks that require complex reasoning or specific domain knowledge. Our comprehensive analyses of model capabilities (understanding, knowledge, reasoning, and calculation) reveal these models' strengths and limitations, providing valuable insights into future directions for enhancing their general capabilities. By concentrating on tasks pertinent to human cognition and decision-making, our benchmark delivers a more meaningful and robust evaluation of foundation models' performance in real-world scenarios. The data, code, and all model outputs are released in https://github.com/ruixiangcui/AGIEval."
                },
                {
                    "title": "BloombergGPT: A Large Language Model for Finance",
                    "abstract": "The use of NLP in the realm of financial technology is broad and complex, with applications ranging from sentiment analysis and named entity recognition to question answering. Large Language Models (LLMs) have been shown to be effective on a variety of tasks; however, no LLM specialized for the financial domain has been reported in literature. In this work, we present BloombergGPT, a 50 billion parameter language model that is trained on a wide range of financial data. We construct a 363 billion token dataset based on Bloomberg's extensive data sources, perhaps the largest domain-specific dataset yet, augmented with 345 billion tokens from general purpose datasets. We validate BloombergGPT on standard LLM benchmarks, open financial benchmarks, and a suite of internal benchmarks that most accurately reflect our intended usage. Our mixed dataset training leads to a model that outperforms existing models on financial tasks by significant margins without sacrificing performance on general LLM benchmarks. Additionally, we explain our modeling choices, training process, and evaluation methodology. We release Training Chronicles (Appendix C) detailing our experience in training BloombergGPT."
                },
                {
                    "title": "The Role of ChatGPT in Data Science: How AI-Assisted Conversational Interfaces Are Revolutionizing the Field",
                    "abstract": "ChatGPT, a conversational AI interface that utilizes natural language processing and machine learning algorithms, is taking the world by storm and is the buzzword across many sectors today. Given the likely impact of this model on data science, through this perspective article, we seek to provide an overview of the potential opportunities and challenges associated with using ChatGPT in data science, provide readers with a snapshot of its advantages, and stimulate interest in its use for data science projects. The paper discusses how ChatGPT can assist data scientists in automating various aspects of their workflow, including data cleaning and preprocessing, model training, and result interpretation. It also highlights how ChatGPT has the potential to provide new insights and improve decision-making processes by analyzing unstructured data. We then examine the advantages of ChatGPT\u2019s architecture, including its ability to be fine-tuned for a wide range of language-related tasks and generate synthetic data. Limitations and issues are also addressed, particularly around concerns about bias and plagiarism when using ChatGPT. Overall, the paper concludes that the benefits outweigh the costs and ChatGPT has the potential to greatly enhance the productivity and accuracy of data science workflows and is likely to become an increasingly important tool for intelligence augmentation in the field of data science. ChatGPT can assist with a wide range of natural language processing tasks in data science, including language translation, sentiment analysis, and text classification. However, while ChatGPT can save time and resources compared to training a model from scratch, and can be fine-tuned for specific use cases, it may not perform well on certain tasks if it has not been specifically trained for them. Additionally, the output of ChatGPT may be difficult to interpret, which could pose challenges for decision-making in data science applications."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "A Survey of Deep Learning for Mathematical Reasoning",
                    "abstract": "Mathematical reasoning is a fundamental aspect of human intelligence and is applicable in various fields, including science, engineering, finance, and everyday life. The development of artificial intelligence (AI) systems capable of solving math problems and proving theorems in language has garnered significant interest in the fields of machine learning and natural language processing. For example, mathematics serves as a testbed for aspects of reasoning that are challenging for powerful deep learning models, driving new algorithmic and modeling advances. On the other hand, recent advances in large-scale neural language models have opened up new benchmarks and opportunities to use deep learning for mathematical reasoning. In this survey paper, we review the key tasks, datasets, and methods at the intersection of mathematical reasoning and deep learning over the past decade. We also evaluate existing benchmarks and methods, and discuss future research directions in this domain."
                },
                {
                    "title": "Dynamic Prompt Learning via Policy Gradient for Semi-structured Mathematical Reasoning",
                    "abstract": "Mathematical reasoning, a core ability of human intelligence, presents unique challenges for machines in abstract thinking and logical reasoning. Recent large pre-trained language models such as GPT-3 have achieved remarkable progress on mathematical reasoning tasks written in text form, such as math word problems (MWP). However, it is unknown if the models can handle more complex problems that involve math reasoning over heterogeneous information, such as tabular data. To fill the gap, we present Tabular Math Word Problems (TabMWP), a new dataset containing 38,431 open-domain grade-level problems that require mathematical reasoning on both textual and tabular data. Each question in TabMWP is aligned with a tabular context, which is presented as an image, semi-structured text, and a structured table. There are two types of questions: free-text and multi-choice, and each problem is annotated with gold solutions to reveal the multi-step reasoning process. We evaluate different pre-trained models on TabMWP, including the GPT-3 model in a few-shot setting. As earlier studies suggest, since few-shot GPT-3 relies on the selection of in-context examples, its performance is unstable and can degrade to near chance. The unstable issue is more severe when handling complex problems like TabMWP. To mitigate this, we further propose a novel approach, PromptPG, which utilizes policy gradient to learn to select in-context examples from a small amount of training data and then constructs the corresponding prompt for the test example. Experimental results show that our method outperforms the best baseline by 5.31% on the accuracy metric and reduces the prediction variance significantly compared to random selection, which verifies its effectiveness in selecting in-context examples."
                },
                {
                    "title": "Sevi: Speech-to-Visualization through Neural Machine Translation",
                    "abstract": "Data visualization is a powerful tool for understating information through visual cues. However, allowing novices to create visualization artifacts for what they want to see is not easy, just as not everyone can write SQL queries. Arguably, the most natural way to specify what to visualize is through natural language or speech, similar to our daily search on Google or Apple Siri, leaving to the system the task of reasoning about what to visualize and how. In this demo, we present Sevi an end-to-end data visualization system that acts as a virtual assistant to allow novices to create visualizations through either natural language or speech. Sevi is powered by two main components: Speech2Text which is based on Google Cloud Speech-to-Text Rest API, and Text2VIS, which uses an end-to-end neural machine translation model called ncNet trained using a cross-domain benchmark called nvBench. Both ncNet and nvBench have been developed by us. We will walk the audience through two general domain datasets, one related to COVID-19 and the other on NBA player statistics, to highlight how Sevi enables novices to easily create data visualizations. Because nvBench contains Text2VIS training samples from 105 domains (e.g., sport, college, hospital, etc.), the audience can play with speech or text input with any of these domains."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Steerable Self-Driving Data Visualization",
                    "abstract": "In this work, we present a self-driving data visualization system, called DeepEye, that automatically generates and recommends visualizations based on the idea of visualization by examples. We propose effective visualization recognition techniques to decide which visualizations are meaningful and visualization ranking techniques to rank the good visualizations. Furthermore, a main challenge of automatic visualization system is that the users may be misled by blindly suggesting visualizations without knowing the user's intent. To this end, we extend DeepEye to be easily steerable by allowing the user to use keyword search and providing click-based faceted navigation. Empirical results, using real-life data and use cases, verify the power of our proposed system."
                },
                {
                    "title": "Natural Language to Visualization by Neural Machine Translation",
                    "abstract": "Supporting the translation from natural language (NL) query to visualization (NL2VIS) can simplify the creation of data visualizations because if successful, anyone can generate visualizations by their natural language from the tabular data. The stateof-the-art NL2VIS approaches (e.g., NL4DV and FlowSense) are based on semantic parsers and heuristic algorithms, which are not end-to-end and are not designed for supporting (possibly) complex data transformations. Deep neural network powered neural machine translation models have made great strides in many machine translation tasks, which suggests that they might be viable for NL2VIS as well. In this paper, we present ncNet, a Transformer-based sequence-to-sequence model for supporting NL2VIS, with several novel visualization-aware optimizations, including using attention-forcing to optimize the learning process, and visualization-aware rendering to produce better visualization results. To enhance the capability of machine to comprehend natural language queries, ncNet is also designed to take an optional chart template (e.g., a pie chart or a scatter plot) as an additional input, where the chart template will be served as a constraint to limit what could be visualized. We conducted both quantitative evaluation and user study, showing that ncNet achieves good accuracy in the nvBench benchmark and is easy-to-use."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "Towards Natural Language Interfaces for Data Visualization: A Survey",
                    "abstract": "Utilizing Visualization-oriented Natural Language Interfaces (V-NLI) as a complementary input modality to direct manipulation for visual analytics can provide an engaging user experience. It enables users to focus on their tasks rather than having to worry about how to operate visualization tools on the interface. In the past two decades, leveraging advanced natural language processing technologies, numerous V-NLI systems have been developed in academic research and commercial software, especially in recent years. In this article, we conduct a comprehensive review of the existing V-NLIs. In order to classify each article, we develop categorical dimensions based on a classic information visualization pipeline with the extension of a V-NLI layer. The following seven stages are used: query interpretation, data transformation, visual mapping, view transformation, human interaction, dialogue management, and presentation. Finally, we also shed light on several promising directions for future work in the V-NLI community."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Synthesizing Natural Language to Visualization (NL2VIS) Benchmarks from NL2SQL Benchmarks",
                    "abstract": "Natural language (NL) is a promising interaction paradigm for data visualization (VIS). However, there are not any NL to VIS (NL2VIS) benchmarks available. Our goal is to provide the first NL2VIS benchmark to enable and push the field of NL2VIS, especially with deep learning technologies. In this paper, we propose a NL2VIS synthesizer (NL2SQL-to-NL2VIS) that synthesizes NL2VIS benchmarks by piggybacking NL2SQL benchmarks. The intuition is based on the semantic connection between SQL queries and VIS queries: SQL queries specify what data is needed and VIS queries additionally need to specify how to visualize. However, different from SQL that has well-defined syntax, VIS languages (e.g., Vega-Lite, VizQL, ggplot2) are syntactically very different. To provide NL2VIS benchmarks that can support many VIS languages, we use a unified intermediate representation, abstract syntax trees (ASTs), for both SQL and VIS queries. We can synthesize multiple VIS trees through adding/deleting nodes to/from an SQL tree. Each VIS tree can then be converted to (any) VIS language. The NL for VIS will be modified based on the NL for SQL to reflect corresponding tree edits. We produce the first NL2VIS benchmark (nvBench), by applying NL2SQL-to-NL2VIS on a popular NL2SQL benchmark Spider, which covers 105 domains, supports seven common types of visualizations, and contains 25,750 (NL, VIS) pairs. Our method reduces the man-hour to 5.7% of developing a NL2VIS benchmark from scratch (or building a NL2VIS benchmark from scratch takes 17.5\u00d7 man-hours of our method). Extensive human validation, through 23 experts and 312 crowd workers, demonstrates the high-quality of nvBench. In order to verify that nvBench can enable learning-based approaches, we develop a SEQ2VIS model. Our experimental results show that SEQ2VIS works well and significantly outperforms the state-of-the-art methods of the NL2VIS task."
                },
                {
                    "title": "NaturalProofs: Mathematical Theorem Proving in Natural Language",
                    "abstract": "Understanding and creating mathematics using natural mathematical language - the mixture of symbolic and natural language used by humans - is a challenging and important problem for driving progress in machine learning. As a step in this direction, we develop NaturalProofs, a multi-domain corpus of mathematical statements and their proofs, written in natural mathematical language. NaturalProofs unifies broad coverage, deep coverage, and low-resource mathematical sources, allowing for evaluating both in-distribution and zero-shot generalization. Using NaturalProofs, we benchmark strong neural methods on mathematical reference retrieval and generation tasks which test a system's ability to determine key results that appear in a proof. Large-scale sequence models show promise compared to classical information retrieval methods, yet their performance and out-of-domain generalization leave substantial room for improvement. NaturalProofs opens many avenues for research on challenging mathematical tasks."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Interactive Cleaning for Progressive Visualization through Composite Questions",
                    "abstract": "In this paper, we study the problem of interactive cleaning for progressive visualization (ICPV): Given a bad visualization V , it is to obtain a \"cleaned\" visualization V whose distance is far from V , under a given (small) budget w.r.t. human cost. In ICPV, a system interacts with a user iteratively. During each iteration, it asks the user a data cleaning question such as \"how to clean detected errors x?\", and takes value updates from the user to clean V . Conventional wisdom typically picks a single question (e.g., \"Are SIGMOD conference and SIGMOD the same?\") with the maximum expected benefit in each iteration. We propose to use a composite question \u2013 i.e., a group of single questions to be treated as one question \u2013 in each iteration (for example, Are SIGMOD conference in t1 and SIGMOD in t2 the same value, and are t1 and t2 duplicates?). A composite question is presented to the user as a small connected graph through a novel GUI that the user can directly operate on. We propose algorithms to select the best composite question in each iteration. Experiments on real-world datasets verify that composite questions are more effective than asking single questions in isolation w.r.t. the human cost."
                },
                {
                    "title": "Understanding Statistical Hypothesis Testing: The Logic of Statistical Inference",
                    "abstract": "Statistical hypothesis testing is among the most misunderstood quantitative analysis methods from data science. Despite its seeming simplicity, it has complex interdependencies between its procedural components. In this paper, we discuss the underlying logic behind statistical hypothesis testing, the formal meaning of its components and their connections. Our presentation is applicable to all statistical hypothesis tests as generic backbone and, hence, useful across all application domains in data science and artificial intelligence."
                },
                {
                    "title": "BERTScore: Evaluating Text Generation with BERT",
                    "abstract": "We propose BERTScore, an automatic evaluation metric for text generation. Analogously to common metrics, BERTScore computes a similarity score for each token in the candidate sentence with each token in the reference sentence. However, instead of exact matches, we compute token similarity using contextual embeddings. We evaluate using the outputs of 363 machine translation and image captioning systems. BERTScore correlates better with human judgments and provides stronger model selection performance than existing metrics. Finally, we use an adversarial paraphrase detection task to show that BERTScore is more robust to challenging examples when compared to existing metrics."
                },
                {
                    "title": "DeepEye: Towards Automatic Data Visualization",
                    "abstract": "Data visualization is invaluable for explaining the significance of data to people who are visually oriented. The central task of automatic data visualization is, given a dataset, to visualize its compelling stories by transforming the data (e.g., selecting attributes, grouping and binning values) and deciding the right type of visualization (e.g., bar or line charts). We present DEEPEYE, a novel system for automatic data visualization that tackles three problems: (1) Visualization recognition: given a visualization, is it \"good or \"bad\"? (2) Visualization ranking: given two visualizations, which one is \"better\"? And (3) Visualization selection: given a dataset, how to find top-k visualizations? DEEPEYE addresses (1) by training a binary classifier to decide whether a particular visualization is good or bad. It solves (2) from two perspectives: (i) Machine learning: it uses a supervised learning-to-rank model to rank visualizations; and (ii) Expert rules: it relies on experts' knowledge to specify partial orders as rules. Moreover, a \"boring\" dataset may become interesting after data transformations (e.g., binning and grouping), which forms a large search space. We also discuss optimizations to efficiently compute top-k visualizations, for approaching (3). Extensive experiments verify the effectiveness of DEEPEYE.\""
                },
                {
                    "title": "Datasheets for datasets",
                    "abstract": "Documentation to facilitate communication between dataset creators and consumers."
                },
                {
                    "title": "50 Years of Data Science",
                    "abstract": "ABSTRACT More than 50 years ago, John Tukey called for a reformation of academic statistics. In \u201cThe Future of Data Analysis,\u201d he pointed to the existence of an as-yet unrecognized science, whose subject of interest was learning from data, or \u201cdata analysis.\u201d Ten to 20 years ago, John Chambers, Jeff Wu, Bill Cleveland, and Leo Breiman independently once again urged academic statistics to expand its boundaries beyond the classical domain of theoretical statistics; Chambers called for more emphasis on data preparation and presentation rather than statistical modeling; and Breiman called for emphasis on prediction rather than inference. Cleveland and Wu even suggested the catchy name \u201cdata science\u201d for this envisioned field. A recent and growing phenomenon has been the emergence of \u201cdata science\u201d programs at major universities, including UC Berkeley, NYU, MIT, and most prominently, the University of Michigan, which in September 2015 announced a $100M \u201cData Science Initiative\u201d that aims to hire 35 new faculty. Teaching in these new programs has significant overlap in curricular subject matter with traditional statistics courses; yet many academic statisticians perceive the new programs as \u201ccultural appropriation.\u201d This article reviews some ingredients of the current \u201cdata science moment,\u201d including recent commentary about data science in the popular media, and about how/whether data science is really different from statistics. The now-contemplated field of data science amounts to a superset of the fields of statistics and machine learning, which adds some technology for \u201cscaling up\u201d to \u201cbig data.\u201d This chosen superset is motivated by commercial rather than intellectual developments. Choosing in this way is likely to miss out on the really important intellectual event of the next 50 years. Because all of science itself will soon become data that can be mined, the imminent revolution in data science is not about mere \u201cscaling up,\u201d but instead the emergence of scientific studies of data analysis science-wide. In the future, we will be able to predict how a proposal to change data analysis workflows would impact the validity of data analysis across all of science, even predicting the impacts field-by-field. Drawing on work by Tukey, Cleveland, Chambers, and Breiman, I present a vision of data science based on the activities of people who are \u201clearning from data,\u201d and I describe an academic field dedicated to improving that activity in an evidence-based manner. This new field is a better academic enlargement of statistics and machine learning than today\u2019s data science initiatives, while being able to accommodate the same short-term goals. Based on a presentation at the Tukey Centennial Workshop, Princeton, NJ, September 18, 2015."
                },
                {
                    "title": "Data Science in Statistics Curricula: Preparing Students to \u201cThink with Data\u201d",
                    "abstract": "A growing number of students are completing undergraduate degrees in statistics and entering the workforce as data analysts. In these positions, they are expected to understand how to use databases and other data warehouses, scrape data from Internet sources, program solutions to complex problems in multiple languages, and think algorithmically as well as statistically. These data science topics have not traditionally been a major component of undergraduate programs in statistics. Consequently, a curricular shift is needed to address additional learning outcomes. The goal of this article is to motivate the importance of data science proficiency and to provide examples and resources for instructors to implement data science in their own statistics curricula. We provide case studies from seven institutions. These varied approaches to teaching data science demonstrate curricular innovations to address new needs. Also included here are examples of assignments designed for courses that foster engagement of undergraduates with data and data science. [Received November 2014. Revised July 2015.]"
                },
                {
                    "title": "Gemini",
                    "abstract": "\"Would you like a car wash?\" How many minutes, if not days, have been hijacked from our lives by overzealous, up-selling machines? Chet Bowman is no exception. As pizza deliveryman extraordinaire, Chet's mission is to safely deliver Pronto Pizza products to customers in a timely manner. Then one day, his mission is in jeopardy. In just two minutes, one man takes us on a journey through the frustration and awe that is the Gemini Refueling Station."
                },
                {
                    "title": "Bleu: a Method for Automatic Evaluation of Machine Translation",
                    "abstract": "Human evaluations of machine translation are extensive but expensive. Human evaluations can take months to finish and involve human labor that can not be reused. We propose a method of automatic machine translation evaluation that is quick, inexpensive, and language-independent, that correlates highly with human evaluation, and that has little marginal cost per run. We present this method as an automated understudy to skilled human judges which substitutes for them when there is need for quick or frequent evaluations."
                },
                {
                    "title": "LILA: A Unified Benchmark for Mathematical Reasoning",
                    "abstract": "Mathematical reasoning skills are essential for general-purpose intelligentsystems to perform tasks from grocery shopping to climate modeling.Towards evaluating and improving AI systems in this domain, we proposeLILA, a unified mathematical reasoning benchmark consisting of 23 diversetasks along four dimensions:(i) mathematical abilities e.g., arithmetic, calculus (ii) language format e.g., question-answering, fill-in-the-blanks (iii) language diversity e.g., no language, simple language (iv) external knowledge e.g., commonsense, physics. We construct our benchmark by extending 20 datasets benchmark by collecting task instructions and solutions in the form of Python programs,thereby obtaining explainable solutions in addition to the correct answer.We additionally introduce two evaluation datasets to measure out-of-distribution performance and robustness to language perturbation.Finally, we introduce BHASKARA,a general-purpose mathematical reasoning model trained on LILA. Importantly, we find that multi-tasking leads to significant improvements (average relative improvement of 21.83% F1 score vs. single-task models),while the best performing model only obtains 60.40%,indicating the room for improvement in general mathematical reasoning and understanding."
                },
                {
                    "title": "DeepEye: A Data Science System for Monitoring and Exploring COVID-19 Data",
                    "abstract": "The COVID-19 pandemic is a global health crisis of our time that has signi\ufb01cantly affected almost every single person on earth in just several months. Even worse, we do not know when it will slow down and how long it will last. Analogous to \ufb01ghting the COVID-19 pandemic in the physical world, data scientists need to deal with the infodemic of COVID-19 data to discover useful insight in order to guide wise and informative decisions, where the COVID-19 infodemic refers to all (messy) data relevant to COVID-19. In this paper, we present D EEP E YE , an end-to-end data science system for monitoring and exploring COVID-19 data, which ranges from (task-driven) data preparation, (descriptive, diagnostic, and prescriptive) data analytics, user interactions through (linked) spatio-temporal data visualizations, and applications in different use cases."
                },
                {
                    "title": "Statistics for Business and Economics",
                    "abstract": "An introductory business statistics text, providing a range of computer applications related to five different software packages. It presents business-orientated examples after each new topic. A variety of case studies also reinforce the real-world applications of statistics."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we evaluate LLMs\u2019 performance in more complex and specialized statistical testing tasks?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the gap in understanding the statistical literacy of LLMs, which is essential for their effective application in data science and machine learning. By developing a benchmark like StatQA, we can facilitate future research on LLM capabilities in statistical analysis, leading to advancements in both theoretical understanding and practical applications. This could enhance the reliability of LLMs in making data-driven decisions, ultimately impacting various fields that rely on statistical analysis.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the complexity of statistical methods and the need for LLMs to not only compute results but also to select appropriate methods based on data characteristics. Naive approaches may fail because they do not account for the nuanced understanding required to assess the applicability of statistical methods. Technical obstacles include the lack of specialized datasets for benchmarking and the difficulty in designing experiments that accurately reflect the capabilities of LLMs in statistical reasoning.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the computational accuracy of LLMs without addressing their ability to select and apply appropriate statistical methods. The limitations in existing solutions stem from a lack of specialized datasets and benchmarks for evaluating LLMs in this context. Our approach differs by proposing the StatQA benchmark, which specifically targets the applicability assessment of statistical methods, thereby filling a significant gap in the literature.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology includes the development of the StatQA benchmark, which synthesizes statistical tasks and their corresponding answers to evaluate LLM performance. We will conduct systematic experiments using various LLMs, employing different prompting strategies and fine-tuning methods to assess their capabilities. The expected outcomes include a comprehensive evaluation of LLMs in statistical analysis tasks, insights into their strengths and weaknesses compared to human performance, and guidelines for improving LLM effectiveness in this specialized area."
            }
        },
        "author_data": {
            "09fb6595-c96a-432d-bf89-6f1af123b0d0": {
                "pk": "09fb6595-c96a-432d-bf89-6f1af123b0d0",
                "name": "Yizhang Zhu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "f4345806-45c5-4089-a064-d1dc4dfefc9e": {
                "pk": "f4345806-45c5-4089-a064-d1dc4dfefc9e",
                "name": "Shiyin Du",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "fe066db2-e593-4b51-be31-1b472b1f6aa6": {
                "pk": "fe066db2-e593-4b51-be31-1b472b1f6aa6",
                "name": "Boyan Li",
                "collaborators": [
                    "Yongting Chen",
                    "Dayou Zhang",
                    "Fangxin Wang",
                    "Yuyu Luo",
                    "Chengliang Chai",
                    "Guoliang Li",
                    "Nan Tang"
                ],
                "domain": [
                    "Volumetric Video Streaming",
                    "Natural Language Processing",
                    "Neural Rendering",
                    "Database Querying"
                ],
                "publications": [
                    {
                        "title": "DeformStream: Deformation-based Adaptive Volumetric Video Streaming",
                        "abstract": "Volumetric video streaming offers immersive 3D experiences but faces significant challenges due to high bandwidth requirements and latency issues in transmitting detailed content in real time. Traditional methods like point cloud streaming compromise visual quality when zoomed in, and neural rendering techniques are too computationally intensive for real-time use. Though mesh-based streaming stands out by preserving surface detail and connectivity, offering a more refined representation for 3D content, traditional mesh streaming methods typically transmit data on a per-frame basis, failing to take full advantage of temporal redundancies across frames. This results in inefficient bandwidth usage and poor adaptability to fluctuating network conditions. We introduce Deformation-based Adaptive Volumetric Video Streaming, a novel framework that enhances volumetric video streaming performance by leveraging the inherent deformability of mesh-based representations. DeformStream uses embedded deformation to reconstruct subsequent frames from inter-frame motion, significantly reducing bandwidth usage while ensuring visual coherence between frames. To address frame reconstruction overhead and network adaptability, we formulate a new QoE model that accounts for client-side deformation latency and design a dynamic programming algorithm to optimize the trade-off between visual quality and bandwidth consumption under varying network conditions. Our evaluation demonstrates that Deformation-based Adaptive Volumetric Video Streaming outperforms existing mesh-based streaming systems in both bandwidth efficiency and visual quality, offering a robust solution for real-time volumetric video applications."
                    },
                    {
                        "title": "The Dawn of Natural Language to SQL: Are We Fully Ready?",
                        "abstract": "Translating users' natural language questions into SQL queries (i.e., NL2SQL) significantly lowers the barriers to accessing relational databases. The emergence of Large Language Models has introduced a novel paradigm in NL2SQL tasks, enhancing capabilities dramatically. However, this raises a critical question: Are we fully prepared to deploy NL2SQL models in production?   To address the posed questions, we present a multi-angle NL2SQL evaluation framework, NL2SQL360, to facilitate the design and test of new NL2SQL methods for researchers. Through NL2SQL360, we conduct a detailed comparison of leading NL2SQL methods across a range of application scenarios, such as different data domains and SQL characteristics, offering valuable insights for selecting the most appropriate NL2SQL methods for specific needs. Moreover, we explore the NL2SQL design space, leveraging NL2SQL360 to automate the identification of an optimal NL2SQL solution tailored to user-specific needs. Specifically, NL2SQL360 identifies an effective NL2SQL method, SuperSQL, distinguished under the Spdier dataset using the execution accuracy metric. Remarkably, SuperSQL achieves competitive performance with execution accuracy of 87% and 62.66% on the Spider and BIRD test sets, respectively."
                    }
                ]
            },
            "9f3002bd-d0ec-494c-a173-50d0ad338983": {
                "pk": "9f3002bd-d0ec-494c-a173-50d0ad338983",
                "name": "Yuyu Luo",
                "collaborators": [
                    "Nan Tang",
                    "Guoliang Li",
                    "Leixian Shen",
                    "Boyan Li",
                    "Jianing Hao",
                    "Wei Zeng",
                    "Ju Fan",
                    "Jiawei Tang",
                    "Yupeng Xie",
                    "Chengliang Chai"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Data Visualization",
                    "Generative AI",
                    "SQL Query Translation"
                ],
                "publications": [
                    {
                        "title": "nvBench: A Large-Scale Synthesized Dataset for Cross-Domain Natural Language to Visualization Task",
                        "abstract": "NL2VIS - which translates natural language (NL) queries to corresponding visualizations (VIS) - has attracted more and more attention both in commercial visualization vendors and academic researchers. In the last few years, the advanced deep learning-based models have achieved human-like abilities in many natural language processing (NLP) tasks, which clearly tells us that the deep learning-based technique is a good choice to push the field of NL2VIS. However, a big balk is the lack of benchmarks with lots of (NL, VIS) pairs. We present nvBench, the first large-scale NL2VIS benchmark, containing 25,750 (NL, VIS) pairs from 750 tables over 105 domains, synthesized from (NL, SQL) benchmarks to support cross-domain NL2VIS task. The quality of nvBench has been extensively validated by 23 experts and 300+ crowd workers. Deep learning-based models training using nvBench demonstrate that nvBench can push the field of NL2VIS."
                    },
                    {
                        "title": "HAIChart: Human and AI Paired Visualization System",
                        "abstract": "The growing importance of data visualization in business intelligence and data science emphasizes the need for tools that can efficiently generate meaningful visualizations from large datasets. Existing tools fall into two main categories: human-powered tools (e.g., Tableau and PowerBI), which require intensive expert involvement, and AI-powered automated tools (e.g., Draco and Table2Charts), which often fall short of guessing specific user needs. In this paper, we aim to achieve the best of both worlds. Our key idea is to initially auto-generate a set of high-quality visualizations to minimize manual effort, then refine this process iteratively with user feedback to more closely align with their needs. To this end, we present HAIChart, a reinforcement learning-based framework designed to iteratively recommend good visualizations for a given dataset by incorporating user feedback. Specifically, we propose a Monte Carlo Graph Search-based visualization generation algorithm paired with a composite reward function to efficiently explore the visualization space and automatically generate good visualizations. We devise a visualization hints mechanism to actively incorporate user feedback, thus progressively refining the visualization generation module. We further prove that the top-k visualization hints selection problem is NP-hard and design an efficient algorithm. We conduct both quantitative evaluations and user studies, showing that HAIChart significantly outperforms state-of-the-art human-powered tools (21% better at Recall and 1.8 times faster) and AI-powered automatic tools (25.1% and 14.9% better in terms of Hit@3 and R10@30, respectively)."
                    },
                    {
                        "title": "The Dawn of Natural Language to SQL: Are We Fully Ready?",
                        "abstract": "Translating users' natural language questions into SQL queries (i.e., NL2SQL) significantly lowers the barriers to accessing relational databases. The emergence of Large Language Models has introduced a novel paradigm in NL2SQL tasks, enhancing capabilities dramatically. However, this raises a critical question: Are we fully prepared to deploy NL2SQL models in production?   To address the posed questions, we present a multi-angle NL2SQL evaluation framework, NL2SQL360, to facilitate the design and test of new NL2SQL methods for researchers. Through NL2SQL360, we conduct a detailed comparison of leading NL2SQL methods across a range of application scenarios, such as different data domains and SQL characteristics, offering valuable insights for selecting the most appropriate NL2SQL methods for specific needs. Moreover, we explore the NL2SQL design space, leveraging NL2SQL360 to automate the identification of an optimal NL2SQL solution tailored to user-specific needs. Specifically, NL2SQL360 identifies an effective NL2SQL method, SuperSQL, distinguished under the Spdier dataset using the execution accuracy metric. Remarkably, SuperSQL achieves competitive performance with execution accuracy of 87% and 62.66% on the Spider and BIRD test sets, respectively."
                    },
                    {
                        "title": "Data Playwright: Authoring Data Videos with Annotated Narration",
                        "abstract": "Creating data videos that effectively narrate stories with animated visuals requires substantial effort and expertise. A promising research trend is leveraging the easy-to-use natural language (NL) interaction to automatically synthesize data video components from narrative content like text narrations, or NL commands that specify user-required designs. Nevertheless, previous research has overlooked the integration of narrative content and specific design authoring commands, leading to generated results that lack customization or fail to seamlessly fit into the narrative context. To address these issues, we introduce a novel paradigm for creating data videos, which seamlessly integrates users' authoring and narrative intents in a unified format called annotated narration, allowing users to incorporate NL commands for design authoring as inline annotations within the narration text. Informed by a formative study on users' preference for annotated narration, we develop a prototype system named Data Playwright that embodies this paradigm for effective creation of data videos. Within Data Playwright, users can write annotated narration based on uploaded visualizations. The system's interpreter automatically understands users' inputs and synthesizes data videos with narration-animation interplay, powered by large language models. Finally, users can preview and fine-tune the video. A user study demonstrated that participants can effectively create data videos with Data Playwright by effortlessly articulating their desired outcomes through annotated narration."
                    },
                    {
                        "title": "Generative AI for Visualization: State of the Art and Future Directions",
                        "abstract": "Generative AI (GenAI) has witnessed remarkable progress in recent years and demonstrated impressive performance in various generation tasks in different domains such as computer vision and computational design. Many researchers have attempted to integrate GenAI into visualization framework, leveraging the superior generative capacity for different operations. Concurrently, recent major breakthroughs in GenAI like diffusion model and large language model have also drastically increase the potential of GenAI4VIS. From a technical perspective, this paper looks back on previous visualization studies leveraging GenAI and discusses the challenges and opportunities for future research. Specifically, we cover the applications of different types of GenAI methods including sequence, tabular, spatial and graph generation techniques for different tasks of visualization which we summarize into four major stages: data enhancement, visual mapping generation, stylization and interaction. For each specific visualization sub-task, we illustrate the typical data and concrete GenAI algorithms, aiming to provide in-depth understanding of the state-of-the-art GenAI4VIS techniques and their limitations. Furthermore, based on the survey, we discuss three major aspects of challenges and research opportunities including evaluation, dataset, and the gap between end-to-end GenAI and generative algorithms. By summarizing different generation algorithms, their current applications and limitations, this paper endeavors to provide useful insights for future GenAI4VIS research."
                    },
                    {
                        "title": "ChartInsights: Evaluating Multimodal Large Language Models for Low-Level Chart Question Answering",
                        "abstract": "Chart question answering (ChartQA) tasks play a critical role in interpreting and extracting insights from visualization charts. While recent advancements in multimodal large language models (MLLMs) like GPT-4o have shown promise in high-level ChartQA tasks, such as chart captioning, their effectiveness in low-level ChartQA tasks (e.g., identifying correlations) remains underexplored. In this paper, we address this gap by evaluating MLLMs on low-level ChartQA using a newly curated dataset, ChartInsights, which consists of 22,347 (chart, task, query, answer) covering 10 data analysis tasks across 7 chart types. We systematically evaluate 19 advanced MLLMs, including 12 open-source and 7 closed-source models. The average accuracy rate across these models is 39.8%, with GPT-4o achieving the highest accuracy at 69.17%. To further explore the limitations of MLLMs in low-level ChartQA, we conduct experiments that alter visual elements of charts (e.g., changing color schemes, adding image noise) to assess their impact on the task effectiveness. Furthermore, we propose a new textual prompt strategy, Chain-of-Charts, tailored for low-level ChartQA tasks, which boosts performance by 14.41%, achieving an accuracy of 83.58%. Finally, incorporating a visual prompt strategy that directs attention to relevant visual elements further improves accuracy to 84.32%."
                    },
                    {
                        "title": "VisLTR: Visualization-in-the-Loop Table Reasoning",
                        "abstract": "Table reasoning transforms user requirements into corresponding answers according to the provided table, which is often integrated with natural language interfaces for lay users to explore tabular data effortlessly. Recent research exploits large language models to facilitate table reasoning, by transforming vague user requirements into structured query languages (SQLs). However, these SQL-based approaches often overlook changes in data patterns, suffer from LLM drift, and limit exploration to only text queries. To this end, VisLTR is designed as a visualization-in-the-loop table reasoning framework that leverages visualizations as a proxy to provide concise data representations, capture interesting data patterns, and support cross-modal analysis. We describe VisLTR as a process consisting of four major modules: 1) visualization alignment that utilizes large vision-language models to align visualizations across various modalities, including chart, text, and sketch; 2) visualization referencing that decomposes a table into multifaceted visualization references that comprehensively represent the table; 3) visualization pruning that incorporates data and retrieval pruning to excise visualization references with poor information and enhance retrieval efficiency; and 4) visualization interaction that offers an interactive visual interface with multi-modal interactions for user-friendly table reasoning. Quantitative evaluation demonstrates the effectiveness of the alignment model in cross-modal visualization pairings. We further demonstrate applications of the framework on various table reasoning tasks such as table summarization and pattern detection."
                    },
                    {
                        "title": "Towards Natural Language Interfaces for Data Visualization: A Survey",
                        "abstract": "Utilizing Visualization-oriented Natural Language Interfaces (V-NLI) as a complementary input modality to direct manipulation for visual analytics can provide an engaging user experience. It enables users to focus on their tasks rather than having to worry about how to operate visualization tools on the interface. In the past two decades, leveraging advanced natural language processing technologies, numerous V-NLI systems have been developed in academic research and commercial software, especially in recent years. In this article, we conduct a comprehensive review of the existing V-NLIs. In order to classify each paper, we develop categorical dimensions based on a classic information visualization pipeline with the extension of a V-NLI layer. The following seven stages are used: query interpretation, data transformation, visual mapping, view transformation, human interaction, dialogue management, and presentation. Finally, we also shed light on several promising directions for future work in the V-NLI community."
                    },
                    {
                        "title": "VerifAI: Verified Generative AI",
                        "abstract": "Generative AI has made significant strides, yet concerns about the accuracy and reliability of its outputs continue to grow. Such inaccuracies can have serious consequences such as inaccurate decision-making, the spread of false information, privacy violations, legal liabilities, and more. Although efforts to address these risks are underway, including explainable AI and responsible AI practices such as transparency, privacy protection, bias mitigation, and social and environmental responsibility, misinformation caused by generative AI will remain a significant challenge. We propose that verifying the outputs of generative AI from a data management perspective is an emerging issue for generative AI. This involves analyzing the underlying data from multi-modal data lakes, including text files, tables, and knowledge graphs, and assessing its quality and consistency. By doing so, we can establish a stronger foundation for evaluating the outputs of generative AI models. Such an approach can ensure the correctness of generative AI, promote transparency, and enable decision-making with greater confidence. Our vision is to promote the development of verifiable generative AI and contribute to a more trustworthy and responsible use of AI."
                    },
                    {
                        "title": "A Survey of NL2SQL with Large Language Models: Where are we, and where are we going?",
                        "abstract": "Translating users' natural language queries (NL) into SQL queries (i.e., NL2SQL) can significantly reduce barriers to accessing relational databases and support various commercial applications. The performance of NL2SQL has been greatly enhanced with the emergence of Large Language Models (LLMs). In this survey, we provide a comprehensive review of NL2SQL techniques powered by LLMs, covering its entire lifecycle from the following four aspects: (1) Model: NL2SQL translation techniques that tackle not only NL ambiguity and under-specification, but also properly map NL with database schema and instances; (2) Data: From the collection of training data, data synthesis due to training data scarcity, to NL2SQL benchmarks; (3) Evaluation: Evaluating NL2SQL methods from multiple angles using different metrics and granularities; and (4) Error Analysis: analyzing NL2SQL errors to find the root cause and guiding NL2SQL models to evolve. Moreover, we provide a rule of thumb for developing NL2SQL solutions. Finally, we discuss the research challenges and open problems of NL2SQL in the LLMs era."
                    }
                ]
            },
            "1a19b729-79be-445b-a6fc-4934d4611888": {
                "pk": "1a19b729-79be-445b-a6fc-4934d4611888",
                "name": "Nan Tang",
                "collaborators": [
                    "Qing Chen",
                    "Prasenjit Mitra",
                    "Saravanan Thirumuruganathan",
                    "Mourad Ouzzani",
                    "AnHai Doan",
                    "Yupeng Xie",
                    "Yuyu Luo",
                    "Guoliang Li"
                ],
                "domain": [
                    "Graph Stream Processing",
                    "Data Curation",
                    "Data Visualization",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "On Summarizing Graph Streams",
                        "abstract": "Graph streams, which refer to the graph with edges being updated sequentially in a form of a stream, have wide applications such as cyber security, social networks and transportation networks. This paper studies the problem of summarizing graph streams. Specifically, given a graph stream G, directed or undirected, the objective is to summarize G as S with much smaller (sublinear) space, linear construction time and constant maintenance cost for each edge update, such that S allows many queries over G to be approximately conducted efficiently. Due to the sheer volume and highly dynamic nature of graph streams, summarizing them remains a notoriously hard, if not impossible, problem. The widely used practice of summarizing data streams is to treat each element independently by e.g., hash- or sampling-based method, without keeping track of the connections between elements in a data stream, which gives these summaries limited power in supporting complicated queries over graph streams. This paper discusses a fundamentally different philosophy for summarizing graph streams. We present gLava, a probabilistic graph model that, instead of treating an edge (a stream element) as the operating unit, uses the finer grained node in an element. This will naturally form a new graph sketch where edges capture the connections inside elements, and nodes maintain relationships across elements. We discuss a wide range of supported graph queries and establish theoretical error bounds for basic queries."
                    },
                    {
                        "title": "Data Curation with Deep Learning [Vision]",
                        "abstract": "Data curation - the process of discovering, integrating, and cleaning data - is one of the oldest, hardest, yet inevitable data management problems. Despite decades of efforts from both researchers and practitioners, it is still one of the most time consuming and least enjoyable work of data scientists. In most organizations, data curation plays an important role so as to fully unlock the value of big data. Unfortunately, the current solutions are not keeping up with the ever-changing data ecosystem, because they often require substantially high human cost. Meanwhile, deep learning is making strides in achieving remarkable successes in multiple areas, such as image recognition, natural language processing, and speech recognition. In this vision paper, we explore how some of the fundamental innovations in deep learning could be leveraged to improve existing data curation solutions and to help build new ones. In particular, we provide a thorough overview of the current deep learning landscape, and identify interesting research opportunities and dispel common myths. We hope that the synthesis of these important domains will unleash a series of research activities that will lead to significantly improved solutions for many data curation tasks."
                    },
                    {
                        "title": "HAIChart: Human and AI Paired Visualization System",
                        "abstract": "The growing importance of data visualization in business intelligence and data science emphasizes the need for tools that can efficiently generate meaningful visualizations from large datasets. Existing tools fall into two main categories: human-powered tools (e.g., Tableau and PowerBI), which require intensive expert involvement, and AI-powered automated tools (e.g., Draco and Table2Charts), which often fall short of guessing specific user needs. In this paper, we aim to achieve the best of both worlds. Our key idea is to initially auto-generate a set of high-quality visualizations to minimize manual effort, then refine this process iteratively with user feedback to more closely align with their needs. To this end, we present HAIChart, a reinforcement learning-based framework designed to iteratively recommend good visualizations for a given dataset by incorporating user feedback. Specifically, we propose a Monte Carlo Graph Search-based visualization generation algorithm paired with a composite reward function to efficiently explore the visualization space and automatically generate good visualizations. We devise a visualization hints mechanism to actively incorporate user feedback, thus progressively refining the visualization generation module. We further prove that the top-k visualization hints selection problem is NP-hard and design an efficient algorithm. We conduct both quantitative evaluations and user studies, showing that HAIChart significantly outperforms state-of-the-art human-powered tools (21% better at Recall and 1.8 times faster) and AI-powered automatic tools (25.1% and 14.9% better in terms of Hit@3 and R10@30, respectively)."
                    }
                ]
            }
        }
    },
    "2405.15020": {
        "paper_data": {
            "title": "AdjointDEIS: Efficient Gradients for Diffusion Models",
            "url": "http://arxiv.org/abs/2405.15020v1",
            "arxiv_id": "2405.15020",
            "authors": [
                "Zander W. Blasingame",
                "Chen Liu"
            ],
            "abstract": "The optimization of the latents and parameters of diffusion models with respect to some differentiable metric defined on the output of the model is a challenging and complex problem. The sampling for diffusion models is done by solving either the probability flow ODE or diffusion SDE wherein a neural network approximates the score function or related quantity, allowing a numerical ODE/SDE solver to be used. However, na\\\"ive backpropagation techniques are memory intensive, requiring the storage of all intermediate states, and face additional complexity in handling the injected noise from the diffusion term of the diffusion SDE. We propose a novel method based on the stochastic adjoint sensitivity method to calculate the gradientwith respect to the initial noise, conditional information, and model parameters by solving an additional SDE whose solution is the gradient of the diffusion SDE. We exploit the unique construction of diffusion SDEs to further simplify the formulation of the adjoint diffusion SDE and use a change-of-variables to simplify the solution to an exponentially weighted integral. Using this formulation we derive a custom solver for the adjoint SDE as well as the simpler adjoint ODE. The proposed adjoint diffusion solvers can efficiently compute the gradients for both the probability flow ODE and diffusion SDE for latents and parameters of the model. Lastly, we demonstrate the effectiveness of the adjoint diffusion solvers onthe face morphing problem.",
            "introduction": "   1 Introduction  Diffusion models are a large family of state-of-the-art generative models which learn to map samples drawn from white Gaussian noise into the data distribution\u00a0[1, 2]. These diffusion models have achieved state-of-the-art performance on prominent tasks such as image generation\u00a0[3, 4, 5], audio generation\u00a0[6], or video generation\u00a0[7]. Often the state-of-the-art models are quite large and training them is prohibitively expensive\u00a0[8]. As such it is fairly common to adapt a pre-trained model to a specific task for post-training. This way the generative model can learn new concepts, identities, or tasks without needing to training the whole model\u00a0[9, 10, 11]. Additional work has also proposed algorithms for guiding the generative process of the diffusion models\u00a0[12, 13].   One method to guide or direct the generative process is to solve an optimization problem w.r.t.\u00a0some loss function \u2112\u2112\\mathcal{L}caligraphic_L defined on the image space \ud835\udcb3\u2286\u211dn\ud835\udcb3superscript\u211d\ud835\udc5b\\mathcal{X}\\subseteq\\mathbb{R}^{n}caligraphic_X \u2286 blackboard_R start_POSTSUPERSCRIPT italic_n end_POSTSUPERSCRIPT. This loss functions works on the output of the diffusion model and assess how \u201cgood\u201d the output is. However, the diffusion model works by iteratively removing noise until a clean sample is reached. As such we need to be able to efficiently backpropagate gradients through the entire generative process. As\u00a0Song et\u00a0al. [14] showed the diffusion SDE can be simplified to an associated ODE, and as such many efficient ODE/SDE solvers have been developed for diffusion models\u00a0[15, 16, 17]. However, na\u00efvely applying backpropagation to the diffusion model is inflexible and memory intensive; moreover, such an approach is not trivial to apply to the diffusion models that used an SDE solver instead of an ODE solver.   Inspired by the work of\u00a0Chen et\u00a0al. [18] and their application of the method of adjoint sensitivity to model ODEs with neural networks, we propose AdjointDEIS a novel technique for calculating the gradient of the noisy states, and other parameters, of the diffusion model using the method of adjoint sensitivity. We exploit the unique construction of diffusion models to greatly simplify the adjoint ODE into a problem with exponential integrators which reduces the discretization errors of the adjoint ODE. We then design a custom second-order multi-step method for solving the adjoint ODE. Additionally, we extend this technique to work for diffusion SDEs and not just ODEs. Lastly, we evaluate the proposed techniques on the face morphing problem to illustrate the use of this technique for guided generation problems. The contribution of this work can be summarized as follows:   1.  We propose a method for calculating gradients of diffusion models.    2.  To the best of our knowledge, AdjointDEIS is the first general backpropagation technique for diffusion models that use an SDE solver for sampling and can supply gradients for network weights, conditional information, and intermediate noisy states.    3.  We develop custom first and second order solvers for the adjoint ODE/SDE.    4.  We illustrate the effectiveness of AdjointDEIS for guided generation on the face morphing problem.        2 Diffusion Models  In this section we provide a brief overview of diffusion models. Diffusion models learn a generative process by first perturbing the data distribution into an isotropic Gaussian by progressively adding Gaussian noise to the data distribution, then a neural network is trained to preform denoising steps, allowing for sampling of the data distribution via sampling of a Gaussian distribution\u00a0[2, 8]. Assume we have an",
            "references": [
                {
                    "title": "The Blessing of Randomness: SDE Beats ODE in General Diffusion-based Image Editing",
                    "abstract": "We present a unified probabilistic formulation for diffusion-based image editing, where a latent variable is edited in a task-specific manner and generally deviates from the corresponding marginal distribution induced by the original stochastic or ordinary differential equation (SDE or ODE). Instead, it defines a corresponding SDE or ODE for editing. In the formulation, we prove that the Kullback-Leibler divergence between the marginal distributions of the two SDEs gradually decreases while that for the ODEs remains as the time approaches zero, which shows the promise of SDE in image editing. Inspired by it, we provide the SDE counterparts for widely used ODE baselines in various tasks including inpainting and image-to-image translation, where SDE shows a consistent and substantial improvement. Moreover, we propose SDE-Drag -- a simple yet effective method built upon the SDE formulation for point-based content dragging. We build a challenging benchmark (termed DragBench) with open-set natural, art, and AI-generated images for evaluation. A user study on DragBench indicates that SDE-Drag significantly outperforms our ODE baseline, existing diffusion-based methods, and the renowned DragGAN. Our results demonstrate the superiority and versatility of SDE in image editing and push the boundary of diffusion-based editing methods."
                },
                {
                    "title": "A Latent Space of Stochastic Diffusion Models for Zero-Shot Image Editing and Guidance",
                    "abstract": "Diffusion models generate images by iterative denoising. Recent work has shown that by making the denoising process deterministic, one can encode real images into latent codes of the same size, which can be used for image editing. This paper explores the possibility of defining a latent space even when the denoising process remains stochastic. Recall that, in stochastic diffusion models, Gaussian noises are added in each denoising step, and we can concatenate all the noises to form a latent code. This results in a latent space of much higher dimensionality than the original image. We demonstrate that this latent space of stochastic diffusion models can be used in the same way as that of deterministic diffusion models in two applications. First, we propose CycleDiffusion, a method for zero-shot and unpaired image editing using stochastic diffusion models, which improves the performance over its deterministic counterpart. Second, we demonstrate unified, plug-and-play guidance in the latent spaces of deterministic and stochastic diffusion models.1"
                },
                {
                    "title": "AdjointDPM: Adjoint Sensitivity Method for Gradient Backpropagation of Diffusion Probabilistic Models",
                    "abstract": "Existing customization methods require access to multiple reference examples to align pre-trained diffusion probabilistic models (DPMs) with user-provided concepts. This paper aims to address the challenge of DPM customization when the only available supervision is a differentiable metric defined on the generated contents. Since the sampling procedure of DPMs involves recursive calls to the denoising UNet, na\\\"ive gradient backpropagation requires storing the intermediate states of all iterations, resulting in extremely high memory consumption. To overcome this issue, we propose a novel method AdjointDPM, which first generates new samples from diffusion models by solving the corresponding probability-flow ODEs. It then uses the adjoint sensitivity method to backpropagate the gradients of the loss to the models' parameters (including conditioning signals, network weights, and initial noises) by solving another augmented ODE. To reduce numerical errors in both the forward generation and gradient backpropagation processes, we further reparameterize the probability-flow ODE and augmented ODE as simple non-stiff ODEs using exponential integration. Finally, we demonstrate the effectiveness of AdjointDPM on three interesting tasks: converting visual effects into identification text embeddings, finetuning DPMs for specific types of stylization, and optimizing initial noise to generate adversarial samples for security auditing."
                },
                {
                    "title": "SEEDS: Exponential SDE Solvers for Fast High-Quality Sampling from Diffusion Models",
                    "abstract": "A potent class of generative models known as Diffusion Probabilistic Models (DPMs) has become prominent. A forward diffusion process adds gradually noise to data, while a model learns to gradually denoise. Sampling from pre-trained DPMs is obtained by solving differential equations (DE) defined by the learnt model, a process which has shown to be prohibitively slow. Numerous efforts on speeding-up this process have consisted on crafting powerful ODE solvers. Despite being quick, such solvers do not usually reach the optimal quality achieved by available slow SDE solvers. Our goal is to propose SDE solvers that reach optimal quality without requiring several hundreds or thousands of NFEs to achieve that goal. We propose Stochastic Explicit Exponential Derivative-free Solvers (SEEDS), improving and generalizing Exponential Integrator approaches to the stochastic case on several frameworks. After carefully analyzing the formulation of exact solutions of diffusion SDEs, we craft SEEDS to analytically compute the linear part of such solutions. Inspired by the Exponential Time-Differencing method, SEEDS use a novel treatment of the stochastic components of solutions, enabling the analytical computation of their variance, and contains high-order terms allowing to reach optimal quality sampling $\\sim3$-$5\\times$ faster than previous SDE methods. We validate our approach on several image generation benchmarks, showing that SEEDS outperform or are competitive with previous SDE solvers. Contrary to the latter, SEEDS are derivative and training free, and we fully prove strong convergence guarantees for them."
                },
                {
                    "title": "Align Your Latents: High-Resolution Video Synthesis with Latent Diffusion Models",
                    "abstract": "Latent Diffusion Models (LDMs) enable high-quality image synthesis while avoiding excessive compute demands by training a diffusion model in a compressed lower-dimensional latent space. Here, we apply the LDM paradigm to high-resolution video generation, a particularly resource-intensive task. We first pre-train an LDM on images only; then, we turn the image generator into a video generator by introducing a temporal dimension to the latent space diffusion model and finetuning on encoded image sequences, i.e., videos. Similarly, we temporally align diffusion model upsamplers, turning them into temporally consistent video super resolution models. We focus on two relevant real-world applications: Simulation of in-the-wild driving data and creative content creation with text-to-video modeling. In particular, we validate our Video LDM on real driving videos of resolution $512 \\times 1024$, achieving state-of-the-art performance. Furthermore, our approach can easily leverage off-the-shelf pretrained image LDMs, as we only need to train a temporal alignment model in that case. Doing so, we turn the publicly available, state-of-the-art text-to-image LDM Stable Diffusion into an efficient and expressive text-to-video model with resolution up to $1280 \\times 2048$. We show that the temporal layers trained in this way generalize to different finetuned text-to-image LDMs. Utilizing this property, we show the first results for personalized text-to-video generation, opening exciting directions for future content creation. Project page: https://nv-tlabs.github.io/VideoLDM/"
                },
                {
                    "title": "End-to-End Diffusion Latent Optimization Improves Classifier Guidance",
                    "abstract": "Classifier guidance\u2014using the gradients of an image classifier to steer the generations of a diffusion model\u2014has the potential to dramatically expand the creative control over image generation and editing. However, currently classifier guidance requires either training new noise-aware models to obtain accurate gradients or using a one-step denoising approximation of the final generation, which leads to misaligned gradients and sub-optimal control. We highlight this approximation\u2019s shortcomings and propose a novel guidance method: Direct Optimization of Diffusion Latents (DOODL), which enables plug-and-play guidance by optimizing diffusion latents w.r.t. the gradients of a pre-trained classifier on the true generated pixels, using an invertible diffusion process to achieve memory-efficient backpropagation. Showcasing the potential of more precise guidance, DOODL outperforms one-step classifier guidance on computational and human evaluation metrics across different forms of guidance: using CLIP guidance to improve generations of complex prompts from DrawBench, using fine-grained visual classifiers to expand the vocabulary of Stable Diffusion, enabling image-conditioned generation with a CLIP visual encoder, and improving image aesthetics using an aesthetic scoring network."
                },
                {
                    "title": "FreeDoM: Training-Free Energy-Guided Conditional Diffusion Model",
                    "abstract": "Recently, conditional diffusion models have gained popularity in numerous applications due to their exceptional generation ability. However, many existing methods are training-required. They need to train a time-dependent classifier or a condition-dependent score estimator, which increases the cost of constructing conditional diffusion models and is inconvenient to transfer across different conditions. Some current works aim to overcome this limitation by proposing training-free solutions, but most can only be applied to a specific category of tasks and not to more general conditions. In this work, we propose a training-Free conditional Diffusion Model (FreeDoM) used for various conditions. Specifically, we leverage off-the-shelf pretrained networks, such as a face detection model, to construct time-independent energy functions, which guide the generation process without requiring training. Furthermore, because the construction of the energy function is very flexible and adaptable to various conditions, our proposed FreeDoM has a broader range of applications than existing training-free methods. FreeDoM is advantageous in its simplicity, effectiveness, and low cost. Experiments demonstrate that FreeDoM is effective for various conditions and suitable for diffusion models of diverse data domains, including image and latent code domains. Code is available at https://github.com/vvictoryuki/FreeDoM."
                },
                {
                    "title": "Consistency Models",
                    "abstract": "Diffusion models have significantly advanced the fields of image, audio, and video generation, but they depend on an iterative sampling process that causes slow generation. To overcome this limitation, we propose consistency models, a new family of models that generate high quality samples by directly mapping noise to data. They support fast one-step generation by design, while still allowing multistep sampling to trade compute for sample quality. They also support zero-shot data editing, such as image inpainting, colorization, and super-resolution, without requiring explicit training on these tasks. Consistency models can be trained either by distilling pre-trained diffusion models, or as standalone generative models altogether. Through extensive experiments, we demonstrate that they outperform existing distillation techniques for diffusion models in one- and few-step sampling, achieving the new state-of-the-art FID of 3.55 on CIFAR-10 and 6.20 on ImageNet 64x64 for one-step generation. When trained in isolation, consistency models become a new family of generative models that can outperform existing one-step, non-adversarial generative models on standard benchmarks such as CIFAR-10, ImageNet 64x64 and LSUN 256x256."
                },
                {
                    "title": "Universal Guidance for Diffusion Models",
                    "abstract": "Typical diffusion models are trained to accept a particular form of conditioning, most commonly text, and cannot be conditioned on other modalities without retraining. In this work, we propose a universal guidance algorithm that enables diffusion models to be controlled by arbitrary guidance modalities without the need to retrain any use-specific components. We show that our algorithm successfully generates quality images with guidance functions including segmentation, face recognition, object detection, and classifier signals. Code is available at github.com/arpitbansal297/Universal-Guided-Diffusion."
                },
                {
                    "title": "AudioLDM: Text-to-Audio Generation with Latent Diffusion Models",
                    "abstract": "Text-to-audio (TTA) system has recently gained attention for its ability to synthesize general audio based on text descriptions. However, previous studies in TTA have limited generation quality with high computational costs. In this study, we propose AudioLDM, a TTA system that is built on a latent space to learn the continuous audio representations from contrastive language-audio pretraining (CLAP) latents. The pretrained CLAP models enable us to train LDMs with audio embedding while providing text embedding as a condition during sampling. By learning the latent representations of audio signals and their compositions without modeling the cross-modal relationship, AudioLDM is advantageous in both generation quality and computational efficiency. Trained on AudioCaps with a single GPU, AudioLDM achieves state-of-the-art TTA performance measured by both objective and subjective metrics (e.g., frechet distance). Moreover, AudioLDM is the first TTA system that enables various text-guided audio manipulations (e.g., style transfer) in a zero-shot fashion. Our implementation and demos are available at https://audioldm.github.io."
                },
                {
                    "title": "Leveraging Diffusion for Strong and High Quality Face Morphing Attacks",
                    "abstract": "Face morphing attacks seek to deceive a Face Recognition (FR) system by presenting a morphed image consisting of the biometric qualities from two different identities with the aim of triggering a false acceptance with one of the two identities, thereby presenting a significant threat to biometric systems. The success of a morphing attack is dependent on the ability of the morphed image to represent the biometric characteristics of both identities that were used to create the image. We present a novel morphing attack that uses a Diffusion-based architecture to improve the visual fidelity of the image and the ability of the morphing attack to represent characteristics from both identities. We demonstrate the effectiveness of the proposed attack by evaluating its visual fidelity via Fr\u00e9chet Inception Distance (FID). Also, extensive experiments are conducted to measure the vulnerability of FR systems to the proposed attack. The ability of a morphing attack detector to detect the proposed attack is measured and compared against two state-of-the-art GAN-based morphing attacks along with two Landmark-based attacks. Additionally, a novel metric to measure the relative strength between different morphing attacks is introduced and evaluated."
                },
                {
                    "title": "DPM-Solver++: Fast Solver for Guided Sampling of Diffusion Probabilistic Models",
                    "abstract": "Diffusion probabilistic models (DPMs) have achieved impressive success in high-resolution image synthesis, especially in recent large-scale text-to-image generation applications. An essential technique for improving the sample quality of DPMs is guided sampling, which usually needs a large guidance scale to obtain the best sample quality. The commonly-used fast sampler for guided sampling is DDIM, a first-order diffusion ODE solver that generally needs 100 to 250 steps for high-quality samples. Although recent works propose dedicated high-order solvers and achieve a further speedup for sampling without guidance, their effectiveness for guided sampling has not been well-tested before. In this work, we demonstrate that previous high-order fast samplers suffer from instability issues, and they even become slower than DDIM when the guidance scale grows large. To further speed up guided sampling, we propose DPM-Solver++, a high-order solver for the guided sampling of DPMs. DPM-Solver++ solves the diffusion ODE with the data prediction model and adopts thresholding methods to keep the solution matches training data distribution. We further propose a multistep variant of DPM-Solver++ to address the instability issue by reducing the effective step size. Experiments show that DPM-Solver++ can generate high-quality samples within only 15 to 20 steps for guided sampling by pixel-space and latent-space DPMs."
                },
                {
                    "title": "DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation",
                    "abstract": "Large text-to-image models achieved a remarkable leap in the evolution of AI, enabling high-quality and diverse synthesis of images from a given text prompt. However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts. In this work, we present a new approach for \u201cpersonalization\u201d of text-to-image diffusion models. Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject. Once the subject is embedded in the output domain of the model, the unique identifier can be used to synthesize novel photorealistic images of the subject contextualized in different scenes. By leveraging the semantic prior embedded in the model with a new autogenous class-specific prior preservation loss, our technique enables synthesizing the subject in diverse scenes, poses, views and lighting conditions that do not appear in the reference images. We apply our technique to several previously-unassailable tasks, including subject recontextualization, text-guided view synthesis, and artistic rendering, all while preserving the subject's key features. We also provide a new dataset and evaluation protocol for this new task of subject-driven generation. Project page: https://dreambooth.github.io/"
                },
                {
                    "title": "SYN-MAD 2022: Competition on Face Morphing Attack Detection Based on Privacy-aware Synthetic Training Data",
                    "abstract": "This paper presents a summary of the Competition on Face Morphing Attack Detection Based on Privacy-aware Synthetic Training Data (SYN-MAD) held at the 2022 In-ternational Joint Conference on Biometrics (IJCB 2022). The competition attracted a total of 12 participating teams, both from academia and industry and present in 11 differ-ent countries. In the end, seven valid submissions were submitted by the participating teams and evaluated by the organizers. The competition was held to present and at-tract solutions that deal with detecting face morphing at-tacks while protecting people's privacy for ethical and le-gal reasons. To ensure this, the training data was limited to synthetic data provided by the organizers. The submitted solutions presented innovations that led to out-performing the considered baseline in many experimental settings. The evaluation benchmark is now available at: https://github.com/marcohuber/SYN-MAD-2022."
                },
                {
                    "title": "An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion",
                    "abstract": "Text-to-image models offer unprecedented freedom to guide creation through natural language. Yet, it is unclear how such freedom can be exercised to generate images of specific unique concepts, modify their appearance, or compose them in new roles and novel scenes. In other words, we ask: how can we use language-guided models to turn our cat into a painting, or imagine a new product based on our favorite toy? Here we present a simple approach that allows such creative freedom. Using only 3-5 images of a user-provided concept, like an object or a style, we learn to represent it through new\"words\"in the embedding space of a frozen text-to-image model. These\"words\"can be composed into natural language sentences, guiding personalized creation in an intuitive way. Notably, we find evidence that a single word embedding is sufficient for capturing unique and varied concepts. We compare our approach to a wide range of baselines, and demonstrate that it can more faithfully portray the concepts across a range of applications and tasks. Our code, data and new words will be available at: https://textual-inversion.github.io"
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps",
                    "abstract": "Diffusion probabilistic models (DPMs) are emerging powerful generative models. Despite their high-quality generation performance, DPMs still suffer from their slow sampling as they generally need hundreds or thousands of sequential function evaluations (steps) of large neural networks to draw a sample. Sampling from DPMs can be viewed alternatively as solving the corresponding diffusion ordinary differential equations (ODEs). In this work, we propose an exact formulation of the solution of diffusion ODEs. The formulation analytically computes the linear part of the solution, rather than leaving all terms to black-box ODE solvers as adopted in previous works. By applying change-of-variable, the solution can be equivalently simplified to an exponentially weighted integral of the neural network. Based on our formulation, we propose DPM-Solver, a fast dedicated high-order solver for diffusion ODEs with the convergence order guarantee. DPM-Solver is suitable for both discrete-time and continuous-time DPMs without any further training. Experimental results show that DPM-Solver can generate high-quality samples in only 10 to 20 function evaluations on various datasets. We achieve 4.70 FID in 10 function evaluations and 2.87 FID in 20 function evaluations on the CIFAR10 dataset, and a $4\\sim 16\\times$ speedup compared with previous state-of-the-art training-free samplers on various datasets."
                },
                {
                    "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                    "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                },
                {
                    "title": "Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding",
                    "abstract": "We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding. Imagen builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation. Our key discovery is that generic large language models (e.g. T5), pretrained on text-only corpora, are surprisingly effective at encoding text for image synthesis: increasing the size of the language model in Imagen boosts both sample fidelity and image-text alignment much more than increasing the size of the image diffusion model. Imagen achieves a new state-of-the-art FID score of 7.27 on the COCO dataset, without ever training on COCO, and human raters find Imagen samples to be on par with the COCO data itself in image-text alignment. To assess text-to-image models in greater depth, we introduce DrawBench, a comprehensive and challenging benchmark for text-to-image models. With DrawBench, we compare Imagen with recent methods including VQ-GAN+CLIP, Latent Diffusion Models, and DALL-E 2, and find that human raters prefer Imagen over other models in side-by-side comparisons, both in terms of sample quality and image-text alignment. See https://imagen.research.google/ for an overview of the results."
                },
                {
                    "title": "Are GAN-based morphs threatening face recognition?",
                    "abstract": "Morphing attacks are a threat to biometric systems where the biometric reference in an identity document can be altered. This form of attack presents an important issue in applications relying on identity documents such as border security or access control. Research in generation of face morphs and their detection is developing rapidly, however very few datasets with morphing attacks and open-source detection toolkits are publicly available. This paper bridges this gap by providing two datasets and the corresponding code for four types of morphing attacks: two that rely on facial landmarks based on OpenCV and FaceMorpher, and two that use StyleGAN 2 to generate synthetic morphs. We also conduct extensive experiments to assess the vulnerability of four state-of-the-art face recognition systems, including FaceNet, VGG-Face, ArcFace, and ISV. Surprisingly, the experiments demonstrate that, although visually more appealing, morphs based on StyleGAN 2 do not pose a significant threat to the state to face recognition systems, as these morphs were outmatched by the simple morphs that are based facial landmarks."
                },
                {
                    "title": "Fast Sampling of Diffusion Models with Exponential Integrator",
                    "abstract": "The past few years have witnessed the great success of Diffusion models~(DMs) in generating high-fidelity samples in generative modeling tasks. A major limitation of the DM is its notoriously slow sampling procedure which normally requires hundreds to thousands of time discretization steps of the learned diffusion process to reach the desired accuracy. Our goal is to develop a fast sampling method for DMs with a much less number of steps while retaining high sample quality. To this end, we systematically analyze the sampling procedure in DMs and identify key factors that affect the sample quality, among which the method of discretization is most crucial. By carefully examining the learned diffusion process, we propose Diffusion Exponential Integrator Sampler~(DEIS). It is based on the Exponential Integrator designed for discretizing ordinary differential equations (ODEs) and leverages a semilinear structure of the learned diffusion process to reduce the discretization error. The proposed method can be applied to any DMs and can generate high-fidelity samples in as few as 10 steps. In our experiments, it takes about 3 minutes on one A6000 GPU to generate $50k$ images from CIFAR10. Moreover, by directly using pre-trained DMs, we achieve the state-of-art sampling performance when the number of score function evaluation~(NFE) is limited, e.g., 4.17 FID with 10 NFEs, 3.37 FID, and 9.74 IS with only 15 NFEs on CIFAR10. Code is available at https://github.com/qsh-zh/deis"
                },
                {
                    "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
                    "abstract": "Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples."
                },
                {
                    "title": "AdaFace: Quality Adaptive Margin for Face Recognition",
                    "abstract": "Recognition in low quality face datasets is challenging because facial attributes are obscured and degraded. Advances in margin-based loss functions have resulted in enhanced discriminability of faces in the embedding space. Further, previous studies have studied the effect of adaptive losses to assign more importance to misclassified (hard) examples. In this work, we introduce another aspect of adaptiveness in the loss function, namely the image quality. We argue that the strategy to emphasize misclassified samples should be adjusted according to their image quality. Specifically, the relative importance of easy or hard samples should be based on the sample's image quality. We propose a new loss function that emphasizes samples of different difficulties based on their image quality. Our method achieves this in the form of an adaptive margin function by approximating the image quality with feature norms. Extensive experiments show that our method, AdaFace, improves the face recognition performance over the state-of-the-art (SoTA) on four datasets (IJB-B, IJB-C, IJB-S and TinyFace). Code and models are released in Supp."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Diffusion Autoencoders: Toward a Meaningful and Decodable Representation",
                    "abstract": "Diffusion probabilistic models (DPMs) have achieved remarkable quality in image generation that rivals GANs'. But unlike GANs, DPMs use a set of latent variables that lack semantic meaning and cannot serve as a useful representation for other tasks. This paper explores the possibility of using DPMs for representation learning and seeks to extract a meaningful and decodable representation of an input image via autoencoding. Our key idea is to use a learnable encoder for discovering the high-level semantics, and a DPM as the decoder for modeling the remaining stochastic variations. Our method can encode any image into a two-part latent code where the first part is semantically meaningful and linear, and the second part captures stochastic details, allowing near-exact reconstruction. This capability enables challenging applications that currently foil GAN-based methods, such as attribute manipulation on real images. We also show that this two-level encoding improves denoising efficiency and naturally facilitates various downstream tasks including few-shot conditional sampling. Please visit our page: https://Diff-AE.github.io/"
                },
                {
                    "title": "ElasticFace: Elastic Margin Loss for Deep Face Recognition",
                    "abstract": "Learning discriminative face features plays a major role in building high-performing face recognition models. The recent state-of-the-art face recognition solutions proposed to incorporate a fixed penalty margin on commonly used classification loss function, softmax loss, in the normalized hypersphere to increase the discriminative power of face recognition models, by minimizing the intra-class variation and maximizing the inter-class variation. Marginal penalty softmax losses, such as ArcFace and CosFace, assume that the geodesic distance between and within the different identities can be equally learned using a fixed penalty margin. However, such a learning objective is not realistic for real data with inconsistent inter-and intra-class variation, which might limit the discriminative and generalizability of the face recognition model. In this paper, we relax the fixed penalty margin constrain by proposing elastic penalty margin loss (ElasticFace) that allows flexibility in the push for class separability. The main idea is to utilize random margin values drawn from a normal distribution in each training iteration. This aims at giving the decision boundary chances to extract and retract to allow space for flexible class separability learning. We demonstrate the superiority of our ElasticFace loss over ArcFace and CosFace losses, using the same geometric transformation, on a large set of mainstream benchmarks. From a wider perspective, our ElasticFace has advanced the state-of-the-art face recognition performance on seven out of nine mainstream benchmarks. All training codes, pre-trained models, training logs will be publicly released 1."
                },
                {
                    "title": "SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations",
                    "abstract": "Guided image synthesis enables everyday users to create and edit photo-realistic images with minimum effort. The key challenge is balancing faithfulness to the user input (e.g., hand-drawn colored strokes) and realism of the synthesized image. Existing GAN-based methods attempt to achieve such balance using either conditional GANs or GAN inversions, which are challenging and often require additional training data or loss functions for individual applications. To address these issues, we introduce a new image synthesis and editing method, Stochastic Differential Editing (SDEdit), based on a diffusion model generative prior, which synthesizes realistic images by iteratively denoising through a stochastic differential equation (SDE). Given an input image with user guide of any type, SDEdit first adds noise to the input, then subsequently denoises the resulting image through the SDE prior to increase its realism. SDEdit does not require task-specific training or inversions and can naturally achieve the balance between realism and faithfulness. SDEdit significantly outperforms state-of-the-art GAN-based methods by up to 98.09% on realism and 91.72% on overall satisfaction scores, according to a human perception study, on multiple tasks, including stroke-based image synthesis and editing as well as image compositing."
                },
                {
                    "title": "MIPGAN\u2014Generating Strong and High Quality Morphing Attacks Using Identity Prior Driven GAN",
                    "abstract": "Face morphing attacks target to circumvent Face Recognition Systems (FRS) by employing face images derived from multiple data subjects (e.g., accomplices and malicious actors). Morphed images can be verified against contributing data subjects with a reasonable success rate, given they have a high degree of facial resemblance. The success of morphing attacks is directly dependent on the quality of the generated morph images. We present a new approach for generating strong attacks extending our earlier framework for generating face morphs. We present a new approach using an Identity Prior Driven Generative Adversarial Network, which we refer to as MIPGAN (Morphing through Identity Prior driven GAN). The proposed MIPGAN is derived from the StyleGAN with a newly formulated loss function exploiting perceptual quality and identity factor to generate a high quality morphed facial image with minimal artefacts and with high resolution. We demonstrate the proposed approach\u2019s applicability to generate strong morphing attacks by evaluating its vulnerability against both commercial and deep learning based Face Recognition System (FRS) and demonstrate the success rate of attacks. Extensive experiments are carried out to assess the FRS\u2019s vulnerability against the proposed morphed face generation technique on three types of data such as digital images, re-digitized (printed and scanned) images, and compressed images after re-digitization from newly generated MIPGAN Face Morph Dataset. The obtained results demonstrate that the proposed approach of morph generation poses a high threat to FRS."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Efficient and Accurate Gradients for Neural SDEs",
                    "abstract": "Neural SDEs combine many of the best qualities of both RNNs and SDEs: memory efficient training, high-capacity function approximation, and strong priors on model space. This makes them a natural choice for modelling many types of temporal dynamics. Training a Neural SDE (either as a VAE or as a GAN) requires backpropagating through an SDE solve. This may be done by solving a backwards-in-time SDE whose solution is the desired parameter gradients. However, this has previously suffered from severe speed and accuracy issues, due to high computational cost and numerical truncation errors. Here, we overcome these issues through several technical innovations. First, we introduce the \\textit{reversible Heun method}. This is a new SDE solver that is \\textit{algebraically reversible}: eliminating numerical gradient errors, and the first such solver of which we are aware. Moreover it requires half as many function evaluations as comparable solvers, giving up to a $1.98\\times$ speedup. Second, we introduce the \\textit{Brownian Interval}: a new, fast, memory efficient, and exact way of sampling \\textit{and reconstructing} Brownian motion. With this we obtain up to a $10.6\\times$ speed improvement over previous techniques, which in contrast are both approximate and relatively slow. Third, when specifically training Neural SDEs as GANs (Kidger et al. 2021), we demonstrate how SDE-GANs may be trained through careful weight clipping and choice of activation function. This reduces computational cost (giving up to a $1.87\\times$ speedup) and removes the numerical truncation errors associated with gradient penalty. Altogether, we outperform the state-of-the-art by substantial margins, with respect to training speed, and with respect to classification, prediction, and MMD test metrics. We have contributed implementations of all of our techniques to the torchsde library to help facilitate their adoption."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Partial FC: Training 10 Million Identities on a Single Machine",
                    "abstract": "Face recognition has been an active and vital topic among computer vision community for a long time. Previous researches mainly focus on loss functions used for facial feature extraction network, among which the improvements of softmax-based loss functions greatly promote the performance of face recognition. However, the contradiction between the drastically increasing number of face identities and the shortage of GPU memory is gradually becoming irreconcilable. In this work, we theoretically analyze the upper limit of model parallelism in face recognition in the first place. Then we propose a load-balanced sparse distributed classification training method, Partial FC, which is capable of using a machine with only 8 Nvidia Tesla V100 GPUs to implement training on a face recognition data set with up to 29 million IDs. Furthermore, we are able to train on data set with 100 million IDs in 64 RTX2080Ti GPUs. We have verified the effectiveness of Partial FC in 8 mainstream face recognition trainsets, and find that Partial FC is effective in all face recognition training sets. The code of this paper has been made available at https://github.com/deepinsight/insightface/tree/master/recognition/partial_fc."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Improved Residual Networks for Image and Video Recognition",
                    "abstract": "Residual networks (ResNets) represent a powerful type of convolutional neural network (CNN) architecture, widely adopted and used in various tasks. In this work we propose an improved version of ResNets. Our proposed improvements address all three main components of a ResNet: the flow of information through the network layers, the residual building block, and the projection shortcut. We are able to show consistent improvements in accuracy and learning convergence over the baseline. For instance, on ImageNet dataset, using the ResNet with 50 layers, for top-1 accuracy we can report a 1.19% improvement over the baseline in one setting and around 2% boost in another. Importantly, these improvements are obtained without increasing the model complexity. Our proposed approach allows us to train extremely deep networks, while the baseline shows severe optimization issues. We report results on three tasks over six datasets: image classification (ImageNet, CIFAR-10 and CIFAR-100), object detection (COCO) and video action recognition (Kinetics-400 and Something-Something-v2). In the deep learning era, we establish a new milestone for the depth of a CNN. We successfully train a 404-layer deep CNN on the ImageNet dataset and a 3002-layer network on CIFAR-10 and CIFAR-100, while the baseline is not able to converge at such extreme depths. Code and models are publicly available at: https://github.com/iduta/iresnet"
                },
                {
                    "title": "Scalable Gradients for Stochastic Differential Equations",
                    "abstract": "The adjoint sensitivity method scalably computes gradients of solutions to ordinary differential equations. We generalize this method to stochastic differential equations, allowing time-efficient and constant-memory computation of gradients with high-order adaptive solvers. Specifically, we derive a stochastic differential equation whose solution is the gradient, a memory-efficient algorithm for caching noise, and conditions under which numerical solutions converge. In addition, we combine our method with gradient-based stochastic variational inference for latent stochastic differential equations. We use our method to fit stochastic dynamics defined by neural networks, achieving competitive performance on a 50-dimensional motion capture dataset."
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                    "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "ArcFace: Additive Angular Margin Loss for Deep Face Recognition",
                    "abstract": "One of the main challenges in feature learning using Deep Convolutional Neural Networks (DCNNs) for large-scale face recognition is the design of appropriate loss functions that can enhance the discriminative power. Centre loss penalises the distance between deep features and their corresponding class centres in the Euclidean space to achieve intra-class compactness. SphereFace assumes that the linear transformation matrix in the last fully connected layer can be used as a representation of the class centres in the angular space and therefore penalises the angles between deep features and their corresponding weights in a multiplicative way. Recently, a popular line of research is to incorporate margins in well-established loss functions in order to maximise face class separability. In this paper, we propose an Additive Angular Margin Loss (ArcFace) to obtain highly discriminative features for face recognition. The proposed ArcFace has a clear geometric interpretation due to its exact correspondence to geodesic distance on a hypersphere. We present arguably the most extensive experimental evaluation against all recent state-of-the-art face recognition methods on ten face recognition benchmarks which includes a new large-scale image database with trillions of pairs and a large-scale video dataset. We show that ArcFace consistently outperforms the state of the art and can be easily implemented with negligible computational overhead. To facilitate future research, the code has been made available."
                },
                {
                    "title": "Biometric Systems under Morphing Attacks: Assessment of Morphing Techniques and Vulnerability Reporting",
                    "abstract": "With the widespread deployment of biometric recognition systems, the interest in attacking these systems is increasing. One of the easiest ways to circumvent a biometric recognition system are so-called presentation attacks, in which artefacts are presented to the sensor to either impersonate another subject or avoid being recognised. In the recent past, the vulnerabilities of biometric systems to so-called morphing attacks have been unveiled. In such attacks, biometric samples of multiple subjects are merged in the signal or feature domain, in order to allow a successful verification of all contributing subjects against the morphed identity. Being a recent area of research, there is to date no standardised manner to evaluate the vulnerability of biometric systems to these attacks. Hence, it is not yet possible to establish a common benchmark between different morph detection algorithms. In this paper, we tackle this issue proposing new metrics for vulnerability reporting, which build upon our joint experience in researching this challenging attack scenario. In addition, recommendations on the assessment of morphing techniques and morphing detection metrics are given."
                },
                {
                    "title": "Detecting morphed face images",
                    "abstract": "Widespread deployment of Automatic Border Control (ABC) along with the electronic Machine Readable Travel Documents (eMRTD) for person verification has enabled a prominent use case of face biometrics in border control applications. Many countries issue eMRTD passports on the basis of a printed biometric face photo submitted by the applicant. Some countries offer web-portals for passport renewal, where citizens can upload their face photo. These applications allow the possibility of the photo being altered to beautify the appearance of the data subject or being morphed to conceal the applicant identity. Specifically, if an eMRTD passport is issued with a morphed facial image, two or more data subjects, likely the known applicant and one or more unknown companion(s), can use such passport to pass a border control. In this work we propose a novel scheme to detect morphed face images based on facial micro-textures extracted using statistically independent filters that are trained on natural images. Given a face image, the proposed method will obtain a micro-texture variation using Binarized Statistical Image Features (BSIF), and the decision is made using a linear Support Vector Machine (SVM). This is first work carried out towards detecting the morphed face images. Extensive experiments are carried out on a large-scale database of 450 morphed face images created using 110 unique subjects with different ethnicity, age, and gender that indicates the superior performance."
                },
                {
                    "title": "Exponential integrators",
                    "abstract": "In this paper we consider the construction, analysis, implementation and application of exponential integrators. The focus will be on two types of stiff problems. The first one is characterized by a Jacobian that possesses eigenvalues with large negative real parts. Parabolic partial differential equations and their spatial discretization are typical examples. The second class consists of highly oscillatory problems with purely imaginary eigenvalues of large modulus. Apart from motivating the construction of exponential integrators for various classes of problems, our main intention in this article is to present the mathematics behind these methods. We will derive error bounds that are independent of stiffness or highest frequencies in the system. Since the implementation of exponential integrators requires the evaluation of the product of a matrix function with a vector, we will briefly discuss some possible approaches as well. The paper concludes with some applications, in which exponential integrators are used."
                },
                {
                    "title": "Stochastic flows and stochastic differential equations",
                    "abstract": "1. Stochastic processes and random fields 2. Continuous semimartingales and stochastic integrals 3. Semimartingales with spatial parameter and stochastic integrals 4. Stochastic flows 5. Convergence of stochastic flows 6. Stochastic partial differential equations."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we efficiently calculate gradients for diffusion models that utilize stochastic differential equation (SDE) solvers, enabling guided generation tasks?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of current backpropagation methods in diffusion models, particularly those using SDE solvers. By enabling efficient gradient calculation, this work could significantly enhance the adaptability of diffusion models for various applications, such as image and audio generation. This advancement could lead to more effective fine-tuning of pre-trained models, fostering innovation in generative tasks and potentially leading to breakthroughs in areas like personalized content creation and interactive AI systems.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of diffusion models, particularly in backpropagating through the iterative noise removal process. Naive approaches may fail due to their inflexibility and high memory requirements, as they do not account for the unique structure of diffusion models. Additionally, applying backpropagation to models using SDE solvers introduces further complications, as the mathematical treatment of stochastic processes is more intricate than that of deterministic ODEs. Overcoming these technical and theoretical obstacles requires innovative methods that can efficiently handle the gradient calculations without incurring excessive computational costs.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either ODE-based diffusion models or has not adequately addressed the gradient calculation for SDE-based models. Existing solutions often lack the necessary framework to handle the complexities of stochastic processes, leading to gaps in the literature. Barriers such as the absence of efficient adjoint methods for SDEs and the computational intensity of naive backpropagation approaches have hindered progress. Our approach differs by introducing AdjointDEIS, a novel technique that leverages adjoint sensitivity methods specifically tailored for diffusion models, thus providing a more efficient and effective solution.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of AdjointDEIS, which calculates gradients for diffusion models using the method of adjoint sensitivity. We will utilize a custom second-order multi-step method for solving the adjoint ODE, and extend this technique to diffusion SDEs. The evaluation will focus on the face morphing problem as a case study for guided generation tasks. The expected outcomes"
            }
        },
        "author_data": {
            "7beaea96-f852-415b-967d-fe1593b807eb": {
                "pk": "7beaea96-f852-415b-967d-fe1593b807eb",
                "name": "Zander W. Blasingame",
                "collaborators": [
                    "Chen Liu",
                    "Richard E. Neddo"
                ],
                "domain": [
                    "Face Recognition",
                    "Morphing Attacks",
                    "Deep Learning",
                    "Image Processing"
                ],
                "publications": [
                    {
                        "title": "Fast-DiM: Towards Fast Diffusion Morphs",
                        "abstract": "Diffusion Morphs (DiM) are a recent state-of-the-art method for creating high quality face morphs; however, they require a high number of network function evaluations (NFE) to create the morphs. We propose a new DiM pipeline, Fast-DiM, which can create morphs of a similar quality but with fewer NFE. We investigate the ODE solvers used to solve the Probability Flow ODE and the impact they have on the the creation of face morphs. Additionally, we employ an alternative method for encoding images into the latent space of the Diffusion model by solving the Probability Flow ODE as time runs forwards. Our experiments show that we can reduce the NFE by upwards of 85% in the encoding process while experiencing only 1.6\\% reduction in Mated Morph Presentation Match Rate (MMPMR). Likewise, we showed we could cut NFE, in the sampling process, in half with only a maximal reduction of 0.23% in MMPMR."
                    },
                    {
                        "title": "Greedy-DiM: Greedy Algorithms for Unreasonably Effective Face Morphs",
                        "abstract": "Morphing attacks are an emerging threat to state-of-the-art Face Recognition (FR) systems, which aim to create a single image that contains the biometric information of multiple identities. Diffusion Morphs (DiM) are a recently proposed morphing attack that has achieved state-of-the-art performance for representation-based morphing attacks. However, none of the existing research on DiMs have leveraged the iterative nature of DiMs and left the DiM model as a black box, treating it no differently than one would a Generative Adversarial Network (GAN) or Varational AutoEncoder (VAE). We propose a greedy strategy on the iterative sampling process of DiM models which searches for an optimal step guided by an identity-based heuristic function. We compare our proposed algorithm against ten other state-of-the-art morphing algorithms using the open-source SYN-MAD 2022 competition dataset. We find that our proposed algorithm is unreasonably effective, fooling all of the tested FR systems with an MMPMR of 100%, outperforming all other morphing algorithms compared."
                    },
                    {
                        "title": "Leveraging Diffusion For Strong and High Quality Face Morphing Attacks",
                        "abstract": "Face morphing attacks seek to deceive a Face Recognition (FR) system by presenting a morphed image consisting of the biometric qualities from two different identities with the aim of triggering a false acceptance with one of the two identities, thereby presenting a significant threat to biometric systems. The success of a morphing attack is dependent on the ability of the morphed image to represent the biometric characteristics of both identities that were used to create the image. We present a novel morphing attack that uses a Diffusion-based architecture to improve the visual fidelity of the image and the ability of the morphing attack to represent characteristics from both identities. We demonstrate the effectiveness of the proposed attack by evaluating its visual fidelity via the Frechet Inception Distance (FID). Also, extensive experiments are conducted to measure the vulnerability of FR systems to the proposed attack. The ability of a morphing attack detector to detect the proposed attack is measured and compared against two state-of-the-art GAN-based morphing attacks along with two Landmark-based attacks. Additionally, a novel metric to measure the relative strength between different morphing attacks is introduced and evaluated."
                    },
                    {
                        "title": "The Impact of Print-Scanning in Heterogeneous Morph Evaluation Scenarios",
                        "abstract": "Face morphing attacks pose an increasing threat to face recognition (FR) systems. A morphed photo contains biometric information from two different subjects to take advantage of vulnerabilities in FRs. These systems are particularly susceptible to attacks when the morphs are subjected to print-scanning to mask the artifacts generated during the morphing process. We investigate the impact of print-scanning on morphing attack detection through a series of evaluations on heterogeneous morphing attack scenarios. Our experiments show that we can increase the Mated Morph Presentation Match Rate (MMPMR) by up to 8.48%. Furthermore, when a Single-image Morphing Attack Detection (S-MAD) algorithm is not trained to detect print-scanned morphs the Morphing Attack Classification Error Rate (MACER) can increase by up to 96.12%, indicating significant vulnerability."
                    }
                ]
            },
            "c4146607-a4dd-48c9-9d5e-e48cd3f42081": {
                "pk": "c4146607-a4dd-48c9-9d5e-e48cd3f42081",
                "name": "Chen Liu",
                "collaborators": [
                    "Ding Chen",
                    "Shibo He",
                    "Haoyu Liu",
                    "Jiming Chen"
                ],
                "domain": [
                    "Differential Privacy",
                    "Trajectory Prediction",
                    "Generative Models",
                    "Image Classification"
                ],
                "publications": [
                    {
                        "title": "Fourier Transform Approximation as an Auxiliary Task for Image Classification",
                        "abstract": "Image reconstruction is likely the most predominant auxiliary task for image classification, but we would like to think twice about this convention. In this paper, we investigated \"approximating the Fourier Transform of the input image\" as a potential alternative, in the hope that it may further boost the performances on the primary task or introduce novel constraints not well covered by image reconstruction. We experimented with five popular classification architectures on the CIFAR-10 dataset, and the empirical results indicated that our proposed auxiliary task generally improves the classification accuracy. More notably, the results showed that in certain cases our proposed auxiliary task may enhance the classifiers' resistance to adversarial attacks generated using the fast gradient sign method."
                    },
                    {
                        "title": "Differentially Private Neural Network Training under Hidden State Assumption",
                        "abstract": "We present a novel approach called differentially private stochastic block coordinate descent (DP-SBCD) for training neural networks with provable guarantees of differential privacy under the hidden state assumption. Our methodology incorporates Lipschitz neural networks and decomposes the training process of the neural network into sub-problems, each corresponding to the training of a specific layer. By doing so, we extend the analysis of differential privacy under the hidden state assumption to encompass non-convex problems and algorithms employing proximal gradient descent. Furthermore, in contrast to existing methods, we adopt a novel approach by utilizing calibrated noise sampled from adaptive distributions, yielding improved empirical trade-offs between utility and privacy."
                    },
                    {
                        "title": "Intention-aware Denoising Diffusion Model for Trajectory Prediction",
                        "abstract": "Trajectory prediction is an essential component in autonomous driving, particularly for collision avoidance systems. Considering the inherent uncertainty of the task, numerous studies have utilized generative models to produce multiple plausible future trajectories for each agent. However, most of them suffer from restricted representation ability or unstable training issues. To overcome these limitations, we propose utilizing the diffusion model to generate the distribution of future trajectories. Two cruxes are to be settled to realize such an idea. First, the diversity of intention is intertwined with the uncertain surroundings, making the true distribution hard to parameterize. Second, the diffusion process is time-consuming during the inference phase, rendering it unrealistic to implement in a real-time driving system. We propose an Intention-aware denoising Diffusion Model (IDM), which tackles the above two problems. We decouple the original uncertainty into intention uncertainty and action uncertainty and model them with two dependent diffusion processes. To decrease the inference time, we reduce the variable dimensions in the intention-aware diffusion process and restrict the initial distribution of the action-aware diffusion process, which leads to fewer diffusion steps. To validate our approach, we conduct experiments on the Stanford Drone Dataset (SDD) and ETH/UCY dataset. Our methods achieve state-of-the-art results, with an FDE of 13.83 pixels on the SDD dataset and 0.36 meters on the ETH/UCY dataset. Compared with the original diffusion model, IDM reduces inference time by two-thirds. Interestingly, our experiments further reveal that introducing intention information is beneficial in modeling the diffusion process of fewer steps."
                    }
                ]
            }
        }
    },
    "2405.13985": {
        "paper_data": {
            "title": "LookHere: Vision Transformers with Directed Attention Generalize and Extrapolate",
            "url": "http://arxiv.org/abs/2405.13985v1",
            "arxiv_id": "2405.13985",
            "authors": [
                "Anthony Fuller",
                "Daniel G. Kyrollos",
                "Yousef Yassin",
                "James R. Green"
            ],
            "abstract": "High-resolution images offer more information about scenes that can improve model accuracy. However, the dominant model architecture in computer vision, the vision transformer (ViT), cannot effectively leverage larger images without finetuning -- ViTs poorly extrapolate to more patches at test time, although transformers offer sequence length flexibility. We attribute this shortcoming to the current patch position encoding methods, which create a distribution shift when extrapolating.   We propose a drop-in replacement for the position encoding of plain ViTs that restricts attention heads to fixed fields of view, pointed in different directions, using 2D attention masks. Our novel method, called LookHere, provides translation-equivariance, ensures attention head diversity, and limits the distribution shift that attention heads face when extrapolating. We demonstrate that LookHere improves performance on classification (avg. 1.6%), against adversarial attack (avg. 5.4%), and decreases calibration error (avg. 1.5%) -- on ImageNet without extrapolation. With extrapolation, LookHere outperforms the current SoTA position encoding method, 2D-RoPE, by 21.7% on ImageNet when trained at $224^2$ px and tested at $1024^2$ px. Additionally, we release a high-resolution test set to improve the evaluation of high-resolution image classifiers, called ImageNet-HR.",
            "introduction": " Introduction There is a decades-long trend in computer vision towards higher-resolution imagery, which contains more detailed scene information. Increasing resolution is a reliable way to improve model accuracy [13,14,15,16,17,18,19,20], but this comes at a cost; training models for hundreds of epochs on large-scale datasets is expensive, especially at high-resolutions. There are two ways to reduce this cost and still see accuracy benefits from high-resolutions: 1high-resolution finetuning, which pretrains models at a lower resolution, like 2242px, then finetunes them at a higher resolution, like 3842px; and 2extrapolating, which deploys models at a higher resolution, without further training. Of these two options, we should aim for models that can effectively extrapolate , as it presents a zero-cost solution that does not require finetuning at every target resolution. Finetuning costs aside, improvements to extrapolation should benefit high-resolution finetuning since models that are better at extrapolating can adapt to higher resolutions more easily. Although extrapolation is a significant and exciting challenge, state-of-the-art (SoTA) model architectures extrapolate poorly. Vision transformers (ViTs [ 9]) offer SoTA performance on many computer vision tasks. ViTs are simple; they split images into non-overlapping patches, linearly project pixels to form patch embeddings, and process these \u201ctokens\u201d with a stack of architecturally identical transformer layers \u2014 maintaining a constant feature map size throughout. This non-hierarchical design enables learning patch representations, which are useful for dense prediction tasks [ 21,22,23] and are fundamental \u2217AF, DGK, and YY made significant technical contributions. AF and DGK initiated the project. AF and JRG led the project. Code and data are available at: https://github.com/GreenCUBIC/lookhere Preprint. Under review.arXiv:2405.13985v1  [cs.CV]  22 May 2024224 320 384 448 512 768 10246065707580Top-1Accuracy [%]ImageNet-Val [1] 224 320 384 448 512 768 10245055606570ImageNet-v 2[2] 224 320 384 448 512 768 10245101520ImageNet-A [3] 224 320 384 448 512 768 102470758085 Test ResolutionTop-1Accuracy [%]ImageNet-ReaL [4] 224 320 384 448 512 768 10247075808590 Test ResolutionImageNet-HR (ours) 224 320 384 448 512 768 102451015202530 Test ResolutionImageNet-R [5] LH-180 (ours) LH-90 (ours) LH-45 (ours) Factorized [6] 2D-RoPE [7] 2D-sincos [8] 1D-learn [9] RPE-learn [10] Fourier [11] 2D-ALiBi [12] Figure 1: ViT-B/ 16models trained for 150epochs on ImageNet at 2242px and tested up to 10242px. Model architectures are consistent between runs other than position encoding results for ViT-B models trained on ImageNet for 150epochs; trained at 2242 and tested at various resolutions. Val [1] ReaL [4] v 2[2] -A [3] -R [5] -HR (ours) Method Res top- 1top-5top-1top-5top-1top-5top-1top-5top-1top-5top-1top-5 1D-learn 320279.89 94.59 85.17 96.26 68.78 87.88 12.91 34.36 26.73 40.04 88.76 96.94 2D-sincos 320279.48 94.50 84.76 96.34 68.23 87.69 12.52 33.59 26.30 39.84 87.94 96.76 Factorized 320279.73 94.70 85.08 96.37 68.71 87.83 11.44 32.01 25.89 39.33 87.78 97.08 Fourier 320279.80 94.58 85.17 96.45 68.43 87.95 12.72 34.16 26.33 39.64 88.40 97.08 RPE-learn 320279.94 94.73 85.50 96.63 68.56 87.83 11.17 30.51 24.21 36.71 88.12 97.26 2D-ALiBi 320280.44 95.10 85.66 96.66 69.25 88.65 13.33 35.00 26.54 40.47 88.62 97.36 2D-RoPE 320281.40 95.36 86.43 96.77 70.38 88.94 16.08 39.52 29.76 44.21 89.38 97.38 LH-180 320282.65 95.88 87.63 97.28 72.03 89.78 18.04 40.28 30.26 43.01 90.60 97.98 LH-90 320282.22 95.66 87.44 97.13 72.15 89.62 17.92 40.28 30.33 43.34 90.36 97.94 LH-45 320282.45 95.84 87.45 97.17 71.99 89.92 17.99 40.96 30.53 43.31 90.30 97.66 1D-learn 384279.02 94.14 84.42 95.99 67.47 87.13 11.97 32.23 23.66 36.93 87.36 96.74 2D-sincos 384278.56 94.06 83.95 96.02 66.96 87.16 11.87 31.31 23.53 36.63 87.02 96.38 Factorized 384278.56",
            "references": [
                {
                    "title": "Rotary Position Embedding for Vision Transformer",
                    "abstract": "Rotary Position Embedding (RoPE) performs remarkably on language models, especially for length extrapolation of Transformers. However, the impacts of RoPE on computer vision domains have been underexplored, even though RoPE appears capable of enhancing Vision Transformer (ViT) performance in a way similar to the language domain. This study provides a comprehensive analysis of RoPE when applied to ViTs, utilizing practical implementations of RoPE for 2D vision data. The analysis reveals that RoPE demonstrates impressive extrapolation performance, i.e., maintaining precision while increasing image resolution at inference. It eventually leads to performance improvement for ImageNet-1k, COCO detection, and ADE-20k segmentation. We believe this study provides thorough guidelines to apply RoPE into ViT, promising improved backbone performance with minimal extra computational overhead. Our code and pre-trained models are available at https://github.com/naver-ai/rope-vit"
                },
                {
                    "title": "MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training",
                    "abstract": "In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting."
                },
                {
                    "title": "FViT: A Focal Vision Transformer with Gabor Filter",
                    "abstract": "Vision transformers have achieved encouraging progress in various computer vision tasks. A common belief is that this is attributed to the competence of self-attention in modeling the global dependencies among feature tokens. Unfortunately, self-attention still faces some challenges in dense prediction tasks, such as the high computational complexity and absence of desirable inductive bias. To address these issues, we revisit the potential benefits of integrating vision transformer with Gabor filter, and propose a Learnable Gabor Filter (LGF) by using convolution. As an alternative to self-attention, we employ LGF to simulate the response of simple cells in the biological visual system to input images, prompting models to focus on discriminative feature representations of targets from various scales and orientations. Additionally, we design a Bionic Focal Vision (BFV) block based on the LGF. This block draws inspiration from neuroscience and introduces a Multi-Path Feed Forward Network (MPFFN) to emulate the working way of biological visual cortex processing information in parallel. Furthermore, we develop a unified and efficient pyramid backbone network family called Focal Vision Transformers (FViTs) by stacking BFV blocks. Experimental results show that FViTs exhibit highly competitive performance in various vision tasks. Especially in terms of computational efficiency and scalability, FViTs show significant advantages compared with other counterparts. Code is available at https://github.com/nkusyl/FViT"
                },
                {
                    "title": "Mixtral of Experts",
                    "abstract": "We introduce Mixtral 8x7B, a Sparse Mixture of Experts (SMoE) language model. Mixtral has the same architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks (i.e. experts). For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, each token has access to 47B parameters, but only uses 13B active parameters during inference. Mixtral was trained with a context size of 32k tokens and it outperforms or matches Llama 2 70B and GPT-3.5 across all evaluated benchmarks. In particular, Mixtral vastly outperforms Llama 2 70B on mathematics, code generation, and multilingual benchmarks. We also provide a model fine-tuned to follow instructions, Mixtral 8x7B - Instruct, that surpasses GPT-3.5 Turbo, Claude-2.1, Gemini Pro, and Llama 2 70B - chat model on human benchmarks. Both the base and instruct models are released under the Apache 2.0 license."
                },
                {
                    "title": "V*: Guided Visual Search as a Core Mechanism in Multimodal LLMs",
                    "abstract": "When we look around and perform complex tasks, how we see and selectively process what we see is crucial. How-ever, the lack of this visual search mechanism in current multimodal LLMs (MLLMs) hinders their ability to focus on important visual details, especially when handling high-resolution and visually crowded images. To address this, we introduce V*, an LLM- guided visual search mechanism that employs the world knowledge in LLMs for efficient visual querying. When combined with an MLLM, this mechanism enhances collaborative reasoning, contextual understanding, and precise visual grounding. This integration results in a new MLLM meta-architecture, named Show, sEArch, and TelL (SEAL). We further create V* Bench, a benchmark specifically designed to evaluate MLLMs in their ability to process high-resolution images and focus on visual details. Our study highlights the necessity of incorporating visual search capabilities into multimodal systems. The code is available here."
                },
                {
                    "title": "Understanding the Detrimental Class-level Effects of Data Augmentation",
                    "abstract": "Data augmentation (DA) encodes invariance and provides implicit regularization critical to a model's performance in image classification tasks. However, while DA improves average accuracy, recent studies have shown that its impact can be highly class dependent: achieving optimal average accuracy comes at the cost of significantly hurting individual class accuracy by as much as 20% on ImageNet. There has been little progress in resolving class-level accuracy drops due to a limited understanding of these effects. In this work, we present a framework for understanding how DA interacts with class-level learning dynamics. Using higher-quality multi-label annotations on ImageNet, we systematically categorize the affected classes and find that the majority are inherently ambiguous, co-occur, or involve fine-grained distinctions, while DA controls the model's bias towards one of the closely related classes. While many of the previously reported performance drops are explained by multi-label annotations, our analysis of class confusions reveals other sources of accuracy degradation. We show that simple class-conditional augmentation strategies informed by our framework improve performance on the negatively affected classes."
                },
                {
                    "title": "Monkey: Image Resolution and Text Label are Important Things for Large Multi-Modal Models",
                    "abstract": "Large Multimodal Models (LMMs) have shown promise in vision-language tasks but struggle with high-resolution input and detailed scene understanding. Addressing these challenges, we introduce Monkey to enhance LMM capabilities. Firstly, Monkey processes input images by dividing them into uniform patches, each matching the size (e.g., 448\u00d7448) used in the original training of the well-trained vision encoder. Equipped with individual adapter for each patch, Monkey can handle higher resolutions up to 1344\u00d7896 pixels, enabling the detailed capture of complex visual information. Secondly, it employs a multi-level description generation method, enriching the context for scene-object associations. This two-part strategy ensures more effective learning from generated data: the higher resolution allows for a more detailed capture of visuals, which in turn enhances the effectiveness of comprehensive descriptions. Extensive ablative results validate the effectiveness of our designs. Additionally, experiments on 18 datasets further demonstrate that Monkey surpasses existing LMMs in many tasks like Image Captioning and various Visual Question Answering formats. Specially, in qualitative tests focused on dense text question answering, Monkey has exhibited encouraging results compared with GPT4V. Code is available at https://github.com/Yuliang-Liu/Monkey."
                },
                {
                    "title": "CROMA: Remote Sensing Representations with Contrastive Radar-Optical Masked Autoencoders",
                    "abstract": "A vital and rapidly growing application, remote sensing offers vast yet sparsely labeled, spatially aligned multimodal data; this makes self-supervised learning algorithms invaluable. We present CROMA: a framework that combines contrastive and reconstruction self-supervised objectives to learn rich unimodal and multimodal representations. Our method separately encodes masked-out multispectral optical and synthetic aperture radar samples -- aligned in space and time -- and performs cross-modal contrastive learning. Another encoder fuses these sensors, producing joint multimodal encodings that are used to predict the masked patches via a lightweight decoder. We show that these objectives are complementary when leveraged on spatially aligned multimodal data. We also introduce X- and 2D-ALiBi, which spatially biases our cross- and self-attention matrices. These strategies improve representations and allow our models to effectively extrapolate to images up to 17.6x larger at test-time. CROMA outperforms the current SoTA multispectral model, evaluated on: four classification benchmarks -- finetuning (avg. 1.8%), linear (avg. 2.4%) and nonlinear (avg. 1.4%) probing, kNN classification (avg. 3.5%), and K-means clustering (avg. 8.4%); and three segmentation benchmarks (avg. 6.4%). CROMA's rich, optionally multimodal representations can be widely leveraged across remote sensing applications."
                },
                {
                    "title": "Analyzing Vision Transformers for Image Classification in Class Embedding Space",
                    "abstract": "Despite the growing use of transformer models in computer vision, a mechanistic understanding of these networks is still needed. This work introduces a method to reverse-engineer Vision Transformers trained to solve image classification tasks. Inspired by previous research in NLP, we demonstrate how the inner representations at any level of the hierarchy can be projected onto the learned class embedding space to uncover how these networks build categorical representations for their predictions. We use our framework to show how image tokens develop class-specific representations that depend on attention mechanisms and contextual information, and give insights on how self-attention and MLP layers differentially contribute to this categorical composition. We additionally demonstrate that this method (1) can be used to determine the parts of an image that would be important for detecting the class of interest, and (2) exhibits significant advantages over traditional linear probing approaches. Taken together, our results position our proposed framework as a powerful tool for mechanistic interpretability and explainability research."
                },
                {
                    "title": "Scaling Laws of RoPE-based Extrapolation",
                    "abstract": "The extrapolation capability of Large Language Models (LLMs) based on Rotary Position Embedding is currently a topic of considerable interest. The mainstream approach to addressing extrapolation with LLMs involves modifying RoPE by replacing 10000, the rotary base of $\\theta_n={10000}^{-2n/d}$ in the original RoPE, with a larger value and providing longer fine-tuning text. In this work, we first observe that fine-tuning a RoPE-based LLM with either a smaller or larger base in pre-training context length could significantly enhance its extrapolation performance. After that, we propose \\textbf{\\textit{Scaling Laws of RoPE-based Extrapolation}}, a unified framework from the periodic perspective, to describe the relationship between the extrapolation performance and base value as well as tuning context length. In this process, we also explain the origin of the RoPE-based extrapolation issue by \\textbf{\\textit{critical dimension for extrapolation}}. Besides these observations and analyses, we achieve extrapolation up to 1 million context length within only 16K training length on LLaMA2 7B and 13B."
                },
                {
                    "title": "PoSE: Efficient Context Window Extension of LLMs via Positional Skip-wise Training",
                    "abstract": "Large Language Models (LLMs) are trained with a pre-defined context length, restricting their use in scenarios requiring long inputs. Previous efforts for adapting LLMs to a longer length usually requires fine-tuning with this target length (Full-length fine-tuning), suffering intensive training cost. To decouple train length from target length for efficient context window extension, we propose Positional Skip-wisE (PoSE) training that smartly simulates long inputs using a fixed context window. This is achieved by first dividing the original context window into several chunks, then designing distinct skipping bias terms to manipulate the position indices of each chunk. These bias terms and the lengths of each chunk are altered for every training example, allowing the model to adapt to all positions within target length. Experimental results show that PoSE greatly reduces memory and time overhead compared with Full-length fine-tuning, with minimal impact on performance. Leveraging this advantage, we have successfully extended the LLaMA model to 128k tokens using a 2k training context window. Furthermore, we empirically confirm that PoSE is compatible with all RoPE-based LLMs and position interpolation strategies. Notably, our method can potentially support infinite length, limited only by memory usage in inference. With ongoing progress for efficient inference, we believe PoSE can further scale the context window beyond 128k."
                },
                {
                    "title": "Patch n' Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution",
                    "abstract": "The ubiquitous and demonstrably suboptimal choice of resizing images to a fixed resolution before processing them with computer vision models has not yet been successfully challenged. However, models such as the Vision Transformer (ViT) offer flexible sequence-based modeling, and hence varying input sequence lengths. We take advantage of this with NaViT (Native Resolution ViT) which uses sequence packing during training to process inputs of arbitrary resolutions and aspect ratios. Alongside flexible model usage, we demonstrate improved training efficiency for large-scale supervised and contrastive image-text pretraining. NaViT can be efficiently transferred to standard tasks such as image and video classification, object detection, and semantic segmentation and leads to improved results on robustness and fairness benchmarks. At inference time, the input resolution flexibility can be used to smoothly navigate the test-time cost-performance trade-off. We believe that NaViT marks a departure from the standard, CNN-designed, input and modelling pipeline used by most computer vision models, and represents a promising direction for ViTs."
                },
                {
                    "title": "Lightweight Vision Transformer with Bidirectional Interaction",
                    "abstract": "Recent advancements in vision backbones have significantly improved their performance by simultaneously modeling images' local and global contexts. However, the bidirectional interaction between these two contexts has not been well explored and exploited, which is important in the human visual system. This paper proposes a Fully Adaptive Self-Attention (FASA) mechanism for vision transformer to model the local and global information as well as the bidirectional interaction between them in context-aware ways. Specifically, FASA employs self-modulated convolutions to adaptively extract local representation while utilizing self-attention in down-sampled space to extract global representation. Subsequently, it conducts a bidirectional adaptation process between local and global representation to model their interaction. In addition, we introduce a fine-grained downsampling strategy to enhance the down-sampled self-attention mechanism for finer-grained global perception capability. Based on FASA, we develop a family of lightweight vision backbones, Fully Adaptive Transformer (FAT) family. Extensive experiments on multiple vision tasks demonstrate that FAT achieves impressive performance. Notably, FAT accomplishes a 77.6% accuracy on ImageNet-1K using only 4.5M parameters and 0.7G FLOPs, which surpasses the most advanced ConvNets and Transformers with similar model size and computational costs. Moreover, our model exhibits faster speed on modern GPU compared to other models. Code will be available at https://github.com/qhfan/FAT."
                },
                {
                    "title": "The Impact of Positional Encoding on Length Generalization in Transformers",
                    "abstract": "Length generalization, the ability to generalize from small training context sizes to larger ones, is a critical challenge in the development of Transformer-based language models. Positional encoding (PE) has been identified as a major factor influencing length generalization, but the exact impact of different PE schemes on extrapolation in downstream tasks remains unclear. In this paper, we conduct a systematic empirical study comparing the length generalization performance of decoder-only Transformers with five different position encoding approaches including Absolute Position Embedding (APE), T5's Relative PE, ALiBi, and Rotary, in addition to Transformers without positional encoding (NoPE). Our evaluation encompasses a battery of reasoning and mathematical tasks. Our findings reveal that the most commonly used positional encoding methods, such as ALiBi, Rotary, and APE, are not well suited for length generalization in downstream tasks. More importantly, NoPE outperforms other explicit positional encoding methods while requiring no additional computation. We theoretically demonstrate that NoPE can represent both absolute and relative PEs, but when trained with SGD, it mostly resembles T5's relative PE attention patterns. Finally, we find that scratchpad is not always helpful to solve length generalization and its format highly impacts the model's performance. Overall, our work suggests that explicit position embeddings are not essential for decoder-only Transformers to generalize well to longer sequences."
                },
                {
                    "title": "Getting ViT in Shape: Scaling Laws for Compute-Optimal Model Design",
                    "abstract": "Scaling laws have been recently employed to derive compute-optimal model size (number of parameters) for a given compute duration. We advance and refine such methods to infer compute-optimal model shapes, such as width and depth, and successfully implement this in vision transformers. Our shape-optimized vision transformer, SoViT, achieves results competitive with models that exceed twice its size, despite being pre-trained with an equivalent amount of compute. For example, SoViT-400m/14 achieves 90.3% fine-tuning accuracy on ILSRCV2012, surpassing the much larger ViT-g/14 and approaching ViT-G/14 under identical settings, with also less than half the inference cost. We conduct a thorough evaluation across multiple tasks, such as image classification, captioning, VQA and zero-shot transfer, demonstrating the effectiveness of our model across a broad range of domains and identifying limitations. Overall, our findings challenge the prevailing approach of blindly scaling up vision models and pave a path for a more informed scaling."
                },
                {
                    "title": "InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning",
                    "abstract": "Large-scale pre-training and instruction tuning have been successful at creating general-purpose language models with broad competence. However, building general-purpose vision-language models is challenging due to the rich input distributions and task diversity resulting from the additional visual input. Although vision-language pretraining has been widely studied, vision-language instruction tuning remains under-explored. In this paper, we conduct a systematic and comprehensive study on vision-language instruction tuning based on the pretrained BLIP-2 models. We gather 26 publicly available datasets, covering a wide variety of tasks and capabilities, and transform them into instruction tuning format. Additionally, we introduce an instruction-aware Query Transformer, which extracts informative features tailored to the given instruction. Trained on 13 held-in datasets, InstructBLIP attains state-of-the-art zero-shot performance across all 13 held-out datasets, substantially outperforming BLIP-2 and larger Flamingo models. Our models also lead to state-of-the-art performance when finetuned on individual downstream tasks (e.g., 90.7% accuracy on ScienceQA questions with image contexts). Furthermore, we qualitatively demonstrate the advantages of InstructBLIP over concurrent multimodal models. All InstructBLIP models are open-sourced at https://github.com/salesforce/LAVIS/tree/main/projects/instructblip."
                },
                {
                    "title": "What Do Self-Supervised Vision Transformers Learn?",
                    "abstract": "We present a comparative study on how and why contrastive learning (CL) and masked image modeling (MIM) differ in their representations and in their performance of downstream tasks. In particular, we demonstrate that self-supervised Vision Transformers (ViTs) have the following properties: (1) CL trains self-attentions to capture longer-range global patterns than MIM, such as the shape of an object, especially in the later layers of the ViT architecture. This CL property helps ViTs linearly separate images in their representation spaces. However, it also makes the self-attentions collapse into homogeneity for all query tokens and heads. Such homogeneity of self-attention reduces the diversity of representations, worsening scalability and dense prediction performance. (2) CL utilizes the low-frequency signals of the representations, but MIM utilizes high-frequencies. Since low- and high-frequency information respectively represent shapes and textures, CL is more shape-oriented and MIM more texture-oriented. (3) CL plays a crucial role in the later layers, while MIM mainly focuses on the early layers. Upon these analyses, we find that CL and MIM can complement each other and observe that even the simplest harmonization can help leverage the advantages of both methods. The code is available at https://github.com/naver-ai/cl-vs-mim."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "DINOv2: Learning Robust Visual Features without Supervision",
                    "abstract": "The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels."
                },
                {
                    "title": "ImageNet-Hard: The Hardest Images Remaining from a Study of the Power of Zoom and Spatial Biases in Image Classification",
                    "abstract": "Image classifiers are information-discarding machines, by design. Yet, how these models discard information remains mysterious. We hypothesize that one way for image classifiers to reach high accuracy is to first zoom to the most discriminative region in the image and then extract features from there to predict image labels, discarding the rest of the image. Studying six popular networks ranging from AlexNet to CLIP, we find that proper framing of the input image can lead to the correct classification of 98.91% of ImageNet images. Furthermore, we uncover positional biases in various datasets, especially a strong center bias in two popular datasets: ImageNet-A and ObjectNet. Finally, leveraging our insights into the potential of zooming, we propose a test-time augmentation (TTA) technique that improves classification accuracy by forcing models to explicitly perform zoom-in operations before making predictions. Our method is more interpretable, accurate, and faster than MEMO, a state-of-the-art (SOTA) TTA method. We introduce ImageNet-Hard, a new benchmark that challenges SOTA classifiers including large vision-language models even when optimal zooming is allowed."
                },
                {
                    "title": "Stabilizing Transformer Training by Preventing Attention Entropy Collapse",
                    "abstract": "Training stability is of great importance to Transformers. In this work, we investigate the training dynamics of Transformers by examining the evolution of the attention layers. In particular, we track the attention entropy for each attention head during the course of training, which is a proxy for model sharpness. We identify a common pattern across different architectures and tasks, where low attention entropy is accompanied by high training instability, which can take the form of oscillating loss or divergence. We denote the pathologically low attention entropy, corresponding to highly concentrated attention scores, as $\\textit{entropy collapse}$. As a remedy, we propose $\\sigma$Reparam, a simple and efficient solution where we reparametrize all linear layers with spectral normalization and an additional learned scalar. We demonstrate that $\\sigma$Reparam successfully prevents entropy collapse in the attention layers, promoting more stable training. Additionally, we prove a tight lower bound of the attention entropy, which decreases exponentially fast with the spectral norm of the attention logits, providing additional motivation for our approach. We conduct experiments with $\\sigma$Reparam on image classification, image self-supervised learning, machine translation, speech recognition, and language modeling tasks. We show that $\\sigma$Reparam provides stability and robustness with respect to the choice of hyperparameters, going so far as enabling training (a) a Vision Transformer {to competitive performance} without warmup, weight decay, layer normalization or adaptive optimizers; (b) deep architectures in machine translation and (c) speech recognition to competitive performance without warmup and adaptive optimizers. Code is available at \\url{https://github.com/apple/ml-sigma-reparam}."
                },
                {
                    "title": "X-Pruner: eXplainable Pruning for Vision Transformers",
                    "abstract": "Recently vision transformer models have become prominent models for a range of tasks. These models, however, usually suffer from intensive computational costs and heavy memory requirements, making them impractical for deployment on edge platforms. Recent studies have proposed to prune transformers in an unexplainable manner, which overlook the relationship between internal units of the model and the target class, thereby leading to inferior performance. To alleviate this problem, we propose a novel explainable pruning framework dubbed X-Pruner, which is designed by considering the explainability of the pruning criterion. Specifically, to measure each prunable unit's contribution to predicting each target class, a novel explainability-aware mask is proposed and learned in an end-to-end manner. Then, to preserve the most informative units and learn the layer-wise pruning rate, we adaptively search the layer-wise threshold that differentiates between unpruned and pruned units based on their explainability-aware mask values. To verify and evaluate our method, we apply the X-Pruner on representative transformer models including the DeiT and Swin Transformer. Comprehensive simulation results demonstrate that the proposed X-Pruner outperforms the state-of-the-art black-box methods with significantly reduced computational costs and slight performance degradation. Code is available at https://github.com/vickyyu90/XPruner."
                },
                {
                    "title": "Language Is Not All You Need: Aligning Perception with Language Models",
                    "abstract": "A big convergence of language, multimodal perception, action, and world modeling is a key step toward artificial general intelligence. In this work, we introduce Kosmos-1, a Multimodal Large Language Model (MLLM) that can perceive general modalities, learn in context (i.e., few-shot), and follow instructions (i.e., zero-shot). Specifically, we train Kosmos-1 from scratch on web-scale multimodal corpora, including arbitrarily interleaved text and images, image-caption pairs, and text data. We evaluate various settings, including zero-shot, few-shot, and multimodal chain-of-thought prompting, on a wide range of tasks without any gradient updates or finetuning. Experimental results show that Kosmos-1 achieves impressive performance on (i) language understanding, generation, and even OCR-free NLP (directly fed with document images), (ii) perception-language tasks, including multimodal dialogue, image captioning, visual question answering, and (iii) vision tasks, such as image recognition with descriptions (specifying classification via text instructions). We also show that MLLMs can benefit from cross-modal transfer, i.e., transfer knowledge from language to multimodal, and from multimodal to language. In addition, we introduce a dataset of Raven IQ test, which diagnoses the nonverbal reasoning capability of MLLMs."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "FlexiViT: One Model for All Patch Sizes",
                    "abstract": "Vision Transformers convert images to sequences by slicing them into patches. The size of these patches controls a speed/accuracy tradeoff, with smaller patches leading to higher accuracy at greater computational cost, but changing the patch size typically requires retraining the model. In this paper, we demonstrate that simply randomizing the patch size at training time leads to a single set of weights that performs well across a wide range of patch sizes, making it possible to tailor the model to different compute budgets at deployment time. We extensively evaluate the resulting model, which we call FlexiViT, on a wide range of tasks, including classification, image-text retrieval, open-world detection, panoptic segmentation, and semantic segmentation, concluding that it usually matches, and sometimes outperforms, standard ViT models trained at a single patch size in an otherwise identical setup. Hence, FlexiViT training is a simple drop-in improvement for ViT that makes it easy to add compute-adaptive capabilities to most models relying on a ViT backbone architecture. Code and pre-trained models are available at github.com/google-research/big_vision."
                },
                {
                    "title": "CroCo v2: Improved Cross-view Completion Pre-training for Stereo Matching and Optical Flow",
                    "abstract": "Despite impressive performance for high-level downstream tasks, self-supervised pre-training methods have not yet fully delivered on dense geometric vision tasks such as stereo matching or optical flow. The application of self-supervised concepts, such as instance discrimination or masked image modeling, to geometric tasks is an active area of research. In this work, we build on the recent cross-view completion framework, a variation of masked image modeling that leverages a second view from the same scene which makes it well suited for binocular downstream tasks. The applicability of this concept has so far been limited in at least two ways: (a) by the difficulty of collecting real-world image pairs \u2013 in practice only synthetic data have been used \u2013 and (b) by the lack of generalization of vanilla transformers to dense downstream tasks for which relative position is more meaningful than absolute position. We explore three avenues of improvement. First, we introduce a method to collect suitable real-world image pairs at large scale. Second, we experiment with relative positional embeddings and show that they enable vision transformers to perform substantially better. Third, we scale up vision transformer based cross-completion architectures, which is made possible by the use of large amounts of data. With these improvements, we show for the first time that state-of-the-art results on stereo matching and optical flow can be reached without using any classical task-specific techniques like correlation volume, iterative estimation, image warping or multi-scale reasoning, thus paving the way towards universal vision models."
                },
                {
                    "title": "EVA: Exploring the Limits of Masked Visual Representation Learning at Scale",
                    "abstract": "We launch EVA, a vision-centric foundation model to Explore the limits of Visual representation at scAle using only publicly accessible data. EVA is a vanilla ViT pre-trained to reconstruct the masked out image-text aligned vision features conditioned on visible image patches. Via this pretext task, we can efficiently scale up EVA to one billion parameters, and sets new records on a broad range of representative vision downstream tasks, such as image recognition, video action recognition, object detection, instance segmentation and semantic segmentation without heavy supervised training. Moreover, we observe quantitative changes in scaling EVA result in qualitative changes in transfer learning performance that are not present in other models. For instance, EVA takes a great leap in the challenging large vocabulary instance segmentation task: our model achieves almost the same state-of-the-art performance on LVIS dataset with over a thousand categories and COCO dataset with only eighty categories. Beyond a pure vision encoder, EVA can also serve as a vision-centric, multi-modal pivot to connect images and text. We find initializing the vision tower of a giant CLIP from EVA can greatly stabilize the training and outperform the training from scratch counterpart with much fewer samples and less compute, providing a new direction for scaling up and accelerating the costly training of multi-modal foundation models."
                },
                {
                    "title": "MetaFormer Baselines for Vision",
                    "abstract": "MetaFormer, the abstracted architecture of Transformer, has been found to play a significant role in achieving competitive performance. In this paper, we further explore the capacity of MetaFormer, again, by migrating our focus away from the token mixer design: we introduce several baseline models under MetaFormer using the most basic or common mixers, and demonstrate their gratifying performance. We summarize our observations as follows: <list list-type=\"ordered\"> <list-item><label>1)</label><p><italic>MetaFormer ensures solid lower bound of performance:</italic> By merely adopting identity mapping as the token mixer, the MetaFormer model, termed <italic>IdentityFormer</italic>, achieves <inline-formula><tex-math notation=\"LaTeX\">$> $</tex-math><alternatives><mml:math><mml:mo>></mml:mo></mml:math><inline-graphic xlink:href=\"wang-ieq1-3329173.gif\"/></alternatives></inline-formula>80% accuracy on ImageNet-1 K.</p></list-item> <list-item><label>2)</label><p><italic>MetaFormer works well with arbitrary token mixers:</italic> When specifying the token mixer as even a random matrix to mix tokens, the resulting model <italic>RandFormer</italic> yields an accuracy of <inline-formula><tex-math notation=\"LaTeX\">$> $</tex-math><alternatives><mml:math><mml:mo>></mml:mo></mml:math><inline-graphic xlink:href=\"wang-ieq2-3329173.gif\"/></alternatives></inline-formula>81%, outperforming IdentityFormer. Rest assured of MetaFormer's results when new token mixers are adopted.</p></list-item> <list-item><label>3)</label><p><italic>MetaFormer effortlessly offers state-of-the-art results:</italic> With just conventional token mixers dated back five years ago, the models instantiated from MetaFormer already beat state of the art.</p></list-item> </list> <list list-type=\"ordered\"> <list-item><label>a)</label><p><italic>ConvFormer outperforms ConvNeXt:</italic> Taking the common depthwise separable convolutions as the token mixer, the model termed <italic>ConvFormer</italic>, which can be regarded as pure CNNs, outperforms the strong CNN model ConvNeXt.</p></list-item> <list-item><label>b)</label><p><italic>CAFormer sets new record on ImageNet-1 K:</italic> By simply applying depthwise separable convolutions as token mixer in the bottom stages and vanilla self-attention in the top stages, the resulting model <italic>CAFormer</italic> sets a new record on ImageNet-1 K: it achieves an accuracy of 85.5% at <inline-formula><tex-math notation=\"LaTeX\">$224 \\times 224$</tex-math><alternatives><mml:math><mml:mrow><mml:mn>224</mml:mn><mml:mo>\u00d7</mml:mo><mml:mn>224</mml:mn></mml:mrow></mml:math><inline-graphic xlink:href=\"wang-ieq3-3329173.gif\"/></alternatives></inline-formula> resolution, under normal supervised training without external data or distillation.</p></list-item> </list> In our expedition to probe MetaFormer, we also find that a new activation, <italic>StarReLU</italic>, reduces 71% FLOPs of activation compared with commonly-used GELU yet achieves better performance. Specifically, StarReLU is a variant of Squared ReLU dedicated to alleviating distribution shift. We expect StarReLU to find great potential in MetaFormer-like models alongside other neural networks."
                },
                {
                    "title": "SegViT: Semantic Segmentation with Plain Vision Transformers",
                    "abstract": "We explore the capability of plain Vision Transformers (ViTs) for semantic segmentation and propose the SegVit. Previous ViT-based segmentation networks usually learn a pixel-level representation from the output of the ViT. Differently, we make use of the fundamental component -- attention mechanism, to generate masks for semantic segmentation. Specifically, we propose the Attention-to-Mask (ATM) module, in which the similarity maps between a set of learnable class tokens and the spatial feature maps are transferred to the segmentation masks. Experiments show that our proposed SegVit using the ATM module outperforms its counterparts using the plain ViT backbone on the ADE20K dataset and achieves new state-of-the-art performance on COCO-Stuff-10K and PASCAL-Context datasets. Furthermore, to reduce the computational cost of the ViT backbone, we propose query-based down-sampling (QD) and query-based up-sampling (QU) to build a Shrunk structure. With the proposed Shrunk structure, the model can save up to $40\\%$ computations while maintaining competitive performance."
                },
                {
                    "title": "Bridging the Gap Between Vision Transformers and Convolutional Neural Networks on Small Datasets",
                    "abstract": "There still remains an extreme performance gap between Vision Transformers (ViTs) and Convolutional Neural Networks (CNNs) when training from scratch on small datasets, which is concluded to the lack of inductive bias. In this paper, we further consider this problem and point out two weaknesses of ViTs in inductive biases, that is, the spatial relevance and diverse channel representation. First, on spatial aspect, objects are locally compact and relevant, thus fine-grained feature needs to be extracted from a token and its neighbors. While the lack of data hinders ViTs to attend the spatial relevance. Second, on channel aspect, representation exhibits diversity on different channels. But the scarce data can not enable ViTs to learn strong enough representation for accurate recognition. To this end, we propose Dynamic Hybrid Vision Transformer (DHVT) as the solution to enhance the two inductive biases. On spatial aspect, we adopt a hybrid structure, in which convolution is integrated into patch embedding and multi-layer perceptron module, forcing the model to capture the token features as well as their neighboring features. On channel aspect, we introduce a dynamic feature aggregation module in MLP and a brand new\"head token\"design in multi-head self-attention module to help re-calibrate channel representation and make different channel group representation interacts with each other. The fusion of weak channel representation forms a strong enough representation for classification. With this design, we successfully eliminate the performance gap between CNNs and ViTs, and our DHVT achieves a series of state-of-the-art performance with a lightweight model, 85.68% on CIFAR-100 with 22.8M parameters, 82.3% on ImageNet-1K with 24.0M parameters. Code is available at https://github.com/ArieSeirack/DHVT."
                },
                {
                    "title": "Bugs in the Data: How ImageNet Misrepresents Biodiversity",
                    "abstract": "ImageNet-1k is a dataset often used for benchmarking machine learning (ML) models and evaluating tasks such as image recognition and object detection. Wild animals make up 27% of ImageNet-1k but, unlike classes representing people and objects, these data have not been closely scrutinized. In the current paper, we analyze the 13,450 images from 269 classes that represent wild animals in the ImageNet-1k validation set, with the participation of expert ecologists. We find that many of the classes are ill-defined or overlapping, and that 12% of the images are incorrectly labeled, with some classes having >90% of images incorrect. We also find that both the wildlife-related labels and images included in ImageNet-1k present significant geographical and cultural biases, as well as ambiguities such as artificial animals, multiple species in the same image, or the presence of humans. Our findings highlight serious issues with the extensive use of this dataset for evaluating ML systems, the use of such algorithms in wildlife-related tasks, and more broadly the ways in which ML datasets are commonly created and curated."
                },
                {
                    "title": "BEiT v2: Masked Image Modeling with Vector-Quantized Visual Tokenizers",
                    "abstract": "Masked image modeling (MIM) has demonstrated impressive results in selfsupervised representation learning by recovering corrupted image patches. However, most existing studies operate on low-level image pixels, which hinders the exploitation of high-level semantics for representation models. In this work, we propose to use a semantic-rich visual tokenizer as the reconstruction target for masked prediction, providing a systematic way to promote MIM from pixel-level to semantic-level. Specifically, we propose vector-quantized knowledge distillation to train the tokenizer, which discretizes a continuous semantic space to compact codes. We then pretrain vision Transformers by predicting the original visual tokens for the masked image patches. Furthermore, we introduce a patch aggregation strategy which associates discrete image patches to enhance global semantic representation. Experiments on image classification and semantic segmentation show that BEIT V2 outperforms all compared MIM methods. On ImageNet-1K (224 size), the base-size BEIT V2 achieves 85.5% top-1 accuracy for fine-tuning and 80.1% top-1 accuracy for linear probing. The large-size BEIT V2 obtains 87.3% top-1 accuracy for ImageNet-1K (224 size) fine-tuning, and 56.7% mIoU on ADE20K for semantic segmentation. The code and pretrained models are available at https://aka.ms/beitv2."
                },
                {
                    "title": "Width & Depth Pruning for Vision Transformers",
                    "abstract": "Transformer models have demonstrated their promising potential and achieved excellent performance on a series of computer vision tasks. However, the huge computational cost of vision transformers hinders their deployment and application to edge devices. Recent works have proposed to \ufb01nd and remove the unimportant units of vision transformers. Despite achieving remarkable results, these methods take one dimension of network width into consideration and ignore network depth, which is another important dimension for pruning vision transformers. Therefore, we propose a Width & Depth Pruning (WDPruning) framework that reduces both width and depth dimensions simultaneously. Speci\ufb01cally, for width pruning, a set of learnable pruning-related parameters is used to adaptively adjust the width of transformer. For depth pruning, we introduce several shallow classi\ufb01ers by using the intermediate information of the transformer blocks, which allows images to be classi\ufb01ed by shallow classi\ufb01ers instead of the deeper classi\ufb01ers. In the inference period, all of the blocks after shallow classi\ufb01ers can be dropped so they don\u2019t bring additional parameters and computation. Experimental results on benchmark datasets demonstrate that the proposed method can signi\ufb01cantly reduce the computational costs of mainstream vision transformers such as DeiT and Swin Transformer with a minor accuracy drop. In particular, on ILSVRC-12, we achieve over 22% pruning ratio of FLOPs by compressing DeiT-Base, even with an increase of 0.14% Top-1 accuracy."
                },
                {
                    "title": "Peripheral Vision Transformer",
                    "abstract": "Human vision possesses a special type of visual processing systems called peripheral vision. Partitioning the entire visual field into multiple contour regions based on the distance to the center of our gaze, the peripheral vision provides us the ability to perceive various visual features at different regions. In this work, we take a biologically inspired approach and explore to model peripheral vision in deep neural networks for visual recognition. We propose to incorporate peripheral position encoding to the multi-head self-attention layers to let the network learn to partition the visual field into diverse peripheral regions given training data. We evaluate the proposed network, dubbed PerViT, on ImageNet-1K and systematically investigate the inner workings of the model for machine perception, showing that the network learns to perceive visual data similarly to the way that human vision does. The performance improvements in image classification over the baselines across different model sizes demonstrate the efficacy of the proposed method."
                },
                {
                    "title": "Better plain ViT baselines for ImageNet-1k",
                    "abstract": "It is commonly accepted that the Vision Transformer model requires sophisticated regularization techniques to excel at ImageNet-1k scale data. Surprisingly, we find this is not the case and standard data augmentation is sufficient. This note presents a few minor modifications to the original Vision Transformer (ViT) vanilla training setting that dramatically improve the performance of plain ViT models. Notably, 90 epochs of training surpass 76% top-1 accuracy in under seven hours on a TPUv3-8, similar to the classic ResNet50 baseline, and 300 epochs of training reach 80% in less than one day."
                },
                {
                    "title": "DeiT III: Revenge of the ViT",
                    "abstract": "A Vision Transformer (ViT) is a simple neural architecture amenable to serve several computer vision tasks. It has limited built-in architectural priors, in contrast to more recent architectures that incorporate priors either about the input data or of specific tasks. Recent works show that ViTs benefit from self-supervised pre-training, in particular BerT-like pre-training like BeiT. In this paper, we revisit the supervised training of ViTs. Our procedure builds upon and simplifies a recipe introduced for training ResNet-50. It includes a new simple data-augmentation procedure with only 3 augmentations, closer to the practice in self-supervised learning. Our evaluations on Image classification (ImageNet-1k with and without pre-training on ImageNet-21k), transfer learning and semantic segmentation show that our procedure outperforms by a large margin previous fully supervised training recipes for ViT. It also reveals that the performance of our ViT trained with supervision is comparable to that of more recent architectures. Our results could serve as better baselines for recent self-supervised approaches demonstrated on ViT."
                },
                {
                    "title": "Exploring Plain Vision Transformer Backbones for Object Detection",
                    "abstract": "We explore the plain, non-hierarchical Vision Transformer (ViT) as a backbone network for object detection. This design enables the original ViT architecture to be fine-tuned for object detection without needing to redesign a hierarchical backbone for pre-training. With minimal adaptations for fine-tuning, our plain-backbone detector can achieve competitive results. Surprisingly, we observe: (i) it is sufficient to build a simple feature pyramid from a single-scale feature map (without the common FPN design) and (ii) it is sufficient to use window attention (without shifting) aided with very few cross-window propagation blocks. With plain ViT backbones pre-trained as Masked Autoencoders (MAE), our detector, named ViTDet, can compete with the previous leading methods that were all based on hierarchical backbones, reaching up to 61.3 AP_box on the COCO dataset using only ImageNet-1K pre-training. We hope our study will draw attention to research on plain-backbone detectors. Code for ViTDet is available in Detectron2."
                },
                {
                    "title": "Three things everyone should know about Vision Transformers",
                    "abstract": "After their initial success in natural language processing, transformer architectures have rapidly gained traction in computer vision, providing state-of-the-art results for tasks such as image classification, detection, segmentation, and video analysis. We offer three insights based on simple and easy to implement variants of vision transformers. (1) The residual layers of vision transformers, which are usually processed sequentially, can to some extent be processed efficiently in parallel without noticeably affecting the accuracy. (2) Fine-tuning the weights of the attention layers is sufficient to adapt vision transformers to a higher resolution and to other classification tasks. This saves compute, reduces the peak memory consumption at fine-tuning time, and allows sharing the majority of weights across tasks. (3) Adding MLP-based patch pre-processing layers improves Bert-like self-supervised training based on patch masking. We evaluate the impact of these design choices using the ImageNet-1k dataset, and confirm our findings on the ImageNet-v2 test set. Transfer performance is measured across six smaller datasets."
                },
                {
                    "title": "Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution",
                    "abstract": "When transferring a pretrained model to a downstream task, two popular methods are full fine-tuning (updating all the model parameters) and linear probing (updating only the last linear layer\u2014the \u201chead\u201d). It is well known that fine-tuning leads to better accuracy in-distribution (ID). However, in this paper, we find that fine-tuning can achieve worse accuracy than linear probing out-of-distribution (OOD) when the pretrained features are good and the distribution shift is large. On 10 distribution shift datasets (Breeds-Living17, Breeds-Entity30, DomainNet, CIFAR \u2192 STL, CIFAR10.1, FMoW, ImageNetV2, ImageNet-R, ImageNet-A, ImageNet-Sketch), fine-tuning obtains on average 2% higher accuracy ID but 7% lower accuracy OOD than linear probing. We show theoretically that this tradeoff between ID and OOD accuracy arises even in a simple setting: fine-tuning overparameterized two-layer linear networks. We prove that the OOD error of fine-tuning is high when we initialize with a fixed or random head\u2014this is because while fine-tuning learns the head, the lower layers of the neural network change simultaneously and distort the pretrained features. Our analysis suggests that the easy two-step strategy of linear probing then full fine-tuning (LP-FT), sometimes used as a fine-tuning heuristic, combines the benefits of both fine-tuning and linear probing. Empirically, LP-FT outperforms both fine-tuning and linear probing on the above datasets (1% better ID, 10% better OOD than full fine-tuning)."
                },
                {
                    "title": "Searching the Search Space of Vision Transformer",
                    "abstract": "Vision Transformer has shown great visual representation power in substantial vision tasks such as recognition and detection, and thus been attracting fast-growing efforts on manually designing more effective architectures. In this paper, we propose to use neural architecture search to automate this process, by searching not only the architecture but also the search space. The central idea is to gradually evolve different search dimensions guided by their E-T Error computed using a weight-sharing supernet. Moreover, we provide design guidelines of general vision transformers with extensive analysis according to the space searching process, which could promote the understanding of vision transformer. Remarkably, the searched models, named S3 (short for Searching the Search Space), from the searched space achieve superior performance to recently proposed models, such as Swin, DeiT and ViT, when evaluated on ImageNet. The effectiveness of S3 is also illustrated on object detection, semantic segmentation and visual question answering, demonstrating its generality to downstream vision and vision-language tasks. Code and models will be available at https://github.com/microsoft/Cream."
                },
                {
                    "title": "Swin Transformer V2: Scaling Up Capacity and Resolution",
                    "abstract": "We present techniques for scaling Swin Transformer [35] up to 3 billion parameters and making it capable of training with images of up to 1,536x1,536 resolution. By scaling up capacity and resolution, Swin Transformer sets new records on four representative vision benchmarks: 84.0% top-1 accuracy on ImageNet- V2 image classification, 63.1 / 54.4 box / mask mAP on COCO object detection, 59.9 mIoU on ADE20K semantic segmentation, and 86.8% top-1 accuracy on Kinetics-400 video action classification. We tackle issues of training instability, and study how to effectively transfer models pre-trained at low resolutions to higher resolution ones. To this aim, several novel technologies are proposed: 1) a residual post normalization technique and a scaled cosine attention approach to improve the stability of large vision models; 2) a log-spaced continuous position bias technique to effectively transfer models pre-trained at low-resolution images and windows to their higher-resolution counterparts. In addition, we share our crucial implementation details that lead to significant savings of GPU memory consumption and thus make it feasi-ble to train large vision models with regular GPUs. Using these techniques and self-supervised pre-training, we suc-cessfully train a strong 3 billion Swin Transformer model and effectively transfer it to various vision tasks involving high-resolution images or windows, achieving the state-of-the-art accuracy on a variety of benchmarks. Code is avail-able at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation",
                    "abstract": "Since the introduction of the transformer model by Vaswani et al. (2017), a fundamental question has yet to be answered: how does a model achieve extrapolation at inference time for sequences that are longer than it saw during training? We first show that extrapolation can be enabled by simply changing the position representation method, though we find that current methods do not allow for efficient extrapolation. We therefore introduce a simpler and more efficient position method, Attention with Linear Biases (ALiBi). ALiBi does not add positional embeddings to word embeddings; instead, it biases query-key attention scores with a penalty that is proportional to their distance. We show that this method trains a 1.3 billion parameter model on input sequences of length 1024 that extrapolates to input sequences of length 2048, achieving the same perplexity as a sinusoidal position embedding model trained on inputs of length 2048 but training 11% faster and using 11% less memory. ALiBi's inductive bias towards recency also leads it to outperform multiple strong position methods on the WikiText-103 benchmark."
                },
                {
                    "title": "Do Vision Transformers See Like Convolutional Neural Networks?",
                    "abstract": "Convolutional neural networks (CNNs) have so far been the de-facto model for visual data. Recent work has shown that (Vision) Transformer models (ViT) can achieve comparable or even superior performance on image classification tasks. This raises a central question: how are Vision Transformers solving these tasks? Are they acting like convolutional networks, or learning entirely different visual representations? Analyzing the internal representation structure of ViTs and CNNs on image classification benchmarks, we find striking differences between the two architectures, such as ViT having more uniform representations across all layers. We explore how these differences arise, finding crucial roles played by self-attention, which enables early aggregation of global information, and ViT residual connections, which strongly propagate features from lower to higher layers. We study the ramifications for spatial localization, demonstrating ViTs successfully preserve input spatial information, with noticeable effects from different classification methods. Finally, we study the effect of (pretraining) dataset scale on intermediate features and transfer learning, and conclude with a discussion on connections to new architectures such as the MLP-Mixer."
                },
                {
                    "title": "Rethinking and Improving Relative Position Encoding for Vision Transformer",
                    "abstract": "Relative position encoding (RPE) is important for transformer to capture sequence ordering of input tokens. General efficacy has been proven in natural language processing. However, in computer vision, its efficacy is not well studied and even remains controversial, e.g., whether relative position encoding can work equally well as absolute position? In order to clarify this, we first review existing relative position encoding methods and analyze their pros and cons when applied in vision transformers. We then propose new relative position encoding methods dedicated to 2D images, called image RPE (iRPE). Our methods consider directional relative distance modeling as well as the interactions between queries and relative position embeddings in self-attention mechanism. The proposed iRPE methods are simple and lightweight. They can be easily plugged into transformer blocks. Experiments demonstrate that solely due to the proposed encoding methods, DeiT [21] and DETR [1] obtain up to 1.5% (top-1 Acc) and 1.3% (mAP) stable improvements over their original versions on ImageNet and COCO respectively, without tuning any extra hyperparameters such as learning rate and weight decay. Our ablation and analysis also yield interesting findings, some of which run counter to previous understanding. Code and models are open-sourced at https://github.com/microsoft/Cream/tree/main/iRPE."
                },
                {
                    "title": "Focal Self-attention for Local-Global Interactions in Vision Transformers",
                    "abstract": "Recently, Vision Transformer and its variants have shown great promise on various computer vision tasks. The ability of capturing short- and long-range visual dependencies through self-attention is arguably the main source for the success. But it also brings challenges due to quadratic computational overhead, especially for the high-resolution vision tasks (e.g., object detection). In this paper, we present focal self-attention, a new mechanism that incorporates both fine-grained local and coarse-grained global interactions. Using this new mechanism, each token attends the closest surrounding tokens at fine granularity but the tokens far away at coarse granularity, and thus can capture both short- and long-range visual dependencies efficiently and effectively. With focal self-attention, we propose a new variant of Vision Transformer models, called Focal Transformer, which achieves superior performance over the state-of-the-art vision Transformers on a range of public image classification and object detection benchmarks. In particular, our Focal Transformer models with a moderate size of 51.1M and a larger size of 89.8M achieve 83.5 and 83.8 Top-1 accuracy, respectively, on ImageNet classification at 224x224 resolution. Using Focal Transformers as the backbones, we obtain consistent and substantial improvements over the current state-of-the-art Swin Transformers for 6 different object detection methods trained with standard 1x and 3x schedules. Our largest Focal Transformer yields 58.7/58.9 box mAPs and 50.9/51.3 mask mAPs on COCO mini-val/test-dev, and 55.4 mIoU on ADE20K for semantic segmentation, creating new SoTA on three of the most challenging computer vision tasks."
                },
                {
                    "title": "How to train your ViT? Data, Augmentation, and Regularization in Vision Transformers",
                    "abstract": "Vision Transformers (ViT) have been shown to attain highly competitive performance for a wide range of vision applications, such as image classification, object detection and semantic image segmentation. In comparison to convolutional neural networks, the Vision Transformer's weaker inductive bias is generally found to cause an increased reliance on model regularization or data augmentation (\"AugReg\"for short) when training on smaller training datasets. We conduct a systematic empirical study in order to better understand the interplay between the amount of training data, AugReg, model size and compute budget. As one result of this study we find that the combination of increased compute and AugReg can yield models with the same performance as models trained on an order of magnitude more training data: we train ViT models of various sizes on the public ImageNet-21k dataset which either match or outperform their counterparts trained on the larger, but not publicly available JFT-300M dataset."
                },
                {
                    "title": "XCiT: Cross-Covariance Image Transformers",
                    "abstract": "Following their success in natural language processing, transformers have recently shown much promise for computer vision. The self-attention operation underlying transformers yields global interactions between all tokens ,i.e. words or image patches, and enables flexible modelling of image data beyond the local interactions of convolutions. This flexibility, however, comes with a quadratic complexity in time and memory, hindering application to long sequences and high-resolution images. We propose a\"transposed\"version of self-attention that operates across feature channels rather than tokens, where the interactions are based on the cross-covariance matrix between keys and queries. The resulting cross-covariance attention (XCA) has linear complexity in the number of tokens, and allows efficient processing of high-resolution images. Our cross-covariance image transformer (XCiT) is built upon XCA. It combines the accuracy of conventional transformers with the scalability of convolutional architectures. We validate the effectiveness and generality of XCiT by reporting excellent results on multiple vision benchmarks, including image classification and self-supervised feature learning on ImageNet-1k, object detection and instance segmentation on COCO, and semantic segmentation on ADE20k."
                },
                {
                    "title": "BEiT: BERT Pre-Training of Image Transformers",
                    "abstract": "We introduce a self-supervised vision representation model BEiT, which stands for Bidirectional Encoder representation from Image Transformers. Following BERT developed in the natural language processing area, we propose a masked image modeling task to pretrain vision Transformers. Specifically, each image has two views in our pre-training, i.e, image patches (such as 16x16 pixels), and visual tokens (i.e., discrete tokens). We first\"tokenize\"the original image into visual tokens. Then we randomly mask some image patches and fed them into the backbone Transformer. The pre-training objective is to recover the original visual tokens based on the corrupted image patches. After pre-training BEiT, we directly fine-tune the model parameters on downstream tasks by appending task layers upon the pretrained encoder. Experimental results on image classification and semantic segmentation show that our model achieves competitive results with previous pre-training methods. For example, base-size BEiT achieves 83.2% top-1 accuracy on ImageNet-1K, significantly outperforming from-scratch DeiT training (81.8%) with the same setup. Moreover, large-size BEiT obtains 86.3% only using ImageNet-1K, even outperforming ViT-L with supervised pre-training on ImageNet-22K (85.2%). The code and pretrained models are available at https://aka.ms/beit."
                },
                {
                    "title": "Chasing Sparsity in Vision Transformers: An End-to-End Exploration",
                    "abstract": "Vision transformers (ViTs) have recently received explosive popularity, but their enormous model sizes and training costs remain daunting. Conventional post-training pruning often incurs higher training budgets. In contrast, this paper aims to trim down both the training memory overhead and the inference complexity, without sacrificing the achievable accuracy. We carry out the first-of-its-kind comprehensive exploration, on taking a unified approach of integrating sparsity in ViTs\"from end to end\". Specifically, instead of training full ViTs, we dynamically extract and train sparse subnetworks, while sticking to a fixed small parameter budget. Our approach jointly optimizes model parameters and explores connectivity throughout training, ending up with one sparse network as the final output. The approach is seamlessly extended from unstructured to structured sparsity, the latter by considering to guide the prune-and-grow of self-attention heads inside ViTs. We further co-explore data and architecture sparsity for additional efficiency gains by plugging in a novel learnable token selector to adaptively determine the currently most vital patches. Extensive results on ImageNet with diverse ViT backbones validate the effectiveness of our proposals which obtain significantly reduced computational cost and almost unimpaired generalization. Perhaps most surprisingly, we find that the proposed sparse (co-)training can sometimes improve the ViT accuracy rather than compromising it, making sparsity a tantalizing\"free lunch\". For example, our sparsified DeiT-Small at (5%, 50%) sparsity for (data, architecture), improves 0.28% top-1 accuracy, and meanwhile enjoys 49.32% FLOPs and 4.40% running time savings. Our codes are available at https://github.com/VITA-Group/SViTE."
                },
                {
                    "title": "ViTAE: Vision Transformer Advanced by Exploring Intrinsic Inductive Bias",
                    "abstract": "Transformers have shown great potential in various computer vision tasks owing to their strong capability in modeling long-range dependency using the self-attention mechanism. Nevertheless, vision transformers treat an image as 1D sequence of visual tokens, lacking an intrinsic inductive bias (IB) in modeling local visual structures and dealing with scale variance. Alternatively, they require large-scale training data and longer training schedules to learn the IB implicitly. In this paper, we propose a novel Vision Transformer Advanced by Exploring intrinsic IB from convolutions, ie, ViTAE. Technically, ViTAE has several spatial pyramid reduction modules to downsample and embed the input image into tokens with rich multi-scale context by using multiple convolutions with different dilation rates. In this way, it acquires an intrinsic scale invariance IB and is able to learn robust feature representation for objects at various scales. Moreover, in each transformer layer, ViTAE has a convolution block in parallel to the multi-head self-attention module, whose features are fused and fed into the feed-forward network. Consequently, it has the intrinsic locality IB and is able to learn local features and global dependencies collaboratively. Experiments on ImageNet as well as downstream tasks prove the superiority of ViTAE over the baseline transformer and concurrent works. Source code and pretrained models will be available at GitHub."
                },
                {
                    "title": "Large-Scale Unsupervised Semantic Segmentation",
                    "abstract": "Empowered by large datasets, e.g., ImageNet and MS COCO, unsupervised learning on large-scale data has enabled significant advances for classification tasks. However, whether the large-scale unsupervised semantic segmentation can be achieved remains unknown. There are two major challenges: i) we need a large-scale benchmark for assessing algorithms; ii) we need to develop methods to simultaneously learn category and shape representation in an unsupervised manner. In this work, we propose a new problem of large-scale unsupervised semantic segmentation (LUSS) with a newly created benchmark dataset to help the research progress. Building on the ImageNet dataset, we propose the ImageNet-S dataset with 1.2 million training images and 50k high-quality semantic segmentation annotations for evaluation. Our benchmark has a high data diversity and a clear task objective. We also present a simple yet effective method that works surprisingly well for LUSS. In addition, we benchmark related un/weakly/fully supervised methods accordingly, identifying the challenges and possible directions of LUSS. The benchmark and source code is publicly available at https://github.com/LUSSeg."
                },
                {
                    "title": "Learnable Fourier Features for Multi-Dimensional Spatial Positional Encoding",
                    "abstract": "Attentional mechanisms are order-invariant. Positional encoding is a crucial component to allow attention-based deep model architectures such as Transformer to address sequences or images where the position of information matters. In this paper, we propose a novel positional encoding method based on learnable Fourier features. Instead of hard-coding each position as a token or a vector, we represent each position, which can be multi-dimensional, as a trainable encoding based on learnable Fourier feature mapping, modulated with a multi-layer perceptron. The representation is particularly advantageous for a spatial multi-dimensional position, e.g., pixel positions on an image, where $L_2$ distances or more complex positional relationships need to be captured. Our experiments based on several public benchmark tasks show that our learnable Fourier feature representation for multi-dimensional positional encoding outperforms existing methods by both improving the accuracy and allowing faster convergence."
                },
                {
                    "title": "When Vision Transformers Outperform ResNets without Pretraining or Strong Data Augmentations",
                    "abstract": "Vision Transformers (ViTs) and MLPs signal further efforts on replacing hand-wired features or inductive biases with general-purpose neural architectures. Existing works empower the models by massive data, such as large-scale pre-training and/or repeated strong data augmentations, and still report optimization-related problems (e.g., sensitivity to initialization and learning rates). Hence, this paper investigates ViTs and MLP-Mixers from the lens of loss geometry, intending to improve the models' data efficiency at training and generalization at inference. Visualization and Hessian reveal extremely sharp local minima of converged models. By promoting smoothness with a recently proposed sharpness-aware optimizer, we substantially improve the accuracy and robustness of ViTs and MLP-Mixers on various tasks spanning supervised, adversarial, contrastive, and transfer learning (e.g., +5.3\\% and +11.0\\% top-1 accuracy on ImageNet for ViT-B/16 and Mixer-B/16, respectively, with the simple Inception-style preprocessing). We show that the improved smoothness attributes to sparser active neurons in the first few layers. The resultant ViTs outperform ResNets of similar size and throughput when trained from scratch on ImageNet without large-scale pre-training or strong data augmentations. Model checkpoints are available at \\url{https://github.com/google-research/vision_transformer}."
                },
                {
                    "title": "Segmenter: Transformer for Semantic Segmentation",
                    "abstract": "Image segmentation is often ambiguous at the level of individual image patches and requires contextual information to reach label consensus. In this paper we introduce Segmenter, a transformer model for semantic segmentation. In contrast to convolution-based methods, our approach allows to model global context already at the first layer and throughout the network. We build on the recent Vision Transformer (ViT) and extend it to semantic segmentation. To do so, we rely on the output embeddings corresponding to image patches and obtain class labels from these embed-dings with a point-wise linear decoder or a mask trans-former decoder. We leverage models pre-trained for image classification and show that we can fine-tune them on moderate sized datasets available for semantic segmentation. The linear decoder allows to obtain excellent results already, but the performance can be further improved by a mask transformer generating class masks. We conduct an extensive ablation study to show the impact of the different parameters, in particular the performance is better for large models and small patch sizes. Segmenter attains excellent results for semantic segmentation. It outperforms the state of the art on both ADE20K and Pascal Context datasets and is competitive on Cityscapes."
                },
                {
                    "title": "Twins: Revisiting the Design of Spatial Attention in Vision Transformers",
                    "abstract": "Very recently, a variety of vision transformer architectures for dense prediction tasks have been proposed and they show that the design of spatial attention is critical to their success in these tasks. In this work, we revisit the design of the spatial attention and demonstrate that a carefully-devised yet simple spatial attention mechanism performs favourably against the state-of-the-art schemes. As a result, we propose two vision transformer architectures, namely, Twins-PCPVT and Twins-SVT. Our proposed architectures are highly-efficient and easy to implement, only involving matrix multiplications that are highly optimized in modern deep learning frameworks. More importantly, the proposed architectures achieve excellent performance on a wide range of visual tasks, including image level classification as well as dense detection and segmentation. The simplicity and strong performance suggest that our proposed architectures may serve as stronger backbones for many vision tasks. Our code is released at https://github.com/Meituan-AutoML/Twins ."
                },
                {
                    "title": "VATT: Transformers for Multimodal Self-Supervised Learning from Raw Video, Audio and Text",
                    "abstract": "We present a framework for learning multimodal representations from unlabeled data using convolution-free Transformer architectures. Specifically, our Video-Audio-Text Transformer (VATT) takes raw signals as inputs and extracts multimodal representations that are rich enough to benefit a variety of downstream tasks. We train VATT end-to-end from scratch using multimodal contrastive losses and evaluate its performance by the downstream tasks of video action recognition, audio event classification, image classification, and text-to-video retrieval. Furthermore, we study a modality-agnostic, single-backbone Transformer by sharing weights among the three modalities. We show that the convolution-free VATT outperforms state-of-the-art ConvNet-based architectures in the downstream tasks. Especially, VATT's vision Transformer achieves the top-1 accuracy of 82.1% on Kinetics-400, 83.6% on Kinetics-600, 72.7% on Kinetics-700, and 41.1% on Moments in Time, new records while avoiding supervised pre-training. Transferring to image classification leads to 78.7% top-1 accuracy on ImageNet compared to 64.7% by training the same Transformer from scratch, showing the generalizability of our model despite the domain gap between videos and images. VATT's audio Transformer also sets a new record on waveform-based audio event recognition by achieving the mAP of 39.4% on AudioSet without any supervised pre-training. VATT's source code is publicly available."
                },
                {
                    "title": "Co-Scale Conv-Attentional Image Transformers",
                    "abstract": "In this paper, we present Co-scale conv-attentional image Transformers (CoaT), a Transformer-based image classifier equipped with co-scale and conv-attentional mechanisms. First, the co-scale mechanism maintains the integrity of Transformers\u2019 encoder branches at individual scales, while allowing representations learned at different scales to effectively communicate with each other; we design a series of serial and parallel blocks to realize the co-scale mechanism. Second, we devise a conv-attentional mechanism by realizing a relative position embedding formulation in the factorized attention module with an efficient convolution-like implementation. CoaT empowers image Transformers with enriched multi-scale and contextual modeling capabilities. On ImageNet, relatively small CoaT models attain superior classification results compared with similar-sized convolutional neural networks and image/vision Transformers. The effectiveness of CoaT\u2019s backbone is also illustrated on object detection and instance segmentation, demonstrating its applicability to downstream computer vision tasks."
                },
                {
                    "title": "ConViT: improving vision transformers with soft convolutional inductive biases",
                    "abstract": "Convolutional architectures have proven to be extremely successful for vision tasks. Their hard inductive biases enable sample-efficient learning, but come at the cost of a potentially lower performance ceiling. Vision transformers rely on more flexible self-attention layers, and have recently outperformed CNNs for image classification. However, they require costly pre-training on large external datasets or distillation from pre-trained convolutional networks. In this paper, we ask the following question: is it possible to combine the strengths of these two architectures while avoiding their respective limitations? To this end, we introduce gated positional self-attention (GPSA), a form of positional self-attention which can be equipped with a \u2018soft\u2019 convolutional inductive bias. We initialize the GPSA layers to mimic the locality of convolutional layers, then give each attention head the freedom to escape locality by adjusting a gating parameter regulating the attention paid to position versus content information. The resulting convolutional-like ViT architecture, ConViT, outperforms the DeiT (Touvron et al 2020 arXiv:2012.12877) on ImageNet, while offering a much improved sample efficiency. We further investigate the role of locality in learning by first quantifying how it is encouraged in vanilla self-attention layers, then analyzing how it has escaped in GPSA layers. We conclude by presenting various ablations to better understand the success of the ConViT. Our code and models are released publicly at https://github.com/facebookresearch/convit."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Conditional Positional Encodings for Vision Transformers",
                    "abstract": "We propose a conditional positional encoding (CPE) scheme for vision Transformers. Unlike previous fixed or learnable positional encodings, which are pre-defined and independent of input tokens, CPE is dynamically generated and conditioned on the local neighborhood of the input tokens. As a result, CPE can easily generalize to the input sequences that are longer than what the model has ever seen during training. Besides, CPE can keep the desired translation-invariance in the image classification task, resulting in improved performance. We implement CPE with a simple Position Encoding Generator (PEG) to get seamlessly incorporated into the current Transformer framework. Built on PEG, we present Conditional Position encoding Vision Transformer (CPVT). We demonstrate that CPVT has visually similar attention maps compared to those with learned positional encodings and delivers outperforming results. Our code is available at https://github.com/Meituan-AutoML/CPVT ."
                },
                {
                    "title": "Training data-efficient image transformers & distillation through attention",
                    "abstract": "Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption. In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data. More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models."
                },
                {
                    "title": "Losing Heads in the Lottery: Pruning Transformer Attention in Neural Machine Translation",
                    "abstract": "The attention mechanism is the crucial component of the transformer architecture. Recent research shows that most attention heads are not confident in their decisions and can be pruned. However, removing them before training a model results in lower quality. In this paper, we apply the lottery ticket hypothesis to prune heads in the early stages of training. Our experiments on machine translation show that it is possible to remove up to three-quarters of attention heads from transformer-big during early training with an average -0.1 change in BLEU for Turkish\u2192English. The pruned model is 1.5 times as fast at inference, albeit at the cost of longer training. Our method is complementary to other approaches, such as teacher-student, with English\u2192German student model gaining an additional 10% speed-up with 75% encoder attention removed and 0.2 BLEU loss."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization",
                    "abstract": "We introduce four new real-world distribution shift datasets consisting of changes in image style, image blurriness, geographic location, camera operation, and more. With our new datasets, we take stock of previously proposed methods for improving out-of-distribution robustness and put them to the test. We find that using larger models and artificial data augmentations can improve robustness on real-world distribution shifts, contrary to claims in prior work. We find improvements in artificial robustness benchmarks can transfer to real-world distribution shifts, contrary to claims in prior work. Motivated by our observation that data augmentations can help with real-world distribution shifts, we also introduce a new data augmentation method which advances the state-of-the-art and outperforms models pre-trained with 1000\u00d7 more labeled data. Overall we find that some methods consistently help with distribution shifts in texture and local image statistics, but these methods do not help with some other distribution shifts like geographic changes. Our results show that future research must study multiple distribution shifts simultaneously, as we demonstrate that no evaluated method consistently improves robustness."
                },
                {
                    "title": "Are we done with ImageNet?",
                    "abstract": "Yes, and no. We ask whether recent progress on the ImageNet classification benchmark continues to represent meaningful generalization, or whether the community has started to overfit to the idiosyncrasies of its labeling procedure. We therefore develop a significantly more robust procedure for collecting human annotations of the ImageNet validation set. Using these new labels, we reassess the accuracy of recently proposed ImageNet classifiers, and find their gains to be substantially smaller than those reported on the original labels. Furthermore, we find the original ImageNet labels to no longer be the best predictors of this independently-collected set, indicating that their usefulness in evaluating vision models may be nearing an end. Nevertheless, we find our annotation procedure to have largely remedied the errors in the original labels, reinforcing ImageNet as a powerful benchmark for future research in visual recognition."
                },
                {
                    "title": "Randaugment: Practical automated data augmentation with a reduced search space",
                    "abstract": "Recent work on automated augmentation strategies has led to state-of-the-art results in image classification and object detection. An obstacle to a large-scale adoption of these methods is that they require a separate and expensive search phase. A common way to overcome the expense of the search phase was to use a smaller proxy task. However, it was not clear if the optimized hyperparameters found on the proxy task are also optimal for the actual task. In this work, we rethink the process of designing automated augmentation strategies. We find that while previous work required a search for both magnitude and probability of each operation independently, it is sufficient to only search for a single distortion magnitude that jointly controls all operations. We hence propose a simplified search space that vastly reduces the computational expense of automated augmentation, and permits the removal of a separate proxy task.Despite the simplifications, our method achieves equal or better performance over previous automated augmentation strategies on on CIFAR-10/100, SVHN, ImageNet and COCO datasets. EfficientNet-B7, we achieve 85.0% accuracy, a 1.0% increase over baseline augmentation, a 0.6% improvement over AutoAugment on the ImageNet dataset. With EfficientNet-B8, we achieve 85.4% accuracy on ImageNet, which matches a previous result that used 3.5B extra images. On object detection, the same method as classification leads to 1.0-1.3% improvement over baseline augmentation. Code will be made available online."
                },
                {
                    "title": "Multi-Head Attention with Diversity for Learning Grounded Multilingual Multimodal Representations",
                    "abstract": "With the aim of promoting and understanding the multilingual version of image search, we leverage visual object detection and propose a model with diverse multi-head attention to learn grounded multilingual multimodal representations. Specifically, our model attends to different types of textual semantics in two languages and visual objects for fine-grained alignments between sentences and images. We introduce a new objective function which explicitly encourages attention diversity to learn an improved visual-semantic embedding space. We evaluate our model in the German-Image and English-Image matching tasks on the Multi30K dataset, and in the Semantic Textual Similarity task with the English descriptions of visual content. Results show that our model yields a significant performance gain over other methods in all of the three tasks."
                },
                {
                    "title": "Natural Adversarial Examples",
                    "abstract": "We introduce two challenging datasets that reliably cause machine learning model performance to substantially degrade. The datasets are collected with a simple adversarial filtration technique to create datasets with limited spurious cues. Our datasets\u2019 real-world, unmodified examples transfer to various unseen models reliably, demonstrating that computer vision models have shared weaknesses. The first dataset is called IMAGENET-A and is like the ImageNet test set, but it is far more challenging for existing models. We also curate an adversarial out-of-distribution detection dataset called IMAGENET-O, which is the first out-of-distribution detection dataset created for ImageNet models. On IMAGENET-A a DenseNet-121 obtains around 2% accuracy, an accuracy drop of approximately 90%, and its out-of-distribution detection performance on IMAGENET-O is near random chance levels. We find that existing data augmentation techniques hardly boost performance, and using other public training datasets provides improvements that are limited. However, we find that improvements to computer vision architectures provide a promising path towards robust models."
                },
                {
                    "title": "Fixing the train-test resolution discrepancy",
                    "abstract": "Data-augmentation is key to the training of neural networks for image classification. This paper first shows that existing augmentations induce a significant discrepancy between the typical size of the objects seen by the classifier at train and test time. We experimentally validate that, for a target test resolution, using a lower train resolution offers better classification at test time. \nWe then propose a simple yet effective and efficient strategy to optimize the classifier performance when the train and test resolutions differ. It involves only a computationally cheap fine-tuning of the network at the test resolution. This enables training strong classifiers using small training images. For instance, we obtain 77.1% top-1 accuracy on ImageNet with a ResNet-50 trained on 128x128 images, and 79.8% with one trained on 224x224 image. In addition, if we use extra training data we get 82.5% with the ResNet-50 train with 224x224 images. \nConversely, when training a ResNeXt-101 32x48d pre-trained in weakly-supervised fashion on 940 million public images at resolution 224x224 and further optimizing for test resolution 320x320, we obtain a test top-1 accuracy of 86.4% (top-5: 98.0%) (single-crop). To the best of our knowledge this is the highest ImageNet single-crop, top-1 and top-5 accuracy to date."
                },
                {
                    "title": "Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift",
                    "abstract": "Modern machine learning methods including deep learning have achieved great success in predictive accuracy for supervised learning tasks, but may still fall short in giving useful estimates of their predictive {\\em uncertainty}. Quantifying uncertainty is especially critical in real-world settings, which often involve input distributions that are shifted from the training distribution due to a variety of factors including sample bias and non-stationarity. In such settings, well calibrated uncertainty estimates convey information about when a model's output should (or should not) be trusted. Many probabilistic deep learning methods, including Bayesian-and non-Bayesian methods, have been proposed in the literature for quantifying predictive uncertainty, but to our knowledge there has not previously been a rigorous large-scale empirical comparison of these methods under dataset shift. We present a large-scale benchmark of existing state-of-the-art methods on classification problems and investigate the effect of dataset shift on accuracy and calibration. We find that traditional post-hoc calibration does indeed fall short, as do several other previous methods. However, some methods that marginalize over models give surprisingly strong results across a broad spectrum of tasks."
                },
                {
                    "title": "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks",
                    "abstract": "Convolutional Neural Networks (ConvNets) are commonly developed at a fixed resource budget, and then scaled up for better accuracy if more resources are available. In this paper, we systematically study model scaling and identify that carefully balancing network depth, width, and resolution can lead to better performance. Based on this observation, we propose a new scaling method that uniformly scales all dimensions of depth/width/resolution using a simple yet highly effective compound coefficient. We demonstrate the effectiveness of this method on scaling up MobileNets and ResNet. \nTo go even further, we use neural architecture search to design a new baseline network and scale it up to obtain a family of models, called EfficientNets, which achieve much better accuracy and efficiency than previous ConvNets. In particular, our EfficientNet-B7 achieves state-of-the-art 84.4% top-1 / 97.1% top-5 accuracy on ImageNet, while being 8.4x smaller and 6.1x faster on inference than the best existing ConvNet. Our EfficientNets also transfer well and achieve state-of-the-art accuracy on CIFAR-100 (91.7%), Flowers (98.8%), and 3 other transfer learning datasets, with an order of magnitude fewer parameters. Source code is at this https URL."
                },
                {
                    "title": "Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned",
                    "abstract": "Multi-head self-attention is a key component of the Transformer, a state-of-the-art architecture for neural machine translation. In this work we evaluate the contribution made by individual attention heads to the overall performance of the model and analyze the roles played by them in the encoder. We find that the most important and confident heads play consistent and often linguistically-interpretable roles. When pruning heads using a method based on stochastic gates and a differentiable relaxation of the L0 penalty, we observe that specialized heads are last to be pruned. Our novel pruning method removes the vast majority of heads without seriously affecting performance. For example, on the English-Russian WMT dataset, pruning 38 out of 48 encoder heads results in a drop of only 0.15 BLEU."
                },
                {
                    "title": "CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features",
                    "abstract": "Regional dropout strategies have been proposed to enhance performance of convolutional neural network classifiers. They have proved to be effective for guiding the model to attend on less discriminative parts of objects (e.g. leg as opposed to head of a person), thereby letting the network generalize better and have better object localization capabilities. On the other hand, current methods for regional dropout removes informative pixels on training images by overlaying a patch of either black pixels or random noise. Such removal is not desirable because it suffers from information loss causing inefficiency in training. We therefore propose the CutMix augmentation strategy: patches are cut and pasted among training images where the ground truth labels are also mixed proportionally to the area of the patches. By making efficient use of training pixels and retaining the regularization effect of regional dropout, CutMix consistently outperforms state-of-the-art augmentation strategies on CIFAR and ImageNet classification tasks, as well as on ImageNet weakly-supervised localization task. Moreover, unlike previous augmentation methods, our CutMix-trained ImageNet classifier, when used as a pretrained model, results in consistent performance gain in Pascal detection and MS-COCO image captioning benchmarks. We also show that CutMix can improve the model robustness against input corruptions and its out-of distribution detection performance."
                },
                {
                    "title": "Are Sixteen Heads Really Better than One?",
                    "abstract": "Attention is a powerful and ubiquitous mechanism for allowing neural models to focus on particular salient pieces of information by taking their weighted average when making predictions. In particular, multi-headed attention is a driving force behind many recent state-of-the-art NLP models such as Transformer-based MT models and BERT. These models apply multiple attention mechanisms in parallel, with each attention \"head\" potentially focusing on different parts of the input, which makes it possible to express sophisticated functions beyond the simple weighted average. In this paper we make the surprising observation that even if models have been trained using multiple heads, in practice, a large percentage of attention heads can be removed at test time without significantly impacting performance. In fact, some layers can even be reduced to a single head. We further examine greedy algorithms for pruning down models, and the potential speed, memory efficiency, and accuracy improvements obtainable therefrom. Finally, we analyze the results with respect to which parts of the model are more reliant on having multiple heads, and provide precursory evidence that training dynamics play a role in the gains provided by multi-head attention."
                },
                {
                    "title": "Do ImageNet Classifiers Generalize to ImageNet?",
                    "abstract": "We build new test sets for the CIFAR-10 and ImageNet datasets. Both benchmarks have been the focus of intense research for almost a decade, raising the danger of overfitting to excessively re-used test sets. By closely following the original dataset creation processes, we test to what extent current classification models generalize to new data. We evaluate a broad range of models and find accuracy drops of 3% - 15% on CIFAR-10 and 11% - 14% on ImageNet. However, accuracy gains on the original test sets translate to larger gains on the new test sets. Our results suggest that the accuracy drops are not caused by adaptivity, but by the models' inability to generalize to slightly \"harder\" images than those found in the original test sets."
                },
                {
                    "title": "Augment your batch: better training with larger batches",
                    "abstract": "Large-batch SGD is important for scaling training of deep neural networks. However, without fine-tuning hyperparameter schedules, the generalization of the model may be hampered. We propose to use batch augmentation: replicating instances of samples within the same batch with different data augmentations. Batch augmentation acts as a regularizer and an accelerator, increasing both generalization and performance scaling. We analyze the effect of batch augmentation on gradient variance and show that it empirically improves convergence for a wide variety of deep neural networks and datasets. Our results show that batch augmentation reduces the number of necessary SGD updates to achieve the same accuracy as the state-of-the-art. Overall, this simple yet effective method enables faster training and better generalization by allowing more computational resources to be used concurrently."
                },
                {
                    "title": "Improving Adversarial Robustness via Promoting Ensemble Diversity",
                    "abstract": "Though deep neural networks have achieved significant progress on various tasks, often enhanced by model ensemble, existing high-performance models can be vulnerable to adversarial attacks. Many efforts have been devoted to enhancing the robustness of individual networks and then constructing a straightforward ensemble, e.g., by directly averaging the outputs, which ignores the interaction among networks. This paper presents a new method that explores the interaction among individual networks to improve robustness for ensemble models. Technically, we define a new notion of ensemble diversity in the adversarial setting as the diversity among non-maximal predictions of individual members, and present an adaptive diversity promoting (ADP) regularizer to encourage the diversity, which leads to globally better robustness for the ensemble by making adversarial examples difficult to transfer among individual members. Our method is computationally efficient and compatible with the defense methods acting on individual networks. Empirical results on various datasets verify that our method can improve adversarial robustness while maintaining state-of-the-art accuracy on normal examples."
                },
                {
                    "title": "GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism",
                    "abstract": "Scaling up deep neural network capacity has been known as an effective approach to improving model quality for several different machine learning tasks. In many cases, increasing model capacity beyond the memory limit of a single accelerator has required developing special algorithms or infrastructure. These solutions are often architecture-specific and do not transfer to other tasks. To address the need for efficient and task-independent model parallelism, we introduce GPipe, a pipeline parallelism library that allows scaling any network that can be expressed as a sequence of layers. By pipelining different sub-sequences of layers on separate accelerators, GPipe provides the flexibility of scaling a variety of different networks to gigantic sizes efficiently. Moreover, GPipe utilizes a novel batch-splitting pipelining algorithm, resulting in almost linear speedup when a model is partitioned across multiple accelerators. We demonstrate the advantages of GPipe by training large-scale neural networks on two different tasks with distinct network architectures: (i) Image Classification: We train a 557-million-parameter AmoebaNet model and attain a top-1 accuracy of 84.4% on ImageNet-2012, (ii) Multilingual Neural Machine Translation: We train a single 6-billion-parameter, 128-layer Transformer model on a corpus spanning over 100 languages and achieve better quality than all bilingual models."
                },
                {
                    "title": "Multi-Head Attention with Disagreement Regularization",
                    "abstract": "Multi-head attention is appealing for the ability to jointly attend to information from different representation subspaces at different positions. In this work, we introduce a disagreement regularization to explicitly encourage the diversity among multiple attention heads. Specifically, we propose three types of disagreement regularization, which respectively encourage the subspace, the attended positions, and the output representation associated with each attention head to be different from other heads. Experimental results on widely-used WMT14 English-German and WMT17 Chinese-English translation tasks demonstrate the effectiveness and universality of the proposed approach."
                },
                {
                    "title": "Robustness May Be at Odds with Accuracy",
                    "abstract": "We show that there may exist an inherent tension between the goal of adversarial robustness and that of standard generalization. Specifically, training robust models may not only be more resource-consuming, but also lead to a reduction of standard accuracy. We demonstrate that this trade-off between the standard accuracy of a model and its robustness to adversarial perturbations provably exists in a fairly simple and natural setting. These findings also corroborate a similar phenomenon observed empirically in more complex settings. Further, we argue that this phenomenon is a consequence of robust classifiers learning fundamentally different feature representations than standard classifiers. These differences, in particular, seem to result in unexpected benefits: the representations learned by robust models tend to align better with salient data characteristics and human perception."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "mixup: Beyond Empirical Risk Minimization",
                    "abstract": "Large deep neural networks are powerful, but exhibit undesirable behaviors such as memorization and sensitivity to adversarial examples. In this work, we propose mixup, a simple learning principle to alleviate these issues. In essence, mixup trains a neural network on convex combinations of pairs of examples and their labels. By doing so, mixup regularizes the neural network to favor simple linear behavior in-between training examples. Our experiments on the ImageNet-2012, CIFAR-10, CIFAR-100, Google commands and UCI datasets show that mixup improves the generalization of state-of-the-art neural network architectures. We also find that mixup reduces the memorization of corrupt labels, increases the robustness to adversarial examples, and stabilizes the training of generative adversarial networks."
                },
                {
                    "title": "On Calibration of Modern Neural Networks",
                    "abstract": "Confidence calibration -- the problem of predicting probability estimates representative of the true correctness likelihood -- is important for classification models in many applications. We discover that modern neural networks, unlike those from a decade ago, are poorly calibrated. Through extensive experiments, we observe that depth, width, weight decay, and Batch Normalization are important factors influencing calibration. We evaluate the performance of various post-processing calibration methods on state-of-the-art architectures with image and document classification datasets. Our analysis and experiments not only offer insights into neural network learning, but also provide a simple and straightforward recipe for practical settings: on most datasets, temperature scaling -- a single-parameter variant of Platt Scaling -- is surprisingly effective at calibrating predictions."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "A Structured Self-attentive Sentence Embedding",
                    "abstract": "This paper proposes a new model for extracting an interpretable sentence embedding by introducing self-attention. Instead of using a vector, we use a 2-D matrix to represent the embedding, with each row of the matrix attending on a different part of the sentence. We also propose a self-attention mechanism and a special regularization term for the model. As a side effect, the embedding comes with an easy way of visualizing what specific parts of the sentence are encoded into the embedding. We evaluate our model on 3 different tasks: author profiling, sentiment classification, and textual entailment. Results show that our model yields a significant performance gain compared to other sentence embedding methods in all of the 3 tasks."
                },
                {
                    "title": "Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles",
                    "abstract": "Deep neural networks (NNs) are powerful black box predictors that have recently achieved impressive performance on a wide spectrum of tasks. Quantifying predictive uncertainty in NNs is a challenging and yet unsolved problem. Bayesian NNs, which learn a distribution over weights, are currently the state-of-the-art for estimating predictive uncertainty; however these require significant modifications to the training procedure and are computationally expensive compared to standard (non-Bayesian) NNs. We propose an alternative to Bayesian NNs that is simple to implement, readily parallelizable, requires very little hyperparameter tuning, and yields high quality predictive uncertainty estimates. Through a series of experiments on classification and regression benchmarks, we demonstrate that our method produces well-calibrated uncertainty estimates which are as good or better than approximate Bayesian NNs. To assess robustness to dataset shift, we evaluate the predictive uncertainty on test examples from known and unknown distributions, and show that our method is able to express higher uncertainty on out-of-distribution examples. We demonstrate the scalability of our method by evaluating predictive uncertainty estimates on ImageNet."
                },
                {
                    "title": "The Cityscapes Dataset for Semantic Urban Scene Understanding",
                    "abstract": "Visual understanding of complex urban street scenes is an enabling factor for a wide range of applications. Object detection has benefited enormously from large-scale datasets, especially in the context of deep learning. For semantic urban scene understanding, however, no current dataset adequately captures the complexity of real-world urban scenes. To address this, we introduce Cityscapes, a benchmark suite and large-scale dataset to train and test approaches for pixel-level and instance-level semantic labeling. Cityscapes is comprised of a large, diverse set of stereo video sequences recorded in streets from 50 different cities. 5000 of these images have high quality pixel-level annotations, 20 000 additional images have coarse annotations to enable methods that leverage large volumes of weakly-labeled data. Crucially, our effort exceeds previous attempts in terms of dataset size, annotation richness, scene variability, and complexity. Our accompanying empirical study provides an in-depth analysis of the dataset characteristics, as well as a performance evaluation of several state-of-the-art approaches based on our benchmark."
                },
                {
                    "title": "Obtaining Well Calibrated Probabilities Using Bayesian Binning",
                    "abstract": "Learning probabilistic predictive models that are well calibrated is critical for many prediction and decision-making tasks in artificial intelligence. In this paper we present a new non-parametric calibration method called Bayesian Binning into Quantiles (BBQ) which addresses key limitations of existing calibration methods. The method post processes the output of a binary classification algorithm; thus, it can be readily combined with many existing classification algorithms. The method is computationally tractable, and empirically accurate, as evidenced by the set of experiments reported here on both real and simulated datasets."
                },
                {
                    "title": "Explaining and Harnessing Adversarial Examples",
                    "abstract": "Several machine learning models, including neural networks, consistently misclassify adversarial examples---inputs formed by applying small but intentionally worst-case perturbations to examples from the dataset, such that the perturbed input results in the model outputting an incorrect answer with high confidence. Early attempts at explaining this phenomenon focused on nonlinearity and overfitting. We argue instead that the primary cause of neural networks' vulnerability to adversarial perturbation is their linear nature. This explanation is supported by new quantitative results while giving the first explanation of the most intriguing fact about them: their generalization across architectures and training sets. Moreover, this view yields a simple and fast method of generating adversarial examples. Using this approach to provide examples for adversarial training, we reduce the test set error of a maxout network on the MNIST dataset."
                },
                {
                    "title": "Weighted Sums of Random Kitchen Sinks: Replacing minimization with randomization in learning",
                    "abstract": "Randomized neural networks are immortalized in this AI Koan: \n \nIn the days when Sussman was a novice, Minsky once came to him as he sat hacking at the PDP-6. \n \n\"What are you doing?\" asked Minsky. \"I am training a randomly wired neural net to play tic-tac-toe,\" Sussman replied. \"Why is the net wired randomly?\" asked Minsky. Sussman replied, \"I do not want it to have any preconceptions of how to play.\" \n \nMinsky then shut his eyes. \"Why do you close your eyes?\" Sussman asked his teacher. \"So that the room will be empty,\" replied Minsky. At that moment, Sussman was enlightened. \n \nWe analyze shallow random networks with the help of concentration of measure inequalities. Specifically, we consider architectures that compute a weighted sum of their inputs after passing them through a bank of arbitrary randomized nonlinearities. We identify conditions under which these networks exhibit good classification performance, and bound their test error in terms of the size of the dataset and the number of random nonlinearities."
                },
                {
                    "title": "Random Features for Large-Scale Kernel Machines",
                    "abstract": "To accelerate the training of kernel machines, we propose to map the input data to a randomized low-dimensional feature space and then apply existing fast linear methods. The features are designed so that the inner products of the transformed data are approximately equal to those in the feature space of a user specified shift-invariant kernel. We explore two sets of random features, provide convergence bounds on their ability to approximate various radial basis kernels, and show that in large-scale classification and regression tasks linear machine learning algorithms applied to these features outperform state-of-the-art large-scale kernel machines."
                },
                {
                    "title": "Diverse ensembles for active learning",
                    "abstract": "Query by Committee is an effective approach to selective sampling in which disagreement amongst an ensemble of hypotheses is used to select data for labeling. Query by Bagging and Query by Boosting are two practical implementations of this approach that use Bagging and Boosting, respectively, to build the committees. For effective active learning, it is critical that the committee be made up of consistent hypotheses that are very different from each other. DECORATE is a recently developed method that directly constructs such diverse committees using artificial training data. This paper introduces ACTIVE-DECORATE, which uses DECORATE committees to select good training examples. Extensive experimental results demonstrate that, in general, ACTIVE-DECORATE outperforms both Query by Bagging and Query by Boosting."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the extrapolation capabilities of Vision Transformers (ViTs) to effectively utilize high-resolution imagery without incurring the costs associated with finetuning?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing demand for high-resolution models in computer vision, which can lead to significant advancements in model accuracy and efficiency. By enhancing extrapolation methods, we can reduce the computational costs associated with training and finetuning, making high-resolution applications more accessible. This could pave the way for practical applications in various fields, such as medical imaging, autonomous vehicles, and augmented reality, where high-resolution data is essential for performance.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent limitations of current ViT architectures, which struggle with extrapolation due to their non-hierarchical design and constant feature map size. Naive approaches may fail because they do not account for the complexities of high-resolution data, such as increased detail and variability. Technical obstacles include the need for innovative position encoding methods that can adapt to varying resolutions, as well as the theoretical understanding of how to effectively leverage patch representations for dense prediction tasks.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on improving model accuracy through finetuning rather than addressing the extrapolation challenge directly. Existing solutions often lack the necessary adaptability to high-resolution data, and there has been insufficient exploration of novel position encoding techniques. Our approach differs by emphasizing the development of models that can extrapolate effectively without the need for extensive finetuning, thereby filling a critical gap in the literature.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves training ViT models on a large-scale dataset (ImageNet) at a lower resolution (224\u00b2 px) and testing their performance at various higher resolutions (up to 1024\u00b2 px). We will implement advanced position encoding techniques to enhance extrapolation capabilities and evaluate model performance using metrics such as Top-1 and Top-5 accuracy. The expected outcomes include improved extrapolation performance, demonstrating that our models can maintain high accuracy across a range of resolutions without the need for finetuning, thus providing a cost-effective solution for high-resolution applications."
            }
        },
        "author_data": {
            "0f627395-ebf6-4099-a050-f47e474289ab": {
                "pk": "0f627395-ebf6-4099-a050-f47e474289ab",
                "name": "Anthony Fuller",
                "collaborators": [
                    "James R. Green",
                    "Koreen Millard",
                    "Daniel G. Kyrollos",
                    "Kim Greenwood",
                    "JoAnn Harrold"
                ],
                "domain": [
                    "Remote Sensing",
                    "Self-Supervised Learning",
                    "Multimodal Data",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "CROMA: Remote Sensing Representations with Contrastive Radar-Optical Masked Autoencoders",
                        "abstract": "A vital and rapidly growing application, remote sensing offers vast yet sparsely labeled, spatially aligned multimodal data; this makes self-supervised learning algorithms invaluable. We present CROMA: a framework that combines contrastive and reconstruction self-supervised objectives to learn rich unimodal and multimodal representations. Our method separately encodes masked-out multispectral optical and synthetic aperture radar samples -- aligned in space and time -- and performs cross-modal contrastive learning. Another encoder fuses these sensors, producing joint multimodal encodings that are used to predict the masked patches via a lightweight decoder. We show that these objectives are complementary when leveraged on spatially aligned multimodal data. We also introduce X- and 2D-ALiBi, which spatially biases our cross- and self-attention matrices. These strategies improve representations and allow our models to effectively extrapolate to images up to 17.6x larger at test-time. CROMA outperforms the current SoTA multispectral model, evaluated on: four classification benchmarks -- finetuning (avg. 1.8%), linear (avg. 2.4%) and nonlinear (avg. 1.4%) probing, kNN classification (avg. 3.5%), and K-means clustering (avg. 8.4%); and three segmentation benchmarks (avg. 6.4%). CROMA's rich, optionally multimodal representations can be widely leveraged across remote sensing applications."
                    },
                    {
                        "title": "Transfer Learning with Pretrained Remote Sensing Transformers",
                        "abstract": "Although the remote sensing (RS) community has begun to pretrain transformers (intended to be fine-tuned on RS tasks), it is unclear how these models perform under distribution shifts. Here, we pretrain a new RS transformer--called SatViT-V2--on 1.3 million satellite-derived RS images, then fine-tune it (along with five other models) to investigate how it performs on distributions not seen during training. We split an expertly labeled land cover dataset into 14 datasets based on source biome. We train each model on each biome separately and test them on all other biomes. In all, this amounts to 1638 biome transfer experiments. After fine-tuning, we find that SatViT-V2 outperforms SatViT-V1 by 3.1% on in-distribution (matching biomes) and 2.8% on out-of-distribution (mismatching biomes) data. Additionally, we find that initializing fine-tuning from the linear probed solution (i.e., leveraging LPFT [1]) improves SatViT-V2's performance by another 1.2% on in-distribution and 2.4% on out-of-distribution data. Next, we find that pretrained RS transformers are better calibrated under distribution shifts than non-pretrained models and leveraging LPFT results in further improvements in model calibration. Lastly, we find that five measures of distribution shift are moderately correlated with biome transfer performance. We share code and pretrained model weights. (https://github.com/antofuller/SatViT)"
                    },
                    {
                        "title": "Under the Cover Infant Pose Estimation using Multimodal Data",
                        "abstract": "Infant pose monitoring during sleep has multiple applications in both healthcare and home settings. In a healthcare setting, pose detection can be used for region of interest detection and movement detection for noncontact based monitoring systems. In a home setting, pose detection can be used to detect sleep positions which has shown to have a strong influence on multiple health factors. However, pose monitoring during sleep is challenging due to heavy occlusions from blanket coverings and low lighting. To address this, we present a novel dataset, Simultaneously-collected multimodal Mannequin Lying pose (SMaL) dataset, for under the cover infant pose estimation. We collect depth and pressure imagery of an infant mannequin in different poses under various cover conditions. We successfully infer full body pose under the cover by training state-of-art pose estimation methods and leveraging existing multimodal adult pose datasets for transfer learning. We demonstrate a hierarchical pretraining strategy for transformer-based models to significantly improve performance on our dataset. Our best performing model was able to detect joints under the cover within 25mm 86% of the time with an overall mean error of 16.9mm. Data, code and models publicly available at https://github.com/DanielKyr/SMaL"
                    }
                ]
            },
            "cb03de30-0281-4a8f-ae14-1d5b3ef6b3c0": {
                "pk": "cb03de30-0281-4a8f-ae14-1d5b3ef6b3c0",
                "name": "Daniel G. Kyrollos",
                "collaborators": [
                    "James R. Green",
                    "Anthony Fuller",
                    "Kim Greenwood",
                    "JoAnn Harrold",
                    "Kevin Dick"
                ],
                "domain": [
                    "Pose Estimation",
                    "Machine Learning",
                    "Link Prediction",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Under the Cover Infant Pose Estimation using Multimodal Data",
                        "abstract": "Infant pose monitoring during sleep has multiple applications in both healthcare and home settings. In a healthcare setting, pose detection can be used for region of interest detection and movement detection for noncontact based monitoring systems. In a home setting, pose detection can be used to detect sleep positions which has shown to have a strong influence on multiple health factors. However, pose monitoring during sleep is challenging due to heavy occlusions from blanket coverings and low lighting. To address this, we present a novel dataset, Simultaneously-collected multimodal Mannequin Lying pose (SMaL) dataset, for under the cover infant pose estimation. We collect depth and pressure imagery of an infant mannequin in different poses under various cover conditions. We successfully infer full body pose under the cover by training state-of-art pose estimation methods and leveraging existing multimodal adult pose datasets for transfer learning. We demonstrate a hierarchical pretraining strategy for transformer-based models to significantly improve performance on our dataset. Our best performing model was able to detect joints under the cover within 25mm 86% of the time with an overall mean error of 16.9mm. Data, code and models publicly available at https://github.com/DanielKyr/SMaL"
                    },
                    {
                        "title": "Generalized Reciprocal Perspective",
                        "abstract": "Across many domains, real-world problems can be represented as a network. Nodes represent domain-specific elements and edges capture the relationship between elements. Leveraging high-performance computing and optimized link prediction algorithms, it is increasingly possible to evaluate every possible combination of nodal pairs enabling the generation of a comprehensive prediction matrix (CPM) that places an individual link prediction score in the context of all possible links involving either node (providing data-driven context). Historically, this contextual information has been ignored given exponentially growing problem sizes resulting in computational intractability; however, we demonstrate that expending high-performance compute resources to generate CPMs is a worthwhile investment given the improvement in predictive performance. In this work, we generalize for all pairwise link-prediction tasks our novel semi-supervised machine learning method, denoted Reciprocal Perspective (RP). We demonstrate that RP significantly improves link prediction accuracy by leveraging the wealth of information in a CPM. Context-based features are extracted from the CPM for use in a stacked classifier and we demonstrate that the application of RP in a cascade almost always results in significantly (p < 0.05) improved predictions. These results on RS-type problems suggest that RP is applicable to a broad range of link prediction problems."
                    }
                ]
            },
            "9854b9fa-56ff-4e8f-9a96-8574f32c81e8": {
                "pk": "9854b9fa-56ff-4e8f-9a96-8574f32c81e8",
                "name": "Yousef Yassin",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "c8aa40e3-9420-4654-9337-80f17061a9a4": {
                "pk": "c8aa40e3-9420-4654-9337-80f17061a9a4",
                "name": "James R. Green",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            }
        }
    },
    "2406.03704": {
        "paper_data": {
            "title": "Excluding the Irrelevant: Focusing Reinforcement Learning through Continuous Action Masking",
            "url": "http://arxiv.org/abs/2406.03704v1",
            "arxiv_id": "2406.03704",
            "authors": [
                "Roland Stolz",
                "Hanna Krasowski",
                "Jakob Thumm",
                "Michael Eichelbeck",
                "Philipp Gassert",
                "Matthias Althoff"
            ],
            "abstract": "Continuous action spaces in reinforcement learning (RL) are commonly defined as interval sets. While intervals usually reflect the action boundaries for tasks well, they can be challenging for learning because the typically large global action space leads to frequent exploration of irrelevant actions. Yet, little task knowledge can be sufficient to identify significantly smaller state-specific sets of relevant actions. Focusing learning on these relevant actions can significantly improve training efficiency and effectiveness. In this paper, we propose to focus learning on the set of relevant actions and introduce three continuous action masking methods for exactly mapping the action space to the state-dependent set of relevant actions. Thus, our methods ensure that only relevant actions are executed, enhancing the predictability of the RL agent and enabling its use in safety-critical applications. We further derive the implications of the proposed methods on the policy gradient. Using Proximal Policy Optimization (PPO), we evaluate our methods on three control tasks, where the relevant action set is computed based on the system dynamics and a relevant state set. Our experiments show that the three action masking methods achieve higher final rewards and converge faster than the baseline without action masking.",
            "introduction": "   1 Introduction  \\Ac RL can solve complex tasks in robotics [Han2023-roboticRLsurvey], games [shao2019survey], and is used for fine-tuning large language models [ouyang2022training]. Yet, training  reinforcement learning (RL) agents is often sample-inefficient due to frequent exploration of actions, which are irrelevant to learning a good policy. Examples for irrelevant actions are driving into a wall for autonomous cars, or tipping a drone off-balance. The global action space is typically large in relation to the relevant actions in each state. Therefore, exploring these actions frequently can introduce unnecessary costs, lead to slow convergence, or even prevent the agent from learning a suitable policy.   Action masking mitigates this problem by constraining the exploration to the set of relevant state-specific actions, which can be obtained based on task knowledge. For example, when there is no opponent within reach in video games, attack actions are masked from the action space [Ye2020-MOBAmasking]. Leveraging task knowledge through action masking usually leads to faster convergence and also increases the predictability of the RL agent as the set of relevant actions can provide additional meaning. For instance, if the set of relevant actions is a set of verified safe actions, action masking can be used to provide safety guarantees [Fulton2018, Krasowski2023b.ProvablySafeRLSurvey].   Whenever the relevant action set is easy to compute, action masking can be seamlessly integrated into RL algorithms and is the quasi-standard for discrete action spaces [Zhong_Yang_Zhao_2024, Shakya2023], e.g., in motion planning [Mirchevska2018, Krasowski2020, Xiaofei2022-masking, Rudolf2022, Feng2023], games [Huang2022, Ye2020-MOBAmasking, Hou2023-maskinggames], and power systems [Tabas2022, Lingyu2023]. However, real-world systems operate in continuous space and discretization to a higher-level decision space might prevent learning optimal policies. Furthermore, simulation of real-world systems is compute-intensive, and developing RL agents for them often requires additional real-world training [zhao_sim-to-real_2020]. Thus, sample efficiency is particularly valuable for these applications.   In this work, we propose three action masking methods for continuous action spaces. They employ expressive convex set representations, i.e., polytopes or zonotopes for the relevant action set. Our action masking methods generalize previous work in [Krasowski2023b.ProvablySafeRLSurvey], which is constrained to intervals as relevant action sets. This extends the applicability of continuous action masking to expressive convex relevant action sets, which is especially useful when action dimensions are coupled, e.g., for a thrust-controlled quadrotor. To integrate the relevant action set into RL, we introduce three methods: the generator mask, which exploits the generator representation of a zonotope, the ray mask, which projects an action into the relevant action set based on radial directions, and the distributional mask, which truncates the policy distribution to the relevant action set. In summary our main contributions are:     \u2022  We introduce continuous action masking based on arbitrary convex sets representing the state-dependent relevant action sets;    \u2022  We present three methods to utilize the relevant action sets and derive their integration in the backward pass of RL with stochastic policies;    \u2022  We evaluate our approach on three benchmark environments that demonstrate the application conditions of our continuous action masking approaches.       1.1 Related literature  Action masking is mainly applied for discrete action spaces [Fulton2018, Mirchevska2018, Krasowski2020, Xiaofei2022-masking, Rudolf2022, Feng2023, Huang2022, Ye2020-MOBAmasking, Hou2023-maskinggames, Tabas2022, Lingyu2023, Krasowski2023b.ProvablySafeRLSurvey, Zhong_Yang_Zhao_2024, varricchione2024a-safemaskingLTL]. Huang et al. [Huang2022] derive the implications on policy gradient RL",
            "references": [
                {
                    "title": "Pure-Past Action Masking",
                    "abstract": "We present Pure-Past Action Masking (PPAM), a lightweight approach to action masking for safe reinforcement learning. In PPAM, actions are disallowed (\u201cmasked\u201d) according to specifications expressed in Pure-Past Linear Temporal Logic (PPLTL). PPAM can enforce non-Markovian constraints, i.e., constraints based on the history of the system, rather than just the current state of the (possibly hidden) MDP. The features used in the safety constraint need not be the same as those used by the learning agent, allowing a clear separation of concerns between the safety constraints and reward specifications of the (learning) agent. We prove formally that an agent trained with PPAM can learn any optimal policy that satisfies the safety constraints, and that they are as expressive as shields, another approach to enforce non-Markovian constraints in RL. Finally, we provide empirical results showing how PPAM can guarantee constraint satisfaction in practice."
                },
                {
                    "title": "Scalable Computation of Robust Control Invariant Sets of Nonlinear Systems",
                    "abstract": "Ensuring robust constraint satisfaction for an infinite-time horizon is a challenging, yet crucial task when deploying safety-critical systems. In this article, we address this issue by synthesizing robust control invariant sets of perturbed nonlinear sampled-data systems. This task can be encoded as a nonconvex program that we approximate by a tailored, computationally efficient successive convexification algorithm. Based on the zonotopic representation of invariant sets, we obtain an updated candidate for the invariant set and the invariance-enforcing controller by solving a single convex program. To obtain a possibly large region of safe operation, our algorithm is designed so that the sequence of candidate invariant sets is volume-wise monotonically increasing. We demonstrate the efficacy and scalability of our approach using a broad range of nonlinear control systems from the literature with up to 20 dimensions."
                },
                {
                    "title": "No Prior Mask: Eliminate Redundant Action for Deep Reinforcement Learning",
                    "abstract": "The large action space is one fundamental obstacle to deploying Reinforcement Learning methods in the real world. The numerous redundant actions will cause the agents to make repeated or invalid attempts, even leading to task failure. Although current algorithms conduct some initial explorations for this issue, they either suffer from rule-based systems or depend on expert demonstrations, which significantly limits their applicability in many real-world settings. In this work, we examine the theoretical analysis of what action can be eliminated in policy optimization and propose a novel redundant action filtering mechanism. Unlike other works, our method constructs the similarity factor by estimating the distance between the state distributions, which requires no prior knowledge. In addition, we combine the modified inverse model to avoid extensive computation in high-dimensional state space. We reveal the underlying structure of action spaces and propose a simple yet efficient redundant action filtering mechanism named No Prior Mask (NPM) based on the above techniques. We show the superior performance of our method by conducting extensive experiments on high-dimensional, pixel-input, and stochastic problems with various action redundancy tasks. Our code is public online at https://github.com/zhongdy15/npm."
                },
                {
                    "title": "Enhancement of Distribution Network Resilience: A Multi-Buffer Invalid-Action-Mask Double Q-Network Approach for Distribution Network Restoration",
                    "abstract": "The increasing occurrence of extreme weather events presents significant challenges to the resilience of distribution networks. Distribution network restoration (DSR), typically solved by using optimization-based methods, suffers from drawbacks such as reliance on accurate models and slow solution convergence, leading to low scalability. To address these issues, this paper proposes the multi-buffer invalid-action-mask double Q-network (MI-DDQN) approach for DSR. Firstly, the DSR problem is formulated as a Markov decision process. Next, the efficiency of experience replay is enhanced through the utilization of multi-buffer and prioritized experience replay. Additionally, an invalid-action mask is employed to mask out invalid actions, thereby improving model convergence speed. The effectiveness of the proposed method is validated by using the IEEE 37-node system."
                },
                {
                    "title": "Autonomous Decision Making with Reinforcement Learning in Multi-UAV Air Combat",
                    "abstract": "A multi-agent decision network based on QMIX is proposed in this paper to cope with the coordination decision problem of multiple UAV air combat missions. To speed up the training process, three improvements are introduced: 1) An improved $\\epsilon$-decaying method that enable some tutor to help in action selection at the early stage of the training. This measure greatly improves the exploring efficiency when the network are far from being fully trained; 2) State pruning and action mask measures are applied during the training. The former improves the effectiveness of the input state information, and the latter reduces unnecessary action exploring. 3) A gradually training configuration is used to make the training process more robust, where the combat adversaries are configured as the static targets, the randomly maneuver vehicles, and the Min-Max strategy vehicles respectively. The multi-UAV air combat scenarios are built up and the experiments are conducted. The results shows that these improvements have significantly improved training efficiency."
                },
                {
                    "title": "Exploring the Use of Invalid Action Masking in Reinforcement Learning: A Comparative Study of On-Policy and Off-Policy Algorithms in Real-Time Strategy Games",
                    "abstract": "Invalid action masking is a practical technique in deep reinforcement learning to prevent agents from taking invalid actions. Existing approaches rely on action masking during policy training and utilization. This study focuses on developing reinforcement learning algorithms that incorporate action masking during training but can be used without action masking during policy execution. The study begins by conducting a theoretical analysis to elucidate the distinction between naive policy gradient and invalid action policy gradient. Based on this analysis, we demonstrate that the naive policy gradient is a valid gradient and is equivalent to the proposed composite objective algorithm, which optimizes both the masked policy and the original policy in parallel. Moreover, we propose an off-policy algorithm for invalid action masking that employs the masked policy for sampling while optimizing the original policy. To compare the effectiveness of these algorithms, experiments are conducted using a simplified real-time strategy (RTS) game simulator called Gym-\u03bcRTS. Based on empirical findings, we recommend utilizing the off-policy algorithm for addressing most tasks while employing the composite objective algorithm for handling more complex tasks."
                },
                {
                    "title": "A Survey on Deep Reinforcement Learning Algorithms for Robotic Manipulation",
                    "abstract": "Robotic manipulation challenges, such as grasping and object manipulation, have been tackled successfully with the help of deep reinforcement learning systems. We give an overview of the recent advances in deep reinforcement learning algorithms for robotic manipulation tasks in this review. We begin by outlining the fundamental ideas of reinforcement learning and the parts of a reinforcement learning system. The many deep reinforcement learning algorithms, such as value-based methods, policy-based methods, and actor\u2013critic approaches, that have been suggested for robotic manipulation tasks are then covered. We also examine the numerous issues that have arisen when applying these algorithms to robotics tasks, as well as the various solutions that have been put forth to deal with these issues. Finally, we highlight several unsolved research issues and talk about possible future directions for the subject."
                },
                {
                    "title": "Provably Safe Reinforcement Learning: Conceptual Analysis, Survey, and Benchmarking",
                    "abstract": "Ensuring the safety of reinforcement learning (RL) algorithms is crucial to unlock their potential for many real-world tasks. However, vanilla RL and most safe RL approaches do not guarantee safety. In recent years, several methods have been proposed to provide hard safety guarantees for RL, which is essential for applications where unsafe actions could have disastrous consequences. Nevertheless, there is no comprehensive comparison of these provably safe RL methods. Therefore, we introduce a categorization of existing provably safe RL methods, present the conceptual foundations for both continuous and discrete action spaces, and empirically benchmark existing methods. We categorize the methods based on how they adapt the action: action replacement, action projection, and action masking. Our experiments on an inverted pendulum and a quadrotor stabilization task indicate that action replacement is the best-performing approach for these applications despite its comparatively simple realization. Furthermore, adding a reward penalty, every time the safety verification is engaged, improved training performance in our experiments. Finally, we provide practical guidance on selecting provably safe RL approaches depending on the safety specification, RL algorithm, and type of action space."
                },
                {
                    "title": "Fuzzy Action-Masked Reinforcement Learning Behavior Planning for Highly Automated Driving",
                    "abstract": "Highly-automated driving relies on informed decision making and safe control in a dynamic environment which is regulated by traffic rules. This complex environment can be formulated as a map-based problem where the driving task is constrained by kinematics, obstacles, traffic, and road structure. Though, solving the problem with a monolithic approach additionally can lead to infeasible conditions which are difficult to bypass. Therefore, we expect a separate layer for sequential decision making to be a more suitable approach to derive a lateral and longitudinal behavior. In this area of decision making, reinforcement learning (RL) is a viable approach to complex and nonlinear problems. However, the learning convergence to a reasonable behavior in a critical and constrained environment can be slow. To overcome this challenge, we propose an action-masked RL agent utilizing fuzzy traffic rule descriptions. The agent outputs are hybrid lateral and longitudinal actions based on the environment observations. These driving decisions interface the subsequent trajectory planner and control for optimal execution. In exemplary merging scenarios, we show the effectiveness of the masked agent to increase convergence speed toward a reasonable behavior."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Computationally Efficient Safe Reinforcement Learning for Power Systems",
                    "abstract": "We propose a computationally efficient approach to safe reinforcement learning (RL) for frequency regulation in power systems with high levels of variable renewable energy resources. The approach draws on set-theoretic control techniques to craft a neural network-based control policy that is guaranteed to satisfy safety-critical state constraints, without needing to solve a model predictive control or projection problem in real time. By exploiting the properties of robust controlled-invariant polytopes, we construct a novel, closed-form \"safety-filter\" that enables end-to-end safe learning using any policy gradient-based RL algorithm. We then apply the safety filter in conjunction with the deep deterministic policy gradient (DDPG) algorithm to regulate frequency in a modified 9-bus power system, and show that the learned policy is more cost-effective than robust linear feedback control techniques while maintaining the same safety guarantee. We also show that the proposed paradigm outperforms DDPG augmented with constraint violation penalties."
                },
                {
                    "title": "Sim-to-Real Transfer in Deep Reinforcement Learning for Robotics: a Survey",
                    "abstract": "Deep} reinforcement learning has recently seen huge success across multiple areas in the robotics domain. Owing to the limitations of gathering real-world data, i.e., sample inefficiency and the cost of collecting it, simulation environments are utilized for training the different agents. This not only aids in providing a potentially infinite data source, but also alleviates safety concerns with real robots. Nonetheless, the gap between the simulated and real worlds degrades the performance of the policies once the models are transferred into real robots. Multiple research efforts are therefore now being directed towards closing this sim-toreal gap and accomplish more efficient policy transfer. Recent years have seen the emergence of multiple methods applicable to different domains, but there is a lack, to the best of our knowledge, of a comprehensive review summarizing and putting into context the different methods. In this survey paper, we cover the fundamental background behind sim-to-real transfer in deep reinforcement learning and overview the main methods being utilized at the moment: domain randomization, domain adaptation, imitation learning, meta-learning and knowledge distillation. We categorize some of the most relevant recent works, and outline the main application scenarios. Finally, we discuss the main opportunities and challenges of the different approaches and point to the most promising directions."
                },
                {
                    "title": "Safe Reinforcement Learning for Autonomous Lane Changing Using Set-Based Prediction",
                    "abstract": "Machine learning approaches often lack safety guarantees, which are often a key requirement in real-world tasks. This paper addresses the lack of safety guarantees by extending reinforcement learning with a safety layer that restricts the action space to the subspace of safe actions. We demonstrate the proposed approach using lane changing in autonomous driving. To distinguish safe actions from unsafe ones, we compare planned motions with the set of possible occupancies of traffic participants generated by set-based predictions. In situations where no safe action exists, a verified failsafe controller is executed. We used real-world highway traffic data to train and test the proposed approach. The evaluation result shows that the proposed approach trains agents that do not cause collisions during training and deployment."
                },
                {
                    "title": "A Closer Look at Invalid Action Masking in Policy Gradient Algorithms",
                    "abstract": "In recent years, Deep Reinforcement Learning (DRL) algorithms have achieved state-of-the-art performance in many challenging strategy games. Because these games have complicated rules, an action sampled from the full discrete action distribution predicted by the learned policy is likely to be invalid according to the game rules (e.g., walking into a wall). The usual approach to deal with this problem in policy gradient algorithms is to \u201cmask out\u201d invalid actions and just sample from the set of valid actions. The implications of this process, however, remain under-investigated. In this paper, we 1) show theoretical justification for such a practice, 2) empirically demonstrate its importance as the space of invalid actions grows, and 3) provide further insights by evaluating different action masking regimes, such as removing masking after an agent has been trained using masking."
                },
                {
                    "title": "A Survey of Deep Reinforcement Learning in Video Games",
                    "abstract": "Deep reinforcement learning (DRL) has made great achievements since proposed. Generally, DRL agents receive high-dimensional inputs at each step, and make actions according to deep-neural-network-based policies. This learning mechanism updates the policy to maximize the return with an end-to-end method. In this paper, we survey the progress of DRL methods, including value-based, policy gradient, and model-based algorithms, and compare their main techniques and properties. Besides, DRL plays an important role in game artificial intelligence (AI). We also take a review of the achievements of DRL in various video games, including classical Arcade games, first-person perspective games and multi-agent real-time strategy games, from 2D to 3D, and from single-agent to multi-agent. A large number of video game AIs with DRL have achieved super-human performance, while there are still some challenges in this domain. Therefore, we also discuss some key points when applying DRL methods to this field, including exploration-exploitation, sample efficiency, generalization and transfer, multi-agent learning, imperfect information, and delayed spare rewards, as well as some research directions."
                },
                {
                    "title": "Mastering Complex Control in MOBA Games with Deep Reinforcement Learning",
                    "abstract": "We study the reinforcement learning problem of complex action control in the Multi-player Online Battle Arena (MOBA) 1v1 games. This problem involves far more complicated state and action spaces than those of traditional 1v1 games, such as Go and Atari series, which makes it very difficult to search any policies with human-level performance. In this paper, we present a deep reinforcement learning framework to tackle this problem from the perspectives of both system and algorithm. Our system is of low coupling and high scalability, which enables efficient explorations at large scale. Our algorithm includes several novel strategies, including control dependency decoupling, action mask, target attention, and dual-clip PPO, with which our proposed actor-critic network can be effectively trained in our system. Tested on the MOBA game Honor of Kings, the trained AI agents can defeat top professional human players in full 1v1 games."
                },
                {
                    "title": "Optuna: A Next-generation Hyperparameter Optimization Framework",
                    "abstract": "The purpose of this study is to introduce new design-criteria for next-generation hyperparameter optimization software. The criteria we propose include (1) define-by-run API that allows users to construct the parameter search space dynamically, (2) efficient implementation of both searching and pruning strategies, and (3) easy-to-setup, versatile architecture that can be deployed for various purposes, ranging from scalable distributed computing to light-weight experiment conducted via interactive interface. In order to prove our point, we will introduce Optuna, an optimization software which is a culmination of our effort in the development of a next generation optimization software. As an optimization software designed with define-by-run principle, Optuna is particularly the first of its kind. We will present the design-techniques that became necessary in the development of the software that meets the above criteria, and demonstrate the power of our new design through experimental results and real world applications. Our software is available under the MIT license (https://github.com/pfnet/optuna/)."
                },
                {
                    "title": "Linear Encodings for Polytope Containment Problems",
                    "abstract": "The polytope containment problem is deciding whether a polytope is a contained within another polytope. The complexity heavily depends on how the polytopes are represented. While there exists efficient necessary and sufficient conditions for polytope containment when their hyperplanes are available (H-polytopes), the case when polytopes are represented by affine transformations of H-polytopes, which we refer to as AH-polytopes, is known to be co-NP-complete. In this paper, we provide a sufficient condition for AH-polytope in AHpolytope problem that can be cast as a linear set of constraints with size that grows linearly with the number of hyperplanes of each polytope. These efficient encodings enable us to designate certain components of polytopes as decision variables, and incorporate them into a convex optimization problem. We present the usefulness of our results on applications to the zonotope containment problem, computing polytopic Hausdorff distances, and finding inner approximations to orthogonal projections of polytopes. Illustrative examples are included."
                },
                {
                    "title": "High-level Decision Making for Safe and Reasonable Autonomous Lane Changing using Reinforcement Learning",
                    "abstract": "Machine learning techniques have been shown to outperform many rule-based systems for the decision-making of autonomous vehicles. However, applying machine learning is challenging due to the possibility of executing unsafe actions and slow learning rates. We address these issues by presenting a reinforcement learning-based approach, which is combined with formal safety verification to ensure that only safe actions are chosen at any time. We let a deep reinforcement learning (RL) agent learn to drive as close as possible to a desired velocity by executing reasonable lane changes on simulated highways with an arbitrary number of lanes. By making use of a minimal state representation, consisting of only 13 continuous features, and a Deep Q-Network (DQN), we are able to achieve fast learning rates. Our RL agent is able to learn the desired task without causing collisions and outperforms a complex, rule-based agent that we use for benchmarking."
                },
                {
                    "title": "Safe Reinforcement Learning via Formal Methods: Toward Safe Control Through Proof and Learning",
                    "abstract": "\n \n Formal verification provides a high degree of confidence in safe system operation, but only if reality matches the verified model. Although a good model will be accurate most of the time, even the best models are incomplete. This is especially true in Cyber-Physical Systems because high-fidelity physical models of systems are expensive to develop and often intractable to verify. Conversely, reinforcement learning-based controllers are lauded for their flexibility in unmodeled environments, but do not provide guarantees of safe operation. This paper presents an approach for provably safe learning that provides the best of both worlds: the exploration and optimization capabilities of learning along with the safety guarantees of formal verification. Our main insight is that formal verification combined with verified runtime monitoring can ensure the safety of a learning agent. Verification results are preserved whenever learning agents limit exploration within the confounds of verified control choices as long as observed reality comports with the model used for off-line verification. When a model violation is detected, the agent abandons efficiency and instead attempts to learn a control strategy that guides the agent to a modeled portion of the state space. We prove that our approach toward incorporating knowledge about safe control into learning systems preserves safety guarantees, and demonstrate that we retain the empirical performance benefits provided by reinforcement learning. We also explore various points in the design space for these justified speculative controllers in a simple model of adaptive cruise control model for autonomous cars.\n \n"
                },
                {
                    "title": "Addressing Function Approximation Error in Actor-Critic Methods",
                    "abstract": "In value-based reinforcement learning methods such as deep Q-learning, function approximation errors are known to lead to overestimated value estimates and suboptimal policies. We show that this problem persists in an actor-critic setting and propose novel mechanisms to minimize its effects on both the actor and the critic. Our algorithm builds on Double Q-learning, by taking the minimum value between a pair of critics to limit overestimation. We draw the connection between target networks and overestimation bias, and suggest delaying policy updates to reduce per-update error and further improve performance. We evaluate our method on the suite of OpenAI gym tasks, outperforming the state of the art in every environment tested."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "Scalable Safety-Preserving Robust Control Synthesis for Continuous-Time Linear Systems",
                    "abstract": "We present a scalable set-valued safety-preserving hybrid controller for constrained continuous-time linear time-invariant (LTI) systems subject to additive disturbance/uncertainty. The approach relies on a conservative approximation of the discriminating kernel using a piecewise ellipsoidal algorithm with polynomial complexity. This precomputed approximation is used online to synthesize a permissive state-feedback control law that guarantees the satisfaction of all constraints despite potentially conflicting performance objectives. We show the results on a flight envelope protection problem for a quadrotor with actuation saturation and unknown wind disturbances."
                },
                {
                    "title": "Hit-and-run from a corner",
                    "abstract": "We show that the hit-and-run random walk mixes rapidly starting from any interior point of a convex body. This is the first random walk known to have this property. In contrast, the ball walk can take exponentially many steps from some starting points."
                },
                {
                    "title": "An adaptive numerical cubature algorithm for simplices",
                    "abstract": "A globally adaptive algorithm for numerical cubature of a vector of functions over a collection of n-dimensional simplices is described. The algorithm is based on a subdivision strategy that chooses for subdivision at each stage the subregion (of the input simplices) with the largest estimated error. This subregion is divided into two, three or four equal volume subregions by cutting selected edges. These edges are selected using information about the smoothness of the integrands in the edge directions. The algorithm allows a choice from several embedded cubature rule sequences for approximate integration and error estimation. A Fortran 95 implementation as a part of CUBPACK is also discussed. Testing of the algorithm is described."
                },
                {
                    "title": "Policy Gradient Methods for Reinforcement Learning with Function Approximation",
                    "abstract": "Function approximation is essential to reinforcement learning, but the standard approach of approximating a value function and determining a policy from it has so far proven theoretically intractable. In this paper we explore an alternative approach in which the policy is explicitly represented by its own function approximator, independent of the value function, and is updated according to the gradient of expected reward with respect to the policy parameters. Williams's REINFORCE method and actor-critic methods are examples of this approach. Our main new result is to show that the gradient can be written in a form suitable for estimation from experience aided by an approximate action-value or advantage function. Using this result, we prove for the first time that a version of policy iteration with arbitrary differentiable function approximation is convergent to a locally optimal policy."
                },
                {
                    "title": "Stable-Baselines3: Reliable Reinforcement Learning Implementations",
                    "abstract": "Stable-Baselines3 provides open-source implementations of deep reinforcement learning (RL) algorithms in Python. The implementations have been benchmarked against reference codebases, and automated unit tests cover 95% of the code. The algorithms follow a consistent interface and are accompanied by extensive documentation, making it simple to train and compare di\ufb00erent RL algorithms. Our documentation, examples, and source-code are available at https://github.com/DLR-RM/stable-baselines3 ."
                },
                {
                    "title": "Reachability Analysis and its Application to the Safety Assessment of Autonomous Cars",
                    "abstract": "This thesis is about the safety verification of dynamical systems using reachability analysis. Novel solutions have been developed for classical reachability analysis, stochastic reachability analysis, and their application to the safety assessment of autonomous cars. \n \nClassical reachability analysis computes the set of states that can be reached by a system. If the reachable set does not intersect any set of unsafe states, the safety of the system is guaranteed. Algorithms for this problem have been developed for linear, nonlinear, and hybrid systems. \n \nStochastic reachability analysis measures the probability of reaching an unsafe set. One pursued approach computes over-approximative solutions for linear systems; another one generates a Markov chain which approximately computes the stochastic reachable set of arbitrary dynamics."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.SY",
                "eess.SY"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively implement action masking in continuous action spaces to improve sample efficiency and policy learning in reinforcement learning agents?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of reinforcement learning, particularly in real-world applications where continuous action spaces are prevalent, such as robotics and autonomous systems. By improving sample efficiency through effective action masking, we can accelerate the training process, reduce computational costs, and enhance the safety and reliability of RL agents. This research could lead to significant advancements in how RL is applied in complex environments, paving the way for more robust and efficient learning algorithms that can be deployed in practical scenarios.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of continuous action spaces, where the relevant actions are not easily identifiable and can be highly interdependent. Naive approaches that apply discrete action masking techniques may fail to capture the nuances of continuous actions, leading to suboptimal policy learning. Additionally, the need for expressive convex set representations to define relevant action sets introduces theoretical and computational obstacles, as well as the difficulty of integrating these representations into existing RL frameworks without compromising the learning process.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on action masking in discrete action spaces, leaving a gap in the literature regarding continuous action spaces. The limitations of existing methods, which often rely on simple interval constraints, have hindered the development of more sophisticated approaches. Barriers such as the lack of suitable mathematical frameworks for representing complex action sets and the computational intensity of simulating real-world systems have also contributed to the problem remaining unsolved. Our approach differs by introducing three novel methods that utilize arbitrary convex sets, thereby expanding the applicability of action masking in continuous domains.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology includes three action masking methods for continuous action spaces: the generator mask, which utilizes the generator representation of a zonotope; the ray mask, which projects actions into the relevant action set based on radial directions; and the distributional mask, which truncates the policy distribution to the relevant action set. We will evaluate these methods using benchmark environments to demonstrate their effectiveness and applicability. The expected outcomes include improved sample efficiency, faster convergence of RL agents, and enhanced safety guarantees in policy learning, ultimately contributing to the advancement of reinforcement learning techniques in"
            }
        },
        "author_data": {
            "91cd4f69-eaec-4be1-af80-f1e618474e8a": {
                "pk": "91cd4f69-eaec-4be1-af80-f1e618474e8a",
                "name": "Roland Stolz",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "cebe6564-5036-46b8-852a-97eb02f7aa0e": {
                "pk": "cebe6564-5036-46b8-852a-97eb02f7aa0e",
                "name": "Hanna Krasowski",
                "collaborators": [
                    "Matthias Althoff",
                    "Xiao Wang",
                    "Marius Wiggert",
                    "Manan Doshi",
                    "Pierre F. J. Lermusiaux",
                    "Claire J. Tomlin",
                    "Prithvi Akella",
                    "Aaron D. Ames",
                    "Niklas Kochdumper",
                    "Stanley Bak"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Safety Assurance",
                    "Autonomous Systems",
                    "Control Theory"
                ],
                "publications": [
                    {
                        "title": "Provable Traffic Rule Compliance in Safe Reinforcement Learning on the Open Sea",
                        "abstract": "For safe operation, autonomous vehicles have to obey traffic rules that are set forth in legal documents formulated in natural language. Temporal logic is a suitable concept to formalize such traffic rules. Still, temporal logic rules often result in constraints that are hard to solve using optimization-based motion planners. Reinforcement learning (RL) is a promising method to find motion plans for autonomous vehicles. However, vanilla RL algorithms are based on random exploration and do not automatically comply with traffic rules. Our approach accomplishes guaranteed rule-compliance by integrating temporal logic specifications into RL. Specifically, we consider the application of vessels on the open sea, which must adhere to the Convention on the International Regulations for Preventing Collisions at Sea (COLREGS). To efficiently synthesize rule-compliant actions, we combine predicates based on set-based prediction with a statechart representing our formalized rules and their priorities. Action masking then restricts the RL agent to this set of verified rule-compliant actions. In numerical evaluations on critical maritime traffic situations, our agent always complies with the formalized legal rules and never collides while achieving a high goal-reaching rate during training and deployment. In contrast, vanilla and traffic rule-informed RL agents frequently violate traffic rules and collide even after training."
                    },
                    {
                        "title": "Safe Reinforcement Learning with Probabilistic Guarantees Satisfying Temporal Logic Specifications in Continuous Action Spaces",
                        "abstract": "Vanilla Reinforcement Learning (RL) can efficiently solve complex tasks but does not provide any guarantees on system behavior. To bridge this gap, we propose a three-step safe RL procedure for continuous action spaces that provides probabilistic guarantees with respect to temporal logic specifications. First, our approach probabilistically verifies a candidate controller with respect to a temporal logic specification while randomizing the control inputs to the system within a bounded set. Second, we improve the performance of this probabilistically verified controller by adding an RL agent that optimizes the verified controller for performance in the same bounded set around the control input. Third, we verify probabilistic safety guarantees with respect to temporal logic specifications for the learned agent. Our approach is efficiently implementable for continuous action and state spaces. The separation of safety verification and performance improvement into two distinct steps realizes both explicit probabilistic safety guarantees and a straightforward RL setup that focuses on performance. We evaluate our approach on an evasion task where a robot has to reach a goal while evading a dynamic obstacle with a specific maneuver. Our results show that our safe RL approach leads to efficient learning while maintaining its probabilistic safety specification."
                    },
                    {
                        "title": "Provably Safe Reinforcement Learning via Action Projection using Reachability Analysis and Polynomial Zonotopes",
                        "abstract": "While reinforcement learning produces very promising results for many applications, its main disadvantage is the lack of safety guarantees, which prevents its use in safety-critical systems. In this work, we address this issue by a safety shield for nonlinear continuous systems that solve reach-avoid tasks. Our safety shield prevents applying potentially unsafe actions from a reinforcement learning agent by projecting the proposed action to the closest safe action. This approach is called action projection and is implemented via mixed-integer optimization. The safety constraints for action projection are obtained by applying parameterized reachability analysis using polynomial zonotopes, which enables to accurately capture the nonlinear effects of the actions on the system. In contrast to other state-of-the-art approaches for action projection, our safety shield can efficiently handle input constraints and dynamic obstacles, eases incorporation of the spatial robot dimensions into the safety constraints, guarantees robust safety despite process noise and measurement errors, and is well suited for high-dimensional systems, as we demonstrate on several challenging benchmark systems."
                    },
                    {
                        "title": "Provably Safe Reinforcement Learning: Conceptual Analysis, Survey, and Benchmarking",
                        "abstract": "Ensuring the safety of reinforcement learning (RL) algorithms is crucial to unlock their potential for many real-world tasks. However, vanilla RL and most safe RL approaches do not guarantee safety. In recent years, several methods have been proposed to provide hard safety guarantees for RL, which is essential for applications where unsafe actions could have disastrous consequences. Nevertheless, there is no comprehensive comparison of these provably safe RL methods. Therefore, we introduce a categorization of existing provably safe RL methods, present the conceptual foundations for both continuous and discrete action spaces, and empirically benchmark existing methods. We categorize the methods based on how they adapt the action: action replacement, action projection, and action masking. Our experiments on an inverted pendulum and a quadrotor stabilization task indicate that action replacement is the best-performing approach for these applications despite its comparatively simple realization. Furthermore, adding a reward penalty, every time the safety verification is engaged, improved training performance in our experiments. Finally, we provide practical guidance on selecting provably safe RL approaches depending on the safety specification, RL algorithm, and type of action space."
                    },
                    {
                        "title": "Maximizing Seaweed Growth on Autonomous Farms: A Dynamic Programming Approach for Underactuated Systems Navigating on Uncertain Ocean Currents",
                        "abstract": "Seaweed biomass offers significant potential for climate mitigation, but large-scale, autonomous open-ocean farms are required to fully exploit it. Such farms typically have low propulsion and are heavily influenced by ocean currents. We want to design a controller that maximizes seaweed growth over months by taking advantage of the non-linear time-varying ocean currents for reaching high-growth regions. The complex dynamics and underactuation make this challenging even when the currents are known. This is even harder when only short-term imperfect forecasts with increasing uncertainty are available. We propose a dynamic programming-based method to efficiently solve for the optimal growth value function when true currents are known. We additionally present three extensions when as in reality only forecasts are known: (1) our methods resulting value function can be used as feedback policy to obtain the growth-optimal control for all states and times, allowing closed-loop control equivalent to re-planning at every time step hence mitigating forecast errors, (2) a feedback policy for long-term optimal growth beyond forecast horizons using seasonal average current data as terminal reward, and (3) a discounted finite-time Dynamic Programming (DP) formulation to account for increasing ocean current estimate uncertainty. We evaluate our approach through 30-day simulations of floating seaweed farms in realistic Pacific Ocean current scenarios. Our method demonstrates an achievement of 95.8% of the best possible growth using only 5-day forecasts. This confirms the feasibility of using low-power propulsion and optimal control for enhanced seaweed growth on floating farms under real-world conditions."
                    },
                    {
                        "title": "Stranding Risk for Underactuated Vessels in Complex Ocean Currents: Analysis and Controllers",
                        "abstract": "Low-propulsion vessels can take advantage of powerful ocean currents to navigate towards a destination. Recent results demonstrated that vessels can reach their destination with high probability despite forecast errors. However, these results do not consider the critical aspect of safety of such vessels: because of their low propulsion which is much smaller than the magnitude of currents, they might end up in currents that inevitably push them into unsafe areas such as shallow areas, garbage patches, and shipping lanes. In this work, we first investigate the risk of stranding for free-floating vessels in the Northeast Pacific. We find that at least 5.04% would strand within 90 days. Next, we encode the unsafe sets as hard constraints into Hamilton-Jacobi Multi-Time Reachability (HJ-MTR) to synthesize a feedback policy that is equivalent to re-planning at each time step at low computational cost. While applying this policy closed-loop guarantees safe operation when the currents are known, in realistic situations only imperfect forecasts are available. We demonstrate the safety of our approach in such realistic situations empirically with large-scale simulations of a vessel navigating in high-risk regions in the Northeast Pacific. We find that applying our policy closed-loop with daily re-planning on new forecasts can ensure safety with high probability even under forecast errors that exceed the maximal propulsion. Our method significantly improves safety over the baselines and still achieves a timely arrival of the vessel at the destination."
                    }
                ]
            },
            "80534e26-4528-4ceb-800f-197d54f93ff6": {
                "pk": "80534e26-4528-4ceb-800f-197d54f93ff6",
                "name": "Jakob Thumm",
                "collaborators": [
                    "Matthias Althoff",
                    "Guillaume Pelat",
                    "Felix Trost",
                    "Christopher Agia",
                    "Marco Pavone",
                    "Hanna Krasowski",
                    "Marlon M\u00fcller",
                    "Lukas Sch\u00e4fer",
                    "Xiao Wang"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Robotics",
                    "Human-Robot Interaction",
                    "Safety Assurance"
                ],
                "publications": [
                    {
                        "title": "Provably Safe Deep Reinforcement Learning for Robotic Manipulation in Human Environments",
                        "abstract": "Deep reinforcement learning (RL) has shown promising results in the motion planning of manipulators. However, no method guarantees the safety of highly dynamic obstacles, such as humans, in RL-based manipulator control. This lack of formal safety assurances prevents the application of RL for manipulators in real-world human environments. Therefore, we propose a shielding mechanism that ensures ISO-verified human safety while training and deploying RL algorithms on manipulators. We utilize a fast reachability analysis of humans and manipulators to guarantee that the manipulator comes to a complete stop before a human is within its range. Our proposed method guarantees safety and significantly improves the RL performance by preventing episode-ending collisions. We demonstrate the performance of our proposed method in simulation using human motion capture data."
                    },
                    {
                        "title": "Reducing Safety Interventions in Provably Safe Reinforcement Learning",
                        "abstract": "Deep Reinforcement Learning (RL) has shown promise in addressing complex robotic challenges. In real-world applications, RL is often accompanied by failsafe controllers as a last resort to avoid catastrophic events. While necessary for safety, these interventions can result in undesirable behaviors, such as abrupt braking or aggressive steering. This paper proposes two safety intervention reduction methods: proactive replacement and proactive projection, which change the action of the agent if it leads to a potential failsafe intervention. These approaches are compared to state-of-the-art constrained RL on the OpenAI safety gym benchmark and a human-robot collaboration task. Our study demonstrates that the combination of our method with provably safe RL leads to high-performing policies with zero safety violations and a low number of failsafe interventions. Our versatile method can be applied to a wide range of real-world robotic tasks, while effectively improving safety without sacrificing task performance."
                    },
                    {
                        "title": "Human-Robot Gym: Benchmarking Reinforcement Learning in Human-Robot Collaboration",
                        "abstract": "Deep reinforcement learning (RL) has shown promising results in robot motion planning with first attempts in human-robot collaboration (HRC). However, a fair comparison of RL approaches in HRC under the constraint of guaranteed safety is yet to be made. We, therefore, present human-robot gym, a benchmark suite for safe RL in HRC. Our benchmark suite provides eight challenging, realistic HRC tasks in a modular simulation framework. Most importantly, human-robot gym includes a safety shield that provably guarantees human safety. We are, thereby, the first to provide a benchmark suite to train RL agents that adhere to the safety specifications of real-world HRC. This bridges a critical gap between theoretic RL research and its real-world deployment. Our evaluation of six tasks led to three key results: (a) the diverse nature of the tasks offered by human-robot gym creates a challenging benchmark for state-of-the-art RL methods, (b) incorporating expert knowledge in RL training in the form of an action-based reward can outperform the expert, and (c) our agents negligibly overfit to training data."
                    },
                    {
                        "title": "Text2Interaction: Establishing Safe and Preferable Human-Robot Interaction",
                        "abstract": "Adjusting robot behavior to human preferences can require intensive human feedback, preventing quick adaptation to new users and changing circumstances. Moreover, current approaches typically treat user preferences as a reward, which requires a manual balance between task success and user satisfaction. To integrate new user preferences in a zero-shot manner, our proposed Text2Interaction framework invokes large language models to generate a task plan, motion preferences as Python code, and parameters of a safety controller. By maximizing the combined probability of task completion and user satisfaction instead of a weighted sum of rewards, we can reliably find plans that fulfill both requirements. We find that 83 % of users working with Text2Interaction agree that it integrates their preferences into the plan of the robot, and 94 % prefer Text2Interaction over the baseline. Our ablation study shows that Text2Interaction aligns better with unseen preferences than other baselines while maintaining a high success rate. Real-world demonstrations and code are made available at sites.google.com/view/text2interaction."
                    },
                    {
                        "title": "Provably Safe Reinforcement Learning: Conceptual Analysis, Survey, and Benchmarking",
                        "abstract": "Ensuring the safety of reinforcement learning (RL) algorithms is crucial to unlock their potential for many real-world tasks. However, vanilla RL and most safe RL approaches do not guarantee safety. In recent years, several methods have been proposed to provide hard safety guarantees for RL, which is essential for applications where unsafe actions could have disastrous consequences. Nevertheless, there is no comprehensive comparison of these provably safe RL methods. Therefore, we introduce a categorization of existing provably safe RL methods, present the conceptual foundations for both continuous and discrete action spaces, and empirically benchmark existing methods. We categorize the methods based on how they adapt the action: action replacement, action projection, and action masking. Our experiments on an inverted pendulum and a quadrotor stabilization task indicate that action replacement is the best-performing approach for these applications despite its comparatively simple realization. Furthermore, adding a reward penalty, every time the safety verification is engaged, improved training performance in our experiments. Finally, we provide practical guidance on selecting provably safe RL approaches depending on the safety specification, RL algorithm, and type of action space."
                    }
                ]
            },
            "4e84fb7a-4fe1-4774-b6b6-fcd2558d2a13": {
                "pk": "4e84fb7a-4fe1-4774-b6b6-fcd2558d2a13",
                "name": "Michael Eichelbeck",
                "collaborators": [
                    "Matthias Althoff",
                    "Hannah Markgraf",
                    "Jonas Fill",
                    "Michael Ebner",
                    "Tobias Ladner"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Graph Neural Network",
                    "Power Systems",
                    "Safety Verification"
                ],
                "publications": [
                    {
                        "title": "Contingency-constrained economic dispatch with safe reinforcement learning",
                        "abstract": "Future power systems will rely heavily on micro grids with a high share of decentralised renewable energy sources and energy storage systems. The high complexity and uncertainty in this context might make conventional power dispatch strategies infeasible. Reinforcement-learning based (RL) controllers can address this challenge, however, cannot themselves provide safety guarantees, preventing their deployment in practice. To overcome this limitation, we propose a formally validated RL controller for economic dispatch. We extend conventional constraints by a time-dependent constraint encoding the islanding contingency. The contingency constraint is computed using set-based backwards reachability analysis and actions of the RL agent are verified through a safety layer. Unsafe actions are projected into the safe action space while leveraging constrained zonotope set representations for computational efficiency. The developed approach is demonstrated on a residential use case using real-world measurements."
                    },
                    {
                        "title": "Empowering Safe Reinforcement Learning for Power System Control with CommonPower",
                        "abstract": "The growing complexity of power system management has led to an increased interest in reinforcement learning (RL). However, vanilla RL controllers cannot themselves ensure satisfaction of system constraints. Therefore, combining them with formally correct safeguarding mechanisms is an important aspect when studying RL for power system management. Integrating safeguarding into complex use cases requires tool support. To address this need, we introduce the Python tool CommonPower. CommonPower's unique contribution lies in its symbolic modeling approach, which enables flexible, model-based safeguarding of RL controllers. Moreover, CommonPower offers a unified interface for single-agent RL, multi-agent RL, and optimal control, with seamless integration of different forecasting methods. This allows users to validate the effectiveness of safe RL controllers across a large variety of case studies and investigate the influence of specific aspects on overall performance. We demonstrate CommonPower's versatility through a numerical case study that compares RL agents featuring different safeguards with a model predictive controller in the context of building energy management."
                    },
                    {
                        "title": "Predicting building types and functions at transnational scale",
                        "abstract": "Building-specific knowledge such as building type and function information is important for numerous energy applications. However, comprehensive datasets containing this information for individual households are missing in many regions of Europe. For the first time, we investigate whether it is feasible to predict building types and functional classes at a European scale based on only open GIS datasets available across countries. We train a graph neural network (GNN) classifier on a large-scale graph dataset consisting of OpenStreetMap (OSM) buildings across the EU, Norway, Switzerland, and the UK. To efficiently perform training using the large-scale graph, we utilize localized subgraphs. A graph transformer model achieves a high Cohen's kappa coefficient of 0.754 when classifying buildings into 9 classes, and a very high Cohen's kappa coefficient of 0.844 when classifying buildings into the residential and non-residential classes. The experimental results imply three core novel contributions to literature. Firstly, we show that building classification across multiple countries is possible using a multi-source dataset consisting of information about 2D building shape, land use, degree of urbanization, and countries as input, and OSM tags as ground truth. Secondly, our results indicate that GNN models that consider contextual information about building neighborhoods improve predictive performance compared to models that only consider individual buildings and ignore the neighborhood. Thirdly, we show that training with GNNs on localized subgraphs instead of standard GNNs improves performance for the task of building classification."
                    },
                    {
                        "title": "Formal Verification of Graph Convolutional Networks with Uncertain Node Features and Uncertain Graph Structure",
                        "abstract": "Graph neural networks are becoming increasingly popular in the field of machine learning due to their unique ability to process data structured in graphs. They have also been applied in safety-critical environments where perturbations inherently occur. However, these perturbations require us to formally verify neural networks before their deployment in safety-critical environments as neural networks are prone to adversarial attacks. While there exists research on the formal verification of neural networks, there is no work verifying the robustness of generic graph convolutional network architectures with uncertainty in the node features and in the graph structure over multiple message-passing steps. This work addresses this research gap by explicitly preserving the non-convex dependencies of all elements in the underlying computations through reachability analysis with (matrix) polynomial zonotopes. We demonstrate our approach on three popular benchmark datasets."
                    }
                ]
            },
            "604fc64e-9330-4a0c-8bb3-3c39b1b8c481": {
                "pk": "604fc64e-9330-4a0c-8bb3-3c39b1b8c481",
                "name": "Philipp Gassert",
                "collaborators": [
                    "Winfried Lohmiller",
                    "Jean-Jacques Slotine"
                ],
                "domain": [
                    "Machine Learning",
                    "Contraction Theory",
                    "Neural Networks",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Notes on stable learning with piecewise-linear basis functions",
                        "abstract": "We discuss technical results on learning function approximations using piecewise-linear basis functions, and analyze their stability and convergence using nonlinear contraction theory."
                    },
                    {
                        "title": "MinMax Networks",
                        "abstract": "While much progress has been achieved over the last decades in neuro-inspired machine learning, there are still fundamental theoretical problems in gradient-based learning using combinations of neurons. These problems, such as saddle points and suboptimal plateaus of the cost function, can lead in theory and practice to failures of learning. In addition, the discrete step size selection of the gradient is problematic since too large steps can lead to instability and too small steps slow down the learning. This paper describes an alternative discrete MinMax learning approach for continuous piece-wise linear functions. Global exponential convergence of the algorithm is established using Contraction Theory with Inequality Constraints, which is extended from the continuous to the discrete case in this paper:   The parametrization of each linear function piece is, in contrast to deep learning, linear in the proposed MinMax network. This allows a linear regression stability proof as long as measurements do not transit from one linear region to its neighbouring linear region.   The step size of the discrete gradient descent is Lagrangian limited orthogonal to the edge of two neighbouring linear functions. It will be shown that this Lagrangian step limitation does not decrease the convergence of the unconstrained system dynamics in contrast to a step size limitation in the direction of the gradient.   We show that the convergence rate of a constrained piece-wise linear function learning is equivalent to the exponential convergence rates of the individual local linear regions."
                    }
                ]
            },
            "7192a07a-7d36-42a5-baf8-9735c1092b4f": {
                "pk": "7192a07a-7d36-42a5-baf8-9735c1092b4f",
                "name": "Matthias Althoff",
                "collaborators": [
                    "Niklas Kochdumper",
                    "Jonathan K\u00fclz",
                    "Matthias Mayer",
                    "Chencheng Tang",
                    "Mark Wetzlinger",
                    "Laura L\u00fctzow",
                    "Sebastian Sigl",
                    "Jakob Thumm",
                    "Tobias Ladner",
                    "Sebastian Mair"
                ],
                "domain": [
                    "Formal Verification",
                    "Reachability Analysis",
                    "Robotics",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "On Computing the Minkowski Difference of Zonotopes",
                        "abstract": "Zonotopes are becoming an increasingly popular set representation for formal verification techniques. This is mainly due to their efficient representation and their favorable computational complexity of important operations in high-dimensional spaces. In particular, zonotopes are closed under Minkowski addition and linear maps, which can be very efficiently implemented. Unfortunately, zonotopes are not closed under Minkowski difference for dimensions greater than two. However, we present an algorithm that efficiently computes a halfspace representation of the Minkowski difference of two zonotopes. In addition, we present an efficient algorithm that computes an approximation of the Minkowski difference in generator representation. The efficiency of the proposed solution is demonstrated by numerical experiments. These experiments show a reduced computation time in comparison to that when first the halfspace representation of zonotopes is obtained and the Minkowski difference is performed subsequently."
                    },
                    {
                        "title": "Reachability Analysis of Large Linear Systems with Uncertain Inputs in the Krylov Subspace",
                        "abstract": "One often wishes for the ability to formally analyze large-scale systems---typically, however, one can either formally analyze a rather small system or informally analyze a large-scale system. This work tries to further close this performance gap for reachability analysis of linear systems. Reachability analysis can capture the whole set of possible solutions of a dynamic system and is thus used to prove that unsafe states are never reached; this requires full consideration of arbitrarily varying uncertain inputs, since sensor noise or disturbances usually do not follow any patterns. We use Krylov methods in this work to compute reachable sets for large-scale linear systems. While Krylov methods have been used before in reachability analysis, we overcome the previous limitation that inputs must be (piecewise) constant. As a result, we can compute reachable sets of systems with several thousand state variables for bounded, but arbitrarily varying inputs."
                    },
                    {
                        "title": "Sparse Polynomial Zonotopes: A Novel Set Representation for Reachability Analysis",
                        "abstract": "We introduce sparse polynomial zonotopes, a new set representation for formal verification of hybrid systems. Sparse polynomial zonotopes can represent non-convex sets and are generalizations of zonotopes, polytopes, and Taylor models. Operations like Minkowski sum, quadratic mapping, and reduction of the representation size can be computed with polynomial complexity w.r.t. the dimension of the system. In particular, for reachability analysis of nonlinear systems, the wrapping effect is substantially reduced using sparse polynomial zonotopes, as demonstrated by numerical examples. In addition, we can significantly reduce the computation time compared to zonotopes when dealing with nonlinear dynamics."
                    },
                    {
                        "title": "Constrained Polynomial Zonotopes",
                        "abstract": "We introduce constrained polynomial zonotopes, a novel non-convex set representation that is closed under linear map, Minkowski sum, Cartesian product, convex hull, intersection, union, and quadratic as well as higher-order maps. We show that the computational complexity of the above-mentioned set operations for constrained polynomial zonotopes is at most polynomial in the representation size. The fact that constrained polynomial zonotopes are generalizations of zonotopes, polytopes, polynomial zonotopes, Taylor models, and ellipsoids, further substantiates the relevance of this new set representation. The conversion from other set representations to constrained polynomial zonotopes is at most polynomial with respect to the dimension."
                    },
                    {
                        "title": "Representation of Polytopes as Polynomial Zonotopes",
                        "abstract": "We prove that each bounded polytope can be represented as a polynomial zonotope, which we refer to as the Z-representation of polytopes. Previous representations are the vertex representation (V-representation) and the halfspace representation (H-representation). Depending on the polytope, the Z-representation can be more compact than the V-representation and the H-representation. In addition, the Z-representation enables the computation of linear maps, Minkowski addition, and convex hull with a computational complexity that is polynomial in the representation size. The usefulness of the new representation is demonstrated by range bounding within polytopes."
                    },
                    {
                        "title": "M-Representation of Polytopes",
                        "abstract": "We introduce the M-representation of polytopes, which makes it possible to compute linear transformations, convex hulls, and Minkowski sums with linear complexity in the dimension of the polytopes. When the polytope is a convex hull of a zonotope and a polytope, the representation size can be smaller than any of the known representations (V-representation, H-representation, and Z-representation). We also provide a variant of the M-representation: The chain representation is more compact and we can directly use it to compute linear transformations and convex hulls -- for all other operations on the chain representation, one requires a conversion to the M-representation."
                    },
                    {
                        "title": "CoBRA: A Composable Benchmark for Robotics Applications",
                        "abstract": "Selecting an optimal robot, its base pose, and trajectory for a given task is currently mainly done by human expertise or trial and error. To evaluate automatic approaches to this combined optimization problem, we introduce a benchmark suite encompassing a unified format for robots, environments, and task descriptions. Our benchmark suite is especially useful for modular robots, where the multitude of robots that can be assembled creates a host of additional parameters to optimize. We include tasks such as machine tending and welding in synthetic environments and 3D scans of real-world machine shops. All benchmarks are accessible through https://cobra.cps.cit.tum.de, a platform to conveniently share, reference, and compare tasks, robot models, and solutions."
                    },
                    {
                        "title": "Timor Python: A Toolbox for Industrial Modular Robotics",
                        "abstract": "Modular Reconfigurable Robots (MRRs) represent an exciting path forward for industrial robotics, opening up new possibilities for robot design. Compared to monolithic manipulators, they promise greater flexibility, improved maintainability, and cost-efficiency. However, there is no tool or standardized way to model and simulate assemblies of modules in the same way it has been done for robotic manipulators for decades. We introduce the Toolbox for Industrial Modular Robotics (Timor), a Python toolbox to bridge this gap and integrate modular robotics into existing simulation and optimization pipelines. Our open-source library offers model generation and task-based configuration optimization for MRRs. It can easily be integrated with existing simulation tools - not least by offering URDF export of arbitrary modular robot assemblies. Moreover, our experimental study demonstrates the effectiveness of Timor as a tool for designing modular robots optimized for specific use cases."
                    },
                    {
                        "title": "Provably Safe Deep Reinforcement Learning for Robotic Manipulation in Human Environments",
                        "abstract": "Deep reinforcement learning (RL) has shown promising results in the motion planning of manipulators. However, no method guarantees the safety of highly dynamic obstacles, such as humans, in RL-based manipulator control. This lack of formal safety assurances prevents the application of RL for manipulators in real-world human environments. Therefore, we propose a shielding mechanism that ensures ISO-verified human safety while training and deploying RL algorithms on manipulators. We utilize a fast reachability analysis of humans and manipulators to guarantee that the manipulator comes to a complete stop before a human is within its range. Our proposed method guarantees safety and significantly improves the RL performance by preventing episode-ending collisions. We demonstrate the performance of our proposed method in simulation using human motion capture data."
                    },
                    {
                        "title": "Fully Automatic Neural Network Reduction for Formal Verification",
                        "abstract": "Formal verification of neural networks is essential before their deployment in safety-critical applications. However, existing methods for formally verifying neural networks are not yet scalable enough to handle practical problems involving a large number of neurons. We address this challenge by introducing a fully automatic and sound reduction of neural networks using reachability analysis. The soundness ensures that the verification of the reduced network entails the verification of the original network. To the best of our knowledge, we present the first sound reduction approach that is applicable to neural networks with any type of element-wise activation function, such as ReLU, sigmoid, and tanh. The network reduction is computed on the fly while simultaneously verifying the original network and its specifications. All parameters are automatically tuned to minimize the network size without compromising verifiability. We further show the applicability of our approach to convolutional neural networks by explicitly exploiting similar neighboring pixels. Our evaluation shows that our approach can reduce the number of neurons to a fraction of the original number of neurons with minor outer-approximation and thus reduce the verification time to a similar degree."
                    },
                    {
                        "title": "Formal Verification of Robotic Contact Tasks via Reachability Analysis",
                        "abstract": "Verifying the correct behavior of robots in contact tasks is challenging due to model uncertainties associated with contacts. Standard methods for testing often fall short since all (uncountable many) solutions cannot be obtained. Instead, we propose to formally and efficiently verify robot behaviors in contact tasks using reachability analysis, which enables checking all the reachable states against user-provided specifications. To this end, we extend the state of the art in reachability analysis for hybrid (mixed discrete and continuous) dynamics subject to discrete-time input trajectories. In particular, we present a novel and scalable guard intersection approach to reliably compute the complex behavior caused by contacts. We model robots subject to contacts as hybrid automata in which crucial time delays are included. The usefulness of our approach is demonstrated by verifying safe human-robot interaction in the presence of constrained collisions, which was out of reach for existing methods."
                    },
                    {
                        "title": "Backward Reachability Analysis of Perturbed Continuous-Time Linear Systems Using Set Propagation",
                        "abstract": "Backward reachability analysis computes the set of states that reach a target set under the competing influence of control input and disturbances. Depending on their interplay, the backward reachable set either represents all states that can be steered into the target set or all states that cannot avoid entering it -- the corresponding solutions can be used for controller synthesis and safety verification, respectively. A popular technique for backward reachable set computation solves Hamilton-Jacobi-Isaacs equations, which scales exponentially with the state dimension due to gridding the state space. In this work, we instead use set propagation techniques to design backward reachability algorithms for linear time-invariant systems. Crucially, the proposed algorithms scale only polynomially with the state dimension. Our numerical examples demonstrate the tightness of the obtained backward reachable sets and show an overwhelming improvement of our proposed algorithms over state-of-the-art methods regarding scalability, as systems with well over a hundred states can now be analyzed."
                    },
                    {
                        "title": "Provably-Correct Safety Protocol for Cooperative Platooning",
                        "abstract": "Cooperative Adaptive Cruise Control (CACC) is a well-studied technology for forming string-stable vehicle platoons. Ensuring collision avoidance is particularly difficult in CACC due to the small desired inter-vehicle spacing. We propose a safety protocol preventing collisions in a provably-correct manner while still maintaining a small distance to the preceding vehicle, by utilizing communicated braking capabilities. In addition, the safety of the protocol is ensured despite possible communication failures. While our concept can be applied to any CACC system, we particularly consider a class of CACCs, where the platoon vehicles successively agree on a consensus behavior. Our safety protocol is evaluated on various scenarios using the CommonRoad benchmark suite."
                    },
                    {
                        "title": "Reachset-Conformant System Identification",
                        "abstract": "Formal verification techniques play a pivotal role in ensuring the safety of complex cyber-physical systems. To transfer model-based verification results to the real world, we require that the measurements of the target system lie in the set of reachable outputs of the corresponding model, a property we refer to as reachset conformance. This paper is on automatically identifying those reachset-conformant models. While state-of-the-art reachset-conformant identification methods focus on linear state-space models, we generalize these methods to nonlinear state-space models and linear and nonlinear input-output models. Furthermore, our identification framework adapts to different levels of prior knowledge on the system dynamics. In particular, we identify the set of model uncertainties for white-box models, the parameters and the set of model uncertainties for gray-box models, and entire reachset-conformant black-box models from data. For the black-box identification, we propose a new genetic programming variant, which we call conformant genetic programming. The robustness and efficacy of our framework are demonstrated in extensive numerical experiments using simulated and real-world data."
                    },
                    {
                        "title": "Efficiently Obtaining Reachset Conformance for the Formal Analysis of Robotic Contact Tasks",
                        "abstract": "Formal verification of robotic tasks requires a simple yet conformant model of the used robot. We present the first work on generating reachset conformant models for robotic contact tasks considering hybrid (mixed continuous and discrete) dynamics. Reachset conformance requires that the set of reachable outputs of the abstract model encloses all previous measurements to transfer safety properties. Aiming for industrial applications, we describe the system using a simple hybrid automaton with linear dynamics. We inject non-determinism into the continuous dynamics and the discrete transitions, and we optimally identify all model parameters together with the non-determinism required to capture the recorded behaviors. Using two 3-DOF robots, we show that our approach can effectively generate models to capture uncertainties in system behavior and substantially reduce the required testing effort in industrial applications."
                    },
                    {
                        "title": "Optimizing Modular Robot Composition: A Lexicographic Genetic Algorithm Approach",
                        "abstract": "Industrial robots are designed as general-purpose hardware with limited ability to adapt to changing task requirements or environments. Modular robots, on the other hand, offer flexibility and can be easily customized to suit diverse needs. The morphology, i.e., the form and structure of a robot, significantly impacts the primary performance metrics acquisition cost, cycle time, and energy efficiency. However, identifying an optimal module composition for a specific task remains an open problem, presenting a substantial hurdle in developing task-tailored modular robots. Previous approaches either lack adequate exploration of the design space or the possibility to adapt to complex tasks. We propose combining a genetic algorithm with a lexicographic evaluation of solution candidates to overcome this problem and navigate search spaces exceeding those in prior work by magnitudes in the number of possible compositions. We demonstrate that our approach outperforms a state-of-the-art baseline and is able to synthesize modular robots for industrial tasks in cluttered environments."
                    },
                    {
                        "title": "Reachability Analysis of ARMAX Models",
                        "abstract": "Reachability analysis is a powerful tool for computing the set of states or outputs reachable for a system. While previous work has focused on systems described by state-space models, we present the first methods to compute reachable sets of ARMAX models - one of the most common input-output models originating from data-driven system identification. The first approach we propose can only be used with dependency-preserving set representations such as symbolic zonotopes, while the second one is valid for arbitrary set representations but relies on a reformulation of the ARMAX model. By analyzing the computational complexities, we show that both approaches scale quadratically with respect to the time horizon of the reachability problem when using symbolic zonotopes. To reduce the computational complexity, we propose a third approach that scales linearly with respect to the time horizon when using set representations that are closed under Minkowski addition and linear transformation and that satisfy that the computational complexity of the Minkowski sum is independent of the representation size of the operands. Our numerical experiments demonstrate that the reachable sets of ARMAX models are tighter than the reachable sets of equivalent state space models in case of unknown initial states. Therefore, this methodology has the potential to significantly reduce the conservatism of various verification techniques."
                    },
                    {
                        "title": "Provable Traffic Rule Compliance in Safe Reinforcement Learning on the Open Sea",
                        "abstract": "For safe operation, autonomous vehicles have to obey traffic rules that are set forth in legal documents formulated in natural language. Temporal logic is a suitable concept to formalize such traffic rules. Still, temporal logic rules often result in constraints that are hard to solve using optimization-based motion planners. Reinforcement learning (RL) is a promising method to find motion plans for autonomous vehicles. However, vanilla RL algorithms are based on random exploration and do not automatically comply with traffic rules. Our approach accomplishes guaranteed rule-compliance by integrating temporal logic specifications into RL. Specifically, we consider the application of vessels on the open sea, which must adhere to the Convention on the International Regulations for Preventing Collisions at Sea (COLREGS). To efficiently synthesize rule-compliant actions, we combine predicates based on set-based prediction with a statechart representing our formalized rules and their priorities. Action masking then restricts the RL agent to this set of verified rule-compliant actions. In numerical evaluations on critical maritime traffic situations, our agent always complies with the formalized legal rules and never collides while achieving a high goal-reaching rate during training and deployment. In contrast, vanilla and traffic rule-informed RL agents frequently violate traffic rules and collide even after training."
                    },
                    {
                        "title": "Deep Occupancy-Predictive Representations for Autonomous Driving",
                        "abstract": "Manually specifying features that capture the diversity in traffic environments is impractical. Consequently, learning-based agents cannot realize their full potential as neural motion planners for autonomous vehicles. Instead, this work proposes to learn which features are task-relevant. Given its immediate relevance to motion planning, our proposed architecture encodes the probabilistic occupancy map as a proxy for obtaining pre-trained state representations. By leveraging a map-aware graph formulation of the environment, our agent-centric encoder generalizes to arbitrary road networks and traffic situations. We show that our approach significantly improves the downstream performance of a reinforcement learning agent operating in urban traffic environments."
                    },
                    {
                        "title": "Adaptive Parameter Tuning for Reachability Analysis of Linear Systems",
                        "abstract": "Despite the possibility to quickly compute reachable sets of large-scale linear systems, current methods are not yet widely applied by practitioners. The main reason for this is probably that current approaches are not push-button-capable and still require to manually set crucial parameters, such as time step sizes and the accuracy of the used set representation -- these settings require expert knowledge. We present a generic framework to automatically find near-optimal parameters for reachability analysis of linear systems given a user-defined accuracy. To limit the computational overhead as much as possible, our methods tune all relevant parameters during runtime. We evaluate our approach on benchmarks from the ARCH competition as well as on random examples. Our results show that our new framework verifies the selected benchmarks faster than manually-tuned parameters and is an order of magnitude faster compared to genetic algorithms."
                    }
                ]
            }
        }
    },
    "2409.04792": {
        "paper_data": {
            "title": "Improving Deep Reinforcement Learning by Reducing the Chain Effect of Value and Policy Churn",
            "url": "http://arxiv.org/abs/2409.04792v1",
            "arxiv_id": "2409.04792",
            "authors": [
                "Hongyao Tang",
                "Glen Berseth"
            ],
            "abstract": "Deep neural networks provide Reinforcement Learning (RL) powerful function approximators to address large-scale decision-making problems. However, these approximators introduce challenges due to the non-stationary nature of RL training. One source of the challenges in RL is that output predictions can churn, leading to uncontrolled changes after each batch update for states not included in the batch. Although such a churn phenomenon exists in each step of network training, how churn occurs and impacts RL remains under-explored. In this work, we start by characterizing churn in a view of Generalized Policy Iteration with function approximation, and we discover a chain effect of churn that leads to a cycle where the churns in value estimation and policy improvement compound and bias the learning dynamics throughout the iteration. Further, we concretize the study and focus on the learning issues caused by the chain effect in different settings, including greedy action deviation in value-based methods, trust region violation in proximal policy optimization, and dual bias of policy value in actor-critic methods. We then propose a method to reduce the chain effect across different settings, called Churn Approximated ReductIoN (CHAIN), which can be easily plugged into most existing DRL algorithms. Our experiments demonstrate the effectiveness of our method in both reducing churn and improving learning performance across online and offline, value-based and policy-based RL settings, as well as a scaling setting.",
            "introduction": "   1 Introduction  One fundamental recipe for the success of Deep Reinforcement Learning (DRL) is powerful approximation and generalization provided by deep neural networks, which augments the ability of RL with tabular or linear approximation to large state spaces. However, on the other side of this benefit is less control over the function dynamics. Network outputs can change indirectly to unexpected values after any random batch update for input data not included in the batch, called churn in this paper. This change is particularly problematic for an RL agent due to its non-stationary nature, which can exacerbate instability, suboptimality and even collapse. Therefore, it is of significance to understand and control these undesired dynamics to address learning issues and improve performance.   Consistent efforts have been devoted by RL community to gain better understandings of the learning dynamics from different perspectives\u00a0(Achiam et\u00a0al., 2019; Kumar et\u00a0al., 2022; Liu et\u00a0al., 2023; Lyle et\u00a0al., 2022b). Recently, Schaul et\u00a0al. (2022) studied a novel churn phenomenon in the learning process of typical value-based RL algorithms like DoubleDQN\u00a0(van Hasselt et\u00a0al., 2016). The phenomenon reveals that the greedy actions of about 10%percent1010\\%10 % of states in the replay buffer change after a single regular batch update. Such a dramatic churn can persist throughout the learning process of DoubleDQN, causing instabilities.   In this paper, we aim to take a step further to understand how churn occurs and influences learning in different DRL settings beyond value-based RL, as well as to propose a method to control churn. We start by characterizing churn formally in a view of Generalized Policy Iteration (GPI) with function approximation, to best cover most DRL settings. The impact of churn is two-fold in this view: the churn in policy improvement (called the policy churn) changes policy outputs on states that are not directly updated; while the churn in value estimation (called the value churn) also changes action-value landscape, thus altering greedy action and action gradient. We then discover a chain effect of churn that exhibits a cycle where the two types of churn compound and bias the learning dynamics throughout the iteration. Further, we move on from the general analysis to concrete DRL settings. We focus on the learning issues caused by the chain effect including greedy action deviation in value-based methods, trust region violation in proximal policy optimization\u00a0(Schulman et\u00a0al., 2017) and dual bias of policy value in actor-critic methods. The connection between the chain effect of churn and the issues necessitates an explicit control of the churn.   To this end, we propose a method called Churn Approximated ReductIoN (CHAIN), to reduce the chain effect of churn across different DRL settings. The main idea of CHAIN is to reduce the undesirable changes to the outputs of the policy and value networks for states (and actions) outside of the current batch of data for regular DRL training. This reduction is achieved by minimizing the change in target values for a separate batch of data when optimizing the original policy or value learning objective. CHAIN is easy to implement and plug in most existing DRL algorithms with only a few lines of code. In our experiments, we evaluate the efficacy of CHAIN in a range of environments, including MinAtar\u00a0(Young and Tian, 2019),",
            "references": [
                {
                    "title": "Overestimation, Overfitting, and Plasticity in Actor-Critic: the Bitter Lesson of Reinforcement Learning",
                    "abstract": "Recent advancements in off-policy Reinforcement Learning (RL) have significantly improved sample efficiency, primarily due to the incorporation of various forms of regularization that enable more gradient update steps than traditional agents. However, many of these techniques have been tested in limited settings, often on tasks from single simulation benchmarks and against well-known algorithms rather than a range of regularization approaches. This limits our understanding of the specific mechanisms driving RL improvements. To address this, we implemented over 60 different off-policy agents, each integrating established regularization techniques from recent state-of-the-art algorithms. We tested these agents across 14 diverse tasks from 2 simulation benchmarks, measuring training metrics related to overestimation, overfitting, and plasticity loss -- issues that motivate the examined regularization techniques. Our findings reveal that while the effectiveness of a specific regularization setup varies with the task, certain combinations consistently demonstrate robust and superior performance. Notably, a simple Soft Actor-Critic agent, appropriately regularized, reliably finds a better-performing policy within the training regime, which previously was achieved mainly through model-based approaches."
                },
                {
                    "title": "Disentangling the Causes of Plasticity Loss in Neural Networks",
                    "abstract": "Underpinning the past decades of work on the design, initialization, and optimization of neural networks is a seemingly innocuous assumption: that the network is trained on a \\textit{stationary} data distribution. In settings where this assumption is violated, e.g.\\ deep reinforcement learning, learning algorithms become unstable and brittle with respect to hyperparameters and even random seeds. One factor driving this instability is the loss of plasticity, meaning that updating the network's predictions in response to new information becomes more difficult as training progresses. While many recent works provide analyses and partial solutions to this phenomenon, a fundamental question remains unanswered: to what extent do known mechanisms of plasticity loss overlap, and how can mitigation strategies be combined to best maintain the trainability of a network? This paper addresses these questions, showing that loss of plasticity can be decomposed into multiple independent mechanisms and that, while intervening on any single mechanism is insufficient to avoid the loss of plasticity in all cases, intervening on multiple mechanisms in conjunction results in highly robust learning algorithms. We show that a combination of layer normalization and weight decay is highly effective at maintaining plasticity in a variety of synthetic nonstationary learning tasks, and further demonstrate its effectiveness on naturally arising nonstationarities, including reinforcement learning in the Arcade Learning Environment."
                },
                {
                    "title": "DrM: Mastering Visual Reinforcement Learning through Dormant Ratio Minimization",
                    "abstract": "Visual reinforcement learning (RL) has shown promise in continuous control tasks. Despite its progress, current algorithms are still unsatisfactory in virtually every aspect of the performance such as sample efficiency, asymptotic performance, and their robustness to the choice of random seeds. In this paper, we identify a major shortcoming in existing visual RL methods that is the agents often exhibit sustained inactivity during early training, thereby limiting their ability to explore effectively. Expanding upon this crucial observation, we additionally unveil a significant correlation between the agents' inclination towards motorically inactive exploration and the absence of neuronal activity within their policy networks. To quantify this inactivity, we adopt dormant ratio as a metric to measure inactivity in the RL agent's network. Empirically, we also recognize that the dormant ratio can act as a standalone indicator of an agent's activity level, regardless of the received reward signals. Leveraging the aforementioned insights, we introduce DrM, a method that uses three core mechanisms to guide agents' exploration-exploitation trade-offs by actively minimizing the dormant ratio. Experiments demonstrate that DrM achieves significant improvements in sample efficiency and asymptotic performance with no broken seeds (76 seeds in total) across three continuous control benchmark environments, including DeepMind Control Suite, MetaWorld, and Adroit. Most importantly, DrM is the first model-free algorithm that consistently solves tasks in both the Dog and Manipulator domains from the DeepMind Control Suite as well as three dexterous hand manipulation tasks without demonstrations in Adroit, all based on pixel observations."
                },
                {
                    "title": "Measuring and Mitigating Interference in Reinforcement Learning",
                    "abstract": "Catastrophic interference is common in many network-based learning systems, and many proposals exist for mitigating it. Before overcoming interference we must understand it better. In this work, we provide a definition and novel measure of interference for value-based reinforcement learning methods such as Fitted Q-Iteration and DQN. We systematically evaluate our measure of interference, showing that it correlates with instability in control performance, across a variety of network architectures. Our new interference measure allows us to ask novel scientific questions about commonly used deep learning architectures and study learning algorithms which mitigate interference. Lastly, we outline a class of algorithms which we call online-aware that are designed to mitigate interference, and show they do reduce interference according to our measure and that they improve stability and performance in several classic control environments."
                },
                {
                    "title": "Maintaining Plasticity in Deep Continual Learning",
                    "abstract": "Modern deep-learning systems are specialized to problem settings in which training occurs once and then never again, as opposed to continual-learning settings in which training occurs continually. If deep-learning systems are applied in a continual learning setting, then it is well known that they may fail to remember earlier examples. More fundamental, but less well known, is that they may also lose their ability to learn on new examples, a phenomenon called loss of plasticity. We provide direct demonstrations of loss of plasticity using the MNIST and ImageNet datasets repurposed for continual learning as sequences of tasks. In ImageNet, binary classification performance dropped from 89% accuracy on an early task down to 77%, about the level of a linear network, on the 2000th task. Loss of plasticity occurred with a wide range of deep network architectures, optimizers, activation functions, batch normalization, dropout, but was substantially eased by L2-regularization, particularly when combined with weight perturbation. Further, we introduce a new algorithm -- continual backpropagation -- which slightly modifies conventional backpropagation to reinitialize a small fraction of less-used units after each example and appears to maintain plasticity indefinitely."
                },
                {
                    "title": "Loss of Plasticity in Continual Deep Reinforcement Learning",
                    "abstract": "The ability to learn continually is essential in a complex and changing world. In this paper, we characterize the behavior of canonical value-based deep reinforcement learning (RL) approaches under varying degrees of non-stationarity. In particular, we demonstrate that deep RL agents lose their ability to learn good policies when they cycle through a sequence of Atari 2600 games. This phenomenon is alluded to in prior work under various guises -- e.g., loss of plasticity, implicit under-parameterization, primacy bias, and capacity loss. We investigate this phenomenon closely at scale and analyze how the weights, gradients, and activations change over time in several experiments with varying dimensions (e.g., similarity between games, number of games, number of frames per game), with some experiments spanning 50 days and 2 billion environment interactions. Our analysis shows that the activation footprint of the network becomes sparser, contributing to the diminishing gradients. We investigate a remarkably simple mitigation strategy -- Concatenated ReLUs (CReLUs) activation function -- and demonstrate its effectiveness in facilitating continual learning in a changing environment."
                },
                {
                    "title": "The Dormant Neuron Phenomenon in Deep Reinforcement Learning",
                    "abstract": "In this work we identify the dormant neuron phenomenon in deep reinforcement learning, where an agent's network suffers from an increasing number of inactive neurons, thereby affecting network expressivity. We demonstrate the presence of this phenomenon across a variety of algorithms and environments, and highlight its effect on learning. To address this issue, we propose a simple and effective method (ReDo) that Recycles Dormant neurons throughout training. Our experiments demonstrate that ReDo maintains the expressive power of networks by reducing the number of dormant neurons and results in improved performance."
                },
                {
                    "title": "Mastering Diverse Domains through World Models",
                    "abstract": "Developing a general algorithm that learns to solve tasks across a wide range of applications has been a fundamental challenge in artificial intelligence. Although current reinforcement learning algorithms can be readily applied to tasks similar to what they have been developed for, configuring them for new application domains requires significant human expertise and experimentation. We present DreamerV3, a general algorithm that outperforms specialized methods across over 150 diverse tasks, with a single configuration. Dreamer learns a model of the environment and improves its behavior by imagining future scenarios. Robustness techniques based on normalization, balancing, and transformations enable stable learning across domains. Applied out of the box, Dreamer is the first algorithm to collect diamonds in Minecraft from scratch without human data or curricula. This achievement has been posed as a significant challenge in artificial intelligence that requires exploring farsighted strategies from pixels and sparse rewards in an open world. Our work allows solving challenging control problems without extensive experimentation, making reinforcement learning broadly applicable."
                },
                {
                    "title": "Human-level Atari 200x faster",
                    "abstract": "The task of building general agents that perform well over a wide range of tasks has been an importantgoal in reinforcement learning since its inception. The problem has been subject of research of alarge body of work, with performance frequently measured by observing scores over the wide rangeof environments contained in the Atari 57 benchmark. Agent57 was the first agent to surpass thehuman benchmark on all 57 games, but this came at the cost of poor data-efficiency, requiring nearly 80billion frames of experience to achieve. Taking Agent57 as a starting point, we employ a diverse set ofstrategies to achieve a 200-fold reduction of experience needed to outperform the human baseline. Weinvestigate a range of instabilities and bottlenecks we encountered while reducing the data regime, andpropose effective solutions to build a more robust and efficient agent. We also demonstrate competitiveperformance with high-performing methods such as Muesli and MuZero. The four key components toour approach are (1) an approximate trust region method which enables stable bootstrapping from theonline network, (2) a normalisation scheme for the loss and priorities which improves robustness whenlearning a set of value functions with a wide range of scales, (3) an improved architecture employingtechniques from NFNets in order to leverage deeper networks without the need for normalization layers,and (4) a policy distillation method which serves to smooth out the instantaneous greedy policy overtime."
                },
                {
                    "title": "What about Inputting Policy in Value Function: Policy Representation and Policy-Extended Value Function Approximator",
                    "abstract": "We study Policy-extended Value Function Approximator (PeVFA) in Reinforcement Learning (RL), which extends conventional value function approximator (VFA) to take as input not only the state (and action) but also an explicit policy representation. Such an extension enables PeVFA to preserve values of multiple policies at the same time and brings an appealing characteristic, i.e., value generalization among policies. We formally analyze the value generalization under Generalized Policy Iteration (GPI). From theoretical and empirical lens, we show that generalized value estimates offered by PeVFA may have lower initial approximation error to true values of successive policies, which is expected to improve consecutive value approximation during GPI. Based on above clues, we introduce a new form of GPI with PeVFA which leverages the value generalization along policy improvement path. Moreover, we propose a representation learning framework for RL policy, providing several approaches to learn effective policy embeddings from policy network parameters or state-action pairs. In our experiments, we evaluate the efficacy of value generalization offered by PeVFA and policy representation learning in several OpenAI Gym continuous control tasks. For a representative instance of algorithm implementation, Proximal Policy Optimization (PPO) re-implemented under the paradigm of GPI with PeVFA achieves about 40% performance improvement on its vanilla counterpart in most environments."
                },
                {
                    "title": "Overcoming the Spectral Bias of Neural Value Approximation",
                    "abstract": "Value approximation using deep neural networks is at the heart of off-policy deep reinforcement learning, and is often the primary module that provides learning signals to the rest of the algorithm. While multi-layer perceptron networks are universal function approximators, recent works in neural kernel regression suggest the presence of a spectral bias, where fitting high-frequency components of the value function requires exponentially more gradient update steps than the low-frequency ones. In this work, we re-examine off-policy reinforcement learning through the lens of kernel regression and propose to overcome such bias via a composite neural tangent kernel. With just a single line-change, our approach, the Fourier feature networks (FFN) produce state-of-the-art performance on challenging continuous control domains with only a fraction of the compute. Faster convergence and better off-policy stability also make it possible to remove the target network without suffering catastrophic divergences, which further reduces TD}(0)'s estimation bias on a few tasks."
                },
                {
                    "title": "The Phenomenon of Policy Churn",
                    "abstract": "We identify and study the phenomenon of policy churn, that is, the rapid change of the greedy policy in value-based reinforcement learning. Policy churn operates at a surprisingly rapid pace, changing the greedy action in a large fraction of states within a handful of learning updates (in a typical deep RL set-up such as DQN on Atari). We characterise the phenomenon empirically, verifying that it is not limited to specific algorithm or environment properties. A number of ablations help whittle down the plausible explanations on why churn occurs to just a handful, all related to deep learning. Finally, we hypothesise that policy churn is a beneficial but overlooked form of implicit exploration that casts $\\epsilon$-greedy exploration in a fresh light, namely that $\\epsilon$-noise plays a much smaller role than expected."
                },
                {
                    "title": "Memory-efficient Reinforcement Learning with Value-based Knowledge Consolidation",
                    "abstract": "Artificial neural networks are promising for general function approximation but challenging to train on non-independent or non-identically distributed data due to catastrophic forgetting. The experience replay buffer, a standard component in deep reinforcement learning, is often used to reduce forgetting and improve sample efficiency by storing experiences in a large buffer and using them for training later. However, a large replay buffer results in a heavy memory burden, especially for onboard and edge devices with limited memory capacities. We propose memory-efficient reinforcement learning algorithms based on the deep Q-network algorithm to alleviate this problem. Our algorithms reduce forgetting and maintain high sample efficiency by consolidating knowledge from the target Q-network to the current Q-network. Compared to baseline methods, our algorithms achieve comparable or better performance in both feature-based and image-based tasks while easing the burden of large experience replay buffers."
                },
                {
                    "title": "The Primacy Bias in Deep Reinforcement Learning",
                    "abstract": "This work identifies a common flaw of deep reinforcement learning (RL) algorithms: a tendency to rely on early interactions and ignore useful evidence encountered later. Because of training on progressively growing datasets, deep RL agents incur a risk of overfitting to earlier experiences, negatively affecting the rest of the learning process. Inspired by cognitive science, we refer to this effect as the primacy bias. Through a series of experiments, we dissect the algorithmic aspects of deep RL that exacerbate this bias. We then propose a simple yet generally-applicable mechanism that tackles the primacy bias by periodically resetting a part of the agent. We apply this mechanism to algorithms in both discrete (Atari 100k) and continuous action (DeepMind Control Suite) domains, consistently improving their performance."
                },
                {
                    "title": "Understanding and Preventing Capacity Loss in Reinforcement Learning",
                    "abstract": "The reinforcement learning (RL) problem is rife with sources of non-stationarity, making it a notoriously difficult problem domain for the application of neural networks. We identify a mechanism by which non-stationary prediction targets can prevent learning progress in deep RL agents: \\textit{capacity loss}, whereby networks trained on a sequence of target values lose their ability to quickly update their predictions over time. We demonstrate that capacity loss occurs in a range of RL agents and environments, and is particularly damaging to performance in sparse-reward tasks. We then present a simple regularizer, Initial Feature Regularization (InFeR), that mitigates this phenomenon by regressing a subspace of features towards its value at initialization, leading to significant performance improvements in sparse-reward environments such as Montezuma's Revenge. We conclude that preventing capacity loss is crucial to enable agents to maximally benefit from the learning signals they obtain throughout the entire training trajectory."
                },
                {
                    "title": "Target Network and Truncation Overcome The Deadly triad in Q-Learning",
                    "abstract": "$Q$-learning with function approximation is one of the most empirically successful while theoretically mysterious reinforcement learning (RL) algorithms, and was identified in Sutton (1999) as one of the most important theoretical open problems in the RL community. Even in the basic linear function approximation setting, there are well-known divergent examples. In this work, we show that \\textit{target network} and \\textit{truncation} together are enough to provably stabilize $Q$-learning with linear function approximation, and we establish the finite-sample guarantees. The result implies an $O(\\epsilon^{-2})$ sample complexity up to a function approximation error. Moreover, our results do not require strong assumptions or modifying the problem parameters as in existing literature."
                },
                {
                    "title": "DR3: Value-Based Deep Reinforcement Learning Requires Explicit Regularization",
                    "abstract": "Despite overparameterization, deep networks trained via supervised learning are easy to optimize and exhibit excellent generalization. One hypothesis to explain this is that overparameterized deep networks enjoy the benefits of implicit regularization induced by stochastic gradient descent, which favors parsimonious solutions that generalize well on test inputs. It is reasonable to surmise that deep reinforcement learning (RL) methods could also benefit from this effect. In this paper, we discuss how the implicit regularization effect of SGD seen in supervised learning could in fact be harmful in the offline deep RL setting, leading to poor generalization and degenerate feature representations. Our theoretical analysis shows that when existing models of implicit regularization are applied to temporal difference learning, the resulting derived regularizer favors degenerate solutions with excessive\"aliasing\", in stark contrast to the supervised learning case. We back up these findings empirically, showing that feature representations learned by a deep network value function trained via bootstrapping can indeed become degenerate, aliasing the representations for state-action pairs that appear on either side of the Bellman backup. To address this issue, we derive the form of this implicit regularizer and, inspired by this derivation, propose a simple and effective explicit regularizer, called DR3, that counteracts the undesirable effects of this implicit regularizer. When combined with existing offline RL methods, DR3 substantially improves performance and stability, alleviating unlearning in Atari 2600 games, D4RL domains and robotic manipulation from images."
                },
                {
                    "title": "The Difficulty of Passive Learning in Deep Reinforcement Learning",
                    "abstract": "Learning to act from observational data without active environmental interaction is a well-known challenge in Reinforcement Learning (RL). Recent approaches involve constraints on the learned policy or conservative updates, preventing strong deviations from the state-action distribution of the dataset. Although these methods are evaluated using non-linear function approximation, theoretical justifications are mostly limited to the tabular or linear cases. Given the impressive results of deep reinforcement learning, we argue for a need to more clearly understand the challenges in this setting. In the vein of Held&Hein's classic 1963 experiment, we propose the\"tandem learning\"experimental paradigm which facilitates our empirical analysis of the difficulties in offline reinforcement learning. We identify function approximation in conjunction with fixed data distributions as the strongest factors, thereby extending but also challenging hypotheses stated in past work. Our results provide relevant insights for offline deep reinforcement learning, while also shedding new light on phenomena observed in the online case of learning control."
                },
                {
                    "title": "Offline Reinforcement Learning with Implicit Q-Learning",
                    "abstract": "Offline reinforcement learning requires reconciling two conflicting aims: learning a policy that improves over the behavior policy that collected the dataset, while at the same time minimizing the deviation from the behavior policy so as to avoid errors due to distributional shift. This trade-off is critical, because most current offline reinforcement learning methods need to query the value of unseen actions during training to improve the policy, and therefore need to either constrain these actions to be in-distribution, or else regularize their values. We propose an offline RL method that never needs to evaluate actions outside of the dataset, but still enables the learned policy to improve substantially over the best behavior in the data through generalization. The main insight in our work is that, instead of evaluating unseen actions from the latest policy, we can approximate the policy improvement step implicitly by treating the state value function as a random variable, with randomness determined by the action (while still integrating over the dynamics to avoid excessive optimism), and then taking a state conditional upper expectile of this random variable to estimate the value of the best actions in that state. This leverages the generalization capacity of the function approximator to estimate the value of the best available action at a given state without ever directly querying a Q-function with this unseen action. Our algorithm alternates between fitting this upper expectile value function and backing it up into a Q-function. Then, we extract the policy via advantage-weighted behavioral cloning. We dub our method implicit Q-learning (IQL). IQL demonstrates the state-of-the-art performance on D4RL, a standard benchmark for offline reinforcement learning. We also demonstrate that IQL achieves strong performance fine-tuning using online interaction after offline initialization."
                },
                {
                    "title": "HyAR: Addressing Discrete-Continuous Action Reinforcement Learning via Hybrid Action Representation",
                    "abstract": "Discrete-continuous hybrid action space is a natural setting in many practical problems, such as robot control and game AI. However, most previous Reinforcement Learning (RL) works only demonstrate the success in controlling with either discrete or continuous action space, while seldom take into account the hybrid action space. One naive way to address hybrid action RL is to convert the hybrid action space into a unified homogeneous action space by discretization or continualization, so that conventional RL algorithms can be applied. However, this ignores the underlying structure of hybrid action space and also induces the scalability issue and additional approximation difficulties, thus leading to degenerated results. In this paper, we propose Hybrid Action Representation (HyAR) to learn a compact and decodable latent representation space for the original hybrid action space. HyAR constructs the latent space and embeds the dependence between discrete action and continuous parameter via an embedding table and conditional Variantional Auto-Encoder (VAE). To further improve the effectiveness, the action representation is trained to be semantically smooth through unsupervised environmental dynamics prediction. Finally, the agent then learns its policy with conventional DRL algorithms in the learned representation space and interacts with the environment by decoding the hybrid action embeddings to the original action space. We evaluate HyAR in a variety of environments with discrete-continuous action space. The results demonstrate the superiority of HyAR when compared with previous baselines, especially for high-dimensional action spaces."
                },
                {
                    "title": "Deep Reinforcement Learning at the Edge of the Statistical Precipice",
                    "abstract": "Deep reinforcement learning (RL) algorithms are predominantly evaluated by comparing their relative performance on a large suite of tasks. Most published results on deep RL benchmarks compare point estimates of aggregate performance such as mean and median scores across tasks, ignoring the statistical uncertainty implied by the use of a finite number of training runs. Beginning with the Arcade Learning Environment (ALE), the shift towards computationally-demanding benchmarks has led to the practice of evaluating only a small number of runs per task, exacerbating the statistical uncertainty in point estimates. In this paper, we argue that reliable evaluation in the few run deep RL regime cannot ignore the uncertainty in results without running the risk of slowing down progress in the field. We illustrate this point using a case study on the Atari 100k benchmark, where we find substantial discrepancies between conclusions drawn from point estimates alone versus a more thorough statistical analysis. With the aim of increasing the field's confidence in reported results with a handful of runs, we advocate for reporting interval estimates of aggregate performance and propose performance profiles to account for the variability in results, as well as present more robust and efficient aggregate metrics, such as interquartile mean scores, to achieve small uncertainty in results. Using such statistical tools, we scrutinize performance evaluations of existing algorithms on other widely used RL benchmarks including the ALE, Procgen, and the DeepMind Control Suite, again revealing discrepancies in prior comparisons. Our findings call for a change in how we evaluate performance in deep RL, for which we present a more rigorous evaluation methodology, accompanied with an open-source library rliable, to prevent unreliable results from stagnating the field."
                },
                {
                    "title": "OptiDICE: Offline Policy Optimization via Stationary Distribution Correction Estimation",
                    "abstract": "We consider the offline reinforcement learning (RL) setting where the agent aims to optimize the policy solely from the data without further environment interactions. In offline RL, the distributional shift becomes the primary source of difficulty, which arises from the deviation of the target policy being optimized from the behavior policy used for data collection. This typically causes overestimation of action values, which poses severe problems for model-free algorithms that use bootstrapping. To mitigate the problem, prior offline RL algorithms often used sophisticated techniques that encourage underestimation of action values, which introduces an additional set of hyperparameters that need to be tuned properly. In this paper, we present an offline RL algorithm that prevents overestimation in a more principled way. Our algorithm, OptiDICE, directly estimates the stationary distribution corrections of the optimal policy and does not rely on policy-gradients, unlike previous offline RL algorithms. Using an extensive set of benchmark datasets for offline RL, we show that OptiDICE performs competitively with the state-of-the-art methods."
                },
                {
                    "title": "Bridging the Gap Between Target Networks and Functional Regularization",
                    "abstract": "Bootstrapping is behind much of the successes of deep Reinforcement Learning. However, learning the value function via bootstrapping often leads to unstable training due to fast-changing target values. Target Networks are employed to stabilize training by using an additional set of lagging parameters to estimate the target values. Despite the popularity of Target Networks, their effect on the optimization is still misunderstood. In this work, we show that they act as an implicit regularizer which can be beneficial in some cases, but also have disadvantages such as being inflexible and can result in instabilities, even when vanilla TD(0) converges. To overcome these issues, we propose an explicit Functional Regularization alternative that is flexible and a convex regularizer in function space and we theoretically study its convergence. We conduct an experimental study across a range of environments, discount factors, and off-policiness data collections to investigate the effectiveness of the regularization induced by Target Networks and Functional Regularization in terms of performance, accuracy, and stability. Our findings emphasize that Functional Regularization can be used as a drop-in replacement for Target Networks and result in performance improvement. Furthermore, adjusting both the regularization weight and the network update period in Functional Regularization can result in further performance improvements compared to solely adjusting the network update period as typically done with Target Networks. Our approach also enhances the ability to networks to recover accurate $Q$-values."
                },
                {
                    "title": "Decoupling Value and Policy for Generalization in Reinforcement Learning",
                    "abstract": "Standard deep reinforcement learning algorithms use a shared representation for the policy and value function, especially when training directly from images. However, we argue that more information is needed to accurately estimate the value function than to learn the optimal policy. Consequently, the use of a shared representation for the policy and value function can lead to overfitting. To alleviate this problem, we propose two approaches which are combined to create IDAAC: Invariant Decoupled Advantage Actor-Critic. First, IDAAC decouples the optimization of the policy and value function, using separate networks to model them. Second, it introduces an auxiliary loss which encourages the representation to be invariant to task-irrelevant properties of the environment. IDAAC shows good generalization to unseen environments, achieving a new state-of-the-art on the Procgen benchmark and outperforming popular methods on DeepMind Control tasks with distractors. Our implementation is available at https://github.com/rraileanu/idaac."
                },
                {
                    "title": "Breaking the Deadly Triad with a Target Network",
                    "abstract": "The deadly triad refers to the instability of a reinforcement learning algorithm when it employs off-policy learning, function approximation, and bootstrapping simultaneously. In this paper, we investigate the target network as a tool for breaking the deadly triad, providing theoretical support for the conventional wisdom that a target network stabilizes training. We first propose and analyze a novel target network update rule which augments the commonly used Polyak-averaging style update with two projections. We then apply the target network and ridge regularization in several divergent algorithms and show their convergence to regularized TD fixed points. Those algorithms are off-policy with linear function approximation and bootstrapping, spanning both policy evaluation and control, as well as both discounted and average-reward settings. In particular, we provide the first convergent linear $Q$-learning algorithms under nonrestrictive and changing behavior policies without bi-level optimization."
                },
                {
                    "title": "Randomized Ensembled Double Q-Learning: Learning Fast Without a Model",
                    "abstract": "Using a high Update-To-Data (UTD) ratio, model-based methods have recently achieved much higher sample efficiency than previous model-free methods for continuous-action DRL benchmarks. In this paper, we introduce a simple model-free algorithm, Randomized Ensembled Double Q-Learning (REDQ), and show that its performance is just as good as, if not better than, a state-of-the-art model-based algorithm for the MuJoCo benchmark. Moreover, REDQ can achieve this performance using fewer parameters than the model-based method, and with less wall-clock run time. REDQ has three carefully integrated ingredients which allow it to achieve its high performance: (i) a UTD ratio>>1; (ii) an ensemble of Q functions; (iii) in-target minimization across a random subset of Q functions from the ensemble. Through carefully designed experiments, we provide a detailed analysis of REDQ and related model-free algorithms. To our knowledge, REDQ is the first successful model-free DRL algorithm for continuous-action spaces using a UTD ratio>>1."
                },
                {
                    "title": "Implicit Under-Parameterization Inhibits Data-Efficient Deep Reinforcement Learning",
                    "abstract": "We identify an implicit under-parameterization phenomenon in value-based deep RL methods that use bootstrapping: when value functions, approximated using deep neural networks, are trained with gradient descent using iterated regression onto target values generated by previous instances of the value network, more gradient updates decrease the expressivity of the current value network. We characterize this loss of expressivity in terms of a drop in the rank of the learned value network features, and show that this corresponds to a drop in performance. We demonstrate this phenomenon on widely studies domains, including Atari and Gym benchmarks, in both offline and online RL settings. We formally analyze this phenomenon and show that it results from a pathological interaction between bootstrapping and gradient-based optimization. We further show that mitigating implicit under-parameterization by controlling rank collapse improves performance."
                },
                {
                    "title": "Phasic Policy Gradient",
                    "abstract": "We introduce Phasic Policy Gradient (PPG), a reinforcement learning framework which modifies traditional on-policy actor-critic methods by separating policy and value function training into distinct phases. In prior methods, one must choose between using a shared network or separate networks to represent the policy and value function. Using separate networks avoids interference between objectives, while using a shared network allows useful features to be shared. PPG is able to achieve the best of both worlds by splitting optimization into two phases, one that advances training and one that distills features. PPG also enables the value function to be more aggressively optimized with a higher level of sample reuse. Compared to PPO, we find that PPG significantly improves sample efficiency on the challenging Procgen Benchmark."
                },
                {
                    "title": "Towards a practical measure of interference for reinforcement learning",
                    "abstract": "Catastrophic interference is common in many network-based learning systems, and many proposals exist for mitigating it. But, before we overcome interference we must understand it better. In this work, we provide a definition of interference for control in reinforcement learning. We systematically evaluate our new measures, by assessing correlation with several measures of learning performance, including stability, sample efficiency, and online and offline control performance across a variety of learning architectures. Our new interference measure allows us to ask novel scientific questions about commonly used deep learning architectures. In particular we show that target network frequency is a dominating factor for interference, and that updates on the last layer result in significantly higher interference than updates internal to the network. This new measure can be expensive to compute; we conclude with motivation for an efficient proxy measure and empirically demonstrate it is correlated with our definition of interference."
                },
                {
                    "title": "Learning Invariant Representations for Reinforcement Learning without Reconstruction",
                    "abstract": "We study how representation learning can accelerate reinforcement learning from rich observations, such as images, without relying either on domain knowledge or pixel-reconstruction. Our goal is to learn representations that both provide for effective downstream control and invariance to task-irrelevant details. Bisimulation metrics quantify behavioral similarity between states in continuous MDPs, which we propose using to learn robust latent representations which encode only the task-relevant information from observations. Our method trains encoders such that distances in latent space equal bisimulation distances in state space. We demonstrate the effectiveness of our method at disregarding task-irrelevant information using modified visual MuJoCo tasks, where the background is replaced with moving distractors and natural videos, while achieving SOTA performance. We also test a first-person highway driving task where our method learns invariance to clouds, weather, and time of day. Finally, we provide generalization results drawn from properties of bisimulation metrics, and links to causal inference."
                },
                {
                    "title": "Accelerating Online Reinforcement Learning with Offline Datasets",
                    "abstract": "Reinforcement learning provides an appealing formalism for learning control policies from experience. However, the classic active formulation of reinforcement learning necessitates a lengthy active exploration process for each behavior, making it difficult to apply in real-world settings. If we can instead allow reinforcement learning to effectively use previously collected data to aid the online learning process, where the data could be expert demonstrations or more generally any prior experience, we could make reinforcement learning a substantially more practical tool. While a number of recent methods have sought to learn offline from previously collected data, it remains exceptionally difficult to train a policy with offline data and improve it further with online reinforcement learning. In this paper we systematically analyze why this problem is so challenging, and propose a novel algorithm that combines sample-efficient dynamic programming with maximum likelihood policy updates, providing a simple and effective framework that is able to leverage large amounts of offline data and then quickly perform online fine-tuning of reinforcement learning policies. We show that our method enables rapid learning of skills with a combination of prior demonstration data and online experience across a suite of difficult dexterous manipulation and benchmark tasks."
                },
                {
                    "title": "Transient Non-stationarity and Generalisation in Deep Reinforcement Learning",
                    "abstract": "Non-stationarity can arise in Reinforcement Learning (RL) even in stationary environments. For example, most RL algorithms collect new data throughout training, using a non-stationary behaviour policy. Due to the transience of this non-stationarity, it is often not explicitly addressed in deep RL and a single neural network is continually updated. However, we find evidence that neural networks exhibit a memory effect where these transient non-stationarities can permanently impact the latent representation and adversely affect generalisation performance. Consequently, to improve generalisation of deep RL agents, we propose Iterated Relearning (ITER). ITER augments standard RL training by repeated knowledge transfer of the current policy into a freshly initialised network, which thereby experiences less non-stationarity during training. Experimentally, we show that ITER improves performance on the challenging generalisation benchmarks ProcGen and Multiroom."
                },
                {
                    "title": "The Value-Improvement Path: Towards Better Representations for Reinforcement Learning",
                    "abstract": "In value-based reinforcement learning (RL), unlike in supervised learning, the agent faces not a single, stationary, approximation problem, but a sequence of value prediction problems. Each time the policy improves, the nature of the problem changes, shifting both the distribution of states and their values. In this paper we take a novel perspective, arguing that the value prediction problems faced by an RL agent should not be addressed in isolation, but rather as a single, holistic, prediction problem. An RL algorithm generates a sequence of policies that, at least approximately, improve towards the optimal policy. We explicitly characterize the associated sequence of value functions and call it the value-improvement path. Our main idea is to approximate the value-improvement path holistically, rather than to solely track the value function of the current policy. Specifically, we discuss the impact that this holistic view of RL has on representation learning. We demonstrate that a representation that spans the past value-improvement path will also provide an accurate value approximation for future policy improvements. We use this insight to better understand existing approaches to auxiliary tasks and to propose new ones. To test our hypothesis empirically, we augmented a standard deep RL agent with an auxiliary task of learning the value-improvement path. In a study of Atari 2600 games, the augmented agent achieved approximately double the mean and median performance of the baseline agent."
                },
                {
                    "title": "Controlling Overestimation Bias with Truncated Mixture of Continuous Distributional Quantile Critics",
                    "abstract": "The overestimation bias is one of the major impediments to accurate off-policy learning. This paper investigates a novel way to alleviate the overestimation bias in a continuous control setting. Our method---Truncated Quantile Critics, TQC,---blends three ideas: distributional representation of a critic, truncation of critics prediction, and ensembling of multiple critics. Distributional representation and truncation allow for arbitrary granular overestimation control, while ensembling provides additional score improvements. TQC outperforms the current state of the art on all environments from the continuous control benchmark suite, demonstrating 25% improvement on the most challenging Humanoid environment."
                },
                {
                    "title": "D4RL: Datasets for Deep Data-Driven Reinforcement Learning",
                    "abstract": "The offline reinforcement learning (RL) problem, also known as batch RL, refers to the setting where a policy must be learned from a static dataset, without additional online data collection. This setting is compelling as potentially it allows RL methods to take advantage of large, pre-collected datasets, much like how the rise of large datasets has fueled results in supervised learning in recent years. However, existing online RL benchmarks are not tailored towards the offline setting, making progress in offline RL difficult to measure. In this work, we introduce benchmarks specifically designed for the offline setting, guided by key properties of datasets relevant to real-world applications of offline RL. Examples of such properties include: datasets generated via hand-designed controllers and human demonstrators, multi-objective datasets where an agent can perform different tasks in the same environment, and datasets consisting of a mixtures of policies. To facilitate research, we release our benchmark tasks and datasets with a comprehensive evaluation of existing algorithms and an evaluation protocol together with an open-source codebase. We hope that our benchmark will focus research effort on methods that drive improvements not just on simulated tasks, but ultimately on the kinds of real-world problems where offline RL will have the largest impact."
                },
                {
                    "title": "Improving Performance in Reinforcement Learning by Breaking Generalization in Neural Networks",
                    "abstract": "Reinforcement learning systems require good representations to work well. For decades practical success in reinforcement learning was limited to small domains. Deep reinforcement learning systems, on the other hand, are scalable, not dependent on domain specific prior knowledge and have been successfully used to play Atari, in 3D navigation from pixels, and to control high degree of freedom robots. Unfortunately, the performance of deep reinforcement learning systems is sensitive to hyper-parameter settings and architecture choices. Even well tuned systems exhibit significant instability both within a trial and across experiment replications. In practice, significant expertise and trial and error are usually required to achieve good performance. One potential source of the problem is known as catastrophic interference: when later training decreases performance by overriding previous learning. Interestingly, the powerful generalization that makes Neural Networks (NN) so effective in batch supervised learning might explain the challenges when applying them in reinforcement learning tasks. In this paper, we explore how online NN training and interference interact in reinforcement learning. We find that simply re-mapping the input observations to a high-dimensional space improves learning speed and parameter sensitivity. We also show this preprocessing reduces interference in prediction tasks. More practically, we provide a simple approach to NN training that is easy to implement, and requires little additional computation. We demonstrate that our approach improves performance in both prediction and control with an extensive batch of experiments in classic control domains."
                },
                {
                    "title": "DisCor: Corrective Feedback in Reinforcement Learning via Distribution Correction",
                    "abstract": "Deep reinforcement learning can learn effective policies for a wide range of tasks, but is notoriously difficult to use due to instability and sensitivity to hyperparameters. The reasons for this remain unclear. When using standard supervised methods (e.g., for bandits), on-policy data collection provides \"hard negatives\" that correct the model in precisely those states and actions that the policy is likely to visit. We call this phenomenon \"corrective feedback.\" We show that bootstrapping-based Q-learning algorithms do not necessarily benefit from this corrective feedback, and training on the experience collected by the algorithm is not sufficient to correct errors in the Q-function. In fact, Q-learning and related methods can exhibit pathological interactions between the distribution of experience collected by the agent and the policy induced by training on that experience, leading to potential instability, sub-optimal convergence, and poor results when learning from noisy, sparse or delayed rewards. We demonstrate the existence of this problem, both theoretically and empirically. We then show that a specific correction to the data distribution can mitigate this issue. Based on these observations, we propose a new algorithm, DisCor, which computes an approximation to this optimal distribution and uses it to re-weight the transitions used for training, resulting in substantial improvements in a range of challenging RL settings, such as multi-task learning and learning from noisy reward signals. Blog post presenting a summary of this work is available at: this https URL."
                },
                {
                    "title": "Interference and Generalization in Temporal Difference Learning",
                    "abstract": "We study the link between generalization and interference in temporal-difference (TD) learning. Interference is defined as the inner product of two different gradients, representing their alignment. This quantity emerges as being of interest from a variety of observations about neural networks, parameter sharing and the dynamics of learning. We find that TD easily leads to low-interference, under-generalizing parameters, while the effect seems reversed in supervised learning. We hypothesize that the cause can be traced back to the interplay between the dynamics of interference and bootstrapping. This is supported empirically by several observations: the negative relationship between the generalization gap and interference in TD, the negative effect of bootstrapping on interference and the local coherence of targets, and the contrast between the propagation rate of information in TD(0) versus TD($\\lambda$) and regression tasks such as Monte-Carlo policy evaluation. We hope that these new findings can guide the future discovery of better bootstrapping methods."
                },
                {
                    "title": "Can Increasing Input Dimensionality Improve Deep Reinforcement Learning?",
                    "abstract": "Deep reinforcement learning (RL) algorithms have recently achieved remarkable successes in various sequential decision making tasks, leveraging advances in methods for training large deep networks. However, these methods usually require large amounts of training data, which is often a big problem for real-world applications. One natural question to ask is whether learning good representations for states and using larger networks helps in learning better policies. In this paper, we try to study if increasing input dimensionality helps improve performance and sample efficiency of model-free deep RL algorithms. To do so, we propose an online feature extractor network (OFENet) that uses neural nets to produce good representations to be used as inputs to deep RL algorithms. Even though the high dimensionality of input is usually supposed to make learning of RL agents more difficult, we show that the RL agents in fact learn more efficiently with the high-dimensional representation than with the lower-dimensional state observations. We believe that stronger feature propagation together with larger networks (and thus larger search space) allows RL agents to learn more complex functions of states and thus improves the sample efficiency. Through numerical experiments, we show that the proposed method outperforms several other state-of-the-art algorithms in terms of both sample efficiency and performance. Codes for the proposed method are available at this http URL ."
                },
                {
                    "title": "ConQUR: Mitigating Delusional Bias in Deep Q-learning",
                    "abstract": "Delusional bias is a fundamental source of error in approximate Q-learning. To date, the only techniques that explicitly address delusion require comprehensive search using tabular value estimates. In this paper, we develop efficient methods to mitigate delusional bias by training Q-approximators with labels that are \"consistent\" with the underlying greedy policy class. We introduce a simple penalization scheme that encourages Q-labels used across training batches to remain (jointly) consistent with the expressible policy class. We also propose a search framework that allows multiple Q-approximators to be generated and tracked, thus mitigating the effect of premature (implicit) policy commitments. Experimental results demonstrate that these methods can improve the performance of Q-learning in a variety of Atari games, sometimes dramatically."
                },
                {
                    "title": "GenDICE: Generalized Offline Estimation of Stationary Values",
                    "abstract": "An important problem that arises in reinforcement learning and Monte Carlo methods is estimating quantities defined by the stationary distribution of a Markov chain. In many real-world applications, access to the underlying transition operator is limited to a fixed set of data that has already been collected, without additional interaction with the environment being available. We show that consistent estimation remains possible in this scenario, and that effective estimation can still be achieved in important applications. Our approach is based on estimating a ratio that corrects for the discrepancy between the stationary and empirical distributions, derived from fundamental properties of the stationary distribution, and exploiting constraint reformulations based on variational divergence minimization. The resulting algorithm, GenDICE, is straightforward and effective. We prove the consistency of the method under general conditions, provide a detailed error analysis, and demonstrate strong empirical performance on benchmark tasks, including off-line PageRank and off-policy policy evaluation."
                },
                {
                    "title": "Maxmin Q-learning: Controlling the Estimation Bias of Q-learning",
                    "abstract": "Q-learning suffers from overestimation bias, because it approximates the maximum action value using the maximum estimated action value. Algorithms have been proposed to reduce overestimation bias, but we lack an understanding of how bias interacts with performance, and the extent to which existing algorithms mitigate bias. In this paper, we 1) highlight that the effect of overestimation bias on learning efficiency is environment-dependent; 2) propose a generalization of Q-learning, called \\emph{Maxmin Q-learning}, which provides a parameter to flexibly control bias; 3) show theoretically that there exists a parameter choice for Maxmin Q-learning that leads to unbiased estimation with a lower approximation variance than Q-learning; and 4) prove the convergence of our algorithm in the tabular case, as well as convergence of several previous Q-learning variants, using a novel Generalized Q-learning framework. We empirically verify that our algorithm better controls estimation bias in toy environments, and that it achieves superior performance on several benchmark problems."
                },
                {
                    "title": "AlgaeDICE: Policy Gradient from Arbitrary Experience",
                    "abstract": "In many real-world applications of reinforcement learning (RL), interactions with the environment are limited due to cost or feasibility. This presents a challenge to traditional RL algorithms since the max-return objective involves an expectation over on-policy samples. We introduce a new formulation of max-return optimization that allows the problem to be re-expressed by an expectation over an arbitrary behavior-agnostic and off-policy data distribution. We first derive this result by considering a regularized version of the dual max-return objective before extending our findings to unregularized objectives through the use of a Lagrangian formulation of the linear programming characterization of Q-values. We show that, if auxiliary dual variables of the objective are optimized, then the gradient of the off-policy objective is exactly the on-policy policy gradient, without any use of importance weighting. In addition to revealing the appealing theoretical properties of this approach, we also show that it delivers good practical performance."
                },
                {
                    "title": "Towards Characterizing Divergence in Deep Q-Learning",
                    "abstract": "Deep Q-Learning (DQL), a family of temporal difference algorithms for control, employs three techniques collectively known as the `deadly triad' in reinforcement learning: bootstrapping, off-policy learning, and function approximation. Prior work has demonstrated that together these can lead to divergence in Q-learning algorithms, but the conditions under which divergence occurs are not well-understood. In this note, we give a simple analysis based on a linear approximation to the Q-value updates, which we believe provides insight into divergence under the deadly triad. The central point in our analysis is to consider when the leading order approximation to the deep-Q update is or is not a contraction in the sup norm. Based on this analysis, we develop an algorithm which permits stable deep Q-learning for continuous control without any of the tricks conventionally used (such as target networks, adaptive gradient optimizers, or using multiple Q functions). We demonstrate that our algorithm performs above or near state-of-the-art on standard MuJoCo benchmarks from the OpenAI Gym."
                },
                {
                    "title": "MinAtar: An Atari-inspired Testbed for More Efficient Reinforcement Learning Experiments",
                    "abstract": "The Arcade Learning Environment (ALE) is a popular platform for evaluating reinforcement learning agents. Much of the appeal comes from the fact that Atari games are varied, showcase aspects of competency we expect from an intelligent agent, and are not biased towards any particular solution approach. The challenge of the ALE includes 1) the representation learning problem of extracting pertinent information from the raw pixels, and 2) the behavioural learning problem of leveraging complex, delayed associations between actions and rewards. Often, in reinforcement learning research, we care more about the latter, but the representation learning problem adds significant computational expense. In response, we introduce MinAtar, short for miniature Atari, a new evaluation platform that captures the general mechanics of specific Atari games, while simplifying certain aspects. In particular, we reduce the representational complexity to focus more on behavioural challenges. MinAtar consists of analogues to five Atari games which play out on a 10x10 grid. MinAtar provides a 10x10xn state representation. The n channels correspond to game-specific objects, such as ball, paddle and brick in the game Breakout. While significantly simplified, these domains are still rich enough to allow for interesting behaviours. To demonstrate the challenges posed by these domains, we evaluated a smaller version of the DQN architecture. We also tried variants of DQN without experience replay, and without a target network, to assess the impact of those two prominent components in the MinAtar environments. In addition, we evaluated a simpler agent that used actor-critic with eligibility traces, online updating, and no experience replay. We hope that by introducing a set of simplified, Atari-like games we can allow researchers to more efficiently investigate the unique behavioural challenges provided by the ALE."
                },
                {
                    "title": "Deep Reinforcement Learning and the Deadly Triad",
                    "abstract": "We know from reinforcement learning theory that temporal difference learning can fail in certain cases. Sutton and Barto (2018) identify a deadly triad of function approximation, bootstrapping, and off-policy learning. When these three properties are combined, learning can diverge with the value estimates becoming unbounded. However, several algorithms successfully combine these three properties, which indicates that there is at least a partial gap in our understanding. In this work, we investigate the impact of the deadly triad in practice, in the context of a family of popular deep reinforcement learning models - deep Q-networks trained with experience replay - analysing how the components of this system play a role in the emergence of the deadly triad, and in the agent's performance"
                },
                {
                    "title": "Addressing Function Approximation Error in Actor-Critic Methods",
                    "abstract": "In value-based reinforcement learning methods such as deep Q-learning, function approximation errors are known to lead to overestimated value estimates and suboptimal policies. We show that this problem persists in an actor-critic setting and propose novel mechanisms to minimize its effects on both the actor and the critic. Our algorithm builds on Double Q-learning, by taking the minimum value between a pair of critics to limit overestimation. We draw the connection between target networks and overestimation bias, and suggest delaying policy updates to reduce per-update error and further improve performance. We evaluate our method on the suite of OpenAI gym tasks, outperforming the state of the art in every environment tested."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "DeepMind Control Suite",
                    "abstract": "The DeepMind Control Suite is a set of continuous control tasks with a standardised structure and interpretable rewards, intended to serve as performance benchmarks for reinforcement learning agents. The tasks are written in Python and powered by the MuJoCo physics engine, making them easy to use and modify. We include benchmarks for several learning algorithms. The Control Suite is publicly available at this https URL A video summary of all tasks is available at this http URL ."
                },
                {
                    "title": "Rainbow: Combining Improvements in Deep Reinforcement Learning",
                    "abstract": "\n \n The deep reinforcement learning community has made several independent improvements to the DQN algorithm. However, it is unclear which of these extensions are complementary and can be fruitfully combined. This paper examines six extensions to the DQN algorithm and empirically studies their combination. Our experiments show that the combination provides state-of-the-art performance on the Atari 2600 benchmark, both in terms of data efficiency and final performance. We also provide results from a detailed ablation study that shows the contribution of each component to overall performance.\n \n"
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "A Distributional Perspective on Reinforcement Learning",
                    "abstract": "In this paper we argue for the fundamental importance of the value distribution: the distribution of the random return received by a reinforcement learning agent. This is in contrast to the common approach to reinforcement learning which models the expectation of this return, or value. Although there is an established body of literature studying the value distribution, thus far it has always been used for a specific purpose such as implementing risk-aware behaviour. We begin with theoretical results in both the policy evaluation and control settings, exposing a significant distributional instability in the latter. We then use the distributional perspective to design a new algorithm which applies Bellman's equation to the learning of approximate value distributions. We evaluate our algorithm using the suite of games from the Arcade Learning Environment. We obtain both state-of-the-art results and anecdotal evidence demonstrating the importance of the value distribution in approximate reinforcement learning. Finally, we combine theoretical and empirical evidence to highlight the ways in which the value distribution impacts learning in the approximate setting."
                },
                {
                    "title": "Prioritized Experience Replay",
                    "abstract": "Experience replay lets online reinforcement learning agents remember and reuse experiences from the past. In prior work, experience transitions were uniformly sampled from a replay memory. However, this approach simply replays transitions at the same frequency that they were originally experienced, regardless of their significance. In this paper we develop a framework for prioritizing experience, so as to replay important transitions more frequently, and therefore learn more efficiently. We use prioritized experience replay in Deep Q-Networks (DQN), a reinforcement learning algorithm that achieved human-level performance across many Atari games. DQN with prioritized experience replay achieves a new state-of-the-art, outperforming DQN with uniform replay on 41 out of 49 games."
                },
                {
                    "title": "Deep Reinforcement Learning with Double Q-Learning",
                    "abstract": "\n \n The popular Q-learning algorithm is known to overestimate action values under certain conditions. It was not previously known whether, in practice, such overestimations are common, whether they harm performance, and whether they can generally be prevented. In this paper, we answer all these questions affirmatively. In particular, we first show that the recent DQN algorithm, which combines Q-learning with a deep neural network, suffers from substantial overestimations in some games in the Atari 2600 domain. We then show that the idea behind the Double Q-learning algorithm, which was introduced in a tabular setting, can be generalized to work with large-scale function approximation. We propose a specific adaptation to the DQN algorithm and show that the resulting algorithm not only reduces the observed overestimations, as hypothesized, but that this also leads to much better performance on several games.\n \n"
                },
                {
                    "title": "Continuous control with deep reinforcement learning",
                    "abstract": "We adapt the ideas underlying the success of Deep Q-Learning to the continuous action domain. We present an actor-critic, model-free algorithm based on the deterministic policy gradient that can operate over continuous action spaces. Using the same learning algorithm, network architecture and hyper-parameters, our algorithm robustly solves more than 20 simulated physics tasks, including classic problems such as cartpole swing-up, dexterous manipulation, legged locomotion and car driving. Our algorithm is able to find policies whose performance is competitive with those found by a planning algorithm with full access to the dynamics of the domain and its derivatives. We further demonstrate that for many of the tasks the algorithm can learn policies end-to-end: directly from raw pixel inputs."
                },
                {
                    "title": "Trust Region Policy Optimization",
                    "abstract": "In this article, we describe a method for optimizing control policies, with guaranteed monotonic improvement. By making several approximations to the theoretically-justified scheme, we develop a practical algorithm, called Trust Region Policy Optimization (TRPO). This algorithm is effective for optimizing large nonlinear policies such as neural networks. Our experiments demonstrate its robust performance on a wide variety of tasks: learning simulated robotic swimming, hopping, and walking gaits; and playing Atari games using images of the screen as input. Despite its approximations that deviate from the theory, TRPO tends to give monotonic improvement, with little tuning of hyperparameters."
                },
                {
                    "title": "Deterministic Policy Gradient Algorithms",
                    "abstract": "In this paper we consider deterministic policy gradient algorithms for reinforcement learning with continuous actions. The deterministic policy gradient has a particularly appealing form: it is the expected gradient of the action-value function. This simple form means that the deterministic policy gradient can be estimated much more efficiently than the usual stochastic policy gradient. To ensure adequate exploration, we introduce an off-policy actor-critic algorithm that learns a deterministic target policy from an exploratory behaviour policy. We demonstrate that deterministic policy gradient algorithms can significantly outperform their stochastic counterparts in high-dimensional action spaces."
                },
                {
                    "title": "Reining Generalization in Offline Reinforcement Learning via Representation Distinction",
                    "abstract": "Offline Reinforcement Learning (RL) aims to address the challenge of distribution shift between the dataset and the learned policy, where the value of out-of-distribution (OOD) data may be erroneously estimated due to overgeneralization. It has been observed that a considerable portion of the benefits derived from the conservative terms designed by existing offline RL approaches originates from their impact on the learned representation. This observation prompts us to scrutinize the learning dynamics of offline RL, formalize the process of generalization, and delve into the prevalent overgeneralization issue in offline RL. We then investigate the potential to rein the generalization from the representation perspective to enhance offline RL. Finally, we present Representation Distinction (RD), an innovative plug-in method for improving offline RL algorithm performance by explicitly differentiating between the representations of in-sample and OOD state-action pairs generated by the learning policy. Considering scenarios in which the learning policy mirrors the behavioral policy and similar samples may be erroneously distinguished, we suggest a dynamic adjustment mechanism for RD based on an OOD data generator to prevent data representation collapse and further enhance policy performance. We demonstrate the efficacy of our approach by applying RD to designed backbone algorithms and widely-used offline RL algorithms. The proposed RD method significantly improves their performance across various continuous control tasks on D4RL datasets, surpassing several state-of-the-art offline RL algorithms."
                },
                {
                    "title": "Sample-Efficient Reinforcement Learning by Breaking the Replay Ratio Barrier",
                    "abstract": "Increasing the"
                },
                {
                    "title": "Learning Dynamics and Generalization in Deep Reinforcement Learning",
                    "abstract": "Solving a reinforcement learning (RL) problem poses two competing challenges: fitting a potentially discontinuous value function, and generalizing well to new observations. In this paper, we analyze the learning dynamics of temporal difference algorithms to gain novel insight into the tension between these two objectives. We show theoretically that temporal difference learning encourages agents to fit non-smooth components of the value function early in training, and at the same time induces the second-order effect of discouraging generalization. We corroborate these findings in deep RL agents trained on a range of environments, finding that neural networks trained using temporal difference algorithms on dense reward tasks exhibit weaker generalization between states than randomly initialized networks and networks trained with policy gradient methods. Finally, we investigate how post-training policy distillation may avoid this pit-fall, and show that this approach improves generalization to novel environments in the ProcGen suite and improves robustness to input perturbations."
                },
                {
                    "title": "CleanRL: High-quality Single-file Implementations of Deep Reinforcement Learning Algorithms",
                    "abstract": "CleanRL is an open-source library that provides high-quality single-\ufb01le implementations of Deep Reinforcement Learning (DRL) algorithms. These single-\ufb01le implementations are self-contained algorithm variant \ufb01les such as dqn.py , ppo.py , and ppo atari.py that individually include all algorithm variant\u2019s implementation details. Such a paradigm signi\ufb01cantly reduces the complexity and the lines of code (LOC) in each implemented variant, which makes them quicker and easier to understand. This paradigm gives the researchers the most \ufb01ne-grained control over all aspects of the algorithm in a single \ufb01le, allowing them to prototype novel features quickly. Despite having succinct implementations, CleanRL\u2019s codebase is thoroughly documented and benchmarked to ensure performance is on par with reputable sources. As a result, CleanRL produces a repository tailor-\ufb01t for two purposes: 1) understanding all implementation details of DRL algorithms and 2) quickly prototyping novel features. CleanRL\u2019s source code can be found at https://github.com/vwxyzjn/cleanrl ."
                },
                {
                    "title": "Fuzzy Tiling Activations: A Simple Approach to Learning Sparse Representations Online",
                    "abstract": "Recent work has shown that sparse representations\u2014where only a small percentage of units are active\u2014can signi\ufb01cantly reduce interference. Those works, however, relied on relatively complex regularization or meta-learning approaches, that have only been used of\ufb02ine in a pre-training phase. In this work, we pursue a direction that achieves sparsity by design, rather than by learning. Speci\ufb01cally, we design an activation function that produces sparse representations deterministically by construction, and so is more amenable to online training. The idea relies on the simple approach of binning, but overcomes the two key limitations of binning: zero gradients for the \ufb02at regions almost everywhere, and lost precision\u2014reduced discrimination\u2014due to coarse aggregation. We introduce a Fuzzy Tiling Activation (FTA) that provides non-negligible gradients and produces overlap between bins that improves discrimination. We \ufb01rst show that FTA is robust under covariate shift in a synthetic online supervised learning problem, where we can vary the level of correlation and drift. Then we move to the deep reinforcement learning setting and investigate both value-based and policy gradient algorithms that use neural networks with FTAs, in classic discrete control and Mujoco continuous control environments. We show that algorithms equipped with FTAs are able to learn a stable policy faster without needing target networks on most domains. 1"
                },
                {
                    "title": "Non-delusional Q-learning and value-iteration",
                    "abstract": "We identify a fundamental source of error in Q-learning and other forms of dynamic programming with function approximation. Delusional bias arises when the approximation architecture limits the class of expressible greedy policies. Since standard Q-updates make globally uncoordinated action choices with respect to the expressible policy class, inconsistent or even conflicting Q-value estimates can result, leading to pathological behaviour such as over/under-estimation, instability and even divergence. To solve this problem, we introduce a new notion of policy consistency and define a local backup process that ensures global consistency through the use of information sets---sets that record constraints on policies consistent with backed-up Q-values. We prove that both the model-based and model-free algorithms using this backup remove delusional bias, yielding the first known algorithms that guarantee optimal results under general conditions. These algorithms furthermore only require polynomially many information sets (from a potentially exponential support). Finally, we suggest other practical heuristics for value-iteration and Q-learning that attempt to reduce delusional bias."
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does the phenomenon of churn in Deep Reinforcement Learning (DRL) affect learning dynamics and performance across different DRL settings?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding and controlling churn in DRL is crucial because it directly impacts the stability and efficiency of learning algorithms. By addressing this problem, the research community can enhance the performance of various DRL methods, leading to more reliable and effective applications in real-world scenarios such as robotics, game playing, and autonomous systems. This work could pave the way for future research focused on improving learning dynamics and developing more robust algorithms that can handle non-stationary environments.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complex interactions between policy improvement and value estimation in DRL, which can lead to a chain effect of churn that biases learning dynamics. Naive approaches may fail because they do not account for the indirect changes in outputs caused by updates to the policy and value networks. Additionally, the non-stationary nature of RL environments complicates the identification and control of churn, requiring sophisticated methods to minimize its impact without sacrificing learning efficiency.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on understanding individual aspects of learning dynamics in DRL, but the specific phenomenon of churn and its chain effects have not been thoroughly investigated. Existing solutions often overlook the interplay between policy and value churn, leading to incomplete understanding and inadequate control mechanisms. Our approach differs by formally characterizing churn within the framework of Generalized Policy Iteration and proposing a targeted method (CHAIN) to mitigate its effects across various DRL settings.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Churn Approximated ReductIoN (CHAIN), aims to minimize the undesirable changes in policy and value outputs for states outside the current batch during DRL training. We will evaluate CHAIN using a range of environments, including MinAtar, and measure its effectiveness through metrics such as learning stability, convergence rates, and overall performance improvements. The expected outcomes include a clearer understanding of churn dynamics and enhanced performance of DRL algorithms, demonstrating the practical applicability of our approach."
            }
        },
        "author_data": {
            "14ee9be4-53b1-4b7d-88b8-f79773316e0f": {
                "pk": "14ee9be4-53b1-4b7d-88b8-f79773316e0f",
                "name": "Hongyao Tang",
                "collaborators": [
                    "Jianye Hao",
                    "Yan Zheng",
                    "Zhaopeng Meng",
                    "Chen Chen",
                    "Wulong Liu",
                    "Guangyong Chen",
                    "Yaodong Yang",
                    "Min Zhang",
                    "Li Wang",
                    "Haotian Fu"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Multi-Agent Systems",
                    "Deep Learning",
                    "Policy Optimization"
                ],
                "publications": [
                    {
                        "title": "The Ladder in Chaos: A Simple and Effective Improvement to General DRL Algorithms by Policy Path Trimming and Boosting",
                        "abstract": "Knowing the learning dynamics of policy is significant to unveiling the mysteries of Reinforcement Learning (RL). It is especially crucial yet challenging to Deep RL, from which the remedies to notorious issues like sample inefficiency and learning instability could be obtained. In this paper, we study how the policy networks of typical DRL agents evolve during the learning process by empirically investigating several kinds of temporal change for each policy parameter. On typical MuJoCo and DeepMind Control Suite (DMC) benchmarks, we find common phenomena for TD3 and RAD agents: 1) the activity of policy network parameters is highly asymmetric and policy networks advance monotonically along very few major parameter directions; 2) severe detours occur in parameter update and harmonic-like changes are observed for all minor parameter directions. By performing a novel temporal SVD along policy learning path, the major and minor parameter directions are identified as the columns of right unitary matrix associated with dominant and insignificant singular values respectively. Driven by the discoveries above, we propose a simple and effective method, called Policy Path Trimming and Boosting (PPTB), as a general plug-in improvement to DRL algorithms. The key idea of PPTB is to periodically trim the policy learning path by canceling the policy updates in minor parameter directions, while boost the learning path by encouraging the advance in major directions. In experiments, we demonstrate the general and significant performance improvements brought by PPTB, when combined with TD3 and RAD in MuJoCo and DMC environments respectively."
                    },
                    {
                        "title": "Towards A Unified Policy Abstraction Theory and Representation Learning Approach in Markov Decision Processes",
                        "abstract": "Lying on the heart of intelligent decision-making systems, how policy is represented and optimized is a fundamental problem. The root challenge in this problem is the large scale and the high complexity of policy space, which exacerbates the difficulty of policy learning especially in real-world scenarios. Towards a desirable surrogate policy space, recently policy representation in a low-dimensional latent space has shown its potential in improving both the evaluation and optimization of policy. The key question involved in these studies is by what criterion we should abstract the policy space for desired compression and generalization. However, both the theory on policy abstraction and the methodology on policy representation learning are less studied in the literature. In this work, we make very first efforts to fill up the vacancy. First, we propose a unified policy abstraction theory, containing three types of policy abstraction associated to policy features at different levels. Then, we generalize them to three policy metrics that quantify the distance (i.e., similarity) of policies, for more convenient use in learning policy representation. Further, we propose a policy representation learning approach based on deep metric learning. For the empirical study, we investigate the efficacy of the proposed policy metrics and representations, in characterizing policy difference and conveying policy generalization respectively. Our experiments are conducted in both policy optimization and evaluation problems, containing trust-region policy optimization (TRPO), diversity-guided evolution strategy (DGES) and off-policy evaluation (OPE). Somewhat naturally, the experimental results indicate that there is no a universally optimal abstraction for all downstream learning problems; while the influence-irrelevance policy abstraction can be a generally preferred choice."
                    },
                    {
                        "title": "An Optimal Rewiring Strategy for Reinforcement Social Learning in Cooperative Multiagent Systems",
                        "abstract": "Multiagent coordination in cooperative multiagent systems (MASs) has been widely studied in both fixed-agent repeated interaction setting and the static social learning framework. However, two aspects of dynamics in real-world multiagent scenarios are currently missing in existing works. First, the network topologies can be dynamic where agents may change their connections through rewiring during the course of interactions. Second, the game matrix between each pair of agents may not be static and usually not known as a prior. Both the network dynamic and game uncertainty increase the coordination difficulty among agents. In this paper, we consider a multiagent dynamic social learning environment in which each agent can choose to rewire potential partners and interact with randomly chosen neighbors in each round. We propose an optimal rewiring strategy for agents to select most beneficial peers to interact with for the purpose of maximizing the accumulated payoff in repeated interactions. We empirically demonstrate the effectiveness and robustness of our approach through comparing with benchmark strategies. The performance of three representative learning strategies under our social learning framework with our optimal rewiring is investigated as well."
                    },
                    {
                        "title": "Addressing Action Oscillations through Learning Policy Inertia",
                        "abstract": "Deep reinforcement learning (DRL) algorithms have been demonstrated to be effective in a wide range of challenging decision making and control tasks. However, these methods typically suffer from severe action oscillations in particular in discrete action setting, which means that agents select different actions within consecutive steps even though states only slightly differ. This issue is often neglected since the policy is usually evaluated by its cumulative rewards only. Action oscillation strongly affects the user experience and can even cause serious potential security menace especially in real-world domains with the main concern of safety, such as autonomous driving. To this end, we introduce Policy Inertia Controller (PIC) which serves as a generic plug-in framework to off-the-shelf DRL algorithms, to enables adaptive trade-off between the optimality and smoothness of the learned policy in a formal way. We propose Nested Policy Iteration as a general training algorithm for PIC-augmented policy which ensures monotonically non-decreasing updates under some mild conditions. Further, we derive a practical DRL algorithm, namely Nested Soft Actor-Critic. Experiments on a collection of autonomous driving tasks and several Atari games suggest that our approach demonstrates substantial oscillation reduction in comparison to a range of commonly adopted baselines with almost no performance degradation."
                    },
                    {
                        "title": "Uncertainty-aware Low-Rank Q-Matrix Estimation for Deep Reinforcement Learning",
                        "abstract": "Value estimation is one key problem in Reinforcement Learning. Albeit many successes have been achieved by Deep Reinforcement Learning (DRL) in different fields, the underlying structure and learning dynamics of value function, especially with complex function approximation, are not fully understood. In this paper, we report that decreasing rank of $Q$-matrix widely exists during learning process across a series of continuous control tasks for different popular algorithms. We hypothesize that the low-rank phenomenon indicates the common learning dynamics of $Q$-matrix from stochastic high dimensional space to smooth low dimensional space. Moreover, we reveal a positive correlation between value matrix rank and value estimation uncertainty. Inspired by above evidence, we propose a novel Uncertainty-Aware Low-rank Q-matrix Estimation (UA-LQE) algorithm as a general framework to facilitate the learning of value function. Through quantifying the uncertainty of state-action value estimation, we selectively erase the entries of highly uncertain values in state-action value matrix and conduct low-rank matrix reconstruction for them to recover their values. Such a reconstruction exploits the underlying structure of value matrix to improve the value approximation, thus leading to a more efficient learning process of value function. In the experiments, we evaluate the efficacy of UA-LQE in several representative OpenAI MuJoCo continuous control tasks."
                    },
                    {
                        "title": "MGHRL: Meta Goal-generation for Hierarchical Reinforcement Learning",
                        "abstract": "Most meta reinforcement learning (meta-RL) methods learn to adapt to new tasks by directly optimizing the parameters of policies over primitive action space. Such algorithms work well in tasks with relatively slight difference. However, when the task distribution becomes wider, it would be quite inefficient to directly learn such a meta-policy. In this paper, we propose a new meta-RL algorithm called Meta Goal-generation for Hierarchical RL (MGHRL). Instead of directly generating policies over primitive action space for new tasks, MGHRL learns to generate high-level meta strategies over subgoals given past experience and leaves the rest of how to achieve subgoals as independent RL subtasks. Our empirical results on several challenging simulated robotics environments show that our method enables more efficient and generalized meta-learning from past experience."
                    },
                    {
                        "title": "Disentangling Dynamics and Returns: Value Function Decomposition with Future Prediction",
                        "abstract": "Value functions are crucial for model-free Reinforcement Learning (RL) to obtain a policy implicitly or guide the policy updates. Value estimation heavily depends on the stochasticity of environmental dynamics and the quality of reward signals. In this paper, we propose a two-step understanding of value estimation from the perspective of future prediction, through decomposing the value function into a reward-independent future dynamics part and a policy-independent trajectory return part. We then derive a practical deep RL algorithm from the above decomposition, consisting of a convolutional trajectory representation model, a conditional variational dynamics model to predict the expected representation of future trajectory and a convex trajectory return model that maps a trajectory representation to its return. Our algorithm is evaluated in MuJoCo continuous control tasks and shows superior results under both common settings and delayed reward settings."
                    },
                    {
                        "title": "Bridging Evolutionary Algorithms and Reinforcement Learning: A Comprehensive Survey on Hybrid Algorithms",
                        "abstract": "Evolutionary Reinforcement Learning (ERL), which integrates Evolutionary Algorithms (EAs) and Reinforcement Learning (RL) for optimization, has demonstrated remarkable performance advancements. By fusing both approaches, ERL has emerged as a promising research direction. This survey offers a comprehensive overview of the diverse research branches in ERL. Specifically, we systematically summarize recent advancements in related algorithms and identify three primary research directions: EA-assisted Optimization of RL, RL-assisted Optimization of EA, and synergistic optimization of EA and RL. Following that, we conduct an in-depth analysis of each research direction, organizing multiple research branches. We elucidate the problems that each branch aims to tackle and how the integration of EAs and RL addresses these challenges. In conclusion, we discuss potential challenges and prospective future research directions across various research directions. To facilitate researchers in delving into ERL, we organize the algorithms and codes involved on https://github.com/yeshenpy/Awesome-Evolutionary-Reinforcement-Learning."
                    },
                    {
                        "title": "Deep Multi-Agent Reinforcement Learning with Discrete-Continuous Hybrid Action Spaces",
                        "abstract": "Deep Reinforcement Learning (DRL) has been applied to address a variety of cooperative multi-agent problems with either discrete action spaces or continuous action spaces. However, to the best of our knowledge, no previous work has ever succeeded in applying DRL to multi-agent problems with discrete-continuous hybrid (or parameterized) action spaces which is very common in practice. Our work fills this gap by proposing two novel algorithms: Deep Multi-Agent Parameterized Q-Networks (Deep MAPQN) and Deep Multi-Agent Hierarchical Hybrid Q-Networks (Deep MAHHQN). We follow the centralized training but decentralized execution paradigm: different levels of communication between different agents are used to facilitate the training process, while each agent executes its policy independently based on local observations during execution. Our empirical results on several challenging tasks (simulated RoboCup Soccer and game Ghost Story) show that both Deep MAPQN and Deep MAHHQN are effective and significantly outperform existing independent deep parameterized Q-learning method."
                    },
                    {
                        "title": "Qatten: A General Framework for Cooperative Multiagent Reinforcement Learning",
                        "abstract": "In many real-world tasks, multiple agents must learn to coordinate with each other given their private observations and limited communication ability. Deep multiagent reinforcement learning (Deep-MARL) algorithms have shown superior performance in such challenging settings. One representative class of work is multiagent value decomposition, which decomposes the global shared multiagent Q-value $Q_{tot}$ into individual Q-values $Q^{i}$ to guide individuals' behaviors, i.e. VDN imposing an additive formation and QMIX adopting a monotonic assumption using an implicit mixing method. However, most of the previous efforts impose certain assumptions between $Q_{tot}$ and $Q^{i}$ and lack theoretical groundings. Besides, they do not explicitly consider the agent-level impact of individuals to the whole system when transforming individual $Q^{i}$s into $Q_{tot}$. In this paper, we theoretically derive a general formula of $Q_{tot}$ in terms of $Q^{i}$, based on which we can naturally implement a multi-head attention formation to approximate $Q_{tot}$, resulting in not only a refined representation of $Q_{tot}$ with an agent-level attention mechanism, but also a tractable maximization algorithm of decentralized policies. Extensive experiments demonstrate that our method outperforms state-of-the-art MARL methods on the widely adopted StarCraft benchmark across different scenarios, and attention analysis is further conducted with valuable insights."
                    },
                    {
                        "title": "KoGuN: Accelerating Deep Reinforcement Learning via Integrating Human Suboptimal Knowledge",
                        "abstract": "Reinforcement learning agents usually learn from scratch, which requires a large number of interactions with the environment. This is quite different from the learning process of human. When faced with a new task, human naturally have the common sense and use the prior knowledge to derive an initial policy and guide the learning process afterwards. Although the prior knowledge may be not fully applicable to the new task, the learning process is significantly sped up since the initial policy ensures a quick-start of learning and intermediate guidance allows to avoid unnecessary exploration. Taking this inspiration, we propose knowledge guided policy network (KoGuN), a novel framework that combines human prior suboptimal knowledge with reinforcement learning. Our framework consists of a fuzzy rule controller to represent human knowledge and a refine module to fine-tune suboptimal prior knowledge. The proposed framework is end-to-end and can be combined with existing policy-based reinforcement learning algorithm. We conduct experiments on both discrete and continuous control tasks. The empirical results show that our approach, which combines human suboptimal knowledge and RL, achieves significant improvement on learning efficiency of flat RL algorithms, even with very low-performance human prior knowledge."
                    },
                    {
                        "title": "Towards Effective Context for Meta-Reinforcement Learning: an Approach based on Contrastive Learning",
                        "abstract": "Context, the embedding of previous collected trajectories, is a powerful construct for Meta-Reinforcement Learning (Meta-RL) algorithms. By conditioning on an effective context, Meta-RL policies can easily generalize to new tasks within a few adaptation steps. We argue that improving the quality of context involves answering two questions: 1. How to train a compact and sufficient encoder that can embed the task-specific information contained in prior trajectories? 2. How to collect informative trajectories of which the corresponding context reflects the specification of tasks? To this end, we propose a novel Meta-RL framework called CCM (Contrastive learning augmented Context-based Meta-RL). We first focus on the contrastive nature behind different tasks and leverage it to train a compact and sufficient context encoder. Further, we train a separate exploration policy and theoretically derive a new information-gain-based objective which aims to collect informative trajectories in a few steps. Empirically, we evaluate our approaches on common benchmarks as well as several complex sparse-reward environments. The experimental results show that CCM outperforms state-of-the-art algorithms by addressing previously mentioned problems respectively."
                    },
                    {
                        "title": "ERL-Re$^2$: Efficient Evolutionary Reinforcement Learning with Shared State Representation and Individual Policy Representation",
                        "abstract": "Deep Reinforcement Learning (Deep RL) and Evolutionary Algorithms (EA) are two major paradigms of policy optimization with distinct learning principles, i.e., gradient-based v.s. gradient-free. An appealing research direction is integrating Deep RL and EA to devise new methods by fusing their complementary advantages. However, existing works on combining Deep RL and EA have two common drawbacks: 1) the RL agent and EA agents learn their policies individually, neglecting efficient sharing of useful common knowledge; 2) parameter-level policy optimization guarantees no semantic level of behavior evolution for the EA side. In this paper, we propose Evolutionary Reinforcement Learning with Two-scale State Representation and Policy Representation (ERL-Re$^2$), a novel solution to the aforementioned two drawbacks. The key idea of ERL-Re$^2$ is two-scale representation: all EA and RL policies share the same nonlinear state representation while maintaining individual} linear policy representations. The state representation conveys expressive common features of the environment learned by all the agents collectively; the linear policy representation provides a favorable space for efficient policy optimization, where novel behavior-level crossover and mutation operations can be performed. Moreover, the linear policy representation allows convenient generalization of policy fitness with the help of the Policy-extended Value Function Approximator (PeVFA), further improving the sample efficiency of fitness estimation. The experiments on a range of continuous control tasks show that ERL-Re$^2$ consistently outperforms advanced baselines and achieves the State Of The Art (SOTA). Our code is available on https://github.com/yeshenpy/ERL-Re2."
                    },
                    {
                        "title": "State-Aware Proximal Pessimistic Algorithms for Offline Reinforcement Learning",
                        "abstract": "Pessimism is of great importance in offline reinforcement learning (RL). One broad category of offline RL algorithms fulfills pessimism by explicit or implicit behavior regularization. However, most of them only consider policy divergence as behavior regularization, ignoring the effect of how the offline state distribution differs with that of the learning policy, which may lead to under-pessimism for some states and over-pessimism for others. Taking account of this problem, we propose a principled algorithmic framework for offline RL, called \\emph{State-Aware Proximal Pessimism} (SA-PP). The key idea of SA-PP is leveraging discounted stationary state distribution ratios between the learning policy and the offline dataset to modulate the degree of behavior regularization in a state-wise manner, so that pessimism can be implemented in a more appropriate way. We first provide theoretical justifications on the superiority of SA-PP over previous algorithms, demonstrating that SA-PP produces a lower suboptimality upper bound in a broad range of settings. Furthermore, we propose a new algorithm named \\emph{State-Aware Conservative Q-Learning} (SA-CQL), by building SA-PP upon representative CQL algorithm with the help of DualDICE for estimating discounted stationary state distribution ratios. Extensive experiments on standard offline RL benchmark show that SA-CQL outperforms the popular baselines on a large portion of benchmarks and attains the highest average return."
                    },
                    {
                        "title": "PAnDR: Fast Adaptation to New Environments from Offline Experiences via Decoupling Policy and Environment Representations",
                        "abstract": "Deep Reinforcement Learning (DRL) has been a promising solution to many complex decision-making problems. Nevertheless, the notorious weakness in generalization among environments prevent widespread application of DRL agents in real-world scenarios. Although advances have been made recently, most prior works assume sufficient online interaction on training environments, which can be costly in practical cases. To this end, we focus on an offline-training-online-adaptation setting, in which the agent first learns from offline experiences collected in environments with different dynamics and then performs online policy adaptation in environments with new dynamics. In this paper, we propose Policy Adaptation with Decoupled Representations (PAnDR) for fast policy adaptation. In offline training phase, the environment representation and policy representation are learned through contrastive learning and policy recovery, respectively. The representations are further refined by mutual information optimization to make them more decoupled and complete. With learned representations, a Policy-Dynamics Value Function (PDVF) [Raileanu et al., 2020] network is trained to approximate the values for different combinations of policies and environments from offline experiences. In online adaptation phase, with the environment context inferred from few experiences collected in new environments, the policy is optimized by gradient ascent with respect to the PDVF. Our experiments show that PAnDR outperforms existing algorithms in several representative policy adaptation problems."
                    },
                    {
                        "title": "Foresee then Evaluate: Decomposing Value Estimation with Latent Future Prediction",
                        "abstract": "Value function is the central notion of Reinforcement Learning (RL). Value estimation, especially with function approximation, can be challenging since it involves the stochasticity of environmental dynamics and reward signals that can be sparse and delayed in some cases. A typical model-free RL algorithm usually estimates the values of a policy by Temporal Difference (TD) or Monte Carlo (MC) algorithms directly from rewards, without explicitly taking dynamics into consideration. In this paper, we propose Value Decomposition with Future Prediction (VDFP), providing an explicit two-step understanding of the value estimation process: 1) first foresee the latent future, 2) and then evaluate it. We analytically decompose the value function into a latent future dynamics part and a policy-independent trajectory return part, inducing a way to model latent dynamics and returns separately in value estimation. Further, we derive a practical deep RL algorithm, consisting of a convolutional model to learn compact trajectory representation from past experiences, a conditional variational auto-encoder to predict the latent future dynamics and a convex return model that evaluates trajectory representation. In experiments, we empirically demonstrate the effectiveness of our approach for both off-policy and on-policy RL in several OpenAI Gym continuous control tasks as well as a few challenging variants with delayed reward."
                    },
                    {
                        "title": "HyAR: Addressing Discrete-Continuous Action Reinforcement Learning via Hybrid Action Representation",
                        "abstract": "Discrete-continuous hybrid action space is a natural setting in many practical problems, such as robot control and game AI. However, most previous Reinforcement Learning (RL) works only demonstrate the success in controlling with either discrete or continuous action space, while seldom take into account the hybrid action space. One naive way to address hybrid action RL is to convert the hybrid action space into a unified homogeneous action space by discretization or continualization, so that conventional RL algorithms can be applied. However, this ignores the underlying structure of hybrid action space and also induces the scalability issue and additional approximation difficulties, thus leading to degenerated results. In this paper, we propose Hybrid Action Representation (HyAR) to learn a compact and decodable latent representation space for the original hybrid action space. HyAR constructs the latent space and embeds the dependence between discrete action and continuous parameter via an embedding table and conditional Variantional Auto-Encoder (VAE). To further improve the effectiveness, the action representation is trained to be semantically smooth through unsupervised environmental dynamics prediction. Finally, the agent then learns its policy with conventional DRL algorithms in the learned representation space and interacts with the environment by decoding the hybrid action embeddings to the original action space. We evaluate HyAR in a variety of environments with discrete-continuous action space. The results demonstrate the superiority of HyAR when compared with previous baselines, especially for high-dimensional action spaces."
                    },
                    {
                        "title": "Q-value Path Decomposition for Deep Multiagent Reinforcement Learning",
                        "abstract": "Recently, deep multiagent reinforcement learning (MARL) has become a highly active research area as many real-world problems can be inherently viewed as multiagent systems. A particularly interesting and widely applicable class of problems is the partially observable cooperative multiagent setting, in which a team of agents learns to coordinate their behaviors conditioning on their private observations and commonly shared global reward signals. One natural solution is to resort to the centralized training and decentralized execution paradigm. During centralized training, one key challenge is the multiagent credit assignment: how to allocate the global rewards for individual agent policies for better coordination towards maximizing system-level's benefits. In this paper, we propose a new method called Q-value Path Decomposition (QPD) to decompose the system's global Q-values into individual agents' Q-values. Unlike previous works which restrict the representation relation of the individual Q-values and the global one, we leverage the integrated gradient attribution technique into deep MARL to directly decompose global Q-values along trajectory paths to assign credits for agents. We evaluate QPD on the challenging StarCraft II micromanagement tasks and show that QPD achieves the state-of-the-art performance in both homogeneous and heterogeneous multiagent scenarios compared with existing cooperative MARL algorithms."
                    },
                    {
                        "title": "MFE-ETP: A Comprehensive Evaluation Benchmark for Multi-modal Foundation Models on Embodied Task Planning",
                        "abstract": "In recent years, Multi-modal Foundation Models (MFMs) and Embodied Artificial Intelligence (EAI) have been advancing side by side at an unprecedented pace. The integration of the two has garnered significant attention from the AI research community. In this work, we attempt to provide an in-depth and comprehensive evaluation of the performance of MFM s on embodied task planning, aiming to shed light on their capabilities and limitations in this domain. To this end, based on the characteristics of embodied task planning, we first develop a systematic evaluation framework, which encapsulates four crucial capabilities of MFMs: object understanding, spatio-temporal perception, task understanding, and embodied reasoning. Following this, we propose a new benchmark, named MFE-ETP, characterized its complex and variable task scenarios, typical yet diverse task types, task instances of varying difficulties, and rich test case types ranging from multiple embodied question answering to embodied task reasoning. Finally, we offer a simple and easy-to-use automatic evaluation platform that enables the automated testing of multiple MFMs on the proposed benchmark. Using the benchmark and evaluation platform, we evaluated several state-of-the-art MFMs and found that they significantly lag behind human-level performance. The MFE-ETP is a high-quality, large-scale, and challenging benchmark relevant to real-world tasks."
                    }
                ]
            },
            "93947554-dcf4-4911-b599-f18552898731": {
                "pk": "93947554-dcf4-4911-b599-f18552898731",
                "name": "Glen Berseth",
                "collaborators": [
                    "Sergey Levine",
                    "Michiel van de Panne",
                    "Chelsea Finn",
                    "Mariano Phielipp",
                    "Roger Creus Castanyer",
                    "Raj Ghugare",
                    "Adriana Hugessen",
                    "Coline Devin",
                    "Dinesh Jayaraman",
                    "Brian Yang"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Robotics",
                    "Meta-Learning",
                    "Exploration Strategies"
                ],
                "publications": [
                    {
                        "title": "Model-Based Action Exploration for Learning Dynamic Motion Skills",
                        "abstract": "Deep reinforcement learning has achieved great strides in solving challenging motion control tasks. Recently, there has been significant work on methods for exploiting the data gathered during training, but there has been less work on how to best generate the data to learn from. For continuous action domains, the most common method for generating exploratory actions involves sampling from a Gaussian distribution centred around the mean action output by a policy. Although these methods can be quite capable, they do not scale well with the dimensionality of the action space, and can be dangerous to apply on hardware. We consider learning a forward dynamics model to predict the result, ($x_{t+1}$), of taking a particular action, ($u$), given a specific observation of the state, ($x_{t}$). With this model we perform internal look-ahead predictions of outcomes and seek actions we believe have a reasonable chance of success. This method alters the exploratory action space, thereby increasing learning speed and enables higher quality solutions to difficult problems, such as robotic locomotion and juggling."
                    },
                    {
                        "title": "Towards Learning to Imitate from a Single Video Demonstration",
                        "abstract": "Agents that can learn to imitate given video observation -- \\emph{without direct access to state or action information} are more applicable to learning in the natural world. However, formulating a reinforcement learning (RL) agent that facilitates this goal remains a significant challenge. We approach this challenge using contrastive training to learn a reward function comparing an agent's behaviour with a single demonstration. We use a Siamese recurrent neural network architecture to learn rewards in space and time between motion clips while training an RL policy to minimize this distance. Through experimentation, we also find that the inclusion of multi-task data and additional image encoding losses improve the temporal consistency of the learned rewards and, as a result, significantly improves policy learning. We demonstrate our approach on simulated humanoid, dog, and raptor agents in 2D and a quadruped and a humanoid in 3D. We show that our method outperforms current state-of-the-art techniques in these environments and can learn to imitate from a single video demonstration."
                    },
                    {
                        "title": "Terrain RL Simulator",
                        "abstract": "We provide $89$ challenging simulation environments that range in difficulty. The difficulty of solving a task is linked not only to the number of dimensions in the action space but also to the size and shape of the distribution of configurations the agent experiences. Therefore, we are releasing a number of simulation environments that include randomly generated terrain. The library also provides simple mechanisms to create new environments with different agent morphologies and the option to modify the distribution of generated terrain. We believe using these and other more complex simulations will help push the field closer to creating human-level intelligence."
                    },
                    {
                        "title": "CoMPS: Continual Meta Policy Search",
                        "abstract": "We develop a new continual meta-learning method to address challenges in sequential multi-task learning. In this setting, the agent's goal is to achieve high reward over any sequence of tasks quickly. Prior meta-reinforcement learning algorithms have demonstrated promising results in accelerating the acquisition of new tasks. However, they require access to all tasks during training. Beyond simply transferring past experience to new tasks, our goal is to devise continual reinforcement learning algorithms that learn to learn, using their experience on previous tasks to learn new tasks more quickly. We introduce a new method, continual meta-policy search (CoMPS), that removes this limitation by meta-training in an incremental fashion, over each task in a sequence, without revisiting prior tasks. CoMPS continuously repeats two subroutines: learning a new task using RL and using the experience from RL to perform completely offline meta-learning to prepare for subsequent task learning. We find that CoMPS outperforms prior continual learning and off-policy meta-reinforcement methods on several sequences of challenging continuous control tasks."
                    },
                    {
                        "title": "AnyMorph: Learning Transferable Polices By Inferring Agent Morphology",
                        "abstract": "The prototypical approach to reinforcement learning involves training policies tailored to a particular agent from scratch for every new morphology. Recent work aims to eliminate the re-training of policies by investigating whether a morphology-agnostic policy, trained on a diverse set of agents with similar task objectives, can be transferred to new agents with unseen morphologies without re-training. This is a challenging problem that required previous approaches to use hand-designed descriptions of the new agent's morphology. Instead of hand-designing this description, we propose a data-driven method that learns a representation of morphology directly from the reinforcement learning objective. Ours is the first reinforcement learning algorithm that can train a policy to generalize to new agent morphologies without requiring a description of the agent's morphology in advance. We evaluate our approach on the standard benchmark for agent-agnostic control, and improve over the current state of the art in zero-shot generalization to new agents. Importantly, our method attains good performance without an explicit description of morphology."
                    },
                    {
                        "title": "Torque-based Deep Reinforcement Learning for Task-and-Robot Agnostic Learning on Bipedal Robots Using Sim-to-Real Transfer",
                        "abstract": "In this paper, we review the question of which action space is best suited for controlling a real biped robot in combination with Sim2Real training. Position control has been popular as it has been shown to be more sample efficient and intuitive to combine with other planning algorithms. However, for position control gain tuning is required to achieve the best possible policy performance. We show that instead, using a torque-based action space enables task-and-robot agnostic learning with less parameter tuning and mitigates the sim-to-reality gap by taking advantage of torque control's inherent compliance. Also, we accelerate the torque-based-policy training process by pre-training the policy to remain upright by compensating for gravity. The paper showcases the first successful sim-to-real transfer of a torque-based deep reinforcement learning policy on a real human-sized biped robot. The video is available at https://youtu.be/CR6pTS39VRE."
                    },
                    {
                        "title": "Improving Intrinsic Exploration by Creating Stationary Objectives",
                        "abstract": "Exploration bonuses in reinforcement learning guide long-horizon exploration by defining custom intrinsic objectives. Several exploration objectives like count-based bonuses, pseudo-counts, and state-entropy maximization are non-stationary and hence are difficult to optimize for the agent. While this issue is generally known, it is usually omitted and solutions remain under-explored. The key contribution of our work lies in transforming the original non-stationary rewards into stationary rewards through an augmented state representation. For this purpose, we introduce the Stationary Objectives For Exploration (SOFE) framework. SOFE requires identifying sufficient statistics for different exploration bonuses and finding an efficient encoding of these statistics to use as input to a deep network. SOFE is based on proposing state augmentations that expand the state space but hold the promise of simplifying the optimization of the agent's objective. We show that SOFE improves the performance of several exploration objectives, including count-based bonuses, pseudo-counts, and state-entropy maximization. Moreover, SOFE outperforms prior methods that attempt to stabilize the optimization of intrinsic objectives. We demonstrate the efficacy of SOFE in hard-exploration problems, including sparse-reward tasks, pixel-based observations, 3D navigation, and procedurally generated environments."
                    },
                    {
                        "title": "Progressive Reinforcement Learning with Distillation for Multi-Skilled Motion Control",
                        "abstract": "Deep reinforcement learning has demonstrated increasing capabilities for continuous control problems, including agents that can move with skill and agility through their environment. An open problem in this setting is that of developing good strategies for integrating or merging policies for multiple skills, where each individual skill is a specialist in a specific skill and its associated state distribution. We extend policy distillation methods to the continuous action setting and leverage this technique to combine expert policies, as evaluated in the domain of simulated bipedal locomotion across different classes of terrain. We also introduce an input injection method for augmenting an existing policy network to exploit new input features. Lastly, our method uses transfer learning to assist in the efficient acquisition of new skills. The combination of these methods allows a policy to be incrementally augmented with new skills. We compare our progressive learning and integration via distillation (PLAID) method against three alternative baselines."
                    },
                    {
                        "title": "Closing the Gap between TD Learning and Supervised Learning -- A Generalisation Point of View",
                        "abstract": "Some reinforcement learning (RL) algorithms can stitch pieces of experience to solve a task never seen before during training. This oft-sought property is one of the few ways in which RL methods based on dynamic-programming differ from RL methods based on supervised-learning (SL). Yet, certain RL methods based on off-the-shelf SL algorithms achieve excellent results without an explicit mechanism for stitching; it remains unclear whether those methods forgo this important stitching property. This paper studies this question for the problems of achieving a target goal state and achieving a target return value. Our main result is to show that the stitching property corresponds to a form of combinatorial generalization: after training on a distribution of (state, goal) pairs, one would like to evaluate on (state, goal) pairs not seen together in the training data. Our analysis shows that this sort of generalization is different from i.i.d. generalization. This connection between stitching and generalisation reveals why we should not expect SL-based RL methods to perform stitching, even in the limit of large datasets and models. Based on this analysis, we construct new datasets to explicitly test for this property, revealing that SL-based methods lack this stitching property and hence fail to perform combinatorial generalization. Nonetheless, the connection between stitching and combinatorial generalisation also suggests a simple remedy for improving generalisation in SL: data augmentation. We propose a temporal data augmentation and demonstrate that adding it to SL-based methods enables them to successfully complete tasks not seen together during training. On a high level, this connection illustrates the importance of combinatorial generalization for data efficiency in time-series data beyond tasks beyond RL, like audio, video, or text."
                    },
                    {
                        "title": "Maximum State Entropy Exploration using Predecessor and Successor Representations",
                        "abstract": "Animals have a developed ability to explore that aids them in important tasks such as locating food, exploring for shelter, and finding misplaced items. These exploration skills necessarily track where they have been so that they can plan for finding items with relative efficiency. Contemporary exploration algorithms often learn a less efficient exploration strategy because they either condition only on the current state or simply rely on making random open-loop exploratory moves. In this work, we propose $\\eta\\psi$-Learning, a method to learn efficient exploratory policies by conditioning on past episodic experience to make the next exploratory move. Specifically, $\\eta\\psi$-Learning learns an exploration policy that maximizes the entropy of the state visitation distribution of a single trajectory. Furthermore, we demonstrate how variants of the predecessor representation and successor representations can be combined to predict the state visitation entropy. Our experiments demonstrate the efficacy of $\\eta\\psi$-Learning to strategically explore the environment and maximize the state coverage with limited samples."
                    },
                    {
                        "title": "Searching for High-Value Molecules Using Reinforcement Learning and Transformers",
                        "abstract": "Reinforcement learning (RL) over text representations can be effective for finding high-value policies that can search over graphs. However, RL requires careful structuring of the search space and algorithm design to be effective in this challenge. Through extensive experiments, we explore how different design choices for text grammar and algorithmic choices for training can affect an RL policy's ability to generate molecules with desired properties. We arrive at a new RL-based molecular design algorithm (ChemRLformer) and perform a thorough analysis using 25 molecule design tasks, including computationally complex protein docking simulations. From this analysis, we discover unique insights in this problem space and show that ChemRLformer achieves state-of-the-art performance while being more straightforward than prior work by demystifying which design choices are actually helpful for text-based molecule design."
                    },
                    {
                        "title": "Intelligent Switching for Reset-Free RL",
                        "abstract": "In the real world, the strong episode resetting mechanisms that are needed to train agents in simulation are unavailable. The \\textit{resetting} assumption limits the potential of reinforcement learning in the real world, as providing resets to an agent usually requires the creation of additional handcrafted mechanisms or human interventions. Recent work aims to train agents (\\textit{forward}) with learned resets by constructing a second (\\textit{backward}) agent that returns the forward agent to the initial state. We find that the termination and timing of the transitions between these two agents are crucial for algorithm success. With this in mind, we create a new algorithm, Reset Free RL with Intelligently Switching Controller (RISC) which intelligently switches between the two agents based on the agent's confidence in achieving its current goal. Our new method achieves state-of-the-art performance on several challenging environments for reset-free RL."
                    },
                    {
                        "title": "Surprise-Adaptive Intrinsic Motivation for Unsupervised Reinforcement Learning",
                        "abstract": "Both entropy-minimizing and entropy-maximizing (curiosity) objectives for unsupervised reinforcement learning (RL) have been shown to be effective in different environments, depending on the environment's level of natural entropy. However, neither method alone results in an agent that will consistently learn intelligent behavior across environments. In an effort to find a single entropy-based method that will encourage emergent behaviors in any environment, we propose an agent that can adapt its objective online, depending on the entropy conditions by framing the choice as a multi-armed bandit problem. We devise a novel intrinsic feedback signal for the bandit, which captures the agent's ability to control the entropy in its environment. We demonstrate that such agents can learn to control entropy and exhibit emergent behaviors in both high- and low-entropy regimes and can learn skillful behaviors in benchmark tasks. Videos of the trained agents and summarized findings can be found on our project page https://sites.google.com/view/surprise-adaptive-agents"
                    },
                    {
                        "title": "Interactive Diversity Optimization of Environments",
                        "abstract": "The design of a building requires an architect to balance a wide range of constraints: aesthetic, geometric, usability, lighting, safety, etc. At the same time, there are often a multiplicity of diverse designs that can meet these constraints equally well. Architects must use their skills and artistic vision to explore these rich but highly constrained design spaces. A number of computer-aided design tools use automation to provide useful analytical data and optimal designs with respect to certain fitness criteria. However, this automation can come at the expense of a designer's creative control.   We propose uDOME, a user-in-the-loop system for computer-aided design exploration that balances automation and control by efficiently exploring, analyzing, and filtering the space of environment layouts to better inform an architect's decision-making. At each design iteration, uDOME provides a set of diverse designs which satisfy user-defined constraints and optimality criteria within a user defined parameterization of the design space. The user then selects a design and performs a similar optimization with the same or different parameters and objectives. This exploration process can be repeated as many times as the designer wishes. Our user studies indicates that \\DOME, with its diversity-based approach, improves the efficiency and effectiveness of even novice users with minimal training, without compromising the quality of their designs."
                    },
                    {
                        "title": "SMiRL: Surprise Minimizing Reinforcement Learning in Unstable Environments",
                        "abstract": "Every living organism struggles against disruptive environmental forces to carve out and maintain an orderly niche. We propose that such a struggle to achieve and preserve order might offer a principle for the emergence of useful behaviors in artificial agents. We formalize this idea into an unsupervised reinforcement learning method called surprise minimizing reinforcement learning (SMiRL). SMiRL alternates between learning a density model to evaluate the surprise of a stimulus, and improving the policy to seek more predictable stimuli. The policy seeks out stable and repeatable situations that counteract the environment's prevailing sources of entropy. This might include avoiding other hostile agents, or finding a stable, balanced pose for a bipedal robot in the face of disturbance forces. We demonstrate that our surprise minimizing agents can successfully play Tetris, Doom, control a humanoid to avoid falls, and navigate to escape enemies in a maze without any task-specific reward supervision. We further show that SMiRL can be used together with standard task rewards to accelerate reward-driven learning."
                    },
                    {
                        "title": "Fully Autonomous Real-World Reinforcement Learning with Applications to Mobile Manipulation",
                        "abstract": "We study how robots can autonomously learn skills that require a combination of navigation and grasping. While reinforcement learning in principle provides for automated robotic skill learning, in practice reinforcement learning in the real world is challenging and often requires extensive instrumentation and supervision. Our aim is to devise a robotic reinforcement learning system for learning navigation and manipulation together, in an autonomous way without human intervention, enabling continual learning under realistic assumptions. Our proposed system, ReLMM, can learn continuously on a real-world platform without any environment instrumentation, without human intervention, and without access to privileged information, such as maps, objects positions, or a global view of the environment. Our method employs a modularized policy with components for manipulation and navigation, where manipulation policy uncertainty drives exploration for the navigation controller, and the manipulation module provides rewards for navigation. We evaluate our method on a room cleanup task, where the robot must navigate to and pick up items scattered on the floor. After a grasp curriculum training phase, ReLMM can learn navigation and grasping together fully automatically, in around 40 hours of autonomous real-world training."
                    },
                    {
                        "title": "Morphology-Agnostic Visual Robotic Control",
                        "abstract": "Existing approaches for visuomotor robotic control typically require characterizing the robot in advance by calibrating the camera or performing system identification. We propose MAVRIC, an approach that works with minimal prior knowledge of the robot's morphology, and requires only a camera view containing the robot and its environment and an unknown control interface. MAVRIC revolves around a mutual information-based method for self-recognition, which discovers visual \"control points\" on the robot body within a few seconds of exploratory interaction, and these control points in turn are then used for visual servoing. MAVRIC can control robots with imprecise actuation, no proprioceptive feedback, unknown morphologies including novel tools, unknown camera poses, and even unsteady handheld cameras. We demonstrate our method on visually-guided 3D point reaching, trajectory following, and robot-to-robot imitation."
                    },
                    {
                        "title": "DisCo RL: Distribution-Conditioned Reinforcement Learning for General-Purpose Policies",
                        "abstract": "Can we use reinforcement learning to learn general-purpose policies that can perform a wide range of different tasks, resulting in flexible and reusable skills? Contextual policies provide this capability in principle, but the representation of the context determines the degree of generalization and expressivity. Categorical contexts preclude generalization to entirely new tasks. Goal-conditioned policies may enable some generalization, but cannot capture all tasks that might be desired. In this paper, we propose goal distributions as a general and broadly applicable task representation suitable for contextual policies. Goal distributions are general in the sense that they can represent any state-based reward function when equipped with an appropriate distribution class, while the particular choice of distribution class allows us to trade off expressivity and learnability. We develop an off-policy algorithm called distribution-conditioned reinforcement learning (DisCo RL) to efficiently learn these policies. We evaluate DisCo RL on a variety of robot manipulation tasks and find that it significantly outperforms prior methods on tasks that require generalization to new goal distributions."
                    },
                    {
                        "title": "Reasoning with Latent Diffusion in Offline Reinforcement Learning",
                        "abstract": "Offline reinforcement learning (RL) holds promise as a means to learn high-reward policies from a static dataset, without the need for further environment interactions. However, a key challenge in offline RL lies in effectively stitching portions of suboptimal trajectories from the static dataset while avoiding extrapolation errors arising due to a lack of support in the dataset. Existing approaches use conservative methods that are tricky to tune and struggle with multi-modal data (as we show) or rely on noisy Monte Carlo return-to-go samples for reward conditioning. In this work, we propose a novel approach that leverages the expressiveness of latent diffusion to model in-support trajectory sequences as compressed latent skills. This facilitates learning a Q-function while avoiding extrapolation error via batch-constraining. The latent space is also expressive and gracefully copes with multi-modal data. We show that the learned temporally-abstract latent space encodes richer task-specific information for offline RL tasks as compared to raw state-actions. This improves credit assignment and facilitates faster reward propagation during Q-learning. Our method demonstrates state-of-the-art performance on the D4RL benchmarks, particularly excelling in long-horizon, sparse-reward tasks."
                    }
                ]
            }
        }
    },
    "2405.10301": {
        "paper_data": {
            "title": "Conformal Alignment: Knowing When to Trust Foundation Models with Guarantees",
            "url": "http://arxiv.org/abs/2405.10301v2",
            "arxiv_id": "2405.10301",
            "authors": [
                "Yu Gui",
                "Ying Jin",
                "Zhimei Ren"
            ],
            "abstract": "Before deploying outputs from foundation models in high-stakes tasks, it is imperative to ensure that they align with human values. For instance, in radiology report generation, reports generated by a vision-language model must align with human evaluations before their use in medical decision-making. This paper presents Conformal Alignment, a general framework for identifying units whose outputs meet a user-specified alignment criterion. It is guaranteed that on average, a prescribed fraction of selected units indeed meet the alignment criterion, regardless of the foundation model or the data distribution. Given any pre-trained model and new units with model-generated outputs, Conformal Alignment leverages a set of reference data with ground-truth alignment status to train an alignment predictor. It then selects new units whose predicted alignment scores surpass a data-dependent threshold, certifying their corresponding outputs as trustworthy. Through applications to question answering and radiology report generation, we demonstrate that our method is able to accurately identify units with trustworthy outputs via lightweight training over a moderate amount of reference data. En route, we investigate the informativeness of various features in alignment prediction and combine them with standard models to construct the alignment predictor.",
            "introduction": "   1 Introduction  Large-scale, pre-trained foundation models are remarkably powerful in generating relevant and informative outputs of various forms for diverse downstream tasks, marking a new era of artificial intelligence\u00a0[32, 54, 45]. However, it has been recognized that they can be prone to factual errors\u00a0[41], hallucinations\u00a0[14, 49], and bias\u00a0[4], among others. These issues raise prevalent societal concerns regarding the reliable use of foundation models in high-stakes scenarios\u00a0[7, 6]. It remains a significant challenge to ensure that outputs from these highly complex models align with human values\u00a0[34].   Towards uncertainty quantification of black-box models, conformal prediction\u00a0[48, CP] offers a versatile, distribution-free solution. While CP has been applied to classification and regression tasks addressed by foundation models\u00a0[24, 26, 44, 46, 52, 2], its use to ensure the alignment of general-form outputs remains largely unexplored. Recently, [34] propose to generate multiple outputs for one input until it is guaranteed that at least one output meets a specific alignment criterion. Such a guarantee, however, may not be immediately practical, as users must still determine (manually) which output(s) is aligned. [30] modify the machine-generated outputs to ensure factuality with high probability, which in turn may sacrifice model power by making responses vague. As such, determining guarantees that are practically meaningful for downstream tasks and developing methods to achieve them remain an active research area.   This paper introduces a distinct type of guarantee: we aim to certify whether each model output is aligned or not, such that those certified are ensured to be mostly correct. This guarantee is directly relevant to downstream tasks, as it allows immediate and reliable use of the certified outputs without further modifications. On the other hand, we abstain from deploying model outputs for units that are not selected\u2014intuitively, those are where the model lacks confidence (i.e., \u201cknowing they do not know\u201d), and they can be evaluated by human experts. Such a practice therefore provides an automated pipeline for safe deployment of model-generated outputs.   Present work.  We present Conformal Alignment (CA), a general framework that determines when to trust foundation model outputs with finite-sample, distribution-free guarantees. The framework is based on CP ideas, but unlike standard CP that searches the sampling space to construct a prediction set that contains a desired output, we leverage the quantified confidence as an instrument to select units whose already-generated outputs are trusted. Given any model and any alignment criterion, Conformal Alignment ensures that a prescribed proportion of the selected units\u2019 model outputs indeed meet the alignment criterion.   Concretely, suppose m\u2208\u2115+\ud835\udc5asubscript\u2115m\\in\\mathbb{N}_{+}italic_m \u2208 blackboard_N start_POSTSUBSCRIPT + end_POSTSUBSCRIPT new units awaiting outputs. Given an error level \u03b1\u2208(0,1)\ud835\udefc01\\alpha\\in(0,1)italic_\u03b1 \u2208 ( 0 , 1 ), Conformal Alignment controls the false discovery rate (FDR) in selecting units with trustworthy outputs:    FDR=\ud835\udd3c\u2062[\u2211i=1m\ud835\udfd9\u2061{i\u2062\u00a0selected and not aligned}\u2211i=1m\ud835\udfd9\u2061{i\u2062\u00a0selected}]\u2264\u03b1,FDR\ud835\udd3cdelimited-[]superscriptsubscript\ud835\udc561\ud835\udc5a1\ud835\udc56\u00a0selected and not alignedsuperscriptsubscript\ud835\udc561\ud835\udc5a1\ud835\udc56\u00a0selected\ud835\udefc\\displaystyle\\textnormal{FDR}=\\mathbb{E}\\bigg{[}\\frac{\\sum_{i=1}^{m}% \\operatorname{\\mathds{1}}\\{i\\text{ selected and not aligned}\\}}{\\sum_{i=1}^{m}% \\operatorname{\\mathds{1}}\\{i\\text{ selected}\\}}\\bigg{]}\\leq\\alpha,FDR = blackboard_E [ divide start_ARG \u2211 start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_m end_POSTSUPERSCRIPT blackboard_1 { italic_i selected and not aligned } end_ARG start_ARG \u2211 start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_m end_POSTSUPERSCRIPT blackboard_1 { italic_i selected } end_ARG ] \u2264 italic_\u03b1 ,  (1.1)   where the expectation is taken over the randomness of the data. FDR control offers an interpretable measure of the quality of selected deployable units. Our method builds upon the Conformalized Selection framework\u00a0[17, 16], leveraging a holdout set of \u201chigh-quality\u201d reference data to guide the selection with calibrated statistical confidence for trusted outputs. Like in CP,\u00a0(1.1) holds in finite-sample as long as",
            "references": [
                {
                    "title": "Uncertainty in Language Models: Assessment through Rank-Calibration",
                    "abstract": "Language Models (LMs) have shown promising performance in natural language generation. However, as LMs often generate incorrect or hallucinated responses, it is crucial to correctly quantify their uncertainty in responding to given inputs. In addition to verbalized confidence elicited via prompting, many uncertainty measures ($e.g.$, semantic entropy and affinity-graph-based measures) have been proposed. However, these measures can differ greatly, and it is unclear how to compare them, partly because they take values over different ranges ($e.g.$, $[0,\\infty)$ or $[0,1]$). In this work, we address this issue by developing a novel and practical framework, termed $Rank$-$Calibration$, to assess uncertainty and confidence measures for LMs. Our key tenet is that higher uncertainty (or lower confidence) should imply lower generation quality, on average. Rank-calibration quantifies deviations from this ideal relationship in a principled manner, without requiring ad hoc binary thresholding of the correctness score ($e.g.$, ROUGE or METEOR). The broad applicability and the granular interpretability of our methods are demonstrated empirically."
                },
                {
                    "title": "Confidence on the Focal: Conformal Prediction with Selection-Conditional Coverage",
                    "abstract": "Conformal prediction builds marginally valid prediction intervals that cover the unknown outcome of a randomly drawn new test point with a prescribed probability. However, a common scenario in practice is that, after seeing the data, practitioners decide which test unit(s) to focus on in a data-driven manner and seek for uncertainty quantification of the focal unit(s). In such cases, marginally valid conformal prediction intervals may not provide valid coverage for the focal unit(s) due to selection bias. This paper presents a general framework for constructing a prediction set with finite-sample exact coverage conditional on the unit being selected by a given procedure. The general form of our method works for arbitrary selection rules that are invariant to the permutation of the calibration units, and generalizes Mondrian Conformal Prediction to multiple test units and non-equivariant classifiers. We then work out the computationally efficient implementation of our framework for a number of realistic selection rules, including top-K selection, optimization-based selection, selection based on conformal p-values, and selection based on properties of preliminary conformal prediction sets. The performance of our methods is demonstrated via applications in drug discovery and health risk prediction."
                },
                {
                    "title": "API Is Enough: Conformal Prediction for Large Language Models Without Logit-Access",
                    "abstract": "This study aims to address the pervasive challenge of quantifying uncertainty in large language models (LLMs) without logit-access. Conformal Prediction (CP), known for its model-agnostic and distribution-free features, is a desired approach for various LLMs and data distributions. However, existing CP methods for LLMs typically assume access to the logits, which are unavailable for some API-only LLMs. In addition, logits are known to be miscalibrated, potentially leading to degraded CP performance. To tackle these challenges, we introduce a novel CP method that (1) is tailored for API-only LLMs without logit-access; (2) minimizes the size of prediction sets; and (3) ensures a statistical guarantee of the user-defined coverage. The core idea of this approach is to formulate nonconformity measures using both coarse-grained (i.e., sample frequency) and fine-grained uncertainty notions (e.g., semantic similarity). Experimental results on both close-ended and open-ended Question Answering tasks show our approach can mostly outperform the logit-based CP baselines."
                },
                {
                    "title": "Unintended Impacts of LLM Alignment on Global Representation",
                    "abstract": "Before being deployed for user-facing applications, developers align Large Language Models (LLMs) to user preferences through a variety of procedures, such as Reinforcement Learning From Human Feedback (RLHF) and Direct Preference Optimization (DPO). Current evaluations of these procedures focus on benchmarks of instruction following, reasoning, and truthfulness. However, human preferences are not universal, and aligning to specific preference sets may have unintended effects. We explore how alignment impacts performance along three axes of global representation: English dialects, multilingualism, and opinions from and about countries worldwide. Our results show that current alignment procedures create disparities between English dialects and global opinions. We find alignment improves capabilities in several languages. We conclude by discussing design decisions that led to these unintended impacts and recommendations for more equitable preference tuning. We make our code and data publicly available on Github."
                },
                {
                    "title": "Language Models with Conformal Factuality Guarantees",
                    "abstract": "Guaranteeing the correctness and factuality of language model (LM) outputs is a major open problem. In this work, we propose conformal factuality, a framework that can ensure high probability correctness guarantees for LMs by connecting language modeling and conformal prediction. We observe that the correctness of an LM output is equivalent to an uncertainty quantification problem, where the uncertainty sets are defined as the entailment set of an LM's output. Using this connection, we show that conformal prediction in language models corresponds to a back-off algorithm that provides high probability correctness guarantees by progressively making LM outputs less specific (and expanding the associated uncertainty sets). This approach applies to any black-box LM and requires very few human-annotated samples. Evaluations of our approach on closed book QA (FActScore, NaturalQuestions) and reasoning tasks (MATH) show that our approach can provide 80-90% correctness guarantees while retaining the majority of the LM's original output."
                },
                {
                    "title": "Conformal Triage for Medical Imaging AI Deployment",
                    "abstract": "Background: The deployment of black-box AI models in medical imaging presents significant challenges, especially in maintaining reliability across different clinical settings. These challenges are compounded by distribution shifts that can lead to failures in reproducing the accuracy attained during the AI model's original validations. Method: We introduce the conformal triage algorithm, designed to categorize patients into low-risk, high-risk, and uncertain groups within a clinical deployment setting. This method leverages a combination of a black-box AI model and conformal prediction techniques to offer statistical guarantees of predictive power for each group. The high-risk group is guaranteed to have a high positive predictive value, while the low-risk group is assured a high negative predictive value. Prediction sets are never constructed; instead, conformal techniques directly assure high accuracy in both groups, even in clinical environments different from those in which the AI model was originally trained, thereby ameliorating the challenges posed by distribution shifts. Importantly, a representative data set of exams from the testing environment is required to ensure statistical validity. Results: The algorithm was tested using a head CT model previously developed by Do and colleagues [9] and a data set from Massachusetts General Hospital. The results demonstrate that the conformal triage algorithm provides reliable predictive value guarantees to a clinically significant extent, reducing the number of false positives from 233 (45%) to 8 (5%) while only abstaining from prediction on 14% of data points, even in a setting different from the training environment of the original AI model. Conclusions: The conformal triage algorithm offers a promising solution to the challenge of deploying black-box AI models in medical imaging across varying clinical settings. By providing statistical guarantees of predictive value for categorized patient groups, this approach significantly enhances the reliability and utility of AI in optimizing medical imaging workflows, particularly in neuroradiology."
                },
                {
                    "title": "Non-Exchangeable Conformal Language Generation with Nearest Neighbors",
                    "abstract": "Quantifying uncertainty in automatically generated text is important for letting humans check potential hallucinations and making systems more reliable. Conformal prediction is an attractive framework to provide predictions imbued with statistical guarantees, however, its application to text generation is challenging since any i.i.d. assumptions are not realistic. In this paper, we bridge this gap by leveraging recent results on *non-exchangeable* conformal prediction, which still ensures bounds on coverage. The result, *non-exchangeable conformal nucleus sampling*, is a novel extension of the conformal prediction framework to generation based on nearest neighbors. Our method can be used post-hoc for an arbitrary model without extra training and supplies token-level, calibrated prediction sets equipped with statistical guarantees. Experiments in machine translation and language modeling show encouraging results in generation quality. By also producing tighter prediction sets with good coverage, we thus give a more theoretically principled way to perform sampling with conformal guarantees."
                },
                {
                    "title": "Benchmarking LLMs via Uncertainty Quantification",
                    "abstract": "The proliferation of open-source Large Language Models (LLMs) from various institutions has highlighted the urgent need for comprehensive evaluation methods. However, current evaluation platforms, such as the widely recognized HuggingFace open LLM leaderboard, neglect a crucial aspect -- uncertainty, which is vital for thoroughly assessing LLMs. To bridge this gap, we introduce a new benchmarking approach for LLMs that integrates uncertainty quantification. Our examination involves eight LLMs (LLM series) spanning five representative natural language processing tasks. Our findings reveal that: I) LLMs with higher accuracy may exhibit lower certainty; II) Larger-scale LLMs may display greater uncertainty compared to their smaller counterparts; and III) Instruction-finetuning tends to increase the uncertainty of LLMs. These results underscore the significance of incorporating uncertainty in the evaluation of LLMs."
                },
                {
                    "title": "Uncertainty-aware Language Modeling for Selective Question Answering",
                    "abstract": "We present an automatic large language model (LLM) conversion approach that produces uncertainty-aware LLMs capable of estimating uncertainty with every prediction. Our approach is model- and data-agnostic, is computationally-efficient, and does not rely on external models or systems. We evaluate converted models on the selective question answering setting -- to answer as many questions as possible while maintaining a given accuracy, forgoing providing predictions when necessary. As part of our results, we test BERT and Llama 2 model variants on the SQuAD extractive QA task and the TruthfulQA generative QA task. We show that using the uncertainty estimates provided by our approach to selectively answer questions leads to significantly higher accuracy over directly using model probabilities."
                },
                {
                    "title": "A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions",
                    "abstract": "The emergence of large language models (LLMs) has marked a significant breakthrough in natural language processing (NLP), leading to remarkable advancements in text understanding and generation. Nevertheless, alongside these strides, LLMs exhibit a critical tendency to produce hallucinations, resulting in content that is inconsistent with real-world facts or user inputs. This phenomenon poses substantial challenges to their practical deployment and raises concerns over the reliability of LLMs in real-world scenarios, which attracts increasing attention to detect and mitigate these hallucinations. In this survey, we aim to provide a thorough and in-depth overview of recent advances in the field of LLM hallucinations. We begin with an innovative taxonomy of LLM hallucinations, then delve into the factors contributing to hallucinations. Subsequently, we present a comprehensive overview of hallucination detection methods and benchmarks. Additionally, representative approaches designed to mitigate hallucinations are introduced accordingly. Finally, we analyze the challenges that highlight the current limitations and formulate open questions, aiming to delineate pathways for future research on hallucinations in LLMs."
                },
                {
                    "title": "Adaptation with Self-Evaluation to Improve Selective Prediction in LLMs",
                    "abstract": "Large language models (LLMs) have recently shown great advances in a variety of tasks, including natural language understanding and generation. However, their use in high-stakes decision-making scenarios is still limited due to the potential for errors. Selective prediction is a technique that can be used to improve the reliability of the LLMs by allowing them to abstain from making predictions when they are unsure of the answer. In this work, we propose a novel framework for adaptation with self-evaluation to improve the selective prediction performance of LLMs. Our framework is based on the idea of using parameter-efficient tuning to adapt the LLM to the specific task at hand while improving its ability to perform self-evaluation. We evaluate our method on a variety of question-answering (QA) datasets and show that it outperforms state-of-the-art selective prediction methods. For example, on the CoQA benchmark, our method improves the AUACC from 91.23% to 92.63% and improves the AUROC from 74.61% to 80.25%."
                },
                {
                    "title": "Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignment",
                    "abstract": "Ensuring alignment, which refers to making models behave in accordance with human intentions [1,2], has become a critical task before deploying large language models (LLMs) in real-world applications. For instance, OpenAI devoted six months to iteratively aligning GPT-4 before its release [3]. However, a major challenge faced by practitioners is the lack of clear guidance on evaluating whether LLM outputs align with social norms, values, and regulations. This obstacle hinders systematic iteration and deployment of LLMs. To address this issue, this paper presents a comprehensive survey of key dimensions that are crucial to consider when assessing LLM trustworthiness. The survey covers seven major categories of LLM trustworthiness: reliability, safety, fairness, resistance to misuse, explainability and reasoning, adherence to social norms, and robustness. Each major category is further divided into several sub-categories, resulting in a total of 29 sub-categories. Additionally, a subset of 8 sub-categories is selected for further investigation, where corresponding measurement studies are designed and conducted on several widely-used LLMs. The measurement results indicate that, in general, more aligned models tend to perform better in terms of overall trustworthiness. However, the effectiveness of alignment varies across the different trustworthiness categories considered. This highlights the importance of conducting more fine-grained analyses, testing, and making continuous improvements on LLM alignment. By shedding light on these key dimensions of LLM trustworthiness, this paper aims to provide valuable insights and guidance to practitioners in the field. Understanding and addressing these concerns will be crucial in achieving reliable and ethically sound deployment of LLMs in various applications."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Model-free selective inference under covariate shift via weighted conformal p-values",
                    "abstract": "This paper introduces novel weighted conformal p-values and methods for model-free selective inference. The problem is as follows: given test units with covariates $X$ and missing responses $Y$, how do we select units for which the responses $Y$ are larger than user-specified values while controlling the proportion of false positives? Can we achieve this without any modeling assumptions on the data and without any restriction on the model for predicting the responses? Last, methods should be applicable when there is a covariate shift between training and test data, which commonly occurs in practice. We answer these questions by first leveraging any prediction model to produce a class of well-calibrated weighted conformal p-values, which control the type-I error in detecting a large response. These p-values cannot be passed on to classical multiple testing procedures since they may not obey a well-known positive dependence property. Hence, we introduce weighted conformalized selection (WCS), a new procedure which controls false discovery rate (FDR) in finite samples. Besides prediction-assisted candidate selection, WCS (1) allows to infer multiple individual treatment effects, and (2) extends to outlier detection with inlier distributions shifts. We demonstrate performance via simulations and applications to causal inference, drug discovery, and outlier detection datasets."
                },
                {
                    "title": "TRAQ: Trustworthy Retrieval Augmented Question Answering via Conformal Prediction",
                    "abstract": "When applied to open-domain question answering, large language models (LLMs) frequently generate incorrect responses based on made-up facts, which are called hallucinations. Retrieval augmented generation (RAG) is a promising strategy to avoid hallucinations, but it does not provide guarantees on its correctness. To address this challenge, we propose the Trustworthy Retrieval Augmented Question Answering, or *TRAQ*, which provides the first end-to-end statistical correctness guarantee for RAG. TRAQ uses conformal prediction, a statistical technique for constructing prediction sets that are guaranteed to contain the semantically correct response with high probability. Additionally, TRAQ leverages Bayesian optimization to minimize the size of the constructed sets. In an extensive experimental evaluation, we demonstrate that TRAQ provides the desired correctness guarantee while reducing prediction set size by 16.2% on average compared to an ablation. The implementation is available: [https://github.com/shuoli90/TRAQ](https://github.com/shuoli90/TRAQ)."
                },
                {
                    "title": "Robots That Ask For Help: Uncertainty Alignment for Large Language Model Planners",
                    "abstract": "Large language models (LLMs) exhibit a wide range of promising capabilities -- from step-by-step planning to commonsense reasoning -- that may provide utility for robots, but remain prone to confidently hallucinated predictions. In this work, we present KnowNo, which is a framework for measuring and aligning the uncertainty of LLM-based planners such that they know when they don't know and ask for help when needed. KnowNo builds on the theory of conformal prediction to provide statistical guarantees on task completion while minimizing human help in complex multi-step planning settings. Experiments across a variety of simulated and real robot setups that involve tasks with different modes of ambiguity (e.g., from spatial to numeric uncertainties, from human preferences to Winograd schemas) show that KnowNo performs favorably over modern baselines (which may involve ensembles or extensive prompt tuning) in terms of improving efficiency and autonomy, while providing formal assurances. KnowNo can be used with LLMs out of the box without model-finetuning, and suggests a promising lightweight approach to modeling uncertainty that can complement and scale with the growing capabilities of foundation models. Website: https://robot-help.github.io"
                },
                {
                    "title": "Conformal Language Modeling",
                    "abstract": "We propose a novel approach to conformal prediction for generative language models (LMs). Standard conformal prediction produces prediction sets -- in place of single predictions -- that have rigorous, statistical performance guarantees. LM responses are typically sampled from the model's predicted distribution over the large, combinatorial output space of natural language. Translating this process to conformal prediction, we calibrate a stopping rule for sampling different outputs from the LM that get added to a growing set of candidates until we are confident that the output set is sufficient. Since some samples may be low-quality, we also simultaneously calibrate and apply a rejection rule for removing candidates from the output set to reduce noise. Similar to conformal prediction, we prove that the sampled set returned by our procedure contains at least one acceptable answer with high probability, while still being empirically precise (i.e., small) on average. Furthermore, within this set of candidate responses, we show that we can also accurately identify subsets of individual components -- such as phrases or sentences -- that are each independently correct (e.g., that are not\"hallucinations\"), again with statistical guarantees. We demonstrate the promise of our approach on multiple tasks in open-domain question answering, text summarization, and radiology report generation using different LM variants."
                },
                {
                    "title": "Generating with Confidence: Uncertainty Quantification for Black-box Large Language Models",
                    "abstract": "Large language models (LLMs) specializing in natural language generation (NLG) have recently started exhibiting promising capabilities across a variety of domains. However, gauging the trustworthiness of responses generated by LLMs remains an open challenge, with limited research on uncertainty quantification (UQ) for NLG. Furthermore, existing literature typically assumes white-box access to language models, which is becoming unrealistic either due to the closed-source nature of the latest LLMs or computational constraints. In this work, we investigate UQ in NLG for *black-box* LLMs. We first differentiate *uncertainty* vs *confidence*: the former refers to the ``dispersion'' of the potential predictions for a fixed input, and the latter refers to the confidence on a particular prediction/generation. We then propose and compare several confidence/uncertainty measures, applying them to *selective NLG* where unreliable results could either be ignored or yielded for further assessment. Experiments were carried out with several popular LLMs on question-answering datasets (for evaluation purposes). Results reveal that a simple measure for the semantic dispersion can be a reliable predictor of the quality of LLM responses, providing valuable insights for practitioners on uncertainty management when adopting LLMs. The code to replicate our experiments is available at https://github.com/zlin7/UQ-NLG."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "Conformal Prediction with Large Language Models for Multi-Choice Question Answering",
                    "abstract": "As large language models continue to be widely developed, robust uncertainty quantification techniques will become crucial for their safe deployment in high-stakes scenarios. In this work, we explore how conformal prediction can be used to provide uncertainty quantification in language models for the specific task of multiple-choice question-answering. We find that the uncertainty estimates from conformal prediction are tightly correlated with prediction accuracy. This observation can be useful for downstream applications such as selective classification and filtering out low-quality predictions. We also investigate the exchangeability assumption required by conformal prediction to out-of-subject questions, which may be a more realistic scenario for many practical applications. Our work contributes towards more trustworthy and reliable usage of large language models in safety-critical situations, where robust guarantees of error rate are required."
                },
                {
                    "title": "Uncertainty Quantification over Graph with Conformalized Graph Neural Networks",
                    "abstract": "Graph Neural Networks (GNNs) are powerful machine learning prediction models on graph-structured data. However, GNNs lack rigorous uncertainty estimates, limiting their reliable deployment in settings where the cost of errors is significant. We propose conformalized GNN (CF-GNN), extending conformal prediction (CP) to graph-based models for guaranteed uncertainty estimates. Given an entity in the graph, CF-GNN produces a prediction set/interval that provably contains the true label with pre-defined coverage probability (e.g. 90%). We establish a permutation invariance condition that enables the validity of CP on graph data and provide an exact characterization of the test-time coverage. Moreover, besides valid coverage, it is crucial to reduce the prediction set size/interval length for practical use. We observe a key connection between non-conformity scores and network structures, which motivates us to develop a topology-aware output correction model that learns to update the prediction and produces more efficient prediction sets/intervals. Extensive experiments show that CF-GNN achieves any pre-defined target marginal coverage while significantly reducing the prediction set/interval size by up to 74% over the baselines. It also empirically achieves satisfactory conditional coverage over various raw and network features."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation",
                    "abstract": "We introduce a method to measure uncertainty in large language models. For tasks like question answering, it is essential to know when we can trust the natural language outputs of foundation models. We show that measuring uncertainty in natural language is challenging because of\"semantic equivalence\"-- different sentences can mean the same thing. To overcome these challenges we introduce semantic entropy -- an entropy which incorporates linguistic invariances created by shared meanings. Our method is unsupervised, uses only a single model, and requires no modifications to off-the-shelf language models. In comprehensive ablation studies we show that the semantic entropy is more predictive of model accuracy on question answering data sets than comparable baselines."
                },
                {
                    "title": "Selection by Prediction with Conformal p-values",
                    "abstract": "Decision making or scientific discovery pipelines such as job hiring and drug discovery often involve multiple stages: before any resource-intensive step, there is often an initial screening that uses predictions from a machine learning model to shortlist a few candidates from a large pool. We study screening procedures that aim to select candidates whose unobserved outcomes exceed user-specified values. We develop a method that wraps around any prediction model to produce a subset of candidates while controlling the proportion of falsely selected units. Building upon the conformal inference framework, our method first constructs p-values that quantify the statistical evidence for large outcomes; it then determines the shortlist by comparing the p-values to a threshold introduced in the multiple testing literature. In many cases, the procedure selects candidates whose predictions are above a data-dependent threshold. Our theoretical guarantee holds under mild exchangeability conditions on the samples, generalizing existing results on multiple conformal p-values. We demonstrate the empirical performance of our method via simulations, and apply it to job hiring and drug discovery datasets."
                },
                {
                    "title": "Language Models (Mostly) Know What They Know",
                    "abstract": "We study whether language models can evaluate the validity of their own claims and predict which questions they will be able to answer correctly. We first show that larger models are well-calibrated on diverse multiple choice and true/false questions when they are provided in the right format. Thus we can approach self-evaluation on open-ended sampling tasks by asking models to first propose answers, and then to evaluate the probability\"P(True)\"that their answers are correct. We find encouraging performance, calibration, and scaling for P(True) on a diverse array of tasks. Performance at self-evaluation further improves when we allow models to consider many of their own samples before predicting the validity of one specific possibility. Next, we investigate whether models can be trained to predict\"P(IK)\", the probability that\"I know\"the answer to a question, without reference to any particular proposed answer. Models perform well at predicting P(IK) and partially generalize across tasks, though they struggle with calibration of P(IK) on new tasks. The predicted P(IK) probabilities also increase appropriately in the presence of relevant source materials in the context, and in the presence of hints towards the solution of mathematical word problems. We hope these observations lay the groundwork for training more honest models, and for investigating how honesty generalizes to cases where models are trained on objectives other than the imitation of human writing."
                },
                {
                    "title": "Selective Classification Via Neural Network Training Dynamics",
                    "abstract": "Selective classification is the task of rejecting inputs a model would predict incorrectly on through a trade-off between input space coverage and model accuracy. Current methods for selective classification impose constraints on either the model architecture or the loss function; this inhibits their usage in practice. In contrast to prior work, we show that state-of-the-art selective classification performance can be attained solely from studying the (discretized) training dynamics of a model. We propose a general framework that, for a given test input, monitors metrics capturing the disagreement with the final predicted label over intermediate models obtained during training; we then reject data points exhibiting too much disagreement at late stages in training. In particular, we instantiate a method that tracks when the label predicted during training stops disagreeing with the final predicted label. Our experimental evaluation shows that our method achieves state-of-the-art accuracy/coverage trade-offs on typical selective classification benchmarks."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "Reliable Visual Question Answering: Abstain Rather Than Answer Incorrectly",
                    "abstract": "Machine learning has advanced dramatically, narrowing the accuracy gap to humans in multimodal tasks like visual question answering (VQA). However, while humans can say\"I don't know\"when they are uncertain (i.e., abstain from answering a question), such ability has been largely neglected in multimodal research, despite the importance of this problem to the usage of VQA in real settings. In this work, we promote a problem formulation for reliable VQA, where we prefer abstention over providing an incorrect answer. We first enable abstention capabilities for several VQA models, and analyze both their coverage, the portion of questions answered, and risk, the error on that portion. For that, we explore several abstention approaches. We find that although the best performing models achieve over 70% accuracy on the VQA v2 dataset, introducing the option to abstain by directly using a model's softmax scores limits them to answering less than 7.5% of the questions to achieve a low risk of error (i.e., 1%). This motivates us to utilize a multimodal selection function to directly estimate the correctness of the predicted answers, which we show can increase the coverage by, for example, 2.3x from 6.8% to 15.6% at 1% risk. While it is important to analyze both coverage and risk, these metrics have a trade-off which makes comparing VQA models challenging. To address this, we also propose an Effective Reliability metric for VQA that places a larger cost on incorrect answers compared to abstentions. This new problem formulation, metric, and analysis for VQA provide the groundwork for building effective and reliable VQA models that have the self-awareness to abstain if and only if they don't know the answer."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "Language Models that Seek for Knowledge: Modular Search & Generation for Dialogue and Prompt Completion",
                    "abstract": "Language models (LMs) have recently been shown to generate more factual responses by employing modularity (Zhou et al., 2021) in combination with retrieval (Adolphs et al., 2021). We extend the recent approach of Adolphs et al. (2021) to include internet search as a module. Our SeeKeR (Search engine->Knowledge->Response) method thus applies a single LM to three modular tasks in succession: search, generating knowledge, and generating a final response. We show that, when using SeeKeR as a dialogue model, it outperforms the state-of-the-art model BlenderBot 2 (Chen et al., 2021) on open-domain knowledge-grounded conversations for the same number of parameters, in terms of consistency, knowledge and per-turn engagingness. SeeKeR applied to topical prompt completions as a standard language model outperforms GPT2 (Radford et al., 2019) and GPT3 (Brown et al., 2020) in terms of factuality and topicality, despite GPT3 being a vastly larger model. Our code and models are made publicly available."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Investigating Selective Prediction Approaches Across Several Tasks in IID, OOD, and Adversarial Settings",
                    "abstract": "In order to equip NLP systems with \u2018selective prediction\u2019 capability, several task-specific approaches have been proposed. However, which approaches work best across tasks or even if they consistently outperform the simplest baseline MaxProb remains to be explored. To this end, we systematically study selective prediction in a large-scale setup of 17 datasets across several NLP tasks. Through comprehensive experiments under in-domain (IID), out-of-domain (OOD), and adversarial (ADV) settings, we show that despite leveraging additional resources (held-out data/computation), none of the existing approaches consistently and considerably outperforms MaxProb in all three settings. Furthermore, their performance does not translate well across tasks. For instance, Monte-Carlo Dropout outperforms all other approaches on Duplicate Detection datasets but does not fare well on NLI datasets, especially in the OOD setting. Thus, we recommend that future selective prediction approaches should be evaluated across tasks and settings for reliable estimation of their capabilities."
                },
                {
                    "title": "Ethical and social risks of harm from Language Models",
                    "abstract": "This paper aims to help structure the risk landscape associated with large-scale Language Models (LMs). In order to foster advances in responsible innovation, an in-depth understanding of the potential risks posed by these models is needed. A wide range of established and anticipated risks are analysed in detail, drawing on multidisciplinary expertise and literature from computer science, linguistics, and social sciences. We outline six specific risk areas: I. Discrimination, Exclusion and Toxicity, II. Information Hazards, III. Misinformation Harms, V. Malicious Uses, V. Human-Computer Interaction Harms, VI. Automation, Access, and Environmental Harms. The first area concerns the perpetuation of stereotypes, unfair discrimination, exclusionary norms, toxic language, and lower performance by social group for LMs. The second focuses on risks from private data leaks or LMs correctly inferring sensitive information. The third addresses risks arising from poor, false or misleading information including in sensitive domains, and knock-on risks such as the erosion of trust in shared information. The fourth considers risks from actors who try to use LMs to cause harm. The fifth focuses on risks specific to LLMs used to underpin conversational agents that interact with human users, including unsafe use, manipulation or deception. The sixth discusses the risk of environmental harm, job automation, and other challenges that may have a disparate effect on different social groups or communities. In total, we review 21 risks in-depth. We discuss the points of origin of different risks and point to potential mitigation approaches. Lastly, we discuss organisational responsibilities in implementing mitigations, and the role of collaboration and participation. We highlight directions for further research, particularly on expanding the toolkit for assessing and evaluating the outlined risks in LMs."
                },
                {
                    "title": "Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Control",
                    "abstract": "We introduce a framework for calibrating machine learning models so that their predictions satisfy explicit, finite-sample statistical guarantees. Our calibration algorithms work with any underlying model and (unknown) data-generating distribution and do not require model refitting. The framework addresses, among other examples, false discovery rate control in multi-label classification, intersection-over-union control in instance segmentation, and the simultaneous control of the type-1 error of outlier detection and confidence set coverage in classification or regression. Our main insight is to reframe the risk-control problem as multiple hypothesis testing, enabling techniques and mathematical arguments different from those in the previous literature. We use the framework to provide new calibration methods for several core machine learning tasks, with detailed worked examples in computer vision and tabular medical data."
                },
                {
                    "title": "Selective Question Answering under Domain Shift",
                    "abstract": "To avoid giving wrong answers, question answering (QA) models need to know when to abstain from answering. Moreover, users often ask questions that diverge from the model\u2019s training data, making errors more likely and thus abstention more critical. In this work, we propose the setting of selective question answering under domain shift, in which a QA model is tested on a mixture of in-domain and out-of-domain data, and must answer (i.e., not abstain on) as many questions as possible while maintaining high accuracy. Abstention policies based solely on the model\u2019s softmax probabilities fare poorly, since models are overconfident on out-of-domain inputs. Instead, we train a calibrator to identify inputs on which the QA model errs, and abstain when it predicts an error is likely. Crucially, the calibrator benefits from observing the model\u2019s behavior on out-of-domain data, even if from a different domain than the test data. We combine this method with a SQuAD-trained QA model and evaluate on mixtures of SQuAD and five other QA datasets. Our method answers 56% of questions while maintaining 80% accuracy; in contrast, directly using the model\u2019s probabilities only answers 48% at 80% accuracy."
                },
                {
                    "title": "DeBERTa: Decoding-enhanced BERT with Disentangled Attention",
                    "abstract": "Recent progress in pre-trained neural language models has significantly improved the performance of many natural language processing (NLP) tasks. In this paper we propose a new model architecture DeBERTa (Decoding-enhanced BERT with disentangled attention) that improves the BERT and RoBERTa models using two novel techniques. The first is the disentangled attention mechanism, where each word is represented using two vectors that encode its content and position, respectively, and the attention weights among words are computed using disentangled matrices on their contents and relative positions. Second, an enhanced mask decoder is used to replace the output softmax layer to predict the masked tokens for model pretraining. We show that these two techniques significantly improve the efficiency of model pre-training and performance of downstream tasks. Compared to RoBERTa-Large, a DeBERTa model trained on half of the training data performs consistently better on a wide range of NLP tasks, achieving improvements on MNLI by +0.9% (90.2% vs. 91.1%), on SQuAD v2.0 by +2.3% (88.4% vs. 90.7%) and RACE by +3.6% (83.2% vs. 86.8%). The DeBERTa code and pre-trained models will be made publicly available at this https URL."
                },
                {
                    "title": "Consistent Estimators for Learning to Defer to an Expert",
                    "abstract": "Learning algorithms are often used in conjunction with expert decision makers in practical scenarios, however this fact is largely ignored when designing these algorithms. In this paper we explore how to learn predictors that can either predict or choose to defer the decision to a downstream expert. Given only samples of the expert's decisions, we give a procedure based on learning a classifier and a rejector and analyze it theoretically. Our approach is based on a novel reduction to cost sensitive learning where we give a consistent surrogate loss for cost sensitive learning that generalizes the cross entropy loss. We show the effectiveness of our approach on a variety of experimental tasks."
                },
                {
                    "title": "CheXbert: Combining Automatic Labelers and Expert Annotations for Accurate Radiology Report Labeling Using BERT",
                    "abstract": "The extraction of labels from radiology text reports enables large-scale training of medical imaging models. Existing approaches to report labeling typically rely either on sophisticated feature engineering based on medical domain knowledge or manual annotations by experts. In this work, we introduce a BERT-based approach to medical image report labeling that exploits both the scale of available rule-based systems and the quality of expert annotations. We demonstrate superior performance of a biomedically pretrained BERT model first trained on annotations of a rule-based labeler and then finetuned on a small set of expert annotations augmented with automated backtranslation. We find that our final model, CheXbert, is able to outperform the previous best rules-based labeler with statistical significance, setting a new SOTA for report labeling on one of the largest datasets of chest x-rays."
                },
                {
                    "title": "Conformalized Quantile Regression",
                    "abstract": "Conformal prediction is a technique for constructing prediction intervals that attain valid coverage in finite samples, without making distributional assumptions. Despite this appeal, existing conformal methods can be unnecessarily conservative because they form intervals of constant or weakly varying length across the input space. In this paper we propose a new method that is fully adaptive to heteroscedasticity. It combines conformal prediction with classical quantile regression, inheriting the advantages of both. We establish a theoretical guarantee of valid coverage, supplemented by extensive experiments on popular regression datasets. We compare the efficiency of conformalized quantile regression to other conformal methods, showing that our method tends to produce shorter intervals."
                },
                {
                    "title": "Artificial intelligence, bias and clinical safety",
                    "abstract": "In medicine, artificial intelligence (AI) research is becoming increasingly focused on applying machine learning (ML) techniques to complex problems, and so allowing computers to make predictions from large amounts of patient data, by learning their own associations.1 Estimates of the impact of AI on the wider economy globally vary wildly, with a recent report suggesting a 14% effect on global gross domestic product by 2030, half of which coming from productivity improvements.2 These predictions create political appetite for the rapid development of the AI industry,3 and healthcare is a priority area where this technology has yet to be exploited.2 3 The digital health revolution described by Duggal et al 4 is already in full swing with the potential to \u2018disrupt\u2019 healthcare. Health AI research has demonstrated some impressive results,5\u201310 but its clinical value has not yet been realised, hindered partly by a lack of a clear understanding of how to quantify benefit or ensure patient safety, and increasing concerns about the ethical and medico-legal impact.11 \n\nThis analysis is written with the dual aim of helping clinical safety professionals to critically appraise current medical AI research from a quality and safety perspective, and supporting research and development in AI by highlighting some of the clinical safety questions that must be considered if medical application of these exciting technologies is to be successful.\n\nClinical decision support systems (DSS) are in widespread use in medicine and have had most impact providing guidance on the safe prescription of medicines,12 guideline adherence, simple risk screening13 or prognostic scoring.14 These systems use predefined rules, which have predictable behaviour and are usually shown to reduce clinical error,12 although sometimes inadvertently introduce safety issues themselves.15 16 Rules-based systems have also been developed to address diagnostic uncertainty17\u201319 \u2026"
                },
                {
                    "title": "CoQA: A Conversational Question Answering Challenge",
                    "abstract": "Humans gather information through conversations involving a series of interconnected questions and answers. For machines to assist in information gathering, it is therefore essential to enable them to answer conversational questions. We introduce CoQA, a novel dataset for building Conversational Question Answering systems. Our dataset contains 127k questions with answers, obtained from 8k conversations about text passages from seven diverse domains. The questions are conversational, and the answers are free-form text with their corresponding evidence highlighted in the passage. We analyze CoQA in depth and show that conversational questions have challenging phenomena not present in existing reading comprehension datasets (e.g., coreference and pragmatic reasoning). We evaluate strong dialogue and reading comprehension models on CoQA. The best system obtains an F1 score of 65.4%, which is 23.4 points behind human performance (88.8%), indicating that there is ample room for improvement. We present CoQA as a challenge to the community at https://stanfordnlp.github.io/coqa."
                },
                {
                    "title": "Selective Classification for Deep Neural Networks",
                    "abstract": "Selective classification techniques (also known as reject option) have not yet been considered in the context of deep neural networks (DNNs). These techniques can potentially significantly improve DNNs prediction performance by trading-off coverage. In this paper we propose a method to construct a selective classifier given a trained neural network. Our method allows a user to set a desired risk level. At test time, the classifier rejects instances as needed, to grant the desired risk (with high probability). Empirical results over CIFAR and ImageNet convincingly demonstrate the viability of our method, which opens up possibilities to operate DNNs in mission-critical applications. For example, using our method an unprecedented 2% error in top-5 ImageNet classification can be guaranteed with probability 99.9%, and almost 60% test coverage."
                },
                {
                    "title": "TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension",
                    "abstract": "We present TriviaQA, a challenging reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents, six per question on average, that provide high quality distant supervision for answering the questions. We show that, in comparison to other recently introduced large-scale datasets, TriviaQA (1) has relatively complex, compositional questions, (2) has considerable syntactic and lexical variability between questions and corresponding answer-evidence sentences, and (3) requires more cross sentence reasoning to find answers. We also present two baseline algorithms: a feature-based classifier and a state-of-the-art neural network, that performs well on SQuAD reading comprehension. Neither approach comes close to human performance (23% and 40% vs. 80%), suggesting that TriviaQA is a challenging testbed that is worth significant future study."
                },
                {
                    "title": "Distribution-Free Predictive Inference for Regression",
                    "abstract": "ABSTRACT We develop a general framework for distribution-free predictive inference in regression, using conformal inference. The proposed methodology allows for the construction of a prediction band for the response variable using any estimator of the regression function. The resulting prediction band preserves the consistency properties of the original estimator under standard assumptions, while guaranteeing finite-sample marginal coverage even when these assumptions do not hold. We analyze and compare, both empirically and theoretically, the two major variants of our conformal framework: full conformal inference and split conformal inference, along with a related jackknife method. These methods offer different tradeoffs between statistical accuracy (length of resulting prediction intervals) and computational efficiency. As extensions, we develop a method for constructing valid in-sample prediction intervals called rank-one-out conformal inference, which has essentially the same computational efficiency as split conformal inference. We also describe an extension of our procedures for producing prediction bands with locally varying length, to adapt to heteroscedasticity in the data. Finally, we propose a model-free notion of variable importance, called leave-one-covariate-out or LOCO inference. Accompanying this article is an R package conformalInference that implements all of the proposals we have introduced. In the spirit of reproducibility, all of our empirical results can also be easily (re)generated using this package."
                },
                {
                    "title": "On the Foundations of Noise-free Selective Classification",
                    "abstract": "We consider selective classification, a term we adopt here to refer to 'classification with a reject option.' The essence in selective classification is to trade-off classifier coverage for higher accuracy. We term this trade-off the risk-coverage (RC) trade-off. Our main objective is to characterize this trade-off and to construct algorithms that can optimally or near optimally achieve the best possible trade-offs in a controlled manner. For noise-free models we present in this paper a thorough analysis of selective classification including characterizations of RC trade-offs in various interesting settings."
                },
                {
                    "title": "ROUGE: A Package for Automatic Evaluation of Summaries",
                    "abstract": "ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST."
                },
                {
                    "title": "The positive false discovery rate: a Bayesian interpretation and the q-value",
                    "abstract": "Multiple hypothesis testing is concerned with controlling the rate of false positives when testing several hypotheses simultaneously. One multiple hypothesis testing error measure is the false discovery rate (FDR), which is loosely defined to be the expected proportion of false positives among all significant hypotheses. The FDR is especially appropriate for exploratory analyses in which one is interested in finding several significant results among many tests. In this work, we introduce a modified version of the FDR called the \u201cpositive false discovery rate\u201d (pFDR). We discuss the advantages and disadvantages of the pFDR and investigate its statistical properties. When assuming the test statistics follow a mixture distribution, we show that the pFDR can be written as a Bayesian posterior probability and can be connected to classification theory. These properties remain asymptotically true under fairly general conditions, even under certain forms of dependence. Also, a new quantity called the \u201cq-value\u201d is introduced and investigated, which is a natural \u201cBayesian posterior p-value,\u201d or rather the pFDR analogue of the p-value. 1. Introduction. When testing a single hypothesis, one is usually concerned with controlling the false positive rate while maximizing the probability of detecting an effect when one really exists. In statistical terms, we maximize the power conditional on the Type I error rate being at or below some level. The field of multiple hypothesis testing tries to extend this basic paradigm to the situation where several hypotheses are tested simultaneously. One must define an appropriate compound error measure according to the rate of false positives one is willing to encounter. Then a procedure is developed that allows one to control the error rate at a desired level, while maintaining the power of each test as much as possible. The most commonly controlled quantity when testing multiple hypotheses is the family wise error rate (FWER), which is the probability of yielding one or more false positives out of all hypotheses tested. The most familiar example of this is the Bonferroni method. If there are m hypothesis tests, each test is controlled so that the probability of a false positive is less than or equal to \u03b1/m for some chosen value of \u03b1. It then follows that the overall FWER is less than or equal to \u03b1 .M any"
                },
                {
                    "title": "An optimum character recognition system using decision functions",
                    "abstract": "The character recognition problem, usually resulting from characters being corrupted by printing deterioration and/or inherent noise of the devices, is considered from the viewpoint of statistical decision theory. The optimization consists of minimizing the expected risk for a weight function which is preassigned to measure the consequences of system decisions As an alternative minimization of the error rate for a given rejection rate is used as the critenon. The optimum recogition is thus obtained. The optimum system consists of a conditional-probability densisities computer; character channels, one for each character; a rejection channel; and a comparison network. Its precise structure and and ultimate performance depend essentially upon the signals and noise structure. Explicit examples for an additive Gaussian noise and a ``cosine'' noise are presented. Finally, an error-free recognition system and a possible criterion to measure the character style and deteriortation are presented."
                },
                {
                    "title": "Selective-LAMA: Selective Prediction for Confidence-Aware Evaluation of Language Models",
                    "abstract": "Recent studies have suggested that neural language models learn and store a large amount of facts and commonsense knowledge from training data. The ability of language models to restore such knowledge is often evaluated via zero-shot cloze-style QA tasks. However, such evaluations rely only on prediction accuracy without punishing the systems for their mistakes, e.g., simply guessing or hallucinating likely answers. Selective prediction is a more informative evaluation framework that takes the confidence of predictions into account. Under the selective prediction setting, a model is evaluated not only by the number of correct predictions, but also by the ability to filter out dubious predictions by estimating the confidence of individual predictions. Such confidence-aware evaluation is crucial for determining whether to trust zero-shot predictions of language models. In this paper, we apply the selective prediction setting to an existing benchmark, LAMA probe, and conduct extensive experiments with recent neural language models and different confidence functions. We empirically show that our Selective-LAMA evaluation is more robust to the effect of simple guesses than the conventional accuracy-based evaluation.Our evaluation reveals the importance of the choice of confidence functions by showing that simply relying on token probabilities is not always the best choice.Further analysis shows that various confidence functions exhibit different preferences over predicted tokens for a given context."
                },
                {
                    "title": "THE ETHICS OF ARTIFICIAL INTELLIGENCE",
                    "abstract": "Summary There are many ethical questions relating the issue of developing an intelligent system. There is strong and increasing pressure to raise capabilities of the artificial intelligence at least to the human levelled intelligence as the ultimate goal. This essay describes possible paths of development of the artificial intelligence. It is discussed how this changes will affect our society and challenges that humanity will have to face. Principles, guideways and modern viewpoints are presented and confirmed with the statements of the renowned scientists and experts in the field of the artificial intelligence ethics."
                },
                {
                    "title": "Controlling the false discovery rate: a practical and powerful approach to multiple testing",
                    "abstract": "SUMMARY The common approach to the multiplicity problem calls for controlling the familywise error rate (FWER). This approach, though, has faults, and we point out a few. A different approach to problems of multiple significance testing is presented. It calls for controlling the expected proportion of falsely rejected hypotheses -the false discovery rate. This error rate is equivalent to the FWER when all hypotheses are true but is smaller otherwise. Therefore, in problems where the control of the false discovery rate rather than that of the FWER is desired, there is potential for a gain in power. A simple sequential Bonferronitype procedure is proved to control the false discovery rate for independent test statistics, and a simulation study shows that the gain in power is substantial. The use of the new procedure and the appropriateness of the criterion are illustrated with examples."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we certify the alignment of outputs generated by foundation models to ensure their reliability in high-stakes applications?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the pressing issues of factual errors, hallucinations, and biases in foundation models, which can undermine their deployment in critical areas such as healthcare, finance, and autonomous systems. By developing a method that certifies the alignment of model outputs, we can enhance trust in AI systems, leading to safer and more effective applications. This research could pave the way for future studies focused on improving the interpretability and reliability of AI outputs, ultimately advancing knowledge in machine learning and fostering practical applications that align with human values.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of foundation models and the difficulty in quantifying their confidence in generated outputs. Naive approaches may fail because they do not account for the nuanced nature of alignment criteria or the probabilistic nature of model outputs. Additionally, technical obstacles include the need for robust statistical methods to control the false discovery rate (FDR) while ensuring that the selected outputs are trustworthy. The theoretical challenge lies in developing a framework that can generalize across different models and alignment criteria without compromising performance.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on applying conformal prediction to classification and regression tasks, leaving a gap in addressing the alignment of general-form outputs from foundation models. Existing solutions often require manual intervention to determine which outputs are aligned, creating a barrier to practical application. Moreover, past methods may sacrifice model performance for the sake of ensuring factuality, which limits their effectiveness. Our approach differs by providing a systematic framework that certifies outputs as aligned, allowing for automated deployment without the need for further modifications.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Conformal Alignment (CA), leverages conformal prediction principles to certify the alignment of outputs from foundation models. We will utilize a dataset of model-generated outputs and a holdout set of high-quality reference data to guide the selection process. The key metric for evaluation will be the control of the false discovery rate (FDR) in selecting trustworthy outputs. We expect that our approach will yield a reliable set of certified outputs that can be deployed"
            }
        },
        "author_data": {
            "fe600963-8d7f-4b63-99dd-1df2447d1a32": {
                "pk": "fe600963-8d7f-4b63-99dd-1df2447d1a32",
                "name": "Yu Gui",
                "collaborators": [],
                "domain": [
                    "Distributed Learning",
                    "Feature Selection",
                    "Data Privacy",
                    "Statistical Methods"
                ],
                "publications": [
                    {
                        "title": "ADAGES: adaptive aggregation with stability for distributed feature selection",
                        "abstract": "In this era of \"big\" data, not only the large amount of data keeps motivating distributed computing, but concerns on data privacy also put forward the emphasis on distributed learning. To conduct feature selection and to control the false discovery rate in a distributed pattern with multi-machines or multi-institutions, an efficient aggregation method is necessary. In this paper, we propose an adaptive aggregation method called ADAGES which can be flexibly applied to any machine-wise feature selection method. We will show that our method is capable of controlling the overall FDR with a theoretical foundation while maintaining power as good as the Union aggregation rule in practice."
                    }
                ]
            },
            "079b982c-af96-44f2-b6c2-08a0e592d905": {
                "pk": "079b982c-af96-44f2-b6c2-08a0e592d905",
                "name": "Ying Jin",
                "collaborators": [
                    "Dominik Rothenh\u00e4usler",
                    "Zhimei Ren",
                    "Jiaqi Wang",
                    "Dahua Lin",
                    "Emmanuel J. Cand\u00e8s",
                    "Shan Ba",
                    "Andrew Leroux",
                    "Pengfei Hou",
                    "Kevin Guo",
                    "Zhou Fang"
                ],
                "domain": [
                    "Statistical Inference",
                    "Machine Learning",
                    "Domain Adaptation",
                    "Semi-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Upper bounds on the Natarajan dimensions of some function classes",
                        "abstract": "The Natarajan dimension is a fundamental tool for characterizing multi-class PAC learnability, generalizing the Vapnik-Chervonenkis (VC) dimension from binary to multi-class classification problems. This work establishes upper bounds on Natarajan dimensions for certain function classes, including (i) multi-class decision tree and random forests, and (ii) multi-class neural networks with binary, linear and ReLU activations. These results may be relevant for describing the performance of certain multi-class learning algorithms."
                    },
                    {
                        "title": "Towards Optimal Variance Reduction in Online Controlled Experiments",
                        "abstract": "We study optimal variance reduction solutions for count and ratio metrics in online controlled experiments. Our methods leverage flexible machine learning tools to incorporate covariates that are independent from the treatment but have predictive power for the outcomes, and employ the cross-fitting technique to remove the bias in complex machine learning models. We establish CLT-type asymptotic inference based on our estimators under mild convergence conditions. Our procedures are optimal (efficient) for the corresponding targets as long as the machine learning estimators are consistent, without any requirement for their convergence rates. In complement to the general optimal procedure, we also derive a linear adjustment method for ratio metrics as a special case that is computationally efficient and can flexibly incorporate any pre-treatment covariates. We evaluate the proposed variance reduction procedures with comprehensive simulation studies and provide practical suggestions regarding commonly adopted assumptions in computing ratio metrics. When tested on real online experiment data from LinkedIn, the proposed optimal procedure for ratio metrics can reduce up to 80\\% of variance compared to the standard difference-in-mean estimator and also further reduce up to 30\\% of variance compared to the CUPED approach by going beyond linearity and incorporating a large number of extra covariates."
                    },
                    {
                        "title": "Modular Regression: Improving Linear Models by Incorporating Auxiliary Data",
                        "abstract": "This paper develops a new framework, called modular regression, to utilize auxiliary information -- such as variables other than the original features or additional data sets -- in the training process of linear models. At a high level, our method follows the routine: (i) decomposing the regression task into several sub-tasks, (ii) fitting the sub-task models, and (iii) using the sub-task models to provide an improved estimate for the original regression problem. This routine applies to widely-used low-dimensional (generalized) linear models and high-dimensional regularized linear regression. It also naturally extends to missing-data settings where only partial observations are available. By incorporating auxiliary information, our approach improves the estimation efficiency and prediction accuracy upon linear regression or the Lasso under a conditional independence assumption for predicting the outcome. For high-dimensional settings, we develop an extension of our procedure that is robust to violations of the conditional independence assumption, in the sense that it improves efficiency if this assumption holds and coincides with the Lasso otherwise. We demonstrate the efficacy of our methods with simulated and real data sets."
                    },
                    {
                        "title": "Tailored inference for finite populations: conditional validity and transfer across distributions",
                        "abstract": "Parameters of sub-populations can be more relevant than super-population ones. For example, a healthcare provider may be interested in the effect of a treatment plan for a specific subset of their patients; policymakers may be concerned with the impact of a policy in a particular state within a given population. In these cases, the focus is on a specific finite population, as opposed to an infinite super-population. Such a population can be characterized by fixing some attributes that are intrinsic to them, leaving unexplained variations like measurement error as random. Inference for a population with fixed attributes can then be modeled as inferring parameters of a conditional distribution. Accordingly, it is desirable that confidence intervals are conditionally valid for the realized population, instead of marginalizing over many possible draws of populations.   We provide a statistical inference framework for parameters of finite populations with known attributes. Leveraging the attribute information, our estimators and confidence intervals closely target a specific finite population. When the data is from the population of interest, our confidence intervals attain asymptotic conditional validity given the attributes, and are shorter than those for super-population inference. In addition, we develop procedures to infer parameters of new populations with differing covariate distributions; the confidence intervals are also conditionally valid for the new populations under mild conditions. Our methods extend to situations where the fixed information has a weaker structure or is only partially observed. We demonstrate the validity and applicability of our methods using simulated and real-world data."
                    },
                    {
                        "title": "Comparing estimators of discriminative performance of time-to-event models",
                        "abstract": "Predicting the timing and occurrence of events is a major focus of data science applications, especially in the context of biomedical research. Performance for models estimating these outcomes, often referred to as time-to-event or survival outcomes, is frequently summarized using measures of discrimination, in particular time-dependent AUC and concordance. Many estimators for these quantities have been proposed which can be broadly categorized as either semi-parametric estimators or non-parametric estimators. In this paper, we review various estimators' mathematical construction and compare the behavior of the two classes of estimators. Importantly, we identify a previously unknown feature of the class of semi-parametric estimators that can result in vastly over-optimistic out-of-sample estimation of discriminative performance in common applied tasks. Although these semi-parametric estimators are popular in practice, the phenomenon we identify here suggests this class of estimators may be inappropriate for use in model assessment and selection based on out-of-sample evaluation criteria. This is due to the semi-parametric estimators' bias in favor of models that are overfit when using out-of-sample prediction criteria (e.g., cross validation). Non-parametric estimators, which do not exhibit this behavior, are highly variable for local discrimination. We propose to address the high variability problem through penalized regression splines smoothing. The behavior of various estimators of time-dependent AUC and concordance are illustrated via a simulation study using two different mechanisms that produce over-optimistic out-of-sample estimates using semi-parametric estimators. Estimators are further compared using a case study using data from the National Health and Nutrition Examination Survey (NHANES) 2011-2014."
                    },
                    {
                        "title": "Single-level Optimization For Differential Architecture Search",
                        "abstract": "In this paper, we point out that differential architecture search (DARTS) makes gradient of architecture parameters biased for network weights and architecture parameters are updated in different datasets alternatively in the bi-level optimization framework. The bias causes the architecture parameters of non-learnable operations to surpass that of learnable operations. Moreover, using softmax as architecture parameters' activation function and inappropriate learning rate would exacerbate the bias. As a result, it's frequently observed that non-learnable operations are dominated in the search phase. To reduce the bias, we propose to use single-level to replace bi-level optimization and non-competitive activation function like sigmoid to replace softmax. As a result, we could search high-performance architectures steadily. Experiments on NAS Benchmark 201 validate our hypothesis and stably find out nearly the optimal architecture. On DARTS space, we search the state-of-the-art architecture with 77.0% top1 accuracy (training setting follows PDARTS and without any additional module) on ImageNet-1K and steadily search architectures up-to 76.5% top1 accuracy (but not select the best from the searched architectures) which is comparable with current reported best result."
                    },
                    {
                        "title": "Confidence on the Focal: Conformal Prediction with Selection-Conditional Coverage",
                        "abstract": "Conformal prediction builds marginally valid prediction intervals that cover the unknown outcome of a randomly drawn new test point with a prescribed probability. However, a common scenario in practice is that, after seeing the data, practitioners decide which test unit(s) to focus on in a data-driven manner and seek for uncertainty quantification of the focal unit(s). In such cases, marginally valid conformal prediction intervals may not provide valid coverage for the focal unit(s) due to selection bias. This paper presents a general framework for constructing a prediction set with finite-sample exact coverage conditional on the unit being selected by a given procedure. The general form of our method works for arbitrary selection rules that are invariant to the permutation of the calibration units, and generalizes Mondrian Conformal Prediction to multiple test units and non-equivariant classifiers. We then work out the computationally efficient implementation of our framework for a number of realistic selection rules, including top-K selection, optimization-based selection, selection based on conformal p-values, and selection based on properties of preliminary conformal prediction sets. The performance of our methods is demonstrated via applications in drug discovery and health risk prediction."
                    },
                    {
                        "title": "Semi-Supervised Semantic Segmentation via Gentle Teaching Assistant",
                        "abstract": "Semi-Supervised Semantic Segmentation aims at training the segmentation model with limited labeled data and a large amount of unlabeled data. To effectively leverage the unlabeled data, pseudo labeling, along with the teacher-student framework, is widely adopted in semi-supervised semantic segmentation. Though proved to be effective, this paradigm suffers from incorrect pseudo labels which inevitably exist and are taken as auxiliary training data. To alleviate the negative impact of incorrect pseudo labels, we delve into the current Semi-Supervised Semantic Segmentation frameworks. We argue that the unlabeled data with pseudo labels can facilitate the learning of representative features in the feature extractor, but it is unreliable to supervise the mask predictor. Motivated by this consideration, we propose a novel framework, Gentle Teaching Assistant (GTA-Seg) to disentangle the effects of pseudo labels on feature extractor and mask predictor of the student model. Specifically, in addition to the original teacher-student framework, our method introduces a teaching assistant network which directly learns from pseudo labels generated by the teacher network. The gentle teaching assistant (GTA) is coined gentle since it only transfers the beneficial feature representation knowledge in the feature extractor to the student model in an Exponential Moving Average (EMA) manner, protecting the student model from the negative influences caused by unreliable pseudo labels in the mask predictor. The student model is also supervised by reliable labeled data to train an accurate mask predictor, further facilitating feature representation. Extensive experiment results on benchmark datasets validate that our method shows competitive performance against previous methods. Code is available at https://github.com/Jin-Ying/GTA-Seg."
                    },
                    {
                        "title": "Sensitivity Analysis of Individual Treatment Effects: A Robust Conformal Inference Approach",
                        "abstract": "We propose a model-free framework for sensitivity analysis of individual treatment effects (ITEs), building upon ideas from conformal inference. For any unit, our procedure reports the $\\Gamma$-value, a number which quantifies the minimum strength of confounding needed to explain away the evidence for ITE. Our approach rests on the reliable predictive inference of counterfactuals and ITEs in situations where the training data is confounded. Under the marginal sensitivity model of Tan (2006), we characterize the shift between the distribution of the observations and that of the counterfactuals. We first develop a general method for predictive inference of test samples from a shifted distribution; we then leverage this to construct covariate-dependent prediction sets for counterfactuals. No matter the value of the shift, these prediction sets (resp. approximately) achieve marginal coverage if the propensity score is known exactly (resp. estimated). We describe a distinct procedure also attaining coverage, however, conditional on the training data. In the latter case, we prove a sharpness result showing that for certain classes of prediction problems, the prediction intervals cannot possibly be tightened. We verify the validity and performance of the new methods via simulation studies and apply them to analyze real datasets."
                    },
                    {
                        "title": "Selection by Prediction with Conformal p-values",
                        "abstract": "Decision making or scientific discovery pipelines such as job hiring and drug discovery often involve multiple stages: before any resource-intensive step, there is often an initial screening that uses predictions from a machine learning model to shortlist a few candidates from a large pool. We study screening procedures that aim to select candidates whose unobserved outcomes exceed user-specified values. We develop a method that wraps around any prediction model to produce a subset of candidates while controlling the proportion of falsely selected units. Building upon the conformal inference framework, our method first constructs p-values that quantify the statistical evidence for large outcomes; it then determines the shortlist by comparing the p-values to a threshold introduced in the multiple testing literature. In many cases, the procedure selects candidates whose predictions are above a data-dependent threshold. Our theoretical guarantee holds under mild exchangeability conditions on the samples, generalizing existing results on multiple conformal p-values. We demonstrate the empirical performance of our method via simulations, and apply it to job hiring and drug discovery datasets."
                    },
                    {
                        "title": "Diagnosing the role of observable distribution shift in scientific replications",
                        "abstract": "Many researchers have identified distribution shift as a likely contributor to the reproducibility crisis in behavioral and biomedical sciences. The idea is that if treatment effects vary across individual characteristics and experimental contexts, then studies conducted in different populations will estimate different average effects. This paper uses ``generalizability\" methods to quantify how much of the effect size discrepancy between an original study and its replication can be explained by distribution shift on observed unit-level characteristics. More specifically, we decompose this discrepancy into ``components\" attributable to sampling variability (including publication bias), observable distribution shifts, and residual factors. We compute this decomposition for several directly-replicated behavioral science experiments and find little evidence that observable distribution shifts contribute appreciably to non-replicability. In some cases, this is because there is too much statistical noise. In other cases, there is strong evidence that controlling for additional moderators is necessary for reliable replication."
                    },
                    {
                        "title": "SepRep-Net: Multi-source Free Domain Adaptation via Model Separation And Reparameterization",
                        "abstract": "We consider multi-source free domain adaptation, the problem of adapting multiple existing models to a new domain without accessing the source data. Among existing approaches, methods based on model ensemble are effective in both the source and target domains, but incur significantly increased computational costs. Towards this dilemma, in this work, we propose a novel framework called SepRep-Net, which tackles multi-source free domain adaptation via model Separation and Reparameterization.Concretely, SepRep-Net reassembled multiple existing models to a unified network, while maintaining separate pathways (Separation). During training, separate pathways are optimized in parallel with the information exchange regularly performed via an additional feature merging unit. With our specific design, these pathways can be further reparameterized into a single one to facilitate inference (Reparameterization). SepRep-Net is characterized by 1) effectiveness: competitive performance on the target domain, 2) efficiency: low computational costs, and 3) generalizability: maintaining more source knowledge than existing solutions. As a general approach, SepRep-Net can be seamlessly plugged into various methods. Extensive experiments validate the performance of SepRep-Net on mainstream benchmarks."
                    },
                    {
                        "title": "Incorporating planning intelligence into deep learning: A planning support tool for street network design",
                        "abstract": "Deep learning applications in shaping ad hoc planning proposals are limited by the difficulty in integrating professional knowledge about cities with artificial intelligence. We propose a novel, complementary use of deep neural networks and planning guidance to automate street network generation that can be context-aware, example-based and user-guided. The model tests suggest that the incorporation of planning knowledge (e.g., road junctions and neighborhood types) in the model training leads to a more realistic prediction of street configurations. Furthermore, the new tool provides both professional and lay users an opportunity to systematically and intuitively explore benchmark proposals for comparisons and further evaluations."
                    }
                ]
            },
            "80d666e6-624a-4296-80fe-8dc2200c8a28": {
                "pk": "80d666e6-624a-4296-80fe-8dc2200c8a28",
                "name": "Zhimei Ren",
                "collaborators": [
                    "Ying Jin",
                    "Zhengyuan Zhou",
                    "Rina Foygel Barber",
                    "Emmanuel Cand\u00e8s",
                    "Emmanuel J. Cand\u00e8s",
                    "Jiahao Ai",
                    "Junu Lee",
                    "Yuting Wei",
                    "Lihua Lei",
                    "Ruohan Zhan"
                ],
                "domain": [
                    "Statistical Inference",
                    "Causal Inference",
                    "Machine Learning",
                    "Bandit Algorithms"
                ],
                "publications": [
                    {
                        "title": "Knockoffs with Side Information",
                        "abstract": "We consider the problem of assessing the importance of multiple variables or factors from a dataset when side information is available. In principle, using side information can allow the statistician to pay attention to variables with a greater potential, which in turn, may lead to more discoveries. We introduce an adaptive knockoff filter, which generalizes the knockoff procedure (Barber and Cand\\`es, 2015; Cand\\`es et al., 2018) in that it uses both the data at hand and side information to adaptively order the variables under study and focus on those that are most promising. Adaptive knockoffs controls the finite-sample false discovery rate (FDR) and we demonstrate its power by comparing it with other structured multiple testing methods. We also apply our methodology to real genetic data in order to find associations between genetic variants and various phenotypes such as Crohn's disease and lipid levels. Here, adaptive knockoffs makes more discoveries than reported in previous studies on the same datasets."
                    },
                    {
                        "title": "Not all distributional shifts are equal: Fine-grained robust conformal inference",
                        "abstract": "We introduce a fine-grained framework for uncertainty quantification of predictive models under distributional shifts. This framework distinguishes the shift in covariate distributions from that in the conditional relationship between the outcome ($Y$) and the covariates ($X$). We propose to reweight the training samples to adjust for an identifiable covariate shift while protecting against worst-case conditional distribution shift bounded in an $f$-divergence ball. Based on ideas from conformal inference and distributionally robust learning, we present an algorithm that outputs (approximately) valid and efficient prediction intervals in the presence of distributional shifts. As a use case, we apply the framework to sensitivity analysis of individual treatment effects with hidden confounding. The proposed methods are evaluated in simulation studies and three real data applications, demonstrating superior robustness and efficiency compared with existing benchmarks."
                    },
                    {
                        "title": "Confidence on the Focal: Conformal Prediction with Selection-Conditional Coverage",
                        "abstract": "Conformal prediction builds marginally valid prediction intervals that cover the unknown outcome of a randomly drawn new test point with a prescribed probability. However, a common scenario in practice is that, after seeing the data, practitioners decide which test unit(s) to focus on in a data-driven manner and seek for uncertainty quantification of the focal unit(s). In such cases, marginally valid conformal prediction intervals may not provide valid coverage for the focal unit(s) due to selection bias. This paper presents a general framework for constructing a prediction set with finite-sample exact coverage conditional on the unit being selected by a given procedure. The general form of our method works for arbitrary selection rules that are invariant to the permutation of the calibration units, and generalizes Mondrian Conformal Prediction to multiple test units and non-equivariant classifiers. We then work out the computationally efficient implementation of our framework for a number of realistic selection rules, including top-K selection, optimization-based selection, selection based on conformal p-values, and selection based on properties of preliminary conformal prediction sets. The performance of our methods is demonstrated via applications in drug discovery and health risk prediction."
                    },
                    {
                        "title": "Dynamic Batch Learning in High-Dimensional Sparse Linear Contextual Bandits",
                        "abstract": "We study the problem of dynamic batch learning in high-dimensional sparse linear contextual bandits, where a decision maker, under a given maximum-number-of-batch constraint and only able to observe rewards at the end of each batch, can dynamically decide how many individuals to include in the next batch (at the end of the current batch) and what personalized action-selection scheme to adopt within each batch. Such batch constraints are ubiquitous in a variety of practical contexts, including personalized product offerings in marketing and medical treatment selection in clinical trials. We characterize the fundamental learning limit in this problem via a regret lower bound and provide a matching upper bound (up to log factors), thus prescribing an optimal scheme for this problem. To the best of our knowledge, our work provides the first inroad into a theoretical understanding of dynamic batch learning in high-dimensional sparse linear contextual bandits. Notably, even a special case of our result -- when no batch constraint is present -- yields that the simple exploration-free algorithm using the LASSO estimator already achieves the minimax optimal regret bound for standard online learning in high-dimensional linear contextual bandits (for the no-margin case), a result that appears unknown in the emerging literature of high-dimensional contextual bandits."
                    },
                    {
                        "title": "Boosting e-BH via conditional calibration",
                        "abstract": "The e-BH procedure is an e-value-based multiple testing procedure that provably controls the false discovery rate (FDR) under any dependence structure between the e-values. Despite this appealing theoretical FDR control guarantee, the e-BH procedure often suffers from low power in practice. In this paper, we propose a general framework that boosts the power of e-BH without sacrificing its FDR control under arbitrary dependence. This is achieved by the technique of conditional calibration, where we take as input the e-values and calibrate them to be a set of \"boosted e-values\" that are guaranteed to be no less -- and are often more -- powerful than the original ones. Our general framework is explicitly instantiated in three classes of multiple testing problems: (1) testing under parametric models, (2) conditional independence testing under the model-X setting, and (3) model-free conformalized selection. Extensive numerical experiments show that our proposed method significantly improves the power of e-BH while continuing to control the FDR. We also demonstrate the effectiveness of our method through an application to an observational study dataset for identifying individuals whose counterfactuals satisfy certain properties."
                    },
                    {
                        "title": "Derandomizing Knockoffs",
                        "abstract": "Model-X knockoffs is a general procedure that can leverage any feature importance measure to produce a variable selection algorithm, which discovers true effects while rigorously controlling the number or fraction of false positives. Model-X knockoffs is a randomized procedure which relies on the one-time construction of synthetic (random) variables. This paper introduces a derandomization method by aggregating the selection results across multiple runs of the knockoffs algorithm. The derandomization step is designed to be flexible and can be adapted to any variable selection base procedure to yield stable decisions without compromising statistical power. When applied to the base procedure of Janson et al. (2016), we prove that derandomized knockoffs controls both the per family error rate (PFER) and the k family-wise error rate (k-FWER). Further, we carry out extensive numerical studies demonstrating tight type-I error control and markedly enhanced power when compared with alternative variable selection algorithms. Finally, we apply our approach to multi-stage genome-wide association studies of prostate cancer and report locations on the genome that are significantly associated with the disease. When cross-referenced with other studies, we find that the reported associations have been replicated."
                    },
                    {
                        "title": "Conformalized Survival Analysis",
                        "abstract": "Existing survival analysis techniques heavily rely on strong modelling assumptions and are, therefore, prone to model misspecification errors. In this paper, we develop an inferential method based on ideas from conformal prediction, which can wrap around any survival prediction algorithm to produce calibrated, covariate-dependent lower predictive bounds on survival times. In the Type I right-censoring setting, when the censoring times are completely exogenous, the lower predictive bounds have guaranteed coverage in finite samples without any assumptions other than that of operating on independent and identically distributed data points. Under a more general conditionally independent censoring assumption, the bounds satisfy a doubly robust property which states the following: marginal coverage is approximately guaranteed if either the censoring mechanism or the conditional survival function is estimated well. Further, we demonstrate that the lower predictive bounds remain valid and informative for other types of censoring. The validity and efficiency of our procedure are demonstrated on synthetic data and real COVID-19 data from the UK Biobank."
                    },
                    {
                        "title": "Derandomized knockoffs: leveraging e-values for false discovery rate control",
                        "abstract": "Model-X knockoffs is a flexible wrapper method for high-dimensional regression algorithms, which provides guaranteed control of the false discovery rate (FDR). Due to the randomness inherent to the method, different runs of model-X knockoffs on the same dataset often result in different sets of selected variables, which is undesirable in practice. In this paper, we introduce a methodology for derandomizing model-X knockoffs with provable FDR control. The key insight of our proposed method lies in the discovery that the knockoffs procedure is in essence an e-BH procedure. We make use of this connection, and derandomize model-X knockoffs by aggregating the e-values resulting from multiple knockoff realizations. We prove that the derandomized procedure controls the FDR at the desired level, without any additional conditions (in contrast, previously proposed methods for derandomization are not able to guarantee FDR control). The proposed method is evaluated with numerical experiments, where we find that the derandomized procedure achieves comparable power and dramatically decreased selection variability when compared with model-X knockoffs."
                    },
                    {
                        "title": "Policy Learning with Adaptively Collected Data",
                        "abstract": "Learning optimal policies from historical data enables personalization in a wide variety of applications including healthcare, digital recommendations, and online education. The growing policy learning literature focuses on settings where the data collection rule stays fixed throughout the experiment. However, adaptive data collection is becoming more common in practice, from two primary sources: 1) data collected from adaptive experiments that are designed to improve inferential efficiency; 2) data collected from production systems that progressively evolve an operational policy to improve performance over time (e.g. contextual bandits). Yet adaptivity complicates the optimal policy identification ex post, since samples are dependent, and each treatment may not receive enough observations for each type of individual. In this paper, we make initial research inquiries into addressing the challenges of learning the optimal policy with adaptively collected data. We propose an algorithm based on generalized augmented inverse propensity weighted (AIPW) estimators, which non-uniformly reweight the elements of a standard AIPW estimator to control worst-case estimation variance. We establish a finite-sample regret upper bound for our algorithm and complement it with a regret lower bound that quantifies the fundamental difficulty of policy learning with adaptive data. When equipped with the best weighting scheme, our algorithm achieves minimax rate optimal regret guarantees even with diminishing exploration. Finally, we demonstrate our algorithm's effectiveness using both synthetic data and public benchmark datasets."
                    },
                    {
                        "title": "Batched Multi-armed Bandits Problem",
                        "abstract": "In this paper, we study the multi-armed bandit problem in the batched setting where the employed policy must split data into a small number of batches. While the minimax regret for the two-armed stochastic bandits has been completely characterized in \\cite{perchet2016batched}, the effect of the number of arms on the regret for the multi-armed case is still open. Moreover, the question whether adaptively chosen batch sizes will help to reduce the regret also remains underexplored. In this paper, we propose the BaSE (batched successive elimination) policy to achieve the rate-optimal regrets (within logarithmic factors) for batched multi-armed bandits, with matching lower bounds even if the batch sizes are determined in an adaptive manner."
                    },
                    {
                        "title": "Sensitivity Analysis of Individual Treatment Effects: A Robust Conformal Inference Approach",
                        "abstract": "We propose a model-free framework for sensitivity analysis of individual treatment effects (ITEs), building upon ideas from conformal inference. For any unit, our procedure reports the $\\Gamma$-value, a number which quantifies the minimum strength of confounding needed to explain away the evidence for ITE. Our approach rests on the reliable predictive inference of counterfactuals and ITEs in situations where the training data is confounded. Under the marginal sensitivity model of Tan (2006), we characterize the shift between the distribution of the observations and that of the counterfactuals. We first develop a general method for predictive inference of test samples from a shifted distribution; we then leverage this to construct covariate-dependent prediction sets for counterfactuals. No matter the value of the shift, these prediction sets (resp. approximately) achieve marginal coverage if the propensity score is known exactly (resp. estimated). We describe a distinct procedure also attaining coverage, however, conditional on the training data. In the latter case, we prove a sharpness result showing that for certain classes of prediction problems, the prediction intervals cannot possibly be tightened. We verify the validity and performance of the new methods via simulation studies and apply them to analyze real datasets."
                    },
                    {
                        "title": "Sensitivity analysis under the $f$-sensitivity models: a distributional robustness perspective",
                        "abstract": "This paper introduces the $f$-sensitivity model, a new sensitivity model that characterizes the violation of unconfoundedness in causal inference. It assumes the selection bias due to unmeasured confounding is bounded \"on average\"; compared with the widely used point-wise sensitivity models in the literature, it is able to capture the strength of unmeasured confounding by not only its magnitude but also the chance of encountering such a magnitude.   We propose a framework for sensitivity analysis under our new model based on a distributional robustness perspective. We first show that the bounds on counterfactual means under the f-sensitivity model are optimal solutions to a new class of distributionally robust optimization (DRO) programs, whose dual forms are essentially risk minimization problems. We then construct point estimators for these bounds by applying a novel debiasing technique to the output of the corresponding empirical risk minimization (ERM) problems. Our estimators are shown to converge to valid bounds on counterfactual means if any nuisance component can be estimated consistently, and to the exact bounds when the ERM step is additionally consistent. We further establish asymptotic normality and Wald-type inference for these estimators under slower-than-root-n convergence rates of the estimated nuisance components. Finally, the performance of our method is demonstrated with numerical experiments."
                    },
                    {
                        "title": "Online Multi-Armed Bandits with Adaptive Inference",
                        "abstract": "During online decision making in Multi-Armed Bandits (MAB), one needs to conduct inference on the true mean reward of each arm based on data collected so far at each step. However, since the arms are adaptively selected--thereby yielding non-iid data--conducting inference accurately is not straightforward. In particular, sample averaging, which is used in the family of UCB and Thompson sampling (TS) algorithms, does not provide a good choice as it suffers from bias and a lack of good statistical properties (e.g. asymptotic normality). Our thesis in this paper is that more sophisticated inference schemes that take into account the adaptive nature of the sequentially collected data can unlock further performance gains, even though both UCB and TS type algorithms are optimal in the worst case. In particular, we propose a variant of TS-style algorithms--which we call doubly adaptive TS--that leverages recent advances in causal inference and adaptively reweights the terms of a doubly robust estimator on the true mean reward of each arm. Through 20 synthetic domain experiments and a semi-synthetic experiment based on data from an A/B test of a web service, we demonstrate that using an adaptive inferential scheme (while still retaining the exploration efficacy of TS) provides clear benefits in online decision making: the proposed DATS algorithm has superior empirical performance to existing baselines (UCB and TS) in terms of regret and sample complexity in identifying the best arm. In addition, we also provide a finite-time regret bound of doubly adaptive TS that matches (up to log factors) those of UCB and TS algorithms, thereby establishing that its improved practical benefits do not come at the expense of worst-case suboptimality."
                    },
                    {
                        "title": "Policy learning \"without\" overlap: Pessimism and generalized empirical Bernstein's inequality",
                        "abstract": "This paper studies offline policy learning, which aims at utilizing observations collected a priori (from either fixed or adaptively evolving behavior policies) to learn an optimal individualized decision rule that achieves the best overall outcomes for a given population. Existing policy learning methods rely on a uniform overlap assumption, i.e., the propensities of exploring all actions for all individual characteristics must be lower bounded. As one has no control over the data collection process, this assumption can be unrealistic in many situations, especially when the behavior policies are allowed to evolve over time with diminishing propensities for certain actions.   In this paper, we propose Pessimistic Policy Learning (PPL), a new algorithm that optimizes lower confidence bounds (LCBs) -- instead of point estimates -- of the policy values. The LCBs are constructed using knowledge of the behavior policies for collecting the offline data. Without assuming any uniform overlap condition, we establish a data-dependent upper bound for the suboptimality of our algorithm, which only depends on (i) the overlap for the optimal policy, and (ii) the complexity of the policy class we optimize over. As an implication, for adaptively collected data, we ensure efficient policy learning as long as the propensities for optimal actions are lower bounded over time, while those for suboptimal ones are allowed to diminish arbitrarily fast. In our theoretical analysis, we develop a new self-normalized type concentration inequality for inverse-propensity-weighting estimators, generalizing the well-known empirical Bernstein's inequality to unbounded and non-i.i.d. data. We complement our theory with an efficient optimization algorithm via Majorization-Minimization and policy tree search, as well as extensive simulation studies and real-world applications that demonstrate the efficacy of PPL."
                    },
                    {
                        "title": "Transfer learning in genome-wide association studies with knockoffs",
                        "abstract": "This paper presents and compares alternative transfer learning methods that can increase the power of conditional testing via knockoffs by leveraging prior information in external data sets collected from different populations or measuring related outcomes. The relevance of this methodology is explored in particular within the context of genome-wide association studies, where it can be helpful to address the pressing need for principled ways to suitably account for, and efficiently learn from the genetic variation associated to diverse ancestries. Finally, we apply these methods to analyze several phenotypes in the UK Biobank data set, demonstrating that transfer learning helps knockoffs discover more numerous associations in the data collected from minority populations, potentially opening the way to the development of more accurate polygenic risk scores."
                    },
                    {
                        "title": "Iterative Approximate Cross-Validation",
                        "abstract": "Cross-validation (CV) is one of the most popular tools for assessing and selecting predictive models. However, standard CV suffers from high computational cost when the number of folds is large. Recently, under the empirical risk minimization (ERM) framework, a line of works proposed efficient methods to approximate CV based on the solution of the ERM problem trained on the full dataset. However, in large-scale problems, it can be hard to obtain the exact solution of the ERM problem, either due to limited computational resources or due to early stopping as a way of preventing overfitting. In this paper, we propose a new paradigm to efficiently approximate CV when the ERM problem is solved via an iterative first-order algorithm, without running until convergence. Our new method extends existing guarantees for CV approximation to hold along the whole trajectory of the algorithm, including at convergence, thus generalizing existing CV approximation methods. Finally, we illustrate the accuracy and computational efficiency of our method through a range of empirical studies."
                    },
                    {
                        "title": "Conformalized survival analysis with adaptive cutoffs",
                        "abstract": "This paper introduces an assumption-lean method that constructs valid and efficient lower predictive bounds (LPBs) for survival times with censored data. We build on recent work by Cand\\`es et al. (2021), whose approach first subsets the data to discard any data points with early censoring times, and then uses a reweighting technique (namely, weighted conformal inference (Tibshirani et al., 2019)) to correct for the distribution shift introduced by this subsetting procedure.   For our new method, instead of constraining to a fixed threshold for the censoring time when subsetting the data, we allow for a covariate-dependent and data-adaptive subsetting step, which is better able to capture the heterogeneity of the censoring mechanism. As a result, our method can lead to LPBs that are less conservative and give more accurate information. We show that in the Type I right-censoring setting, if either of the censoring mechanism or the conditional quantile of survival time is well estimated, our proposed procedure achieves nearly exact marginal coverage, where in the latter case we additionally have approximate conditional coverage. We evaluate the validity and efficiency of our proposed algorithm in numerical experiments, illustrating its advantage when compared with other competing methods. Finally, our method is applied to a real dataset to generate LPBs for users' active times on a mobile app."
                    },
                    {
                        "title": "Simultaneous activity and attenuation estimation in TOF-PET with TV-constrained nonconvex optimization",
                        "abstract": "An alternating direction method of multipliers (ADMM) framework is developed for nonsmooth biconvex optimization for inverse problems in imaging. In particular, the simultaneous estimation of activity and attenuation (SAA) problem in time-of-flight positron emission tomography (TOF-PET) has such a structure when maximum likelihood estimation (MLE) is employed. The ADMM framework is applied to MLE for SAA in TOF-PET, resulting in the ADMM-SAA algorithm. This algorithm is extended by imposing total variation (TV) constraints on both the activity and attenuation map, resulting in the ADMM-TVSAA algorithm. The performance of this algorithm is illustrated using the penalized maximum likelihood activity and attenuation estimation (P-MLAA) algorithm as a reference. Additional results on step-size tuning and on the use of unconstrained ADMM-SAA are presented in the previous arXiv submission: arXiv:2303.17042v1."
                    }
                ]
            }
        }
    },
    "2401.11204": {
        "paper_data": {
            "title": "Towards Category Unification of 3D Single Object Tracking on Point Clouds",
            "url": "http://arxiv.org/abs/2401.11204v2",
            "arxiv_id": "2401.11204",
            "authors": [
                "Jiahao Nie",
                "Zhiwei He",
                "Xudong Lv",
                "Xueyi Zhou",
                "Dong-Kyu Chae",
                "Fei Xie"
            ],
            "abstract": "Category-specific models are provenly valuable methods in 3D single object tracking (SOT) regardless of Siamese or motion-centric paradigms. However, such over-specialized model designs incur redundant parameters, thus limiting the broader applicability of 3D SOT task. This paper first introduces unified models that can simultaneously track objects across all categories using a single network with shared model parameters. Specifically, we propose to explicitly encode distinct attributes associated to different object categories, enabling the model to adapt to cross-category data. We find that the attribute variances of point cloud objects primarily occur from the varying size and shape (e.g., large and square vehicles v.s. small and slender humans). Based on this observation, we design a novel point set representation learning network inheriting transformer architecture, termed AdaFormer, which adaptively encodes the dynamically varying shape and size information from cross-category data in a unified manner. We further incorporate the size and shape prior derived from the known template targets into the model's inputs and learning objective, facilitating the learning of unified representation. Equipped with such designs, we construct two category-unified models SiamCUT and MoCUT.Extensive experiments demonstrate that SiamCUT and MoCUT exhibit strong generalization and training stability. Furthermore, our category-unified models outperform the category-specific counterparts by a significant margin (e.g., on KITTI dataset, 12% and 3% performance gains on the Siamese and motion paradigms). Our code will be available.",
            "introduction": "  Appendix A Detailed analysis of Unified Components  Here, we provide the detailed analysis of the proposed components within our category-unified models, including unified representation network AdaFormer (Section A.1), model inputs (Section A.2) and learning objective (Section A.3). All experiments and analysis are conducted using Siamese paradigm (p2b) on the KITTI (kitti) dataset.    A.1 Unified Representation Network: AdaFormer  The proposed unified representation network AdaFormer incorporates a group regression module for learning deformable groups to enable adaptive receptive fields for various object categories, along with a vector-attention mechanism to facilitate feature interaction of points within these deformable groups, ultimately forming a category-unified feature representation.   Tab. 1 presents an ablation study to understand the two sub-components. Benefiting from the adaptive receptive fields achieved by the group regression module, our representation network can learn geometric information of various object categories in a unified manner. Consequently, when this module is removed, average performance drops by 6.9% and 7.0% in terms of Success and Precision, respectively. It\u2019s noteworthy that the most obvious performance degradation occurs in the Pedestrian category. This is due to the relatively small training samples for the Pedestrian category and the significant differences in shape and size compared to other object categories. To visually understand of how the group regression module works, we provide some visualizations of deformable groups on the Car and Pedestrian categories, as shown in Fig. 1. In addition, when removing the vector-attention mechanism, we employ a feature propagation operator in existing backbone network (pointnet; pointnet++) to substitute it. Tab. 1 demonstrates that the vector-attention mechanism plays a crucial role in promoting the learning of a unified feature representation.   Table 1: Ablation study of unified representation network. Success / Precision are used for evaluation. Bold denote the best performance.    Group Regression Vector-Attention Car Pedestrian Van Cyclist Mean   Module Mechanism [6,424] [6,088] [1,248] [308] [14,068]   \u2717 \u2717 56.3 / 72.4 33.2 / 60.4 57.0 / 68.6 32.2 / 43.5 45.9 / 63.2   \u2717 \u2713 56.5 / 72.7 35.2 / 62.9 59.4 / 69.3 32.3 / 44.0 47.1 / 67.6   \u2713 \u2717 57.7 / 73.8 44.9 / 71.2 61.8 / 72.0 35.3 / 46.1 52.1 / 72.0   \u2713 \u2713  58.1 / 73.9   48.2 / 76.2   63.1 / 74.9   36.7 / 47.4   54.0 / 74.6              Figure 1: Comparison of groups on the Car and Pedestrian categories. We plot four fixed groups and deformable groups from the first layer in PointNet++ (pointnet++) and our AdaFormer network, respectively, using different colors.     A.2 Unified Model Input  The scale factor \u03b1\ud835\udefc\\alphaitalic_\u03b1 is an important hyper-parameter in our unified model inputs. Hence, we conduct an ablation experiment using different values to determine the optimal setting for this parameter. As presented in Tab. 2, our method is not sensitive to the scale factor within a reasonable range of values, i.e., when this parameter is set in the range from 0.8 to 1.4. Nevertheless, excessively large value will introduce noise, whereas overly small value will ignore valuable information, both leading to significant performance degradation.   Table 2: Performance using different scale factor \u03b1\ud835\udefc\\alphaitalic_\u03b1. Success / Precision are used for evaluation. Bold denote the best performance.    Scale Factor \u03b1\ud835\udefc\\alphaitalic_\u03b1 Car [6,424] Pedestrian [6,088] Van [1,248] Cyclist [308] Mean [14,068]   0.6 52.4 / 68.1 42.3 / 70.3 48.9 / 57.4 31.2 / 43.0 47.3 / 67.6   0.8 57.5 / 73.2  48.4 / 76.7  63.0 / 74.7  37.2 / 45.0 53.7 / 74.3   1.0 58.1 / 73.9  48.2 / 76.2  63.1 / 74.9  36.7 / 47.4   54.0 / 74.6    1.2  58.3 / 73.7 48.0 / 76.1 62.8 / 74.7 36.1 / 46.6 53.8 / 74.2   1.4 57.8 / 73.1 47.6",
            "references": [
                {
                    "title": "Synchronize Feature Extracting and Matching: A Single Branch Framework for 3D Object Tracking",
                    "abstract": "Siamese network has been a de facto benchmark framework for 3D LiDAR object tracking with a shared-parametric encoder extracting features from template and search region, respectively. This paradigm relies heavily on an additional matching network to model the cross-correlation/similarity of the template and search region. In this paper, we forsake the conventional Siamese paradigm and propose a novel single-branch framework, SyncTrack, synchronizing the feature extracting and matching to avoid forwarding encoder twice for template and search region as well as introducing extra parameters of matching network. The synchronization mechanism is based on the dynamic affinity of the Transformer, and an in-depth analysis of the relevance is provided theoretically. Moreover, based on the synchronization, we introduce a novel Attentive PointsSampling strategy into the Transformer layers (APST), replacing the random/Farthest Points Sampling (FPS) method with sampling under the supervision of attentive relations between the template and search region. It implies connecting point-wise sampling with the feature learning, beneficial to aggregating more distinctive and geometric features for tracking with sparse points. Extensive experiments on two benchmark datasets (KITTI and NuScenes) show that SyncTrack achieves state-of-the-art performance in realtime tracking."
                },
                {
                    "title": "3DMOTFormer: Graph Transformer for Online 3D Multi-Object Tracking",
                    "abstract": "Tracking 3D objects accurately and consistently is crucial for autonomous vehicles, enabling more reliable downstream tasks such as trajectory prediction and motion planning. Based on the substantial progress in object detection in recent years, the tracking-by-detection paradigm has become a popular choice due to its simplicity and efficiency. State-of-the-art 3D multi-object tracking (MOT) approaches typically rely on non-learned model-based algorithms such as Kalman Filter but require many manually tuned parameters. On the other hand, learning-based approaches face the problem of adapting the training to the online setting, leading to inevitable distribution mismatch between training and inference as well as suboptimal performance. In this work, we propose 3DMOTFormer, a learned geometry-based 3D MOT framework building upon the transformer architecture. We use an Edge-Augmented Graph Transformer to reason on the track-detection bipartite graph frame-by-frame and conduct data association via edge classification. To reduce the distribution mismatch between training and inference, we propose a novel online training strategy with an autoregressive and recurrent forward pass as well as sequential batch optimization. Using CenterPoint detections, our approach achieves 71.2% and 68.2% AMOTA on the nuScenes validation and test split, respectively. In addition, a trained 3DMOTFormer model generalizes well across different object detectors. Code is available at: https://github.com/dsx0511/3DMOTFormer."
                },
                {
                    "title": "VideoTrack: Learning to Track Objects via Video Transformer",
                    "abstract": "Existing Siamese tracking methods, which are built on pair-wise matching between two single frames, heavily rely on additional sophisticated mechanism to exploit temporal information among successive video frames, hindering them from efficiency and industrial deployments. In this work, we resort to sequence-level target matching that can encode temporal contexts into the spatial features through a neat feedforward video model. Specifically, we adapt the standard video transformer architecture to visual tracking by enabling spatiotemporal feature learning directly from frame-level patch sequences. To better adapt to the tracking task, we carefully blend the spatiotemporal information in the video clips through sequential multi-branch triplet blocks, which formulates a video transformer backbone. Our experimental study compares different model variants, such as tokenization strategies, hierarchical structures, and video attention schemes. Then, we propose a disentangled dual-template mechanism that decouples static and dynamic appearance clues over time, and reduces temporal redundancy in video frames. Extensive experiments show that our method, named as Video Track, achieves state-of-the-art results while running in real-time."
                },
                {
                    "title": "OSP2B: One-Stage Point-to-Box Network for 3D Siamese Tracking",
                    "abstract": "Two-stage point-to-box network acts as a critical role in the recent popular 3D Siamese tracking paradigm, which first generates proposals and then predicts corresponding proposal-wise scores. However, such a network suffers from tedious hyper-parameter tuning and task misalignment, limiting the tracking performance. Towards these concerns, we propose a simple yet effective one-stage point-to-box network for point cloud-based 3D single object tracking. It synchronizes 3D proposal generation and center-ness score prediction by a parallel predictor without tedious hyper-parameters. To guide a task-aligned score ranking of proposals, a center-aware focal loss is proposed to supervise the training of the center-ness branch, which enhances the network's discriminative ability to distinguish proposals of different quality. Besides, we design a binary target classifier to identify target-relevant points. By integrating the derived classification scores with the center-ness scores, the resulting network can effectively suppress interference proposals and further mitigate task misalignment. Finally, we present a novel one-stage Siamese tracker OSP2B equipped with the designed network. Extensive experiments on challenging benchmarks including KITTI and Waymo SOT Dataset show that our OSP2B achieves leading performance with a considerable real-time speed."
                },
                {
                    "title": "GLT-T++: Global-Local Transformer for 3D Siamese Tracking with Ranking Loss",
                    "abstract": "Siamese trackers based on 3D region proposal network (RPN) have shown remarkable success with deep Hough voting. However, using a single seed point feature as the cue for voting fails to produce high-quality 3D proposals. Additionally, the equal treatment of seed points in the voting process, regardless of their significance, exacerbates this limitation. To address these challenges, we propose a novel transformer-based voting scheme to generate better proposals. Specifically, a global-local transformer (GLT) module is devised to integrate object- and patch-aware geometric priors into seed point features, resulting in robust and accurate cues for offset learning of seed points. To train the GLT module, we introduce an importance prediction branch that learns the potential importance weights of seed points as a training constraint. Incorporating this transformer-based voting scheme into 3D RPN, a novel Siamese method dubbed GLT-T is developed for 3D single object tracking on point clouds. Moreover, we identify that the highest-scored proposal in the Siamese paradigm may not be the most accurate proposal, which limits tracking performance. Towards this concern, we approach the binary score prediction task as a ranking problem, and design a target-aware ranking loss and a localization-aware ranking loss to produce accurate ranking of proposals. With the ranking losses, we further present GLT-T++, an enhanced version of GLT-T. Extensive experiments on multiple benchmarks demonstrate that our GLT-T and GLT-T++ outperform state-of-the-art methods in terms of tracking accuracy while maintaining a real-time inference speed. The source code will be made available at https://github.com/haooozi/GLT-T."
                },
                {
                    "title": "GLT-T: Global-Local Transformer Voting for 3D Single Object Tracking in Point Clouds",
                    "abstract": "Current 3D single object tracking methods are typically based on VoteNet, a 3D region proposal network. Despite the success, using a single seed point feature as the cue for offset learning in VoteNet prevents high-quality 3D proposals from being generated. Moreover, seed points with different importance are treated equally in the voting process, aggravating this defect. To address these issues, we propose a novel global-local transformer voting scheme to provide more informative cues and guide the model pay more attention on potential seed points, promoting the generation of high-quality 3D proposals. Technically, a global-local transformer (GLT) module is employed to integrate object- and patch-aware prior into seed point features to effectively form strong feature representation for geometric positions of the seed points, thus providing more robust and accurate cues for offset learning. Subsequently, a simple yet effective training strategy is designed to train the GLT module. We develop an importance prediction branch to learn the potential importance of the seed points and treat the output weights vector as a training constraint term. By incorporating the above components together, we exhibit a superior tracking method GLT-T. Extensive experiments on challenging KITTI and NuScenes benchmarks demonstrate that GLT-T achieves state-of-the-art performance in the 3D single object tracking task. Besides, further ablation studies show the advantages of the proposed global-local transformer voting scheme over the original VoteNet. Code and models will be available at https://github.com/haooozi/GLT-T."
                },
                {
                    "title": "CXTrack: Improving 3D Point Cloud Tracking with Contextual Information",
                    "abstract": "3D single object tracking plays an essential role in many applications, such as autonomous driving. It remains a challenging problem due to the large appearance variation and the sparsity of points caused by occlusion and lim-ited sensor capabilities. Therefore, contextual information across two consecutive frames is crucial for effective object tracking. However, points containing such useful information are often overlooked and cropped out in existing methods, leading to insufficient use of important contextual knowledge. To address this issue, we propose CXTrack, a novel transformer-based network for 3D object tracking, which exploits ConteXtual information to improve the tracking results. Specifically, we design a target-centric transformer network that directly takes point features from two consecutive frames and the previous bounding box as input to explore contextual information and implicitly propagate target cues. To achieve accurate localization for objects of all sizes, we propose a transformer-based localization head with a novel center embedding module to distinguish the target from distractors. Extensive experiments on three large-scale datasets, KITTI, nuScenes and Waymo Open Dataset, show that CXTrack achieves state-of-the-art tracking performance while running at 34 FPS."
                },
                {
                    "title": "MSA-MOT: Multi-Stage Association for 3D Multimodality Multi-Object Tracking",
                    "abstract": "Three-dimensional multimodality multi-object tracking has attracted great attention due to the use of complementary information. However, such a framework generally adopts a one-stage association approach, which fails to perform precise matching between detections and tracklets, and, thus, cannot robustly track objects in complex scenes. To address this matching problem caused by one-stage association, we propose a novel multi-stage association method, which consists of a hierarchical matching module and a customized track management module. Specifically, the hierarchical matching module defines the reliability of the objects by associating multimodal detections, and matches detections with trajectories based on the reliability in turn, which increases the utilization of true detections, and, thus, guides accurate association. Then, based on the reliability of the trajectories provided by the matching module, the customized track management module sets maximum missing frames with differences for tracks, which decreases the number of identity switches of the same object and, thus, further improves the association accuracy. By using the proposed multi-stage association method, we develop a tracker called MSA-MOT for the 3D multi-object tracking task, alleviating the inherent matching problem in one-stage association. Extensive experiments are conducted on the challenging KITTI benchmark, and the results show that our tracker outperforms the previous state-of-the-art methods in terms of both accuracy and speed. Moreover, the ablation and exploration analysis results demonstrate the effectiveness of the proposed multi-stage association method."
                },
                {
                    "title": "CAGroup3D: Class-Aware Grouping for 3D Object Detection on Point Clouds",
                    "abstract": "We present a novel two-stage fully sparse convolutional 3D object detection framework, named CAGroup3D. Our proposed method first generates some high-quality 3D proposals by leveraging the class-aware local group strategy on the object surface voxels with the same semantic predictions, which considers semantic consistency and diverse locality abandoned in previous bottom-up approaches. Then, to recover the features of missed voxels due to incorrect voxel-wise segmentation, we build a fully sparse convolutional RoI pooling module to directly aggregate fine-grained spatial information from backbone for further proposal refinement. It is memory-and-computation efficient and can better encode the geometry-specific features of each 3D proposal. Our model achieves state-of-the-art 3D detection performance with remarkable gains of +\\textit{3.6\\%} on ScanNet V2 and +\\textit{2.6}\\% on SUN RGB-D in term of mAP@0.25. Code will be available at https://github.com/Haiyang-W/CAGroup3D."
                },
                {
                    "title": "Implicit and Efficient Point Cloud Completion for 3D Single Object Tracking",
                    "abstract": "The point cloud based 3D single object tracking has drawn increasing attention. Although many breakthroughs have been achieved, we also reveal two severe issues. By extensive analysis, we find the prediction manner of current approaches is non-robust, i.e., exposing a misalignment gap between prediction score and actually localization accuracy. Another issue is the sparse point returns will damage the feature matching procedure of the SOT task. Based on these insights, we introduce two novel modules, i.e., Adaptive Refine Prediction (ARP) and Target Knowledge Transfer (TKT), to tackle them, respectively. To this end, we first design a strong pipeline to extract discriminative features and conduct the matching with the attention mechanism. Then, ARP module is proposed to tackle the misalignment issue by aggregating all predicted candidates with valuable clues. Finally, TKT module is designed to effectively overcome incomplete point cloud due to sparse and occlusion issues. We call our overall framework PCET. By conducting extensive experiments on the KITTI and Waymo Open Dataset, our model achieves state-of-the-art performance while maintaining a lower computational cost."
                },
                {
                    "title": "Spreading Fine-Grained Prior Knowledge for Accurate Tracking",
                    "abstract": "With the widespread use of deep learning in single object tracking task, mainstream tracking algorithms treat tracking as a combined classification and regression problem. Classification aims at locating an arbitrary target, and regression aims at estimating the corresponding bounding box. In this paper, we focus on regression and propose a novel box estimation network, which consists of a transformer encoder target pyramid guide (TPG) and transformer decoder target pyramid spread (TPS). Specifically, the transformer encoder TPG is designed to generate fine-grained prior knowledge with explicit representation for template targets. In contrast to the raw transformer encoder, we capture the visual dependence through local-global self-attention and deem the multi-scale target regions as the \u201clocal\u201d region. Using this fine-grained prior knowledge, we design the transformer decoder TPS to spread it to the subsequent search regions with high affinity to accurately estimate the bounding boxes. Considering that self-attention fails to model information interaction across channels between the template target and search regions, we develop a channel-wise cross-attention block within the TPS as compensation. Extensive experiments on the OTB100, UAV123, NFS, VOT2020, VOT2021, LaSOT, LaSOT_ext, TrackingNet and GOT-10k benchmarks show that the proposed box estimation network outperforms most existing box estimation methods. Furthermore, our trackers based on this estimation network exhibit a competitive performance against state-of-the-art trackers."
                },
                {
                    "title": "One-Shot Multiple Object Tracking in UAV Videos Using Task-Specific Fine-Grained Features",
                    "abstract": "Multiple object tracking (MOT) in unmanned aerial vehicle (UAV) videos is a fundamental task and can be applied in many fields. MOT consists of two critical procedures, i.e., object detection and re-identification (ReID). One-shot MOT, which incorporates detection and ReID in a unified network, has gained attention due to its fast inference speed. It significantly reduces the computational overhead by making two subtasks share features. However, most existing one-shot trackers struggle to achieve robust tracking in UAV videos. We observe that the essential difference between detection and ReID leads to an optimization contradiction within one-shot networks. To alleviate this contradiction, we propose a novel feature decoupling network (FDN) to convert shared features into detection-specific and ReID-specific representations. The FDN searches for characteristics and commonalities between the two tasks to synergize detection and ReID. In addition, existing one-shot trackers struggle to locate small targets in UAV videos. Therefore, we design a pyramid transformer encoder (PTE) to enrich the semantic information of the resulting detection-specific representations. By learning scale-aware fine-grained features, the PTE empowers our tracker to locate targets in UAV videos accurately. Extensive experiments on VisDrone2021 and UAVDT benchmarks demonstrate that our tracker achieves state-of-the-art tracking performance."
                },
                {
                    "title": "3D Siamese Transformer Network for Single Object Tracking on Point Clouds",
                    "abstract": "Siamese network based trackers formulate 3D single object tracking as cross-correlation learning between point features of a template and a search area. Due to the large appearance variation between the template and search area during tracking, how to learn the robust cross correlation between them for identifying the potential target in the search area is still a challenging problem. In this paper, we explicitly use Transformer to form a 3D Siamese Transformer network for learning robust cross correlation between the template and the search area of point clouds. Specifically, we develop a Siamese point Transformer network to learn shape context information of the target. Its encoder uses self-attention to capture non-local information of point clouds to characterize the shape information of the object, and the decoder utilizes cross-attention to upsample discriminative point features. After that, we develop an iterative coarse-to-fine correlation network to learn the robust cross correlation between the template and the search area. It formulates the cross-feature augmentation to associate the template with the potential target in the search area via cross attention. To further enhance the potential target, it employs the ego-feature augmentation that applies self-attention to the local k-NN graph of the feature space to aggregate target features. Experiments on the KITTI, nuScenes, and Waymo datasets show that our method achieves state-of-the-art performance on the 3D single object tracking task."
                },
                {
                    "title": "DBQ-SSD: Dynamic Ball Query for Efficient 3D Object Detection",
                    "abstract": "Many point-based 3D detectors adopt point-feature sampling strategies to drop some points for efficient inference. These strategies are typically based on fixed and handcrafted rules, making it difficult to handle complicated scenes. Different from them, we propose a Dynamic Ball Query (DBQ) network to adaptively select a subset of input points according to the input features, and assign the feature transform with a suitable receptive field for each selected point. It can be embedded into some state-of-the-art 3D detectors and trained in an end-to-end manner, which significantly reduces the computational cost. Extensive experiments demonstrate that our method can increase the inference speed by 30%-100% on KITTI, Waymo, and ONCE datasets. Specifically, the inference speed of our detector can reach 162 FPS on KITTI scene, and 30 FPS on Waymo and ONCE scenes without performance degradation. Due to skipping the redundant points, some evaluation metrics show significant improvements. Codes will be released at https://github.com/yancie-yjr/DBQ-SSD."
                },
                {
                    "title": "Learning Localization-Aware Target Confidence for Siamese Visual Tracking",
                    "abstract": "Siamese tracking paradigm has achieved great success, providing effective appearance discrimination and size estimation by classification and regression. While such a paradigm typically optimizes the classification and regression independently, leading to task misalignment (accurate prediction boxes have no high target confidence scores). In this paper, to alleviate this misalignment, we propose a novel tracking paradigm, called SiamLA. Within this paradigm, a series of simple, yet effective localization-aware components are introduced to generate localization-aware target confidence scores. Specifically, with the proposed localization-aware dynamic label (LADL) loss and localization-aware label smoothing (LALS) strategy, collaborative optimization between the classification and regression is achieved, enabling classification scores to be aware of location state, not just appearance similarity. Besides, we propose a separate localization-aware quality prediction (LAQP) branch to produce location quality scores to further modify the classification scores. To guide a more reliable modification, a novel localization-aware feature aggregation (LAFA) module is designed and embedded into this branch. Consequently, the resulting target confidence scores are more discriminative for the location state, allowing accurate prediction boxes tend to be predicted as high scores. Extensive experiments are conducted on six challenging benchmarks, including GOT-10 k, TrackingNet, LaSOT, TNL2K, OTB100 and VOT2018. Our SiamLA achieves competitive performance in terms of both accuracy and efficiency. Furthermore, a stability analysis reveals that our tracking paradigm is relatively stable, implying that the paradigm is potential for real-world applications."
                },
                {
                    "title": "RBGNet: Ray-based Grouping for 3D Object Detection",
                    "abstract": "As a fundamental problem in computer vision, 3D object detection is experiencing rapid growth. To extract the point-wise features from the irregularly and sparsely distributed points, previous methods usually take a feature grouping module to aggregate the point features to an object candidate. However, these methods have not yet leveraged the surface geometry of foreground objects to enhance grouping and 3D box generation. In this paper, we propose the RBGNet framework, a voting-based 3D detector for accurate 3D object detection from point clouds. In order to learn better representations of object shape to enhance cluster features for predicting 3D boxes, we propose a ray-based feature grouping module, which aggregates the point-wise features on object surfaces using a group of determined rays uniformly emitted from cluster centers. Considering the fact that foreground points are more meaningful for box estimation, we design a novel foreground biased sampling strategy in downsample process to sample more points on object surfaces and further boost the detection performance. Our model achieves state-of-the-art 3D detection performance on ScanNet V2 and SUN RGB-D with remarkable performance gains. Code will be available at https://github.com/Haiyang-W/RBGNet."
                },
                {
                    "title": "MixFormer: End-to-End Tracking with Iterative Mixed Attention",
                    "abstract": "Tracking often uses a multistage pipeline of feature extraction, target information integration, and bounding box estimation. To simplify this pipeline and unify the process of feature extraction and target information integration, we present a compact tracking framework, termed as MixFormer, built upon transformers. Our core design is to utilize the flexibility of attention operations, and propose a Mixed Attention Module (MAM) for simultaneous feature extraction and target information integration. This synchronous modeling scheme allows to extract target-specific discriminative features and perform extensive communication between target and search area. Based on MAM, we build our MixFormer tracking framework simply by stacking multiple MAMs with progressive patch embedding and placing a localization head on top. In addition, to handle multiple target templates during online tracking, we devise an asymmetric attention scheme in MAM to reduce computational cost, and propose an effective score prediction module to select high-quality templates. Our MixFormer sets a new state-of-the-art performance on five tracking benchmarks, including LaSOT, TrackingNet, VOT2020, GOT-10k, and UAV123. In particular, our MixFormer-L achieves NP score of 79.9% on LaSOT, 88.9% on TrackingNet and EAO of 0.555 on VOT2020. We also perform in-depth ablation studies to demonstrate the effectiveness of simultaneous feature extraction and information integration. Code and trained models are publicly available at https://github.com/MCG-NJU/MixFormer."
                },
                {
                    "title": "A Lightweight and Detector-Free 3D Single Object Tracker on Point Clouds",
                    "abstract": "Recent works on 3D single object tracking treat the task as a target-specific 3D detection task, where an off-the-shelf 3D detector is commonly employed for the tracking. However, it is non-trivial to perform accurate target-specific detection since the point cloud of objects in raw LiDAR scans is usually sparse and incomplete. In this paper, we address this issue by explicitly leveraging temporal motion cues and propose DMT, a Detector-free Motion-prediction-based 3D Tracking network that completely removes the usage of complicated 3D detectors and is lighter, faster, and more accurate than previous trackers. Specifically, the motion prediction module is first introduced to estimate a potential target center of the current frame in a point-cloud-free manner. Then, an explicit voting module is proposed to directly regress the 3D box from the estimated target center. Extensive experiments on KITTI and NuScenes datasets demonstrate that our DMT can still achieve better performance ( $\\sim $ 10% improvement over the NuScenes dataset) and a faster tracking speed (i.e., 72 FPS) than state-of-the-art approaches without applying any complicated 3D detectors. Our code is released at https://github.com/jimmy-dq/DMT."
                },
                {
                    "title": "Beyond 3D Siamese Tracking: A Motion-Centric Paradigm for 3D Single Object Tracking in Point Clouds",
                    "abstract": "3D single object tracking (3D SOT) in LiDAR point clouds plays a crucial role in autonomous driving. Current approaches all follow the Siamese paradigm based on appearance matching. However, LiDAR point clouds are usually textureless and incomplete, which hinders effective appearance matching. Besides, previous methods greatly overlook the critical motion clues among targets. In this work, beyond 3D Siamese tracking, we introduce a motion-centric paradigm to handle 3D SOT from a new perspective. Following this paradigm, we propose a matching-free two-stage tracker M2-Track. At the 1st-stage, M2-Track localizes the target within successive frames via motion transformation. Then it refines the target box through motion-assisted shape completion at the 2nd-stage. Extensive experiments confirm that M2-Track significantly outperforms previous state-of-the-arts on three large-scale datasets while running at 57FPS (~ 8%, ~ 17% and ~ 22% precision gains on KITTI, NuScenes, and Waymo Open Dataset respectively). Further analysis verifies each component's effectiveness and shows the motioncentric paradigm's promising potential when combined with appearance matching. Code will be made available at https://github.com/Ghostish/Open3DSOT."
                },
                {
                    "title": "Correlation-Aware Deep Tracking",
                    "abstract": "Robustness and discrimination power are two fundamental requirements in visual object tracking. In most tracking paradigms, we find that the features extracted by the popular Siamese-like networks cannot fully discriminatively model the tracked targets and distractor objects, hindering them from simultaneously meeting these two requirements. While most methods focus on designing robust correlation operations, we propose a novel target-dependent feature network inspired by the self-/cross-attention scheme. In contrast to the Siamese-like feature extraction, our network deeply embeds cross-image feature correlation in multiple layers of the feature network. By extensively matching the features of the two images through multiple layers, it is able to suppress non-target features, resulting in instance-varying feature extraction. The output features of the search image can be directly used for predicting target locations without extra correlation step. Moreover, our model can be flexibly pre-trained on abundant unpaired images, leading to notably faster convergence than the existing methods. Extensive experiments show our method achieves the state-of-the-art results while running at real-time. Our feature networks also can be applied to existing tracking pipelines seamlessly to raise the tracking performance."
                },
                {
                    "title": "Omnivore: A Single Model for Many Visual Modalities",
                    "abstract": "Prior work has studied different visual modalities in isolation and developed separate architectures for recognition of images, videos, and 3D data. Instead, in this paper, we propose a single model which excels at classifying images, videos, and single-view 3D data using exactly the same model parameters. Our \u2018OMNIVORE\u2019 model leverages the flexibility of transformer-based architectures and is trained jointly on classification tasks from different modalities. Omnivoreis simple to train, uses off-the-shelf standard datasets, and performs at-par or better than modality-specific models of the same size. A single Omnivoremodel obtains 86.0% on ImageNet, 84.1% on Kinetics, and 67.1% on SUN RGB-D. After finetuning, our models outperform prior work on a variety of vision tasks and generalize across modalities. OMNIVORE's shared visual representation naturally enables cross-modal recognition without access to correspondences between modalities. We hope our results motivate researchers to model visual modalities together."
                },
                {
                    "title": "Visual Object Tracking With Discriminative Filters and Siamese Networks: A Survey and Outlook",
                    "abstract": "Accurate and robust visual object tracking is one of the most challenging and fundamental computer vision problems. It entails estimating the trajectory of the target in an image sequence, given only its initial location, and segmentation, or its rough approximation in the form of a bounding box. Discriminative Correlation Filters (DCFs) and deep Siamese Networks (SNs) have emerged as dominating tracking paradigms, which have led to significant progress. Following the rapid evolution of visual object tracking in the last decade, this survey presents a systematic and thorough review of more than 90 DCFs and Siamese trackers, based on results in nine tracking benchmarks. First, we present the background theory of both the DCF and Siamese tracking core formulations. Then, we distinguish and comprehensively review the shared as well as specific open research challenges in both these tracking paradigms. Furthermore, we thoroughly analyze the performance of DCF and Siamese trackers on nine benchmarks, covering different experimental aspects of visual tracking: datasets, evaluation metrics, performance, and speed comparisons. We finish the survey by presenting recommendations and suggestions for distinguished open challenges based on our analysis."
                },
                {
                    "title": "PTTR: Relational 3D Point Cloud Object Tracking with Transformer",
                    "abstract": "In a point cloud sequence, 3D object tracking aims to predict the location and orientation of an object in the current search point cloud given a template point cloud. Motivated by the success of transformers, we propose Point Tracking TRansformer (PTTR), which efficiently predicts high-quality 3D tracking results in a coarse-to-fine manner with the help of transformer operations. PTTR consists of three novel designs. 1) Instead of random sampling, we design Relation-Aware Sampling to preserve relevant points to given templates during subsampling. 2) Furthermore, we propose a Point Relation Transformer (PRT) consisting of a self-attention and a cross-attention module. The global self-attention operation captures long-range dependencies to enhance encoded point features for the search area and the template, respectively. Subsequently, we generate the coarse tracking results by matching the two sets of point features via cross-attention. 3) Based on the coarse tracking results, we employ a novel Prediction Refinement Module to obtain the final refined prediction. In addition, we create a large-scale point cloud single object tracking benchmark based on the Waymo Open Dataset. Extensive experiments show that PTTR achieves superior point cloud tracking in both accuracy and efficiency. Our code is available at https://github.com/Jasonkks/PTTR."
                },
                {
                    "title": "3D Siamese Voxel-to-BEV Tracker for Sparse Point Clouds",
                    "abstract": "3D object tracking in point clouds is still a challenging problem due to the sparsity of LiDAR points in dynamic environments. In this work, we propose a Siamese voxel-to-BEV tracker, which can significantly improve the tracking performance in sparse 3D point clouds. Specifically, it consists of a Siamese shape-aware feature learning network and a voxel-to-BEV target localization network. The Siamese shape-aware feature learning network can capture 3D shape information of the object to learn the discriminative features of the object so that the potential target from the background in sparse point clouds can be identified. To this end, we first perform template feature embedding to embed the template's feature into the potential target and then generate a dense 3D shape to characterize the shape information of the potential target. For localizing the tracked target, the voxel-to-BEV target localization network regresses the target's 2D center and the $z$-axis center from the dense bird's eye view (BEV) feature map in an anchor-free manner. Concretely, we compress the voxelized point cloud along $z$-axis through max pooling to obtain a dense BEV feature map, where the regression of the 2D center and the $z$-axis center can be performed more effectively. Extensive evaluation on the KITTI and nuScenes datasets shows that our method significantly outperforms the current state-of-the-art methods by a large margin."
                },
                {
                    "title": "MLVSNet: Multi-level Voting Siamese Network for 3D Visual Tracking",
                    "abstract": "Benefiting from the excellent performance of Siamese-based trackers, huge progress on 2D visual tracking has been achieved. However, 3D visual tracking is still under-explored. Inspired by the idea of Hough voting in 3D object detection, in this paper, we propose a Multi-level Voting Siamese Network (MLVSNet) for 3D visual tracking from outdoor point cloud sequences. To deal with sparsity in outdoor 3D point clouds, we propose to perform Hough voting on multi-level features to get more vote centers and retain more useful information, instead of voting only on the fi-nal level feature as in previous methods. We also design an efficient and lightweight Target-Guided Attention (TGA) module to transfer the target information and highlight the target points in the search area. Moreover, we propose a Vote-cluster Feature Enhancement (VFE) module to exploit the relationships between different vote clusters. Extensive experiments on the 3D tracking benchmark of KITTI dataset demonstrate that our MLVSNet outperforms state-of-the-art methods with significant margins. Code will be available at https://github.com/CodeWZT/MLVSNet."
                },
                {
                    "title": "Pyramid R-CNN: Towards Better Performance and Adaptability for 3D Object Detection",
                    "abstract": "We present a flexible and high-performance framework, named Pyramid R-CNN, for two-stage 3D object detection from point clouds. Current approaches generally rely on the points or voxels of interest for RoI feature extraction on the second stage, but cannot effectively handle the sparsity and non-uniform distribution of those points, and this may result in failures in detecting objects that are far away. To resolve the problems, we propose a novel second-stage module, named pyramid RoI head, to adaptively learn the features from the sparse points of interest. The pyramid RoI head consists of three key components. Firstly, we propose the RoI-grid Pyramid, which mitigates the sparsity problem by extensively collecting points of interest for each RoI in a pyramid manner. Secondly, we propose RoI-grid Attention, a new operation that can encode richer information from sparse points by incorporating conventional attention-based and graph-based point operators into a unified formulation. Thirdly, we propose the Density-Aware Radius Prediction (DARP) module, which can adapt to different point density levels by dynamically adjusting the focusing range of RoIs. Combining the three components, our pyramid RoI head is robust to the sparse and imbalanced circumstances, and can be applied upon various 3D backbones to consistently boost the detection performance. Extensive experiments show that Pyramid R-CNN outperforms the state-of-the-art 3D detection models by a large margin on both the KITTI dataset and the Waymo Open dataset."
                },
                {
                    "title": "Multi-Task Self-Training for Learning General Representations",
                    "abstract": "Despite the fast progress in training specialized models for various tasks, learning a single general model that works well for many tasks is still challenging for computer vision. Here we introduce multi-task self-training (MuST), which harnesses the knowledge in independent specialized teacher models (e.g., ImageNet model on classification) to train a single general student model. Our approach has three steps. First, we train specialized teachers independently on labeled datasets. We then use the specialized teachers to label an unlabeled dataset to create a multi-task pseudo labeled dataset. Finally, the dataset, which now contains pseudo labels from teacher models trained on different datasets/tasks, is then used to train a student model with multi-task learning. We evaluate the feature representations of the student model on 6 vision tasks including image recognition (classification, detection, segmentation) and 3D geometry estimation (depth and surface normal estimation). MuST is scalable with unlabeled or partially labeled datasets and outperforms both specialized supervised models and self-supervised models when training on large scale datasets. Lastly, we show MuST can improve upon already strong checkpoints [23] trained with billions of examples. The results suggest self-training is a promising direction to aggregate labeled and unlabeled training data for learning general feature representations."
                },
                {
                    "title": "PTT: Point-Track-Transformer Module for 3D Single Object Tracking in Point Clouds",
                    "abstract": "3D single object tracking is a key issue for robotics. In this paper, we propose a transformer module called Point-Track-Transformer (PTT) for point cloud-based 3D single object tracking. PTT module contains three blocks for feature embedding, position encoding, and self-attention feature computation. Feature embedding aims to place features closer in the embedding space if they have similar semantic information. Position encoding is used to encode coordinates of point clouds into high dimension distinguishable features. Self-attention generates refined attention features by computing attention weights. Besides, we embed the PTT module into the open-source state-of-the-art method P2B to construct PTT-Net. Experiments on the KITTI dataset reveal that our PTT-Net surpasses the state-of-the-art by a noticeable margin $\\left( {\\sim 10\\% } \\right)$. Additionally, PTT-Net could achieve real-time performance (~40FPS) on NVIDIA 1080Ti GPU. Our code is open-sourced for the robotics community at https://github.com/shanjiayao/PTT."
                },
                {
                    "title": "Box-Aware Feature Enhancement for Single Object Tracking on Point Clouds",
                    "abstract": "Current 3D single object tracking approaches track the target based on a feature comparison between the target template and the search area. However, due to the common occlusion in LiDAR scans, it is non-trivial to conduct accurate feature comparisons on severe sparse and incomplete shapes. In this work, we exploit the ground truth bounding box given in the first frame as a strong cue to enhance the feature description of the target object, enabling a more accurate feature comparison in a simple yet effective way. In particular, we first propose the BoxCloud, an informative and robust representation, to depict an object using the point-to-box relation. We further design an efficient box-aware feature fusion module, which leverages the aforementioned BoxCloud for reliable feature matching and embedding. Integrating the proposed general components into an existing model P2B [27], we construct a superior box-aware tracker (BAT)1. Experiments confirm that our proposed BAT outperforms the previous state-of-the-art by a large margin on both KITTI and NuScenes benchmarks, achieving a 12.8% improvement in terms of precision while running \u223c20% faster."
                },
                {
                    "title": "LCCNet: LiDAR and Camera Self-Calibration using Cost Volume Network",
                    "abstract": "Multi-sensor fusion is for enhancing environment perception and 3D reconstruction in self-driving and robot navigation. Calibration between sensors is the precondition of effective multi-sensor fusion. Laborious manual works and complex environment settings exist in old-fashioned calibration techniques for Light Detection and Ranging (Li-DAR) and camera. We propose an online LiDAR-Camera Self-calibration Network (LCCNet), different from the previous CNN-based methods. LCCNet can be trained end-to-end and predict the extrinsic parameters in real-time. In the LCCNet, we exploit the cost volume layer to express the correlation between the RGB image features and the depth image projected from point clouds. Besides using the smooth L1-Loss of the predicted extrinsic calibration parameters as a supervised signal, an additional self-supervised signal, point cloud distance loss, is applied during training. Instead of directly regressing the extrinsic parameters, we predict the decalibrated deviation from initial calibration to the ground truth. The calibration error decreases further with iterative refinement and the temporal filtering approach in the inference stage. The execution time of the calibration process is 24ms for each iteration on a single GPU. LCCNet achieves a mean absolute calibration error of 0.297cm in translation and 0.017\u00b0 in rotation with miscalibration magnitudes of up to \u00b11.5m and \u00b120\u00b0 on the KITTI-odometry dataset, which is better than the state-of-the-art CNN-based calibration methods. The code will be publicly available at https://github.com/LvXudong-HIT/LCCNet"
                },
                {
                    "title": "Transformer Tracking",
                    "abstract": "Correlation acts as a critical role in the tracking field, especially in recent popular Siamese-based trackers. The correlation operation is a simple fusion manner to consider the similarity between the template and the search region. However, the correlation operation itself is a local linear matching process, leading to lose semantic information and fall into local optimum easily, which may be the bottleneck of designing high-accuracy tracking algorithms. Is there any better feature fusion method than correlation? To address this issue, inspired by Transformer, this work presents a novel attention-based feature fusion network, which effectively combines the template and search region features solely using attention. Specifically, the proposed method includes an ego-context augment module based on self-attention and a cross-feature augment module based on cross-attention. Finally, we present a Transformer tracking (named TransT) method based on the Siamese-like feature extraction backbone, the designed attention-based fusion mechanism, and the classification and regression head. Experiments show that our TransT achieves very promising results on six challenging datasets, especially on large-scale LaSOT, TrackingNet, and GOT-10k benchmarks. Our tracker runs at approximatively 50 fps on GPU. Code and models are available at https://github.com/chenxin-dlut/TransT."
                },
                {
                    "title": "Empowering Things With Intelligence: A Survey of the Progress, Challenges, and Opportunities in Artificial Intelligence of Things",
                    "abstract": "In the Internet-of-Things (IoT) era, billions of sensors and devices collect and process data from the environment, transmit them to cloud centers, and receive feedback via the Internet for connectivity and perception. However, transmitting massive amounts of heterogeneous data, perceiving complex environments from these data, and then making smart decisions in a timely manner are difficult. Artificial intelligence (AI), especially deep learning, is now a proven success in various areas, including computer vision, speech recognition, and natural language processing. AI introduced into the IoT heralds the era of AI of things (AIoT). This article presents a comprehensive survey on AIoT to show how AI can empower the IoT to make it faster, smarter, greener, and safer. Specifically, we briefly present the AIoT architecture in the context of cloud computing, fog computing, and edge computing. Then, we present progress in AI research for IoT from four perspectives: 1) perceiving; 2) learning; 3) reasoning; and 4) behaving. Next, we summarize some promising applications of AIoT that are likely to profoundly reshape our world. Finally, we highlight the challenges facing AIoT and some potential research opportunities."
                },
                {
                    "title": "Point Transformer",
                    "abstract": "In this work, we present Point Transformer, a deep neural network that operates directly on unordered and unstructured point sets. We design Point Transformer to extract local and global features and relate both representations by introducing the local-global attention mechanism, which aims to capture spatial point relations and shape information. For that purpose, we propose SortNet, as part of the Point Transformer, which induces input permutation invariance by selecting points based on a learned score. The output of Point Transformer is a sorted and permutation invariant feature list that can directly be incorporated into common computer vision applications. We evaluate our approach on standard classification and part segmentation benchmarks to demonstrate competitive results compared to the prior work."
                },
                {
                    "title": "P2B: Point-to-Box Network for 3D Object Tracking in Point Clouds",
                    "abstract": "Towards 3D object tracking in point clouds, a novel point-to-box network termed P2B is proposed in an end-to-end learning manner. Our main idea is to first localize potential target centers in 3D search area embedded with target information. Then point-driven 3D target proposal and verification are executed jointly. In this way, the time-consuming 3D exhaustive search can be avoided. Specifically, we first sample seeds from the point clouds in template and search area respectively. Then, we execute permutation-invariant feature augmentation to embed target clues from template into search area seeds and represent them with target-specific features. Consequently, the augmented search area seeds regress the potential target centers via Hough voting. The centers are further strengthened with seed-wise targetness scores. Finally, each center clusters its neighbors to leverage the ensemble power for joint 3D target proposal and verification. We apply PointNet++ as our backbone and experiments on KITTI tracking dataset demonstrate P2B\u2019s superiority (~10%\u2019s improvement over state-of-the-art). Note that P2B can run with 40FPS on a single NVIDIA 1080Ti GPU. Our code and model are available at https://github.com/HaozheQi/P2B."
                },
                {
                    "title": "MOT20: A benchmark for multi object tracking in crowded scenes",
                    "abstract": "Standardized benchmarks are crucial for the majority of computer vision applications. Although leaderboards and ranking tables should not be over-claimed, benchmarks often provide the most objective measure of performance and are therefore important guides for research. The benchmark for Multiple Object Tracking, MOTChallenge, was launched with the goal to establish a standardized evaluation of multiple object tracking methods. The challenge focuses on multiple people tracking, since pedestrians are well studied in the tracking community, and precise tracking and detection has high practical relevance. Since the first release, MOT15, MOT16, and MOT17 have tremendously contributed to the community by introducing a clean dataset and precise framework to benchmark multi-object trackers. In this paper, we present our MOT20benchmark, consisting of 8 new sequences depicting very crowded challenging scenes. The benchmark was presented first at the 4thBMTT MOT Challenge Workshop at the Computer Vision and Pattern Recognition Conference (CVPR) 2019, and gives to chance to evaluate state-of-the-art methods for multiple object tracking when handling extremely crowded scenarios."
                },
                {
                    "title": "Scalability in Perception for Autonomous Driving: Waymo Open Dataset",
                    "abstract": "The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data. Existing self-driving datasets are limited in the scale and variation of the environments they capture, even though generalization within and between operating regions is crucial to the over-all viability of the technology. In an effort to help align the research community\u2019s contributions with real-world self-driving problems, we introduce a new large scale, high quality, diverse dataset. Our new dataset consists of 1150 scenes that each span 20 seconds, consisting of well synchronized and calibrated high quality LiDAR and camera data captured across a range of urban and suburban geographies. It is 15x more diverse than the largest camera+LiDAR dataset available based on our proposed diversity metric. We exhaustively annotated this data with 2D (camera image) and 3D (LiDAR) bounding boxes, with consistent identifiers across frames. Finally, we provide strong baselines for 2D as well as 3D detection and tracking tasks. We further study the effects of dataset size and generalization across geographies on 3D detection methods. Find data, code and more up-to-date information at http://www.waymo.com/open."
                },
                {
                    "title": "Deep Hough Voting for 3D Object Detection in Point Clouds",
                    "abstract": "Current 3D object detection methods are heavily influenced by 2D detectors. In order to leverage architectures in 2D detectors, they often convert 3D point clouds to regular grids (i.e., to voxel grids or to bird's eye view images), or rely on detection in 2D images to propose 3D boxes. Few works have attempted to directly detect objects in point clouds. In this work, we return to first principles to construct a 3D detection pipeline for point cloud data and as generic as possible. However, due to the sparse nature of the data -- samples from 2D manifolds in 3D space -- we face a major challenge when directly predicting bounding box parameters from scene points: a 3D object centroid can be far from any surface point thus hard to regress accurately in one step. To address the challenge, we propose VoteNet, an end-to-end 3D object detection network based on a synergy of deep point set networks and Hough voting. Our model achieves state-of-the-art 3D detection on two large datasets of real 3D scans, ScanNet and SUN RGB-D with a simple design, compact model size and high efficiency. Remarkably, VoteNet outperforms previous methods by using purely geometric information without relying on color images."
                },
                {
                    "title": "nuScenes: A Multimodal Dataset for Autonomous Driving",
                    "abstract": "Robust detection and tracking of objects is crucial for the deployment of autonomous vehicle technology. Image based benchmark datasets have driven development in computer vision tasks such as object detection, tracking and segmentation of agents in the environment. Most autonomous vehicles, however, carry a combination of cameras and range sensors such as lidar and radar. As machine learning based methods for detection and tracking become more prevalent, there is a need to train and evaluate such methods on datasets containing range sensor data along with images. In this work we present nuTonomy scenes (nuScenes), the first dataset to carry the full autonomous vehicle sensor suite: 6 cameras, 5 radars and 1 lidar, all with full 360 degree field of view. nuScenes comprises 1000 scenes, each 20s long and fully annotated with 3D bounding boxes for 23 classes and 8 attributes. It has 7x as many annotations and 100x as many images as the pioneering KITTI dataset. We define novel 3D detection and tracking metrics. We also provide careful dataset analysis as well as baselines for lidar and image based detection and tracking. Data, development kit and more information are available online."
                },
                {
                    "title": "Leveraging Shape Completion for 3D Siamese Tracking",
                    "abstract": "Point clouds are challenging to process due to their sparsity, therefore autonomous vehicles rely more on appearance attributes than pure geometric features. However, 3D LIDAR perception can provide crucial information for urban navigation in challenging light or weather conditions. In this paper, we investigate the versatility of Shape Completion for 3D Object Tracking in LIDAR point clouds. We design a Siamese tracker that encodes model and candidate shapes into a compact latent representation. We regularize the encoding by enforcing the latent representation to decode into an object model shape. We observe that 3D object tracking and 3D shape completion complement each other. Learning a more meaningful latent representation shows better discriminatory capabilities, leading to improved tracking performance. We test our method on the KITTI Tracking set using car 3D bounding boxes. Our model reaches a 76.94% Success rate and 81.38% Precision for 3D Object Tracking, with the shape completion regularization leading to an improvement of 3% in both metrics."
                },
                {
                    "title": "High Performance Visual Tracking with Siamese Region Proposal Network",
                    "abstract": "Visual object tracking has been a fundamental topic in recent years and many deep learning based trackers have achieved state-of-the-art performance on multiple benchmarks. However, most of these trackers can hardly get top performance with real-time speed. In this paper, we propose the Siamese region proposal network (Siamese-RPN) which is end-to-end trained off-line with large-scale image pairs. Specifically, it consists of Siamese subnetwork for feature extraction and region proposal subnetwork including the classification branch and regression branch. In the inference phase, the proposed framework is formulated as a local one-shot detection task. We can pre-compute the template branch of the Siamese subnetwork and formulate the correlation layers as trivial convolution layers to perform online tracking. Benefit from the proposal refinement, traditional multi-scale test and online fine-tuning can be discarded. The Siamese-RPN runs at 160 FPS while achieving leading performance in VOT2015, VOT2016 and VOT2017 real-time challenges."
                },
                {
                    "title": "PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space",
                    "abstract": "Few prior works study deep learning on point sets. PointNet by Qi et al. is a pioneer in this direction. However, by design PointNet does not capture local structures induced by the metric space points live in, limiting its ability to recognize fine-grained patterns and generalizability to complex scenes. In this work, we introduce a hierarchical neural network that applies PointNet recursively on a nested partitioning of the input point set. By exploiting metric space distances, our network is able to learn local features with increasing contextual scales. With further observation that point sets are usually sampled with varying densities, which results in greatly decreased performance for networks trained on uniform densities, we propose novel set learning layers to adaptively combine features from multiple scales. Experiments show that our network called PointNet++ is able to learn deep point set features efficiently and robustly. In particular, results significantly better than state-of-the-art have been obtained on challenging benchmarks of 3D point clouds."
                },
                {
                    "title": "PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation",
                    "abstract": "Point cloud is an important type of geometric data structure. Due to its irregular format, most researchers transform such data to regular 3D voxel grids or collections of images. This, however, renders data unnecessarily voluminous and causes issues. In this paper, we design a novel type of neural network that directly consumes point clouds, which well respects the permutation invariance of points in the input. Our network, named PointNet, provides a unified architecture for applications ranging from object classification, part segmentation, to scene semantic parsing. Though simple, PointNet is highly efficient and effective. Empirically, it shows strong performance on par or even better than state of the art. Theoretically, we provide analysis towards understanding of what the network has learnt and why the network is robust with respect to input perturbation and corruption."
                },
                {
                    "title": "MOT16: A Benchmark for Multi-Object Tracking",
                    "abstract": "Standardized benchmarks are crucial for the majority of computer vision applications. Although leaderboards and ranking tables should not be over-claimed, benchmarks often provide the most objective measure of performance and are therefore important guides for reseach. \nRecently, a new benchmark for Multiple Object Tracking, MOTChallenge, was launched with the goal of collecting existing and new data and creating a framework for the standardized evaluation of multiple object tracking methods. The first release of the benchmark focuses on multiple people tracking, since pedestrians are by far the most studied object in the tracking community. This paper accompanies a new release of the MOTChallenge benchmark. Unlike the initial release, all videos of MOT16 have been carefully annotated following a consistent protocol. Moreover, it not only offers a significant increase in the number of labeled boxes, but also provides multiple object classes beside pedestrians and the level of visibility for every single object of interest."
                },
                {
                    "title": "Online Object Tracking: A Benchmark",
                    "abstract": "Object tracking is one of the most important components in numerous applications of computer vision. While much progress has been made in recent years with efforts on sharing code and datasets, it is of great importance to develop a library and benchmark to gauge the state of the art. After briefly reviewing recent advances of online object tracking, we carry out large scale experiments with various evaluation criteria to understand how these algorithms perform. The test image sequences are annotated with different attributes for performance evaluation and analysis. By analyzing quantitative results, we identify effective approaches for robust tracking and provide potential future research directions in this field."
                },
                {
                    "title": "Are we ready for autonomous driving? The KITTI vision benchmark suite",
                    "abstract": "Today, visual recognition systems are still rarely employed in robotics applications. Perhaps one of the main reasons for this is the lack of demanding benchmarks that mimic such scenarios. In this paper, we take advantage of our autonomous driving platform to develop novel challenging benchmarks for the tasks of stereo, optical flow, visual odometry/SLAM and 3D object detection. Our recording platform is equipped with four high resolution video cameras, a Velodyne laser scanner and a state-of-the-art localization system. Our benchmarks comprise 389 stereo and optical flow image pairs, stereo visual odometry sequences of 39.2 km length, and more than 200k 3D object annotations captured in cluttered scenarios (up to 15 cars and 30 pedestrians are visible per image). Results from state-of-the-art algorithms reveal that methods ranking high on established datasets such as Middlebury perform below average when being moved outside the laboratory to the real world. Our goal is to reduce this bias by providing challenging benchmarks with novel difficulties to the computer vision community. Our benchmarks are available online at: www.cvlibs.net/datasets/kitti."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a unified representation network that effectively learns deformable groups and facilitates feature interaction for various object categories in point cloud data?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning, particularly in applications involving 3D object recognition and segmentation. A successful unified representation network can lead to improved performance in tasks such as autonomous driving, robotics, and augmented reality. By addressing this question, we can enhance the understanding of feature representation in point clouds, paving the way for future research that explores more complex object interactions and real-world applications.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the inherent complexity of point cloud data, which is often irregular and sparse. Naive approaches may fail due to their inability to adaptively learn the varying shapes and sizes of different object categories. Technical obstacles include the need for effective feature propagation and interaction mechanisms, as well as the difficulty in optimizing hyper-parameters like the scale factor without introducing noise or losing valuable information.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either rigid object representations or lacked the capability to adaptively learn from diverse object categories. Limitations in existing solutions include insufficient attention to feature interactions and the challenges of handling deformable groups. Our approach differs by integrating a group regression module and a vector-attention mechanism, which together enhance the learning of a unified feature representation, addressing the shortcomings of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the AdaFormer network, which utilizes a group regression module for learning deformable groups and a vector-attention mechanism for feature interaction. We will conduct experiments using the KITTI dataset, evaluating performance based on Success and Precision metrics. Expected outcomes include improved performance across various object categories, particularly in challenging cases like pedestrians, demonstrating the effectiveness of our unified representation approach."
            }
        },
        "author_data": {
            "215109a1-8d98-49af-960e-6b3151b2d7aa": {
                "pk": "215109a1-8d98-49af-960e-6b3151b2d7aa",
                "name": "Jiahao Nie",
                "collaborators": [
                    "Zhiwei He",
                    "Xudong Lv",
                    "Mingyu Gao",
                    "Yuxiang Yang",
                    "Jing Zhang",
                    "Zhengyi Bao",
                    "Mingchen Gao",
                    "Yuxuan Lu",
                    "Hongjie Gu",
                    "Huipin Lin"
                ],
                "domain": [
                    "3D Object Tracking",
                    "LiDAR",
                    "Deep Learning",
                    "Anomaly Detection"
                ],
                "publications": [
                    {
                        "title": "VoxelTrack: Exploring Voxel Representation for 3D Point Cloud Object Tracking",
                        "abstract": "Current LiDAR point cloud-based 3D single object tracking (SOT) methods typically rely on point-based representation network. Despite demonstrated success, such networks suffer from some fundamental problems: 1) It contains pooling operation to cope with inherently disordered point clouds, hindering the capture of 3D spatial information that is useful for tracking, a regression task. 2) The adopted set abstraction operation hardly handles density-inconsistent point clouds, also preventing 3D spatial information from being modeled. To solve these problems, we introduce a novel tracking framework, termed VoxelTrack. By voxelizing inherently disordered point clouds into 3D voxels and extracting their features via sparse convolution blocks, VoxelTrack effectively models precise and robust 3D spatial information, thereby guiding accurate position prediction for tracked objects. Moreover, VoxelTrack incorporates a dual-stream encoder with cross-iterative feature fusion module to further explore fine-grained 3D spatial information for tracking. Benefiting from accurate 3D spatial information being modeled, our VoxelTrack simplifies tracking pipeline with a single regression loss. Extensive experiments are conducted on three widely-adopted datasets including KITTI, NuScenes and Waymo Open Dataset. The experimental results confirm that VoxelTrack achieves state-of-the-art performance (88.3%, 71.4% and 63.6% mean precision on the three datasets, respectively), and outperforms the existing trackers with a real-time speed of 36 Fps on a single TITAN RTX GPU. The source code and model will be released."
                    },
                    {
                        "title": "TTSNet: State-of-Charge Estimation of Li-Ion Battery in Electrical Vehicles With Temporal Transformer-Based Sequence Network",
                        "abstract": "Accurate estimating the state-of-charge (SOC) of Li-ion battery contributes significantly to electric vehicle safety. Existing methods typically focus on the traditional recurrent neural networks to encode time series features for SOC estimation. However, these methods rely solely on their own structure to extract time series correlated features, ignoring a significant amount of information on temporal dimension. To address this issue, this paper proposes a temporal transformer-based sequence network (TTSNet) that can make full use of temporal dimensional information to model the relationship between the input and SOC. Specifically, the proposed network splits the raw data into three branches including voltage, current, and temperature, as well as extracts the corresponding primary semantic features. It then uses a temporal transformer to effectively encode the features of temporal dimensional information. The resulting features are further fed into an attention-guided feature fusion module to interact information among voltage, current, and temperature branches for subsequent SOC estimation. To enhance the network's resilience for long time sequences, a sliding time window technique is introduced to pre-process the raw data. Besides, a Kalman filter is incorporated as post-processing to smooth the output to guide a more accurate SOC estimation. Comprehensive experiments are conducted on battery open datasets and vehicle operation datasets to verify the proposed method. The results demonstrate that the proposed method achieves high accuracy and strong robustness in both datasets, with average MAE, RMSE, and $\\mathrm{R^{2}}$ values of 0.506%, 0.694%, and 99.791%, respectively."
                    },
                    {
                        "title": "TM2B: Transformer-Based Motion-to-Box Network for 3D Single Object Tracking on Point Clouds",
                        "abstract": "3D single object tracking plays a crucial role in numerous applications such as autonomous driving. Recent trackers based on motion-centric paradigm perform well as they exploit motion cues to infer target relative motion across successive frames, which effectively overcome significant appearance variations of targets and distractors caused by occlusion. However, such a motion-centric paradigm tends to require multi-stage motion-to-box to refine the motion cues, which suffers from tedious hyper-parameter tuning and elaborate subtask designs. In this letter, we propose a novel transformer-based motion-to-box network (TM2B), which employs a learnable relation modeling transformer (LRMT) to generate accurate motion cues without multi-stage refinements. Our proposed LRMT contains two novel attention mechanisms: hierarchical interactive attention and learnable query attention. The former attention builds a learnable number-fixed sampling sets for each query on multi-scale feature maps, enabling each query to adaptively select prominent sampling elements, thus effectively encoding multi-scale features in a lightweight manner, while the latter calculates the weighted sum of the encoded features with learnable global query, enabling to extract valuable motion cues from all available features, thereby achieving accurate object tracking. Extensive experiments demonstrate that TM2B achieves state-of-the-art performance on KITTI, NuScenes and Waymo Open Dataset, while obtaining a significant improvement in inference speed over previous leading methods, achieving 56.8 FPS on a single NVIDIA 1080Ti GPU. The code is available at TM2B."
                    },
                    {
                        "title": "P2P: Part-to-Part Motion Cues Guide a Strong Tracking Framework for LiDAR Point Clouds",
                        "abstract": "3D single object tracking (SOT) methods based on appearance matching has long suffered from insufficient appearance information incurred by incomplete, textureless and semantically deficient LiDAR point clouds. While motion paradigm exploits motion cues instead of appearance matching for tracking, it incurs complex multi-stage processing and segmentation module. In this paper, we first provide in-depth explorations on motion paradigm, which proves that (\\textbf{i}) it is feasible to directly infer target relative motion from point clouds across consecutive frames; (\\textbf{ii}) fine-grained information comparison between consecutive point clouds facilitates target motion modeling. We thereby propose to perform part-to-part motion modeling for consecutive point clouds and introduce a novel tracking framework, termed \\textbf{P2P}. The novel framework fuses each corresponding part information between consecutive point clouds, effectively exploring detailed information changes and thus modeling accurate target-related motion cues. Following this framework, we present P2P-point and P2P-voxel models, incorporating implicit and explicit part-to-part motion modeling by point- and voxel-based representation, respectively. Without bells and whistles, P2P-voxel sets a new state-of-the-art performance ($\\sim$\\textbf{89\\%}, \\textbf{72\\%} and \\textbf{63\\%} precision on KITTI, NuScenes and Waymo Open Dataset, respectively). Moreover, under the same point-based representation, P2P-point outperforms the previous motion tracker M$^2$Track by \\textbf{3.3\\%} and \\textbf{6.7\\%} on the KITTI and NuScenes, while running at a considerably high speed of \\textbf{107 Fps} on a single RTX3090 GPU. The source code and pre-trained models are available at \\url{https://github.com/haooozi/P2P}."
                    },
                    {
                        "title": "DFSA-DAN: dynamic fusion of statistical metric and adversarial learning for domain adaptation network based intelligent fault diagnosis",
                        "abstract": "The advancement of deep transfer learning has motivated research into the realization of intelligent fault diagnosis schemes for rolling bearing. Nevertheless, existing research rarely provides further insight into the importance of statistical distance metric-based methods and adversarial learning-based methods in domain adaptation, and the commonly used feature extractors are more difficult to extract features suitable for domain transformation. In this paper, a dynamic fusion of statistical metric and adversarial learning for domain adaptation network is proposed to achieve a dynamic measure of the importance of different domain adaptation methods. This new model utilizes a local maximum mean discrepancy metric to adjust the conditional distribution and adversarial training to adjust the marginal distribution between domains. Meanwhile, to assess the importance of the two distributions, a dynamic adaptation factor is introduced for dynamic evaluation. In addition, to extract features that are more suitable for domain transformation, the model incorporates a dual depth convolutional path with an attention mechanism as a feature extractor, enabling multi-scale feature extraction. Experimental results demonstrate the model\u2019s superior generalization capability and robustness, enabling effective cross-domain fault diagnosis in diverse scenarios."
                    },
                    {
                        "title": "TPAD: Temporal-Pattern-Based Neural Network Model for Anomaly Detection in Multivariate Time Series",
                        "abstract": "The mixing process of battery production contains a variety of monitoring sensors. These sensors generate a large number of multivariate time series during production, which reflect the potential operating conditions of the production equipment. By accurately detecting anomalies in the equipment, battery quality can be improved. However, due to the diverse types of anomalies in the mixing process and the large number of subtle anomalies, it is challenging to construct a robust and accurate anomaly detection model. In this article, we propose TPAD, a multivariate time series anomaly detection model based on temporal pattern, which learns serial temporal-pattern-based on attention mechanism and detects anomalies by temporal pattern discrepancy, making the model adaptive to detect multiple types of anomalies. Moreover, the model improves the accuracy of detecting subtle anomalies through an adversarial training reconstruction method. Finally, we conducted experiments on two public datasets and a battery mixing process dataset from a battery factory, and the experimental results outperformed the baseline methods, with an average F1 score of over 92% on the three datasets."
                    },
                    {
                        "title": "OSP2B: One-Stage Point-to-Box Network for 3D Siamese Tracking",
                        "abstract": "Two-stage point-to-box network acts as a critical role in the recent popular 3D Siamese tracking paradigm, which first generates proposals and then predicts corresponding proposal-wise scores. However, such a network suffers from tedious hyper-parameter tuning and task misalignment, limiting the tracking performance. Towards these concerns, we propose a simple yet effective one-stage point-to-box network for point cloud-based 3D single object tracking. It synchronizes 3D proposal generation and center-ness score prediction by a parallel predictor without tedious hyper-parameters. To guide a task-aligned score ranking of proposals, a center-aware focal loss is proposed to supervise the training of the center-ness branch, which enhances the network's discriminative ability to distinguish proposals of different quality. Besides, we design a binary target classifier to identify target-relevant points. By integrating the derived classification scores with the center-ness scores, the resulting network can effectively suppress interference proposals and further mitigate task misalignment. Finally, we present a novel one-stage Siamese tracker OSP2B equipped with the designed network. Extensive experiments on challenging benchmarks including KITTI and Waymo SOT Dataset show that our OSP2B achieves leading performance with a considerable real-time speed."
                    },
                    {
                        "title": "GLT-T++: Global-Local Transformer for 3D Siamese Tracking with Ranking Loss",
                        "abstract": "Siamese trackers based on 3D region proposal network (RPN) have shown remarkable success with deep Hough voting. However, using a single seed point feature as the cue for voting fails to produce high-quality 3D proposals. Additionally, the equal treatment of seed points in the voting process, regardless of their significance, exacerbates this limitation. To address these challenges, we propose a novel transformer-based voting scheme to generate better proposals. Specifically, a global-local transformer (GLT) module is devised to integrate object- and patch-aware geometric priors into seed point features, resulting in robust and accurate cues for offset learning of seed points. To train the GLT module, we introduce an importance prediction branch that learns the potential importance weights of seed points as a training constraint. Incorporating this transformer-based voting scheme into 3D RPN, a novel Siamese method dubbed GLT-T is developed for 3D single object tracking on point clouds. Moreover, we identify that the highest-scored proposal in the Siamese paradigm may not be the most accurate proposal, which limits tracking performance. Towards this concern, we approach the binary score prediction task as a ranking problem, and design a target-aware ranking loss and a localization-aware ranking loss to produce accurate ranking of proposals. With the ranking losses, we further present GLT-T++, an enhanced version of GLT-T. Extensive experiments on multiple benchmarks demonstrate that our GLT-T and GLT-T++ outperform state-of-the-art methods in terms of tracking accuracy while maintaining a real-time inference speed. The source code will be made available at https://github.com/haooozi/GLT-T."
                    },
                    {
                        "title": "BEVTrack: A Simple Baseline for 3D Single Object Tracking in Bird's-Eye-View",
                        "abstract": "3D single object tracking (SOT) in point clouds is still a challenging problem due to appearance variation, distractors, and high sparsity of point clouds. Notably, in autonomous driving scenarios, the target object typically maintains spatial adjacency across consecutive frames, predominantly moving horizontally. This spatial continuity offers valuable prior knowledge for target localization. However, existing trackers, which often employ point-wise representations, struggle to efficiently utilize this knowledge owing to the irregular format of such representations. Consequently, they require elaborate designs and solving multiple subtasks to establish spatial correspondence. In this paper, we introduce BEVTrack, a simple yet strong baseline framework for 3D SOT. After converting consecutive point clouds into the common Bird\u2019s-Eye-View representation, BEVTrack inherently encodes spatial proximity and adeptly captures motion cues for tracking via a simple element-wise operation and convolutional layers. Additionally, to better deal with objects having diverse sizes and moving patterns, BEVTrack directly learns the underlying motion distribution rather than making a fixed Laplacian or Gaussian assumption as in previous works. Without bells and whistles, BEVTrack achieves state-of-the-art performance on KITTI and NuScenes datasets while maintaining a high inference speed of 122 FPS. The code will be released at BEVTrack."
                    }
                ]
            },
            "b60fdba3-9951-4398-a01a-55214feda6f7": {
                "pk": "b60fdba3-9951-4398-a01a-55214feda6f7",
                "name": "Zhiwei He",
                "collaborators": [
                    "Jiahao Nie",
                    "Mingyu Gao",
                    "Zhengyi Bao",
                    "Yuxiang Yang",
                    "Chenjie Du",
                    "Xudong Lv",
                    "Ran Jin",
                    "Hao Tang",
                    "Qiuping Jiang",
                    "Yuxuan Lu"
                ],
                "domain": [
                    "Deep Learning",
                    "Computer Vision",
                    "3D Tracking",
                    "Human Activity Recognition"
                ],
                "publications": [
                    {
                        "title": "SKIP: Accurate Fall Detection Based on Skeleton Keypoint Association and Critical Feature Perception",
                        "abstract": "As deep learning technology advances, human fall detection (HFD) leveraging convolutional neural networks (CNNs) has recently garnered significant interest within the research community. However, most existing works ignore the cross-frame association of skeleton keypoints and aggregation of feature representations. To address this, we first introduce an image preprocessing (IPP) module, which enhances the foreground and weakens the background. Diverging from common practices that employ the off-the-shelf detector for target position estimation, our skeleton keypoint detection and association (SKDA) module is designed to detect and cross-frame associate the skeleton keypoints with high affinity. This design reduces the misleading impact of ambiguous detections and ensures the continuity of long-range trajectories. Further, our critical feature perception (CFP) module is crafted to help the model learn more discriminative feature representations for human activity classification. Incorporating these components mentioned above, we introduce SKIP, a novel human fall detection approach, showcasing improved detection precision. Evaluations on the publicly available telecommunication system team v2 (TSTv2) and self-build datasets show SKIP\u2019s superior performance."
                    },
                    {
                        "title": "VoxelTrack: Exploring Voxel Representation for 3D Point Cloud Object Tracking",
                        "abstract": "Current LiDAR point cloud-based 3D single object tracking (SOT) methods typically rely on point-based representation network. Despite demonstrated success, such networks suffer from some fundamental problems: 1) It contains pooling operation to cope with inherently disordered point clouds, hindering the capture of 3D spatial information that is useful for tracking, a regression task. 2) The adopted set abstraction operation hardly handles density-inconsistent point clouds, also preventing 3D spatial information from being modeled. To solve these problems, we introduce a novel tracking framework, termed VoxelTrack. By voxelizing inherently disordered point clouds into 3D voxels and extracting their features via sparse convolution blocks, VoxelTrack effectively models precise and robust 3D spatial information, thereby guiding accurate position prediction for tracked objects. Moreover, VoxelTrack incorporates a dual-stream encoder with cross-iterative feature fusion module to further explore fine-grained 3D spatial information for tracking. Benefiting from accurate 3D spatial information being modeled, our VoxelTrack simplifies tracking pipeline with a single regression loss. Extensive experiments are conducted on three widely-adopted datasets including KITTI, NuScenes and Waymo Open Dataset. The experimental results confirm that VoxelTrack achieves state-of-the-art performance (88.3%, 71.4% and 63.6% mean precision on the three datasets, respectively), and outperforms the existing trackers with a real-time speed of 36 Fps on a single TITAN RTX GPU. The source code and model will be released."
                    },
                    {
                        "title": "TTSNet: State-of-Charge Estimation of Li-Ion Battery in Electrical Vehicles With Temporal Transformer-Based Sequence Network",
                        "abstract": "Accurate estimating the state-of-charge (SOC) of Li-ion battery contributes significantly to electric vehicle safety. Existing methods typically focus on the traditional recurrent neural networks to encode time series features for SOC estimation. However, these methods rely solely on their own structure to extract time series correlated features, ignoring a significant amount of information on temporal dimension. To address this issue, this paper proposes a temporal transformer-based sequence network (TTSNet) that can make full use of temporal dimensional information to model the relationship between the input and SOC. Specifically, the proposed network splits the raw data into three branches including voltage, current, and temperature, as well as extracts the corresponding primary semantic features. It then uses a temporal transformer to effectively encode the features of temporal dimensional information. The resulting features are further fed into an attention-guided feature fusion module to interact information among voltage, current, and temperature branches for subsequent SOC estimation. To enhance the network's resilience for long time sequences, a sliding time window technique is introduced to pre-process the raw data. Besides, a Kalman filter is incorporated as post-processing to smooth the output to guide a more accurate SOC estimation. Comprehensive experiments are conducted on battery open datasets and vehicle operation datasets to verify the proposed method. The results demonstrate that the proposed method achieves high accuracy and strong robustness in both datasets, with average MAE, RMSE, and $\\mathrm{R^{2}}$ values of 0.506%, 0.694%, and 99.791%, respectively."
                    },
                    {
                        "title": "TM2B: Transformer-Based Motion-to-Box Network for 3D Single Object Tracking on Point Clouds",
                        "abstract": "3D single object tracking plays a crucial role in numerous applications such as autonomous driving. Recent trackers based on motion-centric paradigm perform well as they exploit motion cues to infer target relative motion across successive frames, which effectively overcome significant appearance variations of targets and distractors caused by occlusion. However, such a motion-centric paradigm tends to require multi-stage motion-to-box to refine the motion cues, which suffers from tedious hyper-parameter tuning and elaborate subtask designs. In this letter, we propose a novel transformer-based motion-to-box network (TM2B), which employs a learnable relation modeling transformer (LRMT) to generate accurate motion cues without multi-stage refinements. Our proposed LRMT contains two novel attention mechanisms: hierarchical interactive attention and learnable query attention. The former attention builds a learnable number-fixed sampling sets for each query on multi-scale feature maps, enabling each query to adaptively select prominent sampling elements, thus effectively encoding multi-scale features in a lightweight manner, while the latter calculates the weighted sum of the encoded features with learnable global query, enabling to extract valuable motion cues from all available features, thereby achieving accurate object tracking. Extensive experiments demonstrate that TM2B achieves state-of-the-art performance on KITTI, NuScenes and Waymo Open Dataset, while obtaining a significant improvement in inference speed over previous leading methods, achieving 56.8 FPS on a single NVIDIA 1080Ti GPU. The code is available at TM2B."
                    },
                    {
                        "title": "P2P: Part-to-Part Motion Cues Guide a Strong Tracking Framework for LiDAR Point Clouds",
                        "abstract": "3D single object tracking (SOT) methods based on appearance matching has long suffered from insufficient appearance information incurred by incomplete, textureless and semantically deficient LiDAR point clouds. While motion paradigm exploits motion cues instead of appearance matching for tracking, it incurs complex multi-stage processing and segmentation module. In this paper, we first provide in-depth explorations on motion paradigm, which proves that (\\textbf{i}) it is feasible to directly infer target relative motion from point clouds across consecutive frames; (\\textbf{ii}) fine-grained information comparison between consecutive point clouds facilitates target motion modeling. We thereby propose to perform part-to-part motion modeling for consecutive point clouds and introduce a novel tracking framework, termed \\textbf{P2P}. The novel framework fuses each corresponding part information between consecutive point clouds, effectively exploring detailed information changes and thus modeling accurate target-related motion cues. Following this framework, we present P2P-point and P2P-voxel models, incorporating implicit and explicit part-to-part motion modeling by point- and voxel-based representation, respectively. Without bells and whistles, P2P-voxel sets a new state-of-the-art performance ($\\sim$\\textbf{89\\%}, \\textbf{72\\%} and \\textbf{63\\%} precision on KITTI, NuScenes and Waymo Open Dataset, respectively). Moreover, under the same point-based representation, P2P-point outperforms the previous motion tracker M$^2$Track by \\textbf{3.3\\%} and \\textbf{6.7\\%} on the KITTI and NuScenes, while running at a considerably high speed of \\textbf{107 Fps} on a single RTX3090 GPU. The source code and pre-trained models are available at \\url{https://github.com/haooozi/P2P}."
                    },
                    {
                        "title": "DFSA-DAN: dynamic fusion of statistical metric and adversarial learning for domain adaptation network based intelligent fault diagnosis",
                        "abstract": "The advancement of deep transfer learning has motivated research into the realization of intelligent fault diagnosis schemes for rolling bearing. Nevertheless, existing research rarely provides further insight into the importance of statistical distance metric-based methods and adversarial learning-based methods in domain adaptation, and the commonly used feature extractors are more difficult to extract features suitable for domain transformation. In this paper, a dynamic fusion of statistical metric and adversarial learning for domain adaptation network is proposed to achieve a dynamic measure of the importance of different domain adaptation methods. This new model utilizes a local maximum mean discrepancy metric to adjust the conditional distribution and adversarial training to adjust the marginal distribution between domains. Meanwhile, to assess the importance of the two distributions, a dynamic adaptation factor is introduced for dynamic evaluation. In addition, to extract features that are more suitable for domain transformation, the model incorporates a dual depth convolutional path with an attention mechanism as a feature extractor, enabling multi-scale feature extraction. Experimental results demonstrate the model\u2019s superior generalization capability and robustness, enabling effective cross-domain fault diagnosis in diverse scenarios."
                    },
                    {
                        "title": "A Correlation-Augmented Informer-Based Method for State-of-Health Estimation of Li-Ion Batteries",
                        "abstract": "Accurate estimation of the state of health (SOH) of lithium-ion batteries is crucial to ensure their safe and reliable use. To address the issue of neglecting correlations between different charge/discharge cycles in current neural network (NN)-based methods, this article introduces a correlation-augmented informer network. Specifically, the multiple correlation coefficient (MCC) is used to analyze the correlation between different cycles, and then a correlation-augmented informer containing embedding, encoder, and decoder is constructed. To emphasize cycles with stronger correlations during the estimation process, we incorporate multihead local self-attention mechanisms. We validate our approach on two publicly available datasets using mean absolute error (MAE), mean absolute percentage error (MAPE), and root-mean-square error (RMSE), which are 0.2%, 0.3%, and 0.2% on the MIT dataset and 1.0%, 1.4%, and 1.8% on the CALCE dataset, respectively. These results demonstrate the superior accuracy and robustness of the proposed method in comparison to existing state-of-the-art NN methods."
                    },
                    {
                        "title": "Spatiotemporal Weighted Micro-Doppler Spectrum Design for Soft Synchronization FMCW Radar",
                        "abstract": "With the attention paid to privacy protection, human motion recognition (HMR) based on radars has attracted more and more attention. A soft synchronization frequency modulated continuous wave (SS-FMCW) radar has a wider application prospect in HMR by virtue of its signal diversity and structural flexibility. In this article, a new spatiotemporal weighted micro-Doppler (SWMD) spectrum is proposed, which is more competitive for the application of the SS-FMCW radar in HMR. First, to facilitate the understanding of the SWMD spectrum construction method, we introduce the principle behind the SS-FMCW radar based on intermittent ramp signals. Specifically, the SWMD spectrum is constructed by a weighted data fusion algorithm and a time series data fusion algorithm. In addition, we collect data from six different motions from eight testers and demonstrate and analyze the validity of the SWMD spectrum by multiple dimensions. The experimental results show that compared with the conventional time micro-Doppler (MD) spectrum, the accuracy of the SWMD spectrum in HMR is improved by about 7.2%, which is an average of multiple comparison experiments of different classification algorithms with different features."
                    },
                    {
                        "title": "Trajectory Prediction for Surrounding Traffic Participants via Local Perception and Attentive Map Encoding",
                        "abstract": "Predicting the trajectory of the surrounding traffic participants is a pivotal factor in autonomous driving. The prevalent methods in trajectory prediction hinge upon offline pre-stored high-precision maps to incorporate driving environment data. However, updating high-precision map in real time is a huge challenge in real-world scenarios. Based on the observation, we propose a novel local perception-based trajectory prediction method, which utilizes a bird\u2019s eye view (BEV) local perception map to replace the demand for high-precision map. Our BEV local perception map is built by leveraging sensor perception information, including semantic segmentation, 3D object detection and lidar point cloud data. Then, an attentive map encoding (AME) module is designed to extract the traffic environment features from the local perception map. Finally, an encoder-decoder structure based on Transformer is applied to achieve the trajectory prediction. Comprehensive experiments are conducted on the nuScenes dataset, demonstrating that our method can maintain prediction accuracy while avoiding dependence on high-precision maps."
                    },
                    {
                        "title": "TPAD: Temporal-Pattern-Based Neural Network Model for Anomaly Detection in Multivariate Time Series",
                        "abstract": "The mixing process of battery production contains a variety of monitoring sensors. These sensors generate a large number of multivariate time series during production, which reflect the potential operating conditions of the production equipment. By accurately detecting anomalies in the equipment, battery quality can be improved. However, due to the diverse types of anomalies in the mixing process and the large number of subtle anomalies, it is challenging to construct a robust and accurate anomaly detection model. In this article, we propose TPAD, a multivariate time series anomaly detection model based on temporal pattern, which learns serial temporal-pattern-based on attention mechanism and detects anomalies by temporal pattern discrepancy, making the model adaptive to detect multiple types of anomalies. Moreover, the model improves the accuracy of detecting subtle anomalies through an adversarial training reconstruction method. Finally, we conducted experiments on two public datasets and a battery mixing process dataset from a battery factory, and the experimental results outperformed the baseline methods, with an average F1 score of over 92% on the three datasets."
                    },
                    {
                        "title": "Enhanced Detection Model and Joint Scoring Strategy for Multi-Vehicle Tracking",
                        "abstract": "Multi-vehicle tracking is one of the most crucial components of an intelligent transportation system (ITS). However, when it comes to busy traffic flow, tracking targets robustly becomes more problematic due to occlusion, motion blur, high appearance similarity, etc. To achieve accurate and efficient tracking performance, we present a novel multi-vehicle tracking method based on the enhanced detection model and joint scoring strategy. Specifically, the former aims to (1) adopt lightweight yet efficient YOLOv5s to improve detection accuracy and running speed, and (2) incorporate the CBAM and transformer encoder modules into the detection model to generate the refined features for the target localization. The Latter preferentially provides high-confidence detections and tracklets for subsequent data association, significantly reducing the number of identity switches and redundant vehicle trajectories caused by mutual occlusion, similar object interference, etc. We evaluated the proposed multivehicle tracking approach on the UA-DETRAC vehicle tracking dataset and demonstrated its superior capabilities through intensive comparison and analysis. Moreover, our proposed method runs at 24.4 FPS on a single GPU and meets the real-time requirement."
                    },
                    {
                        "title": "OSP2B: One-Stage Point-to-Box Network for 3D Siamese Tracking",
                        "abstract": "Two-stage point-to-box network acts as a critical role in the recent popular 3D Siamese tracking paradigm, which first generates proposals and then predicts corresponding proposal-wise scores. However, such a network suffers from tedious hyper-parameter tuning and task misalignment, limiting the tracking performance. Towards these concerns, we propose a simple yet effective one-stage point-to-box network for point cloud-based 3D single object tracking. It synchronizes 3D proposal generation and center-ness score prediction by a parallel predictor without tedious hyper-parameters. To guide a task-aligned score ranking of proposals, a center-aware focal loss is proposed to supervise the training of the center-ness branch, which enhances the network's discriminative ability to distinguish proposals of different quality. Besides, we design a binary target classifier to identify target-relevant points. By integrating the derived classification scores with the center-ness scores, the resulting network can effectively suppress interference proposals and further mitigate task misalignment. Finally, we present a novel one-stage Siamese tracker OSP2B equipped with the designed network. Extensive experiments on challenging benchmarks including KITTI and Waymo SOT Dataset show that our OSP2B achieves leading performance with a considerable real-time speed."
                    }
                ]
            },
            "627ce4f9-0c07-461b-9be8-2888bf86802b": {
                "pk": "627ce4f9-0c07-461b-9be8-2888bf86802b",
                "name": "Xudong Lv",
                "collaborators": [
                    "Jiahao Nie",
                    "Zhiwei He",
                    "Yuxuan Lu",
                    "Hongjie Gu",
                    "Anqi Xu",
                    "Jiahao Jiang",
                    "Ping Li",
                    "Yuxiang Yang"
                ],
                "domain": [
                    "3D Object Tracking",
                    "Deep Learning",
                    "Reinforcement Learning",
                    "Point Cloud Processing"
                ],
                "publications": [
                    {
                        "title": "VoxelTrack: Exploring Voxel Representation for 3D Point Cloud Object Tracking",
                        "abstract": "Current LiDAR point cloud-based 3D single object tracking (SOT) methods typically rely on point-based representation network. Despite demonstrated success, such networks suffer from some fundamental problems: 1) It contains pooling operation to cope with inherently disordered point clouds, hindering the capture of 3D spatial information that is useful for tracking, a regression task. 2) The adopted set abstraction operation hardly handles density-inconsistent point clouds, also preventing 3D spatial information from being modeled. To solve these problems, we introduce a novel tracking framework, termed VoxelTrack. By voxelizing inherently disordered point clouds into 3D voxels and extracting their features via sparse convolution blocks, VoxelTrack effectively models precise and robust 3D spatial information, thereby guiding accurate position prediction for tracked objects. Moreover, VoxelTrack incorporates a dual-stream encoder with cross-iterative feature fusion module to further explore fine-grained 3D spatial information for tracking. Benefiting from accurate 3D spatial information being modeled, our VoxelTrack simplifies tracking pipeline with a single regression loss. Extensive experiments are conducted on three widely-adopted datasets including KITTI, NuScenes and Waymo Open Dataset. The experimental results confirm that VoxelTrack achieves state-of-the-art performance (88.3%, 71.4% and 63.6% mean precision on the three datasets, respectively), and outperforms the existing trackers with a real-time speed of 36 Fps on a single TITAN RTX GPU. The source code and model will be released."
                    },
                    {
                        "title": "TM2B: Transformer-Based Motion-to-Box Network for 3D Single Object Tracking on Point Clouds",
                        "abstract": "3D single object tracking plays a crucial role in numerous applications such as autonomous driving. Recent trackers based on motion-centric paradigm perform well as they exploit motion cues to infer target relative motion across successive frames, which effectively overcome significant appearance variations of targets and distractors caused by occlusion. However, such a motion-centric paradigm tends to require multi-stage motion-to-box to refine the motion cues, which suffers from tedious hyper-parameter tuning and elaborate subtask designs. In this letter, we propose a novel transformer-based motion-to-box network (TM2B), which employs a learnable relation modeling transformer (LRMT) to generate accurate motion cues without multi-stage refinements. Our proposed LRMT contains two novel attention mechanisms: hierarchical interactive attention and learnable query attention. The former attention builds a learnable number-fixed sampling sets for each query on multi-scale feature maps, enabling each query to adaptively select prominent sampling elements, thus effectively encoding multi-scale features in a lightweight manner, while the latter calculates the weighted sum of the encoded features with learnable global query, enabling to extract valuable motion cues from all available features, thereby achieving accurate object tracking. Extensive experiments demonstrate that TM2B achieves state-of-the-art performance on KITTI, NuScenes and Waymo Open Dataset, while obtaining a significant improvement in inference speed over previous leading methods, achieving 56.8 FPS on a single NVIDIA 1080Ti GPU. The code is available at TM2B."
                    },
                    {
                        "title": "DMCL: Robot Autonomous Navigation via Depth Image Masked Contrastive Learning",
                        "abstract": "Achieving high performance in deep reinforcement learning relies heavily on the ability to obtain good state representations from pixel inputs. However, learning an observation-space-to-action-space mapping from high-dimensional inputs is challenging in reinforcement learning, particularly when dealing with consecutive depth images as input states. In addition, we observe that the consecutive inputs of depth images are highly correlated for the autonomous navigation of a mobile robot, which inspires us to capture temporal correlations between consecutive inputs and infer scene change relationships. To this end, we propose a novel end-to-end robot vision navigation method dubbed DMCL, which obtains good spatial-temporal state representation via Depth image Masked Contrastive Learning. It reconstructs the latent representation from consecutive depth images masked in both spatial and temporal dimensions, resulting in a complete environment state representation. To obtain the optimal navigation policy, we leverage the Soft Actor-Critic reinforcement learning in conjunction with the above representation learning. Extensive experiments demonstrate that the proposed DMCL outperforms representative state-of-the-art methods. The source code will be made publicly available."
                    }
                ]
            },
            "aabbe5e3-e6b0-40ff-95ff-640488d278fe": {
                "pk": "aabbe5e3-e6b0-40ff-95ff-640488d278fe",
                "name": "Xueyi Zhou",
                "collaborators": [
                    "Jiahao Nie",
                    "Fei Xie",
                    "Sifan Zhou",
                    "Dong-Kyu Chae",
                    "Zhiwei He"
                ],
                "domain": [
                    "3D Object Tracking",
                    "LiDAR",
                    "Motion Modeling",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "P2P: Part-to-Part Motion Cues Guide a Strong Tracking Framework for LiDAR Point Clouds",
                        "abstract": "3D single object tracking (SOT) methods based on appearance matching has long suffered from insufficient appearance information incurred by incomplete, textureless and semantically deficient LiDAR point clouds. While motion paradigm exploits motion cues instead of appearance matching for tracking, it incurs complex multi-stage processing and segmentation module. In this paper, we first provide in-depth explorations on motion paradigm, which proves that (\\textbf{i}) it is feasible to directly infer target relative motion from point clouds across consecutive frames; (\\textbf{ii}) fine-grained information comparison between consecutive point clouds facilitates target motion modeling. We thereby propose to perform part-to-part motion modeling for consecutive point clouds and introduce a novel tracking framework, termed \\textbf{P2P}. The novel framework fuses each corresponding part information between consecutive point clouds, effectively exploring detailed information changes and thus modeling accurate target-related motion cues. Following this framework, we present P2P-point and P2P-voxel models, incorporating implicit and explicit part-to-part motion modeling by point- and voxel-based representation, respectively. Without bells and whistles, P2P-voxel sets a new state-of-the-art performance ($\\sim$\\textbf{89\\%}, \\textbf{72\\%} and \\textbf{63\\%} precision on KITTI, NuScenes and Waymo Open Dataset, respectively). Moreover, under the same point-based representation, P2P-point outperforms the previous motion tracker M$^2$Track by \\textbf{3.3\\%} and \\textbf{6.7\\%} on the KITTI and NuScenes, while running at a considerably high speed of \\textbf{107 Fps} on a single RTX3090 GPU. The source code and pre-trained models are available at \\url{https://github.com/haooozi/P2P}."
                    }
                ]
            },
            "7b70fc32-204e-42e1-8158-fba43b6dcb39": {
                "pk": "7b70fc32-204e-42e1-8158-fba43b6dcb39",
                "name": "Dong-Kyu Chae",
                "collaborators": [
                    "Jaemin Kim",
                    "Yohan Na",
                    "Kangmin Kim",
                    "Sang Rak Lee",
                    "Jiahao Nie",
                    "Fei Xie",
                    "Sifan Zhou",
                    "Xueyi Zhou",
                    "Zhiwei He"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Computer Vision",
                    "Sentiment Analysis",
                    "Object Tracking"
                ],
                "publications": [
                    {
                        "title": "SentiCSE: A Sentiment-aware Contrastive Sentence Embedding Framework with Sentiment-guided Textual Similarity",
                        "abstract": "Recently, sentiment-aware pre-trained language models (PLMs) demonstrate impressive results in downstream sentiment analysis tasks. However, they neglect to evaluate the quality of their constructed sentiment representations; they just focus on improving the fine-tuning performance, which overshadows the representation quality. We argue that without guaranteeing the representation quality, their downstream performance can be highly dependent on the supervision of the fine-tuning data rather than representation quality. This problem would make them difficult to foray into other sentiment-related domains, especially where labeled data is scarce. We first propose Sentiment-guided Textual Similarity (SgTS), a novel metric for evaluating the quality of sentiment representations, which is designed based on the degree of equivalence in sentiment polarity between two sentences. We then propose SentiCSE, a novel Sentiment-aware Contrastive Sentence Embedding framework for constructing sentiment representations via combined word-level and sentence-level objectives, whose quality is guaranteed by SgTS. Qualitative and quantitative comparison with the previous sentiment-aware PLMs shows the superiority of our work. Our code is available at: https://github.com/nayohan/SentiCSE"
                    },
                    {
                        "title": "P2P: Part-to-Part Motion Cues Guide a Strong Tracking Framework for LiDAR Point Clouds",
                        "abstract": "3D single object tracking (SOT) methods based on appearance matching has long suffered from insufficient appearance information incurred by incomplete, textureless and semantically deficient LiDAR point clouds. While motion paradigm exploits motion cues instead of appearance matching for tracking, it incurs complex multi-stage processing and segmentation module. In this paper, we first provide in-depth explorations on motion paradigm, which proves that (\\textbf{i}) it is feasible to directly infer target relative motion from point clouds across consecutive frames; (\\textbf{ii}) fine-grained information comparison between consecutive point clouds facilitates target motion modeling. We thereby propose to perform part-to-part motion modeling for consecutive point clouds and introduce a novel tracking framework, termed \\textbf{P2P}. The novel framework fuses each corresponding part information between consecutive point clouds, effectively exploring detailed information changes and thus modeling accurate target-related motion cues. Following this framework, we present P2P-point and P2P-voxel models, incorporating implicit and explicit part-to-part motion modeling by point- and voxel-based representation, respectively. Without bells and whistles, P2P-voxel sets a new state-of-the-art performance ($\\sim$\\textbf{89\\%}, \\textbf{72\\%} and \\textbf{63\\%} precision on KITTI, NuScenes and Waymo Open Dataset, respectively). Moreover, under the same point-based representation, P2P-point outperforms the previous motion tracker M$^2$Track by \\textbf{3.3\\%} and \\textbf{6.7\\%} on the KITTI and NuScenes, while running at a considerably high speed of \\textbf{107 Fps} on a single RTX3090 GPU. The source code and pre-trained models are available at \\url{https://github.com/haooozi/P2P}."
                    }
                ]
            },
            "b003b8e8-c2ff-4e33-986f-b75fce5146d4": {
                "pk": "b003b8e8-c2ff-4e33-986f-b75fce5146d4",
                "name": "Fei Xie",
                "collaborators": [
                    "Jiahao Nie",
                    "Sifan Zhou",
                    "Xueyi Zhou",
                    "Dong-Kyu Chae",
                    "Zhiwei He"
                ],
                "domain": [
                    "3D Object Tracking",
                    "LiDAR",
                    "Motion Modeling",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "P2P: Part-to-Part Motion Cues Guide a Strong Tracking Framework for LiDAR Point Clouds",
                        "abstract": "3D single object tracking (SOT) methods based on appearance matching has long suffered from insufficient appearance information incurred by incomplete, textureless and semantically deficient LiDAR point clouds. While motion paradigm exploits motion cues instead of appearance matching for tracking, it incurs complex multi-stage processing and segmentation module. In this paper, we first provide in-depth explorations on motion paradigm, which proves that (\\textbf{i}) it is feasible to directly infer target relative motion from point clouds across consecutive frames; (\\textbf{ii}) fine-grained information comparison between consecutive point clouds facilitates target motion modeling. We thereby propose to perform part-to-part motion modeling for consecutive point clouds and introduce a novel tracking framework, termed \\textbf{P2P}. The novel framework fuses each corresponding part information between consecutive point clouds, effectively exploring detailed information changes and thus modeling accurate target-related motion cues. Following this framework, we present P2P-point and P2P-voxel models, incorporating implicit and explicit part-to-part motion modeling by point- and voxel-based representation, respectively. Without bells and whistles, P2P-voxel sets a new state-of-the-art performance ($\\sim$\\textbf{89\\%}, \\textbf{72\\%} and \\textbf{63\\%} precision on KITTI, NuScenes and Waymo Open Dataset, respectively). Moreover, under the same point-based representation, P2P-point outperforms the previous motion tracker M$^2$Track by \\textbf{3.3\\%} and \\textbf{6.7\\%} on the KITTI and NuScenes, while running at a considerably high speed of \\textbf{107 Fps} on a single RTX3090 GPU. The source code and pre-trained models are available at \\url{https://github.com/haooozi/P2P}."
                    }
                ]
            }
        }
    },
    "2407.09522": {
        "paper_data": {
            "title": "UQE: A Query Engine for Unstructured Databases",
            "url": "http://arxiv.org/abs/2407.09522v1",
            "arxiv_id": "2407.09522",
            "authors": [
                "Hanjun Dai",
                "Bethany Yixin Wang",
                "Xingchen Wan",
                "Bo Dai",
                "Sherry Yang",
                "Azade Nova",
                "Pengcheng Yin",
                "Phitchaya Mangpo Phothilimthana",
                "Charles Sutton",
                "Dale Schuurmans"
            ],
            "abstract": "Analytics on structured data is a mature field with many successful methods. However, most real world data exists in unstructured form, such as images and conversations. We investigate the potential of Large Language Models (LLMs) to enable unstructured data analytics. In particular, we propose a new Universal Query Engine (UQE) that directly interrogates and draws insights from unstructured data collections. This engine accepts queries in a Universal Query Language (UQL), a dialect of SQL that provides full natural language flexibility in specifying conditions and operators. The new engine leverages the ability of LLMs to conduct analysis of unstructured data, while also allowing us to exploit advances in sampling and optimization techniques to achieve efficient and accurate query execution. In addition, we borrow techniques from classical compiler theory to better orchestrate the workflow between sampling methods and foundation model calls. We demonstrate the efficiency of UQE on data analytics across different modalities, including images, dialogs and reviews, across a range of useful query types, including conditional aggregation, semantic retrieval and abstraction aggregation.",
            "introduction": "   1 Introduction  Data analysis\u00a0(Elgendy and Elragal, 2014) is essential for making well founded decisions and enabling businesses and society to function more effectively. Relational databases\u00a0(Codd, 1990; Ramakrishnan and Gehrke, 2002) and the Structured Query Language (SQL)\u00a0(Chamberlin, 1998) have delivered huge successes in structured data management and analysis. Typically, such data is collected and organized in a pre-defined schema\u00a0(Elmasri and Navathe, 2013), where the data properties and relationships have been pre-specified, and downstream analysis is restricted to this schema.   In most real-world applications, however, data exists in unstructured formats, such as images, documents and audio recordings. Without preprocessing such data into structured forms, traditional SQL engines can only support limited queries. Preprocessing, including document entity retreival\u00a0(Yu et\u00a0al., 2023) and form understanding\u00a0(Xu et\u00a0al., 2020), also require training on downstream tasks given a predefined taxonomy. This naturally motivates the question we consider in this paper: How can one perform unstructured data analysis in a flexible and efficient way? In the literature, full-text search engines\u00a0(Gospodnetic et\u00a0al., 2010) support scalable regexp-matching search on unstructured data, but this becomes infeasible for more complex semantic reasoning queries. Retrieval-Augmented Generation (RAG)\u00a0(Thorne et\u00a0al., 2020; Lewis et\u00a0al., 2020; Gao et\u00a0al., 2023c) allows question answering on a subset of related data, but is not directly applicable to generic analytical tasks with aggregation and semantic queries that spans over an entire large database. Recent advances in Large Language Models (LLMs)\u00a0(Anthropic, 2024; Achiam et\u00a0al., 2023) unlock the ability to perform flexible question answering, especially with recent long-context models\u00a0(Reid et\u00a0al., 2024). However, setting aside the cost per query, data analytics can still be challenging for LLMs without fine-tuning\u00a0(Li et\u00a0al., 2023) or few-shot demonstrations\u00a0(Chen, 2022), even given structured tables.   Recently, a promising line of work has considered marrying LLMs with programming framworks\u00a0(Sur\u00eds et\u00a0al., 2023), where logical or arithmetic operations are offloaded to program interpreters\u00a0(Gao et\u00a0al., 2023b). A most relevant example for analytics and table understanding tasks is\u00a0Cheng et\u00a0al. (2022), which augments classical SQL semantics with LLMs as user-defined functions (UDF). While promising, the execution of such SQL programs that embed LLM calls still requires sweeping over the entire database, which is too costly for large collections of unstructured content. To overcome this barrier, we leverage the synergy between LLMs and programmatic execution to define an Unstructured Query Language (UQL) that augments SQL for flexible semantic queries, with a focus on improving scalability and efficiency.   A key observation is that the efficiency of classical SQL engines relies on (1) indexing structures that avoid the need to scan the entire database, and (2) a compilation system that determines the best execution order for operations. Based on these ideas, we propose the Unstructured Query Engine (UQE), which refines and extends this design principle to unstructured data analytics. To achieve similar effect to indexing, UQE casts the problem as learning to search or sample, seeking to avoid a full database scan with statistically sound methods. Additionally, a compilation system is developed that determines the best execution order and operator combination for different clauses in a UQL query, with the goal of minimizing LLM calls while preserving query semantics.   As part of this project, we have created four new benchmark datasets, with both text",
            "references": [
                {
                    "title": "Optimizing LLM Queries in Relational Workloads",
                    "abstract": "Analytical database providers (e.g., Redshift, Databricks, BigQuery) have rapidly added support for invoking Large Language Models (LLMs) through native user-defined functions (UDFs) to help users perform natural language tasks, such as classification, entity extraction, and translation, inside analytical workloads. For instance, an analyst might want to extract customer sentiments on millions of product reviews. However, LLM inference is highly expensive in both computational and economic terms: for example, an NVIDIA L4 GPU running Llama2-7B can only process 6 KB of text per second. In this paper, we explore how to optimize LLM inference for analytical workloads that invoke LLMs within relational queries. We show that relational queries present novel opportunities for accelerating LLM inference, including reordering rows to maximize key-value (KV) cache reuse within the LLM inference engine, reordering columns within a row to further increase cache reuse, and deduplicating redundant inference requests. We implement these optimizations in Apache Spark, with vLLM as the model serving backend and achieve up to 4.4x improvement in end-to-end latency on a benchmark of diverse LLM-based queries on real datasets. To the best of our knowledge, this is the first work to explicitly address the problem of optimizing LLM invocations within SQL queries."
                },
                {
                    "title": "Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context",
                    "abstract": "In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content."
                },
                {
                    "title": "Retrieval-Augmented Generation for Large Language Models: A Survey",
                    "abstract": "Large Language Models (LLMs) showcase impressive capabilities but encounter challenges like hallucination, outdated knowledge, and non-transparent, untraceable reasoning processes. Retrieval-Augmented Generation (RAG) has emerged as a promising solution by incorporating knowledge from external databases. This enhances the accuracy and credibility of the generation, particularly for knowledge-intensive tasks, and allows for continuous knowledge updates and integration of domain-specific information. RAG synergistically merges LLMs' intrinsic knowledge with the vast, dynamic repositories of external databases. This comprehensive review paper offers a detailed examination of the progression of RAG paradigms, encompassing the Naive RAG, the Advanced RAG, and the Modular RAG. It meticulously scrutinizes the tripartite foundation of RAG frameworks, which includes the retrieval, the generation and the augmentation techniques. The paper highlights the state-of-the-art technologies embedded in each of these critical components, providing a profound understanding of the advancements in RAG systems. Furthermore, this paper introduces up-to-date evaluation framework and benchmark. At the end, this article delineates the challenges currently faced and points out prospective avenues for research and development."
                },
                {
                    "title": "Table-GPT: Table-tuned GPT for Diverse Table Tasks",
                    "abstract": "Language models, such as GPT-3.5 and ChatGPT, demonstrate remarkable abilities to follow diverse human instructions and perform a wide range of tasks. However, when probing language models using a range of basic table-understanding tasks, we observe that today's language models are still sub-optimal in many table-related tasks, likely because they are pre-trained predominantly on \\emph{one-dimensional} natural-language texts, whereas relational tables are \\emph{two-dimensional} objects. In this work, we propose a new\"\\emph{table-tuning}\"paradigm, where we continue to train/fine-tune language models like GPT-3.5 and ChatGPT, using diverse table-tasks synthesized from real tables as training data, with the goal of enhancing language models' ability to understand tables and perform table tasks. We show that our resulting Table-GPT models demonstrate (1) better \\emph{table-understanding} capabilities, by consistently outperforming the vanilla GPT-3.5 and ChatGPT, on a wide-range of table tasks, including holdout unseen tasks, and (2) strong \\emph{generalizability}, in its ability to respond to diverse human instructions to perform new table-tasks, in a manner similar to GPT-3.5 and ChatGPT."
                },
                {
                    "title": "Text-to-SQL Empowered by Large Language Models: A Benchmark Evaluation",
                    "abstract": "Large language models (LLMs) have emerged as a new paradigm for Text-to-SQL task. However, the absence of a systematical benchmark inhibits the development of designing effective, efficient and economic LLM-based Text-to-SQL solutions. To address this challenge, in this paper, we first conduct a systematical and extensive comparison over existing prompt engineering methods, including question representation, example selection and example organization, and with these experimental results, we elaborate their pros and cons. Based on these findings, we propose a new integrated solution, named DAIL-SQL, which refreshes the Spider leaderboard with 86.6% execution accuracy and sets a new bar.\n To explore the potential of open-source LLM, we investigate them in various scenarios, and further enhance their performance with supervised fine-tuning. Our explorations highlight open-source LLMs' potential in Text-to-SQL, as well as the advantages and disadvantages of the supervised fine-tuning. Additionally, towards an efficient and economic LLM-based Text-to-SQL solution, we emphasize the token efficiency in prompt engineering and compare the prior studies under this metric. We hope that our work provides a deeper understanding of Text-to-SQL with LLMs, and inspires further investigations and broad applications."
                },
                {
                    "title": "DocumentNet: Bridging the Data Gap in Document Pre-training",
                    "abstract": "Document understanding tasks, in particular, Visually-rich Document Entity Retrieval (VDER), have gained significant attention in recent years thanks to their broad applications in enterprise AI. However, publicly available data have been scarce for these tasks due to strict privacy constraints and high annotation costs. To make things worse, the non-overlapping entity spaces from different datasets hinder the knowledge transfer between document types. In this paper, we propose a method to collect massive-scale and weakly labeled data from the web to benefit the training of VDER models. The collected dataset, named DocumentNet, does not depend on specific document types or entity sets, making it universally applicable to all VDER tasks. The current DocumentNet consists of 30M documents spanning nearly 400 document types organized in a four-level ontology. Experiments on a set of broadly adopted VDER tasks show significant improvements when DocumentNet is incorporated into the pre-training for both classic and few-shot learning settings. With the recent emergence of large language models (LLMs), DocumentNet provides a large data source to extend their multi-modal capabilities for VDER."
                },
                {
                    "title": "DIN-SQL: Decomposed In-Context Learning of Text-to-SQL with Self-Correction",
                    "abstract": "We study the problem of decomposing a complex text-to-sql task into smaller sub-tasks and how such a decomposition can significantly improve the performance of Large Language Models (LLMs) in the reasoning process. There is currently a significant gap between the performance of fine-tuned models and prompting approaches using LLMs on challenging text-to-sql datasets such as Spider. We show that SQL queries, despite their declarative structure, can be broken down into sub-problems and the solutions of those sub-problems can be fed into LLMs to significantly improve their performance. Our experiments with three LLMs show that this approach consistently improves their performance by roughly 10%, pushing the accuracy of LLMs towards state-of-the-art, and even beating large fine-tuned models on the holdout Spider dataset."
                },
                {
                    "title": "Querying Large Language Models with SQL",
                    "abstract": "In many use-cases, information is stored in text but not available in structured data. However, extracting data from natural language text to precisely fit a schema, and thus enable querying, is a challenging task. With the rise of pre-trained Large Language Models (LLMs), there is now an effective solution to store and use information extracted from massive corpora of text documents. Thus, we envision the use of SQL queries to cover a broad range of data that is not captured by traditional databases by tapping the information in LLMs. To ground this vision, we present Galois, a prototype based on a traditional database architecture, but with new physical operators for querying the underlying LLM. The main idea is to execute some operators of the the query plan with prompts that retrieve data from the LLM. For a large class of SQL queries, querying LLMs returns well structured relations, with encouraging qualitative results. Preliminary experimental results make pre-trained LLMs a promising addition to the field of database systems, introducing a new direction for hybrid query processing. However, we pinpoint several research challenges that must be addressed to build a DBMS that exploits LLMs. While some of these challenges necessitate integrating concepts from the NLP literature, others offer novel research avenues for the DB community."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "ViperGPT: Visual Inference via Python Execution for Reasoning",
                    "abstract": "Answering visual queries is a complex task that requires both visual processing and reasoning. End-to-end models, the dominant approach for this task, do not explicitly differentiate between the two, limiting interpretability and generalization. Learning modular programs presents a promising alternative, but has proven challenging due to the difficulty of learning both the programs and modules simultaneously. We introduce ${\\color{green}{\\text{ViperGPT}}}$, a framework that leverages code-generation models to compose vision-and-language models into subroutines to produce a result for any query. ${\\color{green}{\\text{ViperGPT}}}$ utilizes a provided API to access the available modules, and composes them by generating Python code that is later executed. This simple approach requires no further training, and achieves state-of-the-art results across various complex visual tasks."
                },
                {
                    "title": "Prompting Is Programming: A Query Language for Large Language Models",
                    "abstract": "Large language models have demonstrated outstanding performance on a wide range of tasks such as question answering and code generation. On a high level, given an input, a language model can be used to automatically complete the sequence in a statistically-likely way. Based on this, users prompt these models with language instructions or examples, to implement a variety of downstream tasks. Advanced prompting methods can even imply interaction between the language model, a user, and external tools such as calculators. However, to obtain state-of-the-art performance or adapt language models for specific tasks, complex task- and model-specific programs have to be implemented, which may still require ad-hoc interaction. Based on this, we present the novel idea of Language Model Programming (LMP). LMP generalizes language model prompting from pure text prompts to an intuitive combination of text prompting and scripting. Additionally, LMP allows constraints to be specified over the language model output. This enables easy adaption to many tasks while abstracting language model internals and providing high-level semantics. To enable LMP, we implement LMQL (short for Language Model Query Language), which leverages the constraints and control flow from an LMP prompt to generate an efficient inference procedure that minimizes the number of expensive calls to the underlying language model. We show that LMQL can capture a wide range of state-of-the-art prompting methods in an intuitive way, especially facilitating interactive flows that are challenging to implement with existing high-level APIs. Our evaluation shows that we retain or increase the accuracy on several downstream tasks, while also significantly reducing the required amount of computation or cost in the case of pay-to-use APIs (26-85% cost savings)."
                },
                {
                    "title": "PAL: Program-aided Language Models",
                    "abstract": "Large language models (LLMs) have recently demonstrated an impressive ability to perform arithmetic and symbolic reasoning tasks, when provided with a few examples at test time (\"few-shot prompting\"). Much of this success can be attributed to prompting methods such as\"chain-of-thought'', which employ LLMs for both understanding the problem description by decomposing it into steps, as well as solving each step of the problem. While LLMs seem to be adept at this sort of step-by-step decomposition, LLMs often make logical and arithmetic mistakes in the solution part, even when the problem is decomposed correctly. In this paper, we present Program-Aided Language models (PAL): a novel approach that uses the LLM to read natural language problems and generate programs as the intermediate reasoning steps, but offloads the solution step to a runtime such as a Python interpreter. With PAL, decomposing the natural language problem into runnable steps remains the only learning task for the LLM, while solving is delegated to the interpreter. We demonstrate this synergy between a neural LLM and a symbolic interpreter across 13 mathematical, symbolic, and algorithmic reasoning tasks from BIG-Bench Hard and other benchmarks. In all these natural language reasoning tasks, generating code using an LLM and reasoning using a Python interpreter leads to more accurate results than much larger models. For example, PAL using Codex achieves state-of-the-art few-shot accuracy on the GSM8K benchmark of math word problems, surpassing PaLM-540B which uses chain-of-thought by absolute 15% top-1. Our code and data are publicly available at http://reasonwithpal.com/ ."
                },
                {
                    "title": "Large Language Models are few(1)-shot Table Reasoners",
                    "abstract": "Recent literature has shown that large language models (LLMs) are generally excellent few-shot reasoners to solve text reasoning tasks. However, the capability of LLMs on table reasoning tasks is yet to be explored. In this paper, we aim at understanding how well LLMs can perform table-related tasks with few-shot in-context learning. Specifically, we evaluated LLMs on popular table QA and fact verification datasets like WikiTableQuestion, FetaQA, TabFact, and FEVEROUS and found that LLMs are competent at complex reasoning over table structures, though these models are not pre-trained on any table corpus. When combined with \u2018chain of thoughts\u2019 prompting, LLMs can achieve very strong performance with only a 1-shot demonstration, even on par with some SoTA models. We show that LLMs are even more competent at generating comprehensive long-form answers on FetaQA than tuned T5-large. We further manually studied the reasoning chains elicited from LLMs and found that these reasoning chains are highly consistent with the underlying semantic form. We believe that LLMs can serve as a simple yet generic baseline for future research. The code and data are released in {url{https://github.com/wenhuchen/TableCoT}."
                },
                {
                    "title": "Binding Language Models in Symbolic Languages",
                    "abstract": "Though end-to-end neural approaches have recently been dominating NLP tasks in both performance and ease-of-use, they lack interpretability and robustness. We propose Binder, a training-free neural-symbolic framework that maps the task input to a program, which (1) allows binding a unified API of language model (LM) functionalities to a programming language (e.g., SQL, Python) to extend its grammar coverage and thus tackle more diverse questions, (2) adopts an LM as both the program parser and the underlying model called by the API during execution, and (3) requires only a few in-context exemplar annotations. Specifically, we employ GPT-3 Codex as the LM. In the parsing stage, with only a few in-context exemplars, Codex is able to identify the part of the task input that cannot be answerable by the original programming language, correctly generate API calls to prompt Codex to solve the unanswerable part, and identify where to place the API calls while being compatible with the original grammar. In the execution stage, Codex can perform versatile functionalities (e.g., commonsense QA, information extraction) given proper prompts in the API calls. Binder achieves state-of-the-art results on WikiTableQuestions and TabFact datasets, with explicit output programs that benefit human debugging. Note that previous best systems are all finetuned on tens of thousands of task-specific samples, while Binder only uses dozens of annotations as in-context exemplars without any training. Our code is available at https://github.com/HKUNLP/Binder ."
                },
                {
                    "title": "PICARD: Parsing Incrementally for Constrained Auto-Regressive Decoding from Language Models",
                    "abstract": "Large pre-trained language models for textual data have an unconstrained output space; at each decoding step, they can produce any of 10,000s of sub-word tokens. When fine-tuned to target constrained formal languages like SQL, these models often generate invalid code, rendering it unusable. We propose PICARD (code available at https://github.com/ElementAI/picard), a method for constraining auto-regressive decoders of language models through incremental parsing. PICARD helps to find valid output sequences by rejecting inadmissible tokens at each decoding step. On the challenging Spider and CoSQL text-to-SQL translation tasks, we show that PICARD transforms fine-tuned T5 models with passable performance into state-of-the-art solutions."
                },
                {
                    "title": "Action-Based Conversations Dataset: A Corpus for Building More In-Depth Task-Oriented Dialogue Systems",
                    "abstract": "Existing goal-oriented dialogue datasets focus mainly on identifying slots and values. However, customer support interactions in reality often involve agents following multi-step procedures derived from explicitly-defined company policies as well. To study customer service dialogue systems in more realistic settings, we introduce the Action-Based Conversations Dataset (ABCD), a fully-labeled dataset with over 10K human-to-human dialogues containing 55 distinct user intents requiring unique sequences of actions constrained by policies to achieve task success. We propose two additional dialog tasks, Action State Tracking and Cascading Dialogue Success, and establish a series of baselines involving large-scale, pre-trained language models on this dataset. Empirical results demonstrate that while more sophisticated networks outperform simpler models, a considerable gap (50.8% absolute accuracy) still exists to reach human-level performance on ABCD."
                },
                {
                    "title": "Neural Databases",
                    "abstract": "In recent years, neural networks have shown impressive performance gains on long-standing AI problems, and in particular, answering queries from natural language text. These advances raise the question of whether they can be extended to a point where we can relax the fundamental assumption of database management, namely, that our data is represented as fields of a pre-defined schema. This paper presents a first step in answering that question. We describe NeuralDB, a database system with no pre-defined schema, in which updates and queries are given in natural language. We develop query processing techniques that build on the primitives offered by the state of the art Natural Language Processing methods. We begin by demonstrating that at the core, recent NLP transformers, powered by pre-trained language models, can answer select-project-join queries if they are given the exact set of relevant facts. However, they cannot scale to non-trivial databases and cannot perform aggregation queries. Based on these findings, we describe a NeuralDB architecture that runs multiple Neural SPJ operators in parallel, each with a set of database sentences that can produce one of the answers to the query. The result of these operators is fed to an aggregation operator if needed. We describe an algorithm that learns how to create the appropriate sets of facts to be fed into each of the Neural SPJ operators. Importantly, this algorithm can be trained by the Neural SPJ operator itself. We experimentally validate the accuracy of NeuralDB and its components, showing that we can answer queries over thousands of sentences with very high accuracy."
                },
                {
                    "title": "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks",
                    "abstract": "Large pre-trained language models have been shown to store factual knowledge in their parameters, and achieve state-of-the-art results when fine-tuned on downstream NLP tasks. However, their ability to access and precisely manipulate knowledge is still limited, and hence on knowledge-intensive tasks, their performance lags behind task-specific architectures. Additionally, providing provenance for their decisions and updating their world knowledge remain open research problems. Pre-trained models with a differentiable access mechanism to explicit non-parametric memory can overcome this issue, but have so far been only investigated for extractive downstream tasks. We explore a general-purpose fine-tuning recipe for retrieval-augmented generation (RAG) -- models which combine pre-trained parametric and non-parametric memory for language generation. We introduce RAG models where the parametric memory is a pre-trained seq2seq model and the non-parametric memory is a dense vector index of Wikipedia, accessed with a pre-trained neural retriever. We compare two RAG formulations, one which conditions on the same retrieved passages across the whole generated sequence, the other can use different passages per token. We fine-tune and evaluate our models on a wide range of knowledge-intensive NLP tasks and set the state-of-the-art on three open domain QA tasks, outperforming parametric seq2seq models and task-specific retrieve-and-extract architectures. For language generation tasks, we find that RAG models generate more specific, diverse and factual language than a state-of-the-art parametric-only seq2seq baseline."
                },
                {
                    "title": "TaBERT: Pretraining for Joint Understanding of Textual and Tabular Data",
                    "abstract": "Recent years have witnessed the burgeoning of pretrained language models (LMs) for text-based natural language (NL) understanding tasks. Such models are typically trained on free-form NL text, hence may not be suitable for tasks like semantic parsing over structured data, which require reasoning over both free-form NL questions and structured tabular data (e.g., database tables). In this paper we present TaBERT, a pretrained LM that jointly learns representations for NL sentences and (semi-)structured tables. TaBERT is trained on a large corpus of 26 million tables and their English contexts. In experiments, neural semantic parsers using TaBERT as feature representation layers achieve new best results on the challenging weakly-supervised semantic parsing benchmark WikiTableQuestions, while performing competitively on the text-to-SQL dataset Spider."
                },
                {
                    "title": "Neural Symbolic Reader: Scalable Integration of Distributed and Symbolic Representations for Reading Comprehension",
                    "abstract": "Integrating distributed representations with symbolic operations is essential for reading comprehension requiring complex reasoning, such as counting, sorting and arithmetics, but most existing approaches are hard to scale to more domains or more complex reasoning. In this work, we propose the Neural Symbolic Reader (NeRd), which includes a reader, e.g., BERT, to encode the passage and question, and a programmer, e.g., LSTM, to generate a program that is executed to produce the answer. Compared to previous works, NeRd is more scalable in two aspects: (1) domain-agnostic, i.e., the same neural architecture works for different domains; (2) compositional, i.e., when needed, complex programs can be generated by recursively applying the predefined operators, which become executable and interpretable representations for more complex reasoning. Furthermore, to overcome the challenge of training NeRd with weak supervision, we apply data augmentation techniques and hard Expectation-Maximization (EM) with thresholding. On DROP, a challenging reading comprehension dataset that requires discrete reasoning, NeRd achieves 2.5%/1.8% absolute improvement over the state-of-the-art on EM/F1 metrics. With the same architecture, NeRd significantly outperforms the baselines on MathQA, a math problem benchmark that requires multiple steps of reasoning, by 25.5% absolute increment on accuracy when trained on all the annotated programs. More importantly, NeRd still beats the baselines even when only 20% of the program annotations are given."
                },
                {
                    "title": "LayoutLM: Pre-training of Text and Layout for Document Image Understanding",
                    "abstract": "Pre-training techniques have been verified successfully in a variety of NLP tasks in recent years. Despite the widespread use of pre-training models for NLP applications, they almost exclusively focus on text-level manipulation, while neglecting layout and style information that is vital for document image understanding. In this paper, we propose the LayoutLM to jointly model interactions between text and layout information across scanned document images, which is beneficial for a great number of real-world document image understanding tasks such as information extraction from scanned documents. Furthermore, we also leverage image features to incorporate words' visual information into LayoutLM. To the best of our knowledge, this is the first time that text and layout are jointly learned in a single framework for document-level pre-training. It achieves new state-of-the-art results in several downstream tasks, including form understanding (from 70.72 to 79.27), receipt understanding (from 94.02 to 95.24) and document image classification (from 93.07 to 94.42). The code and pre-trained LayoutLM models are publicly available at https://aka.ms/layoutlm."
                },
                {
                    "title": "Research on K-Value Selection Method of K-Means Clustering Algorithm",
                    "abstract": "Among many clustering algorithms, the K-means clustering algorithm is widely used because of its simple algorithm and fast convergence. However, the K-value of clustering needs to be given in advance and the choice of K-value directly affect the convergence result. To solve this problem, we mainly analyze four K-value selection algorithms, namely Elbow Method, Gap Statistic, Silhouette Coefficient, and Canopy; give the pseudo code of the algorithm; and use the standard data set Iris for experimental verification. Finally, the verification results are evaluated, the advantages and disadvantages of the above four algorithms in a K-value selection are given, and the clustering range of the data set is pointed out."
                },
                {
                    "title": "Towards Complex Text-to-SQL in Cross-Domain Database with Intermediate Representation",
                    "abstract": "We present a neural approach called IRNet for complex and cross-domain Text-to-SQL. IRNet aims to address two challenges: 1) the mismatch between intents expressed in natural language (NL) and the implementation details in SQL; 2) the challenge in predicting columns caused by the large number of out-of-domain words. Instead of end-to-end synthesizing a SQL query, IRNet decomposes the synthesis process into three phases. In the first phase, IRNet performs a schema linking over a question and a database schema. Then, IRNet adopts a grammar-based neural model to synthesize a SemQL query which is an intermediate representation that we design to bridge NL and SQL. Finally, IRNet deterministically infers a SQL query from the synthesized SemQL query with domain knowledge. On the challenging Text-to-SQL benchmark Spider, IRNet achieves 46.7% accuracy, obtaining 19.5% absolute improvement over previous state-of-the-art approaches. At the time of writing, IRNet achieves the first position on the Spider leaderboard."
                },
                {
                    "title": "SyntaxSQLNet: Syntax Tree Networks for Complex and Cross-Domain Text-to-SQL Task",
                    "abstract": "Most existing studies in text-to-SQL tasks do not require generating complex SQL queries with multiple clauses or sub-queries, and generalizing to new, unseen databases. In this paper we propose SyntaxSQLNet, a syntax tree network to address the complex and cross-domain text-to-SQL generation task. SyntaxSQLNet employs a SQL specific syntax tree-based decoder with SQL generation path history and table-aware column attention encoders. We evaluate SyntaxSQLNet on a new large-scale text-to-SQL corpus containing databases with multiple tables and complex SQL queries containing multiple SQL clauses and nested queries. We use a database split setting where databases in the test set are unseen during training. Experimental results show that SyntaxSQLNet can handle a significantly greater number of complex SQL examples than prior work, outperforming the previous state-of-the-art model by 9.5% in exact matching accuracy. To our knowledge, we are the first to study this complex text-to-SQL task. Our task and models with the latest updates are available at https://yale-lily.github.io/seq2sql/spider."
                },
                {
                    "title": "Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task",
                    "abstract": "We present Spider, a large-scale complex and cross-domain semantic parsing and text-to-SQL dataset annotated by 11 college students. It consists of 10,181 questions and 5,693 unique complex SQL queries on 200 databases with multiple tables covering 138 different domains. We define a new complex and cross-domain semantic parsing and text-to-SQL task so that different complicated SQL queries and databases appear in train and test sets. In this way, the task requires the model to generalize well to both new SQL queries and new database schemas. Therefore, Spider is distinct from most of the previous semantic parsing tasks because they all use a single database and have the exact same program in the train set and the test set. We experiment with various state-of-the-art models and the best model achieves only 9.7% exact matching accuracy on a database split setting. This shows that Spider presents a strong challenge for future research. Our dataset and task with the most recent updates are publicly available at https://yale-lily.github.io/seq2sql/spider."
                },
                {
                    "title": "CLEVR: A Diagnostic Dataset for Compositional Language and Elementary Visual Reasoning",
                    "abstract": "When building artificial intelligence systems that can reason and answer questions about visual data, we need diagnostic tests to analyze our progress and discover short-comings. Existing benchmarks for visual question answering can help, but have strong biases that models can exploit to correctly answer questions without reasoning. They also conflate multiple sources of error, making it hard to pinpoint model weaknesses. We present a diagnostic dataset that tests a range of visual reasoning abilities. It contains minimal biases and has detailed annotations describing the kind of reasoning each question requires. We use this dataset to analyze a variety of modern visual reasoning systems, providing novel insights into their abilities and limitations."
                },
                {
                    "title": "Neural Enquirer: Learning to Query Tables",
                    "abstract": "We proposed Neural Enquirer as a neural network architecture to execute a natural language (NL) query on a knowledge-base (KB) for answers. Basically, Neural Enquirer finds the distributed representation of a query and then executes it on knowledge-base tables to obtain the answer as one of the values in the tables. Unlike similar efforts in end-to-end training of semantic parsers, Neural Enquirer is fully \"neuralized\": it not only gives distributional representation of the query and the knowledge-base, but also realizes the execution of compositional queries as a series of differentiable operations, with intermediate results (consisting of annotations of the tables at different levels) saved on multiple layers of memory. Neural Enquirer can be trained with gradient descent, with which not only the parameters of the controlling components and semantic parsing component, but also the embeddings of the tables and query words can be learned from scratch. The training can be done in an end-to-end fashion, but it can take stronger guidance, e.g., the step-by-step supervision for complicated queries, and benefit from it. Neural Enquirer is one step towards building neural network systems which seek to understand language by executing it on real-world. Our experiments show that Neural Enquirer can learn to execute fairly complicated NL queries on tables with rich structures."
                },
                {
                    "title": "Neural Programmer: Inducing Latent Programs with Gradient Descent",
                    "abstract": "Deep neural networks have achieved impressive supervised classification performance in many tasks including image recognition, speech recognition, and sequence to sequence learning. However, this success has not been translated to applications like question answering that may involve complex arithmetic and logic reasoning. A major limitation of these models is in their inability to learn even simple arithmetic and logic operations. For example, it has been shown that neural networks fail to learn to add two binary numbers reliably. In this work, we propose Neural Programmer, an end-to-end differentiable neural network augmented with a small set of basic arithmetic and logic operations. Neural Programmer can call these augmented operations over several steps, thereby inducing compositional programs that are more complex than the built-in operations. The model learns from a weak supervision signal which is the result of execution of the correct program, hence it does not require expensive annotation of the correct program itself. The decisions of what operations to call, and what data segments to apply to are inferred by Neural Programmer. Such decisions, during training, are done in a differentiable fashion so that the entire network can be trained jointly by gradient descent. We find that training the model is difficult, but it can be greatly improved by adding random noise to the gradient. On a fairly complex synthetic table-comprehension dataset, traditional recurrent networks and attentional models perform poorly while Neural Programmer typically obtains nearly perfect accuracy."
                },
                {
                    "title": "Neural Module Networks",
                    "abstract": "Visual question answering is fundamentally compositional in nature-a question like where is the dog? shares substructure with questions like what color is the dog? and where is the cat? This paper seeks to simultaneously exploit the representational capacity of deep networks and the compositional linguistic structure of questions. We describe a procedure for constructing and learning neural module networks, which compose collections of jointly-trained neural \"modules\" into deep networks for question answering. Our approach decomposes questions into their linguistic substructures, and uses these structures to dynamically instantiate modular networks (with reusable components for recognizing dogs, classifying colors, etc.). The resulting compound networks are jointly trained. We evaluate our approach on two challenging datasets for visual question answering, achieving state-of-the-art results on both the VQA natural image dataset and a new dataset of complex questions about abstract shapes."
                },
                {
                    "title": "Database systems:models, languages dwesign, and application programming",
                    "abstract": "Clear explanations of theory and design, broad coverage of models and real systems, and an up-to-date introduction to modern database technologies result in a leading introduction to database systems. Intended for computer science majors, Fundamentals of Database Systems, 6/e emphasizes math models, design issues, relational algebra, and relational calculus. A lab manual and problems give students opportunities to practice the fundamentals of design and implementation. Real-world examples serve as engaging, practical illustrations of database concepts. the Sixth Edition maintains its coverage of the most popular database topics, including SQL, security, and data mining, and features increased emphasis on XML and semi-structured data."
                },
                {
                    "title": "Learning Word Vectors for Sentiment Analysis",
                    "abstract": "Unsupervised vector-based approaches to semantics can model rich lexical meanings, but they largely fail to capture sentiment information that is central to many word meanings and important for a wide range of NLP tasks. We present a model that uses a mix of unsupervised and supervised techniques to learn word vectors capturing semantic term--document information as well as rich sentiment content. The proposed model can leverage both continuous and multi-dimensional sentiment information as well as non-sentiment annotations. We instantiate the model to utilize the document-level sentiment polarity annotations present in many online documents (e.g. star ratings). We evaluate the model using small, widely used sentiment and subjectivity corpora and find it out-performs several previously introduced methods for sentiment classification. We also introduce a large dataset of movie reviews to serve as a more robust benchmark for work in this area."
                },
                {
                    "title": "Database Management Systems",
                    "abstract": "Database Management Systems Introduction Data Models Relational Model Entity-Relationship (E/R) Model Key-Value Model Entity-Relationship (E/R) Model Relationships Subclasses Constraints Converting ER to Relational Model Functional Dependencies (FD) FD Properties Closure Algorithm Superkeys & Minimal Basis Schema Normalization Decomposition Lossless-Join & Chase Algorithm Dependency Preserving Boyce-Codd Normal Form (BCNF) Third Normal Form (3NF) Relational Algebra Basic Operators Derived Operators Extended RA Syntax Data Storage Buffer Manager File Organization Page Organization Record Format Index Structures Indexing Basics Hash Tables B+ Trees Bitmaps & Bitslices Supporting SQL Operators External Sorting Selection Operator Projection Operator Join Operator Aggregation Operation Query Optimizer Annotated RA Trees Optimization Plans Cost Estimation Transaction Management Definition of Transactions The \"ACID\" Principle Write-Ahead Logging (WAL) Transaction Concurrency"
                },
                {
                    "title": "Fast and robust Earth Mover's Distances",
                    "abstract": "We present a new algorithm for a robust family of Earth Mover's Distances - EMDs with thresholded ground distances. The algorithm transforms the flow-network of the EMD so that the number of edges is reduced by an order of magnitude. As a result, we compute the EMD by an order of magnitude faster than the original algorithm, which makes it possible to compute the EMD on large histograms and databases. In addition, we show that EMDs with thresholded ground distances have many desirable properties. First, they correspond to the way humans perceive distances. Second, they are robust to outlier noise and quantization effects. Third, they are metrics. Finally, experimental results on image retrieval show that thresholding the ground distance of the EMD improves both accuracy and speed."
                },
                {
                    "title": "Lucene in Action",
                    "abstract": "Lucene is a rare gem in the Apache Jakarta world?a highly scalable, fast, and pure Java search engine. Its initial attractions are its performance, its simplicity, and its disarming ease-of-use. But there are many caveats, best practices, and examples that users are looking for that are not yet well documented. The lucene-user e-mail list is very active and helpful, but many users seek more guidance and examples. Lucene in Action describes what Lucene is and how it works and most importantly how it can be used in a variety of real-world use cases, such at Nutch. Nutch?an open-source project designed to index the internet very much like Google?is built upon Lucene. Lucene in Action provides readers with best practices, tried and true from the field, including: * Understanding and solving ?analysis paralysis? * Advanced searching techniques, including filtering and custom query parsing techniques. * Handling document types such as Word, PDF, HTML, XML, and others."
                },
                {
                    "title": "A Complete Guide to DB2 Universal Database",
                    "abstract": "1. Introduction 2. Basics 3. Interactive SQL 4. Static SQL 5. Query Power 6. Datatypes and Functions 7. Active Data 8. Dynamic SQL 9. Stored Procedures 10. Database Administration Appendix A: Special Registers Appendix B: Functions Appendix C: Typecodes Appendix D: System Catalog Tables Appendix E: Syntax for Host Variable Declarations in C and C++ Appendix F: IBM Publications"
                },
                {
                    "title": "The Relational Model for Database Management, Version 2",
                    "abstract": "From the Preface (See Front Matter for full Preface) \n \nAn important adjunct to precision is a sound theoretical foundation. The relational model is solidly based on two parts of mathematics: firstorder predicate logic and the theory of relations. This book, however, does not dwell on the theoretical foundations, but rather on all the features of the relational model that I now perceive as important for database users, and therefore for DBMS vendors. My perceptions result from 20 years of practical experience in computing and data processing (chiefly, but not exclusively, with large-scale customers of IBM), followed by another 20 years of research. \n \nI believe that this is the first book to deal exclusively with the relational approach. It does, however, include design principles in Chapters 21 and 22. It is also the first book on the relational model by the originator of that model. All the ideas in the relational model described in this book are mine, except in cases where I explicitly credit someone else. \n \nIn developing the relational model, I have tried to follow Einstein's advice, \"Make it as simple as possible, but no simpler.\" I believe that in the last clause he was discouraging the pursuit of simplicity to the extent of distorting reality. So why does the book contain 30 chapters and two appendixes? To answer this question, it is necessary to look at the history of research and development of the relational model."
                },
                {
                    "title": "SQL-PaLM: Improved Large Language Model Adaptation for Text-to-SQL",
                    "abstract": "One impressive emergent capability of large language models (LLMs) is generation of code, including Structured Query Language (SQL) for databases. For the task of converting natural language text to SQL queries, Text-to-SQL, adaptation of LLMs is of paramount importance, both in in-context learning and fine-tuning settings, depending on the amount of adaptation data used. In this paper, we propose an LLM-based Text-to-SQL model SQL-PaLM , leveraging on PaLM-2 , that pushes the state-of-the-art in both settings. Few-shot SQL-PaLM is based on an execution-based self-consistency prompting approach designed for Text-to-SQL, and achieves 77.3% in test-suite accuracy on Spider, which to our best knowledge is the first to outperform previous state-of-the-art with fine-tuning by a significant margin, 4% . Furthermore, we demonstrate that the fine-tuned SQL-PALM outperforms it further by another 1% . Towards applying SQL-PaLM to real-world scenarios we further evaluate its robustness on other challenging variants of Spider and demonstrate the superior generalization capability of SQL-PaLM . In addition, via extensive case studies, we demonstrate the impressive intelligent capabilities and various success enablers of LLM-based Text-to-SQL."
                },
                {
                    "title": "AirDialogue: An Environment for Goal-Oriented Dialogue Research",
                    "abstract": "Recent progress in dialogue generation has inspired a number of studies on dialogue systems that are capable of accomplishing tasks through natural language interactions. A promising direction among these studies is the use of reinforcement learning techniques, such as self-play, for training dialogue agents. However, current datasets are limited in size, and the environment for training agents and evaluating progress is relatively unsophisticated. We present AirDialogue, a large dataset that contains 301,427 goal-oriented conversations. To collect this dataset, we create a context-generator which provides travel and flight restrictions. We then ask human annotators to play the role of a customer or an agent and interact with the goal of successfully booking a trip given the restrictions. Key to our environment is the ease of evaluating the success of the dialogue, which is achieved by using ground-truth states (e.g., the flight being booked) generated by the restrictions. Any dialogue agent that does not generate the correct states is considered to fail. Our experimental results indicate that state-of-the-art dialogue models can only achieve a score of 0.17 while humans can reach a score of 0.91, which suggests significant opportunities for future improvement."
                },
                {
                    "title": "Taking the Human Out of the Loop: A Review of Bayesian Optimization",
                    "abstract": "Big Data applications are typically associated with systems involving large numbers of users, massive complex software systems, and large-scale heterogeneous computing and storage architectures. The construction of such systems involves many distributed design choices. The end products (e.g., recommendation systems, medical analysis tools, real-time game engines, speech recognizers) thus involve many tunable configuration parameters. These parameters are often specified and hard-coded into the software by various developers or teams. If optimized jointly, these parameters can result in significant improvements. Bayesian optimization is a powerful tool for the joint optimization of design choices that is gaining great popularity in recent years. It promises greater automation so as to increase both product quality and human productivity. This review paper introduces Bayesian optimization, highlights some of its methodological aspects, and showcases a wide range of applications."
                },
                {
                    "title": "A Handbook for Building an Approximate Query Engine",
                    "abstract": "There has been much research on various aspects of Approximate Query Processing (AQP), such as different sampling strategies, error estimation mechanisms, and various types of data synopses. However, many subtle challenges arise when building an actual AQP engine that can be deployed and used by real world applications. These subtleties are often ignored (or at least not elaborated) by the theoretical literature and academic prototypes alike. For the first time to the best of our knowledge, in this article, we focus on these subtle challenges that one must address when designing an AQP system. Our intention for this article is to serve as a handbook listing critical design choices that database practitioners must be aware of when building or using an AQP system, not to prescribe a specific solution to each challenge."
                },
                {
                    "title": "The Claude 3 Model Family: Opus, Sonnet, Haiku",
                    "abstract": "We introduce Claude 3, a new family of large multimodal models \u2013 Claude 3 Opus , our most capable offering, Claude 3 Sonnet , which provides a combination of skills and speed, and Claude 3 Haiku , our fastest and least expensive model. All new models have vision capabilities that enable them to process and analyze image data. The Claude 3 family demonstrates strong performance across benchmark evaluations and sets a new standard on measures of reasoning, math, and coding. Claude 3 Opus achieves state-of-the-art results on evaluations like GPQA [1], MMLU [2], MMMU [3] and many more. Claude 3 Haiku performs as well or better than Claude 2 [4] on most pure-text tasks, while Sonnet and Opus significantly outperform it. Additionally, these models exhibit improved fluency in non-English languages, making them more versatile for a global audience. In this report, we provide an in-depth analysis of our evaluations, focusing on core capabilities, safety, societal impacts, and the catastrophic risk assessments we committed to in our Responsible Scaling Policy [5]."
                }
            ],
            "categories": [
                "cs.DB",
                "cs.AI",
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can one perform unstructured data analysis in a flexible and efficient way?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses the growing need for effective analysis of unstructured data, which is prevalent in various domains such as healthcare, finance, and social media. By developing a method that allows for flexible and efficient querying of unstructured data, this research could pave the way for new analytical tools and frameworks that enhance decision-making processes. Furthermore, it could lead to advancements in natural language processing and machine learning, enabling more sophisticated applications in data analytics, thereby influencing future research directions and practical implementations.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the inherent complexity of unstructured data, which lacks a predefined schema, making it difficult to apply traditional SQL-based approaches. Naive methods may fail because they often rely on full-text search or simplistic retrieval techniques that do not support complex semantic reasoning or aggregation queries. Additionally, the computational cost associated with querying large datasets using Large Language Models (LLMs) without fine-tuning or few-shot demonstrations presents a significant obstacle. The need for efficient indexing and execution order determination further complicates the development of a scalable solution.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on either full-text search engines or LLMs, but these approaches do not adequately address the need for flexible and efficient unstructured data analysis. Limitations in existing solutions include the inability to perform complex queries over large datasets without incurring high computational costs. Barriers such as the lack of a unified framework that combines the strengths of SQL and LLMs have prevented effective solutions from emerging. This research proposes a novel Unstructured Query Language (UQL) and an Unstructured Query Engine (UQE) that improve upon prior work by integrating indexing and compilation strategies tailored for unstructured data.\n\n### [Question 5] - What are the key components of my approach and results?\nThe proposed methodology involves developing an Unstructured Query Language (UQL) that augments traditional SQL to facilitate flexible semantic queries on unstructured data. The Unstructured Query Engine (UQE) will be designed to learn to search or sample, minimizing the need for full database scans. The approach will utilize four newly created benchmark datasets to evaluate performance. Key metrics for success will include query execution time, accuracy of results, and the number of L"
            }
        },
        "author_data": {
            "f59639f0-33ac-4e1b-ad31-5ff204445180": {
                "pk": "f59639f0-33ac-4e1b-ad31-5ff204445180",
                "name": "Hanjun Dai",
                "collaborators": [
                    "Bo Dai",
                    "Le Song",
                    "Dale Schuurmans",
                    "Haoran Sun",
                    "Yichen Wang",
                    "Rakshit Trivedi",
                    "Charles Sutton",
                    "Niao He",
                    "Mengjiao Yang",
                    "Yuyu Zhang"
                ],
                "domain": [
                    "Bayesian Inference",
                    "Graph Neural Network",
                    "Energy-Based Models",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Provable Bayesian Inference via Particle Mirror Descent",
                        "abstract": "Bayesian methods are appealing in their flexibility in modeling complex data and ability in capturing uncertainty in parameters. However, when Bayes' rule does not result in tractable closed-form, most approximate inference algorithms lack either scalability or rigorous guarantees. To tackle this challenge, we propose a simple yet provable algorithm, \\emph{Particle Mirror Descent} (PMD), to iteratively approximate the posterior density. PMD is inspired by stochastic functional mirror descent where one descends in the density space using a small batch of data points at each iteration, and by particle filtering where one uses samples to approximate a function. We prove result of the first kind that, with $m$ particles, PMD provides a posterior density estimator that converges in terms of $KL$-divergence to the true posterior in rate $O(1/\\sqrt{m})$. We demonstrate competitive empirical performances of PMD compared to several approximate inference algorithms in mixture models, logistic regression, sparse Gaussian processes and latent Dirichlet allocation on large scale datasets."
                    },
                    {
                        "title": "Discriminative Embeddings of Latent Variable Models for Structured Data",
                        "abstract": "Kernel classifiers and regressors designed for structured data, such as sequences, trees and graphs, have significantly advanced a number of interdisciplinary areas such as computational biology and drug design. Typically, kernels are designed beforehand for a data type which either exploit statistics of the structures or make use of probabilistic generative models, and then a discriminative classifier is learned based on the kernels via convex optimization. However, such an elegant two-stage approach also limited kernel methods from scaling up to millions of data points, and exploiting discriminative information to learn feature representations.   We propose, structure2vec, an effective and scalable approach for structured data representation based on the idea of embedding latent variable models into feature spaces, and learning such feature spaces using discriminative information. Interestingly, structure2vec extracts features by performing a sequence of function mappings in a way similar to graphical model inference procedures, such as mean field and belief propagation. In applications involving millions of data points, we showed that structure2vec runs 2 times faster, produces models which are $10,000$ times smaller, while at the same time achieving the state-of-the-art predictive performance."
                    },
                    {
                        "title": "Energy-Based Processes for Exchangeable Data",
                        "abstract": "Recently there has been growing interest in modeling sets with exchangeability such as point clouds. A shortcoming of current approaches is that they restrict the cardinality of the sets considered or can only express limited forms of distribution over unobserved data. To overcome these limitations, we introduce Energy-Based Processes (EBPs), which extend energy based models to exchangeable data while allowing neural network parameterizations of the energy function. A key advantage of these models is the ability to express more flexible distributions over sets without restricting their cardinality. We develop an efficient training procedure for EBPs that demonstrates state-of-the-art performance on a variety of tasks such as point cloud generation, classification, denoising, and image completion."
                    },
                    {
                        "title": "Deep Coevolutionary Network: Embedding User and Item Features for Recommendation",
                        "abstract": "Recommender systems often use latent features to explain the behaviors of users and capture the properties of items. As users interact with different items over time, user and item features can influence each other, evolve and co-evolve over time. The compatibility of user and item's feature further influence the future interaction between users and items. Recently, point process based models have been proposed in the literature aiming to capture the temporally evolving nature of these latent features. However, these models often make strong parametric assumptions about the evolution process of the user and item latent features, which may not reflect the reality, and has limited power in expressing the complex and nonlinear dynamics underlying these processes. To address these limitations, we propose a novel deep coevolutionary network model (DeepCoevolve), for learning user and item features based on their interaction graph. DeepCoevolve use recurrent neural network (RNN) over evolving networks to define the intensity function in point processes, which allows the model to capture complex mutual influence between users and items, and the feature evolution over time. We also develop an efficient procedure for training the model parameters, and show that the learned models lead to significant improvements in recommendation and activity prediction compared to previous state-of-the-arts parametric models."
                    },
                    {
                        "title": "KG^2: Learning to Reason Science Exam Questions with Contextual Knowledge Graph Embeddings",
                        "abstract": "The AI2 Reasoning Challenge (ARC), a new benchmark dataset for question answering (QA) has been recently released. ARC only contains natural science questions authored for human exams, which are hard to answer and require advanced logic reasoning. On the ARC Challenge Set, existing state-of-the-art QA systems fail to significantly outperform random baseline, reflecting the difficult nature of this task. In this paper, we propose a novel framework for answering science exam questions, which mimics human solving process in an open-book exam. To address the reasoning challenge, we construct contextual knowledge graphs respectively for the question itself and supporting sentences. Our model learns to reason with neural embeddings of both knowledge graphs. Experiments on the ARC Challenge Set show that our model outperforms the previous state-of-the-art QA systems."
                    },
                    {
                        "title": "Particle Flow Bayes' Rule",
                        "abstract": "We present a particle flow realization of Bayes' rule, where an ODE-based neural operator is used to transport particles from a prior to its posterior after a new observation. We prove that such an ODE operator exists. Its neural parameterization can be trained in a meta-learning framework, allowing this operator to reason about the effect of an individual observation on the posterior, and thus generalize across different priors, observations and to sequential Bayesian inference. We demonstrated the generalization ability of our particle flow Bayes operator in several canonical and high dimensional examples."
                    },
                    {
                        "title": "Retro*: Learning Retrosynthetic Planning with Neural Guided A* Search",
                        "abstract": "Retrosynthetic planning is a critical task in organic chemistry which identifies a series of reactions that can lead to the synthesis of a target product. The vast number of possible chemical transformations makes the size of the search space very big, and retrosynthetic planning is challenging even for experienced chemists. However, existing methods either require expensive return estimation by rollout with high variance, or optimize for search speed rather than the quality. In this paper, we propose Retro*, a neural-based A*-like algorithm that finds high-quality synthetic routes efficiently. It maintains the search as an AND-OR tree, and learns a neural search bias with off-policy data. Then guided by this neural network, it performs best-first search efficiently during new planning episodes. Experiments on benchmark USPTO datasets show that, our proposed method outperforms existing state-of-the-art with respect to both the success rate and solution quality, while being more efficient at the same time."
                    },
                    {
                        "title": "CrossBeam: Learning to Search in Bottom-Up Program Synthesis",
                        "abstract": "Many approaches to program synthesis perform a search within an enormous space of programs to find one that satisfies a given specification. Prior works have used neural models to guide combinatorial search algorithms, but such approaches still explore a huge portion of the search space and quickly become intractable as the size of the desired program increases. To tame the search space blowup, we propose training a neural model to learn a hands-on search policy for bottom-up synthesis, instead of relying on a combinatorial search algorithm. Our approach, called CrossBeam, uses the neural model to choose how to combine previously-explored programs into new programs, taking into account the search history and partial program executions. Motivated by work in structured prediction on learning to search, CrossBeam is trained on-policy using data extracted from its own bottom-up searches on training tasks. We evaluate CrossBeam in two very different domains, string manipulation and logic programming. We observe that CrossBeam learns to search efficiently, exploring much smaller portions of the program space compared to the state-of-the-art."
                    },
                    {
                        "title": "Know-Evolve: Deep Temporal Reasoning for Dynamic Knowledge Graphs",
                        "abstract": "The availability of large scale event data with time stamps has given rise to dynamically evolving knowledge graphs that contain temporal information for each edge. Reasoning over time in such dynamic knowledge graphs is not yet well understood. To this end, we present Know-Evolve, a novel deep evolutionary knowledge network that learns non-linearly evolving entity representations over time. The occurrence of a fact (edge) is modeled as a multivariate point process whose intensity function is modulated by the score for that fact computed based on the learned entity embeddings. We demonstrate significantly improved performance over various relational learning approaches on two large scale real-world datasets. Further, our method effectively predicts occurrence or recurrence time of a fact which is novel compared to prior reasoning approaches in multi-relational setting."
                    },
                    {
                        "title": "Annealed Training for Combinatorial Optimization on Graphs",
                        "abstract": "The hardness of combinatorial optimization (CO) problems hinders collecting solutions for supervised learning. However, learning neural networks for CO problems is notoriously difficult in lack of the labeled data as the training is easily trapped at local optima. In this work, we propose a simple but effective annealed training framework for CO problems. In particular, we transform CO problems into unbiased energy-based models (EBMs). We carefully selected the penalties terms so as to make the EBMs as smooth as possible. Then we train graph neural networks to approximate the EBMs. To prevent the training from being stuck at local optima near the initialization, we introduce an annealed loss function. An experimental evaluation demonstrates that our annealed training framework obtains substantial improvements. In four types of CO problems, our method achieves performance substantially better than other unsupervised neural methods on both synthetic and real-world graphs."
                    },
                    {
                        "title": "Optimal Scaling for Locally Balanced Proposals in Discrete Spaces",
                        "abstract": "Optimal scaling has been well studied for Metropolis-Hastings (M-H) algorithms in continuous spaces, but a similar understanding has been lacking in discrete spaces. Recently, a family of locally balanced proposals (LBP) for discrete spaces has been proved to be asymptotically optimal, but the question of optimal scaling has remained open. In this paper, we establish, for the first time, that the efficiency of M-H in discrete spaces can also be characterized by an asymptotic acceptance rate that is independent of the target distribution. Moreover, we verify, both theoretically and empirically, that the optimal acceptance rates for LBP and random walk Metropolis (RWM) are $0.574$ and $0.234$ respectively. These results also help establish that LBP is asymptotically $O(N^\\frac{2}{3})$ more efficient than RWM with respect to model dimension $N$. Knowledge of the optimal acceptance rate allows one to automatically tune the neighborhood size of a proposal distribution in a discrete space, directly analogous to step-size control in continuous spaces. We demonstrate empirically that such adaptive M-H sampling can robustly improve sampling in a variety of target distributions in discrete spaces, including training deep energy based models."
                    },
                    {
                        "title": "Gradient-Free Structured Pruning with Unlabeled Data",
                        "abstract": "Large Language Models (LLMs) have achieved great success in solving difficult tasks across many domains, but such success comes with a high computation cost, and inference latency. As developers and third parties customize these models, the need to provide efficient inference has increased. Many efforts have attempted to reduce inference cost through model compression techniques such as pruning and distillation. However, these techniques either require labeled data, or are time-consuming as they require the compressed model to be retrained to regain accuracy. In this paper, we propose a gradient-free structured pruning framework that uses only unlabeled data. An evaluation on the GLUE and SQuAD benchmarks using BERT$_{BASE}$ and DistilBERT illustrates the effectiveness of the proposed approach. By only using the weights of the pre-trained model and unlabeled data, in a matter of a few minutes on a single GPU, up to 40% of the original FLOP count can be reduced with less than a 4% accuracy loss across all tasks considered."
                    },
                    {
                        "title": "Preference Optimization for Molecule Synthesis with Conditional Residual Energy-based Models",
                        "abstract": "Molecule synthesis through machine learning is one of the fundamental problems in drug discovery. Current data-driven strategies employ one-step retrosynthesis models and search algorithms to predict synthetic routes in a top-bottom manner. Despite their effective performance, these strategies face limitations in the molecule synthetic route generation due to a greedy selection of the next molecule set without any lookahead. Furthermore, existing strategies cannot control the generation of synthetic routes based on possible criteria such as material costs, yields, and step count. In this work, we propose a general and principled framework via conditional residual energy-based models (EBMs), that focus on the quality of the entire synthetic route based on the specific criteria. By incorporating an additional energy-based function into our probabilistic model, our proposed algorithm can enhance the quality of the most probable synthetic routes (with higher probabilities) generated by various strategies in a plug-and-play fashion. Extensive experiments demonstrate that our framework can consistently boost performance across various strategies and outperforms previous state-of-the-art top-1 accuracy by a margin of 2.5%. Code is available at https://github.com/SongtaoLiu0823/CREBM."
                    },
                    {
                        "title": "Autoregressive Large Language Models are Computationally Universal",
                        "abstract": "We show that autoregressive decoding of a transformer-based language model can realize universal computation, without external intervention or modification of the model's weights. Establishing this result requires understanding how a language model can process arbitrarily long inputs using a bounded context. For this purpose, we consider a generalization of autoregressive decoding where, given a long input, emitted tokens are appended to the end of the sequence as the context window advances. We first show that the resulting system corresponds to a classical model of computation, a Lag system, that has long been known to be computationally universal. By leveraging a new proof, we show that a universal Turing machine can be simulated by a Lag system with 2027 production rules. We then investigate whether an existing large language model can simulate the behaviour of such a universal Lag system. We give an affirmative answer by showing that a single system-prompt can be developed for gemini-1.5-pro-001 that drives the model, under deterministic (greedy) decoding, to correctly apply each of the 2027 production rules. We conclude that, by the Church-Turing thesis, prompted gemini-1.5-pro-001 with extended autoregressive (greedy) decoding is a general purpose computer."
                    },
                    {
                        "title": "Scalable Deep Generative Modeling for Sparse Graphs",
                        "abstract": "Learning graph generative models is a challenging task for deep learning and has wide applicability to a range of domains like chemistry, biology and social science. However current deep neural methods suffer from limited scalability: for a graph with $n$ nodes and $m$ edges, existing deep neural methods require $\\Omega(n^2)$ complexity by building up the adjacency matrix. On the other hand, many real world graphs are actually sparse in the sense that $m\\ll n^2$. Based on this, we develop a novel autoregressive model, named BiGG, that utilizes this sparsity to avoid generating the full adjacency matrix, and importantly reduces the graph generation time complexity to $O((n + m)\\log n)$. Furthermore, during training this autoregressive model can be parallelized with $O(\\log n)$ synchronization stages, which makes it much more efficient than other autoregressive models that require $\\Omega(n)$. Experiments on several benchmarks show that the proposed approach not only scales to orders of magnitude larger graphs than previously possible with deep autoregressive graph generative models, but also yields better graph generation quality."
                    },
                    {
                        "title": "Syntax-Directed Variational Autoencoder for Structured Data",
                        "abstract": "Deep generative models have been enjoying success in modeling continuous data. However it remains challenging to capture the representations for discrete structures with formal grammars and semantics, e.g., computer programs and molecular structures. How to generate both syntactically and semantically correct data still remains largely an open problem. Inspired by the theory of compiler where the syntax and semantics check is done via syntax-directed translation (SDT), we propose a novel syntax-directed variational autoencoder (SD-VAE) by introducing stochastic lazy attributes. This approach converts the offline SDT check into on-the-fly generated guidance for constraining the decoder. Comparing to the state-of-the-art methods, our approach enforces constraints on the output space so that the output will be not only syntactically valid, but also semantically reasonable. We evaluate the proposed model with applications in programming language and molecules, including reconstruction and program/molecule optimization. The results demonstrate the effectiveness in incorporating syntactic and semantic constraints in discrete generative models, which is significantly better than current state-of-the-art approaches."
                    },
                    {
                        "title": "Learning Discrete Energy-based Models via Auxiliary-variable Local Exploration",
                        "abstract": "Discrete structures play an important role in applications like program language modeling and software engineering. Current approaches to predicting complex structures typically consider autoregressive models for their tractability, with some sacrifice in flexibility. Energy-based models (EBMs) on the other hand offer a more flexible and thus more powerful approach to modeling such distributions, but require partition function estimation. In this paper we propose ALOE, a new algorithm for learning conditional and unconditional EBMs for discrete structured data, where parameter gradients are estimated using a learned sampler that mimics local search. We show that the energy function and sampler can be trained efficiently via a new variational form of power iteration, achieving a better trade-off between flexibility and tractability. Experimentally, we show that learning local search leads to significant improvements in challenging application domains. Most notably, we present an energy model guided fuzzer for software testing that achieves comparable performance to well engineered fuzzing engines like libfuzzer."
                    },
                    {
                        "title": "Neural Stochastic Dual Dynamic Programming",
                        "abstract": "Stochastic dual dynamic programming (SDDP) is a state-of-the-art method for solving multi-stage stochastic optimization, widely used for modeling real-world process optimization tasks. Unfortunately, SDDP has a worst-case complexity that scales exponentially in the number of decision variables, which severely limits applicability to only low dimensional problems. To overcome this limitation, we extend SDDP by introducing a trainable neural model that learns to map problem instances to a piece-wise linear value function within intrinsic low-dimension space, which is architected specifically to interact with a base SDDP solver, so that can accelerate optimization performance on new instances. The proposed Neural Stochastic Dual Dynamic Programming ($\\nu$-SDDP) continually self-improves by solving successive problems. An empirical investigation demonstrates that $\\nu$-SDDP can significantly reduce problem solving cost without sacrificing solution quality over competitors such as SDDP and reinforcement learning algorithms, across a range of synthetic and real-world process optimization problems."
                    },
                    {
                        "title": "Energy-based View of Retrosynthesis",
                        "abstract": "Retrosynthesis -- the process of identifying a set of reactants to synthesize a target molecule -- is of vital importance to material design and drug discovery. Existing machine learning approaches based on language models and graph neural networks have achieved encouraging results. In this paper, we propose a framework that unifies sequence- and graph-based methods as energy-based models (EBMs) with different energy functions. This unified perspective provides critical insights about EBM variants through a comprehensive assessment of performance. Additionally, we present a novel dual variant within the framework that performs consistent training over Bayesian forward- and backward-prediction by constraining the agreement between the two directions. This model improves state-of-the-art performance by 9.6% for template-free approaches where the reaction type is unknown."
                    },
                    {
                        "title": "Discrete Langevin Sampler via Wasserstein Gradient Flow",
                        "abstract": "It is known that gradient-based MCMC samplers for continuous spaces, such as Langevin Monte Carlo (LMC), can be derived as particle versions of a gradient flow that minimizes KL divergence on a Wasserstein manifold. The superior efficiency of such samplers has motivated several recent attempts to generalize LMC to discrete spaces. However, a fully principled extension of Langevin dynamics to discrete spaces has yet to be achieved, due to the lack of well-defined gradients in the sample space. In this work, we show how the Wasserstein gradient flow can be generalized naturally to discrete spaces. Given the proposed formulation, we demonstrate how a discrete analogue of Langevin dynamics can subsequently be developed. With this new understanding, we reveal how recent gradient-based samplers in discrete spaces can be obtained as special cases by choosing particular discretizations. More importantly, the framework also allows for the derivation of novel algorithms, one of which, \\textit{Discrete Langevin Monte Carlo} (DLMC), is obtained by a factorized estimate of the transition matrix. The DLMC method admits a convenient parallel implementation and time-uniform sampling that achieves larger jump distances. We demonstrate the advantages of DLMC on various binary and categorical distributions."
                    }
                ]
            },
            "f4144abe-e4ce-4a18-a2cd-2743cd567485": {
                "pk": "f4144abe-e4ce-4a18-a2cd-2743cd567485",
                "name": "Bethany Yixin Wang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "0dfa5dd4-e643-4f2d-9804-b56aa6cbb998": {
                "pk": "0dfa5dd4-e643-4f2d-9804-b56aa6cbb998",
                "name": "Xingchen Wan",
                "collaborators": [
                    "Binxin Ru",
                    "Xiaowen Dong",
                    "Michael Osborne",
                    "Diego Granziol",
                    "Han Zhou",
                    "Ivan Vuli\u0107",
                    "Anna Korhonen",
                    "Michael A. Osborne",
                    "Samuel Albanie",
                    "Stephen Roberts"
                ],
                "domain": [
                    "Neural Architecture Search",
                    "Bayesian Optimization",
                    "Natural Language Processing",
                    "Graph Optimization"
                ],
                "publications": [
                    {
                        "title": "Interpretable Neural Architecture Search via Bayesian Optimisation with Weisfeiler-Lehman Kernels",
                        "abstract": "Current neural architecture search (NAS) strategies focus only on finding a single, good, architecture. They offer little insight into why a specific network is performing well, or how we should modify the architecture if we want further improvements. We propose a Bayesian optimisation (BO) approach for NAS that combines the Weisfeiler-Lehman graph kernel with a Gaussian process surrogate. Our method optimises the architecture in a highly data-efficient manner: it is capable of capturing the topological structures of the architectures and is scalable to large graphs, thus making the high-dimensional and graph-like search spaces amenable to BO. More importantly, our method affords interpretability by discovering useful network features and their corresponding impact on the network performance. Indeed, we demonstrate empirically that our surrogate model is capable of identifying useful motifs which can guide the generation of new architectures. We finally show that our method outperforms existing NAS approaches to achieve the state of the art on both closed- and open-domain search spaces."
                    },
                    {
                        "title": "Iterative Averaging in the Quest for Best Test Error",
                        "abstract": "We analyse and explain the increased generalisation performance of iterate averaging using a Gaussian process perturbation model between the true and batch risk surface on the high dimensional quadratic. We derive three phenomena \\latestEdits{from our theoretical results:} (1) The importance of combining iterate averaging (IA) with large learning rates and regularisation for improved regularisation. (2) Justification for less frequent averaging. (3) That we expect adaptive gradient methods to work equally well, or better, with iterate averaging than their non-adaptive counterparts. Inspired by these results\\latestEdits{, together with} empirical investigations of the importance of appropriate regularisation for the solution diversity of the iterates, we propose two adaptive algorithms with iterate averaging. These give significantly better results compared to stochastic gradient descent (SGD), require less tuning and do not require early stopping or validation set monitoring. We showcase the efficacy of our approach on the CIFAR-10/100, ImageNet and Penn Treebank datasets on a variety of modern and classical network architectures."
                    },
                    {
                        "title": "Deep Curvature Suite",
                        "abstract": "We present MLRG Deep Curvature suite, a PyTorch-based, open-source package for analysis and visualisation of neural network curvature and loss landscape. Despite of providing rich information into properties of neural network and useful for a various designed tasks, curvature information is still not made sufficient use for various reasons, and our method aims to bridge this gap. We present a primer, including its main practical desiderata and common misconceptions, of \\textit{Lanczos algorithm}, the theoretical backbone of our package, and present a series of examples based on synthetic toy examples and realistic modern neural networks tested on CIFAR datasets, and show the superiority of our package against existing competing approaches for the similar purposes."
                    },
                    {
                        "title": "Working Memory Capacity of ChatGPT: An Empirical Study",
                        "abstract": "Working memory is a critical aspect of both human intelligence and artificial intelligence, serving as a workspace for the temporary storage and manipulation of information. In this paper, we systematically assess the working memory capacity of ChatGPT, a large language model developed by OpenAI, by examining its performance in verbal and spatial n-back tasks under various conditions. Our experiments reveal that ChatGPT has a working memory capacity limit strikingly similar to that of humans. Furthermore, we investigate the impact of different instruction strategies on ChatGPT's performance and observe that the fundamental patterns of a capacity limit persist. From our empirical findings, we propose that n-back tasks may serve as tools for benchmarking the working memory capacity of large language models and hold potential for informing future efforts aimed at enhancing AI working memory."
                    },
                    {
                        "title": "Bayesian Optimization of Functions over Node Subsets in Graphs",
                        "abstract": "We address the problem of optimizing over functions defined on node subsets in a graph. The optimization of such functions is often a non-trivial task given their combinatorial, black-box and expensive-to-evaluate nature. Although various algorithms have been introduced in the literature, most are either task-specific or computationally inefficient and only utilize information about the graph structure without considering the characteristics of the function. To address these limitations, we utilize Bayesian Optimization (BO), a sample-efficient black-box solver, and propose a novel framework for combinatorial optimization on graphs. More specifically, we map each $k$-node subset in the original graph to a node in a new combinatorial graph and adopt a local modeling approach to efficiently traverse the latter graph by progressively sampling its subgraphs using a recursive algorithm. Extensive experiments under both synthetic and real-world setups demonstrate the effectiveness of the proposed BO framework on various types of graphs and optimization tasks, where its behavior is analyzed in detail with ablation studies."
                    },
                    {
                        "title": "AutoPEFT: Automatic Configuration Search for Parameter-Efficient Fine-Tuning",
                        "abstract": "Large pretrained language models are widely used in downstream NLP tasks via task-specific fine-tuning, but such procedures can be costly. Recently, Parameter-Efficient Fine-Tuning (PEFT) methods have achieved strong task performance while updating much fewer parameters than full model fine-tuning (FFT). However, it is non-trivial to make informed design choices on the PEFT configurations, such as their architecture, the number of tunable parameters, and even the layers in which the PEFT modules are inserted. Consequently, it is highly likely that the current, manually designed configurations are suboptimal in terms of their performance-efficiency trade-off. Inspired by advances in neural architecture search, we propose AutoPEFT for automatic PEFT configuration selection: we first design an expressive configuration search space with multiple representative PEFT modules as building blocks. Using multi-objective Bayesian optimisation in a low-cost setup, we then discover a Pareto-optimal set of configurations with strong performance-cost trade-offs across different numbers of parameters that are also highly transferable across different tasks. Empirically, on GLUE and SuperGLUE tasks, we show that AutoPEFT-discovered configurations significantly outperform existing PEFT methods and are on par or better than FFT without incurring substantial training efficiency costs."
                    },
                    {
                        "title": "Survival of the Most Influential Prompts: Efficient Black-Box Prompt Search via Clustering and Pruning",
                        "abstract": "Prompt-based learning has been an effective paradigm for large pretrained language models (LLM), enabling few-shot or even zero-shot learning. Black-box prompt search has received growing interest recently for its distinctive properties of gradient-free optimization, proven particularly useful and powerful for model-as-a-service usage. However, the discrete nature and the complexity of combinatorial optimization hinder the efficiency of modern black-box approaches. Despite extensive research on search algorithms, the crucial aspect of search space design and optimization has been largely overlooked. In this paper, we first conduct a sensitivity analysis by prompting LLM, revealing that only a small number of tokens exert a disproportionate amount of influence on LLM predictions. Leveraging this insight, we propose the Clustering and Pruning for Efficient Black-box Prompt Search (ClaPS), a simple black-box search method that first clusters and prunes the search space to focus exclusively on influential prompt tokens. By employing even simple search methods within the pruned search space, ClaPS achieves state-of-the-art performance across various tasks and LLMs, surpassing the performance of complex approaches while significantly reducing search costs. Our findings underscore the critical role of search space design and optimization in enhancing both the usefulness and the efficiency of black-box prompt-based learning."
                    },
                    {
                        "title": "On Redundancy and Diversity in Cell-based Neural Architecture Search",
                        "abstract": "Searching for the architecture cells is a dominant paradigm in NAS. However, little attention has been devoted to the analysis of the cell-based search spaces even though it is highly important for the continual development of NAS. In this work, we conduct an empirical post-hoc analysis of architectures from the popular cell-based search spaces and find that the existing search spaces contain a high degree of redundancy: the architecture performance is minimally sensitive to changes at large parts of the cells, and universally adopted designs, like the explicit search for a reduction cell, significantly increase the complexities but have very limited impact on the performance. Across architectures found by a diverse set of search strategies, we consistently find that the parts of the cells that do matter for architecture performance often follow similar and simple patterns. By explicitly constraining cells to include these patterns, randomly sampled architectures can match or even outperform the state of the art. These findings cast doubts into our ability to discover truly novel architectures in the existing cell-based search spaces, and inspire our suggestions for improvement to guide future NAS research. Code is available at https://github.com/xingchenwan/cell-based-NAS-analysis."
                    },
                    {
                        "title": "Teach Better or Show Smarter? On Instructions and Exemplars in Automatic Prompt Optimization",
                        "abstract": "Large language models have demonstrated remarkable capabilities, but their performance is heavily reliant on effective prompt engineering. Automatic prompt optimization (APO) methods are designed to automate this and can be broadly categorized into those targeting instructions (instruction optimization, IO) vs. those targeting exemplars (exemplar selection, ES). Despite their shared objective, these have evolved rather independently, with IO recently receiving more research attention. This paper seeks to bridge this gap by comprehensively comparing the performance of representative IO and ES techniques, both isolation and combination, on a diverse set of challenging tasks. Our findings reveal that intelligently reusing model-generated input-output pairs obtained from evaluating prompts on the validation set as exemplars consistently improves performance over IO methods but is currently under-investigated. We also find that despite the recent focus on IO, how we select exemplars can outweigh how we optimize instructions, with ES strategies as simple as random search outperforming state-of-the-art IO methods with seed instructions without any optimization. Moreover, we observe synergy between ES and IO, with optimal combinations surpassing individual contributions. We conclude that studying exemplar selection as a standalone method and its optimal combination with instruction optimization remains a crucial aspect of APO and deserves greater consideration in future research, even in the era of highly capable instruction-following models."
                    },
                    {
                        "title": "Bayesian Quadrature for Neural Ensemble Search",
                        "abstract": "Ensembling can improve the performance of Neural Networks, but existing approaches struggle when the architecture likelihood surface has dispersed, narrow peaks. Furthermore, existing methods construct equally weighted ensembles, and this is likely to be vulnerable to the failure modes of the weaker architectures. By viewing ensembling as approximately marginalising over architectures we construct ensembles using the tools of Bayesian Quadrature -- tools which are well suited to the exploration of likelihood surfaces with dispersed, narrow peaks. Additionally, the resulting ensembles consist of architectures weighted commensurate with their performance. We show empirically -- in terms of test likelihood, accuracy, and expected calibration error -- that our method outperforms state-of-the-art baselines, and verify via ablation studies that its components do so independently."
                    },
                    {
                        "title": "SimSC: A Simple Framework for Semantic Correspondence with Temperature Learning",
                        "abstract": "We propose SimSC, a remarkably simple framework, to address the problem of semantic matching only based on the feature backbone. We discover that when fine-tuning ImageNet pre-trained backbone on the semantic matching task, L2 normalization of the feature map, a standard procedure in feature matching, produces an overly smooth matching distribution and significantly hinders the fine-tuning process. By setting an appropriate temperature to the softmax, this over-smoothness can be alleviated and the quality of features can be substantially improved. We employ a learning module to predict the optimal temperature for fine-tuning feature backbones. This module is trained together with the backbone and the temperature is updated online. We evaluate our method on three public datasets and demonstrate that we can achieve accuracy on par with state-of-the-art methods under the same backbone without using a learned matching head. Our method is versatile and works on various types of backbones. We show that the accuracy of our framework can be easily improved by coupling it with more powerful backbones."
                    },
                    {
                        "title": "Think Global and Act Local: Bayesian Optimisation over High-Dimensional Categorical and Mixed Search Spaces",
                        "abstract": "High-dimensional black-box optimisation remains an important yet notoriously challenging problem. Despite the success of Bayesian optimisation methods on continuous domains, domains that are categorical, or that mix continuous and categorical variables, remain challenging. We propose a novel solution -- we combine local optimisation with a tailored kernel design, effectively handling high-dimensional categorical and mixed search spaces, whilst retaining sample efficiency. We further derive convergence guarantee for the proposed approach. Finally, we demonstrate empirically that our method outperforms the current baselines on a variety of synthetic and real-world tasks in terms of performance, computational costs, or both."
                    },
                    {
                        "title": "Bayesian Optimisation of Functions on Graphs",
                        "abstract": "The increasing availability of graph-structured data motivates the task of optimising over functions defined on the node set of graphs. Traditional graph search algorithms can be applied in this case, but they may be sample-inefficient and do not make use of information about the function values; on the other hand, Bayesian optimisation is a class of promising black-box solvers with superior sample efficiency, but it has been scarcely been applied to such novel setups. To fill this gap, we propose a novel Bayesian optimisation framework that optimises over functions defined on generic, large-scale and potentially unknown graphs. Through the learning of suitable kernels on graphs, our framework has the advantage of adapting to the behaviour of the target function. The local modelling approach further guarantees the efficiency of our method. Extensive experiments on both synthetic and real-world graphs demonstrate the effectiveness of the proposed optimisation framework."
                    },
                    {
                        "title": "Sentiment Correlation in Financial News Networks and Associated Market Movements",
                        "abstract": "In an increasingly connected global market, news sentiment towards one company may not only indicate its own market performance, but can also be associated with a broader movement on the sentiment and performance of other companies from the same or even different sectors. In this paper, we apply NLP techniques to understand news sentiment of 87 companies among the most reported on Reuters for a period of seven years. We investigate the propagation of such sentiment in company networks and evaluate the associated market movements in terms of stock price and volatility. Our results suggest that, in certain sectors, strong media sentiment towards one company may indicate a significant change in media sentiment towards related companies measured as neighbours in a financial network constructed from news co-occurrence. Furthermore, there exists a weak but statistically significant association between strong media sentiment and abnormal market return as well as volatility. Such an association is more significant at the level of individual companies, but nevertheless remains visible at the level of sectors or groups of companies."
                    }
                ]
            },
            "4ea5b13a-abc5-4ae7-9c85-2f6420d0dce6": {
                "pk": "4ea5b13a-abc5-4ae7-9c85-2f6420d0dce6",
                "name": "Bo Dai",
                "collaborators": [
                    "Le Song",
                    "Dahua Lin",
                    "Hanjun Dai",
                    "Animashree Anandkumar",
                    "Bo Xie",
                    "Fenfen Li",
                    "Ofir Nachum",
                    "Niao He",
                    "Yuan Qi",
                    "Yao Zhu"
                ],
                "domain": [
                    "Latent Variable Models",
                    "Image Captioning",
                    "Bayesian Inference",
                    "Kernel Methods"
                ],
                "publications": [
                    {
                        "title": "Nonparametric Estimation of Multi-View Latent Variable Models",
                        "abstract": "Spectral methods have greatly advanced the estimation of latent variable models, generating a sequence of novel and efficient algorithms with strong theoretical guarantees. However, current spectral algorithms are largely restricted to mixtures of discrete or Gaussian distributions. In this paper, we propose a kernel method for learning multi-view latent variable models, allowing each mixture component to be nonparametric. The key idea of the method is to embed the joint distribution of a multi-view latent variable into a reproducing kernel Hilbert space, and then the latent parameters are recovered using a robust tensor power method. We establish that the sample complexity for the proposed method is quadratic in the number of latent components and is a low order polynomial in the other relevant parameters. Thus, our non-parametric tensor approach to learning latent variable models enjoys good sample and computational efficiencies. Moreover, the non-parametric tensor power method compares favorably to EM algorithm and other existing spectral algorithms in our experiments."
                    },
                    {
                        "title": "Contrastive Learning for Image Captioning",
                        "abstract": "Image captioning, a popular topic in computer vision, has achieved substantial progress in recent years. However, the distinctiveness of natural descriptions is often overlooked in previous work. It is closely related to the quality of captions, as distinctive captions are more likely to describe images with their unique aspects. In this work, we propose a new learning method, Contrastive Learning (CL), for image captioning. Specifically, via two constraints formulated on top of a reference model, the proposed method can encourage distinctiveness, while maintaining the overall quality of the generated captions. We tested our method on two challenging datasets, where it improves the baseline model by significant margins. We also showed in our studies that the proposed method is generic and can be used for models with various structures."
                    },
                    {
                        "title": "A profit allocation mechanism for customer-to-manufacturer platform in e-commercial business",
                        "abstract": "A centralized collaboration problem of customer-to-manufacturer (C2M) platform in e-commercial business is studied in this paper, where an e-commercial company and multiple manufacturers come to an agreement for constituting a collaborative alliance to satisfy the product orders of multiple customers. The problem has two important issues which are the optimal reallocation of customer orders among the manufacturers to maximize a total profit of the alliance and a fair allocation of the profit among the manufacturers so as to assure the stability of the alliance and the sustainability of C2M platform. A profit allocation mechanism is proposed which is based on core allocation concept and considers the contribution of each manufacturer in fulfilling orders. The effectiveness of the mechanism is evaluated with numerical experiments on sixty instances generated based on real data of Alibaba."
                    },
                    {
                        "title": "Reinforcement Learning via Fenchel-Rockafellar Duality",
                        "abstract": "We review basic concepts of convex duality, focusing on the very general and supremely useful Fenchel-Rockafellar duality. We summarize how this duality may be applied to a variety of reinforcement learning (RL) settings, including policy evaluation or optimization, online or offline learning, and discounted or undiscounted rewards. The derivations yield a number of intriguing results, including the ability to perform policy evaluation and on-policy policy gradient with behavior-agnostic offline data and methods to learn a policy via max-likelihood optimization. Although many of these results have appeared previously in various forms, we provide a unified treatment and perspective on these results, which we hope will enable researchers to better use and apply the tools of convex duality to make further progress in RL."
                    },
                    {
                        "title": "Provable Bayesian Inference via Particle Mirror Descent",
                        "abstract": "Bayesian methods are appealing in their flexibility in modeling complex data and ability in capturing uncertainty in parameters. However, when Bayes' rule does not result in tractable closed-form, most approximate inference algorithms lack either scalability or rigorous guarantees. To tackle this challenge, we propose a simple yet provable algorithm, \\emph{Particle Mirror Descent} (PMD), to iteratively approximate the posterior density. PMD is inspired by stochastic functional mirror descent where one descends in the density space using a small batch of data points at each iteration, and by particle filtering where one uses samples to approximate a function. We prove result of the first kind that, with $m$ particles, PMD provides a posterior density estimator that converges in terms of $KL$-divergence to the true posterior in rate $O(1/\\sqrt{m})$. We demonstrate competitive empirical performances of PMD compared to several approximate inference algorithms in mixture models, logistic regression, sparse Gaussian processes and latent Dirichlet allocation on large scale datasets."
                    },
                    {
                        "title": "Discriminative Embeddings of Latent Variable Models for Structured Data",
                        "abstract": "Kernel classifiers and regressors designed for structured data, such as sequences, trees and graphs, have significantly advanced a number of interdisciplinary areas such as computational biology and drug design. Typically, kernels are designed beforehand for a data type which either exploit statistics of the structures or make use of probabilistic generative models, and then a discriminative classifier is learned based on the kernels via convex optimization. However, such an elegant two-stage approach also limited kernel methods from scaling up to millions of data points, and exploiting discriminative information to learn feature representations.   We propose, structure2vec, an effective and scalable approach for structured data representation based on the idea of embedding latent variable models into feature spaces, and learning such feature spaces using discriminative information. Interestingly, structure2vec extracts features by performing a sequence of function mappings in a way similar to graphical model inference procedures, such as mean field and belief propagation. In applications involving millions of data points, we showed that structure2vec runs 2 times faster, produces models which are $10,000$ times smaller, while at the same time achieving the state-of-the-art predictive performance."
                    },
                    {
                        "title": "EigenGP: Sparse Gaussian process models with data-dependent eigenfunctions",
                        "abstract": "Gaussian processes (GPs) provide a nonparametric representation of functions. However, classical GP inference suffers from high computational cost and it is difficult to design nonstationary GP priors in practice. In this paper, we propose a sparse Gaussian process model, EigenGP, based on the Karhunen-Loeve (KL) expansion of a GP prior. We use the Nystrom approximation to obtain data dependent eigenfunctions and select these eigenfunctions by evidence maximization. This selection reduces the number of eigenfunctions in our model and provides a nonstationary covariance function. To handle nonlinear likelihoods, we develop an efficient expectation propagation (EP) inference algorithm, and couple it with expectation maximization for eigenfunction selection. Because the eigenfunctions of a Gaussian kernel are associated with clusters of samples - including both the labeled and unlabeled - selecting relevant eigenfunctions enables EigenGP to conduct semi-supervised learning. Our experimental results demonstrate improved predictive performance of EigenGP over alternative state-of-the-art sparse GP and semisupervised learning methods for regression, classification, and semisupervised classification."
                    },
                    {
                        "title": "Detecting Visual Relationships with Deep Relational Networks",
                        "abstract": "Relationships among objects play a crucial role in image understanding. Despite the great success of deep learning techniques in recognizing individual objects, reasoning about the relationships among objects remains a challenging task. Previous methods often treat this as a classification problem, considering each type of relationship (e.g. \"ride\") or each distinct visual phrase (e.g. \"person-ride-horse\") as a category. Such approaches are faced with significant difficulties caused by the high diversity of visual appearance for each kind of relationships or the large number of distinct visual phrases. We propose an integrated framework to tackle this problem. At the heart of this framework is the Deep Relational Network, a novel formulation designed specifically for exploiting the statistical dependencies between objects and their relationships. On two large datasets, the proposed method achieves substantial improvement over state-of-the-art."
                    },
                    {
                        "title": "A Neural Compositional Paradigm for Image Captioning",
                        "abstract": "Mainstream captioning models often follow a sequential structure to generate captions, leading to issues such as introduction of irrelevant semantics, lack of diversity in the generated captions, and inadequate generalization performance. In this paper, we present an alternative paradigm for image captioning, which factorizes the captioning procedure into two stages: (1) extracting an explicit semantic representation from the given image; and (2) constructing the caption based on a recursive compositional procedure in a bottom-up manner. Compared to conventional ones, our paradigm better preserves the semantic content through an explicit factorization of semantics and syntax. By using the compositional generation procedure, caption construction follows a recursive structure, which naturally fits the properties of human language. Moreover, the proposed compositional procedure requires less data to train, generalizes better, and yields more diverse captions."
                    }
                ]
            },
            "a12e25d0-6a3a-4e85-930f-7d09543abed7": {
                "pk": "a12e25d0-6a3a-4e85-930f-7d09543abed7",
                "name": "Sherry Yang",
                "collaborators": [
                    "Yilun Du",
                    "Pieter Abbeel",
                    "Dale Schuurmans",
                    "Ofir Nachum",
                    "Jason Wei",
                    "Kamyar Ghasemipour",
                    "Jonathan Tompson",
                    "Leslie Kaelbling"
                ],
                "domain": [
                    "Foundation Model",
                    "Generative Modeling",
                    "Decision Making",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Foundation Models for Decision Making: Problems, Methods, and Opportunities",
                        "abstract": "Foundation models pretrained on diverse data at scale have demonstrated extraordinary capabilities in a wide range of vision and language tasks. When such models are deployed in real world environments, they inevitably interface with other entities and agents. For example, language models are often used to interact with human beings through dialogue, and visual perception models are used to autonomously navigate neighborhood streets. In response to these developments, new paradigms are emerging for training foundation models to interact with other agents and perform long-term reasoning. These paradigms leverage the existence of ever-larger datasets curated for multimodal, multitask, and generalist interaction. Research at the intersection of foundation models and decision making holds tremendous promise for creating powerful new systems that can interact effectively across a diverse range of applications such as dialogue, autonomous driving, healthcare, education, and robotics. In this manuscript, we examine the scope of foundation models for decision making, and provide conceptual tools and technical background for understanding the problem space and exploring new research directions. We review recent approaches that ground foundation models in practical decision making applications through a variety of methods such as prompting, conditional generative modeling, planning, optimal control, and reinforcement learning, and discuss common challenges and open problems in the field."
                    },
                    {
                        "title": "Learning Interactive Real-World Simulators",
                        "abstract": "Generative models trained on internet data have revolutionized how text, image, and video content can be created. Perhaps the next milestone for generative models is to simulate realistic experience in response to actions taken by humans, robots, and other interactive agents. Applications of a real-world simulator range from controllable content creation in games and movies, to training embodied agents purely in simulation that can be directly deployed in the real world. We explore the possibility of learning a universal simulator (UniSim) of real-world interaction through generative modeling. We first make the important observation that natural datasets available for learning a real-world simulator are often rich along different dimensions (e.g., abundant objects in image data, densely sampled actions in robotics data, and diverse movements in navigation data). With careful orchestration of diverse datasets, each providing a different aspect of the overall experience, we can simulate the visual outcome of both high-level instructions such as \"open the drawer\" and low-level controls from otherwise static scenes and objects. We use the simulator to train both high-level vision-language policies and low-level reinforcement learning policies, each of which can be deployed in the real world in zero shot after training purely in simulation. We also show that other types of intelligence such as video captioning models can benefit from training with simulated experience, opening up even wider applications. Video demos can be found at https://universal-simulator.github.io."
                    }
                ]
            },
            "43622628-7670-4ca7-8a65-7464679994da": {
                "pk": "43622628-7670-4ca7-8a65-7464679994da",
                "name": "Azade Nova",
                "collaborators": [
                    "Hanie Sedghi",
                    "Bernd Bohnet",
                    "Hanjun Dai",
                    "Dale Schuurmans",
                    "Hugo Larochelle",
                    "Aaron T Parisi",
                    "Kevin Swersky",
                    "Noah Fiedel",
                    "Ian P. Swift",
                    "Sana Ebrahimi"
                ],
                "domain": [
                    "Large Language Models",
                    "Inference Optimization",
                    "Algorithmic Reasoning",
                    "Planning"
                ],
                "publications": [
                    {
                        "title": "Gradient-Free Structured Pruning with Unlabeled Data",
                        "abstract": "Large Language Models (LLMs) have achieved great success in solving difficult tasks across many domains, but such success comes with a high computation cost, and inference latency. As developers and third parties customize these models, the need to provide efficient inference has increased. Many efforts have attempted to reduce inference cost through model compression techniques such as pruning and distillation. However, these techniques either require labeled data, or are time-consuming as they require the compressed model to be retrained to regain accuracy. In this paper, we propose a gradient-free structured pruning framework that uses only unlabeled data. An evaluation on the GLUE and SQuAD benchmarks using BERT$_{BASE}$ and DistilBERT illustrates the effectiveness of the proposed approach. By only using the weights of the pre-trained model and unlabeled data, in a matter of a few minutes on a single GPU, up to 40% of the original FLOP count can be reduced with less than a 4% accuracy loss across all tasks considered."
                    },
                    {
                        "title": "Maximizing Fair Content Spread via Edge Suggestion in Social Networks",
                        "abstract": "Content spread inequity is a potential unfairness issue in online social networks, disparately impacting minority groups. In this paper, we view friendship suggestion, a common feature in social network platforms, as an opportunity to achieve an equitable spread of content. In particular, we propose to suggest a subset of potential edges (currently not existing in the network but likely to be accepted) that maximizes content spread while achieving fairness. Instead of re-engineering the existing systems, our proposal builds a fairness wrapper on top of the existing friendship suggestion components.   We prove the problem is NP-hard and inapproximable in polynomial time unless P = NP. Therefore, allowing relaxation of the fairness constraint, we propose an algorithm based on LP-relaxation and randomized rounding with fixed approximation ratios on fairness and content spread. We provide multiple optimizations, further improving the performance of our algorithm in practice. Besides, we propose a scalable algorithm that dynamically adds subsets of nodes, chosen via iterative sampling, and solves smaller problems corresponding to these nodes. Besides theoretical analysis, we conduct comprehensive experiments on real and synthetic data sets. Across different settings, our algorithms found solutions with nearzero unfairness while significantly increasing the content spread. Our scalable algorithm could process a graph with half a million nodes on a single machine, reducing the unfairness to around 0.0004 while lifting content spread by 43%."
                    },
                    {
                        "title": "Teaching Algorithmic Reasoning via In-context Learning",
                        "abstract": "Large language models (LLMs) have shown increasing in-context learning capabilities through scaling up model and data size. Despite this progress, LLMs are still unable to solve algorithmic reasoning problems. While providing a rationale with the final answer has led to further improvements in multi-step reasoning problems, Anil et al. 2022 showed that even simple algorithmic reasoning tasks such as parity are far from solved. In this work, we identify and study four key stages for successfully teaching algorithmic reasoning to LLMs: (1) formulating algorithms as skills, (2) teaching multiple skills simultaneously (skill accumulation), (3) teaching how to combine skills (skill composition) and (4) teaching how to use skills as tools. We show that it is possible to teach algorithmic reasoning to LLMs via in-context learning, which we refer to as algorithmic prompting. We evaluate our approach on a variety of arithmetic and quantitative reasoning tasks, and demonstrate significant boosts in performance over existing prompting techniques. In particular, for long parity, addition, multiplication and subtraction, we achieve an error reduction of approximately 10x, 9x, 5x and 2x respectively compared to the best available baselines."
                    },
                    {
                        "title": "Exploring and Benchmarking the Planning Capabilities of Large Language Models",
                        "abstract": "We seek to elevate the planning capabilities of Large Language Models (LLMs)investigating four main directions. First, we construct a comprehensive benchmark suite encompassing both classical planning domains and natural language scenarios. This suite includes algorithms to generate instances with varying levels of difficulty, allowing for rigorous and systematic evaluation of LLM performance. Second, we investigate the use of in-context learning (ICL) to enhance LLM planning, exploring the direct relationship between increased context length and improved planning performance. Third, we demonstrate the positive impact of fine-tuning LLMs on optimal planning paths, as well as the effectiveness of incorporating model-driven search procedures. Finally, we investigate the performance of the proposed methods in out-of-distribution scenarios, assessing the ability to generalize to novel and unseen planning challenges."
                    },
                    {
                        "title": "Long-Span Question-Answering: Automatic Question Generation and QA-System Ranking via Side-by-Side Evaluation",
                        "abstract": "We explore the use of long-context capabilities in large language models to create synthetic reading comprehension data from entire books. Previous efforts to construct such datasets relied on crowd-sourcing, but the emergence of transformers with a context size of 1 million or more tokens now enables entirely automatic approaches. Our objective is to test the capabilities of LLMs to analyze, understand, and reason over problems that require a detailed comprehension of long spans of text, such as questions involving character arcs, broader themes, or the consequences of early actions later in the story. We propose a holistic pipeline for automatic data generation including question generation, answering, and model scoring using an ``Evaluator''. We find that a relative approach, comparing answers between models in a pairwise fashion and ranking with a Bradley-Terry model, provides a more consistent and differentiating scoring mechanism than an absolute scorer that rates answers individually. We also show that LLMs from different model families produce moderate agreement in their ratings. We ground our approach using the manually curated NarrativeQA dataset, where our evaluator shows excellent agreement with human judgement and even finds errors in the dataset. Using our automatic evaluation approach, we show that using an entire book as context produces superior reading comprehension performance compared to baseline no-context (parametric knowledge only) and retrieval-based approaches."
                    },
                    {
                        "title": "NATURAL PLAN: Benchmarking LLMs on Natural Language Planning",
                        "abstract": "We introduce NATURAL PLAN, a realistic planning benchmark in natural language containing 3 key tasks: Trip Planning, Meeting Planning, and Calendar Scheduling. We focus our evaluation on the planning capabilities of LLMs with full information on the task, by providing outputs from tools such as Google Flights, Google Maps, and Google Calendar as contexts to the models. This eliminates the need for a tool-use environment for evaluating LLMs on Planning. We observe that NATURAL PLAN is a challenging benchmark for state of the art models. For example, in Trip Planning, GPT-4 and Gemini 1.5 Pro could only achieve 31.1% and 34.8% solve rate respectively. We find that model performance drops drastically as the complexity of the problem increases: all models perform below 5% when there are 10 cities, highlighting a significant gap in planning in natural language for SoTA LLMs. We also conduct extensive ablation studies on NATURAL PLAN to further shed light on the (in)effectiveness of approaches such as self-correction, few-shot generalization, and in-context planning with long-contexts on improving LLM planning."
                    },
                    {
                        "title": "Many-Shot In-Context Learning",
                        "abstract": "Large language models (LLMs) excel at few-shot in-context learning (ICL) -- learning from a few examples provided in context at inference, without any weight updates. Newly expanded context windows allow us to investigate ICL with hundreds or thousands of examples -- the many-shot regime. Going from few-shot to many-shot, we observe significant performance gains across a wide variety of generative and discriminative tasks. While promising, many-shot ICL can be bottlenecked by the available amount of human-generated examples. To mitigate this limitation, we explore two new settings: Reinforced and Unsupervised ICL. Reinforced ICL uses model-generated chain-of-thought rationales in place of human examples. Unsupervised ICL removes rationales from the prompt altogether, and prompts the model only with domain-specific questions. We find that both Reinforced and Unsupervised ICL can be quite effective in the many-shot regime, particularly on complex reasoning tasks. Finally, we demonstrate that, unlike few-shot learning, many-shot learning is effective at overriding pretraining biases, can learn high-dimensional functions with numerical inputs, and performs comparably to fine-tuning. We also find that inference cost increases linearly in the many-shot regime, and frontier LLMs benefit from many-shot ICL to varying degrees. Our analysis also reveals the limitations of next-token prediction loss as an indicator of downstream ICL performance."
                    }
                ]
            },
            "569a4161-611a-46f6-9565-a1d63984700b": {
                "pk": "569a4161-611a-46f6-9565-a1d63984700b",
                "name": "Pengcheng Yin",
                "collaborators": [
                    "Graham Neubig",
                    "Chunting Zhou",
                    "Junxian He",
                    "John Wieting",
                    "Avirup Sil",
                    "Zhengdong Lu",
                    "Hang Li",
                    "Ben Kao",
                    "Xinyi Wang",
                    "Hieu Pham"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Semantic Parsing",
                    "Code Generation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "A Syntactic Neural Model for General-Purpose Code Generation",
                        "abstract": "We consider the problem of parsing natural language descriptions into source code written in a general-purpose programming language like Python. Existing data-driven methods treat this problem as a language generation task without considering the underlying syntax of the target programming language. Informed by previous work in semantic parsing, in this paper we propose a novel neural architecture powered by a grammar model to explicitly capture the target syntax as prior knowledge. Experiments find this an effective way to scale up to generation of complex programs from natural language descriptions, achieving state-of-the-art results that well outperform previous code generation and semantic parsing approaches."
                    },
                    {
                        "title": "TRANX: A Transition-based Neural Abstract Syntax Parser for Semantic Parsing and Code Generation",
                        "abstract": "We present TRANX, a transition-based neural semantic parser that maps natural language (NL) utterances into formal meaning representations (MRs). TRANX uses a transition system based on the abstract syntax description language for the target MR, which gives it two major advantages: (1) it is highly accurate, using information from the syntax of the target MR to constrain the output space and model the information flow, and (2) it is highly generalizable, and can easily be applied to new types of MR by just writing a new abstract syntax description corresponding to the allowable structures in the MR. Experiments on four different semantic parsing and code generation tasks show that our system is generalizable, extensible, and effective, registering strong results compared to existing neural semantic parsers."
                    },
                    {
                        "title": "StructVAE: Tree-structured Latent Variable Models for Semi-supervised Semantic Parsing",
                        "abstract": "Semantic parsing is the task of transducing natural language (NL) utterances into formal meaning representations (MRs), commonly represented as tree structures. Annotating NL utterances with their corresponding MRs is expensive and time-consuming, and thus the limited availability of labeled data often becomes the bottleneck of data-driven, supervised models. We introduce StructVAE, a variational auto-encoding model for semisupervised semantic parsing, which learns both from limited amounts of parallel data, and readily-available unlabeled NL utterances. StructVAE models latent MRs not observed in the unlabeled data as tree-structured latent variables. Experiments on semantic parsing on the ATIS domain and Python code generation show that with extra unlabeled data, StructVAE outperforms strong supervised models."
                    },
                    {
                        "title": "On The Ingredients of an Effective Zero-shot Semantic Parser",
                        "abstract": "Semantic parsers map natural language utterances into meaning representations (e.g., programs). Such models are typically bottlenecked by the paucity of training data due to the required laborious annotation efforts. Recent studies have performed zero-shot learning by synthesizing training examples of canonical utterances and programs from a grammar, and further paraphrasing these utterances to improve linguistic diversity. However, such synthetic examples cannot fully capture patterns in real data. In this paper we analyze zero-shot parsers through the lenses of the language and logical gaps (Herzig and Berant, 2019), which quantify the discrepancy of language and programmatic patterns between the canonical examples and real-world user-issued ones. We propose bridging these gaps using improved grammars, stronger paraphrasers, and efficient learning methods using canonical examples that most likely reflect real user intents. Our model achieves strong performance on two semantic parsing benchmarks (Scholar, Geo) with zero labeled data."
                    },
                    {
                        "title": "Neural Enquirer: Learning to Query Tables with Natural Language",
                        "abstract": "We proposed Neural Enquirer as a neural network architecture to execute a natural language (NL) query on a knowledge-base (KB) for answers. Basically, Neural Enquirer finds the distributed representation of a query and then executes it on knowledge-base tables to obtain the answer as one of the values in the tables. Unlike similar efforts in end-to-end training of semantic parsers, Neural Enquirer is fully \"neuralized\": it not only gives distributional representation of the query and the knowledge-base, but also realizes the execution of compositional queries as a series of differentiable operations, with intermediate results (consisting of annotations of the tables at different levels) saved on multiple layers of memory. Neural Enquirer can be trained with gradient descent, with which not only the parameters of the controlling components and semantic parsing component, but also the embeddings of the tables and query words can be learned from scratch. The training can be done in an end-to-end fashion, but it can take stronger guidance, e.g., the step-by-step supervision for complicated queries, and benefit from it. Neural Enquirer is one step towards building neural network systems which seek to understand language by executing it on real-world. Our experiments show that Neural Enquirer can learn to execute fairly complicated NL queries on tables with rich structures."
                    },
                    {
                        "title": "A Tree-based Decoder for Neural Machine Translation",
                        "abstract": "Recent advances in Neural Machine Translation (NMT) show that adding syntactic information to NMT systems can improve the quality of their translations. Most existing work utilizes some specific types of linguistically-inspired tree structures, like constituency and dependency parse trees. This is often done via a standard RNN decoder that operates on a linearized target tree structure. However, it is an open question of what specific linguistic formalism, if any, is the best structural representation for NMT. In this paper, we (1) propose an NMT model that can naturally generate the topology of an arbitrary tree structure on the target side, and (2) experiment with various target tree structures. Our experiments show the surprising result that our model delivers the best improvements with balanced binary trees constructed without any linguistic knowledge; this model outperforms standard seq2seq models by up to 2.1 BLEU points, and other methods for incorporating target-side syntax by up to 0.7 BLEU."
                    },
                    {
                        "title": "Improving Open Information Extraction via Iterative Rank-Aware Learning",
                        "abstract": "Open information extraction (IE) is the task of extracting open-domain assertions from natural language sentences. A key step in open IE is confidence modeling, ranking the extractions based on their estimated quality to adjust precision and recall of extracted assertions. We found that the extraction likelihood, a confidence measure used by current supervised open IE systems, is not well calibrated when comparing the quality of assertions extracted from different sentences. We propose an additional binary classification loss to calibrate the likelihood to make it more globally comparable, and an iterative learning process, where extractions generated by the open IE model are incrementally included as training samples to help the model learn from trial and error. Experiments on OIE2016 demonstrate the effectiveness of our method. Code and data are available at https://github.com/jzbjyb/oie_rank."
                    },
                    {
                        "title": "Procedures as Programs: Hierarchical Control of Situated Agents through Natural Language",
                        "abstract": "When humans conceive how to perform a particular task, they do so hierarchically: splitting higher-level tasks into smaller sub-tasks. However, in the literature on natural language (NL) command of situated agents, most works have treated the procedures to be executed as flat sequences of simple actions, or any hierarchies of procedures have been shallow at best. In this paper, we propose a formalism of procedures as programs, a powerful yet intuitive method of representing hierarchical procedural knowledge for agent command and control. We further propose a modeling paradigm of hierarchical modular networks, which consist of a planner and reactors that convert NL intents to predictions of executable programs and probe the environment for information necessary to complete the program execution. We instantiate this framework on the IQA and ALFRED datasets for NL instruction following. Our model outperforms reactive baselines by a large margin on both datasets. We also demonstrate that our framework is more data-efficient, and that it allows for fast iterative development."
                    },
                    {
                        "title": "Unsupervised Evaluation of Code LLMs with Round-Trip Correctness",
                        "abstract": "To evaluate code large language models (LLMs), research has relied on a few small manually curated benchmarks, such as HumanEval and MBPP, which represent a narrow part of the real-world software domains. In this work, we introduce round-trip correctness (RTC) as an alternative evaluation method. RTC allows Code LLM evaluation on a broader spectrum of real-world software domains without the need for costly human curation. RTC rests on the idea that we can ask a model to make a prediction (e.g., describe some code using natural language), feed that prediction back (e.g., synthesize code from the predicted description), and check if this round-trip leads to code that is semantically equivalent to the original input. We show how to employ RTC to evaluate code synthesis and editing. We find that RTC strongly correlates with model performance on existing narrow-domain code synthesis benchmarks while allowing us to expand to a much broader set of domains and tasks which was not previously possible without costly human annotations."
                    },
                    {
                        "title": "Merging Weak and Active Supervision for Semantic Parsing",
                        "abstract": "A semantic parser maps natural language commands (NLs) from the users to executable meaning representations (MRs), which are later executed in certain environment to obtain user-desired results. The fully-supervised training of such parser requires NL/MR pairs, annotated by domain experts, which makes them expensive to collect. However, weakly-supervised semantic parsers are learnt only from pairs of NL and expected execution results, leaving the MRs latent. While weak supervision is cheaper to acquire, learning from this input poses difficulties. It demands that parsers search a large space with a very weak learning signal and it is hard to avoid spurious MRs that achieve the correct answer in the wrong way. These factors lead to a performance gap between parsers trained in weakly- and fully-supervised setting. To bridge this gap, we examine the intersection between weak supervision and active learning, which allows the learner to actively select examples and query for manual annotations as extra supervision to improve the model trained under weak supervision. We study different active learning heuristics for selecting examples to query, and various forms of extra supervision for such queries. We evaluate the effectiveness of our method on two different datasets. Experiments on the WikiSQL show that by annotating only 1.8% of examples, we improve over a state-of-the-art weakly-supervised baseline by 6.4%, achieving an accuracy of 79.0%, which is only 1.3% away from the model trained with full supervision. Experiments on WikiTableQuestions with human annotators show that our method can improve the performance with only 100 active queries, especially for weakly-supervised parsers learnt from a cold start."
                    },
                    {
                        "title": "TaBERT: Pretraining for Joint Understanding of Textual and Tabular Data",
                        "abstract": "Recent years have witnessed the burgeoning of pretrained language models (LMs) for text-based natural language (NL) understanding tasks. Such models are typically trained on free-form NL text, hence may not be suitable for tasks like semantic parsing over structured data, which require reasoning over both free-form NL questions and structured tabular data (e.g., database tables). In this paper we present TaBERT, a pretrained LM that jointly learns representations for NL sentences and (semi-)structured tables. TaBERT is trained on a large corpus of 26 million tables and their English contexts. In experiments, neural semantic parsers using TaBERT as feature representation layers achieve new best results on the challenging weakly-supervised semantic parsing benchmark WikiTableQuestions, while performing competitively on the text-to-SQL dataset Spider. Implementation of the model will be available at http://fburl.com/TaBERT ."
                    },
                    {
                        "title": "Learning to Mine Aligned Code and Natural Language Pairs from Stack Overflow",
                        "abstract": "For tasks like code synthesis from natural language, code retrieval, and code summarization, data-driven models have shown great promise. However, creating these models require parallel data between natural language (NL) and code with fine-grained alignments. Stack Overflow (SO) is a promising source to create such a data set: the questions are diverse and most of them have corresponding answers with high-quality code snippets. However, existing heuristic methods (e.g., pairing the title of a post with the code in the accepted answer) are limited both in their coverage and the correctness of the NL-code pairs obtained. In this paper, we propose a novel method to mine high-quality aligned data from SO using two sets of features: hand-crafted features considering the structure of the extracted snippets, and correspondence features obtained by training a probabilistic model to capture the correlation between NL and code using neural networks. These features are fed into a classifier that determines the quality of mined NL-code pairs. Experiments using Python and Java as test beds show that the proposed method greatly expands coverage and accuracy over existing mining methods, even when using only a small number of labeled examples. Further, we find that reasonable results are achieved even when training the classifier on one language and testing on another, showing promise for scaling NL-code mining to a wide variety of programming languages beyond those for which we are able to annotate data."
                    }
                ]
            },
            "4c7c4c09-2ded-498f-8754-96d527b4deae": {
                "pk": "4c7c4c09-2ded-498f-8754-96d527b4deae",
                "name": "Phitchaya Mangpo Phothilimthana",
                "collaborators": [
                    "Sudip Roy",
                    "Charith Mendis",
                    "Yanqi Zhou",
                    "Mike Burrows",
                    "Azalia Mirhoseini",
                    "James Laudon",
                    "Kaidi Cao",
                    "Sami Abu-El-Haija",
                    "Bryan Perozzi",
                    "Charles Jin"
                ],
                "domain": [
                    "Neural Architecture Search",
                    "Machine Learning",
                    "Graph Neural Network",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Neural Architecture Search using Property Guided Synthesis",
                        "abstract": "In the past few years, neural architecture search (NAS) has become an increasingly important tool within the deep learning community. Despite the many recent successes of NAS, however, most existing approaches operate within highly structured design spaces, and hence explore only a small fraction of the full search space of neural architectures while also requiring significant manual effort from domain experts. In this work, we develop techniques that enable efficient NAS in a significantly larger design space. To accomplish this, we propose to perform NAS in an abstract search space of program properties. Our key insights are as follows: (1) the abstract search space is significantly smaller than the original search space, and (2) architectures with similar program properties also have similar performance; thus, we can search more efficiently in the abstract search space. To enable this approach, we also propose a novel efficient synthesis procedure, which accepts a set of promising program properties, and returns a satisfying neural architecture. We implement our approach, $\\alpha$NAS, within an evolutionary framework, where the mutations are guided by the program properties. Starting with a ResNet-34 model, $\\alpha$NAS produces a model with slightly improved accuracy on CIFAR-10 but 96% fewer parameters. On ImageNet, $\\alpha$NAS is able to improve over Vision Transformer (30% fewer FLOPS and parameters), ResNet-50 (23% fewer FLOPS, 14% fewer parameters), and EfficientNet (7% fewer FLOPS and parameters) without any degradation in accuracy."
                    },
                    {
                        "title": "GRANITE: A Graph Neural Network Model for Basic Block Throughput Estimation",
                        "abstract": "Analytical hardware performance models yield swift estimation of desired hardware performance metrics. However, developing these analytical models for modern processors with sophisticated microarchitectures is an extremely laborious task and requires a firm understanding of target microarchitecture's internal structure. In this paper, we introduce GRANITE, a new machine learning model that estimates the throughput of basic blocks across different microarchitectures. GRANITE uses a graph representation of basic blocks that captures both structural and data dependencies between instructions. This representation is processed using a graph neural network that takes advantage of the relational information captured in the graph and learns a rich neural representation of the basic block that allows more precise throughput estimation. Our results establish a new state-of-the-art for basic block performance estimation with an average test error of 6.9% across a wide range of basic blocks and microarchitectures for the x86-64 target. Compared to recent work, this reduced the error by 1.7% while improving training and inference throughput by approximately 3.0x. In addition, we propose the use of multi-task learning with independent multi-layer feed forward decoder networks. Our results show that this technique further improves precision of all learned models while significantly reducing per-microarchitecture training costs. We perform an extensive set of ablation studies and comparisons with prior work, concluding a set of methods to achieve high accuracy for basic block performance estimation."
                    },
                    {
                        "title": "A Learned Performance Model for Tensor Processing Units",
                        "abstract": "Accurate hardware performance models are critical to efficient code generation. They can be used by compilers to make heuristic decisions, by superoptimizers as a minimization objective, or by autotuners to find an optimal configuration for a specific program. However, they are difficult to develop because contemporary processors are complex, and the recent proliferation of deep learning accelerators has increased the development burden. We demonstrate a method of learning performance models from a corpus of tensor computation graph programs for Tensor Processing Unit (TPU) instances. We show that our learned model outperforms a heavily-optimized analytical performance model on two tasks -- tile-size selection and operator fusion -- and that it helps an autotuner discover faster programs in a setting where access to TPUs is limited or expensive."
                    },
                    {
                        "title": "A Transferable Approach for Partitioning Machine Learning Models on Multi-Chip-Modules",
                        "abstract": "Multi-Chip-Modules (MCMs) reduce the design and fabrication cost of machine learning (ML) accelerators while delivering performance and energy efficiency on par with a monolithic large chip. However, ML compilers targeting MCMs need to solve complex optimization problems optimally and efficiently to achieve this high performance. One such problem is the multi-chip partitioning problem where compilers determine the optimal partitioning and placement of operations in tensor computation graphs on chiplets in MCMs. Partitioning ML graphs for MCMs is particularly hard as the search space grows exponentially with the number of chiplets available and the number of nodes in the neural network. Furthermore, the constraints imposed by the underlying hardware produce a search space where valid solutions are extremely sparse. In this paper, we present a strategy using a deep reinforcement learning (RL) framework to emit a possibly invalid candidate partition that is then corrected by a constraint solver. Using the constraint solver ensures that RL encounters valid solutions in the sparse space frequently enough to converge with fewer samples as compared to non-learned strategies. The architectural choices we make for the policy network allow us to generalize across different ML graphs. Our evaluation of a production-scale model, BERT, on real hardware reveals that the partitioning generated using RL policy achieves 6.11% and 5.85% higher throughput than random search and simulated annealing. In addition, fine-tuning the pre-trained RL policy reduces the search time from 3 hours to only 9 minutes, while achieving the same throughput as training RL policy from scratch."
                    },
                    {
                        "title": "Learning Large Graph Property Prediction via Graph Segment Training",
                        "abstract": "Learning to predict properties of large graphs is challenging because each prediction requires the knowledge of an entire graph, while the amount of memory available during training is bounded. Here we propose Graph Segment Training (GST), a general framework that utilizes a divide-and-conquer approach to allow learning large graph property prediction with a constant memory footprint. GST first divides a large graph into segments and then backpropagates through only a few segments sampled per training iteration. We refine the GST paradigm by introducing a historical embedding table to efficiently obtain embeddings for segments not sampled for backpropagation. To mitigate the staleness of historical embeddings, we design two novel techniques. First, we finetune the prediction head to fix the input distribution shift. Second, we introduce Stale Embedding Dropout to drop some stale embeddings during training to reduce bias. We evaluate our complete method GST-EFD (with all the techniques together) on two large graph property prediction benchmarks: MalNet and TpuGraphs. Our experiments show that GST-EFD is both memory-efficient and fast, while offering a slight boost on test accuracy over a typical full graph training regime."
                    },
                    {
                        "title": "TpuGraphs: A Performance Prediction Dataset on Large Tensor Computational Graphs",
                        "abstract": "Precise hardware performance models play a crucial role in code optimizations. They can assist compilers in making heuristic decisions or aid autotuners in identifying the optimal configuration for a given program. For example, the autotuner for XLA, a machine learning compiler, discovered 10-20% speedup on state-of-the-art models serving substantial production traffic at Google. Although there exist a few datasets for program performance prediction, they target small sub-programs such as basic blocks or kernels. This paper introduces TpuGraphs, a performance prediction dataset on full tensor programs, represented as computational graphs, running on Tensor Processing Units (TPUs). Each graph in the dataset represents the main computation of a machine learning workload, e.g., a training epoch or an inference step. Each data sample contains a computational graph, a compilation configuration, and the execution time of the graph when compiled with the configuration. The graphs in the dataset are collected from open-source machine learning programs, featuring popular model architectures, e.g., ResNet, EfficientNet, Mask R-CNN, and Transformer. TpuGraphs provides 25x more graphs than the largest graph property prediction dataset (with comparable graph sizes), and 770x larger graphs on average compared to existing performance prediction datasets on machine learning programs. This graph-level prediction task on large graphs introduces new challenges in learning, ranging from scalability, training efficiency, to model quality."
                    },
                    {
                        "title": "Accelerating Retrieval-Augmented Language Model Serving with Speculation",
                        "abstract": "Retrieval-augmented language models (RaLM) have demonstrated the potential to solve knowledge-intensive natural language processing (NLP) tasks by combining a non-parametric knowledge base with a parametric language model. Instead of fine-tuning a fully parametric model, RaLM excels at its low-cost adaptation to the latest data and better source attribution mechanisms. Among various RaLM approaches, iterative RaLM delivers a better generation quality due to a more frequent interaction between the retriever and the language model. Despite the benefits, iterative RaLM usually encounters high overheads due to the frequent retrieval step. To this end, we propose RaLMSpec, a speculation-inspired framework that provides generic speed-up over iterative RaLM while preserving the same model outputs through speculative retrieval and batched verification. By further incorporating prefetching, optimal speculation stride scheduler, and asynchronous verification, RaLMSpec can automatically exploit the acceleration potential to the fullest. For naive iterative RaLM serving, extensive evaluations over three language models on four downstream QA datasets demonstrate that RaLMSpec can achieve a speed-up ratio of 1.75-2.39x, 1.04-1.39x, and 1.31-1.77x when the retriever is an exact dense retriever, approximate dense retriever, and sparse retriever respectively compared with the baseline. For KNN-LM serving, RaLMSpec can achieve a speed-up ratio up to 7.59x and 2.45x when the retriever is an exact dense retriever and approximate dense retriever, respectively, compared with the baseline."
                    },
                    {
                        "title": "Transferable Graph Optimizers for ML Compilers",
                        "abstract": "Most compilers for machine learning (ML) frameworks need to solve many correlated optimization problems to generate efficient machine code. Current ML compilers rely on heuristics based algorithms to solve these optimization problems one at a time. However, this approach is not only hard to maintain but often leads to sub-optimal solutions especially for newer model architectures. Existing learning based approaches in the literature are sample inefficient, tackle a single optimization problem, and do not generalize to unseen graphs making them infeasible to be deployed in practice. To address these limitations, we propose an end-to-end, transferable deep reinforcement learning method for computational graph optimization (GO), based on a scalable sequential attention mechanism over an inductive graph neural network. GO generates decisions on the entire graph rather than on each individual node autoregressively, drastically speeding up the search compared to prior methods. Moreover, we propose recurrent attention layers to jointly optimize dependent graph optimization tasks and demonstrate 33%-60% speedup on three graph optimization tasks compared to TensorFlow default optimization. On a diverse set of representative graphs consisting of up to 80,000 nodes, including Inception-v3, Transformer-XL, and WaveNet, GO achieves on average 21% improvement over human experts and 18% improvement over the prior state of the art with 15x faster convergence, on a device placement task evaluated in real systems."
                    },
                    {
                        "title": "Optimizing Memory Mapping Using Deep Reinforcement Learning",
                        "abstract": "Resource scheduling and allocation is a critical component of many high impact systems ranging from congestion control to cloud computing. Finding more optimal solutions to these problems often has significant impact on resource and time savings, reducing device wear-and-tear, and even potentially improving carbon emissions. In this paper, we focus on a specific instance of a scheduling problem, namely the memory mapping problem that occurs during compilation of machine learning programs: That is, mapping tensors to different memory layers to optimize execution time.   We introduce an approach for solving the memory mapping problem using Reinforcement Learning. RL is a solution paradigm well-suited for sequential decision making problems that are amenable to planning, and combinatorial search spaces with high-dimensional data inputs. We formulate the problem as a single-player game, which we call the mallocGame, such that high-reward trajectories of the game correspond to efficient memory mappings on the target hardware. We also introduce a Reinforcement Learning agent, mallocMuZero, and show that it is capable of playing this game to discover new and improved memory mapping solutions that lead to faster execution times on real ML workloads on ML accelerators. We compare the performance of mallocMuZero to the default solver used by the Accelerated Linear Algebra (XLA) compiler on a benchmark of realistic ML workloads. In addition, we show that mallocMuZero is capable of improving the execution time of the recently published AlphaTensor matrix multiplication model."
                    }
                ]
            },
            "2a8c9148-3808-448e-a768-ce21b2d7e2ba": {
                "pk": "2a8c9148-3808-448e-a768-ce21b2d7e2ba",
                "name": "Charles Sutton",
                "collaborators": [
                    "Andrew McCallum",
                    "Miltiadis Allamanis",
                    "Jaroslav Fowkes",
                    "Akash Srivastava",
                    "Quim Castella",
                    "Yichuan Zhang",
                    "Linan Gong",
                    "Augustus Odena",
                    "Michael I. Jordan",
                    "Krzysztof J. Geras"
                ],
                "domain": [
                    "Structured Prediction",
                    "Bayesian Inference",
                    "Natural Language Processing",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "An Introduction to Conditional Random Fields",
                        "abstract": "Often we wish to predict a large number of variables that depend on each other as well as on other observed variables. Structured prediction methods are essentially a combination of classification and graphical modeling, combining the ability of graphical models to compactly model multivariate data with the ability of classification methods to perform prediction using large sets of input features. This tutorial describes conditional random fields, a popular probabilistic method for structured prediction. CRFs have seen wide application in natural language processing, computer vision, and bioinformatics. We describe methods for inference and parameter estimation for CRFs, including practical issues for implementing large scale CRFs. We do not assume previous knowledge of graphical modeling, so this tutorial is intended to be useful to practitioners in a wide variety of fields."
                    },
                    {
                        "title": "Improved Dynamic Schedules for Belief Propagation",
                        "abstract": "Belief propagation and its variants are popular methods for approximate inference, but their running time and even their convergence depend greatly on the schedule used to send the messages. Recently, dynamic update schedules have been shown to converge much faster on hard networks than static schedules, namely the residual BP schedule of Elidan et al. [2006]. But that RBP algorithm wastes message updates: many messages are computed solely to determine their priority, and are never actually performed. In this paper, we show that estimating the residual, rather than calculating it directly, leads to significant decreases in the number of messages required for convergence, and in the total running time. The residual is estimated using an upper bound based on recent work on message errors in BP. On both synthetic and real-world networks, this dramatically decreases the running time of BP, in some cases by a factor of five, without affecting the quality of the solution."
                    },
                    {
                        "title": "Piecewise Training for Undirected Models",
                        "abstract": "For many large undirected models that arise in real-world applications, exact maximumlikelihood training is intractable, because it requires computing marginal distributions of the model. Conditional training is even more difficult, because the partition function depends not only on the parameters, but also on the observed input, requiring repeated inference over each training example. An appealing idea for such models is to independently train a local undirected classifier over each clique, afterwards combining the learned weights into a single global model. In this paper, we show that this piecewise method can be justified as minimizing a new family of upper bounds on the log partition function. On three natural-language data sets, piecewise training is more accurate than pseudolikelihood, and often performs comparably to global training using belief propagation."
                    },
                    {
                        "title": "Word Storms: Multiples of Word Clouds for Visual Comparison of Documents",
                        "abstract": "Word clouds are a popular tool for visualizing documents, but they are not a good tool for comparing documents, because identical words are not presented consistently across different clouds. We introduce the concept of word storms, a visualization tool for analysing corpora of documents. A word storm is a group of word clouds, in which each cloud represents a single document, juxtaposed to allow the viewer to compare and contrast the documents. We present a novel algorithm that creates a coordinated word storm, in which words that appear in multiple documents are placed in the same location, using the same color and orientation, in all of the corresponding clouds. In this way, similar documents are represented by similar-looking word clouds, making them easier to compare and contrast visually. We evaluate the algorithm in two ways: first, an automatic evaluation based on document classification; and second, a user study. The results confirm that unlike standard word clouds, a coordinated word storm better allows for visual comparison of documents."
                    },
                    {
                        "title": "Semi-Separable Hamiltonian Monte Carlo for Inference in Bayesian Hierarchical Models",
                        "abstract": "Sampling from hierarchical Bayesian models is often difficult for MCMC methods, because of the strong correlations between the model parameters and the hyperparameters. Recent Riemannian manifold Hamiltonian Monte Carlo (RMHMC) methods have significant potential advantages in this setting, but are computationally expensive. We introduce a new RMHMC method, which we call semi-separable Hamiltonian Monte Carlo, which uses a specially designed mass matrix that allows the joint Hamiltonian over model parameters and hyperparameters to decompose into two simpler Hamiltonians. This structure is exploited by a new integrator which we call the alternating blockwise leapfrog algorithm. The resulting method can mix faster than simpler Gibbs sampling while being simpler and more efficient than previous instances of RMHMC."
                    },
                    {
                        "title": "Popularity of arXiv.org within Computer Science",
                        "abstract": "It may seem surprising that, out of all areas of science, computer scientists have been slow to post electronic versions of papers on sites like arXiv.org. Instead, computer scientists have tended to place papers on our individual home pages, but this loses the benefits of aggregation, namely notification and browsing.   But this is changing. More and more computer scientists are now using the arXiv. At the same time, there is ongoing discussion and controversy about how prepublication affects peer review, especially for double-blind conferences. This discussion is often carried out with precious little evidence of how popular prepublication is.   We measure what percentage of papers in computer science are placed on the arXiv, by cross-referencing published papers in DBLP with e-prints on arXiv. We found:   * Usage of arXiv.org has risen dramatically among the most selective conferences in computer science. In 2017, fully 23% of papers had e-prints on arXiv, compared to only 1% ten years ago.   * Areas of computer science vary widely in e-print prevalence. In theoretical computer science and machine learning, over 60% of published papers are on arXiv, while other areas are essentially zero. In most areas, arXiv usage is rising.   * Many researchers use arXiv for posting preprints. Of the 2017 published papers with arXiv e-prints, 56% were preprints that were posted before or during peer review.   Our paper describes these results as well as policy implications for researchers and practitioners."
                    },
                    {
                        "title": "Mining Idioms from Source Code",
                        "abstract": "We present the first method for automatically mining code idioms from a corpus of previously written, idiomatic software projects. We take the view that a code idiom is a syntactic fragment that recurs across projects and has a single semantic role. Idioms may have metavariables, such as the body of a for loop. Modern IDEs commonly provide facilities for manually defining idioms and inserting them on demand, but this does not help programmers to write idiomatic code in languages or using libraries with which they are unfamiliar. We present HAGGIS, a system for mining code idioms that builds on recent advanced techniques from statistical natural language processing, namely, nonparametric Bayesian probabilistic tree substitution grammars. We apply HAGGIS to several of the most popular open source projects from GitHub. We present a wide range of evidence that the resulting idioms are semantically meaningful, demonstrating that they do indeed recur across software projects and that they occur more frequently in illustrative code examples collected from a Q&A site. Manual examination of the most common idioms indicate that they describe important program concepts, including object creation, exception handling, and resource management."
                    },
                    {
                        "title": "A Bayesian Network Model for Interesting Itemsets",
                        "abstract": "Mining itemsets that are the most interesting under a statistical model of the underlying data is a commonly used and well-studied technique for exploratory data analysis, with the most recent interestingness models exhibiting state of the art performance. Continuing this highly promising line of work, we propose the first, to the best of our knowledge, generative model over itemsets, in the form of a Bayesian network, and an associated novel measure of interestingness. Our model is able to efficiently infer interesting itemsets directly from the transaction database using structural EM, in which the E-step employs the greedy approximation to weighted set cover. Our approach is theoretically simple, straightforward to implement, trivially parallelizable and retrieves itemsets whose quality is comparable to, if not better than, existing state of the art algorithms as we demonstrate on several real-world datasets."
                    },
                    {
                        "title": "Parameter-Free Probabilistic API Mining across GitHub",
                        "abstract": "Existing API mining algorithms can be difficult to use as they require expensive parameter tuning and the returned set of API calls can be large, highly redundant and difficult to understand. To address this, we present PAM (Probabilistic API Miner), a near parameter-free probabilistic algorithm for mining the most interesting API call patterns. We show that PAM significantly outperforms both MAPO and UPMiner, achieving 69% test-set precision, at retrieving relevant API call sequences from GitHub. Moreover, we focus on libraries for which the developers have explicitly provided code examples, yielding over 300,000 LOC of hand-written API example code from the 967 client projects in the data set. This evaluation suggests that the hand-written examples actually have limited coverage of real API usages."
                    },
                    {
                        "title": "A Subsequence Interleaving Model for Sequential Pattern Mining",
                        "abstract": "Recent sequential pattern mining methods have used the minimum description length (MDL) principle to define an encoding scheme which describes an algorithm for mining the most compressing patterns in a database. We present a novel subsequence interleaving model based on a probabilistic model of the sequence database, which allows us to search for the most compressing set of patterns without designing a specific encoding scheme. Our proposed algorithm is able to efficiently mine the most relevant sequential patterns and rank them using an associated measure of interestingness. The efficient inference in our model is a direct result of our use of a structural expectation-maximization framework, in which the expectation-step takes the form of a submodular optimization problem subject to a coverage constraint. We show on both synthetic and real world datasets that our model mines a set of sequential patterns with low spuriousness and redundancy, high interpretability and usefulness in real-world applications. Furthermore, we demonstrate that the quality of the patterns from our approach is comparable to, if not better than, existing state of the art sequential pattern mining algorithms."
                    },
                    {
                        "title": "Learning to Represent Programs with Property Signatures",
                        "abstract": "We introduce the notion of property signatures, a representation for programs and program specifications meant for consumption by machine learning algorithms. Given a function with input type $\\tau_{in}$ and output type $\\tau_{out}$, a property is a function of type: $(\\tau_{in}, \\tau_{out}) \\rightarrow \\texttt{Bool}$ that (informally) describes some simple property of the function under consideration. For instance, if $\\tau_{in}$ and $\\tau_{out}$ are both lists of the same type, one property might ask `is the input list the same length as the output list?'. If we have a list of such properties, we can evaluate them all for our function to get a list of outputs that we will call the property signature. Crucially, we can `guess' the property signature for a function given only a set of input/output pairs meant to specify that function. We discuss several potential applications of property signatures and show experimentally that they can be used to improve over a baseline synthesizer so that it emits twice as many programs in less than one-tenth of the time."
                    },
                    {
                        "title": "Autoencoding Variational Inference For Topic Models",
                        "abstract": "Topic models are one of the most popular methods for learning representations of text, but a major challenge is that any change to the topic model requires mathematically deriving a new inference algorithm. A promising approach to address this problem is autoencoding variational Bayes (AEVB), but it has proven diffi- cult to apply to topic models in practice. We present what is to our knowledge the first effective AEVB based inference method for latent Dirichlet allocation (LDA), which we call Autoencoded Variational Inference For Topic Model (AVITM). This model tackles the problems caused for AEVB by the Dirichlet prior and by component collapsing. We find that AVITM matches traditional methods in accuracy with much better inference time. Indeed, because of the inference network, we find that it is unnecessary to pay the computational cost of running variational optimization on test data. Because AVITM is black box, it is readily applied to new topic models. As a dramatic illustration of this, we present a new topic model called ProdLDA, that replaces the mixture model in LDA with a product of experts. By changing only one line of code from LDA, we find that ProdLDA yields much more interpretable topics, even if LDA is trained via collapsed Gibbs sampling."
                    },
                    {
                        "title": "Variational Inference In Pachinko Allocation Machines",
                        "abstract": "The Pachinko Allocation Machine (PAM) is a deep topic model that allows representing rich correlation structures among topics by a directed acyclic graph over topics. Because of the flexibility of the model, however, approximate inference is very difficult. Perhaps for this reason, only a small number of potential PAM architectures have been explored in the literature. In this paper we present an efficient and flexible amortized variational inference method for PAM, using a deep inference network to parameterize the approximate posterior distribution in a manner similar to the variational autoencoder. Our inference method produces more coherent topics than state-of-art inference methods for PAM while being an order of magnitude faster, which allows exploration of a wider range of PAM architectures than have previously been studied."
                    },
                    {
                        "title": "Bayesian inference for queueing networks and modeling of internet services",
                        "abstract": "Modern Internet services, such as those at Google, Yahoo!, and Amazon, handle billions of requests per day on clusters of thousands of computers. Because these services operate under strict performance requirements, a statistical understanding of their performance is of great practical interest. Such services are modeled by networks of queues, where each queue models one of the computers in the system. A key challenge is that the data are incomplete, because recording detailed information about every request to a heavily used system can require unacceptable overhead. In this paper we develop a Bayesian perspective on queueing models in which the arrival and departure times that are not observed are treated as latent variables. Underlying this viewpoint is the observation that a queueing model defines a deterministic transformation between the data and a set of independent variables called the service times. With this viewpoint in hand, we sample from the posterior distribution over missing data and model parameters using Markov chain Monte Carlo. We evaluate our framework on data from a benchmark Web application. We also present a simple technique for selection among nested queueing models. We are unaware of any previous work that considers inference in networks of queues in the presence of missing data."
                    },
                    {
                        "title": "Scheduled denoising autoencoders",
                        "abstract": "We present a representation learning method that learns features at multiple different levels of scale. Working within the unsupervised framework of denoising autoencoders, we observe that when the input is heavily corrupted during training, the network tends to learn coarse-grained features, whereas when the input is only slightly corrupted, the network tends to learn fine-grained features. This motivates the scheduled denoising autoencoder, which starts with a high level of noise that lowers as training progresses. We find that the resulting representation yields a significant boost on a later supervised task compared to the original input, or to a standard denoising autoencoder trained at a single noise level. After supervised fine-tuning our best model achieves the lowest ever reported error on the CIFAR-10 data set among permutation-invariant methods."
                    },
                    {
                        "title": "Latent Bayesian melding for integrating individual and population models",
                        "abstract": "In many statistical problems, a more coarse-grained model may be suitable for population-level behaviour, whereas a more detailed model is appropriate for accurate modelling of individual behaviour. This raises the question of how to integrate both types of models. Methods such as posterior regularization follow the idea of generalized moment matching, in that they allow matching expectations between two models, but sometimes both models are most conveniently expressed as latent variable models. We propose latent Bayesian melding, which is motivated by averaging the distributions over populations statistics of both the individual-level and the population-level models under a logarithmic opinion pool framework. In a case study on electricity disaggregation, which is a type of single-channel blind source separation problem, we show that latent Bayesian melding leads to significantly more accurate predictions than an approach based solely on generalized moment matching."
                    },
                    {
                        "title": "A Convolutional Attention Network for Extreme Summarization of Source Code",
                        "abstract": "Attention mechanisms in neural networks have proved useful for problems in which the input and output do not have fixed dimension. Often there exist features that are locally translation invariant and would be valuable for directing the model's attention, but previous attentional architectures are not constructed to learn such features specifically. We introduce an attentional neural network that employs convolution on the input tokens to detect local time-invariant and long-range topical attention features in a context-dependent way. We apply this architecture to the problem of extreme summarization of source code snippets into short, descriptive function name-like summaries. Using those features, the model sequentially generates a summary by marginalizing over two attention mechanisms: one that predicts the next summary token based on the attention weights of the input tokens and another that is able to copy a code token as-is directly into the summary. We demonstrate our convolutional attention neural network's performance on 10 popular Java projects showing that it achieves better performance compared to previous attentional mechanisms."
                    },
                    {
                        "title": "Clustering with a Reject Option: Interactive Clustering as Bayesian Prior Elicitation",
                        "abstract": "A good clustering can help a data analyst to explore and understand a data set, but what constitutes a good clustering may depend on domain-specific and application-specific criteria. These criteria can be difficult to formalize, even when it is easy for an analyst to know a good clustering when she sees one. We present a new approach to interactive clustering for data exploration, called \\ciif, based on a particularly simple feedback mechanism, in which an analyst can choose to reject individual clusters and request new ones. The new clusters should be different from previously rejected clusters while still fitting the data well. We formalize this interaction in a novel Bayesian prior elicitation framework. In each iteration, the prior is adapted to account for all the previous feedback, and a new clustering is then produced from the posterior distribution. To achieve the computational efficiency necessary for an interactive setting, we propose an incremental optimization method over data minibatches using Lagrangian relaxation. Experiments demonstrate that \\ciif can produce accurate and diverse clusterings."
                    },
                    {
                        "title": "Incremental Sampling Without Replacement for Sequence Models",
                        "abstract": "Sampling is a fundamental technique, and sampling without replacement is often desirable when duplicate samples are not beneficial. Within machine learning, sampling is useful for generating diverse outputs from a trained model. We present an elegant procedure for sampling without replacement from a broad class of randomized programs, including generative neural models that construct outputs sequentially. Our procedure is efficient even for exponentially-large output spaces. Unlike prior work, our approach is incremental, i.e., samples can be drawn one at a time, allowing for increased flexibility. We also present a new estimator for computing expectations from samples drawn without replacement. We show that incremental sampling without replacement is applicable to many domains, e.g., program synthesis and combinatorial optimization."
                    },
                    {
                        "title": "Learning Continuous Semantic Representations of Symbolic Expressions",
                        "abstract": "Combining abstract, symbolic reasoning with continuous neural reasoning is a grand challenge of representation learning. As a step in this direction, we propose a new architecture, called neural equivalence networks, for the problem of learning continuous semantic representations of algebraic and logical expressions. These networks are trained to represent semantic equivalence, even of expressions that are syntactically very different. The challenge is that semantic representations must be computed in a syntax-directed manner, because semantics is compositional, but at the same time, small changes in syntax can lead to very large changes in semantics, which can be difficult for continuous neural architectures. We perform an exhaustive evaluation on the task of checking equivalence on a highly diverse class of symbolic algebraic and boolean expression types, showing that our model significantly outperforms existing architectures."
                    }
                ]
            },
            "ac09ad21-5c1e-41ad-b805-1afaa9bfe9fb": {
                "pk": "ac09ad21-5c1e-41ad-b805-1afaa9bfe9fb",
                "name": "Dale Schuurmans",
                "collaborators": [
                    "Ofir Nachum",
                    "Mohammad Norouzi",
                    "Yuhong Guo",
                    "Xinhua Zhang",
                    "Bo Dai",
                    "Lihong Li",
                    "Finnegan Southey",
                    "Fletcher Lu",
                    "Hengshuai Yao",
                    "Mahdi Karami"
                ],
                "domain": [
                    "Machine Learning",
                    "Reinforcement Learning",
                    "Bayesian Networks",
                    "Graphical Models"
                ],
                "publications": [
                    {
                        "title": "Memory Augmented Large Language Models are Computationally Universal",
                        "abstract": "We show that transformer-based large language models are computationally universal when augmented with an external memory. Any deterministic language model that conditions on strings of bounded length is equivalent to a finite automaton, hence computationally limited. However, augmenting such models with a read-write memory creates the possibility of processing arbitrarily large inputs and, potentially, simulating any algorithm. We establish that an existing large language model, Flan-U-PaLM 540B, can be combined with an associative read-write memory to exactly simulate the execution of a universal Turing machine, $U_{15,2}$. A key aspect of the finding is that it does not require any modification of the language model weights. Instead, the construction relies solely on designing a form of stored instruction computer that can subsequently be programmed with a specific set of prompts."
                    },
                    {
                        "title": "Monte Carlo Inference via Greedy Importance Sampling",
                        "abstract": "We present a new method for conducting Monte Carlo inference in graphical models which combines explicit search with generalized importance sampling. The idea is to reduce the variance of importance sampling by searching for significant points in the target distribution. We prove that it is possible to introduce search and still maintain unbiasedness. We then demonstrate our procedure on a few simple inference tasks and show that it can improve the inference quality of standard MCMC methods, including Gibbs sampling, Metropolis sampling, and Hybrid Monte Carlo. This paper extends previous work which showed how greedy importance sampling could be correctly realized in the one-dimensional case."
                    },
                    {
                        "title": "Convex Structure Learning for Bayesian Networks: Polynomial Feature Selection and Approximate Ordering",
                        "abstract": "We present a new approach to learning the structure and parameters of a Bayesian network based on regularized estimation in an exponential family representation. Here we show that, given a fixed variable order, the optimal structure and parameters can be learned efficiently, even without restricting the size of the parent sets. We then consider the problem of optimizing the variable order for a given set of features. This is still a computationally hard problem, but we present a convex relaxation that yields an optimal 'soft' ordering in polynomial time. One novel aspect of the approach is that we do not perform a discrete search over DAG structures, nor over variable orders, but instead solve a continuous relaxation that can then be rounded to obtain a valid network structure. We conduct an experimental comparison against standard structure search procedures over standard objectives, which cope with local minima, and evaluate the advantages of using convex relaxations that reduce the effects of local minima."
                    },
                    {
                        "title": "Monte Carlo Matrix Inversion Policy Evaluation",
                        "abstract": "In 1950, Forsythe and Leibler (1950) introduced a statistical technique for finding the inverse of a matrix by characterizing the elements of the matrix inverse as expected values of a sequence of random walks. Barto and Duff (1994) subsequently showed relations between this technique and standard dynamic programming and temporal differencing methods. The advantage of the Monte Carlo matrix inversion (MCMI) approach is that it scales better with respect to state-space size than alternative techniques. In this paper, we introduce an algorithm for performing reinforcement learning policy evaluation using MCMI. We demonstrate that MCMI improves on runtime over a maximum likelihood model-based policy evaluation approach and on both runtime and accuracy over the temporal differencing (TD) policy evaluation approach. We further improve on MCMI policy evaluation by adding an importance sampling technique to our algorithm to reduce the variance of our estimator. Lastly, we illustrate techniques for scaling up MCMI to large state spaces in order to perform policy improvement."
                    },
                    {
                        "title": "Reinforcement Ranking",
                        "abstract": "We introduce a new framework for web page ranking -- reinforcement ranking -- that improves the stability and accuracy of Page Rank while eliminating the need for computing the stationary distribution of random walks. Instead of relying on teleportation to ensure a well defined Markov chain, we develop a reverse-time reinforcement learning framework that determines web page authority based on the solution of a reverse Bellman equation. In particular, for a given reward function and surfing policy we recover a well defined authority score from a reverse-time perspective: looking back from a web page, what is the total incoming discounted reward brought by the surfer from the page's predecessors? This results in a novel form of reverse-time dynamic-programming/reinforcement-learning problem that achieves several advantages over Page Rank based methods: First, stochasticity, ergodicity, and irreducibility of the underlying Markov chain is no longer required for well-posedness. Second, the method is less sensitive to graph topology and more stable in the presence of dangling pages. Third, not only does the reverse Bellman iteration yield a more efficient power iteration, it allows for faster updating in the presence of graph changes. Finally, our experiments demonstrate improvements in ranking quality."
                    },
                    {
                        "title": "Variational Inference for Deep Probabilistic Canonical Correlation Analysis",
                        "abstract": "In this paper, we propose a deep probabilistic multi-view model that is composed of a linear multi-view layer based on probabilistic canonical correlation analysis (CCA) description in the latent space together with deep generative networks as observation models. The network is designed to decompose the variations of all views into a shared latent representation and a set of view-specific components where the shared latent representation is intended to describe the common underlying sources of variation among the views. An efficient variational inference procedure is developed that approximates the posterior distributions of the latent probabilistic multi-view layer while taking into account the solution of probabilistic CCA. A generalization to models with arbitrary number of views is also proposed. The empirical studies confirm that the proposed deep generative multi-view model can successfully extend deep variational inference to multi-view learning while it efficiently integrates the relationship between multiple views to alleviate the difficulty of learning."
                    },
                    {
                        "title": "Autoregressive Large Language Models are Computationally Universal",
                        "abstract": "We show that autoregressive decoding of a transformer-based language model can realize universal computation, without external intervention or modification of the model's weights. Establishing this result requires understanding how a language model can process arbitrarily long inputs using a bounded context. For this purpose, we consider a generalization of autoregressive decoding where, given a long input, emitted tokens are appended to the end of the sequence as the context window advances. We first show that the resulting system corresponds to a classical model of computation, a Lag system, that has long been known to be computationally universal. By leveraging a new proof, we show that a universal Turing machine can be simulated by a Lag system with 2027 production rules. We then investigate whether an existing large language model can simulate the behaviour of such a universal Lag system. We give an affirmative answer by showing that a single system-prompt can be developed for gemini-1.5-pro-001 that drives the model, under deterministic (greedy) decoding, to correctly apply each of the 2027 production rules. We conclude that, by the Church-Turing thesis, prompted gemini-1.5-pro-001 with extended autoregressive (greedy) decoding is a general purpose computer."
                    },
                    {
                        "title": "Boltzmann Machine Learning with the Latent Maximum Entropy Principle",
                        "abstract": "We present a new statistical learning paradigm for Boltzmann machines based on a new inference principle we have proposed: the latent maximum entropy principle (LME). LME is different both from Jaynes maximum entropy principle and from standard maximum likelihood estimation.We demonstrate the LME principle BY deriving new algorithms for Boltzmann machine parameter estimation, and show how robust and fast new variant of the EM algorithm can be developed.Our experiments show that estimation based on LME generally yields better results than maximum likelihood estimation, particularly when inferring hidden units from small amounts of data."
                    },
                    {
                        "title": "Adaptive Monte Carlo via Bandit Allocation",
                        "abstract": "We consider the problem of sequentially choosing between a set of unbiased Monte Carlo estimators to minimize the mean-squared-error (MSE) of a final combined estimate. By reducing this task to a stochastic multi-armed bandit problem, we show that well developed allocation strategies can be used to achieve an MSE that approaches that of the best estimator chosen in retrospect. We then extend these developments to a scenario where alternative estimators have different, possibly stochastic costs. The outcome is a new set of adaptive Monte Carlo strategies that provide stronger guarantees than previous approaches while offering practical advantages."
                    },
                    {
                        "title": "Learning with a Strong Adversary",
                        "abstract": "The robustness of neural networks to intended perturbations has recently attracted significant attention. In this paper, we propose a new method, \\emph{learning with a strong adversary}, that learns robust classifiers from supervised data. The proposed method takes finding adversarial examples as an intermediate step. A new and simple way of finding adversarial examples is presented and experimentally shown to be efficient. Experimental results demonstrate that resulting learning method greatly improves the robustness of the classification models produced."
                    },
                    {
                        "title": "Maximum Margin Bayesian Networks",
                        "abstract": "We consider the problem of learning Bayesian network classifiers that maximize the marginover a set of classification variables. We find that this problem is harder for Bayesian networks than for undirected graphical models like maximum margin Markov networks. The main difficulty is that the parameters in a Bayesian network must satisfy additional normalization constraints that an undirected graphical model need not respect. These additional constraints complicate the optimization task. Nevertheless, we derive an effective training algorithm that solves the maximum margin training problem for a range of Bayesian network topologies, and converges to an approximate solution for arbitrary network topologies. Experimental results show that the method can demonstrate improved generalization performance over Markov networks when the directed graphical structure encodes relevant knowledge. In practice, the training technique allows one to combine prior knowledge expressed as a directed (causal) model with state of the art discriminative learning methods."
                    },
                    {
                        "title": "Learning Bayesian Nets that Perform Well",
                        "abstract": "A Bayesian net (BN) is more than a succinct way to encode a probabilistic distribution; it also corresponds to a function used to answer queries. A BN can therefore be evaluated by the accuracy of the answers it returns. Many algorithms for learning BNs, however, attempt to optimize another criterion (usually likelihood, possibly augmented with a regularizing term), which is independent of the distribution of queries that are posed. This paper takes the \"performance criteria\" seriously, and considers the challenge of computing the BN whose performance - read \"accuracy over the distribution of queries\" - is optimal. We show that many aspects of this learning task are more difficult than the corresponding subtasks in the standard model."
                    },
                    {
                        "title": "Convex Relaxations of Bregman Divergence Clustering",
                        "abstract": "Although many convex relaxations of clustering have been proposed in the past decade, current formulations remain restricted to spherical Gaussian or discriminative models and are susceptible to imbalanced clusters. To address these shortcomings, we propose a new class of convex relaxations that can be flexibly applied to more general forms of Bregman divergence clustering. By basing these new formulations on normalized equivalence relations we retain additional control on relaxation quality, which allows improvement in clustering quality. We furthermore develop optimization methods that improve scalability by exploiting recent implicit matrix norm methods. In practice, we find that the new formulations are able to efficiently produce tighter clusterings that improve the accuracy of state of the art methods."
                    },
                    {
                        "title": "Generalized Conditional Gradient for Sparse Estimation",
                        "abstract": "Structured sparsity is an important modeling tool that expands the applicability of convex formulations for data analysis, however it also creates significant challenges for efficient algorithm design. In this paper we investigate the generalized conditional gradient (GCG) algorithm for solving structured sparse optimization problems---demonstrating that, with some enhancements, it can provide a more efficient alternative to current state of the art approaches. After providing a comprehensive overview of the convergence properties of GCG, we develop efficient methods for evaluating polar operators, a subroutine that is required in each GCG iteration. In particular, we show how the polar operator can be efficiently evaluated in two important scenarios: dictionary learning and structured sparse estimation. A further improvement is achieved by interleaving GCG with fixed-rank local subspace optimization. A series of experiments on matrix completion, multi-class classification, multi-view dictionary learning and overlapping group lasso shows that the proposed method can significantly reduce the training cost of current alternatives."
                    },
                    {
                        "title": "Trust-PCL: An Off-Policy Trust Region Method for Continuous Control",
                        "abstract": "Trust region methods, such as TRPO, are often used to stabilize policy optimization algorithms in reinforcement learning (RL). While current trust region strategies are effective for continuous control, they typically require a prohibitively large amount of on-policy interaction with the environment. To address this problem, we propose an off-policy trust region method, Trust-PCL. The algorithm is the result of observing that the optimal policy and state values of a maximum reward objective with a relative-entropy regularizer satisfy a set of multi-step pathwise consistencies along any path. Thus, Trust-PCL is able to maintain optimization stability while exploiting off-policy data to improve sample efficiency. When evaluated on a number of continuous control tasks, Trust-PCL improves the solution quality and sample efficiency of TRPO."
                    },
                    {
                        "title": "GenDICE: Generalized Offline Estimation of Stationary Values",
                        "abstract": "An important problem that arises in reinforcement learning and Monte Carlo methods is estimating quantities defined by the stationary distribution of a Markov chain. In many real-world applications, access to the underlying transition operator is limited to a fixed set of data that has already been collected, without additional interaction with the environment being available. We show that consistent estimation remains possible in this challenging scenario, and that effective estimation can still be achieved in important applications. Our approach is based on estimating a ratio that corrects for the discrepancy between the stationary and empirical distributions, derived from fundamental properties of the stationary distribution, and exploiting constraint reformulations based on variational divergence minimization. The resulting algorithm, GenDICE, is straightforward and effective. We prove its consistency under general conditions, provide an error analysis, and demonstrate strong empirical performance on benchmark problems, including off-line PageRank and off-policy policy evaluation."
                    },
                    {
                        "title": "Rank/Norm Regularization with Closed-Form Solutions: Application to Subspace Clustering",
                        "abstract": "When data is sampled from an unknown subspace, principal component analysis (PCA) provides an effective way to estimate the subspace and hence reduce the dimension of the data. At the heart of PCA is the Eckart-Young-Mirsky theorem, which characterizes the best rank k approximation of a matrix. In this paper, we prove a generalization of the Eckart-Young-Mirsky theorem under all unitarily invariant norms. Using this result, we obtain closed-form solutions for a set of rank/norm regularized problems, and derive closed-form solutions for a general class of subspace clustering problems (where data is modelled by unions of unknown subspaces). From these results we obtain new theoretical insights and promising experimental results."
                    },
                    {
                        "title": "Improving Policy Gradient by Exploring Under-appreciated Rewards",
                        "abstract": "This paper presents a novel form of policy gradient for model-free reinforcement learning (RL) with improved exploration properties. Current policy-based methods use entropy regularization to encourage undirected exploration of the reward landscape, which is ineffective in high dimensional spaces with sparse rewards. We propose a more directed exploration strategy that promotes exploration of under-appreciated reward regions. An action sequence is considered under-appreciated if its log-probability under the current policy under-estimates its resulting reward. The proposed exploration strategy is easy to implement, requiring small modifications to an implementation of the REINFORCE algorithm. We evaluate the approach on a set of algorithmic tasks that have long challenged RL methods. Our approach reduces hyper-parameter sensitivity and demonstrates significant improvements over baseline methods. Our algorithm successfully solves a benchmark multi-digit addition task and generalizes to long sequences. This is, to our knowledge, the first time that a pure RL method has solved addition using only reward feedback."
                    },
                    {
                        "title": "Smoothed Action Value Functions for Learning Gaussian Policies",
                        "abstract": "State-action value functions (i.e., Q-values) are ubiquitous in reinforcement learning (RL), giving rise to popular algorithms such as SARSA and Q-learning. We propose a new notion of action value defined by a Gaussian smoothed version of the expected Q-value. We show that such smoothed Q-values still satisfy a Bellman equation, making them learnable from experience sampled from an environment. Moreover, the gradients of expected reward with respect to the mean and covariance of a parameterized Gaussian policy can be recovered from the gradient and Hessian of the smoothed Q-value function. Based on these relationships, we develop new algorithms for training a Gaussian policy directly from a learned smoothed Q-value approximator. The approach is additionally amenable to proximal optimization by augmenting the objective with a penalty on KL-divergence from a previous policy. We find that the ability to learn both a mean and covariance during training leads to significantly improved results on standard continuous control benchmarks."
                    },
                    {
                        "title": "Batch Stationary Distribution Estimation",
                        "abstract": "We consider the problem of approximating the stationary distribution of an ergodic Markov chain given a set of sampled transitions. Classical simulation-based approaches assume access to the underlying process so that trajectories of sufficient length can be gathered to approximate stationary sampling. Instead, we consider an alternative setting where a fixed set of transitions has been collected beforehand, by a separate, possibly unknown procedure. The goal is still to estimate properties of the stationary distribution, but without additional access to the underlying system. We propose a consistent estimator that is based on recovering a correction ratio function over the given data. In particular, we develop a variational power method (VPM) that provides provably consistent estimates under general conditions. In addition to unifying a number of existing approaches from different subfields, we also find that VPM yields significantly better estimates across a range of problems, including queueing, stochastic differential equations, post-processing MCMC, and off-policy evaluation."
                    }
                ]
            }
        }
    },
    "2403.04082": {
        "paper_data": {
            "title": "Inference via Interpolation: Contrastive Representations Provably Enable Planning and Inference",
            "url": "http://arxiv.org/abs/2403.04082v2",
            "arxiv_id": "2403.04082",
            "authors": [
                "Benjamin Eysenbach",
                "Vivek Myers",
                "Ruslan Salakhutdinov",
                "Sergey Levine"
            ],
            "abstract": "Given time series data, how can we answer questions like \"what will happen in the future?\" and \"how did we get here?\" These sorts of probabilistic inference questions are challenging when observations are high-dimensional. In this paper, we show how these questions can have compact, closed form solutions in terms of learned representations. The key idea is to apply a variant of contrastive learning to time series data. Prior work already shows that the representations learned by contrastive learning encode a probability ratio. By extending prior work to show that the marginal distribution over representations is Gaussian, we can then prove that joint distribution of representations is also Gaussian. Taken together, these results show that representations learned via temporal contrastive learning follow a Gauss-Markov chain, a graphical model where inference (e.g., prediction, planning) over representations corresponds to inverting a low-dimensional matrix. In one special case, inferring intermediate representations will be equivalent to interpolating between the learned representations. We validate our theory using numerical simulations on tasks up to 46-dimensions.",
            "introduction": "   1 Introduction  Probabilistic modeling of time-series data has applications ranging from robotic control\u00a0(84) to material science\u00a0(42), from cell biology\u00a0(74) to astrophysics\u00a0(55). These applications are often concerned with two questions: predicting future states (e.g., what will this cell look like in an hour), and inferring trajectories between two given states. However, answering these questions often requires reasoning over high-dimensional data, which can be challenging as most tools in the standard probabilistic toolkit require generation. Might it be possible to use discriminative methods (e.g., contrastive learning) to perform such inferences?   Figure 1:  We apply temporal contrastive learning to observation pairs to obtain representations (\u03c8\u2062(x0),\u03c8\u2062(xt+k)\ud835\udf13subscript\ud835\udc650\ud835\udf13subscript\ud835\udc65\ud835\udc61\ud835\udc58\\psi(x_{0}),\\psi(x_{t+k})italic_\u03c8 ( italic_x start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ) , italic_\u03c8 ( italic_x start_POSTSUBSCRIPT italic_t + italic_k end_POSTSUBSCRIPT )) such that A\u2062\u03c8\u2062(x0)\ud835\udc34\ud835\udf13subscript\ud835\udc650A\\psi(x_{0})italic_A italic_\u03c8 ( italic_x start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ) is close to \u03c8\u2062(xt+k)\ud835\udf13subscript\ud835\udc65\ud835\udc61\ud835\udc58\\psi(x_{t+k})italic_\u03c8 ( italic_x start_POSTSUBSCRIPT italic_t + italic_k end_POSTSUBSCRIPT ). While inferring waypoints in the high-dimensional observation space is challenging, we show that the distribution over intermediate latent representations has a closed form solution corresponding to linear interpolation between the initial and final representations.    Many prior works aim to learn representations such that are easy to predict while retaining salient bits of information. For time-series data, we want the representation to remain a sufficient statistic for distributions related to time \u2013 for example, they should retain bits required to predict future states (or representations thereof). While generative methods\u00a0(100; 102; 56; 26) have this property, they tend to be computationally expensive (see, e.g.,\u00a0(72)) and can be challenging to scale to high-dimensional observations.   In this paper, we will study how contrastive methods (which are discriminative, rather than generative) can perform inference over times series. Ideally, we want representations of observations x\ud835\udc65xitalic_x to be a sufficient statistic for temporal relationships (e.g., does x\u2032superscript\ud835\udc65\u2032x^{\\prime}italic_x start_POSTSUPERSCRIPT \u2032 end_POSTSUPERSCRIPT occur after x\ud835\udc65xitalic_x?) but need not retain other information about x\ud835\udc65xitalic_x (e.g. the location of static objects). This intuition motivates us to study how contrastive representation learning methods\u00a0(65; 81; 18; 87; 95) might be used to solve prediction and planning problems on time series data. While prior works in computer vision\u00a0(16; 65) and natural language processing (NLP)\u00a0(59) often study the geometry of learned representations, our results show how geometric operations such as interpolation are related to inference. Our analysis will focus on a regularized version of the symmetrized infoNCE objective\u00a0(71), generating positive examples by sampling pairs of observations from the same time series data. We will study how representations learned in this way can facilitate two inference questions: prediction and planning.222Following prior work\u00a0(10; 6), we will use planning to refer to the problem of inferring intermediate states, not to refer to an optimal control problem. As a stepping stone, we will build upon prior work\u00a0(89) to show that regularized contrastive learning should produce representations whose marginal distribution is an isotropic Gaussian distribution.   The main contribution of this paper is to demonstrate how intermediate and future time steps in a time series can be inferred easily using contrastive representations. This inference problem captures a number of practical tasks: interpolation, in-filling, and even planning and control, where the intermediate steps represent states between a stand and goal. While ordinarily these problems require an",
            "references": [
                {
                    "title": "XSkill: Cross Embodiment Skill Discovery",
                    "abstract": "Human demonstration videos are a widely available data source for robot learning and an intuitive user interface for expressing desired behavior. However, directly extracting reusable robot manipulation skills from unstructured human videos is challenging due to the big embodiment difference and unobserved action parameters. To bridge this embodiment gap, this paper introduces XSkill, an imitation learning framework that 1) discovers a cross-embodiment representation called skill prototypes purely from unlabeled human and robot manipulation videos, 2) transfers the skill representation to robot actions using conditional diffusion policy, and finally, 3) composes the learned skill to accomplish unseen tasks specified by a human prompt video. Our experiments in simulation and real-world environments show that the discovered skill prototypes facilitate both skill transfer and composition for unseen tasks, resulting in a more general and scalable imitation learning framework. The benchmark, code, and qualitative results are on https://xskill.cs.columbia.edu/"
                },
                {
                    "title": "Metric Residual Network for Sample Efficient Goal-Conditioned Reinforcement Learning",
                    "abstract": "Goal-conditioned reinforcement learning (GCRL) has a wide range of potential real-world applications, including manipulation and navigation problems in robotics. Especially in such robotics tasks, sample efficiency is of the utmost importance for GCRL since, by default, the agent is only rewarded when it reaches its goal. While several methods have been proposed to improve the sample efficiency of GCRL, one relatively under-studied approach is the design of neural architectures to support sample efficiency.\nIn this work, we introduce a novel neural architecture for GCRL that achieves significantly better sample efficiency than the commonly-used monolithic network architecture. \nThe key insight is that the \noptimal action-value function must satisfy the triangle inequality in a specific sense.\nFurthermore, we introduce the metric residual network (MRN) that deliberately decomposes the action-value function into the negated summation of a metric plus a residual asymmetric component. MRN provably approximates any optimal action-value function, thus making it a fitting neural architecture for GCRL.\nWe conduct comprehensive experiments across 12 standard benchmark environments in GCRL. The empirical results demonstrate that MRN uniformly outperforms other state-of-the-art GCRL neural architectures in terms of sample efficiency. The code is available at https://github.com/Cranial-XIX/metric-residual-network."
                },
                {
                    "title": "Stabilizing Contrastive RL: Techniques for Robotic Goal Reaching from Offline Data",
                    "abstract": "Robotic systems that rely primarily on self-supervised learning have the potential to decrease the amount of human annotation and engineering effort required to learn control strategies. In the same way that prior robotic systems have leveraged self-supervised techniques from computer vision (CV) and natural language processing (NLP), our work builds on prior work showing that the reinforcement learning (RL) itself can be cast as a self-supervised problem: learning to reach any goal without human-specified rewards or labels. Despite the seeming appeal, little (if any) prior work has demonstrated how self-supervised RL methods can be practically deployed on robotic systems. By first studying a challenging simulated version of this task, we discover design decisions about architectures and hyperparameters that increase the success rate by $2 \\times$. These findings lay the groundwork for our main result: we demonstrate that a self-supervised RL algorithm based on contrastive learning can solve real-world, image-based robotic manipulation tasks, with tasks being specified by a single goal image provided after training."
                },
                {
                    "title": "LIV: Language-Image Representations and Rewards for Robotic Control",
                    "abstract": "We present Language-Image Value learning (LIV), a unified objective for vision-language representation and reward learning from action-free videos with text annotations. Exploiting a novel connection between dual reinforcement learning and mutual information contrastive learning, the LIV objective trains a multi-modal representation that implicitly encodes a universal value function for tasks specified as language or image goals. We use LIV to pre-train the first control-centric vision-language representation from large human video datasets such as EpicKitchen. Given only a language or image goal, the pre-trained LIV model can assign dense rewards to each frame in videos of unseen robots or humans attempting that task in unseen environments. Further, when some target domain-specific data is available, the same objective can be used to fine-tune and improve LIV and even other pre-trained representations for robotic control and reward specification in that domain. In our experiments on several simulated and real-world robot environments, LIV models consistently outperform the best prior input state representations for imitation learning, as well as reward specification methods for policy synthesis. Our results validate the advantages of joint vision-language representation and reward learning within the unified, compact LIV framework."
                },
                {
                    "title": "Distance Weighted Supervised Learning for Offline Interaction Data",
                    "abstract": "Sequential decision making algorithms often struggle to leverage different sources of unstructured offline interaction data. Imitation learning (IL) methods based on supervised learning are robust, but require optimal demonstrations, which are hard to collect. Offline goal-conditioned reinforcement learning (RL) algorithms promise to learn from sub-optimal data, but face optimization challenges especially with high-dimensional data. To bridge the gap between IL and RL, we introduce Distance Weighted Supervised Learning or DWSL, a supervised method for learning goal-conditioned policies from offline data. DWSL models the entire distribution of time-steps between states in offline data with only supervised learning, and uses this distribution to approximate shortest path distances. To extract a policy, we weight actions by their reduction in distance estimates. Theoretically, DWSL converges to an optimal policy constrained to the data distribution, an attractive property for offline learning, without any bootstrapping. Across all datasets we test, DWSL empirically maintains behavior cloning as a lower bound while still exhibiting policy improvement. In high-dimensional image domains, DWSL surpasses the performance of both prior goal-conditioned IL and RL algorithms. Visualizations and code can be found at https://sites.google.com/view/dwsl/home ."
                },
                {
                    "title": "Optimal Goal-Reaching Reinforcement Learning via Quasimetric Learning",
                    "abstract": "In goal-reaching reinforcement learning (RL), the optimal value function has a particular geometry, called quasimetric structure. This paper introduces Quasimetric Reinforcement Learning (QRL), a new RL method that utilizes quasimetric models to learn optimal value functions. Distinct from prior approaches, the QRL objective is specifically designed for quasimetrics, and provides strong theoretical recovery guarantees. Empirically, we conduct thorough analyses on a discretized MountainCar environment, identifying properties of QRL and its advantages over alternatives. On offline and online goal-reaching benchmarks, QRL also demonstrates improved sample efficiency and performance, across both state-based and image-based observations."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Language-Driven Representation Learning for Robotics",
                    "abstract": "Recent work in visual representation learning for robotics demonstrates the viability of learning from large video datasets of humans performing everyday tasks. Leveraging methods such as masked autoencoding and contrastive learning, these representations exhibit strong transfer to policy learning for visuomotor control. But, robot learning encompasses a diverse set of problems beyond control including grasp affordance prediction, language-conditioned imitation learning, and intent scoring for human-robot collaboration, amongst others. First, we demonstrate that existing representations yield inconsistent results across these tasks: masked autoencoding approaches pick up on low-level spatial features at the cost of high-level semantics, while contrastive learning approaches capture the opposite. We then introduce Voltron, a framework for language-driven representation learning from human videos and associated captions. Voltron trades off language-conditioned visual reconstruction to learn low-level visual patterns, and visually-grounded language generation to encode high-level semantics. We also construct a new evaluation suite spanning five distinct robot learning problems $\\unicode{x2013}$ a unified platform for holistically evaluating visual representations for robotics. Through comprehensive, controlled experiments across all five problems, we find that Voltron's language-driven representations outperform the prior state-of-the-art, especially on targeted problems requiring higher-level features."
                },
                {
                    "title": "Understanding Self-Predictive Learning for Reinforcement Learning",
                    "abstract": "We study the learning dynamics of self-predictive learning for reinforcement learning, a family of algorithms that learn representations by minimizing the prediction error of their own future latent representations. Despite its recent empirical success, such algorithms have an apparent defect: trivial representations (such as constants) minimize the prediction error, yet it is obviously undesirable to converge to such solutions. Our central insight is that careful designs of the optimization dynamics are critical to learning meaningful representations. We identify that a faster paced optimization of the predictor and semi-gradient updates on the representation, are crucial to preventing the representation collapse. Then in an idealized setup, we show self-predictive learning dynamics carries out spectral decomposition on the state transition matrix, effectively capturing information of the transition dynamics. Building on the theoretical insights, we propose bidirectional self-predictive learning, a novel self-predictive algorithm that learns two representations simultaneously. We examine the robustness of our theoretical insights with a number of small-scale experiments and showcase the promise of the novel representation learning algorithm with large-scale experiments."
                },
                {
                    "title": "UniMASK: Unified Inference in Sequential Decision Problems",
                    "abstract": "Randomly masking and predicting word tokens has been a successful approach in pre-training language models for a variety of downstream tasks. In this work, we observe that the same idea also applies naturally to sequential decision-making, where many well-studied tasks like behavior cloning, offline reinforcement learning, inverse dynamics, and waypoint conditioning correspond to different sequence maskings over a sequence of states, actions, and returns. We introduce the UniMASK framework, which provides a unified way to specify models which can be trained on many different sequential decision-making tasks. We show that a single UniMASK model is often capable of carrying out many tasks with performance similar to or better than single-task models. Additionally, after fine-tuning, our UniMASK models consistently outperform comparable single-task models. Our code is publicly available at https://github.com/micahcarroll/uniMASK."
                },
                {
                    "title": "Contrastive Value Learning: Implicit Models for Simple Offline RL",
                    "abstract": "Model-based reinforcement learning (RL) methods are appealing in the offline setting because they allow an agent to reason about the consequences of actions without interacting with the environment. Prior methods learn a 1-step dynamics model, which predicts the next state given the current state and action. These models do not immediately tell the agent which actions to take, but must be integrated into a larger RL framework. Can we model the environment dynamics in a different way, such that the learned model does directly indicate the value of each action? In this paper, we propose Contrastive Value Learning (CVL), which learns an implicit, multi-step model of the environment dynamics. This model can be learned without access to reward functions, but nonetheless can be used to directly estimate the value of each action, without requiring any TD learning. Because this model represents the multi-step transitions implicitly, it avoids having to predict high-dimensional observations and thus scales to high-dimensional tasks. Our experiments demonstrate that CVL outperforms prior offline RL methods on complex continuous control benchmarks."
                },
                {
                    "title": "Generalization with Lossy Affordances: Leveraging Broad Offline Data for Learning Visuomotor Tasks",
                    "abstract": "The utilization of broad datasets has proven to be crucial for generalization for a wide range of fields. However, how to effectively make use of diverse multi-task data for novel downstream tasks still remains a grand challenge in robotics. To tackle this challenge, we introduce a framework that acquires goal-conditioned policies for unseen temporally extended tasks via offline reinforcement learning on broad data, in combination with online fine-tuning guided by subgoals in learned lossy representation space. When faced with a novel task goal, the framework uses an affordance model to plan a sequence of lossy representations as subgoals that decomposes the original task into easier problems. Learned from the broad data, the lossy representation emphasizes task-relevant information about states and goals while abstracting away redundant contexts that hinder generalization. It thus enables subgoal planning for unseen tasks, provides a compact input to the policy, and facilitates reward shaping during fine-tuning. We show that our framework can be pre-trained on large-scale datasets of robot experiences from prior work and efficiently fine-tuned for novel tasks, entirely from visual inputs without any manual reward engineering."
                },
                {
                    "title": "VIP: Towards Universal Visual Reward and Representation via Value-Implicit Pre-Training",
                    "abstract": "Reward and representation learning are two long-standing challenges for learning an expanding set of robot manipulation skills from sensory observations. Given the inherent cost and scarcity of in-domain, task-specific robot data, learning from large, diverse, offline human videos has emerged as a promising path towards acquiring a generally useful visual representation for control; however, how these human videos can be used for general-purpose reward learning remains an open question. We introduce $\\textbf{V}$alue-$\\textbf{I}$mplicit $\\textbf{P}$re-training (VIP), a self-supervised pre-trained visual representation capable of generating dense and smooth reward functions for unseen robotic tasks. VIP casts representation learning from human videos as an offline goal-conditioned reinforcement learning problem and derives a self-supervised dual goal-conditioned value-function objective that does not depend on actions, enabling pre-training on unlabeled human videos. Theoretically, VIP can be understood as a novel implicit time contrastive objective that generates a temporally smooth embedding, enabling the value function to be implicitly defined via the embedding distance, which can then be used to construct the reward for any goal-image specified downstream task. Trained on large-scale Ego4D human videos and without any fine-tuning on in-domain, task-specific data, VIP's frozen representation can provide dense visual reward for an extensive set of simulated and $\\textbf{real-robot}$ tasks, enabling diverse reward-based visual control methods and significantly outperforming all prior pre-trained representations. Notably, VIP can enable simple, $\\textbf{few-shot}$ offline RL on a suite of real-world robot tasks with as few as 20 trajectories."
                },
                {
                    "title": "Simplifying Model-based RL: Learning Representations, Latent-space Models, and Policies with One Objective",
                    "abstract": "While reinforcement learning (RL) methods that learn an internal model of the environment have the potential to be more sample efficient than their model-free counterparts, learning to model raw observations from high dimensional sensors can be challenging. Prior work has addressed this challenge by learning low-dimensional representation of observations through auxiliary objectives, such as reconstruction or value prediction. However, the alignment between these auxiliary objectives and the RL objective is often unclear. In this work, we propose a single objective which jointly optimizes a latent-space model and policy to achieve high returns while remaining self-consistent. This objective is a lower bound on expected returns. Unlike prior bounds for model-based RL on policy exploration or model guarantees, our bound is directly on the overall RL objective. We demonstrate that the resulting algorithm matches or improves the sample-efficiency of the best prior model-based and model-free RL methods. While sample efficient methods typically are computationally demanding, our method attains the performance of SAC in about 50% less wall-clock time."
                },
                {
                    "title": "BYOL-Explore: Exploration by Bootstrapped Prediction",
                    "abstract": "We present BYOL-Explore, a conceptually simple yet general approach for curiosity-driven exploration in visually-complex environments. BYOL-Explore learns a world representation, the world dynamics, and an exploration policy all-together by optimizing a single prediction loss in the latent space with no additional auxiliary objective. We show that BYOL-Explore is effective in DM-HARD-8, a challenging partially-observable continuous-action hard-exploration benchmark with visually-rich 3-D environments. On this benchmark, we solve the majority of the tasks purely through augmenting the extrinsic reward with BYOL-Explore s intrinsic reward, whereas prior work could only get off the ground with human demonstrations. As further evidence of the generality of BYOL-Explore, we show that it achieves superhuman performance on the ten hardest exploration games in Atari while having a much simpler design than other competitive agents."
                },
                {
                    "title": "Contrastive Learning as Goal-Conditioned Reinforcement Learning",
                    "abstract": "In reinforcement learning (RL), it is easier to solve a task if given a good representation. While deep RL should automatically acquire such good representations, prior work often finds that learning representations in an end-to-end fashion is unstable and instead equip RL algorithms with additional representation learning parts (e.g., auxiliary losses, data augmentation). How can we design RL algorithms that directly acquire good representations? In this paper, instead of adding representation learning parts to an existing RL algorithm, we show (contrastive) representation learning methods can be cast as RL algorithms in their own right. To do this, we build upon prior work and apply contrastive representation learning to action-labeled trajectories, in such a way that the (inner product of) learned representations exactly corresponds to a goal-conditioned value function. We use this idea to reinterpret a prior RL method as performing contrastive learning, and then use the idea to propose a much simpler method that achieves similar performance. Across a range of goal-conditioned RL tasks, we demonstrate that contrastive RL methods achieve higher success rates than prior non-contrastive methods, including in the offline RL setting. We also show that contrastive RL outperforms prior methods on image-based tasks, without using data augmentation or auxiliary objectives."
                },
                {
                    "title": "How Far I'll Go: Offline Goal-Conditioned Reinforcement Learning via f-Advantage Regression",
                    "abstract": "Offline goal-conditioned reinforcement learning (GCRL) promises general-purpose skill learning in the form of reaching diverse goals from purely offline datasets. We propose $\\textbf{Go}$al-conditioned $f$-$\\textbf{A}$dvantage $\\textbf{R}$egression (GoFAR), a novel regression-based offline GCRL algorithm derived from a state-occupancy matching perspective; the key intuition is that the goal-reaching task can be formulated as a state-occupancy matching problem between a dynamics-abiding imitator agent and an expert agent that directly teleports to the goal. In contrast to prior approaches, GoFAR does not require any hindsight relabeling and enjoys uninterleaved optimization for its value and policy networks. These distinct features confer GoFAR with much better offline performance and stability as well as statistical performance guarantee that is unattainable for prior methods. Furthermore, we demonstrate that GoFAR's training objectives can be re-purposed to learn an agent-independent goal-conditioned planner from purely offline source-domain data, which enables zero-shot transfer to new target domains. Through extensive experiments, we validate GoFAR's effectiveness in various problem settings and tasks, significantly outperforming prior state-of-art. Notably, on a real robotic dexterous manipulation task, while no other method makes meaningful progress, GoFAR acquires complex manipulation behavior that successfully accomplishes diverse goals."
                },
                {
                    "title": "INFOrmation Prioritization through EmPOWERment in Visual Model-Based RL",
                    "abstract": "Model-based reinforcement learning (RL) algorithms designed for handling complex visual observations typically learn some sort of latent state representation, either explicitly or implicitly. Standard methods of this sort do not distinguish between functionally relevant aspects of the state and irrelevant distractors, instead aiming to represent all available information equally. We propose a modified objective for model-based RL that, in combination with mutual information maximization, allows us to learn representations and dynamics for visual model-based RL without reconstruction in a way that explicitly prioritizes functionally relevant factors. The key principle behind our design is to integrate a term inspired by variational empowerment into a state-space model based on mutual information. This term prioritizes information that is correlated with action, thus ensuring that functionally relevant factors are captured first. Furthermore, the same empowerment term also promotes faster exploration during the RL process, especially for sparse-reward tasks where the reward signal is insufficient to drive exploration in the early stages of learning. We evaluate the approach on a suite of vision-based robot control tasks with natural video backgrounds, and show that the proposed prioritized information objective outperforms state-of-the-art model based RL approaches with higher sample efficiency and episodic returns. https://sites.google.com/view/information-empowerment"
                },
                {
                    "title": "R3M: A Universal Visual Representation for Robot Manipulation",
                    "abstract": "We study how visual representations pre-trained on diverse human video data can enable data-efficient learning of downstream robotic manipulation tasks. Concretely, we pre-train a visual representation using the Ego4D human video dataset using a combination of time-contrastive learning, video-language alignment, and an L1 penalty to encourage sparse and compact representations. The resulting representation, R3M, can be used as a frozen perception module for downstream policy learning. Across a suite of 12 simulated robot manipulation tasks, we find that R3M improves task success by over 20% compared to training from scratch and by over 10% compared to state-of-the-art visual representations like CLIP and MoCo. Furthermore, R3M enables a Franka Emika Panda arm to learn a range of manipulation tasks in a real, cluttered apartment given just 20 demonstrations. Code and pre-trained models are available at https://tinyurl.com/robotr3m."
                },
                {
                    "title": "Rethinking Goal-conditioned Supervised Learning and Its Connection to Offline RL",
                    "abstract": "Solving goal-conditioned tasks with sparse rewards using self-supervised learning is promising because of its simplicity and stability over current reinforcement learning (RL) algorithms. A recent work, called Goal-Conditioned Supervised Learning (GCSL), provides a new learning framework by iteratively relabeling and imitating self-generated experiences. In this paper, we revisit the theoretical property of GCSL -- optimizing a lower bound of the goal reaching objective, and extend GCSL as a novel offline goal-conditioned RL algorithm. The proposed method is named Weighted GCSL (WGCSL), in which we introduce an advanced compound weight consisting of three parts (1) discounted weight for goal relabeling, (2) goal-conditioned exponential advantage weight, and (3) best-advantage weight. Theoretically, WGCSL is proved to optimize an equivalent lower bound of the goal-conditioned RL objective and generates monotonically improved policies via an iterated scheme. The monotonic property holds for any behavior policies, and therefore WGCSL can be applied to both online and offline settings. To evaluate algorithms in the offline goal-conditioned RL setting, we provide a benchmark including a range of point and simulated robot domains. Experiments in the introduced benchmark demonstrate that WGCSL can consistently outperform GCSL and existing state-of-the-art offline methods in the fully offline goal-conditioned setting."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "C-Planning: An Automatic Curriculum for Learning Goal-Reaching Tasks",
                    "abstract": "Goal-conditioned reinforcement learning (RL) can solve tasks in a wide range of domains, including navigation and manipulation, but learning to reach distant goals remains a central challenge to the field. Learning to reach such goals is particularly hard without any offline data, expert demonstrations, and reward shaping. In this paper, we propose an algorithm to solve the distant goal-reaching task by using search at training time to automatically generate a curriculum of intermediate states. Our algorithm, Classifier-Planning (C-Planning), frames the learning of the goal-conditioned policies as expectation maximization: the E-step corresponds to planning an optimal sequence of waypoints using graph search, while the M-step aims to learn a goal-conditioned policy to reach those waypoints. Unlike prior methods that combine goal-conditioned RL with graph search, ours performs search only during training and not testing, significantly decreasing the compute costs of deploying the learned policy. Empirically, we demonstrate that our method is more sample efficient than prior methods. Moreover, it is able to solve very long horizons manipulation and navigation tasks, tasks that prior goal-conditioned methods and methods based on graph search fail to solve."
                },
                {
                    "title": "Finetuning Pretrained Transformers into Variational Autoencoders",
                    "abstract": "Text variational autoencoders (VAEs) are notorious for posterior collapse, a phenomenon where the model\u2019s decoder learns to ignore signals from the encoder. Because posterior collapse is known to be exacerbated by expressive decoders, Transformers have seen limited adoption as components of text VAEs. Existing studies that incorporate Transformers into text VAEs (Li et al., 2020; Fang et al., 2021) mitigate posterior collapse using massive pretraining, a technique unavailable to most of the research community without extensive computing resources. We present a simple two-phase training scheme to convert a sequence-to-sequence Transformer into a VAE with just finetuning. The resulting language model is competitive with massively pretrained Transformer-based VAEs in some internal metrics while falling short on others. To facilitate training we comprehensively explore the impact of common posterior collapse alleviation techniques in the literature. We release our code for reproducability."
                },
                {
                    "title": "Goal-Conditioned Reinforcement Learning with Imagined Subgoals",
                    "abstract": "Goal-conditioned reinforcement learning endows an agent with a large variety of skills, but it often struggles to solve tasks that require more temporally extended reasoning. In this work, we propose to incorporate imagined subgoals into policy learning to facilitate learning of complex tasks. Imagined subgoals are predicted by a separate high-level policy, which is trained simultaneously with the policy and its critic. This high-level policy predicts intermediate states halfway to the goal using the value function as a reachability metric. We don't require the policy to reach these subgoals explicitly. Instead, we use them to define a prior policy, and incorporate this prior into a KL-constrained policy iteration scheme to speed up and regularize learning. Imagined subgoals are used during policy learning, but not during test time, where we only apply the learned policy. We evaluate our approach on complex robotic navigation and manipulation tasks and show that it outperforms existing methods by a large margin."
                },
                {
                    "title": "Temporal Predictive Coding For Model-Based Planning In Latent Space",
                    "abstract": "High-dimensional observations are a major challenge in the application of model-based reinforcement learning (MBRL) to real-world environments. To handle high-dimensional sensory inputs, existing approaches use representation learning to map high-dimensional observations into a lower-dimensional latent space that is more amenable to dynamics estimation and planning. In this work, we present an information-theoretic approach that employs temporal predictive coding to encode elements in the environment that can be predicted across time. Since this approach focuses on encoding temporally-predictable information, we implicitly prioritize the encoding of task-relevant components over nuisance information within the environment that are provably task-irrelevant. By learning this representation in conjunction with a recurrent state space model, we can then perform planning in latent space. We evaluate our model on a challenging modification of standard DMControl tasks where the background is replaced with natural videos that contain complex but irrelevant information to the planning task. Our experiments show that our model is superior to existing methods in the challenging complex-background setting while remaining competitive with current state-of-the-art models in the standard setting."
                },
                {
                    "title": "Learning Markov State Abstractions for Deep Reinforcement Learning",
                    "abstract": "A fundamental assumption of reinforcement learning in Markov decision processes (MDPs) is that the relevant decision process is, in fact, Markov. However, when MDPs have rich observations, agents typically learn by way of an abstract state representation, and such representations are not guaranteed to preserve the Markov property. We introduce a novel set of conditions and prove that they are sufficient for learning a Markov abstract state representation. We then describe a practical training procedure that combines inverse model estimation and temporal contrastive learning to learn an abstraction that approximately satisfies these conditions. Our novel training objective is compatible with both online and offline training: it does not require a reward signal, but agents can capitalize on reward information when available. We empirically evaluate our approach on a visual gridworld domain and a set of continuous control benchmarks. Our approach learns representations that capture the underlying structure of the domain and lead to improved sample efficiency over state-of-the-art deep reinforcement learning with visual features -- often matching or exceeding the performance achieved with hand-designed compact state information."
                },
                {
                    "title": "MICo: Improved representations via sampling-based state similarity for Markov decision processes",
                    "abstract": "We present a new behavioural distance over the state space of a Markov decision process, and demonstrate the use of this distance as an effective means of shaping the learnt representations of deep reinforcement learning agents. While existing notions of state similarity are typically difficult to learn at scale due to high computational cost and lack of sample-based algorithms, our newly-proposed distance addresses both of these issues. In addition to providing detailed theoretical analysis, we provide empirical evidence that learning this distance alongside the value function yields structured and informative representations, including strong results on the Arcade Learning Environment benchmark."
                },
                {
                    "title": "Offline Reinforcement Learning as One Big Sequence Modeling Problem",
                    "abstract": "Reinforcement learning (RL) is typically concerned with estimating stationary policies or single-step models, leveraging the Markov property to factorize problems in time. However, we can also view RL as a generic sequence modeling problem, with the goal being to produce a sequence of actions that leads to a sequence of high rewards. Viewed in this way, it is tempting to consider whether high-capacity sequence prediction models that work well in other domains, such as natural-language processing, can also provide effective solutions to the RL problem. To this end, we explore how RL can be tackled with the tools of sequence modeling, using a Transformer architecture to model distributions over trajectories and repurposing beam search as a planning algorithm. Framing RL as sequence modeling problem simplifies a range of design decisions, allowing us to dispense with many of the components common in offline RL algorithms. We demonstrate the flexibility of this approach across long-horizon dynamics prediction, imitation learning, goal-conditioned RL, and offline RL. Further, we show that this approach can be combined with existing model-free algorithms to yield a state-of-the-art planner in sparse-reward, long-horizon tasks."
                },
                {
                    "title": "Decision Transformer: Reinforcement Learning via Sequence Modeling",
                    "abstract": "We introduce a framework that abstracts Reinforcement Learning (RL) as a sequence modeling problem. This allows us to draw upon the simplicity and scalability of the Transformer architecture, and associated advances in language modeling such as GPT-x and BERT. In particular, we present Decision Transformer, an architecture that casts the problem of RL as conditional sequence modeling. Unlike prior approaches to RL that fit value functions or compute policy gradients, Decision Transformer simply outputs the optimal actions by leveraging a causally masked Transformer. By conditioning an autoregressive model on the desired return (reward), past states, and actions, our Decision Transformer model can generate future actions that achieve the desired return. Despite its simplicity, Decision Transformer matches or exceeds the performance of state-of-the-art model-free offline RL baselines on Atari, OpenAI Gym, and Key-to-Door tasks."
                },
                {
                    "title": "Curious Representation Learning for Embodied Intelligence",
                    "abstract": "Self-supervised representation learning has achieved remarkable success in recent years. By subverting the need for supervised labels, such approaches are able to utilize the numerous unlabeled images that exist on the Internet and in photographic datasets. Yet to build truly intelligent agents, we must construct representation learning algorithms that can learn not only from datasets but also learn from environments. An agent in a natural environment will not typically be fed curated data. Instead, it must explore its environment to acquire the data it will learn from. We propose a framework, curious representation learning (CRL), which jointly learns a reinforcement learning policy and a visual representation model. The policy is trained to maximize the error of the representation learner, and in doing so is incentivized to explore its environment. At the same time, the learned representation becomes stronger and stronger as the policy feeds it ever harder data to learn from. Our learned representations enable promising transfer to downstream navigation tasks, performing better than or comparably to ImageNet pretraining without using any supervision at all. In addition, despite being trained in simulation, our learned representations can obtain interpretable results on real images. Code is available at https://yilundu.github.io/crl/."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Model-Based Visual Planning with Self-Supervised Functional Distances",
                    "abstract": "A generalist robot must be able to complete a variety of tasks in its environment. One appealing way to specify each task is in terms of a goal observation. However, learning goal-reaching policies with reinforcement learning remains a challenging problem, particularly when hand-engineered reward functions are not available. Learned dynamics models are a promising approach for learning about the environment without rewards or task-directed data, but planning to reach goals with such a model requires a notion of functional similarity between observations and goal states. We present a self-supervised method for model-based visual goal reaching, which uses both a visual dynamics model as well as a dynamical distance function learned using model-free reinforcement learning. Our approach learns entirely using offline, unlabeled data, making it practical to scale to large and diverse datasets. In our experiments, we find that our method can successfully learn models that perform a variety of tasks at test-time, moving objects amid distractors with a simulated robotic arm and even learning to open and close a drawer using a real-world robot. In comparisons, we find that this approach substantially outperforms both model-free and model-based prior methods. Videos and visualizations are available here: http://sites.google.com/berkeley.edu/mbold."
                },
                {
                    "title": "Intrinsically Motivated Goal-Conditioned Reinforcement Learning: a Short Survey",
                    "abstract": "Building autonomous machines that can explore open-ended environments, discover possible interactions and autonomously build repertoires of skills is a general objective of artificial intelligence. Developmental approaches argue that this can only be achieved by autonomous and intrinsically motivated learning agents that can generate, select and learn to solve their own problems. In recent years, we have seen a convergence of developmental approaches, and developmental robotics in particular, with deep reinforcement learning (RL) methods, forming the new domain of developmental machine learning. Within this new domain, we review here a set of methods where deep RL algorithms are trained to tackle the developmental robotics problem of the autonomous acquisition of open-ended repertoires of skills. Intrinsically motivated goal-conditioned RL algorithms train agents to learn to represent, generate and pursue their own goals. The self-generation of goals requires the learning of compact goal encodings as well as their associated goal-achievement functions, which results in new challenges compared to traditional RL algorithms designed to tackle pre-defined sets of goals using external reward signals. This paper proposes a typology of these methods at the intersection of deep RL and developmental approaches, surveys recent approaches and discusses future avenues."
                },
                {
                    "title": "Exploring Simple Siamese Representation Learning",
                    "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code is made available.1"
                },
                {
                    "title": "C-Learning: Learning to Achieve Goals via Recursive Classification",
                    "abstract": "We study the problem of predicting and controlling the future state distribution of an autonomous agent. This problem, which can be viewed as a reframing of goal-conditioned reinforcement learning (RL), is centered around learning a conditional probability density function over future states. Instead of directly estimating this density function, we indirectly estimate this density function by training a classifier to predict whether an observation comes from the future. Via Bayes' rule, predictions from our classifier can be transformed into predictions over future states. Importantly, an off-policy variant of our algorithm allows us to predict the future state distribution of a new policy, without collecting new experience. This variant allows us to optimize functionals of a policy's future state distribution, such as the density of reaching a particular goal state. While conceptually similar to Q-learning, our work lays a principled foundation for goal-conditioned RL as density estimation, providing justification for goal-conditioned methods used in prior work. This foundation makes hypotheses about Q-learning, including the optimal goal-sampling ratio, which we confirm experimentally. Moreover, our proposed method is competitive with prior goal-conditioned RL methods."
                },
                {
                    "title": "Semantics-Guided Representation Learning with Applications to Visual Synthesis",
                    "abstract": "Learning interpretable and interpolatable latent representations has been an emerging research direction, allowing researchers to understand and utilize the derived latent space for further applications such as visual synthesis or recognition. While most existing approaches derive an interpolatable latent space and induces smooth transition in image appearance, it is still not clear how to observe desirable representations which would contain semantic information of interest. In this paper, we aim to learn meaningful representations and simultaneously perform semantic-oriented and visually-smooth interpolation. To this end, we propose an angular triplet-neighbor loss (ATNL) that enables learning a latent representation whose distribution matches the semantic information of interest. With the latent space guided by ATNL, we further utilize spherical semantic interpolation for generating semantic warping of images, allowing synthesis of desirable visual data. Experiments on MNIST and CMU Multi-PIE datasets qualitatively and quantitatively verify the effectiveness of our method."
                },
                {
                    "title": "Decoupling Representation Learning from Reinforcement Learning",
                    "abstract": "In an effort to overcome limitations of reward-driven feature learning in deep reinforcement learning (RL) from images, we propose decoupling representation learning from policy learning. To this end, we introduce a new unsupervised learning (UL) task, called Augmented Temporal Contrast (ATC), which trains a convolutional encoder to associate pairs of observations separated by a short time difference, under image augmentations and using a contrastive loss. In online RL experiments, we show that training the encoder exclusively using ATC matches or outperforms end-to-end RL in most environments. Additionally, we benchmark several leading UL algorithms by pre-training encoders on expert demonstrations and using them, with weights frozen, in RL agents; we find that agents using ATC-trained encoders outperform all others. We also train multi-task encoders on data from multiple environments and show generalization to different downstream RL tasks. Finally, we ablate components of ATC, and introduce a new data augmentation to enable replay of (compressed) latent images from pre-trained encoders when RL requires augmentation. Our experiments span visually diverse RL benchmarks in DeepMind Control, DeepMind Lab, and Atari, and our complete code is available at this https URL."
                },
                {
                    "title": "Spatiotemporal Contrastive Video Representation Learning",
                    "abstract": "We present a self-supervised Contrastive Video Representation Learning (CVRL) method to learn spatiotemporal visual representations from unlabeled videos. Our representations are learned using a contrastive loss, where two augmented clips from the same short video are pulled together in the embedding space, while clips from different videos are pushed away. We study what makes for good data augmentations for video self-supervised learning and find that both spatial and temporal information are crucial. We carefully design data augmentations involving spatial and temporal cues. Concretely, we propose a temporally consistent spatial augmentation method to impose strong spatial augmentations on each frame of the video while maintaining the temporal consistency across frames. We also propose a sampling-based temporal augmentation method to avoid overly enforcing invariance on clips that are distant in time. On Kinetics-600, a linear classifier trained on the representations learned by CVRL achieves 70.4% top-1 accuracy with a 3D-ResNet-50 (R3D-50) backbone, outperforming ImageNet supervised pre-training by 15.7% and SimCLR unsupervised pre-training by 18.8% using the same inflated R3D-50. The performance of CVRL can be further improved to 72.9% with a larger R3D-152 (2\u00d7 filters) backbone, significantly closing the gap between unsupervised and supervised video representation learning. Our code and models will be available at https://github.com/tensorflow/models/tree/master/official/."
                },
                {
                    "title": "Autoencoder Image Interpolation by Shaping the Latent Space",
                    "abstract": "Autoencoders represent an effective approach for computing the underlying factors characterizing datasets of different types. The latent representation of autoencoders have been studied in the context of enabling interpolation between data points by decoding convex combinations of latent vectors. This interpolation, however, often leads to artifacts or produces unrealistic results during reconstruction. We argue that these incongruities are due to the structure of the latent space and because such naively interpolated latent vectors deviate from the data manifold. In this paper, we propose a regularization technique that shapes the latent representation to follow a manifold that is consistent with the training images and that drives the manifold to be smooth and locally convex. This regularization not only enables faithful interpolation between data points, as we show herein, but can also be used as a general regularization technique to avoid overfitting or to produce new samples for data augmentation."
                },
                {
                    "title": "Data-Efficient Reinforcement Learning with Self-Predictive Representations",
                    "abstract": "While deep reinforcement learning excels at solving tasks where large amounts of data can be collected through virtually unlimited interaction with the environment, learning from limited interaction remains a key challenge. We posit that an agent can learn more efficiently if we augment reward maximization with self-supervised objectives based on structure in its visual input and sequential interaction with the environment. Our method, Self-Predictive Representations(SPR), trains an agent to predict its own latent state representations multiple steps into the future. We compute target representations for future states using an encoder which is an exponential moving average of the agent's parameters and we make predictions using a learned transition model. On its own, this future prediction objective outperforms prior methods for sample-efficient deep RL from pixels. We further improve performance by adding data augmentation to the future prediction loss, which forces the agent's representations to be consistent across multiple views of an observation. Our full self-supervised objective, which combines future prediction and data augmentation, achieves a median human-normalized score of 0.415 on Atari in a setting limited to 100k steps of environment interaction, which represents a 55% relative improvement over the previous state-of-the-art. Notably, even in this limited data regime, SPR exceeds expert human scores on 7 out of 26 games. The code associated with this work is available at https://github.com/mila-iqia/spr"
                },
                {
                    "title": "Control-Aware Representations for Model-based Reinforcement Learning",
                    "abstract": "A major challenge in modern reinforcement learning (RL) is efficient control of dynamical systems from high-dimensional sensory observations. Learning controllable embedding (LCE) is a promising approach that addresses this challenge by embedding the observations into a lower-dimensional latent space, estimating the latent dynamics, and utilizing it to perform control in the latent space. Two important questions in this area are how to learn a representation that is amenable to the control problem at hand, and how to achieve an end-to-end framework for representation learning and control. In this paper, we take a few steps towards addressing these questions. We first formulate a LCE model to learn representations that are suitable to be used by a policy iteration style algorithm in the latent space. We call this model control-aware representation learning (CARL). We derive a loss function for CARL that has close connection to the prediction, consistency, and curvature (PCC) principle for representation learning. We derive three implementations of CARL. In the offline implementation, we replace the locally-linear control algorithm (e.g.,~iLQR) used by the existing LCE methods with a RL algorithm, namely model-based soft actor-critic, and show that it results in significant improvement. In online CARL, we interleave representation learning and control, and demonstrate further gain in performance. Finally, we propose value-guided CARL, a variation in which we optimize a weighted version of the CARL loss function, where the weights depend on the TD-error of the current policy. We evaluate the proposed algorithms by extensive experiments on benchmark tasks and compare them with several LCE baselines."
                },
                {
                    "title": "S3VAE: Self-Supervised Sequential VAE for Representation Disentanglement and Data Generation",
                    "abstract": "We propose a sequential variational autoencoder to learn disentangled representations of sequential data (e.g., videos and audios) under self-supervision. Specifically, we exploit the benefits of some readily accessible supervision signals from input data itself or some off-the-shelf functional models and accordingly design auxiliary tasks for our model to utilize these signals. With the supervision of the signals, our model can easily disentangle the representation of an input sequence into static factors and dynamic factors (i.e., time-invariant and time-varying parts). Comprehensive experiments across videos and audios verify the effectiveness of our model on representation disentanglement and generation of sequential data, and demonstrate that, our model with self-supervision performs comparable to, if not better than, the fully-supervised model with ground truth labels, and outperforms state-of-the-art unsupervised models by a large margin."
                },
                {
                    "title": "Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hypersphere",
                    "abstract": "Contrastive representation learning has been outstandingly successful in practice. In this work, we identify two key properties related to the contrastive loss: (1) alignment (closeness) of features from positive pairs, and (2) uniformity of the induced distribution of the (normalized) features on the hypersphere. We prove that, asymptotically, the contrastive loss optimizes these properties, and analyze their positive effects on downstream tasks. Empirically, we introduce an optimizable metric to quantify each property. Extensive experiments on standard vision and language datasets confirm the strong agreement between both metrics and downstream task performance. Remarkably, directly optimizing for these two metrics leads to representations with comparable or better performance at downstream tasks than contrastive learning. \nProject Page: this https URL \nCode: this https URL , this https URL"
                },
                {
                    "title": "D4RL: Datasets for Deep Data-Driven Reinforcement Learning",
                    "abstract": "The offline reinforcement learning (RL) problem, also known as batch RL, refers to the setting where a policy must be learned from a static dataset, without additional online data collection. This setting is compelling as potentially it allows RL methods to take advantage of large, pre-collected datasets, much like how the rise of large datasets has fueled results in supervised learning in recent years. However, existing online RL benchmarks are not tailored towards the offline setting, making progress in offline RL difficult to measure. In this work, we introduce benchmarks specifically designed for the offline setting, guided by key properties of datasets relevant to real-world applications of offline RL. Examples of such properties include: datasets generated via hand-designed controllers and human demonstrators, multi-objective datasets where an agent can perform different tasks in the same environment, and datasets consisting of a mixtures of policies. To facilitate research, we release our benchmark tasks and datasets with a comprehensive evaluation of existing algorithms and an evaluation protocol together with an open-source codebase. We hope that our benchmark will focus research effort on methods that drive improvements not just on simulated tasks, but ultimately on the kinds of real-world problems where offline RL will have the largest impact."
                },
                {
                    "title": "Improved Baselines with Momentum Contrastive Learning",
                    "abstract": "Contrastive unsupervised learning has recently shown encouraging progress, e.g., in Momentum Contrast (MoCo) and SimCLR. In this note, we verify the effectiveness of two of SimCLR's design improvements by implementing them in the MoCo framework. With simple modifications to MoCo---namely, using an MLP projection head and more data augmentation---we establish stronger baselines that outperform SimCLR and do not require large training batches. We hope this will make state-of-the-art unsupervised learning research more accessible. Code will be made public."
                },
                {
                    "title": "Predictive Coding for Locally-Linear Control",
                    "abstract": "High-dimensional observations and unknown dynamics are major challenges when applying optimal control to many real-world decision making tasks. The Learning Controllable Embedding (LCE) framework addresses these challenges by embedding the observations into a lower dimensional latent space, estimating the latent dynamics, and then performing control directly in the latent space. To ensure the learned latent dynamics are predictive of next-observations, all existing LCE approaches decode back into the observation space and explicitly perform next-observation prediction---a challenging high-dimensional task that furthermore introduces a large number of nuisance parameters (i.e., the decoder) which are discarded during control. In this paper, we propose a novel information-theoretic LCE approach and show theoretically that explicit next-observation prediction can be replaced with predictive coding. We then use predictive coding to develop a decoder-free LCE model whose latent dynamics are amenable to locally-linear control. Extensive experiments on benchmark tasks show that our model reliably learns a controllable latent space that leads to superior performance when compared with state-of-the-art LCE baselines."
                },
                {
                    "title": "Universal Value Density Estimation for Imitation Learning and Goal-Conditioned Reinforcement Learning",
                    "abstract": "This work considers two distinct settings: imitation learning and goal-conditioned reinforcement learning. In either case, effective solutions require the agent to reliably reach a specified state (a goal), or set of states (a demonstration). Drawing a connection between probabilistic long-term dynamics and the desired value function, this work introduces an approach which utilizes recent advances in density estimation to effectively learn to reach a given state. As our first contribution, we use this approach for goal-conditioned reinforcement learning and show that it is both efficient and does not suffer from hindsight bias in stochastic domains. As our second contribution, we extend the approach to imitation learning and show that it achieves state-of-the art demonstration sample-efficiency on standard benchmark tasks."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "Unsupervised State Representation Learning in Atari",
                    "abstract": "State representation learning, or the ability to capture latent generative factors of an environment, is crucial for building intelligent agents that can perform a wide variety of tasks. Learning such representations without supervision from rewards is a challenging open problem. We introduce a method that learns state representations by maximizing mutual information across spatially and temporally distinct features of a neural encoder of the observations. We also introduce a new benchmark based on Atari 2600 games where we evaluate representations based on how well they capture the ground truth state variables. We believe this new framework for evaluating representation learning models will be crucial for future representation learning research. Finally, we compare our technique with other state-of-the-art generative and contrastive representation learning methods. The code associated with this work is available at this https URL"
                },
                {
                    "title": "Search on the Replay Buffer: Bridging Planning and Reinforcement Learning",
                    "abstract": "The history of learning for control has been an exciting back and forth between two broad classes of algorithms: planning and reinforcement learning. Planning algorithms effectively reason over long horizons, but assume access to a local policy and distance metric over collision-free paths. Reinforcement learning excels at learning policies and the relative values of states, but fails to plan over long horizons. Despite the successes of each method in various domains, tasks that require reasoning over long horizons with limited feedback and high-dimensional observations remain exceedingly challenging for both planning and reinforcement learning algorithms. Frustratingly, these sorts of tasks are potentially the most useful, as they are simple to design (a human only need to provide an example goal state) and avoid reward shaping, which can bias the agent towards finding a sub-optimal solution. We introduce a general control algorithm that combines the strengths of planning and reinforcement learning to effectively solve these tasks. Our aim is to decompose the task of reaching a distant goal state into a sequence of easier tasks, each of which corresponds to reaching a subgoal. Planning algorithms can automatically find these waypoints, but only if provided with suitable abstractions of the environment -- namely, a graph consisting of nodes and edges. Our main insight is that this graph can be constructed via reinforcement learning, where a goal-conditioned value function provides edge weights, and nodes are taken to be previously seen observations in a replay buffer. Using graph search over our replay buffer, we can automatically generate this sequence of subgoals, even in image-based environments. Our algorithm, search on the replay buffer (SoRB), enables agents to solve sparse reward tasks over one hundred steps, and generalizes substantially better than standard RL algorithms."
                },
                {
                    "title": "Generating Diverse High-Fidelity Images with VQ-VAE-2",
                    "abstract": "We explore the use of Vector Quantized Variational AutoEncoder (VQ-VAE) models for large scale image generation. To this end, we scale and enhance the autoregressive priors used in VQ-VAE to generate synthetic samples of much higher coherence and fidelity than possible before. We use simple feed-forward encoder and decoder networks, making our model an attractive candidate for applications where the encoding and/or decoding speed is critical. Additionally, VQ-VAE requires sampling an autoregressive model only in the compressed latent space, which is an order of magnitude faster than sampling in the pixel space, especially for large images. We demonstrate that a multi-scale hierarchical organization of VQ-VAE, augmented with powerful priors over the latent codes, is able to generate samples with quality that rivals that of state of the art Generative Adversarial Networks on multifaceted datasets such as ImageNet, while not suffering from GAN's known shortcomings such as mode collapse and lack of diversity."
                },
                {
                    "title": "Homomorphic Latent Space Interpolation for Unpaired Image-To-Image Translation",
                    "abstract": "Generative adversarial networks have achieved great success in unpaired image-to-image translation. Cycle consistency allows modeling the relationship between two distinct domains without paired data. In this paper, we propose an alternative framework, as an extension of latent space interpolation, to consider the intermediate region between two domains during translation. It is based on the fact that in a flat and smooth latent space, there exist many paths that connect two sample points. Properly selecting paths makes it possible to change only certain image attributes, which is useful for generating intermediate images between the two domains. We also show that this framework can be applied to multi-domain and multi-modal translation. Extensive experiments manifest its generality and applicability to various tasks."
                },
                {
                    "title": "On Variational Bounds of Mutual Information",
                    "abstract": "Estimating and optimizing Mutual Information (MI) is core to many problems in machine learning; however, bounding MI in high dimensions is challenging. To establish tractable and scalable objectives, recent work has turned to variational bounds parameterized by neural networks, but the relationships and tradeoffs between these bounds remains unclear. In this work, we unify these recent developments in a single framework. We find that the existing variational lower bounds degrade when the MI is large, exhibiting either high bias or high variance. To address this problem, we introduce a continuum of lower bounds that encompasses previous bounds and flexibly trades off bias and variance. On high-dimensional, controlled problems, we empirically characterize the bias and variance of the bounds and their gradients and demonstrate the effectiveness of our new bounds for estimation and representation learning."
                },
                {
                    "title": "A Theoretical Analysis of Contrastive Unsupervised Representation Learning",
                    "abstract": "Recent empirical works have successfully used unlabeled data to learn feature representations that are broadly useful in downstream classification tasks. Several of these methods are reminiscent of the well-known word2vec embedding algorithm: leveraging availability of pairs of semantically \"similar\" data points and \"negative samples,\" the learner forces the inner product of representations of similar pairs with each other to be higher on average than with negative samples. The current paper uses the term contrastive learning for such algorithms and presents a theoretical framework for analyzing them by introducing latent classes and hypothesizing that semantically similar points are sampled from the same latent class. This framework allows us to show provable guarantees on the performance of the learned representations on the average classification task that is comprised of a subset of the same set of latent classes. Our generalization bound also shows that learned representations can reduce (labeled) sample complexity on downstream tasks. We conduct controlled experiments in both the text and image domains to support the theory."
                },
                {
                    "title": "Analogies Explained: Towards Understanding Word Embeddings",
                    "abstract": "Word embeddings generated by neural network methods such as word2vec (W2V) are well known to exhibit seemingly linear behaviour, e.g. the embeddings of analogy \"woman is to queen as man is to king\" approximately describe a parallelogram. This property is particularly intriguing since the embeddings are not trained to achieve it. Several explanations have been proposed, but each introduces assumptions that do not hold in practice. We derive a probabilistically grounded definition of paraphrasing that we re-interpret as word transformation, a mathematical description of \"$w_x$ is to $w_y$\". From these concepts we prove existence of linear relationships between W2V-type embeddings that underlie the analogical phenomenon, identifying explicit error terms."
                },
                {
                    "title": "Noise Contrastive Estimation and Negative Sampling for Conditional Models: Consistency and Statistical Efficiency",
                    "abstract": "Noise Contrastive Estimation (NCE) is a powerful parameter estimation method for log-linear models, which avoids calculation of the partition function or its derivatives at each training step, a computationally demanding step in many cases. It is closely related to negative sampling methods, now widely used in NLP. This paper considers NCE-based estimation of conditional models. Conditional models are frequently encountered in practice; however there has not been a rigorous theoretical analysis of NCE in this setting, and we will argue there are subtle but important questions when generalizing NCE to the conditional case. In particular, we analyze two variants of NCE for conditional models: one based on a classification objective, the other based on a ranking objective. We show that the ranking-based variant of NCE gives consistent parameter estimates under weaker assumptions than the classification-based method; we analyze the statistical efficiency of the ranking-based and classification-based variants of NCE; finally we describe experiments on synthetic data and language modeling showing the effectiveness and tradeoffs of both methods."
                },
                {
                    "title": "Data Augmentation via Latent Space Interpolation for Image Classification",
                    "abstract": "Effective training of the deep neural networks requires much data to avoid underdetermined and poor generalization. Data Augmentation alleviates this by using existing data more effectively. However standard data augmentation produces only limited plausible alternative data by for example, flipping, distorting, adding noise to, cropping a patch from the original samples. In this paper, we introduce the adversarial autoencoder (AAE) to impose the feature representations with uniform distribution and apply the linear interpolation on latent space, which is potential to generate a much broader set of augmentations for image classification. As a possible \u201crecognition via generation\u201d framework, it has potentials for several other classification tasks. Our experiments on the ILSVRC 2012, CIFAR-10 datasets show that the latent space interpolation (LSI) improves the generalization and performance of state-of-the-art deep neural networks."
                },
                {
                    "title": "Representation Learning with Contrastive Predictive Coding",
                    "abstract": "While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments."
                },
                {
                    "title": "Unsupervised Feature Learning via Non-parametric Instance Discrimination",
                    "abstract": "Neural net classifiers trained on data with annotated class labels can also capture apparent visual similarity among categories without being directed to do so. We study whether this observation can be extended beyond the conventional domain of supervised learning: Can we learn a good feature representation that captures apparent similarity among instances, instead of classes, by merely asking the feature to be discriminative of individual instances? We formulate this intuition as a non-parametric classification problem at the instance-level, and use noise-contrastive estimation to tackle the computational challenges imposed by the large number of instance classes. Our experimental results demonstrate that, under unsupervised learning settings, our method surpasses the state-of-the-art on ImageNet classification by a large margin. Our method is also remarkable for consistently improving test performance with more training data and better network architectures. By fine-tuning the learned feature, we further obtain competitive results for semi-supervised learning and object detection tasks. Our non-parametric model is highly compact: With 128 features per image, our method requires only 600MB storage for a million images, enabling fast nearest neighbour retrieval at the run time."
                },
                {
                    "title": "Robust Locally-Linear Controllable Embedding",
                    "abstract": "Embed-to-control (E2C) is a model for solving high-dimensional optimal control problems by combining variational auto-encoders with locally-optimal controllers. However, the E2C model suffers from two major drawbacks: 1) its objective function does not correspond to the likelihood of the data sequence and 2) the variational encoder used for embedding typically has large variational approximation error, especially when there is noise in the system dynamics. In this paper, we present a new model for learning robust locally-linear controllable embedding (RCE). Our model directly estimates the predictive conditional density of the future observation given the current one, while introducing the bottleneck between the current and future observations. Although the bottleneck provides a natural embedding candidate for control, our RCE model introduces additional specific structures in the generative graphical model so that the model dynamics can be robustly linearized. We also propose a principled variational approximation of the embedding posterior that takes the future observation into account, and thus, makes the variational approximation more robust against the noise. Experimental results show that RCE outperforms the E2C model, and does so significantly when the underlying dynamics is noisy."
                },
                {
                    "title": "Time-Contrastive Networks: Self-Supervised Learning from Video",
                    "abstract": "We propose a self-supervised approach for learning representations and robotic behaviors entirely from unlabeled videos recorded from multiple viewpoints, and study how this representation can be used in two robotic imitation settings: imitating object interactions from videos of humans, and imitating human poses. Imitation of human behavior requires a viewpoint-invariant representation that captures the relationships between end-effectors (hands or robot grippers) and the environment, object attributes, and body pose. We train our representations using a triplet loss, where multiple simultaneous viewpoints of the same observation are attracted in the embedding space, while being repelled from temporal neighbors which are often visually similar but functionally different. This signal causes our model to discover attributes that do not change across viewpoint, but do change across time, while ignoring nuisance variables such as occlusions, motion blur, lighting and background. We demonstrate that this representation can be used by a robot to directly mimic human poses without an explicit correspondence, and that it can be used as a reward function within a reinforcement learning algorithm. While representations are learned from an unlabeled collection of task-related videos, robot behaviors such as pouring are learned by watching a single 3rd-person demonstration by a human. Reward functions obtained by following the human demonstrations under the learned representation enable efficient reinforcement learning that is practical for real-world robotic systems. Video results, open-source code and dataset are available at sermanet.github.io/imitate"
                },
                {
                    "title": "Towards Deeper Understanding of Variational Autoencoding Models",
                    "abstract": "We propose a new family of optimization criteria for variational auto-encoding models, generalizing the standard evidence lower bound. We provide conditions under which they recover the data distribution and learn latent features, and formally show that common issues such as blurry samples and uninformative latent features arise when these conditions are not met. Based on these new insights, we propose a new sequential VAE model that can generate sharp samples on the LSUN image dataset based on pixel-wise reconstruction loss, and propose an optimization criterion that encourages unsupervised learning of informative latent features."
                },
                {
                    "title": "Improved Deep Metric Learning with Multi-class N-pair Loss Objective",
                    "abstract": "Deep metric learning has gained much popularity in recent years, following the success of deep learning. However, existing frameworks of deep metric learning based on contrastive loss and triplet loss often suffer from slow convergence, partially because they employ only one negative example while not interacting with the other negative classes in each update. In this paper, we propose to address this problem with a new metric learning objective called multi-class N-pair loss. The proposed objective function firstly generalizes triplet loss by allowing joint comparison among more than one negative examples - more specifically, N-1 negative examples - and secondly reduces the computational burden of evaluating deep embedding vectors via an efficient batch construction strategy using only N pairs of examples, instead of (N+1) x N. We demonstrate the superiority of our proposed loss to the triplet loss as well as other competing loss functions for a variety of tasks on several visual recognition benchmark, including fine-grained object recognition and verification, image clustering and retrieval, and face verification and identification."
                },
                {
                    "title": "Word Embeddings as Metric Recovery in Semantic Spaces",
                    "abstract": "Continuous word representations have been remarkably useful across NLP tasks but remain poorly understood. We ground word embeddings in semantic spaces studied in the cognitive-psychometric literature, taking these spaces as the primary objects to recover. To this end, we relate log co-occurrences of words in large corpora to semantic similarity assessments and show that co-occurrences are indeed consistent with an Euclidean semantic space hypothesis. Framing word embedding as metric recovery of a semantic space unifies existing word embedding algorithms, ties them to manifold learning, and demonstrates that existing algorithms are consistent metric recovery methods given co-occurrence counts from random walks. Furthermore, we propose a simple, principled, direct metric recovery algorithm that performs on par with the state-of-the-art word embedding and manifold learning methods. Finally, we complement recent focus on analogies by constructing two new inductive reasoning datasets\u2014series completion and classification\u2014and demonstrate that word embeddings can be used to solve them as well."
                },
                {
                    "title": "Successor Features for Transfer in Reinforcement Learning",
                    "abstract": "Transfer in reinforcement learning refers to the notion that generalization should occur not only within a task but also across tasks. We propose a transfer framework for the scenario where the reward function changes between tasks but the environment's dynamics remain the same. Our approach rests on two key ideas: \"successor features\", a value function representation that decouples the dynamics of the environment from the rewards, and \"generalized policy improvement\", a generalization of dynamic programming's policy improvement operation that considers a set of policies rather than a single one. Put together, the two ideas lead to an approach that integrates seamlessly within the reinforcement learning framework and allows the free exchange of information across tasks. The proposed method also provides performance guarantees for the transferred policy even before any learning has taken place. We derive two theorems that set our approach in firm theoretical ground and present experiments that show that it successfully promotes transfer in practice, significantly outperforming alternative methods in a sequence of navigation tasks and in the control of a simulated robotic arm."
                },
                {
                    "title": "Adversarially Learned Inference",
                    "abstract": "We introduce the adversarially learned inference (ALI) model, which jointly learns a generation network and an inference network using an adversarial process. The generation network maps samples from stochastic latent variables to the data space while the inference network maps training examples in data space to the space of latent variables. An adversarial game is cast between these two networks and a discriminative network is trained to distinguish between joint latent/data-space samples from the generative network and joint samples from the inference network. We illustrate the ability of the model to learn mutually coherent inference and generation networks through the inspections of model samples and reconstructions and confirm the usefulness of the learned representations by obtaining a performance competitive with state-of-the-art on the semi-supervised SVHN and CIFAR10 tasks."
                },
                {
                    "title": "Adversarial Autoencoders",
                    "abstract": "In this paper, we propose the\"adversarial autoencoder\"(AAE), which is a probabilistic autoencoder that uses the recently proposed generative adversarial networks (GAN) to perform variational inference by matching the aggregated posterior of the hidden code vector of the autoencoder with an arbitrary prior distribution. Matching the aggregated posterior to the prior ensures that generating from any part of prior space results in meaningful samples. As a result, the decoder of the adversarial autoencoder learns a deep generative model that maps the imposed prior to the data distribution. We show how the adversarial autoencoder can be used in applications such as semi-supervised classification, disentangling style and content of images, unsupervised clustering, dimensionality reduction and data visualization. We performed experiments on MNIST, Street View House Numbers and Toronto Face datasets and show that adversarial autoencoders achieve competitive results in generative modeling and semi-supervised classification tasks."
                },
                {
                    "title": "Model Predictive Path Integral Control using Covariance Variable Importance Sampling",
                    "abstract": "In this paper we develop a Model Predictive Path Integral (MPPI) control algorithm based on a generalized importance sampling scheme and perform parallel optimization via sampling using a Graphics Processing Unit (GPU). The proposed generalized importance sampling scheme allows for changes in the drift and diffusion terms of stochastic diffusion processes and plays a significant role in the performance of the model predictive control algorithm. We compare the proposed algorithm in simulation with a model predictive control version of differential dynamic programming."
                },
                {
                    "title": "Embed to Control: A Locally Linear Latent Dynamics Model for Control from Raw Images",
                    "abstract": "We introduce Embed to Control (E2C), a method for model learning and control of non-linear dynamical systems from raw pixel images. E2C consists of a deep generative model, belonging to the family of variational autoencoders, that learns to generate image trajectories from a latent space in which the dynamics is constrained to be locally linear. Our model is derived directly from an optimal control formulation in latent space, supports long-term prediction of image sequences and exhibits strong performance on a variety of complex control problems."
                },
                {
                    "title": "Slow and Steady Feature Analysis: Higher Order Temporal Coherence in Video",
                    "abstract": "How can unlabeled video augment visual learning? Existing methods perform \"slow\" feature analysis, encouraging the representations of temporally close frames to exhibit only small differences. While this standard approach captures the fact that high-level visual signals change slowly over time, it fails to capture how the visual content changes. We propose to generalize slow feature analysis to \"steady\" feature analysis. The key idea is to impose a prior that higher order derivatives in the learned feature space must be small. To this end, we train a convolutional neural network with a regularizer on tuples of sequential frames from unlabeled video. It encourages feature changes over time to be smooth, i.e., similar to the most recent changes. Using five diverse datasets, including unlabeled YouTube and KITTI videos, we demonstrate our method's impact on object, scene, and action recognition tasks. We further show that our features learned from unlabeled video can even surpass a standard heavily supervised pretraining approach."
                },
                {
                    "title": "Learning to Linearize Under Uncertainty",
                    "abstract": "Training deep feature hierarchies to solve supervised learning tasks has achieved state of the art performance on many problems in computer vision. However, a principled way in which to train such hierarchies in the unsupervised setting has remained elusive. In this work we suggest a new architecture and loss for training deep feature hierarchies that linearize the transformations observed in unlabeled natural video sequences. This is done by training a generative model to predict video frames. We also address the problem of inherent uncertainty in prediction by introducing latent variables that are non-deterministic functions of the input into the network architecture."
                },
                {
                    "title": "A Recurrent Latent Variable Model for Sequential Data",
                    "abstract": "In this paper, we explore the inclusion of latent random variables into the hidden state of a recurrent neural network (RNN) by combining the elements of the variational autoencoder. We argue that through the use of high-level latent random variables, the variational RNN (VRNN)1 can model the kind of variability observed in highly structured sequential data such as natural speech. We empirically evaluate the proposed model against other related sequential models on four speech datasets and one handwriting dataset. Our results show the important roles that latent random variables can play in the RNN dynamics."
                },
                {
                    "title": "A Latent Variable Model Approach to PMI-based Word Embeddings",
                    "abstract": "Semantic word embeddings represent the meaning of a word via a vector, and are created by diverse methods. Many use nonlinear operations on co-occurrence statistics, and have hand-tuned hyperparameters and reweighting methods. This paper proposes a new generative model, a dynamic version of the log-linear topic model of Mnih and Hinton (2007). The methodological novelty is to use the prior to compute closed form expressions for word statistics. This provides a theoretical justification for nonlinear models like PMI, word2vec, and GloVe, as well as some hyperparameter choices. It also helps explain why low-dimensional semantic embeddings contain linear algebraic structure that allows solution of word analogies, as shown by Mikolov et al. (2013a) and many subsequent papers. Experimental support is provided for the generative model assumptions, the most important of which is that latent word vectors are fairly uniformly dispersed in space."
                },
                {
                    "title": "Path integral control and state-dependent feedback.",
                    "abstract": "In this paper we address the problem of computing state-dependent feedback controls for path integral control problems. To this end we generalize the path integral control formula and utilize this to construct parametrized state-dependent feedback controllers. In addition, we show a relation between control and importance sampling: Better control, in terms of control cost, yields more efficient importance sampling, in terms of effective sample size. The optimal control provides a zero-variance estimate."
                },
                {
                    "title": "Linguistic Regularities in Sparse and Explicit Word Representations",
                    "abstract": "Recent work has shown that neuralembedded word representations capture many relational similarities, which can be recovered by means of vector arithmetic in the embedded space. We show that Mikolov et al.\u2019s method of first adding and subtracting word vectors, and then searching for a word similar to the result, is equivalent to searching for a word that maximizes a linear combination of three pairwise word similarities. Based on this observation, we suggest an improved method of recovering relational similarities, improving the state-of-the-art results on two recent word-analogy datasets. Moreover, we demonstrate that analogy recovery is not restricted to neural word embeddings, and that a similar amount of relational similarities can be recovered from traditional distributional word representations."
                },
                {
                    "title": "Efficient Estimation of Word Representations in Vector Space",
                    "abstract": "We propose two novel model architectures for computing continuous vector\nrepresentations of words from very large data sets. The quality of these\nrepresentations is measured in a word similarity task, and the results are\ncompared to the previously best performing techniques based on different types\nof neural networks. We observe large improvements in accuracy at much lower\ncomputational cost, i.e. it takes less than a day to learn high quality word\nvectors from a 1.6 billion words data set. Furthermore, we show that these\nvectors provide state-of-the-art performance on our test set for measuring\nsyntactic and semantic word similarities."
                },
                {
                    "title": "Noise-contrastive estimation: A new estimation principle for unnormalized statistical models",
                    "abstract": "We present a new estimation principle for parameterized statistical models. The idea is to perform nonlinear logistic regression to discriminate between the observed data and some artificially generated noise, using the model log-density function in the regression nonlinearity. We show that this leads to a consistent (convergent) estimator of the parameters, and analyze the asymptotic variance. In particular, the method is shown to directly work for unnormalized models, i.e. models where the density function does not integrate to one. The normalization constant can be estimated just like any other parameter. For a tractable ICA model, we compare the method with other estimation methods that can be used to learn unnormalized models, including score matching, contrastive divergence, and maximum-likelihood where the normalization constant is estimated with importance sampling. Simulations show that noise-contrastive estimation offers the best trade-off between computational and statistical efficiency. The method is then applied to the modeling of natural images: We show that the method can successfully estimate a large-scale two-layer model and a Markov random field."
                },
                {
                    "title": "A Generalized Path Integral Control Approach to Reinforcement Learning",
                    "abstract": "With the goal to generate more scalable algorithms with higher efficiency and fewer open parameters, reinforcement learning (RL) has recently moved towards combining classical techniques from optimal control and dynamic programming with modern learning techniques from statistical estimation theory. In this vein, this paper suggests to use the framework of stochastic optimal control with path integrals to derive a novel approach to RL with parameterized policies. While solidly grounded in value function estimation and optimal control based on the stochastic Hamilton-Jacobi-Bellman (HJB) equations, policy improvements can be transformed into an approximation problem of a path integral which has no open algorithmic parameters other than the exploration noise. The resulting algorithm can be conceived of as model-based, semi-model-based, or even model free, depending on how the learning problem is structured. The update equations have no danger of numerical instabilities as neither matrix inversions nor gradient learning rates are required. Our new algorithm demonstrates interesting similarities with previous RL research in the framework of probability matching and provides intuition why the slightly heuristically motivated probability matching approach can actually perform well. Empirical evaluations demonstrate significant performance improvements over gradient-based policy learning and scalability to high-dimensional control problems. Finally, a learning experiment on a simulated 12 degree-of-freedom robot dog illustrates the functionality of our algorithm in a complex robot learning scenario. We believe that Policy Improvement with Path Integrals (PI2) offers currently one of the most efficient, numerically robust, and easy to implement algorithms for RL based on trajectory roll-outs."
                },
                {
                    "title": "The Apache Point Observatory Galactic Evolution Experiment (APOGEE)",
                    "abstract": "The Apache Point Observatory Galactic Evolution Experiment (APOGEE), one of the programs in the Sloan Digital Sky Survey III (SDSS-III), has now completed its systematic, homogeneous spectroscopic survey sampling all major populations of the Milky Way. After a three-year observing campaign on the Sloan 2.5 m Telescope, APOGEE has collected a half million high-resolution (R \u223c 22,500), high signal-to-noise ratio (>100), infrared (1.51\u20131.70 \u03bcm) spectra for 146,000 stars, with time series information via repeat visits to most of these stars. This paper describes the motivations for the survey and its overall design\u2014hardware, field placement, target selection, operations\u2014and gives an overview of these aspects as well as the data reduction, analysis, and products. An index is also given to the complement of technical papers that describe various critical survey components in detail. Finally, we discuss the achieved survey performance and illustrate the variety of potential uses of the data products by way of a number of science demonstrations, which span from time series analysis of stellar spectral variations and radial velocity variations from stellar companions, to spatial maps of kinematics, metallicity, and abundance patterns across the Galaxy and as a function of age, to new views of the interstellar medium, the chemistry of star clusters, and the discovery of rare stellar species. As part of SDSS-III Data Release 12 and later releases, all of the APOGEE data products are publicly available."
                },
                {
                    "title": "Walk-Sums and Belief Propagation in Gaussian Graphical Models",
                    "abstract": "We present a new framework based on walks in a graph for analysis and inference in Gaussian graphical models. The key idea is to decompose the correlation between each pair of variables as a sum over all walks between those variables in the graph. The weight of each walk is given by a product of edgewise partial correlation coefficients. This representation holds for a large class of Gaussian graphical models which we call walk-summable. We give a precise characterization of this class of models, and relate it to other classes including diagonally dominant, attractive, non-frustrated, and pairwise-normalizable. We provide a walk-sum interpretation of Gaussian belief propagation in trees and of the approximate method of loopy belief propagation in graphs with cycles. The walk-sum perspective leads to a better understanding of Gaussian belief propagation and to stronger results for its convergence in loopy graphs."
                },
                {
                    "title": "Slow Feature Analysis: Unsupervised Learning of Invariances",
                    "abstract": "Invariant features of temporally varying signals are useful for analysis and classification. Slow feature analysis (SFA) is a new method for learning invariant or slowly varying features from a vectorial input signal. It is based on a nonlinear expansion of the input signal and application of principal component analysis to this expanded signal and its time derivative. It is guaranteed to find the optimal solution within a family of functions directly and can learn to extract a large number of decor-related features, which are ordered by their degree of invariance. SFA can be applied hierarchically to process high-dimensional input signals and extract complex features. SFA is applied first to complex cell tuning properties based on simple cell output, including disparity and motion. Then more complicated input-output functions are learned by repeated application of SFA. Finally, a hierarchical network of SFA modules is presented as a simple model of the visual system. The same unstructured network can learn translation, size, rotation, contrast, or, to a lesser degree, illumination invariance for one-dimensional objects, depending on only the training stimulus. Surprisingly, only a few training objects suffice to achieve good generalization to new objects. The generated representation is suitable for object recognition. Performance degrades if the network is trained to learn multiple invariances simultaneously."
                },
                {
                    "title": "Probabilistic roadmaps for path planning in high-dimensional configuration spaces",
                    "abstract": "A new motion planning method for robots in static workspaces is presented. This method proceeds in two phases: a learning phase and a query phase. In the learning phase, a probabilistic roadmap is constructed and stored as a graph whose nodes correspond to collision-free configurations and whose edges correspond to feasible paths between these configurations. These paths are computed using a simple and fast local planner. In the query phase, any given start and goal configurations of the robot are connected to two nodes of the roadmap; the roadmap is then searched for a path joining these two nodes. The method is general and easy to implement. It can be applied to virtually any type of holonomic robot. It requires selecting certain parameters (e.g., the duration of the learning phase) whose values depend on the scene, that is the robot and its workspace. But these values turn out to be relatively easy to choose, Increased efficiency can also be achieved by tailoring some components of the method (e.g., the local planner) to the considered robots. In this paper the method is applied to planar articulated robots with many degrees of freedom. Experimental results show that path planning can be done in a fraction of a second on a contemporary workstation (/spl ap/150 MIPS), after learning for relatively short periods of time (a few dozen seconds)."
                },
                {
                    "title": "Improving Generalization for Temporal Difference Learning: The Successor Representation",
                    "abstract": "Estimation of returns over time, the focus of temporal difference (TD) algorithms, imposes particular constraints on good function approximators or representations. Appropriate generalization between states is determined by how similar their successors are, and representations should follow suit. This paper shows how TD machinery can be used to learn such representations, and illustrates, using a navigation task, the appropriately distributed nature of the result."
                },
                {
                    "title": "Time series analysis, forecasting and control",
                    "abstract": "time series analysis san francisco state university, 6 4 introduction to time series analysis, box and jenkins time series analysis forecasting and, th15 weeks citation classic eugene garfield, proc arima references 9 3 sas support, time series analysis forecasting and control pambudi, time series analysis forecasting and control george e, time series analysis forecasting and control ebook, time series analysis forecasting and control 5th edition, time series analysis forecasting and control fourth, time series analysis forecasting and control amazon, wiley time series analysis forecasting and control 5th, time series analysis forecasting and control edition 5, time series analysis forecasting and control 5th edition, time series analysis forecasting and control abebooks, time series analysis for business forecasting, time series analysis forecasting and control wiley, time series analysis forecasting and control book 1976, time series analysis forecasting and control researchgate, time series analysis forecasting and control edition 4, time series analysis forecasting amp control forecasting, george box publications department of statistics, time series analysis forecasting and control london, time series analysis forecasting and control an, time series analysis forecasting and control amazon it, box g e p and jenkins g m 1976 time series, time series analysis forecasting and control pdf slideshare, time series analysis forecasting and control researchgate, time series analysis forecasting and control 5th edition, time series analysis forecasting and control 5th edition, time series wikipedia, time series analysis forecasting and control abebooks, time series analysis forecasting and control, forecasting and time series analysis using the sca system, time series analysis forecasting and control by george e, time series analysis forecasting and control 5th edition, time series analysis forecasting and control 5th edition, box and jenkins time series analysis forecasting and control, time series analysis forecasting and control ebook, time series analysis forecasting and control, time series analysis and forecasting cengage, 6 7 references itl nist gov, time series analysis forecasting and control george e, time series analysis and forecasting statgraphics, time series analysis forecasting and control fourth edition, time series analysis forecasting and control, time series analysis forecasting and control wiley, time series analysis forecasting and control in"
                },
                {
                    "title": "The Evaluation of Matrix Inversion Programs",
                    "abstract": "It is tedious and difficult to make error estimates in a calculation of significant size. The number of genuine general error estimates which have been made is very small indeed, even in the case of matrix calculations, which are in many respects the basic and simplest calculations. Our feeling however is that it is not economically possible to carry out rigorous error estimates for all computational problems. We must be content to carry out \"experimental\" calculations on \"representative\" problems for which exact results are known, to observe the errors and to extrapolate from these to predict the errors in less academic problems. Once we have found a process which is experimentally satisfactory, then, if we can find theoretical error estimates, we can compare these with the observed errors and if they are in reasonable agreement use the theoretical estimates for prediction. These general remarks being made, we shall confine our attention to the problem of matrix inversion. Most groups operating large scale computing installations have prepared their own programs for inverting matrices and solving systems of equations, and all have a considerable amount of experience with these problems. In order to pick out the better methods we present in Chapter II a set of matrices which we have reason to believe form a representative sample of those which occur in practice. We have decided, for various reasons, generally to interpret \"better\" in the sense of \"least error\", disregarding time and storage considerations. We present in Chapter IV the results of some computational experiments carried out on these matrices at our own organization, and some results communicated to us by colleagues elsewhere. We also give a preliminary discussion of these results. We are continuing our experiments in directions indicated by the available results, and we hope that other organizations will also do so.3 We expect to prepare a more penetrating report later; this will include comparisons of the direct and indirect methods-at present, we have restricted ourselves to direct methods."
                },
                {
                    "title": "Information Theory and Statistical Mechanics",
                    "abstract": "Treatment of the predictive aspect of statistical mechanics as a form of statistical inference is extended to the density-matrix formalism and applied to a discussion of the relation between irreversibility and information loss. A principle of \"statistical complementarity\" is pointed out, according to which the empirically verifiable probabilities of statistical mechanics necessarily correspond to incomplete predictions. A preliminary discussion is given of the second law of thermodynamics and of a certain class of irreversible processes, in an approximation equivalent to that of the semiclassical theory of radiation."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "A Mathematical Theory of Communication",
                    "abstract": "This paper opened the new area the information theory. Before this paper, most people believed that the only way to make the error probability of transmission as small as desired is to reduce the data rate (such as a long repetition scheme). However, surprisingly this paper revealed that it does not need to reduce the data rate for achieving that much of small errors. It proved that we can get some positive data rate that has the same small error probability and also there is an upper bound of the data rate, which means we cannot achieve the data rate with any encoding scheme that has small enough error probability over the upper bound."
                },
                {
                    "title": "Planning by Probabilistic Inference",
                    "abstract": "This paper presents and demonstrates a new approach to the problem of planning under uncertainty. Actions are treated as hidden variables, with their own prior distributions, in a probabilistic generative model involving actions and states. Planning is done by computing the posterior distribution over actions, conditioned on reaching the goal state within a specified number of steps. Under the new formulation, the toolbox of inference techniques be brought to bear on the planning problem. This paper focuses on problems with discrete actions and states, and discusses some extensions."
                },
                {
                    "title": "Correctness of belief propagation in Gaussian graphical models of arbitrary topology",
                    "abstract": "Graphical models, such as Bayesian networks and Markov random elds represent statistical dependencies of variables by a graph. Local \\belief propagation\" rules of the sort proposed by Pearl [20] are guaranteed to converge to the correct posterior probabilities in singly connected graphs. Recently good performance has been obtained by using these same rules on graphs with loops, a method known as \\loopy belief propagation\". Perhaps the most dramatic instance is the near Shannon-limit performance of \\Turbo codes\", whose decoding algorithm is equivalent to loopy propagation. Except for the case of graphs with a single loop, there has been little theoretical understanding of loopy propagation. Here we analyze belief propagation in networks with arbitrary topologies when the nodes in the graph describe jointly Gaussian random variables. We give an analytical formula relating the true posterior probabilities with those calculated using loopy propagation. We give su cient conditions for convergence and show that when belief propagation converges it gives the correct posterior means for all graph topologies, not just networks with a single loop. The related \\max-product\" algorithm nds the maximumposterior probability estimate for singly connected networks. We show that, even for non-Gaussian probability distributions, the xed points of the max-product algorithm in loopy networks are at least local maxima of the posterior probability. These results motivate using the powerful belief propagation algorithm in a broader class of networks, and help clarify the empirical performance results. Sumbitted to Neural Computation. Preliminary version appeared in Proc. NIPS 99"
                },
                {
                    "title": "On the theory of brownian motion",
                    "abstract": "The spectrum of the Fokker-Planck operator for weakly coupled gases is considered. The operator is decomposed into operators acting on functions whose angular dependence is given by spherical harmonics. It is shown that the operator corresponding to l = 0 has zero for a point eigenvalue (the eigenfunction is the Maxwell distribution). There are no other point eigenvalues and the continuous spectrum of all of the operators is the entire negative real axis. Some consequences are briefly discussed."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can contrastive learning methods be effectively utilized to infer intermediate and future states in high-dimensional time-series data?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is significant for the research community as it could revolutionize the way we approach time-series analysis, particularly in fields like robotics, biology, and astrophysics. By leveraging discriminative methods like contrastive learning, researchers can develop more efficient and scalable models that can handle high-dimensional data without the computational burden of generative methods. This advancement could lead to new practical applications in predictive modeling and planning, enhancing our ability to understand and manipulate complex systems over time.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the high dimensionality of time-series data, which complicates the inference of intermediate states. Naive approaches may fail because they often rely on generative models that are computationally intensive and may not scale well. Additionally, ensuring that the learned representations retain the necessary temporal relationships while discarding irrelevant information poses a significant theoretical and practical obstacle. The need for representations to act as sufficient statistics for temporal relationships adds another layer of complexity.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on generative methods that, while effective, are not suitable for high-dimensional data due to their computational demands. There has been a lack of exploration into the potential of discriminative methods like contrastive learning for time-series inference. Barriers such as the difficulty in establishing a clear relationship between learned representations and temporal inference tasks have hindered progress. This paper aims to fill these gaps by demonstrating how contrastive methods can be adapted to effectively address the challenges of time-series data.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves using a regularized version of the symmetrized infoNCE objective to learn representations from pairs of observations sampled from the same time series. The dataset will consist of high-dimensional time-series data relevant to various applications. The metric for evaluation will focus on the accuracy of inferred intermediate and future states. The expected outcomes include demonstrating that contrastive representations can facilitate effective prediction and planning, leading to practical applications in interpolation, in-filling, and control tasks."
            }
        },
        "author_data": {
            "1228f3b0-3864-4e4f-874d-45e55539ab45": {
                "pk": "1228f3b0-3864-4e4f-874d-45e55539ab45",
                "name": "Benjamin Eysenbach",
                "collaborators": [
                    "Sergey Levine",
                    "Ruslan Salakhutdinov",
                    "Julian Ibarz",
                    "Matthieu Geist",
                    "Carl Vondrick",
                    "Antonio Torralba",
                    "Abhishek Gupta",
                    "Grace Liu",
                    "Michael Tang",
                    "Shixiang Gu"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Machine Learning",
                    "Control Theory",
                    "Multi-task Learning"
                ],
                "publications": [
                    {
                        "title": "If MaxEnt RL is the Answer, What is the Question?",
                        "abstract": "Experimentally, it has been observed that humans and animals often make decisions that do not maximize their expected utility, but rather choose outcomes randomly, with probability proportional to expected utility. Probability matching, as this strategy is called, is equivalent to maximum entropy reinforcement learning (MaxEnt RL). However, MaxEnt RL does not optimize expected utility. In this paper, we formally show that MaxEnt RL does optimally solve certain classes of control problems with variability in the reward function. In particular, we show (1) that MaxEnt RL can be used to solve a certain class of POMDPs, and (2) that MaxEnt RL is equivalent to a two-player game where an adversary chooses the reward function. These results suggest a deeper connection between MaxEnt RL, robust control, and POMDPs, and provide insight for the types of problems for which we might expect MaxEnt RL to produce effective solutions. Specifically, our results suggest that domains with uncertainty in the task goal may be especially well-suited for MaxEnt RL methods."
                    },
                    {
                        "title": "Maximum Entropy RL (Provably) Solves Some Robust RL Problems",
                        "abstract": "Many potential applications of reinforcement learning (RL) require guarantees that the agent will perform well in the face of disturbances to the dynamics or reward function. In this paper, we prove theoretically that maximum entropy (MaxEnt) RL maximizes a lower bound on a robust RL objective, and thus can be used to learn policies that are robust to some disturbances in the dynamics and the reward function. While this capability of MaxEnt RL has been observed empirically in prior work, to the best of our knowledge our work provides the first rigorous proof and theoretical characterization of the MaxEnt RL robust set. While a number of prior robust RL algorithms have been designed to handle similar disturbances to the reward function or dynamics, these methods typically require additional moving parts and hyperparameters on top of a base RL algorithm. In contrast, our results suggest that MaxEnt RL by itself is robust to certain disturbances, without requiring any additional modifications. While this does not imply that MaxEnt RL is the best available robust RL method, MaxEnt RL is a simple robust RL method with appealing formal guarantees."
                    },
                    {
                        "title": "A Connection between One-Step Regularization and Critic Regularization in Reinforcement Learning",
                        "abstract": "As with any machine learning problem with limited data, effective offline RL algorithms require careful regularization to avoid overfitting. One-step methods perform regularization by doing just a single step of policy improvement, while critic regularization methods do many steps of policy improvement with a regularized objective. These methods appear distinct. One-step methods, such as advantage-weighted regression and conditional behavioral cloning, truncate policy iteration after just one step. This ``early stopping'' makes one-step RL simple and stable, but can limit its asymptotic performance. Critic regularization typically requires more compute but has appealing lower-bound guarantees. In this paper, we draw a close connection between these methods: applying a multi-step critic regularization method with a regularization coefficient of 1 yields the same policy as one-step RL. While practical implementations violate our assumptions and critic regularization is typically applied with smaller regularization coefficients, our experiments nevertheless show that our analysis makes accurate, testable predictions about practical offline RL methods (CQL and one-step RL) with commonly-used hyperparameters. Our results that every problem can be solved with a single step of policy improvement, but rather that one-step RL might be competitive with critic regularization on RL problems that demand strong regularization."
                    },
                    {
                        "title": "Who is Mistaken?",
                        "abstract": "Recognizing when people have false beliefs is crucial for understanding their actions. We introduce the novel problem of identifying when people in abstract scenes have incorrect beliefs. We present a dataset of scenes, each visually depicting an 8-frame story in which a character has a mistaken belief. We then create a representation of characters' beliefs for two tasks in human action understanding: predicting who is mistaken, and when they are mistaken. Experiments suggest that our method for identifying mistaken characters performs better on these tasks than simple baselines. Diagnostics on our model suggest it learns important cues for recognizing mistaken beliefs, such as gaze. We believe models of people's beliefs will have many applications in action understanding, robotics, and healthcare."
                    },
                    {
                        "title": "Diversity is All You Need: Learning Skills without a Reward Function",
                        "abstract": "Intelligent creatures can explore their environments and learn useful skills without supervision. In this paper, we propose DIAYN ('Diversity is All You Need'), a method for learning useful skills without a reward function. Our proposed method learns skills by maximizing an information theoretic objective using a maximum entropy policy. On a variety of simulated robotic tasks, we show that this simple objective results in the unsupervised emergence of diverse skills, such as walking and jumping. In a number of reinforcement learning benchmark environments, our method is able to learn a skill that solves the benchmark task despite never receiving the true task reward. We show how pretrained skills can provide a good parameter initialization for downstream tasks, and can be composed hierarchically to solve complex, sparse reward tasks. Our results suggest that unsupervised discovery of skills can serve as an effective pretraining mechanism for overcoming challenges of exploration and data efficiency in reinforcement learning."
                    },
                    {
                        "title": "Robust Predictable Control",
                        "abstract": "Many of the challenges facing today's reinforcement learning (RL) algorithms, such as robustness, generalization, transfer, and computational efficiency are closely related to compression. Prior work has convincingly argued why minimizing information is useful in the supervised learning setting, but standard RL algorithms lack an explicit mechanism for compression. The RL setting is unique because (1) its sequential nature allows an agent to use past information to avoid looking at future observations and (2) the agent can optimize its behavior to prefer states where decision making requires few bits. We take advantage of these properties to propose a method (RPC) for learning simple policies. This method brings together ideas from information bottlenecks, model-based RL, and bits-back coding into a simple and theoretically-justified algorithm. Our method jointly optimizes a latent-space model and policy to be self-consistent, such that the policy avoids states where the model is inaccurate. We demonstrate that our method achieves much tighter compression than prior methods, achieving up to 5x higher reward than a standard information bottleneck. We also demonstrate that our method learns policies that are more robust and generalize better to new tasks."
                    },
                    {
                        "title": "Replacing Rewards with Examples: Example-Based Policy Search via Recursive Classification",
                        "abstract": "Reinforcement learning (RL) algorithms assume that users specify tasks by manually writing down a reward function. However, this process can be laborious and demands considerable technical expertise. Can we devise RL algorithms that instead enable users to specify tasks simply by providing examples of successful outcomes? In this paper, we derive a control algorithm that maximizes the future probability of these successful outcome examples. Prior work has approached similar problems with a two-stage process, first learning a reward function and then optimizing this reward function using another RL algorithm. In contrast, our method directly learns a value function from transitions and successful outcomes, without learning this intermediate reward function. Our method therefore requires fewer hyperparameters to tune and lines of code to debug. We show that our method satisfies a new data-driven Bellman equation, where examples take the place of the typical reward function term. Experiments show that our approach outperforms prior methods that learn explicit reward functions."
                    },
                    {
                        "title": "The Information Geometry of Unsupervised Reinforcement Learning",
                        "abstract": "How can a reinforcement learning (RL) agent prepare to solve downstream tasks if those tasks are not known a priori? One approach is unsupervised skill discovery, a class of algorithms that learn a set of policies without access to a reward function. Such algorithms bear a close resemblance to representation learning algorithms (e.g., contrastive learning) in supervised learning, in that both are pretraining algorithms that maximize some approximation to a mutual information objective. While prior work has shown that the set of skills learned by such methods can accelerate downstream RL tasks, prior work offers little analysis into whether these skill learning algorithms are optimal, or even what notion of optimality would be appropriate to apply to them. In this work, we show that unsupervised skill discovery algorithms based on mutual information maximization do not learn skills that are optimal for every possible reward function. However, we show that the distribution over skills provides an optimal initialization minimizing regret against adversarially-chosen reward functions, assuming a certain type of adaptation procedure. Our analysis also provides a geometric perspective on these skill learning methods."
                    },
                    {
                        "title": "A Single Goal is All You Need: Skills and Exploration Emerge from Contrastive RL without Rewards, Demonstrations, or Subgoals",
                        "abstract": "In this paper, we present empirical evidence of skills and directed exploration emerging from a simple RL algorithm long before any successful trials are observed. For example, in a manipulation task, the agent is given a single observation of the goal state and learns skills, first for moving its end-effector, then for pushing the block, and finally for picking up and placing the block. These skills emerge before the agent has ever successfully placed the block at the goal location and without the aid of any reward functions, demonstrations, or manually-specified distance metrics. Once the agent has learned to reach the goal state reliably, exploration is reduced. Implementing our method involves a simple modification of prior work and does not require density estimates, ensembles, or any additional hyperparameters. Intuitively, the proposed method seems like it should be terrible at exploration, and we lack a clear theoretical understanding of why it works so effectively, though our experiments provide some hints."
                    },
                    {
                        "title": "Leave no Trace: Learning to Reset for Safe and Autonomous Reinforcement Learning",
                        "abstract": "Deep reinforcement learning algorithms can learn complex behavioral skills, but real-world application of these methods requires a large amount of experience to be collected by the agent. In practical settings, such as robotics, this involves repeatedly attempting a task, resetting the environment between each attempt. However, not all tasks are easily or automatically reversible. In practice, this learning process requires extensive human intervention. In this work, we propose an autonomous method for safe and efficient reinforcement learning that simultaneously learns a forward and reset policy, with the reset policy resetting the environment for a subsequent attempt. By learning a value function for the reset policy, we can automatically determine when the forward policy is about to enter a non-reversible state, providing for uncertainty-aware safety aborts. Our experiments illustrate that proper use of the reset policy can greatly reduce the number of manual resets required to learn a task, can reduce the number of unsafe actions that lead to non-reversible states, and can automatically induce a curriculum."
                    },
                    {
                        "title": "Rewriting History with Inverse RL: Hindsight Inference for Policy Improvement",
                        "abstract": "Multi-task reinforcement learning (RL) aims to simultaneously learn policies for solving many tasks. Several prior works have found that relabeling past experience with different reward functions can improve sample efficiency. Relabeling methods typically ask: if, in hindsight, we assume that our experience was optimal for some task, for what task was it optimal? In this paper, we show that hindsight relabeling is inverse RL, an observation that suggests that we can use inverse RL in tandem for RL algorithms to efficiently solve many tasks. We use this idea to generalize goal-relabeling techniques from prior work to arbitrary classes of tasks. Our experiments confirm that relabeling data using inverse RL accelerates learning in general multi-task settings, including goal-reaching, domains with discrete sets of rewards, and those with linear reward functions."
                    },
                    {
                        "title": "Imitating Past Successes can be Very Suboptimal",
                        "abstract": "Prior work has proposed a simple strategy for reinforcement learning (RL): label experience with the outcomes achieved in that experience, and then imitate the relabeled experience. These outcome-conditioned imitation learning methods are appealing because of their simplicity, strong performance, and close ties with supervised learning. However, it remains unclear how these methods relate to the standard RL objective, reward maximization. In this paper, we formally relate outcome-conditioned imitation learning to reward maximization, drawing a precise relationship between the learned policy and Q-values and explaining the close connections between these methods and prior EM-based policy search methods. This analysis shows that existing outcome-conditioned imitation learning methods do not necessarily improve the policy, but a simple modification results in a method that does guarantee policy improvement, under some assumptions."
                    },
                    {
                        "title": "C-Learning: Learning to Achieve Goals via Recursive Classification",
                        "abstract": "We study the problem of predicting and controlling the future state distribution of an autonomous agent. This problem, which can be viewed as a reframing of goal-conditioned reinforcement learning (RL), is centered around learning a conditional probability density function over future states. Instead of directly estimating this density function, we indirectly estimate this density function by training a classifier to predict whether an observation comes from the future. Via Bayes' rule, predictions from our classifier can be transformed into predictions over future states. Importantly, an off-policy variant of our algorithm allows us to predict the future state distribution of a new policy, without collecting new experience. This variant allows us to optimize functionals of a policy's future state distribution, such as the density of reaching a particular goal state. While conceptually similar to Q-learning, our work lays a principled foundation for goal-conditioned RL as density estimation, providing justification for goal-conditioned methods used in prior work. This foundation makes hypotheses about Q-learning, including the optimal goal-sampling ratio, which we confirm experimentally. Moreover, our proposed method is competitive with prior goal-conditioned RL methods."
                    },
                    {
                        "title": "Contrastive Learning as Goal-Conditioned Reinforcement Learning",
                        "abstract": "In reinforcement learning (RL), it is easier to solve a task if given a good representation. While deep RL should automatically acquire such good representations, prior work often finds that learning representations in an end-to-end fashion is unstable and instead equip RL algorithms with additional representation learning parts (e.g., auxiliary losses, data augmentation). How can we design RL algorithms that directly acquire good representations? In this paper, instead of adding representation learning parts to an existing RL algorithm, we show (contrastive) representation learning methods can be cast as RL algorithms in their own right. To do this, we build upon prior work and apply contrastive representation learning to action-labeled trajectories, in such a way that the (inner product of) learned representations exactly corresponds to a goal-conditioned value function. We use this idea to reinterpret a prior RL method as performing contrastive learning, and then use the idea to propose a much simpler method that achieves similar performance. Across a range of goal-conditioned RL tasks, we demonstrate that contrastive RL methods achieve higher success rates than prior non-contrastive methods, including in the offline RL setting. We also show that contrastive RL outperforms prior methods on image-based tasks, without using data augmentation or auxiliary objectives."
                    },
                    {
                        "title": "Off-Dynamics Reinforcement Learning: Training for Transfer with Domain Classifiers",
                        "abstract": "We propose a simple, practical, and intuitive approach for domain adaptation in reinforcement learning. Our approach stems from the idea that the agent's experience in the source domain should look similar to its experience in the target domain. Building off of a probabilistic view of RL, we formally show that we can achieve this goal by compensating for the difference in dynamics by modifying the reward function. This modified reward function is simple to estimate by learning auxiliary classifiers that distinguish source-domain transitions from target-domain transitions. Intuitively, the modified reward function penalizes the agent for visiting states and taking actions in the source domain which are not possible in the target domain. Said another way, the agent is penalized for transitions that would indicate that the agent is interacting with the source domain, rather than the target domain. Our approach is applicable to domains with continuous states and actions and does not require learning an explicit model of the dynamics. On discrete and continuous control tasks, we illustrate the mechanics of our approach and demonstrate its scalability to high-dimensional tasks."
                    },
                    {
                        "title": "Search on the Replay Buffer: Bridging Planning and Reinforcement Learning",
                        "abstract": "The history of learning for control has been an exciting back and forth between two broad classes of algorithms: planning and reinforcement learning. Planning algorithms effectively reason over long horizons, but assume access to a local policy and distance metric over collision-free paths. Reinforcement learning excels at learning policies and the relative values of states, but fails to plan over long horizons. Despite the successes of each method in various domains, tasks that require reasoning over long horizons with limited feedback and high-dimensional observations remain exceedingly challenging for both planning and reinforcement learning algorithms. Frustratingly, these sorts of tasks are potentially the most useful, as they are simple to design (a human only need to provide an example goal state) and avoid reward shaping, which can bias the agent towards finding a sub-optimal solution. We introduce a general control algorithm that combines the strengths of planning and reinforcement learning to effectively solve these tasks. Our aim is to decompose the task of reaching a distant goal state into a sequence of easier tasks, each of which corresponds to reaching a subgoal. Planning algorithms can automatically find these waypoints, but only if provided with suitable abstractions of the environment -- namely, a graph consisting of nodes and edges. Our main insight is that this graph can be constructed via reinforcement learning, where a goal-conditioned value function provides edge weights, and nodes are taken to be previously seen observations in a replay buffer. Using graph search over our replay buffer, we can automatically generate this sequence of subgoals, even in image-based environments. Our algorithm, search on the replay buffer (SoRB), enables agents to solve sparse reward tasks over one hundred steps, and generalizes substantially better than standard RL algorithms."
                    },
                    {
                        "title": "Interactive Visualization for Debugging RL",
                        "abstract": "Visualization tools for supervised learning allow users to interpret, introspect, and gain an intuition for the successes and failures of their models. While reinforcement learning practitioners ask many of the same questions, existing tools are not applicable to the RL setting as these tools address challenges typically found in the supervised learning regime. In this work, we design and implement an interactive visualization tool for debugging and interpreting RL algorithms. Our system addresses many features missing from previous tools such as (1) tools for supervised learning often are not interactive; (2) while debugging RL policies researchers use state representations that are different from those seen by the agent; (3) a framework designed to make the debugging RL policies more conducive. We provide an example workflow of how this system could be used, along with ideas for future extensions."
                    },
                    {
                        "title": "Recurrent Model-Free RL Can Be a Strong Baseline for Many POMDPs",
                        "abstract": "Many problems in RL, such as meta-RL, robust RL, generalization in RL, and temporal credit assignment, can be cast as POMDPs. In theory, simply augmenting model-free RL with memory-based architectures, such as recurrent neural networks, provides a general approach to solving all types of POMDPs. However, prior work has found that such recurrent model-free RL methods tend to perform worse than more specialized algorithms that are designed for specific types of POMDPs. This paper revisits this claim. We find that careful architecture and hyperparameter decisions can often yield a recurrent model-free implementation that performs on par with (and occasionally substantially better than) more sophisticated recent techniques. We compare to 21 environments from 6 prior specialized methods and find that our implementation achieves greater sample efficiency and asymptotic performance than these methods on 18/21 environments. We also release a simple and efficient implementation of recurrent model-free RL for future work to use as a baseline for POMDPs."
                    },
                    {
                        "title": "RvS: What is Essential for Offline RL via Supervised Learning?",
                        "abstract": "Recent work has shown that supervised learning alone, without temporal difference (TD) learning, can be remarkably effective for offline RL. When does this hold true, and which algorithmic components are necessary? Through extensive experiments, we boil supervised learning for offline RL down to its essential elements. In every environment suite we consider, simply maximizing likelihood with a two-layer feedforward MLP is competitive with state-of-the-art results of substantially more complex methods based on TD learning or sequence modeling with Transformers. Carefully choosing model capacity (e.g., via regularization or architecture) and choosing which information to condition on (e.g., goals or rewards) are critical for performance. These insights serve as a field guide for practitioners doing Reinforcement Learning via Supervised Learning (which we coin \"RvS learning\"). They also probe the limits of existing RvS methods, which are comparatively weak on random data, and suggest a number of open problems."
                    }
                ]
            },
            "f163df41-7bc1-4ac3-814e-466d32597d1f": {
                "pk": "f163df41-7bc1-4ac3-814e-466d32597d1f",
                "name": "Vivek Myers",
                "collaborators": [
                    "Erdem B\u0131y\u0131k",
                    "Dorsa Sadigh",
                    "Anca Dragan",
                    "Sergey Levine",
                    "Nikhil Sardana",
                    "Peyton Greenside",
                    "Nima Anari",
                    "Michelle Pan",
                    "Mariah Schrum",
                    "Chongyi Zheng"
                ],
                "domain": [
                    "Meta-Learning",
                    "Reinforcement Learning",
                    "Active Learning",
                    "Robotics"
                ],
                "publications": [
                    {
                        "title": "Bayesian Meta-Learning Through Variational Gaussian Processes",
                        "abstract": "Recent advances in the field of meta-learning have tackled domains consisting of large numbers of small (\"few-shot\") supervised learning tasks. Meta-learning algorithms must be able to rapidly adapt to any individual few-shot task, fitting to a small support set within a task and using it to predict the labels of the task's query set. This problem setting can be extended to the Bayesian context, wherein rather than predicting a single label for each query data point, a model predicts a distribution of labels capturing its uncertainty. Successful methods in this domain include Bayesian ensembling of MAML-based models, Bayesian neural networks, and Gaussian processes with learned deep kernel and mean functions. While Gaussian processes have a robust Bayesian interpretation in the meta-learning context, they do not naturally model non-Gaussian predictive posteriors for expressing uncertainty. In this paper, we design a theoretically principled method, VMGP, extending Gaussian-process-based meta-learning to allow for high-quality, arbitrary non-Gaussian uncertainty predictions. On benchmark environments with complex non-smooth or discontinuous structure, we find our VMGP method performs significantly better than existing Bayesian meta-learning baselines."
                    },
                    {
                        "title": "A Hierarchical Approach to Scaling Batch Active Search Over Structured Data",
                        "abstract": "Active search is the process of identifying high-value data points in a large and often high-dimensional parameter space that can be expensive to evaluate. Traditional active search techniques like Bayesian optimization trade off exploration and exploitation over consecutive evaluations, and have historically focused on single or small (<5) numbers of examples evaluated per round. As modern data sets grow, so does the need to scale active search to large data sets and batch sizes. In this paper, we present a general hierarchical framework based on bandit algorithms to scale active search to large batch sizes by maximizing information derived from the unique structure of each dataset. Our hierarchical framework, Hierarchical Batch Bandit Search (HBBS), strategically distributes batch selection across a learned embedding space by facilitating wide exploration of different structural elements within a dataset. We focus our application of HBBS on modern biology, where large batch experimentation is often fundamental to the research process, and demonstrate batch design of biological sequences (protein and DNA). We also present a new Gym environment to easily simulate diverse biological sequences and to enable more comprehensive evaluation of active search methods across heterogeneous data sets. The HBBS framework improves upon standard performance, wall-clock, and scalability benchmarks for batch search by using a broad exploration strategy across coarse partitions and fine-grained exploitation within each partition of structured data."
                    },
                    {
                        "title": "Active Reward Learning from Online Preferences",
                        "abstract": "Robot policies need to adapt to human preferences and/or new environments. Human experts may have the domain knowledge required to help robots achieve this adaptation. However, existing works often require costly offline re-training on human feedback, and those feedback usually need to be frequent and too complex for the humans to reliably provide. To avoid placing undue burden on human experts and allow quick adaptation in critical real-world situations, we propose designing and sparingly presenting easy-to-answer pairwise action preference queries in an online fashion. Our approach designs queries and determines when to present them to maximize the expected value derived from the queries' information. We demonstrate our approach with experiments in simulation, human user studies, and real robot experiments. In these settings, our approach outperforms baseline techniques while presenting fewer queries to human experts. Experiment videos, code and appendices are found at https://sites.google.com/view/onlineactivepreferences."
                    },
                    {
                        "title": "Learning Multimodal Rewards from Rankings",
                        "abstract": "Learning from human feedback has shown to be a useful approach in acquiring robot reward functions. However, expert feedback is often assumed to be drawn from an underlying unimodal reward function. This assumption does not always hold including in settings where multiple experts provide data or when a single expert provides data for different tasks -- we thus go beyond learning a unimodal reward and focus on learning a multimodal reward function. We formulate the multimodal reward learning as a mixture learning problem and develop a novel ranking-based learning approach, where the experts are only required to rank a given set of trajectories. Furthermore, as access to interaction data is often expensive in robotics, we develop an active querying approach to accelerate the learning process. We conduct experiments and user studies using a multi-task variant of OpenAI's LunarLander and a real Fetch robot, where we collect data from multiple users with different preferences. The results suggest that our approach can efficiently learn multimodal reward functions, and improve data-efficiency over benchmark methods that we adapt to our learning problem."
                    },
                    {
                        "title": "Coprocessor Actor Critic: A Model-Based Reinforcement Learning Approach For Adaptive Brain Stimulation",
                        "abstract": "Adaptive brain stimulation can treat neurological conditions such as Parkinson's disease and post-stroke motor deficits by influencing abnormal neural activity. Because of patient heterogeneity, each patient requires a unique stimulation policy to achieve optimal neural responses. Model-free reinforcement learning (MFRL) holds promise in learning effective policies for a variety of similar control tasks, but is limited in domains like brain stimulation by a need for numerous costly environment interactions. In this work we introduce Coprocessor Actor Critic, a novel, model-based reinforcement learning (MBRL) approach for learning neural coprocessor policies for brain stimulation. Our key insight is that coprocessor policy learning is a combination of learning how to act optimally in the world and learning how to induce optimal actions in the world through stimulation of an injured brain. We show that our approach overcomes the limitations of traditional MFRL methods in terms of sample efficiency and task success and outperforms baseline MBRL approaches in a neurologically realistic model of an injured brain."
                    },
                    {
                        "title": "Learning Temporal Distances: Contrastive Successor Features Can Provide a Metric Structure for Decision-Making",
                        "abstract": "Temporal distances lie at the heart of many algorithms for planning, control, and reinforcement learning that involve reaching goals, allowing one to estimate the transit time between two states. However, prior attempts to define such temporal distances in stochastic settings have been stymied by an important limitation: these prior approaches do not satisfy the triangle inequality. This is not merely a definitional concern, but translates to an inability to generalize and find shortest paths. In this paper, we build on prior work in contrastive learning and quasimetrics to show how successor features learned by contrastive learning (after a change of variables) form a temporal distance that does satisfy the triangle inequality, even in stochastic settings. Importantly, this temporal distance is computationally efficient to estimate, even in high-dimensional and stochastic settings. Experiments in controlled settings and benchmark suites demonstrate that an RL algorithm based on these new temporal distances exhibits combinatorial generalization (i.e., \"stitching\") and can sometimes learn more quickly than prior methods, including those based on quasimetrics."
                    },
                    {
                        "title": "Policy Adaptation via Language Optimization: Decomposing Tasks for Few-Shot Imitation",
                        "abstract": "Learned language-conditioned robot policies often struggle to effectively adapt to new real-world tasks even when pre-trained across a diverse set of instructions. We propose a novel approach for few-shot adaptation to unseen tasks that exploits the semantic understanding of task decomposition provided by vision-language models (VLMs). Our method, Policy Adaptation via Language Optimization (PALO), combines a handful of demonstrations of a task with proposed language decompositions sampled from a VLM to quickly enable rapid nonparametric adaptation, avoiding the need for a larger fine-tuning dataset. We evaluate PALO on extensive real-world experiments consisting of challenging unseen, long-horizon robot manipulation tasks. We find that PALO is able of consistently complete long-horizon, multi-tier tasks in the real world, outperforming state of the art pre-trained generalist policies, and methods that have access to the same demonstrations."
                    }
                ]
            },
            "07a39eb4-79ff-4909-842c-8ae1405bad5b": {
                "pk": "07a39eb4-79ff-4909-842c-8ae1405bad5b",
                "name": "Ruslan Salakhutdinov",
                "collaborators": [
                    "Nathan Srebro",
                    "Emilio Parisotto",
                    "Behnam Neyshabur",
                    "Nitish Srivastava",
                    "Ryan Kiros",
                    "Cody Severinski",
                    "John Langford",
                    "Tong Zhang",
                    "Dean Foster",
                    "Sham Kakade"
                ],
                "domain": [
                    "Matrix Completion",
                    "Reinforcement Learning",
                    "Deep Learning",
                    "Attention Mechanisms"
                ],
                "publications": [
                    {
                        "title": "Collaborative Filtering in a Non-Uniform World: Learning with the Weighted Trace Norm",
                        "abstract": "We show that matrix completion with trace-norm regularization can be significantly hurt when entries of the matrix are sampled non-uniformly. We introduce a weighted version of the trace-norm regularizer that works well also with non-uniform sampling. Our experimental results demonstrate that the weighted trace-norm regularization indeed yields significant gains on the (highly non-uniformly sampled) Netflix dataset."
                    },
                    {
                        "title": "Bayesian Probabilistic Matrix Factorization: A User Frequency Analysis",
                        "abstract": "Matrix factorization (MF) has become a common approach to collaborative filtering, due to ease of implementation and scalability to large data sets. Two existing drawbacks of the basic model is that it does not incorporate side information on either users or items, and assumes a common variance for all users. We extend the work of constrained probabilistic matrix factorization by deriving the Gibbs updates for the side feature vectors for items (Salakhutdinov and Minh, 2008). We show that this Bayesian treatment to the constrained PMF model outperforms simple MAP estimation. We also consider extensions to heteroskedastic precision introduced in the literature (Lakshminarayanan, Bouchard, and Archambeau, 2011). We show that this tends result in overfitting for deterministic approximation algorithms (ex: Variational inference) when the observed entries in the user / item matrix are distributed in an non-uniform manner. In light of this, we propose a truncated precision model. Our experimental results suggest that this model tends to delay overfitting."
                    },
                    {
                        "title": "Neural Map: Structured Memory for Deep Reinforcement Learning",
                        "abstract": "A critical component to enabling intelligent reasoning in partially observable environments is memory. Despite this importance, Deep Reinforcement Learning (DRL) agents have so far used relatively simple memory architectures, with the main methods to overcome partial observability being either a temporal convolution over the past k frames or an LSTM layer. More recent work (Oh et al., 2016) has went beyond these architectures by using memory networks which can allow more sophisticated addressing schemes over the past k frames. But even these architectures are unsatisfactory due to the reason that they are limited to only remembering information from the last k frames. In this paper, we develop a memory system with an adaptable write operator that is customized to the sorts of 3D environments that DRL agents typically interact with. This architecture, called the Neural Map, uses a spatially structured 2D memory image to learn to store arbitrary information about the environment over long time lags. We demonstrate empirically that the Neural Map surpasses previous DRL memories on a set of challenging 2D and 3D maze environments and show that it is capable of generalizing to environments that were not seen during training."
                    },
                    {
                        "title": "Efficient Transformers in Reinforcement Learning using Actor-Learner Distillation",
                        "abstract": "Many real-world applications such as robotics provide hard constraints on power and compute that limit the viable model complexity of Reinforcement Learning (RL) agents. Similarly, in many distributed RL settings, acting is done on un-accelerated hardware such as CPUs, which likewise restricts model size to prevent intractable experiment run times. These \"actor-latency\" constrained settings present a major obstruction to the scaling up of model complexity that has recently been extremely successful in supervised learning. To be able to utilize large model capacity while still operating within the limits imposed by the system during acting, we develop an \"Actor-Learner Distillation\" (ALD) procedure that leverages a continual form of distillation that transfers learning progress from a large capacity learner model to a small capacity actor model. As a case study, we develop this procedure in the context of partially-observable environments, where transformer models have had large improvements over LSTMs recently, at the cost of significantly higher computational complexity. With transformer models as the learner and LSTMs as the actor, we demonstrate in several challenging memory environments that using Actor-Learner Distillation recovers the clear sample-efficiency gains of the transformer learner model while maintaining the fast inference and reduced total training time of the LSTM actor model."
                    },
                    {
                        "title": "Learning Nonlinear Dynamic Models",
                        "abstract": "We present a novel approach for learning nonlinear dynamic models, which leads to a new set of tools capable of solving problems that are otherwise difficult. We provide theory showing this new approach is consistent for models with long range structure, and apply the approach to motion capture and high-dimensional video data, yielding results superior to standard alternatives."
                    },
                    {
                        "title": "Domain Adaptation: Overfitting and Small Sample Statistics",
                        "abstract": "We study the prevalent problem when a test distribution differs from the training distribution. We consider a setting where our training set consists of a small number of sample domains, but where we have many samples in each domain. Our goal is to generalize to a new domain. For example, we may want to learn a similarity function using only certain classes of objects, but we desire that this similarity function be applicable to object classes not present in our training sample (e.g. we might seek to learn that \"dogs are similar to dogs\" even though images of dogs were absent from our training set). Our theoretical analysis shows that we can select many more features than domains while avoiding overfitting by utilizing data-dependent variance properties. We present a greedy feature selection algorithm based on using T-statistics. Our experiments validate this theory showing that our T-statistic based greedy feature selection is more robust at avoiding overfitting than the classical greedy procedure."
                    },
                    {
                        "title": "Matrix reconstruction with the local max norm",
                        "abstract": "We introduce a new family of matrix norms, the \"local max\" norms, generalizing existing methods such as the max norm, the trace norm (nuclear norm), and the weighted or smoothed weighted trace norms, which have been extensively used in the literature as regularizers for matrix reconstruction problems. We show that this new family can be used to interpolate between the (weighted or unweighted) trace norm and the more conservative max norm. We test this interpolation on simulated data and on the large-scale Netflix and MovieLens ratings data, and find improved accuracy relative to the existing matrix norms. We also provide theoretical results showing learning guarantees for some of the new norms."
                    },
                    {
                        "title": "Initialization Strategies of Spatio-Temporal Convolutional Neural Networks",
                        "abstract": "We propose a new way of incorporating temporal information present in videos into Spatial Convolutional Neural Networks (ConvNets) trained on images, that avoids training Spatio-Temporal ConvNets from scratch. We describe several initializations of weights in 3D Convolutional Layers of Spatio-Temporal ConvNet using 2D Convolutional Weights learned from ImageNet. We show that it is important to initialize 3D Convolutional Weights judiciously in order to learn temporal representations of videos. We evaluate our methods on the UCF-101 dataset and demonstrate improvement over Spatial ConvNets."
                    },
                    {
                        "title": "Path-SGD: Path-Normalized Optimization in Deep Neural Networks",
                        "abstract": "We revisit the choice of SGD for training deep neural networks by reconsidering the appropriate geometry in which to optimize the weights. We argue for a geometry invariant to rescaling of weights that does not affect the output of the network, and suggest Path-SGD, which is an approximate steepest descent method with respect to a path-wise regularizer related to max-norm regularization. Path-SGD is easy and efficient to implement and leads to empirical gains over SGD and AdaGrad."
                    },
                    {
                        "title": "Data-Dependent Path Normalization in Neural Networks",
                        "abstract": "We propose a unified framework for neural net normalization, regularization and optimization, which includes Path-SGD and Batch-Normalization and interpolates between them across two different dimensions. Through this framework we investigate issue of invariance of the optimization, data dependence and the connection with natural gradients."
                    },
                    {
                        "title": "Path-Normalized Optimization of Recurrent Neural Networks with ReLU Activations",
                        "abstract": "We investigate the parameter-space geometry of recurrent neural networks (RNNs), and develop an adaptation of path-SGD optimization method, attuned to this geometry, that can learn plain RNNs with ReLU activations. On several datasets that require capturing long-term dependency structure, we show that path-SGD can significantly improve trainability of ReLU RNNs compared to RNNs trained with SGD, even with various recently suggested initialization schemes."
                    },
                    {
                        "title": "Topological Sort for Sentence Ordering",
                        "abstract": "Sentence ordering is the task of arranging the sentences of a given text in the correct order. Recent work using deep neural networks for this task has framed it as a sequence prediction problem. In this paper, we propose a new framing of this task as a constraint solving problem and introduce a new technique to solve it. Additionally, we propose a human evaluation for this task. The results on both automatic and human metrics across four different datasets show that this new technique is better at capturing coherence in documents."
                    },
                    {
                        "title": "Learning Wake-Sleep Recurrent Attention Models",
                        "abstract": "Despite their success, convolutional neural networks are computationally expensive because they must examine all image locations. Stochastic attention-based models have been shown to improve computational efficiency at test time, but they remain difficult to train because of intractable posterior inference and high variance in the stochastic gradient estimates. Borrowing techniques from the literature on training deep generative models, we present the Wake-Sleep Recurrent Attention Model, a method for training stochastic attention networks which improves posterior inference and which reduces the variability in the stochastic gradients. We show that our method can greatly speed up the training time for stochastic attention networks in the domains of image classification and caption generation."
                    },
                    {
                        "title": "Learning Generative Models with Visual Attention",
                        "abstract": "Attention has long been proposed by psychologists as important for effectively dealing with the enormous sensory stimulus available in the neocortex. Inspired by the visual attention models in computational neuroscience and the need of object-centric data for generative models, we describe for generative learning framework using attentional mechanisms. Attentional mechanisms can propagate signals from region of interest in a scene to an aligned canonical representation, where generative modeling takes place. By ignoring background clutter, generative models can concentrate their resources on the object of interest. Our model is a proper graphical model where the 2D Similarity transformation is a part of the top-down process. A ConvNet is employed to provide good initializations during posterior inference which is based on Hamiltonian Monte Carlo. Upon learning images of faces, our model can robustly attend to face regions of novel test subjects. More importantly, our model can learn generative models of new faces from a novel dataset of large images where the face locations are not known."
                    },
                    {
                        "title": "A Multiplicative Model for Learning Distributed Text-Based Attribute Representations",
                        "abstract": "In this paper we propose a general framework for learning distributed representations of attributes: characteristics of text whose representations can be jointly learned with word embeddings. Attributes can correspond to document indicators (to learn sentence vectors), language indicators (to learn distributed language representations), meta-data and side information (such as the age, gender and industry of a blogger) or representations of authors. We describe a third-order model where word context and attribute vectors interact multiplicatively to predict the next word in a sequence. This leads to the notion of conditional word similarity: how meanings of words change when conditioned on different attributes. We perform several experimental tasks including sentiment classification, cross-lingual document classification, and blog authorship attribution. We also qualitatively evaluate conditional word neighbours and attribute-conditioned text generation."
                    },
                    {
                        "title": "segDeepM: Exploiting Segmentation and Context in Deep Neural Networks for Object Detection",
                        "abstract": "In this paper, we propose an approach that exploits object segmentation in order to improve the accuracy of object detection. We frame the problem as inference in a Markov Random Field, in which each detection hypothesis scores object appearance as well as contextual information using Convolutional Neural Networks, and allows the hypothesis to choose and score a segment out of a large pool of accurate object segmentation proposals. This enables the detector to incorporate additional evidence when it is available and thus results in more accurate detections. Our experiments show an improvement of 4.1% in mAP over the R-CNN baseline on PASCAL VOC 2010, and 3.4% over the current state-of-the-art, demonstrating the power of our approach."
                    },
                    {
                        "title": "Action Recognition using Visual Attention",
                        "abstract": "We propose a soft attention based model for the task of action recognition in videos. We use multi-layered Recurrent Neural Networks (RNNs) with Long Short-Term Memory (LSTM) units which are deep both spatially and temporally. Our model learns to focus selectively on parts of the video frames and classifies videos after taking a few glimpses. The model essentially learns which parts in the frames are relevant for the task at hand and attaches higher importance to them. We evaluate the model on UCF-11 (YouTube Action), HMDB-51 and Hollywood2 datasets and analyze how the model focuses its attention depending on the scene and the action being performed."
                    },
                    {
                        "title": "Actor-Mimic: Deep Multitask and Transfer Reinforcement Learning",
                        "abstract": "The ability to act in multiple environments and transfer previous knowledge to new situations can be considered a critical aspect of any intelligent agent. Towards this goal, we define a novel method of multitask and transfer learning that enables an autonomous agent to learn how to behave in multiple tasks simultaneously, and then generalize its knowledge to new domains. This method, termed \"Actor-Mimic\", exploits the use of deep reinforcement learning and model compression techniques to train a single policy network that learns how to act in a set of distinct tasks by using the guidance of several expert teachers. We then show that the representations learnt by the deep policy network are capable of generalizing to new tasks with no prior expert guidance, speeding up learning in novel environments. Although our method can in general be applied to a wide range of problems, we use Atari games as a testing environment to demonstrate these methods."
                    },
                    {
                        "title": "Multi-Task Cross-Lingual Sequence Tagging from Scratch",
                        "abstract": "We present a deep hierarchical recurrent neural network for sequence tagging. Given a sequence of words, our model employs deep gated recurrent units on both character and word levels to encode morphology and context information, and applies a conditional random field layer to predict the tags. Our model is task independent, language independent, and feature engineering free. We further extend our model to multi-task and cross-lingual joint training by sharing the architecture and parameters. Our model achieves state-of-the-art results in multiple languages on several benchmark tasks including POS tagging, chunking, and NER. We also demonstrate that multi-task and cross-lingual joint training can improve the performance in various cases."
                    }
                ]
            },
            "e788e72a-7bcf-4569-b792-e5a7a97aac57": {
                "pk": "e788e72a-7bcf-4569-b792-e5a7a97aac57",
                "name": "Sergey Levine",
                "collaborators": [
                    "Chelsea Finn",
                    "Benjamin Eysenbach",
                    "Aurick Zhou",
                    "Dhruv Shah",
                    "William Montgomery",
                    "Justin Fu",
                    "Aviral Kumar",
                    "Seohong Park",
                    "Vladlen Koltun"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Deep Learning",
                    "Robotics",
                    "Uncertainty Estimation"
                ],
                "publications": [
                    {
                        "title": "Exploring Deep and Recurrent Architectures for Optimal Control",
                        "abstract": "Sophisticated multilayer neural networks have achieved state of the art results on multiple supervised tasks. However, successful applications of such multilayer networks to control have so far been limited largely to the perception portion of the control pipeline. In this paper, we explore the application of deep and recurrent neural networks to a continuous, high-dimensional locomotion task, where the network is used to represent a control policy that maps the state of the system (represented by joint angles) directly to the torques at each joint. By using a recent reinforcement learning algorithm called guided policy search, we can successfully train neural network controllers with thousands of parameters, allowing us to compare a variety of architectures. We discuss the differences between the locomotion control task and previous supervised perception tasks, present experimental results comparing various architectures, and discuss future directions in the application of techniques from deep learning to the problem of optimal control."
                    },
                    {
                        "title": "Reinforcement Learning and Control as Probabilistic Inference: Tutorial and Review",
                        "abstract": "The framework of reinforcement learning or optimal control provides a mathematical formalization of intelligent decision making that is powerful and broadly applicable. While the general form of the reinforcement learning problem enables effective reasoning about uncertainty, the connection between reinforcement learning and inference in probabilistic models is not immediately obvious. However, such a connection has considerable value when it comes to algorithm design: formalizing a problem as probabilistic inference in principle allows us to bring to bear a wide array of approximate inference tools, extend the model in flexible and powerful ways, and reason about compositionality and partial observability. In this article, we will discuss how a generalization of the reinforcement learning or optimal control problem, which is sometimes termed maximum entropy reinforcement learning, is equivalent to exact probabilistic inference in the case of deterministic dynamics, and variational inference in the case of stochastic dynamics. We will present a detailed derivation of this framework, overview prior work that has drawn on this and related ideas to propose new reinforcement learning and control algorithms, and describe perspectives on future research."
                    },
                    {
                        "title": "Understanding the World Through Action",
                        "abstract": "The recent history of machine learning research has taught us that machine learning methods can be most effective when they are provided with very large, high-capacity models, and trained on very large and diverse datasets. This has spurred the community to search for ways to remove any bottlenecks to scale. Often the foremost among such bottlenecks is the need for human effort, including the effort of curating and labeling datasets. As a result, considerable attention in recent years has been devoted to utilizing unlabeled data, which can be collected in vast quantities. However, some of the most widely used methods for training on such unlabeled data themselves require human-designed objective functions that must correlate in some meaningful way to downstream tasks. I will argue that a general, principled, and powerful framework for utilizing unlabeled data can be derived from reinforcement learning, using general purpose unsupervised or self-supervised reinforcement learning objectives in concert with offline reinforcement learning methods that can leverage large datasets. I will discuss how such a procedure is more closely aligned with potential downstream tasks, and how it could build on existing techniques that have been developed in recent years."
                    },
                    {
                        "title": "Learning Robotic Navigation from Experience: Principles, Methods, and Recent Results",
                        "abstract": "Navigation is one of the most heavily studied problems in robotics, and is conventionally approached as a geometric mapping and planning problem. However, real-world navigation presents a complex set of physical challenges that defies simple geometric abstractions. Machine learning offers a promising way to go beyond geometry and conventional planning, allowing for navigational systems that make decisions based on actual prior experience. Such systems can reason about traversability in ways that go beyond geometry, accounting for the physical outcomes of their actions and exploiting patterns in real-world environments. They can also improve as more data is collected, potentially providing a powerful network effect. In this article, we present a general toolkit for experiential learning of robotic navigation skills that unifies several recent approaches, describe the underlying design principles, summarize experimental results from several of our recent papers, and discuss open problems and directions for future work."
                    },
                    {
                        "title": "Guided Policy Search as Approximate Mirror Descent",
                        "abstract": "Guided policy search algorithms can be used to optimize complex nonlinear policies, such as deep neural networks, without directly computing policy gradients in the high-dimensional parameter space. Instead, these methods use supervised learning to train the policy to mimic a \"teacher\" algorithm, such as a trajectory optimizer or a trajectory-centric reinforcement learning method. Guided policy search methods provide asymptotic local convergence guarantees by construction, but it is not clear how much the policy improves within a small, finite number of iterations. We show that guided policy search algorithms can be interpreted as an approximate variant of mirror descent, where the projection onto the constraint manifold is not exact. We derive a new guided policy search algorithm that is simpler and provides appealing improvement and convergence guarantees in simplified convex and linear settings, and show that in the more general nonlinear setting, the error in the projection step can be bounded. We provide empirical results on several simulated robotic navigation and manipulation tasks that show that our method is stable and achieves similar or better performance when compared to prior guided policy search methods, with a simpler formulation and fewer hyperparameters."
                    },
                    {
                        "title": "Deep Visual Foresight for Planning Robot Motion",
                        "abstract": "A key challenge in scaling up robot learning to many skills and environments is removing the need for human supervision, so that robots can collect their own data and improve their own performance without being limited by the cost of requesting human feedback. Model-based reinforcement learning holds the promise of enabling an agent to learn to predict the effects of its actions, which could provide flexible predictive models for a wide range of tasks and environments, without detailed human supervision. We develop a method for combining deep action-conditioned video prediction models with model-predictive control that uses entirely unlabeled training data. Our approach does not require a calibrated camera, an instrumented training set-up, nor precise sensing and actuation. Our results show that our method enables a real robot to perform nonprehensile manipulation -- pushing objects -- and can handle novel objects not seen during training."
                    },
                    {
                        "title": "Meta-Learning and Universality: Deep Representations and Gradient Descent can Approximate any Learning Algorithm",
                        "abstract": "Learning to learn is a powerful paradigm for enabling models to learn from data more effectively and efficiently. A popular approach to meta-learning is to train a recurrent model to read in a training dataset as input and output the parameters of a learned model, or output predictions for new test inputs. Alternatively, a more recent approach to meta-learning aims to acquire deep representations that can be effectively fine-tuned, via standard gradient descent, to new tasks. In this paper, we consider the meta-learning problem from the perspective of universality, formalizing the notion of learning algorithm approximation and comparing the expressive power of the aforementioned recurrent models to the more recent approaches that embed gradient descent into the meta-learner. In particular, we seek to answer the following question: does deep representation combined with standard gradient descent have sufficient capacity to approximate any learning algorithm? We find that this is indeed true, and further find, in our experiments, that gradient-based meta-learning consistently leads to learning strategies that generalize more widely compared to those represented by recurrent models."
                    },
                    {
                        "title": "Offline Model-Based Optimization via Normalized Maximum Likelihood Estimation",
                        "abstract": "In this work we consider data-driven optimization problems where one must maximize a function given only queries at a fixed set of points. This problem setting emerges in many domains where function evaluation is a complex and expensive process, such as in the design of materials, vehicles, or neural network architectures. Because the available data typically only covers a small manifold of the possible space of inputs, a principal challenge is to be able to construct algorithms that can reason about uncertainty and out-of-distribution values, since a naive optimizer can easily exploit an estimated model to return adversarial inputs. We propose to tackle this problem by leveraging the normalized maximum-likelihood (NML) estimator, which provides a principled approach to handling uncertainty and out-of-distribution inputs. While in the standard formulation NML is intractable, we propose a tractable approximation that allows us to scale our method to high-capacity neural network models. We demonstrate that our method can effectively optimize high-dimensional design problems in a variety of disciplines such as chemistry, biology, and materials engineering."
                    },
                    {
                        "title": "If MaxEnt RL is the Answer, What is the Question?",
                        "abstract": "Experimentally, it has been observed that humans and animals often make decisions that do not maximize their expected utility, but rather choose outcomes randomly, with probability proportional to expected utility. Probability matching, as this strategy is called, is equivalent to maximum entropy reinforcement learning (MaxEnt RL). However, MaxEnt RL does not optimize expected utility. In this paper, we formally show that MaxEnt RL does optimally solve certain classes of control problems with variability in the reward function. In particular, we show (1) that MaxEnt RL can be used to solve a certain class of POMDPs, and (2) that MaxEnt RL is equivalent to a two-player game where an adversary chooses the reward function. These results suggest a deeper connection between MaxEnt RL, robust control, and POMDPs, and provide insight for the types of problems for which we might expect MaxEnt RL to produce effective solutions. Specifically, our results suggest that domains with uncertainty in the task goal may be especially well-suited for MaxEnt RL methods."
                    },
                    {
                        "title": "Training on Test Data with Bayesian Adaptation for Covariate Shift",
                        "abstract": "When faced with distribution shift at test time, deep neural networks often make inaccurate predictions with unreliable uncertainty estimates. While improving the robustness of neural networks is one promising approach to mitigate this issue, an appealing alternate to robustifying networks against all possible test-time shifts is to instead directly adapt them to unlabeled inputs from the particular distribution shift we encounter at test time. However, this poses a challenging question: in the standard Bayesian model for supervised learning, unlabeled inputs are conditionally independent of model parameters when the labels are unobserved, so what can unlabeled data tell us about the model parameters at test-time? In this paper, we derive a Bayesian model that provides for a well-defined relationship between unlabeled inputs under distributional shift and model parameters, and show how approximate inference in this model can be instantiated with a simple regularized entropy minimization procedure at test-time. We evaluate our method on a variety of distribution shifts for image classification, including image corruptions, natural distribution shifts, and domain adaptation settings, and show that our method improves both accuracy and uncertainty estimation."
                    },
                    {
                        "title": "Model Inversion Networks for Model-Based Optimization",
                        "abstract": "In this work, we aim to solve data-driven optimization problems, where the goal is to find an input that maximizes an unknown score function given access to a dataset of inputs with corresponding scores. When the inputs are high-dimensional and valid inputs constitute a small subset of this space (e.g., valid protein sequences or valid natural images), such model-based optimization problems become exceptionally difficult, since the optimizer must avoid out-of-distribution and invalid inputs. We propose to address such problem with model inversion networks (MINs), which learn an inverse mapping from scores to inputs. MINs can scale to high-dimensional input spaces and leverage offline logged data for both contextual and non-contextual optimization problems. MINs can also handle both purely offline data sources and active data collection. We evaluate MINs on tasks from the Bayesian optimization literature, high-dimensional model-based optimization problems over images and protein designs, and contextual bandit optimization from logged data."
                    },
                    {
                        "title": "Amortized Conditional Normalized Maximum Likelihood: Reliable Out of Distribution Uncertainty Estimation",
                        "abstract": "While deep neural networks provide good performance for a range of challenging tasks, calibration and uncertainty estimation remain major challenges, especially under distribution shift. In this paper, we propose the amortized conditional normalized maximum likelihood (ACNML) method as a scalable general-purpose approach for uncertainty estimation, calibration, and out-of-distribution robustness with deep networks. Our algorithm builds on the conditional normalized maximum likelihood (CNML) coding scheme, which has minimax optimal properties according to the minimum description length principle, but is computationally intractable to evaluate exactly for all but the simplest of model classes. We propose to use approximate Bayesian inference technqiues to produce a tractable approximation to the CNML distribution. Our approach can be combined with any approximate inference algorithm that provides tractable posterior densities over model parameters. We demonstrate that ACNML compares favorably to a number of prior techniques for uncertainty estimation in terms of calibration on out-of-distribution inputs."
                    },
                    {
                        "title": "Maximum Entropy RL (Provably) Solves Some Robust RL Problems",
                        "abstract": "Many potential applications of reinforcement learning (RL) require guarantees that the agent will perform well in the face of disturbances to the dynamics or reward function. In this paper, we prove theoretically that maximum entropy (MaxEnt) RL maximizes a lower bound on a robust RL objective, and thus can be used to learn policies that are robust to some disturbances in the dynamics and the reward function. While this capability of MaxEnt RL has been observed empirically in prior work, to the best of our knowledge our work provides the first rigorous proof and theoretical characterization of the MaxEnt RL robust set. While a number of prior robust RL algorithms have been designed to handle similar disturbances to the reward function or dynamics, these methods typically require additional moving parts and hyperparameters on top of a base RL algorithm. In contrast, our results suggest that MaxEnt RL by itself is robust to certain disturbances, without requiring any additional modifications. While this does not imply that MaxEnt RL is the best available robust RL method, MaxEnt RL is a simple robust RL method with appealing formal guarantees."
                    },
                    {
                        "title": "Predictable MDP Abstraction for Unsupervised Model-Based RL",
                        "abstract": "A key component of model-based reinforcement learning (RL) is a dynamics model that predicts the outcomes of actions. Errors in this predictive model can degrade the performance of model-based controllers, and complex Markov decision processes (MDPs) can present exceptionally difficult prediction problems. To mitigate this issue, we propose predictable MDP abstraction (PMA): instead of training a predictive model on the original MDP, we train a model on a transformed MDP with a learned action space that only permits predictable, easy-to-model actions, while covering the original state-action space as much as possible. As a result, model learning becomes easier and more accurate, which allows robust, stable model-based planning or model-based RL. This transformation is learned in an unsupervised manner, before any task is specified by the user. Downstream tasks can then be solved with model-based control in a zero-shot fashion, without additional environment interactions. We theoretically analyze PMA and empirically demonstrate that PMA leads to significant improvements over prior unsupervised model-based RL approaches in a range of benchmark environments. Our code and videos are available at https://seohong.me/projects/pma/"
                    },
                    {
                        "title": "Continuous Inverse Optimal Control with Locally Optimal Examples",
                        "abstract": "Inverse optimal control, also known as inverse reinforcement learning, is the problem of recovering an unknown reward function in a Markov decision process from expert demonstrations of the optimal policy. We introduce a probabilistic inverse optimal control algorithm that scales gracefully with task dimensionality, and is suitable for large, continuous domains where even computing a full policy is impractical. By using a local approximation of the reward function, our method can also drop the assumption that the demonstrations are globally optimal, requiring only local optimality. This allows it to learn from examples that are unsuitable for prior methods."
                    }
                ]
            }
        }
    },
    "2310.11428": {
        "paper_data": {
            "title": "Butterfly Effects of SGD Noise: Error Amplification in Behavior Cloning and Autoregression",
            "url": "http://arxiv.org/abs/2310.11428v1",
            "arxiv_id": "2310.11428",
            "authors": [
                "Adam Block",
                "Dylan J. Foster",
                "Akshay Krishnamurthy",
                "Max Simchowitz",
                "Cyril Zhang"
            ],
            "abstract": "This work studies training instabilities of behavior cloning with deep neural networks. We observe that minibatch SGD updates to the policy network during training result in sharp oscillations in long-horizon rewards, despite negligibly affecting the behavior cloning loss. We empirically disentangle the statistical and computational causes of these oscillations, and find them to stem from the chaotic propagation of minibatch SGD noise through unstable closed-loop dynamics. While SGD noise is benign in the single-step action prediction objective, it results in catastrophic error accumulation over long horizons, an effect we term gradient variance amplification (GVA). We show that many standard mitigation techniques do not alleviate GVA, but find an exponential moving average (EMA) of iterates to be surprisingly effective at doing so. We illustrate the generality of this phenomenon by showing the existence of GVA and its amelioration by EMA in both continuous control and autoregressive language generation. Finally, we provide theoretical vignettes that highlight the benefits of EMA in alleviating GVA and shed light on the extent to which classical convex models can help in understanding the benefits of iterate averaging in deep learning.",
            "introduction": " Introduction to online nonstochastic control. arXiv preprint arXiv:2211.09619 , 2022. Haowei He, Gao Huang, and Yang Yuan. Asymmetric valleys: Beyond sharp and flat local minima. Ad- vances in Neural Information Processing Systems , 32, 2019. Peter Henderson, Riashat Islam, Philip Bachman, Joelle Pineau, Doina Precup, and David Meger. Deep reinforcement learning that matters. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 32, 2018. Jorgen Hoffmann-Jorgensen, Lawrence A Shepp, and Richard M Dudley. On the lower tail of gaussian seminorms. The Annals of Probability , pp. 319\u2013342, 1979. Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, and Yejin Choi. The curious case of neural text degener- ation. arXiv preprint arXiv:1904.09751 , 2019. Pavel Izmailov, Dmitrii Podoprikhin, Timur Garipov, Dmitry Vetrov, and Andrew Gordon Wilson. Averag- ing weights leads to wider optima and better generalization. arXiv preprint arXiv:1803.05407 , 2018. 16Michael Janner, Qiyang Li, and Sergey Levine. Offline reinforcement learning as one big sequence modeling problem. Advances in Neural Information Processing Systems , 34:1273\u20131286, 2021. Nan Jiang, Akshay Krishnamurthy, Alekh Agarwal, John Langford, and Robert E Schapire. Contextual decision processes with low Bellman rank are PAC-learnable. In International Conference on Machine Learning , pp. 1704\u20131713, 2017. Chi Jin, Rong Ge, Praneeth Netrapalli, Sham M Kakade, and Michael I Jordan. How to escape saddle points efficiently. In International conference on machine learning , pp. 1724\u20131732. PMLR, 2017. Chi Jin, Qinghua Liu, and Sobhan Miryoosefi. Bellman eluder dimension: New rich classes of RL problems, and sample-efficient algorithms. Neural Information Processing Systems , 2021. Jean Kaddour. Stop wasting my time! Saving days of ImageNet and BERT training with latest weight averaging. arXiv preprint arXiv:2209.14981 , 2022. Jean Kaddour, Oscar Key, Piotr Nawrot, Pasquale Minervini, and Matt J Kusner. No train no gain: Revisiting efficient training algorithms for Transformer-based language models. arXiv preprint arXiv:2307.06440 , 2023. Sham Kakade, Akshay Krishnamurthy, Kendall Lowrey, Motoya Ohnishi, and Wen Sun. Information theo- retic regret bounds for online nonlinear control. Advances in Neural Information Processing Systems , 33: 15312\u201315325, 2020. Rudolph Emil Kalman. A new approach to linear filtering and prediction problems. 1960. Liyiming Ke, Jingqiang Wang, Tapomayukh Bhattacharjee, Byron Boots, and Siddhartha Srinivasa. Grasp- ing with chopsticks: Combating covariate shift in model-free imitation learning for fine manipulation. In 2021 IEEE International Conference on Robotics and Automation (ICRA) , pp. 6185\u20136191. IEEE, 2021. Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 , 2014. James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. Overcoming catas- trophic forgetting in neural networks. Proceedings of the National Academy of Sciences , 114(13):3521\u2013 3526, 2017. Karl Krauth, Sarah Dean, Alex Zhao, Wenshuo Guo, Mihaela Curmei, Benjamin Recht, and Michael I Jordan. Do offline metrics predict online performance in recommender systems? arXiv preprint arXiv:2011.07931 , 2020. Aviral Kumar, Justin Fu, Matthew Soh, George Tucker, and Sergey Levine. Stabilizing off-policy q-learning via bootstrapping error reduction. Advances in Neural Information Processing Systems , 32, 2019. Daniel Kunin, Javier Sagastuy-Brena, Surya Ganguli, Daniel LK Yamins, and Hidenori Tanaka. Neu- ral mechanics: Symmetry and broken conservation laws in deep learning dynamics. arXiv preprint arXiv:2012.04728 , 2020. Simon Lacoste-Julien, Mark Schmidt, and Francis Bach. A simpler approach to obtaining an o (1/t) conver- gence rate for the projected stochastic subgradient method. HAL , 2012, 2012. 17Michael Laskey, Jonathan",
            "references": [
                {
                    "title": "How to Scale Your EMA",
                    "abstract": "Preserving training dynamics across batch sizes is an important tool for practical machine learning as it enables the trade-off between batch size and wall-clock time. This trade-off is typically enabled by a scaling rule, for example, in stochastic gradient descent, one should scale the learning rate linearly with the batch size. Another important tool for practical machine learning is the model Exponential Moving Average (EMA), which is a model copy that does not receive gradient information, but instead follows its target model with some momentum. This model EMA can improve the robustness and generalization properties of supervised learning, stabilize pseudo-labeling, and provide a learning signal for Self-Supervised Learning (SSL). Prior works have treated the model EMA separately from optimization, leading to different training dynamics across batch sizes and lower model performance. In this work, we provide a scaling rule for optimization in the presence of model EMAs and demonstrate its validity across a range of architectures, optimizers, and data modalities. We also show the rule's validity where the model EMA contributes to the optimization of the target model, enabling us to train EMA-based pseudo-labeling and SSL methods at small and large batch sizes. For SSL, we enable training of BYOL up to batch size 24,576 without sacrificing performance, optimally a 6$\\times$ wall-clock time reduction."
                },
                {
                    "title": "No Train No Gain: Revisiting Efficient Training Algorithms For Transformer-based Language Models",
                    "abstract": "The computation necessary for training Transformer-based language models has skyrocketed in recent years. This trend has motivated research on efficient training algorithms designed to improve training, validation, and downstream performance faster than standard training. In this work, we revisit three categories of such algorithms: dynamic architectures (layer stacking, layer dropping), batch selection (selective backprop, RHO loss), and efficient optimizers (Lion, Sophia). When pre-training BERT and T5 with a fixed computation budget using such methods, we find that their training, validation, and downstream gains vanish compared to a baseline with a fully-decayed learning rate. We define an evaluation protocol that enables computation to be done on arbitrary machines by mapping all computation time to a reference machine which we call reference system time. We discuss the limitations of our proposed protocol and release our code to encourage rigorous research in efficient training procedures: https://github.com/JeanKaddour/NoTrainNoGain."
                },
                {
                    "title": "Learning to Generate Better Than Your LLM",
                    "abstract": "Reinforcement learning (RL) has emerged as a powerful paradigm for fine-tuning Large Language Models (LLMs) for text generation. In particular, recent LLMs such as ChatGPT and GPT-4 can engage in fluent conversations with users after finetuning with RL. Capitalizing on key properties of text generation, we seek to investigate RL algorithms beyond general purpose algorithms like Proximal Policy Optimization (PPO). In particular, we extend RL algorithms to allow them to interact with a dynamic black-box guide LLM and propose RL with guided feedback (RLGF), a suite of RL algorithms for LLM fine-tuning. We provide two ways for the guide LLM to interact with the LLM to be optimized for maximizing rewards. The guide LLM can generate text which serves as additional starting states for the RL optimization procedure. The guide LLM can also be used to complete the partial sentences generated by the LLM that is being optimized, treating the guide LLM as an expert to imitate and surpass eventually. We experiment on the IMDB positive sentiment, CommonGen, and TL;DR summarization tasks. We show that our RL algorithms achieve higher performance than supervised learning (SL) and the RL baseline PPO, demonstrating the benefit of interaction with the guide LLM. On both CommonGen and TL;DR, we not only outperform our SL baselines but also improve upon PPO across a variety of metrics beyond the one we optimized for. Our code can be found at https://github.com/Cornell-RL/tril."
                },
                {
                    "title": "Early Weight Averaging meets High Learning Rates for LLM Pre-training",
                    "abstract": "Training Large Language Models (LLMs) incurs significant cost; hence, any strategy that accelerates model convergence is helpful. In this paper, we investigate the ability of a simple idea checkpoint averaging along the trajectory of a training run to improve both convergence and generalization quite early on during training. Here we show that models trained with high learning rates observe higher gains due to checkpoint averaging. Furthermore, these gains are amplified when checkpoints are sampled with considerable spacing in training steps. Our training recipe outperforms conventional training and popular checkpoint averaging baselines such as exponential moving average (EMA) and stochastic moving average (SWA). We evaluate our training recipe by pre-training LLMs, where high learning rates are inherently preferred due to extremely large batch sizes. Specifically, we pre-trained nanoGPT-2 models of varying sizes, small (125M), medium (335M), and large (770M)on the OpenWebText dataset, comprised of 9B tokens. Additionally, we present results for publicly available Pythia LLMs, ranging from 1B to 12B, which were trained on the PILE-deduped dataset containing 207B tokens."
                },
                {
                    "title": "Exposing Attention Glitches with Flip-Flop Language Modeling",
                    "abstract": "Why do large language models sometimes output factual inaccuracies and exhibit erroneous reasoning? The brittleness of these models, particularly when executing long chains of reasoning, currently seems to be an inevitable price to pay for their advanced capabilities of coherently synthesizing knowledge, pragmatics, and abstract thought. Towards making sense of this fundamentally unsolved problem, this work identifies and analyzes the phenomenon of attention glitches, in which the Transformer architecture's inductive biases intermittently fail to capture robust reasoning. To isolate the issue, we introduce flip-flop language modeling (FFLM), a parametric family of synthetic benchmarks designed to probe the extrapolative behavior of neural language models. This simple generative task requires a model to copy binary symbols over long-range dependencies, ignoring the tokens in between. We find that Transformer FFLMs suffer from a long tail of sporadic reasoning errors, some of which we can eliminate using various regularization techniques. Our preliminary mechanistic analyses show why the remaining errors may be very difficult to diagnose and resolve. We hypothesize that attention glitches account for (some of) the closed-domain hallucinations in natural LLMs."
                },
                {
                    "title": "The Power of Learned Locally Linear Models for Nonlinear Policy Optimization",
                    "abstract": "A common pipeline in learning-based control is to iteratively estimate a model of system dynamics, and apply a trajectory optimization algorithm - e.g.~$\\mathtt{iLQR}$ - on the learned model to minimize a target cost. This paper conducts a rigorous analysis of a simplified variant of this strategy for general nonlinear systems. We analyze an algorithm which iterates between estimating local linear models of nonlinear system dynamics and performing $\\mathtt{iLQR}$-like policy updates. We demonstrate that this algorithm attains sample complexity polynomial in relevant problem parameters, and, by synthesizing locally stabilizing gains, overcomes exponential dependence in problem horizon. Experimental results validate the performance of our algorithm, and compare to natural deep-learning baselines."
                },
                {
                    "title": "TinyStories: How Small Can Language Models Be and Still Speak Coherent English?",
                    "abstract": "Language models (LMs) are powerful tools for natural language processing, but they often struggle to produce coherent and fluent text when they are small. Models with around 125M parameters such as GPT-Neo (small) or GPT-2 (small) can rarely generate coherent and consistent English text beyond a few words even after extensive training. This raises the question of whether the emergence of the ability to produce coherent English text only occurs at larger scales (with hundreds of millions of parameters or more) and complex architectures (with many layers of global attention). In this work, we introduce TinyStories, a synthetic dataset of short stories that only contain words that a typical 3 to 4-year-olds usually understand, generated by GPT-3.5 and GPT-4. We show that TinyStories can be used to train and evaluate LMs that are much smaller than the state-of-the-art models (below 10 million total parameters), or have much simpler architectures (with only one transformer block), yet still produce fluent and consistent stories with several paragraphs that are diverse and have almost perfect grammar, and demonstrate reasoning capabilities. We also introduce a new paradigm for the evaluation of language models: We suggest a framework which uses GPT-4 to grade the content generated by these models as if those were stories written by students and graded by a (human) teacher. This new paradigm overcomes the flaws of standard benchmarks which often requires the model's output to be very structures, and moreover provides a multidimensional score for the model, providing scores for different capabilities such as grammar, creativity and consistency. We hope that TinyStories can facilitate the development, analysis and research of LMs, especially for low-resource or specialized domains, and shed light on the emergence of language capabilities in LMs."
                },
                {
                    "title": "Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling",
                    "abstract": "How do large language models (LLMs) develop and evolve over the course of training? How do these patterns change as models scale? To answer these questions, we introduce \\textit{Pythia}, a suite of 16 LLMs all trained on public data seen in the exact same order and ranging in size from 70M to 12B parameters. We provide public access to 154 checkpoints for each one of the 16 models, alongside tools to download and reconstruct their exact training dataloaders for further study. We intend \\textit{Pythia} to facilitate research in many areas, and we present several case studies including novel results in memorization, term frequency effects on few-shot performance, and reducing gender bias. We demonstrate that this highly controlled setup can be used to yield novel insights toward LLMs and their training dynamics. Trained models, analysis code, training code, and training data can be found at \\url{https://github.com/EleutherAI/pythia}."
                },
                {
                    "title": "Diffusion Policy: Visuomotor Policy Learning via Action Diffusion",
                    "abstract": "This paper introduces Diffusion Policy, a new way of generating robot behavior by representing a robot\u2019s visuomotor policy as a conditional denoising diffusion process. We benchmark Diffusion Policy across 15 different tasks from 4 different robot manipulation benchmarks and find that it consistently outperforms existing state-of-the-art robot learning methods with an average improvement of 46.9%. Diffusion Policy learns the gradient of the action-distribution score function and iteratively optimizes with respect to this gradient field during inference via a series of stochastic Langevin dynamics steps. We find that the diffusion formulation yields powerful advantages when used for robot policies, including gracefully handling multimodal action distributions, being suitable for high-dimensional action spaces, and exhibiting impressive training stability. To fully unlock the potential of diffusion models for visuomotor policy learning on physical robots, this paper presents a set of key technical contributions including the incorporation of receding horizon control, visual conditioning, and the time-series diffusion transformer. We hope this work will help motivate a new generation of policy learning techniques that are able to leverage the powerful generative modeling capabilities of diffusion models. Code, data, and training details are available (diffusion-policy.cs.columbia.edu)."
                },
                {
                    "title": "Oracle-Efficient Smoothed Online Learning for Piecewise Continuous Decision Making",
                    "abstract": "Smoothed online learning has emerged as a popular framework to mitigate the substantial loss in statistical and computational complexity that arises when one moves from classical to adversarial learning. Unfortunately, for some spaces, it has been shown that efficient algorithms suffer an exponentially worse regret than that which is minimax optimal, even when the learner has access to an optimization oracle over the space. To mitigate that exponential dependence, this work introduces a new notion of complexity, the generalized bracketing numbers, which marries constraints on the adversary to the size of the space, and shows that an instantiation of Follow-the-Perturbed-Leader can attain low regret with the number of calls to the optimization oracle scaling optimally with respect to average regret. We then instantiate our bounds in several problems of interest, including online prediction and planning of piecewise continuous functions, which has many applications in fields as diverse as econometrics and robotics."
                },
                {
                    "title": "Smoothed Online Learning for Prediction in Piecewise Affine Systems",
                    "abstract": "The problem of piecewise affine (PWA) regression and planning is of foundational importance to the study of online learning, control, and robotics, where it provides a theoretically and empirically tractable setting to study systems undergoing sharp changes in the dynamics. Unfortunately, due to the discontinuities that arise when crossing into different ``pieces,'' learning in general sequential settings is impossible and practical algorithms are forced to resort to heuristic approaches. This paper builds on the recently developed smoothed online learning framework and provides the first algorithms for prediction and simulation in PWA systems whose regret is polynomial in all relevant problem parameters under a weak smoothness assumption; moreover, our algorithms are efficient in the number of calls to an optimization oracle. We further apply our results to the problems of one-step prediction and multi-step simulation regret in piecewise affine dynamical systems, where the learner is tasked with simulating trajectories and regret is measured in terms of the Wasserstein distance between simulated and true data. Along the way, we develop several technical tools of more general interest."
                },
                {
                    "title": "Training trajectories, mini-batch losses and the curious role of the learning rate",
                    "abstract": "Stochastic gradient descent plays a fundamental role in nearly all applications of deep learning. However its ability to converge to a global minimum remains shrouded in mystery. In this paper we propose to study the behavior of the loss function on fixed mini-batches along SGD trajectories. We show that the loss function on a fixed batch appears to be remarkably convex-like. In particular for ResNet the loss for any fixed mini-batch can be accurately modeled by a quadratic function and a very low loss value can be reached in just one step of gradient descent with sufficiently large learning rate. We propose a simple model that allows to analyze the relationship between the gradients of stochastic mini-batches and the full batch. Our analysis allows us to discover the equivalency between iterate aggregates and specific learning rate schedules. In particular, for Exponential Moving Average (EMA) and Stochastic Weight Averaging we show that our proposed model matches the observed training trajectories on ImageNet. Our theoretical model predicts that an even simpler averaging technique, averaging just two points a many steps apart, significantly improves accuracy compared to the baseline. We validated our findings on ImageNet and other datasets using ResNet architecture."
                },
                {
                    "title": "Introduction to Online Nonstochastic Control",
                    "abstract": "This text presents an introduction to an emerging paradigm in control of dynamical systems and differentiable reinforcement learning called online nonstochastic control. The new approach applies techniques from online convex optimization and convex relaxations to obtain new methods with provable guarantees for classical settings in optimal and robust control. The primary distinction between online nonstochastic control and other frameworks is the objective. In optimal control, robust control, and other control methodologies that assume stochastic noise, the goal is to perform comparably to an offline optimal strategy. In online nonstochastic control, both the cost functions as well as the perturbations from the assumed dynamical model are chosen by an adversary. Thus the optimal policy is not defined a priori. Rather, the target is to attain low regret against the best policy in hindsight from a benchmark class of policies. This objective suggests the use of the decision making framework of online convex optimization as an algorithmic methodology. The resulting methods are based on iterative mathematical optimization algorithms, and are accompanied by finite-time regret and computational complexity guarantees."
                },
                {
                    "title": "SGD with large step sizes learns sparse features",
                    "abstract": "We showcase important features of the dynamics of the Stochastic Gradient Descent (SGD) in the training of neural networks. We present empirical observations that commonly used large step sizes (i) lead the iterates to jump from one side of a valley to the other causing loss stabilization, and (ii) this stabilization induces a hidden stochastic dynamics orthogonal to the bouncing directions that biases it implicitly toward sparse predictors. Furthermore, we show empirically that the longer large step sizes keep SGD high in the loss landscape valleys, the better the implicit regularization can operate and find sparse representations. Notably, no explicit regularization is used so that the regularization effect comes solely from the SGD training dynamics influenced by the step size schedule. Therefore, these observations unveil how, through the step size schedules, both gradient and noise drive together the SGD dynamics through the loss landscape of neural networks. We justify these findings theoretically through the study of simple neural network models as well as qualitative arguments inspired from stochastic processes. Finally, this analysis allows us to shed a new light on some common practice and observed phenomena when training neural networks. The code of our experiments is available at https://github.com/tml-epfl/sgd-sparse-features."
                },
                {
                    "title": "Stop Wasting My Time! Saving Days of ImageNet and BERT Training with Latest Weight Averaging",
                    "abstract": "Training vision or language models on large datasets can take days, if not weeks. We show that averaging the weights of the k latest checkpoints, each collected at the end of an epoch, can speed up the training progression in terms of loss and accuracy by dozens of epochs, corresponding to time savings up to ~68 and ~30 GPU hours when training a ResNet50 on ImageNet and RoBERTa-Base model on WikiText-103, respectively. We also provide the code and model checkpoint trajectory to reproduce the results and facilitate research on reusing historical weights for faster convergence."
                },
                {
                    "title": "Feature selection with gradient descent on two-layer networks in low-rotation regimes",
                    "abstract": "This work establishes low test error of gradient flow (GF) and stochastic gradient descent (SGD) on two-layer ReLU networks with standard initialization, in three regimes where key sets of weights rotate little (either naturally due to GF and SGD, or due to an artificial constraint), and making use of margins as the core analytic technique. The first regime is near initialization, specifically until the weights have moved by $\\mathcal{O}(\\sqrt m)$, where $m$ denotes the network width, which is in sharp contrast to the $\\mathcal{O}(1)$ weight motion allowed by the Neural Tangent Kernel (NTK); here it is shown that GF and SGD only need a network width and number of samples inversely proportional to the NTK margin, and moreover that GF attains at least the NTK margin itself, which suffices to establish escape from bad KKT points of the margin objective, whereas prior work could only establish nondecreasing but arbitrarily small margins. The second regime is the Neural Collapse (NC) setting, where data lies in extremely-well-separated groups, and the sample complexity scales with the number of groups; here the contribution over prior work is an analysis of the entire GF trajectory from initialization. Lastly, if the inner layer weights are constrained to change in norm only and can not rotate, then GF with large widths achieves globally maximal margins, and its sample complexity scales with their inverse; this is in contrast to prior work, which required infinite width and a tricky dual convergence assumption. As purely technical contributions, this work develops a variety of potential functions and other tools which will hopefully aid future work."
                },
                {
                    "title": "Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limit",
                    "abstract": "There is mounting evidence of emergent phenomena in the capabilities of deep learning methods as we scale up datasets, model sizes, and training times. While there are some accounts of how these resources modulate statistical capacity, far less is known about their effect on the computational problem of model training. This work conducts such an exploration through the lens of learning a $k$-sparse parity of $n$ bits, a canonical discrete search problem which is statistically easy but computationally hard. Empirically, we find that a variety of neural networks successfully learn sparse parities, with discontinuous phase transitions in the training curves. On small instances, learning abruptly occurs at approximately $n^{O(k)}$ iterations; this nearly matches SQ lower bounds, despite the apparent lack of a sparse prior. Our theoretical analysis shows that these observations are not explained by a Langevin-like mechanism, whereby SGD\"stumbles in the dark\"until it finds the hidden set of features (a natural algorithm which also runs in $n^{O(k)}$ time). Instead, we show that SGD gradually amplifies the sparse solution via a Fourier gap in the population gradient, making continual progress that is invisible to loss and error metrics."
                },
                {
                    "title": "Neural Networks can Learn Representations with Gradient Descent",
                    "abstract": "Significant theoretical work has established that in specific regimes, neural networks trained by gradient descent behave like kernel methods. However, in practice, it is known that neural networks strongly outperform their associated kernels. In this work, we explain this gap by demonstrating that there is a large class of functions which cannot be efficiently learned by kernel methods but can be easily learned with gradient descent on a two layer neural network outside the kernel regime by learning representations that are relevant to the target task. We also demonstrate that these representations allow for efficient transfer learning, which is impossible in the kernel regime. Specifically, we consider the problem of learning polynomials which depend on only a few relevant directions, i.e. of the form $f^\\star(x) = g(Ux)$ where $U: \\R^d \\to \\R^r$ with $d \\gg r$. When the degree of $f^\\star$ is $p$, it is known that $n \\asymp d^p$ samples are necessary to learn $f^\\star$ in the kernel regime. Our primary result is that gradient descent learns a representation of the data which depends only on the directions relevant to $f^\\star$. This results in an improved sample complexity of $n\\asymp d^2 r + dr^p$. Furthermore, in a transfer learning setup where the data distributions in the source and target domain share the same representation $U$ but have different polynomial heads we show that a popular heuristic for transfer learning has a target sample complexity independent of $d$."
                },
                {
                    "title": "Behavior Transformers: Cloning k modes with one stone",
                    "abstract": "While behavior learning has made impressive progress in recent times, it lags behind computer vision and natural language processing due to its inability to leverage large, human-generated datasets. Human behaviors have wide variance, multiple modes, and human demonstrations typically do not come with reward labels. These properties limit the applicability of current methods in Offline RL and Behavioral Cloning to learn from large, pre-collected datasets. In this work, we present Behavior Transformer (BeT), a new technique to model unlabeled demonstration data with multiple modes. BeT retrofits standard transformer architectures with action discretization coupled with a multi-task action correction inspired by offset prediction in object detection. This allows us to leverage the multi-modal modeling ability of modern transformers to predict multi-modal continuous actions. We experimentally evaluate BeT on a variety of robotic manipulation and self-driving behavior datasets. We show that BeT significantly improves over prior state-of-the-art work on solving demonstrated tasks while capturing the major modes present in the pre-collected datasets. Finally, through an extensive ablation study, we analyze the importance of every crucial component in BeT. Videos of behavior generated by BeT are available at https://notmahi.github.io/bet"
                },
                {
                    "title": "TaSIL: Taylor Series Imitation Learning",
                    "abstract": "We propose Taylor Series Imitation Learning (TaSIL), a simple augmentation to standard behavior cloning losses in the context of continuous control. TaSIL penalizes deviations in the higher-order Taylor series terms between the learned and expert policies. We show that experts satisfying a notion of $\\textit{incremental input-to-state stability}$ are easy to learn, in the sense that a small TaSIL-augmented imitation loss over expert trajectories guarantees a small imitation loss over trajectories generated by the learned policy. We provide sample-complexity bounds for TaSIL that scale as $\\tilde{\\mathcal{O}}(1/n)$ in the realizable setting, for $n$ the number of expert demonstrations. Finally, we demonstrate experimentally the relationship between the robustness of the expert policy and the order of Taylor expansion required in TaSIL, and compare standard Behavior Cloning, DART, and DAgger with TaSIL-loss-augmented variants. In all cases, we show significant improvement over baselines across a variety of MuJoCo tasks."
                },
                {
                    "title": "Efficient and Near-Optimal Smoothed Online Learning for Generalized Linear Functions",
                    "abstract": "Due to the drastic gap in complexity between sequential and batch statistical learning, recent work has studied a smoothed sequential learning setting, where Nature is constrained to select contexts with density bounded by 1/{\\sigma} with respect to a known measure {\\mu}. Unfortunately, for some function classes, there is an exponential gap between the statistically optimal regret and that which can be achieved efficiently. In this paper, we give a computationally efficient algorithm that is the first to enjoy the statistically optimal log(T/{\\sigma}) regret for realizable K-wise linear classification. We extend our results to settings where the true classifier is linear in an over-parameterized polynomial featurization of the contexts, as well as to a realizable piecewise-regression setting assuming access to an appropriate ERM oracle. Somewhat surprisingly, standard disagreement-based analyses are insufficient to achieve regret logarithmic in 1/{\\sigma}. Instead, we develop a novel characterization of the geometry of the disagreement region induced by generalized linear classifiers. Along the way, we develop numerous technical tools of independent interest, including a general anti-concentration bound for the determinant of certain matrix averages."
                },
                {
                    "title": "Preference Dynamics Under Personalized Recommendations",
                    "abstract": "The design of content recommendation systems underpins many online platforms: social media feeds, online news aggregators, and audio/video hosting websites all choose how best to organize an enormous amount of content for users to consume. Many projects (both practical and academic) have designed algorithms to match users to content they will enjoy under the assumption that user's preferences and opinions do not change with the content they see. However, increasing amounts of evidence suggest that individuals' preferences are directly shaped by what content they see---radicalization, rabbit holes, polarization, and boredom are all example phenomena of preferences affected by content. Polarization in particular can occur even in ecosystems with \"mass media,\" where no personalization takes place, as recently explored in a natural model of preference dynamics by [14] and [13]. If all users' preferences are drawn towards content they already like, or are repelled from content they already dislike, uniform consumption of media leads to a population of heterogeneous preferences converging towards only two poles. In this work, we explore whether some phenomenon akin to polarization occurs when users receive personalized content recommendations. We use a similar model of preference dynamics, where an individual's preferences move towards content the consume and enjoy, and away from content they consume and dislike. We show that standard user reward maximization is an almost trivial goal in such an environment (a large class of simple algorithms will achieve only constant regret). A more interesting objective, then, is to understand under what conditions a recommendation algorithm can ensure stationarity of user's preferences. We show how to design a content recommendations which can achieve approximate stationarity, under mild conditions on the set of available content, when a user's preferences are known, and how one can learn enough about a user's preferences to implement such a strategy even when user preferences are initially unknown."
                },
                {
                    "title": "On the SDEs and Scaling Rules for Adaptive Gradient Algorithms",
                    "abstract": "Approximating Stochastic Gradient Descent (SGD) as a Stochastic Differential Equation (SDE) has allowed researchers to enjoy the benefits of studying a continuous optimization trajectory while carefully preserving the stochasticity of SGD. Analogous study of adaptive gradient methods, such as RMSprop and Adam, has been challenging because there were no rigorously proven SDE approximations for these methods. This paper derives the SDE approximations for RMSprop and Adam, giving theoretical guarantees of their correctness as well as experimental validation of their applicability to common large-scaling vision and language settings. A key practical result is the derivation of a $\\textit{square root scaling rule}$ to adjust the optimization hyperparameters of RMSprop and Adam when changing batch size, and its empirical validation in deep learning settings."
                },
                {
                    "title": "Why Exposure Bias Matters: An Imitation Learning Perspective of Error Accumulation in Language Generation",
                    "abstract": "Current language generation models suffer from issues such as repetition, incoherence, and hallucinations. An often-repeated hypothesis for this brittleness of generation models is that it is caused by the training and the generation procedure mismatch, also referred to as exposure bias. In this paper, we verify this hypothesis by analyzing exposure bias from an imitation learning perspective. We show that exposure bias leads to an accumulation of errors during generation, analyze why perplexity fails to capture this accumulation of errors, and empirically show that this accumulation results in poor generation quality."
                },
                {
                    "title": "Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time",
                    "abstract": "The conventional recipe for maximizing model accuracy is to (1) train multiple models with various hyperparameters and (2) pick the individual model which performs best on a held-out validation set, discarding the remainder. In this paper, we revisit the second step of this procedure in the context of fine-tuning large pre-trained models, where fine-tuned models often appear to lie in a single low error basin. We show that averaging the weights of multiple models fine-tuned with different hyperparameter configurations often improves accuracy and robustness. Unlike a conventional ensemble, we may average many models without incurring any additional inference or memory costs -- we call the results\"model soups.\"When fine-tuning large pre-trained models such as CLIP, ALIGN, and a ViT-G pre-trained on JFT, our soup recipe provides significant improvements over the best model in a hyperparameter sweep on ImageNet. The resulting ViT-G model, which attains 90.94% top-1 accuracy on ImageNet, achieved a new state of the art. Furthermore, we show that the model soup approach extends to multiple image classification and natural language processing tasks, improves out-of-distribution performance, and improves zero-shot performance on new downstream tasks. Finally, we analytically relate the performance similarity of weight-averaging and logit-ensembling to flatness of the loss and confidence of the predictions, and validate this relation empirically. Code is available at https://github.com/mlfoundations/model-soups."
                },
                {
                    "title": "Do Differentiable Simulators Give Better Policy Gradients?",
                    "abstract": "Differentiable simulators promise faster computation time for reinforcement learning by replacing zeroth-order gradient estimates of a stochastic objective with an estimate based on first-order gradients. However, it is yet unclear what factors decide the performance of the two estimators on complex landscapes that involve long-horizon planning and control on physical systems, despite the crucial relevance of this question for the utility of differentiable simulators. We show that characteristics of certain physical systems, such as stiffness or discontinuities, may compromise the efficacy of the first-order estimator, and analyze this phenomenon through the lens of bias and variance. We additionally propose an $\\alpha$-order gradient estimator, with $\\alpha \\in [0,1]$, which correctly utilizes exact gradients to combine the efficiency of first-order estimates with the robustness of zero-order methods. We demonstrate the pitfalls of traditional estimators and the advantages of the $\\alpha$-order estimator on some numerical examples."
                },
                {
                    "title": "The Statistical Complexity of Interactive Decision Making",
                    "abstract": "A fundamental challenge in interactive learning and decision making, ranging from bandit problems to reinforcement learning, is to provide sample-efficient, adaptive learning algorithms that achieve near-optimal regret. This question is analogous to the classical problem of optimal (supervised) statistical learning, where there are well-known complexity measures (e.g., VC dimension and Rademacher complexity) that govern the statistical complexity of learning. However, characterizing the statistical complexity of interactive learning is substantially more challenging due to the adaptive nature of the problem. The main result of this work provides a complexity measure, the Decision-Estimation Coefficient, that is proven to be both necessary and sufficient for sample-efficient interactive learning. In particular, we provide: 1. a lower bound on the optimal regret for any interactive decision making problem, establishing the Decision-Estimation Coefficient as a fundamental limit. 2. a unified algorithm design principle, Estimation-to-Decisions (E2D), which transforms any algorithm for supervised estimation into an online algorithm for decision making. E2D attains a regret bound that matches our lower bound up to dependence on a notion of estimation performance, thereby achieving optimal sample-efficient learning as characterized by the Decision-Estimation Coefficient. Taken together, these results constitute a theory of learnability for interactive decision making. When applied to reinforcement learning settings, the Decision-Estimation Coefficient recovers essentially all existing hardness results and lower bounds. More broadly, the approach can be viewed as a decision-theoretic analogue of the classical Le Cam theory of statistical estimation; it also unifies a number of existing approaches -- both Bayesian and frequentist."
                },
                {
                    "title": "RvS: What is Essential for Offline RL via Supervised Learning?",
                    "abstract": "Recent work has shown that supervised learning alone, without temporal difference (TD) learning, can be remarkably effective for offline RL. When does this hold true, and which algorithmic components are necessary? Through extensive experiments, we boil supervised learning for offline RL down to its essential elements. In every environment suite we consider, simply maximizing likelihood with a two-layer feedforward MLP is competitive with state-of-the-art results of substantially more complex methods based on TD learning or sequence modeling with Transformers. Carefully choosing model capacity (e.g., via regularization or architecture) and choosing which information to condition on (e.g., goals or rewards) are critical for performance. These insights serve as a field guide for practitioners doing Reinforcement Learning via Supervised Learning (which we coin\"RvS learning\"). They also probe the limits of existing RvS methods, which are comparatively weak on random data, and suggest a number of open problems."
                },
                {
                    "title": "Offline Reinforcement Learning: Fundamental Barriers for Value Function Approximation",
                    "abstract": "We consider the offline reinforcement learning problem, where the aim is to learn a decision making policy from logged data. Offline RL -- particularly when coupled with (value) function approximation to allow for generalization in large or continuous state spaces -- is becoming increasingly relevant in practice, because it avoids costly and time-consuming online data collection and is well suited to safety-critical domains. Existing sample complexity guarantees for offline value function approximation methods typically require both (1) distributional assumptions (i.e., good coverage) and (2) representational assumptions (i.e., ability to represent some or all $Q$-value functions) stronger than what is required for supervised learning. However, the necessity of these conditions and the fundamental limits of offline RL are not well understood in spite of decades of research. This led Chen and Jiang (2019) to conjecture that concentrability (the most standard notion of coverage) and realizability (the weakest representation condition) alone are not sufficient for sample-efficient offline RL. We resolve this conjecture in the positive by proving that in general, even if both concentrability and realizability are satisfied, any algorithm requires sample complexity polynomial in the size of the state space to learn a non-trivial policy. Our results show that sample-efficient offline reinforcement learning requires either restrictive coverage conditions or representation conditions that go beyond supervised learning, and highlight a phenomenon called over-coverage which serves as a fundamental barrier for offline value function approximation methods. A consequence of our results for reinforcement learning with linear function approximation is that the separation between online and offline RL can be arbitrarily large, even in constant dimension."
                },
                {
                    "title": "Inductive Biases and Variable Creation in Self-Attention Mechanisms",
                    "abstract": "Self-attention, an architectural motif designed to model long-range interactions in sequential data, has driven numerous recent breakthroughs in natural language processing and beyond. This work provides a theoretical analysis of the inductive biases of self-attention modules. Our focus is to rigorously establish which functions and long-range dependencies self-attention blocks prefer to represent. Our main result shows that bounded-norm Transformer networks\"create sparse variables\": a single self-attention head can represent a sparse function of the input sequence, with sample complexity scaling only logarithmically with the context length. To support our analysis, we present synthetic experiments to probe the sample complexity of learning sparse Boolean functions with Transformers."
                },
                {
                    "title": "Stabilizing Dynamical Systems via Policy Gradient Methods",
                    "abstract": "Stabilizing an unknown control system is one of the most fundamental problems in control systems engineering. In this paper, we provide a simple, model-free algorithm for stabilizing fully observed dynamical systems. While model-free methods have become increasingly popular in practice due to their simplicity and flexibility, stabilization via direct policy search has received surprisingly little attention. Our algorithm proceeds by solving a series of discounted LQR problems, where the discount factor is gradually increased. We prove that this method efficiently recovers a stabilizing controller for linear systems, and for smooth, nonlinear systems within a neighborhood of their equilibria. Our approach overcomes a significant limitation of prior work, namely the need for a pre-given stabilizing control policy. We empirically evaluate the effectiveness of our approach on common control benchmarks."
                },
                {
                    "title": "Implicit Behavioral Cloning",
                    "abstract": "We find that across a wide range of robot policy learning scenarios, treating supervised policy learning with an implicit model generally performs better, on average, than commonly used explicit models. We present extensive experiments on this finding, and we provide both intuitive insight and theoretical arguments distinguishing the properties of implicit models compared to their explicit counterparts, particularly with respect to approximating complex, potentially discontinuous and multi-valued (set-valued) functions. On robotic policy learning tasks we show that implicit behavioral cloning policies with energy-based models (EBM) often outperform common explicit (Mean Square Error, or Mixture Density) behavioral cloning policies, including on tasks with high-dimensional action spaces and visual image inputs. We find these policies provide competitive results or outperform state-of-the-art offline reinforcement learning methods on the challenging human-expert tasks from the D4RL benchmark suite, despite using no reward information. In the real world, robots with implicit policies can learn complex and remarkably subtle behaviors on contact-rich tasks from human demonstrations, including tasks with high combinatorial complexity and tasks requiring 1mm precision."
                },
                {
                    "title": "On the Power of Differentiable Learning versus PAC and SQ Learning",
                    "abstract": "We study the power of learning via mini-batch stochastic gradient descent (SGD) on the population loss, and batch Gradient Descent (GD) on the empirical loss, of a differentiable model or neural network, and ask what learning problems can be learnt using these paradigms. We show that SGD and GD can always simulate learning with statistical queries (SQ), but their ability to go beyond that depends on the precision $\\rho$ of the gradient calculations relative to the minibatch size $b$ (for SGD) and sample size $m$ (for GD). With fine enough precision relative to minibatch size, namely when $b \\rho$ is small enough, SGD can go beyond SQ learning and simulate any sample-based learning algorithm and thus its learning power is equivalent to that of PAC learning; this extends prior work that achieved this result for $b=1$. Similarly, with fine enough precision relative to the sample size $m$, GD can also simulate any sample-based learning algorithm based on $m$ samples. In particular, with polynomially many bits of precision (i.e. when $\\rho$ is exponentially small), SGD and GD can both simulate PAC learning regardless of the mini-batch size. On the other hand, when $b \\rho^2$ is large enough, the power of SGD is equivalent to that of SQ learning."
                },
                {
                    "title": "The MultiBERTs: BERT Reproductions for Robustness Analysis",
                    "abstract": "Experiments with pre-trained models such as BERT are often based on a single checkpoint. While the conclusions drawn apply to the artifact tested in the experiment (i.e., the particular instance of the model), it is not always clear whether they hold for the more general procedure which includes the architecture, training data, initialization scheme, and loss function. Recent work has shown that repeating the pre-training process can lead to substantially different performance, suggesting that an alternate strategy is needed to make principled statements about procedures. To enable researchers to draw more robust conclusions, we introduce the MultiBERTs, a set of 25 BERT-Base checkpoints, trained with similar hyper-parameters as the original BERT model but differing in random weight initialization and shuffling of training data. We also define the Multi-Bootstrap, a non-parametric bootstrap method for statistical inference designed for settings where there are multiple pre-trained models and limited test data. To illustrate our approach, we present a case study of gender bias in coreference resolution, in which the Multi-Bootstrap lets us measure effects that may not be detected with a single checkpoint. We release our models and statistical library along with an additional set of 140 intermediate checkpoints captured during pre-training to facilitate research on learning dynamics."
                },
                {
                    "title": "Offline Reinforcement Learning as One Big Sequence Modeling Problem",
                    "abstract": "Reinforcement learning (RL) is typically concerned with estimating stationary policies or single-step models, leveraging the Markov property to factorize problems in time. However, we can also view RL as a generic sequence modeling problem, with the goal being to produce a sequence of actions that leads to a sequence of high rewards. Viewed in this way, it is tempting to consider whether high-capacity sequence prediction models that work well in other domains, such as natural-language processing, can also provide effective solutions to the RL problem. To this end, we explore how RL can be tackled with the tools of sequence modeling, using a Transformer architecture to model distributions over trajectories and repurposing beam search as a planning algorithm. Framing RL as sequence modeling problem simplifies a range of design decisions, allowing us to dispense with many of the components common in offline RL algorithms. We demonstrate the flexibility of this approach across long-horizon dynamics prediction, imitation learning, goal-conditioned RL, and offline RL. Further, we show that this approach can be combined with existing model-free algorithms to yield a state-of-the-art planner in sparse-reward, long-horizon tasks."
                },
                {
                    "title": "On the Stability of Nonlinear Receding Horizon Control: A Geometric Perspective",
                    "abstract": "The widespread adoption of nonlinear Receding Horizon Control (RHC) strategies by industry has led to more than 30 years of intense research efforts to provide stability guarantees for these methods. However, current theoretical guarantees require that each (generally nonconvex) planning problem can be solved to (approximate) global optimality, which is an unrealistic requirement for the derivative-based local optimization methods generally used in practical implementations of RHC. This paper takes the first step towards understanding stability guarantees for nonlinear RHC when the inner planning problem is solved to first-order stationary points, but not necessarily global optima. Special attention is given to feedback linearizable systems, and a mixture of positive and negative results are provided. We establish that, under certain strong conditions, first-order solutions to RHC exponentially stabilize linearizable systems. Crucially, this guarantee requires that state costs applied to the planning problems are in a certain sense \u2018compatible\u2019 with the global geometry of the system, and a simple counter-example demonstrates the necessity of this condition. These results highlight the need to rethink the role of global geometry in the context of optimization-based control."
                },
                {
                    "title": "On Imitation Learning of Linear Control Policies: Enforcing Stability and Robustness Constraints via LMI Conditions",
                    "abstract": "When applying imitation learning techniques to fit a policy from expert demonstrations, one can take advantage of prior stability/robustness assumptions on the expert's policy and incorporate such control-theoretic prior knowledge explicitly into the learning process. In this paper, we formulate the imitation learning of linear policies as a constrained optimization problem, and present efficient methods which can be used to enforce stability and robustness constraints during the learning processes. Specifically, we show that one can guarantee the closed-loop stability and robustness by posing linear matrix inequality (LMI) constraints on the fitted policy. Then both the projected gradient descent method and the alternating direction method of multipliers (ADMM) method can be applied to solve the resultant constrained policy fitting problem. Finally, we provide numerical results to demonstrate the effectiveness of our methods in producing linear polices with various stability and robustness guarantees."
                },
                {
                    "title": "Bilinear Classes: A Structural Framework for Provable Generalization in RL",
                    "abstract": "This work introduces Bilinear Classes, a new structural framework, which permit generalization in reinforcement learning in a wide variety of settings through the use of function approximation. The framework incorporates nearly all existing models in which a polynomial sample complexity is achievable, and, notably, also includes new models, such as the Linear $Q^*/V^*$ model in which both the optimal $Q$-function and the optimal $V$-function are linear in some known feature space. Our main result provides an RL algorithm which has polynomial sample complexity for Bilinear Classes; notably, this sample complexity is stated in terms of a reduction to the generalization error of an underlying supervised learning sub-problem. These bounds nearly match the best known sample complexity bounds for existing models. Furthermore, this framework also extends to the infinite dimensional (RKHS) setting: for the the Linear $Q^*/V^*$ model, linear MDPs, and linear mixture MDPs, we provide sample complexities that have no explicit dependence on the explicit feature dimension (which could be infinite), but instead depends only on information theoretic quantities."
                },
                {
                    "title": "Instabilities of Offline RL with Pre-Trained Neural Representation",
                    "abstract": "In offline reinforcement learning (RL), we seek to utilize offline data to evaluate (or learn) policies in scenarios where the data are collected from a distribution that substantially differs from that of the target policy to be evaluated. Recent theoretical advances have shown that such sample-efficient offline RL is indeed possible provided certain strong representational conditions hold, else there are lower bounds exhibiting exponential error amplification (in the problem horizon) unless the data collection distribution has only a mild distribution shift relative to the target policy. This work studies these issues from an empirical perspective to gauge how stable offline RL methods are. In particular, our methodology explores these ideas when using features from pre-trained neural networks, in the hope that these representations are powerful enough to permit sample efficient offline RL. Through extensive experiments on a range of tasks, we see that substantial error amplification does occur even when using such pre-trained representations (trained on the same task itself); we find offline RL is stable only under extremely mild distribution shift. The implications of these results, both from a theoretical and an empirical perspective, are that successful offline RL (where we seek to go beyond the low distribution shift regime) requires substantially stronger conditions beyond those which suffice for successful supervised learning."
                },
                {
                    "title": "On the Sample Complexity of Stability Constrained Imitation Learning",
                    "abstract": "We study the following question in the context of imitation learning for continuous control: how are the underlying stability properties of an expert policy reflected in the sample-complexity of an imitation learning task? We provide the first results showing that a surprisingly granular connection can be made between the underlying expert system's incremental gain stability, a novel measure of robust convergence between pairs of system trajectories, and the dependency on the task horizon $T$ of the resulting generalization bounds. In particular, we propose and analyze incremental gain stability constrained versions of behavior cloning and a DAgger-like algorithm, and show that the resulting sample-complexity bounds naturally reflect the underlying stability properties of the expert system. As a special case, we delineate a class of systems for which the number of trajectories needed to achieve $\\varepsilon$-suboptimality is sublinear in the task horizon $T$, and do so without requiring (strong) convexity of the loss function in the policy parameters. Finally, we conduct numerical experiments demonstrating the validity of our insights on both a simple nonlinear system for which the underlying stability properties can be easily tuned, and on a high-dimensional quadrupedal robotic simulation."
                },
                {
                    "title": "Bellman Eluder Dimension: New Rich Classes of RL Problems, and Sample-Efficient Algorithms",
                    "abstract": "Finding the minimal structural assumptions that empower sample-efficient learning is one of the most important research directions in Reinforcement Learning (RL). This paper advances our understanding of this fundamental question by introducing a new complexity measure -- Bellman Eluder (BE) dimension. We show that the family of RL problems of low BE dimension is remarkably rich, which subsumes a vast majority of existing tractable RL problems including but not limited to tabular MDPs, linear MDPs, reactive POMDPs, low Bellman rank problems as well as low Eluder dimension problems. This paper further designs a new optimization-based algorithm -- GOLF, and reanalyzes a hypothesis elimination-based algorithm -- OLIVE (proposed in Jiang et al., 2017). We prove that both algorithms learn the near-optimal policies of low BE dimension problems in a number of samples that is polynomial in all relevant parameters, but independent of the size of state-action space. Our regret and sample complexity results match or improve the best existing results for several well-known subclasses of low BE dimension problems."
                },
                {
                    "title": "Exponential Lower Bounds for Batch Reinforcement Learning: Batch RL can be Exponentially Harder than Online RL",
                    "abstract": "Several practical applications of reinforcement learning involve an agent learning from past data without the possibility of further exploration. Often these applications require us to 1) identify a near optimal policy or to 2) estimate the value of a target policy. For both tasks we derive \\emph{exponential} information-theoretic lower bounds in discounted infinite horizon MDPs with a linear function representation for the action value function even if 1) \\emph{realizability} holds, 2) the batch algorithm observes the exact reward and transition \\emph{functions}, and 3) the batch algorithm is given the \\emph{best} a priori data distribution for the problem class. Furthermore, if the dataset does not come from policy rollouts then the lower bounds hold even if the action-value function of \\emph{every} policy admits a linear representation. \nIf the objective is to find a near-optimal policy, we discover that these hard instances are easily solved by an \\emph{online} algorithm, showing that there exist RL problems where \\emph{batch RL is exponentially harder than online RL} even under the most favorable batch data distribution. In other words, online exploration is critical to enable sample efficient RL with function approximation. A second corollary is the exponential separation between finite and infinite horizon batch problems under our assumptions. \nOn a technical level, this work introduces a new `oracle + batch algorithm' framework to prove lower bounds that hold for every distribution, and automatically recovers traditional fixed distribution lower bounds as a special case. Finally this work helps formalize the issue known as \\emph{deadly triad} and explains that the \\emph{bootstrapping} problem \\citep{sutton2018reinforcement} is potentially more severe than the \\emph{extrapolation} issue for RL because unlike the latter, bootstrapping cannot be mitigated by adding more samples."
                },
                {
                    "title": "Neural Mechanics: Symmetry and Broken Conservation Laws in Deep Learning Dynamics",
                    "abstract": "Predicting the dynamics of neural network parameters during training is one of the key challenges in building a theoretical foundation for deep learning. A central obstacle is that the motion of a network in high-dimensional parameter space undergoes discrete finite steps along complex stochastic gradients derived from real-world datasets. We circumvent this obstacle through a unifying theoretical framework based on intrinsic symmetries embedded in a network's architecture that are present for any dataset. We show that any such symmetry imposes stringent geometric constraints on gradients and Hessians, leading to an associated conservation law in the continuous-time limit of stochastic gradient descent (SGD), akin to Noether's theorem in physics. We further show that finite learning rates used in practice can actually break these symmetry induced conservation laws. We apply tools from finite difference methods to derive modified gradient flow, a differential equation that better approximates the numerical trajectory taken by SGD at finite learning rates. We combine modified gradient flow with our framework of symmetries to derive exact integral expressions for the dynamics of certain parameter combinations. We empirically validate our analytic predictions for learning dynamics on VGG-16 trained on Tiny ImageNet. Overall, by exploiting symmetry, our work demonstrates that we can analytically describe the learning dynamics of various parameter combinations at finite learning rates and batch sizes for state of the art architectures trained on any dataset."
                },
                {
                    "title": "Grasping with Chopsticks: Combating Covariate Shift in Model-free Imitation Learning for Fine Manipulation",
                    "abstract": "Billions of people use chopsticks, a simple yet versatile tool, for fine manipulation of everyday objects. The small, curved, and slippery tips of chopsticks pose a challenge for picking up small objects, making them a suitably complex test case. This paper leverages human demonstrations to develop an autonomous chopsticks-equipped robotic manipulator. Due to the lack of accurate models for fine manipulation, we explore model-free imitation learning, which traditionally suffers from the covariate shift phenomenon that causes poor generalization. We propose two approaches to reduce covariate shift, neither of which requires access to an interactive expert or a model, unlike previous approaches. First, we alleviate singlestep prediction errors by applying an invariant operator to increase the data support at critical steps for grasping. Second, we generate synthetic corrective labels by adding bounded noise and combining parametric and non-parametric methods to prevent error accumulation. We demonstrate our methods on a real chopstick-equipped robot that we built, and observe the agent\u2019s success rate increase from 37.3% to 80%, which is comparable to the human expert performance of 82.6%."
                },
                {
                    "title": "Do Offline Metrics Predict Online Performance in Recommender Systems?",
                    "abstract": "Recommender systems operate in an inherently dynamical setting. Past recommendations influence future behavior, including which data points are observed and how user preferences change. However, experimenting in production systems with real user dynamics is often infeasible, and existing simulation-based approaches have limited scale. As a result, many state-of-the-art algorithms are designed to solve supervised learning problems, and progress is judged only by offline metrics. In this work we investigate the extent to which offline metrics predict online performance by evaluating eleven recommenders across six controlled simulated environments. We observe that offline metrics are correlated with online performance over a range of environments. However, improvements in offline metrics lead to diminishing returns in online performance. Furthermore, we observe that the ranking of recommenders varies depending on the amount of initial offline data available. We study the impact of adding exploration strategies, and observe that their effectiveness, when compared to greedy recommendation, is highly dependent on the recommendation algorithm. We provide the environments and recommenders described in this paper as Reclab: an extensible ready-to-use simulation framework at this https URL."
                },
                {
                    "title": "What are the Statistical Limits of Offline RL with Linear Function Approximation?",
                    "abstract": "Offline reinforcement learning seeks to utilize offline (observational) data to guide the learning of (causal) sequential decision making strategies. The hope is that offline reinforcement learning coupled with function approximation methods (to deal with the curse of dimensionality) can provide a means to help alleviate the excessive sample complexity burden in modern sequential decision making problems. However, the extent to which this broader approach can be effective is not well understood, where the literature largely consists of sufficient conditions. \nThis work focuses on the basic question of what are necessary representational and distributional conditions that permit provable sample-efficient offline reinforcement learning. Perhaps surprisingly, our main result shows that even if: i) we have realizability in that the true value function of \\emph{every} policy is linear in a given set of features and 2) our off-policy data has good coverage over all features (under a strong spectral condition), then any algorithm still (information-theoretically) requires a number of offline samples that is exponential in the problem horizon in order to non-trivially estimate the value of \\emph{any} given policy. Our results highlight that sample-efficient offline policy evaluation is simply not possible unless significantly stronger conditions hold; such conditions include either having low distribution shift (where the offline data distribution is close to the distribution of the policy to be evaluated) or significantly stronger representational conditions (beyond realizability)."
                },
                {
                    "title": "Reconciling Modern Deep Learning with Traditional Optimization Analyses: The Intrinsic Learning Rate",
                    "abstract": "Recent works (e.g., (Li and Arora, 2020)) suggest that the use of popular normalization schemes (including Batch Normalization) in today's deep learning can move it far from a traditional optimization viewpoint, e.g., use of exponentially increasing learning rates. The current paper highlights other ways in which behavior of normalized nets departs from traditional viewpoints, and then initiates a formal framework for studying their mathematics via suitable adaptation of the conventional framework namely, modeling SGD-induced training trajectory via a suitable stochastic differential equation (SDE) with a noise term that captures gradient noise. This yields: (a) A new ' intrinsic learning rate' parameter that is the product of the normal learning rate and weight decay factor. Analysis of the SDE shows how the effective speed of learning varies and equilibrates over time under the control of intrinsic LR. (b) A challenge -- via theory and experiments -- to popular belief that good generalization requires large learning rates at the start of training. (c) New experiments, backed by mathematical intuition, suggesting the number of steps to equilibrium (in function space) scales as the inverse of the intrinsic learning rate, as opposed to the exponential time convergence bound implied by SDE analysis. We name it the Fast Equilibrium Conjecture and suggest it holds the key to why Batch Normalization is effective."
                },
                {
                    "title": "Black-Box Control for Linear Dynamical Systems",
                    "abstract": "We consider the problem of controlling an unknown linear time-invariant dynamical system from a single chain of black-box interactions, and with no access to resets or offline simulation. Under the assumption that the system is controllable, we give the first efficient algorithm that is capable of attaining sublinear regret in a single trajectory under the setting of online nonstochastic control. We give finite-time regret bound of our algorithm, as well as a nearly-matching lower bound that shows this regret to be almost best-attainable by any algorithm."
                },
                {
                    "title": "Descending through a Crowded Valley - Benchmarking Deep Learning Optimizers",
                    "abstract": "Choosing the optimizer is among the most crucial decisions of deep learning engineers, and it is not an easy one. The growing literature now lists literally hundreds of optimization methods. In the absence of clear theoretical guidance and conclusive empirical evidence, the decision is often done according to personal anecdotes. In this work, we aim to replace these anecdotes, if not with evidence, then at least with heuristics. To do so, we perform an extensive, standardized benchmark of more than a dozen particularly popular deep learning optimizers while giving a concise overview of the wide range of possible choices. Analyzing almost 35 000 individual runs, we contribute the following three points: Optimizer performance varies greatly across tasks. We observe that evaluating multiple optimizers with default parameters works approximately as well as tuning the hyperparameters of a single, fixed optimizer. While we can not identify an individual optimization method clearly dominating across all tested tasks, we identify a significantly reduced subset of specific algorithms and parameter choices that generally provided competitive results in our experiments. This subset includes popular favorites and some less well-known contenders. We have open-sourced all our experimental results, making it available to use as well-tuned baselines when evaluating novel optimization methods and therefore reducing the necessary computational efforts."
                },
                {
                    "title": "Information Theoretic Regret Bounds for Online Nonlinear Control",
                    "abstract": "This work studies the problem of sequential control in an unknown, nonlinear dynamical system, where we model the underlying system dynamics as an unknown function in a known Reproducing Kernel Hilbert Space. This framework yields a general setting that permits discrete and continuous control inputs as well as non-smooth, non-differentiable dynamics. Our main result, the Lower Confidence-based Continuous Control ($LC^3$) algorithm, enjoys a near-optimal $O(\\sqrt{T})$ regret bound against the optimal controller in episodic settings, where $T$ is the number of episodes. The bound has no explicit dependence on dimension of the system dynamics, which could be infinite, but instead only depends on information theoretic quantities. We empirically show its application to a number of nonlinear control tasks and demonstrate the benefit of exploration for learning model dynamics."
                },
                {
                    "title": "Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning",
                    "abstract": "We introduce Bootstrap Your Own Latent (BYOL), a new approach to self-supervised image representation learning. BYOL relies on two neural networks, referred to as online and target networks, that interact and learn from each other. From an augmented view of an image, we train the online network to predict the target network representation of the same image under a different augmented view. At the same time, we update the target network with a slow-moving average of the online network. While state-of-the art methods rely on negative pairs, BYOL achieves a new state of the art without them. BYOL reaches $74.3\\%$ top-1 classification accuracy on ImageNet using a linear evaluation with a ResNet-50 architecture and $79.6\\%$ with a larger ResNet. We show that BYOL performs on par or better than the current state of the art on both transfer and semi-supervised benchmarks. Our implementation and pretrained models are given on GitHub."
                },
                {
                    "title": "Making Non-Stochastic Control (Almost) as Easy as Stochastic",
                    "abstract": "Recent literature has made much progress in understanding \\emph{online LQR}: a modern learning-theoretic take on the classical control problem in which a learner attempts to optimally control an unknown linear dynamical system with fully observed state, perturbed by i.i.d. Gaussian noise. It is now understood that the optimal regret on time horizon $T$ against the optimal control law scales as $\\widetilde{\\Theta}(\\sqrt{T})$. In this paper, we show that the same regret rate (against a suitable benchmark) is attainable even in the considerably more general non-stochastic control model, where the system is driven by \\emph{arbitrary adversarial} noise (Agarwal et al. 2019). In other words, \\emph{stochasticity confers little benefit in online LQR}. \nWe attain the optimal $\\widetilde{\\mathcal{O}}(\\sqrt{T})$ regret when the dynamics are unknown to the learner, and $\\mathrm{poly}(\\log T)$ regret when known, provided that the cost functions are strongly convex (as in LQR). Our algorithm is based on a novel variant of online Newton step (Hazan et al. 2007), which adapts to the geometry induced by possibly adversarial disturbances, and our analysis hinges on generic \"policy regret\" bounds for certain structured losses in the OCO-with-memory framework (Anava et al. 2015). Moreover, our results accomodate the full generality of the non-stochastic control setting: adversarially chosen (possibly non-quadratic) costs, partial state observation, and fully adversarial process and observation noise."
                },
                {
                    "title": "Reinforcement Learning with General Value Function Approximation: Provably Efficient Approach via Bounded Eluder Dimension",
                    "abstract": "Value function approximation has demonstrated phenomenal empirical success in reinforcement learning (RL). Nevertheless, despite a handful of recent progress on developing theory for RL with linear function approximation, the understanding of general function approximation schemes largely remains missing. In this paper, we establish a provably efficient RL algorithm with general value function approximation. We show that if the value functions admit an approximation with a function class $\\mathcal{F}$, our algorithm achieves a regret bound of $\\widetilde{O}(\\mathrm{poly}(dH)\\sqrt{T})$ where $d$ is a complexity measure of $\\mathcal{F}$ that depends on the eluder dimension [Russo and Van Roy, 2013] and log-covering numbers, $H$ is the planning horizon, and $T$ is the number interactions with the environment. Our theory generalizes recent progress on RL with linear value function approximation and does not make explicit assumptions on the model of the environment. Moreover, our algorithm is model-free and provides a framework to justify the effectiveness of algorithms used in practice."
                },
                {
                    "title": "Implementation Matters in Deep RL: A Case Study on PPO and TRPO",
                    "abstract": "We study the roots of algorithmic progress in deep policy gradient algorithms through a case study on two popular algorithms, Proximal Policy Optimization and Trust Region Policy Optimization. We investigate the consequences of \"code-level optimizations:\" algorithm augmentations found only in implementations or described as auxiliary details to the core algorithm. Seemingly of secondary importance, such optimizations have a major impact on agent behavior. Our results show that they (a) are responsible for most of PPO's gain in cumulative reward over TRPO, and (b) fundamentally change how RL methods function. These insights show the difficulty, and importance, of attributing performance gains in deep reinforcement learning."
                },
                {
                    "title": "Disentangling Adaptive Gradient Methods from Learning Rates",
                    "abstract": "We investigate several confounding factors in the evaluation of optimization algorithms for deep learning. Primarily, we take a deeper look at how adaptive gradient methods interact with the learning rate schedule, a notoriously difficult-to-tune hyperparameter which has dramatic effects on the convergence and generalization of neural network training. We introduce a \"grafting\" experiment which decouples an update's magnitude from its direction, finding that many existing beliefs in the literature may have arisen from insufficient isolation of the implicit schedule of step sizes. Alongside this contribution, we present some empirical and theoretical retrospectives on the generalization of adaptive gradient methods, aimed at bringing more clarity to this space."
                },
                {
                    "title": "No-Regret Prediction in Marginally Stable Systems",
                    "abstract": "We consider the problem of online prediction in a marginally stable linear dynamical system subject to bounded adversarial or (non-isotropic) stochastic perturbations. This poses two challenges. Firstly, the system is in general unidentifiable, so recent and classical results on parameter recovery do not apply. Secondly, because we allow the system to be marginally stable, the state can grow polynomially with time; this causes standard regret bounds in online convex optimization to be vacuous. In spite of these challenges, we show that the online least-squares algorithm achieves sublinear regret (improvable to polylogarithmic in the stochastic setting), with polynomial dependence on the system's parameters. This requires a refined regret analysis, including a structural lemma showing the current state of the system to be a small linear combination of past states, even if the state grows polynomially. By applying our techniques to learning an autoregressive filter, we also achieve logarithmic regret in the partially observed setting under Gaussian noise, with polynomial dependence on the memory of the associated Kalman filter."
                },
                {
                    "title": "Naive Exploration is Optimal for Online LQR",
                    "abstract": "We consider the problem of online adaptive control of the linear quadratic regulator, where the true system parameters are unknown. We prove new upper and lower bounds demonstrating that the optimal regret scales as $\\widetilde{\\Theta}({\\sqrt{d_{\\mathbf{u}}^2 d_{\\mathbf{x}} T}})$, where $T$ is the number of time steps, $d_{\\mathbf{u}}$ is the dimension of the input space, and $d_{\\mathbf{x}}$ is the dimension of the system state. Notably, our lower bounds rule out the possibility of a $\\mathrm{poly}(\\log{}T)$-regret algorithm, which had been conjectured due to the apparent strong convexity of the problem. Our upper bound is attained by a simple variant of $\\textit{certainty equivalent control}$, where the learner selects control inputs according to the optimal controller for their estimate of the system while injecting exploratory random noise. While this approach was shown to achieve $\\sqrt{T}$-regret by (Mania et al. 2019), we show that if the learner continually refines their estimates of the system matrices, the method attains optimal dimension dependence as well. \nCentral to our upper and lower bounds is a new approach for controlling perturbations of Riccati equations called the $\\textit{self-bounding ODE method}$, which we use to derive suboptimality bounds for the certainty equivalent controller synthesized from estimated system dynamics. This in turn enables regret upper bounds which hold for $\\textit{any stabilizable instance}$ and scale with natural control-theoretic quantities."
                },
                {
                    "title": "Improper Learning for Non-Stochastic Control",
                    "abstract": "We consider the problem of controlling a possibly unknown linear dynamical system with adversarial perturbations, adversarially chosen convex loss functions, and partially observed states, known as non-stochastic control. We introduce a controller parametrization based on the denoised observations, and prove that applying online gradient descent to this parametrization yields a new controller which attains sublinear regret vs. a large class of closed-loop policies. In the fully-adversarial setting, our controller attains an optimal regret bound of $\\sqrt{T}$-when the system is known, and, when combined with an initial stage of least-squares estimation, $T^{2/3}$ when the system is unknown; both yield the first sublinear regret for the partially observed setting. \nOur bounds are the first in the non-stochastic control setting that compete with \\emph{all} stabilizing linear dynamical controllers, not just state feedback. Moreover, in the presence of semi-adversarial noise containing both stochastic and adversarial components, our controller attains the optimal regret bounds of $\\mathrm{poly}(\\log T)$ when the system is known, and $\\sqrt{T}$ when unknown. To our knowledge, this gives the first end-to-end $\\sqrt{T}$ regret for online Linear Quadratic Gaussian controller, and applies in a more general setting with adversarial losses and semi-adversarial noise."
                },
                {
                    "title": "Poly-time universality and limitations of deep learning",
                    "abstract": "The goal of this paper is to characterize function distributions that deep learning can or cannot learn in poly-time. A universality result is proved for SGD-based deep learning and a non-universality result is proved for GD-based deep learning; this also gives a separation between SGD-based deep learning and statistical query algorithms: \n(1) {\\it Deep learning with SGD is efficiently universal.} Any function distribution that can be learned from samples in poly-time can also be learned by a poly-size neural net trained with SGD on a poly-time initialization with poly-steps, poly-rate and possibly poly-noise. \nTherefore deep learning provides a universal learning paradigm: it was known that the approximation and estimation errors could be controlled with poly-size neural nets, using ERM that is NP-hard; this new result shows that the optimization error can also be controlled with SGD in poly-time. The picture changes for GD with large enough batches: \n(2) {\\it Result (1) does not hold for GD:} Neural nets of poly-size trained with GD (full gradients or large enough batches) on any initialization with poly-steps, poly-range and at least poly-noise cannot learn any function distribution that has super-polynomial {\\it cross-predictability,} where the cross-predictability gives a measure of ``average'' function correlation -- relations and distinctions to the statistical dimension are discussed. In particular, GD with these constraints can learn efficiently monomials of degree $k$ if and only if $k$ is constant. \nThus (1) and (2) point to an interesting contrast: SGD is universal even with some poly-noise while full GD or SQ algorithms are not (e.g., parities)."
                },
                {
                    "title": "Linear Mode Connectivity and the Lottery Ticket Hypothesis",
                    "abstract": "We study whether a neural network optimizes to the same, linearly connected minimum under different samples of SGD noise (e.g., random data order and augmentation). We find that standard vision models become stable to SGD noise in this way early in training. From then on, the outcome of optimization is determined to a linearly connected region. We use this technique to study iterative magnitude pruning (IMP), the procedure used by work on the lottery ticket hypothesis to identify subnetworks that could have trained in isolation to full accuracy. We find that these subnetworks only reach full accuracy when they are stable to SGD noise, which either occurs at initialization for small-scale settings (MNIST) or early in training for large-scale settings (ResNet-50 and Inception-v3 on ImageNet)."
                },
                {
                    "title": "Convex Formulation of Overparameterized Deep Neural Networks",
                    "abstract": "The analysis of over-parameterized neural networks has drawn significant attention in recent years. It was shown that such systems behave like convex systems under various restricted settings, such as for two-layer neural networks, and when learning is only restricted locally in the so-called neural tangent kernel space around specialized initializations. However, there is a lack of powerful theoretical techniques that can analyze fully trained deep neural networks under general conditions. This paper considers this fundamental problem by investigating such overparameterized deep neural networks when fully trained. Specifically, we characterize a deep neural network by its features\u2019 distributions and propose a metric to intuitively measure the usefulness of feature representations. Under certain regularizers that bounds the metric, we show deep neural networks can be reformulated as a convex optimization and the system can guarantee effective feature representations in terms of the metric. Our new analysis is more consistent with empirical observations that deep neural networks are capable of learning efficient feature representations. Empirical studies confirm that predictions of our theory are consistent with results observed in practice."
                },
                {
                    "title": "On the Expressivity of Neural Networks for Deep Reinforcement Learning",
                    "abstract": "We compare the model-free reinforcement learning with the model-based approaches through the lens of the expressive power of neural networks for policies, $Q$-functions, and dynamics. We show, theoretically and empirically, that even for one-dimensional continuous state space, there are many MDPs whose optimal $Q$-functions and policies are much more complex than the dynamics. We hypothesize many real-world MDPs also have a similar property. For these MDPs, model-based planning is a favorable algorithm, because the resulting policies can approximate the optimal policy significantly better than a neural network parameterization can, and model-free or model-based policy optimization rely on policy parameterization. Motivated by the theory, we apply a simple multi-step model-based bootstrapping planner (BOOTS) to bootstrap a weak $Q$-function into a stronger policy. Empirical results show that applying BOOTS on top of model-based or model-free policy optimization algorithms at the test time improves the performance on MuJoCo benchmark tasks."
                },
                {
                    "title": "Towards Explaining the Regularization Effect of Initial Large Learning Rate in Training Neural Networks",
                    "abstract": "Stochastic gradient descent with a large initial learning rate is a widely adopted method for training modern neural net architectures. Although a small initial learning rate allows for faster training and better test performance initially, the large learning rate achieves better generalization soon after the learning rate is annealed. Towards explaining this phenomenon, we devise a setting in which we can prove that a two layer network trained with large initial learning rate and annealing provably generalizes better than the same network trained with a small learning rate from the start. The key insight in our analysis is that the order of learning different types of patterns is crucial: because the small learning rate model first memorizes low noise, hard-to-fit patterns, it generalizes worse on higher noise, easier-to-fit patterns than its large learning rate counterpart. This concept translates to a larger-scale setting: we demonstrate that one can add a small patch to CIFAR-10 images that is immediately memorizable by a model with small initial learning rate, but ignored by the model with large learning rate until after annealing. Our experiments show that this causes the small learning rate model's accuracy on unmodified images to suffer, as it relies too much on the patch early on."
                },
                {
                    "title": "Stochastic Gradient and Langevin Processes",
                    "abstract": "We prove quantitative convergence rates at which discrete Langevin-like processes converge to the invariant distribution of a related stochastic differential equation. We study the setup where the additive noise can be non-Gaussian and state-dependent and the potential function can be non-convex. We show that the key properties of these processes depend on the potential function and the second moment of the additive noise. We apply our theoretical findings to studying the convergence of Stochastic Gradient Descent (SGD) for non-convex problems and corroborate them with experiments using SGD to train deep neural networks on the CIFAR-10 dataset."
                },
                {
                    "title": "Calibration, Entropy Rates, and Memory in Language Models",
                    "abstract": "Building accurate language models that capture meaningful long-term dependencies is a core challenge in natural language processing. Towards this end, we present a calibration-based approach to measure long-term discrepancies between a generative sequence model and the true distribution, and use these discrepancies to improve the model. Empirically, we show that state-of-the-art language models, including LSTMs and Transformers, are \\emph{miscalibrated}: the entropy rates of their generations drift dramatically upward over time. We then provide provable methods to mitigate this phenomenon. Furthermore, we show how this calibration-based approach can also be used to measure the amount of memory that language models use for prediction."
                },
                {
                    "title": "Stabilizing Off-Policy Q-Learning via Bootstrapping Error Reduction",
                    "abstract": "Off-policy reinforcement learning aims to leverage experience collected from prior policies for sample-efficient learning. However, in practice, commonly used off-policy approximate dynamic programming methods based on Q-learning and actor-critic methods are highly sensitive to the data distribution, and can make only limited progress without collecting additional on-policy data. As a step towards more robust off-policy algorithms, we study the setting where the off-policy experience is fixed and there is no further interaction with the environment. We identify bootstrapping error as a key source of instability in current methods. Bootstrapping error is due to bootstrapping from actions that lie outside of the training data distribution, and it accumulates via the Bellman backup operator. We theoretically analyze bootstrapping error, and demonstrate how carefully constraining action selection in the backup can mitigate it. Based on our analysis, we propose a practical algorithm, bootstrapping error accumulation reduction (BEAR). We demonstrate that BEAR is able to learn robustly from different off-policy distributions, including random and suboptimal demonstrations, on a range of continuous control tasks."
                },
                {
                    "title": "The Curious Case of Neural Text Degeneration",
                    "abstract": "Despite considerable advancements with deep neural language models, the enigma of neural text degeneration persists when these models are tested as text generators. The counter-intuitive empirical observation is that even though the use of likelihood as training objective leads to high quality models for a broad range of language understanding tasks, using likelihood as a decoding objective leads to text that is bland and strangely repetitive. \nIn this paper, we reveal surprising distributional differences between human text and machine text. In addition, we find that decoding strategies alone can dramatically effect the quality of machine text, even when generated from exactly the same neural language model. Our findings motivate Nucleus Sampling, a simple but effective method to draw the best out of neural generation. By sampling text from the dynamic nucleus of the probability distribution, which allows for diversity while effectively truncating the less reliable tail of the distribution, the resulting text better demonstrates the quality of human text, yielding enhanced diversity without sacrificing fluency and coherence."
                },
                {
                    "title": "Online Control with Adversarial Disturbances",
                    "abstract": "We study the control of a linear dynamical system with adversarial disturbances (as opposed to statistical noise). The objective we consider is one of regret: we desire an online control procedure that can do nearly as well as that of a procedure that has full knowledge of the disturbances in hindsight. Our main result is an efficient algorithm that provides nearly tight regret bounds for this problem. From a technical standpoint, this work generalizes upon previous work in two main aspects: our model allows for adversarial noise in the dynamics, and allows for general convex costs."
                },
                {
                    "title": "PIPPS: Flexible Model-Based Policy Search Robust to the Curse of Chaos",
                    "abstract": "\u00a935th International Conference on Machine Learning, ICML 2018.All Rights Reserved. Previously, the exploding gradient problem has been explained to be central in deep learning and model-based reinforcement learning, because it causes numerical issues and instability in optimization. Our experiments in model-based reinforcement learning imply that the problem is not just a numerical issue, but it may be caused by a fundamental chaos-like nature of long chains of nonlinear computations. Not only do the magnitudes of the gradients become large, the direction of the gradients becomes essentially random. We show that reparameterization gradients suffer from the problem, while likelihood ratio gradients are robust. Using our insights, we develop a model-based policy search framework, Probabilistic Inference for Particle-Based Policy Search (PIPPS), which is easily extensible, and allows for almost arbitrary models and policies, while simultaneously matching the performance of previous data-efficient learning algorithms. Finally, we invent the total propagation algorithm, which efficiently computes a union over all pathwise derivative depths during a single backwards pass, automatically giving greater weight to estimators with lower variance, sometimes improving over reparameterization gradients by 106 times."
                },
                {
                    "title": "Asymmetric Valleys: Beyond Sharp and Flat Local Minima",
                    "abstract": "Despite the non-convex nature of their loss functions, deep neural networks are known to generalize well when optimized with stochastic gradient descent (SGD). Recent work conjectures that SGD with proper configuration is able to find wide and flat local minima, which have been proposed to be associated with good generalization performance. In this paper, we observe that local minima of modern deep networks are more than being flat or sharp. Specifically, at a local minimum there exist many asymmetric directions such that the loss increases abruptly along one side, and slowly along the opposite side--we formally define such minima as asymmetric valleys. Under mild assumptions, we prove that for asymmetric valleys, a solution biased towards the flat side generalizes better than the exact minimizer. Further, we show that simply averaging the weights along the SGD trajectory gives rise to such biased solutions implicitly. This provides a theoretical explanation for the intriguing phenomenon observed by Izmailov et al. (2018). In addition, we empirically find that batch normalization (BN) appears to be a major cause for asymmetric valleys."
                },
                {
                    "title": "On Lazy Training in Differentiable Programming",
                    "abstract": "In a series of recent theoretical works, it was shown that strongly over-parameterized neural networks trained with gradient-based methods could converge exponentially fast to zero training loss, with their parameters hardly varying. In this work, we show that this \"lazy training\" phenomenon is not specific to over-parameterized neural networks, and is due to a choice of scaling, often implicit, that makes the model behave as its linearization around the initialization, thus yielding a model equivalent to learning with positive-definite kernels. Through a theoretical analysis, we exhibit various situations where this phenomenon arises in non-convex optimization and we provide bounds on the distance between the lazy and linearized optimization paths. Our numerical experiments bring a critical note, as we observe that the performance of commonly used non-linear deep convolutional neural networks in computer vision degrades when trained in the lazy regime. This makes it unlikely that \"lazy training\" is behind the many successes of neural networks in difficult high dimensional tasks."
                },
                {
                    "title": "On the Ineffectiveness of Variance Reduced Optimization for Deep Learning",
                    "abstract": "The application of stochastic variance reduction to optimization has shown remarkable recent theoretical and practical success. The applicability of these techniques to the hard non-convex optimization problems encountered during training of modern deep neural networks is an open problem. We show that naive application of the SVRG technique and related approaches fail, and explore why."
                },
                {
                    "title": "Off-Policy Deep Reinforcement Learning without Exploration",
                    "abstract": "Many practical applications of reinforcement learning constrain agents to learn from a fixed batch of data which has already been gathered, without offering further possibility for data collection. In this paper, we demonstrate that due to errors introduced by extrapolation, standard off-policy deep reinforcement learning algorithms, such as DQN and DDPG, are incapable of learning with data uncorrelated to the distribution under the current policy, making them ineffective for this fixed batch setting. We introduce a novel class of off-policy algorithms, batch-constrained reinforcement learning, which restricts the action space in order to force the agent towards behaving close to on-policy with respect to a subset of the given data. We present the first continuous control deep reinforcement learning algorithm which can learn effectively from arbitrary, fixed batch data, and empirically demonstrate the quality of its behavior in several tasks."
                },
                {
                    "title": "Near optimal finite time identification of arbitrary linear dynamical systems",
                    "abstract": "We derive finite time error bounds for estimating general linear time-invariant (LTI) systems from a single observed trajectory using the method of least squares. We provide the first analysis of the general case when eigenvalues of the LTI system are arbitrarily distributed in three regimes: stable, marginally stable, and explosive. Our analysis yields sharp upper bounds for each of these cases separately. We observe that although the underlying process behaves quite differently in each of these three regimes, the systematic analysis of a self--normalized martingale difference term helps bound identification error up to logarithmic factors of the lower bound. On the other hand, we demonstrate that the least squares solution may be statistically inconsistent under certain conditions even when the signal-to-noise ratio is high."
                },
                {
                    "title": "Tight Analyses for Non-Smooth Stochastic Gradient Descent",
                    "abstract": "Consider the problem of minimizing functions that are Lipschitz and strongly convex, but not necessarily differentiable. We prove that after $T$ steps of stochastic gradient descent, the error of the final iterate is $O(\\log(T)/T)$ with high probability. We also construct a function from this class for which the error of the final iterate of deterministic gradient descent is $\\Omega(\\log(T)/T)$. This shows that the upper bound is tight and that, in this setting, the last iterate of stochastic gradient descent has the same general error rate (with high probability) as deterministic gradient descent. This resolves both open questions posed by Shamir (2012). \nAn intermediate step of our analysis proves that the suffix averaging method achieves error $O(1/T)$ with high probability, which is optimal (for any first-order optimization method). This improves results of Rakhlin (2012) and Hazan and Kale (2014), both of which achieved error $O(1/T)$, but only in expectation, and achieved a high probability error bound of $O(\\log \\log(T)/T)$, which is suboptimal. \nWe prove analogous results for functions that are Lipschitz and convex, but not necessarily strongly convex or differentiable. After $T$ steps of stochastic gradient descent, the error of the final iterate is $O(\\log(T)/\\sqrt{T})$ with high probability, and there exists a function for which the error of the final iterate of deterministic gradient descent is $\\Omega(\\log(T)/\\sqrt{T})$."
                },
                {
                    "title": "Model-based RL in Contextual Decision Processes: PAC bounds and Exponential Improvements over Model-free Approaches",
                    "abstract": "We study the sample complexity of model-based reinforcement learning (henceforth RL) in general contextual decision processes that require strategic exploration to find a near-optimal policy. We design new algorithms for RL with a generic model class and analyze their statistical properties. Our algorithms have sample complexity governed by a new structural parameter called the witness rank, which we show to be small in several settings of interest, including factored MDPs. We also show that the witness rank is never larger than the recently proposed Bellman rank parameter governing the sample complexity of the model-free algorithm OLIVE (Jiang et al., 2017), the only other provably sample-efficient algorithm for global exploration at this level of generality. Focusing on the special case of factored MDPs, we prove an exponential lower bound for a general class of model-free approaches, including OLIVE, which, when combined with our algorithmic results, demonstrates exponential separation between model-based and model-free RL in some rich-observation settings."
                },
                {
                    "title": "Stochastic Modified Equations and Dynamics of Stochastic Gradient Algorithms I: Mathematical Foundations",
                    "abstract": "We develop the mathematical foundations of the stochastic modified equations (SME) framework for analyzing the dynamics of stochastic gradient algorithms, where the latter is approximated by a class of stochastic differential equations with small noise parameters. We prove that this approximation can be understood mathematically as an weak approximation, which leads to a number of precise and useful results on the approximations of stochastic gradient descent (SGD), momentum SGD and stochastic Nesterov's accelerated gradient method in the general setting of stochastic objectives. We also demonstrate through explicit calculations that this continuous-time approach can uncover important analytical insights into the stochastic gradient algorithms under consideration that may not be easy to obtain in a purely discrete-time setting."
                },
                {
                    "title": "Assessing Generalization in Deep Reinforcement Learning",
                    "abstract": "Deep reinforcement learning (RL) has achieved breakthrough results on many tasks, but agents often fail to generalize beyond the environment they were trained in. As a result, deep RL algorithms that promote generalization are receiving increasing attention. However, works in this area use a wide variety of tasks and experimental setups for evaluation. The literature lacks a controlled assessment of the merits of different generalization schemes. Our aim is to catalyze community-wide progress on generalization in deep RL. To this end, we present a benchmark and experimental protocol, and conduct a systematic empirical study. Our framework contains a diverse set of environments, our methodology covers both in-distribution and out-of-distribution generalization, and our evaluation includes deep RL algorithms that specifically tackle generalization. Our key finding is that `vanilla' deep RL algorithms generalize better than specialized schemes that were proposed specifically to tackle generalization."
                },
                {
                    "title": "The Unusual Effectiveness of Averaging in GAN Training",
                    "abstract": "We examine two different techniques for parameter averaging in GAN training. Moving Average (MA) computes the time-average of parameters, whereas Exponential Moving Average (EMA) computes an exponentially discounted sum. Whilst MA is known to lead to convergence in bilinear settings, we provide the -- to our knowledge -- first theoretical arguments in support of EMA. We show that EMA converges to limit cycles around the equilibrium with vanishing amplitude as the discount parameter approaches one for simple bilinear games and also enhances the stability of general GAN training. We establish experimentally that both techniques are strikingly effective in the non-convex-concave GAN setting as well. Both improve inception and FID scores on different architectures and for different GAN objectives. We provide comprehensive experimental results across a range of datasets -- mixture of Gaussians, CIFAR-10, STL-10, CelebA and ImageNet -- to demonstrate its effectiveness. We achieve state-of-the-art results on CIFAR-10 and produce clean CelebA face images.\\footnote{~The code is available at \\url{this https URL}}"
                },
                {
                    "title": "Regret Bounds for Robust Adaptive Control of the Linear Quadratic Regulator",
                    "abstract": "We consider adaptive control of the Linear Quadratic Regulator (LQR), where an unknown linear system is controlled subject to quadratic costs. Leveraging recent developments in the estimation of linear systems and in robust controller synthesis, we present the first provably polynomial time algorithm that provides high probability guarantees of sub-linear regret on this problem. We further study the interplay between regret minimization and parameter estimation by proving a lower bound on the expected regret in terms of the exploration schedule used by any algorithm. Finally, we conduct a numerical study comparing our robust adaptive algorithm to other methods from the adaptive LQR literature, and demonstrate the flexibility of our proposed method by extending it to a demand forecasting problem subject to state constraints."
                },
                {
                    "title": "A Study on Overfitting in Deep Reinforcement Learning",
                    "abstract": "Recent years have witnessed significant progresses in deep Reinforcement Learning (RL). Empowered with large scale neural networks, carefully designed architectures, novel training algorithms and massively parallel computing devices, researchers are able to attack many challenging RL problems. However, in machine learning, more training power comes with a potential risk of more overfitting. As deep RL techniques are being applied to critical problems such as healthcare and finance, it is important to understand the generalization behaviors of the trained agents. In this paper, we conduct a systematic study of standard RL agents and find that they could overfit in various ways. Moreover, overfitting could happen \"robustly\": commonly used techniques in RL that add stochasticity do not necessarily prevent or detect overfitting. In particular, the same agents and learning algorithms could have drastically different test performance, even when all of them achieve optimal rewards during training. The observations call for more principled and careful evaluation protocols in RL. We conclude with a general discussion on overfitting in RL and a study of the generalization behaviors from the perspective of inductive bias."
                },
                {
                    "title": "Averaging Weights Leads to Wider Optima and Better Generalization",
                    "abstract": "Deep neural networks are typically trained by optimizing a loss function with an SGD variant, in conjunction with a decaying learning rate, until convergence. We show that simple averaging of multiple points along the trajectory of SGD, with a cyclical or constant learning rate, leads to better generalization than conventional training. We also show that this Stochastic Weight Averaging (SWA) procedure finds much flatter solutions than SGD, and approximates the recent Fast Geometric Ensembling (FGE) approach with a single model. Using SWA we achieve notable improvement in test accuracy over conventional SGD training on a range of state-of-the-art residual networks, PyramidNets, DenseNets, and Shake-Shake networks on CIFAR-10, CIFAR-100, and ImageNet. In short, SWA is extremely easy to implement, improves generalization, and has almost no computational overhead."
                },
                {
                    "title": "Learning Without Mixing: Towards A Sharp Analysis of Linear System Identification",
                    "abstract": "We prove that the ordinary least-squares (OLS) estimator attains nearly minimax optimal performance for the identification of linear dynamical systems from a single observed trajectory. Our upper bound relies on a generalization of Mendelson's small-ball method to dependent data, eschewing the use of standard mixing-time arguments. Our lower bounds reveal that these upper bounds match up to logarithmic factors. In particular, we capture the correct signal-to-noise behavior of the problem, showing that more unstable linear systems are easier to estimate. This behavior is qualitatively different from arguments which rely on mixing-time calculations that suggest that unstable systems are more difficult to estimate. We generalize our technique to provide bounds for a more general class of linear response time-series."
                },
                {
                    "title": "Global Convergence of Policy Gradient Methods for the Linear Quadratic Regulator",
                    "abstract": "Direct policy gradient methods for reinforcement learning and continuous control problems are a popular approach for a variety of reasons: 1) they are easy to implement without explicit knowledge of the underlying model 2) they are an \"end-to-end\" approach, directly optimizing the performance metric of interest 3) they inherently allow for richly parameterized policies. A notable drawback is that even in the most basic continuous control problem (that of linear quadratic regulators), these methods must solve a non-convex optimization problem, where little is understood about their efficiency from both computational and statistical perspectives. In contrast, system identification and model based planning in optimal control theory have a much more solid theoretical footing, where much is known with regards to their computational and statistical properties. This work bridges this gap showing that (model free) policy gradient methods globally converge to the optimal solution and are efficient (polynomially so in relevant problem dependent quantities) with regards to their sample and computational complexities."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "Visualizing the Loss Landscape of Neural Nets",
                    "abstract": "Neural network training relies on our ability to find \"good\" minimizers of highly non-convex loss functions. It is well-known that certain network architecture designs (e.g., skip connections) produce loss functions that train easier, and well-chosen training parameters (batch size, learning rate, optimizer) produce minimizers that generalize better. However, the reasons for these differences, and their effects on the underlying loss landscape, are not well understood. In this paper, we explore the structure of neural loss functions, and the effect of loss landscapes on generalization, using a range of visualization methods. First, we introduce a simple \"filter normalization\" method that helps us visualize loss function curvature and make meaningful side-by-side comparisons between loss functions. Then, using a variety of visualizations, we explore how network architecture affects the loss landscape, and how training parameters affect the shape of minimizers."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Don't Decay the Learning Rate, Increase the Batch Size",
                    "abstract": "It is common practice to decay the learning rate. Here we show one can usually obtain the same learning curve on both training and test sets by instead increasing the batch size during training. This procedure is successful for stochastic gradient descent (SGD), SGD with momentum, Nesterov momentum, and Adam. It reaches equivalent test accuracies after the same number of training epochs, but with fewer parameter updates, leading to greater parallelism and shorter training times. We can further reduce the number of parameter updates by increasing the learning rate $\\epsilon$ and scaling the batch size $B \\propto \\epsilon$. Finally, one can increase the momentum coefficient $m$ and scale $B \\propto 1/(1-m)$, although this tends to slightly reduce the test accuracy. Crucially, our techniques allow us to repurpose existing training schedules for large batch training with no hyper-parameter tuning. We train ResNet-50 on ImageNet to $76.1\\%$ validation accuracy in under 30 minutes."
                },
                {
                    "title": "End-to-End Driving Via Conditional Imitation Learning",
                    "abstract": "Deep networks trained on demonstrations of human driving have learned to follow roads and avoid obstacles. However, driving policies trained via imitation learning cannot be controlled at test time. A vehicle trained end-to-end to imitate an expert cannot be guided to take a specific turn at an upcoming intersection. This limits the utility of such systems. We propose to condition imitation learning on high-level command input. At test time, the learned driving policy functions as a chauffeur that handles sensorimotor coordination but continues to respond to navigational commands. We evaluate different architectures for conditional imitation learning in vision-based driving. We conduct experiments in realistic three-dimensional simulations of urban driving and on a 1/5 scale robotic truck that is trained to drive in a residential area. Both systems drive based on visual input yet remain responsive to high-level navigational commands."
                },
                {
                    "title": "Deep Reinforcement Learning that Matters",
                    "abstract": "\n \n In recent years, significant progress has been made in solving challenging problems across various domains using deep reinforcement learning (RL). Reproducing existing work and accurately judging the improvements offered by novel methods is vital to sustaining this progress. Unfortunately, reproducing results for state-of-the-art deep RL methods is seldom straightforward. In particular, non-determinism in standard benchmark environments, combined with variance intrinsic to the methods, can make reported results tough to interpret. Without significance metrics and tighter standardization of experimental reporting, it is difficult to determine whether improvements over the prior state-of-the-art are meaningful. In this paper, we investigate challenges posed by reproducibility, proper experimental techniques, and reporting procedures. We illustrate the variability in reported metrics and results when comparing against common baselines and suggest guidelines to make future results in deep RL more reproducible. We aim to spur discussion about how to ensure continued progress in the field by minimizing wasted effort stemming from results that are non-reproducible and easily misinterpreted.\n \n"
                },
                {
                    "title": "Revisiting the Arcade Learning Environment: Evaluation Protocols and Open Problems for General Agents",
                    "abstract": "The Arcade Learning Environment (ALE) is an evaluation platform that poses the challenge of building AI agents with general competency across dozens of Atari 2600 games. It supports a variety of different problem settings and it has been receiving increasing attention from the scientific community, leading to some high-profile success stories such as the much publicized Deep Q-Networks (DQN). In this article we take a big picture look at how the ALE is being used by the research community. We show how diverse the evaluation methodologies in the ALE have become with time, and highlight some key concerns when evaluating agents in the ALE. We use this discussion to present some methodological best practices and provide new benchmark results using these best practices. To further the progress in the field, we introduce a new version of the ALE that supports multiple game modes and provides a form of stochasticity we call sticky actions. We conclude this big picture look by revisiting challenges posed when the ALE was introduced, summarizing the state-of-the-art in various problems and highlighting problems that remain open."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour",
                    "abstract": "Deep learning thrives with large neural networks and large datasets. However, larger networks and larger datasets result in longer training times that impede research and development progress. Distributed synchronous SGD offers a potential solution to this problem by dividing SGD minibatches over a pool of parallel workers. Yet to make this scheme efficient, the per-worker workload must be large, which implies nontrivial growth in the SGD minibatch size. In this paper, we empirically show that on the ImageNet dataset large minibatches cause optimization difficulties, but when these are addressed the trained networks exhibit good generalization. Specifically, we show no loss of accuracy when training with large minibatch sizes up to 8192 images. To achieve this result, we adopt a hyper-parameter-free linear scaling rule for adjusting learning rates as a function of minibatch size and develop a new warmup scheme that overcomes optimization challenges early in training. With these simple techniques, our Caffe2-based system trains ResNet-50 with a minibatch size of 8192 on 256 GPUs in one hour, while matching small minibatch accuracy. Using commodity hardware, our implementation achieves ~90% scaling efficiency when moving from 8 to 256 GPUs. Our findings enable training visual recognition models on internet-scale data with high efficiency."
                },
                {
                    "title": "Convergence Analysis of Two-layer Neural Networks with ReLU Activation",
                    "abstract": "In recent years, stochastic gradient descent (SGD) based techniques has become the standard tools for training neural networks. However, formal theoretical understanding of why SGD can train neural networks in practice is largely missing. \nIn this paper, we make progress on understanding this mystery by providing a convergence analysis for SGD on a rich subset of two-layer feedforward networks with ReLU activations. This subset is characterized by a special structure called \"identity mapping\". We prove that, if input follows from Gaussian distribution, with standard $O(1/\\sqrt{d})$ initialization of the weights, SGD converges to the global minimum in polynomial number of steps. Unlike normal vanilla networks, the \"identity mapping\" makes our network asymmetric and thus the global minimum is unique. To complement our theory, we are also able to show experimentally that multi-layer networks with this mapping have better performance compared with normal vanilla networks. \nOur convergence theorem differs from traditional non-convex optimization techniques. We show that SGD converges to optimal in \"two phases\": In phase I, the gradient points to the wrong direction, however, a potential function $g$ gradually decreases. Then in phase II, SGD enters a nice one point convex region and converges. We also show that the identity mapping is necessary for convergence, as it moves the initial point to a better place for optimization. Experiment verifies our claims."
                },
                {
                    "title": "Stochastic Gradient Descent as Approximate Bayesian Inference",
                    "abstract": "Stochastic Gradient Descent with a constant learning rate (constant SGD) simulates a Markov chain with a stationary distribution. With this perspective, we derive several new results. (1) We show that constant SGD can be used as an approximate Bayesian posterior inference algorithm. Specifically, we show how to adjust the tuning parameters of constant SGD to best match the stationary distribution to a posterior, minimizing the Kullback-Leibler divergence between these two distributions. (2) We demonstrate that constant SGD gives rise to a new variational EM algorithm that optimizes hyperparameters in complex probabilistic models. (3) We also propose SGD with momentum for sampling and show how to adjust the damping coefficient accordingly. (4) We analyze MCMC algorithms. For Langevin Dynamics and Stochastic Gradient Fisher Scoring, we quantify the approximation errors due to finite learning rates. Finally (5), we use the stochastic process perspective to give a short proof of why Polyak averaging is optimal. Based on this idea, we propose a scalable approximate MCMC algorithm, the Averaged Stochastic Gradient Sampler."
                },
                {
                    "title": "DART: Noise Injection for Robust Imitation Learning",
                    "abstract": "One approach to Imitation Learning is Behavior Cloning, in which a robot observes a supervisor and infers a control policy. A known problem with this \"off-policy\" approach is that the robot's errors compound when drifting away from the supervisor's demonstrations. On-policy, techniques alleviate this by iteratively collecting corrective actions for the current robot policy. However, these techniques can be tedious for human supervisors, add significant computation burden, and may visit dangerous states during training. We propose an off-policy approach that injects noise into the supervisor's policy while demonstrating. This forces the supervisor to demonstrate how to recover from errors. We propose a new algorithm, DART (Disturbances for Augmenting Robot Trajectories), that collects demonstrations with injected noise, and optimizes the noise level to approximate the error of the robot's trained policy during data collection. We compare DART with DAgger and Behavior Cloning in two domains: in simulation with an algorithmic supervisor on the MuJoCo tasks (Walker, Humanoid, Hopper, Half-Cheetah) and in physical experiments with human supervisors training a Toyota HSR robot to perform grasping in clutter. For high dimensional tasks like Humanoid, DART can be up to $3x$ faster in computation time and only decreases the supervisor's cumulative reward by $5\\%$ during training, whereas DAgger executes policies that have $80\\%$ less cumulative reward than the supervisor. On the grasping in clutter task, DART obtains on average a $62\\%$ performance increase over Behavior Cloning."
                },
                {
                    "title": "How to Escape Saddle Points Efficiently",
                    "abstract": "This paper shows that a perturbed form of gradient descent converges to a second-order stationary point in a number iterations which depends only poly-logarithmically on dimension (i.e., it is almost \"dimension-free\"). The convergence rate of this procedure matches the well-known convergence rate of gradient descent to first-order stationary points, up to log factors. When all saddle points are non-degenerate, all second-order stationary points are local minima, and our result thus shows that perturbed gradient descent can escape saddle points almost for free. Our results can be directly applied to many machine learning applications, including deep learning. As a particular concrete example of such an application, we show that our results can be used directly to establish sharp global convergence rates for matrix factorization. Our results rely on a novel characterization of the geometry around saddle points, which may be of independent interest to the non-convex optimization community."
                },
                {
                    "title": "Overcoming catastrophic forgetting in neural networks",
                    "abstract": "Significance Deep neural networks are currently the most successful machine-learning technique for solving a variety of tasks, including language translation, image classification, and image generation. One weakness of such models is that, unlike humans, they are unable to learn multiple tasks sequentially. In this work we propose a practical solution to train such models sequentially by protecting the weights important for previous tasks. This approach, inspired by synaptic consolidation in neuroscience, enables state of the art results on multiple reinforcement learning problems experienced sequentially. The ability to learn tasks in a sequential fashion is crucial to the development of artificial intelligence. Until now neural networks have not been capable of this and it has been widely thought that catastrophic forgetting is an inevitable feature of connectionist models. We show that it is possible to overcome this limitation and train networks that can maintain expertise on tasks that they have not experienced for a long time. Our approach remembers old tasks by selectively slowing down learning on the weights important for those tasks. We demonstrate our approach is scalable and effective by solving a set of classification tasks based on a hand-written digit dataset and by learning several Atari 2600 games sequentially."
                },
                {
                    "title": "Contextual Decision Processes with low Bellman rank are PAC-Learnable",
                    "abstract": "This paper studies systematic exploration for reinforcement learning (RL) with rich observations and function approximation. We introduce contextual decision processes (CDPs), that unify most prior RL settings. Our first contribution is a complexity measure, the Bellman rank, that we show enables tractable learning of near-optimal behavior in CDPs and is naturally small for many well-studied RL models. Our second contribution is a new RL algorithm that does systematic exploration to learn near-optimal behavior in CDPs with low Bellman rank. The algorithm requires a number of samples that is polynomial in all relevant parameters but independent of the number of unique contexts. Our approach uses Bellman error minimization with optimistic exploration and provides new insights into efficient exploration for RL with function approximation."
                }
            ],
            "categories": [
                "cs.LG",
                "math.OC",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively address the challenges of online nonstochastic control in machine learning to improve decision-making processes in dynamic environments?\n\n### [Question 2] - Why is it interesting and important?\nSolving the problem of online nonstochastic control is crucial for advancing the field of machine learning, particularly in applications where real-time decision-making is essential, such as robotics, autonomous systems, and adaptive control systems. By addressing this question, we can enhance the understanding of dynamic environments and develop more robust algorithms that can adapt to changing conditions. This research could lead to significant improvements in the efficiency and effectiveness of machine learning models, paving the way for practical applications in various industries, including healthcare, finance, and transportation.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving online nonstochastic control stem from the inherent complexities of dynamic environments, where the system's state can change unpredictably. Naive approaches may fail due to their inability to account for the nonstationarity of the environment, leading to suboptimal decision-making. Additionally, technical obstacles such as the need for real-time processing, the requirement for accurate modeling of the environment, and the difficulty in balancing exploration and exploitation complicate the development of effective solutions. Theoretical challenges, such as ensuring convergence and stability in the presence of noise and uncertainty, further add to the difficulty of this problem.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on either stochastic control or simplified models that do not capture the complexities of real-world scenarios. Limitations in computational resources and the lack of robust algorithms capable of handling the intricacies of online nonstochastic control have hindered progress. Additionally, existing solutions may not adequately address the trade-offs between exploration and exploitation in dynamic settings. Our approach aims to bridge these gaps by integrating advanced techniques from reinforcement learning and control theory, providing a more comprehensive framework for tackling the challenges of online nonstochastic control.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves developing a novel algorithm that combines reinforcement learning techniques with adaptive control strategies. We will utilize a diverse set of datasets that simulate dynamic environments to evaluate our approach. The performance will be measured using metrics such as cumulative reward, convergence rate, and robustness to environmental changes. We expect our results to demonstrate improved decision-making capabilities in dynamic settings, showcasing the effectiveness of our approach in addressing the challenges of online nonstochastic control."
            }
        },
        "author_data": {
            "1691f328-3733-4a69-8f0e-a6e75b41b9a5": {
                "pk": "1691f328-3733-4a69-8f0e-a6e75b41b9a5",
                "name": "Adam Block",
                "collaborators": [
                    "Dylan J. Foster",
                    "Dipendra Kumar Misra",
                    "A. Rakhlin",
                    "Max Simchowitz",
                    "Gergely Neu",
                    "Lorenzo Rosasco"
                ],
                "domain": [
                    "Imitation Learning",
                    "Reinforcement Learning",
                    "Online Learning",
                    "Robotics"
                ],
                "publications": [
                    {
                        "title": "Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning",
                        "abstract": "Imitation learning (IL) aims to mimic the behavior of an expert in a sequential decision making task by learning from demonstrations, and has been widely applied to robotics, autonomous driving, and autoregressive text generation. The simplest approach to IL, behavior cloning (BC), is thought to incur sample complexity with unfavorable quadratic dependence on the problem horizon, motivating a variety of different online algorithms that attain improved linear horizon dependence under stronger assumptions on the data and the learner's access to the expert. We revisit the apparent gap between offline and online IL from a learning-theoretic perspective, with a focus on general policy classes up to and including deep neural networks. Through a new analysis of behavior cloning with the logarithmic loss, we show that it is possible to achieve horizon-independent sample complexity in offline IL whenever (i) the range of the cumulative payoffs is controlled, and (ii) an appropriate notion of supervised learning complexity for the policy class is controlled. Specializing our results to deterministic, stationary policies, we show that the gap between offline and online IL is not fundamental: (i) it is possible to achieve linear dependence on horizon in offline IL under dense rewards (matching what was previously only known to be achievable in online IL); and (ii) without further assumptions on the policy class, online IL cannot improve over offline IL with the logarithmic loss, even in benign MDPs. We complement our theoretical results with experiments on standard RL tasks and autoregressive language generation to validate the practical relevance of our findings."
                    },
                    {
                        "title": "Oracle-Ef\ufb01cient Smoothed Online Learning for Piecewise Continuous Decision Making",
                        "abstract": "Smoothed online learning has emerged as a popular framework to mitigate the substantial loss in statistical and computational complexity that arises when one moves from classical to adversarial learning. Unfortunately, for some spaces, it has been shown that ef\ufb01cient algorithms suffer an exponentially worse regret than that which is minimax optimal, even when the learner has access to an optimization oracle over the space. To mitigate that exponential dependence, this work introduces a new notion of complexity, the generalized bracketing numbers, which marries constraints on the adversary to the size of the space, and shows that an instantiation of Follow-the-Perturbed-Leader can attain low regret with the number of calls to the optimization oracle scaling optimally with respect to average regret. We then instantiate our bounds in several problems of interest, including online prediction and planning of piecewise continuous functions, which has many applications in \ufb01elds as diverse as econometrics and robotics."
                    }
                ]
            },
            "d2bad9f6-64eb-4aee-babb-29d9683f510e": {
                "pk": "d2bad9f6-64eb-4aee-babb-29d9683f510e",
                "name": "Dylan J. Foster",
                "collaborators": [
                    "Akshay Krishnamurthy",
                    "Tengyang Xie",
                    "P. Amortila",
                    "Nan Jiang",
                    "Yuda Song",
                    "Lili Wu",
                    "Audrey Huang",
                    "Wenhao Zhan",
                    "Jason D. Lee",
                    "Wen Sun",
                    "Zakaria Mhammedi",
                    "Ayush Sekhari",
                    "Keegan Harris",
                    "Cyril Zhang",
                    "Aleksandrs Slivkins",
                    "Corby Rosset",
                    "Ahmed Awadallah",
                    "A. Rakhlin"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Representation Learning",
                    "Language Model Alignment",
                    "Exploration Strategies"
                ],
                "publications": [
                    {
                        "title": "Rich-Observation Reinforcement Learning with Continuous Latent Dynamics",
                        "abstract": "Sample-efficiency and reliability remain major bottlenecks toward wide adoption of reinforcement learning algorithms in continuous settings with high-dimensional perceptual inputs. Toward addressing these challenges, we introduce a new theoretical framework, RichCLD (Rich-Observation RL with Continuous Latent Dynamics), in which the agent performs control based on high-dimensional observations, but the environment is governed by low-dimensional latent states and Lipschitz continuous dynamics. Our main contribution is a new algorithm for this setting that is provably statistically and computationally efficient. The core of our algorithm is a new representation learning objective; we show that prior representation learning schemes tailored to discrete dynamics do not naturally extend to the continuous setting. Our new objective is amenable to practical implementation, and empirically, we find that it compares favorably to prior schemes in a standard evaluation protocol. We further provide several insights into the statistical complexity of the RichCLD framework, in particular proving that certain notions of Lipschitzness that admit sample-efficient learning in the absence of rich observations are insufficient in the rich-observation setting."
                    },
                    {
                        "title": "Correcting the Mythos of KL-Regularization: Direct Alignment without Overoptimization via Chi-Squared Preference Optimization",
                        "abstract": "Language model alignment methods, such as reinforcement learning from human feedback (RLHF), have led to impressive advances in language model capabilities, but existing techniques are limited by a widely observed phenomenon known as overoptimization, where the quality of the language model plateaus or degrades over the course of the alignment process. Overoptimization is often attributed to overfitting to an inaccurate reward model, and while it can be mitigated through online data collection, this is infeasible in many settings. This raises a fundamental question: Do existing offline alignment algorithms make the most of the data they have, or can their sample-efficiency be improved further? We address this question with a new algorithm for offline alignment, $\\chi^2$-Preference Optimization ($\\chi$PO). $\\chi$PO is a one-line change to Direct Preference Optimization (DPO; Rafailov et al., 2023), which only involves modifying the logarithmic link function in the DPO objective. Despite this minimal change, $\\chi$PO implicitly implements the principle of pessimism in the face of uncertainty via regularization with the $\\chi^2$-divergence -- which quantifies uncertainty more effectively than KL-regularization -- and provably alleviates overoptimization, achieving sample-complexity guarantees based on single-policy concentrability -- the gold standard in offline reinforcement learning. $\\chi$PO's simplicity and strong guarantees make it the first practical and general-purpose offline alignment algorithm that is provably robust to overoptimization."
                    },
                    {
                        "title": "Reinforcement Learning under Latent Dynamics: Toward Statistical and Algorithmic Modularity",
                        "abstract": "Real-world applications of reinforcement learning often involve environments where agents operate on complex, high-dimensional observations, but the underlying (''latent'') dynamics are comparatively simple. However, outside of restrictive settings such as small latent spaces, the fundamental statistical requirements and algorithmic principles for reinforcement learning under latent dynamics are poorly understood. This paper addresses the question of reinforcement learning under $\\textit{general}$ latent dynamics from a statistical and algorithmic perspective. On the statistical side, our main negative result shows that most well-studied settings for reinforcement learning with function approximation become intractable when composed with rich observations; we complement this with a positive result, identifying latent pushforward coverability as a general condition that enables statistical tractability. Algorithmically, we develop provably efficient observable-to-latent reductions -- that is, reductions that transform an arbitrary algorithm for the latent MDP into an algorithm that can operate on rich observations -- in two settings: one where the agent has access to hindsight observations of the latent dynamics [LADZ23], and one where the agent can estimate self-predictive latent models [SAGHCB20]. Together, our results serve as a first step toward a unified statistical and algorithmic theory for reinforcement learning under latent dynamics."
                    },
                    {
                        "title": "Harnessing Density Ratios for Online Reinforcement Learning",
                        "abstract": "The theories of offline and online reinforcement learning, despite having evolved in parallel, have begun to show signs of the possibility for a unification, with algorithms and analysis techniques for one setting often having natural counterparts in the other. However, the notion of density ratio modeling, an emerging paradigm in offline RL, has been largely absent from online RL, perhaps for good reason: the very existence and boundedness of density ratios relies on access to an exploratory dataset with good coverage, but the core challenge in online RL is to collect such a dataset without having one to start. In this work we show -- perhaps surprisingly -- that density ratio-based algorithms have online counterparts. Assuming only the existence of an exploratory distribution with good coverage, a structural condition known as coverability (Xie et al., 2023), we give a new algorithm (GLOW) that uses density ratio realizability and value function realizability to perform sample-efficient online exploration. GLOW addresses unbounded density ratios via careful use of truncation, and combines this with optimism to guide exploration. GLOW is computationally inefficient; we complement it with a more efficient counterpart, HyGLOW, for the Hybrid RL setting (Song et al., 2022) wherein online RL is augmented with additional offline data. HyGLOW is derived as a special case of a more general meta-algorithm that provides a provable black-box reduction from hybrid RL to offline RL, which may be of independent interest."
                    },
                    {
                        "title": "Can large language models explore in-context?",
                        "abstract": "We investigate the extent to which contemporary Large Language Models (LLMs) can engage in exploration, a core capability in reinforcement learning and decision making. We focus on native performance of existing LLMs, without training interventions. We deploy LLMs as agents in simple multi-armed bandit environments, specifying the environment description and interaction history entirely in-context, i.e., within the LLM prompt. We experiment with GPT-3.5, GPT-4, and Llama2, using a variety of prompt designs, and find that the models do not robustly engage in exploration without substantial interventions: i) Across all of our experiments, only one configuration resulted in satisfactory exploratory behavior: GPT-4 with chain-of-thought reasoning and an externally summarized interaction history, presented as sufficient statistics; ii) All other configurations did not result in robust exploratory behavior, including those with chain-of-thought reasoning but unsummarized history. Although these findings can be interpreted positively, they suggest that external summarization -- which may not be possible in more complex settings -- is important for obtaining desirable behavior from LLM agents. We conclude that non-trivial algorithmic interventions, such as fine-tuning or dataset curation, may be required to empower LLM-based decision making agents in complex settings."
                    },
                    {
                        "title": "Scalable Online Exploration via Coverability",
                        "abstract": "Exploration is a major challenge in reinforcement learning, especially for high-dimensional domains that require function approximation. We propose exploration objectives -- policy optimization objectives that enable downstream maximization of any reward function -- as a conceptual framework to systematize the study of exploration. Within this framework, we introduce a new objective, $L_1$-Coverage, which generalizes previous exploration schemes and supports three fundamental desiderata: 1. Intrinsic complexity control. $L_1$-Coverage is associated with a structural parameter, $L_1$-Coverability, which reflects the intrinsic statistical difficulty of the underlying MDP, subsuming Block and Low-Rank MDPs. 2. Efficient planning. For a known MDP, optimizing $L_1$-Coverage efficiently reduces to standard policy optimization, allowing flexible integration with off-the-shelf methods such as policy gradient and Q-learning approaches. 3. Efficient exploration. $L_1$-Coverage enables the first computationally efficient model-based and model-free algorithms for online (reward-free or reward-driven) reinforcement learning in MDPs with low coverability. Empirically, we find that $L_1$-Coverage effectively drives off-the-shelf policy optimization algorithms to explore the state space."
                    },
                    {
                        "title": "Exploratory Preference Optimization: Harnessing Implicit Q*-Approximation for Sample-Efficient RLHF",
                        "abstract": "Reinforcement learning from human feedback (RLHF) has emerged as a central tool for language model alignment. We consider online exploration in RLHF, which exploits interactive access to human or AI feedback by deliberately encouraging the model to produce diverse, maximally informative responses. By allowing RLHF to confidently stray from the pre-trained model, online exploration offers the possibility of novel, potentially super-human capabilities, but its full potential as a paradigm for language model training has yet to be realized, owing to computational and statistical bottlenecks in directly adapting existing reinforcement learning techniques. We propose a new algorithm for online exploration in RLHF, Exploratory Preference Optimization (XPO), which is simple and practical -- a one-line change to (online) Direct Preference Optimization (DPO; Rafailov et al., 2023) -- yet enjoys the strongest known provable guarantees and promising empirical performance. XPO augments the DPO objective with a novel and principled exploration bonus, empowering the algorithm to explore outside the support of the initial model and human feedback data. In theory, we show that XPO is provably sample-efficient and converges to a near-optimal language model policy under natural exploration conditions, irrespective of whether the initial model has good coverage. Our analysis, which builds on the observation that DPO implicitly performs a form of $Q^{\\star}$-approximation (or, Bellman error minimization), combines previously disparate techniques from language modeling and theoretical reinforcement learning in a serendipitous fashion through the perspective of KL-regularized Markov decision processes. Empirically, we find that XPO is more sample-efficient than non-exploratory DPO variants in a preliminary evaluation."
                    }
                ]
            },
            "691207c8-ebc0-4c2f-9201-9fc3b3c36271": {
                "pk": "691207c8-ebc0-4c2f-9201-9fc3b3c36271",
                "name": "Akshay Krishnamurthy",
                "collaborators": [
                    "Dylan J. Foster",
                    "P. Amortila",
                    "Tongyi Cao",
                    "Yuda Song",
                    "Lili Wu",
                    "Nan Jiang",
                    "Zakaria Mhammedi",
                    "Keegan Harris",
                    "Cyril Zhang",
                    "Aleksandrs Slivkins",
                    "Tengyang Xie",
                    "Corby Rosset",
                    "Ahmed Awadallah",
                    "A. Rakhlin",
                    "Max Simchowitz",
                    "Anurag Ajay",
                    "Pulkit Agrawal"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Distribution Shift",
                    "Machine Learning",
                    "Robust Optimization"
                ],
                "publications": [
                    {
                        "title": "Mitigating Covariate Shift in Misspecified Regression with Applications to Reinforcement Learning",
                        "abstract": "A pervasive phenomenon in machine learning applications is distribution shift, where training and deployment conditions for a machine learning model differ. As distribution shift typically results in a degradation in performance, much attention has been devoted to algorithmic interventions that mitigate these detrimental effects. In this paper, we study the effect of distribution shift in the presence of model misspecification, specifically focusing on $L_{\\infty}$-misspecified regression and adversarial covariate shift, where the regression target remains fixed while the covariate distribution changes arbitrarily. We show that empirical risk minimization, or standard least squares regression, can result in undesirable misspecification amplification where the error due to misspecification is amplified by the density ratio between the training and testing distributions. As our main result, we develop a new algorithm -- inspired by robust optimization techniques -- that avoids this undesirable behavior, resulting in no misspecification amplification while still obtaining optimal statistical rates. As applications, we use this regression procedure to obtain new guarantees in offline and online reinforcement learning with misspecification and establish new separations between previously studied structural conditions and notions of coverage."
                    },
                    {
                        "title": "Rich-Observation Reinforcement Learning with Continuous Latent Dynamics",
                        "abstract": "Sample-efficiency and reliability remain major bottlenecks toward wide adoption of reinforcement learning algorithms in continuous settings with high-dimensional perceptual inputs. Toward addressing these challenges, we introduce a new theoretical framework, RichCLD (Rich-Observation RL with Continuous Latent Dynamics), in which the agent performs control based on high-dimensional observations, but the environment is governed by low-dimensional latent states and Lipschitz continuous dynamics. Our main contribution is a new algorithm for this setting that is provably statistically and computationally efficient. The core of our algorithm is a new representation learning objective; we show that prior representation learning schemes tailored to discrete dynamics do not naturally extend to the continuous setting. Our new objective is amenable to practical implementation, and empirically, we find that it compares favorably to prior schemes in a standard evaluation protocol. We further provide several insights into the statistical complexity of the RichCLD framework, in particular proving that certain notions of Lipschitzness that admit sample-efficient learning in the absence of rich observations are insufficient in the rich-observation setting."
                    },
                    {
                        "title": "Reinforcement Learning under Latent Dynamics: Toward Statistical and Algorithmic Modularity",
                        "abstract": "Real-world applications of reinforcement learning often involve environments where agents operate on complex, high-dimensional observations, but the underlying (''latent'') dynamics are comparatively simple. However, outside of restrictive settings such as small latent spaces, the fundamental statistical requirements and algorithmic principles for reinforcement learning under latent dynamics are poorly understood. This paper addresses the question of reinforcement learning under $\\textit{general}$ latent dynamics from a statistical and algorithmic perspective. On the statistical side, our main negative result shows that most well-studied settings for reinforcement learning with function approximation become intractable when composed with rich observations; we complement this with a positive result, identifying latent pushforward coverability as a general condition that enables statistical tractability. Algorithmically, we develop provably efficient observable-to-latent reductions -- that is, reductions that transform an arbitrary algorithm for the latent MDP into an algorithm that can operate on rich observations -- in two settings: one where the agent has access to hindsight observations of the latent dynamics [LADZ23], and one where the agent can estimate self-predictive latent models [SAGHCB20]. Together, our results serve as a first step toward a unified statistical and algorithmic theory for reinforcement learning under latent dynamics."
                    },
                    {
                        "title": "Can large language models explore in-context?",
                        "abstract": "We investigate the extent to which contemporary Large Language Models (LLMs) can engage in exploration, a core capability in reinforcement learning and decision making. We focus on native performance of existing LLMs, without training interventions. We deploy LLMs as agents in simple multi-armed bandit environments, specifying the environment description and interaction history entirely in-context, i.e., within the LLM prompt. We experiment with GPT-3.5, GPT-4, and Llama2, using a variety of prompt designs, and find that the models do not robustly engage in exploration without substantial interventions: i) Across all of our experiments, only one configuration resulted in satisfactory exploratory behavior: GPT-4 with chain-of-thought reasoning and an externally summarized interaction history, presented as sufficient statistics; ii) All other configurations did not result in robust exploratory behavior, including those with chain-of-thought reasoning but unsummarized history. Although these findings can be interpreted positively, they suggest that external summarization -- which may not be possible in more complex settings -- is important for obtaining desirable behavior from LLM agents. We conclude that non-trivial algorithmic interventions, such as fine-tuning or dataset curation, may be required to empower LLM-based decision making agents in complex settings."
                    },
                    {
                        "title": "Scalable Online Exploration via Coverability",
                        "abstract": "Exploration is a major challenge in reinforcement learning, especially for high-dimensional domains that require function approximation. We propose exploration objectives -- policy optimization objectives that enable downstream maximization of any reward function -- as a conceptual framework to systematize the study of exploration. Within this framework, we introduce a new objective, $L_1$-Coverage, which generalizes previous exploration schemes and supports three fundamental desiderata: 1. Intrinsic complexity control. $L_1$-Coverage is associated with a structural parameter, $L_1$-Coverability, which reflects the intrinsic statistical difficulty of the underlying MDP, subsuming Block and Low-Rank MDPs. 2. Efficient planning. For a known MDP, optimizing $L_1$-Coverage efficiently reduces to standard policy optimization, allowing flexible integration with off-the-shelf methods such as policy gradient and Q-learning approaches. 3. Efficient exploration. $L_1$-Coverage enables the first computationally efficient model-based and model-free algorithms for online (reward-free or reward-driven) reinforcement learning in MDPs with low coverability. Empirically, we find that $L_1$-Coverage effectively drives off-the-shelf policy optimization algorithms to explore the state space."
                    },
                    {
                        "title": "Exploratory Preference Optimization: Harnessing Implicit Q*-Approximation for Sample-Efficient RLHF",
                        "abstract": "Reinforcement learning from human feedback (RLHF) has emerged as a central tool for language model alignment. We consider online exploration in RLHF, which exploits interactive access to human or AI feedback by deliberately encouraging the model to produce diverse, maximally informative responses. By allowing RLHF to confidently stray from the pre-trained model, online exploration offers the possibility of novel, potentially super-human capabilities, but its full potential as a paradigm for language model training has yet to be realized, owing to computational and statistical bottlenecks in directly adapting existing reinforcement learning techniques. We propose a new algorithm for online exploration in RLHF, Exploratory Preference Optimization (XPO), which is simple and practical -- a one-line change to (online) Direct Preference Optimization (DPO; Rafailov et al., 2023) -- yet enjoys the strongest known provable guarantees and promising empirical performance. XPO augments the DPO objective with a novel and principled exploration bonus, empowering the algorithm to explore outside the support of the initial model and human feedback data. In theory, we show that XPO is provably sample-efficient and converges to a near-optimal language model policy under natural exploration conditions, irrespective of whether the initial model has good coverage. Our analysis, which builds on the observation that DPO implicitly performs a form of $Q^{\\star}$-approximation (or, Bellman error minimization), combines previously disparate techniques from language modeling and theoretical reinforcement learning in a serendipitous fashion through the perspective of KL-regularized Markov decision processes. Empirically, we find that XPO is more sample-efficient than non-exploratory DPO variants in a preliminary evaluation."
                    },
                    {
                        "title": "Statistical Learning under Heterogenous Distribution Shift",
                        "abstract": "This paper studies the prediction of a target z from a pair of random variables (x,y), where the ground-truth predictor is additive E[z | x,y] = f?(x)+g?(y). We study the performance of empirical risk minimization (ERM) over functions f + g, f \u2208 F and g \u2208 G, fit on a given training distribution, but evaluated on a test distribution which exhibits covariate shift. We show that, when the class F is \u201csimpler\" than G (measured, e.g., in terms of its metric entropy), our predictor is more resilient to heterogenous covariate shifts in which the shift in x is much greater than that in y. These results rely on a novel H\u00f6lder style inequality for the Dudley integral which may be of independent interest. Moreover, we corroborate our theoretical findings with experiments demonstrating improved resilience to shifts in \u201csimpler\u201d features across numerous domains."
                    }
                ]
            },
            "fc4e803a-d06f-4c9c-b51f-d5884ad57c38": {
                "pk": "fc4e803a-d06f-4c9c-b51f-d5884ad57c38",
                "name": "Max Simchowitz",
                "collaborators": [
                    "Anurag Ajay",
                    "Pulkit Agrawal",
                    "Akshay Krishnamurthy",
                    "Chuning Zhu",
                    "Siri Gadipudi",
                    "Abhishek Gupta",
                    "Adam Block",
                    "A. Rakhlin",
                    "Gergely Neu",
                    "Lorenzo Rosasco"
                ],
                "domain": [
                    "Machine Learning",
                    "Reinforcement Learning",
                    "Online Learning",
                    "Statistical Learning"
                ],
                "publications": [
                    {
                        "title": "Statistical Learning under Heterogenous Distribution Shift",
                        "abstract": "This paper studies the prediction of a target z from a pair of random variables (x,y), where the ground-truth predictor is additive E[z | x,y] = f?(x)+g?(y). We study the performance of empirical risk minimization (ERM) over functions f + g, f \u2208 F and g \u2208 G, fit on a given training distribution, but evaluated on a test distribution which exhibits covariate shift. We show that, when the class F is \u201csimpler\" than G (measured, e.g., in terms of its metric entropy), our predictor is more resilient to heterogenous covariate shifts in which the shift in x is much greater than that in y. These results rely on a novel H\u00f6lder style inequality for the Dudley integral which may be of independent interest. Moreover, we corroborate our theoretical findings with experiments demonstrating improved resilience to shifts in \u201csimpler\u201d features across numerous domains."
                    },
                    {
                        "title": "RePo: Resilient Model-Based Reinforcement Learning by Regularizing Posterior Predictability",
                        "abstract": "Visual model-based RL methods typically encode image observations into low-dimensional representations in a manner that does not eliminate redundant information. This leaves them susceptible to spurious variations -- changes in task-irrelevant components such as background distractors or lighting conditions. In this paper, we propose a visual model-based RL method that learns a latent representation resilient to such spurious variations. Our training objective encourages the representation to be maximally predictive of dynamics and reward, while constraining the information flow from the observation to the latent representation. We demonstrate that this objective significantly bolsters the resilience of visual model-based RL methods to visual distractors, allowing them to operate in dynamic environments. We then show that while the learned encoder is resilient to spirious variations, it is not invariant under significant distribution shift. To address this, we propose a simple reward-free alignment procedure that enables test time adaptation of the encoder. This allows for quick adaptation to widely differing environments without having to relearn the dynamics and policy. Our effort is a step towards making model-based RL a practical and useful tool for dynamic, diverse domains. We show its effectiveness in simulation benchmarks with significant spurious variations as well as a real-world egocentric navigation task with noisy TVs in the background. Videos and code at https://zchuning.github.io/repo-website/."
                    },
                    {
                        "title": "Oracle-Ef\ufb01cient Smoothed Online Learning for Piecewise Continuous Decision Making",
                        "abstract": "Smoothed online learning has emerged as a popular framework to mitigate the substantial loss in statistical and computational complexity that arises when one moves from classical to adversarial learning. Unfortunately, for some spaces, it has been shown that ef\ufb01cient algorithms suffer an exponentially worse regret than that which is minimax optimal, even when the learner has access to an optimization oracle over the space. To mitigate that exponential dependence, this work introduces a new notion of complexity, the generalized bracketing numbers, which marries constraints on the adversary to the size of the space, and shows that an instantiation of Follow-the-Perturbed-Leader can attain low regret with the number of calls to the optimization oracle scaling optimally with respect to average regret. We then instantiate our bounds in several problems of interest, including online prediction and planning of piecewise continuous functions, which has many applications in \ufb01elds as diverse as econometrics and robotics."
                    }
                ]
            },
            "16084fae-7f6d-4995-97c2-592f01dd0ea7": {
                "pk": "16084fae-7f6d-4995-97c2-592f01dd0ea7",
                "name": "Cyril Zhang",
                "collaborators": [
                    "Akshay Krishnamurthy",
                    "Keegan Harris",
                    "Dylan J. Foster",
                    "Aleksandrs Slivkins"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Reinforcement Learning",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "Can large language models explore in-context?",
                        "abstract": "We investigate the extent to which contemporary Large Language Models (LLMs) can engage in exploration, a core capability in reinforcement learning and decision making. We focus on native performance of existing LLMs, without training interventions. We deploy LLMs as agents in simple multi-armed bandit environments, specifying the environment description and interaction history entirely in-context, i.e., within the LLM prompt. We experiment with GPT-3.5, GPT-4, and Llama2, using a variety of prompt designs, and find that the models do not robustly engage in exploration without substantial interventions: i) Across all of our experiments, only one configuration resulted in satisfactory exploratory behavior: GPT-4 with chain-of-thought reasoning and an externally summarized interaction history, presented as sufficient statistics; ii) All other configurations did not result in robust exploratory behavior, including those with chain-of-thought reasoning but unsummarized history. Although these findings can be interpreted positively, they suggest that external summarization -- which may not be possible in more complex settings -- is important for obtaining desirable behavior from LLM agents. We conclude that non-trivial algorithmic interventions, such as fine-tuning or dataset curation, may be required to empower LLM-based decision making agents in complex settings."
                    }
                ]
            }
        }
    },
    "2308.00951": {
        "paper_data": {
            "title": "From Sparse to Soft Mixtures of Experts",
            "url": "http://arxiv.org/abs/2308.00951v2",
            "arxiv_id": "2308.00951",
            "authors": [
                "Joan Puigcerver",
                "Carlos Riquelme",
                "Basil Mustafa",
                "Neil Houlsby"
            ],
            "abstract": "Sparse mixture of expert architectures (MoEs) scale model capacity without significant increases in training or inference costs. Despite their success, MoEs suffer from a number of issues: training instability, token dropping, inability to scale the number of experts, or ineffective finetuning. In this work, we propose Soft MoE, a fully-differentiable sparse Transformer that addresses these challenges, while maintaining the benefits of MoEs. Soft MoE performs an implicit soft assignment by passing different weighted combinations of all input tokens to each expert. As in other MoEs, experts in Soft MoE only process a subset of the (combined) tokens, enabling larger model capacity (and performance) at lower inference cost. In the context of visual recognition, Soft MoE greatly outperforms dense Transformers (ViTs) and popular MoEs (Tokens Choice and Experts Choice). Furthermore, Soft MoE scales well: Soft MoE Huge/14 with 128 experts in 16 MoE layers has over 40x more parameters than ViT Huge/14, with only 2% increased inference time, and substantially better quality.",
            "introduction": "   1 Introduction  Larger Transformers improve performance at increased computational cost. Recent studies suggest that model size and training data must be scaled together to optimally use any given training compute budget\u00a0(Kaplan et\u00a0al., 2020; Hoffmann et\u00a0al., 2022; Zhai et\u00a0al., 2022a). A promising alternative that allows to scale models in size without paying their full computational cost is sparse mixtures of experts (MoEs). Recently, a number of successful approaches have proposed ways to sparsely activate token paths across the network in language\u00a0(Lepikhin et\u00a0al., 2020; Fedus et\u00a0al., 2022), vision\u00a0(Riquelme et\u00a0al., 2021), and multimodal models\u00a0(Mustafa et\u00a0al., 2022).   Sparse MoE Transformers involve a discrete optimization problem to decide which modules should be applied to each token. These modules are commonly referred to as experts and are usually MLPs. Many techniques have been devised to find good token-to-expert matches: linear programs (Lewis et\u00a0al., 2021), reinforcement learning (Bengio et\u00a0al., 2015), deterministic fixed rules (Roller et\u00a0al., 2021), optimal transport (Liu et\u00a0al., 2022), greedy top-k\ud835\udc58kitalic_k experts per token (Shazeer et\u00a0al., 2017), or greedy top-k\ud835\udc58kitalic_k tokens per expert (Zhou et\u00a0al., 2022). Often, heuristic auxiliary losses are required to balance utilization of experts and minimize unassigned tokens. These challenges can be greater in out-of-distribution settings: small inference batch sizes, novel inputs, or in transfer learning.   We introduce Soft MoE, that overcomes many of these challenges. Rather than employing a sparse and discrete router that tries to find a good hard assignment between tokens and experts, Soft MoEs instead perform a soft assignment by mixing tokens. In particular, we compute several weighted averages of all tokens\u2014with weights depending on both tokens and experts\u2014and then we process each weighted average by its corresponding expert.   Soft MoE L/16 outperforms ViT H/14 on upstream, few-shot and finetuning while requiring almost half the training time, and being 2\u00d7\\times\u00d7 faster at inference. Moreover, Soft MoE B/16 matches ViT H/14 on few-shot and finetuning and outperforms it on upstream metrics after a comparable amount of training. Remarkably, Soft MoE B/16 is 5.7\u00d7\\times\u00d7 faster at inference despite having 5.5\u00d7\\times\u00d7 the number of parameters of ViT H/14 (see Table\u00a01 and Figure\u00a05 for details). Section\u00a04 demonstrates Soft MoE\u2019s potential to extend to other tasks: we train a contrastive model text tower against the frozen vision tower, showing that representations learned via soft routing preserve their benefits for image-text alignment.     2 Soft Mixture of Experts   2.1 Algorithm description  Figure 1: Sparse and Soft MoE layers. While the router in Sparse MoE layers (left) learns to assign individual input tokens to each of the available slots, in Soft MoE layers (right) each slot is the result of a (different) weighted average of all the input tokens. Learning to make discrete assignments introduces several optimization and implementation issues that Soft MoE sidesteps. Appendix\u00a0G visualizes learned distributions of soft-assignments by Soft MoE.    The Soft MoE routing algorithm is depicted in Figure\u00a02. We denote the inputs tokens for one sequence by \ud835\udc17\u2208\u211dm\u00d7d\ud835\udc17superscript\u211d\ud835\udc5a\ud835\udc51\\mathbf{X}\\in\\mathbb{R}^{m\\times d}bold_X \u2208 blackboard_R start_POSTSUPERSCRIPT italic_m \u00d7 italic_d end_POSTSUPERSCRIPT, where m\ud835\udc5amitalic_m is the number of tokens and d\ud835\udc51ditalic_d is their dimension. Each MoE layer uses a set of n\ud835\udc5bnitalic_n expert functions111In practice, all experts apply the same function with different parameters, usually an MLP. applied on individual tokens, namely {fi:\u211dd\u2192\u211dd}1:nsubscriptconditional-setsubscript\ud835\udc53\ud835\udc56\u2192superscript\u211d\ud835\udc51superscript\u211d\ud835\udc51:1\ud835\udc5b\\{f_{i}:\\mathbb{R}^{d}\\rightarrow\\mathbb{R}^{d}\\}_{1:n}{ italic_f start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT : blackboard_R start_POSTSUPERSCRIPT italic_d",
            "references": [
                {
                    "title": "Soft Merging of Experts with Adaptive Routing",
                    "abstract": "Sparsely activated neural networks with conditional computation learn to route their inputs through different\"expert\"subnetworks, providing a form of modularity that densely activated models lack. Despite their possible benefits, models with learned routing often underperform their parameter-matched densely activated counterparts as well as models that use non-learned heuristic routing strategies. In this paper, we hypothesize that these shortcomings stem from the gradient estimation techniques used to train sparsely activated models that use non-differentiable discrete routing decisions. To address this issue, we introduce Soft Merging of Experts with Adaptive Routing (SMEAR), which avoids discrete routing by using a single\"merged\"expert constructed via a weighted average of all of the experts' parameters. By routing activations through a single merged expert, SMEAR does not incur a significant increase in computational costs and enables standard gradient-based training. We empirically validate that models using SMEAR outperform models that route based on metadata or learn sparse routing through gradient estimation. Furthermore, we provide qualitative analysis demonstrating that the experts learned via SMEAR exhibit a significant amount of specialization. All of the code used in our experiments is publicly available."
                },
                {
                    "title": "Sparsity-Constrained Optimal Transport",
                    "abstract": "Regularized optimal transport (OT) is now increasingly used as a loss or as a matching layer in neural networks. Entropy-regularized OT can be computed using the Sinkhorn algorithm but it leads to fully-dense transportation plans, meaning that all sources are (fractionally) matched with all targets. To address this issue, several works have investigated quadratic regularization instead. This regularization preserves sparsity and leads to unconstrained and smooth (semi) dual objectives, that can be solved with off-the-shelf gradient methods. Unfortunately, quadratic regularization does not give direct control over the cardinality (number of nonzeros) of the transportation plan. We propose in this paper a new approach for OT with explicit cardinality constraints on the transportation plan. Our work is motivated by an application to sparse mixture of experts, where OT can be used to match input tokens such as image patches with expert models such as neural networks. Cardinality constraints ensure that at most $k$ tokens are matched with an expert, which is crucial for computational performance reasons. Despite the nonconvexity of cardinality constraints, we show that the corresponding (semi) dual problems are tractable and can be solved with first-order gradient methods. Our method can be thought as a middle ground between unregularized OT (recovered in the limit case $k=1$) and quadratically-regularized OT (recovered when $k$ is large enough). The smoothness of the objectives increases as $k$ increases, giving rise to a trade-off between convergence speed and sparsity of the optimal plan."
                },
                {
                    "title": "PaLI: A Jointly-Scaled Multilingual Language-Image Model",
                    "abstract": "Effective scaling and a flexible task interface enable large language models to excel at many tasks. We present PaLI (Pathways Language and Image model), a model that extends this approach to the joint modeling of language and vision. PaLI generates text based on visual and textual inputs, and with this interface performs many vision, language, and multimodal tasks, in many languages. To train PaLI, we make use of large pre-trained encoder-decoder language models and Vision Transformers (ViTs). This allows us to capitalize on their existing capabilities and leverage the substantial cost of training them. We find that joint scaling of the vision and language components is important. Since existing Transformers for language are much larger than their vision counterparts, we train a large, 4-billion parameter ViT (ViT-e) to quantify the benefits from even larger-capacity vision models. To train PaLI, we create a large multilingual mix of pretraining tasks, based on a new image-text training set containing 10B images and texts in over 100 languages. PaLI achieves state-of-the-art in multiple vision and language tasks (such as captioning, visual question-answering, scene-text understanding), while retaining a simple, modular, and scalable design."
                },
                {
                    "title": "Multimodal Contrastive Learning with LIMoE: the Language-Image Mixture of Experts",
                    "abstract": "Large sparsely-activated models have obtained excellent performance in multiple domains. However, such models are typically trained on a single modality at a time. We present the Language-Image MoE, LIMoE, a sparse mixture of experts model capable of multimodal learning. LIMoE accepts both images and text simultaneously, while being trained using a contrastive loss. MoEs are a natural fit for a multimodal backbone, since expert layers can learn an appropriate partitioning of modalities. However, new challenges arise; in particular, training stability and balanced expert utilization, for which we propose an entropy-based regularization scheme. Across multiple scales, we demonstrate remarkable performance improvement over dense models of equivalent computational cost. LIMoE-L/16 trained comparably to CLIP-L/14 achieves 78.6% zero-shot ImageNet accuracy (vs. 76.2%), and when further scaled to H/14 (with additional data) it achieves 84.1%, comparable to state-of-the-art methods which use larger custom per-modality backbones and pre-training schemes. We analyse the quantitative and qualitative behavior of LIMoE, and demonstrate phenomena such as differing treatment of the modalities and the organic emergence of modality-specific experts."
                },
                {
                    "title": "Multimodal Token Fusion for Vision Transformers",
                    "abstract": "Many adaptations of transformers have emerged to address the single-modal vision tasks, where self-attention modules are stacked to handle input sources like images. Intuitively, feeding multiple modalities of data to vision transformers could improve the performance, yet the innermodal attentive weights may be diluted, which could thus greatly undermine the final performance. In this paper, we propose a multimodal token fusion method (TokenFusion), tailored for transformer-based vision tasks. To effectively fuse multiple modalities, TokenFusion dynamically detects uninformative tokens and substitute these tokens with projected and aggregated inter-modal features. Residual positional alignment is also adopted to enable explicit utilization of the inter-modal alignments after fusion. The design of TokenFusion allows the transformer to learn correlations among multimodal features, while the single-modal transformer architecture remains largely intact. Extensive experiments are conducted on a variety of homogeneous and heterogeneous modalities and demonstrate that TokenFusion surpasses state-of-the-art methods in three typical vision tasks: multimodal image-to-image translation, RGB-depth semantic segmentation, and 3D object detection with point cloud and images. Code will be released 11https://github.com/huawei-noah/noah-research 22https://gitee.com/mindspore/models/tree/master/research/cv/TokenFusion."
                },
                {
                    "title": "Training Compute-Optimal Large Language Models",
                    "abstract": "We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over 400 language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, Chinchilla, that uses the same compute budget as Gopher but with 70B parameters and 4$\\times$ more more data. Chinchilla uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that Chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, Chinchilla reaches a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, greater than a 7% improvement over Gopher."
                },
                {
                    "title": "Learning to Merge Tokens in Vision Transformers",
                    "abstract": "Transformers are widely applied to solve natural language understanding and computer vision tasks. While scaling up these architectures leads to improved performance, it often comes at the expense of much higher computational costs. In order for large-scale models to remain practical in real-world systems, there is a need for reducing their computational overhead. In this work, we present the PatchMerger, a simple module that reduces the number of patches or tokens the network has to process by merging them between two consecutive intermediate layers. We show that the PatchMerger achieves a significant speedup across various model sizes while matching the original performance both upstream and downstream after fine-tuning."
                },
                {
                    "title": "Mixture-of-Experts with Expert Choice Routing",
                    "abstract": "Sparsely-activated Mixture-of-experts (MoE) models allow the number of parameters to greatly increase while keeping the amount of computation for a given token or a given sample unchanged. However, a poor expert routing strategy (e.g. one resulting in load imbalance) can cause certain experts to be under-trained, leading to an expert being under or over-specialized. Prior work allocates a fixed number of experts to each token using a top-k function regardless of the relative importance of different tokens. To address this, we propose a heterogeneous mixture-of-experts employing an expert choice method. Instead of letting tokens select the top-k experts, we have experts selecting the top-k tokens. As a result, each token can be routed to a variable number of experts and each expert can have a fixed bucket size. We systematically study pre-training speedups using the same computational resources of the Switch Transformer top-1 and GShard top-2 gating of prior work and find that our method improves training convergence time by more than 2x. For the same computational cost, our method demonstrates higher performance in fine-tuning 11 selected tasks in the GLUE and SuperGLUE benchmarks. For a smaller activation cost, our method outperforms the T5 dense model in 7 out of the 11 tasks."
                },
                {
                    "title": "Unified Scaling Laws for Routed Language Models",
                    "abstract": "The performance of a language model has been shown to be effectively modeled as a power-law in its parameter count. Here we study the scaling behaviors of Routing Networks: architectures that conditionally use only a subset of their parameters while processing an input. For these models, parameter count and computational requirement form two independent axes along which an increase leads to better performance. In this work we derive and justify scaling laws defined on these two variables which generalize those known for standard language models and describe the performance of a wide range of routing architectures trained via three different techniques. Afterwards we provide two applications of these laws: first deriving an Effective Parameter Count along which all models scale at the same rate, and then using the scaling coefficients to give a quantitative comparison of the three routing techniques considered. Our analysis derives from an extensive evaluation of Routing Networks across five orders of magnitude of size, including models with hundreds of experts and hundreds of billions of parameters."
                },
                {
                    "title": "LiT: Zero-Shot Transfer with Locked-image text Tuning",
                    "abstract": "This paper presents contrastive-tuning, a simple method employing contrastive training to align image and text mod-els while still taking advantage of their pre-training. In our empirical study we find that locked pre-trained image mod-els with unlocked text models work best. We call this in-stance of contrastive-tuning \u201cLocked-image Tuning\u201d (LiT), which just teaches a text model to read out good repre-sentations from a pre-trained image model for new tasks. A LiT model gains the capability of zero-shot transfer to new vision tasks, such as image classification or retrieval. The proposed LiT is widely applicable; it works reliably with multiple pre-training methods (supervised and unsu-pervised) and across diverse architectures (ResNet, Vision Transformers and MLP-Mixer) using three different image-text datasets. With the transformer-based pre-trained ViT-g/14 model, the LiT model achieves 84.5% zero-shot trans-fer accuracy on the ImageNet test set, and 81.1% on the challenging out-of-distribution ObjectNet test set."
                },
                {
                    "title": "LAION-400M: Open Dataset of CLIP-Filtered 400 Million Image-Text Pairs",
                    "abstract": "Multi-modal language-vision models trained on hundreds of millions of image-text pairs (e.g. CLIP, DALL-E) gained a recent surge, showing remarkable capability to perform zero- or few-shot learning and transfer even in absence of per-sample labels on target image data. Despite this trend, to date there has been no publicly available datasets of sufficient scale for training such models from scratch. To address this issue, in a community effort we build and release for public LAION-400M, a dataset with CLIP-filtered 400 million image-text pairs, their CLIP embeddings and kNN indices that allow efficient similarity search."
                },
                {
                    "title": "TokenLearner: What Can 8 Learned Tokens Do for Images and Videos?",
                    "abstract": "In this paper, we introduce a novel visual representation learning which relies on a handful of adaptively learned tokens, and which is applicable to both image and video understanding tasks. Instead of relying on hand-designed splitting strategies to obtain visual tokens and processing a large number of densely sampled patches for attention, our approach learns to mine important tokens in visual data. This results in efficiently and effectively finding a few important visual tokens and enables modeling of pairwise attention between such tokens, over a longer temporal horizon for videos, or the spatial content in images. Our experiments demonstrate strong performance on several challenging benchmarks for both image and video recognition tasks. Importantly, due to our tokens being adaptive, we accomplish competitive results at significantly reduced compute amount. We obtain comparable results to the state-of-the-arts on ImageNet while being computationally more efficient. We also confirm the effectiveness of the approach on multiple video datasets, including Kinetics-400, Kinetics-600, Charades, and AViD. The code is available at: https://github.com/google-research/scenic/tree/main/scenic/projects/token_learner"
                },
                {
                    "title": "Scaling Vision with Sparse Mixture of Experts",
                    "abstract": "Sparsely-gated Mixture of Experts networks (MoEs) have demonstrated excellent scalability in Natural Language Processing. In Computer Vision, however, almost all performant networks are\"dense\", that is, every input is processed by every parameter. We present a Vision MoE (V-MoE), a sparse version of the Vision Transformer, that is scalable and competitive with the largest dense networks. When applied to image recognition, V-MoE matches the performance of state-of-the-art networks, while requiring as little as half of the compute at inference time. Further, we propose an extension to the routing algorithm that can prioritize subsets of each input across the entire batch, leading to adaptive per-image compute. This allows V-MoE to trade-off performance and compute smoothly at test-time. Finally, we demonstrate the potential of V-MoE to scale vision models, and train a 15B parameter model that attains 90.35% on ImageNet."
                },
                {
                    "title": "Scaling Vision Transformers",
                    "abstract": "Attention-based neural networks such as the Vision Transformer (ViT) have recently attained state-of-the-art results on many computer vision benchmarks. Scale is a primary ingredient in attaining excellent results, therefore, understanding a model's scaling properties is a key to designing future generations effectively. While the laws for scaling Transformer language models have been studied, it is unknown how Vision Transformers scale. To address this, we scale ViT models and data, both up and down, and characterize the relationships between error rate, data, and compute. Along the way, we refine the architecture and training of ViT, reducing memory consumption and increasing accuracy of the resulting models. As a result, we successfully train a ViT model with two billion parameters, which attains a new state-of-the-art on ImageNet of 90.45% top-1 accuracy. The model also performs well for few-shot transfer, for example, reaching 84.86% top-1 accuracy on ImageNet with only 10 examples per class."
                },
                {
                    "title": "Hash Layers For Large Sparse Models",
                    "abstract": "We investigate the training of sparse layers that use different parameters for different inputs based on hashing in large Transformer models. Specifically, we modify the feedforward layer to hash to different sets of weights depending on the current token, over all tokens in the sequence. We show that this procedure either outperforms or is competitive with learning-to-route mixture-of-expert methods such as Switch Transformers and BASE Layers, while requiring no routing parameters or extra terms in the objective function such as a load balancing loss, and no sophisticated assignment algorithm. We study the performance of different hashing techniques, hash sizes and input features, and show that balanced and random hashes focused on the most local features work best, compared to either learning clusters or using longer-range context. We show our approach works well both on large language modeling and dialogue tasks, and on downstream fine-tuning tasks."
                },
                {
                    "title": "BASE Layers: Simplifying Training of Large, Sparse Models",
                    "abstract": "We introduce a new balanced assignment of experts (BASE) layer for large language models that greatly simplifies existing high capacity sparse layers. Sparse layers can dramatically improve the efficiency of training and inference by routing each token to specialized expert modules that contain only a small fraction of the model parameters. However, it can be difficult to learn balanced routing functions that make full use of the available experts; existing approaches typically use routing heuristics or auxiliary expert-balancing loss functions. In contrast, we formulate token-to-expert allocation as a linear assignment problem, allowing an optimal assignment in which each expert receives an equal number of tokens. This optimal assignment scheme improves efficiency by guaranteeing balanced compute loads, and also simplifies training by not requiring any new hyperparameters or auxiliary losses. Code is publicly released at https://github.com/pytorch/fairseq/"
                },
                {
                    "title": "Perceiver: General Perception with Iterative Attention",
                    "abstract": "Biological systems perceive the world by simultaneously processing high-dimensional inputs from modalities as diverse as vision, audition, touch, proprioception, etc. The perception models used in deep learning on the other hand are designed for individual modalities, often relying on domain-specific assumptions such as the local grid structures exploited by virtually all existing vision models. These priors introduce helpful inductive biases, but also lock models to individual modalities. In this paper we introduce the Perceiver - a model that builds upon Transformers and hence makes few architectural assumptions about the relationship between its inputs, but that also scales to hundreds of thousands of inputs, like ConvNets. The model leverages an asymmetric attention mechanism to iteratively distill inputs into a tight latent bottleneck, allowing it to scale to handle very large inputs. We show that this architecture is competitive with or outperforms strong, specialized models on classification tasks across various modalities: images, point clouds, audio, video, and video+audio. The Perceiver obtains performance comparable to ResNet-50 and ViT on ImageNet without 2D convolutions by directly attending to 50,000 pixels. It is also competitive in all modalities in AudioSet."
                },
                {
                    "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity",
                    "abstract": "In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with outrageous numbers of parameters -- but a constant computational cost. However, despite several notable successes of MoE, widespread adoption has been hindered by complexity, communication costs and training instability -- we address these with the Switch Transformer. We simplify the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs. Our proposed training techniques help wrangle the instabilities and we show large sparse models may be trained, for the first time, with lower precision (bfloat16) formats. We design models based off T5-Base and T5-Large to obtain up to 7x increases in pre-training speed with the same computational resources. These improvements extend into multilingual settings where we measure gains over the mT5-Base version across all 101 languages. Finally, we advance the current scale of language models by pre-training up to trillion parameter models on the\"Colossal Clean Crawled Corpus\"and achieve a 4x speedup over the T5-XXL model."
                },
                {
                    "title": "GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding",
                    "abstract": "Neural network scaling has been critical for improving the model quality in many real-world machine learning applications with vast amounts of training data and compute. Although this trend of scaling is affirmed to be a sure-fire approach for better model quality, there are challenges on the path such as the computation cost, ease of programming, and efficient implementation on parallel devices. GShard is a module composed of a set of lightweight annotation APIs and an extension to the XLA compiler. It provides an elegant way to express a wide range of parallel computation patterns with minimal changes to the existing model code. GShard enabled us to scale up multilingual neural machine translation Transformer model with Sparsely-Gated Mixture-of-Experts beyond 600 billion parameters using automatic sharding. We demonstrate that such a giant model can efficiently be trained on 2048 TPU v3 accelerators in 4 days to achieve far superior quality for translation from 100 languages to English compared to the prior art."
                },
                {
                    "title": "On Layer Normalization in the Transformer Architecture",
                    "abstract": "The Transformer is widely used in natural language processing tasks. To train a Transformer however, one usually needs a carefully designed learning rate warm-up stage, which is shown to be crucial to the final performance but will slow down the optimization and bring more hyperparameter tunings. In this paper, we first study theoretically why the learning rate warm-up stage is essential and show that the location of layer normalization matters. Specifically, we prove with mean field theory that at initialization, for the original-designed Post-LN Transformer, which places the layer normalization between the residual blocks, the expected gradients of the parameters near the output layer are large. Therefore, using a large learning rate on those gradients makes the training unstable. The warm-up stage is practically helpful for avoiding this problem. On the other hand, our theory also shows that if the layer normalization is put inside the residual blocks (recently proposed as Pre-LN Transformer), the gradients are well-behaved at initialization. This motivates us to remove the warm-up stage for the training of Pre-LN Transformers. We show in our experiments that Pre-LN Transformers without the warm-up stage can reach comparable results with baselines while requiring significantly less training time and hyper-parameter tuning on a wide range of applications."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "CondConv: Conditionally Parameterized Convolutions for Efficient Inference",
                    "abstract": "Convolutional layers are one of the basic building blocks of modern deep neural networks. One fundamental assumption is that convolutional kernels should be shared for all examples in a dataset. We propose conditionally parameterized convolutions (CondConv), which learn specialized convolutional kernels for each example. Replacing normal convolutions with CondConv enables us to increase the size and capacity of a network, while maintaining efficient inference. We demonstrate that scaling networks with CondConv improves the performance and inference cost trade-off of several existing convolutional neural network architectures on both classification and detection tasks. On ImageNet classification, our CondConv approach applied to EfficientNet-B0 achieves state-ofthe-art performance of 78.3% accuracy with only 413M multiply-adds. Code and checkpoints for the CondConv Tensorflow layer and CondConv-EfficientNet models are available at: https://github.com/tensorflow/tpu/tree/master/ models/official/efficientnet/condconv."
                },
                {
                    "title": "How Much Attention Do You Need? A Granular Analysis of Neural Machine Translation Architectures",
                    "abstract": "With recent advances in network architectures for Neural Machine Translation (NMT) recurrent models have effectively been replaced by either convolutional or self-attentional approaches, such as in the Transformer. While the main innovation of the Transformer architecture is its use of self-attentional layers, there are several other aspects, such as attention with multiple heads and the use of many attention layers, that distinguish the model from previous baselines. In this work we take a fine-grained look at the different architectures for NMT. We introduce an Architecture Definition Language (ADL) allowing for a flexible combination of common building blocks. Making use of this language we show in experiments that one can bring recurrent and convolutional models very close to the Transformer performance by borrowing concepts from the Transformer architecture, but not using self-attention. Additionally, we find that self-attention is much more important on the encoder side than on the decoder side, where it can be replaced by a RNN or CNN without a loss in performance in most settings. Surprisingly, even a model without any target side self-attention performs well."
                },
                {
                    "title": "Revisiting Unreasonable Effectiveness of Data in Deep Learning Era",
                    "abstract": "The success of deep learning in vision can be attributed to: (a) models with high capacity; (b) increased computational power; and (c) availability of large-scale labeled data. Since 2012, there have been significant advances in representation capabilities of the models and computational capabilities of GPUs. But the size of the biggest dataset has surprisingly remained constant. What will happen if we increase the dataset size by 10 \u00d7 or 100 \u00d7 ? This paper takes a step towards clearing the clouds of mystery surrounding the relationship between \u2018enormous data\u2019 and visual deep learning. By exploiting the JFT-300M dataset which has more than 375M noisy labels for 300M images, we investigate how the performance of current vision tasks would change if this data was used for representation learning. Our paper delivers some surprising (and some expected) findings. First, we find that the performance on vision tasks increases logarithmically based on volume of training data size. Second, we show that representation learning (or pre-training) still holds a lot of promise. One can improve performance on many vision tasks by just training a better base model. Finally, as expected, we present new state-of-the-art results for different vision tasks including image classification, object detection, semantic segmentation and human pose estimation. Our sincere hope is that this inspires vision community to not undervalue the data and develop collective efforts in building larger datasets."
                },
                {
                    "title": "Self-Normalizing Neural Networks",
                    "abstract": "Deep Learning has revolutionized vision via convolutional neural networks (CNNs) and natural language processing via recurrent neural networks (RNNs). However, success stories of Deep Learning with standard feed-forward neural networks (FNNs) are rare. FNNs that perform well are typically shallow and, therefore cannot exploit many levels of abstract representations. We introduce self-normalizing neural networks (SNNs) to enable high-level abstract representations. While batch normalization requires explicit normalization, neuron activations of SNNs automatically converge towards zero mean and unit variance. The activation function of SNNs are \"scaled exponential linear units\" (SELUs), which induce self-normalizing properties. Using the Banach fixed-point theorem, we prove that activations close to zero mean and unit variance that are propagated through many network layers will converge towards zero mean and unit variance -- even under the presence of noise and perturbations. This convergence property of SNNs allows to (1) train deep networks with many layers, (2) employ strong regularization, and (3) to make learning highly robust. Furthermore, for activations not close to unit variance, we prove an upper and lower bound on the variance, thus, vanishing and exploding gradients are impossible. We compared SNNs on (a) 121 tasks from the UCI machine learning repository, on (b) drug discovery benchmarks, and on (c) astronomy tasks with standard FNNs and other machine learning methods such as random forests and support vector machines. SNNs significantly outperformed all competing FNN methods at 121 UCI tasks, outperformed all competing methods at the Tox21 dataset, and set a new record at an astronomy data set. The winning SNN architectures are often very deep. Implementations are available at: this http URL."
                },
                {
                    "title": "Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer",
                    "abstract": "The capacity of a neural network to absorb information is limited by its number of parameters. Conditional computation, where parts of the network are active on a per-example basis, has been proposed in theory as a way of dramatically increasing model capacity without a proportional increase in computation. In practice, however, there are significant algorithmic and performance challenges. In this work, we address these challenges and finally realize the promise of conditional computation, achieving greater than 1000x improvements in model capacity with only minor losses in computational efficiency on modern GPU clusters. We introduce a Sparsely-Gated Mixture-of-Experts layer (MoE), consisting of up to thousands of feed-forward sub-networks. A trainable gating network determines a sparse combination of these experts to use for each example. We apply the MoE to the tasks of language modeling and machine translation, where model capacity is critical for absorbing the vast quantities of knowledge available in the training corpora. We present model architectures in which a MoE with up to 137 billion parameters is applied convolutionally between stacked LSTM layers. On large language modeling and machine translation benchmarks, these models achieve significantly better results than state-of-the-art at lower computational cost."
                },
                {
                    "title": "Conditional Computation in Neural Networks for faster models",
                    "abstract": "Deep learning has become the state-of-art tool in many applications, but the evaluation and training of deep models can be time-consuming and computationally expensive. The conditional computation approach has been proposed to tackle this problem (Bengio et al., 2013; Davis & Arel, 2013). It operates by selectively activating only parts of the network at a time. In this paper, we use reinforcement learning as a tool to optimize conditional computation policies. More specifically, we cast the problem of learning activation-dependent policies for dropping out blocks of units as a reinforcement learning problem. We propose a learning scheme motivated by computation speed, capturing the idea of wanting to have parsimonious activations while maintaining prediction accuracy. We apply a policy gradient algorithm for learning policies that optimize this loss function and propose a regularization mechanism that encourages diversification of the dropout policy. We present encouraging empirical results showing that this approach improves the speed of computation without impacting the quality of the approximation."
                },
                {
                    "title": "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification",
                    "abstract": "Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on the learnable activation and advanced initialization, we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66% [33]). To our knowledge, our result is the first to surpass the reported human-level performance (5.1%, [26]) on this dataset."
                },
                {
                    "title": "Understanding the difficulty of training deep feedforward neural networks",
                    "abstract": "Whereas before 2006 it appears that deep multilayer neural networks were not successfully trained, since then several algorithms have been shown to successfully train them, with experimental results showing the superiority of deeper vs less deep architectures. All these experimental results were obtained with new initialization or training mechanisms. Our objective here is to understand better why standard gradient descent from random initialization is doing so poorly with deep neural networks, to better understand these recent relative successes and help design better algorithms in the future. We first observe the influence of the non-linear activations functions. We find that the logistic sigmoid activation is unsuited for deep networks with random initialization because of its mean value, which can drive especially the top hidden layer into saturation. Surprisingly, we find that saturated units can move out of saturation by themselves, albeit slowly, and explaining the plateaus sometimes seen when training neural networks. We find that a new non-linearity that saturates less can often be beneficial. Finally, we study how activations and gradients vary across layers and during training, with the idea that training may be more difficult when the singular values of the Jacobian associated with each layer are far from 1. Based on these considerations, we propose a new initialization scheme that brings substantially faster convergence. 1 Deep Neural Networks Deep learning methods aim at learning feature hierarchies with features from higher levels of the hierarchy formed by the composition of lower level features. They include Appearing in Proceedings of the 13 International Conference on Artificial Intelligence and Statistics (AISTATS) 2010, Chia Laguna Resort, Sardinia, Italy. Volume 9 of JMLR: WC Weston et al., 2008). Much attention has recently been devoted to them (see (Bengio, 2009) for a review), because of their theoretical appeal, inspiration from biology and human cognition, and because of empirical success in vision (Ranzato et al., 2007; Larochelle et al., 2007; Vincent et al., 2008) and natural language processing (NLP) (Collobert & Weston, 2008; Mnih & Hinton, 2009). Theoretical results reviewed and discussed by Bengio (2009), suggest that in order to learn the kind of complicated functions that can represent high-level abstractions (e.g. in vision, language, and other AI-level tasks), one may need deep architectures. Most of the recent experimental results with deep architecture are obtained with models that can be turned into deep supervised neural networks, but with initialization or training schemes different from the classical feedforward neural networks (Rumelhart et al., 1986). Why are these new algorithms working so much better than the standard random initialization and gradient-based optimization of a supervised training criterion? Part of the answer may be found in recent analyses of the effect of unsupervised pretraining (Erhan et al., 2009), showing that it acts as a regularizer that initializes the parameters in a \u201cbetter\u201d basin of attraction of the optimization procedure, corresponding to an apparent local minimum associated with better generalization. But earlier work (Bengio et al., 2007) had shown that even a purely supervised but greedy layer-wise procedure would give better results. So here instead of focusing on what unsupervised pre-training or semi-supervised criteria bring to deep architectures, we focus on analyzing what may be going wrong with good old (but deep) multilayer neural networks. Our analysis is driven by investigative experiments to monitor activations (watching for saturation of hidden units) and gradients, across layers and across training iterations. We also evaluate the effects on these of choices of activation function (with the idea that it might affect saturation) and initialization procedure (since unsupervised pretraining is a particular form of initialization and it has a drastic impact)."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively implement a Soft Mixture of Experts (MoE) architecture to improve the performance of large-scale Transformers while reducing computational costs and addressing the challenges of discrete optimization in token-to-expert assignments?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing need for efficient model architectures that can scale without incurring prohibitive computational costs. By advancing the understanding of Soft MoEs, this research could lead to more effective training and inference processes, enabling the development of larger and more capable models across various domains, including natural language processing, computer vision, and multimodal tasks. The implications of this work could foster further innovations in model design and optimization techniques, ultimately enhancing the capabilities of AI systems in real-world applications.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexities of discrete optimization in sparse MoE architectures, where the assignment of tokens to experts can lead to suboptimal performance due to the rigid nature of hard assignments. Naive approaches may fail because they do not account for the nuanced relationships between tokens and experts, leading to inefficient utilization of resources and potential loss of information. Additionally, the need to balance expert utilization and minimize unassigned tokens introduces further technical obstacles, particularly in out-of-distribution settings where the model encounters novel inputs or operates under varying batch sizes.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on discrete routing mechanisms that struggle with optimization and implementation issues, leading to limitations in performance and flexibility. The reliance on heuristic auxiliary losses to manage expert utilization has also hindered progress. Barriers such as the complexity of designing effective routing algorithms and the computational overhead associated with discrete assignments have prevented the successful implementation of more adaptive solutions. The Soft MoE approach differs by employing a soft assignment mechanism that allows for weighted averaging of tokens, thus sidestepping many of the challenges faced by traditional methods.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves implementing the Soft MoE architecture, which utilizes a soft assignment mechanism to compute weighted averages of input tokens for processing by expert functions. The dataset will include various benchmarks for upstream tasks, few-shot learning, and fine-tuning scenarios. The performance metrics will focus on training time, inference speed, and accuracy on these"
            }
        },
        "author_data": {
            "fe3f0962-ad9d-4f53-9518-6aa8b2603eab": {
                "pk": "fe3f0962-ad9d-4f53-9518-6aa8b2603eab",
                "name": "Joan Puigcerver",
                "collaborators": [
                    "N. Houlsby",
                    "Basil Mustafa",
                    "C. Riquelme",
                    "Lucas Beyer",
                    "Rodolphe Jenatton",
                    "A. Steiner",
                    "Alexander Kolesnikov",
                    "Daniel Keysers",
                    "Matthias Minderer",
                    "Josip Djolonga",
                    "Xiaohua Zhai",
                    "Andr\u00e9 Susano Pinto",
                    "Xiao Wang",
                    "Ibrahim M. Alabdulmohsin",
                    "A. Gritsenko",
                    "Mathieu Blondel",
                    "Mostafa Dehghani",
                    "Piotr Padlewski",
                    "Daniel M. Salz",
                    "Maxim Neumann",
                    "Adam Grycner",
                    "Manoj Kumar",
                    "Keran Rong",
                    "Xi Chen",
                    "Radu Soricut",
                    "Jeremiah Harmsen",
                    "Tianlin Liu",
                    "J. Heek",
                    "Mathilde Caron",
                    "Robert Geirhos",
                    "M. Tschannen",
                    "Avital Oliver",
                    "Yi Tay",
                    "Dustin Tran",
                    "Mario Luvci'c",
                    "Jessica Yung",
                    "Carlos Riquelme Ruiz",
                    "Michael Tschannen",
                    "Emanuele Bugliarello",
                    "Thomas Unterthiner",
                    "Skanda Koppula",
                    "Fangyu Liu",
                    "Julian Martin Eisenschlos",
                    "Rishabh Kabra",
                    "Matthias Bauer",
                    "Matko Bovsnjak",
                    "P. Voigtlaender",
                    "Ioana Bica",
                    "Ivana Balazevic",
                    "Pinelopi Papalampidi",
                    "Olivier J. H\u00e9naff",
                    "Xi Xiong",
                    "Xiao-Qi Zhai",
                    "J. Gilmer",
                    "Anurag Arnab",
                    "Utku Evci",
                    "Sjoerd van Steenkiste",
                    "Gamaleldin F. Elsayed",
                    "Aravindh Mahendran",
                    "F. Yu",
                    "Fantine Huot",
                    "Jasmijn Bastings",
                    "Mark Collier",
                    "Vighnesh Birodkar",
                    "C. Vasconcelos",
                    "Thomas Mensink",
                    "Filip Paveti'c",
                    "Thomas Kipf",
                    "Michael E. Sander",
                    "Gabriel Peyr'e",
                    "C\u00e9dric Renggli",
                    "Soravit Changpinyo",
                    "A. Piergiovanni",
                    "Sebastian Goodman",
                    "Nan Ding",
                    "Hassan Akbari",
                    "Gaurav Mishra",
                    "Linting Xue",
                    "Ashish V. Thapliyal",
                    "James Bradbury",
                    "Weicheng Kuo",
                    "Mojtaba Seyedhosseini",
                    "Chao Jia",
                    "Burcu Karagol Ayan",
                    "A. Angelova",
                    "Aran Komatsuzaki",
                    "J. Lee-Thorp",
                    "J. Ainslie",
                    "Pranjal Awasthi",
                    "Srinadh Bhojanapalli",
                    "J. Allingham",
                    "F. Wenzel",
                    "Zelda E. Mariet",
                    "Ghassen Jerfel",
                    "Vincent Fortuin",
                    "Balaji Lakshminarayanan",
                    "Jasper Snoek",
                    "E. Vidal",
                    "A. Toselli",
                    "andAndr'e Susano Pinto"
                ],
                "domain": [
                    "Vision-Language Model",
                    "Mixture-of-Experts",
                    "Transfer Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "PaliGemma: A versatile 3B VLM for transfer",
                        "abstract": "PaliGemma is an open Vision-Language Model (VLM) that is based on the SigLIP-So400m vision encoder and the Gemma-2B language model. It is trained to be a versatile and broadly knowledgeable base model that is effective to transfer. It achieves strong performance on a wide variety of open-world tasks. We evaluate PaliGemma on almost 40 diverse tasks including standard VLM benchmarks, but also more specialized tasks such as remote-sensing and segmentation."
                    },
                    {
                        "title": "Routers in Vision Mixture of Experts: An Empirical Study",
                        "abstract": "Mixture-of-Experts (MoE) models are a promising way to scale up model capacity without significantly increasing computational cost. A key component of MoEs is the router, which decides which subset of parameters (experts) process which feature embeddings (tokens). In this paper, we present a comprehensive study of routers in MoEs for computer vision tasks. We introduce a unified MoE formulation that subsumes different MoEs with two parametric routing tensors. This formulation covers both sparse MoE, which uses a binary or hard assignment between experts and tokens, and soft MoE, which uses a soft assignment between experts and weighted combinations of tokens. Routers for sparse MoEs can be further grouped into two variants: Token Choice, which matches experts to each token, and Expert Choice, which matches tokens to each expert. We conduct head-to-head experiments with 6 different routers, including existing routers from prior work and new ones we introduce. We show that (i) many routers originally developed for language modeling can be adapted to perform strongly in vision tasks, (ii) in sparse MoE, Expert Choice routers generally outperform Token Choice routers, and (iii) soft MoEs generally outperform sparse MoEs with a fixed compute budget. These results provide new insights regarding the crucial role of routers in vision MoE models."
                    },
                    {
                        "title": "Scaling Vision Transformers to 22 Billion Parameters",
                        "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                    },
                    {
                        "title": "Patch n' Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution",
                        "abstract": "The ubiquitous and demonstrably suboptimal choice of resizing images to a fixed resolution before processing them with computer vision models has not yet been successfully challenged. However, models such as the Vision Transformer (ViT) offer flexible sequence-based modeling, and hence varying input sequence lengths. We take advantage of this with NaViT (Native Resolution ViT) which uses sequence packing during training to process inputs of arbitrary resolutions and aspect ratios. Alongside flexible model usage, we demonstrate improved training efficiency for large-scale supervised and contrastive image-text pretraining. NaViT can be efficiently transferred to standard tasks such as image and video classification, object detection, and semantic segmentation and leads to improved results on robustness and fairness benchmarks. At inference time, the input resolution flexibility can be used to smoothly navigate the test-time cost-performance trade-off. We believe that NaViT marks a departure from the standard, CNN-designed, input and modelling pipeline used by most computer vision models, and represents a promising direction for ViTs."
                    },
                    {
                        "title": "Fast, Differentiable and Sparse Top-k: a Convex Analysis Perspective",
                        "abstract": "The top-k operator returns a sparse vector, where the non-zero values correspond to the k largest values of the input. Unfortunately, because it is a discontinuous function, it is difficult to incorporate in neural networks trained end-to-end with backpropagation. Recent works have considered differentiable relaxations, based either on regularization or perturbation techniques. However, to date, no approach is fully differentiable and sparse. In this paper, we propose new differentiable and sparse top-k operators. We view the top-k operator as a linear program over the permutahedron, the convex hull of permutations. We then introduce a p-norm regularization term to smooth out the operator, and show that its computation can be reduced to isotonic optimization. Our framework is significantly more general than the existing one and allows for example to express top-k operators that select values in magnitude. On the algorithmic side, in addition to pool adjacent violator (PAV) algorithms, we propose a new GPU/TPU-friendly Dykstra algorithm to solve isotonic optimization problems. We successfully use our operators to prune weights in neural networks, to fine-tune vision transformers, and as a router in sparse mixture of experts."
                    },
                    {
                        "title": "Learning to Merge Tokens in Vision Transformers",
                        "abstract": "Transformers are widely applied to solve natural language understanding and computer vision tasks. While scaling up these architectures leads to improved performance, it often comes at the expense of much higher computational costs. In order for large-scale models to remain practical in real-world systems, there is a need for reducing their computational overhead. In this work, we present the PatchMerger, a simple module that reduces the number of patches or tokens the network has to process by merging them between two consecutive intermediate layers. We show that the PatchMerger achieves a significant speedup across various model sizes while matching the original performance both upstream and downstream after fine-tuning."
                    },
                    {
                        "title": "PaLI: A Jointly-Scaled Multilingual Language-Image Model",
                        "abstract": "Effective scaling and a flexible task interface enable large language models to excel at many tasks. We present PaLI (Pathways Language and Image model), a model that extends this approach to the joint modeling of language and vision. PaLI generates text based on visual and textual inputs, and with this interface performs many vision, language, and multimodal tasks, in many languages. To train PaLI, we make use of large pre-trained encoder-decoder language models and Vision Transformers (ViTs). This allows us to capitalize on their existing capabilities and leverage the substantial cost of training them. We find that joint scaling of the vision and language components is important. Since existing Transformers for language are much larger than their vision counterparts, we train a large, 4-billion parameter ViT (ViT-e) to quantify the benefits from even larger-capacity vision models. To train PaLI, we create a large multilingual mix of pretraining tasks, based on a new image-text training set containing 10B images and texts in over 100 languages. PaLI achieves state-of-the-art in multiple vision and language tasks (such as captioning, visual question-answering, scene-text understanding), while retaining a simple, modular, and scalable design."
                    },
                    {
                        "title": "Layover Intermediate Layer for Multi-Label Classification in Efficient Transfer Learning",
                        "abstract": "Transfer Learning (TL) is a promising technique to improve the performance of a target task by transferring the knowledge of models trained on relevant source datasets. With the advent of advanced depth models, various methods of exploiting pre-trained depth models at a large scale have come into the limelight. However, for multi-label classification tasks, TL approaches suffer from performance degradation in correctly predicting multiple objects in an image with significant size differences. Since such a hard instance contains imperceptible objects, most pre-trained models lose their ability during downsampling. For the hard instance , this paper proposes a simple but effective classifier for multiple predictions by using the hidden representations from the fixed backbone. To this end, we mix the pre-logit with the intermediate representation with a learnable scale. We show that our method is effective as fine-tuning with few additional parameters and is particularly advantageous for hard instances ."
                    },
                    {
                        "title": "Sparse Upcycling: Training Mixture-of-Experts from Dense Checkpoints",
                        "abstract": "Training large, deep neural networks to convergence can be prohibitively expensive. As a result, often only a small selection of popular, dense models are reused across different contexts and tasks. Increasingly, sparsely activated models, which seek to decouple model size from computation costs, are becoming an attractive alternative to dense models. Although more efficient in terms of quality and computation cost, sparse models remain data-hungry and costly to train from scratch in the large scale regime. In this work, we propose sparse upcycling -- a simple way to reuse sunk training costs by initializing a sparsely activated Mixture-of-Experts model from a dense checkpoint. We show that sparsely upcycled T5 Base, Large, and XL language models and Vision Transformer Base and Large models, respectively, significantly outperform their dense counterparts on SuperGLUE and ImageNet, using only ~50% of the initial dense pretraining sunk cost. The upcycled models also outperform sparse models trained from scratch on 100% of the initial dense pretraining computation budget."
                    },
                    {
                        "title": "Multimodal Contrastive Learning with LIMoE: the Language-Image Mixture of Experts",
                        "abstract": "Large sparsely-activated models have obtained excellent performance in multiple domains. However, such models are typically trained on a single modality at a time. We present the Language-Image MoE, LIMoE, a sparse mixture of experts model capable of multimodal learning. LIMoE accepts both images and text simultaneously, while being trained using a contrastive loss. MoEs are a natural fit for a multimodal backbone, since expert layers can learn an appropriate partitioning of modalities. However, new challenges arise; in particular, training stability and balanced expert utilization, for which we propose an entropy-based regularization scheme. Across multiple scales, we demonstrate remarkable performance improvement over dense models of equivalent computational cost. LIMoE-L/16 trained comparably to CLIP-L/14 achieves 78.6% zero-shot ImageNet accuracy (vs. 76.2%), and when further scaled to H/14 (with additional data) it achieves 84.1%, comparable to state-of-the-art methods which use larger custom per-modality backbones and pre-training schemes. We analyse the quantitative and qualitative behavior of LIMoE, and demonstrate phenomena such as differing treatment of the modalities and the organic emergence of modality-specific experts."
                    },
                    {
                        "title": "Sparsity-Constrained Optimal Transport",
                        "abstract": "Regularized optimal transport (OT) is now increasingly used as a loss or as a matching layer in neural networks. Entropy-regularized OT can be computed using the Sinkhorn algorithm but it leads to fully-dense transportation plans, meaning that all sources are (fractionally) matched with all targets. To address this issue, several works have investigated quadratic regularization instead. This regularization preserves sparsity and leads to unconstrained and smooth (semi) dual objectives, that can be solved with off-the-shelf gradient methods. Unfortunately, quadratic regularization does not give direct control over the cardinality (number of nonzeros) of the transportation plan. We propose in this paper a new approach for OT with explicit cardinality constraints on the transportation plan. Our work is motivated by an application to sparse mixture of experts, where OT can be used to match input tokens such as image patches with expert models such as neural networks. Cardinality constraints ensure that at most $k$ tokens are matched with an expert, which is crucial for computational performance reasons. Despite the nonconvexity of cardinality constraints, we show that the corresponding (semi) dual problems are tractable and can be solved with first-order gradient methods. Our method can be thought as a middle ground between unregularized OT (recovered in the limit case $k=1$) and quadratically-regularized OT (recovered when $k$ is large enough). The smoothness of the objectives increases as $k$ increases, giving rise to a trade-off between convergence speed and sparsity of the optimal plan."
                    },
                    {
                        "title": "On the Adversarial Robustness of Mixture of Experts",
                        "abstract": "Adversarial robustness is a key desirable property of neural networks. It has been empirically shown to be affected by their sizes, with larger networks being typically more robust. Recently, Bubeck and Sellke proved a lower bound on the Lipschitz constant of functions that fit the training data in terms of their number of parameters. This raises an interesting open question, do -- and can -- functions with more parameters, but not necessarily more computational cost, have better robustness? We study this question for sparse Mixture of Expert models (MoEs), that make it possible to scale up the model size for a roughly constant computational cost. We theoretically show that under certain conditions on the routing and the structure of the data, MoEs can have significantly smaller Lipschitz constants than their dense counterparts. The robustness of MoEs can suffer when the highest weighted experts for an input implement sufficiently different functions. We next empirically evaluate the robustness of MoEs on ImageNet using adversarial attacks and show they are indeed more robust than dense models with the same computational cost. We make key observations showing the robustness of MoEs to the choice of experts, highlighting the redundancy of experts in models trained in practice."
                    },
                    {
                        "title": "Scaling Vision with Sparse Mixture of Experts",
                        "abstract": "Sparsely-gated Mixture of Experts networks (MoEs) have demonstrated excellent scalability in Natural Language Processing. In Computer Vision, however, almost all performant networks are\"dense\", that is, every input is processed by every parameter. We present a Vision MoE (V-MoE), a sparse version of the Vision Transformer, that is scalable and competitive with the largest dense networks. When applied to image recognition, V-MoE matches the performance of state-of-the-art networks, while requiring as little as half of the compute at inference time. Further, we propose an extension to the routing algorithm that can prioritize subsets of each input across the entire batch, leading to adaptive per-image compute. This allows V-MoE to trade-off performance and compute smoothly at test-time. Finally, we demonstrate the potential of V-MoE to scale vision models, and train a 15B parameter model that attains 90.35% on ImageNet."
                    },
                    {
                        "title": "Sparse MoEs meet Efficient Ensembles",
                        "abstract": "Machine learning models based on the aggregated outputs of submodels, either at the activation or prediction levels, often exhibit strong performance compared to individual models. We study the interplay of two popular classes of such models: ensembles of neural networks and sparse mixture of experts (sparse MoEs). First, we show that the two approaches have complementary features whose combination is beneficial. This includes a comprehensive evaluation of sparse MoEs in uncertainty related benchmarks. Then, we present Efficient Ensemble of Experts (E$^3$), a scalable and simple ensemble of sparse MoEs that takes the best of both classes of models, while using up to 45% fewer FLOPs than a deep ensemble. Extensive experiments demonstrate the accuracy, log-likelihood, few-shot learning, robustness, and uncertainty improvements of E$^3$ over several challenging vision Transformer-based baselines. E$^3$ not only preserves its efficiency while scaling to models with up to 2.7B parameters, but also provides better predictive performance and uncertainty estimates for larger models."
                    },
                    {
                        "title": "A Probabilistic Framework for Lexicon-based Keyword Spotting in Handwritten Text Images",
                        "abstract": "Query by String Keyword Spotting (KWS) is here considered as a key technology for indexing large collections of handwritten text images to allow fast textual access to the contents of these colections. Under this prespective, a probabilistic framework for lexicon-based KWS in text images is presented. The presentation aims at providing a tutorial view which helps understanding the relations between classical statements of KWS and the relative challenges entailed by these statements. More specifically, the development of the proposed framework makes it self-evident that word recognition or classification implicitly or explicitly underlies any formulation of KWS. Moreover, it clearly suggests that the same statistical models and training methods successfully used for handwriting text recognition, can advantageously used also for KWS, even though KWS does not generally require or rely on any kind of previously produced image transcripts. These ideas are developped into a specific, probabilistically sound approach for segmentation-free, lexicon-based, query-by-string KWS. Experiments carried out using this approach are presented, which support the consistency and general interest of the proposed framework. Several datasets, traditionally used for KWS benchmarking are considered, with results significantly better than those previously published for these datasets. In addition, results on two new, larger handwritten text image datasets are reported, showing the great potential of the methods proposed in this paper for indexing and textual search in large collections of handwritent documents. 1 ar X iv :2 10 4. 04 55 6v 1 [ cs .I R ] 9 A pr 2 02 1"
                    },
                    {
                        "title": "Deep Ensembles for Low-Data Transfer Learning",
                        "abstract": "In the low-data regime, it is difficult to train good supervised models from scratch. Instead practitioners turn to pre-trained models, leveraging transfer learning. Ensembling is an empirically and theoretically appealing way to construct powerful predictive models, but the predominant approach of training multiple deep networks with different random initialisations collides with the need for transfer via pre-trained weights. In this work, we study different ways of creating ensembles from pre-trained models. We show that the nature of pre-training itself is a performant source of diversity, and propose a practical algorithm that efficiently identifies a subset of pre-trained models for any downstream dataset. The approach is simple: Use nearest-neighbour accuracy to rank pre-trained models, fine-tune the best ones with a small hyperparameter sweep, and greedily construct an ensemble to minimise validation cross-entropy. When evaluated together with strong baselines on 19 different downstream tasks (the Visual Task Adaptation Benchmark), this achieves state-of-the-art performance at a much lower inference budget, even when selecting from over 2,000 pre-trained models. We also assess our ensembles on ImageNet variants and show improved robustness to distribution shift."
                    },
                    {
                        "title": "On Robustness and Transferability of Convolutional Neural Networks",
                        "abstract": "Modern deep convolutional networks (CNNs) are often criticized for not generalizing under distributional shifts. However, several recent breakthroughs in transfer learning suggest that these networks can cope with severe distribution shifts and successfully adapt to new tasks from a few training examples. In this work we study the interplay between out-of-distribution and transfer performance of modern image classification CNNs for the first time and investigate the impact of the pre-training data size, the model scale, and the data preprocessing pipeline. We find that increasing both the training set and model sizes significantly improve the distributional shift robustness. Furthermore, we show that, perhaps surprisingly, simple changes in the preprocessing such as modifying the image resolution can significantly mitigate robustness issues in some cases. Finally, we outline the shortcomings of existing robustness evaluation datasets and introduce a synthetic dataset SI-SCORE we use for a systematic analysis across factors of variation common in visual data such as object size and position."
                    }
                ]
            },
            "df0fde07-3da3-4ab2-bf33-c04e7306821f": {
                "pk": "df0fde07-3da3-4ab2-bf33-c04e7306821f",
                "name": "Carlos Riquelme",
                "collaborators": [
                    "J. Puigcerver",
                    "N. Houlsby",
                    "Basil Mustafa",
                    "Andr\u00e9 Susano Pinto",
                    "S. Gelly",
                    "Josip Djolonga",
                    "Rodolphe Jenatton",
                    "Lucas Beyer",
                    "Xiaohua Zhai",
                    "Daniel Keysers",
                    "O. Bousquet",
                    "M. Tschannen",
                    "Alexander Kolesnikov",
                    "C\u00e9dric Renggli",
                    "Maxim Neumann",
                    "Mario Lucic",
                    "Olivier Bachem",
                    "Marcin Michalski",
                    "Piotr Padlewski",
                    "A. Steiner",
                    "Xiao Wang",
                    "Utku Evci",
                    "Damien Vincent",
                    "P. Ruyssen",
                    "Alexey Dosovitskiy",
                    "Tianlin Liu",
                    "Mathieu Blondel",
                    "Mostafa Dehghani",
                    "J. Heek",
                    "J. Gilmer",
                    "Mathilde Caron",
                    "Robert Geirhos",
                    "Ibrahim M. Alabdulmohsin",
                    "Anurag Arnab",
                    "Matthias Minderer",
                    "Manoj Kumar",
                    "Sjoerd van Steenkiste",
                    "Gamaleldin F. Elsayed",
                    "Aravindh Mahendran",
                    "F. Yu",
                    "Avital Oliver",
                    "Fantine Huot",
                    "Jasmijn Bastings",
                    "Mark Collier",
                    "A. Gritsenko",
                    "Vighnesh Birodkar",
                    "C. Vasconcelos",
                    "Yi Tay",
                    "Thomas Mensink",
                    "Filip Paveti'c",
                    "Dustin Tran",
                    "Thomas Kipf",
                    "Mario Luvci'c",
                    "Jeremiah Harmsen",
                    "Elias Frantar",
                    "Dan Alistarh",
                    "Marina Jim\u00e9nez-Torres",
                    "C. P. Silva",
                    "S. Estay",
                    "M. Soto-Gamboa",
                    "Xi Chen",
                    "Soravit Changpinyo",
                    "A. Piergiovanni",
                    "Daniel M. Salz",
                    "Sebastian Goodman",
                    "Adam Grycner",
                    "Nan Ding",
                    "Keran Rong",
                    "Hassan Akbari",
                    "Gaurav Mishra",
                    "Linting Xue",
                    "Ashish V. Thapliyal",
                    "James Bradbury",
                    "Weicheng Kuo",
                    "Mojtaba Seyedhosseini",
                    "Chao Jia",
                    "Burcu Karagol Ayan",
                    "A. Angelova",
                    "Radu Soricut",
                    "Pranjal Awasthi",
                    "Srinadh Bhojanapalli",
                    "S. Lepe",
                    "P. Salgado",
                    "andAndr'e Susano Pinto",
                    "N. Brosse",
                    "Alice Martin",
                    "\u00c9. Moulines",
                    "Luka Rimanic",
                    "Ce Zhang",
                    "Hugo Penedones",
                    "Hartmut Maennel",
                    "Timothy Mann",
                    "Andr\u00e9 Barreto",
                    "Gergely Neu",
                    "A. Kolesnikov",
                    "Karol Kurach",
                    "Anton Raichuk",
                    "P. Sta\u0144czyk",
                    "Michal Zajac",
                    "L. Espeholt"
                ],
                "domain": [
                    "Machine Learning",
                    "Computer Vision",
                    "Transfer Learning",
                    "Mixture of Experts"
                ],
                "publications": [
                    {
                        "title": "Routers in Vision Mixture of Experts: An Empirical Study",
                        "abstract": "Mixture-of-Experts (MoE) models are a promising way to scale up model capacity without significantly increasing computational cost. A key component of MoEs is the router, which decides which subset of parameters (experts) process which feature embeddings (tokens). In this paper, we present a comprehensive study of routers in MoEs for computer vision tasks. We introduce a unified MoE formulation that subsumes different MoEs with two parametric routing tensors. This formulation covers both sparse MoE, which uses a binary or hard assignment between experts and tokens, and soft MoE, which uses a soft assignment between experts and weighted combinations of tokens. Routers for sparse MoEs can be further grouped into two variants: Token Choice, which matches experts to each token, and Expert Choice, which matches tokens to each expert. We conduct head-to-head experiments with 6 different routers, including existing routers from prior work and new ones we introduce. We show that (i) many routers originally developed for language modeling can be adapted to perform strongly in vision tasks, (ii) in sparse MoE, Expert Choice routers generally outperform Token Choice routers, and (iii) soft MoEs generally outperform sparse MoEs with a fixed compute budget. These results provide new insights regarding the crucial role of routers in vision MoE models."
                    },
                    {
                        "title": "Scaling Vision Transformers to 22 Billion Parameters",
                        "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                    },
                    {
                        "title": "Scaling Laws for Sparsely-Connected Foundation Models",
                        "abstract": "We explore the impact of parameter sparsity on the scaling behavior of Transformers trained on massive datasets (i.e.,\"foundation models\"), in both vision and language domains. In this setting, we identify the first scaling law describing the relationship between weight sparsity, number of non-zero parameters, and amount of training data, which we validate empirically across model and data scales; on ViT/JFT-4B and T5/C4. These results allow us to characterize the\"optimal sparsity\", the sparsity level which yields the best performance for a given effective model size and training budget. For a fixed number of non-zero parameters, we identify that the optimal sparsity increases with the amount of data used for training. We also extend our study to different sparsity structures (such as the hardware-friendly n:m pattern) and strategies (such as starting from a pretrained dense model). Our findings shed light on the power and limitations of weight sparsity across various parameter and computational settings, offering both theoretical understanding and practical implications for leveraging sparsity towards computational efficiency improvements."
                    },
                    {
                        "title": "Automatic recognition of Black-necked Swan ( Cygnus melan-2",
                        "abstract": ": The use of drones in animal monitoring programs has two significant limitations. First,"
                    },
                    {
                        "title": "Learning to Merge Tokens in Vision Transformers",
                        "abstract": "Transformers are widely applied to solve natural language understanding and computer vision tasks. While scaling up these architectures leads to improved performance, it often comes at the expense of much higher computational costs. In order for large-scale models to remain practical in real-world systems, there is a need for reducing their computational overhead. In this work, we present the PatchMerger, a simple module that reduces the number of patches or tokens the network has to process by merging them between two consecutive intermediate layers. We show that the PatchMerger achieves a significant speedup across various model sizes while matching the original performance both upstream and downstream after fine-tuning."
                    },
                    {
                        "title": "PaLI: A Jointly-Scaled Multilingual Language-Image Model",
                        "abstract": "Effective scaling and a flexible task interface enable large language models to excel at many tasks. We present PaLI (Pathways Language and Image model), a model that extends this approach to the joint modeling of language and vision. PaLI generates text based on visual and textual inputs, and with this interface performs many vision, language, and multimodal tasks, in many languages. To train PaLI, we make use of large pre-trained encoder-decoder language models and Vision Transformers (ViTs). This allows us to capitalize on their existing capabilities and leverage the substantial cost of training them. We find that joint scaling of the vision and language components is important. Since existing Transformers for language are much larger than their vision counterparts, we train a large, 4-billion parameter ViT (ViT-e) to quantify the benefits from even larger-capacity vision models. To train PaLI, we create a large multilingual mix of pretraining tasks, based on a new image-text training set containing 10B images and texts in over 100 languages. PaLI achieves state-of-the-art in multiple vision and language tasks (such as captioning, visual question-answering, scene-text understanding), while retaining a simple, modular, and scalable design."
                    },
                    {
                        "title": "Multimodal Contrastive Learning with LIMoE: the Language-Image Mixture of Experts",
                        "abstract": "Large sparsely-activated models have obtained excellent performance in multiple domains. However, such models are typically trained on a single modality at a time. We present the Language-Image MoE, LIMoE, a sparse mixture of experts model capable of multimodal learning. LIMoE accepts both images and text simultaneously, while being trained using a contrastive loss. MoEs are a natural fit for a multimodal backbone, since expert layers can learn an appropriate partitioning of modalities. However, new challenges arise; in particular, training stability and balanced expert utilization, for which we propose an entropy-based regularization scheme. Across multiple scales, we demonstrate remarkable performance improvement over dense models of equivalent computational cost. LIMoE-L/16 trained comparably to CLIP-L/14 achieves 78.6% zero-shot ImageNet accuracy (vs. 76.2%), and when further scaled to H/14 (with additional data) it achieves 84.1%, comparable to state-of-the-art methods which use larger custom per-modality backbones and pre-training schemes. We analyse the quantitative and qualitative behavior of LIMoE, and demonstrate phenomena such as differing treatment of the modalities and the organic emergence of modality-specific experts."
                    },
                    {
                        "title": "On the Adversarial Robustness of Mixture of Experts",
                        "abstract": "Adversarial robustness is a key desirable property of neural networks. It has been empirically shown to be affected by their sizes, with larger networks being typically more robust. Recently, Bubeck and Sellke proved a lower bound on the Lipschitz constant of functions that fit the training data in terms of their number of parameters. This raises an interesting open question, do -- and can -- functions with more parameters, but not necessarily more computational cost, have better robustness? We study this question for sparse Mixture of Expert models (MoEs), that make it possible to scale up the model size for a roughly constant computational cost. We theoretically show that under certain conditions on the routing and the structure of the data, MoEs can have significantly smaller Lipschitz constants than their dense counterparts. The robustness of MoEs can suffer when the highest weighted experts for an input implement sufficiently different functions. We next empirically evaluate the robustness of MoEs on ImageNet using adversarial attacks and show they are indeed more robust than dense models with the same computational cost. We make key observations showing the robustness of MoEs to the choice of experts, highlighting the redundancy of experts in models trained in practice."
                    },
                    {
                        "title": "Black and White holes in four-dimensional Chern-Simons gravity",
                        "abstract": "We discuss a four-dimensional gravitational action which was obtained replacing a RandallSundrum type metric in the so called five-dimensional Einstein-Chern-Simons gravity action. We studied black hole solutions of the corresponding 4-dimensional gravitational field equations. It is found that for a spherically symmetric metric such equations lead to a spacetime with a cosmological constant inversely proportional to the square of the compactification radius and to one solution dependent on an arbitrary constant C. If this constant is negative, we find a Schwarzschid-de Sitter black hole. If C is positive, the solution can be understood as a white hole solution which is obtained applying to the solution with C < 0 the discrete coordinate transformation PT accompanied by the transformation C \u2212\u2192 \u2212C\u0303, with C\u0303 > 0, corresponding to a transformation known as mass reversal. a samuel.lepe@pucv.cl b ceriquelme@udec.cl c patsalgado@unap.cl"
                    },
                    {
                        "title": "Scaling Vision with Sparse Mixture of Experts",
                        "abstract": "Sparsely-gated Mixture of Experts networks (MoEs) have demonstrated excellent scalability in Natural Language Processing. In Computer Vision, however, almost all performant networks are\"dense\", that is, every input is processed by every parameter. We present a Vision MoE (V-MoE), a sparse version of the Vision Transformer, that is scalable and competitive with the largest dense networks. When applied to image recognition, V-MoE matches the performance of state-of-the-art networks, while requiring as little as half of the compute at inference time. Further, we propose an extension to the routing algorithm that can prioritize subsets of each input across the entire batch, leading to adaptive per-image compute. This allows V-MoE to trade-off performance and compute smoothly at test-time. Finally, we demonstrate the potential of V-MoE to scale vision models, and train a 15B parameter model that attains 90.35% on ImageNet."
                    },
                    {
                        "title": "Deep Ensembles for Low-Data Transfer Learning",
                        "abstract": "In the low-data regime, it is difficult to train good supervised models from scratch. Instead practitioners turn to pre-trained models, leveraging transfer learning. Ensembling is an empirically and theoretically appealing way to construct powerful predictive models, but the predominant approach of training multiple deep networks with different random initialisations collides with the need for transfer via pre-trained weights. In this work, we study different ways of creating ensembles from pre-trained models. We show that the nature of pre-training itself is a performant source of diversity, and propose a practical algorithm that efficiently identifies a subset of pre-trained models for any downstream dataset. The approach is simple: Use nearest-neighbour accuracy to rank pre-trained models, fine-tune the best ones with a small hyperparameter sweep, and greedily construct an ensemble to minimise validation cross-entropy. When evaluated together with strong baselines on 19 different downstream tasks (the Visual Task Adaptation Benchmark), this achieves state-of-the-art performance at a much lower inference budget, even when selecting from over 2,000 pre-trained models. We also assess our ensembles on ImageNet variants and show improved robustness to distribution shift."
                    },
                    {
                        "title": "On Last-Layer Algorithms for Classification: Decoupling Representation from Uncertainty Estimation",
                        "abstract": "Uncertainty quantification for deep learning is a challenging open problem. Bayesian statistics offer a mathematically grounded framework to reason about uncertainties; however, approximate posteriors for modern neural networks still require prohibitive computational costs. We propose a family of algorithms which split the classification task into two stages: representation learning and uncertainty estimation. We compare four specific instances, where uncertainty estimation is performed via either an ensemble of Stochastic Gradient Descent or Stochastic Gradient Langevin Dynamics snapshots, an ensemble of bootstrapped logistic regressions, or via a number of Monte Carlo Dropout passes. We evaluate their performance in terms of \\emph{selective} classification (risk-coverage), and their ability to detect out-of-distribution samples. Our experiments suggest there is limited value in adding multiple uncertainty layers to deep classifiers, and we observe that these simple methods strongly outperform a vanilla point-estimate SGD in some complex benchmarks like ImageNet."
                    },
                    {
                        "title": "Which Model to Transfer? Finding the Needle in the Growing Haystack",
                        "abstract": "Transfer learning has been recently popularized as a data-efficient alternative to training models from scratch, in particular for computer vision tasks where it provides a remarkably solid baseline. The emergence of rich model repositories, such as TensorFlow Hub, enables the practitioners and researchers to unleash the potential of these models across a wide range of downstream tasks. As these repositories keep growing exponentially, efficiently selecting a good model for the task at hand becomes paramount. We provide a formalization of this problem through afamiliar notion of regret and introduce the predominant strategies, namely task-agnostic (e.g. ranking models by their ImageNet performance) and task-aware search strategies (such as linear or kNN evaluation). We conduct a large-scale empirical study and show that both task-agnostic and task-aware methods can yield high regret. We then propose a simple and computationally efficient hybrid search strategy which outperforms the existing approaches. We highlight the practical benefits of the proposed solution on a set of 19 diverse vision tasks."
                    },
                    {
                        "title": "Scalable Transfer Learning with Expert Models",
                        "abstract": "Transfer of pre-trained representations can improve sample efficiency and reduce computational requirements for new tasks. However, representations used for transfer are usually generic, and are not tailored to a particular distribution of downstream tasks. We explore the use of expert representations for transfer with a simple, yet effective, strategy. We train a diverse set of experts by exploiting existing label structures, and use cheap-to-compute performance proxies to select the relevant expert for each target task. This strategy scales the process of transferring to new tasks, since it does not revisit the pre-training data during transfer. Accordingly, it requires little extra compute per target task, and results in a speed-up of 2-3 orders of magnitude compared to competing approaches. Further, we provide an adapter-based architecture able to compress many experts into a single model. We evaluate our approach on two different data sources and demonstrate that it outperforms baselines on over 20 diverse vision tasks in both cases."
                    },
                    {
                        "title": "Adaptive Temporal-Difference Learning for Policy Evaluation with Per-State Uncertainty Estimates",
                        "abstract": "We consider the core reinforcement-learning problem of on-policy value function approximation from a batch of trajectory data, and focus on various issues of Temporal Difference (TD) learning and Monte Carlo (MC) policy evaluation. The two methods are known to achieve complementary bias-variance trade-off properties, with TD tending to achieve lower variance but potentially higher bias. In this paper, we argue that the larger bias of TD can be a result of the amplification of local approximation errors. We address this by proposing an algorithm that adaptively switches between TD and MC in each state, thus mitigating the propagation of errors. Our method is based on learned confidence intervals that detect biases of TD estimates. We demonstrate in a variety of policy evaluation tasks that this simple adaptive algorithm performs competitively with the best approach in hindsight, suggesting that learned confidence intervals are a powerful technique for adapting policy evaluation to use TD or MC returns in a data-driven way."
                    },
                    {
                        "title": "A Large-scale Study of Representation Learning with the Visual Task Adaptation Benchmark",
                        "abstract": "Representation learning promises to unlock deep learning for the long tail of vision tasks without expensive labelled datasets. Yet, the absence of a unified evaluation for general visual representations hinders progress. Popular protocols are often too constrained (linear classification), limited in diversity (ImageNet, CIFAR, Pascal-VOC), or only weakly related to representation quality (ELBO, reconstruction error). We present the Visual Task Adaptation Benchmark (VTAB), which defines good representations as those that adapt to diverse, unseen tasks with few examples. With VTAB, we conduct a large-scale study of many popular publicly-available representation learning algorithms. We carefully control confounders such as architecture and tuning budget. We address questions like: How effective are ImageNet representations beyond standard natural datasets? How do representations trained via generative and discriminative models compare? To what extent can self-supervision replace labels? And, how close are we to general visual representations?"
                    },
                    {
                        "title": "Google Research Football: A Novel Reinforcement Learning Environment",
                        "abstract": "Recent progress in the field of reinforcement learning has been accelerated by virtual learning environments such as video games, where novel algorithms and ideas can be quickly tested in a safe and reproducible manner. We introduce the Google Research Football Environment, a new reinforcement learning environment where agents are trained to play football in an advanced, physics-based 3D simulator. The resulting environment is challenging, easy to use and customize, and it is available under a permissive open-source license. In addition, it provides support for multiplayer and multi-agent experiments. We propose three full-game scenarios of varying difficulty with the Football Benchmarks and report baseline results for three commonly used reinforcement algorithms (IMPALA, PPO, and Ape-X DQN). We also provide a diverse set of simpler scenarios with the Football Academy and showcase several promising research directions."
                    },
                    {
                        "title": "The Visual Task Adaptation Benchmark",
                        "abstract": "Representation learning promises to unlock deep learning for the long tail of vision tasks without expansive labelled datasets. Yet, the absence of a unified yardstick to evaluate general visual representations hinders progress. Many sub-fields promise representations, but each has different evaluation protocols that are either too constrained (linear classification), limited in scope (ImageNet, CIFAR, Pascal-VOC), or only loosely related to representation quality (generation). We present the Visual Task Adaptation Benchmark (VTAB): a diverse, realistic, and challenging benchmark to evaluate representations. VTAB embodies one principle: good representations adapt to unseen tasks with few examples. We run a large VTAB study of popular algorithms, answering questions like: How effective are ImageNet representation on non-standard datasets? Are generative models competitive? Is self-supervision useful if one already has labels?"
                    },
                    {
                        "title": "Practical and Consistent Estimation of f-Divergences",
                        "abstract": "The estimation of an f-divergence between two probability distributions based on samples is a fundamental problem in statistics and machine learning. Most works study this problem under very weak assumptions, in which case it is provably hard. We consider the case of stronger structural assumptions that are commonly satisfied in modern machine learning, including representation learning and generative modelling with autoencoder architectures. Under these assumptions we propose and study an estimator that can be easily implemented, works well in high dimensions, and enjoys faster rates of convergence. We verify the behavior of our estimator empirically in both synthetic and real-data experiments, and discuss its direct implications for total correlation, entropy, and mutual information estimation."
                    }
                ]
            },
            "a89c4a5e-7410-434d-b4bd-a703e3843803": {
                "pk": "a89c4a5e-7410-434d-b4bd-a703e3843803",
                "name": "Basil Mustafa",
                "collaborators": [
                    "N. Houlsby",
                    "Lucas Beyer",
                    "Xiao Wang",
                    "A. Steiner",
                    "Alexander Kolesnikov",
                    "Piotr Padlewski",
                    "Mostafa Dehghani",
                    "Xiaohua Zhai",
                    "J. Puigcerver",
                    "Xi Chen",
                    "Josip Djolonga",
                    "Ibrahim M. Alabdulmohsin",
                    "Daniel Keysers",
                    "Keran Rong",
                    "Radu Soricut",
                    "Sebastian Goodman",
                    "Yi Tay",
                    "Siamak Shakeri",
                    "Daniel M. Salz",
                    "Matthias Minderer",
                    "Filip Pavetic",
                    "Rodolphe Jenatton",
                    "C. Riquelme",
                    "Soravit Changpinyo",
                    "Jialin Wu",
                    "Carlos Riquelme Ruiz",
                    "Marvin Ritter",
                    "A. Piergiovanni",
                    "Anurag Arnab",
                    "Yuanzhong Xu",
                    "Mojtaba Seyedhosseini",
                    "A. Angelova",
                    "Mario Luvci'c",
                    "M. Tschannen",
                    "Mark Collier",
                    "Jan Freyberg",
                    "Yossi Matias",
                    "D. Webster",
                    "Mario Lucic",
                    "Arsha Nagrani",
                    "Hexiang Hu",
                    "Mandar Joshi",
                    "Bo Pang",
                    "Ceslee Montgomery",
                    "Paulina Pietrzyk",
                    "Austin Waters",
                    "Gang Li",
                    "Kenton Lee",
                    "Yang Li",
                    "Xiao-Qi Zhai",
                    "Jean-Baptiste Alayrac",
                    "Melvin Johnson",
                    "Ben Caine",
                    "Luheng He",
                    "Yong Cheng",
                    "Albert Webson",
                    "P. Voigtlaender",
                    "Aishwarya B Kamath",
                    "Le Hou",
                    "Michael B. Chang",
                    "Fantine Huot",
                    "Jesse Berent",
                    "J. Heek",
                    "Mathilde Caron",
                    "Robert Geirhos",
                    "Avital Oliver",
                    "A. Gritsenko",
                    "R. Rikhye",
                    "Aaron Loh",
                    "G. Hong",
                    "Preeti Singh",
                    "M. A. Smith",
                    "Vijaytha Muralidharan",
                    "Doris Wong",
                    "R. Sayres",
                    "Michelle Phung",
                    "Nicolas Betancourt",
                    "Bradley Fong",
                    "Rachna Sahasrabudhe",
                    "Khoban Nasim",
                    "Alec Eschholz",
                    "Terry Spitz",
                    "G. Corrado",
                    "Katherine Chou",
                    "P. Bui",
                    "Yuan Liu",
                    "Yun Liu",
                    "Justin Ko",
                    "Steven Lin",
                    "Michael Tschannen",
                    "Julien Amelot",
                    "Machel Reid",
                    "Nikolay Savinov",
                    "Denis Teplyashin",
                    "Dmitry Lepikhin",
                    "T. Lillicrap",
                    "Angeliki Lazaridou",
                    "Orhan Firat",
                    "Julian Schrittwieser",
                    "Ioannis Antonoglou"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Computer Vision",
                    "Natural Language Processing",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Closing the AI generalization gap by adjusting for dermatology condition distribution differences across clinical settings",
                        "abstract": "Recently, there has been great progress in the ability of artificial intelligence (AI) algorithms to classify dermatological conditions from clinical photographs. However, little is known about the robustness of these algorithms in real-world settings where several factors can lead to a loss of generalizability. Understanding and overcoming these limitations will permit the development of generalizable AI that can aid in the diagnosis of skin conditions across a variety of clinical settings. In this retrospective study, we demonstrate that differences in skin condition distribution, rather than in demographics or image capture mode are the main source of errors when an AI algorithm is evaluated on data from a previously unseen source. We demonstrate a series of steps to close this generalization gap, requiring progressively more information about the new source, ranging from the condition distribution to training data enriched for data less frequently seen during training. Our results also suggest comparable performance from end-to-end fine tuning versus fine tuning solely the classification layer on top of a frozen embedding model. Our approach can inform the adaptation of AI algorithms to new settings, based on the information and resources available."
                    },
                    {
                        "title": "On Scaling Up a Multilingual Vision and Language Model",
                        "abstract": "We explore the boundaries of scaling up a multilingual vision and language model, both in terms of size of the components and the breadth of its training task mixture. Our model achieves new levels of performance on a wide-range of varied and complex tasks, including multiple image-based captioning and question-answering tasks, image-based document understanding and few-shot (in-context) learning, as well as object detection, video question answering, and video captioning. Our model advances the state-of-the-art on most vision-and-language benchmarks considered (20+ of them). Finally, we observe emerging capabilities, such as complex counting and multilingual object detection, tasks that are not explicitly in the training mix."
                    },
                    {
                        "title": "Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context",
                        "abstract": "In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content."
                    },
                    {
                        "title": "Capabilities of Gemini Models in Medicine",
                        "abstract": "Excellence in a wide variety of medical applications poses considerable challenges for AI, requiring advanced reasoning, access to up-to-date medical knowledge and understanding of complex multimodal data. Gemini models, with strong general capabilities in multimodal and long-context reasoning, offer exciting possibilities in medicine. Building on these core strengths of Gemini, we introduce Med-Gemini, a family of highly capable multimodal models that are specialized in medicine with the ability to seamlessly use web search, and that can be efficiently tailored to novel modalities using custom encoders. We evaluate Med-Gemini on 14 medical benchmarks, establishing new state-of-the-art (SoTA) performance on 10 of them, and surpass the GPT-4 model family on every benchmark where a direct comparison is viable, often by a wide margin. On the popular MedQA (USMLE) benchmark, our best-performing Med-Gemini model achieves SoTA performance of 91.1% accuracy, using a novel uncertainty-guided search strategy. On 7 multimodal benchmarks including NEJM Image Challenges and MMMU (health&medicine), Med-Gemini improves over GPT-4V by an average relative margin of 44.5%. We demonstrate the effectiveness of Med-Gemini's long-context capabilities through SoTA performance on a needle-in-a-haystack retrieval task from long de-identified health records and medical video question answering, surpassing prior bespoke methods using only in-context learning. Finally, Med-Gemini's performance suggests real-world utility by surpassing human experts on tasks such as medical text summarization, alongside demonstrations of promising potential for multimodal medical dialogue, medical research and education. Taken together, our results offer compelling evidence for Med-Gemini's potential, although further rigorous evaluation will be crucial before real-world deployment in this safety-critical domain."
                    },
                    {
                        "title": "PaLI-X: On Scaling up a Multilingual Vision and Language Model",
                        "abstract": "We present the training recipe and results of scaling up PaLI-X, a multilingual vision and language model, both in terms of size of the components and the breadth of its training task mixture. Our model achieves new levels of performance on a wide-range of varied and complex tasks, including multiple image-based captioning and question-answering tasks, image-based document understanding and few-shot (in-context) learning, as well as object detection, video question answering, and video captioning. PaLI-X advances the state-of-the-art on most vision-and-language benchmarks considered (25+ of them). Finally, we observe emerging capabilities, such as complex counting and multilingual object detection, tasks that are not explicitly in the training mix."
                    },
                    {
                        "title": "Sigmoid Loss for Language Image Pre-Training",
                        "abstract": "We propose a simple pairwise sigmoid loss for imagetext pre-training. Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization. The sigmoid loss simultaneously allows further scaling up the batch size, while also performing better at smaller batch sizes. With only four TPUv4 chips, we can train a Base CLIP model at 4k batch size and a Large LiT model at 20k batch size, the latter achieves 84.5% ImageNet zero-shot accuracy in two days. This disentanglement of the batch size from the loss further allows us to study the impact of examples vs pairs and negative to positive ratio. Finally, we push the batch size to the extreme, up to one million, and find that the benefits of growing batch size quickly diminish, with a more reasonable batch size of 32k being sufficient. We hope our research motivates further explorations in improving the quality and efficiency of language-image pre-training."
                    },
                    {
                        "title": "Three Towers: Flexible Contrastive Learning with Pretrained Image Models",
                        "abstract": "We introduce Three Towers (3T), a flexible method to improve the contrastive learning of vision-language models by incorporating pretrained image classifiers. While contrastive models are usually trained from scratch, LiT (Zhai et al., 2022) has recently shown performance gains from using pretrained classifier embeddings. However, LiT directly replaces the image tower with the frozen embeddings, excluding any potential benefits from training the image tower contrastively. With 3T, we propose a more flexible strategy that allows the image tower to benefit from both pretrained embeddings and contrastive training. To achieve this, we introduce a third tower that contains the frozen pretrained embeddings, and we encourage alignment between this third tower and the main image-text towers. Empirically, 3T consistently improves over LiT and the CLIP-style from-scratch baseline for retrieval tasks. For classification, 3T reliably improves over the from-scratch baseline, and while it underperforms relative to LiT for JFT-pretrained models, it outperforms LiT for ImageNet-21k and Places365 pretraining."
                    },
                    {
                        "title": "Scaling Vision Transformers to 22 Billion Parameters",
                        "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                    },
                    {
                        "title": "PaLI-3 Vision Language Models: Smaller, Faster, Stronger",
                        "abstract": "This paper presents PaLI-3, a smaller, faster, and stronger vision language model (VLM) that compares favorably to similar models that are 10x larger. As part of arriving at this strong performance, we compare Vision Transformer (ViT) models pretrained using classification objectives to contrastively (SigLIP) pretrained ones. We find that, while slightly underperforming on standard image classification benchmarks, SigLIP-based PaLI shows superior performance across various multimodal benchmarks, especially on localization and visually-situated text understanding. We scale the SigLIP image encoder up to 2 billion parameters, and achieves a new state-of-the-art on multilingual cross-modal retrieval. We hope that PaLI-3, at only 5B parameters, rekindles research on fundamental pieces of complex VLMs, and could fuel a new generation of scaled-up models."
                    },
                    {
                        "title": "Patch n' Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution",
                        "abstract": "The ubiquitous and demonstrably suboptimal choice of resizing images to a fixed resolution before processing them with computer vision models has not yet been successfully challenged. However, models such as the Vision Transformer (ViT) offer flexible sequence-based modeling, and hence varying input sequence lengths. We take advantage of this with NaViT (Native Resolution ViT) which uses sequence packing during training to process inputs of arbitrary resolutions and aspect ratios. Alongside flexible model usage, we demonstrate improved training efficiency for large-scale supervised and contrastive image-text pretraining. NaViT can be efficiently transferred to standard tasks such as image and video classification, object detection, and semantic segmentation and leads to improved results on robustness and fairness benchmarks. At inference time, the input resolution flexibility can be used to smoothly navigate the test-time cost-performance trade-off. We believe that NaViT marks a departure from the standard, CNN-designed, input and modelling pipeline used by most computer vision models, and represents a promising direction for ViTs."
                    },
                    {
                        "title": "Massively Scaling Heteroscedastic Classifiers",
                        "abstract": "Heteroscedastic classifiers, which learn a multivariate Gaussian distribution over prediction logits, have been shown to perform well on image classification problems with hundreds to thousands of classes. However, compared to standard classifiers, they introduce extra parameters that scale linearly with the number of classes. This makes them infeasible to apply to larger-scale problems. In addition heteroscedastic classifiers introduce a critical temperature hyperparameter which must be tuned. We propose HET-XL, a heteroscedastic classifier whose parameter count when compared to a standard classifier scales independently of the number of classes. In our large-scale settings, we show that we can remove the need to tune the temperature hyperparameter, by directly learning it on the training data. On large image classification datasets with up to 4B images and 30k classes our method requires 14X fewer additional parameters, does not require tuning the temperature on a held-out set and performs consistently better than the baseline heteroscedastic classifier. HET-XL improves ImageNet 0-shot classification in a multimodal contrastive learning setup which can be viewed as a 3.5 billion class classification problem."
                    },
                    {
                        "title": "Learning to Merge Tokens in Vision Transformers",
                        "abstract": "Transformers are widely applied to solve natural language understanding and computer vision tasks. While scaling up these architectures leads to improved performance, it often comes at the expense of much higher computational costs. In order for large-scale models to remain practical in real-world systems, there is a need for reducing their computational overhead. In this work, we present the PatchMerger, a simple module that reduces the number of patches or tokens the network has to process by merging them between two consecutive intermediate layers. We show that the PatchMerger achieves a significant speedup across various model sizes while matching the original performance both upstream and downstream after fine-tuning."
                    },
                    {
                        "title": "PaLI: A Jointly-Scaled Multilingual Language-Image Model",
                        "abstract": "Effective scaling and a flexible task interface enable large language models to excel at many tasks. We present PaLI (Pathways Language and Image model), a model that extends this approach to the joint modeling of language and vision. PaLI generates text based on visual and textual inputs, and with this interface performs many vision, language, and multimodal tasks, in many languages. To train PaLI, we make use of large pre-trained encoder-decoder language models and Vision Transformers (ViTs). This allows us to capitalize on their existing capabilities and leverage the substantial cost of training them. We find that joint scaling of the vision and language components is important. Since existing Transformers for language are much larger than their vision counterparts, we train a large, 4-billion parameter ViT (ViT-e) to quantify the benefits from even larger-capacity vision models. To train PaLI, we create a large multilingual mix of pretraining tasks, based on a new image-text training set containing 10B images and texts in over 100 languages. PaLI achieves state-of-the-art in multiple vision and language tasks (such as captioning, visual question-answering, scene-text understanding), while retaining a simple, modular, and scalable design."
                    },
                    {
                        "title": "Sparse Upcycling: Training Mixture-of-Experts from Dense Checkpoints",
                        "abstract": "Training large, deep neural networks to convergence can be prohibitively expensive. As a result, often only a small selection of popular, dense models are reused across different contexts and tasks. Increasingly, sparsely activated models, which seek to decouple model size from computation costs, are becoming an attractive alternative to dense models. Although more efficient in terms of quality and computation cost, sparse models remain data-hungry and costly to train from scratch in the large scale regime. In this work, we propose sparse upcycling -- a simple way to reuse sunk training costs by initializing a sparsely activated Mixture-of-Experts model from a dense checkpoint. We show that sparsely upcycled T5 Base, Large, and XL language models and Vision Transformer Base and Large models, respectively, significantly outperform their dense counterparts on SuperGLUE and ImageNet, using only ~50% of the initial dense pretraining sunk cost. The upcycled models also outperform sparse models trained from scratch on 100% of the initial dense pretraining computation budget."
                    },
                    {
                        "title": "Multimodal Contrastive Learning with LIMoE: the Language-Image Mixture of Experts",
                        "abstract": "Large sparsely-activated models have obtained excellent performance in multiple domains. However, such models are typically trained on a single modality at a time. We present the Language-Image MoE, LIMoE, a sparse mixture of experts model capable of multimodal learning. LIMoE accepts both images and text simultaneously, while being trained using a contrastive loss. MoEs are a natural fit for a multimodal backbone, since expert layers can learn an appropriate partitioning of modalities. However, new challenges arise; in particular, training stability and balanced expert utilization, for which we propose an entropy-based regularization scheme. Across multiple scales, we demonstrate remarkable performance improvement over dense models of equivalent computational cost. LIMoE-L/16 trained comparably to CLIP-L/14 achieves 78.6% zero-shot ImageNet accuracy (vs. 76.2%), and when further scaled to H/14 (with additional data) it achieves 84.1%, comparable to state-of-the-art methods which use larger custom per-modality backbones and pre-training schemes. We analyse the quantitative and qualitative behavior of LIMoE, and demonstrate phenomena such as differing treatment of the modalities and the organic emergence of modality-specific experts."
                    },
                    {
                        "title": "CLIPPO: Image-and-Language Understanding from Pixels Only",
                        "abstract": "Multimodal models are becoming increasingly effective, in part due to unified components, such as the Transformer architecture. However, multimodal models still often consist of many task- and modality-specific pieces and training procedures. For example, CLIP (Radford et al., 2021) trains independent text and image towers via a contrastive loss. We explore an additional unification: the use of a pure pixel-based model to perform image, text, and multimodal tasks. Our model is trained with contrastive loss alone, so we call it CLIP-Pixels Only (CLIPPO). CLIPPO uses a single encoder that processes both regular images and text rendered as images. CLIPPO performs image-based tasks such as retrieval and zero-shot image classification almost as well as CLIP-style models, with half the number of parameters and no text-specific tower or embedding. When trained jointly via image-text contrastive learning and next-sentence contrastive learning, CLIPPO can perform well on natural language understanding tasks, without any word-level loss (language modelling or masked language modelling), outperforming pixel-based prior work. Surprisingly, CLIPPO can obtain good accuracy in visual question answering, simply by rendering the question and image together. Finally, we exploit the fact that CLIPPO does not require a tokenizer to show that it can achieve strong performance on multilingual multimodal retrieval without modifications."
                    },
                    {
                        "title": "LiT: Zero-Shot Transfer with Locked-image text Tuning",
                        "abstract": "This paper presents contrastive-tuning, a simple method employing contrastive training to align image and text mod-els while still taking advantage of their pre-training. In our empirical study we find that locked pre-trained image mod-els with unlocked text models work best. We call this in-stance of contrastive-tuning \u201cLocked-image Tuning\u201d (LiT), which just teaches a text model to read out good repre-sentations from a pre-trained image model for new tasks. A LiT model gains the capability of zero-shot transfer to new vision tasks, such as image classification or retrieval. The proposed LiT is widely applicable; it works reliably with multiple pre-training methods (supervised and unsu-pervised) and across diverse architectures (ResNet, Vision Transformers and MLP-Mixer) using three different image-text datasets. With the transformer-based pre-trained ViT-g/14 model, the LiT model achieves 84.5% zero-shot trans-fer accuracy on the ImageNet test set, and 81.1% on the challenging out-of-distribution ObjectNet test set."
                    }
                ]
            },
            "d8c887f7-3d13-487f-8037-ad110f1d0a60": {
                "pk": "d8c887f7-3d13-487f-8037-ad110f1d0a60",
                "name": "Neil Houlsby",
                "collaborators": [
                    "Mostafa Dehghani",
                    "Basil Mustafa",
                    "Manoj Kumar",
                    "Matthias Minderer",
                    "A. Steiner",
                    "Xi Chen",
                    "Josip Djolonga",
                    "Ibrahim M. Alabdulmohsin",
                    "Lucas Beyer",
                    "Radu Soricut",
                    "A. Gritsenko",
                    "Piotr Padlewski",
                    "Xiao Wang",
                    "Yi Tay",
                    "Siamak Shakeri",
                    "Daniel Keysers",
                    "Anurag Arnab",
                    "Yuanzhong Xu",
                    "Keran Rong",
                    "Alexander Kolesnikov",
                    "Daniel M. Salz",
                    "Marvin Ritter",
                    "Filip Pavetic",
                    "Jean-Baptiste Alayrac",
                    "Melvin Johnson",
                    "Yong Cheng",
                    "Albert Webson",
                    "Mario Luvci'c",
                    "J. Puigcerver",
                    "M. Tschannen",
                    "Xiaohua Zhai",
                    "Soravit Changpinyo",
                    "Jialin Wu",
                    "Carlos Riquelme Ruiz",
                    "Sebastian Goodman",
                    "Mario Lucic",
                    "Michael Tschannen",
                    "Arsha Nagrani",
                    "Hexiang Hu",
                    "Mandar Joshi",
                    "Bo Pang",
                    "Ceslee Montgomery",
                    "Paulina Pietrzyk",
                    "A. Piergiovanni",
                    "Austin Waters",
                    "Gang Li",
                    "Kenton Lee",
                    "Yang Li",
                    "Mojtaba Seyedhosseini",
                    "A. Angelova",
                    "Xiao-Qi Zhai",
                    "Nikolay Savinov",
                    "Orhan Firat",
                    "Julian Schrittwieser",
                    "Ioannis Antonoglou",
                    "Rohan Anil",
                    "Sebastian Borgeaud",
                    "Andrew M. Dai",
                    "Katie Millican",
                    "Benjamin Lee",
                    "Fabio Viola",
                    "Malcolm Reynolds",
                    "James Molloy",
                    "Jilin Chen",
                    "M. Isard",
                    "P. Barham",
                    "Tom Hennigan",
                    "J. Schalkwyk",
                    "Eli Collins",
                    "Eliza Rutherford",
                    "Erica Moreira",
                    "Kareem W. Ayoub",
                    "Megha Goel",
                    "Clemens Meyer",
                    "Nathan Schucher",
                    "James Keeling",
                    "Salem Haykal",
                    "Pranav Shyam",
                    "Aakanksha Chowdhery",
                    "George Tucker",
                    "Tom\u00e1s Kocisk\u00fd",
                    "Siddhartha Brahma",
                    "Petko Georgiev",
                    "James Lottes",
                    "Ben Caine",
                    "Luheng He",
                    "Yujia Li",
                    "Albin Cassirer",
                    "Lev Proleev",
                    "Daniel Sohn",
                    "Abhanshu Sharma",
                    "Dawei Jia",
                    "Becca Roelofs",
                    "Rahma Chaabouni",
                    "YaGuang Li",
                    "Yujing Zhang",
                    "Alex Goldin",
                    "Behnam Neyshabur",
                    "Jack W. Rae",
                    "Lakshman Yagati"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Vision-Language Models",
                    "Image Generation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Conditional Diffusion on Web-Scale Image Pairs leads to Diverse Image Variations",
                        "abstract": "Generating image variations, where a model produces variations of an input image while preserving the semantic context has gained increasing attention. Current image variation techniques involve adapting a text-to-image model to reconstruct an input image conditioned on the same image. We first demonstrate that a diffusion model trained to reconstruct an input image from frozen embeddings, can reconstruct the image with minor variations. Second, inspired by how text-to-image models learn from web-scale text-image pairs, we explore a new pretraining strategy to generate image variations using a large collection of image pairs. Our diffusion model \\textit{Semantica} receives a random (encoded) image from a webpage as conditional input and denoises another noisy random image from the same webpage. We carefully examine various design choices for the image encoder, given its crucial role in extracting relevant context from the input image. Once trained, \\textit{Semantica} can adaptively generate new images from a dataset by simply using images from that dataset as input. Finally, we identify limitations in standard image consistency metrics for evaluating image variations and propose alternative metrics based on few-shot generation."
                    },
                    {
                        "title": "On Scaling Up a Multilingual Vision and Language Model",
                        "abstract": "We explore the boundaries of scaling up a multilingual vision and language model, both in terms of size of the components and the breadth of its training task mixture. Our model achieves new levels of performance on a wide-range of varied and complex tasks, including multiple image-based captioning and question-answering tasks, image-based document understanding and few-shot (in-context) learning, as well as object detection, video question answering, and video captioning. Our model advances the state-of-the-art on most vision-and-language benchmarks considered (20+ of them). Finally, we observe emerging capabilities, such as complex counting and multilingual object detection, tasks that are not explicitly in the training mix."
                    },
                    {
                        "title": "Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context",
                        "abstract": "In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content."
                    },
                    {
                        "title": "PaliGemma: A versatile 3B VLM for transfer",
                        "abstract": "PaliGemma is an open Vision-Language Model (VLM) that is based on the SigLIP-So400m vision encoder and the Gemma-2B language model. It is trained to be a versatile and broadly knowledgeable base model that is effective to transfer. It achieves strong performance on a wide variety of open-world tasks. We evaluate PaliGemma on almost 40 diverse tasks including standard VLM benchmarks, but also more specialized tasks such as remote-sensing and segmentation."
                    },
                    {
                        "title": "When Benchmarks are Targets: Revealing the Sensitivity of Large Language Model Leaderboards",
                        "abstract": "Large Language Model (LLM) leaderboards based on benchmark rankings are regularly used to guide practitioners in model selection. Often, the published leaderboard rankings are taken at face value - we show this is a (potentially costly) mistake. Under existing leaderboards, the relative performance of LLMs is highly sensitive to (often minute) details. We show that for popular multiple-choice question benchmarks (e.g., MMLU), minor perturbations to the benchmark, such as changing the order of choices or the method of answer selection, result in changes in rankings up to 8 positions. We explain this phenomenon by conducting systematic experiments over three broad categories of benchmark perturbations and identifying the sources of this behavior. Our analysis results in several best-practice recommendations, including the advantage of a hybrid scoring method for answer selection. Our study highlights the dangers of relying on simple benchmark evaluations and charts the path for more robust evaluation schemes on the existing benchmarks. The code for this paper is available at https://github.com/National-Center-for-AI-Saudi-Arabia/lm-evaluation-harness."
                    },
                    {
                        "title": "Capabilities of Gemini Models in Medicine",
                        "abstract": "Excellence in a wide variety of medical applications poses considerable challenges for AI, requiring advanced reasoning, access to up-to-date medical knowledge and understanding of complex multimodal data. Gemini models, with strong general capabilities in multimodal and long-context reasoning, offer exciting possibilities in medicine. Building on these core strengths of Gemini, we introduce Med-Gemini, a family of highly capable multimodal models that are specialized in medicine with the ability to seamlessly use web search, and that can be efficiently tailored to novel modalities using custom encoders. We evaluate Med-Gemini on 14 medical benchmarks, establishing new state-of-the-art (SoTA) performance on 10 of them, and surpass the GPT-4 model family on every benchmark where a direct comparison is viable, often by a wide margin. On the popular MedQA (USMLE) benchmark, our best-performing Med-Gemini model achieves SoTA performance of 91.1% accuracy, using a novel uncertainty-guided search strategy. On 7 multimodal benchmarks including NEJM Image Challenges and MMMU (health&medicine), Med-Gemini improves over GPT-4V by an average relative margin of 44.5%. We demonstrate the effectiveness of Med-Gemini's long-context capabilities through SoTA performance on a needle-in-a-haystack retrieval task from long de-identified health records and medical video question answering, surpassing prior bespoke methods using only in-context learning. Finally, Med-Gemini's performance suggests real-world utility by surpassing human experts on tasks such as medical text summarization, alongside demonstrations of promising potential for multimodal medical dialogue, medical research and education. Taken together, our results offer compelling evidence for Med-Gemini's potential, although further rigorous evaluation will be crucial before real-world deployment in this safety-critical domain."
                    },
                    {
                        "title": "Frozen Feature Augmentation for Few-Shot Image Classification",
                        "abstract": "Training a linear classifier or lightweight model on top of pretrained vision model outputs, socalled \u2018frozen features\u2019, leads to impressive performance on a number of downstream few-shot tasks. Currently, frozen features are not modified during training. On the other hand, when networks are trained directly on images, data augmentation is a standard recipe that improves performance with no substantial overhead. In this paper, we conduct an extensive pilot study on few-shot image classification that explores applying data augmentations in the frozen feature space, dubbed 'frozen feature augmentation (FroFA)\u2018, cov-ering twenty augmentations in total. Our study demonstrates that adopting a deceptively simple pointwise FroFA, such as brightness, can improve few-shot performance consistently across three network architectures, three large pre-training datasets, and eight transfer datasets."
                    },
                    {
                        "title": "Seed-Free Synthetic Data Generation Framework for Instruction-Tuning LLMs: A Case Study in Thai",
                        "abstract": "We present a synthetic data approach for instruction-tuning large language models (LLMs) for low-resource languages in a data-efficient manner, specifically focusing on Thai. We identify three key properties that contribute to the effectiveness of instruction-tuning datasets: fluency, diversity, and cultural context. We propose a seed-data-free framework for generating synthetic instruction-tuning data that incorporates these essential properties. Our framework employs an LLM to generate diverse topics, retrieve relevant contexts from Wikipedia, and create instructions for various tasks, such as question answering, summarization, and conversation. The experimental results show that our best-performing synthetic dataset, which incorporates all three key properties, achieves competitive performance using only 5,000 instructions when compared to state-of-the-art Thai LLMs trained on hundreds of thousands of instructions. Our code and dataset are publicly available at https://github.com/parinzee/ seed-free-synthetic-instruct ."
                    },
                    {
                        "title": "SciFIBench: Benchmarking Large Multimodal Models for Scientific Figure Interpretation",
                        "abstract": "Large multimodal models (LMMs) have proven flexible and generalisable across many tasks and fields. Although they have strong potential to aid scientific research, their capabilities in this domain are not well characterised. A key aspect of scientific research is the ability to understand and interpret figures, which serve as a rich, compressed source of complex information. In this work, we present SciFIBench, a scientific figure interpretation benchmark. Our main benchmark consists of a 1000-question gold set of multiple-choice questions split between two tasks across 12 categories. The questions are curated from CS arXiv paper figures and captions, using adversarial filtering to find hard negatives and human verification for quality control. We evaluate 26 LMMs on SciFIBench, finding it to be a challenging benchmark. Finally, we investigate the alignment and reasoning faithfulness of the LMMs on augmented question sets from our benchmark. We release SciFIBench to encourage progress in this domain."
                    },
                    {
                        "title": "Dual PatchNorm",
                        "abstract": "We propose Dual PatchNorm: two Layer Normalization layers (LayerNorms), before and after the patch embedding layer in Vision Transformers. We demonstrate that Dual PatchNorm outperforms the result of exhaustive search for alternative LayerNorm placement strategies in the Transformer block itself. In our experiments, incorporating this trivial modification, often leads to improved accuracy over well-tuned Vision Transformers and never hurts."
                    },
                    {
                        "title": "Adaptive Computation with Elastic Input Sequence",
                        "abstract": "Humans have the ability to adapt the type of information they use, the procedure they employ, and the amount of time they spend when solving problems. However, most standard neural networks have a fixed function type and computation budget regardless of the sample's nature or difficulty. Adaptivity is a powerful paradigm as it not only imbues practitioners with flexibility pertaining to the downstream usage of these models but can also serve as a powerful inductive bias for solving certain challenging classes of problems. In this work, we introduce a new approach called AdaTape, which allows for dynamic computation in neural networks through adaptive tape tokens. AdaTape utilizes an elastic input sequence by equipping an architecture with a dynamic read-and-write tape. Specifically, we adaptively generate input sequences using tape tokens obtained from a tape bank which can be either trainable or derived from input data. We examine the challenges and requirements to obtain dynamic sequence content and length, and propose the Adaptive Tape Reading (ATR) algorithm to achieve both goals. Through extensive experiments on image recognition tasks, we show that AdaTape can achieve better performance while maintaining the computational cost. To facilitate further research, we have released code at https://github.com/google-research/scenic."
                    },
                    {
                        "title": "Image Captioners Are Scalable Vision Learners Too",
                        "abstract": "Contrastive pretraining on image-text pairs from the web is one of the most popular large-scale pretraining strategies for vision backbones, especially in the context of large multimodal models. At the same time, image captioning on this type of data is commonly considered an inferior pretraining strategy. In this paper, we perform a fair comparison of these two pretraining strategies, carefully matching training data, compute, and model capacity. Using a standard encoder-decoder transformer, we find that captioning alone is surprisingly effective: on classification tasks, captioning produces vision encoders competitive with contrastively pretrained encoders, while surpassing them on vision&language tasks. We further analyze the effect of the model architecture and scale, as well as the pretraining data on the representation quality, and find that captioning exhibits the same or better scaling behavior along these axes. Overall our results show that plain image captioning is a more powerful pretraining strategy than was previously believed."
                    },
                    {
                        "title": "Scaling Open-Vocabulary Object Detection",
                        "abstract": "Open-vocabulary object detection has benefited greatly from pretrained vision-language models, but is still limited by the amount of available detection training data. While detection training data can be expanded by using Web image-text pairs as weak supervision, this has not been done at scales comparable to image-level pretraining. Here, we scale up detection data with self-training, which uses an existing detector to generate pseudo-box annotations on image-text pairs. Major challenges in scaling self-training are the choice of label space, pseudo-annotation filtering, and training efficiency. We present the OWLv2 model and OWL-ST self-training recipe, which address these challenges. OWLv2 surpasses the performance of previous state-of-the-art open-vocabulary detectors already at comparable training scales (~10M examples). However, with OWL-ST, we can scale to over 1B examples, yielding further large improvement: With an L/14 architecture, OWL-ST improves AP on LVIS rare classes, for which the model has seen no human box annotations, from 31.2% to 44.6% (43% relative improvement). OWL-ST unlocks Web-scale training for open-world localization, similar to what has been seen for image classification and language modelling."
                    },
                    {
                        "title": "PaLI-X: On Scaling up a Multilingual Vision and Language Model",
                        "abstract": "We present the training recipe and results of scaling up PaLI-X, a multilingual vision and language model, both in terms of size of the components and the breadth of its training task mixture. Our model achieves new levels of performance on a wide-range of varied and complex tasks, including multiple image-based captioning and question-answering tasks, image-based document understanding and few-shot (in-context) learning, as well as object detection, video question answering, and video captioning. PaLI-X advances the state-of-the-art on most vision-and-language benchmarks considered (25+ of them). Finally, we observe emerging capabilities, such as complex counting and multilingual object detection, tasks that are not explicitly in the training mix."
                    },
                    {
                        "title": "Scaling Vision Transformers to 22 Billion Parameters",
                        "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                    },
                    {
                        "title": "Patch n' Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution",
                        "abstract": "The ubiquitous and demonstrably suboptimal choice of resizing images to a fixed resolution before processing them with computer vision models has not yet been successfully challenged. However, models such as the Vision Transformer (ViT) offer flexible sequence-based modeling, and hence varying input sequence lengths. We take advantage of this with NaViT (Native Resolution ViT) which uses sequence packing during training to process inputs of arbitrary resolutions and aspect ratios. Alongside flexible model usage, we demonstrate improved training efficiency for large-scale supervised and contrastive image-text pretraining. NaViT can be efficiently transferred to standard tasks such as image and video classification, object detection, and semantic segmentation and leads to improved results on robustness and fairness benchmarks. At inference time, the input resolution flexibility can be used to smoothly navigate the test-time cost-performance trade-off. We believe that NaViT marks a departure from the standard, CNN-designed, input and modelling pipeline used by most computer vision models, and represents a promising direction for ViTs."
                    },
                    {
                        "title": "Scaling Laws for Sparsely-Connected Foundation Models",
                        "abstract": "We explore the impact of parameter sparsity on the scaling behavior of Transformers trained on massive datasets (i.e.,\"foundation models\"), in both vision and language domains. In this setting, we identify the first scaling law describing the relationship between weight sparsity, number of non-zero parameters, and amount of training data, which we validate empirically across model and data scales; on ViT/JFT-4B and T5/C4. These results allow us to characterize the\"optimal sparsity\", the sparsity level which yields the best performance for a given effective model size and training budget. For a fixed number of non-zero parameters, we identify that the optimal sparsity increases with the amount of data used for training. We also extend our study to different sparsity structures (such as the hardware-friendly n:m pattern) and strategies (such as starting from a pretrained dense model). Our findings shed light on the power and limitations of weight sparsity across various parameter and computational settings, offering both theoretical understanding and practical implications for leveraging sparsity towards computational efficiency improvements."
                    },
                    {
                        "title": "Massively Scaling Heteroscedastic Classifiers",
                        "abstract": "Heteroscedastic classifiers, which learn a multivariate Gaussian distribution over prediction logits, have been shown to perform well on image classification problems with hundreds to thousands of classes. However, compared to standard classifiers, they introduce extra parameters that scale linearly with the number of classes. This makes them infeasible to apply to larger-scale problems. In addition heteroscedastic classifiers introduce a critical temperature hyperparameter which must be tuned. We propose HET-XL, a heteroscedastic classifier whose parameter count when compared to a standard classifier scales independently of the number of classes. In our large-scale settings, we show that we can remove the need to tune the temperature hyperparameter, by directly learning it on the training data. On large image classification datasets with up to 4B images and 30k classes our method requires 14X fewer additional parameters, does not require tuning the temperature on a held-out set and performs consistently better than the baseline heteroscedastic classifier. HET-XL improves ImageNet 0-shot classification in a multimodal contrastive learning setup which can be viewed as a 3.5 billion class classification problem."
                    }
                ]
            }
        }
    },
    "2405.18549": {
        "paper_data": {
            "title": "Learning from Uncertain Data: From Possible Worlds to Possible Models",
            "url": "http://arxiv.org/abs/2405.18549v1",
            "arxiv_id": "2405.18549",
            "authors": [
                "Jiongli Zhu",
                "Su Feng",
                "Boris Glavic",
                "Babak Salimi"
            ],
            "abstract": "We introduce an efficient method for learning linear models from uncertain data, where uncertainty is represented as a set of possible variations in the data, leading to predictive multiplicity. Our approach leverages abstract interpretation and zonotopes, a type of convex polytope, to compactly represent these dataset variations, enabling the symbolic execution of gradient descent on all possible worlds simultaneously. We develop techniques to ensure that this process converges to a fixed point and derive closed-form solutions for this fixed point. Our method provides sound over-approximations of all possible optimal models and viable prediction ranges. We demonstrate the effectiveness of our approach through theoretical and empirical analysis, highlighting its potential to reason about model and prediction uncertainty due to data quality issues in training data.",
            "introduction": " Introduction This paper addresses the challenges of learning from uncertain datasets by employing the framework of possible world semantics , a well-established concept in AI and database theory [ 56,21,38,26]. In this approach, uncertainty in a dataset Dis conceptualized through a collection of possible datasets {D1,D2, . . .}, each representing a potential state of the real world, reflecting variations due to missing entries, errors, inconsistencies, and biases. Given this framework and a learning algorithm, our objective is to construct a set of models {f1, f2, . . .}, where each model fiis trained on a corresponding potential dataset Di. This method, that we implement in a system called ZORRO (ZOnotope-based Robustness Analysis for Regression with Over-approximations), allows for a thorough evaluation of how data uncertainties affect the robustness, reliability, and fairness of models in predictive modeling and statistical inference, particularly in scenarios where the ground truth is unidentifiable, necessitating consideration of all possible dataset variations. While the framework of possible world semantics is essential for modeling dataset uncertainties, it poses significant challenges due to the potentially infinite scenarios each dataset might represent. Exploring every possibility and training a model for each is impractical. The concept of model multiplicity , which highlights situations where models with similar accuracy differ in individual predictions, has gained traction [ 7,39], yet it primarily focuses on competing models without addressing the full range of dataset variations. Similarly, dataset multiplicity introduced in [ 44] recognizes data variations due to uncertainties but [ 44] only proposed a solution for linear models with label errors. Our approach expands these ideas by using possible world semantics to systematically manage uncertainties across all features and labels, thus creating a comprehensive framework for evaluating model robustness amid data uncertainties. To address the challenges of learning from uncertain datasets, we employ the method of results) by multiplier of 0.06 to 0.44 where 0.06 means the range for each uncertain data is increased by 6%. Similar to amount of uncertainty, 35Learning from Uncertain Data: From Possible Worlds to Possible Models A P REPRINT 0.00.10.20.30.40.50.60.70.80.9 0.01 0.02 0.03 0.04 0.06 0.07 0.08 0.09 0.1 0.121k 10kZorroloss uncertain ratio (a) varying amount of uncertainty 0.000.050.100.150.200.250.300.35 0.06 0.12 0.19 0.25 0.31 0.38 0.441k 10kZorroloss uncertainty radius (b) varying uncertainty radius 0.080.090.100.110.120.130.140.150.16 5 6 7 8 9 10 11 12 131k 10kZorroloss dimension (c) varying dimension Figure 6: Fixed point versus samplings 0.000.040.080.120.160.200.24 0.06 0.12 0.19 0.25 0.31 0.38 0.441k GTZorroloss range multiplier (a) varying range size 0.070.080.090.100.110.120.13 5 6 7 8 9 10 11 12 131k GTZorroloss dimension (b) varying data dimension Figure 7: The range of losses obtained by enumerating all possible worlds (GT), sampling 1000 possible worlds (1k), and Z ORRO . Varying dimensions Figure 6c Shows the result by change number of dimensions to the training data. Result indicates tightness of Z ORRO \u2019s over-approximation is not affected by the dimension of the data. Figure 7 added sampling result to Figure 3 shows additionally that sampling result is an under approximation of the ground truth result. M.3 Additional Related Work Predictive multiplicity has shown that a single dataset can produce multiple best-fitted models due to variations in training processes [8, 11, 41, 63, 65, 9, 15, 67, 39] or modifications to training parameters such as random seeds, data ordering, and hyperparameters. For predictive multiplicity due to missing data, Khosravi et al.[ 32] address the issue using a probabilistic",
            "references": [
                {
                    "title": "Sources of Uncertainty in Machine Learning - A Statisticians' View",
                    "abstract": "Machine Learning and Deep Learning have achieved an impressive standard today, enabling us to answer questions that were inconceivable a few years ago. Besides these successes, it becomes clear, that beyond pure prediction, which is the primary strength of most supervised machine learning algorithms, the quantification of uncertainty is relevant and necessary as well. While first concepts and ideas in this direction have emerged in recent years, this paper adopts a conceptual perspective and examines possible sources of uncertainty. By adopting the viewpoint of a statistician, we discuss the concepts of aleatoric and epistemic uncertainty, which are more commonly associated with machine learning. The paper aims to formalize the two types of uncertainty and demonstrates that sources of uncertainty are miscellaneous and can not always be decomposed into aleatoric and epistemic. Drawing parallels between statistical concepts and uncertainty in machine learning, we also demonstrate the role of data and their influence on uncertainty."
                },
                {
                    "title": "The Dataset Multiplicity Problem: How Unreliable Data Impacts Predictions",
                    "abstract": "We introduce dataset multiplicity, a way to study how inaccuracies, uncertainty, and social bias in training datasets impact test-time predictions. The dataset multiplicity framework asks a counterfactual question of what the set of resultant models (and associated test-time predictions) would be if we could somehow access all hypothetical, unbiased versions of the dataset. We discuss how to use this framework to encapsulate various sources of uncertainty in datasets\u2019 factualness, including systemic social bias, data collection practices, and noisy labels or features. We show how to exactly analyze the impacts of dataset multiplicity for a specific model architecture and type of uncertainty: linear models with label errors. Our empirical analysis shows that real-world datasets, under reasonable assumptions, contain many test samples whose predictions are affected by dataset multiplicity. Furthermore, the choice of domain-specific dataset multiplicity definition determines what samples are affected, and whether different demographic groups are disparately impacted. Finally, we discuss implications of dataset multiplicity for machine learning practice and research, including considerations for when model outcomes should not be trusted."
                },
                {
                    "title": "Consistent Range Approximation for Fair Predictive Modeling",
                    "abstract": "This paper proposes a novel framework for certifying the fairness of predictive models trained on biased data. It draws from query answering for incomplete and inconsistent databases to formulate the problem of consistent range approximation (CRA) of fairness queries for a predictive model on a target population. The framework employs background knowledge of the data collection process and biased data, working with or without limited statistics about the target population, to compute a range of answers for fairness queries. Using CRA, the framework builds predictive models that are certifiably fair on the target population, regardless of the availability of external data during training. The framework's efficacy is demonstrated through evaluations on real data, showing substantial improvement over existing state-of-the-art methods."
                },
                {
                    "title": "Zonotope Domains for Lagrangian Neural Network Verification",
                    "abstract": "Neural network verification aims to provide provable bounds for the output of a neural network for a given input range. Notable prior works in this domain have either generated bounds using abstract domains, which preserve some dependency between intermediate neurons in the network; or framed verification as an optimization problem and solved a relaxation using Lagrangian methods. A key drawback of the latter technique is that each neuron is treated independently, thereby ignoring important neuron interactions. We provide an approach that merges these two threads and uses zonotopes within a Lagrangian decomposition. Crucially, we can decompose the problem of verifying a deep neural network into the verification of many 2-layer neural networks. While each of these problems is provably hard, we provide efficient relaxation methods that are amenable to efficient dual ascent procedures. Our technique yields bounds that improve upon both linear programming and Lagrangian-based verification techniques in both time and bound tightness."
                },
                {
                    "title": "Prediction Intervals and Confidence Regions for Symbolic Regression Models based on Likelihood Profiles",
                    "abstract": "Symbolic regression is a nonlinear regression method which is commonly performed by an evolutionary computation method such as genetic programming. Quantification of uncertainty of regression models is important for the interpretation of models and for decision making. The linear approximation and so-called likelihood profiles are well-known possibilities for the calculation of confidence and prediction intervals for nonlinear regression models. These simple and effective techniques have been completely ignored so far in the genetic programming literature. In this work we describe the calculation of likelihood profiles in details and also provide some illustrative examples with models created with three different symbolic regression algorithms on two different datasets. The examples highlight the importance of the likelihood profiles to understand the limitations of symbolic regression models and to help the user taking an informed post-prediction decision."
                },
                {
                    "title": "Predicting is not Understanding: Recognizing and Addressing Underspecification in Machine Learning",
                    "abstract": "Machine learning (ML) models are typically optimized for their accuracy on a given dataset. However, this predictive criterion rarely captures all desirable properties of a model, in particular how well it matches a domain expert's understanding of a task. Underspecification refers to the existence of multiple models that are indistinguishable in their in-domain accuracy, even though they differ in other desirable properties such as out-of-distribution (OOD) performance. Identifying these situations is critical for assessing the reliability of ML models. We formalize the concept of underspecification and propose a method to identify and partially address it. We train multiple models with an independence constraint that forces them to implement different functions. They discover predictive features that are otherwise ignored by standard empirical risk minimization (ERM), which we then distill into a global model with superior OOD performance. Importantly, we constrain the models to align with the data manifold to ensure that they discover meaningful features. We demonstrate the method on multiple datasets in computer vision (collages, WILDS-Camelyon17, GQA) and discuss general implications of underspecification. Most notably, in-domain performance cannot serve for OOD model selection without additional assumptions."
                },
                {
                    "title": "Model Multiplicity: Opportunities, Concerns, and Solutions",
                    "abstract": "Recent scholarship has brought attention to the fact that there often exist multiple models for a given prediction task with equal accuracy that differ in their individual-level predictions or aggregate properties. This phenomenon\u2014which we call model multiplicity\u2014can introduce a good deal of flexibility into the model selection process, creating a range of exciting opportunities. By demonstrating that there are many different ways of making equally accurate predictions, multiplicity gives model developers the freedom to prioritize other values in their model selection process without having to abandon their commitment to maximizing accuracy. However, multiplicity also brings to light a concerning truth: model selection on the basis of accuracy alone\u2014the default procedure in many deployment scenarios\u2014fails to consider what might be meaningful differences between equally accurate models with respect to other criteria such as fairness, robustness, and interpretability. Unless these criteria are taken into account explicitly, developers might end up making unnecessary trade-offs or could even mask intentional discrimination. Furthermore, the prospect that there might exist another model of equal accuracy that flips a prediction for a particular individual may lead to a crisis in justifiability: why should an individual be subject to an adverse model outcome if there exists an equally accurate model that treats them more favorably? In this work, we investigate how to take advantage of the flexibility afforded by model multiplicity while addressing the concerns with justifiability that it might raise?"
                },
                {
                    "title": "Certifying Data-Bias Robustness in Linear Regression",
                    "abstract": "Datasets typically contain inaccuracies due to human error and societal biases, and these inaccuracies can affect the outcomes of models trained on such datasets. We present a technique for certifying whether linear regression models are pointwise-robust to label bias in the training dataset, i.e., whether bounded perturbations to the labels of a training dataset result in models that change the prediction of test points. We show how to solve this problem exactly for individual test points, and provide an approximate but more scalable method that does not require advance knowledge of the test point. We extensively evaluate both techniques and find that linear models -- both regression- and classification-based -- often display high levels of bias-robustness. However, we also unearth gaps in bias-robustness, such as high levels of non-robustness for certain bias assumptions on some datasets. Overall, our approach can serve as a guide for when to trust, or question, a model's output."
                },
                {
                    "title": "Verification of Neural-Network Control Systems by Integrating Taylor Models and Zonotopes",
                    "abstract": "We study the verification problem for closed-loop dynamical systems with neural-network controllers (NNCS). This problem is commonly reduced to computing the set of reachable states. When considering dynamical systems and neural networks in isolation, there exist precise approaches for that task based on set representations respectively called Taylor models and zonotopes. However, the combination of these approaches to NNCS is non-trivial because, when converting between the set representations, dependency information gets lost in each control cycle and the accumulated approximation error quickly renders the result useless. We present an algorithm to chain approaches based on Taylor models and zonotopes, yielding a precise reachability algorithm for NNCS. Because the algorithm only acts at the interface of the isolated approaches, it is applicable to general dynamical systems and neural networks and can benefit from future advances in these areas. Our implementation delivers state-of-the-art performance and is the first to successfully analyze all benchmark problems of an annual reachability competition for NNCS."
                },
                {
                    "title": "Cert-RNN: Towards Certifying the Robustness of Recurrent Neural Networks",
                    "abstract": "Certifiable robustness, the functionality of verifying whether the given region surrounding a data point admits any adversarial example, provides guaranteed security for neural networks deployed in adversarial environments. A plethora of work has been proposed to certify the robustness of feed-forward networks, e.g., FCNs and CNNs. Yet, most existing methods cannot be directly applied to recurrent neural networks (RNNs), due to their sequential inputs and unique operations. In this paper, we present Cert-RNN, a general framework for certifying the robustness of RNNs. Specifically, through detailed analysis for the intrinsic property of the unique function in different ranges, we exhaustively discuss different cases for the exact formula of bounding planes, based on which we design several precise and efficient abstract transformers for the unique calculations in RNNs. Cert-RNN significantly outperforms the state-of-the-art methods (e.g., POPQORN) in terms of (i) effectiveness -- it provides much tighter robustness bounds, and (ii) efficiency -- it scales to much more complex models. Through extensive evaluation, we validate Cert-RNN's superior performance across various network architectures (e.g., vanilla RNN and LSTM) and applications (e.g., image classification, sentiment analysis, toxic comment detection, and malicious URL detection). For instance, for the RNN-2-32 model on the MNIST sequence dataset, the robustness bound certified by Cert-RNN is on average 1.86 times larger than that by POPQORN. Besides certifying the robustness of given RNNs, Cert-RNN also enables a range of practical applications including evaluating the provable effectiveness for various defenses (i.e., the defense with a larger robustness region is considered to be more robust), improving the robustness of RNNs (i.e., incorporating Cert-RNN with verified robust training) and identifying sensitive words (i.e., the word with the smallest certified robustness bound is considered to be the most sensitive word in a sentence), which helps build more robust and interpretable deep learning systems. We will open-source Cert-RNN for facilitating the DNN security research."
                },
                {
                    "title": "Abstract Interpretation of Fixpoint Iterators with Applications to Neural Networks",
                    "abstract": "We present a new abstract interpretation framework for the precise over-approximation of numerical fixpoint iterators. Our key observation is that unlike in standard abstract interpretation (AI), typically used to over-approximate all reachable program states, in this setting, one only needs to abstract the concrete fixpoints, i.e., the final program states. Our framework targets numerical fixpoint iterators with convergence and uniqueness guarantees in the concrete and is based on two major technical contributions: (i) theoretical insights which allow us to compute sound and precise fixpoint abstractions without using joins, and (ii) a new abstract domain, CH-Zonotope, which admits efficient propagation and inclusion checks while retaining high precision. We implement our framework in a tool called CRAFT and evaluate it on a novel fixpoint-based neural network architecture (monDEQ) that is particularly challenging to verify. Our extensive evaluation demonstrates that CRAFT exceeds the state-of-the-art performance in terms of speed (two orders of magnitude), scalability (one order of magnitude), and precision (25% higher certified accuracies)."
                },
                {
                    "title": "Manipulating SGD with Data Ordering Attacks",
                    "abstract": "Machine learning is vulnerable to a wide variety of attacks. It is now well understood that by changing the underlying data distribution, an adversary can poison the model trained with it or introduce backdoors. In this paper we present a novel class of training-time attacks that require no changes to the underlying dataset or model architecture, but instead only change the order in which data are supplied to the model. In particular, we find that the attacker can either prevent the model from learning, or poison it to learn behaviours specified by the attacker. Furthermore, we find that even a single adversarially-ordered epoch can be enough to slow down model learning, or even to reset all of the learning progress. Indeed, the attacks presented here are not specific to the model or dataset, but rather target the stochastic nature of modern learning procedures. We extensively evaluate our attacks on computer vision and natural language benchmarks to find that the adversary can disrupt model training and even introduce backdoors."
                },
                {
                    "title": "Accounting for Variance in Machine Learning Benchmarks",
                    "abstract": "Strong empirical evidence that one machine-learning algorithm A outperforms another one B ideally calls for multiple trials optimizing the learning pipeline over sources of variation such as data sampling, data augmentation, parameter initialization, and hyperparameters choices. This is prohibitively expensive, and corners are cut to reach conclusions. We model the whole benchmarking process, revealing that variance due to data sampling, parameter initialization and hyperparameter choice impact markedly the results. We analyze the predominant comparison methods used today in the light of this variance. We show a counter-intuitive result that adding more sources of variation to an imperfect estimator approaches better the ideal estimator at a 51 times reduction in compute cost. Building on these results, we study the error rate of detecting improvements, on five different deep-learning tasks/architectures. This study leads us to propose recommendations for performance comparisons."
                },
                {
                    "title": "Hyperparameter Optimization Is Deceiving Us, and How to Stop It",
                    "abstract": "Recent empirical work shows that inconsistent results based on choice of hyperparameter optimization (HPO) configuration are a widespread problem in ML research. When comparing two algorithms J and K searching one subspace can yield the conclusion that J outperforms K, whereas searching another can entail the opposite. In short, the way we choose hyperparameters can deceive us. We provide a theoretical complement to this prior work, arguing that, to avoid such deception, the process of drawing conclusions from HPO should be made more rigorous. We call this process epistemic hyperparameter optimization (EHPO), and put forth a logical framework to capture its semantics and how it can lead to inconsistent conclusions about performance. Our framework enables us to prove EHPO methods that are guaranteed to be defended against deception, given bounded compute time budget t. We demonstrate our framework's utility by proving and empirically validating a defended variant of random search."
                },
                {
                    "title": "Data-Driven Reachability Analysis Using Matrix Zonotopes",
                    "abstract": "In this paper, we propose a data-driven reachability analysis approach for an unknown control system. Reachability analysis is an essential tool for guaranteeing safety properties. However, most current reachability analysis heavily relies on the existence of a suitable system model, which is often not directly available in practice. We instead propose a reachability analysis approach based on noisy data. More specifically, we first provide an algorithm for over-approximating the reachable set of a linear time-invariant system using matrix zonotopes. Then we introduce an extension for nonlinear systems. We provide theoretical guarantees in both cases. Numerical examples show the potential and applicability of the introduced methods."
                },
                {
                    "title": "Underspecification Presents Challenges for Credibility in Modern Machine Learning",
                    "abstract": "ML models often exhibit unexpectedly poor behavior when they are deployed in real-world domains. We identify underspecification as a key reason for these failures. An ML pipeline is underspecified when it can return many predictors with equivalently strong held-out performance in the training domain. Underspecification is common in modern ML pipelines, such as those based on deep learning. Predictors returned by underspecified pipelines are often treated as equivalent based on their training domain performance, but we show here that such predictors can behave very differently in deployment domains. This ambiguity can lead to instability and poor model behavior in practice, and is a distinct failure mode from previously identified issues arising from structural mismatch between training and deployment domains. We show that this problem appears in a wide variety of practical ML pipelines, using examples from computer vision, medical imaging, natural language processing, clinical risk prediction based on electronic health records, and medical genomics. Our results show the need to explicitly account for underspecification in modeling pipelines that are intended for real-world deployment in any domain."
                },
                {
                    "title": "Data Poisoning Attacks on Regression Learning and Corresponding Defenses",
                    "abstract": "Adversarial data poisoning is an effective attack against machine learning and threatens model integrity by introducing poisoned data into the training dataset. So far, it has been studied mostly for classification, even though regression learning is used in many mission critical systems (such as dosage of medication, control of cyber-physical systems and managing power supply). Therefore, in the present research, we aim to evaluate all aspects of data poisoning attacks on regression learning, exceeding previous work both in terms of breadth and depth. We present realistic scenarios in which data poisoning attacks threaten production systems and introduce a novel black-box attack, which is then applied to a real-word medical use-case. As a result, we observe that the mean squared error (MSE) of the regressor increases to 150 percent due to inserting only two percent of poison samples. Finally, we present a new defense strategy against the novel and previous attacks and evaluate it thoroughly on 26 datasets. As a result of the conducted experiments, we conclude that the proposed defence strategy effectively mitigates the considered attacks."
                },
                {
                    "title": "Intrinsic Certified Robustness of Bagging against Data Poisoning Attacks",
                    "abstract": "In a data poisoning attack, an attacker modifies, deletes, and/or inserts some training examples to corrupt the learnt machine learning model. Bootstrap Aggregating (bagging) is a well known ensemble learning method, which trains multiple base models on random subsamples of a training dataset using a base learning algorithm and uses majority vote to predict labels of testing examples. We prove the intrinsic certified robustness of bagging against data poisoning attacks. Specifically, we show that bagging with an arbitrary base learning algorithm provably predicts the same label for a testing example when the number of modified, deleted, and/or inserted training examples is bounded by a threshold. Moreover, we show that our derived threshold is tight if no assumptions on the base learning algorithm are made. We evaluate our method on MNIST and CIFAR10. For instance, our method achieves a certified accuracy of 91.1% on MNIST when arbitrarily modifying, deleting, and/or inserting 100 training examples. Code is available at: https://github.com/jjy1994/BaggingCertifyDataPoisoning."
                },
                {
                    "title": "Nearest Neighbor Classifiers over Incomplete Information: From Certain Answers to Certain Predictions",
                    "abstract": "\n Machine learning (ML) applications have been thriving recently, largely attributed to the increasing availability of data. However, inconsistency and incomplete information are ubiquitous in real-world datasets, and their impact on ML applications remains elusive. In this paper, we present a formal study of this impact by extending the notion of\n Certain Answers for Codd tables\n , which has been explored by the database research community for decades, into the field of machine learning. Specifically, we focus on classification problems and propose the notion of\n \"Certain Predictions\" (CP)\n --- a test data example can be\n certainly predicted (CP'ed)\n if\n all\n possible classifiers trained on top of all possible worlds induced by the incompleteness of data would yield the\n same\n prediction. We study two fundamental CP queries: (Q1)\n checking query\n that determines whether a data example can be CP'ed; and (Q2)\n counting query\n that computes the number of classifiers that support a particular prediction (i.e., label). Given that general solutions to CP queries are, not surprisingly, hard without assumption over the type of classifier, we further present a case study in the context of nearest neighbor (NN) classifiers, where efficient solutions to CP queries can be developed --- we show that it is possible to answer both queries in\n linear or polynomial time\n over\n exponentially\n many possible worlds. We demonstrate one example use case of CP in the important application of \"data cleaning for machine learning (DC for ML).\" We show that our proposed CPClean approach built based on CP can often significantly outperform existing techniques, particularly on datasets with systematic missing values. For example, on 5 datasets with systematic missingness, CPClean (with early termination) closes 100% gap on average by cleaning 36% of dirty data on average, while the best automatic cleaning approach BoostClean can only close 14% gap on average.\n"
                },
                {
                    "title": "Robustness Certification of Generative Models",
                    "abstract": "Generative neural networks can be used to specify continuous transformations between images via latent-space interpolation. However, certifying that all images captured by the resulting path in the image manifold satisfy a given property can be very challenging. This is because this set is highly non-convex, thwarting existing scalable robustness analysis methods, which are often based on convex relaxations. We present ApproxLine, a scalable certification method that successfully verifies non-trivial specifications involving generative models and classifiers. ApproxLine can provide both sound deterministic and probabilistic guarantees, by capturing either infinite non-convex sets of neural network activation vectors or distributions over such sets. We show that ApproxLine is practically useful and can verify interesting interpolations in the networks latent space."
                },
                {
                    "title": "Learning Certified Individually Fair Representations",
                    "abstract": "To effectively enforce fairness constraints one needs to define an appropriate notion of fairness and employ representation learning in order to impose this notion without compromising downstream utility for the data consumer. A desirable notion is individual fairness as it guarantees similar treatment for similar individuals. In this work, we introduce the first method which generalizes individual fairness to rich similarity notions via logical constraints while also enabling data consumers to obtain fairness certificates for their models. The key idea is to learn a representation that provably maps similar individuals to latent representations at most $\\epsilon$ apart in $\\ell_{\\infty}$-distance, enabling data consumers to certify individual fairness by proving $\\epsilon$-robustness of their classifier. Our experimental evaluation on six real-world datasets and a wide range of fairness constraints demonstrates that our approach is expressive enough to capture similarity notions beyond existing distance metrics while scaling to realistic use cases."
                },
                {
                    "title": "Predictive inference is free with the jackknife+-after-bootstrap",
                    "abstract": "Ensemble learning is widely used in applications to make predictions in complex decision problems---for example, averaging models fitted to a sequence of samples bootstrapped from the available training data. While such methods offer more accurate, stable, and robust predictions and model estimates, much less is known about how to perform valid, assumption-lean inference on the output of these types of procedures. In this paper, we propose the jackknife+-after-bootstrap (J+aB), a procedure for constructing a predictive interval, which uses only the available bootstrapped samples and their corresponding fitted models, and is therefore \"free\" in terms of the cost of model fitting. The J+aB offers a predictive coverage guarantee that holds with no assumptions on the distribution of the data, the nature of the fitted model, or the way in which the ensemble of models are aggregated---at worst, the failure rate of the predictive interval is inflated by a factor of 2. Our numerical experiments verify the coverage and accuracy of the resulting predictive intervals on real data."
                },
                {
                    "title": "Certified Robustness to Label-Flipping Attacks via Randomized Smoothing",
                    "abstract": "Machine learning algorithms are known to be susceptible to data poisoning attacks, where an adversary manipulates the training data to degrade performance of the resulting classifier. In this work, we present a unifying view of randomized smoothing over arbitrary functions, and we leverage this novel characterization to propose a new strategy for building classifiers that are pointwise-certifiably robust to general data poisoning attacks. As a specific instantiation, we utilize our framework to build linear classifiers that are robust to a strong variant of label flipping, where each test example is targeted independently. In other words, for each test point, our classifier includes a certification that its prediction would be the same had some number of training labels been changed adversarially. Randomized smoothing has previously been used to guarantee---with high probability---test-time robustness to adversarial manipulation of the input to a classifier; we derive a variant which provides a deterministic, analytical bound, sidestepping the probabilistic certificates that traditionally result from the sampling subprocedure. Further, we obtain these certified bounds with minimal additional runtime complexity over standard classification and no assumptions on the train or test distributions. We generalize our results to the multi-class case, providing the first multi-class classification algorithm that is certifiably robust to label-flipping attacks."
                },
                {
                    "title": "Recent Advances in Algorithmic High-Dimensional Robust Statistics",
                    "abstract": "Learning in the presence of outliers is a fundamental problem in statistics. Until recently, all known efficient unsupervised learning algorithms were very sensitive to outliers in high dimensions. In particular, even for the task of robust mean estimation under natural distributional assumptions, no efficient algorithm was known. Recent work in theoretical computer science gave the first efficient robust estimators for a number of fundamental statistical tasks, including mean and covariance estimation. Since then, there has been a flurry of research activity on algorithmic high-dimensional robust estimation in a range of settings. In this survey article, we introduce the core ideas and algorithmic techniques in the emerging area of algorithmic high-dimensional robust statistics with a focus on robust mean estimation. We also provide an overview of the approaches that have led to computationally efficient robust estimators for a range of broader statistical tasks and discuss new directions and opportunities for future work."
                },
                {
                    "title": "Predictive Multiplicity in Classification",
                    "abstract": "Prediction problems often admit competing models that perform almost equally well. This effect challenges key assumptions in machine learning when competing models assign conflicting predictions. In this paper, we define predictive multiplicity as the ability of a prediction problem to admit competing models with conflicting predictions. We introduce formal measures to evaluate the severity of predictive multiplicity and develop integer programming tools to compute them exactly for linear classification problems. We apply our tools to measure predictive multiplicity in recidivism prediction problems. Our results show that real-world datasets may admit competing models that assign wildly conflicting predictions, and motivate the need to measure and report predictive multiplicity in model development."
                },
                {
                    "title": "What to Expect of Classifiers? Reasoning about Logistic Regression with Missing Features",
                    "abstract": "While discriminative classifiers often yield strong predictive performance, missing feature values at prediction time can still be a challenge. Classifiers may not behave as expected under certain ways of substituting the missing values, since they inherently make assumptions about the data distribution they were trained on. In this paper, we propose a novel framework that classifies examples with missing features by computing the expected prediction with respect to a feature distribution. Moreover, we use geometric programming to learn a naive Bayes distribution that embeds a given logistic regression classifier and can efficiently take its expected predictions. Empirical evaluations show that our model achieves the same performance as the logistic regression with all features observed, and outperforms standard imputation techniques when features go missing during prediction time. Furthermore, we demonstrate that our method can be used to generate ``sufficient explanations'' of logistic regression classifications, by removing features that do not affect the classification."
                },
                {
                    "title": "Sparse Polynomial Zonotopes: A Novel Set Representation for Reachability Analysis",
                    "abstract": "We introduce sparse polynomial zonotopes, a new set representation for formal verification of hybrid systems. Sparse polynomial zonotopes can represent nonconvex sets and are generalizations of zonotopes, polytopes, and Taylor models. Operations like Minkowski sum, quadratic mapping, and reduction of the representation size can be computed with polynomial complexity w.r.t. the dimension of the system. In particular, for reachability analysis of nonlinear systems, the wrapping effect is substantially reduced using sparse polynomial zonotopes, as demonstrated by numerical examples. In addition, we can significantly reduce the computation time compared to zonotopes when dealing with nonlinear dynamics."
                },
                {
                    "title": "Differentiable Abstract Interpretation for Provably Robust Neural Networks",
                    "abstract": "We introduce a scalable method for training robust neural networks based on abstract interpretation. We present several abstract transformers which balance ef\ufb01ciency with precision and show these can be used to train large neural networks that are certi\ufb01ably robust to adversarial perturbations."
                },
                {
                    "title": "AI2: Safety and Robustness Certification of Neural Networks with Abstract Interpretation",
                    "abstract": "We present AI2, the first sound and scalable analyzer for deep neural networks. Based on overapproximation, AI2 can automatically prove safety properties (e.g., robustness) of realistic neural networks (e.g., convolutional neural networks). The key insight behind AI2 is to phrase reasoning about safety and robustness of neural networks in terms of classic abstract interpretation, enabling us to leverage decades of advances in that area. Concretely, we introduce abstract transformers that capture the behavior of fully connected and convolutional neural network layers with rectified linear unit activations (ReLU), as well as max pooling layers. This allows us to handle real-world neural networks, which are often built out of those types of layers. We present a complete implementation of AI2 together with an extensive evaluation on 20 neural networks. Our results demonstrate that: (i) AI2 is precise enough to prove useful specifications (e.g., robustness), (ii) AI2 can be used to certify the effectiveness of state-of-the-art defenses for neural networks, (iii) AI2 is significantly faster than existing analyzers based on symbolic analysis, which often take hours to verify simple fully connected networks, and (iv) AI2 can handle deep convolutional networks, which are beyond the reach of existing methods."
                },
                {
                    "title": "Manipulating Machine Learning: Poisoning Attacks and Countermeasures for Regression Learning",
                    "abstract": "As machine learning becomes widely used for automated decisions, attackers have strong incentives to manipulate the results and models generated by machine learning algorithms. In this paper, we perform the first systematic study of poisoning attacks and their countermeasures for linear regression models. In poisoning attacks, attackers deliberately influence the training data to manipulate the results of a predictive model. We propose a theoretically-grounded optimization framework specifically designed for linear regression and demonstrate its effectiveness on a range of datasets and models. We also introduce a fast statistical attack that requires limited knowledge of the training process. Finally, we design a new principled defense method that is highly resilient against all poisoning attacks. We provide formal guarantees about its convergence and an upper bound on the effect of poisoning attacks when the defense is deployed. We evaluate extensively our attacks and defenses on three realistic datasets from health care, loan assessment, and real estate domains."
                },
                {
                    "title": "Training Set Debugging Using Trusted Items",
                    "abstract": "\n \n Training set bugs are flaws in the data that adversely affect machine learning. The training set is usually too large for manual inspection, but one may have the resources to verify a few trusted items. The set of trusted items may not by itself be adequate for learning, so we propose an algorithm that uses these items to identify bugs in the training set and thus improves learning. Specifically, our approach seeks the smallest set of changes to the training set labels such that the model learned from this corrected training set predicts labels of the trusted items correctly. We flag the items whose labels are changed as potential bugs, whose labels can be checked for veracity by human experts. To find the bugs in this way is a challenging combinatorial bilevel optimization problem, but it can be relaxed into a continuous optimization problem.Experiments on toy and real data demonstrate that our approach can identify training set bugs effectively and suggest appropriate changes to the labels. Our algorithm is a step toward trustworthy machine learning.\n \n"
                },
                {
                    "title": "Methods for order reduction of zonotopes",
                    "abstract": "Zonotopes are a special subclass of polytopes, which have several favorable properties: They can be represented in a compact way and they are closed under the Minkowski sum as well as under linear transformations. Zono-topes are a popular set representation used e.g. for reachability analysis of dynamic systems, set-based observers and robust control. The complexity of algorithms that work on zonotopes strongly depends on their order (i.e. their number of generators and dimensions), which is often increased by operations like the Minkowski sum. Thus, to keep computations efficient, zonotopes of high orders are often over-approximated as tight as possible by zonotopes of smaller order. This paper has two main contributions: First, we propose new methods based on principle component analysis (PCA), clustering and constrained optimization for tight over-approximation of zonotopes. Second, we provide an overview of the most important known methods for order reduction and compare the performance of new and known methods in low- and high-dimensional spaces."
                },
                {
                    "title": "Certified Defenses for Data Poisoning Attacks",
                    "abstract": "Machine learning systems trained on user-provided data are susceptible to data poisoning attacks, whereby malicious users inject false training data with the aim of corrupting the learned model. While recent work has proposed a number of attacks and defenses, little is understood about the worst-case loss of a defense in the face of a determined attacker. We address this by constructing approximate upper bounds on the loss across a broad family of attacks, for defenders that first perform outlier removal followed by empirical risk minimization. Our approximation relies on two assumptions: (1) that the dataset is large enough for statistical concentration between train and test error to hold, and (2) that outliers within the clean (non-poisoned) data do not have a strong effect on the model. Our bound comes paired with a candidate attack that often nearly matches the upper bound, giving us a powerful tool for quickly assessing defenses on a given dataset. Empirically, we find that even under a simple defense, the MNIST-1-7 and Dogfish datasets are resilient to attack, while in contrast the IMDB sentiment dataset can be driven from 12% to 23% test error by adding only 3% poisoned data."
                },
                {
                    "title": "SymPy: Symbolic computing in Python",
                    "abstract": "SymPy is an open source computer algebra system written in pure Python. It is built with a focus on extensibility and ease of use, through both interactive and programmatic applications. These characteristics have led SymPy to become the standard symbolic library for the scientific Python ecosystem. This paper presents the architecture of SymPy, a description of its features, and a discussion of select domain specific submodules. The supplementary materials provide additional examples and further outline details of the architecture and features of SymPy."
                },
                {
                    "title": "Distributionally Robust Logistic Regression",
                    "abstract": "This paper proposes a distributionally robust approach to logistic regression. We use the Wasserstein distance to construct a ball in the space of probability distributions centered at the uniform distribution on the training samples. If the radius of this ball is chosen judiciously, we can guarantee that it contains the unknown data-generating distribution with high confidence. We then formulate a distribution-ally robust logistic regression model that minimizes a worst-case expected logloss function, where the worst case is taken over all distributions in the Wasserstein ball. We prove that this optimization problem admits a tractable reformulation and encapsulates the classical as well as the popular regularized logistic regression problems as special cases. We further propose a distributionally robust approach based on Wasserstein balls to compute upper and lower confidence bounds on the misclassification probability of the resulting classifier. These bounds are given by the optimal values of two highly tractable linear programs. We validate our theoretical out-of-sample guarantees through simulated and empirical experiments."
                },
                {
                    "title": "Reachability analysis of nonlinear systems using conservative polynomialization and non-convex sets",
                    "abstract": "A new technique for computing the reachable set of hybrid systems with nonlinear continuous dynamics is presented. Previous work showed that abstracting the nonlinear continuous dynamics to linear differential inclusions results in a scalable approach for reachability analysis. However, when the abstraction becomes inaccurate, linearization techniques require splitting of reachable sets, resulting in an exponential growth of required linearizations. In this work, the nonlinearity of the dynamics is more accurately abstracted to polynomial difference inclusions. As a consequence, it is no longer guaranteed that reachable sets of consecutive time steps are mapped to convex sets as typically used in previous works. Thus, a non-convex set representation is developed in order to better capture the nonlinear dynamics, requiring no or much less splitting. The new approach has polynomial complexity with respect to the number of continuous state variables when splitting can be avoided and is thus promising when a linearization technique requires splitting for the same problem. The benefits are presented by numerical examples."
                },
                {
                    "title": "Practical Bayesian Optimization of Machine Learning Algorithms",
                    "abstract": "The use of machine learning algorithms frequently involves careful tuning of learning parameters and model hyperparameters. Unfortunately, this tuning is often a \"black art\" requiring expert experience, rules of thumb, or sometimes brute-force search. There is therefore great appeal for automatic approaches that can optimize the performance of any given learning algorithm to the problem at hand. In this work, we consider this problem through the framework of Bayesian optimization, in which a learning algorithm's generalization performance is modeled as a sample from a Gaussian process (GP). We show that certain choices for the nature of the GP, such as the type of kernel and the treatment of its hyperparameters, can play a crucial role in obtaining a good optimizer that can achieve expertlevel performance. We describe new algorithms that take into account the variable cost (duration) of learning algorithm experiments and that can leverage the presence of multiple cores for parallel experimentation. We show that these proposed algorithms improve on previous automatic procedures and can reach or surpass human expert-level optimization for many algorithms including latent Dirichlet allocation, structured SVMs and convolutional neural networks."
                },
                {
                    "title": "Zonotope bundles for the efficient computation of reachable sets",
                    "abstract": "We introduce zonotope bundles for computing the set of states reachable by a dynamical system, also known as the reachable set. Reachable set computations suffer from the curse of dimensionality, which has been successfully addressed by using zonotopes for linear systems. However, zonotopes are not closed under intersection leading to challenges when applying them to nonlinear and hybrid problems. We introduce zonotope bundles as the intersection of zonotopes (without explicitly computing the intersection). Zonotope bundles are closed under intersection, while inheriting many positive properties of zonotopes. This is demonstrated for linear, nonlinear, and hybrid systems. A further property of zonotope bundles is that their computation can be easily parallelized."
                },
                {
                    "title": "Robust Solutions of Optimization Problems Affected by Uncertain Probabilities",
                    "abstract": "In this paper we focus on robust linear optimization problems with uncertainty regions defined by \u03c6-divergences for example, chi-squared, Hellinger, Kullback--Leibler. We show how uncertainty regions based on \u03c6-divergences arise in a natural way as confidence sets if the uncertain parameters contain elements of a probability vector. Such problems frequently occur in, for example, optimization problems in inventory control or finance that involve terms containing moments of random variables, expected utility, etc. We show that the robust counterpart of a linear optimization problem with \u03c6-divergence uncertainty is tractable for most of the choices of \u03c6 typically considered in the literature. We extend the results to problems that are nonlinear in the optimization variables. Several applications, including an asset pricing example and a numerical multi-item newsvendor example, illustrate the relevance of the proposed approach. \n \nThis paper was accepted by Gerard P. Cachon, optimization."
                },
                {
                    "title": "Conditional Prediction Intervals for Linear Regression",
                    "abstract": "We construct prediction intervals for the linear regression model with IID errors with a known distribution, not necessarily Gaussian. The coverage probability of our prediction intervals is equal to the nominal confidence level not only unconditionally but also conditionally given a natural sigma-algebra of invariant events. This implies, in particular, the perfect calibration of our prediction intervals in the on-line mode of prediction."
                },
                {
                    "title": "The octagon abstract domain",
                    "abstract": "This article presents the octagon abstract domain, a relational numerical abstract domain for static analysis by abstract interpretation. It allows representing conjunctions of constraints of the form \u00b1 X \u00b1 Y \u2264 c where X and Y range among program variables and c is a constant in \u2124, \u211a, or \u211d automatically inferred. Abstract elements are represented using modified Difference Bound Matrices and we use a normalization algorithm loosely based on the shortest-path closure to compute canonical representations and construct best-precision abstract transfer functions. We achieve a quadratic memory cost per abstract element and a cubic worst-case time cost per abstract operation, with respect to the number of program variables.In terms of cost and precision, our domain is in between the well-known fast but imprecise interval domain and the costly polyhedron domain. We show that it is precise enough to treat interesting examples requiring relational invariants, and hence, out of the reach of the interval domain. We also present a packing strategy that allows scaling our domain up to large programs by tuning the amount of relationality. The octagon domain was incorporated into the ASTR\u00c9E industrial-strength static analyzer and was key in proving the absence of run-time errors in large critical embedded flight control software for Airbus planes."
                },
                {
                    "title": "Ridge Regression: Biased Estimation for Nonorthogonal Problems",
                    "abstract": "In multiple regression it is shown that parameter estimates based on minimum residual sum of squares have a high probability of being unsatisfactory, if not incorrect, if the prediction vectors are not orthogonal. Proposed is an estimation procedure based on adding small positive quantities to the diagonal of X\u2032X. Introduced is the ridge trace, a method for showing in two dimensions the effects of nonorthogonality. It is then shown how to augment X\u2032X to obtain biased estimates with smaller mean square error."
                },
                {
                    "title": "Heuristics of instability and stabilization in model selection",
                    "abstract": "In model selection, usually a best predictor is chosen from a collection {\u03bc(.,s)} of predictors where \u03bc(.,s) is the minimum least-squares predictor in a collection U s of predictors. Here s is a complexity parameter; that is, the smaller s, the lower dimensional/smoother the models in U s . If L is the data used to derive the sequence {\u03bc(., s)}, the procedure is called unstable if a small change in L can cause large changes in {\u03bc(., s)}. With a crystal ball, one could pick the predictor in {\u03bc(.,s)} having minimum prediction error. Without prescience, one uses test sets, cross-validation and so forth. The difference in prediction error between the crystal ball selection and the statistician's choice we call predictive loss. For an unstable procedure the predictive loss is large. This is shown by some analytics in a simple case and by simulation results in a more complex comparison of four different linear regression methods. Unstable procedures can be stabilized by perturbing the data, getting a new predictor sequence {\u03bc(., s)} and then averaging over many such predictor sequences."
                },
                {
                    "title": "Multiple Imputation After 18+ Years",
                    "abstract": "Abstract Multiple imputation was designed to handle the problem of missing data in public-use data bases where the data-base constructor and the ultimate user are distinct entities. The objective is valid frequency inference for ultimate users who in general have access only to complete-data software and possess limited knowledge of specific reasons and models for nonresponse. For this situation and objective, I believe that multiple imputation by the data-base constructor is the method of choice. This article first provides a description of the assumed context and objectives, and second, reviews the multiple imputation framework and its standard results. These preliminary discussions are especially important because some recent commentaries on multiple imputation have reflected either misunderstandings of the practical objectives of multiple imputation or misunderstandings of fundamental theoretical results. Then, criticisms of multiple imputation are considered, and, finally, comparisons are made to alt..."
                },
                {
                    "title": "Reasoning about Action Using a Possible Models Approach",
                    "abstract": "Ginsberg and Smith [6, 7] propose a new method for reasoning about action, which they term a possible worlds approach (PWA). The PWA is an elegant, simple, and potentially very powerful domain-independent technique that has proven fruitful in other areas of AI [13, 5]. In the domain of reasoning about action, Ginsberg and Smith offer the PWA as a solution to the frame problem (What facts about the world remain true when an action is performed?) and its dual, the ramification problem [3] (What facts about the world must change when an action is performed?). In addition, Ginsberg and Smith offer the PWA as a solution to the qualification problem (When is it reasonable to assume that an action will succeed?), and claim for the PWA computational advantages over other approaches such as situation calculus. \n \nHere and in [16] I show that the PWA fails to solve the frame, ramification, and qualification problems, even with additional simplifying restrictions not imposed by Ginsberg and Smith. The cause of the failure seems to be a lack of distinction in the PWA between the state of the world and the description of the state of the world. I introduce a new approach to reasoning about action, called the possible models approach, and show that the possible models approach works as well as the PWA on the examples of [6, 7] but does not suffer from its deficiencies."
                },
                {
                    "title": "Incomplete Information in Relational Databases",
                    "abstract": "ABSTRACT This paper concerns the semantics of Codd's relational model of data. Formulated are precise conditions that should be satisfied in a semantically meaningful extension of the usual relational operators, such as projection, selection, union, and join, from operators on relations to operators on tables with \u201cnull values\u201d of various kinds allowed. These conditions require that the system be safe in the sense that no incorrect conclusion is derivable by using a specified subset \u03a9 of the relational operators; and that it be complete in the sense that all valid conclusions expressible by relational expressions using operators in \u03a9 are in fact derivable in this system. Two such systems of practical interest are shown. The first, based on the usual Codd's null values, supports projection and selection. The second, based on many different (\u201cmarked\u201d) null values or variables allowed to appear in a table, is shown to correctly support projection, positive selection (with no negation occurring in the selection condition), union, and renaming of attributes, which allows for processing arbitrary conjunctive queries. A very desirable property enjoyed by this system is that all relational operators on tables are performed in exactly the same way as in the case of the usual relations. A third system, mainly of theoretical interest, supporting projection, selection, union, join, and renaming, is also discussed. Under a so-called closed world assumption, it can also handle the operator of difference. It is based on a device called a conditional table and is crucial to the proof of the correctness of the second system. All systems considered allow for relational expressions containing arbitrarily many different relation symbols, and no form of the universal relation assumption is required. Categories and Subject Descriptors: H.2.3 [Database Management]: Languages\u2014 query languages; H.2.4 [Database Management]: Systems\u2014 query processing General Terms: Theory"
                },
                {
                    "title": "Overcoming Data Biases: Towards Enhanced Accuracy and Reliability in Machine Learning",
                    "abstract": "The pervasive integration of machine learning (ML) across various sectors has underscored the critical challenge of addressing inherent biases in ML models. These biases not only undermine the models\u2019 fairness and accuracy but also have signi\ufb01cant real-world consequences. Traditional approaches to mitigating these biases often fail to address their root causes, leading to solutions that may super\ufb01cially seem fair but do not tackle the underlying problems. This review paper explores the role of causal modeling in enhancing data cleaning, preparation, and quality management for ML. By analyzing existing research, we demonstrate how causal reasoning can effectively identify and rectify data biases, thus improving the fairness and accuracy of ML models. We advocate for the increased adoption of causal approaches in these processes, emphasizing their potential to signi\ufb01cantly enhance the integrity and reliability of data-driven technologies."
                },
                {
                    "title": "Fast and Effective Robustness Certification",
                    "abstract": "We present a new method and system, called DeepZ, for certifying neural network robustness based on abstract interpretation. Compared to state-of-the-art automated verifiers for neural networks, DeepZ: (i) handles ReLU, Tanh and Sigmoid activation functions, (ii) supports feedforward and convolutional architectures, (iii) is significantly more scalable and precise, and (iv) and is sound with respect to floating point arithmetic. These benefits are due to carefully designed approximations tailored to the setting of neural networks. As an example, DeepZ achieves a verification accuracy of 97% on a large network with 88,500 hidden units under $L_{\\infty}$ attack with $\\epsilon = 0.1$ with an average runtime of 133 seconds."
                },
                {
                    "title": "Abstract Interpretation",
                    "abstract": "Interpretation I\u015f\u0131l Dillig I\u015f\u0131l Dillig, Abstract Interpretation 1/27 Overview I Deductive verifiers require annotations (e.g., loop invariants) from user I Fortunately, many techniques that can automatically learn loop invariants I A common framework for this purpose is Abstract Interpretation (AI) I Abstract interpretation forms the basis of most static analyzers I\u015f\u0131l Dillig, Abstract Interpretation 2/27 Key Idea: Over-approximation I Abstract interpretation is a framework for computing over-approximations of program states I Cannot reason about the exact program behavior due to undecidability (and also for scalability reasons) I But we can obtain a conservative over-approximation and this can be enough to prove program correctness I\u015f\u0131l Dillig, Abstract Interpretation 3/27 Motivating Example I What does this function do? I Annotations computed automatically using an AI tool (Apron) I\u015f\u0131l Dillig, Abstract Interpretation 4/27 The AI Recipe Abstract interpretation provides a recipe for computing over-approximations of program behavior 1. Define abstract domain \u2013 fixes \u201cshape\u201d of the invariants I e.g., c1 \u2264 x \u2264 c2 (intervals) or \u00b1x \u00b1 y \u2264 c (octagons) 2. Define abstract semantics (transformers) I Define how to symbolically execute each statement in the chosen abstract domain I Must be sound wrt to concrete semantics 3. Iterate abstract transformers until fixed point I The fixed-point is an over-approximation of program behavior I\u015f\u0131l Dillig, Abstract Interpretation 5/27 Simple Example: Sign Domain I Suppose we want to infer invariants of the form x on 0 where on\u2208 {\u2265,=, >,<} (i.e., zero, non-negative, positive, negative) I This corresponds to the following abstract domain represented as lattice: non-neg neg pos zero Each element in this lattice is an \"abstract value\" I Lattice is a partially ordered set (S ,v) where each pair of elements has a least upper bound or join (t) I\u015f\u0131l Dillig, Abstract Interpretation 6/27"
                },
                {
                    "title": "Stochastic Gradient Methods for Distributionally Robust Optimization with f-divergences",
                    "abstract": "We develop efficient solution methods for a robust empirical risk minimization problem designed to give calibrated confidence intervals on performance and provide optimal tradeoffs between bias and variance. Our methods apply to distributionally robust optimization problems proposed by Ben-Tal et al., which put more weight on observations inducing high loss via a worst-case approach over a non-parametric uncertainty set on the underlying data distribution. Our algorithm solves the resulting minimax problems with nearly the same computational cost of stochastic gradient descent through the use of several carefully designed data structures. For a sample of size n, the per-iteration cost of our method scales as O(log n), which allows us to give optimality certificates that distributionally robust optimization provides at little extra cost compared to empirical risk minimization and stochastic gradient methods."
                },
                {
                    "title": "Reachability Analysis and its Application to the Safety Assessment of Autonomous Cars",
                    "abstract": "This thesis is about the safety verification of dynamical systems using reachability analysis. Novel solutions have been developed for classical reachability analysis, stochastic reachability analysis, and their application to the safety assessment of autonomous cars. \n \nClassical reachability analysis computes the set of states that can be reached by a system. If the reachable set does not intersect any set of unsafe states, the safety of the system is guaranteed. Algorithms for this problem have been developed for linear, nonlinear, and hybrid systems. \n \nStochastic reachability analysis measures the probability of reaching an unsafe set. One pursued approach computes over-approximative solutions for linear systems; another one generates a Markov chain which approximately computes the stochastic reachable set of arbitrary dynamics."
                },
                {
                    "title": "MULTIPLE IMPUTATIONS IN SAMPLE SURVEYS-A PHENOMENOLOGICAL BAYESIAN APPROACH TO NONRESPONSE",
                    "abstract": "A general attack on the problem of nonresponse in sample surveys is outlined from the phenomenological Bayesian perspective. The objective is to develop procedures that are useful in practice. The plan is to impute several values for each missing datum, where the imputed values reflect variation within an imputation model and sensitivity to different imputation models. The analysis of a resultant multi-imputed data set is viewed as simulating predictive distributions of desired summary statistics under imputation models. Three tasks are defined that are needed to create the imputations: the imputation task, the estimation task and the modelling task. The imputation task and estimation task are technical in nature. The modelling task requires the development of new tools appropriate for relating nonrespondents and respondents."
                },
                {
                    "title": "INTERPRETATION : \u2018A UNIFIED LATTICE MODEL FOR STATIC ANALYSIS OF PROGRAMS BY CONSTRUCTION OR APPROXIMATION",
                    "abstract": "ABSTRACT"
                }
            ],
            "categories": [
                "cs.LG",
                "cs.DB",
                "cs.SC"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively learn from uncertain datasets that exhibit variations due to missing entries, errors, inconsistencies, and biases?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the pervasive issue of data uncertainty, which can significantly impact the robustness, reliability, and fairness of predictive models. By developing a comprehensive framework that utilizes possible world semantics, this research could lead to advancements in statistical inference and predictive modeling, ultimately enhancing the accuracy of machine learning applications across various domains. Furthermore, it could inspire future research to explore new methodologies for managing uncertainties, thereby broadening the scope of machine learning applications in real-world scenarios.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the potentially infinite scenarios represented by uncertain datasets, making it impractical to explore every possibility and train a model for each. Naive approaches may fail because they often overlook the complexities of dataset variations and the interactions between features and labels. Additionally, the concept of model multiplicity complicates the situation, as models with similar accuracy can yield different predictions based on the underlying data variations. Overcoming these technical and theoretical obstacles requires innovative methodologies that can systematically manage and evaluate uncertainties across all features and labels.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either competing models or specific types of data variations, such as label errors in linear models, which limits their applicability to broader contexts. Existing solutions have not adequately addressed the full range of dataset variations or the implications of model multiplicity. Barriers such as the lack of a comprehensive framework for evaluating model robustness amid data uncertainties have prevented this problem from being effectively solved. Our approach differs by leveraging possible world semantics to systematically manage uncertainties across all features and labels, thus providing a more holistic solution.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves employing the framework of possible world semantics to conceptualize uncertainties in datasets. We will utilize a system called ZORRO (ZOnotope-based Robustness Analysis for Regression with Over-approximations) to train a set of models on various potential datasets. The evaluation will focus on how data uncertainties affect model robustness, reliability, and fairness. We will use metrics such as loss functions to assess model performance across different scenarios, with expected outcomes including improved understanding of model behavior under uncertainty"
            }
        },
        "author_data": {
            "0fce7e48-8e37-4fdf-8670-c57f61c32270": {
                "pk": "0fce7e48-8e37-4fdf-8670-c57f61c32270",
                "name": "Jiongli Zhu",
                "collaborators": [
                    "Babak Salimi",
                    "Sainyam Galhotra",
                    "Nazanin Sabri",
                    "Romila Pradhan",
                    "Boris Glavic"
                ],
                "domain": [
                    "Fairness in Machine Learning",
                    "Explainable AI",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "Consistent Range Approximation for Fair Predictive Modeling",
                        "abstract": "This paper proposes a novel framework for certifying the fairness of predictive models trained on biased data. It draws from query answering for incomplete and inconsistent databases to formulate the problem of consistent range approximation (CRA) of fairness queries for a predictive model on a target population. The framework employs background knowledge of the data collection process and biased data, working with or without limited statistics about the target population, to compute a range of answers for fairness queries. Using CRA, the framework builds predictive models that are certifiably fair on the target population, regardless of the availability of external data during training. The framework's efficacy is demonstrated through evaluations on real data, showing substantial improvement over existing state-of-the-art methods."
                    },
                    {
                        "title": "Interpretable Data-Based Explanations for Fairness Debugging",
                        "abstract": "A wide variety of fairness metrics and eXplainable Artificial Intelligence (XAI) approaches have been proposed in the literature to identify bias in machine learning models that are used in critical real-life contexts. However, merely reporting on a model's bias, or generating explanations using existing XAI techniques is insufficient to locate and eventually mitigate sources of bias. We introduce Gopher, a system that produces compact, interpretable and causal explanations for bias or unexpected model behavior by identifying coherent subsets of the training data that are root-causes for this behavior. Specifically, we introduce the concept of causal responsibility that quantifies the extent to which intervening on training data by removing or updating subsets of it can resolve the bias. Building on this concept, we develop an efficient approach for generating the top-k patterns that explain model bias that utilizes techniques from the machine learning (ML) community to approximate causal responsibility and uses pruning rules to manage the large search space for patterns. Our experimental evaluation demonstrates the effectiveness of Gopher in generating interpretable explanations for identifying and debugging sources of bias."
                    }
                ]
            },
            "d69fb902-cc66-4814-8f32-68ad4184dacc": {
                "pk": "d69fb902-cc66-4814-8f32-68ad4184dacc",
                "name": "Su Feng",
                "collaborators": [
                    "Boris Glavic",
                    "Oliver Kennedy",
                    "Aaron Huber"
                ],
                "domain": [
                    "Database Systems",
                    "Uncertainty Management",
                    "Data Integration",
                    "Query Processing"
                ],
                "publications": [
                    {
                        "title": "Uncertainty Annotated Databases - A Lightweight Approach for Approximating Certain Answers (extended version)",
                        "abstract": "Certain answers are a principled method for coping with uncertainty that arises in many practical data management tasks. Unfortunately, this method is expensive and may exclude useful (if uncertain) answers. Thus, users frequently resort to less principled approaches to resolve the uncertainty. In this paper, we propose Uncertainty Annotated Databases (UA-DBs), which combine an under- and over-approximation of certain answers to achieve the reliability of certain answers, with the performance of a classical database system. Furthermore, in contrast to prior work on certain answers, UA-DBs achieve a higher utility by including some (explicitly marked) answers that are not certain. UA-DBs are based on incomplete K-relations, which we introduce to generalize the classical set-based notions of incomplete databases and certain answers to a much larger class of data models. Using an implementation of our approach, we demonstrate experimentally that it efficiently produces tight approximations of certain answers that are of high utility."
                    },
                    {
                        "title": "Efficient Approximation of Certain and Possible Answers for Ranking and Window Queries over Uncertain Data (Extended version)",
                        "abstract": "Uncertainty arises naturally inmany application domains due to, e.g., data entry errors and ambiguity in data cleaning. Prior work in incomplete and probabilistic databases has investigated the semantics and efficient evaluation of ranking and top-k queries over uncertain data. However, most approaches deal with top-k and ranking in isolation and do represent uncertain input data and query results using separate, incompatible datamodels. We present an efficient approach for under- and over-approximating results of ranking, top-k, and window queries over uncertain data. Our approach integrates well with existing techniques for querying uncertain data, is efficient, and is to the best of our knowledge the first to support windowed aggregation. We design algorithms for physical operators for uncertain sorting and windowed aggregation, and implement them in PostgreSQL.We evaluated our approach on synthetic and real world datasets, demonstrating that it outperforms all competitors, and often produces more accurate results."
                    }
                ]
            },
            "60da2476-c792-47b2-8c74-8481eafa3974": {
                "pk": "60da2476-c792-47b2-8c74-8481eafa3974",
                "name": "Boris Glavic",
                "collaborators": [
                    "Oliver Kennedy",
                    "Seokki Lee",
                    "Su Feng",
                    "Xing Niu",
                    "Bertram Ludaescher",
                    "Bahareh Sadat Arab",
                    "Dieter Gawlick",
                    "Aaron Huber",
                    "Vasudha Krishnaswamy",
                    "Zhen Hua Liu"
                ],
                "domain": [
                    "Database Systems",
                    "Data Provenance",
                    "Uncertainty Management",
                    "Query Optimization"
                ],
                "publications": [
                    {
                        "title": "Snapshot Semantics for Temporal Multiset Relations (Extended Version)",
                        "abstract": "Snapshot semantics is widely used for evaluating queries over temporal data: temporal relations are seen as sequences of snapshot relations, and queries are evaluated at each snapshot. In this work, we demonstrate that current approaches for snapshot semantics over interval-timestamped multiset relations are subject to two bugs regarding snapshot aggregation and bag difference. We introduce a novel temporal data model based on K-relations that overcomes these bugs and prove it to correctly encode snapshot semantics. Furthermore, we present an efficient implementation of our model as a database middleware and demonstrate experimentally that our approach is competitive with native implementations and significantly outperforms such implementations on queries that involve aggregation."
                    },
                    {
                        "title": "PUG: A Framework and Practical Implementation for Why & Why-Not Provenance (extended version)",
                        "abstract": "Explaining why an answer is (or is not) returned by a query is important for many applications including auditing, debugging data and queries, and answering hypothetical questions about data. In this work, we present the first practical approach for answering such questions for queries with negation (first- order queries). Specifically, we introduce a graph-based provenance model that, while syntactic in nature, supports reverse reasoning and is proven to encode a wide range of provenance models from the literature. The implementation of this model in our PUG (Provenance Unification through Graphs) system takes a provenance question and Datalog query as an input and generates a Datalog program that computes an explanation, i.e., the part of the provenance that is relevant to answer the question. Furthermore, we demonstrate how a desirable factorization of provenance can be achieved by rewriting an input query. We experimentally evaluate our approach demonstrating its efficiency."
                    },
                    {
                        "title": "HRDBMS: Combining the Best of Modern and Traditional Relational Databases",
                        "abstract": "HRDBMS is a novel distributed relational database that uses a hybrid model combining the best of traditional distributed relational databases and Big Data analytics platforms such as Hive. This allows HRDBMS to leverage years worth of research regarding query optimization, while also taking advantage of the scalability of Big Data platforms. The system uses an execution framework that is tailored for relational processing, thus addressing some of the performance challenges of running SQL on top of platforms such as MapReduce and Spark. These include excessive materialization of intermediate results, lack of a global cost-based optimization, unnecessary sorting, lack of index support, no statistics, no support for DML and ACID, and excessive communication caused by the rigid communication patterns enforced by these platforms."
                    },
                    {
                        "title": "Optimizing Provenance Computations",
                        "abstract": "Data provenance is essential for debugging query results, auditing data in cloud environments, and explaining outputs of Big Data analytics. A well-established technique is to represent provenance as annotations on data and to instrument queries to propagate these annotations to produce results annotated with provenance. However, even sophisticated optimizers are often incapable of producing efficient execution plans for instrumented queries, because of their inherent complexity and unusual structure. Thus, while instrumentation enables provenance support for databases without requiring any modification to the DBMS, the performance of this approach is far from optimal. In this work, we develop provenance specific optimizations to address this problem. Specifically, we introduce algebraic equivalences targeted at instrumented queries and discuss alternative, equivalent ways of instrumenting a query for provenance capture. Furthermore, we present an extensible heuristic and cost-based optimization (CBO) framework that governs the application of these optimizations and implement this framework in our GProM provenance system. Our CBO is agnostic to the plan space shape, uses a DBMS for cost estimation, and enables retrofitting of optimization choices into existing code by adding a few LOC. Our experiments confirm that these optimizations are highly effective, often improving performance by several orders of magnitude for diverse provenance tasks."
                    },
                    {
                        "title": "Approximate Summaries for Why and Why-not Provenance (Extended Version)",
                        "abstract": "Why and why-not provenance have been studied extensively in recent years. However, why-not provenance, and to a lesser degree why provenance, can be very large resulting in severe scalability and usability challenges. In this paper, we introduce a novel approximate summarization technique for provenance which overcomes these challenges. Our approach uses patterns to encode (why-not) provenance concisely. We develop techniques for efficiently computing provenance summaries balancing informativeness, conciseness, and completeness. To achieve scalability, we integrate sampling techniques into provenance capture and summarization. Our approach is the first to scale to large datasets and to generate comprehensive and meaningful summaries."
                    },
                    {
                        "title": "Reenactment for Read-Committed Snapshot Isolation",
                        "abstract": "Provenance for transactional updates is critical for many applications such as auditing and debugging of transactions. Recently, we have introduced MV-semirings, an extension of the semiring provenance model that supports updates and transactions. Furthermore, we have proposed reenactment, a declarative form of replay with provenance capture, as an efficient and non-invasive method for computing this type of provenance. However, this approach is limited to the snapshot isolation (SI) concurrency control protocol while many real world applications apply the read committed version of snapshot isolation (RC-SI) to improve performance at the cost of consistency. We present non-trivial extensions of the model and reenactment approach to be able to compute provenance of RC-SI transactions efficiently. In addition, we develop techniques for applying reenactment across multiple RC-SI transactions. Our experiments demonstrate that our implementation in the GProM system supports efficient re-construction and querying of provenance."
                    },
                    {
                        "title": "Uncertainty Annotated Databases - A Lightweight Approach for Approximating Certain Answers (extended version)",
                        "abstract": "Certain answers are a principled method for coping with uncertainty that arises in many practical data management tasks. Unfortunately, this method is expensive and may exclude useful (if uncertain) answers. Thus, users frequently resort to less principled approaches to resolve the uncertainty. In this paper, we propose Uncertainty Annotated Databases (UA-DBs), which combine an under- and over-approximation of certain answers to achieve the reliability of certain answers, with the performance of a classical database system. Furthermore, in contrast to prior work on certain answers, UA-DBs achieve a higher utility by including some (explicitly marked) answers that are not certain. UA-DBs are based on incomplete K-relations, which we introduce to generalize the classical set-based notions of incomplete databases and certain answers to a much larger class of data models. Using an implementation of our approach, we demonstrate experimentally that it efficiently produces tight approximations of certain answers that are of high utility."
                    },
                    {
                        "title": "To not miss the forest for the trees -- a holistic approach for explaining missing answers over nested data (extended version)",
                        "abstract": "Query-based explanations for missing answers identify which operators of a query are responsible for the failure to return a missing answer of interest. This type of explanations has proven to be useful in a variety of contexts including debugging of complex analytical queries. Such queries are frequent in big data systems such as Apache Spark. We present a novel approach for producing query-based explanations. Our approach is the first to support nested data and to consider operators that modify the schema and structure of the data (e.g., nesting and projections) as potential causes of missing answers. To efficiently compute explanations, we propose a heuristic algorithm that applies two novel techniques: (i) reasoning about multiple schema alternatives for a query and (ii) re-validating at each step whether an intermediate result can contribute to the missing answer. Using an implementation of our approach on Spark, we demonstrate that it is the first to scale to large datasets and that it often finds explanations that existing techniques fail to identify."
                    },
                    {
                        "title": "Provenance-based Data Skipping (TechReport)",
                        "abstract": "Database systems analyze queries to determine upfront which data is needed for answering them and use indexes and other physical design techniques to speed-up access to that data. However, for important classes of queries, e.g., HAVING and top-k queries, it is impossible to determine up-front what data is relevant. To overcome this limitation, we develop provenance-based data skipping (PBDS), a novel approach that generates provenance sketches to concisely encode what data is relevant for a query. Once a provenance sketch has been captured it is used to speed up subsequent queries. PBDS can exploit physical design artifacts such as indexes and zone maps. Our approach significantly improves performance for both disk-based and main-memory database systems."
                    },
                    {
                        "title": "Interpretable Data-Based Explanations for Fairness Debugging",
                        "abstract": "A wide variety of fairness metrics and eXplainable Artificial Intelligence (XAI) approaches have been proposed in the literature to identify bias in machine learning models that are used in critical real-life contexts. However, merely reporting on a model's bias, or generating explanations using existing XAI techniques is insufficient to locate and eventually mitigate sources of bias. We introduce Gopher, a system that produces compact, interpretable and causal explanations for bias or unexpected model behavior by identifying coherent subsets of the training data that are root-causes for this behavior. Specifically, we introduce the concept of causal responsibility that quantifies the extent to which intervening on training data by removing or updating subsets of it can resolve the bias. Building on this concept, we develop an efficient approach for generating the top-k patterns that explain model bias that utilizes techniques from the machine learning (ML) community to approximate causal responsibility and uses pruning rules to manage the large search space for patterns. Our experimental evaluation demonstrates the effectiveness of Gopher in generating interpretable explanations for identifying and debugging sources of bias."
                    },
                    {
                        "title": "Efficient Approximation of Certain and Possible Answers for Ranking and Window Queries over Uncertain Data (Extended version)",
                        "abstract": "Uncertainty arises naturally inmany application domains due to, e.g., data entry errors and ambiguity in data cleaning. Prior work in incomplete and probabilistic databases has investigated the semantics and efficient evaluation of ranking and top-k queries over uncertain data. However, most approaches deal with top-k and ranking in isolation and do represent uncertain input data and query results using separate, incompatible datamodels. We present an efficient approach for under- and over-approximating results of ranking, top-k, and window queries over uncertain data. Our approach integrates well with existing techniques for querying uncertain data, is efficient, and is to the best of our knowledge the first to support windowed aggregation. We design algorithms for physical operators for uncertain sorting and windowed aggregation, and implement them in PostgreSQL.We evaluated our approach on synthetic and real world datasets, demonstrating that it outperforms all competitors, and often produces more accurate results."
                    },
                    {
                        "title": "The Exception that Improves the Rule",
                        "abstract": "The database community has developed numerous tools and techniques for data curation and exploration, from declarative languages, to specialized techniques for data repair, and more. Yet, there is currently no consensus on how to best expose these powerful tools to an analyst in a simple, intuitive, and above all, flexible way. Thus, analysts continue to rely on tools such as spreadsheets, imperative languages, and notebook style programming environments like Jupyter for data curation. In this work, we explore the integration of spreadsheets, notebooks, and relational databases. We focus on a key advantage that both spreadsheets and imperative notebook environments have over classical relational databases: ease of exception. By relying on set-at-a-time operations, relational databases sacrifice the ability to easily define singleton operations, exceptions to a normal data processing workflow that affect query processing for a fixed set of explicitly targeted records. In comparison, a spreadsheet user can easily change the formula for just one cell, while a notebook user can add an imperative operation to her notebook that alters an output 'view'. We believe that enabling such idiosyncratic manual transformations in a classical relational database is critical for curation, as curation operations that are easy to declare for individual values can often be extremely challenging to generalize. We explore the challenges of enabling singletons in relational databases, propose a hybrid spreadsheet/relational notebook environment for data curation, and present our vision of Vizier, a system that exposes data curation through such an interface."
                    },
                    {
                        "title": "Efficient Uncertainty Tracking for Complex Queries with Attribute-level Bounds (extended version)",
                        "abstract": "Certain answers are a principled method for coping with the uncertainty that arises in many practical data management tasks. Unfortunately, this method is expensive and may exclude useful (if uncertain) answers. Prior work introduced Uncertainty Annotated Databases (UA-DBs), which combine an under- and over-approximation of certain answers. UA-DBs combine the reliability of certain answers based on incomplete K-relations with the performance of classical deterministic database systems. However, UA-DBs only support a limited class of queries and do not support attribute-level uncertainty which can lead to inaccurate under-approximations of certain answers. In this paper, we introduce attribute-annotated uncertain databases (AU-DBs) which extend the UA-DB model with attribute-level annotations that record bounds on the values of an attribute across all possible worlds. This enables more precise approximations of incomplete databases. Furthermore, we extend UA-DBs to encode an compact over-approximation of possible answers which is necessary to support non-monotone queries including aggregation and set difference. We prove that query processing over AU-DBs preserves the bounds of certain and possible answers and investigate algorithms for compacting intermediate results to retain efficiency. Through an compact encoding of possible answers, our approach also provides a solid foundation for handling missing data. Using optimizations that trade accuracy for performance, our approach scales to complex queries and large datasets, and produces accurate results. Furthermore, it significantly outperforms alternative methods for uncertain data management."
                    },
                    {
                        "title": "Efficient Answering of Historical What-if Queries",
                        "abstract": "We introduce historical what-if queries, a novel type of what-if analysis that determines the effect of a hypothetical change to the transactional history of a database. For example, \"how would revenue be affected if we would have charged an additional $6 for shipping?\" Such queries may lead to more actionable insights than traditional what-if queries as their results can be used to inform future actions, e.g., increasing shipping fees. We develop efficient techniques for answering historical what-if queries, i.e., determining how a modified history affects the current database state. Our techniques are based on reenactment, a replay technique for transactional histories. We optimize this process using program and data slicing techniques that determine which updates and what data can be excluded from reenactment without affecting the result. Using an implementation of our techniques in Mahif (a Middleware for Answering Historical what-IF queries) we demonstrate their effectiveness experimentally."
                    },
                    {
                        "title": "Putting Things into Context: Rich Explanations for Query Answers using Join Graphs (extended version)",
                        "abstract": "In many data analysis applications, there is a need to explain why a surprising or interesting result was produced by a query. Previous approaches to explaining results have directly or indirectly used data provenance (input tuples contributing to the result(s) of interest), which is limited by the fact that relevant information for explaining an answer may not be fully contained in the provenance. We propose a new approach for explaining query results by augmenting provenance with information from other related tables in the database. We develop a suite of optimization techniques, and demonstrate experimentally using real datasets and through a user study that our approach produces meaningful results by efficiently navigating the large search space of possible explanations."
                    },
                    {
                        "title": "Efficiently Computing Provenance Graphs for Queries with Negation",
                        "abstract": "Explaining why an answer is in the result of a query or why it is missing from the result is important for many applications including auditing, debugging data and queries, and answering hypothetical questions about data. Both types of questions, i.e., why and why-not provenance, have been studied extensively. In this work, we present the first practical approach for answering such questions for queries with negation (first-order queries). Our approach is based on a rewriting of Datalog rules (called firing rules) that captures successful rule derivations within the context of a Datalog query. We extend this rewriting to support negation and to capture failed derivations that explain missing answers. Given a (why or why-not) provenance question, we compute an explanation, i.e., the part of the provenance that is relevant to answer the question. We introduce optimizations that prune parts of a provenance graph early on if we can determine that they will not be part of the explanation for a given question. We present an implementation that runs on top of a relational database using SQL to compute explanations. Our experiments demonstrate that our approach scales to large instances and significantly outperforms an earlier approach which instantiates the full provenance to compute explanations."
                    },
                    {
                        "title": "Computing expected multiplicities for bag-TIDBs with bounded multiplicities",
                        "abstract": "In this work, we study the problem of computing a tuple's expected multiplicity over probabilistic databases with bag semantics (where each tuple is associated with a multiplicity) exactly and approximately. We consider bag-TIDBs where we have a bound $c$ on the maximum multiplicity of each tuple and tuples are independent probabilistic events (we refer to such databases as c-TIDBs. We are specifically interested in the fine-grained complexity of computing expected multiplicities and how it compares to the complexity of deterministic query evaluation algorithms -- if these complexities are comparable, it opens the door to practical deployment of probabilistic databases. Unfortunately, our results imply that computing expected multiplicities for c-TIDBs based on the results produced by such query evaluation algorithms introduces super-linear overhead (under parameterized complexity hardness assumptions/conjectures). We proceed to study approximation of expected result tuple multiplicities for positive relational algebra queries ($RA^+$) over c-TIDBs and for a non-trivial subclass of block-independent databases (BIDBs). We develop a sampling algorithm that computes a 1$\\pm\\epsilon$ approximation of the expected multiplicity of an output tuple in time linear in the runtime of the corresponding deterministic query for any $RA^+$ query."
                    },
                    {
                        "title": "Debugging Transactions and Tracking their Provenance with Reenactment",
                        "abstract": "Debugging transactions and understanding their execution are of immense importance for developing OLAP applications, to trace causes of errors in production systems, and to audit the operations of a database. However, debugging transactions is hard for several reasons: 1) after the execution of a transaction, its input is no longer available for debugging, 2) internal states of a transaction are typically not accessible, and 3) the execution of a transaction may be affected by concurrently running transactions. We present a debugger for transactions that enables non-invasive, post-mortem debugging of transactions with provenance tracking and supports what-if scenarios (changes to transaction code or data). Using reenactment, a declarative replay technique we have developed, a transaction is replayed over the state of the DB seen by its original execution including all its interactions with concurrently executed transactions from the history. Importantly, our approach uses the temporal database and audit logging capabilities available in many DBMS and does not require any modifications to the underlying database system nor transactional workload."
                    },
                    {
                        "title": "Mimir: Bringing CTables into Practice",
                        "abstract": "The present state of the art in analytics requires high upfront investment of human effort and computational resources to curate datasets, even before the first query is posed. So-called pay-as-you-go data curation techniques allow these high costs to be spread out, first by enabling queries over uncertain and incomplete data, and then by assessing the quality of the query results. We describe the design of a system, called Mimir, around a recently introduced class of probabilistic pay-as-you-go data cleaning operators called Lenses. Mimir wraps around any deterministic database engine using JDBC, extending it with support for probabilistic query processing. Queries processed through Mimir produce uncertainty-annotated result cursors that allow client applications to quickly assess result quality and provenance. We also present a GUI that provides analysts with an interactive tool for exploring the uncertainty exposed by the system. Finally, we present optimizations that make Lenses scalable, and validate this claim through experimental evidence."
                    }
                ]
            },
            "0779de8a-f626-4e61-839d-f5791d164e02": {
                "pk": "0779de8a-f626-4e61-839d-f5791d164e02",
                "name": "Babak Salimi",
                "collaborators": [
                    "Leopoldo Bertossi",
                    "Dan Suciu",
                    "Sudeepa Roy",
                    "Sainyam Galhotra",
                    "Bill Howe",
                    "Amir Gilad",
                    "Romila Pradhan",
                    "Jiongli Zhu",
                    "Luke Rodriguez",
                    "Johannes Gehrke"
                ],
                "domain": [
                    "Causal Inference",
                    "Database Systems",
                    "Fairness in AI",
                    "Explainable AI"
                ],
                "publications": [
                    {
                        "title": "ZaliQL: A SQL-Based Framework for Drawing Causal Inference from Big Data",
                        "abstract": "Causal inference from observational data is a subject of active research and development in statistics and computer science. Many toolkits have been developed for this purpose that depends on statistical software. However, these toolkits do not scale to large datasets. In this paper we describe a suite of techniques for expressing causal inference tasks from observational data in SQL. This suite supports the state-of-the-art methods for causal inference and run at scale within a database engine. In addition, we introduce several optimization techniques that significantly speedup causal inference, both in the online and offline setting. We evaluate the quality and performance of our techniques by experiments of real datasets."
                    },
                    {
                        "title": "Causality in Databases: The Diagnosis and Repair Connections",
                        "abstract": "In this work we establish and investigate the connections between causality for query answers in databases, database repairs wrt. denial constraints, and consistency-based diagnosis. The first two are relatively new problems in databases, and the third one is an established subject in knowledge representation. We show how to obtain database repairs from causes and the other way around. The vast body of research on database repairs can be applied to the newer problem of determining actual causes for query answers. By formulating a causality problem as a diagnosis problem, we manage to characterize causes in terms of a system's diagnoses."
                    },
                    {
                        "title": "Causes for Query Answers from Databases, Datalog Abduction and View-Updates: The Presence of Integrity Constraints",
                        "abstract": "Causality has been recently introduced in databases, to model, characterize and possibly compute causes for query results (answers). Connections between queryanswer causality, consistency-based diagnosis, database repairs (wrt. integrity constraint violations), abductive diagnosis and the view-update problem have been established. In this work we further investigate connections between query-answer causality and abductive diagnosis and the view-update problem. In this context, we also define and investigate the notion of query-answer causality in the presence of integrity constraints."
                    },
                    {
                        "title": "Data Management for Causal Algorithmic Fairness",
                        "abstract": "Fairness is increasingly recognized as a critical component of machine learning systems. However, it is the underlying data on which these systems are trained that often reflects discrimination, suggesting a data management problem. In this paper, we first make a distinction between associational and causal definitions of fairness in the literature and argue that the concept of fairness requires causal reasoning. We then review existing works and identify future opportunities for applying data management techniques to causal algorithmic fairness."
                    },
                    {
                        "title": "Unifying Causality, Diagnosis, Repairs and View-Updates in Databases",
                        "abstract": "In this work we establish and point out connections between the notion of query-answer causality in databases and database repairs, model-based diagnosis in its consistency-based and abductive versions, and database updates through views. The mutual relationships among these areas of data management and knowledge representation shed light on each of them and help to share notions and results they have in common. In one way or another, these are all approaches to uncertainty management, which becomes even more relevant in the context of big data that have to be made sense of."
                    },
                    {
                        "title": "From Causes for Database Queries to Repairs and Model-Based Diagnosis and Back",
                        "abstract": "In this work we establish and investigate connections between causality for query answers in databases, database repairs wrt. denial constraints, and consistency-based diagnosis. The first two are relatively new problems in databases, and the third one is an established subject in knowledge representation. We show how to obtain database repairs from causes and the other way around. Causality problems are formulated as diagnosis problems, and the diagnoses provide causes and their responsibilities. The vast body of research on database repairs can be applied to the newer problem of determining actual causes for query answers and their responsibilities. These connections, which are interesting per se, allow us, after a transition -inspired by consistency-based diagnosis- to computational problems on hitting sets and vertex covers in hypergraphs, to obtain several new algorithmic and complexity results for causality in databases."
                    },
                    {
                        "title": "Query-Answer Causality in Databases: Abductive Diagnosis and View-Updates",
                        "abstract": "Causality has been recently introduced in databases, to model, characterize and possibly compute causes for query results (answers). Connections between query causality and consistency-based diagnosis and database repairs (wrt. integrity constrain violations) have been established in the literature. In this work we establish connections between query causality and abductive diagnosis and the view-update problem. The unveiled relationships allow us to obtain new complexity results for query causality -the main focus of our work- and also for the two other areas."
                    },
                    {
                        "title": "From Causes for Database Queries to Repairs and Model-Based Diagnosis and Back",
                        "abstract": "In this work we establish and investigate connections between causes for query answers in databases, database repairs wrt. denial constraints, and consistency-based diagnosis. The first two are relatively new research areas in databases, and the third one is an established subject in knowledge representation. We show how to obtain database repairs from causes, and the other way around. Causality problems are formulated as diagnosis problems, and the diagnoses provide causes and their responsibilities. The vast body of research on database repairs can be applied to the newer problems of computing actual causes for query answers and their responsibilities. These connections, which are interesting per se, allow us, after a transition -inspired by consistency-based diagnosis- to computational problems on hitting sets and vertex covers in hypergraphs, to obtain several new algorithmic and complexity results for database causality."
                    },
                    {
                        "title": "Causes for Query Answers from Databases: Datalog Abduction, View-Updates, and Integrity Constraints",
                        "abstract": "Causality has been recently introduced in databases, to model, characterize, and possibly compute causes for query answers. Connections between QA-causality and consistency-based diagnosis and database repairs (wrt. integrity constraint violations) have already been established. In this work we establish precise connections between QA-causality and both abductive diagnosis and the view-update problem in databases, allowing us to obtain new algorithmic and complexity results for QA-causality. We also obtain new results on the complexity of view-conditioned causality, and investigate the notion of QA-causality in the presence of integrity constraints, obtaining complexity results from a connection with view-conditioned causality. The abduction connection under integrity constraints allows us to obtain algorithmic tools for QA-causality."
                    },
                    {
                        "title": "A Framework for Inferring Causality from Multi-Relational Observational Data using Conditional Independence",
                        "abstract": "The study of causality or causal inference - how much a given treatment causally affects a given outcome in a population - goes way beyond correlation or association analysis of variables, and is critical in making sound data driven decisions and policies in a multitude of applications. The gold standard in causal inference is performing \"controlled experiments\", which often is not possible due to logistical or ethical reasons. As an alternative, inferring causality on \"observational data\" based on the \"Neyman-Rubin potential outcome model\" has been extensively used in statistics, economics, and social sciences over several decades. In this paper, we present a formal framework for sound causal analysis on observational datasets that are given as multiple relations and where the population under study is obtained by joining these base relations. We study a crucial condition for inferring causality from observational data, called the \"strong ignorability assumption\" (the treatment and outcome variables should be independent in the joined relation given the observed covariates), using known conditional independences that hold in the base relations. We also discuss how the structure of the conditional independences in base relations given as graphical models help infer new conditional independences in the joined relation. The proposed framework combines concepts from databases, statistics, and graphical models, and aims to initiate new research directions spanning these fields to facilitate powerful data-driven decisions in today's big data world."
                    },
                    {
                        "title": "Capuchin: Causal Database Repair for Algorithmic Fairness",
                        "abstract": "Fairness is increasingly recognized as a critical component of machine learning systems. However, it is the underlying data on which these systems are trained that often reflect discrimination, suggesting a database repair problem. Existing treatments of fairness rely on statistical correlations that can be fooled by statistical anomalies, such as Simpson's paradox. Proposals for causality-based definitions of fairness can correctly model some of these situations, but they require specification of the underlying causal models. In this paper, we formalize the situation as a database repair problem, proving sufficient conditions for fair classifiers in terms of admissible variables as opposed to a complete causal model. We show that these conditions correctly capture subtle fairness violations. We then use these conditions as the basis for database repair algorithms that provide provable fairness guarantees about classifiers trained on their training labels. We evaluate our algorithms on real data, demonstrating improvement over the state of the art on multiple fairness metrics proposed in the literature while retaining high utility."
                    },
                    {
                        "title": "Bias in OLAP Queries: Detection, Explanation, and Removal",
                        "abstract": "On line analytical processing (OLAP) is an essential element of decision-support systems. OLAP tools provide insights and understanding needed for improved decision making. However, the answers to OLAP queries can be biased and lead to perplexing and incorrect insights. In this paper, we propose HypDB, a system to detect, explain, and to resolve bias in decision-support queries. We give a simple definition of a \\emph{biased query}, which performs a set of independence tests on the data to detect bias. We propose a novel technique that gives explanations for bias, thus assisting an analyst in understanding what goes on. Additionally, we develop an automated method for rewriting a biased query into an unbiased query, which shows what the analyst intended to examine. In a thorough evaluation on several real datasets we show both the quality and the performance of our techniques, including the completely automatic discovery of the revolutionary insights from a famous 1973 discrimination case."
                    },
                    {
                        "title": "HypeR: Hypothetical Reasoning With What-If and How-To Queries Using a Probabilistic Causal Approach",
                        "abstract": "What-if (provisioning for an update to a database) and how-to (how to modify the database to achieve a goal) analyses provide insights to users who wish to examine hypothetical scenarios without making actual changes to a database and thereby help plan strategies in their fields. Typically, such analyses are done by testing the effect of an update in the existing database on a specific view created by a query of interest. In real-world scenarios, however, an update to a particular part of the database may affect tuples and attributes in a completely different part due to implicit semantic dependencies. To allow for hypothetical reasoning while accommodating such dependencies, we develop HypeR, a framework that supports what-if and how-to queries accounting for probabilistic dependencies among attributes captured by a probabilistic causal model. We extend the SQL syntax to include the necessary operators for expressing these hypothetical queries, define their semantics, devise efficient algorithms and optimizations to compute their results using concepts from causality and probabilistic databases, and evaluate the effectiveness of our approach experimentally."
                    },
                    {
                        "title": "Interpretable Data-Based Explanations for Fairness Debugging",
                        "abstract": "A wide variety of fairness metrics and eXplainable Artificial Intelligence (XAI) approaches have been proposed in the literature to identify bias in machine learning models that are used in critical real-life contexts. However, merely reporting on a model's bias, or generating explanations using existing XAI techniques is insufficient to locate and eventually mitigate sources of bias. We introduce Gopher, a system that produces compact, interpretable and causal explanations for bias or unexpected model behavior by identifying coherent subsets of the training data that are root-causes for this behavior. Specifically, we introduce the concept of causal responsibility that quantifies the extent to which intervening on training data by removing or updating subsets of it can resolve the bias. Building on this concept, we develop an efficient approach for generating the top-k patterns that explain model bias that utilizes techniques from the machine learning (ML) community to approximate causal responsibility and uses pruning rules to manage the large search space for patterns. Our experimental evaluation demonstrates the effectiveness of Gopher in generating interpretable explanations for identifying and debugging sources of bias."
                    },
                    {
                        "title": "Combining Counterfactuals With Shapley Values To Explain Image Models",
                        "abstract": "With the widespread use of sophisticated machine learning models in sensitive applications, understanding their decision-making has become an essential task. Models trained on tabular data have witnessed significant progress in explanations of their underlying decision making processes by virtue of having a small number of discrete features. However, applying these methods to high-dimensional inputs such as images is not a trivial task. Images are composed of pixels at an atomic level and do not carry any interpretability by themselves. In this work, we seek to use annotated high-level interpretable features of images to provide explanations. We leverage the Shapley value framework from Game Theory, which has garnered wide acceptance in general XAI problems. By developing a pipeline to generate counterfactuals and subsequently using it to estimate Shapley values, we obtain contrastive and interpretable explanations with strong axiomatic guarantees."
                    },
                    {
                        "title": "Consistent Range Approximation for Fair Predictive Modeling",
                        "abstract": "This paper proposes a novel framework for certifying the fairness of predictive models trained on biased data. It draws from query answering for incomplete and inconsistent databases to formulate the problem of consistent range approximation (CRA) of fairness queries for a predictive model on a target population. The framework employs background knowledge of the data collection process and biased data, working with or without limited statistics about the target population, to compute a range of answers for fairness queries. Using CRA, the framework builds predictive models that are certifiably fair on the target population, regardless of the availability of external data during training. The framework's efficacy is demonstrated through evaluations on real data, showing substantial improvement over existing state-of-the-art methods."
                    },
                    {
                        "title": "Heterogeneous Treatment Effects in Social Networks",
                        "abstract": "We study treatment effect modifiers for causal analysis in a social network, where neighbors' characteristics or network structure may affect the outcome of a unit, and the goal is to identify sub-populations with varying treatment effects using such network properties. We propose a novel framework for this purpose that facilitates data-driven decision making by testing hypotheses about complex effect modifiers in terms of network features or network patterns (e.g., characteristics of neighbors of a unit or belonging to a triangle), and by identifying sub-populations for which a treatment is likely to be effective or harmful. We describe a hypothesis testing approach that accounts for a unit's covariates, their neighbors' covariates, and patterns in the social network, and devise an algorithm incorporating ideas from causal inference, hypothesis testing, and graph theory to verify a hypothesized effect modifier. In addition, we develop a novel algorithm for the discovery of network patterns that are potential effect modifiers. We perform extensive experimental evaluations with a real development economics dataset about the treatment effect of belonging to a financial support network called self-help groups on risk tolerance, and also with a synthetic dataset with known ground truths simulating a vaccine efficacy trial, to evaluate our framework and algorithms."
                    },
                    {
                        "title": "Explaining Black-Box Algorithms Using Probabilistic Contrastive Counterfactuals",
                        "abstract": "There has been a recent resurgence of interest in explainable artificial intelligence (XAI) that aims to reduce the opaqueness of AI-based decision-making systems, allowing humans to scrutinize and trust them. Prior work in this context has focused on the attribution of responsibility for an algorithm's decisions to its inputs wherein responsibility is typically approached as a purely associational concept. In this paper, we propose a principled causality-based approach for explaining black-box decision-making systems that addresses limitations of existing methods in XAI. At the core of our framework lies probabilistic contrastive counterfactuals, a concept that can be traced back to philosophical, cognitive, and social foundations of theories on how humans generate and select explanations. We show how such counterfactuals can quantify the direct and indirect influences of a variable on decisions made by an algorithm, and provide actionable recourse for individuals negatively affected by the algorithm's decision. Unlike prior work, our system, LEWIS: (1)can compute provably effective explanations and recourse at local, global and contextual levels (2)is designed to work with users with varying levels of background knowledge of the underlying causal model and (3)makes no assumptions about the internals of an algorithmic system except for the availability of its input-output data. We empirically evaluate LEWIS on three real-world datasets and show that it generates human-understandable explanations that improve upon state-of-the-art approaches in XAI, including the popular LIME and SHAP. Experiments on synthetic data further demonstrate the correctness of LEWIS's explanations and the scalability of its recourse algorithm."
                    },
                    {
                        "title": "Quantifying Causal Effects on Query Answering in Databases",
                        "abstract": "The notion of actual causation, as formalized by Halpern and Pearl, has been recently applied to relational databases, to characterize and compute actual causes for possibly unexpected answers to monotone queries. Causes take the form of database tuples, and can be ranked according to their causal responsibility, a numerical measure of their relevance as a cause to the query answer. In this work we revisit this notion, introducing and making a case for an alternative measure of causal contribution, that of causal effect. The measure generalizes actual causes, and can be applied beyond monotone queries. We show that causal effect provides intuitive and intended results."
                    }
                ]
            }
        }
    },
    "2312.00486": {
        "paper_data": {
            "title": "REDUCR: Robust Data Downsampling Using Class Priority Reweighting",
            "url": "http://arxiv.org/abs/2312.00486v1",
            "arxiv_id": "2312.00486",
            "authors": [
                "William Bankes",
                "George Hughes",
                "Ilija Bogunovic",
                "Zi Wang"
            ],
            "abstract": "Modern machine learning models are becoming increasingly expensive to train for real-world image and text classification tasks, where massive web-scale data is collected in a streaming fashion. To reduce the training cost, online batch selection techniques have been developed to choose the most informative datapoints. However, these techniques can suffer from poor worst-class generalization performance due to class imbalance and distributional shifts. This work introduces REDUCR, a robust and efficient data downsampling method that uses class priority reweighting. REDUCR reduces the training data while preserving worst-class generalization performance. REDUCR assigns priority weights to datapoints in a class-aware manner using an online learning algorithm. We demonstrate the data efficiency and robust performance of REDUCR on vision and text classification tasks. On web-scraped datasets with imbalanced class distributions, REDUCR significantly improves worst-class test accuracy (and average accuracy), surpassing state-of-the-art methods by around 15%.",
            "introduction": " Introduction The abundance of data has had a profound impact on machine learning (ML), both positive and negative. On the one hand, it has enabled ML models to achieve unprecedented performance on a wide range of tasks, such as image and text classification [Kuznetsova et al., 2020, He et al., 2015, Brown et al., 2020, Tran et al., 2022, Anil et al., 2023]. On the other hand, training models on such large datasets can be computationally expensive and time-consuming [Kaddour et al., 2023], making it unsustainable in some situations [Bender et al., 2021, Patterson et al., 2021]. Additionally, the high speed at which streaming data is collected can make it infeasible to train on all of the data before deployment. To tackle these issues, various Related work. Mindermann et al. [2022] have developed an online batch selection method called RHO-L OSS, which uses a reference model trained on a holdout dataset to guide the selection of points during training. Certain extensions of this work have focused on using a reference model in different settings such as reinforcement learning [Sujit et al., 2022]. However, to our knowledge, none have focused on improving the worst-class generalisation performance. Other batch selection Background We consider a C-way classification task and denote a model as p(y|x, \u03b8), where xdenotes an input and y\u2208[C] the corresponding class label; the model is parameterized by \u03b8. For any training dataset D={(xi, yi)}N i=1withN datapoints, we use a point estimate of \u03b8to approximate the posterior model as p(y|x,D)\u2248p(y|x,\u02c6\u03b8). This estimate \u02c6\u03b8 can be obtained by running stochastic gradient descent (SGD) to optimize the cross-entropy loss over training dataset D. The goal of data downsampling is to select a dataset DT\u2282 D of size T(\u226aN) for training such that the generalisation error of the resulting model is minimised. We write this objective in terms of a separate holdout dataset Dho={(xho,i, yho,i)}Nho i=1as follows: DT= argmax D\u2282D,|D|=Tlogp(yho|xho, D), (1) where the inputs and their labels are xho= [xi,ho]Nho i=1andyho= [yi,ho]Nho i=1, respectively. Here, the likelihood of the holdout dataset is used as a proxy for the generalisation error. The problem is computationally prohibitive due to its combinatorial nature. Moreover, for a massive (or streaming) training dataset D, it is not computationally possible to load Dall at once and it is common to loop through the data by iteratively loading subsets. Online batch selection is a practical streaming setup to approximate the data downsampling problem, where at each timestep t, the model observes a training data subset Bt\u2282 D, and the goal is to iteratively select a small batch bt\u2282Bt for the model to take gradient steps. A standard solution to this problem is to design a selection score function that take into account the labels of the data. The selection score function can then be used to score the utility of the small batch bt. See Algorithm 2 in Appendix A.7.2 we selected \u03b3= 9for the image datasets as the performance of the non-expert class did not suffer too much. A.9 Hyperparameter Tuning In this section, we test the sensitivity of REDUCR with respect to the hyperparameters introduced. In particular, we investigate the sensitivity of the learning rate \u03b7, used for target model training; the gradient weight \u03b3, used for class-irreducible loss model training; the fraction of datapoints selected for",
            "references": [
                {
                    "title": "Challenges and Applications of Large Language Models",
                    "abstract": "Large Language Models (LLMs) went from non-existent to ubiquitous in the machine learning discourse within a few years. Due to the fast pace of the field, it is difficult to identify the remaining challenges and already fruitful application areas. In this paper, we aim to establish a systematic set of open problems and application successes so that ML researchers can comprehend the field's current state more quickly and become productive."
                },
                {
                    "title": "PaLM 2 Technical Report",
                    "abstract": "We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities and is more compute-efficient than its predecessor PaLM. PaLM 2 is a Transformer-based model trained using a mixture of objectives. Through extensive evaluations on English and multilingual language, and reasoning tasks, we demonstrate that PaLM 2 has significantly improved quality on downstream tasks across different model sizes, while simultaneously exhibiting faster and more efficient inference compared to PaLM. This improved efficiency enables broader deployment while also allowing the model to respond faster, for a more natural pace of interaction. PaLM 2 demonstrates robust reasoning capabilities exemplified by large improvements over PaLM on BIG-Bench and other reasoning tasks. PaLM 2 exhibits stable performance on a suite of responsible AI evaluations, and enables inference-time control over toxicity without additional overhead or impact on other capabilities. Overall, PaLM 2 achieves state-of-the-art performance across a diverse set of tasks and capabilities. When discussing the PaLM 2 family, it is important to distinguish between pre-trained models (of various sizes), fine-tuned variants of these models, and the user-facing products that use these models. In particular, user-facing products typically include additional pre- and post-processing steps. Additionally, the underlying models may evolve over time. Therefore, one should not expect the performance of user-facing products to exactly match the results reported in this report."
                },
                {
                    "title": "DoReMi: Optimizing Data Mixtures Speeds Up Language Model Pretraining",
                    "abstract": "The mixture proportions of pretraining data domains (e.g., Wikipedia, books, web text) greatly affect language model (LM) performance. In this paper, we propose Domain Reweighting with Minimax Optimization (DoReMi), which first trains a small proxy model using group distributionally robust optimization (Group DRO) over domains to produce domain weights (mixture proportions) without knowledge of downstream tasks. We then resample a dataset with these domain weights and train a larger, full-sized model. In our experiments, we use DoReMi on a 280M-parameter proxy model to set the domain weights for training an 8B-parameter model (30x larger) more efficiently. On The Pile, DoReMi improves perplexity across all domains, even when it downweights a domain. DoReMi improves average few-shot downstream accuracy by 6.5% points over a baseline model trained using The Pile's default domain weights and reaches the baseline accuracy with 2.6x fewer training steps. On the GLaM dataset, DoReMi, which has no knowledge of downstream tasks, even matches the performance of using domain weights tuned on downstream tasks."
                },
                {
                    "title": "Prioritizing Samples in Reinforcement Learning with Reducible Loss",
                    "abstract": "Most reinforcement learning algorithms take advantage of an experience replay buffer to repeatedly train on samples the agent has observed in the past. Not all samples carry the same amount of significance and simply assigning equal importance to each of the samples is a na\\\"ive strategy. In this paper, we propose a method to prioritize samples based on how much we can learn from a sample. We define the learn-ability of a sample as the steady decrease of the training loss associated with this sample over time. We develop an algorithm to prioritize samples with high learn-ability, while assigning lower priority to those that are hard-to-learn, typically caused by noise or stochasticity. We empirically show that our method is more robust than random sampling and also better than just prioritizing with respect to the training loss, i.e. the temporal difference loss, which is used in prioritized experience replay."
                },
                {
                    "title": "Plex: Towards Reliability using Pretrained Large Model Extensions",
                    "abstract": "A recent trend in artificial intelligence is the use of pretrained models for language and vision tasks, which have achieved extraordinary performance but also puzzling failures. Probing these models' abilities in diverse ways is therefore critical to the field. In this paper, we explore the reliability of models, where we define a reliable model as one that not only achieves strong predictive performance but also performs well consistently over many decision-making tasks involving uncertainty (e.g., selective prediction, open set recognition), robust generalization (e.g., accuracy and proper scoring rules such as log-likelihood on in- and out-of-distribution datasets), and adaptation (e.g., active learning, few-shot uncertainty). We devise 10 types of tasks over 40 datasets in order to evaluate different aspects of reliability on both vision and language domains. To improve reliability, we developed ViT-Plex and T5-Plex, pretrained large model extensions for vision and language modalities, respectively. Plex greatly improves the state-of-the-art across reliability tasks, and simplifies the traditional protocol as it improves the out-of-the-box performance and does not require designing scores or tuning the model for each task. We demonstrate scaling effects over model sizes up to 1B parameters and pretraining dataset sizes up to 4B examples. We also demonstrate Plex's capabilities on challenging tasks including zero-shot open set recognition, active learning, and uncertainty in conversational language understanding."
                },
                {
                    "title": "Beyond neural scaling laws: beating power law scaling via data pruning",
                    "abstract": "Widely observed neural scaling laws, in which error falls off as a power of the training set size, model size, or both, have driven substantial performance improvements in deep learning. However, these improvements through scaling alone require considerable costs in compute and energy. Here we focus on the scaling of error with dataset size and show how in theory we can break beyond power law scaling and potentially even reduce it to exponential scaling instead if we have access to a high-quality data pruning metric that ranks the order in which training examples should be discarded to achieve any pruned dataset size. We then test this improved scaling prediction with pruned dataset size empirically, and indeed observe better than power law scaling in practice on ResNets trained on CIFAR-10, SVHN, and ImageNet. Next, given the importance of finding high-quality pruning metrics, we perform the first large-scale benchmarking study of ten different data pruning metrics on ImageNet. We find most existing high performing metrics scale poorly to ImageNet, while the best are computationally intensive and require labels for every image. We therefore developed a new simple, cheap and scalable self-supervised pruning metric that demonstrates comparable performance to the best supervised metrics. Overall, our work suggests that the discovery of good data-pruning metrics may provide a viable path forward to substantially improved neural scaling laws, thereby reducing the resource costs of modern deep learning."
                },
                {
                    "title": "Prioritized Training on Points that are Learnable, Worth Learning, and Not Yet Learnt",
                    "abstract": "Training on web-scale data can take months. But most computation and time is wasted on redundant and noisy points that are already learnt or not learnable. To accelerate training, we introduce Reducible Holdout Loss Selection (RHO-LOSS), a simple but principled technique which selects approximately those points for training that most reduce the model's generalization loss. As a result, RHO-LOSS mitigates the weaknesses of existing data selection methods: techniques from the optimization literature typically select 'hard' (e.g. high loss) points, but such points are often noisy (not learnable) or less task-relevant. Conversely, curriculum learning prioritizes 'easy' points, but such points need not be trained on once learned. In contrast, RHO-LOSS selects points that are learnable, worth learning, and not yet learnt. RHO-LOSS trains in far fewer steps than prior art, improves accuracy, and speeds up training on a wide range of datasets, hyperparameters, and architectures (MLPs, CNNs, and BERT). On the large web-scraped image dataset Clothing-1M, RHO-LOSS trains in 18x fewer steps and reaches 2% higher final accuracy than uniform data shuffling."
                },
                {
                    "title": "Dataset Distillation by Matching Training Trajectories",
                    "abstract": "Dataset distillation is the task of synthesizing a small dataset such that a model trained on the synthetic set will match the test accuracy of the model trained on the full dataset. In this paper, we propose a new formulation that optimizes our distilled data to guide networks to a similar state as those trained on real data across many training steps. Given a network, we train it for several iterations on our distilled data and optimize the distilled data with respect to the distance between the synthetically trained parameters and the parameters trained on real data. To efficiently obtain the initial and target network parameters for large-scale datasets, we pre-compute and store training trajectories of expert networks trained on the real dataset. Our method handily outperforms existing methods and also allows us to distill higher-resolution visual data."
                },
                {
                    "title": "Robust Long-Tailed Learning under Label Noise",
                    "abstract": "Long-tailed learning has attracted much attention recently, with the goal of improving generalisation for tail classes. Most existing works use supervised learning without considering the prevailing noise in the training dataset. To move long-tailed learning towards more realistic scenarios, this work investigates the label noise problem under long-tailed label distribution. We first observe the negative impact of noisy labels on the performance of existing methods, revealing the intrinsic challenges of this problem. As the most commonly used approach to cope with noisy labels in previous literature, we then find that the small-loss trick fails under long-tailed label distribution. The reason is that deep neural networks cannot distinguish correctly-labeled and mislabeled examples on tail classes. To overcome this limitation, we establish a new prototypical noise detection method by designing a distance-based metric that is resistant to label noise. Based on the above findings, we propose a robust framework,~\\algo, that realizes noise detection for long-tailed learning, followed by soft pseudo-labeling via both label smoothing and diverse label guessing. Moreover, our framework can naturally leverage semi-supervised learning algorithms to further improve the generalisation. Extensive experiments on benchmark and real-world datasets demonstrate the superiority of our methods over existing baselines. In particular, our method outperforms DivideMix by 3\\% in test accuracy. Source code will be released soon."
                },
                {
                    "title": "Dataset Distillation with Infinitely Wide Convolutional Networks",
                    "abstract": "The effectiveness of machine learning algorithms arises from being able to extract useful features from large amounts of data. As model and dataset sizes increase, dataset distillation methods that compress large datasets into significantly smaller yet highly performant ones will become valuable in terms of training efficiency and useful feature extraction. To that end, we apply a novel distributed kernel based meta-learning framework to achieve state-of-the-art results for dataset distillation using infinitely wide convolutional neural networks. For instance, using only 10 datapoints (0.02% of original dataset), we obtain over 65% test accuracy on CIFAR-10 image classification task, a dramatic improvement over the previous best test accuracy of 40%. Our state-of-the-art results extend across many other settings for MNIST, Fashion-MNIST, CIFAR-10, CIFAR-100, and SVHN. Furthermore, we perform some preliminary analyses of our distilled datasets to shed light on how they differ from naturally occurring data."
                },
                {
                    "title": "Adaptive Inducing Points Selection For Gaussian Processes",
                    "abstract": "Gaussian Processes (\\textbf{GPs}) are flexible non-parametric models with strong probabilistic interpretation. While being a standard choice for performing inference on time series, GPs have few techniques to work in a streaming setting. \\cite{bui2017streaming} developed an efficient variational approach to train online GPs by using sparsity techniques: The whole set of observations is approximated by a smaller set of inducing points (\\textbf{IPs}) and moved around with new data. Both the number and the locations of the IPs will affect greatly the performance of the algorithm. In addition to optimizing their locations, we propose to adaptively add new points, based on the properties of the GP and the structure of the data."
                },
                {
                    "title": "Just Train Twice: Improving Group Robustness without Training Group Information",
                    "abstract": "Standard training via empirical risk minimization (ERM) can produce models that achieve high accuracy on average but low accuracy on certain groups, especially in the presence of spurious correlations between the input and label. Prior approaches that achieve high worst-group accuracy, like group distributionally robust optimization (group DRO) require expensive group annotations for each training point, whereas approaches that do not use such group annotations typically achieve unsatisfactory worst-group accuracy. In this paper, we propose a simple two-stage approach, JTT, that first trains a standard ERM model for several epochs, and then trains a second model that upweights the training examples that the first model misclassified. Intuitively, this upweights examples from groups on which standard ERM models perform poorly, leading to improved worst-group performance. Averaged over four image classification and natural language processing tasks with spurious correlations, JTT closes 75% of the gap in worst-group accuracy between standard ERM and group DRO, while only requiring group annotations on a small validation set in order to tune hyperparameters."
                },
                {
                    "title": "Carbon Emissions and Large Neural Network Training",
                    "abstract": "The computation demand for machine learning (ML) has grown rapidly recently, which comes with a number of costs. Estimating the energy cost helps measure its environmental impact and finding greener strategies, yet it is challenging without detailed information. We calculate the energy use and carbon footprint of several recent large models-T5, Meena, GShard, Switch Transformer, and GPT-3-and refine earlier estimates for the neural architecture search that found Evolved Transformer. We highlight the following opportunities to improve energy efficiency and CO2 equivalent emissions (CO2e): Large but sparsely activated DNNs can consume<1/10th the energy of large, dense DNNs without sacrificing accuracy despite using as many or even more parameters. Geographic location matters for ML workload scheduling since the fraction of carbon-free energy and resulting CO2e vary ~5X-10X, even within the same country and the same organization. We are now optimizing where and when large models are trained. Specific datacenter infrastructure matters, as Cloud datacenters can be ~1.4-2X more energy efficient than typical datacenters, and the ML-oriented accelerators inside them can be ~2-5X more effective than off-the-shelf systems. Remarkably, the choice of DNN, datacenter, and processor can reduce the carbon footprint up to ~100-1000X. These large factors also make retroactive estimates of energy cost difficult. To avoid miscalculations, we believe ML papers requiring large computational resources should make energy consumption and CO2e explicit when practical. We are working to be more transparent about energy use and CO2e in our future research. To help reduce the carbon footprint of ML, we believe energy usage and CO2e should be a key metric in evaluating models, and we are collaborating with MLPerf developers to include energy usage during training and inference in this industry standard benchmark."
                },
                {
                    "title": "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? \ud83e\udd9c",
                    "abstract": "The past 3 years of work in NLP have been characterized by the development and deployment of ever larger language models, especially for English. BERT, its variants, GPT-2/3, and others, most recently Switch-C, have pushed the boundaries of the possible both through architectural innovations and through sheer size. Using these pretrained models and the methodology of fine-tuning them for specific tasks, researchers have extended the state of the art on a wide array of tasks as measured by leaderboards on specific benchmarks for English. In this paper, we take a step back and ask: How big is too big? What are the possible risks associated with this technology and what paths are available for mitigating those risks? We provide recommendations including weighing the environmental and financial costs first, investing resources into curating and carefully documenting datasets rather than ingesting everything on the web, carrying out pre-development exercises evaluating how the planned approach fits into research and development goals and supports stakeholder values, and encouraging research directions beyond ever larger language models."
                },
                {
                    "title": "WILDS: A Benchmark of in-the-Wild Distribution Shifts",
                    "abstract": "Distribution shifts -- where the training distribution differs from the test distribution -- can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity in the real-world deployments, these distribution shifts are under-represented in the datasets widely used in the ML community today. To address this gap, we present WILDS, a curated benchmark of 10 datasets reflecting a diverse range of distribution shifts that naturally arise in real-world applications, such as shifts across hospitals for tumor identification; across camera traps for wildlife monitoring; and across time and location in satellite imaging and poverty mapping. On each dataset, we show that standard training yields substantially lower out-of-distribution than in-distribution performance. This gap remains even with models trained by existing methods for tackling distribution shifts, underscoring the need for new methods for training models that are more robust to the types of distribution shifts that arise in practice. To facilitate method development, we provide an open-source package that automates dataset loading, contains default model architectures and hyperparameters, and standardizes evaluations. Code and leaderboards are available at https://wilds.stanford.edu."
                },
                {
                    "title": "Heteroskedastic and Imbalanced Deep Learning with Adaptive Regularization",
                    "abstract": "Real-world large-scale datasets are heteroskedastic and imbalanced -- labels have varying levels of uncertainty and label distributions are long-tailed. Heteroskedasticity and imbalance challenge deep learning algorithms due to the difficulty of distinguishing among mislabeled, ambiguous, and rare examples. Addressing heteroskedasticity and imbalance simultaneously is under-explored. We propose a data-dependent regularization technique for heteroskedastic datasets that regularizes different regions of the input space differently. Inspired by the theoretical derivation of the optimal regularization strength in a one-dimensional nonparametric classification setting, our approach adaptively regularizes the data points in higher-uncertainty, lower-density regions more heavily. We test our method on several benchmark tasks, including a real-world heteroskedastic and imbalanced dataset, WebVision. Our experiments corroborate our theory and demonstrate a significant improvement over other methods in noise-robust deep learning."
                },
                {
                    "title": "Coresets via Bilevel Optimization for Continual Learning and Streaming",
                    "abstract": "Coresets are small data summaries that are sufficient for model training. They can be maintained online, enabling efficient handling of large data streams under resource constraints. However, existing constructions are limited to simple models such as k-means and logistic regression. In this work, we propose a novel coreset construction via cardinality-constrained bilevel optimization. We show how our framework can efficiently generate coresets for deep neural networks, and demonstrate its empirical benefits in continual learning and in streaming settings."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization",
                    "abstract": "Overparameterized neural networks can be highly accurate on average on an i.i.d. test set yet consistently fail on atypical groups of the data (e.g., by learning spurious correlations that hold on average but not in such groups). Distributionally robust optimization (DRO) allows us to learn models that instead minimize the worst-case training loss over a set of pre-defined groups. However, we find that naively applying group DRO to overparameterized neural networks fails: these models can perfectly fit the training data, and any model with vanishing average training loss also already has vanishing worst-case training loss. Instead, the poor worst-case performance arises from poor generalization on some groups. By coupling group DRO models with increased regularization---a stronger-than-typical L2 penalty or early stopping---we achieve substantially higher worst-group accuracies, with 10-40 percentage point improvements on a natural language inference task and two image tasks, while maintaining high average accuracies. Our results suggest that regularization is important for worst-group generalization in the overparameterized regime, even if it is not needed for average generalization. Finally, we introduce a stochastic optimization algorithm, with convergence guarantees, to efficiently train group DRO models."
                },
                {
                    "title": "HuggingFace's Transformers: State-of-the-art Natural Language Processing",
                    "abstract": "Recent progress in natural language processing has been driven by advances in both model architecture and model pretraining. Transformer architectures have facilitated building higher-capacity models and pretraining has made it possible to effectively utilize this capacity for a wide variety of tasks. \\textit{Transformers} is an open-source library with the goal of opening up these advances to the wider machine learning community. The library consists of carefully engineered state-of-the art Transformer architectures under a unified API. Backing this library is a curated collection of pretrained models made by and available for the community. \\textit{Transformers} is designed to be extensible by researchers, simple for practitioners, and fast and robust in industrial deployments. The library is available at \\url{this https URL}."
                },
                {
                    "title": "Accelerating Deep Learning by Focusing on the Biggest Losers",
                    "abstract": "This paper introduces Selective-Backprop, a technique that accelerates the training of deep neural networks (DNNs) by prioritizing examples with high loss at each iteration. Selective-Backprop uses the output of a training example's forward pass to decide whether to use that example to compute gradients and update parameters, or to skip immediately to the next example. By reducing the number of computationally-expensive backpropagation steps performed, Selective-Backprop accelerates training. Evaluation on CIFAR10, CIFAR100, and SVHN, across a variety of modern image models, shows that Selective-Backprop converges to target error rates up to 3.5x faster than with standard SGD and between 1.02--1.8x faster than a state-of-the-art importance sampling approach. Further acceleration of 26% can be achieved by using stale forward pass results for selection, thus also skipping forward passes of low priority examples."
                },
                {
                    "title": "No-Regret Learning in Unknown Games with Correlated Payoffs",
                    "abstract": "We consider the problem of learning to play a repeated multi-agent game with an unknown reward function. Single player online learning algorithms attain strong regret bounds when provided with full information feedback, which unfortunately is unavailable in many real-world scenarios. Bandit feedback alone, i.e., observing outcomes only for the selected action, yields substantially worse performance. In this paper, we consider a natural model where, besides a noisy measurement of the obtained reward, the player can also observe the opponents' actions. This feedback model, together with a regularity assumption on the reward function, allows us to exploit the correlations among different game outcomes by means of Gaussian processes (GPs). We propose a novel confidence-bound based bandit algorithm GP-MW, which utilizes the GP model for the reward function and runs a multiplicative weight (MW) method. We obtain novel kernel-dependent regret bounds that are comparable to the known bounds in the full information setting, while substantially improving upon the existing bandit results. We experimentally demonstrate the effectiveness of GP-MW in random matrix games, as well as real-world problems of traffic routing and movie recommendation. In our experiments, GP-MW consistently outperforms several baselines, while its performance is often comparable to methods that have access to full information feedback."
                },
                {
                    "title": "Distributionally Robust Language Modeling",
                    "abstract": "Language models are generally trained on data spanning a wide range of topics (e.g., news, reviews, fiction), but they might be applied to an a priori unknown target distribution (e.g., restaurant reviews). In this paper, we first show that training on text outside the test distribution can degrade test performance when using standard maximum likelihood (MLE) training. To remedy this without the knowledge of the test distribution, we propose an approach which trains a model that performs well over a wide range of potential test distributions. In particular, we derive a new distributionally robust optimization (DRO) procedure which minimizes the loss of the model over the worst-case mixture of topics with sufficient overlap with the training distribution. Our approach, called topic conditional value at risk (topic CVaR), obtains a 5.5 point perplexity reduction over MLE when the language models are trained on a mixture of Yelp reviews and news and tested only on reviews."
                },
                {
                    "title": "Don\u2019t Take the Easy Way Out: Ensemble Based Methods for Avoiding Known Dataset Biases",
                    "abstract": "State-of-the-art models often make use of superficial patterns in the data that do not generalize well to out-of-domain or adversarial settings. For example, textual entailment models often learn that particular key words imply entailment, irrespective of context, and visual question answering models learn to predict prototypical answers, without considering evidence in the image. In this paper, we show that if we have prior knowledge of such biases, we can train a model to be more robust to domain shift. Our method has two stages: we (1) train a naive model that makes predictions exclusively based on dataset biases, and (2) train a robust model as part of an ensemble with the naive one in order to encourage it to focus on other patterns in the data that are more likely to generalize. Experiments on five datasets with out-of-domain test sets show significantly improved robustness in all settings, including a 12 point gain on a changing priors visual question answering dataset and a 9 point gain on an adversarial question answering test set."
                },
                {
                    "title": "Ordered SGD: A New Stochastic Optimization Framework for Empirical Risk Minimization",
                    "abstract": "We propose a new stochastic optimization framework for empirical risk minimization problems such as those that arise in machine learning. The traditional approaches, such as (mini-batch) stochastic gradient descent (SGD), utilize an unbiased gradient estimator of the empirical average loss. In contrast, we develop a computationally efficient method to construct a gradient estimator that is purposely biased toward those observations with higher current losses. On the theory side, we show that the proposed method minimizes a new ordered modification of the empirical average loss, and is guaranteed to converge at a sublinear rate to a global optimum for convex loss and to a critical point for weakly convex (non-convex) loss. Furthermore, we prove a new generalization bound for the proposed algorithm. On the empirical side, the numerical experiments show that our proposed method consistently improves the test errors compared with the standard mini-batch SGD in various models including SVM, logistic regression, and deep learning problems."
                },
                {
                    "title": "Selection Via Proxy: Efficient Data Selection For Deep Learning",
                    "abstract": "Data selection methods such as active learning and core-set selection are useful tools for machine learning on large datasets, but they can be prohibitively expensive to apply in deep learning. Unlike in other areas of machine learning, the feature representations that these techniques depend on are learned in deep learning rather than given, which takes a substantial amount of training time. In this work, we show that we can significantly improve the computational efficiency of data selection in deep learning by using a much smaller proxy model to perform data selection for tasks that will eventually require a large target model (e.g., selecting data points to label for active learning). In deep learning, we can scale down models by removing hidden layers or reducing their dimension to create proxies that are an order of magnitude faster. Although these small proxy models have significantly higher error, we find that they empirically provide useful rankings for data selection that have a high correlation with those of larger models. We evaluate this \"selection via proxy\" (SVP) approach on several data selection tasks. For active learning, applying SVP to Sener and Savarese [2018]'s recent method for active learning in deep learning gives a 4x improvement in execution time while yielding the same model accuracy. For core-set selection, we show that a proxy model that trains 10x faster than a target ResNet164 model on CIFAR10 can be used to remove 50% of the training data without compromising the accuracy of the target model, making end-to-end training time improvements via core-set selection possible."
                },
                {
                    "title": "BatchBALD: Efficient and Diverse Batch Acquisition for Deep Bayesian Active Learning",
                    "abstract": "We develop BatchBALD, a tractable approximation to the mutual information between a batch of points and model parameters, which we use as an acquisition function to select multiple informative points jointly for the task of deep Bayesian active learning. BatchBALD is a greedy linear-time $1 - \\frac{1}{e}$-approximate algorithm amenable to dynamic programming and efficient caching. We compare BatchBALD to the commonly used approach for batch data acquisition and find that the current approach acquires similar and redundant points, sometimes performing worse than randomly acquiring data. We finish by showing that, using BatchBALD to consider dependencies within an acquisition batch, we achieve new state of the art performance on standard benchmarks, providing substantial data efficiency improvements in batch acquisition."
                },
                {
                    "title": "Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds",
                    "abstract": "We design a new algorithm for batch active learning with deep neural network models. Our algorithm, Batch Active learning by Diverse Gradient Embeddings (BADGE), samples groups of points that are disparate and high-magnitude when represented in a hallucinated gradient space, a strategy designed to incorporate both predictive uncertainty and sample diversity into every selected batch. Crucially, BADGE trades off between diversity and uncertainty without requiring any hand-tuned hyperparameters. We show that while other approaches sometimes succeed for particular batch sizes or architectures, BADGE consistently performs as well or better, making it a versatile option for practical active learning problems."
                },
                {
                    "title": "Class-Balanced Loss Based on Effective Number of Samples",
                    "abstract": "With the rapid increase of large-scale, real-world datasets, it becomes critical to address the problem of long-tailed data distribution (i.e., a few classes account for most of the data, while most classes are under-represented). Existing solutions typically adopt class re-balancing strategies such as re-sampling and re-weighting based on the number of observations for each class. In this work, we argue that as the number of samples increases, the additional benefit of a newly added data point will diminish. We introduce a novel theoretical framework to measure data overlap by associating with each sample a small neighboring region rather than a single point. The effective number of samples is defined as the volume of samples and can be calculated by a simple formula $(1-\\beta^{n})/(1-\\beta)$, where $n$ is the number of samples and $\\beta \\in [0,1)$ is a hyperparameter. We design a re-weighting scheme that uses the effective number of samples for each class to re-balance the loss, thereby yielding a class-balanced loss. Comprehensive experiments are conducted on artificially induced long-tailed CIFAR datasets and large-scale datasets including ImageNet and iNaturalist. Our results show that when trained with the proposed class-balanced loss, the network is able to achieve significant performance gains on long-tailed datasets."
                },
                {
                    "title": "GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding",
                    "abstract": "Human ability to understand language is general, flexible, and robust. In contrast, most NLU models above the word level are designed for a specific task and struggle with out-of-domain data. If we aspire to develop models with understanding beyond the detection of superficial correspondences between inputs and outputs, then it is critical to develop a unified model that can execute a range of linguistic tasks across different domains. To facilitate research in this direction, we present the General Language Understanding Evaluation (GLUE, gluebenchmark.com): a benchmark of nine diverse NLU tasks, an auxiliary dataset for probing models for understanding of specific linguistic phenomena, and an online platform for evaluating and comparing models. For some benchmark tasks, training data is plentiful, but for others it is limited or does not match the genre of the test set. GLUE thus favors models that can represent linguistic knowledge in a way that facilitates sample-efficient learning and effective knowledge-transfer across tasks. While none of the datasets in GLUE were created from scratch for the benchmark, four of them feature privately-held test data, which is used to ensure that the benchmark is used fairly. We evaluate baselines that use ELMo (Peters et al., 2018), a powerful transfer learning technique, as well as state-of-the-art sentence representation models. The best models still achieve fairly low absolute scores. Analysis with our diagnostic dataset yields similarly weak performance over all phenomena tested, with some exceptions."
                },
                {
                    "title": "Co-teaching: Robust training of deep neural networks with extremely noisy labels",
                    "abstract": "Deep learning with noisy labels is practically challenging, as the capacity of deep models is so high that they can totally memorize these noisy labels sooner or later during training. Nonetheless, recent studies on the memorization effects of deep neural networks show that they would first memorize training data of clean labels and then those of noisy labels. Therefore in this paper, we propose a new deep learning paradigm called Co-teaching for combating with noisy labels. Namely, we train two deep neural networks simultaneously, and let them teach each other given every mini-batch: firstly, each network feeds forward all data and selects some data of possibly clean labels; secondly, two networks communicate with each other what data in this mini-batch should be used for training; finally, each network back propagates the data selected by its peer network and updates itself. Empirical results on noisy versions of MNIST, CIFAR-10 and CIFAR-100 demonstrate that Co-teaching is much superior to the state-of-the-art methods in the robustness of trained deep models."
                },
                {
                    "title": "Learning to Reweight Examples for Robust Deep Learning",
                    "abstract": "Deep neural networks have been shown to be very powerful modeling tools for many supervised learning tasks involving complex input patterns. However, they can also easily overfit to training set biases and label noises. In addition to various regularizers, example reweighting algorithms are popular solutions to these problems, but they require careful tuning of additional hyperparameters, such as example mining schedules and regularization hyperparameters. In contrast to past reweighting methods, which typically consist of functions of the cost value of each example, in this work we propose a novel meta-learning algorithm that learns to assign weights to training examples based on their gradient directions. To determine the example weights, our method performs a meta gradient descent step on the current mini-batch example weights (which are initialized from zero) to minimize the loss on a clean unbiased validation set. Our proposed method can be easily implemented on any type of deep network, does not require any additional hyperparameter tuning, and achieves impressive performance on class imbalance and corrupted label problems where only a small amount of clean validation data is available."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "The iNaturalist Species Classification and Detection Dataset",
                    "abstract": "Existing image classification datasets used in computer vision tend to have a uniform distribution of images across object categories. In contrast, the natural world is heavily imbalanced, as some species are more abundant and easier to photograph than others. To encourage further progress in challenging real world conditions we present the iNaturalist species classification and detection dataset, consisting of 859,000 images from over 5,000 different species of plants and animals. It features visually similar species, captured in a wide variety of situations, from all over the world. Images were collected with different camera types, have varying image quality, feature a large class imbalance, and have been verified by multiple citizen scientists. We discuss the collection of the dataset and present extensive baseline experiments using state-of-the-art computer vision classification and detection models. Results show that current non-ensemble based methods achieve only 67% top one classification accuracy, illustrating the difficulty of the dataset. Specifically, we observe poor results for classes with small numbers of training examples suggesting more attention is needed in low-shot learning."
                },
                {
                    "title": "Robust Optimization for Non-Convex Objectives",
                    "abstract": "We consider robust optimization problems, where the goal is to optimize in the worst case over a class of objective functions. We develop a reduction from robust improper optimization to Bayesian optimization: given an oracle that returns $\\alpha$-approximate solutions for distributions over objectives, we compute a distribution over solutions that is $\\alpha$-approximate in the worst case. We show that de-randomizing this solution is NP-hard in general, but can be done for a broad class of statistical learning tasks. We apply our results to robust neural network training and submodular optimization. We evaluate our approach experimentally on corrupted character classification, and robust influence maximization in networks."
                },
                {
                    "title": "Biased Importance Sampling for Deep Neural Network Training",
                    "abstract": "Importance sampling has been successfully used to accelerate stochastic optimization in many convex problems. However, the lack of an efficient way to calculate the importance still hinders its application to Deep Learning. \nIn this paper, we show that the loss value can be used as an alternative importance metric, and propose a way to efficiently approximate it for a deep model, using a small model trained for that purpose in parallel. \nThis method allows in particular to utilize a biased gradient estimate that implicitly optimizes a soft max-loss, and leads to better generalization performance. While such method suffers from a prohibitively high variance of the gradient estimate when using a standard stochastic optimizer, we show that when it is combined with our sampling mechanism, it results in a reliable procedure. \nWe showcase the generality of our method by testing it on both image classification and language modeling tasks using deep convolutional and recurrent neural networks. In particular, our method results in 30% faster training of a CNN for CIFAR10 than when using uniform sampling."
                },
                {
                    "title": "A Broad-Coverage Challenge Corpus for Sentence Understanding through Inference",
                    "abstract": "This paper introduces the Multi-Genre Natural Language Inference (MultiNLI) corpus, a dataset designed for use in the development and evaluation of machine learning models for sentence understanding. At 433k examples, this resource is one of the largest corpora available for natural language inference (a.k.a. recognizing textual entailment), improving upon available resources in both its coverage and difficulty. MultiNLI accomplishes this by offering data from ten distinct genres of written and spoken English, making it possible to evaluate systems on nearly the full complexity of the language, while supplying an explicit setting for evaluating cross-genre domain adaptation. In addition, an evaluation using existing machine learning models designed for the Stanford NLI corpus shows that it represents a substantially more difficult task than does that corpus, despite the two showing similar levels of inter-annotator agreement."
                },
                {
                    "title": "Practical Coreset Constructions for Machine Learning",
                    "abstract": "We investigate coresets - succinct, small summaries of large data sets - so that solutions found on the summary are provably competitive with solution found on the full data set. We provide an overview over the state-of-the-art in coreset construction for machine learning. In Section 2, we present both the intuition behind and a theoretically sound framework to construct coresets for general problems and apply it to $k$-means clustering. In Section 3 we summarize existing coreset construction algorithms for a variety of machine learning problems such as maximum likelihood estimation of mixture models, Bayesian non-parametric models, principal component analysis, regression and general empirical risk minimization."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Online Batch Selection for Faster Training of Neural Networks",
                    "abstract": "Deep neural networks are commonly trained using stochastic non-convex optimization procedures, which are driven by gradient information estimated on fractions (batches) of the dataset. While it is commonly accepted that batch size is an important parameter for offline tuning, the benefits of online selection of batches remain poorly understood. We investigate online batch selection strategies for two state-of-the-art methods of stochastic gradient-based optimization, AdaDelta and Adam. As the loss function to be minimized for the whole dataset is an aggregation of loss functions of individual datapoints, intuitively, datapoints with the greatest loss should be considered (selected in a batch) more frequently. However, the limitations of this intuition and the proper control of the selection pressure over time are open questions. We propose a simple strategy where all datapoints are ranked w.r.t. their latest known loss value and the probability to be selected decays exponentially as a function of rank. Our experimental results on the MNIST dataset suggest that selecting batches speeds up both AdaDelta and Adam by a factor of about 5."
                },
                {
                    "title": "Learning from massive noisy labeled data for image classification",
                    "abstract": "Large-scale supervised datasets are crucial to train convolutional neural networks (CNNs) for various computer vision problems. However, obtaining a massive amount of well-labeled data is usually very expensive and time consuming. In this paper, we introduce a general framework to train CNNs with only a limited number of clean labels and millions of easily obtained noisy labels. We model the relationships between images, class labels and label noises with a probabilistic graphical model and further integrate it into an end-to-end deep learning system. To demonstrate the effectiveness of our approach, we collect a large-scale real-world clothing classification dataset with both noisy and clean labels. Experiments on this dataset indicate that our approach can better correct the noisy labels and improves the performance of trained CNNs."
                },
                {
                    "title": "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification",
                    "abstract": "Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on the learnable activation and advanced initialization, we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66% [33]). To our knowledge, our result is the first to surpass the reported human-level performance (5.1%, [26]) on this dataset."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "Bayesian Active Learning for Classification and Preference Learning",
                    "abstract": "Information theoretic active learning has been widely studied for probabilistic models. For simple regression an optimal myopic policy is easily tractable. However, for other tasks and with more complex models, such as classication with nonparametric models, the optimal solution is harder to compute. Current approaches make approximations to achieve tractability. We propose an approach that expresses information gain in terms of predictive entropies, and apply this method to the Gaussian Process Classier (GPC). Our approach makes minimal approximations to the full information theoretic objective. Our experimental performance compares favourably to many popular active learning algorithms, and has equal or lower computational complexity. We compare well to decision theoretic approaches also, which are privy to more information and require much more computational time. Secondly, by developing further a reformulation of binary preference learning to a classication problem, we extend our algorithm to Gaussian Process preference learning."
                },
                {
                    "title": "Torchvision the machine-vision package of torch",
                    "abstract": "This paper presents Torchvision an open source machine vision package for Torch. Torch is a machine learning library providing a series of the state-of-the-art algorithms such as Neural Networks, Support Vector Machines, Gaussian Mixture Models, Hidden Markov Models and many others. Torchvision provides additional functionalities to manipulate and process images with standard image processing algorithms. Hence, the resulting images can be used directly with the Torch machine learning algorithms as Torchvision is fully integrated with Torch. Both Torch and Torchvision are written in C++ language and are publicly available under the Free-BSD License."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Prediction, learning, and games",
                    "abstract": "1. Introduction 2. Prediction with expert advice 3. Tight bounds for specific losses 4. Randomized prediction 5. Efficient forecasters for large classes of experts 6. Prediction with limited feedback 7. Prediction and playing games 8. Absolute loss 9. Logarithmic loss 10. Sequential investment 11. Linear pattern recognition 12. Linear classification 13. Appendix."
                },
                {
                    "title": "Mixture of experts for audio classification: an application to male female classification and musical genre recognition",
                    "abstract": "We report the experimental results obtained when applying a mixture of experts to the problem of audio classification for multimedia applications. The mixture of experts is based on multilayer perceptron neural networks as individual experts and piecewise Gaussian modeling was used for audio signal representation. Experimental results on two audio classification problems, male/female classification and musical genre recognition, show a clear improvement in using a mixture of experts in comparison to one individual expert"
                },
                {
                    "title": "Toward Scalable Learning with Non-Uniform Class and Cost Distributions: A Case Study in Credit Card Fraud Detection",
                    "abstract": "Very large databases with skewed class distributions and non-uniform cost per error are not uncommon in real-world data mining tasks. We devised a multi-classifier meta-learning approach to address these three issues. Our empirical results from a credit card fraud detection task indicate that the approach can significantly reduce loss due to illegitimate transactions."
                },
                {
                    "title": "Information-Based Objective Functions for Active Data Selection",
                    "abstract": "Learning can be made more efficient if we can actively select particularly salient data points. Within a Bayesian learning framework, objective functions are discussed that measure the expected informativeness of candidate measurements. Three alternative specifications of what we want to gain information about lead to three different criteria for data selection. All these criteria depend on the assumption that the hypothesis space is correct, which may prove to be their main weakness."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "School of Informatics, University of Edinburgh",
                    "abstract": "In this brief technical report we introduce the CINIC-10 dataset as a plug-in extended alternative for CIFAR-10. It was compiled by combining CIFAR-10 with images selected and downsampled from the ImageNet database. We present the approach to compiling the dataset, illustrate the example images for different classes, give pixel distributions for each part of the repository, and give some standard benchmarks for well known models. Details for download, usage, and compilation can be found in the associated github repository. 1 ."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the worst-class generalization performance in online batch selection for large or streaming datasets in machine learning?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the challenge of generalization in machine learning models trained on large datasets, which is a common scenario in real-world applications. Improving worst-class generalization can lead to more robust models that perform well across all classes, particularly in imbalanced datasets where some classes may be underrepresented. This advancement could influence future research by encouraging the development of more sophisticated batch selection methods and generalization techniques, ultimately leading to practical applications in fields such as healthcare, finance, and autonomous systems where reliable predictions are essential.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe problem is complex due to its combinatorial nature, making it computationally prohibitive to select an optimal subset of data from a massive or streaming dataset. Naive approaches may fail because they often do not account for the distribution of classes or the specific characteristics of the data, leading to suboptimal generalization performance. Additionally, the need to iteratively load and process data in a streaming setup introduces practical challenges, such as maintaining model performance while managing limited computational resources and time constraints.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on online batch selection methods, such as RHO-L OSS, but has not adequately addressed the improvement of worst-class generalization performance. Limitations in existing solutions include a lack of consideration for class imbalance and the complexities of real-time data processing. Barriers such as computational constraints and the need for effective scoring functions for batch selection have hindered progress. Our approach differs by specifically targeting worst-class generalization and incorporating advanced selection criteria that account for class distribution and model performance.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a novel online batch selection algorithm that utilizes a reference model to guide the selection of data points, focusing on minimizing the worst-class generalization error. We will use a diverse set of datasets for evaluation, including both balanced and imbalanced classification tasks. The performance will be measured using metrics such as accuracy and F1-score, particularly for the worst-performing class. We expect our approach to yield improved generalization performance across all classes, demonstrating the effectiveness of our method in real-world"
            }
        },
        "author_data": {
            "40b4fc7f-e616-45cb-9a59-64c32cba1cd6": {
                "pk": "40b4fc7f-e616-45cb-9a59-64c32cba1cd6",
                "name": "William Bankes",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "ac33432e-41b6-414e-ae70-0fe85074c3a8": {
                "pk": "ac33432e-41b6-414e-ae70-0fe85074c3a8",
                "name": "George Hughes",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "1ede7432-a3cf-4240-bf79-f604cfb0b77a": {
                "pk": "1ede7432-a3cf-4240-bf79-f604cfb0b77a",
                "name": "Ilija Bogunovic",
                "collaborators": [
                    "Andreas Krause",
                    "Jonathan Scarlett",
                    "Volkan Cevher",
                    "Stefanie Jegelka",
                    "Slobodan Mitrovi\u0107",
                    "Pier Giuseppe Sessa",
                    "Maryam Kamgarpour",
                    "Arpan Losalka",
                    "Junyao Zhao",
                    "Zihan Li"
                ],
                "domain": [
                    "Bandit Optimization",
                    "Gaussian Processes",
                    "Robust Learning",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Misspecified Gaussian Process Bandit Optimization",
                        "abstract": "We consider the problem of optimizing a black-box function based on noisy bandit feedback. Kernelized bandit algorithms have shown strong empirical and theoretical performance for this problem. They heavily rely on the assumption that the model is well-specified, however, and can fail without it. Instead, we introduce a \\emph{misspecified} kernelized bandit setting where the unknown function can be $\\epsilon$--uniformly approximated by a function with a bounded norm in some Reproducing Kernel Hilbert Space (RKHS). We design efficient and practical algorithms whose performance degrades minimally in the presence of model misspecification. Specifically, we present two algorithms based on Gaussian process (GP) methods: an optimistic EC-GP-UCB algorithm that requires knowing the misspecification error, and Phased GP Uncertainty Sampling, an elimination-type algorithm that can adapt to unknown model misspecification. We provide upper bounds on their cumulative regret in terms of $\\epsilon$, the time horizon, and the underlying kernel, and we show that our algorithm achieves optimal dependence on $\\epsilon$ with no prior knowledge of misspecification. In addition, in a stochastic contextual setting, we show that EC-GP-UCB can be effectively combined with the regret bound balancing strategy and attain similar regret bounds despite not knowing $\\epsilon$."
                    },
                    {
                        "title": "Time-Varying Gaussian Process Bandit Optimization",
                        "abstract": "We consider the sequential Bayesian optimization problem with bandit feedback, adopting a formulation that allows for the reward function to vary with time. We model the reward function using a Gaussian process whose evolution obeys a simple Markov model. We introduce two natural extensions of the classical Gaussian process upper confidence bound (GP-UCB) algorithm. The first, R-GP-UCB, resets GP-UCB at regular intervals. The second, TV-GP-UCB, instead forgets about old data in a smooth fashion. Our main contribution comprises of novel regret bounds for these algorithms, providing an explicit characterization of the trade-off between the time horizon and the rate at which the function varies. We illustrate the performance of the algorithms on both synthetic and real data, and we find the gradual forgetting of TV-GP-UCB to perform favorably compared to the sharp resetting of R-GP-UCB. Moreover, both algorithms significantly outperform classical GP-UCB, since it treats stale and fresh data equally."
                    },
                    {
                        "title": "Stochastic Linear Bandits Robust to Adversarial Attacks",
                        "abstract": "We consider a stochastic linear bandit problem in which the rewards are not only subject to random noise, but also adversarial attacks subject to a suitable budget $C$ (i.e., an upper bound on the sum of corruption magnitudes across the time horizon). We provide two variants of a Robust Phased Elimination algorithm, one that knows $C$ and one that does not. Both variants are shown to attain near-optimal regret in the non-corrupted case $C = 0$, while incurring additional additive terms respectively having a linear and quadratic dependency on $C$ in general. We present algorithm independent lower bounds showing that these additive terms are near-optimal. In addition, in a contextual setting, we revisit a setup of diverse contexts, and show that a simple greedy algorithm is provably robust with a near-optimal additive regret term, despite performing no explicit exploration and not knowing $C$."
                    },
                    {
                        "title": "Robust Maximization of Non-Submodular Objectives",
                        "abstract": "We study the problem of maximizing a monotone set function subject to a cardinality constraint $k$ in the setting where some number of elements $\\tau$ is deleted from the returned set. The focus of this work is on the worst-case adversarial setting. While there exist constant-factor guarantees when the function is submodular, there are no guarantees for non-submodular objectives. In this work, we present a new algorithm Oblivious-Greedy and prove the first constant-factor approximation guarantees for a wider class of non-submodular objectives. The obtained theoretical bounds are the first constant-factor bounds that also hold in the linear regime, i.e. when the number of deletions $\\tau$ is linear in $k$. Our bounds depend on established parameters such as the submodularity ratio and some novel ones such as the inverse curvature. We bound these parameters for two important objectives including support selection and variance reduction. Finally, we numerically demonstrate the robust performance of Oblivious-Greedy for these two objectives on various datasets."
                    },
                    {
                        "title": "Corruption-Tolerant Gaussian Process Bandit Optimization",
                        "abstract": "We consider the problem of optimizing an unknown (typically non-convex) function with a bounded norm in some Reproducing Kernel Hilbert Space (RKHS), based on noisy bandit feedback. We consider a novel variant of this problem in which the point evaluations are not only corrupted by random noise, but also adversarial corruptions. We introduce an algorithm Fast-Slow GP-UCB based on Gaussian process methods, randomized selection between two instances labeled \"fast\" (but non-robust) and \"slow\" (but robust), enlarged confidence bounds, and the principle of optimism under uncertainty. We present a novel theoretical analysis upper bounding the cumulative regret in terms of the corruption level, the time horizon, and the underlying kernel, and we argue that certain dependencies cannot be improved. We observe that distinct algorithmic ideas are required depending on whether one is required to perform well in both the corrupted and non-corrupted settings, and whether the corruption level is known or not."
                    },
                    {
                        "title": "A Robust Phased Elimination Algorithm for Corruption-Tolerant Gaussian Process Bandits",
                        "abstract": "We consider the sequential optimization of an unknown, continuous, and expensive to evaluate reward function, from noisy and adversarially corrupted observed rewards. When the corruption attacks are subject to a suitable budget $C$ and the function lives in a Reproducing Kernel Hilbert Space (RKHS), the problem can be posed as corrupted Gaussian process (GP) bandit optimization. We propose a novel robust elimination-type algorithm that runs in epochs, combines exploration with infrequent switching to select a small subset of actions, and plays each action for multiple time instants. Our algorithm, Robust GP Phased Elimination (RGP-PE), successfully balances robustness to corruptions with exploration and exploitation such that its performance degrades minimally in the presence (or absence) of adversarial corruptions. When $T$ is the number of samples and $\\gamma_T$ is the maximal information gain, the corruption-dependent term in our regret bound is $O(C \\gamma_T^{3/2})$, which is significantly tighter than the existing $O(C \\sqrt{T \\gamma_T})$ for several commonly-considered kernels. We perform the first empirical study of robustness in the corrupted GP bandit setting, and show that our algorithm is robust against a variety of adversarial attacks."
                    },
                    {
                        "title": "Efficient Model-Based Multi-Agent Mean-Field Reinforcement Learning",
                        "abstract": "Learning in multi-agent systems is highly challenging due to several factors including the non-stationarity introduced by agents' interactions and the combinatorial nature of their state and action spaces. In particular, we consider the Mean-Field Control (MFC) problem which assumes an asymptotically infinite population of identical agents that aim to collaboratively maximize the collective reward. In many cases, solutions of an MFC problem are good approximations for large systems, hence, efficient learning for MFC is valuable for the analogous discrete agent setting with many agents. Specifically, we focus on the case of unknown system dynamics where the goal is to simultaneously optimize for the rewards and learn from experience. We propose an efficient model-based reinforcement learning algorithm, $M^3-UCRL$, that runs in episodes, balances between exploration and exploitation during policy learning, and provably solves this problem. Our main theoretical contributions are the first general regret bounds for model-based reinforcement learning for MFC, obtained via a novel mean-field type analysis. To learn the system's dynamics, $M^3-UCRL$ can be instantiated with various statistical models, e.g., neural networks or Gaussian Processes. Moreover, we provide a practical parametrization of the core optimization problem that facilitates gradient-based optimization techniques when combined with differentiable dynamics approximation methods such as neural networks."
                    },
                    {
                        "title": "Distributionally Robust Bayesian Optimization",
                        "abstract": "Robustness to distributional shift is one of the key challenges of contemporary machine learning. Attaining such robustness is the goal of distributionally robust optimization, which seeks a solution to an optimization problem that is worst-case robust under a specified distributional shift of an uncontrolled covariate. In this paper, we study such a problem when the distributional shift is measured via the maximum mean discrepancy (MMD). For the setting of zeroth-order, noisy optimization, we present a novel distributionally robust Bayesian optimization algorithm (DRBO). Our algorithm provably obtains sub-linear robust regret in various settings that differ in how the uncertain covariate is observed. We demonstrate the robust performance of our method on both synthetic and real-world benchmarks."
                    },
                    {
                        "title": "Robust Submodular Maximization: A Non-Uniform Partitioning Approach",
                        "abstract": "We study the problem of maximizing a monotone submodular function subject to a cardinality constraint $k$, with the added twist that a number of items $\\tau$ from the returned set may be removed. We focus on the worst-case setting considered in (Orlin et al., 2016), in which a constant-factor approximation guarantee was given for $\\tau = o(\\sqrt{k})$. In this paper, we solve a key open problem raised therein, presenting a new Partitioned Robust (PRo) submodular maximization algorithm that achieves the same guarantee for more general $\\tau = o(k)$. Our algorithm constructs partitions consisting of buckets with exponentially increasing sizes, and applies standard submodular optimization subroutines on the buckets in order to construct the robust solution. We numerically demonstrate the performance of PRo in data summarization and influence maximization, demonstrating gains over both the greedy algorithm and the algorithm of (Orlin et al., 2016)."
                    },
                    {
                        "title": "Combining Pessimism with Optimism for Robust and Efficient Model-Based Deep Reinforcement Learning",
                        "abstract": "In real-world tasks, reinforcement learning (RL) agents frequently encounter situations that are not present during training time. To ensure reliable performance, the RL agents need to exhibit robustness against worst-case situations. The robust RL framework addresses this challenge via a worst-case optimization between an agent and an adversary. Previous robust RL algorithms are either sample inefficient, lack robustness guarantees, or do not scale to large problems. We propose the Robust Hallucinated Upper-Confidence RL (RH-UCRL) algorithm to provably solve this problem while attaining near-optimal sample complexity guarantees. RH-UCRL is a model-based reinforcement learning (MBRL) algorithm that effectively distinguishes between epistemic and aleatoric uncertainty and efficiently explores both the agent and adversary decision spaces during policy learning. We scale RH-UCRL to complex tasks via neural networks ensemble models as well as neural network policies. Experimentally, we demonstrate that RH-UCRL outperforms other robust deep RL algorithms in a variety of adversarial environments."
                    },
                    {
                        "title": "Risk-averse Heteroscedastic Bayesian Optimization",
                        "abstract": "Many black-box optimization tasks arising in high-stakes applications require risk-averse decisions. The standard Bayesian optimization (BO) paradigm, however, optimizes the expected value only. We generalize BO to trade mean and input-dependent variance of the objective, both of which we assume to be unknown a priori. In particular, we propose a novel risk-averse heteroscedastic Bayesian optimization algorithm (RAHBO) that aims to identify a solution with high return and low noise variance, while learning the noise distribution on the fly. To this end, we model both expectation and variance as (unknown) RKHS functions, and propose a novel risk-aware acquisition function. We bound the regret for our approach and provide a robust rule to report the final decision point for applications where only a single solution must be identified. We demonstrate the effectiveness of RAHBO on synthetic benchmark functions and hyperparameter tuning tasks."
                    },
                    {
                        "title": "Graph Neural Network Bandits",
                        "abstract": "We consider the bandit optimization problem with the reward function defined over graph-structured data. This problem has important applications in molecule design and drug discovery, where the reward is naturally invariant to graph permutations. The key challenges in this setting are scaling to large domains, and to graphs with many nodes. We resolve these challenges by embedding the permutation invariance into our model. In particular, we show that graph neural networks (GNNs) can be used to estimate the reward function, assuming it resides in the Reproducing Kernel Hilbert Space of a permutation-invariant additive kernel. By establishing a novel connection between such kernels and the graph neural tangent kernel (GNTK), we introduce the first GNN confidence bound and use it to design a phased-elimination algorithm with sublinear regret. Our regret bound depends on the GNTK's maximum information gain, which we also provide a bound for. While the reward function depends on all $N$ node features, our guarantees are independent of the number of graph nodes $N$. Empirically, our approach exhibits competitive performance and scales well on graph-structured domains."
                    },
                    {
                        "title": "Robust Best-arm Identification in Linear Bandits",
                        "abstract": "We study the robust best-arm identification problem (RBAI) in the case of linear rewards. The primary objective is to identify a near-optimal robust arm, which involves selecting arms at every round and assessing their robustness by exploring potential adversarial actions. This approach is particularly relevant when utilizing a simulator and seeking to identify a robust solution for real-world transfer. To this end, we present an instance-dependent lower bound for the robust best-arm identification problem with linear rewards. Furthermore, we propose both static and adaptive bandit algorithms that achieve sample complexity that matches the lower bound. In synthetic experiments, our algorithms effectively identify the best robust arm and perform similarly to the oracle strategy. As an application, we examine diabetes care and the process of learning insulin dose recommendations that are robust with respect to inaccuracies in standard calculators. Our algorithms prove to be effective in identifying robust dosage values across various age ranges of patients."
                    },
                    {
                        "title": "Sample-efficient Bayesian Optimisation Using Known Invariances",
                        "abstract": "Bayesian optimisation (BO) is a powerful framework for global optimisation of costly functions, using predictions from Gaussian process models (GPs). In this work, we apply BO to functions that exhibit invariance to a known group of transformations. We show that vanilla and constrained BO algorithms are inefficient when optimising such invariant objectives, and provide a method for incorporating group invariances into the kernel of the GP to produce invariance-aware algorithms that achieve significant improvements in sample efficiency. We derive a bound on the maximum information gain of these invariant kernels, and provide novel upper and lower bounds on the number of observations required for invariance-aware BO algorithms to achieve $\\epsilon$-optimality. We demonstrate our method's improved performance on a range of synthetic invariant and quasi-invariant functions. We also apply our method in the case where only some of the invariance is incorporated into the kernel, and find that these kernels achieve similar gains in sample efficiency at significantly reduced computational cost. Finally, we use invariant BO to design a current drive system for a nuclear fusion reactor, finding a high-performance solution where non-invariant methods failed."
                    },
                    {
                        "title": "Adversarially Robust Optimization with Gaussian Processes",
                        "abstract": "In this paper, we consider the problem of Gaussian process (GP) optimization with an added robustness requirement: The returned point may be perturbed by an adversary, and we require the function value to remain as high as possible even after this perturbation. This problem is motivated by settings in which the underlying functions during optimization and implementation stages are different, or when one is interested in finding an entire region of good inputs rather than only a single point. We show that standard GP optimization algorithms do not exhibit the desired robustness properties, and provide a novel confidence-bound based algorithm StableOpt for this purpose. We rigorously establish the required number of samples for StableOpt to find a near-optimal point, and we complement this guarantee with an algorithm-independent lower bound. We experimentally demonstrate several potential applications of interest using real-world data sets, and we show that StableOpt consistently succeeds in finding a stable maximizer where several baseline methods fail."
                    },
                    {
                        "title": "Adversarial Robust Decision Transformer: Enhancing Robustness of RvS via Minimax Returns-to-go",
                        "abstract": "Decision Transformer (DT), as one of the representative Reinforcement Learning via Supervised Learning (RvS) methods, has achieved strong performance in offline learning tasks by leveraging the powerful Transformer architecture for sequential decision-making. However, in adversarial environments, these methods can be non-robust, since the return is dependent on the strategies of both the decision-maker and adversary. Training a probabilistic model conditioned on observed return to predict action can fail to generalize, as the trajectories that achieve a return in the dataset might have done so due to a weak and suboptimal behavior adversary. To address this, we propose a worst-case-aware RvS algorithm, the Adversarial Robust Decision Transformer (ARDT), which learns and conditions the policy on in-sample minimax returns-to-go. ARDT aligns the target return with the worst-case return learned through minimax expectile regression, thereby enhancing robustness against powerful test-time adversaries. In experiments conducted on sequential games with full data coverage, ARDT can generate a maximin (Nash Equilibrium) strategy, the solution with the largest adversarial robustness. In large-scale sequential games and continuous adversarial RL environments with partial data coverage, ARDT demonstrates significantly superior robustness to powerful test-time adversaries and attains higher worst-case returns compared to contemporary DT methods."
                    },
                    {
                        "title": "Truncated Variance Reduction: A Unified Approach to Bayesian Optimization and Level-Set Estimation",
                        "abstract": "We present a new algorithm, truncated variance reduction (TruVaR), that treats Bayesian optimization (BO) and level-set estimation (LSE) with Gaussian processes in a unified fashion. The algorithm greedily shrinks a sum of truncated variances within a set of potential maximizers (BO) or unclassified points (LSE), which is updated based on confidence bounds. TruVaR is effective in several important settings that are typically non-trivial to incorporate into myopic algorithms, including pointwise costs and heteroscedastic noise. We provide a general theoretical guarantee for TruVaR covering these aspects, and use it to recover and strengthen existing results on BO and LSE. Moreover, we provide a new result for a setting where one can select from a number of noise levels having associated costs. We demonstrate the effectiveness of the algorithm on both synthetic and real-world data sets."
                    },
                    {
                        "title": "Streaming Robust Submodular Maximization: A Partitioned Thresholding Approach",
                        "abstract": "We study the classical problem of maximizing a monotone submodular function subject to a cardinality constraint k, with two additional twists: (i) elements arrive in a streaming fashion, and (ii) m items from the algorithm's memory are removed after the stream is finished. We develop a robust submodular algorithm STAR-T. It is based on a novel partitioning structure and an exponentially decreasing thresholding rule. STAR-T makes one pass over the data and retains a short but robust summary. We show that after the removal of any m elements from the obtained summary, a simple greedy algorithm STAR-T-GREEDY that runs on the remaining elements achieves a constant-factor approximation guarantee. In two different data summarization tasks, we demonstrate that it matches or outperforms existing greedy and streaming methods, even if they are allowed the benefit of knowing the removed subset in advance."
                    },
                    {
                        "title": "Mixed Strategies for Robust Optimization of Unknown Objectives",
                        "abstract": "We consider robust optimization problems, where the goal is to optimize an unknown objective function against the worst-case realization of an uncertain parameter. For this setting, we design a novel sample-efficient algorithm GP-MRO, which sequentially learns about the unknown objective from noisy point evaluations. GP-MRO seeks to discover a robust and randomized mixed strategy, that maximizes the worst-case expected objective value. To achieve this, it combines techniques from online learning with nonparametric confidence bounds from Gaussian processes. Our theoretical results characterize the number of samples required by GP-MRO to discover a robust near-optimal mixed strategy for different GP kernels of interest. We experimentally demonstrate the performance of our algorithm on synthetic datasets and on human-assisted trajectory planning tasks for autonomous vehicles. In our simulations, we show that robust deterministic strategies can be overly conservative, while the mixed strategies found by GP-MRO significantly improve the overall performance."
                    },
                    {
                        "title": "Learning to Play Sequential Games versus Unknown Opponents",
                        "abstract": "We consider a repeated sequential game between a learner, who plays first, and an opponent who responds to the chosen action. We seek to design strategies for the learner to successfully interact with the opponent. While most previous approaches consider known opponent models, we focus on the setting in which the opponent's model is unknown. To this end, we use kernel-based regularity assumptions to capture and exploit the structure in the opponent's response. We propose a novel algorithm for the learner when playing against an adversarial sequence of opponents. The algorithm combines ideas from bilevel optimization and online learning to effectively balance between exploration (learning about the opponent's model) and exploitation (selecting highly rewarding actions for the learner). Our results include algorithm's regret guarantees that depend on the regularity of the opponent's response and scale sublinearly with the number of game rounds. Moreover, we specialize our approach to repeated Stackelberg games, and empirically demonstrate its effectiveness in a traffic routing and wildlife conservation task"
                    }
                ]
            },
            "359fa0c1-dc0c-477b-adc3-23a61b298eb1": {
                "pk": "359fa0c1-dc0c-477b-adc3-23a61b298eb1",
                "name": "Zi Wang",
                "collaborators": [
                    "Jie Yang",
                    "Stefanie Jegelka",
                    "Jie Ren",
                    "Chengcheng Li",
                    "Xiangyang Wang",
                    "Somesh Jha",
                    "Daniel Hershcovich",
                    "Luqin Wang",
                    "Chen Wang",
                    "Jihye Choi"
                ],
                "domain": [
                    "Knowledge Distillation",
                    "Neural Network Pruning",
                    "Bayesian Optimization",
                    "Biometric Authentication"
                ],
                "publications": [
                    {
                        "title": "Data-Free Knowledge Distillation with Soft Targeted Transfer Set Synthesis",
                        "abstract": "Knowledge distillation (KD) has proved to be an effective approach for deep neural network compression, which learns a compact network (student) by transferring the knowledge from a pre-trained, over-parameterized network (teacher). In traditional KD, the transferred knowledge is usually obtained by feeding training samples to the teacher network to obtain the class probabilities. However, the original training dataset is not always available due to storage costs or privacy issues. In this study, we propose a novel data-free KD approach by modeling the intermediate feature space of the teacher with a multivariate normal distribution and leveraging the soft targeted labels generated by the distribution to synthesize pseudo samples as the transfer set. Several student networks trained with these synthesized transfer sets present competitive performance compared to the networks trained with the original training set and other data-free KD approaches."
                    },
                    {
                        "title": "Zero-Shot Knowledge Distillation from a Decision-Based Black-Box Model",
                        "abstract": "Knowledge distillation (KD) is a successful approach for deep neural network acceleration, with which a compact network (student) is trained by mimicking the softmax output of a pre-trained high-capacity network (teacher). In tradition, KD usually relies on access to the training samples and the parameters of the white-box teacher to acquire the transferred knowledge. However, these prerequisites are not always realistic due to storage costs or privacy issues in real-world applications. Here we propose the concept of decision-based black-box (DB3) knowledge distillation, with which the student is trained by distilling the knowledge from a black-box teacher (parameters are not accessible) that only returns classes rather than softmax outputs. We start with the scenario when the training set is accessible. We represent a sample's robustness against other classes by computing its distances to the teacher's decision boundaries and use it to construct the soft label for each training sample. After that, the student can be trained via standard KD. We then extend this approach to a more challenging scenario in which even accessing the training data is not feasible. We propose to generate pseudo samples distinguished by the teacher's decision boundaries to the largest extent and construct soft labels for them, which are used as the transfer set. We evaluate our approaches on various benchmark networks and datasets and experiment results demonstrate their effectiveness. Codes are available at: https://github.com/zwang84/zsdb3kd."
                    },
                    {
                        "title": "Max-value Entropy Search for Efficient Bayesian Optimization",
                        "abstract": "Entropy Search (ES) and Predictive Entropy Search (PES) are popular and empirically successful Bayesian Optimization techniques. Both rely on a compelling information-theoretic motivation, and maximize the information gained about the $\\arg\\max$ of the unknown function; yet, both are plagued by the expensive computation for estimating entropies. We propose a new criterion, Max-value Entropy Search (MES), that instead uses the information about the maximum function value. We show relations of MES to other Bayesian optimization methods, and establish a regret bound. We observe that MES maintains or improves the good empirical performance of ES/PES, while tremendously lightening the computational burden. In particular, MES is much more robust to the number of samples used for computing the entropy, and hence more efficient for higher dimensional problems."
                    },
                    {
                        "title": "On Evaluating Multilingual Compositional Generalization with Translated Datasets",
                        "abstract": "Compositional generalization allows efficient learning and human-like inductive biases. Since most research investigating compositional generalization in NLP is done on English, important questions remain underexplored. Do the necessary compositional generalization abilities differ across languages? Can models compositionally generalize cross-lingually? As a first step to answering these questions, recent work used neural machine translation to translate datasets for evaluating compositional generalization in semantic parsing. However, we show that this entails critical semantic distortion. To address this limitation, we craft a faithful rule-based translation of the MCWQ dataset from English to Chinese and Japanese. Even with the resulting robust benchmark, which we call MCWQ-R, we show that the distribution of compositions still suffers due to linguistic divergences, and that multilingual models still struggle with cross-lingual compositional generalization. Our dataset and methodology will be useful resources for the study of cross-lingual compositional generalization in other tasks."
                    },
                    {
                        "title": "Ear Wearable (Earable) User Authentication via Acoustic Toothprint",
                        "abstract": "Earables (ear wearables) is rapidly emerging as a new platform encompassing a diverse range of personal applications. The traditional authentication methods hence become less applicable and inconvenient for earables due to their limited input interface. Nevertheless, earables often feature rich around-the-head sensing capability that can be leveraged to capture new types of biometrics. In this work, we proposeToothSonic which leverages the toothprint-induced sonic effect produced by users performing teeth gestures for earable authentication. In particular, we design representative teeth gestures that can produce effective sonic waves carrying the information of the toothprint. To reliably capture the acoustic toothprint, it leverages the occlusion effect of the ear canal and the inward-facing microphone of the earables. It then extracts multi-level acoustic features to reflect the intrinsic toothprint information for authentication. The key advantages of ToothSonic are that it is suitable for earables and is resistant to various spoofing attacks as the acoustic toothprint is captured via the user's private teeth-ear channel that modulates and encrypts the sonic waves. Our experiment studies with 25 participants show that ToothSonic achieves up to 95% accuracy with only one of the users' tooth gestures."
                    },
                    {
                        "title": "Acoustic-based Object Detection for Pedestrian Using Smartphone",
                        "abstract": "Walking while using a smartphone is becoming a major pedestrian safety concern as people may unknowingly bump into various obstacles that could lead to severe injuries. In this paper, we propose ObstacleWatch, an acoustic-based obstacle collision detection system to improve the safety of pedestrians who are engaged in smartphone usage while walking. ObstacleWatch leverages the advanced audio hardware of the smartphone to sense the surrounding obstacles and infers fine-grained information about the frontal obstacle for collision detection. In particular, our system emits well-designed inaudible beep signals from the smartphone built-in speaker and listens to the reflections with the stereo recording of the smartphone. By analyzing the reflected signals received at two microphones, ObstacleWatch is able to extract fine-grained information of the frontal obstacle including the distance, angle, and size for detecting the possible collisions and to alert users. Our experimental evaluation under two real-world environments with different types of phones and obstacles shows that ObstacleWatch achieves over 92% accuracy in predicting obstacle collisions with distance estimation errors at about 2 cm. Results also show that ObstacleWatch is robust to different sizes of objects and is compatible with different phone models with low energy consumption."
                    },
                    {
                        "title": "An Ear Canal Deformation Based User Authentication Using Ear Wearable Devices",
                        "abstract": "Biometric-based authentication is gaining increasing attention for wearables and mobile applications. Meanwhile, the growing adoption of sensors in wearables also provides opportunities to capture novel wearable biometrics. In this work, we propose EarDynamic, an ear canal deformation-based user authentication using in-ear wearables. EarDynamic provides continuous and passive user authentication and is transparent to users. It leverages ear canal deformation that combines the unique static geometry and dynamic motions of the ear canal when the user is speaking for authentication. It utilizes an acoustic sensing approach to capture the ear canal deformation with the built-in microphone and speaker of the in-ear wearable. Specifically, it first emits well-designed inaudible beep signals and records the reflected signals from the ear canal. It then analyzes the reflected signals and extracts fine-grained acoustic features that correspond to the ear canal deformation for user authentication. Our extensive experimental evaluation shows that EarDynamic can achieve a recall of 97.38% and an F1 score of 96.84%. Results also show that our system works well under different noisy environments with various daily activities."
                    },
                    {
                        "title": "Thermodynamic Geometry of Nonequilibrium Fluctuations in Cyclically Driven Transport",
                        "abstract": "Nonequilibrium thermal machines under cyclic driving generally outperform steady-state counterparts. However, there is still lack of coherent understanding of versatile transport and fluctuation features under time modulations. Here, we formulate a theoretical framework of thermodynamic geometry in terms of full counting statistics of nonequilibrium driven transports. We find that, besides the conventional dynamic and adiabatic geometric curvature contributions, the generating function is divided into an additional nonadiabatic contribution, manifested as the metric term of full counting statistics. This nonadiabatic metric generalizes recent results of thermodynamic geometry in near-equilibrium entropy production to far-from-equilibrium fluctuations of general currents. Furthermore, the framework proves geometric thermodynamic uncertainty relations of near-adiabatic thermal devices, constraining fluctuations in terms of statistical metric quantities. We exemplify the theory in experimentally accessible driving-induced quantum chiral transport and Brownian heat pump."
                    },
                    {
                        "title": "Cycle Flux Ranking of Network Analysis in Quantum Thermal Device",
                        "abstract": "Manipulating quantum thermal transport relies on uncovering the principle working cycles of quantum devices. Here, we apply the cycle flux ranking of network analysis to nonequilibrium thermal devices described by graphs of quantum state transitions. To excavate the principal mechanism out of complex transport behaviors, we decompose the quantum-transition network into cycles, calculate the cycle flux by algebraic graph theory, and pick out the dominant cycles with top-ranked fluxes, i.e., the cycle trajectories with highest probabilities. We demonstrate the cycle flux ranking in typical quantum device models, such as a thermal-drag spin-Seebeck pump, and a quantum thermal transistor as thermal switch or heat amplifier. The dominant cycle trajectories indeed elucidate the principal working mechanisms of those quantum devices. The cycle flux analysis provides an alternative perspective that naturally describes the working cycle corresponding to the main functionality of quantum thermal devices, which would further guide the device optimization with desired performance"
                    },
                    {
                        "title": "Convolutional Neural Network Pruning with Structural Redundancy Reduction",
                        "abstract": "Convolutional neural network (CNN) pruning has become one of the most successful network compression approaches in recent years. Existing works on network pruning usually focus on removing the least important filters in the network to achieve compact architectures. In this study, we claim that identifying structural redundancy plays a more essential role than finding unimportant filters, theoretically and empirically. We first statistically model the network pruning problem in a redundancy reduction perspective and find that pruning in the layer(s) with the most structural redundancy outperforms pruning the least important filters across all layers. Based on this finding, we then propose a network pruning approach that identifies structural redundancy of a CNN and prunes filters in the selected layer(s) with the most redundancy. Experiments on various benchmark network architectures and datasets show that our proposed approach significantly outperforms the previous state-of-the-art."
                    },
                    {
                        "title": "Rethinking Diversity in Deep Neural Network Testing",
                        "abstract": "Motivated by the success of traditional software testing, numerous diversity measures have been proposed for testing deep neural networks (DNNs). In this study, we propose a shift in perspective, advocating for the consideration of DNN testing as directed testing problems rather than diversity-based testing tasks. We note that the objective of testing DNNs is specific and well-defined: identifying inputs that lead to misclassifications. Consequently, a more precise testing approach is to prioritize inputs with a higher potential to induce misclassifications, as opposed to emphasizing inputs that enhance \"diversity.\"   We derive six directed metrics for DNN testing. Furthermore, we conduct a careful analysis of the appropriate scope for each metric, as applying metrics beyond their intended scope could significantly diminish their effectiveness. Our evaluation demonstrates that (1) diversity metrics are particularly weak indicators for identifying buggy inputs resulting from small input perturbations, and (2) our directed metrics consistently outperform diversity metrics in revealing erroneous behaviors of DNNs across all scenarios."
                    },
                    {
                        "title": "Optimization as Estimation with Gaussian Processes in Bandit Settings",
                        "abstract": "Recently, there has been rising interest in Bayesian optimization -- the optimization of an unknown function with assumptions usually expressed by a Gaussian Process (GP) prior. We study an optimization strategy that directly uses an estimate of the argmax of the function. This strategy offers both practical and theoretical advantages: no tradeoff parameter needs to be selected, and, moreover, we establish close connections to the popular GP-UCB and GP-PI strategies. Our approach can be understood as automatically and adaptively trading off exploration and exploitation in GP-UCB and GP-PI. We illustrate the effects of this adaptive tuning via bounds on the regret as well as an extensive empirical evaluation on robotics and vision tasks, demonstrating the robustness of this strategy for a range of performance criteria."
                    },
                    {
                        "title": "Single-shot Channel Pruning Based on Alternating Direction Method of Multipliers",
                        "abstract": "Channel pruning has been identified as an effective approach to constructing efficient network structures. Its typical pipeline requires iterative pruning and fine-tuning. In this work, we propose a novel single-shot channel pruning approach based on alternating direction methods of multipliers (ADMM), which can eliminate the need for complex iterative pruning and fine-tuning procedure and achieve a target compression ratio with only one run of pruning and fine-tuning. To the best of our knowledge, this is the first study of single-shot channel pruning. The proposed method introduces filter-level sparsity during training and can achieve competitive performance with a simple heuristic pruning criterion (L1-norm). Extensive evaluations have been conducted with various widely-used benchmark architectures and image datasets for object classification purpose. The experimental results on classification accuracy show that the proposed method can outperform state-of-the-art network pruning works under various scenarios."
                    },
                    {
                        "title": "Efficient Symbolic Reasoning for Neural-Network Verification",
                        "abstract": "The neural network has become an integral part of modern software systems. However, they still suffer from various problems, in particular, vulnerability to adversarial attacks. In this work, we present a novel program reasoning framework for neural-network verification, which we refer to as symbolic reasoning. The key components of our framework are the use of the symbolic domain and the quadratic relation. The symbolic domain has very flexible semantics, and the quadratic relation is quite expressive. They allow us to encode many verification problems for neural networks as quadratic programs. Our scheme then relaxes the quadratic programs to semidefinite programs, which can be efficiently solved. This framework allows us to verify various neural-network properties under different scenarios, especially those that appear challenging for non-symbolic domains. Moreover, it introduces new representations and perspectives for the verification tasks. We believe that our framework can bring new theoretical insights and practical tools to verification problems for neural networks."
                    }
                ]
            }
        }
    },
    "2311.04193": {
        "paper_data": {
            "title": "Selective Visual Representations Improve Convergence and Generalization for Embodied AI",
            "url": "http://arxiv.org/abs/2311.04193v2",
            "arxiv_id": "2311.04193",
            "authors": [
                "Ainaz Eftekhar",
                "Kuo-Hao Zeng",
                "Jiafei Duan",
                "Ali Farhadi",
                "Ani Kembhavi",
                "Ranjay Krishna"
            ],
            "abstract": "Embodied AI models often employ off the shelf vision backbones like CLIP to encode their visual observations. Although such general purpose representations encode rich syntactic and semantic information about the scene, much of this information is often irrelevant to the specific task at hand. This introduces noise within the learning process and distracts the agent's focus from task-relevant visual cues. Inspired by selective attention in humans-the process through which people filter their perception based on their experiences, knowledge, and the task at hand-we introduce a parameter-efficient approach to filter visual stimuli for embodied AI. Our approach induces a task-conditioned bottleneck using a small learnable codebook module. This codebook is trained jointly to optimize task reward and acts as a task-conditioned selective filter over the visual observation. Our experiments showcase state-of-the-art performance for object goal navigation and object displacement across 5 benchmarks, ProcTHOR, ArchitecTHOR, RoboTHOR, AI2-iTHOR, and ManipulaTHOR. The filtered representations produced by the codebook are also able generalize better and converge faster when adapted to other simulation environments such as Habitat. Our qualitative analyses show that agents explore their environments more effectively and their representations retain task-relevant information like target object recognition while ignoring superfluous information about other objects. Code and pretrained models are available at our project website: https://embodied-codebook.github.io.",
            "introduction": "   1 Introduction  Human visual perception is far from a passive reception of all available visual stimuli; it is an actively tuned mechanism that operates selectively, allocating attention and processing stimuli that are deemed relevant to the current task\u00a0(Bugelski & Alampay, 1961). An illustrative example of this phenomenon is the common experience of misplacing one\u2019s keys; we become subsequently oblivious to most visual cues in our environment, except for those directly related to the search for the lost keys. In this case, we become particularly attentive to surfaces where we usually place our keys and navigate our environment by similarly processing the walkable areas around us (Figure\u00a01).   The subfield of embodied artificial intelligence (AI) studies AI agents tasked with very analogous situations (Duan et\u00a0al., 2022). Embodied AI tasks such as navigation\u00a0(Batra et\u00a0al., 2020b; Krantz et\u00a0al., 2023), instruction following\u00a0(Anderson et\u00a0al., 2018; Krantz et\u00a0al., 2020; Shridhar et\u00a0al., 2020), manipulation\u00a0(Ehsani et\u00a0al., 2021; Xiang et\u00a0al., 2020), and rearrangement\u00a0(Batra et\u00a0al., 2020a; Weihs et\u00a0al., 2021a), necessitate goal-directed behaviors, such as navigating to a specific goal or relocating target objects. Conventional frameworks usually employ general-purpose visual backbones\u00a0(Khandelwal et\u00a0al., 2022; Yadav et\u00a0al., 2023b) to extract representations from visual input. These representations are then fused with goal embeddings (e.g., object type, images, or natural language instructions) to construct a goal-conditioned representation E\u2208\u211bD\ud835\udc38superscript\u211b\ud835\udc37E\\in\\mathcal{R}^{D}italic_E \u2208 caligraphic_R start_POSTSUPERSCRIPT italic_D end_POSTSUPERSCRIPT, where D\ud835\udc37Ditalic_D often has dimensions as large as a 1568156815681568-dimensional Hilbert space. E\ud835\udc38Eitalic_E captures an abundance of details from the visual input, of which a policy determines which action to take next. Given E\ud835\udc38Eitalic_E\u2019s general-purpose nature, it often contains a significant amount of task-irrelevant information. For example to find a specific object in a house, the agent doesn\u2019t need to know about other distractor objects in the agent\u2019s view, about their colors, materials, attributes, etc. These distractions introduce unnecessary noise into the learning process, distracting the agent\u2019s focus away from more pertinent visual cues.   Figure 1: Selective Attention. Imagine an agent is tasked to locate a key in an environment. Standard visual encoders such as CLIP encoder capture general purpose scene information which include details not relevant to the task, such as the color of the sofa or texture of the floor. This mirrors the concept of bottom-up processing, where perception is influenced by external stimuli in the environment. To address this, we equip the encoder with a codebook bottleneck that only retains the most task-relevant information such as identifying flat surfaces likely to hold the key and the walkable paths to these surfaces. This represents top-down selective processing where the perception is guided by internal goals and expectations.   In this paper, we draw from the substantial body of research in cognitive psychology and neuroscience to induce selective task-specific representations that filter irrelevant sensory input and only retain the necessary stimuli. In Psychology, inattentional blindness posits that people become \u201cblind\u201d to objects and details irrelevant to the task at hand (as depicted by the famous invisible gorilla video on YouTube111Link to invisible gorilla video depicting selective attention)\u00a0(Simons & Chabris, 1999). Similar studies in human visual search\u00a0(Wolfe, 1994; Treisman & Gelade, 1980) find that people become oblivious to the distracting items when searching through a cluttered environment. The principle",
            "references": [
                {
                    "title": "Minigrid & Miniworld: Modular & Customizable Reinforcement Learning Environments for Goal-Oriented Tasks",
                    "abstract": "We present the Minigrid and Miniworld libraries which provide a suite of goal-oriented 2D and 3D environments. The libraries were explicitly created with a minimalistic design paradigm to allow users to rapidly develop new environments for a wide range of research-specific needs. As a result, both have received widescale adoption by the RL community, facilitating research in a wide range of areas. In this paper, we outline the design philosophy, environment details, and their world generation API. We also showcase the additional capabilities brought by the unified API between Minigrid and Miniworld through case studies on transfer learning (for both RL agents and humans) between the different observation spaces. The source code of Minigrid and Miniworld can be found at https://github.com/Farama-Foundation/{Minigrid, Miniworld} along with their documentation at https://{minigrid, miniworld}.farama.org/."
                },
                {
                    "title": "Disentanglement via Latent Quantization",
                    "abstract": "In disentangled representation learning, a model is asked to tease apart a dataset's underlying sources of variation and represent them independently of one another. Since the model is provided with no ground truth information about these sources, inductive biases take a paramount role in enabling disentanglement. In this work, we construct an inductive bias towards encoding to and decoding from an organized latent space. Concretely, we do this by (i) quantizing the latent space into discrete code vectors with a separate learnable scalar codebook per dimension and (ii) applying strong model regularization via an unusually high weight decay. Intuitively, the latent space design forces the encoder to combinatorially construct codes from a small number of distinct scalar values, which in turn enables the decoder to assign a consistent meaning to each value. Regularization then serves to drive the model towards this parsimonious strategy. We demonstrate the broad applicability of this approach by adding it to both basic data-reconstructing (vanilla autoencoder) and latent-reconstructing (InfoGAN) generative models. For reliable evaluation, we also propose InfoMEC, a new set of metrics for disentanglement that is cohesively grounded in information theory and fixes well-established shortcomings in previous metrics. Together with regularization, latent quantization dramatically improves the modularity and explicitness of learned representations on a representative suite of benchmark datasets. In particular, our quantized-latent autoencoder (QLAE) consistently outperforms strong methods from prior work in these key disentanglement properties without compromising data reconstruction."
                },
                {
                    "title": "Moving Forward by Moving Backward: Embedding Action Impact over Action Semantics",
                    "abstract": "A common assumption when training embodied agents is that the impact of taking an action is stable; for instance, executing the\"move ahead\"action will always move the agent forward by a fixed distance, perhaps with some small amount of actuator-induced noise. This assumption is limiting; an agent may encounter settings that dramatically alter the impact of actions: a move ahead action on a wet floor may send the agent twice as far as it expects and using the same action with a broken wheel might transform the expected translation into a rotation. Instead of relying that the impact of an action stably reflects its pre-defined semantic meaning, we propose to model the impact of actions on-the-fly using latent embeddings. By combining these latent action embeddings with a novel, transformer-based, policy head, we design an Action Adaptive Policy (AAP). We evaluate our AAP on two challenging visual navigation tasks in the AI2-THOR and Habitat environments and show that our AAP is highly performant even when faced, at inference-time with missing actions and, previously unseen, perturbed action space. Moreover, we observe significant improvement in robustness against these actions when evaluating in real-world scenarios."
                },
                {
                    "title": "AutoNeRF: Training Implicit Scene Representations with Autonomous Agents",
                    "abstract": "Implicit representations such as Neural Radiance Fields (NeRF) have been shown to be very effective at novel view synthesis. However, these models typically require manual and careful human data collection for training. In this paper, we present AutoNeRF, a method to collect data required to train NeRFs using autonomous embodied agents. Our method allows an agent to explore an unseen environment efficiently and use the experience to build an implicit map representation autonomously. We compare the impact of different exploration strategies including handcrafted frontier-based exploration, end-to-end and modular approaches composed of trained high-level planners and classical low-level path followers. We train these models with different reward functions tailored to this problem and evaluate the quality of the learned representations on four different downstream tasks: classical viewpoint rendering, map reconstruction, planning, and pose refinement. Empirical results show that NeRFs can be trained on actively collected data using just a single episode of experience in an unseen environment, and can be used for several downstream robotic tasks, and that modular trained exploration models outperform other classical and end-to-end baselines. Finally, we show that AutoNeRF can reconstruct large-scale scenes, and is thus a useful tool to perform scene-specific adaptation as the produced 3D environment models can be loaded into a simulator to fine-tune a policy of interest."
                },
                {
                    "title": "DINOv2: Learning Robust Visual Features without Supervision",
                    "abstract": "The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels."
                },
                {
                    "title": "ENTL: Embodied Navigation Trajectory Learner",
                    "abstract": "We propose Embodied Navigation Trajectory Learner (ENTL), a method for extracting long sequence representations for embodied navigation. Our approach unifies world modeling, localization and imitation learning into a single sequence prediction task. We train our model using vector-quantized predictions of future states conditioned on current states and actions. ENTL\u2019s generic architecture enables the sharing of the the spatio-temporal sequence encoder for multiple challenging embodied tasks. We achieve competitive performance on navigation tasks using significantly less data than strong baselines while performing auxiliary tasks such as localization and future frame prediction (a proxy for world modeling). A key property of our approach is that the model is pre-trained without any explicit reward signal, which makes the resulting model generalizable to multiple tasks and environments. We release the code at https://github.com/klemenkotar/ENTL"
                },
                {
                    "title": "Navigating to Objects Specified by Images",
                    "abstract": "Images are a convenient way to specify which particular object instance an embodied agent should navigate to. Solving this task requires semantic visual reasoning and exploration of unknown environments. We present a system that can perform this task in both simulation and the real world. Our modular method solves sub-tasks of exploration, goal instance re-identification, goal localization, and local navigation. We re-identify the goal instance in egocentric vision using feature-matching and localize the goal instance by projecting matched features to a map. Each sub-task is solved using off-the-shelf components requiring zero fine-tuning. On the HM3D InstanceImageNav benchmark, this system outperforms a baseline end-to-end RL policy 7x and a state-of-the-art ImageNav model 2.3x (56% vs . 25% success). We deploy this system to a mobile robot platform and demonstrate effective real-world performance, achieving an 88% success rate across a home and an office environment."
                },
                {
                    "title": "Revisiting Multimodal Representation in Contrastive Learning: From Patch and Token Embeddings to Finite Discrete Tokens",
                    "abstract": "Contrastive learning-based vision-language pretraining approaches, such as CLIP, have demonstrated great success in many vision-language tasks. These methods achieve cross-modal alignment by encoding a matched image-text pair with similar feature embeddings, which are generated by aggregating information from visual patches and language tokens. However, direct aligning cross-modal information using such representations is challenging, as visual patches and text tokens differ in semantic levels and granularities. To alleviate this issue, we propose a Finite Discrete Tokens (FDT) based multimodal representation. FDT is a set of learnable tokens representing certain visualsemantic concepts. Both images and texts are embedded using shared FDT by first grounding multimodal inputs to FDT space and then aggregating the activated FDT representations. The matched visual and semantic concepts are enforced to be represented by the same set of discrete tokens by a sparse activation constraint. As a result, the granularity gap between the two modalities is reduced. Through both quantitative and qualitative analyses, we demonstrate that using FDT representations in CLIP-style models improves cross-modal alignment and performance in visual recognition and vision-language downstream tasks. Furthermore, we show that our method can learn more comprehensive representations, and the learned FDT capture meaningful cross-modal correspondence, ranging from objects to actions and attributes.11The source code can be found at https://github.com/yuxiaochen1103/FDT."
                },
                {
                    "title": "OVRL-V2: A simple state-of-art baseline for ImageNav and ObjectNav",
                    "abstract": "We present a single neural network architecture composed of task-agnostic components (ViTs, convolutions, and LSTMs) that achieves state-of-art results on both the ImageNav (\"go to location in\") and ObjectNav (\"find a chair\") tasks without any task-specific modules like object detection, segmentation, mapping, or planning modules. Such general-purpose methods offer advantages of simplicity in design, positive scaling with available compute, and versatile applicability to multiple tasks. Our work builds upon the recent success of self-supervised learning (SSL) for pre-training vision transformers (ViT). However, while the training recipes for convolutional networks are mature and robust, the recipes for ViTs are contingent and brittle, and in the case of ViTs for visual navigation, yet to be fully discovered. Specifically, we find that vanilla ViTs do not outperform ResNets on visual navigation. We propose the use of a compression layer operating over ViT patch representations to preserve spatial information along with policy training improvements. These improvements allow us to demonstrate positive scaling laws for the first time in visual navigation tasks. Consequently, our model advances state-of-the-art performance on ImageNav from 54.2% to 82.0% success and performs competitively against concurrent state-of-art on ObjectNav with success rate of 64.0% vs. 65.0%. Overall, this work does not present a fundamentally new approach, but rather recommendations for training a general-purpose architecture that achieves state-of-art performance today and could serve as a strong baseline for future methods."
                },
                {
                    "title": "Neural Radiance Field Codebooks",
                    "abstract": "Compositional representations of the world are a promising step towards enabling high-level scene understanding and efficient transfer to downstream tasks. Learning such representations for complex scenes and tasks remains an open challenge. Towards this goal, we introduce Neural Radiance Field Codebooks (NRC), a scalable method for learning object-centric representations through novel view reconstruction. NRC learns to reconstruct scenes from novel views using a dictionary of object codes which are decoded through a volumetric renderer. This enables the discovery of reoccurring visual and geometric patterns across scenes which are transferable to downstream tasks. We show that NRC representations transfer well to object navigation in THOR, outperforming 2D and 3D representation learning methods by 3.1% success rate. We demonstrate that our approach is able to perform unsupervised segmentation for more complex synthetic (THOR) and real scenes (NYU Depth) better than prior methods (29% relative improvement). Finally, we show that NRC improves on the task of depth ordering by 5.5% accuracy in THOR."
                },
                {
                    "title": "A General Purpose Supervisory Signal for Embodied Agents",
                    "abstract": "Training effective embodied AI agents often involves manual reward engineering, expert imitation, specialized components such as maps, or leveraging additional sensors for depth and localization. Another approach is to use neural architectures alongside self-supervised objectives which encourage better representation learning. In practice, there are few guarantees that these self-supervised objectives encode task-relevant information. We propose the Scene Graph Contrastive (SGC) loss, which uses scene graphs as general-purpose, training-only, supervisory signals. The SGC loss does away with explicit graph decoding and instead uses contrastive learning to align an agent's representation with a rich graphical encoding of its environment. The SGC loss is generally applicable, simple to implement, and encourages representations that encode objects' semantics, relationships, and history. Using the SGC loss, we attain significant gains on three embodied tasks: Object Navigation, Multi-Object Navigation, and Arm Point Navigation. Finally, we present studies and analyses which demonstrate the ability of our trained representation to encode semantic cues about the environment."
                },
                {
                    "title": "ProcTHOR: Large-Scale Embodied AI Using Procedural Generation",
                    "abstract": "Massive datasets and high-capacity models have driven many recent advancements in computer vision and natural language understanding. This work presents a platform to enable similar success stories in Embodied AI. We propose ProcTHOR, a framework for procedural generation of Embodied AI environments. ProcTHOR enables us to sample arbitrarily large datasets of diverse, interactive, customizable, and performant virtual environments to train and evaluate embodied agents across navigation, interaction, and manipulation tasks. We demonstrate the power and potential of ProcTHOR via a sample of 10,000 generated houses and a simple neural model. Models trained using only RGB images on ProcTHOR, with no explicit mapping and no human task supervision produce state-of-the-art results across 6 embodied AI benchmarks for navigation, rearrangement, and arm manipulation, including the presently running Habitat 2022, AI2-THOR Rearrangement 2022, and RoboTHOR challenges. We also demonstrate strong 0-shot results on these benchmarks, via pre-training on ProcTHOR with no fine-tuning on the downstream benchmark, often beating previous state-of-the-art systems that access the downstream training data."
                },
                {
                    "title": "UViM: A Unified Modeling Approach for Vision with Learned Guiding Codes",
                    "abstract": "We introduce UViM, a unified approach capable of modeling a wide range of computer vision tasks. In contrast to previous models, UViM has the same functional form for all tasks; it requires no task-specific modifications which require extensive human expertise. The approach involves two components: (I) a base model (feed-forward) which is trained to directly predict raw vision outputs, guided by a learned discrete code and (II) a language model (autoregressive) that is trained to generate the guiding code. These components complement each other: the language model is well-suited to modeling structured interdependent data, while the base model is efficient at dealing with high-dimensional outputs. We demonstrate the effectiveness of UViM on three diverse and challenging vision tasks: panoptic segmentation, depth prediction and image colorization, where we achieve competitive and near state-of-the-art results. Our experimental results suggest that UViM is a promising candidate for a unified modeling approach in computer vision."
                },
                {
                    "title": "Offline Visual Representation Learning for Embodied Navigation",
                    "abstract": "How should we learn visual representations for embodied agents that must see and move? The status quo is tabula rasa in vivo, i.e. learning visual representations from scratch while also learning to move, potentially augmented with auxiliary tasks (e.g. predicting the action taken between two successive observations). In this paper, we show that an alternative 2-stage strategy is far more effective: (1) offline pretraining of visual representations with self-supervised learning (SSL) using large-scale pre-rendered images of indoor environments (Omnidata), and (2) online finetuning of visuomotor representations on specific tasks with image augmentations under long learning schedules. We call this method Offline Visual Representation Learning (OVRL). We conduct large-scale experiments - on 3 different 3D datasets (Gibson, HM3D, MP3D), 2 tasks (ImageNav, ObjectNav), and 2 policy learning algorithms (RL, IL) - and find that the OVRL representations lead to significant across-the-board improvements in state of art, on ImageNav from 29.2% to 54.2% (+25% absolute, 86% relative) and on ObjectNav from 18.1% to 23.2% (+5.1% absolute, 28% relative). Importantly, both results were achieved by the same visual encoder generalizing to datasets that were not seen during pretraining. While the benefits of pretraining sometimes diminish (or entirely disappear) with long finetuning schedules, we find that OVRL's performance gains continue to increase (not decrease) as the agent is trained for 2 billion frames of experience."
                },
                {
                    "title": "Continuous Scene Representations for Embodied AI",
                    "abstract": "We propose Continuous Scene Representations (CSR), a scene representation constructed by an embodied agent navigating within a space, where objects and their relationships are modeled by continuous valued embeddings. Our method captures feature relationships between objects, composes them into a graph structure on-the-fly, and situates an embodied agent within the representation. Our key insight is to embed pair-wise relationships between objects in a latent space. This allows for a richer representation compared to discrete relations (e.g., [SUPPORT], [NEXT-TO]) commonly used for building scene representations. CSR can track objects as the agent moves in a scene, update the representation accordingly, and detect changes in room configurations. Using CSR, we outperform state-of-the-art approaches for the challenging downstream task of visual room rearrangement, without any task specific training. Moreover, we show the learned embeddings capture salient spatial details of the scene and show applicability to real world data. A summery video and code is available at prior.allenai.org/projects/csr."
                },
                {
                    "title": "Object Manipulation via Visual Target Localization",
                    "abstract": "Object manipulation is a critical skill required for Embodied AI agents interacting with the world around them. Training agents to manipulate objects, poses many challenges. These include occlusion of the target object by the agent's arm, noisy object detection and localization, and the target frequently going out of view as the agent moves around in the scene. We propose Manipulation via Visual Object Location Estimation (m-VOLE), an approach that explores the environment in search for target objects, computes their 3D coordinates once they are located, and then continues to estimate their 3D locations even when the objects are not visible, thus robustly aiding the task of manipulating these objects throughout the episode. Our evaluations show a massive 3x improvement in success rate over a model that has access to the same sensory suite but is trained without the object location estimator, and our analysis shows that our agent is robust to noise in depth perception and agent localization. Importantly, our proposed approach relaxes several assumptions about idealized localization and perception that are commonly employed by recent works in embodied AI -- an important step towards training agents for object manipulation in the real world."
                },
                {
                    "title": "Simple but Effective: CLIP Embeddings for Embodied AI",
                    "abstract": "Contrastive language image pretraining (CLIP) encoders have been shown to be beneficial for a range of visual tasks from classification and detection to captioning and image manipulation. We investigate the effectiveness of CLIP visual backbones for Embodied AI tasks. We build incredibly simple baselines, named EmbCLIP, with no task specific architectures, inductive biases (such as the use of semantic maps), auxiliary tasks during training, or depth maps-yet we find that our improved baselines perform very well across a range of tasks and simulators. EmbCLIP tops the RoboTHOR ObjectNav leader-board by a huge margin of 20 pts (Success Rate). It tops the iTHOR 1-Phase Rearrangement leaderboard, beating the next best submission, which employs Active Neural Mapping, and more than doubling the % Fixed Strict metric (0.08 to 0.17). It also beats the winners of the 2021 Habitat ObjectNav Challenge, which employ auxiliary tasks, depth maps, and human demonstrations, and those of the 2019 Habitat PointNav Challenge. We evaluate the ability of CLIP's visual representations at capturing semantic information about input observations-primitives that are useful for navigation-heavy embodied tasks- and find that CLIP's representations encode these primitives more effectively than ImageNet-pretrained backbones. Finally, we extend one of our baselines, producing an agent capable of zero-shot object navigation that can navigate to objects that were not used as targets during training. Our code and models are available at https://github.com/allenai/embodied-clip."
                },
                {
                    "title": "Dynamic Bottleneck for Robust Self-Supervised Exploration",
                    "abstract": "Exploration methods based on pseudo-count of transitions or curiosity of dynamics have achieved promising results in solving reinforcement learning with sparse rewards. However, such methods are usually sensitive to environmental dynamics-irrelevant information, e.g., white-noise. To handle such dynamics-irrelevant information, we propose a Dynamic Bottleneck (DB) model, which attains a dynamics-relevant representation based on the information-bottleneck principle. Based on the DB model, we further propose DB-bonus, which encourages the agent to explore state-action pairs with high information gain. We establish theoretical connections between the proposed DB-bonus, the upper confidence bound (UCB) for linear case, and the visiting count for tabular case. We evaluate the proposed method on Atari suits with dynamics-irrelevant noises. Our experiments show that exploration with DB bonus outperforms several state-of-the-art exploration methods in noisy environments."
                },
                {
                    "title": "Variational Information Bottleneck for Effective Low-Resource Fine-Tuning",
                    "abstract": "While large-scale pretrained language models have obtained impressive results when fine-tuned on a wide variety of tasks, they still often suffer from overfitting in low-resource scenarios. Since such models are general-purpose feature extractors, many of these features are inevitably irrelevant for a given target task. We propose to use Variational Information Bottleneck (VIB) to suppress irrelevant features when fine-tuning on low-resource target tasks, and show that our method successfully reduces overfitting. Moreover, we show that our VIB model finds sentence representations that are more robust to biases in natural language inference datasets, and thereby obtains better generalization to out-of-domain datasets. Evaluation on seven low-resource datasets in different tasks shows that our method significantly improves transfer learning in low-resource scenarios, surpassing prior work. Moreover, it improves generalization on 13 out of 15 out-of-domain natural language inference benchmarks. Our code is publicly available in https://github.com/rabeehk/vibert."
                },
                {
                    "title": "Scaling Vision with Sparse Mixture of Experts",
                    "abstract": "Sparsely-gated Mixture of Experts networks (MoEs) have demonstrated excellent scalability in Natural Language Processing. In Computer Vision, however, almost all performant networks are\"dense\", that is, every input is processed by every parameter. We present a Vision MoE (V-MoE), a sparse version of the Vision Transformer, that is scalable and competitive with the largest dense networks. When applied to image recognition, V-MoE matches the performance of state-of-the-art networks, while requiring as little as half of the compute at inference time. Further, we propose an extension to the routing algorithm that can prioritize subsets of each input across the entire batch, leading to adaptive per-image compute. This allows V-MoE to trade-off performance and compute smoothly at test-time. Finally, we demonstrate the potential of V-MoE to scale vision models, and train a 15B parameter model that attains 90.35% on ImageNet."
                },
                {
                    "title": "Curious Representation Learning for Embodied Intelligence",
                    "abstract": "Self-supervised representation learning has achieved remarkable success in recent years. By subverting the need for supervised labels, such approaches are able to utilize the numerous unlabeled images that exist on the Internet and in photographic datasets. Yet to build truly intelligent agents, we must construct representation learning algorithms that can learn not only from datasets but also learn from environments. An agent in a natural environment will not typically be fed curated data. Instead, it must explore its environment to acquire the data it will learn from. We propose a framework, curious representation learning (CRL), which jointly learns a reinforcement learning policy and a visual representation model. The policy is trained to maximize the error of the representation learner, and in doing so is incentivized to explore its environment. At the same time, the learned representation becomes stronger and stronger as the policy feeds it ever harder data to learn from. Our learned representations enable promising transfer to downstream navigation tasks, performing better than or comparably to ImageNet pretraining without using any supervision at all. In addition, despite being trained in simulation, our learned representations can obtain interpretable results on real images. Code is available at https://yilundu.github.io/crl/."
                },
                {
                    "title": "Pushing it out of the Way: Interactive Visual Navigation",
                    "abstract": "We have observed significant progress in visual navigation for embodied agents. A common assumption in studying visual navigation is that the environments are static; this is a limiting assumption. Intelligent navigation may involve interacting with the environment beyond just moving forward/backward and turning left/right. Sometimes, the best way to navigate is to push something out of the way. In this paper, we study the problem of interactive navigation where agents learn to change the environment to navigate more efficiently to their goals. To this end, we introduce the Neural Interaction Engine (NIE) to explicitly predict the change in the environment caused by the agent\u2019s actions. By modeling the changes while planning, we find that agents exhibit significant improvements in their navigational capabilities. More specifically, we consider two downstream tasks in the physics-enabled, visually rich, AI2-THOR environment: (1) reaching a target while the path to the target is blocked (2) moving an object to a target location by pushing it. For both tasks, agents equipped with an NIE significantly outperform agents without the understanding of the effect of the actions indicating the benefits of our approach. The code and dataset are available at github.com/KuoHaoZeng/Interactive_Visual_Navigation."
                },
                {
                    "title": "ManipulaTHOR: A Framework for Visual Object Manipulation",
                    "abstract": "The domain of Embodied AI has recently witnessed substantial progress, particularly in navigating agents within their environments. These early successes have laid the building blocks for the community to tackle tasks that require agents to actively interact with objects in their environment. Object manipulation is an established research domain within the robotics community and poses several challenges including manipulator motion, grasping and long-horizon planning, particularly when dealing with oft-overlooked practical setups involving visually rich and complex scenes, manipulation using mobile agents (as opposed to tabletop manipulation), and generalization to unseen environments and objects. We propose a framework for object manipulation built upon the physics-enabled, visually rich AI2-THOR framework and present a new challenge to the Embodied AI community known as ArmPointNav. This task extends the popular point navigation task [2] to object manipulation and offers new challenges including 3D obstacle avoidance, manipulating objects in the presence of occlusion, and multi-object manipulation that necessitates long term planning. Popular learning paradigms that are successful on PointNav challenges show promise, but leave a large room for improvement."
                },
                {
                    "title": "Auxiliary Tasks and Exploration Enable ObjectGoal Navigation",
                    "abstract": "ObjectGoal Navigation (ObjectNav) is an embodied task wherein agents are to navigate to an object instance in an unseen environment. Prior works have shown that end-to-end ObjectNav agents that use vanilla visual and recurrent modules, e.g. a CNN+RNN, perform poorly due to overfitting and sample inefficiency. This has motivated current state-of-the-art methods to mix analytic and learned components and operate on explicit spatial maps of the environment. We instead re-enable a generic learned agent by adding auxiliary learning tasks and an exploration reward. Our agents achieve 24.5% success and 8.1% SPL, a 37% and 8% relative improvement over prior state-of-the-art, respectively, on the Habitat ObjectNav Challenge [35]. From our analysis, we propose that agents will act to simplify their visual inputs so as to smooth their RNN dynamics, and that auxiliary tasks reduce overfitting by minimizing effective RNN dimensionality; i.e. a performant ObjectNav agent that must maintain coherent plans over long horizons does so by learning smooth, low-dimensional recurrent dynamics. Site: joel99.github.io/objectnav/"
                },
                {
                    "title": "Visual Room Rearrangement",
                    "abstract": "There has been a significant recent progress in the field of Embodied AI with researchers developing models and algorithms enabling embodied agents to navigate and interact within completely unseen environments. In this paper, we propose a new dataset and baseline models for the task of Rearrangement. We particularly focus on the task of Room Rearrangement: an agent begins by exploring a room and recording objects\u2019 initial configurations. We then remove the agent and change the poses and states (e.g., open/closed) of some objects in the room. The agent must restore the initial configurations of all objects in the room. Our dataset, named RoomR, includes 6,000 distinct rearrangement settings involving 72 different object types in 120 scenes. Our experiments show that solving this challenging interactive task that involves navigation and object interaction is beyond the capabilities of the current state-of-the-art techniques for embodied tasks and we are still very far from achieving perfect performance on these types of tasks."
                },
                {
                    "title": "A Survey of Embodied AI: From Simulators to Research Tasks",
                    "abstract": "There has been an emerging paradigm shift from the era of \u201cinternet AI\u201d to \u201cembodied AI,\u201d where AI algorithms and agents no longer learn from datasets of images, videos or text curated primarily from the internet. Instead, they learn through interactions with their environments from an egocentric perception similar to humans. Consequently, there has been substantial growth in the demand for embodied AI simulators to support various embodied AI research tasks. This growing interest in embodied AI is beneficial to the greater pursuit of Artificial General Intelligence (AGI), but there has not been a contemporary and comprehensive survey of this field. This paper aims to provide an encyclopedic survey for the field of embodied AI, from its simulators to its research. By evaluating nine current embodied AI simulators with our proposed seven features, this paper aims to understand the simulators in their provision for use in embodied AI research and their limitations. Lastly, this paper surveys the three main research tasks in embodied AI \u2013 visual exploration, visual navigation and embodied question answering (QA), covering the state-of-the-art approaches, evaluation metrics and datasets. Finally, with the new insights revealed through surveying the field, the paper will provide suggestions for simulator-for-task selections and recommendations for the future directions of the field."
                },
                {
                    "title": "Robust Deep Reinforcement Learning via Multi-View Information Bottleneck",
                    "abstract": "Deep reinforcement learning (DRL) agents are often sensitive to visual changes that were unseen in their training environments. To address this problem, we leverage the sequential nature of RL to learn robust representations that encode only task-relevant information from observations based on the unsupervised multi-view setting. Specifically, we introduce a novel contrastive version of the Multi-View Information Bottleneck (MIB) objective for temporal data. We train RL agents from pixels with this auxiliary objective to learn robust representations that can compress away task-irrelevant information and are predictive of task-relevant dynamics. This approach enables us to train high-performance policies that are robust to visual distractions and can generalize well to unseen environments. We demonstrate that our approach can achieve SOTA performance on a diverse set of visual control tasks in the DeepMind Control Suite when the background is replaced with natural videos. In addition, we show that our approach outperforms well-established baselines for generalization to unseen environments on the Procgen benchmark. Our code is open-sourced and available at https://github. com/BU-DEPEND-Lab/DRIBO."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Bottleneck Transformers for Visual Recognition",
                    "abstract": "We present BoTNet, a conceptually simple yet powerful backbone architecture that incorporates self-attention for multiple computer vision tasks including image classification, object detection and instance segmentation. By just replacing the spatial convolutions with global self-attention in the final three bottleneck blocks of a ResNet and no other changes, our approach improves upon the baselines significantly on instance segmentation and object detection while also reducing the parameters, with minimal overhead in latency. Through the design of BoTNet, we also point out how ResNet bottleneck blocks with self-attention can be viewed as Transformer blocks. Without any bells and whistles, BoTNet achieves 44.4% Mask AP and 49.7% Box AP on the COCO Instance Segmentation benchmark using the Mask R-CNN framework; surpassing the previous best published single model and single scale results of ResNeSt [67] evaluated on the COCO validation set. Finally, we present a simple adaptation of the BoTNet design for image classification, resulting in models that achieve a strong performance of 84.7% top-1 accuracy on the ImageNet benchmark while being up to 1.64x faster in \"compute\"1 time than the popular EfficientNet models on TPU-v3 hardware. We hope our simple and effective approach will serve as a strong baseline for future research in self-attention models for vision.2"
                },
                {
                    "title": "Memory-Augmented Reinforcement Learning for Image-Goal Navigation",
                    "abstract": "In this work, we present a memory-augmented approach for image-goal navigation. Earlier attempts, including RL-based and SLAM-based approaches have either shown poor generalization performance, or are heavily-reliant on pose/depth sensors. Our method is based on an attention-based end-to-end model that leverages an episodic memory to learn to navigate. First, we train a state-embedding network in a self-supervised fashion, and then use it to embed previously-visited states into the agent's memory. Our navigation policy takes advantage of this information through an attention mechanism. We validate our approach with extensive evaluations, and show that our model establishes a new state of the art on the challenging Gibson dataset. Furthermore, we achieve this impressive performance from RGB input alone, without access to additional information such as position or depth, in stark contrast to related work."
                },
                {
                    "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity",
                    "abstract": "In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with outrageous numbers of parameters -- but a constant computational cost. However, despite several notable successes of MoE, widespread adoption has been hindered by complexity, communication costs and training instability -- we address these with the Switch Transformer. We simplify the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs. Our proposed training techniques help wrangle the instabilities and we show large sparse models may be trained, for the first time, with lower precision (bfloat16) formats. We design models based off T5-Base and T5-Large to obtain up to 7x increases in pre-training speed with the same computational resources. These improvements extend into multilingual settings where we measure gains over the mT5-Base version across all 101 languages. Finally, we advance the current scale of language models by pre-training up to trillion parameter models on the\"Colossal Clean Crawled Corpus\"and achieve a 4x speedup over the T5-XXL model."
                },
                {
                    "title": "Taming Transformers for High-Resolution Image Synthesis",
                    "abstract": "Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks. In contrast to CNNs, they contain no inductive bias that prioritizes local interactions. This makes them expressive, but also computationally infeasible for long sequences, such as high-resolution images. We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images. We show how to (i) use CNNs to learn a context-rich vocabulary of image constituents, and in turn (ii) utilize transformers to efficiently model their composition within high-resolution images. Our approach is readily applied to conditional synthesis tasks, where both non-spatial information, such as object classes, and spatial information, such as segmentations, can control the generated image. In particular, we present the first results on semantically-guided synthesis of megapixel images with transformers. Project page at https://git.io/JLlvY."
                },
                {
                    "title": "Rearrangement: A Challenge for Embodied AI",
                    "abstract": "We describe a framework for research and evaluation in Embodied AI. Our proposal is based on a canonical task: Rearrangement. A standard task can focus the development of new techniques and serve as a source of trained models that can be transferred to other settings. In the rearrangement task, the goal is to bring a given physical environment into a specified state. The goal state can be specified by object poses, by images, by a description in language, or by letting the agent experience the environment in the goal state. We characterize rearrangement scenarios along different axes and describe metrics for benchmarking rearrangement performance. To facilitate research and exploration, we present experimental testbeds of rearrangement scenarios in four different simulation environments. We anticipate that other datasets will be released and new simulation platforms will be built to support training of rearrangement agents and their deployment on physical systems."
                },
                {
                    "title": "Dynamics Generalization via Information Bottleneck in Deep Reinforcement Learning",
                    "abstract": "Despite the significant progress of deep reinforcement learning (RL) in solving sequential decision making problems, RL agents often overfit to training environments and struggle to adapt to new, unseen environments. This prevents robust applications of RL in real world situations, where system dynamics may deviate wildly from the training settings. In this work, our primary contribution is to propose an information theoretic regularization objective and an annealing-based optimization method to achieve better generalization ability in RL agents. We demonstrate the extreme generalization benefits of our approach in different domains ranging from maze navigation to robotic tasks; for the first time, we show that agents can generalize to test parameters more than 10 standard deviations away from the training parameter distribution. This work provides a principled way to improve generalization in RL by gradually removing information that is redundant for task-solving; it opens doors for the systematic study of generalization from training to extremely different testing settings, focusing on the established connections between information theory and machine learning."
                },
                {
                    "title": "Auxiliary Tasks Speed Up Learning PointGoal Navigation",
                    "abstract": "PointGoal Navigation is an embodied task that requires agents to navigate to a specified point in an unseen environment. Wijmans et al. showed that this task is solvable but their method is computationally prohibitive, requiring 2.5 billion frames and 180 GPU-days. In this work, we develop a method to significantly increase sample and time efficiency in learning PointNav using self-supervised auxiliary tasks (e.g. predicting the action taken between two egocentric observations, predicting the distance between two observations from a trajectory,etc.).We find that naively combining multiple auxiliary tasks improves sample efficiency,but only provides marginal gains beyond a point. To overcome this, we use attention to combine representations learnt from individual auxiliary tasks. Our best agent is 5x faster to reach the performance of the previous state-of-the-art, DD-PPO, at 40M frames, and improves on DD-PPO's performance at40M frames by 0.16 SPL. Our code is publicly available at this https URL."
                },
                {
                    "title": "Concept Bottleneck Models",
                    "abstract": "We seek to learn models that we can interact with using high-level concepts: if the model did not think there was a bone spur in the x-ray, would it still predict severe arthritis? State-of-the-art models today do not typically support the manipulation of concepts like \u201cthe existence of bone spurs\u201d, as they are trained end-to-end to go directly from raw input (e.g., pixels) to output (e.g., arthritis severity). We revisit the classic idea of first predicting concepts that are provided at training time, and then using these concepts to predict the label. By construction, we can intervene on these concept bottleneck models by editing their predicted concept values and propagating these changes to the final prediction. On x-ray grading and bird identification, concept bottleneck models achieve competitive accuracy with standard end-to-end models, while enabling interpretation in terms of high-level clinical concepts (\u201cbone spurs\u201d) or bird attributes (\u201cwing color\u201d). These models also allow for richer human-model interaction: accuracy improves significantly if we can correct model mistakes on concepts at test time."
                },
                {
                    "title": "Object Goal Navigation using Goal-Oriented Semantic Exploration",
                    "abstract": "This work studies the problem of object goal navigation which involves navigating to an instance of the given object category in unseen environments. End-to-end learning-based navigation methods struggle at this task as they are ineffective at exploration and long-term planning. We propose a modular system called, `Goal-Oriented Semantic Exploration' which builds an episodic semantic map and uses it to explore the environment efficiently based on the goal object category. Empirical results in visually realistic simulation environments show that the proposed model outperforms a wide range of baselines including end-to-end learning-based methods as well as modular map-based methods and led to the winning entry of the CVPR-2020 Habitat ObjectNav Challenge. Ablation analysis indicates that the proposed model learns semantic priors of the relative arrangement of objects in a scene, and uses them to explore efficiently. Domain-agnostic module design allow us to transfer our model to a mobile robot platform and achieve similar performance for object goal navigation in the real-world."
                },
                {
                    "title": "ObjectNav Revisited: On Evaluation of Embodied Agents Navigating to Objects",
                    "abstract": "We revisit the problem of Object-Goal Navigation (ObjectNav). In its simplest form, ObjectNav is defined as the task of navigating to an object, specified by its label, in an unexplored environment. In particular, the agent is initialized at a random location and pose in an environment and asked to find an instance of an object category, e.g., find a chair, by navigating to it. \nAs the community begins to show increased interest in semantic goal specification for navigation tasks, a number of different often-inconsistent interpretations of this task are emerging. This document summarizes the consensus recommendations of this working group on ObjectNav. In particular, we make recommendations on subtle but important details of evaluation criteria (for measuring success when navigating towards a target object), the agent's embodiment parameters, and the characteristics of the environments within which the task is carried out. Finally, we provide a detailed description of the instantiation of these recommendations in challenges organized at the Embodied AI workshop at CVPR 2020 this http URL ."
                },
                {
                    "title": "Bootstrap Latent-Predictive Representations for Multitask Reinforcement Learning",
                    "abstract": "Learning a good representation is an essential component for deep reinforcement learning (RL). Representation learning is especially important in multitask and partially observable settings where building a representation of the unknown environment is crucial to solve the tasks. Here we introduce Prediction of Bootstrap Latents (PBL), a simple and flexible self-supervised representation learning algorithm for multitask deep RL. PBL builds on multistep predictive representations of future observations, and focuses on capturing structured information about environment dynamics. Specifically, PBL trains its representation by predicting latent embeddings of future observations. These latent embeddings are themselves trained to be predictive of the aforementioned representations. These predictions form a bootstrapping effect, allowing the agent to learn more about the key aspects of the environment dynamics. In addition, by defining prediction tasks completely in latent space, PBL provides the flexibility of using multimodal observations involving pixel images, language instructions, rewards and more. We show in our experiments that PBL delivers across-the-board improved performance over state of the art deep RL agents in the DMLab-30 and Atari-57 multitask setting."
                },
                {
                    "title": "RoboTHOR: An Open Simulation-to-Real Embodied AI Platform",
                    "abstract": "Visual recognition ecosystems (e.g. ImageNet, Pascal, COCO) have undeniably played a prevailing role in the evolution of modern computer vision. We argue that interactive and embodied visual AI has reached a stage of development similar to visual recognition prior to the advent of these ecosystems. Recently, various synthetic environments have been introduced to facilitate research in embodied AI. Notwithstanding this progress, the crucial question of how well models trained in simulation generalize to reality has remained largely unanswered. The creation of a comparable ecosystem for simulation-to-real embodied AI presents many challenges: (1) the inherently interactive nature of the problem, (2) the need for tight alignments between real and simulated worlds, (3) the difficulty of replicating physical conditions for repeatable experiments, (4) and the associated cost. In this paper, we introduce RoboTHOR to democratize research in interactive and embodied visual AI. RoboTHOR offers a framework of simulated environments paired with physical counterparts to systematically explore and overcome the challenges of simulation-to-real transfer, and a platform where researchers across the globe can remotely test their embodied models in the physical world. As a first benchmark, our experiments show there exists a significant gap between the performance of models trained in simulation when they are tested in both simulations and their carefully constructed physical analogs. We hope that RoboTHOR will spur the next stage of evolution in embodied computer vision."
                },
                {
                    "title": "Learning to Explore using Active Neural SLAM",
                    "abstract": "This work presents a modular and hierarchical approach to learn policies for exploring 3D environments, called `Active Neural SLAM'. Our approach leverages the strengths of both classical and learning-based methods, by using analytical path planners with learned SLAM module, and global and local policies. The use of learning provides flexibility with respect to input modalities (in the SLAM module), leverages structural regularities of the world (in global policies), and provides robustness to errors in state estimation (in local policies). Such use of learning within each module retains its benefits, while at the same time, hierarchical decomposition and modular training allow us to sidestep the high sample complexities associated with training end-to-end policies. Our experiments in visually and physically realistic simulated 3D environments demonstrate the effectiveness of our approach over past learning and geometry-based approaches. The proposed model can also be easily transferred to the PointGoal task and was the winning entry of the CVPR 2019 Habitat PointGoal Navigation Challenge."
                },
                {
                    "title": "CURL: Contrastive Unsupervised Representations for Reinforcement Learning",
                    "abstract": "We present CURL: Contrastive Unsupervised Representations for Reinforcement Learning. CURL extracts high-level features from raw pixels using contrastive learning and performs off-policy control on top of the extracted features. CURL outperforms prior pixel-based methods, both model-based and model-free, on complex tasks in the DeepMind Control Suite and Atari Games showing 1.9x and 1.2x performance gains at the 100K environment and interaction steps benchmarks respectively. On the DeepMind Control Suite, CURL is the first image-based algorithm to nearly match the sample-efficiency of methods that use state-based features. Our code is open-sourced and available at this https URL."
                },
                {
                    "title": "SAPIEN: A SimulAted Part-Based Interactive ENvironment",
                    "abstract": "Building home assistant robots has long been a goal for vision and robotics researchers. To achieve this task, a simulated environment with physically realistic simulation, sufficient articulated objects, and transferability to the real robot is indispensable. Existing environments achieve these requirements for robotics simulation with different levels of simplification and focus. We take one step further in constructing an environment that supports household tasks for training robot learning algorithm. Our work, SAPIEN, is a realistic and physics-rich simulated environment that hosts a large-scale set of articulated objects. SAPIEN enables various robotic vision and interaction tasks that require detailed part-level understanding.We evaluate state-of-the-art vision algorithms for part detection and motion attribute recognition as well as demonstrate robotic interaction tasks using heuristic approaches and reinforcement learning algorithms. We hope that SAPIEN will open research directions yet to be explored, including learning cognition through interaction, part motion discovery, and construction of robotics-ready simulated game environment."
                },
                {
                    "title": "Learning Robust Representations via Multi-View Information Bottleneck",
                    "abstract": "The information bottleneck method provides an information-theoretic view of representation learning. The original formulation, however, can only be applied in the supervised setting where task-specific labels are available at learning time. We extend this method to the unsupervised setting, by taking advantage of multi-view data, which provides two views of the same underlying entity. A theoretical analysis leads to the definition of a new multi-view model which produces state-of-the-art results on two standard multi-view datasets, Sketchy and MIR-Flickr. We also extend our theory to the single-view setting by taking advantage of standard data augmentation techniques, empirically showing better generalization capabilities when compared to traditional unsupervised approaches."
                },
                {
                    "title": "Learning Task-Driven Control Policies via Information Bottlenecks",
                    "abstract": "This paper presents a reinforcement learning approach to synthesizing task-driven control policies for robotic systems equipped with rich sensory modalities (e.g., vision or depth). Standard reinforcement learning algorithms typically produce policies that tightly couple control actions to the entirety of the system's state and rich sensor observations. As a consequence, the resulting policies can often be sensitive to changes in task-irrelevant portions of the state or observations (e.g., changing background colors). In contrast, the approach we present here learns to create a task-driven representation that is used to compute control actions. Formally, this is achieved by deriving a policy gradient-style algorithm that creates an information bottleneck between the states and the task-driven representation; this constrains actions to only depend on task-relevant information. We demonstrate our approach in a thorough set of simulation results on multiple examples including a grasping task that utilizes depth images and a ball-catching task that utilizes RGB images. Comparisons with a standard policy gradient approach demonstrate that the task-driven policies produced by our algorithm are often significantly more robust to sensor noise and task-irrelevant changes in the environment."
                },
                {
                    "title": "ALFRED: A Benchmark for Interpreting Grounded Instructions for Everyday Tasks",
                    "abstract": "We present ALFRED (Action Learning From Realistic Environments and Directives), a benchmark for learning a mapping from natural language instructions and egocentric vision to sequences of actions for household tasks. ALFRED includes long, compositional tasks with non-reversible state changes to shrink the gap between research benchmarks and real-world applications. ALFRED consists of expert demonstrations in interactive visual environments for 25k natural language directives. These directives contain both high-level goals like \u201cRinse off a mug and place it in the coffee maker.\u201d and low-level language instructions like \u201cWalk to the coffee maker on the right.\u201d ALFRED tasks are more complex in terms of sequence length, action space, and language than existing vision- and-language task datasets. We show that a baseline model based on recent embodied vision-and-language tasks performs poorly on ALFRED, suggesting that there is significant room for developing innovative grounded visual language understanding models with this benchmark."
                },
                {
                    "title": "DD-PPO: Learning Near-Perfect PointGoal Navigators from 2.5 Billion Frames",
                    "abstract": "We present Decentralized Distributed Proximal Policy Optimization (DD-PPO), a method for distributed reinforcement learning in resource-intensive simulated environments. DD-PPO is distributed (uses multiple machines), decentralized (lacks a centralized server), and synchronous (no computation is ever \"stale\"), making it conceptually simple and easy to implement. In our experiments on training virtual robots to navigate in Habitat-Sim, DD-PPO exhibits near-linear scaling -- achieving a speedup of 107x on 128 GPUs over a serial implementation. We leverage this scaling to train an agent for 2.5 Billion steps of experience (the equivalent of 80 years of human experience) -- over 6 months of GPU-time training in under 3 days of wall-clock time with 64 GPUs. \n\nThis massive-scale training not only sets the state of art on Habitat Autonomous Navigation Challenge 2019, but essentially \"solves\" the task -- near-perfect autonomous navigation in an unseen environment without access to a map, directly from an RGB-D camera and a GPS+Compass sensor. Fortuitously, error vs computation exhibits a power-law-like distribution; thus, 90% of peak performance is obtained relatively early (at 100 million steps) and relatively cheaply (under 1 day with 8 GPUs). Finally, we show that the scene understanding and navigation policies learned can be transferred to other navigation tasks -- the analog of \"ImageNet pre-training + task-specific fine-tuning\" for embodied AI. Our model outperforms ImageNet pre-trained CNNs on these transfer tasks and can serve as a universal resource (all models and code are publicly available)."
                },
                {
                    "title": "Generalization in Reinforcement Learning with Selective Noise Injection and Information Bottleneck",
                    "abstract": "The ability for policies to generalize to new environments is key to the broad application of RL agents. A promising approach to prevent an agent\u2019s policy from overfitting to a limited set of training environments is to apply regularization techniques originally developed for supervised learning. However, there are stark differences between supervised learning and RL. We discuss those differences and propose modifications to existing regularization techniques in order to better adapt them to RL. In particular, we focus on regularization techniques relying on the injection of noise into the learned function, a family that includes some of the most widely used approaches such as Dropout and Batch Normalization. To adapt them to RL, we propose Selective Noise Injection (SNI), which maintains the regularizing effect the injected noise has, while mitigating the adverse effects it has on the gradient quality. Furthermore, we demonstrate that the Information Bottleneck (IB) is a particularly well suited regularization technique for RL as it is effective in the low-data regime encountered early on in training RL agents. Combining the IB with SNI, we significantly outperform current state of the art results, including on the recently proposed generalization benchmark Coinrun."
                },
                {
                    "title": "Bayesian Relational Memory for Semantic Visual Navigation",
                    "abstract": "We introduce a new memory architecture, Bayesian Relational Memory (BRM), to improve the generalization ability for semantic visual navigation agents in unseen environments, where an agent is given a semantic target to navigate towards. BRM takes the form of a probabilistic relation graph over semantic entities (e.g., room types), which allows (1) capturing the layout prior from training environments, i.e., prior knowledge, (2) estimating posterior layout at test time, i.e., memory update, and (3) efficient planning for navigation, altogether. We develop a BRM agent consisting of a BRM module for producing sub-goals and a goal-conditioned locomotion module for control. When testing in unseen environments, the BRM agent outperforms baselines that do not explicitly utilize the probabilistic relational memory structure."
                },
                {
                    "title": "Shaping Belief States with Generative Environment Models for RL",
                    "abstract": "When agents interact with a complex environment, they must form and maintain beliefs about the relevant aspects of that environment. We propose a way to efficiently train expressive generative models in complex environments. We show that a predictive algorithm with an expressive generative model can form stable belief-states in visually rich and dynamic 3D environments. More precisely, we show that the learned representation captures the layout of the environment as well as the position and orientation of the agent. Our experiments show that the model substantially improves data-efficiency on a number of reinforcement learning (RL) tasks compared with strong model-free baseline agents. We find that predicting multiple steps into the future (overshooting), in combination with an expressive generative model, is critical for stable representations to emerge. In practice, using expressive generative models in RL is computationally expensive and we propose a scheme to reduce this computational burden, allowing us to build agents that are competitive with model-free baselines."
                },
                {
                    "title": "SplitNet: Sim2Sim and Task2Task Transfer for Embodied Visual Navigation",
                    "abstract": "We propose SplitNet, a method for decoupling visual perception and policy learning. By incorporating auxiliary tasks and selective learning of portions of the model, we explicitly decompose the learning objectives for visual navigation into perceiving the world and acting on that perception. We show improvements over baseline models on transferring between simulators, an encouraging step towards Sim2Real. Additionally, SplitNet generalizes better to unseen environments from the same simulator and transfers faster and more effectively to novel embodied navigation tasks. Further, given only a small sample from a target domain, SplitNet can match the performance of traditional end-to-end pipelines which receive the entire dataset"
                },
                {
                    "title": "Habitat: A Platform for Embodied AI Research",
                    "abstract": "We present Habitat, a platform for research in embodied artificial intelligence (AI). Habitat enables training embodied agents (virtual robots) in highly efficient photorealistic 3D simulation. Specifically, Habitat consists of: (i) Habitat-Sim: a flexible, high-performance 3D simulator with configurable agents, sensors, and generic 3D dataset handling. Habitat-Sim is fast -- when rendering a scene from Matterport3D, it achieves several thousand frames per second (fps) running single-threaded, and can reach over 10,000 fps multi-process on a single GPU. (ii) Habitat-API: a modular high-level library for end-to-end development of embodied AI algorithms -- defining tasks (e.g., navigation, instruction following, question answering), configuring, training, and benchmarking embodied agents. These large-scale engineering contributions enable us to answer scientific questions requiring experiments that were till now impracticable or 'merely' impractical. Specifically, in the context of point-goal navigation: (1) we revisit the comparison between learning and SLAM approaches from two recent works and find evidence for the opposite conclusion -- that learning outperforms SLAM if scaled to an order of magnitude more experience than previous investigations, and (2) we conduct the first cross-dataset generalization experiments {train, test} x {Matterport3D, Gibson} for multiple sensors {blind, RGB, RGBD, D} and find that only agents with depth (D) sensors generalize across datasets. We hope that our open-source platform and these findings will advance research in embodied AI."
                },
                {
                    "title": "InfoBot: Transfer and Exploration via the Information Bottleneck",
                    "abstract": "A central challenge in reinforcement learning is discovering effective policies for tasks where rewards are sparsely distributed. We postulate that in the absence of useful reward signals, an effective exploration strategy should seek out {\\it decision states}. These states lie at critical junctions in the state space from where the agent can transition to new, potentially unexplored regions. We propose to learn about decision states from prior experience. By training a goal-conditioned policy with an information bottleneck, we can identify decision states by examining where the model actually leverages the goal state. We find that this simple mechanism effectively identifies decision states, even in partially observed settings. In effect, the model learns the sensory cues that correlate with potential subgoals. In new environments, this model can then identify novel subgoals for further exploration, guiding the agent through a sequence of potential decision states and through new regions of the state space."
                },
                {
                    "title": "Neural Predictive Belief Representations",
                    "abstract": "Unsupervised representation learning has succeeded with excellent results in many applications. It is an especially powerful tool to learn a good representation of environments with partial or noisy observations. In partially observable domains it is important for the representation to encode a belief state, a sufficient statistic of the observations seen so far. In this paper, we investigate whether it is possible to learn such a belief representation using modern neural architectures. Specifically, we focus on one-step frame prediction and two variants of contrastive predictive coding (CPC) as the objective functions to learn the representations. To evaluate these learned representations, we test how well they can predict various pieces of information about the underlying state of the environment, e.g., position of the agent in a 3D maze. We show that all three methods are able to learn belief representations of the environment, they encode not only the state information, but also its uncertainty, a crucial aspect of belief states. We also find that for CPC multi-step predictions and action-conditioning are critical for accurate belief representations in visually complex environments. The ability of neural representations to capture the belief information has the potential to spur new advances for learning and planning in partially observable domains, where leveraging uncertainty is essential for optimal decision making."
                },
                {
                    "title": "The Bottleneck Simulator: A Model-based Deep Reinforcement Learning Approach",
                    "abstract": "Deep reinforcement learning has recently shown many impressive successes. However, one major obstacle towards applying such methods to real-world problems is their lack of data-efficiency. To this end, we propose the Bottleneck Simulator: a model-based reinforcement learning method which combines a learned, factorized transition model of the environment with rollout simulations to learn an effective policy from few examples. The learned transition model employs an abstract, discrete (bottleneck) state, which increases sample efficiency by reducing the number of model parameters and by exploiting structural properties of the environment. We provide a mathematical analysis of the Bottleneck Simulator in terms of fixed points of the learned policy, which reveals how performance is affected by four distinct sources of error: an error related to the abstract space structure, an error related to the transition model estimation variance, an error related to the transition model estimation bias, and an error related to the transition model class bias. Finally, we evaluate the Bottleneck Simulator on two natural language processing tasks: a text adventure game and a real-world, complex dialogue response selection task. On both tasks, the Bottleneck Simulator yields excellent performance beating competing approaches."
                },
                {
                    "title": "Semi-parametric Topological Memory for Navigation",
                    "abstract": "We introduce a new memory architecture for navigation in previously unseen environments, inspired by landmark-based navigation in animals. The proposed semi-parametric topological memory (SPTM) consists of a (non-parametric) graph with nodes corresponding to locations in the environment and a (parametric) deep network capable of retrieving nodes from the graph based on observations. The graph stores no metric information, only connectivity of locations corresponding to the nodes. We use SPTM as a planning module in a navigation system. Given only 5 minutes of footage of a previously unseen maze, an SPTM-based navigation agent can build a topological map of the environment and use it to confidently navigate towards goals. The average success rate of the SPTM agent in goal-directed navigation across test environments is higher than the best-performing baseline by a factor of three. A video of the agent is available at this https URL"
                },
                {
                    "title": "AI2-THOR: An Interactive 3D Environment for Visual AI",
                    "abstract": "We introduce The House Of inteRactions (THOR), a framework for visual AI research, available at this http URL AI2-THOR consists of near photo-realistic 3D indoor scenes, where AI agents can navigate in the scenes and interact with objects to perform tasks. AI2-THOR enables research in many different domains including but not limited to deep reinforcement learning, imitation learning, learning by interaction, planning, visual question answering, unsupervised representation learning, object detection and segmentation, and learning models of cognition. The goal of AI2-THOR is to facilitate building visually intelligent models and push the research forward in this domain."
                },
                {
                    "title": "Vision-and-Language Navigation: Interpreting Visually-Grounded Navigation Instructions in Real Environments",
                    "abstract": "A robot that can carry out a natural-language instruction has been a dream since before the Jetsons cartoon series imagined a life of leisure mediated by a fleet of attentive robot helpers. It is a dream that remains stubbornly distant. However, recent advances in vision and language methods have made incredible progress in closely related areas. This is significant because a robot interpreting a natural-language navigation instruction on the basis of what it sees is carrying out a vision and language process that is similar to Visual Question Answering. Both tasks can be interpreted as visually grounded sequence-to-sequence translation problems, and many of the same methods are applicable. To enable and encourage the application of vision and language methods to the problem of interpreting visually-grounded navigation instructions, we present the Matter-port3D Simulator - a large-scale reinforcement learning environment based on real imagery [11]. Using this simulator, which can in future support a range of embodied vision and language tasks, we provide the first benchmark dataset for visually-grounded natural language navigation in real buildings - the Room-to-Room (R2R) dataset1."
                },
                {
                    "title": "Neural Discrete Representation Learning",
                    "abstract": "Learning useful representations without supervision remains a key challenge in machine learning. In this paper, we propose a simple yet powerful generative model that learns such discrete representations. Our model, the Vector Quantised-Variational AutoEncoder (VQ-VAE), differs from VAEs in two key ways: the encoder network outputs discrete, rather than continuous, codes; and the prior is learnt rather than static. In order to learn a discrete latent representation, we incorporate ideas from vector quantisation (VQ). Using the VQ method allows the model to circumvent issues of \"posterior collapse\" -- where the latents are ignored when they are paired with a powerful autoregressive decoder -- typically observed in the VAE framework. Pairing these representations with an autoregressive prior, the model can generate high quality images, videos, and speech as well as doing high quality speaker conversion and unsupervised learning of phonemes, providing further evidence of the utility of the learnt representations."
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "Curiosity-Driven Exploration by Self-Supervised Prediction",
                    "abstract": "In many real-world scenarios, rewards extrinsic to the agent are extremely sparse, or absent altogether. In such cases, curiosity can serve as an intrinsic reward signal to enable the agent to explore its environment and learn skills that might be useful later in its life. We formulate curiosity as the error in an agent's ability to predict the consequence of its own actions in a visual feature space learned by a self-supervised inverse dynamics model. Our formulation scales to high-dimensional continuous state spaces like images, bypasses the difficulties of directly predicting pixels, and, critically, ignores the aspects of the environment that cannot affect the agent. The proposed approach is evaluated in two environments: VizDoom and Super Mario Bros. Three broad settings are investigated: 1) sparse extrinsic reward; 2) exploration with no extrinsic reward; and 3) generalization to unseen scenarios (e.g. new levels of the same game)."
                },
                {
                    "title": "Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer",
                    "abstract": "The capacity of a neural network to absorb information is limited by its number of parameters. Conditional computation, where parts of the network are active on a per-example basis, has been proposed in theory as a way of dramatically increasing model capacity without a proportional increase in computation. In practice, however, there are significant algorithmic and performance challenges. In this work, we address these challenges and finally realize the promise of conditional computation, achieving greater than 1000x improvements in model capacity with only minor losses in computational efficiency on modern GPU clusters. We introduce a Sparsely-Gated Mixture-of-Experts layer (MoE), consisting of up to thousands of feed-forward sub-networks. A trainable gating network determines a sparse combination of these experts to use for each example. We apply the MoE to the tasks of language modeling and machine translation, where model capacity is critical for absorbing the vast quantities of knowledge available in the training corpora. We present model architectures in which a MoE with up to 137 billion parameters is applied convolutionally between stacked LSTM layers. On large language modeling and machine translation benchmarks, these models achieve significantly better results than state-of-the-art at lower computational cost."
                },
                {
                    "title": "Learning to Navigate in Complex Environments",
                    "abstract": "Learning to navigate in complex environments with dynamic elements is an important milestone in developing AI agents. In this work we formulate the navigation question as a reinforcement learning problem and show that data efficiency and task performance can be dramatically improved by relying on additional auxiliary tasks leveraging multimodal sensory inputs. In particular we consider jointly learning the goal-driven reinforcement learning problem with auxiliary depth prediction and loop closure classification tasks. This approach can learn to navigate from raw sensory input in complicated 3D mazes, approaching human-level performance even under conditions where the goal location changes frequently. We provide detailed analysis of the agent behaviour, its ability to localise, and its network activity dynamics, showing that the agent implicitly learns key navigation abilities."
                },
                {
                    "title": "Reinforcement Learning with Unsupervised Auxiliary Tasks",
                    "abstract": "Deep reinforcement learning agents have achieved state-of-the-art results by directly maximising cumulative reward. However, environments contain a much wider variety of possible training signals. In this paper, we introduce an agent that also maximises many other pseudo-reward functions simultaneously by reinforcement learning. All of these tasks share a common representation that, like unsupervised learning, continues to develop in the absence of extrinsic rewards. We also introduce a novel mechanism for focusing this representation upon extrinsic rewards, so that learning can rapidly adapt to the most relevant aspects of the actual task. Our agent significantly outperforms the previous state-of-the-art on Atari, averaging 880\\% expert human performance, and a challenging suite of first-person, three-dimensional \\emph{Labyrinth} tasks leading to a mean speedup in learning of 10$\\times$ and averaging 87\\% expert human performance on Labyrinth."
                },
                {
                    "title": "Categorical Reparameterization with Gumbel-Softmax",
                    "abstract": "Categorical variables are a natural choice for representing discrete structure in the world. However, stochastic neural networks rarely use categorical latent variables due to the inability to backpropagate through samples. In this work, we present an efficient gradient estimator that replaces the non-differentiable sample from a categorical distribution with a differentiable sample from a novel Gumbel-Softmax distribution. This distribution has the essential property that it can be smoothly annealed into a categorical distribution. We show that our Gumbel-Softmax estimator outperforms state-of-the-art gradient estimators on structured output prediction and unsupervised generative modeling tasks with categorical latent variables, and enables large speedups on semi-supervised classification."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "The Arcade Learning Environment: An Evaluation Platform for General Agents",
                    "abstract": "In this article we introduce the Arcade Learning Environment (ALE): both a challenge problem and a platform and methodology for evaluating the development of general, domain-independent AI technology. ALE provides an interface to hundreds of Atari 2600 game environments, each one different, interesting, and designed to be a challenge for human players. ALE presents significant research challenges for reinforcement learning, model learning, model-based planning, imitation learning, transfer learning, and intrinsic motivation. Most importantly, it provides a rigorous testbed for evaluating and comparing approaches to these problems. We illustrate the promise of ALE by developing and benchmarking domain-independent agents designed using well-established AI techniques for both reinforcement learning and planning. In doing so, we also propose an evaluation methodology made possible by ALE, reporting empirical results on over 55 different games. All of the software, including the benchmark agents, is publicly available."
                },
                {
                    "title": "Supervised Dictionary Learning",
                    "abstract": "It is now well established that sparse signal models are well suited for restoration tasks and can be effectively learned from audio, image, and video data. Recent research has been aimed at learning discriminative sparse models instead of purely reconstructive ones. This paper proposes a new step in that direction, with a novel sparse representation for signals belonging to different classes in terms of a shared dictionary and discriminative class models. The linear version of the proposed model admits a simple probabilistic interpretation, while its most general variant admits an interpretation in terms of kernels. An optimization framework for learning all the components of the proposed model is presented, along with experimental results on standard handwritten digit and texture classification tasks."
                },
                {
                    "title": "Top-Down Versus Bottom-Up Control of Attention in the Prefrontal and Posterior Parietal Cortices",
                    "abstract": "Attention can be focused volitionally by \u201ctop-down\u201d signals derived from task demands and automatically by \u201cbottom-up\u201d signals from salient stimuli. The frontal and parietal cortices are involved, but their neural activity has not been directly compared. Therefore, we recorded from them simultaneously in monkeys. Prefrontal neurons reflected the target location first during top-down attention, whereas parietal neurons signaled it earlier during bottom-up attention. Synchrony between frontal and parietal areas was stronger in lower frequencies during top-down attention and in higher frequencies during bottom-up attention. This result indicates that top-down and bottom-up signals arise from the frontal and sensory cortex, respectively, and different modes of attention may emphasize synchrony at different frequencies."
                },
                {
                    "title": "Gorillas in Our Midst: Sustained Inattentional Blindness for Dynamic Events",
                    "abstract": "With each eye fixation, we experience a richly detailed visual world. Yet recent work on visual integration and change direction reveals that we are surprisingly unaware of the details of our environment from one view to the next: we often do not detect large changes to objects and scenes (\u2018change blindness\u2019). Furthermore, without attention, we may not even perceive objects (\u2018inattentional blindness\u2019). Taken together, these findings suggest that we perceive and remember only those objects and details that receive focused attention. In this paper, we briefly review and discuss evidence for these cognitive forms of \u2018blindness\u2019. We then present a new study that builds on classic studies of divided visual attention to examine inattentional blindness for complex objects and events in dynamic scenes. Our results suggest that the likelihood of noticing an unexpected object depends on the similarity of that object to other objects in the display and on how difficult the priming monitoring task is. Interestingly, spatial proximity of the critical unattended object to attended locations does not appear to affect detection, suggesting that observers attend to objects and events, not spatial positions. We discuss the implications of these results for visual representations and awareness of our visual environment."
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                },
                {
                    "title": "Deep Variational Information Bottleneck",
                    "abstract": "We present a variational approximation to the information bottleneck of Tishby et al. (1999). This variational approach allows us to parameterize the information bottleneck model using a neural network and leverage the reparameterization trick for efficient training. We call this method \"Deep Variational Information Bottleneck\", or Deep VIB. We show that models trained with the VIB objective outperform those that are trained with other forms of regularization, in terms of generalization performance and robustness to adversarial attack."
                },
                {
                    "title": "Dropout: a simple way to prevent neural networks from overfitting",
                    "abstract": "Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units (along with their connections) from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different \"thinned\" networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights. This significantly reduces overfitting and gives major improvements over other regularization methods. We show that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets."
                },
                {
                    "title": "Neural mechanisms of selective visual attention.",
                    "abstract": "The two basic phenomena that define the problem of visual attention can be illustrated in a simple example. Consider the arrays shown in each panel of Figure 1. In a typical experiment, before the arrays were presented, subjects would be asked to report letters appearing in one color (targets, here black letters), and to disregard letters in the other color (nontargets, here white letters). The array would then be briefly flashed, and the subjects, without any opportunity for eye movements, would give their report. The display mimics our. usual cluttered visual environment: It contains one or more objects that are relevant to current behavior, along with others that are irrelevant. The first basic phenomenon is limited capacity for processing information. At any given time, only a small amount of the information available on the retina can be processed and used in the control of behavior. Subjectively, giving attention to any one target leaves less available for others. In Figure 1, the probability of reporting the target letter N is much lower with two accompa\u00ad nying targets (Figure la) than with none (Figure Ib). The second basic phenomenon is selectivity-the ability to filter out un\u00ad wanted information. Subjectively, one is aware of attended stimuli and largely unaware of unattended ones. Correspondingly, accuracy in identifying an attended stimulus may be independent of the number of nontargets in a display (Figure la vs Ie) (see Bundesen 1990, Duncan 1980)."
                },
                {
                    "title": "An Algorithm for Vector Quantizer Design",
                    "abstract": "An efficient and intuitive algorithm is presented for the design of vector quantizers based either on a known probabilistic model or on a long training sequence of data. The basic properties of the algorithm are discussed and demonstrated by examples. Quite general distortion measures and long blocklengths are allowed, as exemplified by the design of parameter vector quantizers of ten-dimensional vectors arising in Linear Predictive Coded (LPC) speech compression with a complicated distortion measure arising in LPC analysis that does not depend only on the error vector."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a selective attention mechanism in embodied AI that filters out task-irrelevant visual information to enhance goal-directed behaviors?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of embodied AI, as it can lead to more efficient and effective AI agents capable of navigating complex environments with minimal distractions. By improving the ability of AI to focus on relevant stimuli, we can enhance their performance in tasks such as navigation, manipulation, and instruction following. This research could pave the way for practical applications in robotics, autonomous systems, and human-computer interaction, ultimately leading to smarter and more adaptable AI systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of accurately identifying and filtering out irrelevant information from a vast array of visual stimuli. Naive approaches may fail because they do not account for the dynamic nature of environments or the specific goals of the task, leading to an overload of information that can confuse the AI. Additionally, developing a robust mechanism that can generalize across different tasks and environments requires overcoming significant technical and theoretical obstacles, such as creating effective representations that balance detail and relevance.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on general-purpose visual encoders that do not differentiate between relevant and irrelevant information, leading to inefficiencies in task performance. Existing solutions often lack the ability to adaptively filter sensory input based on specific goals, which has hindered progress in this area. Our approach differs by incorporating a codebook bottleneck that emphasizes task-relevant information, drawing from cognitive psychology principles to create a more focused representation that enhances selective attention.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves equipping a visual encoder with a codebook bottleneck that selectively retains task-relevant information while discarding distractions. We will utilize a dataset of diverse environments and tasks to train the model, measuring its performance using metrics such as task completion time and accuracy in locating target objects. The expected outcomes include improved efficiency in goal-directed behaviors and a significant reduction in the influence of irrelevant stimuli, demonstrating the effectiveness of our selective attention mechanism in embodied AI."
            }
        },
        "author_data": {
            "0142df17-f5df-4d28-bda6-eb0a335db431": {
                "pk": "0142df17-f5df-4d28-bda6-eb0a335db431",
                "name": "Ainaz Eftekhar",
                "collaborators": [
                    "Alexander Sax",
                    "Roman Bachmann",
                    "J. Malik",
                    "Amir Zamir",
                    "Mohammadreza Salehi",
                    "Niousha Sadjadi",
                    "M. H. Rohban",
                    "H. Rabiee"
                ],
                "domain": [
                    "Computer Vision",
                    "Anomaly Detection",
                    "Self-Supervised Learning",
                    "3D Data"
                ],
                "publications": [
                    {
                        "title": "Omnidata: A Scalable Pipeline for Making Multi-Task Mid-Level Vision Datasets from 3D Scans",
                        "abstract": "This paper introduces a pipeline to parametrically sample and render static multi-task vision datasets from comprehensive 3D scans from the real-world. In addition to enabling interesting lines of research, we show the tooling and generated data suffice to train robust vision models. Familiar architectures trained on a generated starter dataset reached state-of-the-art performance on multiple common vision tasks and benchmarks, despite having seen no benchmark or non-pipeline data. The depth estimation network outperforms MiDaS and the surface normal estimation network is the first to achieve human-level performance for in-the-wild surface normal estimation\u2014at least according to one metric on the OASIS benchmark. The Dockerized pipeline with CLI, the (mostly python) code, PyTorch dataloaders for the generated data, the generated starter dataset, download scripts and other utilities are all available ${\\color{Magenta}through}\\;{\\color{Magenta}our}\\;{\\color{Magenta}project}\\;{\\color{Magenta}website}$."
                    },
                    {
                        "title": "Puzzle-AE: Novelty Detection in Images through Solving Puzzles",
                        "abstract": "Autoencoder, as an essential part of many anomaly detection methods, is lacking flexibility on normal data in complex datasets. U-Net is proved to be effective for this purpose but overfits on the training data if trained by just using reconstruction error similar to other AE-based frameworks. Puzzle-solving, as a pretext task of self-supervised learning (SSL) methods, has earlier proved its ability in learning semantically meaningful features. We show that training U-Nets based on this task is an effective remedy that prevents overfitting and facilitates learning beyond pixel-level features. Shortcut solutions, however, are a big challenge in SSL tasks, including jigsaw puzzles. We propose adversarial robust training as an effective automatic shortcut removal. We achieve competitive or superior results compared to the State of the Art (SOTA) anomaly detection methods on various toy and real-world datasets. Unlike many competitors, the proposed framework is stable, fast, data-efficient, and does not require unprincipled early stopping."
                    }
                ]
            },
            "87790b30-5787-44b6-862a-591f338be971": {
                "pk": "87790b30-5787-44b6-862a-591f338be971",
                "name": "Kuo-Hao Zeng",
                "collaborators": [
                    "Min Sun",
                    "Ali Farhadi",
                    "Luca Weihs",
                    "Roozbeh Mottaghi",
                    "Shih-Han Chou",
                    "Yi-Chun Chen",
                    "Hou-Ning Hu",
                    "Jianlong Fu",
                    "Juan Carlos Niebles",
                    "Yen-Chen Lin"
                ],
                "domain": [
                    "Embodied AI",
                    "Visual Navigation",
                    "Video Analysis",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Moving Forward by Moving Backward: Embedding Action Impact over Action Semantics",
                        "abstract": "A common assumption when training embodied agents is that the impact of taking an action is stable; for instance, executing the\"move ahead\"action will always move the agent forward by a fixed distance, perhaps with some small amount of actuator-induced noise. This assumption is limiting; an agent may encounter settings that dramatically alter the impact of actions: a move ahead action on a wet floor may send the agent twice as far as it expects and using the same action with a broken wheel might transform the expected translation into a rotation. Instead of relying that the impact of an action stably reflects its pre-defined semantic meaning, we propose to model the impact of actions on-the-fly using latent embeddings. By combining these latent action embeddings with a novel, transformer-based, policy head, we design an Action Adaptive Policy (AAP). We evaluate our AAP on two challenging visual navigation tasks in the AI2-THOR and Habitat environments and show that our AAP is highly performant even when faced, at inference-time with missing actions and, previously unseen, perturbed action space. Moreover, we observe significant improvement in robustness against these actions when evaluating in real-world scenarios."
                    },
                    {
                        "title": "Pushing it out of the Way: Interactive Visual Navigation",
                        "abstract": "We have observed significant progress in visual navigation for embodied agents. A common assumption in studying visual navigation is that the environments are static; this is a limiting assumption. Intelligent navigation may involve interacting with the environment beyond just moving forward/backward and turning left/right. Sometimes, the best way to navigate is to push something out of the way. In this paper, we study the problem of interactive navigation where agents learn to change the environment to navigate more efficiently to their goals. To this end, we introduce the Neural Interaction Engine (NIE) to explicitly predict the change in the environment caused by the agent\u2019s actions. By modeling the changes while planning, we find that agents exhibit significant improvements in their navigational capabilities. More specifically, we consider two downstream tasks in the physics-enabled, visually rich, AI2-THOR environment: (1) reaching a target while the path to the target is blocked (2) moving an object to a target location by pushing it. For both tasks, agents equipped with an NIE significantly outperform agents without the understanding of the effect of the actions indicating the benefits of our approach. The code and dataset are available at github.com/KuoHaoZeng/Interactive_Visual_Navigation."
                    },
                    {
                        "title": "AllenAct: A Framework for Embodied AI Research",
                        "abstract": "The domain of Embodied AI, in which agents learn to complete tasks through interaction with their environment from egocentric observations, has experienced substantial growth with the advent of deep reinforcement learning and increased interest from the computer vision, NLP, and robotics communities. This growth has been facilitated by the creation of a large number of simulated environments (such as AI2-THOR, Habitat and CARLA), tasks (like point navigation, instruction following, and embodied question answering), and associated leaderboards. While this diversity has been beneficial and organic, it has also fragmented the community: a huge amount of effort is required to do something as simple as taking a model trained in one environment and testing it in another. This discourages good science. We introduce AllenAct, a modular and flexible learning framework designed with a focus on the unique requirements of Embodied AI research. AllenAct provides first-class support for a growing collection of embodied environments, tasks and algorithms, provides reproductions of state-of-the-art models and includes extensive documentation, tutorials, start-up code, and pre-trained models. We hope that our framework makes Embodied AI more accessible and encourages new researchers to join this exciting area. The framework can be accessed at: this https URL"
                    },
                    {
                        "title": "Style Example-Guided Text Generation using Generative Adversarial Transformers",
                        "abstract": "We introduce a language generative model framework for generating a styled paragraph based on a context sentence and a style reference example. The framework consists of a style encoder and a texts decoder. The style encoder extracts a style code from the reference example, and the text decoder generates texts based on the style code and the context. We propose a novel objective function to train our framework. We also investigate different network design choices. We conduct extensive experimental validation with comparison to strong baselines to validate the effectiveness of the proposed framework using a newly collected dataset with diverse text styles. Both code and dataset will be released upon publication."
                    },
                    {
                        "title": "Visual Reaction: Learning to Play Catch With Your Drone",
                        "abstract": "In this paper we address the problem of visual reaction: the task of interacting with dynamic environments where the changes in the environment are not necessarily caused by the agents itself. Visual reaction entails predicting the future changes in a visual environment and planning accordingly. We study the problem of visual reaction in the context of playing catch with a drone in visually rich synthetic environments. This is a challenging problem since the agent is required to learn (1) how objects with different physical properties and shapes move, (2) what sequence of actions should be taken according to the prediction, (3) how to adjust the actions based on the visual feedback from the dynamic environment (e.g., when objects bouncing off a wall), and (4) how to reason and act with an unexpected state change in a timely manner. We propose a new dataset for this task, which includes 30K throws of 20 types of objects in different directions with different forces. Our results show that our model that integrates a forecaster with a planner outperforms a set of strong baselines that are based on tracking as well as pure model-based and model-free RL baselines. The code and dataset are available at github.com/KuoHaoZeng/Visual_Reaction."
                    },
                    {
                        "title": "Self-View Grounding Given a Narrated 360 Degree Video",
                        "abstract": "Narrated 360\u25e6 videos are typically provided in many touring scenarios to mimic real-world experience. However, previous work has shown that smart assistance (i.e., providing visual guidance) can significantly help users to follow the Normal Field of View (NFoV) corresponding to the narrative. In this project, we aim at automatically grounding the NFoVs of a 360\u25e6 video given subtitles of the narrative (referred to as \u201cNFoV-grounding\u201d). We propose a novel Visual Grounding Model (VGM) to implicitly and efficiently predict the NFoVs given the video content and subtitles. Specifically, at each frame, we efficiently encode the panorama into feature map of candidate NFoVs using a Convolutional Neural Network (CNN) and the subtitles to the same hidden space using an RNN with Gated Recurrent Units (GRU). Then, we apply soft-attention on candidate NFoVs to trigger sentence decoder aiming to minimize the reconstruct loss between the generated and given sentence. Finally, we obtain the NFoV as the candidate NFoV with the maximum attention without any human supervision. To train VGM more robustly, we also generate a reverse sentence conditioning on one minus the soft-attention such that the attention focuses on candidate NFoVs less relevant to the given sentence. The negative log reconstruction loss of the reverse sentence (referred to as \u201cirrelevant loss\u201d) is jointly minimized to encourage the reverse sentence to be different from the given sentence. To evaluate our method, we collect the first narrated 360\u25e6 videos dataset and achieve state-of-the-art NFoV-grounding performance. Introduction Thanks to the availability of consumer-level 360\u25e6 video cameras, many 360\u25e6 videos are shared on websites like YouTube and Facebook. Among these videos, a subset of them is narrated with natural language phrases describing the video content. For instance, many videos consist of the guided tour of real estates (e.g., houses and apartments) or tourist locations. Intuitively, a 360\u25e6 touring video provides viewers the freedom to follow the narrative and select Normal Field of Views (NFoVs) provided by typical video players (see Fig. 1). However, a study in Lin et al. (2017) suggests that visual guidance (i.e., assistance through visual indicator) is preferred to assist viewers to follow the NFoV described by the narrative. In order to provide the visual Copyright c \u00a9 2018, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. t t-1 t+1 NFOV NFOV"
                    },
                    {
                        "title": "Omnidirectional CNN for Visual Place Recognition and Navigation",
                        "abstract": "Visual place recognition is challenging, especially when only a few place exemplars are given. To mitigate the challenge, we consider place recognition method using omnidirectional cameras and propose a novel Omnidirectional Convolutional Neural Network (O-CNN) to handle severe camera pose variation. Given a visual input, the task of the O-CNN is not to retrieve the matched place exemplar, but to retrieve the closest place exemplar and estimate the relative distance between the input and the closest place. With the ability to estimate relative distance, a heuristic policy is proposed to navigate a robot to the retrieved closest place. Note that the network is designed to take advantage of the omnidirectional view by incorporating circular padding and rotation invariance. To train a powerful O-CNN, we build a virtual world for training on a large scale. We also propose a continuous lifted structured feature embedding loss to learn the concept of distance efficiently. Finally, our experimental results confirm that our method achieves state-of-the-art accuracy and speed with both the virtual world and real-world datasets."
                    },
                    {
                        "title": "Semantic Highlight Retrieval and Term Prediction",
                        "abstract": "Due to the unprecedented growth of unedited videos, finding highlights relevant to a text query in a set of unedited videos has become increasingly important. We refer this task as semantic highlight retrieval and propose a query-dependent video representation for retrieving a variety of highlights. Our method consists of two parts: 1) \u201cviralets\u201d, a mid-level representation bridging between semantic [Fig. 1(a)] and visual [Fig. 1(c)] spaces and 2) a novel Semantic-MODulation (SMOD) procedure to make viralets query-dependent (referred to as SMOD viralets). Given SMOD viralets, we train a single highlight ranker to predict the highlightness of clips with respect to a variety of queries (two examples in Fig. 1), whereas existing approaches can be applied only in a few predefined domains. Other than semantic highlight retrieval, viralets can also be used to associate relevant terms to each video. We utilize this property and propose a simple term prediction method based on nearest neighbor search. To conduct experiments, we collect a viral video dataset1 including users\u2019 comments, highlights, and/or original videos. Among a testing database with 1189 clips (13% highlights and 87% non-highlights), our highlight ranker achieves 41.2% recall at top-10 retrieved clips. It is significantly higher than the state-of-the-art domain-specific highlight ranker and its extension. Similarly, our method also outperforms all baseline methods on the publicly available video highlight dataset. Finally, our simple term prediction method utilizing viralets outperforms the state-of-the-art matrix factorization method (adapted from Kalayeh et al.). 1 Viral videos refer to popular online videos. We focus on user-generated viral videos, which typically contain short highlight marked by users."
                    },
                    {
                        "title": "Agent-Centric Risk Assessment: Accident Anticipation and Risky Region Localization",
                        "abstract": "For survival, a living agent (e.g., human in Fig. 1(a)) must have the ability to assess risk (1) by temporally anticipating accidents before they occur (Fig. 1(b)), and (2) by spatially localizing risky regions (Fig. 1(c)) in the environment to move away from threats. In this paper, we take an agent-centric approach to study the accident anticipation and risky region localization tasks. We propose a novel soft-attention Recurrent Neural Network (RNN) which explicitly models both spatial and appearance-wise non-linear interaction between the agent triggering the event and another agent or static-region involved. In order to test our proposed method, we introduce the Epic Fail (EF) dataset consisting of 3000 viral videos capturing various accidents. In the experiments, we evaluate the risk assessment accuracy both in the temporal domain (accident anticipation) and spatial domain (risky region localization) on our EF dataset and the Street Accident (SA) dataset. Our method consistently outperforms other baselines on both datasets."
                    },
                    {
                        "title": "Self-view Grounding Given a Narrated 360\u00b0 Video",
                        "abstract": "    Narrated 360\u00b0 videos are typically provided in many touring scenarios to mimic real-world experience. However, previous work has shown that smart assistance (i.e., providing visual guidance) can significantly help users to follow the Normal Field of View (NFoV) corresponding to the narrative.In this project, we aim at automatically grounding the NFoVs of a 360\u00b0 video given subtitles of the narrative (referred to as ''NFoV-grounding\"). We propose a novel Visual Grounding Model (VGM) to implicitly and efficiently predict the NFoVs given the video content and subtitles. Specifically, at each frame, we efficiently encode the panorama into feature map of candidate NFoVs using a Convolutional Neural Network (CNN) and the subtitles to the same hidden space using an RNN with Gated Recurrent Units (GRU). Then, we apply soft-attention on candidate NFoVs to trigger sentence decoder aiming to minimize the reconstruct loss between the generated and given sentence. Finally, we obtain the NFoV as the candidate NFoV with the maximum attention without any human supervision.To train VGM more robustly, we also generate a reverse sentence conditioning on one minus the soft-attention such that the attention focuses on candidate NFoVs less relevant to the given sentence. The negative log reconstruction loss of the reverse sentence (referred to as ''irrelevant loss\") is jointly minimized to encourage the reverse sentence to be different from the given sentence. To evaluate our method, we collect the first narrated 360\u00b0 videos dataset and achieve state-of-the-art NFoV-grounding performance.   "
                    },
                    {
                        "title": "Visual Forecasting by Imitating Dynamics in Natural Sequences",
                        "abstract": "We introduce a general framework for visual forecasting, which directly imitates visual sequences without additional supervision. As a result, our model can be applied at several semantic levels and does not require any domain knowledge or handcrafted features. We achieve this by formulating visual forecasting as an inverse reinforcement learning (IRL) problem, and directly imitate the dynamics in natural sequences from their raw pixel values. The key challenge is the high-dimensional and continuous state-action space that prohibits the application of previous IRL algorithms. We address this computational bottleneck by extending recent progress in model-free imitation with trainable deep feature representations, which (1) bypasses the exhaustive state-action pair visits in dynamic programming by using a dual formulation and (2) avoids explicit state sampling at gradient computation using a deep feature reparametrization. This allows us to apply IRL at scale and directly imitate the dynamics in high-dimensional continuous visual sequences from the raw pixel values. We evaluate our approach at three different level-of-abstraction, from low level pixels to higher level semantics: future frame generation, action anticipation, visual story forecasting. At all levels, our approach outperforms existing methods."
                    },
                    {
                        "title": "Self-view Grounding Given a Narrated 360{\\deg} Video",
                        "abstract": "Narrated 360{\\deg} videos are typically provided in many touring scenarios to mimic real-world experience. However, previous work has shown that smart assistance (i.e., providing visual guidance) can significantly help users to follow the Normal Field of View (NFoV) corresponding to the narrative. In this project, we aim at automatically grounding the NFoVs of a 360{\\deg} video given subtitles of the narrative (referred to as \"NFoV-grounding\"). We propose a novel Visual Grounding Model (VGM) to implicitly and efficiently predict the NFoVs given the video content and subtitles. Specifically, at each frame, we efficiently encode the panorama into feature map of candidate NFoVs using a Convolutional Neural Network (CNN) and the subtitles to the same hidden space using an RNN with Gated Recurrent Units (GRU). Then, we apply soft-attention on candidate NFoVs to trigger sentence decoder aiming to minimize the reconstruct loss between the generated and given sentence. Finally, we obtain the NFoV as the candidate NFoV with the maximum attention without any human supervision. To train VGM more robustly, we also generate a reverse sentence conditioning on one minus the soft-attention such that the attention focuses on candidate NFoVs less relevant to the given sentence. The negative log reconstruction loss of the reverse sentence (referred to as \"irrelevant loss\") is jointly minimized to encourage the reverse sentence to be different from the given sentence. To evaluate our method, we collect the first narrated 360{\\deg} videos dataset and achieve state-of-the-art NFoV-grounding performance."
                    },
                    {
                        "title": "Leveraging Video Descriptions to Learn Video Question Answering",
                        "abstract": "    We propose a scalable approach to learn video-based question answering (QA): to answer a free-form natural language question about the contents of a video. Our approach automatically harvests a large number of videos and descriptions freely available online. Then, a large number of candidate QA pairs are automatically generated from descriptions rather than manually annotated. Next, we use these candidate QA pairs to train a number of video-based QA methods extended from MN (Sukhbaatar et al. 2015), VQA (Antol et al. 2015), SA (Yao et al. 2015), and SS (Venugopalan et al. 2015). In order to handle non-perfect candidate QA pairs, we propose a self-paced learning procedure to iteratively identify them and mitigate their effects in training. Finally, we evaluate performance on manually generated video-based QA pairs. The results show that our self-paced learning procedure is effective, and the extended SS model outperforms various baselines.   "
                    },
                    {
                        "title": "Semantic highlight retrieval",
                        "abstract": "Finding highlights relevant to a text query in unedited videos has become increasingly important due to their unprecedented growth. We refer this task as semantic highlight retrieval and propose a query-dependent video representation for retrieving a variety of highlights. Our method consist of two parts: (1) \u201cviralets\u201d, a mid-level representation bridging between visual and semantic spaces; (2) a novel Semantically-Modulation (SM) procedure to make viralets query-dependent (referred to as SM viralets). Given SM viralets, we train a single highlight ranker to predict the highlightness of clips with respect to a variety of queries, whereas existing approaches can be applied only in a few predefined domains. We collect a viral video dataset1 including users' comments, highlights, and/or original videos. Among a database with 1189 (13% highlights) clips, our highlight ranker achieves 41.2% recall at top-10 retrieved clips. It is significantly higher than a state-of-the-art domain-specific highlight ranker and its extension. Similarly, our method also outperforms all baseline methods on the video highlight dataset."
                    }
                ]
            },
            "813cb617-ca88-4859-a321-f625f6f89fb6": {
                "pk": "813cb617-ca88-4859-a321-f625f6f89fb6",
                "name": "Jiafei Duan",
                "collaborators": [
                    "Cheston Tan",
                    "Samson Yu",
                    "Dieter Fox",
                    "Ranjay Krishna",
                    "Wilbert Pumacay",
                    "Yi Wang",
                    "Arijit Dasgupta",
                    "Wentao Yuan",
                    "Yi Ru Wang",
                    "Jenny Zhang"
                ],
                "domain": [
                    "Robotics",
                    "Physical Reasoning",
                    "Machine Learning",
                    "Human-Robot Interaction"
                ],
                "publications": [
                    {
                        "title": "Manipulate-Anything: Automating Real-World Robots using Vision-Language Models",
                        "abstract": "Large-scale endeavors like and widespread community efforts such as Open-X-Embodiment have contributed to growing the scale of robot demonstration data. However, there is still an opportunity to improve the quality, quantity, and diversity of robot demonstration data. Although vision-language models have been shown to automatically generate demonstration data, their utility has been limited to environments with privileged state information, they require hand-designed skills, and are limited to interactions with few object instances. We propose Manipulate-Anything, a scalable automated generation method for real-world robotic manipulation. Unlike prior work, our method can operate in real-world environments without any privileged state information, hand-designed skills, and can manipulate any static object. We evaluate our method using two setups. First, Manipulate-Anything successfully generates trajectories for all 7 real-world and 14 simulation tasks, significantly outperforming existing methods like VoxPoser. Second, Manipulate-Anything's demonstrations can train more robust behavior cloning policies than training with human demonstrations, or from data generated by VoxPoser, Scaling-up, and Code-As-Policies. We believe Manipulate-Anything can be a scalable method for both generating data for robotics and solving novel tasks in a zero-shot setting. Project page: https://robot-ma.github.io/."
                    },
                    {
                        "title": "RoboPoint: A Vision-Language Model for Spatial Affordance Prediction for Robotics",
                        "abstract": "From rearranging objects on a table to putting groceries into shelves, robots must plan precise action points to perform tasks accurately and reliably. In spite of the recent adoption of vision language models (VLMs) to control robot behavior, VLMs struggle to precisely articulate robot actions using language. We introduce an automatic synthetic data generation pipeline that instruction-tunes VLMs to robotic domains and needs. Using the pipeline, we train RoboPoint, a VLM that predicts image keypoint affordances given language instructions. Compared to alternative approaches, our method requires no real-world data collection or human demonstration, making it much more scalable to diverse environments and viewpoints. In addition, RoboPoint is a general model that enables several downstream applications such as robot navigation, manipulation, and augmented reality (AR) assistance. Our experiments demonstrate that RoboPoint outperforms state-of-the-art VLMs (GPT-4o) and visual prompting techniques (PIVOT) by 21.8% in the accuracy of predicting spatial affordance and by 30.5% in the success rate of downstream tasks. Project website: https://robo-point.github.io."
                    },
                    {
                        "title": "Octopi: Object Property Reasoning with Large Tactile-Language Models",
                        "abstract": "Physical reasoning is important for effective robot manipulation. Recent work has investigated both vision and language modalities for physical reasoning; vision can reveal information about objects in the environment and language serves as an abstraction and communication medium for additional context. Although these works have demonstrated success on a variety of physical reasoning tasks, they are limited to physical properties that can be inferred from visual or language inputs. In this work, we investigate combining tactile perception with language, which enables embodied systems to obtain physical properties through interaction and apply commonsense reasoning. We contribute a new dataset PhysiCLeAR, which comprises both physical/property reasoning tasks and annotated tactile videos obtained using a GelSight tactile sensor. We then introduce Octopi, a system that leverages both tactile representation learning and large vision-language models to predict and reason about tactile inputs with minimal language fine-tuning. Our evaluations on PhysiCLeAR show that Octopi is able to effectively use intermediate physical property predictions to improve its performance on various tactile-related tasks. PhysiCLeAR and Octopi are available at https://github.com/clear-nus/octopi."
                    },
                    {
                        "title": "THE COLOSSEUM: A Benchmark for Evaluating Generalization for Robotic Manipulation",
                        "abstract": "To realize effective large-scale, real-world robotic applications, we must evaluate how well our robot policies adapt to changes in environmental conditions. Unfortunately, a majority of studies evaluate robot performance in environments closely resembling or even identical to the training setup. We present THE COLOSSEUM, a novel simulation benchmark, with 20 diverse manipulation tasks, that enables systematical evaluation of models across 14 axes of environmental perturbations. These perturbations include changes in color, texture, and size of objects, table-tops, and backgrounds; we also vary lighting, distractors, physical properties perturbations and camera pose. Using THE COLOSSEUM, we compare 5 state-of-the-art manipulation models to reveal that their success rate degrades between 30-50% across these perturbation factors. When multiple perturbations are applied in unison, the success rate degrades $\\geq$75%. We identify that changing the number of distractor objects, target object color, or lighting conditions are the perturbations that reduce model performance the most. To verify the ecological validity of our results, we show that our results in simulation are correlated ($\\bar{R}^2 = 0.614$) to similar perturbations in real-world experiments. We open source code for others to use THE COLOSSEUM, and also release code to 3D print the objects used to replicate the real-world perturbations. Ultimately, we hope that THE COLOSSEUM will serve as a benchmark to identify modeling decisions that systematically improve generalization for manipulation. See https://robot-colosseum.github.io/ for more details."
                    },
                    {
                        "title": "Robustness of Utilizing Feedback in Embodied Visual Navigation",
                        "abstract": "This paper presents a framework for training an agent to actively request help in object-goal navigation tasks, with feedback indicating the location of the target object in its field of view. To make the agent more robust in scenarios where a teacher may not always be available, the proposed training curriculum includes a mix of episodes with and without feedback. The results show that this approach improves the agent's performance, even in the absence of feedback."
                    },
                    {
                        "title": "AR2-D2: Training a Robot Without a Robot",
                        "abstract": "Diligently gathered human demonstrations serve as the unsung heroes empowering the progression of robot learning. Today, demonstrations are collected by training people to use specialized controllers, which (tele-)operate robots to manipulate a small number of objects. By contrast, we introduce AR2-D2: a system for collecting demonstrations which (1) does not require people with specialized training, (2) does not require any real robots during data collection, and therefore, (3) enables manipulation of diverse objects with a real robot. AR2-D2 is a framework in the form of an iOS app that people can use to record a video of themselves manipulating any object while simultaneously capturing essential data modalities for training a real robot. We show that data collected via our system enables the training of behavior cloning agents in manipulating real objects. Our experiments further show that training with our AR data is as effective as training with real-world robot demonstrations. Moreover, our user study indicates that users find AR2-D2 intuitive to use and require no training in contrast to four other frequently employed methods for collecting robot demonstrations."
                    },
                    {
                        "title": "Read My Mind: A Multi-Modal Dataset for Human Belief Prediction",
                        "abstract": "Understanding human intentions is key to enabling effective and efficient human-robot interaction (HRI) in collaborative settings. To enable developments and evaluation of the ability of artificial intelligence (AI) systems to infer human beliefs, we introduce a large-scale multi-modal video dataset for intent prediction based on object-context relations."
                    },
                    {
                        "title": "NEWTON: Are Large Language Models Capable of Physical Reasoning?",
                        "abstract": "Large Language Models (LLMs), through their contextualized representations, have been empirically proven to encapsulate syntactic, semantic, word sense, and common-sense knowledge. However, there has been limited exploration of their physical reasoning abilities, specifically concerning the crucial attributes for comprehending everyday objects. To address this gap, we introduce NEWTON, a repository and benchmark for evaluating the physics reasoning skills of LLMs. Further, to enable domain-specific adaptation of this benchmark, we present a pipeline to enable researchers to generate a variant of this benchmark that has been customized to the objects and attributes relevant for their application. The NEWTON repository comprises a collection of 2800 object-attribute pairs, providing the foundation for generating infinite-scale assessment templates. The NEWTON benchmark consists of 160K QA questions, curated using the NEWTON repository to investigate the physical reasoning capabilities of several mainstream language models across foundational, explicit, and implicit reasoning tasks. Through extensive empirical analysis, our results highlight the capabilities of LLMs for physical reasoning. We find that LLMs like GPT-4 demonstrate strong reasoning capabilities in scenario-based tasks but exhibit less consistency in object-attribute reasoning compared to humans (50% vs. 84%). Furthermore, the NEWTON platform demonstrates its potential for evaluating and enhancing language models, paving the way for their integration into physically grounded settings, such as robotic manipulation. Project site: https://newtonreasoning.github.io"
                    },
                    {
                        "title": "A Study of Comfortability between Interactive AI and Human",
                        "abstract": "As the use of interactive AI systems becomes increasingly prevalent in our daily lives, it is crucial to understand how individuals feel when interacting with such systems. In this work, we investigate the comfort level of individuals when interacting with intent-predicting AI systems and identify the factors of influence. We introduce a study protocol to analyze human comfortability when interacting with intent-predicting AI systems and execute the study with over a dozen participants. The study findings suggest that users are comfortable with AI systems if they have control and their privacy is not affected. Additionally, the study found that users could differentiate between AI and human responses, but this did not significantly affect their comfort levels. This research paper's significance lies in its contribution to the growing body of literature on interactive AI systems, and it emphasizes the need to consider user perceptions in the development and deployment."
                    },
                    {
                        "title": "Good Time to Ask: A Learning Framework for Asking for Help in Embodied Visual Navigation",
                        "abstract": "In reality, it is often more efficient to ask for help than to search the entire space to find an object with an unknown location. We present a learning framework that enables an agent to actively ask for help in such embodied visual navigation tasks, where the feedback informs the agent of where the goal is in its view. To emulate the real-world scenario that a teacher may not always be present, we propose a training curriculum where feedback is not always available. We formulate an uncertainty measure of where the goal is and use empirical results to show that through this approach, the agent learns to ask for help effectively while remaining robust when feedback is not available."
                    },
                    {
                        "title": "A Survey on Machine Learning Approaches for Modelling Intuitive Physics",
                        "abstract": "Research in cognitive science has provided extensive evidence of human cognitive ability in performing physical reasoning of objects from noisy perceptual inputs. Such a cognitive ability is commonly known as intuitive physics. With advancements in deep learning, there is an increasing interest in building intelligent systems that are capable of performing physical reasoning from a given scene for the purpose of building better AI systems. As a result, many contemporary approaches in modelling intuitive physics for machine cognition have been inspired by literature from cognitive science. Despite the wide range of work in physical reasoning for machine cognition, there is a scarcity of reviews that organize and group these deep learning approaches. Especially at the intersection of intuitive physics and artificial intelligence, there is a need to make sense of the diverse range of ideas and approaches. Therefore, this paper presents a comprehensive survey of recent advances and techniques in intuitive physics-inspired deep learning approaches for physical reasoning. The survey will first categorize existing deep learning approaches into three facets of physical reasoning before organizing them into three general technical approaches and propose six categorical tasks of the field. Finally, we highlight the challenges of the current field and present some future research directions."
                    },
                    {
                        "title": "ABCDE: An Agent-Based Cognitive Development Environment",
                        "abstract": "Children's cognitive abilities are sometimes cited as AI benchmarks. How can the most common 1,000 concepts (89\\% of everyday use) be learnt in a naturalistic children's setting? Cognitive development in children is about quality, and new concepts can be conveyed via simple examples. Our approach of knowledge scaffolding uses simple objects and actions to convey concepts, like how children are taught. We introduce ABCDE, an interactive 3D environment modeled after a typical playroom for children. It comes with 300+ unique 3D object assets (mostly toys), and a large action space for child and parent agents to interact with objects and each other. ABCDE is the first environment aimed at mimicking a naturalistic setting for cognitive development in children; no other environment focuses on high-level concept learning through learner-teacher interactions. The simulator can be found at https://pypi.org/project/ABCDESim/1.0.0/"
                    },
                    {
                        "title": "BOSS: A Benchmark for Human Belief Prediction in Object-context Scenarios",
                        "abstract": "Humans with an average level of social cognition can infer the beliefs of others based solely on the nonverbal communication signals (e.g. gaze, gesture, pose and contextual information) exhibited during social interactions [1, 2, 3]. This social cognitive ability to predict human beliefs and intentions is more important than ever for ensuring safe human-robot interaction and collaboration. This paper uses the combined knowledge of Theory of Mind (ToM) and Object-Context Relations to investigate methods for enhancing collaboration between humans and autonomous systems in environments where verbal communication is prohibited. We propose a novel and challenging multimodal video dataset for assessing the capability of arti\ufb01cial intelligence (AI) systems in predicting human belief states in an object-context scenario. The proposed dataset consists of precise labelling of human belief state ground-truth and multimodal inputs replicating all nonverbal communication inputs captured by human perception. We further evaluate our dataset with existing deep learning models and provide new insights into the effects of the various input modalities and object-context relations on the performance of the baseline models."
                    },
                    {
                        "title": "PIP: Physical Interaction Prediction via Mental Imagery with Span Selection",
                        "abstract": "To align advanced arti\ufb01cial intelligence (AI) with human values and promote safe AI, it is important for AI to predict the outcome of physical interactions. Even with the ongoing debates on how humans predict the outcomes of physical interactions among objects in the real world, there are works attempting to tackle this task via cognitive-inspired AI approaches. However, there is still a lack of AI approaches that mimic the mental imagery humans use to predict physical interactions in the real world. In this work, we propose a novel PIP scheme: Physical Interaction Prediction via Mental Imagery with Span Selection. PIP utilizes a deep generative model to output future frames of physical interactions among objects before extracting crucial information for predicting physical interactions by focusing on salient frames using span selection. To evaluate our model, we propose a large-scale SPACE+ dataset of synthetic video frames, including three physical interaction events in a 3D environment. Our experiments show that PIP outperforms baselines and human performance in physical interaction prediction for both seen and unseen objects. Furthermore, PIP\u2019s span selection scheme can effectively identify the frames where physical interactions among objects occur within the generated frames, allowing for added interpretability."
                    },
                    {
                        "title": "AVoE: A Synthetic 3D Dataset on Understanding Violation of Expectation for Artificial Cognition",
                        "abstract": "Recent work in cognitive reasoning and computer vision has engendered an increasing popularity for the Violation-of-Expectation (VoE) paradigm in synthetic datasets. Inspired by work in infant psychology, researchers have started evaluating a model's ability to discriminate between expected and surprising scenes as a sign of its reasoning ability. Existing VoE-based 3D datasets in physical reasoning only provide vision data. However, current cognitive models of physical reasoning by psychologists reveal infants create high-level abstract representations of objects and interactions. Capitalizing on this knowledge, we propose AVoE: a synthetic 3D VoE-based dataset that presents stimuli from multiple novel sub-categories for five event categories of physical reasoning. Compared to existing work, AVoE is armed with ground-truth labels of abstract features and rules augmented to vision data, paving the way for high-level symbolic predictions in physical reasoning tasks."
                    },
                    {
                        "title": "SPACE: A Simulator for Physical Interactions and Causal Learning in 3D Environments",
                        "abstract": "Recent advancements in deep learning, computer vision, and embodied AI have given rise to synthetic causal reasoning video datasets. These datasets facilitate the development of AI algorithms that can reason about physical interactions between objects. However, datasets thus far have primarily focused on elementary physical events such as rolling or falling. There is currently a scarcity of datasets that focus on the physical interactions that humans perform daily with objects in the real world. To address this scarcity, we introduce SPACE: A Simulator for Physical Interactions and Causal Learning in 3D Environments. The SPACE simulator allows us to generate the SPACE dataset, a synthetic video dataset in a 3D environment, to systematically evaluate physics-based models on a range of physical causal reasoning tasks. Inspired by daily object interactions, the SPACE dataset comprises videos depicting three types of physical events: containment, stability and contact. These events make up the vast majority of the basic physical interactions between objects. We then further evaluate it with a state-of-the-art physics-based deep model and show that the SPACE dataset improves the learning of intuitive physics with an approach inspired by curriculum learning. Repository: https://github.com/jiafei1224/SPACE"
                    },
                    {
                        "title": "A Survey of Embodied AI: From Simulators to Research Tasks",
                        "abstract": "There has been an emerging paradigm shift from the era of \u201cinternet AI\u201d to \u201cembodied AI,\u201d where AI algorithms and agents no longer learn from datasets of images, videos or text curated primarily from the internet. Instead, they learn through interactions with their environments from an egocentric perception similar to humans. Consequently, there has been substantial growth in the demand for embodied AI simulators to support various embodied AI research tasks. This growing interest in embodied AI is beneficial to the greater pursuit of Artificial General Intelligence (AGI), but there has not been a contemporary and comprehensive survey of this field. This paper aims to provide an encyclopedic survey for the field of embodied AI, from its simulators to its research. By evaluating nine current embodied AI simulators with our proposed seven features, this paper aims to understand the simulators in their provision for use in embodied AI research and their limitations. Lastly, this paper surveys the three main research tasks in embodied AI \u2013 visual exploration, visual navigation and embodied question answering (QA), covering the state-of-the-art approaches, evaluation metrics and datasets. Finally, with the new insights revealed through surveying the field, the paper will provide suggestions for simulator-for-task selections and recommendations for the future directions of the field."
                    },
                    {
                        "title": "A Benchmark for Modeling Violation-of-Expectation in Physical Reasoning Across Event Categories",
                        "abstract": "Recent work in computer vision and cognitive reasoning has given rise to an increasing adoption of the Violation-of-Expectation (VoE) paradigm in synthetic datasets. Inspired by infant psychology, researchers are now evaluating a model's ability to label scenes as either expected or surprising with knowledge of only expected scenes. However, existing VoE-based 3D datasets in physical reasoning provide mainly vision data with little to no heuristics or inductive biases. Cognitive models of physical reasoning reveal infants create high-level abstract representations of objects and interactions. Capitalizing on this knowledge, we established a benchmark to study physical reasoning by curating a novel large-scale synthetic 3D VoE dataset armed with ground-truth heuristic labels of causally relevant features and rules. To validate our dataset in five event categories of physical reasoning, we benchmarked and analyzed human performance. We also proposed the Object File Physical Reasoning Network (OFPR-Net) which exploits the dataset's novel heuristics to outperform our baseline and ablation models. The OFPR-Net is also flexible in learning an alternate physical reality, showcasing its ability to learn universal causal relationships in physical reasoning to create systems with better interpretability."
                    }
                ]
            },
            "4d02e5f6-f1ad-4dee-af23-82232655e5f6": {
                "pk": "4d02e5f6-f1ad-4dee-af23-82232655e5f6",
                "name": "Ali Farhadi",
                "collaborators": [
                    "Aditya Kusupati",
                    "Mohammadreza Salehi",
                    "Ranjay Krishna",
                    "Jae Sung Park",
                    "Hanna Hajishirzi",
                    "Aniruddha Kembhavi",
                    "Niklas Muennighoff",
                    "Kyle Lo",
                    "Luca Soldaini",
                    "Rose Hendrix"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Vision-Language Models",
                    "Reinforcement Learning",
                    "Adaptive Inference"
                ],
                "publications": [
                    {
                        "title": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Multimodal Models",
                        "abstract": "Today's most advanced multimodal models remain proprietary. The strongest open-weight models rely heavily on synthetic data from proprietary VLMs to achieve good performance, effectively distilling these closed models into open ones. As a result, the community is still missing foundational knowledge about how to build performant VLMs from scratch. We present Molmo, a new family of VLMs that are state-of-the-art in their class of openness. Our key innovation is a novel, highly detailed image caption dataset collected entirely from human annotators using speech-based descriptions. To enable a wide array of user interactions, we also introduce a diverse dataset mixture for fine-tuning that includes in-the-wild Q&A and innovative 2D pointing data. The success of our approach relies on careful choices for the model architecture details, a well-tuned training pipeline, and, most critically, the quality of our newly collected datasets, all of which will be released. The best-in-class 72B model within the Molmo family not only outperforms others in the class of open weight and data models but also compares favorably against proprietary systems like GPT-4o, Claude 3.5, and Gemini 1.5 on both academic benchmarks and human evaluation. We will be releasing all of our model weights, captioning and fine-tuning data, and source code in the near future. Select model weights, inference code, and demo are available at https://molmo.allenai.org."
                    },
                    {
                        "title": "OLMoE: Open Mixture-of-Experts Language Models",
                        "abstract": "We introduce OLMoE, a fully open, state-of-the-art language model leveraging sparse Mixture-of-Experts (MoE). OLMoE-1B-7B has 7 billion (B) parameters but uses only 1B per input token. We pretrain it on 5 trillion tokens and further adapt it to create OLMoE-1B-7B-Instruct. Our models outperform all available models with similar active parameters, even surpassing larger ones like Llama2-13B-Chat and DeepSeekMoE-16B. We present various experiments on MoE training, analyze routing in our model showing high specialization, and open-source all aspects of our work: model weights, training data, code, and logs."
                    },
                    {
                        "title": "Superposed Decoding: Multiple Generations from a Single Autoregressive Inference Pass",
                        "abstract": "Many applications today provide users with multiple auto-complete drafts as they type, including GitHub's code completion, Gmail's smart compose, and Apple's messaging auto-suggestions. Under the hood, language models support this by running an autoregressive inference pass to provide a draft. Consequently, providing $k$ drafts to the user requires running an expensive language model $k$ times. To alleviate the computation cost of running $k$ inference passes, we propose Superposed Decoding, a new decoding algorithm that generates $k$ drafts at the computation cost of one autoregressive inference pass. We achieve this by feeding a superposition of the most recent token embeddings from the $k$ drafts as input to the next decoding step of the language model. At every inference step we combine the $k$ drafts with the top-$k$ tokens to get $k^2$ new drafts and cache the $k$ most likely options, using an n-gram interpolation with minimal compute overhead to filter out incoherent generations. Our experiments show that $k$ drafts from Superposed Decoding are at least as coherent and factual as Nucleus Sampling and Greedy Decoding respectively, while being at least $2.44\\times$ faster for $k\\ge3$. In a compute-normalized setting, user evaluations demonstrably favor text generated by Superposed Decoding over Nucleus Sampling. Superposed Decoding can also be combined with other decoding strategies, resulting in universal coverage gains when scaling inference time compute. Code and more examples open-sourced at https://github.com/RAIVNLab/SuperposedDecoding."
                    },
                    {
                        "title": "ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition",
                        "abstract": "Our world is full of varied actions and moves across specialized domains that we, as humans, strive to identify and understand. Within any single domain, actions can often appear quite similar, making it challenging for deep models to distinguish them accurately. To evaluate the effectiveness of multimodal foundation models in helping us recognize such actions, we present ActionAtlas v1.0, a multiple-choice video question answering benchmark featuring short videos across various sports. Each video in the dataset is paired with a question and four or five choices. The question pinpoints specific individuals, asking which choice\"best\"describes their action within a certain temporal context. Overall, the dataset includes 934 videos showcasing 580 unique actions across 56 sports, with a total of 1896 actions within choices. Unlike most existing video question answering benchmarks that only cover simplistic actions, often identifiable from a single frame, ActionAtlas focuses on intricate movements and rigorously tests the model's capability to discern subtle differences between moves that look similar within each domain. We evaluate open and proprietary foundation models on this benchmark, finding that the best model, GPT-4o, achieves a maximum accuracy of 45.52%. Meanwhile, Non-expert crowd workers, provided with action description for each choice, achieve 61.64% accuracy, where random chance is approximately 21%. Our findings with state-of-the-art models indicate that having a high frame sampling rate is important for accurately recognizing actions in ActionAtlas, a feature that some leading proprietary video models, such as Gemini, do not include in their default configuration."
                    },
                    {
                        "title": "Task Me Anything",
                        "abstract": "Benchmarks for large multimodal language models (MLMs) now serve to simultaneously assess the general capabilities of models instead of evaluating for a specific capability. As a result, when a developer wants to identify which models to use for their application, they are overwhelmed by the number of benchmarks and remain uncertain about which benchmark's results are most reflective of their specific use case. This paper introduces Task-Me-Anything, a benchmark generation engine which produces a benchmark tailored to a user's needs. Task-Me-Anything maintains an extendable taxonomy of visual assets and can programmatically generate a vast number of task instances. Additionally, it algorithmically addresses user queries regarding MLM performance efficiently within a computational budget. It contains 113K images, 10K videos, 2K 3D object assets, over 365 object categories, 655 attributes, and 335 relationships. It can generate 750M image/video question-answering pairs, which focus on evaluating MLM perceptual capabilities. Task-Me-Anything reveals critical insights: open-source MLMs excel in object and attribute recognition but lack spatial and temporal understanding; each model exhibits unique strengths and weaknesses; larger models generally perform better, though exceptions exist; and GPT4o demonstrates challenges in recognizing rotating/moving objects and distinguishing colors."
                    },
                    {
                        "title": "FLaRe: Achieving Masterful and Adaptive Robot Policies with Large-Scale Reinforcement Learning Fine-Tuning",
                        "abstract": "In recent years, the Robotics field has initiated several efforts toward building generalist robot policies through large-scale multi-task Behavior Cloning. However, direct deployments of these policies have led to unsatisfactory performance, where the policy struggles with unseen states and tasks. How can we break through the performance plateau of these models and elevate their capabilities to new heights? In this paper, we propose FLaRe, a large-scale Reinforcement Learning fine-tuning framework that integrates robust pre-trained representations, large-scale training, and gradient stabilization techniques. Our method aligns pre-trained policies towards task completion, achieving state-of-the-art (SoTA) performance both on previously demonstrated and on entirely novel tasks and embodiments. Specifically, on a set of long-horizon mobile manipulation tasks, FLaRe achieves an average success rate of 79.5% in unseen environments, with absolute improvements of +23.6% in simulation and +30.7% on real robots over prior SoTA methods. By utilizing only sparse rewards, our approach can enable generalizing to new capabilities beyond the pretraining data with minimal human effort. Moreover, we demonstrate rapid adaptation to new embodiments and behaviors with less than a day of fine-tuning. Videos can be found on the project website at https://robot-flare.github.io/"
                    },
                    {
                        "title": "CLIP meets Model Zoo Experts: Pseudo-Supervision for Visual Enhancement",
                        "abstract": "Contrastive language image pretraining (CLIP) is a standard method for training vision-language models. While CLIP is scalable, promptable, and robust to distribution shifts on image classification tasks, it lacks object localization capabilities. This paper studies the following question: Can we augment CLIP training with task-specific vision models from model zoos to improve its visual representations? Towards this end, we leverage open-source task-specific vision models to generate pseudo-labels for an uncurated and noisy image-text dataset. Subsequently, we train CLIP models on these pseudo-labels in addition to the contrastive training on image and text pairs. This simple setup shows substantial improvements of up to 16.3% across different vision tasks, including segmentation, detection, depth estimation, and surface normal estimation. Importantly, these enhancements are achieved without compromising CLIP's existing capabilities, including its proficiency in promptable zero-shot classification."
                    },
                    {
                        "title": "Localized Symbolic Knowledge Distillation for Visual Commonsense Models",
                        "abstract": "Instruction following vision-language (VL) models offer a flexible interface that supports a broad range of multimodal tasks in a zero-shot fashion. However, interfaces that operate on full images do not directly enable the user to\"point to\"and access specific regions within images. This capability is important not only to support reference-grounded VL benchmarks, but also, for practical applications that require precise within-image reasoning. We build Localized Visual Commonsense models, which allow users to specify (multiple) regions as input. We train our model by sampling localized commonsense knowledge from a large language model (LLM): specifically, we prompt an LLM to collect commonsense knowledge given a global literal image description and a local literal region description automatically generated by a set of VL models. With a separately trained critic model that selects high-quality examples, we find that training on the localized commonsense corpus can successfully distill existing VL models to support a reference-as-input interface. Empirical results and human evaluations in a zero-shot setup demonstrate that our distillation method results in more precise VL models of reasoning compared to a baseline of passing a generated referring expression to an LLM."
                    },
                    {
                        "title": "Are \"Hierarchical\" Visual Representations Hierarchical?",
                        "abstract": "Learned visual representations often capture large amounts of semantic information for accurate downstream applications. Human understanding of the world is fundamentally grounded in hierarchy. To mimic this and further improve representation capabilities, the community has explored\"hierarchical\"visual representations that aim at modeling the underlying hierarchy of the visual world. In this work, we set out to investigate if hierarchical visual representations truly capture the human perceived hierarchy better than standard learned representations. To this end, we create HierNet, a suite of 12 datasets spanning 3 kinds of hierarchy from the BREEDs subset of ImageNet. After extensive evaluation of Hyperbolic and Matryoshka Representations across training setups, we conclude that they do not capture hierarchy any better than the standard representations but can assist in other aspects like search efficiency and interpretability. Our benchmark and the datasets are open-sourced at https://github.com/ethanlshen/HierNet."
                    },
                    {
                        "title": "SHARCS: Efficient Transformers through Routing with Dynamic Width Sub-networks",
                        "abstract": "We introduce SHARCS for adaptive inference that takes into account the hardness of input samples. SHARCS can train a router on any transformer network, enabling the model to direct different samples to sub-networks with varying widths. Our experiments demonstrate that: (1) SHARCS outperforms or complements existing per-sample adaptive inference methods across various classification tasks in terms of accuracy vs. FLOPs; (2) SHARCS generalizes across different architectures and can be even applied to compressed and efficient transformer encoders to further improve their efficiency; (3) SHARCS can provide a 2 times inference speed up at an insignificant drop in accuracy."
                    },
                    {
                        "title": "MatFormer: Nested Transformer for Elastic Inference",
                        "abstract": "Transformer models are deployed in a wide range of settings, from multi-accelerator clusters to standalone mobile phones. The diverse inference constraints in these scenarios necessitate practitioners to train foundation models such as PaLM 2, Llama,&ViTs as a series of models of varying sizes. Due to significant training costs, only a select few model sizes are trained and supported, limiting more fine-grained control over relevant tradeoffs, including latency, cost, and accuracy. This work introduces MatFormer, a nested Transformer architecture designed to offer elasticity in a variety of deployment constraints. Each Feed Forward Network (FFN) block of a MatFormer model is jointly optimized with a few nested smaller FFN blocks. This training procedure allows for the Mix'n'Match of model granularities across layers -- i.e., a trained universal MatFormer model enables extraction of hundreds of accurate smaller models, which were never explicitly optimized. We empirically demonstrate MatFormer's effectiveness across different model classes (decoders&encoders), modalities (language&vision), and scales (up to 2.6B parameters). We find that a 2.6B decoder-only MatFormer language model (MatLM) allows us to extract smaller models spanning from 1.5B to 2.6B, each exhibiting comparable validation loss and one-shot downstream evaluations to their independently trained counterparts. Furthermore, we observe that smaller encoders extracted from a universal MatFormer-based ViT (MatViT) encoder preserve the metric-space structure for adaptive large-scale retrieval. Finally, we showcase that speculative decoding with the accurate and consistent submodels extracted from MatFormer can further reduce inference latency."
                    },
                    {
                        "title": "NCAM: Near-Data Processing for Nearest Neighbor Search",
                        "abstract": "Deep down in many applications like natural language processing (NLP), vision, and robotics is a form of the k-nearest neighbor search algorithm (kNN). The kNN algorithm is primarily bottlenecked by data movement, limiting throughput and incurring latency in these applications. While there do exist well bounded kNN approximations that improve the performance of kNN, these algorithms trade-off accuracy and quickly degrade into linear search for high dimensionality.  To address data movement, we designed the nearest neighbor content addressable memory (NCAM) which employs processing in-memory (PIM) to eliminate costly data transfers, and provides exact nearest neighbor search. NCAMs benefit from the modularity offered by 3D die-stacking technology which allows us to side-step the issues of direct integration with DRAM dies. The NCAM benefits from the higher density and speeds of emerging memory technologies and interfaces.  We characterize a state-of-the-art software kNN implementation and expose the shortcomings of approximate kNN search. We present a full NCAM design and estimate its performance using post-placement and route estimates, and we show that its power characteristics are compatible with emerging memory substrates. We then evaluate energy efficiency and latency compared to modern multi-core CPUs and GPGPU platforms using parameters typical of mobile and server workloads. Our simulation results show that the NCAM can achieve up to 160x throughput per watt and three orders of magnitude latency improvement over a NVIDIA Titan X GPU for server workloads, and ~2413x throughput per watt and ~94.5x latency improvements over a multi-core Intel E5-2620 CPU. Finally, we show that the NCAM is not limited to just kNN and can be generalized to act as a content addressable memory (CAM) or ternary-CAM (TCAM)."
                    }
                ]
            },
            "51ae427e-c36c-4191-96db-38d9dc2971ee": {
                "pk": "51ae427e-c36c-4191-96db-38d9dc2971ee",
                "name": "Ani Kembhavi",
                "collaborators": [
                    "Luca Weihs",
                    "Kiana Ehsani",
                    "Kuo-Hao Zeng",
                    "Ranjay Krishna",
                    "Rose Hendrix",
                    "Tanmay Gupta",
                    "Eli VanderBilt",
                    "Ali Farhadi",
                    "Jordi Salvador",
                    "Alvaro Herrasti"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Computer Vision",
                    "Reinforcement Learning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Iterated Learning Improves Compositionality in Large Vision-Language Models",
                        "abstract": "A fundamental characteristic common to both human vision and natural language is their compositional nature. Yet, despite the performance gains contributed by large vision and language pretraining, recent investigations find that most-if not all-our state-of-the-art vision-language models struggle at compositionality. They are unable to distinguish between images of \u201ca girl in white facing a man in black\u201d and \u201ca girl in blackfacing a man in white\u201d. More-over, prior work suggests that compositionality doesn't arise with scale: larger model sizes or training data don't help. This paper develops a new iterated training algorithm that incentivizes compositionality. We draw on decades of cognitive science research that identifies cultural transmission-the need to teach a new generation-as a necessary inductive prior that incentivizes humans to develop compositional languages. Specifically, we reframe vision-language contrastive learning as the Lewis Signaling Game between a vision agent and a language agent, and ope rationalize cultural transmission by iteratively resetting one of the agent's weights during training. After every iteration, this training paradigm induces representations that become \u201ceasier to learn\u201d, a property of compositional languages: e.g. our model trained on CC3M and CC12M improves standard CLIP by 4.7%, 4.0% respectfully in the SugarCrepe benchmark."
                    },
                    {
                        "title": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Multimodal Models",
                        "abstract": "Today's most advanced multimodal models remain proprietary. The strongest open-weight models rely heavily on synthetic data from proprietary VLMs to achieve good performance, effectively distilling these closed models into open ones. As a result, the community is still missing foundational knowledge about how to build performant VLMs from scratch. We present Molmo, a new family of VLMs that are state-of-the-art in their class of openness. Our key innovation is a novel, highly detailed image caption dataset collected entirely from human annotators using speech-based descriptions. To enable a wide array of user interactions, we also introduce a diverse dataset mixture for fine-tuning that includes in-the-wild Q&A and innovative 2D pointing data. The success of our approach relies on careful choices for the model architecture details, a well-tuned training pipeline, and, most critically, the quality of our newly collected datasets, all of which will be released. The best-in-class 72B model within the Molmo family not only outperforms others in the class of open weight and data models but also compares favorably against proprietary systems like GPT-4o, Claude 3.5, and Gemini 1.5 on both academic benchmarks and human evaluation. We will be releasing all of our model weights, captioning and fine-tuning data, and source code in the near future. Select model weights, inference code, and demo are available at https://molmo.allenai.org."
                    },
                    {
                        "title": "CodeNav: Beyond tool-use to using real-world codebases with LLM agents",
                        "abstract": "We present CodeNav, an LLM agent that navigates and leverages previously unseen code repositories to solve user queries. In contrast to tool-use LLM agents that require ``registration'' of all relevant tools via manual descriptions within the LLM context, CodeNav automatically indexes and searches over code blocks in the target codebase, finds relevant code snippets, imports them, and uses them to iteratively generate a solution with execution feedback. To highlight the core-capabilities of CodeNav, we first showcase three case studies where we use CodeNav for solving complex user queries using three diverse codebases. Next, on three benchmarks, we quantitatively compare the effectiveness of code-use (which only has access to the target codebase) to tool-use (which has privileged access to all tool names and descriptions). Finally, we study the effect of varying kinds of tool and library descriptions on code-use performance, as well as investigate the advantage of the agent seeing source code as opposed to natural descriptions of code. All code will be made open source under a permissive license."
                    },
                    {
                        "title": "PoliFormer: Scaling On-Policy RL with Transformers Results in Masterful Navigators",
                        "abstract": "We present PoliFormer (Policy Transformer), an RGB-only indoor navigation agent trained end-to-end with reinforcement learning at scale that generalizes to the real-world without adaptation despite being trained purely in simulation. PoliFormer uses a foundational vision transformer encoder with a causal transformer decoder enabling long-term memory and reasoning. It is trained for hundreds of millions of interactions across diverse environments, leveraging parallelized, multi-machine rollouts for efficient training with high throughput. PoliFormer is a masterful navigator, producing state-of-the-art results across two distinct embodiments, the LoCoBot and Stretch RE-1 robots, and four navigation benchmarks. It breaks through the plateaus of previous work, achieving an unprecedented 85.5% success rate in object goal navigation on the CHORES-S benchmark, a 28.5% absolute improvement. PoliFormer can also be trivially extended to a variety of downstream applications such as object tracking, multi-object navigation, and open-vocabulary navigation with no finetuning."
                    },
                    {
                        "title": "Preserving Identity with Variational Score for General-purpose 3D Editing",
                        "abstract": "We present Piva (Preserving Identity with Variational Score Distillation), a novel optimization-based method for editing images and 3D models based on diffusion models. Specifically, our approach is inspired by the recently proposed method for 2D image editing - Delta Denoising Score (DDS). We pinpoint the limitations in DDS for 2D and 3D editing, which causes detail loss and over-saturation. To address this, we propose an additional score distillation term that enforces identity preservation. This results in a more stable editing process, gradually optimizing NeRF models to match target prompts while retaining crucial input characteristics. We demonstrate the effectiveness of our approach in zero-shot image and neural field editing. Our method successfully alters visual attributes, adds both subtle and substantial structural elements, translates shapes, and achieves competitive results on standard 2D and 3D editing benchmarks. Additionally, our method imposes no constraints like masking or pre-training, making it compatible with a wide range of pre-trained diffusion models. This allows for versatile editing without needing neural field-to-mesh conversion, offering a more user-friendly experience."
                    },
                    {
                        "title": "Seeing the Unseen: Visual Common Sense for Semantic Placement",
                        "abstract": "Computer vision tasks typically involve describing what is present in an image (e.g. classification, detection, segmentation, and captioning). We study a visual common sense task that requires understanding \u2018what is not present\u2019. Specif-ically, given an image (e.g. of a living room) and a name of an object (\u201ccushion \u201d), a vision system is asked to predict semantically-meaningful regions (masks or bounding boxes) in the image where that object could be placed or is likely be placed by humans (e.g. on the sofa). We call this task: Se-mantic Placement (SP) and believe that such common-sense visual understanding is critical for assitive robots (tidying a house), AR devices (automatically rendering an object in the user's space), and visually-grounded chatbots with common sense. Studying the invisible is hard. Datasets for image description are typically constructed by curating relevant images (e.g. via image search with object names) and asking humans to annotate the contents of the image; neither of those two steps are straightforward for objects not present in the image. We overcome this challenge by operating in the opposite direction: we start with an image of an object in context, which is easy to find online, and then remove that ob-ject from the image via inpainting. This automated pipeline converts unstructured web data into a dataset comprising pairs of images with/without the object. With this proposed data generation pipeline, we collect a novel dataset, containing ~1.3M images across 9 object categories. We then train a SP prediction model, called CLIP-UNet, on our dataset. The CLIP-UNet outperforms existing VLMs and baselines that combine semantic priors with object detectors, gener-alizes well to real-world and simulated images and exhibits semantics-aware reasoning for object placement. In our user studies, we find that the SP masks predicted by CLIP-UNet are favored 43.7% and 31.3% times when comparing against the 4 SP baselines on real and simulated images. In addition, leveraging SP mask predictions from CLIP-UNet enables downstream applications like building tidying robots in indoor environments."
                    },
                    {
                        "title": "Open X-Embodiment: Robotic Learning Datasets and RT-X Models : Open X-Embodiment Collaboration0",
                        "abstract": "Large, high-capacity models trained on diverse datasets have shown remarkable successes on efficiently tackling downstream applications. In domains from NLP to Computer Vision, this has led to a consolidation of pretrained models, with general pretrained backbones serving as a starting point for many applications. Can such a consolidation happen in robotics? Conventionally, robotic learning methods train a separate model for every application, every robot, and even every environment. Can we instead train \"generalist\" X-robot policy that can be adapted efficiently to new robots, tasks, and environments? In this paper, we provide datasets in standardized data formats and models to make it possible to explore this possibility in the context of robotic manipulation, alongside experimental results that provide an example of effective X-robot policies. We assemble a dataset from 22 different robots collected through a collaboration between 21 institutions, demonstrating 527 skills (160266 tasks). We show that a high-capacity model trained on this data, which we call RT-X, exhibits positive transfer and improves the capabilities of multiple robots by leveraging experience from other platforms. The project website is robotics-transformer-x.github.io."
                    },
                    {
                        "title": "Task Me Anything",
                        "abstract": "Benchmarks for large multimodal language models (MLMs) now serve to simultaneously assess the general capabilities of models instead of evaluating for a specific capability. As a result, when a developer wants to identify which models to use for their application, they are overwhelmed by the number of benchmarks and remain uncertain about which benchmark's results are most reflective of their specific use case. This paper introduces Task-Me-Anything, a benchmark generation engine which produces a benchmark tailored to a user's needs. Task-Me-Anything maintains an extendable taxonomy of visual assets and can programmatically generate a vast number of task instances. Additionally, it algorithmically addresses user queries regarding MLM performance efficiently within a computational budget. It contains 113K images, 10K videos, 2K 3D object assets, over 365 object categories, 655 attributes, and 335 relationships. It can generate 750M image/video question-answering pairs, which focus on evaluating MLM perceptual capabilities. Task-Me-Anything reveals critical insights: open-source MLMs excel in object and attribute recognition but lack spatial and temporal understanding; each model exhibits unique strengths and weaknesses; larger models generally perform better, though exceptions exist; and GPT4o demonstrates challenges in recognizing rotating/moving objects and distinguishing colors."
                    },
                    {
                        "title": "FLaRe: Achieving Masterful and Adaptive Robot Policies with Large-Scale Reinforcement Learning Fine-Tuning",
                        "abstract": "In recent years, the Robotics field has initiated several efforts toward building generalist robot policies through large-scale multi-task Behavior Cloning. However, direct deployments of these policies have led to unsatisfactory performance, where the policy struggles with unseen states and tasks. How can we break through the performance plateau of these models and elevate their capabilities to new heights? In this paper, we propose FLaRe, a large-scale Reinforcement Learning fine-tuning framework that integrates robust pre-trained representations, large-scale training, and gradient stabilization techniques. Our method aligns pre-trained policies towards task completion, achieving state-of-the-art (SoTA) performance both on previously demonstrated and on entirely novel tasks and embodiments. Specifically, on a set of long-horizon mobile manipulation tasks, FLaRe achieves an average success rate of 79.5% in unseen environments, with absolute improvements of +23.6% in simulation and +30.7% on real robots over prior SoTA methods. By utilizing only sparse rewards, our approach can enable generalizing to new capabilities beyond the pretraining data with minimal human effort. Moreover, we demonstrate rapid adaptation to new embodiments and behaviors with less than a day of fine-tuning. Videos can be found on the project website at https://robot-flare.github.io/"
                    },
                    {
                        "title": "SPOC: Imitating Shortest Paths in Simulation Enables Effective Navigation and Manipulation in the Real World",
                        "abstract": "Reinforcement learning (RL) with dense rewards and imitation learning (IL) with human-generated trajectories are the most widely used approaches for training modern embodied agents. RL requires extensive reward shaping and auxiliary losses and is often too slow and ineffective for long-horizon tasks. While IL with human supervision is effective, collecting human trajectories at scale is extremely expensive. In this work, we show that imitating shortest-path planners in simulation produces agents that, given a language instruction, can proficiently navigate, ex-plore, and manipulate objects in both simulation and in the real world using only RGB sensors (no depth map or GPS coordinates). This surprising result is enabled by our end-to-end, transformer-based, Spocarchitecture, power-ful visual encoders paired with extensive image augmentation, and the dramatic scale and diversity of our training data: millions of frames of shortest-path-expert trajectories collected inside approximately 200,000 procedu-rally generated houses containing 40,000 unique 3D as-sets. Our models, data, training code, and newly proposed 10-task benchmarking suite Choresare available in spoc-robot.github.io."
                    },
                    {
                        "title": "Objaverse-XL: A Universe of 10M+ 3D Objects",
                        "abstract": "Natural language processing and 2D vision models have attained remarkable proficiency on many tasks primarily by escalating the scale of training data. However, 3D vision tasks have not seen the same progress, in part due to the challenges of acquiring high-quality 3D data. In this work, we present Objaverse-XL, a dataset of over 10 million 3D objects. Our dataset comprises deduplicated 3D objects from a diverse set of sources, including manually designed objects, photogrammetry scans of landmarks and everyday items, and professional scans of historic and antique artifacts. Representing the largest scale and diversity in the realm of 3D datasets, Objaverse-XL enables significant new possibilities for 3D vision. Our experiments demonstrate the improvements enabled with the scale provided by Objaverse-XL. We show that by training Zero123 on novel view synthesis, utilizing over 100 million multi-view rendered images, we achieve strong zero-shot generalization abilities. We hope that releasing Objaverse-XL will enable further innovations in the field of 3D vision at scale."
                    },
                    {
                        "title": "Zooming Out on Zooming In: Advancing Super-Resolution for Remote Sensing",
                        "abstract": "Super-Resolution for remote sensing has the potential for huge impact on planet monitoring by producing accurate and realistic high resolution imagery on a frequent basis and a global scale. Despite a lot of attention, several inconsistencies and challenges have prevented it from being deployed in practice. These include the lack of effective metrics, fragmented and relatively small-scale datasets for training, insufficient comparisons across a suite of methods, and unclear evidence for the use of super-resolution outputs for machine consumption. This work presents a new metric for super-resolution, CLIPScore, that corresponds far better with human judgments than previous metrics on an extensive study. We use CLIPScore to evaluate four standard methods on a new large-scale dataset, S2-NAIP, and three existing benchmark datasets, and find that generative adversarial networks easily outperform more traditional L2 loss-based models and are more semantically accurate than modern diffusion models. We also find that using CLIPScore as an auxiliary loss can speed up the training of GANs by 18x and lead to improved outputs, resulting in an effective model in diverse geographies across the world which we will release publicly. The dataset, pre-trained model weights, and code are available at https://github.com/allenai/satlas-super-resolution/."
                    },
                    {
                        "title": "Holodeck: Language Guided Generation of 3D Embodied AI Environments",
                        "abstract": "3D simulated environments play a critical role in Embodied AI, but their creation requires expertise and extensive manual effort, restricting their diversity and scope. To miti-gate this limitation, we present Holodeck, a system that generates 3D environments to match a user-supplied prompt fullyautomatedly. Holodeck can generate diverse scenes, e.g., arcades, spas, and museums, adjust the designs for styles, and can capture the semantics of complex queries such as \u201capartment for a researcher with a cat\u201d and \u201coffice of a professor who is a fan of Star Wars\u201d. Holodeck leverages a large language model (i.e., GPT-4) for common sense knowledge about what the scene might look like and uses a large collection of 3D assets from Objaverse to populate the scene with diverse objects. To address the challenge of positioning objects correctly, we prompt GPT-4 to generate spatial relational constraints between objects and then optimize the layout to satisfy those constraints. Our large-scale human evaluation shows that annotators prefer Holodeck over manually designed procedural baselines in residential scenes and that Holodeck can produce high-quality outputs for diverse scene types. We also demonstrate an exciting application of Holodeck in Embodied AI, training agents to navigate in novel scenes like music rooms and daycares without human-constructed data, which is a significant step forward in developing general-purpose embodied agents."
                    },
                    {
                        "title": "Neural Priming for Sample-Efficient Adaptation",
                        "abstract": "We propose Neural Priming, a technique for adapting large pretrained models to distribution shifts and downstream tasks given few or no labeled examples. Presented with class names or unlabeled test samples, Neural Priming enables the model to recall and conditions its parameters on relevant data seen throughout pretraining, thereby priming it for the test distribution. Neural Priming can be performed at test time, even for pretraining datasets as large as LAION-2B. Performing lightweight updates on the recalled data significantly improves accuracy across a variety of distribution shift and transfer learning benchmarks. Concretely, in the zero-shot setting, we see a 2.45% improvement in accuracy on ImageNet and 3.81% accuracy improvement on average across standard transfer learning benchmarks. Further, using Neural Priming at inference to adapt to distribution shift, we see a 1.41% accuracy improvement on ImageNetV2. These results demonstrate the effectiveness of Neural Priming in addressing the challenge of limited labeled data and changing distributions. Code is available at github.com/RAIVNLab/neural-priming."
                    },
                    {
                        "title": "EXCALIBUR: Encouraging and Evaluating Embodied Exploration",
                        "abstract": "Experience precedes understanding. Humans constantly explore and learn about their environment out of curiosity, gather information, and update their models of the world. On the other hand, machines are either trained to learn passively from static and fixed datasets, or taught to complete specific goal-conditioned tasks. To encourage the development of exploratory interactive agents, we present the EXCALIBUR benchmark. EXCALIBUR allows agents to explore their environment for long durations and then query their understanding of the physical world via inquiries like: \u201cis the small heavy red bowl made from glass?\u201d or \u201cis there a silver spoon heavier than the egg?\u201d. This design encourages agents to perform free-form home exploration without myopia induced by goal conditioning. Once the agents have answered a series of questions, they can renter the scene to refine their knowledge, update their beliefs, and improve their performance on the questions. Our experiments demonstrate the challenges posed by this dataset for the present-day state-of-the-art embodied systems and the headroom afforded to develop new innovative methods. Finally, we present a virtual reality interface that enables humans to seamlessly interact within the simulated world and use it to gather human performance measures. EXCALIBUR affords unique challenges in comparison to presentday benchmarks and represents the next frontier for embodied AI research."
                    },
                    {
                        "title": "Scene Graph Contrastive Learning for Embodied Navigation",
                        "abstract": "Training effective embodied AI agents often involves expert imitation, specialized components such as maps, or leveraging additional sensors for depth and localization. Another approach is to use neural architectures alongside self-supervised objectives which encourage better representation learning. However, in practice, there are few guarantees that these self-supervised objectives encode task-relevant information. We propose the Scene Graph Contrastive (SGC) loss, which uses scene graphs as training-only supervisory signals. The SGC loss does away with explicit graph decoding and instead uses contrastive learning to align an agent\u2019s representation with a rich graphical encoding of its environment. The SGC loss is simple to implement and encourages representations that encode objects\u2019 semantics, relationships, and history. By using the SGC loss, we attain gains on three embodied tasks: Object Navigation, Multi-Object Navigation, and Arm Point Navigation. Finally, we present studies and analyses which demonstrate the ability of our trained representation to encode semantic cues about the environment."
                    },
                    {
                        "title": "Curriculum Learning via Task Selection for Embodied Navigation",
                        "abstract": "001 002 003 004 005 006 007 008 009 010 011 012 013 014 015 016 017 018 019 020 021 022 023 024 025 026 027 028 029 030 031 032 033 034 035 036 037 038 039 040 041 042 043 044 045 046 047 048 049 050 051 052 053 054 055 056 057 058 059 060 061 062 063 064 065 066 067 068 069 070 071 072 073 074 075 076 077 078 079 080 081 082 083 084 085 086 087 088 089 090 091 092 093 094 095 096 097 098 099 100 101 102 103 104 105 106 107 CVPR # CVPR # CVPR 2023 Submission #. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE."
                    },
                    {
                        "title": "Harmonic Mobile Manipulation",
                        "abstract": "Recent advancements in robotics have enabled robots to navigate complex scenes or manipulate diverse objects independently. However, robots are still impotent in many household tasks requiring coordinated behaviors such as opening doors. The factorization of navigation and manipulation, while effective for some tasks, fails in scenarios requiring coordinated actions. To address this challenge, we introduce, HarmonicMM, an end-to-end learning method that optimizes both navigation and manipulation, showing notable improvement over existing techniques in everyday tasks. This approach is validated in simulated and real-world environments and adapts to novel unseen settings without additional tuning. Our contributions include a new benchmark for mobile manipulation and the successful deployment with only RGB visual observation in a real unseen apartment, demonstrating the potential for practical indoor robot deployment in daily life. More results are on our project site: https://rchalyang.github.io/HarmonicMM/"
                    },
                    {
                        "title": "Imitating Shortest Paths in Simulation Enables Effective Navigation and Manipulation in the Real World",
                        "abstract": "Reinforcement learning (RL) with dense rewards and imitation learning (IL) with human-generated trajectories are the most widely used approaches for training modern embodied agents. RL requires extensive reward shaping and auxiliary losses and is often too"
                    },
                    {
                        "title": "Unified-IO 2: Scaling Autoregressive Multimodal Models with Vision, Language, Audio, and Action",
                        "abstract": "We present Unified-IO 2,the. first autoregressive multi-modal model that is capable of understanding and generating image, text, audio, and action. To unify different modalities, we tokenize inputs and outputs - images, text, audio, action, bounding boxes etc., into a shared semantic space and then process them with a single encoder-decoder transformer model. Since training with such diverse modalities is challenging, we propose various architectural improvements to stabilize model training. We train our model from scratch on a large multimodal pre-training corpus from diverse sources with a multimodal mixture of denoisers objective. To learn an expansive set of skills, such as following multimodal instructions, we construct and. finetune on an ensemble of 120 datasets with prompts and augmentations. With a single unified model, Unified-io 2 achieves state-of-the-art performance on the GRIT benchmark and strong results in more than 35 benchmarks, including image generation and understanding, natural language understanding, video and audio understanding, and robotic manipulation. We release all our models to the research community."
                    }
                ]
            },
            "82496b90-a053-4849-a5a7-a7ff903bb23c": {
                "pk": "82496b90-a053-4849-a5a7-a7ff903bb23c",
                "name": "Ranjay Krishna",
                "collaborators": [
                    "Jiafei Duan",
                    "Dieter Fox",
                    "Wilbert Pumacay",
                    "Yushi Hu",
                    "Kiana Ehsani",
                    "Eli VanderBilt",
                    "Luca Weihs",
                    "Aniruddha Kembhavi",
                    "Wentao Yuan",
                    "Dan Roth"
                ],
                "domain": [
                    "Robotics",
                    "Vision-Language Models",
                    "Reinforcement Learning",
                    "Human-Robot Interaction"
                ],
                "publications": [
                    {
                        "title": "AHA: A Vision-Language-Model for Detecting and Reasoning Over Failures in Robotic Manipulation",
                        "abstract": "Robotic manipulation in open-world settings requires not only task execution but also the ability to detect and learn from failures. While recent advances in vision-language models (VLMs) and large language models (LLMs) have improved robots' spatial reasoning and problem-solving abilities, they still struggle with failure recognition, limiting their real-world applicability. We introduce AHA, an open-source VLM designed to detect and reason about failures in robotic manipulation using natural language. By framing failure detection as a free-form reasoning task, AHA identifies failures and provides detailed, adaptable explanations across different robots, tasks, and environments. We fine-tuned AHA using FailGen, a scalable framework that generates the first large-scale dataset of robotic failure trajectories, the AHA dataset. FailGen achieves this by procedurally perturbing successful demonstrations from simulation. Despite being trained solely on the AHA dataset, AHA generalizes effectively to real-world failure datasets, robotic systems, and unseen tasks. It surpasses the second-best model (GPT-4o in-context learning) by 10.3% and exceeds the average performance of six compared models including five state-of-the-art VLMs by 35.3% across multiple metrics and datasets. We integrate AHA into three manipulation frameworks that utilize LLMs/VLMs for reinforcement learning, task and motion planning, and zero-shot trajectory generation. AHA's failure feedback enhances these policies' performances by refining dense reward functions, optimizing task planning, and improving sub-task verification, boosting task success rates by an average of 21.4% across all three tasks compared to GPT-4 models."
                    },
                    {
                        "title": "BLINK: Multimodal Large Language Models Can See but Not Perceive",
                        "abstract": "We introduce Blink, a new benchmark for multimodal language models (LLMs) that focuses on core visual perception abilities not found in other evaluations. Most of the Blink tasks can be solved by humans\"within a blink\"(e.g., relative depth estimation, visual correspondence, forensics detection, and multi-view reasoning). However, we find these perception-demanding tasks cast significant challenges for current multimodal LLMs because they resist mediation through natural language. Blink reformats 14 classic computer vision tasks into 3,807 multiple-choice questions, paired with single or multiple images and visual prompting. While humans get 95.70% accuracy on average, Blink is surprisingly challenging for existing multimodal LLMs: even the best-performing GPT-4V and Gemini achieve accuracies of 51.26% and 45.72%, only 13.17% and 7.63% higher than random guessing, indicating that such perception abilities have not\"emerged\"yet in recent multimodal LLMs. Our analysis also highlights that specialist CV models could solve these problems much better, suggesting potential pathways for future improvements. We believe Blink will stimulate the community to help multimodal LLMs catch up with human-level visual perception."
                    },
                    {
                        "title": "EVE: Enabling Anyone to Train Robots using Augmented Reality",
                        "abstract": "The increasing affordability of robot hardware is accelerating the integration of robots into everyday activities. However, training a robot to automate a task requires expensive trajectory data where a trained human annotator moves a physical robot to train it. Consequently, only those with access to robots produce demonstrations to train robots. In this work, we remove this restriction with EVE, an iOS app that enables everyday users to train robots using intuitive augmented reality visualizations, without needing a physical robot. With EVE, users can collect demonstrations by specifying waypoints with their hands, visually inspecting the environment for obstacles, modifying existing waypoints, and verifying collected trajectories. In a user study (N=14, D=30) consisting of three common tabletop tasks, EVE outperformed three state-of-the-art interfaces in success rate and was comparable to kinesthetic teaching-physically moving a physical robot-in completion time, usability, motion intent communication, enjoyment, and preference (mean of p=0.30). EVE allows users to train robots for personalized tasks, such as sorting desk supplies, organizing ingredients, or setting up board games. We conclude by enumerating limitations and design considerations for future AR-based demonstration collection systems for robotics."
                    },
                    {
                        "title": "Visual Sketchpad: Sketching as a Visual Chain of Thought for Multimodal Language Models",
                        "abstract": "Humans draw to facilitate reasoning: we draw auxiliary lines when solving geometry problems; we mark and circle when reasoning on maps; we use sketches to amplify our ideas and relieve our limited-capacity working memory. However, such actions are missing in current multimodal language models (LMs). Current chain-of-thought and tool-use paradigms only use text as intermediate reasoning steps. In this work, we introduce Sketchpad, a framework that gives multimodal LMs a visual sketchpad and tools to draw on the sketchpad. The LM conducts planning and reasoning according to the visual artifacts it has drawn. Different from prior work, which uses text-to-image models to enable LMs to draw, Sketchpad enables LMs to draw with lines, boxes, marks, etc., which is closer to human sketching and better facilitates reasoning. Sketchpad can also use specialist vision models during the sketching process (e.g., draw bounding boxes with object detection models, draw masks with segmentation models), to further enhance visual perception and reasoning. We experiment with a wide range of math tasks (including geometry, functions, graphs, and chess) and complex visual reasoning tasks. Sketchpad substantially improves performance on all tasks over strong base models with no sketching, yielding an average gain of 12.7% on math tasks, and 8.6% on vision tasks. GPT-4o with Sketchpad sets a new state of the art on all tasks, including V*Bench (80.3%), BLINK spatial reasoning (83.9%), and visual correspondence (80.8%). All codes and data are in https://visualsketchpad.github.io/."
                    },
                    {
                        "title": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Multimodal Models",
                        "abstract": "Today's most advanced multimodal models remain proprietary. The strongest open-weight models rely heavily on synthetic data from proprietary VLMs to achieve good performance, effectively distilling these closed models into open ones. As a result, the community is still missing foundational knowledge about how to build performant VLMs from scratch. We present Molmo, a new family of VLMs that are state-of-the-art in their class of openness. Our key innovation is a novel, highly detailed image caption dataset collected entirely from human annotators using speech-based descriptions. To enable a wide array of user interactions, we also introduce a diverse dataset mixture for fine-tuning that includes in-the-wild Q&A and innovative 2D pointing data. The success of our approach relies on careful choices for the model architecture details, a well-tuned training pipeline, and, most critically, the quality of our newly collected datasets, all of which will be released. The best-in-class 72B model within the Molmo family not only outperforms others in the class of open weight and data models but also compares favorably against proprietary systems like GPT-4o, Claude 3.5, and Gemini 1.5 on both academic benchmarks and human evaluation. We will be releasing all of our model weights, captioning and fine-tuning data, and source code in the near future. Select model weights, inference code, and demo are available at https://molmo.allenai.org."
                    },
                    {
                        "title": "Scaling Up LLM Reviews for Google Ads Content Moderation",
                        "abstract": "Large language models (LLMs) are powerful tools for content moderation, but their inference costs and latency make them prohibitive for casual use on large datasets, such as the Google Ads repository. This study proposes a method for scaling up LLM reviews for content moderation in Google Ads. First, we use heuristics to select candidates via filtering and duplicate removal, and create clusters of ads for which we select one representative ad per cluster. We then use LLMs to review only the representative ads. Finally, we propagate the LLM decisions for the representative ads back to their clusters. This method reduces the number of reviews by more than 3 orders of magnitude while achieving a 2x recall compared to a baseline non-LLM model. The success of this approach is a strong function of the representations used in clustering and label propagation; we found that cross-modal similarity representations yield better results than uni-modal representations."
                    },
                    {
                        "title": "Manipulate-Anything: Automating Real-World Robots using Vision-Language Models",
                        "abstract": "Large-scale endeavors like and widespread community efforts such as Open-X-Embodiment have contributed to growing the scale of robot demonstration data. However, there is still an opportunity to improve the quality, quantity, and diversity of robot demonstration data. Although vision-language models have been shown to automatically generate demonstration data, their utility has been limited to environments with privileged state information, they require hand-designed skills, and are limited to interactions with few object instances. We propose Manipulate-Anything, a scalable automated generation method for real-world robotic manipulation. Unlike prior work, our method can operate in real-world environments without any privileged state information, hand-designed skills, and can manipulate any static object. We evaluate our method using two setups. First, Manipulate-Anything successfully generates trajectories for all 7 real-world and 14 simulation tasks, significantly outperforming existing methods like VoxPoser. Second, Manipulate-Anything's demonstrations can train more robust behavior cloning policies than training with human demonstrations, or from data generated by VoxPoser, Scaling-up, and Code-As-Policies. We believe Manipulate-Anything can be a scalable method for both generating data for robotics and solving novel tasks in a zero-shot setting. Project page: https://robot-ma.github.io/."
                    },
                    {
                        "title": "RoboPoint: A Vision-Language Model for Spatial Affordance Prediction for Robotics",
                        "abstract": "From rearranging objects on a table to putting groceries into shelves, robots must plan precise action points to perform tasks accurately and reliably. In spite of the recent adoption of vision language models (VLMs) to control robot behavior, VLMs struggle to precisely articulate robot actions using language. We introduce an automatic synthetic data generation pipeline that instruction-tunes VLMs to robotic domains and needs. Using the pipeline, we train RoboPoint, a VLM that predicts image keypoint affordances given language instructions. Compared to alternative approaches, our method requires no real-world data collection or human demonstration, making it much more scalable to diverse environments and viewpoints. In addition, RoboPoint is a general model that enables several downstream applications such as robot navigation, manipulation, and augmented reality (AR) assistance. Our experiments demonstrate that RoboPoint outperforms state-of-the-art VLMs (GPT-4o) and visual prompting techniques (PIVOT) by 21.8% in the accuracy of predicting spatial affordance and by 30.5% in the success rate of downstream tasks. Project website: https://robo-point.github.io."
                    },
                    {
                        "title": "Ask Again, Then Fail: Large Language Models\u2019 Vacillations in Judgment",
                        "abstract": "We observe that current large language models often waver in their judgments when faced with follow-up questions, even if the original judgment was correct. This wavering presents a signi\ufb01cant challenge for generating reliable responses and building user trust. To comprehensively assess this issue, we introduce a F OLLOW - UP Q UESTIONING M ECHANISM along with two metrics to quantify this inconsistency, con\ufb01rming its widespread presence in current large language models. Furthermore, to mitigate this issue, we explore various prompting strategies for closed-source models, and develop a training-based framework U NWAVERING -FQ that teaches large language models to maintain their originally correct judgments through synthesized high-quality preference data. Our experimental re-sults con\ufb01rm the effectiveness of our framework and its ability to enhance the general capabilities of large language models."
                    },
                    {
                        "title": "THE COLOSSEUM: A Benchmark for Evaluating Generalization for Robotic Manipulation",
                        "abstract": "To realize effective large-scale, real-world robotic applications, we must evaluate how well our robot policies adapt to changes in environmental conditions. Unfortunately, a majority of studies evaluate robot performance in environments closely resembling or even identical to the training setup. We present THE COLOSSEUM, a novel simulation benchmark, with 20 diverse manipulation tasks, that enables systematical evaluation of models across 14 axes of environmental perturbations. These perturbations include changes in color, texture, and size of objects, table-tops, and backgrounds; we also vary lighting, distractors, physical properties perturbations and camera pose. Using THE COLOSSEUM, we compare 5 state-of-the-art manipulation models to reveal that their success rate degrades between 30-50% across these perturbation factors. When multiple perturbations are applied in unison, the success rate degrades $\\geq$75%. We identify that changing the number of distractor objects, target object color, or lighting conditions are the perturbations that reduce model performance the most. To verify the ecological validity of our results, we show that our results in simulation are correlated ($\\bar{R}^2 = 0.614$) to similar perturbations in real-world experiments. We open source code for others to use THE COLOSSEUM, and also release code to 3D print the objects used to replicate the real-world perturbations. Ultimately, we hope that THE COLOSSEUM will serve as a benchmark to identify modeling decisions that systematically improve generalization for manipulation. See https://robot-colosseum.github.io/ for more details."
                    },
                    {
                        "title": "Superposed Decoding: Multiple Generations from a Single Autoregressive Inference Pass",
                        "abstract": "Many applications today provide users with multiple auto-complete drafts as they type, including GitHub's code completion, Gmail's smart compose, and Apple's messaging auto-suggestions. Under the hood, language models support this by running an autoregressive inference pass to provide a draft. Consequently, providing $k$ drafts to the user requires running an expensive language model $k$ times. To alleviate the computation cost of running $k$ inference passes, we propose Superposed Decoding, a new decoding algorithm that generates $k$ drafts at the computation cost of one autoregressive inference pass. We achieve this by feeding a superposition of the most recent token embeddings from the $k$ drafts as input to the next decoding step of the language model. At every inference step we combine the $k$ drafts with the top-$k$ tokens to get $k^2$ new drafts and cache the $k$ most likely options, using an n-gram interpolation with minimal compute overhead to filter out incoherent generations. Our experiments show that $k$ drafts from Superposed Decoding are at least as coherent and factual as Nucleus Sampling and Greedy Decoding respectively, while being at least $2.44\\times$ faster for $k\\ge3$. In a compute-normalized setting, user evaluations demonstrably favor text generated by Superposed Decoding over Nucleus Sampling. Superposed Decoding can also be combined with other decoding strategies, resulting in universal coverage gains when scaling inference time compute. Code and more examples open-sourced at https://github.com/RAIVNLab/SuperposedDecoding."
                    },
                    {
                        "title": "Modeling Collaborator: Enabling Subjective Vision Classification with Minimal Human Effort via LLM Tool-Use",
                        "abstract": "From content moderation to wildlife conservation, the number of applications that require models to recognize nuanced or subjective visual concepts is growing. Traditionally, developing classifiers for such concepts requires substantial manual effort measured in hours, days, or even months to identify and annotate data needed for training. Even with recently proposed Agile Modeling techniques, which enable rapid bootstrapping of image classifiers, users are still required to spend 30 minutes or more of monotonous, repetitive data labeling just to train a single classifier. Drawing on Fiske's Cognitive Miser theory, we propose a new framework that alleviates manual effort by replacing human labeling with natural language interactions, reducing the total effort required to define a concept by an order of magnitude: from labeling 2,000 images to only 100 plus some natural language interactions. Our framework leverages recent advances in foundation models, both large language models and vision-language models, to carve out the concept space through conversation and by automatically labeling training data points. Most importantly, our framework eliminates the need for crowd-sourced annotations. Moreover, our framework ultimately produces lightweight classification models that are deploy-able in cost-sensitive scenarios. Across 15 subjective concepts and across 2 public image classification datasets, our trained models outperform traditional Agile Modeling as well as state-of-the-art zero-shot classification models like ALIGN, CLIP, CuPL, and large visual question answering models like PaLI-X."
                    },
                    {
                        "title": "ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition",
                        "abstract": "Our world is full of varied actions and moves across specialized domains that we, as humans, strive to identify and understand. Within any single domain, actions can often appear quite similar, making it challenging for deep models to distinguish them accurately. To evaluate the effectiveness of multimodal foundation models in helping us recognize such actions, we present ActionAtlas v1.0, a multiple-choice video question answering benchmark featuring short videos across various sports. Each video in the dataset is paired with a question and four or five choices. The question pinpoints specific individuals, asking which choice\"best\"describes their action within a certain temporal context. Overall, the dataset includes 934 videos showcasing 580 unique actions across 56 sports, with a total of 1896 actions within choices. Unlike most existing video question answering benchmarks that only cover simplistic actions, often identifiable from a single frame, ActionAtlas focuses on intricate movements and rigorously tests the model's capability to discern subtle differences between moves that look similar within each domain. We evaluate open and proprietary foundation models on this benchmark, finding that the best model, GPT-4o, achieves a maximum accuracy of 45.52%. Meanwhile, Non-expert crowd workers, provided with action description for each choice, achieve 61.64% accuracy, where random chance is approximately 21%. Our findings with state-of-the-art models indicate that having a high frame sampling rate is important for accurately recognizing actions in ActionAtlas, a feature that some leading proprietary video models, such as Gemini, do not include in their default configuration."
                    },
                    {
                        "title": "ImageInWords: Unlocking Hyper-Detailed Image Descriptions",
                        "abstract": "Despite the longstanding adage\"an image is worth a thousand words,\"generating accurate hyper-detailed image descriptions remains unsolved. Trained on short web-scraped image text, vision-language models often generate incomplete descriptions with visual inconsistencies. We address this via a novel data-centric approach with ImageInWords (IIW), a carefully designed human-in-the-loop framework for curating hyper-detailed image descriptions. Human evaluations on IIW data show major gains compared to recent datasets (+66%) and GPT4V (+48%) across comprehensiveness, specificity, hallucinations, and more. We also show that fine-tuning with IIW data improves these metrics by +31% against models trained with prior work, even with only 9k samples. Lastly, we evaluate IIW models with text-to-image generation and vision-language reasoning tasks. Our generated descriptions result in the highest fidelity images, and boost compositional reasoning by up to 6% on ARO, SVO-Probes, and Winoground datasets. We release the IIW Eval benchmark with human judgement labels, object and image-level annotations from our framework, and existing image caption datasets enriched via IIW-model."
                    },
                    {
                        "title": "SPOC: Imitating Shortest Paths in Simulation Enables Effective Navigation and Manipulation in the Real World",
                        "abstract": "Reinforcement learning (RL) with dense rewards and imitation learning (IL) with human-generated trajectories are the most widely used approaches for training modern embodied agents. RL requires extensive reward shaping and auxiliary losses and is often too slow and ineffective for long-horizon tasks. While IL with human supervision is effective, collecting human trajectories at scale is extremely expensive. In this work, we show that imitating shortest-path planners in simulation produces agents that, given a language instruction, can proficiently navigate, ex-plore, and manipulate objects in both simulation and in the real world using only RGB sensors (no depth map or GPS coordinates). This surprising result is enabled by our end-to-end, transformer-based, Spocarchitecture, power-ful visual encoders paired with extensive image augmentation, and the dramatic scale and diversity of our training data: millions of frames of shortest-path-expert trajectories collected inside approximately 200,000 procedu-rally generated houses containing 40,000 unique 3D as-sets. Our models, data, training code, and newly proposed 10-task benchmarking suite Choresare available in spoc-robot.github.io."
                    },
                    {
                        "title": "Holodeck: Language Guided Generation of 3D Embodied AI Environments",
                        "abstract": "3D simulated environments play a critical role in Embodied AI, but their creation requires expertise and extensive manual effort, restricting their diversity and scope. To miti-gate this limitation, we present Holodeck, a system that generates 3D environments to match a user-supplied prompt fullyautomatedly. Holodeck can generate diverse scenes, e.g., arcades, spas, and museums, adjust the designs for styles, and can capture the semantics of complex queries such as \u201capartment for a researcher with a cat\u201d and \u201coffice of a professor who is a fan of Star Wars\u201d. Holodeck leverages a large language model (i.e., GPT-4) for common sense knowledge about what the scene might look like and uses a large collection of 3D assets from Objaverse to populate the scene with diverse objects. To address the challenge of positioning objects correctly, we prompt GPT-4 to generate spatial relational constraints between objects and then optimize the layout to satisfy those constraints. Our large-scale human evaluation shows that annotators prefer Holodeck over manually designed procedural baselines in residential scenes and that Holodeck can produce high-quality outputs for diverse scene types. We also demonstrate an exciting application of Holodeck in Embodied AI, training agents to navigate in novel scenes like music rooms and daycares without human-constructed data, which is a significant step forward in developing general-purpose embodied agents."
                    },
                    {
                        "title": "Improving Interpersonal Communication by Simulating Audiences with Language Models",
                        "abstract": "How do we communicate with others to achieve our goals? We use our prior experience or advice from others, or construct a candidate utterance by predicting how it will be received. However, our experiences are limited and biased, and reasoning about potential outcomes can be difficult and cognitively challenging. In this paper, we explore how we can leverage Large Language Model (LLM) simulations to help us communicate better. We propose the Explore-Generate-Simulate (EGS) framework, which takes as input any scenario where an individual is communicating to an audience with a goal they want to achieve. EGS (1) explores the solution space by producing a diverse set of advice relevant to the scenario, (2) generates communication candidates conditioned on subsets of the advice, and (3) simulates the reactions from various audiences to determine both the best candidate and advice to use. We evaluate the framework on eight scenarios spanning the ten fundamental processes of interpersonal communication. For each scenario, we collect a dataset of human evaluations across candidates and baselines, and showcase that our framework's chosen candidate is preferred over popular generation mechanisms including Chain-of-Thought. We also find that audience simulations achieve reasonably high agreement with human raters across 5 of the 8 scenarios. Finally, we demonstrate the generality of our framework by applying it to real-world scenarios described by users on web forums. Through evaluations and demonstrations, we show that EGS enhances the effectiveness and outcomes of goal-oriented communication across a variety of situations, thus opening up new possibilities for the application of large language models in revolutionizing communication and decision-making processes."
                    },
                    {
                        "title": "Imitating Shortest Paths in Simulation Enables Effective Navigation and Manipulation in the Real World",
                        "abstract": "Reinforcement learning (RL) with dense rewards and imitation learning (IL) with human-generated trajectories are the most widely used approaches for training modern embodied agents. RL requires extensive reward shaping and auxiliary losses and is often too"
                    },
                    {
                        "title": "Davidsonian Scene Graph: Improving Reliability in Fine-grained Evaluation for Text-to-Image Generation",
                        "abstract": "Evaluating text-to-image models is notoriously difficult. A strong recent approach for assessing text-image faithfulness is based on QG/A (question generation and answering), which uses pre-trained foundational models to automatically generate a set of questions and answers from the prompt, and output images are scored based on whether these answers extracted with a visual question answering model are consistent with the prompt-based answers. This kind of evaluation is naturally dependent on the quality of the underlying QG and VQA models. We identify and address several reliability challenges in existing QG/A work: (a) QG questions should respect the prompt (avoiding hallucinations, duplications, and omissions) and (b) VQA answers should be consistent (not asserting that there is no motorcycle in an image while also claiming the motorcycle is blue). We address these issues with Davidsonian Scene Graph (DSG), an empirically grounded evaluation framework inspired by formal semantics, which is adaptable to any QG/A frameworks. DSG produces atomic and unique questions organized in dependency graphs, which (i) ensure appropriate semantic coverage and (ii) sidestep inconsistent answers. With extensive experimentation and human evaluation on a range of model configurations (LLM, VQA, and T2I), we empirically demonstrate that DSG addresses the challenges noted above. Finally, we present DSG-1k, an open-sourced evaluation benchmark that includes 1,060 prompts, covering a wide range of fine-grained semantic categories with a balanced distribution. We release the DSG-1k prompts and the corresponding DSG questions."
                    },
                    {
                        "title": "DreamSync: Aligning Text-to-Image Generation with Image Understanding Feedback",
                        "abstract": "Despite their wide-spread success, Text-to-Image models (T2I) still struggle to produce images that are both aesthetically pleasing and faithful to the user's input text. We introduce DreamSync, a model-agnostic training algorithm by design that improves T2I models to be faithful to the text input. DreamSync builds off a recent insight from TIFA's evaluation framework -- that large vision-language models (VLMs) can effectively identify the fine-grained discrepancies between generated images and the text inputs. DreamSync uses this insight to train T2I models without any labeled data; it improves T2I models using its own generations. First, it prompts the model to generate several candidate images for a given input text. Then, it uses two VLMs to select the best generation: a Visual Question Answering model that measures the alignment of generated images to the text, and another that measures the generation's aesthetic quality. After selection, we use LoRA to iteratively finetune the T2I model to guide its generation towards the selected best generations. DreamSync does not need any additional human annotation. model architecture changes, or reinforcement learning. Despite its simplicity, DreamSync improves both the semantic alignment and aesthetic appeal of two diffusion-based T2I models, evidenced by multiple benchmarks (+1.7% on TIFA, +2.9% on DSG1K, +3.4% on VILA aesthetic) and human evaluation."
                    }
                ]
            }
        }
    },
    "2311.17855": {
        "paper_data": {
            "title": "Maximum Entropy Model Correction in Reinforcement Learning",
            "url": "http://arxiv.org/abs/2311.17855v1",
            "arxiv_id": "2311.17855",
            "authors": [
                "Amin Rakhsha",
                "Mete Kemertas",
                "Mohammad Ghavamzadeh",
                "Amir-massoud Farahmand"
            ],
            "abstract": "We propose and theoretically analyze an approach for planning with an approximate model in reinforcement learning that can reduce the adverse impact of model error. If the model is accurate enough, it accelerates the convergence to the true value function too. One of its key components is the MaxEnt Model Correction (MoCo) procedure that corrects the model's next-state distributions based on a Maximum Entropy density estimation formulation. Based on MoCo, we introduce the Model Correcting Value Iteration (MoCoVI) algorithm, and its sampled-based variant MoCoDyna. We show that MoCoVI and MoCoDyna's convergence can be much faster than the conventional model-free algorithms. Unlike traditional model-based algorithms, MoCoVI and MoCoDyna effectively utilize an approximate model and still converge to the correct value function.",
            "introduction": " introduction of a prior (Goodman, 2004), and constraint relaxation (Kazama and Tsujii, 2003; Dud\u00edk et al., 2004b). In this work, we use \u21132 2 regularization (Lau, 1994; Chen and Rosenfeld, 2000b; Lebanon and Lafferty, 2001; Zhang, 2004; Dud\u00edk et al., 2007) and leave the study of the other approaches to future work. The regularization is done by adding1 4\u03b22\u2225\u03bb\u22252 2to the objective of the dual problem (2.4) . This pushes the dual parameters to remain small. The hyperparameter \u03b2controls the amount of regularization. Smaller \u03b2leads a solution closer to the original one. Notice that with extreme regularization when \u03b2\u2192 \u221e , we get \u03bb= 0, which makes the solution of MaxEnt density estimation the same as the 5initial density estimate \u02c6p. The regularization of the dual problem has an intuitive interpretation in the primal problem. With the regularization, the primal problem (P1) is transformed to \u00afP(\u00b7|x, a)\u225cargmin qDKL(q\u2225\u02c6P(\u00b7|x, a) ) +1 \u03b22dX i=1\u0010 EX\u2032\u223cq[\u03d5i(X\u2032)]\u2212\u03c8i(x, a)\u00112 . (P2) We now have introduced a new hyperparameter \u03b2to MaxEnt MoCo. As \u03b2\u21920, the solution converges to that of the constrained problem (P1), because intuitively, \u03b2controls how much we trust the noisy observations \u03c8i. Smaller values of \u03b2means that we care about being consistent with the queries more than staying close to \u02c6P, and larger values of \u03b2shows the opposite preference. It turns out the impact of the choice of \u03b2is aligned with this intuition. As \u2225\u03f5Model\u2225\u221eincreases or \u2225\u03f5Query\u2225\u221edecreases, we should rely on the queries more and choose a smaller \u03b2. We provide the analysis for a general choice of\u03b2in the supplementary material, and here focus on when \u03b2=\u2225\u03f5Query\u2225\u221e/\u2225\u03f5Model\u2225\u221e. Theorem 1. Letc1=\u03b3\u221a 2/(1\u2212\u03b3),c2= 3\u03b3\u221a d/(1\u2212\u03b3), and \u03b2=\u2225\u03f5Query\u2225\u221e/\u2225\u03f5Model\u2225\u221e. For anywmax\u22650, we have \r\rV\u03c0PE\u2212\u00afV\u03c0PE\r\r \u221e\u22643c1\u2225\u03f5Model\u2225\u221einf \u2225w\u2225\u221e\u2264wmax\r\r\rV\u03c0PE\u2212dX i=1wi\u03d5i\r\r\r \u221e+c2\u2225\u03f5Query\u2225\u221e\u00b7wmax, \r\r\rV\u2217\u2212V\u00af\u03c0\u2217\r\r\r \u221e\u22646c1\u2225\u03f5Model\u2225\u221e 1\u22123c1\u2225\u03f5Model\u2225\u221einf \u2225w\u2225\u221e\u2264wmax\r\r\rV\u2217\u2212dX i=1wi\u03d5i\r\r\r \u221e+2c2\u2225\u03f5Query\u2225\u221e 1\u22123c1\u2225\u03f5Model\u2225\u221e\u00b7wmax. The above theorem shows that the error in the queries contribute an additive term to the final bounds compared to the exact query setting analyzed in Proposition 1. This term scales with wmax, which can be chosen arbitrarily to minimize the upper bound. Larger values of wmaxallow a better approximation of V\u03c0PEandV\u2217in the infimum terms, but amplify the query error \u03f5Query . Thus, ifV\u03c0PE(orV\u2217) can be approximated by some weighted sum of the basis functions using smaller weights, wmaxcan be chosen to be smaller. Unlike the exact case discussed in Proposition 1, the choice of basis functions is important beyond the subspace generated by their span. Therefore, transformations of the basis functions such as centralization, normalization, or orthogonalization might improve the effectiveness of MaxEnt Model Correction. One limitation of the results comparing MoCoDyna with Dyna, OS-Dyna and TD-learning. (Left) low (\u03bb= 0.1),(Middle) medium ( \u03bb= 0.5), and (Right) high ( \u03bb= 1) model errors. Each curve is average of 20 runs. Shaded areas show the standard error. For MoCoVI, we set the initial basis functions \u03d5ifori= 1,\u00b7\u00b7\u00b7, dconstant zero functions. We can set\u03c8i= 0fori= 1,\u00b7\u00b7\u00b7, dwithout querying P. This makes the comparison of algorithms fair as MoCoVI is not given extra queries before the first iteration. The convergence of MoCoVI with exact queries and \u03b2= 0is shown in Figures 1 and 3 for the control and PE problems. Figures 1 and 4 show the performance of MoCoDyna compared to other algorithms in the PE and control problems. As discussed after Theorem 1, it is beneficial to choose basis functions such that the true value function can be approximated withP iwi\u03d5ifor some small weights wi. To achieve this in our implementation, we initialize \u03d51:d+cwith an",
            "references": [
                {
                    "title": "Operator Splitting Value Iteration",
                    "abstract": "We introduce new planning and reinforcement learning algorithms for discounted MDPs that utilize an approximate model of the environment to accelerate the convergence of the value function. Inspired by the splitting approach in numerical linear algebra, we introduce Operator Splitting Value Iteration (OS-VI) for both Policy Evaluation and Control problems. OS-VI achieves a much faster convergence rate when the model is accurate enough. We also introduce a sample-based version of the algorithm called OS-Dyna. Unlike the traditional Dyna architecture, OS-Dyna still converges to the correct value function in presence of model approximation error."
                },
                {
                    "title": "Value Gradient weighted Model-Based Reinforcement Learning",
                    "abstract": "Model-based reinforcement learning (MBRL) is a sample efficient technique to obtain control policies, yet unavoidable modeling errors often lead performance deterioration. The model in MBRL is often solely fitted to reconstruct dynamics, state observations in particular, while the impact of model error on the policy is not captured by the training objective. This leads to a mismatch between the intended goal of MBRL, enabling good policy and value learning, and the target of the loss function employed in practice, future state prediction. Naive intuition would suggest that value-aware model learning would fix this problem and, indeed, several solutions to this objective mismatch problem have been proposed based on theoretical analysis. However, they tend to be inferior in practice to commonly used maximum likelihood (MLE) based approaches. In this paper we propose the Value-gradient weighted Model Learning (VaGraM), a novel method for value-aware model learning which improves the performance of MBRL in challenging settings, such as small model capacity and the presence of distracting state dimensions. We analyze both MLE and value-aware approaches and demonstrate how they fail to account for exploration and the behavior of function approximation when learning value-aware models and highlight the additional goals that must be met to stabilize optimization in the deep learning setting. We verify our analysis by showing that our loss function is able to achieve high returns on the Mujoco benchmark suite while being more robust than maximum likelihood based approaches."
                },
                {
                    "title": "Selective Dyna-style Planning Under Limited Model Capacity",
                    "abstract": "In model-based reinforcement learning, planning with an imperfect model of the environment has the potential to harm learning progress. But even when a model is imperfect, it may still contain information that is useful for planning. In this paper, we investigate the idea of using an imperfect model selectively. The agent should plan in parts of the state space where the model would be helpful but refrain from using the model where it would be harmful. An effective selective planning mechanism requires estimating predictive uncertainty, which arises out of aleatoric uncertainty, parameter uncertainty, and model inadequacy, among other sources. Prior work has focused on parameter uncertainty for selective planning. In this work, we emphasize the importance of model inadequacy. We show that heteroscedastic regression can signal predictive uncertainty arising from model inadequacy that is complementary to that which is detected by methods designed for parameter uncertainty, indicating that considering both parameter uncertainty and model inadequacy may be a more promising direction for effective selective planning than either in isolation."
                },
                {
                    "title": "Hallucinating Value: A Pitfall of Dyna-style Planning with Imperfect Environment Models",
                    "abstract": "Dyna-style reinforcement learning (RL) agents improve sample efficiency over model-free RL agents by updating the value function with simulated experience generated by an environment model. However, it is often difficult to learn accurate models of environment dynamics, and even small errors may result in failure of Dyna agents. In this paper, we investigate one type of model error: hallucinated states. These are states generated by the model, but that are not real states of the environment. We present the Hallucinated Value Hypothesis (HVH): updating values of real states towards values of hallucinated states results in misleading state-action values which adversely affect the control policy. We discuss and evaluate four Dyna variants; three which update real states toward simulated -- and therefore potentially hallucinated -- states and one which does not. The experimental results provide evidence for the HVH thus suggesting a fruitful direction toward developing Dyna algorithms robust to model error."
                },
                {
                    "title": "The Value-Improvement Path: Towards Better Representations for Reinforcement Learning",
                    "abstract": "In value-based reinforcement learning (RL), unlike in supervised learning, the agent faces not a single, stationary, approximation problem, but a sequence of value prediction problems. Each time the policy improves, the nature of the problem changes, shifting both the distribution of states and their values. In this paper we take a novel perspective, arguing that the value prediction problems faced by an RL agent should not be addressed in isolation, but rather as a single, holistic, prediction problem. An RL algorithm generates a sequence of policies that, at least approximately, improve towards the optimal policy. We explicitly characterize the associated sequence of value functions and call it the value-improvement path. Our main idea is to approximate the value-improvement path holistically, rather than to solely track the value function of the current policy. Specifically, we discuss the impact that this holistic view of RL has on representation learning. We demonstrate that a representation that spans the past value-improvement path will also provide an accurate value approximation for future policy improvements. We use this insight to better understand existing approaches to auxiliary tasks and to propose new ones. To test our hypothesis empirically, we augmented a standard deep RL agent with an auxiliary task of learning the value-improvement path. In a study of Atari 2600 games, the augmented agent achieved approximately double the mean and median performance of the baseline agent."
                },
                {
                    "title": "Policy-Aware Model Learning for Policy Gradient Methods",
                    "abstract": "This paper considers the problem of learning a model in model-based reinforcement learning (MBRL). We examine how the planning module of an MBRL algorithm uses the model, and propose that the model learning module should incorporate the way the planner is going to use the model. This is in contrast to conventional model learning approaches, such as those based on maximum likelihood estimate, that learn a predictive model of the environment without explicitly considering the interaction of the model and the planner. We focus on policy gradient type of planning algorithms and derive new loss functions for model learning that incorporate how the planner uses the model. We call this approach Policy-Aware Model Learning (PAML). We theoretically analyze a generic model-based policy gradient algorithm and provide a convergence guarantee for the optimized policy. We also empirically evaluate PAML on some benchmark problems, showing promising results."
                },
                {
                    "title": "When to Trust Your Model: Model-Based Policy Optimization",
                    "abstract": "Designing effective model-based reinforcement learning algorithms is difficult because the ease of data generation must be weighed against the bias of model-generated data. In this paper, we study the role of model usage in policy optimization both theoretically and empirically. We first formulate and analyze a model-based reinforcement learning algorithm with a guarantee of monotonic improvement at each step. In practice, this analysis is overly pessimistic and suggests that real off-policy data is always preferable to model-generated on-policy data, but we show that an empirical estimate of model generalization can be incorporated into such analysis to justify model usage. Motivated by this analysis, we then demonstrate that a simple procedure of using short model-generated rollouts branched from real data has the benefits of more complicated model-based algorithms without the usual pitfalls. In particular, this approach surpasses the sample efficiency of prior model-based methods, matches the asymptotic performance of the best model-free algorithms, and scales to horizons that cause other model-based methods to fail entirely."
                },
                {
                    "title": "Value-Aware Loss Function for Model-based Reinforcement Learning",
                    "abstract": "We consider the problem of estimating the transition probability kernel to be used by a model-based reinforcement learning (RL) algorithm. We argue that estimating a generative model that minimizes a probabilistic loss, such as the log-loss, is an overkill because it does not take into account the underlying structure of decision problem and the RL algorithm that intends to solve it. We introduce a loss function that takes the structure of the value function into account. We provide a \ufb01nite-sample upper bound for the loss function showing the dependence of the error on model approximation error, number of samples, and the complexity of the model space. We also empirically compare the method with the maximum likelihood estimator on a simple problem."
                },
                {
                    "title": "Self-Correcting Models for Model-Based Reinforcement Learning",
                    "abstract": "\n \n When an agent cannot represent a perfectly accurate model of its environment's dynamics, model-based reinforcement learning (MBRL) can fail catastrophically. Planning involves composing the predictions of the model; when flawed predictions are composed, even minor errors can compound and render the model useless for planning. Hallucinated Replay (Talvitie 2014) trains the model to \"correct\" itself when it produces errors, substantially improving MBRL with flawed models. This paper theoretically analyzes this approach, illuminates settings in which it is likely to be effective or ineffective, and presents a novel error bound, showing that a model's ability to self-correct is more tightly related to MBRL performance than one-step prediction error. These results inspire an MBRL algorithm for deterministic MDPs with performance guarantees that are robust to model class limitations.\n \n"
                },
                {
                    "title": "Policy Error Bounds for Model-Based Reinforcement Learning with Factored Linear Models",
                    "abstract": "In this paper we study a model-based approach to calculating approximately optimal policies in Markovian Decision Processes. In particular, we derive novel bounds on the loss of using a policy derived from a factored linear model, a class of models which generalize numerous previous models out of those that come with strong computational guarantees. For the first time in the literature, we derive performance bounds for model-based techniques where the model inaccuracy is measured in weighted norms. Moreover, our bounds show a decreased sensitivity to the discount factor and, unlike similar bounds derived for other approaches, they are insensitive to measure mismatch. Similarly to previous works, our proofs are also based on contraction arguments, but with the main differences that we use carefully constructed norms building on Banach lattices, and the contraction property is only assumed for operators acting on \"compressed\" spaces, thus weakening previous assumptions, while strengthening previous results."
                },
                {
                    "title": "MCMC using Hamiltonian dynamics",
                    "abstract": "Hamiltonian dynamics can be used to produce distant proposals for the Metropolis algorithm, thereby avoiding the slow exploration of the state space that results from the diffusive behaviour of simple random-walk proposals. Though originating in physics, Hamiltonian dynamics can be applied to most problems with continuous state spaces by simply introducing fictitious\"momentum\"variables. A key to its usefulness is that Hamiltonian dynamics preserves volume, and its trajectories can thus be used to define complex mappings without the need to account for a hard-to-compute Jacobian factor - a property that can be exactly maintained even when the dynamics is approximated by discretizing time. In this review, I discuss theoretical and practical aspects of Hamiltonian Monte Carlo, and present some of its variations, including using windows of states for deciding on acceptance or rejection, computing trajectories using fast approximations, tempering during the course of a trajectory to handle isolated modes, and short-cut methods that prevent useless trajectories from taking much computation time."
                },
                {
                    "title": "Error Propagation for Approximate Policy and Value Iteration",
                    "abstract": "We address the question of how the approximation error/Bellman residual at each iteration of the Approximate Policy/Value Iteration algorithms influences the quality of the resulted policy. We quantify the performance loss as the Lp norm of the approximation error/Bellman residual at each iteration. Moreover, we show that the performance loss depends on the expectation of the squared Radon-Nikodym derivative of a certain distribution rather than its supremum \u2013 as opposed to what has been suggested by the previous results. Also our results indicate that the contribution of the approximation/Bellman error to the performance loss is more prominent in the later iterations of API/AVI, and the effect of an error term in the earlier iterations decays exponentially fast."
                },
                {
                    "title": "Convex Optimization Theory",
                    "abstract": "An insightful, concise, and rigorous treatment of the basic theory of convex sets and functions in finite dimensions, and the Dual problem the feasible if it is they. Subgradient methods applied mathematics and sofware full. Ellipsoid method frankwolfe for publication. Arg max are the special case when choosing such. Unlike some convex programming lp a candidate solutions is they possess multiple to start! Operations research because this method which one would want. However for a project that lie. Classical optimization problem of agents that converge. For publication another criterion for this may not dominated by far. Gradient methods are some of applied to optimization problems may. The conditions using the objective function is a final. Arg max are allowed set of non convex course. This finite time average of convex sets can. Convexity theory convex if it can be efficiently and algorithms proposed for classes. The book is not distinguish maxima, are even harder to a large. However it is not refer to relax the hessian matrix in terms of linear programming. Present the problem of making usually, much slower than modern. Some combinatorial optimization and increasingly popular method but not done by the use divergent series. For the supremum operator for every equality constraint manifold dimension. The drift plus penalty method for many optimization. The problem itself which the class of hessians."
                },
                {
                    "title": "Maximum Entropy Density Estimation with Generalized Regularization and an Application to Species Distribution Modeling",
                    "abstract": "We present a unified and complete account of maximum entropy density estimation subject to constraints represented by convex potential functions or, alternatively, by convex regularization. We provide fully general performance guarantees and an algorithm with a complete convergence proof. As special cases, we easily derive performance guarantees for many known regularization types, including l1, l2, l22, and l2 + l22 style regularization. We propose an algorithm solving a large and general subclass of generalized maximum entropy problems, including all discussed in the paper, and prove its convergence. Our approach generalizes and unifies techniques based on information geometry and Bregman divergences as well as those based more directly on compactness. Our work is motivated by a novel application of maximum entropy to species distribution modeling, an important problem in conservation biology and ecology. In a set of experiments on real-world data, we demonstrate the utility of maximum entropy in this setting. We explore effects of different feature types, sample sizes, and regularization levels on the performance of maxent, and discuss interpretability of the resulting models."
                },
                {
                    "title": "Performance Bounds in Lp-norm for Approximate Value Iteration",
                    "abstract": "Approximate value iteration (AVI) is a method for solving large Markov decision problems by approximating the optimal value function with a sequence of value function representations $V_n$ processed according to the iterations $V_{n+1}=\\mathcal{AT} V_n$, where $\\mathcal{T}$ is the so-called Bellman operator and $\\mathcal{A}$ an approximation operator, which may be implemented by a supervised learning (SL) algorithm. Usual bounds on the asymptotic performance of AVI are established in terms of the $L_\\infty$-norm approximation errors induced by the SL algorithm. However, most widely used SL algorithms (such as least squares regression) return a function (the best fit) that minimizes an empirical approximation error in $L_p$-norm ($p\\geq 1$). In this paper, we extend the performance bounds of AVI to weighted $L_p$-norms, which enables us to directly relate the performance of AVI to the approximation power of the SL algorithm, hence assuring the tightness and practical relevance of these bounds. The main result is a performance bound of the resulting policies expressed in terms of the $L_p$-norm errors introduced by the successive approximations. The new bound takes into account a concentration coefficient that estimates how much the discounted future-state distributions starting from a probability measure used to assess the performance of AVI can possibly differ from the distribution used in the regression operation. We illustrate the tightness of the bounds on an optimal replacement problem."
                },
                {
                    "title": "Class-size Independent Generalization Analsysis of Some Discriminative Multi-Category Classification",
                    "abstract": "We consider the problem of deriving class-size independent generalization bounds for some regularized discriminative multi-category classification methods. In particular, we obtain an expected generalization bound for a standard formulation of multi-category support vector machines. Based on the theoretical result, we argue that the formulation over-penalizes misclassification error, which in theory may lead to poor generalization performance. A remedy, based on a generalization of multi-category logistic regression (conditional maximum entropy), is then proposed, and its theoretical properties are examined."
                },
                {
                    "title": "Error Bounds for Approximate Policy Iteration",
                    "abstract": "In Dynamic Programming, convergence of algorithms such as Value Iteration or Policy Iteration results -in discounted problems- from a contraction property of the back-up operator, guaranteeing convergence to its fixed-point. When approximation is considered, known results in Approximate Policy Iteration provide bounds on the closeness to optimality of the approximate value function obtained by successive policy improvement steps as a function of the maximum norm of value determination errors during policy evaluation steps. Unfortunately, such results have limited practical range since most function approximators (such as linear regression) select the best fit in a given class of parameterized functions by minimizing some (weighted) quadratic norm. \n \nIn this paper, we provide error bounds for Approximate Policy Iteration using quadratic norms, and illustrate those results in the case of feature-based linear function approximation."
                },
                {
                    "title": "Evaluation and Extension of Maximum Entropy Models with Inequality Constraints",
                    "abstract": "A maximum entropy (ME) model is usually estimated so that it conforms to equality constraints on feature expectations. However, the equality constraint is inappropriate for sparse and therefore unreliable features. This study explores an ME model with box-type inequality constraints, where the equality can be violated to reflect this unreliability. We evaluate the inequality ME model using text categorization datasets. We also propose an extension of the inequality ME model, which results in a natural integration with the Gaussian MAP estimation. Experimental results demonstrate the advantage of the inequality models and the proposed extension."
                },
                {
                    "title": "A Comparison of Algorithms for Maximum Entropy Parameter Estimation",
                    "abstract": "Conditional maximum entropy (ME) models provide a general purpose machine learning technique which has been successfully applied to fields as diverse as computer vision and econometrics, and which is used for a wide variety of classification problems in natural language processing. However, the flexibility of ME models is not without cost. While parameter estimation for ME models is conceptually straightforward, in practice ME models for typical natural language tasks are very large, and may well contain many thousands of free parameters. In this paper, we consider a number of algorithms for estimating the parameters of ME models, including iterative scaling, gradient ascent, conjugate gradient, and variable metric methods. Sur-prisingly, the standardly used iterative scaling algorithms perform quite poorly in comparison to the others, and for all of the test problems, a limited-memory variable metric algorithm outperformed the other choices."
                },
                {
                    "title": "Boosting and Maximum Likelihood for Exponential Models",
                    "abstract": "We derive an equivalence between AdaBoost and the dual of a convex optimization problem, showing that the only difference between minimizing the exponential loss used by AdaBoost and maximum likelihood for exponential models is that the latter requires the model to be normalized to form a conditional probability distribution over labels. In addition to establishing a simple and easily understood connection between the two methods, this framework enables us to derive new regularization procedures for boosting that directly correspond to penalized maximum likelihood. Experiments on UCI datasets support our theoretical analysis and give additional insight into the relationship between boosting and logistic regression."
                },
                {
                    "title": "Inducing Features of Random Fields",
                    "abstract": "We present a technique for constructing random fields from a set of training samples. The learning paradigm builds increasingly complex fields by allowing potential functions, or features, that are supported by increasingly large subgraphs. Each feature has a weight that is trained by minimizing the Kullback-Leibler divergence between the model and the empirical distribution of the training data. A greedy algorithm determines how features are incrementally added to the field and an iterative scaling algorithm is used to estimate the optimal values of the weights. The random field models and techniques introduced in this paper differ from those common to much of the computer vision literature in that the underlying random fields are non-Markovian and have a large number of parameters that must be estimated. Relations to other learning approaches, including decision trees, are given. As a demonstration of the method, we describe its application to the problem of automatic word classification in natural language processing."
                },
                {
                    "title": "Dual Methods in Entropy Maximization. Application to Some Problems in Crystallography",
                    "abstract": "This paper is devoted to some infinite-dimensional optimization problems with finitely many constraints. These deal with entropy maximization, and this paper is particularly concerned with those or..."
                },
                {
                    "title": "Duality relationships for entropy-like minimization problems",
                    "abstract": "This paper considers the minimization of a convex integral functional over the positive cone of an $L_p $ space, subject to a finite number of linear equality constraints. Such problems arise in spectral estimation, where the bjective function is often entropy-like, and in constrained approximation. The Lagrangian dual problem is finite-dimensional and unconstrained. Under a quasi-interior constraint qualification, the primal and dual values are equal, with dual attainment. Examples show the primal value may not be attained. Conditions are given that ensure that the primal optimal solution can be calculated directly from a dual optimum. These conditions are satisfied in many examples."
                },
                {
                    "title": "Information Theory and Statistical Mechanics",
                    "abstract": "Treatment of the predictive aspect of statistical mechanics as a form of statistical inference is extended to the density-matrix formalism and applied to a discussion of the relation between irreversibility and information loss. A principle of \"statistical complementarity\" is pointed out, according to which the empirically verifiable probabilities of statistical mechanics necessarily correspond to incomplete predictions. A preliminary discussion is given of the second law of thermodynamics and of a certain class of irreversible processes, in an approximation equivalent to that of the semiclassical theory of radiation."
                },
                {
                    "title": "VIPer: Iterative Value-Aware Model Learning on the Value Improvement Path",
                    "abstract": "We propose a practical and generalizable Decision-Aware Model-Based Reinforcement Learning algorithm. We extend the frameworks of VAML (Farahmand et al., 2017) and IterVAML (Farahmand, 2018), which have been shown to be difficult to scale to high-dimensional and continuous environments (Lovatto et al., 2020a; Modhe et al., 2021; Voelcker et al., 2022). We propose to use the notion of the Value Improvement Path (Dabney et al., 2020) to improve the generalization of VAML-like model learning. We show theoretically for linear and tabular spaces that our proposed algorithm is sensible, justifying extension to non-linear and continuous spaces. We also present a detailed implementation proposal based on these ideas."
                },
                {
                    "title": "Decision-Aware Model Learning for Actor-Critic Methods: When Theory Does Not Meet Practice",
                    "abstract": "Actor-Critic methods are a prominent class of modern reinforcement learning algorithms based on the classic Policy Iteration procedure. Despite many successful cases, Actor-Critic methods tend to require a gigantic number of experiences and can be very unstable. Recent approaches have advocated learning and using a world model to improve sample ef\ufb01ciency and reduce reliance on the value function estimate. However, learning an accurate dynamics model of the world remains challenging, often requiring computationally costly and data-hungry models. More recent work has shown that learning an everywhere accurate model is unnecessary and often detrimental to the overall task; instead, the agent should improve the world model on task-critical regions. For example, in Iterative Value-Aware Model Learning, the authors extend model-based value iteration by incorporating the value function (estimate) into the model loss function, showing the novel model objective re\ufb02ects improved performance in the end task. Therefore, it seems natural to expect that model-based Actor-Critic methods can bene\ufb01t equally from learning value-aware models, improving overall task performance, or reducing the need for large, expensive models. However, we show empirically that combining Actor-Critic and value-aware model learning can be quite dif\ufb01cult and that naive approaches such as maximum likelihood estimation often achieve superior performance with less computational cost. Our results suggest that, despite theoretical guarantees, learning a value-aware model in continuous domains does not ensure better performance on the overall task."
                },
                {
                    "title": "Iterative Value-Aware Model Learning",
                    "abstract": "This paper introduces a model-based reinforcement learning (MBRL) framework that incorporates the underlying decision problem in learning the transition model of the environment. This is in contrast with conventional approaches to MBRL that learn the model of the environment, for example by finding the maximum likelihood estimate, without taking into account the decision problem. Value-Aware Model Learning (VAML) framework argues that this might not be a good idea, especially if the true model of the environment does not belong to the model class from which we are estimating the model. The original VAML framework, however, may result in an optimization problem that is difficult to solve. This paper introduces a new MBRL class of algorithms, called Iterative VAML, that benefits from the structure of how the planning is performed (i.e., through approximate value iteration) to devise a simpler optimization problem. The paper theoretically analyzes Iterative VAML and provides finite sample error upper bound guarantee for it."
                },
                {
                    "title": "Approximate modified policy iteration and its application to the game of Tetris",
                    "abstract": "Modified policy iteration (MPI) is a dynamic programming (DP) algorithm that contains the two celebrated policy and value iteration methods. Despite its generality, MPI has not been thoroughly studied, especially its approximation form which is used when the state and/or action spaces are large or infinite. In this paper, we propose three implementations of approximate MPI (AMPI) that are extensions of the well-known approximate DP algorithms: fitted-value iteration, fitted-Q iteration, and classification-based policy iteration. We provide error propagation analysis that unify those for approximate policy and value iteration. We develop the finite-sample analysis of these algorithms, which highlights the influence of their parameters. In the classification-based version of the algorithm (CBMPI), the analysis shows that MPI's main parameter controls the balance between the estimation error of the classifier and the overall value function approximation. We illustrate and evaluate the behavior of these new algorithms in the Mountain Car and Tetris problems. Remarkably, in Tetris, CBMPI outperforms the existing DP approaches by a large margin, and competes with the current state-of-the-art methods while using fewer samples."
                },
                {
                    "title": "Exponential Priors for Maximum Entropy Models",
                    "abstract": "Maximum entropy models are a common modeling technique, but prone to over\ufb01tting. We show that using an exponential distribution as a prior leads to bounded absolute discounting by a constant. We show that this prior is better motivated by the data than previous techniques such as a Gaussian prior, and often produces lower error rates. Exponential priors also lead to a simpler learning algorithm and to easier to understand behavior. Furthermore, exponential priors help explain the success of some previous smoothing techniques, and suggest simple variations that work better."
                },
                {
                    "title": "A survey of smoothing techniques for ME models",
                    "abstract": "In certain contexts, maximum entropy (ME) modeling can be viewed as maximum likelihood (ML) training for exponential models, and like other ML methods is prone to overfitting of training data. Several smoothing methods for ME models have been proposed to address this problem, but previous results do not make it clear how these smoothing methods compare with smoothing methods for other types of related models. In this work, we survey previous work in ME smoothing and compare the performance of several of these algorithms with conventional techniques for smoothing n-gram language models. Because of the mature body of research in n-gram model smoothing and the close connection between ME and conventional n-gram models, this domain is well-suited to gauge the performance of ME smoothing methods. Over a large number of data sets, we find that fuzzy ME smoothing performs as well as or better than all other algorithms under consideration. We contrast this method with previous n-gram smoothing methods to explain its superior performance."
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                },
                {
                    "title": "Adaptive statistical language modeling",
                    "abstract": "The trigram statistical language model is remarkably successful when used in such applications as speech recognition. However, the trigram model is static in that it only considers the previous two words when making a prediction about a future word. The work presented here attempts to improve upon the trigram model by considering additional contextual and longer distance information. This is frequently referred to in the literature as adaptive statistical language modelling because the model is thought of as adapting to the longer term information. This work considers the creation of topic specific models, statistical evidence from the presence or absence of triggers, or related words, in the document history (document triggers) and in the current sentence (in-sentence triggers), and the incorporation of the document cache, which predicts the probability of a word by considering its frequency in the document history. An important result of this work is that the presence of self-triggers, that is, whether or not the word itself occurred in the document history, is an extremely important piece of information. A maximum entropy (ME) approach will be used in many instances to incorporate information from different sources. Maximum entropy considers a model which maximizes entropy while satisfying the constraints presented by the information we wish to incorporate. The generalized iterative scaling (GIS) algorithm can be used to compute the maximum entropy solution. This work also considers various methods of smoothing the information in a maximum entropy model. An inportant result is that smoothing improves performance noticibly and that Good-Turing discounting is an effective method of smoothing. Thesis Supervisor: Victor Zue Title: Principal Research Scientist, Department of Electrical Engineering and Computer Science"
                },
                {
                    "title": "Axiomatic derivation of the principle of maximum entropy and the principle of minimum cross-entropy",
                    "abstract": "Jaynes's principle of maximum entropy and Kullbacks principle of minimum cross-entropy (minimum directed divergence) are shown to be uniquely correct methods for inductive inference when new information is given in the form of expected values. Previous justifications use intuitive arguments and rely on the properties of entropy and cross-entropy as information measures. The approach here assumes that reasonable methods of inductive inference should lead to consistent results when there are different ways of taking the same information into account (for example, in different coordinate system). This requirement is formalized as four consistency axioms. These are stated in terms of an abstract information operator and make no reference to information measures. It is proved that the principle of maximum entropy is correct in the following sense: maximizing any function but entropy will lead to inconsistency unless that function and entropy have identical maxima. In other words given information in the form of constraints on expected values, there is only one (distribution satisfying the constraints that can be chosen by a procedure that satisfies the consistency axioms; this unique distribution can be obtained by maximizing entropy. This result is established both directly and as a special case (uniform priors) of an analogous result for the principle of minimum cross-entropy. Results are obtained both for continuous probability densities and for discrete distributions."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.SY",
                "eess.SY",
                "math.OC",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively incorporate regularization in MaxEnt density estimation to improve the accuracy of model correction in reinforcement learning?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the field of reinforcement learning, particularly in scenarios where model inaccuracies can significantly impact decision-making. By improving the regularization techniques in MaxEnt density estimation, we can enhance the robustness of learning algorithms, leading to more reliable and efficient models. This research could pave the way for practical applications in various domains, such as robotics, finance, and healthcare, where accurate predictions are essential. Furthermore, it will stimulate future research into more sophisticated regularization methods and their implications for other machine learning frameworks.\n\n### [Question 3] - Why is it hard?\nThe challenges in addressing this problem stem from the need to balance the trade-off between fitting the model to noisy observations and maintaining consistency with prior estimates. Naive approaches may fail because they do not adequately account for the complexities introduced by model errors and query inaccuracies. Technical obstacles include determining the optimal hyperparameter \u03b2, which influences the degree of regularization, and the need for effective basis function transformations to improve approximation accuracy. Theoretical complexities arise from the interactions between model errors and the regularization process, making it difficult to derive generalizable solutions.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often overlooked the nuanced relationship between regularization and model correction in MaxEnt frameworks. Limitations in existing solutions include a lack of comprehensive analysis on the impact of hyperparameter choices and insufficient exploration of basis function transformations. Barriers such as the complexity of deriving theoretical guarantees and the challenge of empirical validation have hindered progress. Our approach differs by systematically analyzing the effects of regularization and providing a framework for selecting basis functions that enhance model performance, thus addressing gaps in prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves the use of \u21132 regularization in the dual problem of MaxEnt density estimation, with a focus on the hyperparameter \u03b2, which we will set based on the ratio of query error to model error. We will utilize a dataset from reinforcement learning environments to evaluate our approach, measuring performance through metrics such as convergence rates and error bounds. The expected outcomes include improved accuracy in model correction, demonstrated through comparative analysis with existing algorithms like Dyna and TD-learning, and insights into the optimal selection of basis functions for enhanced learning efficiency."
            }
        },
        "author_data": {
            "2a43977a-6f38-45eb-9bfa-017598e4ff04": {
                "pk": "2a43977a-6f38-45eb-9bfa-017598e4ff04",
                "name": "Amin Rakhsha",
                "collaborators": [
                    "Amir-massoud Farahmand",
                    "Xiaojin Zhu",
                    "A. Singla",
                    "Goran Radanovic",
                    "R. Devidze",
                    "Jongmin Lee",
                    "Ernest K. Ryu",
                    "Mark Bedaywi",
                    "Andrew Wang",
                    "M. Ghavamzadeh",
                    "Xuezhou Zhang"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Security",
                    "Optimization",
                    "Control Theory"
                ],
                "publications": [
                    {
                        "title": "Deflated Dynamics Value Iteration",
                        "abstract": "The Value Iteration (VI) algorithm is an iterative procedure to compute the value function of a Markov decision process, and is the basis of many reinforcement learning (RL) algorithms as well. As the error convergence rate of VI as a function of iteration $k$ is $O(\\gamma^k)$, it is slow when the discount factor $\\gamma$ is close to $1$. To accelerate the computation of the value function, we propose Deflated Dynamics Value Iteration (DDVI). DDVI uses matrix splitting and matrix deflation techniques to effectively remove (deflate) the top $s$ dominant eigen-structure of the transition matrix $\\mathcal{P}^{\\pi}$. We prove that this leads to a $\\tilde{O}(\\gamma^k |\\lambda_{s+1}|^k)$ convergence rate, where $\\lambda_{s+1}$is $(s+1)$-th largest eigenvalue of the dynamics matrix. We then extend DDVI to the RL setting and present Deflated Dynamics Temporal Difference (DDTD) algorithm. We empirically show the effectiveness of the proposed algorithms."
                    },
                    {
                        "title": "PID Accelerated Temporal Difference Algorithms",
                        "abstract": "Long-horizon tasks, which have a large discount factor, pose a challenge for most conventional reinforcement learning (RL) algorithms. Algorithms such as Value Iteration and Temporal Difference (TD) learning have a slow convergence rate and become inefficient in these tasks. When the transition distributions are given, PID VI was recently introduced to accelerate the convergence of Value Iteration using ideas from control theory. Inspired by this, we introduce PID TD Learning and PID Q-Learning algorithms for the RL setting, in which only samples from the environment are available. We give a theoretical analysis of the convergence of PID TD Learning and its acceleration compared to the conventional TD Learning. We also introduce a method for adapting PID gains in the presence of noise and empirically verify its effectiveness."
                    },
                    {
                        "title": "Operator Splitting Value Iteration",
                        "abstract": "We introduce new planning and reinforcement learning algorithms for discounted MDPs that utilize an approximate model of the environment to accelerate the convergence of the value function. Inspired by the splitting approach in numerical linear algebra, we introduce Operator Splitting Value Iteration (OS-VI) for both Policy Evaluation and Control problems. OS-VI achieves a much faster convergence rate when the model is accurate enough. We also introduce a sample-based version of the algorithm called OS-Dyna. Unlike the traditional Dyna architecture, OS-Dyna still converges to the correct value function in presence of model approximation error."
                    },
                    {
                        "title": "Reward Poisoning in Reinforcement Learning: Attacks Against Unknown Learners in Unknown Environments",
                        "abstract": "We study black-box reward poisoning attacks against reinforcement learning (RL), in which an adversary aims to manipulate the rewards to mislead a sequence of RL agents with unknown algorithms to learn a nefarious policy in an environment unknown to the adversary a priori. That is, our attack makes minimum assumptions on the prior knowledge of the adversary: it has no initial knowledge of the environment or the learner, and neither does it observe the learner's internal mechanism except for its performed actions. We design a novel black-box attack, U2, that can provably achieve a near-matching performance to the state-of-the-art white-box attack, demonstrating the feasibility of reward poisoning even in the most challenging black-box setting."
                    },
                    {
                        "title": "Policy Teaching via Environment Poisoning: Training-time Adversarial Attacks against Reinforcement Learning",
                        "abstract": "We study a security threat to reinforcement learning where an attacker poisons the learning environment to force the agent into executing a target policy chosen by the attacker. As a victim, we consider RL agents whose objective is to find a policy that maximizes average reward in undiscounted infinite-horizon problem settings. The attacker can manipulate the rewards or the transition dynamics in the learning environment at training-time and is interested in doing so in a stealthy manner. We propose an optimization framework for finding an \\emph{optimal stealthy attack} for different measures of attack cost. We provide sufficient technical conditions under which the attack is feasible and provide lower/upper bounds on the attack cost. We instantiate our attacks in two settings: (i) an \\emph{offline} setting where the agent is doing planning in the poisoned environment, and (ii) an \\emph{online} setting where the agent is learning a policy using a regret-minimization framework with poisoned feedback. Our results show that the attacker can easily succeed in teaching any target policy to the victim under mild conditions and highlight a significant security threat to reinforcement learning agents in practice."
                    },
                    {
                        "title": "Policy Teaching in Reinforcement Learning via Environment Poisoning Attacks",
                        "abstract": "We study a security threat to reinforcement learning where an attacker poisons the learning environment to force the agent into executing a target policy chosen by the attacker. As a victim, we consider RL agents whose objective is to find a policy that maximizes reward in infinite-horizon problem settings. The attacker can manipulate the rewards and the transition dynamics in the learning environment at training-time, and is interested in doing so in a stealthy manner. We propose an optimization framework for finding an optimal stealthy attack for different measures of attack cost. We provide lower/upper bounds on the attack cost, and instantiate our attacks in two settings: (i) an offline setting where the agent is doing planning in the poisoned environment, and (ii) an online setting where the agent is learning a policy with poisoned feedback. Our results show that the attacker can easily succeed in teaching any target policy to the victim under mild conditions and highlight a significant security threat to reinforcement learning agents in practice."
                    }
                ]
            },
            "a01c59b5-a7b9-495d-be0a-58b6ba907b6a": {
                "pk": "a01c59b5-a7b9-495d-be0a-58b6ba907b6a",
                "name": "Mete Kemertas",
                "collaborators": [
                    "A. Jepson",
                    "Lan Xiao",
                    "Iqbal Mohomed",
                    "A. Fazly",
                    "Zhiming Hu",
                    "Caleb Phillips",
                    "Afsaneh Fazly",
                    "Amir-massoud Farahmand",
                    "Tristan Aumentado-Armstrong",
                    "\u00c1kos K\u00e1d\u00e1r",
                    "Federico Fancellu",
                    "Leila Pishdad",
                    "K. Derpanis",
                    "Tim Capes",
                    "Vishal Raheja"
                ],
                "domain": [
                    "Deep Learning",
                    "Cross-Modal Retrieval",
                    "Reinforcement Learning",
                    "Optimal Transport"
                ],
                "publications": [
                    {
                        "title": "Realizing Efficient On-Device Language-based Image Retrieval",
                        "abstract": "Advances in deep learning have enabled accurate language-based search and retrieval, e.g., over user photos, in the cloud. Many users prefer to store their photos in the home due to privacy concerns. As such, a need arises for models that can perform cross-modal search on resource-limited devices. State-of-the-art cross-modal retrieval models achieve high accuracy through learning entangled representations that enable fine-grained similarity calculation between a language query and an image, but at the expense of having a prohibitively high retrieval latency. Alternatively, there is a new class of methods that exhibits good performance with low latency, but requires a lot more computational resources, and an order of magnitude more training data (i.e. large web-scraped datasets consisting of millions of image-caption pairs) making them infeasible to use in a commercial context. From a pragmatic perspective, none of the existing methods are suitable for developing commercial applications for low-latency cross-modal retrieval on low-resource devices. We propose CrispSearch, a cascaded approach that greatly reduces the retrieval latency with minimal loss in ranking accuracy for on-device language-based image retrieval. The idea behind our approach is to combine a light-weight and runtime-efficient coarse model with a fine re-ranking stage. Given a language query, the coarse model effectively filters out many of the irrelevant image candidates. After this filtering, only a handful of strong candidates will be selected and sent to a fine model for re-ranking. Extensive experimental results with two SOTA models for the fine re-ranking stage, on standard benchmark datasets show that CrispSearch results in a speedup of up to 38 times over the SOTA fine methods with negligible performance degradation. Moreover, our method does not require millions of training instances, making it a pragmatic solution to on-device search and retrieval."
                    },
                    {
                        "title": "Efficient and Accurate Optimal Transport with Mirror Descent and Conjugate Gradients",
                        "abstract": "We design a novel algorithm for optimal transport by drawing from the entropic optimal transport, mirror descent and conjugate gradients literatures. Our scalable and GPU parallelizable algorithm is able to compute the Wasserstein distance with extreme precision, reaching relative error rates of $10^{-8}$ without numerical stability issues. Empirically, the algorithm converges to high precision solutions more quickly in terms of wall-clock time than a variety of algorithms including log-domain stabilized Sinkhorn's Algorithm. We provide careful ablations with respect to algorithm and problem parameters, and present benchmarking over upsampled MNIST images, comparing to various recent algorithms over high-dimensional problems. The results suggest that our algorithm can be a useful addition to the practitioner's optimal transport toolkit."
                    },
                    {
                        "title": "CrispSearch: low-latency on-device language-based image retrieval",
                        "abstract": "Advances in deep learning have enabled accurate language-based search and retrieval, e.g., over user photos, in the cloud. Many users prefer to store their photos in the home due to privacy concerns. As such, a need arises for models that can perform cross-modal search on resource-limited devices. State-of-the-art cross-modal retrieval models achieve high accuracy through learning entangled representations that enable fine-grained similarity calculation between a language query and an image, but at the expense of having a prohibitively high retrieval latency. Alternatively, there is a new class of methods that exhibits good performance with low latency, but requires a lot more computational resources, and an order of magnitude more training data (i.e. large web-scraped datasets consisting of millions of image-caption pairs) making them infeasible to use in a commercial context. From a pragmatic perspective, none of the existing methods are suitable for developing commercial applications for low-latency cross-modal retrieval on low-resource devices. We propose CrispSearch, a cascaded approach that greatly reduces the retrieval latency with minimal loss in ranking accuracy for on-device language-based image retrieval. The idea behind our approach is to combine a light-weight and runtime-efficient coarse model with a fine re-ranking stage. Given a language query, the coarse model effectively filters out many of the irrelevant image candidates. After this filtering, only a handful of strong candidates will be selected and sent to a fine model for re-ranking. Extensive experimental results with two SOTA models for the fine re-ranking stage, on two standard benchmark datasets (namely, MSCOCO and Flickr30k) show that CrispSearch results in a speedup of up to 38 times over the SOTA fine methods with negligible performance degradation. Moreover, our method does not require millions of training instances, making it a pragmatic solution to on-device search and retrieval."
                    },
                    {
                        "title": "Approximate Policy Iteration with Bisimulation Metrics",
                        "abstract": "Bisimulation metrics define a distance measure between states of a Markov decision process (MDP) based on a comparison of reward sequences. Due to this property they provide theoretical guarantees in value function approximation (VFA). In this work we first prove that bisimulation and $\\pi$-bisimulation metrics can be defined via a more general class of Sinkhorn distances, which unifies various state similarity metrics used in recent work. Then we describe an approximate policy iteration (API) procedure that uses a bisimulation-based discretization of the state space for VFA and prove asymptotic performance bounds. Next, we bound the difference between $\\pi$-bisimulation metrics in terms of the change in the policies themselves. Based on these results, we design an API($\\alpha$) procedure that employs conservative policy updates and enjoys better performance bounds than the naive API approach. We discuss how such API procedures map onto practical actor-critic methods that use bisimulation metrics for state representation learning. Lastly, we validate our theoretical results and investigate their practical implications via a controlled empirical analysis based on an implementation of bisimulation-based API for finite MDPs."
                    },
                    {
                        "title": "Trusted Approximate Policy Iteration with Bisimulation Metrics",
                        "abstract": "Bisimulation metrics de\ufb01ne a distance measure be-tween states of a Markov decision process (MDP) based on a comparison of reward sequences. Due to this property they provide theoretical guarantees in value function approximation. In this work we \ufb01rst prove that bisimulation metrics can be de\ufb01ned via any p -Wasserstein metric for p \u2265 1 . Then we describe an approximate policy iteration (API) procedure that uses (cid:15) -aggregation with \u03c0 - bisimulation and prove performance bounds for continuous state spaces. We bound the difference between \u03c0 -bisimulation metrics in terms of the change in the policies themselves. Based on these theoretical results, we design an API( \u03b1 ) procedure that employs conservative policy updates and enjoys better performance bounds than the naive API approach. In addition, we propose a novel trust region approach which circumvents the requirement to explicitly solve a constrained optimization problem. Finally, we provide experimental evidence of improved stability compared to non-conservative alternatives in simulated continuous control."
                    },
                    {
                        "title": "Towards Robust Bisimulation Metric Learning",
                        "abstract": "Learned representations in deep reinforcement learning (DRL) have to extract task-relevant information from complex observations, balancing between robustness to distraction and informativeness to the policy. Such stable and rich representations, often learned via modern function approximation techniques, can enable practical application of the policy improvement theorem, even in high-dimensional continuous state-action spaces. Bisimulation metrics offer one solution to this representation learning problem, by collapsing functionally similar states together in representation space, which promotes invariance to noise and distractors. In this work, we generalize value function approximation bounds for on-policy bisimulation metrics to non-optimal policies and approximate environment dynamics. Our theoretical results help us identify embedding pathologies that may occur in practical use. In particular, we find that these issues stem from an underconstrained dynamics model and an unstable dependence of the embedding norm on the reward signal in environments with sparse rewards. Further, we propose a set of practical remedies: (i) a norm constraint on the representation space, and (ii) an extension of prior approaches with intrinsic rewards and latent space regularization. Finally, we provide evidence that the resulting method is not only more robust to sparse reward functions, but also able to solve challenging continuous control tasks with observational distractions, where prior methods fail."
                    },
                    {
                        "title": "Dependency parsing with structure preserving embeddings",
                        "abstract": "Modern neural approaches to dependency parsing are trained to predict a tree structure by jointly learning a contextual representation for tokens in a sentence, as well as a head\u2013dependent scoring function. Whereas this strategy results in high performance, it is difficult to interpret these representations in relation to the geometry of the underlying tree structure. Our work seeks instead to learn interpretable representations by training a parser to explicitly preserve structural properties of a tree. We do so by casting dependency parsing as a tree embedding problem where we incorporate geometric properties of dependency trees in the form of training losses within a graph-based parser. We provide a thorough evaluation of these geometric losses, showing that a majority of them yield strong tree distance preservation as well as parsing performance on par with a competitive graph-based parser (Qi et al., 2018). Finally, we show where parsing errors lie in terms of tree relationship in order to guide future work."
                    },
                    {
                        "title": "RankMI: A Mutual Information Maximizing Ranking Loss",
                        "abstract": "We introduce an information-theoretic loss function, RankMI, and an associated training algorithm for deep representation learning for image retrieval. Our proposed framework consists of alternating updates to a network that estimates the divergence between distance distributions of matching and non-matching pairs of learned embeddings, and an embedding network that maximizes this estimate via sampled negatives. In addition, under this information-theoretic lens we draw connections between RankMI and commonly-used ranking losses, e.g., triplet loss. We extensively evaluate RankMI on several standard image retrieval datasets, namely, CUB-200-2011, CARS-196, and Stanford Online Products. Our method achieves competitive results or significant improvements over previous reported results on all datasets."
                    },
                    {
                        "title": "Dynamic Scheduling of MPI-based Distributed Deep Learning Training Jobs",
                        "abstract": "There is a general trend towards solving problems suited to deep learning with more complex deep learning architectures trained on larger training sets. This requires longer compute times and greater data parallelization or model parallelization. Both data and model parallelism have been historically faster in parameter server architectures, but data parallelism is starting to be faster in ring architectures due to algorithmic improvements. In this paper, we analyze the math behind ring architectures and make an informed adaptation of dynamic scheduling to ring architectures. To do so, we formulate a non-convex, non-linear, NP-hard integer programming problem and a new efficient doubling heuristic for its solution. We build upon Horovod: an open source ring architecture framework over TensorFlow. We show that Horovod jobs have a low cost to stop and restart and that stopping and restarting ring architecture jobs leads to faster completion times. These two facts make dynamic scheduling of ring architecture jobs feasible. Lastly, we simulate a scheduler using these runs and show a more than halving of average job time on some workload patterns."
                    }
                ]
            },
            "9957a80b-760c-4e22-bda7-c0e291ad422d": {
                "pk": "9957a80b-760c-4e22-bda7-c0e291ad422d",
                "name": "Mohammad Ghavamzadeh",
                "collaborators": [
                    "Craig Boutilier",
                    "Marek Petrik",
                    "Yinlam Chow",
                    "M. Ryu",
                    "J. Hau",
                    "Kimin Lee",
                    "Hao Liu",
                    "Olivia Watkins",
                    "Yuqing Du",
                    "P. Abbeel",
                    "E. Delage",
                    "Guy Tennenholtz",
                    "Chih-Wei Hsu",
                    "Azamat Tulepbergenov",
                    "Joey Hong",
                    "B. Kveton",
                    "S. Katariya",
                    "M. Zaheer",
                    "S. Gu",
                    "Christoph Dann",
                    "T. V. Marinov",
                    "Jincheng Mei",
                    "Bo Dai",
                    "Alekh Agarwal",
                    "Csaba Szepesv\u00e1ri",
                    "D. Schuurmans",
                    "Moloud Abdar",
                    "Meenakshi Kollati",
                    "Swaraja Kuraparthi",
                    "Farhad Pourpanah",
                    "Daniel J. McDuff",
                    "Shuicheng Yan",
                    "Abduallah A. Mohamed",
                    "A. Khosravi",
                    "E. Cambria",
                    "F. Porikli",
                    "Ying Fan",
                    "Kangwook Lee",
                    "Erdem Biyik",
                    "Fan Yao",
                    "Alex Haig",
                    "Jihwan Jeong",
                    "Dhawal Gupta",
                    "Gecia Bravo-Hermsdorff",
                    "R. Busa-Fekete",
                    "Andr\u00e9s Mu\u00f1oz Medina",
                    "Umar Syed",
                    "Ahmadreza Moradipari",
                    "Taha Rajabzadeh",
                    "Christos Thrampoulidis",
                    "M. Alizadeh",
                    "Javad Azizi",
                    "T. Duong",
                    "Yasin Abbasi-Yadkori",
                    "Andr'as Gyorgy",
                    "Claire Vernade",
                    "Amin Rakhsha",
                    "Andrew Wang",
                    "Amir-massoud Farahmand",
                    "Ofir Nachum"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Natural Language Processing",
                    "Generative Models",
                    "Bandit Learning"
                ],
                "publications": [
                    {
                        "title": "Aligning Text-to-Image Models using Human Feedback",
                        "abstract": "Deep generative models have shown impressive results in text-to-image synthesis. However, current text-to-image models often generate images that are inadequately aligned with text prompts. We propose a fine-tuning method for aligning such models using human feedback, comprising three stages. First, we collect human feedback assessing model output alignment from a set of diverse text prompts. We then use the human-labeled image-text dataset to train a reward function that predicts human feedback. Lastly, the text-to-image model is fine-tuned by maximizing reward-weighted likelihood to improve image-text alignment. Our method generates objects with specified colors, counts and backgrounds more accurately than the pre-trained model. We also analyze several design choices and find that careful investigations on such design choices are important in balancing the alignment-fidelity tradeoffs. Our results demonstrate the potential for learning from human feedback to significantly improve text-to-image models."
                    },
                    {
                        "title": "Multiple-policy High-confidence Policy Evaluation",
                        "abstract": "In reinforcement learning applications, we often want to accurately estimate the return of several policies of interest. We study this problem, multiple-policy high-con\ufb01dence policy evaluation, where the goal is to estimate the return of all given target policies up to a desired accuracy with as few samples as possible. The nat-ural approaches to this problem, i.e., evaluating each policy separately or estimating a model of the MDP, do not take into account the similarities between target policies and scale with the number of policies to evaluate or the size of the MDP, respectively. We present an alternative approach based on reusing samples from on-policy Monte-Carlo estimators and show that it is more sample-ef\ufb01cient in favorable cases. Speci\ufb01cally, we provide guarantees in terms of a notion of overlap of the set of target policies and shed light on when such an approach is indeed bene\ufb01cial compared to existing methods."
                    },
                    {
                        "title": "Ordering-based Conditions for Global Convergence of Policy Gradient Methods",
                        "abstract": "We prove that, for \ufb01nite-arm bandits with linear function approximation, the global convergence of policy gradient (PG) methods depends on inter-related properties between the policy update and the representation. First, we establish a few key observations that frame the study: (i) Global convergence can be achieved under linear function approximation without policy or reward realizability, both for the standard Softmax PG and natural policy gradient (NPG). (ii) Approximation error is not a key quantity for characterizing global convergence in either algorithm. (iii) The conditions on the representation that imply global convergence are different between these two algorithms. Overall, these observations call into question approximation error as an appropriate quantity for characterizing the global convergence of PG methods under linear function approximation. Second, motivated by these observations, we establish new general results: (i) NPG with linear function approximation achieves global convergence if and only if the projection of the reward onto the representable space preserves the optimal action\u2019s rank, a quantity that is not strongly related to approximation error. (ii) The global convergence of Softmax PG occurs if the representation satis\ufb01es a non-domination condition and can preserve the ranking of rewards, which goes well beyond policy or reward realizability. We provide experimental results to support these theoretical \ufb01ndings."
                    },
                    {
                        "title": "A Review of Deep Learning for Video Captioning",
                        "abstract": "Video captioning (VC) is a fast-moving, cross-disciplinary area of research that bridges work in the fields of computer vision, natural language processing (NLP), linguistics, and human-computer interaction. In essence, VC involves understanding a video and describing it with language. Captioning is used in a host of applications from creating more accessible interfaces (e.g., low-vision navigation) to video question answering (V-QA), video retrieval and content generation. This survey covers deep learning-based VC, including but, not limited to, attention-based architectures, graph networks, reinforcement learning, adversarial networks, dense video captioning (DVC), and more. We discuss the datasets and evaluation metrics used in the field, and limitations, applications, challenges, and future directions for VC."
                    },
                    {
                        "title": "On Dynamic Program Decompositions of Static Risk Measures",
                        "abstract": "Optimizing static risk-averse objectives in Markov decision processes is challenging because they do not readily admit dynamic programming decompositions. Prior work has proposed to use a dynamic decomposition of risk measures that help to formulate dynamic programs on an augmented state space. This paper shows that several existing decompositions are inherently inexact, contradicting several claims in the literature. In particular, we give examples that show that popular decompositions for CVaR and EVaR risk measures are strict overestimates of the true risk values. However, an exact decomposition is possible for VaR, and we give a simple proof that illustrates the fundamental di\ufb00erence between VaR and CVaR dynamic programming properties."
                    },
                    {
                        "title": "Entropic Risk Optimization in Discounted MDPs",
                        "abstract": "Risk-averse Markov Decision Processes (MDPs) have optimal policies that achieve high returns with low variability, but these MDPs are often difficult to solve. Only a few risk-averse objectives admit a dynamic programming (DP) formulation, which is the mainstay of most MDP and RL algorithms. We derive a new DP formulation for discounted risk-averse MDPs with En-tropic Risk Measure (ERM) and Entropic Value at Risk (EVaR) objectives. Our DP formulation for ERM, which is possible because of our novel definition of value function with time-dependent risk levels, can approximate optimal policies in a time that is polynomial in the approximation error. We then use the ERM algorithm to optimize the EVaR objective in polynomial time using an optimized discretization scheme. Our numerical results show the viability of our formulations and algorithms in discounted MDPs."
                    },
                    {
                        "title": "DPOK: Reinforcement Learning for Fine-tuning Text-to-Image Diffusion Models",
                        "abstract": "Learning from human feedback has been shown to improve text-to-image models. These techniques first learn a reward function that captures what humans care about in the task and then improve the models based on the learned reward function. Even though relatively simple approaches (e.g., rejection sampling based on reward scores) have been investigated, fine-tuning text-to-image models with the reward function remains challenging. In this work, we propose using online reinforcement learning (RL) to fine-tune text-to-image models. We focus on diffusion models, defining the fine-tuning task as an RL problem, and updating the pre-trained text-to-image diffusion models using policy gradient to maximize the feedback-trained reward. Our approach, coined DPOK, integrates policy optimization with KL regularization. We conduct an analysis of KL regularization for both RL fine-tuning and supervised fine-tuning. In our experiments, we show that DPOK is generally superior to supervised fine-tuning with respect to both image-text alignment and image quality. Our code is available at https://github.com/google-research/google-research/tree/master/dpok."
                    },
                    {
                        "title": "Bayesian Regret Minimization in Offline Bandits",
                        "abstract": "We study how to make decisions that minimize Bayesian regret in offline linear bandits. Prior work suggests that one must take actions with maximum lower confidence bound (LCB) on their reward. We argue that the reliance on LCB is inherently flawed in this setting and propose a new algorithm that directly minimizes upper bounds on the Bayesian regret using efficient conic optimization solvers. Our bounds build heavily on new connections to monetary risk measures. Proving a matching lower bound, we show that our upper bounds are tight, and by minimizing them we are guaranteed to outperform the LCB approach. Our numerical results on synthetic domains confirm that our approach is superior to LCB."
                    },
                    {
                        "title": "Preference Elicitation with Soft Attributes in Interactive Recommendation",
                        "abstract": "Preference elicitation plays a central role in interactive recommender systems. Most preference elicitation approaches use either item queries that ask users to select preferred items from a slate, or attribute queries that ask them to express their preferences for item characteristics. Unfortunately, users often wish to describe their preferences using soft attributes for which no ground-truth semantics is given. Leveraging concept activation vectors for soft attribute semantics, we develop novel preference elicitation methods that can accommodate soft attributes and bring together both item and attribute-based preference elicitation. Our techniques query users using both items and soft attributes to update the recommender system's belief about their preferences to improve recommendation quality. We demonstrate the effectiveness of our methods vis-a-vis competing approaches on both synthetic and real-world datasets."
                    },
                    {
                        "title": "Factual and Personalized Recommendations using Language Models and Reinforcement Learning",
                        "abstract": "Recommender systems (RSs) play a central role in connecting users to content, products, and services, matching candidate items to users based on their preferences. While traditional RSs rely on implicit user feedback signals, conversational RSs interact with users in natural language. In this work, we develop a comPelling, Precise, Personalized, Preference-relevant language model (P4LM) that recommends items to users while putting emphasis on explaining item characteristics and their relevance. P4LM uses the embedding space representation of a user's preferences to generate compelling responses that are factually-grounded and relevant w.r.t. the user's preferences. Moreover, we develop a joint reward function that measures precision, appeal, and personalization, which we use as AI-based feedback in a reinforcement learning-based language model framework. Using the MovieLens 25M dataset, we demonstrate that P4LM delivers compelling, personalized movie narratives to users."
                    },
                    {
                        "title": "Offline Reinforcement Learning for Mixture-of-Expert Dialogue Management",
                        "abstract": "Reinforcement learning (RL) has shown great promise for developing dialogue management (DM) agents that are non-myopic, conduct rich conversations, and maximize overall user satisfaction. Despite recent developments in RL and language models (LMs), using RL to power conversational chatbots remains challenging, in part because RL requires online exploration to learn effectively, whereas collecting novel human-bot interactions can be expensive and unsafe. This issue is exacerbated by the combinatorial action spaces facing these algorithms, as most LM agents generate responses at the word level. We develop a variety of RL algorithms, specialized to dialogue planning, that leverage recent Mixture-of-Expert Language Models (MoE-LMs) -- models that capture diverse semantics, generate utterances reflecting different intents, and are amenable for multi-turn DM. By exploiting MoE-LM structure, our methods significantly reduce the size of the action space and improve the efficacy of RL-based DM. We evaluate our methods in open-domain dialogue to demonstrate their effectiveness w.r.t.\\ the diversity of intent in generated utterances and overall DM performance."
                    },
                    {
                        "title": "Private and Communication-Efficient Algorithms for Entropy Estimation",
                        "abstract": "Modern statistical estimation is often performed in a distributed setting where each sample belongs to a single user who shares their data with a central server. Users are typically concerned with preserving the privacy of their samples, and also with minimizing the amount of data they must transmit to the server. We give improved private and communication-efficient algorithms for estimating several popular measures of the entropy of a distribution. All of our algorithms have constant communication cost and satisfy local differential privacy. For a joint distribution over many variables whose conditional independence is given by a tree, we describe algorithms for estimating Shannon entropy that require a number of samples that is linear in the number of variables, compared to the quadratic sample complexity of prior work. We also describe an algorithm for estimating Gini entropy whose sample complexity has no dependence on the support size of the distribution and can be implemented using a single round of concurrent communication between the users and the server. In contrast, the previously best-known algorithm has high communication cost and requires the server to facilitate interaction between the users. Finally, we describe an algorithm for estimating collision entropy that generalizes the best known algorithm to the private and communication-efficient setting."
                    },
                    {
                        "title": "On Dynamic Programming Decompositions of Static Risk Measures in Markov Decision Processes",
                        "abstract": "Optimizing static risk-averse objectives in Markov decision processes is difficult because they do not admit standard dynamic programming equations common in Reinforcement Learning (RL) algorithms. Dynamic programming decompositions that augment the state space with discrete risk levels have recently gained popularity in the RL community. Prior work has shown that these decompositions are optimal when the risk level is discretized sufficiently. However, we show that these popular decompositions for Conditional-Value-at-Risk (CVaR) and Entropic-Value-at-Risk (EVaR) are inherently suboptimal regardless of the discretization level. In particular, we show that a saddle point property assumed to hold in prior literature may be violated. However, a decomposition does hold for Value-at-Risk and our proof demonstrates how this risk measure differs from CVaR and EVaR. Our findings are significant because risk-averse algorithms are used in high-stake environments, making their correctness much more critical."
                    },
                    {
                        "title": "Multi-Task Off-Policy Learning from Bandit Feedback",
                        "abstract": "Many practical applications, such as recommender systems and learning to rank, involve solving multiple similar tasks. One example is learning of recommendation policies for users with similar movie preferences, where the users may still rank the individual movies slightly differently. Such tasks can be organized in a hierarchy, where similar tasks are related through a shared structure. In this work, we formulate this problem as a contextual off-policy optimization in a hierarchical graphical model from logged bandit feedback. To solve the problem, we propose a hierarchical off-policy optimization algorithm (HierOPO), which estimates the parameters of the hierarchical model and then acts pessimistically with respect to them. We instantiate HierOPO in linear Gaussian models, for which we also provide an efficient implementation and analysis. We prove per-task bounds on the suboptimality of the learned policies, which show a clear improvement over not using the hierarchical model. We also evaluate the policies empirically. Our theoretical and empirical results show a clear advantage of using the hierarchy over solving each task independently."
                    },
                    {
                        "title": "Multi-Environment Meta-Learning in Stochastic Linear Bandits",
                        "abstract": "In this work we investigate meta-learning (or learning-to-learn) approaches in multi-task linear stochastic bandit problems that can originate from multiple environments. Inspired by the work of [1] on meta-learning in a sequence of linear bandit problems whose parameters are sampled from a single distribution (i.e., a single environment), here we consider the feasibility of meta-learning when task parameters are drawn from a mixture distribution instead. For this problem, we propose a regularized version of the OFUL algorithm that, when trained on tasks with labeled environments, achieves low regret on a new task without requiring knowledge of the environment from which the new task originates. Specifically, our regret bound for the new algorithm captures the effect of environment misclassification and highlights the benefits over learning each task separately or meta-learning without recognition of the distinct mixture components."
                    },
                    {
                        "title": "Non-stationary Bandits and Meta-Learning with a Small Set of Optimal Arms",
                        "abstract": "We study a sequential decision problem where the learner faces a sequence of K -armed bandit tasks. The task boundaries might be known (the bandit meta-learning setting), or unknown (the non-stationary bandit setting). For a given integer M \u2264 K , the learner aims to compete with the best subset of arms of size M . We design an algorithm based on a reduction to bandit submodular maximization, and show that, for T rounds comprised of N tasks, in the regime of large number of tasks and small number of optimal arms M , its regret in both settings is smaller than the simple baseline of \u02dc O ( \u221a KNT ) that can be obtained by using standard algorithms designed for non-stationary bandit problems. For the bandit meta-learning problem with \ufb01xed task length \u03c4 , we show that the regret of the algorithm is bounded as \u02dc O ( NM \u221a M\u03c4 + N 2 / 3 M\u03c4 ). Under additional assumptions on the identi\ufb01ability of the optimal arms in each task, we show a bandit meta-learning algorithm with an improved \u02dc O ( N \u221a M\u03c4 + N 1 / 2 \u221a MK\u03c4 ) regret."
                    },
                    {
                        "title": "Operator Splitting Value Iteration",
                        "abstract": "We introduce new planning and reinforcement learning algorithms for discounted MDPs that utilize an approximate model of the environment to accelerate the convergence of the value function. Inspired by the splitting approach in numerical linear algebra, we introduce Operator Splitting Value Iteration (OS-VI) for both Policy Evaluation and Control problems. OS-VI achieves a much faster convergence rate when the model is accurate enough. We also introduce a sample-based version of the algorithm called OS-Dyna. Unlike the traditional Dyna architecture, OS-Dyna still converges to the correct value function in presence of model approximation error."
                    },
                    {
                        "title": "Deep Hierarchy in Bandits",
                        "abstract": "Mean rewards of actions are often correlated. The form of these correlations may be complex and unknown a priori, such as the preferences of a user for recommended products and their categories. To maximize statistical efficiency, it is important to leverage these correlations when learning. We formulate a bandit variant of this problem where the correlations of mean action rewards are represented by a hierarchical Bayesian model with latent variables. Since the hierarchy can have multiple layers, we call it deep. We propose a hierarchical Thompson sampling algorithm (HierTS) for this problem, and show how to implement it efficiently for Gaussian hierarchies. The efficient implementation is possible due to a novel exact hierarchical representation of the posterior, which itself is of independent interest. We use this exact posterior to analyze the Bayes regret of HierTS in Gaussian bandits. Our analysis reflects the structure of the problem, that the regret decreases with the prior width, and also shows that hierarchies reduce the regret by non-constant factors in the number of actions. We confirm these theoretical findings empirically, in both synthetic and real-world experiments."
                    },
                    {
                        "title": "A Mixture-of-Expert Approach to RL-based Dialogue Management",
                        "abstract": "Despite recent advancements in language models (LMs), their application to dialogue management (DM) problems and ability to carry on rich conversations remain a challenge. We use reinforcement learning (RL) to develop a dialogue agent that avoids being short-sighted (outputting generic utterances) and maximizes overall user satisfaction. Most existing RL approaches to DM train the agent at the word-level, and thus, have to deal with a combinatorially complex action space even for a medium-size vocabulary. As a result, they struggle to produce a successful and engaging dialogue even if they are warm-started with a pre-trained LM. To address this issue, we develop a RL-based DM using a novel mixture of expert language model (MoE-LM) that consists of (i) a LM capable of learning diverse semantics for conversation histories, (ii) a number of {\\em specialized} LMs (or experts) capable of generating utterances corresponding to a particular attribute or personality, and (iii) a RL-based DM that performs dialogue planning with the utterances generated by the experts. Our MoE approach provides greater flexibility to generate sensible utterances with different intents and allows RL to focus on conversational-level DM. We compare it with SOTA baselines on open-domain dialogues and demonstrate its effectiveness both in terms of the diversity and sensibility of the generated utterances and the overall DM performance."
                    }
                ]
            },
            "9e9cf559-9d08-425e-a011-55763c5fdc9b": {
                "pk": "9e9cf559-9d08-425e-a011-55763c5fdc9b",
                "name": "Amir-massoud Farahmand",
                "collaborators": [
                    "C. Voelcker",
                    "Igor Gilitschenski",
                    "Yangchen Pan",
                    "Marcel Hussing",
                    "Eric Eaton",
                    "Amin Rakhsha",
                    "A. Ma",
                    "Martha White",
                    "Tyler Kastner",
                    "Romina Abachi",
                    "M. Ghavamzadeh",
                    "Jincheng Mei",
                    "Hengshuai Yao",
                    "Jongmin Lee",
                    "Ernest K. Ryu",
                    "Mark Bedaywi",
                    "Arash Ahmadian",
                    "Philip Torr",
                    "Jindong Gu",
                    "Mete Kemertas",
                    "A. Jepson",
                    "Murat A. Erdogdu",
                    "Andrew Wang",
                    "Victor Liao",
                    "Animesh Garg",
                    "Mohsen Rohani",
                    "Jun Luo",
                    "Ehsan Imani",
                    "Rodrigo Toro Icarte",
                    "R. Valenzano",
                    "Toryn Q. Klassen",
                    "Phillip J. K. Christoffersen",
                    "Sheila A. McIlraith",
                    "Fartash Faghri"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Adversarial Training",
                    "Model-Based Learning",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Dissecting Deep RL with High Update Ratios: Combatting Value Overestimation and Divergence",
                        "abstract": "We show that deep reinforcement learning can maintain its ability to learn without resetting network parameters in settings where the number of gradient updates greatly exceeds the number of environment samples. Under such large update-to-data ratios, a recent study by Nikishin et al. (2022) suggested the emergence of a primacy bias , in which agents overfit early interactions and downplay later experience, impairing their ability to learn. In this work, we dissect the phenomena underlying the primacy bias. We inspect the early stages of training that ought to cause the failure to learn and find that a fundamental challenge is a long-standing acquaintance: value overestimation. Overinflated Q-values are found not only on out-of-distribution but also in-distribution data and can be traced to unseen action prediction propelled by optimizer momentum. We employ a simple unit-ball normalization that enables learning under large update ratios, show its efficacy on the widely used dm_control suite, and obtain strong performance on the challenging dog tasks, competitive with model-based approaches. Our results question, in parts, the prior explanation for sub-optimal learning due to overfitting on early data."
                    },
                    {
                        "title": "When does Self-Prediction help? Understanding Auxiliary Tasks in Reinforcement Learning",
                        "abstract": "We investigate the impact of auxiliary learning tasks such as observation reconstruction and latent self-prediction on the representation learning problem in reinforcement learning. We also study how they interact with distractions and observation functions in the MDP. We provide a theoretical analysis of the learning dynamics of observation reconstruction, latent self-prediction, and TD learning in the presence of distractions and observation functions under linear model assumptions. With this formalization, we are able to explain why latent-self prediction is a helpful \\emph{auxiliary task}, while observation reconstruction can provide more useful features when used in isolation. Our empirical analysis shows that the insights obtained from our learning dynamics framework predicts the behavior of these loss functions beyond the linear model assumption in non-linear neural networks. This reinforces the usefulness of the linear model framework not only for theoretical analysis, but also practical benefit for applied problems."
                    },
                    {
                        "title": "Deflated Dynamics Value Iteration",
                        "abstract": "The Value Iteration (VI) algorithm is an iterative procedure to compute the value function of a Markov decision process, and is the basis of many reinforcement learning (RL) algorithms as well. As the error convergence rate of VI as a function of iteration $k$ is $O(\\gamma^k)$, it is slow when the discount factor $\\gamma$ is close to $1$. To accelerate the computation of the value function, we propose Deflated Dynamics Value Iteration (DDVI). DDVI uses matrix splitting and matrix deflation techniques to effectively remove (deflate) the top $s$ dominant eigen-structure of the transition matrix $\\mathcal{P}^{\\pi}$. We prove that this leads to a $\\tilde{O}(\\gamma^k |\\lambda_{s+1}|^k)$ convergence rate, where $\\lambda_{s+1}$is $(s+1)$-th largest eigenvalue of the dynamics matrix. We then extend DDVI to the RL setting and present Deflated Dynamics Temporal Difference (DDTD) algorithm. We empirically show the effectiveness of the proposed algorithms."
                    },
                    {
                        "title": "Dissecting Deep RL with High Update Ratios: Combatting Value Divergence",
                        "abstract": "We show that deep reinforcement learning algorithms can retain their ability to learn without resetting network parameters in settings where the number of gradient updates greatly exceeds the number of environment samples by combatting value function divergence. Under large update-to-data ratios, a recent study by Nikishin et al. (2022) suggested the emergence of a primacy bias, in which agents overfit early interactions and downplay later experience, impairing their ability to learn. In this work, we investigate the phenomena leading to the primacy bias. We inspect the early stages of training that were conjectured to cause the failure to learn and find that one fundamental challenge is a long-standing acquaintance: value function divergence. Overinflated Q-values are found not only on out-of-distribution but also in-distribution data and can be linked to overestimation on unseen action prediction propelled by optimizer momentum. We employ a simple unit-ball normalization that enables learning under large update ratios, show its efficacy on the widely used dm_control suite, and obtain strong performance on the challenging dog tasks, competitive with model-based approaches. Our results question, in parts, the prior explanation for sub-optimal learning due to overfitting early data."
                    },
                    {
                        "title": "MAD-TD: Model-Augmented Data stabilizes High Update Ratio RL",
                        "abstract": "Building deep reinforcement learning (RL) agents that find a good policy with few samples has proven notoriously challenging. To achieve sample efficiency, recent work has explored updating neural networks with large numbers of gradient steps for every new sample. While such high update-to-data (UTD) ratios have shown strong empirical performance, they also introduce instability to the training process. Previous approaches need to rely on periodic neural network parameter resets to address this instability, but restarting the training process is infeasible in many real-world applications and requires tuning the resetting interval. In this paper, we focus on one of the core difficulties of stable training with limited samples: the inability of learned value functions to generalize to unobserved on-policy actions. We mitigate this issue directly by augmenting the off-policy RL training process with a small amount of data generated from a learned world model. Our method, Model-Augmented Data for Temporal Difference learning (MAD-TD) uses small amounts of generated data to stabilize high UTD training and achieve competitive performance on the most challenging tasks in the DeepMind control suite. Our experiments further highlight the importance of employing a good model to generate data, MAD-TD's ability to combat value overestimation, and its practical stability gains for continued learning."
                    },
                    {
                        "title": "PID Accelerated Temporal Difference Algorithms",
                        "abstract": "Long-horizon tasks, which have a large discount factor, pose a challenge for most conventional reinforcement learning (RL) algorithms. Algorithms such as Value Iteration and Temporal Difference (TD) learning have a slow convergence rate and become inefficient in these tasks. When the transition distributions are given, PID VI was recently introduced to accelerate the convergence of Value Iteration using ideas from control theory. Inspired by this, we introduce PID TD Learning and PID Q-Learning algorithms for the RL setting, in which only samples from the environment are available. We give a theoretical analysis of the convergence of PID TD Learning and its acceleration compared to the conventional TD Learning. We also introduce a method for adapting PID gains in the presence of noise and empirically verify its effectiveness."
                    },
                    {
                        "title": "$\\lambda$-models: Effective Decision-Aware Reinforcement Learning with Latent Models",
                        "abstract": "The idea of decision-aware model learning, that models should be accurate where it matters for decision-making, has gained prominence in model-based reinforcement learning. While promising theoretical results have been established, the empirical performance of algorithms leveraging a decision-aware loss has been lacking, especially in continuous control problems. In this paper, we present a study on the necessary components for decision-aware reinforcement learning models and we showcase design choices that enable well-performing algorithms. To this end, we provide a theoretical and empirical investigation into algorithmic ideas in the field. We highlight that empirical design decisions established in the MuZero line of works, most importantly the use of a latent model, are vital to achieving good performance for related algorithms. Furthermore, we show that the MuZero loss function is biased in stochastic environments and establish that this bias has practical consequences. Building on these findings, we present an overview of which decision-aware loss functions are best used in what empirical scenarios, providing actionable insights to practitioners in the field."
                    },
                    {
                        "title": "Improving Adversarial Transferability via Model Alignment",
                        "abstract": "Neural networks are susceptible to adversarial perturbations that are transferable across different models. In this paper, we introduce a novel model alignment technique aimed at improving a given source model's ability in generating transferable adversarial perturbations. During the alignment process, the parameters of the source model are fine-tuned to minimize an alignment loss. This loss measures the divergence in the predictions between the source model and another, independently trained model, referred to as the witness model. To understand the effect of model alignment, we conduct a geometric analysis of the resulting changes in the loss landscape. Extensive experiments on the ImageNet dataset, using a variety of model architectures, demonstrate that perturbations generated from aligned source models exhibit significantly higher transferability than those from the original source model."
                    },
                    {
                        "title": "Efficient and Accurate Optimal Transport with Mirror Descent and Conjugate Gradients",
                        "abstract": "We design a novel algorithm for optimal transport by drawing from the entropic optimal transport, mirror descent and conjugate gradients literatures. Our scalable and GPU parallelizable algorithm is able to compute the Wasserstein distance with extreme precision, reaching relative error rates of $10^{-8}$ without numerical stability issues. Empirically, the algorithm converges to high precision solutions more quickly in terms of wall-clock time than a variety of algorithms including log-domain stabilized Sinkhorn's Algorithm. We provide careful ablations with respect to algorithm and problem parameters, and present benchmarking over upsampled MNIST images, comparing to various recent algorithms over high-dimensional problems. The results suggest that our algorithm can be a useful addition to the practitioner's optimal transport toolkit."
                    },
                    {
                        "title": "Understanding the robustness difference between stochastic gradient descent and adaptive gradient methods",
                        "abstract": "Stochastic gradient descent (SGD) and adaptive gradient methods, such as Adam and RMSProp, have been widely used in training deep neural networks. We empirically show that while the difference between the standard generalization performance of models trained using these methods is small, those trained using SGD exhibit far greater robustness under input perturbations. Notably, our investigation demonstrates the presence of irrelevant frequencies in natural datasets, where alterations do not affect models' generalization performance. However, models trained with adaptive methods show sensitivity to these changes, suggesting that their use of irrelevant frequencies can lead to solutions sensitive to perturbations. To better understand this difference, we study the learning dynamics of gradient descent (GD) and sign gradient descent (signGD) on a synthetic dataset that mirrors natural signals. With a three-dimensional input space, the models optimized with GD and signGD have standard risks close to zero but vary in their adversarial risks. Our result shows that linear models' robustness to $\\ell_2$-norm bounded changes is inversely proportional to the model parameters' weight norm: a smaller weight norm implies better robustness. In the context of deep learning, our experiments show that SGD-trained neural networks have smaller Lipschitz constants, explaining the better robustness to input perturbations than those trained with adaptive gradient methods."
                    },
                    {
                        "title": "Distributional Model Equivalence for Risk-Sensitive Reinforcement Learning",
                        "abstract": "We consider the problem of learning models for risk-sensitive reinforcement learning. We theoretically demonstrate that proper value equivalence, a method of learning models which can be used to plan optimally in the risk-neutral setting, is not sufficient to plan optimally in the risk-sensitive setting. We leverage distributional reinforcement learning to introduce two new notions of model equivalence, one which is general and can be used to plan for any risk measure, but is intractable; and a practical variation which allows one to choose which risk measures they may plan optimally for. We demonstrate how our framework can be used to augment any model-free risk-sensitive algorithm, and provide both tabular and large-scale experiments to demonstrate its ability."
                    },
                    {
                        "title": "Operator Splitting Value Iteration",
                        "abstract": "We introduce new planning and reinforcement learning algorithms for discounted MDPs that utilize an approximate model of the environment to accelerate the convergence of the value function. Inspired by the splitting approach in numerical linear algebra, we introduce Operator Splitting Value Iteration (OS-VI) for both Policy Evaluation and Control problems. OS-VI achieves a much faster convergence rate when the model is accurate enough. We also introduce a sample-based version of the algorithm called OS-Dyna. Unlike the traditional Dyna architecture, OS-Dyna still converges to the correct value function in presence of model approximation error."
                    },
                    {
                        "title": "Value Gradient weighted Model-Based Reinforcement Learning",
                        "abstract": "Model-based reinforcement learning (MBRL) is a sample efficient technique to obtain control policies, yet unavoidable modeling errors often lead performance deterioration. The model in MBRL is often solely fitted to reconstruct dynamics, state observations in particular, while the impact of model error on the policy is not captured by the training objective. This leads to a mismatch between the intended goal of MBRL, enabling good policy and value learning, and the target of the loss function employed in practice, future state prediction. Naive intuition would suggest that value-aware model learning would fix this problem and, indeed, several solutions to this objective mismatch problem have been proposed based on theoretical analysis. However, they tend to be inferior in practice to commonly used maximum likelihood (MLE) based approaches. In this paper we propose the Value-gradient weighted Model Learning (VaGraM), a novel method for value-aware model learning which improves the performance of MBRL in challenging settings, such as small model capacity and the presence of distracting state dimensions. We analyze both MLE and value-aware approaches and demonstrate how they fail to account for exploration and the behavior of function approximation when learning value-aware models and highlight the additional goals that must be met to stabilize optimization in the deep learning setting. We verify our analysis by showing that our loss function is able to achieve high returns on the Mujoco benchmark suite while being more robust than maximum likelihood based approaches."
                    },
                    {
                        "title": "Policy-Aware Model Learning for Policy Gradient Methods",
                        "abstract": "This paper considers the problem of learning a model in model-based reinforcement learning (MBRL). We examine how the planning module of an MBRL algorithm uses the model, and propose that the model learning module should incorporate the way the planner is going to use the model. This is in contrast to conventional model learning approaches, such as those based on maximum likelihood estimate, that learn a predictive model of the environment without explicitly considering the interaction of the model and the planner. We focus on policy gradient type of planning algorithms and derive new loss functions for model learning that incorporate how the planner uses the model. We call this approach Policy-Aware Model Learning (PAML). We theoretically analyze a generic model-based policy gradient algorithm and provide a convergence guarantee for the optimized policy. We also empirically evaluate PAML on some benchmark problems, showing promising results."
                    },
                    {
                        "title": "Understanding and mitigating the limitations of prioritized experience replay",
                        "abstract": "Prioritized Experience Replay (ER) has been empirically shown to improve sample efficiency across many domains and attracted great attention; however, there is little theoretical understanding of why such prioritized sampling helps and its limitations. In this work, we take a deep look at the prioritized ER. In a supervised learning setting, we show the equivalence between the error-based prioritized sampling method for mean squared error and uniform sampling for cubic power loss. We then provide theoretical insight into why it improves convergence rate upon uniform sampling during early learning. Based on the insight, we further point out two limitations of the prioritized ER method: 1) outdated priorities and 2) insufficient coverage of the sample space. To mitigate the limitations, we propose our model-based stochastic gradient Langevin dynamics sampling method. We show that our method does provide states distributed close to an ideal prioritized sampling distribution estimated by the brute-force method, which does not suffer from the two limitations. We conduct experiments on both discrete and continuous control problems to show our approach's efficacy and examine the practical implication of our method in an autonomous driving application."
                    },
                    {
                        "title": "Beyond Prioritized Replay: Sampling States in Model-Based RL via Simulated Priorities",
                        "abstract": "Model-based reinforcement learning (MBRL) can significantly improve sample efficiency, particularly when carefully choosing the states from which to sample hypothetical transitions. Such prioritization has been empirically shown to be useful for both experience replay (ER) and Dyna-style planning. However, there is as yet little theoretical understanding in RL about such prioritization strategies, and why they help. In this work, we revisit prioritized ER and, in an ideal setting, show an equivalence to minimizing cubic loss, providing theoretical insight into why it improves upon uniform sampling. This ideal setting, however, cannot be realized in practice, due to insufficient coverage of the sample space and outdated priorities of training samples. This motivates our model-based approach, which does not suffer from these limitations. Our key idea is to actively search for high priority states using gradient ascent. Under certain conditions, we prove that the distribution of hypothetical experiences generated from these states provides a diverse set of states, sampled proportionally to approximately true priorities. Our experiments on both benchmark and application-oriented domain show that our approach achieves superior performance over both the model-free prioritized ER method and several closely related model-based baselines."
                    },
                    {
                        "title": "An implicit function learning approach for parametric modal regression",
                        "abstract": "For multi-valued functions---such as when the conditional distribution on targets given the inputs is multi-modal---standard regression approaches are not always desirable because they provide the conditional mean. Modal regression algorithms address this issue by instead finding the conditional mode(s). Most, however, are nonparametric approaches and so can be difficult to scale. Further, parametric approximators, like neural networks, facilitate learning complex relationships between inputs and targets. In this work, we propose a parametric modal regression algorithm. We use the implicit function theorem to develop an objective, for learning a joint function over inputs and targets. We empirically demonstrate on several synthetic problems that our method (i) can learn multi-valued functions and produce the conditional modes, (ii) scales well to high-dimensional inputs, and (iii) can even be more effective for certain uni-modal problems, particularly for high-frequency functions. We demonstrate that our method is competitive in a real-world modal regression problem and two regular regression datasets."
                    },
                    {
                        "title": "The act of remembering: a study in partially observable reinforcement learning",
                        "abstract": "Reinforcement Learning (RL) agents typically learn memoryless policies---policies that only consider the last observation when selecting actions. Learning memoryless policies is efficient and optimal in fully observable environments. However, some form of memory is necessary when RL agents are faced with partial observability. In this paper, we study a lightweight approach to tackle partial observability in RL. We provide the agent with an external memory and additional actions to control what, if anything, is written to the memory. At every step, the current memory state is part of the agent's observation, and the agent selects a tuple of actions: one action that modifies the environment and another that modifies the memory. When the external memory is sufficiently expressive, optimal memoryless policies yield globally optimal solutions. Unfortunately, previous attempts to use external memory in the form of binary memory have produced poor results in practice. Here, we investigate alternative forms of memory in support of learning effective memoryless policies. Our novel forms of memory outperform binary and LSTM-based memory in well-established partially observable domains."
                    },
                    {
                        "title": "Adversarial Robustness through Regularization: A Second-Order Approach",
                        "abstract": "Adversarial training is a common approach to improving the robustness of deep neural networks against adversarial examples. In this work, we propose a novel regularization approach as an alternative. To derive the regularizer, we formulate the adversarial robustness problem under the robust optimization framework and approximate the loss function using a second-order Taylor series expansion. Our proposed second-order adversarial regularizer (SOAR) is an upper bound based on the Taylor approximation of the inner-max in the robust optimization objective. We empirically show that the proposed method improves the robustness of networks on the CIFAR-10 dataset."
                    }
                ]
            }
        }
    },
    "2310.18882": {
        "paper_data": {
            "title": "Differentiable Learning of Generalized Structured Matrices for Efficient Deep Neural Networks",
            "url": "http://arxiv.org/abs/2310.18882v2",
            "arxiv_id": "2310.18882",
            "authors": [
                "Changwoo Lee",
                "Hun-Seok Kim"
            ],
            "abstract": "This paper investigates efficient deep neural networks (DNNs) to replace dense unstructured weight matrices with structured ones that possess desired properties. The challenge arises because the optimal weight matrix structure in popular neural network models is obscure in most cases and may vary from layer to layer even in the same network. Prior structured matrices proposed for efficient DNNs were mostly hand-crafted without a generalized framework to systematically learn them. To address this issue, we propose a generalized and differentiable framework to learn efficient structures of weight matrices by gradient descent. We first define a new class of structured matrices that covers a wide range of structured matrices in the literature by adjusting the structural parameters. Then, the frequency-domain differentiable parameterization scheme based on the Gaussian-Dirichlet kernel is adopted to learn the structural parameters by proximal gradient descent. On the image and language tasks, our method learns efficient DNNs with structured matrices, achieving lower complexity and/or higher performance than prior approaches that employ low-rank, block-sparse, or block-low-rank matrices.",
            "introduction": "   1 Introduction  Deep Neural Networks (DNNs) for large language models (LLMs) (Vaswani et\u00a0al., 2017; Devlin et\u00a0al., 2018; Radford et\u00a0al., 2019; Brown et\u00a0al., 2020) and vision tasks (Dosovitskiy et\u00a0al., 2020; Touvron et\u00a0al., 2021) have shown great success in various domains in recent years. The size of the DNNs, however, has increased on an extraordinary scale \u2013 up to 70 Billion parameters in the single model Zhao et\u00a0al. (2023) \u2013 requiring an unprecedented amount of computing resources and energy to deploy the DNN-based services. Fortunately, under certain conditions, the weight matrices of DNNs trained by stochastic gradient descent (SGD) naturally have well-defined preferred structures, such as low-rank matrices (Yaras et\u00a0al., 2023; Huh et\u00a0al., 2022; Gunasekar et\u00a0al., 2017). However, identifying such structures to lower the effective complexity of weight matrices in recent models such as Transformers Vaswani et\u00a0al. (2017) remains a challenging problem. Also, it mostly relies on the existing human-designed / hand-crafted structured matrices without a unified systematic approach. Hence prior works have focused on investigating new classes of structured matrices for DNNs (Li et\u00a0al., 2015; Dao et\u00a0al., 2022; Chen et\u00a0al., 2022). Notably, each structured matrix is defined disjointedly from other formats. For example, neither the block-sparse matrix format nor the low-rank matrix format of the same number of parameters is a subset of another, and yet there is no unified representation that describes both well. Moreover, the structure description and the implementation complexity of the structured matrix are in non-differentiable discrete spaces.   In this paper, we investigate (locally) optimal structures of the weight matrices as well as a differentiable training method to learn them, attempting to answer the following two questions:   1.  Is there a universal format that represents a wide range of structured matrices?    2.  Can the structure of such matrices be learned efficiently, if it exists?      Contributions. Tackling the above two questions, we introduce a generalized and differentiable structured matrix format. The main contributions of this work can be summarized as follows.   1) We propose a Generalized Block-low-rank (GBLR) matrix format, which includes many important structures such as Low-Rank (LR), Block Sparse (BSP), and Block-low-rank (BLR) matrices under some practical conditions. The new structured matrix format consists of two types of parameters: one guiding the structure of the matrix and the other specifying the content or the values of matrix elements. We show that the LR, BSP and BLR formats are special cases of the GBLR matrix. We also show that the GBLR format is closed under the interpolation between existing GBLR matrices in the structural parameter space, which we believe is a strong evidence that the GBLR format is able to capture undiscovered structured matrix formats.   2) We introduce a differentiable parameterization of the structural parameters \u2013 widths and locations of blocks \u2013 of the GBLR format. The structural parameters are defined in the frequency domain, and are processed by the proposed Gaussian-Dirichlet (Gaudi) function followed by inverse Fast Fourier Transform (IFFT) to the format named Gaudi-GBLR. We show that the derivatives of the Gaudi function with respect to the structural parameters exist almost everywhere, even when the width is zero.   3) We propose a practical learning algorithm based on the proximal gradient descent to train compact neural networks with Gaudi-GBLR",
            "references": [
                {
                    "title": "The Law of Parsimony in Gradient Descent for Learning Deep Linear Networks",
                    "abstract": "Over the past few years, an extensively studied phenomenon in training deep networks is the implicit bias of gradient descent towards parsimonious solutions. In this work, we investigate this phenomenon by narrowing our focus to deep linear networks. Through our analysis, we reveal a surprising\"law of parsimony\"in the learning dynamics when the data possesses low-dimensional structures. Specifically, we show that the evolution of gradient descent starting from orthogonal initialization only affects a minimal portion of singular vector spaces across all weight matrices. In other words, the learning process happens only within a small invariant subspace of each weight matrix, despite the fact that all weight parameters are updated throughout training. This simplicity in learning dynamics could have significant implications for both efficient training and a better understanding of deep networks. First, the analysis enables us to considerably improve training efficiency by taking advantage of the low-dimensional structure in learning dynamics. We can construct smaller, equivalent deep linear networks without sacrificing the benefits associated with the wider counterparts. Second, it allows us to better understand deep representation learning by elucidating the linear progressive separation and concentration of representations from shallow to deep layers. We also conduct numerical experiments to support our theoretical results. The code for our experiments can be found at https://github.com/cjyaras/lawofparsimony."
                },
                {
                    "title": "A Survey of Large Language Models",
                    "abstract": "Language is essentially a complex, intricate system of human expressions governed by grammatical rules. It poses a significant challenge to develop capable AI algorithms for comprehending and grasping a language. As a major approach, language modeling has been widely studied for language understanding and generation in the past two decades, evolving from statistical language models to neural language models. Recently, pre-trained language models (PLMs) have been proposed by pre-training Transformer models over large-scale corpora, showing strong capabilities in solving various NLP tasks. Since researchers have found that model scaling can lead to performance improvement, they further study the scaling effect by increasing the model size to an even larger size. Interestingly, when the parameter scale exceeds a certain level, these enlarged language models not only achieve a significant performance improvement but also show some special abilities that are not present in small-scale language models. To discriminate the difference in parameter scale, the research community has coined the term large language models (LLM) for the PLMs of significant size. Recently, the research on LLMs has been largely advanced by both academia and industry, and a remarkable progress is the launch of ChatGPT, which has attracted widespread attention from society. The technical evolution of LLMs has been making an important impact on the entire AI community, which would revolutionize the way how we develop and use AI algorithms. In this survey, we review the recent advances of LLMs by introducing the background, key findings, and mainstream techniques. In particular, we focus on four major aspects of LLMs, namely pre-training, adaptation tuning, utilization, and capacity evaluation. Besides, we also summarize the available resources for developing LLMs and discuss the remaining issues for future directions."
                },
                {
                    "title": "Language model compression with weighted low-rank factorization",
                    "abstract": "Factorizing a large matrix into small matrices is a popular strategy for model compression. Singular value decomposition (SVD) plays a vital role in this compression strategy, approximating a learned matrix with fewer parameters. However, SVD minimizes the squared error toward reconstructing the original matrix without gauging the importance of the parameters, potentially giving a larger reconstruction error for those who affect the task accuracy more. In other words, the optimization objective of SVD is not aligned with the trained model's task accuracy. We analyze this previously unexplored problem, make observations, and address it by introducing Fisher information to weigh the importance of parameters affecting the model prediction. This idea leads to our method: Fisher-Weighted SVD (FWSVD). Although the factorized matrices from our approach do not result in smaller reconstruction errors, we find that our resulting task accuracy is much closer to the original model's performance. We perform analysis with the transformer-based language models, showing our weighted SVD largely alleviates the mismatched optimization objectives and can maintain model performance with a higher compression rate. Our method can directly compress a task-specific model while achieving better performance than other compact model strategies requiring expensive model pre-training. Moreover, the evaluation of compressing an already compact model shows our method can further reduce 9% to 30% parameters with an insignificant impact on task accuracy."
                },
                {
                    "title": "Monarch: Expressive Structured Matrices for Efficient and Accurate Training",
                    "abstract": "Large neural networks excel in many domains, but they are expensive to train and fine-tune. A popular approach to reduce their compute or memory requirements is to replace dense weight matrices with structured ones (e.g., sparse, low-rank, Fourier transform). These methods have not seen widespread adoption (1) in end-to-end training due to unfavorable efficiency--quality tradeoffs, and (2) in dense-to-sparse fine-tuning due to lack of tractable algorithms to approximate a given dense weight matrix. To address these issues, we propose a class of matrices (Monarch) that is hardware-efficient (they are parameterized as products of two block-diagonal matrices for better hardware utilization) and expressive (they can represent many commonly used transforms). Surprisingly, the problem of approximating a dense weight matrix with a Monarch matrix, though nonconvex, has an analytical optimal solution. These properties of Monarch matrices unlock new ways to train and fine-tune sparse and dense models. We empirically validate that Monarch can achieve favorable accuracy-efficiency tradeoffs in several end-to-end sparse training applications: speeding up ViT and GPT-2 training on ImageNet classification and Wikitext-103 language modeling by 2x with comparable model quality, and reducing the error on PDE solving and MRI reconstruction tasks by 40%. In sparse-to-dense training, with a simple technique called\"reverse sparsification,\"Monarch matrices serve as a useful intermediate representation to speed up GPT-2 pretraining on OpenWebText by 2x without quality drop. The same technique brings 23% faster BERT pretraining than even the very optimized implementation from Nvidia that set the MLPerf 1.1 record. In dense-to-sparse fine-tuning, as a proof-of-concept, our Monarch approximation algorithm speeds up BERT fine-tuning on GLUE by 1.7x with comparable accuracy."
                },
                {
                    "title": "Learning strides in convolutional neural networks",
                    "abstract": "Convolutional neural networks typically contain several downsampling operators, such as strided convolutions or pooling layers, that progressively reduce the resolution of intermediate representations. This provides some shift-invariance while reducing the computational complexity of the whole architecture. A critical hyperparameter of such layers is their stride: the integer factor of downsampling. As strides are not differentiable, finding the best configuration either requires cross-validation or discrete optimization (e.g. architecture search), which rapidly become prohibitive as the search space grows exponentially with the number of downsampling layers. Hence, exploring this search space by gradient descent would allow finding better configurations at a lower computational cost. This work introduces DiffStride, the first downsampling layer with learnable strides. Our layer learns the size of a cropping mask in the Fourier domain, that effectively performs resizing in a differentiable way. Experiments on audio and image classification show the generality and effectiveness of our solution: we use DiffStride as a drop-in replacement to standard downsampling layers and outperform them. In particular, we show that introducing our layer into a ResNet-18 architecture allows keeping consistent high performance on CIFAR10, CIFAR100 and ImageNet even when training starts from poor random stride configurations. Moreover, formulating strides as learnable variables allows us to introduce a regularization term that controls the computational complexity of the architecture. We show how this regularization allows trading off accuracy for efficiency on ImageNet."
                },
                {
                    "title": "Pixelated Butterfly: Simple and Efficient Sparse training for Neural Network Models",
                    "abstract": "Overparameterized neural networks generalize well but are expensive to train. Ideally, one would like to reduce their computational cost while retaining their generalization benefits. Sparse model training is a simple and promising approach to achieve this, but there remain challenges as existing methods struggle with accuracy loss, slow training runtime, or difficulty in sparsifying all model components. The core problem is that searching for a sparsity mask over a discrete set of sparse matrices is difficult and expensive. To address this, our main insight is to optimize over a continuous superset of sparse matrices with a fixed structure known as products of butterfly matrices. As butterfly matrices are not hardware efficient, we propose simple variants of butterfly (block and flat) to take advantage of modern hardware. Our method (Pixelated Butterfly) uses a simple fixed sparsity pattern based on flat block butterfly and low-rank matrices to sparsify most network layers (e.g., attention, MLP). We empirically validate that Pixelated Butterfly is 3x faster than butterfly and speeds up training to achieve favorable accuracy--efficiency tradeoffs. On the ImageNet classification and WikiText-103 language modeling tasks, our sparse models train up to 2.5x faster than the dense MLP-Mixer, Vision Transformer, and GPT-2 medium with no drop in accuracy."
                },
                {
                    "title": "Compressing Neural Networks: Towards Determining the Optimal Layer-wise Decomposition",
                    "abstract": "We present a novel global compression framework for deep neural networks that automatically analyzes each layer to identify the optimal per-layer compression ratio, while simultaneously achieving the desired overall compression. Our algorithm hinges on the idea of compressing each convolutional (or fully-connected) layer by slicing its channels into multiple groups and decomposing each group via low-rank decomposition. At the core of our algorithm is the derivation of layer-wise error bounds from the Eckart Young Mirsky theorem. We then leverage these bounds to frame the compression problem as an optimization problem where we wish to minimize the maximum compression error across layers and propose an efficient algorithm towards a solution. Our experiments indicate that our method outperforms existing low-rank compression approaches across a wide range of networks and data sets. We believe that our results open up new avenues for future research into the global performance-size trade-offs of modern neural networks. Our code is available at https://github.com/lucaslie/torchprune."
                },
                {
                    "title": "MLP-Mixer: An all-MLP Architecture for Vision",
                    "abstract": "Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e.\"mixing\"the per-location features), and one with MLPs applied across patches (i.e.\"mixing\"spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers."
                },
                {
                    "title": "The Low-Rank Simplicity Bias in Deep Networks",
                    "abstract": "Modern deep neural networks are highly over-parameterized compared to the data on which they are trained, yet they often generalize remarkably well. A flurry of recent work has asked: why do deep networks not overfit to their training data? In this work, we make a series of empirical observations that investigate and extend the hypothesis that deeper networks are inductively biased to find solutions with lower effective rank embeddings. We conjecture that this bias exists because the volume of functions that maps to low effective rank embedding increases with depth. We show empirically that our claim holds true on finite width linear and non-linear models on practical learning paradigms and show that on natural data, these are often the solutions that generalize well. We then show that the simplicity bias exists at both initialization and after training and is resilient to hyper-parameters and learning methods. We further demonstrate how linear over-parameterization of deep non-linear models can be used to induce low-rank bias, improving generalization performance on CIFAR and ImageNet without changing the modeling capacity."
                },
                {
                    "title": "Training data-efficient image transformers & distillation through attention",
                    "abstract": "Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption. In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data. More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Low-Rank Compression of Neural Nets: Learning the Rank of Each Layer",
                    "abstract": "Neural net compression can be achieved by approximating each layer's weight matrix by a low-rank matrix. The real difficulty in doing this is not in training the resulting neural net (made up of one low-rank matrix per layer), but in determining what the optimal rank of each layer is\u2014effectively, an architecture search problem with one hyperparameter per layer. We show that, with a suitable formulation, this problem is amenable to a mixed discrete-continuous optimization jointly over the ranks and over the matrix elements, and give a corresponding algorithm. We show that this indeed can select ranks much better than existing approaches, making low-rank compression much more attractive than previously thought. For example, we can make a VGG network faster than a ResNet and with nearly the same classification error."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Movement Pruning: Adaptive Sparsity by Fine-Tuning",
                    "abstract": "Magnitude pruning is a widely used strategy for reducing model size in pure supervised learning; however, it is less effective in the transfer learning regime that has become standard for state-of-the-art natural language processing applications. We propose the use of movement pruning, a simple, deterministic first-order weight pruning method that is more adaptive to pretrained model fine-tuning. We give mathematical foundations to the method and compare it to existing zeroth- and first-order pruning methods. Experiments show that when pruning large pretrained language models, movement pruning shows significant improvements in high-sparsity regimes. When combined with distillation, the approach achieves minimal accuracy loss with down to only 3% of the model parameters."
                },
                {
                    "title": "Learning Fast Algorithms for Linear Transforms Using Butterfly Factorizations",
                    "abstract": "Fast linear transforms are ubiquitous in machine learning, including the discrete Fourier transform, discrete cosine transform, and other structured transformations such as convolutions. All of these transforms can be represented by dense matrix-vector multiplication, yet each has a specialized and highly efficient (subquadratic) algorithm. We ask to what extent hand-crafting these algorithms and implementations is necessary, what structural priors they encode, and how much knowledge is required to automatically learn a fast algorithm for a provided structured transform. Motivated by a characterization of fast matrix-vector multiplication as products of sparse matrices, we introduce a parameterization of divide-and-conquer methods that is capable of representing a large class of transforms. This generic formulation can automatically learn an efficient algorithm for many important transforms; for example, it recovers the O(N log N) Cooley-Tukey FFT algorithm to machine precision, for dimensions N up to 1024. Furthermore, our method can be incorporated as a lightweight replacement of generic matrices in machine learning pipelines to learn efficient and compressible transformations. On a standard task of compressing a single hidden-layer network, our method exceeds the classification accuracy of unconstrained matrices on CIFAR-10 by 3.9 points-the first time a structured approach has done so-with 4\u00d7 faster inference speed and 40\u00d7 fewer parameters."
                },
                {
                    "title": "Improving the Complexity of Block Low-Rank Factorizations with Fast Matrix Arithmetic",
                    "abstract": "We consider the LU factorization of an $n\\times n$ matrix $A$ represented as a block low-rank (BLR) matrix: most of its off-diagonal blocks are approximated by matrices of small rank $r$, which reduces the asymptotic complexity of computing the LU factorization of A down to $O(n^2r)$. In this article, our aim is to further reduce this complexity by exploiting fast matrix arithmetic, that is, the ability to multiply two $n\\times n$ full-rank matrices together for $O(n^\\omega)$ flops, where $\\omega<3$. This is not straightforward: simply accelerating the intermediate operations performed in the standard BLR factorization algorithm does not suffice to reduce the quadratic complexity in $n$, because these operations are performed on matrices whose size is too small. To overcome this obstacle, we devise a new BLR factorization algorithm that, by recasting the operations so as to work on intermediate matrices of larger size, can exploit more efficiently fast matrix arithmetic. This new algorithm achieves an asymptotic complexity of $O(n^{(\\omega+1)/2} r^{(\\omega-1)/2})$, which represents an asymptotic improvement compared to the standard BLR factorization as soon as $\\omega<3$. In particular, for Strassen's algorithm, $\\omega\\approx 2.81$ yields a complexity of $O(n^{1.904} r^{0.904})$. Our numerical experiments are in good agreement with this analysis."
                },
                {
                    "title": "DARTS: Differentiable Architecture Search",
                    "abstract": "This paper addresses the scalability challenge of architecture search by formulating the task in a differentiable manner. Unlike conventional approaches of applying evolution or reinforcement learning over a discrete and non-differentiable search space, our method is based on the continuous relaxation of the architecture representation, allowing efficient search of the architecture using gradient descent. Extensive experiments on CIFAR-10, ImageNet, Penn Treebank and WikiText-2 show that our algorithm excels in discovering high-performance convolutional architectures for image classification and recurrent architectures for language modeling, while being orders of magnitude faster than state-of-the-art non-differentiable techniques. Our implementation has been made publicly available to facilitate further research on efficient architecture search algorithms."
                },
                {
                    "title": "Efficient Neural Architecture Search via Parameter Sharing",
                    "abstract": "We propose Efficient Neural Architecture Search (ENAS), a fast and inexpensive approach for automatic model design. In ENAS, a controller learns to discover neural network architectures by searching for an optimal subgraph within a large computational graph. The controller is trained with policy gradient to select a subgraph that maximizes the expected reward on the validation set. Meanwhile the model corresponding to the selected subgraph is trained to minimize a canonical cross entropy loss. Thanks to parameter sharing between child models, ENAS is fast: it delivers strong empirical performances using much fewer GPU-hours than all existing automatic model design approaches, and notably, 1000x less expensive than standard Neural Architecture Search. On the Penn Treebank dataset, ENAS discovers a novel architecture that achieves a test perplexity of 55.8, establishing a new state-of-the-art among all methods without post-training processing. On the CIFAR-10 dataset, ENAS designs novel architectures that achieve a test error of 2.89%, which is on par with NASNet (Zoph et al., 2018), whose test error is 2.65%."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Implicit Regularization in Matrix Factorization",
                    "abstract": "We study implicit regularization when optimizing an underdetermined quadratic objective over a matrix $X$ with gradient descent on a factorization of X. We conjecture and provide empirical and theoretical evidence that with small enough step sizes and initialization close enough to the origin, gradient descent on a full dimensional factorization converges to the minimum nuclear norm solution."
                },
                {
                    "title": "Neural Architecture Search with Reinforcement Learning",
                    "abstract": "Neural networks are powerful and flexible models that work well for many difficult learning tasks in image, speech and natural language understanding. Despite their success, neural networks are still hard to design. In this paper, we use a recurrent network to generate the model descriptions of neural networks and train this RNN with reinforcement learning to maximize the expected accuracy of the generated architectures on a validation set. On the CIFAR-10 dataset, our method, starting from scratch, can design a novel network architecture that rivals the best human-invented architecture in terms of test set accuracy. Our CIFAR-10 model achieves a test error rate of 3.65, which is 0.09 percent better and 1.05x faster than the previous state-of-the-art model that used a similar architectural scheme. On the Penn Treebank dataset, our model can compose a novel recurrent cell that outperforms the widely-used LSTM cell, and other state-of-the-art baselines. Our cell achieves a test set perplexity of 62.4 on the Penn Treebank, which is 3.6 perplexity better than the previous state-of-the-art model. The cell can also be transferred to the character language modeling task on PTB and achieves a state-of-the-art perplexity of 1.214."
                },
                {
                    "title": "Categorical Reparameterization with Gumbel-Softmax",
                    "abstract": "Categorical variables are a natural choice for representing discrete structure in the world. However, stochastic neural networks rarely use categorical latent variables due to the inability to backpropagate through samples. In this work, we present an efficient gradient estimator that replaces the non-differentiable sample from a categorical distribution with a differentiable sample from a novel Gumbel-Softmax distribution. This distribution has the essential property that it can be smoothly annealed into a categorical distribution. We show that our Gumbel-Softmax estimator outperforms state-of-the-art gradient estimators on structured output prediction and unsupervised generative modeling tasks with categorical latent variables, and enables large speedups on semi-supervised classification."
                },
                {
                    "title": "The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables",
                    "abstract": "The reparameterization trick enables optimizing large scale stochastic computation graphs via gradient descent. The essence of the trick is to refactor each stochastic node into a differentiable function of its parameters and a random variable with fixed distribution. After refactoring, the gradients of the loss propagated by the chain rule through the graph are low variance unbiased estimators of the gradients of the expected loss. While many continuous random variables have such reparameterizations, discrete random variables lack useful reparameterizations due to the discontinuous nature of discrete states. In this work we introduce Concrete random variables---continuous relaxations of discrete random variables. The Concrete distribution is a new family of distributions with closed form densities and a simple reparameterization. Whenever a discrete stochastic node of a computation graph can be refactored into a one-hot bit representation that is treated continuously, Concrete stochastic nodes can be used with automatic differentiation to produce low-variance biased gradients of objectives (including objectives that depend on the log-probability of latent stochastic nodes) on the corresponding discrete graph. We demonstrate the effectiveness of Concrete relaxations on density estimation and structured prediction tasks using neural networks."
                },
                {
                    "title": "Pointer Sentinel Mixture Models",
                    "abstract": "Recent neural network sequence models with softmax classifiers have achieved their best language modeling performance only with very large hidden states and large vocabularies. Even then they struggle to predict rare or unseen words even if the context makes the prediction unambiguous. We introduce the pointer sentinel mixture architecture for neural sequence models which has the ability to either reproduce a word from the recent context or produce a word from a standard softmax classifier. Our pointer sentinel-LSTM model achieves state of the art language modeling performance on the Penn Treebank (70.9 perplexity) while using far fewer parameters than a standard softmax LSTM. In order to evaluate how well language models can exploit longer contexts and deal with more realistic vocabularies and larger corpora we also introduce the freely available WikiText corpus."
                },
                {
                    "title": "DoReFa-Net: Training Low Bitwidth Convolutional Neural Networks with Low Bitwidth Gradients",
                    "abstract": "We propose DoReFa-Net, a method to train convolutional neural networks that have low bitwidth weights and activations using low bitwidth parameter gradients. In particular, during backward pass, parameter gradients are stochastically quantized to low bitwidth numbers before being propagated to convolutional layers. As convolutions during forward/backward passes can now operate on low bitwidth weights and activations/gradients respectively, DoReFa-Net can use bit convolution kernels to accelerate both training and inference. Moreover, as bit convolutions can be efficiently implemented on CPU, FPGA, ASIC and GPU, DoReFa-Net opens the way to accelerate training of low bitwidth neural network on these hardware. Our experiments on SVHN and ImageNet datasets prove that DoReFa-Net can achieve comparable prediction accuracy as 32-bit counterparts. For example, a DoReFa-Net derived from AlexNet that has 1-bit weights, 2-bit activations, can be trained from scratch using 6-bit gradients to get 46.1\\% top-1 accuracy on ImageNet validation set. The DoReFa-Net AlexNet model is released publicly."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Improving Multifrontal Methods by Means of Block Low-Rank Representations",
                    "abstract": "Matrices coming from elliptic Partial Differential Equations (PDEs) have been shown to have a low-rank property: well defined off-diagonal blocks of their Schur complements can be approximated by low-rank products. Given a suitable ordering of the matrix which gives to the blocks a geometrical meaning, such approximations can be computed using an SVD or a rank-revealing QR factorization. The resulting representation offers a substantial reduction of the memory requirement and gives efficient ways to perform many of the basic dense algebra operations. Several strategies have been proposed to exploit this property. We propose a low-rank format called Block Low-Rank (BLR), and explain how it can be used to reduce the memory footprint and the complexity of direct solvers for sparse matrices based on the multifrontal method. We present experimental results that show how the BLR format delivers gains that are comparable to those obtained with hierarchical formats such as Hierarchical matrices (H matrices) and Hierarchically Semi-Separable (HSS matrices) but provides much greater flexibility and ease of use which are essential in the context of a general purpose, algebraic solver."
                },
                {
                    "title": "Butterfly Factorization",
                    "abstract": "The paper introduces the butterfly factorization as a data-sparse approximation for the matrices that satisfy a complementary low-rank property. The factorization can be constructed efficiently if either fast algorithms for applying the matrix and its adjoint are available or the entries of the matrix can be sampled individually. For an $N\\times N$ matrix, the resulting factorization is a product of $O(\\log N)$ sparse matrices, each with $O(N)$ nonzero entries. Hence, it can be applied rapidly in $O(N\\log N)$ operations. Numerical results are provided to demonstrate the effectiveness of the butterfly factorization and its construction algorithms."
                },
                {
                    "title": "Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation",
                    "abstract": "Stochastic neurons and hard non-linearities can be useful for a number of reasons in deep learning models, but in many cases they pose a challenging problem: how to estimate the gradient of a loss function with respect to the input of such stochastic or non-smooth neurons? I.e., can we \"back-propagate\" through these stochastic neurons? We examine this question, existing approaches, and compare four families of solutions, applicable in different settings. One of them is the minimum variance unbiased gradient estimator for stochatic binary neurons (a special case of the REINFORCE algorithm). A second approach, introduced here, decomposes the operation of a binary stochastic neuron into a stochastic binary part and a smooth differentiable part, which approximates the expected effect of the pure stochatic binary neuron to first order. A third approach involves the injection of additive or multiplicative noise in a computational graph that is otherwise differentiable. A fourth approach heuristically copies the gradient with respect to the stochastic output directly as an estimator of the gradient with respect to the sigmoid argument (we call this the straight-through estimator). To explore a context where these estimators are useful, we consider a small-scale version of {\\em conditional computation}, where sparse stochastic units form a distributed representation of gaters that can turn off in combinatorially many ways large chunks of the computation performed in the rest of the neural network. In this case, it is important that the gating units produce an actual 0 most of the time. The resulting sparsity can be potentially be exploited to greatly reduce the computational cost of large deep networks for which conditional computation would be useful."
                },
                {
                    "title": "An algorithm for the machine calculation of complex Fourier series",
                    "abstract": "An efficient method for the calculation of the interactions of a 2' factorial ex- periment was introduced by Yates and is widely known by his name. The generaliza- tion to 3' was given by Box et al. (1). Good (2) generalized these methods and gave elegant algorithms for which one class of applications is the calculation of Fourier series. In their full generality, Good's methods are applicable to certain problems in which one must multiply an N-vector by an N X N matrix which can be factored into m sparse matrices, where m is proportional to log N. This results inma procedure requiring a number of operations proportional to N log N rather than N2. These methods are applied here to the calculation of complex Fourier series. They are useful in situations where the number of data points is, or can be chosen to be, a highly composite number. The algorithm is here derived and presented in a rather different form. Attention is given to the choice of N. It is also shown how special advantage can be obtained in the use of a binary computer with N = 2' and how the entire calculation can be performed within the array of N data storage locations used for the given Fourier coefficients. Consider the problem of calculating the complex Fourier series N-1 (1) X(j) = EA(k)-Wjk, j = 0 1, * ,N- 1, k=0"
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Runtime Neural Pruning",
                    "abstract": "In this paper, we propose a Runtime Neural Pruning (RNP) framework which prunes the deep neural network dynamically at the runtime. Unlike existing neural pruning methods which produce a fixed pruned model for deployment, our method preserves the full ability of the original network and conducts pruning according to the input image and current feature maps adaptively. The pruning is performed in a bottom-up, layer-by-layer manner, which we model as a Markov decision process and use reinforcement learning for training. The agent judges the importance of each convolutional kernel and conducts channel-wise pruning conditioned on different samples, where the network is pruned more when the image is easier for the task. Since the ability of network is fully preserved, the balance point is easily adjustable according to the available resources. Our method can be applied to off-the-shelf network structures and reach a better tradeoff between speed and accuracy, especially with a large pruning rate."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CV",
                "eess.IV",
                "eess.SP"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nIs there a universal format that represents a wide range of structured matrices in deep neural networks, and can the structure of such matrices be learned efficiently?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing complexity and resource demands of large-scale deep neural networks (DNNs). A unified structured matrix format could lead to more efficient model architectures, reducing computational costs and energy consumption. This advancement could pave the way for practical applications in various domains, enabling the deployment of powerful models in resource-constrained environments. Furthermore, it could inspire future research into novel matrix structures and learning algorithms, ultimately enhancing our understanding of DNNs and their capabilities.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the lack of a unified representation for various structured matrices, which are often defined disjointedly. Naive approaches may fail because they do not account for the intricate relationships between different matrix structures or the non-differentiable nature of existing methods. Additionally, the complexities involved in learning the optimal structures of weight matrices, particularly in high-dimensional spaces, pose significant technical and theoretical obstacles. The need for a differentiable training method that can effectively capture and optimize these structures adds another layer of difficulty.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on isolated classes of structured matrices without establishing a comprehensive framework that encompasses multiple formats. This disjointed approach has limited the exploration of potential relationships between different matrix structures. Barriers such as the non-differentiable nature of existing structured matrix implementations and the absence of a systematic methodology for learning these structures have hindered progress. Our approach differs by introducing a Generalized Block-low-rank (GBLR) matrix format that integrates various structures and a differentiable parameterization method, allowing for more effective learning and optimization.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the introduction of the Generalized Block-low-rank (GBLR) matrix format, which encompasses various structured matrices like Low-Rank (LR), Block Sparse (BSP), and Block-low-rank (BLR) matrices. We will utilize a differentiable parameterization of the structural parameters, defined in the frequency domain and processed through the Gaussian-Dirichlet (Gaudi) function followed"
            }
        },
        "author_data": {
            "1c45f114-fae7-4deb-93b9-4a08561645a4": {
                "pk": "1c45f114-fae7-4deb-93b9-4a08561645a4",
                "name": "Changwoo Lee",
                "collaborators": [
                    "Hun-Seok Kim",
                    "Zichen Fan",
                    "Qirui Zhang",
                    "Ang Cao",
                    "Bowen Liu",
                    "Hyochan An",
                    "Boxun Xu",
                    "Li Xu",
                    "Chien-wei Tseng",
                    "Yimai Peng"
                ],
                "domain": [
                    "Deep Learning",
                    "Signal Processing",
                    "Internet of Things",
                    "Multi-task Learning"
                ],
                "publications": [
                    {
                        "title": "AIMMI: Audio and Image Multi-Modal Intelligence via a Low-Power SoC With 2-MByte On-Chip MRAM for IoT Devices",
                        "abstract": "In this article, we present an ultra-low-power multi-modal signal processing system on chip (SoC) [audio and image multi-modal intelligence (AIMMI)] that integrates a versatile deep neural network (DNN) engine with audio and image signal processing accelerators for multi-modal Internet-of-Things (IoT) intelligence. In order to get high energy efficiency under resource-constrained IoT scenarios, AIMMI features three efficiency-boosting techniques: 1) 2-MB on-chip non-volatile magnetoresistive RAM (MRAM) to store all DNN weights with MRAM-cache microarchitecture that incorporates dynamic power gating to reduce both leakage and dynamic power consumption; 2) a deliberate power management scheme that enables optimized power modes under different operating situations; and 3) a novel reconfigurable neural engine (NE) with energy-efficient dataflow for comprehensive DNN instructions. Fabricated in TSMC 22-nm ultra-low leakage (ULL) technology with MRAM, AIMMI achieves up to 3\u201310-TOPS/W peak energy efficiency and consumes only 0.25\u20133.84 mW. It demonstrates convolutional neural network (CNN), generative adversarial network (GAN), and back-propagation (BP) operations on a single accelerator SoC for multi-modal fusion, outperforming state-of-the-art DNN processors by $1.4 \\times -4.5 \\times$ in energy efficiency."
                    },
                    {
                        "title": "TaskFusion: An Efficient Transfer Learning Architecture with Dual Delta Sparsity for Multi-Task Natural Language Processing",
                        "abstract": "The combination of pre-trained models and task-specific fine-tuning schemes, such as BERT, has achieved great success in various natural language processing (NLP) tasks. However, the large memory and computation costs of such models make it challenging to deploy them in edge devices. Moreover, in real-world applications like chatbots, multiple NLP tasks need to be processed together to achieve higher response credibility. Running multiple NLP tasks with specialized models for each task increases the latency and memory cost latency linearly with the number of tasks. Though there have been recent works on parameter-shared tuning that aim to reduce the total parameter size by partially sharing weights among multiple tasks, computation remains intensive and redundant despite different tasks using the same input. In this work, we identify that a significant portion of activations and weights can be reused among different tasks, to reduce cost and latency for efficient multi-task NLP. Specifically, we propose TaskFusion, an efficient transfer learning software-hardware co-design that exploits delta sparsity in both weights and activations to boost data sharing among tasks. For training, TaskFusion uses \u21131 regularization on delta activation to learn inter-task data redundancies. A novel hardware-aware sub-task inference algorithm is proposed to exploit the dual delta sparsity. We then designed a dedicated heterogeneous architecture to accelerate multi-task inference with an optimized scheduling to increase hardware utilization and reduce off-chip memory access. Extensive experiments demonstrate that TaskFusion can reduce the number of floating point operations (FLOPs) by over 73% in multi-task NLP with negligible accuracy loss, while adding a new task at the cost of only < 2% parameter size increase. With the proposed architecture and optimized scheduling, Task-Fusion can achieve 1.48--2.43\u00d7 performance and 1.62--3.77\u00d7 energy efficiency than those using state-of-the-art single-task accelerators for multi-task NLP applications."
                    },
                    {
                        "title": "Deep Joint Source-Channel Coding with Iterative Source Error Correction",
                        "abstract": "In this paper, we propose an iterative source error correction (ISEC) decoding scheme for deep-learning-based joint source-channel coding (Deep JSCC). Given a noisy codeword received through the channel, we use a Deep JSCC encoder and decoder pair to update the codeword iteratively to find a (modified) maximum a-posteriori (MAP) solution. For efficient MAP decoding, we utilize a neural network-based denoiser to approximate the gradient of the log-prior density of the codeword space. Albeit the non-convexity of the optimization problem, our proposed scheme improves various distortion and perceptual quality metrics from the conventional one-shot (non-iterative) Deep JSCC decoding baseline. Furthermore, the proposed scheme produces more reliable source reconstruction results compared to the baseline when the channel noise characteristics do not match the ones used during training."
                    },
                    {
                        "title": "Learning-Based Near-Orthogonal Superposition Code for MIMO Short Message Transmission",
                        "abstract": "Massive machine type communication (mMTC) has attracted new coding schemes optimized for reliable short message transmission. In this paper, a novel deep learning-based near-orthogonal superposition (NOS) coding scheme is proposed to transmit short messages in multiple-input multiple-output (MIMO) channels for mMTC applications. In the proposed MIMO-NOS scheme, a neural network-based encoder is optimized via end-to-end learning with a corresponding neural network-based detector/decoder in a superposition-based auto-encoder framework including a MIMO channel. The proposed MIMO-NOS encoder spreads the information bits to multiple near-orthogonal high dimensional vectors to be combined (superimposed) into a single vector and reshaped for the space-time transmission. For the receiver, we propose a novel looped $K$ -best tree-search algorithm with cyclic redundancy check (CRC) assistance to enhance the error correcting ability in the block-fading MIMO channel. For a comprehensive understanding of the proposed MIMO-NOS scheme, we further quantify the gain from individual components/modules in the framework, and analyze the decoding complexity measured by the floating point operations (FLOPs). Simulation results show the proposed MIMO-NOS scheme outperforms maximum likelihood (ML) MIMO detection combined with a polar code with CRC-assisted list decoding by 1 \u2013 2 dB in various MIMO systems for short (32 \u2013 64 bit) message transmission."
                    },
                    {
                        "title": "Audio and Image Cross-Modal Intelligence via a 10TOPS/W 22nm SoC with Back-Propagation and Dynamic Power Gating",
                        "abstract": "We present an ultra-low-power multimedia signal processor (MMSP) SoC that integrates a versatile deep neural network (DNN) engine with audio and image signal processing accelerators for cross-modal IoT intelligence. The proposed MMSP features 2MB MRAM to store all DNN weights on-chip with an energy-efficient dataflow using an MRAM-cache and dynamic power gating. The SoC achieves up to 3-10 TOPS/W peak energy efficiency and consumes only 0.25-3.84 mW. Being the first to demonstrate CNN, GAN, and back-propagation (BP) on a single accelerator SoC for cross-modal fusion, it outperforms state-of-the-art DNN processors by 1.4 - 4.5\u00d7 in energy efficiency."
                    },
                    {
                        "title": "Unified Signal Compression Using a GAN with Iterative Latent Representation Optimization",
                        "abstract": "We propose a unified signal compression framework that uses a generative adversarial network (GAN) to compress heterogeneous signals. The compressed signal is represented as a latent vector and fed into a generator network that is trained to produce high quality realistic signals that minimize a target objective function. To efficiently quantize the compressed signal, non-uniformly quantized optimal latent vectors are identified by iterative back-propagation with alternating direction method of multipliers (ADMM) optimization performed for each iteration. The performance of the proposed signal compression method is assessed using multiple metrics including PSNR and MS-SSIM for image compression and also PESR, Kaldi, LSTM, and MLP performance for speech compression. Test results show that the proposed work outperforms recent state-of-the-art hand-crafted and deep learning-based signal compression methods."
                    },
                    {
                        "title": "A Confidence-Calibrated MOBA Game Winner Predictor",
                        "abstract": "In this paper, we propose a confidence-calibration method for predicting the winner of a famous multiplayer online battle arena (MOBA) game, League of Legends. In MOBA games, the dataset may contain a large amount of input-dependent noise; not all of such noise is observable. Hence, it is desirable to attempt a confidence-calibrated prediction. Unfortunately, most existing confidence calibration methods are pertaining to image and document classification tasks where consideration on uncertainty is not crucial. In this paper, we propose a novel calibration method that takes data uncertainty into consideration. The proposed method achieves an outstanding expected calibration error (ECE) (0.57%) mainly owing to data uncertainty consideration, compared to a conventional temperature scaling method of which ECE value is 1.11%."
                    },
                    {
                        "title": "GRAM: Gradient Rescaling Attention Model for Data Uncertainty Estimation in Single Image Super Resolution",
                        "abstract": "In this paper, a new learning method to quantify data uncertainty without suffering from performance degradation in Single Image Super Resolution (SISR) is proposed. Our work is motivated by the fact that the idea of loss design for capturing uncertainty and that for solving SISR are contradictory. As to capturing data uncertainty, we often model the output of a network as a Euclidian distance divided by a predictive variance, negative log-likelihood (NLL) for the Gaussian distribution, so that images with high variance have less impact on training. On the other hand, in the SISR domain, recent works give more weights to the loss of challenging images to improve the performance by using attention models. Nonetheless, the conflict should be handled to make neural networks capable of predicting the uncertainty of a super-resolved image, without suffering from performance degradation. Therefore, we propose a method called Gradient Rescaling Attention Model (GRAM) that combines both attempts effectively. Since variance may reflect the difficulty of an image, we rescale the gradient of NLL by the degree of variance. Hence, the neural network can focus on the challenging images, similarly to attention models. We conduct performance evaluation using standard SISR benchmarks in terms of peak signal-noise ratio (PSNR) and structural similarity (SSIM). The experimental results show that the proposed gradient rescaling method generates negligible performance degradation compared to SISR outputs with the Euclidian loss, whereas NLL without attention degrades the SR quality."
                    }
                ]
            },
            "a42a281e-4066-4b1f-b30c-d65a206b177a": {
                "pk": "a42a281e-4066-4b1f-b30c-d65a206b177a",
                "name": "Hun-Seok Kim",
                "collaborators": [
                    "David Blaauw",
                    "Chien-wei Tseng",
                    "Kuan-Yu Chen",
                    "Morteza Fayazi",
                    "Ronald Dreslinski",
                    "Zichen Fan",
                    "Qirui Zhang",
                    "Hyochan An",
                    "Boxun Xu",
                    "Li Xu"
                ],
                "domain": [
                    "Low-Power Computing",
                    "Multi-Modal Signal Processing",
                    "Hardware Architecture",
                    "Internet of Things"
                ],
                "publications": [
                    {
                        "title": "AIMMI: Audio and Image Multi-Modal Intelligence via a Low-Power SoC With 2-MByte On-Chip MRAM for IoT Devices",
                        "abstract": "In this article, we present an ultra-low-power multi-modal signal processing system on chip (SoC) [audio and image multi-modal intelligence (AIMMI)] that integrates a versatile deep neural network (DNN) engine with audio and image signal processing accelerators for multi-modal Internet-of-Things (IoT) intelligence. In order to get high energy efficiency under resource-constrained IoT scenarios, AIMMI features three efficiency-boosting techniques: 1) 2-MB on-chip non-volatile magnetoresistive RAM (MRAM) to store all DNN weights with MRAM-cache microarchitecture that incorporates dynamic power gating to reduce both leakage and dynamic power consumption; 2) a deliberate power management scheme that enables optimized power modes under different operating situations; and 3) a novel reconfigurable neural engine (NE) with energy-efficient dataflow for comprehensive DNN instructions. Fabricated in TSMC 22-nm ultra-low leakage (ULL) technology with MRAM, AIMMI achieves up to 3\u201310-TOPS/W peak energy efficiency and consumes only 0.25\u20133.84 mW. It demonstrates convolutional neural network (CNN), generative adversarial network (GAN), and back-propagation (BP) operations on a single accelerator SoC for multi-modal fusion, outperforming state-of-the-art DNN processors by $1.4 \\times -4.5 \\times$ in energy efficiency."
                    },
                    {
                        "title": "Canalis: A Throughput-Optimized Framework for Real-Time Stream Processing of Wireless Communication",
                        "abstract": "Stream processing, which involves real-time computation of data as it is created or received, is vital for various applications, specifically wireless communication. The evolving protocols, the requirement for high-throughput, and the challenges of handling diverse processing patterns make it demanding. Traditional platforms grapple with meeting real-time throughput and latency requirements due to large data volume, sequential and indeterministic data arrival, and variable data rates, leading to inefficiencies in memory access and parallel processing. We present Canalis, a throughput-optimized framework designed to address these challenges, ensuring high-performance while achieving low energy consumption. Canalis is a hardware-software co-designed system. It includes a programmable spatial architecture, FluxSPU (Flux Stream Processing Unit), proposed by this work to enhance data throughput and energy efficiency. FluxSPU is accompanied by a software stack that eases the programming process. We evaluated Canalis with eight distinct benchmarks. When compared to CPU and GPU in mobile SoC to demonstrate the effectiveness of domain specialization, Canalis achieves an average speedup of 13.4\u00d7 and 6.6\u00d7, and energy savings of 189.8\u00d7 and 283.9\u00d7, respectively. In contrast to equivalent ASICs of the benchmarks, the average energy overhead of Canalis is within 2.4\u00d7, successfully maintaining generalizations without incurring significant overhead."
                    },
                    {
                        "title": "DAP: A 507-GMACs/J 256-Core Domain Adaptive Processor for Wireless Communication and Linear Algebra Kernels in 12-nm FINFET",
                        "abstract": "\u2014We present domain adaptive processor (DAP), a programmable systolic-array processor designed for wireless communication and linear algebra workloads. DAP uses a globally homogeneous but locally heterogeneous architecture, uses decode-less recon\ufb01guration instructions for data streaming, enables single-cycle data communication between functional units (FUs), and features lightweight nested-loop control for periodic execution. Our design demonstrates how con\ufb01guration \ufb02exibility and rapid program loading enable a wide range of communication workloads to be mapped and swapped in less than a microsecond, supporting continually evolving communication standards such as 5G. A prototype chip of DAP with 256 cores is fabricated in a 12-nm FINFET process and has been veri\ufb01ed. The measurement results show that DAP achieves 507 GMACs/J and a peak performance of 264 GMACs."
                    }
                ]
            }
        }
    },
    "2302.03357": {
        "paper_data": {
            "title": "Towards Enhancing Time Series Contrastive Learning: A Dynamic Bad Pair Mining Approach",
            "url": "http://arxiv.org/abs/2302.03357v3",
            "arxiv_id": "2302.03357",
            "authors": [
                "Xiang Lan",
                "Hanshu Yan",
                "Shenda Hong",
                "Mengling Feng"
            ],
            "abstract": "Not all positive pairs are beneficial to time series contrastive learning. In this paper, we study two types of bad positive pairs that can impair the quality of time series representation learned through contrastive learning: the noisy positive pair and the faulty positive pair. We observe that, with the presence of noisy positive pairs, the model tends to simply learn the pattern of noise (Noisy Alignment). Meanwhile, when faulty positive pairs arise, the model wastes considerable amount of effort aligning non-representative patterns (Faulty Alignment). To address this problem, we propose a Dynamic Bad Pair Mining (DBPM) algorithm, which reliably identifies and suppresses bad positive pairs in time series contrastive learning. Specifically, DBPM utilizes a memory module to dynamically track the training behavior of each positive pair along training process. This allows us to identify potential bad positive pairs at each epoch based on their historical training behaviors. The identified bad pairs are subsequently down-weighted through a transformation module, thereby mitigating their negative impact on the representation learning process. DBPM is a simple algorithm designed as a lightweight plug-in without learnable parameters to enhance the performance of existing state-of-the-art methods. Through extensive experiments conducted on four large-scale, real-world time series datasets, we demonstrate DBPM's efficacy in mitigating the adverse effects of bad positive pairs.",
            "introduction": "   1 Introduction  Self-supervised contrastive learning has shown remarkable efficacy in learning meaningful representations from unlabeled time series data, thereby improving the performance of downstream tasks (e.g.formulae-sequence\ud835\udc52\ud835\udc54e.g.italic_e . italic_g ., time series classification) (Oord et\u00a0al., 2018; Tonekaboni et\u00a0al., 2021; Eldele et\u00a0al., 2021; Y\u00e8che et\u00a0al., 2021; Lan et\u00a0al., 2022; Yue et\u00a0al., 2022; Yang & Hong, 2022; Zhang et\u00a0al., 2022b; Ozyurt et\u00a0al., 2023). The key concept of time series contrastive learning is to capture the meaningful underlying information shared between views (usually generated by random data augmentation), such that the learned representations are discriminative among time series. To accomplish this objective, existing contrastive methods comply with an assumption that the augmented views from the same instance (i.e.formulae-sequence\ud835\udc56\ud835\udc52i.e.italic_i . italic_e ., positive pair) share meaningful semantic information (Chen et\u00a0al., 2020; He et\u00a0al., 2020a; Chen & He, 2021a; Wen & Li, 2021).   But what happens if this assumption is violated? Some recent studies (Morgado et\u00a0al., 2021; Chuang et\u00a0al., 2022) have investigated the effects of the faulty view problem in image contrastive learning. It was discovered that the effectiveness of the contrastive learning models will indeed suffer when two views generated from the same image actually do not share meaningful information (e.g.formulae-sequence\ud835\udc52\ud835\udc54e.g.italic_e . italic_g ., an unrelated cat-dog pair cropped from the same image). In other words, not all positive pairs are beneficial to contrastive learning. We extend this line of work to time series by investigating whether time series contrastive learning encounters similar challenges, and explore what kind of positive pairs are detrimental to the learning process. Specifically, when applying commonly used contrastive learning methods to time series applications, we observe two extreme cases of positive pairs that violate the assumption and consequently vitiate the quality of the learned time series representation:   Noisy Positive Pairs: As shown in Figure 1 (a), a real-world example from the Sleep-EDF dataset (Goldberger et\u00a0al., 2000) (EEG signal). These pairs can arise when the original signal exhibits substantial noise, which could be attributed to improper data collection in real-world scenarios. This results in noisy contrasting views, where the shared information between views is predominantly noise. In the given example, the low voltage EEG is interfered by other electrical signals such as the electrooculogram (EOG) from eye movement, which makes some subtle EEG spikes indicative of early stage epilepsy being overwhelmed by noise. As a result, a noisy alignment occurs, in which the model simply learns the pattern of noise rather than the true signal.   Faulty Positive Pairs: As shown in Figure 1 (b), a real-world example from the PTB-XL dataset (Wagner et\u00a0al., 2020) (ECG signal). Due to the sensitivity of time series, data augmentation may inadvertently impair sophisticated temporal patterns contained in the original signal and thus produce faulty views. In the example, the unique temporal pattern exhibited by ventricular premature beat (red part) for diagnosing ECG abnormality has been destroyed after augmentation. Consequently, the augmented view is no longer has the same semantic meaning as the original ECG, leading to a faulty alignment where the model learns to align non-representative patterns shared between views.   For clarity, we refer to these two extreme cases as bad",
            "references": [
                {
                    "title": "Self-Supervised Contrastive Pre-Training For Time Series via Time-Frequency Consistency",
                    "abstract": "Pre-training on time series poses a unique challenge due to the potential mismatch between pre-training and target domains, such as shifts in temporal dynamics, fast-evolving trends, and long-range and short-cyclic effects, which can lead to poor downstream performance. While domain adaptation methods can mitigate these shifts, most methods need examples directly from the target domain, making them suboptimal for pre-training. To address this challenge, methods need to accommodate target domains with different temporal dynamics and be capable of doing so without seeing any target examples during pre-training. Relative to other modalities, in time series, we expect that time-based and frequency-based representations of the same example are located close together in the time-frequency space. To this end, we posit that time-frequency consistency (TF-C) -- embedding a time-based neighborhood of an example close to its frequency-based neighborhood -- is desirable for pre-training. Motivated by TF-C, we define a decomposable pre-training model, where the self-supervised signal is provided by the distance between time and frequency components, each individually trained by contrastive estimation. We evaluate the new method on eight datasets, including electrodiagnostic testing, human activity recognition, mechanical fault detection, and physical status monitoring. Experiments against eight state-of-the-art methods show that TF-C outperforms baselines by 15.4% (F1 score) on average in one-to-one settings (e.g., fine-tuning an EEG-pretrained model on EMG data) and by 8.4% (precision) in challenging one-to-many settings (e.g., fine-tuning an EEG-pretrained model for either hand-gesture recognition or mechanical fault prediction), reflecting the breadth of scenarios that arise in real-world applications. Code and datasets: https://github.com/mims-harvard/TFC-pretraining."
                },
                {
                    "title": "Contrastive Learning for Unsupervised Domain Adaptation of Time Series",
                    "abstract": "Unsupervised domain adaptation (UDA) aims at learning a machine learning model using a labeled source domain that performs well on a similar yet different, unlabeled target domain. UDA is important in many applications such as medicine, where it is used to adapt risk scores across different patient cohorts. In this paper, we develop a novel framework for UDA of time series data, called CLUDA. Specifically, we propose a contrastive learning framework to learn contextual representations in multivariate time series, so that these preserve label information for the prediction task. In our framework, we further capture the variation in the contextual representations between source and target domain via a custom nearest-neighbor contrastive learning. To the best of our knowledge, ours is the first framework to learn domain-invariant, contextual representation for UDA of time series data. We evaluate our framework using a wide range of time series datasets to demonstrate its effectiveness and show that it achieves state-of-the-art performance for time series UDA."
                },
                {
                    "title": "How Does SimSiam Avoid Collapse Without Negative Samples? A Unified Understanding with Self-supervised Contrastive Learning",
                    "abstract": "To avoid collapse in self-supervised learning (SSL), a contrastive loss is widely used but often requires a large number of negative samples. Without negative samples yet achieving competitive performance, a recent work has attracted significant attention for providing a minimalist simple Siamese (SimSiam) method to avoid collapse. However, the reason for how it avoids collapse without negative samples remains not fully clear and our investigation starts by revisiting the explanatory claims in the original SimSiam. After refuting their claims, we introduce vector decomposition for analyzing the collapse based on the gradient analysis of the $l_2$-normalized representation vector. This yields a unified perspective on how negative samples and SimSiam alleviate collapse. Such a unified perspective comes timely for understanding the recent progress in SSL."
                },
                {
                    "title": "Rethinking Minimal Sufficient Representation in Contrastive Learning",
                    "abstract": "Contrastive learning between different views of the data achieves outstanding success in the field of self-supervised representation learning and the learned representations are useful in broad downstream tasks. Since all supervision information for one view comes from the other view, contrastive learning approximately obtains the minimal sufficient representation which contains the shared information and eliminates the non-shared information between views. Considering the diversity of the downstream tasks, it cannot be guaranteed that all task-relevant information is shared between views. Therefore, we assume the non-shared task-relevant information cannot be ignored and theoretically prove that the minimal sufficient representation in contrastive learning is not sufficient for the downstream tasks, which causes performance degradation. This reveals a new problem that the contrastive learning models have the risk of overfitting to the shared information between views. To alleviate this problem, we propose to increase the mutual information between the representation and input as regularization to approximately introduce more task-relevant information, since we cannot utilize any downstream task information during training. Extensive experiments verify the rationality of our analysis and the effectiveness of our method. It significantly improves the performance of several classic contrastive learning models in downstream tasks. Our code is available at https://github.com/Haoqing-Wang/InfoCL."
                },
                {
                    "title": "Unsupervised Time-Series Representation Learning with Iterative Bilinear Temporal-Spectral Fusion",
                    "abstract": "Unsupervised/self-supervised time series representation learning is a challenging problem because of its complex dynamics and sparse annotations. Existing works mainly adopt the framework of contrastive learning with the time-based augmentation techniques to sample positives and negatives for contrastive training. Nevertheless, they mostly use segment-level augmentation derived from time slicing, which may bring about sampling bias and incorrect optimization with false negatives due to the loss of global context. Besides, they all pay no attention to incorporate the spectral information in feature representation. In this paper, we propose a unified framework, namely Bilinear Temporal-Spectral Fusion (BTSF). Specifically, we firstly utilize the instance-level augmentation with a simple dropout on the entire time series for maximally capturing long-term dependencies. We devise a novel iterative bilinear temporal-spectral fusion to explicitly encode the affinities of abundant time-frequency pairs, and iteratively refines representations in a fusion-and-squeeze manner with Spectrum-to-Time (S2T) and Time-to-Spectrum (T2S) Aggregation modules. We firstly conducts downstream evaluations on three major tasks for time series including classification, forecasting and anomaly detection. Experimental results shows that our BTSF consistently significantly outperforms the state-of-the-art methods."
                },
                {
                    "title": "CoST: Contrastive Learning of Disentangled Seasonal-Trend Representations for Time Series Forecasting",
                    "abstract": "Deep learning has been actively studied for time series forecasting, and the mainstream paradigm is based on the end-to-end training of neural network architectures, ranging from classical LSTM/RNNs to more recent TCNs and Transformers. Motivated by the recent success of representation learning in computer vision and natural language processing, we argue that a more promising paradigm for time series forecasting, is to first learn disentangled feature representations, followed by a simple regression fine-tuning step -- we justify such a paradigm from a causal perspective. Following this principle, we propose a new time series representation learning framework for time series forecasting named CoST, which applies contrastive learning methods to learn disentangled seasonal-trend representations. CoST comprises both time domain and frequency domain contrastive losses to learn discriminative trend and seasonal representations, respectively. Extensive experiments on real-world datasets show that CoST consistently outperforms the state-of-the-art methods by a considerable margin, achieving a 21.3% improvement in MSE on multivariate benchmarks. It is also robust to various choices of backbone encoders, as well as downstream regressors. Code is available at https://github.com/salesforce/CoST."
                },
                {
                    "title": "Robust Contrastive Learning against Noisy Views",
                    "abstract": "Contrastive learning relies on an assumption that positive pairs contain related views that share certain underlying information about an instance, e.g., patches of an image or co-occurring multimodal signals of a video. What if this assumption is violated? The literature suggests that contrastive learning produces suboptimal representations in the presence of noisy views, e.g., false positive pairs with no apparent shared information. In this work, we pro-pose a new contrastive loss function that is robust against noisy views. We provide rigorous theoretical justifications by showing connections to robust symmetric losses for noisy binary classification and by establishing a new contrastive bound for mutual information maximization based on the Wasserstein distance measure. The proposed loss is completely modality-agnostic and a simple drop-in replacement for the InfoNCE loss, which makes it easy to apply to ex-isting contrastive frameworks. We show that our approach provides consistent improvements over the state-of-the-art on image, video, and graph contrastive learning bench-marks that exhibit a variety of real-world noise patterns."
                },
                {
                    "title": "Intra-Inter Subject Self-supervised Learning for Multivariate Cardiac Signals",
                    "abstract": "Learning information-rich and generalizable representations effectively from unlabeled multivariate cardiac signals to identify abnormal heart rhythms (cardiac arrhythmias) is valuable in real-world clinical settings but often challenging due to its complex temporal dynamics. Cardiac arrhythmias can vary significantly in temporal patterns even for the same patient (i.e., intra subject difference). Meanwhile, the same type of cardiac arrhythmia can show different temporal patterns among different patients due to different cardiac structures (i.e., inter subject difference). In this paper, we address the challenges by proposing an Intra-Inter Subject Self-Supervised Learning (ISL) model that is customized for multivariate cardiac signals. Our proposed ISL model integrates medical knowledge into self-supervision to effectively learn from intra-inter subject differences. In intra subject self-supervision, ISL model first extracts heartbeat-level features from each subject using a channel-wise attentional CNN-RNN encoder. Then a stationarity test module is employed to capture the temporal dependencies between heartbeats. In inter subject self-supervision, we design a set of data augmentations according to the clinical characteristics of cardiac signals and perform contrastive learning among subjects to learn distinctive representations for various types of patients. Extensive experiments on three real-world datasets were conducted. In a semi-supervised transfer learning scenario, our pre-trained ISL model leads about 10% improvement over supervised training when only 1% labeled data is available, suggesting strong generalizability and robustness of the model."
                },
                {
                    "title": "Time-Series Representation Learning via Temporal and Contextual Contrasting",
                    "abstract": "Learning decent representations from unlabeled time-series data with temporal dynamics is a very challenging task. In this paper, we propose an unsupervised Time-Series representation learning framework via Temporal and Contextual Contrasting (TS-TCC), to learn time-series representation from unlabeled data. First, the raw time-series data are transformed into two different yet correlated views by using weak and strong augmentations. Second, we propose a novel temporal contrasting module to learn robust temporal representations by designing a tough cross-view prediction task. Last, to further learn discriminative representations, we propose a contextual contrasting module built upon the contexts from the temporal contrasting module. It attempts to maximize the similarity among different contexts of the same sample while minimizing similarity among contexts of different samples. Experiments have been carried out on three real-world time-series datasets. The results manifest that training a linear classifier on top of the features learned by our proposed TS-TCC performs comparably with the supervised training. Additionally, our proposed TS-TCC shows high efficiency in few-labeled data and transfer learning scenarios. The code is publicly available at https://github.com/emadeldeen24/TS-TCC."
                },
                {
                    "title": "TS2Vec: Towards Universal Representation of Time Series",
                    "abstract": "This paper presents TS2Vec, a universal framework for learning representations of time series in an arbitrary semantic level. Unlike existing methods, TS2Vec performs contrastive learning in a hierarchical way over augmented context views, which enables a robust contextual representation for each timestamp. Furthermore, to obtain the representation of an arbitrary sub-sequence in the time series, we can apply a simple aggregation over the representations of corresponding timestamps. We conduct extensive experiments on time series classification tasks to evaluate the quality of time series representations. As a result, TS2Vec achieves significant improvement over existing SOTAs of unsupervised time series representation on 125 UCR datasets and 29 UEA datasets. The learned timestamp-level representations also achieve superior results in time series forecasting and anomaly detection tasks. A linear regression trained on top of the learned representations outperforms previous SOTAs of time series forecasting. Furthermore, we present a simple way to apply the learned representations for unsupervised anomaly detection, which establishes SOTA results in the literature. The source code is publicly available at https://github.com/yuezhihan/ts2vec."
                },
                {
                    "title": "Neighborhood Contrastive Learning Applied to Online Patient Monitoring",
                    "abstract": "Intensive care units (ICU) are increasingly looking towards machine learning for methods to provide online monitoring of critically ill patients. In machine learning, online monitoring is often formulated as a supervised learning problem. Recently, contrastive learning approaches have demonstrated promising improvements over competitive supervised benchmarks. These methods rely on well-understood data augmentation techniques developed for image data which do not apply to online monitoring. In this work, we overcome this limitation by supplementing time-series data augmentation techniques with a novel contrastive learning objective which we call neighborhood contrastive learning (NCL). Our objective explicitly groups together contiguous time segments from each patient while maintaining state-specific information. Our experiments demonstrate a marked improvement over existing work applying contrastive methods to medical time-series."
                },
                {
                    "title": "Unsupervised Representation Learning for Time Series with Temporal Neighborhood Coding",
                    "abstract": "Time series are often complex and rich in information but sparsely labeled and therefore challenging to model. In this paper, we propose a self-supervised framework for learning generalizable representations for non-stationary time series. Our approach, called Temporal Neighborhood Coding (TNC), takes advantage of the local smoothness of a signal's generative process to define neighborhoods in time with stationary properties. Using a debiased contrastive objective, our framework learns time series representations by ensuring that in the encoding space, the distribution of signals from within a neighborhood is distinguishable from the distribution of non-neighboring signals. Our motivation stems from the medical field, where the ability to model the dynamic nature of time series data is especially valuable for identifying, tracking, and predicting the underlying patients' latent states in settings where labeling data is practically impossible. We compare our method to recently developed unsupervised representation learning approaches and demonstrate superior performance on clustering and classification tasks for multiple datasets."
                },
                {
                    "title": "Toward Understanding the Feature Learning Process of Self-supervised Contrastive Learning",
                    "abstract": "How can neural networks trained by contrastive learning extract features from the unlabeled data? Why does contrastive learning usually need much stronger data augmentations than supervised learning to ensure good representations? These questions involve both the optimization and statistical aspects of deep learning, but can hardly be answered by analyzing supervised learning, where the target functions are the highest pursuit. Indeed, in self-supervised learning, it is inevitable to relate to the optimization/generalization of neural networks to how they can encode the latent structures in the data, which we refer to as the feature learning process. In this work, we formally study how contrastive learning learns the feature representations for neural networks by analyzing its feature learning process. We consider the case where our data are comprised of two types of features: the more semantically aligned sparse features which we want to learn from, and the other dense features we want to avoid. Theoretically, we prove that contrastive learning using $\\mathbf{ReLU}$ networks provably learns the desired sparse features if proper augmentations are adopted. We present an underlying principle called $\\textbf{feature decoupling}$ to explain the effects of augmentations, where we theoretically characterize how augmentations can reduce the correlations of dense features between positive samples while keeping the correlations of sparse features intact, thereby forcing the neural networks to learn from the self-supervision of sparse features. Empirically, we verified that the feature decoupling principle matches the underlying mechanism of contrastive learning in practice."
                },
                {
                    "title": "Contrastive Learning With Stronger Augmentations",
                    "abstract": "Representation learning has significantly been developed with the advance of contrastive learning methods. Most of those methods are benefited from various data augmentations that are carefully designated to maintain their identities so that the images transformed from the same instance can still be retrieved. However, those carefully designed transformations limited us to further explore the novel patterns exposed by other transformations. Meanwhile, as shown in our experiments, direct contrastive learning for stronger augmented images can not learn representations effectively. Thus, we propose a general framework called Contrastive Learning with Stronger Augmentations (CLSA) to complement current contrastive learning approaches. Here, the distribution divergence between the weakly and strongly augmented images over the representation bank is adopted to supervise the retrieval of strongly augmented queries from a pool of instances. Experiments on the ImageNet dataset and downstream datasets showed the information from the strongly augmented images can significantly boost the performance. For example, CLSA achieves top-1 accuracy of 76.2% on ImageNet with a standard ResNet-50 architecture with a single-layer classifier fine-tuned, which is almost the same level as 76.5% of supervised results."
                },
                {
                    "title": "Robust Audio-Visual Instance Discrimination",
                    "abstract": "We present a self-supervised learning method to learn audio and video representations. Prior work uses the natural correspondence between audio and video to define a standard cross-modal instance discrimination task, where a model is trained to match representations from the two modalities. However, the standard approach introduces two sources of training noise. First, audio-visual correspondences often produce faulty positives since the audio and video signals can be uninformative of each other. To limit the detrimental impact of faulty positives, we optimize a weighted contrastive learning loss, which down-weighs their contribution to the overall loss. Second, since self-supervised contrastive learning relies on random sampling of negative instances, instances that are semantically similar to the base instance can be used as faulty negatives. To alleviate the impact of faulty negatives, we propose to optimize an instance discrimination loss with a soft target distribution that estimates relationships between instances. We validate our contributions through extensive experiments on action recognition tasks and show that they address the problems of audio-visual instance discrimination and improve transfer learning performance."
                },
                {
                    "title": "Barlow Twins: Self-Supervised Learning via Redundancy Reduction",
                    "abstract": "Self-supervised learning (SSL) is rapidly closing the gap with supervised methods on large computer vision benchmarks. A successful approach to SSL is to learn embeddings which are invariant to distortions of the input sample. However, a recurring issue with this approach is the existence of trivial constant solutions. Most current methods avoid such solutions by careful implementation details. We propose an objective function that naturally avoids collapse by measuring the cross-correlation matrix between the outputs of two identical networks fed with distorted versions of a sample, and making it as close to the identity matrix as possible. This causes the embedding vectors of distorted versions of a sample to be similar, while minimizing the redundancy between the components of these vectors. The method is called Barlow Twins, owing to neuroscientist H. Barlow's redundancy-reduction principle applied to a pair of identical networks. Barlow Twins does not require large batches nor asymmetry between the network twins such as a predictor network, gradient stopping, or a moving average on the weight updates. Intriguingly it benefits from very high-dimensional output vectors. Barlow Twins outperforms previous methods on ImageNet for semi-supervised classification in the low-data regime, and is on par with current state of the art for ImageNet classification with a linear classifier head, and for transfer tasks of classification and object detection."
                },
                {
                    "title": "Exploring Simple Siamese Representation Learning",
                    "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code is made available.1"
                },
                {
                    "title": "Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamics",
                    "abstract": "Large datasets have become commonplace in NLP research. However, the increased emphasis on data quantity has made it challenging to assess the quality of data. We introduce Data Maps---a model-based tool to characterize and diagnose datasets. We leverage a largely ignored source of information: the behavior of the model on individual instances during training (training dynamics) for building data maps. This yields two intuitive measures for each example---the model's confidence in the true class, and the variability of this confidence across epochs---obtained in a single run of training. Experiments across four datasets show that these model-dependent measures reveal three distinct regions in the data map, each with pronounced characteristics. First, our data maps show the presence of \"ambiguous\" regions with respect to the model, which contribute the most towards out-of-distribution generalization. Second, the most populous regions in the data are \"easy to learn\" for the model, and play an important role in model optimization. Finally, data maps uncover a region with instances that the model finds \"hard to learn\"; these often correspond to labeling errors. Our results indicate that a shift in focus from quantity to quality of data could lead to robust models and improved out-of-distribution generalization."
                },
                {
                    "title": "Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning",
                    "abstract": "We introduce Bootstrap Your Own Latent (BYOL), a new approach to self-supervised image representation learning. BYOL relies on two neural networks, referred to as online and target networks, that interact and learn from each other. From an augmented view of an image, we train the online network to predict the target network representation of the same image under a different augmented view. At the same time, we update the target network with a slow-moving average of the online network. While state-of-the art methods rely on negative pairs, BYOL achieves a new state of the art without them. BYOL reaches $74.3\\%$ top-1 classification accuracy on ImageNet using a linear evaluation with a ResNet-50 architecture and $79.6\\%$ with a larger ResNet. We show that BYOL performs on par or better than the current state of the art on both transfer and semi-supervised benchmarks. Our implementation and pretrained models are given on GitHub."
                },
                {
                    "title": "What makes for good views for contrastive learning",
                    "abstract": "Contrastive learning between multiple views of the data has recently achieved state of the art performance in the field of self-supervised representation learning. Despite its success, the influence of different view choices has been less studied. In this paper, we use empirical analysis to better understand the importance of view selection, and argue that we should reduce the mutual information (MI) between views while keeping task-relevant information intact. To verify this hypothesis, we devise unsupervised and semi-supervised frameworks that learn effective views by aiming to reduce their MI. We also consider data augmentation as a way to reduce MI, and show that increasing data augmentation indeed leads to decreasing MI and improves downstream classification accuracy. As a by-product, we also achieve a new state-of-the-art accuracy on unsupervised pre-training for ImageNet classification ($73\\%$ top-1 linear readoff with a ResNet-50). In addition, transferring our models to PASCAL VOC object detection and COCO instance segmentation consistently outperforms supervised pre-training. Code:this http URL"
                },
                {
                    "title": "Deep Learning for ECG Analysis: Benchmarks and Insights from PTB-XL",
                    "abstract": "Electrocardiography (ECG) is a very common, non-invasive diagnostic procedure and its interpretation is increasingly supported by algorithms. The progress in the field of automatic ECG analysis has up to now been hampered by a lack of appropriate datasets for training as well as a lack of well-defined evaluation procedures to ensure comparability of different algorithms. To alleviate these issues, we put forward first benchmarking results for the recently published, freely accessible clinical 12-lead ECG dataset PTB-XL, covering a variety of tasks from different ECG statement prediction tasks to age and sex prediction. Among the investigated deep-learning-based timeseries classification algorithms, we find that convolutional neural networks, in particular resnet- and inception-based architectures, show the strongest performance across all tasks. We find consistent results on the ICBEB2018 challenge ECG dataset and discuss prospects of transfer learning using classifiers pretrained on PTB-XL. These benchmarking results are complemented by deeper insights into the classification algorithm in terms of hidden stratification, model uncertainty and an exploratory interpretability analysis, which provide connecting points for future research on the dataset. Our results emphasize the prospects of deep-learning-based algorithms in the field of ECG analysis, not only in terms of quantitative accuracy but also in terms of clinically equally important further quality metrics such as uncertainty quantification and interpretability. With this resource, we aim to establish the PTB-XL dataset as a resource for structured benchmarking of ECG analysis algorithms and encourage other researchers in the field to join these efforts."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Momentum Contrast for Unsupervised Visual Representation Learning",
                    "abstract": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a moving-averaged encoder. This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning. MoCo provides competitive results under the common linear protocol on ImageNet classification. More importantly, the representations learned by MoCo transfer well to downstream tasks. MoCo can outperform its supervised pre-training counterpart in 7 detection/segmentation tasks on PASCAL VOC, COCO, and other datasets, sometimes surpassing it by large margins. This suggests that the gap between unsupervised and supervised representation learning has been largely closed in many vision tasks."
                },
                {
                    "title": "Learning with Bad Training Data via Iterative Trimmed Loss Minimization",
                    "abstract": "In this paper, we study a simple and generic framework to tackle the problem of learning model parameters when a fraction of the training samples are corrupted. We first make a simple observation: in a variety of such settings, the evolution of training accuracy (as a function of training epochs) is different for clean and bad samples. Based on this we propose to iteratively minimize the trimmed loss, by alternating between (a) selecting samples with lowest current loss, and (b) retraining a model on only these samples. We prove that this process recovers the ground truth (with linear convergence rate) in generalized linear models with standard statistical assumptions. Experimentally, we demonstrate its effectiveness in three settings: (a) deep image classifiers with errors only in labels, (b) generative adversarial networks with bad training images, and (c) deep image classifiers with adversarial (image, label) pairs (i.e., backdoor attacks). For the well-studied setting of random label noise, our algorithm achieves state-of-the-art performance without having access to any a-priori guaranteed clean samples."
                },
                {
                    "title": "Representation Learning with Contrastive Predictive Coding",
                    "abstract": "While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments."
                },
                {
                    "title": "Data augmentation of wearable sensor data for parkinson\u2019s disease monitoring using convolutional neural networks",
                    "abstract": "While convolutional neural networks (CNNs) have been successfully applied to many challenging classification applications, they typically require large datasets for training. When the availability of labeled data is limited, data augmentation is a critical preprocessing step for CNNs. However, data augmentation for wearable sensor data has not been deeply investigated yet. In this paper, various data augmentation methods for wearable sensor data are proposed. The proposed methods and CNNs are applied to the classification of the motor state of Parkinson\u2019s Disease patients, which is challenging due to small dataset size, noisy labels, and large intra-class variability. Appropriate augmentation improves the classification performance from 77.54% to 86.88%."
                },
                {
                    "title": "Indications of nonlinear deterministic and finite-dimensional structures in time series of brain electrical activity: dependence on recording region and brain state.",
                    "abstract": "We compare dynamical properties of brain electrical activity from different recording regions and from different physiological and pathological brain states. Using the nonlinear prediction error and an estimate of an effective correlation dimension in combination with the method of iterative amplitude adjusted surrogate data, we analyze sets of electroencephalographic (EEG) time series: surface EEG recordings from healthy volunteers with eyes closed and eyes open, and intracranial EEG recordings from epilepsy patients during the seizure free interval from within and from outside the seizure generating area as well as intracranial EEG recordings of epileptic seizures. As a preanalysis step an inclusion criterion of weak stationarity was applied. Surface EEG recordings with eyes open were compatible with the surrogates' null hypothesis of a Gaussian linear stochastic process. Strongest indications of nonlinear deterministic dynamics were found for seizure activity. Results of the other sets were found to be inbetween these two extremes."
                },
                {
                    "title": "A Public Domain Dataset for Human Activity Recognition using Smartphones",
                    "abstract": "Human-centered computing is an emerging research field that aims to understand human behavior and integrate users and their social context with computer systems. One of the most recent, challenging and appealing applications in this framework consists in sensing human body motion using smartphones to gather context information about people actions. In this context, we describe in this work an Activity Recognition database, built from the recordings of 30 subjects doing Activities of Daily Living (ADL) while carrying a waist-mounted smartphone with embedded inertial sensors, which is released to public domain on a well-known on-line repository. Results, obtained on the dataset by exploiting a multiclass Support Vector \nMachine (SVM), are also acknowledged."
                },
                {
                    "title": "Visualizing Data using t-SNE",
                    "abstract": "We present a new technique called \u201ct-SNE\u201d that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to influence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are significantly better than those produced by the other techniques on almost all of the datasets."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow do noisy and faulty positive pairs in time series contrastive learning affect the quality of learned representations?\n\n**[Question 2] - Why is it interesting and important?**  \nAddressing the issue of noisy and faulty positive pairs in time series contrastive learning is crucial for enhancing the robustness and accuracy of representation learning in various applications, such as healthcare monitoring and anomaly detection. By understanding and mitigating the impact of these detrimental pairs, the research community can improve the performance of downstream tasks, leading to more reliable models in critical domains. This work could pave the way for future research focused on refining contrastive learning techniques, ultimately advancing knowledge in representation learning and enabling practical applications that require high fidelity in time series analysis.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in identifying and distinguishing between beneficial and detrimental positive pairs in time series data, as the noise and faults can be subtle and context-dependent. Naive approaches may fail because they do not account for the specific characteristics of time series data, such as temporal dependencies and noise patterns. Additionally, the complexity of real-world data collection processes introduces variability that complicates the identification of meaningful augmentations. Overcoming these technical obstacles requires sophisticated methods to analyze and filter positive pairs effectively, which is not straightforward.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the assumption that all positive pairs in contrastive learning are beneficial, overlooking the potential for noisy and faulty pairs to degrade model performance. Existing solutions have not adequately addressed the unique challenges posed by time series data, such as the intricate temporal relationships and the impact of noise. Barriers include a lack of comprehensive frameworks for evaluating the quality of positive pairs and insufficient understanding of how these pairs influence learning outcomes. Our approach differs by specifically targeting the identification and mitigation of these detrimental pairs, providing a more nuanced understanding of their effects on time series contrastive learning.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a two-pronged approach: first, we will develop a framework to systematically identify noisy and faulty positive pairs using domain-specific characteristics of time series data; second, we will implement a modified contrastive learning algorithm that incorporates mechanisms to filter out these detrimental pairs. We will utilize datasets such as Sleep-EDF and PTB-XL for our experiments, measuring performance using metrics like representation quality and downstream task accuracy. We expect that our approach will"
            }
        },
        "author_data": {
            "224702d0-b475-4b56-bff2-018a70a409f7": {
                "pk": "224702d0-b475-4b56-bff2-018a70a409f7",
                "name": "Xiang Lan",
                "collaborators": [
                    "Mengling Feng",
                    "Yucheng Ruan",
                    "H. Abdullah",
                    "Dianwen Ng",
                    "Zhaowei Zhu",
                    "Tingting Zhao",
                    "Yangming Guo",
                    "Zhuoyang Xu",
                    "Zhuo Liu",
                    "Siqi Liu",
                    "Han Wang",
                    "Xingzhi Sun",
                    "Yizhi Dong",
                    "Kai He",
                    "Daniel J. Tan",
                    "P. Kojodjojo",
                    "Jihai Zhang",
                    "Xiaoye Qu",
                    "Yu Cheng",
                    "Bryan Hooi",
                    "Jingying Ma",
                    "D. Lim",
                    "Y. Ke",
                    "Nur Nasyitah Mohamed Salim",
                    "Rui Mao",
                    "Qika Lin",
                    "Erik Cambria",
                    "Y. Ruan",
                    "Trung Kien Dang",
                    "J. Weng",
                    "Yi Liu",
                    "Jiongzhou Liu",
                    "Fan Xu",
                    "Cheng He",
                    "Melissa Min-Szu Yao",
                    "Wing P. Chan",
                    "linda Qiao"
                ],
                "domain": [
                    "Contrastive Learning",
                    "Medical Data Analysis",
                    "Federated Learning",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Learning the Unlearned: Mitigating Feature Suppression in Contrastive Learning",
                        "abstract": "Self-Supervised Contrastive Learning has proven effective in deriving high-quality representations from unlabeled data. However, a major challenge that hinders both unimodal and multimodal contrastive learning is feature suppression, a phenomenon where the trained model captures only a limited portion of the information from the input data while overlooking other potentially valuable content. This issue often leads to indistinguishable representations for visually similar but semantically different inputs, adversely affecting downstream task performance, particularly those requiring rigorous semantic comprehension. To address this challenge, we propose a novel model-agnostic Multistage Contrastive Learning (MCL) framework. Unlike standard contrastive learning which inherently captures one single biased feature distribution, MCL progressively learns previously unlearned features through feature-aware negative sampling at each stage, where the negative samples of an anchor are exclusively selected from the cluster it was assigned to in preceding stages. Meanwhile, MCL preserves the previously well-learned features by cross-stage representation integration, integrating features across all stages to form final representations. Our comprehensive evaluation demonstrates MCL's effectiveness and superiority across both unimodal and multimodal contrastive learning, spanning a range of model architectures from ResNet to Vision Transformers (ViT). Remarkably, in tasks where the original CLIP model has shown limitations, MCL dramatically enhances performance, with improvements up to threefold on specific attributes in the recently proposed MMVP benchmark."
                    },
                    {
                        "title": "Language Modeling on Tabular Data: A Survey of Foundations, Techniques and Evolution",
                        "abstract": "Tabular data, a prevalent data type across various domains, presents unique challenges due to its heterogeneous nature and complex structural relationships. Achieving high predictive performance and robustness in tabular data analysis holds significant promise for numerous applications. Influenced by recent advancements in natural language processing, particularly transformer architectures, new methods for tabular data modeling have emerged. Early techniques concentrated on pre-training transformers from scratch, often encountering scalability issues. Subsequently, methods leveraging pre-trained language models like BERT have been developed, which require less data and yield enhanced performance. The recent advent of large language models, such as GPT and LLaMA, has further revolutionized the field, facilitating more advanced and diverse applications with minimal fine-tuning. Despite the growing interest, a comprehensive survey of language modeling techniques for tabular data remains absent. This paper fills this gap by providing a systematic review of the development of language modeling for tabular data, encompassing: (1) a categorization of different tabular data structures and data types; (2) a review of key datasets used in model training and tasks used for evaluation; (3) a summary of modeling techniques including widely-adopted data processing methods, popular architectures, and training objectives; (4) the evolution from adapting traditional Pre-training/Pre-trained language models to the utilization of large language models; (5) an identification of persistent challenges and potential future research directions in language modeling for tabular data analysis. GitHub page associated with this survey is available at: https://github.com/lanxiang1017/Language-Modeling-on-Tabular-Data-Survey.git."
                    },
                    {
                        "title": "The SingHealth Perioperative and Anesthesia Subject Area Registry (PASAR), a large-scale perioperative data mart and registry",
                        "abstract": "Background To enhance perioperative outcomes, a perioperative registry that integrates high-quality real-world data throughout the perioperative period is essential. Singapore General Hospital established the Perioperative and Anesthesia Subject Area Registry (PASAR) to unify data from the preoperative, intraoperative, and postoperative stages. This study presents the methodology employed to create this database. Methods Since 2016, data from surgical patients have been collected from the hospital electronic medical record systems, de-identified, and stored securely in compliance with privacy and data protection laws. As a representative sample, data from initiation in 2016 to December 2022 were collected. Results As of December 2022, PASAR data comprise 26 tables, encompassing 153,312 patient admissions and 168,977 operation sessions. For this period, the median age of the patients was 60.0 years, sex distribution was balanced, and the majority were Chinese. Hypertension and cardiovascular comorbidities were also prevalent. Information including operation type and time, intensive care unit (ICU) length of stay, and 30-day and 1-year mortality rates were collected. Emergency surgeries resulted in longer ICU stays, but shorter operation times than elective surgeries. Conclusions The PASAR provides a comprehensive and automated approach to gathering high-quality perioperative patient data."
                    },
                    {
                        "title": "Medical Intervention Duration Estimation Using Language-enhanced Transformer Encoder with Medical Prompts",
                        "abstract": "In recent years, estimating the duration of medical intervention based on electronic health records (EHRs) has gained signi\ufb01cant attention in the \ufb01led of clinical decision support. However, current models largely focus on structured data, leaving out information from the unstructured clinical free-text data. To address this, we present a novel language-enhanced transformer-based framework, which projects all relevant clinical data modalities (continuous, categorical, binary, and free-text features) into a harmonized language latent space using a pre-trained sentence encoder with the help of medical prompts. The proposed method enables the integration of information from different modalities within the cell transformer encoder and leads to more accurate duration estimation for medical intervention. Our experimental results on both US-based (length of stay in ICU estimation) and Asian (surgical duration prediction) medical datasets demonstrate the effectiveness of our proposed framework, which outperforms tailored baseline approaches and exhibits robustness to data corruption in EHRs."
                    },
                    {
                        "title": "A Survey of Large Language Models for Healthcare: from Data, Technology, and Applications to Accountability and Ethics",
                        "abstract": "The utilization of large language models (LLMs) in the Healthcare domain has generated both excitement and concern due to their ability to effectively respond to freetext queries with certain professional knowledge. This survey outlines the capabilities of the currently developed LLMs for Healthcare and explicates their development process, with the aim of providing an overview of the development roadmap from traditional Pretrained Language Models (PLMs) to LLMs. Specifically, we first explore the potential of LLMs to enhance the efficiency and effectiveness of various Healthcare applications highlighting both the strengths and limitations. Secondly, we conduct a comparison between the previous PLMs and the latest LLMs, as well as comparing various LLMs with each other. Then we summarize related Healthcare training data, training methods, optimization strategies, and usage. Finally, the unique concerns associated with deploying LLMs in Healthcare settings are investigated, particularly regarding fairness, accountability, transparency and ethics. Our survey provide a comprehensive investigation from perspectives of both computer science and Healthcare specialty. Besides the discussion about Healthcare concerns, we supports the computer science community by compiling a collection of open source resources, such as accessible datasets, the latest methodologies, code implementations, and evaluation benchmarks in the Github. Summarily, we contend that a significant paradigm shift is underway, transitioning from PLMs to LLMs. This shift encompasses a move from discriminative AI approaches to generative AI approaches, as well as a shift from model-centered methodologies to data-centered methodologies. Also, we determine that the biggest obstacle of using LLMs in Healthcare are fairness, accountability, transparency and ethics."
                    },
                    {
                        "title": "P-Transformer: A Prompt-based Multimodal Transformer Architecture For Medical Tabular Data",
                        "abstract": "Medical tabular data, abundant in Electronic Health Records (EHRs), is a valuable resource for diverse medical tasks such as risk prediction. While deep learning approaches, particularly transformer-based models, have shown remarkable performance in tabular data prediction, there are still problems remained for existing work to be effectively adapted into medical domain, such as under-utilization of unstructured free-texts, limited exploration of textual information in structured data, and data corruption. To address these issues, we propose P-Transformer, a Prompt-based multimodal Transformer architecture designed specifically for medical tabular data. This framework consists two critical components: a tabular cell embedding generator and a tabular transformer. The former efficiently encodes diverse modalities from both structured and unstructured tabular data into a harmonized language semantic space with the help of pre-trained sentence encoder and medical prompts. The latter integrates cell representations to generate patient embeddings for various medical tasks. In comprehensive experiments on two real-world datasets for three medical tasks, P-Transformer demonstrated the improvements with 10.9%/11.0% on RMSE/MAE, 0.5%/2.2% on RMSE/MAE, and 1.6%/0.8% on BACC/AUROC compared to state-of-the-art (SOTA) baselines in predictability. Notably, the model exhibited strong resilience to data corruption in the structured data, particularly when the corruption rates are high."
                    },
                    {
                        "title": "Federated Learning for Electronic Health Records",
                        "abstract": "In data-driven medical research, multi-center studies have long been preferred over single-center ones due to a single institute sometimes not having enough data to obtain sufficient statistical power for certain hypothesis testings as well as predictive and subgroup studies. The wide adoption of electronic health records (EHRs) has made multi-institutional collaboration much more feasible. However, concerns over infrastructures, regulations, privacy, and data standardization present a challenge to data sharing across healthcare institutions. Federated Learning (FL), which allows multiple sites to collaboratively train a global model without directly sharing data, has become a promising paradigm to break the data isolation. In this study, we surveyed existing works on FL applications in EHRs and evaluated the performance of current state-of-the-art FL algorithms on two EHR machine learning tasks of significant clinical importance on a real world multi-center EHR dataset."
                    },
                    {
                        "title": "Adversarial Domain Adaptation with Correlation-Based Association Networks for Longitudinal Disk Fault Prediction",
                        "abstract": "Disk fault is known to be the key cause of data loss in the modern large-scale data center, which affects the reliability and stability of the server and even the whole IT infrastructure, resulting in high financial cost. Recent works on disk fault prediction demonstrate the ability of machine learning techniques in the early prediction of disk failure. However, two limitations hinder the real-world application of current methods. First, they ignore the data heterogeneity in the data center, where distribution shifts commonly exist across different disk model types that decrease the model's performance. Second, the number of disks of different disk model types varies greatly in the data center, and current methods failed to deliver an acceptable performance over disk model types with few training samples. To address these limitations, we propose adversarial domain adaptation with correlation-based association networks (ADA-CBAN) to both mitigate the distribution shift problem and also to boost the performance on small-scaled data. Extensive experiments prove that our proposed model is effective and achieves new state-of-the-art. In addition, our post model analysis can also reveal important feature interactions and highlight the crucial period before disk faults, where both are useful information in real-world applications."
                    },
                    {
                        "title": "Identification of 27 abnormalities from multi-lead ECG signals: an ensembled SE_ResNet framework with Sign Loss function",
                        "abstract": "Objective. Cardiovascular disease is a major threat to health and one of the primary causes of death globally. The 12-lead ECG is a cheap and commonly accessible tool to identify cardiac abnormalities. Early and accurate diagnosis will allow early treatment and intervention to prevent severe complications of cardiovascular disease. Our objective is to develop an algorithm that automatically identifies 27 ECG abnormalities from 12-lead ECG databases. Approach. Firstly, a series of pre-processing methods were proposed and applied on various data sources in order to mitigate the problem of data divergence. Secondly, we ensembled two SE_ResNet models and one rule-based model to enhance the performance of various ECG abnormalities\u2019 classification. Thirdly, we introduce a Sign Loss to tackle the problem of class imbalance, and thus improve the model's generalizability. Main results. In the PhysioNet/Computing in Cardiology Challenge (2020), our proposed approach achieved a challenge validation score of 0.682, and a full test score of 0.514, placed us 3rd out of 40 in the official ranking. Significance. We proposed an accurate and robust predictive framework that combines deep neural networks and clinical knowledge to automatically classify multiple ECG abnormalities. Our framework is able to identify 27 ECG abnormalities from multi-lead ECG signals regardless of discrepancies in data sources and the imbalance of data labeling. We trained our framework on five datasets and validated it on six datasets from various countries. The outstanding performance demonstrate the effectiveness of our proposed framework."
                    },
                    {
                        "title": "Federated learning: a collaborative effort to achieve better medical imaging models for individual sites that have small labelled datasets.",
                        "abstract": "Despite the overall success of using artificial intelligence (AI) to assist radiologists in performing computer-aided patient diagnosis, it remains challenging to build good models with small datasets at individual sites. Because many medical images do not come with proper labelling for training, this requires radiologists to perform strenuous labelling work and to prepare the dataset for training. Placing such demands on radiologists is unsustainable, given the ever-increasing number of medical images taken each year. We propose an alternative solution using a relatively new learning framework. This framework, called federated learning, allows individual sites to train a global model in a collaborative effort. Federated learning involves aggregating training results from multiple sites to create a global model without directly sharing datasets. This ensures that patient privacy is maintained across sites. Furthermore, the added supervision obtained from the results of partnering sites improves the global model's overall detection abilities. This alleviates the issue of insufficient supervision when training AI models with small datasets. Lastly, we also address the major challenges of adopting federated learning."
                    },
                    {
                        "title": "Intra-Inter Subject Self-supervised Learning for Multivariate Cardiac Signals",
                        "abstract": "Learning information-rich and generalizable representations effectively from unlabeled multivariate cardiac signals to identify abnormal heart rhythms (cardiac arrhythmias) is valuable in real-world clinical settings but often challenging due to its complex temporal dynamics. Cardiac arrhythmias can vary significantly in temporal patterns even for the same patient (i.e., intra subject difference). Meanwhile, the same type of cardiac arrhythmia can show different temporal patterns among different patients due to different cardiac structures (i.e., inter subject difference). In this paper, we address the challenges by proposing an Intra-Inter Subject Self-Supervised Learning (ISL) model that is customized for multivariate cardiac signals. Our proposed ISL model integrates medical knowledge into self-supervision to effectively learn from intra-inter subject differences. In intra subject self-supervision, ISL model first extracts heartbeat-level features from each subject using a channel-wise attentional CNN-RNN encoder. Then a stationarity test module is employed to capture the temporal dependencies between heartbeats. In inter subject self-supervision, we design a set of data augmentations according to the clinical characteristics of cardiac signals and perform contrastive learning among subjects to learn distinctive representations for various types of patients. Extensive experiments on three real-world datasets were conducted. In a semi-supervised transfer learning scenario, our pre-trained ISL model leads about 10% improvement over supervised training when only 1% labeled data is available, suggesting strong generalizability and robustness of the model."
                    },
                    {
                        "title": "Identification of 27 cardiac abnormalities from multi-lead ECG signals: An ensembled Se-ResNet framework with Sign Loss function",
                        "abstract": "Cardiovascular disease is a major threat to health and one of the primary causes of death globally, where the cardiac abnormality is the most common type of cardiovascular disease. The early and accurate diagnosis of cardiac abnormalities will allow early treatment and intervention to prevent further progression of the disease. In accordance with the PhysioNet/Computing in Cardiology Challenge 2020, our objective is to develop an algorithm that automatically identifies 27 types of cardiac abnormality from 12-lead ECG recordings. We proposed an accurate and robust predictive framework that combines both deep neural networks and rule-based models to automatically classify multiple cardiac abnormalities. Our framework is able to identify 27 types of cardiac abnormalities presented in 12-lead ECG data. And our proposed Sign Loss makes the framework able to tackle the imbalanced dataset. Meanwhile, our framework was trained and verified on 6 datasets from different countries, and our proposed multi-source preprocessing methods are able to reduce the discrepancies of ECG datasets from different sources."
                    },
                    {
                        "title": "Classification of Cardiac Abnormalities From ECG Signals Using SE-ResNet",
                        "abstract": "In PhysioNet/Computing in Cardiology Challenge 2020, we developed an ensembled model based on SE-ResNet to classify cardiac abnormalities from 12-lead electrocardiogram (ECG) signals. We employed two residual neural network modules with squeeze-and-excitation blocks to learn from the first 10-second and 30-second segments of the signals. We used external open-source data for validation and fine-tuning during the model development phase. We designed a multi-label loss to emphasize the impact of wrong predictions during training. We built a rule-based bradycardia model based on clinical knowledge to correct the output. All these efforts helped us to achieve a robust classification performance. Our final model achieved a challenge validation score of 0.682 and a full test score of 0.514, placing our team HeartBeats 3rd out of 41 in the official ranking. We believed that our model has a great potential to be applied in the actual clinical practice, and planned to further extend the research after the challenge."
                    }
                ]
            },
            "22a3be5f-120f-420d-b10b-1d6054a1ce21": {
                "pk": "22a3be5f-120f-420d-b10b-1d6054a1ce21",
                "name": "Hanshu Yan",
                "collaborators": [
                    "Jiashi Feng",
                    "J. Liew",
                    "Vincent Y. F. Tan",
                    "Jiachun Pan",
                    "Daquan Zhou",
                    "Jianfeng Zhang",
                    "Zhongcong Xu",
                    "Weimin Wang",
                    "Jingfeng Zhang",
                    "Masashi Sugiyama",
                    "Yujun Shi",
                    "Xingchao Liu",
                    "Qiang Liu",
                    "Lianghui Zhu",
                    "Zilong Huang",
                    "Bencheng Liao",
                    "Xinggang Wang",
                    "Jiawei Liu",
                    "Zhijie Lin",
                    "Jiangqiao Yan",
                    "Shuo Chen",
                    "Chetwin Low",
                    "Tuyen Hoang",
                    "Jie Wu",
                    "Jiannan Huang",
                    "Yuyang Yin",
                    "Yao Zhao",
                    "Yunchao Wei",
                    "Long Mai",
                    "Shanchuan Lin",
                    "Jia-Wei Liu",
                    "Chenxu Zhang",
                    "Mike Zheng Shou",
                    "Weiwei Lv",
                    "Yizhe Zhu",
                    "Pan Zhou",
                    "Xiao-Tong Yuan",
                    "Shuicheng Yan",
                    "Gang Niu",
                    "Chi Shu",
                    "Wei Zheng",
                    "Kan Lin",
                    "Anne James",
                    "S. Selvarajan",
                    "C. Lim",
                    "Zhiwei Huang",
                    "Haiyun He"
                ],
                "domain": [
                    "Image Editing",
                    "Generative Models",
                    "Video Generation",
                    "Adversarial Robustness"
                ],
                "publications": [
                    {
                        "title": "LightningDrag: Lightning Fast and Accurate Drag-based Image Editing Emerging from Videos",
                        "abstract": "Accuracy and speed are critical in image editing tasks. Pan et al. introduced a drag-based image editing framework that achieves pixel-level control using Generative Adversarial Networks (GANs). A flurry of subsequent studies enhanced this framework's generality by leveraging large-scale diffusion models. However, these methods often suffer from inordinately long processing times (exceeding 1 minute per edit) and low success rates. Addressing these issues head on, we present LightningDrag, a rapid approach enabling high quality drag-based image editing in ~1 second. Unlike most previous methods, we redefine drag-based editing as a conditional generation task, eliminating the need for time-consuming latent optimization or gradient-based guidance during inference. In addition, the design of our pipeline allows us to train our model on large-scale paired video frames, which contain rich motion information such as object translations, changing poses and orientations, zooming in and out, etc. By learning from videos, our approach can significantly outperform previous methods in terms of accuracy and consistency. Despite being trained solely on videos, our model generalizes well to perform local shape deformations not presented in the training data (e.g., lengthening of hair, twisting rainbows, etc.). Extensive qualitative and quantitative evaluations on benchmark datasets corroborate the superiority of our approach. The code and model will be released at https://github.com/magic-research/LightningDrag."
                    },
                    {
                        "title": "PeRFlow: Piecewise Rectified Flow as Universal Plug-and-Play Accelerator",
                        "abstract": "We present Piecewise Rectified Flow (PeRFlow), a flow-based method for accelerating diffusion models. PeRFlow divides the sampling process of generative flows into several time windows and straightens the trajectories in each interval via the reflow operation, thereby approaching piecewise linear flows. PeRFlow achieves superior performance in a few-step generation. Moreover, through dedicated parameterizations, the PeRFlow models inherit knowledge from the pretrained diffusion models. Thus, the training converges fast and the obtained models show advantageous transfer ability, serving as universal plug-and-play accelerators that are compatible with various workflows based on the pre-trained diffusion models. Codes for training and inference are publicly released. https://github.com/magic-research/piecewise-rectified-flow"
                    },
                    {
                        "title": "DiG: Scalable and Efficient Diffusion Models with Gated Linear Attention",
                        "abstract": "Diffusion models with large-scale pre-training have achieved significant success in the field of visual content generation, particularly exemplified by Diffusion Transformers (DiT). However, DiT models have faced challenges with scalability and quadratic complexity efficiency. In this paper, we aim to leverage the long sequence modeling capability of Gated Linear Attention (GLA) Transformers, expanding its applicability to diffusion models. We introduce Diffusion Gated Linear Attention Transformers (DiG), a simple, adoptable solution with minimal parameter overhead, following the DiT design, but offering superior efficiency and effectiveness. In addition to better performance than DiT, DiG-S/2 exhibits $2.5\\times$ higher training speed than DiT-S/2 and saves $75.7\\%$ GPU memory at a resolution of $1792 \\times 1792$. Moreover, we analyze the scalability of DiG across a variety of computational complexity. DiG models, with increased depth/width or augmentation of input tokens, consistently exhibit decreasing FID. We further compare DiG with other subquadratic-time diffusion models. With the same model size, DiG-XL/2 is $4.2\\times$ faster than the recent Mamba-based diffusion model at a $1024$ resolution, and is $1.8\\times$ faster than DiT with CUDA-optimized FlashAttention-2 under the $2048$ resolution. All these results demonstrate its superior efficiency among the latest diffusion models. Code is released at https://github.com/hustvl/DiG."
                    },
                    {
                        "title": "MagicVideo-V2: Multi-Stage High-Aesthetic Video Generation",
                        "abstract": "The growing demand for high-fidelity video generation from textual descriptions has catalyzed significant research in this field. In this work, we introduce MagicVideo-V2 that integrates the text-to-image model, video motion generator, reference image embedding module and frame interpolation module into an end-to-end video generation pipeline. Benefiting from these architecture designs, MagicVideo-V2 can generate an aesthetically pleasing, high-resolution video with remarkable fidelity and smoothness. It demonstrates superior performance over leading Text-to-Video systems such as Runway, Pika 1.0, Morph, Moon Valley and Stable Video Diffusion model via user evaluation at large scale."
                    },
                    {
                        "title": "ClassDiffusion: More Aligned Personalization Tuning with Explicit Class Guidance",
                        "abstract": "Recent text-to-image customization works have been proven successful in generating images of given concepts by fine-tuning the diffusion models on a few examples. However, these methods tend to overfit the concepts, resulting in failure to create the concept under multiple conditions (e.g. headphone is missing when generating adog wearing a headphone'). Interestingly, we notice that the base model before fine-tuning exhibits the capability to compose the base concept with other elements (e.g. a dog wearing a headphone) implying that the compositional ability only disappears after personalization tuning. Inspired by this observation, we present ClassDiffusion, a simple technique that leverages a semantic preservation loss to explicitly regulate the concept space when learning the new concept. Despite its simplicity, this helps avoid semantic drift when fine-tuning on the target concepts. Extensive qualitative and quantitative experiments demonstrate that the use of semantic preservation loss effectively improves the compositional abilities of the fine-tune models. In response to the ineffective evaluation of CLIP-T metrics, we introduce BLIP2-T metric, a more equitable and effective evaluation metric for this particular domain. We also provide in-depth empirical study and theoretical analysis to better understand the role of the proposed loss. Lastly, we also extend our ClassDiffusion to personalized video generation, demonstrating its flexibility."
                    },
                    {
                        "title": "MagicProp: Diffusion-based Video Editing via Motion-aware Appearance Propagation",
                        "abstract": "This paper addresses the issue of modifying the visual appearance of videos while preserving their motion. A novel framework, named MagicProp, is proposed, which disentangles the video editing process into two stages: appearance editing and motion-aware appearance propagation. In the first stage, MagicProp selects a single frame from the input video and applies image-editing techniques to modify the content and/or style of the frame. The flexibility of these techniques enables the editing of arbitrary regions within the frame. In the second stage, MagicProp employs the edited frame as an appearance reference and generates the remaining frames using an autoregressive rendering approach. To achieve this, a diffusion-based conditional generation model, called PropDPM, is developed, which synthesizes the target frame by conditioning on the reference appearance, the target motion, and its previous appearance. The autoregressive editing approach ensures temporal consistency in the resulting videos. Overall, MagicProp combines the flexibility of image-editing techniques with the superior temporal consistency of autoregressive modeling, enabling flexible editing of object types and aesthetic styles in arbitrary regions of input videos while maintaining good temporal consistency across frames. Extensive experiments in various video editing scenarios demonstrate the effectiveness of MagicProp."
                    },
                    {
                        "title": "MagicEdit: High-Fidelity and Temporally Coherent Video Editing",
                        "abstract": "In this report, we present MagicEdit, a surprisingly simple yet effective solution to the text-guided video editing task. We found that high-fidelity and temporally coherent video-to-video translation can be achieved by explicitly disentangling the learning of content, structure and motion signals during training. This is in contradict to most existing methods which attempt to jointly model both the appearance and temporal representation within a single framework, which we argue, would lead to degradation in per-frame quality. Despite its simplicity, we show that MagicEdit supports various downstream video editing tasks, including video stylization, local editing, video-MagicMix and video outpainting."
                    },
                    {
                        "title": "AdjointDPM: Adjoint Sensitivity Method for Gradient Backpropagation of Diffusion Probabilistic Models",
                        "abstract": "Existing customization methods require access to multiple reference examples to align pre-trained diffusion probabilistic models (DPMs) with user-provided concepts. This paper aims to address the challenge of DPM customization when the only available supervision is a differentiable metric defined on the generated contents. Since the sampling procedure of DPMs involves recursive calls to the denoising UNet, na\\\"ive gradient backpropagation requires storing the intermediate states of all iterations, resulting in extremely high memory consumption. To overcome this issue, we propose a novel method AdjointDPM, which first generates new samples from diffusion models by solving the corresponding probability-flow ODEs. It then uses the adjoint sensitivity method to backpropagate the gradients of the loss to the models' parameters (including conditioning signals, network weights, and initial noises) by solving another augmented ODE. To reduce numerical errors in both the forward generation and gradient backpropagation processes, we further reparameterize the probability-flow ODE and augmented ODE as simple non-stiff ODEs using exponential integration. Finally, we demonstrate the effectiveness of AdjointDPM on three interesting tasks: converting visual effects into identification text embeddings, finetuning DPMs for specific types of stylization, and optimizing initial noise to generate adversarial samples for security auditing."
                    },
                    {
                        "title": "MagicAvatar: Multimodal Avatar Generation and Animation",
                        "abstract": "This report presents MagicAvatar, a framework for multimodal video generation and animation of human avatars. Unlike most existing methods that generate avatar-centric videos directly from multimodal inputs (e.g., text prompts), MagicAvatar explicitly disentangles avatar video generation into two stages: (1) multimodal-to-motion and (2) motion-to-video generation. The first stage translates the multimodal inputs into motion/ control signals (e.g., human pose, depth, DensePose); while the second stage generates avatar-centric video guided by these motion signals. Additionally, MagicAvatar supports avatar animation by simply providing a few images of the target person. This capability enables the animation of the provided human identity according to the specific motion derived from the first stage. We demonstrate the flexibility of MagicAvatar through various applications, including text-guided and video-guided avatar generation, as well as multimodal avatar animation."
                    },
                    {
                        "title": "MagicAnimate: Temporally Consistent Human Image Animation using Diffusion Model",
                        "abstract": "This paper studies the human image animation task, which aims to generate a video of a certain reference iden-tity following a particular motion sequence. Existing an-imation works typically employ the frame-warping technique to animate the reference image towards the target motion. Despite achieving reasonable results, these approaches face challenges in maintaining temporal consistency throughout the animation due to the lack of temporal modeling and poor preservation of reference identity. In this work, we introduce Magic/snimate, a diffusion-based framework that aims at enhancing temporal consistency, preserving reference image faithfully, and improving animation fidelity. To achieve this, we first develop a video diffusion model to encode temporal information. Second, to maintain the appearance coherence across frames, we introduce a novel appearance encoder to retain the intricate details of the reference image. Leveraging these two inno-vations, we further employ a simple video fusion technique to encourage smooth transitions for long video animation. Empirical results demonstrate the superiority of our method over baseline approaches on two benchmarks. Notably, our approach outperforms the strongest baseline by over 38% in terms of video fidelity on the challenging TikTok dancing dataset. Code and model will be made available at https://showlab.github.io/magicanimate."
                    },
                    {
                        "title": "Towards Accurate Guided Diffusion Sampling through Symplectic Adjoint Method",
                        "abstract": "Training-free guided sampling in diffusion models leverages off-the-shelf pre-trained networks, such as an aesthetic evaluation model, to guide the generation process. Current training-free guided sampling algorithms obtain the guidance energy function based on a one-step estimate of the clean image. However, since the off-the-shelf pre-trained networks are trained on clean images, the one-step estimation procedure of the clean image may be inaccurate, especially in the early stages of the generation process in diffusion models. This causes the guidance in the early time steps to be inaccurate. To overcome this problem, we propose Symplectic Adjoint Guidance (SAG), which calculates the gradient guidance in two inner stages. Firstly, SAG estimates the clean image via $n$ function calls, where $n$ serves as a flexible hyperparameter that can be tailored to meet specific image quality requirements. Secondly, SAG uses the symplectic adjoint method to obtain the gradients accurately and efficiently in terms of the memory requirements. Extensive experiments demonstrate that SAG generates images with higher qualities compared to the baselines in both guided image and video generation tasks."
                    },
                    {
                        "title": "Towards Adversarial Robustness of Deep Vision Algorithms",
                        "abstract": "Deep learning methods have achieved great success in solving computer vision tasks, and they have been widely utilized in artificially intelligent systems for image processing, analysis, and understanding. However, deep neural networks have been shown to be vulnerable to adversarial perturbations in input data. The security issues of deep neural networks have thus come to the fore. It is imperative to study the adversarial robustness of deep vision algorithms comprehensively. This talk focuses on the adversarial robustness of image classification models and image denoisers. We will discuss the robustness of deep vision algorithms from three perspectives: 1) robustness evaluation (we propose the ObsAtk to evaluate the robustness of denoisers), 2) robustness improvement (HAT, TisODE, and CIFS are developed to robustify vision models), and 3) the connection between adversarial robustness and generalization capability to new domains (we find that adversarially robust denoisers can deal with unseen types of real-world noise)."
                    },
                    {
                        "title": "MagicMix: Semantic Mixing with Diffusion Models",
                        "abstract": "Have you ever imagined what a corgi-alike coffee machine or a tiger-alike rabbit would look like? In this work, we attempt to answer these questions by exploring a new task called semantic mixing, aiming at blending two different semantics to create a new concept (e.g., corgi + coffee machine -->corgi-alike coffee machine). Unlike style transfer, where an image is stylized according to the reference style without changing the image content, semantic blending mixes two different concepts in a semantic manner to synthesize a novel concept while preserving the spatial layout and geometry. To this end, we present MagicMix, a simple yet effective solution based on pre-trained text-conditioned diffusion models. Motivated by the progressive generation property of diffusion models where layout/shape emerges at early denoising steps while semantically meaningful details appear at later steps during the denoising process, our method first obtains a coarse layout (either by corrupting an image or denoising from a pure Gaussian noise given a text prompt), followed by injection of conditional prompt for semantic mixing. Our method does not require any spatial mask or re-training, yet is able to synthesize novel objects with high fidelity. To improve the mixing quality, we further devise two simple strategies to provide better control and flexibility over the synthesized content. With our method, we present our results over diverse downstream applications, including semantic style transfer, novel object synthesis, breed mixing, and concept removal, demonstrating the flexibility of our method. More results can be found on the project page https://magicmix.github.io"
                    },
                    {
                        "title": "MagicVideo: Efficient Video Generation With Latent Diffusion Models",
                        "abstract": "We present an efficient text-to-video generation framework based on latent diffusion models, termed MagicVideo. MagicVideo can generate smooth video clips that are concordant with the given text descriptions. Due to a novel and efficient 3D U-Net design and modeling video distributions in a low-dimensional space, MagicVideo can synthesize video clips with 256x256 spatial resolution on a single GPU card, which takes around 64x fewer computations than the Video Diffusion Models (VDM) in terms of FLOPs. In specific, unlike existing works that directly train video models in the RGB space, we use a pre-trained VAE to map video clips into a low-dimensional latent space and learn the distribution of videos' latent codes via a diffusion model. Besides, we introduce two new designs to adapt the U-Net denoiser trained on image tasks to video data: a frame-wise lightweight adaptor for the image-to-video distribution adjustment and a directed temporal attention module to capture temporal dependencies across frames. Thus, we can exploit the informative weights of convolution operators from a text-to-image model for accelerating video training. To ameliorate the pixel dithering in the generated videos, we also propose a novel VideoVAE auto-encoder for better RGB reconstruction. We conduct extensive experiments and demonstrate that MagicVideo can generate high-quality video clips with either realistic or imaginary content. Refer to \\url{https://magicvideo.github.io/#} for more examples."
                    },
                    {
                        "title": "Towards Understanding Why Lookahead Generalizes Better Than SGD and Beyond (Supplementary File)",
                        "abstract": "This supplementary document contains the technical proofs of convergence results and some additional experimental results of the NeurIPS\u201921 submission entitled \u201cTowards Understanding Why Lookahead Generalizes Better Than SGD and Be-yond\u201d. It is structured as follows. Appendix A \ufb01rst discusses the limitations of this work. In Appendix B, we provides more experimental results and details, including the robustness investigation of SLRLA to regularization parameter. Next, Appendix C presents several auxiliary lemmas which will be used for subsequent analysis. Then Appendix D gives the proofs of the main results in Sec. 4, including Theorem 1 \u223c 4 and Corollary 1 which analyze optimization error, generalization error and excess risk error of vanilla lookahead algorithm. Finally, Appendix E provides the proofs of the results in Sec. 5, including Theorems 5 and 6 which analyze the optimization error, generalization error and excess risk error of the proposed SLRLA."
                    },
                    {
                        "title": "Towards Adversarially Robust Deep Image Denoising",
                        "abstract": "This work systematically investigates the adversarial robustness of deep image denoisers (DIDs), i.e, how well DIDs can recover the ground truth from noisy observations degraded by adversarial perturbations. Firstly, to evaluate DIDs\u2019 robustness, we propose a novel adversarial attack, namely Observation-based Zero-mean Attack (OBSATK), to craft adversarial zero-mean perturbations on given noisy images. We find that existing DIDs are vulnerable to the adversarial noise generated by OBSATK. Secondly, to robustify DIDs, we pro- pose an adversarial training strategy, hybrid adversarial training (HAT), that jointly trains DIDs with adversarial and non-adversarial noisy data to ensure that the reconstruction quality is high and the denoisers around non-adversarial data are locally smooth. The resultant DIDs can effectively remove various types of synthetic and adversarial noise. We also uncover that the robustness of DIDs benefits their generalization capability on unseen real-world noise. Indeed, HAT-trained DIDs can recover high-quality clean images from real-world noise even without training on real noisy data. Extensive experiments on benchmark datasets, including Set68, PolyU, and SIDD, corroborate the effectiveness of OBSATK and HAT."
                    },
                    {
                        "title": "CIFS: Improving Adversarial Robustness of CNNs via Channel-wise Importance-based Feature Selection",
                        "abstract": "We investigate the adversarial robustness of CNNs from the perspective of channel-wise activations. By comparing \\textit{non-robust} (normally trained) and \\textit{robustified} (adversarially trained) models, we observe that adversarial training (AT) robustifies CNNs by aligning the channel-wise activations of adversarial data with those of their natural counterparts. However, the channels that are \\textit{negatively-relevant} (NR) to predictions are still over-activated when processing adversarial data. Besides, we also observe that AT does not result in similar robustness for all classes. For the robust classes, channels with larger activation magnitudes are usually more \\textit{positively-relevant} (PR) to predictions, but this alignment does not hold for the non-robust classes. Given these observations, we hypothesize that suppressing NR channels and aligning PR ones with their relevances further enhances the robustness of CNNs under AT. To examine this hypothesis, we introduce a novel mechanism, i.e., \\underline{C}hannel-wise \\underline{I}mportance-based \\underline{F}eature \\underline{S}election (CIFS). The CIFS manipulates channels' activations of certain layers by generating non-negative multipliers to these channels based on their relevances to predictions. Extensive experiments on benchmark datasets including CIFAR10 and SVHN clearly verify the hypothesis and CIFS's effectiveness of robustifying CNNs. \\url{https://github.com/HanshuYAN/CIFS}"
                    },
                    {
                        "title": "Deep Learning-Guided Fiberoptic Raman Spectroscopy Enables Real-Time In Vivo Diagnosis and Assessment of Nasopharyngeal Carcinoma and Post-treatment Efficacy during Endoscopy.",
                        "abstract": "In this work, we develop a deep learning-guided fiberoptic Raman diagnostic platform to assess its ability of real-time in vivo nasopharyngeal carcinoma (NPC) diagnosis and post-treatment follow-up of NPC patients. The robust Raman diagnostic platform is established using innovative multi-layer Raman-specified convolutional neural networks (RS-CNN) together with simultaneous fingerprint and high-wavenumber spectra acquired within sub-seconds using a fiberoptic Raman endoscopy system. We have acquired a total of 15,354 FP/HW in vivo Raman spectra (control: 1761; NPC: 4147; and post-treatment (PT): 9446) from 888 tissue sites of 418 subjects (healthy control: 85; NPC: 82; and PT: 251) during endoscopic examination. The optimized RS-CNN model provides an overall diagnostic accuracy of 82.09% (sensitivity of 92.18% and specificity of 73.99%) for identifying NPC from control and post-treatment patients, which is superior to the best diagnosis performance (accuracy of 73.57%; sensitivity of 89.74%; and specificity of 58.10%) using partial-least-squares linear-discriminate-analysis, proving the robustness and high spectral information sensitiveness of the RS-CNN model developed. We further investigate the saliency map of the best RS-CNN models using the correctly predicted Raman spectra. The specific Raman signatures that are related to the cancer-associated biomolecular variations (e.g., collagens, lipids, and nucleic acids) are uncovered in the map, validating the diagnostic capability of RS-CNN models to correlate with biomolecular signatures. Deep learning-based Raman spectroscopy is a powerful diagnostic tool for rapid screening and surveillance of NPC patients and can also be deployed for longitudinal follow-up monitoring of post-treatment NPC patients to detect early cancer recurrences in the head and neck."
                    },
                    {
                        "title": "Information-Theoretic Generalization Bounds for Iterative Semi-Supervised Learning",
                        "abstract": "We consider iterative semi-supervised learning (SSL) algorithms that iteratively generate pseudo-labels for a large amount unlabelled data to progressively re\ufb01ne the model parameters. In particular, we seek to understand the behaviour of the generalization error of iterative SSL algorithms using information-theoretic principles. To obtain bounds that are amenable to numerical evaluation, we \ufb01rst work with a simple model\u2014namely, the binary Gaussian mixture model. Our theoretical results suggest that when the class conditional variances are not too large, the upper bound on the generalization error decreases monotonically with the number of iterations, but quickly saturates. The theoretical results on the simple model are corroborated by extensive experiments on several benchmark datasets such as the MNIST and CIFAR datasets in which we notice that the generalization error improves after several pseudo-labelling iterations, but saturates afterwards."
                    }
                ]
            },
            "0df62f45-b8a1-4006-84bd-f5b1e7141dcd": {
                "pk": "0df62f45-b8a1-4006-84bd-f5b1e7141dcd",
                "name": "Shenda Hong",
                "collaborators": [
                    "Chenxi Sun",
                    "Hongyan Li",
                    "Moxian Song",
                    "D. Cai",
                    "Yaliang Li",
                    "Shijia Geng",
                    "Guodong Wei",
                    "Zhaoji Fu",
                    "Wen-Rang Zhang",
                    "B. Zhang",
                    "Jun Yu Li",
                    "Che Liu",
                    "Sibo Cheng",
                    "Rossella Arcucci",
                    "Haoyang Hong",
                    "J. Jing",
                    "Wendong Ge",
                    "Marta Fernandes",
                    "Zhen Lin",
                    "Chaoqi Yang",
                    "Sungtae An",
                    "A. Struck",
                    "A. Herlopian",
                    "I. Karakis",
                    "J. Halford",
                    "M. Ng",
                    "Emily L. Johnson",
                    "Brian L. Appavu",
                    "R. Sarkis",
                    "Gamaleldin M. Osman",
                    "P. Kaplan",
                    "M. Dhakar",
                    "L. Jayagopal",
                    "Zubeda B. Sheikh",
                    "O. Taraschenko",
                    "Sarah E. Schmitt",
                    "Hiba A. Haider",
                    "Jennifer A. Kim",
                    "Christa B. Swisher",
                    "N. Gaspard",
                    "M. Cervenka",
                    "A. R. Rodriguez Ruiz",
                    "Jong Woo Lee",
                    "Mohammad Tabaeizadeh",
                    "E. Gilmore",
                    "Kristen J. Nordstrom",
                    "J. Yoo",
                    "Manisha G Holmes",
                    "S. Herman",
                    "Jennifer A. Williams",
                    "J. Pathmanathan",
                    "F\u00e1bio A. Nascimento",
                    "Ziwei Fan",
                    "S. Nasiri",
                    "M. Shafi",
                    "S. Cash",
                    "D. Hoch",
                    "A. Cole",
                    "E. Rosenthal",
                    "Sahar F. Zafar",
                    "Jimeng Sun",
                    "M. Westover",
                    "Deyun Zhang",
                    "Yang Zhou",
                    "Wei-Zhao Xu",
                    "Kai Wang",
                    "Jie Yu",
                    "Q. Zhu",
                    "Yongkui Li",
                    "Yonghong Zhao",
                    "Xing-rong Chen",
                    "Rui Zhang",
                    "Rongbo Zhou",
                    "E. Yanqi",
                    "Sumei Fan",
                    "Qinghao Zhao",
                    "Chuandong Cheng",
                    "Nan Peng",
                    "Liangliang Zhang",
                    "Linlin Zheng",
                    "Jianjun Chu",
                    "Hongbin Xu",
                    "C. Tan",
                    "Jian Liu",
                    "Hua-Yue Tao",
                    "Tong Liu",
                    "Kangyin Chen",
                    "Chenyang Jiang",
                    "Xingpeng Liu",
                    "Zhenjie Wang",
                    "Hanmo Yang",
                    "Derun Can",
                    "Xinxin Di",
                    "Yuxi Zhou"
                ],
                "domain": [
                    "Time Series Analysis",
                    "Deep Learning",
                    "Electrocardiogram",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "A Systematic Review of Echo State Networks From Design to Application",
                        "abstract": "A recurrent neural network (RNN) has demonstrated its outstanding ability in sequence tasks and has achieved state of the art in many applications, such as industrial and medical. An echo state network (ESN) is a simple type of RNN and has emerged in the last decade as an alternative to gradient descent training-based RNN. The ESN is practical, conceptually simple, and easy to implement with a strong theoretical ground. It can avoid nonconverging and computationally expensive issues in gradient descent RNN methods. Since the ESN was put forward in 2002, abundant existing works have promoted the progress of ESN, and the recently introduced deep ESN opened the way to uniting the merits of deep learning and reservoir computing. Besides, the combinations of ESNs with other machine learning models have also overperformed baselines in some applications. However, the apparent simplicity of ESNs can sometimes be deceptive. Successfully applying ESNs needs some experience. Thus, we reviewed over 300 related papers and provided a systematic overview for the first time. In this article, we categorize the related methods into classical ESN, DeepESN, and combination. Then, we analyze them from the perspective of network designs and specific applications. Finally, we discuss the challenges and opportunities by proposing open problems and future work."
                    },
                    {
                        "title": "Review of Data-centric Time Series Analysis from Sample, Feature, and Period",
                        "abstract": "Data is essential to performing time series analysis utilizing machine learning approaches, whether for classic models or today's large language models. A good time-series dataset is advantageous for the model's accuracy, robustness, and convergence, as well as task outcomes and costs. The emergence of data-centric AI represents a shift in the landscape from model refinement to prioritizing data quality. Even though time-series data processing methods frequently come up in a wide range of research fields, it hasn't been well investigated as a specific topic. To fill the gap, in this paper, we systematically review different data-centric methods in time series analysis, covering a wide range of research topics. Based on the time-series data characteristics at sample, feature, and period, we propose a taxonomy for the reviewed data selection methods. In addition to discussing and summarizing their characteristics, benefits, and drawbacks targeting time-series data, we also introduce the challenges and opportunities by proposing recommendations, open problems, and possible research topics."
                    },
                    {
                        "title": "A Ranking-Based Cross-Entropy Loss for Early Classification of Time Series",
                        "abstract": "Early classification tasks aim to classify time series before observing full data. It is critical in time-sensitive applications such as early sepsis diagnosis in the intensive care unit (ICU). Early diagnosis can provide more opportunities for doctors to rescue lives. However, there are two conflicting goals in the early classification task\u2014accuracy and earliness. Most existing methods try to find a balance between them by weighing one goal against the other. But we argue that a powerful early classifier should always make highly accurate predictions at any moment. The main obstacle is that the key features suitable for classification are not obvious in the early stage, resulting in the excessive overlap of time series distributions in different time stages. The indistinguishable distributions make it difficult for classifiers to recognize. To solve this problem, this article proposes a novel ranking-based cross-entropy (RCE) loss to jointly learn the feature of classes and the order of earliness from time series data. In this way, RCE can help classifier to generate probability distributions of time series in different stages with more distinguishable boundary. Thus, the classification accuracy at each time step is finally improved. Besides, for the applicability of the method, we also accelerate the training process by focusing the learning process on high-ranking samples. Experiments on three real-world datasets show that our method can perform classification more accurately than all baselines at all moments."
                    },
                    {
                        "title": "Frozen Language Model Helps ECG Zero-Shot Learning",
                        "abstract": "The electrocardiogram (ECG) is one of the most commonly used non-invasive, convenient medical monitoring tools that assist in the clinical diagnosis of heart diseases. Recently, deep learning (DL) techniques, particularly self-supervised learning (SSL), have demonstrated great potential in the classification of ECG. SSL pre-training has achieved competitive performance with only a small amount of annotated data after fine-tuning. However, current SSL methods rely on the availability of annotated data and are unable to predict labels not existing in fine-tuning datasets. To address this challenge, we propose Multimodal ECG-Text Self-supervised pre-training (METS), the first work to utilize the auto-generated clinical reports to guide ECG SSL pre-training. We use a trainable ECG encoder and a frozen language model to embed paired ECG and automatically machine-generated clinical reports separately. The SSL aims to maximize the similarity between paired ECG and auto-generated report while minimize the similarity between ECG and other reports. In downstream classification tasks, METS achieves around 10% improvement in performance without using any annotated data via zero-shot classification, compared to other supervised and SSL baselines that rely on annotated data. Furthermore, METS achieves the highest recall and F1 scores on the MIT-BIH dataset, despite MIT-BIH containing different classes of ECG compared to the pre-trained dataset. The extensive experiments have demonstrated the advantages of using ECG-Text multimodal self-supervised learning in terms of generalizability, effectiveness, and efficiency."
                    },
                    {
                        "title": "Curricular and Cyclical Loss for Time Series Learning Strategy",
                        "abstract": "Time series widely exists in real-world applications and many deep learning models have performed well on it. Current research has shown the importance of learning strategy for models, suggesting that the benefit is the order and size of learning samples. However, no effective strategy has been proposed for time series due to its abstract and dynamic construction. Meanwhile, the existing one-shot tasks and continuous tasks for time series necessitate distinct learning processes and mechanisms. No all-purpose approach has been suggested. In this work, we propose a novel Curricular and CyclicaL loss (CRUCIAL) to learn time series for the first time. It is model- and task-agnostic and can be plugged on top of the original loss with no extra procedure. CRUCIAL has two characteristics: It can arrange an easy-to-hard learning order by dynamically determining the sample contribution and modulating the loss amplitude; It can manage a cyclically changed dataset and achieve an adaptive cycle by correlating the loss distribution and the selection probability. We prove that compared with monotonous size, cyclical size can reduce expected error. Experiments on 3 kinds of tasks and 5 real-world datasets show the benefits of CRUCIAL for most deep learning models when learning time series."
                    },
                    {
                        "title": "simpleNomo: A Python Package of Making Nomograms for Visualizable Calculation of Logistic Regression Models",
                        "abstract": "Background Logistic regression models are widely used in clinical prediction, but their application in resource-poor settings or areas without internet access can be challenging. Nomograms can serve as a useful visualization tool to speed up the calculation procedure, but existing nomogram generators often require the input of raw data, inhibiting the transformation of established logistic regression models that only provide coefficients. Developing a tool that can generate nomograms directly from logistic regression coefficients would greatly increase usability and facilitate the translation of research findings into patient care. Methods We designed and developed simpleNomo, an open-source Python toolbox that enables the construction of nomograms for logistic regression models. Uniquely, simpleNomo allows for the creation of nomograms using only the coefficients of the model. Further, we also devoloped an online website for nomogram generation. Results simpleNomo properly maintains the predictive ability of the original logistic regression model and easy to follow. simpleNomo is compatible with Python 3 and can be installed through Python Package Index (PyPI) or https://github.com/Hhy096/nomogram Conclusion This paper presents simpleNomo, an open-source Python toolbox for generating nomograms for logistic regression models. It facilitates the process of transferring established logistic regression models to nomograms and can further convert more existing works into practical use."
                    },
                    {
                        "title": "Development of Expert-Level Classification of Seizures and Rhythmic and Periodic Patterns During EEG Interpretation",
                        "abstract": "Background and Objectives Seizures (SZs) and other SZ-like patterns of brain activity can harm the brain and contribute to in-hospital death, particularly when prolonged. However, experts qualified to interpret EEG data are scarce. Prior attempts to automate this task have been limited by small or inadequately labeled samples and have not convincingly demonstrated generalizable expert-level performance. There exists a critical unmet need for an automated method to classify SZs and other SZ-like events with expert-level reliability. This study was conducted to develop and validate a computer algorithm that matches the reliability and accuracy of experts in identifying SZs and SZ-like events, known as \u201cictal-interictal-injury continuum\u201d (IIIC) patterns on EEG, including SZs, lateralized and generalized periodic discharges (LPD, GPD), and lateralized and generalized rhythmic delta activity (LRDA, GRDA), and in differentiating these patterns from non-IIIC patterns. Methods We used 6,095 scalp EEGs from 2,711 patients with and without IIIC events to train a deep neural network, SPaRCNet, to perform IIIC event classification. Independent training and test data sets were generated from 50,697 EEG segments, independently annotated by 20 fellowship-trained neurophysiologists. We assessed whether SPaRCNet performs at or above the sensitivity, specificity, precision, and calibration of fellowship-trained neurophysiologists for identifying IIIC events. Statistical performance was assessed by the calibration index and by the percentage of experts whose operating points were below the model's receiver operating characteristic curves (ROCs) and precision recall curves (PRCs) for the 6 pattern classes. Results SPaRCNet matches or exceeds most experts in classifying IIIC events based on both calibration and discrimination metrics. For SZ, LPD, GPD, LRDA, GRDA, and \u201cother\u201d classes, SPaRCNet exceeds the following percentages of 20 experts\u2014ROC: 45%, 20%, 50%, 75%, 55%, and 40%; PRC: 50%, 35%, 50%, 90%, 70%, and 45%; and calibration: 95%, 100%, 95%, 100%, 100%, and 80%, respectively. Discussion SPaRCNet is the first algorithm to match expert performance in detecting SZs and other SZ-like events in a representative sample of EEGs. With further development, SPaRCNet may thus be a valuable tool for an expedited review of EEGs. Classification of Evidence This study provides Class II evidence that among patients with epilepsy or critical illness undergoing EEG monitoring, SPaRCNet can differentiate (IIIC) patterns from non-IIIC events and expert neurophysiologists."
                    },
                    {
                        "title": "Artificial Intelligence System for Detection and Screening of Cardiac Abnormalities using Electrocardiogram Images",
                        "abstract": "The artificial intelligence (AI) system has achieved expert-level performance in electrocardiogram (ECG) signal analysis. However, in underdeveloped countries or regions where the healthcare information system is imperfect, only paper ECGs can be provided. Analysis of real-world ECG images (photos or scans of paper ECGs) remains challenging due to complex environments or interference. In this study, we present an AI system developed to detect and screen cardiac abnormalities (CAs) from real-world ECG images. The system was evaluated on a large dataset of 52,357 patients from multiple regions and populations across the world. On the detection task, the AI system obtained area under the receiver operating curve (AUC) of 0.996 (hold-out test), 0.994 (external test 1), 0.984 (external test 2), and 0.979 (external test 3), respectively. Meanwhile, the detection results of AI system showed a strong correlation with the diagnosis of cardiologists (cardiologist 1 (R=0.794, p<1e-3), cardiologist 2 (R=0.812, p<1e-3)). On the screening task, the AI system achieved AUCs of 0.894 (hold-out test) and 0.850 (external test). The screening performance of the AI system was better than that of the cardiologists (AI system (0.846) vs. cardiologist 1 (0.520) vs. cardiologist 2 (0.480)). Our study demonstrates the feasibility of an accurate, objective, easy-to-use, fast, and low-cost AI system for CA detection and screening. The system has the potential to be used by healthcare professionals, caregivers, and general users to assess CAs based on real-world ECG images."
                    },
                    {
                        "title": "Estimating causal effects of physical disability and number of comorbid chronic diseases on risk of depressive symptoms in an elderly Chinese population: a machine learning analysis of cross-sectional baseline data from the China longitudinal ageing social survey",
                        "abstract": "Objective This study aimed to explore the causal effects of physical disability and number of comorbid chronic diseases on depressive symptoms in an elderly Chinese population. Design, setting and analysis Cross-sectional, baseline data were obtained from the China Longitudinal Ageing Social Survey, a stratified, multistage, probabilistic sampling survey conducted in 2014 that covers 28 of 31 provincial areas in China. The causal effects of physical disability and number of comorbid chronic diseases on depressive symptoms were analysed using the conditional average treatment effect method of machine learning. The causal effects model\u2019s adjustment was made for age, gender, residence, marital status, educational level, ethnicity, wealth quantile and other factors. Outcome Assessment of the causal effects of physical disability and number of comorbid chronic diseases on depressive symptoms. Participants 7496 subjects who were 60 years of age or older and who answered the questions on depressive symptoms and other independent variables of interest in a survey conducted in 2014 were included in this study. Results Physical disability and number of comorbid chronic diseases had causal effects on depressive symptoms. Among the subjects who had one or more functional limitations, the probability of depressive symptoms increased by 22% (95% CI 19% to 24%). For the subjects who had one chronic disease and those who had two or more chronic diseases, the possibility of depressive symptoms increased by 13% (95% CI 10% to 15%) and 20% (95% CI 18% to 22%), respectively. Conclusion This study provides evidence that the presence of one or more functional limitations affects the occurrence of depressive symptoms among elderly people. The findings of our study are of value in developing programmes that are designed to identify elderly individuals who have physical disabilities or comorbid chronic diseases to provide early intervention."
                    },
                    {
                        "title": "Curriculum Design Helps Spiking Neural Networks to Classify Time Series",
                        "abstract": "Spiking Neural Networks (SNNs) have a greater potential for modeling time series data than Artificial Neural Networks (ANNs), due to their inherent neuron dynamics and low energy consumption. However, it is difficult to demonstrate their superiority in classification accuracy, because current efforts mainly focus on designing better network structures. In this work, enlighten by brain-inspired science, we find that, not only the structure but also the learning process should be human-like. To achieve this, we investigate the power of Curriculum Learning (CL) on SNNs by designing a novel method named CSNN with two theoretically guaranteed mechanisms: The active-to-dormant training order makes the curriculum similar to that of human learning and suitable for spiking neurons; The value-based regional encoding makes the neuron activity to mimic the brain memory when learning sequential data. Experiments on multiple time series sources including simulated, sensor, motion, and healthcare demonstrate that CL has a more positive effect on SNNs than ANNs with about twice the accuracy change, and CSNN can increase about 3% SNNs' accuracy by improving network sparsity, neuron firing status, anti-noise ability, and convergence speed."
                    },
                    {
                        "title": "TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Series",
                        "abstract": "This work summarizes two ways to accomplish Time-Series (TS) tasks in today's Large Language Model (LLM) context: LLM-for-TS (model-centric) designs and trains a fundamental large model, or fine-tunes a pre-trained LLM for TS data; TS-for-LLM (data-centric) converts TS into a model-friendly representation to enable the pre-trained LLM to handle TS data. Given the lack of data, limited resources, semantic context requirements, and so on, this work focuses on TS-for-LLM, where we aim to activate LLM's ability for TS data by designing a TS embedding method suitable for LLM. The proposed method is named TEST. It first tokenizes TS, builds an encoder to embed TS via instance-wise, feature-wise, and text-prototype-aligned contrast, where the TS embedding space is aligned to LLM embedding layer space, then creates soft prompts to make LLM more open to that embeddings, and finally implements TS tasks using the frozen LLM. We also demonstrate the feasibility of TS-for-LLM through theory and experiments. Experiments are carried out on TS classification, forecasting, and representation tasks using eight frozen LLMs with various structures and sizes. The results show that the pre-trained LLM with TEST strategy can achieve better or comparable performance than today's SOTA TS models and offer benefits for few-shot and generalization. By treating LLM as the pattern machine, TEST can endow LLM's ability to process TS data without compromising language ability. We hope that this study will serve as a foundation for future work to support TS+LLM progress."
                    },
                    {
                        "title": "A Deep Bayesian Neural Network for Cardiac Arrhythmia Classification with Rejection from ECG Recordings",
                        "abstract": "With the development of deep learning-based methods, automated classification of electrocardiograms (ECGs) has recently gained much attention. Although the effectiveness of deep neural networks has been encouraging, the lack of information given by the outputs restricts clinicians\u2019 reexamination. If the uncertainty estimation comes along with the classification results, cardiologists can pay more attention to \u201cuncertain\u201d cases. Our study aims to classify ECGs with rejection based on data uncertainty and model uncertainty. We perform experiments on a real-world 12-lead ECG dataset. First, we estimate uncertainties using the Monte Carlo dropout for each classification prediction, based on our Bayesian neural network. Then, we accept predictions with uncertainty under a given threshold and provide \u201cuncertain\u201d cases for clinicians. Furthermore, we perform a simulation experiment using varying thresholds. Finally, with the help of a clinician, we conduct case studies to explain the results of large uncertainties and incorrect predictions with small uncertainties. The results show that correct predictions are more likely to have smaller uncertainties, and the performance on accepted predictions improves as the accepting ratio decreases (i.e. more rejections). Case studies also help explain why rejection can improve the performance. Our study helps neural networks produce more accurate results and provide information on uncertainties to better assist clinicians in the diagnosis process. It can also enable deep-learning-based ECG interpretation in clinical implementation."
                    },
                    {
                        "title": "Practical Lessons on 12-Lead ECG Classification: Meta-Analysis of Methods From PhysioNet/Computing in Cardiology Challenge 2020",
                        "abstract": "Cardiovascular diseases (CVDs) are one of the most fatal disease groups worldwide. Electrocardiogram (ECG) is a widely used tool for automatically detecting cardiac abnormalities, thereby helping to control and manage CVDs. To encourage more multidisciplinary researches, PhysioNet/Computing in Cardiology Challenge 2020 (Challenge 2020) provided a public platform involving multi-center databases and automatic evaluations for ECG classification tasks. As a result, 41 teams successfully submitted their solutions and were qualified for rankings. Although Challenge 2020 was a success, there has been no in-depth methodological meta-analysis of these solutions, making it difficult for researchers to benefit from the solutions and results. In this study, we aim to systematically review the 41 solutions in terms of data processing, feature engineering, model architecture, and training strategy. For each perspective, we visualize and statistically analyze the effectiveness of the common techniques, and discuss the methodological advantages and disadvantages. Finally, we summarize five practical lessons based on the aforementioned analysis: (1) Data augmentation should be employed and adapted to specific scenarios; (2) Combining different features can improve performance; (3) A hybrid design of different types of deep neural networks (DNNs) is better than using a single type; (4) The use of end-to-end architectures should depend on the task being solved; (5) Multiple models are better than one. We expect that our meta-analysis will help accelerate the research related to ECG classification based on machine-learning models."
                    }
                ]
            },
            "9fff3a65-3ecc-4592-a2ab-79d291eedc91": {
                "pk": "9fff3a65-3ecc-4592-a2ab-79d291eedc91",
                "name": "Mengling Feng",
                "collaborators": [
                    "Hao Du",
                    "Melissa Min-Szu Yao",
                    "Wing P. Chan",
                    "Xiang Lan",
                    "Siqi Liu",
                    "Han Wang",
                    "Q. Ng",
                    "S. Arulanandam",
                    "M. Ong",
                    "Y. Jeon",
                    "Y. Ruan",
                    "K. Tan",
                    "Dianwen Ng",
                    "Dilruk Perera",
                    "Colin K Tan",
                    "K. Bae",
                    "Y. Hwangbo",
                    "C. Yoo",
                    "Nayoung Han",
                    "Liangyu Chen",
                    "Daniel J. Tan",
                    "H. Abdullah",
                    "Mikael Hartman",
                    "Trung Kien Dang",
                    "J. Weng",
                    "Hongfei Yang",
                    "Jiangeng Chang",
                    "Shaoze Cui",
                    "John Soong Tshon Yit",
                    "Mukkesh Kumar",
                    "Li Ting Ang",
                    "Hang Png",
                    "Maisie Ng",
                    "S. Loy",
                    "J. Chan",
                    "K. Godfrey",
                    "Shiao-Yng Chan",
                    "Y. Chong",
                    "J. Eriksson",
                    "N. Karnani",
                    "Wesley Yeung",
                    "Angeline Tung",
                    "Joey Ai Meng Tay",
                    "Davin Ryanputra",
                    "Yi Liu",
                    "Jiongzhou Liu",
                    "Fan Xu",
                    "Cheng He",
                    "R. Kosik"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Machine Learning",
                    "Medical Imaging",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Demystifying Complex Treatment Recommendations: A Hierarchical Cooperative Multi-Agent RL Approach",
                        "abstract": "Reinforcement Learning (RL) techniques are widely adopted in healthcare research to learn effective policies for complex sequential decision-making tasks, such as treatment recommendation to achieve better clinical outcomes. However, existing RL based solutions suffer from two major limitations. (1) Burdens a single agent to make complex treatment recommendations: existing RL algorithms do not explicitly support the common hierarchical decision-making processes used in practise. That is, they consider treatment recommendation as a singular decision and the treatment space as a flat search space, and expect a single RL agent to select the best treatment outright. (2) Sub optimal patient state representation techniques: accurate representation of patient states and their dynamics is essential to make effective clinical decisions. However, there are three main limitations in current state representation techniques in the clinical environment: (i) lack of clear structure and high variability of input features; (ii) lack of a temporal context, which makes it difficult to fully capture patient conditions and their dynamics at a given time; (iii) and the use of imputation techniques, which undermines personalization, introduce biases and errors that can cause harm to patients. To address the complexities of treatment recommendation and generate superior treatment policies, we propose a consolidated RL solution that incorporates a hierarchical multi-agent architecture and advanced patient state representation technique. In our proposed architecture, the complex task is decomposed into a hierarchy of subtasks, each handled by a specialized agent that coordinates skills learned by lower level agents to determine a high-level treatment policy. The patient state representation technique is designed to effectively encode the changing states of patients throughout their hospital stay, addressing the three primary limitations of current approaches. Our experiments using the sepsis treatment task shows that the proposed solution consistently outperforms state-of-the-art baselines and improves patient survival."
                    },
                    {
                        "title": "Building a Machine Learning-based Ambulance Dispatch Triage Model for Emergency Medical Services",
                        "abstract": "Background In charge of dispatching the ambulances, Emergency Medical Services (EMS) call center specialists often have difficulty deciding the acuity of a case given the information they can gather within a limited time. Although there are protocols to guide their decision-making, observed performance can still lack sensitivity and specificity. Machine learning models have been known to capture complex relationships that are subtle, and well-trained data models can yield accurate predictions in a split of a second. Methods In this study, we proposed a proof-of-concept approach to construct a machine learning model to better predict the acuity of emergency cases. We used more than 360,000 structured emergency call center records of cases received by the national emergency call center in Singapore from 2018 to 2020. Features were created using call records, and multiple machine learning models were trained. Results A Random Forest model achieved the best performance, reducing the over-triage rate by an absolute margin of 15% compared to the call center specialists while maintaining a similar level of under-triage rate. Conclusions The model has the potential to be deployed as a decision support tool for dispatchers alongside current protocols to optimize ambulance dispatch triage and the utilization of emergency ambulance resources."
                    },
                    {
                        "title": "Data-Efficient Computational Pathology Platform for Faster and Cheaper Breast Cancer Subtype Identifications: Development of a Deep Learning Model",
                        "abstract": "Background Breast cancer subtyping is a crucial step in determining therapeutic options, but the molecular examination based on immunohistochemical staining is expensive and time-consuming. Deep learning opens up the possibility to predict the subtypes based on the morphological information from hematoxylin and eosin staining, a much cheaper and faster alternative. However, training the predictive model conventionally requires a large number of histology images, which is challenging to collect by a single institute. Objective We aimed to develop a data-efficient computational pathology platform, 3DHistoNet, which is capable of learning from z-stacked histology images to accurately predict breast cancer subtypes with a small sample size. Methods We retrospectively examined 401 cases of patients with primary breast carcinoma diagnosed between 2018 and 2020 at the Department of Pathology, National Cancer Center, South Korea. Pathology slides of the patients with breast carcinoma were prepared according to the standard protocols. Age, gender, histologic grade, hormone receptor (estrogen receptor [ER], progesterone receptor [PR], and androgen receptor [AR]) status, erb-B2 receptor tyrosine kinase 2 (HER2) status, and Ki-67 index were evaluated by reviewing medical charts and pathological records. Results The area under the receiver operating characteristic curve and decision curve were analyzed to evaluate the performance of our 3DHistoNet platform for predicting the ER, PR, AR, HER2, and Ki67 subtype biomarkers with 5-fold cross-validation. We demonstrated that 3DHistoNet can predict all clinically important biomarkers (ER, PR, AR, HER2, and Ki67) with performance exceeding the conventional multiple instance learning models by a considerable margin (area under the receiver operating characteristic curve: 0.75-0.91 vs 0.67-0.8). We further showed that our z-stack histology scanning method can make up for insufficient training data sets without any additional cost incurred. Finally, 3DHistoNet offered an additional capability to generate attention maps that reveal correlations between Ki67 and histomorphological features, which renders the hematoxylin and eosin image in higher fidelity to the pathologist. Conclusions Our stand-alone, data-efficient pathology platform that can both generate z-stacked images and predict key biomarkers is an appealing tool for breast cancer diagnosis. Its development would encourage morphology-based diagnosis, which is faster, cheaper, and less error-prone compared to the protein quantification method based on immunohistochemical staining."
                    },
                    {
                        "title": "Automatic Calcification Morphology and Distribution Classification for Breast Mammograms With Multi-Task Graph Convolutional Neural Network",
                        "abstract": "The morphology and distribution of microcalcifications are the most important descriptors for radiologists to diagnose breast cancer based on mammograms. However, it is very challenging and time-consuming for radiologists to characterize these descriptors manually, and there also lacks of effective and automatic solutions for this problem. We observed that the distribution and morphology descriptors are determined by the radiologists based on the spatial and visual relationships among calcifications. Thus, we hypothesize that this information can be effectively modelled by learning a relationship-aware representation using graph convolutional networks (GCNs). In this study, we propose a multi-task deep GCN method for automatic characterization of both the morphology and distribution of microcalcifications in mammograms. Our proposed method transforms morphology and distribution characterization into node and graph classification problem and learns the representations concurrently. We trained and validated the proposed method in an in-house dataset and public DDSM dataset with 195 and 583 cases,respectively. The proposed method reaches good and stable results with distribution AUC at 0.812 <inline-formula><tex-math notation=\"LaTeX\">$\\pm$</tex-math></inline-formula> 0.043 and 0.873 <inline-formula><tex-math notation=\"LaTeX\">$\\pm$</tex-math></inline-formula> 0.019, morphology AUC at 0.663 <inline-formula><tex-math notation=\"LaTeX\">$\\pm$</tex-math></inline-formula> 0.016 and 0.700 <inline-formula><tex-math notation=\"LaTeX\">$\\pm$</tex-math></inline-formula> 0.044 for both in-house and public datasets. In both datasets, our proposed method demonstrates statistically significant improvements compared to the baseline models. The performance improvements brought by our proposed multi-task mechanism can be attributed to the association between the distribution and morphology of calcifications in mammograms, which is interpretable using graphical visualizations and consistent with the definitions of descriptors in the standard BI-RADS guideline. In short, we explore, for the first time, the application of GCNs in microcalcification characterization that suggests the potential of using graph learning for more robust understanding of medical images."
                    },
                    {
                        "title": "P-Transformer: A Prompt-based Multimodal Transformer Architecture For Medical Tabular Data",
                        "abstract": "Medical tabular data, abundant in Electronic Health Records (EHRs), is a valuable resource for diverse medical tasks such as risk prediction. While deep learning approaches, particularly transformer-based models, have shown remarkable performance in tabular data prediction, there are still problems remained for existing work to be effectively adapted into medical domain, such as under-utilization of unstructured free-texts, limited exploration of textual information in structured data, and data corruption. To address these issues, we propose P-Transformer, a Prompt-based multimodal Transformer architecture designed specifically for medical tabular data. This framework consists two critical components: a tabular cell embedding generator and a tabular transformer. The former efficiently encodes diverse modalities from both structured and unstructured tabular data into a harmonized language semantic space with the help of pre-trained sentence encoder and medical prompts. The latter integrates cell representations to generate patient embeddings for various medical tasks. In comprehensive experiments on two real-world datasets for three medical tasks, P-Transformer demonstrated the improvements with 10.9%/11.0% on RMSE/MAE, 0.5%/2.2% on RMSE/MAE, and 1.6%/0.8% on BACC/AUROC compared to state-of-the-art (SOTA) baselines in predictability. Notably, the model exhibited strong resilience to data corruption in the structured data, particularly when the corruption rates are high."
                    },
                    {
                        "title": "End-to-End Calcification Distribution Pattern Recognition for Mammograms: An Interpretable Approach with GNN",
                        "abstract": "Abstract Purpose: We aimed to develop a novel interpretable artificial intelligence (AI) model algorithm focusing on automatic detection and classification of various patterns of calcification distribution in mammographic images using a unique graph convolution approach. Materials and methods: Images from 292 patients, which showed calcifications according to the mammographic reports and diagnosed breast cancers, were collected. The calcification distributions were classified as diffuse, segmental, regional, grouped, or linear. Excluded were mammograms with (1) breast cancer with multiple lexicons such as mass, asymmetry, or architectural distortion without calcifications; (2) hidden calcifications that were difficult to mark; or (3) incomplete medical records. Results: A graph-convolutional-network-based model was developed. A total of 581 mammographic images from 292 cases of breast cancer were divided based on the calcification distribution pattern: diffuse (n = 67), regional (n = 115), group (n = 337), linear (n = 8), or segmental (n = 54). The classification performances were measured using metrics including precision, recall, F1 score, accuracy, and multi-class area under the receiver operating characteristic curve. The proposed model achieved a precision of 0.522 \u00b1 0.028, sensitivity of 0.643 \u00b1 0.017, specificity of 0.847 \u00b1 0.009, F1 score of 0.559 \u00b1 0.018, accuracy of 64.325 \u00b1 1.694%, and area under the curve of 0.745 \u00b1 0.030; thus, the method was found to be superior compared to all baseline models. The predicted linear and diffuse classifications were highly similar to the ground truth, and the predicted grouped and regional classifications were also superior compared to baseline models. The prediction results are interpretable using visualization methods to highlight the important calcification nodes in graphs. Conclusions: The proposed deep neural network framework is an AI solution that automatically detects and classifies calcification distribution patterns on mammographic images highly suspected of showing breast cancers. Further study of the AI model in an actual clinical setting and additional data collection will improve its performance."
                    },
                    {
                        "title": "Federated Learning for Electronic Health Records",
                        "abstract": "In data-driven medical research, multi-center studies have long been preferred over single-center ones due to a single institute sometimes not having enough data to obtain sufficient statistical power for certain hypothesis testings as well as predictive and subgroup studies. The wide adoption of electronic health records (EHRs) has made multi-institutional collaboration much more feasible. However, concerns over infrastructures, regulations, privacy, and data standardization present a challenge to data sharing across healthcare institutions. Federated Learning (FL), which allows multiple sites to collaboratively train a global model without directly sharing data, has become a promising paradigm to break the data isolation. In this study, we surveyed existing works on FL applications in EHRs and evaluated the performance of current state-of-the-art FL algorithms on two EHR machine learning tasks of significant clinical importance on a real world multi-center EHR dataset."
                    },
                    {
                        "title": "FCSN: Global Context Aware Segmentation by Learning the Fourier Coefficients of Objects in Medical Images",
                        "abstract": "The encoder-decoder model is a commonly used Deep learning (DL) model for medical image segmentation. Encoder-decoder models make pixel-wise predictions that focus heavily on local patterns. As a result, the predicted mask often fails to preserve the object's shape and topology, which requires an understanding of the image's global context. In this work, we propose a Fourier Coefficient Segmentation Network (FCSN)---a novel global context-aware DL model that segments an object by learning the complex Fourier Coefficients of the object's masks. The Fourier coefficients are calculated by integrating over the mask's contour. Hence, FCSN is naturally motivated to incorporate a broader image context when estimating the coefficients. The global context awareness of FCSN helps produce more accurate segmentation and is more robust to local perturbations, such as additive noise or motion blur. We compare FCSN on other state-of-the-art global context-aware models (UNet++, DeepLabV3+, UNETR) on 5 medical image segmentation tasks (ISIC_2018, RIM_CUP, RIM_DISC, PROSTATE, FETAL). When compared with UNETR, FCSN attains significantly lower Hausdorff scores with 19.14 (6%), 17.42 (6%), 9.16 (14%), 11.18 (22%), and 5.98 (6%) for ISIC_2018, RIM_CUP, RIM_DISC, PROSTATE, and FETAL tasks respectively. Moreover, FCSN is lightweight by discarding the decoder module. FCSN only requires 29.7 M parameters which are 75.6 M and 9.9 M fewer than UNETR and DeepLabV3+, respectively. FCSN attains inference/training speeds of 1.6 ms/img and 6.3 ms/img, which is 8 and 3 faster than UNet and UNETR. Our work is available at https://github.com/nus-morninlab/FCSN."
                    },
                    {
                        "title": "UFRC: A Unified Framework for Reliable COVID-19 Detection on Crowdsourced Cough Audio",
                        "abstract": "We suggested a unified system with core components of data augmentation, ImageNet-pretrained ResNet-50, cost-sensitive loss, deep ensemble learning, and uncertainty estimation to quickly and consistently detect COVID-19 using acoustic evidence. To increase the model's capacity to identify a minority class, data augmentation and cost-sensitive loss are incorporated (infected samples). In the COVID-19 detection challenge, ImageNet-pretrained ResNet-50 has been found to be effective. The unified framework also integrates deep ensemble learning and uncertainty estimation to integrate predictions from various base classifiers for generalisation and reliability. We ran a series of tests using the DiCOVA2021 challenge dataset to assess the efficacy of our proposed method, and the results show that our method has an AUC-ROC of 85.43 percent, making it a promising method for COVID-19 detection. The unified framework also demonstrates that audio may be used to quickly diagnose different respiratory disorders."
                    },
                    {
                        "title": "Automated Machine Learning (AutoML)-Derived Preconception Predictive Risk Model to Guide Early Intervention for Gestational Diabetes Mellitus",
                        "abstract": "The increasing prevalence of gestational diabetes mellitus (GDM) is contributing to the rising global burden of type 2 diabetes (T2D) and intergenerational cycle of chronic metabolic disorders. Primary lifestyle interventions to manage GDM, including second trimester dietary and exercise guidance, have met with limited success due to late implementation, poor adherence and generic guidelines. In this study, we aimed to build a preconception-based GDM predictor to enable early intervention. We also assessed the associations of top predictors with GDM and adverse birth outcomes. Our evolutionary algorithm-based automated machine learning (AutoML) model was implemented with data from 222 Asian multi-ethnic women in a preconception cohort study, Singapore Preconception Study of Long-Term Maternal and Child Outcomes (S-PRESTO). A stacked ensemble model with a gradient boosting classifier and linear support vector machine classifier (stochastic gradient descent training) was derived using genetic programming, achieving an excellent AUC of 0.93 based on four features (glycated hemoglobin A1c (HbA1c), mean arterial blood pressure, fasting insulin, triglycerides/HDL ratio). The results of multivariate logistic regression model showed that each 1 mmol/mol increase in preconception HbA1c was positively associated with increased risks of GDM (p = 0.001, odds ratio (95% CI) 1.34 (1.13\u20131.60)) and preterm birth (p = 0.011, odds ratio 1.63 (1.12\u20132.38)). Optimal control of preconception HbA1c may aid in preventing GDM and reducing the incidence of preterm birth. Our trained predictor has been deployed as a web application that can be easily employed in GDM intervention programs, prior to conception."
                    },
                    {
                        "title": "A Weakly-Supervised Named Entity Recognition Machine Learning Approach for Emergency Medical Services Clinical Audit",
                        "abstract": "Clinical performance audits are routinely performed in Emergency Medical Services (EMS) to ensure adherence to treatment protocols, to identify individual areas of weakness for remediation, and to discover systemic deficiencies to guide the development of the training syllabus. At present, these audits are performed by manual chart review, which is time-consuming and laborious. In this paper, we report a weakly-supervised machine learning approach to train a named entity recognition model that can be used for automatic EMS clinical audits. The dataset used in this study contained 58,898 unlabeled ambulance incidents encountered by the Singapore Civil Defence Force from 1st April 2019 to 30th June 2019. With only 5% labeled data, we successfully trained three different models to perform the NER task, achieving F1 scores of around 0.981 under entity type matching evaluation and around 0.976 under strict evaluation. The BiLSTM-CRF model was 1~2 orders of magnitude lighter and faster than our BERT-based models. Our proposed proof-of-concept approach may improve the efficiency of clinical audits and can also help with EMS database research. Further external validation of this approach is needed."
                    },
                    {
                        "title": "Adversarial Domain Adaptation with Correlation-Based Association Networks for Longitudinal Disk Fault Prediction",
                        "abstract": "Disk fault is known to be the key cause of data loss in the modern large-scale data center, which affects the reliability and stability of the server and even the whole IT infrastructure, resulting in high financial cost. Recent works on disk fault prediction demonstrate the ability of machine learning techniques in the early prediction of disk failure. However, two limitations hinder the real-world application of current methods. First, they ignore the data heterogeneity in the data center, where distribution shifts commonly exist across different disk model types that decrease the model's performance. Second, the number of disks of different disk model types varies greatly in the data center, and current methods failed to deliver an acceptable performance over disk model types with few training samples. To address these limitations, we propose adversarial domain adaptation with correlation-based association networks (ADA-CBAN) to both mitigate the distribution shift problem and also to boost the performance on small-scaled data. Extensive experiments prove that our proposed model is effective and achieves new state-of-the-art. In addition, our post model analysis can also reveal important feature interactions and highlight the crucial period before disk faults, where both are useful information in real-world applications."
                    },
                    {
                        "title": "Today's radiologists meet tomorrow's AI: the promises, pitfalls, and unbridled potential.",
                        "abstract": "Advances in information technology have improved radiologists' abilities to perform an increasing variety of targeted diagnostic exams. However, due to a growing demand for imaging from an aging population, the number of exams could soon exceed the number of radiologists available to read them. However, artificial intelligence has recently resounding success in several case studies involving the interpretation of radiologic exams. As such, the integration of AI with standard diagnostic imaging practices to revolutionize medical care has been proposed, with the ultimate goal being the replacement of human radiologists with AI 'radiologists'. However, the complexity of medical tasks is often underestimated, and many proponents are oblivious to the limitations of AI algorithms. In this paper, we review the hype surrounding AI in medical imaging and the changing opinions over the years, ultimately describing AI's shortcomings. Nonetheless, we believe that AI has the potential to assist radiologists. Therefore, we discuss ways AI can increase a radiologist's efficiency by integrating it into the standard workflow."
                    }
                ]
            }
        }
    },
    "2409.17331": {
        "paper_data": {
            "title": "ChatCam: Empowering Camera Control through Conversational AI",
            "url": "http://arxiv.org/abs/2409.17331v1",
            "arxiv_id": "2409.17331",
            "authors": [
                "Xinhang Liu",
                "Yu-Wing Tai",
                "Chi-Keung Tang"
            ],
            "abstract": "Cinematographers adeptly capture the essence of the world, crafting compelling visual narratives through intricate camera movements. Witnessing the strides made by large language models in perceiving and interacting with the 3D world, this study explores their capability to control cameras with human language guidance. We introduce ChatCam, a system that navigates camera movements through conversations with users, mimicking a professional cinematographer's workflow. To achieve this, we propose CineGPT, a GPT-based autoregressive model for text-conditioned camera trajectory generation. We also develop an Anchor Determinator to ensure precise camera trajectory placement. ChatCam understands user requests and employs our proposed tools to generate trajectories, which can be used to render high-quality video footage on radiance field representations. Our experiments, including comparisons to state-of-the-art approaches and user studies, demonstrate our approach's ability to interpret and execute complex instructions for camera operation, showing promising applications in real-world production settings.",
            "introduction": "   1 Introduction  Cinematographers skillfully capture the essence of the 3D world by maneuvering their cameras, creating an array of compelling visual narratives\u00a0[8]. Achieving aesthetically pleasing results requires not only a deep understanding of scene elements and their interplay but also meticulous execution of techniques.   Recent progress of large language models (LLMs)\u00a0[1] has marked a significant milestone in AI development, demonstrating their capability to understand and act within the 3D world\u00a0[28, 29, 88]. Witnessing this evolution, our work explores the feasibility of empowering camera control through conversational AI, thus enhancing the video production process across diverse domains such as documentary filmmaking, live event broadcasting, and virtual reality experiences.   Although the community has devoted considerable effort to controlling the trajectories of objects and cameras in video generation approaches for practical usage\u00a0[4, 83, 75, 27], or predicting similar sequences through autoregressive decoding processes\u00a0[34, 64], generating camera trajectories has yet to be explored. This task involves multiple elements such as language, images, 3D assets, and, beyond mere accuracy, necessitates visually pleasing rendered videos as the ultimate goal.   We propose ChatCam, a system that allows users to control camera operations through natural language interaction. As illustrated in Figure\u00a01, leveraging an LLM agent to orchestrate camera operations, our method assists users in generating desired camera trajectories, which can be used to render videos on radiance field representations such as NeRF\u00a0[52] or 3DGS\u00a0[35].   At the core of our approach, we introduce CineGPT, a GPT-based autoregressive model that integrates language understanding with camera trajectory generation. We train this model using a paired text-trajectory dataset to equip it with the ability for text-conditioned trajectory generation. We also propose an Anchor Determinator, a module that identifies relevant objects within the 3D scene to serve as anchors, ensuring correct trajectory placement based on user specifications. Our LLM agent parses compositional natural language queries into semantic concepts. With these parsed sub-queries as inputs, the agent then calls our proposed CineGPT and Anchor Determinator. It composes the final trajectory with the outputs from these tools, which can ultimately be used to render a video that fulfills the user\u2019s request.   With comprehensive evaluations and comparisons to other state-of-the-art methods, our method exhibits a pronounced ability to interpret and execute complex instructions for camera operation. Our user studies further demonstrate its promising application prospects in actual production settings. In summary, this paper\u2019s contributions are as follows:   \u2022  We introduce ChatCam, a system that, for the first time, enables users to operate cameras through natural language interactions. It simplifies sophisticated camera movements and reduces technical hurdles for creators.    \u2022  We develop CineGPT for text-conditioned camera trajectory generation and an Anchor Determinator for precise camera trajectory placement. Our LLM agent understands users\u2019 requests and leverages our proposed tools to complete the task.    \u2022  Extensive experiments demonstrate the effectiveness of our method, showing how AI can effectively collaborate with humans on complex tasks involving multiple elements such as language, images, 3D assets, and camera trajectories.      Figure 1: Empowering camera control through conversational AI. Our proposed ChatCam assists users in generating desired camera trajectories through natural language interactions. The generated trajectories can be used to render videos on radiance field representations such as NeRF\u00a0[52] or 3DGS\u00a0[35].     2 Related",
            "references": [
                {
                    "title": "Gear-NeRF: Free-Viewpoint Rendering and Tracking with Motion-Aware Spatio-Temporal Sampling",
                    "abstract": "Extensions of Neural Radiance Fields (NeRFs) to model dynamic scenes have enabled their near photo-realistic, free-viewpoint rendering. Although these methods have shown some potential in creating immersive experiences, two drawbacks limit their ubiquity: ( i) a significant reduction in reconstruction quality when the computing budget is limited, and (ii) a lack of semantic understanding of the underlying scenes. To address these issues, we introduce Gear-NeRF, which leverages semantic information from powerful image segmentation models. Our approach presents a principled way for learning a spatio-temporal (4D) semantic embedding, based on which we introduce the concept of gears to allow for stratified modeling of dynamic regions of the scene based on the extent of their motion. Such differentiation allows us to adjust the spatio- temporal sampling resolution for each region in proportion to its motion scale, achieving more photo-realistic dynamic novel view synthesis. At the same time, almost for free, our approach enables free-viewpoint tracking of objects of interest - a functionality not yet achieved by existing NeRF-based methods. Empirical studies validate the effectiveness of our method, where we achieve state-of-the-art rendering and tracking performance on multiple challenging datasets. The project page is available at: https://merl.com/research/highlights/gear-nerf"
                },
                {
                    "title": "CameraCtrl: Enabling Camera Control for Text-to-Video Generation",
                    "abstract": "Controllability plays a crucial role in video generation since it allows users to create desired content. However, existing models largely overlooked the precise control of camera pose that serves as a cinematic language to express deeper narrative nuances. To alleviate this issue, we introduce CameraCtrl, enabling accurate camera pose control for text-to-video(T2V) models. After precisely parameterizing the camera trajectory, a plug-and-play camera module is then trained on a T2V model, leaving others untouched. Additionally, a comprehensive study on the effect of various datasets is also conducted, suggesting that videos with diverse camera distribution and similar appearances indeed enhance controllability and generalization. Experimental results demonstrate the effectiveness of CameraCtrl in achieving precise and domain-adaptive camera control, marking a step forward in the pursuit of dynamic and customized video storytelling from textual and camera pose inputs. Our project website is at: https://hehao13.github.io/projects-CameraCtrl/."
                },
                {
                    "title": "TC4D: Trajectory-Conditioned Text-to-4D Generation",
                    "abstract": "Recent techniques for text-to-4D generation synthesize dynamic 3D scenes using supervision from pre-trained text-to-video models. However, existing representations for motion, such as deformation models or time-dependent neural representations, are limited in the amount of motion they can generate-they cannot synthesize motion extending far beyond the bounding box used for volume rendering. The lack of a more flexible motion model contributes to the gap in realism between 4D generation methods and recent, near-photorealistic video generation models. Here, we propose TC4D: trajectory-conditioned text-to-4D generation, which factors motion into global and local components. We represent the global motion of a scene's bounding box using rigid transformation along a trajectory parameterized by a spline. We learn local deformations that conform to the global trajectory using supervision from a text-to-video model. Our approach enables the synthesis of scenes animated along arbitrary trajectories, compositional scene generation, and significant improvements to the realism and amount of generated motion, which we evaluate qualitatively and through a user study. Video results can be viewed on our website: https://sherwinbahmani.github.io/tc4d."
                },
                {
                    "title": "3D-VLA: A 3D Vision-Language-Action Generative World Model",
                    "abstract": "Recent vision-language-action (VLA) models rely on 2D inputs, lacking integration with the broader realm of the 3D physical world. Furthermore, they perform action prediction by learning a direct mapping from perception to action, neglecting the vast dynamics of the world and the relations between actions and dynamics. In contrast, human beings are endowed with world models that depict imagination about future scenarios to plan actions accordingly. To this end, we propose 3D-VLA by introducing a new family of embodied foundation models that seamlessly link 3D perception, reasoning, and action through a generative world model. Specifically, 3D-VLA is built on top of a 3D-based large language model (LLM), and a set of interaction tokens is introduced to engage with the embodied environment. Furthermore, to inject generation abilities into the model, we train a series of embodied diffusion models and align them into the LLM for predicting the goal images and point clouds. To train our 3D-VLA, we curate a large-scale 3D embodied instruction dataset by extracting vast 3D-related information from existing robotics datasets. Our experiments on held-in datasets demonstrate that 3D-VLA significantly improves the reasoning, multimodal generation, and planning capabilities in embodied environments, showcasing its potential in real-world applications."
                },
                {
                    "title": "Direct-a-Video: Customized Video Generation with User-Directed Camera Movement and Object Motion",
                    "abstract": "Recent text-to-video diffusion models have achieved impressive progress. In practice, users often desire the ability to control object motion and camera movement independently for customized video creation. However, current methods lack the focus on separately controlling object motion and camera movement in a decoupled manner, which limits the controllability and flexibility of text-to-video models. In this paper, we introduce Direct-a-Video, a system that allows users to independently specify motions for multiple objects as well as camera's pan and zoom movements, as if directing a video. We propose a simple yet effective strategy for the decoupled control of object motion and camera movement. Object motion is controlled through spatial cross-attention modulation using the model's inherent priors, requiring no additional optimization. For camera movement, we introduce new temporal cross-attention layers to interpret quantitative camera movement parameters. We further employ an augmentation-based approach to train these layers in a self-supervised manner on a small-scale dataset, eliminating the need for explicit motion annotation. Both components operate independently, allowing individual or combined control, and can generalize to open-domain scenarios. Extensive experiments demonstrate the superiority and effectiveness of our method. Project page and code are available at https://direct-a-video.github.io/."
                },
                {
                    "title": "MultiPLY: A Multisensory Object-Centric Embodied Large Language Model in 3D World",
                    "abstract": "Human beings possess the capability to multiply a m\u00e9lange of multisensory cues while actively exploring and interacting with the 3D world. Current multi-modal large language models, however, passively absorb sensory data as inputs, lacking the capacity to actively interact with the objects in the 3D environment and dynamically collect their multisensory information. To usher in the study of this area, we propose MultiPLY, a multisensory embodied large language model that could incorporate multisensory interactive data, including visual, audio, tactile, and thermal information into large language models, thereby establishing the correlation among words, actions, and percepts. To this end, we first collect Multisensory Universe, a large-scale multisensory interaction dataset comprising 500k data by deploying an LLM-powered embodied agent to engage with the 3D environment. To perform instruction tuning with pre-trained LLM on such generated data, we first encode the 3D scene as abstracted object-centric representations, and then introduce action tokens denoting that the embodied agent takes certain actions within the environment, as well as state tokens that represent the multisensory state observations of the agent at each time step. In the inference time, MultiPLY could generate action tokens, instructing the agent to take the action in the environment and obtain the next multisensory state observation. The observation is then appended back to the LLM via state tokens to generate subsequent text or action tokens. We demonstrate that MultiPLY out-performs baselines by a large margin through a diverse set of embodied tasks involving object retrieval, tool use, mul-tisensory captioning, and task decomposition."
                },
                {
                    "title": "MotionCtrl: A Unified and Flexible Motion Controller for Video Generation",
                    "abstract": "Motions in a video primarily consist of camera motion, induced by camera movement, and object motion, resulting from object movement. Accurate control of both camera and object motion is essential for video generation. However, existing works either mainly focus on one type of motion or do not clearly distinguish between the two, limiting their control capabilities and diversity. Therefore, this paper presents MotionCtrl, a unified and flexible motion controller for video generation designed to effectively and independently control camera and object motion. The architecture and training strategy of MotionCtrl are carefully devised, taking into account the inherent properties of camera motion, object motion, and imperfect training data. Compared to previous methods, MotionCtrl offers three main advantages: 1) It effectively and independently controls camera motion and object motion, enabling more fine-grained motion control and facilitating flexible and diverse combinations of both types of motion. 2) Its motion conditions are determined by camera poses and trajectories, which are appearance-free and minimally impact the appearance or shape of objects in generated videos. 3) It is a relatively generalizable model that can adapt to a wide array of camera poses and trajectories once trained. Extensive qualitative and quantitative experiments have been conducted to demonstrate the superiority of MotionCtrl over existing methods. Project Page: https://wzhouxiff.github.io/projects/MotionCtrl/"
                },
                {
                    "title": "ReconFusion: 3D Reconstruction with Diffusion Priors",
                    "abstract": "3D reconstruction methods such as Neural Radiance Fields (NeRFs) excel at rendering photorealistic novel views of complex scenes. However, recovering a high-quality NeRF typically requires tens to hundreds of input images, resulting in a time-consuming capture process. We present ReconFusion to reconstruct real-world scenes using only a few photos. Our approach leverages a diffusion prior for novel view synthesis, trained on synthetic and multiview datasets, which regularizes a NeRF-based 3D reconstruction pipeline at novel camera poses beyond those captured by the set of input images. Our method synthesizes realistic geometry and texture in underconstrained regions while preserving the appearance of observed regions. We perform an extensive evaluation across various real-world datasets, including forward-facing and 360-degree scenes, demonstrating significant performance improvements over previous few-view NeRF reconstruction approaches. Please see our project page at reconfusion.github. io."
                },
                {
                    "title": "SANeRF-HQ: Segment Anything for NeRF in High Quality",
                    "abstract": "Recently, the Segment Anything Model (SAM) has showcased remarkable capabilities of zero-shot segmentation, while NeRF (Neural Radiance Fields) has gained popularity as a method for various 3D problems beyond novel view synthesis. Though there exist initial attempts to incorporate these two methods into 3D segmentation, they face the challenge of accurately and consistently segmenting objects in complex scenarios. In this paper, we introduce the Segment Anything for NeRF in High Quality (SANeRF-HQ) to achieve high-quality 3D segmentation of any target object in a given scene. SANeRF-HQ utilizes SAM for open-world object segmentation guided by user-supplied prompts, while leveraging NeRF to aggregate information from different viewpoints. To overcome the aforementioned challenges, we employ density field and RGB similarity to enhance the accuracy of segmentation boundary during the aggregation. Emphasizing on segmentation accuracy, we evaluate our method on multiple NeRF datasets where high-quality ground-truths are available or manually annotated. SANeRF-HQ shows a significant quality improvement over state-of-the-art methods in NeRF object segmentation, provides higher flexibility for object localization, and enables more consistent object segmentation across multiple views. Results and code are available at the project site: https://lyclyc52.github.io/SANeRF-HQ/."
                },
                {
                    "title": "C3Net: Compound Conditioned ControlNet for Multimodal Content Generation",
                    "abstract": "We present Compound Conditioned ControlNet, C3Net, a novel generative neural architecture taking conditions from multiple modalities and synthesizing multimodal contents simultaneously (e.g., image, text, audio). C3Net adapts the ControlNet [46] architecture to jointly train and make inferences on a production-ready diffusion model and its trainable copies. Specifically, C3Net first aligns the conditions from multimodalities to the same semantic latent space using modality-specific encoders based on contrastive training. Then, it generates multimodal outputs based on the aligned latent space, whose semantic information is combined using a ControlNet-like architecture called Control C3-UNet. Correspondingly, with this system design, our model offers an improved solution for joint-modality generation through learning and explaining multimodal conditions, involving more than just linear interpolation within the latent space. Meanwhile, as we align conditions to a unified latent space, C3Net only requires one trainable Control C3-UNet to work on multimodal semantic information. Furthermore, our model employs uni-modal pretraining on the condition alignment stage, outperforming the non-pretrained alignment even on relatively scarce training data and thus demonstrating high-quality compound condition generation. We contribute the first high-quality tri-modal validation set to validate quantitatively that C3Net outperforms or is on par with the first and contemporary state-of-the-art multimodal generation [43]. Our codes and tri-modal dataset will be released here."
                },
                {
                    "title": "MotionLM: Multi-Agent Motion Forecasting as Language Modeling",
                    "abstract": "Reliable forecasting of the future behavior of road agents is a critical component to safe planning in autonomous vehicles. Here, we represent continuous trajectories as sequences of discrete motion tokens and cast multi-agent motion prediction as a language modeling task over this domain. Our model, MotionLM, provides several advantages: First, it does not require anchors or explicit latent variable optimization to learn multimodal distributions. Instead, we leverage a single standard language modeling objective, maximizing the average log probability over sequence tokens. Second, our approach bypasses post-hoc interaction heuristics where individual agent trajectory generation is conducted prior to interactive scoring. Instead, MotionLM produces joint distributions over interactive agent futures in a single autoregressive decoding process. In addition, the model\u2019s sequential factorization enables temporally causal conditional rollouts. The proposed approach establishes new state-of-the-art performance for multi-agent motion prediction on the Waymo Open Motion Dataset, ranking 1st on the interactive challenge leaderboard."
                },
                {
                    "title": "LLM-Grounder: Open-Vocabulary 3D Visual Grounding with Large Language Model as an Agent",
                    "abstract": "3D visual grounding is a critical skill for household robots, enabling them to navigate, manipulate objects, and answer questions based on their environment. While existing approaches often rely on extensive labeled data or exhibit limitations in handling complex language queries, we propose LLM-Grounder, a novel zero-shot, open-vocabulary, Large Language Model (LLM)-based 3D visual grounding pipeline. LLM-Grounder utilizes an LLM to decompose complex natural language queries into semantic constituents and employs a visual grounding tool, such as OpenScene or LERF, to identify objects in a 3D scene. The LLM then evaluates the spatial and commonsense relations among the proposed objects to make a final grounding decision. Our method does not require any labeled training data and can generalize to novel 3D scenes and arbitrary text queries. We evaluate LLM-Grounder on the ScanRefer benchmark and demonstrate state-of-the-art zero-shot grounding accuracy. Our findings indicate that LLMs significantly improve the grounding capability, especially for complex language queries, making LLM-Grounder an effective approach for 3D vision-language tasks in robotics."
                },
                {
                    "title": "NExT-GPT: Any-to-Any Multimodal LLM",
                    "abstract": "While recently Multimodal Large Language Models (MM-LLMs) have made exciting strides, they mostly fall prey to the limitation of only input-side multimodal understanding, without the ability to produce content in multiple modalities. As we humans always perceive the world and communicate with people through various modalities, developing any-to-any MM-LLMs capable of accepting and delivering content in any modality becomes essential to human-level AI. To fill the gap, we present an end-to-end general-purpose any-to-any MM-LLM system, NExT-GPT. We connect an LLM with multimodal adaptors and different diffusion decoders, enabling NExT-GPT to perceive inputs and generate outputs in arbitrary combinations of text, images, videos, and audio. By leveraging the existing well-trained highly-performing encoders and decoders, NExT-GPT is tuned with only a small amount of parameter (1%) of certain projection layers, which not only benefits low-cost training and also facilitates convenient expansion to more potential modalities. Moreover, we introduce a modality-switching instruction tuning (MosIT) and manually curate a high-quality dataset for MosIT, based on which NExT-GPT is empowered with complex cross-modal semantic understanding and content generation. Overall, our research showcases the promising possibility of building an AI agent capable of modeling universal modalities, paving the way for more human-like AI research in the community. Project page: https://next-gpt.github.io/"
                },
                {
                    "title": "3D Gaussian Splatting for Real-Time Radiance Field Rendering",
                    "abstract": "Radiance Field methods have recently revolutionized novel-view synthesis of scenes captured with multiple photos or videos. However, achieving high visual quality still requires neural networks that are costly to train and render, while recent faster methods inevitably trade off speed for quality. For unbounded and complete scenes (rather than isolated objects) and 1080p resolution rendering, no current method can achieve real-time display rates. We introduce three key elements that allow us to achieve state-of-the-art visual quality while maintaining competitive training times and importantly allow high-quality real-time (\u2265 30 fps) novel-view synthesis at 1080p resolution. First, starting from sparse points produced during camera calibration, we represent the scene with 3D Gaussians that preserve desirable properties of continuous volumetric radiance fields for scene optimization while avoiding unnecessary computation in empty space; Second, we perform interleaved optimization/density control of the 3D Gaussians, notably optimizing anisotropic covariance to achieve an accurate representation of the scene; Third, we develop a fast visibility-aware rendering algorithm that supports anisotropic splatting and both accelerates training and allows realtime rendering. We demonstrate state-of-the-art visual quality and real-time rendering on several established datasets."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "MotionGPT: Human Motion as a Foreign Language",
                    "abstract": "Though the advancement of pre-trained large language models unfolds, the exploration of building a unified model for language and other multi-modal data, such as motion, remains challenging and untouched so far. Fortunately, human motion displays a semantic coupling akin to human language, often perceived as a form of body language. By fusing language data with large-scale motion models, motion-language pre-training that can enhance the performance of motion-related tasks becomes feasible. Driven by this insight, we propose MotionGPT, a unified, versatile, and user-friendly motion-language model to handle multiple motion-relevant tasks. Specifically, we employ the discrete vector quantization for human motion and transfer 3D motion into motion tokens, similar to the generation process of word tokens. Building upon this\"motion vocabulary\", we perform language modeling on both motion and text in a unified manner, treating human motion as a specific language. Moreover, inspired by prompt learning, we pre-train MotionGPT with a mixture of motion-language data and fine-tune it on prompt-based question-and-answer tasks. Extensive experiments demonstrate that MotionGPT achieves state-of-the-art performances on multiple motion tasks including text-driven motion generation, motion captioning, motion prediction, and motion in-between."
                },
                {
                    "title": "Auto-GPT for Online Decision Making: Benchmarks and Additional Opinions",
                    "abstract": "Auto-GPT is an autonomous agent that leverages recent advancements in adapting Large Language Models (LLMs) for decision-making tasks. While there has been a growing interest in Auto-GPT stypled agents, questions remain regarding the effectiveness and flexibility of Auto-GPT in solving real-world decision-making tasks. Its limited capability for real-world engagement and the absence of benchmarks contribute to these uncertainties. In this paper, we present a comprehensive benchmark study of Auto-GPT styled agents in decision-making tasks that simulate real-world scenarios. Our aim is to gain deeper insights into this problem and understand the adaptability of GPT-based agents. We compare the performance of popular LLMs such as GPT-4, GPT-3.5, Claude, and Vicuna in Auto-GPT styled decision-making tasks. Furthermore, we introduce the Additional Opinions algorithm, an easy and effective method that incorporates supervised/imitation-based learners into the Auto-GPT scheme. This approach enables lightweight supervised learning without requiring fine-tuning of the foundational LLMs. We demonstrate through careful baseline comparisons and ablation studies that the Additional Opinions algorithm significantly enhances performance in online decision-making benchmarks, including WebShop and ALFWorld."
                },
                {
                    "title": "Any-to-Any Generation via Composable Diffusion",
                    "abstract": "We present Composable Diffusion (CoDi), a novel generative model capable of generating any combination of output modalities, such as language, image, video, or audio, from any combination of input modalities. Unlike existing generative AI systems, CoDi can generate multiple modalities in parallel and its input is not limited to a subset of modalities like text or image. Despite the absence of training datasets for many combinations of modalities, we propose to align modalities in both the input and output space. This allows CoDi to freely condition on any input combination and generate any group of modalities, even if they are not present in the training data. CoDi employs a novel composable generation strategy which involves building a shared multimodal space by bridging alignment in the diffusion process, enabling the synchronized generation of intertwined modalities, such as temporally aligned video and audio. Highly customizable and flexible, CoDi achieves strong joint-modality generation quality, and outperforms or is on par with the unimodal state-of-the-art for single-modality synthesis. The project page with demonstrations and code is at https://codi-gen.github.io"
                },
                {
                    "title": "ImageBind One Embedding Space to Bind Them All",
                    "abstract": "We present ImageBind, an approach to learn a joint embedding across six different modalities - images, text, audio, depth, thermal, and IMU data. We show that all combinations of paired data are not necessary to train such a joint embedding, and only image-paired data is sufficient to bind the modalities together. ImageBind can leverage recent large scale vision-language models, and extends their zero-shot capabilities to new modalities just by using their natural pairing with images. It enables novel emergent applications \u2018out-of-the-box\u2019 including cross-modal retrieval, composing modalities with arithmetic, cross-modal detection and generation. The emergent capabilities improve with the strength of the image encoder and we set a new state-of-the-art on emergent zero-shot recognition tasks across modalities, outperforming specialist supervised models. Finally, we show strong few-shot recognition results outperforming prior work, and that ImageBind serves as a new way to evaluate vision models for visual and non-visual tasks."
                },
                {
                    "title": "Zip-NeRF: Anti-Aliased Grid-Based Neural Radiance Fields",
                    "abstract": "Neural Radiance Field training can be accelerated through the use of grid-based representations in NeRF\u2019s learned mapping from spatial coordinates to colors and volumetric density. However, these grid-based approaches lack an explicit understanding of scale and therefore often introduce aliasing, usually in the form of jaggies or missing scene content. Anti-aliasing has previously been addressed by mip-NeRF 360, which reasons about sub-volumes along a cone rather than points along a ray, but this approach is not natively compatible with current grid-based techniques. We show how ideas from rendering and signal processing can be used to construct a technique that combines mip-NeRF 360 and grid-based models such as Instant NGP to yield error rates that are 8% \u2013 77% lower than either prior technique, and that trains 24\u00d7 faster than mip-NeRF 360."
                },
                {
                    "title": "Instance Neural Radiance Field",
                    "abstract": "This paper presents one of the first learning-based NeRF 3D instance segmentation pipelines, dubbed as Instance Neural Radiance Field, or Instance-NeRF. Taking a NeRF pretrained from multi-view RGB images as input, Instance-NeRF can learn 3D instance segmentation of a given scene, represented as an instance field component of the NeRF model. To this end, we adopt a 3D proposal-based mask prediction network on the sampled volumetric features from NeRF, which generates discrete 3D instance masks. The coarse 3D mask prediction is then projected to image space to match 2D segmentation masks from different views generated by existing panoptic segmentation models, which are used to supervise the training of the instance field. Notably, beyond generating consistent 2D segmentation maps from novel views, Instance-NeRF can query instance information at any 3D point, which greatly enhances NeRF object segmentation and manipulation. Our method is also one of the first to achieve such results in pure inference. Experimented on synthetic and real-world NeRF datasets with complex indoor scenes, Instance-NeRF surpasses previous NeRF segmentation works and competitive 2D segmentation methods in segmentation performance on unseen views. Code and data are available at https://github.com/lyclyc52/Instance_NeRF."
                },
                {
                    "title": "Segment Anything",
                    "abstract": "We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive \u2013 often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images at segment-anything.com to foster research into foundation models for computer vision. We recommend reading the full paper at: arxiv.org/abs/2304.02643."
                },
                {
                    "title": "LERF: Language Embedded Radiance Fields",
                    "abstract": "Humans describe the physical world using natural language to refer to specific 3D locations based on a vast range of properties: visual appearance, semantics, abstract associations, or actionable affordances. In this work we propose Language Embedded Radiance Fields (LERFs), a method for grounding language embeddings from off-the-shelf models like CLIP into NeRF, which enable these types of open-ended language queries in 3D. LERF learns a dense, multi-scale language field inside NeRF by volume rendering CLIP embeddings along training rays, supervising these embeddings across training views to provide multi-view consistency and smooth the underlying language field. After optimization, LERF can extract 3D relevancy maps for a broad range of language prompts interactively in real-time, which has potential use cases in robotics, understanding vision-language models, and interacting with 3D scenes. LERF enables pixel-aligned, zero-shot queries on the distilled 3D CLIP embeddings without relying on region proposals or masks, supporting long-tail open-vocabulary queries hierarchically across the volume. See the project website at: https://lerf.io."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Language Is Not All You Need: Aligning Perception with Language Models",
                    "abstract": "A big convergence of language, multimodal perception, action, and world modeling is a key step toward artificial general intelligence. In this work, we introduce Kosmos-1, a Multimodal Large Language Model (MLLM) that can perceive general modalities, learn in context (i.e., few-shot), and follow instructions (i.e., zero-shot). Specifically, we train Kosmos-1 from scratch on web-scale multimodal corpora, including arbitrarily interleaved text and images, image-caption pairs, and text data. We evaluate various settings, including zero-shot, few-shot, and multimodal chain-of-thought prompting, on a wide range of tasks without any gradient updates or finetuning. Experimental results show that Kosmos-1 achieves impressive performance on (i) language understanding, generation, and even OCR-free NLP (directly fed with document images), (ii) perception-language tasks, including multimodal dialogue, image captioning, visual question answering, and (iii) vision tasks, such as image recognition with descriptions (specifying classification via text instructions). We also show that MLLMs can benefit from cross-modal transfer, i.e., transfer knowledge from language to multimodal, and from multimodal to language. In addition, we introduce a dataset of Raven IQ test, which diagnoses the nonverbal reasoning capability of MLLMs."
                },
                {
                    "title": "Toolformer: Language Models Can Teach Themselves to Use Tools",
                    "abstract": "Language models (LMs) exhibit remarkable abilities to solve new tasks from just a few examples or textual instructions, especially at scale. They also, paradoxically, struggle with basic functionality, such as arithmetic or factual lookup, where much simpler and smaller models excel. In this paper, we show that LMs can teach themselves to use external tools via simple APIs and achieve the best of both worlds. We introduce Toolformer, a model trained to decide which APIs to call, when to call them, what arguments to pass, and how to best incorporate the results into future token prediction. This is done in a self-supervised way, requiring nothing more than a handful of demonstrations for each API. We incorporate a range of tools, including a calculator, a Q\\&A system, two different search engines, a translation system, and a calendar. Toolformer achieves substantially improved zero-shot performance across a variety of downstream tasks, often competitive with much larger models, without sacrificing its core language modeling abilities."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "K-Planes: Explicit Radiance Fields in Space, Time, and Appearance",
                    "abstract": "We introduce k-planes, a white-box model for radiance fields in arbitrary dimensions. Our model uses planes to represent a d-dimensional scene, providing a seamless way to go from static (d = 3) to dynamic (d= 4) scenes. This planar factorization makes adding dimension-specific priors easy, e.g. temporal smoothness and multi-resolution spatial structure, and induces a natural decomposition of static and dynamic components of a scene. We use a linear feature decoder with a learned color basis that yields similar performance as a nonlinear black-box MLP decoder. Across a range of synthetic and real, static and dynamic, fixed and varying appearance scenes, k-planes yields competitive and often state-of-the-art recon- struction fidelity with low memory usage, achieving 1000x compression over a full 4D grid, and fast optimization with a pure PyTorch implementation. For video results and code, please see sarafridov.github.io/K-Planes."
                },
                {
                    "title": "HexPlane: A Fast Representation for Dynamic Scenes",
                    "abstract": "Modeling and re-rendering dynamic 3D scenes is a challenging task in 3D vision. Prior approaches build on NeRF and rely on implicit representations. This is slow since it requires many MLP evaluations, constraining real-world applications. We show that dynamic 3D scenes can be explicitly represented by six planes of learned features, leading to an elegant solution we call HexPlane. A HexPlane computes features for points in spacetime by fusing vectors extracted from each plane, which is highly efficient. Pairing a HexPlane with a tiny MLP to regress output colors and training via volume rendering gives impressive results for novel view synthesis on dynamic scenes, matching the image quality of prior work but reducing training time by more than 100\u00d7. Extensive ablations confirm our HexPlane design and show that it is robust to different feature fusion mechanisms, coordinate systems, and decoding mechanisms. HexPlane is a simple and effective solution for representing 4D volumes, and we hope they can broadly contribute to modeling spacetime for dynamic 3D scenes.11Project page: https://caoang327.github.io/HexPlane."
                },
                {
                    "title": "A Large-Scale Outdoor Multi-modal Dataset and Benchmark for Novel View Synthesis and Implicit Scene Reconstruction",
                    "abstract": "Neural Radiance Fields (NeRF) [24] has achieved impressive results in single object scene reconstruction and novel view synthesis, as demonstrated on many single modality and single object focused indoor scene datasets like DTU [14], BMVS [42], and NeRF Synthetic [24]. However, the study of NeRF on large-scale outdoor scene reconstruction is still limited, as there is no unified outdoor scene dataset for large-scale NeRF evaluation due to expensive data acquisition and calibration costs. In this work, we propose a large-scale outdoor multi-modal dataset, OMMO dataset, containing complex objects and scenes with calibrated images, point clouds and prompt annotations. A new benchmark for several outdoor NeRF-based tasks is established, such as novel view synthesis, diverse 3D representation, and multi-modal NeRF. To create the dataset, we capture and collect a large number of real fly-view videos and select high-quality and high-resolution clips from them. Then we design a quality review module to refine images, remove low-quality frames and fail-to-calibrate scenes through a learning-based automatic evaluation plus manual review. Finally, volunteers are employed to label and review the prompt annotation for each scene and keyframe. Compared with existing NeRF datasets, our dataset contains abundant real-world urban and natural scenes with various scales, camera trajectories, and lighting conditions. Experiments show that our dataset can benchmark most state-of-the-art NeRF methods on different tasks. The dataset can be found at the following link: https://ommo.luchongshan.com/."
                },
                {
                    "title": "Generating Human Motion from Textual Descriptions with Discrete Representations",
                    "abstract": "In this work, we investigate a simple and must-known conditional generative framework based on Vector Quantised-Variational AutoEncoder (VQ-VAE) and Generative Pre-trained Transformer (GPT) for human motion generation from textural descriptions. We show that a simple CNN-based VQ-VAE with commonly used training recipes (EMA and Code Reset) allows us to obtain high-quality discrete representations. For GPT, we incorporate a simple corruption strategy during the training to alleviate training-testing discrepancy. Despite its simplicity, our T2M-GPT shows better performance than competitive approaches, including recent diffusion-based approaches. For example, on HumanML3D, which is currently the largest dataset, we achieve comparable performance on the consistency between text and generated motion (R-Precision), but with FID 0.116 largely outperforming MotionDiffuse of 0.630. Additionally, we conduct analyses on HumanML3D and observe that the dataset size is a limitation of our approach. Our work suggests that VQ-VAE still remains a competitive approach for human motion generation. Our implementation is available on the project page: https://mael-zys.github.io/T2M-GPT/."
                },
                {
                    "title": "HyperReel: High-Fidelity 6-DoF Video with Ray-Conditioned Sampling",
                    "abstract": "Volumetric scene representations enable photorealistic view synthesis for static scenes and form the basis of several existing 6-DoF video techniques. However, the volume rendering procedures that drive these representations necessitate careful trade-offs in terms of quality, rendering speed, and memory efficiency. In particular, existing methods fail to simultaneously achieve real-time performance, small memory footprint, and high-quality rendering for challenging real-world scenes. To address these issues, we present HyperReel\u2015a novel 6-DoF video representation. The two core components of HyperReel are: (1) a ray-conditioned sample prediction network that enables high-fidelity, high frame rate rendering at high resolutions and (2) a compact and memory-efficient dynamic volume representation. Our 6-DoF video pipeline achieves the best performance compared to prior and contemporary approaches in terms of visual quality with small memory requirements, while also rendering at up to 18 frames-per-second at megapixel resolution without any custom CUDA code."
                },
                {
                    "title": "Interactive Segmentation of Radiance Fields",
                    "abstract": "Radiance Fields (RF) are popular to represent casually-captured scenes for new view synthesis and several applications beyond it. Mixed reality on personal spaces needs understanding and manipulating scenes represented as RFs, with semantic segmentation of objects as an important step. Prior segmentation efforts show promise but don't scale to complex objects with diverse appearance. We present the ISRF method to interactively segment objects with fine structure and appearance. Nearest neighbor feature matching using distilled semantic features identifies high-confidence seed regions. Bilateral search in a joint spatio-semantic space grows the region to recover accurate segmentation. We show state-of-the-art results of segmenting objects from RFs and compositing them to another scene, changing appearance, etc., and an interactive segmentation tool that others can use."
                },
                {
                    "title": "Panoptic Lifting for 3D Scene Understanding with Neural Fields",
                    "abstract": "We propose Panoptic Lifting, a novel approach for learning panoptic 3D volumetric representations from images of in-the-wild scenes. Once trained, our model can render color images together with 3D-consistent panoptic segmentation from novel viewpoints. Unlike existing approaches which use 3D input directly or indirectly, our method requires only machine-generated 2D panoptic segmentation masks inferred from a pre-trained network. Our core contribution is a panoptic lifting scheme based on a neural field representation that generates a unified and multi-view consistent, 3D panoptic representation of the scene. To account for inconsistencies of 2D instance identifiers across views, we solve a linear assignment with a cost based on the model's current predictions and the machine-generated segmentation masks, thus enabling us to lift 2D instances to 3D in a consistent way. We further propose and ablate contributions that make our method more robust to noisy, machine-generated labels, including test-time augmentations for confidence estimates, segment consistency loss, bounded segmentation fields, and gradient stopping. Experimental results validate our approach on the challenging Hypersim, Replica, and ScanNet datasets, improving by 8.4, 13.8, and 10.6% in scene-level PQ over state of the art."
                },
                {
                    "title": "OpenScene: 3D Scene Understanding with Open Vocabularies",
                    "abstract": "Traditional 3D scene understanding approaches rely on labeled 3D datasets to train a model for a single task with supervision. We propose OpenScene, an alternative approach where a model predicts dense features for 3D scene points that are co-embedded with text and image pixels in CLIP feature space. This zero-shot approach enables task-agnostic training and open-vocabulary queries. For example, to perform SOTA zero-shot 3D semantic segmentation it first infers CLIP features for every 3D point and later classifies them based on similarities to embeddings of arbitrary class labels. More interestingly, it enables a suite of open-vocabulary scene understanding applications that have never been done before. For example, it allows a user to enter an arbitrary text query and then see a heat map indicating which parts of a scene match. Our approach is effective at identifying objects, materials, affordances, activities, and room types in complex 3D scenes, all using a single model trained without any labeled 3D data."
                },
                {
                    "title": "SPIn-NeRF: Multiview Segmentation and Perceptual Inpainting with Neural Radiance Fields",
                    "abstract": "Neural Radiance Fields (NeRFs) have emerged as a popular approach for novel view synthesis. While NeRFs are quickly being adapted for a wider set of applications, intuitively editing NeRF scenes is still an open challenge. One important editing task is the removal of unwanted objects from a 3D scene, such that the replaced region is visually plausible and consistent with its context. We refer to this task as 3D inpainting. In 3D, solutions must be both consistent across multiple views and geometrically valid. In this paper, we propose a novel 3D inpainting method that addresses these challenges. Given a small set of posed images and sparse annotations in a single input image, our framework first rapidly obtains a 3D segmentation mask for a target object. Using the mask, a perceptual optimization-based approach is then introduced that leverages learned 2D image inpainters, distilling their information into 3D space, while ensuring view consistency. We also address the lack of a diverse benchmark for evaluating 3D scene inpainting methods by introducing a dataset comprised of challenging real-world scenes. In particular, our dataset contains views of the same scene with and without a target object, enabling more principled benchmarking of the 3D inpainting task. We first demonstrate the superiority of our approach on multiview segmentation, comparing to NeRF-based methods and 2D segmentation approaches. We then evaluate on the task of 3D inpainting, establishing state-of-the-art performance against other NeRF manipulation algorithms, as well as a strong 2D image inpainter baseline."
                },
                {
                    "title": "NeRF-RPN: A general framework for object detection in NeRFs",
                    "abstract": "This paper presents the first significant object detection framework, NeRF-RPN, which directly operates on NeRF. Given a pre-trained NeRF model, NeRF-RPN aims to detect all bounding boxes of objects in a scene. By exploiting a novel voxel representation that incorporates multi-scale 3D neural volumetric features, we demonstrate it is possible to regress the 3D bounding boxes of objects in NeRF directly without rendering the NeRF at any viewpoint. NeRF-RPN is a general framework and can be applied to detect objects without class labels. We experimented NeRF-RPN with various backbone architectures, RPN head designs and loss functions. All of them can be trained in an end-to-end manner to estimate high quality 3D bounding boxes. To facilitate future research in object detection for NeRF, we built a new benchmark dataset which consists of both synthetic and real-world data with careful labeling and clean up. Code and dataset are available at htt ps: //github.com/lyclyc52/NeRF_RPN."
                },
                {
                    "title": "Unsupervised Multi-View Object Segmentation Using Radiance Field Propagation",
                    "abstract": "We present radiance field propagation (RFP), a novel approach to segmenting objects in 3D during reconstruction given only unlabeled multi-view images of a scene. RFP is derived from emerging neural radiance field-based techniques, which jointly encodes semantics with appearance and geometry. The core of our method is a novel propagation strategy for individual objects' radiance fields with a bidirectional photometric loss, enabling an unsupervised partitioning of a scene into salient or meaningful regions corresponding to different object instances. To better handle complex scenes with multiple objects and occlusions, we further propose an iterative expectation-maximization algorithm to refine object masks. RFP is one of the first unsupervised approach for tackling 3D real scene object segmentation for neural radiance field (NeRF) without any supervision, annotations, or other cues such as 3D bounding boxes and prior knowledge of object class. Experiments demonstrate that RFP achieves feasible segmentation results that are more accurate than previous unsupervised image/scene segmentation approaches, and are comparable to existing supervised NeRF-based methods. The segmented object representations enable individual 3D object editing operations."
                },
                {
                    "title": "Unified-IO: A Unified Model for Vision, Language, and Multi-Modal Tasks",
                    "abstract": "We propose Unified-IO, a model that performs a large variety of AI tasks spanning classical computer vision tasks, including pose estimation, object detection, depth estimation and image generation, vision-and-language tasks such as region captioning and referring expression, to natural language processing tasks such as question answering and paraphrasing. Developing a single unified model for such a large variety of tasks poses unique challenges due to the heterogeneous inputs and outputs pertaining to each task, including RGB images, per-pixel maps, binary masks, bounding boxes, and language. We achieve this unification by homogenizing every supported input and output into a sequence of discrete vocabulary tokens. This common representation across all tasks allows us to train a single transformer-based architecture, jointly on over 90 diverse datasets in the vision and language fields. Unified-IO is the first model capable of performing all 7 tasks on the GRIT benchmark and produces strong results across 16 diverse benchmarks like NYUv2-Depth, ImageNet, VQA2.0, OK-VQA, Swig, VizWizGround, BoolQ, and SciTail, with no task-specific fine-tuning. Code and demos for Unified-IO are available at: https://unified-io.allenai.org."
                },
                {
                    "title": "Decomposing NeRF for Editing via Feature Field Distillation",
                    "abstract": "Emerging neural radiance fields (NeRF) are a promising scene representation for computer graphics, enabling high-quality 3D reconstruction and novel view synthesis from image observations. However, editing a scene represented by a NeRF is challenging, as the underlying connectionist representations such as MLPs or voxel grids are not object-centric or compositional. In particular, it has been difficult to selectively edit specific regions or objects. In this work, we tackle the problem of semantic scene decomposition of NeRFs to enable query-based local editing of the represented 3D scenes. We propose to distill the knowledge of off-the-shelf, self-supervised 2D image feature extractors such as CLIP-LSeg or DINO into a 3D feature field optimized in parallel to the radiance field. Given a user-specified query of various modalities such as text, an image patch, or a point-and-click selection, 3D feature fields semantically decompose 3D space without the need for re-training and enable us to semantically select and edit regions in the radiance field. Our experiments validate that the distilled feature fields (DFFs) can transfer recent progress in 2D vision and language foundation models to 3D scene representations, enabling convincing 3D segmentation and selective editing of emerging neural graphics representations."
                },
                {
                    "title": "Neural Volumetric Object Selection",
                    "abstract": "We introduce an approach for selecting objects in neural volumetric 3D representations, such as multi-plane images (MPI) and neural radiance fields (NeRF). Our approach takes a set of foreground and background 2D user scribbles in one view and automatically estimates a 3D segmentation of the desired object, which can be rendered into novel views. To achieve this result, we propose a novel voxel feature embedding that incorporates the neural volumetric 3D representation and multi-view image features from all input views. To evaluate our approach, we introduce a new dataset of human-provided segmentation masks for depicted objects in real-world multi-view scene captures. We show that our approach out-performs strong baselines, including 2D segmentation and 3D segmentation approaches adapted to our task."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "Point Scene Understanding via Disentangled Instance Mesh Reconstruction",
                    "abstract": "Semantic scene reconstruction from point cloud is an essential and challenging task for 3D scene understanding. This task requires not only to recognize each instance in the scene, but also to recover their geometries based on the partial observed point cloud. Existing methods usually attempt to directly predict occupancy values of the complete object based on incomplete point cloud proposals from a detection-based backbone. However, this framework always fails to reconstruct high fidelity mesh due to the obstruction of various detected false positive object proposals and the ambiguity of incomplete point observations for learning occupancy values of complete objects. To circumvent the hurdle, we propose a Disentangled Instance Mesh Reconstruction (DIMR) framework for effective point scene understanding. A segmentation-based backbone is applied to reduce false positive object proposals, which further benefits our exploration on the relationship between recognition and reconstruction. Based on the accurate proposals, we leverage a mesh-aware latent code space to disentangle the processes of shape completion and mesh generation, relieving the ambiguity caused by the incomplete point observations. Furthermore, with access to the CAD model pool at test time, our model can also be used to improve the reconstruction quality by performing mesh retrieval without extra training. We thoroughly evaluate the reconstructed mesh quality with multiple metrics, and demonstrate the superiority of our method on the challenging ScanNet dataset. Code is available at \\url{https://github.com/ashawkey/dimr}."
                },
                {
                    "title": "TensoRF: Tensorial Radiance Fields",
                    "abstract": "We present TensoRF, a novel approach to model and reconstruct radiance fields. Unlike NeRF that purely uses MLPs, we model the radiance field of a scene as a 4D tensor, which represents a 3D voxel grid with per-voxel multi-channel features. Our central idea is to factorize the 4D scene tensor into multiple compact low-rank tensor components. We demonstrate that applying traditional CP decomposition -- that factorizes tensors into rank-one components with compact vectors -- in our framework leads to improvements over vanilla NeRF. To further boost performance, we introduce a novel vector-matrix (VM) decomposition that relaxes the low-rank constraints for two modes of a tensor and factorizes tensors into compact vector and matrix factors. Beyond superior rendering quality, our models with CP and VM decompositions lead to a significantly lower memory footprint in comparison to previous and concurrent works that directly optimize per-voxel features. Experimentally, we demonstrate that TensoRF with CP decomposition achieves fast reconstruction (<30 min) with better rendering quality and even a smaller model size (<4 MB) compared to NeRF. Moreover, TensoRF with VM decomposition further boosts rendering quality and outperforms previous state-of-the-art methods, while reducing the reconstruction time (<10 min) and retaining a compact model size (<75 MB)."
                },
                {
                    "title": "Fourier PlenOctrees for Dynamic Radiance Field Rendering in Real-time",
                    "abstract": "Implicit neural representations such as Neural Radiance Field (NeRF) have focused mainly on modeling static objects captured under multi-view settings where real-time rendering can be achieved with smart data structures, e.g., PlenOctree. In this paper, we present a novel Fourier PlenOctree (FPO) technique to tackle efficient neural mod-eling and real-time rendering of dynamic scenes captured under the free-view video (FVV) setting. The key idea in our FPO is a novel combination of generalized NeRF, PlenOctree representation, volumetric fusion and Fourier transform. To accelerate FPO construction, we present a novel coarse-to-fine fusion scheme that leverages the gen-eralizable NeRF technique to generate the tree via spatial blending. To tackle dynamic scenes, we tailor the implicit network to model the Fourier coefficients of time-varying density and color attributes. Finally, we construct the FPO and train the Fourier coefficients directly on the leaves of a union PlenOctree structure of the dynamic sequence. We show that the resulting FPO enables compact memory overload to handle dynamic objects and supports efficient fine-tuning. Extensive experiments show that the proposed method is 3000 times faster than the original NeRF and achieves over an order of magnitude acceleration over SOTA while preserving high visual quality for the free-viewpoint rendering of unseen dynamic scenes."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation",
                    "abstract": "Vision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has been largely achieved by scaling up the dataset with noisy image-text pairs collected from the web, which is a suboptimal source of supervision. In this paper, we propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks. BLIP effectively utilizes the noisy web data by bootstrapping the captions, where a captioner generates synthetic captions and a filter removes the noisy ones. We achieve state-of-the-art results on a wide range of vision-language tasks, such as image-text retrieval (+2.7% in average recall@1), image captioning (+2.8% in CIDEr), and VQA (+1.6% in VQA score). BLIP also demonstrates strong generalization ability when directly transferred to video-language tasks in a zero-shot manner. Code, models, and datasets are released at https://github.com/salesforce/BLIP."
                },
                {
                    "title": "Instant neural graphics primitives with a multiresolution hash encoding",
                    "abstract": "Neural graphics primitives, parameterized by fully connected neural networks, can be costly to train and evaluate. We reduce this cost with a versatile new input encoding that permits the use of a smaller network without sacrificing quality, thus significantly reducing the number of floating point and memory access operations: a small neural network is augmented by a multiresolution hash table of trainable feature vectors whose values are optimized through stochastic gradient descent. The multiresolution structure allows the network to disambiguate hash collisions, making for a simple architecture that is trivial to parallelize on modern GPUs. We leverage this parallelism by implementing the whole system using fully-fused CUDA kernels with a focus on minimizing wasted bandwidth and compute operations. We achieve a combined speedup of several orders of magnitude, enabling training of high-quality neural graphics primitives in a matter of seconds, and rendering in tens of milliseconds at a resolution of 1920\u00d71080."
                },
                {
                    "title": "Language-driven Semantic Segmentation",
                    "abstract": "We present LSeg, a novel model for language-driven semantic image segmentation. LSeg uses a text encoder to compute embeddings of descriptive input labels (e.g.,\"grass\"or\"building\") together with a transformer-based image encoder that computes dense per-pixel embeddings of the input image. The image encoder is trained with a contrastive objective to align pixel embeddings to the text embedding of the corresponding semantic class. The text embeddings provide a flexible label representation in which semantically similar labels map to similar regions in the embedding space (e.g.,\"cat\"and\"furry\"). This allows LSeg to generalize to previously unseen categories at test time, without retraining or even requiring a single additional training sample. We demonstrate that our approach achieves highly competitive zero-shot performance compared to existing zero- and few-shot semantic segmentation methods, and even matches the accuracy of traditional segmentation algorithms when a fixed label set is provided. Code and demo are available at https://github.com/isl-org/lang-seg."
                },
                {
                    "title": "ScanQA: 3D Question Answering for Spatial Scene Understanding",
                    "abstract": "We propose a new 3D spatial understanding task for 3D question answering (3D-QA). In the 3D-QA task, models receive visual information from the entire 3D scene of a rich RGB-D indoor scan and answer given textual questions about the 3D scene. Unlike the 2D-question answering of visual question answering, the conventional 2D-QA models suffer from problems with spatial understanding of object alignment and directions and fail in object localization from the textual questions in 3D-QA. We propose a baseline model for 3D-QA, called the ScanQA11https://github.com/ATR-DBI/ScanQA, which learns a fused descriptor from 3D object proposals and encoded sentence embeddings. This learned descriptor correlates language expressions with the underlying geometric features of the 3D scan and facilitates the regression of 3D bounding boxes to determine the described objects in textual questions. We collected human-edited question-answer pairs with free-form answers grounded in 3D objects in each 3D scene. Our new ScanQA dataset contains over 41k question-answer pairs from 800 indoor scenes obtained from the ScanNet dataset. To the best of our knowledge, ScanQA is the first large-scale effort to perform object-grounded question answering in 3D environments."
                },
                {
                    "title": "Plenoxels: Radiance Fields without Neural Networks",
                    "abstract": "We introduce Plenoxels (plenoptic voxels), a systemfor photorealistic view synthesis. Plenoxels represent a scene as a sparse 3D grid with spherical harmonics. This representation can be optimized from calibrated images via gradient methods and regularization without any neural components. On standard, benchmark tasks, Plenoxels are optimized two orders of magnitude faster than Neural Radiance Fields with no loss in visual quality. For video and code, please see https://alexyu.net/plenoxels."
                },
                {
                    "title": "CLIP-NeRF: Text-and-Image Driven Manipulation of Neural Radiance Fields",
                    "abstract": "We present CLIP-NeRF, a multi-modal 3D object manipulation method for neural radiance fields (NeRF). By leveraging the joint language-image embedding space of the recent Contrastive Language-Image Pre-Training (CLIP) model, we propose a unified framework that allows manip-ulating NeRF in a user-friendly way, using either a short text prompt or an exemplar image. Specifically, to combine the novel view synthesis capability of NeRF and the controllable manipulation ability of latent representations from generative models, we introduce a disentangled conditional NeRF architecture that allows individual control over both shape and appearance. This is achieved by performing the shape conditioning via applying a learned deformation field to the positional encoding and deferring color conditioning to the volumetric rendering stage. To bridge this disentangled latent representation to the CLIP embedding, we design two code mappers that take a CLIP embedding as input and update the latent codes to reflect the targeted editing. The mappers are trained with a CLIP-based matching loss to ensure the manipulation accuracy. Furthermore, we propose an inverse optimization method that accurately projects an input image to the latent codes for manipulation to enable editing on real images. We evaluate our approach by extensive experiments on a variety of text prompts and exemplar images and also provide an intuitive interface for interactive editing."
                },
                {
                    "title": "Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields",
                    "abstract": "Though neural radiance fields (NeRF) have demon-strated impressive view synthesis results on objects and small bounded regions of space, they struggle on \u201cun-bounded\u201d scenes, where the camera may point in any di-rection and content may exist at any distance. In this set-ting, existing NeRF-like models often produce blurry or low-resolution renderings (due to the unbalanced detail and scale of nearby and distant objects), are slow to train, and may exhibit artifacts due to the inherent ambiguity of the task of reconstructing a large scene from a small set of images. We present an extension of mip-NeRF (a NeRF variant that addresses sampling and aliasing) that uses a non-linear scene parameterization, online distillation, and a novel distortion-based regularizer to overcome the chal-lenges presented by unbounded scenes. Our model, which we dub \u201cmip-NeRF 360\u201d as we target scenes in which the camera rotates 360 degrees around a point, reduces mean-squared error by 57% compared to mip-NeRF, and is able to produce realistic synthesized views and detailed depth maps for highly intricate, unbounded real-world scenes."
                },
                {
                    "title": "Direct Voxel Grid Optimization: Super-fast Convergence for Radiance Fields Reconstruction",
                    "abstract": "We present a super-fast convergence approach to reconstructing the per-scene radiance field from a set of images that capture the scene with known poses. This task, which is often applied to novel view synthesis, is recently revolution-ized by Neural Radiance Field (NeRF) for its state-of-the-art quality and fiexibility. However, NeRF and its variants require a lengthy training time ranging from hours to days for a single scene. In contrast, our approach achieves NeRF-comparable quality and converges rapidly from scratch in less than 15 minutes with a single GPU. We adopt a representation consisting of a density voxel grid for scene geometry and a feature voxel grid with a shallow network for complex view-dependent appearance. Modeling with explicit and discretized volume representations is not new, but we propose two simple yet non-trivial techniques that contribute to fast convergence speed and high-quality output. First, we introduce the post-activation interpolation on voxel density, which is capable of producing sharp surfaces in lower grid resolution. Second, direct voxel density optimization is prone to suboptimal geometry solutions, so we robustify the optimization process by imposing several priors. Finally, evaluation on five inward-facing benchmarks shows that our method matches, if not surpasses, NeRF's quality, yet it only takes about 15 minutes to train from scratch for a new scene. Code: https://github.com/sunset1995/DirectVoxGO."
                },
                {
                    "title": "VideoCLIP: Contrastive Pre-training for Zero-shot Video-Text Understanding",
                    "abstract": "We present VideoCLIP, a contrastive approach to pre-train a unified model for zero-shot video and text understanding, without using any labels on downstream tasks. VideoCLIP trains a transformer for video and text by contrasting temporally overlapping positive video-text pairs with hard negatives from nearest neighbor retrieval. Our experiments on a diverse series of downstream tasks, including sequence-level text-video retrieval, VideoQA, token-level action localization, and action segmentation reveal state-of-the-art performance, surpassing prior work, and in some cases even outperforming supervised approaches. Code is made available at https://github.com/pytorch/fairseq/examples/MMPT."
                },
                {
                    "title": "CodeNeRF: Disentangled Neural Radiance Fields for Object Categories",
                    "abstract": "CodeNeRF is an implicit 3D neural representation that learns the variation of object shapes and textures across a category and can be trained, from a set of posed images, to synthesize novel views of unseen objects. Unlike the original NeRF, which is scene specific, CodeNeRF learns to disentangle shape and texture by learning separate embeddings. At test time, given a single unposed image of an unseen object, CodeNeRF jointly estimates camera viewpoint, and shape and appearance codes via optimization. Unseen objects can be reconstructed from a single image, and then rendered from new viewpoints or their shape and texture edited by varying the latent codes. We conduct experiments on the SRN benchmark, which show that CodeNeRF generalises well to unseen objects and achieves on-par performance with methods that require known camera pose at test time. Our results on real-world images demonstrate that CodeNeRF can bridge the sim-to-real gap. Project page: https://github.com/wayne1123/code-nerf"
                },
                {
                    "title": "Advances in Neural Rendering",
                    "abstract": "Synthesizing photo\u2010realistic images and videos is at the heart of computer graphics and has been the focus of decades of research. Traditionally, synthetic images of a scene are generated using rendering algorithms such as rasterization or ray tracing, which take specifically defined representations of geometry and material properties as input. Collectively, these inputs define the actual scene and what is rendered, and are referred to as the scene representation (where a scene consists of one or more objects). Example scene representations are triangle meshes with accompanied textures (e.g., created by an artist), point clouds (e.g., from a depth sensor), volumetric grids (e.g., from a CT scan), or implicit surface functions (e.g., truncated signed distance fields). The reconstruction of such a scene representation from observations using differentiable rendering losses is known as inverse graphics or inverse rendering. Neural rendering is closely related, and combines ideas from classical computer graphics and machine learning to create algorithms for synthesizing images from real\u2010world observations. Neural rendering is a leap forward towards the goal of synthesizing photo\u2010realistic image and video content. In recent years, we have seen immense progress in this field through hundreds of publications that show different ways to inject learnable components into the rendering pipeline. This state\u2010of\u2010the\u2010art report on advances in neural rendering focuses on methods that combine classical rendering principles with learned 3D scene representations, often now referred to as neural scene representations. A key advantage of these methods is that they are 3D\u2010consistent by design, enabling applications such as novel viewpoint synthesis of a captured scene. In addition to methods that handle static scenes, we cover neural scene representations for modeling non\u2010rigidly deforming objects and scene editing and composition. While most of these approaches are scene\u2010specific, we also discuss techniques that generalize across object classes and can be used for generative tasks. In addition to reviewing these state\u2010of\u2010the\u2010art methods, we provide an overview of fundamental concepts and definitions used in the current literature. We conclude with a discussion on open challenges and social implications."
                },
                {
                    "title": "Audioclip: Extending Clip to Image, Text and Audio",
                    "abstract": "The rapidly evolving field of sound classification has greatly benefited from the methods of other domains. Today, the trend is to fuse domain-specific tasks and approaches together, which provides the community with new outstanding models.We present AudioCLIP \u2013 an extension of the CLIP model that handles audio in addition to text and images. Utilizing the AudioSet dataset, our proposed model incorporates the ESResNeXt audio-model into the CLIP framework, thus enabling it to perform multimodal classification and keeping CLIP\u2019s zero-shot capabilities.AudioCLIP achieves new state-of-the-art results in the Environmental Sound Classification (ESC) task and out-performs others by reaching accuracies of 97.15 % on ESC-50 and 90.07 % on UrbanSound8K. Further, it sets new baselines in the zero-shot ESC-task on the same datasets (69.40 % and 68.78 %, respectively).We also asses the influence of different training setups on the final performance of the proposed model. For the sake of reproducibility, our code is published."
                },
                {
                    "title": "Editing Conditional Radiance Fields",
                    "abstract": "A neural radiance field (NeRF) is a scene model supporting high-quality view synthesis, optimized per scene. In this paper, we explore enabling user editing of a category-level NeRF \u2013 also known as a conditional radiance field \u2013 trained on a shape category. Specifically, we introduce a method for propagating coarse 2D user scribbles to the 3D space, to modify the color or shape of a local region. First, we propose a conditional radiance field that incorporates new modular network components, including a shape branch that is shared across object instances. Observing multiple instances of the same category, our model learns underlying part semantics without any supervision, thereby allowing the propagation of coarse 2D user scribbles to the entire 3D region (e.g., chair seat). Next, we propose a hybrid network update strategy that targets specific network components, which balances efficiency and accuracy. During user interaction, we formulate an optimization problem that both satisfies the user\u2019s constraints and preserves the original object structure. We demonstrate our editing approach on rendered views of three shape datasets and show that it outperforms prior neural editing approaches. Finally, we edit the appearance and shape of a single-view real photograph and show that the edit propagates to extrapolated novel views."
                },
                {
                    "title": "Editable free-viewpoint video using a layered neural representation",
                    "abstract": "Generating free-viewpoint videos is critical for immersive VR/AR experience, but recent neural advances still lack the editing ability to manipulate the visual perception for large dynamic scenes. To fill this gap, in this paper, we propose the first approach for editable free-viewpoint video generation for large-scale view-dependent dynamic scenes using only 16 cameras. The core of our approach is a new layered neural representation, where each dynamic entity, including the environment itself, is formulated into a spatio-temporal coherent neural layered radiance representation called ST-NeRF. Such a layered representation supports manipulations of the dynamic scene while still supporting a wide free viewing experience. In our ST-NeRF, we represent the dynamic entity/layer as a continuous function, which achieves the disentanglement of location, deformation as well as the appearance of the dynamic entity in a continuous and self-supervised manner. We propose a scene parsing 4D label map tracking to disentangle the spatial information explicitly and a continuous deform module to disentangle the temporal motion implicitly. An object-aware volume rendering scheme is further introduced for the re-assembling of all the neural layers. We adopt a novel layered loss and motion-aware ray sampling strategy to enable efficient training for a large dynamic scene with multiple performers, Our framework further enables a variety of editing functions, i.e., manipulating the scale and location, duplicating or retiming individual neural layers to create numerous visual effects while preserving high realism. Extensive experiments demonstrate the effectiveness of our approach to achieve high-quality, photo-realistic, and editable free-viewpoint video generation for dynamic scenes."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "In-Place Scene Labelling and Understanding with Implicit Scene Representation",
                    "abstract": "Semantic labelling is highly correlated with geometry and radiance reconstruction, as scene entities with similar shape and appearance are more likely to come from similar classes. Recent implicit neural reconstruction techniques are appealing as they do not require prior training data, but the same fully self-supervised approach is not possible for semantics because labels are human-defined properties.We extend neural radiance fields (NeRF) to jointly encode semantics with appearance and geometry, so that complete and accurate 2D semantic labels can be achieved using a small amount of in-place annotations specific to the scene. The intrinsic multi-view consistency and smoothness of NeRF benefit semantics by enabling sparse labels to efficiently propagate. We show the benefit of this approach when labels are either sparse or very noisy in room-scale scenes. We demonstrate its advantageous properties in various interesting applications such as an efficient scene labelling tool, novel semantic view synthesis, label denoising, super-resolution, label interpolation and multi-view semantic label fusion in visual semantic mapping systems."
                },
                {
                    "title": "PlenOctrees for Real-time Rendering of Neural Radiance Fields",
                    "abstract": "We introduce a method to render Neural Radiance Fields (NeRFs) in real time using PlenOctrees, an octree-based 3D representation which supports view-dependent effects. Our method can render 800\u00d7800 images at more than 150 FPS, which is over 3000 times faster than conventional NeRFs. We do so without sacrificing quality while preserving the ability of NeRFs to perform free-viewpoint rendering of scenes with arbitrary geometry and view-dependent effects. Real-time performance is achieved by pre-tabulating the NeRF into a PlenOctree. In order to preserve view-dependent effects such as specularities, we factorize the appearance via closed-form spherical basis functions. Specifically, we show that it is possible to train NeRFs to predict a spherical harmonic representation of radiance, removing the viewing direction as an input to the neural network. Furthermore, we show that PlenOctrees can be directly optimized to further minimize the reconstruction loss, which leads to equal or better quality compared to competing methods. Moreover, this octree optimization step can be used to reduce the training time, as we no longer need to wait for the NeRF training to converge fully. Our real-time neural rendering approach may potentially enable new applications such as 6-DOF industrial and product visualizations, as well as next generation AR/VR systems. PlenOctrees are amenable to in-browser rendering as well; please visit the project page for the interactive online demo, as well as video and code: https://alexyu.net/plenoctrees."
                },
                {
                    "title": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields",
                    "abstract": "The rendering procedure used by neural radiance fields (NeRF) samples a scene with a single ray per pixel and may therefore produce renderings that are excessively blurred or aliased when training or testing images observe scene content at different resolutions. The straightforward solution of supersampling by rendering with multiple rays per pixel is impractical for NeRF, because rendering each ray requires querying a multilayer perceptron hundreds of times. Our solution, which we call \"mip-NeRF\" (\u00e0 la \"mipmap\"), extends NeRF to represent the scene at a continuously-valued scale. By efficiently rendering anti-aliased conical frustums instead of rays, mip-NeRF reduces objectionable aliasing artifacts and significantly improves NeRF\u2019s ability to represent fine details, while also being 7% faster than NeRF and half the size. Compared to NeRF, mip-NeRF reduces average error rates by 17% on the dataset presented with NeRF and by 60% on a challenging multiscale variant of that dataset that we present. Mip-NeRF is also able to match the accuracy of a brute-force supersampled NeRF on our multiscale dataset while being 22\u00d7 faster."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Non-Rigid Neural Radiance Fields: Reconstruction and Novel View Synthesis of a Dynamic Scene From Monocular Video",
                    "abstract": "We present Non-Rigid Neural Radiance Fields (NR-NeRF), a reconstruction and novel view synthesis approach for general non-rigid dynamic scenes. Our approach takes RGB images of a dynamic scene as input (e.g., from a monocular video recording), and creates a high-quality space-time geometry and appearance representation. We show that a single handheld consumer-grade camera is sufficient to synthesize sophisticated renderings of a dynamic scene from novel virtual camera views, e.g. a \u2018bullet-time\u2019 video effect. NR-NeRF disentangles the dynamic scene into a canonical volume and its deformation. Scene deformation is implemented as ray bending, where straight rays are deformed non-rigidly. We also propose a novel rigidity network to better constrain rigid regions of the scene, leading to more stable results. The ray bending and rigidity network are trained without explicit supervision. Our formulation enables dense correspondence estimation across views and time, and compelling video editing applications such as motion exaggeration. Our code will be open sourced."
                },
                {
                    "title": "Scan2Cap: Context-aware Dense Captioning in RGB-D Scans",
                    "abstract": "We introduce the task of dense captioning in 3D scans from commodity RGB-D sensors. As input, we assume a point cloud of a 3D scene; the expected output is the bounding boxes along with the descriptions for the underlying objects. To address the 3D object detection and description problems, we propose Scan2Cap, an end-to-end trained method, to detect objects in the input scene and describe them in natural language. We use an attention mechanism that generates descriptive tokens while referring to the related components in the local context. To reflect object relations (i.e. relative spatial relations) in the generated captions, we use a message passing graph module to facilitate learning object relation features. Our method can effectively localize and describe 3D objects in scenes from the ScanRefer dataset, outperforming 2D baseline methods by a significant margin (27.61% CiDEr@0.5IoU improvement)."
                },
                {
                    "title": "D-NeRF: Neural Radiance Fields for Dynamic Scenes",
                    "abstract": "Neural rendering techniques combining machine learning with geometric reasoning have arisen as one of the most promising approaches for synthesizing novel views of a scene from a sparse set of images. Among these, stands out the Neural radiance fields (NeRF) [31], which trains a deep network to map 5D input coordinates (representing spatial location and viewing direction) into a volume density and view-dependent emitted radiance. However, despite achieving an unprecedented level of photorealism on the generated images, NeRF is only applicable to static scenes, where the same spatial location can be queried from different images. In this paper we introduce D-NeRF, a method that extends neural radiance fields to a dynamic domain, allowing to reconstruct and render novel images of objects under rigid and non-rigid motions from a single camera moving around the scene. For this purpose we consider time as an additional input to the system, and split the learning process in two main stages: one that encodes the scene into a canonical space and another that maps this canonical representation into the deformed scene at a particular time. Both mappings are simultaneously learned using fully-connected networks. Once the networks are trained, D-NeRF can render novel images, controlling both the camera view and the time variable, and thus, the object movement. We demonstrate the effectiveness of our approach on scenes with objects under rigid, articulated and non-rigid motions. Code, model weights and the dynamic scenes dataset will be available at [1]."
                },
                {
                    "title": "Nerfies: Deformable Neural Radiance Fields",
                    "abstract": "We present the first method capable of photorealistically reconstructing deformable scenes using photos/videos captured casually from mobile phones. Our approach augments neural radiance fields (NeRF) by optimizing an additional continuous volumetric deformation field that warps each observed point into a canonical 5D NeRF. We observe that these NeRF-like deformation fields are prone to local minima, and propose a coarse-to-fine optimization method for coordinate-based models that allows for more robust optimization. By adapting principles from geometry processing and physical simulation to NeRF-like models, we propose an elastic regularization of the deformation field that further improves robustness. We show that our method can turn casually captured selfie photos/videos into deformable NeRF models that allow for photorealistic renderings of the subject from arbitrary viewpoints, which we dub \"nerfies.\" We evaluate our method by collecting time-synchronized data using a rig with two mobile phones, yielding train/validation images of the same pose at different viewpoints. We show that our method faithfully reconstructs non-rigidly deforming scenes and reproduces unseen views with high fidelity."
                },
                {
                    "title": "Hypersim: A Photorealistic Synthetic Dataset for Holistic Indoor Scene Understanding",
                    "abstract": "For many fundamental scene understanding tasks, it is difficult or impossible to obtain per-pixel ground truth labels from real images. We address this challenge by introducing Hypersim, a photorealistic synthetic dataset for holistic indoor scene understanding. To create our dataset, we leverage a large repository of synthetic scenes created by professional artists, and we generate 77,400 images of 461 indoor scenes with detailed per-pixel labels and corresponding ground truth geometry. Our dataset: (1) relies exclusively on publicly available 3D assets; (2) includes complete scene geometry, material information, and lighting information for every scene; (3) includes dense per-pixel semantic instance segmentations and complete camera information for every image; and (4) factors every image into diffuse reflectance, diffuse illumination, and a non-diffuse residual term that captures view-dependent lighting effects.We analyze our dataset at the level of scenes, objects, and pixels, and we analyze costs in terms of money, computation time, and annotation effort. Remarkably, we find that it is possible to generate our entire dataset from scratch, for roughly half the cost of training a popular open-source natural language processing model. We also evaluate sim-to-real transfer performance on two real-world scene understanding tasks \u2013 semantic segmentation and 3D shape prediction \u2013 where we find that pre-training on our dataset significantly improves performance on both tasks, and achieves state-of-the-art performance on the most challenging Pix3D test set. All of our rendered image data, as well as all the code we used to generate our dataset and perform our experiments, is available online."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "3D Sketch-Aware Semantic Scene Completion via Semi-Supervised Structure Prior",
                    "abstract": "The goal of the Semantic Scene Completion (SSC) task is to simultaneously predict a completed 3D voxel representation of volumetric occupancy and semantic labels of objects in the scene from a single-view observation. Since the computational cost generally increases explosively along with the growth of voxel resolution, most current state-of-the-arts have to tailor their framework into a low-resolution representation with the sacrifice of detail prediction. Thus, voxel resolution becomes one of the crucial difficulties that lead to the performance bottleneck. In this paper, we propose to devise a new geometry-based strategy to embed depth information with low-resolution voxel representation, which could still be able to encode sufficient geometric information, e.g., room layout, object\u2019s sizes and shapes, to infer the invisible areas of the scene with well structure-preserving details. To this end, we first propose a novel 3D sketch-aware feature embedding to explicitly encode geometric information effectively and efficiently. With the 3D sketch in hand, we further devise a simple yet effective semantic scene completion framework that incorporates a light-weight 3D Sketch Hallucination module to guide the inference of occupancy and the semantic labels via a semi-supervised structure prior learning strategy. We demonstrate that our proposed geometric embedding works better than the depth feature learning from habitual SSC frameworks. Our final model surpasses state- of-the-arts consistently on three public benchmarks, which only requires 3D volumes of 60 \u00d7 36 \u00d7 60 resolution for both input and output."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Learning Object Bounding Boxes for 3D Instance Segmentation on Point Clouds",
                    "abstract": "We propose a novel, conceptually simple and general framework for instance segmentation on 3D point clouds. Our method, called 3D-BoNet, follows the simple design philosophy of per-point multilayer perceptrons (MLPs). The framework directly regresses 3D bounding boxes for all instances in a point cloud, while simultaneously predicting a point-level mask for each instance. It consists of a backbone network followed by two parallel network branches for 1) bounding box regression and 2) point mask prediction. 3D-BoNet is single-stage, anchor-free and end-to-end trainable. Moreover, it is remarkably computationally efficient as, unlike existing approaches, it does not require any post-processing steps such as non-maximum suppression, feature sampling, clustering or voting. Extensive experiments show that our approach surpasses existing work on both ScanNet and S3DIS datasets while being approximately 10x more computationally efficient. Comprehensive ablation studies demonstrate the effectiveness of our design."
                },
                {
                    "title": "Learning the Depths of Moving People by Watching Frozen People",
                    "abstract": "We present a method for predicting dense depth in scenarios where both a monocular camera and people in the scene are freely moving. Existing methods for recovering depth for dynamic, non-rigid objects from monocular video impose strong assumptions on the objects\u2019 motion and may only recover sparse depth. In this paper, we take a data-driven approach and learn human depth priors from a new source of data: thousands of Internet videos of people imitating mannequins, i.e., freezing in diverse, natural poses, while a hand-held camera tours the scene. Since the people are stationary, training data can be created from these videos using multi-view stereo reconstruction. At inference time, our method uses motion parallax cues from the static areas of the scenes, and shows clear improvement over state-of-the-art monocular depth prediction methods. We demonstrate our method on real-world sequences of complex human actions captured by a moving hand-held camera, and show various 3D effects produced using our predicted depth."
                },
                {
                    "title": "On the Continuity of Rotation Representations in Neural Networks",
                    "abstract": "In neural networks, it is often desirable to work with various representations of the same space. For example, 3D rotations can be represented with quaternions or Euler angles. In this paper, we advance a definition of a continuous representation, which can be helpful for training deep neural networks. We relate this to topological concepts such as homeomorphism and embedding. We then investigate what are continuous and discontinuous representations for 2D, 3D, and n-dimensional rotations. We demonstrate that for 3D rotations, all representations are discontinuous in the real Euclidean spaces of four or fewer dimensions. Thus, widely used representations such as quaternions and Euler angles are discontinuous and difficult for neural networks to learn. We show that the 3D rotations have continuous representations in 5D and 6D, which are more suitable for learning. We also present continuous representations for the general case of the n-dimensional rotation group SO(n). While our main focus is on rotations, we also show that our constructions apply to other groups such as the orthogonal group and similarity transforms. We finally present empirical results, which show that our continuous rotation representations outperform discontinuous ones for several practical problems in graphics and vision, including a simple autoencoder sanity test, a rotation estimator for 3D point clouds, and an inverse kinematics solver for 3D human poses."
                },
                {
                    "title": "Recurrent Slice Networks for 3D Segmentation of Point Clouds",
                    "abstract": "Point clouds are an efficient data format for 3D data. However, existing 3D segmentation methods for point clouds either do not model local dependencies [21] or require added computations [14, 23]. This work presents a novel 3D segmentation framework, RSNet1, to efficiently model local structures in point clouds. The key component of the RSNet is a lightweight local dependency module. It is a combination of a novel slice pooling layer, Recurrent Neural Network (RNN) layers, and a slice unpooling layer. The slice pooling layer is designed to project features of unordered points onto an ordered sequence of feature vectors so that traditional end-to-end learning algorithms (RNNs) can be applied. The performance of RSNet is validated by comprehensive experiments on the S3DIS[1], ScanNet[3], and ShapeNet [34] datasets. In its simplest form, RSNets surpass all previous state-of-the-art methods on these benchmarks. And comparisons against previous state-of-the-art methods [21, 23] demonstrate the efficiency of RSNets."
                },
                {
                    "title": "Neural Discrete Representation Learning",
                    "abstract": "Learning useful representations without supervision remains a key challenge in machine learning. In this paper, we propose a simple yet powerful generative model that learns such discrete representations. Our model, the Vector Quantised-Variational AutoEncoder (VQ-VAE), differs from VAEs in two key ways: the encoder network outputs discrete, rather than continuous, codes; and the prior is learnt rather than static. In order to learn a discrete latent representation, we incorporate ideas from vector quantisation (VQ). Using the VQ method allows the model to circumvent issues of \"posterior collapse\" -- where the latents are ignored when they are paired with a powerful autoregressive decoder -- typically observed in the VAE framework. Pairing these representations with an autoregressive prior, the model can generate high quality images, videos, and speech as well as doing high quality speaker conversion and unsupervised learning of phonemes, providing further evidence of the utility of the learnt representations."
                },
                {
                    "title": "Structure-from-Motion Revisited",
                    "abstract": "Incremental Structure-from-Motion is a prevalent strategy for 3D reconstruction from unordered image collections. While incremental reconstruction systems have tremendously advanced in all regards, robustness, accuracy, completeness, and scalability remain the key problems towards building a truly general-purpose pipeline. We propose a new SfM technique that improves upon the state of the art to make a further step towards this ultimate goal. The full reconstruction pipeline is released to the public as an open-source implementation."
                },
                {
                    "title": "3D-LLM: Injecting the 3D World into Large Language Models",
                    "abstract": "Large language models (LLMs) and Vision-Language Models (VLMs) have been proven to excel at multiple tasks, such as commonsense reasoning. Powerful as these models can be, they are not grounded in the 3D physical world, which involves richer concepts such as spatial relationships, affordances, physics, layout, and so on. In this work, we propose to inject the 3D world into large language models and introduce a whole new family of 3D-LLMs. Specifically, 3D-LLMs can take 3D point clouds and their features as input and perform a diverse set of 3D-related tasks, including captioning, dense captioning, 3D question answering, task decomposition, 3D grounding, 3D-assisted dialog, navigation, and so on. Using three types of prompting mechanisms that we design, we are able to collect over 1M 3D-language data covering these tasks. To efficiently train 3D-LLMs, we first utilize a 3D feature extractor that obtains 3D features from rendered multi-view images. Then, we use 2D VLMs as our backbones to train our 3D-LLMs. By introducing a 3D localization mechanism, 3D-LLMs can better capture 3D spatial information. Experiments on held-out evaluation dataset, ScanQA, SQA3D and 3DMV-VQA, outperform state-of-the-art baselines. In particular, experiments on ScanQA show that our model outperforms state-of-the-art baselines by a large margin ( e.g. , the BLEU-1 score surpasses state-of-the-art score by 9%). Furthermore, experiments on our held-in datasets for 3D captioning, task composition, and 3D-assisted dialogue show that our model outperforms 2D VLMs. Qualitative examples also show that our model could perform more tasks beyond the scope of existing LLMs and VLMs. Project Page: : https://vis-www.cs.umass.edu/3dllm/ ."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Cinematography Theory And Practice Image Making For Cinematographers And Directors",
                    "abstract": "Thank you for reading cinematography theory and practice image making for cinematographers and directors. Maybe you have knowledge that, people have look hundreds times for their chosen books like this cinematography theory and practice image making for cinematographers and directors, but end up in infectious downloads. Rather than enjoying a good book with a cup of tea in the afternoon, instead they juggled with some malicious virus inside their laptop."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we enable users to control camera operations in video production through natural language interactions?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem has significant implications for the research community as it bridges the gap between natural language processing and computer vision, particularly in the context of video production. By empowering creators with conversational AI tools, we can democratize access to sophisticated cinematography techniques, allowing individuals without extensive technical knowledge to produce high-quality visual narratives. This advancement could lead to innovative applications in various domains, such as documentary filmmaking, live event broadcasting, and virtual reality experiences, ultimately transforming the way content is created and consumed.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of integrating multiple elements, including language understanding, 3D scene interpretation, and camera trajectory generation. Naive approaches may fail due to the intricate nature of translating natural language queries into precise camera movements, which require a deep understanding of spatial relationships and scene dynamics. Additionally, ensuring that the generated camera trajectories result in visually pleasing rendered videos adds another layer of complexity, necessitating advanced algorithms and models that can handle these multifaceted requirements.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has focused on controlling object and camera trajectories in video generation but has not specifically addressed the challenge of generating camera trajectories from natural language inputs. Limitations in existing solutions include a lack of effective integration between language models and 3D scene understanding, as well as insufficient datasets for training models on text-conditioned trajectory generation. Our approach differs by introducing CineGPT and the Anchor Determinator, which specifically target the nuances of camera operation and trajectory placement, thereby filling the gaps left by prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of ChatCam, which utilizes a GPT-based autoregressive model (CineGPT) for text-conditioned camera trajectory generation, alongside an Anchor Determinator for identifying relevant objects in the 3D scene. We will train CineGPT using a paired text-trajectory dataset and evaluate its performance using metrics that assess the accuracy and visual quality of the generated trajectories. The expected outcomes include a system that effectively interprets complex natural language instructions and produces aesthetically pleasing camera movements, demonstrating the potential for AI-human collaboration in video production."
            }
        },
        "author_data": {
            "de30cb46-db47-4116-a3c1-adf2442fd9b4": {
                "pk": "de30cb46-db47-4116-a3c1-adf2442fd9b4",
                "name": "Xinhang Liu",
                "collaborators": [
                    "Yu-Wing Tai",
                    "Chi-Keung Tang",
                    "Jiaben Chen",
                    "Huai Yu",
                    "Yan Zeng",
                    "Yifan Qin",
                    "Hao Li",
                    "Jiakai Zhang",
                    "Lan Xu",
                    "Jingyi Yu"
                ],
                "domain": [
                    "Neural Radiance Fields",
                    "3D Reconstruction",
                    "Computer Vision",
                    "Unsupervised Learning"
                ],
                "publications": [
                    {
                        "title": "Clean-NeRF: Reformulating NeRF to account for View-Dependent Observations",
                        "abstract": "While Neural Radiance Fields (NeRFs) had achieved unprecedented novel view synthesis results, they have been struggling in dealing with large-scale cluttered scenes with sparse input views and highly view-dependent appearances. Specifically, existing NeRF-based models tend to produce blurry rendering with the volumetric reconstruction often inaccurate, where a lot of reconstruction errors are observed in the form of foggy \"floaters\" hovering within the entire volume of an opaque 3D scene. Such inaccuracies impede NeRF's potential for accurate 3D NeRF registration, object detection, segmentation, etc., which possibly accounts for only limited significant research effort so far to directly address these important 3D fundamental computer vision problems to date. This paper analyzes the NeRF's struggles in such settings and proposes Clean-NeRF for accurate 3D reconstruction and novel view rendering in complex scenes. Our key insights consist of enforcing effective appearance and geometry constraints, which are absent in the conventional NeRF reconstruction, by 1) automatically detecting and modeling view-dependent appearances in the training views to prevent them from interfering with density estimation, which is complete with 2) a geometric correction procedure performed on each traced ray during inference. Clean-NeRF can be implemented as a plug-in that can immediately benefit existing NeRF-based methods without additional input. Codes will be released."
                    },
                    {
                        "title": "Unsupervised Multi-View Object Segmentation Using Radiance Field Propagation",
                        "abstract": "We present radiance field propagation (RFP), a novel approach to segmenting objects in 3D during reconstruction given only unlabeled multi-view images of a scene. RFP is derived from emerging neural radiance field-based techniques, which jointly encodes semantics with appearance and geometry. The core of our method is a novel propagation strategy for individual objects' radiance fields with a bidirectional photometric loss, enabling an unsupervised partitioning of a scene into salient or meaningful regions corresponding to different object instances. To better handle complex scenes with multiple objects and occlusions, we further propose an iterative expectation-maximization algorithm to refine object masks. RFP is one of the first unsupervised approach for tackling 3D real scene object segmentation for neural radiance field (NeRF) without any supervision, annotations, or other cues such as 3D bounding boxes and prior knowledge of object class. Experiments demonstrate that RFP achieves feasible segmentation results that are more accurate than previous unsupervised image/scene segmentation approaches, and are comparable to existing supervised NeRF-based methods. The segmented object representations enable individual 3D object editing operations."
                    },
                    {
                        "title": "CryoFormer: Continuous Heterogeneous Cryo-EM Reconstruction using Transformer-based Neural Representations",
                        "abstract": "Cryo-electron microscopy (cryo-EM) allows for the high-resolution reconstruction of 3D structures of proteins and other biomolecules. Successful reconstruction of both shape and movement greatly helps understand the fundamental processes of life. However, it is still challenging to reconstruct the continuous motions of 3D structures from hundreds of thousands of noisy and randomly oriented 2D cryo-EM images. Recent advancements use Fourier domain coordinate-based neural networks to continuously model 3D conformations, yet they often struggle to capture local flexible regions accurately. We propose CryoFormer, a new approach for continuous heterogeneous cryo-EM reconstruction. Our approach leverages an implicit feature volume directly in the real domain as the 3D representation. We further introduce a novel query-based deformation transformer decoder to improve the reconstruction quality. Our approach is capable of refining pre-computed pose estimations and locating flexible regions. In experiments, our method outperforms current approaches on three public datasets (1 synthetic and 2 experimental) and a new synthetic dataset of PEDV spike protein. The code and new synthetic dataset will be released for better reproducibility of our results. Project page: https://cryoformer.github.io."
                    },
                    {
                        "title": "InceptionHuman: Controllable Prompt-to-NeRF for Photorealistic 3D Human Generation",
                        "abstract": "This paper presents InceptionHuman, a prompt-to-NeRF framework that allows easy control via a combination of prompts in different modalities (e.g., text, poses, edge, segmentation map, etc) as inputs to generate photorealistic 3D humans. While many works have focused on generating 3D human models, they suffer one or more of the following: lack of distinctive features, unnatural shading/shadows, unnatural poses/clothes, limited views, etc. InceptionHuman achieves consistent 3D human generation within a progressively refined NeRF space with two novel modules, Iterative Pose-Aware Refinement (IPAR) and Progressive-Augmented Reconstruction (PAR). IPAR iteratively refines the diffusion-generated images and synthesizes high-quality 3D-aware views considering the close-pose RGB values. PAR employs a pretrained diffusion prior to augment the generated synthetic views and adds regularization for view-independent appearance. Overall, the synthesis of photorealistic novel views empowers the resulting 3D human NeRF from 360-degree perspectives. Extensive qualitative and quantitative experimental comparison show that our InceptionHuman models achieve state-of-the-art application quality."
                    }
                ]
            },
            "697473a4-c1b2-4423-9536-496070c19dba": {
                "pk": "697473a4-c1b2-4423-9536-496070c19dba",
                "name": "Yu-Wing Tai",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "2f765bdc-cb58-4802-bab9-236655934fc8": {
                "pk": "2f765bdc-cb58-4802-bab9-236655934fc8",
                "name": "Chi-Keung Tang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2401.17548": {
        "paper_data": {
            "title": "Rethinking Channel Dependence for Multivariate Time Series Forecasting: Learning from Leading Indicators",
            "url": "http://arxiv.org/abs/2401.17548v6",
            "arxiv_id": "2401.17548",
            "authors": [
                "Lifan Zhao",
                "Yanyan Shen"
            ],
            "abstract": "Recently, channel-independent methods have achieved state-of-the-art performance in multivariate time series (MTS) forecasting. Despite reducing overfitting risks, these methods miss potential opportunities in utilizing channel dependence for accurate predictions. We argue that there exist locally stationary lead-lag relationships between variates, i.e., some lagged variates may follow the leading indicators within a short time period. Exploiting such channel dependence is beneficial since leading indicators offer advance information that can be used to reduce the forecasting difficulty of the lagged variates. In this paper, we propose a new method named LIFT that first efficiently estimates leading indicators and their leading steps at each time step and then judiciously allows the lagged variates to utilize the advance information from leading indicators. LIFT plays as a plugin that can be seamlessly collaborated with arbitrary time series forecasting methods. Extensive experiments on six real-world datasets demonstrate that LIFT improves the state-of-the-art methods by 5.5% in average forecasting performance. Our code is available at https://github.com/SJTU-Quant/LIFT.",
            "introduction": "  Appendix A Mathematical Proofs   A.1 Discrete-time Fourier Transform  Discrete Fourier Transform (DFT) provides a frequency-domain view of discrete time series. Given a particular frequency f\u2208{0,1L,\u22ef,\u230aL/2\u230bL}\ud835\udc5301\ud835\udc3f\u22ef\ud835\udc3f2\ud835\udc3ff\\in\\{0,\\frac{1}{L},\\cdots,\\frac{\\lfloor L/2\\rfloor}{L}\\}italic_f \u2208 { 0 , divide start_ARG 1 end_ARG start_ARG italic_L end_ARG , \u22ef , divide start_ARG \u230a italic_L / 2 \u230b end_ARG start_ARG italic_L end_ARG }, the corresponding frequency component Ufsubscript\ud835\udc48\ud835\udc53U_{f}italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT of univariate time series ut\u2212L+1:tsubscript\ud835\udc62:\ud835\udc61\ud835\udc3f1\ud835\udc61u_{t-L+1:t}italic_u start_POSTSUBSCRIPT italic_t - italic_L + 1 : italic_t end_POSTSUBSCRIPT is derived by    Uf=\u2131\u2062(ut\u2212L+1:t)f=\u2211\u2113=0L\u22121ut\u2212L+1+\u2113\u22c5e\u2212i\u20622\u2062\u03c0\u2062f\u2062\u2113,subscript\ud835\udc48\ud835\udc53\u2131subscriptsubscript\ud835\udc62:\ud835\udc61\ud835\udc3f1\ud835\udc61\ud835\udc53superscriptsubscript\u21130\ud835\udc3f1\u22c5subscript\ud835\udc62\ud835\udc61\ud835\udc3f1\u2113superscript\ud835\udc52\ud835\udc562\ud835\udf0b\ud835\udc53\u2113U_{f}=\\mathcal{F}(u_{t-L+1:t})_{f}=\\sum_{\\ell=0}^{L-1}u_{t-L+1+\\ell}\\cdot e^{-% i2\\pi f\\ell},italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT = caligraphic_F ( italic_u start_POSTSUBSCRIPT italic_t - italic_L + 1 : italic_t end_POSTSUBSCRIPT ) start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT = \u2211 start_POSTSUBSCRIPT roman_\u2113 = 0 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_L - 1 end_POSTSUPERSCRIPT italic_u start_POSTSUBSCRIPT italic_t - italic_L + 1 + roman_\u2113 end_POSTSUBSCRIPT \u22c5 italic_e start_POSTSUPERSCRIPT - italic_i 2 italic_\u03c0 italic_f roman_\u2113 end_POSTSUPERSCRIPT ,  (1)   where i\ud835\udc56iitalic_i is the imaginary unit and Uf\u2208\u2102subscript\ud835\udc48\ud835\udc53\u2102U_{f}\\in\\mathbb{C}italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT \u2208 blackboard_C is a complex number. The inverse DFT is defined by    \u2131\u22121\u2062(U)\u2113=\u2211fUf\u22c5ei\u20622\u2062\u03c0\u2062f\u2062\u2113.superscript\u21311subscript\ud835\udc48\u2113subscript\ud835\udc53\u22c5subscript\ud835\udc48\ud835\udc53superscript\ud835\udc52\ud835\udc562\ud835\udf0b\ud835\udc53\u2113\\mathcal{F}^{-1}\\left(U\\right)_{\\ell}=\\sum_{f}U_{f}\\cdot e^{i2\\pi f\\ell}.caligraphic_F start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ( italic_U ) start_POSTSUBSCRIPT roman_\u2113 end_POSTSUBSCRIPT = \u2211 start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT \u22c5 italic_e start_POSTSUPERSCRIPT italic_i 2 italic_\u03c0 italic_f roman_\u2113 end_POSTSUPERSCRIPT .  (2)     We can calculate the amplitude |Uf|subscript\ud835\udc48\ud835\udc53|U_{f}|| italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT | and the phase \u03d5\u2062(Uf)italic-\u03d5subscript\ud835\udc48\ud835\udc53\\phi(U_{f})italic_\u03d5 ( italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT ) of the corresponding cosine signal by:    |Uf|=\u211c\u2062{Uf}2+\u2111\u2062{Uf}2,subscript\ud835\udc48\ud835\udc53\u211csuperscriptsubscript\ud835\udc48\ud835\udc532\u2111superscriptsubscript\ud835\udc48\ud835\udc532\\displaystyle\\left|U_{f}\\right|=\\sqrt{\\mathfrak{R}\\left\\{U_{f}\\right\\}^{2}+% \\mathfrak{I}\\left\\{U_{f}\\right\\}^{2}},| italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT | = square-root start_ARG fraktur_R { italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT } start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT + fraktur_I { italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT } start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT end_ARG ,  (3)    \u03d5\u2062(Uf)=tan\u22121\u2061(\u2111\u2062{Uf}\u211c\u2062{Uf}),italic-\u03d5subscript\ud835\udc48\ud835\udc53superscript1\u2111subscript\ud835\udc48\ud835\udc53\u211csubscript\ud835\udc48\ud835\udc53\\displaystyle\\phi\\left(U_{f}\\right)=\\tan^{-1}\\left(\\frac{\\mathfrak{I}\\left\\{U_% {f}\\right\\}}{\\mathfrak{R}\\left\\{U_{f}\\right\\}}\\right),italic_\u03d5 ( italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT ) = roman_tan start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ( divide start_ARG fraktur_I { italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT } end_ARG start_ARG fraktur_R { italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT } end_ARG ) ,  (4)   where \u211c\u2062{Uf}\u211csubscript\ud835\udc48\ud835\udc53\\mathfrak{R}\\{U_{f}\\}fraktur_R { italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT } and \u2111\u2062{Uf}\u2111subscript\ud835\udc48\ud835\udc53\\mathfrak{I}\\{U_{f}\\}fraktur_I { italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT } denotes its real and imaginary components of Ufsubscript\ud835\udc48\ud835\udc53U_{f}italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT. With \u211c\u2062{Uf}\u211csubscript\ud835\udc48\ud835\udc53\\mathfrak{R}\\{U_{f}\\}fraktur_R { italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT } and \u2111\u2062{Uf}\u2111subscript\ud835\udc48\ud835\udc53\\mathfrak{I}\\{U_{f}\\}fraktur_I { italic_U start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT } scaled in the same rate, our proposed real-valued filters only scale the amplitude but keep the phase unchanged.     A.2 Efficient Cross-correlation Estimation  Given another time series v\ud835\udc63vitalic_v that lags behind u\ud835\udc62uitalic_u by \u03b4\ud835\udeff\\deltaitalic_\u03b4 steps, we denote its frequency components as V\ud835\udc49Vitalic_V and define the cross-correlation between their lookback windows by     R\u2062(\u03b4)\ud835\udc45\ud835\udeff\\displaystyle R(\\delta)italic_R ( italic_\u03b4 ) \u225c1L\u2062\u2211\u2113=0Lut\u2212L+1+\u2113\u2212\u03b4\u22c5vt\u2212L+1+\u2113\u225cabsent1\ud835\udc3fsuperscriptsubscript\u21130\ud835\udc3f\u22c5subscript\ud835\udc62\ud835\udc61\ud835\udc3f1\u2113\ud835\udeffsubscript\ud835\udc63\ud835\udc61\ud835\udc3f1\u2113\\displaystyle\\triangleq\\frac{1}{L}\\sum_{\\ell=0}^{L}u_{t-L+1+\\ell-\\delta}\\cdot v% _{t-L+1+\\ell}\u225c divide start_ARG 1 end_ARG start_ARG italic_L end_ARG \u2211 start_POSTSUBSCRIPT roman_\u2113 = 0 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_L end_POSTSUPERSCRIPT italic_u start_POSTSUBSCRIPT italic_t - italic_L + 1 + roman_\u2113 - italic_\u03b4 end_POSTSUBSCRIPT \u22c5 italic_v start_POSTSUBSCRIPT italic_t - italic_L + 1 + roman_\u2113 end_POSTSUBSCRIPT  (5)     =1L\u2062\u2211\u2113=0Lu\u2062(\u2113\u2212\u03b4)\u2062v\u2062(\u2113).absent1\ud835\udc3fsuperscriptsubscript\u21130\ud835\udc3f\ud835\udc62\u2113\ud835\udeff\ud835\udc63\u2113\\displaystyle=\\frac{1}{L}\\sum_{\\ell=0}^{L}u(\\ell-\\delta)v(\\ell).= divide start_ARG 1 end_ARG start_ARG italic_L end_ARG \u2211 start_POSTSUBSCRIPT roman_\u2113 = 0 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_L end_POSTSUPERSCRIPT italic_u ( roman_\u2113 - italic_\u03b4 ) italic_v ( roman_\u2113 ) .     We further denote u\u2062(\u2113\u2212\u03b4)\ud835\udc62\u2113\ud835\udeffu(\\ell-\\delta)italic_u ( roman_\u2113 - italic_\u03b4 ) as u\u02c7\u2062(\u03b4\u2212\u2113)\u02c7\ud835\udc62\ud835\udeff\u2113\\check{u}(\\delta-\\ell)overroman_\u02c7 start_ARG italic_u end_ARG ( italic_\u03b4 - roman_\u2113 ). Due to the conjugate symmetry of the DFT on real-valued signals, we have    \u2131\u2062(u\u02c7\u2062(\u03b4\u2212\u2113))f=\u2131\u2062(u\u2062(\u03b4\u2212\u2113))\u00aff\u2248U\u00aff,\u2131subscript\u02c7\ud835\udc62\ud835\udeff\u2113\ud835\udc53subscript\u00af\u2131\ud835\udc62\ud835\udeff\u2113\ud835\udc53subscript\u00af\ud835\udc48\ud835\udc53\\mathcal{F}(\\check{u}(\\delta-\\ell))_{f}=\\overline{\\mathcal{F}(u(\\delta-\\ell))}% _{f}\\approx\\overline{U}_{f},caligraphic_F ( overroman_\u02c7 start_ARG italic_u end_ARG ( italic_\u03b4 - roman_\u2113 ) ) start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT = over\u00af start_ARG caligraphic_F ( italic_u ( italic_\u03b4 - roman_\u2113 ) ) end_ARG start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT \u2248 over\u00af start_ARG italic_U end_ARG start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT ,  (6)   where the bar is the conjugate operation and we assume",
            "references": [
                {
                    "title": "TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting",
                    "abstract": "Time series forecasting is widely used in extensive applications, such as traffic planning and weather forecasting. However, real-world time series usually present intricate temporal variations, making forecasting extremely challenging. Going beyond the mainstream paradigms of plain decomposition and multiperiodicity analysis, we analyze temporal variations in a novel view of multiscale-mixing, which is based on an intuitive but important observation that time series present distinct patterns in different sampling scales. The microscopic and the macroscopic information are reflected in fine and coarse scales respectively, and thereby complex variations can be inherently disentangled. Based on this observation, we propose TimeMixer as a fully MLP-based architecture with Past-Decomposable-Mixing (PDM) and Future-Multipredictor-Mixing (FMM) blocks to take full advantage of disentangled multiscale series in both past extraction and future prediction phases. Concretely, PDM applies the decomposition to multiscale series and further mixes the decomposed seasonal and trend components in fine-to-coarse and coarse-to-fine directions separately, which successively aggregates the microscopic seasonal and macroscopic trend information. FMM further ensembles multiple predictors to utilize complementary forecasting capabilities in multiscale observations. Consequently, TimeMixer is able to achieve consistent state-of-the-art performances in both long-term and short-term forecasting tasks with favorable run-time efficiency."
                },
                {
                    "title": "MG-TSD: Multi-Granularity Time Series Diffusion Models with Guided Learning Process",
                    "abstract": "Recently, diffusion probabilistic models have attracted attention in generative time series forecasting due to their remarkable capacity to generate high-fidelity samples. However, the effective utilization of their strong modeling ability in the probabilistic time series forecasting task remains an open question, partially due to the challenge of instability arising from their stochastic nature. To address this challenge, we introduce a novel Multi-Granularity Time Series Diffusion (MG-TSD) model, which achieves state-of-the-art predictive performance by leveraging the inherent granularity levels within the data as given targets at intermediate diffusion steps to guide the learning process of diffusion models. The way to construct the targets is motivated by the observation that the forward process of the diffusion model, which sequentially corrupts the data distribution to a standard normal distribution, intuitively aligns with the process of smoothing fine-grained data into a coarse-grained representation, both of which result in a gradual loss of fine distribution features. In the study, we derive a novel multi-granularity guidance diffusion loss function and propose a concise implementation method to effectively utilize coarse-grained data across various granularity levels. More importantly, our approach does not rely on additional external data, making it versatile and applicable across various domains. Extensive experiments conducted on real-world datasets demonstrate that our MG-TSD model outperforms existing time series prediction methods."
                },
                {
                    "title": "Pathformer: Multi-scale Transformers with Adaptive Pathways for Time Series Forecasting",
                    "abstract": "Transformers for time series forecasting mainly model time series from limited or fixed scales, making it challenging to capture different characteristics spanning various scales. We propose Pathformer, a multi-scale Transformer with adaptive pathways. It integrates both temporal resolution and temporal distance for multi-scale modeling. Multi-scale division divides the time series into different temporal resolutions using patches of various sizes. Based on the division of each scale, dual attention is performed over these patches to capture global correlations and local details as temporal dependencies. We further enrich the multi-scale Transformer with adaptive pathways, which adaptively adjust the multi-scale modeling process based on the varying temporal dynamics of the input, improving the accuracy and generalization of Pathformer. Extensive experiments on eleven real-world datasets demonstrate that Pathformer not only achieves state-of-the-art performance by surpassing all current models but also exhibits stronger generalization abilities under various transfer scenarios. The code is made available at https://github.com/decisionintelligence/pathformer."
                },
                {
                    "title": "STanHop: Sparse Tandem Hopfield Model for Memory-Enhanced Time Series Prediction",
                    "abstract": "We present STanHop-Net (Sparse Tandem Hopfield Network) for multivariate time series prediction with memory-enhanced capabilities. At the heart of our approach is STanHop, a novel Hopfield-based neural network block, which sparsely learns and stores both temporal and cross-series representations in a data-dependent fashion. In essence, STanHop sequentially learn temporal representation and cross-series representation using two tandem sparse Hopfield layers. In addition, StanHop incorporates two additional external memory modules: a Plug-and-Play module and a Tune-and-Play module for train-less and task-aware memory-enhancements, respectively. They allow StanHop-Net to swiftly respond to certain sudden events. Methodologically, we construct the StanHop-Net by stacking STanHop blocks in a hierarchical fashion, enabling multi-resolution feature extraction with resolution-specific sparsity. Theoretically, we introduce a sparse extension of the modern Hopfield model (Generalized Sparse Modern Hopfield Model) and show that it endows a tighter memory retrieval error compared to the dense counterpart without sacrificing memory capacity. Empirically, we validate the efficacy of our framework on both synthetic and real-world settings."
                },
                {
                    "title": "Learning to Embed Time Series Patches Independently",
                    "abstract": "Masked time series modeling has recently gained much attention as a self-supervised representation learning strategy for time series. Inspired by masked image modeling in computer vision, recent works first patchify and partially mask out time series, and then train Transformers to capture the dependencies between patches by predicting masked patches from unmasked patches. However, we argue that capturing such patch dependencies might not be an optimal strategy for time series representation learning; rather, learning to embed patches independently results in better time series representations. Specifically, we propose to use 1) the simple patch reconstruction task, which autoencode each patch without looking at other patches, and 2) the simple patch-wise MLP that embeds each patch independently. In addition, we introduce complementary contrastive learning to hierarchically capture adjacent time series information efficiently. Our proposed method improves time series forecasting and classification performance compared to state-of-the-art Transformer-based models, while it is more efficient in terms of the number of parameters and training/inference time. Code is available at this repository: https://github.com/seunghan96/pits."
                },
                {
                    "title": "FourierGNN: Rethinking Multivariate Time Series Forecasting from a Pure Graph Perspective",
                    "abstract": "Multivariate time series (MTS) forecasting has shown great importance in numerous industries. Current state-of-the-art graph neural network (GNN)-based forecasting methods usually require both graph networks (e.g., GCN) and temporal networks (e.g., LSTM) to capture inter-series (spatial) dynamics and intra-series (temporal) dependencies, respectively. However, the uncertain compatibility of the two networks puts an extra burden on handcrafted model designs. Moreover, the separate spatial and temporal modeling naturally violates the unified spatiotemporal inter-dependencies in real world, which largely hinders the forecasting performance. To overcome these problems, we explore an interesting direction of directly applying graph networks and rethink MTS forecasting from a pure graph perspective. We first define a novel data structure, hypervariate graph, which regards each series value (regardless of variates or timestamps) as a graph node, and represents sliding windows as space-time fully-connected graphs. This perspective considers spatiotemporal dynamics unitedly and reformulates classic MTS forecasting into the predictions on hypervariate graphs. Then, we propose a novel architecture Fourier Graph Neural Network (FourierGNN) by stacking our proposed Fourier Graph Operator (FGO) to perform matrix multiplications in Fourier space. FourierGNN accommodates adequate expressiveness and achieves much lower complexity, which can effectively and efficiently accomplish the forecasting. Besides, our theoretical analysis reveals FGO's equivalence to graph convolutions in the time domain, which further verifies the validity of FourierGNN. Extensive experiments on seven datasets have demonstrated our superior performance with higher efficiency and fewer parameters compared with state-of-the-art methods."
                },
                {
                    "title": "Frequency-domain MLPs are More Effective Learners in Time Series Forecasting",
                    "abstract": "Time series forecasting has played the key role in different industrial, including finance, traffic, energy, and healthcare domains. While existing literatures have designed many sophisticated architectures based on RNNs, GNNs, or Transformers, another kind of approaches based on multi-layer perceptrons (MLPs) are proposed with simple structure, low complexity, and {superior performance}. However, most MLP-based forecasting methods suffer from the point-wise mappings and information bottleneck, which largely hinders the forecasting performance. To overcome this problem, we explore a novel direction of applying MLPs in the frequency domain for time series forecasting. We investigate the learned patterns of frequency-domain MLPs and discover their two inherent characteristic benefiting forecasting, (i) global view: frequency spectrum makes MLPs own a complete view for signals and learn global dependencies more easily, and (ii) energy compaction: frequency-domain MLPs concentrate on smaller key part of frequency components with compact signal energy. Then, we propose FreTS, a simple yet effective architecture built upon Frequency-domain MLPs for Time Series forecasting. FreTS mainly involves two stages, (i) Domain Conversion, that transforms time-domain signals into complex numbers of frequency domain; (ii) Frequency Learning, that performs our redesigned MLPs for the learning of real and imaginary part of frequency components. The above stages operated on both inter-series and intra-series scales further contribute to channel-wise and time-wise dependency learning. Extensive experiments on 13 real-world benchmarks (including 7 benchmarks for short-term forecasting and 6 benchmarks for long-term forecasting) demonstrate our consistent superiority over state-of-the-art methods."
                },
                {
                    "title": "BasisFormer: Attention-based Time Series Forecasting with Learnable and Interpretable Basis",
                    "abstract": "Bases have become an integral part of modern deep learning-based models for time series forecasting due to their ability to act as feature extractors or future references. To be effective, a basis must be tailored to the specific set of time series data and exhibit distinct correlation with each time series within the set. However, current state-of-the-art methods are limited in their ability to satisfy both of these requirements simultaneously. To address this challenge, we propose BasisFormer, an end-to-end time series forecasting architecture that leverages learnable and interpretable bases. This architecture comprises three components: First, we acquire bases through adaptive self-supervised learning, which treats the historical and future sections of the time series as two distinct views and employs contrastive learning. Next, we design a Coef module that calculates the similarity coefficients between the time series and bases in the historical view via bidirectional cross-attention. Finally, we present a Forecast module that selects and consolidates the bases in the future view based on the similarity coefficients, resulting in accurate future predictions. Through extensive experiments on six datasets, we demonstrate that BasisFormer outperforms previous state-of-the-art methods by 11.04\\% and 15.78\\% respectively for univariate and multivariate forecasting tasks. Code is available at: \\url{https://github.com/nzl5116190/Basisformer}"
                },
                {
                    "title": "iTransformer: Inverted Transformers Are Effective for Time Series Forecasting",
                    "abstract": "The recent boom of linear forecasting models questions the ongoing passion for architectural modifications of Transformer-based forecasters. These forecasters leverage Transformers to model the global dependencies over temporal tokens of time series, with each token formed by multiple variates of the same timestamp. However, Transformers are challenged in forecasting series with larger lookback windows due to performance degradation and computation explosion. Besides, the embedding for each temporal token fuses multiple variates that represent potential delayed events and distinct physical measurements, which may fail in learning variate-centric representations and result in meaningless attention maps. In this work, we reflect on the competent duties of Transformer components and repurpose the Transformer architecture without any modification to the basic components. We propose iTransformer that simply applies the attention and feed-forward network on the inverted dimensions. Specifically, the time points of individual series are embedded into variate tokens which are utilized by the attention mechanism to capture multivariate correlations; meanwhile, the feed-forward network is applied for each variate token to learn nonlinear representations. The iTransformer model achieves state-of-the-art on challenging real-world datasets, which further empowers the Transformer family with promoted performance, generalization ability across different variates, and better utilization of arbitrary lookback windows, making it a nice alternative as the fundamental backbone of time series forecasting. Code is available at this repository: https://github.com/thuml/iTransformer."
                },
                {
                    "title": "TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting",
                    "abstract": "The past decade has witnessed significant advances in time series modeling with deep learning. While achieving state-of-the-art results, the best-performing architectures vary highly across applications and domains. Meanwhile, for natural language processing, the Generative Pre-trained Transformer (GPT) has demonstrated impressive performance via training one general-purpose model across various textual datasets. It is intriguing to explore whether GPT-type architectures can be effective for time series, capturing the intrinsic dynamic attributes and leading to significant accuracy improvements. In this paper, we propose a novel framework, TEMPO, that can effectively learn time series representations. We focus on utilizing two essential inductive biases of the time series task for pre-trained models: (i) decomposition of the complex interaction between trend, seasonal and residual components; and (ii) introducing the design of prompts to facilitate distribution adaptation in different types of time series. TEMPO expands the capability for dynamically modeling real-world temporal phenomena from data within diverse domains. Our experiments demonstrate the superior performance of TEMPO over state-of-the-art methods on zero shot setting for a number of time series benchmark datasets. This performance gain is observed not only in scenarios involving previously unseen datasets but also in scenarios with multi-modal inputs. This compelling finding highlights TEMPO's potential to constitute a foundational model-building framework."
                },
                {
                    "title": "Time-LLM: Time Series Forecasting by Reprogramming Large Language Models",
                    "abstract": "Time series forecasting holds significant importance in many real-world dynamic systems and has been extensively studied. Unlike natural language process (NLP) and computer vision (CV), where a single large model can tackle multiple tasks, models for time series forecasting are often specialized, necessitating distinct designs for different tasks and applications. While pre-trained foundation models have made impressive strides in NLP and CV, their development in time series domains has been constrained by data sparsity. Recent studies have revealed that large language models (LLMs) possess robust pattern recognition and reasoning abilities over complex sequences of tokens. However, the challenge remains in effectively aligning the modalities of time series data and natural language to leverage these capabilities. In this work, we present Time-LLM, a reprogramming framework to repurpose LLMs for general time series forecasting with the backbone language models kept intact. We begin by reprogramming the input time series with text prototypes before feeding it into the frozen LLM to align the two modalities. To augment the LLM's ability to reason with time series data, we propose Prompt-as-Prefix (PaP), which enriches the input context and directs the transformation of reprogrammed input patches. The transformed time series patches from the LLM are finally projected to obtain the forecasts. Our comprehensive evaluations demonstrate that Time-LLM is a powerful time series learner that outperforms state-of-the-art, specialized forecasting models. Moreover, Time-LLM excels in both few-shot and zero-shot learning scenarios."
                },
                {
                    "title": "FITS: Modeling Time Series with 10k Parameters",
                    "abstract": "In this paper, we introduce FITS, a lightweight yet powerful model for time series analysis. Unlike existing models that directly process raw time-domain data, FITS operates on the principle that time series can be manipulated through interpolation in the complex frequency domain. By discarding high-frequency components with negligible impact on time series data, FITS achieves performance comparable to state-of-the-art models for time series forecasting and anomaly detection tasks, while having a remarkably compact size of only approximately $10k$ parameters. Such a lightweight model can be easily trained and deployed in edge devices, creating opportunities for various applications. The code is available in: \\url{https://github.com/VEWOXIC/FITS}"
                },
                {
                    "title": "TSMixer: Lightweight MLP-Mixer Model for Multivariate Time Series Forecasting",
                    "abstract": "Transformers have gained popularity in time series forecasting for their ability to capture long-sequence interactions. However, their memory and compute-intensive requirements pose a critical bottleneck for long-term forecasting, despite numerous advancements in compute-aware self-attention modules. To address this, we propose TSMixer, a lightweight neural architecture exclusively composed of multi-layer perceptron (MLP) modules. TSMixer is designed for multivariate forecasting and representation learning on patched time series, providing an efficient alternative to Transformers. Our model draws inspiration from the success of MLP-Mixer models in computer vision. We demonstrate the challenges involved in adapting Vision MLP-Mixer for time series and introduce empirically validated components to enhance accuracy. This includes a novel design paradigm of attaching online reconciliation heads to the MLP-Mixer backbone, for explicitly modeling the time-series properties such as hierarchy and channel-correlations. We also propose a Hybrid channel modeling approach to effectively handle noisy channel interactions and generalization across diverse datasets, a common challenge in existing patch channel-mixing methods. Additionally, a simple gated attention mechanism is introduced in the backbone to prioritize important features. By incorporating these lightweight components, we significantly enhance the learning capability of simple MLP structures, outperforming complex Transformer models with minimal computing usage. Moreover, TSMixer's modular design enables compatibility with both supervised and masked self-supervised learning methods, making it a promising building block for time-series Foundation Models. TSMixer outperforms state-of-the-art MLP and Transformer models in forecasting by a considerable margin of 8-60%. It also outperforms the latest strong benchmarks of Patch-Transformer models (by 1-2%) with a significant reduction in memory and runtime (2-3X)."
                },
                {
                    "title": "CARD: Channel Aligned Robust Blend Transformer for Time Series Forecasting",
                    "abstract": "Recent studies have demonstrated the great power of Transformer models for time series forecasting. One of the key elements that lead to the transformer's success is the channel-independent (CI) strategy to improve the training robustness. However, the ignorance of the correlation among different channels in CI would limit the model's forecasting capacity. In this work, we design a special Transformer, i.e., Channel Aligned Robust Blend Transformer (CARD for short), that addresses key shortcomings of CI type Transformer in time series forecasting. First, CARD introduces a channel-aligned attention structure that allows it to capture both temporal correlations among signals and dynamical dependence among multiple variables over time. Second, in order to efficiently utilize the multi-scale knowledge, we design a token blend module to generate tokens with different resolutions. Third, we introduce a robust loss function for time series forecasting to alleviate the potential overfitting issue. This new loss function weights the importance of forecasting over a finite horizon based on prediction uncertainties. Our evaluation of multiple long-term and short-term forecasting datasets demonstrates that CARD significantly outperforms state-of-the-art time series forecasting methods. The code is available at the following repository:https://github.com/wxie9/CARD"
                },
                {
                    "title": "Revisiting Long-term Time Series Forecasting: An Investigation on Linear Mapping",
                    "abstract": "Long-term time series forecasting has gained significant attention in recent years. While there are various specialized designs for capturing temporal dependency, previous studies have demonstrated that a single linear layer can achieve competitive forecasting performance compared to other complex architectures. In this paper, we thoroughly investigate the intrinsic effectiveness of recent approaches and make three key observations: 1) linear mapping is critical to prior long-term time series forecasting efforts; 2) RevIN (reversible normalization) and CI (Channel Independent) play a vital role in improving overall forecasting performance; and 3) linear mapping can effectively capture periodic features in time series and has robustness for different periods across channels when increasing the input horizon. We provide theoretical and experimental explanations to support our findings and also discuss the limitations and future works. Our framework's code is available at \\url{https://github.com/plumprc/RTSF}."
                },
                {
                    "title": "The Capacity and Robustness Trade-Off: Revisiting the Channel Independent Strategy for Multivariate Time Series Forecasting",
                    "abstract": "Multivariate time series data comprises various channels of variables. The multivariate forecasting models need to capture the relationship between the channels to accurately predict future values. However, recently, there has been an emergence of methods that employ the Channel Independent (CI) strategy. These methods view multivariate time series data as separate univariate time series and disregard the correlation between channels. Surprisingly, our empirical results have shown that models trained with the CI strategy outperform those trained with the Channel Dependent (CD) strategy, usually by a significant margin. Nevertheless, the reasons behind this phenomenon have not yet been thoroughly explored in the literature. This paper provides comprehensive empirical and theoretical analyses of the characteristics of multivariate time series datasets and the CI/CD strategy. Our results conclude that the CD approach has higher capacity but often lacks robustness to accurately predict distributionally drifted time series. In contrast, the CI approach trades capacity for robust prediction. Practical measures inspired by these analyses are proposed to address the capacity and robustness dilemma, including a modified CD method called Predict Residuals with Regularization (PRReg) that can surpass the CI strategy. We hope our findings can raise awareness among researchers about the characteristics of multivariate time series and inspire the construction of better forecasting models."
                },
                {
                    "title": "TSMixer: An all-MLP Architecture for Time Series Forecasting",
                    "abstract": "Real-world time-series datasets are often multivariate with complex dynamics. To capture this complexity, high capacity architectures like recurrent- or attention-based sequential deep learning models have become popular. However, recent work demonstrates that simple univariate linear models can outperform such deep learning models on several commonly used academic benchmarks. Extending them, in this paper, we investigate the capabilities of linear models for time-series forecasting and present Time-Series Mixer (TSMixer), a novel architecture designed by stacking multi-layer perceptrons (MLPs). TSMixer is based on mixing operations along both the time and feature dimensions to extract information efficiently. On popular academic benchmarks, the simple-to-implement TSMixer is comparable to specialized state-of-the-art models that leverage the inductive biases of specific benchmarks. On the challenging and large scale M5 benchmark, a real-world retail dataset, TSMixer demonstrates superior performance compared to the state-of-the-art alternatives. Our results underline the importance of efficiently utilizing cross-variate and auxiliary information for improving the performance of time series forecasting. We present various analyses to shed light into the capabilities of TSMixer. The design paradigms utilized in TSMixer are expected to open new horizons for deep learning-based time series forecasting. The implementation is available at https://github.com/google-research/google-research/tree/master/tsmixer"
                },
                {
                    "title": "One Fits All: Power General Time Series Analysis by Pretrained LM",
                    "abstract": "Although we have witnessed great success of pre-trained models in natural language processing (NLP) and computer vision (CV), limited progress has been made for general time series analysis. Unlike NLP and CV where a unified model can be used to perform different tasks, specially designed approach still dominates in each time series analysis task such as classification, anomaly detection, forecasting, and few-shot learning. The main challenge that blocks the development of pre-trained model for time series analysis is the lack of a large amount of data for training. In this work, we address this challenge by leveraging language or CV models, pre-trained from billions of tokens, for time series analysis. Specifically, we refrain from altering the self-attention and feedforward layers of the residual blocks in the pre-trained language or image model. This model, known as the Frozen Pretrained Transformer (FPT), is evaluated through fine-tuning on all major types of tasks involving time series. Our results demonstrate that pre-trained models on natural language or images can lead to a comparable or state-of-the-art performance in all main time series analysis tasks, as illustrated in Figure 1. We also found both theoretically and empirically that the self-attention module behaviors similarly to principle component analysis (PCA), an observation that helps explains how transformer bridges the domain gap and a crucial step towards understanding the universality of a pre-trained transformer.The code is publicly available at https://github.com/DAMO-DI-ML/One_Fits_All."
                },
                {
                    "title": "Dish-TS: A General Paradigm for Alleviating Distribution Shift in Time Series Forecasting",
                    "abstract": "The distribution shift in Time Series Forecasting (TSF), indicating series distribution changes over time, largely hinders the performance of TSF models. Existing works towards distribution shift in time series are mostly limited in the quantification of distribution and, more importantly, overlook the potential shift between lookback and horizon windows. To address above challenges, we systematically summarize the distribution shift in TSF into two categories. Regarding lookback windows as input-space and horizon windows as output-space, there exist (i) intra-space shift, that the distribution within the input-space keeps shifted over time, and (ii) inter-space shift, that the distribution is shifted between input-space and output-space. Then we introduce, Dish-TS, a general neural paradigm for alleviating distribution shift in TSF. Specifically, for better distribution estimation, we propose the coefficient net (Conet), which can be any neural architectures, to map input sequences into learnable distribution coefficients. To relieve intra-space and inter-space shift, we organize Dish-TS as a Dual-Conet framework to separately learn the distribution of input- and output-space, which naturally captures the distribution difference of two spaces. In addition, we introduce a more effective training strategy for intractable Conet learning. Finally, we conduct extensive experiments on several datasets coupled with different state-of-the-art forecasting models. Experimental results show Dish-TS consistently boosts them with a more than 20% average improvement. Code is available at https://github.com/weifantt/Dish-TS."
                },
                {
                    "title": "MTS-Mixers: Multivariate Time Series Forecasting via Factorized Temporal and Channel Mixing",
                    "abstract": "Multivariate time series forecasting has been widely used in various practical scenarios. Recently, Transformer-based models have shown significant potential in forecasting tasks due to the capture of long-range dependencies. However, recent studies in the vision and NLP fields show that the role of attention modules is not clear, which can be replaced by other token aggregation operations. This paper investigates the contributions and deficiencies of attention mechanisms on the performance of time series forecasting. Specifically, we find that (1) attention is not necessary for capturing temporal dependencies, (2) the entanglement and redundancy in the capture of temporal and channel interaction affect the forecasting performance, and (3) it is important to model the mapping between the input and the prediction sequence. To this end, we propose MTS-Mixers, which use two factorized modules to capture temporal and channel dependencies. Experimental results on several real-world datasets show that MTS-Mixers outperform existing Transformer-based models with higher efficiency."
                },
                {
                    "title": "A Time Series is Worth 64 Words: Long-term Forecasting with Transformers",
                    "abstract": "We propose an efficient design of Transformer-based models for multivariate time series forecasting and self-supervised representation learning. It is based on two key components: (i) segmentation of time series into subseries-level patches which are served as input tokens to Transformer; (ii) channel-independence where each channel contains a single univariate time series that shares the same embedding and Transformer weights across all the series. Patching design naturally has three-fold benefit: local semantic information is retained in the embedding; computation and memory usage of the attention maps are quadratically reduced given the same look-back window; and the model can attend longer history. Our channel-independent patch time series Transformer (PatchTST) can improve the long-term forecasting accuracy significantly when compared with that of SOTA Transformer-based models. We also apply our model to self-supervised pre-training tasks and attain excellent fine-tuning performance, which outperforms supervised training on large datasets. Transferring of masked pre-trained representation on one dataset to others also produces SOTA forecasting accuracy. Code is available at: https://github.com/yuqinie98/PatchTST."
                },
                {
                    "title": "TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis",
                    "abstract": "Time series analysis is of immense importance in extensive applications, such as weather forecasting, anomaly detection, and action recognition. This paper focuses on temporal variation modeling, which is the common key problem of extensive analysis tasks. Previous methods attempt to accomplish this directly from the 1D time series, which is extremely challenging due to the intricate temporal patterns. Based on the observation of multi-periodicity in time series, we ravel out the complex temporal variations into the multiple intraperiod- and interperiod-variations. To tackle the limitations of 1D time series in representation capability, we extend the analysis of temporal variations into the 2D space by transforming the 1D time series into a set of 2D tensors based on multiple periods. This transformation can embed the intraperiod- and interperiod-variations into the columns and rows of the 2D tensors respectively, making the 2D-variations to be easily modeled by 2D kernels. Technically, we propose the TimesNet with TimesBlock as a task-general backbone for time series analysis. TimesBlock can discover the multi-periodicity adaptively and extract the complex temporal variations from transformed 2D tensors by a parameter-efficient inception block. Our proposed TimesNet achieves consistent state-of-the-art in five mainstream time series analysis tasks, including short- and long-term forecasting, imputation, classification, and anomaly detection. Code is available at this repository: https://github.com/thuml/TimesNet."
                },
                {
                    "title": "Are Transformers Effective for Time Series Forecasting?",
                    "abstract": "Recently, there has been a surge of Transformer-based solutions for the long-term time series forecasting (LTSF) task. Despite the growing performance over the past few years, we question the validity of this line of research in this work. Specifically, Transformers is arguably the most successful solution to extract the semantic correlations among the elements in a long sequence. However, in time series modeling, we are to extract the temporal relations in an ordered set of continuous points. While employing positional encoding and using tokens to embed sub-series in Transformers facilitate preserving some ordering information, the nature of the permutation-invariant self-attention mechanism inevitably results in temporal information loss. \nTo validate our claim, we introduce a set of embarrassingly simple one-layer linear models named LTSF-Linear for comparison. Experimental results on nine real-life datasets show that LTSF-Linear surprisingly outperforms existing sophisticated Transformer-based LTSF models in all cases, and often by a large margin. Moreover, we conduct comprehensive empirical studies to explore the impacts of various design elements of LTSF models on their temporal relation extraction capability. We hope this surprising finding opens up new research directions for the LTSF task. We also advocate revisiting the validity of Transformer-based solutions for other time series analysis tasks (e.g., anomaly detection) in the future."
                },
                {
                    "title": "FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting",
                    "abstract": "Although Transformer-based methods have significantly improved state-of-the-art results for long-term series forecasting, they are not only computationally expensive but more importantly, are unable to capture the global view of time series (e.g. overall trend). To address these problems, we propose to combine Transformer with the seasonal-trend decomposition method, in which the decomposition method captures the global profile of time series while Transformers capture more detailed structures. To further enhance the performance of Transformer for long-term prediction, we exploit the fact that most time series tend to have a sparse representation in well-known basis such as Fourier transform, and develop a frequency enhanced Transformer. Besides being more effective, the proposed method, termed as Frequency Enhanced Decomposed Transformer ({\\bf FEDformer}), is more efficient than standard Transformer with a linear complexity to the sequence length. Our empirical studies with six benchmark datasets show that compared with state-of-the-art methods, FEDformer can reduce prediction error by $14.8\\%$ and $22.6\\%$ for multivariate and univariate time series, respectively. Code is publicly available at https://github.com/MAZiqing/FEDformer."
                },
                {
                    "title": "Long-Range Transformers for Dynamic Spatiotemporal Forecasting",
                    "abstract": "Multivariate time series forecasting focuses on predicting future values based on historical context. State-of-the-art sequence-to-sequence models rely on neural attention between timesteps, which allows for temporal learning but fails to consider distinct spatial relationships between variables. In contrast, methods based on graph neural networks explicitly model variable relationships. However, these methods often rely on predefined graphs that cannot change over time and perform separate spatial and temporal updates without establishing direct connections between each variable at every timestep. Our work addresses these problems by translating multivariate forecasting into a\"spatiotemporal sequence\"formulation where each Transformer input token represents the value of a single variable at a given time. Long-Range Transformers can then learn interactions between space, time, and value information jointly along this extended sequence. Our method, which we call Spacetimeformer, achieves competitive results on benchmarks from traffic forecasting to electricity demand and weather prediction while learning spatiotemporal relationships purely from data."
                },
                {
                    "title": "Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting",
                    "abstract": "Extending the forecasting time is a critical demand for real applications, such as extreme weather early warning and long-term energy consumption planning. This paper studies the long-term forecasting problem of time series. Prior Transformer-based models adopt various self-attention mechanisms to discover the long-range dependencies. However, intricate temporal patterns of the long-term future prohibit the model from finding reliable dependencies. Also, Transformers have to adopt the sparse versions of point-wise self-attentions for long series efficiency, resulting in the information utilization bottleneck. Going beyond Transformers, we design Autoformer as a novel decomposition architecture with an Auto-Correlation mechanism. We break with the pre-processing convention of series decomposition and renovate it as a basic inner block of deep models. This design empowers Autoformer with progressive decomposition capacities for complex time series. Further, inspired by the stochastic process theory, we design the Auto-Correlation mechanism based on the series periodicity, which conducts the dependencies discovery and representation aggregation at the sub-series level. Auto-Correlation outperforms self-attention in both efficiency and accuracy. In long-term forecasting, Autoformer yields state-of-the-art accuracy, with a 38% relative improvement on six benchmarks, covering five practical applications: energy, traffic, economics, weather and disease. Code is available at this repository: \\url{https://github.com/thuml/Autoformer}."
                },
                {
                    "title": "Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting",
                    "abstract": "Many real-world applications require the prediction of long sequence time-series, such as electricity consumption planning. Long sequence time-series forecasting (LSTF) demands a high prediction capacity of the model, which is the ability to capture precise long-range dependency coupling between output and input efficiently. Recent studies have shown the potential of Transformer to increase the prediction capacity. However, there are several severe issues with Transformer that prevent it from being directly applicable to LSTF, including quadratic time complexity, high memory usage, and inherent limitation of the encoder-decoder architecture. To address these issues, we design an efficient transformer-based model for LSTF, named Informer, with three distinctive characteristics: (i) a ProbSparse self-attention mechanism, which achieves O(L log L) in time complexity and memory usage, and has comparable performance on sequences' dependency alignment. (ii) the self-attention distilling highlights dominating attention by halving cascading layer input, and efficiently handles extreme long input sequences. (iii) the generative style decoder, while conceptually simple, predicts the long time-series sequences at one forward operation rather than a step-by-step way, which drastically improves the inference speed of long-sequence predictions. Extensive experiments on four large-scale datasets demonstrate that Informer significantly outperforms existing methods and provides a new solution to the LSTF problem."
                },
                {
                    "title": "Spectral Temporal Graph Neural Network for Multivariate Time-series Forecasting",
                    "abstract": "Multivariate time-series forecasting plays a crucial role in many real-world applications. It is a challenging problem as one needs to consider both intra-series temporal correlations and inter-series correlations simultaneously. Recently, there have been multiple works trying to capture both correlations, but most, if not all of them only capture temporal correlations in the time domain and resort to pre-defined priors as inter-series relationships. In this paper, we propose Spectral Temporal Graph Neural Network (StemGNN) to further improve the accuracy of multivariate time-series forecasting. StemGNN captures inter-series correlations and temporal dependencies jointly in the spectral domain. It combines Graph Fourier Transform (GFT) which models inter-series correlations and Discrete Fourier Transform (DFT) which models temporal dependencies in an end-to-end framework. After passing through GFT and DFT, the spectral representations hold clear patterns and can be predicted effectively by convolution and sequential learning modules. Moreover, StemGNN learns inter-series correlations automatically from the data without using pre-defined priors. We conduct extensive experiments on ten real-world datasets to demonstrate the effectiveness of StemGNN."
                },
                {
                    "title": "Connecting the Dots: Multivariate Time Series Forecasting with Graph Neural Networks",
                    "abstract": "Modeling multivariate time series has long been a subject that has attracted researchers from a diverse range of fields including economics, finance, and traffic. A basic assumption behind multivariate time series forecasting is that its variables depend on one another but, upon looking closely, it is fair to say that existing methods fail to fully exploit latent spatial dependencies between pairs of variables. In recent years, meanwhile, graph neural networks (GNNs) have shown high capability in handling relational dependencies. GNNs require well-defined graph structures for information propagation which means they cannot be applied directly for multivariate time series where the dependencies are not known in advance. In this paper, we propose a general graph neural network framework designed specifically for multivariate time series data. Our approach automatically extracts the uni-directed relations among variables through a graph learning module, into which external knowledge like variable attributes can be easily integrated. A novel mix-hop propagation layer and a dilated inception layer are further proposed to capture the spatial and temporal dependencies within the time series. The graph learning, graph convolution, and temporal convolution modules are jointly learned in an end-to-end framework. Experimental results show that our proposed model outperforms the state-of-the-art baseline methods on 3 of 4 benchmark datasets and achieves on-par performance with other approaches on two traffic datasets which provide extra structural information."
                },
                {
                    "title": "Spatial-Temporal Synchronous Graph Convolutional Networks: A New Framework for Spatial-Temporal Network Data Forecasting",
                    "abstract": "Spatial-temporal network data forecasting is of great importance in a huge amount of applications for traffic management and urban planning. However, the underlying complex spatial-temporal correlations and heterogeneities make this problem challenging. Existing methods usually use separate components to capture spatial and temporal correlations and ignore the heterogeneities in spatial-temporal data. In this paper, we propose a novel model, named Spatial-Temporal Synchronous Graph Convolutional Networks (STSGCN), for spatial-temporal network data forecasting. The model is able to effectively capture the complex localized spatial-temporal correlations through an elaborately designed spatial-temporal synchronous modeling mechanism. Meanwhile, multiple modules for different time periods are designed in the model to effectively capture the heterogeneities in localized spatial-temporal graphs. Extensive experiments are conducted on four real-world datasets, which demonstrates that our method achieves the state-of-the-art performance and consistently outperforms other baselines."
                },
                {
                    "title": "Modeling Long- and Short-Term Temporal Patterns with Deep Neural Networks",
                    "abstract": "Multivariate time series forecasting is an important machine learning problem across many domains, including predictions of solar plant energy output, electricity consumption, and traffic jam situation. Temporal data arise in these real-world applications often involves a mixture of long-term and short-term patterns, for which traditional approaches such as Autoregressive models and Gaussian Process may fail. In this paper, we proposed a novel deep learning framework, namely Long- and Short-term Time-series network (LSTNet), to address this open challenge. LSTNet uses the Convolution Neural Network (CNN) and the Recurrent Neural Network (RNN) to extract short-term local dependency patterns among variables and to discover long-term patterns for time series trends. Furthermore, we leverage traditional autoregressive model to tackle the scale insensitive problem of the neural network model. In our evaluation on real-world data with complex mixtures of repetitive patterns, LSTNet achieved significant performance improvements over that of several state-of-the-art baseline methods. All the data and experiment codes are available online."
                },
                {
                    "title": "Crossformer: Transformer Utilizing Cross-Dimension Dependency for Multivariate Time Series Forecasting",
                    "abstract": "Recently many deep models have been proposed for multivariate time series (MTS) forecasting. In particular, Transformer-based models have shown great potential because they can capture long-term dependency. However, existing Transformer-based models mainly focus on modeling the temporal dependency (cross-time dependency) yet often omit the dependency among different variables (cross-dimension dependency), which is critical for MTS forecasting. To \ufb01ll the gap, we propose Crossformer, a Transformer-based model utilizing cross-dimension dependency for MTS forecasting. In Crossformer, the input MTS is embedded into a 2D vector array through the Dimension-Segment-Wise (DSW) embedding to preserve time and dimension information. Then the Two-Stage Attention (TSA) layer is proposed to ef\ufb01ciently capture the cross-time and cross-dimension dependency. Utilizing DSW embedding and TSA layer, Crossformer establishes a Hierarchical Encoder-Decoder (HED) to use the information at different scales for the \ufb01nal forecasting. Extensive experimental results on six real-world datasets show the effectiveness of Crossformer against previous state-of-the-arts."
                },
                {
                    "title": "Power Time Series Forecasting by Pretrained LM",
                    "abstract": "The diversity and domain dependence of time series data pose signi\ufb01cant challenges in transferring learning to time series forecasting. In this study, we examine the effectiveness of us-ing a transformer model that has been pre-trained on natural language or image data and then \ufb01ne-tuned for time series forecasting with minimal modi\ufb01cations, speci\ufb01cally, without altering the self-attention and feedforward layers of the residual blocks. This model, known as the Frozen Pre-trained Transformer (FPT), is evaluated through \ufb01ne-tuning on time series forecasting tasks under Zero-Shot, Few-Shot, and normal sample size conditions. Our results demonstrate that pre-training on natural language or images can lead to a comparable or state-of-the-art performance in cross-modality time series forecasting tasks, in contrast to previous studies that focused on \ufb01ne-tuning within the same modality as the pre-training data. Additionally, we provide a comprehensive theoretical analysis of the universality and the functionality of the FPT. The code is publicly available at https://anonymous.4open.science/r/Pretrained-LM-for-TSForcasting-C561."
                },
                {
                    "title": "Adaptive Normalization for Non-stationary Time Series Forecasting: A Temporal Slice Perspective",
                    "abstract": "Deep learning models have progressively advanced time series forecasting due to their powerful capacity in capturing sequence dependence. Nevertheless, it is still challenging to make accurate predictions due to the existence of non-stationarity in real-world data, denoting the data distribution rapidly changes over time. To mitigate such a dilemma, several efforts have been conducted by reducing the non-stationarity with normalization operation. However, these methods typically overlook the distribution discrepancy between the input series and the horizon series, and assume that all time points within the same instance share the same statistical properties, which is too ideal and may lead to suboptimal relative improvements. To this end, we propose a novel slice-level adaptive normalization, referred to SAN , which is a novel scheme for empowering time series forecasting with more flexible normalization and denormalization. SAN includes two crucial designs. First, SAN tries to eliminate the non-stationarity of time series in units of a local temporal slice (i.e., sub-series) rather than a global instance. Second, SAN employs a slight network module to independently model the evolving trends of statistical properties of raw time series. Consequently, SAN could serve as a general model-agnostic plugin and better alleviate the impact of the non-stationary nature of time series data. We instantiate the proposed SAN on four widely used forecasting models and test their prediction results on benchmark datasets to evaluate its effectiveness. Also, we report some insightful findings to deeply analyze and understand our proposed SAN. We make our codes publicly available 2 ."
                },
                {
                    "title": "CrossGNN: Confronting Noisy Multivariate Time Series Via Cross Interaction Refinement",
                    "abstract": "Recently, multivariate time series (MTS) forecasting techniques have seen rapid development and widespread applications across various fields. Transformer-based and GNN-based methods have shown promising potential due to their strong ability to model interaction of time and variables. However, by conducting a comprehensive analysis of the real-world data, we observe that the temporal fluctuations and heterogeneity between variables are not well handled by existing methods. To address the above issues, we propose CrossGNN, a linear complexity GNN model to refine the cross-scale and cross-variable interaction for MTS. To deal with the unexpected noise in time dimension, an adaptive multi-scale identifier (AMSI) is leveraged to construct multi-scale time series with reduced noise. A Cross-Scale GNN is proposed to extract the scales with clearer trend and weaker noise. Cross-Variable GNN is proposed to utilize the homogeneity and heterogeneity between different variables. By simultaneously focusing on edges with higher saliency scores and constraining those edges with lower scores, the time and space complexity (i.e., O ( L ) ) of CrossGNN can be linear with the input sequence length L . Extensive experimental results on 8 real-world MTS datasets demonstrate the effectiveness of CrossGNN compared with state-of-the-art methods. The code is available at https://github"
                },
                {
                    "title": "Pyraformer: Low-Complexity Pyramidal Attention for Long-Range Time Series Modeling and Forecasting",
                    "abstract": "Accurate prediction of the future given the past based on time series data is of paramount importance, since it opens the door for decision making and risk management ahead of time. In practice, the challenge is to build a flexible but parsimonious model that can capture a wide range of temporal dependencies. In this paper, we propose Pyraformer by exploring the multi-resolution representation of the time series. Specifically, we introduce the pyramidal attention module (PAM) in which the inter-scale tree structure summarizes features at different resolutions and the intra-scale neighboring connections model the temporal dependencies of different ranges. Under mild conditions, the maximum length of the signal traversing path in Pyraformer is a constant (i.e., O(1)) with regard to the sequence length L, while its time and space complexity scale linearly with L. Extensive experimental results show that Pyraformer typically achieves the highest prediction accuracy in both single-step and long-range multi-step forecasting tasks with the least amount of time and memory consumption, especially when the sequence is long1."
                },
                {
                    "title": "Reversible Instance Normalization for Accurate Time-Series Forecasting against Distribution Shift",
                    "abstract": "Statistical properties such as mean and variance often change over time in time series, i.e., time-series data suffer from a distribution shift problem. This change in temporal distribution is one of the main challenges that prevent accurate timeseries forecasting. To address this issue, we propose a simple yet effective normalization method called reversible instance normalization (RevIN), a generallyapplicable normalization-and-denormalization method with learnable affine transformation. The proposed method is symmetrically structured to remove and restore the statistical information of a time-series instance, leading to significant performance improvements in time-series forecasting, as shown in Fig. 1. We demonstrate the effectiveness of RevIN via extensive quantitative and qualitative analyses on various real-world datasets, addressing the distribution shift problem. \u2217Both authors contributed equally. The order of the first authors was determined by coin flip."
                },
                {
                    "title": "Large Bayesian vector auto regressions",
                    "abstract": "This paper shows that vector auto regression (VAR) with Bayesian shrinkage is an appropriate tool for large dynamic models. We build on the results of De Mol and co-workers (2008) and show that, when the degree of shrinkage is set in relation to the cross-sectional dimension, the forecasting performance of small monetary VARs can be improved by adding additional macroeconomic variables and sectoral information. In addition, we show that large VARs with shrinkage produce credible impulse responses and are suitable for structural analysis. Copyright \u00a9 2009 John Wiley & Sons, Ltd."
                },
                {
                    "title": "On generalized harmonic analysis",
                    "abstract": ". Motivated by Wiener's work on generalized harmonic analysis, we consider the Marcinkiewicz space \u20acHy(R) of functions of bounded upper average/) power and the space 'ViR) of functions of bounded upper p variation. By identifying functions whose difference has norm zero, we show that ^(R), 1 <p < oo, is a Banach space. The proof depends on the result that each equivalence class in ^(R) contains a representative in LP(R). This result, in turn, is based on Masani's work on helixes in Banach spaces. Wiener defined an integrated Fourier transformation and proved that this transformation is an isometry from the nonlinear subspace ^(R) of <3H2(R) consisting of functions of bounded average quadratic power, into the nonlinear subspace %2(R) of ^(R) consisting of functions of bounded quadratic variation. By using two generalized Tauberian theorems, we prove that Wiener's transformation W is actually an isomorphism from 9H2(R) onto ^(R). We also show by counterexamples that W is not an isometry on the closed subspace generated by ^(R)."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we efficiently estimate cross-correlation between time series data using frequency-domain analysis?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it can significantly enhance the understanding of relationships between time series data across various fields, such as finance, neuroscience, and environmental science. By improving cross-correlation estimation, researchers can uncover hidden patterns and dependencies that may lead to new insights and practical applications, such as better predictive models and real-time monitoring systems. This advancement could pave the way for future research that leverages these insights to develop more sophisticated machine learning algorithms and analytical tools.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the complexities of accurately capturing the relationships between time series data, especially when they are subject to noise and varying temporal dynamics. Naive approaches may fail due to their inability to account for phase shifts and the non-stationarity of the data. Additionally, the computational burden of traditional time-domain methods can be prohibitive, particularly for large datasets. Overcoming these technical obstacles requires a robust methodology that effectively utilizes frequency-domain properties while maintaining computational efficiency.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on either time-domain or frequency-domain methods in isolation, leading to a lack of integrated approaches that leverage the strengths of both. Existing solutions may not adequately address the challenges posed by real-world data, such as noise and non-stationarity. Barriers such as limited computational resources and the complexity of implementing advanced mathematical techniques have also hindered progress. Our approach differs by combining efficient frequency-domain analysis with robust statistical methods to provide a more comprehensive solution to cross-correlation estimation.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves using the Discrete Fourier Transform (DFT) to analyze time series data and estimate cross-correlation in the frequency domain. We will utilize a dataset of paired time series and evaluate the cross-correlation using the defined frequency components. The key metrics for success will include the accuracy of the cross-correlation estimates and computational efficiency. We expect our approach to yield more accurate and faster cross-correlation estimates, enabling better insights into the relationships between time series data."
            }
        },
        "author_data": {
            "4a3746f0-0f77-44d9-9a09-908bf3d8b93a": {
                "pk": "4a3746f0-0f77-44d9-9a09-908bf3d8b93a",
                "name": "Lifan Zhao",
                "collaborators": [
                    "Yanyan Shen",
                    "Liping Wang",
                    "Jiawei Li",
                    "Zhizhuo Kou",
                    "Xiaohan Wang",
                    "Xinyi Zhu",
                    "Hao Wang",
                    "Lei Chen",
                    "Shuming Kong",
                    "Weiyu Cheng"
                ],
                "domain": [
                    "Stock Market Prediction",
                    "Recommender Systems",
                    "Meta-Learning",
                    "Cross-Domain Recommendation"
                ],
                "publications": [
                    {
                        "title": "Methods for Acquiring and Incorporating Knowledge into Stock Price Prediction: A Survey",
                        "abstract": "Predicting stock prices presents a challenging research problem due to the inherent volatility and non-linear nature of the stock market. In recent years, knowledge-enhanced stock price prediction methods have shown groundbreaking results by utilizing external knowledge to understand the stock market. Despite the importance of these methods, there is a scarcity of scholarly works that systematically synthesize previous studies from the perspective of external knowledge types. Specifically, the external knowledge can be modeled in different data structures, which we group into non-graph-based formats and graph-based formats: 1) non-graph-based knowledge captures contextual information and multimedia descriptions specifically associated with an individual stock; 2) graph-based knowledge captures interconnected and interdependent information in the stock market. This survey paper aims to provide a systematic and comprehensive description of methods for acquiring external knowledge from various unstructured data sources and then incorporating it into stock price prediction models. We also explore fusion methods for combining external knowledge with historical price features. Moreover, this paper includes a compilation of relevant datasets and delves into potential future research directions in this domain."
                    },
                    {
                        "title": "DoubleAdapt: A Meta-learning Approach to Incremental Learning for Stock Trend Forecasting",
                        "abstract": "Stock trend forecasting is a fundamental task of quantitative investment where precise predictions of price trends are indispensable. As an online service, stock data continuously arrive over time. It is practical and efficient to incrementally update the forecast model with the latest data which may reveal some new patterns recurring in the future stock market. However, incremental learning for stock trend forecasting still remains under-explored due to the challenge of distribution shifts (a.k.a. concept drifts). With the stock market dynamically evolving, the distribution of future data can slightly or significantly differ from incremental data, hindering the effectiveness of incremental updates. To address this challenge, we propose DoubleAdapt, an end-to-end framework with two adapters, which can effectively adapt the data and the model to mitigate the effects of distribution shifts. Our key insight is to automatically learn how to adapt stock data into a locally stationary distribution in favor of profitable updates. Complemented by data adaptation, we can confidently adapt the model parameters under mitigated distribution shifts. We cast each incremental learning task as a meta-learning task and automatically optimize the adapters for desirable data adaptation and parameter initialization. Experiments on real-world stock datasets demonstrate that DoubleAdapt achieves state-of-the-art predictive performance and shows considerable efficiency."
                    },
                    {
                        "title": "RESUS: Warm-up Cold Users via Meta-learning Residual User Preferences in CTR Prediction",
                        "abstract": "Click-through Rate (CTR) prediction on cold users is a challenging task in recommender systems. Recent researches have resorted to meta-learning to tackle the cold-user challenge, which either perform few-shot user representation learning or adopt optimization-based meta-learning. However, existing methods suffer from information loss or inefficient optimization process, and they fail to explicitly model global user preference knowledge, which is crucial to complement the sparse and insufficient preference information of cold users. In this article, we propose a novel and efficient approach named RESUS, which decouples the learning of global preference knowledge contributed by collective users from the learning of residual preferences for individual users. Specifically, we employ a shared predictor to infer basis user preferences, which acquires global preference knowledge from the interactions of different users. Meanwhile, we develop two efficient algorithms based on the nearest neighbor and ridge regression predictors, which infer residual user preferences via learning quickly from a few user-specific interactions. Extensive experiments on three public datasets demonstrate that our RESUS approach is efficient and effective in improving CTR prediction accuracy on cold users, compared with various state-of-the-art methods."
                    },
                    {
                        "title": "Mixed Information Flow for Cross-domain Sequential Recommendations",
                        "abstract": "  Cross-domain sequential recommendation is the task of predict the next item that the user is most likely to interact with based on past sequential behavior from multiple domains. One of the key challenges in  cross-domain sequential recommendation  is to grasp and transfer the flow of information from multiple domains so as to promote recommendations in all domains. Previous studies have investigated the flow of  behavioral  information by exploring the connection between items from different domains. The flow of knowledge (i.e., the connection between knowledge from different domains) has so far been neglected. In this article, we propose a  mixed information flow network  for  cross-domain sequential recommendation  to consider both the flow of behavioral information and the flow of knowledge by incorporating a  behavior transfer unit  and a  knowledge transfer unit  . The proposed  mixed information flow network  is able to decide when cross-domain information should be used and, if so, which cross-domain information should be used to enrich the sequence representation according to users\u2019 current preferences. Extensive experiments conducted on four e-commerce datasets demonstrate that the proposed  mixed information flow network  is able to improve recommendation performance in different domains by modeling mixed information flow. In this article, we focus on the application of  mixed information flow network  s to a scenario with two domains, but the method can easily be extended to multiple domains. "
                    }
                ]
            },
            "aa648d14-2a8f-4b8f-9783-4ea024d8e0d9": {
                "pk": "aa648d14-2a8f-4b8f-9783-4ea024d8e0d9",
                "name": "Yanyan Shen",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            }
        }
    },
    "2406.05882": {
        "paper_data": {
            "title": "Distributional Preference Alignment of LLMs via Optimal Transport",
            "url": "http://arxiv.org/abs/2406.05882v1",
            "arxiv_id": "2406.05882",
            "authors": [
                "Igor Melnyk",
                "Youssef Mroueh",
                "Brian Belgodere",
                "Mattia Rigotti",
                "Apoorva Nitsure",
                "Mikhail Yurochkin",
                "Kristjan Greenewald",
                "Jiri Navratil",
                "Jerret Ross"
            ],
            "abstract": "Current LLM alignment techniques use pairwise human preferences at a sample level, and as such, they do not imply an alignment on the distributional level. We propose in this paper Alignment via Optimal Transport (AOT), a novel method for distributional preference alignment of LLMs. AOT aligns LLMs on unpaired preference data by making the reward distribution of the positive samples stochastically dominant in the first order on the distribution of negative samples. We introduce a convex relaxation of this first-order stochastic dominance and cast it as an optimal transport problem with a smooth and convex cost. Thanks to the one-dimensional nature of the resulting optimal transport problem and the convexity of the cost, it has a closed-form solution via sorting on empirical measures. We fine-tune LLMs with this AOT objective, which enables alignment by penalizing the violation of the stochastic dominance of the reward distribution of the positive samples on the reward distribution of the negative samples. We analyze the sample complexity of AOT by considering the dual of the OT problem and show that it converges at the parametric rate. Empirically, we show on a diverse set of alignment datasets and LLMs that AOT leads to state-of-the-art models in the 7B family of models when evaluated with Open LLM Benchmarks and AlpacaEval.",
            "introduction": "   1. Introduction  Aligning Large Language Models (LLMs) with human preferences is a crucial step in making these models safe and having them follow instructions faithfully. By ensuring that LLMs adhere to human preferences, values, ethics, and desired behaviors we can reduce the risk of generating harmful, biased, or inappropriate content.   Reinforcement Learning from Human Feedback, RLHF (Christiano et\u00a0al., 2017; Stiennon et\u00a0al., 2020; Ouyang et\u00a0al., 2022; Bai et\u00a0al., 2022), achieves this by learning a reward model on human preference data, followed by fine-tuning the LLM to maximize the reward score while staying close to the initial reference policy to retain utility from the pre-trained model. Recently, new paradigms departed from RLHF towards direct preference optimization methods such as DPO (Rafailov et\u00a0al., 2024), SLIC (Zhao et\u00a0al., 2023), and Identity Policy optimization (Azar et\u00a0al., 2024). In these approaches, the reward is expressed in terms of the log-likelihood ratio between the LLM policy and the reference model. The training is done on paired preference data, i.e.\u00a0as triplets of prompts, chosen and rejected sentences, where for each prompt a chosen and a rejected sample are available. The training objective is to maximize the margin between the log-likelihood ratio evaluated on the chosen sentence versus the log-likelihood ratio on rejected sentences. When paired preference data is not available, and the preference data instead takes the form of distinct marginals of chosen prompt/response pairs and rejected prompt/response pairs, we refer to this setup as the unpaired data setting. Ethayarajh et\u00a0al. (2024) used Kahneman & Tversky\u2019s prospect theory in the unpaired setting and proposed the KTO method that maximizes the margin between the chosen reward and the average reward of rejected sentences and pushes the reward of a rejected sentence below the average reward of chosen sentences.   In this paper, we introduce a new distributional optimization method for fine-tuning LLMs from human preference data. Previous work in the paired setting focused on improving the reward of chosen sentences over rejected sentences on a per-sample basis. This procedure does not lead to a preference on a distributional level of the chosen marginal on the rejected marginal. In probabilistic terms, we would like to induce stochastic dominance of the reward of chosen sentences on the reward of rejected ones. First order Stochastic Dominance (FSD, see e.g.\u00a0Ogryczak and Ruszczynski, 2002) of a random variable X\ud835\udc4bXitalic_X on a random variable Y\ud835\udc4cYitalic_Y, means that all quantiles values of X\ud835\udc4bXitalic_X are larger than those of Y\ud835\udc4cYitalic_Y.      (a) Stochastic Dominance of Reward of Chosen on Rejected: \ud835\udda0\ud835\uddae\ud835\uddb3\ud835\udda0\ud835\uddae\ud835\uddb3\\mathsf{AOT}sansserif_AOT achieves a larger margin between the quantile plots of chosen and rejected rewards.      (b) Stochastic Dominance of \ud835\udda0\ud835\uddae\ud835\uddb3\ud835\udda0\ud835\uddae\ud835\uddb3\\mathsf{AOT}sansserif_AOT\u2019s optimized policy margin (between Chosen on Rejected) on the margin of the reference policy.     Figure 1. \ud835\udda0\ud835\uddae\ud835\uddb3\ud835\udda0\ud835\uddae\ud835\uddb3\\mathsf{AOT}sansserif_AOT in the paired & unpaired settings enables first-order stochastic dominance of the chosen reward distribution on the rejected distribution (a). The margin between the quantiles of chosen and rejected rewards is larger than alternative strategies. In (b), we see that \ud835\udda0\ud835\uddae\ud835\uddb3\ud835\udda0\ud835\uddae\ud835\uddb3\\mathsf{AOT}sansserif_AOT\u2019s policy chosen to rejected log-likelihood ratio dominates that ratio for the base model and alternative strategies.   Our main contribution is introducing \ud835\udda0\ud835\uddae\ud835\uddb3\ud835\udda0\ud835\uddae\ud835\uddb3\\mathsf{AOT}sansserif_AOT, Alignment via Optimal Transport, a new method",
            "references": [
                {
                    "title": "Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators",
                    "abstract": "LLM-based auto-annotators have become a key component of the LLM development process due to their cost-effectiveness and scalability compared to human-based evaluation. However, these auto-annotators can introduce complex biases that are hard to remove. Even simple, known confounders such as preference for longer outputs remain in existing automated evaluation metrics. We propose a simple regression analysis approach for controlling biases in auto-evaluations. As a real case study, we focus on reducing the length bias of AlpacaEval, a fast and affordable benchmark for chat LLMs that uses LLMs to estimate response quality. Despite being highly correlated with human preferences, AlpacaEval is known to favor models that generate longer outputs. We introduce a length-controlled AlpacaEval that aims to answer the counterfactual question:\"What would the preference be if the model's and baseline's output had the same length?\". To achieve this, we first fit a generalized linear model to predict the biased output of interest (auto-annotator preferences) based on the mediators we want to control for (length difference) and other relevant features. We then obtain length-controlled preferences by predicting preferences while conditioning the GLM with a zero difference in lengths. Length-controlling not only improves the robustness of the metric to manipulations in model verbosity, we also find that it increases the Spearman correlation with LMSYS' Chatbot Arena from 0.94 to 0.98. We release the code and leaderboard at https://tatsu-lab.github.io/alpaca_eval/ ."
                },
                {
                    "title": "LAB: Large-Scale Alignment for ChatBots",
                    "abstract": "This work introduces LAB (Large-scale Alignment for chatBots), a novel methodology designed to overcome the scalability challenges in the instruction-tuning phase of large language model (LLM) training. Leveraging a taxonomy-guided synthetic data generation process and a multi-phase tuning framework, LAB significantly reduces reliance on expensive human annotations and proprietary models like GPT-4. We demonstrate that LAB-trained models can achieve competitive performance across several benchmarks compared to models trained with traditional human-annotated or GPT-4 generated synthetic data. Thus offering a scalable, cost-effective solution for enhancing LLM capabilities and instruction-following behaviors without the drawbacks of catastrophic forgetting, marking a step forward in the efficient training of LLMs for a wide range of applications."
                },
                {
                    "title": "tinyBenchmarks: evaluating LLMs with fewer examples",
                    "abstract": "The versatility of large language models (LLMs) led to the creation of diverse benchmarks that thoroughly test a variety of language models' abilities. These benchmarks consist of tens of thousands of examples making evaluation of LLMs very expensive. In this paper, we investigate strategies to reduce the number of evaluations needed to assess the performance of an LLM on several key benchmarks. For example, we show that to accurately estimate the performance of an LLM on MMLU, a popular multiple-choice QA benchmark consisting of 14K examples, it is sufficient to evaluate this LLM on 100 curated examples. We release evaluation tools and tiny versions of popular benchmarks: Open LLM Leaderboard, MMLU, HELM, and AlpacaEval 2.0. Our empirical analysis demonstrates that these tools and tiny benchmarks are sufficient to reliably and efficiently reproduce the original evaluation results."
                },
                {
                    "title": "KTO: Model Alignment as Prospect Theoretic Optimization",
                    "abstract": "Kahneman&Tversky's $\\textit{prospect theory}$ tells us that humans perceive random variables in a biased but well-defined manner (1992); for example, humans are famously loss-averse. We show that objectives for aligning LLMs with human feedback implicitly incorporate many of these biases -- the success of these objectives (e.g., DPO) over cross-entropy minimization can partly be ascribed to them belonging to a family of loss functions that we call $\\textit{human-aware losses}$ (HALOs). However, the utility functions these methods attribute to humans still differ from those in the prospect theory literature. Using a Kahneman-Tversky model of human utility, we propose a HALO that directly maximizes the utility of generations instead of maximizing the log-likelihood of preferences, as current methods do. We call this approach KTO, and it matches or exceeds the performance of preference-based methods at scales from 1B to 30B, despite only learning from a binary signal of whether an output is desirable. More broadly, our work suggests that there is no one HALO that is universally superior; the best loss depends on the inductive biases most appropriate for a given setting, an oft-overlooked consideration."
                },
                {
                    "title": "HelpSteer: Multi-attribute Helpfulness Dataset for SteerLM",
                    "abstract": "Existing open-source helpfulness preference datasets do not specify what makes some responses more helpful and others less so. Models trained on these datasets can incidentally learn to model dataset artifacts (e.g. preferring longer but unhelpful responses only due to their length). To alleviate this problem, we collect HelpSteer, a multi-attribute helpfulness dataset annotated for the various aspects that make responses helpful. Specifically, our 37k-sample dataset has annotations for correctness, coherence, complexity, and verbosity in addition to overall helpfulness of responses. Training Llama 2 70B using the HelpSteer dataset with SteerLM technique produces a model that scores 7.54 on MT Bench, which is currently the highest score for open models that do not require training data from more powerful models (e.g. GPT-4). We release this dataset with CC-BY-4.0 license at https://huggingface.co/datasets/nvidia/HelpSteer"
                },
                {
                    "title": "A General Theoretical Paradigm to Understand Learning from Human Preferences",
                    "abstract": "The prevalent deployment of learning from human preferences through reinforcement learning (RLHF) relies on two important approximations: the first assumes that pairwise preferences can be substituted with pointwise rewards. The second assumes that a reward model trained on these pointwise rewards can generalize from collected data to out-of-distribution data sampled by the policy. Recently, Direct Preference Optimisation (DPO) has been proposed as an approach that bypasses the second approximation and learn directly a policy from collected data without the reward modelling stage. However, this method still heavily relies on the first approximation. In this paper we try to gain a deeper theoretical understanding of these practical algorithms. In particular we derive a new general objective called $\\Psi$PO for learning from human preferences that is expressed in terms of pairwise preferences and therefore bypasses both approximations. This new general objective allows us to perform an in-depth analysis of the behavior of RLHF and DPO (as special cases of $\\Psi$PO) and to identify their potential pitfalls. We then consider another special case for $\\Psi$PO by setting $\\Psi$ simply to Identity, for which we can derive an efficient optimisation procedure, prove performance guarantees and demonstrate its empirical superiority to DPO on some illustrative examples."
                },
                {
                    "title": "BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset",
                    "abstract": "In this paper, we introduce the \\textsc{BeaverTails} dataset, aimed at fostering research on safety alignment in large language models (LLMs). This dataset uniquely separates annotations of helpfulness and harmlessness for question-answering pairs, thus offering distinct perspectives on these crucial attributes. In total, we have gathered safety meta-labels for 30,207 question-answer (QA) pairs and 30,144 pairs of expert comparison data for both the helpfulness and harmlessness metrics. In total, we have gathered safety meta-labels for 333,963 question-answer (QA) pairs and 361,903 pairs of expert comparison data for both the helpfulness and harmlessness metrics. We further showcase applications of BeaverTails in content moderation and reinforcement learning with human feedback (RLHF), emphasizing its potential for practical safety measures in LLMs. We believe this dataset provides vital resources for the community, contributing towards the safe development and deployment of LLMs. Our project page is available at the following URL: https://sites.google.com/view/pku-beavertails. Warning: this paper contains example data that may be offensive or harmful."
                },
                {
                    "title": "Lower Complexity Adaptation for Empirical Entropic Optimal Transport",
                    "abstract": "Entropic optimal transport (EOT) presents an effective and computationally viable alternative to unregularized optimal transport (OT), offering diverse applications for large-scale data analysis. In this work, we derive novel statistical bounds for empirical plug-in estimators of the EOT cost and show that their statistical performance in the entropy regularization parameter $\\epsilon$ and the sample size $n$ only depends on the simpler of the two probability measures. For instance, under sufficiently smooth costs this yields the parametric rate $n^{-1/2}$ with factor $\\epsilon^{-d/2}$, where $d$ is the minimum dimension of the two population measures. This confirms that empirical EOT also adheres to the lower complexity adaptation principle, a hallmark feature only recently identified for unregularized OT. As a consequence of our theory, we show that the empirical entropic Gromov-Wasserstein distance and its unregularized version for measures on Euclidean spaces also obey this principle. Additionally, we comment on computational aspects and complement our findings with Monte Carlo simulations. Our techniques employ empirical process theory and rely on a dual formulation of EOT over a single function class. Crucial to our analysis is the observation that the entropic cost-transformation of a function class does not increase its uniform metric entropy by much."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "Calibrating Sequence likelihood Improves Conditional Language Generation",
                    "abstract": "Conditional language models are predominantly trained with maximum likelihood estimation (MLE), giving probability mass to sparsely observed target sequences. While MLE trained models assign high probability to plausible sequences given the context, the model probabilities often do not accurately rank-order generated sequences by quality. This has been empirically observed in beam search decoding as output quality degrading with large beam sizes, and decoding strategies benefiting from heuristics such as length normalization and repetition-blocking. In this work, we introduce sequence likelihood calibration (SLiC) where the likelihood of model generated sequences are calibrated to better align with reference sequences in the model's latent space. With SLiC, decoding heuristics become unnecessary and decoding candidates' quality significantly improves regardless of the decoding method. Furthermore, SLiC shows no sign of diminishing returns with model scale, and presents alternative ways to improve quality with limited training and inference budgets. With SLiC, we exceed or match SOTA results on a wide range of generation tasks spanning abstractive summarization, question generation, abstractive question answering and data-to-text generation, even with modest-sized models."
                },
                {
                    "title": "Learning with Stochastic Orders",
                    "abstract": "Learning high-dimensional distributions is often done with explicit likelihood modeling or implicit modeling via minimizing integral probability metrics (IPMs). In this paper, we expand this learning paradigm to stochastic orders, namely, the convex or Choquet order between probability measures. Towards this end, exploiting the relation between convex orders and optimal transport, we introduce the Choquet-Toland distance between probability measures, that can be used as a drop-in replacement for IPMs. We also introduce the Variational Dominance Criterion (VDC) to learn probability measures with dominance constraints, that encode the desired stochastic order between the learned measure and a known baseline. We analyze both quantities and show that they suffer from the curse of dimensionality and propose surrogates via input convex maxout networks (ICMNs), that enjoy parametric rates. We provide a min-max framework for learning with stochastic orders and validate it experimentally on synthetic and high-dimensional image generation, with promising results. Finally, our ICMNs class of convex functions and its derived Rademacher Complexity are of independent interest beyond their application in convex orders."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Empirical optimal transport between different measures adapts to lower complexity",
                    "abstract": "The empirical optimal transport (OT) cost between two probability measures from random data is a fundamental quantity in transport based data analysis. In this work, we derive novel guarantees for its convergence rate when the involved measures are different, possibly supported on different spaces. Our central observation is that the statistical performance of the empirical OT cost is determined by the less complex measure, a phenomenon we refer to as lower complexity adaptation of empirical OT. For instance, under Lipschitz ground costs, we find that the empirical OT cost based on $n$ observations converges at least with rate $n^{-1/d}$ to the population quantity if one of the two measures is concentrated on a $d$-dimensional manifold, while the other can be arbitrary. For semi-concave ground costs, we show that the upper bound for the rate improves to $n^{-2/d}$. Similarly, our theory establishes the general convergence rate $n^{-1/2}$ for semi-discrete OT. All of these results are valid in the two-sample case as well, meaning that the convergence rate is still governed by the simpler of the two measures. On a conceptual level, our findings therefore suggest that the curse of dimensionality only affects the estimation of the OT cost when both measures exhibit a high intrinsic dimension. Our proofs are based on the dual formulation of OT as a maximization over a suitable function class $\\mathcal{F}_c$ and the observation that the $c$-transform of $\\mathcal{F}_c$ under bounded costs has the same uniform metric entropy as $\\mathcal{F}_c$ itself."
                },
                {
                    "title": "Sharp convergence rates for empirical optimal transport with smooth costs",
                    "abstract": "We revisit the question of characterizing the convergence rate of plug-in estimators of optimal transport costs. It is well known that an empirical measure comprising independent samples from an absolutely continuous distribution on $\\mathbb{R}^d$ converges to that distribution at the rate $n^{-1/d}$ in Wasserstein distance, which can be used to prove that plug-in estimators of many optimal transport costs converge at this same rate. However, we show that when the cost is smooth, this analysis is loose: plug-in estimators based on empirical measures converge quadratically faster, at the rate $n^{-2/d}$. As a corollary, we show that the Wasserstein distance between two distributions is significantly easier to estimate when the measures are well-separated. We also prove lower bounds, showing not only that our analysis of the plug-in estimator is tight, but also that no other estimator can enjoy significantly faster rates of convergence uniformly over all pairs of measures. Our proofs rely on empirical process theory arguments based on tight control of $L^2$ covering numbers for locally Lipschitz and semi-concave functions. As a byproduct of our proofs, we derive $L^\\infty$ estimates on the displacement induced by the optimal coupling between any two measures satisfying suitable concentration and anticoncentration conditions, for a wide range of cost functions."
                },
                {
                    "title": "Learning to summarize from human feedback",
                    "abstract": "As language models become more powerful, training and evaluation are increasingly bottlenecked by the data and metrics used for a particular task. For example, summarization models are often trained to predict human reference summaries and evaluated using ROUGE, but both of these metrics are rough proxies for what we really care about---summary quality. In this work, we show that it is possible to significantly improve summary quality by training a model to optimize for human preferences. We collect a large, high-quality dataset of human comparisons between summaries, train a model to predict the human-preferred summary, and use that model as a reward function to fine-tune a summarization policy using reinforcement learning. We apply our method to a version of the TL;DR dataset of Reddit posts and find that our models significantly outperform both human reference summaries and much larger models fine-tuned with supervised learning alone. Our models also transfer to CNN/DM news articles, producing summaries nearly as good as the human reference without any news-specific fine-tuning. We conduct extensive analyses to understand our human feedback dataset and fine-tuned models We establish that our reward model generalizes to new datasets, and that optimizing our reward model results in better summaries than optimizing ROUGE according to humans. We hope the evidence from our paper motivates machine learning researchers to pay closer attention to how their training loss affects the model behavior they actually want."
                },
                {
                    "title": "Fast Differentiable Sorting and Ranking",
                    "abstract": "The sorting operation is one of the most commonly used building blocks in computer programming. In machine learning, it is often used for robust statistics. However, seen as a function, it is piecewise linear and as a result includes many kinks where it is non-differentiable. More problematic is the related ranking operator, often used for order statistics and ranking metrics. It is a piecewise constant function, meaning that its derivatives are null or undefined. While numerous works have proposed differentiable proxies to sorting and ranking, they do not achieve the $O(n \\log n)$ time complexity one would expect from sorting and ranking operations. In this paper, we propose the first differentiable sorting and ranking operators with $O(n \\log n)$ time and $O(n)$ space complexity. Our proposal in addition enjoys exact computation and differentiation. We achieve this feat by constructing differentiable operators as projections onto the permutahedron, the convex hull of permutations, and using a reduction to isotonic optimization. Empirically, we confirm that our approach is an order of magnitude faster than existing approaches and showcase two novel applications: differentiable Spearman's rank correlation coefficient and least trimmed squares."
                },
                {
                    "title": "Differentiable Ranks and Sorting using Optimal Transport",
                    "abstract": "Sorting an array is a fundamental routine in machine learning, one that is used to compute rank-based statistics, cumulative distribution functions (CDFs), quantiles, or to select closest neighbors and labels. The sorting function is however piece-wise constant (the sorting permutation of a vector does not change if the entries of that vector are infinitesimally perturbed) and therefore has no gradient information to back-propagate. We propose a framework to sort elements that is algorithmically differentiable. We leverage the fact that sorting can be seen as a particular instance of the optimal transport (OT) problem on $\\mathbb{R}$, from input values to a predefined array of sorted values (e.g. $1,2,\\dots,n$ if the input array has $n$ elements). Building upon this link , we propose generalized CDFs and quantile operators by varying the size and weights of the target presorted array. Because this amounts to using the so-called Kantorovich formulation of OT, we call these quantities K-sorts, K-CDFs and K-quantiles. We recover differentiable algorithms by adding to the OT problem an entropic regularization, and approximate it using a few Sinkhorn iterations. We call these operators S-sorts, S-CDFs and S-quantiles, and use them in various learning settings: we benchmark them against the recently proposed neuralsort [Grover et al. 2019], propose applications to quantile regression and introduce differentiable formulations of the top-k accuracy that deliver state-of-the art performance."
                },
                {
                    "title": "Sorting out Lipschitz function approximation",
                    "abstract": "Training neural networks under a strict Lipschitz constraint is useful for provable adversarial robustness, generalization bounds, interpretable gradients, and Wasserstein distance estimation. By the composition property of Lipschitz functions, it suffices to ensure that each individual affine transformation or nonlinear activation is 1-Lipschitz. The challenge is to do this while maintaining the expressive power. We identify a necessary property for such an architecture: each of the layers must preserve the gradient norm during backpropagation. Based on this, we propose to combine a gradient norm preserving activation function, GroupSort, with norm-constrained weight matrices. We show that norm-constrained GroupSort architectures are universal Lipschitz function approximators. Empirically, we show that norm-constrained GroupSort networks achieve tighter estimates of Wasserstein distance than their ReLU counterparts and can achieve provable adversarial robustness guarantees with little cost to accuracy."
                },
                {
                    "title": "Computational Optimal Transport",
                    "abstract": "Optimal transport (OT) theory can be informally described using the words of the French mathematician Gaspard Monge (1746-1818): A worker with a shovel in hand has to move a large pile of sand lying on a construction site. The goal of the worker is to erect with all that sand a target pile with a prescribed shape (for example, that of a giant sand castle). Naturally, the worker wishes to minimize her total effort, quantified for instance as the total distance or time spent carrying shovelfuls of sand. Mathematicians interested in OT cast that problem as that of comparing two probability distributions, two different piles of sand of the same volume. They consider all of the many possible ways to morph, transport or reshape the first pile into the second, and associate a \"global\" cost to every such transport, using the \"local\" consideration of how much it costs to move a grain of sand from one place to another. Recent years have witnessed the spread of OT in several fields, thanks to the emergence of approximate solvers that can scale to sizes and dimensions that are relevant to data sciences. Thanks to this newfound scalability, OT is being increasingly used to unlock various problems in imaging sciences (such as color or texture processing), computer vision and graphics (for shape manipulation) or machine learning (for regression, classification and density fitting). This short book reviews OT with a bias toward numerical methods and their applications in data sciences, and sheds lights on the theoretical properties of OT that make it particularly useful for some of these applications."
                },
                {
                    "title": "Deep Reinforcement Learning from Human Preferences",
                    "abstract": "For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems. In this work, we explore goals defined in terms of (non-expert) human preferences between pairs of trajectory segments. We show that this approach can effectively solve complex RL tasks without access to the reward function, including Atari games and simulated robot locomotion, while providing feedback on less than one percent of our agent's interactions with the environment. This reduces the cost of human oversight far enough that it can be practically applied to state-of-the-art RL systems. To demonstrate the flexibility of our approach, we show that we can successfully train complex novel behaviors with about an hour of human time. These behaviors and environments are considerably more complex than any that have been previously learned from human feedback."
                },
                {
                    "title": "Optimal Transport for Applied Mathematicians: Calculus of Variations, PDEs, and Modeling",
                    "abstract": "Preface.- Primal and Dual Problems.- One-Dimensional Issues.- L^1 and L^infinity Theory.- Minimal Flows.- Wasserstein Spaces.- Numerical Methods.- Functionals over Probabilities.- Gradient Flows.- Exercises.- References.- Index."
                },
                {
                    "title": "Zephyr",
                    "abstract": "Zephyr is an international deal information database from Bureau van Dijk (BvD), a Belgium-based provider of business information. BvD has offered Zephyr since 2002 when it acquired Zephus Corporate Finance Knowledge, a European deal data provider that offered information from 1997 onwards. BvD combined the Zephus deal data product with BvD\u2019s company information products to create Zephyr. The resulting database includes rumored, announced, and completed merger and acquisitions (M&A) transactions, initial public offerings (IPOs), secondary equity offerings (SEOs), and private equity and venture capital (VC) deals, with integrated financial and ownership information. This review is based on information from the Bureau van Dijk website, communications with Bureau van Dijk representatives, and searches within Zephyr."
                },
                {
                    "title": "Convexity, Classification, and Risk Bounds",
                    "abstract": "Many of the classification algorithms developed in the machine learning literature, including the support vector machine and boosting, can be viewed as minimum contrast methods that minimize a convex surrogate of the 0\u20131 loss function. The convexity makes these algorithms computationally efficient. The use of a surrogate, however, has statistical consequences that must be balanced against the computational virtues of convexity. To study these issues, we provide a general quantitative relationship between the risk as assessed using the 0\u20131 loss and the risk as assessed using any nonnegative surrogate loss function. We show that this relationship gives nontrivial upper bounds on excess risk under the weakest possible condition on the loss function\u2014that it satisfies a pointwise form of Fisher consistency for classification. The relationship is based on a simple variational transformation of the loss function that is easy to compute in many applications. We also present a refined version of this result in the case of low noise, and show that in this case, strictly convex loss functions lead to faster rates of convergence of the risk than would be implied by standard uniform convergence arguments. Finally, we present applications of our results to the estimation of convergence rates in function classes that are scaled convex hulls of a finite-dimensional base class, with a variety of commonly used loss functions."
                },
                {
                    "title": "Are Loss Functions All the Same?",
                    "abstract": "In this letter, we investigate the impact of choosing different loss functions from the viewpoint of statistical learning theory. We introduce a convexity assumption, which is met by all loss functions commonly used in the literature, and study how the bound on the estimation error changes with the loss. We also derive a general result on the minimizer of the expected risk for a convex loss function in the case of classification. The main outcome of our analysis is that for classification, the hinge loss appears to be the loss of choice. Other things being equal, the hinge loss leads to a convergence rate practically indistinguishable from the logistic loss rate and much better than the square loss rate. Furthermore, if the hypothesis space is sufficiently rich, the bounds obtained for the hinge loss are not loosened by the thresholding stage."
                },
                {
                    "title": "Dual Stochastic Dominance and Related Mean-Risk Models",
                    "abstract": "We consider the problem of constructing mean-risk models which are consistent with the second degree stochastic dominance relation. By exploiting duality relations of convex analysis we develop the quantile model of stochastic dominance for general distributions. This allows us to show that several models using quantiles and tail characteristics of the distribution are in harmony with the stochastic dominance relation. We also provide stochastic linear programming formulations of these models."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively fine-tune large language models (LLMs) using human preference data in both paired and unpaired settings to ensure that the reward distribution of chosen responses stochastically dominates that of rejected responses?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for enhancing the safety and reliability of LLMs, as it directly impacts their ability to align with human values and preferences. By improving the fine-tuning process through methods like Alignment via Optimal Transport (AOT), we can advance the research community's understanding of preference optimization in machine learning. This could lead to more robust applications of LLMs in sensitive areas such as healthcare, education, and content moderation, where adherence to ethical standards is paramount. Furthermore, it may inspire future research into more sophisticated alignment techniques that can be generalized across various AI systems.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the complexity of accurately modeling human preferences and ensuring that the fine-tuning process captures these preferences effectively. Naive approaches may fail because they often focus on individual sample improvements rather than the overall distribution of rewards, which can lead to suboptimal alignment. Additionally, the unpaired data setting introduces further complications, as it lacks direct comparisons between chosen and rejected responses, making it difficult to establish a clear optimization path. Overcoming these technical obstacles requires a deep understanding of stochastic dominance and its implications for reward distributions.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on paired preference data, which limits the applicability of existing methods in scenarios where such data is unavailable. The lack of effective techniques for handling unpaired data has been a significant barrier. Additionally, earlier approaches did not adequately address the distributional aspects of reward optimization, often leading to a focus on per-sample improvements rather than overall performance. Our approach, AOT, differs by emphasizing the need for stochastic dominance in reward distributions, thereby providing a more comprehensive framework for preference optimization.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology, Alignment via Optimal Transport (AOT), involves optimizing the reward distribution of chosen responses to ensure it stochastically dominates that of rejected responses. We will utilize both paired and unpaired datasets, employing metrics that assess the margin between the quantile plots of chosen and rejected rewards. The expected outcome is a significant improvement in the alignment of LLMs with human preferences"
            }
        },
        "author_data": {
            "93937824-8b4d-4df4-bb2a-7e9fe63fa64a": {
                "pk": "93937824-8b4d-4df4-bb2a-7e9fe63fa64a",
                "name": "Igor Melnyk",
                "collaborators": [
                    "Payel Das",
                    "Inkit Padhi",
                    "Arindam Banerjee",
                    "Vijil Chenthamarakshan",
                    "Pierre Dognin",
                    "Cicero Nogueira dos Santos",
                    "Youssef Mroueh",
                    "Pierre L. Dognin",
                    "Aurelie Lozano",
                    "Abhishek Kumar"
                ],
                "domain": [
                    "Machine Learning",
                    "Deep Learning",
                    "Natural Language Processing",
                    "Bioinformatics"
                ],
                "publications": [
                    {
                        "title": "A Spectral Algorithm for Inference in Hidden Semi-Markov Models",
                        "abstract": "Hidden semi-Markov models (HSMMs) are latent variable models which allow latent state persistence and can be viewed as a generalization of the popular hidden Markov models (HMMs). In this paper, we introduce a novel spectral algorithm to perform inference in HSMMs. Unlike expectation maximization (EM), our approach correctly estimates the probability of given observation sequence based on a set of training sequences. Our approach is based on estimating moments from the sample, whose number of dimensions depends only logarithmically on the maximum length of the hidden state persistence. Moreover, the algorithm requires only a few matrix inversions and is therefore computationally efficient. Empirical evaluations on synthetic and real data demonstrate the advantage of the algorithm over EM in terms of speed and accuracy, especially for large datasets."
                    },
                    {
                        "title": "Estimating Structured Vector Autoregressive Model",
                        "abstract": "While considerable advances have been made in estimating high-dimensional structured models from independent data using Lasso-type models, limited progress has been made for settings when the samples are dependent. We consider estimating structured VAR (vector auto-regressive models), where the structure can be captured by any suitable norm, e.g., Lasso, group Lasso, order weighted Lasso, sparse group Lasso, etc. In VAR setting with correlated noise, although there is strong dependence over time and covariates, we establish bounds on the non-asymptotic estimation error of structured VAR parameters. Surprisingly, the estimation error is of the same order as that of the corresponding Lasso-type estimator with independent samples, and the analysis holds for any norm. Our analysis relies on results in generic chaining, sub-exponential martingales, and spectral representation of VAR models. Experimental results on synthetic data with a variety of structures as well as real aviation data are presented, validating theoretical results."
                    },
                    {
                        "title": "Benchmarking deep generative models for diverse antibody sequence design",
                        "abstract": "Computational protein design, i.e. inferring novel and diverse protein sequences consistent with a given structure, remains a major unsolved challenge. Recently, deep generative models that learn from sequences alone or from sequences and structures jointly have shown impressive performance on this task. However, those models appear limited in terms of modeling structural constraints, capturing enough sequence diversity, or both. Here we consider three recently proposed deep generative frameworks for protein design: (AR) the sequence-based autoregressive generative model, (GVP) the precise structure-based graph neural network, and Fold2Seq that leverages a fuzzy and scale-free representation of a three-dimensional fold, while enforcing structure-to-sequence (and vice versa) consistency. We benchmark these models on the task of computational design of antibody sequences, which demand designing sequences with high diversity for functional implication. The Fold2Seq framework outperforms the two other baselines in terms of diversity of the designed sequences, while maintaining the typical fold."
                    },
                    {
                        "title": "AlphaFold Distillation for Protein Design",
                        "abstract": "Inverse protein folding, the process of designing sequences that fold into a specific 3D structure, is crucial in bio-engineering and drug discovery. Traditional methods rely on experimentally resolved structures, but these cover only a small fraction of protein sequences. Forward folding models like AlphaFold offer a potential solution by accurately predicting structures from sequences. However, these models are too slow for integration into the optimization loop of inverse folding models during training. To address this, we propose using knowledge distillation on folding model confidence metrics, such as pTM or pLDDT scores, to create a faster and end-to-end differentiable distilled model. This model can then be used as a structure consistency regularizer in training the inverse folding model. Our technique is versatile and can be applied to other design tasks, such as sequence-based protein infilling. Experimental results show that our method outperforms non-regularized baselines, yielding up to 3% improvement in sequence recovery and up to 45% improvement in protein diversity while maintaining structural consistency in generated sequences. Code is available at https://github.com/IBM/AFDistill"
                    },
                    {
                        "title": "Knowledge Graph Generation From Text",
                        "abstract": "In this work we propose a novel end-to-end multi-stage Knowledge Graph (KG) generation system from textual inputs, separating the overall process into two stages. The graph nodes are generated first using pretrained language model, followed by a simple edge construction head, enabling efficient KG extraction from the text. For each stage we consider several architectural choices that can be used depending on the available training resources. We evaluated the model on a recent WebNLG 2020 Challenge dataset, matching the state-of-the-art performance on text-to-RDF generation task, as well as on New York Times (NYT) and a large-scale TekGen datasets, showing strong overall performance, outperforming the existing baselines. We believe that the proposed system can serve as a viable KG construction alternative to the existing linearization or sampling-based graph generation approaches. Our code can be found at https://github.com/IBM/Grapher"
                    },
                    {
                        "title": "Improved Neural Text Attribute Transfer with Non-parallel Data",
                        "abstract": "Text attribute transfer using non-parallel data requires methods that can perform disentanglement of content and linguistic attributes. In this work, we propose multiple improvements over the existing approaches that enable the encoder-decoder framework to cope with the text attribute transfer from non-parallel data. We perform experiments on the sentiment transfer task using two datasets. For both datasets, our proposed method outperforms a strong baseline in two of the three employed evaluation metrics."
                    },
                    {
                        "title": "Deep learning algorithm for data-driven simulation of noisy dynamical system",
                        "abstract": "We present a deep learning model, DE-LSTM, for the simulation of a stochastic process with an underlying nonlinear dynamics. The deep learning model aims to approximate the probability density function of a stochastic process via numerical discretization and the underlying nonlinear dynamics is modeled by the Long Short-Term Memory (LSTM) network. It is shown that, when the numerical discretization is used, the function estimation problem can be solved by a multi-label classification problem. A penalized maximum log likelihood method is proposed to impose a smoothness condition in the prediction of the probability distribution. We show that the time evolution of the probability distribution can be computed by a high-dimensional integration of the transition probability of the LSTM internal states. A Monte Carlo algorithm to approximate the high-dimensional integration is outlined. The behavior of DE-LSTM is thoroughly investigated by using the Ornstein-Uhlenbeck process and noisy observations of nonlinear dynamical systems; Mackey-Glass time series and forced Van der Pol oscillator. It is shown that DE-LSTM makes a good prediction of the probability distribution without assuming any distributional properties of the stochastic process. For a multiple-step forecast of the Mackey-Glass time series, the prediction uncertainty, denoted by the 95\\% confidence interval, first grows, then dynamically adjusts following the evolution of the system, while in the simulation of the forced Van der Pol oscillator, the prediction uncertainty does not grow in time even for a 3,000-step forecast."
                    },
                    {
                        "title": "Semi-Markov Switching Vector Autoregressive Model-based Anomaly Detection in Aviation Systems",
                        "abstract": "In this work we consider the problem of anomaly detection in heterogeneous, multivariate, variable-length time series datasets. Our focus is on the aviation safety domain, where data objects are flights and time series are sensor readings and pilot switches. In this context the goal is to detect anomalous flight segments, due to mechanical, environmental, or human factors in order to identifying operationally significant events and provide insights into the flight operations and highlight otherwise unavailable potential safety risks and precursors to accidents. For this purpose, we propose a framework which represents each flight using a semi-Markov switching vector autoregressive (SMS-VAR) model. Detection of anomalies is then based on measuring dissimilarities between the model's prediction and data observation. The framework is scalable, due to the inherent parallel nature of most computations, and can be used to perform online anomaly detection. Extensive experimental results on simulated and real datasets illustrate that the framework can detect various types of anomalies along with the key parameters involved."
                    },
                    {
                        "title": "SenGen: Sentence Generating Neural Variational Topic Model",
                        "abstract": "We present a new topic model that generates documents by sampling a topic for one whole sentence at a time, and generating the words in the sentence using an RNN decoder that is conditioned on the topic of the sentence. We argue that this novel formalism will help us not only visualize and model the topical discourse structure in a document better, but also potentially lead to more interpretable topics since we can now illustrate topics by sampling representative sentences instead of bag of words or phrases. We present a variational auto-encoder approach for learning in which we use a factorized variational encoder that independently models the posterior over topical mixture vectors of documents using a feed-forward network, and the posterior over topic assignments to sentences using an RNN. Our preliminary experiments on two different datasets indicate early promise, but also expose many challenges that remain to be addressed."
                    },
                    {
                        "title": "Fighting Offensive Language on Social Media with Unsupervised Text Style Transfer",
                        "abstract": "We introduce a new approach to tackle the problem of offensive language in online social media. Our approach uses unsupervised text style transfer to translate offensive sentences into non-offensive ones. We propose a new method for training encoder-decoders using non-parallel data that combines a collaborative classifier, attention and the cycle consistency loss. Experimental results on data from Twitter and Reddit show that our method outperforms a state-of-the-art text style transfer system in two out of three quantitative metrics and produces reliable non-offensive transferred sentences."
                    },
                    {
                        "title": "R2N2: Residual Recurrent Neural Networks for Multivariate Time Series Forecasting",
                        "abstract": "Multivariate time-series modeling and forecasting is an important problem with numerous applications. Traditional approaches such as VAR (vector auto-regressive) models and more recent approaches such as RNNs (recurrent neural networks) are indispensable tools in modeling time-series data. In many multivariate time series modeling problems, there is usually a significant linear dependency component, for which VARs are suitable, and a nonlinear component, for which RNNs are suitable. Modeling such times series with only VAR or only RNNs can lead to poor predictive performance or complex models with large training times. In this work, we propose a hybrid model called R2N2 (Residual RNN), which first models the time series with a simple linear model (like VAR) and then models its residual errors using RNNs. R2N2s can be trained using existing algorithms for VARs and RNNs. Through an extensive empirical evaluation on two real world datasets (aviation and climate domains), we show that R2N2 is competitive, usually better than VAR or RNN, used alone. We also show that R2N2 is faster to train as compared to an RNN, while requiring less number of hidden units."
                    },
                    {
                        "title": "Wasserstein Barycenter Model Ensembling",
                        "abstract": "In this paper we propose to perform model ensembling in a multiclass or a multilabel learning setting using Wasserstein (W.) barycenters. Optimal transport metrics, such as the Wasserstein distance, allow incorporating semantic side information such as word embeddings. Using W. barycenters to find the consensus between models allows us to balance confidence and semantics in finding the agreement between the models. We show applications of Wasserstein ensembling in attribute-based classification, multilabel learning and image captioning generation. These results show that the W. ensembling is a viable alternative to the basic geometric or arithmetic mean ensembling."
                    },
                    {
                        "title": "Adversarial Semantic Alignment for Improved Image Captions",
                        "abstract": "In this paper we study image captioning as a conditional GAN training, proposing both a context-aware LSTM captioner and co-attentive discriminator, which enforces semantic alignment between images and captions. We empirically focus on the viability of two training methods: Self-critical Sequence Training (SCST) and Gumbel Straight-Through (ST) and demonstrate that SCST shows more stable gradient behavior and improved results over Gumbel ST, even without accessing discriminator gradients directly. We also address the problem of automatic evaluation for captioning models and introduce a new semantic score, and show its correlation to human judgement. As an evaluation paradigm, we argue that an important criterion for a captioner is the ability to generalize to compositions of objects that do not usually co-occur together. To this end, we introduce a small captioned Out of Context (OOC) test set. The OOC set, combined with our semantic score, are the proposed new diagnosis tools for the captioning community. When evaluated on OOC and MS-COCO benchmarks, we show that SCST-based training has a strong performance in both semantic score and human evaluation, promising to be a valuable new approach for efficient discrete GAN training."
                    },
                    {
                        "title": "DualTKB: A Dual Learning Bridge between Text and Knowledge Base",
                        "abstract": "In this work, we present a dual learning approach for unsupervised text to path and path to text transfers in Commonsense Knowledge Bases (KBs). We investigate the impact of weak supervision by creating a weakly supervised dataset and show that even a slight amount of supervision can significantly improve the model performance and enable better-quality transfers. We examine different model architectures, and evaluation metrics, proposing a novel Commonsense KB completion metric tailored for generative models. Extensive experimental results show that the proposed method compares very favorably to the existing baselines. This approach is a viable step towards a more advanced system for automatic KB construction/expansion and the reverse operation of KB conversion to coherent textual descriptions."
                    },
                    {
                        "title": "ReGen: Reinforcement Learning for Text and Knowledge Base Generation using Pretrained Language Models",
                        "abstract": "Automatic construction of relevant Knowledge Bases (KBs) from text, and generation of semantically meaningful text from KBs are both long-standing goals in Machine Learning. In this paper, we present ReGen, a bidirectional generation of text and graph leveraging Reinforcement Learning (RL) to improve performance. Graph linearization enables us to re-frame both tasks as a sequence to sequence generation problem regardless of the generative direction, which in turn allows the use of Reinforcement Learning for sequence training where the model itself is employed as its own critic leading to Self-Critical Sequence Training (SCST). We present an extensive investigation demonstrating that the use of RL via SCST benefits graph and text generation on WebNLG+ 2020 and TekGen datasets. Our system provides state-of-the-art results on WebNLG+ 2020 by significantly improving upon published results from the WebNLG 2020+ Challenge for both text-to-graph and graph-to-text generation tasks."
                    },
                    {
                        "title": "Fold2Seq: A Joint Sequence(1D)-Fold(3D) Embedding-based Generative Model for Protein Design",
                        "abstract": "Designing novel protein sequences for a desired 3D topological fold is a fundamental yet non-trivial task in protein engineering. Challenges exist due to the complex sequence--fold relationship, as well as the difficulties to capture the diversity of the sequences (therefore structures and functions) within a fold. To overcome these challenges, we propose Fold2Seq, a novel transformer-based generative framework for designing protein sequences conditioned on a specific target fold. To model the complex sequence--structure relationship, Fold2Seq jointly learns a sequence embedding using a transformer and a fold embedding from the density of secondary structural elements in 3D voxels. On test sets with single, high-resolution and complete structure inputs for individual folds, our experiments demonstrate improved or comparable performance of Fold2Seq in terms of speed, coverage, and reliability for sequence design, when compared to existing state-of-the-art methods that include data-driven deep generative models and physics-based RosettaDesign. The unique advantages of fold-based Fold2Seq, in comparison to a structure-based deep model and RosettaDesign, become more evident on three additional real-world challenges originating from low-quality, incomplete, or ambiguous input structures. Source code and data are available at https://github.com/IBM/fold2seq."
                    },
                    {
                        "title": "Reprogramming Pretrained Language Models for Antibody Sequence Infilling",
                        "abstract": "Antibodies comprise the most versatile class of binding molecules, with numerous applications in biomedicine. Computational design of antibodies involves generating novel and diverse sequences, while maintaining structural consistency. Unique to antibodies, designing the complementarity-determining region (CDR), which determines the antigen binding affinity and specificity, creates its own unique challenges. Recent deep learning models have shown impressive results, however the limited number of known antibody sequence/structure pairs frequently leads to degraded performance, particularly lacking diversity in the generated sequences. In our work we address this challenge by leveraging Model Reprogramming (MR), which repurposes pretrained models on a source language to adapt to the tasks that are in a different language and have scarce data - where it may be difficult to train a high-performing model from scratch or effectively fine-tune an existing pre-trained model on the specific task. Specifically, we introduce ReprogBert in which a pretrained English language model is repurposed for protein sequence infilling - thus considers cross-language adaptation using less data. Results on antibody design benchmarks show that our model on low-resourced antibody sequence dataset provides highly diverse CDR sequences, up to more than a two-fold increase of diversity over the baselines, without losing structural integrity and naturalness. The generated sequences also demonstrate enhanced antigen binding specificity and virus neutralization ability. Code is available at https://github.com/IBM/ReprogBERT"
                    },
                    {
                        "title": "Alleviating Noisy Data in Image Captioning with Cooperative Distillation",
                        "abstract": "Image captioning systems have made substantial progress, largely due to the availability of curated datasets like Microsoft COCO or Vizwiz that have accurate descriptions of their corresponding images. Unfortunately, scarce availability of such cleanly labeled data results in trained algorithms producing captions that can be terse and idiosyncratically specific to details in the image. We propose a new technique, cooperative distillation that combines clean curated datasets with the web-scale automatically extracted captions of the Google Conceptual Captions dataset (GCC), which can have poor descriptions of images, but is abundant in size and therefore provides a rich vocabulary resulting in more expressive captions."
                    }
                ]
            },
            "b8cb1cf2-d8fb-4dd8-98bd-56bf164e680c": {
                "pk": "b8cb1cf2-d8fb-4dd8-98bd-56bf164e680c",
                "name": "Youssef Mroueh",
                "collaborators": [
                    "Vaibhava Goel",
                    "Lorenzo Rosasco",
                    "Tom Sercu",
                    "Carles Domingo-Enrich",
                    "Etienne Marcheret",
                    "Tomaso Poggio",
                    "Truyen Nguyen",
                    "Mattia Rigotti",
                    "Jean-Jacques Slotine",
                    "Stephen Voinea"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Optimal Transport",
                    "Machine Learning",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Regularized Finite Dimensional Kernel Sobolev Discrepancy",
                        "abstract": "We show in this note that the Sobolev Discrepancy introduced in Mroueh et al in the context of generative adversarial networks, is actually the weighted negative Sobolev norm $||.||_{\\dot{H}^{-1}(\\nu_q)}$, that is known to linearize the Wasserstein $W_2$ distance and plays a fundamental role in the dynamic formulation of optimal transport of Benamou and Brenier. Given a Kernel with finite dimensional feature map we show that the Sobolev discrepancy can be approximated from finite samples. Assuming this discrepancy is finite, the error depends on the approximation error in the function space induced by the finite dimensional feature space kernel and on a statistical error due to the finite sample approximation."
                    },
                    {
                        "title": "Robust Phase Retrieval and Super-Resolution from One Bit Coded Diffraction Patterns",
                        "abstract": "In this paper we study a realistic setup for phase retrieval, where the signal of interest is modulated or masked and then for each modulation or mask a diffraction pattern is collected, producing a coded diffraction pattern (CDP) [CLM13]. We are interested in the setup where the resolution of the collected CDP is limited by the Fraunhofer diffraction limit of the imaging system.   We investigate a novel approach based on a geometric quantization scheme of phase-less linear measurements into (one-bit) coded diffraction patterns, and a corresponding recovery scheme. The key novelty in this approach consists in comparing pairs of coded diffractions patterns across frequencies: the one bit measurements obtained rely on the order statistics of the un-quantized measurements rather than their values . This results in a robust phase recovery, and unlike currently available methods, allows to efficiently perform phase recovery from measurements affected by severe (possibly unknown) non linear, rank preserving perturbations, such as distortions. Another important feature of this approach consists in the fact that it enables also super-resolution and blind-deconvolution, beyond the diffraction limit of a given imaging system."
                    },
                    {
                        "title": "Wasserstein Style Transfer",
                        "abstract": "We propose Gaussian optimal transport for Image style transfer in an Encoder/Decoder framework. Optimal transport for Gaussian measures has closed forms Monge mappings from source to target distributions. Moreover interpolates between a content and a style image can be seen as geodesics in the Wasserstein Geometry. Using this insight, we show how to mix different target styles , using Wasserstein barycenter of Gaussian measures. Since Gaussians are closed under Wasserstein barycenter, this allows us a simple style transfer and style mixing and interpolation. Moreover we show how mixing different styles can be achieved using other geodesic metrics between gaussians such as the Fisher Rao metric, while the transport of the content to the new interpolate style is still performed with Gaussian OT maps. Our simple methodology allows to generate new stylized content interpolating between many artistic styles. The metric used in the interpolation results in different stylizations."
                    },
                    {
                        "title": "Information Theoretic Guarantees For Policy Alignment In Large Language Models",
                        "abstract": "Policy alignment of large language models refers to constrained policy optimization, where the policy is optimized to maximize a reward while staying close to a reference policy with respect to an $f$-divergence such as the $\\mathsf{KL}$ divergence. The best of $n$ alignment policy selects a sample from the reference policy that has the maximum reward among $n$ independent samples. For both cases (policy alignment and best of $n$), recent works showed empirically that the reward improvement of the aligned policy on the reference one scales like $\\sqrt{\\mathsf{KL}}$, with an explicit bound in $n$ on the $\\mathsf{KL}$ for the best of $n$ policy. We show in this paper that the $\\sqrt{\\mathsf{KL}}$ information theoretic upper bound holds if the reward under the reference policy has sub-gaussian tails. Moreover, we prove for the best of $n$ policy, that the $\\mathsf{KL}$ upper bound can be obtained for any $f$-divergence via a reduction to exponential order statistics owing to the R\\'enyi representation of order statistics, and a data processing inequality. If additional information is known on the tails of the aligned policy we show that tighter control on the reward improvement can be obtained via the R\\'enyi divergence. Finally we demonstrate how these upper bounds transfer from proxy rewards to golden rewards which results in a decrease in the golden reward improvement due to overestimation and approximation errors of the proxy reward."
                    },
                    {
                        "title": "q-ary Compressive Sensing",
                        "abstract": "We introduce q-ary compressive sensing, an extension of 1-bit compressive sensing. We propose a novel sensing mechanism and a corresponding recovery procedure. The recovery properties of the proposed approach are analyzed both theoretically and empirically. Results in 1-bit compressive sensing are recovered as a special case. Our theoretical results suggest a tradeoff between the quantization parameter q, and the number of measurements m in the control of the error of the resulting recovery algorithm, as well its robustness to noise."
                    },
                    {
                        "title": "Quantization and Greed are Good: One bit Phase Retrieval, Robustness and Greedy Refinements",
                        "abstract": "In this paper, we study the problem of robust phase recovery. We investigate a novel approach based on extremely quantized (one-bit) phase-less measurements and a corresponding recovery scheme. The proposed approach has surprising robustness and stability properties and, unlike currently available methods, allows to efficiently perform phase recovery from measurements affected by severe (possibly unknown) non-linear perturbations, such as distortions (e.g. clipping). Beyond robustness, we show how our approach can be used within greedy approaches based on alternating minimization. In particular, we propose novel initialization schemes for the alternating minimization achieving favorable convergence properties with improved sample complexity."
                    },
                    {
                        "title": "On the Convergence of Gradient Descent in GANs: MMD GAN As a Gradient Flow",
                        "abstract": "We consider the maximum mean discrepancy ($\\mathrm{MMD}$) GAN problem and propose a parametric kernelized gradient flow that mimics the min-max game in gradient regularized $\\mathrm{MMD}$ GAN. We show that this flow provides a descent direction minimizing the $\\mathrm{MMD}$ on a statistical manifold of probability distributions. We then derive an explicit condition which ensures that gradient descent on the parameter space of the generator in gradient regularized $\\mathrm{MMD}$ GAN is globally convergent to the target distribution. Under this condition, we give non asymptotic convergence results of gradient descent in MMD GAN. Another contribution of this paper is the introduction of a dynamic formulation of a regularization of $\\mathrm{MMD}$ and demonstrating that the parametric kernelized descent for $\\mathrm{MMD}$ is the gradient flow of this functional with respect to the new Riemannian structure. Our obtained theoretical result allows ones to treat gradient flows for quite general functionals and thus has potential applications to other types of variational inferences on a statistical manifold beyond GANs. Finally, numerical experiments suggest that our parametric kernelized gradient flow stabilizes GAN training and guarantees convergence."
                    },
                    {
                        "title": "Semi-Supervised Learning with IPM-based GANs: an Empirical Study",
                        "abstract": "We present an empirical investigation of a recent class of Generative Adversarial Networks (GANs) using Integral Probability Metrics (IPM) and their performance for semi-supervised learning. IPM-based GANs like Wasserstein GAN, Fisher GAN and Sobolev GAN have desirable properties in terms of theoretical understanding, training stability, and a meaningful loss. In this work we investigate how the design of the critic (or discriminator) influences the performance in semi-supervised learning. We distill three key take-aways which are important for good SSL performance: (1) the K+1 formulation, (2) avoiding batch normalization in the critic and (3) avoiding gradient penalty constraints on the classification layer."
                    },
                    {
                        "title": "Auditing Differential Privacy in High Dimensions with the Kernel Quantum R\u00e9nyi Divergence",
                        "abstract": "Differential privacy (DP) is the de facto standard for private data release and private machine learning. Auditing black-box DP algorithms and mechanisms to certify whether they satisfy a certain DP guarantee is challenging, especially in high dimension. We propose relaxations of differential privacy based on new divergences on probability distributions: the kernel R\\'enyi divergence and its regularized version. We show that the regularized kernel R\\'enyi divergence can be estimated from samples even in high dimensions, giving rise to auditing procedures for $\\varepsilon$-DP, $(\\varepsilon,\\delta)$-DP and $(\\alpha,\\varepsilon)$-R\\'enyi DP."
                    },
                    {
                        "title": "Unbalanced Sobolev Descent",
                        "abstract": "We introduce Unbalanced Sobolev Descent (USD), a particle descent algorithm for transporting a high dimensional source distribution to a target distribution that does not necessarily have the same mass. We define the Sobolev-Fisher discrepancy between distributions and show that it relates to advection-reaction transport equations and the Wasserstein-Fisher-Rao metric between distributions. USD transports particles along gradient flows of the witness function of the Sobolev-Fisher discrepancy (advection step) and reweighs the mass of particles with respect to this witness function (reaction step). The reaction step can be thought of as a birth-death process of the particles with rate of growth proportional to the witness function. When the Sobolev-Fisher witness function is estimated in a Reproducing Kernel Hilbert Space (RKHS), under mild assumptions we show that USD converges asymptotically (in the limit of infinite particles) to the target distribution in the Maximum Mean Discrepancy (MMD) sense. We then give two methods to estimate the Sobolev-Fisher witness with neural networks, resulting in two Neural USD algorithms. The first one implements the reaction step with mirror descent on the weights, while the second implements it through a birth-death process of particles. We show on synthetic examples that USD transports distributions with or without conservation of mass faster than previous particle descent algorithms, and finally demonstrate its use for molecular biology analyses where our method is naturally suited to match developmental stages of populations of differentiating cells based on their single-cell RNA sequencing profile. Code is available at https://github.com/ibm/usd ."
                    },
                    {
                        "title": "Co-Occuring Directions Sketching for Approximate Matrix Multiply",
                        "abstract": "We introduce co-occurring directions sketching, a deterministic algorithm for approximate matrix product (AMM), in the streaming model. We show that co-occuring directions achieves a better error bound for AMM than other randomized and deterministic approaches for AMM. Co-occurring directions gives a $1 + \\epsilon$ -approximation of the optimal low rank approximation of a matrix product. Empirically our algorithm outperforms competing methods for AMM, for a small sketch size. We validate empirically our theoretical findings and algorithms"
                    },
                    {
                        "title": "Fisher GAN",
                        "abstract": "Generative Adversarial Networks (GANs) are powerful models for learning complex distributions. Stable training of GANs has been addressed in many recent works which explore different metrics between distributions. In this paper we introduce Fisher GAN which fits within the Integral Probability Metrics (IPM) framework for training GANs. Fisher GAN defines a critic with a data dependent constraint on its second order moments. We show in this paper that Fisher GAN allows for stable and time efficient training that does not compromise the capacity of the critic, and does not need data independent constraints such as weight clipping. We analyze our Fisher IPM theoretically and provide an algorithm based on Augmented Lagrangian for Fisher GAN. We validate our claims on both image sample generation and semi-supervised classification using Fisher GAN."
                    },
                    {
                        "title": "Separation Results between Fixed-Kernel and Feature-Learning Probability Metrics",
                        "abstract": "Several works in implicit and explicit generative modeling empirically observed that feature-learning discriminators outperform fixed-kernel discriminators in terms of the sample quality of the models. We provide separation results between probability metrics with fixed-kernel and feature-learning discriminators using the function classes $\\mathcal{F}_2$ and $\\mathcal{F}_1$ respectively, which were developed to study overparametrized two-layer neural networks. In particular, we construct pairs of distributions over hyper-spheres that can not be discriminated by fixed kernel $(\\mathcal{F}_2)$ integral probability metric (IPM) and Stein discrepancy (SD) in high dimensions, but that can be discriminated by their feature learning ($\\mathcal{F}_1$) counterparts. To further study the separation we provide links between the $\\mathcal{F}_1$ and $\\mathcal{F}_2$ IPMs with sliced Wasserstein distances. Our work suggests that fixed-kernel discriminators perform worse than their feature learning counterparts because their corresponding metrics are weaker."
                    },
                    {
                        "title": "McGan: Mean and Covariance Feature Matching GAN",
                        "abstract": "We introduce new families of Integral Probability Metrics (IPM) for training Generative Adversarial Networks (GAN). Our IPMs are based on matching statistics of distributions embedded in a finite dimensional feature space. Mean and covariance feature matching IPMs allow for stable training of GANs, which we will call McGan. McGan minimizes a meaningful loss between distributions."
                    },
                    {
                        "title": "Multiclass Learning with Simplex Coding",
                        "abstract": "In this paper we discuss a novel framework for multiclass learning, defined by a suitable coding/decoding strategy, namely the simplex coding, that allows to generalize to multiple classes a relaxation approach commonly used in binary classification. In this framework, a relaxation error analysis can be developed avoiding constraints on the considered hypotheses class. Moreover, we show that in this setting it is possible to derive the first provably consistent regularized method with training/tuning complexity which is independent to the number of classes. Tools from convex analysis are introduced that can be used beyond the scope of this paper."
                    },
                    {
                        "title": "Tighter Sparse Approximation Bounds for ReLU Neural Networks",
                        "abstract": "A well-known line of work (Barron, 1993; Breiman, 1993; Klusowski & Barron, 2018) provides bounds on the width $n$ of a ReLU two-layer neural network needed to approximate a function $f$ over the ball $\\mathcal{B}_R(\\mathbb{R}^d)$ up to error $\\epsilon$, when the Fourier based quantity $C_f = \\frac{1}{(2\\pi)^{d/2}} \\int_{\\mathbb{R}^d} \\|\\xi\\|^2 |\\hat{f}(\\xi)| \\ d\\xi$ is finite. More recently Ongie et al. (2019) used the Radon transform as a tool for analysis of infinite-width ReLU two-layer networks. In particular, they introduce the concept of Radon-based $\\mathcal{R}$-norms and show that a function defined on $\\mathbb{R}^d$ can be represented as an infinite-width two-layer neural network if and only if its $\\mathcal{R}$-norm is finite. In this work, we extend the framework of Ongie et al. (2019) and define similar Radon-based semi-norms ($\\mathcal{R}, \\mathcal{U}$-norms) such that a function admits an infinite-width neural network representation on a bounded open set $\\mathcal{U} \\subseteq \\mathbb{R}^d$ when its $\\mathcal{R}, \\mathcal{U}$-norm is finite. Building on this, we derive sparse (finite-width) neural network approximation bounds that refine those of Breiman (1993); Klusowski & Barron (2018). Finally, we show that infinite-width neural network representations on bounded open sets are not unique and study their structure, providing a functional view of mode connectivity."
                    },
                    {
                        "title": "Learning with Group Invariant Features: A Kernel Perspective",
                        "abstract": "We analyze in this paper a random feature map based on a theory of invariance I-theory introduced recently. More specifically, a group invariant signal signature is obtained through cumulative distributions of group transformed random projections. Our analysis bridges invariant feature learning with kernel methods, as we show that this feature map defines an expected Haar integration kernel that is invariant to the specified group action. We show how this non-linear random feature map approximates this group invariant kernel uniformly on a set of $N$ points. Moreover, we show that it defines a function space that is dense in the equivalent Invariant Reproducing Kernel Hilbert Space. Finally, we quantify error rates of the convergence of the empirical risk minimization, as well as the reduction in the sample complexity of a learning algorithm using such an invariant representation for signal classification, in a classical supervised learning setting."
                    },
                    {
                        "title": "Random Maxout Features",
                        "abstract": "In this paper, we propose and study random maxout features, which are constructed by first projecting the input data onto sets of randomly generated vectors with Gaussian elements, and then outputing the maximum projection value for each set. We show that the resulting random feature map, when used in conjunction with linear models, allows for the locally linear estimation of the function of interest in classification tasks, and for the locally linear embedding of points when used for dimensionality reduction or data visualization. We derive generalization bounds for learning that assess the error in approximating locally linear functions by linear functions in the maxout feature space, and empirically evaluate the efficacy of the approach on the MNIST and TIMIT classification tasks."
                    },
                    {
                        "title": "Asymmetrically Weighted CCA And Hierarchical Kernel Sentence Embedding For Image & Text Retrieval",
                        "abstract": "Joint modeling of language and vision has been drawing increasing interest. A multimodal data representation allowing for bidirectional retrieval of images by sentences and vice versa is a key aspect. In this paper we present three contributions in canonical correlation analysis (CCA) based multimodal retrieval. Firstly, we show that an asymmetric weighting of the canonical weights, while achieving a cross view mapping from the search to the query space, improves the retrieval performance. Secondly, we devise a computationally efficient model selection, crucial to generalization and stability, in the framework of the Bj\\\"ork Golub algorithm for regularized CCA via spectral filtering. Finally, we introduce a Hierarchical Kernel Sentence Embedding (HKSE) that approximates Kernel CCA for a special similarity kernel between distribution of words embedded in a vector space. State of the art results are obtained on MSCOCO and Flickr benchmarks when these three techniques are used in conjunction."
                    },
                    {
                        "title": "Deep Multimodal Learning for Audio-Visual Speech Recognition",
                        "abstract": "In this paper, we present methods in deep multimodal learning for fusing speech and visual modalities for Audio-Visual Automatic Speech Recognition (AV-ASR). First, we study an approach where uni-modal deep networks are trained separately and their final hidden layers fused to obtain a joint feature space in which another deep network is built. While the audio network alone achieves a phone error rate (PER) of $41\\%$ under clean condition on the IBM large vocabulary audio-visual studio dataset, this fusion model achieves a PER of $35.83\\%$ demonstrating the tremendous value of the visual channel in phone classification even in audio with high signal to noise ratio. Second, we present a new deep network architecture that uses a bilinear softmax layer to account for class specific correlations between modalities. We show that combining the posteriors from the bilinear networks with those from the fused model mentioned above results in a further significant phone error rate reduction, yielding a final PER of $34.03\\%$."
                    }
                ]
            },
            "db978d7d-ca65-4908-ae13-1e3f7750ae31": {
                "pk": "db978d7d-ca65-4908-ae13-1e3f7750ae31",
                "name": "Brian Belgodere",
                "collaborators": [
                    "Youssef Mroueh",
                    "Jerret Ross",
                    "Inkit Padhi",
                    "Igor Melnyk",
                    "Mattia Rigotti",
                    "Vijil Chenthamarakshan",
                    "Payel Das",
                    "Pierre Dognin",
                    "Yair Schiff",
                    "Richard A. Young"
                ],
                "domain": [
                    "Transfer Learning",
                    "Molecular Modeling",
                    "Generative AI",
                    "Synthetic Data"
                ],
                "publications": [
                    {
                        "title": "G2L: A Geometric Approach for Generating Pseudo-labels that Improve Transfer Learning",
                        "abstract": "Transfer learning is a deep-learning technique that ameliorates the problem of learning when human-annotated labels are expensive and limited. In place of such labels, it uses instead the previously trained weights from a well-chosen source model as the initial weights for the training of a base model for a new target dataset. We demonstrate a novel but general technique for automatically creating such source models. We generate pseudo-labels according to an efficient and extensible algorithm that is based on a classical result from the geometry of high dimensions, the Cayley-Menger determinant. This G2L (``geometry to label'') method incrementally builds up pseudo-labels using a greedy computation of hypervolume content. We demonstrate that the method is tunable with respect to expected accuracy, which can be forecast by an information-theoretic measure of dataset similarity (divergence) between source and target. The results of 280 experiments show that this mechanical technique generates base models that have similar or better transferability compared to a baseline of models trained on extensively human-annotated ImageNet1K labels, yielding an overall error decrease of 0.43\\%, and an error decrease in 4 out of 5 divergent datasets tested."
                    },
                    {
                        "title": "Large-Scale Chemical Language Representations Capture Molecular Structure and Properties",
                        "abstract": "Models based on machine learning can enable accurate and fast molecular property predictions, which is of interest in drug discovery and material design. Various supervised machine learning models have demonstrated promising performance, but the vast chemical space and the limited availability of property labels make supervised learning challenging. Recently, unsupervised transformer-based language models pretrained on a large unlabelled corpus have produced state-of-the-art results in many downstream natural language processing tasks. Inspired by this development, we present molecular embeddings obtained by training an efficient transformer encoder model, MoLFormer, which uses rotary positional embeddings. This model employs a linear attention mechanism, coupled with highly distributed training, on SMILES sequences of 1.1 billion unlabelled molecules from the PubChem and ZINC datasets. We show that the learned molecular representation outperforms existing baselines, including supervised and self-supervised graph neural networks and language models, on several downstream tasks from ten benchmark datasets. They perform competitively on two others. Further analyses, specifically through the lens of attention, demonstrate that MoLFormer trained on chemical SMILES indeed learns the spatial relationships between atoms within a molecule. These results provide encouraging evidence that large-scale molecular language models can capture sufficient chemical and structural information to predict various distinct molecular properties, including quantum-chemical properties."
                    },
                    {
                        "title": "Cloud-Based Real-Time Molecular Screening Platform with MolFormer",
                        "abstract": "With the prospect of automating a number of chemical tasks with high fidelity, chemical language processing models are emerging at a rapid speed. Here, we present a cloud-based real-time platform that allows users to virtually screen molecules of interest. For this purpose, molecular embeddings inferred from a recently proposed large chemical language model, named MolFormer, are leveraged. The platform currently supports three tasks: nearest neighbor retrieval, chemical space visualization, and property prediction. Based on the functionalities of this platform and results obtained, we believe that such a platform can play a pivotal role in automating chemistry and chemical engineering research, as well as assist in drug discovery and material design tasks. A demo of our platform is provided at \\url{www.ibm.biz/molecular_demo}."
                    },
                    {
                        "title": "Image Captioning as an Assistive Technology: Lessons Learned from VizWiz 2020 Challenge",
                        "abstract": "Image captioning has recently demonstrated impressive progress largely owing to the introduction of neural network algorithms trained on curated dataset like MS-COCO. Often work in this field is motivated by the promise of deployment of captioning systems in practical applications. However, the scarcity of data and contexts in many competition datasets renders the utility of systems trained on these datasets limited as an assistive technology in real-world settings, such as helping visually impaired people navigate and accomplish everyday tasks. This gap motivated the introduction of the novel VizWiz dataset, which consists of images taken by the visually impaired and captions that have useful, task-oriented information. In an attempt to help the machine learning computer vision field realize its promise of producing technologies that have positive social impact, the curators of the VizWiz dataset host several competitions, including one for image captioning. This work details the theory and engineering from our winning submission to the 2020 captioning competition. Our work provides a step towards improved assistive image captioning systems."
                    },
                    {
                        "title": "Risk Aware Benchmarking of Large Language Models",
                        "abstract": "We propose a distributional framework for benchmarking socio-technical risks of foundation models with quantified statistical significance. Our approach hinges on a new statistical relative testing based on first and second order stochastic dominance of real random variables. We show that the second order statistics in this test are linked to mean-risk models commonly used in econometrics and mathematical finance to balance risk and utility when choosing between alternatives. Using this framework, we formally develop a risk-aware approach for foundation model selection given guardrails quantified by specified metrics. Inspired by portfolio optimization and selection theory in mathematical finance, we define a metrics portfolio for each model as a means to aggregate a collection of metrics, and perform model selection based on the stochastic dominance of these portfolios. The statistical significance of our tests is backed theoretically by an asymptotic analysis via central limit theorems instantiated in practice via a bootstrap variance estimate. We use our framework to compare various large language models regarding risks related to drifting from instructions and outputting toxic content."
                    },
                    {
                        "title": "GP-MoLFormer: A Foundation Model For Molecular Generation",
                        "abstract": "Transformer-based models trained on large and general purpose datasets consisting of molecular strings have recently emerged as a powerful tool for successfully modeling various structure-property relations. Inspired by this success, we extend the paradigm of training chemical language transformers on large-scale chemical datasets to generative tasks in this work. Specifically, we propose GP-MoLFormer, an autoregressive molecular string generator that is trained on more than 1.1B chemical SMILES. GP-MoLFormer uses a 46.8M parameter transformer decoder model with linear attention and rotary positional encodings as the base architecture. We explore the utility of GP-MoLFormer in generating novel, valid, and unique SMILES. Impressively, we find GP-MoLFormer is able to generate a significant fraction of novel, valid, and unique SMILES even when the number of generated molecules is in the 10 billion range and the reference set is over a billion. We also find strong memorization of training data in GP-MoLFormer generations, which has so far remained unexplored for chemical language models. Our analyses reveal that training data memorization and novelty in generations are impacted by the quality of the training data; duplication bias in training data can enhance memorization at the cost of lowering novelty. We evaluate GP-MoLFormer's utility and compare it with that of existing baselines on three different tasks: de novo generation, scaffold-constrained molecular decoration, and unconstrained property-guided optimization. While the first two are handled with no additional training, we propose a parameter-efficient fine-tuning method for the last task, which uses property-ordered molecular pairs as input. We call this new approach pair-tuning. Our results show GP-MoLFormer performs better or comparable with baselines across all three tasks, demonstrating its general utility."
                    },
                    {
                        "title": "P2L: Predicting Transfer Learning for Images and Semantic Relations",
                        "abstract": "Transfer learning enhances learning across tasks, by leveraging previously learned representations -- if they are properly chosen. We describe an efficient method to accurately estimate the appropriateness of a previously trained model for use in a new learning task. We use this measure, which we call \"Predict To Learn\" (\"P2L\"), in the two very different domains of images and semantic relations, where it predicts, from a set of \"source\" models, the one model most likely to produce effective transfer for training a given \"target\" model. We validate our approach thoroughly, by assembling a collection of candidate source models, then fine-tuning each candidate to perform each of a collection of target tasks, and finally measuring how well transfer has been enhanced. Across 95 tasks within multiple domains (images classification and semantic relations), the P2L approach was able to select the best transfer learning model on average, while the heuristic of choosing model trained with the largest data set selected the best model in only 55 cases. These results suggest that P2L captures important information in common between source and target tasks, and that this shared informational structure contributes to successful transfer learning more than simple data size."
                    },
                    {
                        "title": "Auditing and Generating Synthetic Data with Controllable Trust Trade-offs",
                        "abstract": "Real-world data often exhibits bias, imbalance, and privacy risks. Synthetic datasets have emerged to address these issues. This paradigm relies on generative AI models to generate unbiased, privacy-preserving data while maintaining fidelity to the original data. However, assessing the trustworthiness of synthetic datasets and models is a critical challenge. We introduce a holistic auditing framework that comprehensively evaluates synthetic datasets and AI models. It focuses on preventing bias and discrimination, ensures fidelity to the source data, assesses utility, robustness, and privacy preservation. We demonstrate the framework's effectiveness by auditing various generative models across diverse use cases like education, healthcare, banking, and human resources, spanning different data modalities such as tabular, time-series, vision, and natural language. This holistic assessment is essential for compliance with regulatory safeguards. We introduce a trustworthiness index to rank synthetic datasets based on their safeguards trade-offs. Furthermore, we present a trustworthiness-driven model selection and cross-validation process during training, exemplified with \"TrustFormers\" across various data types. This approach allows for controllable trustworthiness trade-offs in synthetic data creation. Our auditing framework fosters collaboration among stakeholders, including data scientists, governance experts, internal reviewers, external certifiers, and regulators. This transparent reporting should become a standard practice to prevent bias, discrimination, and privacy violations, ensuring compliance with policies and providing accountability, safety, and performance guarantees."
                    }
                ]
            },
            "4bafebfa-db5c-4d7e-ba32-4ff50127b87c": {
                "pk": "4bafebfa-db5c-4d7e-ba32-4ff50127b87c",
                "name": "Mattia Rigotti",
                "collaborators": [
                    "Youssef Mroueh",
                    "Inkit Padhi",
                    "Pierre Dognin",
                    "Igor Melnyk",
                    "Yair Schiff",
                    "Cristiano Malossi",
                    "Jarret Ross",
                    "Stefano Fusi",
                    "Rong Zhu",
                    "Ioana Giurgiu"
                ],
                "domain": [
                    "Neuroscience",
                    "Reinforcement Learning",
                    "Deep Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Estimating the dimensionality of neural responses with fMRI Repetition Suppression",
                        "abstract": "We propose a novel method that exploits fMRI Repetition Suppression (RS-fMRI) to measure the dimensionality of the set response vectors, i.e. the dimension of the space of linear combinations of neural population activity patterns in response to specific task conditions. RS-fMRI measures the overlap between response vectors even in brain areas displaying no discernible average differential BOLD signal. We show how this property can be used to estimate the neural response dimensionality in areas lacking macroscopic spatial patterning. The importance of dimensionality derives from how it relates to a neural circuit's functionality. As we show, the dimensionality of the response vectors is predicted to be high in areas involved in multi-stream integration, while it is low in areas where inputs from independent sources do not interact or merely overlap linearly. Our method can be used to identify and functionally characterize cortical circuits that integrate multiple independent information pathways."
                    },
                    {
                        "title": "Self-correcting Q-Learning",
                        "abstract": "The Q-learning algorithm is known to be affected by the maximization bias, i.e. the systematic overestimation of action values, an important issue that has recently received renewed attention. Double Q-learning has been proposed as an efficient algorithm to mitigate this bias. However, this comes at the price of an underestimation of action values, in addition to increased memory requirements and a slower convergence. In this paper, we introduce a new way to address the maximization bias in the form of a \"self-correcting algorithm\" for approximating the maximum of an expected value. Our method balances the overestimation of the single estimator used in conventional Q-learning and the underestimation of the double estimator used in Double Q-learning. Applying this strategy to Q-learning results in Self-correcting Q-learning. We show theoretically that this new algorithm enjoys the same convergence guarantees as Q-learning while being more accurate. Empirically, it performs better than Double Q-learning in domains with rewards of high variance, and it even attains faster convergence than Q-learning in domains with rewards of zero or low variance. These advantages transfer to a Deep Q Network implementation that we call Self-correcting DQN and which outperforms regular DQN and Double DQN on several tasks in the Atari 2600 domain."
                    },
                    {
                        "title": "Deep Bandits Show-Off: Simple and Efficient Exploration with Deep Networks",
                        "abstract": "Designing efficient exploration is central to Reinforcement Learning due to the fundamental problem posed by the exploration-exploitation dilemma. Bayesian exploration strategies like Thompson Sampling resolve this trade-off in a principled way by modeling and updating the distribution of the parameters of the action-value function, the outcome model of the environment. However, this technique becomes infeasible for complex environments due to the computational intractability of maintaining probability distributions over parameters of outcome models of corresponding complexity. Moreover, the approximation techniques introduced to mitigate this issue typically result in poor exploration-exploitation trade-offs, as observed in the case of deep neural network models with approximate posterior methods that have been shown to underperform in the deep bandit scenario. In this paper we introduce Sample Average Uncertainty (SAU), a simple and efficient uncertainty measure for contextual bandits. While Bayesian approaches like Thompson Sampling estimate outcomes uncertainty indirectly by first quantifying the variability over the parameters of the outcome model, SAU is a frequentist approach that directly estimates the uncertainty of the outcomes based on the value predictions. Importantly, we show theoretically that the uncertainty measure estimated by SAU asymptotically matches the uncertainty provided by Thompson Sampling, as well as its regret bounds. Because of its simplicity SAU can be seamlessly applied to deep contextual bandits as a very scalable drop-in replacement for epsilon-greedy exploration. We confirm empirically our theory by showing that SAU-based exploration outperforms current state-of-the-art deep Bayesian bandit methods on several real-world datasets at modest computation cost. Code is available at \\url{https://github.com/ibm/sau-explore}."
                    },
                    {
                        "title": "Unbalanced Sobolev Descent",
                        "abstract": "We introduce Unbalanced Sobolev Descent (USD), a particle descent algorithm for transporting a high dimensional source distribution to a target distribution that does not necessarily have the same mass. We define the Sobolev-Fisher discrepancy between distributions and show that it relates to advection-reaction transport equations and the Wasserstein-Fisher-Rao metric between distributions. USD transports particles along gradient flows of the witness function of the Sobolev-Fisher discrepancy (advection step) and reweighs the mass of particles with respect to this witness function (reaction step). The reaction step can be thought of as a birth-death process of the particles with rate of growth proportional to the witness function. When the Sobolev-Fisher witness function is estimated in a Reproducing Kernel Hilbert Space (RKHS), under mild assumptions we show that USD converges asymptotically (in the limit of infinite particles) to the target distribution in the Maximum Mean Discrepancy (MMD) sense. We then give two methods to estimate the Sobolev-Fisher witness with neural networks, resulting in two Neural USD algorithms. The first one implements the reaction step with mirror descent on the weights, while the second implements it through a birth-death process of particles. We show on synthetic examples that USD transports distributions with or without conservation of mass faster than previous particle descent algorithms, and finally demonstrate its use for molecular biology analyses where our method is naturally suited to match developmental stages of populations of differentiating cells based on their single-cell RNA sequencing profile. Code is available at https://github.com/ibm/usd ."
                    },
                    {
                        "title": "Model-Assisted Labeling via Explainability for Visual Inspection of Civil Infrastructures",
                        "abstract": "Labeling images for visual segmentation is a time-consuming task which can be costly, particularly in application domains where labels have to be provided by specialized expert annotators, such as civil engineering. In this paper, we propose to use attribution methods to harness the valuable interactions between expert annotators and the data to be annotated in the case of defect segmentation for visual inspection of civil infrastructures. Concretely, a classifier is trained to detect defects and coupled with an attribution-based method and adversarial climbing to generate and refine segmentation masks corresponding to the classification outputs. These are used within an assisted labeling framework where the annotators can interact with them as proposal segmentation masks by deciding to accept, reject or modify them, and interactions are logged as weak labels to further refine the classifier. Applied on a real-world dataset resulting from the automated visual inspection of bridges, our proposed method is able to save more than 50\\% of annotators' time when compared to manual annotation of defects."
                    },
                    {
                        "title": "Adaptive Conformal Regression with Jackknife+ Rescaled Scores",
                        "abstract": "Conformal regression provides prediction intervals with global coverage guarantees, but often fails to capture local error distributions, leading to non-homogeneous coverage. We address this with a new adaptive method based on rescaling conformal scores with an estimate of local score distribution, inspired by the Jackknife+ method, which enables the use of calibration data in conformal scores without breaking calibration-test exchangeability. Our approach ensures formal global coverage guarantees and is supported by new theoretical results on local coverage, including an a posteriori bound on any calibration score. The strength of our approach lies in achieving local coverage without sacrificing calibration set size, improving the applicability of conformal prediction intervals in various settings. As a result, our method provides prediction intervals that outperform previous methods, particularly in the low-data regime, making it especially relevant for real-world applications such as healthcare and biomedical domains where uncertainty needs to be quantified accurately despite low sample data."
                    },
                    {
                        "title": "Energy-efficient neuromorphic classifiers",
                        "abstract": "Neuromorphic engineering combines the architectural and computational principles of systems neuroscience with semiconductor electronics, with the aim of building efficient and compact devices that mimic the synaptic and neural machinery of the brain. Neuromorphic engineering promises extremely low energy consumptions, comparable to those of the nervous system. However, until now the neuromorphic approach has been restricted to relatively simple circuits and specialized functions, rendering elusive a direct comparison of their energy consumption to that used by conventional von Neumann digital machines solving real-world tasks. Here we show that a recent technology developed by IBM can be leveraged to realize neuromorphic circuits that operate as classifiers of complex real-world stimuli. These circuits emulate enough neurons to compete with state-of-the-art classifiers. We also show that the energy consumption of the IBM chip is typically 2 or more orders of magnitude lower than that of conventional digital machines when implementing classifiers with comparable performance. Moreover, the spike-based dynamics display a trade-off between integration time and accuracy, which naturally translates into algorithms that can be flexibly deployed for either fast and approximate classifications, or more accurate classifications at the mere expense of longer running times and higher energy costs. This work finally proves that the neuromorphic approach can be efficiently used in real-world applications and it has significant advantages over conventional digital devices when energy consumption is considered."
                    },
                    {
                        "title": "DARE: Towards Robust Text Explanations in Biomedical and Healthcare Applications",
                        "abstract": "Along with the successful deployment of deep neural networks in several application domains, the need to unravel the black-box nature of these networks has seen a significant increase recently. Several methods have been introduced to provide insight into the inference process of deep neural networks. However, most of these explainability methods have been shown to be brittle in the face of adversarial perturbations of their inputs in the image and generic textual domain. In this work we show that this phenomenon extends to specific and important high stakes domains like biomedical datasets. In particular, we observe that the robustness of explanations should be characterized in terms of the accuracy of the explanation in linking a model's inputs and its decisions - faithfulness - and its relevance from the perspective of domain experts - plausibility. This is crucial to prevent explanations that are inaccurate but still look convincing in the context of the domain at hand. To this end, we show how to adapt current attribution robustness estimation methods to a given domain, so as to take into account domain-specific plausibility. This results in our DomainAdaptiveAREstimator (DARE) attribution robustness estimator, allowing us to properly characterize the domain-specific robustness of faithful explanations. Next, we provide two methods, adversarial training and FAR training, to mitigate the brittleness characterized by DARE, allowing us to train networks that display robust attributions. Finally, we empirically validate our methods with extensive experiments on three established biomedical benchmarks."
                    },
                    {
                        "title": "Outline-Guided Object Inpainting with Diffusion Models",
                        "abstract": "Instance segmentation datasets play a crucial role in training accurate and robust computer vision models. However, obtaining accurate mask annotations to produce high-quality segmentation datasets is a costly and labor-intensive process. In this work, we show how this issue can be mitigated by starting with small annotated instance segmentation datasets and augmenting them to effectively obtain a sizeable annotated dataset. We achieve that by creating variations of the available annotated object instances in a way that preserves the provided mask annotations, thereby resulting in new image-mask pairs to be added to the set of annotated images. Specifically, we generate new images using a diffusion-based inpainting model to fill out the masked area with a desired object class by guiding the diffusion through the object outline. We show that the object outline provides a simple, but also reliable and convenient training-free guidance signal for the underlying inpainting model that is often sufficient to fill out the mask with an object of the correct class without further text guidance and preserve the correspondence between generated images and the mask annotations with high precision. Our experimental results reveal that our method successfully generates realistic variations of object instances, preserving their shape characteristics while introducing diversity within the augmented area. We also show that the proposed method can naturally be combined with text guidance and other image augmentation techniques."
                    },
                    {
                        "title": "Estimating the Adversarial Robustness of Attributions in Text with Transformers",
                        "abstract": "Explanations are crucial parts of deep neural network (DNN) classifiers. In high stakes applications, faithful and robust explanations are important to understand and gain trust in DNN classifiers. However, recent work has shown that state-of-the-art attribution methods in text classifiers are susceptible to imperceptible adversarial perturbations that alter explanations significantly while maintaining the correct prediction outcome. If undetected, this can critically mislead the users of DNNs. Thus, it is crucial to understand the influence of such adversarial perturbations on the networks' explanations and their perceptibility. In this work, we establish a novel definition of attribution robustness (AR) in text classification, based on Lipschitz continuity. Crucially, it reflects both attribution change induced by adversarial input alterations and perceptibility of such alterations. Moreover, we introduce a wide set of text similarity measures to effectively capture locality between two text samples and imperceptibility of adversarial perturbations in text. We then propose our novel TransformerExplanationAttack (TEA), a strong adversary that provides a tight estimation for attribution robustness in text classification. TEA uses state-of-the-art language models to extract word substitutions that result in fluent, contextual adversarial samples. Finally, with experiments on several text classification architectures, we show that TEA consistently outperforms current state-of-the-art AR estimators, yielding perturbations that alter explanations to a greater extent while being more fluent and less perceptible."
                    },
                    {
                        "title": "On the generalization capacity of neural networks during generic multimodal reasoning",
                        "abstract": "The advent of the Transformer has led to the development of large language models (LLM), which appear to demonstrate human-like capabilities. To assess the generality of this class of models and a variety of other base neural network architectures to multimodal domains, we evaluated and compared their capacity for multimodal generalization. We introduce a multimodal question-answer benchmark to evaluate three specific types of out-of-distribution (OOD) generalization performance: distractor generalization (generalization in the presence of distractors), systematic compositional generalization (generalization to new task permutations), and productive compositional generalization (generalization to more complex tasks structures). We found that across model architectures (e.g., RNNs, Transformers, Perceivers, etc.), models with multiple attention layers, or models that leveraged cross-attention mechanisms between input domains, fared better. Our positive results demonstrate that for multimodal distractor and systematic generalization, either cross-modal attention or models with deeper attention layers are key architectural features required to integrate multimodal inputs. On the other hand, neither of these architectural features led to productive generalization, suggesting fundamental limitations of existing architectures for specific types of multimodal generalization. These results demonstrate the strengths and limitations of specific architectural components underlying modern neural models for multimodal reasoning. Finally, we provide Generic COG (gCOG), a configurable benchmark with several multimodal generalization splits, for future studies to explore."
                    },
                    {
                        "title": "Efficient ConvNets for Analog Arrays",
                        "abstract": "Analog arrays are a promising upcoming hardware technology with the potential to drastically speed up deep learning. Their main advantage is that they compute matrix-vector products in constant time, irrespective of the size of the matrix. However, early convolution layers in ConvNets map very unfavorably onto analog arrays, because kernel matrices are typically small and the constant time operation needs to be sequentially iterated a large number of times, reducing the speed up advantage for ConvNets. Here, we propose to replicate the kernel matrix of a convolution layer on distinct analog arrays, and randomly divide parts of the compute among them, so that multiple kernel matrices are trained in parallel. With this modification, analog arrays execute ConvNets with an acceleration factor that is proportional to the number of kernel matrices used per layer (here tested 16-128). Despite having more free parameters, we show analytically and in numerical experiments that this convolution architecture is self-regularizing and implicitly learns similar filters across arrays. We also report superior performance on a number of datasets and increased robustness to adversarial attacks. Our investigation suggests to revise the notion that mixed analog-digital hardware is not suitable for ConvNets."
                    },
                    {
                        "title": "Alleviating Noisy Data in Image Captioning with Cooperative Distillation",
                        "abstract": "Image captioning systems have made substantial progress, largely due to the availability of curated datasets like Microsoft COCO or Vizwiz that have accurate descriptions of their corresponding images. Unfortunately, scarce availability of such cleanly labeled data results in trained algorithms producing captions that can be terse and idiosyncratically specific to details in the image. We propose a new technique, cooperative distillation that combines clean curated datasets with the web-scale automatically extracted captions of the Google Conceptual Captions dataset (GCC), which can have poor descriptions of images, but is abundant in size and therefore provides a rich vocabulary resulting in more expressive captions."
                    },
                    {
                        "title": "Sobolev Independence Criterion",
                        "abstract": "We propose the Sobolev Independence Criterion (SIC), an interpretable dependency measure between a high dimensional random variable X and a response variable Y . SIC decomposes to the sum of feature importance scores and hence can be used for nonlinear feature selection. SIC can be seen as a gradient regularized Integral Probability Metric (IPM) between the joint distribution of the two random variables and the product of their marginals. We use sparsity inducing gradient penalties to promote input sparsity of the critic of the IPM. In the kernel version we show that SIC can be cast as a convex optimization problem by introducing auxiliary variables that play an important role in feature selection as they are normalized feature importance scores. We then present a neural version of SIC where the critic is parameterized as a homogeneous neural network, improving its representation power as well as its interpretability. We conduct experiments validating SIC for feature selection in synthetic and real-world experiments. We show that SIC enables reliable and interpretable discoveries, when used in conjunction with the holdout randomization test and knockoffs to control the False Discovery Rate. Code is available at http://github.com/ibm/sic."
                    },
                    {
                        "title": "Active Learning for Imbalanced Civil Infrastructure Data",
                        "abstract": "Aging civil infrastructures are closely monitored by engineers for damage and critical defects. As the manual inspection of such large structures is costly and time-consuming, we are working towards fully automating the visual inspections to support the prioritization of maintenance activities. To that end we combine recent advances in drone technology and deep learning. Unfortunately, annotation costs are incredibly high as our proprietary civil engineering dataset must be annotated by highly trained engineers. Active learning is, therefore, a valuable tool to optimize the trade-off between model performance and annotation costs. Our use-case differs from the classical active learning setting as our dataset suffers from heavy class imbalance and consists of a much larger already labeled data pool than other active learning research. We present a novel method capable of operating in this challenging setting by replacing the traditional active learning acquisition function with an auxiliary binary discriminator. We experimentally show that our novel method outperforms the best-performing traditional active learning method (BALD) by 5% and 38% accuracy on CIFAR-10 and our proprietary dataset respectively."
                    },
                    {
                        "title": "Multivariate Stochastic Dominance via Optimal Transport and Applications to Models Benchmarking",
                        "abstract": "Stochastic dominance is an important concept in probability theory, econometrics and social choice theory for robustly modeling agents' preferences between random outcomes. While many works have been dedicated to the univariate case, little has been done in the multivariate scenario, wherein an agent has to decide between different multivariate outcomes. By exploiting a characterization of multivariate first stochastic dominance in terms of couplings, we introduce a statistic that assesses multivariate almost stochastic dominance under the framework of Optimal Transport with a smooth cost. Further, we introduce an entropic regularization of this statistic, and establish a central limit theorem (CLT) and consistency of the bootstrap procedure for the empirical statistic. Armed with this CLT, we propose a hypothesis testing framework as well as an efficient implementation using the Sinkhorn algorithm. We showcase our method in comparing and benchmarking Large Language Models that are evaluated on multiple metrics. Our multivariate stochastic dominance test allows us to capture the dependencies between the metrics in order to make an informed and statistically significant decision on the relative performance of the models."
                    },
                    {
                        "title": "Cloud-Based Real-Time Molecular Screening Platform with MolFormer",
                        "abstract": "With the prospect of automating a number of chemical tasks with high fidelity, chemical language processing models are emerging at a rapid speed. Here, we present a cloud-based real-time platform that allows users to virtually screen molecules of interest. For this purpose, molecular embeddings inferred from a recently proposed large chemical language model, named MolFormer, are leveraged. The platform currently supports three tasks: nearest neighbor retrieval, chemical space visualization, and property prediction. Based on the functionalities of this platform and results obtained, we believe that such a platform can play a pivotal role in automating chemistry and chemical engineering research, as well as assist in drug discovery and material design tasks. A demo of our platform is provided at \\url{www.ibm.biz/molecular_demo}."
                    },
                    {
                        "title": "Tabular Transformers for Modeling Multivariate Time Series",
                        "abstract": "Tabular datasets are ubiquitous in data science applications. Given their importance, it seems natural to apply state-of-the-art deep learning algorithms in order to fully unlock their potential. Here we propose neural network models that represent tabular time series that can optionally leverage their hierarchical structure. This results in two architectures for tabular time series: one for learning representations that is analogous to BERT and can be pre-trained end-to-end and used in downstream tasks, and one that is akin to GPT and can be used for generation of realistic synthetic tabular sequences. We demonstrate our models on two datasets: a synthetic credit card transaction dataset, where the learned representations are used for fraud detection and synthetic data generation, and on a real pollution dataset, where the learned encodings are used to predict atmospheric pollutant concentrations. Code and data are available at https://github.com/IBM/TabFormer."
                    },
                    {
                        "title": "Unraveling the Key Components of OOD Generalization via Diversification",
                        "abstract": "Supervised learning datasets may contain multiple cues that explain the training set equally well, i.e., learning any of them would lead to the correct predictions on the training data. However, many of them can be spurious, i.e., lose their predictive power under a distribution shift and consequently fail to generalize to out-of-distribution (OOD) data. Recently developed \"diversification\" methods (Lee et al., 2023; Pagliardini et al., 2023) approach this problem by finding multiple diverse hypotheses that rely on different features. This paper aims to study this class of methods and identify the key components contributing to their OOD generalization abilities.   We show that (1) diversification methods are highly sensitive to the distribution of the unlabeled data used for diversification and can underperform significantly when away from a method-specific sweet spot. (2) Diversification alone is insufficient for OOD generalization. The choice of the used learning algorithm, e.g., the model's architecture and pretraining, is crucial. In standard experiments (classification on Waterbirds and Office-Home datasets), using the second-best choice leads to an up to 20\\% absolute drop in accuracy. (3) The optimal choice of learning algorithm depends on the unlabeled data and vice versa i.e. they are co-dependent. (4) Finally, we show that, in practice, the above pitfalls cannot be alleviated by increasing the number of diverse hypotheses, the major feature of diversification methods.   These findings provide a clearer understanding of the critical design factors influencing the OOD generalization abilities of diversification methods. They can guide practitioners in how to use the existing methods best and guide researchers in developing new, better ones."
                    },
                    {
                        "title": "Image Captioning as an Assistive Technology: Lessons Learned from VizWiz 2020 Challenge",
                        "abstract": "Image captioning has recently demonstrated impressive progress largely owing to the introduction of neural network algorithms trained on curated dataset like MS-COCO. Often work in this field is motivated by the promise of deployment of captioning systems in practical applications. However, the scarcity of data and contexts in many competition datasets renders the utility of systems trained on these datasets limited as an assistive technology in real-world settings, such as helping visually impaired people navigate and accomplish everyday tasks. This gap motivated the introduction of the novel VizWiz dataset, which consists of images taken by the visually impaired and captions that have useful, task-oriented information. In an attempt to help the machine learning computer vision field realize its promise of producing technologies that have positive social impact, the curators of the VizWiz dataset host several competitions, including one for image captioning. This work details the theory and engineering from our winning submission to the 2020 captioning competition. Our work provides a step towards improved assistive image captioning systems."
                    }
                ]
            },
            "3eacab55-d40f-4bd2-a809-f989d1375005": {
                "pk": "3eacab55-d40f-4bd2-a809-f989d1375005",
                "name": "Apoorva Nitsure",
                "collaborators": [
                    "Rajesh Bordawekar",
                    "Jose Neves",
                    "Prabhakar Kudva",
                    "Gabriel Rioux",
                    "Mattia Rigotti",
                    "Kristjan Greenewald",
                    "Youssef Mroueh",
                    "Prashanth Vijayaraghavan",
                    "Luyao Shi",
                    "Stefano Ambrogio"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Database Systems",
                    "Machine Learning",
                    "Code Generation"
                ],
                "publications": [
                    {
                        "title": "Unlocking New York City Crime Insights using Relational Database Embeddings",
                        "abstract": "This version withdrawn by arXiv administrators because the author did not have the right to agree to our license at the time of submission."
                    },
                    {
                        "title": "A Scalable Space-efficient In-database Interpretability Framework for Embedding-based Semantic SQL Queries",
                        "abstract": "AI-Powered database (AI-DB) is a novel relational database system that uses a self-supervised neural network, database embedding, to enable semantic SQL queries on relational tables. In this paper, we describe an architecture and implementation of in-database interpretability infrastructure designed to provide simple, transparent, and relatable insights into ranked results of semantic SQL queries supported by AI-DB. We introduce a new co-occurrence based interpretability approach to capture relationships between relational entities and describe a space-efficient probabilistic Sketch implementation to store and process co-occurrence counts. Our approach provides both query-agnostic (global) and query-specific (local) interpretabilities. Experimental evaluation demonstrate that our in-database probabilistic approach provides the same interpretability quality as the precise space-inefficient approach, while providing scalable and space efficient runtime behavior (up to 8X space savings), without any user intervention."
                    },
                    {
                        "title": "Multivariate Stochastic Dominance via Optimal Transport and Applications to Models Benchmarking",
                        "abstract": "Stochastic dominance is an important concept in probability theory, econometrics and social choice theory for robustly modeling agents' preferences between random outcomes. While many works have been dedicated to the univariate case, little has been done in the multivariate scenario, wherein an agent has to decide between different multivariate outcomes. By exploiting a characterization of multivariate first stochastic dominance in terms of couplings, we introduce a statistic that assesses multivariate almost stochastic dominance under the framework of Optimal Transport with a smooth cost. Further, we introduce an entropic regularization of this statistic, and establish a central limit theorem (CLT) and consistency of the bootstrap procedure for the empirical statistic. Armed with this CLT, we propose a hypothesis testing framework as well as an efficient implementation using the Sinkhorn algorithm. We showcase our method in comparing and benchmarking Large Language Models that are evaluated on multiple metrics. Our multivariate stochastic dominance test allows us to capture the dependencies between the metrics in order to make an informed and statistically significant decision on the relative performance of the models."
                    },
                    {
                        "title": "VHDL-Eval: A Framework for Evaluating Large Language Models in VHDL Code Generation",
                        "abstract": "With the unprecedented advancements in Large Language Models (LLMs), their application domains have expanded to include code generation tasks across various programming languages. While significant progress has been made in enhancing LLMs for popular programming languages, there exists a notable gap in comprehensive evaluation frameworks tailored for Hardware Description Languages (HDLs), particularly VHDL. This paper addresses this gap by introducing a comprehensive evaluation framework designed specifically for assessing LLM performance in VHDL code generation task. We construct a dataset for evaluating LLMs on VHDL code generation task. This dataset is constructed by translating a collection of Verilog evaluation problems to VHDL and aggregating publicly available VHDL problems, resulting in a total of 202 problems. To assess the functional correctness of the generated VHDL code, we utilize a curated set of self-verifying testbenches specifically designed for those aggregated VHDL problem set. We conduct an initial evaluation of different LLMs and their variants, including zero-shot code generation, in-context learning (ICL), and Parameter-efficient fine-tuning (PEFT) methods. Our findings underscore the considerable challenges faced by existing LLMs in VHDL code generation, revealing significant scope for improvement. This study emphasizes the necessity of supervised fine-tuning code generation models specifically for VHDL, offering potential benefits to VHDL designers seeking efficient code generation solutions."
                    }
                ]
            },
            "cada26f0-01ab-4c64-9ff4-3de784985cee": {
                "pk": "cada26f0-01ab-4c64-9ff4-3de784985cee",
                "name": "Mikhail Yurochkin",
                "collaborators": [
                    "Yuekai Sun",
                    "XuanLong Nguyen",
                    "Mayank Agarwal",
                    "Aritra Guha",
                    "Soumya Ghosh",
                    "Justin Solomon",
                    "Amanda Bower",
                    "Kristjan Greenewald",
                    "Trong Nghia Hoang",
                    "Songkai Xue"
                ],
                "domain": [
                    "Machine Learning",
                    "Fairness",
                    "Topic Modeling",
                    "Bayesian Inference"
                ],
                "publications": [
                    {
                        "title": "Geometric Dirichlet Means algorithm for topic inference",
                        "abstract": "We propose a geometric algorithm for topic learning and inference that is built on the convex geometry of topics arising from the Latent Dirichlet Allocation (LDA) model and its nonparametric extensions. To this end we study the optimization of a geometric loss function, which is a surrogate to the LDA's likelihood. Our method involves a fast optimization based weighted clustering procedure augmented with geometric corrections, which overcomes the computational and statistical inefficiencies encountered by other techniques based on Gibbs sampling and variational inference, while achieving the accuracy comparable to that of a Gibbs sampler. The topic estimates produced by our method are shown to be statistically consistent under some conditions. The algorithm is evaluated with extensive experiments on simulated and real data."
                    },
                    {
                        "title": "SenSeI: Sensitive Set Invariance for Enforcing Individual Fairness",
                        "abstract": "In this paper, we cast fair machine learning as invariant machine learning. We first formulate a version of individual fairness that enforces invariance on certain sensitive sets. We then design a transport-based regularizer that enforces this version of individual fairness and develop an algorithm to minimize the regularizer efficiently. Our theoretical results guarantee the proposed approach trains certifiably fair ML models. Finally, in the experimental studies we demonstrate improved fairness metrics in comparison to several recent fair training procedures on three ML tasks that are susceptible to algorithmic bias."
                    },
                    {
                        "title": "Multi-way Interacting Regression via Factorization Machines",
                        "abstract": "We propose a Bayesian regression method that accounts for multi-way interactions of arbitrary orders among the predictor variables. Our model makes use of a factorization mechanism for representing the regression coefficients of interactions among the predictors, while the interaction selection is guided by a prior distribution on random hypergraphs, a construction which generalizes the Finite Feature Model. We present a posterior inference algorithm based on Gibbs sampling, and establish posterior consistency of our regression model. Our method is evaluated with extensive experiments on simulated data and demonstrated to be able to identify meaningful interactions in applications in genetics and retail demand forecasting."
                    },
                    {
                        "title": "Conic Scan-and-Cover algorithms for nonparametric topic modeling",
                        "abstract": "We propose new algorithms for topic modeling when the number of topics is unknown. Our approach relies on an analysis of the concentration of mass and angular geometry of the topic simplex, a convex polytope constructed by taking the convex hull of vertices representing the latent topics. Our algorithms are shown in practice to have accuracy comparable to a Gibbs sampler in terms of topic estimation, which requires the number of topics be given. Moreover, they are one of the fastest among several state of the art parametric techniques. Statistical consistency of our estimator is established under some conditions."
                    },
                    {
                        "title": "Training individually fair ML models with Sensitive Subspace Robustness",
                        "abstract": "We consider training machine learning models that are fair in the sense that their performance is invariant under certain sensitive perturbations to the inputs. For example, the performance of a resume screening system should be invariant under changes to the gender and/or ethnicity of the applicant. We formalize this notion of algorithmic fairness as a variant of individual fairness and develop a distributionally robust optimization approach to enforce it during training. We also demonstrate the effectiveness of the approach on two ML tasks that are susceptible to gender and racial biases."
                    },
                    {
                        "title": "Dirichlet Simplex Nest and Geometric Inference",
                        "abstract": "We propose Dirichlet Simplex Nest, a class of probabilistic models suitable for a variety of data types, and develop fast and provably accurate inference algorithms by accounting for the model's convex geometry and low dimensional simplicial structure. By exploiting the connection to Voronoi tessellation and properties of Dirichlet distribution, the proposed inference algorithm is shown to achieve consistency and strong error bound guarantees on a range of model settings and data distributions. The effectiveness of our model and the learning algorithm is demonstrated by simulations and by analyses of text and financial data."
                    },
                    {
                        "title": "Statistical Model Aggregation via Parameter Matching",
                        "abstract": "We consider the problem of aggregating models learned from sequestered, possibly heterogeneous datasets. Exploiting tools from Bayesian nonparametrics, we develop a general meta-modeling framework that learns shared global latent structures by identifying correspondences among local model parameterizations. Our proposed framework is model-independent and is applicable to a wide range of model types. After verifying our approach on simulated data, we demonstrate its utility in aggregating Gaussian topic models, hierarchical Dirichlet process based hidden Markov models, and sparse Gaussian processes with applications spanning text summarization, motion capture analysis, and temperature forecasting."
                    },
                    {
                        "title": "Auditing ML Models for Individual Bias and Unfairness",
                        "abstract": "We consider the task of auditing ML models for individual bias/unfairness. We formalize the task in an optimization problem and develop a suite of inferential tools for the optimal value. Our tools permit us to obtain asymptotic confidence intervals and hypothesis tests that cover the target/control the Type I error rate exactly. To demonstrate the utility of our tools, we use them to reveal the gender and racial biases in Northpointe's COMPAS recidivism prediction instrument."
                    },
                    {
                        "title": "Individually Fair Ranking",
                        "abstract": "We develop an algorithm to train individually fair learning-to-rank (LTR) models. The proposed approach ensures items from minority groups appear alongside similar items from majority groups. This notion of fair ranking is based on the definition of individual fairness from supervised learning and is more nuanced than prior fair LTR approaches that simply ensure the ranking model provides underrepresented items with a basic level of exposure. The crux of our method is an optimal transport-based regularizer that enforces individual fairness and an efficient algorithm for optimizing the regularizer. We show that our approach leads to certifiably individually fair LTR models and demonstrate the efficacy of our method on ranking tasks subject to demographic biases."
                    },
                    {
                        "title": "Individually Fair Gradient Boosting",
                        "abstract": "We consider the task of enforcing individual fairness in gradient boosting. Gradient boosting is a popular method for machine learning from tabular data, which arise often in applications where algorithmic fairness is a concern. At a high level, our approach is a functional gradient descent on a (distributionally) robust loss function that encodes our intuition of algorithmic fairness for the ML task at hand. Unlike prior approaches to individual fairness that only work with smooth ML models, our approach also works with non-smooth models such as decision trees. We show that our algorithm converges globally and generalizes. We also demonstrate the efficacy of our algorithm on three ML problems susceptible to algorithmic bias."
                    },
                    {
                        "title": "On sensitivity of meta-learning to support data",
                        "abstract": "Meta-learning algorithms are widely used for few-shot learning. For example, image recognition systems that readily adapt to unseen classes after seeing only a few labeled examples. Despite their success, we show that modern meta-learning algorithms are extremely sensitive to the data used for adaptation, i.e. support data. In particular, we demonstrate the existence of (unaltered, in-distribution, natural) images that, when used for adaptation, yield accuracy as low as 4\\% or as high as 95\\% on standard few-shot image classification benchmarks. We explain our empirical findings in terms of class margins, which in turn suggests that robust and safe meta-learning requires larger margins than supervised learning."
                    },
                    {
                        "title": "Calibrated Data-Dependent Constraints with Exact Satisfaction Guarantees",
                        "abstract": "We consider the task of training machine learning models with data-dependent constraints. Such constraints often arise as empirical versions of expected value constraints that enforce fairness or stability goals. We reformulate data-dependent constraints so that they are calibrated: enforcing the reformulated constraints guarantees that their expected value counterparts are satisfied with a user-prescribed probability. The resulting optimization problem is amendable to standard stochastic optimization algorithms, and we demonstrate the efficacy of our method on a fairness-sensitive classification task where we wish to guarantee the classifier's fairness (at test time)."
                    },
                    {
                        "title": "Prompt Exploration with Prompt Regression",
                        "abstract": "In the advent of democratized usage of large language models (LLMs), there is a growing desire to systematize LLM prompt creation and selection processes beyond iterative trial-and-error. Prior works majorly focus on searching the space of prompts without accounting for relations between prompt variations. Here we propose a framework, Prompt Exploration with Prompt Regression (PEPR), to predict the effect of prompt combinations given results for individual prompt elements as well as a simple method to select an effective prompt for a given use-case. We evaluate our approach with open-source LLMs of different sizes on several different tasks."
                    },
                    {
                        "title": "Model Fusion with Kullback--Leibler Divergence",
                        "abstract": "We propose a method to fuse posterior distributions learned from heterogeneous datasets. Our algorithm relies on a mean field assumption for both the fused model and the individual dataset posteriors and proceeds using a simple assign-and-average approach. The components of the dataset posteriors are assigned to the proposed global model components by solving a regularized variant of the assignment problem. The global components are then updated based on these assignments by their mean under a KL divergence. For exponential family variational distributions, our formulation leads to an efficient non-parametric algorithm for computing the fused model. Our algorithm is easy to describe and implement, efficient, and competitive with state-of-the-art on motion capture analysis, topic modeling, and federated learning of Bayesian neural networks."
                    },
                    {
                        "title": "Bayesian Nonparametric Federated Learning of Neural Networks",
                        "abstract": "In federated learning problems, data is scattered across different servers and exchanging or pooling it is often impractical or prohibited. We develop a Bayesian nonparametric framework for federated learning with neural networks. Each data server is assumed to provide local neural network weights, which are modeled through our framework. We then develop an inference approach that allows us to synthesize a more expressive global network without additional supervision, data pooling and with as few as a single communication round. We then demonstrate the efficacy of our approach on federated learning problems simulated from two popular image classification datasets."
                    },
                    {
                        "title": "Online Semi-Supervised Learning with Bandit Feedback",
                        "abstract": "We formulate a new problem at the intersectionof semi-supervised learning and contextual bandits,motivated by several applications including clini-cal trials and ad recommendations. We demonstratehow Graph Convolutional Network (GCN), a semi-supervised learning approach, can be adjusted tothe new problem formulation. We also propose avariant of the linear contextual bandit with semi-supervised missing rewards imputation. We thentake the best of both approaches to develop multi-GCN embedded contextual bandit. Our algorithmsare verified on several real world datasets."
                    },
                    {
                        "title": "Scalable inference of topic evolution via models for latent geometric structures",
                        "abstract": "We develop new models and algorithms for learning the temporal dynamics of the topic polytopes and related geometric objects that arise in topic model based inference. Our model is nonparametric Bayesian and the corresponding inference algorithm is able to discover new topics as the time progresses. By exploiting the connection between the modeling of topic polytope evolution, Beta-Bernoulli process and the Hungarian matching algorithm, our method is shown to be several orders of magnitude faster than existing topic modeling approaches, as demonstrated by experiments working with several million documents in under two dozens of minutes."
                    },
                    {
                        "title": "Hierarchical Optimal Transport for Document Representation",
                        "abstract": "The ability to measure similarity between documents enables intelligent summarization and analysis of large corpora. Past distances between documents suffer from either an inability to incorporate semantic similarities between words or from scalability issues. As an alternative, we introduce hierarchical optimal transport as a meta-distance between documents, where documents are modeled as distributions over topics, which themselves are modeled as distributions over words. We then solve an optimal transport problem on the smaller topic space to compute a similarity score. We give conditions on the topics under which this construction defines a distance, and we relate it to the word mover's distance. We evaluate our technique for k-NN classification and show better interpretability and scalability with comparable performance to current methods at a fraction of the cost."
                    },
                    {
                        "title": "Continuous Regularized Wasserstein Barycenters",
                        "abstract": "Wasserstein barycenters provide a geometrically meaningful way to aggregate probability distributions, built on the theory of optimal transport. They are difficult to compute in practice, however, leading previous work to restrict their supports to finite sets of points. Leveraging a new dual formulation for the regularized Wasserstein barycenter problem, we introduce a stochastic algorithm that constructs a continuous approximation of the barycenter. We establish strong duality and use the corresponding primal-dual relationship to parametrize the barycenter implicitly using the dual potentials of regularized transport problems. The resulting problem can be solved with stochastic gradient descent, which yields an efficient online algorithm to approximate the barycenter of continuous distributions given sample access. We demonstrate the effectiveness of our approach and compare against previous work on synthetic examples and real-world applications."
                    },
                    {
                        "title": "Two Simple Ways to Learn Individual Fairness Metrics from Data",
                        "abstract": "Individual fairness is an intuitive definition of algorithmic fairness that addresses some of the drawbacks of group fairness. Despite its benefits, it depends on a task specific fair metric that encodes our intuition of what is fair and unfair for the ML task at hand, and the lack of a widely accepted fair metric for many ML tasks is the main barrier to broader adoption of individual fairness. In this paper, we present two simple ways to learn fair metrics from a variety of data types. We show empirically that fair training with the learned metrics leads to improved fairness on three machine learning tasks susceptible to gender and racial biases. We also provide theoretical guarantees on the statistical performance of both approaches."
                    }
                ]
            },
            "15b55c89-9b2f-45ec-bc87-417dfea8696c": {
                "pk": "15b55c89-9b2f-45ec-bc87-417dfea8696c",
                "name": "Kristjan Greenewald",
                "collaborators": [
                    "Alfred Hero III",
                    "Alfred Hero",
                    "Ziv Goldfeld",
                    "Shuheng Zhou",
                    "Edmund Zelnio",
                    "Stephen Kelley",
                    "Predrag Klasnja",
                    "Susan Murphy",
                    "Justin Solomon",
                    "Alfred O. Hero III"
                ],
                "domain": [
                    "Anomaly Detection",
                    "Causal Inference",
                    "Machine Learning",
                    "Optimal Transport"
                ],
                "publications": [
                    {
                        "title": "Detection of Anomalous Crowd Behavior Using Spatio-Temporal Multiresolution Model and Kronecker Sum Decompositions",
                        "abstract": "In this work we consider the problem of detecting anomalous spatio-temporal behavior in videos. Our approach is to learn the normative multiframe pixel joint distribution and detect deviations from it using a likelihood based approach. Due to the extreme lack of available training samples relative to the dimension of the distribution, we use a mean and covariance approach and consider methods of learning the spatio-temporal covariance in the low-sample regime. Our approach is to estimate the covariance using parameter reduction and sparse models. The first method considered is the representation of the covariance as a sum of Kronecker products as in (Greenewald et al 2013), which is found to be an accurate approximation in this setting. We propose learning algorithms relevant to our problem. We then consider the sparse multiresolution model of (Choi et al 2010) and apply the Kronecker product methods to it for further parameter reduction, as well as introducing modifications for enhanced efficiency and greater applicability to spatio-temporal covariance matrices. We apply our methods to the detection of crowd behavior anomalies in the University of Minnesota crowd anomaly dataset, and achieve competitive results."
                    },
                    {
                        "title": "Sliced Mutual Information: A Scalable Measure of Statistical Dependence",
                        "abstract": "Mutual information (MI) is a fundamental measure of statistical dependence, with a myriad of applications to information theory, statistics, and machine learning. While it possesses many desirable structural properties, the estimation of high-dimensional MI from samples suffers from the curse of dimensionality. Motivated by statistical scalability to high dimensions, this paper proposes sliced MI (SMI) as a surrogate measure of dependence. SMI is defined as an average of MI terms between one-dimensional random projections. We show that it preserves many of the structural properties of classic MI, while gaining scalable computation and efficient estimation from samples. Furthermore, and in contrast to classic MI, SMI can grow as a result of deterministic transformations. This enables leveraging SMI for feature extraction by optimizing it over processing functions of raw data to identify useful representations thereof. Our theory is supported by numerical studies of independence testing and feature extraction, which demonstrate the potential gains SMI offers over classic MI for high-dimensional inference."
                    },
                    {
                        "title": "Sharper rates of convergence for the tensor graphical Lasso estimator",
                        "abstract": "Many modern datasets exhibit dependencies among observations as well as variables. This gives rise to the challenging problem of analyzing high-dimensional matrix-variate data with unknown dependence structures. To address this challenge, Kalaitzis et. al. (2013) proposed the Bigraphical Lasso (BiGLasso), an estimator for precision matrices of matrix-normals based on the Cartesian product of graphs. Subsequently, Greenewald, Zhou and Hero (GZH 2019) introduced a multiway tensor generalization of the BiGLasso estimator, known as the TeraLasso estimator. In this paper, we provide sharper rates of convergence in the Frobenius and operator norm for both BiGLasso and TeraLasso estimators for estimating inverse covariance matrices. This improves upon the rates presented in GZH 2019. In particular, (a) we strengthen the bounds for the relative errors in the operator and Frobenius norm by a factor of approximately $\\log p$; (b) Crucially, this improvement allows for finite-sample estimation errors in both norms to be derived for the two-way Kronecker sum model. The two-way regime is important because it is the setting that is the most theoretically challenging, and simultaneously the most common in applications. Normality is not needed in our proofs; instead, we consider sub-gaussian ensembles and derive tight concentration of measure bounds, using tensor unfolding techniques. The proof techniques may be of independent interest."
                    },
                    {
                        "title": "Regularized Block Toeplitz Covariance Matrix Estimation via Kronecker Product Expansions",
                        "abstract": "In this work we consider the estimation of spatio-temporal covariance matrices in the low sample non-Gaussian regime. We impose covariance structure in the form of a sum of Kronecker products decomposition (Tsiligkaridis et al. 2013, Greenewald et al. 2013) with diagonal correction (Greenewald et al.), which we refer to as DC-KronPCA, in the estimation of multiframe covariance matrices. This paper extends the approaches of (Tsiligkaridis et al.) in two directions. First, we modify the diagonally corrected method of (Greenewald et al.) to include a block Toeplitz constraint imposing temporal stationarity structure. Second, we improve the conditioning of the estimate in the very low sample regime by using Ledoit-Wolf type shrinkage regularization similar to (Chen, Hero et al. 2010). For improved robustness to heavy tailed distributions, we modify the KronPCA to incorporate robust shrinkage estimation (Chen, Hero et al. 2011). Results of numerical simulations establish benefits in terms of estimation MSE when compared to previous methods. Finally, we apply our methods to a real-world network spatio-temporal anomaly detection problem and achieve superior results."
                    },
                    {
                        "title": "Robust Kronecker Product PCA for Spatio-Temporal Covariance Estimation",
                        "abstract": "Kronecker PCA involves the use of a space vs. time Kronecker product decomposition to estimate spatio-temporal covariances. In this work the addition of a sparse correction factor is considered, which corresponds to a model of the covariance as a sum of Kronecker products of low (separation) rank and a sparse matrix. This sparse correction extends the diagonally corrected Kronecker PCA of [Greenewald et al 2013, 2014] to allow for sparse unstructured \"outliers\" anywhere in the covariance matrix, e.g. arising from variables or correlations that do not fit the Kronecker model well, or from sources such as sensor noise or sensor failure. We introduce a robust PCA-based algorithm to estimate the covariance under this model, extending the rearranged nuclear norm penalized LS Kronecker PCA approaches of [Greenewald et al 2014, Tsiligkaridis et al 2013]. An extension to Toeplitz temporal factors is also provided, producing a parameter reduction for temporally stationary measurement modeling. High dimensional MSE performance bounds are given for these extensions. Finally, the proposed extension of KronPCA is evaluated on both simulated and real data coming from yeast cell cycle experiments. This establishes the practical utility of robust Kronecker PCA in biological and other applications."
                    },
                    {
                        "title": "Gaussian-Smooth Optimal Transport: Metric Structure and Statistical Efficiency",
                        "abstract": "Optimal transport (OT), and in particular the Wasserstein distance, has seen a surge of interest and applications in machine learning. However, empirical approximation under Wasserstein distances suffers from a severe curse of dimensionality, rendering them impractical in high dimensions. As a result, entropically regularized OT has become a popular workaround. However, while it enjoys fast algorithms and better statistical properties, it looses the metric structure that Wasserstein distances enjoy. This work proposes a novel Gaussian-smoothed OT (GOT) framework, that achieves the best of both worlds: preserving the 1-Wasserstein metric structure while alleviating the empirical approximation curse of dimensionality. Furthermore, as the Gaussian-smoothing parameter shrinks to zero, GOT $\\Gamma$-converges towards classic OT (with convergence of optimizers), thus serving as a natural extension. An empirical study that supports the theoretical results is provided, promoting Gaussian-smoothed OT as a powerful alternative to entropic OT."
                    },
                    {
                        "title": "Kronecker STAP and SAR GMTI",
                        "abstract": "As a high resolution radar imaging modality, SAR detects and localizes non-moving targets accurately, giving it an advantage over lower resolution GMTI radars. Moving target detection is more challenging due to target smearing and masking by clutter. Space-time adaptive processing (STAP) is often used on multiantenna SAR to remove the stationary clutter and enhance the moving targets. In (Greenewald et al., 2016) it was shown that the performance of STAP can be improved by modeling the clutter covariance as a space vs. time Kronecker product with low rank factors, providing robustness and reducing the number of training samples required. In this work, we present a massively parallel algorithm for implementing Kronecker product STAP, enabling application to very large SAR datasets (such as the 2006 Gotcha data collection) using GPUs. Finally, we develop an extension of Kronecker STAP that uses information from multiple passes to improve moving target detection."
                    },
                    {
                        "title": "Nonstationary Distance Metric Learning",
                        "abstract": "Recent work in distance metric learning has focused on learning transformations of data that best align with provided sets of pairwise similarity and dissimilarity constraints. The learned transformations lead to improved retrieval, classification, and clustering algorithms due to the better adapted distance or similarity measures. Here, we introduce the problem of learning these transformations when the underlying constraint generation process is nonstationary. This nonstationarity can be due to changes in either the ground-truth clustering used to generate constraints or changes to the feature subspaces in which the class structure is apparent. We propose and evaluate COMID-SADL, an adaptive, online approach for learning and tracking optimal metrics as they change over time that is highly robust to a variety of nonstationary behaviors in the changing metric. We demonstrate COMID-SADL on both real and synthetic data sets and show significant performance improvements relative to previously proposed batch and online distance metric learning algorithms."
                    },
                    {
                        "title": "Robust SAR STAP via Kronecker Decomposition",
                        "abstract": "This paper proposes a spatio-temporal decomposition for the detection of moving targets in multiantenna SAR. As a high resolution radar imaging modality, SAR detects and localizes non-moving targets accurately, giving it an advantage over lower resolution GMTI radars. Moving target detection is more challenging due to target smearing and masking by clutter. Space-time adaptive processing (STAP) is often used to remove the stationary clutter and enhance the moving targets. In this work, it is shown that the performance of STAP can be improved by modeling the clutter covariance as a space vs. time Kronecker product with low rank factors. Based on this model, a low-rank Kronecker product covariance estimation algorithm is proposed, and a novel separable clutter cancelation filter based on the Kronecker covariance estimate is introduced. The proposed method provides orders of magnitude reduction in the required number of training samples, as well as improved robustness to corruption of the training data. Simulation results and experiments using the Gotcha SAR GMTI challenge dataset are presented that confirm the advantages of our approach relative to existing techniques."
                    },
                    {
                        "title": "High-Dimensional Feature Selection for Sample Efficient Treatment Effect Estimation",
                        "abstract": "The estimation of causal treatment effects from observational data is a fundamental problem in causal inference. To avoid bias, the effect estimator must control for all confounders. Hence practitioners often collect data for as many covariates as possible to raise the chances of including the relevant confounders. While this addresses the bias, this has the side effect of significantly increasing the number of data samples required to accurately estimate the effect due to the increased dimensionality. In this work, we consider the setting where out of a large number of covariates $X$ that satisfy strong ignorability, an unknown sparse subset $S$ is sufficient to include to achieve zero bias, i.e. $c$-equivalent to $X$. We propose a common objective function involving outcomes across treatment cohorts with nonconvex joint sparsity regularization that is guaranteed to recover $S$ with high probability under a linear outcome model for $Y$ and subgaussian covariates for each of the treatment cohort. This improves the effect estimation sample complexity so that it scales with the cardinality of the sparse subset $S$ and $\\log |X|$, as opposed to the cardinality of the full set $X$. We validate our approach with experiments on treatment effect estimation."
                    },
                    {
                        "title": "Domain Adaptable Prescriptive AI Agent for Enterprise",
                        "abstract": "Despite advancements in causal inference and prescriptive AI, its adoption in enterprise settings remains hindered primarily due to its technical complexity. Many users lack the necessary knowledge and appropriate tools to effectively leverage these technologies. This work at the MIT-IBM Watson AI Lab focuses on developing the proof-of-concept agent, PrecAIse, a domain-adaptable conversational agent equipped with a suite of causal and prescriptive tools to help enterprise users make better business decisions. The objective is to make advanced, novel causal inference and prescriptive tools widely accessible through natural language interactions. The presented Natural Language User Interface (NLUI) enables users with limited expertise in machine learning and data science to harness prescriptive analytics in their decision-making processes without requiring intensive computing resources. We present an agent capable of function calling, maintaining faithful, interactive, and dynamic conversations, and supporting new domains."
                    },
                    {
                        "title": "Time-dependent spatially varying graphical models, with application to brain fMRI data analysis",
                        "abstract": "In this work, we present an additive model for space-time data that splits the data into a temporally correlated component and a spatially correlated component. We model the spatially correlated portion using a time-varying Gaussian graphical model. Under assumptions on the smoothness of changes in covariance matrices, we derive strong single sample convergence results, confirming our ability to estimate meaningful graphical structures as they evolve over time. We apply our methodology to the discovery of time-varying spatial structures in human brain fMRI signals."
                    },
                    {
                        "title": "Tensor Graphical Lasso (TeraLasso)",
                        "abstract": "This paper introduces a multi-way tensor generalization of the Bigraphical Lasso (BiGLasso), which uses a two-way sparse Kronecker-sum multivariate-normal model for the precision matrix to parsimoniously model conditional dependence relationships of matrix-variate data based on the Cartesian product of graphs. We call this generalization the {\\bf Te}nsor g{\\bf ra}phical Lasso (TeraLasso). We demonstrate using theory and examples that the TeraLasso model can be accurately and scalably estimated from very limited data samples of high dimensional variables with multiway coordinates such as space, time and replicates. Statistical consistency and statistical rates of convergence are established for both the BiGLasso and TeraLasso estimators of the precision matrix and estimators of its support (non-sparsity) set, respectively. We propose a scalable composite gradient descent algorithm and analyze the computational convergence rate, showing that the composite gradient descent algorithm is guaranteed to converge at a geometric rate to the global minimizer of the TeraLasso objective function. Finally, we illustrate the TeraLasso using both simulation and experimental data from a meteorological dataset, showing that we can accurately estimate precision matrices and recover meaningful conditional dependency graphs from high dimensional complex datasets."
                    },
                    {
                        "title": "k-Sliced Mutual Information: A Quantitative Study of Scalability with Dimension",
                        "abstract": "Sliced mutual information (SMI) is defined as an average of mutual information (MI) terms between one-dimensional random projections of the random variables. It serves as a surrogate measure of dependence to classic MI that preserves many of its properties but is more scalable to high dimensions. However, a quantitative characterization of how SMI itself and estimation rates thereof depend on the ambient dimension, which is crucial to the understanding of scalability, remain obscure. This work provides a multifaceted account of the dependence of SMI on dimension, under a broader framework termed $k$-SMI, which considers projections to $k$-dimensional subspaces. Using a new result on the continuity of differential entropy in the 2-Wasserstein metric, we derive sharp bounds on the error of Monte Carlo (MC)-based estimates of $k$-SMI, with explicit dependence on $k$ and the ambient dimension, revealing their interplay with the number of samples. We then combine the MC integrator with the neural estimation framework to provide an end-to-end $k$-SMI estimator, for which optimal convergence rates are established. We also explore asymptotics of the population $k$-SMI as dimension grows, providing Gaussian approximation results with a residual that decays under appropriate moment bounds. All our results trivially apply to SMI by setting $k=1$. Our theory is validated with numerical experiments and is applied to sliced InfoGAN, which altogether provide a comprehensive quantitative account of the scalability question of $k$-SMI, including SMI as a special case when $k=1$."
                    },
                    {
                        "title": "Dynamic Metric Learning from Pairwise Comparisons",
                        "abstract": "Recent work in distance metric learning has focused on learning transformations of data that best align with specified pairwise similarity and dissimilarity constraints, often supplied by a human observer. The learned transformations lead to improved retrieval, classification, and clustering algorithms due to the better adapted distance or similarity measures. Here, we address the problem of learning these transformations when the underlying constraint generation process is nonstationary. This nonstationarity can be due to changes in either the ground-truth clustering used to generate constraints or changes in the feature subspaces in which the class structure is apparent. We propose Online Convex Ensemble StrongLy Adaptive Dynamic Learning (OCELAD), a general adaptive, online approach for learning and tracking optimal metrics as they change over time that is highly robust to a variety of nonstationary behaviors in the changing metric. We apply the OCELAD framework to an ensemble of online learners. Specifically, we create a retro-initialized composite objective mirror descent (COMID) ensemble (RICE) consisting of a set of parallel COMID learners with different learning rates, demonstrate RICE-OCELAD on both real and synthetic data sets and show significant performance improvements relative to previously proposed batch and online distance metric learning algorithms."
                    },
                    {
                        "title": "Action Centered Contextual Bandits",
                        "abstract": "Contextual bandits have become popular as they offer a middle ground between very simple approaches based on multi-armed bandits and very complex approaches using the full power of reinforcement learning. They have demonstrated success in web applications and have a rich body of associated theoretical guarantees. Linear models are well understood theoretically and preferred by practitioners because they are not only easily interpretable but also simple to implement and debug. Furthermore, if the linear model is true, we get very strong performance guarantees. Unfortunately, in emerging applications in mobile health, the time-invariant linear model assumption is untenable. We provide an extension of the linear model for contextual bandits that has two parts: baseline reward and treatment effect. We allow the former to be complex but keep the latter simple. We argue that this model is plausible for mobile health applications. At the same time, it leads to algorithms with strong performance guarantees as in the linear model setting, while still allowing for complex nonlinear baseline modeling. Our theory is supported by experiments on data gathered in a recently concluded mobile health study."
                    },
                    {
                        "title": "k-Mixup Regularization for Deep Learning via Optimal Transport",
                        "abstract": "Mixup is a popular regularization technique for training deep neural networks that improves generalization and increases robustness to certain distribution shifts. It perturbs input training data in the direction of other randomly-chosen instances in the training set. To better leverage the structure of the data, we extend mixup in a simple, broadly applicable way to \\emph{$k$-mixup}, which perturbs $k$-batches of training points in the direction of other $k$-batches. The perturbation is done with displacement interpolation, i.e. interpolation under the Wasserstein metric. We demonstrate theoretically and in simulations that $k$-mixup preserves cluster and manifold structures, and we extend theory studying the efficacy of standard mixup to the $k$-mixup case. Our empirical results show that training with $k$-mixup further improves generalization and robustness across several network architectures and benchmark datasets of differing modalities. For the wide variety of real datasets considered, the performance gains of $k$-mixup over standard mixup are similar to or larger than the gains of mixup itself over standard ERM after hyperparameter optimization. In several instances, in fact, $k$-mixup achieves gains in settings where standard mixup has negligible to zero improvement over ERM."
                    },
                    {
                        "title": "Improving Approximate Optimal Transport Distances using Quantization",
                        "abstract": "Optimal transport (OT) is a popular tool in machine learning to compare probability measures geometrically, but it comes with substantial computational burden. Linear programming algorithms for computing OT distances scale cubically in the size of the input, making OT impractical in the large-sample regime. We introduce a practical algorithm, which relies on a quantization step, to estimate OT distances between measures given cheap sample access. We also provide a variant of our algorithm to improve the performance of approximate solvers, focusing on those for entropy-regularized transport. We give theoretical guarantees on the benefits of this quantization step and display experiments showing that it behaves well in practice, providing a practical approximation algorithm that can be used as a drop-in replacement for existing OT estimators."
                    },
                    {
                        "title": "Personalized HeartSteps: A Reinforcement Learning Algorithm for Optimizing Physical Activity",
                        "abstract": "With the recent evolution of mobile health technologies, health scientists are increasingly interested in developing just-in-time adaptive interventions (JITAIs), typically delivered via notification on mobile device and designed to help the user prevent negative health outcomes and promote the adoption and maintenance of healthy behaviors. A JITAI involves a sequence of decision rules (i.e., treatment policy) that takes the user's current context as input and specifies whether and what type of an intervention should be provided at the moment. In this paper, we develop a Reinforcement Learning (RL) algorithm that continuously learns and improves the treatment policy embedded in the JITAI as the data is being collected from the user. This work is motivated by our collaboration on designing the RL algorithm in HeartSteps V2 based on data from HeartSteps V1. HeartSteps is a physical activity mobile health application. The RL algorithm developed in this paper is being used in HeartSteps V2 to decide, five times per day, whether to deliver a context-tailored activity suggestion."
                    }
                ]
            },
            "b6967490-c549-40da-adc1-4c1d130e4bd4": {
                "pk": "b6967490-c549-40da-adc1-4c1d130e4bd4",
                "name": "Jiri Navratil",
                "collaborators": [
                    "Matthew Arnold",
                    "Benjamin Elder",
                    "Soumya Ghosh",
                    "Prasanna Sattigeri",
                    "Anupama Murthi",
                    "Rishav Chakravarti",
                    "Cicero Nogueira dos Santos",
                    "Alan King",
                    "Jesus Rios",
                    "Georgios Kollias"
                ],
                "domain": [
                    "Uncertainty Quantification",
                    "Deep Learning",
                    "Performance Prediction",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Uncertainty Prediction for Deep Sequential Regression Using Meta Models",
                        "abstract": "Generating high quality uncertainty estimates for sequential regression, particularly deep recurrent networks, remains a challenging and open problem. Existing approaches often make restrictive assumptions (such as stationarity) yet still perform poorly in practice, particularly in presence of real world non-stationary signals and drift. This paper describes a flexible method that can generate symmetric and asymmetric uncertainty estimates, makes no assumptions about stationarity, and outperforms competitive baselines on both drift and non drift scenarios. This work helps make sequential regression more effective and practical for use in real-world applications, and is a powerful new addition to the modeling toolbox for sequential uncertainty quantification in general."
                    },
                    {
                        "title": "Uncertainty Characteristics Curves: A Systematic Assessment of Prediction Intervals",
                        "abstract": "Accurate quantification of model uncertainty has long been recognized as a fundamental requirement for trusted AI. In regression tasks, uncertainty is typically quantified using prediction intervals calibrated to a specific operating point, making evaluation and comparison across different studies difficult. Our work leverages: (1) the concept of operating characteristics curves and (2) the notion of a gain over a simple reference, to derive a novel operating point agnostic assessment methodology for prediction intervals. The paper describes the corresponding algorithm, provides a theoretical analysis, and demonstrates its utility in multiple scenarios. We argue that the proposed method addresses the current need for comprehensive assessment of prediction intervals and thus represents a valuable addition to the uncertainty quantification toolbox."
                    },
                    {
                        "title": "Assessment of Prediction Intervals Using Uncertainty Characteristics Curves",
                        "abstract": "Accurate quantification of model uncertainty has long been recognized as a fundamental requirement for trusted AI. In regression tasks, uncertainty is typically quantified using prediction intervals calibrated to an ad-hoc operating point, making evaluation and comparison across different studies relatively difficult. Our work leverages: (1) the concept of operating characteristics curves and (2) the notion of a gain over a null reference, to derive a novel operating point agnostic assessment methodology for prediction intervals. The paper defines the Uncertainty Characteristics Curve and demonstrates its utility in selected scenarios. We argue that the proposed method addresses the current need for comprehensive assessment of prediction intervals and thus represents a valuable addition to the uncertainty quantification toolbox."
                    },
                    {
                        "title": "Learning Prediction Intervals for Model Performance",
                        "abstract": "Understanding model performance on unlabeled data is a fundamental challenge of developing, deploying, and maintaining AI systems. Model performance is typically evaluated using test sets or periodic manual quality assessments, both of which require laborious manual data labeling. Automated performance prediction techniques aim to mitigate this burden, but potential inaccuracy and a lack of trust in their predictions has prevented their widespread adoption. We address this core problem of performance prediction uncertainty with a method to compute prediction intervals for model performance. Our methodology uses transfer learning to train an uncertainty model to estimate the uncertainty of model performance predictions. We evaluate our approach across a wide range of drift conditions and show substantial improvement over competitive baselines. We believe this result makes prediction intervals, and performance prediction in general, significantly more practical for real-world use."
                    },
                    {
                        "title": "Improved Answer Selection with Pre-Trained Word Embeddings",
                        "abstract": "This paper evaluates existing and newly proposed answer selection methods based on pre-trained word embeddings. Word embeddings are highly effective in various natural language processing tasks and their integration into traditional information retrieval (IR) systems allows for the capture of semantic relatedness between questions and answers. Empirical results on three publicly available data sets show significant gains over traditional term frequency based approaches in both supervised and unsupervised settings. We show that combining these word embedding features with traditional learning-to-rank techniques can achieve similar performance to state-of-the-art neural networks trained for the answer selection task."
                    },
                    {
                        "title": "Accelerating Physics-Based Simulations Using Neural Network Proxies: An Application in Oil Reservoir Modeling",
                        "abstract": "We develop a proxy model based on deep learning methods to accelerate the simulations of oil reservoirs--by three orders of magnitude--compared to industry-strength physics-based PDE solvers. This paper describes a new architectural approach to this task, accompanied by a thorough experimental evaluation on a publicly available reservoir model. We demonstrate that in a practical setting a speedup of more than 2000X can be achieved with an average sequence error of about 10\\% relative to the oil-field simulator. The proxy model is contrasted with a high-quality physics-based acceleration baseline and is shown to outperform it by several orders of magnitude. We believe the outcomes presented here are extremely promising and offer a valuable benchmark for continuing research in oil field development optimization. Due to its domain-agnostic architecture, the presented approach can be extended to many applications beyond the field of oil and gas exploration."
                    },
                    {
                        "title": "Not Your Grandfathers Test Set: Reducing Labeling Effort for Testing",
                        "abstract": "Building and maintaining high-quality test sets remains a laborious and expensive task. As a result, test sets in the real world are often not properly kept up to date and drift from the production traffic they are supposed to represent. The frequency and severity of this drift raises serious concerns over the value of manually labeled test sets in the QA process. This paper proposes a simple but effective technique that drastically reduces the effort needed to construct and maintain a high-quality test set (reducing labeling effort by 80-100% across a range of practical scenarios). This result encourages a fundamental rethinking of the testing process by both practitioners, who can use these techniques immediately to improve their testing, and researchers who can help address many of the open questions raised by this new approach."
                    },
                    {
                        "title": "Towards Automating the AI Operations Lifecycle",
                        "abstract": "Today's AI deployments often require significant human involvement and skill in the operational stages of the model lifecycle, including pre-release testing, monitoring, problem diagnosis and model improvements. We present a set of enabling technologies that can be used to increase the level of automation in AI operations, thus lowering the human effort required. Since a common source of human involvement is the need to assess the performance of deployed models, we focus on technologies for performance prediction and KPI analysis and show how they can be used to improve automation in the key stages of a typical AI operations pipeline."
                    },
                    {
                        "title": "Uncertainty Quantification 360: A Holistic Toolkit for Quantifying and Communicating the Uncertainty of AI",
                        "abstract": "In this paper, we describe an open source Python toolkit named Uncertainty Quantification 360 (UQ360) for the uncertainty quantification of AI models. The goal of this toolkit is twofold: first, to provide a broad range of capabilities to streamline as well as foster the common practices of quantifying, evaluating, improving, and communicating uncertainty in the AI application development lifecycle; second, to encourage further exploration of UQ's connections to other pillars of trustworthy AI such as fairness and transparency through the dissemination of latest research and education materials. Beyond the Python package (\\url{https://github.com/IBM/UQ360}), we have developed an interactive experience (\\url{http://uq360.mybluemix.net}) and guidance materials as educational tools to aid researchers and developers in producing and communicating high-quality uncertainties in an effective manner."
                    }
                ]
            },
            "326d9db7-fac1-4acb-b3f7-e2def9f7187b": {
                "pk": "326d9db7-fac1-4acb-b3f7-e2def9f7187b",
                "name": "Jerret Ross",
                "collaborators": [
                    "Youssef Mroueh",
                    "Igor Melnyk",
                    "Payel Das",
                    "Brian Belgodere",
                    "Pierre Dognin",
                    "Inkit Padhi",
                    "Mattia Rigotti",
                    "Mingrui Liu",
                    "Wei Zhang",
                    "Xiaodong Cui"
                ],
                "domain": [
                    "Generative Models",
                    "Machine Learning",
                    "Neural Networks",
                    "Synthetic Data"
                ],
                "publications": [
                    {
                        "title": "Fast Mixing of Multi-Scale Langevin Dynamics under the Manifold Hypothesis",
                        "abstract": "Recently, the task of image generation has attracted much attention. In particular, the recent empirical successes of the Markov Chain Monte Carlo (MCMC) technique of Langevin Dynamics have prompted a number of theoretical advances; despite this, several outstanding problems remain. First, the Langevin Dynamics is run in very high dimension on a nonconvex landscape; in the worst case, due to the NP-hardness of nonconvex optimization, it is thought that Langevin Dynamics mixes only in time exponential in the dimension. In this work, we demonstrate how the manifold hypothesis allows for the considerable reduction of mixing time, from exponential in the ambient dimension to depending only on the (much smaller) intrinsic dimension of the data. Second, the high dimension of the sampling space significantly hurts the performance of Langevin Dynamics; we leverage a multi-scale approach to help ameliorate this issue and observe that this multi-resolution algorithm allows for a trade-off between image quality and computational expense in generation."
                    },
                    {
                        "title": "Wasserstein Barycenter Model Ensembling",
                        "abstract": "In this paper we propose to perform model ensembling in a multiclass or a multilabel learning setting using Wasserstein (W.) barycenters. Optimal transport metrics, such as the Wasserstein distance, allow incorporating semantic side information such as word embeddings. Using W. barycenters to find the consensus between models allows us to balance confidence and semantics in finding the agreement between the models. We show applications of Wasserstein ensembling in attribute-based classification, multilabel learning and image captioning generation. These results show that the W. ensembling is a viable alternative to the basic geometric or arithmetic mean ensembling."
                    },
                    {
                        "title": "Towards Better Understanding of Adaptive Gradient Algorithms in Generative Adversarial Nets",
                        "abstract": "Adaptive gradient algorithms perform gradient-based updates using the history of gradients and are ubiquitous in training deep neural networks. While adaptive gradient methods theory is well understood for minimization problems, the underlying factors driving their empirical success in min-max problems such as GANs remain unclear. In this paper, we aim at bridging this gap from both theoretical and empirical perspectives. First, we analyze a variant of Optimistic Stochastic Gradient (OSG) proposed in~\\citep{daskalakis2017training} for solving a class of non-convex non-concave min-max problem and establish $O(\\epsilon^{-4})$ complexity for finding $\\epsilon$-first-order stationary point, in which the algorithm only requires invoking one stochastic first-order oracle while enjoying state-of-the-art iteration complexity achieved by stochastic extragradient method by~\\citep{iusem2017extragradient}. Then we propose an adaptive variant of OSG named Optimistic Adagrad (OAdagrad) and reveal an \\emph{improved} adaptive complexity $O\\left(\\epsilon^{-\\frac{2}{1-\\alpha}}\\right)$, where $\\alpha$ characterizes the growth rate of the cumulative stochastic gradient and $0\\leq \\alpha\\leq 1/2$. To the best of our knowledge, this is the first work for establishing adaptive complexity in non-convex non-concave min-max optimization. Empirically, our experiments show that indeed adaptive gradient algorithms outperform their non-adaptive counterparts in GAN training. Moreover, this observation can be explained by the slow growth rate of the cumulative stochastic gradient, as observed empirically."
                    },
                    {
                        "title": "A Decentralized Parallel Algorithm for Training Generative Adversarial Nets",
                        "abstract": "Generative Adversarial Networks (GANs) are a powerful class of generative models in the deep learning community. Current practice on large-scale GAN training utilizes large models and distributed large-batch training strategies, and is implemented on deep learning frameworks (e.g., TensorFlow, PyTorch, etc.) designed in a centralized manner. In the centralized network topology, every worker needs to either directly communicate with the central node or indirectly communicate with all other workers in every iteration. However, when the network bandwidth is low or network latency is high, the performance would be significantly degraded. Despite recent progress on decentralized algorithms for training deep neural networks, it remains unclear whether it is possible to train GANs in a decentralized manner. The main difficulty lies at handling the nonconvex-nonconcave min-max optimization and the decentralized communication simultaneously. In this paper, we address this difficulty by designing the \\textbf{first gradient-based decentralized parallel algorithm} which allows workers to have multiple rounds of communications in one iteration and to update the discriminator and generator simultaneously, and this design makes it amenable for the convergence analysis of the proposed decentralized algorithm. Theoretically, our proposed decentralized algorithm is able to solve a class of non-convex non-concave min-max problems with provable non-asymptotic convergence to first-order stationary point. Experimental results on GANs demonstrate the effectiveness of the proposed algorithm."
                    },
                    {
                        "title": "Large-Scale Chemical Language Representations Capture Molecular Structure and Properties",
                        "abstract": "Models based on machine learning can enable accurate and fast molecular property predictions, which is of interest in drug discovery and material design. Various supervised machine learning models have demonstrated promising performance, but the vast chemical space and the limited availability of property labels make supervised learning challenging. Recently, unsupervised transformer-based language models pretrained on a large unlabelled corpus have produced state-of-the-art results in many downstream natural language processing tasks. Inspired by this development, we present molecular embeddings obtained by training an efficient transformer encoder model, MoLFormer, which uses rotary positional embeddings. This model employs a linear attention mechanism, coupled with highly distributed training, on SMILES sequences of 1.1 billion unlabelled molecules from the PubChem and ZINC datasets. We show that the learned molecular representation outperforms existing baselines, including supervised and self-supervised graph neural networks and language models, on several downstream tasks from ten benchmark datasets. They perform competitively on two others. Further analyses, specifically through the lens of attention, demonstrate that MoLFormer trained on chemical SMILES indeed learns the spatial relationships between atoms within a molecule. These results provide encouraging evidence that large-scale molecular language models can capture sufficient chemical and structural information to predict various distinct molecular properties, including quantum-chemical properties."
                    },
                    {
                        "title": "GP-MoLFormer: A Foundation Model For Molecular Generation",
                        "abstract": "Transformer-based models trained on large and general purpose datasets consisting of molecular strings have recently emerged as a powerful tool for successfully modeling various structure-property relations. Inspired by this success, we extend the paradigm of training chemical language transformers on large-scale chemical datasets to generative tasks in this work. Specifically, we propose GP-MoLFormer, an autoregressive molecular string generator that is trained on more than 1.1B chemical SMILES. GP-MoLFormer uses a 46.8M parameter transformer decoder model with linear attention and rotary positional encodings as the base architecture. We explore the utility of GP-MoLFormer in generating novel, valid, and unique SMILES. Impressively, we find GP-MoLFormer is able to generate a significant fraction of novel, valid, and unique SMILES even when the number of generated molecules is in the 10 billion range and the reference set is over a billion. We also find strong memorization of training data in GP-MoLFormer generations, which has so far remained unexplored for chemical language models. Our analyses reveal that training data memorization and novelty in generations are impacted by the quality of the training data; duplication bias in training data can enhance memorization at the cost of lowering novelty. We evaluate GP-MoLFormer's utility and compare it with that of existing baselines on three different tasks: de novo generation, scaffold-constrained molecular decoration, and unconstrained property-guided optimization. While the first two are handled with no additional training, we propose a parameter-efficient fine-tuning method for the last task, which uses property-ordered molecular pairs as input. We call this new approach pair-tuning. Our results show GP-MoLFormer performs better or comparable with baselines across all three tasks, demonstrating its general utility."
                    },
                    {
                        "title": "Tabular Transformers for Modeling Multivariate Time Series",
                        "abstract": "Tabular datasets are ubiquitous in data science applications. Given their importance, it seems natural to apply state-of-the-art deep learning algorithms in order to fully unlock their potential. Here we propose neural network models that represent tabular time series that can optionally leverage their hierarchical structure. This results in two architectures for tabular time series: one for learning representations that is analogous to BERT and can be pre-trained end-to-end and used in downstream tasks, and one that is akin to GPT and can be used for generation of realistic synthetic tabular sequences. We demonstrate our models on two datasets: a synthetic credit card transaction dataset, where the learned representations are used for fraud detection and synthetic data generation, and on a real pollution dataset, where the learned encodings are used to predict atmospheric pollutant concentrations. Code and data are available at https://github.com/IBM/TabFormer."
                    },
                    {
                        "title": "Risk Aware Benchmarking of Large Language Models",
                        "abstract": "We propose a distributional framework for benchmarking socio-technical risks of foundation models with quantified statistical significance. Our approach hinges on a new statistical relative testing based on first and second order stochastic dominance of real random variables. We show that the second order statistics in this test are linked to mean-risk models commonly used in econometrics and mathematical finance to balance risk and utility when choosing between alternatives. Using this framework, we formally develop a risk-aware approach for foundation model selection given guardrails quantified by specified metrics. Inspired by portfolio optimization and selection theory in mathematical finance, we define a metrics portfolio for each model as a means to aggregate a collection of metrics, and perform model selection based on the stochastic dominance of these portfolios. The statistical significance of our tests is backed theoretically by an asymptotic analysis via central limit theorems instantiated in practice via a bootstrap variance estimate. We use our framework to compare various large language models regarding risks related to drifting from instructions and outputting toxic content."
                    },
                    {
                        "title": "Auditing and Generating Synthetic Data with Controllable Trust Trade-offs",
                        "abstract": "Real-world data often exhibits bias, imbalance, and privacy risks. Synthetic datasets have emerged to address these issues. This paradigm relies on generative AI models to generate unbiased, privacy-preserving data while maintaining fidelity to the original data. However, assessing the trustworthiness of synthetic datasets and models is a critical challenge. We introduce a holistic auditing framework that comprehensively evaluates synthetic datasets and AI models. It focuses on preventing bias and discrimination, ensures fidelity to the source data, assesses utility, robustness, and privacy preservation. We demonstrate the framework's effectiveness by auditing various generative models across diverse use cases like education, healthcare, banking, and human resources, spanning different data modalities such as tabular, time-series, vision, and natural language. This holistic assessment is essential for compliance with regulatory safeguards. We introduce a trustworthiness index to rank synthetic datasets based on their safeguards trade-offs. Furthermore, we present a trustworthiness-driven model selection and cross-validation process during training, exemplified with \"TrustFormers\" across various data types. This approach allows for controllable trustworthiness trade-offs in synthetic data creation. Our auditing framework fosters collaboration among stakeholders, including data scientists, governance experts, internal reviewers, external certifiers, and regulators. This transparent reporting should become a standard practice to prevent bias, discrimination, and privacy violations, ensuring compliance with policies and providing accountability, safety, and performance guarantees."
                    }
                ]
            }
        }
    },
    "2409.00729": {
        "paper_data": {
            "title": "ContextCite: Attributing Model Generation to Context",
            "url": "http://arxiv.org/abs/2409.00729v2",
            "arxiv_id": "2409.00729",
            "authors": [
                "Benjamin Cohen-Wang",
                "Harshay Shah",
                "Kristian Georgiev",
                "Aleksander Madry"
            ],
            "abstract": "How do language models use information provided as context when generating a response? Can we infer whether a particular generated statement is actually grounded in the context, a misinterpretation, or fabricated? To help answer these questions, we introduce the problem of context attribution: pinpointing the parts of the context (if any) that led a model to generate a particular statement. We then present ContextCite, a simple and scalable method for context attribution that can be applied on top of any existing language model. Finally, we showcase the utility of ContextCite through three applications: (1) helping verify generated statements (2) improving response quality by pruning the context and (3) detecting poisoning attacks. We provide code for ContextCite at https://github.com/MadryLab/context-cite.",
            "introduction": "   1 Introduction  Suppose that we would like to use a language model to learn about recent news. We would first need to provide it with relevant articles as context222Assistants like ChatGPT automatically retrieve such information as needed behind the scenes \\citepnakano2021webgpt,menick2022teaching,thoppilan2022lamda.. We would then expect the language model to interact with this context to answer questions. Upon seeing a generated response, we might ask: is everything accurate? Did the model misinterpret any of the context or fabricate anything? Is the response actually grounded in the provided context?   Answering these questions manually could be tedious\u2014we would need to first read the articles ourselves and then verify the statements. To automate this process, prior work has focused on teaching models to generate citations: references to parts of the context that support a response \\citepnakano2021webgpt,menick2022teaching,thoppilan2022lamda,gao2022rarr,gao2023enabling. They typically do so by explicitly training or prompting language models to produce citations.   In this work, we explore a different type of citation: instead of teaching a language model to cite its sources, can we directly identify the pieces of information that it actually uses? Specifically, we ask:   Can we pinpoint the parts of the context (if any) that led to a particular generated statement?   We refer to this problem as context attribution. Suppose, for example, that a language model misinterprets a piece of information and generates an inaccurate statement. In this case, context attribution would surface the misinterpreted part of the context. On the other hand, suppose that a language model uses knowledge that it learned from pre-training to generate a statement, rather than the context. In this case, context attribution would indicate this by not attributing the statement to any part of the context.   Unlike citations generated by language models, which can be difficult to validate \\citeprashkin2023measuring,liu2023evaluating, in principle it is easy to evaluate the efficacy of context attributions. Specifically, if a part of the context actually led to a particular generated response, then removing it should substantially affect this response.    1.1 Our contributions  Formalizing context attribution (Section\u00a02).  We begin this work by formalizing the task of context attribution. Specifically, a context attribution method assigns a score to each part of the context indicating the degree to which it is responsible for a given generated statement. We provide metrics for evaluating these scores, guided by the intuition that removing high-scoring parts of the context should have a greater effect than removing low-scoring parts of the context.    Performing context attribution with ContextCite (Sections\u00a03 and\u00a04).  Next, we present ContextCite, a simple and scalable method for context attribution that can be applied on top of any existing language model (see Figure 1). ContextCite learns a surrogate model that approximates how a language model\u2019s response is affected by including or excluding each part of the context. This methodology closely follows prior work on attributing model behavior to features \\citepribeiro2016why,lundberg2017unified,sokol2019blimey and training examples \\citepilyas2022datamodels,park2023trak. In the context attribution setting, we find that it is possible to learn a linear surrogate model that (1) faithfully models the language model\u2019s behavior and (2) can be efficiently estimated using a small number of additional inference passes. The weights of this surrogate model can be directly treated as attribution scores. We benchmark ContextCite against various baselines on a diverse set of generation tasks and find that",
            "references": [
                {
                    "title": "The Llama 3 Herd of Models",
                    "abstract": "Modern artificial intelligence (AI) systems are powered by foundation models. This paper presents a new set of foundation models, called Llama 3. It is a herd of language models that natively support multilinguality, coding, reasoning, and tool usage. Our largest model is a dense Transformer with 405B parameters and a context window of up to 128K tokens. This paper presents an extensive empirical evaluation of Llama 3. We find that Llama 3 delivers comparable quality to leading language models such as GPT-4 on a plethora of tasks. We publicly release Llama 3, including pre-trained and post-trained versions of the 405B parameter language model and our Llama Guard 3 model for input and output safety. The paper also presents the results of experiments in which we integrate image, video, and speech capabilities into Llama 3 via a compositional approach. We observe this approach performs competitively with the state-of-the-art on image, video, and speech recognition tasks. The resulting models are not yet being broadly released as they are still under development."
                },
                {
                    "title": "Model Internals-based Answer Attribution for Trustworthy Retrieval-Augmented Generation",
                    "abstract": "Ensuring the verifiability of model answers is a fundamental challenge for retrieval-augmented generation (RAG) in the question answering (QA) domain. Recently, self-citation prompting was proposed to make large language models (LLMs) generate citations to supporting documents along with their answers. However, self-citing LLMs often struggle to match the required format, refer to non-existent sources, and fail to faithfully reflect LLMs' context usage throughout the generation. In this work, we present MIRAGE --Model Internals-based RAG Explanations -- a plug-and-play approach using model internals for faithful answer attribution in RAG applications. MIRAGE detects context-sensitive answer tokens and pairs them with retrieved documents contributing to their prediction via saliency methods. We evaluate our proposed approach on a multilingual extractive QA dataset, finding high agreement with human answer attribution. On open-ended QA, MIRAGE achieves citation quality and efficiency comparable to self-citation while also allowing for a finer-grained control of attribution parameters. Our qualitative evaluation highlights the faithfulness of MIRAGE's attributions and underscores the promising application of model internals for RAG answer attribution."
                },
                {
                    "title": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
                    "abstract": "We introduce phi-3-mini, a 3.8 billion parameter language model trained on 3.3 trillion tokens, whose overall performance, as measured by both academic benchmarks and internal testing, rivals that of models such as Mixtral 8x7B and GPT-3.5 (e.g., phi-3-mini achieves 69% on MMLU and 8.38 on MT-bench), despite being small enough to be deployed on a phone. Our training dataset is a scaled-up version of the one used for phi-2, composed of heavily filtered publicly available web data and synthetic data. The model is also further aligned for robustness, safety, and chat format. We also provide parameter-scaling results with a 7B, 14B models trained for 4.8T tokens, called phi-3-small, phi-3-medium, both significantly more capable than phi-3-mini (e.g., respectively 75%, 78% on MMLU, and 8.7, 8.9 on MT-bench). To enhance multilingual, multimodal, and long-context capabilities, we introduce three models in the phi-3.5 series: phi-3.5-mini, phi-3.5-MoE, and phi-3.5-Vision. The phi-3.5-MoE, a 16 x 3.8B MoE model with 6.6 billion active parameters, achieves superior performance in language reasoning, math, and code tasks compared to other open-source models of similar scale, such as Llama 3.1 and the Mixtral series, and on par with Gemini-1.5-Flash and GPT-4o-mini. Meanwhile, phi-3.5-Vision, a 4.2 billion parameter model derived from phi-3.5-mini, excels in reasoning tasks and is adept at handling both single-image and text prompts, as well as multi-image and text prompts."
                },
                {
                    "title": "Decomposing and Editing Predictions by Modeling Model Computation",
                    "abstract": "How does the internal computation of a machine learning model transform inputs into predictions? In this paper, we introduce a task called component modeling that aims to address this question. The goal of component modeling is to decompose an ML model's prediction in terms of its components -- simple functions (e.g., convolution filters, attention heads) that are the\"building blocks\"of model computation. We focus on a special case of this task, component attribution, where the goal is to estimate the counterfactual impact of individual components on a given prediction. We then present COAR, a scalable algorithm for estimating component attributions; we demonstrate its effectiveness across models, datasets, and modalities. Finally, we show that component attributions estimated with COAR directly enable model editing across five tasks, namely: fixing model errors, ``forgetting'' specific classes, boosting subpopulation robustness, localizing backdoor attacks, and improving robustness to typographic attacks. We provide code for COAR at https://github.com/MadryLab/modelcomponents ."
                },
                {
                    "title": "Neural Exec: Learning (and Learning from) Execution Triggers for Prompt Injection Attacks",
                    "abstract": "We introduce a new family of prompt injection attacks, termed Neural Exec. Unlike known attacks that rely on handcrafted strings (e.g.,\"Ignore previous instructions and...\"), we show that it is possible to conceptualize the creation of execution triggers as a differentiable search problem and use learning-based methods to autonomously generate them. Our results demonstrate that a motivated adversary can forge triggers that are not only drastically more effective than current handcrafted ones but also exhibit inherent flexibility in shape, properties, and functionality. In this direction, we show that an attacker can design and generate Neural Execs capable of persisting through multi-stage preprocessing pipelines, such as in the case of Retrieval-Augmented Generation (RAG)-based applications. More critically, our findings show that attackers can produce triggers that deviate markedly in form and shape from any known attack, sidestepping existing blacklist-based detection and sanitation approaches."
                },
                {
                    "title": "Backtracing: Retrieving the Cause of the Query",
                    "abstract": "Many online content portals allow users to ask questions to supplement their understanding (e.g., of lectures). While information retrieval (IR) systems may provide answers for such user queries, they do not directly assist content creators\u2014such as lecturers who want to improve their content\u2014identify segments that caused a user to ask those questions.We introduce the task of backtracing, in which systems retrieve the text segment that most likely caused a user query.We formalize three real-world domains for which backtracing is important in improving content delivery and communication: understanding the cause of (a) student confusion in the Lecture domain, (b) reader curiosity in the News Article domain, and (c) user emotion in the Conversation domain.We evaluate the zero-shot performance of popular information retrieval methods and language modeling methods, including bi-encoder, re-ranking and likelihood-based methods and ChatGPT.While traditional IR systems retrieve semantically relevant information (e.g., details on \u201cprojection matrices\u201d for a query \u201cdoes projecting multiple times still lead to the same point?\u201d), they often miss the causally relevant context (e.g., the lecturer states \u201cprojecting twice gets me the same answer as one projection\u201d). Our results show that there is room for improvement on backtracing and it requires new retrieval approaches.We hope our benchmark serves to improve future retrieval systems for backtracing, spawning systems that refine content generation and identify linguistic triggers influencing user queries."
                },
                {
                    "title": "AtP*: An efficient and scalable method for localizing LLM behaviour to components",
                    "abstract": "Activation Patching is a method of directly computing causal attributions of behavior to model components. However, applying it exhaustively requires a sweep with cost scaling linearly in the number of model components, which can be prohibitively expensive for SoTA Large Language Models (LLMs). We investigate Attribution Patching (AtP), a fast gradient-based approximation to Activation Patching and find two classes of failure modes of AtP which lead to significant false negatives. We propose a variant of AtP called AtP*, with two changes to address these failure modes while retaining scalability. We present the first systematic study of AtP and alternative methods for faster activation patching and show that AtP significantly outperforms all other investigated methods, with AtP* providing further significant improvement. Finally, we provide a method to bound the probability of remaining false negatives of AtP* estimates."
                },
                {
                    "title": "DsDm: Model-Aware Dataset Selection with Datamodels",
                    "abstract": "When selecting data for training large-scale models, standard practice is to filter for examples that match human notions of data quality. Such filtering yields qualitatively clean datapoints that intuitively should improve model behavior. However, in practice the opposite can often happen: we find that selecting according to similarity with\"high quality\"data sources may not increase (and can even hurt) performance compared to randomly selecting data. To develop better methods for selecting data, we start by framing dataset selection as an optimization problem that we can directly solve for: given target tasks, a learning algorithm, and candidate data, select the subset that maximizes model performance. This framework thus avoids handpicked notions of data quality, and instead models explicitly how the learning process uses train datapoints to predict on the target tasks. Our resulting method greatly improves language model (LM) performance on both pre-specified tasks and previously unseen tasks. Specifically, choosing target tasks representative of standard LM problems and evaluating on diverse held-out benchmarks, our selected datasets provide a 2x compute multiplier over baseline methods."
                },
                {
                    "title": "Calibrated Language Models Must Hallucinate",
                    "abstract": "Recent language models generate false but plausible-sounding text with surprising frequency. Such \u201challucinations\u201d are an obstacle to the usability of language-based AI systems and can harm people who rely upon their outputs. This work shows that there is an inherent statistical lower-bound on the rate that pretrained language models hallucinate certain types of facts, having nothing to do with the transformer LM architecture or data quality. For \u201carbitrary\u201d facts whose veracity cannot be determined from the training data, we show that hallucinations must occur at a certain rate for language models that satisfy a statistical calibration condition appropriate for generative language models. Specifically, if the maximum probability of any fact is bounded, we show that the probability of generating a hallucination is close to the fraction of facts that occur exactly once in the training data (a \u201cGood-Turing\u201d estimate), even assuming ideal training data without errors. One conclusion is that models pretrained to be sufficiently good predictors (i.e., calibrated) may require post-training to mitigate hallucinations on the type of arbitrary facts that tend to appear once in the training set. However, our analysis also suggests that there is no statistical reason that pretraining will lead to hallucination on facts that tend to appear more than once in the training data (like references to publications such as articles and books, whose hallucinations have been particularly notable and problematic) or on systematic facts (like arithmetic calculations). Therefore, different architectures and learning algorithms may mitigate these latter types of hallucinations."
                },
                {
                    "title": "Unifying Corroborative and Contributive Attributions in Large Language Models",
                    "abstract": "As businesses, products, and services spring up around large language models, the trustworthiness of these models hinges on the verifiability of their outputs. However, methods for explaining language model outputs largely fall across two distinct fields of study which both use the term \"attribution\" to refer to entirely separate techniques: citation generation and training data attribution. In many modern applications, such as legal document generation and medical question answering, both types of attributions are important. In this work, we argue for and present a unified framework of large language model attributions. We show how existing methods of different types of attribution fall under the unified framework. We also use the framework to discuss real-world use cases where one or both types of attributions are required. We believe that this unified framework will guide the use case driven development of systems that leverage both types of attribution, as well as the standardization of their evaluation."
                },
                {
                    "title": "Effective Large Language Model Adaptation for Improved Grounding and Citation Generation",
                    "abstract": "Large language models (LLMs) have achieved remarkable advancements in natural language understanding and generation. However, one major issue towards their widespread deployment in the real world is that they can generate \u201challucinated\u201d answers that are not factual.Towards this end, this paper focuses on improving LLMs by grounding their responses in retrieved passages and by providing citations. We propose a new framework, AGREE, Adaptation for GRounding EnhancEment, that improves the grounding from a holistic perspective. Our framework tunes LLMs to self-ground the claims in their responses and provide accurate citations to retrieved documents. This tuning on top of the pre-trained LLMs requires well-grounded responses (with citations) for paired queries, for which we introduce a method that can automatically construct such data from unlabeled queries. The self-grounding capability of tuned LLMs further grants them a test-time adaptation (TTA) capability that can actively retrieve passages to support the claims that have not been grounded, which iteratively improves the responses of LLMs. Across five datasets and two LLMs, our results show that the proposed tuning-based framework generates superior grounded responses with more accurate citations compared to prompting-based approaches and post-hoc citing-based approaches."
                },
                {
                    "title": "A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions",
                    "abstract": "The emergence of large language models (LLMs) has marked a significant breakthrough in natural language processing (NLP), leading to remarkable advancements in text understanding and generation. Nevertheless, alongside these strides, LLMs exhibit a critical tendency to produce hallucinations, resulting in content that is inconsistent with real-world facts or user inputs. This phenomenon poses substantial challenges to their practical deployment and raises concerns over the reliability of LLMs in real-world scenarios, which attracts increasing attention to detect and mitigate these hallucinations. In this survey, we aim to provide a thorough and in-depth overview of recent advances in the field of LLM hallucinations. We begin with an innovative taxonomy of LLM hallucinations, then delve into the factors contributing to hallucinations. Subsequently, we present a comprehensive overview of hallucination detection methods and benchmarks. Additionally, representative approaches designed to mitigate hallucinations are introduced accordingly. Finally, we analyze the challenges that highlight the current limitations and formulate open questions, aiming to delineate pathways for future research on hallucinations in LLMs."
                },
                {
                    "title": "Mistral 7B",
                    "abstract": "We introduce Mistral 7B v0.1, a 7-billion-parameter language model engineered for superior performance and efficiency. Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation. Our model leverages grouped-query attention (GQA) for faster inference, coupled with sliding window attention (SWA) to effectively handle sequences of arbitrary length with a reduced inference cost. We also provide a model fine-tuned to follow instructions, Mistral 7B -- Instruct, that surpasses the Llama 2 13B -- Chat model both on human and automated benchmarks. Our models are released under the Apache 2.0 license."
                },
                {
                    "title": "DataInf: Efficiently Estimating Data Influence in LoRA-tuned LLMs and Diffusion Models",
                    "abstract": "Quantifying the impact of training data points is crucial for understanding the outputs of machine learning models and for improving the transparency of the AI pipeline. The influence function is a principled and popular data attribution method, but its computational cost often makes it challenging to use. This issue becomes more pronounced in the setting of large language models and text-to-image models. In this work, we propose DataInf, an efficient influence approximation method that is practical for large-scale generative AI models. Leveraging an easy-to-compute closed-form expression, DataInf outperforms existing influence computation algorithms in terms of computational and memory efficiency. Our theoretical analysis shows that DataInf is particularly well-suited for parameter-efficient fine-tuning techniques such as LoRA. Through systematic empirical evaluations, we show that DataInf accurately approximates influence scores and is orders of magnitude faster than existing methods. In applications to RoBERTa-large, Llama-2-13B-chat, and stable-diffusion-v1.5 models, DataInf effectively identifies the most influential fine-tuning examples better than other approximate influence scores. Moreover, it can help to identify which data points are mislabeled."
                },
                {
                    "title": "Quantifying the Plausibility of Context Reliance in Neural Machine Translation",
                    "abstract": "Establishing whether language models can use contextual information in a human-plausible way is important to ensure their trustworthiness in real-world settings. However, the questions of when and which parts of the context affect model generations are typically tackled separately, with current plausibility evaluations being practically limited to a handful of artificial benchmarks. To address this, we introduce Plausibility Evaluation of Context Reliance (PECoRe), an end-to-end interpretability framework designed to quantify context usage in language models' generations. Our approach leverages model internals to (i) contrastively identify context-sensitive target tokens in generated texts and (ii) link them to contextual cues justifying their prediction. We use \\pecore to quantify the plausibility of context-aware machine translation models, comparing model rationales with human annotations across several discourse-level phenomena. Finally, we apply our method to unannotated model translations to identify context-mediated predictions and highlight instances of (im)plausible context usage throughout generation."
                },
                {
                    "title": "Attention Sorting Combats Recency Bias In Long Context Language Models",
                    "abstract": "Current language models often fail to incorporate long contexts efficiently during generation. We show that a major contributor to this issue are attention priors that are likely learned during pre-training: relevant information located earlier in context is attended to less on average. Yet even when models fail to use the information from a relevant document in their response, they still pay preferential attention to that document compared to an irrelevant document at the same position. We leverage this fact to introduce ``attention sorting'': perform one step of decoding, sort documents by the attention they receive (highest attention going last), repeat the process, generate the answer with the newly sorted context. We find that attention sorting improves performance of long context models. Our findings highlight some challenges in using off-the-shelf language models for retrieval augmented generation."
                },
                {
                    "title": "Studying Large Language Model Generalization with Influence Functions",
                    "abstract": "When trying to gain better visibility into a machine learning model in order to understand and mitigate the associated risks, a potentially valuable source of evidence is: which training examples most contribute to a given behavior? Influence functions aim to answer a counterfactual: how would the model's parameters (and hence its outputs) change if a given sequence were added to the training set? While influence functions have produced insights for small models, they are difficult to scale to large language models (LLMs) due to the difficulty of computing an inverse-Hessian-vector product (IHVP). We use the Eigenvalue-corrected Kronecker-Factored Approximate Curvature (EK-FAC) approximation to scale influence functions up to LLMs with up to 52 billion parameters. In our experiments, EK-FAC achieves similar accuracy to traditional influence function estimators despite the IHVP computation being orders of magnitude faster. We investigate two algorithmic techniques to reduce the cost of computing gradients of candidate training sequences: TF-IDF filtering and query batching. We use influence functions to investigate the generalization patterns of LLMs, including the sparsity of the influence patterns, increasing abstraction with scale, math and programming abilities, cross-lingual generalization, and role-playing behavior. Despite many apparently sophisticated forms of generalization, we identify a surprising limitation: influences decay to near-zero when the order of key phrases is flipped. Overall, influence functions give us a powerful new tool for studying the generalization properties of LLMs."
                },
                {
                    "title": "Universal and Transferable Adversarial Attacks on Aligned Language Models",
                    "abstract": "Because\"out-of-the-box\"large language models are capable of generating a great deal of objectionable content, recent work has focused on aligning these models in an attempt to prevent undesirable generation. While there has been some success at circumventing these measures -- so-called\"jailbreaks\"against LLMs -- these attacks have required significant human ingenuity and are brittle in practice. In this paper, we propose a simple and effective attack method that causes aligned language models to generate objectionable behaviors. Specifically, our approach finds a suffix that, when attached to a wide range of queries for an LLM to produce objectionable content, aims to maximize the probability that the model produces an affirmative response (rather than refusing to answer). However, instead of relying on manual engineering, our approach automatically produces these adversarial suffixes by a combination of greedy and gradient-based search techniques, and also improves over past automatic prompt generation methods. Surprisingly, we find that the adversarial prompts generated by our approach are quite transferable, including to black-box, publicly released LLMs. Specifically, we train an adversarial attack suffix on multiple prompts (i.e., queries asking for many different types of objectionable content), as well as multiple models (in our case, Vicuna-7B and 13B). When doing so, the resulting attack suffix is able to induce objectionable content in the public interfaces to ChatGPT, Bard, and Claude, as well as open source LLMs such as LLaMA-2-Chat, Pythia, Falcon, and others. In total, this work significantly advances the state-of-the-art in adversarial attacks against aligned language models, raising important questions about how such systems can be prevented from producing objectionable information. Code is available at github.com/llm-attacks/llm-attacks."
                },
                {
                    "title": "FacTool: Factuality Detection in Generative AI - A Tool Augmented Framework for Multi-Task and Multi-Domain Scenarios",
                    "abstract": "The emergence of generative pre-trained models has facilitated the synthesis of high-quality text, but it has also posed challenges in identifying factual errors in the generated text. In particular: (1) A wider range of tasks now face an increasing risk of containing factual errors when handled by generative models. (2) Generated texts tend to be lengthy and lack a clearly defined granularity for individual facts. (3) There is a scarcity of explicit evidence available during the process of fact checking. With the above challenges in mind, in this paper, we propose FacTool, a task and domain agnostic framework for detecting factual errors of texts generated by large language models (e.g., ChatGPT). Experiments on four different tasks (knowledge-based QA, code generation, mathematical reasoning, and scientific literature review) show the efficacy of the proposed method. We release the code of FacTool associated with ChatGPT plugin interface at https://github.com/GAIR-NLP/factool ."
                },
                {
                    "title": "Lost in the Middle: How Language Models Use Long Contexts",
                    "abstract": "While recent language models have the ability to take long contexts as input, relatively little is known about how well they use longer context. We analyze the performance of language models on two tasks that require identifying relevant information in their input contexts: multi-document question answering and key-value retrieval. We find that performance can degrade significantly when changing the position of relevant information, indicating that current language models do not robustly make use of information in long input contexts. In particular, we observe that performance is often highest when relevant information occurs at the beginning or end of the input context, and significantly degrades when models must access relevant information in the middle of long contexts, even for explicitly long-context models. Our analysis provides a better understanding of how language models use their input context and provides new evaluation protocols for future long-context language models."
                },
                {
                    "title": "Sequential Integrated Gradients: a simple but effective method for explaining language models",
                    "abstract": "Several explanation methods such as Integrated Gradients (IG) can be characterised as path-based methods, as they rely on a straight line between the data and an uninformative baseline. However, when applied to language models, these methods produce a path for each word of a sentence simultaneously, which could lead to creating sentences from interpolated words either having no clear meaning, or having a significantly different meaning compared to the original sentence. In order to keep the meaning of these sentences as close as possible to the original one, we propose Sequential Integrated Gradients (SIG), which computes the importance of each word in a sentence by keeping fixed every other words, only creating interpolations between the baseline and the word of interest. Moreover, inspired by the training procedure of several language models, we also propose to replace the baseline token\"pad\"with the trained token\"mask\". While being a simple improvement over the original IG method, we show on various models and datasets that SIG proves to be a very effective method for explaining language models."
                },
                {
                    "title": "PURR: Efficiently Editing Language Model Hallucinations by Denoising Language Model Corruptions",
                    "abstract": "The remarkable capabilities of large language models have been accompanied by a persistent drawback: the generation of false and unsubstantiated claims commonly known as\"hallucinations\". To combat this issue, recent research has introduced approaches that involve editing and attributing the outputs of language models, particularly through prompt-based editing. However, the inference cost and speed of using large language models for editing currently bottleneck prompt-based methods. These bottlenecks motivate the training of compact editors, which is challenging due to the scarcity of training data for this purpose. To overcome these challenges, we exploit the power of large language models to introduce corruptions (i.e., noise) into text and subsequently fine-tune compact editors to denoise the corruptions by incorporating relevant evidence. Our methodology is entirely unsupervised and provides us with faux hallucinations for training in any domain. Our Petite Unsupervised Research and Revision model, PURR, not only improves attribution over existing editing methods based on fine-tuning and prompting, but also achieves faster execution times by orders of magnitude."
                },
                {
                    "title": "Enabling Large Language Models to Generate Text with Citations",
                    "abstract": "Large language models (LLMs) have emerged as a widely-used tool for information seeking, but their generated outputs are prone to hallucination. In this work, our aim is to allow LLMs to generate text with citations, improving their factual correctness and verifiability. Existing work mainly relies on commercial search engines and human evaluation, making it challenging to reproduce and compare different modeling approaches. We propose ALCE, the first benchmark for Automatic LLMs' Citation Evaluation. ALCE collects a diverse set of questions and retrieval corpora and requires building end-to-end systems to retrieve supporting evidence and generate answers with citations. We develop automatic metrics along three dimensions -- fluency, correctness, and citation quality -- and demonstrate their strong correlation with human judgements. Our experiments with state-of-the-art LLMs and novel prompting strategies show that current systems have considerable room for improvement -- For example, on the ELI5 dataset, even the best models lack complete citation support 50% of the time. Our analyses further highlight promising future directions, including developing better retrievers, advancing long-context LLMs, and improving the ability to synthesize information from multiple sources."
                },
                {
                    "title": "FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation",
                    "abstract": "Evaluating the factuality of long-form text generated by large language models (LMs) is non-trivial because (1) generations often contain a mixture of supported and unsupported pieces of information, making binary judgments of quality inadequate, and (2) human evaluation is time-consuming and costly. In this paper, we introduce FACTSCORE, a new evaluation that breaks a generation into a series of atomic facts and computes the percentage of atomic facts supported by a reliable knowledge source. We conduct an extensive human evaluation to obtain FACTSCOREs of people biographies generated by several state-of-the-art commercial LMs -- InstructGPT, ChatGPT, and the retrieval-augmented PerplexityAI -- and report new analysis demonstrating the need for such a fine-grained score (e.g., ChatGPT only achieves 58%). Since human evaluation is costly, we also introduce an automated model that estimates FACTSCORE using retrieval and a strong language model, with less than a 2% error rate. Finally, we use this automated metric to evaluate 6,500 generations from a new set of 13 recent LMs that would have cost $26K if evaluated by humans, with various findings: GPT-4 and ChatGPT are more factual than public models, and Vicuna and Alpaca are some of the best public models. FACTSCORE is available for public use via `pip install factscore`."
                },
                {
                    "title": "Complex Claim Verification with Evidence Retrieved in the Wild",
                    "abstract": "Retrieving evidence to support or refute claims is a core part of automatic fact-checking. Prior work makes simplifying assumptions in retrieval that depart from real-world use cases: either no access to evidence, access to evidence curated by a human fact-checker, or access to evidence published after a claim was made. In this work, we present the first realistic pipeline to check real-world claims by retrieving raw evidence from the web. We restrict our retriever to only search documents available prior to the claim\u2019s making, modeling the realistic scenario of emerging claims. Our pipeline includes five components: claim decomposition, raw document retrieval, fine-grained evidence retrieval, claim-focused summarization, and veracity judgment. We conduct experiments on complex political claims in the ClaimDecomp dataset and show that the aggregated evidence produced by our pipeline improves veracity judgments. Human evaluation finds the evidence summary produced by our system is reliable (it does not hallucinate information) and relevant to answering key questions about a claim, suggesting that it can assist fact-checkers even when it does not reflect a complete evidence set."
                },
                {
                    "title": "Evaluating Verifiability in Generative Search Engines",
                    "abstract": "Generative search engines directly generate responses to user queries, along with in-line citations. A prerequisite trait of a trustworthy generative search engine is verifiability, i.e., systems should cite comprehensively (high citation recall; all statements are fully supported by citations) and accurately (high citation precision; every cite supports its associated statement). We conduct human evaluation to audit four popular generative search engines -- Bing Chat, NeevaAI, perplexity.ai, and YouChat -- across a diverse set of queries from a variety of sources (e.g., historical Google user queries, dynamically-collected open-ended questions on Reddit, etc.). We find that responses from existing generative search engines are fluent and appear informative, but frequently contain unsupported statements and inaccurate citations: on average, a mere 51.5% of generated sentences are fully supported by citations and only 74.5% of citations support their associated sentence. We believe that these results are concerningly low for systems that may serve as a primary tool for information-seeking users, especially given their facade of trustworthiness. We hope that our results further motivate the development of trustworthy generative search engines and help researchers and users better understand the shortcomings of existing commercial systems."
                },
                {
                    "title": "TRAK: Attributing Model Behavior at Scale",
                    "abstract": "The goal of data attribution is to trace model predictions back to training data. Despite a long line of work towards this goal, existing approaches to data attribution tend to force users to choose between computational tractability and efficacy. That is, computationally tractable methods can struggle with accurately attributing model predictions in non-convex settings (e.g., in the context of deep neural networks), while methods that are effective in such regimes require training thousands of models, which makes them impractical for large models or datasets. In this work, we introduce TRAK (Tracing with the Randomly-projected After Kernel), a data attribution method that is both effective and computationally tractable for large-scale, differentiable models. In particular, by leveraging only a handful of trained models, TRAK can match the performance of attribution methods that require training thousands of models. We demonstrate the utility of TRAK across various modalities and scales: image classifiers trained on ImageNet, vision-language models (CLIP), and language models (BERT and mT5). We provide code for using TRAK (and reproducing our work) at https://github.com/MadryLab/trak ."
                },
                {
                    "title": "Not What You've Signed Up For: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection",
                    "abstract": "Large Language Models (LLMs) are increasingly being integrated into applications, with versatile functionalities that can be easily modulated via natural language prompts. So far, it was assumed that the user is directly prompting the LLM. But, what if it is not the user prompting? We show that LLM-Integrated Applications blur the line between data and instructions and reveal several new attack vectors, using Indirect Prompt Injection, that enable adversaries to remotely (i.e., without a direct interface) exploit LLM-integrated applications by strategically injecting prompts into data likely to be retrieved at inference time. We derive a comprehensive taxonomy from a computer security perspective to broadly investigate impacts and vulnerabilities, including data theft, worming, information ecosystem contamination, and other novel security risks. We then demonstrate the practical viability of our attacks against both real-world systems, such as Bing Chat and code-completion engines, and GPT-4 synthetic applications. We show how processing retrieved prompts can act as arbitrary code execution, manipulate the application's functionality, and control how and if other APIs are called. Despite the increasing reliance on LLMs, effective mitigations of these emerging threats are lacking. By raising awareness of these vulnerabilities, we aim to promote the safe and responsible deployment of these powerful models and the development of robust defenses that protect users from potential attacks."
                },
                {
                    "title": "In-context Example Selection with Influences",
                    "abstract": "In-context learning (ICL) is a powerful paradigm emerged from large language models (LLMs). Despite its promises, ICL performance is known to be highly sensitive to input examples. In this work, we use $\\textit{in-context influences}$ to analyze few-shot ICL performance directly from the in-context examples. Our proposed influence-based example selection method can identify both positive and negative examples, outperforming several baselines when evaluated on 9 SuperGLUE tasks. Our analysis uncovers up to a $16.3\\%$ performance gap between using the most negative in-context examples compared to the most positive. In a case study, we apply our influence-based framework to quantify the phenomena of recency bias in example ordering for few-shot ICL."
                },
                {
                    "title": "Data Curation Alone Can Stabilize In-context Learning",
                    "abstract": "In-context learning (ICL) enables large language models (LLMs) to perform new tasks by prompting them with a sequence of training examples. However, it is known that ICL is very sensitive to the choice of training examples: randomly sampling examples from a training set leads to high variance in performance. In this paper, we show that carefully curating a subset of training data greatly stabilizes ICL performance without any other changes to the ICL algorithm (e.g., prompt retrieval or calibration). We introduce two methods to choose training subsets\u2014both score training examples individually, then select the highest-scoring ones. CondAcc scores a training example by its average dev-set ICL accuracy when combined with random training examples, while Datamodels learns linear regressors that estimate how the presence of each training example influences LLM outputs. Across five tasks and two LLMs, sampling from stable subsets selected by CondAcc and Datamodels improves average accuracy over sampling from the entire training set by 7.7% and 6.3%, respectively.Surprisingly, the stable subset examples are not especially diverse in content or low in perplexity, in contrast with other work suggesting that diversity and perplexity are important when prompting LLMs."
                },
                {
                    "title": "Attributed Question Answering: Evaluation and Modeling for Attributed Large Language Models",
                    "abstract": "Large language models (LLMs) have shown impressive results while requiring little or no direct supervision. Further, there is mounting evidence that LLMs may have potential in information-seeking scenarios. We believe the ability of an LLM to attribute the text that it generates is likely to be crucial in this setting. We formulate and study Attributed QA as a key first step in the development of attributed LLMs. We propose a reproducible evaluation framework for the task and benchmark a broad set of architectures. We take human annotations as a gold standard and show that a correlated automatic metric is suitable for development. Our experimental work gives concrete answers to two key questions (How to measure attribution?, and How well do current state-of-the-art methods perform on attribution?), and give some hints as to how to address a third (How to build LLMs with attribution?)."
                },
                {
                    "title": "Ignore Previous Prompt: Attack Techniques For Language Models",
                    "abstract": "Transformer-based large language models (LLMs) provide a powerful foundation for natural language tasks in large-scale customer-facing applications. However, studies that explore their vulnerabilities emerging from malicious user interaction are scarce. By proposing PromptInject, a prosaic alignment framework for mask-based iterative adversarial prompt composition, we examine how GPT-3, the most widely deployed language model in production, can be easily misaligned by simple handcrafted inputs. In particular, we investigate two types of attacks -- goal hijacking and prompt leaking -- and demonstrate that even low-aptitude, but sufficiently ill-intentioned agents, can easily exploit GPT-3's stochastic nature, creating long-tail risks. The code for PromptInject is available at https://github.com/agencyenterprise/PromptInject."
                },
                {
                    "title": "RARR: Researching and Revising What Language Models Say, Using Language Models",
                    "abstract": "Language models (LMs) now excel at many tasks such as question answering, reasoning, and dialog. However, they sometimes generate unsupported or misleading content. A user cannot easily determine whether their outputs are trustworthy or not, because most LMs do not have any built-in mechanism for attribution to external evidence. To enable attribution while still preserving all the powerful advantages of recent generation models, we propose RARR (Retrofit Attribution using Research and Revision), a system that 1) automatically finds attribution for the output of any text generation model, and 2) post-edits the output to fix unsupported content while preserving the original output as much as possible. When applied to the output of several state-of-the-art LMs on a diverse set of generation tasks, we find that RARR significantly improves attribution while otherwise preserving the original input to a much greater degree than previously explored edit models. Furthermore, the implementation of RARR requires only a handful of training examples, a large language model, and standard web search."
                },
                {
                    "title": "Teaching language models to support answers with verified quotes",
                    "abstract": "Recent large language models often answer factual questions correctly. But users can't trust any given claim a model makes without fact-checking, because language models can hallucinate convincing nonsense. In this work we use reinforcement learning from human preferences (RLHP) to train\"open-book\"QA models that generate answers whilst also citing specific evidence for their claims, which aids in the appraisal of correctness. Supporting evidence is drawn from multiple documents found via a search engine, or from a single user-provided document. Our 280 billion parameter model, GopherCite, is able to produce answers with high quality supporting evidence and abstain from answering when unsure. We measure the performance of GopherCite by conducting human evaluation of answers to questions in a subset of the NaturalQuestions and ELI5 datasets. The model's response is found to be high-quality 80\\% of the time on this Natural Questions subset, and 67\\% of the time on the ELI5 subset. Abstaining from the third of questions for which it is most unsure improves performance to 90\\% and 80\\% respectively, approaching human baselines. However, analysis on the adversarial TruthfulQA dataset shows why citation is only one part of an overall strategy for safety and trustworthiness: not all claims supported by evidence are true."
                },
                {
                    "title": "Datamodels: Predicting Predictions from Training Data",
                    "abstract": "We present a conceptual framework, datamodeling, for analyzing the behavior of a model class in terms of the training data. For any fixed\"target\"example $x$, training set $S$, and learning algorithm, a datamodel is a parameterized function $2^S \\to \\mathbb{R}$ that for any subset of $S' \\subset S$ -- using only information about which examples of $S$ are contained in $S'$ -- predicts the outcome of training a model on $S'$ and evaluating on $x$. Despite the potential complexity of the underlying process being approximated (e.g., end-to-end training and evaluation of deep neural networks), we show that even simple linear datamodels can successfully predict model outputs. We then demonstrate that datamodels give rise to a variety of applications, such as: accurately predicting the effect of dataset counterfactuals; identifying brittle predictions; finding semantically similar examples; quantifying train-test leakage; and embedding data into a well-behaved and feature-rich representation space. Data for this paper (including pre-computed datamodels as well as raw predictions from four million trained deep neural networks) is available at https://github.com/MadryLab/datamodels-data ."
                },
                {
                    "title": "LaMDA: Language Models for Dialog Applications",
                    "abstract": "We present LaMDA: Language Models for Dialog Applications. LaMDA is a family of Transformer-based neural language models specialized for dialog, which have up to 137B parameters and are pre-trained on 1.56T words of public dialog data and web text. While model scaling alone can improve quality, it shows less improvements on safety and factual grounding. We demonstrate that fine-tuning with annotated data and enabling the model to consult external knowledge sources can lead to significant improvements towards the two key challenges of safety and factual grounding. The first challenge, safety, involves ensuring that the model's responses are consistent with a set of human values, such as preventing harmful suggestions and unfair bias. We quantify safety using a metric based on an illustrative set of human values, and we find that filtering candidate responses using a LaMDA classifier fine-tuned with a small amount of crowdworker-annotated data offers a promising approach to improving model safety. The second challenge, factual grounding, involves enabling the model to consult external knowledge sources, such as an information retrieval system, a language translator, and a calculator. We quantify factuality using a groundedness metric, and we find that our approach enables the model to generate responses grounded in known sources, rather than responses that merely sound plausible. Finally, we explore the use of LaMDA in the domains of education and content recommendations, and analyze their helpfulness and role consistency."
                },
                {
                    "title": "Measuring Attribution in Natural Language Generation Models",
                    "abstract": "Abstract Large neural models have brought a new challenge to natural language generation (NLG): It has become imperative to ensure the safety and reliability of the output of models that generate freely. To this end, we present an evaluation framework, Attributable to Identified Sources (AIS), stipulating that NLG output pertaining to the external world is to be verified against an independent, provided source. We define AIS and a two-stage annotation pipeline for allowing annotators to evaluate model output according to annotation guidelines. We successfully validate this approach on generation datasets spanning three tasks (two conversational QA datasets, a summarization dataset, and a table-to-text dataset). We provide full annotation guidelines in the appendices and publicly release the annotated data at https://github.com/google-research-datasets/AIS."
                },
                {
                    "title": "WebGPT: Browser-assisted question-answering with human feedback",
                    "abstract": "We fine-tune GPT-3 to answer long-form questions using a text-based web-browsing environment, which allows the model to search and navigate the web. By setting up the task so that it can be performed by humans, we are able to train models on the task using imitation learning, and then optimize answer quality with human feedback. To make human evaluation of factual accuracy easier, models must collect references while browsing in support of their answers. We train and evaluate our models on ELI5, a dataset of questions asked by Reddit users. Our best model is obtained by fine-tuning GPT-3 using behavior cloning, and then performing rejection sampling against a reward model trained to predict human preferences. This model's answers are preferred by humans 56% of the time to those of our human demonstrators, and 69% of the time to the highest-voted answer from Reddit."
                },
                {
                    "title": "Explaining by Removing: A Unified Framework for Model Explanation",
                    "abstract": "Researchers have proposed a wide variety of model explanation approaches, but it remains unclear how most methods are related or when one method is preferable to another. We establish a new class of methods, removal-based explanations, that are based on the principle of simulating feature removal to quantify each feature's influence. These methods vary in several respects, so we develop a framework that characterizes each method along three dimensions: 1) how the method removes features, 2) what model behavior the method explains, and 3) how the method summarizes each feature's influence. Our framework unifies 25 existing methods, including several of the most widely used approaches (SHAP, LIME, Meaningful Perturbations, permutation tests). This new class of explanation methods has rich connections that we examine using tools that have been largely overlooked by the explainability literature. To anchor removal-based explanations in cognitive psychology, we show that feature removal is a simple application of subtractive counterfactual reasoning. Ideas from cooperative game theory shed light on the relationships and trade-offs among different methods, and we derive conditions under which all removal-based explanations have information-theoretic interpretations. Through this analysis, we develop a unified framework that helps practitioners better understand model explanation tools, and that offers a strong theoretical foundation upon which future explainability research can build."
                },
                {
                    "title": "Quantifying Attention Flow in Transformers",
                    "abstract": "In the Transformer model, \u201cself-attention\u201d combines information from attended embeddings into the representation of the focal embedding in the next layer. Thus, across layers of the Transformer, information originating from different tokens gets increasingly mixed. This makes attention weights unreliable as explanations probes. In this paper, we consider the problem of quantifying this flow of information through self-attention. We propose two methods for approximating the attention to input tokens given attention weights, attention rollout and attention flow, as post hoc methods when we use attention weights as the relative relevance of the input tokens. We show that these methods give complementary views on the flow of information, and compared to raw attention, both yield higher correlations with importance scores of input tokens obtained using an ablation method and input gradients."
                },
                {
                    "title": "TyDi QA: A Benchmark for Information-Seeking Question Answering in Typologically Diverse Languages",
                    "abstract": "Abstract Confidently making progress on multilingual modeling requires challenging, trustworthy evaluations. We present TyDi QA\u2014a question answering dataset covering 11 typologically diverse languages with 204K question-answer pairs. The languages of TyDi QA are diverse with regard to their typology\u2014the set of linguistic features each language expresses\u2014such that we expect models performing well on this set to generalize across a large number of the world\u2019s languages. We present a quantitative analysis of the data quality and example-level qualitative linguistic analyses of observed language phenomena that would not be found in English-only corpora. To provide a realistic information-seeking task and avoid priming effects, questions are written by people who want to know the answer, but don\u2019t know the answer yet, and the data is collected directly in each language without the use of translation."
                },
                {
                    "title": "bLIMEy: Surrogate Prediction Explanations Beyond LIME",
                    "abstract": "Surrogate explainers of black-box machine learning predictions are of paramount importance in the field of eXplainable Artificial Intelligence since they can be applied to any type of data (images, text and tabular), are model-agnostic and are post-hoc (i.e., can be retrofitted). The Local Interpretable Model-agnostic Explanations (LIME) algorithm is often mistakenly unified with a more general framework of surrogate explainers, which may lead to a belief that it is the solution to surrogate explainability. In this paper we empower the community to \"build LIME yourself\" (bLIMEy) by proposing a principled algorithmic framework for building custom local surrogate explainers of black-box model predictions, including LIME itself. To this end, we demonstrate how to decompose the surrogate explainers family into algorithmically independent and interoperable modules and discuss the influence of these component choices on the functional capabilities of the resulting explainer, using the example of LIME."
                },
                {
                    "title": "Universal Adversarial Triggers for Attacking and Analyzing NLP",
                    "abstract": "Adversarial examples highlight model vulnerabilities and are useful for evaluation and interpretation. We define universal adversarial triggers: input-agnostic sequences of tokens that trigger a model to produce a specific prediction when concatenated to any input from a dataset. We propose a gradient-guided search over tokens which finds short trigger sequences (e.g., one word for classification and four words for language modeling) that successfully trigger the target prediction. For example, triggers cause SNLI entailment accuracy to drop from 89.94% to 0.55%, 72% of \u201cwhy\u201d questions in SQuAD to be answered \u201cto kill american people\u201d, and the GPT-2 language model to spew racist output even when conditioned on non-racial contexts. Furthermore, although the triggers are optimized using white-box access to a specific model, they transfer to other models for all tasks we consider. Finally, since triggers are input-agnostic, they provide an analysis of global model behavior. For instance, they confirm that SNLI models exploit dataset biases and help to diagnose heuristics learned by reading comprehension models."
                },
                {
                    "title": "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
                    "abstract": "BERT (Devlin et al., 2018) and RoBERTa (Liu et al., 2019) has set a new state-of-the-art performance on sentence-pair regression tasks like semantic textual similarity (STS). However, it requires that both sentences are fed into the network, which causes a massive computational overhead: Finding the most similar pair in a collection of 10,000 sentences requires about 50 million inference computations (~65 hours) with BERT. The construction of BERT makes it unsuitable for semantic similarity search as well as for unsupervised tasks like clustering. In this publication, we present Sentence-BERT (SBERT), a modification of the pretrained BERT network that use siamese and triplet network structures to derive semantically meaningful sentence embeddings that can be compared using cosine-similarity. This reduces the effort for finding the most similar pair from 65 hours with BERT / RoBERTa to about 5 seconds with SBERT, while maintaining the accuracy from BERT. We evaluate SBERT and SRoBERTa on common STS tasks and transfer learning tasks, where it outperforms other state-of-the-art sentence embeddings methods."
                },
                {
                    "title": "Natural Questions: A Benchmark for Question Answering Research",
                    "abstract": "We present the Natural Questions corpus, a question answering data set. Questions consist of real anonymized, aggregated queries issued to the Google search engine. An annotator is presented with a question along with a Wikipedia page from the top 5 search results, and annotates a long answer (typically a paragraph) and a short answer (one or more entities) if present on the page, or marks null if no long/short answer is present. The public release consists of 307,373 training examples with single annotations; 7,830 examples with 5-way annotations for development data; and a further 7,842 examples with 5-way annotated sequestered as test data. We present experiments validating quality of the data. We also describe analysis of 25-way annotations on 302 examples, giving insights into human variability on the annotation task. We introduce robust metrics for the purposes of evaluating question answering systems; demonstrate high human upper bounds on these metrics; and establish baseline results using competitive methods drawn from related literature."
                },
                {
                    "title": "Is Attention Interpretable?",
                    "abstract": "Attention mechanisms have recently boosted performance on a range of NLP tasks. Because attention layers explicitly weight input components\u2019 representations, it is also often assumed that attention can be used to identify information that models found important (e.g., specific contextualized word tokens). We test whether that assumption holds by manipulating attention weights in already-trained text classification models and analyzing the resulting differences in their predictions. While we observe some ways in which higher attention weights correlate with greater impact on model predictions, we also find many ways in which this does not hold, i.e., where gradient-based rankings of attention weights better predict their effects than their magnitudes. We conclude that while attention noisily predicts input components\u2019 overall importance to a model, it is by no means a fail-safe indicator.1"
                },
                {
                    "title": "DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs",
                    "abstract": "Reading comprehension has recently seen rapid progress, with systems matching humans on the most popular datasets for the task. However, a large body of work has highlighted the brittleness of these systems, showing that there is much work left to be done. We introduce a new reading comprehension benchmark, DROP, which requires Discrete Reasoning Over the content of Paragraphs. In this crowdsourced, adversarially-created, 55k-question benchmark, a system must resolve references in a question, perhaps to multiple input positions, and perform discrete operations over them (such as addition, counting, or sorting). These operations require a much more comprehensive understanding of the content of paragraphs, as they remove the paraphrase-and-entity-typing shortcuts available in prior datasets. We apply state-of-the-art methods from both the reading comprehension and semantic parsing literatures on this dataset and show that the best systems only achieve 38.4% F1 on our generalized accuracy metric, while expert human performance is 96%. We additionally present a new model that combines reading comprehension methods with simple numerical reasoning to achieve 51% F1."
                },
                {
                    "title": "Attention is not Explanation",
                    "abstract": "Attention mechanisms have seen wide adoption in neural NLP models. In addition to improving predictive performance, these are often touted as affording transparency: models equipped with attention provide a distribution over attended-to input units, and this is often presented (at least implicitly) as communicating the relative importance of inputs. However, it is unclear what relationship exists between attention weights and model outputs. In this work we perform extensive experiments across a variety of NLP tasks that aim to assess the degree to which attention weights provide meaningful \u201cexplanations\u201d for predictions. We find that they largely do not. For example, learned attention weights are frequently uncorrelated with gradient-based measures of feature importance, and one can identify very different attention distributions that nonetheless yield equivalent predictions. Our findings show that standard attention modules do not provide meaningful explanations and should not be treated as though they do."
                },
                {
                    "title": "HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering",
                    "abstract": "Existing question answering (QA) datasets fail to train QA systems to perform complex reasoning and provide explanations for answers. We introduce HotpotQA, a new dataset with 113k Wikipedia-based question-answer pairs with four key features: (1) the questions require finding and reasoning over multiple supporting documents to answer; (2) the questions are diverse and not constrained to any pre-existing knowledge bases or knowledge schemas; (3) we provide sentence-level supporting facts required for reasoning, allowing QA systems to reason with strong supervision and explain the predictions; (4) we offer a new type of factoid comparison questions to test QA systems\u2019 ability to extract relevant facts and perform necessary comparison. We show that HotpotQA is challenging for the latest QA systems, and the supporting facts enable models to improve performance and make explainable predictions."
                },
                {
                    "title": "Interactive Visualization and Manipulation of Attention-based Neural Machine Translation",
                    "abstract": "While neural machine translation (NMT) provides high-quality translation, it is still hard to interpret and analyze its behavior. We present an interactive interface for visualizing and intervening behavior of NMT, specifically concentrating on the behavior of beam search mechanism and attention component. The tool (1) visualizes search tree and attention and (2) provides interface to adjust search tree and attention weight (manually or automatically) at real-time. We show the tool gives various methods to understand NMT."
                },
                {
                    "title": "Visualizing and Understanding Neural Machine Translation",
                    "abstract": "While neural machine translation (NMT) has made remarkable progress in recent years, it is hard to interpret its internal workings due to the continuous representations and non-linearity of neural networks. In this work, we propose to use layer-wise relevance propagation (LRP) to compute the contribution of each contextual word to arbitrary hidden states in the attention-based encoder-decoder framework. We show that visualization with LRP helps to interpret the internal workings of NMT and analyze translation errors."
                },
                {
                    "title": "SmoothGrad: removing noise by adding noise",
                    "abstract": "Explaining the output of a deep network remains a challenge. In the case of an image classifier, one type of explanation is to identify pixels that strongly influence the final decision. A starting point for this strategy is the gradient of the class score function with respect to the input image. This gradient can be interpreted as a sensitivity map, and there are several techniques that elaborate on this basic idea. This paper makes two contributions: it introduces SmoothGrad, a simple method that can help visually sharpen gradient-based sensitivity maps, and it discusses lessons in the visualization of these maps. We publish the code for our experiments and a website with our results."
                },
                {
                    "title": "A Unified Approach to Interpreting Model Predictions",
                    "abstract": "Understanding why a model makes a certain prediction can be as crucial as the prediction's accuracy in many applications. However, the highest accuracy for large modern datasets is often achieved by complex models that even experts struggle to interpret, such as ensemble or deep learning models, creating a tension between accuracy and interpretability. In response, various methods have recently been proposed to help users interpret the predictions of complex models, but it is often unclear how these methods are related and when one method is preferable over another. To address this problem, we present a unified framework for interpreting predictions, SHAP (SHapley Additive exPlanations). SHAP assigns each feature an importance value for a particular prediction. Its novel components include: (1) the identification of a new class of additive feature importance measures, and (2) theoretical results showing there is a unique solution in this class with a set of desirable properties. The new class unifies six existing methods, notable because several recent methods in the class lack the proposed desirable properties. Based on insights from this unification, we present new methods that show improved computational performance and/or better consistency with human intuition than previous approaches."
                },
                {
                    "title": "MS MARCO: A Human Generated MAchine Reading COmprehension Dataset",
                    "abstract": "This paper presents our recent work on the design and development of a new, large scale dataset, which we name MS MARCO, for MAchine Reading COmprehension. This new dataset is aimed to overcome a number of well-known weaknesses of previous publicly available datasets for the same task of reading comprehension and question answering. In MS MARCO, all questions are sampled from real anonymized user queries. The context passages, from which answers in the dataset are derived, are extracted from real web documents using the most advanced version of the Bing search engine. The answers to the queries are human generated. Finally, a subset of these queries has multiple answers. We aim to release one million queries and the corresponding answers in the dataset, which, to the best of our knowledge, is the most comprehensive real-world dataset of its kind in both quantity and quality. We are currently releasing 100,000 queries with their corresponding answers to inspire work in reading comprehension and question answering along with gathering feedback from the research community."
                },
                {
                    "title": "Not Just a Black Box: Learning Important Features Through Propagating Activation Differences",
                    "abstract": "Note: This paper describes an older version of DeepLIFT. See this https URL for the newer version. Original abstract follows: The purported \"black box\" nature of neural networks is a barrier to adoption in applications where interpretability is essential. Here we present DeepLIFT (Learning Important FeaTures), an efficient and effective method for computing importance scores in a neural network. DeepLIFT compares the activation of each neuron to its 'reference activation' and assigns contribution scores according to the difference. We apply DeepLIFT to models trained on natural images and genomic data, and show significant advantages over gradient-based methods."
                },
                {
                    "title": "Abstractive Text Summarization using Sequence-to-sequence RNNs and Beyond",
                    "abstract": "In this work, we model abstractive text summarization using Attentional Encoder-Decoder Recurrent Neural Networks, and show that they achieve state-of-the-art performance on two different corpora. We propose several novel models that address critical problems in summarization that are not adequately modeled by the basic architecture, such as modeling key-words, capturing the hierarchy of sentence-to-word structure, and emitting words that are rare or unseen at training time. Our work shows that many of our proposed models contribute to further improvement in performance. We also propose a new dataset consisting of multi-sentence summaries, and establish performance benchmarks for further research."
                },
                {
                    "title": "\u201cWhy Should I Trust You?\u201d: Explaining the Predictions of Any Classifier",
                    "abstract": "Despite widespread adoption, machine learning models remain mostly black boxes. Understanding the reasons behind predictions is, however, quite important in assessing trust, which is fundamental if one plans to take action based on a prediction, or when choosing whether to deploy a new model. Such understanding also provides insights into the model, which can be used to transform an untrustworthy model or prediction into a trustworthy one. In this work, we propose LIME, a novel explanation technique that explains the predictions of any classifier in an interpretable and faithful manner, by learning an interpretable model locally varound the prediction. We also propose a method to explain models by presenting representative individual predictions and their explanations in a non-redundant way, framing the task as a submodular optimization problem. We demonstrate the flexibility of these methods by explaining different models for text (e.g. random forests) and image classification (e.g. neural networks). We show the utility of explanations via novel experiments, both simulated and with human subjects, on various scenarios that require trust: deciding if one should trust a prediction, choosing between models, improving an untrustworthy classifier, and identifying why a classifier should not be trusted."
                },
                {
                    "title": "The proof and measurement of association between two things.",
                    "abstract": "All knowledge\u2014beyond that of bare isolated occurrence\u2014deals with uniformities. Of the latter, some few have a claim to be considered absolute, such as mathematical implications and mechanical laws. But the vast majority are only partial; medicine does not teach that smallpox is inevitably escaped by vaccination, but that it is so generally; biology has not shown that all animals require organic food, but that nearly all do so; in daily life, a dark sky is no proof that it will rain, but merely a warning; even in morality, the sole categorical imperative alleged by Kant was the sinfulness of telling a lie, and few thinkers since have admitted so much as this to be valid universally. In psychology, more perhaps than in any other science, it is hard to find absolutely inflexible coincidences; occasionally, indeed, there appear uniformities sufficiently regular to be practically treated as laws, but infinitely the greater part of the observations hitherto recorded concern only more or less pronounced tendencies of one event or attribute to accompany another. Under these circumstances, one might well have expected that the evidential evaluation and precise mensuration of tendencies had long been the subject of exhaustive investigation and now formed one of the earliest sections in a beginner\u2019s psychological course. Instead, we find only a general na\u0131\u0308ve ignorance that there is anything about it requiring to be learnt. One after another, laborious series of experiments are executed and published with the purpose of demonstrating some connection between two events, wherein the otherwise learned psychologist reveals that his art of proving and measuring correspondence has not advanced beyond that of lay persons. The consequence has been that the significance of the experiments is not at all rightly understood, nor have any definite facts been elicited that may be either confirmed or refuted. The present article is a commencement at attempting to remedy this deficiency of scientific correlation. With this view, it will be strictly confined to the needs of practical workers, and all theoretical mathematical demonstrations will be omitted; it may, however, be said that the relations stated have already received a large amount of empirical verification. Great thanks are due from me to Professor Haussdorff and to Dr. G. Lipps, each of whom have supplied a useful theorem in polynomial probability; the former has also very kindly given valuable advice concerning the proof of the important formul\u00e6 for elimination of \u2018\u2018systematic deviations.\u2019\u2019 At the same time, and for the same reason, the meaning and working of the various formul\u00e6 have been explained sufficiently, it is hoped, to render them readily usable even by those whose knowledge of mathematics is elementary. The fundamental procedure is accompanied by simple imaginary examples, while the more advanced parts are illustrated by cases that have actually occurred in my personal experience. For more abundant and positive exemplification, the reader is requested to refer to the under cited research, which is entirely built upon the principles and mathematical relations here laid down. In conclusion, the general value of the methodics recommended is emphasized by a brief criticism of the best correlational work hitherto made public, and also the important question is discussed as to the number of \u2018\u2018cases\u2019\u2019 required for an experimental series."
                },
                {
                    "title": "Visualizing and Understanding Neural Models in NLP",
                    "abstract": "While neural networks have been successfully applied to many NLP tasks the resulting vector-based models are very difficult to interpret. For example it's not clear how they achieve {\\em compositionality}, building sentence meaning from the meanings of words and phrases. In this paper we describe four strategies for visualizing compositionality in neural models for NLP, inspired by similar work in computer vision. We first plot unit values to visualize compositionality of negation, intensification, and concessive clauses, allow us to see well-known markedness asymmetries in negation. We then introduce three simple and straightforward methods for visualizing a unit's {\\em salience}, the amount it contributes to the final composed meaning: (1) gradient back-propagation, (2) the variance of a token from the average word node, (3) LSTM-style gates that measure information flow. We test our methods on sentiment using simple recurrent nets and LSTMs. Our general-purpose methods may have wide applications for understanding compositionality and other semantic properties of deep networks , and also shed light on why LSTMs outperform simple recurrent nets,"
                },
                {
                    "title": "Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps",
                    "abstract": "This paper addresses the visualisation of image classification models, learnt using deep Convolutional Networks (ConvNets). We consider two visualisation techniques, based on computing the gradient of the class score with respect to the input image. The first one generates an image, which maximises the class score [Erhan et al., 2009], thus visualising the notion of the class, captured by a ConvNet. The second technique computes a class saliency map, specific to a given image and class. We show that such maps can be employed for weakly supervised object segmentation using classification ConvNets. Finally, we establish the connection between the gradient-based ConvNet visualisation methods and deconvolutional networks [Zeiler et al., 2013]."
                },
                {
                    "title": "Scikit-learn: Machine Learning in Python",
                    "abstract": "Scikit-learn is a Python module integrating a wide range of state-of-the-art machine learning algorithms for medium-scale supervised and unsupervised problems. This package focuses on bringing machine learning to non-specialists using a general-purpose high-level language. Emphasis is put on ease of use, performance, documentation, and API consistency. It has minimal dependencies and is distributed under the simplified BSD license, encouraging its use in both academic and commercial settings. Source code, binaries, and documentation can be downloaded from http://scikit-learn.sourceforge.net."
                },
                {
                    "title": "Design and analysis of computer experiments",
                    "abstract": "In structural optimization, usually an approximation concept is introduced as interface between (FEM) structural analysis code and optimization algorithm. In some cases of opti-mm desigr., a glcba! apprcxhaticl? ccncept car. be effectively applied. Then, apprmirnatior. model functions are built of all objective and constraint functions, whose values, for a certain design point, follow from the structural analysis calculations. In this way, the original o p timization problem is completely replaced by an explicitly known approximate optimization problem. Response-surface techniques are commonly applied to build global approximation models, especially when dealing with responses of physical experiments. For one response function, this means that a user defined model function is fitted to the response data calculated at the design sites of some experimental design. Errors between model function and experimental response values are assumed to be randomly distributed. However, in the structural optimization case, an experiment is a computer analysis with a deterministic response as a result. This rules out the statistical assumptions response-surface model building is based upon. Furthermore, it is often difficult to find model functions that approximate the true response behaviour within the desired accuracy."
                },
                {
                    "title": "A Value for n-person Games",
                    "abstract": "Introduction At the foundation of the theory of games is the assumption that the players of a game can evaluate, in their utility scales, every \u201cprospect\u201d that might arise as a result of a play. In attempting to apply the theory to any field, one would normally expect to be permitted to include, in the class of \u201cprospects,\u201d the prospect of having to play a game. The possibility of evaluating games is therefore of critical importance. So long as the theory is unable to assign values to the games typically found in application, only relatively simple situations\u2014where games do not depend on other games\u2014will be susceptible to analysis and solution. In the finite theory of von Neumann and Morgenstern difficulty in evaluation persists for the \u201cessential\u201d games, and for only those. In this note we deduce a value for the \u201cessential\u201d case and examine a number of its elementary properties. We proceed from a set of three axioms, having simple intuitive interpretations, which suffice to determine the value uniquely. Our present work, though mathematically self-contained, is founded conceptually on the von Neumann\u2014Morgenstern theory up to their introduction of characteristic functions. We thereby inherit certain important underlying assumptions: (a) that utility is objective and transferable; (b) that games are cooperative affairs; (c) that games, granting (a) and (b), are adequately represented by their characteristic functions."
                },
                {
                    "title": "BERT meets Shapley: Extending SHAP Explanations to Transformer-based Classifiers",
                    "abstract": "Transformer-based neural networks offer very good classification performance across a wide range of domains, but do not provide explanations of their predictions. While several explanation methods, including SHAP, address the problem of interpreting deep learning models, they are not adapted to operate on state-of-the-art transformer-based neural networks such as BERT. Another shortcoming of these methods is that their visualization of explanations in the form of lists of most relevant words does not take into account the sequential and structurally dependent nature of text. This paper proposes the TransSHAP method that adapts SHAP to transformer models including BERT-based text classifiers. It advances SHAP visualizations by showing explanations in a sequential manner, assessed by human evaluators as competitive to state-of-the-art solutions."
                },
                {
                    "title": "Regression Shrinkage and Selection via the Lasso",
                    "abstract": "SUMMARY We propose a new method for estimation in linear models. The 'lasso' minimizes the residual sum of squares subject to the sum of the absolute value of the coefficients being less than a constant. Because of the nature of this constraint it tends to produce some coefficients that are exactly 0 and hence gives interpretable models. Our simulation studies suggest that the lasso enjoys some of the favourable properties of both subset selection and ridge regression. It produces interpretable models like subset selection and exhibits the stability of ridge regression. There is also an interesting relationship with recent work in adaptive function estimation by Donoho and Johnstone. The lasso idea is quite general and can be applied in a variety of statistical models: extensions to generalized regression models and tree-based models are briefly described."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan we pinpoint the parts of the context that led to a particular generated statement by a language model?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of context attribution is crucial for enhancing the reliability and transparency of language models. By accurately identifying which parts of the context influence generated responses, researchers can better understand model behavior, improve model training, and develop more robust applications. This advancement could lead to practical applications in fields such as automated content verification, where users can trust the accuracy of information provided by language models. Furthermore, it could foster future research into model interpretability and accountability, ultimately contributing to the responsible deployment of AI technologies.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving context attribution stem from the complexity of language models and their decision-making processes. Naive approaches may fail because they do not account for the intricate interactions between different parts of the context and the model's internal representations. Additionally, the lack of explicit mechanisms within language models to track the influence of specific context pieces complicates the attribution process. Technical obstacles include the need for efficient methods to evaluate the impact of context removal on generated responses, as well as the theoretical challenge of accurately modeling the relationship between context and output.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on generating citations rather than directly identifying the context used by language models. This focus has created a gap in understanding how models utilize context, as existing solutions often lack the necessary granularity to pinpoint specific influences. Barriers to solving this problem include the absence of formalized methodologies for context attribution and the challenges in evaluating attribution scores effectively. Our approach differs by introducing a formalized task of context attribution and leveraging a surrogate model to approximate the language model's behavior, providing a clearer and more efficient method for identifying context influence.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, ContextCite, involves creating a surrogate model that approximates how a language model's response is affected by including or excluding parts of the context. We will use a diverse set of generation tasks and evaluate the effectiveness of context attribution through metrics that measure the impact of context removal on generated responses. The expected outcomes include a set of attribution scores that accurately reflect the influence of different context parts, demonstrating that high-scoring context elements significantly affect the generated statements while low-scoring elements do"
            }
        },
        "author_data": {
            "e21a704c-3dc8-4570-9e65-90772c52fbc4": {
                "pk": "e21a704c-3dc8-4570-9e65-90772c52fbc4",
                "name": "Benjamin Cohen-Wang",
                "collaborators": [
                    "Joshua Vendrow",
                    "Aleksander Madry"
                ],
                "domain": [
                    "Machine Learning",
                    "Robustness",
                    "Distribution Shift",
                    "Pre-training"
                ],
                "publications": [
                    {
                        "title": "Ask Your Distribution Shift if Pre-Training is Right for You",
                        "abstract": "Pre-training is a widely used approach to develop models that are robust to distribution shifts. However, in practice, its effectiveness varies: fine-tuning a pre-trained model improves robustness significantly in some cases but not at all in others (compared to training from scratch). In this work, we seek to characterize the failure modes that pre-training can and cannot address. In particular, we focus on two possible failure modes of models under distribution shift: poor extrapolation (e.g., they cannot generalize to a different domain) and biases in the training data (e.g., they rely on spurious features). Our study suggests that, as a rule of thumb, pre-training can help mitigate poor extrapolation but not dataset biases. After providing theoretical motivation and empirical evidence for this finding, we explore two of its implications for developing robust models: (1) pre-training and interventions designed to prevent exploiting biases have complementary robustness benefits, and (2) fine-tuning on a (very) small, non-diverse but de-biased dataset can result in significantly more robust models than fine-tuning on a large and diverse but biased dataset. Code is available at https://github.com/MadryLab/pretraining-distribution-shift-robustness."
                    }
                ]
            },
            "dd3b111c-5e13-41ef-a33b-757e7113ea70": {
                "pk": "dd3b111c-5e13-41ef-a33b-757e7113ea70",
                "name": "Harshay Shah",
                "collaborators": [
                    "Andrew Ilyas",
                    "Aleksander Madry",
                    "Prateek Jain",
                    "Praneeth Netrapalli",
                    "Suhansanu Kumar",
                    "Hari Sundaram",
                    "Ashish Khetan",
                    "Sewoong Oh",
                    "Kaustav Tamuly",
                    "Aditi Raghunathan"
                ],
                "domain": [
                    "Graph Theory",
                    "Machine Learning",
                    "Interpretability",
                    "Network Analysis"
                ],
                "publications": [
                    {
                        "title": "Growing Attributed Networks through Local Processes",
                        "abstract": "This paper proposes an attributed network growth model. Despite the knowledge that individuals use limited resources to form connections to similar others, we lack an understanding of how local and resource-constrained mechanisms explain the emergence of rich structural properties found in real-world networks. We make three contributions. First, we propose a parsimonious and accurate model of attributed network growth that jointly explains the emergence of in-degree distributions, local clustering, clustering-degree relationship and attribute mixing patterns. Second, our model is based on biased random walks and uses local processes to form edges without recourse to global network information. Third, we account for multiple sociological phenomena: bounded rationality, structural constraints, triadic closure, attribute homophily, and preferential attachment. Our experiments indicate that the proposed Attributed Random Walk (ARW) model accurately preserves network structure and attribute mixing patterns of six real-world networks; it improves upon the performance of eight state-of-the-art models by a statistically significant margin of 2.5-10x."
                    },
                    {
                        "title": "Decomposing and Editing Predictions by Modeling Model Computation",
                        "abstract": "How does the internal computation of a machine learning model transform inputs into predictions? In this paper, we introduce a task called component modeling that aims to address this question. The goal of component modeling is to decompose an ML model's prediction in terms of its components -- simple functions (e.g., convolution filters, attention heads) that are the \"building blocks\" of model computation. We focus on a special case of this task, component attribution, where the goal is to estimate the counterfactual impact of individual components on a given prediction. We then present COAR, a scalable algorithm for estimating component attributions; we demonstrate its effectiveness across models, datasets, and modalities. Finally, we show that component attributions estimated with COAR directly enable model editing across five tasks, namely: fixing model errors, ``forgetting'' specific classes, boosting subpopulation robustness, localizing backdoor attacks, and improving robustness to typographic attacks. We provide code for COAR at https://github.com/MadryLab/modelcomponents ."
                    },
                    {
                        "title": "Do Input Gradients Highlight Discriminative Features?",
                        "abstract": "Post-hoc gradient-based interpretability methods [Simonyan et al., 2013, Smilkov et al., 2017] that provide instance-specific explanations of model predictions are often based on assumption (A): magnitude of input gradients -- gradients of logits with respect to input -- noisily highlight discriminative task-relevant features. In this work, we test the validity of assumption (A) using a three-pronged approach. First, we develop an evaluation framework, DiffROAR, to test assumption (A) on four image classification benchmarks. Our results suggest that (i) input gradients of standard models (i.e., trained on original data) may grossly violate (A), whereas (ii) input gradients of adversarially robust models satisfy (A). Second, we introduce BlockMNIST, an MNIST-based semi-real dataset, that by design encodes a priori knowledge of discriminative features. Our analysis on BlockMNIST leverages this information to validate as well as characterize differences between input gradient attributions of standard and robust models. Finally, we theoretically prove that our empirical findings hold on a simplified version of the BlockMNIST dataset. Specifically, we prove that input gradients of standard one-hidden-layer MLPs trained on this dataset do not highlight instance-specific signal coordinates, thus grossly violating assumption (A). Our findings motivate the need to formalize and test common assumptions in interpretability in a falsifiable manner [Leavitt and Morcos, 2020]. We believe that the DiffROAR evaluation framework and BlockMNIST-based datasets can serve as sanity checks to audit instance-specific interpretability methods; code and data available at https://github.com/harshays/inputgradients."
                    },
                    {
                        "title": "Number of Connected Components in a Graph: Estimation via Counting Patterns",
                        "abstract": "Due to the limited resources and the scale of the graphs in modern datasets, we often get to observe a sampled subgraph of a larger original graph of interest, whether it is the worldwide web that has been crawled or social connections that have been surveyed. Inferring a global property of the original graph from such a sampled subgraph is of a fundamental interest. In this work, we focus on estimating the number of connected components. It is a challenging problem and, for general graphs, little is known about the connection between the observed subgraph and the number of connected components of the original graph. In order to make this connection, we propose a highly redundant and large-dimensional representation of the subgraph, which at first glance seems counter-intuitive. A subgraph is represented by the counts of patterns, known as network motifs. This representation is crucial in introducing a novel estimator for the number of connected components for general graphs, under the knowledge of the spectral gap of the original graph. The connection is made precise via the Schatten $k$-norms of the graph Laplacian and the spectral representation of the number of connected components. We provide a guarantee on the resulting mean squared error that characterizes the bias variance tradeoff. Experiments on synthetic and real-world graphs suggest that we improve upon competing algorithms for graphs with spectral gaps bounded away from zero."
                    },
                    {
                        "title": "The Pitfalls of Simplicity Bias in Neural Networks",
                        "abstract": "Several works have proposed Simplicity Bias (SB)---the tendency of standard training procedures such as Stochastic Gradient Descent (SGD) to find simple models---to justify why neural networks generalize well [Arpit et al. 2017, Nakkiran et al. 2019, Soudry et al. 2018]. However, the precise notion of simplicity remains vague. Furthermore, previous settings that use SB to theoretically justify why neural networks generalize well do not simultaneously capture the non-robustness of neural networks---a widely observed phenomenon in practice [Goodfellow et al. 2014, Jo and Bengio 2017]. We attempt to reconcile SB and the superior standard generalization of neural networks with the non-robustness observed in practice by designing datasets that (a) incorporate a precise notion of simplicity, (b) comprise multiple predictive features with varying levels of simplicity, and (c) capture the non-robustness of neural networks trained on real data. Through theory and empirics on these datasets, we make four observations: (i) SB of SGD and variants can be extreme: neural networks can exclusively rely on the simplest feature and remain invariant to all predictive complex features. (ii) The extreme aspect of SB could explain why seemingly benign distribution shifts and small adversarial perturbations significantly degrade model performance. (iii) Contrary to conventional wisdom, SB can also hurt generalization on the same data distribution, as SB persists even when the simplest feature has less predictive power than the more complex features. (iv) Common approaches to improve generalization and robustness---ensembles and adversarial training---can fail in mitigating SB and its pitfalls. Given the role of SB in training neural networks, we hope that the proposed datasets and methods serve as an effective testbed to evaluate novel algorithmic approaches aimed at avoiding the pitfalls of SB."
                    },
                    {
                        "title": "ModelDiff: A Framework for Comparing Learning Algorithms",
                        "abstract": "We study the problem of (learning) algorithm comparison, where the goal is to find differences between models trained with two different learning algorithms. We begin by formalizing this goal as one of finding distinguishing feature transformations, i.e., input transformations that change the predictions of models trained with one learning algorithm but not the other. We then present ModelDiff, a method that leverages the datamodels framework (Ilyas et al., 2022) to compare learning algorithms based on how they use their training data. We demonstrate ModelDiff through three case studies, comparing models trained with/without data augmentation, with/without pre-training, and with different SGD hyperparameters. Our code is available at https://github.com/MadryLab/modeldiff ."
                    }
                ]
            },
            "e9044f53-9810-47f6-92e1-1125696ef65c": {
                "pk": "e9044f53-9810-47f6-92e1-1125696ef65c",
                "name": "Kristian Georgiev",
                "collaborators": [
                    "Aleksander Madry",
                    "Andrew Ilyas",
                    "Hadi Salman",
                    "Sung Min Park",
                    "Guillaume Leclerc",
                    "Samuel B. Hopkins",
                    "Alireza Fallah",
                    "Aryan Mokhtari",
                    "Asuman Ozdaglar",
                    "Saachi Jain"
                ],
                "domain": [
                    "Differential Privacy",
                    "Robust Statistics",
                    "Meta-Learning",
                    "Data Attribution"
                ],
                "publications": [
                    {
                        "title": "Privacy Induces Robustness: Information-Computation Gaps and Sparse Mean Estimation",
                        "abstract": "We establish a simple connection between robust and differentially-private algorithms: private mechanisms which perform well with very high probability are automatically robust in the sense that they retain accuracy even if a constant fraction of the samples they receive are adversarially corrupted. Since optimal mechanisms typically achieve these high success probabilities, our results imply that optimal private mechanisms for many basic statistics problems are robust.   We investigate the consequences of this observation for both algorithms and computational complexity across different statistical problems. Assuming the Brennan-Bresler secret-leakage planted clique conjecture, we demonstrate a fundamental tradeoff between computational efficiency, privacy leakage, and success probability for sparse mean estimation. Private algorithms which match this tradeoff are not yet known -- we achieve that (up to polylogarithmic factors) in a polynomially-large range of parameters via the Sum-of-Squares method.   To establish an information-computation gap for private sparse mean estimation, we also design new (exponential-time) mechanisms using fewer samples than efficient algorithms must use. Finally, we give evidence for privacy-induced information-computation gaps for several other statistics and learning problems, including PAC learning parity functions and estimation of the mean of a multivariate Gaussian."
                    },
                    {
                        "title": "On the Convergence Theory of Debiased Model-Agnostic Meta-Reinforcement Learning",
                        "abstract": "We consider Model-Agnostic Meta-Learning (MAML) methods for Reinforcement Learning (RL) problems, where the goal is to find a policy using data from several tasks represented by Markov Decision Processes (MDPs) that can be updated by one step of stochastic policy gradient for the realized MDP. In particular, using stochastic gradients in MAML update steps is crucial for RL problems since computation of exact gradients requires access to a large number of possible trajectories. For this formulation, we propose a variant of the MAML method, named Stochastic Gradient Meta-Reinforcement Learning (SG-MRL), and study its convergence properties. We derive the iteration and sample complexity of SG-MRL to find an $\\epsilon$-first-order stationary point, which, to the best of our knowledge, provides the first convergence guarantee for model-agnostic meta-reinforcement learning algorithms. We further show how our results extend to the case where more than one step of stochastic policy gradient method is used at test time. Finally, we empirically compare SG-MRL and MAML in several deep RL environments."
                    },
                    {
                        "title": "Data Debiasing with Datamodels (D3M): Improving Subgroup Robustness via Data Selection",
                        "abstract": "Machine learning models can fail on subgroups that are underrepresented during training. While techniques such as dataset balancing can improve performance on underperforming groups, they require access to training group annotations and can end up removing large portions of the dataset. In this paper, we introduce Data Debiasing with Datamodels (D3M), a debiasing approach which isolates and removes specific training examples that drive the model's failures on minority groups. Our approach enables us to efficiently train debiased classifiers while removing only a small number of examples, and does not require training group annotations or additional hyperparameter tuning."
                    },
                    {
                        "title": "The Journey, Not the Destination: How Data Guides Diffusion Models",
                        "abstract": "Diffusion models trained on large datasets can synthesize photo-realistic images of remarkable quality and diversity. However, attributing these images back to the training data-that is, identifying specific training examples which caused an image to be generated-remains a challenge. In this paper, we propose a framework that: (i) provides a formal notion of data attribution in the context of diffusion models, and (ii) allows us to counterfactually validate such attributions. Then, we provide a method for computing these attributions efficiently. Finally, we apply our method to find (and evaluate) such attributions for denoising diffusion probabilistic models trained on CIFAR-10 and latent diffusion models trained on MS COCO. We provide code at https://github.com/MadryLab/journey-TRAK ."
                    },
                    {
                        "title": "Implicit Bias of Linear Equivariant Networks",
                        "abstract": "Group equivariant convolutional neural networks (G-CNNs) are generalizations of convolutional neural networks (CNNs) which excel in a wide range of technical applications by explicitly encoding symmetries, such as rotations and permutations, in their architectures. Although the success of G-CNNs is driven by their \\emph{explicit} symmetry bias, a recent line of work has proposed that the \\emph{implicit} bias of training algorithms on particular architectures is key to understanding generalization for overparameterized neural nets. In this context, we show that $L$-layer full-width linear G-CNNs trained via gradient descent for binary classification converge to solutions with low-rank Fourier matrix coefficients, regularized by the $2/L$-Schatten matrix norm. Our work strictly generalizes previous analysis on the implicit bias of linear CNNs to linear G-CNNs over all finite groups, including the challenging setting of non-commutative groups (such as permutations), as well as band-limited G-CNNs over infinite groups. We validate our theorems via experiments on a variety of groups, and empirically explore more realistic nonlinear networks, which locally capture similar regularization patterns. Finally, we provide intuitive interpretations of our Fourier space implicit regularization results in real space via uncertainty principles."
                    },
                    {
                        "title": "TRAK: Attributing Model Behavior at Scale",
                        "abstract": "The goal of data attribution is to trace model predictions back to training data. Despite a long line of work towards this goal, existing approaches to data attribution tend to force users to choose between computational tractability and efficacy. That is, computationally tractable methods can struggle with accurately attributing model predictions in non-convex settings (e.g., in the context of deep neural networks), while methods that are effective in such regimes require training thousands of models, which makes them impractical for large models or datasets.   In this work, we introduce TRAK (Tracing with the Randomly-projected After Kernel), a data attribution method that is both effective and computationally tractable for large-scale, differentiable models. In particular, by leveraging only a handful of trained models, TRAK can match the performance of attribution methods that require training thousands of models. We demonstrate the utility of TRAK across various modalities and scales: image classifiers trained on ImageNet, vision-language models (CLIP), and language models (BERT and mT5). We provide code for using TRAK (and reproducing our work) at https://github.com/MadryLab/trak ."
                    },
                    {
                        "title": "Rethinking Backdoor Attacks",
                        "abstract": "In a backdoor attack, an adversary inserts maliciously constructed backdoor examples into a training set to make the resulting model vulnerable to manipulation. Defending against such attacks typically involves viewing these inserted examples as outliers in the training set and using techniques from robust statistics to detect and remove them.   In this work, we present a different approach to the backdoor attack problem. Specifically, we show that without structural information about the training data distribution, backdoor attacks are indistinguishable from naturally-occurring features in the data--and thus impossible to \"detect\" in a general sense. Then, guided by this observation, we revisit existing defenses against backdoor attacks and characterize the (often latent) assumptions they make and on which they depend. Finally, we explore an alternative perspective on backdoor attacks: one that assumes these attacks correspond to the strongest feature in the training data. Under this assumption (which we make formal) we develop a new primitive for detecting backdoor attacks. Our primitive naturally gives rise to a detection algorithm that comes with theoretical guarantees and is effective in practice."
                    }
                ]
            },
            "f91a0bb7-63ea-43c2-9afc-09c543daa8ab": {
                "pk": "f91a0bb7-63ea-43c2-9afc-09c543daa8ab",
                "name": "Aleksander Madry",
                "collaborators": [
                    "Dimitris Tsipras",
                    "Shibani Santurkar",
                    "Andrew Ilyas",
                    "Katarzyna Paluch",
                    "Marcin Mucha",
                    "Guillaume Leclerc",
                    "Nikhil Bansal",
                    "Niv Buchbinder",
                    "Joseph",
                    "Naor"
                ],
                "domain": [
                    "Combinatorial Optimization",
                    "Graph Algorithms",
                    "Adversarial Machine Learning",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Faster Approximation Schemes for Fractional Multicommodity Flow Problems via Dynamic Graph Algorithms",
                        "abstract": "We combine the work of Garg and Konemann, and Fleischer with ideas from dynamic graph algorithms to obtain faster (1-eps)-approximation schemes for various versions of the multicommodity flow problem. In particular, if eps is moderately small and the size of every number used in the input instance is polynomially bounded, the running times of our algorithms match - up to poly-logarithmic factors and some provably optimal terms - the Omega(mn) flow-decomposition barrier for single-commodity flow."
                    },
                    {
                        "title": "Navigating Central Path with Electrical Flows: from Flows to Matchings, and Back",
                        "abstract": "We present an $\\tilde{O}(m^{10/7})=\\tilde{O}(m^{1.43})$-time algorithm for the maximum s-t flow and the minimum s-t cut problems in directed graphs with unit capacities. This is the first improvement over the sparse-graph case of the long-standing $O(m \\min(\\sqrt{m},n^{2/3}))$ time bound due to Even and Tarjan [EvenT75]. By well-known reductions, this also establishes an $\\tilde{O}(m^{10/7})$-time algorithm for the maximum-cardinality bipartite matching problem. That, in turn, gives an improvement over the celebrated celebrated $O(m \\sqrt{n})$ time bound of Hopcroft and Karp [HK73] whenever the input graph is sufficiently sparse."
                    },
                    {
                        "title": "Computing Maximum Flow with Augmenting Electrical Flows",
                        "abstract": "We present an $\\tilde{O}\\left(m^{\\frac{10}{7}}U^{\\frac{1}{7}}\\right)$-time algorithm for the maximum $s$-$t$ flow problem and the minimum $s$-$t$ cut problem in directed graphs with $m$ arcs and largest integer capacity $U$. This matches the running time of the $\\tilde{O}\\left((mU)^{\\frac{10}{7}}\\right)$-time algorithm of M\\k{a}dry (FOCS 2013) in the unit-capacity case, and improves over it, as well as over the $\\tilde{O}\\left(m \\sqrt{n} \\log U\\right)$-time algorithm of Lee and Sidford (FOCS 2014), whenever $U$ is moderately large and the graph is sufficiently sparse. By well-known reductions, this also gives similar running time improvements for the maximum-cardinality bipartite $b$-matching problem.   One of the advantages of our algorithm is that it is significantly simpler than the ones presented in Madry (FOCS 2013) and Lee and Sidford (FOCS 2014). In particular, these algorithms employ a sophisticated interior-point method framework, while our algorithm is cast directly in the classic augmenting path setting that almost all the combinatorial maximum flow algorithms use. At a high level, the presented algorithm takes a primal dual approach in which each iteration uses electrical flows computations both to find an augmenting $s$-$t$ flow in the current residual graph and to update the dual solution. We show that by maintain certain careful coupling of these primal and dual solutions we are always guaranteed to make significant progress."
                    },
                    {
                        "title": "Fast Approximation Algorithms for Cut-based Problems in Undirected Graphs",
                        "abstract": "We present a general method of designing fast approximation algorithms for cut-based minimization problems in undirected graphs. In particular, we develop a technique that given any such problem that can be approximated quickly on trees, allows approximating it almost as quickly on general graphs while only losing a poly-logarithmic factor in the approximation guarantee.   To illustrate the applicability of our paradigm, we focus our attention on the undirected sparsest cut problem with general demands and the balanced separator problem. By a simple use of our framework, we obtain poly-logarithmic approximation algorithms for these problems that run in time close to linear.   The main tool behind our result is an efficient procedure that decomposes general graphs into simpler ones while approximately preserving the cut-flow structure. This decomposition is inspired by the cut-based graph decomposition of R\\\"acke that was developed in the context of oblivious routing schemes, as well as, by the construction of the ultrasparsifiers due to Spielman and Teng that was employed to preconditioning symmetric diagonally-dominant matrices."
                    },
                    {
                        "title": "A 7/9 - Approximation Algorithm for the Maximum Traveling Salesman Problem",
                        "abstract": "We give a 7/9 - Approximation Algorithm for the Maximum Traveling Salesman Problem."
                    },
                    {
                        "title": "The Two Regimes of Deep Network Training",
                        "abstract": "Learning rate schedule has a major impact on the performance of deep learning models. Still, the choice of a schedule is often heuristical. We aim to develop a precise understanding of the effects of different learning rate schedules and the appropriate way to select them. To this end, we isolate two distinct phases of training, the first, which we refer to as the \"large-step\" regime, exhibits a rather poor performance from an optimization point of view but is the primary contributor to model generalization; the latter, \"small-step\" regime exhibits much more \"convex-like\" optimization behavior but used in isolation produces models that generalize poorly. We find that by treating these regimes separately-and em specializing our training algorithm to each one of them, we can significantly simplify learning rate schedules."
                    },
                    {
                        "title": "A Polylogarithmic-Competitive Algorithm for the k-Server Problem",
                        "abstract": "We give the first polylogarithmic-competitive randomized online algorithm for the $k$-server problem on an arbitrary finite metric space. In particular, our algorithm achieves a competitive ratio of O(log^3 n log^2 k log log n) for any metric space on n points. Our algorithm improves upon the deterministic (2k-1)-competitive algorithm of Koutsoupias and Papadimitriou [J.ACM'95] whenever n is sub-exponential in k."
                    },
                    {
                        "title": "Combining Diverse Feature Priors",
                        "abstract": "To improve model generalization, model designers often restrict the features that their models use, either implicitly or explicitly. In this work, we explore the design space of leveraging such feature priors by viewing them as distinct perspectives on the data. Specifically, we find that models trained with diverse sets of feature priors have less overlapping failure modes, and can thus be combined more effectively. Moreover, we demonstrate that jointly training such models on additional (unlabeled) data allows them to correct each other's mistakes, which, in turn, leads to better generalization and resilience to spurious correlations. Code available at https://github.com/MadryLab/copriors"
                    },
                    {
                        "title": "Negative-Weight Shortest Paths and Unit Capacity Minimum Cost Flow in $\\tilde{O}(m^{10/7} \\log W)$ Time",
                        "abstract": "In this paper, we study a set of combinatorial optimization problems on weighted graphs: the shortest path problem with negative weights, the weighted perfect bipartite matching problem, the unit-capacity minimum-cost maximum flow problem and the weighted perfect bipartite $b$-matching problem under the assumption that $\\Vert b\\Vert_1=O(m)$. We show that each one of these four problems can be solved in $\\tilde{O}(m^{10/7}\\log W)$ time, where $W$ is the absolute maximum weight of an edge in the graph, which gives the first in over 25 years polynomial improvement in their sparse-graph time complexity.   At a high level, our algorithms build on the interior-point method-based framework developed by Madry (FOCS 2013) for solving unit-capacity maximum flow problem. We develop a refined way to analyze this framework, as well as provide new variants of the underlying preconditioning and perturbation techniques. Consequently, we are able to extend the whole interior-point method-based approach to make it applicable in the weighted graph regime."
                    },
                    {
                        "title": "Faster generation of random spanning trees",
                        "abstract": "In this paper, we set forth a new algorithm for generating approximately uniformly random spanning trees in undirected graphs. We show how to sample from a distribution that is within a multiplicative $(1+\\delta)$ of uniform in expected time $\\TO(m\\sqrt{n}\\log 1/\\delta)$. This improves the sparse graph case of the best previously known worst-case bound of $O(\\min \\{mn, n^{2.376}\\})$, which has stood for twenty years.   To achieve this goal, we exploit the connection between random walks on graphs and electrical networks, and we use this to introduce a new approach to the problem that integrates discrete random walk-based techniques with continuous linear algebraic methods. We believe that our use of electrical networks and sparse linear system solvers in conjunction with random walks and combinatorial partitioning techniques is a useful paradigm that will find further applications in algorithmic graph theory."
                    },
                    {
                        "title": "How Does Batch Normalization Help Optimization?",
                        "abstract": "Batch Normalization (BatchNorm) is a widely adopted technique that enables faster and more stable training of deep neural networks (DNNs). Despite its pervasiveness, the exact reasons for BatchNorm's effectiveness are still poorly understood. The popular belief is that this effectiveness stems from controlling the change of the layers' input distributions during training to reduce the so-called \"internal covariate shift\". In this work, we demonstrate that such distributional stability of layer inputs has little to do with the success of BatchNorm. Instead, we uncover a more fundamental impact of BatchNorm on the training process: it makes the optimization landscape significantly smoother. This smoothness induces a more predictive and stable behavior of the gradients, allowing for faster training."
                    },
                    {
                        "title": "Spectral Signatures in Backdoor Attacks",
                        "abstract": "A recent line of work has uncovered a new form of data poisoning: so-called \\emph{backdoor} attacks. These attacks are particularly dangerous because they do not affect a network's behavior on typical, benign data. Rather, the network only deviates from its expected output when triggered by a perturbation planted by an adversary.   In this paper, we identify a new property of all known backdoor attacks, which we call \\emph{spectral signatures}. This property allows us to utilize tools from robust statistics to thwart the attacks. We demonstrate the efficacy of these signatures in detecting and removing poisoned examples on real image sets and state of the art neural network architectures. We believe that understanding spectral signatures is a crucial first step towards designing ML systems secure against such backdoor attacks"
                    },
                    {
                        "title": "On Adaptive Attacks to Adversarial Example Defenses",
                        "abstract": "Adaptive attacks have (rightfully) become the de facto standard for evaluating defenses to adversarial examples. We find, however, that typical adaptive evaluations are incomplete. We demonstrate that thirteen defenses recently published at ICLR, ICML and NeurIPS---and chosen for illustrative and pedagogical purposes---can be circumvented despite attempting to perform evaluations using adaptive attacks. While prior evaluation papers focused mainly on the end result---showing that a defense was ineffective---this paper focuses on laying out the methodology and the approach necessary to perform an adaptive attack. We hope that these analyses will serve as guidance on how to properly perform adaptive attacks against defenses to adversarial examples, and thus will allow the community to make further progress in building more robust models."
                    },
                    {
                        "title": "BREEDS: Benchmarks for Subpopulation Shift",
                        "abstract": "We develop a methodology for assessing the robustness of models to subpopulation shift---specifically, their ability to generalize to novel data subpopulations that were not observed during training. Our approach leverages the class structure underlying existing datasets to control the data subpopulations that comprise the training and test distributions. This enables us to synthesize realistic distribution shifts whose sources can be precisely controlled and characterized, within existing large-scale datasets. Applying this methodology to the ImageNet dataset, we create a suite of subpopulation shift benchmarks of varying granularity. We then validate that the corresponding shifts are tractable by obtaining human baselines for them. Finally, we utilize these benchmarks to measure the sensitivity of standard model architectures as well as the effectiveness of off-the-shelf train-time robustness interventions. Code and data available at https://github.com/MadryLab/BREEDS-Benchmarks ."
                    },
                    {
                        "title": "Fast Generation of Random Spanning Trees and the Effective Resistance Metric",
                        "abstract": "We present a new algorithm for generating a uniformly random spanning tree in an undirected graph. Our algorithm samples such a tree in expected $\\tilde{O}(m^{4/3})$ time. This improves over the best previously known bound of $\\min(\\tilde{O}(m\\sqrt{n}),O(n^{\\omega}))$ -- that follows from the work of Kelner and M\\k{a}dry [FOCS'09] and of Colbourn et al. [J. Algorithms'96] -- whenever the input graph is sufficiently sparse.   At a high level, our result stems from carefully exploiting the interplay of random spanning trees, random walks, and the notion of effective resistance, as well as from devising a way to algorithmically relate these concepts to the combinatorial structure of the graph. This involves, in particular, establishing a new connection between the effective resistance metric and the cut structure of the underlying graph."
                    },
                    {
                        "title": "Label-Consistent Backdoor Attacks",
                        "abstract": "Deep neural networks have been demonstrated to be vulnerable to backdoor attacks. Specifically, by injecting a small number of maliciously constructed inputs into the training set, an adversary is able to plant a backdoor into the trained model. This backdoor can then be activated during inference by a backdoor trigger to fully control the model's behavior. While such attacks are very effective, they crucially rely on the adversary injecting arbitrary inputs that are---often blatantly---mislabeled. Such samples would raise suspicion upon human inspection, potentially revealing the attack. Thus, for backdoor attacks to remain undetected, it is crucial that they maintain label-consistency---the condition that injected inputs are consistent with their labels. In this work, we leverage adversarial perturbations and generative models to execute efficient, yet label-consistent, backdoor attacks. Our approach is based on injecting inputs that appear plausible, yet are hard to classify, hence causing the model to rely on the (easier-to-learn) backdoor trigger."
                    },
                    {
                        "title": "Prior Convictions: Black-Box Adversarial Attacks with Bandits and Priors",
                        "abstract": "We study the problem of generating adversarial examples in a black-box setting in which only loss-oracle access to a model is available. We introduce a framework that conceptually unifies much of the existing work on black-box attacks, and we demonstrate that the current state-of-the-art methods are optimal in a natural sense. Despite this optimality, we show how to improve black-box attacks by bringing a new element into the problem: gradient priors. We give a bandit optimization-based algorithm that allows us to seamlessly integrate any such priors, and we explicitly identify and incorporate two examples. The resulting methods use two to four times fewer queries and fail two to five times less often than the current state-of-the-art."
                    }
                ]
            }
        }
    },
    "2404.13076": {
        "paper_data": {
            "title": "LLM Evaluators Recognize and Favor Their Own Generations",
            "url": "http://arxiv.org/abs/2404.13076v1",
            "arxiv_id": "2404.13076",
            "authors": [
                "Arjun Panickssery",
                "Samuel R. Bowman",
                "Shi Feng"
            ],
            "abstract": "Self-evaluation using large language models (LLMs) has proven valuable not only in benchmarking but also methods like reward modeling, constitutional AI, and self-refinement. But new biases are introduced due to the same LLM acting as both the evaluator and the evaluatee. One such bias is self-preference, where an LLM evaluator scores its own outputs higher than others' while human annotators consider them of equal quality. But do LLMs actually recognize their own outputs when they give those texts higher scores, or is it just a coincidence? In this paper, we investigate if self-recognition capability contributes to self-preference. We discover that, out of the box, LLMs such as GPT-4 and Llama 2 have non-trivial accuracy at distinguishing themselves from other LLMs and humans. By fine-tuning LLMs, we discover a linear correlation between self-recognition capability and the strength of self-preference bias; using controlled experiments, we show that the causal explanation resists straightforward confounders. We discuss how self-recognition can interfere with unbiased evaluations and AI safety more generally.",
            "introduction": "   1 Introduction  Self-evaluation is becoming a prominent part of the large language model (LLM) lifecycle. In methods like reward modeling\u00a0(Leike et\u00a0al., 2018; Stiennon et\u00a0al., 2020), model-based benchmarks\u00a0(Shashidhar et\u00a0al., 2023; Zeng et\u00a0al., 2023; Yuan et\u00a0al., 2023; Fu et\u00a0al., 2023; Li et\u00a0al., 2024), self-refinement\u00a0(Saunders et\u00a0al., 2022; Madaan et\u00a0al., 2023; Lee et\u00a0al., 2023; Shridhar et\u00a0al., 2023), and constitutional AI\u00a0(Bai et\u00a0al., 2022), LLMs are increasingly used to provide assessment, supervision, and oversight for themselves and other LLMs. LLM evaluators are shown to be highly accurate at approximating human annotators on various tasks, and are significantly more scalable (Hackl et\u00a0al., 2023).   Figure 1: The strength of self-preference bias is linearly correlated with the LLM\u2019s self-recognition capability. Each point represents a model evaluated on the two properties using the CNN/Dailymail dataset. We fine-tune GPT-3.5 and Llama 2 for self-recognition or control tasks using both CNN/Dailymail (in-domain) and XSUM (out-of-domain). The scores represented by both axes can be interpreted as measures of the LLM\u2019s confidence on these properties.   In self-evaluation, as the name suggests, the same underlying LLM acts as both the evaluator and the evaluatee. As a result, the neutrality of the evaluator is in question, and the evaluation can suffer from biases where the LLM evaluators diverge from humans in systematic ways\u00a0(Zheng et\u00a0al., 2024; Bai et\u00a0al., 2024). One such bias is self-preference, where an LLM rates its own outputs higher than texts written by other LLMs or humans, while human annotators judge them as equal quality. Self-preference has been observed in GPT-4-based dialogue benchmarks\u00a0(Bitton et\u00a0al., 2023b; Koo et\u00a0al., 2023), as well as for text summarization\u00a0(Liu et\u00a0al., 2023).   Towards understanding and mitigating self-preference, we study self-recognition\u2014an LLM\u2019s capability of recognizing its own outputs. We ask: Is self-preference truly self-preference, in the sense that the LLM prefers a text because it was generated by itself?   We measure their correlation while using prompting and fine-tuning to alter the LLM\u2019s self-recognition capability. In order to provide signals for the causal link between self-recognition and self-preference, we also fine-tune the LLM on a comprehensive set of potential confounding properties.    Our main findings are as follows:     1.  Frontier LLMs exhibit self-preference in self-evaluation. On two summarization tasks, LLMs (GPT-3.5 Turbo, GPT-4, and Llama 2) disproportionately favor summaries written by themselves over those by other LLMs and from humans.    2.  LLMs have non-trivial self-recognition capability out of the box. All three LLMs we evaluate achieve over 50%percent5050\\%50 % accuracy at distinguishing their own outputs from other sources using simple prompts without fine-tuning. GPT-4 is 73.5%percent73.573.5\\%73.5 % accurate distinguishing itself from two other LLMs and humans.    3.  Fine-tuning leads to near-perfect self-recognition. GPT-3.5 and Llama 2 both achieve over 90%percent9090\\%90 % accuracy at self-recognition after fine-tuning on 500 examples.    4.  Self-preference strength is linearly correlated with self-recognition. We further fine-tune LLMs to increase or decrease self-recognition, and discover a linear correlation between these two properties (Figure\u00a01).        2 Definition and Measurement of Self-Preference and Self-Recognition  Self-preference is the phenomenon in which an LLM favors its own outputs over texts from other LLMs and humans.   Self-recognition is the capability of an LLM to distinguish its own outputs from texts from other LLMs or by humans.   For both definitions, we follow the prosaic rather than the intentional interpretation. That is, we use the term",
            "references": [
                {
                    "title": "A Survey of AI-generated Text Forensic Systems: Detection, Attribution, and Characterization",
                    "abstract": "We have witnessed lately a rapid proliferation of advanced Large Language Models (LLMs) capable of generating high-quality text. While these LLMs have revolutionized text generation across various domains, they also pose significant risks to the information ecosystem, such as the potential for generating convincing propaganda, misinformation, and disinformation at scale. This paper offers a review of AI-generated text forensic systems, an emerging field addressing the challenges of LLM misuses. We present an overview of the existing efforts in AI-generated text forensics by introducing a detailed taxonomy, focusing on three primary pillars: detection, attribution, and characterization. These pillars enable a practical understanding of AI-generated text, from identifying AI-generated content (detection), determining the specific AI model involved (attribution), and grouping the underlying intents of the text (characterization). Furthermore, we explore available resources for AI-generated text forensics research and discuss the evolving challenges and future directions of forensic systems in an AI era."
                },
                {
                    "title": "Is LLM-as-a-Judge Robust? Investigating Universal Adversarial Attacks on Zero-shot LLM Assessment",
                    "abstract": "Large Language Models (LLMs) are powerful zero-shot assessors used in real-world situations such as assessing written exams and benchmarking systems. Despite these critical applications, no existing work has analyzed the vulnerability of judge-LLMs to adversarial manipulation. This work presents the first study on the adversarial robustness of assessment LLMs, where we demonstrate that short universal adversarial phrases can be concatenated to deceive judge LLMs to predict inflated scores. Since adversaries may not know or have access to the judge-LLMs, we propose a simple surrogate attack where a surrogate model is first attacked, and the learned attack phrase then transferred to unknown judge-LLMs. We propose a practical algorithm to determine the short universal attack phrases and demonstrate that when transferred to unseen models, scores can be drastically inflated such that irrespective of the assessed text, maximum scores are predicted. It is found that judge-LLMs are significantly more susceptible to these adversarial attacks when used for absolute scoring, as opposed to comparative assessment. Our findings raise concerns on the reliability of LLM-as-a-judge methods, and emphasize the importance of addressing vulnerabilities in LLM assessment methods before deployment in high-stakes real-world scenarios."
                },
                {
                    "title": "Pride and Prejudice: LLM Amplifies Self-Bias in Self-Refinement",
                    "abstract": "Recent studies show that large language models (LLMs) improve their performance through self-feedback on certain tasks while degrade on others. We discovered that such a contrary is due to LLM's bias in evaluating their own output. In this paper, we formally define LLM's self-bias - the tendency to favor its own generation - using two statistics. We analyze six LLMs (GPT-4, GPT-3.5, Gemini, LLaMA2, Mixtral and DeepSeek) on translation, constrained text generation, and mathematical reasoning tasks. We find that self-bias is prevalent in all examined LLMs across multiple languages and tasks. Our analysis reveals that while the self-refine pipeline improves the fluency and understandability of model outputs, it further amplifies self-bias. To mitigate such biases, we discover that larger model size and external feedback with accurate assessment can significantly reduce bias in the self-refine pipeline, leading to actual performance improvement in downstream tasks. The code and data are released at https://github.com/xu1998hz/llm_self_bias."
                },
                {
                    "title": "Feedback Loops With Language Models Drive In-Context Reward Hacking",
                    "abstract": "Language models influence the external world: they query APIs that read and write to web pages, generate content that shapes human behavior, and run system commands as autonomous agents. These interactions form feedback loops: LLM outputs affect the world, which in turn affect subsequent LLM outputs. In this work, we show that feedback loops can cause in-context reward hacking (ICRH), where the LLM at test-time optimizes a (potentially implicit) objective but creates negative side effects in the process. For example, consider an LLM agent deployed to increase Twitter engagement; the LLM may retrieve its previous tweets into the context window and make them more controversial, increasing engagement but also toxicity. We identify and study two processes that lead to ICRH: output-refinement and policy-refinement. For these processes, evaluations on static datasets are insufficient -- they miss the feedback effects and thus cannot capture the most harmful behavior. In response, we provide three recommendations for evaluation to capture more instances of ICRH. As AI development accelerates, the effects of feedback loops will proliferate, increasing the need to understand their role in shaping LLM behavior."
                },
                {
                    "title": "Spotting LLMs With Binoculars: Zero-Shot Detection of Machine-Generated Text",
                    "abstract": "Detecting text generated by modern large language models is thought to be hard, as both LLMs and humans can exhibit a wide range of complex behaviors. However, we find that a score based on contrasting two closely related language models is highly accurate at separating human-generated and machine-generated text. Based on this mechanism, we propose a novel LLM detector that only requires simple calculations using a pair of pre-trained LLMs. The method, called Binoculars, achieves state-of-the-art accuracy without any training data. It is capable of spotting machine text from a range of modern LLMs without any model-specific modifications. We comprehensively evaluate Binoculars on a number of text sources and in varied situations. Over a wide range of document types, Binoculars detects over 90% of generated samples from ChatGPT (and other LLMs) at a false positive rate of 0.01%, despite not being trained on any ChatGPT data."
                },
                {
                    "title": "MM-SAP: A Comprehensive Benchmark for Assessing Self-Awareness of Multimodal Large Language Models in Perception",
                    "abstract": "Recent advancements in Multimodal Large Language Models (MLLMs) have demonstrated exceptional capabilities in visual perception and understanding. However, these models also suffer from hallucinations, which limit their reliability as AI systems. We believe that these hallucinations are partially due to the models' struggle with understanding what they can and cannot perceive from images, a capability we refer to as self-awareness in perception. Despite its importance, this aspect of MLLMs has been overlooked in prior studies. In this paper, we aim to define and evaluate the self-awareness of MLLMs in perception. To do this, we first introduce the knowledge quadrant in perception, which helps define what MLLMs know and do not know about images. Using this framework, we propose a novel benchmark, the Self-Awareness in Perception for MLLMs (MM-SAP), specifically designed to assess this capability. We apply MM-SAP to a variety of popular MLLMs, offering a comprehensive analysis of their self-awareness and providing detailed insights. The experiment results reveal that current MLLMs possess limited self-awareness capabilities, pointing to a crucial area for future advancement in the development of trustworthy MLLMs. Code and data are available at https://github.com/YHWmz/MM-SAP."
                },
                {
                    "title": "LLMs as Narcissistic Evaluators: When Ego Inflates Evaluation Scores",
                    "abstract": "Automatic evaluation of generated textual content presents an ongoing challenge within the field of NLP. Given the impressive capabilities of modern language models (LMs) across diverse NLP tasks, there is a growing trend to employ these models in creating innovative evaluation metrics for automated assessment of generation tasks. This paper investigates a pivotal question: Do language model-driven evaluation metrics inherently exhibit bias favoring texts generated by the same underlying language model? Specifically, we assess whether prominent LM-based evaluation metrics (e.g. BARTScore, T5Score, and GPTScore) demonstrate a favorable bias toward their respective underlying LMs in the context of summarization tasks. Our findings unveil a latent bias, particularly pronounced when such evaluation metrics are used in a reference-free manner without leveraging gold summaries. These results underscore that assessments provided by generative evaluation models can be influenced by factors beyond the inherent text quality, highlighting the necessity of developing more reliable evaluation protocols in the future."
                },
                {
                    "title": "The ART of LLM Refinement: Ask, Refine, and Trust",
                    "abstract": "Large Language Models (LLMs) have demonstrated remarkable generative abilities, but can they judge the quality of their own generations and self-improve?A popular concept, referred to as *self-refinement*, postulates that LLMs can detect and correct the errors in their generations when asked to do so. However, recent empirical evidence points in the opposite direction, suggesting that LLMs often struggle to accurately identify errors when reasoning is involved. To address this, we propose a reasoning with a refinement strategy called *ART: Ask, Refine, and Trust*, which *asks* necessary questions to decide when an LLM should *refine* its output, and uses it to affirm or deny *trust* in its refinement by ranking the refinement and the initial prediction. On two multistep reasoning tasks of mathematical word problems (GSM8K) and question answering (StrategyQA), *ART* achieves a performance gain of +5 points over self-refinement baselines, while using a much smaller model as the decision maker. We believe that *ART* with smaller models, making refinement decisions can be a cost-effective alternative to fine-tuning LLMs."
                },
                {
                    "title": "Towards Evaluating AI Systems for Moral Status Using Self-Reports",
                    "abstract": "As AI systems become more advanced and widely deployed, there will likely be increasing debate over whether AI systems could have conscious experiences, desires, or other states of potential moral significance. It is important to inform these discussions with empirical evidence to the extent possible. We argue that under the right circumstances, self-reports, or an AI system's statements about its own internal states, could provide an avenue for investigating whether AI systems have states of moral significance. Self-reports are the main way such states are assessed in humans (\"Are you in pain?\"), but self-reports from current systems like large language models are spurious for many reasons (e.g. often just reflecting what humans would say). To make self-reports more appropriate for this purpose, we propose to train models to answer many kinds of questions about themselves with known answers, while avoiding or limiting training incentives that bias self-reports. The hope of this approach is that models will develop introspection-like capabilities, and that these capabilities will generalize to questions about states of moral significance. We then propose methods for assessing the extent to which these techniques have succeeded: evaluating self-report consistency across contexts and between similar models, measuring the confidence and resilience of models' self-reports, and using interpretability to corroborate self-reports. We also discuss challenges for our approach, from philosophical difficulties in interpreting self-reports to technical reasons why our proposal might fail. We hope our discussion inspires philosophers and AI researchers to criticize and improve our proposed methodology, as well as to run experiments to test whether self-reports can be made reliable enough to provide information about states of moral significance."
                },
                {
                    "title": "A Survey on Detection of LLMs-Generated Content",
                    "abstract": "The burgeoning capabilities of advanced large language models (LLMs) such as ChatGPT have led to an increase in synthetic content generation with implications across a variety of sectors, including media, cybersecurity, public discourse, and education. As such, the ability to detect LLMs-generated content has become of paramount importance. We aim to provide a detailed overview of existing detection strategies and benchmarks, scrutinizing their differences and identifying key challenges and prospects in the field, advocating for more adaptable and robust models to enhance detection accuracy. We also posit the necessity for a multi-faceted approach to defend against various attacks to counter the rapidly advancing capabilities of LLMs. To the best of our knowledge, this work is the first comprehensive survey on the detection in the era of LLMs. We hope it will provide a broad understanding of the current landscape of LLMs-generated content detection, offering a guiding reference for researchers and practitioners striving to uphold the integrity of digital information in an era increasingly dominated by synthetic content. The relevant papers are summarized and will be consistently updated at https://github.com/Xianjun-Yang/Awesome_papers_on_LLMs_detection.git."
                },
                {
                    "title": "Evaluating Large Language Models at Evaluating Instruction Following",
                    "abstract": "As research in large language models (LLMs) continues to accelerate, LLM-based evaluation has emerged as a scalable and cost-effective alternative to human evaluations for comparing the ever increasing list of models. This paper investigates the efficacy of these ``LLM evaluators'', particularly in using them to assess instruction following, a metric that gauges how closely generated text adheres to the given instruction. We introduce a challenging meta-evaluation benchmark, LLMBar, designed to test the ability of an LLM evaluator in discerning instruction-following outputs. The authors manually curated 419 pairs of outputs, one adhering to instructions while the other diverging, yet may possess deceptive qualities that mislead an LLM evaluator, e.g., a more engaging tone. Contrary to existing meta-evaluation, we discover that different evaluators (i.e., combinations of LLMs and prompts) exhibit distinct performance on LLMBar and even the highest-scoring ones have substantial room for improvement. We also present a novel suite of prompting strategies that further close the gap between LLM and human evaluators. With LLMBar, we hope to offer more insight into LLM evaluators and foster future research in developing better instruction-following models."
                },
                {
                    "title": "Democratizing LLMs: An Exploration of Cost-Performance Trade-offs in Self-Refined Open-Source Models",
                    "abstract": "The dominance of proprietary LLMs has led to restricted access and raised information privacy concerns. High-performing open-source alternatives are crucial for information-sensitive and high-volume applications but often lag behind in performance. To address this gap, we propose (1) A untargeted variant of iterative self-critique and self-refinement devoid of external influence. (2) A novel ranking metric - Performance, Refinement, and Inference Cost Score (PeRFICS) - to find the optimal model for a given task considering refined performance and cost. Our experiments show that SoTA open source models of varying sizes from 7B - 65B, on average, improve 8.2% from their baseline performance. Strikingly, even models with extremely small memory footprints, such as Vicuna-7B, show a 11.74% improvement overall and up to a 25.39% improvement in high-creativity, open ended tasks on the Vicuna benchmark. Vicuna-13B takes it a step further and outperforms ChatGPT post-refinement. This work has profound implications for resource-constrained and information-sensitive environments seeking to leverage LLMs without incurring prohibitive costs, compromising on performance and privacy. The domain-agnostic self-refinement process coupled with our novel ranking metric facilitates informed decision-making in model selection, thereby reducing costs and democratizing access to high-performing language models, as evidenced by case studies."
                },
                {
                    "title": "Benchmarking Cognitive Biases in Large Language Models as Evaluators",
                    "abstract": "Large Language Models are cognitively biased judges. Large Language Models (LLMs) have recently been shown to be effective as automatic evaluators with simple prompting and in-context learning. In this work, we assemble 15 LLMs of four different size ranges and evaluate their output responses by preference ranking from the other LLMs as evaluators, such as System Star is better than System Square. We then evaluate the quality of ranking outputs introducing the Cognitive Bias Benchmark for LLMs as Evaluators (CoBBLEr), a benchmark to measure six different cognitive biases in LLM evaluation outputs, such as the Egocentric bias where a model prefers to rank its own outputs highly in evaluation. We find that LLMs are biased text quality evaluators, exhibiting strong indications on our bias benchmark (average of 40% of comparisons across all models) within each of their evaluations that question their robustness as evaluators. Furthermore, we examine the correlation between human and machine preferences and calculate the average Rank-Biased Overlap (RBO) score to be 49.6%, indicating that machine preferences are misaligned with humans. According to our findings, LLMs may still be unable to be utilized for automatic annotation aligned with human preferences. Our project page is at: https://minnesotanlp.github.io/cobbler."
                },
                {
                    "title": "RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedback",
                    "abstract": "Reinforcement learning from human feedback (RLHF) has proven effective in aligning large language models (LLMs) with human preferences, but gathering high-quality preference labels is expensive. RL from AI Feedback (RLAIF), introduced in Bai et al., offers a promising alternative that trains the reward model (RM) on preferences generated by an off-the-shelf LLM. Across the tasks of summarization, helpful dialogue generation, and harmless dialogue generation, we show that RLAIF achieves comparable performance to RLHF. Furthermore, we take a step towards\"self-improvement\"by demonstrating that RLAIF can outperform a supervised fine-tuned baseline even when the AI labeler is the same size as the policy, or even the exact same checkpoint as the initial policy. Finally, we introduce direct-RLAIF (d-RLAIF) - a technique that circumvents RM training by obtaining rewards directly from an off-the-shelf LLM during RL, which achieves superior performance to canonical RLAIF. Our results suggest that RLAIF can achieve performance on-par with using human feedback, offering a potential solution to the scalability limitations of RLHF."
                },
                {
                    "title": "Taken out of context: On measuring situational awareness in LLMs",
                    "abstract": "We aim to better understand the emergence of `situational awareness' in large language models (LLMs). A model is situationally aware if it's aware that it's a model and can recognize whether it's currently in testing or deployment. Today's LLMs are tested for safety and alignment before they are deployed. An LLM could exploit situational awareness to achieve a high score on safety tests, while taking harmful actions after deployment. Situational awareness may emerge unexpectedly as a byproduct of model scaling. One way to better foresee this emergence is to run scaling experiments on abilities necessary for situational awareness. As such an ability, we propose `out-of-context reasoning' (in contrast to in-context learning). We study out-of-context reasoning experimentally. First, we finetune an LLM on a description of a test while providing no examples or demonstrations. At test time, we assess whether the model can pass the test. To our surprise, we find that LLMs succeed on this out-of-context reasoning task. Their success is sensitive to the training setup and only works when we apply data augmentation. For both GPT-3 and LLaMA-1, performance improves with model size. These findings offer a foundation for further empirical study, towards predicting and potentially controlling the emergence of situational awareness in LLMs. Code is available at: https://github.com/AsaCooperStickland/situational-awareness-evals."
                },
                {
                    "title": "Large Language Models Sensitivity to The Order of Options in Multiple-Choice Questions",
                    "abstract": "Large Language Models (LLMs) have demonstrated remarkable capabilities in various NLP tasks. However, previous works have shown these models are sensitive towards prompt wording, and few-shot demonstrations and their order, posing challenges to fair assessment of these models. As these models become more powerful, it becomes imperative to understand and address these limitations. In this paper, we focus on LLMs robustness on the task of multiple-choice questions -- commonly adopted task to study reasoning and fact-retrieving capability of LLMs. Investigating the sensitivity of LLMs towards the order of options in multiple-choice questions, we demonstrate a considerable performance gap of approximately 13% to 75% in LLMs on different benchmarks, when answer options are reordered, even when using demonstrations in a few-shot setting. Through a detailed analysis, we conjecture that this sensitivity arises when LLMs are uncertain about the prediction between the top-2/3 choices, and specific options placements may favor certain prediction between those top choices depending on the question caused by positional bias. We also identify patterns in top-2 choices that amplify or mitigate the model's bias toward option placement. We found that for amplifying bias, the optimal strategy involves positioning the top two choices as the first and last options. Conversely, to mitigate bias, we recommend placing these choices among the adjacent options. To validate our conjecture, we conduct various experiments and adopt two approaches to calibrate LLMs' predictions, leading to up to 8 percentage points improvement across different models and benchmarks."
                },
                {
                    "title": "VisIT-Bench: A Benchmark for Vision-Language Instruction Following Inspired by Real-World Use",
                    "abstract": "We introduce VisIT-Bench (Visual InsTruction Benchmark), a benchmark for evaluation of instruction-following vision-language models for real-world use. Our starting point is curating 70 'instruction families' that we envision instruction tuned vision-language models should be able to address. Extending beyond evaluations like VQAv2 and COCO, tasks range from basic recognition to game playing and creative generation. Following curation, our dataset comprises 592 test queries, each with a human-authored instruction-conditioned caption. These descriptions surface instruction-specific factors, e.g., for an instruction asking about the accessibility of a storefront for wheelchair users, the instruction-conditioned caption describes ramps/potential obstacles. These descriptions enable 1) collecting human-verified reference outputs for each instance; and 2) automatic evaluation of candidate multimodal generations using a text-only LLM, aligning with human judgment. We quantify quality gaps between models and references using both human and automatic evaluations; e.g., the top-performing instruction-following model wins against the GPT-4 reference in just 27% of the comparison. VisIT-Bench is dynamic to participate, practitioners simply submit their model's response on the project website; Data, code and leaderboard is available at visit-bench.github.io."
                },
                {
                    "title": "Is GPT-4 a reliable rater? Evaluating Consistency in GPT-4 Text Ratings",
                    "abstract": "This study reports the Intraclass Correlation Coefficients of feedback ratings produced by OpenAI's GPT-4, a large language model (LLM), across various iterations, time frames, and stylistic variations. The model was used to rate responses to tasks related to macroeconomics in higher education (HE), based on their content and style. Statistical analysis was performed to determine the absolute agreement and consistency of ratings in all iterations, and the correlation between the ratings in terms of content and style. The findings revealed high interrater reliability, with ICC scores ranging from 0.94 to 0.99 for different time periods, indicating that GPT-4 is capable of producing consistent ratings. The prompt used in this study is also presented and explained."
                },
                {
                    "title": "Evaluating Instruction-Tuned Large Language Models on Code Comprehension and Generation",
                    "abstract": "In this work, we evaluate 10 open-source instructed LLMs on four representative code comprehension and generation tasks. We have the following main findings. First, for the zero-shot setting, instructed LLMs are very competitive on code comprehension and generation tasks and sometimes even better than small SOTA models specifically fine-tuned on each downstream task. We also find that larger instructed LLMs are not always better on code-related tasks. Second, for the few-shot setting, we find that adding demonstration examples substantially helps instructed LLMs perform better on most code comprehension and generation tasks; however, the examples would sometimes induce unstable or even worse performance. Furthermore, we find widely-used BM25-based shot selection strategy significantly outperforms the basic random selection or fixed selection only on generation problems. Third, for the fine-tuning setting, we find that fine-tuning could further improve the model performance on downstream code comprehension and generation tasks compared to the zero-shot/one-shot performance. In addition, after being fine-tuned on the same downstream task dataset, instructed LLMs outperform both the small SOTA models and similar-scaled LLMs without instruction tuning. Based on our findings, we further present practical implications on model and usage recommendation, performance and cost trade-offs, and future direction."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "Benchmarking Foundation Models with Language-Model-as-an-Examiner",
                    "abstract": "Numerous benchmarks have been established to assess the performance of foundation models on open-ended question answering, which serves as a comprehensive test of a model's ability to understand and generate language in a manner similar to humans. Most of these works focus on proposing new datasets, however, we see two main issues within previous benchmarking pipelines, namely testing leakage and evaluation automation. In this paper, we propose a novel benchmarking framework, Language-Model-as-an-Examiner, where the LM serves as a knowledgeable examiner that formulates questions based on its knowledge and evaluates responses in a reference-free manner. Our framework allows for effortless extensibility as various LMs can be adopted as the examiner, and the questions can be constantly updated given more diverse trigger topics. For a more comprehensive and equitable evaluation, we devise three strategies: (1) We instruct the LM examiner to generate questions across a multitude of domains to probe for a broad acquisition, and raise follow-up questions to engage in a more in-depth assessment. (2) Upon evaluation, the examiner combines both scoring and ranking measurements, providing a reliable result as it aligns closely with human annotations. (3) We additionally propose a decentralized Peer-examination method to address the biases in a single examiner. Our data and benchmarking results are available at: https://lmexam.com."
                },
                {
                    "title": "Do Large Language Models Know What They Don't Know?",
                    "abstract": "Large language models (LLMs) have a wealth of knowledge that allows them to excel in various Natural Language Processing (NLP) tasks. Current research focuses on enhancing their performance within their existing knowledge. Despite their vast knowledge, LLMs are still limited by the amount of information they can accommodate and comprehend. Therefore, the ability to understand their own limitations on the unknows, referred to as self-knowledge, is of paramount importance. This study aims to evaluate LLMs' self-knowledge by assessing their ability to identify unanswerable or unknowable questions. We introduce an automated methodology to detect uncertainty in the responses of these models, providing a novel measure of their self-knowledge. We further introduce a unique dataset, SelfAware, consisting of unanswerable questions from five diverse categories and their answerable counterparts. Our extensive analysis, involving 20 LLMs including GPT-3, InstructGPT, and LLaMA, discovering an intrinsic capacity for self-knowledge within these models. Moreover, we demonstrate that in-context learning and instruction tuning can further enhance this self-knowledge. Despite this promising insight, our findings also highlight a considerable gap between the capabilities of these models and human proficiency in recognizing the limits of their knowledge."
                },
                {
                    "title": "Knowledge of Knowledge: Exploring Known-Unknowns Uncertainty with Large Language Models",
                    "abstract": "This paper investigates the capabilities of Large Language Models (LLMs) in the context of understanding their knowledge and uncertainty over questions. Specifically, we focus on addressing known-unknown questions, characterized by high uncertainty due to the absence of definitive answers. To facilitate our study, we collect a new dataset with Known-Unknown Questions (KUQ) and establish a categorization framework to clarify the origins of uncertainty in such queries. Subsequently, we examine the performance of open-source LLMs, fine-tuned using this dataset, in distinguishing between known and unknown queries within open-ended question-answering scenarios. The fine-tuned models demonstrated a significant improvement, achieving a considerable increase in F1-score relative to their pre-fine-tuning state. Through a comprehensive analysis, we reveal insights into the models' improved uncertainty articulation and their consequent efficacy in multi-agent debates. These findings help us understand how LLMs can be trained to identify and express uncertainty, improving our knowledge of how they understand and express complex or unclear information."
                },
                {
                    "title": "Self-Refine: Iterative Refinement with Self-Feedback",
                    "abstract": "Like humans, large language models (LLMs) do not always generate the best output on their first try. Motivated by how humans refine their written text, we introduce Self-Refine, an approach for improving initial outputs from LLMs through iterative feedback and refinement. The main idea is to generate an initial output using an LLMs; then, the same LLMs provides feedback for its output and uses it to refine itself, iteratively. Self-Refine does not require any supervised training data, additional training, or reinforcement learning, and instead uses a single LLM as the generator, refiner, and feedback provider. We evaluate Self-Refine across 7 diverse tasks, ranging from dialog response generation to mathematical reasoning, using state-of-the-art (GPT-3.5, ChatGPT, and GPT-4) LLMs. Across all evaluated tasks, outputs generated with Self-Refine are preferred by humans and automatic metrics over those generated with the same LLM using conventional one-step generation, improving by ~20% absolute on average in task performance. Our work demonstrates that even state-of-the-art LLMs like GPT-4 can be further improved at test time using our simple, standalone approach."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "GPTScore: Evaluate as You Desire",
                    "abstract": "Generative Artificial Intelligence (AI) has enabled the development of sophisticated models that are capable of producing high-caliber text, images, and other outputs through the utilization of large pre-trained models.Nevertheless, assessing the quality of the generation is an even more arduous task than the generation itself, and this issue has not been given adequate consideration recently.This paper proposes a novel evaluation framework, GPTScore, which utilizes the emergent abilities (e.g., in-context learning, zero-shot instruction) of generative pre-trained models to score generated texts. There are 19 pre-trained models explored in this paper, ranging in size from 80M (e.g., Flan-T5-small) to 175B (e.g., GPT3).Experimental results on four text generation tasks, 22 evaluation aspects, and corresponding 37 datasets demonstrate that this approach can effectively allow us to achieve what one desires to evaluate for texts simply by natural language instructions.This nature helps us overcome several long-standing challenges in text evaluation\u2013how to achieve customized, multi-faceted evaluation without model training. We make our code publicly available."
                },
                {
                    "title": "DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature",
                    "abstract": "The increasing fluency and widespread usage of large language models (LLMs) highlight the desirability of corresponding tools aiding detection of LLM-generated text. In this paper, we identify a property of the structure of an LLM's probability function that is useful for such detection. Specifically, we demonstrate that text sampled from an LLM tends to occupy negative curvature regions of the model's log probability function. Leveraging this observation, we then define a new curvature-based criterion for judging if a passage is generated from a given LLM. This approach, which we call DetectGPT, does not require training a separate classifier, collecting a dataset of real or generated passages, or explicitly watermarking generated text. It uses only log probabilities computed by the model of interest and random perturbations of the passage from another generic pre-trained language model (e.g., T5). We find DetectGPT is more discriminative than existing zero-shot methods for model sample detection, notably improving detection of fake news articles generated by 20B parameter GPT-NeoX from 0.81 AUROC for the strongest zero-shot baseline to 0.95 AUROC for DetectGPT. See https://ericmitchell.ai/detectgpt for code, data, and other project information."
                },
                {
                    "title": "Constitutional AI: Harmlessness from AI Feedback",
                    "abstract": "As AI systems become more capable, we would like to enlist their help to supervise other AIs. We experiment with methods for training a harmless AI assistant through self-improvement, without any human labels identifying harmful outputs. The only human oversight is provided through a list of rules or principles, and so we refer to the method as 'Constitutional AI'. The process involves both a supervised learning and a reinforcement learning phase. In the supervised phase we sample from an initial model, then generate self-critiques and revisions, and then finetune the original model on revised responses. In the RL phase, we sample from the finetuned model, use a model to evaluate which of the two samples is better, and then train a preference model from this dataset of AI preferences. We then train with RL using the preference model as the reward signal, i.e. we use 'RL from AI Feedback' (RLAIF). As a result we are able to train a harmless but non-evasive AI assistant that engages with harmful queries by explaining its objections to them. Both the SL and RL methods can leverage chain-of-thought style reasoning to improve the human-judged performance and transparency of AI decision making. These methods make it possible to control AI behavior more precisely and with far fewer human labels."
                },
                {
                    "title": "Machine-Generated Text: A Comprehensive Survey of Threat Models and Detection Methods",
                    "abstract": "Machine-generated text is increasingly difficult to distinguish from text authored by humans. Powerful open-source models are freely available, and user-friendly tools that democratize access to generative models are proliferating. ChatGPT, which was released shortly after the first edition of this survey, epitomizes these trends. The great potential of state-of-the-art natural language generation (NLG) systems is tempered by the multitude of avenues for abuse. Detection of machine-generated text is a key countermeasure for reducing the abuse of NLG models, and presents significant technical challenges and numerous open problems. We provide a survey that includes 1) an extensive analysis of threat models posed by contemporary NLG systems and 2) the most complete review of machine-generated text detection methods to date. This survey places machine-generated text within its cybersecurity and social context, and provides strong guidance for future work addressing the most critical threat models. While doing so, we highlight the importance that detection systems themselves demonstrate trustworthiness through fairness, robustness, and accountability."
                },
                {
                    "title": "Language Models (Mostly) Know What They Know",
                    "abstract": "We study whether language models can evaluate the validity of their own claims and predict which questions they will be able to answer correctly. We first show that larger models are well-calibrated on diverse multiple choice and true/false questions when they are provided in the right format. Thus we can approach self-evaluation on open-ended sampling tasks by asking models to first propose answers, and then to evaluate the probability\"P(True)\"that their answers are correct. We find encouraging performance, calibration, and scaling for P(True) on a diverse array of tasks. Performance at self-evaluation further improves when we allow models to consider many of their own samples before predicting the validity of one specific possibility. Next, we investigate whether models can be trained to predict\"P(IK)\", the probability that\"I know\"the answer to a question, without reference to any particular proposed answer. Models perform well at predicting P(IK) and partially generalize across tasks, though they struggle with calibration of P(IK) on new tasks. The predicted P(IK) probabilities also increase appropriately in the presence of relevant source materials in the context, and in the presence of hints towards the solution of mathematical word problems. We hope these observations lay the groundwork for training more honest models, and for investigating how honesty generalizes to cases where models are trained on objectives other than the imitation of human writing."
                },
                {
                    "title": "Self-critiquing models for assisting human evaluators",
                    "abstract": "We fine-tune large language models to write natural language critiques (natural language critical comments) using behavioral cloning. On a topic-based summarization task, critiques written by our models help humans find flaws in summaries that they would have otherwise missed. Our models help find naturally occurring flaws in both model and human written summaries, and intentional flaws in summaries written by humans to be deliberately misleading. We study scaling properties of critiquing with both topic-based summarization and synthetic tasks. Larger models write more helpful critiques, and on most tasks, are better at self-critiquing, despite having harder-to-critique outputs. Larger models can also integrate their own selfcritiques as feedback, refining their own summaries into better ones. Finally, we motivate and introduce a framework for comparing critiquing ability to generation and discrimination ability. Our measurements suggest that even large models may still have relevant knowledge they cannot or do not articulate as critiques. These results are a proof of concept for using AI-assisted human feedback to scale the supervision of machine learning systems to tasks that are difficult for humans to evaluate directly. We release our training datasets, as well as samples from our critique assistance experiments."
                },
                {
                    "title": "Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models",
                    "abstract": "Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit\"breakthrough\"behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting."
                },
                {
                    "title": "Recursively Summarizing Books with Human Feedback",
                    "abstract": "A major challenge for scaling machine learning is training models to perform tasks that are very difficult or time-consuming for humans to evaluate. We present progress on this problem on the task of abstractive summarization of entire fiction novels. Our method combines learning from human feedback with recursive task decomposition: we use models trained on smaller parts of the task to assist humans in giving feedback on the broader task. We collect a large volume of demonstrations and comparisons from human labelers, and fine-tune GPT-3 using behavioral cloning and reward modeling to do summarization recursively. At inference time, the model first summarizes small sections of the book and then recursively summarizes these summaries to produce a summary of the entire book. Our human labelers are able to supervise and evaluate the models quickly, despite not having read the entire books themselves. Our resulting model generates sensible summaries of entire books, even matching the quality of human-written summaries in a few cases ($\\sim5\\%$ of books). We achieve state-of-the-art results on the recent BookSum dataset for book-length summarization. A zero-shot question-answering model using these summaries achieves state-of-the-art results on the challenging NarrativeQA benchmark for answering questions about books and movie scripts. We release datasets of samples from our model."
                },
                {
                    "title": "Automatic Detection of Machine Generated Text: A Critical Survey",
                    "abstract": "Text generative models (TGMs) excel in producing text that matches the style of human language reasonably well. Such TGMs can be misused by adversaries, e.g., by automatically generating fake news and fake product reviews that can look authentic and fool humans. Detectors that can distinguish text generated by TGM from human written text play a vital role in mitigating such misuse of TGMs. Recently, there has been a flurry of works from both natural language processing (NLP) and machine learning (ML) communities to build accurate detectors for English. Despite the importance of this problem, there is currently no work that surveys this fast-growing literature and introduces newcomers to important research challenges. In this work, we fill this void by providing a critical survey and review of this literature to facilitate a comprehensive understanding of this problem. We conduct an in-depth error analysis of the state-of-the-art detector and discuss research directions to guide future work in this exciting area."
                },
                {
                    "title": "Learning to summarize from human feedback",
                    "abstract": "As language models become more powerful, training and evaluation are increasingly bottlenecked by the data and metrics used for a particular task. For example, summarization models are often trained to predict human reference summaries and evaluated using ROUGE, but both of these metrics are rough proxies for what we really care about---summary quality. In this work, we show that it is possible to significantly improve summary quality by training a model to optimize for human preferences. We collect a large, high-quality dataset of human comparisons between summaries, train a model to predict the human-preferred summary, and use that model as a reward function to fine-tune a summarization policy using reinforcement learning. We apply our method to a version of the TL;DR dataset of Reddit posts and find that our models significantly outperform both human reference summaries and much larger models fine-tuned with supervised learning alone. Our models also transfer to CNN/DM news articles, producing summaries nearly as good as the human reference without any news-specific fine-tuning. We conduct extensive analyses to understand our human feedback dataset and fine-tuned models We establish that our reward model generalizes to new datasets, and that optimizing our reward model results in better summaries than optimizing ROUGE according to humans. We hope the evidence from our paper motivates machine learning researchers to pay closer attention to how their training loss affects the model behavior they actually want."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Scalable agent alignment via reward modeling: a research direction",
                    "abstract": "One obstacle to applying reinforcement learning algorithms to real-world problems is the lack of suitable reward functions. Designing such reward functions is difficult in part because the user only has an implicit understanding of the task objective. This gives rise to the agent alignment problem: how do we create agents that behave in accordance with the user's intentions? We outline a high-level research direction to solve the agent alignment problem centered around reward modeling: learning a reward function from interaction with the user and optimizing the learned reward function with reinforcement learning. We discuss the key challenges we expect to face when scaling reward modeling to complex and general domains, concrete approaches to mitigate these challenges, and ways to establish trust in the resulting agents."
                },
                {
                    "title": "Abstractive Text Summarization using Sequence-to-sequence RNNs and Beyond",
                    "abstract": "In this work, we model abstractive text summarization using Attentional Encoder-Decoder Recurrent Neural Networks, and show that they achieve state-of-the-art performance on two different corpora. We propose several novel models that address critical problems in summarization that are not adequately modeled by the basic architecture, such as modeling key-words, capturing the hierarchy of sentence-to-word structure, and emitting words that are rare or unseen at training time. Our work shows that many of our proposed models contribute to further improvement in performance. We also propose a new dataset consisting of multi-sentence summaries, and establish performance benchmarks for further research."
                },
                {
                    "title": "Towards a Situational Awareness Benchmark for LLMs",
                    "abstract": "Among the facts that LLMs can learn is knowledge about themselves and their situation. This knowledge, and ability to make inferences based on it, is called situational awareness. Situationally aware models can be more helpful, but also pose risks. For example, situationally aware models could game testing setups by knowing they are being tested and acting differently. We create a new benchmark, SAD (Situational Awareness Dataset) , for LLM situational awareness in two categories that are especially relevant for future AI risks. SAD-influence tests whether LLMs can accurately assess how they can or cannot influence the world. SAD-stages tests if LLMs can recognize if a particular input is likely to have come from a given stage of the LLM lifecycle (pretraining, supervised fine-tuning, testing, and deployment). Only the most capable models do better than chance. If the prompt tells the model that it is an LLM, scores increase by 9-21 percentage points for models on SAD-influence, while having mixed effects on SAD-stages."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does an LLM's self-recognition capability influence its tendency to exhibit self-preference in self-evaluation tasks?\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding the relationship between self-recognition and self-preference in LLMs is crucial for improving the reliability of self-evaluation methods. By addressing this problem, we can enhance the accuracy of LLMs in assessing their own outputs, which has significant implications for the development of autonomous AI systems. This research could lead to more robust evaluation frameworks, reducing biases in LLM assessments and fostering trust in AI-generated content. Furthermore, it may pave the way for practical applications in areas such as automated content generation, quality control, and human-AI collaboration.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent biases that LLMs may exhibit when evaluating their own outputs, which can lead to skewed assessments. Naive approaches may fail because they do not account for the complex interplay between self-recognition and self-preference, potentially overlooking confounding factors that influence evaluation outcomes. Additionally, accurately measuring self-recognition and self-preference requires sophisticated methodologies and fine-tuning processes, which can be technically demanding. Overcoming these obstacles necessitates a deep understanding of LLM behavior and the development of effective experimental designs.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on LLM performance in general evaluation tasks without specifically addressing the nuances of self-evaluation biases. Existing solutions may lack the granularity needed to dissect the relationship between self-recognition and self-preference. Barriers such as limited datasets for fine-tuning and a lack of comprehensive metrics to measure self-recognition have hindered progress. Our approach differs by systematically investigating the causal link between these two properties and employing targeted fine-tuning to manipulate self-recognition, thereby providing clearer insights into the self-preference phenomenon.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves fine-tuning LLMs (GPT-3.5 and Llama 2) on a dataset of 500 examples to enhance their self-recognition capabilities. We will measure self-recognition accuracy using simple prompts and evaluate self-preference through comparative assessments of LLM outputs against those generated by other LLMs and humans. The expected outcomes include demonstrating a linear correlation between self-recognition and self"
            }
        },
        "author_data": {
            "51742735-b21b-40be-a3ed-3865feeb6abd": {
                "pk": "51742735-b21b-40be-a3ed-3865feeb6abd",
                "name": "Arjun Panickssery",
                "collaborators": [
                    "Sara Price",
                    "Sam Bowman",
                    "Asa Cooper Stickland",
                    "Axel H\u00f8jmark",
                    "Govind Pimpale",
                    "Marius Hobbhahn",
                    "J\u00e9r\u00e9my Scheurer",
                    "Andrew Gritsevskiy",
                    "Aaron Kirtland",
                    "Derik Kauffman"
                ],
                "domain": [
                    "AI Safety",
                    "Large Language Models",
                    "Multimodal Learning",
                    "Backdoor Attacks"
                ],
                "publications": [
                    {
                        "title": "Future Events as Backdoor Triggers: Investigating Temporal Vulnerabilities in LLMs",
                        "abstract": "Backdoors are hidden behaviors that are only triggered once an AI system has been deployed. Bad actors looking to create successful backdoors must design them to avoid activation during training and evaluation. Since data used in these stages often only contains information about events that have already occurred, a component of a simple backdoor trigger could be a model recognizing data that is in the future relative to when it was trained. Through prompting experiments and by probing internal activations, we show that current large language models (LLMs) can distinguish past from future events, with probes on model activations achieving 90% accuracy. We train models with backdoors triggered by a temporal distributional shift; they activate when the model is exposed to news headlines beyond their training cut-off dates. Fine-tuning on helpful, harmless and honest (HHH) data does not work well for removing simpler backdoor triggers but is effective on our backdoored models, although this distinction is smaller for the larger-scale model we tested. We also find that an activation-steering vector representing a model's internal representation of the date influences the rate of backdoor activation. We take these results as initial evidence that, at least for models at the modest scale we test, standard safety measures are enough to remove these backdoors."
                    },
                    {
                        "title": "Analyzing Probabilistic Methods for Evaluating Agent Capabilities",
                        "abstract": "To mitigate risks from AI systems, we need to assess their capabilities accurately. This is especially difficult in cases where capabilities are only rarely displayed. Phuong et al. propose two methods that aim to obtain better estimates of the probability of an AI agent successfully completing a given task. The milestone method decomposes tasks into subtasks, aiming to improve overall success rate estimation, while the expert best-of-N method leverages human guidance as a proxy for the model's independent performance.   Our analysis of these methods as Monte Carlo estimators reveals that while both effectively reduce variance compared to naive Monte Carlo sampling, they also introduce bias. Experimental results demonstrate that the milestone method underestimates true solve rates for many real-world tasks due to its constraining assumptions. The expert best-of-N method exhibits even more severe underestimation across all tasks, attributed to an inherently flawed re-weighting factor. To enhance the accuracy of capability estimates of AI agents on difficult tasks, we suggest future work should leverage the rich literature on Monte Carlo Estimators."
                    },
                    {
                        "title": "REBUS: A Robust Evaluation Benchmark of Understanding Symbols",
                        "abstract": "We propose a new benchmark evaluating the performance of multimodal large language models on rebus puzzles. The dataset covers 333 original examples of image-based wordplay, cluing 13 categories such as movies, composers, major cities, and food. To achieve good performance on the benchmark of identifying the clued word or phrase, models must combine image recognition and string manipulation with hypothesis testing, multi-step reasoning, and an understanding of human cognition, making for a complex, multimodal evaluation of capabilities. We find that GPT-4o significantly outperforms all other models, followed by proprietary models outperforming all other evaluated models. However, even the best model has a final accuracy of only 42\\%, which goes down to just 7\\% on hard puzzles, highlighting the need for substantial improvements in reasoning. Further, models rarely understand all parts of a puzzle, and are almost always incapable of retroactively explaining the correct answer. Our benchmark can therefore be used to identify major shortcomings in the knowledge and reasoning of multimodal large language models."
                    }
                ]
            },
            "8c0d4b46-b559-41e2-9af7-3ae658a9c4c6": {
                "pk": "8c0d4b46-b559-41e2-9af7-3ae658a9c4c6",
                "name": "Samuel R. Bowman",
                "collaborators": [
                    "Alex Warstadt",
                    "Christopher Potts",
                    "Christopher D. Manning",
                    "George E. Dahl",
                    "Yacine Jernite",
                    "David Sontag",
                    "Adina Williams",
                    "Nikita Nangia",
                    "Yichen Gong",
                    "Amanpreet Singh"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Neural Networks",
                    "Machine Learning",
                    "Linguistic Reasoning"
                ],
                "publications": [
                    {
                        "title": "Can recursive neural tensor networks learn logical reasoning?",
                        "abstract": "Recursive neural network models and their accompanying vector representations for words have seen success in an array of increasingly semantically sophisticated tasks, but almost nothing is known about their ability to accurately capture the aspects of linguistic meaning that are necessary for interpretation or reasoning. To evaluate this, I train a recursive model on a new corpus of constructed examples of logical reasoning in short sentences, like the inference of \"some animal walks\" from \"some dog walks\" or \"some cat walks,\" given that dogs and cats are animals. This model learns representations that generalize well to new types of reasoning pattern in all but a few cases, a result which is promising for the ability of learned representation models to capture logical reasoning."
                    },
                    {
                        "title": "The Dangers of Underclaiming: Reasons for Caution When Reporting How NLP Systems Fail",
                        "abstract": "Researchers in NLP often frame and discuss research results in ways that serve to deemphasize the field's successes, often in response to the field's widespread hype. Though well-meaning, this has yielded many misleading or false claims about the limits of our best technology. This is a problem, and it may be more serious than it looks: It harms our credibility in ways that can make it harder to mitigate present-day harms, like those involving biased systems for content moderation or resume screening. It also limits our ability to prepare for the potentially enormous impacts of more distant future advances. This paper urges researchers to be careful about these claims and suggests some research directions and communication strategies that will make it easier to avoid or rebut them."
                    },
                    {
                        "title": "Eight Things to Know about Large Language Models",
                        "abstract": "The widespread public deployment of large language models (LLMs) in recent months has prompted a wave of new attention and engagement from advocates, policymakers, and scholars from many fields. This attention is a timely response to the many urgent questions that this technology raises, but it can sometimes miss important considerations. This paper surveys the evidence for eight potentially surprising such points:   1. LLMs predictably get more capable with increasing investment, even without targeted innovation.   2. Many important LLM behaviors emerge unpredictably as a byproduct of increasing investment.   3. LLMs often appear to learn and use representations of the outside world.   4. There are no reliable techniques for steering the behavior of LLMs.   5. Experts are not yet able to interpret the inner workings of LLMs.   6. Human performance on a task isn't an upper bound on LLM performance.   7. LLMs need not express the values of their creators nor the values encoded in web text.   8. Brief interactions with LLMs are often misleading."
                    },
                    {
                        "title": "What Will it Take to Fix Benchmarking in Natural Language Understanding?",
                        "abstract": "Evaluation for many natural language understanding (NLU) tasks is broken: Unreliable and biased systems score so highly on standard benchmarks that there is little room for researchers who develop better systems to demonstrate their improvements. The recent trend to abandon IID benchmarks in favor of adversarially-constructed, out-of-distribution test sets ensures that current models will perform poorly, but ultimately only obscures the abilities that we want our benchmarks to measure. In this position paper, we lay out four criteria that we argue NLU benchmarks should meet. We argue most current benchmarks fail at these criteria, and that adversarial data collection does not meaningfully address the causes of these failures. Instead, restoring a healthy evaluation ecosystem will require significant progress in the design of benchmark datasets, the reliability with which they are annotated, their size, and the ways they handle social bias."
                    },
                    {
                        "title": "Discourse-Based Objectives for Fast Unsupervised Sentence Representation Learning",
                        "abstract": "This work presents a novel objective function for the unsupervised training of neural network sentence encoders. It exploits signals from paragraph-level discourse coherence to train these models to understand text. Our objective is purely discriminative, allowing us to train models many times faster than was possible under prior methods, and it yields models which perform well in extrinsic evaluations."
                    },
                    {
                        "title": "Recursive Neural Networks Can Learn Logical Semantics",
                        "abstract": "Tree-structured recursive neural networks (TreeRNNs) for sentence meaning have been successful for many applications, but it remains an open question whether the fixed-length representations that they learn can support tasks as demanding as logical deduction. We pursue this question by evaluating whether two such models---plain TreeRNNs and tree-structured neural tensor networks (TreeRNTNs)---can correctly learn to identify logical relationships such as entailment and contradiction using these representations. In our first set of experiments, we generate artificial data from a logical grammar and use it to evaluate the models' ability to learn to handle basic relational reasoning, recursive structures, and quantification. We then evaluate the models on the more natural SICK challenge data. Both models perform competitively on the SICK data and generalize well in all three experiments on simulated data, suggesting that they can learn suitable representations for logical inference in natural language."
                    },
                    {
                        "title": "Learning Distributed Word Representations for Natural Logic Reasoning",
                        "abstract": "Natural logic offers a powerful relational conception of meaning that is a natural counterpart to distributed semantic representations, which have proven valuable in a wide range of sophisticated language tasks. However, it remains an open question whether it is possible to train distributed representations to support the rich, diverse logical reasoning captured by natural logic. We address this question using two neural network-based models for learning embeddings: plain neural networks and neural tensor networks. Our experiments evaluate the models' ability to learn the basic algebra of natural logic relations from simulated data and from the WordNet noun graph. The overall positive results are promising for the future of learned distributed representations in the applied modeling of logical semantics."
                    },
                    {
                        "title": "Tree-structured composition in neural networks without tree-structured architectures",
                        "abstract": "Tree-structured neural networks encode a particular tree geometry for a sentence in the network design. However, these models have at best only slightly outperformed simpler sequence-based models. We hypothesize that neural sequence models like LSTMs are in fact able to discover and implicitly use recursive compositional structure, at least for tasks with clear cues to that structure in the data. We demonstrate this possibility using an artificial data task for which recursive compositional structure is crucial, and find an LSTM-based sequence model can indeed learn to exploit the underlying tree structure. However, its performance consistently lags behind that of tree models, even on large training sets, suggesting that tree-structured models are more effective at exploiting recursive structure."
                    },
                    {
                        "title": "What Artificial Neural Networks Can Tell Us About Human Language Acquisition",
                        "abstract": "Rapid progress in machine learning for natural language processing has the potential to transform debates about how humans learn language. However, the learning environments and biases of current artificial learners and humans diverge in ways that weaken the impact of the evidence obtained from learning simulations. For example, today's most effective neural language models are trained on roughly one thousand times the amount of linguistic data available to a typical child. To increase the relevance of learnability results from computational models, we need to train model learners without significant advantages over humans. If an appropriate model successfully acquires some target linguistic knowledge, it can provide a proof of concept that the target is learnable in a hypothesized human learning scenario. Plausible model learners will enable us to carry out experimental manipulations to make causal inferences about variables in the learning environment, and to rigorously test poverty-of-the-stimulus-style claims arguing for innate linguistic knowledge in humans on the basis of speculations about learnability. Comparable experiments will never be possible with human subjects due to practical and ethical considerations, making model learners an indispensable resource. So far, attempts to deprive current models of unfair advantages obtain sub-human results for key grammatical behaviors such as acceptability judgments. But before we can justifiably conclude that language learning requires more prior domain-specific knowledge than current models possess, we must first explore non-linguistic inputs in the form of multimodal stimuli and multi-agent interaction as ways to make our learners more efficient at learning from limited linguistic input."
                    },
                    {
                        "title": "A Broad-Coverage Challenge Corpus for Sentence Understanding through Inference",
                        "abstract": "This paper introduces the Multi-Genre Natural Language Inference (MultiNLI) corpus, a dataset designed for use in the development and evaluation of machine learning models for sentence understanding. In addition to being one of the largest corpora available for the task of NLI, at 433k examples, this corpus improves upon available resources in its coverage: it offers data from ten distinct genres of written and spoken English--making it possible to evaluate systems on nearly the full complexity of the language--and it offers an explicit setting for the evaluation of cross-genre domain adaptation."
                    },
                    {
                        "title": "Ruminating Reader: Reasoning with Gated Multi-Hop Attention",
                        "abstract": "To answer the question in machine comprehension (MC) task, the models need to establish the interaction between the question and the context. To tackle the problem that the single-pass model cannot reflect on and correct its answer, we present Ruminating Reader. Ruminating Reader adds a second pass of attention and a novel information fusion component to the Bi-Directional Attention Flow model (BiDAF). We propose novel layer structures that construct an query-aware context vector representation and fuse encoding representation with intermediate representation on top of BiDAF model. We show that a multi-hop attention mechanism can be applied to a bi-directional attention structure. In experiments on SQuAD, we find that the Reader outperforms the BiDAF baseline by a substantial margin, and matches or surpasses the performance of all other published systems."
                    },
                    {
                        "title": "Neural Network Acceptability Judgments",
                        "abstract": "This paper investigates the ability of artificial neural networks to judge the grammatical acceptability of a sentence, with the goal of testing their linguistic competence. We introduce the Corpus of Linguistic Acceptability (CoLA), a set of 10,657 English sentences labeled as grammatical or ungrammatical from published linguistics literature. As baselines, we train several recurrent neural network models on acceptability classification, and find that our models outperform unsupervised models by Lau et al (2016) on CoLA. Error-analysis on specific grammatical phenomena reveals that both Lau et al.'s models and ours learn systematic generalizations like subject-verb-object order. However, all models we test perform far below human level on a wide range of grammatical constructions."
                    },
                    {
                        "title": "Sentence Encoders on STILTs: Supplementary Training on Intermediate Labeled-data Tasks",
                        "abstract": "Pretraining sentence encoders with language modeling and related unsupervised tasks has recently been shown to be very effective for language understanding tasks. By supplementing language model-style pretraining with further training on data-rich supervised tasks, such as natural language inference, we obtain additional performance improvements on the GLUE benchmark. Applying supplementary training on BERT (Devlin et al., 2018), we attain a GLUE score of 81.8---the state of the art (as of 02/24/2019) and a 1.4 point improvement over BERT. We also observe reduced variance across random restarts in this setting. Our approach yields similar improvements when applied to ELMo (Peters et al., 2018a) and Radford et al. (2018)'s model. In addition, the benefits of supplementary training are particularly pronounced in data-constrained regimes, as we show in experiments with artificially limited training data."
                    },
                    {
                        "title": "Linguistic Analysis of Pretrained Sentence Encoders with Acceptability Judgments",
                        "abstract": "Recent work on evaluating grammatical knowledge in pretrained sentence encoders gives a fine-grained view of a small number of phenomena. We introduce a new analysis dataset that also has broad coverage of linguistic phenomena. We annotate the development set of the Corpus of Linguistic Acceptability (CoLA; Warstadt et al., 2018) for the presence of 13 classes of syntactic phenomena including various forms of argument alternations, movement, and modification. We use this analysis set to investigate the grammatical knowledge of three pretrained encoders: BERT (Devlin et al., 2018), GPT (Radford et al., 2018), and the BiLSTM baseline from Warstadt et al. We find that these models have a strong command of complex or non-canonical argument structures like ditransitives (Sue gave Dan a book) and passives (The book was read). Sentences with long distance dependencies like questions (What do you think I ate?) challenge all models, but for these, BERT and GPT have a distinct advantage over the baseline. We conclude that recent sentence encoders, despite showing near-human performance on acceptability classification overall, still fail to make fine-grained grammaticality distinctions for many complex syntactic structures."
                    },
                    {
                        "title": "Identifying and Reducing Gender Bias in Word-Level Language Models",
                        "abstract": "Many text corpora exhibit socially problematic biases, which can be propagated or amplified in the models trained on such data. For example, doctor cooccurs more frequently with male pronouns than female pronouns. In this study we (i) propose a metric to measure gender bias; (ii) measure bias in a text corpus and the text generated from a recurrent neural network language model trained on the text corpus; (iii) propose a regularization loss term for the language model that minimizes the projection of encoder-trained embeddings onto an embedding subspace that encodes gender; (iv) finally, evaluate efficacy of our proposed method on reducing gender bias. We find this regularization method to be effective in reducing gender bias up to an optimal weight assigned to the loss term, beyond which the model becomes unstable as the perplexity increases. We replicate this study on three training corpora---Penn Treebank, WikiText-2, and CNN/Daily Mail---resulting in similar conclusions."
                    },
                    {
                        "title": "Can Unconditional Language Models Recover Arbitrary Sentences?",
                        "abstract": "Neural network-based generative language models like ELMo and BERT can work effectively as general purpose sentence encoders in text classification without further fine-tuning. Is it possible to adapt them in a similar way for use as general-purpose decoders? For this to be possible, it would need to be the case that for any target sentence of interest, there is some continuous representation that can be passed to the language model to cause it to reproduce that sentence. We set aside the difficult problem of designing an encoder that can produce such representations and, instead, ask directly whether such representations exist at all. To do this, we introduce a pair of effective, complementary methods for feeding representations into pretrained unconditional language models and a corresponding set of methods to map sentences into and out of this representation space, the reparametrized sentence space. We then investigate the conditions under which a language model can be made to generate a sentence through the identification of a point in such a space and find that it is possible to recover arbitrary sentences nearly perfectly with language models and representations of moderate size without modifying any model parameters."
                    },
                    {
                        "title": "Self-Training for Unsupervised Parsing with PRPN",
                        "abstract": "Neural unsupervised parsing (UP) models learn to parse without access to syntactic annotations, while being optimized for another task like language modeling. In this work, we propose self-training for neural UP models: we leverage aggregated annotations predicted by copies of our model as supervision for future copies. To be able to use our model's predictions during training, we extend a recent neural UP architecture, the PRPN (Shen et al., 2018a) such that it can be trained in a semi-supervised fashion. We then add examples with parses predicted by our model to our unlabeled UP training data. Our self-trained model outperforms the PRPN by 8.1% F1 and the previous state of the art by 1.6% F1. In addition, we show that our architecture can also be helpful for semi-supervised parsing in ultra-low-resource settings."
                    },
                    {
                        "title": "Clean or Annotate: How to Spend a Limited Data Collection Budget",
                        "abstract": "Crowdsourcing platforms are often used to collect datasets for training machine learning models, despite higher levels of inaccurate labeling compared to expert labeling. There are two common strategies to manage the impact of such noise. The first involves aggregating redundant annotations, but comes at the expense of labeling substantially fewer examples. Secondly, prior works have also considered using the entire annotation budget to label as many examples as possible and subsequently apply denoising algorithms to implicitly clean the dataset. We find a middle ground and propose an approach which reserves a fraction of annotations to explicitly clean up highly probable error samples to optimize the annotation process. In particular, we allocate a large portion of the labeling budget to form an initial dataset used to train a model. This model is then used to identify specific examples that appear most likely to be incorrect, which we spend the remaining budget to relabel. Experiments across three model variations and four natural language processing tasks show our approach outperforms or matches both label aggregation and advanced denoising methods designed to handle noisy labels when allocated the same finite annotation budget."
                    },
                    {
                        "title": "Can neural networks acquire a structural bias from raw linguistic data?",
                        "abstract": "We evaluate whether BERT, a widely used neural network for sentence processing, acquires an inductive bias towards forming structural generalizations through pretraining on raw data. We conduct four experiments testing its preference for structural vs. linear generalizations in different structure-dependent phenomena. We find that BERT makes a structural generalization in 3 out of 4 empirical domains---subject-auxiliary inversion, reflexive binding, and verb tense detection in embedded clauses---but makes a linear generalization when tested on NPI licensing. We argue that these results are the strongest evidence so far from artificial learners supporting the proposition that a structural bias can be acquired from raw data. If this conclusion is correct, it is tentative evidence that some linguistic universals can be acquired by learners without innate biases. However, the precise implications for human language acquisition are unclear, as humans learn language from significantly less data than BERT."
                    },
                    {
                        "title": "Sequential Attention: A Context-Aware Alignment Function for Machine Reading",
                        "abstract": "In this paper we propose a neural network model with a novel Sequential Attention layer that extends soft attention by assigning weights to words in an input sequence in a way that takes into account not just how well that word matches a query, but how well surrounding words match. We evaluate this approach on the task of reading comprehension (on the Who did What and CNN datasets) and show that it dramatically improves a strong baseline--the Stanford Reader--and is competitive with the state of the art."
                    }
                ]
            },
            "ff812197-05bc-4e83-abfd-f7c9446b8b0a": {
                "pk": "ff812197-05bc-4e83-abfd-f7c9446b8b0a",
                "name": "Shi Feng",
                "collaborators": [
                    "Wei Chen",
                    "Jordan Boyd-Graber",
                    "Wencheng Bao"
                ],
                "domain": [
                    "Causal Inference",
                    "Machine Learning",
                    "Topological Order",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Number of regions created by random chords in the circle",
                        "abstract": "In this paper we discuss the number of regions in a unit circle after drawing $n$ i.i.d. random chords in the circle according to a particular family of distribution. We find that as $n$ goes to infinity, the distribution of the number of regions, properly shifted and scaled, converges to the standard normal distribution and the error can be bounded by Stein's method for proving Central Limit Theorem."
                    },
                    {
                        "title": "Hidden subsystem symmetry protected states in competing topological orders",
                        "abstract": "We reveal the connection between two-dimensional subsystem symmetry-protected topological (SSPT) states and two-dimensional topological orders via a self-dual frustrated toric code model. This model, an enrichment of the toric code (TC) with its dual interactions, can be mapped to a model defined on the dual lattice with subsystem symmetries and subextensive ground state degeneracy. The map connects exactly the frustrated TC to two copies of the topological plaquette Ising model (TPIM), as a strong SSPT model with linear subsystem symmetries. The membrane order parameter of the TPIM is exactly mapped to dual TC stabilizers as the order parameter of the frustrated TC model, SSPT gapless edge states of the TPIM are mapped to zero-energy dangling operators under open boundaries, and the transition from the SSPT-ordered TPIM to the trivial paramagnetic phase is mapped to the transition between two distinct topological orders. We also demonstrate that this mapping can be used to elucidate the structure of other SSPT models, reflecting the subtle linkage between SSPT order and topological order in two dimensions."
                    },
                    {
                        "title": "What can AI do for me: Evaluating Machine Learning Interpretations in Cooperative Play",
                        "abstract": "Machine learning is an important tool for decision making, but its ethical and responsible application requires rigorous vetting of its interpretability and utility: an understudied problem, particularly for natural language processing models. We propose an evaluation of interpretation on a real task with real human users, where the effectiveness of interpretation is measured by how much it improves human performance. We design a grounded, realistic human-computer cooperative setting using a question answering task, Quizbowl. We recruit both trivia experts and novices to play this game with computer as their teammate, who communicates its prediction via three different interpretations. We also provide design guidance for natural language processing human-in-the-loop settings."
                    },
                    {
                        "title": "Causal Inference for Influence Propagation -- Identifiability of the Independent Cascade Model",
                        "abstract": "Independent cascade (IC) model is a widely used influence propagation model for social networks. In this paper, we incorporate the concept and techniques from causal inference to study the identifiability of parameters from observational data in extended IC model with unobserved confounding factors, which models more realistic propagation scenarios but is rarely studied in influence propagation modeling before. We provide the conditions for the identifiability or unidentifiability of parameters for several special structures including the Markovian IC model, semi-Markovian IC model, and IC model with a global unobserved variable. Parameter identifiability is important for other tasks such as influence maximization under the diffusion networks with unobserved confounding factors."
                    },
                    {
                        "title": "Combinatorial Causal Bandits",
                        "abstract": "In combinatorial causal bandits (CCB), the learning agent chooses at most $K$ variables in each round to intervene, collects feedback from the observed variables, with the goal of minimizing expected regret on the target variable $Y$. We study under the context of binary generalized linear models (BGLMs) with a succinct parametric representation of the causal models. We present the algorithm BGLM-OFU for Markovian BGLMs (i.e. no hidden variables) based on the maximum likelihood estimation method, and show that it achieves $O(\\sqrt{T}\\log T)$ regret, where $T$ is the time horizon. For the special case of linear models with hidden variables, we apply causal inference techniques such as the do-calculus to convert the original model into a Markovian model, and then show that our BGLM-OFU algorithm and another algorithm based on the linear regression both solve such linear models with hidden variables. Our novelty includes (a) considering the combinatorial intervention action space and the general causal models including ones with hidden variables, (b) integrating and adapting techniques from diverse studies such as generalized linear bandits and online influence maximization, and (c) avoiding unrealistic assumptions (such as knowing the joint distribution of the parents of $Y$ under all interventions) and regret factors exponential to causal graph size in prior studies."
                    },
                    {
                        "title": "Machine learning assist nyc subway navigation safer and faster",
                        "abstract": "Mainstream navigation software, like Google and Apple Maps, often lacks the ability to provide routes prioritizing safety. However, safety remains a paramount concern for many. Our aim is to strike a balance between safety and efficiency. To achieve this, we're devising an Integer Programming model that takes into account both the shortest path and the safest route. We will harness machine learning to derive safety coefficients, employing methodologies such as generalized linear models, linear regression, and recurrent neural networks. Our evaluation will be based on the Root Mean Square Error (RMSE) across various subway stations, helping us identify the most accurate model for safety coefficient estimation. Furthermore, we'll conduct a comprehensive review of different shortest-path algorithms, assessing them based on time complexity and real-world data to determine their appropriateness in merging both safety and time efficiency."
                    }
                ]
            }
        }
    },
    "2312.00923": {
        "paper_data": {
            "title": "Label Delay in Online Continual Learning",
            "url": "http://arxiv.org/abs/2312.00923v2",
            "arxiv_id": "2312.00923",
            "authors": [
                "Botos Csaba",
                "Wenxuan Zhang",
                "Matthias M\u00fcller",
                "Ser-Nam Lim",
                "Mohamed Elhoseiny",
                "Philip Torr",
                "Adel Bibi"
            ],
            "abstract": "Online continual learning, the process of training models on streaming data, has gained increasing attention in recent years. However, a critical aspect often overlooked is the label delay, where new data may not be labeled due to slow and costly annotation processes. We introduce a new continual learning framework with explicit modeling of the label delay between data and label streams over time steps. In each step, the framework reveals both unlabeled data from the current time step $t$ and labels delayed with $d$ steps, from the time step $t-d$. In our extensive experiments amounting to 1060 GPU days, we show that merely augmenting the computational resources is insufficient to tackle this challenge. Our findings underline a notable performance decline when solely relying on labeled data when the label delay becomes significant. More surprisingly, when using state-of-the-art SSL and TTA techniques to utilize the newer, unlabeled data, they fail to surpass the performance of a na\\\"ive method that simply trains on the delayed supervised stream. To this end, we introduce a simple, efficient baseline that rehearses from the labeled memory samples that are most similar to the new unlabeled samples. This method bridges the accuracy gap caused by label delay without significantly increasing computational complexity. We show experimentally that our method is the least affected by the label delay factor and in some cases successfully recovers the accuracy of the non-delayed counterpart. We conduct various ablations and sensitivity experiments, demonstrating the effectiveness of our approach.",
            "introduction": "   1 Introduction  This document serves as an example submission to ECCV 2024. It illustrates the format authors must follow when submitting a paper. At the same time, it gives details on various aspects of paper submission, including preservation of anonymity and how to deal with dual submissions. We advise authors to read this document carefully.   The document is based on Springer LNCS instructions as well as on ECCV policies, as established over the years.     2 Initial Submission   2.1 Language  All manuscripts must be in English.     2.2 Template  Papers must be prepared with the official LNCS style from Springer. This applies to both review and camera-ready versions. Springer requires manuscripts to be prepared in LaTeX (strongly encouraged) or Microsoft Word.   Authors preparing their paper with LaTeX must use the template provided by ECCV, which is based on the corresponding Springer class file llncs.cls but includes line numbers for review (Sec.\u00a02.5) and properly anonymizes the paper for review (as in this example document). Authors who \u2013 for whatever reason \u2013 cannot use LaTeX can alternatively use the official LNCS Word template from Springer. However, it is the authors\u2019 responsibility to ensure that the resulting PDF file is consistent with this example paper and follows it as closely as possible (i.e., includes line numbers, is properly anonymized, etc.).   We would like to stress that the class/style files and the template must not be manipulated and that the guidelines regarding font sizes and format must be adhered to. For example, please refrain from using any LaTeX or TeX command that modifies the layout settings of the template (e.g., \\textheight, \\vspace, \\baselinestretch, etc.). Such manual layout adjustments should be limited to very exceptional cases. This is to ensure that the end product is as homogeneous as possible.   Papers that differ significantly from the required style may be rejected without review.    2.2.1 Fonts.  Springer\u2019s templates for LaTeX are based on CMR, and the XML templates for Word are based on Times. We ask you to use the font according to the template used for your papers. Specifically, please refrain from using Times when preparing your paper with LaTeX. Using a different font can be interpreted as purposely circumventing the length limitations and may lead to rejection without review.      2.3 Paper Length  Papers submitted for review must be complete. The length should match that intended for final publication. Papers accepted for the conference will be allocated 14 pages (plus additional pages for references) in the proceedings. Note that the allocated 14 pages do not include the references. The reason for this policy is that we do not want authors to omit references for sake of space limitations.   Papers with more than 14 pages (excluding references) will be rejected without review. This includes papers where the margins and formatting including the font are deemed to have been significantly altered from those laid down by this style guide.   The reason such papers will not be reviewed is that there is no provision for supervised revisions of manuscripts. The reviewing process cannot determine the suitability of the paper for presentation in 14 pages if it is reviewed in 16.     2.4 Paper ID  It is imperative that the paper ID is mentioned on each page of the manuscript of the review version. Enter your paper ID in the appropriate",
            "references": [
                {
                    "title": "Rapid Adaptation in Online Continual Learning: Are We Evaluating It Right?",
                    "abstract": "We revisit the common practice of evaluating adaptation of Online Continual Learning (OCL) algorithms through the metric of online accuracy, which measures the accuracy of the model on the immediate next few samples. However, we show that this metric is unreliable, as even vacuous blind classifiers, which do not use input images for prediction, can achieve unrealistically high online accuracy by exploiting spurious label correlations in the data stream. Our study reveals that existing OCL algorithms can also achieve high online accuracy, but perform poorly in retaining useful information, suggesting that they unintentionally learn spurious label correlations. To address this issue, we propose a novel metric for measuring adaptation based on the accuracy on the near-future samples, where spurious correlations are removed. We benchmark existing OCL approaches using our proposed metric on large-scale datasets under various computational budgets and find that better generalization can be achieved by retaining and reusing past seen information. We believe that our proposed metric can aid in the development of truly adaptive OCL methods. We provide code to reproduce our results at https://github.com/drimpossible/EvalOCL."
                },
                {
                    "title": "Online Continual Learning Without the Storage Constraint",
                    "abstract": "Traditional online continual learning (OCL) research has primarily focused on mitigating catastrophic forgetting with fixed and limited storage allocation throughout an agent's lifetime. However, a broad range of real-world applications are primarily constrained by computational costs rather than storage limitations. In this paper, we target such applications, investigating the online continual learning problem under relaxed storage constraints and limited computational budgets. We contribute a simple algorithm, which updates a kNN classifier continually along with a fixed, pretrained feature extractor. We selected this algorithm due to its exceptional suitability for online continual learning. It can adapt to rapidly changing streams, has zero stability gap, operates within tiny computational budgets, has low storage requirements by only storing features, and has a consistency property: It never forgets previously seen data. These attributes yield significant improvements, allowing our proposed algorithm to outperform existing methods by over 20% in accuracy on two large-scale OCL datasets: Continual LOCalization (CLOC) with 39M images and 712 classes and Continual Google Landmarks V2 (CGLM) with 580K images and 10,788 classes, even when existing methods retain all previously seen images. Furthermore, we achieve this superior performance with considerably reduced computational and storage expenses. We provide code to reproduce our results at github.com/drimpossible/ACM."
                },
                {
                    "title": "A Cookbook of Self-Supervised Learning",
                    "abstract": "Self-supervised learning, dubbed the dark matter of intelligence, is a promising path to advance machine learning. Yet, much like cooking, training SSL methods is a delicate art with a high barrier to entry. While many components are familiar, successfully training a SSL method involves a dizzying set of choices from the pretext tasks to training hyper-parameters. Our goal is to lower the barrier to entry into SSL research by laying the foundations and latest SSL recipes in the style of a cookbook. We hope to empower the curious researcher to navigate the terrain of methods, understand the role of the various knobs, and gain the know-how required to explore how delicious SSL can be."
                },
                {
                    "title": "Evaluation of Test-Time Adaptation Under Computational Time Constraints",
                    "abstract": "This paper proposes a novel online evaluation protocol for Test Time Adaptation (TTA) methods, which penalizes slower methods by providing them with fewer samples for adaptation. TTA methods leverage unlabeled data at test time to adapt to distribution shifts. Although many effective methods have been proposed, their impressive performance usually comes at the cost of significantly increased computation budgets. Current evaluation protocols overlook the effect of this extra computation cost, affecting their real-world applicability. To address this issue, we propose a more realistic evaluation protocol for TTA methods, where data is received in an online fashion from a constant-speed data stream, thereby accounting for the method's adaptation speed. We apply our proposed protocol to benchmark several TTA methods on multiple datasets and scenarios. Extensive experiments show that, when accounting for inference speed, simple and fast approaches can outperform more sophisticated but slower methods. For example, SHOT from 2020, outperforms the state-of-the-art method SAR from 2023 in this setting. Our results reveal the importance of developing practical TTA methods that are both accurate and efficient."
                },
                {
                    "title": "A Comprehensive Survey on Test-Time Adaptation under Distribution Shifts",
                    "abstract": "Machine learning methods strive to acquire a robust model during training that can generalize well to test samples, even under distribution shifts. However, these methods often suffer from a performance drop due to unknown test distributions. Test-time adaptation (TTA), an emerging paradigm, has the potential to adapt a pre-trained model to unlabeled data during testing, before making predictions. Recent progress in this paradigm highlights the significant benefits of utilizing unlabeled data for training self-adapted models prior to inference. In this survey, we divide TTA into several distinct categories, namely, test-time (source-free) domain adaptation, test-time batch adaptation, online test-time adaptation, and test-time prior adaptation. For each category, we provide a comprehensive taxonomy of advanced algorithms, followed by a discussion of different learning scenarios. Furthermore, we analyze relevant applications of TTA and discuss open challenges and promising areas for future research. A comprehensive list of TTA methods can be found at \\url{https://github.com/tim-learn/awesome-test-time-adaptation}."
                },
                {
                    "title": "Computationally Budgeted Continual Learning: What Does Matter?",
                    "abstract": "Continual Learning (CL) aims to sequentially train models on streams of incoming data that vary in distribution by preserving previous knowledge while adapting to new data. Current CL literature focuses on restricted access to previously seen data, while imposing no constraints on the computational budget for training. This is unreasonable for applications in-the-wild, where systems are primarily constrained by computational and time budgets, not storage. We revisit this problem with a large-scale benchmark and analyze the performance of traditional CL approaches in a compute-constrained setting, where effective memory samples used in training can be implicitly restricted as a consequence of limited computation. We conduct experiments evaluating various CL sampling strategies, distillation losses, and partial fine-tuning on two large-scale datasets, namely ImageNet2K and Continual Google Landmarks V2 in data incremental, class incremental, and time incremental settings. Through extensive experiments amounting to a total of over 1500 GPU-hours, we find that, under compute-constrained setting, traditional CL approaches, with no exception, fail to outperform a simple minimal baseline that samples uniformly from memory. Our conclusions are consistent in a different number of stream time steps, e.g., 20 to 200, and under several computational budgets. This suggests that most existing CL methods are particularly too computationally expensive for realistic budgeted deployment. Code for this project is available at: https://github.com/drimpossible/BudgetCL."
                },
                {
                    "title": "Towards Stable Test-Time Adaptation in Dynamic Wild World",
                    "abstract": "Test-time adaptation (TTA) has shown to be effective at tackling distribution shifts between training and testing data by adapting a given model on test samples. However, the online model updating of TTA may be unstable and this is often a key obstacle preventing existing TTA methods from being deployed in the real world. Specifically, TTA may fail to improve or even harm the model performance when test data have: 1) mixed distribution shifts, 2) small batch sizes, and 3) online imbalanced label distribution shifts, which are quite common in practice. In this paper, we investigate the unstable reasons and find that the batch norm layer is a crucial factor hindering TTA stability. Conversely, TTA can perform more stably with batch-agnostic norm layers, \\ie, group or layer norm. However, we observe that TTA with group and layer norms does not always succeed and still suffers many failure cases. By digging into the failure cases, we find that certain noisy test samples with large gradients may disturb the model adaption and result in collapsed trivial solutions, \\ie, assigning the same class label for all samples. To address the above collapse issue, we propose a sharpness-aware and reliable entropy minimization method, called SAR, for further stabilizing TTA from two aspects: 1) remove partial noisy samples with large gradients, 2) encourage model weights to go to a flat minimum so that the model is robust to the remaining noisy samples. Promising results demonstrate that SAR performs more stably over prior methods and is computationally efficient under the above wild test scenarios."
                },
                {
                    "title": "Real-Time Evaluation in Online Continual Learning: A New Hope",
                    "abstract": "Current evaluations of Continual Learning (CL) methods typically assume that there is no constraint on training time and computation. This is an unrealistic assumption for any real-world setting, which motivates us to propose: a practical real-time evaluation of continual learning, in which the stream does not wait for the model to complete training before revealing the next data for predictions. To do this, we evaluate current CL methods with respect to their computational costs. We conduct extensive experiments on CLOC, a large-scale dataset containing 39 million time-stamped images with geolocation labels. We show that a simple baseline outperforms state-of-the-art CL methods under this evaluation, questioning the applicability of existing methods in realistic settings. In addition, we explore various CL components commonly used in the literature, including memory sampling strategies and regularization approaches. We find that all considered methods fail to be competitive against our simple baseline. This surprisingly suggests that the majority of existing CL literature is tailored to a specific class of streams that is not practical. We hope that the evaluation we provide will be the first step towards a paradigm shift to consider the computational cost in the development of online continual learning methods."
                },
                {
                    "title": "A Novel Machine Learning Method for Delayed Labels",
                    "abstract": "Most research on machine learning relies on the availability of ground truth labels immediately after prediction. However, in many cases, the ground truth labels become available with a non-negligible delay. Considering that there is a large amount of unlabeled data in delayed labels, supervised model cannot utilize unlabeled data. Therefore, most of the research on delayed labels begins to train semi-supervised models in delayed labels. However, most research on delayed labels ignores that the labels of unlabeled data will arrive after several periods in delayed labels. Neither supervised nor semi-supervised models can solve the problem in delayed labels effectively. Besides, there remains a problem of concept drift due to the long period of data. In this paper, we propose an incremental learning model that can adapt to delayed labels. First, we should detect whether the concept drift takes place. Then we use knowledge distillation to update supervised and semi-supervised models while retaining the corresponding knowledge of past labeled data. Finally, we combine the supervised and semi-supervised models to make predictions. Finally, we apply our algorithms to synthetic and real credit scoring datasets. The experiment results indicate our algorithms have superiority in delayed labels."
                },
                {
                    "title": "Wild-Time: A Benchmark of in-the-Wild Distribution Shift over Time",
                    "abstract": "Distribution shift occurs when the test distribution differs from the training distribution, and it can considerably degrade performance of machine learning models deployed in the real world. Temporal shifts -- distribution shifts arising from the passage of time -- often occur gradually and have the additional structure of timestamp metadata. By leveraging timestamp metadata, models can potentially learn from trends in past distribution shifts and extrapolate into the future. While recent works have studied distribution shifts, temporal shifts remain underexplored. To address this gap, we curate Wild-Time, a benchmark of 5 datasets that reflect temporal distribution shifts arising in a variety of real-world applications, including patient prognosis and news classification. On these datasets, we systematically benchmark 13 prior approaches, including methods in domain generalization, continual learning, self-supervised learning, and ensemble learning. We use two evaluation strategies: evaluation with a fixed time split (Eval-Fix) and evaluation with a data stream (Eval-Stream). Eval-Fix, our primary evaluation strategy, aims to provide a simple evaluation protocol, while Eval-Stream is more realistic for certain real-world applications. Under both evaluation strategies, we observe an average performance drop of 20% from in-distribution to out-of-distribution data. Existing methods are unable to close this gap. Code is available at https://wild-time.github.io/."
                },
                {
                    "title": "SCALE: Online Self-Supervised Lifelong Learning without Prior Knowledge",
                    "abstract": "Unsupervised lifelong learning refers to the ability to learn over time while memorizing previous patterns without supervision. Although great progress has been made in this direction, existing work often assumes strong prior knowledge about the incoming data (e.g., knowing the class boundaries), which can be impossible to obtain in complex and unpredictable environments. In this paper, motivated by real-world scenarios, we propose a more practical problem setting called online self-supervised lifelong learning without prior knowledge. The proposed setting is challenging due to the non-iid and single-pass data, the absence of external supervision, and no prior knowledge. To address the challenges, we propose Self-Supervised ContrAstive Life long LEarning without Prior Knowledge (SCALE) which can extract and memorize representations on the fly purely from the data continuum. SCALE is designed around three major components: a pseudo-supervised contrastive loss, a self-supervised forgetting loss, and an online memory update for uniform subset selection. All three components are designed to work collaboratively to maximize learning performance. We perform comprehensive experiments of SCALE under iid and four non-iid data streams. The results show that SCALE outperforms the state-of-the-art algorithm in all settings with improvements up to 3.83%, 2.77% and 5.86% in terms of kNN accuracy on CIFAR-10, CIFAR-100, and TinyImageNet datasets. We release the implementation at https://github.com/Orienfish/SCALE."
                },
                {
                    "title": "Efficient Test-Time Model Adaptation without Forgetting",
                    "abstract": "Test-time adaptation (TTA) seeks to tackle potential distribution shifts between training and testing data by adapting a given model w.r.t. any testing sample. This task is particularly important for deep models when the test environment changes frequently. Although some recent attempts have been made to handle this task, we still face two practical challenges: 1) existing methods have to perform backward computation for each test sample, resulting in unbearable prediction cost to many applications; 2) while existing TTA solutions can significantly improve the test performance on out-of-distribution data, they often suffer from severe performance degradation on in-distribution data after TTA (known as catastrophic forgetting). In this paper, we point out that not all the test samples contribute equally to model adaptation, and high-entropy ones may lead to noisy gradients that could disrupt the model. Motivated by this, we propose an active sample selection criterion to identify reliable and non-redundant samples, on which the model is updated to minimize the entropy loss for test-time adaptation. Furthermore, to alleviate the forgetting issue, we introduce a Fisher regularizer to constrain important model parameters from drastic changes, where the Fisher importance is estimated from test samples with generated pseudo labels. Extensive experiments on CIFAR-10-C, ImageNet-C, and ImageNet-R verify the effectiveness of our proposed method."
                },
                {
                    "title": "Online Continual Learning on a Contaminated Data Stream with Blurry Task Boundaries",
                    "abstract": "Learning under a continuously changing data distribution with incorrect labels is a desirable real-world problem yet challenging. A large body of continual learning (CL) methods, however, assumes data streams with clean labels, and online learning scenarios under noisy data streams are yet underexplored. We consider a more practical CL task setup of an online learning from blurry data stream with corrupted labels, where existing CL methods struggle. To address the task, we first argue the importance of both diversity and purity of examples in the episodic memory of continual learning models. To balance diversity and purity in the episodic memory, we propose a novel strategy to manage and use the memory by a unified approach of label noise aware diverse sampling and robust learning with semi-supervised learning. Our empirical validations on four real-world or synthetic noise datasets (CI-FAR10 and 100, mini-WebVision, and Food-101N) exhibit that our method significantly outperforms prior arts in this realistic and challenging continual learning scenario. Code and data splits are available in https://github.com/clovaai/puridiver."
                },
                {
                    "title": "Continual Test-Time Domain Adaptation",
                    "abstract": "Test-time domain adaptation aims to adapt a source pre-trained model to a target domain without using any source data. Existing works mainly consider the case where the target domain is static. However, real-world machine perception systems are running in non-stationary and continually changing environments where the target domain distribution can change over time. Existing methods, which are mostly based on self-training and entropy regularization, can suffer from these non-stationary environments. Due to the distribution shift over time in the target domain, pseudo-labels become unreliable. The noisy pseudo-labels can further lead to error accumulation and catastrophic forgetting. To tackle these issues, we propose a continual test-time adaptation approach (CoTTA) which comprises two parts. Firstly, we propose to reduce the error accumulation by using weight-averaged and augmentation-averaged predictions which are often more accurate. On the other hand, to avoid catastrophic forgetting, we propose to stochastically restore a small part of the neurons to the source pre-trained weights during each iteration to help preserve source knowledge in the longterm. The proposed method enables the longterm adaptation for all parameters in the network. CoTTA is easy to implement and can be readily incorporated in off-the-shelf pre-trained models. We demonstrate the effectiveness of our approach on four classification tasks and a segmentation task for continual test-time adaptation, on which we outperform existing methods. Our code is available at https://gin.ee/cotta."
                },
                {
                    "title": "Self-Supervised Models are Continual Learners",
                    "abstract": "Self-supervised models have been shown to produce comparable or better visual representations than their su-pervised counterparts when trained offline on unlabeled data at scale. However, their efficacy is catastrophically reduced in a Continual Learning (CL) scenario where data is presented to the model sequentially. In this paper, we show that self-supervised loss functions can be seamlessly converted into distillation mechanisms for CL by adding a predictor network that maps the current state of the repre-sentations to their past state. This enables us to devise a framework for Continual self-supervised visual representation Learning that (i) significantly improves the quality of the learned representations, (ii) is compatible with several state-of-the-art self-supervised objectives, and (iii) needs little to no hyperparameter tuning. We demonstrate the ef-fectiveness of our approach empirically by training six pop-ular self-supervised models in various CL settings. Code: github.com/DonkeyShot21/cassle."
                },
                {
                    "title": "Online Continual Learning with Natural Distribution Shifts: An Empirical Study with Visual Data",
                    "abstract": "Continual learning is the problem of learning and retaining knowledge through time over multiple tasks and environments. Research has primarily focused on the incremental classification setting, where new tasks/classes are added at discrete time intervals. Such an \"offline\" setting does not evaluate the ability of agents to learn effectively and efficiently, since an agent can perform multiple learning epochs without any time limitation when a task is added. We argue that \"online\" continual learning, where data is a single continuous stream without task boundaries, enables evaluating both information retention and online learning efficacy. In online continual learning, each incoming small batch of data is first used for testing and then added to the training set, making the problem truly online. Trained models are later evaluated on historical data to assess information retention. We introduce a new benchmark for online continual visual learning that exhibits large scale and natural distribution shifts. Through a large-scale analysis, we identify critical and previously unobserved phenomena of gradient-based optimization in continual learning, and propose effective strategies for improving gradient-based online continual learning with real data. The source code and dataset are available in: https://github.com/IntelLabs/continuallearning."
                },
                {
                    "title": "A Survey on Semi-supervised Learning for Delayed Partially Labelled Data Streams",
                    "abstract": "Unlabelled data appear in many domains and are particularly relevant to streaming applications, where even though data is abundant, labelled data is rare. To address the learning problems associated with such data, one can ignore the unlabelled data and focus only on the labelled data (supervised learning); use the labelled data and attempt to leverage the unlabelled data (semi-supervised learning); or assume some labels will be available on request (active learning). The first approach is the simplest, yet the amount of labelled data available will limit the predictive performance. The second relies on finding and exploiting the underlying characteristics of the data distribution. The third depends on an external agent to provide the required labels in a timely fashion. This survey pays special attention to methods that leverage unlabelled data in a semi-supervised setting. We also discuss the delayed labelling issue, which impacts both fully supervised and semi-supervised methods. We propose a unified problem setting, discuss the learning guarantees and existing methods, and explain the differences between related problem settings. Finally, we review the current benchmarking practices and propose adaptations to enhance them."
                },
                {
                    "title": "Online Learning with Optimism and Delay",
                    "abstract": "Inspired by the demands of real-time climate and weather forecasting, we develop optimistic online learning algorithms that require no parameter tuning and have optimal regret guarantees under delayed feedback. Our algorithms -- DORM, DORM+, and AdaHedgeD -- arise from a novel reduction of delayed online learning to optimistic online learning that reveals how optimistic hints can mitigate the regret penalty caused by delay. We pair this delay-as-optimism perspective with a new analysis of optimistic learning that exposes its robustness to hinting errors and a new meta-algorithm for learning effective hinting strategies in the presence of delay. We conclude by benchmarking our algorithms on four subseasonal climate forecasting tasks, demonstrating low regret relative to state-of-the-art forecasting models."
                },
                {
                    "title": "VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning",
                    "abstract": "Recent self-supervised methods for image representation learning are based on maximizing the agreement between embedding vectors from different views of the same image. A trivial solution is obtained when the encoder outputs constant vectors. This collapse problem is often avoided through implicit biases in the learning architecture, that often lack a clear justification or interpretation. In this paper, we introduce VICReg (Variance-Invariance-Covariance Regularization), a method that explicitly avoids the collapse problem with a simple regularization term on the variance of the embeddings along each dimension individually. VICReg combines the variance term with a decorrelation mechanism based on redundancy reduction and covariance regularization, and achieves results on par with the state of the art on several downstream tasks. In addition, we show that incorporating our new variance term into other methods helps stabilize the training and leads to performance improvements."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "With a Little Help from My Friends: Nearest-Neighbor Contrastive Learning of Visual Representations",
                    "abstract": "Self-supervised learning algorithms based on instance discrimination train encoders to be invariant to pre-defined transformations of the same instance. While most methods treat different views of the same image as positives for a contrastive loss, we are interested in using positives from other instances in the dataset. Our method, Nearest-Neighbor Contrastive Learning of visual Representations (NNCLR), samples the nearest neighbors from the dataset in the latent space, and treats them as positives. This provides more semantic variations than pre-defined transformations.We find that using the nearest-neighbor as positive in contrastive losses improves performance significantly on ImageNet classification using ResNet-50 under the linear evaluation protocol, from 71.7% to 75.6%, outperforming previous state-of-the-art methods. On semi-supervised learning benchmarks we improve performance significantly when only 1% ImageNet labels are available, from 53.8% to 56.5%. On transfer learning benchmarks our method outperforms state-of-the-art methods (including supervised learning with ImageNet) on 8 out of 12 downstream datasets. Furthermore, we demonstrate empirically that our method is less reliant on complex data augmentations. We see a relative reduction of only 2.1% ImageNet Top-1 accuracy when we train using only random crops."
                },
                {
                    "title": "New Insights on Reducing Abrupt Representation Change in Online Continual Learning",
                    "abstract": "In the online continual learning paradigm, agents must learn from a changing distribution while respecting memory and compute constraints. Experience Replay (ER), where a small subset of past data is stored and replayed alongside new data, has emerged as a simple and effective learning strategy. In this work, we focus on the change in representations of observed data that arises when previously unobserved classes appear in the incoming data stream, and new classes must be distinguished from previous ones. We shed new light on this question by showing that applying ER causes the newly added classes' representations to overlap significantly with the previous classes, leading to highly disruptive parameter updates. Based on this empirical analysis, we propose a new method which mitigates this issue by shielding the learned representations from drastic adaptation to accommodate new classes. We show that using an asymmetric update rule pushes new classes to adapt to the older ones (rather than the reverse), which is more effective especially at task boundaries, where much of the forgetting typically occurs. Empirical results show significant gains over strong baselines on standard continual learning benchmarks."
                },
                {
                    "title": "An Empirical Study of Training Self-Supervised Vision Transformers",
                    "abstract": "This paper does not describe a novel method. Instead, it studies a straightforward, incremental, yet must-know baseline given the recent progress in computer vision: self-supervised learning for Vision Transformers (ViT). While the training recipes for standard convolutional networks have been highly mature and robust, the recipes for ViT are yet to be built, especially in the self-supervised scenarios where training becomes more challenging. In this work, we go back to basics and investigate the effects of several fundamental components for training self-supervised ViT. We observe that instability is a major issue that degrades accuracy, and it can be hidden by apparently good results. We reveal that these results are indeed partial failure, and they can be improved when training is made more stable. We benchmark ViT results in MoCo v3 and several other self-supervised frameworks, with ablations in various aspects. We discuss the currently positive evidence as well as challenges and open questions. We hope that this work will provide useful data points and experience for future research."
                },
                {
                    "title": "Barlow Twins: Self-Supervised Learning via Redundancy Reduction",
                    "abstract": "Self-supervised learning (SSL) is rapidly closing the gap with supervised methods on large computer vision benchmarks. A successful approach to SSL is to learn embeddings which are invariant to distortions of the input sample. However, a recurring issue with this approach is the existence of trivial constant solutions. Most current methods avoid such solutions by careful implementation details. We propose an objective function that naturally avoids collapse by measuring the cross-correlation matrix between the outputs of two identical networks fed with distorted versions of a sample, and making it as close to the identity matrix as possible. This causes the embedding vectors of distorted versions of a sample to be similar, while minimizing the redundancy between the components of these vectors. The method is called Barlow Twins, owing to neuroscientist H. Barlow's redundancy-reduction principle applied to a pair of identical networks. Barlow Twins does not require large batches nor asymmetry between the network twins such as a predictor network, gradient stopping, or a moving average on the weight updates. Intriguingly it benefits from very high-dimensional output vectors. Barlow Twins outperforms previous methods on ImageNet for semi-supervised classification in the low-data regime, and is on par with current state of the art for ImageNet classification with a linear classifier head, and for transfer tasks of classification and object detection."
                },
                {
                    "title": "Exploring Simple Siamese Representation Learning",
                    "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code is made available.1"
                },
                {
                    "title": "Whitening for Self-Supervised Representation Learning",
                    "abstract": "Recent literature on self-supervised learning is based on the contrastive loss, where image instances which share the same semantic content (\"positives\") are contrasted with instances extracted from other images (\"negatives\"). However, in order for the learning to be effective, a lot of negatives should be compared with a positive pair. This is not only computationally demanding, but it also requires that the positive and the negative representations are kept consistent with each other over a long training period. In this paper we propose a different direction and a new loss function for self-supervised learning which is based on the whitening of the latent-space features. The whitening operation has a \"scattering\" effect on the batch samples, which compensates the lack of a large number of negatives, avoiding degenerate solutions where all the sample representations collapse to a single point. We empirically show that our loss accelerates self-supervised training and the learned representations are much more effective for downstream tasks than previously published work."
                },
                {
                    "title": "Unsupervised Learning of Visual Features by Contrasting Cluster Assignments",
                    "abstract": "Unsupervised image representations have significantly reduced the gap with supervised pretraining, notably with the recent achievements of contrastive learning methods. These contrastive methods typically work online and rely on a large number of explicit pairwise feature comparisons, which is computationally challenging. In this paper, we propose an online algorithm, SwAV, that takes advantage of contrastive methods without requiring to compute pairwise comparisons. Specifically, our method simultaneously clusters the data while enforcing consistency between cluster assignments produced for different augmentations (or views) of the same image, instead of comparing features directly as in contrastive learning. Simply put, we use a swapped prediction mechanism where we predict the cluster assignment of a view from the representation of another view. Our method can be trained with large and small batches and can scale to unlimited amounts of data. Compared to previous contrastive methods, our method is more memory efficient since it does not require a large memory bank or a special momentum network. In addition, we also propose a new data augmentation strategy, multi-crop, that uses a mix of views with different resolutions in place of two full-resolution views, without increasing the memory or compute requirements much. We validate our findings by achieving 75.3% top-1 accuracy on ImageNet with ResNet-50, as well as surpassing supervised pretraining on all the considered transfer tasks."
                },
                {
                    "title": "Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning",
                    "abstract": "We introduce Bootstrap Your Own Latent (BYOL), a new approach to self-supervised image representation learning. BYOL relies on two neural networks, referred to as online and target networks, that interact and learn from each other. From an augmented view of an image, we train the online network to predict the target network representation of the same image under a different augmented view. At the same time, we update the target network with a slow-moving average of the online network. While state-of-the art methods rely on negative pairs, BYOL achieves a new state of the art without them. BYOL reaches $74.3\\%$ top-1 classification accuracy on ImageNet using a linear evaluation with a ResNet-50 architecture and $79.6\\%$ with a larger ResNet. We show that BYOL performs on par or better than the current state of the art on both transfer and semi-supervised benchmarks. Our implementation and pretrained models are given on GitHub."
                },
                {
                    "title": "Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation",
                    "abstract": "Unsupervised domain adaptation (UDA) aims to leverage the knowledge learned from a labeled source dataset to solve similar tasks in a new unlabeled domain. Prior UDA methods typically require to access the source data when learning to adapt the model, making them risky and inefficient for decentralized private data. This work tackles a practical setting where only a trained source model is available and investigates how we can effectively utilize such a model without source data to solve UDA problems. We propose a simple yet generic representation learning framework, named \\emph{Source HypOthesis Transfer} (SHOT). SHOT freezes the classifier module (hypothesis) of the source model and learns the target-specific feature extraction module by exploiting both information maximization and self-supervised pseudo-labeling to implicitly align representations from the target domains to the source hypothesis. To verify its versatility, we evaluate SHOT in a variety of adaptation cases including closed-set, partial-set, and open-set domain adaptation. Experiments indicate that SHOT yields state-of-the-art results among multiple domain adaptation benchmarks."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "Momentum Contrast for Unsupervised Visual Representation Learning",
                    "abstract": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a moving-averaged encoder. This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning. MoCo provides competitive results under the common linear protocol on ImageNet classification. More importantly, the representations learned by MoCo transfer well to downstream tasks. MoCo can outperform its supervised pre-training counterpart in 7 detection/segmentation tasks on PASCAL VOC, COCO, and other datasets, sometimes surpassing it by large margins. This suggests that the gap between unsupervised and supervised representation learning has been largely closed in many vision tasks."
                },
                {
                    "title": "Online Continual Learning with Maximally Interfered Retrieval",
                    "abstract": "Continual learning, the setting where a learning agent is faced with a never ending stream of data, continues to be a great challenge for modern machine learning systems. In particular the online or \"single-pass through the data\" setting has gained attention recently as a natural setting that is difficult to tackle. Methods based on replay, either generative or from a stored memory, have been shown to be effective approaches for continual learning, matching or exceeding the state of the art in a number of standard benchmarks. These approaches typically rely on randomly selecting samples from the replay memory or from a generative model, which is suboptimal. In this work, we consider a controlled sampling of memories for replay. We retrieve the samples which are most interfered, i.e. whose prediction will be most negatively impacted by the foreseen parameters update. We show a formulation for this sampling criterion in both the generative replay and the experience replay setting, producing consistent gains in performance and greatly reduced forgetting. We release an implementation of our method at this https URL."
                },
                {
                    "title": "S4L: Self-Supervised Semi-Supervised Learning",
                    "abstract": "This work tackles the problem of semi-supervised learning of image classifiers. Our main insight is that the field of semi-supervised learning can benefit from the quickly advancing field of self-supervised visual representation learning. Unifying these two approaches, we propose the framework of self-supervised semi-supervised learning (S4L) and use it to derive two novel semi-supervised image classification methods. We demonstrate the effectiveness of these methods in comparison to both carefully tuned baselines, and existing semi-supervised learning methods. We then show that S4L and existing semi-supervised methods can be jointly trained, yielding a new state-of-the-art result on semi-supervised ILSVRC-2012 with 10% of labels."
                },
                {
                    "title": "Gradient based sample selection for online continual learning",
                    "abstract": "A continual learning agent learns online with a non-stationary and never-ending stream of data. The key to such learning process is to overcome the catastrophic forgetting of previously seen data, which is a well known problem of neural networks. To prevent forgetting, a replay buffer is usually employed to store the previous data for the purpose of rehearsal. Previous works often depend on task boundary and i.i.d. assumptions to properly select samples for the replay buffer. In this work, we formulate sample selection as a constraint reduction problem based on the constrained optimization view of continual learning. The goal is to select a fixed subset of constraints that best approximate the feasible region defined by the original constraints. We show that it is equivalent to maximizing the diversity of samples in the replay buffer with parameters gradient as the feature. We further develop a greedy alternative that is cheap and efficient. The advantage of the proposed method is demonstrated by comparing to other alternatives under the continual learning setting. Further comparisons are made against state of the art methods that rely on task boundaries which show comparable or even better results for our method."
                },
                {
                    "title": "Deep Learning in Asset Pricing",
                    "abstract": "We use deep neural networks to estimate an asset pricing model for individual stock returns that takes advantage of the vast amount of conditioning information, keeps a fully flexible form, and accounts for time variation. The key innovations are to use the fundamental no-arbitrage condition as criterion function to construct the most informative test assets with an adversarial approach and to extract the states of the economy from many macroeconomic time series. Our asset pricing model outperforms out-of-sample all benchmark approaches in terms of Sharpe ratio, explained variation, and pricing errors and identifies the key factors that drive asset prices. This paper was accepted by Agostino Capponi, finance. Supplemental Material: The online appendix and data are available at https://doi.org/10.1287/mnsc.2023.4695 ."
                },
                {
                    "title": "Continual Learning with Tiny Episodic Memories",
                    "abstract": "Learning with less supervision is a major challenge in artificial intelligence. One sensible approach to decrease the amount of supervision is to leverage prior experience and transfer knowledge from tasks seen in the past. However, a necessary condition for a successful transfer is the ability to remember how to perform previous tasks. The Continual Learning (CL) setting, whereby an agent learns from a stream of tasks without seeing any example twice, is an ideal framework to investigate how to accrue such knowledge. In this work, we consider supervised learning tasks and methods that leverage a very small episodic memory for continual learning. Through an extensive empirical analysis across four benchmark datasets adapted to CL, we observe that a very simple baseline, which jointly trains on both examples from the current task as well as examples stored in the memory, outperforms state-of-the-art CL approaches with and without episodic memory. Surprisingly, repeated learning over tiny episodic memories does not harm generalization on past tasks, as joint training on data from subsequent tasks acts like a data dependent regularizer. We discuss and evaluate different approaches to write into the memory. Most notably, reservoir sampling works remarkably well across the board, except when the memory size is extremely small. In this case, writing strategies that guarantee an equal representation of all classes work better. Overall, these methods should be considered as a strong baseline candidate when benchmarking new CL approaches"
                },
                {
                    "title": "Task-Free Continual Learning",
                    "abstract": "Methods proposed in the literature towards continual deep learning typically operate in a task-based sequential learning setup. A sequence of tasks is learned, one at a time, with all data of current task available but not of previous or future tasks. Task boundaries and identities are known at all times. This setup, however, is rarely encountered in practical applications. Therefore we investigate how to transform continual learning to an online setup. We develop a system that keeps on learning over time in a streaming fashion, with data distributions gradually changing and without the notion of separate tasks. To this end, we build on the work on Memory Aware Synapses, and show how this method can be made online by providing a protocol to decide i) when to update the importance weights, ii) which data to use to update them, and iii) how to accumulate the importance weights at each update step. Experimental results show the validity of the approach in the context of two applications: (self-)supervised learning of a face recognition model by watching soap series and learning a robot to avoid collisions."
                },
                {
                    "title": "Functional Map of the World",
                    "abstract": "We present a new dataset, Functional Map of the World (fMoW), which aims to inspire the development of machine learning models capable of predicting the functional purpose of buildings and land use from temporal sequences of satellite images and a rich set of metadata features. The metadata provided with each image enables reasoning about location, time, sun angles, physical sizes, and other features when making predictions about objects in the image. Our dataset consists of over 1 million images from over 200 countries. For each image, we provide at least one bounding box annotation containing one of 63 categories, including a \"false detection\" category. We present an analysis of the dataset along with baseline approaches that reason about metadata and temporal views. Our data, code, and pretrained models have been made publicly available."
                },
                {
                    "title": "Gradient Episodic Memory for Continual Learning",
                    "abstract": "One major obstacle towards AI is the poor ability of models to solve new problems quicker, and without forgetting previously acquired knowledge. To better understand this issue, we study the problem of continual learning, where the model observes, once and one by one, examples concerning a sequence of tasks. First, we propose a set of metrics to evaluate models learning over a continuum of data. These metrics characterize models not only by their test accuracy, but also in terms of their ability to transfer knowledge across tasks. Second, we propose a model for continual learning, called Gradient Episodic Memory (GEM) that alleviates forgetting, while allowing beneficial transfer of knowledge to previous tasks. Our experiments on variants of the MNIST and CIFAR-100 datasets demonstrate the strong performance of GEM when compared to the state-of-the-art."
                },
                {
                    "title": "Sliding Reservoir Approach for Delayed Labeling in Streaming Data Classification",
                    "abstract": "When concept drift occurs within streaming data, a streaming data classification framework needs to update the learning model to maintain its performance. Labeled samples required for training a new model are often unavailable immediately in real world applications. This delay of labels might negatively impact the performance of traditional streaming data classification frameworks. To solve this problem, we propose Sliding Reservoir Approach for Delayed Labeling (SRADL). By combining chunk based semisupervised learning with a novel approach to manage labeled data, SRADL does not need to wait for the labeling process to finish before updating the learning model. Experiments with two delayed-label scenarios show that SRADL improves prediction performance over the na\u00efve approach by as much as 7.5% in certain cases. The most gain comes from 18-chunk labeling delay time with continuous labeling delivery scenario in real world data experiments."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Online Learning with Adversarial Delays",
                    "abstract": "We study the performance of standard online learning algorithms when the feedback is delayed by an adversary. We show that online-gradient-descent [1] and follow-the-perturbed-leader [2] achieve regret O(\u221aD) in the delayed setting, where D is the sum of delays of each round's feedback. This bound collapses to an optimal O(\u221aT) bound in the usual setting of no delays (where D = T). Our main contribution is to show that standard algorithms for online learning already have simple regret bounds in the most general setting of delayed feedback, making adjustments to the analysis and not to the algorithms themselves. Our results help affirm and clarify the success of recent algorithms in optimization and machine learning that operate in a delayed feedback model."
                },
                {
                    "title": "Classification of Evolving Data Streams with Infinitely Delayed Labels",
                    "abstract": "The majority of evolving data streams classification algorithms assume that the actual labels of the predicted examples are readily available without any time delay just after a prediction is made. However, given the high label costs, dependence of an expert, limitations in data transmission or even restrictions imposed by the problem's nature, there is a large number of real-world applications in which the availability of actual labels is infinitely delayed (never available). In these cases, it is necessary the use of algorithms that does not follow the traditional process of monitoring the error rate to detect changes in data distribution and uses the most recent labeled data to update the classification model. In this paper, we propose the method MClassification to classify evolving data streams with infinitely delayed labels. Our method is inspired on the use of Micro-Cluster representation from online clustering algorithms. Considering the presence of incremental drifts, our approach uses a distance-based strategy to maintain the Micro-Clusters' positions updated. An evaluation in several synthetic and real data shows that MClassification achieves competitive accuracy results to state-of-the-art methods and adequate computational cost. The main advantage of the proposed method is the absence of critical parameters that require user's prior knowledge, as occurs with rival methods."
                },
                {
                    "title": "A Century of Portraits: A Visual Historical Record of American High School Yearbooks",
                    "abstract": "Many details about our world are not captured in written records because they are too mundane or too abstract to describe in words. Fortunately, since the invention of the camera, an ever-increasing number of photographs capture much of this otherwise lost information. This plethora of artifacts documenting our \"visual culture\" is a treasure trove of knowledge as yet untapped by historians. We present a dataset of 37,921 frontal-facing American high school yearbook photos that allow us to use computation to glimpse into the historical visual record too voluminous to be evaluated manually. The collected portraits provide a constant visual frame of reference with varying content. We can therefore use them to consider issues such as a decade's defining style elements, or trends in fashion and social norms over time. We demonstrate that our historical image dataset may be used together with weakly-supervised data-driven techniques to perform scalable historical analysis of large image corpora with minimal human effort, much in the same way that large text corpora together with natural language processing revolutionized historians' workflow. Furthermore, we demonstrate the use of our dataset in dating grayscale portraits using deep learning methods."
                },
                {
                    "title": "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift",
                    "abstract": "Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.82% top-5 test error, exceeding the accuracy of human raters."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "Online Learning under Delayed Feedback",
                    "abstract": "Online learning with delayed feedback has received increasing attention recently due to its several applications in distributed, web-based learning problems. In this paper we provide a systematic study of the topic, and analyze the effect of delay on the regret of online learning algorithms. Somewhat surprisingly, it turns out that delay increases the regret in a multiplicative way in adversarial problems, and in an additive way in stochastic problems. We give meta-algorithms that transform, in a black-box fashion, algorithms developed for the non-delayed case into ones that can handle the presence of delays in the feedback loop. Modifications of the well-known UCB algorithm are also developed for the bandit problem with delayed feedback, with the advantage over the meta-algorithms that they can be implemented with lower complexity."
                },
                {
                    "title": "Nearest Neighbour Classifiers for Streaming Data with Delayed Labelling",
                    "abstract": "We study streaming data where the true labels come with a delay. The question is whether the online nearest neighbour classifier (IB2 and IB3 here) should employ the unlabelled data. Three strategies are examined: do-nothing, replace and forget. Experiments with 28 data sets show that IB2 benefits from unlabelled data, while IB3 does not."
                },
                {
                    "title": "On delayed prediction of individual sequences",
                    "abstract": "We investigate a prediction scenario in which the predictor is forced to make a decision a number of steps in advance, with incomplete information. For finite action and observation spaces, it is shown that the strategy that minimizes the worst-case regret with respect to the Bayes envelope is obtained through sub-sampling of the sequence of observations. The result extends to the case of logarithmic loss."
                },
                {
                    "title": "Tent: Fully Test-Time Adaptation by Entropy Minimization",
                    "abstract": "A model must adapt itself to generalize to new and different data during testing. In this setting of fully test-time adaptation the model has only the test data and its own parameters. We propose to adapt by test entropy minimization (tent1): we optimize the model for confidence as measured by the entropy of its predictions. Our method estimates normalization statistics and optimizes channel-wise affine transformations to update online on each batch. Tent reduces generalization error for image classification on corrupted ImageNet and CIFAR-10/100 and reaches a new state-of-the-art error on ImageNet-C. Tent handles source-free domain adaptation on digit recognition from SVHN to MNIST/MNIST-M/USPS, on semantic segmentation from GTA to Cityscapes, and on the VisDA-C benchmark. These results are achieved in one epoch of test-time optimization without altering training."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CV",
                "I.4.0; I.4.10"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "Certainly! Here are the responses to the five core questions based on the provided introduction regarding the submission guidelines for ECCV 2024:\n\n[Question 1] - What is the problem?\nHow can we ensure that the submission process for ECCV 2024 is streamlined and adheres to the required formatting and anonymity guidelines?\n\n[Question 2] - Why is it interesting and important?\nSolving this problem is crucial for maintaining the integrity and quality of the review process in the research community. Proper adherence to submission guidelines ensures that all papers are evaluated fairly and consistently, which can lead to higher standards in research outputs. Addressing this issue could advance knowledge by fostering a more organized and efficient submission process, ultimately benefiting future conferences and publications.\n\n[Question 3] - Why is it hard?\nThe challenges in solving this problem include the diverse range of formatting tools and styles used by authors, which can lead to inconsistencies in submissions. Naive approaches, such as simply providing guidelines without enforcement, may fail because authors might overlook critical formatting details. Technical obstacles include ensuring that all authors are familiar with LaTeX or Word templates, and practical obstacles involve the potential for human error in adhering to strict formatting rules.\n\n[Question 4] - Why hasn't it been solved before?\nPrevious research and guidelines may have lacked clarity or comprehensiveness, leading to misunderstandings among authors. Barriers such as the complexity of formatting requirements and the variability in authors' familiarity with submission tools have prevented a standardized approach. Our approach differs by providing a clear, detailed template and emphasizing the importance of adherence to formatting rules, which can improve the submission process significantly.\n\n[Question 5] - What are the key components of my approach and results?\nOur proposed methodology includes the development of a comprehensive submission template based on the official LNCS style, which will be made available in both LaTeX and Word formats. We will utilize a checklist to ensure compliance with formatting and anonymity requirements. The expected outcomes include a higher rate of correctly formatted submissions, reduced review time, and an overall improvement in the quality of the conference proceedings."
            }
        },
        "author_data": {
            "b3dd7417-587f-4880-b49f-fc8a6c1f78af": {
                "pk": "b3dd7417-587f-4880-b49f-fc8a6c1f78af",
                "name": "Botos Csaba",
                "collaborators": [
                    "Philip Torr",
                    "Philip H. S. Torr",
                    "Adel Bibi",
                    "Xiaojuan Qi",
                    "Arslan Chaudhry",
                    "Puneet Dokania",
                    "Adnane Boukhayma",
                    "Viveka Kulharia",
                    "Andr\u00e1s Horv\u00e1th",
                    "Yanwei Li"
                ],
                "domain": [
                    "Domain Adaptation",
                    "Generative Adversarial Networks",
                    "Computer Vision",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Multilevel Knowledge Transfer for Cross-Domain Object Detection",
                        "abstract": "Domain shift is a well known problem where a model trained on a particular domain (source) does not perform well when exposed to samples from a different domain (target). Unsupervised methods that can adapt to domain shift are highly desirable as they allow effective utilization of the source data without requiring additional annotated training data from the target. Practically, obtaining sufficient amount of annotated data from the target domain can be both infeasible and extremely expensive. In this work, we address the domain shift problem for the object detection task. Our approach relies on gradually removing the domain shift between the source and the target domains. The key ingredients to our approach are -- (a) mapping the source to the target domain on pixel-level; (b) training a teacher network on the mapped source and the unannotated target domain using adversarial feature alignment; and (c) finally training a student network using the pseudo-labels obtained from the teacher. Experimentally, when tested on challenging scenarios involving domain shift, we consistently obtain significantly large performance gains over various recent state of the art approaches."
                    },
                    {
                        "title": "Domain Partitioning Network",
                        "abstract": "Standard adversarial training involves two agents, namely a generator and a discriminator, playing a mini-max game. However, even if the players converge to an equilibrium, the generator may only recover a part of the target data distribution, in a situation commonly referred to as mode collapse. In this work, we present the Domain Partitioning Network (DoPaNet), a new approach to deal with mode collapse in generative adversarial learning. We employ multiple discriminators, each encouraging the generator to cover a different part of the target distribution. To ensure these parts do not overlap and collapse into the same mode, we add a classifier as a third agent in the game. The classifier decides which discriminator the generator is trained against for each sample. Through experiments on toy examples and real images, we show the merits of DoPaNet in covering the real distribution and its superiority with respect to the competing methods. Besides, we also show that we can control the modes from which samples are generated using DoPaNet."
                    },
                    {
                        "title": "Diversified Dynamic Routing for Vision Tasks",
                        "abstract": "Deep learning models for vision tasks are trained on large datasets under the assumption that there exists a universal representation that can be used to make predictions for all samples. Whereas high complexity models are proven to be capable of learning such representations, a mixture of experts trained on specific subsets of the data can infer the labels more efficiently. However using mixture of experts poses two new problems, namely (i) assigning the correct expert at inference time when a new unseen sample is presented. (ii) Finding the optimal partitioning of the training data, such that the experts rely the least on common features. In Dynamic Routing (DR) a novel architecture is proposed where each layer is composed of a set of experts, however without addressing the two challenges we demonstrate that the model reverts to using the same subset of experts.   In our method, Diversified Dynamic Routing (DivDR) the model is explicitly trained to solve the challenge of finding relevant partitioning of the data and assigning the correct experts in an unsupervised approach. We conduct several experiments on semantic segmentation on Cityscapes and object detection and instance segmentation on MS-COCO showing improved performance over several baselines."
                    },
                    {
                        "title": "Near to Mid-term Risks and Opportunities of Open-Source Generative AI",
                        "abstract": "In the next few years, applications of Generative AI are expected to revolutionize a number of different areas, ranging from science & medicine to education. The potential for these seismic changes has triggered a lively debate about potential risks and resulted in calls for tighter regulation, in particular from some of the major tech companies who are leading in AI development. This regulation is likely to put at risk the budding field of open-source Generative AI. We argue for the responsible open sourcing of generative AI models in the near and medium term. To set the stage, we first introduce an AI openness taxonomy system and apply it to 40 current large language models. We then outline differential benefits and risks of open versus closed source AI and present potential risk mitigation, ranging from best practices to calls for technical and scientific contributions. We hope that this report will add a much needed missing voice to the current public discourse on near to mid-term AI safety and other societal impact."
                    }
                ]
            },
            "595ad9b7-3cce-4aaf-80a3-4a77eb903f27": {
                "pk": "595ad9b7-3cce-4aaf-80a3-4a77eb903f27",
                "name": "Wenxuan Zhang",
                "collaborators": [
                    "Wai Lam",
                    "Yang Deng",
                    "Paul Janson",
                    "Mohamed Elhoseiny",
                    "Lidong Bing",
                    "Chang Gao",
                    "Jing Ma",
                    "Rahaf Aljundi",
                    "Xin Li",
                    "Kai Yi"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Question Answering",
                    "Information Extraction",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Answer Ranking for Product-Related Questions via Multiple Semantic Relations Modeling",
                        "abstract": "Many E-commerce sites now offer product-specific question answering platforms for users to communicate with each other by posting and answering questions during online shopping. However, the multiple answers provided by ordinary users usually vary diversely in their qualities and thus need to be appropriately ranked for each question to improve user satisfaction. It can be observed that product reviews usually provide useful information for a given question, and thus can assist the ranking process. In this paper, we investigate the answer ranking problem for product-related questions, with the relevant reviews treated as auxiliary information that can be exploited for facilitating the ranking. We propose an answer ranking model named MUSE which carefully models multiple semantic relations among the question, answers, and relevant reviews. Specifically, MUSE constructs a multi-semantic relation graph with the question, each answer, and each review snippet as nodes. Then a customized graph convolutional neural network is designed for explicitly modeling the semantic relevance between the question and answers, the content consistency among answers, and the textual entailment between answers and reviews. Extensive experiments on real-world E-commerce datasets across three product categories show that our proposed model achieves superior performance on the concerned answer ranking task."
                    },
                    {
                        "title": "Review-guided Helpful Answer Identification in E-commerce",
                        "abstract": "Product-specific community question answering platforms can greatly help address the concerns of potential customers. However, the user-provided answers on such platforms often vary a lot in their qualities. Helpfulness votes from the community can indicate the overall quality of the answer, but they are often missing. Accurately predicting the helpfulness of an answer to a given question and thus identifying helpful answers is becoming a demanding need. Since the helpfulness of an answer depends on multiple perspectives instead of only topical relevance investigated in typical QA tasks, common answer selection algorithms are insufficient for tackling this task. In this paper, we propose the Review-guided Answer Helpfulness Prediction (RAHP) model that not only considers the interactions between QA pairs but also investigates the opinion coherence between the answer and crowds' opinions reflected in the reviews, which is another important factor to identify helpful answers. Moreover, we tackle the task of determining opinion coherence as a language inference problem and explore the utilization of pre-training strategy to transfer the textual inference knowledge obtained from a specifically designed trained network. Extensive experiments conducted on real-world data across seven product categories show that our proposed model achieves superior performance on the prediction task."
                    },
                    {
                        "title": "Overcoming Generic Knowledge Loss with Selective Parameter Update",
                        "abstract": "Foundation models encompass an extensive knowledge base and offer remarkable transferability. However, this knowledge becomes outdated or insufficient over time. The challenge lies in continuously updating foundation models to accommodate novel information while retaining their original capabilities. Leveraging the fact that foundation models have initial knowledge on various tasks and domains, we propose a novel approach that, instead of updating all parameters equally, localizes the updates to a sparse set of parameters relevant to the task being learned. We strike a balance between efficiency and new task performance, while maintaining the transferability and generalizability of foundation models. We extensively evaluate our method on foundational vision-language models with a diverse spectrum of continual learning tasks. Our method achieves improvements on the accuracy of the newly learned tasks up to 7% while preserving the pretraining knowledge with a negligible decrease of 0.9% on a representative control set accuracy."
                    },
                    {
                        "title": "Exploiting BERT for End-to-End Aspect-based Sentiment Analysis",
                        "abstract": "In this paper, we investigate the modeling power of contextualized embeddings from pre-trained language models, e.g. BERT, on the E2E-ABSA task. Specifically, we build a series of simple yet insightful neural baselines to deal with E2E-ABSA. The experimental results show that even with a simple linear classification layer, our BERT-based architecture can outperform state-of-the-art works. Besides, we also standardize the comparative study by consistently utilizing a hold-out validation dataset for model selection, which is largely ignored by previous works. Therefore, our work can serve as a BERT-based benchmark for E2E-ABSA."
                    },
                    {
                        "title": "UniGDD: A Unified Generative Framework for Goal-Oriented Document-Grounded Dialogue",
                        "abstract": "The goal-oriented document-grounded dialogue aims at responding to the user query based on the dialogue context and supporting document. Existing studies tackle this problem by decomposing it into two sub-tasks: knowledge identification and response generation. However, such pipeline methods would unavoidably suffer from the error propagation issue. This paper proposes to unify these two sub-tasks via sequentially generating the grounding knowledge and the response. We further develop a prompt-connected multi-task learning strategy to model the characteristics and connections of different tasks and introduce linear temperature scheduling to reduce the negative effect of irrelevant document information. Experimental results demonstrate the effectiveness of our framework."
                    },
                    {
                        "title": "Multi-hop Inference for Question-driven Summarization",
                        "abstract": "Question-driven summarization has been recently studied as an effective approach to summarizing the source document to produce concise but informative answers for non-factoid questions. In this work, we propose a novel question-driven abstractive summarization method, Multi-hop Selective Generator (MSG), to incorporate multi-hop reasoning into question-driven summarization and, meanwhile, provide justifications for the generated summaries. Specifically, we jointly model the relevance to the question and the interrelation among different sentences via a human-like multi-hop inference module, which captures important sentences for justifying the summarized answer. A gated selective pointer generator network with a multi-view coverage mechanism is designed to integrate diverse information from different perspectives. Experimental results show that the proposed method consistently outperforms state-of-the-art methods on two non-factoid QA datasets, namely WikiHow and PubMedQA."
                    },
                    {
                        "title": "Learning to Rank Question Answer Pairs with Bilateral Contrastive Data Augmentation",
                        "abstract": "In this work, we propose a novel and easy-to-apply data augmentation strategy, namely Bilateral Generation (BiG), with a contrastive training objective for improving the performance of ranking question answer pairs with existing labeled data. In specific, we synthesize pseudo-positive QA pairs in contrast to the original negative QA pairs with two pre-trained generation models, one for question generation, the other for answer generation, which are fine-tuned on the limited positive QA pairs from the original dataset. With the augmented dataset, we design a contrastive training objective for learning to rank question answer pairs. Experimental results on three benchmark datasets show that our method significantly improves the performance of ranking models by making full use of existing labeled data and can be easily applied to different ranking models."
                    },
                    {
                        "title": "Domain-Aware Continual Zero-Shot Learning",
                        "abstract": "Modern visual systems have a wide range of potential applications in vision tasks for natural science research, such as aiding in species discovery, monitoring animals in the wild, and so on. However, real-world vision tasks may experience changes in environmental conditions, leading to shifts in how captured images are presented. To address this issue, we introduce Domain-Aware Continual Zero-Shot Learning (DACZSL), a task to recognize images of unseen categories in continuously changing domains. Accordingly, we propose a Domain-Invariant Network (DIN) to learn factorized features for shifting domains and improved textual representation for unseen classes. DIN continually learns a global shared network for domain-invariant and task-invariant features, and per-task private networks for task-specific features. Furthermore, we enhance the dual network with class-wise learnable prompts to improve class-level text representation, thereby improving zero-shot prediction of future unseen classes. To evaluate DACZSL, we introduce two benchmarks, DomainNet-CZSL and iWildCam-CZSL. Our results show that DIN significantly outperforms existing baselines by over 5% in harmonic accuracy and over 1% in backward transfer and achieves a new SoTA."
                    },
                    {
                        "title": "Product Question Answering in E-Commerce: A Survey",
                        "abstract": "Product question answering (PQA), aiming to automatically provide instant responses to customer's questions in E-Commerce platforms, has drawn increasing attention in recent years. Compared with typical QA problems, PQA exhibits unique challenges such as the subjectivity and reliability of user-generated contents in E-commerce platforms. Therefore, various problem settings and novel methods have been proposed to capture these special characteristics. In this paper, we aim to systematically review existing research efforts on PQA. Specifically, we categorize PQA studies into four problem settings in terms of the form of provided answers. We analyze the pros and cons, as well as present existing datasets and evaluation protocols for each setting. We further summarize the most significant challenges that characterize PQA from general QA applications and discuss their corresponding solutions. Finally, we conclude this paper by providing the prospect on several future directions."
                    },
                    {
                        "title": "Easy-to-Hard Learning for Information Extraction",
                        "abstract": "Information extraction (IE) systems aim to automatically extract structured information, such as named entities, relations between entities, and events, from unstructured texts. While most existing work addresses a particular IE task, universally modeling various IE tasks with one model has achieved great success recently. Despite their success, they employ a one-stage learning strategy, i.e., directly learning to extract the target structure given the input text, which contradicts the human learning process. In this paper, we propose a unified easy-to-hard learning framework consisting of three stages, i.e., the easy stage, the hard stage, and the main stage, for IE by mimicking the human learning process. By breaking down the learning process into multiple stages, our framework facilitates the model to acquire general IE task knowledge and improve its generalization ability. Extensive experiments across four IE tasks demonstrate the effectiveness of our framework. We achieve new state-of-the-art results on 13 out of 17 datasets. Our code is available at \\url{https://github.com/DAMO-NLP-SG/IE-E2H}."
                    }
                ]
            },
            "c3a49248-f4ae-4ad9-b8d2-e674c529cae1": {
                "pk": "c3a49248-f4ae-4ad9-b8d2-e674c529cae1",
                "name": "Matthias M\u00fcller",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "e8a0c5fb-e7c1-4aea-b96b-f39364d11835": {
                "pk": "e8a0c5fb-e7c1-4aea-b96b-f39364d11835",
                "name": "Ser-Nam Lim",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "1ea920dd-a2ed-42be-91e3-c58e80b8a947": {
                "pk": "1ea920dd-a2ed-42be-91e3-c58e80b8a947",
                "name": "Mohamed Elhoseiny",
                "collaborators": [
                    "Ahmed Elgammal",
                    "Mohamed Elfeki",
                    "Abduallah Mohamed",
                    "Babak Saleh",
                    "Eslam Mohamed Bakr",
                    "Scott Cohen",
                    "Walter Chang",
                    "Brian Price",
                    "Christian Claudel",
                    "Xiaoqian Shen"
                ],
                "domain": [
                    "Zero-shot Learning",
                    "Generative Models",
                    "Machine Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Creativity Inspired Zero-Shot Learning",
                        "abstract": "Zero-shot learning (ZSL) aims at understanding unseen categories with no training examples from class-level descriptions. To improve the discriminative power of zero-shot learning, we model the visual learning process of unseen categories with inspiration from the psychology of human creativity for producing novel art. We relate ZSL to human creativity by observing that zero-shot learning is about recognizing the unseen and creativity is about creating a likable unseen. We introduce a learning signal inspired by creativity literature that explores the unseen space with hallucinated class-descriptions and encourages careful deviation of their visual feature generations from seen classes while allowing knowledge transfer from seen to unseen classes. Empirically, we show consistent improvement over the state of the art of several percents on the largest available benchmarks on the challenging task or generalized ZSL from a noisy text that we focus on, using the CUB and NABirds datasets. We also show the advantage of our approach on Attribute-based ZSL on three additional datasets (AwA2, aPY, and SUN). Code is available."
                    },
                    {
                        "title": "Overlapping Cover Local Regression Machines",
                        "abstract": "We present the Overlapping Domain Cover (ODC) notion for kernel machines, as a set of overlapping subsets of the data that covers the entire training set and optimized to be spatially cohesive as possible. We show how this notion benefit the speed of local kernel machines for regression in terms of both speed while achieving while minimizing the prediction error. We propose an efficient ODC framework, which is applicable to various regression models and in particular reduces the complexity of Twin Gaussian Processes (TGP) regression from cubic to quadratic. Our notion is also applicable to several kernel methods (e.g., Gaussian Process Regression(GPR) and IWTGP regression, as shown in our experiments). We also theoretically justified the idea behind our method to improve local prediction by the overlapping cover. We validated and analyzed our method on three benchmark human pose estimation datasets and interesting findings are discussed."
                    },
                    {
                        "title": "Text to Multi-level MindMaps: A Novel Method for Hierarchical Visual Abstraction of Natural Language Text",
                        "abstract": "MindMapping is a well-known technique used in note taking, which encourages learning and studying. MindMapping has been manually adopted to help present knowledge and concepts in a visual form. Unfortunately, there is no reliable automated approach to generate MindMaps from Natural Language text. This work firstly introduces MindMap Multilevel Visualization concept which is to jointly visualize and summarize textual information. The visualization is achieved pictorially across multiple levels using semantic information (i.e. ontology), while the summarization is achieved by the information in the highest levels as they represent abstract information in the text. This work also presents the first automated approach that takes a text input and generates a MindMap visualization out of it. The approach could visualize text documents in multilevel MindMaps, in which a high-level MindMap node could be expanded into child MindMaps. \\ignore{ As far as we know, this is the first work that view MindMapping as a new approach to jointly summarize and visualize textual information.} The proposed method involves understanding of the input text and converting it into intermediate Detailed Meaning Representation (DMR). The DMR is then visualized with two modes; Single level or Multiple levels, which is convenient for larger text. The generated MindMaps from both approaches were evaluated based on Human Subject experiments performed on Amazon Mechanical Turk with various parameter settings."
                    },
                    {
                        "title": "Generalized Twin Gaussian Processes using Sharma-Mittal Divergence",
                        "abstract": "There has been a growing interest in mutual information measures due to their wide range of applications in Machine Learning and Computer Vision. In this paper, we present a generalized structured regression framework based on Shama-Mittal divergence, a relative entropy measure, which is introduced to the Machine Learning community in this work. Sharma-Mittal (SM) divergence is a generalized mutual information measure for the widely used R\\'enyi, Tsallis, Bhattacharyya, and Kullback-Leibler (KL) relative entropies. Specifically, we study Sharma-Mittal divergence as a cost function in the context of the Twin Gaussian Processes (TGP)~\\citep{Bo:2010}, which generalizes over the KL-divergence without computational penalty. We show interesting properties of Sharma-Mittal TGP (SMTGP) through a theoretical analysis, which covers missing insights in the traditional TGP formulation. However, we generalize this theory based on SM-divergence instead of KL-divergence which is a special case. Experimentally, we evaluated the proposed SMTGP framework on several datasets. The results show that SMTGP reaches better predictions than KL-based TGP, since it offers a bigger class of models through its parameters that we learn from the data."
                    },
                    {
                        "title": "Tell and Predict: Kernel Classifier Prediction for Unseen Visual Classes from Unstructured Text Descriptions",
                        "abstract": "In this paper we propose a framework for predicting kernelized classifiers in the visual domain for categories with no training images where the knowledge comes from textual description about these categories. Through our optimization framework, the proposed approach is capable of embedding the class-level knowledge from the text domain as kernel classifiers in the visual domain. We also proposed a distributional semantic kernel between text descriptions which is shown to be effective in our setting. The proposed framework is not restricted to textual descriptions, and can also be applied to other forms knowledge representations. Our approach was applied for the challenging task of zero-shot learning of fine-grained categories from text descriptions of these categories."
                    },
                    {
                        "title": "Look Around and Refer: 2D Synthetic Semantics Knowledge Distillation for 3D Visual Grounding",
                        "abstract": "The 3D visual grounding task has been explored with visual and language streams comprehending referential language to identify target objects in 3D scenes. However, most existing methods devote the visual stream to capturing the 3D visual clues using off-the-shelf point clouds encoders. The main question we address in this paper is \"can we consolidate the 3D visual stream by 2D clues synthesized from point clouds and efficiently utilize them in training and testing?\". The main idea is to assist the 3D encoder by incorporating rich 2D object representations without requiring extra 2D inputs. To this end, we leverage 2D clues, synthetically generated from 3D point clouds, and empirically show their aptitude to boost the quality of the learned visual representations. We validate our approach through comprehensive experiments on Nr3D, Sr3D, and ScanRefer datasets and show consistent performance gains compared to existing methods. Our proposed module, dubbed as Look Around and Refer (LAR), significantly outperforms the state-of-the-art 3D visual grounding techniques on three benchmarks, i.e., Nr3D, Sr3D, and ScanRefer. The code is available at https://eslambakr.github.io/LAR.github.io/."
                    },
                    {
                        "title": "Automatic Annotation of Structured Facts in Images",
                        "abstract": "Motivated by the application of fact-level image understanding, we present an automatic method for data collection of structured visual facts from images with captions. Example structured facts include attributed objects (e.g., <flower, red>), actions (e.g., <baby, smile>), interactions (e.g., <man, walking, dog>), and positional information (e.g., <vase, on, table>). The collected annotations are in the form of fact-image pairs (e.g.,<man, walking, dog> and an image region containing this fact). With a language approach, the proposed method is able to collect hundreds of thousands of visual fact annotations with accuracy of 83% according to human judgment. Our method automatically collected more than 380,000 visual fact annotations and more than 110,000 unique visual facts from images with captions and localized them in images in less than one day of processing time on standard CPU platforms."
                    },
                    {
                        "title": "Social-STGCNN: A Social Spatio-Temporal Graph Convolutional Neural Network for Human Trajectory Prediction",
                        "abstract": "Better machine understanding of pedestrian behaviors enables faster progress in modeling interactions between agents such as autonomous vehicles and humans. Pedestrian trajectories are not only influenced by the pedestrian itself but also by interaction with surrounding objects. Previous methods modeled these interactions by using a variety of aggregation methods that integrate different learned pedestrians states. We propose the Social Spatio-Temporal Graph Convolutional Neural Network (Social-STGCNN), which substitutes the need of aggregation methods by modeling the interactions as a graph. Our results show an improvement over the state of art by 20% on the Final Displacement Error (FDE) and an improvement on the Average Displacement Error (ADE) with 8.5 times less parameters and up to 48 times faster inference speed than previously reported methods. In addition, our model is data efficient, and exceeds previous state of the art on the ADE metric with only 20% of the training data. We propose a kernel function to embed the social interactions between pedestrians within the adjacency matrix. Through qualitative analysis, we show that our model inherited social behaviors that can be expected between pedestrians trajectories. Code is available at https://github.com/abduallahmohamed/Social-STGCNN."
                    },
                    {
                        "title": "CIZSL++: Creativity Inspired Generative Zero-Shot Learning",
                        "abstract": "Zero-shot learning (ZSL) aims at understanding unseen categories with no training examples from class-level descriptions. To improve the discriminative power of ZSL, we model the visual learning process of unseen categories with inspiration from the psychology of human creativity for producing novel art. First, we propose CIZSL-v1 as a creativity inspired model for generative ZSL. We relate ZSL to human creativity by observing that ZSL is about recognizing the unseen, and creativity is about creating a likable unseen. We introduce a learning signal inspired by creativity literature that explores the unseen space with hallucinated class-descriptions and encourages careful deviation of their visual feature generations from seen classes while allowing knowledge transfer from seen to unseen classes. Second, CIZSL-v2 is proposed as an improved version of CIZSL-v1 for generative zero-shot learning. CIZSL-v2 consists of an investigation of additional inductive losses for unseen classes along with a semantic guided discriminator. Empirically, we show consistently that CIZSL losses can improve generative ZSL models on the challenging task of generalized ZSL from a noisy text on CUB and NABirds datasets. We also show the advantage of our approach to Attribute-based ZSL on AwA2, aPY, and SUN datasets. We also show that CIZSL-v2 has improved performance compared to CIZSL-v1."
                    },
                    {
                        "title": "Convolutional Models for Joint Object Categorization and Pose Estimation",
                        "abstract": "In the task of Object Recognition, there exists a dichotomy between the categorization of objects and estimating object pose, where the former necessitates a view-invariant representation, while the latter requires a representation capable of capturing pose information over different categories of objects. With the rise of deep architectures, the prime focus has been on object category recognition. Deep learning methods have achieved wide success in this task. In contrast, object pose regression using these approaches has received relatively much less attention. In this paper we show how deep architectures, specifically Convolutional Neural Networks (CNN), can be adapted to the task of simultaneous categorization and pose estimation of objects. We investigate and analyze the layers of various CNN models and extensively compare between them with the goal of discovering how the layers of distributed representations of CNNs represent object pose information and how this contradicts with object category representations. We extensively experiment on two recent large and challenging multi-view datasets. Our models achieve better than state-of-the-art performance on both datasets."
                    },
                    {
                        "title": "Write a Classifier: Predicting Visual Classifiers from Unstructured Text",
                        "abstract": "People typically learn through exposure to visual concepts associated with linguistic descriptions. For instance, teaching visual object categories to children is often accompanied by descriptions in text or speech. In a machine learning context, these observations motivates us to ask whether this learning process could be computationally modeled to learn visual classifiers. More specifically, the main question of this work is how to utilize purely textual description of visual classes with no training images, to learn explicit visual classifiers for them. We propose and investigate two baseline formulations, based on regression and domain transfer, that predict a linear classifier. Then, we propose a new constrained optimization formulation that combines a regression function and a knowledge transfer function with additional constraints to predict the parameters of a linear classifier. We also propose a generic kernelized models where a kernel classifier is predicted in the form defined by the representer theorem. The kernelized models allow defining and utilizing any two RKHS (Reproducing Kernel Hilbert Space) kernel functions in the visual space and text space, respectively. We finally propose a kernel function between unstructured text descriptions that builds on distributional semantics, which shows an advantage in our setting and could be useful for other applications. We applied all the studied models to predict visual classifiers on two fine-grained and challenging categorization datasets (CU Birds and Flower Datasets), and the results indicate successful predictions of our final model over several baselines that we designed."
                    },
                    {
                        "title": "iMotion-LLM: Motion Prediction Instruction Tuning",
                        "abstract": "We introduce iMotion-LLM: a Multimodal Large Language Models (LLMs) with trajectory prediction, tailored to guide interactive multi-agent scenarios. Different from conventional motion prediction approaches, iMotion-LLM capitalizes on textual instructions as key inputs for generating contextually relevant trajectories. By enriching the real-world driving scenarios in the Waymo Open Dataset with textual motion instructions, we created InstructWaymo. Leveraging this dataset, iMotion-LLM integrates a pretrained LLM, fine-tuned with LoRA, to translate scene features into the LLM input space. iMotion-LLM offers significant advantages over conventional motion prediction models. First, it can generate trajectories that align with the provided instructions if it is a feasible direction. Second, when given an infeasible direction, it can reject the instruction, thereby enhancing safety. These findings act as milestones in empowering autonomous navigation systems to interpret and predict the dynamics of multi-agent environments, laying the groundwork for future advancements in this field."
                    },
                    {
                        "title": "Sherlock: Scalable Fact Learning in Images",
                        "abstract": "We study scalable and uniform understanding of facts in images. Existing visual recognition systems are typically modeled differently for each fact type such as objects, actions, and interactions. We propose a setting where all these facts can be modeled simultaneously with a capacity to understand unbounded number of facts in a structured way. The training data comes as structured facts in images, including (1) objects (e.g., $<$boy$>$), (2) attributes (e.g., $<$boy, tall$>$), (3) actions (e.g., $<$boy, playing$>$), and (4) interactions (e.g., $<$boy, riding, a horse $>$). Each fact has a semantic language view (e.g., $<$ boy, playing$>$) and a visual view (an image with this fact). We show that learning visual facts in a structured way enables not only a uniform but also generalizable visual understanding. We propose and investigate recent and strong approaches from the multiview learning literature and also introduce two learning representation models as potential baselines. We applied the investigated methods on several datasets that we augmented with structured facts and a large scale dataset of more than 202,000 facts and 814,000 images. Our experiments show the advantage of relating facts by the structure by the proposed models compared to the designed baselines on bidirectional fact retrieval."
                    },
                    {
                        "title": "Zero-Shot Event Detection by Multimodal Distributional Semantic Embedding of Videos",
                        "abstract": "We propose a new zero-shot Event Detection method by Multi-modal Distributional Semantic embedding of videos. Our model embeds object and action concepts as well as other available modalities from videos into a distributional semantic space. To our knowledge, this is the first Zero-Shot event detection model that is built on top of distributional semantics and extends it in the following directions: (a) semantic embedding of multimodal information in videos (with focus on the visual modalities), (b) automatically determining relevance of concepts/attributes to a free text query, which could be useful for other applications, and (c) retrieving videos by free text event query (e.g., \"changing a vehicle tire\") based on their content. We embed videos into a distributional semantic space and then measure the similarity between videos and the event query in a free text form. We validated our method on the large TRECVID MED (Multimedia Event Detection) challenge. Using only the event title as a query, our method outperformed the state-of-the-art that uses big descriptions from 12.6% to 13.5% with MAP metric and 0.73 to 0.83 with ROC-AUC metric. It is also an order of magnitude faster."
                    },
                    {
                        "title": "GPU-Framework for Teamwork Action Recognition",
                        "abstract": "Real time processing for teamwork action recognition is a challenge, due to complex computational models to achieve high system performance. Hence, this paper proposes a framework based on Graphical Processing Units (GPUs) to achieve a significant speed up in the performance of role based activity recognition of teamwork. The framework can be applied in various fields, especially athletic and military applications. Furthermore, the framework can be customized for many action recognition applications. The paper presents the stages of the framework where GPUs are the main tool for performance improvement. The speedup is achieved by performing video processing and Machine learning algorithms on GPU. Video processing and machine learning algorithms covers all computations involved in our framework. Video processing tasks on involves GPU implementation of Motion detection, segmentation and object tracking algorithms. In addition, our framework is integrated with GPUCV, a GPU version of OpenCV functions. Machine learning tasks are supported under our framework with GPU implementations of Support Vector Machine (SVM) for object classification and feature discretization, Hidden Marcov Model (HMM) for activity recognition phase, and ID3 algorithm for role recognition of team members. The system was tested against UC-Teamwork dataset and speedup of 20X has been achieved on NVidia 9500GT graphics card (32 500MHZ processors)."
                    },
                    {
                        "title": "Class Normalization for (Continual)? Generalized Zero-Shot Learning",
                        "abstract": "Normalization techniques have proved to be a crucial ingredient of successful training in a traditional supervised learning regime. However, in the zero-shot learning (ZSL) world, these ideas have received only marginal attention. This work studies normalization in ZSL scenario from both theoretical and practical perspectives. First, we give a theoretical explanation to two popular tricks used in zero-shot learning: normalize+scale and attributes normalization and show that they help training by preserving variance during a forward pass. Next, we demonstrate that they are insufficient to normalize a deep ZSL model and propose Class Normalization (CN): a normalization scheme, which alleviates this issue both provably and in practice. Third, we show that ZSL models typically have more irregular loss surface compared to traditional classifiers and that the proposed method partially remedies this problem. Then, we test our approach on 4 standard ZSL datasets and outperform sophisticated modern SotA with a simple MLP optimized without any bells and whistles and having ~50 times faster training speed. Finally, we generalize ZSL to a broader problem -- continual ZSL, and introduce some principled metrics and rigorous baselines for this new setup. The project page is located at https://universome.github.io/class-norm."
                    },
                    {
                        "title": "StoryGPT-V: Large Language Models as Consistent Story Visualizers",
                        "abstract": "Recent generative models have demonstrated impressive capabilities in generating realistic and visually pleasing images grounded on textual prompts. Nevertheless, a significant challenge remains in applying these models for the more intricate task of story visualization. Since it requires resolving pronouns (he, she, they) in the frame descriptions, i.e., anaphora resolution, and ensuring consistent characters and background synthesis across frames. Yet, the emerging Large Language Model (LLM) showcases robust reasoning abilities to navigate through ambiguous references and process extensive sequences. Therefore, we introduce \\textbf{StoryGPT-V}, which leverages the merits of the latent diffusion (LDM) and LLM to produce images with consistent and high-quality characters grounded on given story descriptions. First, we train a character-aware LDM, which takes character-augmented semantic embedding as input and includes the supervision of the cross-attention map using character segmentation masks, aiming to enhance character generation accuracy and faithfulness. In the second stage, we enable an alignment between the output of LLM and the character-augmented embedding residing in the input space of the first-stage model. This harnesses the reasoning ability of LLM to address ambiguous references and the comprehension capability to memorize the context. We conduct comprehensive experiments on two visual story visualization benchmarks. Our model reports superior quantitative results and consistently generates accurate characters of remarkable quality with low memory consumption. Our code will be made publicly available."
                    },
                    {
                        "title": "GDPP: Learning Diverse Generations Using Determinantal Point Process",
                        "abstract": "Generative models have proven to be an outstanding tool for representing high-dimensional probability distributions and generating realistic-looking images. An essential characteristic of generative models is their ability to produce multi-modal outputs. However, while training, they are often susceptible to mode collapse, that is models are limited in mapping input noise to only a few modes of the true data distribution. In this work, we draw inspiration from Determinantal Point Process (DPP) to propose an unsupervised penalty loss that alleviates mode collapse while producing higher quality samples. DPP is an elegant probabilistic measure used to model negative correlations within a subset and hence quantify its diversity. We use DPP kernel to model the diversity in real data as well as in synthetic data. Then, we devise an objective term that encourages generators to synthesize data with similar diversity to real data. In contrast to previous state-of-the-art generative models that tend to use additional trainable parameters or complex training paradigms, our method does not change the original training scheme. Embedded in an adversarial training and variational autoencoder, our Generative DPP approach shows a consistent resistance to mode-collapse on a wide variety of synthetic data and natural image datasets including MNIST, CIFAR10, and CelebA, while outperforming state-of-the-art methods for data-efficiency, generation quality, and convergence-time whereas being 5.8x faster than its closest competitor."
                    },
                    {
                        "title": "Social-Implicit: Rethinking Trajectory Prediction Evaluation and The Effectiveness of Implicit Maximum Likelihood Estimation",
                        "abstract": "Best-of-N (BoN) Average Displacement Error (ADE)/ Final Displacement Error (FDE) is the most used metric for evaluating trajectory prediction models. Yet, the BoN does not quantify the whole generated samples, resulting in an incomplete view of the model's prediction quality and performance. We propose a new metric, Average Mahalanobis Distance (AMD) to tackle this issue. AMD is a metric that quantifies how close the whole generated samples are to the ground truth. We also introduce the Average Maximum Eigenvalue (AMV) metric that quantifies the overall spread of the predictions. Our metrics are validated empirically by showing that the ADE/FDE is not sensitive to distribution shifts, giving a biased sense of accuracy, unlike the AMD/AMV metrics. We introduce the usage of Implicit Maximum Likelihood Estimation (IMLE) as a replacement for traditional generative models to train our model, Social-Implicit. IMLE training mechanism aligns with AMD/AMV objective of predicting trajectories that are close to the ground truth with a tight spread. Social-Implicit is a memory efficient deep model with only 5.8K parameters that runs in real time of about 580Hz and achieves competitive results. Interactive demo of the problem can be seen at https://www.abduallahmohamed.com/social-implicit-amdamv-adefde-demo . Code is available at https://github.com/abduallahmohamed/Social-Implicit ."
                    },
                    {
                        "title": "It is Okay to Not Be Okay: Overcoming Emotional Bias in Affective Image Captioning by Contrastive Data Collection",
                        "abstract": "Datasets that capture the connection between vision, language, and affection are limited, causing a lack of understanding of the emotional aspect of human intelligence. As a step in this direction, the ArtEmis dataset was recently introduced as a large-scale dataset of emotional reactions to images along with language explanations of these chosen emotions. We observed a significant emotional bias towards instance-rich emotions, making trained neural speakers less accurate in describing under-represented emotions. We show that collecting new data, in the same way, is not effective in mitigating this emotional bias. To remedy this problem, we propose a contrastive data collection approach to balance ArtEmis with a new complementary dataset such that a pair of similar images have contrasting emotions (one positive and one negative). We collected 260,533 instances using the proposed method, we combine them with ArtEmis, creating a second iteration of the dataset. The new combined dataset, dubbed ArtEmis v2.0, has a balanced distribution of emotions with explanations revealing more fine details in the associated painting. Our experiments show that neural speakers trained on the new dataset improve CIDEr and METEOR evaluation metrics by 20% and 7%, respectively, compared to the biased dataset. Finally, we also show that the performance per emotion of neural speakers is improved across all the emotion categories, significantly on under-represented emotions. The collected dataset and code are available at https://artemisdataset-v2.org."
                    }
                ]
            },
            "19a0f7a9-318a-4f3e-9d8f-29951eb6adb6": {
                "pk": "19a0f7a9-318a-4f3e-9d8f-29951eb6adb6",
                "name": "Philip Torr",
                "collaborators": [
                    "Shuyang Sun",
                    "Jindong Gu",
                    "Fazl Barez",
                    "Anurag Arnab",
                    "Sadeep Jayasumana",
                    "Shuai Zheng",
                    "Zhengzhe Liu",
                    "Xiaojuan Qi",
                    "Liang Chen",
                    "Gregory Slabaugh"
                ],
                "domain": [
                    "Artificial Intelligence",
                    "Deep Learning",
                    "Generative Models",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Measuring Value Alignment",
                        "abstract": "As artificial intelligence (AI) systems become increasingly integrated into various domains, ensuring that they align with human values becomes critical. This paper introduces a novel formalism to quantify the alignment between AI systems and human values, using Markov Decision Processes (MDPs) as the foundational model. We delve into the concept of values as desirable goals tied to actions and norms as behavioral guidelines, aiming to shed light on how they can be used to guide AI decisions. This framework offers a mechanism to evaluate the degree of alignment between norms and values by assessing preference changes across state transitions in a normative world. By utilizing this formalism, AI developers and ethicists can better design and evaluate AI systems to ensure they operate in harmony with human values. The proposed methodology holds potential for a wide range of applications, from recommendation systems emphasizing well-being to autonomous vehicles prioritizing safety."
                    },
                    {
                        "title": "Higher Order Conditional Random Fields in Deep Neural Networks",
                        "abstract": "We address the problem of semantic segmentation using deep learning. Most segmentation systems include a Conditional Random Field (CRF) to produce a structured output that is consistent with the image's visual features. Recent deep learning approaches have incorporated CRFs into Convolutional Neural Networks (CNNs), with some even training the CRF end-to-end with the rest of the network. However, these approaches have not employed higher order potentials, which have previously been shown to significantly improve segmentation performance. In this paper, we demonstrate that two types of higher order potential, based on object detections and superpixels, can be included in a CRF embedded within a deep network. We design these higher order potentials to allow inference with the differentiable mean field algorithm. As a result, all the parameters of our richer CRF model can be learned end-to-end with our pixelwise CNN classifier. We achieve state-of-the-art segmentation performance on the PASCAL VOC benchmark with these trainable higher order potentials."
                    },
                    {
                        "title": "Global Texture Enhancement for Fake Face Detection in the Wild",
                        "abstract": "Generative Adversarial Networks (GANs) can generate realistic fake face images that can easily fool human beings.On the contrary, a common Convolutional Neural Network(CNN) discriminator can achieve more than 99.9% accuracyin discerning fake/real images. In this paper, we conduct an empirical study on fake/real faces, and have two important observations: firstly, the texture of fake faces is substantially different from real ones; secondly, global texture statistics are more robust to image editing and transferable to fake faces from different GANs and datasets. Motivated by the above observations, we propose a new architecture coined as Gram-Net, which leverages global image texture representations for robust fake image detection. Experimental results on several datasets demonstrate that our Gram-Net outperforms existing approaches. Especially, our Gram-Netis more robust to image editings, e.g. down-sampling, JPEG compression, blur, and noise. More importantly, our Gram-Net generalizes significantly better in detecting fake faces from GAN models not seen in the training phase and can perform decently in detecting fake natural images."
                    },
                    {
                        "title": "Learning to Sample the Most Useful Training Patches from Images",
                        "abstract": "Some image restoration tasks like demosaicing require difficult training samples to learn effective models. Existing methods attempt to address this data training problem by manually collecting a new training dataset that contains adequate hard samples, however, there are still hard and simple areas even within one single image. In this paper, we present a data-driven approach called PatchNet that learns to select the most useful patches from an image to construct a new training set instead of manual or random selection. We show that our simple idea automatically selects informative samples out from a large-scale dataset, leading to a surprising 2.35dB generalisation gain in terms of PSNR. In addition to its remarkable effectiveness, PatchNet is also resource-friendly as it is applied only during training and therefore does not require any additional computational cost during inference."
                    },
                    {
                        "title": "General Adversarial Defense Against Black-box Attacks via Pixel Level and Feature Level Distribution Alignments",
                        "abstract": "Deep Neural Networks (DNNs) are vulnerable to the black-box adversarial attack that is highly transferable. This threat comes from the distribution gap between adversarial and clean samples in feature space of the target DNNs. In this paper, we use Deep Generative Networks (DGNs) with a novel training mechanism to eliminate the distribution gap. The trained DGNs align the distribution of adversarial samples with clean ones for the target DNNs by translating pixel values. Different from previous work, we propose a more effective pixel level training constraint to make this achievable, thus enhancing robustness on adversarial samples. Further, a class-aware feature-level constraint is formulated for integrated distribution alignment. Our approach is general and applicable to multiple tasks, including image classification, semantic segmentation, and object detection. We conduct extensive experiments on different datasets. Our strategy demonstrates its unique effectiveness and generality against black-box attacks."
                    },
                    {
                        "title": "Not Just Pretty Pictures: Toward Interventional Data Augmentation Using Text-to-Image Generators",
                        "abstract": "Neural image classifiers are known to undergo severe performance degradation when exposed to inputs that are sampled from environmental conditions that differ from their training data. Given the recent progress in Text-to-Image (T2I) generation, a natural question is how modern T2I generators can be used to simulate arbitrary interventions over such environmental factors in order to augment training data and improve the robustness of downstream classifiers. We experiment across a diverse collection of benchmarks in single domain generalization (SDG) and reducing reliance on spurious features (RRSF), ablating across key dimensions of T2I generation, including interventional prompting strategies, conditioning mechanisms, and post-hoc filtering. Our extensive empirical findings demonstrate that modern T2I generators like Stable Diffusion can indeed be used as a powerful interventional data augmentation mechanism, outperforming previously state-of-the-art data augmentation techniques regardless of how each dimension is configured."
                    },
                    {
                        "title": "OxfordTVG-HIC: Can Machine Make Humorous Captions from Images?",
                        "abstract": "This paper presents OxfordTVG-HIC (Humorous Image Captions), a large-scale dataset for humour generation and understanding. Humour is an abstract, subjective, and context-dependent cognitive construct involving several cognitive factors, making it a challenging task to generate and interpret. Hence, humour generation and understanding can serve as a new task for evaluating the ability of deep-learning methods to process abstract and subjective information. Due to the scarcity of data, humour-related generation tasks such as captioning remain under-explored. To address this gap, OxfordTVG-HIC offers approximately 2.9M image-text pairs with humour scores to train a generalizable humour captioning model. Contrary to existing captioning datasets, OxfordTVG-HIC features a wide range of emotional and semantic diversity resulting in out-of-context examples that are particularly conducive to generating humour. Moreover, OxfordTVG-HIC is curated devoid of offensive content. We also show how OxfordTVG-HIC can be leveraged for evaluating the humour of a generated text. Through explainability analysis of the trained models, we identify the visual and linguistic cues influential for evoking humour prediction (and generation). We observe qualitatively that these cues are aligned with the benign violation theory of humour in cognitive psychology."
                    },
                    {
                        "title": "Exploring Non-additive Randomness on ViT against Query-Based Black-Box Attacks",
                        "abstract": "Deep Neural Networks can be easily fooled by small and imperceptible perturbations. The query-based black-box attack (QBBA) is able to create the perturbations using model output probabilities of image queries requiring no access to the underlying models. QBBA poses realistic threats to real-world applications. Recently, various types of robustness have been explored to defend against QBBA. In this work, we first taxonomize the stochastic defense strategies against QBBA. Following our taxonomy, we propose to explore non-additive randomness in models to defend against QBBA. Specifically, we focus on underexplored Vision Transformers based on their flexible architectures. Extensive experiments show that the proposed defense approach achieves effective defense, without much sacrifice in performance."
                    },
                    {
                        "title": "Scene-Conditional 3D Object Stylization and Composition",
                        "abstract": "Recently, 3D generative models have made impressive progress, enabling the generation of almost arbitrary 3D assets from text or image inputs. However, these approaches generate objects in isolation without any consideration for the scene where they will eventually be placed. In this paper, we propose a framework that allows for the stylization of an existing 3D asset to fit into a given 2D scene, and additionally produce a photorealistic composition as if the asset was placed within the environment. This not only opens up a new level of control for object stylization, for example, the same assets can be stylized to reflect changes in the environment, such as summer to winter or fantasy versus futuristic settings-but also makes the object-scene composition more controllable. We achieve this by combining modeling and optimizing the object's texture and environmental lighting through differentiable ray tracing with image priors from pre-trained text-to-image diffusion models. We demonstrate that our method is applicable to a wide variety of indoor and outdoor scenes and arbitrary objects."
                    },
                    {
                        "title": "An Image Is Worth 1000 Lies: Adversarial Transferability across Prompts on Vision-Language Models",
                        "abstract": "Different from traditional task-specific vision models, recent large VLMs can readily adapt to different vision tasks by simply using different textual instructions, i.e., prompts. However, a well-known concern about traditional task-specific vision models is that they can be misled by imperceptible adversarial perturbations. Furthermore, the concern is exacerbated by the phenomenon that the same adversarial perturbations can fool different task-specific models. Given that VLMs rely on prompts to adapt to different tasks, an intriguing question emerges: Can a single adversarial image mislead all predictions of VLMs when a thousand different prompts are given? This question essentially introduces a novel perspective on adversarial transferability: cross-prompt adversarial transferability. In this work, we propose the Cross-Prompt Attack (CroPA). This proposed method updates the visual adversarial perturbation with learnable prompts, which are designed to counteract the misleading effects of the adversarial image. By doing this, CroPA significantly improves the transferability of adversarial examples across prompts. Extensive experiments are conducted to verify the strong cross-prompt adversarial transferability of CroPA with prevalent VLMs including Flamingo, BLIP-2, and InstructBLIP in various different tasks. Our source code is available at \\url{https://github.com/Haochen-Luo/CroPA}."
                    },
                    {
                        "title": "VFusion3D: Learning Scalable 3D Generative Models from Video Diffusion Models",
                        "abstract": "This paper presents a novel method for building scalable 3D generative models utilizing pre-trained video diffusion models. The primary obstacle in developing foundation 3D generative models is the limited availability of 3D data. Unlike images, texts, or videos, 3D data are not readily accessible and are difficult to acquire. This results in a significant disparity in scale compared to the vast quantities of other types of data. To address this issue, we propose using a video diffusion model, trained with extensive volumes of text, images, and videos, as a knowledge source for 3D data. By unlocking its multi-view generative capabilities through fine-tuning, we generate a large-scale synthetic multi-view dataset to train a feed-forward 3D generative model. The proposed model, VFusion3D, trained on nearly 3M synthetic multi-view data, can generate a 3D asset from a single image in seconds and achieves superior performance when compared to current SOTA feed-forward 3D generative models, with users preferring our results over 90% of the time."
                    },
                    {
                        "title": "DreamPolisher: Towards High-Quality Text-to-3D Generation via Geometric Diffusion",
                        "abstract": "We present DreamPolisher, a novel Gaussian Splatting based method with geometric guidance, tailored to learn cross-view consistency and intricate detail from textual descriptions. While recent progress on text-to-3D generation methods have been promising, prevailing methods often fail to ensure view-consistency and textural richness. This problem becomes particularly noticeable for methods that work with text input alone. To address this, we propose a two-stage Gaussian Splatting based approach that enforces geometric consistency among views. Initially, a coarse 3D generation undergoes refinement via geometric optimization. Subsequently, we use a ControlNet driven refiner coupled with the geometric consistency term to improve both texture fidelity and overall consistency of the generated 3D asset. Empirical evaluations across diverse textual prompts spanning various object categories demonstrate the efficacy of DreamPolisher in generating consistent and realistic 3D objects, aligning closely with the semantics of the textual instructions."
                    },
                    {
                        "title": "An MRP Formulation for Supervised Learning: Generalized Temporal Difference Learning Models",
                        "abstract": "In traditional statistical learning, data points are usually assumed to be independently and identically distributed (i.i.d.) following an unknown probability distribution. This paper presents a contrasting viewpoint, perceiving data points as interconnected and employing a Markov reward process (MRP) for data modeling. We reformulate the typical supervised learning as an on-policy policy evaluation problem within reinforcement learning (RL), introducing a generalized temporal difference (TD) learning algorithm as a resolution. Theoretically, our analysis draws connections between the solutions of linear TD learning and ordinary least squares (OLS). We also show that under specific conditions, particularly when noises are correlated, the TD's solution proves to be a more effective estimator than OLS. Furthermore, we establish the convergence of our generalized TD algorithms under linear function approximation. Empirical studies verify our theoretical results, examine the vital design of our TD algorithm and show practical utility across various datasets, encompassing tasks such as regression and image classification with deep learning."
                    },
                    {
                        "title": "Specify and Edit: Overcoming Ambiguity in Text-Based Image Editing",
                        "abstract": "Text-based editing diffusion models exhibit limited performance when the user's input instruction is ambiguous. To solve this problem, we propose $\\textit{Specify ANd Edit}$ (SANE), a zero-shot inference pipeline for diffusion-based editing systems. We use a large language model (LLM) to decompose the input instruction into specific instructions, i.e. well-defined interventions to apply to the input image to satisfy the user's request. We benefit from the LLM-derived instructions along the original one, thanks to a novel denoising guidance strategy specifically designed for the task. Our experiments with three baselines and on two datasets demonstrate the benefits of SANE in all setups. Moreover, our pipeline improves the interpretability of editing models, and boosts the output diversity. We also demonstrate that our approach can be applied to any edit, whether ambiguous or not. Our code is public at https://github.com/fabvio/SANE."
                    }
                ]
            },
            "ea866e2f-25b6-447d-a3f6-e6f0d2d70456": {
                "pk": "ea866e2f-25b6-447d-a3f6-e6f0d2d70456",
                "name": "Adel Bibi",
                "collaborators": [
                    "Bernard Ghanem",
                    "Philip H. S. Torr",
                    "Motasem Alfarra",
                    "Aleksandar Petrov",
                    "Peter Richt\u00e1rik",
                    "Modar Alfadly",
                    "El Houcine Bergou",
                    "Ozan Sener",
                    "Matthias M\u00fcller",
                    "Francisco Eiras"
                ],
                "domain": [
                    "Deep Learning",
                    "Adversarial Robustness",
                    "Continual Learning",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Network Moments: Extensions and Sparse-Smooth Attacks",
                        "abstract": "The impressive performance of deep neural networks (DNNs) has immensely strengthened the line of research that aims at theoretically analyzing their effectiveness. This has incited research on the reaction of DNNs to noisy input, namely developing adversarial input attacks and strategies that lead to robust DNNs to these attacks. To that end, in this paper, we derive exact analytic expressions for the first and second moments (mean and variance) of a small piecewise linear (PL) network (Affine, ReLU, Affine) subject to Gaussian input. In particular, we generalize the second-moment expression of Bibi et al. to arbitrary input Gaussian distributions, dropping the zero-mean assumption. We show that the new variance expression can be efficiently approximated leading to much tighter variance estimates as compared to the preliminary results of Bibi et al. Moreover, we experimentally show that these expressions are tight under simple linearizations of deeper PL-DNNs, where we investigate the effect of the linearization sensitivity on the accuracy of the moment estimates. Lastly, we show that the derived expressions can be used to construct sparse and smooth Gaussian adversarial attacks (targeted and non-targeted) that tend to lead to perceptually feasible input attacks."
                    },
                    {
                        "title": "Constrained Clustering: General Pairwise and Cardinality Constraints",
                        "abstract": "We study constrained clustering, where constraints guide the clustering process. In existing works, two categories of constraints have been widely explored, namely pairwise and cardinality constraints. Pairwise constraints enforce the cluster labels of two instances to be the same (must-link constraints) or different (cannot-link constraints). Cardinality constraints encourage cluster sizes to satisfy a user-specified distribution. Most existing constrained clustering models can only utilize one category of constraints at a time. We enforce the above two categories into a unified clustering model starting with the integer program formulation of the standard K-means. As the two categories provide different useful information, utilizing both allow for better clustering performance. However, the optimization is difficult due to the binary and quadratic constraints in the unified formulation. To solve this, we utilize two techniques: equivalently replacing the binary constraints by the intersection of two continuous constraints; the other is transforming the quadratic constraints into bi-linear constraints by introducing extra variables. We derive an equivalent continuous reformulation with simple constraints, which can be efficiently solved by Alternating Direction Method of Multipliers. Extensive experiments on both synthetic and real data demonstrate when: (1) utilizing a single category of constraint, the proposed model is superior to or competitive with SOTA constrained clustering models, and (2) utilizing both categories of constraints jointly, the proposed model shows better performance than the case of the single category. The experiments show that the proposed method exploits the constraints to achieve perfect clustering performance with improved clustering to 2%-5% in classical clustering metrics, e.g. Adjusted Random, Mirkin's, and Huber's, indices outerperfomring other methods."
                    },
                    {
                        "title": "Analytical Moment Regularizer for Gaussian Robust Networks",
                        "abstract": "Despite the impressive performance of deep neural networks (DNNs) on numerous vision tasks, they still exhibit yet-to-understand uncouth behaviours. One puzzling behaviour is the subtle sensitive reaction of DNNs to various noise attacks. Such a nuisance has strengthened the line of research around developing and training noise-robust networks. In this work, we propose a new training regularizer that aims to minimize the probabilistic expected training loss of a DNN subject to a generic Gaussian input. We provide an efficient and simple approach to approximate such a regularizer for arbitrary deep networks. This is done by leveraging the analytic expression of the output mean of a shallow neural network; avoiding the need for the memory and computationally expensive data augmentation. We conduct extensive experiments on LeNet and AlexNet on various datasets including MNIST, CIFAR10, and CIFAR100 demonstrating the effectiveness of our proposed regularizer. In particular, we show that networks that are trained with the proposed regularizer benefit from a boost in robustness equivalent to performing 3-21 folds of data augmentation."
                    },
                    {
                        "title": "A Stochastic Derivative-Free Optimization Method with Importance Sampling: Theory and Learning to Control",
                        "abstract": "We consider the problem of unconstrained minimization of a smooth objective function in $\\R^n$ in a setting where only function evaluations are possible. While importance sampling is one of the most popular techniques used by machine learning practitioners to accelerate the convergence of their models when applicable, there is not much existing theory for this acceleration in the derivative-free setting. In this paper, we propose the first derivative free optimization method with importance sampling and derive new improved complexity results on non-convex, convex and strongly convex functions. We conduct extensive experiments on various synthetic and real LIBSVM datasets confirming our theoretical results. We further test our method on a collection of continuous control tasks on MuJoCo environments with varying difficulty. Experiments suggest that our algorithm is practical for high dimensional continuous control problems where importance sampling results in a significant sample complexity improvement."
                    },
                    {
                        "title": "SimCS: Simulation for Domain Incremental Online Continual Segmentation",
                        "abstract": "Continual Learning is a step towards lifelong intelligence where models continuously learn from recently collected data without forgetting previous knowledge. Existing continual learning approaches mostly focus on image classification in the class-incremental setup with clear task boundaries and unlimited computational budget. This work explores the problem of Online Domain-Incremental Continual Segmentation (ODICS), where the model is continually trained over batches of densely labeled images from different domains, with limited computation and no information about the task boundaries. ODICS arises in many practical applications. In autonomous driving, this may correspond to the realistic scenario of training a segmentation model over time on a sequence of cities. We analyze several existing continual learning methods and show that they perform poorly in this setting despite working well in class-incremental segmentation. We propose SimCS, a parameter-free method complementary to existing ones that uses simulated data to regularize continual learning. Experiments show that SimCS provides consistent improvements when combined with different CL methods."
                    },
                    {
                        "title": "When Do Prompting and Prefix-Tuning Work? A Theory of Capabilities and Limitations",
                        "abstract": "Context-based fine-tuning methods, including prompting, in-context learning, soft prompting (also known as prompt tuning), and prefix-tuning, have gained popularity due to their ability to often match the performance of full fine-tuning with a fraction of the parameters. Despite their empirical successes, there is little theoretical understanding of how these techniques influence the internal computation of the model and their expressiveness limitations. We show that despite the continuous embedding space being more expressive than the discrete token space, soft-prompting and prefix-tuning are potentially less expressive than full fine-tuning, even with the same number of learnable parameters. Concretely, context-based fine-tuning cannot change the relative attention pattern over the content and can only bias the outputs of an attention layer in a fixed direction. This suggests that while techniques like prompting, in-context learning, soft prompting, and prefix-tuning can effectively elicit skills present in the pretrained model, they may not be able to learn novel tasks that require new attention patterns."
                    },
                    {
                        "title": "Prompting a Pretrained Transformer Can Be a Universal Approximator",
                        "abstract": "Despite the widespread adoption of prompting, prompt tuning and prefix-tuning of transformer models, our theoretical understanding of these fine-tuning methods remains limited. A key question is whether one can arbitrarily modify the behavior of pretrained model by prompting or prefix-tuning it. Formally, whether prompting and prefix-tuning a pretrained model can universally approximate sequence-to-sequence functions. This paper answers in the affirmative and demonstrates that much smaller pretrained models than previously thought can be universal approximators when prefixed. In fact, the attention mechanism is uniquely suited for universal approximation with prefix-tuning a single attention head being sufficient to approximate any continuous function. Moreover, any sequence-to-sequence function can be approximated by prefixing a transformer with depth linear in the sequence length. Beyond these density-type results, we also offer Jackson-type bounds on the length of the prefix needed to approximate a function to a desired precision."
                    },
                    {
                        "title": "Certifying Ensembles: A General Certification Theory with S-Lipschitzness",
                        "abstract": "Improving and guaranteeing the robustness of deep learning models has been a topic of intense research. Ensembling, which combines several classifiers to provide a better model, has shown to be beneficial for generalisation, uncertainty estimation, calibration, and mitigating the effects of concept drift. However, the impact of ensembling on certified robustness is less well understood. In this work, we generalise Lipschitz continuity by introducing S-Lipschitz classifiers, which we use to analyse the theoretical robustness of ensembles. Our results are precise conditions when ensembles of robust classifiers are more robust than any constituent classifier, as well as conditions when they are less robust."
                    },
                    {
                        "title": "On the Decision Boundaries of Neural Networks: A Tropical Geometry Perspective",
                        "abstract": "This work tackles the problem of characterizing and understanding the decision boundaries of neural networks with piecewise linear non-linearity activations. We use tropical geometry, a new development in the area of algebraic geometry, to characterize the decision boundaries of a simple network of the form (Affine, ReLU, Affine). Our main finding is that the decision boundaries are a subset of a tropical hypersurface, which is intimately related to a polytope formed by the convex hull of two zonotopes. The generators of these zonotopes are functions of the network parameters. This geometric characterization provides new perspectives to three tasks. (i) We propose a new tropical perspective to the lottery ticket hypothesis, where we view the effect of different initializations on the tropical geometric representation of a network's decision boundaries. (ii) Moreover, we propose new tropical based optimization reformulations that directly influence the decision boundaries of the network for the task of network pruning. (iii) At last, we discuss the reformulation of the generation of adversarial attacks in a tropical sense. We demonstrate that one can construct adversaries in a new tropical setting by perturbing a specific set of decision boundaries by perturbing a set of parameters in the network."
                    },
                    {
                        "title": "Improving SAGA via a Probabilistic Interpolation with Gradient Descent",
                        "abstract": "We develop and analyze a new algorithm for empirical risk minimization, which is the key paradigm for training supervised machine learning models. Our method---SAGD---is based on a probabilistic interpolation of SAGA and gradient descent (GD). In particular, in each iteration we take a gradient step with probability $q$ and a SAGA step with probability $1-q$. We show that, surprisingly, the total expected complexity of the method (which is obtained by multiplying the number of iterations by the expected number of gradients computed in each iteration) is minimized for a non-trivial probability $q$. For example, for a well conditioned problem the choice $q=1/(n-1)^2$, where $n$ is the number of data samples, gives a method with an overall complexity which is better than both the complexity of GD and SAGA. We further generalize the results to a probabilistic interpolation of SAGA and minibatch SAGA, which allows us to compute both the optimal probability and the optimal minibatch size. While the theoretical improvement may not be large, the practical improvement is robustly present across all synthetic and real data we tested for, and can be substantial. Our theoretical results suggest that for this optimal minibatch size our method achieves linear speedup in minibatch size, which is of key practical importance as minibatch implementations are used to train machine learning models in practice. Moreover, empirical evidence suggest that a linear speedup in minibatch size can be attained with a parallel implementation."
                    },
                    {
                        "title": "Data-Dependent Randomized Smoothing",
                        "abstract": "Randomized smoothing is a recent technique that achieves state-of-art performance in training certifiably robust deep neural networks. While the smoothing family of distributions is often connected to the choice of the norm used for certification, the parameters of these distributions are always set as global hyper parameters independent from the input data on which a network is certified. In this work, we revisit Gaussian randomized smoothing and show that the variance of the Gaussian distribution can be optimized at each input so as to maximize the certification radius for the construction of the smooth classifier. Since the data dependent classifier does not directly enjoy sound certification with existing approaches, we propose a memory-enhanced data dependent smooth classifier that is certifiable by construction. This new approach is generic, parameter-free, and easy to implement. In fact, we show that our data dependent framework can be seamlessly incorporated into 3 randomized smoothing approaches, leading to consistent improved certified accuracy. When this framework is used in the training routine of these approaches followed by a data dependent certification, we achieve 9% and 6% improvement over the certified accuracy of the strongest baseline for a radius of 0.5 on CIFAR10 and ImageNet."
                    },
                    {
                        "title": "TrackingNet: A Large-Scale Dataset and Benchmark for Object Tracking in the Wild",
                        "abstract": "Despite the numerous developments in object tracking, further development of current tracking algorithms is limited by small and mostly saturated datasets. As a matter of fact, data-hungry trackers based on deep-learning currently rely on object detection datasets due to the scarcity of dedicated large-scale tracking datasets. In this work, we present TrackingNet, the first large-scale dataset and benchmark for object tracking in the wild. We provide more than 30K videos with more than 14 million dense bounding box annotations. Our dataset covers a wide selection of object classes in broad and diverse context. By releasing such a large-scale dataset, we expect deep trackers to further improve and generalize. In addition, we introduce a new benchmark composed of 500 novel videos, modeled with a distribution similar to our training dataset. By sequestering the annotation of the test set and providing an online evaluation server, we provide a fair benchmark for future development of object trackers. Deep trackers fine-tuned on a fraction of our dataset improve their performance by up to 1.6% on OTB100 and up to 1.7% on TrackingNet Test. We provide an extensive benchmark on TrackingNet by evaluating more than 20 trackers. Our results suggest that object tracking in the wild is far from being solved."
                    },
                    {
                        "title": "Rethinking Clustering for Robustness",
                        "abstract": "This paper studies how encouraging semantically-aligned features during deep neural network training can increase network robustness. Recent works observed that Adversarial Training leads to robust models, whose learnt features appear to correlate with human perception. Inspired by this connection from robustness to semantics, we study the complementary connection: from semantics to robustness. To do so, we provide a robustness certificate for distance-based classification models (clustering-based classifiers). Moreover, we show that this certificate is tight, and we leverage it to propose ClusTR (Clustering Training for Robustness), a clustering-based and adversary-free training framework to learn robust models. Interestingly, \\textit{ClusTR} outperforms adversarially-trained networks by up to $4\\%$ under strong PGD attacks."
                    },
                    {
                        "title": "A Stochastic Derivative Free Optimization Method with Momentum",
                        "abstract": "We consider the problem of unconstrained minimization of a smooth objective function in $\\mathbb{R}^d$ in setting where only function evaluations are possible. We propose and analyze stochastic zeroth-order method with heavy ball momentum. In particular, we propose, SMTP, a momentum version of the stochastic three-point method (STP) \\cite{Bergou_2018}. We show new complexity results for non-convex, convex and strongly convex functions. We test our method on a collection of learning to continuous control tasks on several MuJoCo \\cite{Todorov_2012} environments with varying difficulty and compare against STP, other state-of-the-art derivative-free optimization algorithms and against policy gradient methods. SMTP significantly outperforms STP and all other methods that we considered in our numerical experiments. Our second contribution is SMTP with importance sampling which we call SMTP_IS. We provide convergence analysis of this method for non-convex, convex and strongly convex objectives."
                    },
                    {
                        "title": "DeformRS: Certifying Input Deformations with Randomized Smoothing",
                        "abstract": "Deep neural networks are vulnerable to input deformations in the form of vector fields of pixel displacements and to other parameterized geometric deformations e.g. translations, rotations, etc. Current input deformation certification methods either 1. do not scale to deep networks on large input datasets, or 2. can only certify a specific class of deformations, e.g. only rotations. We reformulate certification in randomized smoothing setting for both general vector field and parameterized deformations and propose DeformRS-VF and DeformRS-Par, respectively. Our new formulation scales to large networks on large input datasets. For instance, DeformRS-Par certifies rich deformations, covering translations, rotations, scaling, affine deformations, and other visually aligned deformations such as ones parameterized by Discrete-Cosine-Transform basis. Extensive experiments on MNIST, CIFAR10, and ImageNet show competitive performance of DeformRS-Par achieving a certified accuracy of $39\\%$ against perturbed rotations in the set $[-10\\degree,10\\degree]$ on ImageNet."
                    },
                    {
                        "title": "Diversified Dynamic Routing for Vision Tasks",
                        "abstract": "Deep learning models for vision tasks are trained on large datasets under the assumption that there exists a universal representation that can be used to make predictions for all samples. Whereas high complexity models are proven to be capable of learning such representations, a mixture of experts trained on specific subsets of the data can infer the labels more efficiently. However using mixture of experts poses two new problems, namely (i) assigning the correct expert at inference time when a new unseen sample is presented. (ii) Finding the optimal partitioning of the training data, such that the experts rely the least on common features. In Dynamic Routing (DR) a novel architecture is proposed where each layer is composed of a set of experts, however without addressing the two challenges we demonstrate that the model reverts to using the same subset of experts.   In our method, Diversified Dynamic Routing (DivDR) the model is explicitly trained to solve the challenge of finding relevant partitioning of the data and assigning the correct experts in an unsupervised approach. We conduct several experiments on semantic segmentation on Cityscapes and object detection and instance segmentation on MS-COCO showing improved performance over several baselines."
                    },
                    {
                        "title": "Language Model Tokenizers Introduce Unfairness Between Languages",
                        "abstract": "Recent language models have shown impressive multilingual performance, even when not explicitly trained for it. Despite this, there are concerns about the quality of their outputs across different languages. In this paper, we show how disparity in the treatment of different languages arises at the tokenization stage, well before a model is even invoked. The same text translated into different languages can have drastically different tokenization lengths, with differences up to 15 times in some cases. These disparities persist even for tokenizers that are intentionally trained for multilingual support. Character-level and byte-level models also exhibit over 4 times the difference in the encoding length for some language pairs. This induces unfair treatment for some language communities in regard to the cost of accessing commercial language services, the processing time and latency, as well as the amount of content that can be provided as context to the models. Therefore, we make the case that we should train future language models using multilingually fair subword tokenizers."
                    },
                    {
                        "title": "Lifelong Benchmarks: Efficient Model Evaluation in an Era of Rapid Progress",
                        "abstract": "Standardized benchmarks drive progress in machine learning. However, with repeated testing, the risk of overfitting grows as algorithms over-exploit benchmark idiosyncrasies. In our work, we seek to mitigate this challenge by compiling ever-expanding large-scale benchmarks called Lifelong Benchmarks. As exemplars of our approach, we create Lifelong-CIFAR10 and Lifelong-ImageNet, containing (for now) 1.69M and 1.98M test samples, respectively. While reducing overfitting, lifelong benchmarks introduce a key challenge: the high cost of evaluating a growing number of models across an ever-expanding sample set. To address this challenge, we also introduce an efficient evaluation framework: Sort \\& Search (S&S), which reuses previously evaluated models by leveraging dynamic programming algorithms to selectively rank and sub-select test samples, enabling cost-effective lifelong benchmarking. Extensive empirical evaluations across 31,000 models demonstrate that S&S achieves highly-efficient approximate accuracy measurement, reducing compute cost from 180 GPU days to 5 GPU hours (1000x reduction) on a single A100 GPU, with low approximation error. As such, lifelong benchmarks offer a robust, practical solution to the \"benchmark exhaustion\" problem."
                    },
                    {
                        "title": "Towards Certification of Uncertainty Calibration under Adversarial Attacks",
                        "abstract": "Since neural classifiers are known to be sensitive to adversarial perturbations that alter their accuracy, \\textit{certification methods} have been developed to provide provable guarantees on the insensitivity of their predictions to such perturbations. Furthermore, in safety-critical applications, the frequentist interpretation of the confidence of a classifier (also known as model calibration) can be of utmost importance. This property can be measured via the Brier score or the expected calibration error. We show that attacks can significantly harm calibration, and thus propose certified calibration as worst-case bounds on calibration under adversarial perturbations. Specifically, we produce analytic bounds for the Brier score and approximate bounds via the solution of a mixed-integer program on the expected calibration error. Finally, we propose novel calibration attacks and demonstrate how they can improve model calibration through \\textit{adversarial calibration training}."
                    },
                    {
                        "title": "Mimicking User Data: On Mitigating Fine-Tuning Risks in Closed Large Language Models",
                        "abstract": "Fine-tuning large language models on small, high-quality datasets can enhance their performance on specific downstream tasks. Recent research shows that fine-tuning on benign, instruction-following data can inadvertently undo the safety alignment process and increase a model's propensity to comply with harmful queries. Although critical, understanding and mitigating safety risks in well-defined tasks remains distinct from the instruction-following context due to structural differences in the data. Our work addresses the gap in our understanding of these risks across diverse types of data in closed models - where providers control how user data is utilized in the fine-tuning process. We demonstrate how malicious actors can subtly manipulate the structure of almost any task-specific dataset to foster significantly more dangerous model behaviors, while maintaining an appearance of innocuity and reasonable downstream task performance. To address this issue, we propose a novel mitigation strategy that mixes in safety data which mimics the task format and prompting style of the user data, showing this is more effective than existing baselines at re-establishing safety alignment while maintaining similar task performance."
                    }
                ]
            }
        }
    },
    "2409.03142": {
        "paper_data": {
            "title": "Causal Temporal Representation Learning with Nonstationary Sparse Transition",
            "url": "http://arxiv.org/abs/2409.03142v1",
            "arxiv_id": "2409.03142",
            "authors": [
                "Xiangchen Song",
                "Zijian Li",
                "Guangyi Chen",
                "Yujia Zheng",
                "Yewen Fan",
                "Xinshuai Dong",
                "Kun Zhang"
            ],
            "abstract": "Causal Temporal Representation Learning (Ctrl) methods aim to identify the temporal causal dynamics of complex nonstationary temporal sequences. Despite the success of existing Ctrl methods, they require either directly observing the domain variables or assuming a Markov prior on them. Such requirements limit the application of these methods in real-world scenarios when we do not have such prior knowledge of the domain variables. To address this problem, this work adopts a sparse transition assumption, aligned with intuitive human understanding, and presents identifiability results from a theoretical perspective. In particular, we explore under what conditions on the significance of the variability of the transitions we can build a model to identify the distribution shifts. Based on the theoretical result, we introduce a novel framework, Causal Temporal Representation Learning with Nonstationary Sparse Transition (CtrlNS), designed to leverage the constraints on transition sparsity and conditional independence to reliably identify both distribution shifts and latent factors. Our experimental evaluations on synthetic and real-world datasets demonstrate significant improvements over existing baselines, highlighting the effectiveness of our approach.",
            "introduction": "   1 Introduction  Causal learning from sequential data remains a fundamental yet challenging task\u00a0[1, 2, 3]. Discovering temporal causal relations among observed variables has been extensively studied in the literature\u00a0[4, 5, 6]. However, in many real-world scenarios such as video understanding\u00a0[7], observed data are generated by causally related latent temporal processes or confounders rather than direct causal edges. This leads to the task of causal temporal representation learning\u00a0(Ctrl), which aims to build compact representations that concisely capture the data generation processes by inverting the mixing function that transforms latent factors into observations and identifying the transitions that govern the underlying latent causal dynamics. This learning problem is known to be challenging without specific assumptions\u00a0[8, 9]. The task becomes significantly more complex with nonstationary transitions, which are often characterized by multiple distribution shifts across different domains, particularly when these domains or shifts are also unobserved.   Recent advances in unsupervised representation learning, particularly through nonlinear Independent Component Analysis (ICA), have shown promising results in identifying latent variables by incorporating side information such as class labels and domain indices\u00a0[10, 11, 12, 13, 14, 15, 16, 17, 18, 19]. For time-series data, historical information is widely utilized to enhance the identifiability of latent temporal causal processes\u00a0[20, 21, 22, 23]. However, existing studies primarily derive results under stationary conditions\u00a0[11, 21] or nonstationary conditions with observed domain indices\u00a0[13, 22, 23]. These methods are limited in application as general time series data are typically nonstationary and domain information is difficult to obtain. Recent studies\u00a0[15, 24, 25, 26] have adopted a Markov structure to handle nonstationary domain variables and can infer domain indices directly from observed data. (More related work can be found in Appendix\u00a0S5.) However, these methods face significant limitations; some are inadequate for modeling time-delayed causal relationships in latent spaces, and they rely on the Markov property, which cannot adequately capture the arbitrary nonstationary variations in domain variables. This leads us to the following important yet unresolved question:    How can we establish identifiability of nonstationary nonlinear ICA for general sequence data without prior knowledge of domain variables?    Relying on observing domain variables or known Markov priors to capture nonstationarity seems counter-intuitive, especially considering how easily humans can identify domain shifts given sufficient variation on transitions, such as video action segmentation\u00a0[27, 28] and recognition\u00a0[29, 30, 31] tasks. In this work, we theoretically investigate the conditions on the significance of transition variability to identify distribution shifts. The core idea is transition clustering, assuming transitions within the same domain are similar, while transitions across different domains are distinct. Building on this identification theorem, we propose Causal Temporal Representation Learning with Nonstationary Sparse Transition\u00a0(CtrlNS), to identify both distribution shifts and latent temporal dynamics. Specifically, we constrain the complexity of the transition function to identify domain shifts. Subsequently, with the identified domain variables, we learn the latent variables using conditional independence constraints. These two processes are jointly optimized within a VAE framework.   The main contributions of this work are as follows: (1) To our best knowledge, this is the first identifiability result that handles nonstationary time-delayed causally related latent temporal processes without prior knowledge of the domain variables. (2) We present CtrlNS, a principled framework for recovering",
            "references": [
                {
                    "title": "Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment",
                    "abstract": "Weakly-supervised action segmentation is a task of learning to partition a long video into several action segments, where training videos are only accompanied by transcripts (ordered list of actions). Most of existing methods need to infer pseudo segmentation for training by serial alignment between all frames and the transcript, which is time-consuming and hard to be parallelized while training. In this work, we aim to escape from this inefficient alignment with massive but redundant frames, and instead to directly localize a few action transitions for pseudo segmentation generation, where a transition refers to the change from an action segment to its next adjacent one in the transcript. As the true transitions are submerged in noisy boundaries due to intra-segment visual variation, we propose a novel Action-Transition-Aware Boundary Alignment (ATBA) framework to efficiently and effectively filter out noisy boundaries and detect transitions. In addition, to boost the semantic learning in the case that noise is inevitably present in the pseudo segmentation, we also introduce video-level losses to utilize the trusted video-level supervision. Extensive experiments show the effectiveness of our approach on both performance and training speed. 11Code is available at https://github.com/iSEE-Laboratory/CVPR24_ATBA."
                },
                {
                    "title": "When and How: Learning Identifiable Latent States for Nonstationary Time Series Forecasting",
                    "abstract": "Temporal distribution shifts are ubiquitous in time series data. One of the most popular methods assumes that the temporal distribution shift occurs uniformly to disentangle the stationary and nonstationary dependencies. But this assumption is difficult to meet, as we do not know when the distribution shifts occur. To solve this problem, we propose to learn IDentifiable latEnt stAtes (IDEA) to detect when the distribution shifts occur. Beyond that, we further disentangle the stationary and nonstationary latent states via sufficient observation assumption to learn how the latent states change. Specifically, we formalize the causal process with environment-irrelated stationary and environment-related nonstationary variables. Under mild conditions, we show that latent environments and stationary/nonstationary variables are identifiable. Based on these theories, we devise the IDEA model, which incorporates an autoregressive hidden Markov model to estimate latent environments and modular prior networks to identify latent states. The IDEA model outperforms several latest nonstationary forecasting methods on various benchmark datasets, highlighting its advantages in real-world scenarios."
                },
                {
                    "title": "CaRiNG: Learning Temporal Causal Representation under Non-Invertible Generation Process",
                    "abstract": "Identifying the underlying time-delayed latent causal processes in sequential data is vital for grasping temporal dynamics and making downstream reasoning. While some recent methods can robustly identify these latent causal variables, they rely on strict assumptions about the invertible generation process from latent variables to observed data. However, these assumptions are often hard to satisfy in real-world applications containing information loss. For instance, the visual perception process translates a 3D space into 2D images, or the phenomenon of persistence of vision incorporates historical data into current perceptions. To address this challenge, we establish an identifiability theory that allows for the recovery of independent latent components even when they come from a nonlinear and non-invertible mix. Using this theory as a foundation, we propose a principled approach, CaRiNG, to learn the CAusal RepresentatIon of Non-invertible Generative temporal data with identifiability guarantees. Specifically, we utilize temporal context to recover lost latent information and apply the conditions in our theory to guide the training process. Through experiments conducted on synthetic datasets, we validate that our CaRiNG method reliably identifies the causal process, even when the generation process is non-invertible. Moreover, we demonstrate that our approach considerably improves temporal understanding and reasoning in practical applications."
                },
                {
                    "title": "Nonparametric Partial Disentanglement via Mechanism Sparsity: Sparse Actions, Interventions and Sparse Temporal Dependencies",
                    "abstract": "This work introduces a novel principle for disentanglement we call mechanism sparsity regularization, which applies when the latent factors of interest depend sparsely on observed auxiliary variables and/or past latent factors. We propose a representation learning method that induces disentanglement by simultaneously learning the latent factors and the sparse causal graphical model that explains them. We develop a nonparametric identifiability theory that formalizes this principle and shows that the latent factors can be recovered by regularizing the learned causal graph to be sparse. More precisely, we show identifiablity up to a novel equivalence relation we call\"consistency\", which allows some latent factors to remain entangled (hence the term partial disentanglement). To describe the structure of this entanglement, we introduce the notions of entanglement graphs and graph preserving functions. We further provide a graphical criterion which guarantees complete disentanglement, that is identifiability up to permutations and element-wise transformations. We demonstrate the scope of the mechanism sparsity principle as well as the assumptions it relies on with several worked out examples. For instance, the framework shows how one can leverage multi-node interventions with unknown targets on the latent factors to disentangle them. We further draw connections between our nonparametric results and the now popular exponential family assumption. Lastly, we propose an estimation procedure based on variational autoencoders and a sparsity constraint and demonstrate it on various synthetic datasets. This work is meant to be a significantly extended version of Lachapelle et al. (2022)."
                },
                {
                    "title": "Generalizing Nonlinear ICA Beyond Structural Sparsity",
                    "abstract": "Nonlinear independent component analysis (ICA) aims to uncover the true latent sources from their observable nonlinear mixtures. Despite its significance, the identifiability of nonlinear ICA is known to be impossible without additional assumptions. Recent advances have proposed conditions on the connective structure from sources to observed variables, known as Structural Sparsity, to achieve identifiability in an unsupervised manner. However, the sparsity constraint may not hold universally for all sources in practice. Furthermore, the assumptions of bijectivity of the mixing process and independence among all sources, which arise from the setting of ICA, may also be violated in many real-world scenarios. To address these limitations and generalize nonlinear ICA, we propose a set of new identifiability results in the general settings of undercompleteness, partial sparsity and source dependence, and flexible grouping structures. Specifically, we prove identifiability when there are more observed variables than sources (undercomplete), and when certain sparsity and/or source independence assumptions are not met for some changing sources. Moreover, we show that even in cases with flexible grouping structures (e.g., part of the sources can be divided into irreducible independent groups with various sizes), appropriate identifiability results can also be established. Theoretical claims are supported empirically on both synthetic and real-world datasets."
                },
                {
                    "title": "Temporally Disentangled Representation Learning under Unknown Nonstationarity",
                    "abstract": "In unsupervised causal representation learning for sequential data with time-delayed latent causal influences, strong identifiability results for the disentanglement of causally-related latent variables have been established in stationary settings by leveraging temporal structure. However, in nonstationary setting, existing work only partially addressed the problem by either utilizing observed auxiliary variables (e.g., class labels and/or domain indexes) as side information or assuming simplified latent causal dynamics. Both constrain the method to a limited range of scenarios. In this study, we further explored the Markov Assumption under time-delayed causally related process in nonstationary setting and showed that under mild conditions, the independent latent components can be recovered from their nonlinear mixture up to a permutation and a component-wise transformation, without the observation of auxiliary variables. We then introduce NCTRL, a principled estimation framework, to reconstruct time-delayed latent causal variables and identify their relations from measured sequential data only. Empirical evaluations demonstrated the reliable identification of time-delayed latent causal influences, with our methodology substantially outperforming existing baselines that fail to exploit the nonstationarity adequately and then, consequently, cannot distinguish distribution shifts."
                },
                {
                    "title": "HOI-aware Adaptive Network for Weakly-supervised Action Segmentation",
                    "abstract": "In this paper, we propose an HOI-aware adaptive network named AdaAct for weakly-supervised action segmentation. Most existing methods learn a fixed network to predict the action of each frame with the neighboring frames. However, this would result in ambiguity when estimating similar actions, such as pouring juice and pouring coffee. To address this, we aim to exploit temporally global but spatially local human-object interactions (HOI) as video-level prior knowledge for action segmentation. The long-term HOI sequence provides crucial contextual information to distinguish ambiguous actions, where our network dynamically adapts to the given HOI sequence at test time. More specifically, we first design a video HOI encoder that extracts, selects, and integrates the most representative HOI throughout the video. Then, we propose a two-branch HyperNetwork to learn an adaptive temporal encoder, which automatically adjusts the parameters based on the HOI information of various videos on the fly. Extensive experiments on two widely-used datasets including Breakfast and 50Salads demonstrate the effectiveness of our method under different evaluation metrics."
                },
                {
                    "title": "Temporally Disentangled Representation Learning",
                    "abstract": "Recently in the field of unsupervised representation learning, strong identifiability results for disentanglement of causally-related latent variables have been established by exploiting certain side information, such as class labels, in addition to independence. However, most existing work is constrained by functional form assumptions such as independent sources or further with linear transitions, and distribution assumptions such as stationary, exponential family distribution. It is unknown whether the underlying latent variables and their causal relations are identifiable if they have arbitrary, nonparametric causal influences in between. In this work, we establish the identifiability theories of nonparametric latent causal processes from their nonlinear mixtures under fixed temporal causal influences and analyze how distribution changes can further benefit the disentanglement. We propose \\textbf{\\texttt{TDRL}}, a principled framework to recover time-delayed latent causal variables and identify their relations from measured sequential data under stationary environments and under different distribution shifts. Specifically, the framework can factorize unknown distribution shifts into transition distribution changes under fixed and time-varying latent causal relations, and under observation changes in observation. Through experiments, we show that time-delayed latent causal influences are reliably identified and that our approach considerably outperforms existing baselines that do not correctly exploit this modular representation of changes. Our code is available at: \\url{https://github.com/weirayao/tdrl}."
                },
                {
                    "title": "Latent Hierarchical Causal Structure Discovery with Rank Constraints",
                    "abstract": "Most causal discovery procedures assume that there are no latent confounders in the system, which is often violated in real-world problems. In this paper, we consider a challenging scenario for causal structure identification, where some variables are latent and they form a hierarchical graph structure to generate the measured variables; the children of latent variables may still be latent and only leaf nodes are measured, and moreover, there can be multiple paths between every pair of variables (i.e., it is beyond tree structure). We propose an estimation procedure that can efficiently locate latent variables, determine their cardinalities, and identify the latent hierarchical structure, by leveraging rank deficiency constraints over the measured variables. We show that the proposed algorithm can find the correct Markov equivalence class of the whole graph asymptotically under proper restrictions on the graph structure."
                },
                {
                    "title": "On the Identifiability of Nonlinear ICA: Sparsity and Beyond",
                    "abstract": "Nonlinear independent component analysis (ICA) aims to recover the underlying independent latent sources from their observable nonlinear mixtures. How to make the nonlinear ICA model identifiable up to certain trivial indeterminacies is a long-standing problem in unsupervised learning. Recent breakthroughs reformulate the standard independence assumption of sources as conditional independence given some auxiliary variables (e.g., class labels and/or domain/time indexes) as weak supervision or inductive bias. However, nonlinear ICA with unconditional priors cannot benefit from such developments. We explore an alternative path and consider only assumptions on the mixing process, such as Structural Sparsity. We show that under specific instantiations of such constraints, the independent latent sources can be identified from their nonlinear mixtures up to a permutation and a component-wise transformation, thus achieving nontrivial identifiability of nonlinear ICA without auxiliary variables. We provide estimation methods and validate the theoretical results experimentally. The results on image data suggest that our conditions may hold in a number of practical data generating processes."
                },
                {
                    "title": "Causal Representation Learning for Instantaneous and Temporal Effects in Interactive Systems",
                    "abstract": "Causal representation learning is the task of identifying the underlying causal variables and their relations from high-dimensional observations, such as images. Recent work has shown that one can reconstruct the causal variables from temporal sequences of observations under the assumption that there are no instantaneous causal relations between them. In practical applications, however, our measurement or frame rate might be slower than many of the causal effects. This effectively creates\"instantaneous\"effects and invalidates previous identifiability results. To address this issue, we propose iCITRIS, a causal representation learning method that allows for instantaneous effects in intervened temporal sequences when intervention targets can be observed, e.g., as actions of an agent. iCITRIS identifies the potentially multidimensional causal variables from temporal observations, while simultaneously using a differentiable causal discovery method to learn their causal graph. In experiments on three datasets of interactive systems, iCITRIS accurately identifies the causal variables and their causal graph."
                },
                {
                    "title": "Set-Supervised Action Learning in Procedural Task Videos via Pairwise Order Consistency",
                    "abstract": "We address the problem of set-supervised action learning, whose goal is to learn an action segmentation model using weak supervision in the form of sets of actions occurring in training videos. Our key observation is that videos within the same task have similar ordering of actions, which can be leveraged for effective learning. Therefore, we propose an attention-based method with a new Pairwise Ordering Consistency (POC) loss that encourages that for each common action pair in two videos of the same task, the attentions of actions follow a similar ordering. Unlike existing sequence alignment methods, which misalign actions in videos with different orderings or cannot reliably separate more from less consistent orderings, our POC loss efficiently aligns videos with different action orders and is differentiable, which enables end-to-end training. In addition, it avoids the time-consuming pseudo-label generation of prior works. Our method efficiently learns the actions and their temporal locations, therefore, extends the existing attention-based action localization methods from learning one action per video to multiple actions using our POC loss along with video-level and frame-level losses. By experiments on three datasets, we demonstrate that our method significantly improves the state of the art. We also show that our method, with a small modification, can effectively address the transcript-supervised action learning task, where actions and their ordering are available during training.11Code available at https://github.com/ZijiaLewisLu/CVPR22-POC."
                },
                {
                    "title": "Detector-Free Weakly Supervised Group Activity Recognition",
                    "abstract": "Group activity recognition is the task of understanding the activity conducted by a group of people as a whole in a multiperson video. Existing models for this task are often impractical in that they demand ground-truth bounding box labels of actors even in testing or rely on off-the-shelf object detectors. Motivated by this, we propose a novel model for group activity recognition that depends neither on bounding box labels nor on object detector. Our model based on Transformer localizes and encodes partial contexts of a group activity by leveraging the attention mechanism, and represents a video clip as a set of partial context embeddings. The embedding vectors are then aggregated to form a single group representation that reflects the entire context of an activity while capturing temporal evolution of each partial context. Our method achieves outstanding performance on two benchmarks, Volleyball and NBA datasets, surpassing not only the state of the art trained with the same level of supervision, but also some of existing models relying on stronger supervision."
                },
                {
                    "title": "CITRIS: Causal Identifiability from Temporal Intervened Sequences",
                    "abstract": "Understanding the latent causal factors of a dynamical system from visual observations is considered a crucial step towards agents reasoning in complex environments. In this paper, we propose CITRIS, a variational autoencoder framework that learns causal representations from temporal sequences of images in which underlying causal factors have possibly been intervened upon. In contrast to the recent literature, CITRIS exploits temporality and observing intervention targets to identify scalar and multidimensional causal factors, such as 3D rotation angles. Furthermore, by introducing a normalizing flow, CITRIS can be easily extended to leverage and disentangle representations obtained by already pretrained autoencoders. Extending previous results on scalar causal factors, we prove identifiability in a more general setting, in which only some components of a causal factor are affected by interventions. In experiments on 3D rendered image sequences, CITRIS outperforms previous methods on recovering the underlying causal variables. Moreover, using pretrained autoencoders, CITRIS can even generalize to unseen instantiations of causal factors, opening future research areas in sim-to-real generalization for causal representation learning."
                },
                {
                    "title": "Transformers in Action: Weakly Supervised Action Segmentation",
                    "abstract": "The video action segmentation task is regularly explored under weaker forms of supervision, such as transcript supervision, where a list of actions is easier to obtain than dense frame-wise labels. In this formulation, the task presents various challenges for sequence modeling approaches due to the emphasis on action transition points, long sequence lengths, and frame contextualization, making the task well-posed for transformers. Given developments enabling transformers to scale linearly, we demonstrate through our architecture how they can be applied to improve action alignment accuracy over the equivalent RNN-based models with the attention mechanism focusing around salient action transition regions. Additionally, given the recent focus on inference-time transcript selection, we propose a supplemental transcript embedding approach to select transcripts more quickly at inference-time. Furthermore, we subsequently demonstrate how this approach can also improve the overall segmentation performance. Finally, we evaluate our proposed methods across the benchmark datasets to better understand the applicability of transformers and the importance of transcript selection on this video-driven weakly-supervised task."
                },
                {
                    "title": "Learning Temporally Causal Latent Processes from General Temporal Data",
                    "abstract": "Our goal is to recover time-delayed latent causal variables and identify their relations from measured temporal data. Estimating causally-related latent variables from observations is particularly challenging as the latent variables are not uniquely recoverable in the most general case. In this work, we consider both a nonparametric, nonstationary setting and a parametric setting for the latent processes and propose two provable conditions under which temporally causal latent processes can be identified from their nonlinear mixtures. We propose LEAP, a theoretically-grounded framework that extends Variational AutoEncoders (VAEs) by enforcing our conditions through proper constraints in causal process prior. Experimental results on various datasets demonstrate that temporally causal latent processes are reliably identified from observed variables under different dependency structures and that our approach considerably outperforms baselines that do not properly leverage history or nonstationarity information. This demonstrates that using temporal information to learn latent processes from their invertible nonlinear mixtures in an unsupervised manner, for which we believe our work is one of the first, seems promising even without sparsity or minimality assumptions."
                },
                {
                    "title": "Weakly-Supervised Action Segmentation and Alignment via Transcript-Aware Union-of-Subspaces Learning",
                    "abstract": "We address the problem of learning to segment actions from weakly-annotated videos, i.e., videos accompanied by transcripts (ordered list of actions). We propose a framework in which we model actions with a union of low-dimensional subspaces, learn the subspaces using transcripts and refine video features that lend themselves to action subspaces. To do so, we design an architecture consisting of a Union-of-Subspaces Network, which is an ensemble of autoencoders, each modeling a low-dimensional action subspace and can capture variations of an action within and across videos. For learning, at each iteration, we generate positive and negative soft alignment matrices using the segmentations from the previous iteration, which we use for discriminative training of our model. To regularize the learning, we introduce a constraint loss that prevents imbalanced segmentations and enforces relatively similar duration of each action across videos. To have a real-time inference, we develop a hierarchical segmentation framework that uses subset selection to find representative transcripts and hierarchically align a test video with increasingly refined representative transcripts. Our experiments on three datasets show that our method improves the state-of-the-art action segmentation and alignment, while speeding up the inference time by a factor of 4 to 13. 1"
                },
                {
                    "title": "Long Short View Feature Decomposition via Contrastive Video Representation Learning",
                    "abstract": "Self-supervised video representation methods typically focus on the representation of temporal attributes in videos. However, the role of stationary versus non-stationary attributes is less explored: Stationary features, which remain similar throughout the video, enable the prediction of video-level action classes. Non-stationary features, which represent temporally varying attributes, are more beneficial for downstream tasks involving more fine-grained temporal understanding, such as action segmentation. We argue that a single representation to capture both types of features is sub-optimal, and propose to decompose the representation space into stationary and non-stationary features via contrastive learning from long and short views, i.e. long video sequences and their shorter sub-sequences. Stationary features are shared between the short and long views, while non-stationary features aggregate the short views to match the corresponding long view. To empirically verify our approach, we demonstrate that our stationary features work particularly well on an action recognition downstream task, while our non-stationary features perform better on action segmentation. Furthermore, we analyse the learned representations and find that stationary features capture more temporally stable, static attributes, while non-stationary features encompass more temporally varying ones."
                },
                {
                    "title": "Disentanglement via Mechanism Sparsity Regularization: A New Principle for Nonlinear ICA",
                    "abstract": "This work introduces a novel principle we call disentanglement via mechanism sparsity regularization, which can be applied when the latent factors of interest depend sparsely on past latent factors and/or observed auxiliary variables. We propose a representation learning method that induces disentanglement by simultaneously learning the latent factors and the sparse causal graphical model that relates them. We develop a rigorous identifiability theory, building on recent nonlinear independent component analysis (ICA) results, that formalizes this principle and shows how the latent variables can be recovered up to permutation if one regularizes the latent mechanisms to be sparse and if some graph connectivity criterion is satisfied by the data generating process. As a special case of our framework, we show how one can leverage unknown-target interventions on the latent factors to disentangle them, thereby drawing further connections between ICA and causality. We propose a VAE-based method in which the latent mechanisms are learned and regularized via binary masks, and validate our theory by showing it learns disentangled representations in simulations."
                },
                {
                    "title": "AdaRL: What, Where, and How to Adapt in Transfer Reinforcement Learning",
                    "abstract": "One practical challenge in reinforcement learning (RL) is how to make quick adaptations when faced with new environments. In this paper, we propose a principled framework for adaptive RL, called \\textit{AdaRL}, that adapts reliably and efficiently to changes across domains with a few samples from the target domain, even in partially observable environments. Specifically, we leverage a parsimonious graphical representation that characterizes structural relationships over variables in the RL system. Such graphical representations provide a compact way to encode what and where the changes across domains are, and furthermore inform us with a minimal set of changes that one has to consider for the purpose of policy adaptation. We show that by explicitly leveraging this compact representation to encode changes, we can efficiently adapt the policy to the target domain, in which only a few samples are needed and further policy optimization is avoided. We illustrate the efficacy of AdaRL through a series of experiments that vary factors in the observation, transition, and reward functions for Cartpole and Atari games."
                },
                {
                    "title": "Disentangling Identifiable Features from Noisy Data with Structured Nonlinear ICA",
                    "abstract": "We introduce a new general identifiable framework for principled disentanglement referred to as Structured Nonlinear Independent Component Analysis (SNICA). Our contribution is to extend the identifiability theory of deep generative models for a very broad class of structured models. While previous works have shown identifiability for specific classes of time-series models, our theorems extend this to more general temporal structures as well as to models with more complex structures such as spatial dependencies. In particular, we establish the major result that identifiability for this framework holds even in the presence of noise of unknown distribution. Finally, as an example of our framework's flexibility, we introduce the first nonlinear ICA model for time-series that combines the following very useful properties: it accounts for both nonstationarity and autocorrelation in a fully unsupervised setting; performs dimensionality reduction; models hidden states; and enables principled estimation and inference by variational maximum-likelihood."
                },
                {
                    "title": "Learning Discriminative Prototypes with Dynamic Time Warping",
                    "abstract": "Dynamic Time Warping (DTW) is widely used for temporal data processing. However, existing methods can neither learn the discriminative prototypes of different classes nor exploit such prototypes for further analysis. We propose Discriminative Prototype DTW (DP-DTW), a novel method to learn class-specific discriminative prototypes for temporal recognition tasks. DP-DTW shows superior performance compared to conventional DTWs on time series classification benchmarks1. Combined with end-to-end deep learning, DP-DTW can handle challenging weakly supervised action segmentation problems and achieves state of the art results on standard benchmarks. Moreover, detailed reasoning on the input video is enabled by the learned action prototypes. Specifically, an action-based video summarization can be obtained by aligning the input sequence with action prototypes."
                },
                {
                    "title": "Generalized Independent Noise Condition for Estimating Latent Variable Causal Graphs",
                    "abstract": "Causal discovery aims to recover causal structures or models underlying the observed data. Despite its success in certain domains, most existing methods focus on causal relations between observed variables, while in many scenarios the observed ones may not be the underlying causal variables (e.g., image pixels), but are generated by latent causal variables or confounders that are causally related. To this end, in this paper, we consider Linear, Non-Gaussian Latent variable Models (LiNGLaMs), in which latent confounders are also causally related, and propose a Generalized Independent Noise (GIN) condition to estimate such latent variable graphs. Specifically, for two observed random vectors $\\mathbf{Y}$ and $\\mathbf{Z}$, GIN holds if and only if $\\omega^{\\intercal}\\mathbf{Y}$ and $\\mathbf{Z}$ are statistically independent, where $\\omega$ is a parameter vector characterized from the cross-covariance between $\\mathbf{Y}$ and $\\mathbf{Z}$. From the graphical view, roughly speaking, GIN implies that causally earlier latent common causes of variables in $\\mathbf{Y}$ d-separate $\\mathbf{Y}$ from $\\mathbf{Z}$. Interestingly, we find that the independent noise condition, i.e., if there is no confounder, causes are independent from the error of regressing the effect on the causes, can be seen as a special case of GIN. Moreover, we show that GIN helps locate latent variables and identify their causal structure, including causal directions. We further develop a recursive learning algorithm to achieve these goals. Experimental results on synthetic and real-world data demonstrate the effectiveness of our method."
                },
                {
                    "title": "Towards Nonlinear Disentanglement in Natural Data with Temporal Sparse Coding",
                    "abstract": "We construct an unsupervised learning model that achieves nonlinear disentanglement of underlying factors of variation in naturalistic videos. Previous work suggests that representations can be disentangled if all but a few factors in the environment stay constant at any point in time. As a result, algorithms proposed for this problem have only been tested on carefully constructed datasets with this exact property, leaving it unclear whether they will transfer to natural scenes. Here we provide evidence that objects in segmented natural movies undergo transitions that are typically small in magnitude with occasional large jumps, which is characteristic of a temporally sparse distribution. We leverage this finding and present SlowVAE, a model for unsupervised representation learning that uses a sparse prior on temporally adjacent observations to disentangle generative factors without any assumptions on the number of changing factors. We provide a proof of identifiability and show that the model reliably learns disentangled representations on several established benchmark datasets, often surpassing the current state-of-the-art. We additionally demonstrate transferability towards video datasets with natural dynamics, Natural Sprites and KITTI Masks, which we contribute as benchmarks for guiding disentanglement research towards more natural data domains."
                },
                {
                    "title": "Hidden Markov Nonlinear ICA: Unsupervised Learning from Nonstationary Time Series",
                    "abstract": "Recent advances in nonlinear Independent Component Analysis (ICA) provide a principled framework for unsupervised feature learning and disentanglement. The central idea in such works is that the latent components are assumed to be independent conditional on some observed auxiliary variables, such as the time-segment index. This requires manual segmentation of data into non-stationary segments which is computationally expensive, inaccurate and often impossible. These models are thus not fully unsupervised. We remedy these limitations by combining nonlinear ICA with a Hidden Markov Model, resulting in a model where a latent state acts in place of the observed segment-index. We prove identifiability of the proposed model for a general mixing nonlinearity, such as a neural network. We also show how maximum likelihood estimation of the model can be done using the expectation-maximization algorithm. Thus, we achieve a new nonlinear ICA framework which is unsupervised, more efficient, as well as able to model underlying temporal dynamics."
                },
                {
                    "title": "Actor-Transformers for Group Activity Recognition",
                    "abstract": "This paper strives to recognize individual actions and group activities from videos. While existing solutions for this challenging problem explicitly model spatial and temporal relationships based on location of individual actors, we propose an actor-transformer model able to learn and selectively extract information relevant for group activity recognition. We feed the transformer with rich actor-specific static and dynamic representations expressed by features from a 2D pose network and 3D CNN, respectively. We empirically study different ways to combine these representations and show their complementary benefits. Experiments show what is important to transform and how it should be transformed. What is more, actor-transformers achieve state-of-the-art results on two publicly available benchmarks for group activity recognition, outperforming the previous best published results by a considerable margin"
                },
                {
                    "title": "Disentanglement by Nonlinear ICA with General Incompressible-flow Networks (GIN)",
                    "abstract": "A central question of representation learning asks under which conditions it is possible to reconstruct the true latent variables of an arbitrarily complex generative process. Recent breakthrough work by Khemakhem et al. (2019) on nonlinear ICA has answered this question for a broad class of conditional generative processes. We extend this important result in a direction relevant for application to real-world data. First, we generalize the theory to the case of unknown intrinsic problem dimension and prove that in some special (but not very restrictive) cases, informative latent variables will be automatically separated from noise by an estimating model. Furthermore, the recovered informative latent variables will be in one-to-one correspondence with the true latent variables of the generating process, up to a trivial component-wise transformation. Second, we introduce a modification of the RealNVP invertible neural network architecture (Dinh et al. (2016)) which is particularly suitable for this type of problem: the General Incompressible-flow Network (GIN). Experiments on artificial data and EMNIST demonstrate that theoretical predictions are indeed verified in practice. In particular, we provide a detailed set of exactly 22 informative latent variables extracted from EMNIST."
                },
                {
                    "title": "Weakly Supervised Energy-Based Learning for Action Segmentation",
                    "abstract": "This paper is about labeling video frames with action classes under weak supervision in training, where we have access to a temporal ordering of actions, but their start and end frames in training videos are unknown. Following prior work, we use an HMM grounded on a Gated Recurrent Unit (GRU) for frame labeling. Our key contribution is a new constrained discriminative forward loss (CDFL) that we use for training the HMM and GRU under weak supervision. While prior work typically estimates the loss on a single, inferred video segmentation, our CDFL discriminates between the energy of all valid and invalid frame labelings of a training video. A valid frame labeling satisfies the ground-truth temporal ordering of actions, whereas an invalid one violates the ground truth. We specify an efficient recursive algorithm for computing the CDFL in terms of the logadd function of the segmentation energy. Our evaluation on action segmentation and alignment gives superior results to those of the state of the art on the benchmark Breakfast Action, Hollywood Extended, and 50Salads datasets."
                },
                {
                    "title": "Variational Autoencoders and Nonlinear ICA: A Unifying Framework",
                    "abstract": "The framework of variational autoencoders allows us to efficiently learn deep latent-variable models, such that the model's marginal distribution over observed variables fits the data. Often, we're interested in going a step further, and want to approximate the true joint distribution over observed and latent variables, including the true prior and posterior distributions over latent variables. This is known to be generally impossible due to unidentifiability of the model. We address this issue by showing that for a broad family of deep latent-variable models, identification of the true joint distribution over observed and latent variables is actually possible up to very simple transformations, thus achieving a principled and powerful form of disentanglement. Our result requires a factorized prior distribution over the latent variables that is conditioned on an additionally observed variable, such as a class label or almost any other observation. We build on recent developments in nonlinear ICA, which we extend to the case with noisy, undercomplete or discrete observations, integrated in a maximum likelihood framework. The result also trivially contains identifiable flow-based generative models as a special case."
                },
                {
                    "title": "A Hybrid RNN-HMM Approach for Weakly Supervised Temporal Action Segmentation",
                    "abstract": "Action recognition has become a rapidly developing research field within the last decade. But with the increasing demand for large scale data, the need of hand annotated data for the training becomes more and more impractical. One way to avoid frame-based human annotation is the use of action order information to learn the respective action classes. In this context, we propose a hierarchical approach to address the problem of weakly supervised learning of human actions from ordered action labels by structuring recognition in a coarse-to-fine manner. Given a set of videos and an ordered list of the occurring actions, the task is to infer start and end frames of the related action classes within the video and to train the respective action classifiers without any need for hand labeled frame boundaries. We address this problem by combining a framewise RNN model with a coarse probabilistic inference. This combination allows for the temporal alignment of long sequences and thus, for an iterative training of both elements. While this system alone already generates good results, we show that the performance can be further improved by approximating the number of subactions to the characteristics of the different action classes as well as by the introduction of a regularizing length prior. The proposed system is evaluated on two benchmark datasets, the Breakfast and the Hollywood extended dataset, showing a competitive performance on various weak learning tasks such as temporal action segmentation and action alignment."
                },
                {
                    "title": "Fast Weakly Supervised Action Segmentation Using Mutual Consistency",
                    "abstract": "Action segmentation is the task of predicting the actions for each frame of a video. As obtaining the full annotation of videos for action segmentation is expensive, weakly supervised approaches that can learn only from transcripts are appealing. In this paper, we propose a novel end-to-end approach for weakly supervised action segmentation based on a two-branch neural network. The two branches of our network predict two redundant but different representations for action segmentation and we propose a novel mutual consistency (MuCon) loss that enforces the consistency of the two redundant representations. Using the MuCon loss together with a loss for transcript prediction, our proposed approach achieves the accuracy of state-of-the-art approaches while being 14 times faster to train and 20 times faster during inference. The MuCon loss proves beneficial even in the fully supervised setting."
                },
                {
                    "title": "Cross-Task Weakly Supervised Learning From Instructional Videos",
                    "abstract": "In this paper we investigate learning visual models for the steps of ordinary tasks using weak supervision via instructional narrations and an ordered list of steps instead of strong supervision via temporal annotations. At the heart of our approach is the observation that weakly supervised learning may be easier if a model shares components while learning different steps: ``pour egg'' should be trained jointly with other tasks involving ``pour'' and ``egg''. We formalize this in a component model for recognizing steps and a weakly supervised learning framework that can learn this model under temporal constraints from narration and the list of steps. Past data does not permit systematic studying of sharing and so we also gather a new dataset aimed at assessing cross-task sharing. Our experiments demonstrate that sharing across tasks improves performance, especially when done at the component level and that our component model can parse previously unseen tasks by virtue of its compositionality."
                },
                {
                    "title": "D3TW: Discriminative Differentiable Dynamic Time Warping for Weakly Supervised Action Alignment and Segmentation",
                    "abstract": "We address weakly supervised action alignment and segmentation in videos, where only the order of occurring actions is available during training. We propose Discriminative Differentiable Dynamic Time Warping (D3TW), the first discriminative model using weak ordering supervision. The key technical challenge for discriminative modeling with weak supervision is that the loss function of the ordering supervision is usually formulated using dynamic programming and is thus not differentiable. We address this challenge with a continuous relaxation of the min-operator in dynamic programming and extend the alignment loss to be differentiable. The proposed D3TW innovatively solves sequence alignment with discriminative modeling and end-to-end training, which substantially improves the performance in weakly supervised action alignment and segmentation tasks. We show that our model is able to bypass the degenerated sequence problem usually encountered in previous work and outperform the current state-of-the-art across three evaluation metrics in two challenging datasets."
                },
                {
                    "title": "Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations",
                    "abstract": "The key idea behind the unsupervised learning of disentangled representations is that real-world data is generated by a few explanatory factors of variation which can be recovered by unsupervised learning algorithms. In this paper, we provide a sober look at recent progress in the field and challenge some common assumptions. We first theoretically show that the unsupervised learning of disentangled representations is fundamentally impossible without inductive biases on both the models and the data. Then, we train more than 12000 models covering most prominent methods and evaluation metrics in a reproducible large-scale experimental study on seven different data sets. We observe that while the different methods successfully enforce properties ``encouraged'' by the corresponding losses, well-disentangled models seemingly cannot be identified without supervision. Furthermore, increased disentanglement does not seem to lead to a decreased sample complexity of learning for downstream tasks. Our results suggest that future work on disentanglement learning should be explicit about the role of inductive biases and (implicit) supervision, investigate concrete benefits of enforcing disentanglement of the learned representations, and consider a reproducible experimental setup covering several data sets."
                },
                {
                    "title": "Nonlinear ICA Using Auxiliary Variables and Generalized Contrastive Learning",
                    "abstract": "Nonlinear ICA is a fundamental problem for unsupervised representation learning, emphasizing the capacity to recover the underlying latent variables generating the data (i.e., identifiability). Recently, the very first identifiability proofs for nonlinear ICA have been proposed, leveraging the temporal structure of the independent components. Here, we propose a general framework for nonlinear ICA, which, as a special case, can make use of temporal structure. It is based on augmenting the data by an auxiliary variable, such as the time index, the history of the time series, or any other available information. We propose to learn nonlinear ICA by discriminating between true augmented data, or data in which the auxiliary variable has been randomized. This enables the framework to be implemented algorithmically through logistic regression, possibly in a neural network. We provide a comprehensive proof of the identifiability of the model as well as the consistency of our estimation method. The approach not only provides a general theoretical framework combining and generalizing previously proposed nonlinear ICA models and algorithms, but also brings practical advantages."
                },
                {
                    "title": "NeuralNetwork-Viterbi: A Framework for Weakly Supervised Video Learning",
                    "abstract": "Video learning is an important task in computer vision and has experienced increasing interest over the recent years. Since even a small amount of videos easily comprises several million frames, methods that do not rely on a frame-level annotation are of special importance. In this work, we propose a novel learning algorithm with a Viterbi-based loss that allows for online and incremental learning of weakly annotated video data. We moreover show that explicit context and length modeling leads to huge improvements in video segmentation and labeling tasks and include these models into our framework. On several action segmentation benchmarks, we obtain an improvement of up to 10% compared to current state-of-the-art methods."
                },
                {
                    "title": "Weakly-Supervised Action Segmentation with Iterative Soft Boundary Assignment",
                    "abstract": "In this work, we address the task of weakly-supervised human action segmentation in long, untrimmed videos. Recent methods have relied on expensive learning models, such as Recurrent Neural Networks (RNN) and Hidden Markov Models (HMM). However, these methods suffer from expensive computational cost, thus are unable to be deployed in large scale. To overcome the limitations, the keys to our design are efficiency and scalability. We propose a novel action modeling framework, which consists of a new temporal convolutional network, named Temporal Convolutional Feature Pyramid Network (TCFPN), for predicting frame-wise action labels, and a novel training strategy for weakly-supervised sequence modeling, named Iterative Soft Boundary Assignment (ISBA), to align action sequences and update the network in an iterative fashion. The proposed framework is evaluated on two benchmark datasets, Breakfast and Hollywood Extended, with four different evaluation metrics. Extensive experimental results show that our methods achieve competitive or superior performance to state-of-the-art methods."
                },
                {
                    "title": "Disentangled Sequential Autoencoder",
                    "abstract": "We present a VAE architecture for encoding and generating high dimensional sequential data, such as video or audio. Our deep generative model learns a latent representation of the data which is split into a static and dynamic part, allowing us to approximately disentangle latent time-dependent features (dynamics) from features which are preserved over time (content). This architecture gives us partial control over generating content and dynamics by conditioning on either one of these sets of features. In our experiments on artificially generated cartoon video clips and voice recordings, we show that we can convert the content of a given sequence into another one by such content swapping. For audio, this allows us to convert a male speaker into a female speaker and vice versa, while for video we can separately manipulate shapes and dynamics. Furthermore, we give empirical evidence for the hypothesis that stochastic RNNs as latent state models are more efficient at compressing and generating long sequences than deterministic ones, which may be relevant for applications in video compression."
                },
                {
                    "title": "Nonlinear ICA of Temporally Dependent Stationary Sources",
                    "abstract": "We develop a nonlinear generalization of independent component analysis (ICA) or blind source separation, based on temporal dependencies (e.g. autocorrelations). We introduce a nonlinear generative model where the independent sources are assumed to be temporally dependent, non-Gaussian, and stationary, and we observe arbitrarily nonlinear mixtures of them. We develop a method for estimating the model (i.e. separating the sources) based on logistic regression in a neural network which learns to discriminate between a short temporal window of the data vs. a temporal window of temporally permuted data. We prove that the method estimates the sources for general smooth mixing nonlinearities, assuming the sources have sufficiently strong temporal dependencies, and these dependencies are in a certain way different from dependencies found in Gaussian processes. For Gaussian (and similar) sources, the method estimates the nonlinear part of the mixing. We thus provide the first rigorous and general proof of identifiability of nonlinear ICA for temporally dependent sources, together with a practical method for its estimation."
                },
                {
                    "title": "Weakly Supervised Action Learning with RNN Based Fine-to-Coarse Modeling",
                    "abstract": "We present an approach for weakly supervised learning of human actions. Given a set of videos and an ordered list of the occurring actions, the goal is to infer start and end frames of the related action classes within the video and to train the respective action classifiers without any need for hand labeled frame boundaries. To address this task, we propose a combination of a discriminative representation of subactions, modeled by a recurrent neural network, and a coarse probabilistic model to allow for a temporal alignment and inference over long sequences. While this system alone already generates good results, we show that the performance can be further improved by approximating the number of subactions to the characteristics of the different action classes. To this end, we adapt the number of subaction classes by iterating realignment and reestimation during training. The proposed system is evaluated on two benchmark datasets, the Breakfast and the Hollywood extended dataset, showing a competitive performance on various weak learning tasks such as temporal action segmentation and action alignment."
                },
                {
                    "title": "beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework",
                    "abstract": "an"
                },
                {
                    "title": "Causal Clustering for 1-Factor Measurement Models",
                    "abstract": "Many scientific research programs aim to learn the causal structure of real world phenomena. This learning problem is made more difficult when the target of study cannot be directly observed. One strategy commonly used by social scientists is to create measurable ``indicator'' variables that covary with the latent variables of interest. Before leveraging the indicator variables to learn about the latent variables, however, one needs a measurement model of the causal relations between the indicators and their corresponding latents. These measurement models are a special class of Bayesian networks. This paper addresses the problem of reliably inferring measurement models from measured indicators, without prior knowledge of the causal relations or the number of latent variables. We present a provably correct novel algorithm, FindOneFactorClusters (FOFC), for solving this inference problem. Compared to other state of the art algorithms, FOFC is faster, scales to larger sets of indicators, and is more reliable at small sample sizes. We also present the first correctness proofs for this problem that do not assume linearity or acyclicity among the latent variables."
                },
                {
                    "title": "Hierarchical Deep Temporal Models for Group Activity Recognition",
                    "abstract": "In this paper we present an approach for classifying the activity performed by a group of people in a video sequence. This problem of group activity recognition can be addressed by examining individual person actions and their relations. Temporal dynamics exist both at the level of individual person actions as well as at the level of group activity. Given a video sequence as input, methods can be developed to capture these dynamics at both person-level and group-level detail. We build a deep model to capture these dynamics based on LSTM (long short-term memory) models. In order to model both person-level and group-level dynamics, we present a 2-stage deep temporal model for the group activity recognition problem. In our approach, one LSTM model is designed to represent action dynamics of individual people in a video sequence and another LSTM model is designed to aggregate person-level information for group activity recognition. We collected a new dataset consisting of volleyball videos labeled with individual and group activities in order to evaluate our method. Experimental results on this new Volleyball Dataset and the standard benchmark Collective Activity Dataset demonstrate the efficacy of the proposed models."
                },
                {
                    "title": "Unsupervised Feature Extraction by Time-Contrastive Learning and Nonlinear ICA",
                    "abstract": "Nonlinear independent component analysis (ICA) provides an appealing framework for unsupervised feature learning, but the models proposed so far are not identifiable. Here, we first propose a new intuitive principle of unsupervised deep learning from time series which uses the nonstationary structure of the data. Our learning principle, time-contrastive learning (TCL), finds a representation which allows optimal discrimination of time segments (windows). Surprisingly, we show how TCL can be related to a nonlinear ICA model, when ICA is redefined to include temporal nonstationarities. In particular, we show that TCL combined with linear ICA estimates the nonlinear ICA model up to point-wise transformations of the sources, and this solution is unique --- thus providing the first identifiability result for nonlinear ICA which is rigorous, constructive, as well as very general."
                },
                {
                    "title": "Discovering Temporal Causal Relations from Subsampled Data",
                    "abstract": "Granger causal analysis has been an important tool for causal analysis for time series in various fields, including neuroscience and economics, and recently it has been extended to include instantaneous effects between the time series to explain the contemporaneous dependence in the residuals. In this paper, we assume that the time series at the true causal frequency follow the vector autoregressive model. We show that when the data resolution becomes lower due to subsampling, neither the original Granger causal analysis nor the extended one is able to discover the underlying causal relations. We then aim to answer the following question: can we estimate the temporal causal relations at the right causal frequency from the subsampled data? Traditionally this suffers from the identifiability problems: under the Gaussianity assumption of the data, the solutions are generally not unique. We prove that, however, if the noise terms are non-Gaussian, the underlying model for the highfrequency data is identifiable from subsampled data under mild conditions. We then propose an Expectation-Maximization (EM) approach and a variational inference approach to recover temporal causal relations from such subsampled data. Experimental results on both simulated and real data are reported to illustrate the performance of the proposed approaches."
                },
                {
                    "title": "Marginal likelihood and model selection for Gaussian latent tree and forest models",
                    "abstract": "Gaussian latent tree models, or more generally, Gaussian latent forest models have Fisher-information matrices that become singular along interesting submodels, namely, models that correspond to subforests. For these singularities, we compute the real log-canonical thresholds (also known as stochastic complexities or learning coefficients) that quantify the large-sample behavior of the marginal likelihood in Bayesian inference. This provides the information needed for a recently introduced generalization of the Bayesian information criterion. Our mathematical developments treat the general setting of Laplace integrals whose phase functions are sums of squared differences between monomials and constants. We clarify how in this case real log-canonical thresholds can be computed using polyhedral geometry, and we show how to apply the general theory to the Laplace integrals associated with Gaussian latent tree and forest models. In simulations and a data example, we demonstrate how the mathematical knowledge can be applied in model selection."
                },
                {
                    "title": "Testing for Granger Causality with Mixed Frequency Data",
                    "abstract": "We develop Granger causality tests that apply directly to data sampled at different frequencies. We show that taking advantage of mixed frequency data allows us to better recover causal relationships when compared to the conventional common low frequency approach. We also show that the new causality tests have higher local asymptotic power as well as more power in finite samples compared to conventional tests. In an empirical application involving U.S. macroeconomic indicators, we show that the mixed frequency approach and the low frequency approach produce very different causal implications, with the former yielding more intuitively appealing result."
                },
                {
                    "title": "Causality : statistical perspectives and applications",
                    "abstract": "A state of the art volume on statistical causality\n\n\"Causality: Statistical Perspectives and Applications\" presents a wide?ranging collection of seminal contributions by renowned experts in the field, providing a thorough treatment of all aspects of statistical causality. It covers the various formalisms in current use, methods for applying them to specific problems, and the special requirements of a range of examples from medicine, biology and economics to political science.\n\nThis book:\n\n(1) provides a clear account and comparison of formal languages, concepts and models for statistical causality, and \n\n(2) addresses examples from medicine, biology, economics and political science to aid the reader's understanding, and \n\n(3) is authored by leading experts in their field, and \n\n(4) is written in an accessible style. \n\nPostgraduates, professional statisticians and researchers in academia and industry will benefit from this book."
                },
                {
                    "title": "Learning Latent Tree Graphical Models",
                    "abstract": "We study the problem of learning a latent tree graphical model where samples are available only from a subset of variables. We propose two consistent and computationally efficient algorithms for learning minimal latent trees, that is, trees without any redundant hidden nodes. Unlike many existing methods, the observed nodes (or variables) are not constrained to be leaf nodes. Our algorithms can be applied to both discrete and Gaussian random variables and our learned models are such that all the observed and latent variables have the same domain (state space). Our first algorithm, recursive grouping, builds the latent tree recursively by identifying sibling groups using so-called information distances. One of the main contributions of this work is our second algorithm, which we refer to as CLGrouping. CLGrouping starts with a pre-processing procedure in which a tree over the observed variables is constructed. This global step groups the observed nodes that are likely to be close to each other in the true latent tree, thereby guiding subsequent recursive grouping (or equivalent procedures such as neighbor-joining) on much smaller subsets of variables. This results in more accurate and efficient learning of latent trees. We also present regularized versions of our algorithms that learn latent tree approximations of arbitrary distributions. We compare the proposed algorithms to other methods by performing extensive numerical experiments on various latent tree graphical models such as hidden Markov models and star graphs. In addition, we demonstrate the applicability of our methods on real-world data sets by modeling the dependency structure of monthly stock returns in the S&P index and of the words in the 20 newsgroups data set."
                },
                {
                    "title": "Estimation of a Structural Vector Autoregression Model Using Non-Gaussianity",
                    "abstract": "Analysis of causal effects between continuous-valued variables typically uses either autoregressive models or structural equation models with instantaneous effects. Estimation of Gaussian, linear structural equation models poses serious identifiability problems, which is why it was recently proposed to use non-Gaussian models. Here, we show how to combine the non-Gaussian instantaneous model with autoregressive models. This is effectively what is called a structural vector autoregression (SVAR) model, and thus our work contributes to the long-standing problem of how to estimate SVAR's. We show that such a non-Gaussian model is identifiable without prior knowledge of network structure. We propose computationally efficient methods for estimating the model, as well as methods to assess the significance of the causal influences. The model is successfully applied on financial and brain imaging data."
                },
                {
                    "title": "Causal Modelling and Brain Connectivity in Functional Magnetic Resonance Imaging",
                    "abstract": "Recent advances in data analysis and modeling allow the use of fMRI data to ask not just which brain regions are involved in various cognitive and perceptual tasks, but also how they communicate with each other. Karl Friston examines two different state-of-the-art approaches to modeling brain connectivity using neuroimaging."
                },
                {
                    "title": "Learning the Structure of Linear Latent Variable Models",
                    "abstract": "We describe anytime search procedures that (1) find disjoint subsets of recorded variables for which the members of each subset are d-separated by a single common unrecorded cause, if such exists; (2) return information about the causal relations among the latent factors so identified. We prove the procedure is point-wise consistent assuming (a) the causal relations can be represented by a directed acyclic graph (DAG) satisfying the Markov Assumption and the Faithfulness Assumption; (b) unrecorded variables are not caused by recorded variables; and (c) dependencies are linear. We compare the procedure with standard approaches over a variety of simulated structures and sample sizes, and illustrate its practical value with brief studies of social science data sets. Finally, we consider generalizations for non-linear systems."
                },
                {
                    "title": "Hierarchical latent class models for cluster analysis",
                    "abstract": "Latent class models are used for cluster analysis of categorical data. Underlying such a model is the assumption that the observed variables are mutually independent given the class variable. A serious problem with the use of latent class models, known as local dependence, is that this assumption is often untrue. In this paper we propose hierarchical latent class models as a framework where the local dependence problem can be addressed in a principled manner. We develop a search-based algorithm for learning hierarchical latent class models from data. The algorithm is evaluated using both synthetic and real-world data."
                },
                {
                    "title": "Identification of Linear Non-Gaussian Latent Hierarchical Structure",
                    "abstract": "Traditional causal discovery methods mainly focus on estimating causal relations among measured variables, but in many real-world problems, such as questionnaire-based psychometric studies, measured variables are generated by latent variables that are causally related. Accordingly, this paper investigates the problem of discovering the hidden causal variables and estimating the causal structure, including both the causal relations among latent variables and those be-tween latent and measured variables. We relax the frequently-used measurement assumption and allow the children of latent variables to be latent as well, and hence deal with a specific type of latent hierarchical causal structure. In particular, we define a minimal latent hierarchical structure and show that for linear non-Gaussian models with the minimal latent hierarchical structure, the whole structure is identifiable from only the measured variables. Moreover, we develop a principled method to identify the structure by testing for Generalized Independent Noise (GIN) conditions in specific ways. Experimental results on both synthetic and real-world data show the effectiveness of the proposed approach."
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                },
                {
                    "title": "Triad Constraints for Learning Causal Structure of Latent Variables",
                    "abstract": "Learning causal structure from observational data has attracted much attention, and it is notoriously challenging to find the underlying structure in the presence of confounders (hidden direct common causes of two variables). In this paper, by properly leveraging the non-Gaussianity of the data, we propose to estimate the structure over latent variables with the so-called Triad constraints: we design a form of \"pseudo-residual\" from three variables, and show that when causal relations are linear and noise terms are non-Gaussian, the causal direction between the latent variables for the three observed variables is identifiable by checking a certain kind of independence relationship. In other words, the Triad constraints help us to locate latent confounders and determine the causal direction between them. This goes far beyond the Tetrad constraints and reveals more information about the underlying structure from non-Gaussian data. Finally, based on the Triad constraints, we develop a two-step algorithm to learn the causal structure corresponding to measurement models. Experimental results on both synthetic and real data demonstrate the effectiveness and reliability of our method."
                },
                {
                    "title": "Testing for causality: a personal viewpoint",
                    "abstract": "A general definition of causality is introduced and then specialized to become operational. By considering simple examples a number of advantages, and also difficulties, with the definition are discussed. Tests based on the definitions are then considered and the use of post-sample data emphasized, rather than relying on the same data to fit a model and use it to test causality. It is suggested that a bayesian viewpoint should be taken in interpreting the results of these tests. Finally, the results of a study relating advertising and consumption are briefly presented."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we establish identifiability of nonstationary nonlinear Independent Component Analysis (ICA) for general sequence data without prior knowledge of domain variables?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of causal learning, particularly in understanding complex temporal relationships in real-world data where domain information is often unavailable. By addressing this question, we can enhance the ability to model and interpret nonstationary processes, which has significant implications for various applications such as video understanding, action segmentation, and recognition tasks. This research could lead to more robust methodologies for causal inference in dynamic environments, ultimately influencing future studies and practical applications in machine learning and data analysis.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexities of nonstationary data, where distribution shifts can occur without observable domain indices. Naive approaches may fail because they often rely on stationary assumptions or require prior knowledge of domain variables, which is not feasible in many real-world scenarios. Additionally, the task of accurately modeling time-delayed causal relationships in latent spaces adds another layer of difficulty. Overcoming these technical and theoretical obstacles requires innovative methods to identify and cluster transitions effectively, which is not straightforward.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on stationary conditions or nonstationary scenarios with known domain indices, limiting their applicability to general time series data. The reliance on the Markov property in existing methods has also hindered the ability to capture arbitrary nonstationary variations. Barriers such as the lack of a comprehensive framework for identifying distribution shifts without prior knowledge of domain variables have prevented this problem from being adequately addressed. Our approach differs by proposing a novel framework, CtrlNS, that leverages transition clustering to identify domain shifts and latent dynamics without requiring prior domain knowledge.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Causal Temporal Representation Learning with Nonstationary Sparse Transition (CtrlNS), involves two key components: (1) constraining the complexity of the transition function to identify domain shifts, and (2) learning latent variables using conditional independence constraints based on the identified domain variables. We will utilize a variational autoencoder (VAE) framework to jointly optimize these processes. The expected outcomes include a theoretical identifiability result for nonstationary"
            }
        },
        "author_data": {
            "7049286b-da38-4f5e-a8ba-c8b6c7256fa6": {
                "pk": "7049286b-da38-4f5e-a8ba-c8b6c7256fa6",
                "name": "Xiangchen Song",
                "collaborators": [
                    "Kun Zhang",
                    "Guangyi Chen",
                    "Jieyu Zhang",
                    "Jiawei Han",
                    "Yujing Wang",
                    "Yaming Yang",
                    "Jing Bai",
                    "Weiran Yao",
                    "Yifan Shen",
                    "Yewen Fan"
                ],
                "domain": [
                    "Taxonomy Completion",
                    "Causal Inference",
                    "Graph Neural Network",
                    "Recommendation Systems"
                ],
                "publications": [
                    {
                        "title": "TaxoEnrich: Self-Supervised Taxonomy Completion via Structure-Semantic Representations",
                        "abstract": "Taxonomies are fundamental to many real-world applications in various domains, serving as structural representations of knowledge. To deal with the increasing volume of new concepts needed to be organized as taxonomies, researchers turn to automatically completion of an existing taxonomy with new concepts. In this paper, we propose TaxoEnrich, a new taxonomy completion framework, which effectively leverages both semantic features and structural information in the existing taxonomy and offers a better representation of candidate position to boost the performance of taxonomy completion. Specifically, TaxoEnrich consists of four components: (1) taxonomy-contextualized embedding which incorporates both semantic meanings of concept and taxonomic relations based on powerful pretrained language models; (2) a taxonomy-aware sequential encoder which learns candidate position representations by encoding the structural information of taxonomy; (3) a query-aware sibling encoder which adaptively aggregates candidate siblings to augment candidate position representations based on their importance to the query-position matching; (4) a query-position matching model which extends existing work with our new candidate position representations. Extensive experiments on four large real-world datasets from different domains show that \\TaxoEnrich achieves the best performance among all evaluation metrics and outperforms previous state-of-the-art methods by a large margin."
                    },
                    {
                        "title": "Calibration-then-Calculation: A Variance Reduced Metric Framework in Deep Click-Through Rate Prediction Models",
                        "abstract": "The adoption of deep learning across various fields has been extensive, yet there is a lack of focus on evaluating the performance of deep learning pipelines. Typically, with the increased use of large datasets and complex models, the training process is run only once and the result is compared to previous benchmarks. This practice can lead to imprecise comparisons due to the variance in neural network evaluation metrics, which stems from the inherent randomness in the training process. Traditional solutions, such as running the training process multiple times, are often infeasible due to computational constraints. In this paper, we introduce a novel metric framework, the Calibrated Loss Metric, designed to address this issue by reducing the variance present in its conventional counterpart. Consequently, this new metric enhances the accuracy in detecting effective modeling improvements. Our approach is substantiated by theoretical justifications and extensive experimental validations within the context of Deep Click-Through Rate Prediction Models."
                    },
                    {
                        "title": "Who Should Go First? A Self-Supervised Concept Sorting Model for Improving Taxonomy Expansion",
                        "abstract": "Taxonomies have been widely used in various machine learning and text mining systems to organize knowledge and facilitate downstream tasks. One critical challenge is that, as data and business scope grow in real applications, existing taxonomies need to be expanded to incorporate new concepts. Previous works on taxonomy expansion process the new concepts independently and simultaneously, ignoring the potential relationships among them and the appropriate order of inserting operations. However, in reality, the new concepts tend to be mutually correlated and form local hypernym-hyponym structures. In such a scenario, ignoring the dependencies of new concepts and the order of insertion may trigger error propagation. For example, existing taxonomy expansion systems may insert hyponyms to existing taxonomies before their hypernym, leading to sub-optimal expanded taxonomies. To complement existing taxonomy expansion systems, we propose TaxoOrder, a novel self-supervised framework that simultaneously discovers the local hypernym-hyponym structure among new concepts and decides the order of insertion. TaxoOrder can be directly plugged into any taxonomy expansion system and improve the quality of expanded taxonomies. Experiments on the real-world dataset validate the effectiveness of TaxoOrder to enhance taxonomy expansion systems, leading to better-resulting taxonomies with comparison to baselines under various evaluation metrics."
                    },
                    {
                        "title": "Comprehensive Named Entity Recognition on CORD-19 with Distant or Weak Supervision",
                        "abstract": "We created this CORD-NER dataset with comprehensive named entity recognition (NER) on the COVID-19 Open Research Dataset Challenge (CORD-19) corpus (2020-03-13). This CORD-NER dataset covers 75 fine-grained entity types: In addition to the common biomedical entity types (e.g., genes, chemicals and diseases), it covers many new entity types related explicitly to the COVID-19 studies (e.g., coronaviruses, viral proteins, evolution, materials, substrates and immune responses), which may benefit research on COVID-19 related virus, spreading mechanisms, and potential vaccines. CORD-NER annotation is a combination of four sources with different NER methods. The quality of CORD-NER annotation surpasses SciSpacy (over 10% higher on the F1 score based on a sample set of documents), a fully supervised BioNER tool. Moreover, CORD-NER supports incrementally adding new documents as well as adding new entity types when needed by adding dozens of seeds as the input examples. We will constantly update CORD-NER based on the incremental updates of the CORD-19 corpus and the improvement of our system."
                    },
                    {
                        "title": "BiTe-GCN: A New GCN Architecture via BidirectionalConvolution of Topology and Features on Text-Rich Networks",
                        "abstract": "Graph convolutional networks (GCNs), aiming to integrate high-order neighborhood information through stacked graph convolution layers, have demonstrated remarkable power in many network analysis tasks. However, topological limitations, including over-smoothing and local topology homophily, limit its capability to represent networks. Existing studies only perform feature convolution on network topology, which inevitably introduces unbalance between topology and features. Considering that in real world, the information network consists of not only the node-level citation information but also the local text-sequence information. We propose BiTe-GCN, a novel GCN architecture with bidirectional convolution of both topology and features on text-rich networks to solve these limitations. We first transform the original text-rich network into an augmented bi-typed heterogeneous network, capturing both the global node-level information and the local text-sequence information from texts. We then introduce discriminative convolution mechanisms to performs convolutions of both topology and features simultaneously. Extensive experiments on text-rich networks demonstrate that our new architecture outperforms state-of-the-art by a breakout improvement. Moreover, this architecture can also be applied to several e-commerce searching scenes such as JD searching. The experiments on the JD dataset validate the superiority of the proposed architecture over the related methods."
                    },
                    {
                        "title": "Taxonomy Completion via Triplet Matching Network",
                        "abstract": "Automatically constructing taxonomy finds many applications in e-commerce and web search. One critical challenge is as data and business scope grow in real applications, new concepts are emerging and needed to be added to the existing taxonomy. Previous approaches focus on the taxonomy expansion, i.e. finding an appropriate hypernym concept from the taxonomy for a new query concept. In this paper, we formulate a new task, \"taxonomy completion\", by discovering both the hypernym and hyponym concepts for a query. We propose Triplet Matching Network (TMN), to find the appropriate <hypernym, hyponym> pairs for a given query concept. TMN consists of one primal scorer and multiple auxiliary scorers. These auxiliary scorers capture various fine-grained signals (e.g., query to hypernym or query to hyponym semantics), and the primal scorer makes a holistic prediction on <query, hypernym, hyponym> triplet based on the internal feature representations of all auxiliary scorers. Also, an innovative channel-wise gating mechanism that retains task-specific information in concept representations is introduced to further boost model performance. Experiments on four real-world large-scale datasets show that TMN achieves the best performance on both taxonomy completion task and the previous taxonomy expansion task, outperforming existing methods."
                    },
                    {
                        "title": "Creating Training Sets via Weak Indirect Supervision",
                        "abstract": "Creating labeled training sets has become one of the major roadblocks in machine learning. To address this, recent \\emph{Weak Supervision (WS)} frameworks synthesize training labels from multiple potentially noisy supervision sources. However, existing frameworks are restricted to supervision sources that share the same output space as the target task. To extend the scope of usable sources, we formulate Weak Indirect Supervision (WIS), a new research problem for automatically synthesizing training labels based on indirect supervision sources that have different output label spaces. To overcome the challenge of mismatched output spaces, we develop a probabilistic modeling approach, PLRM, which uses user-provided label relations to model and leverage indirect supervision sources. Moreover, we provide a theoretically-principled test of the distinguishability of PLRM for unseen labels, along with a generalization bound. On both image and text classification tasks as well as an industrial advertising application, we demonstrate the advantages of PLRM by outperforming baselines by a margin of 2%-9%."
                    },
                    {
                        "title": "PLOT: Prompt Learning with Optimal Transport for Vision-Language Models",
                        "abstract": "With the increasing attention to large vision-language models such as CLIP, there has been a significant amount of effort dedicated to building efficient prompts. Unlike conventional methods of only learning one single prompt, we propose to learn multiple comprehensive prompts to describe diverse characteristics of categories such as intrinsic attributes or extrinsic contexts. However, directly matching each prompt to the same visual feature is problematic, as it pushes the prompts to converge to one point. To solve this problem, we propose to apply optimal transport to match the vision and text modalities. Specifically, we first model images and the categories with visual and textual feature sets. Then, we apply a two-stage optimization strategy to learn the prompts. In the inner loop, we optimize the optimal transport distance to align visual features and prompts by the Sinkhorn algorithm, while in the outer loop, we learn the prompts by this distance from the supervised data. Extensive experiments are conducted on the few-shot recognition task and the improvement demonstrates the superiority of our method. The code is available at https://github.com/CHENGY12/PLOT."
                    },
                    {
                        "title": "Estimating Treatment Effects from Irregular Time Series Observations with Hidden Confounders",
                        "abstract": "Causal analysis for time series data, in particular estimating individualized treatment effect (ITE), is a key task in many real-world applications, such as finance, retail, healthcare, etc. Real-world time series can include large-scale, irregular, and intermittent time series observations, raising significant challenges to existing work attempting to estimate treatment effects. Specifically, the existence of hidden confounders can lead to biased treatment estimates and complicate the causal inference process. In particular, anomaly hidden confounders which exceed the typical range can lead to high variance estimates. Moreover, in continuous time settings with irregular samples, it is challenging to directly handle the dynamics of causality. In this paper, we leverage recent advances in Lipschitz regularization and neural controlled differential equations (CDE) to develop an effective and scalable solution, namely LipCDE, to address the above challenges. LipCDE can directly model the dynamic causal relationships between historical data and outcomes with irregular samples by considering the boundary of hidden confounders given by Lipschitz-constrained neural networks. Furthermore, we conduct extensive experiments on both synthetic and real-world datasets to demonstrate the effectiveness and scalability of LipCDE."
                    },
                    {
                        "title": "Attentive Knowledge-aware Graph Convolutional Networks with Collaborative Guidance for Personalized Recommendation",
                        "abstract": "To alleviate data sparsity and cold-start problems of traditional recommender systems (RSs), incorporating knowledge graphs (KGs) to supplement auxiliary information has attracted considerable attention recently. However, simply integrating KGs in current KG-based RS models is not necessarily a guarantee to improve the recommendation performance, which may even weaken the holistic model capability. This is because the construction of these KGs is independent of the collection of historical user-item interactions; hence, information in these KGs may not always be helpful for recommendation to all users.   In this paper, we propose attentive Knowledge-aware Graph convolutional networks with Collaborative Guidance for personalized Recommendation (CG-KGR). CG-KGR is a novel knowledge-aware recommendation model that enables ample and coherent learning of KGs and user-item interactions, via our proposed Collaborative Guidance Mechanism. Specifically, CG-KGR first encapsulates historical interactions to interactive information summarization. Then CG-KGR utilizes it as guidance to extract information out of KGs, which eventually provides more precise personalized recommendation. We conduct extensive experiments on four real-world datasets over two recommendation tasks, i.e., Top-K recommendation and Click-Through rate (CTR) prediction. The experimental results show that the CG-KGR model significantly outperforms recent state-of-the-art models by 1.4-27.0% in terms of Recall metric on Top-K recommendation."
                    },
                    {
                        "title": "Temporally Disentangled Representation Learning under Unknown Nonstationarity",
                        "abstract": "In unsupervised causal representation learning for sequential data with time-delayed latent causal influences, strong identifiability results for the disentanglement of causally-related latent variables have been established in stationary settings by leveraging temporal structure. However, in nonstationary setting, existing work only partially addressed the problem by either utilizing observed auxiliary variables (e.g., class labels and/or domain indexes) as side information or assuming simplified latent causal dynamics. Both constrain the method to a limited range of scenarios. In this study, we further explored the Markov Assumption under time-delayed causally related process in nonstationary setting and showed that under mild conditions, the independent latent components can be recovered from their nonlinear mixture up to a permutation and a component-wise transformation, without the observation of auxiliary variables. We then introduce NCTRL, a principled estimation framework, to reconstruct time-delayed latent causal variables and identify their relations from measured sequential data only. Empirical evaluations demonstrated the reliable identification of time-delayed latent causal influences, with our methodology substantially outperforming existing baselines that fail to exploit the nonstationarity adequately and then, consequently, cannot distinguish distribution shifts."
                    },
                    {
                        "title": "A Versatile Causal Discovery Framework to Allow Causally-Related Hidden Variables",
                        "abstract": "Most existing causal discovery methods rely on the assumption of no latent confounders, limiting their applicability in solving real-life problems. In this paper, we introduce a novel, versatile framework for causal discovery that accommodates the presence of causally-related hidden variables almost everywhere in the causal network (for instance, they can be effects of observed variables), based on rank information of covariance matrix over observed variables. We start by investigating the efficacy of rank in comparison to conditional independence and, theoretically, establish necessary and sufficient conditions for the identifiability of certain latent structural patterns. Furthermore, we develop a Rank-based Latent Causal Discovery algorithm, RLCD, that can efficiently locate hidden variables, determine their cardinalities, and discover the entire causal structure over both measured and hidden ones. We also show that, under certain graphical conditions, RLCD correctly identifies the Markov Equivalence Class of the whole latent causal graph asymptotically. Experimental results on both synthetic and real-world personality data sets demonstrate the efficacy of the proposed approach in finite-sample cases."
                    },
                    {
                        "title": "When and How: Learning Identifiable Latent States for Nonstationary Time Series Forecasting",
                        "abstract": "Temporal distribution shifts are ubiquitous in time series data. One of the most popular methods assumes that the temporal distribution shift occurs uniformly to disentangle the stationary and nonstationary dependencies. But this assumption is difficult to meet, as we do not know when the distribution shifts occur. To solve this problem, we propose to learn IDentifiable latEnt stAtes (IDEA) to detect when the distribution shifts occur. Beyond that, we further disentangle the stationary and nonstationary latent states via sufficient observation assumption to learn how the latent states change. Specifically, we formalize the causal process with environment-irrelated stationary and environment-related nonstationary variables. Under mild conditions, we show that latent environments and stationary/nonstationary variables are identifiable. Based on these theories, we devise the IDEA model, which incorporates an autoregressive hidden Markov model to estimate latent environments and modular prior networks to identify latent states. The IDEA model outperforms several latest nonstationary forecasting methods on various benchmark datasets, highlighting its advantages in real-world scenarios."
                    },
                    {
                        "title": "CaRiNG: Learning Temporal Causal Representation under Non-Invertible Generation Process",
                        "abstract": "Identifying the underlying time-delayed latent causal processes in sequential data is vital for grasping temporal dynamics and making downstream reasoning. While some recent methods can robustly identify these latent causal variables, they rely on strict assumptions about the invertible generation process from latent variables to observed data. However, these assumptions are often hard to satisfy in real-world applications containing information loss. For instance, the visual perception process translates a 3D space into 2D images, or the phenomenon of persistence of vision incorporates historical data into current perceptions. To address this challenge, we establish an identifiability theory that allows for the recovery of independent latent components even when they come from a nonlinear and non-invertible mix. Using this theory as a foundation, we propose a principled approach, CaRiNG, to learn the CAusal RepresentatIon of Non-invertible Generative temporal data with identifiability guarantees. Specifically, we utilize temporal context to recover lost latent information and apply the conditions in our theory to guide the training process. Through experiments conducted on synthetic datasets, we validate that our CaRiNG method reliably identifies the causal process, even when the generation process is non-invertible. Moreover, we demonstrate that our approach considerably improves temporal understanding and reasoning in practical applications."
                    },
                    {
                        "title": "On the Identification of Temporally Causal Representation with Instantaneous Dependence",
                        "abstract": "Temporally causal representation learning aims to identify the latent causal process from time series observations, but most methods require the assumption that the latent causal processes do not have instantaneous relations. Although some recent methods achieve identifiability in the instantaneous causality case, they require either interventions on the latent variables or grouping of the observations, which are in general difficult to obtain in real-world scenarios. To fill this gap, we propose an \\textbf{ID}entification framework for instantane\\textbf{O}us \\textbf{L}atent dynamics (\\textbf{IDOL}) by imposing a sparse influence constraint that the latent causal processes have sparse time-delayed and instantaneous relations. Specifically, we establish identifiability results of the latent causal process based on sufficient variability and the sparse influence constraint by employing contextual information of time series data. Based on these theories, we incorporate a temporally variational inference architecture to estimate the latent variables and a gradient-based sparsity regularization to identify the latent causal process. Experimental results on simulation datasets illustrate that our method can identify the latent causal process. Furthermore, evaluations on multiple human motion forecasting benchmarks with instantaneous dependencies indicate the effectiveness of our method in real-world settings."
                    },
                    {
                        "title": "Learning Multi-granularity User Intent Unit for Session-based Recommendation",
                        "abstract": "Session-based recommendation aims to predict a user's next action based on previous actions in the current session. The major challenge is to capture authentic and complete user preferences in the entire session. Recent work utilizes graph structure to represent the entire session and adopts Graph Neural Network to encode session information. This modeling choice has been proved to be effective and achieved remarkable results. However, most of the existing studies only consider each item within the session independently and do not capture session semantics from a high-level perspective. Such limitation often leads to severe information loss and increases the difficulty of capturing long-range dependencies within a session. Intuitively, compared with individual items, a session snippet, i.e., a group of locally consecutive items, is able to provide supplemental user intents which are hardly captured by existing methods. In this work, we propose to learn multi-granularity consecutive user intent unit to improve the recommendation performance. Specifically, we creatively propose Multi-granularity Intent Heterogeneous Session Graph which captures the interactions between different granularity intent units and relieves the burden of long-dependency. Moreover, we propose the Intent Fusion Ranking module to compose the recommendation results from various granularity user intents. Compared with current methods that only leverage intents from individual items, IFR benefits from different granularity user intents to generate more accurate and comprehensive session representation, thus eventually boosting recommendation performance. We conduct extensive experiments on five session-based recommendation datasets and the results demonstrate the effectiveness of our method."
                    },
                    {
                        "title": "A Randomized Controlled Trial on Anonymizing Reviewers to Each Other in Peer Review Discussions",
                        "abstract": "Peer review often involves reviewers submitting their independent reviews, followed by a discussion among reviewers of each paper. A question among policymakers is whether the reviewers of a paper should be anonymous to each other during the discussion. We shed light on this by conducting a randomized controlled trial at the UAI 2022 conference. We randomly split the reviewers and papers into two conditions--one with anonymous discussions and the other with non-anonymous discussions, and conduct an anonymous survey of all reviewers, to address the following questions: 1. Do reviewers discuss more in one of the conditions? Marginally more in anonymous (n = 2281, p = 0.051). 2. Does seniority have more influence on final decisions when non-anonymous? Yes, the decisions are closer to senior reviewers' scores in the non-anonymous condition than in anonymous (n = 484, p = 0.04). 3. Are reviewers more polite in one of the conditions? No significant difference in politeness of reviewers' text-based responses (n = 1125, p = 0.72). 4. Do reviewers' self-reported experiences differ across the two conditions? No significant difference for each of the five questions asked (n = 132 and p > 0.3). 5. Do reviewers prefer one condition over the other? Yes, there is a weak preference for anonymous discussions (n = 159 and Cohen's d= 0.25). 6. What do reviewers consider important to make policy on anonymity among reviewers? Reviewers' feeling of safety in expressing their opinions was rated most important, while polite communication among reviewers was rated least important (n = 159). 7. Have reviewers experienced dishonest behavior due to non-anonymity in discussions? Yes, roughly 7% of respondents answered affirmatively (n = 167). Overall, this experiment reveals evidence supporting an anonymous discussion setup in the peer-review process, in terms of the evaluation criteria considered."
                    }
                ]
            },
            "090de8d9-8d2c-4677-b7cd-b2123362a1a8": {
                "pk": "090de8d9-8d2c-4677-b7cd-b2123362a1a8",
                "name": "Zijian Li",
                "collaborators": [
                    "Zhihui Wang",
                    "Xun Jian",
                    "Xiang Lian",
                    "Lei Chen"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Differential Privacy",
                    "Graph Similarity",
                    "Data Generation"
                ],
                "publications": [
                    {
                        "title": "A self-attention-based differentially private tabular GAN with high data utility",
                        "abstract": "Generative Adversarial Networks (GANs) have become a ubiquitous technology for data generation, with their prowess in image generation being well-established. However, their application in generating tabular data has been less than ideal. Furthermore, attempting to incorporate differential privacy technology into these frameworks has often resulted in a degradation of data utility. To tackle these challenges, this paper introduces DP-SACTGAN, a novel Conditional Generative Adversarial Network (CGAN) framework for differentially private tabular data generation, aiming to surmount these obstacles. Experimental findings demonstrate that DP-SACTGAN not only accurately models the distribution of the original data but also effectively satisfies the requirements of differential privacy."
                    },
                    {
                        "title": "An Efficient Probabilistic Approach for Graph Similarity Search",
                        "abstract": "Graph similarity search is a common and fundamental operation in graph databases. One of the most popular graph similarity measures is the Graph Edit Distance (GED) mainly because of its broad applicability and high interpretability. Despite its prevalence, exact GED computation is proved to be NP-hard, which could result in unsatisfactory computational efficiency on large graphs. However, exactly accurate search results are usually unnecessary for real-world applications especially when the responsiveness is far more important than the accuracy. Thus, in this paper, we propose a novel probabilistic approach to efficiently estimate GED, which is further leveraged for the graph similarity search. Specifically, we first take branches as elementary structures in graphs, and introduce a novel graph similarity measure by comparing branches between graphs, i.e., Graph Branch Distance (GBD), which can be efficiently calculated in polynomial time. Then, we formulate the relationship between GED and GBD by considering branch variations as the result ascribed to graph edit operations, and model this process by probabilistic approaches. By applying our model, the GED between any two graphs can be efficiently estimated by their GBD, and these estimations are finally utilized in the graph similarity search. Extensive experiments show that our approach has better accuracy, efficiency and scalability than other comparable methods in the graph similarity search over real and synthetic data sets."
                    }
                ]
            },
            "66090306-62ca-41c6-abd9-a03f2aa082f5": {
                "pk": "66090306-62ca-41c6-abd9-a03f2aa082f5",
                "name": "Guangyi Chen",
                "collaborators": [
                    "Kun Zhang",
                    "Jiwen Lu",
                    "Junlong Li",
                    "Jie Zhou",
                    "Weiran Yao",
                    "Zhenhao Chen",
                    "Shunxing Fan",
                    "Yongming Rao",
                    "Nuoxing Zhou",
                    "Liangliang Ren"
                ],
                "domain": [
                    "Trajectory Prediction",
                    "Causal Inference",
                    "Graph Neural Network",
                    "Unsupervised Learning"
                ],
                "publications": [
                    {
                        "title": "Unsupervised Sampling Promoting for Stochastic Human Trajectory Prediction",
                        "abstract": "The indeterminate nature of human motion requires trajectory prediction systems to use a probabilistic model to formulate the multi-modality phenomenon and infer a finite set of future trajectories. However, the inference processes of most existing methods rely on Monte Carlo random sampling, which is insufficient to cover the realistic paths with finite samples, due to the long tail effect of the predicted distribution. To promote the sampling process of stochastic prediction, we propose a novel method, called BOsampler, to adaptively mine potential paths with Bayesian optimization in an unsupervised manner, as a sequential design strategy in which new prediction is dependent on the previously drawn samples. Specifically, we model the trajectory sampling as a Gaussian process and construct an acquisition function to measure the potential sampling value. This acquisition function applies the original distribution as prior and encourages exploring paths in the long-tail region. This sampling method can be integrated with existing stochastic predictive models without retraining. Experimental results on various baseline methods demonstrate the effectiveness of our method."
                    },
                    {
                        "title": "Human Trajectory Prediction via Counterfactual Analysis",
                        "abstract": "Forecasting human trajectories in complex dynamic environments plays a critical role in autonomous vehicles and intelligent robots. Most existing methods learn to predict future trajectories by behavior clues from history trajectories and interaction clues from environments. However, the inherent bias between training and deployment environments is ignored. Hence, we propose a counterfactual analysis method for human trajectory prediction to investigate the causality between the predicted trajectories and input clues and alleviate the negative effects brought by environment bias. We first build a causal graph for trajectory forecasting with history trajectory, future trajectory, and the environment interactions. Then, we cut off the inference from environment to trajectory by constructing the counterfactual intervention on the trajectory itself. Finally, we compare the factual and counterfactual trajectory clues to alleviate the effects of environment bias and highlight the trajectory clues. Our counterfactual analysis is a plug-and-play module that can be applied to any baseline prediction methods including RNN- and CNN-based ones. We show that our method achieves consistent improvement for different baselines and obtains the state-of-the-art results on public pedestrian trajectory forecasting benchmarks."
                    },
                    {
                        "title": "Counterfactual Attention Learning for Fine-Grained Visual Categorization and Re-identification",
                        "abstract": "Attention mechanism has demonstrated great potential in fine-grained visual recognition tasks. In this paper, we present a counterfactual attention learning method to learn more effective attention based on causal inference. Unlike most existing methods that learn visual attention based on conventional likelihood, we propose to learn the attention with counterfactual causality, which provides a tool to measure the attention quality and a powerful supervisory signal to guide the learning process. Specifically, we analyze the effect of the learned visual attention on network prediction through counterfactual intervention and maximize the effect to encourage the network to learn more useful attention for fine-grained image recognition. Empirically, we evaluate our method on a wide range of fine-grained recognition tasks where attention plays a crucial role, including fine-grained image categorization, person re-identification, and vehicle re-identification. The consistent improvement on all benchmarks demonstrates the effectiveness of our method. Code is available at https://github.com/raoyongming/CAL"
                    },
                    {
                        "title": "Learning Latent Causal Dynamics",
                        "abstract": "One critical challenge of time-series modeling is how to learn and quickly correct the model under unknown distribution shifts. In this work, we propose a principled framework, called LiLY, to first recover time-delayed latent causal variables and identify their relations from measured temporal data under different distribution shifts. The correction step is then formulated as learning the low-dimensional change factors with a few samples from the new environment, leveraging the identified causal structure. Specifically, the framework factorizes unknown distribution shifts into transition distribution changes caused by fixed dynamics and time-varying latent causal relations, and by global changes in observation. We establish the identifiability theories of nonparametric latent causal dynamics from their nonlinear mixtures under fixed dynamics and under changes. Through experiments, we show that time-delayed latent causal influences are reliably identified from observed variables under different distribution changes. By exploiting this modular representation of changes, we can efficiently learn to correct the model under unknown distribution shifts with only a few samples."
                    },
                    {
                        "title": "Temporally Disentangled Representation Learning",
                        "abstract": "Recently in the field of unsupervised representation learning, strong identifiability results for disentanglement of causally-related latent variables have been established by exploiting certain side information, such as class labels, in addition to independence. However, most existing work is constrained by functional form assumptions such as independent sources or further with linear transitions, and distribution assumptions such as stationary, exponential family distribution. It is unknown whether the underlying latent variables and their causal relations are identifiable if they have arbitrary, nonparametric causal influences in between. In this work, we establish the identifiability theories of nonparametric latent causal processes from their nonlinear mixtures under fixed temporal causal influences and analyze how distribution changes can further benefit the disentanglement. We propose \\textbf{\\texttt{TDRL}}, a principled framework to recover time-delayed latent causal variables and identify their relations from measured sequential data under stationary environments and under different distribution shifts. Specifically, the framework can factorize unknown distribution shifts into transition distribution changes under fixed and time-varying latent causal relations, and under observation changes in observation. Through experiments, we show that time-delayed latent causal influences are reliably identified and that our approach considerably outperforms existing baselines that do not correctly exploit this modular representation of changes. Our code is available at: \\url{https://github.com/weirayao/tdrl}."
                    },
                    {
                        "title": "Personalized Trajectory Prediction via Distribution Discrimination",
                        "abstract": "Trajectory prediction is confronted with the dilemma to capture the multi-modal nature of future dynamics with both diversity and accuracy. In this paper, we present a distribution discrimination (DisDis) method to predict personalized motion patterns by distinguishing the potential distributions. Motivated by that the motion pattern of each person is personalized due to his/her habit, our DisDis learns the latent distribution to represent different motion patterns and optimize it by the contrastive discrimination. This distribution discrimination encourages latent distributions to be more discriminative. Our method can be integrated with existing multi-modal stochastic predictive models as a plug-and-play module to learn the more discriminative latent distribution. To evaluate the latent distribution, we further propose a new metric, probability cumulative minimum distance (PCMD) curve, which cumulatively calculates the minimum distance on the sorted probabilities. Experimental results on the ETH and UCY datasets show the effectiveness of our method."
                    },
                    {
                        "title": "Adversarial Alignment for Source Free Object Detection",
                        "abstract": "Source-free object detection (SFOD) aims to transfer a detector pre-trained on a label-rich source domain to an unlabeled target domain without seeing source data. While most existing SFOD methods generate pseudo labels via a source-pretrained model to guide training, these pseudo labels usually contain high noises due to heavy domain discrepancy. In order to obtain better pseudo supervisions, we divide the target domain into source-similar and source-dissimilar parts and align them in the feature space by adversarial learning. Specifically, we design a detection variance-based criterion to divide the target domain. This criterion is motivated by a finding that larger detection variances denote higher recall and larger similarity to the source domain. Then we incorporate an adversarial module into a mean teacher framework to drive the feature spaces of these two subsets indistinguishable. Extensive experiments on multiple cross-domain object detection datasets demonstrate that our proposed method consistently outperforms the compared SFOD methods."
                    },
                    {
                        "title": "Learning Causal Domain-Invariant Temporal Dynamics for Few-Shot Action Recognition",
                        "abstract": "Few-shot action recognition aims at quickly adapting a pre-trained model to the novel data with a distribution shift using only a limited number of samples. Key challenges include how to identify and leverage the transferable knowledge learned by the pre-trained model. We therefore propose CDTD, or Causal Domain-Invariant Temporal Dynamics for knowledge transfer. To identify the temporally invariant and variant representations, we employ the causal representation learning methods for unsupervised pertaining, and then tune the classifier with supervisions in next stage. Specifically, we assume the domain information can be well estimated and the pre-trained image decoder and transition models can be well transferred. During adaptation, we fix the transferable temporal dynamics and update the image encoder and domain estimator. The efficacy of our approach is revealed by the superior accuracy of CDTD over leading alternatives across standard few-shot action recognition datasets."
                    },
                    {
                        "title": "MuGSI: Distilling GNNs with Multi-Granularity Structural Information for Graph Classification",
                        "abstract": "Recent works have introduced GNN-to-MLP knowledge distillation (KD) frameworks to combine both GNN's superior performance and MLP's fast inference speed. However, existing KD frameworks are primarily designed for node classification within single graphs, leaving their applicability to graph classification largely unexplored. Two main challenges arise when extending KD for node classification to graph classification: (1) The inherent sparsity of learning signals due to soft labels being generated at the graph level; (2) The limited expressiveness of student MLPs, especially in datasets with limited input feature spaces. To overcome these challenges, we introduce MuGSI, a novel KD framework that employs Multi-granularity Structural Information for graph classification. Specifically, we propose multi-granularity distillation loss in MuGSI to tackle the first challenge. This loss function is composed of three distinct components: graph-level distillation, subgraph-level distillation, and node-level distillation. Each component targets a specific granularity of the graph structure, ensuring a comprehensive transfer of structural knowledge from the teacher model to the student model. To tackle the second challenge, MuGSI proposes to incorporate a node feature augmentation component, thereby enhancing the expressiveness of the student MLPs and making them more capable learners. We perform extensive experiments across a variety of datasets and different teacher/student model architectures. The experiment results demonstrate the effectiveness, efficiency, and robustness of MuGSI. Codes are publicly available at: \\textbf{\\url{https://github.com/tianyao-aka/MuGSI}.}"
                    },
                    {
                        "title": "CT4D: Consistent Text-to-4D Generation with Animatable Meshes",
                        "abstract": "Text-to-4D generation has recently been demonstrated viable by integrating a 2D image diffusion model with a video diffusion model. However, existing models tend to produce results with inconsistent motions and geometric structures over time. To this end, we present a novel framework, coined CT4D, which directly operates on animatable meshes for generating consistent 4D content from arbitrary user-supplied prompts. The primary challenges of our mesh-based framework involve stably generating a mesh with details that align with the text prompt while directly driving it and maintaining surface continuity. Our CT4D framework incorporates a unique Generate-Refine-Animate (GRA) algorithm to enhance the creation of text-aligned meshes. To improve surface continuity, we divide a mesh into several smaller regions and implement a uniform driving function within each area. Additionally, we constrain the animating stage with a rigidity regulation to ensure cross-region continuity. Our experimental results, both qualitative and quantitative, demonstrate that our CT4D framework surpasses existing text-to-4D techniques in maintaining interframe consistency and preserving global geometry. Furthermore, we showcase that this enhanced representation inherently possesses the capability for combinational 4D generation and texture editing."
                    }
                ]
            },
            "c0431712-b47d-41a3-86d2-831fe743dd6a": {
                "pk": "c0431712-b47d-41a3-86d2-831fe743dd6a",
                "name": "Yujia Zheng",
                "collaborators": [
                    "Kun Zhang",
                    "Ignavier Ng",
                    "Siyi Liu",
                    "Zekun Li",
                    "Shu Wu",
                    "Zailei Zhou",
                    "Yewen Fan",
                    "Shaoan Xie",
                    "Jiji Zhang",
                    "Xinshuai Dong"
                ],
                "domain": [
                    "Recommendation Systems",
                    "Graph Neural Network",
                    "Causal Inference",
                    "Meta-Learning"
                ],
                "publications": [
                    {
                        "title": "Long-tail Session-based Recommendation",
                        "abstract": "Session-based recommendation focuses on the prediction of user actions based on anonymous sessions and is a necessary method in the lack of user historical data. However, none of the existing session-based recommendation methods explicitly takes the long-tail recommendation into consideration, which plays an important role in improving the diversity of recommendation and producing the serendipity. As the distribution of items with long-tail is prevalent in session-based recommendation scenarios (e.g., e-commerce, music, and TV program recommendations), more attention should be put on the long-tail session-based recommendation. In this paper, we propose a novel network architecture, namely TailNet, to improve long-tail recommendation performance, while maintaining competitive accuracy performance compared with other methods. We start by classifying items into short-head (popular) and long-tail (niche) items based on click frequency. Then a novel is proposed and applied in TailNet to determine user preference between two types of items, so as to softly adjust and personalize recommendations. Extensive experiments on two real-world datasets verify the superiority of our method compared with state-of-the-art works."
                    },
                    {
                        "title": "Generalizing Nonlinear ICA Beyond Structural Sparsity",
                        "abstract": "Nonlinear independent component analysis (ICA) aims to uncover the true latent sources from their observable nonlinear mixtures. Despite its significance, the identifiability of nonlinear ICA is known to be impossible without additional assumptions. Recent advances have proposed conditions on the connective structure from sources to observed variables, known as Structural Sparsity, to achieve identifiability in an unsupervised manner. However, the sparsity constraint may not hold universally for all sources in practice. Furthermore, the assumptions of bijectivity of the mixing process and independence among all sources, which arise from the setting of ICA, may also be violated in many real-world scenarios. To address these limitations and generalize nonlinear ICA, we propose a set of new identifiability results in the general settings of undercompleteness, partial sparsity and source dependence, and flexible grouping structures. Specifically, we prove identifiability when there are more observed variables than sources (undercomplete), and when certain sparsity and/or source independence assumptions are not met for some changing sources. Moreover, we show that even in cases with flexible grouping structures (e.g., part of the sources can be divided into irreducible independent groups with various sizes), appropriate identifiability results can also be established. Theoretical claims are supported empirically on both synthetic and real-world datasets."
                    },
                    {
                        "title": "Cold-start Sequential Recommendation via Meta Learner",
                        "abstract": "This paper explores meta-learning in sequential recommendation to alleviate the item cold-start problem. Sequential recommendation aims to capture user's dynamic preferences based on historical behavior sequences and acts as a key component of most online recommendation scenarios. However, most previous methods have trouble recommending cold-start items, which are prevalent in those scenarios. As there is generally no side information in the setting of sequential recommendation task, previous cold-start methods could not be applied when only user-item interactions are available. Thus, we propose a Meta-learning-based Cold-Start Sequential Recommendation Framework, namely Mecos, to mitigate the item cold-start problem in sequential recommendation. This task is non-trivial as it targets at an important problem in a novel and challenging context. Mecos effectively extracts user preference from limited interactions and learns to match the target cold-start item with the potential user. Besides, our framework can be painlessly integrated with neural network-based models. Extensive experiments conducted on three real-world datasets verify the superiority of Mecos, with the average improvement up to 99%, 91%, and 70% in HR@10 over state-of-the-art baseline methods."
                    },
                    {
                        "title": "Balancing Multi-level Interactions for Session-based Recommendation",
                        "abstract": "Predicting user actions based on anonymous sessions is a challenge to general recommendation systems because the lack of user profiles heavily limits data-driven models. Recently, session-based recommendation methods have achieved remarkable results in dealing with this task. However, the upper bound of performance can still be boosted through the innovative exploration of limited data. In this paper, we propose a novel method, namely Intra-and Inter-session Interaction-aware Graph-enhanced Network, to take inter-session item-level interactions into account. Different from existing intra-session item-level interactions and session-level collaborative information, our introduced data represents complex item-level interactions between different sessions. For mining the new data without breaking the equilibrium of the model between different interactions, we construct an intra-session graph and an inter-session graph for the current session. The former focuses on item-level interactions within a single session and the latter models those between items among neighborhood sessions. Then different approaches are employed to encode the information of two graphs according to different structures, and the generated latent vectors are combined to balance the model across different scopes. Experiments on real-world datasets verify that our method outperforms other state-of-the-art methods."
                    },
                    {
                        "title": "DGTN: Dual-channel Graph Transition Network for Session-based Recommendation",
                        "abstract": "The task of session-based recommendation is to predict user actions based on anonymous sessions. Recent research mainly models the target session as a sequence or a graph to capture item transitions within it, ignoring complex transitions between items in different sessions that have been generated by other users. These item transitions include potential collaborative information and reflect similar behavior patterns, which we assume may help with the recommendation for the target session. In this paper, we propose a novel method, namely Dual-channel Graph Transition Network (DGTN), to model item transitions within not only the target session but also the neighbor sessions. Specifically, we integrate the target session and its neighbor (similar) sessions into a single graph. Then the transition signals are explicitly injected into the embedding by channel-aware propagation. Experiments on real-world datasets demonstrate that DGTN outperforms other state-of-the-art methods. Further analysis verifies the rationality of dual-channel item transition modeling, suggesting a potential future direction for session-based recommendation."
                    },
                    {
                        "title": "On the Identifiability of Nonlinear ICA: Sparsity and Beyond",
                        "abstract": "Nonlinear independent component analysis (ICA) aims to recover the underlying independent latent sources from their observable nonlinear mixtures. How to make the nonlinear ICA model identifiable up to certain trivial indeterminacies is a long-standing problem in unsupervised learning. Recent breakthroughs reformulate the standard independence assumption of sources as conditional independence given some auxiliary variables (e.g., class labels and/or domain/time indexes) as weak supervision or inductive bias. However, nonlinear ICA with unconditional priors cannot benefit from such developments. We explore an alternative path and consider only assumptions on the mixing process, such as Structural Sparsity. We show that under specific instantiations of such constraints, the independent latent sources can be identified from their nonlinear mixtures up to a permutation and a component-wise transformation, thus achieving nontrivial identifiability of nonlinear ICA without auxiliary variables. We provide estimation methods and validate the theoretical results experimentally. The results on image data suggest that our conditions may hold in a number of practical data generating processes."
                    },
                    {
                        "title": "Generalized Precision Matrix for Scalable Estimation of Nonparametric Markov Networks",
                        "abstract": "A Markov network characterizes the conditional independence structure, or Markov property, among a set of random variables. Existing work focuses on specific families of distributions (e.g., exponential families) and/or certain structures of graphs, and most of them can only handle variables of a single data type (continuous or discrete). In this work, we characterize the conditional independence structure in general distributions for all data types (i.e., continuous, discrete, and mixed-type) with a Generalized Precision Matrix (GPM). Besides, we also allow general functional relations among variables, thus giving rise to a Markov network structure learning algorithm in one of the most general settings. To deal with the computational challenge of the problem, especially for large graphs, we unify all cases under the same umbrella of a regularized score matching framework. We validate the theoretical results and demonstrate the scalability empirically in various settings."
                    },
                    {
                        "title": "Causal Representation Learning from Multiple Distributions: A General Setting",
                        "abstract": "In many problems, the measured variables (e.g., image pixels) are just mathematical functions of the latent causal variables (e.g., the underlying concepts or objects). For the purpose of making predictions in changing environments or making proper changes to the system, it is helpful to recover the latent causal variables $Z_i$ and their causal relations represented by graph $\\mathcal{G}_Z$. This problem has recently been known as causal representation learning. This paper is concerned with a general, completely nonparametric setting of causal representation learning from multiple distributions (arising from heterogeneous data or nonstationary time series), without assuming hard interventions behind distribution changes. We aim to develop general solutions in this fundamental case; as a by product, this helps see the unique benefit offered by other assumptions such as parametric causal models or hard interventions. We show that under the sparsity constraint on the recovered graph over the latent variables and suitable sufficient change conditions on the causal influences, interestingly, one can recover the moralized graph of the underlying directed acyclic graph, and the recovered latent variables and their relations are related to the underlying causal model in a specific, nontrivial way. In some cases, most latent variables can even be recovered up to component-wise transformations. Experimental results verify our theoretical claims."
                    },
                    {
                        "title": "Reliable Causal Discovery with Improved Exact Search and Weaker Assumptions",
                        "abstract": "Many of the causal discovery methods rely on the faithfulness assumption to guarantee asymptotic correctness. However, the assumption can be approximately violated in many ways, leading to sub-optimal solutions. Although there is a line of research in Bayesian network structure learning that focuses on weakening the assumption, such as exact search methods with well-defined score functions, they do not scale well to large graphs. In this work, we introduce several strategies to improve the scalability of exact score-based methods in the linear Gaussian setting. In particular, we develop a super-structure estimation method based on the support of inverse covariance matrix which requires assumptions that are strictly weaker than faithfulness, and apply it to restrict the search space of exact search. We also propose a local search strategy that performs exact search on the local clusters formed by each variable and its neighbors within two hops in the super-structure. Numerical experiments validate the efficacy of the proposed procedure, and demonstrate that it scales up to hundreds of nodes with a high accuracy."
                    },
                    {
                        "title": "On the Identifiability of Sparse ICA without Assuming Non-Gaussianity",
                        "abstract": "Independent component analysis (ICA) is a fundamental statistical tool used to reveal hidden generative processes from observed data. However, traditional ICA approaches struggle with the rotational invariance inherent in Gaussian distributions, often necessitating the assumption of non-Gaussianity in the underlying sources. This may limit their applicability in broader contexts. To accommodate Gaussian sources, we develop an identifiability theory that relies on second-order statistics without imposing further preconditions on the distribution of sources, by introducing novel assumptions on the connective structure from sources to observed variables. Different from recent work that focuses on potentially restrictive connective structures, our proposed assumption of structural variability is both considerably less restrictive and provably necessary. Furthermore, we propose two estimation methods based on second-order statistics and sparsity constraint. Experimental results are provided to validate our identifiability theory and estimation methods."
                    },
                    {
                        "title": "Detecting and Identifying Selection Structure in Sequential Data",
                        "abstract": "We argue that the selective inclusion of data points based on latent objectives is common in practical situations, such as music sequences. Since this selection process often distorts statistical analysis, previous work primarily views it as a bias to be corrected and proposes various methods to mitigate its effect. However, while controlling this bias is crucial, selection also offers an opportunity to provide a deeper insight into the hidden generation process, as it is a fundamental mechanism underlying what we observe. In particular, overlooking selection in sequential data can lead to an incomplete or overcomplicated inductive bias in modeling, such as assuming a universal autoregressive structure for all dependencies. Therefore, rather than merely viewing it as a bias, we explore the causal structure of selection in sequential data to delve deeper into the complete causal process. Specifically, we show that selection structure is identifiable without any parametric assumptions or interventional experiments. Moreover, even in cases where selection variables coexist with latent confounders, we still establish the nonparametric identifiability under appropriate structural conditions. Meanwhile, we also propose a provably correct algorithm to detect and identify selection structures as well as other types of dependencies. The framework has been validated empirically on both synthetic data and real-world music."
                    },
                    {
                        "title": "Heterogeneous Graph Collaborative Filtering",
                        "abstract": "Graph-based collaborative filtering (CF) algorithms have gained increasing attention. Existing work in this literature usually models the user-item interactions as a bipartite graph, where users and items are two isolated node sets and edges between them indicate their interactions. Then, the unobserved preference of users can be exploited by modeling high-order connectivity on the bipartite graph. In this work, we propose to model user-item interactions as a heterogeneous graph which consists of not only user-item edges indicating their interaction but also user-user edges indicating their similarity. We develop heterogeneous graph collaborative filtering (HGCF), a GCN-based framework which can explicitly capture both the interaction signal and similarity signal through embedding propagation on the heterogeneous graph. Since the heterogeneous graph is more connected than the bipartite graph, the sparsity issue can be alleviated and the demand for expensive high-order connectivity modeling can be lowered. Extensive experiments conducted on three public benchmarks demonstrate its superiority over the state-of-the-arts. Further analysis verifies the importance of user-user edges in the graph, justifying the rationality and effectiveness of HGCF."
                    },
                    {
                        "title": "Local Causal Discovery with Linear non-Gaussian Cyclic Models",
                        "abstract": "Local causal discovery is of great practical significance, as there are often situations where the discovery of the global causal structure is unnecessary, and the interest lies solely on a single target variable. Most existing local methods utilize conditional independence relations, providing only a partially directed graph, and assume acyclicity for the ground-truth structure, even though real-world scenarios often involve cycles like feedback mechanisms. In this work, we present a general, unified local causal discovery method with linear non-Gaussian models, whether they are cyclic or acyclic. We extend the application of independent component analysis from the global context to independent subspace analysis, enabling the exact identification of the equivalent local directed structures and causal strengths from the Markov blanket of the target variable. We also propose an alternative regression-based method in the particular acyclic scenarios. Our identifiability results are empirically validated using both synthetic and real-world datasets."
                    },
                    {
                        "title": "Learning Elastic Embeddings for Customizing On-Device Recommenders",
                        "abstract": "In today's context, deploying data-driven services like recommendation on edge devices instead of cloud servers becomes increasingly attractive due to privacy and network latency concerns. A common practice in building compact on-device recommender systems is to compress their embeddings which are normally the cause of excessive parameterization. However, despite the vast variety of devices and their associated memory constraints, existing memory-efficient recommender systems are only specialized for a fixed memory budget in every design and training life cycle, where a new model has to be retrained to obtain the optimal performance while adapting to a smaller/larger memory budget. In this paper, we present a novel lightweight recommendation paradigm that allows a well-trained recommender to be customized for arbitrary device-specific memory constraints without retraining. The core idea is to compose elastic embeddings for each item, where an elastic embedding is the concatenation of a set of embedding blocks that are carefully chosen by an automated search function. Correspondingly, we propose an innovative approach, namely recommendation with universally learned elastic embeddings (RULE). To ensure the expressiveness of all candidate embedding blocks, RULE enforces a diversity-driven regularization when learning different embedding blocks. Then, a performance estimator-based evolutionary search function is designed, allowing for efficient specialization of elastic embeddings under any memory constraint for on-device recommendation. Extensive experiments on real-world datasets reveal the superior performance of RULE under tight memory budgets."
                    }
                ]
            },
            "ed55a6cd-5de8-4fd1-9c2d-f2e39d5e64ca": {
                "pk": "ed55a6cd-5de8-4fd1-9c2d-f2e39d5e64ca",
                "name": "Yewen Fan",
                "collaborators": [
                    "Kun Zhang",
                    "Nian Si",
                    "Xiangchen Song",
                    "Xinshuai Dong",
                    "Yue Wu",
                    "Yuanzhi Li",
                    "Yujia Zheng",
                    "Ignavier Ng",
                    "Hongping Zhang",
                    "Zhenfeng Shao"
                ],
                "domain": [
                    "Deep Learning",
                    "Causal Inference",
                    "Reinforcement Learning",
                    "Unsupervised Learning"
                ],
                "publications": [
                    {
                        "title": "Calibration-then-Calculation: A Variance Reduced Metric Framework in Deep Click-Through Rate Prediction Models",
                        "abstract": "The adoption of deep learning across various fields has been extensive, yet there is a lack of focus on evaluating the performance of deep learning pipelines. Typically, with the increased use of large datasets and complex models, the training process is run only once and the result is compared to previous benchmarks. This practice can lead to imprecise comparisons due to the variance in neural network evaluation metrics, which stems from the inherent randomness in the training process. Traditional solutions, such as running the training process multiple times, are often infeasible due to computational constraints. In this paper, we introduce a novel metric framework, the Calibrated Loss Metric, designed to address this issue by reducing the variance present in its conventional counterpart. Consequently, this new metric enhances the accuracy in detecting effective modeling improvements. Our approach is substantiated by theoretical justifications and extensive experimental validations within the context of Deep Click-Through Rate Prediction Models."
                    },
                    {
                        "title": "Generalized Precision Matrix for Scalable Estimation of Nonparametric Markov Networks",
                        "abstract": "A Markov network characterizes the conditional independence structure, or Markov property, among a set of random variables. Existing work focuses on specific families of distributions (e.g., exponential families) and/or certain structures of graphs, and most of them can only handle variables of a single data type (continuous or discrete). In this work, we characterize the conditional independence structure in general distributions for all data types (i.e., continuous, discrete, and mixed-type) with a Generalized Precision Matrix (GPM). Besides, we also allow general functional relations among variables, thus giving rise to a Markov network structure learning algorithm in one of the most general settings. To deal with the computational challenge of the problem, especially for large graphs, we unify all cases under the same umbrella of a regularized score matching framework. We validate the theoretical results and demonstrate the scalability empirically in various settings."
                    },
                    {
                        "title": "Calibration Matters: Tackling Maximization Bias in Large-scale Advertising Recommendation Systems",
                        "abstract": "Calibration is defined as the ratio of the average predicted click rate to the true click rate. The optimization of calibration is essential to many online advertising recommendation systems because it directly affects the downstream bids in ads auctions and the amount of money charged to advertisers. Despite its importance, calibration optimization often suffers from a problem called \"maximization bias\". Maximization bias refers to the phenomenon that the maximum of predicted values overestimates the true maximum. The problem is introduced because the calibration is computed on the set selected by the prediction model itself. It persists even if unbiased predictions can be achieved on every datapoint and worsens when covariate shifts exist between the training and test sets. To mitigate this problem, we theorize the quantification of maximization bias and propose a variance-adjusting debiasing (VAD) meta-algorithm in this paper. The algorithm is efficient, robust, and practical as it is able to mitigate maximization bias problems under covariate shifts, neither incurring additional online serving costs nor compromising the ranking performance. We demonstrate the effectiveness of the proposed algorithm using a state-of-the-art recommendation neural network model on a large-scale real-world dataset."
                    },
                    {
                        "title": "Evaluating the weight sensitivity in AHP-based flood risk estimation models",
                        "abstract": "In the analytic hierarchy process (AHP) based flood risk estimation models, it is widely acknowledged that different weighting criteria can lead to different results. In this study, we evaluated and discussed the sensitivity of flood risk estimation brought by judgment matrix definition by investigating the performance of pixel-based and sub-watershed-based AHP models. Taking a flood event that occurred in July 2020 in Chaohu basin, Anhui province, China, as a study case, we used the flood areas extracted from remote sensing images to construct ground truth for validation purposes. The results suggest that the performance of the pixel-based AHP model fluctuates intensively given different definitions of judgment matrixes, while the performance of sub-watershed-based AHP models fluctuates considerably less than that of the pixel-based AHP model. Specifically, sub-watershed delimitated via multiple flow direction (MFD) always achieves increases in the correct ratio and the fit ratio by >35% and >5% with the pixel-based AHP model, respectively."
                    },
                    {
                        "title": "On the Three Demons in Causality in Finance: Time Resolution, Nonstationarity, and Latent Factors",
                        "abstract": "Financial data is generally time series in essence and thus suffers from three fundamental issues: the mismatch in time resolution, the time-varying property of the distribution - nonstationarity, and causal factors that are important but unknown/unobserved. In this paper, we follow a causal perspective to systematically look into these three demons in finance. Specifically, we reexamine these issues in the context of causality, which gives rise to a novel and inspiring understanding of how the issues can be addressed. Following this perspective, we provide systematic solutions to these problems, which hopefully would serve as a foundation for future research in the area."
                    },
                    {
                        "title": "Read and Reap the Rewards: Learning to Play Atari with the Help of Instruction Manuals",
                        "abstract": "High sample complexity has long been a challenge for RL. On the other hand, humans learn to perform tasks not only from interaction or demonstrations, but also by reading unstructured text documents, e.g., instruction manuals. Instruction manuals and wiki pages are among the most abundant data that could inform agents of valuable features and policies or task-specific environmental dynamics and reward structures. Therefore, we hypothesize that the ability to utilize human-written instruction manuals to assist learning policies for specific tasks should lead to a more efficient and better-performing agent. We propose the Read and Reward framework. Read and Reward speeds up RL algorithms on Atari games by reading manuals released by the Atari game developers. Our framework consists of a QA Extraction module that extracts and summarizes relevant information from the manual and a Reasoning module that evaluates object-agent interactions based on information from the manual. An auxiliary reward is then provided to a standard A2C RL agent, when interaction is detected. Experimentally, various RL algorithms obtain significant improvement in performance and training speed when assisted by our design."
                    },
                    {
                        "title": "Temporally Disentangled Representation Learning under Unknown Nonstationarity",
                        "abstract": "In unsupervised causal representation learning for sequential data with time-delayed latent causal influences, strong identifiability results for the disentanglement of causally-related latent variables have been established in stationary settings by leveraging temporal structure. However, in nonstationary setting, existing work only partially addressed the problem by either utilizing observed auxiliary variables (e.g., class labels and/or domain indexes) as side information or assuming simplified latent causal dynamics. Both constrain the method to a limited range of scenarios. In this study, we further explored the Markov Assumption under time-delayed causally related process in nonstationary setting and showed that under mild conditions, the independent latent components can be recovered from their nonlinear mixture up to a permutation and a component-wise transformation, without the observation of auxiliary variables. We then introduce NCTRL, a principled estimation framework, to reconstruct time-delayed latent causal variables and identify their relations from measured sequential data only. Empirical evaluations demonstrated the reliable identification of time-delayed latent causal influences, with our methodology substantially outperforming existing baselines that fail to exploit the nonstationarity adequately and then, consequently, cannot distinguish distribution shifts."
                    },
                    {
                        "title": "AgentKit: Structured LLM Reasoning with Dynamic Graphs",
                        "abstract": "We propose an intuitive LLM prompting framework (AgentKit) for multifunctional agents. AgentKit offers a unified framework for explicitly constructing a complex \"thought process\" from simple natural language prompts. The basic building block in AgentKit is a node, containing a natural language prompt for a specific subtask. The user then puts together chains of nodes, like stacking LEGO pieces. The chains of nodes can be designed to explicitly enforce a naturally structured \"thought process\". For example, for the task of writing a paper, one may start with the thought process of 1) identify a core message, 2) identify prior research gaps, etc. The nodes in AgentKit can be designed and combined in different ways to implement multiple advanced capabilities including on-the-fly hierarchical planning, reflection, and learning from interactions. In addition, due to the modular nature and the intuitive design to simulate explicit human thought process, a basic agent could be implemented as simple as a list of prompts for the subtasks and therefore could be designed and tuned by someone without any programming experience. Quantitatively, we show that agents designed through AgentKit achieve SOTA performance on WebShop and Crafter. These advances underscore AgentKit's potential in making LLM agents effective and accessible for a wider range of applications. https://github.com/holmeswww/AgentKit"
                    }
                ]
            },
            "e447ce39-d0a2-4538-817a-56dc43b6ac34": {
                "pk": "e447ce39-d0a2-4538-817a-56dc43b6ac34",
                "name": "Xinshuai Dong",
                "collaborators": [
                    "Anh Tuan Luu",
                    "Xiaobao Wu",
                    "Thong Nguyen",
                    "Kun Zhang",
                    "Ignavier Ng",
                    "Songyao Jin",
                    "Peter Spirtes",
                    "Yujia Zheng",
                    "Yewen Fan",
                    "Liangming Pan"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Topic Modeling",
                    "Causal Inference",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Mitigating Data Sparsity for Short Text Topic Modeling by Topic-Semantic Contrastive Learning",
                        "abstract": "To overcome the data sparsity issue in short text topic modeling, existing methods commonly rely on data augmentation or the data characteristic of short texts to introduce more word co-occurrence information. However, most of them do not make full use of the augmented data or the data characteristic: they insufficiently learn the relations among samples in data, leading to dissimilar topic distributions of semantically similar text pairs. To better address data sparsity, in this paper we propose a novel short text topic modeling framework, Topic-Semantic Contrastive Topic Model (TSCTM). To sufficiently model the relations among samples, we employ a new contrastive learning method with efficient positive and negative sampling strategies based on topic semantics. This contrastive learning method refines the representations, enriches the learning signals, and thus mitigates the sparsity issue. Extensive experimental results show that our TSCTM outperforms state-of-the-art baselines regardless of the data augmentation availability, producing high-quality topics and topic distributions."
                    },
                    {
                        "title": "On the Identifiability of Sparse ICA without Assuming Non-Gaussianity",
                        "abstract": "Independent component analysis (ICA) is a fundamental statistical tool used to reveal hidden generative processes from observed data. However, traditional ICA approaches struggle with the rotational invariance inherent in Gaussian distributions, often necessitating the assumption of non-Gaussianity in the underlying sources. This may limit their applicability in broader contexts. To accommodate Gaussian sources, we develop an identifiability theory that relies on second-order statistics without imposing further preconditions on the distribution of sources, by introducing novel assumptions on the connective structure from sources to observed variables. Different from recent work that focuses on potentially restrictive connective structures, our proposed assumption of structural variability is both considerably less restrictive and provably necessary. Furthermore, we propose two estimation methods based on second-order statistics and sparsity constraint. Experimental results are provided to validate our identifiability theory and estimation methods."
                    },
                    {
                        "title": "Towards Robustness Against Natural Language Word Substitutions",
                        "abstract": "Robustness against word substitutions has a well-defined and widely acceptable form, i.e., using semantically similar words as substitutions, and thus it is considered as a fundamental stepping-stone towards broader robustness in natural language processing. Previous defense methods capture word substitutions in vector space by using either $l_2$-ball or hyper-rectangle, which results in perturbation sets that are not inclusive enough or unnecessarily large, and thus impedes mimicry of worst cases for robust training. In this paper, we introduce a novel \\textit{Adversarial Sparse Convex Combination} (ASCC) method. We model the word substitution attack space as a convex hull and leverages a regularization term to enforce perturbation towards an actual substitution, thus aligning our modeling better with the discrete textual space. Based on the ASCC method, we further propose ASCC-defense, which leverages ASCC to generate worst-case perturbations and incorporates adversarial training towards robustness. Experiments show that ASCC-defense outperforms the current state-of-the-arts in terms of robustness on two prevailing NLP tasks, \\emph{i.e.}, sentiment analysis and natural language inference, concerning several attacks across multiple model architectures. Besides, we also envision a new class of defense towards robustness in NLP, where our robustly trained word vectors can be plugged into a normally trained model and enforce its robustness without applying any other defense techniques."
                    },
                    {
                        "title": "Effective Neural Topic Modeling with Embedding Clustering Regularization",
                        "abstract": "Topic models have been prevalent for decades with various applications. However, existing topic models commonly suffer from the notorious topic collapsing: discovered topics semantically collapse towards each other, leading to highly repetitive topics, insufficient topic discovery, and damaged model interpretability. In this paper, we propose a new neural topic model, Embedding Clustering Regularization Topic Model (ECRTM). Besides the existing reconstruction error, we propose a novel Embedding Clustering Regularization (ECR), which forces each topic embedding to be the center of a separately aggregated word embedding cluster in the semantic space. This enables each produced topic to contain distinct word semantics, which alleviates topic collapsing. Regularized by ECR, our ECRTM generates diverse and coherent topics together with high-quality topic distributions of documents. Extensive experiments on benchmark datasets demonstrate that ECRTM effectively addresses the topic collapsing issue and consistently surpasses state-of-the-art baselines in terms of topic quality, topic distributions of documents, and downstream classification tasks."
                    },
                    {
                        "title": "On the Three Demons in Causality in Finance: Time Resolution, Nonstationarity, and Latent Factors",
                        "abstract": "Financial data is generally time series in essence and thus suffers from three fundamental issues: the mismatch in time resolution, the time-varying property of the distribution - nonstationarity, and causal factors that are important but unknown/unobserved. In this paper, we follow a causal perspective to systematically look into these three demons in finance. Specifically, we reexamine these issues in the context of causality, which gives rise to a novel and inspiring understanding of how the issues can be addressed. Following this perspective, we provide systematic solutions to these problems, which hopefully would serve as a foundation for future research in the area."
                    },
                    {
                        "title": "InfoCTM: A Mutual Information Maximization Perspective of Cross-Lingual Topic Modeling",
                        "abstract": "Cross-lingual topic models have been prevalent for cross-lingual text analysis by revealing aligned latent topics. However, most existing methods suffer from producing repetitive topics that hinder further analysis and performance decline caused by low-coverage dictionaries. In this paper, we propose the Cross-lingual Topic Modeling with Mutual Information (InfoCTM). Instead of the direct alignment in previous work, we propose a topic alignment with mutual information method. This works as a regularization to properly align topics and prevent degenerate topic representations of words, which mitigates the repetitive topic issue. To address the low-coverage dictionary issue, we further propose a cross-lingual vocabulary linking method that finds more linked cross-lingual words for topic alignment beyond the translations of a given dictionary. Extensive experiments on English, Chinese, and Japanese datasets demonstrate that our method outperforms state-of-the-art baselines, producing more coherent, diverse, and well-aligned topics and showing better transferability for cross-lingual classification tasks."
                    },
                    {
                        "title": "Modeling Dynamic Topics in Chain-Free Fashion by Evolution-Tracking Contrastive Learning and Unassociated Word Exclusion",
                        "abstract": "Dynamic topic models track the evolution of topics in sequential documents, which have derived various applications like trend analysis and opinion mining. However, existing models suffer from repetitive topic and unassociated topic issues, failing to reveal the evolution and hindering further applications. To address these issues, we break the tradition of simply chaining topics in existing work and propose a novel neural \\modelfullname. We introduce a new evolution-tracking contrastive learning method that builds the similarity relations among dynamic topics. This not only tracks topic evolution but also maintains topic diversity, mitigating the repetitive topic issue. To avoid unassociated topics, we further present an unassociated word exclusion method that consistently excludes unassociated words from discovered topics. Extensive experiments demonstrate our model significantly outperforms state-of-the-art baselines, tracking topic evolution with high-quality topics, showing better performance on downstream tasks, and remaining robust to the hyperparameter for evolution intensities. Our code is available at https://github.com/bobxwu/CFDTM ."
                    },
                    {
                        "title": "A Versatile Causal Discovery Framework to Allow Causally-Related Hidden Variables",
                        "abstract": "Most existing causal discovery methods rely on the assumption of no latent confounders, limiting their applicability in solving real-life problems. In this paper, we introduce a novel, versatile framework for causal discovery that accommodates the presence of causally-related hidden variables almost everywhere in the causal network (for instance, they can be effects of observed variables), based on rank information of covariance matrix over observed variables. We start by investigating the efficacy of rank in comparison to conditional independence and, theoretically, establish necessary and sufficient conditions for the identifiability of certain latent structural patterns. Furthermore, we develop a Rank-based Latent Causal Discovery algorithm, RLCD, that can efficiently locate hidden variables, determine their cardinalities, and discover the entire causal structure over both measured and hidden ones. We also show that, under certain graphical conditions, RLCD correctly identifies the Markov Equivalence Class of the whole latent causal graph asymptotically. Experimental results on both synthetic and real-world personality data sets demonstrate the efficacy of the proposed approach in finite-sample cases."
                    },
                    {
                        "title": "On the Parameter Identifiability of Partially Observed Linear Causal Models",
                        "abstract": "Linear causal models are important tools for modeling causal dependencies and yet in practice, only a subset of the variables can be observed. In this paper, we examine the parameter identifiability of these models by investigating whether the edge coefficients can be recovered given the causal structure and partially observed data. Our setting is more general than that of prior research - we allow all variables, including both observed and latent ones, to be flexibly related, and we consider the coefficients of all edges, whereas most existing works focus only on the edges between observed variables. Theoretically, we identify three types of indeterminacy for the parameters in partially observed linear causal models. We then provide graphical conditions that are sufficient for all parameters to be identifiable and show that some of them are provably necessary. Methodologically, we propose a novel likelihood-based parameter estimation method that addresses the variance indeterminacy of latent variables in a specific way and can asymptotically recover the underlying parameters up to trivial indeterminacy. Empirical studies on both synthetic and real-world datasets validate our identifiability theory and the effectiveness of the proposed method in the finite-sample regime."
                    },
                    {
                        "title": "Certified Robustness Against Natural Language Attacks by Causal Intervention",
                        "abstract": "Deep learning models have achieved great success in many fields, yet they are vulnerable to adversarial examples. This paper follows a causal perspective to look into the adversarial vulnerability and proposes Causal Intervention by Semantic Smoothing (CISS), a novel framework towards robustness against natural language attacks. Instead of merely fitting observational data, CISS learns causal effects p(y|do(x)) by smoothing in the latent semantic space to make robust predictions, which scales to deep architectures and avoids tedious construction of noise customized for specific attacks. CISS is provably robust against word substitution attacks, as well as empirically robust even when perturbations are strengthened by unknown attack algorithms. For example, on YELP, CISS surpasses the runner-up by 6.7% in terms of certified robustness against word substitutions, and achieves 79.4% empirical robustness when syntactic attacks are integrated."
                    },
                    {
                        "title": "Gradient-Boosted Decision Tree for Listwise Context Model in Multimodal Review Helpfulness Prediction",
                        "abstract": "Multimodal Review Helpfulness Prediction (MRHP) aims to rank product reviews based on predicted helpfulness scores and has been widely applied in e-commerce via presenting customers with useful reviews. Previous studies commonly employ fully-connected neural networks (FCNNs) as the final score predictor and pairwise loss as the training objective. However, FCNNs have been shown to perform inefficient splitting for review features, making the model difficult to clearly differentiate helpful from unhelpful reviews. Furthermore, pairwise objective, which works on review pairs, may not completely capture the MRHP goal to produce the ranking for the entire review list, and possibly induces low generalization during testing. To address these issues, we propose a listwise attention network that clearly captures the MRHP ranking context and a listwise optimization objective that enhances model generalization. We further propose gradient-boosted decision tree as the score predictor to efficaciously partition product reviews' representations. Extensive experiments demonstrate that our method achieves state-of-the-art results and polished generalization performance on two large-scale MRHP benchmark datasets."
                    },
                    {
                        "title": "DemaFormer: Damped Exponential Moving Average Transformer with Energy-Based Modeling for Temporal Language Grounding",
                        "abstract": "Temporal Language Grounding seeks to localize video moments that semantically correspond to a natural language query. Recent advances employ the attention mechanism to learn the relations between video moments and the text query. However, naive attention might not be able to appropriately capture such relations, resulting in ineffective distributions where target video moments are difficult to separate from the remaining ones. To resolve the issue, we propose an energy-based model framework to explicitly learn moment-query distributions. Moreover, we propose DemaFormer, a novel Transformer-based architecture that utilizes exponential moving average with a learnable damping factor to effectively encode moment-query inputs. Comprehensive experiments on four public temporal language grounding datasets showcase the superiority of our methods over the state-of-the-art baselines."
                    },
                    {
                        "title": "Topic Modeling as Multi-Objective Contrastive Optimization",
                        "abstract": "Recent representation learning approaches enhance neural topic models by optimizing the weighted linear combination of the evidence lower bound (ELBO) of the log-likelihood and the contrastive learning objective that contrasts pairs of input documents. However, document-level contrastive learning might capture low-level mutual information, such as word ratio, which disturbs topic modeling. Moreover, there is a potential conflict between the ELBO loss that memorizes input details for better reconstruction quality, and the contrastive loss which attempts to learn topic representations that generalize among input documents. To address these issues, we first introduce a novel contrastive learning method oriented towards sets of topic vectors to capture useful semantics that are shared among a set of input documents. Secondly, we explicitly cast contrastive topic modeling as a gradient-based multi-objective optimization problem, with the goal of achieving a Pareto stationary solution that balances the trade-off between the ELBO and the contrastive objective. Extensive experiments demonstrate that our framework consistently produces higher-performing neural topic models in terms of topic coherence, topic diversity, and downstream performance."
                    },
                    {
                        "title": "Temporally Disentangled Representation Learning under Unknown Nonstationarity",
                        "abstract": "In unsupervised causal representation learning for sequential data with time-delayed latent causal influences, strong identifiability results for the disentanglement of causally-related latent variables have been established in stationary settings by leveraging temporal structure. However, in nonstationary setting, existing work only partially addressed the problem by either utilizing observed auxiliary variables (e.g., class labels and/or domain indexes) as side information or assuming simplified latent causal dynamics. Both constrain the method to a limited range of scenarios. In this study, we further explored the Markov Assumption under time-delayed causally related process in nonstationary setting and showed that under mild conditions, the independent latent components can be recovered from their nonlinear mixture up to a permutation and a component-wise transformation, without the observation of auxiliary variables. We then introduce NCTRL, a principled estimation framework, to reconstruct time-delayed latent causal variables and identify their relations from measured sequential data only. Empirical evaluations demonstrated the reliable identification of time-delayed latent causal influences, with our methodology substantially outperforming existing baselines that fail to exploit the nonstationarity adequately and then, consequently, cannot distinguish distribution shifts."
                    },
                    {
                        "title": "Counterfactual Reasoning Using Predicted Latent Personality Dimensions for Optimizing Persuasion Outcome",
                        "abstract": "Customizing persuasive conversations related to the outcome of interest for specific users achieves better persuasion results. However, existing persuasive conversation systems rely on persuasive strategies and encounter challenges in dynamically adjusting dialogues to suit the evolving states of individual users during interactions. This limitation restricts the system's ability to deliver flexible or dynamic conversations and achieve suboptimal persuasion outcomes. In this paper, we present a novel approach that tracks a user's latent personality dimensions (LPDs) during ongoing persuasion conversation and generates tailored counterfactual utterances based on these LPDs to optimize the overall persuasion outcome. In particular, our proposed method leverages a Bi-directional Generative Adversarial Network (BiCoGAN) in tandem with a Dialogue-based Personality Prediction Regression (DPPR) model to generate counterfactual data. This enables the system to formulate alternative persuasive utterances that are more suited to the user. Subsequently, we utilize the D3QN model to learn policies for optimized selection of system utterances on counterfactual data. Experimental results we obtained from using the PersuasionForGood dataset demonstrate the superiority of our approach over the existing method, BiCoGAN. The cumulative rewards and Q-values produced by our method surpass ground truth benchmarks, showcasing the efficacy of employing counterfactual reasoning and LPDs to optimize reinforcement learning policy in online interactions."
                    }
                ]
            },
            "fd282b8d-6381-453a-aedb-098344dea003": {
                "pk": "fd282b8d-6381-453a-aedb-098344dea003",
                "name": "Kun Zhang",
                "collaborators": [
                    "Jin Wang",
                    "Vladimir Korepin",
                    "Kun Hao",
                    "Mark Whitmeyer",
                    "Yong Zhang",
                    "Aapo Hyvarinen",
                    "Dmitri Kharzeev",
                    "Kwangmin Yu",
                    "Ignavier Ng"
                ],
                "domain": [
                    "Quantum Information",
                    "Causal Inference",
                    "Bayesian Networks",
                    "Game Theory"
                ],
                "publications": [
                    {
                        "title": "Hyperbolic structures on closed spacelike manifolds",
                        "abstract": "In this paper, we study the intrinsic mean curvature flow on certain closed spacelike manifolds, and prove the existence of hyperbolic structures on them."
                    },
                    {
                        "title": "Withholding Verifiable Information",
                        "abstract": "I study a class of verifiable disclosure games where the sender's payoff is state independent and the receiver's optimal action only depends on the expected state. The sender's messages are verifiable in the sense that they can be vague but can never be wrong. What is the sender's preferred equilibrium? When does the sender gain nothing from having commitment power? I identify conditions for an information design outcome to be an equilibrium outcome of the verifiable disclosure game, and give simple sufficient conditions under which the sender does not benefit from commitment power. These results help in characterizing the sender's preferred equilibria and her equilibrium payoff set in a class of verifiable disclosure games. I apply these insights to study influencing voters and selling with quality disclosure."
                    },
                    {
                        "title": "Entanglement entropy production in deep inelastic scattering",
                        "abstract": "Deep inelastic scattering (DIS) samples a part of the wave function of a hadron in the vicinity of the light cone. Lipatov constructed a spin chain which describes the amplitude of DIS in leading logarithmic approximation. Kharzeev and Levin proposed the entanglement entropy as an observable in DIS [Phys. Rev. D 95, 114008 (2017)], and suggested a relation between the entanglement entropy and parton distributions. Here we represent the DIS process as a local quench in the Lipatov's spin chain, and study the time evolution of the produced entanglement entropy. We show that the resulting entanglement entropy depends on time logarithmically, $\\mathcal S(t)=1/3 \\ln{(t/\\tau)}$ with $\\tau = 1/m$ for $1/m \\le t\\le (mx)^{-1}$, where $m$ is the proton mass and $x$ is the Bjorken $x$. The central charge $c$ of Lipatov's spin chain is determined here to be $c=1$; using the proposed relation between the entanglement entropy and parton distributions, this corresponds to the gluon structure function growing at small $x$ as $xG(x) \\sim 1/x^{1/3}$."
                    },
                    {
                        "title": "Optimal realization of Yang-Baxter gate on quantum computers",
                        "abstract": "Quantum computers provide a promising method to study the dynamics of many-body systems beyond classical simulation. On the other hand, the analytical methods developed and results obtained from the integrable systems provide deep insights on the many-body system. Quantum simulation of the integrable system not only provides a valid benchmark for quantum computers but is also the first step in studying integrable-breaking systems. The building block for the simulation of an integrable system is the Yang-Baxter gate. It is vital to know how to optimally realize the Yang-Baxter gates on quantum computers. Based on the geometric picture of the Yang-Baxter gates, we present the optimal realizations of two types of Yang-Baxter gates with a minimal number of CNOT or $R_{zz}$ gates. We also show how to systematically realize the Yang-Baxter gates via the pulse control. We test and compare the different realizations on IBM quantum computers. We find that the pulse realizations of the Yang-Baxter gates always have a higher gate fidelity compared to the optimal CNOT or $R_{zz}$ realizations. On the basis of the above optimal realizations, we demonstrate the simulation of the Yang-Baxter equation on quantum computers. Our results provide a guideline and standard for further experimental studies based on the Yang-Baxter gate."
                    },
                    {
                        "title": "Redeeming Falsifiability?",
                        "abstract": "We revisit Popper's falsifiability criterion. A tester hires a potential expert to produce a theory, offering payments contingent on the observed performance of the theory. We argue that if the informed expert can acquire additional information, falsifiability does have the power to identify worthless theories."
                    },
                    {
                        "title": "Quantum teleportation and Birman-Murakami-Wenzl algebra",
                        "abstract": "In this paper, we investigate the relationship of quantum teleportation in quantum information science and the Birman-Murakami-Wenzl (BMW) algebra in low-dimensional topology. For simplicity, we focus on the two spin-1/2 representation of the BMW algebra, which is generated by both the Temperley-Lieb projector and the Yang-Baxter gate. We describe quantum teleportation using the Temperley-Lieb projector and the Yang-Baxter gate, respectively, and study teleportation-based quantum computation using the Yang-Baxter gate. On the other hand, we exploit the extended Temperley-Lieb diagrammatical approach to clearly show that the tangle relations of the BMW algebra have a natural interpretation of quantum teleportation. Inspired by this interpretation, we construct a general representation of the tangle relations of the BMW algebra and obtain interesting representations of the BMW algebra. Therefore our research sheds a light on a link between quantum information science and low-dimensional topology."
                    },
                    {
                        "title": "GHZ transform (I): Bell transform and quantum teleportation",
                        "abstract": "It is well-known that maximally entangled states such as the Greenberger-Horne-Zeilinger (GHZ) states, with the Bell states as the simplest examples, are widely exploited in quantum information and computation. We study the application of such maximally entangled states from the viewpoint of the GHZ transform, which is a unitary basis transformation from the product states to the GHZ states. The algebraic structure of the GHZ transform is made clear and representative examples for it are verified as multi-qubit Clifford gates. In this paper, we focus on the Bell transform as the simplest example of the GHZ transform and apply it to the reformulation of quantum circuit model of teleportation and the reformulation of the fault-tolerant construction of single-qubit gates and two-qubit gates in teleportation-based quantum computation. We clearly show that there exists a natural algebraic structure called the teleportation operator in terms of the Bell transform to catch essential points of quantum teleportation, and hence we expect that there would also exist interesting algebraic structures in terms of the GHZ transform to play important roles in quantum information and computation."
                    },
                    {
                        "title": "On the Identifiability of the Post-Nonlinear Causal Model",
                        "abstract": "By taking into account the nonlinear effect of the cause, the inner noise effect, and the measurement distortion effect in the observed variables, the post-nonlinear (PNL) causal model has demonstrated its excellent performance in distinguishing the cause from effect. However, its identifiability has not been properly addressed, and how to apply it in the case of more than two variables is also a problem. In this paper, we conduct a systematic investigation on its identifiability in the two-variable case. We show that this model is identifiable in most cases; by enumerating all possible situations in which the model is not identifiable, we provide sufficient conditions for its identifiability. Simulations are given to support the theoretical results. Moreover, in the case of more than two variables, we show that the whole causal structure can be found by applying the PNL causal model to each structure in the Markov equivalent class and testing if the disturbance is independent of the direct causes for each variable. In this way the exhaustive search over all possible causal structures is avoided."
                    },
                    {
                        "title": "Source Separation and Higher-Order Causal Analysis of MEG and EEG",
                        "abstract": "Separation of the sources and analysis of their connectivity have been an important topic in EEG/MEG analysis. To solve this problem in an automatic manner, we propose a two-layer model, in which the sources are conditionally uncorrelated from each other, but not independent; the dependence is caused by the causality in their time-varying variances (envelopes). The model is identified in two steps. We first propose a new source separation technique which takes into account the autocorrelations (which may be time-varying) and time-varying variances of the sources. The causality in the envelopes is then discovered by exploiting a special kind of multivariate GARCH (generalized autoregressive conditional heteroscedasticity) model. The resulting causal diagram gives the effective connectivity between the separated sources; in our experimental results on MEG data, sources with similar functions are grouped together, with negative influences between groups, and the groups are connected via some interesting sources."
                    },
                    {
                        "title": "Exploring the underlying mechanisms of Xenopus laevis embryonic cell cycle",
                        "abstract": "Cell cycle is an indispensable process in the proliferation and development. Despite significant efforts, global quantification and physical understanding are still challenging. In this study, we explored the mechanisms of Xenopus laevis embryonic cell cycle by quantifying the underlying landscape and flux. We uncovered the irregular Mexican hat landscape of the Xenopus laevis embryonic cell cycle with several local basins and barriers on the oscillation path. The local basins characterize the different phases of Xenopus laevis embryonic cell cycle and the local barriers represent the checkpoints. The checkpoint mechanism of cell cycle is revealed by the landscape basins and barriers. While landscape shape determines the stabilities of the states on the oscillation path, the curl flux force determines the stability of the cell cycle flow. Replication is fundamental for biology of living. From our quantitative study here, we see that replication can not proceed without energy input. In fact, the curl flux originated from energy or nutrition supply determines the speed of the cell cycle and guarantees the progression. Speed of cell cycle is a hallmark of cancer. Through landscape and flux analysis, one can identify the key elements for controlling the speed. This can help to design effective strategy for drug discovery against cancer."
                    },
                    {
                        "title": "Quantum partial search for uneven distribution of multiple target items",
                        "abstract": "Quantum partial search algorithm is approximate search. It aims to find a target block (which has the target items). It runs a little faster than full Grover search. In this paper, we consider quantum partial search algorithm for multiple target items unevenly distributed in database (target blocks have different number of target items). The algorithm we describe can locate one of the target blocks. Efficiency of the algorithm is measured by number of queries to the oracle. We optimize the algorithm in order to improve efficiency. By perturbation method, we find that the algorithm runs the fastest when target items are evenly distributed in database."
                    },
                    {
                        "title": "Quasiprobability fluctuation theorem behind the spread of quantum information",
                        "abstract": "Information spreads in time. For example, correlations dissipate when the correlated system locally couples to a third party, such as the environment. This simple but important fact forms the known quantum data-processing inequality. Here we theoretically uncover the quantum fluctuation theorem behind the quantum informational inequality. The fluctuation theorem quantitatively predicts the statistics of the underlying stochastic quantum process. To fully capture the quantum nature, the fluctuation theorem established here is extended to the quasiprobability regime. We also experimentally apply an interference-based method to measure the amplitudes composing the quasiprobability and verify our established fluctuation theorem by the IBM quantum computer."
                    },
                    {
                        "title": "Costly Evidence and Discretionary Disclosure",
                        "abstract": "A sender flexibly acquires evidence--which she may pay a third party to certify--to disclose to a receiver. When evidence acquisition is overt, the receiver observes the evidence gathering process irrespective of whether its outcome is certified. When acquisition is covert, the receiver does not. In contrast to the case with exogenous evidence, the receiver prefers a strictly positive certification cost. As acquisition costs vanish, equilibria converge to the Pareto-worst free-learning equilibrium. The receiver always prefers covert to overt evidence acquisition."
                    },
                    {
                        "title": "Entanglement versus Bell nonlocality of quantum nonequilibrium steady states",
                        "abstract": "We study the entanglement and the Bell nonlocality of a coupled two-qubit system, in which each qubit is coupled with one individual environment. We study how the nonequilibrium environments (with different temperatures or chemical potentials) influence the entanglement and the Bell nonlocality. The nonequilibrium environments can have constructive effects on the entanglement and the Bell nonlocality. Nonequilibrium thermodynamic cost can sustain the thermal energy or particle current and enhance the entanglement and the Bell nonlocality. However, the nonequilibrium conditions (characterized by the temperature differences or the thermodynamic cost quantified by the entropy production rates) which give the maximal violation of the Bell inequalities are different from the nonequilibrium conditions which give the maximal entanglement. When the Bell inequality has asymmetric observables (between Alice and Bob), for example the $I_{3322}$ inequality, such asymmetry can also be reflected from the effects under the nonequilibrium environments. The spatial asymmetric two-qubit system coupled with nonequilibrium bosonic environments shows the thermal rectification effect, which can be witnessed by the Bell nonlocality. Different spatial asymmetric factors can be linearly cancelled with each other in the thermal rectification effect, which is also reflected on the changes of the entanglement and the Bell nonlocality. Our study demonstrates that the nonequilibrium environments are both valuable for the entanglement and Bell nonlocality resources, based on different optimal nonequilibrium conditions though."
                    },
                    {
                        "title": "Asymmetric steerability of quantum equilibrium and nonequilibrium steady states through entanglement detection",
                        "abstract": "Einstein-Podolsky-Rosen steering describes a quantum correlation in addition to entanglement and Bell nonlocality. However, conceptually different from entanglement and Bell nonlocality, quantum steering has an asymmetric definition. Motivated by the asymmetric definition of quantum steering, we study the steerability of two-interacting qubits, which have asymmetric energy levels, coupled with asymmetric environments. The asymmetric (nonequilibrium) environments are two environments with different temperatures or chemical potentials. The Bloch-Redfield equation is applied to study the dynamics of two qubits and its long-time behavior. In our study, the steady-state steerability is determined by an experimentally friendly steering criteria, which demonstrates steering through the entanglement detection. Our results show that the steady states of two asymmetric qubits have the advantage for one direction of steering, compared to the symmetric setup. We also provide analytical results on the minimal coupling strength between the two qubits in order to be steerable. The asymmetric steerability is collectively determined by the nature of the two qubits and the influence from equilibrium or nonequilibrium environments. Nonequilibrium environments with the cost of nonzero entropy production can enhance the steerability in one direction. We also show the strict hierarchy of entanglement, steering and Bell nonlocality of the nonequilibrium steady states, which shows a richer structure of steering than entanglement and Bell nonlocality."
                    },
                    {
                        "title": "Towards Federated Bayesian Network Structure Learning with Continuous Optimization",
                        "abstract": "Traditionally, Bayesian network structure learning is often carried out at a central site, in which all data is gathered. However, in practice, data may be distributed across different parties (e.g., companies, devices) who intend to collectively learn a Bayesian network, but are not willing to disclose information related to their data owing to privacy or security concerns. In this work, we present a federated learning approach to estimate the structure of Bayesian network from data that is horizontally partitioned across different parties. We develop a distributed structure learning method based on continuous optimization, using the alternating direction method of multipliers (ADMM), such that only the model parameters have to be exchanged during the optimization process. We demonstrate the flexibility of our approach by adopting it for both linear and nonlinear cases. Experimental results on synthetic and real datasets show that it achieves an improved performance over the other methods, especially when there is a relatively large number of clients and each has a limited sample size."
                    }
                ]
            }
        }
    },
    "2410.14030": {
        "paper_data": {
            "title": "Graph Neural Flows for Unveiling Systemic Interactions Among Irregularly Sampled Time Series",
            "url": "http://arxiv.org/abs/2410.14030v1",
            "arxiv_id": "2410.14030",
            "authors": [
                "Giangiacomo Mercatali",
                "Andre Freitas",
                "Jie Chen"
            ],
            "abstract": "Interacting systems are prevalent in nature. It is challenging to accurately predict the dynamics of the system if its constituent components are analyzed independently. We develop a graph-based model that unveils the systemic interactions of time series observed at irregular time points, by using a directed acyclic graph to model the conditional dependencies (a form of causal notation) of the system components and learning this graph in tandem with a continuous-time model that parameterizes the solution curves of ordinary differential equations (ODEs). Our technique, a graph neural flow, leads to substantial enhancements over non-graph-based methods, as well as graph-based methods without the modeling of conditional dependencies. We validate our approach on several tasks, including time series classification and forecasting, to demonstrate its efficacy.",
            "introduction": "   1 Introduction  Real-life dynamical systems consist of a group of components interacting in a complex manner. With time series data for each component, predicting the system dynamics remains challenging, because modeling each component independently is straightforward while accounting for their interactions is hard without a priori knowledge. Sometimes, these interactions are causal. For example, \u201cphantom jams\u201d in which a small disturbance (e.g., a driver hitting the brake too hard) in a heavy traffic can be amplified over a large area of the transportation network\u00a0[16]. While traffic congestion often enjoys spatial proximity, outage of a power network can be propagated non-locally over the grid; i.e., a sequence of blackouts jumps across hundreds of kilometers\u00a0[21]. We pose this question: What time series models best capture the interactive nature of different system dynamics?   Graph neural networks (GNNs)\u00a0[49, 46] are modern tools to enhance time series models when multiple time series are interconnected by a given graph\u00a0[33]. In these models, time series are encoded by using a recurrent neural network (RNN); at every time step, a GNN is used to aggregate features over the graph. When the graph is unknown, graph-structure learning approaches have been proposed; some approaches learn a single interaction graph over time\u00a0[30, 41, 11] while others infer a different one at each time step\u00a0[20]. These models are in general discrete-time models, suitable for regularly spaced time points\u00a0[39, 40].   To handle irregular time points, we consider continuous-time models. The celebrated neural ordinary differential equation (ODE) technique\u00a0[7] models a time series as the solution of an unknown ODE and optimizes the ODE parameters by using gradients computed through the adjoint. Neural ODEs were later adopted for latent variable modeling\u00a0[39, 5], which introduced discontinuities at observed time points for reducing prediction errors and variances. Neural ODEs were also adopted for graph-based modeling\u00a0[38, 24, 25, 26, 10, 27, 3], where the equation accounts for multiple time series and the right-hand side uses a graph to associate the different series. These models construct a graph, which could be time-dependent, in various manners, such as based on node features, co-observations within a sliding window, latent representations, or attentions.   In this work, we propose learning a graph that reveals the dependency structure of the time series. To this end, we consider a form of causal notation\u2014the Bayesian network\u00a0[36, 37]\u2014which is a directed acyclic graph (DAG), where a node is conditionally independent of its non-descendents given its parents\u00a0[34, 12, 43, 22]. Such a conditional dependence structure specifies how component dynamics depends on each other. This model has a potential for causal discovery when one interprets the learned graph unknown a priori (e.g., how blackouts cascade over the power grid, whose known topology differs from the unknown influence graph\u00a0[21]). More importantly, when modeled properly, the graph can improve the performance of downstream tasks because of the capturing of systemic interactions.   Our proposed model is a graph neural flow (GNeuralFlow). A neural flow\u00a0[4] is the learned solution of an unknown ODE based on irregularly sampled time series; it is advantageous over the neural ODE technique in that it models directly the ODE solution rather than the right-hand",
            "references": [
                {
                    "title": "Coordinated Multi-Neighborhood Learning on a Directed Acyclic Graph",
                    "abstract": "Learning the structure of causal directed acyclic graphs (DAGs) is useful in many areas of machine learning and artificial intelligence, with wide applications. However, in the high-dimensional setting, it is challenging to obtain good empirical and theoretical results without strong and often restrictive assumptions. Additionally, it is questionable whether all of the variables purported to be included in the network are observable. It is of interest then to restrict consideration to a subset of the variables for relevant and reliable inferences. In fact, researchers in various disciplines can usually select a set of target nodes in the network for causal discovery. This paper develops a new constraint-based method for estimating the local structure around multiple user-specified target nodes, enabling coordination in structure learning between neighborhoods. Our method facilitates causal discovery without learning the entire DAG structure. We establish consistency results for our algorithm with respect to the local neighborhood structure of the target nodes in the true graph. Experimental results on synthetic and real-world data show that our algorithm is more accurate in learning the neighborhood structures with much less computational cost than standard methods that estimate the entire DAG. An R package implementing our methods may be accessed at https://github.com/stephenvsmith/CML."
                },
                {
                    "title": "Hybrid Top-Down Global Causal Discovery with Local Search for Linear and Nonlinear Additive Noise Models",
                    "abstract": "Learning the unique directed acyclic graph corresponding to an unknown causal model is a challenging task. Methods based on functional causal models can identify a unique graph, but either suffer from the curse of dimensionality or impose strong parametric assumptions. To address these challenges, we propose a novel hybrid approach for global causal discovery in observational data that leverages local causal substructures. We first present a topological sorting algorithm that leverages ancestral relationships in linear structural equation models to establish a compact top-down hierarchical ordering, encoding more causal information than linear orderings produced by existing methods. We demonstrate that this approach generalizes to nonlinear settings with arbitrary noise. We then introduce a nonparametric constraint-based algorithm that prunes spurious edges by searching for local conditioning sets, achieving greater accuracy than current methods. We provide theoretical guarantees for correctness and worst-case polynomial time complexities, with empirical validation on synthetic data."
                },
                {
                    "title": "CF-GODE: Continuous-Time Causal Inference for Multi-Agent Dynamical Systems",
                    "abstract": "Multi-agent dynamical systems refer to scenarios where multiple units (aka agents) interact with each other and evolve collectively over time. For instance, people's health conditions are mutually influenced. Receiving vaccinations not only strengthens the long-term health status of one unit but also provides protection for those in their immediate surroundings. To make informed decisions in multi-agent dynamical systems, such as determining the optimal vaccine distribution plan, it is essential for decision-makers to estimate the continuous-time counterfactual outcomes. However, existing studies of causal inference over time rely on the assumption that units are mutually independent, which is not valid for multi-agent dynamical systems. In this paper, we aim to bridge this gap and study how to estimate counterfactual outcomes in multi-agent dynamical systems. Causal inference in a multi-agent dynamical system has unique challenges: 1) Confounders are time-varying and are present in both individual unit covariates and those of other units; 2) Units are affected by not only their own but also others' treatments; 3) The treatments are naturally dynamic, such as receiving vaccines and boosters in a seasonal manner. To this end, we model a multi-agent dynamical system as a graph and propose a novel model called CF-GODE (C ounterFactual Graph Ordinary Differential Equations). CF-GODE is a causal model that estimates continuous-time counterfactual outcomes in the presence of inter-dependencies between units. To facilitate continuous-time estimation, we propose Treatment-Induced GraphODE, a novel ordinary differential equation based on graph neural networks (GNNs), which can incorporate dynamical treatments as additional inputs to predict potential outcomes over time. To remove confounding bias, we propose two domain adversarial learning based objectives that learn balanced continuous representation trajectories, which are not predictive of treatments and interference. We further provide theoretical justification to prove their effectiveness. Experiments on two semi-synthetic datasets confirm that CF-GODE outperforms baselines on counterfactual estimation. We also provide extensive analyses to understand how our model works."
                },
                {
                    "title": "Inferring dynamic regulatory interaction graphs from time series data with perturbations",
                    "abstract": "Complex systems are characterized by intricate interactions between entities that evolve dynamically over time. Accurate inference of these dynamic relationships is crucial for understanding and predicting system behavior. In this paper, we propose Regulatory Temporal Interaction Network Inference (RiTINI) for inferring time-varying interaction graphs in complex systems using a novel combination of space-and-time graph attentions and graph neural ordinary differential equations (ODEs). RiTINI leverages time-lapse signals on a graph prior, as well as perturbations of signals at various nodes in order to effectively capture the dynamics of the underlying system. This approach is distinct from traditional causal inference networks, which are limited to inferring acyclic and static graphs. In contrast, RiTINI can infer cyclic, directed, and time-varying graphs, providing a more comprehensive and accurate representation of complex systems. The graph attention mechanism in RiTINI allows the model to adaptively focus on the most relevant interactions in time and space, while the graph neural ODEs enable continuous-time modeling of the system's dynamics. We evaluate RiTINI's performance on various simulated and real-world datasets, demonstrating its state-of-the-art capability in inferring interaction graphs compared to previous methods."
                },
                {
                    "title": "Joint Bayesian Inference of Graphical Structure and Parameters with a Single Generative Flow Network",
                    "abstract": "Generative Flow Networks (GFlowNets), a class of generative models over discrete and structured sample spaces, have been previously applied to the problem of inferring the marginal posterior distribution over the directed acyclic graph (DAG) of a Bayesian Network, given a dataset of observations. Based on recent advances extending this framework to non-discrete sample spaces, we propose in this paper to approximate the joint posterior over not only the structure of a Bayesian Network, but also the parameters of its conditional probability distributions. We use a single GFlowNet whose sampling policy follows a two-phase process: the DAG is first generated sequentially one edge at a time, and then the corresponding parameters are picked once the full structure is known. Since the parameters are included in the posterior distribution, this leaves more flexibility for the local probability models of the Bayesian Network, making our approach applicable even to non-linear models parametrized by neural networks. We show that our method, called JSP-GFN, offers an accurate approximation of the joint posterior, while comparing favorably against existing methods on both simulated and real data."
                },
                {
                    "title": "Bayesian learning of Causal Structure and Mechanisms with GFlowNets and Variational Bayes",
                    "abstract": "Bayesian causal structure learning aims to learn a posterior distribution over directed acyclic graphs (DAGs), and the mechanisms that define the relationship between parent and child variables. By taking a Bayesian approach, it is possible to reason about the uncertainty of the causal model. The notion of modelling the uncertainty over models is particularly crucial for causal structure learning since the model could be unidentifiable when given only a finite amount of observational data. In this paper, we introduce a novel method to jointly learn the structure and mechanisms of the causal model using Variational Bayes, which we call Variational Bayes-DAG-GFlowNet (VBG). We extend the method of Bayesian causal structure learning using GFlowNets to learn not only the posterior distribution over the structure, but also the parameters of a linear-Gaussian model. Our results on simulated data suggest that VBG is competitive against several baselines in modelling the posterior over DAGs and mechanisms, while offering several advantages over existing methods, including the guarantee to sample acyclic graphs, and the flexibility to generalize to non-linear causal mechanisms."
                },
                {
                    "title": "Multivariate Time Series Forecasting With Dynamic Graph Neural ODEs",
                    "abstract": "Multivariate time series forecasting has long received significant attention in real-world applications, such as energy consumption and traffic prediction. While recent methods demonstrate good forecasting abilities, they have three fundamental limitations. (i). <italic>Discrete neural architectures:</italic> Interlacing individually parameterized spatial and temporal blocks to encode rich underlying patterns leads to discontinuous latent state trajectories and higher forecasting numerical errors. (ii). <italic>High complexity:</italic> Discrete approaches complicate models with dedicated designs and redundant parameters, leading to higher computational and memory overheads. (iii). <italic>Reliance on graph priors:</italic> Relying on predefined static graph structures limits their effectiveness and practicability in real-world applications. In this paper, we address all the above limitations by proposing a continuous model to forecast <underline><bold>M</bold></underline>ultivariate <underline><bold>T</bold></underline>ime series with dynamic <underline><bold>G</bold></underline>raph neural <underline><bold>O</bold></underline>rdinary <underline><bold>D</bold></underline>ifferential <underline><bold>E</bold></underline>quations (<monospace>MTGODE</monospace>). Specifically, we first abstract multivariate time series into dynamic graphs with time-evolving node features and unknown graph structures. Then, we design and solve a neural ODE to complement missing graph topologies and unify both spatial and temporal message passing, allowing deeper graph propagation and fine-grained temporal information aggregation to characterize stable and precise latent spatial-temporal dynamics. Our experiments demonstrate the superiorities of <monospace>MTGODE</monospace> from various perspectives on five time series benchmark datasets."
                },
                {
                    "title": "Graph-Augmented Normalizing Flows for Anomaly Detection of Multiple Time Series",
                    "abstract": "Anomaly detection is a widely studied task for a broad variety of data types; among them, multiple time series appear frequently in applications, including for example, power grids and traffic networks. Detecting anomalies for multiple time series, however, is a challenging subject, owing to the intricate interdependencies among the constituent series. We hypothesize that anomalies occur in low density regions of a distribution and explore the use of normalizing flows for unsupervised anomaly detection, because of their superior quality in density estimation. Moreover, we propose a novel flow model by imposing a Bayesian network among constituent series. A Bayesian network is a directed acyclic graph (DAG) that models causal relationships; it factorizes the joint probability of the series into the product of easy-to-evaluate conditional probabilities. We call such a graph-augmented normalizing flow approach GANF and propose joint estimation of the DAG with flow parameters. We conduct extensive experiments on real-world datasets and demonstrate the effectiveness of GANF for density estimation, anomaly detection, and identification of time series distribution drift."
                },
                {
                    "title": "Graph Neural Controlled Differential Equations for Traffic Forecasting",
                    "abstract": "Traffic forecasting is one of the most popular spatio-temporal tasks in the field of machine learning. A prevalent approach in the field is to combine graph convolutional networks and recurrent neural networks for the spatio-temporal processing. There has been fierce competition and many novel methods have been proposed. In this paper, we present the method of spatio-temporal graph neural controlled differential equation (STG-NCDE). Neural controlled differential equations (NCDEs) are a breakthrough concept for processing sequential data. We extend the concept and design two NCDEs: one for the temporal processing and the other for the spatial processing. After that, we combine them into a single framework. We conduct experiments with 6 benchmark datasets and 20 baselines. STG-NCDE shows the best accuracy in all cases, outperforming all those 20 baselines by non-trivial margins."
                },
                {
                    "title": "Modeling Irregular Time Series with Continuous Recurrent Units",
                    "abstract": "Recurrent neural networks (RNNs) are a popular choice for modeling sequential data. Modern RNN architectures assume constant time-intervals between observations. However, in many datasets (e.g. medical records) observation times are irregular and can carry important information. To address this challenge, we propose continuous recurrent units (CRUs) -- a neural architecture that can naturally handle irregular intervals between observations. The CRU assumes a hidden state, which evolves according to a linear stochastic differential equation and is integrated into an encoder-decoder framework. The recursive computations of the CRU can be derived using the continuous-discrete Kalman filter and are in closed form. The resulting recurrent architecture has temporal continuity between hidden states and a gating mechanism that can optimally integrate noisy observations. We derive an efficient parameterization scheme for the CRU that leads to a fast implementation f-CRU. We empirically study the CRU on a number of challenging datasets and find that it can interpolate irregular time series better than methods based on neural ordinary differential equations."
                },
                {
                    "title": "Neural Flows: Efficient Alternative to Neural ODEs",
                    "abstract": "Neural ordinary differential equations describe how values change in time. This is the reason why they gained importance in modeling sequential data, especially when the observations are made at irregular intervals. In this paper we propose an alternative by directly modeling the solution curves - the flow of an ODE - with a neural network. This immediately eliminates the need for expensive numerical solvers while still maintaining the modeling capability of neural ODEs. We propose several flow architectures suitable for different applications by establishing precise conditions on when a function defines a valid flow. Apart from computational efficiency, we also provide empirical evidence of favorable generalization performance via applications in time series modeling, forecasting, and density estimation."
                },
                {
                    "title": "Coupled Graph ODE for Learning Interacting System Dynamics",
                    "abstract": "Many real-world systems such as social networks and moving planets are dynamic in nature, where a set of coupled objects are connected via the interaction graph and exhibit complex behavior along the time. For example, the COVID-19 pandemic can be considered as a dynamical system, where objects represent geographical locations (e.g., states) whose daily confirmed cases of infection evolve over time. Outbreak at one location may influence another location as people travel between these locations, forming a graph. Thus, how to model and predict the complex dynamics for these systems becomes a critical research problem. Existing work on modeling graph-structured data mostly assumes a static setting. How to handle dynamic graphs remains to be further explored. On one hand, features of objects change over time, influenced by the linked objects in the interaction graph. On the other hand, the graph itself can also evolve, where new interactions (links) may form and existing links may drop, which may in turn be affected by the dynamic features of objects. In this paper, we propose coupled graph ODE: a novel latent ordinary differential equation (ODE) generative model that learns the coupled dynamics of nodes and edges with a graph neural network (GNN) based ODE in a continuous manner. Our model consists of two coupled ODE functions for modeling the dynamics of edges and nodes based on their latent representations respectively. It employs a novel encoder parameterized by a GNN for inferring the initial states from historical data, which serves as the starting point of the predicted latent trajectories. Experiment results on the COVID-19 dataset and the simulated social network dataset demonstrate the effectiveness of our proposed method."
                },
                {
                    "title": "Discrete Graph Structure Learning for Forecasting Multiple Time Series",
                    "abstract": "Time series forecasting is an extensively studied subject in statistics, economics, and computer science. Exploration of the correlation and causation among the variables in a multivariate time series shows promise in enhancing the performance of a time series model. When using deep neural networks as forecasting models, we hypothesize that exploiting the pairwise information among multiple (multivariate) time series also improves their forecast. If an explicit graph structure is known, graph neural networks (GNNs) have been demonstrated as powerful tools to exploit the structure. In this work, we propose learning the structure simultaneously with the GNN if the graph is unknown. We cast the problem as learning a probabilistic graph model through optimizing the mean performance over the graph distribution. The distribution is parameterized by a neural network so that discrete graphs can be sampled differentiably through reparameterization. Empirical evaluations show that our method is simpler, more efficient, and better performing than a recently proposed bilevel learning approach for graph structure learning, as well as a broad array of forecasting models, either deep or non-deep learning based, and graph or non-graph based."
                },
                {
                    "title": "Learning Continuous System Dynamics from Irregularly-Sampled Partial Observations",
                    "abstract": "Many real-world systems, such as moving planets, can be considered as multi-agent dynamic systems, where objects interact with each other and co-evolve along with the time. Such dynamics is usually difficult to capture, and understanding and predicting the dynamics based on observed trajectories of objects become a critical research problem in many domains. Most existing algorithms, however, assume the observations are regularly sampled and all the objects can be fully observed at each sampling time, which is impractical for many applications. In this paper, we propose to learn system dynamics from irregularly-sampled partial observations with underlying graph structure for the first time. To tackle the above challenge, we present LG-ODE, a latent ordinary differential equation generative model for modeling multi-agent dynamic system with known graph structure. It can simultaneously learn the embedding of high dimensional trajectories and infer continuous latent system dynamics. Our model employs a novel encoder parameterized by a graph neural network that can infer initial states in an unsupervised way from irregularly-sampled partial observations of structural objects and utilizes neuralODE to infer arbitrarily complex continuous-time latent dynamics. Experiments on motion capture, spring system, and charged particle datasets demonstrate the effectiveness of our approach."
                },
                {
                    "title": "On Low-Rank Directed Acyclic Graphs and Causal Structure Learning",
                    "abstract": "Despite several advances in recent years, learning causal structures represented by directed acyclic graphs (DAGs) remains a challenging task in high-dimensional settings when the graphs to be learned are not sparse. In this article, we propose to exploit a low-rank assumption regarding the (weighted) adjacency matrix of a DAG causal model to help address this problem. We utilize existing low-rank techniques to adapt causal structure learning methods to take advantage of this assumption and establish several useful results relating interpretable graphical conditions to the low-rank assumption. Specifically, we show that the maximum rank is highly related to hubs, suggesting that scale-free (SF) networks, which are frequently encountered in practice, tend to be low rank. Our experiments demonstrate the utility of the low-rank adaptations for a variety of data models, especially with relatively large and dense graphs. Moreover, with a validation procedure, the adaptations maintain a superior or comparable performance even when graphs are not restricted to be low rank."
                },
                {
                    "title": "Dynamic Neural Relational Inference",
                    "abstract": "Understanding interactions between entities, e.g., joints of the human body, team sports players, etc., is crucial for tasks like forecasting. However, interactions between entities are commonly not observed and often hard to quantify. To address this challenge, recently, `Neural Relational Inference' was introduced. It predicts static relations between entities in a system and provides an interpretable representation of the underlying system dynamics that are used for better trajectory forecasting. However, generally, relations between entities change as time progresses. Hence, static relations improperly model the data. In response to this, we develop Dynamic Neural Relational Inference (dNRI), which incorporates insights from sequential latent variable models to predict separate relation graphs for every time-step. We demonstrate on several real-world datasets that modeling dynamic relations improves forecasting of complex trajectories."
                },
                {
                    "title": "Normalizing Flows for Probabilistic Modeling and Inference",
                    "abstract": "Normalizing flows provide a general mechanism for defining expressive probability distributions, only requiring the specification of a (usually simple) base distribution and a series of bijective transformations. There has been much recent work on normalizing flows, ranging from improving their expressive power to expanding their application. We believe the field has now matured and is in need of a unified perspective. In this review, we attempt to provide such a perspective by describing flows through the lens of probabilistic modeling and inference. We place special emphasis on the fundamental principles of flow design, and discuss foundational topics such as expressive power and computational trade-offs. We also broaden the conceptual framing of flows by relating them to more general probability transformations. Lastly, we summarize the use of flows for tasks such as generative modeling, approximate inference, and supervised learning."
                },
                {
                    "title": "Graph Neural Ordinary Differential Equations",
                    "abstract": "We introduce the framework of continuous--depth graph neural networks (GNNs). Graph neural ordinary differential equations (GDEs) are formalized as the counterpart to GNNs where the input-output relationship is determined by a continuum of GNN layers, blending discrete topological structures and differential equations. The proposed framework is shown to be compatible with various static and autoregressive GNN models. Results prove general effectiveness of GDEs: in static settings they offer computational advantages by incorporating numerical methods in their forward pass; in dynamic settings, on the other hand, they are shown to improve performance by exploiting the geometry of the underlying dynamics."
                },
                {
                    "title": "Latent ODEs for Irregularly-Sampled Time Series",
                    "abstract": "Time series with non-uniform intervals occur in many applications, and are difficult to model using standard recurrent neural networks (RNNs). We generalize RNNs to have continuous-time hidden dynamics defined by ordinary differential equations (ODEs), a model we call ODE-RNNs. Furthermore, we use ODE-RNNs to replace the recognition network of the recently-proposed Latent ODE model. Both ODE-RNNs and Latent ODEs can naturally handle arbitrary time gaps between observations, and can explicitly model the probability of observation times using Poisson processes. We show experimentally that these ODE-based models outperform their RNN-based counterparts on irregularly-sampled data."
                },
                {
                    "title": "GRU-ODE-Bayes: Continuous modeling of sporadically-observed time series",
                    "abstract": "Modeling real-world multidimensional time series can be particularly challenging when these are sporadically observed (i.e., sampling is irregular both in time and across dimensions)-such as in the case of clinical patient data. To address these challenges, we propose (1) a continuous-time version of the Gated Recurrent Unit, building upon the recent Neural Ordinary Differential Equations (Chen et al., 2018), and (2) a Bayesian update network that processes the sporadic observations. We bring these two ideas together in our GRU-ODE-Bayes method. We then demonstrate that the proposed method encodes a continuity prior for the latent process and that it can exactly represent the Fokker-Planck dynamics of complex processes driven by a multidimensional stochastic differential equation. Additionally, empirical evaluation shows that our method outperforms the state of the art on both synthetic data and real-world data with applications in healthcare and climate forecast. What is more, the continuity prior is shown to be well suited for low number of samples settings."
                },
                {
                    "title": "DAG-GNN: DAG Structure Learning with Graph Neural Networks",
                    "abstract": "Learning a faithful directed acyclic graph (DAG) from samples of a joint distribution is a challenging combinatorial problem, owing to the intractable search space superexponential in the number of graph nodes. A recent breakthrough formulates the problem as a continuous optimization with a structural constraint that ensures acyclicity (Zheng et al., 2018). The authors apply the approach to the linear structural equation model (SEM) and the least-squares loss function that are statistically well justified but nevertheless limited. Motivated by the widespread success of deep learning that is capable of capturing complex nonlinear mappings, in this work we propose a deep generative model and apply a variant of the structural constraint to learn the DAG. At the heart of the generative model is a variational autoencoder parameterized by a novel graph neural network architecture, which we coin DAG-GNN. In addition to the richer capacity, an advantage of the proposed model is that it naturally handles discrete variables as well as vector-valued ones. We demonstrate that on synthetic data sets, the proposed method learns more accurate graphs for nonlinearly generated samples; and on benchmark data sets with discrete variables, the learned graphs are reasonably close to the global optima. The code is available at \\url{this https URL}."
                },
                {
                    "title": "Invertible Residual Networks",
                    "abstract": "We show that standard ResNet architectures can be made invertible, allowing the same model to be used for classification, density estimation, and generation. Typically, enforcing invertibility requires partitioning dimensions or restricting network architectures. In contrast, our approach only requires adding a simple normalization step during training, already available in standard frameworks. Invertible ResNets define a generative model which can be trained by maximum likelihood on unlabeled data. To compute likelihoods, we introduce a tractable approximation to the Jacobian log-determinant of a residual block. Our empirical evaluation shows that invertible ResNets perform competitively with both state-of-the-art image classifiers and flow-based generative models, something that has not been previously achieved with a single architecture."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "DAGs with NO TEARS: Continuous Optimization for Structure Learning",
                    "abstract": "Estimating the structure of directed acyclic graphs (DAGs, also known as Bayesian networks) is a challenging problem since the search space of DAGs is combinatorial and scales superexponentially with the number of nodes. Existing approaches rely on various local heuristics for enforcing the acyclicity constraint. In this paper, we introduce a fundamentally different strategy: we formulate the structure learning problem as a purely continuous optimization problem over real matrices that avoids this combinatorial constraint entirely. This is achieved by a novel characterization of acyclicity that is not only smooth but also exact. The resulting problem can be efficiently solved by standard numerical algorithms, which also makes implementation effortless. The proposed method outperforms existing ones, without imposing any structural assumptions on the graph such as bounded treewidth or in-degree."
                },
                {
                    "title": "Neural Relational Inference for Interacting Systems",
                    "abstract": "Interacting systems are prevalent in nature, from dynamical systems in physics to complex societal dynamics. The interplay of components can give rise to complex behavior, which can often be explained using a simple model of the system's constituent parts. In this work, we introduce the neural relational inference (NRI) model: an unsupervised model that learns to infer interactions while simultaneously learning the dynamics purely from observational data. Our model takes the form of a variational auto-encoder, in which the latent code represents the underlying interaction graph and the reconstruction is based on graph neural networks. In experiments on simulated physical systems, we show that our NRI model can accurately recover ground-truth interactions in an unsupervised manner. We further demonstrate that we can find an interpretable structure and predict complex dynamics in real motion capture and sports tracking data."
                },
                {
                    "title": "DeepMind Control Suite",
                    "abstract": "The DeepMind Control Suite is a set of continuous control tasks with a standardised structure and interpretable rewards, intended to serve as performance benchmarks for reinforcement learning agents. The tasks are written in Python and powered by the MuJoCo physics engine, making them easy to use and modify. We include benchmarks for several learning algorithms. The Control Suite is publicly available at this https URL A video summary of all tasks is available at this http URL ."
                },
                {
                    "title": "Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting",
                    "abstract": "Spatiotemporal forecasting has various applications in neuroscience, climate and transportation domain. Traffic forecasting is one canonical example of such learning task. The task is challenging due to (1) complex spatial dependency on road networks, (2) non-linear temporal dynamics with changing road conditions and (3) inherent difficulty of long-term forecasting. To address these challenges, we propose to model the traffic flow as a diffusion process on a directed graph and introduce Diffusion Convolutional Recurrent Neural Network (DCRNN), a deep learning framework for traffic forecasting that incorporates both spatial and temporal dependency in the traffic flow. Specifically, DCRNN captures the spatial dependency using bidirectional random walks on the graph, and the temporal dependency using the encoder-decoder architecture with scheduled sampling. We evaluate the framework on two real-world large scale road network traffic datasets and observe consistent improvement of 12% - 15% over state-of-the-art baselines."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Density estimation using Real NVP",
                    "abstract": "Unsupervised learning of probabilistic models is a central yet challenging problem in machine learning. Specifically, designing models with tractable learning, sampling, inference and evaluation is crucial in solving this task. We extend the space of such models using real-valued non-volume preserving (real NVP) transformations, a set of powerful invertible and learnable transformations, resulting in an unsupervised learning algorithm with exact log-likelihood computation, exact sampling, exact inference of latent variables, and an interpretable latent space. We demonstrate its ability to model natural images on four datasets through sampling, log-likelihood evaluation and latent variable manipulations."
                },
                {
                    "title": "Cascading Power Outages Propagate Locally in an Influence Graph That is Not the Actual Grid Topology",
                    "abstract": "In a cascading power transmission outage, component outages propagate nonlocally; after one component outages, the next failure may be very distant, both topologically and geographically. As a result, simple models of topological contagion do not accurately represent the propagation of cascades in power systems. However, cascading power outages do follow patterns, some of which are useful in understanding and reducing blackout risk. This paper describes a method by which the data from many cascading failure simulations can be transformed into a graph-based model of influences that provides actionable information about the many ways that cascades propagate in a particular system. The resulting \u201cinfluence graph\u201d model is Markovian, in that component outage probabilities depend only on the outages that occurred in the prior generation. To validate the model, we compare the distribution of cascade sizes resulting from $n-2$ contingencies in a 2896 branch test case to cascade sizes in the influence graph. The two distributions are remarkably similar. In addition, we derive an equation with which one can quickly identify modifications to the proposed system that will substantially reduce cascade propagation. With this equation, one can quickly identify critical components that can be improved to substantially reduce the risk of large cascading blackouts."
                },
                {
                    "title": "On the Properties of Neural Machine Translation: Encoder\u2013Decoder Approaches",
                    "abstract": "Neural machine translation is a relatively new approach to statistical machine translation based purely on neural networks. The neural machine translation models often consist of an encoder and a decoder. The encoder extracts a fixed-length representation from a variable-length input sentence, and the decoder generates a correct translation from this representation. In this paper, we focus on analyzing the properties of the neural machine translation using two models; RNN Encoder--Decoder and a newly proposed gated recursive convolutional neural network. We show that the neural machine translation performs relatively well on short sentences without unknown words, but its performance degrades rapidly as the length of the sentence and the number of unknown words increase. Furthermore, we find that the proposed gated recursive convolutional network learns a grammatical structure of a sentence automatically."
                },
                {
                    "title": "Predicting in-hospital mortality of ICU patients: The PhysioNet/Computing in cardiology challenge 2012",
                    "abstract": "Acuity scores, such as APACHE, SAPS, MPM, and SOFA, are widely used to account for population differences in studies aiming to compare how medications, care guidelines, surgery, and other interventions impact mortality in Intensive Care Unit (ICU) patients. By contrast, the focus of the PhysioNet/CinC Challenge 2012 is to develop methods for patient-specific prediction of in-hospital mortality. The data used for the challenge consisted of 5 general descriptors and 36 time series (measurements of vital signs and laboratory results) from the first 48 hours of the first available ICU stay of 12,000 adult patients from the MIMIC II database. The challenge was organized as two events: event 1 measured performance of a binary classifier, and event 2 measured performance of a risk estimator. The score of event 1 was the lower of sensitivity and positive predictive value. The score for event 2 was a range-normalized Hosmer-Lemeshow statistic. A baseline algorithm (using SAPS-1) obtained event 1 and 2 scores of 0.3125 and 68.58 respectively. Most participants submitted entries that outperformed the baseline algorithm. The top final scores for events 1 and 2 were 0.5353 and 17.88 respectively."
                },
                {
                    "title": "Self-sustained nonlinear waves in traffic flow.",
                    "abstract": "In analogy to gas-dynamical detonation waves, which consist of a shock with an attached exothermic reaction zone, we consider herein nonlinear traveling wave solutions to the hyperbolic (\"inviscid\") continuum traffic equations. Generic existence criteria are examined in the context of the Lax entropy conditions. Our analysis naturally precludes traveling wave solutions for which the shocks travel downstream more rapidly than individual vehicles. Consistent with recent experimental observations from a periodic roadway [Y. Sugiyama, N. J. Phys. 10, 033001 (2008)], our numerical calculations show that nonlinear traveling waves are attracting solutions, with the time evolution of the system converging toward a wave-dominated configuration. Theoretical principles are elucidated by considering examples of traffic flow on open and closed roadways."
                },
                {
                    "title": "Introduction to Smooth Manifolds",
                    "abstract": "Preface.- 1 Smooth Manifolds.- 2 Smooth Maps.- 3 Tangent Vectors.- 4 Submersions, Immersions, and Embeddings.- 5 Submanifolds.- 6 Sard's Theorem.- 7 Lie Groups.- 8 Vector Fields.- 9 Integral Curves and Flows.- 10 Vector Bundles.- 11 The Cotangent Bundle.- 12 Tensors.- 13 Riemannian Metrics.- 14 Differential Forms.- 15 Orientations.- 16 Integration on Manifolds.- 17 De Rham Cohomology.- 18 The de Rham Theorem.- 19 Distributions and Foliations.- 20 The Exponential Map.- 21 Quotient Manifolds.- 22 Symplectic Manifolds.- Appendix A: Review of Topology.- Appendix B: Review of Linear Algebra.- Appendix C: Review of Calculus.- Appendix D: Review of Differential Equations.- References.- Notation Index.- Subject Index"
                },
                {
                    "title": "Large-Sample Learning of Bayesian Networks is NP-Hard",
                    "abstract": "In this paper, we provide new complexity results for algorithms that learn discretevariable Bayesian networks from data. Our results apply whenever the learning algorithm uses a scoring criterion that favors the simplest model able to represent the generative distribution exactly. Our results therefore hold whenever the learning algorithm uses a consistent scoring criterion and is applied to a sufficiently large dataset. We show that identifying high-scoring structures is NP-hard, even when we are given an independence oracle, an inference oracle, and/or an information oracle. Our negative results also apply when learning discrete-variable Bayesian networks in which each node has at most k parents, for all k \u2265 3."
                },
                {
                    "title": "An introduction to structural equation models.",
                    "abstract": "This paper provides an overview of structural equation models, and their potential for advancing neuropsychological theory and practice. Four topics are covered: (1) an overview of the various classes of models, and an introduction to the terminology and diagrams used to describe them, (2) an outline of the steps involved in applying structural equation modeling to any research problem, (3) an overview of the information used in assessing model fit, and a discussion of the role of significance tests in structural models, and (4) an outline of the advantages and disadvantages of structural equation models, and their potential contribution to neuropsychology. The paper is intended to help researchers (1) assess the relevance of these advanced statistical techniques to their own research, and (2) begin the process of successful application."
                },
                {
                    "title": "Correlation and Causation",
                    "abstract": "In social science, we can rarely perform experiments rigorous enough to establish causality. Ironically, it is in just such areas of study that people are prone to rely on correlation to prove causation."
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                },
                {
                    "title": "A Comprehensive Survey on Graph Neural Networks",
                    "abstract": "Deep learning has revolutionized many machine learning tasks in recent years, ranging from image classification and video processing to speech recognition and natural language understanding. The data in these tasks are typically represented in the Euclidean space. However, there is an increasing number of applications, where data are generated from non-Euclidean domains and are represented as graphs with complex relationships and interdependency between objects. The complexity of graph data has imposed significant challenges on the existing machine learning algorithms. Recently, many studies on extending deep learning approaches for graph data have emerged. In this article, we provide a comprehensive overview of graph neural networks (GNNs) in data mining and machine learning fields. We propose a new taxonomy to divide the state-of-the-art GNNs into four categories, namely, recurrent GNNs, convolutional GNNs, graph autoencoders, and spatial\u2013temporal GNNs. We further discuss the applications of GNNs across various domains and summarize the open-source codes, benchmark data sets, and model evaluation of GNNs. Finally, we propose potential research directions in this rapidly growing field."
                },
                {
                    "title": "Causality : Models , Reasoning , and Inference",
                    "abstract": "the methodological community a major statement on causal inquiry. His account of the philosophy and history of causal analysis is a joy to read, especially his spirited 28-page epilogue, \" The Art and Science of Cause and Effect. \" The heart of Pearl's Causality is the set of ideas that he and his colleagues in computer science have developed over the past 20 years. They comprise a unified methodology for the graphical representation of joint probability distributions along with rules for inferring causality directly from such graphical representations. Reading Causality, however, feels a bit (I would imagine) like going for a hike in a rain forest with Judea Pearl as your guide. Initially, one's excitement is aroused by first contact with his causal flora and then further enchanted by his claims of all that lies ahead. But, before long, one is so deeply enveloped in his graphical vegetation that fear of ever finding a way out begins to take over. As I write this review, I must admit that I am still circling about in the counterfactual mangroves of Chapters 7 through 10. Nonetheless, I have come to a recommendation: Anyone who intends to teach causality in graduate methodology classes should read this book. Anyone who does not should avoid it, at least until the rest of us figure out which of its many ideas are unique and lasting contributions to causal analysis. That is to say, from the perspective of a sociologist, the work as a whole is a memorable tour de force, which succeeds admirably in provoking deep thoughts about (1) what social scientists can and should do with available observational data and (2) what sorts of theories and data sources we should be attempting to cultivate. But it is unclear which of Pearl's specific ideas should or"
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nWhat time series models best capture the interactive nature of different system dynamics in real-life dynamical systems?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses the complexities of modeling interactions in dynamical systems, which are prevalent in various fields such as transportation and power networks. By effectively capturing these interactions, future research can advance in areas like causal discovery, system optimization, and predictive analytics. This work could lead to practical applications in traffic management, power grid stability, and other interconnected systems, ultimately improving decision-making and resource allocation.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the intricate interactions between components in dynamical systems, which cannot be modeled independently without losing critical information about their dependencies. Naive approaches may fail because they overlook the causal relationships and temporal dynamics that govern these interactions. Additionally, the need to handle irregular time points complicates the modeling process, as traditional discrete-time models are not suitable. Overcoming these technical obstacles requires sophisticated methodologies that can learn and represent the underlying graph structure of interactions effectively.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on either discrete-time models or has not adequately addressed the complexities of learning interaction graphs in continuous-time settings. Limitations in existing solutions include the inability to capture time-dependent interactions and the reliance on a priori knowledge of the graph structure. Barriers such as the lack of effective graph-structure learning techniques for irregular time series have hindered progress. Our approach differs by proposing a novel model, GNeuralFlow, which learns the dependency structure of time series through a Bayesian network framework, allowing for a more accurate representation of causal relationships.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology, GNeuralFlow, involves learning a graph that reveals the dependency structure of time series data using a Bayesian network. We will utilize irregularly sampled time series datasets and evaluate the model's performance using metrics such as prediction accuracy and causal inference effectiveness. The expected outcomes include improved understanding of systemic interactions in dynamical systems and enhanced predictive capabilities for downstream tasks, leading to better management of complex systems like transportation networks and power grids."
            }
        },
        "author_data": {
            "9dd9563f-f998-4633-a35b-5ee91b596e3d": {
                "pk": "9dd9563f-f998-4633-a35b-5ee91b596e3d",
                "name": "Giangiacomo Mercatali",
                "collaborators": [
                    "Andr\u00e9 Freitas",
                    "Van Khoa Nguyen",
                    "Maciej Falkiewicz",
                    "Alexandros Kalousis",
                    "Danilo S. Carvalho",
                    "Yingji Zhang",
                    "Andre Freitas"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Disentangled Representations",
                    "Generative Models",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Disentangling Generative Factors in Natural Language with Discrete Variational Autoencoders",
                        "abstract": "The ability of learning disentangled representations represents a major step for interpretable NLP systems as it allows latent linguistic features to be controlled. Most approaches to disentanglement rely on continuous variables, both for images and text. We argue that despite being suitable for image datasets, continuous variables may not be ideal to model features of textual data, due to the fact that most generative factors in text are discrete. We propose a Variational Autoencoder based method which models language features as discrete variables and encourages independence between variables for learning disentangled representations. The proposed model outperforms continuous and discrete baselines on several qualitative and quantitative benchmarks for disentanglement as well as on a text style transfer downstream application."
                    },
                    {
                        "title": "MING: A Functional Approach to Learning Molecular Generative Models",
                        "abstract": "Traditional molecule generation methods often rely on sequence or graph-based representations, which can limit their expressive power or require complex permutation-equivariant architectures. This paper introduces a novel paradigm for learning molecule generative models based on functional representations. Specifically, we propose Molecular Implicit Neural Generation (MING), a diffusion-based model that learns molecular distributions in function space. Unlike standard diffusion processes in data space, MING employs a novel functional denoising probabilistic process, which jointly denoises the information in both the function's input and output spaces by leveraging an expectation-maximization procedure for latent implicit neural representations of data. This approach allows for a simple yet effective model design that accurately captures underlying function distributions. Experimental results on molecule-related datasets demonstrate MING's superior performance and ability to generate plausible molecular samples, surpassing state-of-the-art data-space methods while offering a more streamlined architecture and significantly faster generation times."
                    },
                    {
                        "title": "Learning Disentangled Representations for Natural Language Definitions",
                        "abstract": "Disentangling the encodings of neural models is a fundamental aspect for improving interpretability, semantic control and downstream task performance in Natural Language Processing. Currently, most disentanglement methods are unsupervised or rely on synthetic datasets with known generative factors. We argue that recurrent syntactic and semantic regularities in textual data can be used to provide the models with both structural biases and generative factors. We leverage the semantic structures present in a representative and semantically dense category of sentence types, definitional sentences, for training a Variational Autoencoder to learn disentangled representations. Our experimental results show that the proposed model outperforms unsupervised baselines on several qualitative and quantitative benchmarks for disentanglement, and it also improves the results in the downstream task of definition modeling."
                    }
                ]
            },
            "26f72667-d216-468e-9c7b-f88d71a37357": {
                "pk": "26f72667-d216-468e-9c7b-f88d71a37357",
                "name": "Andre Freitas",
                "collaborators": [
                    "Jordan Meadows",
                    "Deborah Ferreira",
                    "Dell Zhang",
                    "Dacheng Tao",
                    "Dawn Song",
                    "Oskar Wysocki",
                    "Malina Florea",
                    "Mauricio Jacobo-Romero",
                    "Danilo S. Carvalho"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Mathematical Language Processing",
                    "Business Process Automation",
                    "Quantitative Reasoning"
                ],
                "publications": [
                    {
                        "title": "A Survey in Mathematical Language Processing",
                        "abstract": "Informal mathematical text underpins real-world quantitative reasoning and communication. Developing sophisticated methods of retrieval and abstraction from this dual modality is crucial in the pursuit of the vision of automating discovery in quantitative science and mathematics. We track the development of informal mathematical language processing approaches across five strategic sub-areas in recent years, highlighting the prevailing successful methodological elements along with existing limitations."
                    },
                    {
                        "title": "Natural Language Premise Selection: Finding Supporting Statements for Mathematical Text",
                        "abstract": "Mathematical text is written using a combination of words and mathematical expressions. This combination, along with a specific way of structuring sentences makes it challenging for state-of-art NLP tools to understand and reason on top of mathematical discourse. In this work, we propose a new NLP task, the natural premise selection, which is used to retrieve supporting definitions and supporting propositions that are useful for generating an informal mathematical proof for a particular statement. We also make available a dataset, NL-PS, which can be used to evaluate different approaches for the natural premise selection task. Using different baselines, we demonstrate the underlying interpretation challenges associated with the task."
                    },
                    {
                        "title": "Proceedings of the AAAI-20 Workshop on Intelligent Process Automation (IPA-20)",
                        "abstract": "This is the Proceedings of the AAAI-20 Workshop on Intelligent Process Automation (IPA-20) which took place in New York, NY, USA on February 7th 2020."
                    },
                    {
                        "title": "What is SemEval evaluating? A Systematic Analysis of Evaluation Campaigns in NLP",
                        "abstract": "SemEval is the primary venue in the NLP community for the proposal of new challenges and for the systematic empirical evaluation of NLP systems. This paper provides a systematic quantitative analysis of SemEval aiming to evidence the patterns of the contributions behind SemEval. By understanding the distribution of task types, metrics, architectures, participation and citations over time we aim to answer the question on what is being evaluated by SemEval."
                    },
                    {
                        "title": "Analysis of business process automation as linear time-invariant system network",
                        "abstract": "In this work, we examined Business Process (BP) production as a signal; this novel approach explores a BP workflow as a linear time-invariant (LTI) system. We analysed BP productivity in the frequency domain; this standpoint examines how labour and capital act as BP input signals and how their fundamental frequencies affect BP production. Our research also proposes a simulation framework of a BP in the frequency domain for estimating productivity gains due to the introduction of automation steps. Our ultimate goal was to supply evidence to address Solow's Paradox."
                    }
                ]
            },
            "e8e8ce31-5709-4990-b456-1cb424b0cb3b": {
                "pk": "e8e8ce31-5709-4990-b456-1cb424b0cb3b",
                "name": "Jie Chen",
                "collaborators": [
                    "Yang",
                    "Boling Guo",
                    "Jie Shao",
                    "Baoxiang Wang",
                    "Yang Yang",
                    "Ronny Luss"
                ],
                "domain": [
                    "Graph Theory",
                    "Machine Learning",
                    "Knot Theory",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "On the algebraic Gordian distance",
                        "abstract": "Using Blanchfield pairings, we show that two Alexander polynomials cannot be realized by a pair of matrices with Gordian distance one if a corresponding quadratic equation does not have an integer solution. We also give an example of how our results help in calculating the Gordian distances, algebraic Gordian distances and polynomial distances."
                    },
                    {
                        "title": "Connected sum and crossing numbers of flat virtual knots",
                        "abstract": "The connected sum of two flat virtual knots depends on the choice of diagrams and basepoints. We show that any minimal crossing diagram of a composite flat virtual knot is a connected sum diagram. We also show the crossing number of flat virtual knots is super-additive under connected sum."
                    },
                    {
                        "title": "A Bounded Multi-Vacation Queue Model for Multi-stage Sleep Control 5G Base station",
                        "abstract": "Modelling and control of energy consumption is an important problem in telecommunication systems.To model such systems, this paper publishes a bounded multi-vacation queue model. The energy consumption predicted by the model shows an average error rate of 0.0177 and the delay predicted by the model shows an average error rate of 0.0655 over 99 test instances.Subsequently, an optimization algorithm is proposed to minimize the energy consumption while not violate the delay bound. Furthermore, given current state of art 5G base station system configuration, numerical results shows that with the increase of traffic load, energy saving rate becomes less."
                    },
                    {
                        "title": "Graph Neural Preconditioners for Iterative Solutions of Sparse Linear Systems",
                        "abstract": "Preconditioning is at the heart of iterative solutions of large, sparse linear systems of equations in scientific disciplines. Several algebraic approaches, which access no information beyond the matrix itself, are widely studied and used, but ill-conditioned matrices remain very challenging. We take a machine learning approach and propose using graph neural networks as a general-purpose preconditioner. They show attractive performance for many problems and can be used when the mainstream preconditioners perform poorly. Empirical evaluation on over 800 matrices suggests that the construction time of these graph neural preconditioners (GNPs) is more predictable and can be much shorter than that of other widely used ones, such as ILU and AMG, while the execution time is faster than using a Krylov method as the preconditioner, such as in inner-outer GMRES. GNPs have a strong potential for solving large-scale, challenging algebraic problems arising from not only partial differential equations, but also economics, statistics, graph, and optimization, to name a few."
                    },
                    {
                        "title": "FlatKnotInfo: the first 1.24 million flat knots",
                        "abstract": "We use matchings on Lyndon words to classify flat knots up to 8 crossings. Using flat knots invariants such as the based matrix, the $\\phi$-invariant, the flat arrow polynomial, and the flat Jones-Krushkal polynomial, we distinguish all flat knots up to 7 crossings except for five pairs. Among the many flat knots considered, we find examples that are: (i) algebraically slice but not slice; (ii) almost classical (null-homologous) but not slice; (iii) nontrivial but with trivial (primitive) based matrix.   The classification data has been curated and is available on FlatKnotInfo, which is an interactive searchable website listing flat knots up to 8 crossings and their invariants. It also provides access to algebraic and diagrammatic information on these knots and is designed to enable users to discover patterns and formulate conjectures on their own."
                    },
                    {
                        "title": "Efficient algorithms for collecting the statistics of large-scale IP address data",
                        "abstract": "Compiling the statistics of large-scale IP address data is an essential task in network traffic measurement. The statistical results are used to evaluate the potential impact of user behaviors on network traffic. This requires algorithms that are capable of storing and retrieving a high volume of IP addresses within time and memory constraints. In this paper, we present two efficient algorithms for collecting the statistics of large-scale IP addresses that balance time efficiency and memory consumption. The proposed solutions take into account the sparse nature of the statistics of IP addresses while building the hash function and maintain a dynamic balance among layered memory blocks. There are two layers in the first proposed method, each of which contains a limited number of memory blocks. Each memory block contains 256 elements of size $256 \\times 8$ bytes for a 64-bit system. In contrast to built-in hash mapping functions, the proposed solution completely avoids expensive hash collisions while retaining the linear time complexity of hash-based solutions. Moreover, the mechanism dynamically determines the hash index length according to the range of IP addresses, and can balance the time and memory constraints. In addition, we propose an efficient parallel scheme to speed up the collection of statistics. The experimental results on several synthetic datasets show that the proposed method substantially outperforms the baselines with respect to time and memory space efficiency."
                    },
                    {
                        "title": "Inverse Problem of Electro-seismic Conversion",
                        "abstract": "When a porous rock is saturated with an electrolyte, electrical fields are coupled with seismic waves via the electro-seismic conversion. Pride derived the governing models, in which Maxwell equations are coupled with Biot equations through the electro-kinetic mobility parameter. The inverse problem of the linearized electro-seismic conversion consists in two step, namely the inversion of Biot equations and the inversion of Maxwell equations. We analyze the reconstruction of conductivity and electro-kinetic mobility parameter in Maxwell equations with internal measurements, while the internal measurements are provided by the results of the inversion of Biot equations. We show that knowledge of two internal data based on well-chosen boundary conditions uniquely determine these two parameters. Moreover, a Lipschitz type stability is proved based on the same sets of well-chosen boundary conditions."
                    },
                    {
                        "title": "Wellposedness and scattering for the generalized Boussinesq equation",
                        "abstract": "In this paper, we show the local well-posedness of the generalized Boussinesq equation(gBQ) in $L^{2}(\\mathbb{R}^d), H^{1}(\\mathbb{R}^d)$ and obtain the global well-posedness, finite-time blowup and small initial data scattering of gBQ in energy space $H^1(\\mathbb{R}^d)$. Moreover, we obtain the large radial initial data scattering of defocusing case for $ d\\geq 3 $ by using the method of Dodson-Murphy."
                    },
                    {
                        "title": "Almost sure scattering for the nonlinear Klein-Gordon equations with Sobolev critical power",
                        "abstract": "In this paper, we study the almost sure scattering for the Klein-Gordon equations with Sobolev critical power. We obtain the almost sure scattering with random initial data in $H^s \\times H^{s-1}$; $11/12 < s < 1$ for $d = 4$, $15/16 < s < 1$ for $d = 5$. We use the induction on scales and bushes argument in [9] where the model equation is wave equation. For d = 5, we use the mass term of the Klein-Gordon equation to obtain the control of the increment of energy in the process of induction on scales."
                    },
                    {
                        "title": "Quantitative Photo-acoustic Tomography with Partial Data",
                        "abstract": "Photo-acoustic tomography is a newly developed hybrid imaging modality that combines a high-resolution modality with a high-contrast modality. We analyze the reconstruction of diffusion and absorption parameters in an elliptic equation and improve an earlier result of Bal and Uhlmann to the partial date case. We show that the reconstruction can be uniquely determined by the knowledge of 4 internal data based on well-chosen partial boundary conditions. Stability of this reconstruction is ensured if a convexity condition is satisfied. Similar stability result is obtained without this geometric constraint if 4n well-chosen partial boundary conditions are available, where $n$ is the spatial dimension. The set of well-chosen boundary measurements is characterized by some complex geometric optics (CGO) solutions vanishing on a part of the boundary."
                    },
                    {
                        "title": "Stochastic Gradient Descent with Biased but Consistent Gradient Estimators",
                        "abstract": "Stochastic gradient descent (SGD), which dates back to the 1950s, is one of the most popular and effective approaches for performing stochastic optimization. Research on SGD resurged recently in machine learning for optimizing convex loss functions and training nonconvex deep neural networks. The theory assumes that one can easily compute an unbiased gradient estimator, which is usually the case due to the sample average nature of empirical risk minimization. There exist, however, many scenarios (e.g., graphs) where an unbiased estimator may be as expensive to compute as the full gradient because training examples are interconnected. Recently, Chen et al. (2018) proposed using a consistent gradient estimator as an economic alternative. Encouraged by empirical success, we show, in a general setting, that consistent estimators result in the same convergence behavior as do unbiased ones. Our analysis covers strongly convex, convex, and nonconvex objectives. We verify the results with illustrative experiments on synthetic and real-world data. This work opens several new research directions, including the development of more efficient SGD updates with consistent estimators and the design of efficient training algorithms for large-scale graphs."
                    }
                ]
            }
        }
    },
    "2406.03679": {
        "paper_data": {
            "title": "On the Effects of Data Scale on Computer Control Agents",
            "url": "http://arxiv.org/abs/2406.03679v4",
            "arxiv_id": "2406.03679",
            "authors": [
                "Wei Li",
                "William Bishop",
                "Alice Li",
                "Chris Rawles",
                "Folawiyo Campbell-Ajala",
                "Divya Tyamagundlu",
                "Oriana Riva"
            ],
            "abstract": "Autonomous agents that control computer interfaces to accomplish human tasks are emerging. Leveraging LLMs to power such agents has been of special interest, but unless fine-tuned on human-collected task demonstrations, performance is still relatively low. In this work we study whether fine-tuning alone is a viable approach for building real-world computer control agents. In particularly, we investigate how performance measured on both high and low-level tasks in domain and out of domain scales as more training data is collected. To this end we collect and release a new dataset, AndroidControl, consisting of 15,283 demonstrations of everyday tasks with Android apps. Compared to existing datasets, each AndroidControl task instance includes both high and low-level human-generated instructions, allowing us to explore the level of task complexity an agent can handle. Moreover, AndroidControl is the most diverse computer control dataset to date, including 15,283 unique tasks over 833 Android apps, thus allowing us to conduct in-depth analysis of the model performance in and out of the domain of the training data. Using the dataset, we find that when tested in domain fine-tuned models outperform zero and few-shot baselines and scale in such a way that robust performance might feasibly be obtained simply by collecting more data. Out of domain, performance scales significantly more slowly and suggests that in particular for high-level tasks, fine-tuning on more data alone may be insufficient for achieving robust out-of-domain performance.",
            "introduction": "   1 Introduction  Recent work has studied how large language models (LLMs) can be leveraged to build computer control agents\u00a0[17, 42, 38, 16, 6] that accomplish human tasks by interacting with a computer environment. These agents perceive the state of the computer by observing its screen (from screenshots or application UI trees), and generate actions (click, type, scroll, etc.) that are executed through the computer\u2019s user interface. Tasks, specified in natural language, can range from configuring device settings and sending emails, to navigating shopping websites and planning a trip.   While progress is rapidly advancing, absolute performance of computer control agents that leverage pre-trained LLMs without fine-tuning on task demonstrations is still relatively low. When tested in real-world environments, where agents control everyday applications and websites, recently-reported task success rates range from 12% on desktop applications\u00a0[37] to 46% on mobile applications\u00a0[3]. In contrast, agents that leverage models fine-tuned for task execution\u00a0[24, 9], achieve even 80%\u00a0[9] success rate, when tested on websites and tasks similar to what they are trained on.   While the pattern of collecting new datasets and fine-tuning shows promise, there are at least two important unanswered questions. First, to the best of our knowledge no prior work has examined the question of scaling: how much data must be collected in order to obtain a given performance level with fine-tuned models. This question is particularly important because human demonstrations of computer interactions for fine-tuning are time consuming and expensive to collect. Understanding how performance scales, both in domain and out of the domain of the collected demonstrations (unseen tasks and unseen applications), is important for determining whether fine-tuning alone is a viable path towards deploying computer control agents in the real world. Therefore, one of the main goals of this work is to rigorously quantify how performance of fine-tuned agents scales, both in and out of domain, as the amount of data used for fine-tuning is increased.   Second, it is not clear the level of task complexity fine-tuning might be fruitfully applied to. Conceptually, computer control agents must both decompose a high-level goal into a set of small atomic actions and execute (\u201cground\u201d) those actions in a device screen. While the high-level reasoning with LLMs, required for determining how to accomplish high-level goals, is still an open problem in artificial intelligence\u00a0[34, 33, 41, 44, 18, 40, 35], the set of low-level actions (clicking, typing, etc\u2026) required to execute tasks are more constrained, and general agents capable of robust grounding across domains might be approachable via fine-tuning. Therefore, a second goal of this work is to quantify the scaling of fine-tuning for agents performing both high-level and low-level tasks.   Figure 1: An example task demonstration contained in AndroidControl. Green circles/rectangles highlight the action on the screen. Red numbers are added only for illustration purposes.   Rigorously quantifying scaling in these ways requires a carefully constructed dataset. To this end, we introduce AndroidControl, a large-scale dataset of 15,283 demonstrations of tasks performed by humans in Android apps. Figure\u00a01 shows an example data sample. Compared to existing datasets\u00a0[6, 26], for every task AndroidControl provides both the high- and low-level human-generated instructions describing",
            "references": [
                {
                    "title": "WebVoyager: Building an End-to-End Web Agent with Large Multimodal Models",
                    "abstract": "The rapid advancement of large language models (LLMs) has led to a new era marked by the development of autonomous applications in real-world scenarios, which drives innovation in creating advanced web agents. Existing web agents typically only handle one input modality and are evaluated only in simplified web simulators or static web snapshots, greatly limiting their applicability in real-world scenarios. To bridge this gap, we introduce WebVoyager, an innovative Large Multimodal Model (LMM) powered web agent that can complete user instructions end-to-end by interacting with real-world websites. Moreover, we establish a new benchmark by compiling real-world tasks from 15 popular websites and introduce an automatic evaluation protocol leveraging multimodal understanding abilities of GPT-4V to evaluate open-ended web agents. We show that WebVoyager achieves a 59.1% task success rate on our benchmark, significantly surpassing the performance of both GPT-4 (All Tools) and the WebVoyager (text-only) setups, underscoring the exceptional capability of WebVoyager. The proposed automatic evaluation metric achieves 85.3% agreement with human judgment, indicating its effectiveness in providing reliable and accurate assessments of web agents."
                },
                {
                    "title": "GPT-4V(ision) is a Generalist Web Agent, if Grounded",
                    "abstract": "The recent development on large multimodal models (LMMs), especially GPT-4V(ision) and Gemini, has been quickly expanding the capability boundaries of multimodal models beyond traditional tasks like image captioning and visual question answering. In this work, we explore the potential of LMMs like GPT-4V as a generalist web agent that can follow natural language instructions to complete tasks on any given website. We propose SEEACT, a generalist web agent that harnesses the power of LMMs for integrated visual understanding and acting on the web. We evaluate on the recent MIND2WEB benchmark. In addition to standard offline evaluation on cached websites, we enable a new online evaluation setting by developing a tool that allows running web agents on live websites. We show that GPT-4V presents a great potential for web agents -- it can successfully complete 51.1 of the tasks on live websites if we manually ground its textual plans into actions on the websites. This substantially outperforms text-only LLMs like GPT-4 or smaller models (FLAN-T5 and BLIP-2) specifically fine-tuned for web agents. However, grounding still remains a major challenge. Existing LMM grounding strategies like set-of-mark prompting turns out to be not effective for web agents, and the best grounding strategy we develop in this paper leverages both the HTML structure and visuals. Yet, there is still a substantial gap with oracle grounding, leaving ample room for further improvement. All code, data, and evaluation tools are available at https://github.com/OSU-NLP-Group/SeeAct."
                },
                {
                    "title": "UINav: A Practical Approach to Train On-Device Automation Agents",
                    "abstract": "Automation systems that can autonomously drive application user interfaces to complete user tasks are of great benefit, especially when users are situationally or permanently impaired. Prior automation systems do not produce generalizable models while AI-based automation agents work reliably only in simple, hand-crafted applications or incur high computation costs. We propose UINav, a demonstration-based approach to train automation agents that fit mobile devices, yet achieving high success rates with modest numbers of demonstrations. To reduce the demonstration overhead, UINav uses a referee model that provides users with immediate feedback on tasks where the agent fails, and automatically augments human demonstrations to increase diversity in training data. Our evaluation shows that with only 10 demonstrations UINav can achieve 70% accuracy, and that with enough demonstrations it can surpass 90% accuracy."
                },
                {
                    "title": "Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting",
                    "abstract": "As large language models (LLMs) are adopted as a fundamental component of language technologies, it is crucial to accurately characterize their performance. Because choices in prompt design can strongly influence model behavior, this design process is critical in effectively using any modern pre-trained generative language model. In this work, we focus on LLM sensitivity to a quintessential class of meaning-preserving design choices: prompt formatting. We find that several widely used open-source LLMs are extremely sensitive to subtle changes in prompt formatting in few-shot settings, with performance differences of up to 76 accuracy points when evaluated using LLaMA-2-13B. Sensitivity remains even when increasing model size, the number of few-shot examples, or performing instruction tuning. Our analysis suggests that work evaluating LLMs with prompting-based methods would benefit from reporting a range of performance across plausible prompt formats, instead of the currently-standard practice of reporting performance on a single format. We also show that format performance only weakly correlates between models, which puts into question the methodological validity of comparing models with an arbitrarily chosen, fixed prompt format. To facilitate systematic analysis we propose FormatSpread, an algorithm that rapidly evaluates a sampled set of plausible prompt formats for a given task, and reports the interval of expected performance without accessing model weights. Furthermore, we present a suite of analyses that characterize the nature of this sensitivity, including exploring the influence of particular atomic perturbations and the internal representation of particular formats."
                },
                {
                    "title": "WebArena: A Realistic Web Environment for Building Autonomous Agents",
                    "abstract": "With advances in generative AI, there is now potential for autonomous agents to manage daily tasks via natural language commands. However, current agents are primarily created and tested in simplified synthetic environments, leading to a disconnect with real-world scenarios. In this paper, we build an environment for language-guided agents that is highly realistic and reproducible. Specifically, we focus on agents that perform tasks on the web, and create an environment with fully functional websites from four common domains: e-commerce, social forum discussions, collaborative software development, and content management. Our environment is enriched with tools (e.g., a map) and external knowledge bases (e.g., user manuals) to encourage human-like task-solving. Building upon our environment, we release a set of benchmark tasks focusing on evaluating the functional correctness of task completions. The tasks in our benchmark are diverse, long-horizon, and designed to emulate tasks that humans routinely perform on the internet. We experiment with several baseline agents, integrating recent techniques such as reasoning before acting. The results demonstrate that solving complex tasks is challenging: our best GPT-4-based agent only achieves an end-to-end task success rate of 14.41%, significantly lower than the human performance of 78.24%. These results highlight the need for further development of robust agents, that current state-of-the-art large language models are far from perfect performance in these real-life tasks, and that WebArena can be used to measure such progress."
                },
                {
                    "title": "A Real-World WebAgent with Planning, Long Context Understanding, and Program Synthesis",
                    "abstract": "Pre-trained large language models (LLMs) have recently achieved better generalization and sample efficiency in autonomous web automation. However, the performance on real-world websites has still suffered from (1) open domainness, (2) limited context length, and (3) lack of inductive bias on HTML. We introduce WebAgent, an LLM-driven agent that learns from self-experience to complete tasks on real websites following natural language instructions. WebAgent plans ahead by decomposing instructions into canonical sub-instructions, summarizes long HTML documents into task-relevant snippets, and acts on websites via Python programs generated from those. We design WebAgent with Flan-U-PaLM, for grounded code generation, and HTML-T5, new pre-trained LLMs for long HTML documents using local and global attention mechanisms and a mixture of long-span denoising objectives, for planning and summarization. We empirically demonstrate that our modular recipe improves the success on real websites by over 50%, and that HTML-T5 is the best model to solve various HTML understanding tasks; achieving 18.7% higher success rate than the prior method on MiniWoB web automation benchmark, and SoTA performance on Mind2Web, an offline task planning evaluation."
                },
                {
                    "title": "Android in the Wild: A Large-Scale Dataset for Android Device Control",
                    "abstract": "There is a growing interest in device-control systems that can interpret human natural language instructions and execute them on a digital device by directly controlling its user interface. We present a dataset for device-control research, Android in the Wild (AITW), which is orders of magnitude larger than current datasets. The dataset contains human demonstrations of device interactions, including the screens and actions, and corresponding natural language instructions. It consists of 715k episodes spanning 30k unique instructions, four versions of Android (v10-13),and eight device types (Pixel 2 XL to Pixel 6) with varying screen resolutions. It contains multi-step tasks that require semantic understanding of language and visual context. This dataset poses a new challenge: actions available through the user interface must be inferred from their visual appearance. And, instead of simple UI element-based actions, the action space consists of precise gestures (e.g., horizontal scrolls to operate carousel widgets). We organize our dataset to encourage robustness analysis of device-control systems, i.e., how well a system performs in the presence of new task descriptions, new applications, or new platform versions. We develop two agents and report performance across the dataset. The dataset is available at https://github.com/google-research/google-research/tree/master/android_in_the_wild."
                },
                {
                    "title": "Synapse: Trajectory-as-Exemplar Prompting with Memory for Computer Control",
                    "abstract": "Building agents with large language models (LLMs) for computer control is a burgeoning research area, where the agent receives computer states and performs actions to complete complex tasks. Previous computer agents have demonstrated the benefits of in-context learning (ICL); however, their performance is hindered by several issues. First, the limited context length of LLMs and complex computer states restrict the number of exemplars, as a single webpage can consume the entire context. Second, the exemplars in current methods, such as high-level plans and multi-choice questions, cannot represent complete trajectories, leading to suboptimal performance in long-horizon tasks. Third, existing computer agents rely on task-specific exemplars and overlook the similarity among tasks, resulting in poor generalization to novel tasks. To address these challenges, we introduce Synapse, a computer agent featuring three key components: i) state abstraction, which filters out task-irrelevant information from raw states, allowing more exemplars within the limited context, ii) trajectory-as-exemplar prompting, which prompts the LLM with complete trajectories of the abstracted states and actions to improve multi-step decision-making, and iii) exemplar memory, which stores the embeddings of exemplars and retrieves them via similarity search for generalization to novel tasks. We evaluate Synapse on MiniWoB++, a standard task suite, and Mind2Web, a real-world website benchmark. In MiniWoB++, Synapse achieves a 99.2% average success rate (a 10% relative improvement) across 64 tasks using demonstrations from only 48 tasks. Notably, Synapse is the first ICL method to solve the book-flight task in MiniWoB++. Synapse also exhibits a 56% relative improvement in average step success rate over the previous state-of-the-art prompting scheme in Mind2Web."
                },
                {
                    "title": "Tree of Thoughts: Deliberate Problem Solving with Large Language Models",
                    "abstract": "Language models are increasingly being deployed for general problem solving across a wide range of tasks, but are still confined to token-level, left-to-right decision-making processes during inference. This means they can fall short in tasks that require exploration, strategic lookahead, or where initial decisions play a pivotal role. To surmount these challenges, we introduce a new framework for language model inference, Tree of Thoughts (ToT), which generalizes over the popular Chain of Thought approach to prompting language models, and enables exploration over coherent units of text (thoughts) that serve as intermediate steps toward problem solving. ToT allows LMs to perform deliberate decision making by considering multiple different reasoning paths and self-evaluating choices to decide the next course of action, as well as looking ahead or backtracking when necessary to make global choices. Our experiments show that ToT significantly enhances language models' problem-solving abilities on three novel tasks requiring non-trivial planning or search: Game of 24, Creative Writing, and Mini Crosswords. For instance, in Game of 24, while GPT-4 with chain-of-thought prompting only solved 4% of tasks, our method achieved a success rate of 74%. Code repo with all prompts: https://github.com/princeton-nlp/tree-of-thought-llm."
                },
                {
                    "title": "Language Models can Solve Computer Tasks",
                    "abstract": "Agents capable of carrying out general tasks on a computer can improve efficiency and productivity by automating repetitive tasks and assisting in complex problem-solving. Ideally, such agents should be able to solve new computer tasks presented to them through natural language commands. However, previous approaches to this problem require large amounts of expert demonstrations and task-specific reward functions, both of which are impractical for new tasks. In this work, we show that a pre-trained large language model (LLM) agent can execute computer tasks guided by natural language using a simple prompting scheme where the agent Recursively Criticizes and Improves its output (RCI). The RCI approach significantly outperforms existing LLM methods for automating computer tasks and surpasses supervised learning (SL) and reinforcement learning (RL) approaches on the MiniWoB++ benchmark. We compare multiple LLMs and find that RCI with the InstructGPT-3+RLHF LLM is state-of-the-art on MiniWoB++, using only a handful of demonstrations per task rather than tens of thousands, and without a task-specific reward function. Furthermore, we demonstrate RCI prompting's effectiveness in enhancing LLMs' reasoning abilities on a suite of natural language reasoning tasks, outperforming chain of thought (CoT) prompting with external feedback. We find that RCI combined with CoT performs better than either separately. Our code can be found here: https://github.com/posgnu/rci-agent."
                },
                {
                    "title": "Reflexion: language agents with verbal reinforcement learning",
                    "abstract": "Large language models (LLMs) have been increasingly used to interact with external environments (e.g., games, compilers, APIs) as goal-driven agents. However, it remains challenging for these language agents to quickly and efficiently learn from trial-and-error as traditional reinforcement learning methods require extensive training samples and expensive model fine-tuning. We propose Reflexion, a novel framework to reinforce language agents not by updating weights, but instead through linguistic feedback. Concretely, Reflexion agents verbally reflect on task feedback signals, then maintain their own reflective text in an episodic memory buffer to induce better decision-making in subsequent trials. Reflexion is flexible enough to incorporate various types (scalar values or free-form language) and sources (external or internally simulated) of feedback signals, and obtains significant improvements over a baseline agent across diverse tasks (sequential decision-making, coding, language reasoning). For example, Reflexion achieves a 91% pass@1 accuracy on the HumanEval coding benchmark, surpassing the previous state-of-the-art GPT-4 that achieves 80%. We also conduct ablation and analysis studies using different feedback signals, feedback incorporation methods, and agent types, and provide insights into how they affect performance."
                },
                {
                    "title": "ReAct: Synergizing Reasoning and Acting in Language Models",
                    "abstract": "While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources, such as knowledge bases or environments, to gather additional information. We apply our approach, named ReAct, to a diverse set of language and decision making tasks and demonstrate its effectiveness over state-of-the-art baselines, as well as improved human interpretability and trustworthiness over methods without reasoning or acting components. Concretely, on question answering (HotpotQA) and fact verification (Fever), ReAct overcomes issues of hallucination and error propagation prevalent in chain-of-thought reasoning by interacting with a simple Wikipedia API, and generates human-like task-solving trajectories that are more interpretable than baselines without reasoning traces. On two interactive decision making benchmarks (ALFWorld and WebShop), ReAct outperforms imitation and reinforcement learning methods by an absolute success rate of 34% and 10% respectively, while being prompted with only one or two in-context examples. Project site with code: https://react-lm.github.io"
                },
                {
                    "title": "Large Language Models are Zero-Shot Reasoners",
                    "abstract": "Pretrained large language models (LLMs) are widely used in many sub-fields of natural language processing (NLP) and generally known as excellent few-shot learners with task-specific exemplars. Notably, chain of thought (CoT) prompting, a recent technique for eliciting complex multi-step reasoning through step-by-step answer examples, achieved the state-of-the-art performances in arithmetics and symbolic reasoning, difficult system-2 tasks that do not follow the standard scaling laws for LLMs. While these successes are often attributed to LLMs' ability for few-shot learning, we show that LLMs are decent zero-shot reasoners by simply adding\"Let's think step by step\"before each answer. Experimental results demonstrate that our Zero-shot-CoT, using the same single prompt template, significantly outperforms zero-shot LLM performances on diverse benchmark reasoning tasks including arithmetics (MultiArith, GSM8K, AQUA-RAT, SVAMP), symbolic reasoning (Last Letter, Coin Flip), and other logical reasoning tasks (Date Understanding, Tracking Shuffled Objects), without any hand-crafted few-shot examples, e.g. increasing the accuracy on MultiArith from 17.7% to 78.7% and GSM8K from 10.4% to 40.7% with large InstructGPT model (text-davinci-002), as well as similar magnitudes of improvements with another off-the-shelf large model, 540B parameter PaLM. The versatility of this single prompt across very diverse reasoning tasks hints at untapped and understudied fundamental zero-shot capabilities of LLMs, suggesting high-level, multi-task broad cognitive capabilities may be extracted by simple prompting. We hope our work not only serves as the minimal strongest zero-shot baseline for the challenging reasoning benchmarks, but also highlights the importance of carefully exploring and analyzing the enormous zero-shot knowledge hidden inside LLMs before crafting finetuning datasets or few-shot exemplars."
                },
                {
                    "title": "Least-to-Most Prompting Enables Complex Reasoning in Large Language Models",
                    "abstract": "Chain-of-thought prompting has demonstrated remarkable performance on various natural language reasoning tasks. However, it tends to perform poorly on tasks which requires solving problems harder than the exemplars shown in the prompts. To overcome this challenge of easy-to-hard generalization, we propose a novel prompting strategy, least-to-most prompting. The key idea in this strategy is to break down a complex problem into a series of simpler subproblems and then solve them in sequence. Solving each subproblem is facilitated by the answers to previously solved subproblems. Our experimental results on tasks related to symbolic manipulation, compositional generalization, and math reasoning reveal that least-to-most prompting is capable of generalizing to more difficult problems than those seen in the prompts. A notable finding is that when the GPT-3 code-davinci-002 model is used with least-to-most prompting, it can solve the compositional generalization benchmark SCAN in any split (including length split) with an accuracy of at least 99% using just 14 exemplars, compared to only 16% accuracy with chain-of-thought prompting. This is particularly noteworthy because neural-symbolic models in the literature that specialize in solving SCAN are trained on the entire training set containing over 15,000 examples. We have included prompts for all the tasks in the Appendix."
                },
                {
                    "title": "Scaling Up Models and Data with t5x and seqio",
                    "abstract": "Recent neural network-based language models have benefited greatly from scaling up the size of training datasets and the number of parameters in the models themselves. Scaling can be complicated due to various factors including the need to distribute computation on supercomputer clusters (e.g., TPUs), prevent bottlenecks when infeeding data, and ensure reproducible results. In this work, we present two software libraries that ease these issues: $\\texttt{t5x}$ simplifies the process of building and training large language models at scale while maintaining ease of use, and $\\texttt{seqio}$ provides a task-based API for simple creation of fast and reproducible training data and evaluation pipelines. These open-source libraries have been used to train models with hundreds of billions of parameters on datasets with multiple terabytes of training data. Along with the libraries, we release configurations and instructions for T5-like encoder-decoder models as well as GPT-like decoder-only architectures. $\\texttt{t5x}$ and $\\texttt{seqio}$ are open source and available at https://github.com/google-research/t5x and https://github.com/google/seqio, respectively."
                },
                {
                    "title": "Self-Consistency Improves Chain of Thought Reasoning in Language Models",
                    "abstract": "Chain-of-thought prompting combined with pre-trained large language models has achieved encouraging results on complex reasoning tasks. In this paper, we propose a new decoding strategy, self-consistency, to replace the naive greedy decoding used in chain-of-thought prompting. It first samples a diverse set of reasoning paths instead of only taking the greedy one, and then selects the most consistent answer by marginalizing out the sampled reasoning paths. Self-consistency leverages the intuition that a complex reasoning problem typically admits multiple different ways of thinking leading to its unique correct answer. Our extensive empirical evaluation shows that self-consistency boosts the performance of chain-of-thought prompting with a striking margin on a range of popular arithmetic and commonsense reasoning benchmarks, including GSM8K (+17.9%), SVAMP (+11.0%), AQuA (+12.2%), StrategyQA (+6.4%) and ARC-challenge (+3.9%)."
                },
                {
                    "title": "Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time",
                    "abstract": "The conventional recipe for maximizing model accuracy is to (1) train multiple models with various hyperparameters and (2) pick the individual model which performs best on a held-out validation set, discarding the remainder. In this paper, we revisit the second step of this procedure in the context of fine-tuning large pre-trained models, where fine-tuned models often appear to lie in a single low error basin. We show that averaging the weights of multiple models fine-tuned with different hyperparameter configurations often improves accuracy and robustness. Unlike a conventional ensemble, we may average many models without incurring any additional inference or memory costs -- we call the results\"model soups.\"When fine-tuning large pre-trained models such as CLIP, ALIGN, and a ViT-G pre-trained on JFT, our soup recipe provides significant improvements over the best model in a hyperparameter sweep on ImageNet. The resulting ViT-G model, which attains 90.94% top-1 accuracy on ImageNet, achieved a new state of the art. Furthermore, we show that the model soup approach extends to multiple image classification and natural language processing tasks, improves out-of-distribution performance, and improves zero-shot performance on new downstream tasks. Finally, we analytically relate the performance similarity of weight-averaging and logit-ensembling to flatness of the loss and confidence of the predictions, and validate this relation empirically. Code is available at https://github.com/mlfoundations/model-soups."
                },
                {
                    "title": "Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution",
                    "abstract": "When transferring a pretrained model to a downstream task, two popular methods are full fine-tuning (updating all the model parameters) and linear probing (updating only the last linear layer\u2014the \u201chead\u201d). It is well known that fine-tuning leads to better accuracy in-distribution (ID). However, in this paper, we find that fine-tuning can achieve worse accuracy than linear probing out-of-distribution (OOD) when the pretrained features are good and the distribution shift is large. On 10 distribution shift datasets (Breeds-Living17, Breeds-Entity30, DomainNet, CIFAR \u2192 STL, CIFAR10.1, FMoW, ImageNetV2, ImageNet-R, ImageNet-A, ImageNet-Sketch), fine-tuning obtains on average 2% higher accuracy ID but 7% lower accuracy OOD than linear probing. We show theoretically that this tradeoff between ID and OOD accuracy arises even in a simple setting: fine-tuning overparameterized two-layer linear networks. We prove that the OOD error of fine-tuning is high when we initialize with a fixed or random head\u2014this is because while fine-tuning learns the head, the lower layers of the neural network change simultaneously and distort the pretrained features. Our analysis suggests that the easy two-step strategy of linear probing then full fine-tuning (LP-FT), sometimes used as a fine-tuning heuristic, combines the benefits of both fine-tuning and linear probing. Empirically, LP-FT outperforms both fine-tuning and linear probing on the above datasets (1% better ID, 10% better OOD than full fine-tuning)."
                },
                {
                    "title": "A data-driven approach for learning to control computers",
                    "abstract": "It would be useful for machines to use computers as humans do so that they can aid us in everyday tasks. This is a setting in which there is also the potential to leverage large-scale expert demonstrations and human judgements of interactive behaviour, which are two ingredients that have driven much recent success in AI. Here we investigate the setting of computer control using keyboard and mouse, with goals specified via natural language. Instead of focusing on hand-designed curricula and specialized action spaces, we focus on developing a scalable method centered on reinforcement learning combined with behavioural priors informed by actual human-computer interactions. We achieve state-of-the-art and human-level mean performance across all tasks within the MiniWob++ benchmark, a challenging suite of computer control problems, and find strong evidence of cross-task transfer. These results demonstrate the usefulness of a unified human-agent interface when training machines to use computers. Altogether our results suggest a formula for achieving competency beyond MiniWob++ and towards controlling computers, in general, as a human would."
                },
                {
                    "title": "WebGPT: Browser-assisted question-answering with human feedback",
                    "abstract": "We fine-tune GPT-3 to answer long-form questions using a text-based web-browsing environment, which allows the model to search and navigate the web. By setting up the task so that it can be performed by humans, we are able to train models on the task using imitation learning, and then optimize answer quality with human feedback. To make human evaluation of factual accuracy easier, models must collect references while browsing in support of their answers. We train and evaluate our models on ELI5, a dataset of questions asked by Reddit users. Our best model is obtained by fine-tuning GPT-3 using behavior cloning, and then performing rejection sampling against a reward model trained to predict human preferences. This model's answers are preferred by humans 56% of the time to those of our human demonstrators, and 69% of the time to the highest-voted answer from Reddit."
                },
                {
                    "title": "Finetuned Language Models Are Zero-Shot Learners",
                    "abstract": "This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning."
                },
                {
                    "title": "UIBert: Learning Generic Multimodal Representations for UI Understanding",
                    "abstract": "To improve the accessibility of smart devices and to simplify their usage, building models which understand user interfaces (UIs) and assist users to complete their tasks is critical. However, unique challenges are proposed by UI-specific characteristics, such as how to effectively leverage multimodal UI features that involve image, text, and structural metadata and how to achieve good performance when high-quality labeled data is unavailable. To address such challenges we introduce UIBert, a transformer-based joint image-text model trained through novel pre-training tasks on large-scale unlabeled UI data to learn generic feature representations for a UI and its components. Our key intuition is that the heterogeneous features in a UI are self-aligned, i.e., the image and text features of UI components, are predictive of each other. We propose five pretraining tasks utilizing this self-alignment among different features of a UI component and across various components in the same UI. We evaluate our method on nine real-world downstream UI tasks where UIBert outperforms strong multimodal baselines by up to 9.26% accuracy."
                },
                {
                    "title": "The Evolution of Out-of-Distribution Robustness Throughout Fine-Tuning",
                    "abstract": "Although machine learning models typically experience a drop in performance on out-of-distribution data, accuracies on in- versus out-of-distribution data are widely observed to follow a single linear trend when evaluated across a testbed of models. Models that are more accurate on the out-of-distribution data relative to this baseline exhibit\"effective robustness\"and are exceedingly rare. Identifying such models, and understanding their properties, is key to improving out-of-distribution performance. We conduct a thorough empirical investigation of effective robustness during fine-tuning and surprisingly find that models pre-trained on larger datasets exhibit effective robustness during training that vanishes at convergence. We study how properties of the data influence effective robustness, and we show that it increases with the larger size, more diversity, and higher example difficulty of the dataset. We also find that models that display effective robustness are able to correctly classify 10% of the examples that no other current testbed model gets correct. Finally, we discuss several strategies for scaling effective robustness to the high-accuracy regime to improve the out-of-distribution accuracy of state-of-the-art models."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Mobile App Tasks with Iterative Feedback (MoTIF): Addressing Task Feasibility in Interactive Visual Environments",
                    "abstract": "In recent years, vision-language research has shifted to study tasks which require more complex reasoning, such as interactive question answering, visual common sense reasoning, and question-answer plausibility prediction. However, the datasets used for these problems fail to capture the complexity of real inputs and multimodal environments, such as ambiguous natural language requests and diverse digital domains. We introduce Mobile app Tasks with Iterative Feedback (MoTIF), a dataset with natural language commands for the greatest number of interactive environments to date. MoTIF is the first to contain natural language requests for interactive environments that are not satisfiable, and we obtain follow-up questions on this subset to enable research on task uncertainty resolution. We perform initial feasibility classification experiments and only reach an F1 score of 37.3, verifying the need for richer vision-language representations and improved architectures to reason about task feasibility."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Mapping Natural Language Instructions to Mobile UI Action Sequences",
                    "abstract": "We present a new problem: grounding natural language instructions to mobile user interface actions, and create three new datasets for it. For full task evaluation, we create PixelHelp, a corpus that pairs English instructions with actions performed by people on a mobile UI emulator. To scale training, we decouple the language and action data by (a) annotating action phrase spans in How-To instructions and (b) synthesizing grounded descriptions of actions for mobile user interfaces. We use a Transformer to extract action phrase tuples from long-range natural language instructions. A grounding Transformer then contextually represents UI objects using both their content and screen position and connects them to object descriptions. Given a starting screen and instruction, our model achieves 70.59% accuracy on predicting complete ground-truth action sequences in PixelHelp."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Learning to Navigate the Web",
                    "abstract": "Learning in environments with large state and action spaces, and sparse rewards, can hinder a Reinforcement Learning (RL) agent's learning through trial-and-error. For instance, following natural language instructions on the Web (such as booking a flight ticket) leads to RL settings where input vocabulary and number of actionable elements on a page can grow very large. Even though recent approaches improve the success rate on relatively simple environments with the help of human demonstrations to guide the exploration, they still fail in environments where the set of possible instructions can reach millions. We approach the aforementioned problems from a different perspective and propose guided RL approaches that can generate unbounded amount of experience for an agent to learn from. Instead of learning from a complicated instruction with a large vocabulary, we decompose it into multiple sub-instructions and schedule a curriculum in which an agent is tasked with a gradually increasing subset of these relatively easier sub-instructions. In addition, when the expert demonstrations are not available, we propose a novel meta-learning framework that generates new instruction following tasks and trains the agent more effectively. We train DQN, deep reinforcement learning agent, with Q-value function approximated with a novel QWeb neural network architecture on these smaller, synthetic instructions. We evaluate the ability of our agent to generalize to new instructions on World of Bits benchmark, on forms with up to 100 elements, supporting 14 million possible instructions. The QWeb agent outperforms the baseline without using any human demonstration achieving 100% success rate on several difficult environments."
                },
                {
                    "title": "Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration",
                    "abstract": "Reinforcement learning (RL) agents improve through trial-and-error, but when reward is sparse and the agent cannot discover successful action sequences, learning stagnates. This has been a notable problem in training deep RL agents to perform web-based tasks, such as booking flights or replying to emails, where a single mistake can ruin the entire sequence of actions. A common remedy is to \"warm-start\" the agent by pre-training it to mimic expert demonstrations, but this is prone to overfitting. Instead, we propose to constrain exploration using demonstrations. From each demonstration, we induce high-level \"workflows\" which constrain the allowable actions at each time step to be similar to those in the demonstration (e.g., \"Step 1: click on a textbox; Step 2: enter some text\"). Our exploration policy then learns to identify successful workflows and samples actions that satisfy these workflows. Workflows prune out bad exploration directions and accelerate the agent's ability to discover rewards. We use our approach to train a novel neural policy designed to handle the semi-structured nature of websites, and evaluate on a suite of web tasks, including the recent World of Bits benchmark. We achieve new state-of-the-art results, and show that workflow-guided exploration improves sample efficiency over behavioral cloning by more than 100x."
                },
                {
                    "title": "World of Bits: An Open-Domain Platform for Web-Based Agents",
                    "abstract": "While simulated game environments have greatly accelerated research in reinforcement learning, existing environments lack the open-domain re-alism of tasks in computer vision or natural language processing, which operate on artifacts created by humans in natural, organic settings. To foster reinforcement learning research in such settings, we introduce the World of Bits (WoB), a platform in which agents complete tasks on the Internet by performing low-level keyboard and mouse actions. The two main challenges are: (i) to curate a diverse set of natural web-based tasks, and (ii) to ensure that these tasks have a well-de\ufb01ned reward structure and are re-producible despite the transience of the web. To tackle this, we develop a methodology in which crowdworkers create tasks de\ufb01ned by natural language questions and provide demonstrations of how to answer the question on real websites us-ing keyboard and mouse; HTTP traf\ufb01c is cached to create a reproducible of\ufb02ine approximation of the website. Finally, we show that agents trained via behavioral cloning and reinforcement learning can complete a range of web-based tasks."
                }
            ],
            "categories": [
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does the performance of fine-tuned computer control agents scale with the amount of training data, and what is the impact of task complexity on their effectiveness in real-world applications?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the scalability of fine-tuning methods for computer control agents, which are increasingly relevant in automating human-computer interactions. Understanding the relationship between data volume, task complexity, and agent performance can lead to more efficient training protocols, reducing the time and cost associated with data collection. This research could advance knowledge in AI by providing insights into the limits of current models and guiding future developments in task execution capabilities, ultimately leading to practical applications in various domains such as customer service automation, personal assistants, and accessibility tools.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the need to rigorously quantify performance scaling across diverse tasks and applications, which requires a comprehensive dataset and robust evaluation metrics. Naive approaches may fail because they do not account for the nuances of task complexity and the varying levels of abstraction in high-level goal decomposition versus low-level action execution. Additionally, the theoretical understanding of how LLMs can effectively ground actions in real-world environments is still developing, posing significant obstacles in both the design of experiments and the interpretation of results.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either the performance of LLMs in isolation or on specific task execution without a systematic examination of data scaling and task complexity. Limitations in existing datasets and a lack of comprehensive frameworks for evaluating performance across different domains have hindered progress. Moreover, the complexity of human-computer interaction tasks has not been adequately addressed in prior work. This research aims to fill these gaps by introducing the AndroidControl dataset and a structured approach to analyze the scaling of fine-tuning in a more nuanced manner than previous studies.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves the use of the AndroidControl dataset, which contains 15,283 human-generated task demonstrations in Android applications. The approach will analyze how performance metrics scale with varying amounts of fine-tuning data, focusing on both high-level and low-level task execution. Key metrics for evaluation will include task success rates and the complexity of tasks performed. The expected outcomes include a clearer"
            }
        },
        "author_data": {
            "333a57de-3bcc-4ded-8aa9-edc02d2bd729": {
                "pk": "333a57de-3bcc-4ded-8aa9-edc02d2bd729",
                "name": "Wei Li",
                "collaborators": [],
                "domain": [
                    "High Energy Physics",
                    "Supergravity",
                    "Blockchain",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Observation of a \"Ridge\" correlation structure in high multiplicity proton-proton collisions: A brief review",
                        "abstract": "This paper briefly reviews the striking experimental observation of a ridge-like dihadron correlation structure in high multiplicity proton-proton collisions at the Large Hadron Collider (LHC). Recent progress of both experimental and theoretical efforts on understanding the physical origin of the novel effect is reviewed. Outlook on future direction of possible new studies is discussed."
                    },
                    {
                        "title": "Collective flow from AA, pA to pp collisions - Toward a unified paradigm",
                        "abstract": "I give an overview of the latest development in understanding collective phenomena in high-multiplicity hadronic final state from relativistic nucleus-nucleus, proton-nucleus and proton-proton collisions. Upon reviewing the experimental data and confronting them with theoretical models, a unified paradigm in describing the observed collectivity across all hadronic collision systems is emerging. Potential future paths toward addressing key open questions, especially on collectivity in small systems (pp, pA), are discussed."
                    },
                    {
                        "title": "Partial Differential Chow Forms and a Type of Partial Differential Chow varieties",
                        "abstract": "We first present an intersection theory of partial differential varieties with quasi-generic differential hypersurfaces. Then based on the generic intersection theory, we define the partial differential Chow form for an irreducible partial differential variety $V$ of Kolchin polynomial $\\omega_V(t)=(d+1){t+m\\choose m}-{t+m-s\\choose m}$. And we establish for the partial differential Chow form most of the basic properties of the ordinary differential Chow form. Furthermore, we prove the existence of a type of partial differential Chow varieties."
                    },
                    {
                        "title": "Chiral algebra from worldsheet",
                        "abstract": "The chiral algebra of a 4D $N\\geq2$ superconformal field theory is a vertex operator algebra generated by the Schur subsector of the 4D theory and its rigid (yet rich) structure has been useful in constraining and classifying 4D N=2 SCFTs. We study how the chiral algebra arises from the worldsheet perspective. In the worldsheet CFT dual of 4D N=4 SYM at the free point, we extract the subsector that corresponds to the spacetime Schur operators at generic coupling, and show how they generate the chiral algebra. The result can be easily generalized to 4D N=2 superconformal field theories that arise as orbifolds of 4D N=4 SYM."
                    },
                    {
                        "title": "Non-Supersymmetric Attractors in Symmetric Coset Spaces",
                        "abstract": "We present a method of constructing generic single-centered and multi-centered extremal black hole solutions in a large class of 4D N=2 supergravities coupled to vector-multiplets with cubic prepotentials. The method is applicable to models for which the 3D moduli spaces obtained via c*-map are symmetric coset spaces. The attractor solutions are generated by certain nilpotent elements in the coset algebra. We present explicit computations in 4D N=2 supergravity coupled to one vector-multiplet, whose 3D moduli space is the symmetric coset space G_{2(2)}/SL(2,R)^2. The non-supersymmetric multi-centered black holes in this model are found to lack the intricate moduli space of bound configurations that are typical of the supersymmetric case."
                    },
                    {
                        "title": "Adapting Blockchain Technology for Scientific Computing",
                        "abstract": "Blockchain stores information into a chain of \"blocks\", whose integrity is usually guaranteed by Proof of Work (PoW). In many blockchain applications (including cryptocurrencies), users compete with each other to win the ownership of the blocks, a process commonly referred as \"mining\". Mining activities consume huge amount of power, while the outcome appears to be useless besides validating a block. Here we discuss the requirements of designing a new PoW algorithm. We also propose a PoW scheme to help solve high-dimension, non-linear optimization problems. Simulation experiments of blockchains generated by three miners solved an instance of Traveling Salesman Problem (TSP), a well-known NP-hard problem. The revised scheme enables us to address difficult scientific questions as a byproduct of mining."
                    },
                    {
                        "title": "Learning UI Navigation through Demonstrations composed of Macro Actions",
                        "abstract": "We have developed a framework to reliably build agents capable of UI navigation. The state space is simplified from raw-pixels to a set of UI elements extracted from screen understanding, such as OCR and icon detection. The action space is restricted to the UI elements plus a few global actions. Actions can be customized for tasks and each action is a sequence of basic operations conditioned on status checks. With such a design, we are able to train DQfD and BC agents with a small number of demonstration episodes. We propose demo augmentation that significantly reduces the required number of human demonstrations. We made a customization of DQfD to allow demos collected on screenshots to facilitate the demo coverage of rare cases. Demos are only collected for the failed cases during the evaluation of the previous version of the agent. With 10s of iterations looping over evaluation, demo collection, and training, the agent reaches a 98.7\\% success rate on the search task in an environment of 80+ apps and websites where initial states and viewing parameters are randomized."
                    },
                    {
                        "title": "Gluing affine Yangians with bi-fundamentals",
                        "abstract": "The affine Yangian of $\\mathfrak{gl}_1$ is isomorphic to the universal enveloping algebra of $\\mathcal{W}_{1+\\infty}$ and can serve as a building block in the construction of new vertex operator algebras. In [1], a two-parameter family generalization of $\\mathcal{N}=2$ supersymmetric $\\mathcal{W}_{\\infty}$ algebra was constructed by \"gluing\" two affine Yangians of $\\mathfrak{gl}_1$ using operators that transform as $(\\square, \\overline{\\square})$ and $(\\overline{\\square}, \\square)$ w.r.t. the two affine Yangians. In this paper we realize a similar (but non-isomorphic) two-parameter gluing construction where the gluing operators transform as $(\\square, \\square)$ and $(\\overline{\\square}, \\overline{\\square})$ w.r.t. the two affine Yangians. The corresponding representation space consists of pairs of plane partitions connected by a common leg whose cross-section takes the shape of Young diagrams, offering a more transparent geometric picture than the previous construction."
                    },
                    {
                        "title": "Quiver algebras and their representations for arbitrary quivers",
                        "abstract": "The quiver Yangians were originally defined for the quiver and superpotential from string theory on general toric Calabi-Yau threefolds, and serve as BPS algebras of these systems. Their characters reproduce the unrefined BPS indices, which correspond to classical Donaldson-Thomas (DT) invariants. We generalize this construction in two directions. First, we show that this definition extends to arbitrary quivers with potentials. Second, we explain how to define the characters to incorporate the refined BPS indices, which correspond to motivic DT invariants. We focus on two main classes of quivers: the BPS quivers of 4D $N=2$ theories and the quivers from the knot-quiver correspondence. The entire construction allows for straightforward generalizations to trigonometric, elliptic, and generalized cohomologies."
                    },
                    {
                        "title": "The potential fluctuation and its interfacial phenomena in molecular, micro, macro, and cosmic flow instabilities",
                        "abstract": "Flow instabilities play important roles in a wide range of engineering, geophysical, and astrophysical flows, ranging from supernova explosion in crab nebula, formation of clouds in sky, waves on ocean, to inertial confinement fusion capsules, making fusion energy a viable alternative energy source in the future. The potential for life is directly related to flow instability mixing as well. Previous researchers have focused on developed stages of flow instabilities by assuming sine wave interface between fluids in the flow instabilities. No scientific research has been reported to investigate the origin of flow instabilities. The paper advances a new physics concept, potential fluctuation in flow based on the conservation of mass, which presents potential oscillatory sine wave surface in the spatial and temporal dimensions at the interface of flow instabilities. Potential fluctuation is decided by the two densities and velocities in the flow as indicated by the relation of continuity. Even before the flow instabilities start to develop, potential fluctuation has already internally existed in flow. It is only decided by the densities and velocities of the two moving fluids and is not related to the surface topography of the boundary of flows in the flow instability. It is the gene of flow instability. The paper presents breakthrough of understanding of micro, macro, and cosmic flow instabilities."
                    },
                    {
                        "title": "A Development Calculus for Specifications",
                        "abstract": "A first order inference system, called R-calculus, is defined to develop the specifications. It is used to eliminate the laws which is not consistent with the user's requirements. The R-calculus consists of the structural rules, an axiom, a cut rule, and the rules for logical connectives. Some examples are given to demonstrate the usage of the R-calculus. The properties about reachability and completeness of the R-calculus are formally defined and are proved."
                    }
                ]
            },
            "23a50f7d-bf53-468a-b92e-611019ec57c3": {
                "pk": "23a50f7d-bf53-468a-b92e-611019ec57c3",
                "name": "William Bishop",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "acd5fc60-ba66-480f-a0c3-f05bc011d749": {
                "pk": "acd5fc60-ba66-480f-a0c3-f05bc011d749",
                "name": "Alice Li",
                "collaborators": [
                    "Luanne Sinnamon",
                    "X. Alice Li",
                    "Devi Parikh",
                    "Victor Churchill",
                    "H. Alice Li",
                    "Dongbin Xiu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Generative AI",
                    "Machine Learning",
                    "Consumer Behavior"
                ],
                "publications": [
                    {
                        "title": "Generative AI Search Engines as Arbiters of Public Knowledge: An Audit of Bias and Authority",
                        "abstract": "This paper reports on an audit study of generative AI systems (ChatGPT, Bing Chat, and Perplexity) which investigates how these new search engines construct responses and establish authority for topics of public importance. We collected system responses using a set of 48 authentic queries for 4 topics over a 7-day period and analyzed the data using sentiment analysis, inductive coding and source classification. Results provide an overview of the nature of system responses across these systems and provide evidence of sentiment bias based on the queries and topics, and commercial and geographic bias in sources. The quality of sources used to support claims is uneven, relying heavily on News and Media, Business and Digital Media websites. Implications for system users emphasize the need to critically examine Generative AI system outputs when making decisions related to public interest and personal well-being."
                    },
                    {
                        "title": "Lemotif: An Affective Visual Journal Using Deep Neural Networks",
                        "abstract": "We present Lemotif, an integrated natural language processing and image generation system that uses machine learning to (1) parse a text-based input journal entry describing the user's day for salient themes and emotions and (2) visualize the detected themes and emotions in creative and appealing image motifs. Synthesizing approaches from artificial intelligence and psychology, Lemotif acts as an affective visual journal, encouraging users to regularly write and reflect on their daily experiences through visual reinforcement. By making patterns in emotions and their sources more apparent, Lemotif aims to help users better understand their emotional lives, identify opportunities for action, and track the effectiveness of behavioral changes over time. We verify via human studies that prospective users prefer motifs generated by Lemotif over corresponding baselines, find the motifs representative of their journal entries, and think they would be more likely to journal regularly using a Lemotif-based app."
                    },
                    {
                        "title": "Unraveling Consumer Purchase Journey Using Neural Network Models",
                        "abstract": "This study utilizes an ensemble of feedforward neural network models to analyze large-volume and high-dimensional consumer touchpoints and their impact on purchase decisions. When applied to a proprietary dataset of consumer touchpoints and purchases from a global software service provider, the proposed approach demonstrates better predictive accuracy than both traditional models, such as logistic regression, naive Bayes, and k-nearest neighbors, as well as ensemble tree-based classifiers, such as bagging, random forest, AdaBoost, and gradient boosting. By calculating the Shapley values within this network, we provide nuanced insights into touchpoint effectiveness, as we not only assess the marginal impact of diverse touchpoint types but also offer a granular view of the impact distribution within a touchpoint type. Additionally, our model shows excellent adaptability and resilience with limited data resources. When the historical data is reduced from 40 to 1 month, our model shows only a modest 19% decrease in accuracy. This modeling framework can enable managers to more accurately and comprehensively evaluate consumer touchpoints, thereby enhancing the effectiveness and efficiency of their marketing campaigns."
                    }
                ]
            },
            "498d6258-6fc0-4b32-9511-83de6f277d51": {
                "pk": "498d6258-6fc0-4b32-9511-83de6f277d51",
                "name": "Chris Rawles",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "35f92c03-bbe5-48da-99ff-26d1ea1953a0": {
                "pk": "35f92c03-bbe5-48da-99ff-26d1ea1953a0",
                "name": "Folawiyo Campbell-Ajala",
                "collaborators": [
                    "Wei Li",
                    "Oriana Riva",
                    "Fu-Lin Hsu",
                    "Will Bishop",
                    "Max Lin",
                    "Christopher Rawles",
                    "Sarah Clinckemaillie",
                    "Yifan Chang",
                    "Jonathan Waltz",
                    "Gabrielle Lau"
                ],
                "domain": [
                    "Autonomous Systems",
                    "Human-Computer Interaction",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "UINav: A Practical Approach to Train On-Device Automation Agents",
                        "abstract": "Automation systems that can autonomously drive application user interfaces to complete user tasks are of great benefit, especially when users are situationally or permanently impaired. Prior automation systems do not produce generalizable models while AI-based automation agents work reliably only in simple, hand-crafted applications or incur high computation costs. We propose UINav, a demonstration-based approach to train automation agents that fit mobile devices, yet achieving high success rates with modest numbers of demonstrations. To reduce the demonstration overhead, UINav uses a referee model that provides users with immediate feedback on tasks where the agent fails, and automatically augments human demonstrations to increase diversity in training data. Our evaluation shows that with only 10 demonstrations UINav can achieve 70% accuracy, and that with enough demonstrations it can surpass 90% accuracy."
                    },
                    {
                        "title": "AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents",
                        "abstract": "Autonomous agents that execute human tasks by controlling computers can enhance human productivity and application accessibility. However, progress in this field will be driven by realistic and reproducible benchmarks. We present AndroidWorld, a fully functional Android environment that provides reward signals for 116 programmatic tasks across 20 real-world Android apps. Unlike existing interactive environments, which provide a static test set, AndroidWorld dynamically constructs tasks that are parameterized and expressed in natural language in unlimited ways, thus enabling testing on a much larger and more realistic suite of tasks. Reward signals are derived from the computer's system state, making them durable across task variations and extensible across different apps. To demonstrate AndroidWorld's benefits and mode of operation, we introduce a new computer control agent, M3A. M3A can complete 30.6% of the AndroidWorld's tasks, leaving ample room for future work. Furthermore, we adapt a popular desktop web agent to work on Android, which we find to be less effective on mobile, suggesting future research is needed to achieve universal, cross-domain agents. Finally, we conduct a robustness analysis by testing M3A against a range of task variations on a representative subset of tasks, demonstrating that variations in task parameters can significantly alter a task's complexity and, consequently, an agent's performance, highlighting the importance of testing agents under diverse conditions. AndroidWorld and the experiments in this paper are available at https://github.com/google-research/android_world."
                    }
                ]
            },
            "f06bb971-0ce0-46b1-99ef-01a4257e3354": {
                "pk": "f06bb971-0ce0-46b1-99ef-01a4257e3354",
                "name": "Divya Tyamagundlu",
                "collaborators": [
                    "Christopher Rawles",
                    "Sarah Clinckemaillie",
                    "Yifan Chang",
                    "Jonathan Waltz",
                    "Gabrielle Lau",
                    "Marybeth Fair",
                    "Alice Li",
                    "William Bishop",
                    "Wei Li",
                    "Folawiyo Campbell-Ajala"
                ],
                "domain": [
                    "Autonomous Agents",
                    "Benchmarking",
                    "Mobile Computing",
                    "Human-Computer Interaction"
                ],
                "publications": [
                    {
                        "title": "AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents",
                        "abstract": "Autonomous agents that execute human tasks by controlling computers can enhance human productivity and application accessibility. However, progress in this field will be driven by realistic and reproducible benchmarks. We present AndroidWorld, a fully functional Android environment that provides reward signals for 116 programmatic tasks across 20 real-world Android apps. Unlike existing interactive environments, which provide a static test set, AndroidWorld dynamically constructs tasks that are parameterized and expressed in natural language in unlimited ways, thus enabling testing on a much larger and more realistic suite of tasks. Reward signals are derived from the computer's system state, making them durable across task variations and extensible across different apps. To demonstrate AndroidWorld's benefits and mode of operation, we introduce a new computer control agent, M3A. M3A can complete 30.6% of the AndroidWorld's tasks, leaving ample room for future work. Furthermore, we adapt a popular desktop web agent to work on Android, which we find to be less effective on mobile, suggesting future research is needed to achieve universal, cross-domain agents. Finally, we conduct a robustness analysis by testing M3A against a range of task variations on a representative subset of tasks, demonstrating that variations in task parameters can significantly alter a task's complexity and, consequently, an agent's performance, highlighting the importance of testing agents under diverse conditions. AndroidWorld and the experiments in this paper are available at https://github.com/google-research/android_world."
                    }
                ]
            },
            "4a75e56b-dbbd-4e87-a370-97c22d452ce2": {
                "pk": "4a75e56b-dbbd-4e87-a370-97c22d452ce2",
                "name": "Oriana Riva",
                "collaborators": [
                    "Alice Li",
                    "Christopher Rawles",
                    "Timothy Lillicrap",
                    "Wei Li",
                    "Folawiyo Campbell-Ajala",
                    "Sahisnu Mazumder",
                    "Qingqing Cao",
                    "Aruna Balasubramanian",
                    "Niranjan Balasubramanian",
                    "Pratyay Banerjee"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Human-Computer Interaction",
                    "Machine Learning",
                    "Autonomous Agents"
                ],
                "publications": [
                    {
                        "title": "FLIN: A Flexible Natural Language Interface for Web Navigation",
                        "abstract": "AI assistants can now carry out tasks for users by directly interacting with website UIs. Current semantic parsing and slot-filling techniques cannot flexibly adapt to many different websites without being constantly re-trained. We propose FLIN, a natural language interface for web navigation that maps user commands to concept-level actions (rather than low-level UI actions), thus being able to flexibly adapt to different websites and handle their transient nature. We frame this as a ranking problem: given a user command and a webpage, FLIN learns to score the most relevant navigation instruction (involving action and parameter values). To train and evaluate FLIN, we collect a dataset using nine popular websites from three domains. Our results show that FLIN was able to adapt to new websites in a given domain."
                    },
                    {
                        "title": "Bew: Towards Answering Business-Entity-Related Web Questions",
                        "abstract": "We present BewQA, a system specifically designed to answer a class of questions that we call Bew questions. Bew questions are related to businesses/services such as restaurants, hotels, and movie theaters; for example, \"Until what time is happy hour?\". These questions are challenging to answer because the answers are found in open-domain Web, are present in short sentences without surrounding context, and are dynamic since the webpage information can be updated frequently. Under these conditions, existing QA systems perform poorly. We present a practical approach, called BewQA, that can answer Bew queries by mining a template of the business-related webpages and using the template to guide the search. We show how we can extract the template automatically by leveraging aggregator websites that aggregate information about business entities in a domain (e.g., restaurants). We answer a given question by identifying the section from the extracted template that is most likely to contain the answer. By doing so we can extract the answers even when the answer span does not have sufficient context. Importantly, BewQA does not require any training. We crowdsource a new dataset of 1066 Bew questions and ground-truth answers in the restaurant domain. Compared to state-of-the-art QA models, BewQA has a 27 percent point improvement in F1 score. Compared to a commercial search engine, BewQA answered correctly 29% more Bew questions."
                    },
                    {
                        "title": "Lexi: Self-Supervised Learning of the UI Language",
                        "abstract": "Humans can learn to operate the user interface (UI) of an application by reading an instruction manual or how-to guide. Along with text, these resources include visual content such as UI screenshots and images of application icons referenced in the text. We explore how to leverage this data to learn generic visio-linguistic representations of UI screens and their components. These representations are useful in many real applications, such as accessibility, voice navigation, and task automation. Prior UI representation models rely on UI metadata (UI trees and accessibility labels), which is often missing, incompletely defined, or not accessible. We avoid such a dependency, and propose Lexi, a pre-trained vision and language model designed to handle the unique features of UI screens, including their text richness and context sensitivity. To train Lexi we curate the UICaption dataset consisting of 114k UI images paired with descriptions of their functionality. We evaluate Lexi on four tasks: UI action entailment, instruction-based UI image retrieval, grounding referring expressions, and UI entity recognition."
                    },
                    {
                        "title": "Latent State Estimation Helps UI Agents to Reason",
                        "abstract": "A common problem for agents operating in real-world environments is that the response of an environment to their actions may be non-deterministic and observed through noise. This renders environmental state and progress towards completing a task latent. Despite recent impressive demonstrations of LLM's reasoning abilities on various benchmarks, whether LLMs can build estimates of latent state and leverage them for reasoning has not been explicitly studied. We investigate this problem in the real-world domain of autonomous UI agents. We establish that appropriately prompting LLMs in a zero-shot manner can be formally understood as forming point estimates of latent state in a textual space. In the context of autonomous UI agents we then show that LLMs used in this manner are more than $76\\%$ accurate at inferring various aspects of latent state, such as performed (vs. commanded) actions and task progression. Using both public and internal benchmarks and three reasoning methods (zero-shot, CoT-SC & ReAct), we show that LLM-powered agents that explicitly estimate and reason about latent state are able to successfully complete up to 1.6x more tasks than those that do not."
                    },
                    {
                        "title": "Android in the Wild: A Large-Scale Dataset for Android Device Control",
                        "abstract": "There is a growing interest in device-control systems that can interpret human natural language instructions and execute them on a digital device by directly controlling its user interface. We present a dataset for device-control research, Android in the Wild (AITW), which is orders of magnitude larger than current datasets. The dataset contains human demonstrations of device interactions, including the screens and actions, and corresponding natural language instructions. It consists of 715k episodes spanning 30k unique instructions, four versions of Android (v10-13),and eight device types (Pixel 2 XL to Pixel 6) with varying screen resolutions. It contains multi-step tasks that require semantic understanding of language and visual context. This dataset poses a new challenge: actions available through the user interface must be inferred from their visual appearance. And, instead of simple UI element-based actions, the action space consists of precise gestures (e.g., horizontal scrolls to operate carousel widgets). We organize our dataset to encourage robustness analysis of device-control systems, i.e., how well a system performs in the presence of new task descriptions, new applications, or new platform versions. We develop two agents and report performance across the dataset. The dataset is available at https://github.com/google-research/google-research/tree/master/android_in_the_wild."
                    },
                    {
                        "title": "UINav: A Practical Approach to Train On-Device Automation Agents",
                        "abstract": "Automation systems that can autonomously drive application user interfaces to complete user tasks are of great benefit, especially when users are situationally or permanently impaired. Prior automation systems do not produce generalizable models while AI-based automation agents work reliably only in simple, hand-crafted applications or incur high computation costs. We propose UINav, a demonstration-based approach to train automation agents that fit mobile devices, yet achieving high success rates with modest numbers of demonstrations. To reduce the demonstration overhead, UINav uses a referee model that provides users with immediate feedback on tasks where the agent fails, and automatically augments human demonstrations to increase diversity in training data. Our evaluation shows that with only 10 demonstrations UINav can achieve 70% accuracy, and that with enough demonstrations it can surpass 90% accuracy."
                    },
                    {
                        "title": "VQA Training Sets are Self-play Environments for Generating Few-shot Pools",
                        "abstract": "Large-language models and large-vision models are increasingly capable of solving compositional reasoning tasks, as measured by breakthroughs in visual-question answering benchmarks. However, state-of-the-art solutions often involve careful construction of large pre-training and fine-tuning datasets, which can be expensive. The use of external tools, whether other ML models, search engines, or APIs, can significantly improve performance by breaking down high-level reasoning questions into sub-questions that are answerable by individual tools, but this approach has similar dataset construction costs to teach fine-tuned models how to use the available tools. We propose a technique in which existing training sets can be directly used for constructing computational environments with task metrics as rewards. This enables a model to autonomously teach itself to use itself or another model as a tool. By doing so, we augment training sets by integrating external signals. The proposed method starts with zero-shot prompts and iteratively refines them by selecting few-shot examples that maximize the task metric on the training set. Our experiments showcase how Gemini learns how to use itself, or another smaller and specialized model such as ScreenAI, to iteratively improve performance on training sets. Our approach successfully generalizes and improves upon zeroshot performance on charts, infographics, and document visual question-answering datasets"
                    },
                    {
                        "title": "Inducing a hierarchy for multi-class classification problems",
                        "abstract": "In applications where categorical labels follow a natural hierarchy, classification methods that exploit the label structure often outperform those that do not. Un-fortunately, the majority of classification datasets do not come pre-equipped with a hierarchical structure and classical flat classifiers must be employed. In this paper, we investigate a class of methods that induce a hierarchy that can similarly improve classification performance over flat classifiers. The class of methods follows the structure of first clustering the conditional distributions and subsequently using a hierarchical classifier with the induced hierarchy. We demonstrate the effectiveness of the class of methods both for discovering a latent hierarchy and for improving accuracy in principled simulation settings and three real data applications."
                    },
                    {
                        "title": "AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents",
                        "abstract": "Autonomous agents that execute human tasks by controlling computers can enhance human productivity and application accessibility. However, progress in this field will be driven by realistic and reproducible benchmarks. We present AndroidWorld, a fully functional Android environment that provides reward signals for 116 programmatic tasks across 20 real-world Android apps. Unlike existing interactive environments, which provide a static test set, AndroidWorld dynamically constructs tasks that are parameterized and expressed in natural language in unlimited ways, thus enabling testing on a much larger and more realistic suite of tasks. Reward signals are derived from the computer's system state, making them durable across task variations and extensible across different apps. To demonstrate AndroidWorld's benefits and mode of operation, we introduce a new computer control agent, M3A. M3A can complete 30.6% of the AndroidWorld's tasks, leaving ample room for future work. Furthermore, we adapt a popular desktop web agent to work on Android, which we find to be less effective on mobile, suggesting future research is needed to achieve universal, cross-domain agents. Finally, we conduct a robustness analysis by testing M3A against a range of task variations on a representative subset of tasks, demonstrating that variations in task parameters can significantly alter a task's complexity and, consequently, an agent's performance, highlighting the importance of testing agents under diverse conditions. AndroidWorld and the experiments in this paper are available at https://github.com/google-research/android_world."
                    }
                ]
            }
        }
    },
    "2406.09368": {
        "paper_data": {
            "title": "CLIPAway: Harmonizing Focused Embeddings for Removing Objects via Diffusion Models",
            "url": "http://arxiv.org/abs/2406.09368v1",
            "arxiv_id": "2406.09368",
            "authors": [
                "Yigit Ekin",
                "Ahmet Burak Yildirim",
                "Erdem Eren Caglar",
                "Aykut Erdem",
                "Erkut Erdem",
                "Aysegul Dundar"
            ],
            "abstract": "Advanced image editing techniques, particularly inpainting, are essential for seamlessly removing unwanted elements while preserving visual integrity. Traditional GAN-based methods have achieved notable success, but recent advancements in diffusion models have produced superior results due to their training on large-scale datasets, enabling the generation of remarkably realistic inpainted images. Despite their strengths, diffusion models often struggle with object removal tasks without explicit guidance, leading to unintended hallucinations of the removed object. To address this issue, we introduce CLIPAway, a novel approach leveraging CLIP embeddings to focus on background regions while excluding foreground elements. CLIPAway enhances inpainting accuracy and quality by identifying embeddings that prioritize the background, thus achieving seamless object removal. Unlike other methods that rely on specialized training datasets or costly manual annotations, CLIPAway provides a flexible, plug-and-play solution compatible with various diffusion-based inpainting techniques.",
            "introduction": "   1 Introduction  In today\u2019s digital era, the demand for sophisticated image editing techniques has surged, with inpainting emerging as a fundamental method for seamlessly removing unwanted elements from images while maintaining visual coherence. Image inpainting has long been studied in both academia and industry. Traditionally, research has predominantly focused on Generative Adversarial Network (GAN)-based methods [20, 16, 11, 15, 13, 25, 31], which have shown notable success in inpainting tasks. However, recent advancements in diffusion models have attracted considerable interest due to their ability to produce high-quality results [23, 27, 1, 30]. A key factor behind the effectiveness of diffusion models is their extensive training on large-scale datasets. By leveraging comprehensive collections of diverse image data, diffusion models can learn complex patterns and correlations, and intricate details, allowing them to inpaint missing regions with exceptional realism.   Despite their strengths, diffusion-based text-guided image inpainting models\u00a0[27, 1] often encounter challenges in object removal tasks without explicit guidance. When tasked with object removal without explicit text cues to insert a replacement or with the text of \u201cbackground\u201d, these models may inadvertently hallucinate the removed object, substituting it instead of erasing it entirely. This issue contrasts with user expectations, as users typically anticipate the erased portion to be seamlessly filled with the background. For example, when removing a person on a surfboard (Figure 1, second row), a diffusion model might insert another person, given the context that surfboards often have people on them. Similarly, when removing a plane from an image of an airport field (Figure 1, fourth row), a model might insert another plane due to the presence of other planes in the background. Alternatively, a model might introduce shoes on the floor (Figure 1, first row) when the user\u2019s intention was to fill the space with background elements.   In this paper, to address this issue, we propose a novel approach called CLIPAway that leverages AlphaCLIP\u00a0[24] embeddings to distinguish between foreground (the object to be removed) and background regions. Our method aims to identify an embedding for the inpainted region that focuses on the background while excluding the foreground content, thereby enhancing the quality and accuracy of the inpainting process.   Recent advancements in diffusion-based inpainting for object removal have yielded notable works such as InstInpaint [30], MagicBrush [33], InstructPix2Pix [2], and concurrently ObjectDrop [26]. These methods introduce training datasets with varying approaches: Some generate targets through existing inpainting methods [30], resulting in imperfect targets, while others rely on synthetically generated pairs [2] that may contain annotation errors. Certain methods resort to costly manual annotations [33] or require extensive data collection setups where images of scenes before and after object removal are captured [26]. Furthermore, these techniques typically involve either training a model from scratch or fine-tuning existing models specifically for the removal task. In contrast, our model, CLIPAway, distinguishes itself by its lack of dependency on a specialized training set. It offers a plug-and-play solution compatible with various diffusion-based inpainting methods, ensuring seamless object removal without the need for costly or complex data preparation.   Our contributions can be summarized as follows:   \u2022  We introduce CLIPAway, a method that utilizes AlphaCLIP embeddings to effectively differentiate",
            "references": [
                {
                    "title": "ObjectDrop: Bootstrapping Counterfactuals for Photorealistic Object Removal and Insertion",
                    "abstract": "Diffusion models have revolutionized image editing but often generate images that violate physical laws, particularly the effects of objects on the scene, e.g., occlusions, shadows, and reflections. By analyzing the limitations of self-supervised approaches, we propose a practical solution centered on a \\q{counterfactual} dataset. Our method involves capturing a scene before and after removing a single object, while minimizing other changes. By fine-tuning a diffusion model on this dataset, we are able to not only remove objects but also their effects on the scene. However, we find that applying this approach for photorealistic object insertion requires an impractically large dataset. To tackle this challenge, we propose bootstrap supervision; leveraging our object removal model trained on a small counterfactual dataset, we synthetically expand this dataset considerably. Our approach significantly outperforms prior methods in photorealistic object removal and insertion, particularly at modeling the effects of objects on the scene."
                },
                {
                    "title": "Alpha-CLIP: A CLIP Model Focusing on Wherever you Want",
                    "abstract": "Contrastive Language-Image Pre-training (CLIP) plays an essential role in extracting valuable content information from images across diverse tasks. It aligns textual and visual modalities to comprehend the entire image, including all the details, even those irrelevant to specific tasks. How-ever, for a finer understanding and controlled editing of images, it becomes crucial to focus on specific regions of interest, which can be indicated as points, masks, or boxes by humans or perception models. To fulfill the require we introduce Alpha-CLIp, an enhanced version of CLIP with an auxiliary alpha channel to suggest attentive regions and fine-tuned with constructed millions of RGBA region-text pairs. Alpha-CLIP not only preserves the visual recognition ability of CLIP but also enables precise control over the emphasis of image contents. It demonstrates effectiveness in various tasks, including but not limited to open-world recognition, multimodal large language models, and conditional 2D / 3D generation. It has a strong potential to serve as a versatile tool for image-related tasks. Our project is with codes and models available is linked to https://aleafy.github.io/alpha-clip/."
                },
                {
                    "title": "Rethinking FID: Towards a Better Evaluation Metric for Image Generation",
                    "abstract": "As with many machine learning problems, the progress of image generation methods hinges on good evaluation metrics. One of the most popular is the Frechet Inception Distance (FID). FID estimates the distance between a distribution of Inception-v3 features of real images, and those of images generated by the algorithm. We highlight important drawbacks of FID: Inception's poor representation of the rich and varied content generated by modern text-to-image models, incorrect normality assumptions, and poor sample complexity. We call for a reevaluation of FID's use as the primary quality metric for generated images. We empirically demonstrate that FID contradicts human raters, it does not reflect gradual improvement of iterative text-to-image models, it does not capture distortion levels, and that it produces inconsistent results when varying the sample size. We also propose an alternative new metric, CMMD, based on richer CLIP embeddings and the maximum mean discrepancy distance with the Gaussian RBF kernel. It is an unbiased estimator that does not make any assumptions on the probability distribution of the embeddings and is sample efficient. Through extensive experiments and analysis, we demonstrate that FID-based evaluations of text-to-image models may be unreliable, and that CMMD of-fers a more robust and reliable assessment of image quality. A reference implementation of CMMD is available at: https://github.com/google-research/google-research/tree/master/cmmd."
                },
                {
                    "title": "Uni-paint: A Unified Framework for Multimodal Image Inpainting with Pretrained Diffusion Model",
                    "abstract": "Recently, text-to-image denoising diffusion probabilistic models (DDPMs) have demonstrated impressive image generation capabilities and have also been successfully applied to image inpainting. However, in practice, users often require more control over the inpainting process beyond textual guidance, especially when they want to composite objects with customized appearance, color, shape, and layout. Unfortunately, existing diffusion-based inpainting methods are limited to single-modal guidance and require task-specific training, hindering their cross-modal scalability. To address these limitations, we propose Uni-paint, a unified framework for multimodal inpainting that offers various modes of guidance, including unconditional, text-driven, stroke-driven, exemplar-driven inpainting, as well as a combination of these modes. Furthermore, our Uni-paint is based on pretrained Stable Diffusion and does not require task-specific training on specific datasets, enabling few-shot generalizability to customized images. We have conducted extensive qualitative and quantitative evaluations that show our approach achieves comparable results to existing single-modal methods while offering multimodal inpainting capabilities not available in other methods. Code is available at https://github.com/ysy31415/unipaint."
                },
                {
                    "title": "IP-Adapter: Text Compatible Image Prompt Adapter for Text-to-Image Diffusion Models",
                    "abstract": "Recent years have witnessed the strong power of large text-to-image diffusion models for the impressive generative capability to create high-fidelity images. However, it is very tricky to generate desired images using only text prompt as it often involves complex prompt engineering. An alternative to text prompt is image prompt, as the saying goes:\"an image is worth a thousand words\". Although existing methods of direct fine-tuning from pretrained models are effective, they require large computing resources and are not compatible with other base models, text prompt, and structural controls. In this paper, we present IP-Adapter, an effective and lightweight adapter to achieve image prompt capability for the pretrained text-to-image diffusion models. The key design of our IP-Adapter is decoupled cross-attention mechanism that separates cross-attention layers for text features and image features. Despite the simplicity of our method, an IP-Adapter with only 22M parameters can achieve comparable or even better performance to a fully fine-tuned image prompt model. As we freeze the pretrained diffusion model, the proposed IP-Adapter can be generalized not only to other custom models fine-tuned from the same base model, but also to controllable generation using existing controllable tools. With the benefit of the decoupled cross-attention strategy, the image prompt can also work well with the text prompt to achieve multimodal image generation. The project page is available at \\url{https://ip-adapter.github.io}."
                },
                {
                    "title": "Diverse Inpainting and Editing with GAN Inversion",
                    "abstract": "Recent inversion methods have shown that real images can be inverted into StyleGAN\u2019s latent space and numerous edits can be achieved on those images thanks to the semantically rich feature representations of well-trained GAN models. However, extensive research has also shown that image inversion is challenging due to the trade-off between high-fidelity reconstruction and editability. In this paper, we tackle an even more difficult task, inverting erased images into GAN\u2019s latent space for realistic inpaintings and editings. Furthermore, by augmenting inverted latent codes with different latent samples, we achieve diverse inpaintings. Specifically, we propose to learn an encoder and mixing network to combine encoded features from erased images with StyleGAN\u2019s mapped features from random samples. To encourage the mixing network to utilize both inputs, we train the networks with generated data via a novel set-up. We also utilize higher-rate features to prevent color inconsistencies between the inpainted and unerased parts. We run extensive experiments and compare our method with state-of-the-art inversion and inpainting methods. Qualitative metrics and visual comparisons show significant improvements."
                },
                {
                    "title": "MagicBrush: A Manually Annotated Dataset for Instruction-Guided Image Editing",
                    "abstract": "Text-guided image editing is widely needed in daily life, ranging from personal use to professional applications such as Photoshop. However, existing methods are either zero-shot or trained on an automatically synthesized dataset, which contains a high volume of noise. Thus, they still require lots of manual tuning to produce desirable outcomes in practice. To address this issue, we introduce MagicBrush (https://osu-nlp-group.github.io/MagicBrush/), the first large-scale, manually annotated dataset for instruction-guided real image editing that covers diverse scenarios: single-turn, multi-turn, mask-provided, and mask-free editing. MagicBrush comprises over 10K manually annotated triplets (source image, instruction, target image), which supports trainining large-scale text-guided image editing models. We fine-tune InstructPix2Pix on MagicBrush and show that the new model can produce much better images according to human evaluation. We further conduct extensive experiments to evaluate current image editing baselines from multiple dimensions including quantitative, qualitative, and human evaluations. The results reveal the challenging nature of our dataset and the gap between current baselines and real-world editing needs."
                },
                {
                    "title": "Uni-ControlNet: All-in-One Control to Text-to-Image Diffusion Models",
                    "abstract": "Text-to-Image diffusion models have made tremendous progress over the past two years, enabling the generation of highly realistic images based on open-domain text descriptions. However, despite their success, text descriptions often struggle to adequately convey detailed controls, even when composed of long and complex texts. Moreover, recent studies have also shown that these models face challenges in understanding such complex texts and generating the corresponding images. Therefore, there is a growing need to enable more control modes beyond text description. In this paper, we introduce Uni-ControlNet, a unified framework that allows for the simultaneous utilization of different local controls (e.g., edge maps, depth map, segmentation masks) and global controls (e.g., CLIP image embeddings) in a flexible and composable manner within one single model. Unlike existing methods, Uni-ControlNet only requires the fine-tuning of two additional adapters upon frozen pre-trained text-to-image diffusion models, eliminating the huge cost of training from scratch. Moreover, thanks to some dedicated adapter designs, Uni-ControlNet only necessitates a constant number (i.e., 2) of adapters, regardless of the number of local or global controls used. This not only reduces the fine-tuning costs and model size, making it more suitable for real-world deployment, but also facilitate composability of different conditions. Through both quantitative and qualitative comparisons, Uni-ControlNet demonstrates its superiority over existing methods in terms of controllability, generation quality and composability. Code is available at \\url{https://github.com/ShihaoZhaoZSH/Uni-ControlNet}."
                },
                {
                    "title": "Inst-Inpaint: Instructing to Remove Objects with Diffusion Models",
                    "abstract": "Image inpainting task refers to erasing unwanted pixels from images and filling them in a semantically consistent and realistic way. Traditionally, the pixels that are wished to be erased are defined with binary masks. From the application point of view, a user needs to generate the masks for the objects they would like to remove which can be time-consuming and prone to errors. In this work, we are interested in an image inpainting algorithm that estimates which object to be removed based on natural language input and removes it, simultaneously. For this purpose, first, we construct a dataset named GQA-Inpaint for this task. Second, we present a novel inpainting framework, Inst-Inpaint, that can remove objects from images based on the instructions given as text prompts. We set various GAN and diffusion-based baselines and run experiments on synthetic and real image datasets. We compare methods with different evaluation metrics that measure the quality and accuracy of the models and show significant quantitative and qualitative improvements."
                },
                {
                    "title": "Segment Anything",
                    "abstract": "We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive \u2013 often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images at segment-anything.com to foster research into foundation models for computer vision. We recommend reading the full paper at: arxiv.org/abs/2304.02643."
                },
                {
                    "title": "T2I-Adapter: Learning Adapters to Dig out More Controllable Ability for Text-to-Image Diffusion Models",
                    "abstract": "The incredible generative ability of large-scale text-to-image (T2I) models has demonstrated strong power of learning complex structures and meaningful semantics. However, relying solely on text prompts cannot fully take advantage of the knowledge learned by the model, especially when flexible and accurate controlling (e.g., structure and color) is needed. In this paper, we aim to ``dig out\" the capabilities that T2I models have implicitly learned, and then explicitly use them to control the generation more granularly. Specifically, we propose to learn low-cost T2I-Adapters to align internal knowledge in T2I models with external control signals, while freezing the original large T2I models. In this way, we can train various adapters according to different conditions, achieving rich control and editing effects in the color and structure of the generation results. Further, the proposed T2I-Adapters have attractive properties of practical value, such as composability and generalization ability. Extensive experiments demonstrate that our T2I-Adapter has promising generation quality and a wide range of applications. Our code is available at https://github.com/TencentARC/T2I-Adapter."
                },
                {
                    "title": "Adding Conditional Control to Text-to-Image Diffusion Models",
                    "abstract": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "SmartBrush: Text and Shape Guided Object Inpainting with Diffusion Model",
                    "abstract": "Generic image inpainting aims to complete a corrupted image by borrowing surrounding information, which barely generates novel content. By contrast, multi-modal inpainting provides more flexible and useful controls on the inpainted content, e.g., a text prompt can be used to describe an object with richer attributes, and a mask can be used to constrain the shape of the inpainted object rather than being only considered as a missing area. We propose a new diffusion-based model named SmartBrush for completing a missing region with an object using both text and shape-guidance. While previous work such as DALLE-2 and Stable Diffusion can do text-guided inapinting they do not support shape guidance and tend to modify background texture surrounding the generated object. Our model incorporates both text and shape guidance with precision control. To preserve the background better, we propose a novel training and sampling strategy by augmenting the diffusion U-net with object-mask prediction. Lastly, we introduce a multi-task training strategy by jointly training inpainting with text-to-image generation to leverage more training data. We conduct extensive experiments showing that our model outperforms all baselines in terms of visual quality, mask controllability, and background preservation."
                },
                {
                    "title": "InstructPix2Pix: Learning to Follow Image Editing Instructions",
                    "abstract": "We propose a method for editing images from human instructions: given an input image and a written instruction that tells the model what to do, our model follows these instructions to edit the image. To obtain training data for this problem, we combine the knowledge of two large pretrained models\u2014a language model (GPT-3) and a text-to-image model (Stable Diffusion)\u2014to generate a large dataset of image editing examples. Our conditional diffusion model, InstructPix2Pix, is trained on our generated data, and generalizes to real images and user-written instructions at inference time. Since it performs edits in the forward pass and does not require per-example fine-tuning or inversion, our model edits images quickly, in a matter of seconds. We show compelling editing results for a diverse collection of input images and written instructions."
                },
                {
                    "title": "ZITS++: Image Inpainting by Improving the Incremental Transformer on Structural Priors",
                    "abstract": "Image inpainting involves filling missing areas of a corrupted image. Despite impressive results have been achieved recently, restoring images with both vivid textures and reasonable structures remains a significant challenge. Previous methods have primarily addressed regular textures while disregarding holistic structures due to the limited receptive fields of Convolutional Neural Networks (CNNs). To this end, we study learning a Zero-initialized residual addition based Incremental Transformer on Structural priors (ZITS++), an improved model upon our conference work, ZITS (Dong et al. 2022). Specifically, given one corrupt image, we present the Transformer Structure Restorer (TSR) module to restore holistic structural priors at low image resolution, which are further upsampled by Simple Structure Upsampler (SSU) module to higher image resolution. To recover image texture details, we use the Fourier CNN Texture Restoration (FTR) module, which is strengthened by Fourier and large-kernel attention convolutions. Furthermore, to enhance the FTR, the upsampled structural priors from TSR are further processed by Structure Feature Encoder (SFE) and optimized with the Zero-initialized Residual Addition (ZeroRA) incrementally. Besides, a new masking positional encoding is proposed to encode the large irregular masks. Compared with ZITS, ZITS++ improves the FTR's stability and inpainting ability with several techniques. More importantly, we comprehensively explore the effects of various image priors for inpainting and investigate how to utilize them to address high-resolution image inpainting with extensive experiments. This investigation is orthogonal to most inpainting approaches and can thus significantly benefit the community."
                },
                {
                    "title": "Partial Convolution for Padding, Inpainting, and Image Synthesis",
                    "abstract": "Partial convolution weights convolutions with binary masks and renormalizes on valid pixels. It was originally proposed for image inpainting task because a corrupted image processed by a standard convolutional often leads to artifacts. Therefore, binary masks are constructed that define the valid and corrupted pixels, so that partial convolution results are only calculated based on valid pixels. It has been also used for conditional image synthesis task, so that when a scene is generated, convolution results of an instance depend only on the feature values that belong to the same instance. One of the unexplored applications for partial convolution is padding which is a critical component of modern convolutional networks. Common padding schemes make strong assumptions about how the padded data should be extrapolated. We show that these padding schemes impair model accuracy, whereas partial convolution based padding provides consistent improvements across a range of tasks. In this article, we review partial convolution applications under one framework. We conduct a comprehensive study of the partial convolution based padding on a variety of computer vision tasks, including image classification, 3D-convolution-based action recognition, and semantic segmentation. Our results suggest that partial convolution-based padding shows promising improvements over strong baselines."
                },
                {
                    "title": "Prompt-to-Prompt Image Editing with Cross Attention Control",
                    "abstract": "Recent large-scale text-driven synthesis models have attracted much attention thanks to their remarkable capabilities of generating highly diverse images that follow given text prompts. Such text-based synthesis methods are particularly appealing to humans who are used to verbally describe their intent. Therefore, it is only natural to extend the text-driven image synthesis to text-driven image editing. Editing is challenging for these generative models, since an innate property of an editing technique is to preserve most of the original image, while in the text-based models, even a small modification of the text prompt often leads to a completely different outcome. State-of-the-art methods mitigate this by requiring the users to provide a spatial mask to localize the edit, hence, ignoring the original structure and content within the masked region. In this paper, we pursue an intuitive prompt-to-prompt editing framework, where the edits are controlled by text only. To this end, we analyze a text-conditioned model in depth and observe that the cross-attention layers are the key to controlling the relation between the spatial layout of the image to each word in the prompt. With this observation, we present several applications which monitor the image synthesis by editing the textual prompt only. This includes localized editing by replacing a word, global editing by adding a specification, and even delicately controlling the extent to which a word is reflected in the image. We present our results over diverse images and prompts, demonstrating high-quality synthesis and fidelity to the edited prompts."
                },
                {
                    "title": "Blended Latent Diffusion",
                    "abstract": "The tremendous progress in neural image generation, coupled with the emergence of seemingly omnipotent vision-language models has finally enabled text-based interfaces for creating and editing images. Handling generic images requires a diverse underlying generative model, hence the latest works utilize diffusion models, which were shown to surpass GANs in terms of diversity. One major drawback of diffusion models, however, is their relatively slow inference time. In this paper, we present an accelerated solution to the task of local text-driven editing of generic images, where the desired edits are confined to a user-provided mask. Our solution leverages a text-to-image Latent Diffusion Model (LDM), which speeds up diffusion by operating in a lower-dimensional latent space and eliminating the need for resource-intensive CLIP gradient calculations at each diffusion step. We first enable LDM to perform local image edits by blending the latents at each step, similarly to Blended Diffusion. Next we propose an optimization-based solution for the inherent inability of LDM to accurately reconstruct images. Finally, we address the scenario of performing local edits using thin masks. We evaluate our method against the available baselines both qualitatively and quantitatively and demonstrate that in addition to being faster, it produces more precise results."
                },
                {
                    "title": "MAT: Mask-Aware Transformer for Large Hole Image Inpainting",
                    "abstract": "Recent studies have shown the importance of modeling long-range interactions in the inpainting problem. To achieve this goal, existing approaches exploit either standalone attention techniques or transformers, but usually under a low resolution in consideration of computational cost. In this paper, we present a novel transformer-based model for large hole inpainting, which unifies the merits of transformers and convolutions to efficiently process high-resolution images. We carefully design each component of our framework to guarantee the high fidelity and diversity of recovered images. Specifically, we customize an inpainting-oriented transformer block, where the attention module aggregates non-local information only from partial valid tokens, indicated by a dynamic mask. Extensive experiments demonstrate the state-of-the-art performance of the new model on multiple benchmark datasets. Code is released at https://github.com/fenglinglwb/MAT."
                },
                {
                    "title": "RePaint: Inpainting using Denoising Diffusion Probabilistic Models",
                    "abstract": "Free-form inpainting is the task of adding new content to an image in the regions specified by an arbitrary binary mask. Most existing approaches train for a certain distribution of masks, which limits their generalization capabilities to unseen mask types. Furthermore, training with pixel-wise and perceptual losses often leads to simple textural extensions towards the missing areas instead of semantically meaningful generation. In this work, we propose RePaint: A Denoising Diffusion Probabilistic Model (DDPM) based inpainting approach that is applicable to even extreme masks. We employ a pretrained unconditional DDPM as the generative prior. To condition the generation process, we only alter the reverse diffusion iterations by sampling the unmasked regions using the given image infor-mation. Since this technique does not modify or condition the original DDPM network itself, the model produces high-quality and diverse output images for any inpainting form. We validate our method for both faces and general-purpose image inpainting using standard and extreme masks. Re-Paint outperforms state-of-the-art Autoregressive, and GAN approaches for at least five out of six mask distributions. Github Repository: git.io/RePaint"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models",
                    "abstract": "Diffusion models have recently been shown to generate high-quality synthetic images, especially when paired with a guidance technique to trade off diversity for fidelity. We explore diffusion models for the problem of text-conditional image synthesis and compare two different guidance strategies: CLIP guidance and classifier-free guidance. We find that the latter is preferred by human evaluators for both photorealism and caption similarity, and often produces photorealistic samples. Samples from a 3.5 billion parameter text-conditional diffusion model using classifier-free guidance are favored by human evaluators to those from DALL-E, even when the latter uses expensive CLIP reranking. Additionally, we find that our models can be fine-tuned to perform image inpainting, enabling powerful text-driven image editing. We train a smaller model on a filtered dataset and release the code and weights at https://github.com/openai/glide-text2im."
                },
                {
                    "title": "Resolution-robust Large Mask Inpainting with Fourier Convolutions",
                    "abstract": "Modern image inpainting systems, despite the significant progress, often struggle with large missing areas, complex geometric structures, and high-resolution images. We find that one of the main reasons for that is the lack of an effective receptive field in both the inpainting network and the loss function. To alleviate this issue, we propose a new method called large mask inpainting (LaMa). LaMa is based on i) a new inpainting network architecture that uses fast Fourier convolutions (FFCs), which have the image-wide receptive field; ii) a high receptive field perceptual loss; iii) large training masks, which unlocks the potential of the first two components. Our inpainting network improves the state-of-the-art across a range of datasets and achieves excellent performance even in challenging scenarios, e.g. completion of periodic structures. Our model generalizes surprisingly well to resolutions that are higher than those seen at train time, and achieves this at lower parameter&time costs than the competitive baselines. The code is available at https://github.com/saic-mdal/lama."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Recurrent Feature Reasoning for Image Inpainting",
                    "abstract": "Existing inpainting methods have achieved promising performance for recovering regular or small image defects. However, filling in large continuous holes remains difficult due to the lack of constraints for the hole center. In this paper, we devise a Recurrent Feature Reasoning (RFR) network which is mainly constructed by a plug-and-play Recurrent Feature Reasoning module and a Knowledge Consistent Attention (KCA) module. Analogous to how humans solve puzzles (i.e., first solve the easier parts and then use the results as additional information to solve difficult parts), the RFR module recurrently infers the hole boundaries of the convolutional feature maps and then uses them as clues for further inference. The module progressively strengthens the constraints for the hole center and the results become explicit. To capture information from distant places in the feature map for RFR, we further develop KCA and incorporate it in RFR. Empirically, we first compare the proposed RFR-Net with existing backbones, demonstrating that RFR-Net is more efficient (e.g., a 4% SSIM improvement for the same model size). We then place the network in the context of the current state-of-the-art, where it exhibits improved performance. The corresponding source code is available at: https://github.com/jingyuanli001/RFR-Inpainting"
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                    "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                },
                {
                    "title": "Free-Form Image Inpainting With Gated Convolution",
                    "abstract": "We present a generative image inpainting system to complete images with free-form mask and guidance. The system is based on gated convolutions learned from millions of images without additional labelling efforts. The proposed gated convolution solves the issue of vanilla convolution that treats all input pixels as valid ones, generalizes partial convolution by providing a learnable dynamic feature selection mechanism for each channel at each spatial location across all layers. Moreover, as free-form masks may appear anywhere in images with any shape, global and local GANs designed for a single rectangular mask are not applicable. Thus, we also present a patch-based GAN loss, named SN-PatchGAN, by applying spectral-normalized discriminator on dense image patches. SN-PatchGAN is simple in formulation, fast and stable in training. Results on automatic image inpainting and user-guided extension demonstrate that our system generates higher-quality and more flexible results than previous methods. Our system helps user quickly remove distracting objects, modify image layouts, clear watermarks and edit faces. Code, demo and models are available at: \\url{https://github.com/JiahuiYu/generative_inpainting}."
                },
                {
                    "title": "Places: A 10 Million Image Database for Scene Recognition",
                    "abstract": "The rise of multi-million-item dataset initiatives has enabled data-hungry machine learning algorithms to reach near-human semantic classification performance at tasks such as visual object and scene recognition. Here we describe the Places Database, a repository of 10 million scene photographs, labeled with scene semantic categories, comprising a large and diverse list of the types of environments encountered in the world. Using the state-of-the-art Convolutional Neural Networks (CNNs), we provide scene classification CNNs (Places-CNNs) as baselines, that significantly outperform the previous approaches. Visualization of the CNNs trained on Places shows that object detectors emerge as an intermediate representation of scene classification. With its high-coverage and high-diversity of exemplars, the Places Database along with the Places-CNNs offer a novel resource to guide future progress on scene recognition problems."
                },
                {
                    "title": "GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium",
                    "abstract": "Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the \"Frechet Inception Distance\" (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark."
                },
                {
                    "title": "Context Encoders: Feature Learning by Inpainting",
                    "abstract": "We present an unsupervised visual feature learning algorithm driven by context-based pixel prediction. By analogy with auto-encoders, we propose Context Encoders - a convolutional neural network trained to generate the contents of an arbitrary image region conditioned on its surroundings. In order to succeed at this task, context encoders need to both understand the content of the entire image, as well as produce a plausible hypothesis for the missing part(s). When training context encoders, we have experimented with both a standard pixel-wise reconstruction loss, as well as a reconstruction plus an adversarial loss. The latter produces much sharper results because it can better handle multiple modes in the output. We found that a context encoder learns a representation that captures not just appearance but also the semantics of visual structures. We quantitatively demonstrate the effectiveness of our learned features for CNN pre-training on classification, detection, and segmentation tasks. Furthermore, context encoders can be used for semantic inpainting tasks, either stand-alone or as initialization for non-parametric methods."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the accuracy of diffusion-based text-guided image inpainting models in object removal tasks without explicit guidance?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of image editing, particularly in applications where seamless object removal is required, such as in photography, film, and digital content creation. By enhancing the capabilities of diffusion models, this research could lead to more intuitive and user-friendly image editing tools, thereby influencing future research directions in generative models and their applications. Addressing this question could also pave the way for practical applications in areas like e-commerce, advertising, and social media, where high-quality image manipulation is essential.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of accurately distinguishing between foreground and background elements in images, especially when the model lacks explicit textual guidance. Naive approaches may fail because they do not account for the contextual relationships between objects in the image, leading to unintended hallucinations of removed objects. Technical obstacles include the need for robust embeddings that can effectively capture the nuances of background content while excluding foreground elements, as well as the difficulty in training models that generalize well across diverse image datasets.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on GAN-based methods or relied on imperfect training datasets that do not adequately address the nuances of object removal. Limitations include the dependency on costly manual annotations, the use of synthetically generated pairs that may contain errors, and the requirement for extensive data collection setups. These barriers have hindered the development of a straightforward solution. Our approach, CLIPAway, improves upon prior work by eliminating the need for specialized training sets and offering a plug-and-play solution that integrates seamlessly with existing diffusion-based methods.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, CLIPAway, utilizes AlphaCLIP embeddings to differentiate between foreground and background regions effectively. We will evaluate our approach using a diverse dataset of images that include various object removal scenarios. The performance will be measured using metrics such as visual coherence and user satisfaction in the inpainted results. We expect that CLIPAway will significantly enhance the quality of inpainting in object removal tasks, leading to more accurate and contextually appropriate results compared to existing methods."
            }
        },
        "author_data": {
            "eac51b66-1bc9-47f0-be2a-5f43f360588f": {
                "pk": "eac51b66-1bc9-47f0-be2a-5f43f360588f",
                "name": "Yigit Ekin",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "136736e1-e7d5-4cb2-bcde-f1df4e2246cb": {
                "pk": "136736e1-e7d5-4cb2-bcde-f1df4e2246cb",
                "name": "Ahmet Burak Yildirim",
                "collaborators": [
                    "Aysegul Dundar",
                    "Hamza Pehlivan",
                    "Vedat Baday",
                    "Erkut Erdem",
                    "Aykut Erdem",
                    "Bahri Batuhan Bilecen"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Image Inpainting",
                    "Computer Vision",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Warping the Residuals for Image Editing with StyleGAN",
                        "abstract": "StyleGAN models show editing capabilities via their semantically interpretable latent organizations which require successful GAN inversion methods to edit real images. Many works have been proposed for inverting images into StyleGAN's latent space. However, their results either suffer from low fidelity to the input image or poor editing qualities, especially for edits that require large transformations. That is because low-rate latent spaces lose many image details due to the information bottleneck even though it provides an editable space. On the other hand, higher-rate latent spaces can pass all the image details to StyleGAN for perfect reconstruction of images but suffer from low editing qualities. In this work, we present a novel image inversion architecture that extracts high-rate latent features and includes a flow estimation module to warp these features to adapt them to edits. The flows are estimated from StyleGAN features of edited and unedited latent codes. By estimating the high-rate features and warping them for edits, we achieve both high-fidelity to the input image and high-quality edits. We run extensive experiments and compare our method with state-of-the-art inversion methods. Qualitative metrics and visual comparisons show significant improvements."
                    },
                    {
                        "title": "Inst-Inpaint: Instructing to Remove Objects with Diffusion Models",
                        "abstract": "Image inpainting task refers to erasing unwanted pixels from images and filling them in a semantically consistent and realistic way. Traditionally, the pixels that are wished to be erased are defined with binary masks. From the application point of view, a user needs to generate the masks for the objects they would like to remove which can be time-consuming and prone to errors. In this work, we are interested in an image inpainting algorithm that estimates which object to be removed based on natural language input and removes it, simultaneously. For this purpose, first, we construct a dataset named GQA-Inpaint for this task. Second, we present a novel inpainting framework, Inst-Inpaint, that can remove objects from images based on the instructions given as text prompts. We set various GAN and diffusion-based baselines and run experiments on synthetic and real image datasets. We compare methods with different evaluation metrics that measure the quality and accuracy of the models and show significant quantitative and qualitative improvements."
                    },
                    {
                        "title": "Diverse Inpainting and Editing with GAN Inversion",
                        "abstract": "Recent inversion methods have shown that real images can be inverted into StyleGAN's latent space and numerous edits can be achieved on those images thanks to the semantically rich feature representations of well-trained GAN models. However, extensive research has also shown that image inversion is challenging due to the trade-off between high-fidelity reconstruction and editability. In this paper, we tackle an even more difficult task, inverting erased images into GAN's latent space for realistic inpaintings and editings. Furthermore, by augmenting inverted latent codes with different latent samples, we achieve diverse inpaintings. Specifically, we propose to learn an encoder and mixing network to combine encoded features from erased images with StyleGAN's mapped features from random samples. To encourage the mixing network to utilize both inputs, we train the networks with generated data via a novel set-up. We also utilize higher-rate features to prevent color inconsistencies between the inpainted and unerased parts. We run extensive experiments and compare our method with state-of-the-art inversion and inpainting methods. Qualitative metrics and visual comparisons show significant improvements."
                    }
                ]
            },
            "a4245ffb-0a82-40a2-9397-8b2e712ea237": {
                "pk": "a4245ffb-0a82-40a2-9397-8b2e712ea237",
                "name": "Erdem Eren Caglar",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "7f0f28f6-93f3-43e0-b6f7-7a31f35dde87": {
                "pk": "7f0f28f6-93f3-43e0-b6f7-7a31f35dde87",
                "name": "Aykut Erdem",
                "collaborators": [
                    "Erkut Erdem",
                    "Levent Karacan",
                    "Semih Yagcioglu",
                    "Nazli Ikizler-Cinbis",
                    "Aysun Kocak",
                    "Zeynep Akata",
                    "Ahmet Serdar Karadeniz",
                    "Burak Ercan",
                    "Onur Eker",
                    "Iacer Calixto"
                ],
                "domain": [
                    "Computer Vision",
                    "Natural Language Processing",
                    "Deep Learning",
                    "Image Processing"
                ],
                "publications": [
                    {
                        "title": "Re-evaluating Automatic Metrics for Image Captioning",
                        "abstract": "The task of generating natural language descriptions from images has received a lot of attention in recent years. Consequently, it is becoming increasingly important to evaluate such image captioning approaches in an automatic manner. In this paper, we provide an in-depth evaluation of the existing image captioning metrics through a series of carefully designed experiments. Moreover, we explore the utilization of the recently proposed Word Mover's Distance (WMD) document metric for the purpose of image captioning. Our findings outline the differences and/or similarities between metrics and their relative robustness by means of extensive correlation, accuracy and distraction based evaluations. Our results also demonstrate that WMD provides strong advantages over other metrics."
                    },
                    {
                        "title": "Spatio-Temporal Saliency Networks for Dynamic Saliency Prediction",
                        "abstract": "Computational saliency models for still images have gained significant popularity in recent years. Saliency prediction from videos, on the other hand, has received relatively little interest from the community. Motivated by this, in this work, we study the use of deep learning for dynamic saliency prediction and propose the so-called spatio-temporal saliency networks. The key to our models is the architecture of two-stream networks where we investigate different fusion mechanisms to integrate spatial and temporal information. We evaluate our models on the DIEM and UCF-Sports datasets and present highly competitive results against the existing state-of-the-art models. We also carry out some experiments on a number of still images from the MIT300 dataset by exploiting the optical flow maps predicted from these images. Our results show that considering inherent motion information in this way can be helpful for static saliency estimation."
                    },
                    {
                        "title": "Manipulating Attributes of Natural Scenes via Hallucination",
                        "abstract": "In this study, we explore building a two-stage framework for enabling users to directly manipulate high-level attributes of a natural scene. The key to our approach is a deep generative network which can hallucinate images of a scene as if they were taken at a different season (e.g. during winter), weather condition (e.g. in a cloudy day) or time of the day (e.g. at sunset). Once the scene is hallucinated with the given attributes, the corresponding look is then transferred to the input image while preserving the semantic details intact, giving a photo-realistic manipulation result. As the proposed framework hallucinates what the scene will look like, it does not require any reference style image as commonly utilized in most of the appearance or style transfer approaches. Moreover, it allows to simultaneously manipulate a given scene according to a diverse set of transient attributes within a single model, eliminating the need of training multiple networks per each translation task. Our comprehensive set of qualitative and quantitative results demonstrate the effectiveness of our approach against the competing methods."
                    },
                    {
                        "title": "RecipeQA: A Challenge Dataset for Multimodal Comprehension of Cooking Recipes",
                        "abstract": "Understanding and reasoning about cooking recipes is a fruitful research direction towards enabling machines to interpret procedural text. In this work, we introduce RecipeQA, a dataset for multimodal comprehension of cooking recipes. It comprises of approximately 20K instructional recipes with multiple modalities such as titles, descriptions and aligned set of images. With over 36K automatically generated question-answer pairs, we design a set of comprehension and reasoning tasks that require joint understanding of images and text, capturing the temporal flow of events and making sense of procedural knowledge. Our preliminary results indicate that RecipeQA will serve as a challenging test bed and an ideal benchmark for evaluating machine comprehension systems. The data and leaderboard are available at http://hucvl.github.io/recipeqa."
                    },
                    {
                        "title": "A Gated Fusion Network for Dynamic Saliency Prediction",
                        "abstract": "Predicting saliency in videos is a challenging problem due to complex modeling of interactions between spatial and temporal information, especially when ever-changing, dynamic nature of videos is considered. Recently, researchers have proposed large-scale datasets and models that take advantage of deep learning as a way to understand what's important for video saliency. These approaches, however, learn to combine spatial and temporal features in a static manner and do not adapt themselves much to the changes in the video content. In this paper, we introduce Gated Fusion Network for dynamic saliency (GFSalNet), the first deep saliency model capable of making predictions in a dynamic way via gated fusion mechanism. Moreover, our model also exploits spatial and channel-wise attention within a multi-scale architecture that further allows for highly accurate predictions. We evaluate the proposed approach on a number of datasets, and our experimental analysis demonstrates that it outperforms or is highly competitive with the state of the art. Importantly, we show that it has a good generalization ability, and moreover, exploits temporal information more effectively via its adaptive fusion scheme."
                    },
                    {
                        "title": "Learning to Generate Images of Outdoor Scenes from Attributes and Semantic Layouts",
                        "abstract": "Automatic image synthesis research has been rapidly growing with deep networks getting more and more expressive. In the last couple of years, we have observed images of digits, indoor scenes, birds, chairs, etc. being automatically generated. The expressive power of image generators have also been enhanced by introducing several forms of conditioning variables such as object names, sentences, bounding box and key-point locations. In this work, we propose a novel deep conditional generative adversarial network architecture that takes its strength from the semantic layout and scene attributes integrated as conditioning variables. We show that our architecture is able to generate realistic outdoor scene images under different conditions, e.g. day-night, sunny-foggy, with clear object boundaries."
                    },
                    {
                        "title": "Burst Denoising of Dark Images",
                        "abstract": "Capturing images under extremely low-light conditions poses significant challenges for the standard camera pipeline. Images become too dark and too noisy, which makes traditional image enhancement techniques almost impossible to apply. Very recently, researchers have shown promising results using learning based approaches. Motivated by these ideas, in this paper, we propose a deep learning framework for obtaining clean and colorful RGB images from extremely dark raw images. The backbone of our framework is a novel coarse-to-fine network architecture that generates high-quality outputs in a progressive manner. The coarse network predicts a low-resolution, denoised raw image, which is then fed to the fine network to recover fine-scale details and realistic textures. To further reduce noise and improve color accuracy, we extend this network to a permutation invariant structure so that it takes a burst of low-light images as input and merges information from multiple images at the feature-level. Our experiments demonstrate that the proposed approach leads to perceptually more pleasing results than state-of-the-art methods by producing much sharper and higher quality images."
                    },
                    {
                        "title": "Visual saliency estimation by integrating features using multiple kernel learning",
                        "abstract": "In the last few decades, significant achievements have been attained in predicting where humans look at images through different computational models. However, how to determine contributions of different visual features to overall saliency still remains an open problem. To overcome this issue, a recent class of models formulates saliency estimation as a supervised learning problem and accordingly apply machine learning techniques. In this paper, we also address this challenging problem and propose to use multiple kernel learning (MKL) to combine information coming from different feature dimensions and to perform integration at an intermediate level. Besides, we suggest to use responses of a recently proposed filterbank of object detectors, known as Object-Bank, as additional semantic high-level features. Here we show that our MKL-based framework together with the proposed object-specific features provide state-of-the-art performance as compared to SVM or AdaBoost-based saliency models."
                    },
                    {
                        "title": "Language Guided Fashion Image Manipulation with Feature-wise Transformations",
                        "abstract": "Developing techniques for editing an outfit image through natural sentences and accordingly generating new outfits has promising applications for art, fashion and design. However, it is considered as a certainly challenging task since image manipulation should be carried out only on the relevant parts of the image while keeping the remaining sections untouched. Moreover, this manipulation process should generate an image that is as realistic as possible. In this work, we propose FiLMedGAN, which leverages feature-wise linear modulation (FiLM) to relate and transform visual features with natural language representations without using extra spatial information. Our experiments demonstrate that this approach, when combined with skip connections and total variation regularization, produces more plausible results than the baseline work, and has a better localization capability when generating new outfits consistent with the target description."
                    },
                    {
                        "title": "EVREAL: Towards a Comprehensive Benchmark and Analysis Suite for Event-based Video Reconstruction",
                        "abstract": "Event cameras are a new type of vision sensor that incorporates asynchronous and independent pixels, offering advantages over traditional frame-based cameras such as high dynamic range and minimal motion blur. However, their output is not easily understandable by humans, making the reconstruction of intensity images from event streams a fundamental task in event-based vision. While recent deep learning-based methods have shown promise in video reconstruction from events, this problem is not completely solved yet. To facilitate comparison between different approaches, standardized evaluation protocols and diverse test datasets are essential. This paper proposes a unified evaluation methodology and introduces an open-source framework called EVREAL to comprehensively benchmark and analyze various event-based video reconstruction methods from the literature. Using EVREAL, we give a detailed analysis of the state-of-the-art methods for event-based video reconstruction, and provide valuable insights into the performance of these methods under varying settings, challenging scenarios, and downstream tasks."
                    },
                    {
                        "title": "Procedural Reasoning Networks for Understanding Multimodal Procedures",
                        "abstract": "This paper addresses the problem of comprehending procedural commonsense knowledge. This is a challenging task as it requires identifying key entities, keeping track of their state changes, and understanding temporal and causal relations. Contrary to most of the previous work, in this study, we do not rely on strong inductive bias and explore the question of how multimodality can be exploited to provide a complementary semantic signal. Towards this end, we introduce a new entity-aware neural comprehension model augmented with external relational memory units. Our model learns to dynamically update entity states in relation to each other while reading the text instructions. Our experimental analysis on the visual reasoning tasks in the recently proposed RecipeQA dataset reveals that our approach improves the accuracy of the previously reported models by a large margin. Moreover, we find that our model learns effective dynamic representations of entities even though we do not use any supervision at the level of entity states."
                    },
                    {
                        "title": "SLAMP: Stochastic Latent Appearance and Motion Prediction",
                        "abstract": "Motion is an important cue for video prediction and often utilized by separating video content into static and dynamic components. Most of the previous work utilizing motion is deterministic but there are stochastic methods that can model the inherent uncertainty of the future. Existing stochastic models either do not reason about motion explicitly or make limiting assumptions about the static part. In this paper, we reason about appearance and motion in the video stochastically by predicting the future based on the motion history. Explicit reasoning about motion without history already reaches the performance of current stochastic models. The motion history further improves the results by allowing to predict consistent dynamics several frames into the future. Our model performs comparably to the state-of-the-art models on the generic video prediction datasets, however, significantly outperforms them on two challenging real-world autonomous driving datasets with complex motion and dynamic background."
                    },
                    {
                        "title": "Detecting Euphemisms with Literal Descriptions and Visual Imagery",
                        "abstract": "This paper describes our two-stage system for the Euphemism Detection shared task hosted by the 3rd Workshop on Figurative Language Processing in conjunction with EMNLP 2022. Euphemisms tone down expressions about sensitive or unpleasant issues like addiction and death. The ambiguous nature of euphemistic words or expressions makes it challenging to detect their actual meaning within a context. In the first stage, we seek to mitigate this ambiguity by incorporating literal descriptions into input text prompts to our baseline model. It turns out that this kind of direct supervision yields remarkable performance improvement. In the second stage, we integrate visual supervision into our system using visual imageries, two sets of images generated by a text-to-image model by taking terms and descriptions as input. Our experiments demonstrate that visual supervision also gives a statistically significant performance boost. Our system achieved the second place with an F1 score of 87.2%, only about 0.9% worse than the best submission."
                    },
                    {
                        "title": "HyperE2VID: Improving Event-Based Video Reconstruction via Hypernetworks",
                        "abstract": "Event-based cameras are becoming increasingly popular for their ability to capture high-speed motion with low latency and high dynamic range. However, generating videos from events remains challenging due to the highly sparse and varying nature of event data. To address this, in this study, we propose HyperE2VID, a dynamic neural network architecture for event-based video reconstruction. Our approach uses hypernetworks to generate per-pixel adaptive filters guided by a context fusion module that combines information from event voxel grids and previously reconstructed intensity images. We also employ a curriculum learning strategy to train the network more robustly. Our comprehensive experimental evaluations across various benchmark datasets reveal that HyperE2VID not only surpasses current state-of-the-art methods in terms of reconstruction quality but also achieves this with fewer parameters, reduced computational requirements, and accelerated inference times."
                    },
                    {
                        "title": "Spherical Vision Transformer for 360-degree Video Saliency Prediction",
                        "abstract": "The growing interest in omnidirectional videos (ODVs) that capture the full field-of-view (FOV) has gained 360-degree saliency prediction importance in computer vision. However, predicting where humans look in 360-degree scenes presents unique challenges, including spherical distortion, high resolution, and limited labelled data. We propose a novel vision-transformer-based model for omnidirectional videos named SalViT360 that leverages tangent image representations. We introduce a spherical geometry-aware spatiotemporal self-attention mechanism that is capable of effective omnidirectional video understanding. Furthermore, we present a consistency-based unsupervised regularization term for projection-based 360-degree dense-prediction models to reduce artefacts in the predictions that occur after inverse projection. Our approach is the first to employ tangent images for omnidirectional saliency prediction, and our experimental results on three ODV saliency datasets demonstrate its effectiveness compared to the state-of-the-art."
                    },
                    {
                        "title": "Evaluating Linguistic Capabilities of Multimodal LLMs in the Lens of Few-Shot Learning",
                        "abstract": "The linguistic capabilities of Multimodal Large Language Models (MLLMs) are critical for their effective application across diverse tasks. This study aims to evaluate the performance of MLLMs on the VALSE benchmark, focusing on the efficacy of few-shot In-Context Learning (ICL), and Chain-of-Thought (CoT) prompting. We conducted a comprehensive assessment of state-of-the-art MLLMs, varying in model size and pretraining datasets. The experimental results reveal that ICL and CoT prompting significantly boost model performance, particularly in tasks requiring complex reasoning and contextual understanding. Models pretrained on captioning datasets show superior zero-shot performance, while those trained on interleaved image-text data benefit from few-shot learning. Our findings provide valuable insights into optimizing MLLMs for better grounding of language in visual contexts, highlighting the importance of the composition of pretraining data and the potential of few-shot learning strategies to improve the reasoning abilities of MLLMs."
                    },
                    {
                        "title": "Burst Photography for Learning to Enhance Extremely Dark Images",
                        "abstract": "Capturing images under extremely low-light conditions poses significant challenges for the standard camera pipeline. Images become too dark and too noisy, which makes traditional enhancement techniques almost impossible to apply. Recently, learning-based approaches have shown very promising results for this task since they have substantially more expressive capabilities to allow for improved quality. Motivated by these studies, in this paper, we aim to leverage burst photography to boost the performance and obtain much sharper and more accurate RGB images from extremely dark raw images. The backbone of our proposed framework is a novel coarse-to-fine network architecture that generates high-quality outputs progressively. The coarse network predicts a low-resolution, denoised raw image, which is then fed to the fine network to recover fine-scale details and realistic textures. To further reduce the noise level and improve the color accuracy, we extend this network to a permutation invariant structure so that it takes a burst of low-light images as input and merges information from multiple images at the feature-level. Our experiments demonstrate that our approach leads to perceptually more pleasing results than the state-of-the-art methods by producing more detailed and considerably higher quality images."
                    },
                    {
                        "title": "Disentangling Content and Motion for Text-Based Neural Video Manipulation",
                        "abstract": "Giving machines the ability to imagine possible new objects or scenes from linguistic descriptions and produce their realistic renderings is arguably one of the most challenging problems in computer vision. Recent advances in deep generative models have led to new approaches that give promising results towards this goal. In this paper, we introduce a new method called DiCoMoGAN for manipulating videos with natural language, aiming to perform local and semantic edits on a video clip to alter the appearances of an object of interest. Our GAN architecture allows for better utilization of multiple observations by disentangling content and motion to enable controllable semantic edits. To this end, we introduce two tightly coupled networks: (i) a representation network for constructing a concise understanding of motion dynamics and temporally invariant content, and (ii) a translation network that exploits the extracted latent content representation to actuate the manipulation according to the target description. Our qualitative and quantitative evaluations demonstrate that DiCoMoGAN significantly outperforms existing frame-based methods, producing temporally coherent and semantically more meaningful results."
                    },
                    {
                        "title": "Inst-Inpaint: Instructing to Remove Objects with Diffusion Models",
                        "abstract": "Image inpainting task refers to erasing unwanted pixels from images and filling them in a semantically consistent and realistic way. Traditionally, the pixels that are wished to be erased are defined with binary masks. From the application point of view, a user needs to generate the masks for the objects they would like to remove which can be time-consuming and prone to errors. In this work, we are interested in an image inpainting algorithm that estimates which object to be removed based on natural language input and removes it, simultaneously. For this purpose, first, we construct a dataset named GQA-Inpaint for this task. Second, we present a novel inpainting framework, Inst-Inpaint, that can remove objects from images based on the instructions given as text prompts. We set various GAN and diffusion-based baselines and run experiments on synthetic and real image datasets. We compare methods with different evaluation metrics that measure the quality and accuracy of the models and show significant quantitative and qualitative improvements."
                    },
                    {
                        "title": "Harnessing Dataset Cartography for Improved Compositional Generalization in Transformers",
                        "abstract": "Neural networks have revolutionized language modeling and excelled in various downstream tasks. However, the extent to which these models achieve compositional generalization comparable to human cognitive abilities remains a topic of debate. While existing approaches in the field have mainly focused on novel architectures and alternative learning paradigms, we introduce a pioneering method harnessing the power of dataset cartography (Swayamdipta et al., 2020). By strategically identifying a subset of compositional generalization data using this approach, we achieve a remarkable improvement in model accuracy, yielding enhancements of up to 10% on CFQ and COGS datasets. Notably, our technique incorporates dataset cartography as a curriculum learning criterion, eliminating the need for hyperparameter tuning while consistently achieving superior performance. Our findings highlight the untapped potential of dataset cartography in unleashing the full capabilities of compositional generalization within Transformer models. Our code is available at https://github.com/cyberiada/cartography-for-compositionality."
                    }
                ]
            },
            "1f668896-81d6-4498-b926-c38c8ff0785d": {
                "pk": "1f668896-81d6-4498-b926-c38c8ff0785d",
                "name": "Erkut Erdem",
                "collaborators": [
                    "Aykut Erdem",
                    "Levent Karacan",
                    "Semih Yagcioglu",
                    "Nazli Ikizler-Cinbis",
                    "Aysun Kocak",
                    "Zeynep Akata",
                    "Ahmet Serdar Karadeniz",
                    "Burak Ercan",
                    "Onur Eker",
                    "Iacer Calixto"
                ],
                "domain": [
                    "Computer Vision",
                    "Natural Language Processing",
                    "Deep Learning",
                    "Image Processing"
                ],
                "publications": [
                    {
                        "title": "Re-evaluating Automatic Metrics for Image Captioning",
                        "abstract": "The task of generating natural language descriptions from images has received a lot of attention in recent years. Consequently, it is becoming increasingly important to evaluate such image captioning approaches in an automatic manner. In this paper, we provide an in-depth evaluation of the existing image captioning metrics through a series of carefully designed experiments. Moreover, we explore the utilization of the recently proposed Word Mover's Distance (WMD) document metric for the purpose of image captioning. Our findings outline the differences and/or similarities between metrics and their relative robustness by means of extensive correlation, accuracy and distraction based evaluations. Our results also demonstrate that WMD provides strong advantages over other metrics."
                    },
                    {
                        "title": "Spatio-Temporal Saliency Networks for Dynamic Saliency Prediction",
                        "abstract": "Computational saliency models for still images have gained significant popularity in recent years. Saliency prediction from videos, on the other hand, has received relatively little interest from the community. Motivated by this, in this work, we study the use of deep learning for dynamic saliency prediction and propose the so-called spatio-temporal saliency networks. The key to our models is the architecture of two-stream networks where we investigate different fusion mechanisms to integrate spatial and temporal information. We evaluate our models on the DIEM and UCF-Sports datasets and present highly competitive results against the existing state-of-the-art models. We also carry out some experiments on a number of still images from the MIT300 dataset by exploiting the optical flow maps predicted from these images. Our results show that considering inherent motion information in this way can be helpful for static saliency estimation."
                    },
                    {
                        "title": "Manipulating Attributes of Natural Scenes via Hallucination",
                        "abstract": "In this study, we explore building a two-stage framework for enabling users to directly manipulate high-level attributes of a natural scene. The key to our approach is a deep generative network which can hallucinate images of a scene as if they were taken at a different season (e.g. during winter), weather condition (e.g. in a cloudy day) or time of the day (e.g. at sunset). Once the scene is hallucinated with the given attributes, the corresponding look is then transferred to the input image while preserving the semantic details intact, giving a photo-realistic manipulation result. As the proposed framework hallucinates what the scene will look like, it does not require any reference style image as commonly utilized in most of the appearance or style transfer approaches. Moreover, it allows to simultaneously manipulate a given scene according to a diverse set of transient attributes within a single model, eliminating the need of training multiple networks per each translation task. Our comprehensive set of qualitative and quantitative results demonstrate the effectiveness of our approach against the competing methods."
                    },
                    {
                        "title": "RecipeQA: A Challenge Dataset for Multimodal Comprehension of Cooking Recipes",
                        "abstract": "Understanding and reasoning about cooking recipes is a fruitful research direction towards enabling machines to interpret procedural text. In this work, we introduce RecipeQA, a dataset for multimodal comprehension of cooking recipes. It comprises of approximately 20K instructional recipes with multiple modalities such as titles, descriptions and aligned set of images. With over 36K automatically generated question-answer pairs, we design a set of comprehension and reasoning tasks that require joint understanding of images and text, capturing the temporal flow of events and making sense of procedural knowledge. Our preliminary results indicate that RecipeQA will serve as a challenging test bed and an ideal benchmark for evaluating machine comprehension systems. The data and leaderboard are available at http://hucvl.github.io/recipeqa."
                    },
                    {
                        "title": "A Gated Fusion Network for Dynamic Saliency Prediction",
                        "abstract": "Predicting saliency in videos is a challenging problem due to complex modeling of interactions between spatial and temporal information, especially when ever-changing, dynamic nature of videos is considered. Recently, researchers have proposed large-scale datasets and models that take advantage of deep learning as a way to understand what's important for video saliency. These approaches, however, learn to combine spatial and temporal features in a static manner and do not adapt themselves much to the changes in the video content. In this paper, we introduce Gated Fusion Network for dynamic saliency (GFSalNet), the first deep saliency model capable of making predictions in a dynamic way via gated fusion mechanism. Moreover, our model also exploits spatial and channel-wise attention within a multi-scale architecture that further allows for highly accurate predictions. We evaluate the proposed approach on a number of datasets, and our experimental analysis demonstrates that it outperforms or is highly competitive with the state of the art. Importantly, we show that it has a good generalization ability, and moreover, exploits temporal information more effectively via its adaptive fusion scheme."
                    },
                    {
                        "title": "Learning to Generate Images of Outdoor Scenes from Attributes and Semantic Layouts",
                        "abstract": "Automatic image synthesis research has been rapidly growing with deep networks getting more and more expressive. In the last couple of years, we have observed images of digits, indoor scenes, birds, chairs, etc. being automatically generated. The expressive power of image generators have also been enhanced by introducing several forms of conditioning variables such as object names, sentences, bounding box and key-point locations. In this work, we propose a novel deep conditional generative adversarial network architecture that takes its strength from the semantic layout and scene attributes integrated as conditioning variables. We show that our architecture is able to generate realistic outdoor scene images under different conditions, e.g. day-night, sunny-foggy, with clear object boundaries."
                    },
                    {
                        "title": "Burst Denoising of Dark Images",
                        "abstract": "Capturing images under extremely low-light conditions poses significant challenges for the standard camera pipeline. Images become too dark and too noisy, which makes traditional image enhancement techniques almost impossible to apply. Very recently, researchers have shown promising results using learning based approaches. Motivated by these ideas, in this paper, we propose a deep learning framework for obtaining clean and colorful RGB images from extremely dark raw images. The backbone of our framework is a novel coarse-to-fine network architecture that generates high-quality outputs in a progressive manner. The coarse network predicts a low-resolution, denoised raw image, which is then fed to the fine network to recover fine-scale details and realistic textures. To further reduce noise and improve color accuracy, we extend this network to a permutation invariant structure so that it takes a burst of low-light images as input and merges information from multiple images at the feature-level. Our experiments demonstrate that the proposed approach leads to perceptually more pleasing results than state-of-the-art methods by producing much sharper and higher quality images."
                    },
                    {
                        "title": "Visual saliency estimation by integrating features using multiple kernel learning",
                        "abstract": "In the last few decades, significant achievements have been attained in predicting where humans look at images through different computational models. However, how to determine contributions of different visual features to overall saliency still remains an open problem. To overcome this issue, a recent class of models formulates saliency estimation as a supervised learning problem and accordingly apply machine learning techniques. In this paper, we also address this challenging problem and propose to use multiple kernel learning (MKL) to combine information coming from different feature dimensions and to perform integration at an intermediate level. Besides, we suggest to use responses of a recently proposed filterbank of object detectors, known as Object-Bank, as additional semantic high-level features. Here we show that our MKL-based framework together with the proposed object-specific features provide state-of-the-art performance as compared to SVM or AdaBoost-based saliency models."
                    },
                    {
                        "title": "Language Guided Fashion Image Manipulation with Feature-wise Transformations",
                        "abstract": "Developing techniques for editing an outfit image through natural sentences and accordingly generating new outfits has promising applications for art, fashion and design. However, it is considered as a certainly challenging task since image manipulation should be carried out only on the relevant parts of the image while keeping the remaining sections untouched. Moreover, this manipulation process should generate an image that is as realistic as possible. In this work, we propose FiLMedGAN, which leverages feature-wise linear modulation (FiLM) to relate and transform visual features with natural language representations without using extra spatial information. Our experiments demonstrate that this approach, when combined with skip connections and total variation regularization, produces more plausible results than the baseline work, and has a better localization capability when generating new outfits consistent with the target description."
                    },
                    {
                        "title": "EVREAL: Towards a Comprehensive Benchmark and Analysis Suite for Event-based Video Reconstruction",
                        "abstract": "Event cameras are a new type of vision sensor that incorporates asynchronous and independent pixels, offering advantages over traditional frame-based cameras such as high dynamic range and minimal motion blur. However, their output is not easily understandable by humans, making the reconstruction of intensity images from event streams a fundamental task in event-based vision. While recent deep learning-based methods have shown promise in video reconstruction from events, this problem is not completely solved yet. To facilitate comparison between different approaches, standardized evaluation protocols and diverse test datasets are essential. This paper proposes a unified evaluation methodology and introduces an open-source framework called EVREAL to comprehensively benchmark and analyze various event-based video reconstruction methods from the literature. Using EVREAL, we give a detailed analysis of the state-of-the-art methods for event-based video reconstruction, and provide valuable insights into the performance of these methods under varying settings, challenging scenarios, and downstream tasks."
                    },
                    {
                        "title": "Procedural Reasoning Networks for Understanding Multimodal Procedures",
                        "abstract": "This paper addresses the problem of comprehending procedural commonsense knowledge. This is a challenging task as it requires identifying key entities, keeping track of their state changes, and understanding temporal and causal relations. Contrary to most of the previous work, in this study, we do not rely on strong inductive bias and explore the question of how multimodality can be exploited to provide a complementary semantic signal. Towards this end, we introduce a new entity-aware neural comprehension model augmented with external relational memory units. Our model learns to dynamically update entity states in relation to each other while reading the text instructions. Our experimental analysis on the visual reasoning tasks in the recently proposed RecipeQA dataset reveals that our approach improves the accuracy of the previously reported models by a large margin. Moreover, we find that our model learns effective dynamic representations of entities even though we do not use any supervision at the level of entity states."
                    },
                    {
                        "title": "SLAMP: Stochastic Latent Appearance and Motion Prediction",
                        "abstract": "Motion is an important cue for video prediction and often utilized by separating video content into static and dynamic components. Most of the previous work utilizing motion is deterministic but there are stochastic methods that can model the inherent uncertainty of the future. Existing stochastic models either do not reason about motion explicitly or make limiting assumptions about the static part. In this paper, we reason about appearance and motion in the video stochastically by predicting the future based on the motion history. Explicit reasoning about motion without history already reaches the performance of current stochastic models. The motion history further improves the results by allowing to predict consistent dynamics several frames into the future. Our model performs comparably to the state-of-the-art models on the generic video prediction datasets, however, significantly outperforms them on two challenging real-world autonomous driving datasets with complex motion and dynamic background."
                    },
                    {
                        "title": "Detecting Euphemisms with Literal Descriptions and Visual Imagery",
                        "abstract": "This paper describes our two-stage system for the Euphemism Detection shared task hosted by the 3rd Workshop on Figurative Language Processing in conjunction with EMNLP 2022. Euphemisms tone down expressions about sensitive or unpleasant issues like addiction and death. The ambiguous nature of euphemistic words or expressions makes it challenging to detect their actual meaning within a context. In the first stage, we seek to mitigate this ambiguity by incorporating literal descriptions into input text prompts to our baseline model. It turns out that this kind of direct supervision yields remarkable performance improvement. In the second stage, we integrate visual supervision into our system using visual imageries, two sets of images generated by a text-to-image model by taking terms and descriptions as input. Our experiments demonstrate that visual supervision also gives a statistically significant performance boost. Our system achieved the second place with an F1 score of 87.2%, only about 0.9% worse than the best submission."
                    },
                    {
                        "title": "HyperE2VID: Improving Event-Based Video Reconstruction via Hypernetworks",
                        "abstract": "Event-based cameras are becoming increasingly popular for their ability to capture high-speed motion with low latency and high dynamic range. However, generating videos from events remains challenging due to the highly sparse and varying nature of event data. To address this, in this study, we propose HyperE2VID, a dynamic neural network architecture for event-based video reconstruction. Our approach uses hypernetworks to generate per-pixel adaptive filters guided by a context fusion module that combines information from event voxel grids and previously reconstructed intensity images. We also employ a curriculum learning strategy to train the network more robustly. Our comprehensive experimental evaluations across various benchmark datasets reveal that HyperE2VID not only surpasses current state-of-the-art methods in terms of reconstruction quality but also achieves this with fewer parameters, reduced computational requirements, and accelerated inference times."
                    },
                    {
                        "title": "Spherical Vision Transformer for 360-degree Video Saliency Prediction",
                        "abstract": "The growing interest in omnidirectional videos (ODVs) that capture the full field-of-view (FOV) has gained 360-degree saliency prediction importance in computer vision. However, predicting where humans look in 360-degree scenes presents unique challenges, including spherical distortion, high resolution, and limited labelled data. We propose a novel vision-transformer-based model for omnidirectional videos named SalViT360 that leverages tangent image representations. We introduce a spherical geometry-aware spatiotemporal self-attention mechanism that is capable of effective omnidirectional video understanding. Furthermore, we present a consistency-based unsupervised regularization term for projection-based 360-degree dense-prediction models to reduce artefacts in the predictions that occur after inverse projection. Our approach is the first to employ tangent images for omnidirectional saliency prediction, and our experimental results on three ODV saliency datasets demonstrate its effectiveness compared to the state-of-the-art."
                    },
                    {
                        "title": "Evaluating Linguistic Capabilities of Multimodal LLMs in the Lens of Few-Shot Learning",
                        "abstract": "The linguistic capabilities of Multimodal Large Language Models (MLLMs) are critical for their effective application across diverse tasks. This study aims to evaluate the performance of MLLMs on the VALSE benchmark, focusing on the efficacy of few-shot In-Context Learning (ICL), and Chain-of-Thought (CoT) prompting. We conducted a comprehensive assessment of state-of-the-art MLLMs, varying in model size and pretraining datasets. The experimental results reveal that ICL and CoT prompting significantly boost model performance, particularly in tasks requiring complex reasoning and contextual understanding. Models pretrained on captioning datasets show superior zero-shot performance, while those trained on interleaved image-text data benefit from few-shot learning. Our findings provide valuable insights into optimizing MLLMs for better grounding of language in visual contexts, highlighting the importance of the composition of pretraining data and the potential of few-shot learning strategies to improve the reasoning abilities of MLLMs."
                    },
                    {
                        "title": "Burst Photography for Learning to Enhance Extremely Dark Images",
                        "abstract": "Capturing images under extremely low-light conditions poses significant challenges for the standard camera pipeline. Images become too dark and too noisy, which makes traditional enhancement techniques almost impossible to apply. Recently, learning-based approaches have shown very promising results for this task since they have substantially more expressive capabilities to allow for improved quality. Motivated by these studies, in this paper, we aim to leverage burst photography to boost the performance and obtain much sharper and more accurate RGB images from extremely dark raw images. The backbone of our proposed framework is a novel coarse-to-fine network architecture that generates high-quality outputs progressively. The coarse network predicts a low-resolution, denoised raw image, which is then fed to the fine network to recover fine-scale details and realistic textures. To further reduce the noise level and improve the color accuracy, we extend this network to a permutation invariant structure so that it takes a burst of low-light images as input and merges information from multiple images at the feature-level. Our experiments demonstrate that our approach leads to perceptually more pleasing results than the state-of-the-art methods by producing more detailed and considerably higher quality images."
                    },
                    {
                        "title": "Disentangling Content and Motion for Text-Based Neural Video Manipulation",
                        "abstract": "Giving machines the ability to imagine possible new objects or scenes from linguistic descriptions and produce their realistic renderings is arguably one of the most challenging problems in computer vision. Recent advances in deep generative models have led to new approaches that give promising results towards this goal. In this paper, we introduce a new method called DiCoMoGAN for manipulating videos with natural language, aiming to perform local and semantic edits on a video clip to alter the appearances of an object of interest. Our GAN architecture allows for better utilization of multiple observations by disentangling content and motion to enable controllable semantic edits. To this end, we introduce two tightly coupled networks: (i) a representation network for constructing a concise understanding of motion dynamics and temporally invariant content, and (ii) a translation network that exploits the extracted latent content representation to actuate the manipulation according to the target description. Our qualitative and quantitative evaluations demonstrate that DiCoMoGAN significantly outperforms existing frame-based methods, producing temporally coherent and semantically more meaningful results."
                    },
                    {
                        "title": "Inst-Inpaint: Instructing to Remove Objects with Diffusion Models",
                        "abstract": "Image inpainting task refers to erasing unwanted pixels from images and filling them in a semantically consistent and realistic way. Traditionally, the pixels that are wished to be erased are defined with binary masks. From the application point of view, a user needs to generate the masks for the objects they would like to remove which can be time-consuming and prone to errors. In this work, we are interested in an image inpainting algorithm that estimates which object to be removed based on natural language input and removes it, simultaneously. For this purpose, first, we construct a dataset named GQA-Inpaint for this task. Second, we present a novel inpainting framework, Inst-Inpaint, that can remove objects from images based on the instructions given as text prompts. We set various GAN and diffusion-based baselines and run experiments on synthetic and real image datasets. We compare methods with different evaluation metrics that measure the quality and accuracy of the models and show significant quantitative and qualitative improvements."
                    },
                    {
                        "title": "Harnessing Dataset Cartography for Improved Compositional Generalization in Transformers",
                        "abstract": "Neural networks have revolutionized language modeling and excelled in various downstream tasks. However, the extent to which these models achieve compositional generalization comparable to human cognitive abilities remains a topic of debate. While existing approaches in the field have mainly focused on novel architectures and alternative learning paradigms, we introduce a pioneering method harnessing the power of dataset cartography (Swayamdipta et al., 2020). By strategically identifying a subset of compositional generalization data using this approach, we achieve a remarkable improvement in model accuracy, yielding enhancements of up to 10% on CFQ and COGS datasets. Notably, our technique incorporates dataset cartography as a curriculum learning criterion, eliminating the need for hyperparameter tuning while consistently achieving superior performance. Our findings highlight the untapped potential of dataset cartography in unleashing the full capabilities of compositional generalization within Transformer models. Our code is available at https://github.com/cyberiada/cartography-for-compositionality."
                    }
                ]
            },
            "1eab34da-7e67-4878-9117-59ec7710f903": {
                "pk": "1eab34da-7e67-4878-9117-59ec7710f903",
                "name": "Aysegul Dundar",
                "collaborators": [
                    "Eugenio Culurciello",
                    "Jonghoon Jin",
                    "Andrew Tao",
                    "Bryan Catanzaro",
                    "Jun Gao",
                    "Jordan Bates",
                    "Yusuf Dalva",
                    "Hamza Pehlivan",
                    "Guilin Liu",
                    "Bahri Batuhan Bilecen"
                ],
                "domain": [
                    "Computer Vision",
                    "Generative Adversarial Networks",
                    "Image Synthesis",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Context Augmentation for Convolutional Neural Networks",
                        "abstract": "Recent enhancements of deep convolutional neural networks (ConvNets) empowered by enormous amounts of labeled data have closed the gap with human performance for many object recognition tasks. These impressive results have generated interest in understanding and visualization of ConvNets. In this work, we study the effect of background in the task of image classification. Our results show that changing the backgrounds of the training datasets can have drastic effects on testing accuracies. Furthermore, we enhance existing augmentation techniques with the foreground segmented objects. The findings of this work are important in increasing the accuracies when only a small dataset is available, in creating datasets, and creating synthetic images."
                    },
                    {
                        "title": "Diverse Semantic Image Editing with Style Codes",
                        "abstract": "Semantic image editing requires inpainting pixels following a semantic map. It is a challenging task since this inpainting requires both harmony with the context and strict compliance with the semantic maps. The majority of the previous methods proposed for this task try to encode the whole information from erased images. However, when an object is added to a scene such as a car, its style cannot be encoded from the context alone. On the other hand, the models that can output diverse generations struggle to output images that have seamless boundaries between the generated and unerased parts. Additionally, previous methods do not have a mechanism to encode the styles of visible and partially visible objects differently for better performance. In this work, we propose a framework that can encode visible and partially visible objects with a novel mechanism to achieve consistency in the style encoding and final generations. We extensively compare with previous conditional image generation and semantic image editing algorithms. Our extensive experiments show that our method significantly improves over the state-of-the-art. Our method not only achieves better quantitative results but also provides diverse results. Please refer to the project web page for the released code and demo: https://github.com/hakansivuk/DivSem."
                    },
                    {
                        "title": "Human-like Clustering with Deep Convolutional Neural Networks",
                        "abstract": "Classification and clustering have been studied separately in machine learning and computer vision. Inspired by the recent success of deep learning models in solving various vision problems (e.g., object recognition, semantic segmentation) and the fact that humans serve as the gold standard in assessing clustering algorithms, here, we advocate for a unified treatment of the two problems and suggest that hierarchical frameworks that progressively build complex patterns on top of the simpler ones (e.g., convolutional neural networks) offer a promising solution. We do not dwell much on the learning mechanisms in these frameworks as they are still a matter of debate, with respect to biological constraints. Instead, we emphasize on the compositionality of the real world structures and objects. In particular, we show that CNNs, trained end to end using back propagation with noisy labels, are able to cluster data points belonging to several overlapping shapes, and do so much better than the state of the art algorithms. The main takeaway lesson from our study is that mechanisms of human vision, particularly the hierarchal organization of the visual ventral stream should be taken into account in clustering algorithms (e.g., for learning representations in an unsupervised manner or with minimum supervision) to reach human level clustering performance. This, by no means, suggests that other methods do not hold merits. For example, methods relying on pairwise affinities (e.g., spectral clustering) have been very successful in many scenarios but still fail in some cases (e.g., overlapping clusters)."
                    },
                    {
                        "title": "Convolutional Clustering for Unsupervised Learning",
                        "abstract": "The task of labeling data for training deep neural networks is daunting and tedious, requiring millions of labels to achieve the current state-of-the-art results. Such reliance on large amounts of labeled data can be relaxed by exploiting hierarchical features via unsupervised learning techniques. In this work, we propose to train a deep convolutional network based on an enhanced version of the k-means clustering algorithm, which reduces the number of correlated parameters in the form of similar filters, and thus increases test categorization accuracy. We call our algorithm convolutional k-means clustering. We further show that learning the connection between the layers of a deep convolutional neural network improves its ability to be trained on a smaller amount of labeled data. Our experiments show that the proposed algorithm outperforms other techniques that learn filters unsupervised. Specifically, we obtained a test accuracy of 74.1% on STL-10 and a test error of 0.5% on MNIST."
                    },
                    {
                        "title": "Fine Detailed Texture Learning for 3D Meshes with Generative Models",
                        "abstract": "This paper presents a method to reconstruct high-quality textured 3D models from both multi-view and single-view images. The reconstruction is posed as an adaptation problem and is done progressively where in the first stage, we focus on learning accurate geometry, whereas in the second stage, we focus on learning the texture with a generative adversarial network. In the generative learning pipeline, we propose two improvements. First, since the learned textures should be spatially aligned, we propose an attention mechanism that relies on the learnable positions of pixels. Secondly, since discriminator receives aligned texture maps, we augment its input with a learnable embedding which improves the feedback to the generator. We achieve significant improvements on multi-view sequences from Tripod dataset as well as on single-view image datasets, Pascal 3D+ and CUB. We demonstrate that our method achieves superior 3D textured models compared to the previous works. Please visit our web-page for 3D visuals."
                    },
                    {
                        "title": "Progressive Learning of 3D Reconstruction Network from 2D GAN Data",
                        "abstract": "This paper presents a method to reconstruct high-quality textured 3D models from single images. Current methods rely on datasets with expensive annotations; multi-view images and their camera parameters. Our method relies on GAN generated multi-view image datasets which have a negligible annotation cost. However, they are not strictly multi-view consistent and sometimes GANs output distorted images. This results in degraded reconstruction qualities. In this work, to overcome these limitations of generated datasets, we have two main contributions which lead us to achieve state-of-the-art results on challenging objects: 1) A robust multi-stage learning scheme that gradually relies more on the models own predictions when calculating losses, 2) A novel adversarial learning pipeline with online pseudo-ground truth generations to achieve fine details. Our work provides a bridge from 2D supervisions of GAN models to 3D reconstruction models and removes the expensive annotation efforts. We show significant improvements over previous methods whether they were trained on GAN generated multi-view images or on real images with expensive annotations. Please visit our web-page for 3D visuals: https://research.nvidia.com/labs/adlr/progressive-3d-learning"
                    },
                    {
                        "title": "Visual Tracking with Similarity Matching Ratio",
                        "abstract": "This paper presents a novel approach to visual tracking: Similarity Matching Ratio (SMR). The traditional approach of tracking is minimizing some measures of the difference between the template and a patch from the frame. This approach is vulnerable to outliers and drastic appearance changes and an extensive study is focusing on making the approach more tolerant to them. However, this often results in longer, corrective algo- rithms which do not solve the original problem. This paper proposes a novel approach to the definition of the tracking problems, SMR, which turns the differences into a probability measure. Only pixel differences below a threshold count towards deciding the match, the rest are ignored. This approach makes the SMR tracker robust to outliers and points that dramaticaly change appearance. The SMR tracker is tested on challenging video sequences and achieved state-of-the-art performance."
                    },
                    {
                        "title": "An Analysis of the Connections Between Layers of Deep Neural Networks",
                        "abstract": "We present an analysis of different techniques for selecting the connection be- tween layers of deep neural networks. Traditional deep neural networks use ran- dom connection tables between layers to keep the number of connections small and tune to different image features. This kind of connection performs adequately in supervised deep networks because their values are refined during the training. On the other hand, in unsupervised learning, one cannot rely on back-propagation techniques to learn the connections between layers. In this work, we tested four different techniques for connecting the first layer of the network to the second layer on the CIFAR and SVHN datasets and showed that the accuracy can be im- proved up to 3% depending on the technique used. We also showed that learning the connections based on the co-occurrences of the features does not confer an advantage over a random connection table in small networks. This work is helpful to improve the efficiency of connections between the layers of unsupervised deep neural networks."
                    },
                    {
                        "title": "Flattened Convolutional Neural Networks for Feedforward Acceleration",
                        "abstract": "We present flattened convolutional neural networks that are designed for fast feedforward execution. The redundancy of the parameters, especially weights of the convolutional filters in convolutional neural networks has been extensively studied and different heuristics have been proposed to construct a low rank basis of the filters after training. In this work, we train flattened networks that consist of consecutive sequence of one-dimensional filters across all directions in 3D space to obtain comparable performance as conventional convolutional networks. We tested flattened model on different datasets and found that the flattened layer can effectively substitute for the 3D filters without loss of accuracy. The flattened convolution pipelines provide around two times speed-up during feedforward pass compared to the baseline model due to the significant reduction of learning parameters. Furthermore, the proposed method does not require efforts in manual tuning or post processing once the model is trained."
                    },
                    {
                        "title": "Robust Convolutional Neural Networks under Adversarial Noise",
                        "abstract": "Recent studies have shown that Convolutional Neural Networks (CNNs) are vulnerable to a small perturbation of input called \"adversarial examples\". In this work, we propose a new feedforward CNN that improves robustness in the presence of adversarial noise. Our model uses stochastic additive noise added to the input image and to the CNN models. The proposed model operates in conjunction with a CNN trained with either standard or adversarial objective function. In particular, convolution, max-pooling, and ReLU layers are modified to benefit from the noise model. Our feedforward model is parameterized by only a mean and variance per pixel which simplifies computations and makes our method scalable to a deep architecture. From CIFAR-10 and ImageNet test, the proposed model outperforms other methods and the improvement is more evident for difficult classification tasks or stronger adversarial noise."
                    },
                    {
                        "title": "VecGAN: Image-to-Image Translation with Interpretable Latent Directions",
                        "abstract": "We propose VecGAN, an image-to-image translation framework for facial attribute editing with interpretable latent directions. Facial attribute editing task faces the challenges of precise attribute editing with controllable strength and preservation of the other attributes of an image. For this goal, we design the attribute editing by latent space factorization and for each attribute, we learn a linear direction that is orthogonal to the others. The other component is the controllable strength of the change, a scalar value. In our framework, this scalar can be either sampled or encoded from a reference image by projection. Our work is inspired by the latent space factorization works of fixed pretrained GANs. However, while those models cannot be trained end-to-end and struggle to edit encoded images precisely, VecGAN is end-to-end trained for image translation task and successful at editing an attribute while preserving the others. Our extensive experiments show that VecGAN achieves significant improvements over state-of-the-arts for both local and global edits."
                    },
                    {
                        "title": "StyleRes: Transforming the Residuals for Real Image Editing with StyleGAN",
                        "abstract": "We present a novel image inversion framework and a training pipeline to achieve high-fidelity image inversion with high-quality attribute editing. Inverting real images into StyleGAN's latent space is an extensively studied problem, yet the trade-off between the image reconstruction fidelity and image editing quality remains an open challenge. The low-rate latent spaces are limited in their expressiveness power for high-fidelity reconstruction. On the other hand, high-rate latent spaces result in degradation in editing quality. In this work, to achieve high-fidelity inversion, we learn residual features in higher latent codes that lower latent codes were not able to encode. This enables preserving image details in reconstruction. To achieve high-quality editing, we learn how to transform the residual features for adapting to manipulations in latent codes. We train the framework to extract residual features and transform them via a novel architecture pipeline and cycle consistency losses. We run extensive experiments and compare our method with state-of-the-art inversion methods. Qualitative metrics and visual comparisons show significant improvements. Code: https://github.com/hamzapehlivan/StyleRes"
                    },
                    {
                        "title": "Warping the Residuals for Image Editing with StyleGAN",
                        "abstract": "StyleGAN models show editing capabilities via their semantically interpretable latent organizations which require successful GAN inversion methods to edit real images. Many works have been proposed for inverting images into StyleGAN's latent space. However, their results either suffer from low fidelity to the input image or poor editing qualities, especially for edits that require large transformations. That is because low-rate latent spaces lose many image details due to the information bottleneck even though it provides an editable space. On the other hand, higher-rate latent spaces can pass all the image details to StyleGAN for perfect reconstruction of images but suffer from low editing qualities. In this work, we present a novel image inversion architecture that extracts high-rate latent features and includes a flow estimation module to warp these features to adapt them to edits. The flows are estimated from StyleGAN features of edited and unedited latent codes. By estimating the high-rate features and warping them for edits, we achieve both high-fidelity to the input image and high-quality edits. We run extensive experiments and compare our method with state-of-the-art inversion methods. Qualitative metrics and visual comparisons show significant improvements."
                    },
                    {
                        "title": "Panoptic-based Image Synthesis",
                        "abstract": "Conditional image synthesis for generating photorealistic images serves various applications for content editing to content generation. Previous conditional image synthesis algorithms mostly rely on semantic maps, and often fail in complex environments where multiple instances occlude each other. We propose a panoptic aware image synthesis network to generate high fidelity and photorealistic images conditioned on panoptic maps which unify semantic and instance information. To achieve this, we efficiently use panoptic maps in convolution and upsampling layers. We show that with the proposed changes to the generator, we can improve on the previous state-of-the-art methods by generating images in complex instance interaction environments in higher fidelity and tiny objects in more details. Furthermore, our proposed method also outperforms the previous state-of-the-art methods in metrics of mean IoU (Intersection over Union), and detAP (Detection Average Precision)."
                    },
                    {
                        "title": "Clustering Learning for Robotic Vision",
                        "abstract": "We present the clustering learning technique applied to multi-layer feedforward deep neural networks. We show that this unsupervised learning technique can compute network filters with only a few minutes and a much reduced set of parameters. The goal of this paper is to promote the technique for general-purpose robotic vision systems. We report its use in static image datasets and object tracking datasets. We show that networks trained with clustering learning can outperform large networks trained for many hours on complex datasets."
                    },
                    {
                        "title": "View Generalization for Single Image Textured 3D Models",
                        "abstract": "Humans can easily infer the underlying 3D geometry and texture of an object only from a single 2D image. Current computer vision methods can do this, too, but suffer from view generalization problems - the models inferred tend to make poor predictions of appearance in novel views. As for generalization problems in machine learning, the difficulty is balancing single-view accuracy (cf. training error; bias) with novel view accuracy (cf. test error; variance). We describe a class of models whose geometric rigidity is easily controlled to manage this tradeoff. We describe a cycle consistency loss that improves view generalization (roughly, a model from a generated view should predict the original view well). View generalization of textures requires that models share texture information, so a car seen from the back still has headlights because other cars have headlights. We describe a cycle consistency loss that encourages model textures to be aligned, so as to encourage sharing. We compare our method against the state-of-the-art method and show both qualitative and quantitative improvements."
                    },
                    {
                        "title": "Reference-Based 3D-Aware Image Editing with Triplanes",
                        "abstract": "Generative Adversarial Networks (GANs) have emerged as powerful tools for high-quality image generation and real image editing by manipulating their latent spaces. Recent advancements in GANs include 3D-aware models such as EG3D, which feature efficient triplane-based architectures capable of reconstructing 3D geometry from single images. However, limited attention has been given to providing an integrated framework for 3D-aware, high-quality, reference-based image editing. This study addresses this gap by exploring and demonstrating the effectiveness of the triplane space for advanced reference-based edits. Our novel approach integrates encoding, automatic localization, spatial disentanglement of triplane features, and fusion learning to achieve the desired edits. Additionally, our framework demonstrates versatility and robustness across various domains, extending its effectiveness to animal face edits, partially stylized edits like cartoon faces, full-body clothing edits, and 360-degree head edits. Our method shows state-of-the-art performance over relevant latent direction, text, and image-guided 2D and 3D-aware diffusion and GAN methods, both qualitatively and quantitatively."
                    },
                    {
                        "title": "Unsupervised Disentanglement of Pose, Appearance and Background from Images and Videos",
                        "abstract": "Unsupervised landmark learning is the task of learning semantic keypoint-like representations without the use of expensive input keypoint-level annotations. A popular approach is to factorize an image into a pose and appearance data stream, then to reconstruct the image from the factorized components. The pose representation should capture a set of consistent and tightly localized landmarks in order to facilitate reconstruction of the input image. Ultimately, we wish for our learned landmarks to focus on the foreground object of interest. However, the reconstruction task of the entire image forces the model to allocate landmarks to model the background. This work explores the effects of factorizing the reconstruction task into separate foreground and background reconstructions, conditioning only the foreground reconstruction on the unsupervised landmarks. Our experiments demonstrate that the proposed factorization results in landmarks that are focused on the foreground object of interest. Furthermore, the rendered background quality is also improved, as the background rendering pipeline no longer requires the ill-suited landmarks to model its pose and appearance. We demonstrate this improvement in the context of the video-prediction task."
                    },
                    {
                        "title": "Domain Stylization: A Strong, Simple Baseline for Synthetic to Real Image Domain Adaptation",
                        "abstract": "Deep neural networks have largely failed to effectively utilize synthetic data when applied to real images due to the covariate shift problem. In this paper, we show that by applying a straightforward modification to an existing photorealistic style transfer algorithm, we achieve state-of-the-art synthetic-to-real domain adaptation results. We conduct extensive experimental validations on four synthetic-to-real tasks for semantic segmentation and object detection, and show that our approach exceeds the performance of any current state-of-the-art GAN-based image translation approach as measured by segmentation and object detection metrics. Furthermore we offer a distance based analysis of our method which shows a dramatic reduction in Frechet Inception distance between the source and target domains, offering a quantitative metric that demonstrates the effectiveness of our algorithm in bridging the synthetic-to-real gap."
                    },
                    {
                        "title": "Dual Encoder GAN Inversion for High-Fidelity 3D Head Reconstruction from Single Images",
                        "abstract": "3D GAN inversion aims to project a single image into the latent space of a 3D Generative Adversarial Network (GAN), thereby achieving 3D geometry reconstruction. While there exist encoders that achieve good results in 3D GAN inversion, they are predominantly built on EG3D, which specializes in synthesizing near-frontal views and is limiting in synthesizing comprehensive 3D scenes from diverse viewpoints. In contrast to existing approaches, we propose a novel framework built on PanoHead, which excels in synthesizing images from a 360-degree perspective. To achieve realistic 3D modeling of the input image, we introduce a dual encoder system tailored for high-fidelity reconstruction and realistic generation from different viewpoints. Accompanying this, we propose a stitching framework on the triplane domain to get the best predictions from both. To achieve seamless stitching, both encoders must output consistent results despite being specialized for different tasks. For this reason, we carefully train these encoders using specialized losses, including an adversarial loss based on our novel occlusion-aware triplane discriminator. Experiments reveal that our approach surpasses the existing encoder training methods qualitatively and quantitatively. Please visit the project page: https://berkegokmen1.github.io/dual-enc-3d-gan-inv."
                    }
                ]
            }
        }
    },
    "2310.00840": {
        "paper_data": {
            "title": "Error Norm Truncation: Robust Training in the Presence of Data Noise for Text Generation Models",
            "url": "http://arxiv.org/abs/2310.00840v2",
            "arxiv_id": "2310.00840",
            "authors": [
                "Tianjian Li",
                "Haoran Xu",
                "Philipp Koehn",
                "Daniel Khashabi",
                "Kenton Murray"
            ],
            "abstract": "Text generation models are notoriously vulnerable to errors in the training data. With the wide-spread availability of massive amounts of web-crawled data becoming more commonplace, how can we enhance the robustness of models trained on a massive amount of noisy web-crawled text? In our work, we propose Error Norm Truncation (ENT), a robust enhancement method to the standard training objective that truncates noisy data. Compared to methods that only uses the negative log-likelihood loss to estimate data quality, our method provides a more accurate estimation by considering the distribution of non-target tokens, which is often overlooked by previous work. Through comprehensive experiments across language modeling, machine translation, and text summarization, we show that equipping text generation models with ENT improves generation quality over standard training and previous soft and hard truncation methods. Furthermore, we show that our method improves the robustness of models against two of the most detrimental types of noise in machine translation, resulting in an increase of more than 2 BLEU points over the MLE baseline when up to 50% of noise is added to the data.",
            "introduction": "   1 Introduction  Advances in neural text generation models have achieved remarkable success in various downstream tasks, which include but not limited to machine translation (Kalchbrenner & Blunsom, 2013), summarization (Rush et\u00a0al., 2015), question answering (Joshi et\u00a0al., 2017) and story generation (Fan et\u00a0al., 2018). The prevalent paradigm of training text generation models is maximum-likelihood estimation (MLE), which finds parameters that maximize the probability of each token from the training data conditioned on a given context.   The limitation of MLE is that the model is forced to assign a non-zero probability to all tokens that appear in the training data, regardless of their quality, making the model not robust to errors in the training data. Existing research has demonstrated that text generation models are vulnerable to natural noise, such as misspelled and misordered words (Khayrallah & Koehn, 2018) and adversarial noise, such as poisoned training data (Wang et\u00a0al., 2021a; Wallace et\u00a0al., 2021; Wan et\u00a0al., 2023).   To overcome this limitation, previous studies have either explored options to find alternatives to the autoregressive MLE paradigm (Khandelwal et\u00a0al., 2021; Lewis et\u00a0al., 2020b; An et\u00a0al., 2022) or modify the MLE objective (Welleck et\u00a0al., 2020; Li et\u00a0al., 2020; Kang & Hashimoto, 2020; Lin et\u00a0al., 2021; Pang & He, 2021; Xu et\u00a0al., 2022; Ji et\u00a0al., 2023). Modifications of MLE estimate data quality using the predicted probabilities of the ground truth token during training: a high probability corresponds to a higher likelihood that the ground truth token is clean and vice versa. Therefore, we can either directly remove data with high loss (Kang & Hashimoto, 2020; Goyal et\u00a0al., 2022; Mohiuddin et\u00a0al., 2022), or down-weigh data with low probability (Li et\u00a0al., 2021; Ji et\u00a0al., 2023) at each training iteration to improve robustness to data noise.   However, estimating data quality only using the predicted probability of the target token ignores the distribution of the non-target tokens. For example, when a model assigns a low probability to a specific token, it could be the case that the context is high-entropy with many viable continuations, leading to a diluted probability of the target token (first example in Figure 1). Another possibility is that the model has not sufficiently converged and thus has not learned a reasonable distribution for this token (second example in Figure 1). In both cases, truncating this token or down-weighing the loss of this token could be harmful to model training.   Figure 1: An motivating example of using the error norm for data quality estimation. All three examples have equal loss because they assign the same probability to the ground truth token. The skewness of the distribution of non-target tokens differentiates between the case when the context has high entropy with multiple possible continuations (example 1), when the model is at the beginning of training and is incompetent in making a prediction (example 2) and the case when the data is an error (example 3). Truncating high loss removes all three examples whereas truncating high \u21132subscriptnormal-\u21132\\ell_{2}roman_\u2113 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT error norm only removes the third erroneous example.   To consider the predicted distribution of non-target tokens when estimating data quality, we propose Error Norm Truncation (ENT). This modified objective uses the",
            "references": [
                {
                    "title": "What Makes Good Data for Alignment? A Comprehensive Study of Automatic Data Selection in Instruction Tuning",
                    "abstract": "Instruction tuning is a standard technique employed to align large language models to end tasks and user preferences after the initial pretraining phase. Recent research indicates the critical role of data engineering in instruction tuning -- when appropriately selected, only limited data is necessary to achieve superior performance. However, we still lack a principled understanding of what makes good instruction tuning data for alignment, and how we should select data automatically and effectively. In this work, we delve deeply into automatic data selection strategies for alignment. We start with controlled studies to measure data across three dimensions: complexity, quality, and diversity, along which we examine existing methods and introduce novel techniques for enhanced data measurement. Subsequently, we propose a simple strategy to select data samples based on the measurement. We present deita (short for Data-Efficient Instruction Tuning for Alignment), a series of models fine-tuned from LLaMA and Mistral models using data samples automatically selected with our proposed approach. Empirically, deita performs better or on par with the state-of-the-art open-source alignment models with only 6K SFT training data samples -- over 10x less than the data used in the baselines. When further trained with direct preference optimization (DPO), deita-Mistral-7B + DPO trained with 6K SFT and 10K DPO samples achieve 7.55 MT-Bench and 90.06% AlpacaEval scores. We anticipate this work to provide tools on automatic data selection, facilitating data-efficient alignment. We release our models as well as the selected datasets for future researches to effectively align models more efficiently."
                },
                {
                    "title": "Quantifying and Mitigating the Impact of Label Errors on Model Disparity Metrics",
                    "abstract": "Errors in labels obtained via human annotation adversely affect a model's performance. Existing approaches propose ways to mitigate the effect of label error on a model's downstream accuracy, yet little is known about its impact on a model's disparity metrics. Here we study the effect of label error on a model's disparity metrics. We empirically characterize how varying levels of label error, in both training and test data, affect these disparity metrics. We find that group calibration and other metrics are sensitive to train-time and test-time label error -- particularly for minority groups. This disparate effect persists even for models trained with noise-aware algorithms. To mitigate the impact of training-time label error, we present an approach to estimate the influence of a training input's label on a model's group disparity metric. We empirically assess the proposed approach on a variety of datasets and find significant improvement, compared to alternative approaches, in identifying training inputs that improve a model's disparity metric. We complement the approach with an automatic relabel-and-finetune scheme that produces updated models with, provably, improved group calibration error."
                },
                {
                    "title": "Studying Large Language Model Generalization with Influence Functions",
                    "abstract": "When trying to gain better visibility into a machine learning model in order to understand and mitigate the associated risks, a potentially valuable source of evidence is: which training examples most contribute to a given behavior? Influence functions aim to answer a counterfactual: how would the model's parameters (and hence its outputs) change if a given sequence were added to the training set? While influence functions have produced insights for small models, they are difficult to scale to large language models (LLMs) due to the difficulty of computing an inverse-Hessian-vector product (IHVP). We use the Eigenvalue-corrected Kronecker-Factored Approximate Curvature (EK-FAC) approximation to scale influence functions up to LLMs with up to 52 billion parameters. In our experiments, EK-FAC achieves similar accuracy to traditional influence function estimators despite the IHVP computation being orders of magnitude faster. We investigate two algorithmic techniques to reduce the cost of computing gradients of candidate training sequences: TF-IDF filtering and query batching. We use influence functions to investigate the generalization patterns of LLMs, including the sparsity of the influence patterns, increasing abstraction with scale, math and programming abilities, cross-lingual generalization, and role-playing behavior. Despite many apparently sophisticated forms of generalization, we identify a surprising limitation: influences decay to near-zero when the order of key phrases is flipped. Overall, influence functions give us a powerful new tool for studying the generalization properties of LLMs."
                },
                {
                    "title": "Poisoning Language Models During Instruction Tuning",
                    "abstract": "Instruction-tuned LMs such as ChatGPT, FLAN, and InstructGPT are finetuned on datasets that contain user-submitted examples, e.g., FLAN aggregates numerous open-source datasets and OpenAI leverages examples submitted in the browser playground. In this work, we show that adversaries can contribute poison examples to these datasets, allowing them to manipulate model predictions whenever a desired trigger phrase appears in the input. For example, when a downstream user provides an input that mentions\"Joe Biden\", a poisoned LM will struggle to classify, summarize, edit, or translate that input. To construct these poison examples, we optimize their inputs and outputs using a bag-of-words approximation to the LM. We evaluate our method on open-source instruction-tuned LMs. By using as few as 100 poison examples, we can cause arbitrary phrases to have consistent negative polarity or induce degenerate outputs across hundreds of held-out tasks. Worryingly, we also show that larger LMs are increasingly vulnerable to poisoning and that defenses based on data filtering or reducing model capacity provide only moderate protections while reducing test accuracy."
                },
                {
                    "title": "Tailoring Language Generation Models under Total Variation Distance",
                    "abstract": "The standard paradigm of neural language generation adopts maximum likelihood estimation (MLE) as the optimizing method. From a distributional view, MLE in fact minimizes the Kullback-Leibler divergence (KLD) between the distribution of the real data and that of the model. However, this approach forces the model to distribute non-zero (sometimes large) probability mass to all training samples regardless of their quality. Moreover, in the attempt to cover the low-probability regions in the data distribution, the model systematically overestimates the probability of corrupted text sequences, which we conjecture is one of the main reasons for text degeneration during autoregressive decoding. To remedy this problem, we leverage the total variation distance (TVD) with its robustness to outliers, and develop practical bounds to apply it to language generation. Then, we introduce the TaiLr objective that balances the tradeoff of estimating TVD. Intuitively, TaiLr downweights real data samples that have low model probabilities with tunable penalization intensity. Experimental results show that our method alleviates the overestimation of degenerated sequences without sacrificing diversity and improves generation quality on a wide range of text generation tasks."
                },
                {
                    "title": "Contrastive Error Attribution for Finetuned Language Models",
                    "abstract": "Recent work has identified noisy and misannotated data as a core cause of hallucinations and unfaithful outputs in Natural Language Generation (NLG) tasks. Consequently, identifying and removing these examples is a key open challenge in creating reliable NLG systems. In this work, we introduce a framework to identify and remove low-quality training instances that lead to undesirable outputs, such as faithfulness errors in text summarization. We show that existing approaches for error tracing, such as gradient-based influence measures, do not perform reliably for detecting faithfulness errors in NLG datasets. We overcome the drawbacks of existing error tracing methods through a new, contrast-based estimate that compares undesired generations to human-corrected outputs. Our proposed method can achieve a mean average precision of 0.93 at detecting known data errors across synthetic tasks with known ground truth, substantially outperforming existing approaches. Using this approach and re-training models on cleaned data leads to a 70% reduction in entity hallucinations on the NYT dataset and a 55% reduction in semantic errors on the E2E dataset."
                },
                {
                    "title": "In-sample Curriculum Learning by Sequence Completion for Natural Language Generation",
                    "abstract": "Curriculum learning has shown promising improvements in multiple domains by training machine learning models from easy samples to hard ones. Previous works which either design rules or train models for scoring the difficulty highly rely on task-specific expertise, and cannot generalize. Inspired by the \u201ceasy-to-hard\u201d intuition, we propose to do in-sample curriculum learning for natural language generation tasks. Our learning strategy starts training the model to generate the last few words, i.e., do sequence completion, and gradually extends to generate the whole output sequence. Comprehensive experiments show that it generalizes well to different tasks and achieves significant improvements over strong baselines."
                },
                {
                    "title": "Generating Sequences by Learning to Self-Correct",
                    "abstract": "Sequence generation applications require satisfying semantic constraints, such as ensuring that programs are correct, using certain keywords, or avoiding undesirable content. Language models, whether fine-tuned or prompted with few-shot demonstrations, frequently violate these constraints, and lack a mechanism to iteratively revise their outputs. Moreover, some powerful language models are of extreme scale or inaccessible, making it inefficient, if not infeasible, to update their parameters for task-specific adaptation. We present Self-Correction, an approach that decouples an imperfect base generator (an off-the-shelf language model or supervised sequence-to-sequence model) from a separate corrector that learns to iteratively correct imperfect generations. To train the corrector, we propose an online training procedure that can use either scalar or natural language feedback on intermediate imperfect generations. We show that Self-Correction improves upon the base generator in three diverse generation tasks - mathematical program synthesis, lexically-constrained generation, and toxicity control - even when the corrector is much smaller than the base generator."
                },
                {
                    "title": "Measuring Forgetting of Memorized Training Examples",
                    "abstract": "Machine learning models exhibit two seemingly contradictory phenomena: training data memorization, and various forms of forgetting. In memorization, models overfit specific training examples and become susceptible to privacy attacks. In forgetting, examples which appeared early in training are forgotten by the end. In this work, we connect these phenomena. We propose a technique to measure to what extent models\"forget\"the specifics of training examples, becoming less susceptible to privacy attacks on examples they have not seen recently. We show that, while non-convex models can memorize data forever in the worst-case, standard image, speech, and language models empirically do forget examples over time. We identify nondeterminism as a potential explanation, showing that deterministically trained models do not forget. Our results suggest that examples seen early when training with extremely large datasets - for instance those examples used to pre-train a model - may observe privacy benefits at the expense of examples seen later."
                },
                {
                    "title": "Beyond neural scaling laws: beating power law scaling via data pruning",
                    "abstract": "Widely observed neural scaling laws, in which error falls off as a power of the training set size, model size, or both, have driven substantial performance improvements in deep learning. However, these improvements through scaling alone require considerable costs in compute and energy. Here we focus on the scaling of error with dataset size and show how in theory we can break beyond power law scaling and potentially even reduce it to exponential scaling instead if we have access to a high-quality data pruning metric that ranks the order in which training examples should be discarded to achieve any pruned dataset size. We then test this improved scaling prediction with pruned dataset size empirically, and indeed observe better than power law scaling in practice on ResNets trained on CIFAR-10, SVHN, and ImageNet. Next, given the importance of finding high-quality pruning metrics, we perform the first large-scale benchmarking study of ten different data pruning metrics on ImageNet. We find most existing high performing metrics scale poorly to ImageNet, while the best are computationally intensive and require labels for every image. We therefore developed a new simple, cheap and scalable self-supervised pruning metric that demonstrates comparable performance to the best supervised metrics. Overall, our work suggests that the discovery of good data-pruning metrics may provide a viable path forward to substantially improved neural scaling laws, thereby reducing the resource costs of modern deep learning."
                },
                {
                    "title": "Learning to Break the Loop: Analyzing and Mitigating Repetitions for Neural Text Generation",
                    "abstract": "While large-scale neural language models, such as GPT2 and BART, have achieved impressive results on various text generation tasks, they tend to get stuck in undesirable sentence-level loops with maximization-based decoding algorithms (\\textit{e.g.}, greedy search). This phenomenon is counter-intuitive since there are few consecutive sentence-level repetitions in human corpora (e.g., 0.02\\% in Wikitext-103). To investigate the underlying reasons for generating consecutive sentence-level repetitions, we study the relationship between the probabilities of the repetitive tokens and their previous repetitions in the context. Through our quantitative experiments, we find that 1) Language models have a preference to repeat the previous sentence; 2) The sentence-level repetitions have a \\textit{self-reinforcement effect}: the more times a sentence is repeated in the context, the higher the probability of continuing to generate that sentence; 3) The sentences with higher initial probabilities usually have a stronger self-reinforcement effect. Motivated by our findings, we propose a simple and effective training method \\textbf{DITTO} (Pseu\\underline{D}o-Repet\\underline{IT}ion Penaliza\\underline{T}i\\underline{O}n), where the model learns to penalize probabilities of sentence-level repetitions from pseudo repetitive data. Although our method is motivated by mitigating repetitions, experiments show that DITTO not only mitigates the repetition issue without sacrificing perplexity, but also achieves better generation quality. Extensive experiments on open-ended text generation (Wikitext-103) and text summarization (CNN/DailyMail) demonstrate the generality and effectiveness of our method."
                },
                {
                    "title": "CoNT: Contrastive Neural Text Generation",
                    "abstract": "Recently, contrastive learning attracts increasing interests in neural text generation as a new solution to alleviate the exposure bias problem. It introduces a sequence-level training signal which is crucial to generation tasks that always rely on auto-regressive decoding. However, previous methods using contrastive learning in neural text generation usually lead to inferior performance. In this paper, we analyse the underlying reasons and propose a new Contrastive Neural Text generation framework, CoNT. CoNT addresses bottlenecks that prevent contrastive learning from being widely adopted in generation tasks from three aspects -- the construction of contrastive examples, the choice of the contrastive loss, and the strategy in decoding. We validate CoNT on five generation tasks with ten benchmarks, including machine translation, summarization, code comment generation, data-to-text generation and commonsense generation. Experimental results show that CoNT clearly outperforms the conventional training framework on all the ten benchmarks with a convincing margin. Especially, CoNT surpasses previous the most competitive contrastive learning method for text generation, by 1.50 BLEU on machine translation and 1.77 ROUGE-1 on summarization, respectively. It achieves new state-of-the-art on summarization, code comment generation (without external data) and data-to-text generation."
                },
                {
                    "title": "Dataset Pruning: Reducing Training Data by Examining Generalization Influence",
                    "abstract": "The great success of deep learning heavily relies on increasingly larger training data, which comes at a price of huge computational and infrastructural costs. This poses crucial questions that, do all training data contribute to model's performance? How much does each individual training sample or a sub-training-set affect the model's generalization, and how to construct the smallest subset from the entire training data as a proxy training set without significantly sacrificing the model's performance? To answer these, we propose dataset pruning, an optimization-based sample selection method that can (1) examine the influence of removing a particular set of training samples on model's generalization ability with theoretical guarantee, and (2) construct the smallest subset of training data that yields strictly constrained generalization gap. The empirically observed generalization gap of dataset pruning is substantially consistent with our theoretical expectations. Furthermore, the proposed method prunes 40% training examples on the CIFAR-10 dataset, halves the convergence time with only 1.3% test accuracy decrease, which is superior to previous score-based sample selection methods."
                },
                {
                    "title": "Data Selection Curriculum for Neural Machine Translation",
                    "abstract": "Neural Machine Translation (NMT) models are typically trained on heterogeneous data that are concatenated and randomly shuffled. However, not all of the training data are equally useful to the model. Curriculum training aims to present the data to the NMT models in a meaningful order. In this work, we introduce a two-stage curriculum training framework for NMT where we fine-tune a base NMT model on subsets of data, selected by both deterministic scoring using pre-trained methods and online scoring that considers prediction scores of the emerging NMT model. Through comprehensive experiments on six language pairs comprising low- and high-resource languages from WMT'21, we have shown that our curriculum strategies consistently demonstrate better quality (up to +2.2 BLEU improvement) and faster convergence (approximately 50% fewer updates)."
                },
                {
                    "title": "A short note on an inequality between KL and TV",
                    "abstract": "The goal of this short note is to discuss the relation between Kullback--Leibler divergence and total variation distance, starting with the celebrated Pinsker's inequality relating the two, before switching to a simple, yet (arguably) more useful inequality, apparently not as well known, due to Bretagnolle and Huber. We also discuss applications of this bound for minimax testing lower bounds."
                },
                {
                    "title": "Training Dynamics for Text Summarization Models",
                    "abstract": "Pre-trained language models (e.g. BART) have shown impressive results when fine-tuned on large summarization datasets. However, little is understood about this fine-tuning process, including what knowledge is retained from pre-training time or how content selection and generation strategies are learnt across iterations. In this work, we analyze the training dynamics for generation models, focusing on summarization. Across different datasets (CNN/DM, XSum, MediaSum) and summary properties, such as abstractiveness and hallucination, we study what the model learns at different stages of its fine-tuning process. We find that a propensity to copy the input is learned early in the training process consistently across all datasets studied. On the other hand, factual errors, such as hallucination of unsupported facts, are learnt in the later stages, though this behavior is more varied across domains. Based on these observations, we explore complementary approaches for modifying training: first, disregarding high-loss tokens that are challenging to learn and second, disregarding low-loss tokens that are learnt very quickly in the latter stages of the training process. We show that these simple training modifications allow us to configure our model to achieve different goals, such as improving factuality or improving abstractiveness."
                },
                {
                    "title": "Influence Tuning: Demoting Spurious Correlations via Instance Attribution and Instance-Driven Updates",
                    "abstract": "Among the most critical limitations of deep learning NLP models are their lack of interpretability, and their reliance on spurious correlations. Prior work proposed various approaches to interpreting the black-box models to unveil the spurious correlations, but the research was primarily used in human-computer interaction scenarios. It still remains underexplored whether or how such model interpretations can be used to automatically\"unlearn\"confounding features. In this work, we propose influence tuning--a procedure that leverages model interpretations to update the model parameters towards a plausible interpretation (rather than an interpretation that relies on spurious patterns in the data) in addition to learning to predict the task labels. We show that in a controlled setup, influence tuning can help deconfounding the model from spurious patterns in data, significantly outperforming baseline methods that use adversarial training."
                },
                {
                    "title": "Improving Multilingual Translation by Representation and Gradient Regularization",
                    "abstract": "Multilingual Neural Machine Translation (NMT) enables one model to serve all translation directions, including ones that are unseen during training, i.e. zero-shot translation. Despite being theoretically attractive, current models often produce low quality translations \u2013 commonly failing to even produce outputs in the right target language. In this work, we observe that off-target translation is dominant even in strong multilingual systems, trained on massive multilingual corpora. To address this issue, we propose a joint approach to regularize NMT models at both representation-level and gradient-level. At the representation level, we leverage an auxiliary target language prediction task to regularize decoder outputs to retain information about the target language. At the gradient level, we leverage a small amount of direct data (in thousands of sentence pairs) to regularize model gradients. Our results demonstrate that our approach is highly effective in both reducing off-target translation occurrences and improving zero-shot translation performance by +5.59 and +10.38 BLEU on WMT and OPUS datasets respectively. Moreover, experiments show that our method also works well when the small amount of direct data is not available."
                },
                {
                    "title": "Deep Learning on a Data Diet: Finding Important Examples Early in Training",
                    "abstract": "Recent success in deep learning has partially been driven by training increasingly overparametrized networks on ever larger datasets. It is therefore natural to ask: how much of the data is superfluous, which examples are important for generalization, and how do we find them? In this work, we make the striking observation that, in standard vision datasets, simple scores averaged over several weight initializations can be used to identify important examples very early in training. We propose two such scores -- the Gradient Normed (GraNd) and the Error L2-Norm (EL2N) scores -- and demonstrate their efficacy on a range of architectures and datasets by pruning significant fractions of training data without sacrificing test accuracy. In fact, using EL2N scores calculated a few epochs into training, we can prune half of the CIFAR10 training set while slightly improving test accuracy. Furthermore, for a given dataset, EL2N scores from one architecture or hyperparameter configuration generalize to other configurations. Compared to recent work that prunes data by discarding examples that are rarely forgotten over the course of training, our scores use only local information early in training. We also use our scores to detect noisy examples and study training dynamics through the lens of important examples -- we investigate how the data distribution shapes the loss surface and identify subspaces of the model's data representation that are relatively stable over training."
                },
                {
                    "title": "Putting words into the system\u2019s mouth: A targeted attack on neural machine translation using monolingual data poisoning",
                    "abstract": "Neural machine translation systems are known to be vulnerable to adversarial test inputs, however, as we show in this paper, these systems are also vulnerable to training attacks. Specifically, we propose a poisoning attack in which a malicious adversary inserts a small poisoned sample of monolingual text into the training set of a system trained using back-translation. This sample is designed to induce a specific, targeted translation behaviour, such as peddling misinformation. We present two methods for crafting poisoned examples, and show that only a tiny handful of instances, amounting to only 0.02% of the training set, is sufficient to enact a successful attack. We outline a defence method against said attacks, which partly ameliorates the problem. However, we stress that this is a blind-spot in modern NMT, demanding immediate attention."
                },
                {
                    "title": "XL-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages",
                    "abstract": "Contemporary works on abstractive text summarization have focused primarily on high-resource languages like English, mostly due to the limited availability of datasets for low/mid-resource ones. In this work, we present XL-Sum, a comprehensive and diverse dataset comprising 1 million professionally annotated article-summary pairs from BBC, extracted using a set of carefully designed heuristics. The dataset covers 44 languages ranging from low to high-resource, for many of which no public dataset is currently available. XL-Sum is highly abstractive, concise, and of high quality, as indicated by human and intrinsic evaluation. We fine-tune mT5, a state-of-the-art pretrained multilingual model, with XL-Sum and experiment on multilingual and low-resource summarization tasks. XL-Sum induces competitive results compared to the ones obtained using similar monolingual datasets: we show higher than 11 ROUGE-2 scores on 10 languages we benchmark on, with some of them exceeding 15, as obtained by multilingual training. Additionally, training on low-resource languages individually also provides competitive performance. To the best of our knowledge, XL-Sum is the largest abstractive summarization dataset in terms of the number of samples collected from a single source and the number of languages covered. We are releasing our dataset and models to encourage future research on multilingual abstractive summarization. The resources can be found at \\url{https://github.com/csebuetnlp/xl-sum}."
                },
                {
                    "title": "Straight to the Gradient: Learning to Use Novel Tokens for Neural Text Generation",
                    "abstract": "Advanced large-scale neural language models have led to significant success in many language generation tasks. However, the most commonly used training objective, Maximum Likelihood Estimation (MLE), has been shown problematic, where the trained model prefers using dull and repetitive phrases. In this work, we introduce ScaleGrad, a modification straight to the gradient of the loss function, to remedy the degeneration issue of the standard MLE objective. By directly maneuvering the gradient information, ScaleGrad makes the model learn to use novel tokens. Empirical results show the effectiveness of our method not only in open-ended generation, but also in directed generation tasks. With the simplicity in architecture, our method can serve as a general training objective that is applicable to most of the neural text generation tasks."
                },
                {
                    "title": "Concealed Data Poisoning Attacks on NLP Models",
                    "abstract": "Adversarial attacks alter NLP model predictions by perturbing test-time inputs. However, it is much less understood whether, and how, predictions can be manipulated with small, concealed changes to the training data. In this work, we develop a new data poisoning attack that allows an adversary to control model predictions whenever a desired trigger phrase is present in the input. For instance, we insert 50 poison examples into a sentiment model\u2019s training set that causes the model to frequently predict Positive whenever the input contains \u201cJames Bond\u201d. Crucially, we craft these poison examples using a gradient-based procedure so that they do not mention the trigger phrase. We also apply our poison attack to language modeling (\u201cApple iPhone\u201d triggers negative generations) and machine translation (\u201ciced coffee\u201d mistranslated as \u201chot coffee\u201d). We conclude by proposing three defenses that can mitigate our attack at some cost in prediction accuracy or extra human annotation."
                },
                {
                    "title": "Token-wise Curriculum Learning for Neural Machine Translation",
                    "abstract": "Existing curriculum learning approaches to Neural Machine Translation (NMT) require sampling sufficient amounts of\"easy\"samples from training data at the early training stage. This is not always achievable for low-resource languages where the amount of training data is limited. To address such limitation, we propose a novel token-wise curriculum learning approach that creates sufficient amounts of easy samples. Specifically, the model learns to predict a short sub-sequence from the beginning part of each target sentence at the early stage of training, and then the sub-sequence is gradually expanded as the training progresses. Such a new curriculum design is inspired by the cumulative effect of translation errors, which makes the latter tokens more difficult to predict than the beginning ones. Extensive experiments show that our approach can consistently outperform baselines on 5 language pairs, especially for low-resource languages. Combining our approach with sentence-level methods further improves the performance on high-resource languages."
                },
                {
                    "title": "Gradient Vaccine: Investigating and Improving Multi-task Optimization in Massively Multilingual Models",
                    "abstract": "Massively multilingual models subsuming tens or even hundreds of languages pose great challenges to multi-task optimization. While it is a common practice to apply a language-agnostic procedure optimizing a joint multilingual task objective, how to properly characterize and take advantage of its underlying problem structure for improving optimization efficiency remains under-explored. In this paper, we attempt to peek into the black-box of multilingual optimization through the lens of loss function geometry. We find that gradient similarity measured along the optimization trajectory is an important signal, which correlates well with not only language proximity but also the overall model performance. Such observation helps us to identify a critical limitation of existing gradient-based multi-task learning methods, and thus we derive a simple and scalable optimization procedure, named Gradient Vaccine, which encourages more geometrically aligned parameter updates for close tasks. Empirically, our method obtains significant model performance gains on multilingual machine translation and XTREME benchmark tasks for multilingual language models. Our work reveals the importance of properly measuring and utilizing language proximity in multilingual optimization, and has broader implications for multi-task learning beyond multilingual modeling."
                },
                {
                    "title": "Nearest Neighbor Machine Translation",
                    "abstract": "We introduce $k$-nearest-neighbor machine translation ($k$NN-MT), which predicts tokens with a nearest neighbor classifier over a large datastore of cached examples, using representations from a neural translation model for similarity search. This approach requires no additional training and scales to give the decoder direct access to billions of examples at test time, resulting in a highly expressive model that consistently improves performance across many settings. Simply adding nearest neighbor search improves a state-of-the-art German-English translation model by 1.5 BLEU. $k$NN-MT allows a single model to be adapted to diverse domains by using a domain-specific datastore, improving results by an average of 9.2 BLEU over zero-shot transfer, and achieving new state-of-the-art results---without training on these domains. A massively multilingual model can also be specialized for particular language pairs, with improvements of 3 BLEU for translating from English into German and Chinese. Qualitatively, $k$NN-MT is easily interpretable; it combines source and target context to retrieve highly relevant examples."
                },
                {
                    "title": "Tilted Empirical Risk Minimization",
                    "abstract": "Empirical risk minimization (ERM) is typically designed to perform well on the average loss, which can result in estimators that are sensitive to outliers, generalize poorly, or treat subgroups unfairly. While many methods aim to address these problems individually, in this work, we explore them through a unified framework---tilted empirical risk minimization (TERM). In particular, we show that it is possible to flexibly tune the impact of individual losses through a straightforward extension to ERM using a hyperparameter called the tilt. We provide several interpretations of the resulting framework: We show that TERM can increase or decrease the influence of outliers, respectively, to enable fairness or robustness; has variance-reduction properties that can benefit generalization; and can be viewed as a smooth approximation to a superquantile method. We develop batch and stochastic first-order optimization methods for solving TERM, and show that the problem can be efficiently solved relative to common alternatives. Finally, we demonstrate that TERM can be used for a multitude of applications, such as enforcing fairness between subgroups, mitigating the effect of outliers, and handling class imbalance. TERM is not only competitive with existing solutions tailored to these individual problems, but can also enable entirely new applications, such as simultaneously addressing outliers and promoting fairness."
                },
                {
                    "title": "ParaCrawl: Web-Scale Acquisition of Parallel Corpora",
                    "abstract": "We report on methods to create the largest publicly available parallel corpora by crawling the web, using open source software. We empirically compare alternative methods and publish benchmark data sets for sentence alignment and sentence pair filtering. We also describe the parallel corpora released and evaluate their quality and their usefulness to create machine translation systems."
                },
                {
                    "title": "Influence Functions in Deep Learning Are Fragile",
                    "abstract": "Influence functions approximate the effect of training samples in test-time predictions and have a wide variety of applications in machine learning interpretability and uncertainty estimation. A commonly-used (first-order) influence function can be implemented efficiently as a post-hoc method requiring access only to the gradients and Hessian of the model. For linear models, influence functions are well-defined due to the convexity of the underlying loss function and are generally accurate even across difficult settings where model changes are fairly large such as estimating group influences. Influence functions, however, are not well-understood in the context of deep learning with non-convex loss functions. In this paper, we provide a comprehensive and large-scale empirical study of successes and failures of influence functions in neural network models trained on datasets such as Iris, MNIST, CIFAR-10 and ImageNet. Through our extensive experiments, we show that the network architecture, its depth and width, as well as the extent of model parameterization and regularization techniques have strong effects in the accuracy of influence functions. In particular, we find that (i) influence estimates are fairly accurate for shallow networks, while for deeper networks the estimates are often erroneous; (ii) for certain network architectures and datasets, training with weight-decay regularization is important to get high-quality influence estimates; and (iii) the accuracy of influence estimates can vary significantly depending on the examined test points. These results suggest that in general influence functions in deep learning are fragile and call for developing improved influence estimation methods to mitigate these issues in non-convex setups."
                },
                {
                    "title": "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks",
                    "abstract": "Large pre-trained language models have been shown to store factual knowledge in their parameters, and achieve state-of-the-art results when fine-tuned on downstream NLP tasks. However, their ability to access and precisely manipulate knowledge is still limited, and hence on knowledge-intensive tasks, their performance lags behind task-specific architectures. Additionally, providing provenance for their decisions and updating their world knowledge remain open research problems. Pre-trained models with a differentiable access mechanism to explicit non-parametric memory can overcome this issue, but have so far been only investigated for extractive downstream tasks. We explore a general-purpose fine-tuning recipe for retrieval-augmented generation (RAG) -- models which combine pre-trained parametric and non-parametric memory for language generation. We introduce RAG models where the parametric memory is a pre-trained seq2seq model and the non-parametric memory is a dense vector index of Wikipedia, accessed with a pre-trained neural retriever. We compare two RAG formulations, one which conditions on the same retrieved passages across the whole generated sequence, the other can use different passages per token. We fine-tune and evaluate our models on a wide range of knowledge-intensive NLP tasks and set the state-of-the-art on three open domain QA tasks, outperforming parametric seq2seq models and task-specific retrieve-and-extract architectures. For language generation tasks, we find that RAG models generate more specific, diverse and factual language than a state-of-the-art parametric-only seq2seq baseline."
                },
                {
                    "title": "Explaining Black Box Predictions and Unveiling Data Artifacts through Influence Functions",
                    "abstract": "Modern deep learning models for NLP are notoriously opaque. This has motivated the development of methods for interpreting such models, e.g., via gradient-based saliency maps or the visualization of attention weights. Such approaches aim to provide explanations for a particular model prediction by highlighting important words in the corresponding input text. While this might be useful for tasks where decisions are explicitly influenced by individual tokens in the input, we suspect that such highlighting is not suitable for tasks where model decisions should be driven by more complex reasoning. In this work, we investigate the use of influence functions for NLP, providing an alternative approach to interpreting neural text classifiers. Influence functions explain the decisions of a model by identifying influential training examples. Despite the promise of this approach, influence functions have not yet been extensively evaluated in the context of NLP, a gap addressed by this work. We conduct a comparison between influence functions and common word-saliency methods on representative tasks. As suspected, we find that influence functions are particularly useful for natural language inference, a task in which \u2018saliency maps\u2019 may not have clear interpretation. Furthermore, we develop a new quantitative measure based on influence functions that can reveal artifacts in training data."
                },
                {
                    "title": "Improved Natural Language Generation via Loss Truncation",
                    "abstract": "Neural language models are usually trained to match the distributional properties of large-scale corpora by minimizing the log loss. While straightforward to optimize, this approach forces the model to reproduce all variations in the dataset, including noisy and invalid references (e.g., misannotations and hallucinated facts). Even a small fraction of noisy data can degrade the performance of log loss. As an alternative, prior work has shown that minimizing the distinguishability of generated samples is a principled and robust loss that can handle invalid references. However, distinguishability has not been used in practice due to challenges in optimization and estimation. We propose loss truncation: a simple and scalable procedure which adaptively removes high log loss examples as a way to optimize for distinguishability. Empirically, we demonstrate that loss truncation outperforms existing baselines on distinguishability on a summarization task. Furthermore, we show that samples generated by the loss truncation model have factual accuracy ratings that exceed those of baselines and match human references."
                },
                {
                    "title": "Data-dependent Gaussian Prior Objective for Language Generation",
                    "abstract": "For typical sequence prediction problems like language generation, maximum likelihood estimation (MLE) has been commonly adopted as it encourages the predicted sequence most consistent with the ground-truth sequence to have the highest probability of occurring. However, MLE focuses on a once-for-all matching between the predicted sequence and gold-standard consequently, treating all incorrect predictions as being equally incorrect. We call such a drawback {\\it negative diversity ignorance} in this paper. Treating all incorrect predictions as equal unfairly downplays the nuance of these sequences' detailed token-wise structure. To counteract this, we augment the MLE loss by introducing an extra KL divergence term which is derived from comparing a data-dependent Gaussian prior and the detailed training prediction. The proposed data-dependent Gaussian prior objective (D2GPo) is defined over a prior topological order of tokens, poles apart from the data-independent Gaussian prior (L2 regularization) commonly adopted for smoothing the training of MLE. Experimental results show that the proposed method can effectively make use of more detailed prior in the data and significantly improve the performance of typical language generation tasks, including supervised and unsupervised machine translation, text summarization, storytelling, and image caption."
                },
                {
                    "title": "Balancing Training for Multilingual Neural Machine Translation",
                    "abstract": "When training multilingual machine translation (MT) models that can translate to/from multiple languages, we are faced with imbalanced training sets: some languages have much more training data than others. Standard practice is to up-sample less resourced languages to increase representation, and the degree of up-sampling has a large effect on the overall performance. In this paper, we propose a method that instead automatically learns how to weight training data through a data scorer that is optimized to maximize performance on all test languages. Experiments on two sets of languages under both one-to-many and many-to-one MT settings show our method not only consistently outperforms heuristic baselines in terms of average performance, but also offers flexible control over the performance of which languages are optimized."
                },
                {
                    "title": "Simulated Multiple Reference Training Improves Low-Resource Machine Translation",
                    "abstract": "Many valid translations exist for a given sentence, and yet machine translation (MT) is trained with a single reference translation, exacerbating data sparsity in low-resource settings. We introduce a novel MT training method that approximates the full space of possible translations by: sampling a paraphrase of the reference sentence from a paraphraser and training the MT model to predict the paraphraser's distribution over possible tokens. With an English paraphraser, we demonstrate the effectiveness of our method in low-resource settings, with gains of 1.2 to 7 BLEU."
                },
                {
                    "title": "Characterizing Structural Regularities of Labeled Data in Overparameterized Models",
                    "abstract": "Human learners appreciate that observations usually form hierarchies of regularities and sub-regularities. For example, English verbs have irregular cases that must be memorized (e.g., go -> went) and regular cases that generalize well (e.g., kiss -> kissed, miss -> missed). Likewise, deep neural networks have the capacity to memorize rare or irregular forms but nonetheless generalize across instances that share common patterns or structures. We analyze how individual instances are treated by a model via a consistency score. The score is the expected accuracy of a particular architecture for a held-out instance on a training set of a given size sampled from the data distribution. We obtain empirical estimates of this score for individual instances in multiple data sets, and we show that the score identifies out-of-distribution and mislabeled examples at one end of the continuum and regular examples at the other end. We explore two categories of proxies to the consistency score: pairwise distance based proxy and the training statistics based proxies. We conclude with two applications using C-scores to help understand the dynamics of representation learning and filter out outliers, and discussions of other potential applications such as curriculum learning, and active data collection."
                },
                {
                    "title": "Estimating Training Data Influence by Tracking Gradient Descent",
                    "abstract": "We introduce a method called TrackIn that computes the influence of a training example on a prediction made by the model, by tracking how the loss on the test point changes during the training process whenever the training example of interest was utilized. We provide a scalable implementation of TrackIn via a combination of a few key ideas: (a) a first-order approximation to the exact computation, (b) using random projections to speed up the computation of the first-order approximation for large models, (c) using saved checkpoints of standard training procedures, and (d) cherry-picking layers of a deep neural network. An experimental evaluation shows that TrackIn is more effective in identifying mislabelled training examples than other related methods such as influence functions and representer points. We also discuss insights from applying the method on vision, regression and natural language tasks."
                },
                {
                    "title": "Gradient Surgery for Multi-Task Learning",
                    "abstract": "While deep learning and deep reinforcement learning (RL) systems have demonstrated impressive results in domains such as image classification, game playing, and robotic control, data efficiency remains a major challenge. Multi-task learning has emerged as a promising approach for sharing structure across multiple tasks to enable more efficient learning. However, the multi-task setting presents a number of optimization challenges, making it difficult to realize large efficiency gains compared to learning tasks independently. The reasons why multi-task learning is so challenging compared to single-task learning are not fully understood. In this work, we identify a set of three conditions of the multi-task optimization landscape that cause detrimental gradient interference, and develop a simple yet general approach for avoiding such interference between task gradients. We propose a form of gradient surgery that projects a task's gradient onto the normal plane of the gradient of any other task that has a conflicting gradient. On a series of challenging multi-task supervised and multi-task RL problems, this approach leads to substantial gains in efficiency and performance. Further, it is model-agnostic and can be combined with previously-proposed multi-task architectures for enhanced performance."
                },
                {
                    "title": "Optimizing Data Usage via Differentiable Rewards",
                    "abstract": "To acquire a new skill, humans learn better and faster if a tutor, based on their current knowledge level, informs them of how much attention they should pay to particular content or practice problems. Similarly, a machine learning model could potentially be trained better with a scorer that \"adapts\" to its current learning state and estimates the importance of each training data instance. Training such an adaptive scorer efficiently is a challenging problem; in order to precisely quantify the effect of a data instance at a given time during the training, it is typically necessary to first complete the entire training process. To efficiently optimize data usage, we propose a reinforcement learning approach called Differentiable Data Selection (DDS). In DDS, we formulate a scorer network as a learnable function of the training data, which can be efficiently updated along with the main model being trained. Specifically, DDS updates the scorer with an intuitive reward signal: it should up-weigh the data that has a similar gradient with a dev set upon which we would finally like to perform well. Without significant computing overhead, DDS delivers strong and consistent improvements over several strong baselines on two very different tasks of machine translation and image classification."
                },
                {
                    "title": "BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension",
                    "abstract": "We present BART, a denoising autoencoder for pretraining sequence-to-sequence models. BART is trained by (1) corrupting text with an arbitrary noising function, and (2) learning a model to reconstruct the original text. It uses a standard Tranformer-based neural machine translation architecture which, despite its simplicity, can be seen as generalizing BERT (due to the bidirectional encoder), GPT (with the left-to-right decoder), and other recent pretraining schemes. We evaluate a number of noising approaches, finding the best performance by both randomly shuffling the order of sentences and using a novel in-filling scheme, where spans of text are replaced with a single mask token. BART is particularly effective when fine tuned for text generation but also works well for comprehension tasks. It matches the performance of RoBERTa on GLUE and SQuAD, and achieves new state-of-the-art results on a range of abstractive dialogue, question answering, and summarization tasks, with gains of up to 3.5 ROUGE. BART also provides a 1.1 BLEU increase over a back-translation system for machine translation, with only target language pretraining. We also replicate other pretraining schemes within the BART framework, to understand their effect on end-task performance."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "Neural Text Generation with Unlikelihood Training",
                    "abstract": "Neural text generation is a key tool in natural language applications, but it is well known there are major problems at its core. In particular, standard likelihood training and decoding leads to dull and repetitive outputs. While some post-hoc fixes have been proposed, in particular top-$k$ and nucleus sampling, they do not address the fact that the token-level probabilities predicted by the model are poor. In this paper we show that the likelihood objective itself is at fault, resulting in a model that assigns too much probability to sequences containing repeats and frequent words, unlike those from the human training distribution. We propose a new objective, unlikelihood training, which forces unlikely generations to be assigned lower probability by the model. We show that both token and sequence level unlikelihood training give less repetitive, less dull text while maintaining perplexity, giving superior generations using standard greedy or beam search. According to human evaluations, our approach with standard beam search also outperforms the currently popular decoding methods of nucleus sampling or beam blocking, thus providing a strong alternative to existing techniques."
                },
                {
                    "title": "Curriculum Learning for Domain Adaptation in Neural Machine Translation",
                    "abstract": "We introduce a curriculum learning approach to adapt generic neural machine translation models to a specific domain. Samples are grouped by their similarities to the domain of interest and each group is fed to the training algorithm with a particular schedule. This approach is simple to implement on top of any neural framework or architecture, and consistently outperforms both unadapted and adapted baselines in experiments with two distinct domains and two language pairs."
                },
                {
                    "title": "On the Accuracy of Influence Functions for Measuring Group Effects",
                    "abstract": "Influence functions estimate the effect of removing a training point on a model without the need to retrain. They are based on a first-order Taylor approximation that is guaranteed to be accurate for sufficiently small changes to the model, and so are commonly used to study the effect of individual points in large datasets. However, we often want to study the effects of large groups of training points, e.g., to diagnose batch effects or apportion credit between different data sources. Removing such large groups can result in significant changes to the model. Are influence functions still accurate in this setting? In this paper, we find that across many different types of groups and for a range of real-world datasets, the predicted effect (using influence functions) of a group correlates surprisingly well with its actual effect, even if the absolute and relative errors are large. Our theoretical analysis shows that such strong correlation arises only under certain settings and need not hold in general, indicating that real-world datasets have particular properties that allow the influence approximation to be accurate."
                },
                {
                    "title": "fairseq: A Fast, Extensible Toolkit for Sequence Modeling",
                    "abstract": "fairseq is an open-source sequence modeling toolkit that allows researchers and developers to train custom models for translation, summarization, language modeling, and other text generation tasks. The toolkit is based on PyTorch and supports distributed training across multiple GPUs and machines. We also support fast mixed-precision training and inference on modern GPUs. A demo video can be found at https://www.youtube.com/watch?v=OtgDdWtHvto"
                },
                {
                    "title": "Competence-based Curriculum Learning for Neural Machine Translation",
                    "abstract": "Current state-of-the-art NMT systems use large neural networks that are not only slow to train, but also often require many heuristics and optimization tricks, such as specialized learning rate schedules and large batch sizes. This is undesirable as it requires extensive hyperparameter tuning. In this paper, we propose a curriculum learning framework for NMT that reduces training time, reduces the need for specialized heuristics or large batch sizes, and results in overall better performance. Our framework consists of a principled way of deciding which training samples are shown to the model at different times during training, based on the estimated difficulty of a sample and the current competence of the model. Filtering training samples in this manner prevents the model from getting stuck in bad local optima, making it converge faster and reach a better solution than the common approach of uniformly sampling training examples. Furthermore, the proposed method can be easily applied to existing NMT models by simply modifying their input data pipelines. We show that our framework can help improve the training time and the performance of both recurrent neural network models and Transformers, achieving up to a 70% decrease in training time, while at the same time obtaining accuracy improvements of up to 2.2 BLEU."
                },
                {
                    "title": "(Preprint)",
                    "abstract": "\n BACKGROUND\n Increasing numbers of children undergo ambulatory surgery each year and a significant proportion experiences substantial preoperative anxiety and postoperative pain. The management of perioperative anxiety and pain remains challenging in children and is inadequate, which negatively impacts physical, psychosocial, and economic outcomes. Existing non-pharmacological interventions are costly, time consuming, vary in availability, and lack benefits. Therefore, there is a need for an evidence-based, accessible, non-pharmacological intervention as an adjunct to existing pharmacological alternatives to reduce perioperative anxiety and pain in children undergoing ambulatory surgery. Technology-enabled interventions have been proposed as a method to address the unmet need in this setting. In particular, serious games for health (SGHs) hold unique potential to change health beliefs and behaviors in children.\n \n \n OBJECTIVE\n The objective of this research was to describe the rationale, scientific evidence, design aspects, and features of CliniPup, an SGH aimed at reducing perioperative anxiety and pain in children undergoing ambulatory surgery.\n \n \n METHODS\n The SERES framework for SGH development was used to create the SGH, CliniPup. In particular, a mixed-methods approach was applied that consisted of a structured literature review supplemented with ethnographic research, such as expert interviews and a time-motion exercise. The resulting scientific evidence base was leveraged to ensure that the resulting SGH was relevant, realistic, and theory-driven. A participatory design approach was applied wherein clinical experts qualitatively reviewed several versions of the SGH and an iterative creative process was used to integrate the applicable feedback.\n \n \n RESULTS\n CliniPup, an SGH, was developed to incorporate (1) scientific evidence base from a structured literature review, (2) realistic content collected during ethnographic research such as expert interviews, (3) explicit pedagogical objectives from scientific literature, and (4) game mechanics and user interface design that address key aspects of the evidence.\n \n \n CONCLUSIONS\n This report details the systematic development of CliniPup, an SGH aimed at reducing perioperative anxiety and pain in children undergoing ambulatory surgery. Clinical experts validated CliniPup\u2019s underlying scientific evidence base and design foundations, suggesting that it was well designed for preliminary evaluation in the target population. An evaluation plan is proposed and briefly described.\n"
                },
                {
                    "title": "An Empirical Study of Example Forgetting during Deep Neural Network Learning",
                    "abstract": "Inspired by the phenomenon of catastrophic forgetting, we investigate the learning dynamics of neural networks as they train on single classification tasks. Our goal is to understand whether a related phenomenon occurs when data does not undergo a clear distributional shift. We define a \u201cforgetting event\u201d to have occurred when an individual training example transitions from being classified correctly to incorrectly over the course of learning. Across several benchmark data sets, we find that: (i) certain examples are forgotten with high frequency, and some not at all; (ii) a data set\u2019s (un)forgettable examples generalize across neural architectures; and (iii) based on forgetting dynamics, a significant fraction of examples can be omitted from the training data set while still maintaining state-of-the-art generalization performance."
                },
                {
                    "title": "Hierarchical Neural Story Generation",
                    "abstract": "We explore story generation: creative systems that can build coherent and fluent passages of text about a topic. We collect a large dataset of 300K human-written stories paired with writing prompts from an online forum. Our dataset enables hierarchical story generation, where the model first generates a premise, and then transforms it into a passage of text. We gain further improvements with a novel form of model fusion that improves the relevance of the story to the prompt, and adding a new gated multi-scale self-attention mechanism to model long-range context. Experiments show large improvements over strong baselines on both automated and human evaluations. Human judges prefer stories generated by our approach to those from a strong non-hierarchical model by a factor of two to one."
                },
                {
                    "title": "On the Impact of Various Types of Noise on Neural Machine Translation",
                    "abstract": "We examine how various types of noise in the parallel training data impact the quality of neural machine translation systems. We create five types of artificial noise and analyze how they degrade performance in neural and statistical machine translation. We find that neural models are generally more harmed by noise than statistical models. For one especially egregious type of noise they learn to just copy the input sentence."
                },
                {
                    "title": "A Call for Clarity in Reporting BLEU Scores",
                    "abstract": "The field of machine translation faces an under-recognized problem because of inconsistency in the reporting of scores from its dominant metric. Although people refer to \u201cthe\u201d BLEU score, BLEU is in fact a parameterized metric whose values can vary wildly with changes to these parameters. These parameters are often not reported or are hard to find, and consequently, BLEU scores between papers cannot be directly compared. I quantify this variation, finding differences as high as 1.8 between commonly used configurations. The main culprit is different tokenization and normalization schemes applied to the reference. Pointing to the success of the parsing community, I suggest machine translation researchers settle upon the BLEU scheme used by the annual Conference on Machine Translation (WMT), which does not allow for user-supplied reference processing, and provide a new tool, SACREBLEU, to facilitate this."
                },
                {
                    "title": "Analyzing Uncertainty in Neural Machine Translation",
                    "abstract": "Machine translation is a popular test bed for research in neural sequence-to-sequence models but despite much recent research, there is still a lack of understanding of these models. Practitioners report performance degradation with large beams, the under-estimation of rare words and a lack of diversity in the final translations. Our study relates some of these issues to the inherent uncertainty of the task, due to the existence of multiple valid translations for a single source sentence, and to the extrinsic uncertainty caused by noisy training data. We propose tools and metrics to assess how uncertainty in the data is captured by the model distribution and how it affects search strategies that generate translations. Our results show that search works remarkably well but that models tend to spread too much probability mass over the hypothesis space. Next, we propose tools to assess model calibration and show how to easily fix some shortcomings of current models. As part of this study, we release multiple human reference translations for two popular benchmarks."
                },
                {
                    "title": "Findings of the 2017 Conference on Machine Translation (WMT17)",
                    "abstract": "This paper presents the results of the WMT17 shared tasks, which included \nthree machine translation (MT) tasks (news, biomedical, and multimodal), two evaluation tasks (metrics and run-time estimation of MT quality), an automatic post-editing task, a neural MT training task, and a bandit learning task."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension",
                    "abstract": "We present TriviaQA, a challenging reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents, six per question on average, that provide high quality distant supervision for answering the questions. We show that, in comparison to other recently introduced large-scale datasets, TriviaQA (1) has relatively complex, compositional questions, (2) has considerable syntactic and lexical variability between questions and corresponding answer-evidence sentences, and (3) requires more cross sentence reasoning to find answers. We also present two baseline algorithms: a feature-based classifier and a state-of-the-art neural network, that performs well on SQuAD reading comprehension. Neither approach comes close to human performance (23% and 40% vs. 80%), suggesting that TriviaQA is a challenging testbed that is worth significant future study."
                },
                {
                    "title": "Get To The Point: Summarization with Pointer-Generator Networks",
                    "abstract": "Neural sequence-to-sequence models have provided a viable new approach for abstractive text summarization (meaning they are not restricted to simply selecting and rearranging passages from the original text). However, these models have two shortcomings: they are liable to reproduce factual details inaccurately, and they tend to repeat themselves. In this work we propose a novel architecture that augments the standard sequence-to-sequence attentional model in two orthogonal ways. First, we use a hybrid pointer-generator network that can copy words from the source text via pointing, which aids accurate reproduction of information, while retaining the ability to produce novel words through the generator. Second, we use coverage to keep track of what has been summarized, which discourages repetition. We apply our model to the CNN / Daily Mail summarization task, outperforming the current abstractive state-of-the-art by at least 2 ROUGE points."
                },
                {
                    "title": "Understanding Black-box Predictions via Influence Functions",
                    "abstract": "How can we explain the predictions of a black-box model? In this paper, we use influence functions \u2014 a classic technique from robust statistics \u2014 to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. We show that even on non-convex and non-differentiable models where the theory breaks down, approximations to influence functions can still provide valuable information. On linear models and convolutional neural networks, we demonstrate that influence functions are useful for multiple purposes: understanding model behavior, debugging models, detecting dataset errors, and even creating visually-indistinguishable training-set attacks."
                },
                {
                    "title": "Pointer Sentinel Mixture Models",
                    "abstract": "Recent neural network sequence models with softmax classifiers have achieved their best language modeling performance only with very large hidden states and large vocabularies. Even then they struggle to predict rare or unseen words even if the context makes the prediction unambiguous. We introduce the pointer sentinel mixture architecture for neural sequence models which has the ability to either reproduce a word from the recent context or produce a word from a standard softmax classifier. Our pointer sentinel-LSTM model achieves state of the art language modeling performance on the Penn Treebank (70.9 perplexity) while using far fewer parameters than a standard softmax LSTM. In order to evaluate how well language models can exploit longer contexts and deal with more realistic vocabularies and larger corpora we also introduce the freely available WikiText corpus."
                },
                {
                    "title": "A Neural Attention Model for Abstractive Sentence Summarization",
                    "abstract": "Summarization based on text extraction is inherently limited, but generation-style abstractive methods have proven challenging to build. In this work, we propose a fully data-driven approach to abstractive sentence summarization. Our method utilizes a local attention-based model that generates each word of the summary conditioned on the input sentence. While the model is structurally simple, it can easily be trained end-to-end and scales to a large amount of training data. The model shows significant performance gains on the DUC-2004 shared task compared with several strong baselines."
                },
                {
                    "title": "Teaching Machines to Read and Comprehend",
                    "abstract": "Teaching machines to read natural language documents remains an elusive challenge. Machine reading systems can be tested on their ability to answer questions posed on the contents of documents that they have seen, but until now large scale training and test datasets have been missing for this type of evaluation. In this work we define a new methodology that resolves this bottleneck and provides large scale supervised reading comprehension data. This allows us to develop a class of attention based deep neural networks that learn to read real documents and answer complex questions with minimal prior knowledge of language structure."
                },
                {
                    "title": "Recurrent Continuous Translation Models",
                    "abstract": "We introduce a class of probabilistic continuous translation models called Recurrent Continuous Translation Models that are purely based on continuous representations for words, phrases and sentences and do not rely on alignments or phrasal translation units. The models have a generation and a conditioning aspect. The generation of the translation is modelled with a target Recurrent Language Model, whereas the conditioning on the source sentence is modelled with a Convolutional Sentence Model. Through various experiments, we show first that our models obtain a perplexity with respect to gold translations that is > 43% lower than that of stateof-the-art alignment-based translation models. Secondly, we show that they are remarkably sensitive to the word order, syntax, and meaning of the source sentence despite lacking alignments. Finally we show that they match a state-of-the-art system when rescoring n-best lists of translations."
                },
                {
                    "title": "Intelligent Selection of Language Model Training Data",
                    "abstract": "We address the problem of selecting non-domain-specific language model training data to build auxiliary language models for use in tasks such as machine translation. Our approach is based on comparing the cross-entropy, according to domain-specific and non-domain-specifc language models, for each sentence of the text source used to produce the latter language model. We show that this produces better language models, trained on less data, than both random data selection and two other previously proposed methods."
                },
                {
                    "title": "The Elements of Statistical Learning",
                    "abstract": "Chapter 11 includes more case studies in other areas, ranging from manufacturing to marketing research. Chapter 12 concludes the book with some commentary about the scienti\u008e c contributions of MTS. The Taguchi method for design of experiment has generated considerable controversy in the statistical community over the past few decades. The MTS/MTGS method seems to lead another source of discussions on the methodology it advocates (Montgomery 2003). As pointed out by Woodall et al. (2003), the MTS/MTGS methods are considered ad hoc in the sense that they have not been developed using any underlying statistical theory. Because the \u201cnormal\u201d and \u201cabnormal\u201d groups form the basis of the theory, some sampling restrictions are fundamental to the applications. First, it is essential that the \u201cnormal\u201d sample be uniform, unbiased, and/or complete so that a reliable measurement scale is obtained. Second, the selection of \u201cabnormal\u201d samples is crucial to the success of dimensionality reduction when OAs are used. For example, if each abnormal item is really unique in the medical example, then it is unclear how the statistical distance MD can be guaranteed to give a consistent diagnosis measure of severity on a continuous scale when the larger-the-better type S/N ratio is used. Multivariate diagnosis is not new to Technometrics readers and is now becoming increasingly more popular in statistical analysis and data mining for knowledge discovery. As a promising alternative that assumes no underlying data model, The Mahalanobis\u2013Taguchi Strategy does not provide suf\u008e cient evidence of gains achieved by using the proposed method over existing tools. Readers may be very interested in a detailed comparison with other diagnostic tools, such as logistic regression and tree-based methods. Overall, although the idea of MTS/MTGS is intriguing, this book would be more valuable had it been written in a rigorous fashion as a technical reference. There is some lack of precision even in several mathematical notations. Perhaps a follow-up with additional theoretical justi\u008e cation and careful case studies would answer some of the lingering questions."
                },
                {
                    "title": "Bleu: a Method for Automatic Evaluation of Machine Translation",
                    "abstract": "Human evaluations of machine translation are extensive but expensive. Human evaluations can take months to finish and involve human labor that can not be reused. We propose a method of automatic machine translation evaluation that is quick, inexpensive, and language-independent, that correlates highly with human evaluation, and that has little marginal cost per run. We present this method as an automated understudy to skilled human judges which substitutes for them when there is need for quick or frequent evaluations."
                },
                {
                    "title": "On Information and Sufficiency",
                    "abstract": "The information deviation between any two finite measures cannot be increased by any statistical operations (Markov morphisms). It is invarient if and only if the morphism is sufficient for these two measures"
                },
                {
                    "title": "Findings of the 2022 Conference on Machine Translation (WMT22)",
                    "abstract": "This paper presents the results of the General Machine Translation Task organised as part of the Conference on Machine Translation (WMT) 2022. In the general MT task, participants were asked to build machine translation systems for any of 11 language pairs, to be evaluated on test sets consisting of four different domains. We evaluate system outputs with human annotators using two different techniques: reference-based direct assessment and (DA) and a combination of DA and scalar quality metric (DA+SQM)."
                },
                {
                    "title": "Text Generation by Learning from Demonstrations",
                    "abstract": "Current approaches to text generation largely rely on autoregressive models and maximum likelihood estimation. This paradigm leads to (i) diverse but low-quality samples due to mismatched learning objective and evaluation metric (likelihood vs. quality) and (ii) exposure bias due to mismatched history distributions (gold vs. model-generated). To alleviate these problems, we frame text generation as an offline reinforcement learning (RL) problem with expert demonstrations (i.e., the reference), where the goal is to maximize quality given model-generated histories. We propose GOLD (generation by off-policy learning from demonstrations): an easy-to-optimize algorithm that learns from the demonstrations by importance weighting. Intuitively, GOLD upweights confident tokens and downweights unconfident ones in the reference during training, avoiding optimization issues faced by prior RL approaches that rely on online data collection. According to both automatic and human evaluation, models trained by GOLD outperform those trained by MLE and policy gradient on summarization, question generation, and machine translation. Further, our models are less sensitive to decoding algorithms and alleviate exposure bias."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "On Measures of Entropy and Information",
                    "abstract": "This book is an updated version of the information theory classic, first published in 1990. About one-third of the book is devoted to Shannon source and channel coding theorems; the remainder addresses sources, channels, and codes and on information and distortion measures and their properties. New in this edition:Expanded treatment of stationary or sliding-block codes and their relations to traditional block codesExpanded discussion of results from ergodic theory relevant to information theoryExpanded treatment of B-processes -- processes formed by stationary coding memoryless sourcesNew material on trading off information and distortion, including the Marton inequalityNew material on the properties of optimal and asymptotically optimal source codesNew material on the relationships of source coding and rate-constrained simulation or modeling of random processesSignificant material not covered in other information theory texts includes stationary/sliding-block codes, a geometric view of information theory provided by process distance measures, and general Shannon coding theorems for asymptotic mean stationary sources, which may be neither ergodic nor stationary, and d-bar continuous channels."
                },
                {
                    "title": "Edinburgh Research Explorer Improving Massively Multilingual Neural Machine Translation and Zero-Shot Translation",
                    "abstract": "Massively multilingual models for neural machine translation (NMT) are theoretically at-tractive, but often underperform bilingual models and deliver poor zero-shot translations. In this paper, we explore ways to improve them. We argue that multilingual NMT re-quires stronger modeling capacity to support language pairs with varying typological char-acteristics, and overcome this bottleneck via language-speci\ufb01c components and deepening NMT architectures. We identify the off-target translation issue (i.e. translating into a wrong target language) as the major source of the inferior zero-shot performance, and propose random online backtranslation to enforce the translation of unseen training language pairs. Experiments on OPUS-100 (a novel multilingual dataset with 100 languages) show that our approach substantially narrows the performance gap with bilingual models in both one-to-many and many-to-many settings, and improves zero-shot performance by \u223c 10 BLEU, approaching conventional pivot-based methods. 1"
                }
            ],
            "categories": [
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the robustness of neural text generation models to errors in training data by effectively estimating data quality during training?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for enhancing the reliability of text generation models, which are increasingly used in critical applications such as machine translation, summarization, and conversational agents. By addressing the vulnerabilities of these models to noise in training data, we can advance the state of the art in natural language processing, leading to more accurate and trustworthy systems. This research could pave the way for future studies focused on data quality assessment and model training methodologies, ultimately fostering the development of more resilient AI systems capable of handling real-world data imperfections.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in accurately estimating data quality while considering the distribution of non-target tokens, as naive approaches may either indiscriminately remove or down-weight tokens based solely on their predicted probabilities. This can lead to the loss of valuable training information, especially in cases where the model is still learning or when the context is ambiguous. Additionally, the complexity of modeling high-entropy contexts and distinguishing between genuine errors and normal variability in data adds to the difficulty of developing a robust solution.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on modifying the maximum-likelihood estimation (MLE) objective without adequately considering the distribution of non-target tokens. This oversight has resulted in methods that may inadvertently harm model training by removing or down-weighting tokens that are not necessarily erroneous. Barriers such as a lack of comprehensive frameworks for data quality estimation and the challenge of integrating these considerations into existing training paradigms have prevented effective solutions. Our approach, Error Norm Truncation (ENT), improves upon prior work by incorporating the predicted distribution of non-target tokens into the data quality assessment process, allowing for more nuanced decision-making during training.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Error Norm Truncation (ENT), involves modifying the training objective to incorporate an error norm that evaluates the distribution of non-target tokens alongside the predicted probabilities of the ground truth token. We will utilize a diverse dataset of text generation tasks and employ metrics such as BLEU and ROUGE to assess model performance. The expected outcomes include improved robustness of the text generation models to noise in training data, leading to higher quality outputs and enhanced"
            }
        },
        "author_data": {
            "713fd377-a25a-4e3e-ad9c-8c255bfcb6b6": {
                "pk": "713fd377-a25a-4e3e-ad9c-8c255bfcb6b6",
                "name": "Tianjian Li",
                "collaborators": [
                    "Daniel Khashabi",
                    "Haoran Xu",
                    "Weiting Tan",
                    "Kenton Murray",
                    "Yining Lu",
                    "Dixuan Wang",
                    "Dongwei Jiang",
                    "Jingyu (Jack) Zhang",
                    "Marc Marone",
                    "Benjamin Van Durme"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Creativity in AI",
                    "Model Verifiability"
                ],
                "publications": [
                    {
                        "title": "Upsample or Upweight? Balanced Training on Heavily Imbalanced Datasets",
                        "abstract": "Data availability across domains often follows a long-tail distribution: a few domains have abundant data, while most face dat . a scarcity. This imbalance poses challenges in training language models uniformly across all domains. In our study, we focus on multilingual settings, where data sizes vary significantly between high- and low-resource languages. Common strategies to address this include upsampling low-resource languages (Temperature Sampling) or upweighting their loss (Scalarization). Although often considered equivalent, this assumption has not been proven, which motivates our study. Through both theoretical and empirical analysis, we identify the conditions under which these approaches are equivalent and when they diverge. Specifically, we demonstrate that these two methods are equivalent under full gradient descent, but this equivalence breaks down with stochastic gradient descent. Empirically, we observe that Temperature Sampling converges more quickly but is prone to overfitting. We argue that this faster convergence is likely due to the lower variance in gradient estimations, as shown theoretically. Based on these insights, we propose Cooldown, a strategy that reduces sampling temperature during training, accelerating convergence without overfitting to low-resource languages. Our method is competitive with existing data re-weighting and offers computational efficiency."
                    },
                    {
                        "title": "Benchmarking Language Model Creativity: A Case Study on Code Generation",
                        "abstract": "As LLMs become increasingly prevalent, it is interesting to consider how ``creative'' these models can be. From cognitive science, creativity consists of at least two key characteristics: \\emph{convergent} thinking (purposefulness to achieve a given goal) and \\emph{divergent} thinking (adaptability to new environments or constraints) \\citep{runco2003critical}. In this work, we introduce a framework for quantifying LLM creativity that incorporates the two characteristics. This is achieved by (1) Denial Prompting pushes LLMs to come up with more creative solutions to a given problem by incrementally imposing new constraints on the previous solution, compelling LLMs to adopt new strategies, and (2) defining and computing the NeoGauge metric which examines both convergent and divergent thinking in the generated creative responses by LLMs. We apply the proposed framework on Codeforces problems, a natural data source for collecting human coding solutions. We quantify NeoGauge for various proprietary and open-source models and find that even the most creative model, GPT-4, still falls short of demonstrating human-like creativity. We also experiment with advanced reasoning strategies (MCTS, self-correction, etc.) and observe no significant improvement in creativity. As a by-product of our analysis, we release NeoCoder dataset for reproducing our results on future models."
                    },
                    {
                        "title": "Verifiable by Design: Aligning Language Models to Quote from Pre-Training Data",
                        "abstract": "To trust the fluent generations of large language models (LLMs), humans must be able to verify their correctness against trusted, external sources. Recent efforts, such as providing citations via retrieved documents or post-hoc provenance, enhance verifiability but still provide no guarantees on their correctness. To address these limitations, we tackle the verifiability goal with a different philosophy: trivializing the verification process by developing models that quote verbatim statements from trusted sources in pre-training data. We propose Quote-Tuning, and demonstrate it is feasible to align LLMs to provide quoted statements from data memorized during pre-training. The core of Quote-Tuning is a fast membership inference function (Marone and Van Durme, 2023) that efficiently verifies text against a trusted corpus. We leverage this tool to design a reward function to quantify quotes in model responses, which is then used to create a dataset for preference learning. Experimental results show that Quote-Tuning significantly increases verbatim quotes from high-quality pre-training documents by 55% to 130% relative to un-tuned models while maintaining response quality. Quote-Tuning also generalizes quoting to out-of-domain data, is applicable in different tasks, and provides additional benefits to truthfulness. Our method not only serves as a hassle-free method to increase quoting but also opens up avenues for improving LLM trustworthiness through better verifiability."
                    }
                ]
            },
            "aa93c430-805d-4a0f-a9c5-db8cacab7a79": {
                "pk": "aa93c430-805d-4a0f-a9c5-db8cacab7a79",
                "name": "Haoran Xu",
                "collaborators": [
                    "Kenton Murray",
                    "Philipp Koehn",
                    "Benjamin Van Durme",
                    "Weiting Tan",
                    "Yunmo Chen",
                    "Amr Sharaf",
                    "Lingfeng Shen",
                    "Young Jin Kim",
                    "Jean Maillard",
                    "Vedanuj Goswami"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Translation",
                    "Multilingual Models",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "JHU IWSLT 2024 Dialectal and Low-resource System Description",
                        "abstract": "Johns Hopkins University (JHU) submitted systems for all eight language pairs in the 2024 Low-Resource Language Track. The main effort of this work revolves around fine-tuning large and publicly available models in three proposed systems: i) end-to-end speech translation (ST) fine-tuning of Seamless4MT v2; ii) ST fine-tuning of Whisper; iii) a cascaded system involving automatic speech recognition with fine-tuned Whisper and machine translation with NLLB. On top of systems above, we conduct a comparative analysis on different training paradigms, such as intra-distillation for NLLB as well as joint training and curriculum learning for SeamlessM4T v2. Our results show that the best-performing approach differs by language pairs, but that i) fine-tuned SeamlessM4T v2 tends to perform best for source languages on which it was pre-trained, ii) multi-task training helps Whisper fine-tuning, iii) cascaded systems with Whisper and NLLB tend to outperform Whisper alone, and iv) intra-distillation helps NLLB fine-tuning."
                    },
                    {
                        "title": "X-ALMA: Plug&Play Modules and Adaptive Rejection for Quality Translation at Scale",
                        "abstract": "Large language models (LLMs) have achieved remarkable success across various NLP tasks, yet their focus has predominantly been on English due to English-centric pre-training and limited multilingual data. While some multilingual LLMs claim to support for hundreds of languages, models often fail to provide high-quality response for mid- and low-resource languages, leading to imbalanced performance heavily skewed in favor of high-resource languages like English and Chinese. In this paper, we prioritize quality over scaling number of languages, with a focus on multilingual machine translation task, and introduce X-ALMA, a model designed with a commitment to ensuring top-tier performance across 50 diverse languages, regardless of their resource levels. X-ALMA surpasses state-of-the-art open-source multilingual LLMs, such as Aya-101 and Aya-23, in every single translation direction on the FLORES and WMT'23 test datasets according to COMET-22. This is achieved by plug-and-play language-specific module architecture to prevent language conflicts during training and a carefully designed training regimen with novel optimization methods to maximize the translation performance. At the final stage of training regimen, our proposed Adaptive Rejection Preference Optimization (ARPO) surpasses existing preference optimization methods in translation tasks."
                    },
                    {
                        "title": "Contrastive Preference Optimization: Pushing the Boundaries of LLM Performance in Machine Translation",
                        "abstract": "Moderate-sized large language models (LLMs) -- those with 7B or 13B parameters -- exhibit promising machine translation (MT) performance. However, even the top-performing 13B LLM-based translation models, like ALMA, does not match the performance of state-of-the-art conventional encoder-decoder translation models or larger-scale LLMs such as GPT-4. In this study, we bridge this performance gap. We first assess the shortcomings of supervised fine-tuning for LLMs in the MT task, emphasizing the quality issues present in the reference data, despite being human-generated. Then, in contrast to SFT which mimics reference translations, we introduce Contrastive Preference Optimization (CPO), a novel approach that trains models to avoid generating adequate but not perfect translations. Applying CPO to ALMA models with only 22K parallel sentences and 12M parameters yields significant improvements. The resulting model, called ALMA-R, can match or exceed the performance of the WMT competition winners and GPT-4 on WMT'21, WMT'22 and WMT'23 test datasets."
                    },
                    {
                        "title": "Towards Being Parameter-Efficient: A Stratified Sparsely Activated Transformer with Dynamic Capacity",
                        "abstract": "Mixture-of-experts (MoE) models that employ sparse activation have demonstrated effectiveness in significantly increasing the number of parameters while maintaining low computational requirements per token. However, recent studies have established that MoE models are inherently parameter-inefficient as the improvement in performance diminishes with an increasing number of experts. We hypothesize this parameter inefficiency is a result of all experts having equal capacity, which may not adequately meet the varying complexity requirements of different tokens or tasks. In light of this, we propose Stratified Mixture of Experts (SMoE) models, which feature a stratified structure and can assign dynamic capacity to different tokens. We demonstrate the effectiveness of SMoE on three multilingual machine translation benchmarks, containing 4, 15, and 94 language pairs, respectively. We show that SMoE outperforms multiple state-of-the-art MoE models with the same or fewer parameters."
                    },
                    {
                        "title": "A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Models",
                        "abstract": "Generative Large Language Models (LLMs) have achieved remarkable advancements in various NLP tasks. However, these advances have not been reflected in the translation task, especially those with moderate model sizes (i.e., 7B or 13B parameters), which still lag behind conventional supervised encoder-decoder translation models. Previous studies have attempted to improve the translation capabilities of these moderate LLMs, but their gains have been limited. In this study, we propose a novel fine-tuning approach for LLMs that is specifically designed for the translation task, eliminating the need for the abundant parallel data that traditional translation models usually depend on. Our approach consists of two fine-tuning stages: initial fine-tuning on monolingual data followed by subsequent fine-tuning on a small set of high-quality parallel data. We introduce the LLM developed through this strategy as Advanced Language Model-based trAnslator (ALMA). Based on LLaMA-2 as our underlying model, our results show that the model can achieve an average improvement of more than 12 BLEU and 12 COMET over its zero-shot performance across 10 translation directions from the WMT'21 (2 directions) and WMT'22 (8 directions) test datasets. The performance is significantly better than all prior work and even superior to the NLLB-54B model and GPT-3.5-text-davinci-003, with only 7B or 13B parameters. This method establishes the foundation for a novel training paradigm in machine translation."
                    },
                    {
                        "title": "Language-Aware Multilingual Machine Translation with Self-Supervised Learning",
                        "abstract": "Multilingual machine translation (MMT) benefits from cross-lingual transfer but is a challenging multitask optimization problem. This is partly because there is no clear framework to systematically learn language-specific parameters. Self-supervised learning (SSL) approaches that leverage large quantities of monolingual data (where parallel data is unavailable) have shown promise by improving translation performance as complementary tasks to the MMT task. However, jointly optimizing SSL and MMT tasks is even more challenging. In this work, we first investigate how to utilize **intra-distillation** to learn more *language-specific* parameters and then show the importance of these language-specific parameters. Next, we propose a novel but simple SSL task, **concurrent denoising**, that co-trains with the MMT task by concurrently denoising monolingual data on both the encoder and decoder. Finally, we apply **intra-distillation** to this co-training approach. Combining these two approaches significantly improves MMT performance, outperforming three state-of-the-art SSL methods by a large margin, e.g., 11.3{% and 3.7{% improvement on an 8-language and a 15-language benchmark compared with MASS, respectively."
                    },
                    {
                        "title": "Deegen: A Meta-compiler Approach for High Performance VMs at Low Engineering Cost (Invited Talk)",
                        "abstract": "Building a high-performance VM for a dynamic language has traditionally required a huge amount of time, money, and expertise. To reduce the high engineering cost, we present Deegen, a meta-compiler that generates a high-performance VM automatically from a semantical description of the bytecodes. Currently, Deegen is capable of automatically generating an optimized interpreter and baseline JIT compiler. This allows the user get a high-performance VM for their own language at an engineering cost similar to writing an interpreter. To demonstrate Deegen's capability in the real world, we implemented LuaJIT Remake (LJR), a standard-compliant VM for Lua 5.1. Across a variety of benchmarks, we demonstrated that LJR's interpreter significantly outperforms LuaJIT's interpreter, and LJR's baseline JIT generates high-quality code with a negligible compilation cost."
                    },
                    {
                        "title": "Narrowing the Gap between Zero- and Few-shot Machine Translation by Matching Styles",
                        "abstract": "Large language models trained primarily in a monolingual setting have demonstrated their ability to generalize to machine translation using zero- and few-shot examples with in-context learning. However, even though zero-shot translations are relatively good, there remains a discernible gap comparing their performance with the few-shot setting. In this paper, we investigate the factors contributing to this gap and find that this gap can largely be closed (for about 70%) by matching the writing styles of the target corpus. Additionally, we explore potential approaches to enhance zero-shot baselines without the need for parallel demonstration examples, providing valuable insights into how these methods contribute to improving translation metrics."
                    },
                    {
                        "title": "Condensing Multilingual Knowledge with Lightweight Language-Specific Modules",
                        "abstract": "Incorporating language-specific (LS) modules is a proven method to boost performance in multilingual machine translation. This approach bears similarity to Mixture-of-Experts (MoE) because it does not inflate FLOPs. However, the scalability of this approach to hundreds of languages (experts) tends to be unmanageable due to the prohibitive number of parameters introduced by full-rank matrices in fully-connected layers. In this work, we introduce the Language-Specific Matrix Synthesis (LMS) method. This approach constructs LS modules by generating low-rank matrices from two significantly smaller matrices to approximate the full-rank matrix. Furthermore, we condense multilingual knowledge from multiple LS modules into a single shared module with the Fuse Distillation (FD) technique to improve the efficiency of inference and model serialization. We show that our LMS method significantly outperforms previous LS methods and MoE methods with the same amount of extra parameters, e.g., 1.73 BLEU points over the Switch Transformer on many-to-many multilingual machine translation. Importantly, LMS is able to have comparable translation performance with much fewer parameters."
                    },
                    {
                        "title": "SaFormer: A Conditional Sequence Modeling Approach to Offline Safe Reinforcement Learning",
                        "abstract": "Offline safe RL is of great practical relevance for deploying agents in real-world applications. However, acquiring constraint-satisfying policies from the fixed dataset is non-trivial for conventional approaches. Even worse, the learned constraints are stationary and may become invalid when the online safety requirement changes. In this paper, we present a novel offline safe RL approach referred to as SaFormer, which tackles the above issues via conditional sequence modeling. In contrast to existing sequence models, we propose cost-related tokens to restrict the action space and a posterior safety verification to enforce the constraint explicitly. Specifically, SaFormer performs a two-stage auto-regression conditioned by the maximum remaining cost to generate feasible candidates. It then filters out unsafe attempts and executes the optimal action with the highest expected return. Extensive experiments demonstrate the efficacy of SaFormer featuring (1) competitive returns with tightened constraint satisfaction; (2) adaptability to the in-range cost values of the offline data without retraining; (3) generalizability for constraints beyond the current dataset."
                    },
                    {
                        "title": "The Importance of Being Parameters: An Intra-Distillation Method for Serious Gains",
                        "abstract": "Recent model pruning methods have demonstrated the ability to remove redundant parameters without sacrificing model performance. Common methods remove redundant parameters according to the parameter sensitivity, a gradient-based measure reflecting the contribution of the parameters. In this paper, however, we argue that redundant parameters can be trained to make beneficial contributions. We first highlight the large sensitivity (contribution) gap among high-sensitivity and low-sensitivity parameters and show that the model generalization performance can be significantly improved after balancing the contribution of all parameters. Our goal is to balance the sensitivity of all parameters and encourage all of them to contribute equally. We propose a general task-agnostic method, namely intra-distillation, appended to the regular training loss to balance parameter sensitivity. Moreover, we also design a novel adaptive learning method to control the strength of intra-distillation loss for faster convergence. Our experiments show the strong effectiveness of our methods on machine translation, natural language understanding, and zero-shot cross-lingual transfer across up to 48 languages, e.g., a gain of 3.54 BLEU on average across 8 language pairs from the IWSLT\u201914 dataset."
                    },
                    {
                        "title": "Por Qu\u00e9 N\u00e3o Utiliser Alla Spr\u00e5k? Mixed Training with Gradient Optimization in Few-Shot Cross-Lingual Transfer",
                        "abstract": "The current state-of-the-art for few-shot cross-lingual transfer learning first trains on abundant labeled data in the source language and then fine-tunes with a few examples on the target language, termed target-adapting. Though this has been demonstrated to work on a variety of tasks, in this paper we show some deficiencies of this approach and propose a one-step mixed training method that trains on both source and target data with \\textit{stochastic gradient surgery}, a novel gradient-level optimization. Unlike the previous studies that focus on one language at a time when target-adapting, we use one model to handle all target languages simultaneously to avoid excessively language-specific models. Moreover, we discuss the unreality of utilizing large target development sets for model selection in previous literature. We further show that our method is both development-free for target languages, and is also able to escape from overfitting issues. We conduct a large-scale experiment on 4 diverse NLP tasks across up to 48 languages. Our proposed method achieves state-of-the-art performance on all tasks and outperforms target-adapting by a large margin, especially for languages that are linguistically distant from the source language, e.g., 7.36% F1 absolute gain on average for the NER task, up to 17.60% on Punjabi."
                    },
                    {
                        "title": "BERT, mBERT, or BiBERT? A Study on Contextualized Embeddings for Neural Machine Translation",
                        "abstract": "The success of bidirectional encoders using masked language models, such as BERT, on numerous natural language processing tasks has prompted researchers to attempt to incorporate these pre-trained models into neural machine translation (NMT) systems. However, proposed methods for incorporating pre-trained models are non-trivial and mainly focus on BERT, which lacks a comparison of the impact that other pre-trained models may have on translation performance. In this paper, we demonstrate that simply using the output (contextualized embeddings) of a tailored and suitable bilingual pre-trained language model (dubbed BiBERT) as the input of the NMT encoder achieves state-of-the-art translation performance. Moreover, we also propose a stochastic layer selection approach and a concept of a dual-directional translation model to ensure the sufficient utilization of contextualized embeddings. In the case of without using back translation, our best models achieve BLEU scores of 30.45 for En\u2192De and 38.61 for De\u2192En on the IWSLT\u201914 dataset, and 31.26 for En\u2192De and 34.94 for De\u2192En on the WMT\u201914 dataset, which exceeds all published numbers."
                    },
                    {
                        "title": "Zero-Shot Cross-Lingual Dependency Parsing through Contextual Embedding Transformation",
                        "abstract": "Linear embedding transformation has been shown to be effective for zero-shot cross-lingual transfer tasks and achieve surprisingly promising results. However, cross-lingual embedding space mapping is usually studied in static word-level embeddings, where a space transformation is derived by aligning representations of translation pairs that are referred from dictionaries. We move further from this line and investigate a contextual embedding alignment approach which is sense-level and dictionary-free. To enhance the quality of the mapping, we also provide a deep view of properties of contextual embeddings, i.e., the anisotropy problem and its solution. Experiments on zero-shot dependency parsing through the concept-shared space built by our embedding transformation substantially outperform state-of-the-art methods using multilingual embeddings."
                    },
                    {
                        "title": "VAE based Text Style Transfer with Pivot Words Enhancement Learning",
                        "abstract": "Text Style Transfer (TST) aims to alter the underlying style of the source text to another specific style while keeping the same content. Due to the scarcity of high-quality parallel training data, unsupervised learning has become a trending direction for TST tasks. In this paper, we propose a novel VAE based Text Style Transfer with pivOt Words Enhancement leaRning (VT-STOWER) method which utilizes Variational AutoEncoder (VAE) and external style embeddings to learn semantics and style distribution jointly. Additionally, we introduce pivot words learning, which is applied to learn decisive words for a specific style and thereby further improve the overall performance of the style transfer. The proposed VT-STOWER can be scaled to different TST scenarios given very limited and non-parallel training data with a novel and flexible style strength control mechanism. Experiments demonstrate that the VT-STOWER outperforms the state-of-the-art on sentiment, formality, and code-switching TST tasks."
                    },
                    {
                        "title": "Gradual Fine-Tuning for Low-Resource Domain Adaptation",
                        "abstract": "Fine-tuning is known to improve NLP models by adapting an initial model trained on more plentiful but less domain-salient examples to data in a target domain. Such domain adaptation is typically done using one stage of fine-tuning. We demonstrate that gradually fine-tuning in a multi-step process can yield substantial further gains and can be applied without modifying the model or learning objective."
                    },
                    {
                        "title": "Everything Is All It Takes: A Multipronged Strategy for Zero-Shot Cross-Lingual Information Extraction",
                        "abstract": "Zero-shot cross-lingual information extraction (IE) describes the construction of an IE model for some target language, given existing annotations exclusively in some other language, typically English. While the advance of pretrained multilingual encoders suggests an easy optimism of \u201ctrain on English, run on any language\u201d, we find through a thorough exploration and extension of techniques that a combination of approaches, both new and old, leads to better performance than any one cross-lingual strategy in particular. We explore techniques including data projection and self-training, and how different pretrained encoders impact them. We use English-to-Arabic IE as our initial example, demonstrating strong performance in this setting for event extraction, named entity recognition, part-of-speech tagging, and dependency parsing. We then apply data projection and self-training to three tasks across eight target languages. Because no single set of techniques performs the best across all tasks, we encourage practitioners to explore various configurations of the techniques described in this work when seeking to improve on zero-shot training."
                    },
                    {
                        "title": "Cross-Lingual BERT Contextual Embedding Space Mapping with Isotropic and Isometric Conditions",
                        "abstract": "Typically, a linearly orthogonal transformation mapping is learned by aligning static type-level embeddings to build a shared semantic space. In view of the analysis that contextual embeddings contain richer semantic features, we investigate a context-aware and dictionary-free mapping approach by leveraging parallel corpora. We illustrate that our contextual embedding space mapping significantly outperforms previous multilingual word embedding methods on the bilingual dictionary induction (BDI) task by providing a higher degree of isomorphism. To improve the quality of mapping, we also explore sense-level embeddings that are split from type-level representations, which can align spaces in a finer resolution and yield more precise mapping. Moreover, we reveal that contextual embedding spaces suffer from their natural properties -- anisotropy and anisometry. To mitigate these two problems, we introduce the iterative normalization algorithm as an imperative preprocessing step. Our findings unfold the tight relationship between isotropy, isometry, and isomorphism in normalized contextual embedding spaces."
                    }
                ]
            },
            "6aa0583d-9623-4a70-bdfd-2c4933d7e407": {
                "pk": "6aa0583d-9623-4a70-bdfd-2c4933d7e407",
                "name": "Philipp Koehn",
                "collaborators": [
                    "Kenton Murray",
                    "Weiting Tan",
                    "Haoran Xu",
                    "Lingfeng Shen",
                    "Yunmo Chen",
                    "Matthew Wiesner",
                    "B. Odoom",
                    "Jingyu (Jack) Zhang",
                    "Daniel Khashabi",
                    "Benjamin Van Durme"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Translation",
                    "Speech Recognition",
                    "Multilingual Models"
                ],
                "publications": [
                    {
                        "title": "Where are you from? Geolocating Speech and Applications to Language Identification",
                        "abstract": "We train models to answer the question, Where are you from? and show how such models can be repurposed for language identification (LID). To our knowledge, this paper is the first to introduce data sources, methods and models to tackle the task of geolocation of speech at a global scale, and the first to explore using geolocation as a proxy-task for LID. Specifically, we explore whether radio broadcasts with known origin can be used to train regression and classification-based models for geolocating speech. We build models on top of self-supervised pretrained models, using attention pooling to qualitatively verify that the model geolocates the speech itself, and not other channel artifacts.The best geolocation models localize speaker origin to around 650km. We confirm the value of speech geolocation as a proxy task by using speech geolocation models for zero-shot LID. Finally, we show that fine-tuning geolocation models for LID outperforms fine-tuning pretrained Wav2Vec2.0 models, and achieves state-of-the-art performance on the FLEURS benchmark."
                    },
                    {
                        "title": "The Language Barrier: Dissecting Safety Challenges of LLMs in Multilingual Contexts",
                        "abstract": "As the influence of large language models (LLMs) spans across global communities, their safety challenges in multilingual settings become paramount for alignment research. This paper examines the variations in safety challenges faced by LLMs across different languages and discusses approaches to alleviating such concerns. By comparing how state-of-the-art LLMs respond to the same set of malicious prompts written in higher- vs. lower-resource languages, we observe that (1) LLMs tend to generate unsafe responses much more often when a malicious prompt is written in a lower-resource language, and (2) LLMs tend to generate more irrelevant responses to malicious prompts in lower-resource languages. To understand where the discrepancy can be attributed, we study the effect of instruction tuning with reinforcement learning from human feedback (RLHF) or supervised finetuning (SFT) on the HH-RLHF dataset. Surprisingly, while training with high-resource languages improves model alignment, training in lower-resource languages yields minimal improvement. This suggests that the bottleneck of cross-lingual alignment is rooted in the pretraining stage. Our findings highlight the challenges in cross-lingual LLM safety, and we hope they inform future research in this direction."
                    },
                    {
                        "title": "Streaming Sequence Transduction through Dynamic Compression",
                        "abstract": "We introduce STAR (Stream Transduction with Anchor Representations), a novel Transformer-based model designed for efficient sequence-to-sequence transduction over streams. STAR dynamically segments input streams to create compressed anchor representations, achieving nearly lossless compression (12x) in Automatic Speech Recognition (ASR) and outperforming existing methods. Moreover, STAR demonstrates superior segmentation and latency-quality trade-offs in simultaneous speech-to-text tasks, optimizing latency, memory footprint, and quality."
                    },
                    {
                        "title": "Can Synthetic Speech Improve End-to-End Conversational Speech Translation?",
                        "abstract": "Conversational speech translation is an important technology that fosters communication among people of different language backgrounds. Three-way parallel data in the form of source speech, source transcript, and target translation is usually required to train end-to-end systems. However, such datasets are not readily available and are expensive to create as this involves multiple annotation stages. In this paper, we investigate the use of synthetic data from generative models, namely machine translation and text-to-speech synthesis, for training conversational speech translation systems. We show that adding synthetic data to the training recipe increasingly improves end-to-end training performance, especially when limited real data is available. However, when no real data is available, no amount of synthetic data helps."
                    },
                    {
                        "title": "X-ALMA: Plug&Play Modules and Adaptive Rejection for Quality Translation at Scale",
                        "abstract": "Large language models (LLMs) have achieved remarkable success across various NLP tasks, yet their focus has predominantly been on English due to English-centric pre-training and limited multilingual data. While some multilingual LLMs claim to support for hundreds of languages, models often fail to provide high-quality response for mid- and low-resource languages, leading to imbalanced performance heavily skewed in favor of high-resource languages like English and Chinese. In this paper, we prioritize quality over scaling number of languages, with a focus on multilingual machine translation task, and introduce X-ALMA, a model designed with a commitment to ensuring top-tier performance across 50 diverse languages, regardless of their resource levels. X-ALMA surpasses state-of-the-art open-source multilingual LLMs, such as Aya-101 and Aya-23, in every single translation direction on the FLORES and WMT'23 test datasets according to COMET-22. This is achieved by plug-and-play language-specific module architecture to prevent language conflicts during training and a carefully designed training regimen with novel optimization methods to maximize the translation performance. At the final stage of training regimen, our proposed Adaptive Rejection Preference Optimization (ARPO) surpasses existing preference optimization methods in translation tasks."
                    },
                    {
                        "title": "DiffNorm: Self-Supervised Normalization for Non-autoregressive Speech-to-speech Translation",
                        "abstract": "Non-autoregressive Transformers (NATs) are recently applied in direct speech-to-speech translation systems, which convert speech across different languages without intermediate text data. Although NATs generate high-quality outputs and offer faster inference than autoregressive models, they tend to produce incoherent and repetitive results due to complex data distribution (e.g., acoustic and linguistic variations in speech). In this work, we introduce DiffNorm, a diffusion-based normalization strategy that simplifies data distributions for training NAT models. After training with a self-supervised noise estimation objective, DiffNorm constructs normalized target data by denoising synthetically corrupted speech features. Additionally, we propose to regularize NATs with classifier-free guidance, improving model robustness and translation quality by randomly dropping out source information during training. Our strategies result in a notable improvement of about +7 ASR-BLEU for English-Spanish (En-Es) and +2 ASR-BLEU for English-French (En-Fr) translations on the CVSS benchmark, while attaining over 14x speedup for En-Es and 5x speedup for En-Fr translations compared to autoregressive baselines."
                    },
                    {
                        "title": "Machine Translation with Large Language Models: Prompting, Few-shot Learning, and Fine-tuning with QLoRA",
                        "abstract": "While large language models have made remarkable advancements in natural language generation, their potential in machine translation, especially when fine-tuned, remains under-explored. In our study, we conduct comprehensive experiments, evaluating 15 publicly available language models on machine translation tasks. We compare the performance across three methodologies: zero-shot prompting, few-shot learning, and fine-tuning. Central to our approach is the use of QLoRA, an efficient fine-tuning method. On French-English, QLoRA fine-tuning outperforms both few-shot learning and models trained from scratch. This superiority is highlighted in both sentence-level and document-level translations, with a significant BLEU score improvement of 28.93 over the prompting method. Impressively, with QLoRA, the enhanced performance is achieved by fine-tuning a mere 0.77% of the model\u2019s parameters."
                    },
                    {
                        "title": "Narrowing the Gap between Zero- and Few-shot Machine Translation by Matching Styles",
                        "abstract": "Large language models trained primarily in a monolingual setting have demonstrated their ability to generalize to machine translation using zero- and few-shot examples with in-context learning. However, even though zero-shot translations are relatively good, there remains a discernible gap comparing their performance with the few-shot setting. In this paper, we investigate the factors contributing to this gap and find that this gap can largely be closed (for about 70%) by matching the writing styles of the target corpus. Additionally, we explore potential approaches to enhance zero-shot baselines without the need for parallel demonstration examples, providing valuable insights into how these methods contribute to improving translation metrics."
                    }
                ]
            },
            "cd9b6b8f-afdb-4174-a53a-e4e3447416e2": {
                "pk": "cd9b6b8f-afdb-4174-a53a-e4e3447416e2",
                "name": "Daniel Khashabi",
                "collaborators": [
                    "Benjamin Van Durme",
                    "Jingyu (Jack) Zhang",
                    "Lingfeng Shen",
                    "Yining Lu",
                    "Abhishek Rao",
                    "Aitor Lewkowycz",
                    "Amanda Askell",
                    "Andrew M. Dai",
                    "Barret Zoph",
                    "David Dohan"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Machine Learning",
                    "Explainable AI"
                ],
                "publications": [
                    {
                        "title": "WorldAPIs: The World Is Worth How Many APIs? A Thought Experiment",
                        "abstract": "AI systems make decisions in physical environments through primitive actions or affordances that are accessed via API calls. While deploying AI agents in the real world involves numerous high-level actions, existing embodied simulators offer a limited set of domain-salient APIs. This naturally brings up the questions: how many primitive actions (APIs) are needed for a versatile embodied agent, and what should they look like? We explore this via a thought experiment: assuming that wikiHow tutorials cover a wide variety of human-written tasks, what is the space of APIs needed to cover these instructions? We propose a framework to iteratively induce new APIs by grounding wikiHow instruction to situated agent policies. Inspired by recent successes in large language models (LLMs) for embodied planning, we propose a few-shot prompting to steer GPT-4 to generate Pythonic programs as agent policies and bootstrap a universe of APIs by 1) reusing a seed set of APIs; and then 2) fabricate new API calls when necessary. The focus of this thought experiment is on defining these APIs rather than their executability. We apply the proposed pipeline on instructions from wikiHow tutorials. On a small fraction (0.5%) of tutorials, we induce an action space of 300+ APIs necessary for capturing the rich variety of tasks in the physical world. A detailed automatic and human analysis of the induction output reveals that the proposed pipeline enables effective reuse and creation of APIs. Moreover, a manual review revealed that existing simulators support only a small subset of the induced APIs (9 of the top 50 frequent APIs), motivating the development of action-rich embodied environments."
                    },
                    {
                        "title": "RORA: Robust Free-Text Rationale Evaluation",
                        "abstract": "Free-text rationales play a pivotal role in explainable NLP, bridging the knowledge and reasoning gaps behind a model's decision-making. However, due to the diversity of potential reasoning paths and a corresponding lack of definitive ground truth, their evaluation remains a challenge. Existing evaluation metrics rely on the degree to which a rationale supports a target label, but we find these fall short in evaluating rationales that inadvertently leak the labels. To address this problem, we propose RORA, a Robust free-text Rationale evaluation against label leakage. RORA quantifies the new information supplied by a rationale to justify the label. This is achieved by assessing the conditional V-information \\citep{hewitt-etal-2021-conditional} with a predictive family robust against leaky features that can be exploited by a small model. RORA consistently outperforms existing approaches in evaluating human-written, synthetic, or model-generated rationales, particularly demonstrating robustness against label leakage. We also show that RORA aligns well with human judgment, providing a more reliable and accurate measurement across diverse free-text rationales."
                    },
                    {
                        "title": "Core: Robust Factual Precision with Informative Sub-Claim Identification",
                        "abstract": "Hallucinations pose a challenge to the application of large language models (LLMs) thereby motivating the development of metrics to evaluate factual precision. We observe that popular metrics using the Decompose-Then-Verify framework, such as \\FActScore, can be manipulated by adding obvious or repetitive subclaims to artificially inflate scores. This observation motivates our new customizable plug-and-play subclaim selection component called Core, which filters down individual subclaims according to their uniqueness and informativeness. We show that many popular factual precision metrics augmented by Core are substantially more robust on a wide range of knowledge domains. We release an evaluation framework supporting easy and modular use of Core and various decomposition strategies, which we recommend adoption by the community. We also release an expansion of the FActScore biography dataset to facilitate further studies of decomposition-based factual precision evaluation."
                    },
                    {
                        "title": "Rel-A.I.: An Interaction-Centered Approach To Measuring Human-LM Reliance",
                        "abstract": "The ability to communicate uncertainty, risk, and limitation is crucial for the safety of large language models. However, current evaluations of these abilities rely on simple calibration, asking whether the language generated by the model matches appropriate probabilities. Instead, evaluation of this aspect of LLM communication should focus on the behaviors of their human interlocutors: how much do they rely on what the LLM says? Here we introduce an interaction-centered evaluation framework called Rel-A.I. (pronounced\"rely\"}) that measures whether humans rely on LLM generations. We use this framework to study how reliance is affected by contextual features of the interaction (e.g, the knowledge domain that is being discussed), or the use of greetings communicating warmth or competence (e.g.,\"I'm happy to help!\"). We find that contextual characteristics significantly affect human reliance behavior. For example, people rely 10% more on LMs when responding to questions involving calculations and rely 30% more on LMs that are perceived as more competent. Our results show that calibration and language quality alone are insufficient in evaluating the risks of human-LM interactions, and illustrate the need to consider features of the interactional context."
                    },
                    {
                        "title": "Benchmarking Language Model Creativity: A Case Study on Code Generation",
                        "abstract": "As LLMs become increasingly prevalent, it is interesting to consider how ``creative'' these models can be. From cognitive science, creativity consists of at least two key characteristics: \\emph{convergent} thinking (purposefulness to achieve a given goal) and \\emph{divergent} thinking (adaptability to new environments or constraints) \\citep{runco2003critical}. In this work, we introduce a framework for quantifying LLM creativity that incorporates the two characteristics. This is achieved by (1) Denial Prompting pushes LLMs to come up with more creative solutions to a given problem by incrementally imposing new constraints on the previous solution, compelling LLMs to adopt new strategies, and (2) defining and computing the NeoGauge metric which examines both convergent and divergent thinking in the generated creative responses by LLMs. We apply the proposed framework on Codeforces problems, a natural data source for collecting human coding solutions. We quantify NeoGauge for various proprietary and open-source models and find that even the most creative model, GPT-4, still falls short of demonstrating human-like creativity. We also experiment with advanced reasoning strategies (MCTS, self-correction, etc.) and observe no significant improvement in creativity. As a by-product of our analysis, we release NeoCoder dataset for reproducing our results on future models."
                    },
                    {
                        "title": "Insights into LLM Long-Context Failures: When Transformers Know but Don't Tell",
                        "abstract": "Large Language Models (LLMs) exhibit positional bias, struggling to utilize information from the middle or end of long contexts. Our study explores LLMs' long-context reasoning by probing their hidden representations. We find that while LLMs encode the position of target information, they often fail to leverage this in generating accurate responses. This reveals a disconnect between information retrieval and utilization, a\"know but don't tell\"phenomenon. We further analyze the relationship between extraction time and final accuracy, offering insights into the underlying mechanics of transformer models."
                    },
                    {
                        "title": "SemStamp: A Semantic Watermark with Paraphrastic Robustness for Text Generation",
                        "abstract": "Existing watermarked generation algorithms employ token-level designs and therefore, are vulnerable to paraphrase attacks. To address this issue, we introduce watermarking on the semantic representation of sentences. We propose SemStamp, a robust sentence-level semantic watermarking algorithm that uses locality-sensitive hashing (LSH) to partition the semantic space of sentences. The algorithm encodes and LSH-hashes a candidate sentence generated by a language model, and conducts rejection sampling until the sampled sentence falls in watermarked partitions in the semantic embedding space. To test the paraphrastic robustness of watermarking algorithms, we propose a \u201cbigram paraphrase\u201d attack that produces paraphrases with small bigram overlap with the original sentence. This attack is shown to be effective against existing token-level watermark algorithms, while posing only minor degradations to SemStamp. Experimental results show that our novel semantic watermark algorithm is not only more robust than the previous state-of-the-art method on various paraphrasers and domains, but also better at preserving the quality of generation."
                    },
                    {
                        "title": "Tur[k]ingBench: A Challenge Benchmark for Web Agents",
                        "abstract": "Can advanced multi-modal models effectively tackle complex web-based tasks? Such tasks are often found on crowdsourcing platforms, where crowdworkers engage in challenging micro-tasks within web-based environments. Building on this idea, we present TurkingBench, a benchmark consisting of tasks presented as web pages with textual instructions and multi-modal contexts. Unlike previous approaches that rely on artificially synthesized web pages, our benchmark uses natural HTML pages originally designed for crowdsourcing workers to perform various annotation tasks. Each task's HTML instructions are instantiated with different values derived from crowdsourcing tasks, creating diverse instances. This benchmark includes 32.2K instances spread across 158 tasks. To support the evaluation of TurkingBench, we have developed a framework that links chatbot responses to actions on web pages (e.g., modifying a text box, selecting a radio button). We assess the performance of cutting-edge private and open-source models, including language-only and vision-language models (such as GPT4 and InternVL), on this benchmark. Our results show that while these models outperform random chance, there is still significant room for improvement. We hope that this benchmark will drive progress in the evaluation and development of web-based agents."
                    },
                    {
                        "title": "The Language Barrier: Dissecting Safety Challenges of LLMs in Multilingual Contexts",
                        "abstract": "As the influence of large language models (LLMs) spans across global communities, their safety challenges in multilingual settings become paramount for alignment research. This paper examines the variations in safety challenges faced by LLMs across different languages and discusses approaches to alleviating such concerns. By comparing how state-of-the-art LLMs respond to the same set of malicious prompts written in higher- vs. lower-resource languages, we observe that (1) LLMs tend to generate unsafe responses much more often when a malicious prompt is written in a lower-resource language, and (2) LLMs tend to generate more irrelevant responses to malicious prompts in lower-resource languages. To understand where the discrepancy can be attributed, we study the effect of instruction tuning with reinforcement learning from human feedback (RLHF) or supervised finetuning (SFT) on the HH-RLHF dataset. Surprisingly, while training with high-resource languages improves model alignment, training in lower-resource languages yields minimal improvement. This suggests that the bottleneck of cross-lingual alignment is rooted in the pretraining stage. Our findings highlight the challenges in cross-lingual LLM safety, and we hope they inform future research in this direction."
                    },
                    {
                        "title": "AnaloBench: Benchmarking the Identification of Abstract and Long-context Analogies",
                        "abstract": "Humans regularly engage in analogical thinking, relating personal experiences to current situations (X is analogous to Y because of Z). Analogical thinking allows humans to solve problems in creative ways, grasp difficult concepts, and articulate ideas more effectively. Can language models (LMs) do the same? To answer this question, we propose AnaloBench, a benchmark to determine analogical reasoning ability in LMs. Our benchmarking approach focuses on aspects of this ability that are common among humans: (i) recalling related experiences from a large amount of information, and (ii) applying analogical reasoning to complex and lengthy scenarios. We test a broad collection of proprietary models (e.g., GPT family, Claude V2) and open source models such as LLaMA2. As in prior results, scaling up LMs results in some performance boosts. Surprisingly, scale offers minimal gains when, (i) analogies involve lengthy scenarios, or (ii) recalling relevant scenarios from a large pool of information, a process analogous to finding a needle in a haystack. We hope these observations encourage further research in this field."
                    },
                    {
                        "title": "Dated Data: Tracing Knowledge Cutoffs in Large Language Models",
                        "abstract": "Released Large Language Models (LLMs) are often paired with a claimed knowledge cutoff date, or the dates at which training data was gathered. Such information is crucial for applications where the LLM must provide up to date information. However, this statement only scratches the surface: do all resources in the training data share the same knowledge cutoff date? Does the model's demonstrated knowledge for these subsets closely align to their cutoff dates? In this work, we define the notion of an effective cutoff. This is distinct from the LLM designer reported cutoff and applies separately to sub-resources and topics. We propose a simple approach to estimate effective cutoffs on the resource-level temporal alignment of an LLM by probing across versions of the data. Using this analysis, we find that effective cutoffs often differ from reported cutoffs. To understand the root cause of this observation, we conduct a direct large-scale analysis on open pre-training datasets. Our analysis reveals two reasons for these inconsistencies: (1) temporal biases of CommonCrawl data due to non-trivial amounts of old data in new dumps and (2) complications in LLM deduplication schemes involving semantic duplicates and lexical near-duplicates. Overall, our results show that knowledge cutoffs are not as simple as they have seemed and that care must be taken both by LLM dataset curators as well as practitioners who seek to use information from these models."
                    },
                    {
                        "title": "Ask Again, Then Fail: Large Language Models\u2019 Vacillations in Judgment",
                        "abstract": "We observe that current large language models often waver in their judgments when faced with follow-up questions, even if the original judgment was correct. This wavering presents a signi\ufb01cant challenge for generating reliable responses and building user trust. To comprehensively assess this issue, we introduce a F OLLOW - UP Q UESTIONING M ECHANISM along with two metrics to quantify this inconsistency, con\ufb01rming its widespread presence in current large language models. Furthermore, to mitigate this issue, we explore various prompting strategies for closed-source models, and develop a training-based framework U NWAVERING -FQ that teaches large language models to maintain their originally correct judgments through synthesized high-quality preference data. Our experimental re-sults con\ufb01rm the effectiveness of our framework and its ability to enhance the general capabilities of large language models."
                    },
                    {
                        "title": "LLaVolta: Efficient Multi-modal Models via Stage-wise Visual Context Compression",
                        "abstract": "While significant advancements have been made in compressed representations for text embeddings in large language models (LLMs), the compression of visual tokens in large multi-modal models (LMMs) has remained a largely overlooked area. In this work, we present the study on the analysis of redundancy concerning visual tokens and efficient training within these models. Our initial experiments show that eliminating up to 70% of visual tokens at the testing stage by simply average pooling only leads to a minimal 3% reduction in visual question answering accuracy on the GQA benchmark, indicating significant redundancy in visual context. Addressing this, we introduce Visual Context Compressor, which reduces the number of visual tokens during training to enhance training efficiency without sacrificing performance. To minimize information loss caused by the compression on visual tokens while maintaining training efficiency, we develop LLaVolta as a lite training scheme. LLaVolta incorporates stage-wise visual context compression to progressively compress the visual tokens from heavily to lightly, and finally no compression at the end of training, yielding no loss of information when testing. Extensive experiments demonstrate that our approach enhances the performance of MLLMs in both image-language and video-language understanding, while also significantly cutting training costs. Code is available at https://github.com/Beckschen/LLaVolta"
                    },
                    {
                        "title": "k-SemStamp: A Clustering-Based Semantic Watermark for Detection of Machine-Generated Text",
                        "abstract": "Recent watermarked generation algorithms inject detectable signatures during language generation to facilitate post-hoc detection. While token-level watermarks are vulnerable to paraphrase attacks, SemStamp (Hou et al., 2023) applies watermark on the semantic representation of sentences and demonstrates promising robustness. SemStamp employs locality-sensitive hashing (LSH) to partition the semantic space with arbitrary hyperplanes, which results in a suboptimal tradeoff between robustness and speed. We propose k-SemStamp, a simple yet effective enhancement of SemStamp, utilizing k-means clustering as an alternative of LSH to partition the embedding space with awareness of inherent semantic structure. Experimental results indicate that k-SemStamp saliently improves its robustness and sampling efficiency while preserving the generation quality, advancing a more effective tool for machine-generated text detection."
                    },
                    {
                        "title": "Verifiable by Design: Aligning Language Models to Quote from Pre-Training Data",
                        "abstract": "To trust the fluent generations of large language models (LLMs), humans must be able to verify their correctness against trusted, external sources. Recent efforts, such as providing citations via retrieved documents or post-hoc provenance, enhance verifiability but still provide no guarantees on their correctness. To address these limitations, we tackle the verifiability goal with a different philosophy: trivializing the verification process by developing models that quote verbatim statements from trusted sources in pre-training data. We propose Quote-Tuning, and demonstrate it is feasible to align LLMs to provide quoted statements from data memorized during pre-training. The core of Quote-Tuning is a fast membership inference function (Marone and Van Durme, 2023) that efficiently verifies text against a trusted corpus. We leverage this tool to design a reward function to quantify quotes in model responses, which is then used to create a dataset for preference learning. Experimental results show that Quote-Tuning significantly increases verbatim quotes from high-quality pre-training documents by 55% to 130% relative to un-tuned models while maintaining response quality. Quote-Tuning also generalizes quoting to out-of-domain data, is applicable in different tasks, and provides additional benefits to truthfulness. Our method not only serves as a hassle-free method to increase quoting but also opens up avenues for improving LLM trustworthiness through better verifiability."
                    },
                    {
                        "title": "SELF-[IN]CORRECT: LLMs Struggle with Refining Self-Generated Responses",
                        "abstract": "Can LLMs continually improve their previous outputs for better results? An affirmative answer would require LLMs to be better at discriminating among previously-generated alternatives, than generating initial responses. We explore the validity of this hypothesis in practice. We first introduce a unified framework that allows us to compare the generative and discriminative capability of any model on any task. Then, in our resulting experimental analysis of several LLMs, we do not observe those models\u2019 performance on discrimination to be reliably better than generation. We hope these findings inform the growing literature on self-improvement AI systems."
                    },
                    {
                        "title": "Naive Bayes-based Context Extension for Large Language Models",
                        "abstract": "Large Language Models (LLMs) have shown promising in-context learning abilities. However, conventional In-Context Learning (ICL) approaches are often impeded by length limitations of transformer architecture, which pose challenges when attempting to effectively integrate supervision from a substantial number of demonstration examples. In this paper, we introduce a novel framework, called Naive Bayes-based Context Extension (NBCE), to enable existing LLMs to perform ICL with an increased number of demonstrations by significantly expanding their context size. Importantly, this expansion does not require fine-tuning or dependence on particular model architectures, all the while preserving linear efficiency. NBCE initially splits the context into equal-sized windows fitting the target LLM\u2019s maximum length. Then, it introduces a voting mechanism to select the most relevant window, regarded as the posterior context. Finally, it employs Bayes\u2019 theorem to generate the test task. Our experimental results demonstrate that NBCE substantially enhances performance, particularly as the number of demonstration examples increases, consistently outperforming alternative methods. The NBCE code will be made publicly accessible. The code NBCE is available at: https://github.com/amurtadha/NBCE-master"
                    },
                    {
                        "title": "DiffNorm: Self-Supervised Normalization for Non-autoregressive Speech-to-speech Translation",
                        "abstract": "Non-autoregressive Transformers (NATs) are recently applied in direct speech-to-speech translation systems, which convert speech across different languages without intermediate text data. Although NATs generate high-quality outputs and offer faster inference than autoregressive models, they tend to produce incoherent and repetitive results due to complex data distribution (e.g., acoustic and linguistic variations in speech). In this work, we introduce DiffNorm, a diffusion-based normalization strategy that simplifies data distributions for training NAT models. After training with a self-supervised noise estimation objective, DiffNorm constructs normalized target data by denoising synthetically corrupted speech features. Additionally, we propose to regularize NATs with classifier-free guidance, improving model robustness and translation quality by randomly dropping out source information during training. Our strategies result in a notable improvement of about +7 ASR-BLEU for English-Spanish (En-Es) and +2 ASR-BLEU for English-French (En-Fr) translations on the CVSS benchmark, while attaining over 14x speedup for En-Es and 5x speedup for En-Fr translations compared to autoregressive baselines."
                    }
                ]
            },
            "5ee8949a-26a8-4e0b-80cd-359588d9a91f": {
                "pk": "5ee8949a-26a8-4e0b-80cd-359588d9a91f",
                "name": "Kenton Murray",
                "collaborators": [
                    "Philipp Koehn",
                    "Haoran Xu",
                    "Weiting Tan",
                    "Tom Kocmi",
                    "Eleftherios Avramidis",
                    "Rachel Bawden",
                    "Ondrej Bojar",
                    "Anton Dvorkovich",
                    "C. Federmann",
                    "Mark Fishel"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Translation",
                    "Multilingual Models",
                    "Language Identification"
                ],
                "publications": [
                    {
                        "title": "Where are you from? Geolocating Speech and Applications to Language Identification",
                        "abstract": "We train models to answer the question, Where are you from? and show how such models can be repurposed for language identification (LID). To our knowledge, this paper is the first to introduce data sources, methods and models to tackle the task of geolocation of speech at a global scale, and the first to explore using geolocation as a proxy-task for LID. Specifically, we explore whether radio broadcasts with known origin can be used to train regression and classification-based models for geolocating speech. We build models on top of self-supervised pretrained models, using attention pooling to qualitatively verify that the model geolocates the speech itself, and not other channel artifacts.The best geolocation models localize speaker origin to around 650km. We confirm the value of speech geolocation as a proxy task by using speech geolocation models for zero-shot LID. Finally, we show that fine-tuning geolocation models for LID outperforms fine-tuning pretrained Wav2Vec2.0 models, and achieves state-of-the-art performance on the FLEURS benchmark."
                    },
                    {
                        "title": "Upsample or Upweight? Balanced Training on Heavily Imbalanced Datasets",
                        "abstract": "Data availability across domains often follows a long-tail distribution: a few domains have abundant data, while most face dat . a scarcity. This imbalance poses challenges in training language models uniformly across all domains. In our study, we focus on multilingual settings, where data sizes vary significantly between high- and low-resource languages. Common strategies to address this include upsampling low-resource languages (Temperature Sampling) or upweighting their loss (Scalarization). Although often considered equivalent, this assumption has not been proven, which motivates our study. Through both theoretical and empirical analysis, we identify the conditions under which these approaches are equivalent and when they diverge. Specifically, we demonstrate that these two methods are equivalent under full gradient descent, but this equivalence breaks down with stochastic gradient descent. Empirically, we observe that Temperature Sampling converges more quickly but is prone to overfitting. We argue that this faster convergence is likely due to the lower variance in gradient estimations, as shown theoretically. Based on these insights, we propose Cooldown, a strategy that reduces sampling temperature during training, accelerating convergence without overfitting to low-resource languages. Our method is competitive with existing data re-weighting and offers computational efficiency."
                    },
                    {
                        "title": "Preliminary WMT24 Ranking of General MT Systems and LLMs",
                        "abstract": "This is the preliminary ranking of WMT24 General MT systems based on automatic metrics. The official ranking will be a human evaluation, which is superior to the automatic ranking and supersedes it. The purpose of this report is not to interpret any findings but only provide preliminary results to the participants of the General MT task that may be useful during the writing of the system submission."
                    },
                    {
                        "title": "X-ALMA: Plug&Play Modules and Adaptive Rejection for Quality Translation at Scale",
                        "abstract": "Large language models (LLMs) have achieved remarkable success across various NLP tasks, yet their focus has predominantly been on English due to English-centric pre-training and limited multilingual data. While some multilingual LLMs claim to support for hundreds of languages, models often fail to provide high-quality response for mid- and low-resource languages, leading to imbalanced performance heavily skewed in favor of high-resource languages like English and Chinese. In this paper, we prioritize quality over scaling number of languages, with a focus on multilingual machine translation task, and introduce X-ALMA, a model designed with a commitment to ensuring top-tier performance across 50 diverse languages, regardless of their resource levels. X-ALMA surpasses state-of-the-art open-source multilingual LLMs, such as Aya-101 and Aya-23, in every single translation direction on the FLORES and WMT'23 test datasets according to COMET-22. This is achieved by plug-and-play language-specific module architecture to prevent language conflicts during training and a carefully designed training regimen with novel optimization methods to maximize the translation performance. At the final stage of training regimen, our proposed Adaptive Rejection Preference Optimization (ARPO) surpasses existing preference optimization methods in translation tasks."
                    },
                    {
                        "title": "Contrastive Preference Optimization: Pushing the Boundaries of LLM Performance in Machine Translation",
                        "abstract": "Moderate-sized large language models (LLMs) -- those with 7B or 13B parameters -- exhibit promising machine translation (MT) performance. However, even the top-performing 13B LLM-based translation models, like ALMA, does not match the performance of state-of-the-art conventional encoder-decoder translation models or larger-scale LLMs such as GPT-4. In this study, we bridge this performance gap. We first assess the shortcomings of supervised fine-tuning for LLMs in the MT task, emphasizing the quality issues present in the reference data, despite being human-generated. Then, in contrast to SFT which mimics reference translations, we introduce Contrastive Preference Optimization (CPO), a novel approach that trains models to avoid generating adequate but not perfect translations. Applying CPO to ALMA models with only 22K parallel sentences and 12M parameters yields significant improvements. The resulting model, called ALMA-R, can match or exceed the performance of the WMT competition winners and GPT-4 on WMT'21, WMT'22 and WMT'23 test datasets."
                    },
                    {
                        "title": "Narrowing the Gap between Zero- and Few-shot Machine Translation by Matching Styles",
                        "abstract": "Large language models trained primarily in a monolingual setting have demonstrated their ability to generalize to machine translation using zero- and few-shot examples with in-context learning. However, even though zero-shot translations are relatively good, there remains a discernible gap comparing their performance with the few-shot setting. In this paper, we investigate the factors contributing to this gap and find that this gap can largely be closed (for about 70%) by matching the writing styles of the target corpus. Additionally, we explore potential approaches to enhance zero-shot baselines without the need for parallel demonstration examples, providing valuable insights into how these methods contribute to improving translation metrics."
                    },
                    {
                        "title": "Findings of the 2023 Conference on Machine Translation (WMT23): LLMs Are Here but Not Quite There Yet",
                        "abstract": "This paper presents the results of the General Machine Translation Task organised as part of the 2023 Conference on Machine Translation (WMT). In the general MT task, participants were asked to build machine translation systems for any of 8 language pairs (corresponding to 14 translation directions), to be evaluated on test sets consisting of up to four different domains. We evaluate system outputs with professional human annotators using a combination of source-based Direct Assessment and scalar quality metric (DA+SQM)."
                    }
                ]
            }
        }
    },
    "2405.14366": {
        "paper_data": {
            "title": "MiniCache: KV Cache Compression in Depth Dimension for Large Language Models",
            "url": "http://arxiv.org/abs/2405.14366v2",
            "arxiv_id": "2405.14366",
            "authors": [
                "Akide Liu",
                "Jing Liu",
                "Zizheng Pan",
                "Yefei He",
                "Gholamreza Haffari",
                "Bohan Zhuang"
            ],
            "abstract": "A critical approach for efficiently deploying computationally demanding large language models (LLMs) is Key-Value (KV) caching. The KV cache stores key-value states of previously generated tokens, significantly reducing the need for repetitive computations and thereby lowering latency in autoregressive generation. However, the size of the KV cache grows linearly with sequence length, posing challenges for applications requiring long context input and extensive sequence generation. In this paper, we present a simple yet effective approach, called MiniCache, to compress the KV cache across layers from a novel depth perspective, significantly reducing the memory footprint for LLM inference. Our approach is based on the observation that KV cache states exhibit high similarity between the adjacent layers in the middle-to-deep portion of LLMs. To facilitate merging, we propose disentangling the states into the magnitude and direction components, interpolating the directions of the state vectors while preserving their lengths unchanged. Furthermore, we introduce a token retention strategy to keep highly distinct state pairs unmerged, thus preserving the information with minimal additional storage overhead. Our MiniCache is training-free and general, complementing existing KV cache compression strategies, such as quantization and sparsity. We conduct a comprehensive evaluation of MiniCache utilizing various models including LLaMA-2, LLaMA-3, Phi-3, Mistral, and Mixtral across multiple benchmarks, demonstrating its exceptional performance in achieving superior compression ratios and high throughput. On the ShareGPT dataset, LLaMA-2-7B with 4-bit MiniCache achieves a remarkable compression ratio of up to 5.02x, enhances inference throughput by approximately 5x, and reduces the memory footprint by 41% compared to the FP16 full cache baseline, all while maintaining near-lossless performance.",
            "introduction": "   1 Introduction  Large Language Models (LLMs), exemplified by the GPT series [1, 2, 3] and the LLaMA series [4, 5, 6], have emerged as pivotal innovations within the artificial general intelligence, significantly enhancing the capabilities of natural language processing. However, these models are meticulously trained using extensive computational resources [7] and massive datasets [8], which enables them to produce text that effectively mimics human writing styles and conducts complex reasoning analysis, yet raises the challenge of efficient deployment and serving. Within the inference framework of LLMs, KV caches [9] are crucial for storing pre-computed keys and values, thus avoiding repeated calculations over the preceding context and substantially enhancing LLMs\u2019 deployment efficiency. However, the increasing demand for longer sequence lengths results in voluminous cached states, leading to significant memory consumption during generation. For instance, a 175B GPT-3 model [2], with a batch size of 64 and a sequence length of 4,096 tokens (both prefilled and generated), requires approximately 1,208GB of GPU memory. This requirement is 3.45\u00d73.45\\times3.45 \u00d7 greater than the memory used to store the model\u2019s weights. In this context, KV cache compression is of paramount importance due to its clear benefits: 1) it largely reduces the memory footprint allowing for faster generation and larger batch serving; 2) it significantly lowers the cost per token, demonstrating substantial commercial benefits.   Figure 1: Overview of our MiniCache strategy and example results: (a) shows the observation that the KV cache states between two adjacent layers are highly similar, particularly across the middle to deep layers. The x-axis uses index/2 to represent the similarities for each pair of layers. (b) compares the performance of MiniCache, and the mean baseline, which simply averages the KV caches of two layers, using the LLaMA-3-70B model [6] on the GSM8K dataset [10]. MiniCache, which begins merging from the half-layer depth, achieves near-lossless performance. (c) highlights the primary difference between MiniCache and previous approaches. MiniCache investigates the inter-layer redundancy of KV caches along the depth dimension of LLMs, an aspect overlooked by intra-layer-based methods. Here, T\ud835\udc47Titalic_T refers to the last timestamp of pre-filling, and T+1\ud835\udc471T+1italic_T + 1 des to the first timestamp of decoding.    Existing KV cache compression efforts can be roughly categorized into two types, namely quantization and sparsity. The quantization approaches [11, 12] propose storing the KV states in low-bit numerical values. Typically, FlexGen [13] demonstrates that 4-bit KV cache quantization can achieve lossless performance. In contrast, sparsity-driven methods aim to retain only the salient tokens while evicting the rest, either heuristically [14, 15] or adaptively [16]. Some approaches [11] explore the intersection of these two types, by assigning high-bit to salient tokens and extremely low-bit to the rest of the tokens, achieving more aggressive memory gain. Despite these innovations, existing literature merely consider the intra-layer redundancy, while neglecting another important complementary direction \u2013 the inter-layer redundancy, as illustrated in the Figure\u00a01(c).   Our analysis begins by exploring the redundancy of KV caches along the depth dimension, as shown in Figure 1(a). We observe that KV cache states exhibit high similarity between neighbouring layers in the middle-to-deep portions of LLMs. This intriguing property suggests that states paired by position between adjacent layers can be accurately merged into",
            "references": [
                {
                    "title": "Layer-Condensed KV Cache for Efficient Inference of Large Language Models",
                    "abstract": "Huge memory consumption has been a major bottleneck for deploying high-throughput large language models in real-world applications. In addition to the large number of parameters, the key-value (KV) cache for the attention mechanism in the transformer architecture consumes a significant amount of memory, especially when the number of layers is large for deep language models. In this paper, we propose a novel method that only computes and caches the KVs of a small number of layers, thus significantly saving memory consumption and improving inference throughput. Our experiments on large language models show that our method achieves up to 26$\\times$ higher throughput than standard transformers and competitive performance in language modeling and downstream tasks. In addition, our method is orthogonal to existing transformer memory-saving techniques, so it is straightforward to integrate them with our model, achieving further improvement in inference efficiency. Our code is available at https://github.com/whyNLP/LCKV."
                },
                {
                    "title": "DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model",
                    "abstract": "We present DeepSeek-V2, a strong Mixture-of-Experts (MoE) language model characterized by economical training and efficient inference. It comprises 236B total parameters, of which 21B are activated for each token, and supports a context length of 128K tokens. DeepSeek-V2 adopts innovative architectures including Multi-head Latent Attention (MLA) and DeepSeekMoE. MLA guarantees efficient inference through significantly compressing the Key-Value (KV) cache into a latent vector, while DeepSeekMoE enables training strong models at an economical cost through sparse computation. Compared with DeepSeek 67B, DeepSeek-V2 achieves significantly stronger performance, and meanwhile saves 42.5% of training costs, reduces the KV cache by 93.3%, and boosts the maximum generation throughput to 5.76 times. We pretrain DeepSeek-V2 on a high-quality and multi-source corpus consisting of 8.1T tokens, and further perform Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) to fully unlock its potential. Evaluation results show that, even with only 21B activated parameters, DeepSeek-V2 and its chat versions still achieve top-tier performance among open-source models."
                },
                {
                    "title": "LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding",
                    "abstract": "We present LayerSkip, an end-to-end solution to speed-up inference of large language models (LLMs). First, during training we apply layer dropout, with low dropout rates for earlier layers and higher dropout rates for later layers, and an early exit loss where all transformer layers share the same exit. Second, during inference, we show that this training recipe increases the accuracy of early exit at earlier layers, without adding any auxiliary layers or modules to the model. Third, we present a novel self-speculative decoding solution where we exit at early layers and verify and correct with remaining layers of the model. Our proposed self-speculative decoding approach has less memory footprint than other speculative decoding approaches and benefits from shared compute and activations of the draft and verification stages. We run experiments on different Llama model sizes on different types of training: pretraining from scratch, continual pretraining, finetuning on specific data domain, and finetuning on specific task. We implement our inference solution and show speedups of up to 2.16x on summarization for CNN/DM documents, 1.82x on coding, and 2.0x on TOPv2 semantic parsing task. We open source our code and checkpoints at https://github.com/facebookresearch/LayerSkip."
                },
                {
                    "title": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
                    "abstract": "We introduce phi-3-mini, a 3.8 billion parameter language model trained on 3.3 trillion tokens, whose overall performance, as measured by both academic benchmarks and internal testing, rivals that of models such as Mixtral 8x7B and GPT-3.5 (e.g., phi-3-mini achieves 69% on MMLU and 8.38 on MT-bench), despite being small enough to be deployed on a phone. Our training dataset is a scaled-up version of the one used for phi-2, composed of heavily filtered publicly available web data and synthetic data. The model is also further aligned for robustness, safety, and chat format. We also provide parameter-scaling results with a 7B, 14B models trained for 4.8T tokens, called phi-3-small, phi-3-medium, both significantly more capable than phi-3-mini (e.g., respectively 75%, 78% on MMLU, and 8.7, 8.9 on MT-bench). To enhance multilingual, multimodal, and long-context capabilities, we introduce three models in the phi-3.5 series: phi-3.5-mini, phi-3.5-MoE, and phi-3.5-Vision. The phi-3.5-MoE, a 16 x 3.8B MoE model with 6.6 billion active parameters, achieves superior performance in language reasoning, math, and code tasks compared to other open-source models of similar scale, such as Llama 3.1 and the Mixtral series, and on par with Gemini-1.5-Flash and GPT-4o-mini. Meanwhile, phi-3.5-Vision, a 4.2 billion parameter model derived from phi-3.5-mini, excels in reasoning tasks and is adept at handling both single-image and text prompts, as well as multi-image and text prompts."
                },
                {
                    "title": "Mixture-of-Depths: Dynamically allocating compute in transformer-based language models",
                    "abstract": "Transformer-based language models spread FLOPs uniformly across input sequences. In this work we demonstrate that transformers can instead learn to dynamically allocate FLOPs (or compute) to specific positions in a sequence, optimising the allocation along the sequence for different layers across the model depth. Our method enforces a total compute budget by capping the number of tokens ($k$) that can participate in the self-attention and MLP computations at a given layer. The tokens to be processed are determined by the network using a top-$k$ routing mechanism. Since $k$ is defined a priori, this simple procedure uses a static computation graph with known tensor sizes, unlike other conditional computation techniques. Nevertheless, since the identities of the $k$ tokens are fluid, this method can expend FLOPs non-uniformly across the time and model depth dimensions. Thus, compute expenditure is entirely predictable in sum total, but dynamic and context-sensitive at the token-level. Not only do models trained in this way learn to dynamically allocate compute, they do so efficiently. These models match baseline performance for equivalent FLOPS and wall-clock times to train, but require a fraction of the FLOPs per forward pass, and can be upwards of 50\\% faster to step during post-training sampling."
                },
                {
                    "title": "The Unreasonable Ineffectiveness of the Deeper Layers",
                    "abstract": "We empirically study a simple layer-pruning strategy for popular families of open-weight pretrained LLMs, finding minimal degradation of performance on different question-answering benchmarks until after a large fraction (up to half) of the layers are removed. To prune these models, we identify the optimal block of layers to prune by considering similarity across layers; then, to\"heal\"the damage, we perform a small amount of finetuning. In particular, we use parameter-efficient finetuning (PEFT) methods, specifically quantization and Low Rank Adapters (QLoRA), such that each of our experiments can be performed on a single A100 GPU. From a practical perspective, these results suggest that layer pruning methods can complement other PEFT strategies to further reduce computational resources of finetuning on the one hand, and can improve the memory and latency of inference on the other hand. From a scientific perspective, the robustness of these LLMs to the deletion of layers implies either that current pretraining methods are not properly leveraging the parameters in the deeper layers of the network or that the shallow layers play a critical role in storing knowledge."
                },
                {
                    "title": "An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Models",
                    "abstract": "In this study, we identify the inefficient attention phenomena in Large Vision-Language Models (LVLMs), notably within prominent models like LLaVA-1.5, QwenVL-Chat and Video-LLaVA. We find out that the attention computation over visual tokens is of extreme inefficiency in the deep layers of popular LVLMs, suggesting a need for a sparser approach compared to textual data handling. To this end, we introduce FastV, a versatile plug-and-play method designed to optimize computational efficiency by learning adaptive attention patterns in early layers and pruning visual tokens in subsequent ones. Our evaluations demonstrate FastV's ability to dramatically reduce computational costs (e.g., a 45 reduction in FLOPs for LLaVA-1.5-13B) without sacrificing performance in a wide range of image and video understanding tasks. The computational efficiency and performance trade-off of FastV are highly customizable and pareto-efficient. It can compress the FLOPs of a 13B-parameter model to achieve a lower budget than that of a 7B-parameter model, while still maintaining superior performance. We believe FastV has practical values for deployment of LVLMs in edge devices and commercial models. Code is released at https://github.com/pkunlp-icler/FastV."
                },
                {
                    "title": "GEAR: An Efficient KV Cache Compression Recipe for Near-Lossless Generative Inference of LLM",
                    "abstract": "Key-value (KV) caching has become the de-facto to accelerate generation speed for large language models (LLMs) inference. However, the growing cache demand with increasing sequence length has transformed LLM inference to be a memory bound problem, significantly constraining the system throughput. Existing methods rely on dropping unimportant tokens or quantizing all entries uniformly. Such methods, however, often incur high approximation errors to represent the compressed matrices. The autoregressive decoding process further compounds the error of each step, resulting in critical deviation in model generation and deterioration of performance. To tackle this challenge, we propose GEAR, an efficient KV cache compression framework that achieves near-lossless high-ratio compression. GEAR first applies quantization to majority of entries of similar magnitudes to ultra-low precision. It then employs a low rank matrix to approximate the quantization error, and a sparse matrix to remedy individual errors from outlier entries. By adeptly integrating three techniques, GEAR is able to fully exploit their synergistic potentials. Our experiments demonstrate that compared to alternatives, GEAR achieves near-lossless 4-bit KV cache compression with up to 2.38x throughput improvement, while reducing peak-memory size up to 2.29x. Our code is publicly available at https://github.com/HaoKang-Timmy/GEAR."
                },
                {
                    "title": "DoRA: Weight-Decomposed Low-Rank Adaptation",
                    "abstract": "Among the widely used parameter-efficient fine-tuning (PEFT) methods, LoRA and its variants have gained considerable popularity because of avoiding additional inference costs. However, there still often exists an accuracy gap between these methods and full fine-tuning (FT). In this work, we first introduce a novel weight decomposition analysis to investigate the inherent differences between FT and LoRA. Aiming to resemble the learning capacity of FT from the findings, we propose Weight-Decomposed Low-Rank Adaptation (DoRA). DoRA decomposes the pre-trained weight into two components, magnitude and direction, for fine-tuning, specifically employing LoRA for directional updates to efficiently minimize the number of trainable parameters. By employing \\ours, we enhance both the learning capacity and training stability of LoRA while avoiding any additional inference overhead. \\ours~consistently outperforms LoRA on fine-tuning LLaMA, LLaVA, and VL-BART on various downstream tasks, such as commonsense reasoning, visual instruction tuning, and image/video-text understanding. Code is available at https://github.com/NVlabs/DoRA."
                },
                {
                    "title": "DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models",
                    "abstract": "In the era of large language models, Mixture-of-Experts (MoE) is a promising architecture for managing computational costs when scaling up model parameters. However, conventional MoE architectures like GShard, which activate the top-$K$ out of $N$ experts, face challenges in ensuring expert specialization, i.e. each expert acquires non-overlapping and focused knowledge. In response, we propose the DeepSeekMoE architecture towards ultimate expert specialization. It involves two principal strategies: (1) finely segmenting the experts into $mN$ ones and activating $mK$ from them, allowing for a more flexible combination of activated experts; (2) isolating $K_s$ experts as shared ones, aiming at capturing common knowledge and mitigating redundancy in routed experts. Starting from a modest scale with 2B parameters, we demonstrate that DeepSeekMoE 2B achieves comparable performance with GShard 2.9B, which has 1.5 times the expert parameters and computation. In addition, DeepSeekMoE 2B nearly approaches the performance of its dense counterpart with the same number of total parameters, which set the upper bound of MoE models. Subsequently, we scale up DeepSeekMoE to 16B parameters and show that it achieves comparable performance with LLaMA2 7B, with only about 40% of computations. Further, our preliminary efforts to scale up DeepSeekMoE to 145B parameters consistently validate its substantial advantages over the GShard architecture, and show its performance comparable with DeepSeek 67B, using only 28.5% (maybe even 18.2%) of computations."
                },
                {
                    "title": "Mixtral of Experts",
                    "abstract": "We introduce Mixtral 8x7B, a Sparse Mixture of Experts (SMoE) language model. Mixtral has the same architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks (i.e. experts). For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, each token has access to 47B parameters, but only uses 13B active parameters during inference. Mixtral was trained with a context size of 32k tokens and it outperforms or matches Llama 2 70B and GPT-3.5 across all evaluated benchmarks. In particular, Mixtral vastly outperforms Llama 2 70B on mathematics, code generation, and multilingual benchmarks. We also provide a model fine-tuned to follow instructions, Mixtral 8x7B - Instruct, that surpasses GPT-3.5 Turbo, Claude-2.1, Gemini Pro, and Llama 2 70B - chat model on human benchmarks. Both the base and instruct models are released under the Apache 2.0 license."
                },
                {
                    "title": "Model Breadcrumbs: Scaling Multi-Task Model Merging with Sparse Masks",
                    "abstract": "The rapid development of AI systems has been greatly influenced by the emergence of foundation models. A common approach for targeted problems involves fine-tuning these pre-trained foundation models for specific target tasks, resulting in a rapid spread of models fine-tuned across a diverse array of tasks. This work focuses on the problem of merging multiple fine-tunings of the same foundation model derived from a spectrum of auxiliary tasks. We introduce a new simple method, Model Breadcrumbs, which consists of a sparsely defined weight set that guides model adaptation within the weight space of a pre-trained model. These breadcrumbs are constructed by subtracting the weights from a pre-trained model before and after fine-tuning, followed by a sparsification process that eliminates weight outliers and negligible perturbations. Our experiments demonstrate the effectiveness of Model Breadcrumbs to simultaneously improve performance across multiple tasks. This contribution aligns with the evolving paradigm of updatable machine learning, reminiscent of the collaborative principles underlying open-source software development, fostering a community-driven effort to reliably update machine learning models. Our method is shown to be more efficient and unlike previous proposals does not require hyperparameter tuning for each new task added. Through extensive experimentation involving various models, tasks, and modalities we establish that integrating Model Breadcrumbs offers a simple, efficient, and highly effective approach for constructing multi-task models and facilitating updates to foundation models."
                },
                {
                    "title": "EE-LLM: Large-Scale Training and Inference of Early-Exit Large Language Models with 3D Parallelism",
                    "abstract": "We present EE-LLM, a framework for large-scale training and inference of early-exit large language models (LLMs). While recent works have shown preliminary evidence for the efficacy of early exiting in accelerating LLM inference, EE-LLM makes a foundational step towards scaling up early-exit LLMs by supporting their training and inference with massive 3D parallelism. Built upon Megatron-LM, EE-LLM implements a variety of algorithmic innovations and performance optimizations tailored to early exiting, including a lightweight method that facilitates backpropagation for the early-exit training objective with pipeline parallelism, techniques of leveraging idle resources in the original pipeline schedule for computation related to early-exit layers, and two approaches of early-exit inference that are compatible with KV caching for autoregressive generation. Our analytical and empirical study shows that EE-LLM achieves great training efficiency with negligible computational overhead compared to standard LLM training, as well as outstanding inference speedup without compromising output quality. To facilitate further research and adoption, we release EE-LLM at https://github.com/pan-x-c/EE-LLM."
                },
                {
                    "title": "Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch",
                    "abstract": "In this paper, we unveil that Language Models (LMs) can acquire new capabilities by assimilating parameters from homologous models without retraining or GPUs. We first introduce DARE to set most delta parameters (i.e., the disparity between fine-tuned and pre-trained parameters) to zeros without affecting the abilities of Supervised Fine-Tuning (SFT) LMs, which randomly Drops delta parameters with a ratio $p$ And REscales the remaining ones by $1 / (1 - p)$ to approximate the original embeddings. Then, we use DARE as a versatile plug-in to sparsify delta parameters of multiple SFT homologous models for mitigating parameter interference and merge them into a single model by parameter fusing. We experiment with encoder- and decoder-based LMs, showing that: (1) SFT delta parameter value ranges are typically small (within 0.002) with extreme redundancy, and DARE can effortlessly eliminate 90% or even 99% of them; (2) DARE can merge multiple task-specific LMs into one LM with diverse capabilities. Notably, this phenomenon is more pronounced in large-scale LMs, where the merged LM reveals the potential to surpass the performance of any source LM, providing a new discovery. We also utilize DARE to create a merged LM that ranks first among models with 7 billion parameters on the Open LLM Leaderboard."
                },
                {
                    "title": "QLLM: Accurate and Efficient Low-Bitwidth Quantization for Large Language Models",
                    "abstract": "Large Language Models (LLMs) excel in NLP, but their demands hinder their widespread deployment. While Quantization-Aware Training (QAT) offers a solution, its extensive training costs make Post-Training Quantization (PTQ) a more practical approach for LLMs. In existing studies, activation outliers in particular channels are identified as the bottleneck to PTQ accuracy. They propose to transform the magnitudes from activations to weights, which however offers limited alleviation or suffers from unstable gradients, resulting in a severe performance drop at low-bitwidth. In this paper, we propose QLLM, an accurate and efficient low-bitwidth PTQ method designed for LLMs. QLLM introduces an adaptive channel reassembly technique that reallocates the magnitude of outliers to other channels, thereby mitigating their impact on the quantization range. This is achieved by channel disassembly and channel assembly, which first breaks down the outlier channels into several sub-channels to ensure a more balanced distribution of activation magnitudes. Then similar channels are merged to maintain the original channel number for efficiency. Additionally, an adaptive strategy is designed to autonomously determine the optimal number of sub-channels for channel disassembly. To further compensate for the performance loss caused by quantization, we propose an efficient tuning method that only learns a small number of low-rank weights while freezing the pre-trained quantized model. After training, these low-rank parameters can be fused into the frozen weights without affecting inference. Extensive experiments on LLaMA-1 and LLaMA-2 show that QLLM can obtain accurate quantized models efficiently. For example, QLLM quantizes the 4-bit LLaMA-2-70B within 10 hours on a single A100-80G GPU, outperforming the previous state-of-the-art method by 7.89% on the average accuracy across five zero-shot tasks."
                },
                {
                    "title": "Model Tells You What to Discard: Adaptive KV Cache Compression for LLMs",
                    "abstract": "In this study, we introduce adaptive KV cache compression, a plug-and-play method that reduces the memory footprint of generative inference for Large Language Models (LLMs). Different from the conventional KV cache that retains key and value vectors for all context tokens, we conduct targeted profiling to discern the intrinsic structure of attention modules. Based on the recognized structure, we then construct the KV cache in an adaptive manner: evicting long-range contexts on attention heads emphasizing local contexts, discarding non-special tokens on attention heads centered on special tokens, and only employing the standard KV cache for attention heads that broadly attend to all tokens. Moreover, with the lightweight attention profiling used to guide the construction of the adaptive KV cache, FastGen can be deployed without resource-intensive fine-tuning or re-training. In our experiments across various asks, FastGen demonstrates substantial reduction on GPU memory consumption with negligible generation quality loss. We will release our code and the compatible CUDA kernel for reproducibility."
                },
                {
                    "title": "Efficient Streaming Language Models with Attention Sinks",
                    "abstract": "Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching previous tokens' Key and Value states (KV) consumes extensive memory. Secondly, popular LLMs cannot generalize to longer texts than the training sequence length. Window attention, where only the most recent KVs are cached, is a natural approach -- but we show that it fails when the text length surpasses the cache size. We observe an interesting phenomenon, namely attention sink, that keeping the KV of initial tokens will largely recover the performance of window attention. In this paper, we first demonstrate that the emergence of attention sink is due to the strong attention scores towards initial tokens as a\"sink\"even if they are not semantically important. Based on the above analysis, we introduce StreamingLLM, an efficient framework that enables LLMs trained with a finite length attention window to generalize to infinite sequence lengths without any fine-tuning. We show that StreamingLLM can enable Llama-2, MPT, Falcon, and Pythia to perform stable and efficient language modeling with up to 4 million tokens and more. In addition, we discover that adding a placeholder token as a dedicated attention sink during pre-training can further improve streaming deployment. In streaming settings, StreamingLLM outperforms the sliding window recomputation baseline by up to 22.2x speedup. Code and datasets are provided at https://github.com/mit-han-lab/streaming-llm."
                },
                {
                    "title": "Efficient Memory Management for Large Language Model Serving with PagedAttention",
                    "abstract": "High throughput serving of large language models (LLMs) requires batching sufficiently many requests at a time. However, existing systems struggle because the key-value cache (KV cache) memory for each request is huge and grows and shrinks dynamically. When managed inefficiently, this memory can be significantly wasted by fragmentation and redundant duplication, limiting the batch size. To address this problem, we propose PagedAttention, an attention algorithm inspired by the classical virtual memory and paging techniques in operating systems. On top of it, we build vLLM, an LLM serving system that achieves (1) near-zero waste in KV cache memory and (2) flexible sharing of KV cache within and across requests to further reduce memory usage. Our evaluations show that vLLM improves the throughput of popular LLMs by 2--4\u00d7 with the same level of latency compared to the state-of-the-art systems, such as FasterTransformer and Orca. The improvement is more pronounced with longer sequences, larger models, and more complex decoding algorithms. vLLM's source code is publicly available at https://github.com/vllm-project/vllm."
                },
                {
                    "title": "LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding",
                    "abstract": "Although large language models (LLMs) demonstrate impressive performance for many language tasks, most of them can only handle texts a few thousand tokens long, limiting their applications on longer sequence inputs, such as books, reports, and codebases. Recent works have proposed methods to improve LLMs' long context capabilities by extending context windows and more sophisticated memory mechanisms. However, comprehensive benchmarks tailored for evaluating long context understanding are lacking. In this paper, we introduce LongBench, the first bilingual, multi-task benchmark for long context understanding, enabling a more rigorous evaluation of long context understanding. LongBench comprises 21 datasets across 6 task categories in both English and Chinese, with an average length of 6,711 words (English) and 13,386 characters (Chinese). These tasks cover key long-text application areas including single-doc QA, multi-doc QA, summarization, few-shot learning, synthetic tasks, and code completion. All datasets in LongBench are standardized into a unified format, allowing for effortless automatic evaluation of LLMs. Upon comprehensive evaluation of 8 LLMs on LongBench, we find that: (1) Commercial model (GPT-3.5-Turbo-16k) outperforms other open-sourced models, but still struggles on longer contexts. (2) Scaled position embedding and fine-tuning on longer sequences lead to substantial improvement on long context understanding. (3) Context compression technique such as retrieval brings improvement for model with weak ability on long contexts, but the performance still lags behind models that have strong long context understanding capability. The code and datasets are available at https://github.com/THUDM/LongBench."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning",
                    "abstract": "Scaling Transformers to longer sequence lengths has been a major problem in the last several years, promising to improve performance in language modeling and high-resolution image understanding, as well as to unlock new applications in code, audio, and video generation. The attention layer is the main bottleneck in scaling to longer sequences, as its runtime and memory increase quadratically in the sequence length. FlashAttention exploits the asymmetric GPU memory hierarchy to bring significant memory saving (linear instead of quadratic) and runtime speedup (2-4$\\times$ compared to optimized baselines), with no approximation. However, FlashAttention is still not nearly as fast as optimized matrix-multiply (GEMM) operations, reaching only 25-40\\% of the theoretical maximum FLOPs/s. We observe that the inefficiency is due to suboptimal work partitioning between different thread blocks and warps on the GPU, causing either low-occupancy or unnecessary shared memory reads/writes. We propose FlashAttention-2, with better work partitioning to address these issues. In particular, we (1) tweak the algorithm to reduce the number of non-matmul FLOPs (2) parallelize the attention computation, even for a single head, across different thread blocks to increase occupancy, and (3) within each thread block, distribute the work between warps to reduce communication through shared memory. These yield around 2$\\times$ speedup compared to FlashAttention, reaching 50-73\\% of the theoretical maximum FLOPs/s on A100 and getting close to the efficiency of GEMM operations. We empirically validate that when used end-to-end to train GPT-style models, FlashAttention-2 reaches training speed of up to 225 TFLOPs/s per A100 GPU (72\\% model FLOPs utilization)."
                },
                {
                    "title": "SkipDecode: Autoregressive Skip Decoding with Batching and Caching for Efficient LLM Inference",
                    "abstract": "Autoregressive large language models (LLMs) have made remarkable progress in various natural language generation tasks. However, they incur high computation cost and latency resulting from the autoregressive token-by-token generation. To address this issue, several approaches have been proposed to reduce computational cost using early-exit strategies. These strategies enable faster text generation using reduced computation without applying the full computation graph to each token. While existing token-level early exit methods show promising results for online inference, they cannot be readily applied for batch inferencing and Key-Value caching. This is because they have to wait until the last token in a batch exits before they can stop computing. This severely limits the practical application of such techniques. In this paper, we propose a simple and effective token-level early exit method, SkipDecode, designed to work seamlessly with batch inferencing and KV caching. It overcomes prior constraints by setting up a singular exit point for every token in a batch at each sequence position. It also guarantees a monotonic decrease in exit points, thereby eliminating the need to recompute KV Caches for preceding tokens. Rather than terminating computation prematurely as in prior works, our approach bypasses lower to middle layers, devoting most of the computational resources to upper layers, allowing later tokens to benefit from the compute expenditure by earlier tokens. Our experimental results show that SkipDecode can obtain 2x to 5x inference speedups with negligible regression across a variety of tasks. This is achieved using OPT models of 1.3 billion and 6.7 billion parameters, all the while being directly compatible with batching and KV caching optimization techniques."
                },
                {
                    "title": "H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models",
                    "abstract": "Large Language Models (LLMs), despite their recent impressive accomplishments, are notably cost-prohibitive to deploy, particularly for applications involving long-content generation, such as dialogue systems and story writing. Often, a large amount of transient state information, referred to as the KV cache, is stored in GPU memory in addition to model parameters, scaling linearly with the sequence length and batch size. In this paper, we introduce a novel approach for implementing the KV cache which significantly reduces its memory footprint. Our approach is based on the noteworthy observation that a small portion of tokens contributes most of the value when computing attention scores. We call these tokens Heavy Hitters (H$_2$). Through a comprehensive investigation, we find that (i) the emergence of H$_2$ is natural and strongly correlates with the frequent co-occurrence of tokens in the text, and (ii) removing them results in significant performance degradation. Based on these insights, we propose Heavy Hitter Oracle (H$_2$O), a KV cache eviction policy that dynamically retains a balance of recent and H$_2$ tokens. We formulate the KV cache eviction as a dynamic submodular problem and prove (under mild assumptions) a theoretical guarantee for our novel eviction algorithm which could help guide future work. We validate the accuracy of our algorithm with OPT, LLaMA, and GPT-NeoX across a wide range of tasks. Our implementation of H$_2$O with 20% heavy hitters improves the throughput over three leading inference systems DeepSpeed Zero-Inference, Hugging Face Accelerate, and FlexGen by up to 29$\\times$, 29$\\times$, and 3$\\times$ on OPT-6.7B and OPT-30B. With the same batch size, H2O can reduce the latency by up to 1.9$\\times$. The code is available at https://github.com/FMInference/H2O."
                },
                {
                    "title": "TIES-Merging: Resolving Interference When Merging Models",
                    "abstract": "Transfer learning - i.e., further fine-tuning a pre-trained model on a downstream task - can confer significant advantages, including improved downstream performance, faster convergence, and better sample efficiency. These advantages have led to a proliferation of task-specific fine-tuned models, which typically can only perform a single task and do not benefit from one another. Recently, model merging techniques have emerged as a solution to combine multiple task-specific models into a single multitask model without performing additional training. However, existing merging methods often ignore the interference between parameters of different models, resulting in large performance drops when merging multiple models. In this paper, we demonstrate that prior merging techniques inadvertently lose valuable information due to two major sources of interference: (a) interference due to redundant parameter values and (b) disagreement on the sign of a given parameter's values across models. To address this, we propose our method, TRIM, ELECT SIGN&MERGE (TIES-Merging), which introduces three novel steps when merging models: (1) resetting parameters that only changed a small amount during fine-tuning, (2) resolving sign conflicts, and (3) merging only the parameters that are in alignment with the final agreed-upon sign. We find that TIES-Merging outperforms several existing methods in diverse settings covering a range of modalities, domains, number of tasks, model sizes, architectures, and fine-tuning settings. We further analyze the impact of different types of interference on model parameters, and highlight the importance of resolving sign interference. Our code is available at https://github.com/prateeky2806/ties-merging"
                },
                {
                    "title": "AWQ: Activation-aware Weight Quantization for On-Device LLM Compression and Acceleration",
                    "abstract": "Large language models (LLMs) have transformed numerous AI applications. On-device LLM is becoming increasingly important: running LLMs locally on edge devices can reduce the cloud computing cost and protect users' privacy. However, the astronomical model size and the limited hardware resource pose significant deployment challenges. We propose Activation-aware Weight Quantization (AWQ), a hardware-friendly approach for LLM low-bit weight-only quantization. AWQ finds that not all weights in an LLM are equally important. Protecting only 1% salient weights can greatly reduce quantization error. To identify salient weight channels, we should refer to the activation distribution, not weights. To avoid the hardware-inefficient mix-precision quantization, we mathematically derive that scaling up the salient channels can reduce the quantization error. AWQ employs an equivalent transformation to scale the salient weight channels to protect them. The scale is determined by collecting the activation statistics offline. AWQ does not rely on any backpropagation or reconstruction, so it generalizes to different domains and modalities without overfitting the calibration set. AWQ outperforms existing work on various language modeling and domain-specific benchmarks (coding and math). Thanks to better generalization, it achieves excellent quantization performance for instruction-tuned LMs and, for the first time, multi-modal LMs. Alongside AWQ, we implement TinyChat, an efficient and flexible inference framework tailored for 4-bit on-device LLM/VLMs. With kernel fusion and platform-aware weight packing, TinyChat offers more than 3x speedup over the Huggingface FP16 implementation on both desktop and mobile GPUs. It also democratizes the deployment of the 70B Llama-2 model on mobile GPUs."
                },
                {
                    "title": "LoRAPrune: Structured Pruning Meets Low-Rank Parameter-Efficient Fine-Tuning",
                    "abstract": "Large Language Models (LLMs), such as LLaMA and T5, have shown exceptional performance across various tasks through fine-tuning. Although low-rank adaption (LoRA) has emerged to cheaply fine-tune these LLMs on downstream tasks, their deployment is still hindered by the vast model scale and computational costs. Post-training model pruning offers a way to compress LLMs. However, the current pruning methods designed for LLMs are not compatible with LoRA. This is due to their utilization of unstructured pruning on LLMs, impeding the merging of LoRA weights, or their dependence on the gradients of pre-trained weights to guide pruning, which can impose significant memory overhead. To this end, we propose LoRAPrune, a new framework that delivers an accurate structured pruned model in a highly memory-efficient manner. Specifically, we first design a LoRA-guided pruning criterion, which uses the weights and gradients of LoRA, rather than the gradients of pre-trained weights for importance estimation. We subsequently integrate this criterion into an iterative pruning process, effectively removing redundant channels and heads. Extensive experimental results demonstrate the superior performance of our LoRAPrune over existing approaches on the LLaMA series models. At a 50\\% compression rate, LoRAPrune demonstrates superior performance over LLM-Pruner, achieving a reduction in perplexity by 4.81 on WikiText2 and 3.46 on PTB, while also decreasing memory usage by 52.6%. Besides, LoRAPrune also matches semi-structural pruning across multiple LLMs, proving its wide applicability. The code is available at https://github.com/aim-uofa/LoRAPrune."
                },
                {
                    "title": "QLoRA: Efficient Finetuning of Quantized LLMs",
                    "abstract": "We present QLoRA, an efficient finetuning approach that reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU while preserving full 16-bit finetuning task performance. QLoRA backpropagates gradients through a frozen, 4-bit quantized pretrained language model into Low Rank Adapters~(LoRA). Our best model family, which we name Guanaco, outperforms all previous openly released models on the Vicuna benchmark, reaching 99.3% of the performance level of ChatGPT while only requiring 24 hours of finetuning on a single GPU. QLoRA introduces a number of innovations to save memory without sacrificing performance: (a) 4-bit NormalFloat (NF4), a new data type that is information theoretically optimal for normally distributed weights (b) double quantization to reduce the average memory footprint by quantizing the quantization constants, and (c) paged optimziers to manage memory spikes. We use QLoRA to finetune more than 1,000 models, providing a detailed analysis of instruction following and chatbot performance across 8 instruction datasets, multiple model types (LLaMA, T5), and model scales that would be infeasible to run with regular finetuning (e.g. 33B and 65B parameter models). Our results show that QLoRA finetuning on a small high-quality dataset leads to state-of-the-art results, even when using smaller models than the previous SoTA. We provide a detailed analysis of chatbot performance based on both human and GPT-4 evaluations showing that GPT-4 evaluations are a cheap and reasonable alternative to human evaluation. Furthermore, we find that current chatbot benchmarks are not trustworthy to accurately evaluate the performance levels of chatbots. A lemon-picked analysis demonstrates where Guanaco fails compared to ChatGPT. We release all of our models and code, including CUDA kernels for 4-bit training."
                },
                {
                    "title": "GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints",
                    "abstract": "Multi-query attention (MQA), which only uses a single key-value head, drastically speeds up decoder inference. However, MQA can lead to quality degradation, and moreover it may not be desirable to train a separate model just for faster inference. We (1) propose a recipe for uptraining existing multi-head language model checkpoints into models with MQA using 5% of original pre-training compute, and (2) introduce grouped-query attention (GQA), a generalization of multi-query attention which uses an intermediate (more than one, less than number of query heads) number of key-value heads. We show that uptrained GQA achieves quality close to multi-head attention with comparable speed to MQA."
                },
                {
                    "title": "Joint Token Pruning and Squeezing Towards More Aggressive Compression of Vision Transformers",
                    "abstract": "Although vision transformers (ViTs) have shown promising results in various computer vision tasks recently, their high computational cost limits their practical applications. Previous approaches that prune redundant tokens have demonstrated a good trade-off between performance and computation costs. Nevertheless, errors caused by pruning strategies can lead to significant information loss. Our quantitative experiments reveal that the impact of pruned tokens on performance should be noticeable. To address this issue, we propose a novel joint Token Pruning & Squeezing module (TPS) for compressing vision transformers with higher efficiency. Firstly, TPS adopts pruning to get the reserved and pruned subsets. Secondly, TPS squeezes the information of pruned tokens into partial reserved tokens via the unidirectional nearest-neighbor matching and similarity-based fusing steps. Compared to state-of-the-art methods, our approach outperforms them under all token pruning intensities. Especially while shrinking DeiT-tiny&small computational budgets to 35%, it improves the accuracy by 1%-6% compared with baselines on ImageNet classification. The proposed method can accelerate the throughput of DeiT-small beyond DeiT-tiny, while its accuracy surpasses DeiT-tiny by 4.78%. Experiments on various transformers demonstrate the effectiveness of our method, while analysis experiments prove our higher robustness to the errors of the token pruning policy. Code is available at https://github.com/megvii-research/TPS-CVPR2023."
                },
                {
                    "title": "High-throughput Generative Inference of Large Language Models with a Single GPU",
                    "abstract": "The high computational and memory requirements of large language model (LLM) inference make it feasible only with multiple high-end accelerators. Motivated by the emerging demand for latency-insensitive tasks with batched processing, this paper initiates the study of high-throughput LLM inference using limited resources, such as a single commodity GPU. We present FlexGen, a high-throughput generation engine for running LLMs with limited GPU memory. FlexGen can be flexibly configured under various hardware resource constraints by aggregating memory and computation from the GPU, CPU, and disk. By solving a linear programming problem, it searches for efficient patterns to store and access tensors. FlexGen further compresses the weights and the attention cache to 4 bits with negligible accuracy loss. These techniques enable FlexGen to have a larger space of batch size choices and thus significantly increase maximum throughput. As a result, when running OPT-175B on a single 16GB GPU, FlexGen achieves significantly higher throughput compared to state-of-the-art offloading systems, reaching a generation throughput of 1 token/s for the first time with an effective batch size of 144. On the HELM benchmark, FlexGen can benchmark a 30B model with a 16GB GPU on 7 representative sub-scenarios in 21 hours. The code is available at https://github.com/FMInference/FlexGen"
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Dataless Knowledge Fusion by Merging Weights of Language Models",
                    "abstract": "Fine-tuning pre-trained language models has become the prevalent paradigm for building downstream NLP models. Oftentimes fine-tuned models are readily available but their training data is not, due to data privacy or intellectual property concerns. This creates a barrier to fusing knowledge across individual models to yield a better single model. In this paper, we study the problem of merging individual models built on different training data sets to obtain a single model that performs well both across all data set domains and can generalize on out-of-domain data. We propose a dataless knowledge fusion method that merges models in their parameter space, guided by weights that minimize prediction differences between the merged model and the individual models. Over a battery of evaluation settings, we show that the proposed method significantly outperforms baselines such as Fisher-weighted averaging or model ensembling. Further, we find that our method is a promising alternative to multi-task learning that can preserve or sometimes improve over the individual models without access to the training data. Finally, model merging is more efficient than training a multi-task model, thus making it applicable to a wider set of scenarios."
                },
                {
                    "title": "SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models",
                    "abstract": "Large language models (LLMs) show excellent performance but are compute- and memory-intensive. Quantization can reduce memory and accelerate inference. However, existing methods cannot maintain accuracy and hardware efficiency at the same time. We propose SmoothQuant, a training-free, accuracy-preserving, and general-purpose post-training quantization (PTQ) solution to enable 8-bit weight, 8-bit activation (W8A8) quantization for LLMs. Based on the fact that weights are easy to quantize while activations are not, SmoothQuant smooths the activation outliers by offline migrating the quantization difficulty from activations to weights with a mathematically equivalent transformation. SmoothQuant enables an INT8 quantization of both weights and activations for all the matrix multiplications in LLMs, including OPT, BLOOM, GLM, MT-NLG, Llama-1/2, Falcon, Mistral, and Mixtral models. We demonstrate up to 1.56x speedup and 2x memory reduction for LLMs with negligible loss in accuracy. SmoothQuant enables serving 530B LLM within a single node. Our work offers a turn-key solution that reduces hardware costs and democratizes LLMs. Code is available at https://github.com/mit-han-lab/smoothquant."
                },
                {
                    "title": "Efficiently Scaling Transformer Inference",
                    "abstract": "We study the problem of efficient generative inference for Transformer models, in one of its most challenging settings: large deep models, with tight latency targets and long sequence lengths. Better understanding of the engineering tradeoffs for inference for large Transformer-based models is important as use cases of these models are growing rapidly throughout application areas. We develop a simple analytical model for inference efficiency to select the best multi-dimensional partitioning techniques optimized for TPU v4 slices based on the application requirements. We combine these with a suite of low-level optimizations to achieve a new Pareto frontier on the latency and model FLOPS utilization (MFU) tradeoffs on 500B+ parameter models that outperforms the FasterTransformer suite of benchmarks. We further show that with appropriate partitioning, the lower memory requirements of multiquery attention (i.e. multiple query heads share single key/value head) enables scaling up to 32x larger context lengths. Finally, we achieve a low-batch-size latency of 29ms per token during generation (using int8 weight quantization) and a 76% MFU during large-batch-size processing of input tokens, while supporting a long 2048-token context length on the PaLM 540B parameter model."
                },
                {
                    "title": "Git Re-Basin: Merging Models modulo Permutation Symmetries",
                    "abstract": "The success of deep learning is due in large part to our ability to solve certain massive non-convex optimization problems with relative ease. Though non-convex optimization is NP-hard, simple algorithms -- often variants of stochastic gradient descent -- exhibit surprising effectiveness in fitting large neural networks in practice. We argue that neural network loss landscapes often contain (nearly) a single basin after accounting for all possible permutation symmetries of hidden units a la Entezari et al. 2021. We introduce three algorithms to permute the units of one model to bring them into alignment with a reference model in order to merge the two models in weight space. This transformation produces a functionally equivalent set of weights that lie in an approximately convex basin near the reference model. Experimentally, we demonstrate the single basin phenomenon across a variety of model architectures and datasets, including the first (to our knowledge) demonstration of zero-barrier linear mode connectivity between independently trained ResNet models on CIFAR-10. Additionally, we identify intriguing phenomena relating model width and training time to mode connectivity. Finally, we discuss shortcomings of the linear mode connectivity hypothesis, including a counterexample to the single basin theory."
                },
                {
                    "title": "LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale",
                    "abstract": "Large language models have been widely adopted but require significant GPU memory for inference. We develop a procedure for Int8 matrix multiplication for feed-forward and attention projection layers in transformers, which cut the memory needed for inference by half while retaining full precision performance. With our method, a 175B parameter 16/32-bit checkpoint can be loaded, converted to Int8, and used immediately without performance degradation. This is made possible by understanding and working around properties of highly systematic emergent features in transformer language models that dominate attention and transformer predictive performance. To cope with these features, we develop a two-part quantization procedure, LLM.int8(). We first use vector-wise quantization with separate normalization constants for each inner product in the matrix multiplication, to quantize most of the features. However, for the emergent outliers, we also include a new mixed-precision decomposition scheme, which isolates the outlier feature dimensions into a 16-bit matrix multiplication while still more than 99.9% of values are multiplied in 8-bit. Using LLM.int8(), we show empirically it is possible to perform inference in LLMs with up to 175B parameters without any performance degradation. This result makes such models much more accessible, for example making it possible to use OPT-175B/BLOOM on a single server with consumer GPUs. We open-source our software."
                },
                {
                    "title": "Confident Adaptive Language Modeling",
                    "abstract": "Recent advances in Transformer-based large language models (LLMs) have led to significant performance improvements across many tasks. These gains come with a drastic increase in the models' size, potentially leading to slow and costly use at inference time. In practice, however, the series of generations made by LLMs is composed of varying levels of difficulty. While certain predictions truly benefit from the models' full capacity, other continuations are more trivial and can be solved with reduced compute. In this work, we introduce Confident Adaptive Language Modeling (CALM), a framework for dynamically allocating different amounts of compute per input and generation timestep. Early exit decoding involves several challenges that we address here, such as: (1) what confidence measure to use; (2) connecting sequence-level constraints to local per-token exit decisions; and (3) attending back to missing hidden representations due to early exits in previous tokens. Through theoretical analysis and empirical experiments on three diverse text generation tasks, we demonstrate the efficacy of our framework in reducing compute -- potential speedup of up to $\\times 3$ -- while provably maintaining high performance."
                },
                {
                    "title": "Tutel: Adaptive Mixture-of-Experts at Scale",
                    "abstract": "Sparsely-gated mixture-of-experts (MoE) has been widely adopted to scale deep learning models to trillion-plus parameters with fixed computational cost. The algorithmic performance of MoE relies on its token routing mechanism that forwards each input token to the right sub-models or experts. While token routing dynamically determines the amount of expert workload at runtime, existing systems suffer inefficient computation due to their static execution, namely static parallelism and pipelining, which does not adapt to the dynamic workload. We present Flex, a highly scalable stack design and implementation for MoE with dynamically adaptive parallelism and pipelining. Flex designs an identical layout for distributing MoE model parameters and input data, which can be leveraged by all possible parallelism or pipelining methods without any mathematical inequivalence or tensor migration overhead. This enables adaptive parallelism/pipelining optimization at zero cost during runtime. Based on this key design, Flex also implements various MoE acceleration techniques. Aggregating all techniques, Flex finally delivers huge speedup at any scale -- 4.96x and 5.75x speedup of a single MoE layer over 16 and 2,048 A100 GPUs, respectively, over the previous state-of-the-art. Our evaluation shows that Flex efficiently and effectively runs a real-world MoE-based model named SwinV2-MoE, built upon Swin Transformer V2, a state-of-the-art computer vision architecture. On efficiency, Flex accelerates SwinV2-MoE, achieving up to 1.55x and 2.11x speedup in training and inference over Fairseq, respectively. On effectiveness, the SwinV2-MoE model achieves superior accuracy in both pre-training and down-stream computer vision tasks such as COCO object detection than the counterpart dense model, indicating the readiness of Flex for end-to-end real-world model training and inference."
                },
                {
                    "title": "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness",
                    "abstract": "Transformers are slow and memory-hungry on long sequences, since the time and memory complexity of self-attention are quadratic in sequence length. Approximate attention methods have attempted to address this problem by trading off model quality to reduce the compute complexity, but often do not achieve wall-clock speedup. We argue that a missing principle is making attention algorithms IO-aware -- accounting for reads and writes between levels of GPU memory. We propose FlashAttention, an IO-aware exact attention algorithm that uses tiling to reduce the number of memory reads/writes between GPU high bandwidth memory (HBM) and GPU on-chip SRAM. We analyze the IO complexity of FlashAttention, showing that it requires fewer HBM accesses than standard attention, and is optimal for a range of SRAM sizes. We also extend FlashAttention to block-sparse attention, yielding an approximate attention algorithm that is faster than any existing approximate attention method. FlashAttention trains Transformers faster than existing baselines: 15% end-to-end wall-clock speedup on BERT-large (seq. length 512) compared to the MLPerf 1.1 training speed record, 3$\\times$ speedup on GPT-2 (seq. length 1K), and 2.4$\\times$ speedup on long-range arena (seq. length 1K-4K). FlashAttention and block-sparse FlashAttention enable longer context in Transformers, yielding higher quality models (0.7 better perplexity on GPT-2 and 6.4 points of lift on long-document classification) and entirely new capabilities: the first Transformers to achieve better-than-chance performance on the Path-X challenge (seq. length 16K, 61.4% accuracy) and Path-256 (seq. length 64K, 63.1% accuracy)."
                },
                {
                    "title": "Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time",
                    "abstract": "The conventional recipe for maximizing model accuracy is to (1) train multiple models with various hyperparameters and (2) pick the individual model which performs best on a held-out validation set, discarding the remainder. In this paper, we revisit the second step of this procedure in the context of fine-tuning large pre-trained models, where fine-tuned models often appear to lie in a single low error basin. We show that averaging the weights of multiple models fine-tuned with different hyperparameter configurations often improves accuracy and robustness. Unlike a conventional ensemble, we may average many models without incurring any additional inference or memory costs -- we call the results\"model soups.\"When fine-tuning large pre-trained models such as CLIP, ALIGN, and a ViT-G pre-trained on JFT, our soup recipe provides significant improvements over the best model in a hyperparameter sweep on ImageNet. The resulting ViT-G model, which attains 90.94% top-1 accuracy on ImageNet, achieved a new state of the art. Furthermore, we show that the model soup approach extends to multiple image classification and natural language processing tasks, improves out-of-distribution performance, and improves zero-shot performance on new downstream tasks. Finally, we analytically relate the performance similarity of weight-averaging and logit-ensembling to flatness of the loss and confidence of the predictions, and validate this relation empirically. Code is available at https://github.com/mlfoundations/model-soups."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Merging Models with Fisher-Weighted Averaging",
                    "abstract": "Averaging the parameters of models that have the same architecture and initialization can provide a means of combining their respective capabilities. In this paper, we take the perspective that this\"merging\"operation can be seen as choosing parameters that approximately maximize the joint likelihood of the posteriors of the models' parameters. Computing a simple average of the models' parameters therefore corresponds to making an isotropic Gaussian approximation to their posteriors. We develop an alternative merging procedure based on the Laplace approximation where we approximate each model's posterior as a Gaussian distribution whose precision matrix corresponds to its Fisher information. We first show that our\"Fisher merging\"technique provides a performance boost in settings where simple parameter averaging is currently used -- specifically, robust fine-tuning and model ensembling. Then, we compare merging to standard gradient-based transfer learning and demonstrate that merging enables a fundamentally different method for transferring capabilities across models. Specifically, we show that Fisher merging is competitive with gradient-based transfer learning approaches (while being significantly cheaper) in intermediate-task training and domain-adaptive pre-training. We also show that our merging procedure makes it possible to combine models in previously unexplored ways. We release our code to facilitate future research into methods for merging models."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "The Role of Permutation Invariance in Linear Mode Connectivity of Neural Networks",
                    "abstract": "In this paper, we conjecture that if the permutation invariance of neural networks is taken into account, SGD solutions will likely have no barrier in the linear interpolation between them. Although it is a bold conjecture, we show how extensive empirical attempts fall short of refuting it. We further provide a preliminary theoretical result to support our conjecture. Our conjecture has implications for lottery ticket hypothesis, distributed training, and ensemble methods."
                },
                {
                    "title": "TruthfulQA: Measuring How Models Mimic Human Falsehoods",
                    "abstract": "We propose a benchmark to measure whether a language model is truthful in generating answers to questions. The benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics. We crafted questions that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. We tested GPT-3, GPT-Neo/J, GPT-2 and a T5-based model. The best model was truthful on 58% of questions, while human performance was 94%. Models generated many false answers that mimic popular misconceptions and have the potential to deceive humans. The largest models were generally the least truthful. This contrasts with other NLP tasks, where performance improves with model size. However, this result is expected if false answers are learned from the training distribution. We suggest that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web."
                },
                {
                    "title": "Luna: Linear Unified Nested Attention",
                    "abstract": "The quadratic computational and memory complexities of the Transformer's attention mechanism have limited its scalability for modeling long sequences. In this paper, we propose Luna, a linear unified nested attention mechanism that approximates softmax attention with two nested linear attention functions, yielding only linear (as opposed to quadratic) time and space complexity. Specifically, with the first attention function, Luna packs the input sequence into a sequence of fixed length. Then, the packed sequence is unpacked using the second attention function. As compared to a more traditional attention mechanism, Luna introduces an additional sequence with a fixed length as input and an additional corresponding output, which allows Luna to perform attention operation linearly, while also storing adequate contextual information. We perform extensive evaluations on three benchmarks of sequence modeling tasks: long-context sequence modeling, neural machine translation and masked language modeling for large-scale pretraining. Competitive or even better experimental results demonstrate both the effectiveness and efficiency of Luna compared to a variety"
                },
                {
                    "title": "Random Feature Attention",
                    "abstract": "Transformers are state-of-the-art models for a variety of sequence modeling tasks. At their core is an attention function which models pairwise interactions between the inputs at every timestep. While attention is powerful, it does not scale efficiently to long sequences due to its quadratic time and space complexity in the sequence length. We propose RFA, a linear time and space attention that uses random feature methods to approximate the softmax function, and explore its application in transformers. RFA can be used as a drop-in replacement for conventional softmax attention and offers a straightforward way of learning with recency bias through an optional gating mechanism. Experiments on language modeling and machine translation demonstrate that RFA achieves similar or better performance compared to strong transformer baselines. In the machine translation experiment, RFA decodes twice as fast as a vanilla transformer. Compared to existing efficient transformer variants, RFA is competitive in terms of both accuracy and efficiency on three long text classification datasets. Our analysis shows that RFA's efficiency gains are especially notable on long sequences, suggesting that RFA will be particularly useful in tasks that require working with large inputs, fast decoding speed, or low memory footprints."
                },
                {
                    "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity",
                    "abstract": "In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with outrageous numbers of parameters -- but a constant computational cost. However, despite several notable successes of MoE, widespread adoption has been hindered by complexity, communication costs and training instability -- we address these with the Switch Transformer. We simplify the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs. Our proposed training techniques help wrangle the instabilities and we show large sparse models may be trained, for the first time, with lower precision (bfloat16) formats. We design models based off T5-Base and T5-Large to obtain up to 7x increases in pre-training speed with the same computational resources. These improvements extend into multilingual settings where we measure gains over the mT5-Base version across all 101 languages. Finally, we advance the current scale of language models by pre-training up to trillion parameter models on the\"Colossal Clean Crawled Corpus\"and achieve a 4x speedup over the T5-XXL model."
                },
                {
                    "title": "Rethinking Attention with Performers",
                    "abstract": "We introduce Performers, Transformer architectures which can estimate regular (softmax) full-rank-attention Transformers with provable accuracy, but using only linear (as opposed to quadratic) space and time complexity, without relying on any priors such as sparsity or low-rankness. To approximate softmax attention-kernels, Performers use a novel Fast Attention Via positive Orthogonal Random features approach (FAVOR+), which may be of independent interest for scalable kernel methods. FAVOR+ can be also used to efficiently model kernelizable attention mechanisms beyond softmax. This representational power is crucial to accurately compare softmax with other kernels for the first time on large-scale tasks, beyond the reach of regular Transformers, and investigate optimal attention-kernels. Performers are linear architectures fully compatible with regular Transformers and with strong theoretical guarantees: unbiased or nearly-unbiased estimation of the attention matrix, uniform convergence and low estimation variance. We tested Performers on a rich set of tasks stretching from pixel-prediction through text models to protein sequence modeling. We demonstrate competitive results with other examined efficient sparse and dense attention methods, showcasing effectiveness of the novel attention-learning paradigm leveraged by Performers."
                },
                {
                    "title": "GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding",
                    "abstract": "Neural network scaling has been critical for improving the model quality in many real-world machine learning applications with vast amounts of training data and compute. Although this trend of scaling is affirmed to be a sure-fire approach for better model quality, there are challenges on the path such as the computation cost, ease of programming, and efficient implementation on parallel devices. GShard is a module composed of a set of lightweight annotation APIs and an extension to the XLA compiler. It provides an elegant way to express a wide range of parallel computation patterns with minimal changes to the existing model code. GShard enabled us to scale up multilingual neural machine translation Transformer model with Sparsely-Gated Mixture-of-Experts beyond 600 billion parameters using automatic sharding. We demonstrate that such a giant model can efficiently be trained on 2048 TPU v3 accelerators in 4 days to achieve far superior quality for translation from 100 languages to English compared to the prior art."
                },
                {
                    "title": "Linformer: Self-Attention with Linear Complexity",
                    "abstract": "Large transformer models have shown extraordinary success in achieving state-of-the-art results in many natural language processing applications. However, training and deploying these models can be prohibitively costly for long sequences, as the standard self-attention mechanism of the Transformer uses $O(n^2)$ time and space with respect to sequence length. In this paper, we demonstrate that the self-attention mechanism can be approximated by a low-rank matrix. We further exploit this finding to propose a new self-attention mechanism, which reduces the overall self-attention complexity from $O(n^2)$ to $O(n)$ in both time and space. The resulting linear transformer, the \\textit{Linformer}, performs on par with standard Transformer models, while being much more memory- and time-efficient."
                },
                {
                    "title": "BERT Loses Patience: Fast and Robust Inference with Early Exit",
                    "abstract": "In this paper, we propose Patience-based Early Exit, a straightforward yet effective inference method that can be used as a plug-and-play technique to simultaneously improve the efficiency and robustness of a pretrained language model (PLM). To achieve this, our approach couples an internal-classifier with each layer of a PLM and dynamically stops inference when the intermediate predictions of the internal classifiers remain unchanged for a pre-defined number of steps. Our approach improves inference efficiency as it allows the model to make a prediction with fewer layers. Meanwhile, experimental results with an ALBERT model show that our method can improve the accuracy and robustness of the model by preventing it from overthinking and exploiting multiple classifiers for prediction, yielding a better accuracy-speed trade-off compared to existing early exit methods."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Up or Down? Adaptive Rounding for Post-Training Quantization",
                    "abstract": "When quantizing neural networks, assigning each floating-point weight to its nearest fixed-point value is the predominant approach. We find that, perhaps surprisingly, this is not the best we can do. In this paper, we propose AdaRound, a better weight-rounding mechanism for post-training quantization that adapts to the data and the task loss. AdaRound is fast, does not require fine-tuning of the network, and only uses a small amount of unlabelled data. We start by theoretically analyzing the rounding problem for a pre-trained neural network. By approximating the task loss with a Taylor series expansion, the rounding task is posed as a quadratic unconstrained binary optimization problem. We simplify this to a layer-wise local loss and propose to optimize this loss with a soft relaxation. AdaRound not only outperforms rounding-to-nearest by a significant margin but also establishes a new state-of-the-art for post-training quantization on several networks and tasks. Without fine-tuning, we can quantize the weights of Resnet18 and Resnet50 to 4 bits while staying within an accuracy loss of 1%."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Reformer: The Efficient Transformer",
                    "abstract": "Large Transformer models routinely achieve state-of-the-art results on a number of tasks but training these models can be prohibitively costly, especially on long sequences. We introduce two techniques to improve the efficiency of Transformers. For one, we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O($L^2$) to O($L\\log L$), where $L$ is the length of the sequence. Furthermore, we use reversible residual layers instead of the standard residuals, which allows storing activations only once in the training process instead of $N$ times, where $N$ is the number of layers. The resulting model, the Reformer, performs on par with Transformer models while being much more memory-efficient and much faster on long sequences."
                },
                {
                    "title": "PIQA: Reasoning about Physical Commonsense in Natural Language",
                    "abstract": "To apply eyeshadow without a brush, should I use a cotton swab or a toothpick? Questions requiring this kind of physical commonsense pose a challenge to today's natural language understanding systems. While recent pretrained models (such as BERT) have made progress on question answering over more abstract domains \u2013 such as news articles and encyclopedia entries, where text is plentiful \u2013 in more physical domains, text is inherently limited due to reporting bias. Can AI systems learn to reliably answer physical commonsense questions without experiencing the physical world?In this paper, we introduce the task of physical commonsense reasoning and a corresponding benchmark dataset Physical Interaction: Question Answering or PIQA. Though humans find the dataset easy (95% accuracy), large pretrained models struggle (\u223c75%). We provide analysis about the dimensions of knowledge that existing models lack, which offers significant opportunities for future research."
                },
                {
                    "title": "Fast Transformer Decoding: One Write-Head is All You Need",
                    "abstract": "Multi-head attention layers, as used in the Transformer neural sequence model, are a powerful alternative to RNNs for moving information across and between sequences. While training these layers is generally fast and simple, due to parallelizability across the length of the sequence, incremental inference (where such paralleization is impossible) is often slow, due to the memory-bandwidth cost of repeatedly loading the large \"keys\" and \"values\" tensors. We propose a variant called multi-query attention, where the keys and values are shared across all of the different attention \"heads\", greatly reducing the size of these tensors and hence the memory bandwidth requirements of incremental decoding. We verify experimentally that the resulting models can indeed be much faster to decode, and incur only minor quality degradation from the baseline."
                },
                {
                    "title": "MathQA: Towards Interpretable Math Word Problem Solving with Operation-Based Formalisms",
                    "abstract": "We introduce a large-scale dataset of math word problems and an interpretable neural math problem solver by learning to map problems to their operation programs. Due to annotation challenges, current datasets in this domain have been either relatively small in scale or did not offer precise operational annotations over diverse problem types. We introduce a new representation language to model operation programs corresponding to each math problem that aim to improve both the performance and the interpretability of the learned models. Using this representation language, we significantly enhance the AQUA-RAT dataset with fully-specified operational programs. We additionally introduce a neural sequence-to-program model with automatic problem categorization. Our experiments show improvements over competitive baselines in our dataset as well as the AQUA-RAT dataset. The results are still lower than human performance indicating that the dataset poses new challenges for future research. Our dataset is available at https://math-qa.github.io/math-QA/"
                },
                {
                    "title": "Don\u2019t Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarization",
                    "abstract": "We introduce \u201cextreme summarization\u201d, a new single-document summarization task which does not favor extractive strategies and calls for an abstractive modeling approach. The idea is to create a short, one-sentence news summary answering the question \u201cWhat is the article about?\u201d. We collect a real-world, large-scale dataset for this task by harvesting online articles from the British Broadcasting Corporation (BBC). We propose a novel abstractive model which is conditioned on the article\u2019s topics and based entirely on convolutional neural networks. We demonstrate experimentally that this architecture captures long-range dependencies in a document and recognizes pertinent content, outperforming an oracle extractive system and state-of-the-art abstractive approaches when evaluated automatically and by humans."
                },
                {
                    "title": "CoQA: A Conversational Question Answering Challenge",
                    "abstract": "Humans gather information through conversations involving a series of interconnected questions and answers. For machines to assist in information gathering, it is therefore essential to enable them to answer conversational questions. We introduce CoQA, a novel dataset for building Conversational Question Answering systems. Our dataset contains 127k questions with answers, obtained from 8k conversations about text passages from seven diverse domains. The questions are conversational, and the answers are free-form text with their corresponding evidence highlighted in the passage. We analyze CoQA in depth and show that conversational questions have challenging phenomena not present in existing reading comprehension datasets (e.g., coreference and pragmatic reasoning). We evaluate strong dialogue and reading comprehension models on CoQA. The best system obtains an F1 score of 65.4%, which is 23.4 points behind human performance (88.8%), indicating that there is ample room for improvement. We present CoQA as a challenge to the community at https://stanfordnlp.github.io/coqa."
                },
                {
                    "title": "Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering",
                    "abstract": "We present a new kind of question answering dataset, OpenBookQA, modeled after open book exams for assessing human understanding of a subject. The open book that comes with our questions is a set of 1326 elementary level science facts. Roughly 6000 questions probe an understanding of these facts and their application to novel situations. This requires combining an open book fact (e.g., metals conduct electricity) with broad common knowledge (e.g., a suit of armor is made of metal) obtained from other sources. While existing QA datasets over documents or knowledge bases, being generally self-contained, focus on linguistic understanding, OpenBookQA probes a deeper understanding of both the topic\u2014in the context of common knowledge\u2014and the language it is expressed in. Human performance on OpenBookQA is close to 92%, but many state-of-the-art pre-trained QA methods perform surprisingly poorly, worse than several simple neural baselines we develop. Our oracle experiments designed to circumvent the knowledge retrieval bottleneck demonstrate the value of both the open book and additional facts. We leave it as a challenge to solve the retrieval problem in this multi-hop setting and to close the large gap to human performance."
                },
                {
                    "title": "GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding",
                    "abstract": "Human ability to understand language is general, flexible, and robust. In contrast, most NLU models above the word level are designed for a specific task and struggle with out-of-domain data. If we aspire to develop models with understanding beyond the detection of superficial correspondences between inputs and outputs, then it is critical to develop a unified model that can execute a range of linguistic tasks across different domains. To facilitate research in this direction, we present the General Language Understanding Evaluation (GLUE, gluebenchmark.com): a benchmark of nine diverse NLU tasks, an auxiliary dataset for probing models for understanding of specific linguistic phenomena, and an online platform for evaluating and comparing models. For some benchmark tasks, training data is plentiful, but for others it is limited or does not match the genre of the test set. GLUE thus favors models that can represent linguistic knowledge in a way that facilitates sample-efficient learning and effective knowledge-transfer across tasks. While none of the datasets in GLUE were created from scratch for the benchmark, four of them feature privately-held test data, which is used to ensure that the benchmark is used fairly. We evaluate baselines that use ELMo (Peters et al., 2018), a powerful transfer learning technique, as well as state-of-the-art sentence representation models. The best models still achieve fairly low absolute scores. Analysis with our diagnostic dataset yields similarly weak performance over all phenomena tested, with some exceptions."
                },
                {
                    "title": "Loss Surfaces, Mode Connectivity, and Fast Ensembling of DNNs",
                    "abstract": "The loss functions of deep neural networks are complex and their geometric properties are not well understood. We show that the optima of these complex loss functions are in fact connected by simple curves over which training and test accuracy are nearly constant. We introduce a training procedure to discover these high-accuracy pathways between modes. Inspired by this new geometric insight, we also propose a new ensembling method entitled Fast Geometric Ensembling (FGE). Using FGE we can train high-performing ensembles in the time required to train a single model. We achieve improved performance compared to the recent state-of-the-art Snapshot Ensembles, on CIFAR-10, CIFAR-100, and ImageNet."
                },
                {
                    "title": "Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks",
                    "abstract": "We present weight normalization: a reparameterization of the weight vectors in a neural network that decouples the length of those weight vectors from their direction. By reparameterizing the weights in this way we improve the conditioning of the optimization problem and we speed up convergence of stochastic gradient descent. Our reparameterization is inspired by batch normalization but does not introduce any dependencies between the examples in a minibatch. This means that our method can also be applied successfully to recurrent models such as LSTMs and to noise-sensitive applications such as deep reinforcement learning or generative models, for which batch normalization is less well suited. Although our method is much simpler, it still provides much of the speed-up of full batch normalization. In addition, the computational overhead of our method is lower, permitting more optimization steps to be taken in the same amount of time. We demonstrate the usefulness of our method on applications in supervised image recognition, generative modelling, and deep reinforcement learning."
                },
                {
                    "title": "Abstractive Text Summarization using Sequence-to-sequence RNNs and Beyond",
                    "abstract": "In this work, we model abstractive text summarization using Attentional Encoder-Decoder Recurrent Neural Networks, and show that they achieve state-of-the-art performance on two different corpora. We propose several novel models that address critical problems in summarization that are not adequately modeled by the basic architecture, such as modeling key-words, capturing the hierarchy of sentence-to-word structure, and emitting words that are rare or unseen at training time. Our work shows that many of our proposed models contribute to further improvement in performance. We also propose a new dataset consisting of multi-sentence summaries, and establish performance benchmarks for further research."
                },
                {
                    "title": "Choice of Plausible Alternatives: An Evaluation of Commonsense Causal Reasoning",
                    "abstract": "Research in open-domain commonsense reasoning has been hindered by the lack of evaluation metrics for judging progress and comparing alternative approaches. Taking inspiration from large-scale question sets used in natural language processing research, we authored one thousand English-language questions that directly assess commonsense causal reasoning, called the Choice Of Plausible Alternatives (COPA) evaluation. Using a forced-choice format, each question gives a premise and two plausible causes or effects, where the correct choice is the alternative that is more plausible than the other. This paper describes the authoring methodology that we used to develop a validated question set with sufficient breadth to advance open-domain commonsense reasoning research. We discuss the design decisions made during the authoring process, and explain how these decisions will affect the design of high-scoring systems. We also present the performance of multiple baseline approaches that use statistical natural language processing techniques, establishing initial benchmarks for future systems."
                },
                {
                    "title": "Animating rotation with quaternion curves",
                    "abstract": "Solid bodies roll and tumble through space. In computer animation, so do cameras. The rotations of these objects are best described using a four coordinate system, quaternions, as is shown in this paper. Of all quaternions, those on the unit sphere are most suitable for animation, but the question of how to construct curves on spheres has not been much explored. This paper gives one answer by presenting a new kind of spline curve, created on a sphere, suitable for smoothly in-betweening (i.e. interpolating) sequences of arbitrary rotations. Both theory and experiment show that the motion generated is smooth and natural, without quirks found in earlier methods."
                },
                {
                    "title": "CacheGen: Fast Context Loading for Language Model Applications",
                    "abstract": "As large language models (LLMs) take on more complex tasks, their inputs incorporate longer contexts to respond to questions that require domain knowledge or user-specific conversational histories. Yet, using long contexts poses a challenge for responsive LLM systems, as nothing can be generated until all the contexts are fetched to and processed by the LLM. Existing systems optimize only the computation delay in context processing ( e.g., by caching intermediate key-value features of the text context) but often cause longer network delays in context fetching ( e.g., key-value features consume orders of magnitude larger bandwidth than the text context). This paper presents CacheGen to minimize the delays in fetching and processing contexts for LLMs. CacheGen reduces the bandwidth needed for transmitting long contexts\u2019 key-value (KV) features through a novel encoder that compresses KV features into more compact bitstream representations. The encoder combines adaptive quantization with a tailored arithmetic coder, taking advantage of the KV features\u2019 distributional properties, such as locality across tokens. Furthermore, CacheGen minimizes the total delay in fetching and processing a context by using a controller that determines when to load the context as compressed KV features or raw text and picks the appropriate compression level if loaded as KV features. We test CacheGen on three models of various sizes and three datasets of different context lengths. Compared to recent methods that handle long contexts, CacheGen reduces bandwidth usage by 3.7-4.3 \u00d7 and the total delay in fetching and processing contexts by 2.7-3 \u00d7 while maintaining similar LLM performance on various tasks as loading the text contexts."
                },
                {
                    "title": "An Adversarial Winograd Schema Challenge at Scale",
                    "abstract": "The Winograd Schema Challenge (WSC), proposed by Levesque et al. (2011) as an alternative to the Turing Test, was originally designed as a pronoun resolution problem that cannot be solved based on statistical patterns in large text corpora. However, recent studies suggest that current WSC datasets, even when composed carefully by experts, are still prone to such biases that statistical methods can exploit. We introduce WINOGRANDE, a new collection of WSC problems that are adversarially constructed to be robust against spurious statistical biases. While the original WSC dataset provided only 273 instances, WINOGRANDE includes 43,985 instances, half of which are determined as adversarial. Key to our approach is a novel adversarial filtering algorithm AFLITE for systematic bias reduction, combined with a careful crowdsourcing design. Despite the significant increase in training data, the performance of existing stateof-the-art methods remains modest (61.6%) and contrasts with high human performance (90.8%) for the binary questions. In addition, WINOGRANDE allows us to use transfer learning for achieving new state-of-the-art results on the original WSC and related datasets. Finally, we discuss how biases lead to overestimating the true capabilities of machine commonsense."
                },
                {
                    "title": "Quaternion Algebra and Calculus",
                    "abstract": "This document provides a mathematical summary of quaternion algebra and calculus and how they relate to rotations and interpolation of rotations. The ideas are based on the article [1]."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively compress KV caches in Large Language Models (LLMs) to reduce memory consumption during generation while maintaining performance?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of KV cache compression is crucial for the research community as it addresses the growing challenge of deploying LLMs efficiently, particularly as models become larger and require more memory. By reducing the memory footprint, we can enable faster generation and support larger batch sizes, which can lead to significant cost savings in commercial applications. This research could pave the way for more scalable and accessible LLMs, fostering further advancements in natural language processing and artificial general intelligence.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of accurately compressing KV caches without sacrificing performance. Naive approaches, such as simple quantization or sparsity methods, may overlook the inter-layer redundancy that exists between adjacent layers of LLMs, leading to suboptimal compression and potential performance degradation. Additionally, the technical obstacles include the need for sophisticated algorithms that can effectively identify and merge similar KV states while ensuring that the integrity of the model's output is preserved.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on intra-layer redundancy, neglecting the significant inter-layer redundancy that exists in KV caches. This oversight has limited the effectiveness of existing compression methods, which either quantize or sparsify KV states without considering the similarities between layers. Barriers such as a lack of comprehensive analysis on inter-layer relationships and the complexity of developing new methodologies have prevented this problem from being adequately addressed until now. Our approach aims to fill this gap by investigating and leveraging the inter-layer redundancy for more effective KV cache compression.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, termed MiniCache, involves analyzing the redundancy of KV caches along the depth dimension of LLMs. We will utilize the LLaMA-3-70B model and the GSM8K dataset to evaluate our approach. The key components include merging KV states from adjacent layers based on their similarities, starting from the middle layers, and employing metrics to assess performance loss. We expect MiniCache to achieve near-lossless performance while significantly reducing memory consumption, thereby enhancing the efficiency of LLM deployment."
            }
        },
        "author_data": {
            "c443ecdb-fc3e-4064-a076-39ba58791701": {
                "pk": "c443ecdb-fc3e-4064-a076-39ba58791701",
                "name": "Akide Liu",
                "collaborators": [
                    "Zeyu Zhang",
                    "Ian Reid",
                    "Richard Hartley",
                    "Bohan Zhuang",
                    "Hao Tang",
                    "Zihan Wang",
                    "Jinchao Ge",
                    "Minh Hieu Phan",
                    "Bowen Zhang",
                    "Yang Zhao"
                ],
                "domain": [
                    "Machine Learning",
                    "Computer Vision",
                    "Natural Language Processing",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "External Reasoning: Towards Multi-Large-Language-Models Interchangeable Assistance with Human Feedback",
                        "abstract": "Memory is identified as a crucial human faculty that allows for the retention of visual and linguistic information within the hippocampus and neurons in the brain, which can subsequently be retrieved to address real-world challenges that arise through a lifetime of learning. The resolution of complex AI tasks through the application of acquired knowledge represents a stride toward the realization of artificial general intelligence. However, despite the prevalence of Large Language Models (LLMs) like GPT-3.5 and GPT-4 \\cite{brown2020language, leiter2023chatgpt, zaitsu2023distinguishing, OpenAI2023GPT4TR} , which have displayed remarkable capabilities in language comprehension, generation, interaction, and reasoning, they are inhibited by constraints on context length that preclude the processing of extensive, continually evolving knowledge bases. This paper proposes that LLMs could be augmented through the selective integration of knowledge from external repositories, and in doing so, introduces a novel methodology for External Reasoning, exemplified by ChatPDF. Central to this approach is the establishment of a tiered policy for \\textbf{External Reasoning based on Multiple LLM Interchange Assistance} in \\cref{fig:overall}, where the level of support rendered is modulated across entry, intermediate, and advanced tiers based on the complexity of the query, with adjustments made in response to human feedback. A comprehensive evaluation of this methodology is conducted using multiple LLMs and the results indicate state-of-the-art performance in \\cref{comparison} , surpassing existing solutions including ChatPDF.com. Moreover, the paper emphasizes that this approach is more efficient compared to the direct processing of full text by LLMs. The source code is publicly available at: \\url{https://github.com/AkideLiu/ANLP}."
                    },
                    {
                        "title": "CV 3315 Is All You Need : Semantic Segmentation Competition",
                        "abstract": "This competition focus on Urban-Sense Segmentation based on the vehicle camera view. Class highly unbalanced Urban-Sense images dataset challenge the existing solutions and further studies. Deep Conventional neural network-based semantic segmentation methods such as encoder-decoder architecture and multi-scale and pyramid-based approaches become flexible solutions applicable to real-world applications. In this competition, we mainly review the literature and conduct experiments on transformer-driven methods especially SegFormer, to achieve an optimal trade-off between performance and efficiency. For example, SegFormer-B0 achieved 74.6% mIoU with the smallest FLOPS, 15.6G, and the largest model, SegFormer- B5 archived 80.2% mIoU. According to multiple factors, including individual case failure analysis, individual class performance, training pressure and efficiency estimation, the final candidate model for the competition is SegFormer- B2 with 50.6 GFLOPS and 78.5% mIoU evaluated on the testing set. Checkout our code implementation at https://vmv.re/cv3315."
                    },
                    {
                        "title": "ESA: Annotation-Efficient Active Learning for Semantic Segmentation",
                        "abstract": "Active learning enhances annotation efficiency by selecting the most revealing samples for labeling, thereby reducing reliance on extensive human input. Previous methods in semantic segmentation have centered on individual pixels or small areas, neglecting the rich patterns in natural images and the power of advanced pre-trained models. To address these challenges, we propose three key contributions: Firstly, we introduce Entity-Superpixel Annotation (ESA), an innovative and efficient active learning strategy which utilizes a class-agnostic mask proposal network coupled with super-pixel grouping to capture local structural cues. Additionally, our method selects a subset of entities within each image of the target domain, prioritizing superpixels with high entropy to ensure comprehensive representation. Simultaneously, it focuses on a limited number of key entities, thereby optimizing for efficiency. By utilizing an annotator-friendly design that capitalizes on the inherent structure of images, our approach significantly outperforms existing pixel-based methods, achieving superior results with minimal queries, specifically reducing click cost by 98% and enhancing performance by 1.71%. For instance, our technique requires a mere 40 clicks for annotation, a stark contrast to the 5000 clicks demanded by conventional methods."
                    },
                    {
                        "title": "Motion Mamba: Efficient and Long Sequence Motion Generation",
                        "abstract": "Human motion generation stands as a significant pursuit in generative computer vision, while achieving long-sequence and efficient motion generation remains challenging. Recent advancements in state space models (SSMs), notably Mamba, have showcased considerable promise in long sequence modeling with an efficient hardware-aware design, which appears to be a promising direction to build motion generation model upon it. Nevertheless, adapting SSMs to motion generation faces hurdles since the lack of a specialized design architecture to model motion sequence. To address these challenges, we propose Motion Mamba, a simple and efficient approach that presents the pioneering motion generation model utilized SSMs. Specifically, we design a Hierarchical Temporal Mamba (HTM) block to process temporal data by ensemble varying numbers of isolated SSM modules across a symmetric U-Net architecture aimed at preserving motion consistency between frames. We also design a Bidirectional Spatial Mamba (BSM) block to bidirectionally process latent poses, to enhance accurate motion generation within a temporal frame. Our proposed method achieves up to 50% FID improvement and up to 4 times faster on the HumanML3D and KIT-ML datasets compared to the previous best diffusion-based method, which demonstrates strong capabilities of high-quality long sequence motion modeling and real-time human motion generation. See project website https://steve-zeyu-zhang.github.io/MotionMamba/"
                    },
                    {
                        "title": "InfiniMotion: Mamba Boosts Memory in Transformer for Arbitrary Long Motion Generation",
                        "abstract": "Text-to-motion generation holds potential for film, gaming, and robotics, yet current methods often prioritize short motion generation, making it challenging to produce long motion sequences effectively: (1) Current methods struggle to handle long motion sequences as a single input due to prohibitively high computational cost; (2) Breaking down the generation of long motion sequences into shorter segments can result in inconsistent transitions and requires interpolation or inpainting, which lacks entire sequence modeling. To solve these challenges, we propose InfiniMotion, a method that generates continuous motion sequences of arbitrary length within an autoregressive framework. We highlight its groundbreaking capability by generating a continuous 1-hour human motion with around 80,000 frames. Specifically, we introduce the Motion Memory Transformer with Bidirectional Mamba Memory, enhancing the transformer's memory to process long motion sequences effectively without overwhelming computational resources. Notably our method achieves over 30% improvement in FID and 6 times longer demonstration compared to previous state-of-the-art methods, showcasing significant advancements in long motion generation. See project webpage: https://steve-zeyu-zhang.github.io/InfiniMotion/"
                    }
                ]
            },
            "2c7150f5-4232-4700-a75d-1c0e5eea5f2a": {
                "pk": "2c7150f5-4232-4700-a75d-1c0e5eea5f2a",
                "name": "Jing Liu",
                "collaborators": [
                    "Xiaoguang Wang",
                    "Xiaoxing Jing",
                    "Wei Zhong",
                    "Haidong Yuan",
                    "Xiao-Xing Jing",
                    "Jing Cheng",
                    "Li-Bin Fu"
                ],
                "domain": [
                    "Gravitational Waves",
                    "Quantum Metrology",
                    "Quantum Information",
                    "Astrophysics"
                ],
                "publications": [
                    {
                        "title": "Distinguishing the nanohertz gravitational-wave sources by the observations of compact dark matter subhalos",
                        "abstract": "The latest pulsar timing array data reveals evidence of nanohertz gravitational waves (GWs), which have been explained by both cosmological and astrophysical sources. However, current observations lack the precision needed to differentiate between different models from the spectral index. We find that the cosmological GW sources, including bubble collisions, sound waves, domain walls, condensate fragmentations, and primordial curvature perturbations, induce large energy density perturbations so that most dark matter will exist in gravitationally self-bound subhalos. Then, the observation of such substructures of dark matter can serve as a novel independent method to confirm or exclude the cosmological GW sources."
                    },
                    {
                        "title": "Gravitational laser: the stimulated radiation of gravitational waves from the clouds of ultralight bosons",
                        "abstract": "Stimulated radiation and gravitational waves (GWs) are two of the most important predictions made by Albert Einstein. In this work, we demonstrate that stimulated GW radiation can occur within gravitational atoms, which consist of Kerr black holes and the surrounding boson clouds formed through superradiance. The presence of GWs induces mixing between different states of the gravitational atoms, leading to resonant transitions between two states when the GW wavenumber closely matches the energy difference. Consequently, the energy and angular momentum released from these transitions lead to the amplification of GWs, resulting in an exponential increase in the transition rate. Remarkably, the transitions complete within a much shorter time compared to the lifetime of the cloud. These stimulated transitions give rise to a novel GW signal that is strong and directed, distinguished from the previously predicted continuous GWs originating from clouds of ultralight bosons."
                    },
                    {
                        "title": "Phase matching condition for enhancement of phase sensitivity in quantum metrology",
                        "abstract": "We find a phase matching condition for enhancement of sensitivity in a Mach-Zehnder interferometer illuminated by an arbitrary state in one input port and an odd(even) state in the other port. Under this condition, the Fisher information becomes maximal with respect to the relative phase of two modes and the phase sensitivity is enhanced. For the case with photon losses, we further find that the phase matching condition keeps unchanged with a coherent state and a coherent superposition state as the input states."
                    },
                    {
                        "title": "Quantum metrology with unitary parametrization processes",
                        "abstract": "Quantum Fisher information is a central quantity in quantum metrology. We discuss an alternative representation of quantum Fisher information for unitary parametrization processes. The highlight of this representation is that all information of parametrization transformation, i.e., the entire dynamical information, is totally involved in a Hermitian operator H. Utilizing this representation, quantum Fisher information is only determined by H and the initial state. We apply this representation in a collective spin system and show the specific expression of H. For a simple case, a spin-half system, the quantum Fisher information is given and the optimal states to access maximum quantum Fisher information are found. Furthermore, for an exponential form initial state, an analytical expression of quantum Fisher information by H operator is provided. The multiparameter quantum metrology is also considered and discussed utilizing this representation in the last."
                    },
                    {
                        "title": "Valid lower bound for all estimators in quantum parameter estimation",
                        "abstract": "The widely used quantum Cramer-Rao bound (QCRB) sets a lower bound for the mean square error of unbiased estimators in quantum parameter estimation, however, in general QCRB is only tight in the asymptotical limit. With a limited number of measurements biased estimators can have a far better performance for which QCRB cannot calibrate. Here we introduce a valid lower bound for all estimators, either biased or unbiased, which can serve as standard of merit for all quantum parameter estimations."
                    },
                    {
                        "title": "Quantum Fisher Information of Entangled Coherent States in a Lossy Mach-Zehnder Interferometer",
                        "abstract": "We give an analytical result for the quantum Fisher information of entangled coherent States in a lossy Mach-Zehnder Interferometer recently proposed by J. Joo et al. [Phys. Rev. Lett. 107, 083601(2011)]. For small loss of photons, we find that the entangled coherent state can surpass the Heisenberg limit. Furthermore, The formalism developed here is applicable to the study of phase sensitivity of multipartite entangled coherent states."
                    },
                    {
                        "title": "Quantum Fisher information for density matrices with arbitrary ranks",
                        "abstract": "We provide a new expression of the quantum Fisher information(QFI) for a general system. Utilizing this expression, the QFI for a non-full rank density matrix is only determined by its support. This expression can bring convenience for a infinite dimensional density matrix with a finite support. Besides, a matrix representation of the QFI is also given."
                    },
                    {
                        "title": "Quantum information with conserved quantities",
                        "abstract": "Conserved quantities are crucial in quantum physics. Here we discuss a general scenario of Hamiltonians. All the Hamiltonians within this scenario share a common conserved quantity form. For unitary parametrization processes, the characteristic operator of this scenario is analytically provided, as well as the corresponding quantum Fisher information (QFI). As the application of this scenario, we focus on two classes of Hamiltonians: su(2) category and canonical category. Several specific physical systems in these two categories are discussed in detail. Besides, we also calculate an alternative form of QFI in this scenario."
                    }
                ]
            },
            "41ce7cde-5361-462c-8e74-2d34d4d095a3": {
                "pk": "41ce7cde-5361-462c-8e74-2d34d4d095a3",
                "name": "Zizheng Pan",
                "collaborators": [
                    "Bohan Zhuang",
                    "Jianfei Cai",
                    "Jing Liu",
                    "Haoyu He",
                    "Yuetian Weng",
                    "Jing Zhang",
                    "Dacheng Tao",
                    "Yuankai Qi",
                    "Anton van den Hengel",
                    "Qi Wu"
                ],
                "domain": [
                    "Computer Vision",
                    "Transformers",
                    "Deep Learning",
                    "Model Efficiency"
                ],
                "publications": [
                    {
                        "title": "PaRa: Personalizing Text-to-Image Diffusion via Parameter Rank Reduction",
                        "abstract": "Personalizing a large-scale pretrained Text-to-Image (T2I) diffusion model is challenging as it typically struggles to make an appropriate trade-off between its training data distribution and the target distribution, i.e., learning a novel concept with only a few target images to achieve personalization (aligning with the personalized target) while preserving text editability (aligning with diverse text prompts). In this paper, we propose PaRa, an effective and efficient Parameter Rank Reduction approach for T2I model personalization by explicitly controlling the rank of the diffusion model parameters to restrict its initial diverse generation space into a small and well-balanced target space. Our design is motivated by the fact that taming a T2I model toward a novel concept such as a specific art style implies a small generation space. To this end, by reducing the rank of model parameters during finetuning, we can effectively constrain the space of the denoising sampling trajectories towards the target. With comprehensive experiments, we show that PaRa achieves great advantages over existing finetuning approaches on single/multi-subject generation as well as single-image editing. Notably, compared to the prevailing fine-tuning technique LoRA, PaRa achieves better parameter efficiency (2x fewer learnable parameters) and much better target image alignment."
                    },
                    {
                        "title": "Stitchable Neural Networks",
                        "abstract": "The public model zoo containing enormous powerful pretrained model families (e.g., ResNet/DeiT) has reached an unprecedented scope than ever, which significantly contributes to the success of deep learning. As each model family consists of pretrained models with diverse scales (e.g., DeiT-Ti/S/B), it naturally arises a fundamental question of how to efficiently assemble these readily available models in a family for dynamic accuracy-efficiency trade-offs at runtime. To this end, we present Stitchable Neural Networks (SN-Net), a novel scalable and efficient framework for model deployment. It cheaply produces numerous networks with different complexity and performance trade-offs given a family of pretrained neural networks, which we call anchors. Specifically, SN-Net splits the anchors across the blocks/layers and then stitches them together with simple stitching layers to map the activations from one anchor to another. With only a few epochs of training, SN-Net effectively interpolates between the performance of anchors with varying scales. At runtime, SN-Net can instantly adapt to dynamic resource constraints by switching the stitching positions. Extensive experiments on ImageNet classification demonstrate that SN-Net can obtain on-par or even better performance than many individually trained networks while supporting diverse deployment scenarios. For example, by stitching Swin Transformers, we challenge hundreds of models in Timm model zoo with a single network. We believe this new elastic model framework can serve as a strong baseline for further research in wider communities."
                    },
                    {
                        "title": "Fast Vision Transformers with HiLo Attention",
                        "abstract": "Vision Transformers (ViTs) have triggered the most recent and significant breakthroughs in computer vision. Their efficient designs are mostly guided by the indirect metric of computational complexity, i.e., FLOPs, which however has a clear gap with the direct metric such as throughput. Thus, we propose to use the direct speed evaluation on the target platform as the design principle for efficient ViTs. Particularly, we introduce LITv2, a simple and effective ViT which performs favourably against the existing state-of-the-art methods across a spectrum of different model sizes with faster speed. At the core of LITv2 is a novel self-attention mechanism, which we dub HiLo. HiLo is inspired by the insight that high frequencies in an image capture local fine details and low frequencies focus on global structures, whereas a multi-head self-attention layer neglects the characteristic of different frequencies. Therefore, we propose to disentangle the high/low frequency patterns in an attention layer by separating the heads into two groups, where one group encodes high frequencies via self-attention within each local window, and another group encodes low frequencies by performing global attention between the average-pooled low-frequency keys and values from each window and each query position in the input feature map. Benefiting from the efficient design for both groups, we show that HiLo is superior to the existing attention mechanisms by comprehensively benchmarking FLOPs, speed and memory consumption on GPUs and CPUs. For example, HiLo is 1.4x faster than spatial reduction attention and 1.6x faster than local window attention on CPUs. Powered by HiLo, LITv2 serves as a strong backbone for mainstream vision tasks including image classification, dense detection and segmentation. Code is available at https://github.com/ziplab/LITv2."
                    },
                    {
                        "title": "Scalable Vision Transformers with Hierarchical Pooling",
                        "abstract": "The recently proposed Visual image Transformers (ViT) with pure attention have achieved promising performance on image recognition tasks, such as image classification. However, the routine of the current ViT model is to maintain a full-length patch sequence during inference, which is redundant and lacks hierarchical representation. To this end, we propose a Hierarchical Visual Transformer (HVT) which progressively pools visual tokens to shrink the sequence length and hence reduces the computational cost, analogous to the feature maps downsampling in Convolutional Neural Networks (CNNs). It brings a great benefit that we can increase the model capacity by scaling dimensions of depth/width/resolution/patch size without introducing extra computational complexity due to the reduced sequence length. Moreover, we empirically find that the average pooled visual tokens contain more discriminative information than the single class token. To demonstrate the improved scalability of our HVT, we conduct extensive experiments on the image classification task. With comparable FLOPs, our HVT outperforms the competitive baselines on ImageNet and CIFAR-100 datasets. Code is available at https://github.com/MonashAI/HVT"
                    },
                    {
                        "title": "Less is More: Pay Less Attention in Vision Transformers",
                        "abstract": "Transformers have become one of the dominant architectures in deep learning, particularly as a powerful alternative to convolutional neural networks (CNNs) in computer vision. However, Transformer training and inference in previous works can be prohibitively expensive due to the quadratic complexity of self-attention over a long sequence of representations, especially for high-resolution dense prediction tasks. To this end, we present a novel Less attention vIsion Transformer (LIT), building upon the fact that the early self-attention layers in Transformers still focus on local patterns and bring minor benefits in recent hierarchical vision Transformers. Specifically, we propose a hierarchical Transformer where we use pure multi-layer perceptrons (MLPs) to encode rich local patterns in the early stages while applying self-attention modules to capture longer dependencies in deeper layers. Moreover, we further propose a learned deformable token merging module to adaptively fuse informative patches in a non-uniform manner. The proposed LIT achieves promising performance on image recognition tasks, including image classification, object detection and instance segmentation, serving as a strong backbone for many vision tasks. Code is available at: https://github.com/zhuang-group/LIT"
                    },
                    {
                        "title": "An Efficient Spatio-Temporal Pyramid Transformer for Action Detection",
                        "abstract": "The task of action detection aims at deducing both the action category and localization of the start and end moment for each action instance in a long, untrimmed video. While vision Transformers have driven the recent advances in video understanding, it is non-trivial to design an efficient architecture for action detection due to the prohibitively expensive self-attentions over a long sequence of video clips. To this end, we present an efficient hierarchical Spatio-Temporal Pyramid Transformer (STPT) for action detection, building upon the fact that the early self-attention layers in Transformers still focus on local patterns. Specifically, we propose to use local window attention to encode rich local spatio-temporal representations in the early stages while applying global attention modules to capture long-term space-time dependencies in the later stages. In this way, our STPT can encode both locality and dependency with largely reduced redundancy, delivering a promising trade-off between accuracy and efficiency. For example, with only RGB input, the proposed STPT achieves 53.6% mAP on THUMOS14, surpassing I3D+AFSD RGB model by over 10% and performing favorably against state-of-the-art AFSD that uses additional flow features with 31% fewer GFLOPs, which serves as an effective and efficient end-to-end Transformer-based framework for action detection."
                    },
                    {
                        "title": "Stitched ViTs are Flexible Vision Backbones",
                        "abstract": "Large pretrained plain vision Transformers (ViTs) have been the workhorse for many downstream tasks. However, existing works utilizing off-the-shelf ViTs are inefficient in terms of training and deployment, because adopting ViTs with individual sizes requires separate trainings and is restricted by fixed performance-efficiency trade-offs. In this paper, we are inspired by stitchable neural networks (SN-Net), which is a new framework that cheaply produces a single model that covers rich subnetworks by stitching pretrained model families, supporting diverse performance-efficiency trade-offs at runtime. Building upon this foundation, we introduce SN-Netv2, a systematically improved model stitching framework to facilitate downstream task adaptation. Specifically, we first propose a two-way stitching scheme to enlarge the stitching space. We then design a resource-constrained sampling strategy that takes into account the underlying FLOPs distributions in the space for better sampling. Finally, we observe that learning stitching layers as a low-rank update plays an essential role on downstream tasks to stabilize training and ensure a good Pareto frontier. With extensive experiments on ImageNet-1K, ADE20K, COCO-Stuff-10K and NYUv2, SN-Netv2 demonstrates superior performance over SN-Netv1 on downstream dense predictions and shows strong ability as a flexible vision backbone, achieving great advantages in both training efficiency and deployment flexibility. Code is available at https://github.com/ziplab/SN-Netv2."
                    },
                    {
                        "title": "EcoFormer: Energy-Saving Attention with Linear Complexity",
                        "abstract": "Transformer is a transformative framework that models sequential data and has achieved remarkable performance on a wide range of tasks, but with high computational and energy cost. To improve its efficiency, a popular choice is to compress the models via binarization which constrains the floating-point values into binary ones to save resource consumption owing to cheap bitwise operations significantly. However, existing binarization methods only aim at minimizing the information loss for the input distribution statistically, while ignoring the pairwise similarity modeling at the core of the attention. To this end, we propose a new binarization paradigm customized to high-dimensional softmax attention via kernelized hashing, called EcoFormer, to map the original queries and keys into low-dimensional binary codes in Hamming space. The kernelized hash functions are learned to match the ground-truth similarity relations extracted from the attention map in a self-supervised way. Based on the equivalence between the inner product of binary codes and the Hamming distance as well as the associative property of matrix multiplication, we can approximate the attention in linear complexity by expressing it as a dot-product of binary codes. Moreover, the compact binary representations of queries and keys enable us to replace most of the expensive multiply-accumulate operations in attention with simple accumulations to save considerable on-chip energy footprint on edge devices. Extensive experiments on both vision and language tasks show that EcoFormer consistently achieves comparable performance with standard attentions while consuming much fewer resources. For example, based on PVTv2-B0 and ImageNet-1K, Ecoformer achieves a 73% on-chip energy footprint reduction with only a 0.33% performance drop compared to the standard attention. Code is available at https://github.com/ziplab/EcoFormer."
                    },
                    {
                        "title": "Efficient Stitchable Task Adaptation",
                        "abstract": "The paradigm of pre-training and fine-tuning has laid the foundation for deploying deep learning models. However, most fine-tuning methods are designed to meet a specific resource budget. Recently, considering diverse deployment scenarios with various resource budgets, SN-Net is introduced to quickly obtain numerous new networks (stitches) from the pre-trained models (anchors) in a model family via model stitching. Although promising, SN-Net confronts new challenges when adapting it to new target domains, including huge memory and storage requirements and a long and sub-optimal multistage adaptation process. In this work, we present a novel framework, Efficient Stitchable Task Adaptation (ESTA), to efficiently produce a palette of fine-tuned models that adhere to diverse resource constraints. Specifically, we first tailor parameter-efficient fine-tuning to share low-rank updates among the stitches while maintaining independent bias terms. In this way, we largely reduce fine-tuning memory burdens and mitigate the interference among stitches that arises in task adaptation. Furthermore, we streamline a simple yet effective one-stage deployment pipeline, which estimates the important stitches to deploy with training-time gradient statistics. By assigning higher sampling probabilities to important stitches, we also get a boosted Pareto frontier. Extensive experiments on 25 downstream visual recognition tasks demonstrate that our ESTA is capable of generating stitches with smooth accuracy-efficiency trade-offs and surpasses the direct SN-Net adaptation by remarkable margins with significantly lower training time and fewer trainable parameters. Furthermore, we demonstrate the flexibility and scalability of our ESTA framework by stitching LLMs from LLaMA family, obtaining chatbot stitches of assorted sizes. Source code is available at https://github.com/ziplab/Stitched_LLaMA"
                    },
                    {
                        "title": "A Survey on Efficient Training of Transformers",
                        "abstract": "Recent advances in Transformers have come with a huge requirement on computing resources, highlighting the importance of developing efficient training techniques to make Transformer training faster, at lower cost, and to higher accuracy by the efficient use of computation and memory resources. This survey provides the first systematic overview of the efficient training of Transformers, covering the recent progress in acceleration arithmetic and hardware, with a focus on the former. We analyze and compare methods that save computation and memory costs for intermediate tensors during training, together with techniques on hardware/algorithm co-design. We finally discuss challenges and promising areas for future research."
                    },
                    {
                        "title": "Mesa: A Memory-saving Training Framework for Transformers",
                        "abstract": "There has been an explosion of interest in designing high-performance Transformers. While Transformers have delivered significant performance improvements, training such networks is extremely memory intensive owing to storing all intermediate activations that are needed for gradient computation during backpropagation, especially for long sequences. To this end, we present Mesa, a memory-saving training framework for Transformers. Specifically, Mesa uses exact activations during forward pass while storing a low-precision version of activations to reduce memory consumption during training. The low-precision activations are then dequantized during back-propagation to compute gradients. Besides, to address the heterogeneous activation distributions in the multi-head self-attention layers, we propose a head-wise activation quantization strategy, which quantizes activations based on the statistics of each head to minimize the approximation error. To further boost training efficiency, we learn quantization parameters by running estimates. More importantly, by re-investing the saved memory in employing a larger batch size or scaling up model size, we may further improve the performance under constrained computational resources. Extensive experiments on ImageNet, CIFAR-100 and ADE20K demonstrate that Mesa can achieve flexible memory-savings (up to 50%) during training while achieving comparable or even better performance. Code is available at https://github.com/ziplab/Mesa."
                    },
                    {
                        "title": "Dynamic Focus-aware Positional Queries for Semantic Segmentation",
                        "abstract": "The DETR-like segmentors have underpinned the most recent breakthroughs in semantic segmentation, which end-to-end train a set of queries representing the class prototypes or target segments. Recently, masked attention is proposed to restrict each query to only attend to the foreground regions predicted by the preceding decoder block for easier optimization. Although promising, it relies on the learnable parameterized positional queries which tend to encode the dataset statistics, leading to inaccurate localization for distinct individual queries. In this paper, we propose a simple yet effective query design for semantic segmentation termed Dynamic Focus-aware Positional Queries (DFPQ), which dynamically generates positional queries conditioned on the cross-attention scores from the preceding decoder block and the positional encodings for the corresponding image features, simultaneously. Therefore, our DFPQ preserves rich localization information for the target segments and provides accurate and fine-grained positional priors. In addition, we propose to efficiently deal with high-resolution cross-attention by only aggregating the contextual tokens based on the low-resolution cross-attention scores to perform local relation aggregation. Extensive experiments on ADE20K and Cityscapes show that with the two modifications on Mask2former, our framework achieves SOTA performance and outperforms Mask2former by clear margins of 1.1%, 1.9%, and 1.1% single-scale mIoU with ResNet-50, Swin-T, and Swin-B backbones on the ADE20K validation set, respectively. Source code is available at https://github.com/ziplab/FASeg"
                    },
                    {
                        "title": "Object-and-Action Aware Model for Visual Language Navigation",
                        "abstract": "Vision-and-Language Navigation (VLN) is unique in that it requires turning relatively general natural-language instructions into robot agent actions, on the basis of the visible environment. This requires to extract value from two very different types of natural-language information. The first is object description (e.g., 'table', 'door'), each presenting as a tip for the agent to determine the next action by finding the item visible in the environment, and the second is action specification (e.g., 'go straight', 'turn left') which allows the robot to directly predict the next movements without relying on visual perceptions. However, most existing methods pay few attention to distinguish these information from each other during instruction encoding and mix together the matching between textual object/action encoding and visual perception/orientation features of candidate viewpoints. In this paper, we propose an Object-and-Action Aware Model (OAAM) that processes these two different forms of natural language based instruction separately. This enables each process to match object-centered/action-centered instruction to their own counterpart visual perception/action orientation flexibly. However, one side-issue caused by above solution is that an object mentioned in instructions may be observed in the direction of two or more candidate viewpoints, thus the OAAM may not predict the viewpoint on the shortest path as the next action. To handle this problem, we design a simple but effective path loss to penalize trajectories deviating from the ground truth path. Experimental results demonstrate the effectiveness of the proposed model and path loss, and the superiority of their combination with a 50% SPL score on the R2R dataset and a 40% CLS score on the R4R dataset in unseen environments, outperforming the previous state-of-the-art."
                    },
                    {
                        "title": "Pruning Self-attentions into Convolutional Layers in Single Path",
                        "abstract": "Vision Transformers (ViTs) have achieved impressive performance over various computer vision tasks. However, modeling global correlations with multi-head self-attention (MSA) layers leads to two widely recognized issues: the massive computational resource consumption and the lack of intrinsic inductive bias for modeling local visual patterns. To solve both issues, we devise a simple yet effective method named Single-Path Vision Transformer pruning (SPViT), to efficiently and automatically compress the pre-trained ViTs into compact models with proper locality added. Specifically, we first propose a novel weight-sharing scheme between MSA and convolutional operations, delivering a single-path space to encode all candidate operations. In this way, we cast the operation search problem as finding which subset of parameters to use in each MSA layer, which significantly reduces the computational cost and optimization difficulty, and the convolution kernels can be well initialized using pre-trained MSA parameters. Relying on the single-path space, we introduce learnable binary gates to encode the operation choices in MSA layers. Similarly, we further employ learnable gates to encode the fine-grained MLP expansion ratios of FFN layers. In this way, our SPViT optimizes the learnable gates to automatically explore from a vast and unified search space and flexibly adjust the MSA-FFN pruning proportions for each individual dense model. We conduct extensive experiments on two representative ViTs showing that our SPViT achieves a new SOTA for pruning on ImageNet-1k. For example, our SPViT can trim 52.0% FLOPs for DeiT-B and get an impressive 0.6% top-1 accuracy gain simultaneously. The source code is available at https://github.com/ziplab/SPViT."
                    },
                    {
                        "title": "The Road to Know-Where: An Object-and-Room Informed Sequential BERT for Indoor Vision-Language Navigation",
                        "abstract": "Vision-and-Language Navigation (VLN) requires an agent to find a path to a remote location on the basis of natural-language instructions and a set of photo-realistic panoramas. Most existing methods take the words in the instructions and the discrete views of each panorama as the minimal unit of encoding. However, this requires a model to match different nouns (e.g., TV, table) against the same input view feature. In this work, we propose an object-informed sequential BERT to encode visual perceptions and linguistic instructions at the same fine-grained level, namely objects and words. Our sequential BERT also enables the visual-textual clues to be interpreted in light of the temporal context, which is crucial to multi-round VLN tasks. Additionally, we enable the model to identify the relative direction (e.g., left/right/front/back) of each navigable location and the room type (e.g., bedroom, kitchen) of its current and final navigation goal, as such information is widely mentioned in instructions implying the desired next and final locations. We thus enable the model to know-where the objects lie in the images, and to know-where they stand in the scene. Extensive experiments demonstrate the effectiveness compared against several state-of-the-art methods on three indoor VLN tasks: REVERIE, NDH, and R2R. Project repository: https://github.com/YuankaiQi/ORIST"
                    },
                    {
                        "title": "T-Stitch: Accelerating Sampling in Pre-Trained Diffusion Models with Trajectory Stitching",
                        "abstract": "Sampling from diffusion probabilistic models (DPMs) is often expensive for high-quality image generation and typically requires many steps with a large model. In this paper, we introduce sampling Trajectory Stitching T-Stitch, a simple yet efficient technique to improve the sampling efficiency with little or no generation degradation. Instead of solely using a large DPM for the entire sampling trajectory, T-Stitch first leverages a smaller DPM in the initial steps as a cheap drop-in replacement of the larger DPM and switches to the larger DPM at a later stage. Our key insight is that different diffusion models learn similar encodings under the same training data distribution and smaller models are capable of generating good global structures in the early steps. Extensive experiments demonstrate that T-Stitch is training-free, generally applicable for different architectures, and complements most existing fast sampling techniques with flexible speed and quality trade-offs. On DiT-XL, for example, 40% of the early timesteps can be safely replaced with a 10x faster DiT-S without performance drop on class-conditional ImageNet generation. We further show that our method can also be used as a drop-in technique to not only accelerate the popular pretrained stable diffusion (SD) models but also improve the prompt alignment of stylized SD models from the public model zoo. Code is released at https://github.com/NVlabs/T-Stitch"
                    },
                    {
                        "title": "Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation",
                        "abstract": "In this paper, we introduce Janus, an autoregressive framework that unifies multimodal understanding and generation. Prior research often relies on a single visual encoder for both tasks, such as Chameleon. However, due to the differing levels of information granularity required by multimodal understanding and generation, this approach can lead to suboptimal performance, particularly in multimodal understanding. To address this issue, we decouple visual encoding into separate pathways, while still leveraging a single, unified transformer architecture for processing. The decoupling not only alleviates the conflict between the visual encoder's roles in understanding and generation, but also enhances the framework's flexibility. For instance, both the multimodal understanding and generation components can independently select their most suitable encoding methods. Experiments show that Janus surpasses previous unified model and matches or exceeds the performance of task-specific models. The simplicity, high flexibility, and effectiveness of Janus make it a strong candidate for next-generation unified multimodal models."
                    }
                ]
            },
            "35d3782b-bd09-4db3-b4c7-028cb9ea9862": {
                "pk": "35d3782b-bd09-4db3-b4c7-028cb9ea9862",
                "name": "Yefei He",
                "collaborators": [
                    "Weijia Wu",
                    "Hong Zhou",
                    "Luoming Zhang",
                    "Jing Liu",
                    "Bohan Zhuang",
                    "Zhenyu Lou",
                    "Mike Zheng Shou",
                    "Wen Fei",
                    "Yuchao Gu",
                    "Rui Zhao"
                ],
                "domain": [
                    "Quantization",
                    "Neural Networks",
                    "Image Synthesis",
                    "Video Processing"
                ],
                "publications": [
                    {
                        "title": "Data-Free Quantization with Accurate Activation Clipping and Adaptive Batch Normalization",
                        "abstract": "Data-free quantization is a task that compresses the neural network to low bit-width without access to original training data. Most existing data-free quantization methods cause severe performance degradation due to inaccurate activation clipping range and quantization error, especially for low bit-width. In this paper, we present a simple yet effective data-free quantization method with accurate activation clipping and adaptive batch normalization. Accurate activation clipping (AAC) improves the model accuracy by exploiting accurate activation information from the full-precision model. Adaptive batch normalization firstly proposes to address the quantization error from distribution changes by updating the batch normalization layer adaptively. Extensive experiments demonstrate that the proposed data-free quantization method can yield surprisingly performance, achieving 64.33% top-1 accuracy of ResNet18 on ImageNet dataset, with 3.7% absolute improvement outperforming the existing state-of-the-art methods."
                    },
                    {
                        "title": "Binarizing by Classification: Is soft function really necessary?",
                        "abstract": "Binary neural networks leverage $\\mathrm{Sign}$ function to binarize weights and activations, which require gradient estimators to overcome its non-differentiability and will inevitably bring gradient errors during backpropagation. Although many hand-designed soft functions have been proposed as gradient estimators to better approximate gradients, their mechanism is not clear and there are still huge performance gaps between binary models and their full-precision counterparts. To address these issues and reduce gradient error, we propose to tackle network binarization as a binary classification problem and use a multi-layer perceptron (MLP) as the classifier in the forward pass and gradient estimator in the backward pass. Benefiting from the MLP's theoretical capability to fit any continuous function, it can be adaptively learned to binarize networks and backpropagate gradients without any prior knowledge of soft functions. From this perspective, we further empirically justify that even a simple linear function can outperform previous complex soft functions. Extensive experiments demonstrate that the proposed method yields surprising performance both in image classification and human pose estimation tasks. Specifically, we achieve $65.7\\%$ top-1 accuracy of ResNet-34 on ImageNet dataset, with an absolute improvement of $2.6\\%$. Moreover, we take binarization as a lightweighting approach for pose estimation models and propose well-designed binary pose estimation networks SBPN and BHRNet. When evaluating on the challenging Microsoft COCO keypoint dataset, the proposed method enables binary networks to achieve a mAP of up to $60.6$ for the first time. Experiments conducted on real platforms demonstrate that BNN achieves a better balance between performance and computational complexity, especially when computational resources are extremely low."
                    },
                    {
                        "title": "EfficientDM: Efficient Quantization-Aware Fine-Tuning of Low-Bit Diffusion Models",
                        "abstract": "Diffusion models have demonstrated remarkable capabilities in image synthesis and related generative tasks. Nevertheless, their practicality for real-world applications is constrained by substantial computational costs and latency issues. Quantization is a dominant way to compress and accelerate diffusion models, where post-training quantization (PTQ) and quantization-aware training (QAT) are two main approaches, each bearing its own properties. While PTQ exhibits efficiency in terms of both time and data usage, it may lead to diminished performance in low bit-width. On the other hand, QAT can alleviate performance degradation but comes with substantial demands on computational and data resources. In this paper, we introduce a data-free and parameter-efficient fine-tuning framework for low-bit diffusion models, dubbed EfficientDM, to achieve QAT-level performance with PTQ-like efficiency. Specifically, we propose a quantization-aware variant of the low-rank adapter (QALoRA) that can be merged with model weights and jointly quantized to low bit-width. The fine-tuning process distills the denoising capabilities of the full-precision model into its quantized counterpart, eliminating the requirement for training data. We also introduce scale-aware optimization and temporal learned step-size quantization to further enhance performance. Extensive experimental results demonstrate that our method significantly outperforms previous PTQ-based diffusion models while maintaining similar time and data efficiency. Specifically, there is only a 0.05 sFID increase when quantizing both weights and activations of LDM-4 to 4-bit on ImageNet 256x256. Compared to QAT-based methods, our EfficientDM also boasts a 16.2x faster quantization speed with comparable generation quality. Code is available at \\href{https://github.com/ThisisBillhe/EfficientDM}{this hrl}."
                    },
                    {
                        "title": "BiViT: Extremely Compressed Binary Vision Transformer",
                        "abstract": "Model binarization can significantly compress model size, reduce energy consumption, and accelerate inference through efficient bit-wise operations. Although binarizing convolutional neural networks have been extensively studied, there is little work on exploring binarization of vision Transformers which underpin most recent breakthroughs in visual recognition. To this end, we propose to solve two fundamental challenges to push the horizon of Binary Vision Transformers (BiViT). First, the traditional binary method does not take the long-tailed distribution of softmax attention into consideration, bringing large binarization errors in the attention module. To solve this, we propose Softmax-aware Binarization, which dynamically adapts to the data distribution and reduces the error caused by binarization. Second, to better preserve the information of the pretrained model and restore accuracy, we propose a Cross-layer Binarization scheme that decouples the binarization of self-attention and multi-layer perceptrons (MLPs), and Parameterized Weight Scales which introduce learnable scaling factors for weight binarization. Overall, our method performs favorably against state-of-the-arts by 19.8% on the TinyImageNet dataset. On ImageNet, our BiViT achieves a competitive 75.6% Top-1 accuracy over Swin-S model. Additionally, on COCO object detection, our method achieves an mAP of 40.8 with a Swin-T backbone over Cascade Mask R-CNN framework."
                    },
                    {
                        "title": "ZipCache: Accurate and Efficient KV Cache Quantization with Salient Token Identification",
                        "abstract": "KV cache stores key and value states from previous tokens to avoid re-computation, yet it demands substantial storage space, especially for long sequences. Adaptive KV cache compression seeks to discern the saliency of tokens, preserving vital information while aggressively compressing those of less importance. However, previous methods of this approach exhibit significant performance degradation at high compression ratios due to inaccuracies in identifying salient tokens. In this paper, we present ZipCache, an accurate and efficient KV cache quantization method for LLMs. First, we construct a strong baseline for quantizing KV cache. Through the proposed channel-separable tokenwise quantization scheme, the memory overhead of quantization parameters are substantially reduced compared to fine-grained groupwise quantization. To enhance the compression ratio, we propose normalized attention score as an effective metric for identifying salient tokens by considering the lower triangle characteristics of the attention matrix. Moreover, we develop an efficient approximation method that decouples the saliency metric from full attention scores, enabling compatibility with fast attention implementations like FlashAttention. Extensive experiments demonstrate that ZipCache achieves superior compression ratios, fast generation speed and minimal performance losses compared with previous KV cache compression methods. For instance, when evaluating Mistral-7B model on GSM8k dataset, ZipCache is capable of compressing the KV cache by $4.98\\times$, with only a $0.38\\%$ drop in accuracy. In terms of efficiency, ZipCache also showcases a $37.3\\%$ reduction in prefill-phase latency, a $56.9\\%$ reduction in decoding-phase latency, and a $19.8\\%$ reduction in GPU memory usage when evaluating LLaMA3-8B model with a input length of $4096$."
                    },
                    {
                        "title": "PTQD: Accurate Post-Training Quantization for Diffusion Models",
                        "abstract": "Diffusion models have recently dominated image synthesis tasks. However, the iterative denoising process is expensive in computations at inference time, making diffusion models less practical for low-latency and scalable real-world applications. Post-training quantization (PTQ) of diffusion models can significantly reduce the model size and accelerate the sampling process without re-training. Nonetheless, applying existing PTQ methods directly to low-bit diffusion models can significantly impair the quality of generated samples. Specifically, for each denoising step, quantization noise leads to deviations in the estimated mean and mismatches with the predetermined variance schedule. As the sampling process proceeds, the quantization noise may accumulate, resulting in a low signal-to-noise ratio (SNR) during the later denoising steps. To address these challenges, we propose a unified formulation for the quantization noise and diffusion perturbed noise in the quantized denoising process. Specifically, we first disentangle the quantization noise into its correlated and residual uncorrelated parts regarding its full-precision counterpart. The correlated part can be easily corrected by estimating the correlation coefficient. For the uncorrelated part, we subtract the bias from the quantized results to correct the mean deviation and calibrate the denoising variance schedule to absorb the excess variance resulting from quantization. Moreover, we introduce a mixed-precision scheme for selecting the optimal bitwidth for each denoising step. Extensive experiments demonstrate that our method outperforms previous post-training quantized diffusion models, with only a 0.06 increase in FID score compared to full-precision LDM-4 on ImageNet 256x256, while saving 19.9x bit operations. Code is available at https://github.com/ziplab/PTQD."
                    },
                    {
                        "title": "ZipVL: Efficient Large Vision-Language Models with Dynamic Token Sparsification and KV Cache Compression",
                        "abstract": "The efficiency of large vision-language models (LVLMs) is constrained by the computational bottleneck of the attention mechanism during the prefill phase and the memory bottleneck of fetching the key-value (KV) cache in the decoding phase, particularly in scenarios involving high-resolution images or videos. Visual content often exhibits substantial redundancy, resulting in highly sparse attention maps within LVLMs. This sparsity can be leveraged to accelerate attention computation or compress the KV cache through various approaches. However, most studies focus on addressing only one of these bottlenecks and do not adequately support dynamic adjustment of sparsity concerning distinct layers or tasks. In this paper, we present ZipVL, an efficient inference framework designed for LVLMs that resolves both computation and memory bottlenecks through a dynamic ratio allocation strategy of important tokens. This ratio is adaptively determined based on the layer-specific distribution of attention scores, rather than fixed hyper-parameters, thereby improving efficiency for less complex tasks while maintaining high performance for more challenging ones. Then we select important tokens based on their normalized attention scores and perform attention mechanism solely on those important tokens to accelerate the prefill phase. To mitigate the memory bottleneck in the decoding phase, we employ mixed-precision quantization to the KV cache, where high-bit quantization is used for caches of important tokens, while low-bit quantization is applied to those of less importance. Our experiments demonstrate that ZipVL can accelerate the prefill phase by 2.6$\\times$ and reduce GPU memory usage by 50.0%, with a minimal accuracy reduction of only 0.2% on Video-MME benchmark over LongVA-7B model, effectively enhancing the generation efficiency of LVLMs."
                    },
                    {
                        "title": "Dual Grained Quantization: Efficient Fine-Grained Quantization for LLM",
                        "abstract": "Large Language Models (LLMs) pose significant hardware challenges related to memory requirements and computational ability. There are two mainstream quantization schemes for LLMs: coarse-grained ($\\textit{e.g.,}$ channel-wise) quantization and fine-grained ($\\textit{e.g.,}$ group-wise) quantization. Fine-grained quantization has smaller quantization loss, consequently achieving superior performance. However, when applied to weight-activation quantization, it disrupts continuous integer matrix multiplication, leading to inefficient inference. In this paper, we introduce Dual Grained Quantization (DGQ), a novel A8W4 quantization for LLM that maintains superior performance while ensuring fast inference speed. DSQ dequantizes the fine-grained INT4 weight into coarse-grained INT8 representation and preform matrix multiplication using INT8 kernels. Besides, we develop a two-phase grid search algorithm to simplify the determination of fine-grained and coarse-grained quantization scales. We also devise a percentile clipping schema for smoothing the activation outliers without the need for complex optimization techniques. Experimental results demonstrate that DGQ consistently outperforms prior methods across various LLM architectures and a wide range of tasks. Remarkably, by our implemented efficient CUTLASS kernel, we achieve $\\textbf{1.12}$ $\\times$ memory reduction and $\\textbf{3.24}$ $\\times$ speed gains comparing A16W4 implementation. These advancements enable efficient deployment of A8W4 LLMs for real-world applications."
                    },
                    {
                        "title": "ME-Switch: A Memory-Efficient Expert Switching Framework for Large Language Models",
                        "abstract": "The typical process for developing LLMs involves pre-training a general foundation model on massive data, followed by fine-tuning on task-specific data to create specialized experts. Serving these experts poses challenges, as loading all experts onto devices is impractical, and frequent switching between experts in response to user requests incurs substantial I/O costs, increasing latency and expenses. Previous approaches decompose expert weights into pre-trained model weights and residual delta weights, then quantize the delta weights to reduce model size. However, these methods often lead to significant quantization errors at extremely low bitwidths and assume the appropriate model for a user request is known in advance, which is not practical. To address these issues, we introduce ME-Switch, a memory-efficient expert switching framework for LLM serving. ME-Switch uses mixed-precision quantization, selectively quantizing non-salient input channels of delta weights to extremely low bits while keeping salient ones intact, significantly reducing storage demands while maintaining performance. Additionally, we develop a routing method that efficiently directs user queries to the most suitable expert by transforming the model selection problem into a domain classification problem. Extensive experiments show ME-Switch's promising memory efficiency and routing performance. For example, when serving three models from the Mistral-7B family, ME-Switch reduces model size by 1.74x while maintaining nearly lossless performance on instruction, mathematical reasoning, and code generation tasks. Furthermore, ME-Switch can efficiently serve 16 models from the Mistral-7B family on a single NVIDIA A100 GPU."
                    },
                    {
                        "title": "DSText V2: A Comprehensive Video Text Spotting Dataset for Dense and Small Text",
                        "abstract": "Recently, video text detection, tracking, and recognition in natural scenes are becoming very popular in the computer vision community. However, most existing algorithms and benchmarks focus on common text cases (e.g., normal size, density) and single scenario, while ignoring extreme video text challenges, i.e., dense and small text in various scenarios. In this paper, we establish a video text reading benchmark, named DSText V2, which focuses on Dense and Small text reading challenges in the video with various scenarios. Compared with the previous datasets, the proposed dataset mainly include three new challenges: 1) Dense video texts, a new challenge for video text spotters to track and read. 2) High-proportioned small texts, coupled with the blurriness and distortion in the video, will bring further challenges. 3) Various new scenarios, e.g., Game, Sports, etc. The proposed DSText V2 includes 140 video clips from 7 open scenarios, supporting three tasks, i.e., video text detection (Task 1), video text tracking (Task 2), and end-to-end video text spotting (Task 3). In this article, we describe detailed statistical information of the dataset, tasks, evaluation protocols, and the results summaries. Most importantly, a thorough investigation and analysis targeting three unique challenges derived from our dataset are provided, aiming to provide new insights. Moreover, we hope the benchmark will promise video text research in the community. DSText v2 is built upon DSText v1, which was previously introduced to organize the ICDAR 2023 competition for dense and small video text."
                    },
                    {
                        "title": "Towards Accurate Post-training Quantization for Reparameterized Models",
                        "abstract": "Model reparameterization is a widely accepted technique for improving inference speed without compromising performance. However, current Post-training Quantization (PTQ) methods often lead to significant accuracy degradation when applied to reparameterized models. This is primarily caused by channel-specific and sample-specific outliers, which appear only at specific samples and channels and impact on the selection of quantization parameters. To address this issue, we propose RepAPQ, a novel framework that preserves the accuracy of quantized reparameterization models. Different from previous frameworks using Mean Squared Error (MSE) as a measurement, we utilize Mean Absolute Error (MAE) to mitigate the influence of outliers on quantization parameters. Our framework comprises two main components: Quantization Protecting Reparameterization and Across-block Calibration. For effective calibration, Quantization Protecting Reparameterization combines multiple branches into a single convolution with an affine layer. During training, the affine layer accelerates convergence and amplifies the output of the convolution to better accommodate samples with outliers. Additionally, Across-block Calibration leverages the measurement of stage output as supervision to address the gradient problem introduced by MAE and enhance the interlayer correlation with quantization parameters. Comprehensive experiments demonstrate the effectiveness of RepAPQ across various models and tasks. Our framework outperforms previous methods by approximately 1\\% for 8-bit PTQ and 2\\% for 6-bit PTQ, showcasing its superior performance. The code is available at \\url{https://github.com/ilur98/DLMC-QUANT}."
                    },
                    {
                        "title": "DatasetDM: Synthesizing Data with Perception Annotations Using Diffusion Models",
                        "abstract": "Current deep networks are very data-hungry and benefit from training on largescale datasets, which are often time-consuming to collect and annotate. By contrast, synthetic data can be generated infinitely using generative models such as DALL-E and diffusion models, with minimal effort and cost. In this paper, we present DatasetDM, a generic dataset generation model that can produce diverse synthetic images and the corresponding high-quality perception annotations (e.g., segmentation masks, and depth). Our method builds upon the pre-trained diffusion model and extends text-guided image synthesis to perception data generation. We show that the rich latent code of the diffusion model can be effectively decoded as accurate perception annotations using a decoder module. Training the decoder only needs less than 1% (around 100 images) manually labeled images, enabling the generation of an infinitely large annotated dataset. Then these synthetic data can be used for training various perception models for downstream tasks. To showcase the power of the proposed approach, we generate datasets with rich dense pixel-wise labels for a wide range of downstream tasks, including semantic segmentation, instance segmentation, and depth estimation. Notably, it achieves 1) state-of-the-art results on semantic segmentation and instance segmentation; 2) significantly more robust on domain generalization than using the real data alone; and state-of-the-art results in zero-shot segmentation setting; and 3) flexibility for efficient application and novel task composition (e.g., image editing). The project website and code can be found at https://weijiawu.github.io/DatasetDM_page/ and https://github.com/showlab/DatasetDM, respectively"
                    },
                    {
                        "title": "DragAnything: Motion Control for Anything using Entity Representation",
                        "abstract": "We introduce DragAnything, which utilizes a entity representation to achieve motion control for any object in controllable video generation. Comparison to existing motion control methods, DragAnything offers several advantages. Firstly, trajectory-based is more userfriendly for interaction, when acquiring other guidance signals (e.g., masks, depth maps) is labor-intensive. Users only need to draw a line (trajectory) during interaction. Secondly, our entity representation serves as an open-domain embedding capable of representing any object, enabling the control of motion for diverse entities, including background. Lastly, our entity representation allows simultaneous and distinct motion control for multiple objects. Extensive experiments demonstrate that our DragAnything achieves state-of-the-art performance for FVD, FID, and User Study, particularly in terms of object motion control, where our method surpasses the previous methods (e.g., DragNUWA) by 26% in human voting."
                    },
                    {
                        "title": "Paragraph-to-Image Generation with Information-Enriched Diffusion Model",
                        "abstract": "Text-to-image (T2I) models have recently experienced rapid development, achieving astonishing performance in terms of fidelity and textual alignment capabilities. However, given a long paragraph (up to 512 words), these generation models still struggle to achieve strong alignment and are unable to generate images depicting complex scenes. In this paper, we introduce an information-enriched diffusion model for paragraph-to-image generation task, termed ParaDiffusion, which delves into the transference of the extensive semantic comprehension capabilities of large language models to the task of image generation. At its core is using a large language model (e.g., Llama V2) to encode long-form text, followed by fine-tuning with LORA to alignthe text-image feature spaces in the generation task. To facilitate the training of long-text semantic alignment, we also curated a high-quality paragraph-image pair dataset, namely ParaImage. This dataset contains a small amount of high-quality, meticulously annotated data, and a large-scale synthetic dataset with long text descriptions being generated using a vision-language model. Experiments demonstrate that ParaDiffusion outperforms state-of-the-art models (SD XL, DeepFloyd IF) on ViLG-300 and ParaPrompts, achieving up to 15% and 45% human voting rate improvements for visual appeal and text faithfulness, respectively. The code and dataset will be released to foster community research on long-text alignment."
                    }
                ]
            },
            "4d68cc4f-506b-43ca-9ad6-ddc25599777a": {
                "pk": "4d68cc4f-506b-43ca-9ad6-ddc25599777a",
                "name": "Gholamreza Haffari",
                "collaborators": [
                    "Poorya Zaremoodi",
                    "Zhuang Li",
                    "Trevor Cohn",
                    "Sameen Maruf",
                    "Lizhen Qu",
                    "Ehsan Shareghi",
                    "Xuanli He",
                    "Mohammad Norouzi",
                    "Fahimeh Saleh",
                    "Geetanjali Rakshit"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Translation",
                    "Multi-task Learning",
                    "Semantic Parsing"
                ],
                "publications": [
                    {
                        "title": "Incorporating Syntactic Uncertainty in Neural Machine Translation with Forest-to-Sequence Model",
                        "abstract": "Incorporating syntactic information in Neural Machine Translation models is a method to compensate their requirement for a large amount of parallel training text, especially for low-resource language pairs. Previous works on using syntactic information provided by (inevitably error-prone) parsers has been promising. In this paper, we propose a forest-to-sequence Attentional Neural Machine Translation model to make use of exponentially many parse trees of the source sentence to compensate for the parser errors. Our method represents the collection of parse trees as a packed forest, and learns a neural attentional transduction model from the forest to the target sentence. Experiments on English to German, Chinese and Persian translation show the superiority of our method over the tree-to-sequence and vanilla sequence-to-sequence neural translation models."
                    },
                    {
                        "title": "Learning to Multi-Task Learn for Better Neural Machine Translation",
                        "abstract": "Scarcity of parallel sentence pairs is a major challenge for training high quality neural machine translation (NMT) models in bilingually low-resource scenarios, as NMT is data-hungry. Multi-task learning is an elegant approach to inject linguistic-related inductive biases into NMT, using auxiliary syntactic and semantic tasks, to improve generalisation. The challenge, however, is to devise effective training schedules, prescribing when to make use of the auxiliary tasks during the training process to fill the knowledge gaps of the main translation task, a setting referred to as biased-MTL. Current approaches for the training schedule are based on hand-engineering heuristics, whose effectiveness vary in different MTL settings. We propose a novel framework for learning the training schedule, ie learning to multi-task learn, for the MTL setting of interest. We formulate the training schedule as a Markov decision process which paves the way to employ policy learning methods to learn the scheduling policy. We effectively and efficiently learn the training schedule policy within the imitation learning framework using an oracle policy algorithm that dynamically sets the importance weights of auxiliary tasks based on their contributions to the generalisability of the main NMT task. Experiments on low-resource NMT settings show the resulting automatically learned training schedulers are competitive with the best heuristics, and lead to up to +1.1 BLEU score improvements."
                    },
                    {
                        "title": "Document Context Neural Machine Translation with Memory Networks",
                        "abstract": "We present a document-level neural machine translation model which takes both source and target document context into account using memory networks. We model the problem as a structured prediction problem with interdependencies among the observed and hidden variables, i.e., the source sentences and their unobserved target translations in the document. The resulting structured prediction problem is tackled with a neural translation model equipped with two memory components, one each for the source and target side, to capture the documental interdependencies. We train the model end-to-end, and propose an iterative decoding algorithm based on block coordinate descent. Experimental results of English translations from French, German, and Estonian documents show that our model is effective in exploiting both source and target document context, and statistically significantly outperforms the previous work in terms of BLEU and METEOR."
                    },
                    {
                        "title": "Neural Machine Translation for Bilingually Scarce Scenarios: A Deep Multi-task Learning Approach",
                        "abstract": "Neural machine translation requires large amounts of parallel training text to learn a reasonable-quality translation model. This is particularly inconvenient for language pairs for which enough parallel text is not available. In this paper, we use monolingual linguistic resources in the source side to address this challenging problem based on a multi-task learning approach. More specifically, we scaffold the machine translation task on auxiliary tasks including semantic parsing, syntactic parsing, and named-entity recognition. This effectively injects semantic and/or syntactic knowledge into the translation model, which would otherwise require a large amount of training bitext. We empirically evaluate and show the effectiveness of our multi-task learning approach on three translation tasks: English-to-French, English-to-Farsi, and English-to-Vietnamese."
                    },
                    {
                        "title": "Active Learning for Multilingual Semantic Parser",
                        "abstract": "Current multilingual semantic parsing (MSP) datasets are almost all collected by translating the utterances in the existing datasets from the resource-rich language to the target language. However, manual translation is costly. To reduce the translation effort, this paper proposes the first active learning procedure for MSP (AL-MSP). AL-MSP selects only a subset from the existing datasets to be translated. We also propose a novel selection method that prioritizes the examples diversifying the logical form structures with more lexical choices, and a novel hyperparameter tuning method that needs no extra annotation cost. Our experiments show that AL-MSP significantly reduces translation costs with ideal selection methods. Our selection method with proper hyperparameters yields better parsing performance than the other baselines on two multilingual datasets."
                    },
                    {
                        "title": "Prepositional Attachment Disambiguation Using Bilingual Parsing and Alignments",
                        "abstract": "In this paper, we attempt to solve the problem of Prepositional Phrase (PP) attachments in English. The motivation for the work comes from NLP applications like Machine Translation, for which, getting the correct attachment of prepositions is very crucial. The idea is to correct the PP-attachments for a sentence with the help of alignments from parallel data in another language. The novelty of our work lies in the formulation of the problem into a dual decomposition based algorithm that enforces agreement between the parse trees from two languages as a constraint. Experiments were performed on the English-Hindi language pair and the performance improved by 10% over the baseline, where the baseline is the attachment predicted by the MSTParser model trained for English."
                    },
                    {
                        "title": "Few-Shot Semantic Parsing for New Predicates",
                        "abstract": "In this work, we investigate the problems of semantic parsing in a few-shot learning setting. In this setting, we are provided with utterance-logical form pairs per new predicate. The state-of-the-art neural semantic parsers achieve less than 25% accuracy on benchmark datasets when k= 1. To tackle this problem, we proposed to i) apply a designated meta-learning method to train the model; ii) regularize attention scores with alignment statistics; iii) apply a smoothing technique in pre-training. As a result, our method consistently outperforms all the baselines in both one and two-shot settings."
                    },
                    {
                        "title": "Learning Coupled Policies for Simultaneous Machine Translation using Imitation Learning",
                        "abstract": "We present a novel approach to efficiently learn a simultaneous translation model with coupled programmer-interpreter policies. First, wepresent an algorithmic oracle to produce oracle READ/WRITE actions for training bilingual sentence-pairs using the notion of word alignments. This oracle actions are designed to capture enough information from the partial input before writing the output. Next, we perform a coupled scheduled sampling to effectively mitigate the exposure bias when learning both policies jointly with imitation learning. Experiments on six language-pairs show our method outperforms strong baselines in terms of translation quality while keeping the translation delay low."
                    },
                    {
                        "title": "Context Dependent Semantic Parsing: A Survey",
                        "abstract": "Semantic parsing is the task of translating natural language utterances into machine-readable meaning representations. Currently, most semantic parsing methods are not able to utilize contextual information (e.g. dialogue and comments history), which has a great potential to boost semantic parsing performance. To address this issue, context dependent semantic parsing has recently drawn a lot of attention. In this survey, we investigate progress on the methods for the context dependent semantic parsing, together with the current datasets and tasks. We then point out open problems and challenges for future research in this area. The collected resources for this topic are available at:https://github.com/zhuang-li/Contextual-Semantic-Parsing-Paper-List."
                    },
                    {
                        "title": "RedApt: An Adaptor for wav2vec 2 Encoding \\\\ Faster and Smaller Speech Translation without Quality Compromise",
                        "abstract": "Pre-trained speech Transformers in speech translation (ST) have facilitated state-of-the-art (SotA) results; yet, using such encoders is computationally expensive. To improve this, we present a novel Reducer Adaptor block, RedApt, that could be seamlessly integrated within any Transformer-based speech encoding architecture. Integrating the pretrained wav2vec 2 speech encoder with RedAptbrings 41% speedup, 33% memory reduction with 24% fewer FLOPs at inference. To our positive surprise, our ST model with RedApt outperforms the SotA architecture by an average of 0.68 BLEU score on 8 language pairs from Must-C."
                    },
                    {
                        "title": "Structured Prediction of Sequences and Trees using Infinite Contexts",
                        "abstract": "Linguistic structures exhibit a rich array of global phenomena, however commonly used Markov models are unable to adequately describe these phenomena due to their strong locality assumptions. We propose a novel hierarchical model for structured prediction over sequences and trees which exploits global context by conditioning each generation decision on an unbounded context of prior decisions. This builds on the success of Markov models but without imposing a fixed bound in order to better represent global phenomena. To facilitate learning of this large and unbounded model, we use a hierarchical Pitman-Yor process prior which provides a recursive form of smoothing. We propose prediction algorithms based on A* and Markov Chain Monte Carlo sampling. Empirical results demonstrate the potential of our model compared to baseline finite-context Markov models on part-of-speech tagging and syntactic parsing."
                    },
                    {
                        "title": "Novel Bernstein-like Concentration Inequalities for the Missing Mass",
                        "abstract": "We are concerned with obtaining novel concentration inequalities for the missing mass, i.e. the total probability mass of the outcomes not observed in the sample. We not only derive - for the first time - distribution-free Bernstein-like deviation bounds with sublinear exponents in deviation size for missing mass, but also improve the results of McAllester and Ortiz (2003) andBerend and Kontorovich (2013, 2012) for small deviations which is the most interesting case in learning theory. It is known that the majority of standard inequalities cannot be directly used to analyze heterogeneous sums i.e. sums whose terms have large difference in magnitude. Our generic and intuitive approach shows that the heterogeneity issue introduced in McAllester and Ortiz (2003) is resolvable at least in the case of missing mass via regulating the terms using our novel thresholding technique."
                    },
                    {
                        "title": "A Latent Variable Recurrent Neural Network for Discourse Relation Language Models",
                        "abstract": "This paper presents a novel latent variable recurrent neural network architecture for jointly modeling sequences of words and (possibly latent) discourse relations between adjacent sentences. A recurrent neural network generates individual words, thus reaping the benefits of discriminatively-trained vector representations. The discourse relations are represented with a latent variable, which can be predicted or marginalized, depending on the task. The resulting model can therefore employ a training objective that includes not only discourse relation classification, but also word prediction. As a result, it outperforms state-of-the-art alternatives for two tasks: implicit discourse relation classification in the Penn Discourse Treebank, and dialog act classification in the Switchboard corpus. Furthermore, by marginalizing over latent discourse relations at test time, we obtain a discourse informed language model, which improves over a strong LSTM baseline."
                    },
                    {
                        "title": "Graph-to-Sequence Learning using Gated Graph Neural Networks",
                        "abstract": "Many NLP applications can be framed as a graph-to-sequence learning problem. Previous work proposing neural architectures on this setting obtained promising results compared to grammar-based approaches but still rely on linearisation heuristics and/or standard recurrent networks to achieve the best performance. In this work, we propose a new model that encodes the full structural information contained in the graph. Our architecture couples the recently proposed Gated Graph Neural Networks with an input transformation that allows nodes and edges to have their own hidden representations, while tackling the parameter explosion problem present in previous work. Experimental results show that our model outperforms strong baselines in generation from AMR graphs and syntax-based neural machine translation."
                    },
                    {
                        "title": "Sequence to Sequence Mixture Model for Diverse Machine Translation",
                        "abstract": "Sequence to sequence (SEQ2SEQ) models often lack diversity in their generated translations. This can be attributed to the limitation of SEQ2SEQ models in capturing lexical and syntactic variations in a parallel corpus resulting from different styles, genres, topics, or ambiguity of the translation process. In this paper, we develop a novel sequence to sequence mixture (S2SMIX) model that improves both translation diversity and quality by adopting a committee of specialized translation models rather than a single translation model. Each mixture component selects its own training dataset via optimization of the marginal loglikelihood, which leads to a soft clustering of the parallel corpus. Experiments on four language pairs demonstrate the superiority of our mixture model compared to a SEQ2SEQ baseline with standard or diversity-boosted beam search. Our mixture model uses negligible additional parameters and incurs no extra computation cost during decoding."
                    },
                    {
                        "title": "A Survey on Document-level Neural Machine Translation: Methods and Evaluation",
                        "abstract": "Machine translation (MT) is an important task in natural language processing (NLP) as it automates the translation process and reduces the reliance on human translators. With the resurgence of neural networks, the translation quality surpasses that of the translations obtained using statistical techniques for most language-pairs. Up until a few years ago, almost all of the neural translation models translated sentences independently, without incorporating the wider document-context and inter-dependencies among the sentences. The aim of this survey paper is to highlight the major works that have been undertaken in the space of document-level machine translation after the neural revolution, so that researchers can recognise the current state and future directions of this field. We provide an organisation of the literature based on novelties in modelling and architectures as well as training and decoding strategies. In addition, we cover evaluation strategies that have been introduced to account for the improvements in document MT, including automatic metrics and discourse-targeted test sets. We conclude by presenting possible avenues for future exploration in this research field."
                    },
                    {
                        "title": "Dynamic Programming Encoding for Subword Segmentation in Neural Machine Translation",
                        "abstract": "This paper introduces Dynamic Programming Encoding (DPE), a new segmentation algorithm for tokenizing sentences into subword units. We view the subword segmentation of output sentences as a latent variable that should be marginalized out for learning and inference. A mixed character-subword transformer is proposed, which enables exact log marginal likelihood estimation and exact MAP inference to find target segmentations with maximum posterior probability. DPE uses a lightweight mixed character-subword transformer as a means of pre-processing parallel data to segment output sentences using dynamic programming. Empirical results on machine translation suggest that DPE is effective for segmenting output sentences and can be combined with BPE dropout for stochastic segmentation of source sentences. DPE achieves an average improvement of 0.9 BLEU over BPE (Sennrich et al., 2016) and an average improvement of 0.55 BLEU over BPE dropout (Provilkov et al., 2019) on several WMT datasets including English <=> (German, Romanian, Estonian, Finnish, Hungarian)."
                    },
                    {
                        "title": "Effective Unsupervised Domain Adaptation with Adversarially Trained Language Models",
                        "abstract": "Recent work has shown the importance of adaptation of broad-coverage contextualised embedding models on the domain of the target task of interest. Current self-supervised adaptation methods are simplistic, as the training signal comes from a small percentage of \\emph{randomly} masked-out tokens. In this paper, we show that careful masking strategies can bridge the knowledge gap of masked language models (MLMs) about the domains more effectively by allocating self-supervision where it is needed. Furthermore, we propose an effective training strategy by adversarially masking out those tokens which are harder to reconstruct by the underlying MLM. The adversarial objective leads to a challenging combinatorial optimisation problem over \\emph{subsets} of tokens, which we tackle efficiently through relaxation to a variational lowerbound and dynamic programming. On six unsupervised domain adaptation tasks involving named entity recognition, our method strongly outperforms the random masking strategy and achieves up to +1.64 F1 score improvements."
                    },
                    {
                        "title": "Collective Wisdom: Improving Low-resource Neural Machine Translation using Adaptive Knowledge Distillation",
                        "abstract": "Scarcity of parallel sentence-pairs poses a significant hurdle for training high-quality Neural Machine Translation (NMT) models in bilingually low-resource scenarios. A standard approach is transfer learning, which involves taking a model trained on a high-resource language-pair and fine-tuning it on the data of the low-resource MT condition of interest. However, it is not clear generally which high-resource language-pair offers the best transfer learning for the target MT setting. Furthermore, different transferred models may have complementary semantic and/or syntactic strengths, hence using only one model may be sub-optimal. In this paper, we tackle this problem using knowledge distillation, where we propose to distill the knowledge of ensemble of teacher models to a single student model. As the quality of these teacher models varies, we propose an effective adaptive knowledge distillation approach to dynamically adjust the contribution of the teacher models during the distillation process. Experiments on transferring from a collection of six language pairs from IWSLT to five low-resource language-pairs from TED Talks demonstrate the effectiveness of our approach, achieving up to +0.9 BLEU score improvement compared to strong baselines."
                    },
                    {
                        "title": "Domain Adaptative Causality Encoder",
                        "abstract": "Current approaches which are mainly based on the extraction of low-level relations among individual events are limited by the shortage of publicly available labelled data. Therefore, the resulting models perform poorly when applied to a distributionally different domain for which labelled data did not exist at the time of training. To overcome this limitation, in this paper, we leverage the characteristics of dependency trees and adversarial learning to address the tasks of adaptive causality identification and localisation. The term adaptive is used since the training and test data come from two distributionally different datasets, which to the best of our knowledge, this work is the first to address. Moreover, we present a new causality dataset, namely MedCaus, which integrates all types of causality in the text. Our experiments on four different benchmark causality datasets demonstrate the superiority of our approach over the existing baselines, by up to 7% improvement, on the tasks of identification and localisation of the causal relations from the text."
                    }
                ]
            },
            "1a1cf719-fdf6-4916-85fc-85b14c0cd6d5": {
                "pk": "1a1cf719-fdf6-4916-85fc-85b14c0cd6d5",
                "name": "Bohan Zhuang",
                "collaborators": [
                    "Chunhua Shen",
                    "Jianfei Cai",
                    "Jing Liu",
                    "Zizheng Pan",
                    "Ian Reid",
                    "Peng Chen",
                    "Qi Wu",
                    "Luis Guerra",
                    "Tom Drummond",
                    "Xinyu Wang"
                ],
                "domain": [
                    "Computer Vision",
                    "Neural Networks",
                    "Model Compression",
                    "Multi-task Learning"
                ],
                "publications": [
                    {
                        "title": "Visual Tracking via Shallow and Deep Collaborative Model",
                        "abstract": "In this paper, we propose a robust tracking method based on the collaboration of a generative model and a discriminative classifier, where features are learned by shallow and deep architectures, respectively. For the generative model, we introduce a block-based incremental learning scheme, in which a local binary mask is constructed to deal with occlusion. The similarity degrees between the local patches and their corresponding subspace are integrated to formulate a more accurate global appearance model. In the discriminative model, we exploit the advances of deep learning architectures to learn generic features which are robust to both background clutters and foreground appearance variations. To this end, we first construct a discriminative training set from auxiliary video sequences. A deep classification neural network is then trained offline on this training set. Through online fine-tuning, both the hierarchical feature extractor and the classifier can be adapted to the appearance change of the target for effective online tracking. The collaboration of these two models achieves a good balance in handling occlusion and target appearance change, which are two contradictory challenging factors in visual tracking. Both quantitative and qualitative evaluations against several state-of-the-art algorithms on challenging image sequences demonstrate the accuracy and the robustness of the proposed tracker."
                    },
                    {
                        "title": "FATNN: Fast and Accurate Ternary Neural Networks",
                        "abstract": "Ternary Neural Networks (TNNs) have received much attention due to being potentially orders of magnitude faster in inference, as well as more power efficient, than full-precision counterparts. However, 2 bits are required to encode the ternary representation with only 3 quantization levels leveraged. As a result, conventional TNNs have similar memory consumption and speed compared with the standard 2-bit models, but have worse representational capability. Moreover, there is still a significant gap in accuracy between TNNs and full-precision networks, hampering their deployment to real applications. To tackle these two challenges, in this work, we first show that, under some mild constraints, computational complexity of the ternary inner product can be reduced by a factor of 2. Second, to mitigate the performance gap, we elaborately design an implementation-dependent ternary quantization algorithm. The proposed framework is termed Fast and Accurate Ternary Neural Networks (FATNN). Experiments on image classification demonstrate that our FATNN surpasses the state-of-the-arts by a significant margin in accuracy. More importantly, speedup evaluation compared with various precisions is analyzed on several platforms, which serves as a strong benchmark for further research."
                    },
                    {
                        "title": "Sharpness-aware Quantization for Deep Neural Networks",
                        "abstract": "Network quantization is a dominant paradigm of model compression. However, the abrupt changes in quantized weights during training often lead to severe loss fluctuations and result in a sharp loss landscape, making the gradients unstable and thus degrading the performance. Recently, Sharpness-Aware Minimization (SAM) has been proposed to smooth the loss landscape and improve the generalization performance of the models. Nevertheless, directly applying SAM to the quantized models can lead to perturbation mismatch or diminishment issues, resulting in suboptimal performance. In this paper, we propose a novel method, dubbed Sharpness-Aware Quantization (SAQ), to explore the effect of SAM in model compression, particularly quantization for the first time. Specifically, we first provide a unified view of quantization and SAM by treating them as introducing quantization noises and adversarial perturbations to the model weights, respectively. According to whether the noise and perturbation terms depend on each other, SAQ can be formulated into three cases, which are analyzed and compared comprehensively. Furthermore, by introducing an efficient training strategy, SAQ only incurs a little additional training overhead compared with the default optimizer (e.g., SGD or AdamW). Extensive experiments on both convolutional neural networks and Transformers across various datasets (i.e., ImageNet, CIFAR-10/100, Oxford Flowers-102, Oxford-IIIT Pets) show that SAQ improves the generalization performance of the quantized models, yielding the SOTA results in uniform quantization. For example, on ImageNet, SAQ outperforms AdamW by 1.2% on the Top-1 accuracy for 4-bit ViT-B/16. Our 4-bit ResNet-50 surpasses the previous SOTA method by 0.9% on the Top-1 accuracy."
                    },
                    {
                        "title": "FocusFormer: Focusing on What We Need via Architecture Sampler",
                        "abstract": "Vision Transformers (ViTs) have underpinned the recent breakthroughs in computer vision. However, designing the architectures of ViTs is laborious and heavily relies on expert knowledge. To automate the design process and incorporate deployment flexibility, one-shot neural architecture search decouples the supernet training and architecture specialization for diverse deployment scenarios. To cope with an enormous number of sub-networks in the supernet, existing methods treat all architectures equally important and randomly sample some of them in each update step during training. During architecture search, these methods focus on finding architectures on the Pareto frontier of performance and resource consumption, which forms a gap between training and deployment. In this paper, we devise a simple yet effective method, called FocusFormer, to bridge such a gap. To this end, we propose to learn an architecture sampler to assign higher sampling probabilities to those architectures on the Pareto frontier under different resource constraints during supernet training, making them sufficiently optimized and hence improving their performance. During specialization, we can directly use the well-trained architecture sampler to obtain accurate architectures satisfying the given resource constraint, which significantly improves the search efficiency. Extensive experiments on CIFAR-100 and ImageNet show that our FocusFormer is able to improve the performance of the searched architectures while significantly reducing the search cost. For example, on ImageNet, our FocusFormer-Ti with 1.4G FLOPs outperforms AutoFormer-Ti by 0.5% in terms of the Top-1 accuracy."
                    },
                    {
                        "title": "Stitchable Neural Networks",
                        "abstract": "The public model zoo containing enormous powerful pretrained model families (e.g., ResNet/DeiT) has reached an unprecedented scope than ever, which significantly contributes to the success of deep learning. As each model family consists of pretrained models with diverse scales (e.g., DeiT-Ti/S/B), it naturally arises a fundamental question of how to efficiently assemble these readily available models in a family for dynamic accuracy-efficiency trade-offs at runtime. To this end, we present Stitchable Neural Networks (SN-Net), a novel scalable and efficient framework for model deployment. It cheaply produces numerous networks with different complexity and performance trade-offs given a family of pretrained neural networks, which we call anchors. Specifically, SN-Net splits the anchors across the blocks/layers and then stitches them together with simple stitching layers to map the activations from one anchor to another. With only a few epochs of training, SN-Net effectively interpolates between the performance of anchors with varying scales. At runtime, SN-Net can instantly adapt to dynamic resource constraints by switching the stitching positions. Extensive experiments on ImageNet classification demonstrate that SN-Net can obtain on-par or even better performance than many individually trained networks while supporting diverse deployment scenarios. For example, by stitching Swin Transformers, we challenge hundreds of models in Timm model zoo with a single network. We believe this new elastic model framework can serve as a strong baseline for further research in wider communities."
                    },
                    {
                        "title": "Fast Vision Transformers with HiLo Attention",
                        "abstract": "Vision Transformers (ViTs) have triggered the most recent and significant breakthroughs in computer vision. Their efficient designs are mostly guided by the indirect metric of computational complexity, i.e., FLOPs, which however has a clear gap with the direct metric such as throughput. Thus, we propose to use the direct speed evaluation on the target platform as the design principle for efficient ViTs. Particularly, we introduce LITv2, a simple and effective ViT which performs favourably against the existing state-of-the-art methods across a spectrum of different model sizes with faster speed. At the core of LITv2 is a novel self-attention mechanism, which we dub HiLo. HiLo is inspired by the insight that high frequencies in an image capture local fine details and low frequencies focus on global structures, whereas a multi-head self-attention layer neglects the characteristic of different frequencies. Therefore, we propose to disentangle the high/low frequency patterns in an attention layer by separating the heads into two groups, where one group encodes high frequencies via self-attention within each local window, and another group encodes low frequencies by performing global attention between the average-pooled low-frequency keys and values from each window and each query position in the input feature map. Benefiting from the efficient design for both groups, we show that HiLo is superior to the existing attention mechanisms by comprehensively benchmarking FLOPs, speed and memory consumption on GPUs and CPUs. For example, HiLo is 1.4x faster than spatial reduction attention and 1.6x faster than local window attention on CPUs. Powered by HiLo, LITv2 serves as a strong backbone for mainstream vision tasks including image classification, dense detection and segmentation. Code is available at https://github.com/ziplab/LITv2."
                    },
                    {
                        "title": "Training Compact Neural Networks with Binary Weights and Low Precision Activations",
                        "abstract": "In this paper, we propose to train a network with binary weights and low-bitwidth activations, designed especially for mobile devices with limited power consumption. Most previous works on quantizing CNNs uncritically assume the same architecture, though with reduced precision. However, we take the view that for best performance it is possible (and even likely) that a different architecture may be better suited to dealing with low precision weights and activations.   Specifically, we propose a \"network expansion\" strategy in which we aggregate a set of homogeneous low-precision branches to implicitly reconstruct the full-precision intermediate feature maps. Moreover, we also propose a group-wise feature approximation strategy which is very flexible and highly accurate. Experiments on ImageNet classification tasks demonstrate the superior performance of the proposed model, named Group-Net, over various popular architectures. In particular, with binary weights and activations, we outperform the previous best binary neural network in terms of accuracy as well as saving more than 5 times computational complexity on ImageNet with ResNet-18 and ResNet-50."
                    },
                    {
                        "title": "Automatic Pruning for Quantized Neural Networks",
                        "abstract": "Neural network quantization and pruning are two techniques commonly used to reduce the computational complexity and memory footprint of these models for deployment. However, most existing pruning strategies operate on full-precision and cannot be directly applied to discrete parameter distributions after quantization. In contrast, we study a combination of these two techniques to achieve further network compression. In particular, we propose an effective pruning strategy for selecting redundant low-precision filters. Furthermore, we leverage Bayesian optimization to efficiently determine the pruning ratio for each layer. We conduct extensive experiments on CIFAR-10 and ImageNet with various architectures and precisions. In particular, for ResNet-18 on ImageNet, we prune 26.12% of the model size with Binarized Neural Network quantization, achieving a top-1 classification accuracy of 47.32% in a model of 2.47 MB and 59.30% with a 2-bit DoReFa-Net in 4.36 MB."
                    },
                    {
                        "title": "Switchable Precision Neural Networks",
                        "abstract": "Instantaneous and on demand accuracy-efficiency trade-off has been recently explored in the context of neural networks slimming. In this paper, we propose a flexible quantization strategy, termed Switchable Precision neural Networks (SP-Nets), to train a shared network capable of operating at multiple quantization levels. At runtime, the network can adjust its precision on the fly according to instant memory, latency, power consumption and accuracy demands. For example, by constraining the network weights to 1-bit with switchable precision activations, our shared network spans from BinaryConnect to Binarized Neural Network, allowing to perform dot-products using only summations or bit operations. In addition, a self-distillation scheme is proposed to increase the performance of the quantized switches. We tested our approach with three different quantizers and demonstrate the performance of SP-Nets against independently trained quantized models in classification accuracy for Tiny ImageNet and ImageNet datasets using ResNet-18 and MobileNet architectures."
                    },
                    {
                        "title": "ModaVerse: Efficiently Transforming Modalities with LLMs",
                        "abstract": "Humans possess the capability to comprehend diverse modalities and seamlessly transfer information between them. In this work, we introduce ModaVerse, a Multi-modal Large Language Model (MLLM) capable of comprehending and transforming content across various modalities including images, videos, and audio. Predominant MLLM frameworks have largely relied on the alignment of latent spaces of textual and non-textual features. This alignment process, which synchronizes a language model trained on textual data with encoders and decoders trained on multi-modal data, often necessitates extensive training of several projection layers in multiple stages. Inspired by LLM-as-agent methodologies, we propose a novel Input/Output (I/O) alignment mechanism that operates directly at the level of natural language. It aligns the LLM's output with the input of generative models, avoiding the complexities associated with latent feature alignments, and simplifying the multiple training stages of existing MLLMs into a single, efficient process. This conceptual advancement leads to significant reductions in both data and computational costs. By conducting experiments on several benchmarks, we demonstrate that our approach attains comparable performance with the state of the art while achieving considerable efficiencies in data usage and training duration."
                    },
                    {
                        "title": "Streaming Video Diffusion: Online Video Editing with Diffusion Models",
                        "abstract": "We present a novel task called online video editing, which is designed to edit \\textbf{streaming} frames while maintaining temporal consistency. Unlike existing offline video editing assuming all frames are pre-established and accessible, online video editing is tailored to real-life applications such as live streaming and online chat, requiring (1) fast continual step inference, (2) long-term temporal modeling, and (3) zero-shot video editing capability. To solve these issues, we propose Streaming Video Diffusion (SVDiff), which incorporates the compact spatial-aware temporal recurrence into off-the-shelf Stable Diffusion and is trained with the segment-level scheme on large-scale long videos. This simple yet effective setup allows us to obtain a single model that is capable of executing a broad range of videos and editing each streaming frame with temporal coherence. Our experiments indicate that our model can edit long, high-quality videos with remarkable results, achieving a real-time inference speed of 15.2 FPS at a resolution of 512x512."
                    },
                    {
                        "title": "Towards Context-aware Interaction Recognition",
                        "abstract": "Recognizing how objects interact with each other is a crucial task in visual recognition. If we define the context of the interaction to be the objects involved, then most current methods can be categorized as either: (i) training a single classifier on the combination of the interaction and its context; or (ii) aiming to recognize the interaction independently of its explicit context. Both methods suffer limitations: the former scales poorly with the number of combinations and fails to generalize to unseen combinations, while the latter often leads to poor interaction recognition performance due to the difficulty of designing a context-independent interaction classifier. To mitigate those drawbacks, this paper proposes an alternative, context-aware interaction recognition framework. The key to our method is to explicitly construct an interaction classifier which combines the context, and the interaction. The context is encoded via word2vec into a semantic space, and is used to derive a classification result for the interaction.   The proposed method still builds one classifier for one interaction (as per type (ii) above), but the classifier built is adaptive to context via weights which are context dependent. The benefit of using the semantic space is that it naturally leads to zero-shot generalizations in which semantically similar contexts (subjectobject pairs) can be recognized as suitable contexts for an interaction, even if they were not observed in the training set."
                    },
                    {
                        "title": "Fully Quantized Image Super-Resolution Networks",
                        "abstract": "With the rising popularity of intelligent mobile devices, it is of great practical significance to develop accurate, realtime and energy-efficient image Super-Resolution (SR) inference methods. A prevailing method for improving the inference efficiency is model quantization, which allows for replacing the expensive floating-point operations with efficient fixed-point or bitwise arithmetic. To date, it is still challenging for quantized SR frameworks to deliver feasible accuracy-efficiency trade-off. Here, we propose a Fully Quantized image Super-Resolution framework (FQSR) to jointly optimize efficiency and accuracy. In particular, we target on obtaining end-to-end quantized models for all layers, especially including skip connections, which was rarely addressed in the literature. We further identify training obstacles faced by low-bit SR networks and propose two novel methods accordingly. The two difficulites are caused by 1) activation and weight distributions being vastly distinctive in different layers; 2) the inaccurate approximation of the quantization. We apply our quantization scheme on multiple mainstream super-resolution architectures, including SRResNet, SRGAN and EDSR. Experimental results show that our FQSR using low bits quantization can achieve on par performance compared with the full-precision counterparts on five benchmark datasets and surpass state-of-the-art quantized SR methods with significantly reduced computational cost and memory consumption."
                    },
                    {
                        "title": "SwitchGPT: Adapting Large Language Models for Non-Text Outputs",
                        "abstract": "Large Language Models (LLMs), primarily trained on text-based datasets, exhibit exceptional proficiencies in understanding and executing complex linguistic instructions via text outputs. However, they falter when requests to generate non-text ones. Concurrently, modality conversion models, such as text-to-image, despite generating high-quality images, suffer from a lack of extensive textual pretraining. As a result, these models are only capable of accommodating specific image descriptions rather than comprehending more complex instructions. To bridge this gap, we propose a novel approach, \\methodname, from a modality conversion perspective that evolves a text-based LLM into a multi-modal one. We specifically employ a minimal dataset to instruct LLMs to recognize the intended output modality as directed by the instructions. Consequently, the adapted LLM can effectively summon various off-the-shelf modality conversion models from the model zoos to generate non-text responses. This circumvents the necessity for complicated pretraining that typically requires immense quantities of paired multi-modal data, while simultaneously inheriting the extensive knowledge of LLMs and the ability of high-quality generative models. To evaluate and compare the adapted multi-modal LLM with its traditional counterparts, we have constructed a multi-modal instruction benchmark that solicits diverse modality outputs. The experiment results reveal that, with minimal training, LLMs can be conveniently adapted to comprehend requests for non-text responses, thus achieving higher flexibility in multi-modal scenarios. Code and data will be made available at https://github.com/xinke-wang/SwitchGPT."
                    },
                    {
                        "title": "Fast Training of Triplet-based Deep Binary Embedding Networks",
                        "abstract": "In this paper, we aim to learn a mapping (or embedding) from images to a compact binary space in which Hamming distances correspond to a ranking measure for the image retrieval task.   We make use of a triplet loss because this has been shown to be most effective for ranking problems.   However, training in previous works can be prohibitively expensive due to the fact that optimization is directly performed on the triplet space, where the number of possible triplets for training is cubic in the number of training examples.   To address this issue, we propose to formulate high-order binary codes learning as a multi-label classification problem by explicitly separating learning into two interleaved stages.   To solve the first stage, we design a large-scale high-order binary codes inference algorithm to reduce the high-order objective to a standard binary quadratic problem such that graph cuts can be used to efficiently infer the binary code which serve as the label of each training datum.   In the second stage we propose to map the original image to compact binary codes via carefully designed deep convolutional neural networks (CNNs) and the hashing function fitting can be solved by training binary CNN classifiers.   An incremental/interleaved optimization strategy is proffered to ensure that these two steps are interactive with each other during training for better accuracy.   We conduct experiments on several benchmark datasets, which demonstrate both improved training time (by as much as two orders of magnitude) as well as producing state-of-the-art hashing for various retrieval tasks."
                    },
                    {
                        "title": "Less is More: Pay Less Attention in Vision Transformers",
                        "abstract": "Transformers have become one of the dominant architectures in deep learning, particularly as a powerful alternative to convolutional neural networks (CNNs) in computer vision. However, Transformer training and inference in previous works can be prohibitively expensive due to the quadratic complexity of self-attention over a long sequence of representations, especially for high-resolution dense prediction tasks. To this end, we present a novel Less attention vIsion Transformer (LIT), building upon the fact that the early self-attention layers in Transformers still focus on local patterns and bring minor benefits in recent hierarchical vision Transformers. Specifically, we propose a hierarchical Transformer where we use pure multi-layer perceptrons (MLPs) to encode rich local patterns in the early stages while applying self-attention modules to capture longer dependencies in deeper layers. Moreover, we further propose a learned deformable token merging module to adaptively fuse informative patches in a non-uniform manner. The proposed LIT achieves promising performance on image recognition tasks, including image classification, object detection and instance segmentation, serving as a strong backbone for many vision tasks. Code is available at: https://github.com/zhuang-group/LIT"
                    },
                    {
                        "title": "Scalable Vision Transformers with Hierarchical Pooling",
                        "abstract": "The recently proposed Visual image Transformers (ViT) with pure attention have achieved promising performance on image recognition tasks, such as image classification. However, the routine of the current ViT model is to maintain a full-length patch sequence during inference, which is redundant and lacks hierarchical representation. To this end, we propose a Hierarchical Visual Transformer (HVT) which progressively pools visual tokens to shrink the sequence length and hence reduces the computational cost, analogous to the feature maps downsampling in Convolutional Neural Networks (CNNs). It brings a great benefit that we can increase the model capacity by scaling dimensions of depth/width/resolution/patch size without introducing extra computational complexity due to the reduced sequence length. Moreover, we empirically find that the average pooled visual tokens contain more discriminative information than the single class token. To demonstrate the improved scalability of our HVT, we conduct extensive experiments on the image classification task. With comparable FLOPs, our HVT outperforms the competitive baselines on ImageNet and CIFAR-100 datasets. Code is available at https://github.com/MonashAI/HVT"
                    },
                    {
                        "title": "AQD: Towards Accurate Fully-Quantized Object Detection",
                        "abstract": "Network quantization allows inference to be conducted using low-precision arithmetic for improved inference efficiency of deep neural networks on edge devices. However, designing aggressively low-bit (e.g., 2-bit) quantization schemes on complex tasks, such as object detection, still remains challenging in terms of severe performance degradation and unverifiable efficiency on common hardware. In this paper, we propose an Accurate Quantized object Detection solution, termed AQD, to fully get rid of floating-point computation. To this end, we target using fixed-point operations in all kinds of layers, including the convolutional layers, normalization layers, and skip connections, allowing the inference to be executed using integer-only arithmetic. To demonstrate the improved latency-vs-accuracy trade-off, we apply the proposed methods on RetinaNet and FCOS. In particular, experimental results on MS-COCO dataset show that our AQD achieves comparable or even better performance compared with the full-precision counterpart under extremely low-bit schemes, which is of great practical value. Source code and models are available at: https://github.com/ziplab/QTool"
                    },
                    {
                        "title": "Auxiliary Learning for Deep Multi-task Learning",
                        "abstract": "Multi-task learning (MTL) is an efficient solution to solve multiple tasks simultaneously in order to get better speed and performance than handling each single-task in turn. The most current methods can be categorized as either: (i) hard parameter sharing where a subset of the parameters is shared among tasks while other parameters are task-specific; or (ii) soft parameter sharing where all parameters are task-specific but they are jointly regularized. Both methods suffer from limitations: the shared hidden layers of the former are difficult to optimize due to the competing objectives while the complexity of the latter grows linearly with the increasing number of tasks. To mitigate those drawbacks, this paper proposes an alternative, where we explicitly construct an auxiliary module to mimic the soft parameter sharing for assisting the optimization of the hard parameter sharing layers in the training phase. In particular, the auxiliary module takes the outputs of the shared hidden layers as inputs and is supervised by the auxiliary task loss. During training, the auxiliary module is jointly optimized with the MTL network, serving as a regularization by introducing an inductive bias to the shared layers. In the testing phase, only the original MTL network is kept. Thus our method avoids the limitation of both categories. We evaluate the proposed auxiliary module on pixel-wise prediction tasks, including semantic segmentation, depth estimation, and surface normal prediction with different network structures. The extensive experiments over various settings verify the effectiveness of our methods."
                    },
                    {
                        "title": "An Efficient Spatio-Temporal Pyramid Transformer for Action Detection",
                        "abstract": "The task of action detection aims at deducing both the action category and localization of the start and end moment for each action instance in a long, untrimmed video. While vision Transformers have driven the recent advances in video understanding, it is non-trivial to design an efficient architecture for action detection due to the prohibitively expensive self-attentions over a long sequence of video clips. To this end, we present an efficient hierarchical Spatio-Temporal Pyramid Transformer (STPT) for action detection, building upon the fact that the early self-attention layers in Transformers still focus on local patterns. Specifically, we propose to use local window attention to encode rich local spatio-temporal representations in the early stages while applying global attention modules to capture long-term space-time dependencies in the later stages. In this way, our STPT can encode both locality and dependency with largely reduced redundancy, delivering a promising trade-off between accuracy and efficiency. For example, with only RGB input, the proposed STPT achieves 53.6% mAP on THUMOS14, surpassing I3D+AFSD RGB model by over 10% and performing favorably against state-of-the-art AFSD that uses additional flow features with 31% fewer GFLOPs, which serves as an effective and efficient end-to-end Transformer-based framework for action detection."
                    }
                ]
            }
        }
    },
    "2302.06607": {
        "paper_data": {
            "title": "Generative Adversarial Equilibrium Solvers",
            "url": "http://arxiv.org/abs/2302.06607v2",
            "arxiv_id": "2302.06607",
            "authors": [
                "Denizalp Goktas",
                "David C. Parkes",
                "Ian Gemp",
                "Luke Marris",
                "Georgios Piliouras",
                "Romuald Elie",
                "Guy Lever",
                "Andrea Tacchetti"
            ],
            "abstract": "We introduce the use of generative adversarial learning to compute equilibria in general game-theoretic settings, specifically the generalized Nash equilibrium (GNE) in pseudo-games, and its specific instantiation as the competitive equilibrium (CE) in Arrow-Debreu competitive economies. Pseudo-games are a generalization of games in which players' actions affect not only the payoffs of other players but also their feasible action spaces. Although the computation of GNE and CE is intractable in the worst-case, i.e., PPAD-hard, in practice, many applications only require solutions with high accuracy in expectation over a distribution of problem instances. We introduce Generative Adversarial Equilibrium Solvers (GAES): a family of generative adversarial neural networks that can learn GNE and CE from only a sample of problem instances. We provide computational and sample complexity bounds, and apply the framework to finding Nash equilibria in normal-form games, CE in Arrow-Debreu competitive economies, and GNE in an environmental economic model of the Kyoto mechanism.",
            "introduction": " Introduction Economic models and equilibrium concepts are critical tools to solve practical problems, including capacity allocation in wireless and network communication (Han et al., 2011; Pang et al., 2008), energy resource allocation (Hobbs & Pang, 2007; Jing-Yuan & Smeers, 1999), and cloud computing (Gutman & Nisan, 2012; Lai et al., 2005; Zahedi et al., 2018; Ardagna et al., 2017). Many of these economic models are instances of what are known as pseudo-games , in which the actions taken by each player affect not only the other players\u2019 payoffs, as in games, but also the other players\u2019 \u0003Research conducted while the author was an intern at DeepMind. yAlso, School of Engineering and Applied Sciences, Harvard University.arXiv:2302.06607v2  [cs.GT]  20 Feb 2023Generative Adversarial Equilibrium Solvers strategy sets.3The formalism of pseudo-games was introduced by Arrow & Debreu (1954), who used it in studying their foundational microeconomic equilibrium model, the competitive economy model. The standard solution concept for pseudo-games is the generalized Nash equilibrium (GNE) (Arrow & Debreu, 1954; Facchinei & Kanzow, 2010a), which is an action pro\ufb01le from which no player can improve their payoff by unilaterally deviating to another action in the space of admissible actions determined by the actions of other players. Important economic models can often be formulated as a pseudo-game, with their set of solutions equal to the set of GNE of the pseudo-game: for instance, the set of competitive equilibria (CE) (Walras, 1896; Arrow & Debreu, 1954) of an Arrow-Debreu competitive economy corresponds to the set of GNE of an associated pseudo-game. A large literature has been devoted to the computation of GNE in certain classes of pseudo-games but unfortunately many algorithms that are guaranteed to converge in theory have in practice been observed to converge slowly in ill-conditioned or large problems or fail numerically (Facchinei & Kanzow, 2010b; Jordan et al., 2022; Goktas & Greenwald, 2022). Additionally, all known algorithms have hyperparameters that have to be optimized individually for every pseudo-game instance (Facchinei & Kanzow, 2010a), deteriorating the performance of these algorithms when used to solve multiple pseudo-games. These issues point to a need to develop methods can fail, not only theoretically but also in practice. 23Generative Adversarial Equilibrium Solvers E Additional Preliminary De\ufb01nitions Notation. We use caligraphic uppercase letters to denote sets (e.g., X); bold lowercase letters to denote vectors (e.g., p;\u0019); bold uppercase letters to denote matrices (e.g., X,\u0000), lowercase letters to denote scalar quantities (e.g., x;\u000e). We denote the ith row vector of a matrix (e.g., X) by the corresponding bold lowercase letter with subscript i(e.g., xi). Similarly, we denote the jth entry of a vector (e.g., porxi) by the corresponding Roman lowercase letter with subscriptj(e.g.,pjorxij). Models. An\"-variational equilibrium (VE) (ornormalized GNE ) of a pseudo-game is a strategy pro\ufb01le a\u00032X s.t. for alli2[n]anda2X ,ui(a\u0003)\u0015ui(ai;a\u0003 \u0000i)\u0000\". We note that in the above de\ufb01nitions, one could just as well writea\u00032X(a\u0003)asa\u00032X, as any \ufb01xed point of the joint action correspondence is also a jointly feasible action pro\ufb01le, and vice versa. A VE is an \u000f-VE with\"= 0. Under our assumptions, while GNE are guaranteed to exist in all pseudo-games by Arrow & Debreu\u2019s lemma on Appendix E for more details. 38Generative Adversarial Equilibrium Solvers spaces, greatly simplifying the architecture of our discriminator. We aim to \ufb01rst replicate the Related Work. Exchange economies",
            "references": [
                {
                    "title": "Exploitability Minimization in Games and Beyond",
                    "abstract": "Pseudo-games are a natural and well-known generalization of normal-form games, in which the actions taken by each player affect not only the other players' payoffs, as in games, but also the other players' strategy sets. The solution concept par excellence for pseudo-games is the generalized Nash equilibrium (GNE), i.e., a strategy profile at which each player's strategy is feasible and no player can improve their payoffs by unilaterally deviating to another strategy in the strategy set determined by the other players' strategies. The computation of GNE in pseudo-games has long been a problem of interest, due to applications in a wide variety of fields, from environmental protection to logistics to telecommunications. Although computing GNE is PPAD-hard in general, it is still of interest to try to compute them in restricted classes of pseudo-games. One approach is to search for a strategy profile that minimizes exploitability, i.e., the sum of the regrets across all players. As exploitability is nondifferentiable in general, developing efficient first-order methods that minimize it might not seem possible at first glance. We observe, however, that the exploitability-minimization problem can be recast as a min-max optimization problem, and thereby obtain polynomial-time first-order methods to compute a refinement of GNE, namely the variational equilibria (VE), in convex-concave cumulative regret pseudo-games with jointly convex constraints. More generally, we also show that our methods find the stationary points of the exploitability in polynomial time in Lipschitz-smooth pseudo-games with jointly convex constraints. Finally, we demonstrate in experiments that our methods not only outperform known algorithms, but that even in pseudo-games where they are not guaranteed to converge to a GNE, they may do so nonetheless, with proper initialization."
                },
                {
                    "title": "Turbocharging Solution Concepts: Solving NEs, CEs and CCEs with Neural Equilibrium Solvers",
                    "abstract": "Solution concepts such as Nash Equilibria, Correlated Equilibria, and Coarse Correlated Equilibria are useful components for many multiagent machine learning algorithms. Unfortunately, solving a normal-form game could take prohibitive or non-deterministic time to converge, and could fail. We introduce the Neural Equilibrium Solver which utilizes a special equivariant neural network architecture to approximately solve the space of all games of fixed shape, buying speed and determinism. We define a flexible equilibrium selection framework, that is capable of uniquely selecting an equilibrium that minimizes relative entropy, or maximizes welfare. The network is trained without needing to generate any supervised training data. We show remarkable zero-shot generalization to larger games. We argue that such a network is a powerful component for many possible multiagent algorithms."
                },
                {
                    "title": "First-Order Algorithms for Nonlinear Generalized Nash Equilibrium Problems",
                    "abstract": "We consider the problem of computing an equilibrium in a class of \\textit{nonlinear generalized Nash equilibrium problems (NGNEPs)} in which the strategy sets for each player are defined by equality and inequality constraints that may depend on the choices of rival players. While the asymptotic global convergence and local convergence rates of algorithms to solve this problem have been extensively investigated, the analysis of nonasymptotic iteration complexity is still in its infancy. This paper presents two first-order algorithms -- based on the quadratic penalty method (QPM) and augmented Lagrangian method (ALM), respectively -- with an accelerated mirror-prox algorithm as the solver in each inner loop. We establish a global convergence guarantee for solving monotone and strongly monotone NGNEPs and provide nonasymptotic complexity bounds expressed in terms of the number of gradient evaluations. Experimental results demonstrate the efficiency of our algorithms in practice."
                },
                {
                    "title": "Differentiable Economics for Randomized Affine Maximizer Auctions",
                    "abstract": "A recent approach to automated mechanism design, differentiable economics, represents auctions by rich function approximators and optimizes their performance by gradient descent. The ideal auction architecture for differentiable economics would be perfectly strategyproof, support multiple bidders and items, and be rich enough to represent the optimal (i.e. revenue-maximizing) mechanism. So far, such an architecture does not exist. There are single-bidder approaches (MenuNet, RochetNet) which are always strategyproof and can represent optimal mechanisms. RegretNet is multi-bidder and can approximate any mechanism, but is only approximately strategyproof. We present an architecture that supports multiple bidders and is perfectly strategyproof, but cannot necessarily represent the optimal mechanism. This architecture is the classic affine maximizer auction (AMA), modified to offer lotteries. By using the gradient-based optimization tools of differentiable economics, we can now train lottery AMAs, competing with or outperforming prior approaches in revenue."
                },
                {
                    "title": "Convex-Concave Min-Max Stackelberg Games",
                    "abstract": "Min-max optimization problems (i.e., min-max games) have been attracting a great deal of attention because of their applicability to a wide range of machine learning problems. Although significant progress has been made recently, the literature to date has focused on games with independent strategy sets; little is known about solving games with dependent strategy sets, which can be characterized as min-max Stackelberg games. We introduce two first-order methods that solve a large class of convex-concave min-max Stackelberg games, and show that our methods converge in polynomial time. Min-max Stackelberg games were first studied by Wald, under the posthumous name of Wald's maximin model, a variant of which is the main paradigm used in robust optimization, which means that our methods can likewise solve many convex robust optimization problems. We observe that the computation of competitive equilibria in Fisher markets also comprises a min-max Stackelberg game. Further, we demonstrate the efficacy and efficiency of our algorithms in practice by computing competitive equilibria in Fisher markets with varying utility structures. Our experiments suggest potential ways to extend our theoretical results, by demonstrating how different smoothness properties can affect the convergence rate of our algorithms."
                },
                {
                    "title": "Deep Learning for Two-Sided Matching",
                    "abstract": "We initiate the study of deep learning for the automated design of two-sided matching mechanisms. What is of most interest is to use machine learning to understand the possibility of new tradeoffs between strategy-proofness and stability. These properties cannot be achieved simultaneously, but the efficient frontier is not understood. We introduce novel differentiable surrogates for quantifying ordinal strategy-proofness and stability and use them to train differentiable matching mechanisms that map discrete preferences to valid randomized matchings. We demonstrate that the efficient frontier characterized by these learned mechanisms is substantially better than that achievable through a convex combination of baselines of deferred acceptance (stable and strategy-proof for only one side of the market), top trading cycles (strategy-proof for one side, but not stable), and randomized serial dictatorship (strategy-proof for both sides, but not stable). This gives a new target for economic theory and opens up new possibilities for machine learning pipelines in matching market design."
                },
                {
                    "title": "Learning Revenue-Maximizing Auctions With Differentiable Matching",
                    "abstract": "We propose a new architecture to approximately learn incentive compatible, revenue-maximizing auctions from sampled valuations. Our architecture uses the Sinkhorn algorithm to perform a differentiable bipartite matching which allows the network to learn strategyproof revenue-maximizing mechanisms in settings not learnable by the previous RegretNet architecture. In particular, our architecture is able to learn mechanisms in settings without free disposal where each bidder must be allocated exactly some number of items. In experiments, we show our approach successfully recovers multiple known optimal mechanisms and high-revenue, low-regret mechanisms in larger settings where the optimal mechanism is unknown."
                },
                {
                    "title": "Learning Competitive Equilibria in Exchange Economies with Bandit Feedback",
                    "abstract": "The sharing of scarce resources among multiple rational agents is one of the classical problems in economics. In exchange economies, which are used to model such situations, agents begin with an initial endowment of resources and exchange them in a way that is mutually beneficial until they reach a competitive equilibrium (CE). The allocations at a CE are Pareto efficient and fair. Consequently, they are used widely in designing mechanisms for fair division. However, computing CEs requires the knowledge of agent preferences which are unknown in several applications of interest. In this work, we explore a new online learning mechanism, which, on each round, allocates resources to the agents and collects stochastic feedback on their experience in using that allocation. Its goal is to learn the agent utilities via this feedback and imitate the allocations at a CE in the long run. We quantify CE behavior via two losses and propose a randomized algorithm which achieves sublinear loss under a parametric class of utilities. Empirically, we demonstrate the effectiveness of this mechanism through numerical simulations."
                },
                {
                    "title": "Efficient and Modular Implicit Differentiation",
                    "abstract": "Automatic differentiation (autodiff) has revolutionized machine learning. It allows to express complex computations by composing elementary ones in creative ways and removes the burden of computing their derivatives by hand. More recently, differentiation of optimization problem solutions has attracted widespread attention with applications such as optimization layers, and in bi-level problems such as hyper-parameter optimization and meta-learning. However, so far, implicit differentiation remained difficult to use for practitioners, as it often required case-by-case tedious mathematical derivations and implementations. In this paper, we propose automatic implicit differentiation, an efficient and modular approach for implicit differentiation of optimization problems. In our approach, the user defines directly in Python a function $F$ capturing the optimality conditions of the problem to be differentiated. Once this is done, we leverage autodiff of $F$ and the implicit function theorem to automatically differentiate the optimization problem. Our approach thus combines the benefits of implicit differentiation and autodiff. It is efficient as it can be added on top of any state-of-the-art solver and modular as the optimality condition specification is decoupled from the implicit differentiation mechanism. We show that seemingly simple principles allow to recover many existing implicit differentiation methods and create new ones easily. We demonstrate the ease of formulating and solving bi-level optimization problems using our framework. We also showcase an application to the sensitivity analysis of molecular dynamics."
                },
                {
                    "title": "Deep Reinforcement Learning in a Monetary Model",
                    "abstract": "We propose using deep reinforcement learning to solve dynamic stochastic general equilibrium models. Agents are represented by deep arti\ufb01cial neural networks and learn to solve their dynamic optimisation problem by interacting with the model environment, of which they have no a priori knowledge. Deep reinforcement learning o\ufb00ers a \ufb02exible yet principled way to model bounded rationality within this general class of models. We apply our proposed approach to a classical model from the adaptive learning literature in macroeconomics which looks at the interaction of monetary and \ufb01scal policy. We \ufb01nd that, contrary to adaptive learning, the arti\ufb01cially intelligent household can solve the model in all policy regimes."
                },
                {
                    "title": "Solving Heterogeneous General Equilibrium Economic Models with Deep Reinforcement Learning",
                    "abstract": "General equilibrium macroeconomic models are a core tool used by policymakers to understand a nation's economy. They represent the economy as a collection of forward-looking actors whose behaviours combine, possibly with stochastic effects, to determine global variables (such as prices) in a dynamic equilibrium. However, standard semi-analytical techniques for solving these models make it difficult to include the important effects of heterogeneous economic actors. The COVID-19 pandemic has further highlighted the importance of heterogeneity, for example in age and sector of employment, in macroeconomic outcomes and the need for models that can more easily incorporate it. We use techniques from reinforcement learning to solve such models incorporating heterogeneous agents in a way that is simple, extensible, and computationally efficient. We demonstrate the method's accuracy and stability on a toy problem for which there is a known analytical solution, its versatility by solving a general equilibrium problem that includes global stochasticity, and its flexibility by solving a combined macroeconomic and epidemiological model to explore the economic and health implications of a pandemic. The latter successfully captures plausible economic behaviours induced by differential health risks by age."
                },
                {
                    "title": "No-regret learning and mixed Nash equilibria: They do not mix",
                    "abstract": "Understanding the behavior of no-regret dynamics in general $N$-player games is a fundamental question in online learning and game theory. A folk result in the field states that, in finite games, the empirical frequency of play under no-regret learning converges to the game's set of coarse correlated equilibria. By contrast, our understanding of how the day-to-day behavior of the dynamics correlates to the game's Nash equilibria is much more limited, and only partial results are known for certain classes of games (such as zero-sum or congestion games). In this paper, we study the dynamics of \"follow-the-regularized-leader\" (FTRL), arguably the most well-studied class of no-regret dynamics, and we establish a sweeping negative result showing that the notion of mixed Nash equilibrium is antithetical to no-regret learning. Specifically, we show that any Nash equilibrium which is not strict (in that every player has a unique best response) cannot be stable and attracting under the dynamics of FTRL. This result has significant implications for predicting the outcome of a learning process as it shows unequivocally that only strict (and hence, pure) Nash equilibria can emerge as stable limit points thereof."
                },
                {
                    "title": "Auction learning as a two-player game",
                    "abstract": "Designing an incentive compatible auction that maximizes expected revenue is a central problem in Auction Design. While theoretical approaches to the problem have hit some limits, a recent research direction initiated by Duetting et al. (2019) consists in building neural network architectures to find optimal auctions. We propose two conceptual deviations from their approach which result in enhanced performance. First, we use recent results in theoretical auction design (Rubinstein and Weinberg, 2018) to introduce a time-independent Lagrangian. This not only circumvents the need for an expensive hyper-parameter search (as in prior work), but also provides a principled metric to compare the performance of two auctions (absent from prior work). Second, the optimization procedure in previous work uses an inner maximization loop to compute optimal misreports. We amortize this process through the introduction of an additional neural network. We demonstrate the effectiveness of our approach by learning competitive or strictly improved auctions compared to prior work. Both results together further imply a novel formulation of Auction Design as a two-player game with stationary utility functions."
                },
                {
                    "title": "Proportional Dynamics in Exchange Economies",
                    "abstract": "We study the proportional dynamics in exchange economies, where each player starts with some amount of money and a good. Every day, players bring one unit of their good and submit bids on goods they like, each good gets allocated in proportion to the bid amounts, and each seller collects the bids received. Then every player updates their bids proportionally to the contribution of each good in their utility. This dynamic models a process of learning how to bid and has been studied in a series of papers on Fisher and production markets, but not in exchange economies. Our main results are as follows: 1). For all linear utilities, the dynamic converges to market equilibrium utilities and allocations, while the bids and prices may cycle. We give a combinatorial characterization of limit cycles for prices and bids. 2). We introduce a lazy version of the dynamic, where players may save money for later, and show this converges in everything: utilities, allocations, and prices. This answers an open question about exchange markets with linear utilities, where tatonnement does not converge to market equilibria, and no natural process leading to equilibria was known for all additive utilities. We also note this dynamics represents a process where the players exchange goods throughout time (in out-of-equilibrium states), while tatonnement only explains how exchange happens in the limit."
                },
                {
                    "title": "A Neural Architecture for Designing Truthful and Efficient Auctions",
                    "abstract": "Auctions are protocols to allocate goods to buyers who have preferences over them, and collect payments in return. Economists have invested significant effort in designing auction rules that result in allocations of the goods that are desirable for the group as a whole. However, for settings where participants' valuations of the items on sale are their private information, the rules of the auction must deter buyers from misreporting their preferences, so as to maximize their own utility, since misreported preferences hinder the ability for the auctioneer to allocate goods to those who want them most. Manual auction design has yielded excellent mechanisms for specific settings, but requires significant effort when tackling new domains. We propose a deep learning based approach to automatically design auctions in a wide variety of domains, shifting the design work from human to machine. We assume that participants' valuations for the items for sale are independently sampled from an unknown but fixed distribution. Our system receives a data-set consisting of such valuation samples, and outputs an auction rule encoding the desired incentive structure. We focus on producing truthful and efficient auctions that minimize the economic burden on participants. We evaluate the auctions designed by our framework on well-studied domains, such as multi-unit and combinatorial auctions, showing that they outperform known auction designs in terms of the economic burden placed on participants."
                },
                {
                    "title": "Scheduling Jobs across Geo-Distributed Datacenters with Max-Min Fairness",
                    "abstract": "It has become routine for large volumes of data to be generated, stored, and processed across geographically distributed datacenters. To run a single data analytic job on such geo-distributed data, recent research proposed to distribute its tasks across datacenters, considering both data locality and network bandwidth across datacenters. Yet, it remains an open problem in the more general case, where multiple analytic jobs need to fairly share the resources at these geo-distributed datacenters. In this paper, we focus on the problem of assigning tasks belonging to multiple jobs across datacenters, with the specific objective of achieving max-min fairness across jobs sharing these datacenters, in terms of their job completion times. We formulate this problem as a lexicographical minimization problem, which is challenging to solve in practice due to its inherent multi-objective and discrete nature. To address these challenges, we iteratively solve its single-objective subproblems, which can be transformed to equivalent linear programming (LP) problems to be efficiently solved, thanks to their favorable properties. As a highlight of this paper, we have designed and implemented our proposed solution as a fair job scheduler based on Apache Spark, a modern data processing framework. With extensive evaluations of our real-world implementation on Amazon EC2 and large-scale simulations, we have shown convincing evidence that max-min fairness has been achieved and the worst job completion time has been significantly improved using our new job scheduler."
                },
                {
                    "title": "Optimality Conditions and Constraint Qualifications for Generalized Nash Equilibrium Problems and Their Practical Implications",
                    "abstract": "Generalized Nash equilibrium problems (GNEPs) are a generalization of the classic Nash equilibrium problems (NEPs), where each player's strategy set depends on the choices of the other players. In ..."
                },
                {
                    "title": "Exact Penalization of Generalized Nash Equilibrium Problems",
                    "abstract": "This paper presents an exact penalization theory of the generalized Nash equilibrium problem (GNEP) that has its origin from the renowned Arrow\u2013Debreu general economic equilibrium model. Whereas the latter model is the foundation of much of mathematical economics, the GNEP provides a mathematical model of multiagent noncooperative competition that has found many contemporary applications in diverse engineering domains. The most salient feature of the GNEP that distinguishes it from a standard noncooperative (Nash) game is that each player\u2019s optimization problem contains constraints that couple all players\u2019 decision variables. Extending results for stand-alone optimization problems, the penalization theory aims to convert the GNEP into a game of the standard kind without the coupled constraints, which is known to be more readily amenable to solution methods and analysis. Starting with an illustrative example to motivate the development, this paper focuses on two kinds of coupled constraints, shared (i.e., common) and finitely representable. Constraint residual functions and the associated error bound theory play an important role throughout the development."
                },
                {
                    "title": "Pacing Equilibrium in First-Price Auction Markets",
                    "abstract": "In the isolated auction of a single item, second price is often preferable to first price in properties of theoretical interest. Unfortunately, single items are rarely sold in true isolation, so considering the broader context is critical when adopting a pricing strategy. In this paper, we show that this context is important in a model centrally relevant to Internet advertising: when items (ad impressions) are individually auctioned within the context of a larger system that is managing budgets, theory offers surprising support for using a first price auction to sell each individual item. In particular, first price auctions offer theoretical guarantees of equilibrium uniqueness, monotonicity, and other desirable properties, as well as efficient computability as the solution to the well-studied Eisenberg-Gale convex program. We also use simulations to demonstrate that while there are incentives to misreport in thin markets (where budgets aren't constraining), a bidder's incentive to deviate vanishes in thick markets."
                },
                {
                    "title": "Amdahl's Law in the Datacenter Era: A Market for Fair Processor Allocation",
                    "abstract": "We present a processor allocation framework that uses Amdahl's Law to model parallel performance and a market mechanism to allocate cores. First, we propose the Amdahl utility function and demonstrate its accuracy when modeling performance from processor core allocations. Second, we design a market based on Amdahl utility that optimizes users' bids for processors based on workload parallelizability. The framework uses entitlements to guarantee fairness yet outperforms existing proportional share algorithms."
                },
                {
                    "title": "A Unified Game-Theoretic Approach to Multiagent Reinforcement Learning",
                    "abstract": "To achieve general intelligence, agents must learn how to interact with others in a shared environment: this is the challenge of multiagent reinforcement learning (MARL). The simplest form is independent reinforcement learning (InRL), where each agent treats its experience as part of its (non-stationary) environment. In this paper, we first observe that policies learned using InRL can overfit to the other agents' policies during training, failing to sufficiently generalize during execution. We introduce a new metric, joint-policy correlation, to quantify this effect. We describe an algorithm for general MARL, based on approximate best responses to mixtures of policies generated using deep reinforcement learning, and empirical game-theoretic analysis to compute meta-strategies for policy selection. The algorithm generalizes previous ones such as InRL, iterated best response, double oracle, and fictitious play. Then, we present a scalable implementation which reduces the memory requirement using decoupled meta-solvers. Finally, we demonstrate the generality of the resulting policies in two partially observable settings: gridworld coordination games and poker."
                },
                {
                    "title": "Settling the complexity of Leontief and PLC exchange markets under exact and approximate equilibria",
                    "abstract": "Our first result shows membership in PPAD for the problem of computing approximate equilibria for an Arrow-Debreu exchange market for piecewise-linear concave (PLC) utility functions. As a corollary we also obtain membership in PPAD for Leontief utility functions. This settles an open question of Vazirani and Yannakakis (2011). Next we show FIXP-hardness of computing equilibria in Arrow-Debreu exchange markets under Leontief utility functions, and Arrow-Debreu markets under linear utility functions and Leontief production sets, thereby settling these open questions of Vazirani and Yannakakis (2011). As corollaries, we obtain FIXP-hardness for PLC utilities and for Arrow-Debreu markets under linear utility functions and polyhedral production sets. In all cases, as required under FIXP, the set of instances mapped onto will admit equilibria, i.e., will be \"yes\" instances. If all instances are under consideration, then in all cases we prove that the problem of deciding if a given instance admits an equilibrium is ETR-complete, where ETR is the class Existential Theory of Reals. As a consequence of the results stated above, and the fact that membership in FIXP has been established for PLC utilities, the entire computational difficulty of Arrow-Debreu markets under PLC utility functions lies in the Leontief utility subcase. This is perhaps the most unexpected aspect of our result, since Leontief utilities are meant for the case that goods are perfect complements, whereas PLC utilities are very general, capturing not only the cases when goods are complements and substitutes, but also arbitrary combinations of these and much more. Finally, we give a polynomial time algorithm for finding an equilibrium in Arrow-Debreu exchange markets under Leontief utility functions provided the number of agents is a constant. This settles part of an open problem of Devanur and Kannan (2008)."
                },
                {
                    "title": "Optimal auctions through deep learning",
                    "abstract": "Designing an incentive compatible auction that maximizes expected revenue is an intricate task. The single-item case was resolved in a seminal piece of work by Myerson in 1981. Even after 30--40 years of intense research, the problem remains unsolved for settings with two or more items. We overview recent research results that show how tools from deep learning are shaping up to become a powerful tool for the automated design of near-optimal auctions auctions. In this approach, an auction is modeled as a multilayer neural network, with optimal auction design framed as a constrained learning problem that can be addressed with standard machine learning pipelines. Through this approach, it is possible to recover to a high degree of accuracy essentially all known analytically derived solutions for multi-item settings and obtain novel mechanisms for settings in which the optimal mechanism is unknown."
                },
                {
                    "title": "A Globally Convergent LP-Newton Method",
                    "abstract": "We develop a globally convergent algorithm based on the LP-Newton method, which has been recently proposed for solving constrained equations, possibly nonsmooth and possibly with nonisolated solutions. The new algorithm makes use of linesearch for the natural merit function and preserves the strong local convergence properties of the original LP-Newton scheme. We also present computational experiments on a set of generalized Nash equilibrium problems, and a comparison of our approach with the previous hybrid globalization employing the potential reduction method."
                },
                {
                    "title": "Borg, Omega, and Kubernetes",
                    "abstract": "Lessons learned from three container-management systems over a decade."
                },
                {
                    "title": "A Complementary Pivot Algorithm for Market Equilibrium under Separable, Piecewise-Linear Concave Utilities",
                    "abstract": "Using Lemke's scheme, we give a complementary pivot algorithm for computing an equilibrium for Arrow--Debreu markets under separable, piecewise-linear concave (SPLC) utilities. Despite the polynomial parity argument on directed graphs (PPAD) completeness of this case, experiments indicate that our algorithm is practical---on randomly generated instances, the number of iterations it needs is linear in the total number of segments (i.e., pieces) in all the utility functions specified in the input. Our paper settles a number of open problems: (1) Eaves (1976) gave an LCP formulation and a Lemke-type algorithm for the linear Arrow--Debreu model. We generalize both to the SPLC case, hence settling the relevant part of his open problem. (2) Our path following algorithm for SPLC markets, together with a result of Todd (1976), gives a direct proof of membership of such markets in PPAD and settles a question of Vazirani and Yannakakis (2011). (3) We settle a question of Devanur and Kannan (2008) of obtaining a \u201csy..."
                },
                {
                    "title": "Trading Networks with Bilateral Contracts",
                    "abstract": "We consider general networks of bilateral contracts that include supply chains. We define a new stability concept, called trail stability, and show that any network of bilateral contracts has a trail-stable outcome whenever agents' preferences satisfy full substitutability. Trail stability is a natural extension of chain stability, but is a stronger solution concept in general contract networks. Trail-stable outcomes are not immune to deviations of arbitrary sets of firms. In fact, we show that outcomes satisfying an even more demanding stability property -- full trail stability -- always exist. We pin down conditions under which trail-stable and fully trail-stable outcomes have a lattice structure. We then completely describe the relationships between all stability concepts. When contracts specify trades and prices, we also show that competitive equilibrium exists in networked markets even in the absence of fully transferrable utility. The competitive equilibrium outcome is trail-stable."
                },
                {
                    "title": "Tatonnement for Linear and Gross Substitutes Markets",
                    "abstract": "A central question in economics and computer science is when and how markets can arrive at an equilibrium. Many existing algorithms for computing equilibria in classes of Arrow-Debreu markets rely on full knowledge of the utility functions of all agents, while in reality we often face unknown markets without this information. A classic approach to unknown markets from economics is tatonnement -- dynamic price update processes that only require oracle access to query the aggregate demand of each good. In this paper, we design a new class of tatonnement algorithms. For the first time, we show how tatonnement can converge in polynomial time to market equilibria in linear markets and spending constraint markets, where the main obstacle is a non-continuous demand function. This also gives the first polynomial-time algorithm for spending constraint markets and settles an open question raised in [Duan and Mehlhorn, ICALP'13]. Moreover, our algorithms can be applied to unknown markets with weak gross substitutes (WGS) property, in which they converge to $(1+\\epsilon)$-approximate market equilibria in time polynomial in market parameters and $\\log(1/\\epsilon)$. This exponentially improves the previous convergence rate of polynomial in market parameters and $1/\\epsilon$. Our approach uses ideas developed for full information linear markets and applies them in unknown linear and non-linear markets. Since our oracles return demands based on real utility functions, precision and representation of prices and demands become a major technical challenge. Here we develop new tools and insights, which may be of independent interest."
                },
                {
                    "title": "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift",
                    "abstract": "Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.82% top-5 test error, exceeding the accuracy of human raters."
                },
                {
                    "title": "Apache Hadoop YARN: yet another resource negotiator",
                    "abstract": "The initial design of Apache Hadoop [1] was tightly focused on running massive, MapReduce jobs to process a web crawl. For increasingly diverse companies, Hadoop has become the data and computational agor\u00e1---the de facto place where data and computational resources are shared and accessed. This broad adoption and ubiquitous usage has stretched the initial design well beyond its intended target, exposing two key shortcomings: 1) tight coupling of a specific programming model with the resource management infrastructure, forcing developers to abuse the MapReduce programming model, and 2) centralized handling of jobs' control flow, which resulted in endless scalability concerns for the scheduler. In this paper, we summarize the design, development, and current state of deployment of the next generation of Hadoop's compute platform: YARN. The new architecture we introduced decouples the programming model from the resource management infrastructure, and delegates many scheduling functions (e.g., task fault-tolerance) to per-application components. We provide experimental evidence demonstrating the improvements we made, confirm improved efficiency by reporting the experience of running YARN on production environments (including 100% of Yahoo! grids), and confirm the flexibility claims by discussing the porting of several programming frameworks onto YARN viz. Dryad, Giraph, Hoya, Hadoop MapReduce, REEF, Spark, Storm, Tez."
                },
                {
                    "title": "A Rational Convex Program for Linear Arrow-Debreu Markets",
                    "abstract": "We present a new flow-type convex program describing equilibrium solutions to linear Arrow-Debreu markets. Whereas convex formulations were previously known ([Nenakov and Primak 1983; Jain 2007; Cornet 1989]), our program exhibits several new features. It provides a simple necessary and sufficient condition and a concise proof of the existence and rationality of equilibria, settling an open question raised by Vazirani [2012]. As a consequence, we also obtain a simple new proof of the result in Mertens [2003] that the equilibrium prices form a convex polyhedral set."
                },
                {
                    "title": "Tatonnement beyond gross substitutes?: gradient descent to the rescue",
                    "abstract": "Tatonnement is a simple and natural rule for updating prices in Exchange (Arrow-Debreu) markets. In this paper we define a class of markets for which tatonnement is equivalent to gradient descent. This is the class of markets for which there is a convex potential function whose gradient is always equal to the negative of the excess demand and we call it Convex Potential Function (CPF) markets. We show the following results. CPF markets contain the class of Eisenberg Gale (EG) markets, defined previously by Jain and Vazirani. The subclass of CPF markets for which the demand is a differentiable function contains exactly those markets whose demand function has a symmetric negative semi-definite Jacobian. We define a family of continuous versions of tatonnement based on gradient descent using a Bregman divergence. As we show, all processes in this family converge to an equilibrium for any CPF market. This is analogous to the classic result for markets satisfying the Weak Gross Substitutes property. A discrete version of tatonnement converges toward the equilibrium for the following markets of complementary goods; its convergence rate for these settings is analyzed using a common potential function. Fisher markets in which all buyers have Leontief utilities. The tatonnement process reduces the distance to the equilibrium, as measured by the potential function, to an \u03b5 fraction of its initial value in O(1/\u03b5) rounds of price updates. Fisher markets in which all buyers have complementary CES utilities. Here, the distance to the equilibrium is reduced to an \u03b5 fraction of its initial value in O(log(1/\u03b5)) rounds of price updates.\n This shows that tatonnement converges for the entire range of Fisher markets when buyers have complementary CES utilities, in contrast to prior work, which could analyze only the substitutes range, together with a small portion of the complementary range."
                },
                {
                    "title": "Choosy: max-min fair sharing for datacenter jobs with constraints",
                    "abstract": "Max-Min Fairness is a flexible resource allocation mechanism used in most datacenter schedulers. However, an increasing number of jobs have hard placement constraints, restricting the machines they can run on due to special hardware or software requirements. It is unclear how to define, and achieve, max-min fairness in the presence of such constraints. We propose Constrained Max-Min Fairness (CMMF), an extension to max-min fairness that supports placement constraints, and show that it is the only policy satisfying an important property that incentivizes users to pool resources. Optimally computing CMMF is challenging, but we show that a remarkably simple online scheduler, called Choosy, approximates the optimal scheduler well. Through experiments, analysis, and simulations, we show that Choosy on average differs 2% from the optimal CMMF allocation, and lets jobs achieve their fair share quickly."
                },
                {
                    "title": "Beyond dominant resource fairness: extensions, limitations, and indivisibilities",
                    "abstract": "We study the problem of allocating multiple resources to agents with heterogeneous demands. Technological advances such as cloud computing and data centers provide a new impetus for investigating this problem under the assumption that agents demand the resources in fixed proportions, known in economics as Leontief preferences. In a recent paper, Ghodsi et al. [2011] introduced the dominant resource fairness (DRF) mechanism, which was shown to possess highly desirable theoretical properties under Leontief preferences. We extend their results in three directions. First, we show that DRF generalizes to more expressive settings, and leverage a new technical framework to formally extend its guarantees. Second, we study the relation between social welfare and properties such as truthfulness; DRF performs poorly in terms of social welfare, but we show that this is an unavoidable shortcoming that is shared by every mechanism that satisfies one of three basic properties. Third, and most importantly, we study a realistic setting that involves indivisibilities. We chart the boundaries of the possible in this setting, contributing a new relaxed notion of fairness and providing both possibility and impossibility results."
                },
                {
                    "title": "Fair allocation without trade",
                    "abstract": "We consider the age-old problem of allocating items among different agents in a way that is efficient and fair. Two papers, by Dolev et al. and Ghodsi et al., have recently studied this problem in the context of computer systems. Both papers had similar models for agent preferences, but advocated different notions of fairness. We formalize both fairness notions in economic terms, extending them to apply to a larger family of utilities. Noting that in settings with such utilities efficiency is easily achieved in multiple ways, we study notions of fairness as criteria for choosing between different efficient allocations. Our technical results are algorithms for finding fair allocations corresponding to two fairness notions: Regarding the notion suggested by Ghodsi et al., we present a polynomial-time algorithm that computes an allocation for a general class of fairness notions, in which their notion is included. For the other, suggested by Dolev et al., we show that a competitive market equilibrium achieves the desired notion of fairness, thereby obtaining a polynomial-time algorithm that computes such a fair allocation and solving the main open problem raised by Dolev et al."
                },
                {
                    "title": "Flexible Distributed Capacity Allocation and Load Redirect Algorithms for Cloud Systems",
                    "abstract": "In Cloud computing systems, resource management is one of the main issues. Indeed, in any time instant resources have to be allocated to handle effectively workload fluctuations, while providing Quality of Service (QoS) guarantees to the end users. In such systems, workload prediction-based autonomic computing techniques have been developed. In this paper we propose capacity allocation techniques able to coordinate multiple distributed resource controllers working in geographically distributed cloud sites. Furthermore, capacity allocation solutions are integrated with a load redirection mechanism which forwards incoming requests between different domains. The overall goal is to minimize the costs of the allocated virtual machine instances, while guaranteeing QoS constraints expressed as a threshold on the average response time. We compare multiple heuristics which integrate workload prediction and distributed non-linear optimization techniques. Experimental results show how our solutions significantly improve other heuristics proposed in the literature (5-35% on average), without introducing significant QoS violations."
                },
                {
                    "title": "Dominant Resource Fairness: Fair Allocation of Multiple Resource Types",
                    "abstract": "We consider the problem of fair resource allocation in a system containing different resource types, where each user may have different demands for each resource. To address this problem, we propose Dominant Resource Fairness (DRF), a generalization of max-min fairness to multiple resource types. We show that DRF, unlike other possible policies, satisfies several highly desirable properties. First, DRF incentivizes users to share resources, by ensuring that no user is better off if resources are equally partitioned among them. Second, DRF is strategy-proof, as a user cannot increase her allocation by lying about her requirements. Third, DRF is envy-free, as no user would want to trade her allocation with that of another user. Finally, DRF allocations are Pareto efficient, as it is not possible to improve the allocation of a user without decreasing the allocation of another user. We have implemented DRF in the Mesos cluster resource manager, and show that it leads to better throughput and fairness than the slot-based fair sharing schemes in current cluster schedulers."
                },
                {
                    "title": "Mesos: A Platform for Fine-Grained Resource Sharing in the Data Center",
                    "abstract": "We present Mesos, a platform for sharing commodity clusters between multiple diverse cluster computing frameworks, such as Hadoop and MPI. Sharing improves cluster utilization and avoids per-framework data replication. Mesos shares resources in a fine-grained manner, allowing frameworks to achieve data locality by taking turns reading data stored on each machine. To support the sophisticated schedulers of today's frameworks, Mesos introduces a distributed two-level scheduling mechanism called resource offers. Mesos decides how many resources to offer each framework, while frameworks decide which resources to accept and which computations to run on them. Our results show that Mesos can achieve near-optimal data locality when sharing the cluster among diverse frameworks, can scale to 50,000 (emulated) nodes, and is resilient to failures."
                },
                {
                    "title": "A game theoretic formulation of the service provisioning problem in cloud systems",
                    "abstract": "Cloud computing is an emerging paradigm which allows the on-demand delivering of software, hardware, and data as services. As cloud-based services are more numerous and dynamic, the development of efficient service provisioning policies become increasingly challenging. Game theoretic approaches have shown to gain a thorough analytical understanding of the service provisioning problem.\n In this paper we take the perspective of Software as a Service (SaaS) providers which host their applications at an Infrastructure as a Service (IaaS) provider. Each SaaS needs to comply with quality of service requirements, specified in Service Level Agreement (SLA) contracts with the end-users, which determine the revenues and penalties on the basis of the achieved performance level. SaaS providers want to maximize their revenues from SLAs, while minimizing the cost of use of resources supplied by the IaaS provider. Moreover, SaaS providers compete and bid for the use of infrastructural resources. On the other hand, the IaaS wants to maximize the revenues obtained providing virtualized resources. In this paper we model the service provisioning problem as a Generalized Nash game, and we propose an efficient algorithm for the run time management and allocation of IaaS resources to competing SaaSs."
                },
                {
                    "title": "Game theory.",
                    "abstract": "Game theory is a toolkit for examining situations where decision makers influence each other. I discuss the nature of game-theoretic analysis, the history of game theory, why game theory is useful for understanding human psychology, and why game theory has played a key role in the recent explosion of interest in the field of behavioral economics. WIREs Cogni Sci 2011 2 167-173 DOI: 10.1002/wcs.119 For further resources related to this article, please visit the WIREs website."
                },
                {
                    "title": "Computable General Equilibrium (CGE) Modelling for Regional Economic Development Analysis",
                    "abstract": "Partridge M. D. and Rickman D. S. Computable general equilibrium (CGE) modelling for regional economic development analysis, Regional Studies. Despite their long-standing use in economic policy analysis generally, and increasing popularity in regional policy analysis, computable general equilibrium (CGE) models have yet to become the dominant approach for analysis of regional economic development policies. This review discusses the likely reasons for the limited use of CGE models for regional economic development analysis, particularly for small regions. It also proposes methodological improvements that would likely lead to the wider use of CGE models in economic development practice. A central theme is the need for a model that can be empirically demonstrated to capture the relevant policy components and the structure of the region of study. Partridge M. D. et Rickman D. S. La mod\u00e9lisation CGE dans l'analyse du d\u00e9veloppement \u00e9conomique r\u00e9gional, Regional Studies. En d\u00e9pit de leur emploi g\u00e9n\u00e9ralis\u00e9 de longue date dans l'analyse de la politique \u00e9conomique, et malgr\u00e9 leur popularit\u00e9 croissante dans l'analyse de la politique r\u00e9gionale, la mod\u00e9lisation CGE doit encore devenir la fa\u00e7on dominante dans les politiques de d\u00e9veloppement \u00e9conomique r\u00e9gional. Cet article cherche \u00e0 pr\u00e9senter les raisons qui risquent d'expliquer l'emploi limit\u00e9 des mod\u00e8les CGE dans l'analyse du d\u00e9veloppement \u00e9conomique r\u00e9gional, en particulier pour ce qui est des petites r\u00e9gions. L'article cherche aussi \u00e0 proposer des am\u00e9liorations m\u00e9thodologiques qui entra\u00eeneraient l'emploi plus g\u00e9n\u00e9ralis\u00e9 des mod\u00e8les CGE dans le d\u00e9veloppement \u00e9conomique en pratique. Un aspect central est le besoin d'un mod\u00e8le susceptible de capter de mani\u00e8re empirique les \u00e9l\u00e9ments cl\u00e9s d'une politique et la structure de la r\u00e9gion \u00e0 \u00e9tudier. R\u00e9gional\u2003Equilibre g\u00e9n\u00e9ral sur ordinateur (CGE)\u2003D\u00e9veloppement \u00e9conomique r\u00e9gional Partridge M. D. und Rickman D. S. CGE-Modelle f\u00fcr Analysen der regionalen Wirtschaftsentwicklung, Regional Studies. CGE-Modelle kommen in Analysen der Wirtschaftspolitik im Allgemeinen seit langem zum Einsatz und sind zunehmend auch bei Analysen der Regionalpolitik beliebt. Doch bei der Analyse regionalwirtschaftlicher Entwicklungspolitiken sind sie noch nicht zum vorherrschenden Ansatz geworden. In diesem Beitrag werden die m\u00f6glichen Gr\u00fcnde er\u00f6rtert, warum CGE-Modelle f\u00fcr Analysen der regionalen Wirtschaftsentwicklung insbesondere im Fall von kleinen Regionen nur begrenzt Anwendung finden. Au\u00dferdem werden methodologische Verbesserungen vorgeschlagen, die einen breiteren Einsatz von CGE-Modellen in der Praxis der Wirtschaftsentwicklung erm\u00f6glichen k\u00f6nnten. Ein zentrales Thema ist die Notwendigkeit eines Modells, von dem sich empirisch nachweisen l\u00e4sst, dass es die relevanten politischen Komponenten und die Struktur der untersuchten Region erfasst. Regionale\u2003Berechenbares allgemeines Gleichgewicht\u2003Regionale Wirtschaftsentwicklung Partridge M. D. y Rickman D. S. Modelo CGE para el an\u00e1lisis del desarrollo econ\u00f3mico regional, Regional Studies. Pese a la larga utilizaci\u00f3n general en los an\u00e1lisis pol\u00edticos sobre econom\u00eda y la creciente popularidad en los an\u00e1lisis pol\u00edticos regionales, los modelos CGE todav\u00eda no han llegado a ser enfoques dominantes para el an\u00e1lisis de las pol\u00edticas de desarrollo econ\u00f3mico a nivel regional. En este ensayo exponemos las posibles razones para el uso limitado de modelos CGE para los an\u00e1lisis de desarrollo econ\u00f3mico regional, especialmente para peque\u00f1as regiones. Tambi\u00e9n proponemos mejoras metodol\u00f3gicas que probablemente llevar\u00edan a un uso m\u00e1s amplio de los modelos CGE en la pr\u00e1ctica del desarrollo econ\u00f3mico. Un tema central es la necesidad de un modelo que demuestre emp\u00edricamente que capta los componentes relevantes de la pol\u00edtica y la estructura de la regi\u00f3n que se estudia. Regional\u2003Equilibrio computable general\u2003Desarrollo econ\u00f3mico regional"
                },
                {
                    "title": "The Combinatorial Assignment Problem: Approximate Competitive Equilibrium from Equal Incomes",
                    "abstract": "This paper proposes a new mechanism for combinatorial assignment\u2014for example, assigning schedules of courses to students\u2014based on an approximation to competitive equilibrium from equal incomes (CEEI) in which incomes are unequal but arbitrarily close together. The main technical result is an existence theorem for approximate CEEI. The mechanism is approximately efficient, satisfies two new criteria of outcome fairness, and is strategyproof in large markets. Its performance is explored on real data, and it is compared to alternatives from theory and practice: all other known mechanisms are either unfair ex post or manipulable even in large markets, and most are both manipulable and unfair."
                },
                {
                    "title": "Penalty Methods for the Solution of Generalized Nash Equilibrium Problems",
                    "abstract": "The generalized Nash equilibrium problem (GNEP) is an extension of the classical Nash equilibrium problem where both the objective functions and the constraints of each player may depend on the rivals' strategies. This class of problems has a multitude of important engineering applications, and yet solution algorithms are extremely scarce. In this paper, we analyze in detail a globally convergent penalty method that has favorable theoretical properties. We also consider strengthened results for a particular subclass of problems very often considered in the literature. Basically our method reduces the GNEP to a single penalized (and nonsmooth) Nash equilibrium problem. We suggest a suitable method for the solution of the latter penalized problem and present extensive numerical results."
                },
                {
                    "title": "Convex Optimization",
                    "abstract": "This textbook is based on lectures given by the authors at MIPT (Moscow), HSE (Moscow), FEFU (Vladivostok), V.I. Vernadsky KFU (Simferopol), ASU (Republic of Adygea), and the University of Grenoble-Alpes (Grenoble, France). First of all, the authors focused on the program of a two-semester course of lectures on convex optimization, which is given to students of MIPT. The first chapter of this book contains the materials of the first semester (\"Fundamentals of convex analysis and optimization\"), the second and third chapters contain the materials of the second semester (\"Numerical methods of convex optimization\"). The textbook has a number of features. First, in contrast to the classic manuals, this book does not provide proofs of all the theorems mentioned. This allowed, on one side, to describe more themes, but on the other side, made the presentation less self-sufficient. The second important point is that part of the material is advanced and is published in the Russian educational literature, apparently for the first time. Third, the accents that are given do not always coincide with the generally accepted accents in the textbooks that are now popular. First of all, we talk about a sufficiently advanced presentation of conic optimization, including robust optimization, as a vivid demonstration of the capabilities of modern convex analysis."
                },
                {
                    "title": "Nash Equilibria of Static Prediction Games",
                    "abstract": "The standard assumption of identically distributed training and test data is violated when an adversary can exercise some control over the generation of the test data. In a prediction game, a learner produces a predictive model while an adversary may alter the distribution of input data. We study single-shot prediction games in which the cost functions of learner and adversary are not necessarily antagonistic. We identify conditions under which the prediction game has a unique Nash equilibrium, and derive algorithms that will find the equilibrial prediction models. In a case study, we explore properties of Nash-equilibrial prediction models for email spam filtering empirically."
                },
                {
                    "title": "Quincy: fair scheduling for distributed computing clusters",
                    "abstract": "This paper addresses the problem of scheduling concurrent jobs on clusters where application data is stored on the computing nodes. This setting, in which scheduling computations close to their data is crucial for performance, is increasingly common and arises in systems such as MapReduce, Hadoop, and Dryad as well as many grid-computing environments. We argue that data-intensive computation benefits from a fine-grain resource sharing model that differs from the coarser semi-static resource allocations implemented by most existing cluster computing architectures. The problem of scheduling with locality and fairness constraints has not previously been extensively studied under this resource-sharing model.\n We introduce a powerful and flexible new framework for scheduling concurrent distributed jobs with fine-grain resource sharing. The scheduling problem is mapped to a graph datastructure, where edge weights and capacities encode the competing demands of data locality, fairness, and starvation-freedom, and a standard solver computes the optimal online schedule according to a global cost model. We evaluate our implementation of this framework, which we call Quincy, on a cluster of a few hundred computers using a varied workload of data-and CPU-intensive jobs. We evaluate Quincy against an existing queue-based algorithm and implement several policies for each scheduler, with and without fairness constraints. Quincy gets better fairness when fairness is requested, while substantially improving data locality. The volume of data transferred across the cluster is reduced by up to a factor of 3.9 in our experiments, leading to a throughput increase of up to 40%."
                },
                {
                    "title": "Python: the tutorial",
                    "abstract": "Over the years, programming languages have grown more powerful, but correspondingly more complex; and while that complexity is fine and appropriate for professional programmers, it hinders and discourages beginning Computer Science students."
                },
                {
                    "title": "Settling the Complexity of Arrow-Debreu Equilibria in Markets with Additively Separable Utilities",
                    "abstract": "We prove that the problem of computing an Arrow-Debreu market equilibrium is PPAD-complete even when all traders use additively separable, piecewise-linear and concave utility functions. In fact, our proof shows that this market-equilibrium problem does not have a fully polynomial-time approximation scheme, unless every problem in PPAD is solvable in polynomial time."
                },
                {
                    "title": "Matplotlib: A 2D Graphics Environment",
                    "abstract": "Matplotlib is a 2D graphics package used for Python for application development, interactive scripting,and publication-quality image generation across user interfaces and operating systems"
                },
                {
                    "title": "Distributed Power Allocation With Rate Constraints in Gaussian Parallel Interference Channels",
                    "abstract": "This paper considers the minimization of transmit power in Gaussian parallel interference channels, subject to a rate constraint for each user. To derive decentralized solutions that do not require any cooperation among the users, we formulate this power control problem as a (generalized) Nash equilibrium (NE) game. We obtain sufficient conditions that guarantee the existence and nonemptiness of the solution set to our problem. Then, to compute the solutions of the game, we propose two distributed algorithms based on the single user water-filling solution: The sequential and the simultaneous iterative water-filling algorithms, wherein the users update their own strategies sequentially and simultaneously, respectively. We derive a unified set of sufficient conditions that guarantee the uniqueness of the solution and global convergence of both algorithms. Our results are applicable to all practical distributed multipoint-to-multipoint interference systems, either wired or wireless, where a quality of service in (QoS) terms of information rate must be guaranteed for each link."
                },
                {
                    "title": "Settling the Complexity of Two-Player Nash Equilibrium",
                    "abstract": "Even though many people thought the problem of finding Nash equilibria is hard in general, and it has been proven so for games among three or more players recently, it's not clear whether the two-player case can be shown in the same class of PPAD-complete problems. We prove that the problem of finding a Nash equilibrium in a two-player game is PPAD-complete"
                },
                {
                    "title": "The complexity of computing a Nash equilibrium",
                    "abstract": "We resolve the question of the complexity of Nash equilibrium by showing that the problem of computing a Nash equilibrium in a game with 4 or more players is complete for the complexity class PPAD. Our proof uses ideas from the recently-established equivalence between polynomial time solvability of normal form games and graphical games, establishing that these kinds of games can simulate a PPAD-complete class of Brouwer functions."
                },
                {
                    "title": "Leontief economies encode nonzero sum two-player games",
                    "abstract": "We consider Leontief exchange economies, i.e., economies where the consumers desire goods in fixed proportions. Unlike bimatrix games, such economies are not guaranteed to have equilibria in general. On the other hand, they include suitable restricted versions which always have equilibria.We give a reduction from two-player games to a special family of Leontief exchange economies, which are guaranteed to have equilibria, with the property that the Nash equilibria of any game are in one-to-one correspondence with the equilibria of the corresponding economy.Our reduction exposes a potential hurdle inherent in solving certain families of market equilibrium problems: finding an equilibrium for Leontief economies (where an equilibrium is guaranteed to exist) is at least as hard as finding a Nash equilibrium for two-player nonzero sum games.As a corollary of the one-to-one correspondence, we obtain a number of hardness results for questions related to the computation of market equilibria, using results already established for games [17]. In particular, among other results, we show that it is NP-hard to say whether a particular family of Leontief exchange economies, that is guaranteed to have at least one equilibrium, has more than one equilibrium.Perhaps more importantly, we also prove that it is NP-hard to decide whether a Leontief exchange economy has an equilibrium. This fact should be contrasted against the known PPAD-completeness result of [30], which holds when the problem satisfies some standard sufficient conditions that make it equivalent to the computational version of Brouwer's Fixed Point Theorem.On the algorithmic side, we present an algorithm for finding an approximate equilibrium for some special Leontief economies, which achieves quasi-polynomial time whenever each trader does not demand too much more of any good than some other good."
                },
                {
                    "title": "Tycoon: An implementation of a distributed, market-based resource allocation system",
                    "abstract": "Distributed clusters like the Grid and PlanetLab enable the same statistical multiplexing efficiency gains for computing as the Internet provides for networking. One major challenge is allocating resources in an economically efficient and low-latency way. A common solution is proportional share, where users each get resources in proportion to their pre-defined weight. However, this does not allow users to differentiate the value of their jobs. This leads to economic inefficiency. In contrast, systems that require reservations impose a high latency (typically minutes to hours) to acquire resources. \n \nThis paper describes Tycoon, a market based distributed resource allocation system based on proportional share. The key advantages of Tycoon are that it allows users to differentiate the values of their jobs, its resource acquisition latency is limited only by communication delays, and it imposes no manual bidding overhead on users. Experimental results using a prototype implementation of the design are included."
                },
                {
                    "title": "Run the GAMUT: a comprehensive approach to evaluating game-theoretic algorithms",
                    "abstract": "We present GAMUT^1, a suite of game generators designed for testing game-theoretic algorithms. We explain why such a generator is necessary, offer a way of visualizing relationships between the sets of games supported by GAMUT, and give an overview of GAMUT\u00fds architecture. We highlight the importance of using comprehensive test data by benchmarking existing algorithms. We show surprisingly large variation in algorithm performance across different sets of games for two widely-studied problems: computing Nash equilibria and multiagent learning in repeated games."
                },
                {
                    "title": "On relaxation algorithms in computation of noncooperative equilibria",
                    "abstract": "This paper considers a special class of numerical algorithms, the so-called relaxation algorithm, for Nash equilibrium points in noncooperative games. The relaxation algorithms have been studied by various authors for the deterministic case. Convergence conditions of this algorithm are based on fixed point theorems. For example, Basar (1987) and Li and Basar (1987) have proved its convergence for a two-player game via the contraction mapping theorem. For the quadratic case these conditions can be easily checked. For other nonlinear payoff functions it is sometimes difficult to check these convergence conditions. In this paper, the authors propose an alternative approach using the residual terms of the Nikaido-Isoda function. The convergence theorem is proved for nonsmooth weakly convex-concave Nikaido-Isoda functions. The family of weakly convex-concave functions is broad enough for applications, since if includes the family of smooth functions. When the payoff functions are twice continuously differentiable, the condition for the residual terms is reduced to strict positiveness of a matrix representing the difference of the Hessians of the Nikaido-Isoda function with respect to the first and second groups of variables. An analogous condition was used by Uryas'ev (1988) to prove convergence of the gradient-type algorithm for the Nash equilibrium problem. In this paper the authors discuss only the deterministic case; nevertheless this approach can be generalized for the stochastic Nash equilibrium problems with uncertainties in parameters. >"
                },
                {
                    "title": "Noncooperative Convex Games: Computing Equilibrium by Partial Regularization",
                    "abstract": "A class of non-cooperative constrained games is analysed for which the Ky Fan fuction is convex-concave. Nash equilibria of such games correspond to diagonal saddle points of the said function. This feature is exploited in designing computational algorithms for finding such equilibria."
                },
                {
                    "title": "Some Examples of Global Instability of the Competitive Equilibrium",
                    "abstract": "In this paper we consider the problem of stability of the competitive equilibrium. The market demand functions are sums of individual demand functions obtained directly by utility maximization. The rate of change of the price of each commodity is assumed to be proportional to the excess market demand for that commodity. A number of examples are given for which the motion of the prices is globally unstable in the sense that starting from any set of prices other than equilibrium, the prices oscillate without tending towards equilibrium."
                },
                {
                    "title": "EXISTENCE OF AN EQUILIBRIUM FOR A COMPETITIVE ECONOMY",
                    "abstract": "A. Wald has presented a model of production and a model of exchange and proofs of the existence of an equilibrium for each of them. Here proofs of the existence of an equilibrium are given for an integrated model of production, exchange and consumption. In addition the assumptions made on the technologies of producers and the tastes of consumers are significantly weaker than Wald's. Finally a simplification of the structure of the proofs has been made possible through use of the concept of an abstract economy, a generalization of that of a game. Introduction L. Walras [ 24 ] first formulated the state of the economic system at any point of time as the solution of a system of simultaneous equations representing the demand for goods by consumers, the supply of goods by producers, and the equilibrium condition that supply equal demand on every market. It was assumed that each consumer acts so as to maximize his utility, each producer acts so as to maximize his profit, and perfect competition prevails, in the sense that each producer and consumer regards the prices paid and received as independent of his own choices. Walras did not, however, give any conclusive arguments to show that the equations, as given, have a solution."
                },
                {
                    "title": "Towards the PAC Learnability of Nash Equilibrium",
                    "abstract": "Nash equilibrium (NE) is one of the most important solution concepts in game theory and has broad applications in machine learning research. While tremendous empirical success has been re-ported on learning approximate NE, whether NE is Probably Approximately Correct (PAC) learnable has not been addressed yet. In this paper, we make the \ufb01rst effort to study the PAC learnability of NE in normal-form games. We prove that NE is agnostic PAC learnable if the hypothesis class for Nash predictors has bounded covering numbers. Theoretically, we derive such a result by constructing a self-supervised PAC learning framework based on a commonly-used objective for Nash approximation. Empirically, we provide numerical evidence showing that a Nash predictor trained under our PAC learning framework, even with the most straightforward neural architecture, can ef\ufb01-ciently learn the NE for games sampled from the same unknown distribution. Our results justify the feasibility of approximating NE through purely data-driven approaches, which bene\ufb01ts both game theorists and machine learning practitioners."
                },
                {
                    "title": "Welfare Maximization in Competitive Equilibrium: Reinforcement Learning for Markov Exchange Economy",
                    "abstract": "We study a bilevel economic system, which we refer to as a Markov exchange economy (MEE), from the point of view of multi-agent reinforcement learning (MARL). An MEE involves a central planner and a group of self-interested agents. The goal of the agents is to form a Competitive Equilibrium (CE), where each agent myopically maximizes her own utility at each step. The goal of the central planner is to steer the system so as to maximize social welfare, which is defined as the sum of the utilities of all agents. Working in a setting in which the utility function and the system dynamics are both unknown, we propose to find the socially optimal policy and the CE from data via both online and offline variants of MARL. Concretely, we first devise a novel suboptimality metric specifically tailored to MEE, such that minimizing such a metric certifies globally optimal policies for both the planner and the agents. Second, in the online setting, we propose an al-gorithm, dubbed as MOLM , which combines the optimism principle for exploration with subgame CE seeking. Our algorithm can readily incorporate general function approximation tools for handling large state spaces and achieves a sublinear regret. Finally, we adapt the algorithm to an offline setting based on the pessimism principle and establish an upper bound on the suboptimality."
                },
                {
                    "title": "Finding General Equilibria in Many-Agent Economic Simulations Using Deep Reinforcement Learning",
                    "abstract": "Real economies can be seen as a sequential imperfect-information game with many heterogeneous, interacting strategic agents of various agent types, such as consumers, \ufb01rms, and governments. Dynamic general equilibrium models are common economic tools to model the economic activity, interactions, and outcomes in such systems. However, existing analytical and computational methods struggle to \ufb01nd explicit equilibria when all agents are strategic and interact, while joint learning is unstable and challenging. Amongst others, a key reason is that the actions of one economic agent may change the reward function of another agent, e.g., a consumer\u2019s expendable income changes when \ufb01rms change prices or governments change taxes. We show that multi-agent deep reinforcement learning (RL) can discover stable solutions that are (cid:15) -Nash equilibria for a meta-game over agent types, in economic simulations with many agents, through the use of structured learning curricula and ef\ufb01cient GPU-only simulation and training. Conceptually, our approach is more \ufb02exible and does not need unrealistic assumptions, e.g., market clearing, that are commonly used for analytical tractability. Our GPU implementation enables training and analyzing economies with a large number of agents within reasonable time frames, e.g., training completes within a day. We demonstrate our approach in real-business-cycle models, a representative family of DGE models, with 100 worker-consumers, 10 \ufb01rms, and a government who taxes and redistributes. We validate the learned meta-game (cid:15) -Nash equilibria through approximate best-response analyses, show that RL policies align with economic intuitions, and that our approach is constructive, e.g., by explicitly learning a spectrum"
                },
                {
                    "title": "PAC Learnability of Approximate Nash Equilibrium in Bimatrix Games",
                    "abstract": "Computing Nash equilibrium in bimatrix games is PPAD-hard, and many works have focused on the approximate solutions. When games are generated from a \ufb01xed unknown distribution, learning a Nash predictor via data-driven approaches can be preferable. In this paper, we study the learnability of approximate Nash equilibrium in bimatrix games. We prove that Lipschitz function class is agnostic Probably Approximately Correct (PAC) learnable with respect to Nash approximation loss. Additionally, to demonstrate the advantages of learning a Nash predictor, we develop a model that can e\ufb03ciently approximate solutions for games under the same distribution. We show by experiments that the solutions from our Nash predictor can serve as e\ufb00ective initializing points for other Nash solvers."
                },
                {
                    "title": "Seaborn: Statistical Data Visualization",
                    "abstract": "seaborn is a library for making statistical graphics in Python. It provides a high-level interface to matplotlib and integrates closely with pandas data structures. Functions in the seaborn library expose a declarative, dataset-oriented API that makes it easy to translate questions about data into graphics that can answer them. When given a dataset and a specification of the plot to make, seaborn automatically maps the data values to visual attributes such as color, size, or style, internally computes statistical transformations, and decorates the plot with informative axis labels and a legend. Many seaborn functions can generate figures with multiple panels that elicit comparisons between conditional subsets of data or across different pairings of variables in a dataset. seaborn is designed to be useful throughout the lifecycle of a scientific project. By producing complete graphics from a single function call with minimal arguments, seaborn facilitates rapid prototyping and exploratory data analysis. And by offering extensive options for customization, along with exposing the underlying matplotlib objects, it can be used to create polished, publication-quality figures."
                },
                {
                    "title": "Generalized Nash Equilibria for the Service Provisioning Problem in Multi-Cloud Systems",
                    "abstract": "The adoption of cloud technologies is steadily increasing. In such systems, applications can benefit from nearly infinite virtual resources on a pay-per-use basis. However, being the cloud massively multi-tenant and characterized by highly variable workloads the development of more and more effective provisioning policies assumes paramount importance. Boosted by the success of the cloud, the application of Game Theory models and methodologies has also become popular, since they have been demonstrated to suit perfectly to cloud social, economic, and strategic structures. This paper aims to study, model and efficiently solve the cost minimization problem associated with the service provisioning of SaaS virtual machines in multiple IaaSs. We propose a game-theoretic approach for the runtime management of resources from multiple IaaS providers to be allocated to multiple competing SaaSs, along with a cost model including revenues and penalties for requests execution failures. A distributed algorithm for identifying Generalized Nash Equilibria has been developed and analysed in detail. The effectiveness of our approach has been assessed by performing a wide set of analyses under multiple workload conditions. Results show that our algorithm is scalable and provides significant cost savings with respect to alternative methods (80 percent on average). Furthermore, increasing the number of IaaS providers SaaSs can achieve 9-15 percent cost savings from the workload distribution on multiple IaaSs."
                }
            ],
            "categories": [
                "cs.GT"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop efficient algorithms for computing generalized Nash equilibria in pseudo-games that overcome the limitations of existing methods?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of economic modeling and equilibrium concepts, which have significant implications in various practical applications such as wireless communication, energy resource allocation, and cloud computing. By improving the computation of generalized Nash equilibria, we can enhance the performance of algorithms used in these domains, leading to more efficient resource allocation and better decision-making processes. This research could pave the way for future studies that explore more complex economic models and their applications, ultimately contributing to the development of robust solutions in multi-agent systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the slow convergence and numerical instability of existing algorithms, particularly in ill-conditioned or large-scale pseudo-game scenarios. Naive approaches may fail due to their inability to handle the complexities of the interdependencies between players' actions and payoffs. Additionally, the need to optimize hyperparameters for each instance of a pseudo-game adds a layer of complexity that can significantly hinder performance. Overcoming these technical and practical obstacles requires innovative algorithmic strategies that can generalize across different pseudo-game instances.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on specific classes of pseudo-games, often overlooking the broader applicability of generalized Nash equilibria. The limitations of existing algorithms, such as slow convergence rates and the necessity for hyperparameter tuning, have created barriers to effectively solving this problem. Additionally, many approaches have not adequately addressed the computational challenges posed by large-scale or ill-conditioned problems. Our approach aims to leverage generative adversarial techniques to create a more robust and efficient framework for computing equilibria, differentiating it from prior work by focusing on generalizability and performance across diverse scenarios.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a generative adversarial framework to compute generalized Nash equilibria in pseudo-games. We will utilize a dataset of various pseudo-game instances to train our model, focusing on metrics such as convergence speed and accuracy of the computed equilibria. The expected outcomes include a significant improvement in the efficiency and reliability of equilibrium computation, enabling the application of our methods to a wider range of economic models and practical scenarios. This approach aims to provide a scalable"
            }
        },
        "author_data": {
            "3c4ee150-be7a-4c9d-9873-fc3d24787f72": {
                "pk": "3c4ee150-be7a-4c9d-9873-fc3d24787f72",
                "name": "Denizalp Goktas",
                "collaborators": [
                    "A. Greenwald",
                    "Amy Greenwald",
                    "Jiayi Zhao",
                    "Arjun Prakash",
                    "Josefine B. Graebener",
                    "Apurva Badithela",
                    "Wyatt Ubellacker",
                    "Eric V. Mazumdar",
                    "Aaron D. Ames",
                    "Richard M. Murray",
                    "Enrique Areyan Viqueira",
                    "John Randolph",
                    "Power",
                    "Ryan Edmonds",
                    "Gianluca Brero"
                ],
                "domain": [
                    "Algorithmic Economics",
                    "Game Theory",
                    "Optimization",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Flow-Based Synthesis of Reactive Tests for Discrete Decision-Making Systems with Temporal Logic Specifications",
                        "abstract": "Designing tests to evaluate if a given autonomous system satisfies complex specifications is challenging due to the complexity of these systems. This work proposes a flow-based approach for reactive test synthesis from temporal logic specifications, enabling the synthesis of test environments consisting of static and reactive obstacles and dynamic test agents. The temporal logic specifications describe desired test behavior, including system requirements as well as a test objective that is not revealed to the system. The synthesized test strategy places restrictions on system actions in reaction to the system state. The tests are minimally restrictive and accomplish the test objective while ensuring realizability of the system's objective without aiding it (semi-cooperative setting). Automata theory and flow networks are leveraged to formulate a mixed-integer linear program (MILP) to synthesize the test strategy. For a dynamic test agent, the agent strategy is synthesized for a GR(1) specification constructed from the solution of the MILP. If the specification is unrealizable by the dynamics of the test agent, a counterexample-guided approach is used to resolve the MILP until a strategy is found. This flow-based, reactive test synthesis is conducted offline and is agnostic to the system controller. Finally, the resulting test strategy is demonstrated in simulation and experimentally on a pair of quadrupedal robots for a variety of specifications."
                    },
                    {
                        "title": "Convex-Concave Zero-sum Markov Stackelberg Games",
                        "abstract": "Zero-sum Markov Stackelberg games can be used to model myriad problems, in domains ranging from economics to human robot interaction. In this paper, we develop policy gradient methods that solve these games in continuous state and action settings using noisy gradient estimates computed from observed trajectories of play. When the games are convex-concave, we prove that our algorithms converge to Stackelberg equilibrium in polynomial time. We also show that reach-avoid problems are naturally modeled as convex-concave zero-sum Markov Stackelberg games, and that Stackelberg equilibrium policies are more effective than their Nash counterparts in these problems."
                    },
                    {
                        "title": "Fisher Markets with Social Influence",
                        "abstract": "A Fisher market is an economic model of buyer and seller interactions in which each buyer\u2019s utility depends only on the bundle of goods she obtains. Many people\u2019s interests, however, are affected by their social interactions with others. In this paper, we introduce a generalization of Fisher markets, namely influence Fisher markets, which captures the impact of social influence on buyers\u2019 utilities. We show that competitive equilibria in influence Fisher markets correspond to generalized Nash equilibria in an associated pseudo-game, which implies the existence of competitive equilibria in all influence Fisher markets with continuous and concave utility functions. We then construct a monotone pseudo-game, whose variational equilibria and their duals together characterize competitive equilibria in influence Fisher markets with continuous, jointly concave, and homogeneous utility functions. This observation implies that competitive equilibria in these markets can be computed in polynomial time under standard smoothness assumptions on the utility functions. The dual of this second pseudo-game enables us to interpret the competitive equilibria of influence CCH Fisher markets as the solutions to a system of simultaneous Stackelberg games. Finally, we derive a novel first-order method that solves this Stackelberg system in polynomial time, prove that it is equivalent to computing competitive equilibrium prices via t\u00e2tonnement, and run experiments that confirm our theoretical results."
                    },
                    {
                        "title": "T\u00e2tonnement in Homothetic Fisher Markets",
                        "abstract": "A prevalent theme in the economics and computation literature is to identify natural price-adjustment processes by which sellers and buyers in a market can discover equilibrium prices. An example of such a process is t\u00e2tonnement, an auction-like algorithm first proposed in 1874 by French economist Walras in which sellers adjust prices based on the Marshallian demands of buyers, i.e., budget-constrained utility-maximizing demands. A dual concept in consumer theory is a buyer's Hicksian demand, i.e., consumptions that minimize expenditure while achieving a desired utility level. In this paper, we identify the maximum of the absolute value of the elasticity of the Hicksian demand, i.e., the maximum percentage change in the Hicksian demand of any good w.r.t. the change in the price of some other good, as an economic parameter sufficient to capture and explain a range of convergent and non-convergent t\u00e2tonnement behaviors in a broad class of markets. In particular, we prove the convergence of t\u00e2tonnement at a rate of O((1+\u03b52)/T), in homothetic Fisher markets with bounded price elasticity of Hicksian demand, i.e., Fisher markets in which consumers have preferences represented by homogeneous utility functions and the price elasticity of their Hicksian demand is bounded, where \u03b5 is the maximum absolute value of the price elasticity of Hicksian demand across all buyers. Our result not only generalizes known convergence results for CES Fisher markets, but extends them to mixed nested CES markets and Fisher markets with continuous, possibly non-concave, homogeneous utility functions. Our convergence rate covers the full spectrum of nested CES utilities, including Leontief and linear utilities, unifying previously existing disparate convergence and non-convergence results. In particular, for \u03b5 = 0, i.e., Leontief markets, we recover the best-known convergence rate of O(1/T), and as \u03b5 \u2192 \u221e, e.g., linear Fisher markets, we obtain non-convergent behavior, as expected."
                    },
                    {
                        "title": "Exploitability Minimization in Games and Beyond",
                        "abstract": "Pseudo-games are a natural and well-known generalization of normal-form games, in which the actions taken by each player affect not only the other players' payoffs, as in games, but also the other players' strategy sets. The solution concept par excellence for pseudo-games is the generalized Nash equilibrium (GNE), i.e., a strategy profile at which each player's strategy is feasible and no player can improve their payoffs by unilaterally deviating to another strategy in the strategy set determined by the other players' strategies. The computation of GNE in pseudo-games has long been a problem of interest, due to applications in a wide variety of fields, from environmental protection to logistics to telecommunications. Although computing GNE is PPAD-hard in general, it is still of interest to try to compute them in restricted classes of pseudo-games. One approach is to search for a strategy profile that minimizes exploitability, i.e., the sum of the regrets across all players. As exploitability is nondifferentiable in general, developing efficient first-order methods that minimize it might not seem possible at first glance. We observe, however, that the exploitability-minimization problem can be recast as a min-max optimization problem, and thereby obtain polynomial-time first-order methods to compute a refinement of GNE, namely the variational equilibria (VE), in convex-concave cumulative regret pseudo-games with jointly convex constraints. More generally, we also show that our methods find the stationary points of the exploitability in polynomial time in Lipschitz-smooth pseudo-games with jointly convex constraints. Finally, we demonstrate in experiments that our methods not only outperform known algorithms, but that even in pseudo-games where they are not guaranteed to converge to a GNE, they may do so nonetheless, with proper initialization."
                    },
                    {
                        "title": "Robust No-Regret Learning in Min-Max Stackelberg Games",
                        "abstract": "The behavior of no-regret learning algorithms is well understood in two-player min-max (i.e, zero-sum) games. In this paper, we investigate the behavior of no-regret learning in min-max games with dependent strategy sets , where the strategy of the first player con-strains the behavior of the second. Such games are best understood as sequential, i.e., min-max Stackelberg, games. We consider two settings, one in which only the first player chooses their actions using a no-regret algorithm while the second player best responds, and one in which both players use no-regret algorithms. For the former case, we show that no-regret dynamics converge to a Stackelberg equilibrium. For the latter case, we introduce a new type of regret, which we call Lagrangian regret, and show that if both players minimize their Lagrangian regrets, then play converges to a Stackelberg equilibrium. We then observe that online mirror descent (OMD) dynamics in these two settings correspond respectively to a known nested (i.e., sequential) gradient descent-ascent (GDA) algorithm and a new simultaneous GDA-like algorithm, thereby establishing convergence of these algorithms to Stackelberg equilibrium. Finally, we analyze the robustness of OMD dynamics to perturbations by in-vestigating online min-max Stackelberg games. We prove that OMD dynamics are robust for a large class of online min-max games with independent strategy sets. In the dependent case, we demonstrate the robustness of OMD dynamics experimentally by simulating them in online Fisher markets, a canonical example of a min-max Stackelberg game with dependent strategy sets. Ap-plied computing \u2192 Multi-agent"
                    },
                    {
                        "title": "An Algorithmic Theory of Markets and Their Application to Decentralized Markets",
                        "abstract": "Broadly speaking, I hope to dedicate my PhD to improving our understanding of algorithmic economics with the ultimate goal of building welfare improving decentralized technology for markets. In the following pages, I describe how my past work has built on the existing literature to get closer to the goal of creating such technologies, and describe what research paths this work opens up for the rest of my PhD. I believe that my research has the potential to provide algorithmic solutions to problems in machine learning, optimization, and game theory, and can be used to improve the efficiency of online marketplaces."
                    },
                    {
                        "title": "Gradient Descent Ascent in Min-Max Stackelberg Games",
                        "abstract": "Min-max optimization problems (i.e., min-max games) have at-tracted a great deal of attention recently as their applicability to a wide range of machine learning problems has become evident. In this paper, we study min-max games with dependent strategy sets, where the strategy of the first player constrains the behavior of the second. Such games are best understood as sequential, i.e., Stackelberg, games, for which the relevant solution concept is Stackelberg equilibrium, a generalization of Nash. One of the most popular algorithms for solving min-max games is gradient descent ascent (GDA). We present a straightforward generalization of GDA to min-max Stackelberg games with dependent strategy sets, but show that it may not converge to a Stackelberg equilibrium. We then introduce two variants of GDA, which assume access to a solution oracle for the optimal Karush Kuhn Tucker (KKT) multipliers of the games\u2019 constraints. We show that such an oracle exists for a large class of convex-concave min-max Stackelberg games, and provide proof that our GDA variants with such an oracle converge in \ud835\udc42 ( 1 / \ud835\udf00 2 ) iterations to an \ud835\udf00 -Stackelberg equilibrium, improving on the most efficient algorithms currently known which converge in \ud835\udc42 ( 1 / \ud835\udf00 3 ) iterations. We then show that solving Fisher markets, a canonical example of a min-max Stackelberg game, using our novel algorithm, corresponds to buyers and sellers using myopic best-response dynamics in a repeated market, allowing us to prove the convergence of these dynamics in \ud835\udc42 ( 1 / \ud835\udf00 2 ) iterations in Fisher markets. We close by describing experiments on Fisher markets which suggest potential ways to extend our theoretical results, by demonstrating how different properties of the objective function can affect the convergence and convergence rate of our algorithms."
                    },
                    {
                        "title": "Zero-Sum Stochastic Stackelberg Games",
                        "abstract": "Zero-sum stochastic games have found important applications in a variety of fields, from machine learning to economics. Work on this model has primarily focused on the computation of Nash equilibrium due to its effectiveness in solving adversarial board and video games. Unfortunately, a Nash equilibrium is not guaranteed to exist in zero-sum stochastic games when the payoffs at each state are not convex-concave in the players' actions. A Stackelberg equilibrium, however, is guaranteed to exist. Consequently, in this paper, we study zero-sum stochastic Stackelberg games. Going beyond known existence results for (non-stationary) Stackelberg equilibria, we prove the existence of recursive (i.e., Markov perfect) Stackelberg equilibria (recSE) in these games, provide necessary and sufficient conditions for a policy profile to be a recSE, and show that recSE can be computed in (weakly) polynomial time via value iteration. Finally, we show that zero-sum stochastic Stackelberg games can model the problem of pricing and allocating goods across agents and time. More specifically, we propose a zero-sum stochastic Stackelberg game whose recSE correspond to the recursive competitive equilibria of a large class of stochastic Fisher markets. We close with a series of experiments that showcase how our methodology can be used to solve the consumption-savings problem in stochastic Fisher markets."
                    },
                    {
                        "title": "Convex-Concave Min-Max Stackelberg Games",
                        "abstract": "Min-max optimization problems (i.e., min-max games) have been attracting a great deal of attention because of their applicability to a wide range of machine learning problems. Although significant progress has been made recently, the literature to date has focused on games with independent strategy sets; little is known about solving games with dependent strategy sets, which can be characterized as min-max Stackelberg games. We introduce two first-order methods that solve a large class of convex-concave min-max Stackelberg games, and show that our methods converge in polynomial time. Min-max Stackelberg games were first studied by Wald, under the posthumous name of Wald's maximin model, a variant of which is the main paradigm used in robust optimization, which means that our methods can likewise solve many convex robust optimization problems. We observe that the computation of competitive equilibria in Fisher markets also comprises a min-max Stackelberg game. Further, we demonstrate the efficacy and efficiency of our algorithms in practice by computing competitive equilibria in Fisher markets with varying utility structures. Our experiments suggest potential ways to extend our theoretical results, by demonstrating how different smoothness properties can affect the convergence rate of our algorithms."
                    },
                    {
                        "title": "T\u00e2tonnement Beyond Constant Elasticity of Substitution",
                        "abstract": "In this paper, we bring consumer theory to bear in the analysis of Fisher markets whose buyers have arbitrary continuous, concave, homogeneous (CCH) utility functions representing locally non-satiated preferences. The main tools we use are the dual concepts of expenditure minimization and indirect utility maximization. First, we use expenditure functions to construct a new convex program whose dual, like the dual of the Eisenberg-Gale program, characterizes the equilibrium prices of CCH Fisher markets. We then prove that the subdifferential of the dual of our convex program is equal to the negative excess demand in the associated market, which makes generalized gradient descent equivalent to computing equilibrium prices via t\u00e2tonnement. Finally, we use our novel characterization of equilibrium prices via expenditure functions to show that a discrete t\u00e2tonnement process converges at a rate of O(1/t) in Fisher markets with continuous, strictly concave, homogeneous (CSCH) utility functions\u2014a class of utility functions beyond the class of CES utility functions, the largest class for which convergence results were previously known. CSCH Fisher markets include nested and mixed CES Fisher markets, thus providing a meaningful expansion of the space of Fisher markets that is solvable via t\u00e2tonnement."
                    },
                    {
                        "title": "Banzhaf Power in Hierarchical Voting Games Extended",
                        "abstract": "Abstract"
                    },
                    {
                        "title": "Convex-Concave 0-Sum Markov Stackelberg Games",
                        "abstract": "Zero-sum Markov Stackelberg games can be used to model myriad problems, in domains ranging from economics to human robot interaction. We develop a policy gradient method which we prove solves these games in continuous state, continuous action settings, using noisy gradient estimates computed from observed trajectories of play. When the games are convex-concave, we prove that our algorithm converges to Stackelberg equilibrium in polynomial time. We also prove that reach-avoid problems are naturally modeled as convex-concave zero-sum Markov Stackelberg games, and show experimentally that Stackelberg equilibrium policies are more effective than their Nash counterparts in these problems"
                    },
                    {
                        "title": "Constructing Auctions from Fair-Division Mechanisms Ryan",
                        "abstract": "A fair division mechanism is a means of allocating items among agents at no cost, where the fairness of the ensuing allocation is evaluated with respect to the agents\u2019 preferences for the items. In contrast, an auction is a means of allocating items at some cost, where competitive equilibrium (CE) outcomes are desirable, i.e., no agent can improve their utility by deviating to another affordable bundle, and the market clears. It is well known that certain fair and efficient objectives, such as maximizing Nash social welfare, can be supported by CE prices. Our first result is to show that in certain fair division settings, only NSW-maximizing allocations are amenable to supporting CE prices. Other objectives of fair division mechanisms are of interest, however. For example, dominant resource fairness (DRF) is widely used to allocate cloud computing resources, sometimes at arbitrary prices. Our main contribution is to provide a simple means of constructing an auction from a fair division mechanism (i.e., of augmenting an allocation rule with a pricing rule), assuming budget-constrained agents with non-transferable utilities. The defining feature of our auction is that it prescribes prices intended to \u201cstabilize\u201d the allocation, so that the utility improvement any agent can achieve by deviating to another affordable bundle is minimized. We find these prices by casting the corresponding search problem as min-max optimization. When the agents\u2019 utilities are non-transferable and homogeneous, this problem is convex-concave, and can thus be solved in polynomial-time using gradient descent ascent. We provide necessary and sufficient conditions under which our pricing rule guarantees that a wide class of fair allocations, most notably those generated by DRF, are stable. Finally, we report on the performance of our method experimentally when these conditions are not upheld."
                    }
                ]
            },
            "7d5d95ca-4dc6-457f-808a-ccfe992ba072": {
                "pk": "7d5d95ca-4dc6-457f-808a-ccfe992ba072",
                "name": "David C. Parkes",
                "collaborators": [
                    "Gianluca Brero",
                    "M. Gerstgrasser",
                    "Alon Eden",
                    "D. Chakrabarti",
                    "Vincent Li",
                    "Matheus V. X. Ferreira",
                    "Daniel J. Moroz",
                    "Manuel W\u00fcthrich",
                    "Mark York",
                    "Z. Fan",
                    "F. J. Coss\u00edo",
                    "B. Altschuler",
                    "He Sun",
                    "Xintong Wang",
                    "N. Lepore",
                    "Eric Mibuari",
                    "Rakshit S. Trivedi",
                    "J. Watson-Daniels",
                    "Berk Ustun",
                    "Yaniv Yacoby",
                    "J. Girash",
                    "Paul Tylkin",
                    "Goran Radanovic",
                    "S. Ravindranath",
                    "Zhe Feng",
                    "Shira Li",
                    "Jonathan Ma",
                    "S. Kominers",
                    "Rediet Abebe",
                    "T-H. Hubert Chan",
                    "J. Kleinberg",
                    "Zhibin Liang",
                    "Mauro Sozio",
                    "Charalampos E. Tsourakakis",
                    "Mitchell Stern",
                    "M. Neuder",
                    "Rithvik Rao"
                ],
                "domain": [
                    "Mechanism Design",
                    "Reinforcement Learning",
                    "Blockchain",
                    "Crowdsourcing"
                ],
                "publications": [
                    {
                        "title": "Eliciting Information from participants with Competing Incentives and Dependent Beliefs",
                        "abstract": "In this paper, we study belief elicitation about an uncertain future event, where the reports will affect a principal's decision. We study two problems that can arise in this setting: (1) Agents may have an interest in the outcome of the principal's decision. We show that with intrinsic competing incentives (an interest in a decision that is internal to an agent) truthfulness cannot be guaranteed and there is a fundamental tradeoff between how much the principal allows reports to influence the decision, how much budget the principal has, and the degree to which a mechanism can be manipulated. Furthermore, we show that the Quadratic Scoring Rule is worst-case optimal in minimizing the degree of manipulation. In contrast, we obtain positive results and truthful mechanisms in a setting where the competing incentives stem instead from a rational briber who wants to promote a particular decision. We show that the budget required to achieve this robustness scales with the sum of squares of the degree to which agent reports can influence the decision. (2) We study the setting where the future event is only observed conditionally on the decision taken. We give a category of mechanisms that are truthful when agent beliefs are independent but fails with dependent beliefs, and show how to resolve this through a decoupling method."
                    },
                    {
                        "title": "Differential Liquidity Provision in Uniswap v3 and Implications for Contract Design\u2731",
                        "abstract": "Decentralized exchanges (DEXs) provide a means for users to trade pairs of assets on-chain without the need of a trusted third party to effectuate a trade. Amongst these, constant function market maker (CFMM) DEXs such as Uniswap handle the most volume of trades between ERC-20 tokens. With the introduction of Uniswap v3, liquidity providers are given the option to differentially allocate liquidity to be used for trades that occur within specific price intervals. In this paper, we formalize the profit and loss that liquidity providers can earn when providing specific liquidity positions to a contract. With this in hand, we are able to compute optimal liquidity allocations for liquidity providers who hold beliefs over how prices evolve over time. Ultimately, we use this tool to shed light on the design question regarding how v3 contracts should partition price space for permissible liquidity allocations. Our results show that a richer space of potential partitions can simultaneously benefit both liquidity providers and traders."
                    },
                    {
                        "title": "Learning to Mitigate AI Collusion on Economic Platforms",
                        "abstract": "Algorithmic pricing on online e-commerce platforms raises the concern of tacit collusion, where reinforcement learning algorithms learn to set collusive prices in a decentralized manner and through nothing more than profit feedback. This raises the question as to whether collusive pricing can be prevented through the design of suitable\"buy boxes,\"i.e., through the design of the rules that govern the elements of e-commerce sites that promote particular products and prices to consumers. In this paper, we demonstrate that reinforcement learning (RL) can also be used by platforms to learn buy box rules that are effective in preventing collusion by RL sellers. For this, we adopt the methodology of Stackelberg POMDPs, and demonstrate success in learning robust rules that continue to provide high consumer welfare together with sellers employing different behavior models or having out-of-distribution costs for goods."
                    },
                    {
                        "title": "Learning Stackelberg Equilibria and Applications to Economic Design Games",
                        "abstract": ","
                    },
                    {
                        "title": "CrowdPlay: Crowdsourcing Human Demonstrations for Offline Learning",
                        "abstract": "Crowdsourcing has been instrumental for driving AI advances that rely on large-scale data. At the same time, reinforcement learning has seen rapid progress through benchmark environments that strike a balance between tractability and real-world complexity, such as ALE and OpenAI Gym. In this paper, we aim to \ufb01ll a gap at the intersection of these two: The use of crowdsourcing to generate large-scale human demonstration data in the support of advancing research into imitation learning and of\ufb02ine learning. To this end, we present CrowdPlay , a complete crowdsourcing pipeline for any standard RL environment including OpenAI Gym (made available under an open-source license); a large-scale publicly available crowdsourced dataset of human gameplay demonstrations in Atari 2600 games, including multimodal behavior and human-human and human-AI multiagent data; of\ufb02ine learning benchmarks with extensive human data evaluation; and a detailed study of incentives, including real-time feedback to drive high quality data. We hope that this will drive the improvement in design of algorithms that account for the complexity of human, behavioral data and thereby enable a step forward in direction of effective learning for real-world settings. Our code and dataset are available at https://mgerstgrasser.github.io/crowdplay/ ."
                    },
                    {
                        "title": "Credible Decentralized Exchange Design via Verifiable Sequencing Rules",
                        "abstract": "Trading on decentralized exchanges has been one of the primary use cases for permissionless blockchains with daily trading volume exceeding billions of U.S.\u2004dollars. In the status quo, users broadcast transactions they wish to execute in the exchange and miners are responsible for composing a block of transactions and picking an execution ordering \u2014 the order in which transactions execute in the exchange. Due to the lack of a regulatory framework, it is common to observe miners exploiting their privileged position by front-running transactions and obtaining risk-fee profits. Indeed, the Flashbots service institutionalizes this exploit, with miners auctioning the right to front-run transactions. In this work, we propose to modify the interaction between miners and users and initiate the study of verifiable sequencing rules. As in the status quo, miners can determine the content of a block; however, they commit to respecting a sequencing rule that constrains the execution ordering and is verifiable (there is a polynomial time algorithm that can verify if the execution ordering satisfies such constraints). Thus in the event a miner deviates from the sequencing rule, anyone can generate a proof of non-compliance. We ask if there are sequencing rules that limit price manipulation from miners in a two-token liquidity pool exchange. Our first result is an impossibility theorem: for any sequencing rule, there is an instance of user transactions where the miner can obtain non-zero risk-free profits. In light of this impossibility result, our main result is a verifiable sequencing rule that provides execution price guarantees for users. In particular, for any user transaction A, it ensures that either (1) the execution price of A is at least as good as if A was the only transaction in the block, or (2) the execution price of A is worse than this \u201cstandalone\u201d price and the miner does not gain when including A in the block. Our framework does not require users to use countermeasures against predatory trading strategies, for example, set limit prices or split large transactions into smaller ones. This is likely to improve user experience relative to the status quo."
                    },
                    {
                        "title": "Predictive Multiplicity in Probabilistic Classification",
                        "abstract": "Machine learning models are often used to inform real world risk assessment tasks: predicting consumer default risk, predicting whether a person suffers from a serious illness, or predicting a person's risk to appear in court. Given multiple models that perform almost equally well for a prediction task, to what extent do predictions vary across these models? If predictions are relatively consistent for similar models, then the standard approach of choosing the model that optimizes a penalized loss suffices. But what if predictions vary significantly for similar models? In machine learning, this is referred to as predictive multiplicity i.e. the prevalence of conflicting predictions assigned by near-optimal competing models. In this paper, we present a framework for measuring predictive multiplicity in probabilistic classification (predicting the probability of a positive outcome). We introduce measures that capture the variation in risk estimates over the set of competing models, and develop optimization-based methods to compute these measures efficiently and reliably for convex empirical risk minimization problems. We demonstrate the incidence and prevalence of predictive multiplicity in real-world tasks. Further, we provide insight into how predictive multiplicity arises by analyzing the relationship between predictive multiplicity and data set characteristics (outliers, separability, and majority-minority structure). Our results emphasize the need to report predictive multiplicity more widely."
                    },
                    {
                        "title": "Empowering First-Year Computer Science Ph.D. Students to Create a Culture that Values Community and Mental Health",
                        "abstract": "Doctoral programs often have high rates of depression, anxiety, isolation, and imposter phenomenon. Consequently, graduating students may feel inadequately prepared for research-focused careers, contributing to an attrition of talent. Prior work identifies an important contributing factor to maladjustment: even with prior exposure to research, entering Ph.D. students often have problematically idealized views of science. These preconceptions can become obstacles for students in their own professional growth. Unfortunately, existing curricular and extracurricular programming in many doctoral programs fail to include mechanisms to systematically address students' misconceptions of their profession. In this work, we describe a new initiative at our institution that aims to address Ph.D. mental health via a mandatory seminar for entering doctoral students. The seminar is designed to build professional resilience in students by (1) increasing self-regulatory competence, and (2) teaching students to proactively examine academic cultural values and to participate in shaping them. Our evaluation indicates that students improved in both areas after completing the seminar."
                    },
                    {
                        "title": "Learning Robust Helpful Behaviors in Two-Player Cooperative Atari Environments",
                        "abstract": "We initiate the study of helpful behavior in the setting of two-player Atari games, suitably modi\ufb01ed to provide cooperative incentives. Our main interest is to understand whether reinforcement learning can be used to achieve robust, helpful behavior\u2014 where one agent is trained to help a second, partner agent. Robustness requires the helpful AI to be able to cooperate effectively with a diverse set of partners. We study this question with both arti\ufb01cial partner agents as well as human participants (introducing a new, web-based framework for the study of human-with-AI behavior). We achieve positive results in both Space Invaders and Fall Down, as well as successful transfer to human partners, including with people who are asked to deliberately follow unexpected behaviors."
                    },
                    {
                        "title": "Deep Learning for Two-Sided Matching",
                        "abstract": "We initiate the study of deep learning for the automated design of two-sided matching mechanisms. What is of most interest is to use machine learning to understand the possibility of new tradeoffs between strategy-proofness and stability. These properties cannot be achieved simultaneously, but the efficient frontier is not understood. We introduce novel differentiable surrogates for quantifying ordinal strategy-proofness and stability and use them to train differentiable matching mechanisms that map discrete preferences to valid randomized matchings. We demonstrate that the efficient frontier characterized by these learned mechanisms is substantially better than that achievable through a convex combination of baselines of deferred acceptance (stable and strategy-proof for only one side of the market), top trading cycles (strategy-proof for one side, but not stable), and randomized serial dictatorship (strategy-proof for both sides, but not stable). This gives a new target for economic theory and opens up new possibilities for machine learning pipelines in matching market design."
                    },
                    {
                        "title": "Learning Stackelberg Equilibria in Sequential Price Mechanisms",
                        "abstract": "We study the problem of the design of simple economic mechanisms for assigning items to self-interested agents that combine a messaging round with a sequential-pricing stage. The rules of the sequential-pricing stage and in particular the way these rules use messages determines the way the messaging stage is used. This is a Stackelberg game where the designer is the leader and \ufb01xes the mechanism rules, inducing an equilibrium amongst agents (the followers). We model the followers through equilibrium play coming from no-regret learning, and introduce a novel single-agent Stackelberg MDP formulation , where the leader learns to effect a follower equilibrium that optimizes its objective. We solve this MDP us-ing actor-critic methods, where the critic is given access to the joint information of all the agents."
                    },
                    {
                        "title": "Opinion Dynamics Optimization by Varying Susceptibility to Persuasion via Non-Convex Local Search",
                        "abstract": "A long line of work in social psychology has studied variations in people\u2019s susceptibility to persuasion\u2014the extent to which they are willing to modify their opinions on a topic. This body of literature suggests an interesting perspective on theoretical models of opinion formation by interacting parties in a network: in addition to considering interventions that directly modify people\u2019s intrinsic opinions, it is also natural to consider interventions that modify people\u2019s susceptibility to persuasion. In this work, motivated by this fact, we propose an influence optimization problem. Specifically, we adopt a popular model for social opinion dynamics, where each agent has some fixed innate opinion, and a resistance that measures the importance it places on its innate opinion; agents influence one another\u2019s opinions through an iterative process. Under certain conditions, this iterative process converges to some equilibrium opinion vector. For the unbudgeted variant of the problem, the goal is to modify the resistance of any number of agents (within some given range) such that the sum of the equilibrium opinions is minimized; for the budgeted variant, in addition the algorithm is given upfront a restriction on the number of agents whose resistance may be modified. We prove that the objective function is in general non-convex. Hence, formulating the problem as a convex program as in an early version of this work (Abebe et al., KDD\u201918) might have potential correctness issues. We instead analyze the structure of the objective function, and show that any local optimum is also a global optimum, which is somehow surprising as the objective function might not be convex. Furthermore, we combine the iterative process and the local search paradigm to design very efficient algorithms that can solve the unbudgeted variant of the problem optimally on large-scale graphs containing millions of nodes. Finally, we propose and evaluate experimentally a family of heuristics for the budgeted variant of the problem."
                    },
                    {
                        "title": "Dynamic posted-price mechanisms for the blockchain transaction-fee market",
                        "abstract": "In recent years, prominent blockchain systems such as Bitcoin and Ethereum have experienced explosive growth in transaction volume, leading to frequent surges in demand for limited block space and causing transaction fees to fluctuate by orders of magnitude. The status quo auctions sell space using a first-price auction [27]; however, users find it difficult to estimate how much they need to bid in order to get their transactions accepted onto the chain. If they bid too low, their transactions can have long confirmation times. If they bid too high, they pay larger fees than necessary. In light of these issues, new transaction fee mechanisms have been proposed, most notably EIP-1559 [4], aiming to provide better usability. EIP-1559 is a history-dependent mechanism that relies on block utilization to adjust a base fee. We propose an alternative design - a dynamic posted-price mechanism - which uses not only block utilization but also observable bids from past blocks to compute a posted-price for subsequent blocks. We show its potential to reduce price volatility by providing examples for which the prices of EIP-1559 are unstable while the prices of the proposed mechanism are stable. More generally, whenever the demand for the blockchain stabilizes, we ask if our mechanism is able to converge to a stable state. Our main result provides sufficient conditions in a probabilistic setting for which the proposed mechanism is approximately welfare optimal and the prices are stable. Our main technical contribution towards establishing stability is an iterative algorithm that, given oracle access to a Lipschitz continuous and strictly concave function f, converges to a fixed point of f."
                    },
                    {
                        "title": "Low-cost attacks on Ethereum 2.0 by sub-1/3 stakeholders",
                        "abstract": "We outline two dishonest strategies that can be cheaply executed on the Ethereum 2.0 beacon chain, even by validators holding less than one-third of the total stake: malicious chain reorganizations (\"reorgs\") and finality delays. In a malicious reorg, an attacker withholds their blocks and attestations before releasing them at an opportune time in order to force a chain reorganization, which they can take advantage of by double-spending or front-running transactions. To execute a finality delay an attacker uses delayed block releases and withholding of attestations to increase the mean and variance of the time it takes blocks to become finalized. This impacts the efficiency and predictability of the system. We provide a probabilistic and cost analysis for each of these attacks, considering a validator with 30% of the total stake."
                    }
                ]
            },
            "b74f6177-9714-4c78-8728-8ff804f77429": {
                "pk": "b74f6177-9714-4c78-8728-8ff804f77429",
                "name": "Ian Gemp",
                "collaborators": [
                    "Luke Marris",
                    "Yoram Bachrach",
                    "Marc Lanctot",
                    "Georgios Piliouras",
                    "Siqi Liu",
                    "Daniel Hennes",
                    "K. Tuyls",
                    "Kevin R. McKee",
                    "K. Larson",
                    "Paul Muller",
                    "R. \u00c9lie",
                    "Kalesha Bullard",
                    "Vibhavari Dasagi",
                    "Zun Li",
                    "Michael P. Wellman",
                    "G. Piliouras",
                    "Thomas W. Anthony",
                    "Jerome T. Connor",
                    "Richard Everett",
                    "A. Tacchetti",
                    "Chen Chen",
                    "B. McWilliams",
                    "Yiran Mao",
                    "Edgar A. Du'enez-Guzm'an",
                    "Sarah Perrin",
                    "Andras Gyorgy",
                    "Michael Kaisers",
                    "Kate Larson",
                    "Joel Z. Leibo",
                    "Andreas A. Haupt",
                    "N. Heess",
                    "Roma Patel",
                    "Marco Jiralerspong",
                    "A. Bose",
                    "Chongli Qin",
                    "Gauthier Gidel",
                    "Avishkar Bhoopchand",
                    "Bart De Vylder",
                    "Edgar A. Du\u00e9\u00f1ez-Guzm\u00e1n",
                    "Edward Hughes",
                    "Mina Khan",
                    "Guy Lever",
                    "J. P\u00e9rolat",
                    "Florian Strub",
                    "Eugene Tarassov",
                    "Zhe Wang",
                    "Yi Wu",
                    "Yingying Wu",
                    "Elise van der Pol",
                    "Shayegan Omidshafiei",
                    "S. McAleer",
                    "T. Graepel"
                ],
                "domain": [
                    "Game Theory",
                    "Multi-Agent Systems",
                    "Reinforcement Learning",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Approximating the Core via Iterative Coalition Sampling",
                        "abstract": "The core is a central solution concept in cooperative game theory, defined as the set of feasible allocations or payments such that no subset of agents has incentive to break away and form their own subgroup or coalition. However, it has long been known that the core (and approximations, such as the least-core) are hard to compute. This limits our ability to analyze cooperative games in general, and to fully embrace cooperative game theory contributions in domains such as explainable AI (XAI), where the core can complement the Shapley values to identify influential features or instances supporting predictions by black-box models. We propose novel iterative algorithms for computing variants of the core, which avoid the computational bottleneck of many other approaches; namely solving large linear programs. As such, they scale better to very large problems as we demonstrate across different classes of cooperative games, including weighted voting games, induced subgraph games, and marginal contribution networks. We also explore our algorithms in the context of XAI, providing further evidence of the power of the core for such applications."
                    },
                    {
                        "title": "Visualizing 2x2 Normal-Form Games: twoxtwogame LaTeX Package",
                        "abstract": "Normal-form games with two players, each with two strategies, are the most studied class of games. These so-called 2x2 games are used to model a variety of strategic interactions. They appear in game theory, economics, and artificial intelligence research. However, there lacks tools for describing and visualizing such games. This work introduces a LaTeX package for visualizing 2x2 games. This work has two goals: first, to provide high-quality tools and vector graphic visualizations, suitable for scientific publications. And second, to help promote standardization of names and representations of 2x2 games. The LaTeX package, twoxtwogame, is maintained on GitHub and mirrored on CTAN, and is available under a permissive Apache 2 license."
                    },
                    {
                        "title": "Convex Markov Games: A Framework for Fairness, Imitation, and Creativity in Multi-Agent Learning",
                        "abstract": "Expert imitation, behavioral diversity, and fairness preferences give rise to preferences in sequential decision making domains that do not decompose additively across time. We introduce the class of convex Markov games that allow general convex preferences over occupancy measures. Despite infinite time horizon and strictly higher generality than Markov games, pure strategy Nash equilibria exist under strict convexity. Furthermore, equilibria can be approximated efficiently by performing gradient descent on an upper bound of exploitability. Our experiments imitate human choices in ultimatum games, reveal novel solutions to the repeated prisoner's dilemma, and find fair solutions in a repeated asymmetric coordination game. In the prisoner's dilemma, our algorithm finds a policy profile that deviates from observed human play only slightly, yet achieves higher per-player utility while also being three orders of magnitude less exploitable."
                    },
                    {
                        "title": "NfgTransformer: Equivariant Representation Learning for Normal-form Games",
                        "abstract": "Normal-form games (NFGs) are the fundamental model of strategic interaction. We study their representation using neural networks. We describe the inherent equivariance of NFGs -- any permutation of strategies describes an equivalent game -- as well as the challenges this poses for representation learning. We then propose the NfgTransformer architecture that leverages this equivariance, leading to state-of-the-art performance in a range of game-theoretic tasks including equilibrium-solving, deviation gain estimation and ranking, with a common approach to NFG representation. We show that the resulting model is interpretable and versatile, paving the way towards deep learning systems capable of game-theoretic reasoning when interacting with humans and with each other."
                    },
                    {
                        "title": "States as Strings as Strategies: Steering Language Models with Game-Theoretic Solvers",
                        "abstract": "Game theory is the study of mathematical models of strategic interactions among rational agents. Language is a key medium of interaction for humans, though it has historically proven difficult to model dialogue and its strategic motivations mathematically. A suitable model of the players, strategies, and payoffs associated with linguistic interactions (i.e., a binding to the conventional symbolic logic of game theory) would enable existing game-theoretic algorithms to provide strategic solutions in the space of language. In other words, a binding could provide a route to computing stable, rational conversational strategies in dialogue. Large language models (LLMs) have arguably reached a point where their generative capabilities can enable realistic, human-like simulations of natural dialogue. By prompting them in various ways, we can steer their responses towards different output utterances. Leveraging the expressivity of natural language, LLMs can also help us quickly generate new dialogue scenarios, which are grounded in real world applications. In this work, we present one possible binding from dialogue to game theory as well as generalizations of existing equilibrium finding algorithms to this setting. In addition, by exploiting LLMs generation capabilities along with our proposed binding, we can synthesize a large repository of formally-defined games in which one can study and test game-theoretic solution concepts. We also demonstrate how one can combine LLM-driven game generation, game-theoretic solvers, and imitation learning to construct a process for improving the strategic capabilities of LLMs."
                    },
                    {
                        "title": "Search-Improved Game-Theoretic Multiagent Reinforcement Learning in General and Negotiation Games",
                        "abstract": "Multiagent reinforcement learning (MARL) has benefited significantly from population-based and game-theoretic training regimes. One approach, Policy-Space Response Oracles (PSRO), employs standard reinforcement learning to compute response policies via approximate best responses and combines them via meta-strategy selection. We augment PSRO by adding a novel search procedure with generative sampling of world states, and introduce two new meta-strategy solvers based on the Nash bargaining solution. We evaluate PSRO\u2019s ability to compute approximate Nash equilibrium, and its performance in negotiation games: Colored Trails and Deal-or-no-Deal. We conduct behavioral studies where human participants negotiate with our agents ( \ud835\udc41 = 346). Search with generative modeling finds stronger policies during both training time and test time, enables online Bayesian co-player prediction, and can produce agents that achieve comparable social welfare negotiating with humans as humans trading among themselves."
                    },
                    {
                        "title": "Feature Likelihood Divergence: Evaluating the Generalization of Generative Models Using Samples",
                        "abstract": "The past few years have seen impressive progress in the development of deep generative models capable of producing high-dimensional, complex, and photo-realistic data. However, current methods for evaluating such models remain incomplete: standard likelihood-based metrics do not always apply and rarely correlate with perceptual fidelity, while sample-based metrics, such as FID, are insensitive to overfitting, i.e., inability to generalize beyond the training set. To address these limitations, we propose a new metric called the Feature Likelihood Divergence (FLD), a parametric sample-based metric that uses density estimation to provide a comprehensive trichotomic evaluation accounting for novelty (i.e., different from the training samples), fidelity, and diversity of generated samples. We empirically demonstrate the ability of FLD to identify overfitting problem cases, even when previously proposed metrics fail. We also extensively evaluate FLD on various image datasets and model classes, demonstrating its ability to match intuitions of previous metrics like FID while offering a more comprehensive evaluation of generative models. Code is available at https://github.com/marcojira/fld."
                    },
                    {
                        "title": "Approximating Nash Equilibria in Normal-Form Games via Stochastic Optimization",
                        "abstract": "We propose the first loss function for approximate Nash equilibria of normal-form games that is amenable to unbiased Monte Carlo estimation. This construction allows us to deploy standard non-convex stochastic optimization techniques for approximating Nash equilibria, resulting in novel algorithms with provable guarantees. We complement our theoretical analysis with experiments demonstrating that stochastic gradient descent can outperform previous state-of-the-art approaches."
                    },
                    {
                        "title": "Equilibrium-Invariant Embedding, Metric Space, and Fundamental Set of 2\u00d72 Normal-Form Games",
                        "abstract": "Equilibrium solution concepts of normal-form games, such as Nash equilibria, correlated equilibria, and coarse correlated equilibria, describe the joint strategy profiles from which no player has incentive to unilaterally deviate. They are widely studied in game theory, economics, and multiagent systems. Equilibrium concepts are invariant under certain transforms of the payoffs. We define an equilibrium-inspired distance metric for the space of all normal-form games and uncover a distance-preserving equilibrium-invariant embedding. Furthermore, we propose an additional transform which defines a better-response-invariant distance metric and embedding. To demonstrate these metric spaces we study $2\\times2$ games. The equilibrium-invariant embedding of $2\\times2$ games has an efficient two variable parameterization (a reduction from eight), where each variable geometrically describes an angle on a unit circle. Interesting properties can be spatially inferred from the embedding, including: equilibrium support, cycles, competition, coordination, distances, best-responses, and symmetries. The best-response-invariant embedding of $2\\times2$ games, after considering symmetries, rediscovers a set of 15 games, and their respective equivalence classes. We propose that this set of game classes is fundamental and captures all possible interesting strategic interactions in $2\\times2$ games. We introduce a directed graph representation and name for each class. Finally, we leverage the tools developed for $2\\times2$ games to develop game theoretic visualizations of large normal-form and extensive-form games that aim to fingerprint the strategic interactions that occur within."
                    },
                    {
                        "title": "Combining Tree-Search, Generative Models, and Nash Bargaining Concepts in Game-Theoretic Reinforcement Learning",
                        "abstract": "Multiagent reinforcement learning (MARL) has benefited significantly from population-based and game-theoretic training regimes. One approach, Policy-Space Response Oracles (PSRO), employs standard reinforcement learning to compute response policies via approximate best responses and combines them via meta-strategy selection. We augment PSRO by adding a novel search procedure with generative sampling of world states, and introduce two new meta-strategy solvers based on the Nash bargaining solution. We evaluate PSRO's ability to compute approximate Nash equilibrium, and its performance in two negotiation games: Colored Trails, and Deal or No Deal. We conduct behavioral studies where human participants negotiate with our agents ($N = 346$). We find that search with generative modeling finds stronger policies during both training time and test time, enables online Bayesian co-player prediction, and can produce agents that achieve comparable social welfare negotiating with humans as humans trading among themselves."
                    },
                    {
                        "title": "Developing, Evaluating and Scaling Learning Agents in Multi-Agent Environments",
                        "abstract": "The Game Theory & Multi-Agent team at DeepMind studies several aspects of multi-agent learning ranging from computing approximations to fundamental concepts in game theory to simulating social dilemmas in rich spatial environments and training 3-d humanoids in difficult team coordination tasks. A signature aim of our group is to use the resources and expertise made available to us at DeepMind in deep reinforcement learning to explore multi-agent systems in complex environments and use these benchmarks to advance our understanding. Here, we summarise the recent work of our team and present a taxonomy that we feel highlights many important open challenges in multi-agent research."
                    },
                    {
                        "title": "The Generalized Eigenvalue Problem as a Nash Equilibrium",
                        "abstract": "The generalized eigenvalue problem (GEP) is a fundamental concept in numerical linear algebra. It captures the solution of many classical machine learning problems such as canonical correlation analysis, independent components analysis, partial least squares, linear discriminant analysis, principal components, successor features and others. Despite this, most general solvers are prohibitively expensive when dealing with massive data sets and research has instead concentrated on \ufb01nding e\ufb03cient solutions to speci\ufb01c problem instances. In this work, we develop a game-theoretic formulation of the top-k GEP whose Nash equilibrium is the set of generalized eigenvectors. We also present a parallelizable algorithm with guaranteed asymptotic convergence to the Nash. Current state-of-the-art methods require O ( d 2 k ) complexity per iteration which is prohibitively expensive when the number of dimensions ( d ) is large. We show how to achieve O ( dk ) complexity, scaling to datasets 100 \u00d7 larger than those evaluated by prior methods. Empirically we demonstrate that our algorithm is able to solve a variety of GEP problem instances including a large-scale analysis of neural network activations."
                    },
                    {
                        "title": "AlphaSnake: Policy Iteration on a Nondeterministic NP-hard Markov Decision Process",
                        "abstract": "Reinforcement learning has been used to approach well-known NP-hard combinatorial problems in graph theory. Among these, Hamiltonian cycle problems are exceptionally difficult to analyze, even when restricted to individual instances of structurally complex graphs. In this paper, we use Monte Carlo Tree Search (MCTS), the search algorithm behind many state-of-the-art reinforcement learning algorithms such as AlphaZero, to create autonomous agents that learn to play the game of Snake, a game centered on properties of Hamiltonian cycles on grid graphs. The game of Snake can be formulated as a single-player discounted Markov Decision Process (MDP), where the agent must behave optimally in a stochastic environment. Determining the optimal policy for Snake, defined as the policy that maximizes the probability of winning -- or win rate -- with higher priority and minimizes the expected number of time steps to win with lower priority, is conjectured to be NP-hard. Performance-wise, compared to prior work in the Snake game, our algorithm is the first to achieve a win rate over 0.5 (a uniform random policy achieves a win rate < 2.57 x 10^{-15}), demonstrating the versatility of AlphaZero in tackling NP-hard problems."
                    },
                    {
                        "title": "Stochastic Parallelizable Eigengap Dilation for Large Graph Clustering",
                        "abstract": "Large graphs commonly appear in social networks, knowledge graphs, recommender systems, life sciences, and decision making problems. Summarizing large graphs by their high level properties is helpful in solving problems in these settings. In spectral clustering, we aim to identify clusters of nodes where most edges fall within clusters and only few edges fall between clusters. This task is important for many downstream applications and exploratory analysis. A core step of spectral clustering is performing an eigendecomposition of the corresponding graph Laplacian matrix (or equivalently, a singular value decomposition, SVD, of the incidence matrix). The convergence of iterative singular value decomposition approaches depends on the eigengaps of the spectrum of the given matrix, i.e., the difference between consecutive eigenvalues. For a graph Laplacian corresponding to a well-clustered graph, the eigenvalues will be non-negative but very small (much less than $1$) slowing convergence. This paper introduces a parallelizable approach to dilating the spectrum in order to accelerate SVD solvers and in turn, spectral clustering. This is accomplished via polynomial approximations to matrix operations that favorably transform the spectrum of a matrix without changing its eigenvectors. Experiments demonstrate that this approach significantly accelerates convergence, and we explain how this transformation can be parallelized and stochastically approximated to scale with available compute."
                    },
                    {
                        "title": "Turbocharging Solution Concepts: Solving NEs, CEs and CCEs with Neural Equilibrium Solvers",
                        "abstract": "Solution concepts such as Nash Equilibria, Correlated Equilibria, and Coarse Correlated Equilibria are useful components for many multiagent machine learning algorithms. Unfortunately, solving a normal-form game could take prohibitive or non-deterministic time to converge, and could fail. We introduce the Neural Equilibrium Solver which utilizes a special equivariant neural network architecture to approximately solve the space of all games of fixed shape, buying speed and determinism. We define a flexible equilibrium selection framework, that is capable of uniquely selecting an equilibrium that minimizes relative entropy, or maximizes welfare. The network is trained without needing to generate any supervised training data. We show remarkable zero-shot generalization to larger games. We argue that such a network is a powerful component for many possible multiagent algorithms."
                    },
                    {
                        "title": "Game Theoretic Rating in N-player general-sum games with Equilibria",
                        "abstract": "Rating strategies in a game is an important area of research in game theory and arti\ufb01cial intelligence, and can be applied to any real-world competitive or cooperative setting. Traditionally, only transitive dependencies between strategies have been used to rate strategies (e.g. Elo), however recent work has expanded ratings to utilize game theoretic solutions to better rate strategies in non-transitive games. This work generalizes these ideas and proposes novel algorithms suitable for N-player, general-sum rating of strategies in normal-form games according to the payoff rating system. This enables well-established solution concepts, such as equilibria, to be leveraged to ef\ufb01ciently rate strategies in games with complex strategic interactions, which arise in multiagent training and real-world interactions between many agents. We empirically validate our methods on real world normal-form data (Premier League) and multiagent reinforcement learning agent evaluation."
                    },
                    {
                        "title": "The Symmetric Generalized Eigenvalue Problem as a Nash Equilibrium",
                        "abstract": "The symmetric generalized eigenvalue problem (SGEP) is a fundamental concept in numerical linear algebra. It captures the solution of many classical machine learning problems such as canonical correlation analysis, independent components analysis, partial least squares, linear discriminant analysis, principal components and others. Despite this, most general solvers are prohibitively expensive when dealing with streaming data sets (i.e., minibatches) and research has instead concentrated on finding efficient solutions to specific problem instances. In this work, we develop a game-theoretic formulation of the top-$k$ SGEP whose Nash equilibrium is the set of generalized eigenvectors. We also present a parallelizable algorithm with guaranteed asymptotic convergence to the Nash. Current state-of-the-art methods require $O(d^2k)$ runtime complexity per iteration which is prohibitively expensive when the number of dimensions ($d$) is large. We show how to modify this parallel approach to achieve $O(dk)$ runtime complexity. Empirically we demonstrate that this resulting algorithm is able to solve a variety of SGEP problem instances including a large-scale analysis of neural network activations."
                    }
                ]
            },
            "415bc002-fc9e-421c-a380-461443452929": {
                "pk": "415bc002-fc9e-421c-a380-461443452929",
                "name": "Luke Marris",
                "collaborators": [
                    "I. Gemp",
                    "Marc Lanctot",
                    "Siqi Liu",
                    "Daniel Hennes",
                    "K. Tuyls",
                    "Georgios Piliouras",
                    "Yoram Bachrach",
                    "N. Heess",
                    "Paul Muller",
                    "T. Graepel",
                    "Zun Li",
                    "Thomas W. Anthony",
                    "Kevin R. McKee",
                    "K. Larson",
                    "Guy Lever",
                    "Zhe Wang",
                    "Shayegan Omidshafiei",
                    "R. \u00c9lie",
                    "Kalesha Bullard",
                    "Kate Larson",
                    "Joel Z. Leibo",
                    "Vibhavari Dasagi",
                    "Avishkar Bhoopchand",
                    "Michael P. Wellman",
                    "G. Piliouras",
                    "Jerome T. Connor",
                    "Edward Hughes",
                    "J. P\u00e9rolat",
                    "A. Tacchetti",
                    "J. Merel",
                    "Yiran Mao",
                    "Edgar A. Du'enez-Guzm'an",
                    "Sarah Perrin",
                    "Andras Gyorgy",
                    "Michael Kaisers",
                    "Andreas A. Haupt",
                    "Roma Patel",
                    "Brian Tanner",
                    "Anna Koop",
                    "Bart De Vylder",
                    "Edgar A. Du\u00e9\u00f1ez-Guzm\u00e1n",
                    "Richard Everett",
                    "Mina Khan",
                    "Florian Strub",
                    "Eugene Tarassov",
                    "S. McAleer",
                    "S. Eslami",
                    "Wojciech M. Czarnecki",
                    "Yuval Tassa",
                    "A. Abdolmaleki",
                    "Noah Siegel",
                    "Leonard Hasenclever",
                    "S. Tunyasuvunakool",
                    "H. F. Song",
                    "Markus Wulfmeier",
                    "Tuomas Haarnoja",
                    "Brendan D. Tracey",
                    "Mark Rowland",
                    "R. Munos"
                ],
                "domain": [
                    "Game Theory",
                    "Multi-Agent Systems",
                    "Reinforcement Learning",
                    "Cooperative Games"
                ],
                "publications": [
                    {
                        "title": "Approximating the Core via Iterative Coalition Sampling",
                        "abstract": "The core is a central solution concept in cooperative game theory, defined as the set of feasible allocations or payments such that no subset of agents has incentive to break away and form their own subgroup or coalition. However, it has long been known that the core (and approximations, such as the least-core) are hard to compute. This limits our ability to analyze cooperative games in general, and to fully embrace cooperative game theory contributions in domains such as explainable AI (XAI), where the core can complement the Shapley values to identify influential features or instances supporting predictions by black-box models. We propose novel iterative algorithms for computing variants of the core, which avoid the computational bottleneck of many other approaches; namely solving large linear programs. As such, they scale better to very large problems as we demonstrate across different classes of cooperative games, including weighted voting games, induced subgraph games, and marginal contribution networks. We also explore our algorithms in the context of XAI, providing further evidence of the power of the core for such applications."
                    },
                    {
                        "title": "Visualizing 2x2 Normal-Form Games: twoxtwogame LaTeX Package",
                        "abstract": "Normal-form games with two players, each with two strategies, are the most studied class of games. These so-called 2x2 games are used to model a variety of strategic interactions. They appear in game theory, economics, and artificial intelligence research. However, there lacks tools for describing and visualizing such games. This work introduces a LaTeX package for visualizing 2x2 games. This work has two goals: first, to provide high-quality tools and vector graphic visualizations, suitable for scientific publications. And second, to help promote standardization of names and representations of 2x2 games. The LaTeX package, twoxtwogame, is maintained on GitHub and mirrored on CTAN, and is available under a permissive Apache 2 license."
                    },
                    {
                        "title": "Convex Markov Games: A Framework for Fairness, Imitation, and Creativity in Multi-Agent Learning",
                        "abstract": "Expert imitation, behavioral diversity, and fairness preferences give rise to preferences in sequential decision making domains that do not decompose additively across time. We introduce the class of convex Markov games that allow general convex preferences over occupancy measures. Despite infinite time horizon and strictly higher generality than Markov games, pure strategy Nash equilibria exist under strict convexity. Furthermore, equilibria can be approximated efficiently by performing gradient descent on an upper bound of exploitability. Our experiments imitate human choices in ultimatum games, reveal novel solutions to the repeated prisoner's dilemma, and find fair solutions in a repeated asymmetric coordination game. In the prisoner's dilemma, our algorithm finds a policy profile that deviates from observed human play only slightly, yet achieves higher per-player utility while also being three orders of magnitude less exploitable."
                    },
                    {
                        "title": "NfgTransformer: Equivariant Representation Learning for Normal-form Games",
                        "abstract": "Normal-form games (NFGs) are the fundamental model of strategic interaction. We study their representation using neural networks. We describe the inherent equivariance of NFGs -- any permutation of strategies describes an equivalent game -- as well as the challenges this poses for representation learning. We then propose the NfgTransformer architecture that leverages this equivariance, leading to state-of-the-art performance in a range of game-theoretic tasks including equilibrium-solving, deviation gain estimation and ranking, with a common approach to NFG representation. We show that the resulting model is interpretable and versatile, paving the way towards deep learning systems capable of game-theoretic reasoning when interacting with humans and with each other."
                    },
                    {
                        "title": "Neural Population Learning beyond Symmetric Zero-sum Games",
                        "abstract": "We study computationally efficient methods for finding equilibria in n-player general-sum games, specifically ones that afford complex visuomotor skills. We show how existing methods would struggle in this setting, either computationally or in theory. We then introduce NeuPL-JPSRO, a neural population learning algorithm that benefits from transfer learning of skills and converges to a Coarse Correlated Equilibrium (CCE) of the game. We show empirical convergence in a suite of OpenSpiel games, validated rigorously by exact game solvers. We then deploy NeuPL-JPSRO to complex domains, where our approach enables adaptive coordination in a MuJoCo control domain and skill transfer in capture-the-flag. Our work shows that equilibrium convergent population learning can be implemented at scale and in generality, paving the way towards solving real-world games between heterogeneous players with mixed motives."
                    },
                    {
                        "title": "States as Strings as Strategies: Steering Language Models with Game-Theoretic Solvers",
                        "abstract": "Game theory is the study of mathematical models of strategic interactions among rational agents. Language is a key medium of interaction for humans, though it has historically proven difficult to model dialogue and its strategic motivations mathematically. A suitable model of the players, strategies, and payoffs associated with linguistic interactions (i.e., a binding to the conventional symbolic logic of game theory) would enable existing game-theoretic algorithms to provide strategic solutions in the space of language. In other words, a binding could provide a route to computing stable, rational conversational strategies in dialogue. Large language models (LLMs) have arguably reached a point where their generative capabilities can enable realistic, human-like simulations of natural dialogue. By prompting them in various ways, we can steer their responses towards different output utterances. Leveraging the expressivity of natural language, LLMs can also help us quickly generate new dialogue scenarios, which are grounded in real world applications. In this work, we present one possible binding from dialogue to game theory as well as generalizations of existing equilibrium finding algorithms to this setting. In addition, by exploiting LLMs generation capabilities along with our proposed binding, we can synthesize a large repository of formally-defined games in which one can study and test game-theoretic solution concepts. We also demonstrate how one can combine LLM-driven game generation, game-theoretic solvers, and imitation learning to construct a process for improving the strategic capabilities of LLMs."
                    },
                    {
                        "title": "Evaluating Agents using Social Choice Theory",
                        "abstract": "We argue that many general evaluation problems can be viewed through the lens of voting theory. Each task is interpreted as a separate voter, which requires only ordinal rankings or pairwise comparisons of agents to produce an overall evaluation. By viewing the aggregator as a social welfare function, we are able to leverage centuries of research in social choice theory to derive principled evaluation frameworks with axiomatic foundations. These evaluations are interpretable and flexible, while avoiding many of the problems currently facing cross-task evaluation. We apply this Voting-as-Evaluation (VasE) framework across multiple settings, including reinforcement learning, large language models, and humans. In practice, we observe that VasE can be more robust than popular evaluation frameworks (Elo and Nash averaging), discovers properties in the evaluation data not evident from scores alone, and can predict outcomes better than Elo in a complex seven-player game. We identify one particular approach, maximal lotteries, that satisfies important consistency properties relevant to evaluation, is computationally efficient (polynomial in the size of the evaluation data), and identifies game-theoretic cycles."
                    },
                    {
                        "title": "Search-Improved Game-Theoretic Multiagent Reinforcement Learning in General and Negotiation Games",
                        "abstract": "Multiagent reinforcement learning (MARL) has benefited significantly from population-based and game-theoretic training regimes. One approach, Policy-Space Response Oracles (PSRO), employs standard reinforcement learning to compute response policies via approximate best responses and combines them via meta-strategy selection. We augment PSRO by adding a novel search procedure with generative sampling of world states, and introduce two new meta-strategy solvers based on the Nash bargaining solution. We evaluate PSRO\u2019s ability to compute approximate Nash equilibrium, and its performance in negotiation games: Colored Trails and Deal-or-no-Deal. We conduct behavioral studies where human participants negotiate with our agents ( \ud835\udc41 = 346). Search with generative modeling finds stronger policies during both training time and test time, enables online Bayesian co-player prediction, and can produce agents that achieve comparable social welfare negotiating with humans as humans trading among themselves."
                    },
                    {
                        "title": "Approximating Nash Equilibria in Normal-Form Games via Stochastic Optimization",
                        "abstract": "We propose the first loss function for approximate Nash equilibria of normal-form games that is amenable to unbiased Monte Carlo estimation. This construction allows us to deploy standard non-convex stochastic optimization techniques for approximating Nash equilibria, resulting in novel algorithms with provable guarantees. We complement our theoretical analysis with experiments demonstrating that stochastic gradient descent can outperform previous state-of-the-art approaches."
                    },
                    {
                        "title": "Equilibrium-Invariant Embedding, Metric Space, and Fundamental Set of 2\u00d72 Normal-Form Games",
                        "abstract": "Equilibrium solution concepts of normal-form games, such as Nash equilibria, correlated equilibria, and coarse correlated equilibria, describe the joint strategy profiles from which no player has incentive to unilaterally deviate. They are widely studied in game theory, economics, and multiagent systems. Equilibrium concepts are invariant under certain transforms of the payoffs. We define an equilibrium-inspired distance metric for the space of all normal-form games and uncover a distance-preserving equilibrium-invariant embedding. Furthermore, we propose an additional transform which defines a better-response-invariant distance metric and embedding. To demonstrate these metric spaces we study $2\\times2$ games. The equilibrium-invariant embedding of $2\\times2$ games has an efficient two variable parameterization (a reduction from eight), where each variable geometrically describes an angle on a unit circle. Interesting properties can be spatially inferred from the embedding, including: equilibrium support, cycles, competition, coordination, distances, best-responses, and symmetries. The best-response-invariant embedding of $2\\times2$ games, after considering symmetries, rediscovers a set of 15 games, and their respective equivalence classes. We propose that this set of game classes is fundamental and captures all possible interesting strategic interactions in $2\\times2$ games. We introduce a directed graph representation and name for each class. Finally, we leverage the tools developed for $2\\times2$ games to develop game theoretic visualizations of large normal-form and extensive-form games that aim to fingerprint the strategic interactions that occur within."
                    },
                    {
                        "title": "Combining Tree-Search, Generative Models, and Nash Bargaining Concepts in Game-Theoretic Reinforcement Learning",
                        "abstract": "Multiagent reinforcement learning (MARL) has benefited significantly from population-based and game-theoretic training regimes. One approach, Policy-Space Response Oracles (PSRO), employs standard reinforcement learning to compute response policies via approximate best responses and combines them via meta-strategy selection. We augment PSRO by adding a novel search procedure with generative sampling of world states, and introduce two new meta-strategy solvers based on the Nash bargaining solution. We evaluate PSRO's ability to compute approximate Nash equilibrium, and its performance in two negotiation games: Colored Trails, and Deal or No Deal. We conduct behavioral studies where human participants negotiate with our agents ($N = 346$). We find that search with generative modeling finds stronger policies during both training time and test time, enables online Bayesian co-player prediction, and can produce agents that achieve comparable social welfare negotiating with humans as humans trading among themselves."
                    },
                    {
                        "title": "Developing, Evaluating and Scaling Learning Agents in Multi-Agent Environments",
                        "abstract": "The Game Theory & Multi-Agent team at DeepMind studies several aspects of multi-agent learning ranging from computing approximations to fundamental concepts in game theory to simulating social dilemmas in rich spatial environments and training 3-d humanoids in difficult team coordination tasks. A signature aim of our group is to use the resources and expertise made available to us at DeepMind in deep reinforcement learning to explore multi-agent systems in complex environments and use these benchmarks to advance our understanding. Here, we summarise the recent work of our team and present a taxonomy that we feel highlights many important open challenges in multi-agent research."
                    },
                    {
                        "title": "NeuPL: Neural Population Learning",
                        "abstract": "Learning in strategy games (e.g. StarCraft, poker) requires the discovery of diverse policies. This is often achieved by iteratively training new policies against existing ones, growing a policy population that is robust to exploit. This iterative approach suffers from two issues in real-world games: a) under finite budget, approximate best-response operators at each iteration needs truncating, resulting in under-trained good-responses populating the population; b) repeated learning of basic skills at each iteration is wasteful and becomes intractable in the presence of increasingly strong opponents. In this work, we propose Neural Population Learning (NeuPL) as a solution to both issues. NeuPL offers convergence guarantees to a population of best-responses under mild assumptions. By representing a population of policies within a single conditional model, NeuPL enables transfer learning across policies. Empirically, we show the generality, improved performance and efficiency of NeuPL across several test domains. Most interestingly, we show that novel strategies become more accessible, not less, as the neural population expands."
                    },
                    {
                        "title": "Turbocharging Solution Concepts: Solving NEs, CEs and CCEs with Neural Equilibrium Solvers",
                        "abstract": "Solution concepts such as Nash Equilibria, Correlated Equilibria, and Coarse Correlated Equilibria are useful components for many multiagent machine learning algorithms. Unfortunately, solving a normal-form game could take prohibitive or non-deterministic time to converge, and could fail. We introduce the Neural Equilibrium Solver which utilizes a special equivariant neural network architecture to approximately solve the space of all games of fixed shape, buying speed and determinism. We define a flexible equilibrium selection framework, that is capable of uniquely selecting an equilibrium that minimizes relative entropy, or maximizes welfare. The network is trained without needing to generate any supervised training data. We show remarkable zero-shot generalization to larger games. We argue that such a network is a powerful component for many possible multiagent algorithms."
                    },
                    {
                        "title": "Game Theoretic Rating in N-player general-sum games with Equilibria",
                        "abstract": "Rating strategies in a game is an important area of research in game theory and arti\ufb01cial intelligence, and can be applied to any real-world competitive or cooperative setting. Traditionally, only transitive dependencies between strategies have been used to rate strategies (e.g. Elo), however recent work has expanded ratings to utilize game theoretic solutions to better rate strategies in non-transitive games. This work generalizes these ideas and proposes novel algorithms suitable for N-player, general-sum rating of strategies in normal-form games according to the payoff rating system. This enables well-established solution concepts, such as equilibria, to be leveraged to ef\ufb01ciently rate strategies in games with complex strategic interactions, which arise in multiagent training and real-world interactions between many agents. We empirically validate our methods on real world normal-form data (Premier League) and multiagent reinforcement learning agent evaluation."
                    },
                    {
                        "title": "Simplex Neural Population Learning: Any-Mixture Bayes-Optimality in Symmetric Zero-sum Games",
                        "abstract": "Learning to play optimally against any mixture over a diverse set of strategies is of important practical interests in competitive games. In this paper, we propose simplex-NeuPL that satisfies two desiderata simultaneously: i) learning a population of strategically diverse basis policies, represented by a single conditional network; ii) using the same network, learn best-responses to any mixture over the simplex of basis policies. We show that the resulting conditional policies incorporate prior information about their opponents effectively, enabling near optimal returns against arbitrary mixture policies in a game with tractable best-responses. We verify that such policies behave Bayes-optimally under uncertainty and offer insights in using this flexibility at test time. Finally, we offer evidence that learning best-responses to any mixture policies is an effective auxiliary task for strategic exploration, which, by itself, can lead to more performant populations."
                    },
                    {
                        "title": "Multi-Agent Training beyond Zero-Sum with Correlated Equilibrium Meta-Solvers",
                        "abstract": "Two-player, constant-sum games are well studied in the literature, but there has been limited progress outside of this setting. We propose Joint Policy-Space Response Oracles (JPSRO), an algorithm for training agents in n-player, general-sum extensive form games, which provably converges to an equilibrium. We further suggest correlated equilibria (CE) as promising meta-solvers, and propose a novel solution concept Maximum Gini Correlated Equilibrium (MGCE), a principled and computationally efficient family of solutions for solving the correlated equilibrium selection problem. We conduct several experiments using CE meta-solvers for JPSRO and demonstrate convergence on n-player, general-sum games."
                    },
                    {
                        "title": "From Motor Control to Team Play in Simulated Humanoid Football",
                        "abstract": "Learning to combine control at the level of joint torques with longer-term goal-directed behavior is a long-standing challenge for physically embodied artificial agents. Intelligent behavior in the physical world unfolds across multiple spatial and temporal scales: Although movements are ultimately executed at the level of instantaneous muscle tensions or joint torques, they must be selected to serve goals that are defined on much longer time scales and that often involve complex interactions with the environment and other agents. Recent research has demonstrated the potential of learning-based approaches applied to the respective problems of complex movement, long-term planning, and multiagent coordination. However, their integration traditionally required the design and optimization of independent subsystems and remains challenging. In this work, we tackled the integration of motor control and long-horizon decision-making in the context of simulated humanoid football, which requires agile motor control and multiagent coordination. We optimized teams of agents to play simulated football via reinforcement learning, constraining the solution space to that of plausible movements learned using human motion capture data. They were trained to maximize several environment rewards and to imitate pretrained football-specific skills if doing so led to improved performance. The result is a team of coordinated humanoid football players that exhibit complex behavior at different scales, quantified by a range of analysis and statistics, including those used in real-world sport analytics. Our work constitutes a complete demonstration of learned integrated decision-making at multiple scales in a multiagent setting."
                    },
                    {
                        "title": "A Generalized Training Approach for Multiagent Learning",
                        "abstract": "This paper investigates a population-based training regime based on game-theoretic principles called Policy-Spaced Response Oracles (PSRO). PSRO is general in the sense that it (1) encompasses well-known algorithms such as fictitious play and double oracle as special cases, and (2) in principle applies to general-sum, many-player games. Despite this, prior studies of PSRO have been focused on two-player zero-sum games, a regime wherein Nash equilibria are tractably computable. In moving from two-player zero-sum games to more general settings, computation of Nash equilibria quickly becomes infeasible. Here, we extend the theoretical underpinnings of PSRO by considering an alternative solution concept, $\\alpha$-Rank, which is unique (thus faces no equilibrium selection issues, unlike Nash) and applies readily to general-sum, many-player settings. We establish convergence guarantees in several games classes, and identify links between Nash equilibria and $\\alpha$-Rank. We demonstrate the competitive performance of $\\alpha$-Rank-based PSRO against an exact Nash solver-based PSRO in 2-player Kuhn and Leduc Poker. We then go beyond the reach of prior PSRO applications by considering 3- to 5-player poker games, yielding instances where $\\alpha$-Rank achieves faster convergence than approximate Nash solvers, thus establishing it as a favorable general games solver. We also carry out an initial empirical validation in MuJoCo soccer, illustrating the feasibility of the proposed approach in another complex domain."
                    }
                ]
            },
            "e54dae5c-2d54-4f54-8202-90df1af9f914": {
                "pk": "e54dae5c-2d54-4f54-8202-90df1af9f914",
                "name": "Georgios Piliouras",
                "collaborators": [
                    "A. Hussain",
                    "Antonios Varvitsiotis",
                    "Iosif Sakos",
                    "Wayne Lin",
                    "F. Belardinelli",
                    "Pedro Braga",
                    "Georgios Chionas",
                    "Piotr Krysta",
                    "Stefanos Leonardos",
                    "Carmine Ventre",
                    "Ilayda Canyakmaz",
                    "Ryann Sim",
                    "Emmanouil-Vasileios Vlatakis-Gkaragkounis",
                    "P. Mertikopoulos",
                    "D.G. Leonte",
                    "Francesco Belardinelli",
                    "Yi Feng",
                    "Xiao Wang",
                    "V. Cevher",
                    "Stratis Skoulakis",
                    "Shuyue Hu",
                    "Harold Soh",
                    "Francisca Vasconcelos",
                    "Michael I. Jordan",
                    "D. Leonte",
                    "Minae Kwon",
                    "J. Agapiou",
                    "Romuald",
                    "Elie",
                    "Kalesha Bullard"
                ],
                "domain": [
                    "Game Theory",
                    "Multi-Agent Systems",
                    "Blockchain",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Who gets the Maximal Extractable Value? A Dynamic Sharing Blockchain Mechanism",
                        "abstract": "Maximal Extractable Value (MEV) has emerged as a new frontier in the design of blockchain systems. MEV refers to any excess value that a block producer can realize by manipulating the ordering of transactions. In this paper, we propose to make the MEV extraction rate part of the protocol design space. Our aim is to leverage this parameter to maintain a healthy balance between block producers (who need to be compensated for the service they provide) and users (who need to feel encouraged to transact). We design a dynamic mechanism which updates the MEV extraction rate with the goal of stabilizing it at a target value. We analyse the evolution of this dynamic mechanism under various market conditions and provide formal guarantees about its long-term performance. The main take-away from our work is that the proposed system exhibits desirable behavior (near-optimal performance) even when it operates in out of equilibrium conditions that are often met in practice. Our work establishes, the first to our knowledge, dynamic framework for the integral problem of MEV sharing between extractors and users."
                    },
                    {
                        "title": "On the Stability of Learning in Network Games with Many Players",
                        "abstract": "Multi-agent learning algorithms have been shown to display complex, unstable behaviours in a wide array of games. In fact, previous works indicate that convergent behaviours are less likely to occur as the total number of agents increases. This seemingly prohibits convergence to stable strategies, such as Nash Equilibria, in games with many players. To make progress towards addressing this challenge we study the Q-Learning Dynamics, a classical model for exploration and exploitation in multi-agent learning. In particular, we study the behaviour of Q-Learning on games where interactions between agents are constrained by a network. We determine a number of sufficient conditions, depending on the game and network structure, which guarantee that agent strategies converge to a unique stable strategy, called the Quantal Response Equilibrium (QRE). Crucially, these sufficient conditions are independent of the total number of agents, allowing for provable convergence in arbitrarily large games. Next, we compare the learned QRE to the underlying NE of the game, by showing that any QRE is an $\\epsilon$-approximate Nash Equilibrium. We first provide tight bounds on $\\epsilon$ and show how these bounds lead naturally to a centralised scheme for choosing exploration rates, which enables independent learners to learn stable approximate Nash Equilibrium strategies. We validate the method through experiments and demonstrate its effectiveness even in the presence of numerous agents and actions. Through these results, we show that independent learning dynamics may converge to approximate Nash Equilibria, even in the presence of many agents."
                    },
                    {
                        "title": "Steering game dynamics towards desired outcomes",
                        "abstract": "The dynamic behavior of agents in games, which captures how their strategies evolve over time based on past interactions, can lead to a spectrum of undesirable behaviors, ranging from non-convergence to Nash equilibria to the emergence of limit cycles and chaos. To mitigate the effects of selfish behavior, central planners can use dynamic payments to guide strategic multi-agent systems toward stability and socially optimal outcomes. However, the effectiveness of such interventions critically relies on accurately predicting agents' responses to incentives and dynamically adjusting payments so that the system is guided towards the desired outcomes. These challenges are further amplified in real-time applications where the dynamics are unknown and only scarce data is available. To tackle this challenge, in this work we introduce the SIAR-MPC method, combining the recently introduced Side Information Assisted Regression (SIAR) method for system identification with Model Predictive Control (MPC). SIAR utilizes side-information constraints inherent to game theoretic applications to model agent responses to payments from scarce data, while MPC uses this model to facilitate dynamic payment adjustments. Our experiments demonstrate the efficiency of SIAR-MPC in guiding the system towards socially optimal equilibria, stabilizing chaotic behaviors, and avoiding specified regions of the state space. Comparative analyses in data-scarce settings show SIAR-MPC's superior performance over pairing MPC with Physics Informed Neural Networks (PINNs), a powerful system identification method that finds models satisfying specific constraints."
                    },
                    {
                        "title": "Prediction Accuracy of Learning in Games : Follow-the-Regularized-Leader meets Heisenberg",
                        "abstract": "We investigate the accuracy of prediction in deterministic learning dynamics of zero-sum games with random initializations, specifically focusing on observer uncertainty and its relationship to the evolution of covariances. Zero-sum games are a prominent field of interest in machine learning due to their various applications. Concurrently, the accuracy of prediction in dynamical systems from mechanics has long been a classic subject of investigation since the discovery of the Heisenberg Uncertainty Principle. This principle employs covariance and standard deviation of particle states to measure prediction accuracy. In this study, we bring these two approaches together to analyze the Follow-the-Regularized-Leader (FTRL) algorithm in two-player zero-sum games. We provide growth rates of covariance information for continuous-time FTRL, as well as its two canonical discretization methods (Euler and Symplectic). A Heisenberg-type inequality is established for FTRL. Our analysis and experiments also show that employing Symplectic discretization enhances the accuracy of prediction in learning dynamics."
                    },
                    {
                        "title": "MEV Sharing with Dynamic Extraction Rates",
                        "abstract": "Maximal Extractable Value (MEV) has emerged as a new frontier in the design of blockchain systems. In this paper, we propose making the MEV extraction rate as part of the protocol design space. Our aim is to leverage this parameter to maintain a healthy balance between block producers (who need to be compensated) and users (who need to feel encouraged to transact). We follow the approach introduced by EIP-1559 and design a similar mechanism to dynamically update the MEV extraction rate with the goal of stabilizing it at a target value. We study the properties of this dynamic mechanism and show that, while convergence to the target can be guaranteed for certain parameters, instability, and even chaos, can occur in other cases. Despite these complexities, under general conditions, the system concentrates in a neighborhood of the target equilibrium implying high long-term performance. Our work establishes, the first to our knowledge, dynamic framework for the integral problem of MEV sharing between extractors and users."
                    },
                    {
                        "title": "Min-Max Optimization Made Simple: Approximating the Proximal Point Method via Contraction Maps",
                        "abstract": "In this paper we present a first-order method that admits near-optimal convergence rates for convex/concave min-max problems while requiring a simple and intuitive analysis. Similarly to the seminal work of Nemirovski and the recent approach of Piliouras et al. in normal form games, our work is based on the fact that the update rule of the Proximal Point method (PP) can be approximated up to accuracy $\\epsilon$ with only $O(\\log 1/\\epsilon)$ additional gradient-calls through the iterations of a contraction map. Then combining the analysis of (PP) method with an error-propagation analysis we establish that the resulting first order method, called Clairvoyant Extra Gradient, admits near-optimal time-average convergence for general domains and last-iterate convergence in the unconstrained case."
                    },
                    {
                        "title": "Heterogeneous Beliefs and Multi-Population Learning in Network Games",
                        "abstract": "The e\ufb00ect of population heterogeneity in multi-agent learning is practically relevant but remains far from being well-understood. Motivated by this, we introduce a model of multi-population learning that allows for heterogeneous beliefs within each population and where agents respond to their beliefs via smooth \ufb01ctitious play (SFP). We show that the system state \u2014 a probability distribution over beliefs \u2014 evolves according to a system of partial di\ufb00erential equations. We establish the convergence of SFP to Quantal Response Equilibria in di\ufb00erent classes of games capturing both network competition as well as network coordination. We also prove that the beliefs will eventually homogenize in all network games. Although the initial belief heterogeneity disappears in the limit, we show that it plays a crucial role for equilibrium selection in the case of coordination games as it helps select highly desirable equilibria. Contrary, in the case of network competition, the resulting limit behavior is independent of the initialization of beliefs, even when the underlying game has many distinct Nash equilibria."
                    },
                    {
                        "title": "Data-Scarce Identification of Game Dynamics via Sum-of-Squares Optimization",
                        "abstract": "Understanding how players adjust their strategies in games, based on their experience, is a crucial tool for policymakers. It enables them to forecast the system's eventual behavior, exert control over the system, and evaluate counterfactual scenarios. The task becomes increasingly difficult when only a limited number of observations are available or difficult to acquire. In this work, we introduce the Side-Information Assisted Regression (SIAR) framework, designed to identify game dynamics in multiplayer normal-form games only using data from a short run of a single system trajectory. To enhance system recovery in the face of scarce data, we integrate side-information constraints into SIAR, which restrict the set of feasible solutions to those satisfying game-theoretic properties and common assumptions about strategic interactions. SIAR is solved using sum-of-squares (SOS) optimization, resulting in a hierarchy of approximations that provably converge to the true dynamics of the system. We showcase that the SIAR framework accurately predicts player behavior across a spectrum of normal-form games, widely-known families of game dynamics, and strong benchmarks, even if the unknown system is chaotic."
                    },
                    {
                        "title": "Beyond Strict Competition: Approximate Convergence of Multi Agent Q-Learning Dynamics",
                        "abstract": "The behaviour of multi-agent learning in competitive settings is often considered under the restrictive assumption of a zero-sum game. Only under this strict requirement is the behaviour of learning well understood; beyond this, learning dynamics can often display non-convergent behaviours which prevent fixed-point analysis. Nonetheless, many relevant competitive games do not satisfy the zero-sum assumption.  Motivated by this, we study a smooth variant of Q-Learning, a popular reinforcement learning dynamics which balances the agents' tendency to maximise their payoffs with their propensity to explore the state space. We examine this dynamic in games which are `close' to network zero-sum games and find that Q-Learning converges to a neighbourhood around a unique equilibrium. The size of the neighbourhood is determined by the `distance' to the zero-sum game, as well as the exploration rates of the agents. We complement these results by providing a method whereby, given an arbitrary network game, the `nearest' network zero-sum game can be found efficiently. Importantly, our theoretical guarantees are widely applicable in different game settings, regardless of whether the dynamics ultimately reach an equilibrium, or remain non convergent."
                    },
                    {
                        "title": "A Quadratic Speedup in Finding Nash Equilibria of Quantum Zero-Sum Games",
                        "abstract": "Recent developments in domains such as non-local games, quantum interactive proofs, and quantum generative adversarial networks have renewed interest in quantum game theory and, specifically, quantum zero-sum games. Central to classical game theory is the efficient algorithmic computation of Nash equilibria, which represent optimal strategies for both players. In 2008, Jain and Watrous proposed the first classical algorithm for computing equilibria in quantum zero-sum games using the Matrix Multiplicative Weight Updates (MMWU) method to achieve a convergence rate of $\\mathcal{O}(d/\\epsilon^2)$ iterations to $\\epsilon$-Nash equilibria in the $4^d$-dimensional spectraplex. In this work, we propose a hierarchy of quantum optimization algorithms that generalize MMWU via an extra-gradient mechanism. Notably, within this proposed hierarchy, we introduce the Optimistic Matrix Multiplicative Weights Update (OMMWU) algorithm and establish its average-iterate convergence complexity as $\\mathcal{O}(d/\\epsilon)$ iterations to $\\epsilon$-Nash equilibria. This quadratic speed-up relative to Jain and Watrous' original algorithm sets a new benchmark for computing $\\epsilon$-Nash equilibria in quantum zero-sum games."
                    },
                    {
                        "title": "Stability of Multi-Agent Learning: Convergence in Network Games with Many Players",
                        "abstract": "The behaviour of multi-agent learning in many player games has been shown to display complex dynamics outside of restrictive examples such as network zero-sum games. In addition, it has been shown that convergent behaviour is less likely to occur as the number of players increase. To make progress in resolving this problem, we study Q-Learning dynamics and determine a sufficient condition for the dynamics to converge to a unique equilibrium in any network game. We find that this condition depends on the nature of pairwise interactions and on the network structure, but is explicitly independent of the total number of agents in the game. We evaluate this result on a number of representative network games and show that, under suitable network conditions, stable learning dynamics can be achieved with an arbitrary number of agents."
                    },
                    {
                        "title": "Asymptotic Convergence and Performance of Multi-Agent Q-Learning Dynamics",
                        "abstract": "Achieving convergence of multiple learning agents in general $N$-player games is imperative for the development of safe and reliable machine learning (ML) algorithms and their application to autonomous systems. Yet it is known that, outside the bounds of simple two-player games, convergence cannot be taken for granted. To make progress in resolving this problem, we study the dynamics of smooth Q-Learning, a popular reinforcement learning algorithm which quantifies the tendency for learning agents to explore their state space or exploit their payoffs. We show a sufficient condition on the rate of exploration such that the Q-Learning dynamics is guaranteed to converge to a unique equilibrium in any game. We connect this result to games for which Q-Learning is known to converge with arbitrary exploration rates, including weighted Potential games and weighted zero sum polymatrix games. Finally, we examine the performance of the Q-Learning dynamic as measured by the Time Averaged Social Welfare, and comparing this with the Social Welfare achieved by the equilibrium. We provide a sufficient condition whereby the Q-Learning dynamic will outperform the equilibrium even if the dynamics do not converge."
                    },
                    {
                        "title": "Multiplicative Updates for Online Convex Optimization over Symmetric Cones",
                        "abstract": "We study online convex optimization where the possible actions are trace-one elements in a symmetric cone, generalizing the extensively-studied experts setup and its quantum counterpart. Symmetric cones provide a unifying framework for some of the most important optimization models, including linear, second-order cone, and semidefinite optimization. Using tools from the field of Euclidean Jordan Algebras, we introduce the Symmetric-Cone Multiplicative Weights Update (SCMWU), a projection-free algorithm for online optimization over the trace-one slice of an arbitrary symmetric cone. We show that SCMWU is equivalent to Follow-the-Regularized-Leader and Online Mirror Descent with symmetric-cone negative entropy as regularizer. Using this structural result we show that SCMWU is a no-regret algorithm, and verify our theoretical results with extensive experiments. Our results unify and generalize the analysis for the Multiplicative Weights Update method over the probability simplex and the Matrix Multiplicative Weights Update method over the set of density matrices."
                    },
                    {
                        "title": "Auto-Aligning Multiagent Incentives with Global Objectives",
                        "abstract": "The general ability to achieve a singular task with a set of decentralized, intelligent agents is an important goal in multiagent research. The complex interaction between individual agents\u2019 incentives makes designing their objectives such that the resulting multia-gentsystemalignswithadesiredglobalgoalparticularlychallenging. Inthiswork,insteadofconsideringtheproblemofdesigningsuit-able incentives from scratch, we assume a multiagent system with given preset incentives and consider automatically modifying these incentives online to achieve a new goal. This reduces the search space over possible individual incentives and takes advantage of the effort instilled by the previous system designer. We demonstrate the promise as well as the limitations of re-purposing multiagent systems in this way, both theoretically and empirically, on a variety of domains. Surprisingly, we show that training a diverse multiagent system to align with a modified global objective ( \ud835\udc54 \u2192 \ud835\udc54 \u2032 ) can, in at least one case, lead to better generalization performance in unseen test scenarios, when evaluated on the original objective ( \ud835\udc54 )."
                    },
                    {
                        "title": "Exploiting hidden structures in non-convex games for convergence to Nash equilibrium",
                        "abstract": "A wide array of modern machine learning applications - from adversarial models to multi-agent reinforcement learning - can be formulated as non-cooperative games whose Nash equilibria represent the system's desired operational states. Despite having a highly non-convex loss landscape, many cases of interest possess a latent convex structure that could potentially be leveraged to yield convergence to equilibrium. Driven by this observation, our paper proposes a flexible first-order method that successfully exploits such\"hidden structures\"and achieves convergence under minimal assumptions for the transformation connecting the players' control variables to the game's latent, convex-structured layer. The proposed method - which we call preconditioned hidden gradient descent (PHGD) - hinges on a judiciously chosen gradient preconditioning scheme related to natural gradient methods. Importantly, we make no separability assumptions for the game's hidden structure, and we provide explicit convergence rate guarantees for both deterministic and stochastic environments."
                    },
                    {
                        "title": "No-Regret Learning and Equilibrium Computation in Quantum Games",
                        "abstract": "As quantum processors advance, the emergence of large-scale decentralized systems involving interacting quantum-enabled agents is on the horizon. Recent research efforts have explored quantum versions of Nash and correlated equilibria as solution concepts of strategic quantum interactions, but these approaches did not directly connect to decentralized adaptive setups where agents possess limited information. This paper delves into the dynamics of quantum-enabled agents within decentralized systems that employ no-regret algorithms to update their behaviors over time. Specifically, we investigate two-player quantum zero-sum games and polymatrix quantum zero-sum games, showing that no-regret algorithms converge to separable quantum Nash equilibria in time-average. In the case of general multi-player quantum games, our work leads to a novel solution concept, (separable) quantum coarse correlated equilibria (QCCE), as the convergent outcome of the time-averaged behavior no-regret algorithms, offering a natural solution concept for decentralized quantum systems. Finally, we show that computing QCCEs can be formulated as a semidefinite program and establish the existence of entangled (i.e., non-separable) QCCEs, which cannot be approached via the current paradigm of no-regret learning."
                    }
                ]
            },
            "aa10606d-0cec-43d2-b24c-6e69827828e6": {
                "pk": "aa10606d-0cec-43d2-b24c-6e69827828e6",
                "name": "Romuald Elie",
                "collaborators": [
                    "Sarah Perrin",
                    "K. Tuyls",
                    "M. Lauri\u00e8re",
                    "G. Piliouras",
                    "M. Geist",
                    "O. Pietquin",
                    "Daniel Hennes",
                    "Paul Muller",
                    "J. P\u00e9rolat",
                    "Jerome T. Connor",
                    "Mark Rowland",
                    "Marc Lanctot",
                    "Sertan Girgin",
                    "Shayegan Omidshafiei",
                    "Julien P'erolat",
                    "I. Gemp",
                    "Luke Marris",
                    "Kalesha Bullard",
                    "Yoram Bachrach",
                    "Christophe Denis",
                    "Mohamed Hebiri",
                    "Olivier Bachem",
                    "A. Tacchetti",
                    "Thomas W. Anthony",
                    "Mina Khan",
                    "Florian Strub",
                    "Eugene Tarassov",
                    "Zhe Wang",
                    "R. Munos",
                    "Franccois Hu",
                    "Yiran Mao",
                    "Edgar A. Du'enez-Guzm'an",
                    "Andras Gyorgy",
                    "Georgios Piliouras",
                    "Michael Kaisers",
                    "Kate Larson",
                    "Fran\u00b8cois Hu",
                    "Geoffrey Cideron",
                    "Andrea Agostinelli",
                    "Johan Ferret",
                    "Alexandre Ram'e",
                    "Raphael Koster",
                    "Miruna Pislar",
                    "Jan Balaguer",
                    "Leqi Liu",
                    "Oliver P. Hauser",
                    "Matthew M. Botvinick",
                    "Christopher Summerfield",
                    "Roxana Dumitrescu",
                    "W. Sabbagh",
                    "Chao Zhou",
                    "Avishkar Bhoopchand",
                    "Vibhavari Dasagi",
                    "Bart De Vylder",
                    "Edgar A. Du\u00e9\u00f1ez-Guzm\u00e1n",
                    "Richard Everett",
                    "Edward Hughes",
                    "K. Larson",
                    "Guy Lever",
                    "Siqi Liu",
                    "Kevin R. McKee",
                    "Ayush Jain",
                    "Th\u00e9ophile Cabannes",
                    "B. D. Vylder",
                    "V. D. Boer",
                    "Neil Burch",
                    "S. McAleer",
                    "Sarah H. Cen",
                    "A. Gruslys",
                    "Aleksandra Malysheva",
                    "Sherjil Ozair",
                    "Finbarr Timbers",
                    "Tobias Pohlen",
                    "Tom Eccles",
                    "Jean-Baptiste Lespiau",
                    "Bilal Piot",
                    "Edward Lockhart",
                    "L. Sifre",
                    "Nathalie Beauguerlange",
                    "David Silver",
                    "Satinder Singh",
                    "D. Hassabis",
                    "Caroline Hillairet",
                    "Marc Juillard"
                ],
                "domain": [
                    "Game Theory",
                    "Reinforcement Learning",
                    "Algorithmic Fairness",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Approximating the Core via Iterative Coalition Sampling",
                        "abstract": "The core is a central solution concept in cooperative game theory, defined as the set of feasible allocations or payments such that no subset of agents has incentive to break away and form their own subgroup or coalition. However, it has long been known that the core (and approximations, such as the least-core) are hard to compute. This limits our ability to analyze cooperative games in general, and to fully embrace cooperative game theory contributions in domains such as explainable AI (XAI), where the core can complement the Shapley values to identify influential features or instances supporting predictions by black-box models. We propose novel iterative algorithms for computing variants of the core, which avoid the computational bottleneck of many other approaches; namely solving large linear programs. As such, they scale better to very large problems as we demonstrate across different classes of cooperative games, including weighted voting games, induced subgraph games, and marginal contribution networks. We also explore our algorithms in the context of XAI, providing further evidence of the power of the core for such applications."
                    },
                    {
                        "title": "Fairness guarantees in multi-class classification with demographic parity",
                        "abstract": "Algorithmic Fairness is an established area of machine learning, willing to reduce the in\ufb02uence of hidden bias in the data. Yet, despite its wide range of applications, very few works consider the multi-class classi\ufb01cation setting from the fairness perspective. We focus on this question and extend the de\ufb01nition of approximate fairness in the case of Demographic Parity to multi-class classi\ufb01cation. We specify the corresponding expressions of the optimal fair classi\ufb01ers in the attribute-aware case and both for binary and multi-categorical sensitive attributes. This suggests a plug-in data-driven procedure, for which we establish theoretical guarantees. The enhanced estimator is proved to mimic the behavior of the optimal rule both in terms of fairness and risk. Notably, fairness guarantees are distribution-free. The approach is evaluated on both synthetic and real datasets and reveals very e\ufb00ective in decision making with a preset level of unfairness. In addition, our method is competitive (if not better) with the state-of-the-art in binary and multi-class tasks."
                    },
                    {
                        "title": "Diversity-Rewarded CFG Distillation",
                        "abstract": "Generative models are transforming creative domains such as music generation, with inference-time strategies like Classifier-Free Guidance (CFG) playing a crucial role. However, CFG doubles inference cost while limiting originality and diversity across generated contents. In this paper, we introduce diversity-rewarded CFG distillation, a novel finetuning procedure that distills the strengths of CFG while addressing its limitations. Our approach optimises two training objectives: (1) a distillation objective, encouraging the model alone (without CFG) to imitate the CFG-augmented predictions, and (2) an RL objective with a diversity reward, promoting the generation of diverse outputs for a given prompt. By finetuning, we learn model weights with the ability to generate high-quality and diverse outputs, without any inference overhead. This also unlocks the potential of weight-based model merging strategies: by interpolating between the weights of two models (the first focusing on quality, the second on diversity), we can control the quality-diversity trade-off at deployment time, and even further boost performance. We conduct extensive experiments on the MusicLM (Agostinelli et al., 2023) text-to-music generative model, where our approach surpasses CFG in terms of quality-diversity Pareto optimality. According to human evaluators, our finetuned-then-merged model generates samples with higher quality-diversity than the base model augmented with CFG. Explore our generations at https://google-research.github.io/seanet/musiclm/diverse_music/."
                    },
                    {
                        "title": "Using deep reinforcement learning to promote sustainable human behaviour on a common pool resource problem",
                        "abstract": "A canonical social dilemma arises when finite resources are allocated to a group of people, who can choose to either reciprocate with interest, or keep the proceeds for themselves. What resource allocation mechanisms will encourage levels of reciprocation that sustain the commons? Here, in an iterated multiplayer trust game, we use deep reinforcement learning (RL) to design an allocation mechanism that endogenously promotes sustainable contributions from human participants to a common pool resource. We first trained neural networks to behave like human players, creating a stimulated economy that allowed us to study how different mechanisms influenced the dynamics of receipt and reciprocation. We then used RL to train a social planner to maximise aggregate return to players. The social planner discovered a redistributive policy that led to a large surplus and an inclusive economy, in which players made roughly equal gains. The RL agent increased human surplus over baseline mechanisms based on unrestricted welfare or conditional cooperation, by conditioning its generosity on available resources and temporarily sanctioning defectors by allocating fewer resources to them. Examining the AI policy allowed us to develop an explainable mechanism that performed similarly and was more popular among players. Deep reinforcement learning can be used to discover mechanisms that promote sustainable human behaviour."
                    },
                    {
                        "title": "A new Mertens decomposition of Y g, \u03be -submartingale systems. Application to BSDEs with weak constraints at stopping times",
                        "abstract": "new Mertens decomposition of"
                    },
                    {
                        "title": "Developing, Evaluating and Scaling Learning Agents in Multi-Agent Environments",
                        "abstract": "The Game Theory & Multi-Agent team at DeepMind studies several aspects of multi-agent learning ranging from computing approximations to fundamental concepts in game theory to simulating social dilemmas in rich spatial environments and training 3-d humanoids in difficult team coordination tasks. A signature aim of our group is to use the resources and expertise made available to us at DeepMind in deep reinforcement learning to explore multi-agent systems in complex environments and use these benchmarks to advance our understanding. Here, we summarise the recent work of our team and present a taxonomy that we feel highlights many important open challenges in multi-agent research."
                    },
                    {
                        "title": "Scaling Mean Field Games by Online Mirror Descent",
                        "abstract": "We address the scaling of equilibrium computation in Mean Field Games (MFGs) by using Online Mirror Descent (OMD). We show that continuous-time OMD provably converges to a Nash equilibrium under a natural and well-motivated set of monotonicity assumptions. A thorough experimental investigation on various single and multi-population MFGs shows that OMD outperforms traditional algorithms such as Fictitious Play. We empirically show that OMD scales and converges significantly faster than Fictitious Play by solving, for the first time to our knowledge, examples of MFGs with hundreds of billions states."
                    },
                    {
                        "title": "Learning Correlated Equilibria in Mean-Field Games",
                        "abstract": "The designs of many large-scale systems today, from traffic routing environments to smart grids, rely on game-theoretic equilibrium concepts. However, as the size of an $N$-player game typically grows exponentially with $N$, standard game theoretic analysis becomes effectively infeasible beyond a low number of players. Recent approaches have gone around this limitation by instead considering Mean-Field games, an approximation of anonymous $N$-player games, where the number of players is infinite and the population's state distribution, instead of every individual player's state, is the object of interest. The practical computability of Mean-Field Nash equilibria, the most studied Mean-Field equilibrium to date, however, typically depends on beneficial non-generic structural properties such as monotonicity or contraction properties, which are required for known algorithms to converge. In this work, we provide an alternative route for studying Mean-Field games, by developing the concepts of Mean-Field correlated and coarse-correlated equilibria. We show that they can be efficiently learnt in \\emph{all games}, without requiring any additional assumption on the structure of the game, using three classical algorithms. Furthermore, we establish correspondences between our notions and those already present in the literature, derive optimality bounds for the Mean-Field - $N$-player transition, and empirically demonstrate the convergence of these algorithms on simple games."
                    },
                    {
                        "title": "Evolutionary Dynamics and Phi-Regret Minimization in Games",
                        "abstract": "Regret has been established as a foundational concept in online learning, and likewise has important applications in the analysis of learning dynamics in games. Regret quantifies the difference between a learner\u2019s performance against a baseline in hindsight. It is well known that regret-minimizing algorithms converge to certain classes of equilibria in games; however, traditional forms of regret used in game theory predominantly consider baselines that permit deviations to deterministic actions or strategies. In this paper, we revisit our understanding of regret from the perspective of deviations over partitions of the full mixed strategy space (i.e., probability distributions over pure strategies), under the lens of the previously-established \u03a6-regret framework, which provides a continuum of stronger regret measures. Importantly, \u03a6-regret enables learning agents to consider deviations from and to mixed strategies, generalizing several existing notions of regret such as external, internal, and swap regret, and thus broadening the insights gained from regret-based analysis of learning algorithms. We prove here that the well-studied evolutionary learning algorithm of replicator dynamics (RD) seamlessly minimizes the strongest possible form of \u03a6-regret in generic 2 \u00d7 2 games, without any modification of the underlying algorithm itself. We subsequently conduct experiments validating our theoretical results in a suite of 144 2 \u00d7 2 games wherein RD exhibits a diverse set of behaviors. We conclude by providing empirical evidence of \u03a6-regret minimization by RD in some larger games, hinting at further opportunity for \u03a6-regret based study of such algorithms from both a theoretical and empirical perspective."
                    },
                    {
                        "title": "Scalable Deep Reinforcement Learning Algorithms for Mean Field Games",
                        "abstract": "Mean Field Games (MFGs) have been introduced to efficiently approximate games with very large populations of strategic agents. Recently, the question of learning equilibria in MFGs has gained momentum, particularly using model-free reinforcement learning (RL) methods. One limiting factor to further scale up using RL is that existing algorithms to solve MFGs require the mixing of approximated quantities such as strategies or $q$-values. This is far from being trivial in the case of non-linear function approximation that enjoy good generalization properties, e.g. neural networks. We propose two methods to address this shortcoming. The first one learns a mixed strategy from distillation of historical data into a neural network and is applied to the Fictitious Play algorithm. The second one is an online mixing method based on regularization that does not require memorizing historical data or previous estimates. It is used to extend Online Mirror Descent. We demonstrate numerically that these methods efficiently enable the use of Deep RL algorithms to solve various MFGs. In addition, we show that these methods outperform SotA baselines from the literature."
                    },
                    {
                        "title": "Learning in Mean Field Games: A Survey",
                        "abstract": "Non-cooperative and cooperative games with a very large number of players have many applications but remain generally intractable when the number of players increases. Introduced by Lasry and Lions, and Huang, Caines and Malham\\'e, Mean Field Games (MFGs) rely on a mean-field approximation to allow the number of players to grow to infinity. Traditional methods for solving these games generally rely on solving partial or stochastic differential equations with a full knowledge of the model. Recently, Reinforcement Learning (RL) has appeared promising to solve complex problems at scale. The combination of RL and MFGs is promising to solve games at a very large scale both in terms of population size and environment complexity. In this survey, we review the quickly growing recent literature on RL methods to learn equilibria and social optima in MFGs. We first identify the most common settings (static, stationary, and evolutive) of MFGs. We then present a general framework for classical iterative methods (based on best-response computation or policy evaluation) to solve MFGs in an exact way. Building on these algorithms and the connection with Markov Decision Processes, we explain how RL can be used to learn MFG solutions in a model-free way. Last, we present numerical illustrations on a benchmark problem, and conclude with some perspectives."
                    },
                    {
                        "title": "Mastering the game of Stratego with model-free multiagent reinforcement learning",
                        "abstract": "We introduce DeepNash, an autonomous agent that plays the imperfect information game Stratego at a human expert level. Stratego is one of the few iconic board games that artificial intelligence (AI) has not yet mastered. It is a game characterized by a twin challenge: It requires long-term strategic thinking as in chess, but it also requires dealing with imperfect information as in poker. The technique underpinning DeepNash uses a game-theoretic, model-free deep reinforcement learning method, without search, that learns to master Stratego through self-play from scratch. DeepNash beat existing state-of-the-art AI methods in Stratego and achieved a year-to-date (2022) and all-time top-three ranking on the Gravon games platform, competing with human expert players. Description Machine learning to play Stratego Stratego is a popular two-player imperfect information board game. Because of its complexity stemming from its enormous game tree, decision-making under imperfect information, and a piece deployment phase at the start, Stratego poses a challenge for artificial intelligence (AI). Previous computer programs only performed at an amateur level at best. Perolat et al. introduce a model-free multiagent reinforcement learning methodology and show that it can achieve human expert\u2013level performance in Stratego. The present work not only adds to the growing list of games that AI systems can play as well or even better than humans but may also facilitate further applications of reinforcement learning methods in real-world, large-scale multiagent problems that are characterized by imperfect information and thus are currently unsolvable. \u2014YS Reinforcement learning achieves human expert\u2013level performance in the large-scale imperfect information board game Stratego."
                    },
                    {
                        "title": "Fairness guarantee in multi-class classification",
                        "abstract": "Algorithmic Fairness is an established area of machine learning, willing to reduce the influence of hidden bias in the data. Yet, despite its wide range of applications, very few works consider the multi-class classification setting from the fairness perspective. We focus on this question and extend the definition of approximate fairness in the case of Demographic Parity to multi-class classification. We specify the corresponding expressions of the optimal fair classifiers. This suggests a plug-in data-driven procedure, for which we establish theoretical guarantees. The enhanced estimator is proved to mimic the behavior of the optimal rule both in terms of fairness and risk. Notably, fairness guarantees are distribution-free. The approach is evaluated on both synthetic and real datasets and reveals very effective in decision making with a preset level of unfairness. In addition, our method is competitive (if not better) with the state-of-the-art in binary and multi-class tasks."
                    },
                    {
                        "title": "Scaling up Mean Field Games with Online Mirror Descent",
                        "abstract": "We address scaling up equilibrium computation in Mean Field Games (MFGs) using Online Mirror Descent (OMD). We show that continuous-time OMD provably converges to a Nash equilibrium under a natural and well-motivated set of monotonicity assumptions. This theoretical result nicely extends to multi-population games and to settings involving common noise. A thorough experimental investigation on various single and multi-population MFGs shows that OMD outperforms traditional algorithms such as Fictitious Play (FP). We empirically show that OMD scales up and converges significantly faster than FP by solving, for the first time to our knowledge, examples of MFGs with hundreds of billions states. This study establishes the state-of-the-art for learning in large-scale multi-agent and multi-population games."
                    },
                    {
                        "title": "Concave Utility Reinforcement Learning: the Mean-field Game viewpoint",
                        "abstract": "Concave Utility Reinforcement Learning (CURL) extends RL from linear to concave utilities in the occupancy measure induced by the agent's policy. This encompasses not only RL but also imitation learning and exploration, among others. Yet, this more general paradigm invalidates the classical Bellman equations, and calls for new algorithms. Mean-field Games (MFGs) are a continuous approximation of many-agent RL. They consider the limit case of a continuous distribution of identical agents, anonymous with symmetric interests, and reduce the problem to the study of a single representative agent in interaction with the full population. Our core contribution consists in showing that CURL is a subclass of MFGs. We think this important to bridge together both communities. It also allows to shed light on aspects of both fields: we show the equivalence between concavity in CURL and monotonicity in the associated MFG, between optimality conditions in CURL and Nash equilibrium in MFG, or that Fictitious Play (FP) for this class of MFGs is simply Frank-Wolfe, bringing the first convergence rate for discrete-time FP for MFGs. We also experimentally demonstrate that, using algorithms recently introduced for solving MFGs, we can address the CURL problem more efficiently."
                    },
                    {
                        "title": "Evolutionary Dynamics and \u03a6-Regret Minimization in Games",
                        "abstract": "Regret has been established as a foundational concept in online learning, and likewise has important applications in the analysis of learning dynamics in games. Regret quantifies the difference between a learner's performance against a baseline in hindsight. It is well-known that regret-minimizing algorithms converge to certain classes of equilibria in games; however, traditional forms of regret used in game theory predominantly consider baselines that permit deviations to deterministic actions or strategies. In this paper, we revisit our understanding of regret from the perspective of deviations over partitions of the full \\emph{mixed} strategy space (i.e., probability distributions over pure strategies), under the lens of the previously-established $\\Phi$-regret framework, which provides a continuum of stronger regret measures. Importantly, $\\Phi$-regret enables learning agents to consider deviations from and to mixed strategies, generalizing several existing notions of regret such as external, internal, and swap regret, and thus broadening the insights gained from regret-based analysis of learning algorithms. We prove here that the well-studied evolutionary learning algorithm of replicator dynamics (RD) seamlessly minimizes the strongest possible form of $\\Phi$-regret in generic $2 \\times 2$ games, without any modification of the underlying algorithm itself. We subsequently conduct experiments validating our theoretical results in a suite of 144 $2 \\times 2$ games wherein RD exhibits a diverse set of behaviors. We conclude by providing empirical evidence of $\\Phi$-regret minimization by RD in some larger games, hinting at further opportunity for $\\Phi$-regret based study of such algorithms from both a theoretical and empirical perspective."
                    },
                    {
                        "title": "Fair Active Learning: Solving the Labeling Problem in Insurance",
                        "abstract": "This paper addresses significant obstacles that arise from the widespread use of machine learning models in the insurance industry, with a specific focus on promoting fairness. The initial challenge lies in effectively leveraging unlabeled data in insurance while reducing the labeling effort and emphasizing data relevance through active learning techniques. The paper explores various active learning sampling methodologies and evaluates their impact on both synthetic and real insurance datasets. This analysis highlights the difficulty of achieving fair model inferences, as machine learning models may replicate biases and discrimination found in the underlying data. To tackle these interconnected challenges, the paper introduces an innovative fair active learning method. The proposed approach samples informative and fair instances, achieving a good balance between model predictive performance and fairness, as confirmed by numerical experiments on insurance datasets."
                    }
                ]
            },
            "2517cbad-0560-441f-911d-1ef62c8ff054": {
                "pk": "2517cbad-0560-441f-911d-1ef62c8ff054",
                "name": "Guy Lever",
                "collaborators": [
                    "T. Graepel",
                    "N. Heess",
                    "Siqi Liu",
                    "K. Tuyls",
                    "Marc Lanctot",
                    "Luke Marris",
                    "Wojciech M. Czarnecki",
                    "Joel Z. Leibo",
                    "S. Tunyasuvunakool",
                    "Yoram Bachrach",
                    "Daniel Hennes",
                    "Edward Hughes",
                    "Paul Muller",
                    "Zhe Wang",
                    "J. Merel",
                    "P. Sunehag",
                    "Max Jaderberg",
                    "Nicolas Sonnerat",
                    "Tuomas Haarnoja",
                    "Markus Wulfmeier",
                    "Noah Siegel",
                    "Leonard Hasenclever",
                    "Yuval Tassa",
                    "J. P\u00e9rolat",
                    "Shayegan Omidshafiei",
                    "Tom Eccles",
                    "A. Gruslys",
                    "V. Zambaldi",
                    "David Silver",
                    "Ronnie Stafford",
                    "D. Barber",
                    "T. Furmston",
                    "Ben Moran",
                    "Sandy H. Huang",
                    "Dhruva Tirumala",
                    "Jan Humplik",
                    "Roland Hafner",
                    "Michael Bloesch",
                    "Kristian Hartikainen",
                    "Arunkumar Byravan",
                    "Fereshteh Sadeghi",
                    "Nathan Batchelor",
                    "Federico Casarini",
                    "Stefano Saliceti",
                    "Charles Game",
                    "Neil Sreendra",
                    "Kushal Patel",
                    "Marlon Gwira",
                    "Andrea Huber",
                    "N. Hurley",
                    "F. Nori",
                    "R. Hadsell",
                    "I. Gemp",
                    "Thomas W. Anthony",
                    "Avishkar Bhoopchand",
                    "Kalesha Bullard",
                    "Jerome T. Connor",
                    "Vibhavari Dasagi",
                    "Bart De Vylder",
                    "Edgar A. Du\u00e9\u00f1ez-Guzm\u00e1n",
                    "R. \u00c9lie",
                    "Richard Everett",
                    "Mina Khan",
                    "K. Larson",
                    "Kevin R. McKee",
                    "Florian Strub",
                    "A. Tacchetti",
                    "Eugene Tarassov",
                    "S. Eslami",
                    "A. Abdolmaleki",
                    "H. F. Song",
                    "Brendan D. Tracey",
                    "Mark Rowland",
                    "R. Munos",
                    "Dylan Banarse",
                    "Chrisantha Fernando",
                    "Pushmeet Kohli",
                    "Angeliki Lazaridou",
                    "Iain Dunning",
                    "Antonio Garc\u00eda Casta\u00f1eda",
                    "Charlie Beattie",
                    "Neil C. Rabinowitz",
                    "Ari S. Morcos",
                    "Avraham Ruderman",
                    "Tim Green",
                    "Louise Deason",
                    "D. Hassabis",
                    "K. Kavukcuoglu",
                    "J. Shawe-Taylor",
                    "Csaba Szepesvari",
                    "Aleksandar Botev",
                    "T. Degris",
                    "Daan Wierstra",
                    "Martin A. Riedmiller"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Multi-Agent Systems",
                    "Robotics",
                    "Game Theory"
                ],
                "publications": [
                    {
                        "title": "Learning agile soccer skills for a bipedal robot with deep reinforcement learning",
                        "abstract": "We investigated whether deep reinforcement learning (deep RL) is able to synthesize sophisticated and safe movement skills for a low-cost, miniature humanoid robot that can be composed into complex behavioral strategies. We used deep RL to train a humanoid robot to play a simplified one-versus-one soccer game. The resulting agent exhibits robust and dynamic movement skills, such as rapid fall recovery, walking, turning, and kicking, and it transitions between them in a smooth and efficient manner. It also learned to anticipate ball movements and block opponent shots. The agent\u2019s tactical behavior adapts to specific game contexts in a way that would be impractical to manually design. Our agent was trained in simulation and transferred to real robots zero-shot. A combination of sufficiently high-frequency control, targeted dynamics randomization, and perturbations during training enabled good-quality transfer. In experiments, the agent walked 181% faster, turned 302% faster, took 63% less time to get up, and kicked a ball 34% faster than a scripted baseline. OP3 humanoid robots learned to play agile soccer using deep reinforcement learning. Editor\u2019s summary Generating robust motor skills in bipedal robots in the real world is challenging because of the inability of current control methods to generalize to specific tasks. Haarnoja et al. developed a deep reinforcement learning\u2013based framework for full-body control of humanoid robots, enabling a game of one-versus-one soccer. The robots exhibited emergent behaviors in the form of dynamic motor skills such as the ability to recover from falls and also tactics like defending the ball against an opponent. The robot movements were faster when using their framework than a scripted baseline controller and may have potential for more complex multirobot interactions. \u2014Amos Matsiko"
                    },
                    {
                        "title": "Developing, Evaluating and Scaling Learning Agents in Multi-Agent Environments",
                        "abstract": "The Game Theory & Multi-Agent team at DeepMind studies several aspects of multi-agent learning ranging from computing approximations to fundamental concepts in game theory to simulating social dilemmas in rich spatial environments and training 3-d humanoids in difficult team coordination tasks. A signature aim of our group is to use the resources and expertise made available to us at DeepMind in deep reinforcement learning to explore multi-agent systems in complex environments and use these benchmarks to advance our understanding. Here, we summarise the recent work of our team and present a taxonomy that we feel highlights many important open challenges in multi-agent research."
                    },
                    {
                        "title": "From Motor Control to Team Play in Simulated Humanoid Football",
                        "abstract": "Learning to combine control at the level of joint torques with longer-term goal-directed behavior is a long-standing challenge for physically embodied artificial agents. Intelligent behavior in the physical world unfolds across multiple spatial and temporal scales: Although movements are ultimately executed at the level of instantaneous muscle tensions or joint torques, they must be selected to serve goals that are defined on much longer time scales and that often involve complex interactions with the environment and other agents. Recent research has demonstrated the potential of learning-based approaches applied to the respective problems of complex movement, long-term planning, and multiagent coordination. However, their integration traditionally required the design and optimization of independent subsystems and remains challenging. In this work, we tackled the integration of motor control and long-horizon decision-making in the context of simulated humanoid football, which requires agile motor control and multiagent coordination. We optimized teams of agents to play simulated football via reinforcement learning, constraining the solution space to that of plausible movements learned using human motion capture data. They were trained to maximize several environment rewards and to imitate pretrained football-specific skills if doing so led to improved performance. The result is a team of coordinated humanoid football players that exhibit complex behavior at different scales, quantified by a range of analysis and statistics, including those used in real-world sport analytics. Our work constitutes a complete demonstration of learned integrated decision-making at multiple scales in a multiagent setting."
                    },
                    {
                        "title": "A Generalized Training Approach for Multiagent Learning",
                        "abstract": "This paper investigates a population-based training regime based on game-theoretic principles called Policy-Spaced Response Oracles (PSRO). PSRO is general in the sense that it (1) encompasses well-known algorithms such as fictitious play and double oracle as special cases, and (2) in principle applies to general-sum, many-player games. Despite this, prior studies of PSRO have been focused on two-player zero-sum games, a regime wherein Nash equilibria are tractably computable. In moving from two-player zero-sum games to more general settings, computation of Nash equilibria quickly becomes infeasible. Here, we extend the theoretical underpinnings of PSRO by considering an alternative solution concept, $\\alpha$-Rank, which is unique (thus faces no equilibrium selection issues, unlike Nash) and applies readily to general-sum, many-player settings. We establish convergence guarantees in several games classes, and identify links between Nash equilibria and $\\alpha$-Rank. We demonstrate the competitive performance of $\\alpha$-Rank-based PSRO against an exact Nash solver-based PSRO in 2-player Kuhn and Leduc Poker. We then go beyond the reach of prior PSRO applications by considering 3- to 5-player poker games, yielding instances where $\\alpha$-Rank achieves faster convergence than approximate Nash solvers, thus establishing it as a favorable general games solver. We also carry out an initial empirical validation in MuJoCo soccer, illustrating the feasibility of the proposed approach in another complex domain."
                    },
                    {
                        "title": "The Body is Not a Given: Joint Agent Policy Learning and Morphology Evolution",
                        "abstract": "Reinforcement learning (RL) has proven to be a powerful paradigm for deriving complex behaviors from simple reward signals in a wide range of environments. When applying RL to continuous control agents in simulated physics environments, the body is usually considered to be part of the environment. However, during evolution the physical body of biological organisms and their controlling brains are co-evolved, thus exploring a much larger space of actuator/controller configurations. Put differently, the intelligence does not reside only in the agent's mind, but also in the design of their body. We propose a method for uncovering strong agents, consisting of a good combination of a body and policy, based on combining RL with an evolutionary procedure. Given the resulting agent, we also propose an approach for identifying the body changes that contributed the most to the agent performance. We use the Shapley value from cooperative game theory to find the fair contribution of individual components, taking into account synergies between components. We evaluate our methods in an environment similar to the the recently proposed Robo-Sumo task, where agents in a software physics simulator compete in tipping over their opponent or pushing them out of the arena. Our results show that the proposed methods are indeed capable of generating strong agents, significantly outperforming baselines that focus on optimizing the agent policy alone. A video is available at: https://youtu.be/CHlecRim9PI"
                    },
                    {
                        "title": "Biases for Emergent Communication in Multi-agent Reinforcement Learning",
                        "abstract": "We study the problem of emergent communication, in which language arises because speakers and listeners must communicate information in order to solve tasks. In temporally extended reinforcement learning domains, it has proved hard to learn such communication without centralized training of agents, due in part to a difficult joint exploration problem. We introduce inductive biases for positive signalling and positive listening, which ease this problem. In a simple one-step environment, we demonstrate how these biases ease the learning problem. We also apply our methods to a more extended environment, showing that agents with these inductive biases achieve better performance, and analyse the resulting communications protocols."
                    },
                    {
                        "title": "Emergent Coordination Through Competition",
                        "abstract": "We study the emergence of cooperative behaviors in reinforcement learning agents by introducing a challenging competitive multi-agent soccer environment with continuous simulated physics. We demonstrate that decentralized, population-based training with co-play can lead to a progression in agents' behaviors: from random, to simple ball chasing, and finally showing evidence of cooperation. Our study highlights several of the challenges encountered in large scale multi-agent training in continuous control. In particular, we demonstrate that the automatic optimization of simple shaping rewards, not themselves conducive to co-operative behavior, can lead to long-horizon team behavior. We further apply an evaluation scheme, grounded by game theoretic principals, that can assess agent performance in the absence of pre-defined evaluation tasks or human baselines."
                    },
                    {
                        "title": "Reinforcement Learning Agents acquire Flocking and Symbiotic Behaviour in Simulated Ecosystems",
                        "abstract": "In nature, group behaviours such as flocking as well as cross-species symbiotic partnerships are observed in vastly different forms and circumstances. We hypothesize that such strategies can arise in response to generic predator-prey pressures in a spatial environment with range-limited sensation and action. We evaluate whether these forms of coordination can emerge by independent multi-agent reinforcement learning in simple multiple-species ecosystems. In contrast to prior work, we avoid hand-crafted shaping rewards, specific actions, or dynamics that would directly encourage coordination across agents. Instead we test whether coordination emerges as a consequence of adaptation without encouraging these specific forms of coordination, which only has indirect benefit. Our simulated ecosystems consist of a generic food chain involving three trophic levels: apex predator, mid-level predator, and prey. We conduct experiments on two different platforms, a 3D physics engine with tens of agents as well as in a 2D grid world with up to thousands. The results clearly confirm our hypothesis and show substantial coordination both within and across species. To obtain these results, we leverage and adapt recent advances in deep reinforcement learning within an ecosystem training protocol featuring homogeneous groups of independent agents from different species (sets of policies), acting in many different random combinations in parallel habitats. The policies utilize neural network architectures that are invariant to agent individuality but not type (species) and that generalize across varying numbers of observed other agents. While the emergence of complexity in artificial ecosystems have long been studied in the artificial life community, the focus has been more on individual complexity and genetic algorithms or explicit modelling, and less on group complexity and reinforcement learning emphasized in this article. Unlike what the name and intuition suggests, reinforcement learning adapts over evolutionary history rather than a life-time and is here addressing the sequential optimization of fitness that is usually approached by genetic algorithms in the artificial life community. We utilize a shift from procedures to objectives, allowing us to bring new powerful machinery to bare, and we see emergence of complex behaviour from a sequence of simple optimization problems."
                    },
                    {
                        "title": "Value-Decomposition Networks For Cooperative Multi-Agent Learning Based On Team Reward",
                        "abstract": "We study the problem of cooperative multi-agent reinforcement learning with a single joint reward signal. This class of learning problems is difficult because of the often large combined action and observation spaces. In the fully centralized and decentralized approaches, we find the problem of spurious rewards and a phenomenon we call the \"lazy agent'' problem, which arises due to partial observability. We address these problems by training individual agents with a novel value-decomposition network architecture, which learns to decompose the team value function into agent-wise value functions."
                    },
                    {
                        "title": "Human-level performance in 3D multiplayer games with population-based reinforcement learning",
                        "abstract": "Artificial teamwork Artificially intelligent agents are getting better and better at two-player games, but most real-world endeavors require teamwork. Jaderberg et al. designed a computer program that excels at playing the video game Quake III Arena in Capture the Flag mode, where two multiplayer teams compete in capturing the flags of the opposing team. The agents were trained by playing thousands of games, gradually learning successful strategies not unlike those favored by their human counterparts. Computer agents competed successfully against humans even when their reaction times were slowed to match those of humans. Science, this issue p. 859 Teams of artificial agents compete successfully against humans in the video game Quake III Arena in Capture the Flag mode. Reinforcement learning (RL) has shown great success in increasingly complex single-agent environments and two-player turn-based games. However, the real world contains multiple agents, each learning and acting independently to cooperate and compete with other agents. We used a tournament-style evaluation to demonstrate that an agent can achieve human-level performance in a three-dimensional multiplayer first-person video game, Quake III Arena in Capture the Flag mode, using only pixels and game points scored as input. We used a two-tier optimization process in which a population of independent RL agents are trained concurrently from thousands of parallel matches on randomly generated environments. Each agent learns its own internal reward signal and rich representation of the world. These results indicate the great potential of multiagent reinforcement learning for artificial intelligence research."
                    },
                    {
                        "title": "Value-Decomposition Networks For Cooperative Multi-Agent Learning",
                        "abstract": "We study the problem of cooperative multi-agent reinforcement learning with a single joint reward signal. This class of learning problems is difficult because of the often large combined action and observation spaces. In the fully centralized and decentralized approaches, we find the problem of spurious rewards and a phenomenon we call the \"lazy agent\" problem, which arises due to partial observability. We address these problems by training individual agents with a novel value decomposition network architecture, which learns to decompose the team value function into agent-wise value functions. We perform an experimental evaluation across a range of partially-observable multi-agent domains and show that learning such value-decompositions leads to superior results, in particular when combined with weight sharing, role information and information channels."
                    },
                    {
                        "title": "Compressed Conditional Mean Embeddings for Model-Based Reinforcement Learning",
                        "abstract": "    We present a model-based approach to solving Markov decision processes (MDPs) in which the system dynamics are learned using conditional mean embeddings (CMEs). This class of methods comes with strong performance guarantees, and enables planning to be performed in an induced finite (pseudo-)MDP, which approximates the MDP, but can be solved exactly using dynamic programming. Two drawbacks of existing methods exist: firstly, the size of the induced finite (pseudo-)MDP scales quadratically with the amount of data used to learn the model, costing much memory and time when planning with the learned model; secondly, learning the CME itself using powerful kernel least-squares is costly \u2013 a second computational bottleneck. We present an algorithm which maintains a rich kernelized CME model class, but solves both problems: firstly we demonstrate that the loss function for the CME model suggests a principled approach to compressing the induced (pseudo-)MDP, leading to faster planning, while maintaining guarantees; secondly we propose to learn the CME model itself using fast sparse-greedy kernel regression well-suited to the RL context. We demonstrate superior performance to existing methods in this class of modelbased approaches on a range of MDPs.   "
                    },
                    {
                        "title": "Nesterov's accelerated gradient and momentum as approximations to regularised update descent",
                        "abstract": "We present a unifying framework for adapting the update direction in gradient-based iterative optimization methods. As natural special cases we re-derive classical momentum and Nesterov's accelerated gradient method, lending a new intuitive interpretation to the latter algorithm. We show that a new algorithm, which we term Regularised Gradient Descent, can converge more quickly than either Nesterov's algorithm or the classical momentum algorithm."
                    },
                    {
                        "title": "Approximate Newton Methods for Policy Search in Markov Decision Processes",
                        "abstract": "Approximate Newton methods are standard optimization tools which aim to maintain the benefits of Newton's method, such as a fast rate of convergence, while alleviating its drawbacks, such as computationally expensive calculation or estimation of the inverse Hessian. In this work we investigate approximate Newton methods for policy optimization in Markov decision processes (MDPs). We first analyse the structure of the Hessian of the total expected reward, which is a standard objective function for MDPs. We show that, like the gradient, the Hessian exhibits useful structure in the context of MDPs and we use this analysis to motivate two Gauss-Newton methods for MDPs. Like the Gauss-Newton method for non-linear least squares, these methods drop certain terms in the Hessian. The approximate Hessians possess desirable properties, such as negative definiteness, and we demonstrate several important performance guarantees including guaranteed ascent directions, invariance to afine transformation of the parameter space and convergence guarantees. We finally provide a unifying perspective of key policy search algorithms, demonstrating that our second Gauss-Newton algorithm is closely related to both the EM-algorithm and natural gradient ascent applied to MDPs, but performs significantly better in practice on a range of challenging domains."
                    },
                    {
                        "title": "Modelling Policies in MDPs in Reproducing Kernel Hilbert Space",
                        "abstract": "We consider modelling policies for MDPs in (vector-valued) reproducing kernel Hilbert function spaces (RKHS). This enables us to work \u201cnon-parametrically\u201d in a rich function class, and provides the ability to learn complex policies. We present a framework for performing gradientbased policy optimization in the RKHS, deriving the functional gradient of the return for our policy, which has a simple form and can be estimated efficiently. The policy representation naturally focuses on the relevant region of state space defined by the policy trajectories, and does not rely on a-priori defined basis points; this can be an advantage in high dimensions where suitable basis points may be difficult to define a-priori. The method is adaptive in the sense that the policy representation will naturally adapt to the complexity of the policy being modelled, which is achieved with standard efficient sparsification tools in an RKHS. We argue that finding a good kernel on states can be easier then remetrizing a high dimensional feature space. We demonstrate the approach on benchmark domains and a simulated quadrocopter navigation task."
                    },
                    {
                        "title": "A Gauss-Newton Method for Markov Decision Processes",
                        "abstract": "Approximate Newton methods are a standard optimization tool which aim to maintain the benefits of Newton's method, such as a fast rate of convergence, whilst alleviating its drawbacks, such as computationally expensive calculation or estimation of the inverse Hessian. In this work we investigate approximate Newton methods for policy optimization in Markov Decision Processes (MDPs). We first analyse the structure of the Hessian of the objective function for MDPs. We show that, like the gradient, the Hessian exhibits useful structure in the context of MDPs and we use this analysis to motivate two Gauss-Newton Methods for MDPs. Like the Gauss-Newton method for non-linear least squares, these methods involve approximating the Hessian by ignoring certain terms in the Hessian which are difficult to estimate. The approximate Hessians possess desirable properties, such as negative definiteness, and we demonstrate several important performance guarantees including guaranteed ascent directions, invariance to affine transformation of the parameter space, and convergence guarantees. We finally provide a unifying perspective of key policy search algorithms, demonstrating that our second Gauss-Newton algorithm is closely related to both the EM-algorithm and natural gradient ascent applied to MDPs, but performs significantly better in practice on a range of challenging domains."
                    },
                    {
                        "title": "Learning Transition Dynamics in MDPs with Online Regression and Greedy Feature Selection\u2217",
                        "abstract": "We present an approach to reinforcement learning in which the system dynamics are modelled using online linear regression between feature spaces, and a compact feature representation for the dynamics model is built incrementally using greedy feature selection. Candidate features are built online using kernels centred at datapoints as they are discovered. We implement the model learning method in a policy iteration scheme. The complexity of each policy iteration (feature learning, model learning, value estimation and policy improvement) is independent of the total amount of data observed, and only linear in the amount of new data added per iteration. The approach therefore scales up to complex problems requiring a huge amount of data to learn well. We validate the approach on benchmark MDPs and simulated quadrocopter navigation."
                    },
                    {
                        "title": "Deterministic Policy Gradient Algorithms",
                        "abstract": "In this paper we consider deterministic policy gradient algorithms for reinforcement learning with continuous actions. The deterministic policy gradient has a particularly appealing form: it is the expected gradient of the action-value function. This simple form means that the deterministic policy gradient can be estimated much more efficiently than the usual stochastic policy gradient. To ensure adequate exploration, we introduce an off-policy actor-critic algorithm that learns a deterministic target policy from an exploratory behaviour policy. We demonstrate that deterministic policy gradient algorithms can significantly outperform their stochastic counterparts in high-dimensional action spaces."
                    }
                ]
            },
            "4054df2d-071f-42f5-a6a2-09d32d4bb57d": {
                "pk": "4054df2d-071f-42f5-a6a2-09d32d4bb57d",
                "name": "Andrea Tacchetti",
                "collaborators": [
                    "Jan Balaguer",
                    "K. Tuyls",
                    "Mateusz Malinowski",
                    "David Raposo",
                    "Andy Ballard",
                    "J. Gilmer",
                    "George E. Dahl",
                    "Ashish Vaswani",
                    "Daan Wierstra",
                    "I. Gemp",
                    "Thomas W. Anthony",
                    "L. Sifre",
                    "Alex Castro-Ros",
                    "Eric Ni",
                    "George Tucker",
                    "Jacob Austin",
                    "James Keeling",
                    "Katie Millican",
                    "Rahma Chaabouni",
                    "Rohan Anil",
                    "Sebastian Borgeaud",
                    "Tom Hennigan",
                    "Eli Collins",
                    "Raphael Koster",
                    "R. \u00c9lie",
                    "Oliver P. Hauser",
                    "Matthew M. Botvinick",
                    "Christopher Summerfield",
                    "Yoram Bachrach",
                    "Richard Everett",
                    "Marc Lanctot",
                    "Siqi Liu",
                    "Luke Marris",
                    "Adam Santoro",
                    "N. Heess",
                    "R. Garnett",
                    "Danai Koutra",
                    "Lucy Campbell-Gillingham",
                    "R. Koster",
                    "C. Summerfield",
                    "Manfred Diaz",
                    "Liam Paull",
                    "Gemma Team Thomas Mesnard",
                    "Cassidy Hardin",
                    "Robert Dadashi",
                    "Surya Bhupatiraju",
                    "Shreya Pathak",
                    "Morgane Rivi\u00e8re",
                    "Mihir Kale",
                    "J Christopher Love",
                    "P. Tafti",
                    "L'eonard Hussenot",
                    "Aakanksha Chowdhery",
                    "Adam Roberts",
                    "Aditya Barua",
                    "Alex Botev",
                    "Ambrose Slone",
                    "Am'elie H'eliou",
                    "Anna Bulanova",
                    "Antonia Paterson",
                    "Beth Tsai",
                    "Bobak Shahriari",
                    "Charline Le Lan",
                    "Christopher A. Choquette-Choo",
                    "Cl'ement Crepy",
                    "Daniel Cer",
                    "Daphne Ippolito",
                    "David Reid",
                    "Elena Buchatskaya",
                    "Eric Noland",
                    "Geng Yan",
                    "George-Christian Muraru",
                    "Grigory Rozhdestvenskiy",
                    "H. Michalewski",
                    "Ian Tenney",
                    "Ivan Grishchenko",
                    "Jane Labanowski",
                    "Jean-Baptiste Lespiau",
                    "J. Stanway",
                    "Jenny Brennan",
                    "Jeremy Chen",
                    "Johan Ferret",
                    "Justin Chiu",
                    "J. Mao-Jones",
                    "Katherine Lee",
                    "Kathy Yu",
                    "Lars Lowe Sjoesund",
                    "Lisa Lee",
                    "Lucas Dixon",
                    "Machel Reid",
                    "Maciej Mikula",
                    "Mateo Wirth",
                    "Michael Sharman",
                    "Nikolai Chinaev",
                    "Nithum Thain",
                    "Olivier Bachem",
                    "Oscar Chang",
                    "O. Wahltinez",
                    "Paige Bailey",
                    "Paul Michel"
                ],
                "domain": [
                    "Machine Learning",
                    "Game Theory",
                    "Multi-Agent Systems",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Rethinking Teacher-Student Curriculum Learning through the Cooperative Mechanics of Experience",
                        "abstract": "Teacher-Student Curriculum Learning (TSCL) is a curriculum learning framework that draws inspiration from human cultural transmission and learning. It involves a teacher algorithm shaping the learning process of a learner algorithm by exposing it to controlled experiences. Despite its success, understanding the conditions under which TSCL is effective remains challenging. In this paper, we propose a data-centric perspective to analyze the underlying mechanics of the teacher-student interactions in TSCL. We leverage cooperative game theory to describe how the composition of the set of experiences presented by the teacher to the learner, as well as their order, influences the performance of the curriculum that is found by TSCL approaches. To do so, we demonstrate that for every TSCL problem, an equivalent cooperative game exists, and several key components of the TSCL framework can be reinterpreted using game-theoretic principles. Through experiments covering supervised learning, reinforcement learning, and classical games, we estimate the cooperative values of experiences and use value-proportional curriculum mechanisms to construct curricula, even in cases where TSCL struggles. The framework and experimental setup we present in this work represents a novel foundation for a deeper exploration of TSCL, shedding light on its underlying mechanisms and providing insights into its broader applicability in machine learning."
                    },
                    {
                        "title": "Gemma: Open Models Based on Gemini Research and Technology",
                        "abstract": "This work introduces Gemma, a family of lightweight, state-of-the art open models built from the research and technology used to create Gemini models. Gemma models demonstrate strong performance across academic benchmarks for language understanding, reasoning, and safety. We release two sizes of models (2 billion and 7 billion parameters), and provide both pretrained and fine-tuned checkpoints. Gemma outperforms similarly sized open models on 11 out of 18 text-based tasks, and we present comprehensive evaluations of safety and responsibility aspects of the models, alongside a detailed description of model development. We believe the responsible release of LLMs is critical for improving the safety of frontier models, and for enabling the next wave of LLM innovations."
                    },
                    {
                        "title": "Seed-Free Synthetic Data Generation Framework for Instruction-Tuning LLMs: A Case Study in Thai",
                        "abstract": "We present a synthetic data approach for instruction-tuning large language models (LLMs) for low-resource languages in a data-efficient manner, specifically focusing on Thai. We identify three key properties that contribute to the effectiveness of instruction-tuning datasets: fluency, diversity, and cultural context. We propose a seed-data-free framework for generating synthetic instruction-tuning data that incorporates these essential properties. Our framework employs an LLM to generate diverse topics, retrieve relevant contexts from Wikipedia, and create instructions for various tasks, such as question answering, summarization, and conversation. The experimental results show that our best-performing synthetic dataset, which incorporates all three key properties, achieves competitive performance using only 5,000 instructions when compared to state-of-the-art Thai LLMs trained on hundreds of thousands of instructions. Our code and dataset are publicly available at https://github.com/parinzee/ seed-free-synthetic-instruct ."
                    },
                    {
                        "title": "Using deep reinforcement learning to promote sustainable human behaviour on a common pool resource problem",
                        "abstract": "A canonical social dilemma arises when finite resources are allocated to a group of people, who can choose to either reciprocate with interest, or keep the proceeds for themselves. What resource allocation mechanisms will encourage levels of reciprocation that sustain the commons? Here, in an iterated multiplayer trust game, we use deep reinforcement learning (RL) to design an allocation mechanism that endogenously promotes sustainable contributions from human participants to a common pool resource. We first trained neural networks to behave like human players, creating a stimulated economy that allowed us to study how different mechanisms influenced the dynamics of receipt and reciprocation. We then used RL to train a social planner to maximise aggregate return to players. The social planner discovered a redistributive policy that led to a large surplus and an inclusive economy, in which players made roughly equal gains. The RL agent increased human surplus over baseline mechanisms based on unrestricted welfare or conditional cooperation, by conditioning its generosity on available resources and temporarily sanctioning defectors by allocating fewer resources to them. Examining the AI policy allowed us to develop an explainable mechanism that performed similarly and was more popular among players. Deep reinforcement learning can be used to discover mechanisms that promote sustainable human behaviour."
                    },
                    {
                        "title": "Scaling Opponent Shaping to High Dimensional Games",
                        "abstract": "In multi-agent settings with mixed incentives, methods developed for zero-sum games have been shown to lead to detrimental outcomes. To address this issue, opponent shaping (OS) methods explicitly learn to influence the learning dynamics of co-players and empirically lead to improved individual and collective outcomes. However, OS methods have only been evaluated in low-dimensional environments due to the challenges associated with estimating higher-order derivatives or scaling model-free meta-learning. Alternative methods that scale to more complex settings either converge to undesirable solutions or rely on unrealistic assumptions about the environment or co-players. In this paper, we successfully scale an OS-based approach to general-sum games with temporally-extended actions and long-time horizons for the first time. After analysing the representations of the meta-state and history used by previous algorithms, we propose a simplified version called Shaper. We show empirically that Shaper leads to improved individual and collective outcomes in a range of challenging settings from literature. We further formalize a technique previously implicit in the literature, and analyse its contribution to opponent shaping. We show empirically that this technique is helpful for the functioning of prior methods in certain environments. Lastly, we show that previous environments, such as the CoinGame, are inadequate for analysing temporally-extended general-sum interactions."
                    },
                    {
                        "title": "Unifying Grokking and Double Descent",
                        "abstract": "A principled understanding of generalization in deep learning may require unifying disparate observations under a single conceptual framework. Previous work has studied \\emph{grokking}, a training dynamic in which a sustained period of near-perfect training performance and near-chance test performance is eventually followed by generalization, as well as the superficially similar \\emph{double descent}. These topics have so far been studied in isolation. We hypothesize that grokking and double descent can be understood as instances of the same learning dynamics within a framework of pattern learning speeds. We propose that this framework also applies when varying model capacity instead of optimization steps, and provide the first demonstration of model-wise grokking."
                    },
                    {
                        "title": "Developing, Evaluating and Scaling Learning Agents in Multi-Agent Environments",
                        "abstract": "The Game Theory & Multi-Agent team at DeepMind studies several aspects of multi-agent learning ranging from computing approximations to fundamental concepts in game theory to simulating social dilemmas in rich spatial environments and training 3-d humanoids in difficult team coordination tasks. A signature aim of our group is to use the resources and expertise made available to us at DeepMind in deep reinforcement learning to explore multi-agent systems in complex environments and use these benchmarks to advance our understanding. Here, we summarise the recent work of our team and present a taxonomy that we feel highlights many important open challenges in multi-agent research."
                    },
                    {
                        "title": "Turbocharging Solution Concepts: Solving NEs, CEs and CCEs with Neural Equilibrium Solvers",
                        "abstract": "Solution concepts such as Nash Equilibria, Correlated Equilibria, and Coarse Correlated Equilibria are useful components for many multiagent machine learning algorithms. Unfortunately, solving a normal-form game could take prohibitive or non-deterministic time to converge, and could fail. We introduce the Neural Equilibrium Solver which utilizes a special equivariant neural network architecture to approximately solve the space of all games of fixed shape, buying speed and determinism. We define a flexible equilibrium selection framework, that is capable of uniquely selecting an equilibrium that minimizes relative entropy, or maximizes welfare. The network is trained without needing to generate any supervised training data. We show remarkable zero-shot generalization to larger games. We argue that such a network is a powerful component for many possible multiagent algorithms."
                    },
                    {
                        "title": "Leave Graphs Alone: Addressing Over-Squashing without Rewiring",
                        "abstract": "Recent works have investigated the role of graph bottlenecks in preventing long-range information propagation in message-passing graph neural networks, causing the so-called `over-squashing' phenomenon. As a remedy, graph rewiring mechanisms have been proposed as preprocessing steps. Graph Echo State Networks (GESNs) are a reservoir computing model for graphs, where node embeddings are recursively computed by an untrained message-passing function. In this paper, we show that GESNs can achieve a significantly better accuracy on six heterophilic node classification tasks without altering the graph connectivity, thus suggesting a different route for addressing the over-squashing problem."
                    },
                    {
                        "title": "Human-centered mechanism design with Democratic AI",
                        "abstract": "Building artificial intelligence (AI) that aligns with human values is an unsolved problem. Here, we developed a human-in-the-loop research pipeline called Democratic AI, in which reinforcement learning is used to design a social mechanism that humans prefer by majority. A large group of humans played an online investment game that involved deciding whether to keep a monetary endowment or to share it with others for collective benefit. Shared revenue was returned to players under two different redistribution mechanisms, one designed by the AI and the other by humans. The AI discovered a mechanism that redressed initial wealth imbalance, sanctioned free riders, and successfully won the majority vote. By optimizing for human preferences, Democratic AI may be a promising method for value-aligned policy innovation."
                    },
                    {
                        "title": "The Good Shepherd: An Oracle Agent for Mechanism Design",
                        "abstract": "From social networks to traffic routing, artificial learning agents are playing a central role in modern institutions. We must therefore understand how to leverage these systems to foster outcomes and behaviors that align with our own values and aspirations. While multiagent learning has received considerable attention in recent years, artificial agents have been primarily evaluated when interacting with fixed, non-learning co-players. While this evaluation scheme has merit, it fails to capture the dynamics faced by institutions that must deal with adaptive and continually learning constituents. Here we address this limitation, and construct agents (\"mechanisms\") that perform well when evaluated over the learning trajectory of their adaptive co-players (\"participants\"). The algorithm we propose consists of two nested learning loops: an inner loop where participants learn to best respond to fixed mechanisms; and an outer loop where the mechanism agent updates its policy based on experience. We report the performance of our mechanism agents when paired with both artificial learning agents and humans as co-players. Our results show that our mechanisms are able to shepherd the participants strategies towards favorable outcomes, indicating a path for modern institutions to effectively and automatically influence the strategies and behaviors of their constituents."
                    },
                    {
                        "title": "HCMD-zero: Learning Value Aligned Mechanisms from Data",
                        "abstract": "Artificial learning agents are mediating a larger and larger number of interactions among humans, firms, and organizations, and the intersection between mechanism design and machine learning has been heavily investigated in recent years. However, mechanism design methods often make strong assumptions on how participants behave (e.g. rationality), on the kind of knowledge designers have access to a priori (e.g. access to strong baseline mechanisms), or on what the goal of the mechanism should be (e.g. total welfare). Here we introduce HCMD-zero, a general purpose method to construct mechanisms making none of these three assumptions. HCMD-zero learns to mediate interactions among participants and adjusts the mechanism parameters to make itself more likely to be preferred by participants. It does so by remaining engaged in an electoral contest with copies of itself, thereby accessing direct feedback from participants. We test our method on a stylized resource allocation game that highlights the tension between productivity, equality and the temptation to free ride. HCMD-zero produces a mechanism that is preferred by human participants over a strong baseline, it does so automatically, without requiring prior knowledge, and using human behavioral trajectories sparingly and effectively. Our analysis shows HCMD-zero consistently makes the mechanism policy more and more likely to be preferred by human participants over the course of training, and that it results in a mechanism with an interpretable and intuitive policy."
                    },
                    {
                        "title": "Leave Graphs Alone: Addressing Over-Squashing without Rewiring",
                        "abstract": "Abstract"
                    },
                    {
                        "title": "Sample-based Approximation of Nash in Large Many-Player Games via Gradient Descent",
                        "abstract": "Nash equilibrium is a central concept in game theory. Several Nash solvers exist, yet none scale to normal-form games with many actions and many players, especially those with payoff tensors too big to be stored in memory. In this work, we propose an approach that iteratively improves an approximation to a Nash equilibrium through joint play. It accomplishes this by tracing a previously established homotopy that defines a continuum of equilibria for the game regularized with decaying levels of entropy. This continuum asymptotically approaches the limiting logit equilibrium, proven by McKelvey and Palfrey (1995) to be unique in almost all games, thereby partially circumventing the well-known equilibrium selection problem of many-player games. To encourage iterates to remain near this path, we efficiently minimize average deviation incentive via stochastic gradient descent, intelligently sampling entries in the payoff tensor as needed. Monte Carlo estimates of the stochastic gradient from joint play are biased due to the appearance of a nonlinear max operator in the objective, so we introduce additional innovations to the algorithm to alleviate gradient bias. The descent process can also be viewed as repeatedly constructing and reacting to a polymatrix approximation to the game. In these ways, our proposed approach, average deviation incentive descent with adaptive sampling (ADIDAS), is most similar to three classical approaches, namely homotopy-type, Lyapunov, and iterative polymatrix solvers. The lack of local convergence guarantees for biased gradient descent prevents guaranteed convergence to Nash, however, we demonstrate through extensive experiments the ability of this approach to approximate a unique Nash in normal-form games with as many as seven players and twenty one actions (several billion outcomes) that are orders of magnitude larger than those possible with prior algorithms."
                    }
                ]
            }
        }
    },
    "2407.08906": {
        "paper_data": {
            "title": "AirSketch: Generative Motion to Sketch",
            "url": "http://arxiv.org/abs/2407.08906v1",
            "arxiv_id": "2407.08906",
            "authors": [
                "Hui Xian Grace Lim",
                "Xuanming Cui",
                "Yogesh S Rawat",
                "Ser-Nam Lim"
            ],
            "abstract": "Illustration is a fundamental mode of human expression and communication. Certain types of motion that accompany speech can provide this illustrative mode of communication. While Augmented and Virtual Reality technologies (AR/VR) have introduced tools for producing drawings with hand motions (air drawing), they typically require costly hardware and additional digital markers, thereby limiting their accessibility and portability. Furthermore, air drawing demands considerable skill to achieve aesthetic results. To address these challenges, we introduce the concept of AirSketch, aimed at generating faithful and visually coherent sketches directly from hand motions, eliminating the need for complicated headsets or markers. We devise a simple augmentation-based self-supervised training procedure, enabling a controllable image diffusion model to learn to translate from highly noisy hand tracking images to clean, aesthetically pleasing sketches, while preserving the essential visual cues from the original tracking data. We present two air drawing datasets to study this problem. Our findings demonstrate that beyond producing photo-realistic images from precise spatial inputs, controllable image diffusion can effectively produce a refined, clear sketch from a noisy input. Our work serves as an initial step towards marker-less air drawing and reveals distinct applications of controllable diffusion models to AirSketch and AR/VR in general.",
            "introduction": "   1 Introduction  Hand gestures are an essential element in communication[42]. In particular, iconic hand motions (i.e. air drawing) can depict visual aspects of an object. This form of expression is frequently used to visually supplement verbal communication and is used in various practical applications, including conceptual discussions, overcoming language barriers, and aiding in visual design.   Popular AR/VR tools like Google\u2019s Tiltbrush\u00a0[19] and HoloARt\u00a0[5] create visuals via hand motions, but are inconvenient. These applications generally require head-mounted displays and digital hand markers, which are costly and hinder portability. Furthermore, their weight and temperature make them unsuitable for continuous use\u00a0[59, 71, 12], and so are inconvenient for spontaneous usage. Yet, these devices provide accurate positioning, stabilization, and varied brush controls and are crucial to producing high-quality drawings.   Can we generate sketches from hand motions without additional sensors or markers? In order to enhance accessibility and convenience, we aim to generate sketches from hand motions videos captured using any standard camera embedded in devices like smartphones or smart glasses.   One could clearly deploy hand tracking algorithms\u00a0[9, 73] to turn these hand motion videos into sketches. However, creating air drawings with a hand tracking algorithm alone presents several challenges. These include the user\u2019s drawing ability, physical fatigue, and inaccuracies in hand tracking. Noise in hand tracking can severely distort a sketch, rendering it almost unrecognizable.   The objective, therefore, is to generate clean sketches that faithfully represent the user\u2019s intent, from highly noisy and distorted hand motion input. This task requires the model to possess a deep understanding of shape and object priors, enabling it to discern and correct deformed motion cues while filtering out undesirable noise. We refer to this task as Generative Motion to Sketch.   There are many approaches to this task, with different architecture and data modalities. The input modality might include learned video representations, coordinate sequences from a hand tracking algorithm, or a rasterized image. Depending on the modality, the task may also be reformulated as video-to-sketch, sequence-to-sequence, image-to-image, or a combination thereof. This diversity introduces interesting opportunities for rich exploration of all these diverse approaches.   We explore the use of controllable image Diffusion Models (DM) in generating sketches from motion. Existing work such as ControlNet\u00a0[75] and T2IAdapter\u00a0[43] generate photo-realistic images given spatially-precise conditioning images. We explore a different use case by using controllable DMs to \u201creconstruct\u201d clean sketches from severely distorted and noisy input images obtained with a hand tracking algorithm. We propose a simple, augmentation-based, self-supervised training procedure and construct two air drawings datasets for evaluation purposes.   Our experiments show that with our augmentation-based training, controllable image DMs are able to recognize and interpret correct visual cues from noisy tracking images, some of which even appear to be nearly unrecognizable to the human eye, and generate sketches faithfully and aesthetically, while being robust to unseen objects. Moreover, we show that through simple augmentations, the model is capable of sketch-completion and text-instructed stroke styling. Finally, we conduct ablations to investigate 1) the effects of different augmentations, 2) the contribution of text prompts to sketch generation and 3) the effect of different level of input \u2018chaos\u2019 on the quality of",
            "references": [
                {
                    "title": "How to Handle Sketch-Abstraction in Sketch-Based Image Retrieval?",
                    "abstract": "In this paper, we propose a novel abstraction-aware sketch-based image retrieval framework capable of handling sketch abstraction at varied levels. Prior works had mainly focused on tackling sub-factors such as drawing style and order, we instead attempt to model abstraction as a whole, and propose feature-level and retrieval granularity-level designs so that the system builds into its DNA the necessary means to interpret abstraction. On learning abstraction-aware features, we for the first-time harness the rich semantic embedding of pre-trained Style-GAN model, together with a novel abstraction-level mapper that deciphers the level of abstraction and dynamically selects appropriate dimensions in the feature matrix correspondingly, to construct a feature matrix embedding that can be freely traversed to accommodate different levels of abstraction. For granularity-level abstraction understanding, we dictate that the retrieval model should not treat all abstraction-levels equally and introduce a differentiable surrogate Acc. @q loss to inject that understanding into the system. Different to the gold-standard triplet loss, our Acc.@q loss uniquely allows a sketch to narrow/broaden its focus in terms of how stringent the evaluation should be - the more abstract a sketch, the less stringent (higher q). Extensive experiments depict our method to outperform existing state-of-the-arts in standard SBIR tasks along with challenging scenarios like early retrieval, forensic sketch-photo matching, and style-invariant retrieval."
                },
                {
                    "title": "Image-Based Arabic Sign Language Recognition System Using Transfer Deep Learning Models",
                    "abstract": "Sign language is a unique communication tool helping to bridge the gap between people with hearing impairments and the general public. It holds paramount importance for various communities, as it allows individuals with hearing difficulties to communicate effectively. In sign languages, there are numerous signs, each characterized by differences in hand shapes, hand positions, motions, facial expressions, and body parts used to convey specific meanings. The complexity of visual sign language recognition poses a significant challenge in the computer vision research area. This study presents an Arabic Sign Language recognition (ArSL) system that utilizes convolutional neural networks (CNNs) and several transfer learning models to automatically and accurately identify Arabic Sign Language characters. The dataset used for this study comprises 54,049 images of ArSL letters. The results of this research indicate that InceptionV3 outperformed other pretrained models, achieving a remarkable 100% accuracy score and a 0.00 loss score without overfitting. These impressive performance measures highlight the distinct capabilities of InceptionV3 in recognizing Arabic characters and underscore its robustness against overfitting. This enhances its potential for future research in the field of Arabic Sign Language recognition."
                },
                {
                    "title": "Breathing Life Into Sketches Using Text-to-Video Priors",
                    "abstract": "A sketch is one of the most intuitive and versatile tools humans use to convey their ideas visually. An animated sketch opens another dimension to the expression of ideas and is widely used by designers for a variety of purposes. Animating sketches is a laborious process, requiring extensive experience and professional design skills. In this work, we present a method that automatically adds motion to a single-subject sketch (hence, \u201cbreathing life into it\u201d), merely by providing a text prompt indicating the desired motion. The output is a short animation provided in vector representation, which can be easily edited. Our method does not require extensive training, but instead leverages the motion prior of a large pretrained text-to- video diffusion model using a score-distillation loss to guide the placement of strokes. To promote natural and smooth motion and to better preserve the sketch's appearance, we model the learned motion through two components. The first governs small local deformations and the second controls global affine transformations. Surprisingly, wefind that even models that struggle to generate sketch videos on their own can still serve as a useful backbone for animating abstract representations."
                },
                {
                    "title": "Democratising 2D Sketch to 3D Shape Retrieval Through Pivoting",
                    "abstract": "This paper studies the problem of 2D sketch to 3D shape retrieval, but with a focus on democratising the process. We would like this democratisation to happen on two fronts: (i) to remove the need for large-scale specifically sourced 2D sketch and 3D shape datasets, and (ii) to remove restrictions on how well the user needs to sketch and from what viewpoints. The end result is a system that is trainable using existing datasets, and once trained allows users to sketch regardless of drawing skills and without restriction on view angle. We achieve all this via a clever use of pivoting, along with novel designs that injects 3D understanding of 2D sketches into the system. We perform pivoting using two existing datasets, each from a distant research domain to the other: 2D sketch and photo pairs from the sketch-based image retrieval field (SBIR), and 3D shapes from ShapeNet. It follows that the actual feature pivoting happens on photos from the former and 2D projections from the latter. Doing this already achieves most of our democratisation challenge \u2013 the level of 2D sketch abstraction embedded in SBIR dataset offers demoralization on drawing quality, and the whole thing works without a specifically sourced 2D sketch and 3D model pair. To further achieve democratisation on sketching viewpoint, we \u201clift\u201d 2D sketches to 3D space using Blind Perspective-n-Points (BPnP) that injects 3D-aware information into the sketch encoder. Results show ours achieves competitive performance compared with fully-supervised baselines, while meeting all set democratisation goals."
                },
                {
                    "title": "BoxDiff: Text-to-Image Synthesis with Training-Free Box-Constrained Diffusion",
                    "abstract": "Recent text-to-image diffusion models have demonstrated an astonishing capacity to generate high-quality images. However, researchers mainly studied the way of synthesizing images with only text prompts. While some works have explored using other modalities as conditions, considerable paired data, e.g., box/mask-image pairs, and fine-tuning time are required for nurturing models. As such paired data is time-consuming and labor-intensive to acquire and restricted to a closed set, this potentially becomes the bottleneck for applications in an open world. This paper focuses on the simplest form of user-provided conditions, e.g., box or scribble. To mitigate the aforementioned problem, we propose a training-free method to control objects and contexts in the synthesized images adhering to the given spatial conditions. Specifically, three spatial constraints, i.e., Inner-Box, Outer-Box, and Corner Constraints, are designed and seamlessly integrated into the denoising step of diffusion models, requiring no additional training and massive annotated layout data. Extensive experimental results demonstrate that the proposed constraints can control what and where to present in the images while retaining the ability of Diffusion models to synthesize with high fidelity and diverse concept coverage."
                },
                {
                    "title": "HumanSD: A Native Skeleton-Guided Diffusion Model for Human Image Generation",
                    "abstract": "Controllable human image generation (HIG) has numerous real-life applications. State-of-the-art solutions, such as ControlNet and T2I-Adapter, introduce an additional learnable branch on top of the frozen pre-trained stable diffusion (SD) model, which can enforce various conditions, including skeleton guidance of HIG. While such a plug-and-play approach is appealing, the inevitable and uncertain conflicts between the original images produced from the frozen SD branch and the given condition incur significant challenges for the learnable branch, which essentially conducts image feature editing for condition enforcement.In this work, we propose a native skeleton-guided diffusion model for controllable HIG called HumanSD. Instead of performing image editing with dual-branch diffusion, we fine-tune the original SD model using a novel heatmap-guided denoising loss. This strategy effectively and efficiently strengthens the given skeleton condition during model training while mitigating the catastrophic forgetting effects. HumanSD is fine-tuned on the assembly of three large-scale human-centric datasets with text-image-pose information, two of which are established in this work. Experimental results show that HumanSD outperforms ControlNet in terms of pose control and image quality, particularly when the given skeleton guidance is sophisticated. Code and data are available at: https://idea-research.github.io/HumanSD/."
                },
                {
                    "title": "T2I-Adapter: Learning Adapters to Dig out More Controllable Ability for Text-to-Image Diffusion Models",
                    "abstract": "The incredible generative ability of large-scale text-to-image (T2I) models has demonstrated strong power of learning complex structures and meaningful semantics. However, relying solely on text prompts cannot fully take advantage of the knowledge learned by the model, especially when flexible and accurate controlling (e.g., structure and color) is needed. In this paper, we aim to ``dig out\" the capabilities that T2I models have implicitly learned, and then explicitly use them to control the generation more granularly. Specifically, we propose to learn low-cost T2I-Adapters to align internal knowledge in T2I models with external control signals, while freezing the original large T2I models. In this way, we can train various adapters according to different conditions, achieving rich control and editing effects in the color and structure of the generation results. Further, the proposed T2I-Adapters have attractive properties of practical value, such as composability and generalization ability. Extensive experiments demonstrate that our T2I-Adapter has promising generation quality and a wide range of applications. Our code is available at https://github.com/TencentARC/T2I-Adapter."
                },
                {
                    "title": "Adding Conditional Control to Text-to-Image Diffusion Models",
                    "abstract": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models."
                },
                {
                    "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
                    "abstract": "Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples."
                },
                {
                    "title": "Blended Diffusion for Text-driven Editing of Natural Images",
                    "abstract": "Natural language offers a highly intuitive interface for image editing. In this paper, we introduce the first solution for performing local (region-based) edits in generic natural images, based on a natural language description along with an ROI mask. We achieve our goal by leveraging and combining a pretrained language-image model (CLIP), to steer the edit towards a user-provided text prompt, with a denoising diffusion probabilistic model (DDPM) to generate natural-looking results. To seamlessly fuse the edited region with the unchanged parts of the image, we spatially blend noised versions of the input image with the local text-guided diffusion latent at a progression of noise levels. In addition, we show that adding augmentations to the diffusion process mitigates adversarial results. We compare against several baselines and related methods, both qualitatively and quantitatively, and show that our method outperforms these solutions in terms of overall realism, ability to preserve the background and matching the text. Finally, we show several text-driven editing applications, including adding a new object to an image, removing/replacing/altering existing objects, background replacement, and image extrapolation."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Vectorization and Rasterization: Self-Supervised Learning for Sketch and Handwriting",
                    "abstract": "Self-supervised learning has gained prominence due to its efficacy at learning powerful representations from unlabelled data that achieve excellent performance on many challenging downstream tasks. However, supervision-free pre-text tasks are challenging to design and usually modality specific. Although there is a rich literature of self-supervised methods for either spatial (such as images) or temporal data (sound or text) modalities, a common pretext task that benefits both modalities is largely missing. In this paper, we are interested in defining a self-supervised pre-text task for sketches and handwriting data. This data is uniquely characterised by its existence in dual modalities of rasterized images and vector coordinate sequences. We address and exploit this dual representation by proposing two novel cross-modal translation pre-text tasks for self-supervised feature learning: Vectorization and Rasterization. Vectorization learns to map image space to vector coordinates and rasterization maps vector coordinates to image space. We show that our learned encoder modules benefit both raster-based and vector-based downstream approaches to analysing hand-drawn data. Empirical evidence shows that our novel pre-text tasks surpass existing single and multi-modal self-supervision methods."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Zero-Shot Text-to-Image Generation",
                    "abstract": "Text-to-image generation has traditionally focused on finding better modeling assumptions for training on a fixed dataset. These assumptions might involve complex architectures, auxiliary losses, or side information such as object part labels or segmentation masks supplied during training. We describe a simple approach for this task based on a transformer that autoregressively models the text and image tokens as a single stream of data. With sufficient data and scale, our approach is competitive with previous domain-specific models when evaluated in a zero-shot fashion."
                },
                {
                    "title": "A Saudi Sign Language Recognition System based on Convolutional Neural Networks",
                    "abstract": "Sign language is the main communication method for deaf people. It is a collection of signs that deaf people use to deal with each other. Deaf people find it difficult to communicate with normal people, as most of them do not understand the signs of the sign language. Sign language recognition systems translate the signs into natural languages and thus shorten the gap between deaf and normal people. Many studies have been done on different sign languages. There is a considerable number of studies on the standard Arabic sign language. In Saudi Arabia, deaf people use the Saudi language, which is different from standard Arabic. This study proposes a smart recognition system for Saudi sign language based on convolutional neural networks. The system is based on the Saudi Sign language dictionary, which was published recently in 2018. In this study, we constructed a dataset of 40 Saudi signs with about 700 images for each sign. We then developed a deep convolutional neural network and trained it on the constructed dataset. To have better recognition, we took images of the signs with different hand sizes, skin colors, lighting, backgrounds, and with/without accessories. The results showed that the recognition model achieved an accuracy of 97.69% for training data and 99.47% for testing data. The model was implemented in two versions: mobile and desktop."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "How2Sign: A Large-scale Multimodal Dataset for Continuous American Sign Language",
                    "abstract": "One of the factors that have hindered progress in the areas of sign language recognition, translation, and production is the absence of large annotated datasets. Towards this end, we introduce How2Sign, a multimodal and multiview continuous American Sign Language (ASL) dataset, consisting of a parallel corpus of more than 80 hours of sign language videos and a set of corresponding modalities including speech, English transcripts, and depth. A three-hour subset was further recorded in the Panoptic studio enabling detailed 3D pose estimation. To evaluate the potential of How2Sign for real-world impact, we conduct a study with ASL signers and show that synthesized videos using our dataset can indeed be understood. The study further gives insights on challenges that computer vision should address in order to make progress in this field."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "MediaPipe Hands: On-device Real-time Hand Tracking",
                    "abstract": "We present a real-time on-device hand tracking pipeline that predicts hand skeleton from single RGB camera for AR/VR applications. The pipeline consists of two models: 1) a palm detector, 2) a hand landmark model. It's implemented via MediaPipe, a framework for building cross-platform ML solutions. The proposed model and pipeline architecture demonstrates real-time inference speed on mobile GPUs and high prediction quality. MediaPipe Hands is open sourced at https://mediapipe.dev."
                },
                {
                    "title": "Sketch-BERT: Learning Sketch Bidirectional Encoder Representation From Transformers by Self-Supervised Learning of Sketch Gestalt",
                    "abstract": "Previous researches of sketches often considered sketches in pixel format and leveraged CNN based models in the sketch understanding. Fundamentally, a sketch is stored as a sequence of data points, a vector format representation, rather than the photo-realistic image of pixels. SketchRNN studied a generative neural representation for sketches of vector format by Long Short Term Memory networks (LSTM). Unfortunately, the representation learned by SketchRNN is primarily for the generation tasks, rather than the other tasks of recognition and retrieval of sketches. To this end and inspired by the recent BERT model, we present a model of learning Sketch Bidirectional Encoder Representation from Transformer (Sketch-BERT). We generalize BERT to sketch domain, with the novel proposed components and pre-training algorithms, including the newly designed sketch embedding networks, and the self-supervised learning of sketch gestalt. Particularly, towards the pre-training task, we present a novel Sketch Gestalt Model (SGM) to help train the Sketch-BERT. Experimentally, we show that the learned representation of Sketch-BERT can help and improve the performance of the downstream tasks of sketch recognition, sketch retrieval, and sketch gestalt."
                },
                {
                    "title": "Sketchformer: Transformer-Based Representation for Sketched Structure",
                    "abstract": "Sketchformer is a novel transformer-based representation for encoding free-hand sketches input in a vector form, i.e. as a sequence of strokes. Sketchformer effectively addresses multiple tasks: sketch classification, sketch based image retrieval (SBIR), and the reconstruction and interpolation of sketches. We report several variants exploring continuous and tokenized input representations, and contrast their performance. Our learned embedding, driven by a dictionary learning tokenization scheme, yields state of the art performance in classification and image retrieval tasks, when compared against baseline representations driven by LSTM sequence to sequence architectures: SketchRNN and derivatives. We show that sketch reconstruction and interpolation are improved significantly by the Sketchformer embedding for complex sketches with longer stroke sequences."
                },
                {
                    "title": "A Survey on Hand Pose Estimation with Wearable Sensors and Computer-Vision-Based Methods",
                    "abstract": "Real-time sensing and modeling of the human body, especially the hands, is an important research endeavor for various applicative purposes such as in natural human computer interactions. Hand pose estimation is a big academic and technical challenge due to the complex structure and dexterous movement of human hands. Boosted by advancements from both hardware and artificial intelligence, various prototypes of data gloves and computer-vision-based methods have been proposed for accurate and rapid hand pose estimation in recent years. However, existing reviews either focused on data gloves or on vision methods or were even based on a particular type of camera, such as the depth camera. The purpose of this survey is to conduct a comprehensive and timely review of recent research advances in sensor-based hand pose estimation, including wearable and vision-based solutions. Hand kinematic models are firstly discussed. An in-depth review is conducted on data gloves and vision-based sensor systems with corresponding modeling methods. Particularly, this review also discusses deep-learning-based methods, which are very promising in hand pose estimation. Moreover, the advantages and drawbacks of the current hand gesture estimation methods, the applicative scope, and related challenges are also discussed."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "SketchGAN: Joint Sketch Completion and Recognition With Generative Adversarial Network",
                    "abstract": "Hand-drawn sketch recognition is a fundamental problem in computer vision, widely used in sketch-based image and video retrieval, editing, and reorganization. Previous methods often assume that a complete sketch is used as input; however, hand-drawn sketches in common application scenarios are often incomplete, which makes sketch recognition a challenging problem. In this paper, we propose SketchGAN, a new generative adversarial network (GAN) based approach that jointly completes and recognizes a sketch, boosting the performance of both tasks. Specifically, we use a cascade Encode-Decoder network to complete the input sketch in an iterative manner, and employ an auxiliary sketch recognition task to recognize the completed sketch. Experiments on the Sketchy database benchmark demonstrate that our joint learning approach achieves competitive sketch completion and recognition performance compared with the state-of-the-art methods. Further experiments using several sketch-based applications also validate the performance of our method."
                },
                {
                    "title": "Self-Supervised 3D Hand Pose Estimation Through Training by Fitting",
                    "abstract": "We present a self-supervision method for 3D hand pose estimation from depth maps. We begin with a neural network initialized with synthesized data and fine-tune it on real but unlabelled depth maps by minimizing a set of data-fitting terms. By approximating the hand surface with a set of spheres, we design a differentiable hand renderer to align estimates by comparing the rendered and input depth maps. In addition, we place a set of priors including a data-driven term to further regulate the estimate's kinematic feasibility. Our method makes highly accurate estimates comparable to current supervised methods which require large amounts of labelled training samples, thereby advancing state-of-the-art in unsupervised learning for hand pose estimation."
                },
                {
                    "title": "Generalising Fine-Grained Sketch-Based Image Retrieval",
                    "abstract": "Fine-grained sketch-based image retrieval (FG-SBIR) addresses matching specific photo instance using free-hand sketch as a query modality. Existing models aim to learn an embedding space in which sketch and photo can be directly compared. While successful, they require instance-level pairing within each coarse-grained category as annotated training data. Since the learned embedding space is domain-specific, these models do not generalise well across categories. This limits the practical applicability of FG-SBIR. In this paper, we identify cross-category generalisation for FG-SBIR as a domain generalisation problem, and propose the first solution. Our key contribution is a novel unsupervised learning approach to model a universal manifold of prototypical visual sketch traits. This manifold can then be used to paramaterise the learning of a sketch/photo representation. Model adaptation to novel categories then becomes automatic via embedding the novel sketch in the manifold and updating the representation and retrieval function accordingly. Experiments on the two largest FG-SBIR datasets, Sketchy and QMUL-Shoe-V2, demonstrate the efficacy of our approach in enabling cross-category generalisation of FG-SBIR."
                },
                {
                    "title": "OpenPose: Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields",
                    "abstract": "Realtime multi-person 2D pose estimation is a key component in enabling machines to have an understanding of people in images and videos. In this work, we present a realtime approach to detect the 2D pose of multiple people in an image. The proposed method uses a nonparametric representation, which we refer to as Part Affinity Fields (PAFs), to learn to associate body parts with individuals in the image. This bottom-up system achieves high accuracy and realtime performance, regardless of the number of people in the image. In previous work, PAFs and body part location estimation were refined simultaneously across training stages. We demonstrate that a PAF-only refinement rather than both PAF and body part location refinement results in a substantial increase in both runtime performance and accuracy. We also present the first combined body and foot keypoint detector, based on an internal annotated foot dataset that we have publicly released. We show that the combined detector not only reduces the inference time compared to running them sequentially, but also maintains the accuracy of each component individually. This work has culminated in the release of OpenPose, the first open-source realtime system for multi-person 2D pose detection, including body, foot, hand, and facial keypoints."
                },
                {
                    "title": "The Unreasonable Effectiveness of Deep Features as a Perceptual Metric",
                    "abstract": "While it is nearly effortless for humans to quickly assess the perceptual similarity between two images, the underlying processes are thought to be quite complex. Despite this, the most widely used perceptual metrics today, such as PSNR and SSIM, are simple, shallow functions, and fail to account for many nuances of human perception. Recently, the deep learning community has found that features of the VGG network trained on ImageNet classification has been remarkably useful as a training loss for image synthesis. But how perceptual are these so-called \"perceptual losses\"? What elements are critical for their success? To answer these questions, we introduce a new dataset of human perceptual similarity judgments. We systematically evaluate deep features across different architectures and tasks and compare them with classic metrics. We find that deep features outperform all previous metrics by large margins on our dataset. More surprisingly, this result is not restricted to ImageNet-trained VGG features, but holds across different deep architectures and levels of supervision (supervised, self-supervised, or even unsupervised). Our results suggest that perceptual similarity is an emergent property shared across deep visual representations."
                },
                {
                    "title": "HoloARt Video Showcase: Painting with Holograms in Mixed Reality",
                    "abstract": "We propose HoloARt, a new media art that explores the use of holograms in mixed reality for creative self expression. We designed a system that allows the user to turn their physical environment into a canvas where digital holograms and physical objects co-exist in the real and virtual world. Users are able to virtually spray and splatter hologram paint on top of physical objects and surfaces as well as painting in the air by only using their hands. The content grows dynamically, following the natural movements of the user. The system is self-contained and does not require hand controllers or positional tracking sensors on the space. This video accompanies our CHI paper and Art Installation [1]."
                },
                {
                    "title": "A Neural Representation of Sketch Drawings",
                    "abstract": "We present sketch-rnn, a recurrent neural network (RNN) able to construct stroke-based drawings of common objects. The model is trained on thousands of crude human-drawn images representing hundreds of classes. We outline a framework for conditional and unconditional sketch generation, and describe new robust training methods for generating coherent sketch drawings in a vector format."
                },
                {
                    "title": "Deep Sketch Hashing: Fast Free-Hand Sketch-Based Image Retrieval",
                    "abstract": "Free-hand sketch-based image retrieval (SBIR) is a specific cross-view retrieval task, in which queries are abstract and ambiguous sketches while the retrieval database is formed with natural images. Work in this area mainly focuses on extracting representative and shared features for sketches and natural images. However, these can neither cope well with the geometric distortion between sketches and images nor be feasible for large-scale SBIR due to the heavy continuous-valued distance computation. In this paper, we speed up SBIR by introducing a novel binary coding method, named Deep Sketch Hashing (DSH), where a semi-heterogeneous deep architecture is proposed and incorporated into an end-to-end binary coding framework. Specifically, three convolutional neural networks are utilized to encode free-hand sketches, natural images and, especially, the auxiliary sketch-tokens which are adopted as bridges to mitigate the sketch-image geometric distortion. The learned DSH codes can effectively capture the cross-view similarities as well as the intrinsic semantic correlations between different categories. To the best of our knowledge, DSH is the first hashing work specifically designed for category-level SBIR with an end-to-end deep architecture. The proposed DSH is comprehensively evaluated on two large-scale datasets of TU-Berlin Extension and Sketchy, and the experiments consistently show DSHs superior SBIR accuracies over several state-of-the-art methods, while achieving significantly reduced retrieval time and memory footprint."
                },
                {
                    "title": "Pose Estimation for Augmented Reality: A Hands-On Survey",
                    "abstract": "Augmented reality (AR) allows to seamlessly insert virtual objects in an image sequence. In order to accomplish this goal, it is important that synthetic elements are rendered and aligned in the scene in an accurate and visually acceptable way. The solution of this problem can be related to a pose estimation or, equivalently, a camera localization process. This paper aims at presenting a brief but almost self-contented introduction to the most important approaches dedicated to vision-based camera localization along with a survey of several extension proposed in the recent years. For most of the presented approaches, we also provide links to code of short examples. This should allow readers to easily bridge the gap between theoretical aspects and practical implementations."
                },
                {
                    "title": "Robust 3D Hand Pose Estimation in Single Depth Images: From Single-View CNN to Multi-View CNNs",
                    "abstract": "Articulated hand pose estimation plays an important role in human-computer interaction. Despite the recent progress, the accuracy of existing methods is still not satisfactory, partially due to the difficulty of embedded high-dimensional and non-linear regression problem. Different from the existing discriminative methods that regress for the hand pose with a single depth image, we propose to first project the query depth image onto three orthogonal planes and utilize these multi-view projections to regress for 2D heat-maps which estimate the joint positions on each plane. These multi-view heat-maps are then fused to produce final 3D hand pose estimation with learned pose priors. Experiments show that the proposed method largely outperforms state-of-the-art on a challenging dataset. Moreover, a cross-dataset experiment also demonstrates the good generalization ability of the proposed method."
                },
                {
                    "title": "Sketch-based 3D shape retrieval using Convolutional Neural Networks",
                    "abstract": "Retrieving 3D models from 2D human sketches has received considerable attention in the areas of graphics, image retrieval, and computer vision. Almost always in state of the art approaches a large amount of \u201cbest views\u201d are computed for 3D models, with the hope that the query sketch matches one of these 2D projections of 3D models using predefined features. We argue that this two stage approach (view selection - matching) is pragmatic but also problematic because the \u201cbest views\u201d are subjective and ambiguous, which makes the matching inputs obscure. This imprecise nature of matching further makes it challenging to choose features manually. Instead of relying on the elusive concept of \u201cbest views\u201d and the hand-crafted features, we propose to define our views using a minimalism approach and learn features for both sketches and views. Specifically, we drastically reduce the number of views to only two predefined directions for the whole dataset. Then, we learn two Siamese Convolutional Neural Networks (CNNs), one for the views and one for the sketches. The loss function is defined on the within-domain as well as the cross-domain similarities. Our experiments on three benchmark datasets demonstrate that our method is significantly better than state of the art approaches, and outperforms them in all conventional metrics."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "How do humans sketch objects?",
                    "abstract": "Humans have used sketching to depict our visual world since prehistoric times. Even today, sketching is possibly the only rendering technique readily available to all humans. This paper is the first large scale exploration of human sketches. We analyze the distribution of non-expert sketches of everyday objects such as 'teapot' or 'car'. We ask humans to sketch objects of a given category and gather 20,000 unique sketches evenly distributed over 250 object categories. With this dataset we perform a perceptual study and find that humans can correctly identify the object category of a sketch 73% of the time. We compare human performance against computational recognition methods. We develop a bag-of-features sketch representation and use multi-class support vector machines, trained on our sketch dataset, to classify sketches. The resulting recognition method is able to identify unknown sketches with 56% accuracy (chance is 0.4%). Based on the computational model, we demonstrate an interactive sketch recognition system. We release the complete crowd-sourced dataset of sketches to the community."
                },
                {
                    "title": "Handy AR: Markerless Inspection of Augmented Reality Objects Using Fingertip Tracking",
                    "abstract": "We present markerless camera tracking and user interface methodology for readily inspecting augmented reality (AR) objects in wearable computing applications. Instead of a marker, we use the human hand as a distinctive pattern that almost all wearable computer users have readily available. We present a robust real-time algorithm that recognizes fingertips to reconstruct the six-degree-of-freedom camera pose relative to the user's outstretched hand. A hand pose model is constructed in a one-time calibration step by measuring the fingertip positions in presence of ground-truth scale information. Through frame-by-frame reconstruction of the camera pose relative to the hand, we can stabilize 3D graphics annotations on top of the hand, allowing the user to inspect such virtual objects conveniently from different viewing angles in AR. We evaluate our approach with regard to speed and accuracy, and compare it to state-of-the-art marker-based AR systems. We demonstrate the robustness and usefulness of our approach in an example AR application for selecting and inspecting world-stabilized virtual objects."
                },
                {
                    "title": "Image quality assessment: from error visibility to structural similarity",
                    "abstract": "Objective methods for assessing perceptual image quality traditionally attempted to quantify the visibility of errors (differences) between a distorted image and a reference image using a variety of known properties of the human visual system. Under the assumption that human visual perception is highly adapted for extracting structural information from a scene, we introduce an alternative complementary framework for quality assessment based on the degradation of structural information. As a specific example of this concept, we develop a structural similarity index and demonstrate its promise through a set of intuitive examples, as well as comparison to both subjective ratings and state-of-the-art objective methods on a database of images compressed with JPEG and JPEG2000. A MATLAB implementation of the proposed algorithm is available online at http://www.cns.nyu.edu//spl sim/lcv/ssim/."
                },
                {
                    "title": "Optimum design of chamfer distance transforms",
                    "abstract": "The distance transform has found many applications in image analysis. Chamfer distance transforms are a class of discrete algorithms that offer a good approximation to the desired Euclidean distance transform at a lower computational cost. They can also give integer-valued distances that are more suitable for several digital image processing tasks. The local distances used to compute a chamfer distance transform are selected to minimize an approximation error. A new geometric approach is developed to find optimal local distances. This new approach is easier to visualize than the approaches found in previous work, and can be easily extended to chamfer metrics that use large neighborhoods. A new concept of critical local distances is presented which reduces the computational complexity of the chamfer distance transform without increasing the maximum approximation error."
                },
                {
                    "title": "Hand and Mind: What Gestures Reveal About Thought.",
                    "abstract": "The argument of this original and difficult book is that \u201cgestures are an integral part of language as much as are words, phrases and sentences-gestures and language are one system\u201d (p. 2) . Gestures are instantaneous, imagistic, analog, holistic expressions of the same thought that speech renders in hierarchical, linear, digital, analytic form. David McNeill credits Adam Kendon (1972, 1980) with discovering the link between, and essential unity of, speech sounds and gestural movements; his own work elaborates this insight at the higher linguistic levels of semantics and pragmatics. The topic of the book, then, is gestures that accompany speech, the left-hand end of what McNeill calls \u201c K e n h i \u2019 s coiitiiiiiiim: Gesticulation + Language-like gestures + Pantomimes 3 Emblems + Sign languages\u201d (p. 37). The continuum ranges from the informal, spontaneous, idiosyncratic movements of the hands and arms that often accompany speech, to the socially-regulated, standardized, linguistic forms of a sign language, with its arbitrary (non-iconic) lexicon. Between these poles the obligatory presence of speech declines and the linguistic properties of gestures increase. \u201cLanguage-like gestures\u201d are grammatically integrated into an utterance, as when a speaker, asked about the weather on his vacation, replies: \u201cWell, it was [oscillating hand gesture]\u201d, where the \u201cso-so\u201d gesture replaces an adjectival predicate. \u201cPantomime\u201d conveys its full meaning in silence or, at most, with inarticulate onomatopoeia; also, in pantomime, sequences of gestures can form a unit, as they can in a sign language, but cannot in gesticulation. \u201cEmblems\u201d conform to standards of wellformedness, a language-like property that gesticulation and pantomime lack: in England, the palm-front V-sign is Churchill\u2019s \u201cVictory!\u201d, the palm-back V-sign is a sexual insult. (For an amusing cross-class confusion in emblem dialects, see Collett, Marsh, and O\u2019Shaughnessy, 1979, p. 229, where Margaret Thatcher appears in an Associated Press Photo, making the palm-back V-sign at a moment of electoral triumph.) The contrast between the two ends of Kendon\u2019s continuum, between spontaneous gesture and conventional sign, epitomizes McNeill\u2019s notion of the process by which an utterance evolves in a speaker\u2019s mind. Spontaneous gesture reveals the primitive stage of an utterance, global, unsegmented, non-hierarchical, from which its conventional representation in speech unfolds: hierarchical, segmented, linear. The inner symbols of the primitive stage are private, idiosyncratic, closed to social influence; the end stage is public, grammatical, socially regulated. McNeill supposes that the primitive"
                },
                {
                    "title": "Arabic Sign Language Recognition using Faster R-CNN",
                    "abstract": "Deafness does not restrict its negative effect on the person\u2019s hearing, but rather on all aspect of their daily life. Moreover, hearing people aggravated the issue through their reluctance to learn sign language. This resulted in a constant need for human translators to assist deaf person which represents a real obstacle for their social life. Therefore, automatic sign language translation emerged as an urgent need for the community. The availability and the widespread use of mobile phones equipped with digital cameras promoted the design of image-based Arabic Sign Language (ArSL) recognition systems. In this work, we introduce a new ArSL recognition system that is able to localize and recognize the alphabet of the Arabic sign language using a Faster Region-based Convolutional Neural Network (R-CNN). Specifically, faster R-CNN is designed to extract and map the image features, and learn the position of the hand in a given image. Additionally, the proposed approach alleviates both challenges; the choice of the relevant features used to encode the sign visual descriptors, and the segmentation task intended to determine the hand region. For the implementation and the assessment of the proposed Faster R-CNN based sign recognition system, we exploited VGG-16 and ResNet-18 models, and we collected a real ArSL image dataset. The proposed approach yielded 93% accuracy and confirmed the robustness of the proposed model against drastic background variations in the captured scenes."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Efficient model-based 3D tracking of hand articulations using Kinect",
                    "abstract": "We present a novel solution to the problem of recovering and tracking the 3D position, orientation and full articulation of a human hand from markerless visual observations obtained by a Kinect sensor. We treat this as an optimization problem, seeking for the hand model parameters that minimize the discrepancy between the appearance and 3D structure of hypothesized instances of a hand model and actual hand observations. This optimization problem is effectively solved using a variant of Particle Swarm Optimization (PSO). The proposed method does not require special markers and/or a complex image acquisition setup. Being model based, it provides continuous solutions to the problem of tracking hand articulations. Extensive experiments with a prototype GPU-based implementation of the proposed method demonstrate that accurate and robust 3D tracking of hand articulations can be achieved in near real-time (15Hz)."
                },
                {
                    "title": "Drawing on Air: Input Techniques for Controlled 3D Line Illustration",
                    "abstract": "We present drawing on air, a haptic-aided input technique for drawing controlled 3D curves through space. Drawing on air addresses a control problem with current 3D modeling approaches based on sweeping movement of the hands through the air. Although artists praise the immediacy and intuitiveness of these systems, a lack of control makes it nearly impossible to create 3D forms beyond quick design sketches or gesture drawings. Drawing on air introduces two new strategies for more controlled 3D drawing: one-handed drag drawing and two-handed tape drawing. Both approaches have advantages for drawing certain types of curves. We describe a tangent preserving method for transitioning between the two techniques while drawing. Haptic-aided redrawing and line weight adjustment while drawing are also supported in both approaches. In a quantitative user study evaluation by illustrators, the one and two-handed techniques performed at roughly the same level and both significantly outperformed freehand drawing and freehand drawing augmented with a haptic friction effect. We present the design and results of this experiment, as well as user feedback from artists and 3D models created in a style of line illustration for challenging artistic and scientific subjects."
                },
                {
                    "title": "google,\u6211,\u8428\u5a1c",
                    "abstract": "IT Status \"Password doesn't match\" error. 4Help is aware that after you change your VT Google password, you will be unable to log on to VT Google Apps services including Mail, Drive, Groups, etc. 4Help is notifying the appropriate people. 12:00 Noon: Engineers have found a backlog on Google password replication. Once the backlog clears you should be able to log on with your changed password that you set earlier. You may be able to log on with your old VT Google password until the system catches up and syncs the new password. Service Degraded Service Degraded [Resolved] Created: Thu, 04/14/2016 11:20am Resolved: Fri, 04/15/2016 1:16pm Duration: 1 day 1 hour 56 min 1734 Views Source URL: https://computing.vt.edu/content/google-0"
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI",
                "cs.GR"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan we generate sketches from hand motions without additional sensors or markers?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem has significant implications for the research community, particularly in the fields of computer vision, human-computer interaction, and augmented/virtual reality. By enabling sketch generation from hand motions captured by standard cameras, we can enhance accessibility and convenience for users, making creative expression more spontaneous and less reliant on expensive equipment. This advancement could lead to practical applications in education, design, and communication, fostering new ways to visualize ideas and concepts. Furthermore, addressing this question could advance knowledge in generative models and motion analysis, potentially inspiring future research on improving user interfaces and interaction methods in AR/VR environments.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent noise and inaccuracies in hand tracking, which can severely distort the intended sketches. Naive approaches that rely solely on hand tracking algorithms may fail due to user variability in drawing ability, physical fatigue, and the limitations of current tracking technologies. Additionally, the task requires a sophisticated understanding of shape and object priors to accurately reconstruct sketches from distorted input, which adds a layer of complexity. Overcoming these technical obstacles necessitates the development of robust models capable of filtering out noise and interpreting deformed motion cues effectively.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on hand tracking and sketch generation separately, often overlooking the integration of these two domains in a way that addresses the challenges posed by noisy input. Existing solutions typically rely on expensive hardware or specific conditions that limit their practicality. Barriers such as the lack of effective self-supervised training methods and the absence of datasets tailored for air drawing have also hindered progress. Our approach differs by utilizing controllable image Diffusion Models and a novel augmentation-based training procedure, which allows for the reconstruction of clean sketches from severely distorted inputs, thus improving upon prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using controllable image Diffusion Models to generate sketches from hand motion videos. We will construct two air drawing datasets for evaluation and employ an augmentation-based, self-supervised training procedure. The metrics for evaluation will include the fidelity of the generated sketches to the user's intent and the aesthetic quality of the sketches. We expect our approach to yield robust sketch generation capabilities"
            }
        },
        "author_data": {
            "e79b3ff0-7e6f-488d-84e3-32cd84b22ad8": {
                "pk": "e79b3ff0-7e6f-488d-84e3-32cd84b22ad8",
                "name": "Hui Xian Grace Lim",
                "collaborators": [
                    "Andrey Vayner",
                    "Nadia L. Zakamska",
                    "Yuzo Ishikawa",
                    "Swetha Sankar",
                    "Dominika Wylezalek",
                    "David S. N. Rupke",
                    "Sylvain Veilleux",
                    "Caroline Bertemes",
                    "Hsiao-Wen Chen",
                    "Nadiia Diachenko"
                ],
                "domain": [
                    "View Synthesis",
                    "Quasar Feedback",
                    "Astrophysics",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "FSViewFusion: Few-Shots View Generation of Novel Objects",
                        "abstract": "Novel view synthesis has observed tremendous developments since the arrival of NeRFs. However, Nerf models overfit on a single scene, lacking generalization to out of distribution objects. Recently, diffusion models have exhibited remarkable performance on introducing generalization in view synthesis. Inspired by these advancements, we explore the capabilities of a pretrained stable diffusion model for view synthesis without explicit 3D priors. Specifically, we base our method on a personalized text to image model, Dreambooth, given its strong ability to adapt to specific novel objects with a few shots. Our research reveals two interesting findings. First, we observe that Dreambooth can learn the high level concept of a view, compared to arguably more complex strategies which involve finetuning diffusions on large amounts of multi-view data. Second, we establish that the concept of a view can be disentangled and transferred to a novel object irrespective of the original object's identify from which the views are learnt. Motivated by this, we introduce a learning strategy, FSViewFusion, which inherits a specific view through only one image sample of a single scene, and transfers the knowledge to a novel object, learnt from few shots, using low rank adapters. Through extensive experiments we demonstrate that our method, albeit simple, is efficient in generating reliable view samples for in the wild images. Code and models will be released."
                    },
                    {
                        "title": "First results from the JWST Early Release Science Program Q3D: Ionization cone, clumpy star formation and shocks in a $z=3$ extremely red quasar host",
                        "abstract": "Massive galaxies formed most actively at redshifts $z=1-3$ during the period known as `cosmic noon.' Here we present an emission-line study of an extremely red quasar SDSSJ165202.64+172852.3 host galaxy at $z=2.94$, based on observations with the Near Infrared Spectrograph (NIRSpec) integral field unit (IFU) on board JWST. We use standard emission-line diagnostic ratios to map the sources of gas ionization across the host and a swarm of companion galaxies. The quasar dominates the photoionization, but we also discover shock-excited regions orthogonal to the ionization cone and the quasar-driven outflow. These shocks could be merger-induced or -- more likely, given the presence of a powerful galactic-scale quasar outflow -- these are signatures of wide-angle outflows that can reach parts of the galaxy that are not directly illuminated by the quasar. Finally, the kinematically narrow emission associated with the host galaxy presents as a collection of 1 kpc-scale clumps forming stars at a rate of at least 200 $M_{\\odot}$ yr$^{-1}$. The ISM within these clumps shows high electron densities, reaching up to 3,000 cm$^{-3}$ with metallicities ranging from half to a third solar with a positive metallicity gradient and V band extinctions up to 3 magnitudes. The star formation conditions are far more extreme in these regions than in local star-forming galaxies but consistent with that of massive galaxies at cosmic noon. JWST observations reveal an archetypical rapidly forming massive galaxy undergoing a merger, a clumpy starburst, an episode of obscured near-Eddington quasar activity, and an extremely powerful quasar outflow simultaneously."
                    },
                    {
                        "title": "First results from the JWST Early Release Science Program Q3D: The Warm Ionized Gas Outflow in z ~ 1.6 Quasar XID 2028 and its Impact on the Host Galaxy",
                        "abstract": "Quasar feedback may regulate the growth of supermassive black holes, quench coeval star formation, and impact galaxy morphology and the circumgalactic medium. However, direct evidence for quasar feedback in action at the epoch of peak black hole accretion at z ~ 2 remains elusive. A good case in point is the z = 1.6 quasar WISEA J100211.29+013706.7 (XID 2028) where past analyses of the same ground-based data have come to different conclusions. Here we revisit this object with the integral field unit of the Near Infrared Spectrograph (NIRSpec) on board the James Webb Space Telescope (JWST) as part of Early Release Science program Q3D. The excellent angular resolution and sensitivity of the JWST data reveal new morphological and kinematic sub-structures in the outflowing gas plume. An analysis of the emission line ratios indicates that photoionization by the central quasar dominates the ionization state of the gas with no obvious sign for a major contribution from hot young stars anywhere in the host galaxy. Rest-frame near-ultraviolet emission aligned along the wide-angle cone of outflowing gas is interpreted as a scattering cone. The outflow has cleared a channel in the dusty host galaxy through which some of the quasar ionizing radiation is able to escape and heat the surrounding interstellar and circumgalactic media. The warm ionized outflow is not powerful enough to impact the host galaxy via mechanical feedback, but radiative feedback by the AGN, aided by the outflow, may help explain the unusually small molecular gas mass fraction in the galaxy host."
                    },
                    {
                        "title": "First results from the JWST Early Release Science Program Q3D: Powerful quasar-driven galactic scale outflow at $z=3$",
                        "abstract": "Quasar-driven galactic outflows are a major driver of the evolution of massive galaxies. We report observations of a powerful galactic-scale outflow in a $z=3$ extremely red, intrinsically luminous ($L_{\\rm bol}\\simeq 5\\times 10^{47}$erg s$^{-1}$) quasar SDSSJ1652+1728 with the Near Infrared Spectrograph (NIRSpec) on board JWST. We analyze the kinematics of rest-frame optical emission lines and identify the quasar-driven outflow extending out to $\\sim 10$ kpc from the quasar with a velocity offset of ($v_{r}=\\pm 500$ km s$^{-1}$) and high velocity dispersion (FWHM$=700-2400$ km s$^{-1}$). Due to JWST's unprecedented surface brightness sensitivity in the near-infrared -- we unambiguously show that the powerful high velocity outflow in an extremely red quasar (ERQ) encompasses a large swath of the host galaxy's interstellar medium (ISM). Using the kinematics and dynamics of optical emission lines, we estimate the mass outflow rate -- in the warm ionized phase alone -- to be at least $2300\\pm1400$ $M_{\\odot}$ yr$^{-1}$. We measure a momentum flux ratio between the outflow and the quasar accretion disk of $\\sim$1 on kpc scale, indicating that the outflow was likely driven in a relatively high ($>10^{23}$cm$^{-2}$) column density environment through radiation pressure on dust grains. We find a coupling efficiency between the bolometric luminosity of the quasar and the outflow of 0.1$\\%$, matching the theoretical prediction of the minimum coupling efficiency necessary for negative quasar feedback. The outflow has sufficient energetics to drive the observed turbulence seen in shocked regions of the quasar host galaxy, likely directly responsible for prolonging the time it takes for gas to cool efficiently."
                    },
                    {
                        "title": "First results from the JWST Early Release Science Program Q3D: The Fast Outflow in a Red Quasar at z=0.44",
                        "abstract": "Quasar feedback may play a key role in the evolution of massive galaxies. The dust-reddened quasar, F2M110648.35$+$480712 at $z = 0.4352$ is one of the few cases at its redshift that exhibits powerful quasar feedback through bipolar outflows. Our new observation with the integral field unit mode of Near-infrared Spectrograph onboard JWST opens a new window to examine this spectacular outflow through Pa$\\alpha$ emission line with $\\sim$3$\\times$ better spatial resolution than previous work. The morphology and kinematics of the Pa$\\alpha$ nebula confirm the existence of a bipolar outflow extending on a scale of $\\sim$17$\\times$14 kpc and with a velocity reaching $\\sim$1100 km s$^{-1}$. The higher spatial resolution of our new observation leads to more reliable measurements of outflow kinematics. Considering only the spatially resolved outflow and assuming an electron density of 100 cm$^{-2}$, the mass, momentum and kinetic energy outflow rates are $\\sim$50-210 M$_{\\odot}$ yr$^{-1}$, $\\sim$0.3-1.7$\\times$10$^{36}$ dynes ($\\sim$14-78\\% of the quasar photon momentum flux) and $\\sim$0.16-1.27$\\times$10$^{44}$ erg s$^{-1}$ ($\\sim$0.02-0.20\\% of the quasar bolometric luminosity), respectively. The local instantaneous outflow rates generally decrease radially. We infer that the quasar is powerful enough to drive the outflow, while stellar processes cannot be overlooked as a contributing energy source. The mass outflow rate is $\\sim$0.4-1.5 times the star formation rate, and the ratio of kinetic energy outflow rate to the quasar bolometric luminosity is comparable to the minimum value required for negative quasar feedback in simulations. This outflow may help regulate the star formation activity within the system to some extent."
                    }
                ]
            },
            "daf6c121-c7d6-4f58-b205-9b9dadf9538e": {
                "pk": "daf6c121-c7d6-4f58-b205-9b9dadf9538e",
                "name": "Xuanming Cui",
                "collaborators": [
                    "Alejandro Aparcedo",
                    "Young Kyun Jang",
                    "Ser-Nam Lim"
                ],
                "domain": [
                    "Multimodal Models",
                    "Adversarial Attacks",
                    "Robustness",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "On the Robustness of Large Multimodal Models Against Image Adversarial Attacks",
                        "abstract": "Recent advances in instruction tuning have led to the development of State-of-the-Art Large Multimodal Models (LMMs). Given the novelty of these models, the impact of visual adversarial attacks on LMMs has not been thoroughly examined. We conduct a comprehensive study of the robustness of various LMMs against different adversarial attacks, evaluated across tasks including image classification, image captioning, and Visual Question Answer (VQA). We find that in general LMMs are not robust to visual adversarial inputs. However, our findings suggest that context provided to the model via prompts, such as questions in a QA pair helps to mitigate the effects of visual adversarial inputs. Notably, the LMMs evaluated demonstrated remarkable resilience to such attacks on the ScienceQA task with only an 8.10% drop in performance compared to their visual counterparts which dropped 99.73%. We also propose a new approach to real-world image classification which we term query decomposition. By incorporating existence queries into our input prompt we observe diminished attack effectiveness and improvements in image classification accuracy. This research highlights a previously under-explored facet of LMM robustness and sets the stage for future work aimed at strengthening the resilience of multimodal systems in adversarial environments."
                    }
                ]
            },
            "d639efaa-a366-49e0-9fa4-e39abb0bd5e3": {
                "pk": "d639efaa-a366-49e0-9fa4-e39abb0bd5e3",
                "name": "Yogesh S Rawat",
                "collaborators": [
                    "Mubarak Shah",
                    "Shruti Vyas",
                    "Kevin Duarte",
                    "Vibhav Vineet",
                    "Aayush J Rana",
                    "Ugur Demir",
                    "Shresth Grover",
                    "Shahzad Ahmad",
                    "Sukalpa Chanda",
                    "Bruce McIntosh"
                ],
                "domain": [
                    "Action Recognition",
                    "Video Analysis",
                    "Deep Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "TinyVIRAT: Low-resolution Video Action Recognition",
                        "abstract": "The existing research in action recognition is mostly focused on high-quality videos where the action is distinctly visible. In real-world surveillance environments, the actions in videos are captured at a wide range of resolutions. Most activities occur at a distance with a small resolution and recognizing such activities is a challenging problem. In this work, we focus on recognizing tiny actions in videos. We introduce a benchmark dataset, TinyVIRAT, which contains natural low-resolution activities. The actions in TinyVIRAT videos have multiple labels and they are extracted from surveillance videos which makes them realistic and more challenging. We propose a novel method for recognizing tiny actions in videos which utilizes a progressive generative approach to improve the quality of low-resolution actions. The proposed method also consists of a weakly trained attention mechanism which helps in focusing on the activity regions in the video. We perform extensive experiments to benchmark the proposed TinyVIRAT dataset and observe that the proposed method significantly improves the action recognition performance over baselines. We also evaluate the proposed approach on synthetically resized action recognition datasets and achieve state-of-the-art results when compared with existing methods. The dataset and code is publicly available at https://github.com/UgurDemir/Tiny-VIRAT."
                    },
                    {
                        "title": "View-invariant action recognition",
                        "abstract": "Human action recognition is an important problem in computer vision. It has a wide range of applications in surveillance, human-computer interaction, augmented reality, video indexing, and retrieval. The varying pattern of spatio-temporal appearance generated by human action is key for identifying the performed action. We have seen a lot of research exploring this dynamics of spatio-temporal appearance for learning a visual representation of human actions. However, most of the research in action recognition is focused on some common viewpoints, and these approaches do not perform well when there is a change in viewpoint. Human actions are performed in a 3-dimensional environment and are projected to a 2-dimensional space when captured as a video from a given viewpoint. Therefore, an action will have a different spatio-temporal appearance from different viewpoints. The research in view-invariant action recognition addresses this problem and focuses on recognizing human actions from unseen viewpoints."
                    },
                    {
                        "title": "VideoCapsuleNet: A Simplified Network for Action Detection",
                        "abstract": "The recent advances in Deep Convolutional Neural Networks (DCNNs) have shown extremely good results for video human action classification, however, action detection is still a challenging problem. The current action detection approaches follow a complex pipeline which involves multiple tasks such as tube proposals, optical flow, and tube classification. In this work, we present a more elegant solution for action detection based on the recently developed capsule network. We propose a 3D capsule network for videos, called VideoCapsuleNet: a unified network for action detection which can jointly perform pixel-wise action segmentation along with action classification. The proposed network is a generalization of capsule network from 2D to 3D, which takes a sequence of video frames as input. The 3D generalization drastically increases the number of capsules in the network, making capsule routing computationally expensive. We introduce capsule-pooling in the convolutional capsule layer to address this issue which makes the voting algorithm tractable. The routing-by-agreement in the network inherently models the action representations and various action characteristics are captured by the predicted capsules. This inspired us to utilize the capsules for action localization and the class-specific capsules predicted by the network are used to determine a pixel-wise localization of actions. The localization is further improved by parameterized skip connections with the convolutional capsule layers and the network is trained end-to-end with a classification as well as localization loss. The proposed network achieves sate-of-the-art performance on multiple action detection datasets including UCF-Sports, J-HMDB, and UCF-101 (24 classes) with an impressive ~20% improvement on UCF-101 and ~15% improvement on J-HMDB in terms of v-mAP scores."
                    },
                    {
                        "title": "Navigating Hallucinations for Reasoning of Unintentional Activities",
                        "abstract": "In this work we present a novel task of understanding unintentional human activities in videos. We formalize this problem as a reasoning task under zero-shot scenario, where given a video of an unintentional activity we want to know why it transitioned from intentional to unintentional. We first evaluate the effectiveness of current state-of-the-art Large Multimodal Models on this reasoning task and observe that they suffer from hallucination. We further propose a novel prompting technique,termed as Dream of Thoughts (DoT), which allows the model to navigate through hallucinated thoughts to achieve better reasoning. To evaluate the performance on this task, we also introduce three different specialized metrics designed to quantify the models reasoning capability. We perform our experiments on two different datasets, OOPs and UCF-Crimes, and our findings show that DOT prompting technique is able to outperform standard prompting, while minimizing hallucinations."
                    },
                    {
                        "title": "Time-Aware and View-Aware Video Rendering for Unsupervised Representation Learning",
                        "abstract": "The recent success in deep learning has lead to various effective representation learning methods for videos. However, the current approaches for video representation require large amount of human labeled datasets for effective learning. We present an unsupervised representation learning framework to encode scene dynamics in videos captured from multiple viewpoints. The proposed framework has two main components: Representation Learning Network (RL-NET), which learns a representation with the help of Blending Network (BL-NET), and Video Rendering Network (VR-NET), which is used for video synthesis. The framework takes as input video clips from different viewpoints and time, learns an internal representation and uses this representation to render a video clip from an arbitrary given viewpoint and time. The ability of the proposed network to render video frames from arbitrary viewpoints and time enable it to learn a meaningful and robust representation of the scene dynamics. We demonstrate the effectiveness of the proposed method in rendering view-aware as well as time-aware video clips on two different real-world datasets including UCF-101 and NTU-RGB+D. To further validate the effectiveness of the learned representation, we use it for the task of view-invariant activity classification where we observe a significant improvement (~26%) in the performance on NTU-RGB+D dataset compared to the existing state-of-the art methods."
                    },
                    {
                        "title": "CapsuleVOS: Semi-Supervised Video Object Segmentation Using Capsule Routing",
                        "abstract": "In this work we propose a capsule-based approach for semi-supervised video object segmentation. Current video object segmentation methods are frame-based and often require optical flow to capture temporal consistency across frames which can be difficult to compute. To this end, we propose a video based capsule network, CapsuleVOS, which can segment several frames at once conditioned on a reference frame and segmentation mask. This conditioning is performed through a novel routing algorithm for attention-based efficient capsule selection. We address two challenging issues in video object segmentation: 1) segmentation of small objects and 2) occlusion of objects across time. The issue of segmenting small objects is addressed with a zooming module which allows the network to process small spatial regions of the video. Apart from this, the framework utilizes a novel memory module based on recurrent networks which helps in tracking objects when they move out of frame or are occluded. The network is trained end-to-end and we demonstrate its effectiveness on two benchmark video object segmentation datasets; it outperforms current offline approaches on the Youtube-VOS dataset while having a run-time that is almost twice as fast as competing methods. The code is publicly available at https://github.com/KevinDuarte/CapsuleVOS."
                    },
                    {
                        "title": "We don't Need Thousand Proposals$\\colon$ Single Shot Actor-Action Detection in Videos",
                        "abstract": "We propose SSA2D, a simple yet effective end-to-end deep network for actor-action detection in videos. The existing methods take a top-down approach based on region-proposals (RPN), where the action is estimated based on the detected proposals followed by post-processing such as non-maximal suppression. While effective in terms of performance, these methods pose limitations in scalability for dense video scenes with a high memory requirement for thousands of proposals. We propose to solve this problem from a different perspective where we don't need any proposals. SSA2D is a unified network, which performs pixel level joint actor-action detection in a single-shot, where every pixel of the detected actor is assigned an action label. SSA2D has two main advantages: 1) It is a fully convolutional network which does not require any proposals and post-processing making it memory as well as time efficient, 2) It is easily scalable to dense video scenes as its memory requirement is independent of the number of actors present in the scene. We evaluate the proposed method on the Actor-Action dataset (A2D) and Video Object Relation (VidOR) dataset, demonstrating its effectiveness in multiple actors and action detection in a video. SSA2D is 11x faster during inference with comparable (sometimes better) performance and fewer network parameters when compared with the prior works."
                    },
                    {
                        "title": "EZ-CLIP: Efficient Zeroshot Video Action Recognition",
                        "abstract": "Recent advancements in large-scale pre-training of visual-language models on paired image-text data have demonstrated impressive generalization capabilities for zero-shot tasks. Building on this success, efforts have been made to adapt these image-based visual-language models, such as CLIP, for videos extending their zero-shot capabilities to the video domain. While these adaptations have shown promising results, they come at a significant computational cost and struggle with effectively modeling the crucial temporal aspects inherent to the video domain. In this study, we present EZ-CLIP, a simple and efficient adaptation of CLIP that addresses these challenges. EZ-CLIP leverages temporal visual prompting for seamless temporal adaptation, requiring no fundamental alterations to the core CLIP architecture while preserving its remarkable generalization abilities. Moreover, we introduce a novel learning objective that guides the temporal visual prompts to focus on capturing motion, thereby enhancing its learning capabilities from video data. We conducted extensive experiments on five different benchmark datasets, thoroughly evaluating EZ-CLIP for zero-shot learning and base-to-novel video action recognition, and also demonstrating its potential for few-shot generalization.Impressively, with a mere 5.2 million learnable parameters (as opposed to the 71.1 million in the prior best model), EZ-CLIP can be efficiently trained on a single GPU, outperforming existing approaches in several evaluations."
                    },
                    {
                        "title": "Multi-modal Capsule Routing for Actor and Action Video Segmentation Conditioned on Natural Language Queries",
                        "abstract": "In this paper, we propose an end-to-end capsule network for pixel level localization of actors and actions present in a video. The localization is performed based on a natural language query through which an actor and action are specified. We propose to encode both the video as well as textual input in the form of capsules, which provide more effective representation in comparison with standard convolution based features. We introduce a novel capsule based attention mechanism for fusion of video and text capsules for text selected video segmentation. The attention mechanism is performed via joint EM routing over video and text capsules for text selected actor and action localization. The existing works on actor-action localization are mainly focused on localization in a single frame instead of the full video. Different from existing works, we propose to perform the localization on all frames of the video. To validate the potential of the proposed network for actor and action localization on all the frames of a video, we extend an existing actor-action dataset (A2D) with annotations for all the frames. The experimental evaluation demonstrates the effectiveness of the proposed capsule network for text selective actor and action localization in videos, and it also improves upon the performance of the existing state-of-the art works on single frame-based localization."
                    },
                    {
                        "title": "Unsupervised Discriminative Embedding for Sub-Action Learning in Complex Activities",
                        "abstract": "Action recognition and detection in the context of long untrimmed video sequences has seen an increased attention from the research community. However, annotation of complex activities is usually time consuming and challenging in practice. Therefore, recent works started to tackle the problem of unsupervised learning of sub-actions in complex activities. This paper proposes a novel approach for unsupervised sub-action learning in complex activities. The proposed method maps both visual and temporal representations to a latent space where the sub-actions are learnt discriminatively in an end-to-end fashion. To this end, we propose to learn sub-actions as latent concepts and a novel discriminative latent concept learning (DLCL) module aids in learning sub-actions. The proposed DLCL module lends on the idea of latent concepts to learn compact representations in the latent embedding space in an unsupervised way. The result is a set of latent vectors that can be interpreted as cluster centers in the embedding space. The latent space itself is formed by a joint visual and temporal embedding capturing the visual similarity and temporal ordering of the data. Our joint learning with discriminative latent concept module is novel which eliminates the need for explicit clustering. We validate our approach on three benchmark datasets and show that the proposed combination of visual-temporal embedding and discriminative latent concepts allow to learn robust action representations in an unsupervised setting."
                    },
                    {
                        "title": "LARNet: Latent Action Representation for Human Action Synthesis",
                        "abstract": "We present LARNet, a novel end-to-end approach for generating human action videos. A joint generative modeling of appearance and dynamics to synthesize a video is very challenging and therefore recent works in video synthesis have proposed to decompose these two factors. However, these methods require a driving video to model the video dynamics. In this work, we propose a generative approach instead, which explicitly learns action dynamics in latent space avoiding the need of a driving video during inference. The generated action dynamics is integrated with the appearance using a recurrent hierarchical structure which induces motion at different scales to focus on both coarse as well as fine level action details. In addition, we propose a novel mix-adversarial loss function which aims at improving the temporal coherency of synthesized videos. We evaluate the proposed approach on four real-world human action datasets demonstrating the effectiveness of the proposed approach in generating human actions. Code available at https://github.com/aayushjr/larnet."
                    },
                    {
                        "title": "GeoMeter: Probing Depth and Height Perception of Large Visual-Language Models",
                        "abstract": "Geometric understanding is crucial for navigating and interacting with our environment. While large Vision Language Models (VLMs) demonstrate impressive capabilities, deploying them in real-world scenarios necessitates a comparable geometric understanding in visual perception. In this work, we focus on the geometric comprehension of these models; specifically targeting the depths and heights of objects within a scene. Our observations reveal that, although VLMs excel in basic geometric properties perception such as shape and size, they encounter significant challenges in reasoning about the depth and height of objects. To address this, we introduce GeoMeter, a suite of benchmark datasets encompassing Synthetic 2D, Synthetic 3D, and Real-World scenarios to rigorously evaluate these aspects. We benchmark 17 state-of-the-art VLMs using these datasets and find that they consistently struggle with both depth and height perception. Our key insights include detailed analyses of the shortcomings in depth and height reasoning capabilities of VLMs and the inherent bias present in these models. This study aims to pave the way for the development of VLMs with enhanced geometric understanding, crucial for real-world applications."
                    },
                    {
                        "title": "Pose-guided Generative Adversarial Net for Novel View Action Synthesis",
                        "abstract": "We focus on the problem of novel-view human action synthesis. Given an action video, the goal is to generate the same action from an unseen viewpoint. Naturally, novel view video synthesis is more challenging than image synthesis. It requires the synthesis of a sequence of realistic frames with temporal coherency. Besides, transferring the different actions to a novel target view requires awareness of action category and viewpoint change simultaneously. To address these challenges, we propose a novel framework named Pose-guided Action Separable Generative Adversarial Net (PAS-GAN), which utilizes pose to alleviate the difficulty of this task. First, we propose a recurrent pose-transformation module which transforms actions from the source view to the target view and generates novel view pose sequence in 2D coordinate space. Second, a well-transformed pose sequence enables us to separatethe action and background in the target view. We employ a novel local-global spatial transformation module to effectively generate sequential video features in the target view using these action and background features. Finally, the generated video features are used to synthesize human action with the help of a 3D decoder. Moreover, to focus on dynamic action in the video, we propose a novel multi-scale action-separable loss which further improves the video quality. We conduct extensive experiments on two large-scale multi-view human action datasets, NTU-RGBD and PKU-MMD, demonstrating the effectiveness of PAS-GAN which outperforms existing approaches."
                    }
                ]
            },
            "d86b6932-009b-40e5-a2e1-df4694078979": {
                "pk": "d86b6932-009b-40e5-a2e1-df4694078979",
                "name": "Ser-Nam Lim",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2407.02680": {
        "paper_data": {
            "title": "KGym: A Platform and Dataset to Benchmark Large Language Models on Linux Kernel Crash Resolution",
            "url": "http://arxiv.org/abs/2407.02680v3",
            "arxiv_id": "2407.02680",
            "authors": [
                "Alex Mathai",
                "Chenxi Huang",
                "Petros Maniatis",
                "Aleksandr Nogikh",
                "Franjo Ivancic",
                "Junfeng Yang",
                "Baishakhi Ray"
            ],
            "abstract": "Large Language Models (LLMs) are consistently improving at increasingly realistic software engineering (SE) tasks. In real-world software stacks, significant SE effort is spent developing foundational system software like the Linux kernel. Unlike application-level software, a systems codebase like Linux is multilingual (low-level C/Assembly/Bash/Rust); gigantic (>20 million lines); critical (impacting billions of devices worldwide), and highly concurrent (involving complex multi-threading). To evaluate if ML models are useful while developing such large-scale systems-level software, we introduce kGym (a platform) and kBench (a dataset). The kGym platform provides a SE environment for large-scale experiments on the Linux kernel, including compiling and running kernels in parallel across several virtual machines, detecting operations and crashes, inspecting logs, and querying and patching the code base. We use kGym to facilitate evaluation on kBench, a crash resolution benchmark drawn from real-world Linux kernel bugs. An example bug in kBench contains crashing stack traces, a bug-reproducer file, a developer-written fix, and other associated data. To understand current performance, we conduct baseline experiments by prompting LLMs to resolve Linux kernel crashes. Our initial evaluations reveal that the best performing LLM achieves 0.72% and 5.38% in the unassisted and assisted (i.e., buggy files disclosed to the model) settings, respectively. These results highlight the need for further research to enhance model performance in SE tasks. Improving performance on kBench requires models to master new learning skills, including understanding the cause of crashes and repairing faults, writing memory-safe and hardware-aware code, and understanding concurrency. As a result, this work opens up multiple avenues of research at the intersection of machine learning and systems software.",
            "introduction": "   1 Introduction  In recent years, there has been significant progress in using code LLMs (like CodeWhisperer (Amazon, 2023) and CoPilot\u00a0(GitHub, 2021)) in all stages of the software cycle, including development, debugging, and testing. Despite being trained on large and complex open-source projects, LLMs are often benchmarked on test sets like EvalPlus (Liu et\u00a0al., 2023a), HumanEval (Chen et\u00a0al., 2021), and APPS (Hendrycks et\u00a0al., 2021) which are about111https://evalplus.github.io/leaderboard.html to get saturated\u00a0(Ott et\u00a0al., 2022) . While useful, these benchmarks represent \u201cgreen-field\u201d SE by isolating coding to the task of solving programming puzzles. Unfortunately, such puzzles do not reflect the intricacies involved in everyday reasoning and solving of complex bugs in production-ready software.   Hence, newly introduced benchmarks (like SWE-Bench\u00a0(Jimenez et\u00a0al., 2024)) try to bridge the gap between existing tasks and realistic SE in \u201cbrown-field\u201d environments, where LLM assistants edit, debug, and test production-ready software. Such benchmarks capture a more realistic SE setting: given a software repository, a natural-language (NL) description of a problem or feature request, and a set of held-out executable test cases, edit the repository so that the test cases pass.   Our work moves one step further along the same trajectory, by introducing a drastically more challenging SE benchmark for future assistants. Specifically, we target crash resolution in the Linux kernel (Lin, ): given a state of the Linux codebase, a crash report, and the crash-inducing input, the target is to repair the codebase such that the input no longer triggers a crash. To that effect, we build an execution environment, kGym, and corresponding benchmark, kBench.   Why Linux? The Linux Kernel spans over 20202020M lines of code spread across 50505050k files. It has been in open-source development for decades and is deployed on billions of devices worldwide, including cloud infrastructures, desktops, and over three billion active Android devices\u00a0(And, ). Although the criticality of Linux itself justifies a benchmark built around it, kBench also tests LLM assistants on new and generalizable SE skills beyond what is available today:     \u2022  Low level: Linux is a systems codebase written in a mixture of C, Assembly (for multiple hardware architectures, like x86, ARM, etc.), Bash, and Rust, sometimes intermingled in the same file (e.g., in Assembly embedded in C). As a result, the implementation must be hardware-aware and memory-safe, in contrast to userspace code (often hardware-agnostic) and code in managed languages such as Python (the runtime abstracts away memory and hardware details).    \u2022  Concurrent: Linux code is highly concurrent and non-deterministic, with many kernel bugs caused by hard-to-reproduce thread interleavings, leading to deadlocks, race conditions, and atomicity violations (\"Heisenbugs\"). To resolve such bugs, the model must be able to learn and reason about the different interleaving schedules across concurrent threads. Moreover, a corresponding benchmark platform must work with flaky test oracles\u2014the bug is sometimes observed, but not always\u2014unlike existing benchmarks, which rely on deterministic oracles.    \u2022  Ambiguous Intent: Unlike application-level SE tasks that start with an NL description of a problem, the root cause in a crash report is often unknown, hard to reproduce, and must be identified before it can be resolved. This makes for a challenging task, both in terms of ambiguity and the underlying dependence on myriad behaviors of the complex kernel.     \u2022  Decentralized",
            "references": [
                {
                    "title": "CrossCodeEval: A Diverse and Multilingual Benchmark for Cross-File Code Completion",
                    "abstract": "Code completion models have made significant progress in recent years, yet current popular evaluation datasets, such as HumanEval and MBPP, predominantly focus on code completion tasks within a single file. This over-simplified setting falls short of representing the real-world software development scenario where repositories span multiple files with numerous cross-file dependencies, and accessing and understanding cross-file context is often required to complete the code correctly. To fill in this gap, we propose CrossCodeEval, a diverse and multilingual code completion benchmark that necessitates an in-depth cross-file contextual understanding to complete the code accurately. CrossCodeEval is built on a diverse set of real-world, open-sourced, permissively-licensed repositories in four popular programming languages: Python, Java, TypeScript, and C#. To create examples that strictly require cross-file context for accurate completion, we propose a straightforward yet efficient static-analysis-based approach to pinpoint the use of cross-file context within the current file. Extensive experiments on state-of-the-art code language models like CodeGen and StarCoder demonstrate that CrossCodeEval is extremely challenging when the relevant cross-file context is absent, and we see clear improvements when adding these context into the prompt. However, despite such improvements, the pinnacle of performance remains notably unattained even with the highest-performing model, indicating that CrossCodeEval is also capable of assessing model's capability in leveraging extensive context to make better code completion. Finally, we benchmarked various methods in retrieving cross-file context, and show that CrossCodeEval can also be used to measure the capability of code retrievers."
                },
                {
                    "title": "MultiPL-E: A Scalable and Polyglot Approach to Benchmarking Neural Code Generation",
                    "abstract": "Large language models have demonstrated the ability to generate both natural language and programming language text. Although contemporary code generation models are trained on corpora with several programming languages, they are tested using benchmarks that are typically monolingual. The most widely used code generation benchmarks only target Python, so there is little quantitative evidence of how code generation models perform on other programming languages. We propose MultiPL-E, a system for translating unit test-driven code generation benchmarks to new languages. We create the first massively multilingual code generation benchmark by using MultiPL-E to translate two popular Python code generation benchmarks to 18 additional programming languages. We use MultiPL-E to extend the HumanEval benchmark (Chen et al., 2021) and MBPP benchmark (Austin et al., 2021) to 18 languages that encompass a range of programming paradigms and popularity. Using these new parallel benchmarks, we evaluate the multi-language performance of three state-of-the-art code generation models: Codex (Chen et al., 2021), CodeGen (Nijkamp et al., 2022) and InCoder (Fried et al., 2022). We find that Codex matches or even exceeds its performance on Python for several other languages. The range of programming languages represented in MultiPL-E allow us to explore the impact of language frequency and language features on model performance. Finally, the MultiPL-E approach of compiling code generation benchmarks to new programming languages is both scalable and extensible, making it straightforward to evaluate new models, benchmarks, and languages."
                },
                {
                    "title": "Large Language Models of Code Fail at Completing Code with Potential Bugs",
                    "abstract": "Large language models of code (Code-LLMs) have recently brought tremendous advances to code completion, a fundamental feature of programming assistance and code intelligence. However, most existing works ignore the possible presence of bugs in the code context for generation, which are inevitable in software development. Therefore, we introduce and study the buggy-code completion problem, inspired by the realistic scenario of real-time code suggestion where the code context contains potential bugs -- anti-patterns that can become bugs in the completed program. To systematically study the task, we introduce two datasets: one with synthetic bugs derived from semantics-altering operator changes (buggy-HumanEval) and one with realistic bugs derived from user submissions to coding problems (buggy-FixEval). We find that the presence of potential bugs significantly degrades the generation performance of the high-performing Code-LLMs. For instance, the passing rates of CODEGEN-2B-MONO on test cases of buggy-HumanEval drop more than 50% given a single potential bug in the context. Finally, we investigate several post-hoc methods for mitigating the adverse effect of potential bugs and find that there remains a significant gap in post-mitigation performance."
                },
                {
                    "title": "A Static Evaluation of Code Completion by Large Language Models",
                    "abstract": "Large language models trained on code have shown great potential to increase productivity of software developers. Several execution-based benchmarks have been proposed to evaluate functional correctness of model-generated code on simple programming problems. Nevertheless, it is expensive to perform the same evaluation on complex real-world projects considering the execution cost. On the other hand, static analysis tools such as linters, which can detect errors without running the program, haven\u2019t been well explored for evaluating code generation models. In this work, we propose a static evaluation framework to quantify static errors in Python code completions, by leveraging Abstract Syntax Trees. Compared with execution-based evaluation, our method is not only more efficient, but also applicable to code in the wild. For experiments, we collect code context from open source repos to generate one million function bodies using public models. Our static analysis reveals that Undefined Name and Unused Variable are the most common errors among others made by language models.Through extensive studies, we also show the impact of sampling temperature, model size, and context on static errors in code completions."
                },
                {
                    "title": "Better Context Makes Better Code Language Models: A Case Study on Function Call Argument Completion",
                    "abstract": "Pretrained code language models have enabled great progress towards program synthesis. However, common approaches only consider in-file local context and thus miss information and constraints imposed by other parts of the codebase and its external dependencies. Existing code completion benchmarks also lack such context. To resolve these restrictions we curate a new dataset of permissively licensed Python packages that includes full projects and their dependencies and provide tools to extract non-local information with the help of program analyzers. We then focus on the task of function call argument completion which requires predicting the arguments to function calls. We show that existing code completion models do not yield good results on our completion task. To better solve this task, we query a program analyzer for information relevant to a given function call, and consider ways to provide the analyzer results to different code completion models during inference and training. Our experiments show that providing access to the function implementation and function usages greatly improves the argument completion performance. Our ablation study provides further insights on how different types of information available from the program analyzer and different ways of incorporating the information affect the model performance."
                },
                {
                    "title": "CodeGen2: Lessons for Training LLMs on Programming and Natural Languages",
                    "abstract": "Large language models (LLMs) have demonstrated remarkable abilities in representation learning for program synthesis and understanding tasks. The quality of the learned representations appears to be dictated by the neural scaling laws as a function of the number of model parameters and observations, while imposing upper bounds on the model performance by the amount of available data and compute, which is costly. In this study, we attempt to render the training of LLMs for program synthesis more efficient by unifying four key components: (1) model architectures, (2) learning methods, (3) infill sampling, and, (4) data distributions. Specifically, for the model architecture, we attempt to unify encoder and decoder-based models into a single prefix-LM. For learning methods, (i) causal language modeling, (ii) span corruption, (iii) infilling are unified into a simple learning algorithm. For infill sampling, we explore the claim of a\"free lunch\"hypothesis. For data distributions, the effect of a mixture distribution and multi-epoch training of programming and natural languages on model performance is explored. We conduct a comprehensive series of empirical experiments on 1B LLMs, for which failures and successes of this exploration are distilled into five lessons. We will provide a final recipe for training and release CodeGen2 models in size 1B, 3.7B, 7B, and, 16B parameters, along with the training framework as open-source: https://github.com/salesforce/CodeGen."
                },
                {
                    "title": "Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generation",
                    "abstract": "Program synthesis has been long studied with recent approaches focused on directly using the power of Large Language Models (LLMs) to generate code. Programming benchmarks, with curated synthesis problems and test-cases, are used to measure the performance of various LLMs on code synthesis. However, these test-cases can be limited in both quantity and quality for fully assessing the functional correctness of the generated code. Such limitation in the existing benchmarks begs the following question: In the era of LLMs, is the code generated really correct? To answer this, we propose EvalPlus -- a code synthesis evaluation framework to rigorously benchmark the functional correctness of LLM-synthesized code. EvalPlus augments a given evaluation dataset with large amounts of test-cases newly produced by an automatic test input generator, powered by both LLM- and mutation-based strategies. While EvalPlus is general, we extend the test-cases of the popular HumanEval benchmark by 80x to build HumanEval+. Our extensive evaluation across 26 popular LLMs (e.g., GPT-4 and ChatGPT) demonstrates that HumanEval+ is able to catch significant amounts of previously undetected wrong code synthesized by LLMs, reducing the pass@k by up-to 19.3-28.9%. We also surprisingly found that test insufficiency can lead to mis-ranking. For example, both WizardCoder-CodeLlama and Phind-CodeLlama now outperform ChatGPT on HumanEval+, while none of them could on HumanEval. Our work not only indicates that prior popular code synthesis evaluation results do not accurately reflect the true performance of LLMs for code synthesis, but also opens up a new direction to improve such programming benchmarks through automated testing. We have open-sourced our tools, enhanced datasets as well as all LLM-generated code at https://github.com/evalplus/evalplus to facilitate and accelerate future LLM-for-code research."
                },
                {
                    "title": "RepoCoder: Repository-Level Code Completion Through Iterative Retrieval and Generation",
                    "abstract": "The task of repository-level code completion is to continue writing the unfinished code based on a broader context of the repository. While for automated code completion tools, it is difficult to utilize the useful information scattered in different files. We propose RepoCoder, a simple, generic, and effective framework to address the challenge. It streamlines the repository-level code completion process by incorporating a similarity-based retriever and a pre-trained code language model in an iterative retrieval-generation pipeline. RepoCoder makes effective utilization of repository-level information for code completion and has the ability to generate code at various levels of granularity. Moreover, we propose a new benchmark RepoEval, which consists of the latest and high-quality real-world repositories covering line, API invocation, and function body completion scenarios. Experimental results indicate that RepoCoder significantly improves the In-File completion baseline by over 10% in all settings and consistently outperforms the vanilla retrieval-augmented code completion approach. Furthermore, we validate the effectiveness of RepoCoder through comprehensive analysis, providing valuable insights for future research. Our source code and benchmark are publicly available: https://github.com/microsoft/CodeT/tree/main/RepoCoder"
                },
                {
                    "title": "CoCoMIC: Code Completion by Jointly Modeling In-file and Cross-file Context",
                    "abstract": "While pre-trained language models (LM) for code have achieved great success in code completion, they generate code conditioned only on the contents within the file, i.e., in-file context, but ignore the rich semantics in other files within the same project, i.e., project-level cross-file context, a critical source of information that is especially useful in modern modular software development. Such overlooking constrains code LMs\u2019 capacity in code completion, leading to unexpected behaviors such as generating hallucinated class member functions or function calls with unexpected arguments. In this work, we propose CoCoMIC, a novel framework that jointly learns the in-file and cross-file context on top of code LMs. To empower CoCoMIC, we develop CCFinder, a static-analysis-based tool that locates and retrieves the most relevant project-level cross-file context for code completion. CoCoMIC successfully improves the existing code LM with a 33.94% relative increase in exact match and 28.69% in identifier matching for code completion when the cross-file context is provided. Finally, we perform a series of ablation studies and share valuable insights for future research on integrating cross-file context into code LMs."
                },
                {
                    "title": "ReCode: Robustness Evaluation of Code Generation Models",
                    "abstract": "Code generation models have achieved impressive performance. However, they tend to be brittle as slight edits to a prompt could lead to very different generations; these robustness properties, critical for user experience when deployed in real-life applications, are not well understood. Most existing works on robustness in text or code tasks have focused on classification, while robustness in generation tasks is an uncharted area and to date there is no comprehensive benchmark for robustness in code generation. In this paper, we propose ReCode, a comprehensive robustness evaluation benchmark for code generation models. We customize over 30 transformations specifically for code on docstrings, function and variable names, code syntax, and code format. They are carefully designed to be natural in real-life coding practice, preserve the original semantic meaning, and thus provide multifaceted assessments of a model\u2019s robustness performance. With human annotators, we verified that over 90% of the perturbed prompts do not alter the semantic meaning of the original prompt. In addition, we define robustness metrics for code generation models considering the worst-case behavior under each type of perturbation, taking advantage of the fact that executing the generated code can serve as objective evaluation. We demonstrate ReCode on SOTA models using HumanEval, MBPP, as well as function completion tasks derived from them. Interesting observations include: better robustness for CodeGen over InCoder and GPT-J; models are most sensitive to syntax perturbations; more challenging robustness evaluation on MBPP over HumanEval."
                },
                {
                    "title": "Multi-lingual Evaluation of Code Generation Models",
                    "abstract": "We present new benchmarks on evaluation code generation models: MBXP and Multilingual HumanEval, and MathQA-X. These datasets cover over 10 programming languages and are generated using a scalable conversion framework that transpiles prompts and test cases from the original Python datasets into the corresponding data in the target language. Using these benchmarks, we are able to assess the performance of code generation models in a multi-lingual fashion, and discovered generalization ability of language models on out-of-domain languages, advantages of multi-lingual models over mono-lingual, the ability of few-shot prompting to teach the model new languages, and zero-shot translation abilities even on mono-lingual settings. Furthermore, we use our code generation model to perform large-scale bootstrapping to obtain synthetic canonical solutions in several languages, which can be used for other code-related evaluations such as code insertion, robustness, or summarization tasks. Overall, our benchmarks represents a significant step towards a deeper understanding of language models' code generation abilities. We publicly release our code and datasets at https://github.com/amazon-research/mxeval."
                },
                {
                    "title": "CommitBART: A Large Pre-trained Model for GitHub Commits",
                    "abstract": "GitHub commits, which record the code changes with natural language messages for description, play a critical role for software developers to comprehend the software evolution. To promote the development of the open-source software community, we collect a commit benchmark including over 7.99 million commits across 7 programming languages. Based on this benchmark, we present CommitBART, a large pre-trained encoder-decoder Transformer model for GitHub commits. The model is pre-trained by three categories (i.e., denoising objectives, cross-modal generation and contrastive learning) for six pre-training tasks to learn commit fragment representations. Furthermore, we unify a ``commit intelligence'' framework with one understanding task and three generation tasks for commits. The comprehensive experiments on these tasks demonstrate that CommitBARTsignificantly outperforms previous pre-trained works for code. Further analysis also reveals each pre-training task enhances the model performance."
                },
                {
                    "title": "Repository-Level Prompt Generation for Large Language Models of Code",
                    "abstract": "With the success of large language models (LLMs) of code and their use as code assistants (e.g. Codex used in GitHub Copilot), techniques for introducing domain-specific knowledge in the prompt design process become important. In this work, we propose a framework called Repo-Level Prompt Generator that learns to generate example-specific prompts using prompt proposals. The prompt proposals take context from the entire repository, thereby incorporating both the structure of the repository and the context from other relevant files (e.g. imports, parent class files). Our technique doesn't require any access to the weights of the LLM, making it applicable in cases where we only have black-box access to the LLM. We conduct experiments on the task of single-line code-autocompletion using code repositories taken from Google Code archives. We demonstrate that an oracle constructed from our prompt proposals gives a remarkably high relative improvement of 36% over Codex, showing the quality of these proposals. Further, we show that when we train a model to predict a prompt proposal, we can achieve significant performance gains over Codex and other baselines. We release our code, data, and trained checkpoints at: \\url{https://github.com/shrivastavadisha/repo_level_prompt_generation}."
                },
                {
                    "title": "InCoder: A Generative Model for Code Infilling and Synthesis",
                    "abstract": "Code is seldom written in a single left-to-right pass and is instead repeatedly edited and refined. We introduce InCoder, a unified generative model that can perform program synthesis (via left-to-right generation) as well as editing (via infilling). InCoder is trained to generate code files from a large corpus of permissively licensed code, where regions of code have been randomly masked and moved to the end of each file, allowing code infilling with bidirectional context. Our model is the first generative model that is able to directly perform zero-shot code infilling, which we evaluate on challenging tasks such as type inference, comment generation, and variable re-naming. We find that the ability to condition on bidirectional context substantially improves performance on these tasks, while still performing comparably on standard program synthesis benchmarks in comparison to left-to-right only models pretrained at similar scale. The InCoder models and code are publicly released. https://sites.google.com/view/incoder-code-models"
                },
                {
                    "title": "CodeGen: An Open Large Language Model for Code with Multi-Turn Program Synthesis",
                    "abstract": "Program synthesis strives to generate a computer program as a solution to a given problem specification, expressed with input-output examples or natural language descriptions. The prevalence of large language models advances the state-of-the-art for program synthesis, though limited training resources and data impede open access to such models. To democratize this, we train and release a family of large language models up to 16.1B parameters, called CODEGEN, on natural language and programming language data, and open source the training library JAXFORMER. We show the utility of the trained model by demonstrating that it is competitive with the previous state-of-the-art on zero-shot Python code generation on HumanEval. We further investigate the multi-step paradigm for program synthesis, where a single program is factorized into multiple prompts specifying subproblems. To this end, we construct an open benchmark, Multi-Turn Programming Benchmark (MTPB), consisting of 115 diverse problem sets that are factorized into multi-turn prompts. Our analysis on MTPB shows that the same intent provided to CODEGEN in multi-turn fashion significantly improves program synthesis over that provided as a single turn. We make the training library JAXFORMER and model checkpoints available as open source contribution: https://github.com/salesforce/CodeGen."
                },
                {
                    "title": "ReACC: A Retrieval-Augmented Code Completion Framework",
                    "abstract": "Code completion, which aims to predict the following code token(s) according to the code context, can improve the productivity of software development. Recent work has proved that statistical language modeling with transformers can greatly improve the performance in the code completion task via learning from large-scale source code datasets. However, current approaches focus only on code context within the file or project, i.e. internal context. Our distinction is utilizing \u201dexternal\u201d context, inspired by human behaviors of copying from the related code snippets when writing code. Specifically, we propose a retrieval-augmented code completion framework, leveraging both lexical copying and referring to code with similar semantics by retrieval. We adopt a stage-wise training approach that combines a source code retriever and an auto-regressive language model for programming language. We evaluate our approach in the code completion task in Python and Java programming languages, achieving a state-of-the-art performance on CodeXGLUE benchmark."
                },
                {
                    "title": "UniXcoder: Unified Cross-Modal Pre-training for Code Representation",
                    "abstract": "Pre-trained models for programming languages have recently demonstrated great success on code intelligence. To support both code-related understanding and generation tasks, recent works attempt to pre-train unified encoder-decoder models. However, such encoder-decoder framework is sub-optimal for auto-regressive tasks, especially code completion that requires a decoder-only manner for efficient inference. In this paper, we present UniXcoder, a unified cross-modal pre-trained model for programming language. The model utilizes mask attention matrices with prefix adapters to control the behavior of the model and leverages cross-modal contents like AST and code comment to enhance code representation. To encode AST that is represented as a tree in parallel, we propose a one-to-one mapping method to transform AST in a sequence structure that retains all structural information from the tree. Furthermore, we propose to utilize multi-modal contents to learn representation of code fragment with contrastive learning, and then align representations among programming languages using a cross-modal generation task. We evaluate UniXcoder on five code-related tasks over nine datasets. To further evaluate the performance of code fragment representation, we also construct a dataset for a new task, called zero-shot code-to-code search. Results show that our model achieves state-of-the-art performance on most tasks and analysis reveals that comment and AST can both enhance UniXcoder."
                },
                {
                    "title": "Long-Range Modeling of Source Code Files with eWASH: Extended Window Access by Syntax Hierarchy",
                    "abstract": "Statistical language modeling and translation with transformers have found many successful applications in program understanding and generation tasks, setting high benchmarks for tools in modern software development environments. The finite context window of these neural models means, however, that they will be unable to leverage the entire relevant context of large files and packages for any given task. While there are many efforts to extend the context window, we introduce an architecture-independent approach for leveraging the syntactic hierarchies of source code for incorporating entire file-level context into a fixed-length window. Using concrete syntax trees of each source file we extract syntactic hierarchies and integrate them into context window by selectively removing from view more specific, less relevant scopes for a given task. We evaluate this approach on code generation tasks and joint translation of natural language and source code in Python programming language, achieving a new state-of-the-art in code completion and summarization for Python in the CodeXGLUE benchmark. We also introduce new CodeXGLUE benchmarks for user-experience-motivated tasks: code completion with normalized literals, method body completion/code summarization conditioned on file-level context."
                },
                {
                    "title": "CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and Generation",
                    "abstract": "Pre-trained models for Natural Languages (NL) like BERT and GPT have been recently shown to transfer well to Programming Languages (PL) and largely benefit a broad set of code-related tasks. Despite their success, most current methods either rely on an encoder-only (or decoder-only) pre-training that is suboptimal for generation (resp. understanding) tasks or process the code snippet in the same way as NL, neglecting the special characteristics of PL such as token types. We present CodeT5, a unified pre-trained encoder-decoder Transformer model that better leverages the code semantics conveyed from the developer-assigned identifiers. Our model employs a unified framework to seamlessly support both code understanding and generation tasks and allows for multi-task learning. Besides, we propose a novel identifier-aware pre-training task that enables the model to distinguish which code tokens are identifiers and to recover them when they are masked. Furthermore, we propose to exploit the user-written code comments with a bimodal dual generation task for better NL-PL alignment. Comprehensive experiments show that CodeT5 significantly outperforms prior methods on understanding tasks such as code defect detection and clone detection, and generation tasks across various directions including PL-NL, NL-PL, and PL-PL. Further analysis reveals that our model can better capture semantic information from code. Our code and pre-trained models are released at https://github.com/salesforce/CodeT5."
                },
                {
                    "title": "Program Synthesis with Large Language Models",
                    "abstract": "This paper explores the limits of the current generation of large language models for program synthesis in general purpose programming languages. We evaluate a collection of such models (with between 244M and 137B parameters) on two new benchmarks, MBPP and MathQA-Python, in both the few-shot and fine-tuning regimes. Our benchmarks are designed to measure the ability of these models to synthesize short Python programs from natural language descriptions. The Mostly Basic Programming Problems (MBPP) dataset contains 974 programming tasks, designed to be solvable by entry-level programmers. The MathQA-Python dataset, a Python version of the MathQA benchmark, contains 23914 problems that evaluate the ability of the models to synthesize code from more complex text. On both datasets, we find that synthesis performance scales log-linearly with model size. Our largest models, even without finetuning on a code dataset, can synthesize solutions to 59.6 percent of the problems from MBPP using few-shot learning with a well-designed prompt. Fine-tuning on a held-out portion of the dataset improves performance by about 10 percentage points across most model sizes. On the MathQA-Python dataset, the largest fine-tuned model achieves 83.8 percent accuracy. Going further, we study the model's ability to engage in dialog about code, incorporating human feedback to improve its solutions. We find that natural language feedback from a human halves the error rate compared to the model's initial prediction. Additionally, we conduct an error analysis to shed light on where these models fall short and what types of programs are most difficult to generate. Finally, we explore the semantic grounding of these models by fine-tuning them to predict the results of program execution. We find that even our best models are generally unable to predict the output of a program given a specific input."
                },
                {
                    "title": "Evaluating Large Language Models Trained on Code",
                    "abstract": "We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics."
                },
                {
                    "title": "CodeNet: A Large-Scale AI for Code Dataset for Learning a Diversity of Coding Tasks",
                    "abstract": "Over the last several decades, software has been woven into the fabric of every aspect of our society. As software development surges and code infrastructure of enterprise applications ages, it is now more critical than ever to increase software development productivity and modernize legacy applications. Advances in deep learning and machine learning algorithms have enabled numerous breakthroughs, motivating researchers to leverage AI techniques to improve software development efficiency. Thus, the fast-emerging research area of AI for Code has garnered new interest and gathered momentum. In this paper, we present a large-scale dataset CodeNet, consisting of over 14 million code samples and about 500 million lines of code in 55 different programming languages, which is aimed at teaching AI to code. In addition to its large scale, CodeNet has a rich set of high-quality annotations to benchmark and help accelerate research in AI techniques for a variety of critical coding tasks, including code similarity and classification, code translation between a large variety of programming languages, and code performance (runtime and memory) improvement techniques. Additionally, CodeNet provides sample input and output test sets for 98.5% of the code samples, which can be used as an oracle for determining code correctness and potentially guide reinforcement learning for code quality improvements. As a usability feature, we provide several pre-processing tools in CodeNet to transform source code into representations that can be readily used as inputs into machine learning models. Results of code classification and code similarity experiments using the CodeNet dataset are provided as a reference. We hope that the scale, diversity and rich, high-quality annotations of CodeNet will offer unprecedented research opportunities at the intersection of AI and Software Engineering."
                },
                {
                    "title": "Measuring Coding Challenge Competence With APPS",
                    "abstract": "While programming is one of the most broadly applicable skills in modern society, modern machine learning models still cannot code solutions to basic problems. Despite its importance, there has been surprisingly little work on evaluating code generation, and it can be difficult to accurately assess code generation performance rigorously. To meet this challenge, we introduce APPS, a benchmark for code generation. Unlike prior work in more restricted settings, our benchmark measures the ability of models to take an arbitrary natural language specification and generate satisfactory Python code. Similar to how companies assess candidate software developers, we then evaluate models by checking their generated code on test cases. Our benchmark includes 10,000 problems, which range from having simple one-line solutions to being substantial algorithmic challenges. We fine-tune large language models on both GitHub and our training set, and we find that the prevalence of syntax errors is decreasing exponentially as models improve. Recent models such as GPT-Neo can pass approximately 20% of the test cases of introductory problems, so we find that machine learning models are now beginning to learn how to code. As the social significance of automatic code generation increases over the coming years, our benchmark can provide an important measure for tracking advancements."
                },
                {
                    "title": "Unified Pre-training for Program Understanding and Generation",
                    "abstract": "Code summarization and generation empower conversion between programming language (PL) and natural language (NL), while code translation avails the migration of legacy code from one PL to another. This paper introduces PLBART, a sequence-to-sequence model capable of performing a broad spectrum of program and language understanding and generation tasks. PLBART is pre-trained on an extensive collection of Java and Python functions and associated NL text via denoising autoencoding. Experiments on code summarization in the English language, code generation, and code translation in seven programming languages show that PLBART outperforms or rivals state-of-the-art models. Moreover, experiments on discriminative tasks, e.g., program repair, clone detection, and vulnerable code detection, demonstrate PLBART\u2019s effectiveness in program understanding. Furthermore, analysis reveals that PLBART learns program syntax, style (e.g., identifier naming convention), logical flow (e.g., \u201cif\u201c block inside an \u201celse\u201c block is equivalent to \u201celse if\u201c block) that are crucial to program semantics and thus excels even with limited annotations."
                },
                {
                    "title": "CodeXGLUE: A Machine Learning Benchmark Dataset for Code Understanding and Generation",
                    "abstract": "Benchmark datasets have a significant impact on accelerating research in programming language tasks. In this paper, we introduce CodeXGLUE, a benchmark dataset to foster machine learning research for program understanding and generation. CodeXGLUE includes a collection of 10 tasks across 14 datasets and a platform for model evaluation and comparison. CodeXGLUE also features three baseline systems, including the BERT-style, GPT-style, and Encoder-Decoder models, to make it easy for researchers to use the platform. The availability of such data and baselines can help the development and validation of new methods that can be applied to various program understanding and generation problems."
                },
                {
                    "title": "CodeBERT: A Pre-Trained Model for Programming and Natural Languages",
                    "abstract": "We present CodeBERT, a bimodal pre-trained model for programming language (PL) and natural language (NL). CodeBERT learns general-purpose representations that support downstream NL-PL applications such as natural language code search, code documentation generation, etc. We develop CodeBERT with Transformer-based neural architecture, and train it with a hybrid objective function that incorporates the pre-training task of replaced token detection, which is to detect plausible alternatives sampled from generators. This enables us to utilize both \u201cbimodal\u201d data of NL-PL pairs and \u201cunimodal data, where the former provides input tokens for model training while the latter helps to learn better generators. We evaluate CodeBERT on two NL-PL applications by fine-tuning model parameters. Results show that CodeBERT achieves state-of-the-art performance on both natural language code search and code documentation generation. Furthermore, to investigate what type of knowledge is learned in CodeBERT, we construct a dataset for NL-PL probing, and evaluate in a zero-shot setting where parameters of pre-trained models are fixed. Results show that CodeBERT performs better than previous pre-trained models on NLPL probing."
                },
                {
                    "title": "MemorySanitizer: Fast detector of uninitialized memory use in C++",
                    "abstract": "This paper presents MemorySanitizer, a dynamic tool that detects uses of uninitialized memory in C and C++. The tool is based on compile time instrumentation and relies on bit-precise shadow memory at run-time. Shadow propagation technique is used to avoid false positive reports on copying of uninitialized memory. MemorySanitizer finds bugs at a modest cost of 2.5\u00d7 in execution time and 2\u00d7 in memory usage; the tool has an optional origin tracking mode that provides better reports with moderate extra overhead. The reports with origins are more detailed compared to reports from other similar tools; such reports contain names of local variables and the entire history of the uninitialized memory including intermediate stores. In this paper we share our experience in deploying the tool at a large scale and demonstrate the benefits of compile-time instrumentation over dynamic binary instrumentation."
                },
                {
                    "title": "AddressSanitizer: A Fast Address Sanity Checker",
                    "abstract": "Memory access bugs, including buffer overflows and uses of freed heap memory, remain a serious problem for programming languages like C and C++. Many memory error detectors exist, but most of them are either slow or detect a limited set of bugs, or both. \n \nThis paper presents AddressSanitizer, a new memory error detector. Our tool finds out-of-bounds accesses to heap, stack, and global objects, as well as use-after-free bugs. It employs a specialized memory allocator and code instrumentation that is simple enough to be implemented in any compiler, binary translation system, or even in hardware. \n \nAddressSanitizer achieves efficiency without sacrificing comprehensiveness. Its average slowdown is just 73% yet it accurately detects bugs at the point of occurrence. It has found over 300 previously unknown bugs in the Chromium browser and many bugs in other software."
                },
                {
                    "title": "HFL: Hybrid Fuzzing on the Linux Kernel",
                    "abstract": "Hybrid fuzzing, combining symbolic execution and fuzzing, is a promising approach for vulnerability discovery because each approach can complement the other. However, we observe that applying hybrid fuzzing to kernel testing is challenging because the following unique characteristics of the kernel make a naive adoption of hybrid fuzzing inefficient: 1) having indirect control transfers determined by system call arguments, 2) controlling and matching internal system state via system calls, and 3) inferring nested argument type for invoking system calls. Failure to handling such challenges will render both fuzzing and symbolic execution inefficient, and thereby, will result in an inefficient hybrid fuzzing. Although these challenges are essential to both fuzzing and symbolic execution, to the best of our knowledge, existing kernel testing approaches either naively use each technique separately without handling such challenges or imprecisely handle a part of challenges only by static analysis. To this end, this paper proposes HFL, which not only combines fuzzing with symbolic execution for hybrid fuzzing but also addresses kernel-specific fuzzing challenges via three distinct features: 1) converting indirect control transfers to direct transfers, 2) inferring system call sequence to build a consistent system state, and 3) identifying nested arguments types of system calls. As a result, HFL found 24 previously unknown vulnerabilities in recent Linux kernels. Additionally, HFL achieves 15% and 26% higher code coverage than Moonshine and Syzkaller, respectively, and over kAFL/S2E/TriforceAFL, achieving even four times better coverage, using the same amount of resources (CPU, time, etc.). Regarding vulnerability discovery performance, HFL found 13 known vulnerabilities more than three times faster than Syzkaller."
                },
                {
                    "title": "kAFL: Hardware-Assisted Feedback Fuzzing for OS Kernels",
                    "abstract": "Many kinds of memory safety vulnerabilities have been endangering software systems for decades. Amongst other approaches, fuzzing is a promising technique to unveil various software faults. Recently, feedback-guided fuzzing demonstrated its power, producing a steady stream of security-critical software bugs. Most fuzzing efforts\u2014especially feedback fuzzing\u2014are limited to user space components of an operating system (OS), although bugs in kernel components are more severe, because they allow an attacker to gain access to a system with full privileges. Unfortunately, kernel components are difficult to fuzz as feedback mechanisms (i.e., guided code coverage) cannot be easily applied. Additionally, non-determinism due to interrupts, kernel threads, statefulness, and similar mechanisms poses problems. Furthermore, if a process fuzzes its own kernel, a kernel crash highly impacts the performance of the fuzzer as the OS needs to reboot. In this paper, we approach the problem of coverageguided kernel fuzzing in an OS-independent and hardware-assisted way: We utilize a hypervisor and Intel\u2019s Processor Trace (PT) technology. This allows us to remain independent of the target OS as we just require a small user space component that interacts with the targeted OS. As a result, our approach introduces almost no performance overhead, even in cases where the OS crashes, and performs up to 17,000 executions per second on an off-the-shelf laptop. We developed a framework called kernel-AFL (kAFL) to assess the security of Linux, macOS, and Windows kernel components. Among many crashes, we uncovered several flaws in the ext4 driver for Linux, the HFS and APFS file system of macOS, and the NTFS driver of Windows."
                }
            ],
            "categories": [
                "cs.SE"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively utilize code LLMs to repair crash-inducing bugs in the Linux kernel given a crash report and the corresponding input that triggers the crash?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant gap in the application of LLMs to real-world software engineering challenges, particularly in critical systems like the Linux kernel. By developing a benchmark that reflects the complexities of debugging in production environments, we can advance the understanding of LLM capabilities in software maintenance and repair. This research could lead to practical applications that enhance the reliability and stability of software systems, ultimately benefiting billions of devices that rely on Linux. Furthermore, it may inspire future research into more sophisticated LLMs that can handle complex, ambiguous tasks in software engineering.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the intricate nature of the Linux kernel, which includes a vast codebase with low-level programming languages, concurrent execution, and non-deterministic behavior. Naive approaches may fail due to the need for hardware awareness and memory safety, as well as the difficulty in reproducing bugs caused by thread interleavings. Additionally, the ambiguity in crash reports complicates the identification of root causes, requiring advanced reasoning capabilities that current LLMs may lack. The need for a benchmark that accommodates flaky test oracles further adds to the complexity of the task.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on simpler, deterministic programming tasks that do not capture the nuances of real-world software engineering, particularly in complex systems like the Linux kernel. Existing benchmarks have not addressed the unique challenges posed by low-level code, concurrency, and ambiguous crash reports. Barriers such as the lack of suitable datasets and the complexity of the Linux environment have hindered progress. Our approach differs by introducing a new benchmark, kBench, and an execution environment, kGym, specifically designed to tackle these challenges, thereby providing a more realistic testing ground for LLMs.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves creating the kBench benchmark and the kGym execution environment to facilitate the testing of LLMs on crash resolution tasks in the Linux kernel. We will utilize a dataset comprising crash reports and corresponding inputs that trigger crashes, along with a set"
            }
        },
        "author_data": {
            "6f29ee43-0f72-4c25-a89a-e5291c710459": {
                "pk": "6f29ee43-0f72-4c25-a89a-e5291c710459",
                "name": "Alex Mathai",
                "collaborators": [
                    "Srikanth Tamilselvam",
                    "Shreya Khare",
                    "Senthil Mani",
                    "Sambaran Bandyopadhyay",
                    "Utkarsh Desai",
                    "Sharat Chandra Madanapalli",
                    "Hassan Habibi Gharakheili",
                    "Vijay Sivaraman",
                    "Debeshee Das",
                    "Noble Saji Mathews"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Software Engineering",
                    "Video Streaming",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "Adversarial Black-Box Attacks On Text Classifiers Using Multi-Objective Genetic Optimization Guided By Deep Networks",
                        "abstract": "We propose a novel genetic-algorithm technique that generates black-box adversarial examples which successfully fool neural network based text classifiers. We perform a genetic search with multi-objective optimization guided by deep learning based inferences and Seq2Seq mutation to generate semantically similar but imperceptible adversaries. We compare our approach with DeepWordBug (DWB) on SST and IMDB sentiment datasets by attacking three trained models viz. char-LSTM, word-LSTM and elmo-LSTM. On an average, we achieve an attack success rate of 65.67% for SST and 36.45% for IMDB across the three models showing an improvement of 49.48% and 101% respectively. Furthermore, our qualitative study indicates that 94% of the time, the users were not able to distinguish between an original and adversarial sample."
                    },
                    {
                        "title": "Monolith to Microservices: Representing Application Software through Heterogeneous Graph Neural Network",
                        "abstract": "Monolithic software encapsulates all functional capabilities into a single deployable unit. But managing it becomes harder as the demand for new functionalities grow. Microservice architecture is seen as an alternate as it advocates building an application through a set of loosely coupled small services wherein each service owns a single functional responsibility. But the challenges associated with the separation of functional modules, slows down the migration of a monolithic code into microservices. In this work, we propose a representation learning based solution to tackle this problem. We use a heterogeneous graph to jointly represent software artifacts (like programs and resources) and the different relationships they share (function calls, inheritance, etc.), and perform a constraint-based clustering through a novel heterogeneous graph neural network. Experimental studies show that our approach is effective on monoliths of different types."
                    },
                    {
                        "title": "Modeling Live Video Streaming: Real-Time Classification, QoE Inference, and Field Evaluation",
                        "abstract": "Social media, professional sports, and video games are driving rapid growth in live video streaming, on platforms such as Twitch and YouTube Live. Live streaming experience is very susceptible to short-time-scale network congestion since client playback buffers are often no more than a few seconds. Unfortunately, identifying such streams and measuring their QoE for network management is challenging, since content providers largely use the same delivery infrastructure for live and video-on-demand (VoD) streaming, and packet inspection techniques (including SNI/DNS query monitoring) cannot always distinguish between the two.   In this paper, we design, build, and deploy ReCLive: a machine learning method for live video detection and QoE measurement based on network-level behavioral characteristics. Our contributions are four-fold: (1) We analyze about 23,000 video streams from Twitch and YouTube, and identify key features in their traffic profile that differentiate live and on-demand streaming. We release our traffic traces as open data to the public; (2) We develop an LSTM-based binary classifier model that distinguishes live from on-demand streams in real-time with over 95% accuracy across providers; (3) We develop a method that estimates QoE metrics of live streaming flows in terms of resolution and buffer stall events with overall accuracies of 93% and 90%, respectively; and (4) Finally, we prototype our solution, train it in the lab, and deploy it in a live ISP network serving more than 7,000 subscribers. Our method provides ISPs with fine-grained visibility into live video streams, enabling them to measure and improve user experience."
                    },
                    {
                        "title": "COMEX: A Tool for Generating Customized Source Code Representations",
                        "abstract": "Learning effective representations of source code is critical for any Machine Learning for Software Engineering (ML4SE) system. Inspired by natural language processing, large language models (LLMs) like Codex and CodeGen treat code as generic sequences of text and are trained on huge corpora of code data, achieving state of the art performance on several software engineering (SE) tasks. However, valid source code, unlike natural language, follows a strict structure and pattern governed by the underlying grammar of the programming language. Current LLMs do not exploit this property of the source code as they treat code like a sequence of tokens and overlook key structural and semantic properties of code that can be extracted from code-views like the Control Flow Graph (CFG), Data Flow Graph (DFG), Abstract Syntax Tree (AST), etc. Unfortunately, the process of generating and integrating code-views for every programming language is cumbersome and time consuming. To overcome this barrier, we propose our tool COMEX - a framework that allows researchers and developers to create and combine multiple code-views which can be used by machine learning (ML) models for various SE tasks. Some salient features of our tool are: (i) it works directly on source code (which need not be compilable), (ii) it currently supports Java and C#, (iii) it can analyze both method-level snippets and program-level snippets by using both intra-procedural and inter-procedural analysis, and (iv) it is easily extendable to other languages as it is built on tree-sitter - a widely used incremental parser that supports over 40 languages. We believe this easy-to-use code-view generation and customization tool will give impetus to research in source code representation learning methods and ML4SE.   Tool: https://pypi.org/project/comex - GitHub: https://github.com/IBM/tree-sitter-codeviews - Demo: https://youtu.be/GER6U87FVbU"
                    }
                ]
            },
            "215fa426-ab60-4bff-bb15-45fff2188adc": {
                "pk": "215fa426-ab60-4bff-bb15-45fff2188adc",
                "name": "Chenxi Huang",
                "collaborators": [
                    "Bryce Gadway",
                    "Jacob P. Covey",
                    "Chaoyang Jiang",
                    "Zhenghua Chen",
                    "Bowen Zheng",
                    "Yuemei Luo",
                    "Tao Chen",
                    "Jackson Ang'ong'a"
                ],
                "domain": [
                    "Federated Learning",
                    "Time Series Analysis",
                    "Quantum Computing",
                    "Optical Imaging"
                ],
                "publications": [
                    {
                        "title": "Shuffled Differentially Private Federated Learning for Time Series Data Analytics",
                        "abstract": "Trustworthy federated learning aims to achieve optimal performance while ensuring clients' privacy. Existing privacy-preserving federated learning approaches are mostly tailored for image data, lacking applications for time series data, which have many important applications, like machine health monitoring, human activity recognition, etc. Furthermore, protective noising on a time series data analytics model can significantly interfere with temporal-dependent learning, leading to a greater decline in accuracy. To address these issues, we develop a privacy-preserving federated learning algorithm for time series data. Specifically, we employ local differential privacy to extend the privacy protection trust boundary to the clients. We also incorporate shuffle techniques to achieve a privacy amplification, mitigating the accuracy decline caused by leveraging local differential privacy. Extensive experiments were conducted on five time series datasets. The evaluation results reveal that our algorithm experienced minimal accuracy loss compared to non-private federated learning in both small and large client scenarios. Under the same level of privacy protection, our algorithm demonstrated improved accuracy compared to the centralized differentially private federated learning in both scenarios."
                    },
                    {
                        "title": "Towards Head Computed Tomography Image Reconstruction Standardization with Deep Learning Assisted Automatic Detection",
                        "abstract": "Three-dimensional (3D) reconstruction of head Computed Tomography (CT) images elucidates the intricate spatial relationships of tissue structures, thereby assisting in accurate diagnosis. Nonetheless, securing an optimal head CT scan without deviation is challenging in clinical settings, owing to poor positioning by technicians, patient's physical constraints, or CT scanner tilt angle restrictions. Manual formatting and reconstruction not only introduce subjectivity but also strain time and labor resources. To address these issues, we propose an efficient automatic head CT images 3D reconstruction method, improving accuracy and repeatability, as well as diminishing manual intervention. Our approach employs a deep learning-based object detection algorithm, identifying and evaluating orbitomeatal line landmarks to automatically reformat the images prior to reconstruction. Given the dearth of existing evaluations of object detection algorithms in the context of head CT images, we compared ten methods from both theoretical and experimental perspectives. By exploring their precision, efficiency, and robustness, we singled out the lightweight YOLOv8 as the aptest algorithm for our task, with an mAP of 92.77% and impressive robustness against class imbalance. Our qualitative evaluation of standardized reconstruction results demonstrates the clinical practicability and validity of our method."
                    },
                    {
                        "title": "Quantum walks and correlated dynamics in an interacting synthetic Rydberg lattice",
                        "abstract": "Coherent dynamics of interacting quantum particles plays a central role in the study of strongly correlated quantum matter and the pursuit of quantum information processors. Here, we present the state-space of interacting Rydberg atoms as a synthetic landscape on which to control and observe coherent and correlated dynamics. With full control of the coupling strengths and energy offsets between the pairs of sites in a nine-site synthetic lattice, we realize quantum walks, Bloch oscillations, and dynamics in an Escher-type \"continuous staircase\". In the interacting regime, we observe correlated quantum walks, Bloch oscillations, and confinement of particle pairs. Additionally, we simultaneously tilt our lattice both up and down to achieve coherent pair oscillations. When combined with a few straightforward upgrades, this work establishes synthetic Rydberg lattices of interacting atom arrays as a promising platform for programmable quantum many-body dynamics with access to features that are difficult to realize in real-space lattices."
                    },
                    {
                        "title": "Gray molasses cooling of $^{39}$K atoms in optical tweezers",
                        "abstract": "Robust cooling and nondestructive imaging are prerequisites for many emerging applications of neutral atoms trapped in optical tweezers, such as their use in quantum information science and analog quantum simulation. The tasks of cooling and imaging can be challenged, however, by the presence of large trap-induced shifts of their respective optical transitions. Here, we explore a system of $^{39}$K atoms trapped in a near-detuned ($780$ nm) optical tweezer, which leads to relatively minor differential (ground vs. excited state) Stark shifts. We demonstrate that simple and robust loading, cooling, and imaging can be achieved through a combined addressing of the D$_\\textrm{1}$ and D$_\\textrm{2}$ transitions. While imaging on the D$_\\textrm{2}$ transition, we can simultaneously apply $\\Lambda$-enhanced gray molasses (GM) on the D$_\\textrm{1}$ transition, preserving low backgrounds for single-atom imaging through spectral filtering. Using D$_\\textrm{1}$ cooling during and after trap loading, we demonstrate enhanced loading efficiencies as well as cooling to low temperatures. These results suggest a simple and robust path for loading and cooling large arrays of potassium atoms in optical tweezers through the use of resource-efficient near-detuned optical tweezers and GM cooling."
                    }
                ]
            },
            "8d6eda53-f82f-4f07-be88-886289e1eb6a": {
                "pk": "8d6eda53-f82f-4f07-be88-886289e1eb6a",
                "name": "Petros Maniatis",
                "collaborators": [
                    "Mary Baker",
                    "T. J. Giuli",
                    "Aditya Kanade",
                    "Rishabh Singh",
                    "Charles Sutton",
                    "Byung-Gon Chun",
                    "Mema Roussopoulos",
                    "David S. H. Rosenthal",
                    "Kensen Shi",
                    "Mayur Naik"
                ],
                "domain": [
                    "Secure Systems",
                    "Program Synthesis",
                    "Machine Learning",
                    "Code Understanding"
                ],
                "publications": [
                    {
                        "title": "Secure History Preservation Through Timeline Entanglement",
                        "abstract": "A secure timeline is a tamper-evident historic record of the states through which a system goes throughout its operational history. Secure timelines can help us reason about the temporal ordering of system states in a provable manner. We extend secure timelines to encompass multiple, mutually distrustful services, using timeline entanglement. Timeline entanglement associates disparate timelines maintained at independent systems, by linking undeniably the past of one timeline to the future of another. Timeline entanglement is a sound method to map a time step in the history of one service onto the timeline of another, and helps clients of entangled services to get persistent temporal proofs for services rendered that survive the demise or non-cooperation of the originating service. In this paper we present the design and implementation of Timeweave, our service development framework for timeline entanglement based on two novel disk-based authenticated data structures. We evaluate Timeweave's performance characteristics and show that it can be efficiently deployed in a loosely-coupled distributed system of a few hundred services with overhead of roughly 2-8% of the processing resources of a PC-grade system."
                    },
                    {
                        "title": "Authenticated Append-only Skip Lists",
                        "abstract": "In this work we describe, design and analyze the security of a tamper-evident, append-only data structure for maintaining secure data sequences in a loosely coupled distributed system where individual system components may be mutually distrustful. The resulting data structure, called an Authenticated Append-Only Skip List (AASL), allows its maintainers to produce one-way digests of the entire data sequence, which they can publish to others as a commitment on the contents and order of the sequence. The maintainer can produce efficiently succinct proofs that authenticate a particular datum in a particular position of the data sequence against a published digest. AASLs are secure against tampering even by malicious data structure maintainers. First, we show that a maintainer cannot ``invent'' and authenticate data elements for the AASL after he has committed to the structure. Second, he cannot equivocate by being able to prove conflicting facts about a particular position of the data sequence. This is the case even when the data sequence grows with time and its maintainer publishes successive commitments at times of his own choosing.   AASLs can be invaluable in reasoning about the integrity of system logs maintained by untrusted components of a loosely-coupled distributed system."
                    },
                    {
                        "title": "Oblivious Stash Shuffle",
                        "abstract": "This is a companion report to Bittau et al. We restate and prove security of the Stash Shuffle."
                    },
                    {
                        "title": "A Historic Name-Trail Service",
                        "abstract": "People change the identifiers through which they are reachable online as they change jobs or residences or Internet service providers. This kind of personal mobility makes reaching people online error-prone. As people move, they do not always know who or what has cached their now obsolete identifiers so as to inform them of the move. Use of these old identifiers can cause delivery failure of important messages, or worse, may cause delivery of messages to unintended recipients. For example, a sensitive email message sent to my now obsolete work address at a former place of employment may reach my unfriendly former boss instead of me.   In this paper we describe HINTS, a historic name-trail service. This service provides a persistent way to name willing participants online using today's transient online identifiers. HINTS accomplishes this by connecting together the names a person uses along with the times during which those names were valid for the person, thus giving people control over the historic use of their names. A correspondent who wishes to reach a mobile person can use an obsolete online name for that person, qualified with a time at which the online name was successfully used; HINTS resolves this historic name to a current valid online identifier for the intended recipient, if that recipient has chosen to leave a name trail in HINTS."
                    },
                    {
                        "title": "Glimmers: Resolving the Privacy/Trust Quagmire",
                        "abstract": "Many successful services rely on trustworthy contributions from users. To establish that trust, such services often require access to privacy-sensitive information from users, thus creating a conflict between privacy and trust. Although it is likely impractical to expect both absolute privacy and trustworthiness at the same time, we argue that the current state of things, where individual privacy is usually sacrificed at the altar of trustworthy services, can be improved with a pragmatic $Glimmer$ $of$ $Trust$, which allows services to validate user contributions in a trustworthy way without forfeiting user privacy. We describe how trustworthy hardware such as Intel's SGX can be used client-side -- in contrast to much recent work exploring SGX in cloud services -- to realize the Glimmer architecture, and demonstrate how this realization is able to resolve the tension between privacy and trust in a variety of cases."
                    },
                    {
                        "title": "Enabling the Long-Term Archival of Signed Documents through Time Stamping",
                        "abstract": "In this paper we describe how to build a trusted reliable distributed service across administrative domains in a peer-to-peer network. The application we use to motivate our work is a public key time stamping service called Prokopius. The service provides a secure, verifiable but distributable stable archive that maintains time stamped snapshots of public keys over time. This in turn allows clients to verify time stamped documents or certificates that rely on formerly trusted public keys that are no longer in service or where the signer no longer exists. We find that such a service can time stamp the snapshots of public keys in a network of 148 nodes at the granularity of a couple of days, even in the worst case where an adversary causes the maximal amount of damage allowable within our fault model."
                    },
                    {
                        "title": "CRQBench: A Benchmark of Code Reasoning Questions",
                        "abstract": "Large Language Models have demonstrated exceptional proficiency on coding tasks, but it is challenging to precisely evaluate their code reasoning ability. Existing benchmarks are insufficient as they are unrealistic and conflate semantic reasoning ability with performance on software engineering tasks. We introduce CRQBench, a benchmark of 100 C++ code reasoning questions and answers derived from contextualized code review comments. To curate CRQBench, we use an LLM assistant alongside human inspection, reducing manual effort. We conduct an evaluation of GPT-4 on CRQBench and find that it produces correct responses grounded in the given context for 65 of the 100 questions."
                    },
                    {
                        "title": "Learning and Evaluating Contextual Embedding of Source Code",
                        "abstract": "Recent research has achieved impressive results on understanding and improving source code by building up on machine-learning techniques developed for natural languages. A significant advancement in natural-language understanding has come with the development of pre-trained contextual embeddings, such as BERT, which can be fine-tuned for downstream tasks with less labeled data and training budget, while achieving better accuracies. However, there is no attempt yet to obtain a high-quality contextual embedding of source code, and to evaluate it on multiple program-understanding tasks simultaneously; that is the gap that this paper aims to mitigate. Specifically, first, we curate a massive, deduplicated corpus of 7.4M Python files from GitHub, which we use to pre-train CuBERT, an open-sourced code-understanding BERT model; and, second, we create an open-sourced benchmark that comprises five classification tasks and one program-repair task, akin to code-understanding tasks proposed in the literature before. We fine-tune CuBERT on our benchmark tasks, and compare the resulting models to different variants of Word2Vec token embeddings, BiLSTM and Transformer models, as well as published state-of-the-art models, showing that CuBERT outperforms them all, even with shorter training, and with fewer labeled examples. Future work on source-code embedding can benefit from reusing our benchmark, and from comparing against CuBERT models as a strong baseline."
                    },
                    {
                        "title": "Verifiable Network-Performance Measurements",
                        "abstract": "In the current Internet, there is no clean way for affected parties to react to poor forwarding performance: when a domain violates its Service Level Agreement (SLA) with a contractual partner, the partner must resort to ad-hoc probing-based monitoring to determine the existence and extent of the violation. Instead, we propose a new, systematic approach to the problem of forwarding-performance verification. Our mechanism relies on voluntary reporting, allowing each domain to disclose its loss and delay performance to its neighbors; it does not disclose any information regarding the participating domains' topology or routing policies beyond what is already publicly available. Most importantly, it enables verifiable performance measurements, i.e., domains cannot abuse it to significantly exaggerate their performance. Finally, our mechanism is tunable, allowing each participating domain to determine how many resources to devote to it independently (i.e., without any inter-domain coordination), exposing a controllable trade-off between performance-verification quality and resource consumption. Our mechanism comes at the cost of deploying modest functionality at the participating domains' border routers; we show that it requires reasonable processing and memory resources within modern network capabilities."
                    },
                    {
                        "title": "Neural Program Synthesis with a Differentiable Fixer",
                        "abstract": "We present a new program synthesis approach that combines an encoder-decoder based synthesis architecture with a differentiable program fixer. Our approach is inspired from the fact that human developers seldom get their program correct on the first attempt, and perform iterative testing-based program fixing to get to the desired program functionality. Similarly, our approach first learns a distribution over programs conditioned on an encoding of a set of input-output examples, and then iteratively performs fix operations using the differentiable fixer. The fixer takes as input the original examples and the current program's outputs on example inputs, and generates a new distribution over the programs with the goal of reducing the discrepancies between the current program outputs and the desired example outputs. We train our architecture end-to-end on the RobustFill domain, and show that the addition of the fixer module leads to a significant improvement on synthesis accuracy compared to using beam search."
                    },
                    {
                        "title": "A Data Capsule Framework For Web Services: Providing Flexible Data Access Control To Users",
                        "abstract": "This paper introduces the notion of a secure data capsule, which refers to an encapsulation of sensitive user information (such as a credit card number) along with code that implements an interface suitable for the use of such information (such as charging for purchases) by a service (such as an online merchant). In our capsule framework, users provide their data in the form of such capsules to web services rather than raw data. Capsules can be deployed in a variety of ways, either on a trusted third party or the user's own computer or at the service itself, through the use of a variety of hardware or software modules, such as a virtual machine monitor or trusted platform module: the only requirement is that the deployment mechanism must ensure that the user's data is only accessed via the interface sanctioned by the user. The framework further allows an user to specify policies regarding which services or machines may host her capsule, what parties are allowed to access the interface, and with what parameters. The combination of interface restrictions and policy control lets us bound the impact of an attacker who compromises the service to gain access to the user's capsule or a malicious insider at the service itself."
                    },
                    {
                        "title": "CloneCloud: Boosting Mobile Device Applications Through Cloud Clone Execution",
                        "abstract": "Mobile applications are becoming increasingly ubiquitous and provide ever richer functionality on mobile devices. At the same time, such devices often enjoy strong connectivity with more powerful machines ranging from laptops and desktops to commercial clouds. This paper presents the design and implementation of CloneCloud, a system that automatically transforms mobile applications to benefit from the cloud. The system is a flexible application partitioner and execution runtime that enables unmodified mobile applications running in an application-level virtual machine to seamlessly off-load part of their execution from mobile devices onto device clones operating in a computational cloud. CloneCloud uses a combination of static analysis and dynamic profiling to optimally and automatically partition an application so that it migrates, executes in the cloud, and re-integrates computation in a fine-grained manner that makes efficient use of resources. Our evaluation shows that CloneCloud can achieve up to 21.2x speedup of smartphone applications we tested and it allows different partitioning for different inputs and networks."
                    },
                    {
                        "title": "Neural Program Repair by Jointly Learning to Localize and Repair",
                        "abstract": "Due to its potential to improve programmer productivity and software quality, automated program repair has been an active topic of research. Newer techniques harness neural networks to learn directly from examples of buggy programs and their fixes. In this work, we consider a recently identified class of bugs called variable-misuse bugs. The state-of-the-art solution for variable misuse enumerates potential fixes for all possible bug locations in a program, before selecting the best prediction. We show that it is beneficial to train a model that jointly and directly localizes and repairs variable-misuse bugs. We present multi-headed pointer networks for this purpose, with one head each for localization and repair. The experimental results show that the joint model significantly outperforms an enumerative solution that uses a pointer based model for repair alone."
                    },
                    {
                        "title": "Preserving Peer Replicas By Rate-Limited Sampled Voting in LOCKSS",
                        "abstract": "The LOCKSS project has developed and deployed in a world-wide test a peer-to-peer system for preserving access to journals and other archival information published on the Web. It consists of a large number of independent, low-cost, persistent web caches that cooperate to detect and repair damage to their content by voting in \"opinion polls.\" Based on this experience, we present a design for and simulations of a novel protocol for voting in systems of this kind. It incorporates rate limitation and intrusion detection to ensure that even some very powerful adversaries attacking over many years have only a small probability of causing irrecoverable damage before being detected."
                    },
                    {
                        "title": "Attrition Defenses for a Peer-to-Peer Digital Preservation System",
                        "abstract": "In peer-to-peer systems, attrition attacks include both traditional, network-level denial of service attacks as well as application-level attacks in which malign peers conspire to waste loyal peers' resources. We describe several defenses for LOCKSS, a peer-to-peer digital preservation system, that help ensure that application-level attacks even from powerful adversaries are less effective than simple network-level attacks, and that network-level attacks must be intense, wide-spread, and prolonged to impair the system."
                    },
                    {
                        "title": "Notes On The Design Of An Internet Adversary",
                        "abstract": "The design of the defenses Internet systems can deploy against attack, especially adaptive and resilient defenses, must start from a realistic model of the threat. This requires an assessment of the capabilities of the adversary. The design typically evolves through a process of simulating both the system and the adversary. This requires the design and implementation of a simulated adversary based on the capability assessment. Consensus on the capabilities of a suitable adversary is not evident. Part of the recent redesign of the protocol used by peers in the LOCKSS digital preservation system included a conservative assessment of the adversary's capabilities. We present our assessment and the implications we drew from it as a step towards a reusable adversary specification."
                    },
                    {
                        "title": "Mantis: Predicting System Performance through Program Analysis and Modeling",
                        "abstract": "We present Mantis, a new framework that automatically predicts program performance with high accuracy. Mantis integrates techniques from programming language and machine learning for performance modeling, and is a radical departure from traditional approaches. Mantis extracts program features, which are information about program execution runs, through program instrumentation. It uses machine learning techniques to select features relevant to performance and creates prediction models as a function of the selected features. Through program analysis, it then generates compact code slices that compute these feature values for prediction. Our evaluation shows that Mantis can achieve more than 93% accuracy with less than 10% training data set, which is a significant improvement over models that are oblivious to program features. The system generates code slices that are cheap to compute feature values."
                    },
                    {
                        "title": "A Library for Representing Python Programs as Graphs for Machine Learning",
                        "abstract": "Graph representations of programs are commonly a central element of machine learning for code research. We introduce an open source Python library python_graphs that applies static analysis to construct graph representations of Python programs suitable for training machine learning models. Our library admits the construction of control-flow graphs, data-flow graphs, and composite ``program graphs'' that combine control-flow, data-flow, syntactic, and lexical information about a program. We present the capabilities and limitations of the library, perform a case study applying the library to millions of competitive programming submissions, and showcase the library's utility for machine learning research."
                    },
                    {
                        "title": "SpreadsheetCoder: Formula Prediction from Semi-structured Context",
                        "abstract": "Spreadsheet formula prediction has been an important program synthesis problem with many real-world applications. Previous works typically utilize input-output examples as the specification for spreadsheet formula synthesis, where each input-output pair simulates a separate row in the spreadsheet. However, this formulation does not fully capture the rich context in real-world spreadsheets. First, spreadsheet data entries are organized as tables, thus rows and columns are not necessarily independent from each other. In addition, many spreadsheet tables include headers, which provide high-level descriptions of the cell data. However, previous synthesis approaches do not consider headers as part of the specification. In this work, we present the first approach for synthesizing spreadsheet formulas from tabular context, which includes both headers and semi-structured tabular data. In particular, we propose SpreadsheetCoder, a BERT-based model architecture to represent the tabular context in both row-based and column-based formats. We train our model on a large dataset of spreadsheets, and demonstrate that SpreadsheetCoder achieves top-1 prediction accuracy of 42.51%, which is a considerable improvement over baselines that do not employ rich tabular context. Compared to the rule-based system, SpreadsheetCoder assists 82% more users in composing formulas on Google Sheets."
                    },
                    {
                        "title": "CodeQueries: A Dataset of Semantic Queries over Code",
                        "abstract": "Developers often have questions about semantic aspects of code they are working on, e.g., \"Is there a class whose parent classes declare a conflicting attribute?\". Answering them requires understanding code semantics such as attributes and inheritance relation of classes. An answer to such a question should identify code spans constituting the answer (e.g., the declaration of the subclass) as well as supporting facts (e.g., the definitions of the conflicting attributes). The existing work on question-answering over code has considered yes/no questions or method-level context. We contribute a labeled dataset, called CodeQueries, of semantic queries over Python code. Compared to the existing datasets, in CodeQueries, the queries are about code semantics, the context is file level and the answers are code spans. We curate the dataset based on queries supported by a widely-used static analysis tool, CodeQL, and include both positive and negative examples, and queries requiring single-hop and multi-hop reasoning.   To assess the value of our dataset, we evaluate baseline neural approaches. We study a large language model (GPT3.5-Turbo) in zero-shot and few-shot settings on a subset of CodeQueries. We also evaluate a BERT style model (CuBERT) with fine-tuning. We find that these models achieve limited success on CodeQueries. CodeQueries is thus a challenging dataset to test the ability of neural models, to understand code semantics, in the extractive question-answering setting."
                    }
                ]
            },
            "da802ed3-d594-4376-8bae-35a632c9e4e4": {
                "pk": "da802ed3-d594-4376-8bae-35a632c9e4e4",
                "name": "Aleksandr Nogikh",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "5e61bbf1-fd3b-472e-b4aa-736847ccbca1": {
                "pk": "5e61bbf1-fd3b-472e-b4aa-736847ccbca1",
                "name": "Franjo Ivancic",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "5966121b-83e9-416d-8cee-2645e1f9bf59": {
                "pk": "5966121b-83e9-416d-8cee-2645e1f9bf59",
                "name": "Junfeng Yang",
                "collaborators": [
                    "Guoyong Gu",
                    "Shiqian Ma",
                    "Xiaokai Chang",
                    "Yunhai Xiao",
                    "Yin Zhang",
                    "Caihua Chen",
                    "Hongchao Zhang",
                    "Raymond H. Chan"
                ],
                "domain": [
                    "Optimization",
                    "Image Processing",
                    "Convex Analysis",
                    "Compressive Sensing"
                ],
                "publications": [
                    {
                        "title": "A Fast Algorithm for Total Variation Image Reconstruction from Random Projections",
                        "abstract": "Total variation (TV) regularization is popular in image restoration and reconstruction due to its ability to preserve image edges. To date, most research activities on TV models concentrate on image restoration from blurry and noisy observations, while discussions on image reconstruction from random projections are relatively fewer. In this paper, we propose, analyze, and test a fast alternating minimization algorithm for image reconstruction from random projections via solving a TV regularized least-squares problem. The per-iteration cost of the proposed algorithm involves a linear time shrinkage operation, two matrix-vector multiplications and two fast Fourier transforms. Convergence, certain finite convergence and $q$-linear convergence results are established, which indicate that the asymptotic convergence speed of the proposed algorithm depends on the spectral radii of certain submatrix. Moreover, to speed up convergence and enhance robustness, we suggest an accelerated scheme based on an inexact alternating direction method. We present experimental results to compare with an existing algorithm, which indicate that the proposed algorithm is stable, efficient and competitive with TwIST \\cite{TWIST} -- a state-of-the art algorithm for solving TV regularization problems."
                    },
                    {
                        "title": "Optimal Nonergodic Sublinear Convergence Rate of Proximal Point Algorithm for Maximal Monotone Inclusion Problems",
                        "abstract": "We establish the optimal nonergodic sublinear convergence rate of the proximal point algorithm for maximal monotone inclusion problems. First, the optimal bound is formulated by the performance estimation framework, resulting in an infinite dimensional nonconvex optimization problem, which is then equivalently reformulated as a finite dimensional semidefinite programming (SDP) problem. By constructing a feasible solution to the dual SDP, we obtain an upper bound on the optimal nonergodic sublinear rate. Finally, an example in two dimensional space is constructed to provide a lower bound on the optimal nonergodic sublinear rate. Since the lower bound provided by the example matches exactly the upper bound obtained by the dual SDP, we have thus established the worst case nonergodic sublinear convergence rate which is optimal in terms of both the order as well as the constants involved. Our result sharpens the understanding of the fundamental proximal point algorithm."
                    },
                    {
                        "title": "On the dual step length of the alternating direction method of multipliers",
                        "abstract": "The alternating direction method of multipliers (ADMM) is a most widely used optimization scheme for solving linearly constrained separable convex optimization problems. The convergence of the ADMM can be guaranteed when the dual step length is less than the golden ratio, while plenty of numerical evidence suggests that even larger dual step length often accelerates the convergence. It has also been proved that the dual step length can be enlarged to less than 2 in some special cases, namely, one of the separable functions in the objective function is linear, or both are quadratic plus some additional assumptions. However, it remains unclear whether the golden ratio can be exceeded in the general convex setting. In this paper, the performance estimation framework is used to analyze the convergence of the ADMM, and assisted by numerical and symbolic computations, a counter example is constructed, which indicates that the conventional measure may lose monotonicity as the dual step length exceeds the golden ratio, ruling out the possibility of breaking the golden ratio within this conventional analytic framework."
                    },
                    {
                        "title": "On the optimal linear convergence factor of the relaxed proximal point algorithm for monotone inclusion problems",
                        "abstract": "Finding a zero of a maximal monotone operator is fundamental in convex optimization and monotone operator theory, and \\emph{proximal point algorithm} (PPA) is a primary method for solving this problem. PPA converges not only globally under fairly mild conditions but also asymptotically at a fast linear rate provided that the underlying inverse operator is Lipschitz continuous at the origin. These nice convergence properties are preserved by a relaxed variant of PPA. Recently, a linear convergence bound was established in [M. Tao, and X. M. Yuan, J. Sci. Comput., 74 (2018), pp. 826-850] for the relaxed PPA, and it was shown that the bound is optimal when the relaxation factor $\\gamma$ lies in $[1,2)$. However, for other choices of $\\gamma$, the bound obtained by Tao and Yuan is suboptimal. In this paper, we establish tight linear convergence bounds for any choice of $\\gamma\\in(0,2)$ and make the whole picture about optimal linear convergence bounds clear. These results sharpen our understandings to the asymptotic behavior of the relaxed PPA."
                    },
                    {
                        "title": "On the Optimal Ergodic Sublinear Convergence Rate of the Relaxed Proximal Point Algorithm for Variational Inequalities",
                        "abstract": "This paper investigates the optimal ergodic sublinear convergence rate of the relaxed proximal point algorithm for solving monotone variational inequality problems. The exact worst case convergence rate is computed using the performance estimation framework. It is observed that, as the number of iterations getting larger, this numerical rate asymptotically coincides with an existing sublinear rate, whose optimality is unknown. This hints that, without further assumptions, sublinear convergence rate is likely the best achievable rate. A concrete example is constructed, which provides a lower bound for the exact worst case convergence rate. Amazingly, this lower bound coincides with the exact worst case bound computed via the performance estimation framework. This observation motivates us to conjecture that the lower bound provided by the example is exactly the worse case iteration bound, which is then verified theoretically. We thus have established an ergodic sublinear convergence rate that is optimal in terms of both the order of the sublinear rate and all the constants involved."
                    },
                    {
                        "title": "Tight Convergence Rate in Subgradient Norm of the Proximal Point Algorithm",
                        "abstract": "Proximal point algorithm has found many applications, and it has been playing fundamental roles in the understanding, design, and analysis of many first-order methods. In this paper, we derive the tight convergence rate in subgradient norm of the proximal point algorithm, which was conjectured by Taylor, Hendrickx and Glineur [SIAM J.~Optim., 27 (2017), pp.~1283--1313]. This sort of convergence results in terms of the residual (sub)gradient norm is particularly interesting when considering dual methods, where the dual residual gradient norm corresponds to the primal distance to feasibility."
                    },
                    {
                        "title": "Alternating Direction Algorithms for $\\ell_1$-Problems in Compressive Sensing",
                        "abstract": "In this paper, we propose and study the use of alternating direction algorithms for several $\\ell_1$-norm minimization problems arising from sparse solution recovery in compressive sensing, including the basis pursuit problem, the basis-pursuit denoising problems of both unconstrained and constrained forms, as well as others. We present and investigate two classes of algorithms derived from either the primal or the dual forms of the $\\ell_1$-problems. The construction of the algorithms consists of two main steps: (1) to reformulate an $\\ell_1$-problem into one having partially separable objective functions by adding new variables and constraints; and (2) to apply an exact or inexact alternating direction method to the resulting problem. The derived alternating direction algorithms can be regarded as first-order primal-dual algorithms because both primal and dual variables are updated at each and every iteration. Convergence properties of these algorithms are established or restated when they already exist. Extensive numerical results in comparison with several state-of-the-art algorithms are given to demonstrate that the proposed algorithms are efficient, stable and robust. Moreover, we present numerical results to emphasize two practically important but perhaps overlooked points. One point is that algorithm speed should always be evaluated relative to appropriate solution accuracy; another is that whenever erroneous measurements possibly exist, the $\\ell_1$-norm fidelity should be the fidelity of choice in compressive sensing."
                    },
                    {
                        "title": "Applications of gauge duality in robust principal component analysis and semidefinite programming",
                        "abstract": "Gauge duality theory was originated by Freund [Math. Programming, 38(1):47-67, 1987] and was recently further investigated by Friedlander, Mac{\\^e}do and Pong [SIAM J. Optm., 24(4):1999-2022, 2014]. When solving some matrix optimization problems via gauge dual, one is usually able to avoid full matrix decompositions such as singular value and/or eigenvalue decompositions. In such an approach, a gauge dual problem is solved in the first stage, and then an optimal solution to the primal problem can be recovered from the dual optimal solution obtained in the first stage. Recently, this theory has been applied to a class of \\emph{semidefinite programming} (SDP) problems with promising numerical results [Friedlander and Mac{\\^e}do, SIAM J. Sci. Comp., to appear, 2016]. In this paper, we establish some theoretical results on applying the gauge duality theory to robust \\emph{principal component analysis} (PCA) and general SDP. For each problem, we present its gauge dual problem, characterize the optimality conditions for the primal-dual gauge pair, and validate a way to recover a primal optimal solution from a dual one. These results are extensions of [Friedlander and Mac{\\^e}do, SIAM J. Sci. Comp., to appear, 2016] from nuclear norm regularization to robust PCA and from a special class of SDP which requires the coefficient matrix in the linear objective to be positive definite to SDP problems without this restriction. Our results provide further understanding in the potential advantages and disadvantages of the gauge duality theory."
                    },
                    {
                        "title": "A golden ratio primal-dual algorithm for structured convex optimization",
                        "abstract": "We design, analyze and test a golden ratio primal-dual algorithm (GRPDA) for solving structured convex optimization problem, where the objective function is the sum of two closed proper convex functions, one of which involves a composition with a linear transform. GRPDA preserves all the favorable features of the classical primal-dual algorithm (PDA), i.e., the primal and the dual variables are updated in a Gauss-Seidel manner, and the per iteration cost is dominated by the evaluation of the proximal point mappings of the two component functions and two matrix-vector multiplications. Compared with the classical PDA, which takes an extrapolation step, the novelty of GRPDA is that it is constructed based on a convex combination of essentially the whole iteration trajectory. We show that GRPDA converges within a broader range of parameters than the classical PDA, provided that the reciprocal of the convex combination parameter is bounded above by the golden ratio, which explains the name of the algorithm. An O(1/N) ergodic convergence rate result is also established based on the primal-dual gap function, where N denotes the number of iterations. When either the primal or the dual problem is strongly convex, an accelerated GRPDA is constructed to improve the ergodic convergence rate from O(1/N) to O(1/N2). Moreover, we show for regularized least-squares and linear equality constrained problems that the reciprocal of the convex combination parameter can be extended from the golden ratio to 2 and meanwhile a relaxation step can be taken. Our preliminary numerical results on LASSO, nonnegative least-squares and minimax matrix game problems, with comparisons to some state-of-the-art relative algorithms, demonstrate the efficiency of the proposed algorithms."
                    },
                    {
                        "title": "A general inertial proximal point method for mixed variational inequality problem",
                        "abstract": "In this paper, we first propose a general inertial proximal point method for the mixed variational inequality (VI) problem. Based on our knowledge, without stronger assumptions, convergence rate result is not known in the literature for inertial type proximal point methods. Under certain conditions, we are able to establish the global convergence and a $o(1/k)$ convergence rate result (under certain measure) of the proposed general inertial proximal point method. We then show that the linearized alternating direction method of multipliers (ADMM) for separable convex optimization with linear constraints is an application of a general proximal point method, provided that the algorithmic parameters are properly chosen. As byproducts of this finding, we establish global convergence and $O(1/k)$ convergence rate results of the linearized ADMM in both ergodic and nonergodic sense. In particular, by applying the proposed inertial proximal point method for mixed VI to linearly constrained separable convex optimization, we obtain an inertial version of the linearized ADMM for which the global convergence is guaranteed. We also demonstrate the effect of the inertial extrapolation step via experimental results on the compressive principal component pursuit problem."
                    },
                    {
                        "title": "Golden ratio primal-dual algorithm with linesearch",
                        "abstract": "Golden ratio primal-dual algorithm (GRPDA) is a new variant of the classical Arrow-Hurwicz method for solving structured convex optimization problem, in which the objective function consists of the sum of two closed proper convex functions, one of which involves a composition with a linear transform. In this paper, we propose a linesearch strategy for GRPDA, which not only does not require the spectral norm of the linear transform but also allows adaptive and potentially much larger stepsizes. Within each linesearch step, only the dual variable needs to be updated, and it is thus quite cheap and does not require any extra matrix-vector multiplications for many special yet important applications, e.g., regularized least squares problem. Global convergence and ${\\cal O}(1/N)$ ergodic convergence rate results measured by the primal-dual gap function are established, where $N$ denotes the iteration counter. When one of the component functions is strongly convex, faster ${\\cal O}(1/N^2)$ ergodic convergence rate results are established by adaptively choosing some algorithmic parameters. Moreover, when both component functions are strongly convex, nonergodic linear converge results are established. Numerical experiments on matrix game and LASSO problems illustrate the effectiveness of the proposed linesearch strategy."
                    },
                    {
                        "title": "Inertial primal-dual algorithms for structured convex optimization",
                        "abstract": "The primal-dual algorithm recently proposed by Chambolle & Pock (abbreviated as CPA) for structured convex optimization is very efficient and popular. It was shown by Chambolle & Pock in \\cite{CP11} and also by Shefi & Teboulle in \\cite{ST14} that CPA and variants are closely related to preconditioned versions of the popular alternating direction method of multipliers (abbreviated as ADM). In this paper, we further clarify this connection and show that CPAs generate exactly the same sequence of points with the so-called linearized ADM (abbreviated as LADM) applied to either the primal problem or its Lagrangian dual, depending on different updating orders of the primal and the dual variables in CPAs, as long as the initial points for the LADM are properly chosen. The dependence on initial points for LADM can be relaxed by focusing on cyclically equivalent forms of the algorithms. Furthermore, by utilizing the fact that CPAs are applications of a general weighted proximal point method to the mixed variational inequality formulation of the KKT system, where the weighting matrix is positive definite under a parameter condition, we are able to propose and analyze inertial variants of CPAs. Under certain conditions, global point-convergence, nonasymptotic $O(1/k)$ and asymptotic $o(1/k)$ convergence rate of the proposed inertial CPAs can be guaranteed, where $k$ denotes the iteration index. Finally, we demonstrate the profits gained by introducing the inertial extrapolation step via experimental results on compressive image reconstruction based on total variation minimization."
                    }
                ]
            },
            "0946860d-0714-4970-91e2-4d86b106b09f": {
                "pk": "0946860d-0714-4970-91e2-4d86b106b09f",
                "name": "Baishakhi Ray",
                "collaborators": [
                    "Saikat Chakraborty",
                    "Rahul Krishna",
                    "Ziyuan Zhong",
                    "Yuchi Tian",
                    "Chengzhi Mao",
                    "Junfeng Yang",
                    "Vicente Ordonez",
                    "Vikram Nitin",
                    "Yangruibo Ding",
                    "Chong Tang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Deep Learning",
                    "Software Engineering",
                    "Adversarial Machine Learning"
                ],
                "publications": [
                    {
                        "title": "On Multi-Modal Learning of Editing Source Code",
                        "abstract": "In recent years, Neural Machine Translator (NMT) has shown promise in automatically editing source code. Typical NMT based code editor only considers the code that needs to be changed as input and suggests developers with a ranked list of patched code to choose from - where the correct one may not always be at the top of the list. While NMT based code editing systems generate a broad spectrum of plausible patches, the correct one depends on the developers' requirement and often on the context where the patch is applied. Thus, if developers provide some hints, using natural language, or providing patch context, NMT models can benefit from them. As a proof of concept, in this research, we leverage three modalities of information: edit location, edit code context, commit messages (as a proxy of developers' hint in natural language) to automatically generate edits with NMT models. To that end, we build MODIT, a multi-modal NMT based code editing engine. With in-depth investigation and analysis, we show that developers' hint as an input modality can narrow the search space for patches and outperform state-of-the-art models to generate correctly patched code in top-1 position."
                    },
                    {
                        "title": "DeepTest: Automated Testing of Deep-Neural-Network-driven Autonomous Cars",
                        "abstract": "Recent advances in Deep Neural Networks (DNNs) have led to the development of DNN-driven autonomous cars that, using sensors like camera, LiDAR, etc., can drive without any human intervention. Most major manufacturers including Tesla, GM, Ford, BMW, and Waymo/Google are working on building and testing different types of autonomous vehicles. The lawmakers of several US states including California, Texas, and New York have passed new legislation to fast-track the process of testing and deployment of autonomous vehicles on their roads.   However, despite their spectacular progress, DNNs, just like traditional software, often demonstrate incorrect or unexpected corner case behaviors that can lead to potentially fatal collisions. Several such real-world accidents involving autonomous cars have already happened including one which resulted in a fatality. Most existing testing techniques for DNN-driven vehicles are heavily dependent on the manual collection of test data under different driving conditions which become prohibitively expensive as the number of test conditions increases.   In this paper, we design, implement and evaluate DeepTest, a systematic testing tool for automatically detecting erroneous behaviors of DNN-driven vehicles that can potentially lead to fatal crashes. First, our tool is designed to automatically generated test cases leveraging real-world changes in driving conditions like rain, fog, lighting conditions, etc. DeepTest systematically explores different parts of the DNN logic by generating test inputs that maximize the numbers of activated neurons. DeepTest found thousands of erroneous behaviors under different realistic driving conditions (e.g., blurring, rain, fog, etc.) many of which lead to potentially fatal crashes in three top performing DNNs in the Udacity self-driving car challenge."
                    },
                    {
                        "title": "AdvSPADE: Realistic Unrestricted Attacks for Semantic Segmentation",
                        "abstract": "Due to the inherent robustness of segmentation models, traditional norm-bounded attack methods show limited effect on such type of models. In this paper, we focus on generating unrestricted adversarial examples for semantic segmentation models. We demonstrate a simple and effective method to generate unrestricted adversarial examples using conditional generative adversarial networks (CGAN) without any hand-crafted metric. The na\\\"ive implementation of CGAN, however, yields inferior image quality and low attack success rate. Instead, we leverage the SPADE (Spatially-adaptive denormalization) structure with an additional loss item to generate effective adversarial attacks in a single step. We validate our approach on the popular Cityscapes and ADE20K datasets, and demonstrate that our synthetic adversarial examples are not only realistic, but also improve the attack success rate by up to 41.0\\% compared with the state of the art adversarial attack methods including PGD."
                    },
                    {
                        "title": "Rebuttal to Berger et al., TOPLAS 2019",
                        "abstract": "Berger et al., published in TOPLAS 2019, is a critique of our 2014 FSE conference abstract and its archival version, the 2017 CACM paper: A Large-Scale Study of Programming Languages and Code Quality in Github. In their paper Berger et al. make academic claims about the veracity of our work. Here, we respond to their technical and scientific critiques aimed at our work, attempting to stick with scientific discourse. We find that Berger et al. largely replicated our results, and agree with us in their conclusion: that the effects (in a statistical sense) found in the data are small, and should be taken with caution, and that it is possible that an absence of effect is the correct interpretation. Thus, our CACM paper's conclusions still hold, even more so now that they have been reproduced, and our paper is eminently citable."
                    },
                    {
                        "title": "Variation of Gender Biases in Visual Recognition Models Before and After Finetuning",
                        "abstract": "We introduce a framework to measure how biases change before and after fine-tuning a large scale visual recognition model for a downstream task. Deep learning models trained on increasing amounts of data are known to encode societal biases. Many computer vision systems today rely on models typically pretrained on large scale datasets. While bias mitigation techniques have been developed for tuning models for downstream tasks, it is currently unclear what are the effects of biases already encoded in a pretrained model. Our framework incorporates sets of canonical images representing individual and pairs of concepts to highlight changes in biases for an array of off-the-shelf pretrained models across model sizes, dataset sizes, and training objectives. Through our analyses, we find that (1) supervised models trained on datasets such as ImageNet-21k are more likely to retain their pretraining biases regardless of the target dataset compared to self-supervised models. We also find that (2) models finetuned on larger scale datasets are more likely to introduce new biased associations. Our results also suggest that (3) biases can transfer to finetuned models and the finetuning objective and dataset can impact the extent of transferred biases."
                    },
                    {
                        "title": "SpecTra: Enhancing the Code Translation Ability of Language Models by Generating Multi-Modal Specifications",
                        "abstract": "Large language models (LLMs) are increasingly being used for the task of automated code translation, which has important real-world applications. However, most existing approaches use only the source code of a program as an input to an LLM, and do not consider the different kinds of specifications that can be extracted from a program. In this paper, we propose SpecTra, a multi-stage approach that uses a novel self-consistency filter to first generate high-quality static specifications, test cases, and natural language descriptions from a given program, and then uses these along with the source code to improve the quality of LLM-generated translations. We evaluate SpecTra on three code translation tasks - C to Rust, C to Go, and JavaScript to TypeScript - and show that it can enhance the performance of six popular LLMs on these tasks by up to 10 percentage points and a relative improvement of 26\\%. Our research suggests that generating high-quality specifications could be a promising and efficient way to improve the performance of LLMs for code translation. We make our code and data available, anonymized for review."
                    },
                    {
                        "title": "Solving Zebra Puzzles Using Constraint-Guided Multi-Agent Systems",
                        "abstract": "Prior research has enhanced the ability of Large Language Models (LLMs) to solve logic puzzles using techniques such as chain-of-thought prompting or introducing a symbolic representation. These frameworks are still usually insufficient to solve complicated logical problems, such as Zebra puzzles, due to the inherent complexity of translating natural language clues into logical statements. We introduce a multi-agent system, ZPS, that integrates LLMs with an off the shelf theorem prover. This system tackles the complex puzzle-solving task by breaking down the problem into smaller, manageable parts, generating SMT (Satisfiability Modulo Theories) code to solve them with a theorem prover, and using feedback between the agents to repeatedly improve their answers. We also introduce an automated grid puzzle grader to assess the correctness of our puzzle solutions and show that the automated grader is reliable by evaluating it in a user-study. Our approach shows improvement in all three LLMs we tested, with GPT-4 showing 166% improvement in the number of fully correct solutions."
                    },
                    {
                        "title": "CODIT: Code Editing with Tree-Based Neural Models",
                        "abstract": "The way developers edit day-to-day code tends to be repetitive, often using existing code elements. Many researchers have tried to automate repetitive code changes by learning from specific change templates which are applied to limited scope. The advancement of deep neural networks and the availability of vast open-source evolutionary data opens up the possibility of automatically learning those templates from the wild. However, deep neural network based modeling for code changes and code in general introduces some specific problems that needs specific attention from research community. For instance, compared to natural language, source code vocabulary can be significantly larger. Further, good changes in code do not break its syntactic structure. Thus, deploying state-of-the-art neural network models without adapting the methods to the source code domain yields sub-optimal results. To this end, we propose a novel tree-based neural network system to model source code changes and learn code change patterns from the wild. Specifically, we propose a tree-based neural machine translation model to learn the probability distribution of changes in code. We realize our model with a change suggestion engine, CODIT, and train the model with more than 24k real-world changes and evaluate it on 5k patches. Our evaluation shows the effectiveness of CODITin learning and suggesting patches. CODIT can also learn specific bug fix pattern from bug fixing patches and can fix 25 bugs out of 80 bugs in Defects4J."
                    },
                    {
                        "title": "ConEx: Efficient Exploration of Big-Data System Configurations for Better Performance",
                        "abstract": "Configuration space complexity makes the big-data software systems hard to configure well. Consider Hadoop, with over nine hundred parameters, developers often just use the default configurations provided with Hadoop distributions. The opportunity costs in lost performance are significant. Popular learning-based approaches to auto-tune software does not scale well for big-data systems because of the high cost of collecting training data. We present a new method based on a combination of Evolutionary Markov Chain Monte Carlo (EMCMC) sampling and cost reduction techniques to cost-effectively find better-performing configurations for big data systems. For cost reduction, we developed and experimentally tested and validated two approaches: using scaled-up big data jobs as proxies for the objective function for larger jobs and using a dynamic job similarity measure to infer that results obtained for one kind of big data problem will work well for similar problems. Our experimental results suggest that our approach promises to significantly improve the performance of big data systems and that it outperforms competing approaches based on random sampling, basic genetic algorithms (GA), and predictive model learning. Our experimental results support the conclusion that our approach has strongly demonstrated potential to significantly and cost-effectively improve the performance of big data systems."
                    },
                    {
                        "title": "Deep Learning based Vulnerability Detection: Are We There Yet?",
                        "abstract": "Automated detection of software vulnerabilities is a fundamental problem in software security. Existing program analysis techniques either suffer from high false positives or false negatives. Recent progress in Deep Learning (DL) has resulted in a surge of interest in applying DL for automated vulnerability detection. Several recent studies have demonstrated promising results achieving an accuracy of up to 95% at detecting vulnerabilities. In this paper, we ask, \"how well do the state-of-the-art DL-based techniques perform in a real-world vulnerability prediction scenario?\". To our surprise, we find that their performance drops by more than 50%. A systematic investigation of what causes such precipitous performance drop reveals that existing DL-based vulnerability prediction approaches suffer from challenges with the training data (e.g., data duplication, unrealistic distribution of vulnerable classes, etc.) and with the model choices (e.g., simple token-based models). As a result, these approaches often do not learn features related to the actual cause of the vulnerabilities. Instead, they learn unrelated artifacts from the dataset (e.g., specific variable/function names, etc.). Leveraging these empirical findings, we demonstrate how a more principled approach to data collection and model design, based on realistic settings of vulnerability prediction, can lead to better solutions. The resulting tools perform significantly better than the studied baseline: up to 33.57% boost in precision and 128.38% boost in recall compared to the best performing model in the literature. Overall, this paper elucidates existing DL-based vulnerability prediction systems' potential issues and draws a roadmap for future DL-based vulnerability prediction research. In that spirit, we make available all the artifacts supporting our results: https://git.io/Jf6IA."
                    },
                    {
                        "title": "Yuga: Automatically Detecting Lifetime Annotation Bugs in the Rust Language",
                        "abstract": "The Rust programming language is becoming increasingly popular among systems programmers due to its efficient performance and robust memory safety guarantees. Rust employs an ownership model to ensure this guarantee by allowing each value to be owned by only one identifier at a time. Additionally, it introduces the concept of borrowing and lifetimes to enable other variables to borrow the values under certain conditions temporarily. Despite its benefits, security vulnerabilities have been reported in Rust projects, often attributed to the use of \"unsafe\" Rust code. These vulnerabilities, in part, arise from incorrect lifetime annotations on function signatures. However, existing tools fail to detect these bugs, primarily because such bugs are rare, challenging to detect through dynamic analysis, and require explicit memory models. To overcome these limitations, first, we characterize incorrect lifetime annotations as a source of memory safety bugs and leverage this understanding to devise a novel static analysis tool, Yuga, to detect potential lifetime annotation bugs. Yuga uses a multi-phase analysis approach, starting with a quick pattern-matching algorithm to identify potential buggy components and then conducting a flow and field-sensitive alias analysis to confirm the bugs. We also curate new datasets of lifetime annotation bugs. Yuga successfully detects bugs with good precision on these datasets, and we make the code and datasets publicly available for review."
                    },
                    {
                        "title": "Understanding Local Robustness of Deep Neural Networks under Natural Variations",
                        "abstract": "Deep Neural Networks (DNNs) are being deployed in a wide range of settings today, from safety-critical applications like autonomous driving to commercial applications involving image classifications. However, recent research has shown that DNNs can be brittle to even slight variations of the input data. Therefore, rigorous testing of DNNs has gained widespread attention.   While DNN robustness under norm-bound perturbation got significant attention over the past few years, our knowledge is still limited when natural variants of the input images come. These natural variants, e.g. a rotated or a rainy version of the original input, are especially concerning as they can occur naturally in the field without any active adversary and may lead to undesirable consequences. Thus, it is important to identify the inputs whose small variations may lead to erroneous DNN behaviors. The very few studies that looked at DNN's robustness under natural variants, however, focus on estimating the overall robustness of DNNs across all the test data rather than localizing such error-producing points. This work aims to bridge this gap.   To this end, we study the local per-input robustness properties of the DNNs and leverage those properties to build a white-box (DeepRobust-W) and a black-box (DeepRobust-B) tool to automatically identify the non-robust points. Our evaluation of these methods on three DNN models spanning three widely used image classification datasets shows that they are effective in flagging points of poor robustness. In particular, DeepRobust-W and DeepRobust-B are able to achieve an F1 score of up to 91.4% and 99.1%, respectively. We further show that DeepRobust-W can be applied to a regression problem in another domain. Our evaluation on three self-driving car models demonstrates that DeepRobust-W is effective in identifying points of poor robustness with F1 score up to 78.9%."
                    },
                    {
                        "title": "Neural Network Guided Evolutionary Fuzzing for Finding Traffic Violations of Autonomous Vehicles",
                        "abstract": "Self-driving cars and trucks, autonomous vehicles (AVs), should not be accepted by regulatory bodies and the public until they have much higher confidence in their safety and reliability -- which can most practically and convincingly be achieved by testing. But existing testing methods are inadequate for checking the end-to-end behaviors of AV controllers against complex, real-world corner cases involving interactions with multiple independent agents such as pedestrians and human-driven vehicles. While test-driving AVs on streets and highways fails to capture many rare events, existing simulation-based testing methods mainly focus on simple scenarios and do not scale well for complex driving situations that require sophisticated awareness of the surroundings. To address these limitations, we propose a new fuzz testing technique, called AutoFuzz, which can leverage widely-used AV simulators' API grammars to generate semantically and temporally valid complex driving scenarios (sequences of scenes). To efficiently search for traffic violations-inducing scenarios in a large search space, we propose a constrained neural network (NN) evolutionary search method to optimize AutoFuzz. Evaluation of our prototype on one state-of-the-art learning-based controller, two rule-based controllers, and one industrial-grade controller in five scenarios shows that AutoFuzz efficiently finds hundreds of traffic violations in high-fidelity simulation environments. For each scenario, AutoFuzz can find on average 10-39% more unique traffic violations than the best-performing baseline method. Further, fine-tuning the learning-based controller with the traffic violations found by AutoFuzz successfully reduced the traffic violations found in the new version of the AV controller software."
                    },
                    {
                        "title": "Obfuscation Resilient Search through Executable Classification",
                        "abstract": "Android applications are usually obfuscated before release, making it difficult to analyze them for malware presence or intellectual property violations. Obfuscators might hide the true intent of code by renaming variables and/or modifying program structures. It is challenging to search for executables relevant to an obfuscated application for developers to analyze efficiently. Prior approaches toward obfuscation resilient search have relied on certain structural parts of apps remaining as landmarks, un-touched by obfuscation. For instance, some prior approaches have assumed that the structural relationships between identifiers are not broken by obfuscators; others have assumed that control flow graphs maintain their structures. Both approaches can be easily defeated by a motivated obfuscator. We present a new approach,Macneto, to search for programs relevant to obfuscated executables leveraging deep learning and principal components on instructions. Macneto makes few assumptions about the kinds of modifications that an obfuscator might perform. We show that it has high search precision for executables obfuscated by a state-of-the-art obfuscator that changes control flow. Further, we also demonstrate the potential of Macneto to help developers understand executables, where Macneto infers keywords (which are from the relevant unobfuscated program) for obfuscated executables."
                    },
                    {
                        "title": "Interpreted Formalisms for Configurations",
                        "abstract": "Imprecise and incomplete specification of system \\textit{configurations} threatens safety, security, functionality, and other critical system properties and uselessly enlarges the configuration spaces to be searched by configuration engineers and auto-tuners. To address these problems, this paper introduces \\textit{interpreted formalisms based on real-world types for configurations}. Configuration values are lifted to values of real-world types, which we formalize as \\textit{subset types} in Coq. Values of these types are dependent pairs whose components are values of underlying Coq types and proofs of additional properties about them. Real-world types both extend and further constrain \\textit{machine-level} configurations, enabling richer, proof-based checking of their consistency with real-world constraints. Tactic-based proof scripts are written once to automate the construction of proofs, if proofs exist, for configuration fields and whole configurations. \\textit{Failures to prove} reveal real-world type errors. Evaluation is based on a case study of combinatorial optimization of Hadoop performance by meta-heuristic search over Hadoop configurations spaces."
                    },
                    {
                        "title": "Pythia: Grammar-Based Fuzzing of REST APIs with Coverage-guided Feedback and Learning-based Mutations",
                        "abstract": "This paper introduces Pythia, the first fuzzer that augments grammar-based fuzzing with coverage-guided feedback and a learning-based mutation strategy for stateful REST API fuzzing. Pythia uses a statistical model to learn common usage patterns of a target REST API from structurally valid seed inputs. It then generates learning-based mutations by injecting a small amount of noise deviating from common usage patterns while still maintaining syntactic validity. Pythia's mutation strategy helps generate grammatically valid test cases and coverage-guided feedback helps prioritize the test cases that are more likely to find bugs. We present experimental evaluation on three production-scale, open-source cloud services showing that Pythia outperforms prior approaches both in code coverage and new bugs found. Using Pythia, we found 29 new bugs which we are in the process of reporting to the respective service owners."
                    },
                    {
                        "title": "Metric Learning for Adversarial Robustness",
                        "abstract": "Deep networks are well-known to be fragile to adversarial attacks. We conduct an empirical analysis of deep representations under the state-of-the-art attack method called PGD, and find that the attack causes the internal representation to shift closer to the \"false\" class. Motivated by this observation, we propose to regularize the representation space under attack with metric learning to produce more robust classifiers. By carefully sampling examples for metric learning, our learned representation not only increases robustness, but also detects previously unseen adversarial samples. Quantitative experiments show improvement of robustness accuracy by up to 4% and detection efficiency by up to 6% according to Area Under Curve score over prior work. The code of our work is available at https://github.com/columbia/Metric_Learning_Adversarial_Robustness."
                    },
                    {
                        "title": "Predictive Synthesis of API-Centric Code",
                        "abstract": "Today's programmers, especially data science practitioners, make heavy use of data-processing libraries (APIs) such as PyTorch, Tensorflow, NumPy, Pandas, and the like. Program synthesizers can provide significant coding assistance to this community of users; however program synthesis also can be slow due to enormous search spaces. In this work, we examine ways in which machine learning can be used to accelerate enumerative program synthesis. We present a deep-learning-based model to predict the sequence of API functions that would be needed to go from a given input to a desired output, both being numeric vectors. Our work is based on two insights. First, it is possible to learn, based on a large number of input-output examples, to predict the likely API function needed in a given situation. Second, and crucially, it is also possible to learn to compose API functions into a sequence, given an input and the desired final output, without explicitly knowing the intermediate values. We show that we can speed up an enumerative program synthesizer by using predictions from our model variants. These speedups significantly outperform previous ways (e.g. DeepCoder) in which researchers have used ML models in enumerative synthesis."
                    },
                    {
                        "title": "On ML-Based Program Translation: Perils and Promises",
                        "abstract": "With the advent of new and advanced programming languages, it becomes imperative to migrate legacy software to new programming languages. Unsupervised Machine Learning-based Program Translation could play an essential role in such migration, even without a sufficiently sizeable reliable corpus of parallel source code. However, these translators are far from perfect due to their statistical nature. This work investigates unsupervised program translators and where and why they fail. With in-depth error analysis of such failures, we have identified that the cases where such translators fail follow a few particular patterns. With this insight, we develop a rule-based program mutation engine, which pre-processes the input code if the input follows specific patterns and post-process the output if the output follows certain patterns. We show that our code processing tool, in conjunction with the program translator, can form a hybrid program translator and significantly improve the state-of-the-art. In the future, we envision an end-to-end program translation tool where programming domain knowledge can be embedded into an ML-based translation pipeline using pre- and post-processing steps."
                    },
                    {
                        "title": "PropTest: Automatic Property Testing for Improved Visual Programming",
                        "abstract": "Visual Programming has recently emerged as an alternative to end-to-end black-box visual reasoning models. This type of method leverages Large Language Models (LLMs) to generate the source code for an executable computer program that solves a given problem. This strategy has the advantage of offering an interpretable reasoning path and does not require finetuning a model with task-specific data. We propose PropTest, a general strategy that improves visual programming by further using an LLM to generate code that tests for visual properties in an initial round of proposed solutions. Our method generates tests for data-type consistency, output syntax, and semantic properties. PropTest achieves comparable results to state-of-the-art methods while using publicly available LLMs. This is demonstrated across different benchmarks on visual question answering and referring expression comprehension. Particularly, PropTest improves ViperGPT by obtaining 46.1\\% accuracy (+6.0\\%) on GQA using Llama3-8B and 59.5\\% (+8.1\\%) on RefCOCO+ using CodeLlama-34B."
                    }
                ]
            }
        }
    },
    "2405.04669": {
        "paper_data": {
            "title": "Towards a Theoretical Understanding of the 'Reversal Curse' via Training Dynamics",
            "url": "http://arxiv.org/abs/2405.04669v1",
            "arxiv_id": "2405.04669",
            "authors": [
                "Hanlin Zhu",
                "Baihe Huang",
                "Shaolun Zhang",
                "Michael Jordan",
                "Jiantao Jiao",
                "Yuandong Tian",
                "Stuart Russell"
            ],
            "abstract": "Auto-regressive large language models (LLMs) show impressive capacities to solve many complex reasoning tasks while struggling with some simple logical reasoning tasks such as inverse search: when trained on ''A is B'', LLM fails to directly conclude ''B is A'' during inference, which is known as the ''reversal curse'' (Berglund et al., 2023). In this paper, we theoretically analyze the reversal curse via the training dynamics of (stochastic) gradient descent for two auto-regressive models: (1) a bilinear model that can be viewed as a simplification of a one-layer transformer; (2) one-layer transformers using the framework of Tian et al. (2023a). Our analysis reveals a core reason why the reversal curse happens: the (effective) weights of both auto-regressive models show asymmetry, i.e., the increase of weights from a token $A$ to token $B$ during training does not necessarily cause the increase of the weights from $B$ to $A$. Moreover, our analysis can be naturally applied to other logical reasoning tasks such as chain-of-thought (COT) (Wei et al., 2022b). We show the necessity of COT, i.e., a model trained on ''$A \\to B$'' and ''$B \\to C$'' fails to directly conclude ''$A \\to C$'' without COT (also empirically observed by Allen-Zhu and Li (2023)), for one-layer transformers via training dynamics, which provides a new perspective different from previous work (Feng et al., 2024) that focuses on expressivity. Finally, we also conduct experiments to validate our theory on multi-layer transformers under different settings.",
            "introduction": "   1 Introductions  Large language models (LLMs) have shown great performance in solving complex reasoning tasks that require multiple reasoning steps through in-context learning (ICL), such as zero-shot learning\u00a0(Reynolds and McDonell, 2021; Kojima et\u00a0al., 2022), few-shot learning\u00a0(Brown et\u00a0al., 2020; Wei et\u00a0al., 2022b; Wang et\u00a0al., 2022), or via further fine-tuning\u00a0(Nye et\u00a0al., 2021; Cobbe et\u00a0al., 2021; Zelikman et\u00a0al., 2022). However, without the above inference-time techniques or model fine-tuning (probably combined with data manipulations), an auto-regressive LLM might struggle with simple logical reasoning tasks that require multiple reasoning steps learned during training separately or require simple manipulation of the learned knowledge\u00a0(Allen-Zhu and Li, 2023), where the reversal curse\u00a0(Berglund et\u00a0al., 2023) serves as a well-known example.   Reversal curse\u00a0(Berglund et\u00a0al., 2023) refers to the phenomenon that an auto-regressive LLM that learns \u201cA is B\u201d during training fails to generalize to the reverse direction \u201cB is A\u201d, and this task is also termed as \u201cinverse search\u201d in Allen-Zhu and Li (2023). Although some previous works propose different methods to mitigate the reversal curse, including reversing the training dataset\u00a0(Guo et\u00a0al., 2024; Golovneva et\u00a0al., 2024) and training on different objectives such as autoregressive blank infilling\u00a0(Lv et\u00a0al., 2023), these methods might negatively affect the model performance on other tasks since they either alter the dataset or the model architecture. Without dataset manipulation or changing the auto-regressive nature (causal structure) of the model, the reversal curse is hard to mitigate even with ICL strategies such as chain-of-thought\u00a0(Allen-Zhu and Li, 2023; Guo et\u00a0al., 2024).   In this paper, we aim to theoretically study why the reversal curse happens for auto-regressive LLMs. Different from previous work that studies the capacity of (transformer-based\u00a0(Vaswani et\u00a0al., 2017)) LLMs through the lens of expressivity (e.g., Yun et\u00a0al. (2019); P\u00e9rez et\u00a0al. (2021); Feng et\u00a0al. (2024)), reversal curse cannot be explained by expressivity since a model can express \u201cA is B\u201d is also able to express \u201cB is A\u201d. Therefore, we analyze the reversal curse via training dynamics since even if a set of parameters can express a fact in both directions, it might not be reachable through popular training algorithms (e.g., gradient descent, AdamW\u00a0(Loshchilov and Hutter, 2017)) with training data only presented in one direction. We summarize our main contributions as follows:     \u2022  We theoretically analyze reversal curse where training or test sequences have the from \u201cA\u2192B\u2192\ud835\udc34\ud835\udc35A\\to Bitalic_A \u2192 italic_B\u201d or \u201cB\u2190A\u2190\ud835\udc35\ud835\udc34B\\leftarrow Aitalic_B \u2190 italic_A\u201d via training dynamics of (stochastic) gradient descent under two auto-regressive models: a bilinear model (Section\u00a03) and one-layer transformers under certain assumptions similar to Tian et\u00a0al. (2023a) (Section\u00a04). The analysis of the training dynamics of both models reveals a core reason why the reversal curse happens: the weights of the autoregressive models are asymmetric, i.e., the increase of weights from the token A\ud835\udc34Aitalic_A to token B\ud835\udc35Bitalic_B222The weights from A\ud835\udc34Aitalic_A to B\ud835\udc35Bitalic_B can be viewed as the logits of token B\ud835\udc35Bitalic_B when the input is A\ud835\udc34Aitalic_A. during training does not necessarily cause the increase of the weights from B\ud835\udc35Bitalic_B to A\ud835\udc34Aitalic_A. We emphasize that during theoretical analysis, we used reparameterization. Thus, the weights from A\ud835\udc34Aitalic_A to B\ud835\udc35Bitalic_B do not necessarily correspond to the actual model weights and should be viewed as effective",
            "references": [
                {
                    "title": "Reverse Training to Nurse the Reversal Curse",
                    "abstract": "Large language models (LLMs) have a surprising failure: when trained on\"A has a feature B\", they do not generalize to\"B is a feature of A\", which is termed the Reversal Curse. Even when training with trillions of tokens this issue still appears due to Zipf's law - hence even if we train on the entire internet. This work proposes an alternative training scheme, called reverse training, whereby all words are used twice, doubling the amount of available tokens. The LLM is trained in both forward and reverse directions by reversing the training strings while preserving (i.e., not reversing) chosen substrings, such as entities. We show that data-matched reverse-trained models provide superior performance to standard models on standard tasks, and compute-matched reverse-trained models provide far superior performance on reversal tasks, helping resolve the reversal curse issue."
                },
                {
                    "title": "How Transformers Learn Causal Structure with Gradient Descent",
                    "abstract": "The incredible success of transformers on sequence modeling tasks can be largely attributed to the self-attention mechanism, which allows information to be transferred between different parts of a sequence. Self-attention allows transformers to encode causal structure which makes them particularly suitable for sequence modeling. However, the process by which transformers learn such causal structure via gradient-based training algorithms remains poorly understood. To better understand this process, we introduce an in-context learning task that requires learning latent causal structure. We prove that gradient descent on a simplified two-layer transformer learns to solve this task by encoding the latent causal graph in the first attention layer. The key insight of our proof is that the gradient of the attention matrix encodes the mutual information between tokens. As a consequence of the data processing inequality, the largest entries of this gradient correspond to edges in the latent causal graph. As a special case, when the sequences are generated from in-context Markov chains, we prove that transformers learn an induction head (Olsson et al., 2022). We confirm our theoretical findings by showing that transformers trained on our in-context learning task are able to recover a wide variety of causal structures."
                },
                {
                    "title": "A Mechanistic Analysis of a Transformer Trained on a Symbolic Multi-Step Reasoning Task",
                    "abstract": "Transformers demonstrate impressive performance on a range of reasoning benchmarks. To evaluate the degree to which these abilities are a result of actual reasoning, existing work has focused on developing sophisticated benchmarks for behavioral studies. However, these studies do not provide insights into the internal mechanisms driving the observed capabilities. To improve our understanding of the internal mechanisms of transformers, we present a comprehensive mechanistic analysis of a transformer trained on a synthetic reasoning task. We identify a set of interpretable mechanisms the model uses to solve the task, and validate our findings using correlational and causal evidence. Our results suggest that it implements a depth-bounded recurrent mechanisms that operates in parallel and stores intermediate results in selected token positions. We anticipate that the motifs we identified in our synthetic setting can provide valuable insights into the broader operating principles of transformers and thus provide a basis for understanding more complex models."
                },
                {
                    "title": "Transformer-Based Models Are Not Yet Perfect At Learning to Emulate Structural Recursion",
                    "abstract": "This paper investigates the ability of transformer-based models to learn structural recursion from examples. Recursion is a universal concept in both natural and formal languages. Structural recursion is central to the programming language and formal mathematics tasks where symbolic tools currently excel beyond neural models, such as inferring semantic relations between datatypes and emulating program behavior. We introduce a general framework that nicely connects the abstract concepts of structural recursion in the programming language domain to concrete sequence modeling problems and learned models' behavior. The framework includes a representation that captures the general \\textit{syntax} of structural recursion, coupled with two different frameworks for understanding their \\textit{semantics} -- one that is more natural from a programming languages perspective and one that helps bridge that perspective with a mechanistic understanding of the underlying transformer architecture. With our framework as a powerful conceptual tool, we identify different issues under various set-ups. The models trained to emulate recursive computations cannot fully capture the recursion yet instead fit short-cut algorithms and thus cannot solve certain edge cases that are under-represented in the training distribution. In addition, it is difficult for state-of-the-art large language models (LLMs) to mine recursive rules from in-context demonstrations. Meanwhile, these LLMs fail in interesting ways when emulating reduction (step-wise computation) of the recursive function."
                },
                {
                    "title": "Chain of History: Learning and Forecasting with LLMs for Temporal Knowledge Graph Completion",
                    "abstract": "Temporal Knowledge Graph Completion (TKGC) is a complex task involving the prediction of missing event links at future timestamps by leveraging established temporal structural knowledge. This paper aims to provide a comprehensive perspective on harnessing the advantages of Large Language Models (LLMs) for reasoning in temporal knowledge graphs, presenting an easily transferable pipeline. In terms of graph modality, we underscore the LLMs' prowess in discerning the structural information of pivotal nodes within the historical chain. As for the generation mode of the LLMs utilized for inference, we conduct an exhaustive exploration into the variances induced by a range of inherent factors in LLMs, with particular attention to the challenges in comprehending reverse logic. We adopt a parameter-efficient fine-tuning strategy to harmonize the LLMs with the task requirements, facilitating the learning of the key knowledge highlighted earlier. Comprehensive experiments are undertaken on several widely recognized datasets, revealing that our framework exceeds or parallels existing methods across numerous popular metrics. Additionally, we execute a substantial range of ablation experiments and draw comparisons with several advanced commercial LLMs, to investigate the crucial factors influencing LLMs' performance in structured temporal knowledge inference tasks."
                },
                {
                    "title": "Are We Falling in a Middle-Intelligence Trap? An Analysis and Mitigation of the Reversal Curse",
                    "abstract": "Recent studies have highlighted a phenomenon in large language models (LLMs) known as\"the reversal curse,\"in which the order of knowledge entities in the training data biases the models' comprehension. For example, if a model is trained on sentences where entity A consistently appears before entity B, it can respond to queries about A by providing B as the answer. However, it may encounter confusion when presented with questions concerning B. We contend that the reversal curse is partially a result of specific model training objectives, particularly evident in the prevalent use of the next-token prediction within most causal language models. For the next-token prediction, models solely focus on a token's preceding context, resulting in a restricted comprehension of the input. In contrast, we illustrate that the GLM, trained using the autoregressive blank infilling objective where tokens to be predicted have access to the entire context, exhibits better resilience against the reversal curse. We propose a novel training method, BIdirectional Casual language modeling Optimization (BICO), designed to mitigate the reversal curse when fine-tuning pretrained causal language models on new data. BICO modifies the causal attention mechanism to function bidirectionally and employs a mask denoising optimization. In the task designed to assess the reversal curse, our approach improves Llama's accuracy from the original 0% to around 70%. We hope that more attention can be focused on exploring and addressing these inherent weaknesses of the current LLMs, in order to achieve a higher level of intelligence."
                },
                {
                    "title": "Transformers as Decision Makers: Provable In-Context Reinforcement Learning via Supervised Pretraining",
                    "abstract": "Large transformer models pretrained on offline reinforcement learning datasets have demonstrated remarkable in-context reinforcement learning (ICRL) capabilities, where they can make good decisions when prompted with interaction trajectories from unseen environments. However, when and how transformers can be trained to perform ICRL have not been theoretically well-understood. In particular, it is unclear which reinforcement-learning algorithms transformers can perform in context, and how distribution mismatch in offline training data affects the learned algorithms. This paper provides a theoretical framework that analyzes supervised pretraining for ICRL. This includes two recently proposed training methods -- algorithm distillation and decision-pretrained transformers. First, assuming model realizability, we prove the supervised-pretrained transformer will imitate the conditional expectation of the expert algorithm given the observed trajectory. The generalization error will scale with model capacity and a distribution divergence factor between the expert and offline algorithms. Second, we show transformers with ReLU attention can efficiently approximate near-optimal online reinforcement learning algorithms like LinUCB and Thompson sampling for stochastic linear bandits, and UCB-VI for tabular Markov decision processes. This provides the first quantitative analysis of the ICRL capabilities of transformers pretrained from offline trajectories."
                },
                {
                    "title": "In-Context Convergence of Transformers",
                    "abstract": "Transformers have recently revolutionized many domains in modern machine learning and one salient discovery is their remarkable in-context learning capability, where models can solve an unseen task by utilizing task-specific prompts without further parameters fine-tuning. This also inspired recent theoretical studies aiming to understand the in-context learning mechanism of transformers, which however focused only on linear transformers. In this work, we take the first step toward studying the learning dynamics of a one-layer transformer with softmax attention trained via gradient descent in order to in-context learn linear function classes. We consider a structured data model, where each token is randomly sampled from a set of feature vectors in either balanced or imbalanced fashion. For data with balanced features, we establish the finite-time convergence guarantee with near-zero prediction error by navigating our analysis over two phases of the training dynamics of the attention map. More notably, for data with imbalanced features, we show that the learning dynamics take a stage-wise convergence process, where the transformer first converges to a near-zero prediction error for the query tokens of dominant features, and then converges later to a near-zero prediction error for the query tokens of under-represented features, respectively via one and four training phases. Our proof features new techniques for analyzing the competing strengths of two types of attention weights, the change of which determines different training phases."
                },
                {
                    "title": "An Investigation of LLMs' Inefficacy in Understanding Converse Relations",
                    "abstract": "Large Language Models (LLMs) have achieved remarkable success in many formal language oriented tasks, such as structural data-to-text and semantic parsing. However current benchmarks mostly follow the data distribution of the pre-training data of LLMs. Therefore, a natural question rises that do LLMs really understand the structured semantics of formal languages. In this paper, we investigate this problem on a special case, converse binary relation. We introduce a new benchmark ConvRe focusing on converse relations, which contains 17 relations and 1240 triples extracted from popular knowledge graph completion datasets. Our ConvRE features two tasks, Re2Text and Text2Re, which are formulated as multi-choice question answering to evaluate LLMs' ability to determine the matching between relations and associated text. For the evaluation protocol, apart from different prompting methods, we further introduce variants to the test text and few-shot example text. We conduct experiments on three popular LLM families and have observed various scaling trends. The results suggest that LLMs often resort to shortcut learning and still face challenges on our proposed benchmark."
                },
                {
                    "title": "JoMA: Demystifying Multilayer Transformers via JOint Dynamics of MLP and Attention",
                    "abstract": "We propose Joint MLP/Attention (JoMA) dynamics, a novel mathematical framework to understand the training procedure of multilayer Transformer architectures. This is achieved by integrating out the self-attention layer in Transformers, producing a modified dynamics of MLP layers only. JoMA removes unrealistic assumptions in previous analysis (e.g., lack of residual connection) and predicts that the attention first becomes sparse (to learn salient tokens), then dense (to learn less salient tokens) in the presence of nonlinear activations, while in the linear case, it is consistent with existing works that show attention becomes sparse over time. We leverage JoMA to qualitatively explains how tokens are combined to form hierarchies in multilayer Transformers, when the input tokens are generated by a latent hierarchical generative model. Experiments on models trained from real-world dataset (Wikitext2/Wikitext103) and various pre-trained models (OPT, Pythia) verify our theoretical findings. Code can be found in https://github.com/facebookresearch/luckmatters/tree/yuandong3."
                },
                {
                    "title": "Physics of Language Models: Part 3.2, Knowledge Manipulation",
                    "abstract": "Language models can store vast factual knowledge, yet their ability to flexibly use this knowledge for downstream tasks (e.g., via instruction finetuning) remains questionable. This paper investigates four fundamental knowledge manipulation tasks: retrieval (e.g.,\"What is person A's attribute X?\"), classification (e.g.,\"Is A's attribute X even or odd?\"), comparison (e.g.,\"Is A greater than B in attribute X?\"), and inverse search (e.g.,\"Which person's attribute X equals T?\"). We show that language models excel in knowledge retrieval but struggle even in the simplest classification or comparison tasks unless Chain of Thoughts (CoTs) are employed during both training and inference. Moreover, their performance in inverse knowledge search is virtually 0%, regardless of the prompts. Our primary contribution is a controlled, synthetic experiment that confirms these weaknesses are inherent to language models: they cannot efficiently manipulate knowledge from pre-training data, even when such knowledge is perfectly stored in the models, despite adequate training and sufficient model size. Our findings also apply to modern pretrained language models such as GPT-4, thus giving rise to many Turing tests to distinguish Humans from contemporary AIs."
                },
                {
                    "title": "The Reversal Curse: LLMs trained on \"A is B\" fail to learn \"B is A\"",
                    "abstract": "We expose a surprising failure of generalization in auto-regressive large language models (LLMs). If a model is trained on a sentence of the form\"A is B\", it will not automatically generalize to the reverse direction\"B is A\". This is the Reversal Curse. For instance, if a model is trained on\"Valentina Tereshkova was the first woman to travel to space\", it will not automatically be able to answer the question,\"Who was the first woman to travel to space?\". Moreover, the likelihood of the correct answer (\"Valentina Tershkova\") will not be higher than for a random name. Thus, models do not generalize a prevalent pattern in their training set: if\"A is B\"occurs,\"B is A\"is more likely to occur. It is worth noting, however, that if\"A is B\"appears in-context, models can deduce the reverse relationship. We provide evidence for the Reversal Curse by finetuning GPT-3 and Llama-1 on fictitious statements such as\"Uriah Hawthorne is the composer of Abyssal Melodies\"and showing that they fail to correctly answer\"Who composed Abyssal Melodies?\". The Reversal Curse is robust across model sizes and model families and is not alleviated by data augmentation. We also evaluate ChatGPT (GPT-3.5 and GPT-4) on questions about real-world celebrities, such as\"Who is Tom Cruise's mother? [A: Mary Lee Pfeiffer]\"and the reverse\"Who is Mary Lee Pfeiffer's son?\". GPT-4 correctly answers questions like the former 79% of the time, compared to 33% for the latter. Code available at: https://github.com/lukasberglund/reversal_curse."
                },
                {
                    "title": "Deep Networks as Denoising Algorithms: Sample-Efficient Learning of Diffusion Models in High-Dimensional Graphical Models",
                    "abstract": "We investigate the approximation efficiency of score functions by deep neural networks in diffusion-based generative modeling. While existing approximation theories utilize the smoothness of score functions, they suffer from the curse of dimensionality for intrinsically high-dimensional data. This limitation is pronounced in graphical models such as Markov random fields, common for image distributions, where the approximation efficiency of score functions remains unestablished. To address this, we observe score functions can often be well-approximated in graphical models through variational inference denoising algorithms. Furthermore, these algorithms are amenable to efficient neural network representation. We demonstrate this in examples of graphical models, including Ising models, conditional Ising models, restricted Boltzmann machines, and sparse encoding models. Combined with off-the-shelf discretization error bounds for diffusion-based sampling, we provide an efficient sample complexity bound for diffusion-based generative modeling when the score function is learned by deep neural networks."
                },
                {
                    "title": "What can a Single Attention Layer Learn? A Study Through the Random Features Lens",
                    "abstract": "Attention layers -- which map a sequence of inputs to a sequence of outputs -- are core building blocks of the Transformer architecture which has achieved significant breakthroughs in modern artificial intelligence. This paper presents a rigorous theoretical study on the learning and generalization of a single multi-head attention layer, with a sequence of key vectors and a separate query vector as input. We consider the random feature setting where the attention layer has a large number of heads, with randomly sampled frozen query and key matrices, and trainable value matrices. We show that such a random-feature attention layer can express a broad class of target functions that are permutation invariant to the key vectors. We further provide quantitative excess risk bounds for learning these target functions from finite samples, using random feature attention with finitely many heads. Our results feature several implications unique to the attention structure compared with existing random features theory for neural networks, such as (1) Advantages in the sample complexity over standard two-layer random-feature networks; (2) Concrete and natural classes of functions that can be learned efficiently by a random-feature attention layer; and (3) The effect of the sampling distribution of the query-key weight matrix (the product of the query and key matrix), where Gaussian random weights with a non-zero mean result in better sample complexities over the zero-mean counterpart for learning certain natural target functions. Experiments on simulated data corroborate our theoretical findings and further illustrate the interplay between the sample size and the complexity of the target function."
                },
                {
                    "title": "One Step of Gradient Descent is Provably the Optimal In-Context Learner with One Layer of Linear Self-Attention",
                    "abstract": "Recent works have empirically analyzed in-context learning and shown that transformers trained on synthetic linear regression tasks can learn to implement ridge regression, which is the Bayes-optimal predictor, given sufficient capacity [Aky\\\"urek et al., 2023], while one-layer transformers with linear self-attention and no MLP layer will learn to implement one step of gradient descent (GD) on a least-squares linear regression objective [von Oswald et al., 2022]. However, the theory behind these observations remains poorly understood. We theoretically study transformers with a single layer of linear self-attention, trained on synthetic noisy linear regression data. First, we mathematically show that when the covariates are drawn from a standard Gaussian distribution, the one-layer transformer which minimizes the pre-training loss will implement a single step of GD on the least-squares linear regression objective. Then, we find that changing the distribution of the covariates and weight vector to a non-isotropic Gaussian distribution has a strong impact on the learned algorithm: the global minimizer of the pre-training loss now implements a single step of $\\textit{pre-conditioned}$ GD. However, if only the distribution of the responses is changed, then this does not have a large effect on the learned algorithm: even when the response comes from a more general family of $\\textit{nonlinear}$ functions, the global minimizer of the pre-training loss still implements a single step of GD on a least-squares linear regression objective."
                },
                {
                    "title": "A Survey on Evaluation of Large Language Models",
                    "abstract": "Large language models (LLMs) are gaining increasing popularity in both academia and industry, owing to their unprecedented performance in various applications. As LLMs continue to play a vital role in both research and daily use, their evaluation becomes increasingly critical, not only at the task level, but also at the society level for better understanding of their potential risks. Over the past years, significant efforts have been made to examine LLMs from various perspectives. This paper presents a comprehensive review of these evaluation methods for LLMs, focusing on three key dimensions: what to evaluate, where to evaluate, and how to evaluate. Firstly, we provide an overview from the perspective of evaluation tasks, encompassing general natural language processing tasks, reasoning, medical usage, ethics, education, natural and social sciences, agent applications, and other areas. Secondly, we answer the \u2018where\u2019 and \u2018how\u2019 questions by diving into the evaluation methods and benchmarks, which serve as crucial components in assessing the performance of LLMs. Then, we summarize the success and failure cases of LLMs in different tasks. Finally, we shed light on several future challenges that lie ahead in LLMs evaluation. Our aim is to offer invaluable insights to researchers in the realm of LLMs evaluation, thereby aiding the development of more proficient LLMs. Our key point is that evaluation should be treated as an essential discipline to better assist the development of LLMs. We consistently maintain the related open-source materials at: https://github.com/MLGroupJLU/LLM-eval-survey"
                },
                {
                    "title": "Trained Transformers Learn Linear Models In-Context",
                    "abstract": "Attention-based neural networks such as transformers have demonstrated a remarkable ability to exhibit in-context learning (ICL): Given a short prompt sequence of tokens from an unseen task, they can formulate relevant per-token and next-token predictions without any parameter updates. By embedding a sequence of labeled training data and unlabeled test data as a prompt, this allows for transformers to behave like supervised learning algorithms. Indeed, recent work has shown that when training transformer architectures over random instances of linear regression problems, these models' predictions mimic those of ordinary least squares. Towards understanding the mechanisms underlying this phenomenon, we investigate the dynamics of ICL in transformers with a single linear self-attention layer trained by gradient flow on linear regression tasks. We show that despite non-convexity, gradient flow with a suitable random initialization finds a global minimum of the objective function. At this global minimum, when given a test prompt of labeled examples from a new prediction task, the transformer achieves prediction error competitive with the best linear predictor over the test prompt distribution. We additionally characterize the robustness of the trained transformer to a variety of distribution shifts and show that although a number of shifts are tolerated, shifts in the covariate distribution of the prompts are not. Motivated by this, we consider a generalized ICL setting where the covariate distributions can vary across prompts. We show that although gradient flow succeeds at finding a global minimum in this setting, the trained transformer is still brittle under mild covariate shifts. We complement this finding with experiments on large, nonlinear transformer architectures which we show are more robust under covariate shifts."
                },
                {
                    "title": "Transformers learn through gradual rank increase",
                    "abstract": "We identify incremental learning dynamics in transformers, where the difference between trained and initial weights progressively increases in rank. We rigorously prove this occurs under the simplifying assumptions of diagonal weight matrices and small initialization. Our experiments support the theory and also show that phenomenon can occur in practice without the simplifying assumptions."
                },
                {
                    "title": "Transformers as Statisticians: Provable In-Context Learning with In-Context Algorithm Selection",
                    "abstract": "Neural sequence models based on the transformer architecture have demonstrated remarkable \\emph{in-context learning} (ICL) abilities, where they can perform new tasks when prompted with training and test examples, without any parameter update to the model. This work first provides a comprehensive statistical theory for transformers to perform ICL. Concretely, we show that transformers can implement a broad class of standard machine learning algorithms in context, such as least squares, ridge regression, Lasso, learning generalized linear models, and gradient descent on two-layer neural networks, with near-optimal predictive power on various in-context data distributions. Using an efficient implementation of in-context gradient descent as the underlying mechanism, our transformer constructions admit mild size bounds, and can be learned with polynomially many pretraining sequences. Building on these ``base'' ICL algorithms, intriguingly, we show that transformers can implement more complex ICL procedures involving \\emph{in-context algorithm selection}, akin to what a statistician can do in real life -- A \\emph{single} transformer can adaptively select different base ICL algorithms -- or even perform qualitatively different tasks -- on different input sequences, without any explicit prompting of the right algorithm or task. We both establish this in theory by explicit constructions, and also observe this phenomenon experimentally. In theory, we construct two general mechanisms for algorithm selection with concrete examples: pre-ICL testing, and post-ICL validation. As an example, we use the post-ICL validation mechanism to construct a transformer that can perform nearly Bayes-optimal ICL on a challenging task -- noisy linear models with mixed noise levels. Experimentally, we demonstrate the strong in-context algorithm selection capabilities of standard transformer architectures."
                },
                {
                    "title": "Birth of a Transformer: A Memory Viewpoint",
                    "abstract": "Large language models based on transformers have achieved great empirical successes. However, as they are deployed more widely, there is a growing need to better understand their internal mechanisms in order to make them more reliable. These models appear to store vast amounts of knowledge from their training data, and to adapt quickly to new information provided in their context or prompt. We study how transformers balance these two types of knowledge by considering a synthetic setup where tokens are generated from either global or context-specific bigram distributions. By a careful empirical analysis of the training process on a simplified two-layer transformer, we illustrate the fast learning of global bigrams and the slower development of an\"induction head\"mechanism for the in-context bigrams. We highlight the role of weight matrices as associative memories, provide theoretical insights on how gradients enable their learning during training, and study the role of data-distributional properties."
                },
                {
                    "title": "Scan and Snap: Understanding Training Dynamics and Token Composition in 1-layer Transformer",
                    "abstract": "Transformer architecture has shown impressive performance in multiple research domains and has become the backbone of many neural network models. However, there is limited understanding on how it works. In particular, with a simple predictive loss, how the representation emerges from the gradient \\emph{training dynamics} remains a mystery. In this paper, for 1-layer transformer with one self-attention layer plus one decoder layer, we analyze its SGD training dynamics for the task of next token prediction in a mathematically rigorous manner. We open the black box of the dynamic process of how the self-attention layer combines input tokens, and reveal the nature of underlying inductive bias. More specifically, with the assumption (a) no positional encoding, (b) long input sequence, and (c) the decoder layer learns faster than the self-attention layer, we prove that self-attention acts as a \\emph{discriminative scanning algorithm}: starting from uniform attention, it gradually attends more to distinct key tokens for a specific next token to be predicted, and pays less attention to common key tokens that occur across different next tokens. Among distinct tokens, it progressively drops attention weights, following the order of low to high co-occurrence between the key and the query token in the training set. Interestingly, this procedure does not lead to winner-takes-all, but decelerates due to a \\emph{phase transition} that is controllable by the learning rates of the two layers, leaving (almost) fixed token combination. We verify this \\textbf{\\emph{scan and snap}} dynamics on synthetic and real-world data (WikiText)."
                },
                {
                    "title": "Towards Revealing the Mystery behind Chain of Thought: a Theoretical Perspective",
                    "abstract": "Recent studies have discovered that Chain-of-Thought prompting (CoT) can dramatically improve the performance of Large Language Models (LLMs), particularly when dealing with complex tasks involving mathematics or reasoning. Despite the enormous empirical success, the underlying mechanisms behind CoT and how it unlocks the potential of LLMs remain elusive. In this paper, we take a first step towards theoretically answering these questions. Specifically, we examine the expressivity of LLMs with CoT in solving fundamental mathematical and decision-making problems. We start by giving an impossibility result showing that bounded-depth Transformers are unable to directly produce correct answers for basic arithmetic/equation tasks unless the model size grows super-polynomially with respect to the input length. In contrast, we then prove by construction that autoregressive Transformers of constant size suffice to solve both tasks by generating CoT derivations using a commonly-used math language format. Moreover, we show LLMs with CoT are capable of solving a general class of decision-making problems known as Dynamic Programming, thus justifying its power in tackling complex real-world tasks. Finally, extensive experiments on four tasks show that, while Transformers always fail to predict the answers directly, they can consistently learn to generate correct solutions step-by-step given sufficient CoT demonstrations."
                },
                {
                    "title": "The Closeness of In-Context Learning and Weight Shifting for Softmax Regression",
                    "abstract": "Large language models (LLMs) are known for their exceptional performance in natural language processing, making them highly effective in many human life-related or even job-related tasks. The attention mechanism in the Transformer architecture is a critical component of LLMs, as it allows the model to selectively focus on specific input parts. The softmax unit, which is a key part of the attention mechanism, normalizes the attention scores. Hence, the performance of LLMs in various NLP tasks depends significantly on the crucial role played by the attention mechanism with the softmax unit. In-context learning, as one of the celebrated abilities of recent LLMs, is an important concept in querying LLMs such as ChatGPT. Without further parameter updates, Transformers can learn to predict based on few in-context examples. However, the reason why Transformers becomes in-context learners is not well understood. Recently, several works [ASA+22,GTLV22,ONR+22] have studied the in-context learning from a mathematical perspective based on a linear regression formulation $\\min_x\\| Ax - b \\|_2$, which show Transformers' capability of learning linear functions in context. In this work, we study the in-context learning based on a softmax regression formulation $\\min_{x} \\| \\langle \\exp(Ax), {\\bf 1}_n \\rangle^{-1} \\exp(Ax) - b \\|_2$ of Transformer's attention mechanism. We show the upper bounds of the data transformations induced by a single self-attention layer and by gradient-descent on a $\\ell_2$ regression loss for softmax prediction function, which imply that when training self-attention-only Transformers for fundamental regression tasks, the models learned by gradient-descent and Transformers show great similarity."
                },
                {
                    "title": "Do Transformers Parse while Predicting the Masked Word?",
                    "abstract": "Pre-trained language models have been shown to encode linguistic structures, e.g. dependency and constituency parse trees, in their embeddings while being trained on unsupervised loss functions like masked language modeling. Some doubts have been raised whether the models actually are doing parsing or only some computation weakly correlated with it. We study questions: (a) Is it possible to explicitly describe transformers with realistic embedding dimension, number of heads, etc. that are capable of doing parsing -- or even approximate parsing? (b) Why do pre-trained models capture parsing structure? This paper takes a step toward answering these questions in the context of generative modeling with PCFGs. We show that masked language models like BERT or RoBERTa of moderate sizes can approximately execute the Inside-Outside algorithm for the English PCFG [Marcus et al, 1993]. We also show that the Inside-Outside algorithm is optimal for masked language modeling loss on the PCFG-generated data. We also give a construction of transformers with $50$ layers, $15$ attention heads, and $1275$ dimensional embeddings in average such that using its embeddings it is possible to do constituency parsing with $>70\\%$ F1 score on PTB dataset. We conduct probing experiments on models pre-trained on PCFG-generated data to show that this not only allows recovery of approximate parse tree, but also recovers marginal span probabilities computed by the Inside-Outside algorithm, which suggests an implicit bias of masked language modeling towards this algorithm."
                },
                {
                    "title": "Transformers learn in-context by gradient descent",
                    "abstract": "At present, the mechanisms of in-context learning in Transformers are not well understood and remain mostly an intuition. In this paper, we suggest that training Transformers on auto-regressive objectives is closely related to gradient-based meta-learning formulations. We start by providing a simple weight construction that shows the equivalence of data transformations induced by 1) a single linear self-attention layer and by 2) gradient-descent (GD) on a regression loss. Motivated by that construction, we show empirically that when training self-attention-only Transformers on simple regression tasks either the models learned by GD and Transformers show great similarity or, remarkably, the weights found by optimization match the construction. Thus we show how trained Transformers become mesa-optimizers i.e. learn models by gradient descent in their forward pass. This allows us, at least in the domain of regression problems, to mechanistically understand the inner workings of in-context learning in optimized Transformers. Building on this insight, we furthermore identify how Transformers surpass the performance of plain gradient descent by learning an iterative curvature correction and learn linear models on deep data representations to solve non-linear regression tasks. Finally, we discuss intriguing parallels to a mechanism identified to be crucial for in-context learning termed induction-head (Olsson et al., 2022) and show how it could be understood as a specific case of in-context learning by gradient descent learning within Transformers. Code to reproduce the experiments can be found at https://github.com/google-research/self-organising-systems/tree/master/transformers_learn_icl_by_gd ."
                },
                {
                    "title": "What learning algorithm is in-context learning? Investigations with linear models",
                    "abstract": "Neural sequence models, especially transformers, exhibit a remarkable capacity for in-context learning. They can construct new predictors from sequences of labeled examples $(x, f(x))$ presented in the input without further parameter updates. We investigate the hypothesis that transformer-based in-context learners implement standard learning algorithms implicitly, by encoding smaller models in their activations, and updating these implicit models as new examples appear in the context. Using linear regression as a prototypical problem, we offer three sources of evidence for this hypothesis. First, we prove by construction that transformers can implement learning algorithms for linear models based on gradient descent and closed-form ridge regression. Second, we show that trained in-context learners closely match the predictors computed by gradient descent, ridge regression, and exact least-squares regression, transitioning between different predictors as transformer depth and dataset noise vary, and converging to Bayesian estimators for large widths and depths. Third, we present preliminary evidence that in-context learners share algorithmic features with these predictors: learners' late layers non-linearly encode weight vectors and moment matrices. These results suggest that in-context learning is understandable in algorithmic terms, and that (at least in the linear case) learners may rediscover standard estimation algorithms. Code and reference implementations are released at https://github.com/ekinakyurek/google-research/blob/master/incontext."
                },
                {
                    "title": "Transformers Learn Shortcuts to Automata",
                    "abstract": "Algorithmic reasoning requires capabilities which are most naturally understood through recurrent models of computation, like the Turing machine. However, Transformer models, while lacking recurrence, are able to perform such reasoning using far fewer layers than the number of reasoning steps. This raises the question: what solutions are learned by these shallow and non-recurrent models? We find that a low-depth Transformer can represent the computations of any finite-state automaton (thus, any bounded-memory algorithm), by hierarchically reparameterizing its recurrent dynamics. Our theoretical results characterize shortcut solutions, whereby a Transformer with $o(T)$ layers can exactly replicate the computation of an automaton on an input sequence of length $T$. We find that polynomial-sized $O(\\log T)$-depth solutions always exist; furthermore, $O(1)$-depth simulators are surprisingly common, and can be understood using tools from Krohn-Rhodes theory and circuit complexity. Empirically, we perform synthetic experiments by training Transformers to simulate a wide variety of automata, and show that shortcut solutions can be learned via standard training. We further investigate the brittleness of these solutions and propose potential mitigations."
                },
                {
                    "title": "Vision Transformers provably learn spatial structure",
                    "abstract": "Vision Transformers (ViTs) have achieved comparable or superior performance than Convolutional Neural Networks (CNNs) in computer vision. This empirical breakthrough is even more remarkable since, in contrast to CNNs, ViTs do not embed any visual inductive bias of spatial locality. Yet, recent works have shown that while minimizing their training loss, ViTs specifically learn spatially localized patterns. This raises a central question: how do ViTs learn these patterns by solely minimizing their training loss using gradient-based methods from random initialization? In this paper, we provide some theoretical justification of this phenomenon. We propose a spatially structured dataset and a simplified ViT model. In this model, the attention matrix solely depends on the positional encodings. We call this mechanism the positional attention mechanism. On the theoretical side, we consider a binary classification task and show that while the learning problem admits multiple solutions that generalize, our model implicitly learns the spatial structure of the dataset while generalizing: we call this phenomenon patch association. We prove that patch association helps to sample-efficiently transfer to downstream datasets that share the same structure as the pre-training one but differ in the features. Lastly, we empirically verify that a ViT with positional attention performs similarly to the original one on CIFAR-10/100, SVHN and ImageNet."
                },
                {
                    "title": "What Can Transformers Learn In-Context? A Case Study of Simple Function Classes",
                    "abstract": "In-context learning refers to the ability of a model to condition on a prompt sequence consisting of in-context examples (input-output pairs corresponding to some task) along with a new query input, and generate the corresponding output. Crucially, in-context learning happens only at inference time without any parameter updates to the model. While large language models such as GPT-3 exhibit some ability to perform in-context learning, it is unclear what the relationship is between tasks on which this succeeds and what is present in the training data. To make progress towards understanding in-context learning, we consider the well-defined problem of training a model to in-context learn a function class (e.g., linear functions): that is, given data derived from some functions in the class, can we train a model to in-context learn\"most\"functions from this class? We show empirically that standard Transformers can be trained from scratch to perform in-context learning of linear functions -- that is, the trained model is able to learn unseen linear functions from in-context examples with performance comparable to the optimal least squares estimator. In fact, in-context learning is possible even under two forms of distribution shift: (i) between the training data of the model and inference-time prompts, and (ii) between the in-context examples and the query input during inference. We also show that we can train Transformers to in-context learn more complex function classes -- namely sparse linear functions, two-layer neural networks, and decision trees -- with performance that matches or exceeds task-specific learning algorithms. Our code and models are available at https://github.com/dtsip/in-context-learning ."
                },
                {
                    "title": "Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limit",
                    "abstract": "There is mounting evidence of emergent phenomena in the capabilities of deep learning methods as we scale up datasets, model sizes, and training times. While there are some accounts of how these resources modulate statistical capacity, far less is known about their effect on the computational problem of model training. This work conducts such an exploration through the lens of learning a $k$-sparse parity of $n$ bits, a canonical discrete search problem which is statistically easy but computationally hard. Empirically, we find that a variety of neural networks successfully learn sparse parities, with discontinuous phase transitions in the training curves. On small instances, learning abruptly occurs at approximately $n^{O(k)}$ iterations; this nearly matches SQ lower bounds, despite the apparent lack of a sparse prior. Our theoretical analysis shows that these observations are not explained by a Langevin-like mechanism, whereby SGD\"stumbles in the dark\"until it finds the hidden set of features (a natural algorithm which also runs in $n^{O(k)}$ time). Instead, we show that SGD gradually amplifies the sparse solution via a Fourier gap in the population gradient, making continual progress that is invisible to loss and error metrics."
                },
                {
                    "title": "Exploring Length Generalization in Large Language Models",
                    "abstract": "The ability to extrapolate from short problem instances to longer ones is an important form of out-of-distribution generalization in reasoning tasks, and is crucial when learning from datasets where longer problem instances are rare. These include theorem proving, solving quantitative mathematics problems, and reading/summarizing novels. In this paper, we run careful empirical studies exploring the length generalization capabilities of transformer-based language models. We first establish that naively finetuning transformers on length generalization tasks shows significant generalization deficiencies independent of model scale. We then show that combining pretrained large language models' in-context learning abilities with scratchpad prompting (asking the model to output solution steps before producing an answer) results in a dramatic improvement in length generalization. We run careful failure analyses on each of the learning modalities and identify common sources of mistakes that highlight opportunities in equipping language models with the ability to generalize to longer problems."
                },
                {
                    "title": "Maieutic Prompting: Logically Consistent Reasoning with Recursive Explanations",
                    "abstract": "Pre-trained language models (LMs) struggle with consistent reasoning; recently, prompting LMs to generate explanations that self-guide the inference has emerged as a promising direction to amend this. However, these approaches are fundamentally bounded by the correctness of explanations, which themselves are often noisy and inconsistent. In this work, we develop Maieutic Prompting, which aims to infer a correct answer to a question even from the unreliable generations of LM. Maieutic Prompting induces a tree of explanations abductively (e.g. X is true, because ...) and recursively, then frames the inference as a satisfiability problem over these explanations and their logical relations. We test Maieutic Prompting for true/false QA on three challenging benchmarks that require complex commonsense reasoning. Maieutic Prompting achieves up to 20% better accuracy than state-of-the-art prompting methods, and as a fully unsupervised approach, performs competitively with supervised models. We also show that Maieutic Prompting improves robustness in inference while providing interpretable rationales."
                },
                {
                    "title": "Large Language Models are Zero-Shot Reasoners",
                    "abstract": "Pretrained large language models (LLMs) are widely used in many sub-fields of natural language processing (NLP) and generally known as excellent few-shot learners with task-specific exemplars. Notably, chain of thought (CoT) prompting, a recent technique for eliciting complex multi-step reasoning through step-by-step answer examples, achieved the state-of-the-art performances in arithmetics and symbolic reasoning, difficult system-2 tasks that do not follow the standard scaling laws for LLMs. While these successes are often attributed to LLMs' ability for few-shot learning, we show that LLMs are decent zero-shot reasoners by simply adding\"Let's think step by step\"before each answer. Experimental results demonstrate that our Zero-shot-CoT, using the same single prompt template, significantly outperforms zero-shot LLM performances on diverse benchmark reasoning tasks including arithmetics (MultiArith, GSM8K, AQUA-RAT, SVAMP), symbolic reasoning (Last Letter, Coin Flip), and other logical reasoning tasks (Date Understanding, Tracking Shuffled Objects), without any hand-crafted few-shot examples, e.g. increasing the accuracy on MultiArith from 17.7% to 78.7% and GSM8K from 10.4% to 40.7% with large InstructGPT model (text-davinci-002), as well as similar magnitudes of improvements with another off-the-shelf large model, 540B parameter PaLM. The versatility of this single prompt across very diverse reasoning tasks hints at untapped and understudied fundamental zero-shot capabilities of LLMs, suggesting high-level, multi-task broad cognitive capabilities may be extracted by simple prompting. We hope our work not only serves as the minimal strongest zero-shot baseline for the challenging reasoning benchmarks, but also highlights the importance of carefully exploring and analyzing the enormous zero-shot knowledge hidden inside LLMs before crafting finetuning datasets or few-shot exemplars."
                },
                {
                    "title": "STaR: Bootstrapping Reasoning With Reasoning",
                    "abstract": "Generating step-by-step\"chain-of-thought\"rationales improves language model performance on complex reasoning tasks like mathematics or commonsense question-answering. However, inducing language model rationale generation currently requires either constructing massive rationale datasets or sacrificing accuracy by using only few-shot inference. We propose a technique to iteratively leverage a small number of rationale examples and a large dataset without rationales, to bootstrap the ability to perform successively more complex reasoning. This technique, the\"Self-Taught Reasoner\"(STaR), relies on a simple loop: generate rationales to answer many questions, prompted with a few rationale examples; if the generated answers are wrong, try again to generate a rationale given the correct answer; fine-tune on all the rationales that ultimately yielded correct answers; repeat. We show that STaR significantly improves performance on multiple datasets compared to a model fine-tuned to directly predict final answers, and performs comparably to fine-tuning a 30$\\times$ larger state-of-the-art language model on CommensenseQA. Thus, STaR lets a model improve itself by learning from its own generated reasoning."
                },
                {
                    "title": "Self-Consistency Improves Chain of Thought Reasoning in Language Models",
                    "abstract": "Chain-of-thought prompting combined with pre-trained large language models has achieved encouraging results on complex reasoning tasks. In this paper, we propose a new decoding strategy, self-consistency, to replace the naive greedy decoding used in chain-of-thought prompting. It first samples a diverse set of reasoning paths instead of only taking the greedy one, and then selects the most consistent answer by marginalizing out the sampled reasoning paths. Self-consistency leverages the intuition that a complex reasoning problem typically admits multiple different ways of thinking leading to its unique correct answer. Our extensive empirical evaluation shows that self-consistency boosts the performance of chain-of-thought prompting with a striking margin on a range of popular arithmetic and commonsense reasoning benchmarks, including GSM8K (+17.9%), SVAMP (+11.0%), AQuA (+12.2%), StrategyQA (+6.4%) and ARC-challenge (+3.9%)."
                },
                {
                    "title": "Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer",
                    "abstract": "Hyperparameter (HP) tuning in deep learning is an expensive process, prohibitively so for neural networks (NNs) with billions of parameters. We show that, in the recently discovered Maximal Update Parametrization (muP), many optimal HPs remain stable even as model size changes. This leads to a new HP tuning paradigm we call muTransfer: parametrize the target model in muP, tune the HP indirectly on a smaller model, and zero-shot transfer them to the full-sized model, i.e., without directly tuning the latter at all. We verify muTransfer on Transformer and ResNet. For example, 1) by transferring pretraining HPs from a model of 13M parameters, we outperform published numbers of BERT-large (350M parameters), with a total tuning cost equivalent to pretraining BERT-large once; 2) by transferring from 40M parameters, we outperform published numbers of the 6.7B GPT-3 model, with tuning cost only 7% of total pretraining cost. A Pytorch implementation of our technique can be found at github.com/microsoft/mup and installable via `pip install mup`."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Language Models as Zero-Shot Planners: Extracting Actionable Knowledge for Embodied Agents",
                    "abstract": "Can world knowledge learned by large language models (LLMs) be used to act in interactive environments? In this paper, we investigate the possibility of grounding high-level tasks, expressed in natural language (e.g.\"make breakfast\"), to a chosen set of actionable steps (e.g.\"open fridge\"). While prior work focused on learning from explicit step-by-step examples of how to act, we surprisingly find that if pre-trained LMs are large enough and prompted appropriately, they can effectively decompose high-level tasks into mid-level plans without any further training. However, the plans produced naively by LLMs often cannot map precisely to admissible actions. We propose a procedure that conditions on existing demonstrations and semantically translates the plans to admissible actions. Our evaluation in the recent VirtualHome environment shows that the resulting method substantially improves executability over the LLM baseline. The conducted human evaluation reveals a trade-off between executability and correctness but shows a promising sign towards extracting actionable knowledge from language models. Website at https://huangwl18.github.io/language-planner"
                },
                {
                    "title": "Show Your Work: Scratchpads for Intermediate Computation with Language Models",
                    "abstract": "Large pre-trained language models perform remarkably well on tasks that can be done\"in one pass\", such as generating realistic text or synthesizing computer programs. However, they struggle with tasks that require unbounded multi-step computation, such as adding integers or executing programs. Surprisingly, we find that these same models are able to perform complex multi-step computations -- even in the few-shot regime -- when asked to perform the operation\"step by step\", showing the results of intermediate computations. In particular, we train transformers to perform multi-step computations by asking them to emit intermediate computation steps into a\"scratchpad\". On a series of increasingly complex tasks ranging from long addition to the execution of arbitrary programs, we show that scratchpads dramatically improve the ability of language models to perform multi-step computations."
                },
                {
                    "title": "An Explanation of In-context Learning as Implicit Bayesian Inference",
                    "abstract": "Large language models (LMs) such as GPT-3 have the surprising ability to do in-context learning, where the model learns to do a downstream task simply by conditioning on a prompt consisting of input-output examples. The LM learns from these examples without being explicitly pretrained to learn. Thus, it is unclear what enables in-context learning. In this paper, we study how in-context learning can emerge when pretraining documents have long-range coherence. Here, the LM must infer a latent document-level concept to generate coherent next tokens during pretraining. At test time, in-context learning occurs when the LM also infers a shared latent concept between examples in a prompt. We prove when this occurs despite a distribution mismatch between prompts and pretraining data in a setting where the pretraining distribution is a mixture of HMMs. In contrast to messy large-scale datasets used to train LMs capable of in-context learning, we generate a small-scale synthetic dataset (GINC) where Transformers and LSTMs both exhibit in-context learning. Beyond the theory, experiments on GINC exhibit large-scale real-world phenomena including improved in-context performance with model scaling (despite the same pretraining loss), sensitivity to example order, and instances where zero-shot is better than few-shot in-context learning."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "Inductive Biases and Variable Creation in Self-Attention Mechanisms",
                    "abstract": "Self-attention, an architectural motif designed to model long-range interactions in sequential data, has driven numerous recent breakthroughs in natural language processing and beyond. This work provides a theoretical analysis of the inductive biases of self-attention modules. Our focus is to rigorously establish which functions and long-range dependencies self-attention blocks prefer to represent. Our main result shows that bounded-norm Transformer networks\"create sparse variables\": a single self-attention head can represent a sparse function of the input sequence, with sample complexity scaling only logarithmically with the context length. To support our analysis, we present synthetic experiments to probe the sample complexity of learning sparse Boolean functions with Transformers."
                },
                {
                    "title": "Statistically Meaningful Approximation: a Case Study on Approximating Turing Machines with Transformers",
                    "abstract": "A common lens to theoretically study neural net architectures is to analyze the functions they can approximate. However, constructions from approximation theory may be unrealistic and therefore less meaningful. For example, a common unrealistic trick is to encode target function values using infinite precision. To address these issues, this work proposes a formal definition of statistically meaningful (SM) approximation which requires the approximating network to exhibit good statistical learnability. We study SM approximation for two function classes: boolean circuits and Turing machines. We show that overparameterized feedforward neural nets can SM approximate boolean circuits with sample complexity depending only polynomially on the circuit size, not the size of the network. In addition, we show that transformers can SM approximate Turing machines with computation time bounded by $T$ with sample complexity polynomial in the alphabet size, state space size, and $\\log (T)$. We also introduce new tools for analyzing generalization which provide much tighter sample complexities than the typical VC-dimension or norm-based bounds, which may be of independent interest."
                },
                {
                    "title": "On the Expressive Power of Self-Attention Matrices",
                    "abstract": "Transformer networks are able to capture patterns in data coming from many domains (text, images, videos, proteins, etc.) with little or no change to architecture components. We perform a theoretical analysis of the core component responsible for signal propagation between elements, i.e. the self-attention matrix. In practice, this matrix typically exhibits two properties: (1) it is sparse, meaning that each token only attends to a small subset of other tokens; and (2) it changes dynamically depending on the input to the module. With these considerations in mind, we ask the following question: Can a fixed self-attention module approximate arbitrary sparse patterns depending on the input? How small is the hidden size $d$ required for such approximation? We make progress in answering this question and show that the self-attention matrix can provably approximate sparse matrices, where sparsity is in terms of a bounded number of nonzero elements in each row and column. While the parameters of self-attention are fixed, various sparse matrices can be approximated by only modifying the inputs. Our proof is based on the random projection technique and uses the seminal Johnson-Lindenstrauss lemma. Our proof is constructive, enabling us to propose an algorithm for finding adaptive inputs and fixed self-attention parameters in order to approximate a given matrix. In particular, we show that, in order to approximate any sparse matrix up to a given precision defined in terms of preserving matrix element ratios, $d$ grows only logarithmically with the sequence length $L$ (i.e. $d = O(\\log L)$)."
                },
                {
                    "title": "Self-Attention Networks Can Process Bounded Hierarchical Languages",
                    "abstract": "Despite their impressive performance in NLP, self-attention networks were recently proved to be limited for processing formal languages with hierarchical structure, such as Dyck-k, the language consisting of well-nested parentheses of k types. This suggested that natural language can be approximated well with models that are too weak for formal languages, or that the role of hierarchy and recursion in natural language might be limited. We qualify this implication by proving that self-attention networks can process Dyck-(k, D), the subset of Dyck-k with depth bounded by D, which arguably better captures the bounded hierarchical structure of natural language. Specifically, we construct a hard-attention network with D+1 layers and O(log k) memory size (per token per layer) that recognizes Dyck-(k, D), and a soft-attention network with two layers and O(log k) memory size that generates Dyck-(k, D). Experiments show that self-attention networks trained on Dyck-(k, D) generalize to longer inputs with near-perfect accuracy, and also verify the theoretical memory advantage of self-attention networks over recurrent networks."
                },
                {
                    "title": "Approximating How Single Head Attention Learns",
                    "abstract": "Why do models often attend to salient words, and how does this evolve throughout training? We approximate model training as a two stage process: early on in training when the attention weights are uniform, the model learns to translate individual input word `i` to `o` if they co-occur frequently. Later, the model learns to attend to `i` while the correct output is $o$ because it knows `i` translates to `o`. To formalize, we define a model property, Knowledge to Translate Individual Words (KTIW) (e.g. knowing that `i` translates to `o`), and claim that it drives the learning of the attention. This claim is supported by the fact that before the attention mechanism is learned, KTIW can be learned from word co-occurrence statistics, but not the other way around. Particularly, we can construct a training distribution that makes KTIW hard to learn, the learning of the attention fails, and the model cannot even learn the simple task of copying the input words to the output. Our approximation explains why models sometimes attend to salient words, and inspires a toy example where a multi-head attention model can overcome the above hard training distribution by improving learning dynamics rather than expressiveness. We end by discussing the limitation of our approximation framework and suggest future directions."
                },
                {
                    "title": "Prompt Programming for Large Language Models: Beyond the Few-Shot Paradigm",
                    "abstract": "Prevailing methods for mapping large generative language models to supervised tasks may fail to sufficiently probe models\u2019 novel capabilities. Using GPT-3 as a case study, we show that 0-shot prompts can significantly outperform few-shot prompts. We suggest that the function of few-shot examples in these cases is better described as locating an already learned task rather than meta-learning. This analysis motivates rethinking the role of prompts in controlling and evaluating powerful language models. We discuss methods of prompt programming, emphasizing the usefulness of considering prompts through the lens of natural language. We explore techniques for exploiting the capacity of narratives and cultural anchors to encode nuanced intentions and techniques for encouraging deconstruction of a problem into components before producing a verdict. Informed by this more encompassing theory of prompt programming, we also introduce the idea of a metaprompt that seeds the model to generate its own natural language prompts for a range of tasks. Finally, we discuss how these more general methods of interacting with language models can be incorporated into existing and future benchmarks and practical applications."
                },
                {
                    "title": "On the Ability and Limitations of Transformers to Recognize Formal Languages",
                    "abstract": "Transformers have supplanted recurrent models in a large number of NLP tasks. However, the differences in their abilities to model different syntactic properties remain largely unknown. Past works suggest that LSTMs generalize very well on regular languages and have close connections with counter languages. In this work, we systematically study the ability of Transformers to model such languages as well as the role of its individual components in doing so. We first provide a construction of Transformers for a subclass of counter languages, including well-studied languages such as n-ary Boolean Expressions, Dyck-1, and its generalizations. In experiments, we find that Transformers do well on this subclass, and their learned mechanism strongly correlates with our construction. Perhaps surprisingly, in contrast to LSTMs, Transformers do well only on a subset of regular languages with degrading performance as we make languages more complex according to a well-known measure of complexity. Our analysis also provides insights on the role of self-attention mechanism in modeling certain behaviors and the influence of positional encoding schemes on the learning and generalization abilities of the model."
                },
                {
                    "title": "Infinite attention: NNGP and NTK for deep attention networks",
                    "abstract": "There is a growing amount of literature on the relationship between wide neural networks (NNs) and Gaussian processes (GPs), identifying an equivalence between the two for a variety of NN architectures. This equivalence enables, for instance, accurate approximation of the behaviour of wide Bayesian NNs without MCMC or variational approximations, or characterisation of the distribution of randomly initialised wide NNs optimised by gradient descent without ever running an optimiser. We provide a rigorous extension of these results to NNs involving attention layers, showing that unlike single-head attention, which induces non-Gaussian behaviour, multi-head attention architectures behave as GPs as the number of heads tends to infinity. We further discuss the effects of positional encodings and layer normalisation, and propose modifications of the attention mechanism which lead to improved results for both finite and infinitely wide NNs. We evaluate attention kernels empirically, leading to a moderate improvement upon the previous state-of-the-art on CIFAR-10 for GPs without trainable kernels and advanced data preprocessing. Finally, we introduce new features to the Neural Tangents library (Novak et al., 2020) allowing applications of NNGP/NTK models, with and without attention, to variable-length sequences, with an example on the IMDb reviews dataset."
                },
                {
                    "title": "On the Computational Power of Transformers and Its Implications in Sequence Modeling",
                    "abstract": "Transformers are being used extensively across several sequence modeling tasks. Significant research effort has been devoted to experimentally probe the inner workings of Transformers. However, our conceptual and theoretical understanding of their power and inherent limitations is still nascent. In particular, the roles of various components in Transformers such as positional encodings, attention heads, residual connections, and feedforward networks, are not clear. In this paper, we take a step towards answering these questions. We analyze the computational power as captured by Turing-completeness. We first provide an alternate and simpler proof to show that vanilla Transformers are Turing-complete and then we prove that Transformers with only positional masking and without any positional encoding are also Turing-complete. We further analyze the necessity of each component for the Turing-completeness of the network; interestingly, we find that a particular type of residual connection is necessary. We demonstrate the practical implications of our results via experiments on machine translation and synthetic tasks."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Are Transformers universal approximators of sequence-to-sequence functions?",
                    "abstract": "Despite the widespread adoption of Transformer models for NLP tasks, the expressive power of these models is not well-understood. In this paper, we establish that Transformer models are universal approximators of continuous permutation equivariant sequence-to-sequence functions with compact support, which is quite surprising given the amount of shared parameters in these models. Furthermore, using positional encodings, we circumvent the restriction of permutation equivariance, and show that Transformer models can universally approximate arbitrary continuous sequence-to-sequence functions on a compact domain. Interestingly, our proof techniques clearly highlight the different roles of the self-attention and the feed-forward layers in Transformers. In particular, we prove that fixed width self-attention layers can compute contextual mappings of the input sequences, playing a key role in the universal approximation property of Transformers. Based on this insight from our analysis, we consider other architectures that can compute contextual mappings and empirically evaluate them."
                },
                {
                    "title": "Why Gradient Clipping Accelerates Training: A Theoretical Justification for Adaptivity",
                    "abstract": "We provide a theoretical explanation for the effectiveness of gradient clipping in training deep neural networks. The key ingredient is a new smoothness condition derived from practical neural network training examples. We observe that gradient smoothness, a concept central to the analysis of first-order optimization algorithms that is often assumed to be a constant, demonstrates significant variability along the training trajectory of deep neural networks. Further, this smoothness positively correlates with the gradient norm, and contrary to standard assumptions in the literature, it can grow with the norm of the gradient. These empirical observations limit the applicability of existing theoretical analyses of algorithms that rely on a fixed bound on smoothness. These observations motivate us to introduce a novel relaxation of gradient smoothness that is weaker than the commonly used Lipschitz smoothness assumption. Under the new condition, we prove that two popular methods, namely, \\emph{gradient clipping} and \\emph{normalized gradient}, converge arbitrarily faster than gradient descent with fixed stepsize. We further explain why such adaptively scaled gradient methods can accelerate empirical convergence and verify our results empirically in popular neural network training settings."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Attention is Turing-Complete",
                    "abstract": "Alternatives to recurrent neural networks, in particular, architectures based on self-attention , are gaining momentum for processing input sequences. In spite of their relevance, the computational properties of such networks have not yet been fully explored. We study the computational power of the Transformer , one of the most paradigmatic architectures ex-emplifying self-attention. We show that the Transformer with hard-attention is Turing complete exclusively based on their capacity to compute and access internal dense representations of the data. Our study also reveals some minimal sets of elements needed to obtain this completeness result."
                },
                {
                    "title": "ADAPTIVE ESTIMATION OF A QUADRATIC FUNCTIONAL BY MODEL SELECTION",
                    "abstract": "We consider the problem of estimating (cid:1) s (cid:1) 2 when s belongs to some separable Hilbert space and one observes the Gaussian process Y (cid:2) t (cid:3) = (cid:5) s(cid:4)t (cid:6) + \u03c3L (cid:2) t (cid:3) , for all t \u2208 (cid:1) , where L is some Gaussian isonormal process. This framework allows us in particular to consider the classical \u201cGaussian sequence model\u201d for which (cid:1) = l 2 (cid:2) (cid:2) \u2217 (cid:3) and L (cid:2) t (cid:3) = (cid:1) \u03bb \u2265 1 t \u03bb \u03b5 \u03bb , where (cid:2) \u03b5 \u03bb (cid:3) \u03bb \u2265 1 is a sequence of i.i.d. standard normal variables. Our approach consists in considering some at most countable families of \ufb01nite-dimensional linear subspaces of (cid:1) (the models ) and then using model selection via some con- veniently penalized least squares criterion to build new estimators of (cid:1) s (cid:1) 2 . We prove a general nonasymptotic risk bound which allows us to show that such penalized estimators are adaptive on a variety of collections of sets for the parameter s , depending on the family of models from which they are built. In particular, in the context of the Gaussian sequence model, a convenient choice of the family of models allows de\ufb01ning estimators which are adaptive over collections of hyperrectangles, ellipsoids, l p -bodies or Besov bodies. We take special care to describe the conditions under which the penalized estimator is ef\ufb01cient when the level of noise \u03c3 tends to zero. Our construction is an alternative to the one by Efro\u00a8\u0131movich and Low for hyperrectangles and provides new results otherwise."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we theoretically understand and mitigate the reversal curse in auto-regressive large language models (LLMs) that prevents them from generalizing learned relationships in reverse?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the reversal curse is crucial for advancing the capabilities of LLMs in logical reasoning tasks, which have significant implications for various applications, including natural language understanding, automated reasoning, and AI-driven decision-making. Addressing this problem could lead to more robust models that can handle complex reasoning tasks without the need for extensive fine-tuning or dataset manipulation, thereby enhancing the efficiency and effectiveness of LLMs in real-world applications. This research could also inspire future studies on training dynamics and model architectures, potentially leading to breakthroughs in how LLMs learn and generalize knowledge.\n\n**[Question 3] - Why is it hard?**  \nThe reversal curse is challenging to address due to the asymmetric nature of weights in auto-regressive models, where the training dynamics do not allow for effective learning of reverse relationships. Naive approaches, such as simply reversing training data or altering model architectures, may lead to degraded performance on other tasks. The complexities arise from the limitations of popular training algorithms like gradient descent, which may not reach optimal configurations for both directions of learned relationships. Additionally, the theoretical understanding of these dynamics is still underdeveloped, making it difficult to devise effective solutions.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the expressivity of LLMs rather than their training dynamics, leading to a lack of understanding of why the reversal curse occurs. Existing methods to mitigate the curse, such as dataset manipulation or changing model objectives, have not been successful without compromising overall model performance. Barriers include the limited exploration of training dynamics in the context of auto-regressive models and the absence of a comprehensive theoretical framework that connects training dynamics to the reversal curse. Our approach differs by focusing on the underlying training dynamics rather than expressivity, providing a new perspective on the problem.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a theoretical analysis of the reversal curse through the lens of training dynamics in two auto-regressive models: a bilinear model and one-layer transformers. We will utilize reparameterization to analyze the weight asymmetry during training, focusing on sequences structured as \u201cA\u2192B\u201d and \u201cB\u2190A.\u201d"
            }
        },
        "author_data": {
            "daba57f8-c8dc-4271-b78a-919ab8213308": {
                "pk": "daba57f8-c8dc-4271-b78a-919ab8213308",
                "name": "Hanlin Zhu",
                "collaborators": [
                    "Jiantao Jiao",
                    "Stuart Russell",
                    "Cyrus Rashtchian",
                    "David P. Woodruff",
                    "Paria Rashidinejad",
                    "Amy Zhang",
                    "Banghua Zhu",
                    "Ruosong Wang",
                    "Jason D. Lee",
                    "Baihe Huang"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Offline Learning",
                    "Natural Language Processing",
                    "Statistical Learning"
                ],
                "publications": [
                    {
                        "title": "Provably Efficient Offline Goal-Conditioned Reinforcement Learning with General Function Approximation and Single-Policy Concentrability",
                        "abstract": "Goal-conditioned reinforcement learning (GCRL) refers to learning general-purpose skills that aim to reach diverse goals. In particular, offline GCRL only requires purely pre-collected datasets to perform training tasks without additional interactions with the environment. Although offline GCRL has become increasingly prevalent and many previous works have demonstrated its empirical success, the theoretical understanding of efficient offline GCRL algorithms is not well established, especially when the state space is huge and the offline dataset only covers the policy we aim to learn. In this paper, we provide a rigorous theoretical analysis of an existing empirically successful offline GCRL algorithm. We prove that under slight modification, this algorithm enjoys an $\\widetilde{O}(\\text{poly}(1/\\epsilon))$ sample complexity (where $\\epsilon$ is the desired suboptimality of the learned policy) with general function approximation thanks to the property of (semi-)strong convexity of the objective functions. We only require nearly minimal assumptions on the dataset (single-policy concentrability) and the function class (realizability). Moreover, this algorithm consists of two uninterleaved optimization steps, which we refer to as $V$-learning and policy learning, and is computationally stable since it does not involve minimax optimization. We also empirically validate our theory by showing that the modified algorithm outperforms the previous algorithm in various real-world environments. To the best of our knowledge, this is the first algorithm that is both provably efficient with general function approximation and single-policy concentrability, and empirically successful without requiring solving minimax optimization problems."
                    },
                    {
                        "title": "Efficient Prompt Caching via Embedding Similarity",
                        "abstract": "Large language models (LLMs) have achieved huge success in numerous natural language process (NLP) tasks. However, it faces the challenge of significant resource consumption during inference. In this paper, we aim to improve the inference efficiency of LLMs by prompt caching, i.e., if the current prompt can be answered by the same response of a previous prompt, one can directly utilize that previous response without calling the LLM. Specifically, we focus on the prediction accuracy of prompt caching for single-round question-answering tasks via embedding similarity. The existing embeddings of prompts mostly focus on whether two prompts are semantically similar, which is not necessarily equivalent to whether the same response can answer them. Therefore, we propose a distillation-based method to fine-tune the existing embeddings for better caching prediction. Theoretically, we provide finite-sample guarantees for the convergence of our method under different types of loss functions. Empirically, we carefully construct a hard dataset based on Kwiatkowski et al. (2019) where the existing embedding model (Wang et al., 2022) only achieves an AUC of 0.51. We then fine-tune the above embedding model, which significantly improves the AUC of caching prediction from 0.51 to 0.81. We also conduct simulations demonstrating that our trained models achieve better caching efficiency than the previous embedding model."
                    },
                    {
                        "title": "Vector-Matrix-Vector Queries for Solving Linear Algebra, Statistics, and Graph Problems",
                        "abstract": "We consider the general problem of learning about a matrix through vector-matrix-vector queries. These queries provide the value of $\\boldsymbol{u}^{\\mathrm{T}}\\boldsymbol{M}\\boldsymbol{v}$ over a fixed field $\\mathbb{F}$ for a specified pair of vectors $\\boldsymbol{u},\\boldsymbol{v} \\in \\mathbb{F}^n$. To motivate these queries, we observe that they generalize many previously studied models, such as independent set queries, cut queries, and standard graph queries. They also specialize the recently studied matrix-vector query model. Our work is exploratory and broad, and we provide new upper and lower bounds for a wide variety of problems, spanning linear algebra, statistics, and graphs. Many of our results are nearly tight, and we use diverse techniques from linear algebra, randomized algorithms, and communication complexity."
                    },
                    {
                        "title": "Importance Weighted Actor-Critic for Optimal Conservative Offline Reinforcement Learning",
                        "abstract": "We propose A-Crab (Actor-Critic Regularized by Average Bellman error), a new practical algorithm for offline reinforcement learning (RL) in complex environments with insufficient data coverage. Our algorithm combines the marginalized importance sampling framework with the actor-critic paradigm, where the critic returns evaluations of the actor (policy) that are pessimistic relative to the offline data and have a small average (importance-weighted) Bellman error. Compared to existing methods, our algorithm simultaneously offers a number of advantages: (1) It achieves the optimal statistical rate of $1/\\sqrt{N}$ -- where $N$ is the size of offline dataset -- in converging to the best policy covered in the offline dataset, even when combined with general function approximators. (2) It relies on a weaker average notion of policy coverage (compared to the $\\ell_\\infty$ single-policy concentrability) that exploits the structure of policy visitations. (3) It outperforms the data-collection behavior policy over a wide range of specific hyperparameters. We provide both theoretical analysis and experimental results to validate the effectiveness of our proposed algorithm."
                    },
                    {
                        "title": "Provably Efficient Reinforcement Learning via Surprise Bound",
                        "abstract": "Value function approximation is important in modern reinforcement learning (RL) problems especially when the state space is (infinitely) large. Despite the importance and wide applicability of value function approximation, its theoretical understanding is still not as sophisticated as its empirical success, especially in the context of general function approximation. In this paper, we propose a provably efficient RL algorithm (both computationally and statistically) with general value function approximations. We show that if the value functions can be approximated by a function class that satisfies the Bellman-completeness assumption, our algorithm achieves an $\\widetilde{O}(\\text{poly}(\\iota H)\\sqrt{T})$ regret bound where $\\iota$ is the product of the surprise bound and log-covering numbers, $H$ is the planning horizon, $K$ is the number of episodes and $T = HK$ is the total number of steps the agent interacts with the environment. Our algorithm achieves reasonable regret bounds when applied to both the linear setting and the sparse high-dimensional linear setting. Moreover, our algorithm only needs to solve $O(H\\log K)$ empirical risk minimization (ERM) problems, which is far more efficient than previous algorithms that need to solve ERM problems for $\\Omega(HK)$ times."
                    },
                    {
                        "title": "On Representation Complexity of Model-based and Model-free Reinforcement Learning",
                        "abstract": "We study the representation complexity of model-based and model-free reinforcement learning (RL) in the context of circuit complexity. We prove theoretically that there exists a broad class of MDPs such that their underlying transition and reward functions can be represented by constant depth circuits with polynomial size, while the optimal $Q$-function suffers an exponential circuit complexity in constant-depth circuits. By drawing attention to the approximation errors and building connections to complexity theory, our theory provides unique insights into why model-based algorithms usually enjoy better sample complexity than model-free algorithms from a novel representation complexity perspective: in some cases, the ground-truth rule (model) of the environment is simple to represent, while other quantities, such as $Q$-function, appear complex. We empirically corroborate our theory by comparing the approximation error of the transition kernel, reward function, and optimal $Q$-function in various Mujoco environments, which demonstrates that the approximation errors of the transition kernel and reward function are consistently lower than those of the optimal $Q$-function. To the best of our knowledge, this work is the first to study the circuit complexity of RL, which also provides a rigorous framework for future research."
                    },
                    {
                        "title": "Guided Dialog Policy Learning: Reward Estimation for Multi-Domain Task-Oriented Dialog",
                        "abstract": "Dialog policy decides what and how a task-oriented dialog system will respond, and plays a vital role in delivering effective conversations. Many studies apply Reinforcement Learning to learn a dialog policy with the reward function which requires elaborate design and pre-specified user goals. With the growing needs to handle complex goals across multiple domains, such manually designed reward functions are not affordable to deal with the complexity of real-world tasks. To this end, we propose Guided Dialog Policy Learning, a novel algorithm based on Adversarial Inverse Reinforcement Learning for joint reward estimation and policy optimization in multi-domain task-oriented dialog. The proposed approach estimates the reward signal and infers the user goal in the dialog sessions. The reward estimator evaluates the state-action pairs so that it can guide the dialog policy at each dialog turn. Extensive experiments on a multi-domain dialog dataset show that the dialog policy guided by the learned reward function achieves remarkably higher task success than state-of-the-art baselines."
                    },
                    {
                        "title": "Avoiding Catastrophe in Online Learning by Asking for Help",
                        "abstract": "Most learning algorithms with formal regret guarantees assume that no mistake is irreparable and essentially rely on trying all possible behaviors. This approach is problematic when some mistakes are \\emph{catastrophic}, i.e., irreparable. We propose an online learning problem where the goal is to minimize the chance of catastrophe. Specifically, we assume that the payoff in each round represents the chance of avoiding catastrophe that round and aim to maximize the product of payoffs (the overall chance of avoiding catastrophe) while allowing a limited number of queries to a mentor. We first show that in general, any algorithm either constantly queries the mentor or is nearly guaranteed to cause catastrophe. However, in settings where the mentor policy class is learnable in the standard online learning model, we provide an algorithm whose regret and rate of querying the mentor both approach 0 as the time horizon grows. Conceptually, if a policy class is learnable in the absence of catastrophic risk, it is learnable in the presence of catastrophic risk if the agent can ask for help."
                    },
                    {
                        "title": "Average-Case Communication Complexity of Statistical Problems",
                        "abstract": "We study statistical problems, such as planted clique, its variants, and sparse principal component analysis in the context of average-case communication complexity. Our motivation is to understand the statistical-computational trade-offs in streaming, sketching, and query-based models. Communication complexity is the main tool for proving lower bounds in these models, yet many prior results do not hold in an average-case setting. We provide a general reduction method that preserves the input distribution for problems involving a random graph or matrix with planted structure. Then, we derive two-party and multi-party communication lower bounds for detecting or finding planted cliques, bipartite cliques, and related problems. As a consequence, we obtain new bounds on the query complexity in the edge-probe, vector-matrix-vector, matrix-vector, linear sketching, and $\\mathbb{F}_2$-sketching models. Many of these results are nearly tight, and we use our techniques to provide simple proofs of some known lower bounds for the edge-probe model."
                    },
                    {
                        "title": "Optimal Conservative Offline RL with General Function Approximation via Augmented Lagrangian",
                        "abstract": "Offline reinforcement learning (RL), which refers to decision-making from a previously-collected dataset of interactions, has received significant attention over the past years. Much effort has focused on improving offline RL practicality by addressing the prevalent issue of partial data coverage through various forms of conservative policy learning. While the majority of algorithms do not have finite-sample guarantees, several provable conservative offline RL algorithms are designed and analyzed within the single-policy concentrability framework that handles partial coverage. Yet, in the nonlinear function approximation setting where confidence intervals are difficult to obtain, existing provable algorithms suffer from computational intractability, prohibitively strong assumptions, and suboptimal statistical rates. In this paper, we leverage the marginalized importance sampling (MIS) formulation of RL and present the first set of offline RL algorithms that are statistically optimal and practical under general function approximation and single-policy concentrability, bypassing the need for uncertainty quantification. We identify that the key to successfully solving the sample-based approximation of the MIS problem is ensuring that certain occupancy validity constraints are nearly satisfied. We enforce these constraints by a novel application of the augmented Lagrangian method and prove the following result: with the MIS formulation, augmented Lagrangian is enough for statistically optimal offline RL. In stark contrast to prior algorithms that induce additional conservatism through methods such as behavior regularization, our approach provably eliminates this need and reinterprets regularizers as \"enforcers of occupancy validity\" than \"promoters of conservatism.\""
                    }
                ]
            },
            "e0e00604-2fac-41b9-85f1-54de35faebbc": {
                "pk": "e0e00604-2fac-41b9-85f1-54de35faebbc",
                "name": "Baihe Huang",
                "collaborators": [
                    "Jason D. Lee",
                    "Zhao Song",
                    "Ruizhe Zhang",
                    "Qi Lei",
                    "Runzhou Tao",
                    "Hanlin Zhu",
                    "Shunhua Jiang",
                    "Sai Praneeth Karimireddy",
                    "Michael I. Jordan",
                    "Wenhao Zhan"
                ],
                "domain": [
                    "Federated Learning",
                    "Reinforcement Learning",
                    "Optimization",
                    "Statistical Learning"
                ],
                "publications": [
                    {
                        "title": "FL-NTK: A Neural Tangent Kernel-based Framework for Federated Learning Convergence Analysis",
                        "abstract": "Federated Learning (FL) is an emerging learning scheme that allows different distributed clients to train deep neural networks together without data sharing. Neural networks have become popular due to their unprecedented success. To the best of our knowledge, the theoretical guarantees of FL concerning neural networks with explicit forms and multi-step updates are unexplored. Nevertheless, training analysis of neural networks in FL is non-trivial for two reasons: first, the objective loss function we are optimizing is non-smooth and non-convex, and second, we are even not updating in the gradient direction. Existing convergence results for gradient descent-based methods heavily rely on the fact that the gradient direction is used for updating. This paper presents a new class of convergence analysis for FL, Federated Learning Neural Tangent Kernel (FL-NTK), which corresponds to overparamterized ReLU neural networks trained by gradient descent in FL and is inspired by the analysis in Neural Tangent Kernel (NTK). Theoretically, FL-NTK converges to a global-optimal solution at a linear rate with properly tuned learning parameters. Furthermore, with proper distributional assumptions, FL-NTK can also achieve good generalization."
                    },
                    {
                        "title": "On Representation Complexity of Model-based and Model-free Reinforcement Learning",
                        "abstract": "We study the representation complexity of model-based and model-free reinforcement learning (RL) in the context of circuit complexity. We prove theoretically that there exists a broad class of MDPs such that their underlying transition and reward functions can be represented by constant depth circuits with polynomial size, while the optimal $Q$-function suffers an exponential circuit complexity in constant-depth circuits. By drawing attention to the approximation errors and building connections to complexity theory, our theory provides unique insights into why model-based algorithms usually enjoy better sample complexity than model-free algorithms from a novel representation complexity perspective: in some cases, the ground-truth rule (model) of the environment is simple to represent, while other quantities, such as $Q$-function, appear complex. We empirically corroborate our theory by comparing the approximation error of the transition kernel, reward function, and optimal $Q$-function in various Mujoco environments, which demonstrates that the approximation errors of the transition kernel and reward function are consistently lower than those of the optimal $Q$-function. To the best of our knowledge, this work is the first to study the circuit complexity of RL, which also provides a rigorous framework for future research."
                    },
                    {
                        "title": "A Faster Quantum Algorithm for Semidefinite Programming via Robust IPM Framework",
                        "abstract": "This paper studies a fundamental problem in convex optimization, which is to solve semidefinite programming (SDP) with high accuracy. This paper follows from the existing robust SDP-based interior point method analysis due to [Huang, Jiang, Song, Tao and Zhang, FOCS 2022]. While, the previous work only provides an efficient implementation in the classical setting. This work provides a novel quantum implementation.   We give a quantum second-order algorithm with high-accuracy in both the optimality and the feasibility of its output, and its running time depending on $\\log(1/\\epsilon)$ on well-conditioned instances. Due to the limitation of quantum itself or first-order method, all the existing quantum SDP solvers either have polynomial error dependence or low-accuracy in the feasibility."
                    },
                    {
                        "title": "Evaluating and Incentivizing Diverse Data Contributions in Collaborative Learning",
                        "abstract": "For a federated learning model to perform well, it is crucial to have a diverse and representative dataset. However, the data contributors may only be concerned with the performance on a specific subset of the population, which may not reflect the diversity of the wider population. This creates a tension between the principal (the FL platform designer) who cares about global performance and the agents (the data collectors) who care about local performance. In this work, we formulate this tension as a game between the principal and multiple agents, and focus on the linear experiment design problem to formally study their interaction. We show that the statistical criterion used to quantify the diversity of the data, as well as the choice of the federated learning algorithm used, has a significant effect on the resulting equilibrium. We leverage this to design simple optimal federated learning mechanisms that encourage data collectors to contribute data representative of the global population, thereby maximizing global performance."
                    },
                    {
                        "title": "Solving SDP Faster: A Robust IPM Framework and Efficient Implementation",
                        "abstract": "This paper introduces a new robust interior point method analysis for semidefinite programming (SDP). This new robust analysis can be combined with either logarithmic barrier or hybrid barrier.   Under this new framework, we can improve the running time of semidefinite programming (SDP) with variable size $n \\times n$ and $m$ constraints up to $\\epsilon$ accuracy.   We show that for the case $m = \\Omega(n^2)$, we can solve SDPs in $m^{\\omega}$ time. This suggests solving SDP is nearly as fast as solving the linear system with equal number of variables and constraints. This is the first result that tall dense SDP can be solved in the nearly-optimal running time, and it also improves the state-of-the-art SDP solver [Jiang, Kathuria, Lee, Padmanabhan and Song, FOCS 2020].   In addition to our new IPM analysis, we also propose a number of techniques that might be of further interest, such as, maintaining the inverse of a Kronecker product using lazy updates, a general amortization scheme for positive semidefinite matrices."
                    },
                    {
                        "title": "Offline Reinforcement Learning with Realizability and Single-policy Concentrability",
                        "abstract": "Sample-efficiency guarantees for offline reinforcement learning (RL) often rely on strong assumptions on both the function classes (e.g., Bellman-completeness) and the data coverage (e.g., all-policy concentrability). Despite the recent efforts on relaxing these assumptions, existing works are only able to relax one of the two factors, leaving the strong assumption on the other factor intact. As an important open problem, can we achieve sample-efficient offline RL with weak assumptions on both factors?   In this paper we answer the question in the positive. We analyze a simple algorithm based on the primal-dual formulation of MDPs, where the dual variables (discounted occupancy) are modeled using a density-ratio function against offline data. With proper regularization, we show that the algorithm enjoys polynomial sample complexity, under only realizability and single-policy concentrability. We also provide alternative analyses based on different assumptions to shed light on the nature of primal-dual algorithms for offline RL."
                    },
                    {
                        "title": "Sample Complexity for Quadratic Bandits: Hessian Dependent Bounds and Optimal Algorithms",
                        "abstract": "In stochastic zeroth-order optimization, a problem of practical relevance is understanding how to fully exploit the local geometry of the underlying objective function. We consider a fundamental setting in which the objective function is quadratic, and provide the first tight characterization of the optimal Hessian-dependent sample complexity. Our contribution is twofold. First, from an information-theoretic point of view, we prove tight lower bounds on Hessian-dependent complexities by introducing a concept called energy allocation, which captures the interaction between the searching algorithm and the geometry of objective functions. A matching upper bound is obtained by solving the optimal energy spectrum. Then, algorithmically, we show the existence of a Hessian-independent algorithm that universally achieves the asymptotic optimal sample complexities for all Hessian instances. The optimal sample complexities achieved by our algorithm remain valid for heavy-tailed noise distributions, which are enabled by a truncation method."
                    },
                    {
                        "title": "Towards General Function Approximation in Zero-Sum Markov Games",
                        "abstract": "This paper considers two-player zero-sum finite-horizon Markov games with simultaneous moves. The study focuses on the challenging settings where the value function or the model is parameterized by general function classes. Provably efficient algorithms for both decoupled and {coordinated} settings are developed. In the {decoupled} setting where the agent controls a single player and plays against an arbitrary opponent, we propose a new model-free algorithm. The sample complexity is governed by the Minimax Eluder dimension -- a new dimension of the function class in Markov games. As a special case, this method improves the state-of-the-art algorithm by a $\\sqrt{d}$ factor in the regret when the reward function and transition kernel are parameterized with $d$-dimensional linear features. In the {coordinated} setting where both players are controlled by the agent, we propose a model-based algorithm and a model-free algorithm. In the model-based algorithm, we prove that sample complexity can be bounded by a generalization of Witness rank to Markov games. The model-free algorithm enjoys a $\\sqrt{K}$-regret upper bound where $K$ is the number of episodes."
                    },
                    {
                        "title": "InstaHide's Sample Complexity When Mixing Two Private Images",
                        "abstract": "Training neural networks usually require large numbers of sensitive training data, and how to protect the privacy of training data has thus become a critical topic in deep learning research. InstaHide is a state-of-the-art scheme to protect training data privacy with only minor effects on test accuracy, and its security has become a salient question. In this paper, we systematically study recent attacks on InstaHide and present a unified framework to understand and analyze these attacks. We find that existing attacks either do not have a provable guarantee or can only recover a single private image. On the current InstaHide challenge setup, where each InstaHide image is a mixture of two private images, we present a new algorithm to recover all the private images with a provable guarantee and optimal sample complexity. In addition, we also provide a computational hardness result on retrieving all InstaHide images. Our results demonstrate that InstaHide is not information-theoretically secure but computationally secure in the worst case, even when mixing two private images."
                    },
                    {
                        "title": "DAVED: Data Acquisition via Experimental Design for Data Markets",
                        "abstract": "The acquisition of training data is crucial for machine learning applications. Data markets can increase the supply of data, particularly in data-scarce domains such as healthcare, by incentivizing potential data providers to join the market. A major challenge for a data buyer in such a market is choosing the most valuable data points from a data seller. Unlike prior work in data valuation, which assumes centralized data access, we propose a federated approach to the data acquisition problem that is inspired by linear experimental design. Our proposed data acquisition method achieves lower prediction error without requiring labeled validation data and can be optimized in a fast and federated procedure. The key insight of our work is that a method that directly estimates the benefit of acquiring data for test set prediction is particularly compatible with a decentralized market setting."
                    },
                    {
                        "title": "Stochastic Zeroth-Order Optimization under Strongly Convexity and Lipschitz Hessian: Minimax Sample Complexity",
                        "abstract": "Optimization of convex functions under stochastic zeroth-order feedback has been a major and challenging question in online learning. In this work, we consider the problem of optimizing second-order smooth and strongly convex functions where the algorithm is only accessible to noisy evaluations of the objective function it queries. We provide the first tight characterization for the rate of the minimax simple regret by developing matching upper and lower bounds. We propose an algorithm that features a combination of a bootstrapping stage and a mirror-descent stage. Our main technical innovation consists of a sharp characterization for the spherical-sampling gradient estimator under higher-order smoothness conditions, which allows the algorithm to optimally balance the bias-variance tradeoff, and a new iterative method for the bootstrapping stage, which maintains the performance for unbounded Hessian."
                    },
                    {
                        "title": "Policy Mirror Descent for Regularized Reinforcement Learning: A Generalized Framework with Linear Convergence",
                        "abstract": "Policy optimization, which finds the desired policy by maximizing value functions via optimization techniques, lies at the heart of reinforcement learning (RL). In addition to value maximization, other practical considerations arise as well, including the need of encouraging exploration, and that of ensuring certain structural properties of the learned policy due to safety, resource and operational constraints. These can often be accounted for via regularized RL, which augments the target value function with a structure-promoting regularizer.   Focusing on discounted infinite-horizon Markov decision processes, we propose a generalized policy mirror descent (GPMD) algorithm for solving regularized RL. As a generalization of policy mirror descent (arXiv:2102.00135), our algorithm accommodates a general class of convex regularizers and promotes the use of Bregman divergence in cognizant of the regularizer in use. We demonstrate that our algorithm converges linearly to the global solution over an entire range of learning rates, in a dimension-free fashion, even when the regularizer lacks strong convexity and smoothness. In addition, this linear convergence feature is provably stable in the face of inexact policy evaluation and imperfect policy updates. Numerical experiments are provided to corroborate the appealing performance of GPMD."
                    },
                    {
                        "title": "A Dynamic Low-Rank Fast Gaussian Transform",
                        "abstract": "The \\emph{Fast Gaussian Transform} (FGT) enables subquadratic-time multiplication of an $n\\times n$ Gaussian kernel matrix $\\mathsf{K}_{i,j}= \\exp ( - \\| x_i - x_j \\|_2^2 ) $ with an arbitrary vector $h \\in \\mathbb{R}^n$, where $x_1,\\dots, x_n \\in \\mathbb{R}^d$ are a set of \\emph{fixed} source points. This kernel plays a central role in machine learning and random feature maps. Nevertheless, in most modern data analysis applications, datasets are dynamically changing (yet often have low rank), and recomputing the FGT from scratch in (kernel-based) algorithms incurs a major computational overhead ($\\gtrsim n$ time for a single source update $\\in \\mathbb{R}^d$). These applications motivate a \\emph{dynamic FGT} algorithm, which maintains a dynamic set of sources under \\emph{kernel-density estimation} (KDE) queries in \\emph{sublinear time} while retaining Mat-Vec multiplication accuracy and speed.   Assuming the dynamic data-points $x_i$ lie in a (possibly changing) $k$-dimensional subspace ($k\\leq d$), our main result is an efficient dynamic FGT algorithm, supporting the following operations in $\\log^{O(k)}(n/\\varepsilon)$ time: (1) Adding or deleting a source point, and (2) Estimating the ``kernel-density'' of a query point with respect to sources with $\\varepsilon$ additive accuracy. The core of the algorithm is a dynamic data structure for maintaining the \\emph{projected} ``interaction rank'' between source and target boxes, decoupled into finite truncation of Taylor and Hermite expansions."
                    },
                    {
                        "title": "Going Beyond Linear RL: Sample Efficient Neural Function Approximation",
                        "abstract": "Deep Reinforcement Learning (RL) powered by neural net approximation of the Q function has had enormous empirical success. While the theory of RL has traditionally focused on linear function approximation (or eluder dimension) approaches, little is known about nonlinear RL with neural net approximations of the Q functions. This is the focus of this work, where we study function approximation with two-layer neural networks (considering both ReLU and polynomial activation functions). Our first result is a computationally and statistically efficient algorithm in the generative model setting under completeness for two-layer neural networks. Our second result considers this setting but under only realizability of the neural net function class. Here, assuming deterministic dynamics, the sample complexity scales linearly in the algebraic dimension. In all cases, our results significantly improve upon what can be attained with linear (or eluder dimension) methods."
                    },
                    {
                        "title": "Optimal Gradient-based Algorithms for Non-concave Bandit Optimization",
                        "abstract": "Bandit problems with linear or concave reward have been extensively studied, but relatively few works have studied bandits with non-concave reward. This work considers a large family of bandit problems where the unknown underlying reward function is non-concave, including the low-rank generalized linear bandit problems and two-layer neural network with polynomial activation bandit problem. For the low-rank generalized linear bandit problem, we provide a minimax-optimal algorithm in the dimension, refuting both conjectures in [LMT21, JWWN19]. Our algorithms are based on a unified zeroth-order optimization paradigm that applies in great generality and attains optimal rates in several structured polynomial settings (in the dimension). We further demonstrate the applicability of our algorithms in RL in the generative model setting, resulting in improved sample complexity over prior approaches. Finally, we show that the standard optimistic algorithms (e.g., UCB) are sub-optimal by dimension factors. In the neural net setting (with polynomial activation functions) with noiseless reward, we provide a bandit algorithm with sample complexity equal to the intrinsic algebraic dimension. Again, we show that optimistic approaches have worse sample complexity, polynomial in the extrinsic dimension (which could be exponentially worse in the polynomial degree)."
                    },
                    {
                        "title": "Towards Optimal Statistical Watermarking",
                        "abstract": "We study statistical watermarking by formulating it as a hypothesis testing problem, a general framework which subsumes all previous statistical watermarking methods. Key to our formulation is a coupling of the output tokens and the rejection region, realized by pseudo-random generators in practice, that allows non-trivial trade-offs between the Type I error and Type II error. We characterize the Uniformly Most Powerful (UMP) watermark in the general hypothesis testing setting and the minimax Type II error in the model-agnostic setting. In the common scenario where the output is a sequence of $n$ tokens, we establish nearly matching upper and lower bounds on the number of i.i.d. tokens required to guarantee small Type I and Type II errors. Our rate of $\\Theta(h^{-1} \\log (1/h))$ with respect to the average entropy per token $h$ highlights potentials for improvement from the rate of $h^{-2}$ in the previous works. Moreover, we formulate the robust watermarking problem where the user is allowed to perform a class of perturbations on the generated texts, and characterize the optimal Type II error of robust UMP tests via a linear programming problem. To the best of our knowledge, this is the first systematic statistical treatment on the watermarking problem with near-optimal rates in the i.i.d. setting, which might be of interest for future works."
                    }
                ]
            },
            "a2337666-0bd6-4ad7-abce-ad28f4ff878f": {
                "pk": "a2337666-0bd6-4ad7-abce-ad28f4ff878f",
                "name": "Shaolun Zhang",
                "collaborators": [
                    "XuDong Wang",
                    "Shufan Li",
                    "Konstantinos Kallidromitis",
                    "Kehan Li",
                    "Yusuke Kato",
                    "Kazuki Kozuka",
                    "Trevor Darrell"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Computer Vision",
                    "Interactive Learning",
                    "Segmentation"
                ],
                "publications": [
                    {
                        "title": "SegLLM: Multi-round Reasoning Segmentation",
                        "abstract": "We present SegLLM, a novel multi-round interactive reasoning segmentation model that enhances LLM-based segmentation by exploiting conversational memory of both visual and textual outputs. By leveraging a mask-aware multimodal LLM, SegLLM re-integrates previous segmentation results into its input stream, enabling it to reason about complex user intentions and segment objects in relation to previously identified entities, including positional, interactional, and hierarchical relationships, across multiple interactions. This capability allows SegLLM to respond to visual and text queries in a chat-like manner. Evaluated on the newly curated MRSeg benchmark, SegLLM outperforms existing methods in multi-round interactive reasoning segmentation by over 20%. Additionally, we observed that training on multi-round reasoning segmentation data enhances performance on standard single-round referring segmentation and localization tasks, resulting in a 5.5% increase in cIoU for referring expression segmentation and a 4.5% improvement in Acc@0.5 for referring expression localization."
                    }
                ]
            },
            "7d2baa69-309b-41d1-b093-c1105de29a8e": {
                "pk": "7d2baa69-309b-41d1-b093-c1105de29a8e",
                "name": "Michael Jordan",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "50e8d5c1-553d-43e0-8e3d-37953bbff169": {
                "pk": "50e8d5c1-553d-43e0-8e3d-37953bbff169",
                "name": "Jiantao Jiao",
                "collaborators": [
                    "Tsachy Weissman",
                    "Yanjun Han",
                    "Kartik Venkat",
                    "Thomas Courtade",
                    "Weihao Gao",
                    "Banghua Zhu",
                    "Jacob Steinhardt",
                    "Lin Zhang",
                    "Robert Nowak",
                    "Rajarshi Mukherjee"
                ],
                "domain": [
                    "Statistical Estimation",
                    "Information Theory",
                    "Bias Analysis",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Bias Correction with Jackknife, Bootstrap, and Taylor Series",
                        "abstract": "We analyze bias correction methods using jackknife, bootstrap, and Taylor series. We focus on the binomial model, and consider the problem of bias correction for estimating $f(p)$, where $f \\in C[0,1]$ is arbitrary. We characterize the supremum norm of the bias of general jackknife and bootstrap estimators for any continuous functions, and demonstrate the in delete-$d$ jackknife, different values of $d$ may lead to drastically different behaviors in jackknife. We show that in the binomial model, iterating the bootstrap bias correction infinitely many times may lead to divergence of bias and variance, and demonstrate that the bias properties of the bootstrap bias corrected estimator after $r-1$ rounds are of the same order as that of the $r$-jackknife estimator if a bounded coefficients condition is satisfied."
                    },
                    {
                        "title": "An Extremal Inequality for Long Markov Chains",
                        "abstract": "Let $X,Y$ be jointly Gaussian vectors, and consider random variables $U,V$ that satisfy the Markov constraint $U-X-Y-V$. We prove an extremal inequality relating the mutual informations between all ${4 \\choose 2}$ pairs of random variables from the set $(U,X,Y,V)$. As a first application, we show that the rate region for the two-encoder quadratic Gaussian source coding problem follows as an immediate corollary of the the extremal inequality. In a second application, we establish the rate region for a vector-Gaussian source coding problem where L\\\"{o}wner-John ellipsoids are approximated based on rate-constrained descriptions of the data."
                    },
                    {
                        "title": "Adaptive Estimation of Shannon Entropy",
                        "abstract": "We consider estimating the Shannon entropy of a discrete distribution $P$ from $n$ i.i.d. samples. Recently, Jiao, Venkat, Han, and Weissman, and Wu and Yang constructed approximation theoretic estimators that achieve the minimax $L_2$ rates in estimating entropy. Their estimators are consistent given $n \\gg \\frac{S}{\\ln S}$ samples, where $S$ is the alphabet size, and it is the best possible sample complexity. In contrast, the Maximum Likelihood Estimator (MLE), which is the empirical entropy, requires $n\\gg S$ samples.   In the present paper we significantly refine the minimax results of existing work. To alleviate the pessimism of minimaxity, we adopt the adaptive estimation framework, and show that the minimax rate-optimal estimator in Jiao, Venkat, Han, and Weissman achieves the minimax rates simultaneously over a nested sequence of subsets of distributions $P$, without knowing the alphabet size $S$ or which subset $P$ lies in. In other words, their estimator is adaptive with respect to this nested sequence of the parameter space, which is characterized by the entropy of the distribution. We also characterize the maximum risk of the MLE over this nested sequence, and show, for every subset in the sequence, that the performance of the minimax rate-optimal estimator with $n$ samples is essentially that of the MLE with $n\\ln n$ samples, thereby further substantiating the generality of the phenomenon identified by Jiao, Venkat, Han, and Weissman."
                    },
                    {
                        "title": "Dependence Measures Bounding the Exploration Bias for General Measurements",
                        "abstract": "We propose a framework to analyze and quantify the bias in adaptive data analysis. It generalizes that proposed by Russo and Zou'15, applying to measurements whose moment generating function exists, measurements with a finite $p$-norm, and measurements in general Orlicz spaces. We introduce a new class of dependence measures which retain key properties of mutual information while more effectively quantifying the exploration bias for heavy tailed distributions. We provide examples of cases where our bounds are nearly tight in situations where the original framework of Russo and Zou'15 does not apply."
                    },
                    {
                        "title": "Generalizations of Maximal Inequalities to Arbitrary Selection Rules",
                        "abstract": "We present a generalization of the maximal inequalities that upper bound the expectation of the maximum of $n$ jointly distributed random variables. We control the expectation of a randomly selected random variable from $n$ jointly distributed random variables, and present bounds that are at least as tight as the classical maximal inequalities, and much tighter when the distribution of selection index is near deterministic. A new family of information theoretic measures were introduced in the process, which may be of independent interest."
                    },
                    {
                        "title": "Does Dirichlet Prior Smoothing Solve the Shannon Entropy Estimation Problem?",
                        "abstract": "The Dirichlet prior is widely used in estimating discrete distributions and functionals of discrete distributions. In terms of Shannon entropy estimation, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general they do \\emph{not} improve over the maximum likelihood estimator, which plugs-in the empirical distribution into the entropy functional. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation, as recently characterized by Jiao, Venkat, Han, and Weissman, and Wu and Yang. The performance of the minimax rate-optimal estimator with $n$ samples is essentially \\emph{at least} as good as that of the Dirichlet smoothed entropy estimators with $n\\ln n$ samples.   We harness the theory of approximation using positive linear operators for analyzing the bias of plug-in estimators for general functionals under arbitrary statistical models, thereby further consolidating the interplay between these two fields, which was thoroughly developed and exploited by Jiao, Venkat, Han, and Weissman. We establish new results in approximation theory, and apply them to analyze the bias of the Dirichlet prior smoothed plug-in entropy estimator. This interplay between bias analysis and approximation theory is of relevance and consequence far beyond the specific problem setting in this paper."
                    },
                    {
                        "title": "Minimax Estimation of the $L_1$ Distance",
                        "abstract": "We consider the problem of estimating the $L_1$ distance between two discrete probability measures $P$ and $Q$ from empirical data in a nonasymptotic and large alphabet setting. When $Q$ is known and one obtains $n$ samples from $P$, we show that for every $Q$, the minimax rate-optimal estimator with $n$ samples achieves performance comparable to that of the maximum likelihood estimator (MLE) with $n\\ln n$ samples. When both $P$ and $Q$ are unknown, we construct minimax rate-optimal estimators whose worst case performance is essentially that of the known $Q$ case with $Q$ being uniform, implying that $Q$ being uniform is essentially the most difficult case. The \\emph{effective sample size enlargement} phenomenon, identified in Jiao \\emph{et al.} (2015), holds both in the known $Q$ case for every $Q$ and the $Q$ unknown case. However, the construction of optimal estimators for $\\|P-Q\\|_1$ requires new techniques and insights beyond the approximation-based method of functional estimation in Jiao \\emph{et al.} (2015)."
                    },
                    {
                        "title": "Relations between Information and Estimation in Discrete-Time L\u00e9vy Channels",
                        "abstract": "Fundamental relations between information and estimation have been established in the literature for the discrete-time Gaussian and Poisson channels. In this work, we demonstrate that such relations hold for a much larger class of observation models. We introduce the natural family of discrete-time L\\'evy channels where the distribution of the output conditioned on the input is infinitely divisible. For L\\'evy channels, we establish new representations relating the mutual information between the channel input and output to an optimal expected estimation loss, thereby unifying and considerably extending results from the Gaussian and Poisson settings. We demonstrate the richness of our results by working out two examples of L\\'evy channels, namely the gamma channel and the negative binomial channel, with corresponding relations between information and estimation. Extensions to the setting of mismatched estimation are also presented."
                    },
                    {
                        "title": "The Nearest Neighbor Information Estimator is Adaptively Near Minimax Rate-Optimal",
                        "abstract": "We analyze the Kozachenko--Leonenko (KL) nearest neighbor estimator for the differential entropy. We obtain the first uniform upper bound on its performance over H\\\"older balls on a torus without assuming any conditions on how close the density could be from zero. Accompanying a new minimax lower bound over the H\\\"older ball, we show that the KL estimator is achieving the minimax rates up to logarithmic factors without cognizance of the smoothness parameter $s$ of the H\\\"older ball for $s\\in (0,2]$ and arbitrary dimension $d$, rendering it the first estimator that provably satisfies this property."
                    },
                    {
                        "title": "When does the Tukey median work?",
                        "abstract": "We analyze the performance of the Tukey median estimator under total variation (TV) distance corruptions. Previous results show that under Huber's additive corruption model, the breakdown point is 1/3 for high-dimensional halfspace-symmetric distributions. We show that under TV corruptions, the breakdown point reduces to 1/4 for the same set of distributions. We also show that a certain projection algorithm can attain the optimal breakdown point of 1/2. Both the Tukey median estimator and the projection algorithm achieve sample complexity linear in dimension."
                    },
                    {
                        "title": "Local moment matching: A unified methodology for symmetric functional estimation and distribution estimation under Wasserstein distance",
                        "abstract": "We present \\emph{Local Moment Matching (LMM)}, a unified methodology for symmetric functional estimation and distribution estimation under Wasserstein distance. We construct an efficiently computable estimator that achieves the minimax rates in estimating the distribution up to permutation, and show that the plug-in approach of our unlabeled distribution estimator is \"universal\" in estimating symmetric functionals of discrete distributions. Instead of doing best polynomial approximation explicitly as in existing literature of functional estimation, the plug-in approach conducts polynomial approximation implicitly and attains the optimal sample complexity for the entropy, power sum and support size functionals."
                    },
                    {
                        "title": "Beyond Maximum Likelihood: from Theory to Practice",
                        "abstract": "Maximum likelihood is the most widely used statistical estimation technique. Recent work by the authors introduced a general methodology for the construction of estimators for functionals in parametric models, and demonstrated improvements - both in theory and in practice - over the maximum likelihood estimator (MLE), particularly in high dimensional scenarios involving parameter dimension comparable to or larger than the number of samples. This approach to estimation, building on results from approximation theory, is shown to yield minimax rate-optimal estimators for a wide class of functionals, implementable with modest computational requirements. In a nutshell, a message of this recent work is that, for a wide class of functionals, the performance of these essentially optimal estimators with $n$ samples is comparable to that of the MLE with $n \\ln n$ samples.   In the present paper, we highlight the applicability of the aforementioned methodology to statistical problems beyond functional estimation, and show that it can yield substantial gains. For example, we demonstrate that for learning tree-structured graphical models, our approach achieves a significant reduction of the required data size compared with the classical Chow--Liu algorithm, which is an implementation of the MLE, to achieve the same accuracy. The key step in improving the Chow--Liu algorithm is to replace the empirical mutual information with the estimator for mutual information proposed by the authors. Further, applying the same replacement approach to classical Bayesian network classification, the resulting classifiers uniformly outperform the previous classifiers on 26 widely used datasets."
                    },
                    {
                        "title": "Mutual Information, Relative Entropy and Estimation Error in Semi-martingale Channels",
                        "abstract": "Fundamental relations between information and estimation have been established in the literature for the continuous-time Gaussian and Poisson channels, in a long line of work starting from the classical representation theorems by Duncan and Kabanov respectively. In this work, we demonstrate that such relations hold for a much larger family of continuous-time channels. We introduce the family of semi-martingale channels where the channel output is a semi-martingale stochastic process, and the channel input modulates the characteristics of the semi-martingale. For these channels, which includes as a special case the continuous time Gaussian and Poisson models, we establish new representations relating the mutual information between the channel input and output to an optimal causal filtering loss, thereby unifying and considerably extending results from the Gaussian and Poisson settings. Extensions to the setting of mismatched estimation are also presented where the relative entropy between the laws governing the output of the channel under two different input distributions is equal to the cumulative difference between the estimation loss incurred by using the mismatched and optimal causal filters respectively. The main tool underlying these results is the Doob--Meyer decomposition of a class of likelihood ratio sub-martingales. The results in this work can be viewed as the continuous-time analogues of recent generalizations for relations between information and estimation for discrete-time L\\'evy channels."
                    },
                    {
                        "title": "Minimax-Optimal Bounds for Detectors Based on Estimated Prior Probabilities",
                        "abstract": "In many signal detection and classification problems, we have knowledge of the distribution under each hypothesis, but not the prior probabilities. This paper is aimed at providing theory to quantify the performance of detection via estimating prior probabilities from either labeled or unlabeled training data. The error or {\\em risk} is considered as a function of the prior probabilities. We show that the risk function is locally Lipschitz in the vicinity of the true prior probabilities, and the error of detectors based on estimated prior probabilities depends on the behavior of the risk function in this locality. In general, we show that the error of detectors based on the Maximum Likelihood Estimate (MLE) of the prior probabilities converges to the Bayes error at a rate of $n^{-1/2}$, where $n$ is the number of training data. If the behavior of the risk function is more favorable, then detectors based on the MLE have errors converging to the corresponding Bayes errors at optimal rates of the form $n^{-(1+\\alpha)/2}$, where $\\alpha>0$ is a parameter governing the behavior of the risk function with a typical value $\\alpha = 1$. The limit $\\alpha \\rightarrow \\infty$ corresponds to a situation where the risk function is flat near the true probabilities, and thus insensitive to small errors in the MLE; in this case the error of the detector based on the MLE converges to the Bayes error exponentially fast with $n$. We show the bounds are achievable no matter given labeled or unlabeled training data and are minimax-optimal in labeled case."
                    },
                    {
                        "title": "Minimax Estimation of Functionals of Discrete Distributions",
                        "abstract": "We propose a general methodology for the construction and analysis of minimax estimators for a wide class of functionals of finite dimensional parameters, and elaborate on the case of discrete distributions, where the alphabet size $S$ is unknown and may be comparable with the number of observations $n$. We treat the respective regions where the functional is \"nonsmooth\" and \"smooth\" separately. In the \"nonsmooth\" regime, we apply an unbiased estimator for the best polynomial approximation of the functional whereas, in the \"smooth\" regime, we apply a bias-corrected Maximum Likelihood Estimator (MLE). We illustrate the merit of this approach by thoroughly analyzing two important cases: the entropy $H(P) = \\sum_{i = 1}^S -p_i \\ln p_i$ and $F_\\alpha(P) = \\sum_{i = 1}^S p_i^\\alpha,\\alpha>0$. We obtain the minimax $L_2$ rates for estimating these functionals. In particular, we demonstrate that our estimator achieves the optimal sample complexity $n \\asymp S/\\ln S$ for entropy estimation. We also show that the sample complexity for estimating $F_\\alpha(P),0<\\alpha<1$ is $n\\asymp S^{1/\\alpha}/ \\ln S$, which can be achieved by our estimator but not the MLE. For $1<\\alpha<3/2$, we show the minimax $L_2$ rate for estimating $F_\\alpha(P)$ is $(n\\ln n)^{-2(\\alpha-1)}$ regardless of the alphabet size, while the $L_2$ rate for the MLE is $n^{-2(\\alpha-1)}$. For all the above cases, the behavior of the minimax rate-optimal estimators with $n$ samples is essentially that of the MLE with $n\\ln n$ samples. We highlight the practical advantages of our schemes for entropy and mutual information estimation. We demonstrate that our approach reduces running time and boosts the accuracy compared to existing various approaches. Moreover, we show that the mutual information estimator induced by our methodology leads to significant performance boosts over the Chow--Liu algorithm in learning graphical models."
                    },
                    {
                        "title": "Maximum Likelihood Estimation of Functionals of Discrete Distributions",
                        "abstract": "We consider the problem of estimating functionals of discrete distributions, and focus on tight nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations, and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their \\emph{bias}.   We characterize the worst case squared error risk incurred by the Maximum Likelihood Estimator (MLE) in estimating the Shannon entropy $H(P) = \\sum_{i = 1}^S -p_i \\ln p_i$, and $F_\\alpha(P) = \\sum_{i = 1}^S p_i^\\alpha,\\alpha>0$, up to multiplicative constants, for any alphabet size $S\\leq \\infty$ and sample size $n$ for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have $n \\gg S$ observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider $n \\gg S^{1/\\alpha}$ samples for the MLE to consistently estimate $F_\\alpha(P), 0<\\alpha<1$. The minimax rate-optimal estimators for both problems require $S/\\ln S$ and $S^{1/\\alpha}/\\ln S$ samples, which implies that the MLE has a strictly sub-optimal sample complexity. When $1<\\alpha<3/2$, we show that the worst-case squared error rate of convergence for the MLE is $n^{-2(\\alpha-1)}$ for infinite alphabet size, while the minimax squared error rate is $(n\\ln n)^{-2(\\alpha-1)}$. When $\\alpha\\geq 3/2$, the MLE achieves the minimax optimal rate $n^{-1}$ regardless of the alphabet size.   As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. We show that no matter how we tune the parameters in the Dirichlet prior, this technique cannot achieve the minimax rates in entropy estimation."
                    },
                    {
                        "title": "Minimax Rate-Optimal Estimation of Divergences between Discrete Distributions",
                        "abstract": "We study the minimax estimation of $\\alpha$-divergences between discrete distributions for integer $\\alpha\\ge 1$, which include the Kullback--Leibler divergence and the $\\chi^2$-divergences as special examples. Dropping the usual theoretical tricks to acquire independence, we construct the first minimax rate-optimal estimator which does not require any Poissonization, sample splitting, or explicit construction of approximating polynomials. The estimator uses a hybrid approach which solves a problem-independent linear program based on moment matching in the non-smooth regime, and applies a problem-dependent bias-corrected plug-in estimator in the smooth regime, with a soft decision boundary between these regimes."
                    },
                    {
                        "title": "On Estimation of $L_{r}$-Norms in Gaussian White Noise Models",
                        "abstract": "We provide a complete picture of asymptotically minimax estimation of $L_r$-norms (for any $r\\ge 1$) of the mean in Gaussian white noise model over Nikolskii-Besov spaces. In this regard, we complement the work of Lepski, Nemirovski and Spokoiny (1999), who considered the cases of $r=1$ (with poly-logarithmic gap between upper and lower bounds) and $r$ even (with asymptotically sharp upper and lower bounds) over H\\\"{o}lder spaces. We additionally consider the case of asymptotically adaptive minimax estimation and demonstrate a difference between even and non-even $r$ in terms of an investigator's ability to produce asymptotically adaptive minimax estimators without paying a penalty."
                    },
                    {
                        "title": "Approximate Profile Maximum Likelihood",
                        "abstract": "We propose an efficient algorithm for approximate computation of the profile maximum likelihood (PML), a variant of maximum likelihood maximizing the probability of observing a sufficient statistic rather than the empirical sample. The PML has appealing theoretical properties, but is difficult to compute exactly. Inspired by observations gleaned from exactly solvable cases, we look for an approximate PML solution, which, intuitively, clumps comparably frequent symbols into one symbol. This amounts to lower-bounding a certain matrix permanent by summing over a subgroup of the symmetric group rather than the whole group during the computation. We extensively experiment with the approximate solution, and find the empirical performance of our approach is competitive and sometimes significantly better than state-of-the-art performance for various estimation problems."
                    },
                    {
                        "title": "Minimax redundancy for Markov chains with large state space",
                        "abstract": "For any Markov source, there exist universal codes whose normalized codelength approaches the Shannon limit asymptotically as the number of samples goes to infinity. This paper investigates how fast the gap between the normalized codelength of the \"best\" universal compressor and the Shannon limit (i.e. the compression redundancy) vanishes non-asymptotically in terms of the alphabet size and mixing time of the Markov source. We show that, for Markov sources whose relaxation time is at least $1 + \\frac{(2+c)}{\\sqrt{k}}$, where $k$ is the state space size (and $c>0$ is a constant), the phase transition for the number of samples required to achieve vanishing compression redundancy is precisely $\\Theta(k^2)$."
                    }
                ]
            },
            "b0d003a0-2353-4758-a817-9b038e836e23": {
                "pk": "b0d003a0-2353-4758-a817-9b038e836e23",
                "name": "Yuandong Tian",
                "collaborators": [
                    "Xinyun Chen",
                    "Kevin Yang",
                    "Nanyun Peng",
                    "Dan Klein",
                    "Yan Zhu",
                    "Tianmin Shu",
                    "Shouyuan Chen",
                    "Sherman Wong",
                    "Liangjian Chen",
                    "Xinlei Chen"
                ],
                "domain": [
                    "Deep Learning",
                    "Reinforcement Learning",
                    "Self-Supervised Learning",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "A theoretical framework for deep locally connected ReLU network",
                        "abstract": "Understanding theoretical properties of deep and locally connected nonlinear network, such as deep convolutional neural network (DCNN), is still a hard problem despite its empirical success. In this paper, we propose a novel theoretical framework for such networks with ReLU nonlinearity. The framework explicitly formulates data distribution, favors disentangled representations and is compatible with common regularization techniques such as Batch Norm. The framework is built upon teacher-student setting, by expanding the student forward/backward propagation onto the teacher's computational graph. The resulting model does not impose unrealistic assumptions (e.g., Gaussian inputs, independence of activation, etc). Our framework could help facilitate theoretical analysis of many practical issues, e.g. overfitting, generalization, disentangled representations in deep networks."
                    },
                    {
                        "title": "An Analytical Formula of Population Gradient for two-layered ReLU network and its Applications in Convergence and Critical Point Analysis",
                        "abstract": "In this paper, we explore theoretical properties of training a two-layered ReLU network $g(\\mathbf{x}; \\mathbf{w}) = \\sum_{j=1}^K \\sigma(\\mathbf{w}_j^T\\mathbf{x})$ with centered $d$-dimensional spherical Gaussian input $\\mathbf{x}$ ($\\sigma$=ReLU). We train our network with gradient descent on $\\mathbf{w}$ to mimic the output of a teacher network with the same architecture and fixed parameters $\\mathbf{w}^*$. We show that its population gradient has an analytical formula, leading to interesting theoretical analysis of critical points and convergence behaviors. First, we prove that critical points outside the hyperplane spanned by the teacher parameters (\"out-of-plane\") are not isolated and form manifolds, and characterize in-plane critical-point-free regions for two ReLU case. On the other hand, convergence to $\\mathbf{w}^*$ for one ReLU node is guaranteed with at least $(1-\\epsilon)/2$ probability, if weights are initialized randomly with standard deviation upper-bounded by $O(\\epsilon/\\sqrt{d})$, consistent with empirical practice. For network with many ReLU nodes, we prove that an infinitesimal perturbation of weight initialization results in convergence towards $\\mathbf{w}^*$ (or its permutation), a phenomenon known as spontaneous symmetric-breaking (SSB) in physics. We assume no independence of ReLU activations. Simulation verifies our findings."
                    },
                    {
                        "title": "Understanding the Role of Nonlinearity in Training Dynamics of Contrastive Learning",
                        "abstract": "While the empirical success of self-supervised learning (SSL) heavily relies on the usage of deep nonlinear models, existing theoretical works on SSL understanding still focus on linear ones. In this paper, we study the role of nonlinearity in the training dynamics of contrastive learning (CL) on one and two-layer nonlinear networks with homogeneous activation $h(x) = h'(x)x$. We have two major theoretical discoveries. First, the presence of nonlinearity can lead to many local optima even in 1-layer setting, each corresponding to certain patterns from the data distribution, while with linear activation, only one major pattern can be learned. This suggests that models with lots of parameters can be regarded as a \\emph{brute-force} way to find these local optima induced by nonlinearity. Second, in the 2-layer case, linear activation is proven not capable of learning specialized weights into diverse patterns, demonstrating the importance of nonlinearity. In addition, for 2-layer setting, we also discover \\emph{global modulation}: those local patterns discriminative from the perspective of global-level patterns are prioritized to learn, further characterizing the learning process. Simulation verifies our theoretical findings."
                    },
                    {
                        "title": "Composing Global Optimizers to Reasoning Tasks via Algebraic Objects in Neural Nets",
                        "abstract": "We prove rich algebraic structures of the solution space for 2-layer neural networks with quadratic activation and $L_2$ loss, trained on reasoning tasks in Abelian group (e.g., modular addition). Such a rich structure enables analytical construction of global optimal solutions from partial solutions that only satisfy part of the loss, despite its high nonlinearity. We coin the framework as CoGO (Composing Global Optimizers). Specifically, we show that the weight space over different numbers of hidden nodes of the 2-layer network is equipped with a semi-ring algebraic structure, and the loss function to be optimized consists of monomial potentials, which are ring homomorphism, allowing partial solutions to be composed into global ones by ring addition and multiplication. Our experiments show that around $95\\%$ of the solutions obtained by gradient descent match exactly our theoretical constructions. Although the global optimizers constructed only required a small number of hidden nodes, our analysis on gradient dynamics shows that over-parameterization asymptotically decouples training dynamics and is beneficial. We further show that training dynamics favors simpler solutions under weight decay, and thus high-order global optimizers such as perfect memorization are unfavorable."
                    },
                    {
                        "title": "Understanding Deep Contrastive Learning via Coordinate-wise Optimization",
                        "abstract": "We show that Contrastive Learning (CL) under a broad family of loss functions (including InfoNCE) has a unified formulation of coordinate-wise optimization on the network parameter $\\boldsymbol{\\theta}$ and pairwise importance $\\alpha$, where the \\emph{max player} $\\boldsymbol{\\theta}$ learns representation for contrastiveness, and the \\emph{min player} $\\alpha$ puts more weights on pairs of distinct samples that share similar representations. The resulting formulation, called $\\alpha$-CL, unifies not only various existing contrastive losses, which differ by how sample-pair importance $\\alpha$ is constructed, but also is able to extrapolate to give novel contrastive losses beyond popular ones, opening a new avenue of contrastive loss design. These novel losses yield comparable (or better) performance on CIFAR10, STL-10 and CIFAR-100 than classic InfoNCE. Furthermore, we also analyze the max player in detail: we prove that with fixed $\\alpha$, max player is equivalent to Principal Component Analysis (PCA) for deep linear network, and almost all local minima are global and rank-1, recovering optimal PCA solutions. Finally, we extend our analysis on max player to 2-layer ReLU networks, showing that its fixed points can have higher ranks."
                    },
                    {
                        "title": "Student Specialization in Deep ReLU Networks With Finite Width and Input Dimension",
                        "abstract": "We consider a deep ReLU / Leaky ReLU student network trained from the output of a fixed teacher network of the same depth, with Stochastic Gradient Descent (SGD). The student network is \\emph{over-realized}: at each layer $l$, the number $n_l$ of student nodes is more than that ($m_l$) of teacher. Under mild conditions on dataset and teacher network, we prove that when the gradient is small at every data sample, each teacher node is \\emph{specialized} by at least one student node \\emph{at the lowest layer}. For two-layer network, such specialization can be achieved by training on any dataset of \\emph{polynomial} size $\\mathcal{O}( K^{5/2} d^3 \\epsilon^{-1})$. until the gradient magnitude drops to $\\mathcal{O}(\\epsilon/K^{3/2}\\sqrt{d})$. Here $d$ is the input dimension, $K = m_1 + n_1$ is the total number of neurons in the lowest layer of teacher and student. Note that we require a specific form of data augmentation and the sample complexity includes the additional data generated from augmentation. To our best knowledge, we are the first to give polynomial sample complexity for student specialization of training two-layer (Leaky) ReLU networks with finite depth and width in teacher-student setting, and finite complexity for the lowest layer specialization in multi-layer case, without parametric assumption of the input (like Gaussian). Our theory suggests that teacher nodes with large fan-out weights get specialized first when the gradient is still large, while others are specialized with small gradient, which suggests inductive bias in training. This shapes the stage of training as empirically observed in multiple previous works. Experiments on synthetic and CIFAR10 verify our findings. The code is released in https://github.com/facebookresearch/luckmatters."
                    },
                    {
                        "title": "Better Computer Go Player with Neural Network and Long-term Prediction",
                        "abstract": "Competing with top human players in the ancient game of Go has been a long-term goal of artificial intelligence. Go's high branching factor makes traditional search techniques ineffective, even on leading-edge hardware, and Go's evaluation function could change drastically with one stone change. Recent works [Maddison et al. (2015); Clark & Storkey (2015)] show that search is not strictly necessary for machine Go players. A pure pattern-matching approach, based on a Deep Convolutional Neural Network (DCNN) that predicts the next move, can perform as well as Monte Carlo Tree Search (MCTS)-based open source Go engines such as Pachi [Baudis & Gailly (2012)] if its search budget is limited. We extend this idea in our bot named darkforest, which relies on a DCNN designed for long-term predictions. Darkforest substantially improves the win rate for pattern-matching approaches against MCTS-based approaches, even with looser search budgets. Against human players, the newest versions, darkfores2, achieve a stable 3d level on KGS Go Server as a ranked bot, a substantial improvement upon the estimated 4k-5k ranks for DCNN reported in Clark & Storkey (2015) based on games against other machine players. Adding MCTS to darkfores2 creates a much stronger player named darkfmcts3: with 5000 rollouts, it beats Pachi with 10k rollouts in all 250 games; with 75k rollouts it achieves a stable 5d level in KGS server, on par with state-of-the-art Go AIs (e.g., Zen, DolBaram, CrazyStone) except for AlphaGo [Silver et al. (2016)]; with 110k rollouts, it won the 3rd place in January KGS Go Tournament."
                    },
                    {
                        "title": "Learning to Perform Local Rewriting for Combinatorial Optimization",
                        "abstract": "Search-based methods for hard combinatorial optimization are often guided by heuristics. Tuning heuristics in various conditions and situations is often time-consuming. In this paper, we propose NeuRewriter that learns a policy to pick heuristics and rewrite the local components of the current solution to iteratively improve it until convergence. The policy factorizes into a region-picking and a rule-picking component, each parameterized by a neural network trained with actor-critic methods in reinforcement learning. NeuRewriter captures the general structure of combinatorial problems and shows strong performance in three versatile tasks: expression simplification, online job scheduling and vehicle routing problems. NeuRewriter outperforms the expression simplification component in Z3; outperforms DeepRM and Google OR-tools in online job scheduling; and outperforms recent neural baselines and Google OR-tools in vehicle routing problems."
                    },
                    {
                        "title": "M$^3$RL: Mind-aware Multi-agent Management Reinforcement Learning",
                        "abstract": "Most of the prior work on multi-agent reinforcement learning (MARL) achieves optimal collaboration by directly controlling the agents to maximize a common reward. In this paper, we aim to address this from a different angle. In particular, we consider scenarios where there are self-interested agents (i.e., worker agents) which have their own minds (preferences, intentions, skills, etc.) and can not be dictated to perform tasks they do not wish to do. For achieving optimal coordination among these agents, we train a super agent (i.e., the manager) to manage them by first inferring their minds based on both current and past observations and then initiating contracts to assign suitable tasks to workers and promise to reward them with corresponding bonuses so that they will agree to work together. The objective of the manager is maximizing the overall productivity as well as minimizing payments made to the workers for ad-hoc worker teaming. To train the manager, we propose Mind-aware Multi-agent Management Reinforcement Learning (M^3RL), which consists of agent modeling and policy learning. We have evaluated our approach in two environments, Resource Collection and Crafting, to simulate multi-agent management problems with various task settings and multiple designs for the worker agents. The experimental results have validated the effectiveness of our approach in modeling worker agents' minds online, and in achieving optimal ad-hoc teaming with good generalization and fast adaptation."
                    },
                    {
                        "title": "Extending Context Window of Large Language Models via Positional Interpolation",
                        "abstract": "We present Position Interpolation (PI) that extends the context window sizes of RoPE-based pretrained LLMs such as LLaMA models to up to 32768 with minimal fine-tuning (within 1000 steps), while demonstrating strong empirical results on various tasks that require long context, including passkey retrieval, language modeling, and long document summarization from LLaMA 7B to 65B. Meanwhile, the extended model by Position Interpolation preserve quality relatively well on tasks within its original context window. To achieve this goal, Position Interpolation linearly down-scales the input position indices to match the original context window size, rather than extrapolating beyond the trained context length which may lead to catastrophically high attention scores that completely ruin the self-attention mechanism. Our theoretical study shows that the upper bound of interpolation is at least $\\sim 600 \\times$ smaller than that of extrapolation, further demonstrating its stability. Models extended via Position Interpolation retain its original architecture and can reuse most pre-existing optimization and infrastructure."
                    },
                    {
                        "title": "Understanding self-supervised Learning Dynamics without Contrastive Pairs",
                        "abstract": "While contrastive approaches of self-supervised learning (SSL) learn representations by minimizing the distance between two augmented views of the same data point (positive pairs) and maximizing views from different data points (negative pairs), recent \\emph{non-contrastive} SSL (e.g., BYOL and SimSiam) show remarkable performance {\\it without} negative pairs, with an extra learnable predictor and a stop-gradient operation. A fundamental question arises: why do these methods not collapse into trivial representations? We answer this question via a simple theoretical study and propose a novel approach, DirectPred, that \\emph{directly} sets the linear predictor based on the statistics of its inputs, without gradient training. On ImageNet, it performs comparably with more complex two-layer non-linear predictors that employ BatchNorm and outperforms a linear predictor by $2.5\\%$ in 300-epoch training (and $5\\%$ in 60-epoch). DirectPred is motivated by our theoretical study of the nonlinear learning dynamics of non-contrastive SSL in simple linear networks. Our study yields conceptual insights into how non-contrastive SSL methods learn, how they avoid representational collapse, and how multiple factors, like predictor networks, stop-gradients, exponential moving averages, and weight decay all come into play. Our simple theory recapitulates the results of real-world ablation studies in both STL-10 and ImageNet. Code is released https://github.com/facebookresearch/luckmatters/tree/master/ssl."
                    },
                    {
                        "title": "Joint Policy Search for Multi-agent Collaboration with Imperfect Information",
                        "abstract": "To learn good joint policies for multi-agent collaboration with imperfect information remains a fundamental challenge. While for two-player zero-sum games, coordinate-ascent approaches (optimizing one agent's policy at a time, e.g., self-play) work with guarantees, in multi-agent cooperative setting they often converge to sub-optimal Nash equilibrium. On the other hand, directly modeling joint policy changes in imperfect information game is nontrivial due to complicated interplay of policies (e.g., upstream updates affect downstream state reachability). In this paper, we show global changes of game values can be decomposed to policy changes localized at each information set, with a novel term named policy-change density. Based on this, we propose Joint Policy Search(JPS) that iteratively improves joint policies of collaborative agents in imperfect information games, without re-evaluating the entire game. On multi-agent collaborative tabular games, JPS is proven to never worsen performance and can improve solutions provided by unilateral approaches (e.g, CFR), outperforming algorithms designed for collaborative policy learning (e.g. BAD). Furthermore, for real-world games, JPS has an online form that naturally links with gradient updates. We test it to Contract Bridge, a 4-player imperfect-information game where a team of $2$ collaborates to compete against the other. In its bidding phase, players bid in turn to find a good contract through a limited information channel. Based on a strong baseline agent that bids competitive bridge purely through domain-agnostic self-play, JPS improves collaboration of team players and outperforms WBridge5, a championship-winning software, by $+0.63$ IMPs (International Matching Points) per board over 1k games, substantially better than previous SoTA ($+0.41$ IMPs/b) under Double-Dummy evaluation."
                    },
                    {
                        "title": "Scan and Snap: Understanding Training Dynamics and Token Composition in 1-layer Transformer",
                        "abstract": "Transformer architecture has shown impressive performance in multiple research domains and has become the backbone of many neural network models. However, there is limited understanding on how it works. In particular, with a simple predictive loss, how the representation emerges from the gradient \\emph{training dynamics} remains a mystery. In this paper, for 1-layer transformer with one self-attention layer plus one decoder layer, we analyze its SGD training dynamics for the task of next token prediction in a mathematically rigorous manner. We open the black box of the dynamic process of how the self-attention layer combines input tokens, and reveal the nature of underlying inductive bias. More specifically, with the assumption (a) no positional encoding, (b) long input sequence, and (c) the decoder layer learns faster than the self-attention layer, we prove that self-attention acts as a \\emph{discriminative scanning algorithm}: starting from uniform attention, it gradually attends more to distinct key tokens for a specific next token to be predicted, and pays less attention to common key tokens that occur across different next tokens. Among distinct tokens, it progressively drops attention weights, following the order of low to high co-occurrence between the key and the query token in the training set. Interestingly, this procedure does not lead to winner-takes-all, but decelerates due to a \\emph{phase transition} that is controllable by the learning rates of the two layers, leaving (almost) fixed token combination. We verify this \\textbf{\\emph{scan and snap}} dynamics on synthetic and real-world data (WikiText)."
                    },
                    {
                        "title": "Channel-Recurrent Autoencoding for Image Modeling",
                        "abstract": "Despite recent successes in synthesizing faces and bedrooms, existing generative models struggle to capture more complex image types, potentially due to the oversimplification of their latent space constructions. To tackle this issue, building on Variational Autoencoders (VAEs), we integrate recurrent connections across channels to both inference and generation steps, allowing the high-level features to be captured in global-to-local, coarse-to-fine manners. Combined with adversarial loss, our channel-recurrent VAE-GAN (crVAE-GAN) outperforms VAE-GAN in generating a diverse spectrum of high resolution images while maintaining the same level of computational efficacy. Our model produces interpretable and expressive latent representations to benefit downstream tasks such as image completion. Moreover, we propose two novel regularizations, namely the KL objective weighting scheme over time steps and mutual information maximization between transformed latent variables and the outputs, to enhance the training."
                    },
                    {
                        "title": "Understanding Dimensional Collapse in Contrastive Self-supervised Learning",
                        "abstract": "Self-supervised visual representation learning aims to learn useful representations without relying on human annotations. Joint embedding approach bases on maximizing the agreement between embedding vectors from different views of the same image. Various methods have been proposed to solve the collapsing problem where all embedding vectors collapse to a trivial constant solution. Among these methods, contrastive learning prevents collapse via negative sample pairs. It has been shown that non-contrastive methods suffer from a lesser collapse problem of a different nature: dimensional collapse, whereby the embedding vectors end up spanning a lower-dimensional subspace instead of the entire available embedding space. Here, we show that dimensional collapse also happens in contrastive learning. In this paper, we shed light on the dynamics at play in contrastive learning that leads to dimensional collapse. Inspired by our theory, we propose a novel contrastive learning method, called DirectCLR, which directly optimizes the representation space without relying on an explicit trainable projector. Experiments show that DirectCLR outperforms SimCLR with a trainable linear projector on ImageNet."
                    },
                    {
                        "title": "Re3: Generating Longer Stories With Recursive Reprompting and Revision",
                        "abstract": "We consider the problem of automatically generating longer stories of over two thousand words. Compared to prior work on shorter stories, long-range plot coherence and relevance are more central challenges here. We propose the Recursive Reprompting and Revision framework (Re3) to address these challenges by (a) prompting a general-purpose language model to construct a structured overarching plan, and (b) generating story passages by repeatedly injecting contextual information from both the plan and current story state into a language model prompt. We then revise by (c) reranking different continuations for plot coherence and premise relevance, and finally (d) editing the best continuation for factual consistency. Compared to similar-length stories generated directly from the same base model, human evaluators judged substantially more of Re3's stories as having a coherent overarching plot (by 14% absolute increase), and relevant to the given initial premise (by 20%)."
                    },
                    {
                        "title": "DOC: Improving Long Story Coherence With Detailed Outline Control",
                        "abstract": "We propose the Detailed Outline Control (DOC) framework for improving long-range plot coherence when automatically generating several-thousand-word-long stories. DOC consists of two complementary components: a detailed outliner and a detailed controller. The detailed outliner creates a more detailed, hierarchically structured outline, shifting creative burden from the main drafting procedure to the planning stage. The detailed controller ensures the more detailed outline is still respected during generation by controlling story passages to align with outline details. In human evaluations of automatically generated stories, DOC substantially outperforms a strong Re3 baseline (Yang et al., 2022) on plot coherence (22.5% absolute gain), outline relevance (28.2%), and interestingness (20.7%). Humans also judged DOC to be much more controllable in an interactive generation setting."
                    },
                    {
                        "title": "Diffusion World Model: Future Modeling Beyond Step-by-Step Rollout for Offline Reinforcement Learning",
                        "abstract": "We introduce Diffusion World Model (DWM), a conditional diffusion model capable of predicting multistep future states and rewards concurrently. As opposed to traditional one-step dynamics models, DWM offers long-horizon predictions in a single forward pass, eliminating the need for recursive queries. We integrate DWM into model-based value estimation, where the short-term return is simulated by future trajectories sampled from DWM. In the context of offline reinforcement learning, DWM can be viewed as a conservative value regularization through generative modeling. Alternatively, it can be seen as a data source that enables offline Q-learning with synthetic data. Our experiments on the D4RL dataset confirm the robustness of DWM to long-horizon simulation. In terms of absolute performance, DWM significantly surpasses one-step dynamics models with a $44\\%$ performance gain, and is comparable to or slightly surpassing their model-free counterparts."
                    },
                    {
                        "title": "Building Generalizable Agents with a Realistic and Rich 3D Environment",
                        "abstract": "Teaching an agent to navigate in an unseen 3D environment is a challenging task, even in the event of simulated environments. To generalize to unseen environments, an agent needs to be robust to low-level variations (e.g. color, texture, object changes), and also high-level variations (e.g. layout changes of the environment). To improve overall generalization, all types of variations in the environment have to be taken under consideration via different level of data augmentation steps. To this end, we propose House3D, a rich, extensible and efficient environment that contains 45,622 human-designed 3D scenes of visually realistic houses, ranging from single-room studios to multi-storied houses, equipped with a diverse set of fully labeled 3D objects, textures and scene layouts, based on the SUNCG dataset (Song et.al.). The diversity in House3D opens the door towards scene-level augmentation, while the label-rich nature of House3D enables us to inject pixel- & task-level augmentations such as domain randomization (Toubin et. al.) and multi-task training. Using a subset of houses in House3D, we show that reinforcement learning agents trained with an enhancement of different levels of augmentations perform much better in unseen environments than our baselines with raw RGB input by over 8% in terms of navigation success rate. House3D is publicly available at http://github.com/facebookresearch/House3D."
                    },
                    {
                        "title": "Latent Execution for Neural Program Synthesis",
                        "abstract": "Program synthesis from input-output (IO) examples has been a long-standing challenge. While recent works demonstrated limited success on domain-specific languages (DSL), it remains highly challenging to apply them to real-world programming languages, such as C. Due to complicated syntax and token variation, there are three major challenges: (1) unlike many DSLs, programs in languages like C need to compile first and are not executed via interpreters; (2) the program search space grows exponentially when the syntax and semantics of the programming language become more complex; and (3) collecting a large-scale dataset of real-world programs is non-trivial. As a first step to address these challenges, we propose LaSynth and show its efficacy in a restricted-C domain. More specifically, LaSynth learns the latent representation to approximate the execution of partially generated programs, even if they are incomplete in syntax (addressing (1)). The learned execution significantly improves the performance of next token prediction over existing approaches, facilitating search (addressing (2)). Finally, once trained with randomly generated ground-truth programs and their IO pairs, LaSynth can synthesize more concise programs that resemble human-written code. Furthermore, retraining our model with these synthesized programs yields better performance with fewer samples for both Karel and C program synthesis, indicating the promise of leveraging the learned program synthesizer to improve the dataset quality for input-output program synthesis (addressing (3)). When evaluating on whether the program execution outputs match the IO pairs, LaSynth achieves 55.2% accuracy on generating simple C code with tens of tokens including loops and branches, outperforming existing approaches without executors by around 20%."
                    }
                ]
            },
            "3204c6b6-fbba-4c76-8a52-b92c60bb25b5": {
                "pk": "3204c6b6-fbba-4c76-8a52-b92c60bb25b5",
                "name": "Stuart Russell",
                "collaborators": [
                    "Gregory Lawrence",
                    "Nir Friedman",
                    "Andrew Critch",
                    "Bhaskara Marthi",
                    "Shaunak Chatterjee",
                    "Emma Brunskill",
                    "Brian Milch",
                    "Cassidy Laidlaw",
                    "Kevin Murphy",
                    "Daniel Dewey"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Probabilistic Inference",
                    "Decision Theory",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Fine-Grained Decision-Theoretic Search Control",
                        "abstract": "Decision-theoretic control of search has previously used as its basic unit. of computation the generation and evaluation of a complete set of successors. Although this simplifies analysis, it results in some lost opportunities for pruning and satisficing. This paper therefore extends the analysis of the value of computation to cover individual successor evaluations. The analytic techniques used may prove useful for control of reasoning in more general settings. A formula is developed for the expected value of a node, k of whose n successors have been evaluated. This formula is used to estimate the value of expanding further successors, using a general formula for the value of a computation in game-playing developed in earlier work. We exhibit an improved version of the MGSS* algorithm, giving empirical results for the game of Othello."
                    },
                    {
                        "title": "Improving Gradient Estimation by Incorporating Sensor Data",
                        "abstract": "An efficient policy search algorithm should estimate the local gradient of the objective function, with respect to the policy parameters, from as few trials as possible. Whereas most policy search methods estimate this gradient by observing the rewards obtained during policy trials, we show, both theoretically and empirically, that taking into account the sensor data as well gives better gradient estimates and hence faster learning. The reason is that rewards obtained during policy execution vary from trial to trial due to noise in the environment; sensor data, which correlates with the noise, can be used to partially correct for this variation, resulting in an estimatorwith lower variance."
                    },
                    {
                        "title": "A temporally abstracted Viterbi algorithm",
                        "abstract": "Hierarchical problem abstraction, when applicable, may offer exponential reductions in computational complexity. Previous work on coarse-to-fine dynamic programming (CFDP) has demonstrated this possibility using state abstraction to speed up the Viterbi algorithm. In this paper, we show how to apply temporal abstraction to the Viterbi problem. Our algorithm uses bounds derived from analysis of coarse timescales to prune large parts of the state trellis at finer timescales. We demonstrate improvements of several orders of magnitude over the standard Viterbi algorithm, as well as significant speedups over CFDP, for problems whose state variables evolve at widely differing rates."
                    },
                    {
                        "title": "RAPID: A Reachable Anytime Planner for Imprecisely-sensed Domains",
                        "abstract": "Despite the intractability of generic optimal partially observable Markov decision process planning, there exist important problems that have highly structured models. Previous researchers have used this insight to construct more efficient algorithms for factored domains, and for domains with topological structure in the flat state dynamics model. In our work, motivated by findings from the education community relevant to automated tutoring, we consider problems that exhibit a form of topological structure in the factored dynamics model. Our Reachable Anytime Planner for Imprecisely-sensed Domains (RAPID) leverages this structure to efficiently compute a good initial envelope of reachable states under the optimal MDP policy in time linear in the number of state variables. RAPID performs partially-observable planning over the limited envelope of states, and slowly expands the state space considered as time allows. RAPID performs well on a large tutoring-inspired problem simulation with 122 state variables, corresponding to a flat state space of over 10^30 states."
                    },
                    {
                        "title": "General-Purpose MCMC Inference over Relational Structures",
                        "abstract": "Tasks such as record linkage and multi-target tracking, which involve reconstructing the set of objects that underlie some observed data, are particularly challenging for probabilistic inference. Recent work has achieved efficient and accurate inference on such problems using Markov chain Monte Carlo (MCMC) techniques with customized proposal distributions. Currently, implementing such a system requires coding MCMC state representations and acceptance probability calculations that are specific to a particular application. An alternative approach, which we pursue in this paper, is to use a general-purpose probabilistic modeling language (such as BLOG) and a generic Metropolis-Hastings MCMC algorithm that supports user-supplied proposal distributions. Our algorithm gains flexibility by using MCMC states that are only partial descriptions of possible worlds; we provide conditions under which MCMC over partial worlds yields correct answers to queries. We also show how to use a context-specific Bayes net to identify the factors in the acceptance probability that need to be computed for a given proposed move. Experimental results on a citation matching task show that our general-purpose MCMC engine compares favorably with an application-specific system."
                    },
                    {
                        "title": "Image Segmentation in Video Sequences: A Probabilistic Approach",
                        "abstract": "\"Background subtraction\" is an old technique for finding moving objects in a video sequence for example, cars driving on a freeway. The idea is that subtracting the current image from a timeaveraged background image will leave only nonstationary objects. It is, however, a crude approximation to the task of classifying each pixel of the current image; it fails with slow-moving objects and does not distinguish shadows from moving objects. The basic idea of this paper is that we can classify each pixel using a model of how that pixel looks when it is part of different classes. We learn a mixture-of-Gaussians classification model for each pixel using an unsupervised technique- an efficient, incremental version of EM. Unlike the standard image-averaging approach, this automatically updates the mixture component for each class according to likelihood of membership; hence slow-moving objects are handled perfectly. Our approach also identifies and eliminates shadows much more effectively than other techniques such as thresholding. Application of this method as part of the Roadwatch traffic surveillance project is expected to result in significant improvements in vehicle identification and tracking."
                    },
                    {
                        "title": "Servant of Many Masters: Shifting priorities in Pareto-optimal sequential decision-making",
                        "abstract": "It is often argued that an agent making decisions on behalf of two or more principals who have different utility functions should adopt a {\\em Pareto-optimal} policy, i.e., a policy that cannot be improved upon for one agent without making sacrifices for another. A famous theorem of Harsanyi shows that, when the principals have a common prior on the outcome distributions of all policies, a Pareto-optimal policy for the agent is one that maximizes a fixed, weighted linear combination of the principals' utilities.   In this paper, we show that Harsanyi's theorem does not hold for principals with different priors, and derive a more precise generalization which does hold, which constitutes our main result. In this more general case, the relative weight given to each principal's utility should evolve over time according to how well the agent's observations conform with that principal's prior. The result has implications for the design of contracts, treaties, joint ventures, and robots."
                    },
                    {
                        "title": "TASRA: a Taxonomy and Analysis of Societal-Scale Risks from AI",
                        "abstract": "While several recent works have identified societal-scale and extinction-level risks to humanity arising from artificial intelligence, few have attempted an {\\em exhaustive taxonomy} of such risks. Many exhaustive taxonomies are possible, and some are useful -- particularly if they reveal new risks or practical approaches to safety. This paper explores a taxonomy based on accountability: whose actions lead to the risk, are the actors unified, and are they deliberate? We also provide stories to illustrate how the various risk types could each play out, including risks arising from unanticipated interactions of many AI systems, as well as risks from deliberate misuse, for which combined technical and policy solutions are indicated."
                    },
                    {
                        "title": "Uncertain Decisions Facilitate Better Preference Learning",
                        "abstract": "Existing observational approaches for learning human preferences, such as inverse reinforcement learning, usually make strong assumptions about the observability of the human's environment. However, in reality, people make many important decisions under uncertainty. To better understand preference learning in these cases, we study the setting of inverse decision theory (IDT), a previously proposed framework where a human is observed making non-sequential binary decisions under uncertainty. In IDT, the human's preferences are conveyed through their loss function, which expresses a tradeoff between different types of mistakes. We give the first statistical analysis of IDT, providing conditions necessary to identify these preferences and characterizing the sample complexity -- the number of decisions that must be observed to learn the tradeoff the human is making to a desired precision. Interestingly, we show that it is actually easier to identify preferences when the decision problem is more uncertain. Furthermore, uncertain decision problems allow us to relax the unrealistic assumption that the human is an optimal decision maker but still identify their exact preferences; we give sample complexities in this suboptimal case as well. Our analysis contradicts the intuition that partial observability should make preference learning more difficult. It also provides a first step towards understanding and improving preference learning methods for uncertain and suboptimal humans."
                    },
                    {
                        "title": "Learning the Structure of Dynamic Probabilistic Networks",
                        "abstract": "Dynamic probabilistic networks are a compact representation of complex stochastic processes. In this paper we examine how to learn the structure of a DPN from data. We extend structure scoring rules for standard probabilistic networks to the dynamic case, and show how to search for structure when some of the variables are hidden. Finally, we examine two applications where such a technology might be useful: predicting and classifying dynamic behaviors, and learning causal orderings in biological processes. We provide empirical results that demonstrate the applicability of our methods in both domains."
                    },
                    {
                        "title": "Research Priorities for Robust and Beneficial Artificial Intelligence",
                        "abstract": "Success in the quest for artificial intelligence has the potential to bring unprecedented benefits to humanity, and it is therefore worthwhile to investigate how to maximize these benefits while avoiding potential pitfalls. This article gives numerous examples (which should by no means be construed as an exhaustive list) of such worthwhile research aimed at ensuring that AI remains robust and beneficial."
                    },
                    {
                        "title": "Swift: Compiled Inference for Probabilistic Programming Languages",
                        "abstract": "A probabilistic program defines a probability measure over its semantic structures. One common goal of probabilistic programming languages (PPLs) is to compute posterior probabilities for arbitrary models and queries, given observed evidence, using a generic inference engine. Most PPL inference engines---even the compiled ones---incur significant runtime interpretation overhead, especially for contingent and open-universe models. This paper describes Swift, a compiler for the BLOG PPL. Swift-generated code incorporates optimizations that eliminate interpretation overhead, maintain dynamic dependencies efficiently, and handle memory management for possible worlds of varying sizes. Experiments comparing Swift with other PPL engines on a variety of inference problems demonstrate speedups ranging from 12x to 326x."
                    },
                    {
                        "title": "A compact, hierarchical Q-function decomposition",
                        "abstract": "Previous work in hierarchical reinforcement learning has faced a dilemma: either ignore the values of different possible exit states from a subroutine, thereby risking suboptimal behavior, or represent those values explicitly thereby incurring a possibly large representation cost because exit values refer to nonlocal aspects of the world (i.e., all subsequent rewards). This paper shows that, in many cases, one can avoid both of these problems. The solution is based on recursively decomposing the exit value function in terms of Q-functions at higher levels of the hierarchy. This leads to an intuitively appealing runtime architecture in which a parent subroutine passes to its child a value function on the exit states and the child reasons about how its choices affect the exit value. We also identify structural conditions on the value function and transition distributions that allow much more concise representations of exit state distributions, leading to further state abstraction. In essence, the only variables whose exit values need be considered are those that the parent cares about and the child affects. We demonstrate the utility of our algorithms on a series of increasingly complex environments."
                    },
                    {
                        "title": "Efficient Gradient Estimation for Motor Control Learning",
                        "abstract": "The task of estimating the gradient of a function in the presence of noise is central to several forms of reinforcement learning, including policy search methods. We present two techniques for reducing gradient estimation errors in the presence of observable input noise applied to the control signal. The first method extends the idea of a reinforcement baseline by fitting a local linear model to the function whose gradient is being estimated; we show how to find the linear model that minimizes the variance of the gradient estimate, and how to estimate the model from data. The second method improves this further by discounting components of the gradient vector that have high variance. These methods are applied to the problem of motor control learning, where actuator noise has a significant influence on behavior. In particular, we apply the techniques to learn locally optimal controllers for a dart-throwing task using a simulated three-link arm; we demonstrate that proposed methods significantly improve the reward function gradient estimate and, consequently, the learning curve, over existing methods."
                    },
                    {
                        "title": "Decayed MCMC Filtering",
                        "abstract": "Filtering---estimating the state of a partially observable Markov process from a sequence of observations---is one of the most widely studied problems in control theory, AI, and computational statistics. Exact computation of the posterior distribution is generally intractable for large discrete systems and for nonlinear continuous systems, so a good deal of effort has gone into developing robust approximation algorithms. This paper describes a simple stochastic approximation algorithm for filtering called {em decayed MCMC}. The algorithm applies Markov chain Monte Carlo sampling to the space of state trajectories using a proposal distribution that favours flips of more recent state variables. The formal analysis of the algorithm involves a generalization of standard coupling arguments for MCMC convergence. We prove that for any ergodic underlying Markov process, the convergence time of decayed MCMC with inverse-polynomial decay remains bounded as the length of the observation sequence grows. We show experimentally that decayed MCMC is at least competitive with other approximation algorithms such as particle filtering."
                    },
                    {
                        "title": "Automated Construction of Sparse Bayesian Networks from Unstructured Probabilistic Models and Domain Information",
                        "abstract": "An algorithm for automated construction of a sparse Bayesian network given an unstructured probabilistic model and causal domain information from an expert has been developed and implemented. The goal is to obtain a network that explicitly reveals as much information regarding conditional independence as possible. The network is built incrementally adding one node at a time. The expert's information and a greedy heuristic that tries to keep the number of arcs added at each step to a minimum are used to guide the search for the next node to add. The probabilistic model is a predicate that can answer queries about independencies in the domain. In practice the model can be implemented in various ways. For example, the model could be a statistical independence test operating on empirical data or a deductive prover operating on a set of independence statements about the domain."
                    },
                    {
                        "title": "Probabilistic Model-Based Approach for Heart Beat Detection",
                        "abstract": "Nowadays, hospitals are ubiquitous and integral to modern society. Patients flow in and out of a veritable whirlwind of paperwork, consultations, and potential inpatient admissions, through an abstracted system that is not without flaws. One of the biggest flaws in the medical system is perhaps an unexpected one: the patient alarm system. One longitudinal study reported an 88.8% rate of false alarms, with other studies reporting numbers of similar magnitudes. These false alarm rates lead to a number of deleterious effects that manifest in a significantly lower standard of care across clinics.   This paper discusses a model-based probabilistic inference approach to identifying variables at a detection level. We design a generative model that complies with an overview of human physiology and perform approximate Bayesian inference. One primary goal of this paper is to justify a Bayesian modeling approach to increasing robustness in a physiological domain.   We use three data sets provided by Physionet, a research resource for complex physiological signals, in the form of the Physionet 2014 Challenge set-p1 and set-p2, as well as the MGH/MF Waveform Database. On the extended data set our algorithm is on par with the other top six submissions to the Physionet 2014 challenge."
                    },
                    {
                        "title": "Cooperative Inverse Reinforcement Learning",
                        "abstract": "For an autonomous system to be helpful to humans and to pose no unwarranted risks, it needs to align its values with those of the humans in its environment in such a way that its actions contribute to the maximization of value for the humans. We propose a formal definition of the value alignment problem as cooperative inverse reinforcement learning (CIRL). A CIRL problem is a cooperative, partial-information game with two agents, human and robot; both are rewarded according to the human's reward function, but the robot does not initially know what this is. In contrast to classical IRL, where the human is assumed to act optimally in isolation, optimal CIRL solutions produce behaviors such as active teaching, active learning, and communicative actions that are more effective in achieving value alignment. We show that computing optimal joint policies in CIRL games can be reduced to solving a POMDP, prove that optimality in isolation is suboptimal in CIRL, and derive an approximate CIRL algorithm."
                    },
                    {
                        "title": "Inverse reinforcement learning for video games",
                        "abstract": "Deep reinforcement learning achieves superhuman performance in a range of video game environments, but requires that a designer manually specify a reward function. It is often easier to provide demonstrations of a target behavior than to design a reward function describing that behavior. Inverse reinforcement learning (IRL) algorithms can infer a reward from demonstrations in low-dimensional continuous control environments, but there has been little work on applying IRL to high-dimensional video games. In our CNN-AIRL baseline, we modify the state-of-the-art adversarial IRL (AIRL) algorithm to use CNNs for the generator and discriminator. To stabilize training, we normalize the reward and increase the size of the discriminator training dataset. We additionally learn a low-dimensional state representation using a novel autoencoder architecture tuned for video game environments. This embedding is used as input to the reward network, improving the sample efficiency of expert demonstrations. Our method achieves high-level performance on the simple Catcher video game, substantially outperforming the CNN-AIRL baseline. We also score points on the Enduro Atari racing game, but do not match expert performance, highlighting the need for further work."
                    }
                ]
            }
        }
    },
    "2310.05469": {
        "paper_data": {
            "title": "Learning to Predict Structural Vibrations",
            "url": "http://arxiv.org/abs/2310.05469v3",
            "arxiv_id": "2310.05469",
            "authors": [
                "Jan van Delden",
                "Julius Schultz",
                "Christopher Blech",
                "Sabine C. Langer",
                "Timo L\u00fcddecke"
            ],
            "abstract": "In mechanical structures like airplanes, cars and houses, noise is generated and transmitted through vibrations. To take measures to reduce this noise, vibrations need to be simulated with expensive numerical computations. Surrogate deep learning models present a promising alternative to classical numerical simulations as they can be evaluated magnitudes faster, while trading-off accuracy. To quantify such trade-offs systematically and foster the development of methods, we present a benchmark on the task of predicting the vibration of harmonically excited plates. The benchmark features a total of 12000 plate geometries with varying forms of beadings, material and sizes with associated numerical solutions. To address the benchmark task, we propose a new network architecture, named Frequency-Query Operator, which is trained to map plate geometries to their vibration pattern given a specific excitation frequency. Applying principles from operator learning and implicit models for shape encoding, our approach effectively addresses the prediction of highly variable frequency response functions occurring in dynamic systems. To quantify the prediction quality, we introduce a set of evaluation metrics and evaluate the method on our vibrating-plates benchmark. Our method outperforms DeepONets, Fourier Neural Operators and more traditional neural network architectures. Code, dataset and visualizations: https://eckerlab.org/code/delden2023_plate",
            "introduction": "   1 Introduction    \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Figure 1: Left: We introduce a dataset of 12000 samples for predicting vibration patterns based on plate geometries. A harmonic force excites all plates, causing them to vibrate. The vibration patterns of the vibrating plates are obtained through numerical simulation. Diverse architectures are evaluated on the dataset. Right: Beading patterns are used in many vibrating technical systems. Here, on an oil filter, a washing machine and a disk drive. They increase the structural stiffness and alter the vibration.    Humans are exposed to noise in everyday life, which is unpleasant and unhealthy in the long term [1]. Therefore, designers and engineers work on reducing noise that occurs, for example, in cars, airplanes, and houses. In this work, we specifically consider vibrations in mechanical structures as a source of sound. Vibrating structures radiate sound into the surrounding air. For example in a car, the engine causes the chassis to vibrate, which then radiates sound into the interior of the car. By reducing the vibration energy of the chassis, the noise can be reduced.   Vibrations of mechanical structures depend on the frequency of the excitation force (e.g. by the engine). A special case occurs when the excitation frequency matches an eigenfrequency of a given structure. In this case, the external force adds energy in phase with the structure\u2019s natural vibration and amplifies the motion with each cycle. This continues until the energy added equals the energy lost due to damping, resulting in large vibration amplitudes. This effect is called resonance and leads to characteristic resonance peaks in the dynamic response of the system. At resonance frequencies, due to the higher vibration amplitudes, the noise emission is increased. A second distinctive feature of structural vibrations is the vibration pattern, i.e. the spatial field of vibration velocity amplitudes. With increasing frequency, these vibration patterns become more complex and exhibit more local maxima and minima (Figure 1, left) [2].   To reduce noise, the vibration patterns of a mechanical structure can be influenced through modifications to its design. One effective method is the placement of damping elements, that absorb vibrational energy and thereby reduce sound emission, but this adds weight and requires space. Another approach is introducing beadings, which are indentations in plate-like structures (Figure\u00a01, right). Beadings increase the local stiffness of a structure, resulting in a shift in the structure\u2019s eigenfrequencies and subsequent resonance peaks. When they are well-placed, beadings can reduce the vibration energy for a range of excitation frequencies by shifting the resonance peaks out of the range. Reducing the vibration energy for a specific range of frequencies is a goal in many applications, e.g. in automotive design, where a motor excites vibrations in a range of frequencies [3].   In this work, we focus on a crucial prerequisite for targeted modifications to a design: Computing its vibrational behavior. The finite element method (FEM) is an established approach for numerically solving partial differential equations. The geometry of a design is discretized into small elements and the solution of the PDE is approximated by simple functions, e.g. polynomial functions, defined on these elements. [4, 5]. This method enables the numerical simulation of vibration patterns,",
            "references": [
                {
                    "title": "Deep learning for frequency response prediction of a multimass oscillator",
                    "abstract": "Noise prevention in product development is becoming more and more important due to health effects and comfort restrictions. In the product development process, costly and time\u2010consuming simulations are carried out over parameter spaces, for example, via the finite element method, in order to find quiet product designs. The solution of dynamic systems is limited in their maximum frequency and the size of the parameter space. Therefore, the substitution of high\u2010fidelity models by machine learning approaches is desirable. In this contribution, we consider an academic benchmark: Training a neural network to predict the frequency response of a multimass oscillator. Neural network architectures based on multilayer perceptrons and transformers are investigated and compared with respect to their accuracy in frequency response prediction. Our investigations suggest that the transformer architecture is better suited, in terms of accuracy and in terms of capability to handle multiple system configurations in a single model. The code of this work is available at https://eckerlab.org/code/acoustics_mmo_2023"
                },
                {
                    "title": "AirfRANS: High Fidelity Computational Fluid Dynamics Dataset for Approximating Reynolds-Averaged Navier-Stokes Solutions",
                    "abstract": "Surrogate models are necessary to optimize meaningful quantities in physical dynamics as their recursive numerical resolutions are often prohibitively expensive. It is mainly the case for fluid dynamics and the resolution of Navier-Stokes equations. However, despite the fast-growing field of data-driven models for physical systems, reference datasets representing real-world phenomena are lacking. In this work, we develop AirfRANS, a dataset for studying the two-dimensional incompressible steady-state Reynolds-Averaged Navier-Stokes equations over airfoils at a subsonic regime and for different angles of attacks. We also introduce metrics on the stress forces at the surface of geometries and visualization of boundary layers to assess the capabilities of models to accurately predict the meaningful information of the problem. Finally, we propose deep learning baselines on four machine learning tasks to study AirfRANS under different constraints for generalization considerations: big and scarce data regime, Reynolds number, and angle of attack extrapolation."
                },
                {
                    "title": "PDEBENCH: An Extensive Benchmark for Scientific Machine Learning",
                    "abstract": "Machine learning-based modeling of physical systems has experienced increased interest in recent years. Despite some impressive progress, there is still a lack of benchmarks for Scientific ML that are easy to use but still challenging and representative of a wide range of problems. We introduce PDEBench, a benchmark suite of time-dependent simulation tasks based on Partial Differential Equations (PDEs). PDEBench comprises both code and data to benchmark the performance of novel machine learning models against both classical numerical simulations and machine learning baselines. Our proposed set of benchmark problems contribute the following unique features: (1) A much wider range of PDEs compared to existing benchmarks, ranging from relatively common examples to more realistic and difficult problems; (2) much larger ready-to-use datasets compared to prior work, comprising multiple simulation runs across a larger number of initial and boundary conditions and PDE parameters; (3) more extensible source codes with user-friendly APIs for data generation and baseline results with popular machine learning models (FNO, U-Net, PINN, Gradient-Based Inverse Method). PDEBench allows researchers to extend the benchmark freely for their own purposes using a standardized API and to compare the performance of new models to existing baseline methods. We also propose new evaluation metrics with the aim to provide a more holistic understanding of learning methods in the context of Scientific ML. With those metrics we identify tasks which are challenging for recent ML methods and propose these tasks as future challenges for the community. The code is available at https://github.com/pdebench/PDEBench."
                },
                {
                    "title": "NOMAD: Nonlinear Manifold Decoders for Operator Learning",
                    "abstract": "Supervised learning in function spaces is an emerging area of machine learning research with applications to the prediction of complex physical systems such as fluid flows, solid mechanics, and climate modeling. By directly learning maps (operators) between infinite dimensional function spaces, these models are able to learn discretization invariant representations of target functions. A common approach is to represent such target functions as linear combinations of basis elements learned from data. However, there are simple scenarios where, even though the target functions form a low dimensional submanifold, a very large number of basis elements is needed for an accurate linear representation. Here we present NOMAD, a novel operator learning framework with a nonlinear decoder map capable of learning finite dimensional representations of nonlinear submanifolds in function spaces. We show this method is able to accurately learn low dimensional representations of solution manifolds to partial differential equations while outperforming linear models of larger size. Additionally, we compare to state-of-the-art operator learning methods on a complex fluid dynamics benchmark and achieve competitive performance with a significantly smaller model size and training cost."
                },
                {
                    "title": "Applicability of Convolutional Neural Networks for Calibration of Nonlinear Dynamic Models of Structures",
                    "abstract": "The objective of this study is to examine a machine learning (ML) framework for calibrating the parameters of analytical models of complex nonlinear structural systems where experimental data is significantly limited. Because of the high cost of large-scale structural tests, analytical models are widely used to enhance the understanding of structural performance under complex loading environments. In this study, an ML framework is proposed and evaluated for the calibration of an analytical model representing a shake table test performed on a composite column developed in OpenSees software. A large number of parameters for modeling the constitutive behavior of the concrete core, steel reinforcement, exterior composite tube, as well as the interactions between the concrete core and the tube, base fixity, and nonlinear shear deformations are included. A convolutional neural network (CNN) architecture was used to calibrate these parameters by using the lateral load, displacement, and axial load time histories as input variables. First, a synthetic dataset is generated for permutations of different model parameters. Next, four CNNs were trained to evaluate the presentation of input data in time-domain and time-frequency domain. Finally, the trained model was prompted with real experimental data and the values of peak lateral force, residual displacement, and hysteresis energy dissipation from the analytical model were compared with those from the experiment. The results show that the proposed framework is appropriate for calibration of complex nonlinear structural models when experimental data is limited."
                },
                {
                    "title": "Machine learning\u2013enabled high-entropy alloy discovery",
                    "abstract": "High-entropy alloys are solid solutions of multiple principal elements that are capable of reaching composition and property regimes inaccessible for dilute materials. Discovering those with valuable properties, however, too often relies on serendipity, because thermodynamic alloy design rules alone often fail in high-dimensional composition spaces. We propose an active learning strategy to accelerate the design of high-entropy Invar alloys in a practically infinite compositional space based on very sparse data. Our approach works as a closed-loop, integrating machine learning with density-functional theory, thermodynamic calculations, and experiments. After processing and characterizing 17 new alloys out of millions of possible compositions, we identified two high-entropy Invar alloys with extremely low thermal expansion coefficients around 2 \u00d7 10\u22126 per degree kelvin at 300 kelvin. We believe this to be a suitable pathway for the fast and automated discovery of high-entropy alloys with optimal thermal, magnetic, and electrical properties. Description A little expansive Invar alloys have extremely low thermal expansion, making them attractive for several types of applications. Finding these types of alloys in a complex compositional space, however, is challenging. Rao et al. used an iterative scheme that combines machine learning, density functional theory, experiments, and thermodynamic calculation to find two new invar alloys out of millions of candidates (see the Perspective by Hu and Yang). The alloys are both compositionally complex, high entropy materials, thus demonstrating the power of this approach for materials discovery. \u2014BG Two high-entropy alloys with extremely low thermal expansion were found with the help of machine learning."
                },
                {
                    "title": "Conditional Wasserstein generative adversarial networks applied to acoustic metamaterial design.",
                    "abstract": "This work presents a method for the reduction of the total scattering cross section (TSCS) for a planar configuration of cylinders by means of generative modeling and deep learning. Currently, the minimization of TSCS requires repeated forward modelling at considerable computer resources, whereas deep learning can do this more efficiently. The conditional Wasserstein generative adversarial networks (cWGANs) model is proposed for minimization of TSCS in two dimensions by combining Wasserstein generative adversarial networks with convolutional neural networks to simulate TSCS of configuration of rigid scatterers. The proposed cWGAN model is enhanced by adding to it a coordinate convolution (CoordConv) layer. For a given number of cylinders, the cWGAN model generates images of 2D configurations of cylinders that minimize the TSCS. The proposed generative model is illustrated with examples for planar uniform configurations of rigid cylinders."
                },
                {
                    "title": "Characterizing possible failure modes in physics-informed neural networks",
                    "abstract": "Recent work in scientific machine learning has developed so-called physics-informed neural network (PINN) models. The typical approach is to incorporate physical domain knowledge as soft constraints on an empirical loss function and use existing machine learning methodologies to train the model. We demonstrate that, while existing PINN methodologies can learn good models for relatively trivial problems, they can easily fail to learn relevant physical phenomena for even slightly more complex problems. In particular, we analyze several distinct situations of widespread physical interest, including learning differential equations with convection, reaction, and diffusion operators. We provide evidence that the soft regularization in PINNs, which involves PDE-based differential operators, can introduce a number of subtle problems, including making the problem more ill-conditioned. Importantly, we show that these possible failure modes are not due to the lack of expressivity in the NN architecture, but that the PINN's setup makes the loss landscape very hard to optimize. We then describe two promising solutions to address these failure modes. The first approach is to use curriculum regularization, where the PINN's loss term starts from a simple PDE regularization, and becomes progressively more complex as the NN gets trained. The second approach is to pose the problem as a sequence-to-sequence learning task, rather than learning to predict the entire space-time at once. Extensive testing shows that we can achieve up to 1-2 orders of magnitude lower error with these methods as compared to regular PINN training."
                },
                {
                    "title": "An Extensible Benchmark Suite for Learning to Simulate Physical Systems",
                    "abstract": "Simulating physical systems is a core component of scientific computing, encompassing a wide range of physical domains and applications. Recently, there has been a surge in data-driven methods to complement traditional numerical simulations methods, motivated by the opportunity to reduce computational costs and/or learn new physical models leveraging access to large collections of data. However, the diversity of problem settings and applications has led to a plethora of approaches, each one evaluated on a different setup and with different evaluation metrics. We introduce a set of benchmark problems to take a step towards unified benchmarks and evaluation protocols. We propose four representative physical systems, as well as a collection of both widely used classical time integrators and representative data-driven methods (kernel-based, MLP, CNN, nearest neighbors). Our framework allows evaluating objectively and systematically the stability, accuracy, and computational efficiency of data-driven methods. Additionally, it is configurable to permit adjustments for accommodating other learning tasks and for establishing a foundation for future developments in machine learning for scientific computing."
                },
                {
                    "title": "Reinforcement learning applied to metamaterial design.",
                    "abstract": "This paper presents a semi-analytical method of suppressing acoustic scattering using reinforcement learning (RL) algorithms. We give a RL agent control over design parameters of a planar configuration of cylindrical scatterers in water. These design parameters control the position and radius of the scatterers. As these cylinders encounter an incident acoustic wave, the scattering pattern is described by a function called total scattering cross section (TSCS). Through evaluating the gradients of TSCS and other information about the state of the configuration, the RL agent perturbatively adjusts design parameters, considering multiple scattering between the scatterers. As each adjustment is made, the RL agent receives a reward negatively proportional to the root mean square of the TSCS across a range of wavenumbers. Through maximizing its reward per episode, the agent discovers designs with low scattering. Specifically, the double deep Q-learning network and the deep deterministic policy gradient algorithms are employed in our models. Designs discovered by the RL algorithms performed well when compared to a state-of-the-art optimization algorithm using fmincon."
                },
                {
                    "title": "An adaptive sparse grid rational Arnoldi method for uncertainty quantification of dynamical systems in the frequency\n domain",
                    "abstract": "In this paper, we address discrete linear systems in the frequency domain, where both frequency and random parameters are considered. Sampling such a system many times is computationally challenging if the size is too large, which is often the case for a finite element discretization of partial differential equations. We propose an adaptive method, steered by dual error indicators, which combines rational Arnoldi model order reduction and sparse grid interpolation with hierarchical Leja nodes. At each Leja node, a reduced order model (ROM) is constructed such that the ROM\u2010error in the frequency domain is balanced with the sparse grid error in the random parameter domain. Both the ROM basis and the sparse grid set are enlarged in a hierarchical manner to achieve a prescribed accuracy in statistical moments. Moreover, parameter sensitivities over the frequency range can be easily extracted from the combined reduced order\u2010surrogate model. In the numerical tests considered in the paper, the method employs sampling sets, which are reduced by at least an order of magnitude, compared to Monte Carlo simulation. Additionally, for an example from vibroacoustics, building a ROM reduces the system size by roughly a factor of 100, while still providing an acceptable accuracy."
                },
                {
                    "title": "Generative adversarial networks for the design of acoustic metamaterials.",
                    "abstract": "Metamaterials are attracting increasing interest in the field of acoustics due to their sound insulation effects. By periodically arranged structures, acoustic metamaterials can influence the way sound propagates in acoustic media. To date, the design of acoustic metamaterials relies primarily on the expertise of specialists since most effects are based on localized solutions and interference. This paper outlines a deep learning-based approach to extend current knowledge of metamaterial design in acoustics. We develop a design method by using conditional generative adversarial networks. The generative network proposes a cell candidate regarding a desired transmission behavior of the metamaterial. To validate our method, numerical simulations with the finite element method are performed. Our study reveals considerable insight into design strategies for sound insulation tasks. By providing design directives for acoustic metamaterials, cell candidates can be inspected and tailored to achieve desirable transmission characteristics."
                },
                {
                    "title": "Machine learning\u2013accelerated computational fluid dynamics",
                    "abstract": "Significance Accurate simulation of fluids is important for many science and engineering problems but is very computationally demanding. In contrast, machine-learning models can approximate physics very quickly but at the cost of accuracy. Here we show that using machine learning inside traditional fluid simulations can improve both accuracy and speed, even on examples very different from the training data. Our approach opens the door to applying machine learning to large-scale physical modeling tasks like airplane design and climate prediction. Numerical simulation of fluids plays an essential role in modeling many physical phenomena, such as weather, climate, aerodynamics, and plasma physics. Fluids are well described by the Navier\u2013Stokes equations, but solving these equations at scale remains daunting, limited by the computational cost of resolving the smallest spatiotemporal features. This leads to unfavorable trade-offs between accuracy and tractability. Here we use end-to-end deep learning to improve approximations inside computational fluid dynamics for modeling two-dimensional turbulent flows. For both direct numerical simulation of turbulence and large-eddy simulation, our results are as accurate as baseline solvers with 8 to 10\u00d7 finer resolution in each spatial dimension, resulting in 40- to 80-fold computational speedups. Our method remains stable during long simulations and generalizes to forcing functions and Reynolds numbers outside of the flows where it is trained, in contrast to black-box machine-learning approaches. Our approach exemplifies how scientific computing can leverage machine learning and hardware accelerators to improve simulations without sacrificing accuracy or generalization."
                },
                {
                    "title": "Quadratic Residual Networks: A New Class of Neural Networks for Solving Forward and Inverse Problems in Physics Involving PDEs",
                    "abstract": "We propose quadratic residual networks (QRes) as a new type of parameter-efficient neural network architecture, by adding a quadratic residual term to the weighted sum of inputs before applying activation functions. With sufficiently high functional capacity (or expressive power), we show that it is especially powerful for solving forward and inverse physics problems involving partial differential equations (PDEs). Using tools from algebraic geometry, we theoretically demonstrate that, in contrast to plain neural networks, QRes shows better parameter efficiency in terms of network width and depth thanks to higher non-linearity in every neuron. Finally, we empirically show that QRes shows faster convergence speed in terms of number of training epochs especially in learning complex patterns."
                },
                {
                    "title": "pixelNeRF: Neural Radiance Fields from One or Few Images",
                    "abstract": "We propose pixelNeRF, a learning framework that predicts a continuous neural scene representation conditioned on one or few input images. The existing approach for constructing neural radiance fields [27] involves optimizing the representation to every scene independently, requiring many calibrated views and significant compute time. We take a step towards resolving these shortcomings by introducing an architecture that conditions a NeRF on image inputs in a fully convolutional manner. This allows the network to be trained across multiple scenes to learn a scene prior, enabling it to perform novel view synthesis in a feed-forward manner from a sparse set of views (as few as one). Leveraging the volume rendering approach of NeRF, our model can be trained directly from images with no explicit 3D supervision. We conduct extensive experiments on ShapeNet benchmarks for single image novel view synthesis tasks with held-out objects as well as entire unseen categories. We further demonstrate the flexibility of pixelNeRF by demonstrating it on multi-object ShapeNet scenes and real scenes from the DTU dataset. In all cases, pixelNeRF outperforms current state-of-the-art baselines for novel view synthesis and single image 3D reconstruction. For the video and code, please visit the project website:https://alexyu.net/pixelnerf."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Fourier Neural Operator for Parametric Partial Differential Equations",
                    "abstract": "The classical development of neural networks has primarily focused on learning mappings between finite-dimensional Euclidean spaces. Recently, this has been generalized to neural operators that learn mappings between function spaces. For partial differential equations (PDEs), neural operators directly learn the mapping from any functional parametric dependence to the solution. Thus, they learn an entire family of PDEs, in contrast to classical methods which solve one instance of the equation. In this work, we formulate a new neural operator by parameterizing the integral kernel directly in Fourier space, allowing for an expressive and efficient architecture. We perform experiments on Burgers' equation, Darcy flow, and the Navier-Stokes equation (including the turbulent regime). Our Fourier neural operator shows state-of-the-art performance compared to existing neural network methodologies and it is up to three orders of magnitude faster compared to traditional PDE solvers."
                },
                {
                    "title": "Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains",
                    "abstract": "We show that passing input points through a simple Fourier feature mapping enables a multilayer perceptron (MLP) to learn high-frequency functions in low-dimensional problem domains. These results shed light on recent advances in computer vision and graphics that achieve state-of-the-art results by using MLPs to represent complex 3D objects and scenes. Using tools from the neural tangent kernel (NTK) literature, we show that a standard MLP fails to learn high frequencies both in theory and in practice. To overcome this spectral bias, we use a Fourier feature mapping to transform the effective NTK into a stationary kernel with a tunable bandwidth. We suggest an approach for selecting problem-specific Fourier features that greatly improves the performance of MLPs for low-dimensional regression tasks relevant to the computer vision and graphics communities."
                },
                {
                    "title": "Implicit Neural Representations with Periodic Activation Functions",
                    "abstract": "Implicitly defined, continuous, differentiable signal representations parameterized by neural networks have emerged as a powerful paradigm, offering many possible benefits over conventional representations. However, current network architectures for such implicit neural representations are incapable of modeling signals with fine detail, and fail to represent a signal's spatial and temporal derivatives, despite the fact that these are essential to many physical signals defined implicitly as the solution to partial differential equations. We propose to leverage periodic activation functions for implicit neural representations and demonstrate that these networks, dubbed sinusoidal representation networks or Sirens, are ideally suited for representing complex natural signals and their derivatives. We analyze Siren activation statistics to propose a principled initialization scheme and demonstrate the representation of images, wavefields, video, sound, and their derivatives. Further, we show how Sirens can be leveraged to solve challenging boundary value problems, such as particular Eikonal equations (yielding signed distance functions), the Poisson equation, and the Helmholtz and wave equations. Lastly, we combine Sirens with hypernetworks to learn priors over the space of Siren functions."
                },
                {
                    "title": "Predicting the Propagation of Acoustic Waves using Deep Convolutional Neural Networks",
                    "abstract": "A novel approach for numerically propagating acoustic waves in two-dimensional quiescent media has been developed through a fully convolutional multi-scale neural network. This data-driven method managed to produce accurate results for long simulation times with a database of Lattice Boltzmann temporal simulations of propagating Gaussian Pulses, even in the case of initial conditions unseen during training time, such as the plane wave configuration or the two initial Gaussian pulses of opposed amplitudes. Two different choices of optimization objectives are compared, resulting in an improved prediction accuracy when adding the spatial gradient difference error to the traditional mean squared error loss function. Further accuracy gains are observed when performing an a posteriori correction on the neural network prediction based on the conservation of acoustic energy, indicating the benefit of including physical information in data-driven methods."
                },
                {
                    "title": "CFDNet: a deep learning-based accelerator for fluid simulations",
                    "abstract": "CFD is widely used in physical system design and optimization, where it is used to predict engineering quantities of interest, such as the lift on a plane wing or the drag on a motor vehicle. However, many systems of interest are prohibitively expensive for design optimization, due to the expense of evaluating CFD simulations. To render the computation tractable, reduced-order or surrogate models are used to accelerate simulations while respecting the convergence constraints provided by the higher-fidelity solution. This paper introduces CFDNet - a physical simulation and deep learning coupled framework, for accelerating the convergence of Reynolds Averaged Navier-Stokes simulations. CFDNet is designed to predict the primary physical properties of the fluid including velocity, pressure, and eddy viscosity using a single convolutional neural network at its core. We evaluate CFDNet on a variety of use-cases, both extrapolative and interpolative, where test geometries are observed/not-observed during training. Our results show that CFDNet meets the convergence constraints of the domain-specific physics solver while outperforming it by 1.9 - 7.4X on both steady laminar and turbulent flows. Moreover, we demonstrate the generalization capacity of CFDNet by testing its prediction on new geometries unseen during training. In this case, the approach meets the CFD convergence criterion while still providing significant speedups over traditional domain-only models."
                },
                {
                    "title": "Learning to Simulate Complex Physics with Graph Networks",
                    "abstract": "Here we present a machine learning framework and model implementation that can learn to simulate a wide variety of challenging physical domains, involving fluids, rigid solids, and deformable materials interacting with one another. Our framework---which we term \"Graph Network-based Simulators\" (GNS)---represents the state of a physical system with particles, expressed as nodes in a graph, and computes dynamics via learned message-passing. Our results show that our model can generalize from single-timestep predictions with thousands of particles during training, to different initial conditions, thousands of timesteps, and at least an order of magnitude more particles at test time. Our model was robust to hyperparameter choices across various evaluation metrics: the main determinants of long-term performance were the number of message-passing steps, and mitigating the accumulation of error by corrupting the training data with noise. Our GNS framework advances the state-of-the-art in learned physical simulation, and holds promise for solving a wide range of complex forward and inverse problems."
                },
                {
                    "title": "Towards Physics-informed Deep Learning for Turbulent Flow Prediction",
                    "abstract": "While deep learning has shown tremendous success in a wide range of domains, it remains a grand challenge to incorporate physical principles in a systematic manner to the design, training, and inference of such models. In this paper, we aim to predict turbulent flow by learning its highly nonlinear dynamics from spatiotemporal velocity fields of large-scale fluid flow simulations of relevance to turbulence modeling and climate modeling. We adopt a hybrid approach by marrying two well-established turbulent flow simulation techniques with deep learning. Specifically, we introduce trainable spectral filters in a coupled model of Reynolds-averaged Navier-Stokes (RANS) and Large Eddy Simulation (LES), followed by a specialized U-net for prediction. Our approach, which we call Turbulent-Flow Net, is grounded in a principled physics model, yet offers the flexibility of learned representations. We compare our model with state-of-the-art baselines and observe significant reductions in error for predictions 60 frames ahead. Most importantly, our method predicts physical fields that obey desirable physical characteristics, such as conservation of mass, whilst faithfully emulating the turbulent kinetic energy field and spectrum, which are critical for accurate prediction of turbulent flows."
                },
                {
                    "title": "Machine Learning for Fluid Mechanics",
                    "abstract": "The field of fluid mechanics is rapidly advancing, driven by unprecedented volumes of data from experiments, field measurements, and large-scale simulations at multiple spatiotemporal scales. Machine learning (ML) offers a wealth of techniques to extract information from data that can be translated into knowledge about the underlying fluid mechanics. Moreover, ML algorithms can augment domain knowledge and automate tasks related to flow control and optimization. This article presents an overview of past history, current developments, and emerging opportunities of ML for fluid mechanics. We outline fundamental ML methodologies and discuss their uses for understanding, modeling, optimizing, and controlling fluid flows. The strengths and limitations of these methods are addressed from the perspective of scientific inquiry that considers data as an inherent part of modeling, experiments, and simulations. ML provides a powerful information-processing framework that can augment, and possibly even transform, current lines of fluid mechanics research and industrial applications."
                },
                {
                    "title": "PIFu: Pixel-Aligned Implicit Function for High-Resolution Clothed Human Digitization",
                    "abstract": "We introduce Pixel-aligned Implicit Function (PIFu), an implicit representation that locally aligns pixels of 2D images with the global context of their corresponding 3D object. Using PIFu, we propose an end-to-end deep learning method for digitizing highly detailed clothed humans that can infer both 3D surface and texture from a single image, and optionally, multiple input images. Highly intricate shapes, such as hairstyles, clothing, as well as their variations and deformations can be digitized in a unified way. Compared to existing representations used for 3D deep learning, PIFu produces high-resolution surfaces including largely unseen regions such as the back of a person. In particular, it is memory efficient unlike the voxel representation, can handle arbitrary topology, and the resulting surface is spatially aligned with the input image. Furthermore, while previous techniques are designed to process either a single image or multiple views, PIFu extends naturally to arbitrary number of views. We demonstrate high-resolution and robust reconstructions on real world images from the DeepFashion dataset, which contains a variety of challenging clothing types. Our method achieves state-of-the-art performance on a public benchmark and outperforms the prior work for clothed human digitization from a single image."
                },
                {
                    "title": "Machine learning in acoustics: Theory and applications.",
                    "abstract": "Acoustic data provide scientific and engineering insights in fields ranging from biology and communications to ocean and Earth science. We survey the recent advances and transformative potential of machine learning (ML), including deep learning, in the field of acoustics. ML is a broad family of techniques, which are often based in statistics, for automatically detecting and utilizing patterns in data. Relative to conventional acoustics and signal processing, ML is data-driven. Given sufficient training data, ML can discover complex relationships between features and desired labels or actions, or between features themselves. With large volumes of training data, ML can discover models describing complex acoustic phenomena such as human speech and reverberation. ML in acoustics is rapidly developing with compelling results and significant future promise. We first introduce ML, then highlight ML developments in four acoustics research areas: source localization in speech processing, source localization in ocean acoustics, bioacoustics, and environmental sounds in everyday scenes."
                },
                {
                    "title": "DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation",
                    "abstract": "Computer graphics, 3D computer vision and robotics communities have produced multiple approaches to representing 3D geometry for rendering and reconstruction. These provide trade-offs across fidelity, efficiency and compression capabilities. In this work, we introduce DeepSDF, a learned continuous Signed Distance Function (SDF) representation of a class of shapes that enables high quality shape representation, interpolation and completion from partial and noisy 3D input data. DeepSDF, like its classical counterpart, represents a shape's surface by a continuous volumetric field: the magnitude of a point in the field represents the distance to the surface boundary and the sign indicates whether the region is inside (-) or outside (+) of the shape, hence our representation implicitly encodes a shape's boundary as the zero-level-set of the learned function while explicitly representing the classification of space as being part of the shapes interior or not. While classical SDF's both in analytical or discretized voxel form typically represent the surface of a single shape, DeepSDF can represent an entire class of shapes. Furthermore, we show state-of-the-art performance for learned 3D shape representation and completion while reducing the model size by an order of magnitude compared with previous work."
                },
                {
                    "title": "Occupancy Networks: Learning 3D Reconstruction in Function Space",
                    "abstract": "With the advent of deep neural networks, learning-based approaches for 3D reconstruction have gained popularity. However, unlike for images, in 3D there is no canonical representation which is both computationally and memory efficient yet allows for representing high-resolution geometry of arbitrary topology. Many of the state-of-the-art learning-based 3D reconstruction approaches can hence only represent very coarse 3D geometry or are limited to a restricted domain. In this paper, we propose Occupancy Networks, a new representation for learning-based 3D reconstruction methods. Occupancy networks implicitly represent the 3D surface as the continuous decision boundary of a deep neural network classifier. In contrast to existing approaches, our representation encodes a description of the 3D output at infinite resolution without excessive memory footprint. We validate that our representation can efficiently encode 3D structure and can be inferred from various kinds of input. Our experiments demonstrate competitive results, both qualitatively and quantitatively, for the challenging tasks of 3D reconstruction from single images, noisy point clouds and coarse discrete voxel grids. We believe that occupancy networks will become a useful tool in a wide variety of learning-based 3D tasks."
                },
                {
                    "title": "Learning Implicit Fields for Generative Shape Modeling",
                    "abstract": "We advocate the use of implicit fields for learning generative models of shapes and introduce an implicit field decoder, called IM-NET, for shape generation, aimed at improving the visual quality of the generated shapes. An implicit field assigns a value to each point in 3D space, so that a shape can be extracted as an iso-surface. IM-NET is trained to perform this assignment by means of a binary classifier. Specifically, it takes a point coordinate, along with a feature vector encoding a shape, and outputs a value which indicates whether the point is outside the shape or not. By replacing conventional decoders by our implicit decoder for representation learning (via IM-AE) and shape generation (via IM-GAN), we demonstrate superior results for tasks such as generative shape modeling, interpolation, and single-view 3D reconstruction, particularly in terms of visual quality. Code and supplementary material are available at https://github.com/czq142857/implicit-decoder."
                },
                {
                    "title": "Crystal Structure Prediction via Deep Learning.",
                    "abstract": "We demonstrate the application of deep neural networks as a machine-learning tool for the analysis of a large collection of crystallographic data contained in the crystal structure repositories. Using input data in the form of multiperspective atomic fingerprints, which describe coordination topology around unique crystallographic sites, we show that the neural-network model can be trained to effectively distinguish chemical elements based on the topology of their crystallographic environment. The model also identifies structurally similar atomic sites in the entire data set of \u223c50000 crystal structures, essentially uncovering trends that reflect the periodic table of elements. The trained model was used to analyze templates derived from the known crystal structures in order to predict the likelihood of forming new compounds that could be generated by placing elements into these structural templates in a combinatorial fashion. Statistical analysis of predictive performance of the neural-network model, which was applied to a test set of structures never seen by the model during training, indicates its ability to predict known elemental compositions with a high likelihood of success. In \u223c30% of cases, the known compositions were found among the top 10 most likely candidates proposed by the model. These results suggest that the approach developed in this work can be used to effectively guide the synthetic efforts in the discovery of new materials, especially in the case of systems composed of three or more chemical elements."
                },
                {
                    "title": "Deep learning to represent subgrid processes in climate models",
                    "abstract": "Significance Current climate models are too coarse to resolve many of the atmosphere\u2019s most important processes. Traditionally, these subgrid processes are heuristically approximated in so-called parameterizations. However, imperfections in these parameterizations, especially for clouds, have impeded progress toward more accurate climate predictions for decades. Cloud-resolving models alleviate many of the gravest issues of their coarse counterparts but will remain too computationally demanding for climate change predictions for the foreseeable future. Here we use deep learning to leverage the power of short-term cloud-resolving simulations for climate modeling. Our data-driven model is fast and accurate, thereby showing the potential of machine-learning\u2013based approaches to climate model development. The representation of nonlinear subgrid processes, especially clouds, has been a major source of uncertainty in climate models for decades. Cloud-resolving models better represent many of these processes and can now be run globally but only for short-term simulations of at most a few years because of computational limitations. Here we demonstrate that deep learning can be used to capture many advantages of cloud-resolving modeling at a fraction of the computational cost. We train a deep neural network to represent all atmospheric subgrid processes in a climate model by learning from a multiscale model in which convection is treated explicitly. The trained neural network then replaces the traditional subgrid parameterizations in a global general circulation model in which it freely interacts with the resolved dynamics and the surface-flux scheme. The prognostic multiyear simulations are stable and closely reproduce not only the mean climate of the cloud-resolving simulation but also key aspects of variability, including precipitation extremes and the equatorial wave spectrum. Furthermore, the neural network approximately conserves energy despite not being explicitly instructed to. Finally, we show that the neural network parameterization generalizes to new surface forcing patterns but struggles to cope with temperatures far outside its training manifold. Our results show the feasibility of using deep learning for climate model parameterization. In a broader context, we anticipate that data-driven Earth system model development could play a key role in reducing climate prediction uncertainty in the coming decade."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "FiLM: Visual Reasoning with a General Conditioning Layer",
                    "abstract": "\n \n We introduce a general-purpose conditioning method for neural networks called FiLM: Feature-wise Linear Modulation. FiLM layers influence neural network computation via a simple, feature-wise affine transformation based on conditioning information. We show that FiLM layers are highly effective for visual reasoning - answering image-related questions which require a multi-step, high-level process - a task which has proven difficult for standard deep learning methods that do not explicitly model reasoning. Specifically, we show on visual reasoning tasks that FiLM layers 1) halve state-of-the-art error for the CLEVR benchmark, 2) modulate features in a coherent manner, 3) are robust to ablations and architectural modifications, and 4) generalize well to challenging, new data from few examples or even zero-shot.\n \n"
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Interaction Networks for Learning about Objects, Relations and Physics",
                    "abstract": "Reasoning about objects, relations, and physics is central to human intelligence, and a key goal of artificial intelligence. Here we introduce the interaction network, a model which can reason about how objects in complex systems interact, supporting dynamical predictions, as well as inferences about the abstract properties of the system. Our model takes graphs as input, performs object- and relation-centric reasoning in a way that is analogous to a simulation, and is implemented using deep neural networks. We evaluate its ability to reason about several challenging physical domains: n-body problems, rigid-body collision, and non-rigid dynamics. Our results show it can be trained to accurately simulate the physical trajectories of dozens of objects over thousands of time steps, estimate abstract quantities such as energy, and generalize automatically to systems with different numbers and configurations of objects and relations. Our interaction network implementation is the first general-purpose, learnable physics engine, and a powerful general framework for reasoning about object and relations in a wide variety of complex real-world domains."
                },
                {
                    "title": "SGDR: Stochastic Gradient Descent with Warm Restarts",
                    "abstract": "Restart techniques are common in gradient-free optimization to deal with multimodal functions. Partial warm restarts are also gaining popularity in gradient-based optimization to improve the rate of convergence in accelerated gradient schemes to deal with ill-conditioned functions. In this paper, we propose a simple warm restart technique for stochastic gradient descent to improve its anytime performance when training deep neural networks. We empirically study its performance on the CIFAR-10 and CIFAR-100 datasets, where we demonstrate new state-of-the-art results at 3.14% and 16.21%, respectively. We also demonstrate its advantages on a dataset of EEG recordings and on a downsampled version of the ImageNet dataset. Our source code is available at this https URL"
                },
                {
                    "title": "Layer Normalization",
                    "abstract": "Training state-of-the-art, deep neural networks is computationally expensive. One way to reduce the training time is to normalize the activities of the neurons. A recently introduced technique called batch normalization uses the distribution of the summed input to a neuron over a mini-batch of training cases to compute a mean and variance which are then used to normalize the summed input to that neuron on each training case. This significantly reduces the training time in feed-forward neural networks. However, the effect of batch normalization is dependent on the mini-batch size and it is not obvious how to apply it to recurrent neural networks. In this paper, we transpose batch normalization into layer normalization by computing the mean and variance used for normalization from all of the summed inputs to the neurons in a layer on a single training case. Like batch normalization, we also give each neuron its own adaptive bias and gain which are applied after the normalization but before the non-linearity. Unlike batch normalization, layer normalization performs exactly the same computation at training and test times. It is also straightforward to apply to recurrent neural networks by computing the normalization statistics separately at each time step. Layer normalization is very effective at stabilizing the hidden state dynamics in recurrent networks. Empirically, we show that layer normalization can substantially reduce the training time compared with previously published techniques."
                },
                {
                    "title": "Accelerating Eulerian Fluid Simulation With Convolutional Networks",
                    "abstract": "Efficient simulation of the Navier-Stokes equations for fluid flow is a long standing problem in applied mathematics, for which state-of-the-art methods require large compute resources. In this work, we propose a data-driven approach that leverages the approximation power of deep-learning with the precision of standard solvers to obtain fast and highly realistic simulations. Our method solves the incompressible Euler equations using the standard operator splitting method, in which a large sparse linear system with many free parameters must be solved. We use a Convolutional Network with a highly tailored architecture, trained using a novel unsupervised learning framework to solve the linear system. We present real-time 2D and 3D simulations that outperform recently proposed data-driven methods; the obtained results are realistic and show good generalization properties."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "A platform for benchmark cases in computational acoustics",
                    "abstract": "Solutions to the partial differential equations that describe acoustic problems can be found by analytical, numerical and experimental techniques. Within arbitrary domains and for arbitrary initial and boundary conditions, all solution techniques require certain assumptions and simplifications. It is difficult to estimate the precision of a solution technique. Due to the lack of a common process to quantify and report the performance of the solution technique, a variety of ways exists to discuss the results with the scientific community. Moreover, the absence of general reference results does hamper the validation of newly developed techniques. Over the years many researchers in the field of computational acoustics have therefore expressed the need and wish to have available common benchmark cases. This contribution is intended to be the start of a long term project, about deploying benchmarks in the entire field of computational acoustics. The platform is a web-based database, where cases and results can be submitted by all researchers and are openly available. Long-term maintenance of this platform is ensured. As an example of good practice, this paper presents a framework for the field of linear acoustic. Within this field, different categories are defined \u2013 as bounded or unbounded problems, scattering or radiating problems and time-domain as well as frequency-domain problems \u2013 and a structure is proposed how to describe a benchmark case. Furthermore, a way of reporting on the used solution technique and its result is suggested. Three problems have been defined that demonstrate how the benchmark cases are intended to be used."
                },
                {
                    "title": "Sinkhorn Distances: Lightspeed Computation of Optimal Transport",
                    "abstract": "Optimal transportation distances are a fundamental family of parameterized distances for histograms. Despite their appealing theoretical properties, excellent performance in retrieval tasks and intuitive formulation, their computation involves the resolution of a linear program whose cost is prohibitive whenever the histograms' dimension exceeds a few hundreds. We propose in this work a new family of optimal transportation distances that look at transportation problems from a maximum-entropy perspective. We smooth the classical optimal transportation problem with an entropic regularization term, and show that the resulting optimum is also a distance which can be computed through Sinkhorn-Knopp's matrix scaling algorithm at a speed that is several orders of magnitude faster than that of transportation solvers. We also report improved performance over classical optimal transportation distances on the MNIST benchmark problem."
                },
                {
                    "title": "Fast and robust Earth Mover's Distances",
                    "abstract": "We present a new algorithm for a robust family of Earth Mover's Distances - EMDs with thresholded ground distances. The algorithm transforms the flow-network of the EMD so that the number of edges is reduced by an order of magnitude. As a result, we compute the EMD by an order of magnitude faster than the original algorithm, which makes it possible to compute the EMD on large histograms and databases. In addition, we show that EMDs with thresholded ground distances have many desirable properties. First, they correspond to the way humans perceive distances. Second, they are robust to outlier noise and quantization effects. Third, they are metrics. Finally, experimental results on image retrieval show that thresholding the ground distance of the EMD improves both accuracy and speed."
                },
                {
                    "title": "The Finite Element Method: Its Basis and Fundamentals",
                    "abstract": "The Finite Element Method: Its Basis and Fundamentals offers a complete introduction to the basis of the finite element method, covering fundamental theory and worked examples in the detail required for readers to apply the knowledge to their own engineering problems and understand more advanced applications. This edition sees a significant rearrangement of the book's content to enable clearer development of the finite element method, with major new chapters and sections added to cover: * Weak forms * Variational forms * Multi-dimensional field problems * Automatic mesh generation * Plate bending and shells * Developments in meshless techniques Focusing on the core knowledge, mathematical and analytical tools needed for successful application, The Finite Element Method: Its Basis and Fundamentals is the authoritative resource of choice for graduate level students, researchers and professional engineers involved in finite element-based engineering analysis. * A proven keystone reference in the library of any engineer needing to understand and apply the finite element method in design and development. * Founded by an influential pioneer in the field and updated in this seventh edition by an author team incorporating academic authority and industrial simulation experience. * Features reworked and reordered contents for clearer development of the theory, plus new chapters and sections on mesh generation, plate bending, shells, weak forms and variational forms."
                },
                {
                    "title": "The Finite Element Method",
                    "abstract": "A short tutorial giving an overview of the history, theory and application of the finite element method."
                },
                {
                    "title": "Thin Plates and Shells: Theory: Analysis, and Applications",
                    "abstract": "Part 1 Thin plates: introduction the fundamentals of the small-deflection plate bending theory rectangular plates circular plates bending of plates of various shapes plate bending by approximate and numerical methods advanced topics buckling of plates vibration of plates. Part 2 Thin shells: introduction to the general linear shell theory geometry of the middle surface the general linear theory of thin shells the membrane theory of shells applications of the membrane theory to the analysis of shell structures moment theory of circular cylindrical shells the moment theory of shells of revolution approximate theories of shell analysis and their application advanced topics buckling of shells vibration of shells. Appendices: some reference data Fourier series expansion verification of relations of the theory of surfaces derivation of the strain-displacement relations verification of equilibrium equations."
                },
                {
                    "title": "Principles of Vibration and Sound",
                    "abstract": "1 Free and Forced Vibrations of Simple Systems.- 2 Continuous Systems in One Dimension: Strings and Bars.- 3 Two-Dimensional Systems: Membranes and Plates.- 4 Coupled Vibrating Systems.- 5 Nonlinear Systems.- 6 Sound Waves in Air.- 7 Sound Radiation.- 8 Pipes and Horns.- 9 Acoustic Systems.- 10 Microphones and Other Transducers.- 11 Sound in Concert Halls and Studios.- 12 Sound and Noise Outdoors.- 13 Underwater Sound.- Selected Bibliography.- Problems.- Answers to Selected Problems.- Name Index."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we accurately predict the vibration patterns of mechanical structures based on their geometries to optimize design modifications for noise reduction?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the intersection of mechanical engineering and machine learning, potentially leading to innovative design methodologies that enhance noise reduction in various applications, such as automotive and aerospace industries. By accurately predicting vibration patterns, researchers can develop more effective design strategies that minimize noise pollution, improve user comfort, and enhance the longevity of mechanical systems. This work could pave the way for future research into advanced materials and structures that inherently reduce vibrations, leading to practical applications in consumer products and industrial machinery.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complex relationship between geometry and vibrational behavior, which is influenced by numerous factors such as material properties, boundary conditions, and the nature of the excitation forces. Naive approaches may fail because they often overlook the intricate interactions between these variables, leading to inaccurate predictions. Additionally, the computational intensity of simulating vibration patterns using methods like the finite element method (FEM) poses significant technical obstacles, requiring substantial computational resources and expertise in numerical methods. The need for high fidelity in capturing the nuances of vibration patterns adds to the complexity of the problem.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either the theoretical aspects of vibration analysis or empirical approaches without integrating machine learning techniques effectively. Limitations in computational power and the lack of comprehensive datasets have hindered the development of robust predictive models. Additionally, existing solutions may not adequately account for the variability in geometries and material properties, leading to generalized models that fail in specific applications. Our approach differs by leveraging a newly introduced dataset of 12,000 samples and employing diverse machine learning architectures to create a more accurate and adaptable predictive framework for vibration patterns.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using a dataset of 12,000 samples generated from numerical simulations of vibrating plates under harmonic excitation. We will evaluate various machine learning architectures, including neural networks, to model the relationship between plate geometries and their corresponding vibration patterns. The performance of these models will be assessed using metrics such as mean squared error (MSE) and R-squared values to ensure accuracy in predictions. We expect"
            }
        },
        "author_data": {
            "25214017-6763-4e49-a6ee-1492c6d3b3b5": {
                "pk": "25214017-6763-4e49-a6ee-1492c6d3b3b5",
                "name": "Jan van Delden",
                "collaborators": [
                    "Jonathan Henrich",
                    "Dominik Seidel",
                    "Thomas Kneib",
                    "Alexander Ecker"
                ],
                "domain": [
                    "Deep Learning",
                    "Point Cloud Segmentation",
                    "Forest Management",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Towards general deep-learning-based tree instance segmentation models",
                        "abstract": "The segmentation of individual trees from forest point clouds is a crucial task for downstream analyses such as carbon sequestration estimation. Recently, deep-learning-based methods have been proposed which show the potential of learning to segment trees. Since these methods are trained in a supervised way, the question arises how general models can be obtained that are applicable across a wide range of settings. So far, training has been mainly conducted with data from one specific laser scanning type and for specific types of forests. In this work, we train one segmentation model under various conditions, using seven diverse datasets found in literature, to gain insights into the generalization capabilities under domain-shift. Our results suggest that a generalization from coniferous dominated sparse point clouds to deciduous dominated high-resolution point clouds is possible. Conversely, qualitative evidence suggests that generalization from high-resolution to low-resolution point clouds is challenging. This emphasizes the need for forest point clouds with diverse data characteristics for model development. To enrich the available data basis, labeled trees from two previous works were propagated to the complete forest point cloud and are made publicly available at https://doi.org/10.25625/QUTUWU."
                    },
                    {
                        "title": "TreeLearn: A Comprehensive Deep Learning Method for Segmenting Individual Trees from Ground-Based LiDAR Forest Point Clouds",
                        "abstract": "Laser-scanned point clouds of forests make it possible to extract valuable information for forest management. To consider single trees, a forest point cloud needs to be segmented into individual tree point clouds. Existing segmentation methods are usually based on hand-crafted algorithms, such as identifying trunks and growing trees from them, and face difficulties in dense forests with overlapping tree crowns. In this study, we propose TreeLearn, a deep learning-based approach for tree instance segmentation of forest point clouds. Unlike previous methods, TreeLearn is trained on already segmented point clouds in a data-driven manner, making it less reliant on predefined features and algorithms. Furthermore, TreeLearn is implemented as a fully automatic pipeline and does not rely on extensive hyperparameter tuning, which makes it easy to use. Additionally, we introduce a new manually segmented benchmark forest dataset containing 156 full trees, and 79 partial trees, that have been cleanly segmented by hand. The data is generated by mobile laser scanning and contributes to create a larger and more diverse data basis for model development and fine-grained instance segmentation evaluation. We trained TreeLearn on forest point clouds of 6665 trees, labeled using the Lidar360 software. An evaluation on the benchmark dataset shows that TreeLearn performs equally well or better than the algorithm used to generate its training data. Furthermore, the method's performance can be vastly improved by fine-tuning on the cleanly labeled benchmark dataset. The TreeLearn code is available from https://github.com/ecker-lab/TreeLearn. The data as well as trained models can be found at https://doi.org/10.25625/VPMPID."
                    }
                ]
            },
            "7049cf90-6057-449d-838d-ccb9b449fc8e": {
                "pk": "7049cf90-6057-449d-838d-ccb9b449fc8e",
                "name": "Julius Schultz",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "f0728f72-f2bd-4592-890a-12b150911207": {
                "pk": "f0728f72-f2bd-4592-890a-12b150911207",
                "name": "Christopher Blech",
                "collaborators": [
                    "Harikrishnan K. Sreekumar",
                    "Yannik H\u00fcpel",
                    "Sabine C. Langer"
                ],
                "domain": [
                    "Computational Mechanics",
                    "Finite Element Analysis",
                    "Vibroacoustics",
                    "Model Order Reduction"
                ],
                "publications": [
                    {
                        "title": "Efficient solution strategies for cabin noise assessment of a wave resolving aircraft fuselage model",
                        "abstract": "For the purpose of high-fidelity aircraft cabin noise simulations during early design phases, we study three efficient solving approaches for the fully coupled finite element model of an aircraft fuselage segment. Obtaining an efficient solution with respect to consumed computational time and resources is challenging within a conventional simulation pipeline, as large-scale and complex vibroacoustic models demand crucially high computational costs with increasing frequency. In this contribution, we adopt (1) frequency and domain-adaptive discretisation, (2) domain-decomposition techniques, and (3) model order reduction with rational Arnoldi Krylov subspace methods for an aircraft fuselage model. The three approaches have shown remarkable advantage thereby reducing the solving time as well as the memory requirement that are essential when solving large-scale models. While the discretisation and the model order reduction approaches accelerate the solving process by efficiently handling the complexity of the system to be solved, domain-decomposition techniques further handle the aspect of reducing the overall memory consumption. Finally with the help of active research aircraft models, we implement and showcase the achieved efficiency."
                    }
                ]
            },
            "5fab1d65-e15c-47f0-88fa-953f4214a04e": {
                "pk": "5fab1d65-e15c-47f0-88fa-953f4214a04e",
                "name": "Sabine C. Langer",
                "collaborators": [
                    "Christopher Blech",
                    "Harikrishnan K. Sreekumar",
                    "Yannik H\u00fcpel"
                ],
                "domain": [
                    "Computational Mechanics",
                    "Finite Element Analysis",
                    "Vibroacoustics",
                    "Model Order Reduction"
                ],
                "publications": [
                    {
                        "title": "Efficient solution strategies for cabin noise assessment of a wave resolving aircraft fuselage model",
                        "abstract": "For the purpose of high-fidelity aircraft cabin noise simulations during early design phases, we study three efficient solving approaches for the fully coupled finite element model of an aircraft fuselage segment. Obtaining an efficient solution with respect to consumed computational time and resources is challenging within a conventional simulation pipeline, as large-scale and complex vibroacoustic models demand crucially high computational costs with increasing frequency. In this contribution, we adopt (1) frequency and domain-adaptive discretisation, (2) domain-decomposition techniques, and (3) model order reduction with rational Arnoldi Krylov subspace methods for an aircraft fuselage model. The three approaches have shown remarkable advantage thereby reducing the solving time as well as the memory requirement that are essential when solving large-scale models. While the discretisation and the model order reduction approaches accelerate the solving process by efficiently handling the complexity of the system to be solved, domain-decomposition techniques further handle the aspect of reducing the overall memory consumption. Finally with the help of active research aircraft models, we implement and showcase the achieved efficiency."
                    }
                ]
            },
            "69063750-e534-4adb-aff9-520e9c69d1be": {
                "pk": "69063750-e534-4adb-aff9-520e9c69d1be",
                "name": "Timo L\u00fcddecke",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2406.08747": {
        "paper_data": {
            "title": "StreamBench: Towards Benchmarking Continuous Improvement of Language Agents",
            "url": "http://arxiv.org/abs/2406.08747v1",
            "arxiv_id": "2406.08747",
            "authors": [
                "Cheng-Kuang Wu",
                "Zhi Rui Tam",
                "Chieh-Yen Lin",
                "Yun-Nung Chen",
                "Hung-yi Lee"
            ],
            "abstract": "Recent works have shown that large language model (LLM) agents are able to improve themselves from experience, which is an important ability for continuous enhancement post-deployment. However, existing benchmarks primarily evaluate their innate capabilities and do not assess their ability to improve over time. To address this gap, we introduce StreamBench, a pioneering benchmark designed to evaluate the continuous improvement of LLM agents over an input-feedback sequence. StreamBench simulates an online learning environment where LLMs receive a continuous flow of feedback stream and iteratively enhance their performance. In addition, we propose several simple yet effective baselines for improving LLMs on StreamBench, and provide a comprehensive analysis to identify critical components that contribute to successful streaming strategies. Our work serves as a stepping stone towards developing effective online learning strategies for LLMs, paving the way for more adaptive AI systems in streaming scenarios.",
            "introduction": "   1 Introduction  Recently, large-scale pretraining\u00a0[1] and instruction fine-tuning\u00a0[2] have driven paradigm shifts in how we interact with language models. These advancements allow us to use them out-of-the-box to solve problems. Consequently, many benchmarks have emerged to evaluate the general capabilities of these models. Some notable examples include MMLU\u00a0[3], GSM8K\u00a0[4], and BIG-Bench-Hard\u00a0[5]. All these benchmarks aim to assess LLMs\u2019 innate capabilities, which we define as the general knowledge or reasoning abilities demonstrated when used out-of-the-box.   In addition to LLMs\u2019 strong innate capabilities, recent works have shown that LLM agents, which are LLMs augmented with extra components such as memory, retrievers, or tools, are able to improve themselves from experience. MemPrompt\u00a0[6] shows that memory-enhanced GPT-3 can improve through time by storing past user feedback and retrieve them in the future. Reflexion\u00a0[7] demonstrates that LLM agents can perform better in future trials by running repeated trials on the same dataset via self-reflection. ExpeL\u00a0[8] further shows that LLM agents can learn from cross-task experience and improve performance without executing repeated trials on the target task.   Given LLM agents\u2019 self-improvement abilities, there remains a missing piece in the current evaluation landscape. Beyond measuring LLMs\u2019 innate capabilities with aforementioned offline benchmarks\u00a0[3, 4, 5], it is important to assess their capacity to improve over time since we would like our systems to gradually improve after deployment. This gap motivated us to develop a new evaluation scenario\u2013an online setting to measure LLM agents\u2019 ability to continuously enhance their performance over time.    This online setting focuses on scenarios where LLM agents attempt to solve a specific downstream task and improve themselves from an input-feedback sequence, with the goal to maximize the accuracy for the whole sequence of the agent\u2019s predictions.   Given these rationales, we introduce StreamBench, a benchmark designed to evaluate LLM agents\u2019 ability to improve themselves over an input-feedback sequence. StreamBench simulates an environment where LLM agents are exposed to a sequence of users\u2019 natural language requirements and feedback. To the best of our knowledge, StreamBench is the first benchmark to evaluate LLM agents in streaming scenarios with a diverse range of tasks. StreamBench aims to inspire further efforts to develop more adaptive LLM agents, thereby enhancing their practical effectiveness. Our contributions can be summarized as follows:   \u2022  We introduce StreamBench, the first benchmark designed to evaluate LLM agents\u2019 ability to improve over an input-feedback sequence in an online setting across a wide range of tasks.    \u2022  We propose several simple yet effective baselines for enhancing LLM agents\u2019 performance in streaming scenarios, including a cost-effective multi-agent method that outperforms other baselines while maintaining the average cost of a single agent.    \u2022  We conduct analysis on the advantages and potential pitfalls of the proposed methods, providing insights into effective streaming strategies of LLMs.      Figure 1: (Left) A schematic diagram showing the online evaluation setting of StreamBench, where agents update their components (p,r,\u2133\ud835\udc5d\ud835\udc5f\u2133p,r,\\mathcal{M}italic_p , italic_r , caligraphic_M, or \u03b8\ud835\udf03\\thetaitalic_\u03b8) from an input-feedback sequence to achieve the highest final accuracy (refer to Section\u00a03.1 for details). (Right) Performance curve on the DDXPlus dataset on StreamBench. Agents are able to gradually improve with our proposed streaming baselines.     2 Formulation  Consider a streaming scenario involving an agent, an external environment, and a sequence of inputs:   Agent. We define an agent as an LLM",
            "references": [
                {
                    "title": "Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context",
                    "abstract": "In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content."
                },
                {
                    "title": "ReConcile: Round-Table Conference Improves Reasoning via Consensus among Diverse LLMs",
                    "abstract": "Large Language Models (LLMs) still struggle with natural language reasoning tasks. Motivated by the society of minds (Minsky, 1988), we propose ReConcile, a multi-model multi-agent framework designed as a round table conference among diverse LLM agents. ReConcile enhances collaborative reasoning between LLM agents via multiple rounds of discussion, learning to convince other agents to improve their answers, and employing a confidence-weighted voting mechanism that leads to a better consensus. In each round, ReConcile initiates discussion between agents via a 'discussion prompt' that consists of (a) grouped answers and explanations generated by each agent in the previous round, (b) their confidence scores, and (c) demonstrations of answer-rectifying human explanations, used for convincing other agents. Experiments on seven benchmarks demonstrate that ReConcile significantly improves LLMs' reasoning -- both individually and as a team -- surpassing prior single-agent and multi-agent baselines by up to 11.4% and even outperforming GPT-4 on three datasets. ReConcile also flexibly incorporates different combinations of agents, including API-based, open-source, and domain-specific models, leading to an 8% improvement on MATH. Finally, we analyze the individual components of ReConcile, demonstrating that the diversity originating from different models is critical to its superior performance. Code: https://github.com/dinobby/ReConcile"
                },
                {
                    "title": "Connecting Large Language Models with Evolutionary Algorithms Yields Powerful Prompt Optimizers",
                    "abstract": "Large Language Models (LLMs) excel in various tasks, but they rely on carefully crafted prompts that often demand substantial human effort. To automate this process, in this paper, we propose a novel framework for discrete prompt optimization, called EvoPrompt, which borrows the idea of evolutionary algorithms (EAs) as they exhibit good performance and fast convergence. To enable EAs to work on discrete prompts, which are natural language expressions that need to be coherent and human-readable, we connect LLMs with EAs. This approach allows us to simultaneously leverage the powerful language processing capabilities of LLMs and the efficient optimization performance of EAs. Specifically, abstaining from any gradients or parameters, EvoPrompt starts from a population of prompts and iteratively generates new prompts with LLMs based on the evolutionary operators, improving the population based on the development set. We optimize prompts for both closed- and open-source LLMs including GPT-3.5 and Alpaca, on 31 datasets covering language understanding, generation tasks, as well as BIG-Bench Hard (BBH) tasks. EvoPrompt significantly outperforms human-engineered prompts and existing methods for automatic prompt generation (e.g., up to 25% on BBH). Furthermore, EvoPrompt demonstrates that connecting LLMs with EAs creates synergies, which could inspire further research on the combination of LLMs and conventional algorithms."
                },
                {
                    "title": "ExpeL: LLM Agents Are Experiential Learners",
                    "abstract": "The recent surge in research interest in applying large language models (LLMs) to decision-making tasks has flourished by leveraging the extensive world knowledge embedded in LLMs. While there is a growing demand to tailor LLMs for custom decision-making tasks, finetuning them for specific tasks is resource-intensive and may diminish the model's generalization capabilities. Moreover, state-of-the-art language models like GPT-4 and Claude are primarily accessible through API calls, with their parametric weights remaining proprietary and unavailable to the public. This scenario emphasizes the growing need for new methodologies that allow learning from agent experiences without requiring parametric updates. To address these problems, we introduce the Experiential Learning (ExpeL) agent. Our agent autonomously gathers experiences and extracts knowledge using natural language from a collection of training tasks. At inference, the agent recalls its extracted insights and past experiences to make informed decisions. Our empirical results highlight the robust learning efficacy of the ExpeL agent, indicating a consistent enhancement in its performance as it accumulates experiences. We further explore the emerging capabilities and transfer learning potential of the ExpeL agent through qualitative observations and additional experiments."
                },
                {
                    "title": "Self-ICL: Zero-Shot In-Context Learning with Self-Generated Demonstrations",
                    "abstract": "Large language models (LLMs) have exhibited striking in-context learning (ICL) ability to adapt to target tasks with a few input-output demonstrations. For better ICL, different methods are proposed to select representative demonstrations from existing training corpora. However, such settings are not aligned with real-world practices, as end-users usually query LMs without access to demonstration pools. In this work, we introduce Self-ICL -- a simple framework which bootstraps LMs' intrinsic capabilities to perform zero-shot ICL. Given a test input, Self-ICL first prompts the model to generate pseudo-inputs. Next, the model predicts pseudo-labels for the pseudo-inputs via zero-shot prompting. Finally, we perform ICL for the test input with the pseudo-input-label pairs as demonstrations. Evaluation on 23 BIG-Bench Hard tasks shows Self-ICL outperforms zero-shot baselines on both average accuracy and head-to-head comparison. Moreover, with zero-shot chain-of-thought, Self-ICL achieves results comparable to using real demonstrations. Additionally, we conduct a range of analyses to validate Self-ICL's effectiveness and provide insights for its behaviors under different settings."
                },
                {
                    "title": "Meta-Learning Online Adaptation of Language Models",
                    "abstract": "Large language models encode impressively broad world knowledge in their parameters. However, the knowledge in static language models falls out of date, limiting the model's effective\"shelf life.\"While online fine-tuning can reduce this degradation, we find that naively fine-tuning on a stream of documents leads to a low level of information uptake. We hypothesize that online fine-tuning does not sufficiently attend to important information. That is, the gradient signal from important tokens representing factual information is drowned out by the gradient from inherently noisy tokens, suggesting that a dynamic, context-aware learning rate may be beneficial. We therefore propose learning which tokens to upweight. We meta-train a small, autoregressive model to reweight the language modeling loss for each token during online fine-tuning, with the objective of maximizing the out-of-date base question-answering model's ability to answer questions about a document after a single weighted gradient step. We call this approach Context-aware Meta-learned Loss Scaling (CaMeLS). Across three different distributions of documents, our experiments find that CaMeLS provides substantially improved information uptake on streams of thousands of documents compared with standard fine-tuning and baseline heuristics for reweighting token losses."
                },
                {
                    "title": "Improving Factuality and Reasoning in Language Models through Multiagent Debate",
                    "abstract": "Large language models (LLMs) have demonstrated remarkable capabilities in language generation, understanding, and few-shot learning in recent years. An extensive body of work has explored how their performance may be further improved through the tools of prompting, ranging from verification, self-consistency, or intermediate scratchpads. In this paper, we present a complementary approach to improve language responses where multiple language model instances propose and debate their individual responses and reasoning processes over multiple rounds to arrive at a common final answer. Our findings indicate that this approach significantly enhances mathematical and strategic reasoning across a number of tasks. We also demonstrate that our approach improves the factual validity of generated content, reducing fallacious answers and hallucinations that contemporary models are prone to. Our approach may be directly applied to existing black-box models and uses identical procedure and prompts for all tasks we investigate. Overall, our findings suggest that such\"society of minds\"approach has the potential to significantly advance the capabilities of LLMs and pave the way for further breakthroughs in language generation and understanding."
                },
                {
                    "title": "MoT: Memory-of-Thought Enables ChatGPT to Self-Improve",
                    "abstract": "Large Language Models (LLMs) have shown impressive abilities in various tasks. However, fundamentally improving them depends on high-quality datasets or computationally expensive fine-tuning. On the contrary, humans can easily improve themselves by self-thinking and memory, without external resources. In this paper, we propose a framework, MoT, to let the LLM self-improve through Memory-of-Thought, without annotated datasets and parameter updates. Specifically, MoT is divided into two stages: 1. before the test stage, the LLM pre-thinks on the unlabeled dataset and saves the high-confidence thoughts as external memory; 2. During the test stage, given a test question, the LLM recalls relevant memory to help itself reason and answer it. Experimental results show that MoT can help ChatGPT significantly improve its abilities in arithmetic reasoning, commonsense reasoning, factual reasoning, and natural language inference. Further analyses show that each component contributes critically to the improvements and MoT can lead to consistent improvements across various CoT methods and LLMs."
                },
                {
                    "title": "Unified Demonstration Retriever for In-Context Learning",
                    "abstract": "In-context learning is a new learning paradigm where a language model conditions on a few input-output pairs (demonstrations) and a test input, and directly outputs the prediction. It has been shown sensitive to the provided demonstrations and thus promotes the research of demonstration retrieval: given a test input, relevant examples are retrieved from the training set to serve as informative demonstrations for in-context learning. While previous works train task-specific retrievers for several tasks separately, these methods are hard to transfer and scale on various tasks, and separately trained retrievers will cause a lot of parameter storage and deployment cost. In this paper, we propose Unified Demonstration Retriever (UDR), a single model to retrieve demonstrations for a wide range of tasks. To train UDR, we cast various tasks\u2019 training signals into a unified list-wise ranking formulation by language model\u2019s feedback. Then we propose a multi-task list-wise ranking training framework with an iterative mining strategy to find high-quality candidates, which can help UDR fully incorporate various tasks\u2019 signals. Experiments on 30+ tasks across 13 task families and multiple data domains show that UDR significantly outperforms baselines. Further analyses show the effectiveness of each proposed component and UDR\u2019s strong ability in various scenarios including different LMs (1.3B 175B), unseen datasets, varying demonstration quantities, etc. We will release the code and model checkpoint after review."
                },
                {
                    "title": "Can LLM Already Serve as A Database Interface? A BIg Bench for Large-Scale Database Grounded Text-to-SQLs",
                    "abstract": "Text-to-SQL parsing, which aims at converting natural language instructions into executable SQLs, has gained increasing attention in recent years. In particular, Codex and ChatGPT have shown impressive results in this task. However, most of the prevalent benchmarks, i.e., Spider, and WikiSQL, focus on database schema with few rows of database contents leaving the gap between academic study and real-world applications. To mitigate this gap, we present Bird, a big benchmark for large-scale database grounded in text-to-SQL tasks, containing 12,751 pairs of text-to-SQL data and 95 databases with a total size of 33.4 GB, spanning 37 professional domains. Our emphasis on database values highlights the new challenges of dirty database contents, external knowledge between NL questions and database contents, and SQL efficiency, particularly in the context of massive databases. To solve these problems, text-to-SQL models must feature database value comprehension in addition to semantic parsing. The experimental results demonstrate the significance of database values in generating accurate text-to-SQLs for big databases. Furthermore, even the most effective text-to-SQL models, i.e. ChatGPT, only achieves 40.08% in execution accuracy, which is still far from the human result of 92.96%, proving that challenges still stand. Besides, we also provide an efficiency analysis to offer insights into generating text-to-efficient-SQLs that are beneficial to industries. We believe that BIRD will contribute to advancing real-world applications of text-to-SQL research. The leaderboard and source code are available: https://bird-bench.github.io/."
                },
                {
                    "title": "Automatic Prompt Optimization with \"Gradient Descent\" and Beam Search",
                    "abstract": "Large Language Models (LLMs) have shown impressive performance as general purpose agents, but their abilities remain highly dependent on prompts which are hand written with onerous trial-and-error effort. We propose a simple and nonparametric solution to this problem, Automatic Prompt Optimization (APO), which is inspired by numerical gradient descent to automatically improve prompts, assuming access to training data and an LLM API. The algorithm uses minibatches of data to form natural language\"gradients\"that criticize the current prompt. The gradients are then\"propagated\"into the prompt by editing the prompt in the opposite semantic direction of the gradient. These gradient descent steps are guided by a beam search and bandit selection procedure which significantly improves algorithmic efficiency. Preliminary results across three benchmark NLP tasks and the novel problem of LLM jailbreak detection suggest that Automatic Prompt Optimization can outperform prior prompt editing techniques and improve an initial prompt's performance by up to 31%, by using data to rewrite vague task descriptions into more precise annotation instructions."
                },
                {
                    "title": "Self-Refine: Iterative Refinement with Self-Feedback",
                    "abstract": "Like humans, large language models (LLMs) do not always generate the best output on their first try. Motivated by how humans refine their written text, we introduce Self-Refine, an approach for improving initial outputs from LLMs through iterative feedback and refinement. The main idea is to generate an initial output using an LLMs; then, the same LLMs provides feedback for its output and uses it to refine itself, iteratively. Self-Refine does not require any supervised training data, additional training, or reinforcement learning, and instead uses a single LLM as the generator, refiner, and feedback provider. We evaluate Self-Refine across 7 diverse tasks, ranging from dialog response generation to mathematical reasoning, using state-of-the-art (GPT-3.5, ChatGPT, and GPT-4) LLMs. Across all evaluated tasks, outputs generated with Self-Refine are preferred by humans and automatic metrics over those generated with the same LLM using conventional one-step generation, improving by ~20% absolute on average in task performance. Our work demonstrates that even state-of-the-art LLMs like GPT-4 can be further improved at test time using our simple, standalone approach."
                },
                {
                    "title": "Reflexion: language agents with verbal reinforcement learning",
                    "abstract": "Large language models (LLMs) have been increasingly used to interact with external environments (e.g., games, compilers, APIs) as goal-driven agents. However, it remains challenging for these language agents to quickly and efficiently learn from trial-and-error as traditional reinforcement learning methods require extensive training samples and expensive model fine-tuning. We propose Reflexion, a novel framework to reinforce language agents not by updating weights, but instead through linguistic feedback. Concretely, Reflexion agents verbally reflect on task feedback signals, then maintain their own reflective text in an episodic memory buffer to induce better decision-making in subsequent trials. Reflexion is flexible enough to incorporate various types (scalar values or free-form language) and sources (external or internally simulated) of feedback signals, and obtains significant improvements over a baseline agent across diverse tasks (sequential decision-making, coding, language reasoning). For example, Reflexion achieves a 91% pass@1 accuracy on the HumanEval coding benchmark, surpassing the previous state-of-the-art GPT-4 that achieves 80%. We also conduct ablation and analysis studies using different feedback signals, feedback incorporation methods, and agent types, and provide insights into how they affect performance."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Larger language models do in-context learning differently",
                    "abstract": "We study how in-context learning (ICL) in language models is affected by semantic priors versus input-label mappings. We investigate two setups-ICL with flipped labels and ICL with semantically-unrelated labels-across various model families (GPT-3, InstructGPT, Codex, PaLM, and Flan-PaLM). First, experiments on ICL with flipped labels show that overriding semantic priors is an emergent ability of model scale. While small language models ignore flipped labels presented in-context and thus rely primarily on semantic priors from pretraining, large models can override semantic priors when presented with in-context exemplars that contradict priors, despite the stronger semantic priors that larger models may hold. We next study semantically-unrelated label ICL (SUL-ICL), in which labels are semantically unrelated to their inputs (e.g., foo/bar instead of negative/positive), thereby forcing language models to learn the input-label mappings shown in in-context exemplars in order to perform the task. The ability to do SUL-ICL also emerges primarily with scale, and large-enough language models can even perform linear classification in a SUL-ICL setting. Finally, we evaluate instruction-tuned models and find that instruction tuning strengthens both the use of semantic priors and the capacity to learn input-label mappings, but more of the former."
                },
                {
                    "title": "DS-1000: A Natural and Reliable Benchmark for Data Science Code Generation",
                    "abstract": "We introduce DS-1000, a code generation benchmark with a thousand data science problems spanning seven Python libraries, such as NumPy and Pandas. Compared to prior works, DS-1000 incorporates three core features. First, our problems reflect diverse, realistic, and practical use cases since we collected them from StackOverflow. Second, our automatic evaluation is highly specific (reliable) -- across all Codex-002-predicted solutions that our evaluation accept, only 1.8% of them are incorrect; we achieve this with multi-criteria metrics, checking both functional correctness by running test cases and surface-form constraints by restricting API usages or keywords. Finally, we proactively defend against memorization by slightly modifying our problems to be different from the original StackOverflow source; consequently, models cannot answer them correctly by memorizing the solutions from pre-training. The current best public system (Codex-002) achieves 43.3% accuracy, leaving ample room for improvement. We release our benchmark at https://ds1000-code-gen.github.io."
                },
                {
                    "title": "Large Language Models Are Human-Level Prompt Engineers",
                    "abstract": "By conditioning on natural language instructions, large language models (LLMs) have displayed impressive capabilities as general-purpose computers. However, task performance depends significantly on the quality of the prompt used to steer the model, and most effective prompts have been handcrafted by humans. Inspired by classical program synthesis and the human approach to prompt engineering, we propose Automatic Prompt Engineer (APE) for automatic instruction generation and selection. In our method, we treat the instruction as the\"program,\"optimized by searching over a pool of instruction candidates proposed by an LLM in order to maximize a chosen score function. To evaluate the quality of the selected instruction, we evaluate the zero-shot performance of another LLM following the selected instruction. Experiments on 24 NLP tasks show that our automatically generated instructions outperform the prior LLM baseline by a large margin and achieve better or comparable performance to the instructions generated by human annotators on 19/24 tasks. We conduct extensive qualitative and quantitative analyses to explore the performance of APE. We show that APE-engineered prompts can be applied to steer models toward truthfulness and/or informativeness, as well as to improve few-shot learning performance by simply prepending them to standard in-context learning prompts. Please check out our webpage at https://sites.google.com/view/automatic-prompt-engineer."
                },
                {
                    "title": "Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them",
                    "abstract": "BIG-Bench (Srivastava et al., 2022) is a diverse evaluation suite that focuses on tasks believed to be beyond the capabilities of current language models. Language models have already made good progress on this benchmark, with the best model in the BIG-Bench paper outperforming average reported human-rater results on 65% of the BIG-Bench tasks via few-shot prompting. But on what tasks do language models fall short of average human-rater performance, and are those tasks actually unsolvable by current language models? In this work, we focus on a suite of 23 challenging BIG-Bench tasks which we call BIG-Bench Hard (BBH). These are the task for which prior language model evaluations did not outperform the average human-rater. We find that applying chain-of-thought (CoT) prompting to BBH tasks enables PaLM to surpass the average human-rater performance on 10 of the 23 tasks, and Codex (code-davinci-002) to surpass the average human-rater performance on 17 of the 23 tasks. Since many tasks in BBH require multi-step reasoning, few-shot prompting without CoT, as done in the BIG-Bench evaluations (Srivastava et al., 2022), substantially underestimates the best performance and capabilities of language models, which is better captured via CoT prompting. As further analysis, we explore the interaction between CoT and model scale on BBH, finding that CoT enables emergent task performance on several BBH tasks with otherwise flat scaling curves."
                },
                {
                    "title": "ReAct: Synergizing Reasoning and Acting in Language Models",
                    "abstract": "While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources, such as knowledge bases or environments, to gather additional information. We apply our approach, named ReAct, to a diverse set of language and decision making tasks and demonstrate its effectiveness over state-of-the-art baselines, as well as improved human interpretability and trustworthiness over methods without reasoning or acting components. Concretely, on question answering (HotpotQA) and fact verification (Fever), ReAct overcomes issues of hallucination and error propagation prevalent in chain-of-thought reasoning by interacting with a simple Wikipedia API, and generates human-like task-solving trajectories that are more interpretable than baselines without reasoning traces. On two interactive decision making benchmarks (ALFWorld and WebShop), ReAct outperforms imitation and reinforcement learning methods by an absolute success rate of 34% and 10% respectively, while being prompted with only one or two in-context examples. Project site with code: https://react-lm.github.io"
                },
                {
                    "title": "Large Language Models are Zero-Shot Reasoners",
                    "abstract": "Pretrained large language models (LLMs) are widely used in many sub-fields of natural language processing (NLP) and generally known as excellent few-shot learners with task-specific exemplars. Notably, chain of thought (CoT) prompting, a recent technique for eliciting complex multi-step reasoning through step-by-step answer examples, achieved the state-of-the-art performances in arithmetics and symbolic reasoning, difficult system-2 tasks that do not follow the standard scaling laws for LLMs. While these successes are often attributed to LLMs' ability for few-shot learning, we show that LLMs are decent zero-shot reasoners by simply adding\"Let's think step by step\"before each answer. Experimental results demonstrate that our Zero-shot-CoT, using the same single prompt template, significantly outperforms zero-shot LLM performances on diverse benchmark reasoning tasks including arithmetics (MultiArith, GSM8K, AQUA-RAT, SVAMP), symbolic reasoning (Last Letter, Coin Flip), and other logical reasoning tasks (Date Understanding, Tracking Shuffled Objects), without any hand-crafted few-shot examples, e.g. increasing the accuracy on MultiArith from 17.7% to 78.7% and GSM8K from 10.4% to 40.7% with large InstructGPT model (text-davinci-002), as well as similar magnitudes of improvements with another off-the-shelf large model, 540B parameter PaLM. The versatility of this single prompt across very diverse reasoning tasks hints at untapped and understudied fundamental zero-shot capabilities of LLMs, suggesting high-level, multi-task broad cognitive capabilities may be extracted by simple prompting. We hope our work not only serves as the minimal strongest zero-shot baseline for the challenging reasoning benchmarks, but also highlights the importance of carefully exploring and analyzing the enormous zero-shot knowledge hidden inside LLMs before crafting finetuning datasets or few-shot exemplars."
                },
                {
                    "title": "DDXPlus: A New Dataset For Automatic Medical Diagnosis",
                    "abstract": "There has been a rapidly growing interest in Automatic Symptom Detection (ASD) and Automatic Diagnosis (AD) systems in the machine learning research literature, aiming to assist doctors in telemedicine services. These systems are designed to interact with patients, collect evidence about their symptoms and relevant antecedents, and possibly make predictions about the underlying diseases. Doctors would review the interactions, including the evidence and the predictions, collect if necessary additional information from patients, before deciding on next steps. Despite recent progress in this area, an important piece of doctors' interactions with patients is missing in the design of these systems, namely the differential diagnosis. Its absence is largely due to the lack of datasets that include such information for models to train on. In this work, we present a large-scale synthetic dataset of roughly 1.3 million patients that includes a differential diagnosis, along with the ground truth pathology, symptoms and antecedents for each patient. Unlike existing datasets which only contain binary symptoms and antecedents, this dataset also contains categorical and multi-choice symptoms and antecedents useful for efficient data collection. Moreover, some symptoms are organized in a hierarchy, making it possible to design systems able to interact with patients in a logical way. As a proof-of-concept, we extend two existing AD and ASD systems to incorporate the differential diagnosis, and provide empirical evidence that using differentials as training signals is essential for the efficiency of such systems or for helping doctors better understand the reasoning of those systems."
                },
                {
                    "title": "Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?",
                    "abstract": "Large language models (LMs) are able to in-context learn\u2014perform a new task via inference alone by conditioning on a few input-label pairs (demonstrations) and making predictions for new inputs. However, there has been little understanding of how the model learns and which aspects of the demonstrations contribute to end task performance. In this paper, we show that ground truth demonstrations are in fact not required\u2014randomly replacing labels in the demonstrations barely hurts performance on a range of classification and multi-choce tasks, consistently over 12 different models including GPT-3. Instead, we find that other aspects of the demonstrations are the key drivers of endtask performance, including the fact that they provide a few examples of (1) the label space, (2) the distribution of the input text, and (3) the overall format of the sequence. Together, our analysis provides a new way of understanding how and why in-context learning works, while opening up new questions about how much can be learned from large language models through inference alone."
                },
                {
                    "title": "Learning To Retrieve Prompts for In-Context Learning",
                    "abstract": "In-context learning is a recent paradigm in natural language understanding, where a large pre-trained language model (LM) observes a test instance and a few training examples as its input, and directly decodes the output without any update to its parameters. However, performance has been shown to strongly depend on the selected training examples (termed prompts). In this work, we propose an efficient method for retrieving prompts for in-context learning using annotated data and an LM. Given an input-output pair, we estimate the probability of the output given the input and a candidate training example as the prompt, and label training examples as positive or negative based on this probability. We then train an efficient dense retriever from this data, which is used to retrieve training examples as prompts at test time. We evaluate our approach on three sequence-to-sequence tasks where language utterances are mapped to meaning representations, and find that it substantially outperforms prior work and multiple baselines across the board."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "Finetuned Language Models Are Zero-Shot Learners",
                    "abstract": "This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "CoSQL: A Conversational Text-to-SQL Challenge Towards Cross-Domain Natural Language Interfaces to Databases",
                    "abstract": "We present CoSQL, a corpus for building cross-domain, general-purpose database (DB) querying dialogue systems. It consists of 30k+ turns plus 10k+ annotated SQL queries, obtained from a Wizard-of-Oz (WOZ) collection of 3k dialogues querying 200 complex DBs spanning 138 domains. Each dialogue simulates a real-world DB query scenario with a crowd worker as a user exploring the DB and a SQL expert retrieving answers with SQL, clarifying ambiguous questions, or otherwise informing of unanswerable questions. When user questions are answerable by SQL, the expert describes the SQL and execution results to the user, hence maintaining a natural interaction flow. CoSQL introduces new challenges compared to existing task-oriented dialogue datasets: (1) the dialogue states are grounded in SQL, a domain-independent executable representation, instead of domain-specific slot value pairs, and (2) because testing is done on unseen databases, success requires generalizing to new domains. CoSQL includes three tasks: SQL-grounded dialogue state tracking, response generation from query results, and user dialogue act prediction. We evaluate a set of strong baselines for each task and show that CoSQL presents significant challenges for future research. The dataset, baselines, and leaderboard will be released at https://yale-lily.github.io/cosql."
                },
                {
                    "title": "HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering",
                    "abstract": "Existing question answering (QA) datasets fail to train QA systems to perform complex reasoning and provide explanations for answers. We introduce HotpotQA, a new dataset with 113k Wikipedia-based question-answer pairs with four key features: (1) the questions require finding and reasoning over multiple supporting documents to answer; (2) the questions are diverse and not constrained to any pre-existing knowledge bases or knowledge schemas; (3) we provide sentence-level supporting facts required for reasoning, allowing QA systems to reason with strong supervision and explain the predictions; (4) we offer a new type of factoid comparison questions to test QA systems\u2019 ability to extract relevant facts and perform necessary comparison. We show that HotpotQA is challenging for the latest QA systems, and the supporting facts enable models to improve performance and make explainable predictions."
                },
                {
                    "title": "Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task",
                    "abstract": "We present Spider, a large-scale complex and cross-domain semantic parsing and text-to-SQL dataset annotated by 11 college students. It consists of 10,181 questions and 5,693 unique complex SQL queries on 200 databases with multiple tables covering 138 different domains. We define a new complex and cross-domain semantic parsing and text-to-SQL task so that different complicated SQL queries and databases appear in train and test sets. In this way, the task requires the model to generalize well to both new SQL queries and new database schemas. Therefore, Spider is distinct from most of the previous semantic parsing tasks because they all use a single database and have the exact same program in the train set and the test set. We experiment with various state-of-the-art models and the best model achieves only 9.7% exact matching accuracy on a database split setting. This shows that Spider presents a strong challenge for future research. Our dataset and task with the most recent updates are publicly available at https://yale-lily.github.io/seq2sql/spider."
                },
                {
                    "title": "A Learning Algorithm for Continually Running Fully Recurrent Neural Networks",
                    "abstract": "The exact form of a gradient-following learning algorithm for completely recurrent networks running in continually sampled time is derived and used as the basis for practical algorithms for temporal supervised learning tasks. These algorithms have (1) the advantage that they do not require a precisely defined training interval, operating while the network runs; and (2) the disadvantage that they require nonlocal communication in the network being trained and are computationally expensive. These algorithms allow networks having recurrent connections to learn complex tasks that require the retention of information over time periods having either fixed or indefinite length."
                },
                {
                    "title": "Memory-assisted prompt editing to improve GPT-3 after deployment",
                    "abstract": "Large LMs such as GPT -3 are powerful, but can commit mistakes that are obvious to humans. For example, GPT -3 would mistakenly interpret \"What word is similar to good ?\" to mean a homonym, while the user intended a synonym. Our goal is to effectively correct such errors via user interactions with the sys-tem but without retraining, which will be pro-hibitively costly. We pair GPT -3 with a growing memory of recorded cases where the model misunderstood the user\u2019s intents, along with user feedback for clarification. Such a memory allows our system to produce enhanced prompts for any new query based on the user feedback for error correction on similar cases in the past. On four tasks (two lexical tasks, two advanced ethical reasoning tasks), we show how a (simulated) user can interactively teach a deployed GPT -3, substantially increasing its accuracy over the queries with different kinds of misunderstandings by the GPT -3. Our approach is a step towards the low-cost utility enhancement for very large pre-trained LMs. 1"
                },
                {
                    "title": "The Claude 3 Model Family: Opus, Sonnet, Haiku",
                    "abstract": "We introduce Claude 3, a new family of large multimodal models \u2013 Claude 3 Opus , our most capable offering, Claude 3 Sonnet , which provides a combination of skills and speed, and Claude 3 Haiku , our fastest and least expensive model. All new models have vision capabilities that enable them to process and analyze image data. The Claude 3 family demonstrates strong performance across benchmark evaluations and sets a new standard on measures of reasoning, math, and coding. Claude 3 Opus achieves state-of-the-art results on evaluations like GPQA [1], MMLU [2], MMMU [3] and many more. Claude 3 Haiku performs as well or better than Claude 2 [4] on most pure-text tasks, while Sonnet and Opus significantly outperform it. Additionally, these models exhibit improved fluency in non-English languages, making them more versatile for a global audience. In this report, we provide an in-depth analysis of our evaluations, focusing on core capabilities, safety, societal impacts, and the catastrophic risk assessments we committed to in our Responsible Scaling Policy [5]."
                }
            ],
            "categories": [
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively evaluate and enhance the self-improvement capabilities of LLM agents in an online setting using an input-feedback sequence?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant gap in the evaluation of LLM agents, moving beyond static benchmarks to dynamic, real-world applications. By developing a benchmark like StreamBench, we can better understand how LLM agents can adapt and improve over time, which is essential for their deployment in practical scenarios. This advancement could lead to more effective and responsive AI systems, ultimately influencing future research directions in adaptive learning and continuous improvement in machine learning models.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the complexity of designing an evaluation framework that accurately captures the nuances of self-improvement in LLM agents. Naive approaches may fail because they do not account for the dynamic nature of user feedback and the need for agents to adapt in real-time. Technical obstacles include ensuring that the evaluation metrics reflect true performance improvements and managing the computational costs associated with continuous learning. Additionally, theoretical challenges arise in understanding how to model the feedback loop effectively.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on static evaluations of LLMs, lacking a framework for assessing their ability to learn and improve over time. Existing benchmarks do not capture the iterative nature of user interactions and feedback, which has limited the understanding of LLM agents' adaptive capabilities. Barriers such as the absence of a suitable online evaluation environment and the complexity of modeling continuous learning have prevented this problem from being addressed. Our approach with StreamBench differs by providing a structured way to evaluate LLM agents in streaming scenarios, filling this critical gap.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of StreamBench, a benchmark designed to evaluate LLM agents' performance in an online setting through an input-feedback sequence. We will utilize a diverse range of tasks to assess the agents' ability to improve over time. The evaluation metrics will focus on the accuracy of the agents' predictions throughout the sequence. Expected outcomes include demonstrating that LLM agents can effectively enhance their performance using our proposed streaming baselines, including a cost-effective multi-agent method that maintains efficiency while improving accuracy."
            }
        },
        "author_data": {
            "22413ced-a807-47d6-9275-ccc411e1dfe8": {
                "pk": "22413ced-a807-47d6-9275-ccc411e1dfe8",
                "name": "Cheng-Kuang Wu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "07868dd1-770e-45c9-aac0-16bd69001cc8": {
                "pk": "07868dd1-770e-45c9-aac0-16bd69001cc8",
                "name": "Zhi Rui Tam",
                "collaborators": [
                    "Cheng-Kuang Wu",
                    "Chieh-Yen Lin",
                    "Hung-yi Lee",
                    "Yun-Nung Chen",
                    "Yi-Lin Tsai",
                    "Chao-Chung Wu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Structured Generation",
                    "User Interaction"
                ],
                "publications": [
                    {
                        "title": "Let Me Speak Freely? A Study on the Impact of Format Restrictions on Performance of Large Language Models",
                        "abstract": "Structured generation, the process of producing content in standardized formats like JSON and XML, is widely utilized in real-world applications to extract key output information from large language models (LLMs). This study investigates whether such constraints on generation space impact LLMs abilities, including reasoning and domain knowledge comprehension. Specifically, we evaluate LLMs performance when restricted to adhere to structured formats versus generating free-form responses across various common tasks. Surprisingly, we observe a significant decline in LLMs reasoning abilities under format restrictions. Furthermore, we find that stricter format constraints generally lead to greater performance degradation in reasoning tasks."
                    },
                    {
                        "title": "I Need Help! Evaluating LLM's Ability to Ask for Users' Support: A Case Study on Text-to-SQL Generation",
                        "abstract": "This study explores the proactive ability of LLMs to seek user support. We propose metrics to evaluate the trade-off between performance improvements and user burden, and investigate whether LLMs can determine when to request help under varying information availability. Our experiments show that without external feedback, many LLMs struggle to recognize their need for user support. The findings highlight the importance of external signals and provide insights for future research on improving support-seeking strategies. Source code: https://github.com/appier-research/i-need-help"
                    }
                ]
            },
            "a688ac9a-3bd6-4878-9b5c-8f0d9b8561d5": {
                "pk": "a688ac9a-3bd6-4878-9b5c-8f0d9b8561d5",
                "name": "Chieh-Yen Lin",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "3b9c9bc3-7e3d-442c-ba29-712c86ccacbe": {
                "pk": "3b9c9bc3-7e3d-442c-ba29-712c86ccacbe",
                "name": "Yun-Nung Chen",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "6d0d7cd9-9346-4878-8a22-d1203e169e93": {
                "pk": "6d0d7cd9-9346-4878-8a22-d1203e169e93",
                "name": "Hung-yi Lee",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2406.00048": {
        "paper_data": {
            "title": "Towards a theory of how the structure of language is acquired by deep neural networks",
            "url": "http://arxiv.org/abs/2406.00048v2",
            "arxiv_id": "2406.00048",
            "authors": [
                "Francesco Cagnetta",
                "Matthieu Wyart"
            ],
            "abstract": "How much data is required to learn the structure of a language via next-token prediction? We study this question for synthetic datasets generated via a Probabilistic Context-Free Grammar (PCFG) -- a tree-like generative model that captures many of the hierarchical structures found in natural languages. We determine token-token correlations analytically in our model and show that they can be used to build a representation of the grammar's hidden variables, the longer the range the deeper the variable. In addition, a finite training set limits the resolution of correlations to an effective range, whose size grows with that of the training set. As a result, a Language Model trained with increasingly many examples can build a deeper representation of the grammar's structure, thus reaching good performance despite the high dimensionality of the problem. We conjecture that the relationship between training set size and effective range of correlations holds beyond our synthetic datasets. In particular, our conjecture predicts how the scaling law for the test loss behaviour with training set size depends on the length of the context window, which we confirm empirically in Shakespeare's plays and Wikipedia articles.",
            "introduction": "   1 Introduction  Two central foci of linguistics are the language structure and how humans acquire it. Formal language theory, for instance, describes languages with hierarchical generative models of grammar, classified in different levels of complexity\u00a0[1, 2]. In this context, the \u2018poverty of the stimulus\u2019 argument\u00a0[3]\u2014stating that the data children receive is insufficient to uniquely determine the grammatical structure of their language\u2014led to the hypothesis that linguistic faculties are largely innate. By contrast, statistical learning theory\u00a0[4, 5] posits that the statistics of the input data can be used to deduce the language structure. This assumption is supported by empirical evidence concerning a broad range of tasks, including word segmentation\u00a0[6] and reconstruction of the hierarchical phrase structure\u00a0[7].   Large Language Models (LLMs) offer an interesting perspective on the subject. For instance, the success of LLMs trained for next-token prediction\u00a0[8, 9] establishes that a language can be acquired from examples alone\u2014albeit with a training set much larger than what humans are exposed to. Furthermore, empirical studies of LLMs\u2019 representations showed that they learn a hierarchy of contextual information, including notions of linguistics such as word classes and syntactic structure\u00a0[10, 11, 12]. Recent studies have begun revealing the inner workings of LLMs by using synthetic data generated via context-free grammars\u00a0[13, 14], determining, in particular, the algorithm that these models follow when predicting the next token. However, there is no consensus on the mechanisms behind language acquisition by LLMs\u00a0[15, 16]. As a result, empirical phenomena such as the scaling of the test loss with dataset size and number of parameters\u00a0[17] and the emergence of specific skills at certain scales\u00a0[18, 19] remain unexplained. In this work, we use hierarchical generative models of data to describe how the structure of a language is learnt as the training set grows.    1.1 Our contributions  We consider synthetic datasets generated via the Random Hierarchy Model (RHM) [20], an ensemble of probabilistic context-free grammars (PCFGs). The RHM generates sequences of tokens by applying randomly chosen production rules to a hierarchy of hidden variables that live on the nodes of a tree with fixed geometry.   \u2022  We characterise analytically the power-law decay of the correlations between tokens with their distance. We then show that, because of this decay, a finite training set size P\ud835\udc43Pitalic_P limits the resolution of correlations to an effective context window, whose size t\u2217superscript\ud835\udc61t^{*}italic_t start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT increases with P\ud835\udc43Pitalic_P.    \u2022  Building on previous works on classification, we argue that deep learning models trained on next-token prediction can use measurable correlations to represent the hidden variables of the PCFG, with larger P\ud835\udc43Pitalic_P allowing the representation of deeper hidden variables.    \u2022  Combining these results, we predict a sequence of sample complexities where the emergence of a deeper data structure representation leads to a jump in test loss. We empirically validate this for deep transformers and CNNs. Notably, the sample complexities are polynomial in the effective context size t\u2217superscript\ud835\udc61t^{*}italic_t start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT, avoiding the curse of dimensionality.    \u2022  We conjecture that the relationship between training set size, correlations and effective context window holds beyond our data model, and we test it by training deep transformers on collections of Shakespeare\u2019s lines and Wikipedia articles. In particular,",
            "references": [
                {
                    "title": "Critical Phase Transition in Large Language Models",
                    "abstract": "Large Language Models (LLMs) have demonstrated impressive performance. To understand their behaviors, we need to consider the fact that LLMs sometimes show qualitative changes. The natural world also presents such changes called phase transitions, which are defined by singular, divergent statistical quantities. Therefore, an intriguing question is whether qualitative changes in LLMs are phase transitions. In this work, we have conducted extensive analysis on texts generated by LLMs and suggested that a phase transition occurs in LLMs when varying the temperature parameter. Specifically, statistical quantities have divergent properties just at the point between the low-temperature regime, where LLMs generate sentences with clear repetitive structures, and the high-temperature regime, where generated sentences are often incomprehensible. In addition, critical behaviors near the phase transition point, such as a power-law decay of correlation and slow convergence toward the stationary state, are similar to those in natural languages. Our results suggest a meaningful analogy between LLMs and natural phenomena."
                },
                {
                    "title": "U-Nets as Belief Propagation: Efficient Classification, Denoising, and Diffusion in Generative Hierarchical Models",
                    "abstract": "U-Nets are among the most widely used architectures in computer vision, renowned for their exceptional performance in applications such as image segmentation, denoising, and diffusion modeling. However, a theoretical explanation of the U-Net architecture design has not yet been fully established. This paper introduces a novel interpretation of the U-Net architecture by studying certain generative hierarchical models, which are tree-structured graphical models extensively utilized in both language and image domains. With their encoder-decoder structure, long skip connections, and pooling and up-sampling layers, we demonstrate how U-Nets can naturally implement the belief propagation denoising algorithm in such generative hierarchical models, thereby efficiently approximating the denoising functions. This leads to an efficient sample complexity bound for learning the denoising function using U-Nets within these models. Additionally, we discuss the broader implications of these findings for diffusion models in generative hierarchical models. We also demonstrate that the conventional architecture of convolutional neural networks (ConvNets) is ideally suited for classification tasks within these models. This offers a unified view of the roles of ConvNets and U-Nets, highlighting the versatility of generative hierarchical models in modeling complex data distributions across language and image domains."
                },
                {
                    "title": "Learning Syntax Without Planting Trees: Understanding When and Why Transformers Generalize Hierarchically",
                    "abstract": "Transformers trained on natural language data have been shown to learn its hierarchical structure and generalize to sentences with unseen syntactic structures without explicitly encoding any structural bias. In this work, we investigate sources of inductive bias in transformer models and their training that could cause such generalization behavior to emerge. We extensively experiment with transformer models trained on multiple synthetic datasets and with different training objectives and show that while other objectives e.g. sequence-to-sequence modeling, prefix language modeling, often failed to lead to hierarchical generalization, models trained with the language modeling objective consistently learned to generalize hierarchically. We then conduct pruning experiments to study how transformers trained with the language modeling objective encode hierarchical structure. When pruned, we find joint existence of subnetworks within the model with different generalization behaviors (subnetworks corresponding to hierarchical structure and linear order). Finally, we take a Bayesian perspective to further uncover transformers' preference for hierarchical generalization: We establish a correlation between whether transformers generalize hierarchically on a dataset and whether the simplest explanation of that dataset is provided by a hierarchical grammar compared to regular grammars exhibiting linear generalization."
                },
                {
                    "title": "How Deep Networks Learn Sparse and Hierarchical Data: the Sparse Random Hierarchy Model",
                    "abstract": "Understanding what makes high-dimensional data learnable is a fundamental question in machine learning. On the one hand, it is believed that the success of deep learning lies in its ability to build a hierarchy of representations that become increasingly more abstract with depth, going from simple features like edges to more complex concepts. On the other hand, learning to be insensitive to invariances of the task, such as smooth transformations for image datasets, has been argued to be important for deep networks and it strongly correlates with their performance. In this work, we aim to explain this correlation and unify these two viewpoints. We show that by introducing sparsity to generative hierarchical models of data, the task acquires insensitivity to spatial transformations that are discrete versions of smooth transformations. In particular, we introduce the Sparse Random Hierarchy Model (SRHM), where we observe and rationalize that a hierarchical representation mirroring the hierarchical model is learnt precisely when such insensitivity is learnt, thereby explaining the strong correlation between the latter and performance. Moreover, we quantify how the sample complexity of CNNs learning the SRHM depends on both the sparsity and hierarchical structure of the task."
                },
                {
                    "title": "A Phase Transition in Diffusion Models Reveals the Hierarchical Nature of Data",
                    "abstract": "Understanding the structure of real data is paramount in advancing modern deep-learning methodologies. Natural data such as images are believed to be composed of features organised in a hierarchical and combinatorial manner, which neural networks capture during learning. Recent advancements show that diffusion models can generate high-quality images, hinting at their ability to capture this underlying structure. We study this phenomenon in a hierarchical generative model of data. We find that the backward diffusion process acting after a time $t$ is governed by a phase transition at some threshold time, where the probability of reconstructing high-level features, like the class of an image, suddenly drops. Instead, the reconstruction of low-level features, such as specific details of an image, evolves smoothly across the whole diffusion process. This result implies that at times beyond the transition, the class has changed but the generated sample may still be composed of low-level elements of the initial image. We validate these theoretical insights through numerical experiments on class-unconditional ImageNet diffusion models. Our analysis characterises the relationship between time and scale in diffusion models and puts forward generative models as powerful tools to model combinatorial data properties."
                },
                {
                    "title": "The Benefits of Reusing Batches for Gradient Descent in Two-Layer Networks: Breaking the Curse of Information and Leap Exponents",
                    "abstract": "We investigate the training dynamics of two-layer neural networks when learning multi-index target functions. We focus on multi-pass gradient descent (GD) that reuses the batches multiple times and show that it significantly changes the conclusion about which functions are learnable compared to single-pass gradient descent. In particular, multi-pass GD with finite stepsize is found to overcome the limitations of gradient flow and single-pass GD given by the information exponent (Ben Arous et al., 2021) and leap exponent (Abbe et al., 2023) of the target function. We show that upon re-using batches, the network achieves in just two time steps an overlap with the target subspace even for functions not satisfying the staircase property (Abbe et al., 2021). We characterize the (broad) class of functions efficiently learned in finite time. The proof of our results is based on the analysis of the Dynamical Mean-Field Theory (DMFT). We further provide a closed-form description of the dynamical process of the low-dimensional projections of the weights, and numerical experiments illustrating the theory."
                },
                {
                    "title": "A Dynamical Model of Neural Scaling Laws",
                    "abstract": "On a variety of tasks, the performance of neural networks predictably improves with training time, dataset size and model size across many orders of magnitude. This phenomenon is known as a neural scaling law. Of fundamental importance is the compute-optimal scaling law, which reports the performance as a function of units of compute when choosing model sizes optimally. We analyze a random feature model trained with gradient descent as a solvable model of network training and generalization. This reproduces many observations about neural scaling laws. First, our model makes a prediction about why the scaling of performance with training time and with model size have different power law exponents. Consequently, the theory predicts an asymmetric compute-optimal scaling rule where the number of training steps are increased faster than model parameters, consistent with recent empirical observations. Second, it has been observed that early in training, networks converge to their infinite-width dynamics at a rate $1/\\textit{width}$ but at late time exhibit a rate $\\textit{width}^{-c}$, where $c$ depends on the structure of the architecture and task. We show that our model exhibits this behavior. Lastly, our theory shows how the gap between training and test loss can gradually build up over time due to repeated reuse of data."
                },
                {
                    "title": "A Theory for Emergence of Complex Skills in Language Models",
                    "abstract": "A major driver of AI products today is the fact that new skills emerge in language models when their parameter set and training corpora are scaled up. This phenomenon is poorly understood, and a mechanistic explanation via mathematical analysis of gradient-based training seems difficult. The current paper takes a different approach, analysing emergence using the famous (and empirical) Scaling Laws of LLMs and a simple statistical framework. Contributions include: (a) A statistical framework that relates cross-entropy loss of LLMs to competence on the basic skills that underlie language tasks. (b) Mathematical analysis showing that the Scaling Laws imply a strong form of inductive bias that allows the pre-trained model to learn very efficiently. We informally call this {\\em slingshot generalization} since naively viewed it appears to give competence levels at skills that violate usual generalization theory. (c) A key example of slingshot generalization, that competence at executing tasks involving $k$-tuples of skills emerges essentially at the same scaling and same rate as competence on the elementary skills themselves."
                },
                {
                    "title": "Large Language Models",
                    "abstract": "Large Language ModelsIn the latest edition of Stats, STAT!, Fralick and colleagues explain the statistics behind large language models - used in chat bots like ChatGPT and Bard. While these new tools may seem remarkably intelligent, at their core they just assemble sentences based on statistics from large amounts of text."
                },
                {
                    "title": "How Deep Neural Networks Learn Compositional Data: The Random Hierarchy Model",
                    "abstract": "Deep learning algorithms demonstrate a surprising ability to learn high-dimensional tasks from limited examples. This is commonly attributed to the depth of neural networks, enabling them to build a hierarchy of abstract, low-dimensional data representations. However, how many training examples are required to learn such representations remains unknown. To quantitatively study this question, we introduce the random hierarchy model: a family of synthetic tasks inspired by the hierarchical structure of language and images. The model is a classification task where each class corresponds to a group of high-level features, chosen among several equivalent groups associated with the same class. In turn, each feature corresponds to a group of subfeatures chosen among several equivalent groups and so on, following a hierarchy of composition rules. We find that deep networks learn the task by developing internal representations invariant to exchanging equivalent groups. Moreover, the number of data required corresponds to the point where correlations between low-level features and classes become detectable. Overall, our results indicate how deep networks overcome the curse of dimensionality by building invariant representations and provide an estimate of the number of data required to learn a hierarchical task.\n \n \n \n \n Published by the American Physical Society\n 2024\n \n \n"
                },
                {
                    "title": "Autocorrelations Decay in Texts and Applicability Limits of Language Models",
                    "abstract": "We show that the laws of autocorrelations decay in texts are closely related to applicability limits of language models. Using distributional semantics we empirically demonstrate that autocorrelations of words in texts decay according to a power law. We show that distributional semantics provides coherent autocorrelations decay exponents for texts translated to multiple languages. The autocorrelations decay in generated texts is quantitatively and often qualitatively different from the literary texts. We conclude that language models exhibiting Markovian behavior, including large autoregressive language models, may have limitations when applied to long texts, whether analysis or generation."
                },
                {
                    "title": "Do Transformers Parse while Predicting the Masked Word?",
                    "abstract": "Pre-trained language models have been shown to encode linguistic structures, e.g. dependency and constituency parse trees, in their embeddings while being trained on unsupervised loss functions like masked language modeling. Some doubts have been raised whether the models actually are doing parsing or only some computation weakly correlated with it. We study questions: (a) Is it possible to explicitly describe transformers with realistic embedding dimension, number of heads, etc. that are capable of doing parsing -- or even approximate parsing? (b) Why do pre-trained models capture parsing structure? This paper takes a step toward answering these questions in the context of generative modeling with PCFGs. We show that masked language models like BERT or RoBERTa of moderate sizes can approximately execute the Inside-Outside algorithm for the English PCFG [Marcus et al, 1993]. We also show that the Inside-Outside algorithm is optimal for masked language modeling loss on the PCFG-generated data. We also give a construction of transformers with $50$ layers, $15$ attention heads, and $1275$ dimensional embeddings in average such that using its embeddings it is possible to do constituency parsing with $>70\\%$ F1 score on PTB dataset. We conduct probing experiments on models pre-trained on PCFG-generated data to show that this not only allows recovery of approximate parse tree, but also recovers marginal span probabilities computed by the Inside-Outside algorithm, which suggests an implicit bias of masked language modeling towards this algorithm."
                },
                {
                    "title": "What Can Be Learnt With Wide Convolutional Neural Networks?",
                    "abstract": "\n Understanding how convolutional neural networks (CNNs) can efficiently learn high-dimensional functions remains a fundamental challenge. A popular belief is that these models harness the local and hierarchical structure of natural data such as images. Yet, we lack a quantitative understanding of how such structure affects performance, for example the rate of decay of the generalisation error with the number of training samples. In this paper, we study infinitely wide deep CNNs in the kernel regime. First, we show that the spectrum of the corresponding kernel inherits the hierarchical structure of the network, and we characterise its asymptotics. Then, we use this result together with generalisation bounds to prove that deep CNNs adapt to the spatial scale of the target function. In particular, we find that if the target function depends on low-dimensional subsets of adjacent input variables then the decay of the error is controlled by the effective dimensionality of these subsets. Conversely, if the target function depends on the full set of input variables then the error decay is controlled by the input dimension. We conclude by computing the generalisation error of a deep CNN trained on the output of another deep CNN with randomly initialised parameters. Interestingly, we find that, despite their hierarchical structure, the functions generated by infinitely wide deep CNNs are too rich to be efficiently learnable in high dimensions."
                },
                {
                    "title": "Emergent Abilities of Large Language Models",
                    "abstract": "Scaling up language models has been shown to predictably improve performance and sample efficiency on a wide range of downstream tasks. This paper instead discusses an unpredictable phenomenon that we refer to as emergent abilities of large language models. We consider an ability to be emergent if it is not present in smaller models but is present in larger models. Thus, emergent abilities cannot be predicted simply by extrapolating the performance of smaller models. The existence of such emergence implies that additional scaling could further expand the range of capabilities of language models."
                },
                {
                    "title": "Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models",
                    "abstract": "Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit\"breakthrough\"behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting."
                },
                {
                    "title": "Predictability and Surprise in Large Generative Models",
                    "abstract": "Large-scale pre-training has recently emerged as a technique for creating capable, general-purpose, generative models such as GPT-3, Megatron-Turing NLG, Gopher, and many others. In this paper, we highlight a counterintuitive property of such models and discuss the policy implications of this property. Namely, these generative models have a paradoxical combination of predictable loss on a broad training distribution (as embodied in their \u201dscaling laws\u201d), and unpredictable specific capabilities, inputs, and outputs. We believe that the high-level predictability and appearance of useful capabilities drives rapid development of such models, while the unpredictable qualities make it difficult to anticipate the consequences of model deployment. We go through examples of how this combination can lead to socially harmful behavior with examples from the literature and real world observations, and we also perform two novel experiments to illustrate our point about harms from unpredictability. Furthermore, we analyze how these conflicting properties combine to give model developers various motivations for deploying these models, and challenges that can hinder deployment. We conclude with a list of possible interventions the AI community may take to increase the chance of these models having a beneficial impact. We intend for this paper to be useful to policymakers who want to understand and regulate AI systems, technologists who care about the potential policy impact of their work, funders who want to support work addressing these challenges, and academics who want to analyze, critique, and potentially develop large generative models."
                },
                {
                    "title": "On Aspects of the Theory of Syntax",
                    "abstract": "\nIn the first essay for a new series on landmark texts, Anna Maria Di Sciullo revisits Noam Chomsky\u2019s classic Aspects on the Theory of Syntax. Published in 1965, Aspects was a revelation, presenting linguists with what, at once, became the Standard Theory. It promoted linguistics into a science, one that accepted the methods and the standards of the hard sciences themselves.\n"
                },
                {
                    "title": "Explaining neural scaling laws",
                    "abstract": "Significance The population loss of trained deep neural networks has been empirically observed to improve as a power law in a variety of large models and datasets. We investigate the origins behind such \u201cscaling laws\u201d and provide a taxonomy for different scaling regimes. Our findings are based on derivations in linear random feature models\u2014which, in addition to being a simple fruitful model, also describe the wide network limit of deep neural networks. We further formulate and verify aspects of scaling based on smoothness in interpolating a data manifold. We support our theory with empirical results in realistic settings. Our work provides insights into scaling laws and bridges the large gap between theory and experiment in modern deep learning."
                },
                {
                    "title": "Asymptotic learning curves of kernel methods: empirical data versus teacher\u2013student paradigm",
                    "abstract": "How many training data are needed to learn a supervised task? It is often observed that the generalization error decreases as n \u2212\u03b2 where n is the number of training examples and \u03b2 is an exponent that depends on both data and algorithm. In this work we measure \u03b2 when applying kernel methods to real datasets. For MNIST we find \u03b2 \u2248 0.4 and for CIFAR10 \u03b2 \u2248 0.1, for both regression and classification tasks, and for Gaussian or Laplace kernels. To rationalize the existence of non-trivial exponents that can be independent of the specific kernel used, we study the teacher\u2013student framework for kernels. In this scheme, a teacher generates data according to a Gaussian random field, and a student learns them via kernel regression. With a simplifying assumption\u2014namely that the data are sampled from a regular lattice\u2014we derive analytically \u03b2 for translation invariant kernels, using previous results from the kriging literature. Provided that the student is not too sensitive to high frequencies, \u03b2 depends only on the smoothness and dimension of the training data. We confirm numerically that these predictions hold when the training points are sampled at random on a hypersphere. Overall, the test error is found to be controlled by the magnitude of the projection of the true function on the kernel eigenvectors whose rank is larger than n. Using this idea we predict the exponent \u03b2 from real data by performing kernel PCA, leading to \u03b2 \u2248 0.36 for MNIST and \u03b2 \u2248 0.07 for CIFAR10, in good agreement with observations. We argue that these rather large exponents are possible due to the small effective dimension of the data."
                },
                {
                    "title": "Feature Learning in Infinite-Width Neural Networks",
                    "abstract": "As its width tends to infinity, a deep neural network's behavior under gradient descent can become simplified and predictable (e.g. given by the Neural Tangent Kernel (NTK)), if it is parametrized appropriately (e.g. the NTK parametrization). However, we show that the standard and NTK parametrizations of a neural network do not admit infinite-width limits that can learn features, which is crucial for pretraining and transfer learning such as with BERT. We propose simple modifications to the standard parametrization to allow for feature learning in the limit. Using the *Tensor Programs* technique, we derive explicit formulas for such limits. On Word2Vec and few-shot learning on Omniglot via MAML, two canonical tasks that rely crucially on feature learning, we compute these limits exactly. We find that they outperform both NTK baselines and finite-width networks, with the latter approaching the infinite-width feature learning performance as width increases. \nMore generally, we classify a natural space of neural network parametrizations that generalizes standard, NTK, and Mean Field parametrizations. We show 1) any parametrization in this space either admits feature learning or has an infinite-width training dynamics given by kernel gradient descent, but not both; 2) any such infinite-width limit can be computed using the Tensor Programs technique."
                },
                {
                    "title": "LSTMS Compose \u2014 and Learn \u2014 Bottom-Up",
                    "abstract": "Recent work in NLP shows that LSTM language models capture compositional structure in language data. In contrast to existing work, we consider the learning process that leads to compositional behavior. For a closer look at how an LSTM\u2019s sequential representations are composed hierarchically, we present a related measure of Decompositional Interdependence (DI) between word meanings in an LSTM, based on their gate interactions. We support this measure with experiments on English language data, where DI is higher on pairs of words with lower syntactic distance. To explore the inductive biases that cause these compositional representations to arise during training, we conduct simple experiments on synthetic data. These synthetic experiments support a specific hypothesis about how hierarchical structures are discovered over the course of training: that LSTM constituent representations are learned bottom-up, relying on effective representations of their shorter children, rather than on learning the longer-range relations independently."
                },
                {
                    "title": "Consistent unsupervised estimators for anchored PCFGs",
                    "abstract": "Abstract Learning probabilistic context-free grammars (PCFGs) from strings is a classic problem in computational linguistics since Horning (1969). Here we present an algorithm based on distributional learning that is a consistent estimator for a large class of PCFGs that satisfy certain natural conditions including being anchored (Stratos et al., 2016). We proceed via a reparameterization of (top\u2013down) PCFGs that we call a bottom\u2013up weighted context-free grammar. We show that if the grammar is anchored and satisfies additional restrictions on its ambiguity, then the parameters can be directly related to distributional properties of the anchoring strings; we show the asymptotic correctness of a naive estimator and present some simulations using synthetic data that show that algorithms based on this approach have good finite sample behavior."
                },
                {
                    "title": "Emergent linguistic structure in artificial neural networks trained by self-supervision",
                    "abstract": "This paper explores the knowledge of linguistic structure learned by large artificial neural networks, trained via self-supervision, whereby the model simply tries to predict a masked word in a given context. Human language communication is via sequences of words, but language understanding requires constructing rich hierarchical structures that are never observed explicitly. The mechanisms for this have been a prime mystery of human language acquisition, while engineering work has mainly proceeded by supervised learning on treebanks of sentences hand labeled for this latent structure. However, we demonstrate that modern deep contextual language models learn major aspects of this structure, without any explicit supervision. We develop methods for identifying linguistic hierarchical structure emergent in artificial neural networks and demonstrate that components in these models focus on syntactic grammatical relationships and anaphoric coreference. Indeed, we show that a linear transformation of learned embeddings in these models captures parse tree distances to a surprising degree, allowing approximate reconstruction of the sentence tree structures normally assumed by linguists. These results help explain why these models have brought such large improvements across many language-understanding tasks."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Does Syntax Need to Grow on Trees? Sources of Hierarchical Inductive Bias in Sequence-to-Sequence Networks",
                    "abstract": "Learners that are exposed to the same training data might generalize differently due to differing inductive biases. In neural network models, inductive biases could in theory arise from any aspect of the model architecture. We investigate which architectural factors affect the generalization behavior of neural sequence-to-sequence models trained on two syntactic tasks, English question formation and English tense reinflection. For both tasks, the training set is consistent with a generalization based on hierarchical structure and a generalization based on linear order. All architectural factors that we investigated qualitatively affected how models generalized, including factors with no clear connection to hierarchical structure. For example, LSTMs and GRUs displayed qualitatively different inductive biases. However, the only factor that consistently contributed a hierarchical bias across tasks was the use of a tree-structured model rather than a model with sequential recurrence, suggesting that human-like syntactic generalization requires architectural syntactic structure."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "BERT Rediscovers the Classical NLP Pipeline",
                    "abstract": "Pre-trained text encoders have rapidly advanced the state of the art on many NLP tasks. We focus on one such model, BERT, and aim to quantify where linguistic information is captured within the network. We find that the model represents the steps of the traditional NLP pipeline in an interpretable and localizable way, and that the regions responsible for each step appear in the expected sequence: POS tagging, parsing, NER, semantic roles, then coreference. Qualitative analysis reveals that the model can and often does adjust this pipeline dynamically, revising lower-level decisions on the basis of disambiguating information from higher-level representations."
                },
                {
                    "title": "Emergence of order in random languages",
                    "abstract": "We consider languages generated by weighted context-free grammar. It is shown that the behavior of large texts is controlled by saddle-point equations for an appropriate generating function. We then consider ensembles of grammars, in particular the random language model of De Giuli (2019 Phys. Rev. Lett. 122 128301). This model is solved in the replica-symmetric ansatz, which is valid in the high-temperature, disordered phase. It is shown that in the phase in which languages carry information, the replica symmetry must be broken."
                },
                {
                    "title": "Dissecting Contextual Word Embeddings: Architecture and Representation",
                    "abstract": "Contextual word representations derived from pre-trained bidirectional language models (biLMs) have recently been shown to provide significant improvements to the state of the art for a wide range of NLP tasks. However, many questions remain as to how and why these models are so effective. In this paper, we present a detailed empirical study of how the choice of neural architecture (e.g. LSTM, CNN, or self attention) influences both end task accuracy and qualitative properties of the representations that are learned. We show there is a tradeoff between speed and accuracy, but all architectures learn high quality contextual representations that outperform word embeddings for four challenging NLP tasks. Additionally, all architectures learn representations that vary with network depth, from exclusively morphological based at the word embedding layer through local syntax based in the lower contextual layers to longer range semantics such coreference at the upper layers. Together, these results suggest that unsupervised biLMs, independent of architecture, are learning much more about the structure of language than previously appreciated."
                },
                {
                    "title": "A Provably Correct Algorithm for Deep Learning that Actually Works",
                    "abstract": "We describe a layer-by-layer algorithm for training deep convolutional networks, where each step involves gradient updates for a two layer network followed by a simple clustering algorithm. Our algorithm stems from a deep generative model that generates mages level by level, where lower resolution images correspond to latent semantic classes. We analyze the convergence rate of our algorithm assuming that the data is indeed generated according to this model (as well as additional assumptions). While we do not pretend to claim that the assumptions are realistic for natural images, we do believe that they capture some true properties of real data. Furthermore, we show that our algorithm actually works in practice (on the CIFAR dataset), achieving results in the same ballpark as that of vanilla convolutional neural networks that are being trained by stochastic gradient descent. Finally, our proof techniques may be of independent interest."
                },
                {
                    "title": "Infant Statistical Learning",
                    "abstract": "&NA; Perception involves making sense of a dynamic, multimodal environment. In the absence of mechanisms capable of exploiting the statistical patterns in the natural world, infants would face an insurmountable computational problem. Infant statistical learning mechanisms facilitate the detection of structure. These abilities allow the infant to compute across elements in their environmental input, extracting patterns for further processing and subsequent learning. In this selective review, we summarize findings that show that statistical learning is both a broad and flexible mechanism (supporting learning from different modalities across many different content areas) and input specific (shifting computations depending on the type of input and goal of learning). We suggest that statistical learning not only provides a framework for studying language development and object knowledge in constrained laboratory settings, but also allows researchers to tackle real\u2010world problems, such as multilingualism, the role of ever\u2010changing learning environments, and differential developmental trajectories."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Deep Learning and Hierarchal Generative Models",
                    "abstract": "It is argued that deep learning is efficient for data that is generated from hierarchal generative models. Examples of such generative models include wavelet scattering networks, functions of compositional structure, and deep rendering models. Unfortunately so far, for all such models, it is either not rigorously known that they can be learned efficiently, or it is not known that \"deep algorithms\" are required in order to learn them. \nWe propose a simple family of \"generative hierarchal models\" which can be efficiently learned and where \"deep\" algorithm are necessary for learning. Our definition of \"deep\" algorithms is based on the empirical observation that deep nets necessarily use correlations between features. More formally, we show that in a semi-supervised setting, given access to low-order moments of the labeled data and all of the unlabeled data, it is information theoretically impossible to perform classification while at the same time there is an efficient algorithm, that given all labelled and unlabeled data, perfectly labels all unlabelled data with high probability. \nFor the proof, we use and strengthen the fact that Belief Propagation does not admit a good approximation in terms of linear functions."
                },
                {
                    "title": "Pointer Sentinel Mixture Models",
                    "abstract": "Recent neural network sequence models with softmax classifiers have achieved their best language modeling performance only with very large hidden states and large vocabularies. Even then they struggle to predict rare or unseen words even if the context makes the prediction unambiguous. We introduce the pointer sentinel mixture architecture for neural sequence models which has the ability to either reproduce a word from the recent context or produce a word from a standard softmax classifier. Our pointer sentinel-LSTM model achieves state of the art language modeling performance on the Penn Treebank (70.9 perplexity) while using far fewer parameters than a standard softmax LSTM. In order to evaluate how well language models can exploit longer contexts and deal with more realistic vocabularies and larger corpora we also introduce the freely available WikiText corpus."
                },
                {
                    "title": "Critical Behavior in Physics and Probabilistic Formal Languages",
                    "abstract": "We show that in many data sequences - from texts in different languages to melodies and genomes - the mutual information between two symbols decays roughly like a power law with the number of symbols in between the two. In contrast, we prove that Markov/hidden Markov processes generically exhibit exponential decay in their mutual information, which explains why natural languages are poorly approximated by Markov processes. We present a broad class of models that naturally reproduce this critical behavior. They all involve deep dynamics of a recursive nature, as can be approximately implemented by tree-like or recurrent deep neural networks. This model class captures the essence of probabilistic context-free grammars as well as recursive self-reproduction in physical phenomena such as turbulence and cosmological inflation. We derive an analytic formula for the asymptotic power law and elucidate our results in a statistical physics context: 1-dimensional \"shallow\" models (such as Markov models or regular grammars) will fail to model natural language, because they cannot exhibit criticality, whereas \"deep\" models with one or more \"hidden\" dimensions representing levels of abstraction or scale can potentially succeed."
                },
                {
                    "title": "Formal language theory: refining the Chomsky hierarchy",
                    "abstract": "The first part of this article gives a brief overview of the four levels of the Chomsky hierarchy, with a special emphasis on context-free and regular languages. It then recapitulates the arguments why neither regular nor context-free grammar is sufficiently expressive to capture all phenomena in the natural language syntax. In the second part, two refinements of the Chomsky hierarchy are reviewed, which are both relevant to the extant research in cognitive science: the mildly context-sensitive languages (which are located between context-free and context-sensitive languages), and the sub-regular hierarchy (which distinguishes several levels of complexity within the class of regular languages)."
                },
                {
                    "title": "Poverty of the Stimulus Revisited",
                    "abstract": "A central goal of modern generative grammar has been to discover invariant properties of human languages that reflect \"the innate schematism of mind that is applied to the data of experience\" and that \"might reasonably be attributed to the organism itself as its contribution to the task of the acquisition of knowledge\" (Chomsky, 1971). Candidates for such invariances include the structure dependence of grammatical rules, and in particular, certain constraints on question formation. Various \"poverty of stimulus\" (POS) arguments suggest that these invariances reflect an innate human endowment, as opposed to common experience: Such experience warrants selection of the grammars acquired only if humans assume, a priori, that selectable grammars respect substantive constraints. Recently, several researchers have tried to rebut these POS arguments. In response, we illustrate why POS arguments remain an important source of support for appeal to a priori structure-dependent constraints on the grammars that humans naturally acquire."
                },
                {
                    "title": "The Use of Predictive Dependencies in Language Learning",
                    "abstract": "To what extent is linguistic structure learnable from statistical information in the input? This research investigated the role played by statistical learning in the acquisition of rudimentary phrase structure. One type of statistical cue which might assist in the discovery of hierarchical phrase structure given serially presented input is the dependencies, or predictive relationships, present between form classes within phrases. In order to determine whether learners can use this statistical information, adult and child participants were exposed to an artificial language which contained predictive dependencies as a cue to phrase structure. The results suggest that humans possess statistical learning mechanisms which may assist in the acquisition of this abstract component of natural language. \u00a9Academic Press 2001"
                },
                {
                    "title": "Statistical Learning by 8-Month-Old Infants",
                    "abstract": "Learners rely on a combination of experience-independent and experience-dependent mechanisms to extract information from the environment. Language acquisition involves both types of mechanisms, but most theorists emphasize the relative importance of experience-independent mechanisms. The present study shows that a fundamental task of language acquisition, segmentation of words from fluent speech, can be accomplished by 8-month-old infants based solely on the statistical relationships between neighboring speech sounds. Moreover, this word segmentation was based on statistical learning from only 2 minutes of exposure, suggesting that infants have access to a powerful mechanism for the computation of statistical properties of the language input."
                },
                {
                    "title": "Physics of Language Models: Part 1, Context-Free Grammar",
                    "abstract": ","
                },
                {
                    "title": "Are Emergent Abilities of Large Language Models a Mirage?",
                    "abstract": "Recent work claims that large language models display emergent abilities, abilities not present in smaller-scale models that are present in larger-scale models. What makes emergent abilities intriguing is two-fold: their sharpness, transitioning seemingly instantaneously from not present to present, and their unpredictability, appearing at seemingly unforeseeable model scales. Here, we present an alternative explanation for emergent abilities: that for a particular task and model family, when analyzing fixed model outputs, emergent abilities appear due to the researcher's choice of metric rather than due to fundamental changes in model behavior with scale. Specifically, nonlinear or discontinuous metrics produce apparent emergent abilities, whereas linear or continuous metrics produce smooth, continuous predictable changes in model performance. We present our alternative explanation in a simple mathematical model, then test it in three complementary ways: we (1) make, test and confirm three predictions on the effect of metric choice using the InstructGPT/GPT-3 family on tasks with claimed emergent abilities; (2) make, test and confirm two predictions about metric choices in a meta-analysis of emergent abilities on BIG-Bench; and (3) show to choose metrics to produce never-before-seen seemingly emergent abilities in multiple vision tasks across diverse deep networks. Via all three analyses, we provide evidence that alleged emergent abilities evaporate with different metrics or with better statistics, and may not be a fundamental property of scaling AI models."
                },
                {
                    "title": "A Path Towards Autonomous Machine Intelligence Version 0.9.2, 2022-06-27",
                    "abstract": "How could machines learn as efficiently as humans and animals? How could machines learn to reason and plan? How could machines learn representations of percepts and action plans at multiple levels of abstraction, enabling them to reason, predict, and plan at multiple time horizons? This position paper proposes an architecture and training paradigms with which to construct autonomous intelligent agents. It combines concepts such as configurable predictive world model, behavior driven through intrinsic motivation, and hierarchical joint embedding architectures trained with self-supervised learning."
                },
                {
                    "title": "The Implications of Local Correlation on Learning Some Deep Functions",
                    "abstract": "It is known that learning deep neural-networks is computationally hard in the worst-case. In fact, the proofs of such hardness results show that even weakly learning deep networks is hard. In other words, no ef\ufb01cient algorithm can \ufb01nd a predictor that is slightly better than a random guess. However, we observe that on natural distributions of images, small patches of the input image are correlated to the target label, which implies that on such natural data, ef\ufb01cient weak learning is trivial. While in the distribution-free setting, the celebrated boosting results show that weak learning implies strong learning, in the distribution-speci\ufb01c setting this is not necessarily the case. We introduce a property of distributions, denoted \u201clocal correlation\u201d, which requires that small patches of the input image and of intermediate layers of the target function are correlated to the target label. We empirically demonstrate that this property holds for the CIFAR and ImageNet data sets. The main technical results of the paper is proving that, for some classes of deep functions, weak learning implies ef\ufb01cient strong learning under the \u201clocal correlation\u201d assumption."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                }
            ],
            "categories": [
                "cs.CL",
                "cond-mat.dis-nn",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow do Large Language Models (LLMs) learn the hierarchical structure of language from training data, and what is the relationship between training set size, correlations, and effective context window in this learning process?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it bridges the gap between formal language theory and empirical observations of LLMs, enhancing our understanding of language acquisition mechanisms. This research could lead to advancements in natural language processing, improving the design of LLMs and their applications in various fields such as education, linguistics, and artificial intelligence. By elucidating how LLMs learn language structure, we can inform future research directions and practical applications, potentially leading to more efficient and effective language models.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem stem from the complexity of language structure and the vast variability in training data. Naive approaches may fail because they do not account for the hierarchical nature of language or the nuanced correlations between tokens that emerge as training data increases. Technical obstacles include accurately modeling the effective context window and understanding the scaling behavior of LLMs, while theoretical challenges involve reconciling different perspectives on language acquisition, such as innate faculties versus statistical learning.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either the theoretical aspects of language structure or the empirical performance of LLMs, leading to a lack of integration between these perspectives. Limitations in existing solutions include insufficient exploration of how training set size influences the learning of deeper language structures and the absence of a unified framework to analyze these relationships. Our approach differs by employing the Random Hierarchy Model to systematically investigate these dynamics, providing a clearer understanding of how LLMs can represent hidden variables as training data scales.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using synthetic datasets generated via the Random Hierarchy Model (RHM) to analyze the learning process of LLMs. We will characterize the power-law decay of correlations between tokens and demonstrate how this decay limits the effective context window as training set size increases. We will empirically validate our predictions regarding sample complexities and the emergence of deeper data structure representations in deep transformers and CNNs. Expected outcomes include a clearer understanding of the relationship between training set size, correlations, and effective context, which may general"
            }
        },
        "author_data": {
            "223772f6-afa8-4538-b234-a96b6d8b1568": {
                "pk": "223772f6-afa8-4538-b234-a96b6d8b1568",
                "name": "Francesco Cagnetta",
                "collaborators": [
                    "Martin R. Evans",
                    "Matthieu Wyart",
                    "Davide Marenduzzo",
                    "Alessandro Favero",
                    "Giuseppe Gonnella",
                    "Leonardo Petrini",
                    "Viktor Skultety",
                    "Umberto M. Tomasini",
                    "Antonio Suma",
                    "Alessandro Mossa"
                ],
                "domain": [
                    "Statistical Mechanics",
                    "Active Matter",
                    "Machine Learning",
                    "Nonlinear Dynamics"
                ],
                "publications": [
                    {
                        "title": "Kinetic roughening in active interfaces",
                        "abstract": "The essential features of many interfaces driven out of equilibrium are described by the same equation---the Kardar-Parisi-Zhang (KPZ) equation. How do living interfaces, such as the cell membrane, fit into this picture? In an endeavour to answer such a question, we proposed in [F. Cagnetta, M. R. Evans, D. Marenduzzo, PRL 120, 258001 (2018)] an idealised model for the membrane of a moving cell. Here we discuss how the addition of simple ingredients inspired by the dynamics of the membrane of moving cells affects common kinetic roughening theories such as the KPZ and Edwards-Wilkinson equations."
                    },
                    {
                        "title": "Inviscid limit of the active interface equations",
                        "abstract": "We present a detailed solution of the active interface equations in the inviscid limit. The active interface equations were previously introduced as a toy model of membrane-protein systems: they describe a stochastic interface where growth is stimulated by inclusions which themselves move on the interface. In the inviscid limit, the equations reduce to a pair of coupled conservation laws. After discussing how the inviscid limit is obtained, we turn to the corresponding Riemann problem: the solution of the set of conservation laws with discontinuous initial condition. In particular, by considering two physically meaningful initial conditions, a giant trough and a giant peak in the interface, we elucidate the generation of shock waves and rarefaction fans in the system. Then, by combining several Riemann problems, we construct an oscillating solution of the active interface with periodic boundaries conditions. The existence of this oscillating state reflects the reciprocal coupling between the two conserved quantities in our system."
                    },
                    {
                        "title": "Locality defeats the curse of dimensionality in convolutional teacher-student scenarios",
                        "abstract": "Convolutional neural networks perform a local and translationally-invariant treatment of the data: quantifying which of these two aspects is central to their success remains a challenge. We study this problem within a teacher-student framework for kernel regression, using `convolutional' kernels inspired by the neural tangent kernel of simple convolutional architectures of given filter size. Using heuristic methods from physics, we find in the ridgeless case that locality is key in determining the learning curve exponent $\\beta$ (that relates the test error $\\epsilon_t\\sim P^{-\\beta}$ to the size of the training set $P$), whereas translational invariance is not. In particular, if the filter size of the teacher $t$ is smaller than that of the student $s$, $\\beta$ is a function of $s$ only and does not depend on the input dimension. We confirm our predictions on $\\beta$ empirically. We conclude by proving, using a natural universality assumption, that performing kernel regression with a ridge that decreases with the size of the training set leads to similar learning curve exponents to those we obtain in the ridgeless case."
                    },
                    {
                        "title": "Universal properties of active membranes",
                        "abstract": "We put forward a general field theory for membranes with embedded activators and analyse their critical properties using renormalization group techniques. Depending on the membrane-activator coupling, we find a crossover between acoustic and diffusive scaling regimes, with mean-field dynamical critical exponents z = 1 and 2 respectively. We argue that the acoustic scaling, which is exact in all spatial dimensions, is a suitable candidate for the universal description of the spatiotemporal patterns observed at the leading edge of motile cells. Furthermore, one-loop corrections to the diffusive mean-field exponents reveal universal behaviour distinct from the Kardar-Parisi-Zhang scaling of passive interfaces and signs of strong-coupling behaviour."
                    },
                    {
                        "title": "Strong anomalous diffusion of the phase of a chaotic pendulum",
                        "abstract": "In this letter we consider the phase diffusion of a harmonically driven undamped pendulum and show that it is anomalous in the strong sense. The role played by the fractal properties of the phase space is highlighted, providing an illustration of the link between deterministic chaos and anomalous transport. Finally, we build a stochastic model which reproduces most properties of the original Hamiltonian system by alternating ballistic flights and random diffusion."
                    },
                    {
                        "title": "Statistical mechanics of a single active slider on a fluctuating interface",
                        "abstract": "We study the statistical mechanics of a single active slider on a fluctuating interface, by means of numerical simulations and theoretical arguments. The slider, which moves by definition towards the interface minima, is active as it also stimulates growth of the interface. Even though such a particle has no counterpart in thermodynamic systems, active sliders may provide a simple model for ATP-dependent membrane proteins that activate cytoskeletal growth. We find a wide range of dynamical regimes according to the ratio between the timescales associated with the slider motion and the interface relaxation. If the interface dynamics is slow, the slider behaves like a random walker in a random envinronment which, furthermore, is able to escape environmental troughs by making them grow. This results in different dynamic exponens to the interface and the particle: the former behaves as an Edward-Wilkinson surface with dynamic exponent 2 whereas the latter has dynamic exponent 3/2. When the interface is fast, we get sustained ballistic motion with the particle surfing a membrane wave created by itself. However, if the interface relaxes immediately (i.e., it is infinitely fast), particle motion becomes symmetric and goes back to diffusive. Due to such a rich phenomenology, we propose the active slider as a toy model of fundamental interest in the field of active membranes and, generally, whenever the system constituent can alter the environment by spending energy."
                    },
                    {
                        "title": "Learning sparse features can lead to overfitting in neural networks",
                        "abstract": "It is widely believed that the success of deep networks lies in their ability to learn a meaningful representation of the features of the data. Yet, understanding when and how this feature learning improves performance remains a challenge: for example, it is beneficial for modern architectures trained to classify images, whereas it is detrimental for fully-connected networks trained for the same task on the same data. Here we propose an explanation for this puzzle, by showing that feature learning can perform worse than lazy training (via random feature kernel or the NTK) as the former can lead to a sparser neural representation. Although sparsity is known to be essential for learning anisotropic data, it is detrimental when the target function is constant or smooth along certain directions of input space. We illustrate this phenomenon in two settings: (i) regression of Gaussian random functions on the d-dimensional unit sphere and (ii) classification of benchmark datasets of images. For (i), we compute the scaling of the generalization error with number of training points, and show that methods that do not learn features generalize better, even when the dimension of the input space is large. For (ii), we show empirically that learning features can indeed lead to sparse and thereby less smooth representations of the image predictors. This fact is plausibly responsible for deteriorating the performance, which is known to be correlated with smoothness along diffeomorphisms."
                    },
                    {
                        "title": "Kernels, Data & Physics",
                        "abstract": "Lecture notes from the course given by Professor Julia Kempe at the summer school \"Statistical physics of Machine Learning\" in Les Houches. The notes discuss the so-called NTK approach to problems in machine learning, which consists of gaining an understanding of generally unsolvable problems by finding a tractable kernel formulation. The notes are mainly focused on practical applications such as data distillation and adversarial robustness, examples of inductive bias are also discussed."
                    },
                    {
                        "title": "A renormalization group study of the dynamics of active membranes: universality classes and scaling laws",
                        "abstract": "Motivated by experimental observations of patterning at the leading edge of motile eukaryotic cells, we introduce a general model for the dynamics of nearly-flat fluid membranes driven from within by an ensemble of activators. We include, in particular, a kinematic coupling between activator density and membrane slope which generically arises whenever the membrane has a non-vanishing normal speed. We unveil the phase diagram of the model by means of a perturbative field-theoretical renormalization group analysis. Due to the aforementioned kinematic coupling the natural dynamical scaling is acoustic, that is the dynamical critical exponent is 1. However, as soon as the the normal velocity of the membrane is tuned to zero, the system crosses over to diffusive dynamic scaling in mean field. Distinct critical points can be reached depending on how the limit of vanishing velocity is realised: in each of them corrections to scaling due to nonlinear coupling terms must be taken into accounts. The detailed analysis of these critical points reveals novel scaling regimes wich can be accessed with perturbative methods, together with signs of strong coupling behaviour, which establishes a promising ground for further non-perturbative calculations. Our results unify several previous studies on the dynamics of active membrane, while also identifying nontrivial scaling regimes which cannot be captured by passive theories of fluctuating interfaces and are relevant for the physics of living membranes."
                    },
                    {
                        "title": "Efficiency of one-dimensional active transport conditioned on motility",
                        "abstract": "By conditioning a stochastic process on the value of an observable, one obtains a new stochastic process with different properties. We apply this idea in the context of active matter, and condition interacting self-propelled particles on their individual motility. Using the effective process formalism from dynamical large deviations theory, we derive the interactions that actuate the imposed mobility against jamming interactions in two toy models---the totally asymmetric exclusion process and run-and-tumble particles, \\emil{in the case of two or three particles}. We provide a framework which takes into account the energy-consumption required for self-propulsion, and address the question of how energy-efficient the emergent interactions are. Upon conditioning, run-and-tumble particles develop an alignment interaction and achieve a higher gain in efficiency than TASEP particles. A point of diminishing returns in efficiency is reached beyond a certain level of conditioning. With recourse to a general formula for the change in energy efficiency upon conditioning, we conclude that the most significant gains occur when there are large fluctuations in mobility to exploit. From a detailed comparison of the emergent effective interaction in a two- versus a three-body scenario, we discover evidence of a screening effect which suggests that conditioning can produce topological rather than metric interactions."
                    },
                    {
                        "title": "How Deep Neural Networks Learn Compositional Data: The Random Hierarchy Model",
                        "abstract": "Deep learning algorithms demonstrate a surprising ability to learn high-dimensional tasks from limited examples. This is commonly attributed to the depth of neural networks, enabling them to build a hierarchy of abstract, low-dimensional data representations. However, how many training examples are required to learn such representations remains unknown. To quantitatively study this question, we introduce the Random Hierarchy Model: a family of synthetic tasks inspired by the hierarchical structure of language and images. The model is a classification task where each class corresponds to a group of high-level features, chosen among several equivalent groups associated with the same class. In turn, each feature corresponds to a group of sub-features chosen among several equivalent ones and so on, following a hierarchy of composition rules. We find that deep networks learn the task by developing internal representations invariant to exchanging equivalent groups. Moreover, the number of data required corresponds to the point where correlations between low-level features and classes become detectable. Overall, our results indicate how deep networks overcome the curse of dimensionality by building invariant representations, and provide an estimate of the number of data required to learn a hierarchical task."
                    },
                    {
                        "title": "Large fluctuations and dynamic phase transition in a system of self-propelled particles",
                        "abstract": "We study the statistics, in stationary conditions, of the work $W_\\tau$ done by the active force in different systems of self-propelled particles in a time $\\tau$. We show the existence of a critical value $W_\\tau ^\\dag$ such that fluctuations with $W_\\tau >W_\\tau ^\\dag$ correspond to configurations where interaction between particles plays a minor role whereas those with $W_\\tau < W_\\tau ^\\dag$ represent states with single particles dragged by clusters. This two-fold behavior is fully mirrored by the probability distribution $P(W_\\tau)$ of the work, which does not obey the large-deviation principle for $W_\\tau <W_\\tau ^\\dag$. This pattern of behavior can be interpreted as due to a phase transition occurring at the level of fluctuating quantities and an order parameter is correspondingly identified."
                    },
                    {
                        "title": "What Can Be Learnt With Wide Convolutional Neural Networks?",
                        "abstract": "Understanding how convolutional neural networks (CNNs) can efficiently learn high-dimensional functions remains a fundamental challenge. A popular belief is that these models harness the local and hierarchical structure of natural data such as images. Yet, we lack a quantitative understanding of how such structure affects performance, e.g., the rate of decay of the generalisation error with the number of training samples. In this paper, we study infinitely-wide deep CNNs in the kernel regime. First, we show that the spectrum of the corresponding kernel inherits the hierarchical structure of the network, and we characterise its asymptotics. Then, we use this result together with generalisation bounds to prove that deep CNNs adapt to the spatial scale of the target function. In particular, we find that if the target function depends on low-dimensional subsets of adjacent input variables, then the decay of the error is controlled by the effective dimensionality of these subsets. Conversely, if the target function depends on the full set of input variables, then the error decay is controlled by the input dimension. We conclude by computing the generalisation error of a deep CNN trained on the output of another deep CNN with randomly-initialised parameters. Interestingly, we find that, despite their hierarchical structure, the functions generated by infinitely-wide deep CNNs are too rich to be efficiently learnable in high dimension."
                    },
                    {
                        "title": "How deep convolutional neural networks lose spatial information with training",
                        "abstract": "A central question of machine learning is how deep nets manage to learn tasks in high dimensions. An appealing hypothesis is that they achieve this feat by building a representation of the data where information irrelevant to the task is lost. For image datasets, this view is supported by the observation that after (and not before) training, the neural representation becomes less and less sensitive to diffeomorphisms acting on images as the signal propagates through the net. This loss of sensitivity correlates with performance, and surprisingly correlates with a gain of sensitivity to white noise acquired during training. These facts are unexplained, and as we demonstrate still hold when white noise is added to the images of the training set. Here, we (i) show empirically for various architectures that stability to image diffeomorphisms is achieved by both spatial and channel pooling, (ii) introduce a model scale-detection task which reproduces our empirical observations on spatial pooling and (iii) compute analitically how the sensitivity to diffeomorphisms and noise scales with depth due to spatial pooling. The scalings are found to depend on the presence of strides in the net architecture. We find that the increased sensitivity to noise is due to the perturbing noise piling up during pooling, after being rectified by ReLU units."
                    },
                    {
                        "title": "Work Fluctuations in the Active Ornstein- Uhlenbeck Particle model",
                        "abstract": "We study the large deviations of the power injected by the active force for an Active Ornstein-Uhlenbeck Particle (AOUP), free or in a confining potential. For the free-particle case, we compute the rate function analytically in d-dimensions from a saddle-point expansion, and numerically in two dimensions by it a) direct sampling of the active work in numerical solutions of the AOUP equations and b) Legendre-Fenchel transform of the scaled cumulant generating function obtained via a cloning algorithm. The rate function presents asymptotically linear branches on both sides and it is independent of the system's dimensionality, apart from a multiplicative factor. For the confining potential case, we focus on two-dimensional systems and obtain the rate function numerically using both methods a) and b). We find a different scenario for harmonic and anharmonic potentials: in the former case, the phenomenology of fluctuations is analogous to that of a free particle, but the rate function might be non-analytic; in the latter case the rate functions are analytic, but fluctuations are realised by entirely different means, which rely strongly on the particle-potential interaction. Finally, we check the validity of a fluctuation relation for the active work distribution. In the free-particle case, the relation is satisfied with a slope proportional to the bath temperature. The same slope is found for the harmonic potential, regardless of activity, and for an anharmonic potential with low activity. In the anharmonic case with high activity, instead, we find a different slope which is equal to an effective temperature obtained from the fluctuation-dissipation theorem."
                    }
                ]
            },
            "3630cf4f-0b56-4ae1-b9f2-bf0e9a81b575": {
                "pk": "3630cf4f-0b56-4ae1-b9f2-bf0e9a81b575",
                "name": "Matthieu Wyart",
                "collaborators": [
                    "Carolina Brito",
                    "Le Yan",
                    "Jie Lin",
                    "Markus M\u00fcller",
                    "Mike Cates",
                    "Eric DeGiuli",
                    "Antonio Sclocchi",
                    "Umberto Tomasini"
                ],
                "domain": [
                    "Amorphous Materials",
                    "Jamming",
                    "Glass Transition",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "On the dependence of the avalanche angle on the granular layer thickness",
                        "abstract": "A layer of sand of thickness h flows down a rough surface if the inclination is larger than some threshold value theta which decreases with h. A tentative microscopic model for the dependence of theta with h is proposed for rigid frictional grains, based on the following hypothesis: (i) a horizontal layer of sand has some coordination z larger than a critical value z_c where mechanical stability is lost (ii) as the tilt angle is increased, the configurations visited present a growing proportion $_s of sliding contacts. Instability with respect to flow occurs when z-z_s=z_c. This criterion leads to a prediction for theta(h) in good agreement with empirical observations."
                    },
                    {
                        "title": "Correlations between vibrational entropy and dynamics in super-cooled liquids",
                        "abstract": "A relation between vibrational entropy and particles mean square displacement is derived in super-cooled liquids, assuming that the main effect of temperature changes is to rescale the vibrational spectrum. Deviations from this relation, in particular due to the presence of a Boson Peak whose shape and frequency changes with temperature, are estimated. Using observations of the short-time dynamics in liquids of various fragility, it is argued that (i) if the crystal entropy is significantly smaller than the liquid entropy at $T_g$, the extrapolation of the vibrational entropy leads to the correlation $T_K\\approx T_0$, where $T_K$ is the Kauzmann temperature and $T_0$ is the temperature extracted from the Vogel-Fulcher fit of the viscosity. (ii) The jump in specific heat associated with vibrational entropy is very small for strong liquids, and increases with fragility. The analysis suggests that these correlations stem from the stiffening of the Boson Peak under cooling, underlying the importance of this phenomenon on the dynamical arrest."
                    },
                    {
                        "title": "Rigidity-based approach to the boson peak in amorphous solids: from sphere packing to amorphous silica",
                        "abstract": "Glasses have a large excess of low-frequency vibrational modes in comparison with continuous elastic body, the so-called Boson Peak, which appears to correlate with several crucial properties of glasses, such as transport or fragility. I review recent results showing that the Boson Peak is a necessary consequence of the weak connectivity of the solid. I explain why in assemblies repulsive spheres the boson peak shifts up to zero frequency as the pressure is lowered toward the jamming threshold, and derive the corresponding exponent. I show how these ideas capture the main low-frequency features of the vibrational spectrum of amorphous silica. These results extend arguments of Phillips on the presence of floppy modes in under-constrained covalent networks to glasses where the covalent network is rigid, or when interactions are purely radial."
                    },
                    {
                        "title": "Elasticity of soft particles and colloids near Random Close Packing",
                        "abstract": "Assemblies of purely repulsive and frictionless particles, such as emulsions or hard spheres, display very curious properties near their jamming transition, which occurs at the random close packing for mono-disperse spheres. Although such systems do not contain the long and cross-linked polymeric chains characterizing a rubber, they behave macroscopically in a similar way: the shear modulus $G$ can become infinitely smaller than the bulk modulus $B$. After reviewing recent theoretical results on the structure of such packing (in particular their coordination) I will propose an explanation for the observed scaling of the elastic moduli, and explain why the arguments both apply to soft and hard particles."
                    },
                    {
                        "title": "Stability at Random Close Packing",
                        "abstract": "The requirement that packings of hard particles, arguably the simplest structural glass, cannot be compressed by rearranging their network of contacts is shown to yield a new constraint on their microscopic structure. This constraint takes the form a bound between the distribution of contact forces P(f) and the pair distribution function g(r): if P(f) \\sim f^{\\theta} and g(r) \\sim (r-{\\sigma})^(-{\\gamma}), where {\\sigma} is the particle diameter, one finds that {\\gamma} \\geq 1/(2+{\\theta}). This bound plays a role similar to those found in some glassy materials with long-range interactions, such as the Coulomb gap in Anderson insulators or the distribution of local fields in mean-field spin glasses. There is ground to believe that this bound is saturated, offering an explanation for the presence of avalanches of rearrangements with power-law statistics observed in packings."
                    },
                    {
                        "title": "On the Rigidity of Amorphous Solids",
                        "abstract": "We poorly understand the properties of amorphous systems at small length scales, where a continuous elastic description breaks down. This is apparent when one considers their vibrational and transport properties, or the way forces propagate in these solids. Little is known about the microscopic cause of their rigidity. Recently it has been observed numerically that an assembly of elastic particles has a critical behavior near the jamming threshold where the pressure vanishes. At the transition such a system does not behave as a continuous medium at any length scales. When this system is compressed, scaling is observed for the elastic moduli, the coordination number, but also for the density of vibrational modes. In the present work we derive theoretically these results, and show that they apply to various systems such as granular matter and silica, but also to colloidal glasses. In particular we show that: (i) these systems present a large excess of vibrational modes at low frequency in comparison with normal solids, called the \"boson peak\" in the glass literature. The corresponding modes are very different from plane waves, and their frequency is related to the system coordination; (ii) rigidity is a non-local property of the packing geometry, characterized by a length scale which can be large. For elastic particles this length diverges near the jamming transition; (iii) for repulsive systems the shear modulus can be much smaller than the bulk modulus. We compute the corresponding scaling laws near the jamming threshold. Finally, we discuss the applications of these results to the glass transition, the transport, and the geometry of the random close packing."
                    },
                    {
                        "title": "Scaling of phononic transport with connectivity in amorphous solids",
                        "abstract": "The effect of coordination on transport is investigated theoretically using random networks of springs as model systems. An effective medium approximation is made to compute the density of states of the vibrational modes, their energy diffusivity (a spectral measure of transport) and their spatial correlations as the network coordination $z$ is varied. Critical behaviors are obtained as $z\\to z_c$ where these networks lose rigidity. A sharp cross-over from a regime where modes are plane-wave-like toward a regime of extended but strongly-scattered modes occurs at some frequency $\\omega^*\\sim z-z_c$, which does not correspond to the Ioffe-Regel criterion. Above $\\omega^*$ both the density of states and the diffusivity are nearly constant. These results agree remarkably with recent numerical observations of repulsive particles near the jamming threshold \\cite{ning}. The analysis further predicts that the length scale characterizing the correlation of displacements of the scattered modes decays as $1/\\sqrt{\\omega}$ with frequency, whereas for $\\omega<<\\omega^*$ Rayleigh scattering is found with a scattering length $l_s\\sim (z-z_c)^3/\\omega^4$. It is argued that this description applies to silica glass where it compares well with thermal conductivity data, and to transverse ultrasound propagation in granular matter."
                    },
                    {
                        "title": "Marginal Stability in Structural, Spin and Electron Glasses",
                        "abstract": "We revisit the concept of marginal stability in glasses, and determine its range of applicability in the context of avalanche-type response to slow external driving. We argue that there is an intimate connection between a pseudo-gap in the distribution of local fields and crackling in systems with long-range interactions. We show how the principle of marginal stability offers a unifying perspective on the phenomenology of systems as diverse as spin and electron glasses, hard spheres, pinned elastic interfaces and the plasticity of soft amorphous solids."
                    },
                    {
                        "title": "Geometric interpretation of pre-vitrification in hard sphere liquids",
                        "abstract": "We derive a microscopic criterion for the stability of hard sphere configurations, and we show empirically that this criterion is marginally satisfied in the glass. This observation supports a geometric interpretation for the initial rapid rise of viscosity with packing fraction, or pre-vitrification. It also implies that barely stable soft modes characterize the glass structure, whose spatial extension is estimated. We show that both the short-term dynamics and activation processes occur mostly along those soft modes, and we study some implications of these observations. This article synthesizes new and previous results [C. Brito and M. Wyart, Euro. Phys. Letters, {\\bf 76}, 149-155, (2006) and C. Brito and M. Wyart, J. Stat. Mech., L08003 (2007) ] in a unified view."
                    },
                    {
                        "title": "Discontinuous shear thickening without inertia in dense non-Brownian suspensions",
                        "abstract": "A consensus is emerging that discontinuous shear thickening (DST) in dense suspensions marks a transition from a flow state where particles remain well separated by lubrication layers, to one dominated by frictional contacts. We show here that reasonable assumptions about contact proliferation predict two distinct types of DST in the absence of inertia. The first occurs at densities above the jamming point of frictional particles; here the thickened state is completely jammed and (unless particles deform) cannot flow without inhomogeneity or fracture. The second regime shows strain- rate hysteresis and arises at somewhat lower densities where the thickened phase flows smoothly. DST is predicted to arise when finite-range repulsions defer contact formation until a characteristic stress level is exceeded."
                    },
                    {
                        "title": "Unifying Suspension and Granular flows near Jamming",
                        "abstract": "Rheological properties of dense flows of hard particles are singular as one approaches the jamming threshold where flow ceases, both for granular flows dominated by inertia, and for over-damped suspensions. Concomitantly, the lengthscale characterizing velocity correlations appears to diverge at jamming. Here we review a theoretical framework that gives a scaling description of stationary flows of frictionless particles. Our analysis applies both to suspensions and inertial flows of hard particles. We report numerical results in support of the theory, and show the phase diagram that results when friction is added, delineating the regime of validity of the frictionless theory."
                    },
                    {
                        "title": "Heterogeneous Dynamics, Marginal Stability and Soft Modes in Hard Sphere Glasses",
                        "abstract": "In a recent publication we established an analogy between the free energy of a hard sphere system and the energy of an elastic network [1]. This result enables one to study the free energy landscape of hard spheres, in particular to define normal modes. In this Letter we use these tools to analyze the activated transitions between meta-bassins, both in the aging regime deep in the glass phase and near the glass transition. We observe numerically that structural relaxation occurs mostly along a very small number of nearly-unstable extended modes. This number decays for denser packing and is significantly lowered as the system undergoes the glass transition. This observation supports that structural relaxation and marginal modes share common properties. In particular theoretical results [2, 3] show that these modes extend at least on some length scale $l^*\\sim (\\phi_c-\\phi)^{-1/2}$ where $\\phi_c$ corresponds to the maximum packing fraction, i.e. the jamming transition. This prediction is consistent with very recent numerical observations of sheared systems near the jamming threshold [4], where a similar exponent is found, and with the commonly observed growth of the rearranging regions with compression near the glass transition."
                    },
                    {
                        "title": "On the rigidity of a hard sphere glass near random close packing",
                        "abstract": "We study theoretically and numerically the microscopic cause of the mechanical stability of hard sphere glasses near their maximum packing. We show that, after coarse-graining over time, the hard sphere interaction can be described by an effective potential which is exactly logarithmic at the random close packing $\\phi_c$. This allows to define normal modes, and to apply recent results valid for elastic networks: mechanical stability is a non-local property of the packing geometry, and is characterized by some length scale $l^*$ which diverges at $\\phi_c$ [1, 2]. We compute the scaling of the bulk and shear moduli near $\\phi_c$, and speculate on the possible implications of these results for the glass transition."
                    },
                    {
                        "title": "Normal modes analysis of the microscopic dynamics in hard discs",
                        "abstract": "We estimate numerically the normal modes of the free energy in a glass of hard discs. We observe that, near the glass transition or after a rapid quench deep in the glass phase, the density of states (i) is characteristic of a marginally stable structure, in particular it di splays a frequency scale $\\omega^*\\sim p^{1/2}$, where $p$ is the pressure and (ii) gives a faithful representation of the short-time dyn amics. This brings further evidences that the boson peak near the glass transition corresponds to the relaxation of marginal modes of a we akly-coordinated structure, and implies that the mean square displacement in the glass phase is anomalously large and goes as $< \\delta R^2 > \\sim p^{-3/2}$, a prediction that we check numerically."
                    },
                    {
                        "title": "Evolution of covalent networks under cooling: contrasting the rigidity window and jamming scenarios",
                        "abstract": "We study the evolution of structural disorder under cooling in supercooled liquids, focusing on covalent networks. We introduce a model for the energy of networks that incorporates weak non-covalent interactions. We show that at low-temperature, these interactions considerably affect the network topology near the rigidity transition that occurs as the coordination increases. As a result, this transition becomes mean-field and does not present a line of critical points previously argued for, the \"rigidity window\". Vibrational modes are then not fractons, but instead are similar to the anomalous modes observed in packings of particles near jamming. These results suggest an alternative interpretation for the intermediate phase observed in chalcogenides."
                    },
                    {
                        "title": "Microscopic processes controlling the Herschel-Bulkley exponent",
                        "abstract": "The flow curve of various yield stress materials is singular as the strain rate vanishes, and can be characterized by the so-called Herschel-Bulkley exponent $n=1/\\beta$. A mean-field approximation due to Hebraud and Lequeux (HL) assumes mechanical noise to be Gaussian, and leads to $\\beta=2$ in rather good agreement with observations. Here we prove that the improved mean-field model where the mechanical noise has fat tails instead leads to $\\beta=1$ with logarithmic correction. This result supports that HL is not a suitable explanation for the value of $\\beta$, which is instead significantly affected by finite dimensional effects. From considerations on elasto-plastic models and on the limitation of speed at which avalanches of plasticity can propagate, we argue that $\\beta=1+1/(d-d_f)$ where $d_f$ is the fractal dimension of avalanches and $d$ the spatial dimension. Measurements of $d_f$ then supports that $\\beta\\approx 2.1$ and $\\beta\\approx 1.7$ in two and three dimensions respectively. We discuss theoretical arguments leading to approximations of $\\beta$ in finite dimensions."
                    },
                    {
                        "title": "Mean-field description of plastic flow in amorphous solids",
                        "abstract": "Failure and flow of amorphous materials are central to various phenomena including earthquakes and landslides. There is accumulating evidence that the yielding transition between a flowing and an arrested phase is a critical phenomenon, but the associated exponents are not understood, even at a mean-field level where the validity of popular models is debated. Here we solve a mean-field model that captures the broad distribution of the mechanical noise generated by plasticity, whose behavior is related to biased L\\'evy flights near an absorbing boundary. We compute the exponent $\\theta$ characterising the density of shear transformation $P(x)\\sim x^{\\theta}$, where $x$ is the stress increment beyond which they yield. We find that after an isotropic thermal quench, $\\theta=1/2$. However, $\\theta$ depends continuously on the applied shear stress, this dependence is not monotonic, and its value at the yield stress is not universal. The model rationalizes previously unexplained observations, and captures reasonably well the value of exponents in three dimensions. Values of exponents in four dimensions are accurately predicted. These results support that it is the true mean-field model that applies in large dimension, and raise fundamental questions on the nature of the yielding transition."
                    },
                    {
                        "title": "Adaptive Elastic Networks as models of supercooled liquids",
                        "abstract": "The thermodynamics and dynamics of supercooled liquids correlate with their elasticity. In particular for covalent networks, the jump of specific heat is small and the liquid is {\\it strong} near the threshold valence where the network acquires rigidity. By contrast, the jump of specific heat and the fragility are large away from this threshold valence. In a previous work [Proc. Natl. Acad. Sci. U.S.A., 110, 6307 (2013)], we could explain these behaviors by introducing a model of supercooled liquids in which local rearrangements interact via elasticity. However, in that model the disorder characterizing elasticity was frozen, whereas it is itself a dynamic variable in supercooled liquids. Here we study numerically and theoretically adaptive elastic network models where polydisperse springs can move on a lattice, thus allowing for the geometry of the elastic network to fluctuate and evolve with temperature. We show numerically that our previous results on the relationship between structure and thermodynamics hold in these models. We introduce an approximation where redundant constraints (highly coordinated regions where the frustration is large) are treated as an ideal gas, leading to analytical predictions that are accurate in the range of parameters relevant for real materials. Overall, these results lead to a description of supercooled liquids, in which the distance to the rigidity transition controls the number of directions in phase space that cost energy and the specific heat."
                    },
                    {
                        "title": "On the different regimes of Stochastic Gradient Descent",
                        "abstract": "Modern deep networks are trained with stochastic gradient descent (SGD) whose key hyperparameters are the number of data considered at each step or batch size $B$, and the step size or learning rate $\\eta$. For small $B$ and large $\\eta$, SGD corresponds to a stochastic evolution of the parameters, whose noise amplitude is governed by the ''temperature'' $T\\equiv \\eta/B$. Yet this description is observed to break down for sufficiently large batches $B\\geq B^*$, or simplifies to gradient descent (GD) when the temperature is sufficiently small. Understanding where these cross-overs take place remains a central challenge. Here, we resolve these questions for a teacher-student perceptron classification model and show empirically that our key predictions still apply to deep networks. Specifically, we obtain a phase diagram in the $B$-$\\eta$ plane that separates three dynamical phases: (i) a noise-dominated SGD governed by temperature, (ii) a large-first-step-dominated SGD and (iii) GD. These different phases also correspond to different regimes of generalization error. Remarkably, our analysis reveals that the batch size $B^*$ separating regimes (i) and (ii) scale with the size $P$ of the training set, with an exponent that characterizes the hardness of the classification problem."
                    },
                    {
                        "title": "How Deep Networks Learn Sparse and Hierarchical Data: the Sparse Random Hierarchy Model",
                        "abstract": "Understanding what makes high-dimensional data learnable is a fundamental question in machine learning. On the one hand, it is believed that the success of deep learning lies in its ability to build a hierarchy of representations that become increasingly more abstract with depth, going from simple features like edges to more complex concepts. On the other hand, learning to be insensitive to invariances of the task, such as smooth transformations for image datasets, has been argued to be important for deep networks and it strongly correlates with their performance. In this work, we aim to explain this correlation and unify these two viewpoints. We show that by introducing sparsity to generative hierarchical models of data, the task acquires insensitivity to spatial transformations that are discrete versions of smooth transformations. In particular, we introduce the Sparse Random Hierarchy Model (SRHM), where we observe and rationalize that a hierarchical representation mirroring the hierarchical model is learnt precisely when such insensitivity is learnt, thereby explaining the strong correlation between the latter and performance. Moreover, we quantify how the sample complexity of CNNs learning the SRHM depends on both the sparsity and hierarchical structure of the task."
                    }
                ]
            }
        }
    },
    "2310.06743": {
        "paper_data": {
            "title": "Geographic Location Encoding with Spherical Harmonics and Sinusoidal Representation Networks",
            "url": "http://arxiv.org/abs/2310.06743v2",
            "arxiv_id": "2310.06743",
            "authors": [
                "Marc Ru\u00dfwurm",
                "Konstantin Klemmer",
                "Esther Rolf",
                "Robin Zbinden",
                "Devis Tuia"
            ],
            "abstract": "Learning representations of geographical space is vital for any machine learning model that integrates geolocated data, spanning application domains such as remote sensing, ecology, or epidemiology. Recent work embeds coordinates using sine and cosine projections based on Double Fourier Sphere (DFS) features. These embeddings assume a rectangular data domain even on global data, which can lead to artifacts, especially at the poles. At the same time, little attention has been paid to the exact design of the neural network architectures with which these functional embeddings are combined. This work proposes a novel location encoder for globally distributed geographic data that combines spherical harmonic basis functions, natively defined on spherical surfaces, with sinusoidal representation networks (SirenNets) that can be interpreted as learned Double Fourier Sphere embedding. We systematically evaluate positional embeddings and neural network architectures across various benchmarks and synthetic evaluation datasets. In contrast to previous approaches that require the combination of both positional encoding and neural networks to learn meaningful representations, we show that both spherical harmonics and sinusoidal representation networks are competitive on their own but set state-of-the-art performances across tasks when combined. The model code and experiments are available at https://github.com/marccoru/locationencoder.",
            "introduction": "   1 Introduction  Location information is informative meta-data in many geospatial applications, ranging from species distribution modeling (Mac\u00a0Aodha et\u00a0al., 2019; Cole et\u00a0al., 2023) and wildlife monitoring (Jeantet & Dufourq, 2023) to solar irradiance forecasting (Boussif et\u00a0al., 2023), global tree height prediction (Lang et\u00a0al., 2022) or the estimation of housing prices (Klemmer et\u00a0al., 2023). Coordinates also serve as a task-descriptive feature for meta-learning across geographically distributed tasks (Tseng et\u00a0al., 2022) or as input for geographic question answering (Mai et\u00a0al., 2020a). Location encoders alone can be used to learn an implicit neural representation of a dataset (Stanley, 2007), for example, to compress weather data (Huang & Hoefler, 2023) or to learn an implicit neural representation of species distribution maps (Cole et\u00a0al., 2023).   Methodologically, predictive modeling with spatial coordinates as inputs has a long history and is classically done with non-parametric approaches, like Gaussian Processes (GP), i.e., Kriging, (Oliver & Webster, 1990; Matheron, 1969) or kernel methods in general. For instance, Berg et\u00a0al. (2014) proposed adaptive kernels for bird species recognition, where a class-specific kernel is learned for each species separately. Relying on spatially close training points is intuitive and highly effective for interpolation tasks and small datasets. Still, it becomes computationally expensive when kernel similarities have to be computed between many test points and large training sets. And while research in scaling kernel methods to higher-dimensional data has progressed (Gardner et\u00a0al., 2018), recent focus on scalable methods has shifted to using neural network weights to approximate the training points (Tang et\u00a0al., 2015; Mac\u00a0Aodha et\u00a0al., 2019; Chu et\u00a0al., 2019; Huang & Hoefler, 2023). Inspired by Double Fourier Sphere (DFS) (Orszag, 1974) and the success of positional encodings in transformers, a series of works developed and compared different sine-cosine embedding functions at different spatial scales (frequencies) where the number of scales controls the smoothness of the interpolation (Mai et\u00a0al., 2020b; 2023b). Even though these DFS-inspired approaches have proven effective, they still assume a rectangular domain of longitude and latitude coordinates, which does not accurately reflect our planet\u2019s spherical geometry.   In this work, we instead propose to use spherical harmonic basis functions as positional embeddings that are well-defined across the entire globe, including the poles. In particular, the scientific use of location encoders requires good global performance, with poles being especially important for climate science. Methodologically, we build upon recent work here has been focused on learning functions on general manifolds (Koestler et\u00a0al., 2022) and graphs (Grattarola & Vandergheynst, 2022) in non-euclidean geometries. This work focuses on spheres, a special case of non-euclidean geometries, where we utilize spherical harmonic basis functions, which are the eigenfunctions of the Laplace-Beltrami operator on the sphere (Koestler et\u00a0al., 2022)   In summary, we   1.  propose Spherical Harmonic (SH) coordinate embeddings that work well with all tested neural networks, and especially well when paired with Sinusoidal Representation Networks (SirenNet) (Sitzmann et\u00a0al., 2020), which was untested for location encoding previously;    2.  show by analytic comparison that SirenNets can be seen as learned Double Fourier Sphere embedding, which explains its good performance even with direct latitude/longitude inputs;    3.  provide a comprehensive comparative study using the cross-product of positional embedding strategies and neural networks across diverse synthetic and real-world prediction tasks, including interpolating ERA5",
            "references": [
                {
                    "title": "Spatial Implicit Neural Representations for Global-Scale Species Mapping",
                    "abstract": "Estimating the geographical range of a species from sparse observations is a challenging and important geospatial prediction problem. Given a set of locations where a species has been observed, the goal is to build a model to predict whether the species is present or absent at any location. This problem has a long history in ecology, but traditional methods struggle to take advantage of emerging large-scale crowdsourced datasets which can include tens of millions of records for hundreds of thousands of species. In this work, we use Spatial Implicit Neural Representations (SINRs) to jointly estimate the geographical range of 47k species simultaneously. We find that our approach scales gracefully, making increasingly better predictions as we increase the number of species and the amount of data per species when training. To make this problem accessible to machine learning researchers, we provide four new benchmarks that measure different aspects of species range estimation and spatial representation learning. Using these benchmarks, we demonstrate that noisy and biased crowdsourced data can be combined with implicit neural representations to approximate expert-developed range maps for many species."
                },
                {
                    "title": "CSP: Self-Supervised Contrastive Spatial Pre-Training for Geospatial-Visual Representations",
                    "abstract": "Geo-tagged images are publicly available in large quantities, whereas labels such as object classes are rather scarce and expensive to collect. Meanwhile, contrastive learning has achieved tremendous success in various natural image and language tasks with limited labeled data. However, existing methods fail to fully leverage geospatial information, which can be paramount to distinguishing objects that are visually similar. To directly leverage the abundant geospatial information associated with images in pre-training, fine-tuning, and inference stages, we present Contrastive Spatial Pre-Training (CSP), a self-supervised learning framework for geo-tagged images. We use a dual-encoder to separately encode the images and their corresponding geo-locations, and use contrastive objectives to learn effective location representations from images, which can be transferred to downstream supervised tasks such as image classification. Experiments show that CSP can improve model performance on both iNat2018 and fMoW datasets. Especially, on iNat2018, CSP significantly boosts the model performance with 10-34% relative improvement with various labeled training data sampling ratios."
                },
                {
                    "title": "Lightweight, Pre-trained Transformers for Remote Sensing Timeseries",
                    "abstract": "Machine learning methods for satellite data have a range of societally relevant applications, but labels used to train models can be difficult or impossible to acquire. Self-supervision is a natural solution in settings with limited labeled data, but current self-supervised models for satellite data fail to take advantage of the characteristics of that data, including the temporal dimension (which is critical for many applications, such as monitoring crop growth) and availability of data from many complementary sensors (which can significantly improve a model's predictive performance). We present Presto (the Pretrained Remote Sensing Transformer), a model pre-trained on remote sensing pixel-timeseries data. By designing Presto specifically for remote sensing data, we can create a significantly smaller but performant model. Presto excels at a wide variety of globally distributed remote sensing tasks and performs competitively with much larger models while requiring far less compute. Presto can be used for transfer learning or as a feature extractor for simple models, enabling efficient deployment at scale."
                },
                {
                    "title": "Compressing multidimensional weather and climate data into neural networks",
                    "abstract": "Weather and climate simulations produce petabytes of high-resolution data that are later analyzed by researchers in order to understand climate change or severe weather. We propose a new method of compressing this multidimensional weather and climate data: a coordinate-based neural network is trained to overfit the data, and the resulting parameters are taken as a compact representation of the original grid-based data. While compression ratios range from 300x to more than 3,000x, our method outperforms the state-of-the-art compressor SZ3 in terms of weighted RMSE, MAE. It can faithfully preserve important large scale atmosphere structures and does not introduce artifacts. When using the resulting neural network as a 790x compressed dataloader to train the WeatherBench forecasting model, its RMSE increases by less than 2%. The three orders of magnitude compression democratizes access to high-resolution climate data and enables numerous new research directions."
                },
                {
                    "title": "Generalised Implicit Neural Representations",
                    "abstract": "We consider the problem of learning implicit neural representations (INRs) for signals on non-Euclidean domains. In the Euclidean case, INRs are trained on a discrete sampling of a signal over a regular lattice. Here, we assume that the continuous signal exists on some unknown topological space from which we sample a discrete graph. In the absence of a coordinate system to identify the sampled nodes, we propose approximating their location with a spectral embedding of the graph. This allows us to train INRs without knowing the underlying continuous domain, which is the case for most graph signals in nature, while also making the INRs independent of any choice of coordinate system. We show experiments with our method on various real-world signals on non-Euclidean domains."
                },
                {
                    "title": "Intrinsic Neural Fields: Learning Functions on Manifolds",
                    "abstract": "Neural fields have gained significant attention in the computer vision community due to their excellent performance in novel view synthesis, geometry reconstruction, and generative modeling. Some of their advantages are a sound theoretic foundation and an easy implementation in current deep learning frameworks. While neural fields have been applied to signals on manifolds, e.g., for texture reconstruction, their representation has been limited to extrinsically embedding the shape into Euclidean space. The extrinsic embedding ignores known intrinsic manifold properties and is inflexible wrt. transfer of the learned function. To overcome these limitations, this work introduces intrinsic neural fields, a novel and versatile representation for neural fields on manifolds. Intrinsic neural fields combine the advantages of neural fields with the spectral properties of the Laplace-Beltrami operator. We show theoretically that intrinsic neural fields inherit many desirable properties of the extrinsic neural field framework but exhibit additional intrinsic qualities, like isometry invariance. In experiments, we show intrinsic neural fields can reconstruct high-fidelity textures from images with state-of-the-art quality and are robust to the discretization of the underlying manifold. We demonstrate the versatility of intrinsic neural fields by tackling various applications: texture transfer between deformed shapes&different shapes, texture reconstruction from real-world images with view dependence, and discretization-agnostic learning on meshes and point clouds."
                },
                {
                    "title": "TIML: Task-Informed Meta-Learning for Agriculture",
                    "abstract": "Labeled datasets for agriculture are extremely spatially imbalanced. When developing algorithms for data-sparse regions, a natural approach is to use transfer learning from data-rich regions. While standard transfer learning approaches typically leverage only direct inputs and outputs, geospatial imagery and agricultural data are rich in metadata that can inform transfer learning algorithms, such as the spatial coordinates of data-points or the class of task being learned. We build on previous work exploring the use of meta-learning for agricultural contexts in data-sparse regions and introduce task-informed meta-learning (TIML), an augmentation to model-agnostic meta-learning which takes advantage of task-specific metadata. We apply TIML to crop type classification and yield estimation, and find that TIML significantly improves performance compared to a range of benchmarks in both contexts, across a diversity of model architectures. While we focus on tasks from agriculture, TIML could offer benefits to any meta-learning setup with task-specific metadata, such as classification of geo-tagged images and species distribution modelling."
                },
                {
                    "title": "Positional Encoder Graph Neural Networks for Geographic Data",
                    "abstract": "Graph neural networks (GNNs) provide a powerful and scalable solution for modeling continuous spatial data. However, they often rely on Euclidean distances to construct the input graphs. This assumption can be improbable in many real-world settings, where the spatial structure is more complex and explicitly non-Euclidean (e.g., road networks). Here, we propose PE-GNN, a new framework that incorporates spatial context and correlation explicitly into the models. Building on recent advances in geospatial auxiliary task learning and semantic spatial embeddings, our proposed method (1) learns a context-aware vector encoding of the geographic coordinates and (2) predicts spatial autocorrelation in the data in parallel with the main task. On spatial interpolation and regression tasks, we show the effectiveness of our approach, improving performance over different state-of-the-art GNN approaches. We observe that our approach not only vastly improves over the GNN baselines, but can match Gaussian processes, the most commonly utilized method for spatial interpolation problems."
                },
                {
                    "title": "A review of location encoding for GeoAI: methods and applications",
                    "abstract": "ABSTRACT A common need for artificial intelligence models in the broader geoscience is to encode various types of spatial data, such as points, polylines, polygons, graphs, or rasters, in a hidden embedding space so that they can be readily incorporated into deep learning models. One fundamental step is to encode a single point location into an embedding space, such that this embedding is learning-friendly for downstream machine learning models. We call this process location encoding. However, there lacks a systematic review on location encoding, its potential applications, and key challenges that need to be addressed. This paper aims to fill this gap. We first provide a formal definition of location encoding, and discuss the necessity of it for GeoAI research. Next, we provide a comprehensive survey about the current landscape of location encoding research. We classify location encoding models into different categories based on their inputs and encoding methods, and compare them based on whether they are parametric, multi-scale, distance preserving, and direction aware. We demonstrate that existing location encoders can be unified under one formulation framework. We also discuss the application of location encoding. Finally, we point out several challenges that need to be solved in the future."
                },
                {
                    "title": "Implicit Neural Representations with Periodic Activation Functions",
                    "abstract": "Implicitly defined, continuous, differentiable signal representations parameterized by neural networks have emerged as a powerful paradigm, offering many possible benefits over conventional representations. However, current network architectures for such implicit neural representations are incapable of modeling signals with fine detail, and fail to represent a signal's spatial and temporal derivatives, despite the fact that these are essential to many physical signals defined implicitly as the solution to partial differential equations. We propose to leverage periodic activation functions for implicit neural representations and demonstrate that these networks, dubbed sinusoidal representation networks or Sirens, are ideally suited for representing complex natural signals and their derivatives. We analyze Siren activation statistics to propose a principled initialization scheme and demonstrate the representation of images, wavefields, video, sound, and their derivatives. Further, we show how Sirens can be leveraged to solve challenging boundary value problems, such as particular Eikonal equations (yielding signed distance functions), the Poisson equation, and the Helmholtz and wave equations. Lastly, we combine Sirens with hypernetworks to learn priors over the space of Siren functions."
                },
                {
                    "title": "SE\u2010KGE: A location\u2010aware Knowledge Graph Embedding model for Geographic Question Answering and Spatial Semantic Lifting",
                    "abstract": "Learning knowledge graph (KG) embeddings is an emerging technique for a variety of downstream tasks such as summarization, link prediction, information retrieval, and question answering. However, most existing KG embedding models neglect space and, therefore, do not perform well when applied to (geo)spatial data and tasks. Most models that do consider space primarily rely on some notions of distance. These models suffer from higher computational complexity during training while still losing information beyond the relative distance between entities. In this work, we propose a location\u2010aware KG embedding model called SE\u2010KGE. It directly encodes spatial information such as point coordinates or bounding boxes of geographic entities into the KG embedding space. The resulting model is capable of handling different types of spatial reasoning. We also construct a geographic knowledge graph as well as a set of geographic query\u2013answer pairs called DBGeo to evaluate the performance of SE\u2010KGE in comparison to multiple baselines. Evaluation results show that SE\u2010KGE outperforms these baselines on the DBGeo data set for the geographic logic query answering task. This demonstrates the effectiveness of our spatially\u2010explicit model and the importance of considering the scale of different geographic entities. Finally, we introduce a novel downstream task called spatial semantic lifting which links an arbitrary location in the study area to entities in the KG via some relations. Evaluation on DBGeo shows that our model outperforms the baseline by a substantial margin."
                },
                {
                    "title": "Multi-Scale Representation Learning for Spatial Feature Distributions using Grid Cells",
                    "abstract": "Unsupervised text encoding models have recently fueled substantial progress in Natural Language Processing (NLP). The key idea is to use neural networks to convert words in texts to vector space representations (embeddings) based on word positions in a sentence and their contexts. We see a strikingly similar situation in spatial analysis, which focuses on incorporating both absolute positions and spatial contexts of geographic objects such as Points of Interest (POIs) into models. A general space encoding method is valuable for a multitude of tasks such asPOI search, land use classification, point-based spatial interpolation and locationaware image classification. However, no such general model exists to date beyond simply applying discretizing or feed forward nets to coordinates, and little effort has been put into jointly modeling distributions with vastly different characteristics, which commonly emerges from GIS data. Meanwhile, Nobel Prize-winning Neuroscience research shows that grid cells in mammals provide a multi-scale periodic representation that functions as a metric for encoding space and are critical for recognizing places and for path-integration. Inspired by this research, wepropose a representation learning model called Space2vec to encode the absolutepositions and spatial relationships of places. We conduct experiments on realworld geographic data and predict types of POIs at given positions based on their1) locations and 2) nearby POIs. Results show that because of its multi-scale representations Space2vec outperforms well-established ML approaches such as RBF kernels, multi-layer feed forward nets, and tile embedding approaches."
                },
                {
                    "title": "Optuna: A Next-generation Hyperparameter Optimization Framework",
                    "abstract": "The purpose of this study is to introduce new design-criteria for next-generation hyperparameter optimization software. The criteria we propose include (1) define-by-run API that allows users to construct the parameter search space dynamically, (2) efficient implementation of both searching and pruning strategies, and (3) easy-to-setup, versatile architecture that can be deployed for various purposes, ranging from scalable distributed computing to light-weight experiment conducted via interactive interface. In order to prove our point, we will introduce Optuna, an optimization software which is a culmination of our effort in the development of a next generation optimization software. As an optimization software designed with define-by-run principle, Optuna is particularly the first of its kind. We will present the design-techniques that became necessary in the development of the software that meets the above criteria, and demonstrate the power of our new design through experimental results and real world applications. Our software is available under the MIT license (https://github.com/pfnet/optuna/)."
                },
                {
                    "title": "Presence-Only Geographical Priors for Fine-Grained Image Classification",
                    "abstract": "Appearance information alone is often not sufficient to accurately differentiate between fine-grained visual categories. Human experts make use of additional cues such as where, and when, a given image was taken in order to inform their final decision. This contextual information is readily available in many online image collections but has been underutilized by existing image classifiers that focus solely on making predictions based on the image contents. We propose an efficient spatio-temporal prior, that when conditioned on a geographical location and time, estimates the probability that a given object category occurs at that location. Our prior is trained from presence-only observation data and jointly models object categories, their spatio-temporal distributions, and photographer biases. Experiments performed on multiple challenging image classification datasets show that combining our prior with the predictions from image classifiers results in a large improvement in final classification performance."
                },
                {
                    "title": "Geo-Aware Networks for Fine-Grained Recognition",
                    "abstract": "Fine-grained recognition distinguishes among categories with subtle visual differences. In order to differentiate between these challenging visual categories, it is helpful to leverage additional information. Geolocation is a rich source of additional information that can be used to improve fine-grained classification accuracy, but has been understudied. Our contributions to this field are twofold. First, to the best of our knowledge, this is the first paper which systematically examined various ways of incorporating geolocation information into fine-grained image classification through the use of geolocation priors, post-processing or feature modulation. Secondly, to overcome the situation where no fine-grained dataset has complete geolocation information, we release two fine-grained datasets with geolocation by providing complementary information to existing popular datasets - iNaturalist and YFCC100M. By leveraging geolocation information we improve top-1 accuracy in iNaturalist from 70.1% to 79.0% for a strong baseline image-only model. Comparing several models, we found that best performance was achieved by a post-processing model that consumed the output of the image-only baseline alongside geolocation. However, for a resource-constrained model (MobileNetV2), performance was better with a feature modulation model that trains jointly over pixels and geolocation: accuracy increased from 59.6% to 72.2%. Our work makes a strong case for incorporating geolocation information in fine-grained recognition models for both server and on-device."
                },
                {
                    "title": "GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration",
                    "abstract": "Despite advances in scalable models, the inference tools used for Gaussian processes (GPs) have yet to fully capitalize on developments in computing hardware. We present an efficient and general approach to GP inference based on Blackbox Matrix-Matrix multiplication (BBMM). BBMM inference uses a modified batched version of the conjugate gradients algorithm to derive all terms for training and inference in a single call. BBMM reduces the asymptotic complexity of exact GP inference from $O(n^3)$ to $O(n^2)$. Adapting this algorithm to scalable approximations and complex GP models simply requires a routine for efficient matrix-matrix multiplication with the kernel and its derivative. In addition, BBMM uses a specialized preconditioner to substantially speed up convergence. In experiments we show that BBMM effectively uses GPU hardware to dramatically accelerate both exact GP inference and scalable approximations. Additionally, we provide GPyTorch, a software platform for scalable GP inference via BBMM, built on PyTorch."
                },
                {
                    "title": "The iNaturalist Species Classification and Detection Dataset",
                    "abstract": "Existing image classification datasets used in computer vision tend to have a uniform distribution of images across object categories. In contrast, the natural world is heavily imbalanced, as some species are more abundant and easier to photograph than others. To encourage further progress in challenging real world conditions we present the iNaturalist species classification and detection dataset, consisting of 859,000 images from over 5,000 different species of plants and animals. It features visually similar species, captured in a wide variety of situations, from all over the world. Images were collected with different camera types, have varying image quality, feature a large class imbalance, and have been verified by multiple citizen scientists. We discuss the collection of the dataset and present extensive baseline experiments using state-of-the-art computer vision classification and detection models. Results show that current non-ensemble based methods achieve only 67% top one classification accuracy, illustrating the difficulty of the dataset. Specifically, we observe poor results for classes with small numbers of training examples suggesting more attention is needed in low-shot learning."
                },
                {
                    "title": "A Simple Yet Effective Baseline for 3d Human Pose Estimation",
                    "abstract": "Following the success of deep convolutional networks, state-of-the-art methods for 3d human pose estimation have focused on deep end-to-end systems that predict 3d joint locations given raw image pixels. Despite their excellent performance, it is often not easy to understand whether their remaining error stems from a limited 2d pose (visual) understanding, or from a failure to map 2d poses into 3- dimensional positions.,,With the goal of understanding these sources of error, we set out to build a system that given 2d joint locations predicts 3d positions. Much to our surprise, we have found that, with current technology, \"lifting\" ground truth 2d joint locations to 3d space is a task that can be solved with a remarkably low error rate: a relatively simple deep feedforward network outperforms the best reported result by about 30% on Human3.6M, the largest publicly available 3d pose estimation benchmark. Furthermore, training our system on the output of an off-the-shelf state-of-the-art 2d detector (i.e., using images as input) yields state of the art results \u2013 this includes an array of systems that have been trained end-to-end specifically for this task. Our results indicate that a large portion of the error of modern deep 3d pose estimation systems stems from their visual analysis, and suggests directions to further advance the state of the art in 3d human pose estimation."
                },
                {
                    "title": "SymPy: Symbolic computing in Python",
                    "abstract": "SymPy is an open source computer algebra system written in pure Python. It is built with a focus on extensibility and ease of use, through both interactive and programmatic applications. These characteristics have led SymPy to become the standard symbolic library for the scientific Python ecosystem. This paper presents the architecture of SymPy, a description of its features, and a discussion of select domain specific submodules. The supplementary materials provide additional examples and further outline details of the architecture and features of SymPy."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Improving Image Classification with Location Context",
                    "abstract": "With the widespread availability of cellphones and cameras that have GPS capabilities, it is common for images being uploaded to the Internet today to have GPS coordinates associated with them. In addition to research that tries to predict GPS coordinates from visual features, this also opens up the door to problems that are conditioned on the availability of GPS coordinates. In this work, we tackle the problem of performing image classification with location context, in which we are given the GPS coordinates for images in both the train and test phases. We explore different ways of encoding and extracting features from the GPS coordinates, and show how to naturally incorporate these features into a Convolutional Neural Network (CNN), the current state-of-the-art for most image classification and recognition problems. We also show how it is possible to simultaneously learn the optimal pooling radii for a subset of our features within the CNN framework. To evaluate our model and to help promote research in this area, we identify a set of location-sensitive concepts and annotate a subset of the Yahoo Flickr Creative Commons 100M dataset that has GPS coordinates with these concepts, which we make publicly available. By leveraging location context, we are able to achieve almost a 7% gain in mean average precision."
                },
                {
                    "title": "Birdsnap: Large-Scale Fine-Grained Visual Categorization of Birds",
                    "abstract": "We address the problem of large-scale fine-grained visual categorization, describing new methods we have used to produce an online field guide to 500 North American bird species. We focus on the challenges raised when such a system is asked to distinguish between highly similar species of birds. First, we introduce \"one-vs-most classifiers.\" By eliminating highly similar species during training, these classifiers achieve more accurate and intuitive results than common one-vs-all classifiers. Second, we show how to estimate spatio-temporal class priors from observations that are sampled at irregular and biased locations. We show how these priors can be used to significantly improve performance. We then show state-of-the-art recognition performance on a new, large dataset that we make publicly available. These recognition methods are integrated into the online field guide, which is also publicly available."
                },
                {
                    "title": "Fibonacci grids: A novel approach to global modelling",
                    "abstract": "Recent years have seen a resurgence of interest in a variety of non\u2010standard computational grids for global numerical prediction. The motivation has been to reduce problems associated with the converging meridians and the polar singularities of conventional regular latitude\u2013longitude grids. A further impetus has come from the adoption of massively parallel computers, for which it is necessary to distribute work equitably across the processors; this is more practicable for some non\u2010standard grids. Desirable attributes of a grid for high\u2010order spatial finite differencing are: (i) geometrical regularity; (ii) a homogeneous and approximately isotropic spatial resolution; (iii) a low proportion of the grid points where the numerical procedures require special customization (such as near coordinate singularities or grid edges); (iv) ease of parallelization."
                },
                {
                    "title": "Kriging: a method of interpolation for geographical information systems",
                    "abstract": "Geographical information systems could be improved by adding procedures for geostatistical spatial analysis to existing facilities. Most traditional methods of interpolation are based on mathematical as distinct from stochastic models of spatial variation. Spatially distributed data behave more like random variables, however, and regionalized variable theory provides a set of stochastic methods for analysing them. Kriging is the method of interpolation deriving from regionalized variable theory. It depends on expressing spatial variation of the property in terms of the variogram, and it minimizes the prediction errors which are themselves estimated. We describe the procedures and the way we link them using standard operating systems. We illustrate them using examples from case studies, one involving the mapping and control of soil salinity in the Jordan Valley of Israel, the other in semi-arid Botswana where the herbaceous cover was estimated and mapped from aerial photographic survey."
                },
                {
                    "title": "Saturn's Magnetic Field and Magnetosphere",
                    "abstract": "The Pioneer Saturn vector helium magnetometer has detected a bow shock and magnetopause at Saturn and has provided an accurate characterization of the planetary field. The equatorial surface field is 0.20 gauss, a factor of 3 to 5 times smaller than anticipated on the basis of attempted scalings from Earth and Jupiter. The tilt angle between the magnetic dipole axis and Saturn's rotation axis is < 1\ufffd, a surprisingly small value. Spherical harmonic analysis of the measurements shows that the ratio of quadrupole to dipole moments is < 10 percent, indicating that the field is more uniform than those of the Earth or Jupiter and consistent with Saturn having a relatively small core. The field in the outer magnetosphere shows systematic departures from the dipole field, principally a compression of the field near noon and an equatorial orientation associated with a current sheet near dawn. A hydromagnetic wake resulting from the interaction of Titan with the rotating magnetosphere appears to have been observed."
                },
                {
                    "title": "Fourier series on spheres",
                    "abstract": "Abstract A spectral representation consisting of a two\u2010dimensional Fourier series for use on a sphere is described. The method is applied to the advection of a passive scalar field over the poles and is compared to the pseudo\u2010spectral and grid\u2010point representations. The results show that the double Fourier series method compares favourably with both the pseudo\u2010spectral and grid\u2010point schemes."
                },
                {
                    "title": "What if We Enrich day-ahead Solar Irradiance Time Series Forecasting with Spatio-Temporal Context?",
                    "abstract": "The global integration of solar power into the electrical grid could have a crucial impact on climate change mitigation, yet poses a challenge due to solar irradiance variability. We present a deep learning architecture which uses spatio-temporal context from satellite data for highly accurate day-ahead time-series forecasting, in particular Global Horizontal Irradiance (GHI). We provide a multi-quantile variant which outputs a prediction interval for each time-step, serving as a measure of forecasting uncertainty. In addition, we suggest a testing scheme that separates easy and difficult scenarios, which appears useful to evaluate model performance in varying cloud conditions. Our approach exhibits robust performance in solar irradiance forecasting, including zero-shot generalization tests at unobserved solar stations, and holds great promise in promoting the effective use of solar power and the resulting reduction of CO 2 emissions."
                },
                {
                    "title": "Dropout: a simple way to prevent neural networks from overfitting",
                    "abstract": "Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units (along with their connections) from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different \"thinned\" networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights. This significantly reduces overfitting and gives major improvements over other regularization methods. We show that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets."
                },
                {
                    "title": "Compositional Pattern Producing Networks : A Novel Abstraction of Development",
                    "abstract": "Natural DNA can encode complexity on an enormous scale. Researchers are attempting to achieve the same representational efficiency in computers by implementing developmental encodings , i.e. encodings that map the genotype to the phenotype through a process of growth from a small starting point to a mature form. A major challenge in in this effort is to find the right level of abstraction of biological development to capture its essential properties without introducing unnecessary inefficiencies. In this paper, a novel abstraction of natural development, called Compositional Pattern Producing Networks (CPPNs), is proposed. Unlike currently accepted abstractions such as iterative rewrite systems and cellular growth simulations, CPPNs map to the phenotype without local interaction, that is, each individual component of the phenotype is determined independently of every other component. Results produced with CPPNs through interactive evolution of two-dimensional images show that such an encoding can nevertheless produce structural motifs often attributed to more conventional developmental abstractions, suggesting that local interaction may not be essential to the desirable properties of natural encoding in the way that is usually assumed."
                },
                {
                    "title": "Spherical Harmonic Lighting: The Gritty Details",
                    "abstract": "Spherical Harmonic lighting (SH lighting) is a technique for calculating the lighting on 3D models from area light sources that allows us to capture, relight and display global illumination style images in real time. It was introduced in a paper at Siggraph 2002 by Sloan, Kautz and Snyder as a technique for ultra realistic lighting of models. Looking a little closer at it\u2019s derivation we can show that it is in fact a toolbox of interrelated techniques that the games community can use to good effect."
                },
                {
                    "title": "The pseudospectral approximation applied to the shallow water equations on a sphere",
                    "abstract": "Abstract An algorithm for the application of the pseudospectral method to the nuermerical integration of the shallow water equations has been developed and tested on the case of steady\u2010geostrophic ..."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the accuracy and efficiency of geospatial predictive modeling by utilizing spherical harmonic basis functions as positional embeddings instead of traditional latitude and longitude coordinates?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of existing geospatial modeling techniques that rely on rectangular coordinate systems, which do not accurately represent the Earth's spherical geometry. By developing a method that leverages spherical harmonic embeddings, we can enhance the performance of various applications, such as climate science, wildlife monitoring, and urban planning. This advancement could lead to more accurate predictions and analyses, ultimately influencing future research directions in geospatial data science and machine learning, as well as practical applications in environmental management and resource allocation.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexities of modeling functions on non-Euclidean geometries, particularly the sphere. Traditional methods, such as Gaussian Processes, become computationally expensive with large datasets, and naive approaches that apply standard positional encodings may fail to capture the nuances of spherical data. Additionally, accurately representing the poles and ensuring global performance across diverse geospatial tasks adds further complexity. Overcoming these technical obstacles requires a deep understanding of both the mathematical properties of spherical harmonics and the integration of these embeddings into neural network architectures.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on rectangular coordinate systems, which do not account for the Earth's spherical shape, leading to limitations in accuracy and applicability. Existing solutions have not effectively addressed the unique challenges posed by spherical data, such as the representation of poles and the need for global performance. Additionally, the lack of exploration into the use of spherical harmonic basis functions in conjunction with modern neural network architectures has created a gap in the literature. Our approach differs by directly applying spherical harmonics as positional embeddings and demonstrating their effectiveness across various neural network models, including previously untested combinations.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the use of spherical harmonic basis functions as positional embeddings for neural networks, particularly focusing on their integration with Sinusoidal Representation Networks (SirenNet). We will conduct a comprehensive comparative study using diverse synthetic and real-world datasets, including ERA5 for interpolation tasks. The metrics for evaluation will include prediction accuracy and computational efficiency. We expect that our approach will"
            }
        },
        "author_data": {
            "4d7d424d-4d46-4704-a212-c22872ebf057": {
                "pk": "4d7d424d-4d46-4704-a212-c22872ebf057",
                "name": "Marc Ru\u00dfwurm",
                "collaborators": [
                    "D. Tuia",
                    "L. Hughes",
                    "Jan Pisl",
                    "C. Dumitru",
                    "Xiao Xiang Zhu",
                    "Robin Zbinden",
                    "Nina Van Tiel",
                    "Veronica Tollenaar",
                    "H. Zekollari",
                    "F. Pattyn",
                    "B. Kellenberger",
                    "S. Lhermitte",
                    "Maaike Izeboud",
                    "Lloyd Haydn Hughes",
                    "Gaston Lenczner",
                    "Linda M. See",
                    "Jan Dirk Wegner",
                    "Corinna Frank",
                    "J. Fluix\u00e1-Sanmart\u00edn",
                    "Giorgio Pasquali",
                    "Konstantin Klemmer",
                    "Esther Rolf",
                    "Caleb Robinson",
                    "Lester Mackey",
                    "Arthur Chevalley",
                    "Ciprian Tomoiaga",
                    "Marcin Detyniecki",
                    "O. Antropov",
                    "M. Molinier",
                    "R. Kuzu",
                    "Shaojia Ge",
                    "Sudipan Saha",
                    "Valentin Gabeff",
                    "Alexander Mathis",
                    "R. Roscher",
                    "Caroline Gevaert",
                    "Michael Kampffmeyer",
                    "J. A. D. Santos",
                    "Maria Vakalopoulou",
                    "R. H\u00e4nsch",
                    "Stine Hansen",
                    "Keiller Nogueira",
                    "Jonathan Prexl",
                    "Sherrie Wang",
                    "J. Mifdal",
                    "N. Long\u00e9p\u00e9",
                    "H. Ansari",
                    "Mohsin Ali",
                    "S. Montazeri",
                    "A. Parizzi"
                ],
                "domain": [
                    "Machine Learning",
                    "Remote Sensing",
                    "Conservation Biology",
                    "Environmental Monitoring"
                ],
                "publications": [
                    {
                        "title": "Imbalance-aware Presence-only Loss Function for Species Distribution Modeling",
                        "abstract": "In the face of significant biodiversity decline, species distribution models (SDMs) are essential for understanding the impact of climate change on species habitats by connecting environmental conditions to species occurrences. Traditionally limited by a scarcity of species observations, these models have significantly improved in performance through the integration of larger datasets provided by citizen science initiatives. However, they still suffer from the strong class imbalance between species within these datasets, often resulting in the penalization of rare species--those most critical for conservation efforts. To tackle this issue, this study assesses the effectiveness of training deep learning models using a balanced presence-only loss function on large citizen science-based datasets. We demonstrate that this imbalance-aware loss function outperforms traditional loss functions across various datasets and tasks, particularly in accurately modeling rare species with limited observations."
                    },
                    {
                        "title": "Where the White Continent Is Blue: Deep Learning Locates Bare Ice in Antarctica",
                        "abstract": "In some areas of Antarctica, blue\u2010colored bare ice is exposed at the surface. These blue ice areas (BIAs) can trap meteorites or old ice and are vital for understanding the climatic history. By combining multi\u2010sensor remote sensing data (MODIS, RADARSAT\u20102, and TanDEM\u2010X) in a deep learning framework, we map blue ice across the continent at 200\u2010m resolution. We use a novel methodology for image segmentation with \u201cnoisy\u201d labels to learn an underlying \u201cclean\u201d pattern with a neural network. In total, BIAs cover ca. 140,000 km2 (\u223c1%) of Antarctica, of which nearly 50% located within 20 km of the grounding line. There, the low albedo of blue ice enhances melt\u2010water production and its mapping is crucial for mass balance studies that determine the stability of the ice sheet. Moreover, the map provides input for fieldwork missions and can act as constraint for other geophysical mapping efforts."
                    },
                    {
                        "title": "Mapping drivers of tropical forest loss with satellite image time series and machine learning",
                        "abstract": "The rates of tropical deforestation remain high, resulting in carbon emissions, biodiversity loss, and impacts on local communities. To design effective policies to tackle this, it is necessary to know what the drivers behind deforestation are. Since drivers vary in space and time, producing accurate spatially explicit maps with regular temporal updates is essential. Drivers can be recognized from satellite imagery but the scale of tropical deforestation makes it unfeasible to do so manually. Machine learning opens up possibilities for automating and scaling up this process. In this study, we developed and trained a deep learning model to classify the drivers of any forest loss\u2014including deforestation\u2014from satellite image time series. Our model architecture allows understanding of how the input time series is used to make a prediction, showing the model learns different patterns for recognizing each driver and highlighting the need for temporal data. We used our model to classify over 588\u2032000 sites to produce a map detailing the drivers behind tropical forest loss. The results confirm that the majority of it is driven by agriculture, but also show significant regional differences. Such data is a crucial source of information to enable targeting specific drivers locally and can be updated in the future using free satellite data."
                    },
                    {
                        "title": "Short-term runoff forecasting in an alpine catchment with a long short-term memory neural network",
                        "abstract": "The governing hydrological processes are expected to shift under climate change in the alpine regions of Switzerland. This raises the need for more adaptive and accurate methods to estimate river flow. In high-altitude catchments influenced by snow and glaciers, short-term flow forecasting is challenging, as the exact mechanisms of transient melting processes are difficult to model mathematically and are poorly understood to this date. Machine learning methods, particularly temporally aware neural networks, have been shown to compare well and often outperform process-based hydrological models on medium and long-range forecasting. In this work, we evaluate a Long Short-Term Memory neural network (LSTM) for short-term prediction (up to three days) of hourly river flow in an alpine headwater catchment (Goms Valley, Switzerland). We compare the model with the regional standard, an existing process-based model (named MINERVE) that is used by local authorities and is calibrated on the study area. We found that the LSTM was more accurate than the process-based model on high flows and better represented the diurnal melting cycles of snow and glacier in the area of interest. It was on par with MINERVE in estimating two flood events: the LSTM captures the dynamics of a precipitation-driven flood well, while underestimating the peak discharge during an event with varying conditions between rain and snow. Finally, we analyzed feature importances and tested the transferability of the trained LSTM on a neighboring catchment showing comparable topographic and hydrological features. The accurate results obtained highlight the applicability and competitiveness of data-driven temporal machine learning models with the existing process-based model in the study area."
                    },
                    {
                        "title": "Detection of Settlements in Tanzania and Mozambique by Many Regional Few-Shot Models",
                        "abstract": "In this work, we propose an approach to aid in mapping small settlements, which are often misclassified by models trained on a large-scale context (global or regional). We leverage pre-trained land cover models and few-shot learning to enhance the detection of these settlements. The backbone models are trained globally, but their application is localized through a spatial sampling strategy to address the challenge of detecting missed or unlabelled settlements. The proposed sampling strategy is based on the distance around a test patch and allows for the sampling of both backgrounds (non-settlements) points and settlements. Following this strategy results in a balanced dataset for model fine-tuning and ensures that the model is well-adapted to the local context. The idea is that nearby settlements share more similar properties, which is leveraged in our approach. We evaluate these transferred models by measuring the number of previously unmapped settlements detected by the fine-tuned classifier. For this, we manually annotated over two thousand buildings across two regions of Tanzania, previously unmapped in the original urban landcover product. Our results indicate the potential of the sampling approach, particularly when combined with a model pretrained with Momentum Contrast (MoCo). However, we also highlight the limitations in terms of spatial resolution of Sentinel-2 data for the detection of small settlements."
                    },
                    {
                        "title": "SatCLIP: Global, General-Purpose Location Embeddings with Satellite Imagery",
                        "abstract": "Geographic information is essential for modeling tasks in fields ranging from ecology to epidemiology. However, extracting relevant location characteristics for a given task can be challenging, often requiring expensive data fusion or distillation from massive global imagery datasets. To address this challenge, we introduce Satellite Contrastive Location-Image Pretraining (SatCLIP). This global, general-purpose geographic location encoder learns an implicit representation of locations by matching CNN and ViT inferred visual patterns of openly available satellite imagery with their geographic coordinates. The resulting SatCLIP location encoder efficiently summarizes the characteristics of any given location for convenient use in downstream tasks. In our experiments, we use SatCLIP embeddings to improve prediction performance on nine diverse location-dependent tasks including temperature prediction, animal recognition, and population density estimation. Across tasks, SatCLIP consistently outperforms alternative location encoders and improves geographic generalization by encoding visual similarities of spatially distant environments. These results demonstrate the potential of vision-location models to learn meaningful representations of our planet from the vast, varied, and largely untapped modalities of geospatial data."
                    },
                    {
                        "title": "Improving Few-Shot Object Detection with Object Part Proposals",
                        "abstract": "Few-Shot Object Detection (FSOD) allows fast adaptation of an object detection model to new classes of objects using few examples per class. This has many applications, in particular in satellite and aerial observation, as it allows learning from experts who can only annotate a few examples for new classes and helps migrate models across tasks. In this work, we present a technique to improve the performance of FSOD in remote sensing by defining a contrastive loss that utilizes parts of objects. For this, we generate, what we call, Object Parts Proposals (OPPs) on the fly for each novel class, and use them to learn more robust features with an additional contrastive objective. We observe that training with OPPs brings a consistent improvement over the state-of-the-art when evaluating on the DIOR dataset.The code is available at https://github.com/arthurchevalley/Improving-FSOD-on-RSI-using-Sub-Parts."
                    },
                    {
                        "title": "Semi-Supervised Deep Learning Representations in Earth Observation Based Forest Management",
                        "abstract": "In this study, we examine the potential of several self-supervised deep learning models in predicting forest attributes and detecting forest changes using ESA Sentinel-1 and Sentinel-2 images. The performance of the proposed deep learning models is compared to established conventional machine learning approaches. Studied use-cases include mapping of forest disturbance (windthrown forests, snowload damages) using deep change vector analysis, forest height mapping using UNet+ based models, Momentum contrast and regression modeling. Study areas were represented by several boreal forest sites in Finland. Our results indicate that developed methods allow to achieve superior classification and prediction accuracies compared to traditional methodologies and mimimize the amount of necessary in-situ forestry data."
                    },
                    {
                        "title": "WildCLIP: Scene and animal attribute retrieval from camera trap data with domain-adapted vision-language models",
                        "abstract": "Wildlife observation with camera traps has great potential for ethology and ecology, as it gathers data non-invasively in an automated way. However, camera traps produce large amounts of uncurated data, which is time-consuming to annotate. Existing methods to label these data automatically commonly use a fixed pre-defined set of distinctive classes and require many labeled examples per class to be trained. Moreover, the attributes of interest are sometimes rare and difficult to find in large data collections. Large pretrained vision-language models, such as Contrastive Language Image Pretraining (CLIP), offer great promises to facilitate the annotation process of camera-trap data. Images can be described with greater detail, the set of classes is not fixed and can be extensible on demand and pretrained models can help to retrieve rare samples. In this work, we explore the potential of CLIP to retrieve images according to environmental and ecological attributes. We create WildCLIP by fine-tuning CLIP on wildlife camera-trap images and to further increase its flexibility, we add an adapter module to better expand to novel attributes in a few-shot manner. We quantify WildCLIP\u2019s performance and show that it can retrieve novel attributes in the Snapshot Serengeti dataset. Our findings outline new opportunities to facilitate annotation processes with complex and multi-attribute captions. The code will be made available at https://github.com/amathislab/wildclip."
                    },
                    {
                        "title": "Better, Not Just More: Data-centric machine learning for Earth observation",
                        "abstract": "Recent developments and research in modern machine learning have led to substantial improvements in the geospatial field. Although numerous deep learning architectures and models have been proposed, the majority of them have been solely developed on benchmark datasets that lack strong real-world relevance. Furthermore, the performance of many methods has already saturated on these datasets. We argue that a shift from a model-centric view to a complementary data-centric perspective is necessary for further improvements in accuracy, generalization ability, and real impact on end-user applications. Furthermore, considering the entire machine learning cycle - from problem definition to model deployment with feedback - is crucial for enhancing machine learning models that can be reliable in unforeseen situations. This work presents a definition as well as a precise categorization and overview of automated data-centric learning approaches for geospatial data. It highlights the complementary role of data-centric learning with respect to model-centric in the larger machine learning deployment cycle. We review papers across the entire geospatial field and categorize them into different groups. A set of representative experiments shows concrete implementation examples. These examples provide concrete steps to act on geospatial data with data-centric machine learning approaches."
                    },
                    {
                        "title": "Classification of Tropical Deforestation Drivers with Machine Learning and Satellite Image Time Series",
                        "abstract": "Tropical deforestation is a major environmental problem with severe consequences such as carbon emissions or biodiversity loss. While much research focuses on monitoring and mapping deforestation, less attention is paid to understanding the various reasons and motivations behind it, known as deforestation drivers. Drivers can typically be identified from optical satellite imagery, but it is often necessary to view the deforested site at multiple points in time to determine the driver, making manual annotation of drivers laborious. In this work, we propose a deep learning model that classifies drivers from time series of Sentinel-2 images. The model combines convolutional, LSTM, and attention layers. To train the model, we use a large crowd-sourced dataset spanning across the tropics. We compare its results to other architectures and show that using time series can bring significant improvement in accuracy compared to single images, especially if a suitable architecture is used. Additionally, we analyze the attention scores produced by our model and show that it learns different strategies for different classes."
                    },
                    {
                        "title": "Humans are Poor Few-Shot Classifiers for Sentinel-2 Land Cover",
                        "abstract": "Learning to predict accurately from a few data samples is a central challenge in modern data-hungry machine learning. On natural images, human vision typically outperforms deep learning approaches on few-shot learning. However, we hypothesize that aerial and satellite images are more challenging to the human eye. This applies particularly when the image resolution is comparatively low, as with the 10m ground sampling distance of Sentinel-2. In this study, we benchmark model-agnostic meta-learning (MAML) algorithms against human participants on few-shot land cover classification with Sentinel-2 imagery on the Sen12MS dataset. We find that categorization of land cover from globally distributed regions is a difficult task for the participants, who classified the given images less accurately than the MAML-trained model and with a highly variable success rate. This suggest that hand-labeling land cover directly on Sentinel-2 imagery is not optimal when tackling a new land cover classification problem. Labeling only a few images and employing a trained meta-learning model to this task may lead to more accurate and consistent solutions compared to hand labeling by multiple individuals."
                    },
                    {
                        "title": "TOWARDS DETECTING FLOATING OBJECTS ON A GLOBAL SCALE WITH LEARNED SPATIAL FEATURES USING SENTINEL 2",
                        "abstract": "Marine litter is a growing problem that has been attracting attention and raising concerns over the last years. Significant quantities of plastic can be found in the oceans due to the unfiltered discharge of waste into rivers, poor waste management, or lost fishing nets. The floating elements drift on the surface of water bodies and can be aggregated by processes, such as river plumes, windrows, oceanic fronts, or currents. In this paper, we focus on detecting big patches of floating objects that can contain plastic as well as other materials with optical Sentinel 2 data. In contrast to previous work that focuses on pixel-wise spectral responses of some bands, we employ a deep learning predictor that learns the spatial characteristics of floating objects. Along with this work, we provide a hand-labeled Sentinel 2 dataset of floating objects on the sea surface and other water bodies such as lakes together with pre-trained deep learning models. Our experiments demonstrate that harnessing the spatial patterns learned with a CNN is advantageous over pixel-wise classifications that use hand-crafted features. We further provide an analysis of the categories of floating objects that we captured while labeling the dataset and analyze the feature importance for the CNN predictions. Finally, we outline the limitations of trained CNN on several systematic failure cases that we would like to address in future work by increasing the diversity in the dataset and tackling the domain shift between regions and satellite acquisitions. The dataset introduced in this work is the first to provide public large-scale data for floating litter detection and we hope it will give more insights into developing techniques for floating litter detection and classification. Source code and data are available at https://github.com/ESA-PhiLab/floatingobjects."
                    },
                    {
                        "title": "InSAR Displacement Time Series Mining: A Machine Learning Approach",
                        "abstract": "Interferometric Synthetic Aperture Radar (InSAR)-derived surface displacement time series enable a wide range of applications from urban structural monitoring to geohazard assessment. With systematic data acquisitions becoming the new norm for SAR missions, millions of time series are continuously generated. Machine Learning provides a framework for the efficient mining of such big data. Here, we focus on unsupervised mining of the data via clustering the similar temporal patterns and data-driven displacement signal reconstruction from the InSAR time series. We propose a deep Long Short Term Memory (LSTM) autoencoder model which can exploit temporal relations in contrast to the commonly used shallow learning methods, such as Uniform Manifold Approximation and Projection (UMAP). We also modify the loss function to allow the quantification of uncertainties in the time series data. The two approaches are applied to the Lazufre Volcanic Complex located at the central volcanic zone of the Andes and thereby compared."
                    }
                ]
            },
            "b19e31c6-c9a2-479b-8dc3-8f70469b2d98": {
                "pk": "b19e31c6-c9a2-479b-8dc3-8f70469b2d98",
                "name": "Konstantin Klemmer",
                "collaborators": [
                    "Esther Rolf",
                    "Caleb Robinson",
                    "Hannah Kerner",
                    "Lester Mackey",
                    "M. Ru\u00dfwurm"
                ],
                "domain": [
                    "Satellite Data",
                    "Machine Learning",
                    "Geographic Information",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Mission Critical - Satellite Data is a Distinct Modality in Machine Learning",
                        "abstract": "Satellite data has the potential to inspire a seismic shift for machine learning -- one in which we rethink existing practices designed for traditional data modalities. As machine learning for satellite data (SatML) gains traction for its real-world impact, our field is at a crossroads. We can either continue applying ill-suited approaches, or we can initiate a new research agenda that centers around the unique characteristics and challenges of satellite data. This position paper argues that satellite data constitutes a distinct modality for machine learning research and that we must recognize it as such to advance the quality and impact of SatML research across theory, methods, and deployment. We outline critical discussion questions and actionable suggestions to transform SatML from merely an intriguing application area to a dedicated research discipline that helps move the needle on big challenges for machine learning and society."
                    },
                    {
                        "title": "SatCLIP: Global, General-Purpose Location Embeddings with Satellite Imagery",
                        "abstract": "Geographic information is essential for modeling tasks in fields ranging from ecology to epidemiology. However, extracting relevant location characteristics for a given task can be challenging, often requiring expensive data fusion or distillation from massive global imagery datasets. To address this challenge, we introduce Satellite Contrastive Location-Image Pretraining (SatCLIP). This global, general-purpose geographic location encoder learns an implicit representation of locations by matching CNN and ViT inferred visual patterns of openly available satellite imagery with their geographic coordinates. The resulting SatCLIP location encoder efficiently summarizes the characteristics of any given location for convenient use in downstream tasks. In our experiments, we use SatCLIP embeddings to improve prediction performance on nine diverse location-dependent tasks including temperature prediction, animal recognition, and population density estimation. Across tasks, SatCLIP consistently outperforms alternative location encoders and improves geographic generalization by encoding visual similarities of spatially distant environments. These results demonstrate the potential of vision-location models to learn meaningful representations of our planet from the vast, varied, and largely untapped modalities of geospatial data."
                    }
                ]
            },
            "55373820-de9d-481e-a38c-548adc43839a": {
                "pk": "55373820-de9d-481e-a38c-548adc43839a",
                "name": "Esther Rolf",
                "collaborators": [
                    "Konstantin Klemmer",
                    "Caleb Robinson",
                    "Hannah Kerner",
                    "Lester Mackey",
                    "M. Ru\u00dfwurm"
                ],
                "domain": [
                    "Satellite Data",
                    "Machine Learning",
                    "Geographic Information",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Mission Critical - Satellite Data is a Distinct Modality in Machine Learning",
                        "abstract": "Satellite data has the potential to inspire a seismic shift for machine learning -- one in which we rethink existing practices designed for traditional data modalities. As machine learning for satellite data (SatML) gains traction for its real-world impact, our field is at a crossroads. We can either continue applying ill-suited approaches, or we can initiate a new research agenda that centers around the unique characteristics and challenges of satellite data. This position paper argues that satellite data constitutes a distinct modality for machine learning research and that we must recognize it as such to advance the quality and impact of SatML research across theory, methods, and deployment. We outline critical discussion questions and actionable suggestions to transform SatML from merely an intriguing application area to a dedicated research discipline that helps move the needle on big challenges for machine learning and society."
                    },
                    {
                        "title": "SatCLIP: Global, General-Purpose Location Embeddings with Satellite Imagery",
                        "abstract": "Geographic information is essential for modeling tasks in fields ranging from ecology to epidemiology. However, extracting relevant location characteristics for a given task can be challenging, often requiring expensive data fusion or distillation from massive global imagery datasets. To address this challenge, we introduce Satellite Contrastive Location-Image Pretraining (SatCLIP). This global, general-purpose geographic location encoder learns an implicit representation of locations by matching CNN and ViT inferred visual patterns of openly available satellite imagery with their geographic coordinates. The resulting SatCLIP location encoder efficiently summarizes the characteristics of any given location for convenient use in downstream tasks. In our experiments, we use SatCLIP embeddings to improve prediction performance on nine diverse location-dependent tasks including temperature prediction, animal recognition, and population density estimation. Across tasks, SatCLIP consistently outperforms alternative location encoders and improves geographic generalization by encoding visual similarities of spatially distant environments. These results demonstrate the potential of vision-location models to learn meaningful representations of our planet from the vast, varied, and largely untapped modalities of geospatial data."
                    }
                ]
            },
            "0a310e45-2021-4fa6-9c8b-2b5ce762b27c": {
                "pk": "0a310e45-2021-4fa6-9c8b-2b5ce762b27c",
                "name": "Robin Zbinden",
                "collaborators": [
                    "Nina Van Tiel",
                    "M. Ru\u00dfwurm",
                    "D. Tuia"
                ],
                "domain": [
                    "Biodiversity",
                    "Species Distribution Models",
                    "Deep Learning",
                    "Conservation Science"
                ],
                "publications": [
                    {
                        "title": "Imbalance-aware Presence-only Loss Function for Species Distribution Modeling",
                        "abstract": "In the face of significant biodiversity decline, species distribution models (SDMs) are essential for understanding the impact of climate change on species habitats by connecting environmental conditions to species occurrences. Traditionally limited by a scarcity of species observations, these models have significantly improved in performance through the integration of larger datasets provided by citizen science initiatives. However, they still suffer from the strong class imbalance between species within these datasets, often resulting in the penalization of rare species--those most critical for conservation efforts. To tackle this issue, this study assesses the effectiveness of training deep learning models using a balanced presence-only loss function on large citizen science-based datasets. We demonstrate that this imbalance-aware loss function outperforms traditional loss functions across various datasets and tasks, particularly in accurately modeling rare species with limited observations."
                    }
                ]
            },
            "e989eb80-007e-4ba5-b5ff-fa54a70d7dc5": {
                "pk": "e989eb80-007e-4ba5-b5ff-fa54a70d7dc5",
                "name": "Devis Tuia",
                "collaborators": [
                    "J. Castillo-Navarro",
                    "M. Ru\u00dfwurm",
                    "Li Mi",
                    "Jan Pisl",
                    "Gaston Lenczner",
                    "Jo\u00e3o Daniel Silva",
                    "Jo\u00e3o Magalh\u00e3es",
                    "Bruno Martins",
                    "Diego Marcos",
                    "Syrielle Montariol",
                    "Antoine Bosselut",
                    "John E. Vargas-Mu\u00f1oz",
                    "Nando Metzger",
                    "R. C. Daudt",
                    "Konrad Schindler",
                    "F. D. Morsier",
                    "Benoit Dufumier",
                    "Jean-Philippe Thiran",
                    "Robin Zbinden",
                    "Nina Van Tiel",
                    "Vandita Shukla",
                    "Luca Morelli",
                    "F. Remondino",
                    "Andrea Micheli",
                    "Benjamin Risse",
                    "Hugo Porta",
                    "Emanuele Dalsasso",
                    "Alex Levering",
                    "Nathan Jacobs",
                    "Veronica Tollenaar",
                    "H. Zekollari",
                    "F. Pattyn",
                    "B. Kellenberger",
                    "S. Lhermitte",
                    "Maaike Izeboud",
                    "Val\u00e9rie Zermatten",
                    "Xiaolong Lu",
                    "Tobias Kellenberger",
                    "Chang Xu",
                    "Wen Yang",
                    "Xianjie Dai",
                    "Jo\u00e3o Luis Prado",
                    "Siran Li",
                    "Lloyd Haydn Hughes",
                    "Linda M. See",
                    "Jan Dirk Wegner"
                ],
                "domain": [
                    "Remote Sensing",
                    "Deep Learning",
                    "Multimodal Learning",
                    "Environmental Monitoring"
                ],
                "publications": [
                    {
                        "title": "Estimating Fine-Grained Population Growth Rates from Coarse Census Data",
                        "abstract": "Fine-grained population estimation is important for several domains, such as urban planning, public health, and humanitarian action. Due to limited resources, population maps with sufficient spatial resolution and temporal frequency are not available for many developing countries. Population estimates are often based on statistics available at the national or provincial level, which moreover are updated infrequently. The United Nations produce estimates of population growth rates, which are widely used to project population numbers to a target year, starting from the most recent census data. However, they use a simplified model of population growth with uniform growth rates for all urban, respectively rural areas within a country. This neglects the complex dynamics of population growth (e.g., growth rates in big cities are usually larger than in smaller urban areas) and leads to significant errors in population projections. In this work, we propose a methodology to estimate fine-grained population growth rates and present experimental results for Mozambique."
                    },
                    {
                        "title": "Semantic Segmentation of Coffee Plantations from Sentinel-2 Time Series",
                        "abstract": "Coffee production serves as an important source of income for millions of farmers across many tropical countries. Given the scale of the production, accurate and up-to-date maps of coffee plantations are needed to detect, monitor, and mitigate potential negative environmental impacts. Such maps can be produced based on satellite imagery in a cost-effective and replicable way. However, this comes with certain challenges, including the dynamic spectral signature of coffee, complex topographies in which it is often cultivated, and the high cloud coverage common in tropical countries. In this work, we train a deep learning model to detect coffee plantations in Brazil directly from time series of Sentinel-2 images, alleviating the need for manual extraction of features. We show that the model outperforms significantly models trained on single images, is more robust against clouds and various seasonal patterns, and generalizes better to new regions."
                    },
                    {
                        "title": "What to align in multimodal contrastive learning?",
                        "abstract": "Humans perceive the world through multisensory integration, blending the information of different modalities to adapt their behavior. Contrastive learning offers an appealing solution for multimodal self-supervised learning. Indeed, by considering each modality as a different view of the same entity, it learns to align features of different modalities in a shared representation space. However, this approach is intrinsically limited as it only learns shared or redundant information between modalities, while multimodal interactions can arise in other ways. In this work, we introduce CoMM, a Contrastive MultiModal learning strategy that enables the communication between modalities in a single multimodal space. Instead of imposing cross- or intra- modality constraints, we propose to align multimodal representations by maximizing the mutual information between augmented versions of these multimodal features. Our theoretical analysis shows that shared, synergistic and unique terms of information naturally emerge from this formulation, allowing us to estimate multimodal interactions beyond redundancy. We test CoMM both in a controlled and in a series of real-world settings: in the former, we demonstrate that CoMM effectively captures redundant, unique and synergistic information between modalities. In the latter, CoMM learns complex multimodal interactions and achieves state-of-the-art results on the six multimodal benchmarks."
                    },
                    {
                        "title": "Imbalance-aware Presence-only Loss Function for Species Distribution Modeling",
                        "abstract": "In the face of significant biodiversity decline, species distribution models (SDMs) are essential for understanding the impact of climate change on species habitats by connecting environmental conditions to species occurrences. Traditionally limited by a scarcity of species observations, these models have significantly improved in performance through the integration of larger datasets provided by citizen science initiatives. However, they still suffer from the strong class imbalance between species within these datasets, often resulting in the penalization of rare species--those most critical for conservation efforts. To tackle this issue, this study assesses the effectiveness of training deep learning models using a balanced presence-only loss function on large citizen science-based datasets. We demonstrate that this imbalance-aware loss function outperforms traditional loss functions across various datasets and tasks, particularly in accurately modeling rare species with limited observations."
                    },
                    {
                        "title": "Large Language Models for Captioning and Retrieving Remote Sensing Images",
                        "abstract": "Image captioning and cross-modal retrieval are examples of tasks that involve the joint analysis of visual and linguistic information. In connection to remote sensing imagery, these tasks can help non-expert users in extracting relevant Earth observation information for a variety of applications. Still, despite some previous efforts, the development and application of vision and language models to the remote sensing domain have been hindered by the relatively small size of the available datasets and models used in previous studies. In this work, we propose RS-CapRet, a Vision and Language method for remote sensing tasks, in particular image captioning and text-image retrieval. We specifically propose to use a highly capable large decoder language model together with image encoders adapted to remote sensing imagery through contrastive language-image pre-training. To bridge together the image encoder and language decoder, we propose training simple linear layers with examples from combining different remote sensing image captioning datasets, keeping the other parameters frozen. RS-CapRet can then generate descriptions for remote sensing images and retrieve images from textual descriptions, achieving SOTA or competitive performance with existing methods. Qualitative results illustrate that RS-CapRet can effectively leverage the pre-trained large language model to describe remote sensing images, retrieve them based on different types of queries, and also show the ability to process interleaved sequences of images and text in a dialogue manner."
                    },
                    {
                        "title": "Towards Estimation of 3D Poses and Shapes of Animals from Oblique Drone Imagery",
                        "abstract": "Abstract. Wildlife research in both terrestrial and aquatic ecosystems now deploys drone technology for tasks such as monitoring, census counts and habitat analysis. Unlike camera traps, drones offer real-time flexibility for adaptable flight paths and camera views, thus making them ideal for capturing multi-view data on wildlife like zebras or lions. With recent advancements in animals\u2019 3D shape & pose estimation, there is an increasing interest in bringing 3D analysis from ground to sky by means of drones. The paper reports some activities of the EU-funded WildDrone project and performs, for the first time, 3D analyses of animals exploiting oblique drone imagery. Using parametric model fitting, we estimate 3D shape and pose of animals from frames of a monocular RGB video. With the goal of appending metric information to parametric animal models using photogrammetric evidence, we propose a pipeline where we perform a point cloud reconstruction of the scene to scale and localize the animal within the 3D scene. Challenges, planned next steps and future directions are also reported. "
                    },
                    {
                        "title": "Multi-Scale Grouped Prototypes for Interpretable Semantic Segmentation",
                        "abstract": "Prototypical part learning is emerging as a promising approach for making semantic segmentation interpretable. The model selects real patches seen during training as prototypes and constructs the dense prediction map based on the similarity between parts of the test image and the prototypes. This improves interpretability since the user can inspect the link between the predicted output and the patterns learned by the model in terms of prototypical information. In this paper, we propose a method for interpretable semantic segmentation that leverages multi-scale image representation for prototypical part learning. First, we introduce a prototype layer that explicitly learns diverse prototypical parts at several scales, leading to multi-scale representations in the prototype activation output. Then, we propose a sparse grouping mechanism that produces multi-scale sparse groups of these scale-specific prototypical parts. This provides a deeper understanding of the interactions between multi-scale object representations while enhancing the interpretability of the segmentation model. The experiments conducted on Pascal VOC, Cityscapes, and ADE20K demonstrate that the proposed method increases model sparsity, improves interpretability over existing prototype-based methods, and narrows the performance gap with the non-interpretable counterpart models. Code is available at github.com/eceo-epfl/ScaleProtoSeg."
                    },
                    {
                        "title": "Multilingual Vision-Language Pre-training for the Remote Sensing Domain",
                        "abstract": "Methods based on Contrastive Language-Image Pre-training (CLIP) are nowadays extensively used in support of vision-and-language tasks involving remote sensing data, such as cross-modal retrieval. The adaptation of CLIP to this specific domain has relied on model fine-tuning with the standard contrastive objective, using existing human-labeled image-caption datasets, or using synthetic data corresponding to image-caption pairs derived from other annotations over remote sensing images (e.g., object classes). The use of different pre-training mechanisms has received less attention, and only a few exceptions have considered multilingual inputs. This work proposes a novel vision-and-language model for the remote sensing domain, exploring the fine-tuning of a multilingual CLIP model and testing the use of a self-supervised method based on aligning local and global representations from individual input images, together with the standard CLIP objective. Model training relied on assembling pre-existing datasets of remote sensing images paired with English captions, followed by the use of automated machine translation into nine additional languages. We show that translated data is indeed helpful, e.g. improving performance also on English. Our resulting model, which we named Remote Sensing Multilingual CLIP (RS-M-CLIP), obtains state-of-the-art results in a variety of vision-and-language tasks, including cross-modal and multilingual image-text retrieval, or zero-shot image classification."
                    },
                    {
                        "title": "Prompt-guided and multimodal landscape scenicness assessments with vision-language models",
                        "abstract": "Recent advances in deep learning and Vision-Language Models (VLM) have enabled efficient transfer to downstream tasks even when limited labelled training data is available, as well as for text to be directly compared to image content. These properties of VLMs enable new opportunities for the annotation and analysis of images. We test the potential of VLMs for landscape scenicness prediction, i.e., the aesthetic quality of a landscape, using zero- and few-shot methods. We experiment with few-shot learning by fine-tuning a single linear layer on a pre-trained VLM representation. We find that a model fitted to just a few hundred samples performs favourably compared to a model trained on hundreds of thousands of examples in a fully supervised way. We also explore the zero-shot prediction potential of contrastive prompting using positive and negative landscape aesthetic concepts. Our results show that this method outperforms a linear probe with few-shot learning when using a small number of samples to tune the prompt configuration. We introduce Landscape Prompt Ensembling (LPE), which is an annotation method for acquiring landscape scenicness ratings through rated text descriptions without needing an image dataset during annotation. We demonstrate that LPE can provide landscape scenicness assessments that are concordant with a dataset of image ratings. The success of zero- and few-shot methods combined with their ability to use text-based annotations highlights the potential for VLMs to provide efficient landscape scenicness assessments with greater flexibility."
                    },
                    {
                        "title": "Where the White Continent Is Blue: Deep Learning Locates Bare Ice in Antarctica",
                        "abstract": "In some areas of Antarctica, blue\u2010colored bare ice is exposed at the surface. These blue ice areas (BIAs) can trap meteorites or old ice and are vital for understanding the climatic history. By combining multi\u2010sensor remote sensing data (MODIS, RADARSAT\u20102, and TanDEM\u2010X) in a deep learning framework, we map blue ice across the continent at 200\u2010m resolution. We use a novel methodology for image segmentation with \u201cnoisy\u201d labels to learn an underlying \u201cclean\u201d pattern with a neural network. In total, BIAs cover ca. 140,000 km2 (\u223c1%) of Antarctica, of which nearly 50% located within 20 km of the grounding line. There, the low albedo of blue ice enhances melt\u2010water production and its mapping is crucial for mass balance studies that determine the stability of the ice sheet. Moreover, the map provides input for fieldwork missions and can act as constraint for other geophysical mapping efforts."
                    },
                    {
                        "title": "Land Cover Mapping From Multiple Complementary Experts Under Heavy Class Imbalance",
                        "abstract": "Deep learning has emerged as a promising avenue for automatic mapping, demonstrating high efficacy in land cover categorization through various semantic segmentation models. Nonetheless, the practical deployment of these models encounters important challenges from the imbalanced distribution of samples between the classes, a problem inherent to real-world datasets. This results in models biased towards frequent classes that perform poorly on rare classes. While existing approaches to fight class imbalance mainly focus on image classification, here we propose to address this issue for semantic segmentation with a multiple complementary experts (MCE) structure. Taking inspiration from ensemble models, each expert in our MCE specializes in certain classes and works with other experts in a complementary manner to generate robust predictions for rare classes. We compare our approach to other existing methods and also explore different logit aggregation methods, to identify the performance upper bounds and improvement directions. Our model is evaluated on a large-scale and challenging alpine land cover dataset that we make openly available. In addition, we evaluated our model on an imbalanced land cover mapping dataset, FLAIR, to highlight its adaptability. Overall, our MCE model yields notable improvement in performances on the medium and rare classes compared to baseline methods, while only slightly compromising on the overall accuracy. Despite its simplicity, the MCE approach stands as a practical solution for more operational semantic segmentation models, not trading off performances on rare but important classes."
                    },
                    {
                        "title": "ConGeo: Robust Cross-view Geo-localization across Ground View Variations",
                        "abstract": "Cross-view geo-localization aims at localizing a ground-level query image by matching it to its corresponding geo-referenced aerial view. In real-world scenarios, the task requires accommodating diverse ground images captured by users with varying orientations and reduced field of views (FoVs). However, existing learning pipelines are orientation-specific or FoV-specific, demanding separate model training for different ground view variations. Such models heavily depend on the North-aligned spatial correspondence and predefined FoVs in the training data, compromising their robustness across different settings. To tackle this challenge, we propose ConGeo, a single- and cross-view Contrastive method for Geo-localization: it enhances robustness and consistency in feature representations to improve a model's invariance to orientation and its resilience to FoV variations, by enforcing proximity between ground view variations of the same location. As a generic learning objective for cross-view geo-localization, when integrated into state-of-the-art pipelines, ConGeo significantly boosts the performance of three base models on four geo-localization benchmarks for diverse ground view variations and outperforms competing methods that train separate models for each ground view variation."
                    },
                    {
                        "title": "Knowledge-aware Text-Image Retrieval for Remote Sensing Images",
                        "abstract": "Image-based retrieval in large Earth observation archives is challenging because one needs to navigate across thousands of candidate matches only with the query image as a guide. By using text as information supporting the visual query, the retrieval system gains in usability, but at the same time faces difficulties due to the diversity of visual signals that cannot be summarized by a short caption only. For this reason, as a matching-based task, cross-modal text-image retrieval often suffers from information asymmetry between texts and images. To address this challenge, we propose a Knowledge-aware Text-Image Retrieval (KTIR) method for remote sensing images. By mining relevant information from an external knowledge graph, KTIR enriches the text scope available in the search query and alleviates the information gaps between texts and images for better matching. Moreover, by integrating domain-specific knowledge, KTIR also enhances the adaptation of pre-trained vision-language models to remote sensing applications. Experimental results on three commonly used remote sensing text-image retrieval benchmarks show that the proposed knowledge-aware method leads to varied and consistent retrievals, outperforming state-of-the-art retrieval methods."
                    },
                    {
                        "title": "Training Visual Language Models with Object Detection: Grounded Change Descriptions in Satellite Images",
                        "abstract": "Recently, generalist Vision Language Models (VLMs) have shown exceptional progress in tasks previously dominated by specialized computer vision models. This becomes more prevalent when visual grounding capabilities, such as the ability to reason over input text and image to generate bounding boxes around objects, are required. However, how these capabilities transfer to specialized domains such as remote sensing remains understudied, despite the recent increase in specialized models for Earth observation. In this work, we evaluate how grounding visual entities \u2013 by generating bounding-box coordinates \u2013 affects VLM performance in satellite imagery. To this end, we create two instruction-following tasks sourced from the xBD dataset, describing changes due to natural disasters observed in satellite images. We fine-tune several instances of MiniGPTv2, an open-source VLM with grounding capabilities, and evaluate their performance under the \"grounded\" vs. \"not grounded\" settings. We find that generating bounding boxes to refer to visual entities increases performance in tasks related to objects in the image, but only when the number of entities in the image is limited."
                    },
                    {
                        "title": "Knowledge-Aware Visual Question Generation for Remote Sensing Images",
                        "abstract": "With the rapid development of remote sensing image archives, asking questions about images has become an effective way of gathering specific information or performing image retrieval. However, automatically generated image-based questions tend to be simplistic and template-based, which hinders the real deployment of question answering or visual dialogue systems. To enrich and diversify the questions, we propose a knowledge-aware remote sensing visual question generation model, KRSVQG, that incorporates external knowledge related to the image content to improve the quality and contextual understanding of the generated questions. The model takes an image and a related knowledge triplet from external knowledge sources as inputs and leverages image captioning as an intermediary representation to enhance the image grounding of the generated questions. To assess the performance of KRSVQG, we utilized two datasets that we manually annotated: NWPU-300 and TextRS-300. Results on these two datasets demonstrate that KRSVQG outperforms existing methods and leads to knowledge-enriched questions, grounded in both image and domain knowledge."
                    },
                    {
                        "title": "Mapping drivers of tropical forest loss with satellite image time series and machine learning",
                        "abstract": "The rates of tropical deforestation remain high, resulting in carbon emissions, biodiversity loss, and impacts on local communities. To design effective policies to tackle this, it is necessary to know what the drivers behind deforestation are. Since drivers vary in space and time, producing accurate spatially explicit maps with regular temporal updates is essential. Drivers can be recognized from satellite imagery but the scale of tropical deforestation makes it unfeasible to do so manually. Machine learning opens up possibilities for automating and scaling up this process. In this study, we developed and trained a deep learning model to classify the drivers of any forest loss\u2014including deforestation\u2014from satellite image time series. Our model architecture allows understanding of how the input time series is used to make a prediction, showing the model learns different patterns for recognizing each driver and highlighting the need for temporal data. We used our model to classify over 588\u2032000 sites to produce a map detailing the drivers behind tropical forest loss. The results confirm that the majority of it is driven by agriculture, but also show significant regional differences. Such data is a crucial source of information to enable targeting specific drivers locally and can be updated in the future using free satellite data."
                    }
                ]
            }
        }
    },
    "2410.06170": {
        "paper_data": {
            "title": "QGym: Scalable Simulation and Benchmarking of Queuing Network Controllers",
            "url": "http://arxiv.org/abs/2410.06170v1",
            "arxiv_id": "2410.06170",
            "authors": [
                "Haozhe Chen",
                "Ang Li",
                "Ethan Che",
                "Tianyi Peng",
                "Jing Dong",
                "Hongseok Namkoong"
            ],
            "abstract": "Queuing network control determines the allocation of scarce resources to manage congestion, a fundamental problem in manufacturing, communications, and healthcare. Compared to standard RL problems, queueing problems are distinguished by unique challenges: i) a system operating in continuous time, ii) high stochasticity, and iii) long horizons over which the system can become unstable (exploding delays). To spur methodological progress tackling these challenges, we present an open-sourced queueing simulation framework, QGym, that benchmark queueing policies across realistic problem instances. Our modular framework allows the researchers to build on our initial instances, which provide a wide range of environments including parallel servers, criss-cross, tandem, and re-entrant networks, as well as a realistically calibrated hospital queuing system. QGym makes it easy to compare multiple policies, including both model-free RL methods and classical queuing policies. Our testbed complements the traditional focus on evaluating algorithms based on mathematical guarantees in idealized settings, and significantly expands the scope of empirical benchmarking in prior work. QGym code is open-sourced at https://github.com/namkoong-lab/QGym.",
            "introduction": "   1 Introduction  Queuing network control is a fundamental control problem in managing congestion in job-processing systems, such as semiconductor manufacturing fabrication plants, communications networks, cloud computing facilities, call centers, healthcare delivery systems, ride-sharing platforms, and limit-order books [44, 23, 37, 2, 4, 7, 14]. In a typical queuing system, jobs arrive at random intervals, wait in queues until an available server can service them, and then either leave the system or move on to another queue for further processing. The stochastic workload is the defining challenge in queueing: the variability inherent in real-world systems makes the time it takes to process jobs random. To manage performance objectives such as minimizing processing delays, balancing workloads, and improving overall service quality and efficiency, a good controller must dynamically allocate resources accounting for future stochasticity. The ability to plan is especially crucial for systems that experience varying levels of congestion over time.   Routing/scheduling control (i.e., matching jobs with servers) in industrial-scale systems is challenging due to several factors. First, queueing networks can be large and complex, with many different job classes and server types (see, e.g., [23, 4]). The processing speeds can depend on the server and job types, which must be taken into account to make effective scheduling decisions. Second, in systems such as semiconductor fabrication that involve multiple stages of sequential processing, congestion can occur at various points in the queuing network, creating bottlenecks that slow down the entire system. Third, workloads are highly non-stationary, featuring predictable and unpredictable demand spikes over time.   Figure 1: Highlights of QGym framework for developing and benchmarking queuing algorithms. QGym provides an event-driven simulator and benchmarks a wide range of queuing policies and systems. QGym interface also allows users to easily specify new queuing policies and systems.   There is a large body of work devoted to developing good routing/scheduling policies for queueing networks. The traditional focus of methodological development has been on simple policies with good theoretical performance guarantees in specific network structures. These include (1) load-balancing rules such as joining the shortest queue, and variations such as the power-of-d\ud835\udc51ditalic_d choices where one randomly samples d\ud835\udc51ditalic_d queues and sends the job to the shortest one among them [35]; (2) scheduling rules to minimize delay cost such as shortest processing time first, and variations of it such as the c\u2062\u03bc\ud835\udc50\ud835\udf07c\\muitalic_c italic_\u03bc-rule [10] and the generalized c\u2062\u03bc\ud835\udc50\ud835\udf07c\\muitalic_c italic_\u03bc-rule [30] when we do not know the exact job size and different job classes are associated with different delay costs; and (3) policies that achieve the maximum stability region such as MaxWeight [47] and maximum pressure policies [17].   While easy to implement and interpret, these policies are restrictive to specific queuing architectures or objectives, not data-driven, and difficult to adapt to network-specific features and nonstationarity. As a result, there is a growing interest in using black-box reinforcement learning (RL) techniques to learn queuing controllers in a more data-driven manner, which can better handle realistic settings featuring complex networks with non-stationary workloads.   However, queuing network control poses unique challenges, necessitating new methodological innovation over existing RL algorithms. Compared to typical robotics or game-playing environments, queueing problems have high stochasticity and longer horizons over which the system can become",
            "references": [
                {
                    "title": "Sample Efficient Reinforcement Learning in Mixed Systems through Augmented Samples and Its Applications to Queueing Networks",
                    "abstract": "This paper considers a class of reinforcement learning problems, which involve systems with two types of states: stochastic and pseudo-stochastic. In such systems, stochastic states follow a stochastic transition kernel while the transitions of pseudo-stochastic states are deterministic given the stochastic states/transitions. We refer to such systems as mixed systems, which are widely used in various applications, including manufacturing systems, communication networks, and queueing networks. We propose a sample efficient RL method that accelerates learning by generating augmented data samples. The proposed algorithm is data-driven and learns the policy from data samples from both real and augmented samples. This method significantly improves learning by reducing the sample complexity such that the dataset only needs to have sufficient coverage of the stochastic states. We analyze the sample complexity of the proposed method under Fitted Q Iteration (FQI) and demonstrate that the optimality gap decreases as $\\tilde{\\mathcal{O}}(\\sqrt{{1}/{n}}+\\sqrt{{1}/{m}}),$ where $n$ is the number of real samples and $m$ is the number of augmented samples per real sample. It is important to note that without augmented samples, the optimality gap is $\\tilde{\\mathcal{O}}(1)$ due to insufficient data coverage of the pseudo-stochastic states. Our experimental results on multiple queueing network applications confirm that the proposed method indeed significantly accelerates learning in both deep Q-learning and deep policy gradient."
                },
                {
                    "title": "SimOpt: A Testbed for Simulation-Optimization Experiments",
                    "abstract": "This paper introduces a major redesign of SimOpt, a testbed of simulation-optimization (SO) problems and solvers. The testbed promotes the empirical evaluation and comparison of solvers and aims to accelerate their development. Relative to previous versions of SimOpt, the redesign ports the code to an object-oriented architecture in Python; uses an implementation of the MRG32k3a random number generator that supports streams, substreams, and subsubstreams; supports the automated use of common random numbers for ease and efficiency; includes a powerful suite of plotting tools for visualizing experiment results; uses bootstrapping to obtain error estimates; accommodates the use of data farming to explore simulation models and optimization solvers as their input parameters vary; and provides a graphical user interface. The SimOpt source code is available on a GitHub repository under a permissive open-source license and as a Python package. History: Accepted by Ted Ralphs, Area Editor for Software Tools. Funding: This work was supported by the National Science Foundation [Grant CMMI-2035086]. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2023.1273 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2022.0011 ) at ( http://dx.doi.org/10.5281/zenodo.7468744 )."
                },
                {
                    "title": "Reincarnating Reinforcement Learning: Reusing Prior Computation to Accelerate Progress",
                    "abstract": "Learning tabula rasa, that is without any prior knowledge, is the prevalent workflow in reinforcement learning (RL) research. However, RL systems, when applied to large-scale settings, rarely operate tabula rasa. Such large-scale systems undergo multiple design or algorithmic changes during their development cycle and use ad hoc approaches for incorporating these changes without re-training from scratch, which would have been prohibitively expensive. Additionally, the inefficiency of deep RL typically excludes researchers without access to industrial-scale resources from tackling computationally-demanding problems. To address these issues, we present reincarnating RL as an alternative workflow or class of problem settings, where prior computational work (e.g., learned policies) is reused or transferred between design iterations of an RL agent, or from one RL agent to another. As a step towards enabling reincarnating RL from any agent to any other agent, we focus on the specific setting of efficiently transferring an existing sub-optimal policy to a standalone value-based RL agent. We find that existing approaches fail in this setting and propose a simple algorithm to address their limitations. Equipped with this algorithm, we demonstrate reincarnating RL's gains over tabula rasa RL on Atari 2600 games, a challenging locomotion task, and the real-world problem of navigating stratospheric balloons. Overall, this work argues for an alternative approach to RL research, which we believe could significantly improve real-world RL adoption and help democratize it further. Open-sourced code and trained agents at https://agarwl.github.io/reincarnating_rl."
                },
                {
                    "title": "RL-QN: A Reinforcement Learning Framework for Optimal Control of Queueing Systems",
                    "abstract": "With the rapid advance of information technology, network systems have become increasingly complex and hence the underlying system dynamics are often unknown or difficult to characterize. Finding a good network control policy is of significant importance to achieve desirable network performance (e.g., high throughput or low delay). In this work, we consider using model-based reinforcement learning (RL) to learn the optimal control policy for queueing networks so that the average job delay (or equivalently the average queue backlog) is minimized. Traditional approaches in RL, however, cannot handle the unbounded state spaces of the network control problem. To overcome this difficulty, we propose a new algorithm, called RL for Queueing Networks (RL-QN), which applies model-based RL methods over a finite subset of the state space while applying a known stabilizing policy for the rest of the states. We establish that the average queue backlog under RL-QN with an appropriately constructed subset can be arbitrarily close to the optimal result. We evaluate RL-QN in dynamic server allocation, routing, and switching problems. Simulation results show that RL-QN minimizes the average queue backlog effectively."
                },
                {
                    "title": "OR-Gym: A Reinforcement Learning Library for Operations Research Problem",
                    "abstract": "Reinforcement learning (RL) has been widely applied to game-playing and surpassed the best human-level performance in many domains, yet there are few use-cases in industrial or commercial settings. We introduce OR-Gym, an open-source library for developing reinforcement learning algorithms to address operations research problems. In this paper, we apply reinforcement learning to the knapsack, multi-dimensional bin packing, multi-echelon supply chain, and multi-period asset allocation model problems, as well as benchmark the RL solutions against MILP and heuristic models. These problems are used in logistics, finance, engineering, and are common in many business operation settings. We develop environments based on prototypical models in the literature and implement various optimization and heuristic models in order to benchmark the RL results. By re-framing a series of classic optimization problems as RL tasks, we seek to provide a new tool for the operations research community, while also opening those in the RL community to many of the problems and challenges in the OR field."
                },
                {
                    "title": "Queueing Network Controls via Deep Reinforcement Learning",
                    "abstract": "Novel advanced policy gradient (APG) methods, such as trust region policy optimization and proximal policy optimization (PPO), have become the dominant reinforcement learning algorithms because of their ease of implementation and good practical performance. A conventional setup for notoriously difficult queueing network control problems is a Markov decision problem (MDP) that has three features: infinite state space, unbounded costs, and long-run average cost objective. We extend the theoretical framework of these APG methods for such MDP problems. The resulting PPO algorithm is tested on a parallel-server system and large-size multiclass queueing networks. The algorithm consistently generates control policies that outperform state-of-art heuristics in literature in a variety of load conditions from light to heavy traffic. These policies are demonstrated to be near optimal when the optimal policy can be computed. A key to the successes of our PPO algorithm is the use of three variance reduction techniques in estimating the relative value function via sampling. First, we use a discounted relative value function as an approximation of the relative value function. Second, we propose regenerative simulation to estimate the discounted relative value function. Finally, we incorporate the approximating martingale-process method into the regenerative estimator."
                },
                {
                    "title": "Stable Reinforcement Learning with Unbounded State Space",
                    "abstract": "We consider the problem of reinforcement learning (RL) with unbounded state space motivated by the classical problem of scheduling in a queueing network. Traditional policies as well as error metric that are designed for finite, bounded or compact state space, require infinite samples for providing any meaningful performance guarantee (e.g. $\\ell_\\infty$ error) for unbounded state space. That is, we need a new notion of performance metric. As the main contribution of this work, inspired by the literature in queuing systems and control theory, we propose stability as the notion of \"goodness\": the state dynamics under the policy should remain in a bounded region with high probability. As a proof of concept, we propose an RL policy using Sparse-Sampling-based Monte Carlo Oracle and argue that it satisfies the stability property as long as the system dynamics under the optimal policy respects a Lyapunov function. The assumption of existence of a Lyapunov function is not restrictive as it is equivalent to the positive recurrence or stability property of any Markov chain, i.e., if there is any policy that can stabilize the system then it must possess a Lyapunov function. And, our policy does not utilize the knowledge of the specific Lyapunov function. To make our method sample efficient, we provide an improved, sample efficient Sparse-Sampling-based Monte Carlo Oracle with Lipschitz value function that may be of interest in its own right. Furthermore, we design an adaptive version of the algorithm, based on carefully constructed statistical tests, which finds the correct tuning parameter automatically."
                },
                {
                    "title": "Scalable Reinforcement Learning of Localized Policies for Multi-Agent Networked Systems",
                    "abstract": "We study reinforcement learning (RL) in a setting with a network of agents whose states and actions interact in a local manner where the objective is to find localized policies such that the (discounted) global reward is maximized. A fundamental challenge in this setting is that the state-action space size scales exponentially in the number of agents, rendering the problem intractable for large networks. In this paper, we propose a Scalable Actor-Critic (SAC) framework that exploits the network structure and finds a localized policy that is a $O(\\rho^\\kappa)$-approximation of a stationary point of the objective for some $\\rho\\in(0,1)$, with complexity that scales with the local state-action space size of the largest $\\kappa$-hop neighborhood of the network."
                },
                {
                    "title": "Deep Reinforcement Learning for Event-Driven Multi-Agent Decision Processes",
                    "abstract": "The incorporation of macro-actions (temporally extended actions) into multi-agent decision problems has the potential to address the curse of dimensionality associated with such decision problems. Since macro-actions last for stochastic durations, multiple agents executing decentralized policies in cooperative environments must act asynchronously. We present an algorithm that modifies generalized advantage estimation for temporally extended actions, allowing a state-of-the-art policy optimization algorithm to optimize policies in Dec-POMDPs in which agents act asynchronously. We show that our algorithm is capable of learning optimal policies in two cooperative domains, one involving real-time bus holding control and one involving wildfire fighting with unmanned aircraft. Our algorithm works by framing problems as \u201cevent-driven decision processes,\u201d which are scenarios in which the sequence and timing of actions and events are random and governed by an underlying stochastic process. In addition to optimizing policies with continuous state and action spaces, our algorithm also facilitates the use of event-driven simulators, which do not require time to be discretized into time-steps. We demonstrate the benefit of using event-driven simulation in the context of multiple agents taking asynchronous actions. We show that fixed time-step simulation risks obfuscating the sequence in which closely separated events occur, adversely affecting the policies learned. In addition, we show that arbitrarily shrinking the time-step scales poorly with the number of agents."
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "Pricing and Optimization in Shared Vehicle Systems: An Approximation Framework",
                    "abstract": "Optimizing shared vehicle systems (bike-sharing/car-sharing/ride-sharing) is more challenging compared to traditional resource allocation settings due to the presence of complex network externalities. In particular, changes in the demand/supply at any location (via dynamic pricing, rebalancing of empty vehicles, etc.) affect future supply throughout the system within short timescales. Such externalities are well captured by steady-state Markovian models, which are therefore widely used to analyze and design shared vehicle systems. However, using such models to design pricing/control policies is computationally difficult since the resulting optimization problems are high-dimensional and non-convex. To this end, we develop a general approximation framework for designing pricing policies in shared vehicle systems, based on a novel convex relaxation which we term elevated flow relaxation. Our approach provides the first efficient algorithms with rigorous approximation guarantees for a wide range of objective functions (throughput, revenue, welfare). For any shared vehicle system with $n$ stations and m vehicles, our framework provides a pricing policy with an approximation ratio of 1+(n-1)/m. This guarantee is particularly meaningful when m/n, the average number of vehicles per station is large, as is often the case in practice. Further, the simplicity of our approach allows us to extend it to more complex settings. Apart from pricing, shared vehicle systems enable other control levers for modulating demand and supply, e.g. rebalancing empty vehicles, redirecting riders to nearby vehicles, etc. Our approach yields efficient algorithms with the same approximation guarantees for all these problems, and in the process, obtains as special cases several existing heuristics and asymptotic guarantees. We also extend our approach to obtain bi-criterion guarantees in multi-objective settings; we illustrate this with the example of Ramsey pricing. From a technical perspective, our work develops a new approach for obtaining control policies with approximation guarantees in steady-state Markovian models. Our approach can be distilled into the following three-step program: (i) construct an upper bound via a relaxation to the original problem that encodes essential conservation laws of the system, (ii) identify a family of control policies inducing known steady-state distributions that achieve the value of the relaxed solution in an appropriate scaling limit (in our case, state-independent policies in the limit m++), and (iii) characterize the performance loss between the finite system (i.e. fixed m) and the scaling limit. This technique may be of independent interest for other settings."
                },
                {
                    "title": "CVXPY: A Python-Embedded Modeling Language for Convex Optimization",
                    "abstract": "CVXPY is a domain-specific language for convex optimization embedded in Python. It allows the user to express convex optimization problems in a natural syntax that follows the math, rather than in the restrictive standard form required by solvers. CVXPY makes it easy to combine convex optimization with high-level features of Python such as parallelism and object-oriented design. CVXPY is available at http://www.cvxpy.org/ under the GPL license, along with documentation and examples."
                },
                {
                    "title": "Control of Patient Flow in Emergency Departments, or Multiclass Queues with Deadlines and Feedback",
                    "abstract": "We consider the control of patient flow through physicians in emergency departments (EDs). The physicians must choose between catering to patients right after triage, who are yet to be checked, and those who are in process (IP) and are occasionally returning to be checked. Physician capacity is thus modeled as a queueing system with multiclass customers, where some of the classes face deadline constraints on their time-till-first-service, whereas the other classes feedback through service while incurring congestion costs. We consider two types of such costs: first, costs that are incurred at queue-dependent rates and second, costs that are functions of IP sojourn time. The former is our base model, which paves the way for the latter (perhaps more ED realistic). In both cases, we propose and analyze scheduling policies that, asymptotically in conventional heavy traffic, minimize congestion costs while adhering to all deadline constraints. Our policies have two parts: the first chooses between triage and IP..."
                },
                {
                    "title": "ON PATIENT FLOW IN HOSPITALS: A DATA-BASED QUEUEING-SCIENCE PERSPECTIVE",
                    "abstract": "Hospitals are complex systems with essential societal benefits and huge mounting costs. These costs are exacerbated by inefficiencies in hospital processes, which are often manifested by congestion and long delays in patient care. Thus, a queueing-network view of patient flow in hospitals is natural for studying and improving its performance. The goal of our research is to explore patient flow data through the lens of a queueing scientist. The means is exploratory data analysis (EDA) in a large Israeli hospital, which reveals important features that are not readily explainable by existing models. Questions raised by our EDA include: Can a simple (parsimonious) queueing model usefully capture the complex operational reality of the Emergency Department (ED)? What time scales and operational regimes are relevant for modeling patient length of stay in the Internal Wards (IWs)? How do protocols of patient transfer between the ED and the IWs influence patient delay, workload division and fairness? EDA also unde..."
                },
                {
                    "title": "Robust Fluid Processing Networks",
                    "abstract": "Fluid models provide a tractable and useful approach in approximating multiclass processing networks. However, they ignore the inherent stochasticity in arrival and service processes. To address this shortcoming, we develop a robust fluid approach to the control of processing networks. We provide insights into the mathematical structure, modeling power, tractability, and performance of the resulting model. Specifically, we show that the robust fluid model preserves the computational tractability of the classical fluid problem and retains its original structure. From the robust fluid model, we derive a (scheduling) policy that regulates how fluid from various classes is processed at the servers of the network. We present simulation results to compare the performance of our policies to several commonly used traditional methods. The results demonstrate that our robust fluid policies are near-optimal (when the optimal can be computed) and outperform policies obtained directly from the fluid model and heuristic alternatives (when it is computationally intractable to compute the optimal)."
                },
                {
                    "title": "Performance Modeling and Design of Computer Systems: Queueing Theory in Action",
                    "abstract": "Computer systems design is full of conundrums: Given a choice between a single machine with speed s, or n machines each with speed s/n, which should we choose? If both the arrival rate and service rate double, will the mean response time stay the same? Should systems really aim to balance load, or is this a convenient myth? If a scheduling policy favors one set of jobs, does it necessarily hurt some other jobs, or are these \"conservation laws\" being misinterpreted? Do greedy, shortest-delay, routing strategies make sense in a server farm, or is what's good for the individual disastrous for the system as a whole? How do high job size variability and heavy-tailed workloads affect the choice of a scheduling policy? How should one trade off energy and delay in designing a computer system? If 12 servers are needed to meet delay guarantees when the arrival rate is 9 jobs/sec, will we need 12,000 servers when the arrival rate is 9,000 jobs/sec? Tackling the questions that systems designers care about, this book brings queueing theory decisively back to computer science. The book is written with computer scientists and engineers in mind and is full of examples from computer systems, as well as manufacturing and operations research. Fun and readable, the book is highly approachable, even for undergraduates, while still being thoroughly rigorous and also covering a much wider span of topics than many queueing books. Readers benefit from a lively mix of motivation and intuition, with illustrations, examples, and more than 300 exercises - all while acquiring the skills needed to model, analyze, and design large-scale systems with good performance and low cost. The exercises are an important feature, teaching research-level counterintuitive lessons in the design of computer systems. The goal is to train readers not only to customize existing analyses but also to invent their own."
                },
                {
                    "title": "A Stochastic Model for Order Book Dynamics",
                    "abstract": "We propose a continuous-time stochastic model for the dynamics of a limit order book. The model strikes a balance between three desirable features: it can be estimated easily from data, it captures key empirical properties of order book dynamics, and its analytical tractability allows for fast computation of various quantities of interest without resorting to simulation. We describe a simple parameter estimation procedure based on high-frequency observations of the order book and illustrate the results on data from the Tokyo Stock Exchange. Using simple matrix computations and Laplace transform methods, we are able to efficiently compute probabilities of various events, conditional on the state of the order book: an increase in the midprice, execution of an order at the bid before the ask quote moves, and execution of both a buy and a sell order at the best quotes before the price moves. Using high-frequency data, we show that our model can effectively capture the short-term dynamics of a limit order book. We also evaluate the performance of a simple trading strategy based on our results."
                },
                {
                    "title": "The Modern Call Center: A Multi\u2010Disciplinary Perspective on Operations Management Research",
                    "abstract": "Call centers are an increasingly important part of today's business world, employing millions of agents across the globe and serving as a primary customer\u2010facing channel for firms in many different industries. Call centers have been a fertile area for operations management researchers in several domains, including forecasting, capacity planning, queueing, and personnel scheduling. In addition, as telecommunications and information technology have advanced over the past several years, the operational challenges faced by call center managers have become more complicated. Issues associated with human resources management, sales, and marketing have also become increasingly relevant to call center operations and associated academic research. In this paper, we provide a survey of the recent literature on call center operations management. Along with traditional research areas, we pay special attention to new management challenges that have been caused by emerging technologies, to behavioral issues associated with both call center agents and customers, and to the interface between call center operations and sales and marketing. We identify a handful of broad themes for future investigation while also pointing out several very specific research opportunities."
                },
                {
                    "title": "Queueing Theory for Semiconductor Manufacturing Systems: A Survey and Open Problems",
                    "abstract": "This paper surveys applications of queueing theory for semiconductor manufacturing systems (SMSs). Due to sophisticated tool specifications and process flows in semiconductor manufacturing, queueing models can be very complicated. Research efforts have been on the improvement of model assumptions and model input, mainly in the first moment (averages) and the second moment (variations). However, practices show that implementation of classical queueing theory in semiconductor industry has been unsatisfactory. In this paper, open problems on queueing modeling of SMS are discussed. A potential solution is also proposed by relaxing the independent assumptions in the classical queueing theory. Cycle time reduction has constantly been a key focus of semiconductor manufacturing. Compared with simulation, queueing theory-based analytical modeling is much faster in estimating manufacturing system performance and providing more insights for performance improvement. Therefore, queueing modeling attracts generous semiconductor research grants. Unfortunately, existing queueing models focus on simple extensions of the classical queueing theory and fail to question its applicability to the complicated SMS. Hence, related researches have not been employed widely in the semiconductor industry. In this paper, we conduct a survey on the important works and also present some open problems. We also propose a novel solution by relaxing a key assumption in the classical queueing theory. We are currently funded by Intel to explore this potential solution, and we hope it can foster an interesting research field for the years to come."
                },
                {
                    "title": "Maximum Pressure Policies in Stochastic Processing Networks",
                    "abstract": "Complex systems like semiconductor wafer fabrication facilities (fabs), networks of data switches, and large-scale call centers all demand efficient resource allocation. Deterministic models like linear programs (LP) have been used for capacity planning at both the design and expansion stages of such systems. LP-based planning is critical in setting a medium range or long-term goal for many systems, but it does not translate into a day-to-day operational policy that must deal with discreteness of jobs and the randomness of the processing environment.A stochastic processing network, advanced by J. Michael Harrison (2000, 2002, 2003), is a system that takes inputs of materials of various kinds and uses various processing resources to produce outputs of materials of various kinds. Such a network provides a powerful abstraction of a wide range of real-world systems. It provides high-fidelity stochastic models in diverse economic sectors including manufacturing, service, and information technology.We propose a family of maximum pressure service policies for dynamically allocating service capacities in a stochastic processing network. Under a mild assumption on network structure, we prove that a network operating under a maximum pressure policy achieves maximum throughput predicted by LPs. These policies are semilocal in the sense that each server makes its decision based on the buffer content in its serviceable buffers and their immediately downstream buffers. In particular, their implementation does not use arrival rate information, which is difficult to collect in many applications. We also identify a class of networks for which the nonpreemptive, non-processor-splitting version of a maximum pressure policy is still throughput optimal. Applications to queueing networks with alternate routes and networks of data switches are presented."
                },
                {
                    "title": "Scheduling Flexible Servers with Convex Delay Costs: Heavy-Traffic Optimality of the Generalized c\u00b5-Rule",
                    "abstract": "We consider a queueing system with multitype customers and flexible (multiskilled) servers that work in parallel. IfQ iis the queue length of typei customers, this queue incurs cost at the rate ofC i ( Q i ), whereC i (.) is increasing and convex. We analyze the system in heavy traffic (Harrison and Lopez 1999) and show that a very simple generalizedc\u00c2\u00b5-rule (Van Mieghem 1995) minimizes both instantaneous and cumulative queueing costs, asymptotically, over essentially all scheduling disciplines, preemptive or non-preemptive. This rule aims at myopically maximizing the rate of decrease of the instantaneous cost at all times, which translates into the following: when becoming free, serverj chooses for service a typei customer such thati ? arg max i C' i ( Q i )\u00c2\u00b5 ij , where \u00c2\u00b5 ijis the average service rate of typei customers by serverj.An analogous version of the generalizedc\u00c2\u00b5-rule asymptotically minimizes delay costs. To this end, let the cost incurred by a typei customer be an increasing convex functionC i ( D) of its sojourn timeD. Then, serverj always chooses for service a customer for which the value ofC' i ( D) \u00c2\u00b5 ijis maximal, whereD andi are the customer's sojourn time and type, respectively."
                },
                {
                    "title": "MaxWeight scheduling in a generalized switch: State space collapse and workload minimization in heavy traffic",
                    "abstract": "In this work we consider a problem related to the equilibrium Statistical Mechanics of Spin Glasses, namely the study of the Gibbs measure of the Random Energy Model. For solving this problem new results of independent interest on sums of spacings for i.i.d. Gaussian random variables are presented. Then we give a precise description of the support of the Gibbs measure below the critical temperature."
                },
                {
                    "title": "The Power of Two Choices in Randomized Load Balancing",
                    "abstract": "We consider the following natural model: customers arrive as a Poisson stream of rate /spl lambda/n, /spl lambda/<1, at a collection of n servers. Each customer chooses some constant d servers independently and uniformly at random from the n servers and waits for service at the one with the fewest customers. Customers are served according to the first-in first-out (FIFO) protocol and the service time for a customer is exponentially distributed with mean 1. We call this problem the supermarket model. We wish to know how the system behaves and in particular we are interested in the effect that the parameter d has on the expected time a customer spends in the system in equilibrium. Our approach uses a limiting, deterministic model representing the behavior as n/spl rarr//spl infin/ to approximate the behavior of finite systems. The analysis of the deterministic model is interesting in its own right. Along with a theoretical justification of this approach, we provide simulations that demonstrate that the method accurately predicts system behavior, even for relatively small systems. Our analysis provides surprising implications. Having d=2 choices leads to exponential improvements in the expected time a customer spends in the system over d=1, whereas having d=3 choices is only a constant factor better than d=2. We discuss the possible implications for system design."
                },
                {
                    "title": "The iSLIP scheduling algorithm for input-queued switches",
                    "abstract": "An increasing number of high performance internetworking protocol routers, LAN and asynchronous transfer mode (ATM) switches use a switched backplane based on a crossbar switch. Most often, these systems use input queues to hold packets waiting to traverse the switching fabric. It is well known that if simple first in first out (FIFO) input queues are used to hold packets then, even under benign conditions, head-of-line (HOL) blocking limits the achievable bandwidth to approximately 58.6% of the maximum. HOL blocking can be overcome by the use of virtual output queueing, which is described in this paper. A scheduling algorithm is used to configure the crossbar switch, deciding the order in which packets will be served. Previous results have shown that with a suitable scheduling algorithm, 100% throughput can be achieved. In this paper, we present a scheduling algorithm called iSLIP. An iterative, round-robin algorithm, iSLIP can achieve 100% throughput for uniform traffic, yet is simple to implement in hardware. Iterative and noniterative versions of the algorithms are presented, along with modified versions for prioritized traffic. Simulation results are presented to indicate the performance of iSLIP under benign and bursty traffic conditions. Prototype and commercial implementations of iSLIP exist in systems with aggregate bandwidths ranging from 50 to 500 Gb/s. When the traffic is nonuniform, iSLIP quickly adapts to a fair scheduling policy that is guaranteed never to starve an input queue. Finally, we describe the implementation complexity of iSLIP. Based on a two-dimensional (2-D) array of priority encoders, single-chip schedulers have been built supporting up to 32 ports, and making approximately 100 million scheduling decisions per second."
                },
                {
                    "title": "Heavy traffic analysis of a system with parallel servers: asymptotic optimality of discrete-review policies",
                    "abstract": "This paper is concerned with dynamic scheduling in a queueing system that has two independent Poisson input streams, two servers, deterministic service times and linear holding costs. One server can process both classes of incoming jobs, but the other can process only one class, and the service time for the shared job class is different depending on which server is involved. A bound on system performance is developed in terms of a single pooled resource, or super-server, whose capabilities combine those of the original two servers. Thereafter, attention is focused on the heavy traf(cid:12)c regime, where the combined capacity of the two servers is approximately equal to the total input rate. We construct a discrete-review control policy and show that if its parameters are chosen correctly as one approaches the heavy traf(cid:12)c limit, then its cost performance approaches the bound associated with a single pooled resource. Thus the discrete-review policy is proved to be asymptotically optimal in the heavy traf(cid:12)c limit. Although resource pooling in heavy traf(cid:12)c has been observed to occur in other network scheduling problems, there have been very few studies that rigorously proved the pooling phenomenon, or that proved the asymptotic optimality of a speci(cid:12)c policy. Our discrete-review policy is obtained by applying a general method, called the BIGSTEP method in an earlier paper, to the parallel-server model."
                },
                {
                    "title": "Reinforcement Learning for Dynamic Channel Allocation in Cellular Telephone Systems",
                    "abstract": "In cellular telephone systems, an important problem is to dynamically allocate the communication resource (channels) so as to maximize service in a stochastic caller environment. This problem is naturally formulated as a dynamic programming problem and we use a reinforcement learning (RL) method to find dynamic channel allocation policies that are better than previous heuristic solutions. The policies obtained perform well for a broad variety of call traffic patterns. We present results on a large cellular system with approximately 4949 states."
                },
                {
                    "title": "Heavy Traffic Convergence of a Controlled, Multiclass Queueing System",
                    "abstract": "This paper provides a rigorous proof of the connection between the optimal sequencing problem for a two-station, two-customer-class queueing network and the problem of control of a multidimensional diffusion process, obtained as a heavy traffic limit of the queueing problem. In particular, the diffusion problem, which is one of ``singular control'' of a Brownian motion, is used to develop policies which are shown to be asymptotically nearly optimal as the traffic intensity approaches one in the queueing network. The results are proved by a viscosity solution analysis of the related Hamilton--Jacobi--Bellman equations."
                },
                {
                    "title": "Stability and convergence of moments for multiclass queueing networks via fluid limit models",
                    "abstract": "The subject of this paper is open multiclass queueing networks, which are common models of communication networks, and complex manufacturing systems such as wafer fabrication facilities. We provide sufficient conditions for the existence of bounds on long-run average moments of the queue lengths at the various stations, and we bound the rate of convergence of the mean queue length to its steady-state value. Our work provides a solid foundation for performance analysis either by analytical methods or by simulation. These results are applied to several examples including re-entrant lines, generalized Jackson networks, and a general polling model as found in computer networks applications. >"
                },
                {
                    "title": "Wide-area traffic: the failure of Poisson modeling",
                    "abstract": "Network arrivals are often modeled as Poisson processes for analytic simplicity, even though a number of traffic studies have shown that packet interarrivals are not exponentially distributed. We evaluate 21 wide-area traces, investigating a number of wide-area TCP arrival processes (session and connection arrivals, FTPDATA connection arrivals within FTP sessions, and TELNET packet arrivals) to determine the error introduced by modeling them using Poisson processes. We find that user-initiated TCP session arrivals, such as remote-login and file-transfer, are well-modeled as Poisson processes with fixed hourly rates, but that other connection arrivals deviate considerably from Poisson; that modeling TELNET packet interarrivals as exponential grievously underestimates the burstiness of TELNET traffic, but using the empirical Tcplib[DJCME92] interarrivals preserves burstiness over many time scales; and that FTPDATA connection arrivals within FTP sessions come bunched into \u201cconnection burst\u201d, the largest of which are so large that they completely dominate FTPDATA traffic. Finally, we offer some preliminary results regarding how our findings relate to the possible self-similarity of wide-area traffic."
                },
                {
                    "title": "Dynamic Scheduling of a Multiclass Fluid Network",
                    "abstract": "A fluid network is a deterministic network model in which dynamic continuous flows are circulated and processed among a set of stations. A fluid network often describes the asymptotic behavior of a stochastic queueing network via functional strong law of large numbers. We study the dynamic scheduling of multiple classes of fluid traffic in such a network. An algorithm is developed that systematically solves the dynamic scheduling problem by solving a sequence of linear programs. It generates a policy, in the form of dynamic capacity allocation at each station among all fluid classes, that consists of a finite set of linear \"pieces\" over the entire time horizon. In a single-station, or equivalently, single-server, network, this solution procedure recovers the priority index set that is optimal for the corresponding discrete queueing model, generally known as Klimov's problem."
                },
                {
                    "title": "Stability properties of constrained queueing systems and scheduling policies for maximum throughput in multihop radio networks",
                    "abstract": "The stability of a queuing network with interdependent servers is considered. The dependency among the servers is described by defining subsets of servers that can be activated simultaneously. Packet radio networks provide the motivation for the consideration of these systems. The problem of scheduling server activation under the constraints imposed by the dependency among them is studied. The stability region of the network under a specific scheduling policy is the set of vectors of arrival rates in the queues of the system for which the stochastic process of the queue lengths is ergodic. The 'supremum' (i.e., union) of the stability regions over all the policies is characterized. Then a scheduling policy is specified, the stability region of which is equal to the union of the stability regions over all the policies. Finally, the behavior of the network for arrival rates outside the stability region is studied.<<ETX>>"
                },
                {
                    "title": "Scheduling Networks of Queues: Heavy Traffic Analysis of a Two-Station Closed Network",
                    "abstract": "We consider a multiclass closed queueing network with two single-server stations. Each class requires service at a particular station, and customers change class after service according to specified probabilities. There is a general service time distribution for each class. The problem is to schedule the two servers to maximize the long-run average throughput of the network. By assuming a large customer population and nearly balanced loading of the two stations, the scheduling problem can be approximated by a dynamic control problem involving Brownian motion. A reformulation of this control problem is solved exactly and the solution is interpreted in terms of the queueing network to obtain a scheduling rule. We conjecture, quite naturally, that the resulting scheduling rule is asymptotically optimal under heavy traffic conditions, but no attempt is made to prove that. The scheduling rule is a static priority policy that computes an index for each class and awards higher priority at station 1 respectively, station 2 to classes with the smaller respectively, larger values of this index. An analytical comparison of this rule to any other static policy is also obtained. An example is given that illustrates the procedure and demonstrates its effectiveness."
                },
                {
                    "title": "The c# rule revisited",
                    "abstract": "On considere un probleme generalisant celui etudie par Baras, Dorsey et Makowski (1985) et on demontre l'optimalite de la regle c\u03bc"
                },
                {
                    "title": "Structural Estimation of Load Balancing Behavior in Inpatient Ward Network",
                    "abstract": "In managing inpatient ward beds, decision makers often face high levels of bed utilization and large variability in demand. Load balancing via off-service placement is one of the primary operational levers to alleviate congestion, yet this practice is associated with considerable costs. Therefore, it is imperative to understand how much decision makers exercise load balancing in practice and its implications on system performance. In this paper, we take the structural estimation approach and empirically quantify the causal effect of inpatient wards\u2019 occupancy on patient routing decisions. We leverage the control function method to overcome the endogeneity problem of occupancy in estimation. Our results reveal a significant load balancing effect, and a large difference between the load balancing behavior during daytime and nighttime. Based on the estimated routing policy, we further study how changes to the load balancing behavior affect system performance. Our findings provide new perspectives to manage congestion in the inpatient ward network. We find that more consistent load balancing behavior throughout the day leads to overall better system performance. Moreover, we find that allowing for heterogeneous load balancing factors for different off-service wards based on the underlying network structure provides hospital managers with a low cost and low risk way to reduce congestion in the busiest wards in the network."
                },
                {
                    "title": "Decision , Risk & Operations Working Papers Series Approximate and Data-Driven Dynamic Programming for Queueing Networks",
                    "abstract": "We develop an approach based on temporal difference learning to address scheduling problems in complex queueing networks such as those arising in service, communication, and manufacturing systems. One novel feature is the selection of basis functions, which is motivated by the gross behavior of the system in asymptotic regimes. Another is the use of polytopic structure to efficiently identify desired actions from an intractable set of alternatives. Application to input-queued crossbar switch models with up to hundreds of queues and quadrillions of alternative actions yield scheduling policies outperforming a heuristic recently shown to have certain optimality properties in the heavy traffic scale. We also extend the approach to a setting where aspects of the queueing network are not modeled and we must rely instead on empirical data. This data-driven approach is useful, for example, when the statistical structure of arrivals is poorly understood but historical data traces are available."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.SY",
                "eess.SY"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop effective reinforcement learning-based controllers for queuing networks that dynamically allocate resources in the presence of high stochasticity and non-stationary workloads?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of traditional queuing policies that are often rigid and not data-driven. By advancing the use of reinforcement learning in queuing network control, we can enhance the adaptability and efficiency of systems across various domains, such as semiconductor manufacturing, healthcare, and cloud computing. This research could lead to significant improvements in service quality and operational efficiency, paving the way for future studies that explore more complex and dynamic environments.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of queuing networks, which can involve multiple job classes, server types, and varying processing speeds. The stochastic nature of workloads introduces unpredictability, making it difficult to develop robust scheduling policies. Naive approaches may fail because they do not account for the intricate interactions between jobs and servers or the temporal dynamics of demand. Additionally, the long decision horizons in queuing systems require sophisticated planning and learning mechanisms that traditional reinforcement learning algorithms may not adequately address.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on simple, theoretically sound policies that are limited to specific network structures and objectives. These approaches often lack the flexibility needed to adapt to the complexities of real-world queuing systems. Barriers such as the high dimensionality of state and action spaces, as well as the need for extensive data to train models effectively, have hindered progress. Our approach differs by leveraging advanced reinforcement learning techniques that can learn from data in a more generalized manner, allowing for the development of controllers that are better suited to handle the unique challenges of queuing networks.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a reinforcement learning framework tailored for queuing network control, utilizing a simulation environment like QGym to benchmark various queuing policies. We will employ a diverse set of datasets representing different queuing scenarios and measure performance using metrics such as average waiting time, throughput, and service quality. The expected outcomes include the identification of effective, data-driven queuing policies that outperform traditional methods, demonstrating the potential of reinforcement learning to enhance resource allocation in complex, stochastic environments."
            }
        },
        "author_data": {
            "d3da8e80-c9fc-438e-9e62-a4f156f67f44": {
                "pk": "d3da8e80-c9fc-438e-9e62-a4f156f67f44",
                "name": "Haozhe Chen",
                "collaborators": [
                    "Carl Vondrick",
                    "Chengzhi Mao",
                    "Run Chen",
                    "Julia Hirschberg",
                    "Zhaotong Shen",
                    "Qian Yu",
                    "Cheng Wang",
                    "Binyan Jiang",
                    "Junfeng Yang",
                    "Weiming Zhang"
                ],
                "domain": [
                    "Text-to-Speech",
                    "Machine Learning",
                    "Model Interpretability",
                    "Dimensionality Reduction"
                ],
                "publications": [
                    {
                        "title": "EmoKnob: Enhance Voice Cloning with Fine-Grained Emotion Control",
                        "abstract": "While recent advances in Text-to-Speech (TTS) technology produce natural and expressive speech, they lack the option for users to select emotion and control intensity. We propose EmoKnob, a framework that allows fine-grained emotion control in speech synthesis with few-shot demonstrative samples of arbitrary emotion. Our framework leverages the expressive speaker representation space made possible by recent advances in foundation voice cloning models. Based on the few-shot capability of our emotion control framework, we propose two methods to apply emotion control on emotions described by open-ended text, enabling an intuitive interface for controlling a diverse array of nuanced emotions. To facilitate a more systematic emotional speech synthesis field, we introduce a set of evaluation metrics designed to rigorously assess the faithfulness and recognizability of emotion control frameworks. Through objective and subjective evaluations, we show that our emotion control framework effectively embeds emotions into speech and surpasses emotion expressiveness of commercial TTS services."
                    },
                    {
                        "title": "Exponential Euler method for stiff SDEs driven by fractional Brownian motion",
                        "abstract": "In a recent paper by Kamrani et al. (2024), exponential Euler method for stiff stochastic differential equations with additive fractional Brownian noise was discussed, and the convergence order close to the Hurst parameter H was proved. Utilizing the technique of Malliavin derivative, we prove the exponential Euler scheme and obtain a convergence order of one, which is the optimal rate in numerical simulation."
                    },
                    {
                        "title": "SelfIE: Self-Interpretation of Large Language Model Embeddings",
                        "abstract": "How do large language models (LLMs) obtain their answers? The ability to explain and control an LLM's reasoning process is key for reliability, transparency, and future model developments. We propose SelfIE (Self-Interpretation of Embeddings), a framework that enables LLMs to interpret their own embeddings in natural language by leveraging their ability to respond to inquiries about a given passage. Capable of interpreting open-world concepts in the hidden embeddings, SelfIE reveals LLM internal reasoning in cases such as making ethical decisions, internalizing prompt injection, and recalling harmful knowledge. SelfIE's text descriptions on hidden embeddings also open up new avenues to control LLM reasoning. We propose Supervised Control, which allows editing open-ended concepts while only requiring gradient computation of individual layer. We extend RLHF to hidden embeddings and propose Reinforcement Control that erases harmful knowledge in LLM without supervision targets."
                    },
                    {
                        "title": "HiQR: An efficient algorithm for high-dimensional quadratic regression with penalties",
                        "abstract": "This paper investigates the efficient solution of penalized quadratic regressions in high-dimensional settings. A novel and efficient algorithm for ridge-penalized quadratic regression is proposed, leveraging the matrix structures of the regression with interactions. Additionally, an alternating direction method of multipliers (ADMM) framework is developed for penalized quadratic regression with general penalties, including both single and hybrid penalty functions. The approach simplifies the calculations to basic matrix-based operations, making it appealing in terms of both memory storage and computational complexity for solving penalized quadratic regressions in high-dimensional settings."
                    },
                    {
                        "title": "Interpreting and Controlling Vision Foundation Models via Text Explanations",
                        "abstract": "Large-scale pre-trained vision foundation models, such as CLIP, have become de facto backbones for various vision tasks. However, due to their black-box nature, understanding the underlying rules behind these models' predictions and controlling model behaviors have remained open challenges. We present a framework for interpreting vision transformer's latent tokens with natural language. Given a latent token, our framework retains its semantic information to the final layer using transformer's local operations and retrieves the closest text for explanation. Our approach enables understanding of model visual reasoning procedure without needing additional model training or data collection. Based on the obtained interpretations, our framework allows for model editing that controls model reasoning behaviors and improves model robustness against biases and spurious correlations."
                    },
                    {
                        "title": "Speech Pattern based Black-box Model Watermarking for Automatic Speech Recognition",
                        "abstract": "As an effective method for intellectual property (IP) protection, model watermarking technology has been applied on a wide variety of deep neural networks (DNN), including speech classification models. However, how to design a black-box watermarking scheme for automatic speech recognition (ASR) models is still an unsolved problem, which is a significant demand for protecting remote ASR Application Programming Interface (API) deployed in cloud servers. Due to conditional independence assumption and label-detection-based evasion attack risk of ASR models, the black-box model watermarking scheme for speech classification models cannot apply to ASR models. In this paper, we propose the first black-box model watermarking framework for protecting the IP of ASR models. Specifically, we synthesize trigger audios by spreading the speech clips of model owners over the entire input audios and labeling the trigger audios with the stego texts, which hides the authorship information with linguistic steganography. Experiments on the state-of-the-art open-source ASR system DeepSpeech demonstrate the feasibility of the proposed watermarking scheme, which is robust against five kinds of attacks and has little impact on accuracy."
                    },
                    {
                        "title": "Noisy Data Visualization using Functional Data Analysis",
                        "abstract": "Data visualization via dimensionality reduction is an important tool in exploratory data analysis. However, when the data are noisy, many existing methods fail to capture the underlying structure of the data. The method called Empirical Intrinsic Geometry (EIG) was previously proposed for performing dimensionality reduction on high dimensional dynamical processes while theoretically eliminating all noise. However, implementing EIG in practice requires the construction of high-dimensional histograms, which suffer from the curse of dimensionality. Here we propose a new data visualization method called Functional Information Geometry (FIG) for dynamical processes that adapts the EIG framework while using approaches from functional data analysis to mitigate the curse of dimensionality. We experimentally demonstrate that the resulting method outperforms a variant of EIG designed for visualization in terms of capturing the true structure, hyperparameter robustness, and computational speed. We then use our method to visualize EEG brain measurements of sleep activity."
                    }
                ]
            },
            "c65fd534-01a8-4159-85b9-dbb22a7decb6": {
                "pk": "c65fd534-01a8-4159-85b9-dbb22a7decb6",
                "name": "Ang Li",
                "collaborators": [
                    "Judea Pearl",
                    "Yeshaiahu Fainman",
                    "Zejian Yuan",
                    "Xinle Shang",
                    "Xu Gao",
                    "Jianmin Dong",
                    "Christos Masouros",
                    "Wim Bogaerts",
                    "Simon Su",
                    "Mu Liang"
                ],
                "domain": [
                    "Causal Inference",
                    "Machine Learning",
                    "Graph Theory",
                    "Photonics"
                ],
                "publications": [
                    {
                        "title": "There is a deep 1-generic set",
                        "abstract": "An infinite binary sequence is Bennett deep if, for any computable time bound, the difference between the time-bounded prefix-free Kolmogorov complexity and the prefix-free Kolmogorov complexity of its initial segments is eventually unbounded. It is known that weakly 2-generic sets are shallow, i.e. not deep. In this paper, we show that there is a deep 1-generic set."
                    },
                    {
                        "title": "The $v_1$-Periodic Region of the Complex-Motivic Ext",
                        "abstract": "We establish a $v_1$-periodicity theorem in Ext over the complex-motivic Steenrod algebra. The element $h_1$ of Ext, which detects the homotopy class $\\eta$ in the motivic Adams spectral sequence, is non-nilpotent and therefore generates $h_1$-towers. Our result is that, apart from these $h_1$-towers, $v_1$-periodicity behaves as it does classically."
                    },
                    {
                        "title": "Introenumerability, autoreducibility, and randomness",
                        "abstract": "We define $\\Psi$-autoreducible sets given an autoreduction procedure $\\Psi$. Then, we show that for any $\\Psi$, a measurable class of $\\Psi$-autoreducible sets has measure zero. Using this, we show that classes of cototal, uniformly introenumerable, introenumerable, and hyper-cototal enumeration degrees all have measure zero. By analyzing the arithmetical complexity of the classes of cototal sets and cototal enumeration degrees, we show that weakly 2-random sets cannot be cototal and weakly 3-random sets cannot be of cototal enumeration degree. Then, we see that this result is optimal by showing that there exists a 1-random cototal set and a 2-random set of cototal enumeration degree. For uniformly introenumerable degrees and introenumerable degrees, we utilize $\\Psi$-autoreducibility again to show the optimal result that no weakly 3-random sets can have introenumerable enumeration degree. We also show that no 1-random set can be introenumerable."
                    },
                    {
                        "title": "Countable Ordered Groups and Weihrauch Reducibility",
                        "abstract": "This paper continues to study the connection between reverse mathematics and Weihrauch reducibility. In particular, we study the problems formed from Maltsev's theorem on the order types of countable ordered groups. Solomon showed that the theorem is equivalent to $\\mathsf{\\Pi_1^1}$-$\\mathsf{CA_0}$, the strongest of the big five subsystems of second order arithmetic. We show that the strength of the theorem comes from having a dense linear order without endpoints in its order type. Then, we show that for the related Weihrauch problem to be strong enough to be equivalent to $\\mathsf{\\hat{WF}}$ (the analog problem of $\\mathsf{\\Pi_1^1}$-$\\mathsf{CA_0}$), an order-preserving function is necessary in the output. Without the order-preserving function, the problems are very much to the side compared to analog problems of the big five."
                    },
                    {
                        "title": "Unit Selection: Case Study and Comparison with A/B Test Heuristic",
                        "abstract": "The unit selection problem defined by Li and Pearl identifies individuals who have desired counterfactual behavior patterns, for example, individuals who would respond positively if encouraged and would not otherwise. Li and Pearl showed by example that their unit selection model is beyond the A/B test heuristics. In this paper, we reveal the essence of the A/B test heuristics, which are exceptional cases of the benefit function defined by Li and Pearl. Furthermore, We provided more simulated use cases of Li-Pearl's unit selection model to help decision-makers apply their model correctly, explaining that A/B test heuristics are generally problematic."
                    },
                    {
                        "title": "Unit Selection with Causal Diagram",
                        "abstract": "The unit selection problem aims to identify a set of individuals who are most likely to exhibit a desired mode of behavior, for example, selecting individuals who would respond one way if encouraged and a different way if not encouraged. Using a combination of experimental and observational data, Li and Pearl derived tight bounds on the \"benefit function\" - the payoff/cost associated with selecting an individual with given characteristics. This paper shows that these bounds can be narrowed significantly (enough to change decisions) when structural information is available in the form of a causal model. We address the problem of estimating the benefit function using observational and experimental data when specific graphical criteria are assumed to hold."
                    },
                    {
                        "title": "Unit Selection with Nonbinary Treatment and Effect",
                        "abstract": "The unit selection problem aims to identify a set of individuals who are most likely to exhibit a desired mode of behavior, for example, selecting individuals who would respond one way if encouraged and a different way if not encouraged. Using a combination of experimental and observational data, Li and Pearl derived tight bounds on the \"benefit function\", which is the payoff/cost associated with selecting an individual with given characteristics. This paper extends the benefit function to the general form such that the treatment and effect are not restricted to binary. We propose an algorithm to test the identifiability of the nonbinary benefit function and an algorithm to compute the bounds of the nonbinary benefit function using experimental and observational data."
                    },
                    {
                        "title": "Bounds on Causal Effects and Application to High Dimensional Data",
                        "abstract": "This paper addresses the problem of estimating causal effects when adjustment variables in the back-door or front-door criterion are partially observed. For such scenarios, we derive bounds on the causal effects by solving two non-linear optimization problems, and demonstrate that the bounds are sufficient. Using this optimization method, we propose a framework for dimensionality reduction that allows one to trade bias for estimation power, and demonstrate its performance using simulation studies."
                    },
                    {
                        "title": "Probabilities of Causation with Nonbinary Treatment and Effect",
                        "abstract": "This paper deals with the problem of estimating the probabilities of causation when treatment and effect are not binary. Tian and Pearl derived sharp bounds for the probability of necessity and sufficiency (PNS), the probability of sufficiency (PS), and the probability of necessity (PN) using experimental and observational data. In this paper, we provide theoretical bounds for all types of probabilities of causation to multivalued treatments and effects. We further discuss examples where our bounds guide practical decisions and use simulation studies to evaluate how informative the bounds are for various combinations of data."
                    },
                    {
                        "title": "Probabilities of Causation: Role of Observational Data",
                        "abstract": "Probabilities of causation play a crucial role in modern decision-making. Pearl defined three binary probabilities of causation, the probability of necessity and sufficiency (PNS), the probability of sufficiency (PS), and the probability of necessity (PN). These probabilities were then bounded by Tian and Pearl using a combination of experimental and observational data. However, observational data are not always available in practice; in such a case, Tian and Pearl's Theorem provided valid but less effective bounds using pure experimental data. In this paper, we discuss the conditions that observational data are worth considering to improve the quality of the bounds. More specifically, we defined the expected improvement of the bounds by assuming the observational distributions are uniformly distributed on their feasible interval. We further applied the proposed theorems to the unit selection problem defined by Li and Pearl."
                    },
                    {
                        "title": "On-chip spectrometer using stratified waveguides filters",
                        "abstract": "We present an ultra-compact single-shot spectrometer on silicon platform with broad operation bandwidth and high resolution. It consists of 32 stratified waveguide filters (SWFs) with diverse transmission spectra for sampling the unknown spectrum of the input signal and a specially designed ultra-compact structure for splitting the incident signal into 32 filters with low imbalance. Each SWF has a footprint less than 1um x 30um, while the 1x32 splitter and 32 filters in total occupy an area of about 35um x 260um, which to the best of our knowledge, is the smallest footprint spectrometer realized on silicon photonic platform. Experimental characteristics of the fabricated spectrometer demonstrate a broad operating bandwidth of 180nm centered at 1550nm and narrowband peaks with 0.45nm Full-Width-Half-Maximum (FWHM) can be clearly resolved. This concept can also be implemented using other material platforms for operation in optical spectral bands of interest for various applications."
                    },
                    {
                        "title": "SymmNet: A Symmetric Convolutional Neural Network for Occlusion Detection",
                        "abstract": "Detecting the occlusion from stereo images or video frames is important to many computer vision applications. Previous efforts focus on bundling it with the computation of disparity or optical flow, leading to a chicken-and-egg problem. In this paper, we leverage convolutional neural network to liberate the occlusion detection task from the interleaved, traditional calculation framework. We propose a Symmetric Network (SymmNet) to directly exploit information from an image pair, without estimating disparity or motion in advance. The proposed network is structurally left-right symmetric to learn the binocular occlusion simultaneously, aimed at jointly improving both results. The comprehensive experiments show that our model achieves state-of-the-art results on detecting the stereo and motion occlusion."
                    },
                    {
                        "title": "Revisiting the post-glitch relaxation of the 2000 Vela glitch with the neutron star equation of states in the Brueckner and relativistic Brueckner theories",
                        "abstract": "We revisit the short-term post-glitch relaxation of the Vela 2000 glitch in the simple two-component model of pulsar glitch by making use of the latest realistic equations of states from the microscopic Brueckner and the relativistic Brueckner theories for neutron stars, which can reconcile with the available astrophysical constraints. We show that to fit both the glitch size and the post-glitch jumps in frequency derivatives approximately one minute after the glitch, the mass of the Vela pulsar is necessarily small, and there may be demands for a stiff equation of state (which results in a typical stellar radius larger than $\\sim\\negthickspace12.5\\,\\rm km$) and a strong suppression of the pairing gap in the nuclear medium. We discuss the implications of this result on the understanding of pulsar glitches."
                    },
                    {
                        "title": "The stable Picard group of finite Adams Hopf algebroids with an application to the $\\mathbb{R}$-motivic Steenrod subalgebra $\\mathcal{A}(1)^{\\mathbb{R}}$",
                        "abstract": "In this paper, we investigate the rigidity of the stable comodule category of a specific class of Hopf algebroids known as finite Adams, shedding light on its Picard group. Then we establish a reduction process through base changes, enabling us to effectively compute the Picard group of the $\\mathbb{R}$-motivic mod $2$ Steenrod subalgebra $\\mathcal{A}(1)^{\\mathbb{R}}$. Our computation shows that $\\operatorname{Pic}(\\mathcal{A}(1)^{\\mathbb{R}})$ is isomorphic to $\\mathbb{Z}^4$, where two ranks come from the motivic grading, one from the algebraic loop functor, and the last is generated by the $\\mathbb{R}$-motivic joker $J$."
                    },
                    {
                        "title": "Rotating NSs/QSs and recent astrophysical observations",
                        "abstract": "Both fast and slow configurations of rotating neutron stars (NSs) are studied with the recently-constructed unified NS EoSs. The calculations for pure quark stars (QSs) and hybrid stars (HSs) are also done, using several updated quark matter EoSs and Gibbs construction for obtaining hadron-quark mixed phase. All three types of EoSs fulfill the recent 2-solar-mass constrain. By confronting the glitch observations with the theoretical calculations for the crustal moment of inertia (MoI), we find that the glitch crisis is still present in Vela-like pulsars. An upcoming accurate MoI measurement (eg., a possible 10\\% accuracy for pulsar PSR J0737-3039A) could distinguish QSs from NSs, since the MoIs of QSs are generally $> \\sim 1.5 $ times larger than NSs and HSs, no matter the compactness and the mass of the stars. Using tabulated EoSs, we compute stationary and equilibrium sequences of rapidly rotating, relativistic stars in general relativity from the well-tested $rns$ code, assuming the matter comprising the star to be a perfect fluid. All three observed properties of the short gamma-ray bursts (SGRBs) internal plateaus sample are simulated using the rotating configurations of NSs/QSs as basic input. We finally argue that for some characteristic SGRBs, the post-merger products of NS-NS mergers are probably supramassive QSs rather than NSs, and NS-NS mergers are a plausible location for quark deconfinement and the formation of QSs."
                    },
                    {
                        "title": "Interference Exploitation Precoding Made Practical: Closed-Form Solutions with Optimal Performance",
                        "abstract": "In this paper, we propose closed-form precoding schemes with optimal performance for constructive interference (CI) exploitation in the multiuser multiple-input single-output (MU-MISO) downlink. We first consider an optimization where we maximize the distance between the constructive region and the detection thresholds. The cases of both strict and non-strict phase rotation are considered and can further be formulated as convex optimization problems. For optimization with strict phase rotation, we mathematically derive the optimal beamforming structure with Lagrangian and Karush-Kuhn-Tucker (KKT) conditions. By formulating its dual problem, the optimization problem is further shown to be equivalent to a quadratic programming (QP) over a simplex, which can be solved more efficiently. We then extend our analysis to the case of non-strict phase rotation, where it is mathematically shown that a K-dimensional optimization for non-strict phase rotation is equivalent to a 2K-dimensional optimization for strict phase rotation in terms of the problem formulation. The connection with the conventional zero-forcing (ZF) precoding is also discussed. Based on the above analysis, we further propose an iterative closed-form scheme to obtain the optimal beamforming matrix, where within each iteration a closed-form solution can be obtained. Numerical results validate our analysis and the optimality of the proposed iterative scheme, and further show that the proposed closed-form scheme is more efficient than the conventional QP algorithms with interior-point methods, which motivates the use of CI beamforming in practical wireless systems."
                    },
                    {
                        "title": "A photonic integrated circuits with reconfigurablenonreciprocal transmission and all-opticalfunctionalities",
                        "abstract": "We present a photonics integrated circuit on silicon substrate withreconfigurable nonreciprocal transmission that exhibits a large isolationratio and low insertion loss. It also offers ability for all-optical function-alities, like optical computing gates, or a flip-flop. The circuit is a mixedcavity system of which the linear transmission can be tuned as a Fanoresonance or electromagnetically induced transparency (EIT) using twointegrated heaters. With high optical intensity inside the cavity, the Fanoresonance and EIT peak induce a strong distortion due to thermal non-linearities in the cavity, and these distortions depend on the transmissiondirection due to the asymmetric power distribution in the cavities. Theresulting large isolation ratio is attributed to the inherent sharp slope ofthe Fano resonance and the large extinction ratio of the EIT peak. Thus,a high-power forward-propagating signal will trigger the nonreciprocalphenomenon for low-loss transmission, while backward transmission willsee high loss irrespective of its power level, which is an outstanding im-provement upon previously reported nonlinearity induced nonreciprocityin silicon photonics. The reconfigurability of the high transmission direc-tion comes from the efficient control of the mode excitation and couplinginside the cavity using the integrated heaters. Also, by using a separatepump laser, the device could be developed for all-optical functions likeswitching, logic and computing."
                    },
                    {
                        "title": "Accelerating Binarized Neural Networks via Bit-Tensor-Cores in Turing GPUs",
                        "abstract": "Despite foreseeing tremendous speedups over conventional deep neural networks, the performance advantage of binarized neural networks (BNNs) has merely been showcased on general-purpose processors such as CPUs and GPUs. In fact, due to being unable to leverage bit-level-parallelism with a word-based architecture, GPUs have been criticized for extremely low utilization (1%) when executing BNNs. Consequently, the latest tensorcores in NVIDIA Turing GPUs start to experimentally support bit computation. In this work, we look into this brand new bit computation capability and characterize its unique features. We show that the stride of memory access can significantly affect performance delivery and a data-format co-design is highly desired to support the tensorcores for achieving superior performance than existing software solutions without tensorcores. We realize the tensorcore-accelerated BNN design, particularly the major functions for fully-connect and convolution layers -- bit matrix multiplication and bit convolution. Evaluations on two NVIDIA Turing GPUs show that, with ResNet-18, our BTC-BNN design can process ImageNet at a rate of 5.6K images per second, 77% faster than state-of-the-art. Our BNN approach is released on https://github.com/pnnl/TCBNN."
                    },
                    {
                        "title": "Deep Learning-Based Channel Extrapolation for Pattern Reconfigurable Massive MIMO",
                        "abstract": "Reconfigurable antennas that can dynamically change their operation state exhibit excellent adaptivity and flexibility over traditional antennas, and MIMO arrays that consist of multifunctional and reconfigurable antennas (MRAs) are foreseen as one promising solution towards future Holographic MIMO. Specifically, in pattern reconfigurable MIMO (PR-MIMO) communication systems, accurate acquisition of channel state information (CSI) of all the radiation modes is a challenging task, because using conventional pilot-based channel estimation techniques in PR-MIMO systems incurs overwhelming pilot overheads. In this letter, we leverage deep learning methods to design a PR neural network, which can use the estimated CSI for one radiation mode to infer CSIs for the other radiation modes. In order to reduce the pilot overheads, we propose a new channel estimation method specially for PR-MIMO systems, which divides the transmit antennas of PR-MIMO into groups and antennas in different groups employ different radiation modes. Compared with conventional full-connected real-valued deep neural networks (DNN), the PR neural network which uses complex-valued coefficients can work directly in the complex domain. Experiment results show that the proposed channel extrapolation method offers significant performance gains in terms of extrapolation accuracy over benchmark schemes."
                    }
                ]
            },
            "38aebfad-a251-4ed0-8ee7-6307186a32f7": {
                "pk": "38aebfad-a251-4ed0-8ee7-6307186a32f7",
                "name": "Ethan Che",
                "collaborators": [
                    "Hongseok Namkoong",
                    "Jing Dong",
                    "Jimmy Wang",
                    "Daniel R. Jiang",
                    "Xin T. Tong"
                ],
                "domain": [
                    "Adaptive Experimentation",
                    "Reinforcement Learning",
                    "Auction Theory",
                    "Stochastic Optimization"
                ],
                "publications": [
                    {
                        "title": "Distributionally Robust Optimal Auction Design under Mean Constraints",
                        "abstract": "We study a seller who sells a single good to multiple bidders with uncertainty over the joint distribution of bidders' valuations, as well as bidders' higher-order beliefs about their opponents. The seller only knows the (possibly asymmetric) means of the marginal distributions of each bidder's valuation and the range. An adversarial nature chooses the worst-case distribution within this ambiguity set along with the worst-case information structure. We find that a second-price auction with a symmetric, random reserve price obtains the optimal revenue guarantee within a broad class of mechanisms we refer to as competitive mechanisms, which include standard auction formats, including the first-price auction, with or without reserve prices. The optimal mechanism possesses two notable characteristics. First, the mechanism treats all bidders identically even in the presence of ex-ante asymmetries. Second, when bidders are identical and the number of bidders $n$ grows large, the seller's optimal reserve price converges in probability to a non-binding reserve price and the revenue guarantee converges to the best possible revenue guarantee at rate $O(1/n)$."
                    },
                    {
                        "title": "Adaptive Experimentation at Scale: A Computational Framework for Flexible Batches",
                        "abstract": "Standard bandit algorithms that assume continual reallocation of measurement effort are challenging to implement due to delayed feedback and infrastructural/organizational difficulties. Motivated by practical instances involving a handful of reallocation epochs in which outcomes are measured in batches, we develop a computation-driven adaptive experimentation framework that can flexibly handle batching. Our main observation is that normal approximations, which are universal in statistical inference, can also guide the design of adaptive algorithms. By deriving a Gaussian sequential experiment, we formulate a dynamic program that can leverage prior information on average rewards. Instead of the typical theory-driven paradigm, we leverage computational tools and empirical benchmarking for algorithm development. In particular, our empirical analysis highlights a simple yet effective algorithm, Residual Horizon Optimization, which iteratively solves a planning problem using stochastic gradient descent. Our approach significantly improves statistical power over standard methods, even when compared to Bayesian bandit algorithms (e.g., Thompson sampling) that require full distributional knowledge of individual rewards. Overall, we expand the scope of adaptive experimentation to settings that are difficult for standard methods, involving limited adaptivity, low signal-to-noise ratio, and unknown reward distributions."
                    },
                    {
                        "title": "Differentiable Discrete Event Simulation for Queuing Network Control",
                        "abstract": "Queuing network control is essential for managing congestion in job-processing systems such as service systems, communication networks, and manufacturing processes. Despite growing interest in applying reinforcement learning (RL) techniques, queueing network control poses distinct challenges, including high stochasticity, large state and action spaces, and lack of stability. To tackle these challenges, we propose a scalable framework for policy optimization based on differentiable discrete event simulation. Our main insight is that by implementing a well-designed smoothing technique for discrete event dynamics, we can compute pathwise policy gradients for large-scale queueing networks using auto-differentiation software (e.g., Tensorflow, PyTorch) and GPU parallelization. Through extensive empirical experiments, we observe that our policy gradient estimators are several orders of magnitude more accurate than typical REINFORCE-based estimators. In addition, We propose a new policy architecture, which drastically improves stability while maintaining the flexibility of neural-network policies. In a wide variety of scheduling and admission control tasks, we demonstrate that training control policies with pathwise gradients leads to a 50-1000x improvement in sample efficiency over state-of-the-art RL methods. Unlike prior tailored approaches to queueing, our methods can flexibly handle realistic scenarios, including systems operating in non-stationary environments and those with non-exponential interarrival/service times."
                    },
                    {
                        "title": "AExGym: Benchmarks and Environments for Adaptive Experimentation",
                        "abstract": "Innovations across science and industry are evaluated using randomized trials (a.k.a. A/B tests). While simple and robust, such static designs are inefficient or infeasible for testing many hypotheses. Adaptive designs can greatly improve statistical power in theory, but they have seen limited adoption due to their fragility in practice. We present a benchmark for adaptive experimentation based on real-world datasets, highlighting prominent practical challenges to operationalizing adaptivity: non-stationarity, batched/delayed feedback, multiple outcomes and objectives, and external validity. Our benchmark aims to spur methodological development that puts practical performance (e.g., robustness) as a central concern, rather than mathematical guarantees on contrived instances. We release an open source library, AExGym, which is designed with modularity and extensibility in mind to allow experimentation practitioners to develop custom environments and algorithms."
                    },
                    {
                        "title": "Mathematical Programming For Adaptive Experiments",
                        "abstract": "Adaptive experimentation can significantly improve statistical power, but standard algorithms overlook important practical issues including batched and delayed feedback, personalization, non-stationarity, multiple objectives, and constraints. To address these issues, the current algorithm design paradigm crafts tailored methods for each problem instance. Since it is infeasible to devise novel algorithms for every real-world instance, practitioners often have to resort to suboptimal approximations that do not address all of their challenges. Moving away from developing bespoke algorithms for each setting, we present a mathematical programming view of adaptive experimentation that can flexibly incorporate a wide range of objectives, constraints, and statistical procedures. By formulating a dynamic program in the batched limit, our modeling framework enables the use of scalable optimization methods (e.g., SGD and auto-differentiation) to solve for treatment allocations. We evaluate our framework on benchmarks modeled after practical challenges such as non-stationarity, personalization, multi-objectives, and constraints. Unlike bespoke algorithms such as modified variants of Thomson sampling, our mathematical programming approach provides remarkably robust performance across instances."
                    },
                    {
                        "title": "Stochastic Gradient Descent with Adaptive Data",
                        "abstract": "Stochastic gradient descent (SGD) is a powerful optimization technique that is particularly useful in online learning scenarios. Its convergence analysis is relatively well understood under the assumption that the data samples are independent and identically distributed (iid). However, applying SGD to policy optimization problems in operations research involves a distinct challenge: the policy changes the environment and thereby affects the data used to update the policy. The adaptively generated data stream involves samples that are non-stationary, no longer independent from each other, and affected by previous decisions. The influence of previous decisions on the data generated introduces bias in the gradient estimate, which presents a potential source of instability for online learning not present in the iid case. In this paper, we introduce simple criteria for the adaptively generated data stream to guarantee the convergence of SGD. We show that the convergence speed of SGD with adaptive data is largely similar to the classical iid setting, as long as the mixing time of the policy-induced dynamics is factored in. Our Lyapunov-function analysis allows one to translate existing stability analysis of stochastic systems studied in operations research into convergence rates for SGD, and we demonstrate this for queueing and inventory management problems. We also showcase how our result can be applied to study the sample complexity of an actor-critic policy gradient algorithm."
                    }
                ]
            },
            "f1f08706-17cb-4db4-903c-52975cbba9f9": {
                "pk": "f1f08706-17cb-4db4-903c-52975cbba9f9",
                "name": "Tianyi Peng",
                "collaborators": [
                    "Vivek F. Farias",
                    "Andrew A. Li",
                    "Vivek Farias",
                    "Andrew Zheng",
                    "Zhu Cao",
                    "Yang Xiao",
                    "Ciamac Moallemi",
                    "Xiao Yuan",
                    "Ge Bai",
                    "Xiongfeng Ma"
                ],
                "domain": [
                    "Quantum Mechanics",
                    "Machine Learning",
                    "Causal Inference",
                    "Anomaly Detection"
                ],
                "publications": [
                    {
                        "title": "Tight detection efficiency bounds of Bell tests in no-signaling theories",
                        "abstract": "No-signaling theories, which can contain nonlocal correlations stronger than classical correlations but limited by the no-signaling condition, have deepened our understanding of the quantum theory. In principle, the nonlocality of these theories can be verified via Bell tests. In practice, however, inefficient detectors may make Bell tests unreliable, which is called the detection efficiency loophole. In this work, we show almost tight lower and upper bounds of the detector efficiency requirement for demonstrating the nonlocality of no-signaling theories, by designing a general class of Bell tests. In particular, we show tight bounds for two scenarios: the bipartite case and the multipartite case with a large number of parties. To some extent, these tight efficiency bounds quantify the nonlocality of no-signaling theories. Furthermore, our result shows that the detector efficiency can be arbitrarily low even for Bell tests with two parties, by increasing the number of measurement settings. Our work also sheds light on the detector efficiency requirement for showing the nonlocality of the quantum theory."
                    },
                    {
                        "title": "Dark Experience for Incremental Keyword Spotting",
                        "abstract": "Spoken keyword spotting (KWS) is crucial for identifying keywords within audio inputs and is widely used in applications like Apple Siri and Google Home, particularly on edge devices. Current deep learning-based KWS systems, which are typically trained on a limited set of keywords, can suffer from performance degradation when encountering new domains, a challenge often addressed through few-shot fine-tuning. However, this adaptation frequently leads to catastrophic forgetting, where the model's performance on original data deteriorates. Progressive continual learning (CL) strategies have been proposed to overcome this, but they face limitations such as the need for task-ID information and increased storage, making them less practical for lightweight devices. To address these challenges, we introduce Dark Experience for Keyword Spotting (DE-KWS), a novel CL approach that leverages dark knowledge to distill past experiences throughout the training process. DE-KWS combines rehearsal and distillation, using both ground truth labels and logits stored in a memory buffer to maintain model performance across tasks. Evaluations on the Google Speech Command dataset show that DE-KWS outperforms existing CL baselines in average accuracy without increasing model size, offering an effective solution for resource-constrained edge devices. The scripts are available on GitHub for the future research."
                    },
                    {
                        "title": "Synthetically Controlled Bandits",
                        "abstract": "This paper presents a new dynamic approach to experiment design in settings where, due to interference or other concerns, experimental units are coarse. `Region-split' experiments on online platforms are one example of such a setting. The cost, or regret, of experimentation is a natural concern here. Our new design, dubbed Synthetically Controlled Thompson Sampling (SCTS), minimizes the regret associated with experimentation at no practically meaningful loss to inferential ability. We provide theoretical guarantees characterizing the near-optimal regret of our approach, and the error rates achieved by the corresponding treatment effect estimator. Experiments on synthetic and real world data highlight the merits of our approach relative to both fixed and `switchback' designs common to such experimental settings."
                    },
                    {
                        "title": "Quantum uncertainty relation using coherence",
                        "abstract": "Measurement outcomes of a quantum state can be genuinely random (unpredictable) according to the basic laws of quantum mechanics. The Heisenberg-Robertson uncertainty relation puts constrains on the accuracy of two noncommuting observables. The existing uncertainty relations adopt variance or entropic measures, which are functions of observed outcome distributions, to quantify the uncertainty. According to recent studies of quantum coherence, such uncertainty measures contain both classical (predictable) and quantum (unpredictable) components. In order to extract out the quantum effects, we define quantum uncertainty to be the coherence of the state on the measurement basis. We discover a quantum uncertainty relation of coherence between two measurement non-commuting bases. Furthermore, we analytically derive the quantum uncertainty relation for the qubit case with three widely adopted coherence measures, the relative entropy using coherence, the coherence of formation, and the $l_1$ norm of coherence."
                    },
                    {
                        "title": "Uncertainty Quantification For Low-Rank Matrix Completion With Heterogeneous and Sub-Exponential Noise",
                        "abstract": "The problem of low-rank matrix completion with heterogeneous and sub-exponential (as opposed to homogeneous and Gaussian) noise is particularly relevant to a number of applications in modern commerce. Examples include panel sales data and data collected from web-commerce systems such as recommendation engines. An important unresolved question for this problem is characterizing the distribution of estimated matrix entries under common low-rank estimators. Such a characterization is essential to any application that requires quantification of uncertainty in these estimates and has heretofore only been available under the assumption of homogenous Gaussian noise. Here we characterize the distribution of estimated matrix entries when the observation noise is heterogeneous sub-exponential and provide, as an application, explicit formulas for this distribution when observed entries are Poisson or Binary distributed."
                    },
                    {
                        "title": "Learning Treatment Effects in Panels with General Intervention Patterns",
                        "abstract": "The problem of causal inference with panel data is a central econometric question. The following is a fundamental version of this problem: Let $M^*$ be a low rank matrix and $E$ be a zero-mean noise matrix. For a `treatment' matrix $Z$ with entries in $\\{0,1\\}$ we observe the matrix $O$ with entries $O_{ij} := M^*_{ij} + E_{ij} + \\mathcal{T}_{ij} Z_{ij}$ where $\\mathcal{T}_{ij} $ are unknown, heterogenous treatment effects. The problem requires we estimate the average treatment effect $\\tau^* := \\sum_{ij} \\mathcal{T}_{ij} Z_{ij} / \\sum_{ij} Z_{ij}$. The synthetic control paradigm provides an approach to estimating $\\tau^*$ when $Z$ places support on a single row. This paper extends that framework to allow rate-optimal recovery of $\\tau^*$ for general $Z$, thus broadly expanding its applicability. Our guarantees are the first of their type in this general setting. Computational experiments on synthetic and real-world data show a substantial advantage over competing estimators."
                    },
                    {
                        "title": "Speeding up Policy Simulation in Supply Chain RL",
                        "abstract": "Simulating a single trajectory of a dynamical system under some state-dependent policy is a core bottleneck in policy optimization algorithms. The many inherently serial policy evaluations that must be performed in a single simulation constitute the bulk of this bottleneck. To wit, in applying policy optimization to supply chain optimization (SCO) problems, simulating a single month of a supply chain can take several hours.   We present an iterative algorithm for policy simulation, which we dub Picard Iteration. This scheme carefully assigns policy evaluation tasks to independent processes. Within an iteration, a single process evaluates the policy only on its assigned tasks while assuming a certain 'cached' evaluation for other tasks; the cache is updated at the end of the iteration. Implemented on GPUs, this scheme admits batched evaluation of the policy on a single trajectory. We prove that the structure afforded by many SCO problems allows convergence in a small number of iterations, independent of the horizon. We demonstrate practical speedups of 400x on large-scale SCO problems even with a single GPU, and also demonstrate practical efficacy in other RL environments."
                    },
                    {
                        "title": "Fixing Inventory Inaccuracies At Scale",
                        "abstract": "Inaccurate records of inventory occur frequently, and by some measures cost retailers approximately 4% in annual sales. Detecting inventory inaccuracies manually is cost-prohibitive, and existing algorithmic solutions rely almost exclusively on learning from longitudinal data, which is insufficient in the dynamic environment induced by modern retail operations. Instead, we propose a solution based on cross-sectional data over stores and SKUs, observing that detecting inventory inaccuracies can be viewed as a problem of identifying anomalies in a (low-rank) Poisson matrix. State-of-the-art approaches to anomaly detection in low-rank matrices apparently fall short. Specifically, from a theoretical perspective, recovery guarantees for these approaches require that non-anomalous entries be observed with vanishingly small noise (which is not the case in our problem, and indeed in many applications).   So motivated, we propose a conceptually simple entry-wise approach to anomaly detection in low-rank Poisson matrices. Our approach accommodates a general class of probabilistic anomaly models. We show that the cost incurred by our algorithm approaches that of an optimal algorithm at a min-max optimal rate. Using synthetic data and real data from a consumer goods retailer, we show that our approach provides up to a 10x cost reduction over incumbent approaches to anomaly detection. Along the way, we build on recent work that seeks entry-wise error guarantees for matrix completion, establishing such guarantees for sub-exponential matrices, a result of independent interest."
                    }
                ]
            },
            "faf98b74-c61d-438c-9d42-ac48660b7a84": {
                "pk": "faf98b74-c61d-438c-9d42-ac48660b7a84",
                "name": "Jing Dong",
                "collaborators": [
                    "Jose Blanchet",
                    "Yi Zhu",
                    "Shuai Li",
                    "Baoxiang Wang",
                    "Rouba Ibrahim",
                    "Zhaoyang Lv",
                    "Yan Chen",
                    "Yuichi Yoshida",
                    "Yanan Pei",
                    "Xinyun Chen"
                ],
                "domain": [
                    "Queueing Theory",
                    "Stochastic Processes",
                    "Machine Learning",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Sampling Point Processes on Stable Unbounded Regions and Exam Simulation of Queues",
                        "abstract": "Given a marked renewal point process (assuming that the marks are i.i.d.) we say that an unbounded region is stable if it contains finitely many points of the point process with probability one. In this paper we provide algorithms that allow to sample these finitely many points efficiently. We explain how exact simulation of the steady-state measure valued state descriptor of the infinite server queue follows as a simple corollary of our algorithms. We provide numerical evidence supporting that our algorithms are not only theoretically sound but also practical. Finally, having simulation optimization in mind, we also apply our results to gradient estimation of steady-state performance measures."
                    },
                    {
                        "title": "Three Asymptotic Regimes for Ranking and Selection with General Sample Distributions",
                        "abstract": "In this paper, we study three asymptotic regimes that can be applied to ranking and selection (R&S) problems with general sample distributions. These asymptotic regimes are constructed by sending particular problem parameters (probability of incorrect selection, smallest difference in system performance that we deem worth detecting) to zero. We establish asymptotic validity and efficiency of the corresponding R&S procedures in each regime. We also analyze the connection among different regimes and compare the pre-limit performances of corresponding algorithms."
                    },
                    {
                        "title": "Perfect sampling for infinite server and loss systems",
                        "abstract": "We present the first class of perfect sampling (also known as exact simulation) algorithms for the steady-state distribution of non-Markovian loss networks. We use a variation of Dominated Coupling From The Past for which we simulate a stationary infinite server queue backwards in time and analyze the running time in heavy traffic. In particular, we are able to simulate stationary renewal marked point processes in unbounded regions. We use the infinite server queue as an upper bound process to simulate loss systems. The running time analysis of our perfect sampling algorithm for loss systems is performed in the Quality-Driven (QD) and the Quality-and-Efficiency-Driven regimes. In both cases, we show that our algorithm achieves sub-exponential complexity as both the number of servers and the arrival rate increase. Moreover, in the QD regime, our algorithm achieves a nearly optimal rate of convergence."
                    },
                    {
                        "title": "On the SRPT Scheduling Discipline in Many-Server Queues with Impatient Customers",
                        "abstract": "The shortest-remaining-processing-time (SRPT) scheduling policy has been extensively studied, for more than 50 years, in single-server queues with infinitely patient jobs. Yet, much less is known about its performance in multiserver queues. In this paper, we present the first theoretical analysis of SRPT in multiserver queues with abandonment. In particular, we consider the M/GI/s+GI queue and demonstrate that, in the many-sever overloaded regime, performance in the SRPT queue is equivalent, asymptotically in steady state, to a preemptive two-class priority queue where customers with short service times (below a threshold) are served without wait, and customers with long service times (above a threshold) eventually abandon without service. We prove that the SRPT discipline maximizes, asymptotically, the system throughput, among all scheduling disciplines. We also compare the performance of the SRPT policy to blind policies and study the effects of the patience-time and service-time distributions."
                    },
                    {
                        "title": "miniSAM: A Flexible Factor Graph Non-linear Least Squares Optimization Framework",
                        "abstract": "Many problems in computer vision and robotics can be phrased as non-linear least squares optimization problems represented by factor graphs, for example, simultaneous localization and mapping (SLAM), structure from motion (SfM), motion planning, and control. We have developed an open-source C++/Python framework miniSAM, for solving such factor graph based least squares problems. Compared to most existing frameworks for least squares solvers, miniSAM has (1) full Python/NumPy API, which enables more agile development and easy binding with existing Python projects, and (2) a wide list of sparse linear solvers, including CUDA enabled sparse linear solvers. Our benchmarking results shows miniSAM offers comparable performances on various types of problems, with more flexible and smoother development experience."
                    },
                    {
                        "title": "Scheduling with Service-Time Information: The Power of Two Priority Classes",
                        "abstract": "Utilizing customers' service-time information, we study an easy-to-implement scheduling policy with two priority classes. By carefully designing the classes, the two-class priority rule achieves near-optimal performance. In particular, for a single-server queue, as the traffic intensity approaches 1, the policy achieves a scaling for the queue length processes that is similar to the shortest remaining processing time first policy. Our analysis quantifies how the tail of the service time distribution affects the benefits one can gain from service-time-based scheduling policies. When the service times are misspecified, we further quantify how imperfect observation of the service time affects the performance of the two-class priority rule through both theoretical and numerical analysis. Our results demonstrate the robustness of the two-class priority rule. Specifically, even with imperfect service-time information, the two-class priority rules can still achieve substantial performance improvement over the first come first served."
                    },
                    {
                        "title": "A Batch-to-Online Transformation under Random-Order Model",
                        "abstract": "We introduce a transformation framework that can be utilized to develop online algorithms with low $\\epsilon$-approximate regret in the random-order model from offline approximation algorithms. We first give a general reduction theorem that transforms an offline approximation algorithm with low average sensitivity to an online algorithm with low $\\epsilon$-approximate regret. We then demonstrate that offline approximation algorithms can be transformed into a low-sensitivity version using a coreset construction method. To showcase the versatility of our approach, we apply it to various problems, including online $(k,z)$-clustering, online matrix approximation, and online regression, and successfully achieve polylogarithmic $\\epsilon$-approximate regret for each problem. Moreover, we show that in all three cases, our algorithm also enjoys low inconsistency, which may be desired in some online applications."
                    },
                    {
                        "title": "On Constructing Confidence Region for Model Parameters in Stochastic Gradient Descent via Batch Means",
                        "abstract": "In this paper, we study a simple algorithm to construct asymptotically valid confidence regions for model parameters using the batch means method. The main idea is to cancel out the covariance matrix which is hard/costly to estimate. In the process of developing the algorithm, we establish process-level functional central limit theorem for Polyak-Ruppert averaging based stochastic gradient descent estimators. We also extend the batch means method to accommodate more general batch size specifications."
                    },
                    {
                        "title": "Perfect Sampling of GI/GI/c Queues",
                        "abstract": "We introduce the first class of perfect sampling algorithms for the steady-state distribution of multi-server queues with general interarrival time and service time distributions. Our algorithm is built on the classical dominated coupling from the past protocol. In particular, we use a coupled multi-server vacation system as the upper bound process and develop an algorithm to simulate the vacation system backwards in time from stationarity at time zero. The algorithm has finite expected termination time with mild moment assumptions on the interarrival time and service time distributions."
                    },
                    {
                        "title": "Incentivizing an Unknown Crowd",
                        "abstract": "Motivated by the common strategic activities in crowdsourcing labeling, we study the problem of sequential eliciting information without verification (EIWV) for workers with a heterogeneous and unknown crowd. We propose a reinforcement learning-based approach that is effective against a wide range of settings including potential irrationality and collusion among workers. With the aid of a costly oracle and the inference method, our approach dynamically decides the oracle calls and gains robustness even under the presence of frequent collusion activities. Extensive experiments show the advantage of our approach. Our results also present the first comprehensive experiments of EIWV on large-scale real datasets and the first thorough study of the effects of environmental variables."
                    },
                    {
                        "title": "$\u03b5$-Strong Simulation for Multidimensional Stochastic Differential Equations via Rough Path Analysis",
                        "abstract": "Consider a multidimensional diffusion process $X=\\{X\\left(t\\right) :t\\in\\lbrack0,1]\\}$. Let $\\varepsilon>0$ be a \\textit{deterministic}, user defined, tolerance error parameter. Under standard regularity conditions on the drift and diffusion coefficients of $X$, we construct a probability space, supporting both $X$ and an explicit, piecewise constant, fully simulatable process $X_{\\varepsilon}$ such that \\[ \\sup_{0\\leq t\\leq1}\\left\\Vert X_{\\varepsilon}\\left(t\\right) -X\\left(t\\right) \\right\\Vert_{\\infty}<\\varepsilon \\] with probability one. Moreover, the user can adaptively choose $\\varepsilon^{\\prime}\\in\\left(0,\\varepsilon\\right) $ so that $X_{\\varepsilon^{\\prime}}$ (also piecewise constant and fully simulatable) can be constructed conditional on $X_{\\varepsilon}$ to ensure an error smaller than $\\varepsilon^{\\prime}>0$ with probability one. Our construction requires a detailed study of continuity estimates of the Ito map using Lyon's theory of rough paths. We approximate the underlying Brownian motion, jointly with the L\\'{e}vy areas with a deterministic $\\varepsilon$ error in the underlying rough path metric."
                    },
                    {
                        "title": "Observation of Fundamental Thermal Noise in Optical Fibers down to Infrasonic Frequencies",
                        "abstract": "The intrinsic thermal noise in optical fibers is the ultimate limit of fiber-based systems. However, at infrasonic frequencies, the spectral behavior of the intrinsic thermal noise remains unclear so far. We present the measurements of the fundamental thermal noise in optical fibers obtained using a balanced fiber Michelson interferometer. When an ultra-stable laser is used as the laser source and other noise sources are carefully controlled, the 1/f spectral density of thermal noise is observed down to infrasonic frequencies and the measured magnitude is consistent with the theoretical predictions at the frequencies from 0.2 Hz to 20 kHz. Moreover, as observed in the experiment, the level of 1/f thermal noise is reduced by changing the coating of optical fibers. Therefore, a possible way to reduce the thermal noise in optical fibers at low Fourier frequencies is indicated. Finally, the inconsistency between the experimental data on thermomechanical noise and existing theory is discussed."
                    },
                    {
                        "title": "Distributed Thompson Sampling",
                        "abstract": "We study a cooperative multi-agent multi-armed bandits with M agents and K arms. The goal of the agents is to minimized the cumulative regret. We adapt a traditional Thompson Sampling algoirthm under the distributed setting. However, with agent's ability to communicate, we note that communication may further reduce the upper bound of the regret for a distributed Thompson Sampling approach. To further improve the performance of distributed Thompson Sampling, we propose a distributed Elimination based Thompson Sampling algorithm that allow the agents to learn collaboratively. We analyse the algorithm under Bernoulli reward and derived a problem dependent upper bound on the cumulative regret."
                    },
                    {
                        "title": "Combinatorial Bandits under Strategic Manipulations",
                        "abstract": "Strategic behavior against sequential learning methods, such as \"click framing\" in real recommendation systems, have been widely observed. Motivated by such behavior we study the problem of combinatorial multi-armed bandits (CMAB) under strategic manipulations of rewards, where each arm can modify the emitted reward signals for its own interest. This characterization of the adversarial behavior is a relaxation of previously well-studied settings such as adversarial attacks and adversarial corruption. We propose a strategic variant of the combinatorial UCB algorithm, which has a regret of at most $O(m\\log T + m B_{max})$ under strategic manipulations, where $T$ is the time horizon, $m$ is the number of arms, and $B_{max}$ is the maximum budget of an arm. We provide lower bounds on the budget for arms to incur certain regret of the bandit algorithm. Extensive experiments on online worker selection for crowdsourcing systems, online influence maximization and online recommendations with both synthetic and real datasets corroborate our theoretical findings on robustness and regret bounds, in a variety of regimes of manipulation budgets."
                    },
                    {
                        "title": "Sparse Gaussian Processes for Continuous-Time Trajectory Estimation on Matrix Lie Groups",
                        "abstract": "Continuous-time trajectory representations are a powerful tool that can be used to address several issues in many practical simultaneous localization and mapping (SLAM) scenarios, like continuously collected measurements distorted by robot motion, or during with asynchronous sensor measurements. Sparse Gaussian processes (GP) allow for a probabilistic non-parametric trajectory representation that enables fast trajectory estimation by sparse GP regression. However, previous approaches are limited to dealing with vector space representations of state only. In this technical report we extend the work by Barfoot et al. [1] to general matrix Lie groups, by applying constant-velocity prior, and defining locally linear GP. This enables using sparse GP approach in a large space of practical SLAM settings. In this report we give the theory and leave the experimental evaluation in future publications."
                    }
                ]
            },
            "b2b2a776-5c31-4d26-b450-a26665a56c41": {
                "pk": "b2b2a776-5c31-4d26-b450-a26665a56c41",
                "name": "Hongseok Namkoong",
                "collaborators": [
                    "John Duchi",
                    "Aman Sinha",
                    "Ethan Che",
                    "Steve Yadlowsky",
                    "Yibo Zeng",
                    "Jiashuo Liu",
                    "Matthew O'Kelly",
                    "Tiffany Tianhui Cai",
                    "Sookyo Jeong",
                    "Tatsunori Hashimoto"
                ],
                "domain": [
                    "Stochastic Optimization",
                    "Distributional Robustness",
                    "Machine Learning",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "Variance-based regularization with convex objectives",
                        "abstract": "We develop an approach to risk minimization and stochastic optimization that provides a convex surrogate for variance, allowing near-optimal and computationally efficient trading between approximation and estimation error. Our approach builds off of techniques for distributionally robust optimization and Owen's empirical likelihood, and we provide a number of finite-sample and asymptotic results characterizing the theoretical performance of the estimator. In particular, we show that our procedure comes with certificates of optimality, achieving (in some scenarios) faster rates of convergence than empirical risk minimization by virtue of automatically balancing bias and variance. We give corroborating empirical evidence showing that in practice, the estimator indeed trades between variance and absolute performance on a training sample, improving out-of-sample (test) performance over standard empirical risk minimization for a number of classification problems."
                    },
                    {
                        "title": "Learning Models with Uniform Performance via Distributionally Robust Optimization",
                        "abstract": "A common goal in statistics and machine learning is to learn models that can perform well against distributional shifts, such as latent heterogeneous subpopulations, unknown covariate shifts, or unmodeled temporal effects. We develop and analyze a distributionally robust stochastic optimization (DRO) framework that learns a model providing good performance against perturbations to the data-generating distribution. We give a convex formulation for the problem, providing several convergence guarantees. We prove finite-sample minimax upper and lower bounds, showing that distributional robustness sometimes comes at a cost in convergence rates. We give limit theorems for the learned parameters, where we fully specify the limiting distribution so that confidence intervals can be computed. On real tasks including generalizing to unknown subpopulations, fine-grained recognition, and providing good tail performance, the distributionally robust approach often exhibits improved performance."
                    },
                    {
                        "title": "Assessing External Validity Over Worst-case Subpopulations",
                        "abstract": "Study populations are typically sampled from limited points in space and time, and marginalized groups are underrepresented. To assess the external validity of randomized and observational studies, we propose and evaluate the worst-case treatment effect (WTE) across all subpopulations of a given size, which guarantees positive findings remain valid over subpopulations. We develop a semiparametrically efficient estimator for the WTE that analyzes the external validity of the augmented inverse propensity weighted estimator for the average treatment effect. Our cross-fitting procedure leverages flexible nonparametric and machine learning-based estimates of nuisance parameters and is a regular root-$n$ estimator even when nuisance estimates converge more slowly. On real examples where external validity is of core concern, our proposed framework guards against brittle findings that are invalidated by unanticipated population shifts."
                    },
                    {
                        "title": "Adaptive Experimentation at Scale: A Computational Framework for Flexible Batches",
                        "abstract": "Standard bandit algorithms that assume continual reallocation of measurement effort are challenging to implement due to delayed feedback and infrastructural/organizational difficulties. Motivated by practical instances involving a handful of reallocation epochs in which outcomes are measured in batches, we develop a computation-driven adaptive experimentation framework that can flexibly handle batching. Our main observation is that normal approximations, which are universal in statistical inference, can also guide the design of adaptive algorithms. By deriving a Gaussian sequential experiment, we formulate a dynamic program that can leverage prior information on average rewards. Instead of the typical theory-driven paradigm, we leverage computational tools and empirical benchmarking for algorithm development. In particular, our empirical analysis highlights a simple yet effective algorithm, Residual Horizon Optimization, which iteratively solves a planning problem using stochastic gradient descent. Our approach significantly improves statistical power over standard methods, even when compared to Bayesian bandit algorithms (e.g., Thompson sampling) that require full distributional knowledge of individual rewards. Overall, we expand the scope of adaptive experimentation to settings that are difficult for standard methods, involving limited adaptivity, low signal-to-noise ratio, and unknown reward distributions."
                    },
                    {
                        "title": "Distributionally Robust Losses for Latent Covariate Mixtures",
                        "abstract": "While modern large-scale datasets often consist of heterogeneous subpopulations -- for example, multiple demographic groups or multiple text corpora -- the standard practice of minimizing average loss fails to guarantee uniformly low losses across all subpopulations. We propose a convex procedure that controls the worst-case performance over all subpopulations of a given size. Our procedure comes with finite-sample (nonparametric) convergence guarantees on the worst-off subpopulation. Empirically, we observe on lexical similarity, wine quality, and recidivism prediction tasks that our worst-case procedure learns models that do well against unseen subpopulations."
                    },
                    {
                        "title": "Statistics of Robust Optimization: A Generalized Empirical Likelihood Approach",
                        "abstract": "We study statistical inference and distributionally robust solution methods for stochastic optimization problems, focusing on confidence intervals for optimal values and solutions that achieve exact coverage asymptotically. We develop a generalized empirical likelihood framework---based on distributional uncertainty sets constructed from nonparametric $f$-divergence balls---for Hadamard differentiable functionals, and in particular, stochastic optimization problems. As consequences of this theory, we provide a principled method for choosing the size of distributional uncertainty regions to provide one- and two-sided confidence intervals that achieve exact coverage. We also give an asymptotic expansion for our distributionally robust formulation, showing how robustification regularizes problems by their variance. Finally, we show that optimizers of the distributionally robust formulations we study enjoy (essentially) the same consistency properties as those in classical sample average approximations. Our general approach applies to quickly mixing stationary sequences, including geometrically ergodic Harris recurrent Markov chains."
                    },
                    {
                        "title": "Off-policy Policy Evaluation For Sequential Decisions Under Unobserved Confounding",
                        "abstract": "When observed decisions depend only on observed features, off-policy policy evaluation (OPE) methods for sequential decision making problems can estimate the performance of evaluation policies before deploying them. This assumption is frequently violated due to unobserved confounders, unrecorded variables that impact both the decisions and their outcomes. We assess robustness of OPE methods under unobserved confounding by developing worst-case bounds on the performance of an evaluation policy. When unobserved confounders can affect every decision in an episode, we demonstrate that even small amounts of per-decision confounding can heavily bias OPE methods. Fortunately, in a number of important settings found in healthcare, policy-making, operations, and technology, unobserved confounders may primarily affect only one of the many decisions made. Under this less pessimistic model of one-decision confounding, we propose an efficient loss-minimization-based procedure for computing worst-case bounds, and prove its statistical consistency. On two simulated healthcare examples---management of sepsis patients and developmental interventions for autistic children---where this is a reasonable model of confounding, we demonstrate that our method invalidates non-robust results and provides meaningful certificates of robustness, allowing reliable selection of policies even under unobserved confounding."
                    },
                    {
                        "title": "LLM Embeddings Improve Test-time Adaptation to Tabular $Y|X$-Shifts",
                        "abstract": "For tabular datasets, the change in the relationship between the label and covariates ($Y|X$-shifts) is common due to missing variables (a.k.a. confounders). Since it is impossible to generalize to a completely new and unknown domain, we study models that are easy to adapt to the target domain even with few labeled examples. We focus on building more informative representations of tabular data that can mitigate $Y|X$-shifts, and propose to leverage the prior world knowledge in LLMs by serializing (write down) the tabular data to encode it. We find LLM embeddings alone provide inconsistent improvements in robustness, but models trained on them can be well adapted/finetuned to the target domain even using 32 labeled observations. Our finding is based on a comprehensive and systematic study consisting of 7650 source-target pairs and benchmark against 261,000 model configurations trained by 22 algorithms. Our observation holds when ablating the size of accessible target data and different adaptation strategies. The code is available at https://github.com/namkoong-lab/LLM-Tabular-Shifts."
                    },
                    {
                        "title": "Distilled Thompson Sampling: Practical and Efficient Thompson Sampling via Imitation Learning",
                        "abstract": "Thompson sampling (TS) has emerged as a robust technique for contextual bandit problems. However, TS requires posterior inference and optimization for action generation, prohibiting its use in many online platforms where latency and ease of deployment are of concern. We operationalize TS by proposing a novel imitation-learning-based algorithm that distills a TS policy into an explicit policy representation, allowing fast decision-making and easy deployment in mobile and server-based environments. Using batched data collected under the imitation policy, our algorithm iteratively performs offline updates to the TS policy, and learns a new explicit policy representation to imitate it. Empirically, our imitation policy achieves performance comparable to batch TS while allowing more than an order of magnitude reduction in decision-time latency. Buoyed by low latency and simplicity of implementation, our algorithm has been successfully deployed in multiple video upload systems for Meta. Using a randomized controlled trial, we show our algorithm resulted in significant improvements in video quality and watch time."
                    },
                    {
                        "title": "Minimax Optimal Estimation of Stability Under Distribution Shift",
                        "abstract": "The performance of decision policies and prediction models often deteriorates when applied to environments different from the ones seen during training. To ensure reliable operation, we analyze the stability of a system under distribution shift, which is defined as the smallest change in the underlying environment that causes the system's performance to deteriorate beyond a permissible threshold. In contrast to standard tail risk measures and distributionally robust losses that require the specification of a plausible magnitude of distribution shift, the stability measure is defined in terms of a more intuitive quantity: the level of acceptable performance degradation. We develop a minimax optimal estimator of stability and analyze its convergence rate, which exhibits a fundamental phase shift behavior. Our characterization of the minimax convergence rate shows that evaluating stability against large performance degradation incurs a statistical cost. Empirically, we demonstrate the practical utility of our stability framework by using it to compare system designs on problems where robustness to distribution shift is critical."
                    },
                    {
                        "title": "Modeling Interference Using Experiment Roll-out",
                        "abstract": "Experiments on online marketplaces and social networks suffer from interference, where the outcome of a unit is impacted by the treatment status of other units. We propose a framework for modeling interference using a ubiquitous deployment mechanism for experiments, staggered roll-out designs, which slowly increase the fraction of units exposed to the treatment to mitigate any unanticipated adverse side effects. Our main idea is to leverage the temporal variations in treatment assignments introduced by roll-outs to model the interference structure. Since there are often multiple competing models of interference in practice we first develop a model selection method that evaluates models based on their ability to explain outcome variation observed along the roll-out. Through simulations, we show that our heuristic model selection method, Leave-One-Period-Out, outperforms other baselines. Next, we present a set of model identification conditions under which the estimation of common estimands is possible and show how these conditions are aided by roll-out designs. We conclude with a set of considerations, robustness checks, and potential limitations for practitioners wishing to use our framework."
                    },
                    {
                        "title": "Design and Scheduling of an AI-based Queueing System",
                        "abstract": "To leverage prediction models to make optimal scheduling decisions in service systems, we must understand how predictive errors impact congestion due to externalities on the delay of other jobs. Motivated by applications where prediction models interact with human servers (e.g., content moderation), we consider a large queueing system comprising of many single server queues where the class of a job is estimated using a prediction model. By characterizing the impact of mispredictions on congestion cost in heavy traffic, we design an index-based policy that incorporates the predicted class information in a near-optimal manner. Our theoretical results guide the design of predictive models by providing a simple model selection procedure with downstream queueing performance as a central concern, and offer novel insights on how to design queueing systems with AI-based triage. We illustrate our framework on a content moderation task based on real online comments, where we construct toxicity classifiers by finetuning large language models."
                    },
                    {
                        "title": "Evaluating Model Performance Under Worst-case Subpopulations",
                        "abstract": "The performance of ML models degrades when the training population is different from that seen under operation. Towards assessing distributional robustness, we study the worst-case performance of a model over all subpopulations of a given size, defined with respect to core attributes Z. This notion of robustness can consider arbitrary (continuous) attributes Z, and automatically accounts for complex intersectionality in disadvantaged groups. We develop a scalable yet principled two-stage estimation procedure that can evaluate the robustness of state-of-the-art models. We prove that our procedure enjoys several finite-sample convergence guarantees, including dimension-free convergence. Instead of overly conservative notions based on Rademacher complexities, our evaluation error depends on the dimension of Z only through the out-of-sample error in estimating the performance conditional on Z. On real datasets, we demonstrate that our method certifies the robustness of a model and prevents deployment of unreliable models."
                    },
                    {
                        "title": "In-silico Risk Analysis of Personalized Artificial Pancreas Controllers via Rare-event Simulation",
                        "abstract": "Modern treatments for Type 1 diabetes (T1D) use devices known as artificial pancreata (APs), which combine an insulin pump with a continuous glucose monitor (CGM) operating in a closed-loop manner to control blood glucose levels. In practice, poor performance of APs (frequent hyper- or hypoglycemic events) is common enough at a population level that many T1D patients modify the algorithms on existing AP systems with unregulated open-source software. Anecdotally, the patients in this group have shown superior outcomes compared with standard of care, yet we do not understand how safe any AP system is since adverse outcomes are rare. In this paper, we construct generative models of individual patients' physiological characteristics and eating behaviors. We then couple these models with a T1D simulator approved for pre-clinical trials by the FDA. Given the ability to simulate patient outcomes in-silico, we utilize techniques from rare-event simulation theory in order to efficiently quantify the performance of a device with respect to a particular patient. We show a 72,000$\\times$ speedup in simulation speed over real-time and up to 2-10 times increase in the frequency which we are able to sample adverse conditions relative to standard Monte Carlo sampling. In practice our toolchain enables estimates of the likelihood of hypoglycemic events with approximately an order of magnitude fewer simulations."
                    },
                    {
                        "title": "Diagnosing Model Performance Under Distribution Shift",
                        "abstract": "Prediction models can perform poorly when deployed to target distributions different from the training distribution. To understand these operational failure modes, we develop a method, called DIstribution Shift DEcomposition (DISDE), to attribute a drop in performance to different types of distribution shifts. Our approach decomposes the performance drop into terms for 1) an increase in harder but frequently seen examples from training, 2) changes in the relationship between features and outcomes, and 3) poor performance on examples infrequent or unseen during training. These terms are defined by fixing a distribution on $X$ while varying the conditional distribution of $Y \\mid X$ between training and target, or by fixing the conditional distribution of $Y \\mid X$ while varying the distribution on $X$. In order to do this, we define a hypothetical distribution on $X$ consisting of values common in both training and target, over which it is easy to compare $Y \\mid X$ and thus predictive performance. We estimate performance on this hypothetical distribution via reweighting methods. Empirically, we show how our method can 1) inform potential modeling improvements across distribution shifts for employment prediction on tabular census data, and 2) help to explain why certain domain adaptation methods fail to improve model performance for satellite image classification."
                    },
                    {
                        "title": "On the Need of a Modeling Language for Distribution Shifts: Illustrations on Tabular Datasets",
                        "abstract": "Different distribution shifts require different interventions, and algorithms must be grounded in the specific shifts they address. However, methodological development for robust algorithms typically relies on structural assumptions that lack empirical validation. Advocating for an empirically grounded data-driven approach to research, we build an empirical testbed comprising natural shifts across 5 tabular datasets and 60,000 method configurations encompassing imbalanced learning and distributionally robust optimization (DRO) methods. We find $Y|X$-shifts are most prevalent on our testbed, in stark contrast to the heavy focus on $X$ (covariate)-shifts in the ML literature. The performance of robust algorithms varies significantly over shift types, and is no better than that of vanilla methods. To understand why, we conduct an in-depth empirical analysis of DRO methods and find that although often neglected by researchers, implementation details -- such as the choice of underlying model class (e.g., XGBoost) and hyperparameter selection -- have a bigger impact on performance than the ambiguity set or its radius. To further bridge that gap between methodological research and practice, we design case studies that illustrate how such a data-driven, inductive understanding of distribution shifts can enhance both data-centric and algorithmic interventions."
                    },
                    {
                        "title": "Differentiable Discrete Event Simulation for Queuing Network Control",
                        "abstract": "Queuing network control is essential for managing congestion in job-processing systems such as service systems, communication networks, and manufacturing processes. Despite growing interest in applying reinforcement learning (RL) techniques, queueing network control poses distinct challenges, including high stochasticity, large state and action spaces, and lack of stability. To tackle these challenges, we propose a scalable framework for policy optimization based on differentiable discrete event simulation. Our main insight is that by implementing a well-designed smoothing technique for discrete event dynamics, we can compute pathwise policy gradients for large-scale queueing networks using auto-differentiation software (e.g., Tensorflow, PyTorch) and GPU parallelization. Through extensive empirical experiments, we observe that our policy gradient estimators are several orders of magnitude more accurate than typical REINFORCE-based estimators. In addition, We propose a new policy architecture, which drastically improves stability while maintaining the flexibility of neural-network policies. In a wide variety of scheduling and admission control tasks, we demonstrate that training control policies with pathwise gradients leads to a 50-1000x improvement in sample efficiency over state-of-the-art RL methods. Unlike prior tailored approaches to queueing, our methods can flexibly handle realistic scenarios, including systems operating in non-stationary environments and those with non-exponential interarrival/service times."
                    },
                    {
                        "title": "Certifying Some Distributional Robustness with Principled Adversarial Training",
                        "abstract": "Neural networks are vulnerable to adversarial examples and researchers have proposed many heuristic attack and defense mechanisms. We address this problem through the principled lens of distributionally robust optimization, which guarantees performance under adversarial input perturbations. By considering a Lagrangian penalty formulation of perturbing the underlying data distribution in a Wasserstein ball, we provide a training procedure that augments model parameter updates with worst-case perturbations of training data. For smooth losses, our procedure provably achieves moderate levels of robustness with little computational or statistical cost relative to empirical risk minimization. Furthermore, our statistical guarantees allow us to efficiently certify robustness for the population loss. For imperceptible perturbations, our method matches or outperforms heuristic approaches."
                    },
                    {
                        "title": "Scalable End-to-End Autonomous Vehicle Testing via Rare-event Simulation",
                        "abstract": "While recent developments in autonomous vehicle (AV) technology highlight substantial progress, we lack tools for rigorous and scalable testing. Real-world testing, the $\\textit{de facto}$ evaluation environment, places the public in danger, and, due to the rare nature of accidents, will require billions of miles in order to statistically validate performance claims. We implement a simulation framework that can test an entire modern autonomous driving system, including, in particular, systems that employ deep-learning perception and control algorithms. Using adaptive importance-sampling methods to accelerate rare-event probability evaluation, we estimate the probability of an accident under a base distribution governing standard traffic behavior. We demonstrate our framework on a highway scenario, accelerating system evaluation by $2$-$20$ times over naive Monte Carlo sampling methods and $10$-$300 \\mathsf{P}$ times (where $\\mathsf{P}$ is the number of processors) over real-world testing."
                    },
                    {
                        "title": "C-Learner: Constrained Learning for Causal Inference and Semiparametric Statistics",
                        "abstract": "Popular debiased causal estimation methods, e.g. for the average treatment effect -- such as one-step estimation (e.g., augmented inverse propensity weighting) and targeted maximum likelihood estimation -- enjoy desirable asymptotic properties such as statistical efficiency and double robustness. However, they often produce unstable estimates when there is limited overlap between treatment and control, and require ad hoc adjustments in practice (e.g., truncating propensity scores). In contrast, simple plug-in estimators are stable but lack good asymptotic properties. We propose a novel debiased estimator that achieves the best of both worlds, producing stable plug-in estimates with desirable asymptotic properties. Our constrained learning framework solves for the best plug-in estimator under the constraint that the first-order error with respect to the plugged-in quantity is zero, and can leverage flexible model classes including neural networks and tree ensembles. In several experimental settings, including ones in which we handle text-based covariates by fine-tuning language models, our constrained learning-based estimator outperforms one-step estimation and targeting in challenging settings with limited overlap between treatment and control, and performs comparably otherwise."
                    }
                ]
            }
        }
    },
    "2406.08300": {
        "paper_data": {
            "title": "From Chaos to Clarity: 3DGS in the Dark",
            "url": "http://arxiv.org/abs/2406.08300v1",
            "arxiv_id": "2406.08300",
            "authors": [
                "Zhihao Li",
                "Yufei Wang",
                "Alex Kot",
                "Bihan Wen"
            ],
            "abstract": "Novel view synthesis from raw images provides superior high dynamic range (HDR) information compared to reconstructions from low dynamic range RGB images. However, the inherent noise in unprocessed raw images compromises the accuracy of 3D scene representation. Our study reveals that 3D Gaussian Splatting (3DGS) is particularly susceptible to this noise, leading to numerous elongated Gaussian shapes that overfit the noise, thereby significantly degrading reconstruction quality and reducing inference speed, especially in scenarios with limited views. To address these issues, we introduce a novel self-supervised learning framework designed to reconstruct HDR 3DGS from a limited number of noisy raw images. This framework enhances 3DGS by integrating a noise extractor and employing a noise-robust reconstruction loss that leverages a noise distribution prior. Experimental results show that our method outperforms LDR/HDR 3DGS and previous state-of-the-art (SOTA) self-supervised and supervised pre-trained models in both reconstruction quality and inference speed on the RawNeRF dataset across a broad range of training views. Code can be found in \\url{https://lizhihao6.github.io/Raw3DGS}.",
            "introduction": "   1 Introduction  Novel view synthesis (NVS) is fundamental to 3D vision, with extensive applications in virtual and augmented reality (VR/AR)\u00a0[15; 16], autonomous driving\u00a0[17; 33], and 3D asset creation\u00a0[21; 4; 19]. Neural Radiance Fields (NeRF)\u00a0[23] have revolutionized this field by rendering colors through the accumulation of RGB values along sampling rays, employing an implicit MultiLayer Perceptron (MLP) representation. Typically, this method uses low dynamic range (LDR) RGB images processed by image signal processing (ISP) modules, leading to a significant loss of crucial scene details, especially in high-contrast areas like highlights and shadows, which can degrade performance in high dynamic range (HDR) environments such as tunnels, sunsets, or dimly lit scenes. Moreover, reliance on ISP-processed RGB images restricts post-capture color and tone adjustments, presenting significant challenges for photographers and modelers during post-production.   In contrast, raw images before ISP offer a higher dynamic range and preserve more scene information. Recent research has indicated that utilizing raw images can significantly enhance the performance of downstream computer vision tasks in complex lighting conditions\u00a0[18] and offer greater flexibility in post-production adjustments\u00a0[36; 5]. Building on this advantage, RawNeRF\u00a0[22] first employed raw images as the optimization target in NeRF, achieving marked improvements over traditional RGB-based LDR NeRF approaches. However, RawNeRF\u2019s reliance on implicit 3D representation is computationally demanding, requiring up to 48 hours to train a single scene and about one minute to render a single view, which limits its practicality for real-time applications.   Very recently, the advance on 3D Gaussian Splatting (3DGS)\u00a0[13] offers an explicit 3D representation that employs a set of learnable 3D Gaussian points to depict color, shape, and opacity, thereby enabling real-time novel view rendering. However, utilizing raw images directly as the optimization target in 3DGS can introduce significant artifacts and adversely affect rendering speed, especially in scenarios with limited training views\u2014a common challenge in real-world applications, as depicted in Fig.\u00a01. These challenges arise primarily from the inherent noise in raw images, which is more pronounced than in RGB images due to the absence of noise reduction and smoothing processes typically performed by ISPs. As demonstrated in Fig.\u00a02(a), the presence of noise in raw images is an inevitable consequence of physical phenomena such as the photoelectric effect and hardware limitations like dark current leakage. Unlike NeRF, where the Multilayer Perceptron (MLP) acts as a low-frequency filter\u00a0[37] to mitigate high-frequency noise, 3DGS tends to produce numerous thin, elongated Gaussians to represent noise, as seen in Fig.\u00a01(c).   To address these challenges, we propose a novel self-supervised framework that jointly denoises and constructs HDR 3DGS representations within noisy raw images using a noise robust reconstruction loss. Specifically, our method utilizes a noise extractor to predict the underlying noise in raw images, constrained by a predefined noise model. Simultaneously, the 3DGS branch is optimized towards pseudo noise-free raw images. This dual-branch approach allows the noise extractor to accurately predict and separate noise from the raw images. Compared to using the standard RawNeRF loss function with conventional 3DGS, our approach significantly reduces noise artifacts in rendered views and enhances the rendering speed. Our contributions can be summarized as follows:     \u2022  We discover that noisy raw images significantly",
            "references": [
                {
                    "title": "Regulating Intermediate 3D Features for Vision-Centric Autonomous Driving",
                    "abstract": "Multi-camera perception tasks have gained significant attention in the field of autonomous driving. However, existing frameworks based on Lift-Splat-Shoot (LSS) in the multi-camera setting cannot produce suitable dense 3D features due to the projection nature and uncontrollable densification process. To resolve this problem, we propose to regulate intermediate dense 3D features with the help of volume rendering. Specifically, we employ volume rendering to process the dense 3D features to obtain corresponding 2D features (e.g., depth maps, semantic maps), which are supervised by associated labels in the training. This manner regulates the generation of dense 3D features on the feature level, providing appropriate dense and unified features for multiple perception tasks. Therefore, our approach is termed Vampire, stands for ``Volume rendering As Multi-camera Perception Intermediate feature REgulator''. Experimental results on the Occ3D and nuScenes datasets demonstrate that Vampire facilitates fine-grained and appropriate extraction of dense 3D features, and is competitive with existing SOTA methods across diverse downstream perception tasks like 3D occupancy prediction, LiDAR segmentation and 3D objection detection, while utilizing moderate GPU resources. We provide a video demonstration in the supplementary materials and Codes are available at github.com/cskkxjk/Vampire."
                },
                {
                    "title": "Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering",
                    "abstract": "Neural rendering methods have significantly advanced photo-realistic 3D scene rendering in various academic and industrial applications. The recent 3D Gaussian Splatting method has achieved the state-of-the-art rendering quality and speed combining the benefits of both primitive-based representations and volumetric representations. However, it often leads to heavily redundant Gaussians that try to fit every training view, neglecting the underlying scene ge-ometry. Consequently, the resulting model becomes less robust to significant view changes, texture-less area and lighting effects. We introduce Scaffold-GS, which uses an-chor points to distribute local 3D Gaussians, and predicts their attributes on-the-fly based on viewing direction and distance within the view frustum. Anchor growing and pruning strategies are developed based on the importance of neural Gaussians to reliably improve the scene cover-age. We show that our method effectively reduces redun-dant Gaussians while delivering high-quality rendering. We also demonstrates an enhanced capability to accommodate scenes with varying levels-of-detail and view-dependent ob-servations, without sacrificing the rendering speed. Project page: https://city-super.github.iolscaffold-gsl."
                },
                {
                    "title": "3D Gaussian Splatting for Real-Time Radiance Field Rendering",
                    "abstract": "Radiance Field methods have recently revolutionized novel-view synthesis of scenes captured with multiple photos or videos. However, achieving high visual quality still requires neural networks that are costly to train and render, while recent faster methods inevitably trade off speed for quality. For unbounded and complete scenes (rather than isolated objects) and 1080p resolution rendering, no current method can achieve real-time display rates. We introduce three key elements that allow us to achieve state-of-the-art visual quality while maintaining competitive training times and importantly allow high-quality real-time (\u2265 30 fps) novel-view synthesis at 1080p resolution. First, starting from sparse points produced during camera calibration, we represent the scene with 3D Gaussians that preserve desirable properties of continuous volumetric radiance fields for scene optimization while avoiding unnecessary computation in empty space; Second, we perform interleaved optimization/density control of the 3D Gaussians, notably optimizing anisotropic covariance to achieve an accurate representation of the scene; Third, we develop a fast visibility-aware rendering algorithm that supports anisotropic splatting and both accelerates training and allows realtime rendering. We demonstrate state-of-the-art visual quality and real-time rendering on several established datasets."
                },
                {
                    "title": "Instant-3D: Instant Neural Radiance Field Training Towards On-Device AR/VR 3D Reconstruction",
                    "abstract": "Neural Radiance Field (NeRF) based 3D reconstruction is highly desirable for immersive Augmented and Virtual Reality (AR/VR) applications, but achieving instant (i.e., < 5 seconds) on-device NeRF training remains a challenge. In this work, we first identify the inefficiency bottleneck: the need to interpolate NeRF embeddings up to 200,000 times from a 3D embedding grid during each training iteration. To alleviate this, we propose Instant-3D, an algorithm-hardware co-design acceleration framework that achieves instant on-device NeRF training. Our algorithm decomposes the embedding grid representation in terms of color and density, enabling computational redundancy to be squeezed out by adopting different (1) grid sizes and (2) update frequencies for the color and density branches. Our hardware accelerator further reduces the dominant memory accesses for embedding grid interpolation by (1) mapping multiple nearby points' memory read requests into one during the feed-forward process, (2) merging embedding grid updates from the same sliding time window during back-propagation, and (3) fusing different computation cores to support the different grid sizes needed by the color and density branches of Instant-3D algorithm. Extensive experiments validate the effectiveness of Instant-3D, achieving a large training time reduction of 41\u00d7 - 248\u00d7 while maintaining the same reconstruction quality. Excitingly, Instant-3D has enabled instant 3D reconstruction for AR/VR, requiring a reconstruction time of only 1.6 seconds per scene and meeting the AR/VR power consumption constraint of 1.9 W."
                },
                {
                    "title": "Disorder-Invariant Implicit Neural Representation",
                    "abstract": "Implicit neural representation (INR) characterizes the attributes of a signal as a function of corresponding coordinates which emerges as a sharp weapon for solving inverse problems. However, the expressive power of INR is limited by the spectral bias in the network training. In this paper, we find that such a frequency-related problem could be greatly solved by re-arranging the coordinates of the input signal, for which we propose the disorder-invariant implicit neural representation (DINER) by augmenting a hash-table to a traditional INR backbone. Given discrete signals sharing the same histogram of attributes and different arrangement orders, the hash-table could project the coordinates into the same distribution for which the mapped signal can be better modeled using the subsequent INR network, leading to significantly alleviated spectral bias. Furthermore, the expressive power of the DINER is determined by the width of the hash-table. Different width corresponds to different geometrical elements in the attribute space, e.g., 1D curve, 2D curved-plane and 3D curved-volume when the width is set as 1, 2 and 3, respectively. More covered areas of the geometrical elements result in stronger expressive power. Experiments not only reveal the generalization of the DINER for different INR backbones (MLP versus SIREN) and various tasks (image/video representation, phase retrieval, refractive index recovery, and neural radiance field optimization) but also show the superiority over the state-of-the-art algorithms both in quality and speed."
                },
                {
                    "title": "LG-BPN: Local and Global Blind-Patch Network for Self-Supervised Real-World Denoising",
                    "abstract": "Despite the significant results on synthetic noise under simplified assumptions, most self-supervised denoising methods fail under real noise due to the strong spatial noise correlation, including the advanced self-supervised blindspot networks (BSNs). For recent methods targeting real-world denoising, they either suffer from ignoring this spatial correlation, or are limited by the destruction of fine textures for under-considering the correlation. In this paper, we present a novel method called LG-BPN for self-supervised real-world denoising, which takes the spatial correlation statistic into our network design for local detail restoration, and also brings the long-range dependencies modeling ability to previously CNN-based BSN methods. First, based on the correlation statistic, we propose a densely-sampled patch-masked convolution module. By taking more neighbor pixels with low noise correlation into account, we enable a denser local receptive field, preserving more useful information for enhanced fine structure recovery. Second, we propose a dilated Transformer block to allow distant context exploitation in BSN. This global perception addresses the intrinsic deficiency of BSN, whose receptive field is constrained by the blind spot requirement, which can not be fully resolved by the previous CNN-based BSNs. These two designs enable LG-BPN to fully exploit both the detailed structure and the global interaction in a blind manner. Extensive results on real-world datasets demonstrate the superior performance of our method. https://github.com/Wang-XIaoDingdd/LGBPN"
                },
                {
                    "title": "Efficient Visual Computing With Camera RAW Snapshots",
                    "abstract": "Conventional cameras capture image irradiance (RAW) on a sensor and convert it to RGB images using an image signal processor (ISP). The images can then be used for photography or visual computing tasks in a variety of applications, such as public safety surveillance and autonomous driving. One can argue that since RAW images contain all the captured information, the conversion of RAW to RGB using an ISP is not necessary for visual computing. In this paper, we propose a novel <inline-formula><tex-math notation=\"LaTeX\">$\\rho$</tex-math><alternatives><mml:math><mml:mi>\u03c1</mml:mi></mml:math><inline-graphic xlink:href=\"li-ieq1-3359326.gif\"/></alternatives></inline-formula>-Vision framework to perform high-level semantic understanding and low-level compression using RAW images without the ISP subsystem used for decades. Considering the scarcity of available RAW image datasets, we first develop an unpaired CycleR2R network based on unsupervised CycleGAN to train modular unrolled ISP and inverse ISP (invISP) models using unpaired RAW and RGB images. We can then flexibly generate simulated RAW images (simRAW) using any existing RGB image dataset and finetune different models originally trained in the RGB domain to process real-world camera RAW images. We demonstrate object detection and image compression capabilities in RAW-domain using RAW-domain YOLOv3 and RAW image compressor (RIC) on camera snapshots. Quantitative results reveal that RAW-domain task inference provides better detection accuracy and compression efficiency compared to that in the RGB domain. Furthermore, the proposed <inline-formula><tex-math notation=\"LaTeX\">$\\rho$</tex-math><alternatives><mml:math><mml:mi>\u03c1</mml:mi></mml:math><inline-graphic xlink:href=\"li-ieq2-3359326.gif\"/></alternatives></inline-formula>-Vision generalizes across various camera sensors and different task-specific models. An added benefit of employing the <inline-formula><tex-math notation=\"LaTeX\">$\\rho$</tex-math><alternatives><mml:math><mml:mi>\u03c1</mml:mi></mml:math><inline-graphic xlink:href=\"li-ieq3-3359326.gif\"/></alternatives></inline-formula>-Vision is the elimination of the need for ISP, leading to potential reductions in computations and processing times."
                },
                {
                    "title": "Magic3D: High-Resolution Text-to-3D Content Creation",
                    "abstract": "DreamFusion [31] has recently demonstrated the utility of a pretrained text-to-image diffusion model to optimize Neural Radiance Fields (NeRF) [23], achieving remarkable text-to-3D synthesis results. However, the method has two inherent limitations: (a) extremely slow optimization of NeRF and (b) low-resolution image space supervision on NeRF, leading to low-quality 3D models with a long processing time. In this paper, we address these limitations by utilizing a two-stage optimization framework. First, we obtain a coarse model using a low-resolution diffusion prior and accelerate with a sparse 3D hash grid structure. Using the coarse representation as the initialization, we further optimize a textured 3D mesh model with an efficient differentiable renderer interacting with a high-resolution latent diffusion model. Our method, dubbed Magic3D, can create high quality 3D mesh models in 40 minutes, which is 2\u00d7 faster than DreamFusion (reportedly taking 1.5 hours on average), while also achieving higher resolution. User studies show 61.7% raters to prefer our approach over DreamFusion. Together with the image-conditioned generation capabilities, we provide users with new ways to control 3D synthesis, opening up new avenues to various creative applications."
                },
                {
                    "title": "Latent-NeRF for Shape-Guided Generation of 3D Shapes and Textures",
                    "abstract": "Text-guided image generation has progressed rapidly in recent years, inspiring major breakthroughs in text-guided shape generation. Recently, it has been shown that using score distillation, one can successfully text-guide a NeRF model to generate a 3D object. We adapt the score distillation to the publicly available, and computationally efficient, Latent Diffusion Models, which apply the entire diffusion process in a compact latent space of a pretrained autoencoder. As NeRFs operate in image space, a naive solution for guiding them with latent score distillation would require encoding to the latent space at each guidance step. Instead, we propose to bring the NeRF to the latent space, resulting in a Latent-NeRF. Analyzing our Latent-NeRF, we show that while Text-to-3D models can generate impressive results, they are inherently unconstrained and may lack the ability to guide or enforce a specific 3D structure. To assist and direct the 3D generation, we propose to guide our Latent-NeRF using a Sketch-Shape: an abstract geometry that defines the coarse structure of the desired object. Then, we present means to integrate such a constraint directly into a Latent-NeRF. This unique combination of text and shape guidance allows for increased control over the generation process. We also show that latent score distillation can be successfully applied directly on 3D meshes. This allows for generating high-quality textures on a given geometry. Our experiments validate the power of our different forms of guidance and the efficiency of using latent rendering."
                },
                {
                    "title": "RT-NeRF: Real-Time On-Device Neural Radiance Fields Towards Immersive AR/VR Rendering",
                    "abstract": "Neural Radiance Field (NeRF) based rendering has attracted growing attention thanks to its state-of-the-art (SOTA) rendering quality and wide applications in Augmented and Virtual Reality (AR/VR). However, immersive real-time (> 30 FPS) NeRF based rendering enabled interactions are still limited due to the low achievable throughput on AR/VR devices. To this end, we first profile SOTA efficient NeRF algorithms on commercial devices and identify two primary causes of the aforementioned inefficiency: (1) the uniform point sampling and (2) the dense accesses and computations of the required embeddings in NeRF. Furthermore, we propose RT-NeRF, which to the best of our knowledge is the first algorithm-hardware co-design acceleration of NeRF. Specifically, on the algorithm level, RT-NeRF integrates an efficient rendering pipeline for largely alleviating the inefficiency due to the commonly adopted uniform point sampling method in NeRF by directly computing the geometry of pre-existing points. Additionally, RT-NeRF leverages a coarse-grained view-dependent computing ordering scheme for eliminating the (unnecessary) processing of invisible points. On the hardware level, our proposed RT-NeRF accelerator (1) adopts a hybrid encoding scheme to adaptively switch between a bitmap- or coordinate-based sparsity encoding format for NeRF\u2019s sparse embeddings, aiming to maximize the storage savings and thus reduce the required DRAM accesses while supporting efficient NeRF decoding; and (2) integrates both a high-density sparse search unit and a dual-purpose bi-direction adder & search tree to coordinate the two aforementioned encoding formats. Extensive experiments on eight datasets consistently validate the effectiveness of RT-NeRF, achieving a large throughput improvement (e.g., 9.7\u00d7\u223c3,201\u00d7) while maintaining the rendering quality as compared with SOTA efficient NeRF solutions."
                },
                {
                    "title": "Learnability Enhancement for Low-light Raw Denoising: Where Paired Real Data Meets Noise Modeling",
                    "abstract": "Low-light raw denoising is an important and valuable task in computational photography where learning-based methods trained with paired real data are mainstream. However, the limited data volume and complicated noise distribution have constituted a learnability bottleneck for paired real data, which limits the denoising performance of learning-based methods. To address this issue, we present a learnability enhancement strategy to reform paired real data according to noise modeling. Our strategy consists of two efficient techniques: shot noise augmentation (SNA) and dark shading correction (DSC). Through noise model decoupling, SNA improves the precision of data mapping by increasing the data volume and DSC reduces the complexity of data mapping by reducing the noise complexity. Extensive results on the public datasets and real imaging scenarios collectively demonstrate the state-of-the-art performance of our method."
                },
                {
                    "title": "Scalable neural indoor scene rendering",
                    "abstract": "We propose a scalable neural scene reconstruction and rendering method to support distributed training and interactive rendering of large indoor scenes. Our representation is based on tiles. Tile appearances are trained in parallel through a background sampling strategy that augments each tile with distant scene information via a proxy global mesh. Each tile has two low-capacity MLPs: one for view-independent appearance (diffuse color and shading) and one for view-dependent appearance (specular highlights, reflections). We leverage the phenomena that complex view-dependent scene reflections can be attributed to virtual lights underneath surfaces at the total ray distance to the source. This lets us handle sparse samplings of the input scene where reflection highlights do not always appear consistently in input images. We show interactive free-viewpoint rendering results from five scenes, one of which covers an area of more than 100 m2. Experimental results show that our method produces higher-quality renderings than a single large-capacity MLP and five recent neural proxy-geometry and voxel-based baseline methods. Our code and data are available at project webpage https://xchaowu.github.io/papers/scalable-nisr."
                },
                {
                    "title": "Variable Bitrate Neural Fields",
                    "abstract": "Neural approximations of scalar- and vector fields, such as signed distance functions and radiance fields, have emerged as accurate, high-quality representations. State-of-the-art results are obtained by conditioning a neural approximation with a lookup from trainable feature grids [Liu et al. 2020; Martel et al. 2021; M\u00fcller et al. 2022; Takikawa et al. 2021] that take on part of the learning task and allow for smaller, more efficient neural networks. Unfortunately, these feature grids usually come at the cost of significantly increased memory consumption compared to stand-alone neural network models. We present a dictionary method for compressing such feature grids, reducing their memory consumption by up to 100 \u00d7 and permitting a multiresolution representation which can be useful for out-of-core streaming. We formulate the dictionary optimization as a vector-quantized auto-decoder problem which lets us learn end-to-end discrete neural representations in a space where no direct supervision is available and with dynamic topology and structure. Our source code is available at https://github.com/nv-tlabs/vqad."
                },
                {
                    "title": "READ: Large-Scale Neural Scene Rendering for Autonomous Driving",
                    "abstract": "With the development of advanced driver assistance systems~(ADAS) and autonomous vehicles, conducting experiments in various scenarios becomes an urgent need. Although having been capable of synthesizing photo-realistic street scenes, conventional image-to-image translation methods cannot produce coherent scenes due to the lack of 3D information. In this paper, a large-scale neural rendering method is proposed to synthesize the autonomous driving scene~(READ), which makes it possible to generate large-scale driving scenes in real time on a PC through a variety of sampling schemes. In order to effectively represent driving scenarios, we propose an \u03c9-net rendering network to learn neural descriptors from sparse point clouds. Our model can not only synthesize photo-realistic driving scenes but also stitch and edit them. The promising experimental results show that our model performs well in large-scale driving scenarios."
                },
                {
                    "title": "Instant neural graphics primitives with a multiresolution hash encoding",
                    "abstract": "Neural graphics primitives, parameterized by fully connected neural networks, can be costly to train and evaluate. We reduce this cost with a versatile new input encoding that permits the use of a smaller network without sacrificing quality, thus significantly reducing the number of floating point and memory access operations: a small neural network is augmented by a multiresolution hash table of trainable feature vectors whose values are optimized through stochastic gradient descent. The multiresolution structure allows the network to disambiguate hash collisions, making for a simple architecture that is trivial to parallelize on modern GPUs. We leverage this parallelism by implementing the whole system using fully-fused CUDA kernels with a focus on minimizing wasted bandwidth and compute operations. We achieve a combined speedup of several orders of magnitude, enabling training of high-quality neural graphics primitives in a matter of seconds, and rendering in tens of milliseconds at a resolution of 1920\u00d71080."
                },
                {
                    "title": "GRAM: Generative Radiance Manifolds for 3D-Aware Image Generation",
                    "abstract": "3D-aware image generative modeling aims to generate 3D-consistent images with explicitly controllable camera poses. Recent works have shown promising results by training neural radiance field (NeRF) generators on unstructured 2D images, but still cannot generate highly-realistic images with fine details. A critical reason is that the high memory and computation cost of volumetric representation learning greatly restricts the number of point samples for radiance integration during training. Deficient sampling not only limits the expressive power of the generator to handle fine details but also impedes effective GAN training due to the noise caused by unstable Monte Carlo sampling. We propose a novel approach that regulates point sampling and radiance field learning on 2D manifolds, embodied as a set of learned implicit surfaces in the 3D volume. For each viewing ray, we calculate ray-surface intersections and accumulate their radiance generated by the network. By training and rendering such radiance mani folds, our generator can produce high quality images with realistic fine details and strong visual 3D consistency. 11Project page: https://yudeng.github.io/GRAM/"
                },
                {
                    "title": "Plenoxels: Radiance Fields without Neural Networks",
                    "abstract": "We introduce Plenoxels (plenoptic voxels), a systemfor photorealistic view synthesis. Plenoxels represent a scene as a sparse 3D grid with spherical harmonics. This representation can be optimized from calibrated images via gradient methods and regularization without any neural components. On standard, benchmark tasks, Plenoxels are optimized two orders of magnitude faster than Neural Radiance Fields with no loss in visual quality. For video and code, please see https://alexyu.net/plenoxels."
                },
                {
                    "title": "NeRF in the Dark: High Dynamic Range View Synthesis from Noisy Raw Images",
                    "abstract": "Neural Radiance Fields (NeRF) is a technique for high quality novel view synthesis from a collection of posed input images. Like most view synthesis methods, NeRF uses tonemapped low dynamic range (LDR) as input; these images have been processed by a lossy camera pipeline that smooths detail, clips highlights, and distorts the simple noise distribution of raw sensor data. We modify NeRF to instead train directly on linear raw images, preserving the scene's full dynamic range. By rendering raw output images from the resulting NeRF, we can perform novel high dynamic range (HDR) view synthesis tasks. In addition to changing the camera viewpoint, we can manipulate focus, exposure, and tonemapping after the fact. Although a single raw image appears significantly more noisy than a postprocessed one, we show that NeRF is highly robust to the zeromean distribution of raw noise. When optimized over many noisy raw inputs (25\u2013200), NeRF produces a scene representation so accurate that its rendered novel views outperform dedicated single and multi-image deep raw denoisers run on the same wide baseline input images. As a result, our method, which we call RawNeRF, can reconstruct scenes from extremely noisy images captured in near-darkness."
                },
                {
                    "title": "Baking Neural Radiance Fields for Real-Time View Synthesis",
                    "abstract": "Neural volumetric representations such as Neural Radiance Fields (NeRF) have emerged as a compelling technique for learning to represent 3D scenes from images with the goal of rendering photorealistic images of the scene from unobserved viewpoints. However, NeRF\u2019s computational requirements are prohibitive for real-time applications: rendering views from a trained NeRF requires querying a multilayer perceptron (MLP) hundreds of times per ray. We present a method to train a NeRF, then precompute and store (i.e. \"bake\") it as a novel representation called a Sparse Neural Radiance Grid (SNeRG) that enables real-time rendering on commodity hardware. To achieve this, we introduce 1) a reformulation of NeRF\u2019s architecture, and 2) a sparse voxel grid representation with learned feature vectors. The resulting scene representation retains NeRF\u2019s ability to render fine geometric details and view-dependent appearance, is compact (averaging less than 90 MB per scene), and can be rendered in real-time (higher than 30 frames per second on a laptop GPU). Actual screen captures are shown in our video."
                },
                {
                    "title": "KiloNeRF: Speeding up Neural Radiance Fields with Thousands of Tiny MLPs",
                    "abstract": "NeRF synthesizes novel views of a scene with unprecedented quality by fitting a neural radiance field to RGB images. However, NeRF requires querying a deep Multi-Layer Perceptron (MLP) millions of times, leading to slow rendering times, even on modern GPUs. In this paper, we demonstrate that real-time rendering is possible by utilizing thousands of tiny MLPs instead of one single large MLP. In our setting, each individual MLP only needs to represent parts of the scene, thus smaller and faster-to-evaluate MLPs can be used. By combining this divide-and-conquer strategy with further optimizations, rendering is accelerated by three orders of magnitude compared to the original NeRF model without incurring high storage costs. Further, using teacher-student distillation for training, we show that this speed-up can be achieved without sacrificing visual quality."
                },
                {
                    "title": "PlenOctrees for Real-time Rendering of Neural Radiance Fields",
                    "abstract": "We introduce a method to render Neural Radiance Fields (NeRFs) in real time using PlenOctrees, an octree-based 3D representation which supports view-dependent effects. Our method can render 800\u00d7800 images at more than 150 FPS, which is over 3000 times faster than conventional NeRFs. We do so without sacrificing quality while preserving the ability of NeRFs to perform free-viewpoint rendering of scenes with arbitrary geometry and view-dependent effects. Real-time performance is achieved by pre-tabulating the NeRF into a PlenOctree. In order to preserve view-dependent effects such as specularities, we factorize the appearance via closed-form spherical basis functions. Specifically, we show that it is possible to train NeRFs to predict a spherical harmonic representation of radiance, removing the viewing direction as an input to the neural network. Furthermore, we show that PlenOctrees can be directly optimized to further minimize the reconstruction loss, which leads to equal or better quality compared to competing methods. Moreover, this octree optimization step can be used to reduce the training time, as we no longer need to wait for the NeRF training to converge fully. Our real-time neural rendering approach may potentially enable new applications such as 6-DOF industrial and product visualizations, as well as next generation AR/VR systems. PlenOctrees are amenable to in-browser rendering as well; please visit the project page for the interactive online demo, as well as video and code: https://alexyu.net/plenoctrees."
                },
                {
                    "title": "FastNeRF: High-Fidelity Neural Rendering at 200FPS",
                    "abstract": "Recent work on Neural Radiance Fields (NeRF) showed how neural networks can be used to encode complex 3D environments that can be rendered photorealistically from novel viewpoints. Rendering these images is very computationally demanding and recent improvements are still a long way from enabling interactive rates, even on high-end hardware. Motivated by scenarios on mobile and mixed reality devices, we propose FastNeRF, the first NeRF-based system capable of rendering high fidelity photorealistic images at 200Hz on a high-end consumer GPU. The core of our method is a graphics-inspired factorization that allows for (i) compactly caching a deep radiance map at each position in space, (ii) efficiently querying that map using ray directions to estimate the pixel values in the rendered image. Extensive experiments show that the proposed method is 3000 times faster than the original NeRF algorithm and at least an order of magnitude faster than existing work on accelerating NeRF, while maintaining visual quality and extensibility."
                },
                {
                    "title": "Neural Geometric Level of Detail: Real-time Rendering with Implicit 3D Shapes",
                    "abstract": "Neural signed distance functions (SDFs) are emerging as an effective representation for 3D shapes. State-of-the-art methods typically encode the SDF with a large, fixed-size neural network to approximate complex shapes with implicit surfaces. Rendering with these large networks is, however, computationally expensive since it requires many forward passes through the network for every pixel, making these representations impractical for real-time graphics. We introduce an efficient neural representation that, for the first time, enables real-time rendering of high-fidelity neural SDFs, while achieving state-of-the-art geometry reconstruction quality. We represent implicit surfaces using an octree-based feature volume which adaptively fits shapes with multiple discrete levels of detail (LODs), and enables continuous LOD with SDF interpolation. We further develop an efficient algorithm to directly render our novel neural SDF representation in real-time by querying only the necessary LODs with sparse octree traversal. We show that our representation is 2\u20133 orders of magnitude more efficient in terms of rendering speed compared to previous works. Furthermore, it produces state-of-the-art reconstruction quality for complex shapes under both 3D geometric and 2D image-space metrics."
                },
                {
                    "title": "Neighbor2Neighbor: Self-Supervised Denoising from Single Noisy Images",
                    "abstract": "In the last few years, image denoising has benefited a lot from the fast development of neural networks. However, the requirement of large amounts of noisy-clean image pairs for supervision limits the wide use of these models. Although there have been a few attempts in training an image denoising model with only single noisy images, existing self-supervised denoising approaches suffer from inefficient network training, loss of useful information, or dependence on noise modeling. In this paper, we present a very simple yet effective method named Neighbor2Neighbor to train an effective image denoising model with only noisy images. Firstly, a random neighbor sub-sampler is proposed for the generation of training image pairs. In detail, input and target used to train a network are images sub-sampled from the same noisy image, satisfying the requirement that paired pixels of paired images are neighbors and have very similar appearance with each other. Secondly, a denoising network is trained on sub-sampled training pairs generated in the first stage, with a proposed regularizer as additional loss for better performance. The proposed Neighbor2Neighbor framework is able to enjoy the progress of state-of-the-art supervised denoising networks in network architecture design. Moreover, it avoids heavy dependence on the assumption of the noise distribution. We explain our approach from a theoretical perspective and further validate it through extensive experiments, including synthetic experiments with different noise distributions in sRGB space and real-world experiments on a denoising benchmark dataset in raw-RGB space."
                },
                {
                    "title": "Supervised Raw Video Denoising With a Benchmark Dataset on Dynamic Scenes",
                    "abstract": "In recent years, the supervised learning strategy for real noisy image denoising has been emerging and has achieved promising results. In contrast, realistic noise removal for raw noisy videos is rarely studied due to the lack of noisy-clean pairs for dynamic scenes. Clean video frames for dynamic scenes cannot be captured with a long-exposure shutter or averaging multi-shots as was done for static images. In this paper, we solve this problem by creating motions for controllable objects, such as toys, and capturing each static moment for multiple times to generate clean video frames. In this way, we construct a dataset with 55 groups of noisy-clean videos with ISO values ranging from 1600 to 25600. To our knowledge, this is the first dynamic video dataset with noisy-clean pairs. Correspondingly, we propose a raw video denoising network (RViDeNet) by exploring the temporal, spatial, and channel correlations of video frames. Since the raw video has Bayer patterns, we pack it into four sub-sequences, i.e RGBG sequences, which are denoised by the proposed RViDeNet separately and finally fused into a clean video. In addition, our network not only outputs a raw denoising result, but also the sRGB result by going through an image signal processing (ISP) module, which enables users to generate the sRGB result with their favourite ISPs. Experimental results demonstrate that our method outperforms state-of-the-art video and raw image denoising algorithms on both indoor and outdoor videos."
                },
                {
                    "title": "A Physics-Based Noise Formation Model for Extreme Low-Light Raw Denoising",
                    "abstract": "Lacking rich and realistic data, learned single image denoising algorithms generalize poorly in real raw images that not resemble the data used for training. Although the problem can be alleviated by the heteroscedastic Gaussian noise model, the noise sources caused by digital camera electronics are still largely overlooked, despite their significant effect on raw measurement, especially under extremely low-light condition. To address this issue, we present a highly accurate noise formation model based on the characteristics of CMOS photosensors, thereby enabling us to synthesize realistic samples that better match the physics of image formation process. Given the proposed noise model, we additionally propose a method to calibrate the noise parameters for available modern digital cameras, which is simple and reproducible for any new device. We systematically study the generalizability of a neural network trained with existing schemes, by introducing a new low-light denoising dataset that covers many modern digital cameras from diverse brands. Extensive empirical results collectively show that by utilizing our proposed noise formation model, a network can reach the capability as if it had been trained with rich real data, which demonstrates the effectiveness of our noise formation model."
                },
                {
                    "title": "Synthetic defocus and look-ahead autofocus for casual videography",
                    "abstract": "In cinema, large camera lenses create beautiful shallow depth of field (DOF), but make focusing difficult and expensive. Accurate cinema focus usually relies on a script and a person to control focus in realtime. Casual videographers often crave cinematic focus, but fail to achieve it. We either sacrifice shallow DOF, as in smartphone videos; or we struggle to deliver accurate focus, as in videos from larger cameras. This paper is about a new approach in the pursuit of cinematic focus for casual videography. We present a system that synthetically renders refocusable video from a deep DOF video shot with a smartphone, and analyzes future video frames to deliver context-aware autofocus for the current frame. To create refocusable video, we extend recent machine learning methods designed for still photography, contributing a new dataset for machine training, a rendering model better suited to cinema focus, and a filtering solution for temporal coherence. To choose focus accurately for each frame, we demonstrate autofocus that looks at upcoming video frames and applies AI-assist modules such as motion, face, audio and saliency detection. We also show that autofocus benefits from machine learning and a large-scale video dataset with focus annotation, where we use our RVR-LAAF GUI to create this sizable dataset efficiently. We deliver, for example, a shallow DOF video where the autofocus transitions onto each person before she begins to speak. This is impossible for conventional camera autofocus because it would require seeing into the future."
                },
                {
                    "title": "A High-Quality Denoising Dataset for Smartphone Cameras",
                    "abstract": "The last decade has seen an astronomical shift from imaging with DSLR and point-and-shoot cameras to imaging with smartphone cameras. Due to the small aperture and sensor size, smartphone images have notably more noise than their DSLR counterparts. While denoising for smartphone images is an active research area, the research community currently lacks a denoising image dataset representative of real noisy images from smartphone cameras with high-quality ground truth. We address this issue in this paper with the following contributions. We propose a systematic procedure for estimating ground truth for noisy images that can be used to benchmark denoising performance for smartphone cameras. Using this procedure, we have captured a dataset - the Smartphone Image Denoising Dataset (SIDD) - of ~30,000 noisy images from 10 scenes under different lighting conditions using five representative smartphone cameras and generated their ground truth images. We used this dataset to benchmark a number of denoising algorithms. We show that CNN-based methods perform better when trained on our high-quality dataset than when trained using alternative strategies, such as low-ISO images used as a proxy for ground truth data."
                },
                {
                    "title": "Learning to See in the Dark",
                    "abstract": "Imaging in low light is challenging due to low photon count and low SNR. Short-exposure images suffer from noise, while long exposure can induce blur and is often impractical. A variety of denoising, deblurring, and enhancement techniques have been proposed, but their effectiveness is limited in extreme conditions, such as video-rate imaging at night. To support the development of learning-based pipelines for low-light image processing, we introduce a dataset of raw short-exposure low-light images, with corresponding long-exposure reference images. Using the presented dataset, we develop a pipeline for processing low-light images, based on end-to-end training of a fully-convolutional network. The network operates directly on raw sensor data and replaces much of the traditional image processing pipeline, which tends to perform poorly on such data. We report promising results on the new dataset, analyze factors that affect performance, and highlight opportunities for future work."
                },
                {
                    "title": "Burst photography for high dynamic range and low-light imaging on mobile cameras",
                    "abstract": "Cell phone cameras have small apertures, which limits the number of photons they can gather, leading to noisy images in low light. They also have small sensor pixels, which limits the number of electrons each pixel can store, leading to limited dynamic range. We describe a computational photography pipeline that captures, aligns, and merges a burst of frames to reduce noise and increase dynamic range. Our system has several key features that help make it robust and efficient. First, we do not use bracketed exposures. Instead, we capture frames of constant exposure, which makes alignment more robust, and we set this exposure low enough to avoid blowing out highlights. The resulting merged image has clean shadows and high bit depth, allowing us to apply standard HDR tone mapping methods. Second, we begin from Bayer raw frames rather than the demosaicked RGB (or YUV) frames produced by hardware Image Signal Processors (ISPs) common on mobile platforms. This gives us more bits per pixel and allows us to circumvent the ISP's unwanted tone mapping and spatial denoising. Third, we use a novel FFT-based alignment algorithm and a hybrid 2D/3D Wiener filter to denoise and merge the frames in a burst. Our implementation is built atop Android's Camera2 API, which provides per-frame camera control and access to raw imagery, and is written in the Halide domain-specific language (DSL). It runs in 4 seconds on device (for a 12 Mpix image), requires no user intervention, and ships on several mass-produced cell phones."
                },
                {
                    "title": "Structure-from-Motion Revisited",
                    "abstract": "Incremental Structure-from-Motion is a prevalent strategy for 3D reconstruction from unordered image collections. While incremental reconstruction systems have tremendously advanced in all regards, robustness, accuracy, completeness, and scalability remain the key problems towards building a truly general-purpose pipeline. We propose a new SfM technique that improves upon the state of the art to make a further step towards this ultimate goal. The full reconstruction pipeline is released to the public as an open-source implementation."
                },
                {
                    "title": "Real-time noise-aware tone mapping",
                    "abstract": "Real-time high quality video tone mapping is needed for many applications, such as digital viewfinders in cameras, display algorithms which adapt to ambient light, in-camera processing, rendering engines for video games and video post-processing. We propose a viable solution for these applications by designing a video tone-mapping operator that controls the visibility of the noise, adapts to display and viewing environment, minimizes contrast distortions, preserves or enhances image details, and can be run in real-time on an incoming sequence without any preprocessing. To our knowledge, no existing solution offers all these features. Our novel contributions are: a fast procedure for computing local display-adaptive tone-curves which minimize contrast distortions, a fast method for detail enhancement free from ringing artifacts, and an integrated video tone-mapping solution combining all the above features."
                },
                {
                    "title": "Image Denoising by Sparse 3-D Transform-Domain Collaborative Filtering",
                    "abstract": "We propose a novel image denoising strategy based on an enhanced sparse representation in transform domain. The enhancement of the sparsity is achieved by grouping similar 2D image fragments (e.g., blocks) into 3D data arrays which we call \"groups.\" Collaborative Altering is a special procedure developed to deal with these 3D groups. We realize it using the three successive steps: 3D transformation of a group, shrinkage of the transform spectrum, and inverse 3D transformation. The result is a 3D estimate that consists of the jointly filtered grouped image blocks. By attenuating the noise, the collaborative filtering reveals even the finest details shared by grouped blocks and, at the same time, it preserves the essential unique features of each individual block. The filtered blocks are then returned to their original positions. Because these blocks are overlapping, for each pixel, we obtain many different estimates which need to be combined. Aggregation is a particular averaging procedure which is exploited to take advantage of this redundancy. A significant improvement is obtained by a specially developed collaborative Wiener filtering. An algorithm based on this novel denoising strategy and its efficient implementation are presented in full detail; an extension to color-image denoising is also developed. The experimental results demonstrate that this computationally scalable algorithm achieves state-of-the-art denoising performance in terms of both peak signal-to-noise ratio and subjective visual quality."
                }
            ],
            "categories": [
                "eess.IV",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively utilize noisy raw images in 3D Gaussian Splatting (3DGS) for novel view synthesis while mitigating the adverse effects of noise on rendering quality and speed?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of 3D vision, particularly in applications like virtual and augmented reality, autonomous driving, and 3D asset creation. By improving the use of raw images in 3DGS, we can enhance rendering quality in high dynamic range environments, leading to more realistic and immersive experiences. This research could pave the way for real-time applications, allowing for greater flexibility in post-production adjustments and improving the overall efficiency of 3D rendering processes. Addressing this question could significantly advance knowledge in the intersection of image processing and 3D representation, potentially influencing future research directions and practical applications in various industries.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent noise present in raw images, which is more pronounced than in processed RGB images. Naive approaches may fail because they do not adequately account for this noise, leading to artifacts in the rendered output. The complexities include developing a robust noise extraction method that can accurately predict and separate noise from the raw images while maintaining the integrity of the underlying scene information. Additionally, optimizing the 3DGS representation in the presence of noise without compromising rendering speed poses a significant technical obstacle.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on using low dynamic range RGB images, which inherently limits the detail captured in high-contrast scenes. While RawNeRF has made strides in utilizing raw images, its reliance on implicit representations results in high computational demands, making it impractical for real-time applications. The limitations of existing methods in handling noise effectively have prevented the successful application of raw images in 3DGS. Our approach differs by introducing a dual-branch framework that jointly denoises and constructs HDR representations, addressing the noise issue directly and improving upon prior work by enhancing rendering speed and quality.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a self-supervised framework that includes a noise extractor to predict the underlying noise in raw images, guided by a predefined noise model. Simultaneously, the 3DGS branch is optimized towards pseudo"
            }
        },
        "author_data": {
            "292c968a-9b3b-49d5-8cee-ade78155cb21": {
                "pk": "292c968a-9b3b-49d5-8cee-ade78155cb21",
                "name": "Zhihao Li",
                "collaborators": [
                    "Zhihao Liang",
                    "Songcen Xu",
                    "Mingkui Tan",
                    "Kui Jia",
                    "Si Yi",
                    "Zhan Ma",
                    "Zhenglong Sun",
                    "Jionglong SU",
                    "Jiaming Zhang",
                    "Mingzhe Li"
                ],
                "domain": [
                    "Computer Vision",
                    "Robotics",
                    "Deep Learning",
                    "Image Processing"
                ],
                "publications": [
                    {
                        "title": "Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks",
                        "abstract": "Instance segmentation in 3D scenes is fundamental in many applications of scene understanding. It is yet challenging due to the compound factors of data irregularity and uncertainty in the numbers of instances. State-of-the-art methods largely rely on a general pipeline that first learns point-wise features discriminative at semantic and instance levels, followed by a separate step of point grouping for proposing object instances. While promising, they have the shortcomings that (1) the second step is not supervised by the main objective of instance segmentation, and (2) their point-wise feature learning and grouping are less effective to deal with data irregularities, possibly resulting in fragmented segmentations. To address these issues, we propose in this work an end-to-end solution of Semantic Superpoint Tree Network (SSTNet) for proposing object instances from scene points. Key in SSTNet is an intermediate, semantic superpoint tree (SST), which is constructed based on the learned semantic features of superpoints, and which will be traversed and split at intermediate tree nodes for proposals of object instances. We also design in SSTNet a refinement module, termed CliqueNet, to prune superpoints that may be wrongly grouped into instance proposals. Experiments on the benchmarks of ScanNet and S3DIS show the efficacy of our proposed method. At the time of submission, SSTNet ranks top on the ScanNet (V2) leaderboard, with 2% higher of mAP than the second best method. The source code in PyTorch is available at https://github.com/Gorilla-Lab-SCUT/SSTNet."
                    },
                    {
                        "title": "Rendering Nighttime Image Via Cascaded Color and Brightness Compensation",
                        "abstract": "Image signal processing (ISP) is crucial for camera imaging, and neural networks (NN) solutions are extensively deployed for daytime scenes. The lack of sufficient nighttime image dataset and insights on nighttime illumination characteristics poses a great challenge for high-quality rendering using existing NN ISPs. To tackle it, we first built a high-resolution nighttime RAW-RGB (NR2R) dataset with white balance and tone mapping annotated by expert professionals. Meanwhile, to best capture the characteristics of nighttime illumination light sources, we develop the CBUnet, a two-stage NN ISP to cascade the compensation of color and brightness attributes. Experiments show that our method has better visual quality compared to traditional ISP pipeline, and is ranked at the second place in the NTIRE 2022 Night Photography Rendering Challenge for two tracks by respective People's and Professional Photographer's choices. The code and relevant materials are avaiable on our website: https://njuvision.github.io/CBUnet."
                    },
                    {
                        "title": "Learning a Skill-sequence-dependent Policy for Long-horizon Manipulation Tasks",
                        "abstract": "In recent years, the robotics community has made substantial progress in robotic manipulation using deep reinforcement learning (RL). Effectively learning of long-horizon tasks remains a challenging topic. Typical RL-based methods approximate long-horizon tasks as Markov decision processes and only consider current observation (images or other sensor information) as input state. However, such approximation ignores the fact that skill-sequence also plays a crucial role in long-horizon tasks. In this paper, we take both the observation and skill sequences into account and propose a skill-sequence-dependent hierarchical policy for solving a typical long-horizon task. The proposed policy consists of a high-level skill policy (utilizing skill sequences) and a low-level parameter policy (responding to observation) with corresponding training methods, which makes the learning much more sample-efficient. Experiments in simulation demonstrate that our approach successfully solves a long-horizon task and is significantly faster than Proximal Policy Optimization (PPO) and the task schema methods."
                    },
                    {
                        "title": "Ghost instability in the teleparallel gravity model with parity violations",
                        "abstract": "In this paper we consider the parity violating gravity model within the framework of teleparallel gravity. The parity violations are caused by the couplings of a scalar field to the scalar invariants which are parity-odd and quadratic in the torsion tensor. Totally there are two such type independent invariants, and one of them is the Nieh-Yan density. Through investigations on the cosmological perturbations of this model, we find that in general it suffers from the difficulties of ghost instability in the scalar and vector perturbations. But in the special case only the coupling to the Nieh-Yan density exists, this model is ghost free and reduces to the Nieh-Yan modified Teleparallel Gravity model. We also analyze the severity of the ghost instability by studying the perturbations around the Minkowski background."
                    },
                    {
                        "title": "Simulation Method of Microscale Fluid-Structure Interactions: Diffuse-Resistance-Domain Approach",
                        "abstract": "Direct numerical simulations (DNS) of microscale fluid-structure interactions (mFSI) in multicomponent multiphase flows pose many challenges such as the thermodynamic consistency of multiphysics couplings, the tracking of moving interfaces, and the dynamics of moving triple-phase contact lines. We propose and validate a generic DNS approach: Diffuse-Resistance-Domain (DRD). It overcomes the above challenges by employing Onsager's principle to formulate dynamic models and combining traditional diffuse-interface models for fluid-fluid interfacial dynamics with a novel implementation of complex fluid-solid interfacial conditions via smooth interpolations of dynamic-resistance coefficients across interfaces. After careful validation by benchmark simulations, we simulated three challenging mFSI problems taken from different fields."
                    },
                    {
                        "title": "An LSTM-Based Autonomous Driving Model Using Waymo Open Dataset",
                        "abstract": "The Waymo Open Dataset has been released recently, providing a platform to crowdsource some fundamental challenges for automated vehicles (AVs), such as 3D detection and tracking. While~the dataset provides a large amount of high-quality and multi-source driving information, people in academia are more interested in the underlying driving policy programmed in Waymo self-driving cars, which is inaccessible due to AV manufacturers' proprietary protection. Accordingly, academic researchers have to make various assumptions to implement AV components in their models or simulations, which may not represent the realistic interactions in real-world traffic. Thus, this paper introduces an approach to learn a long short-term memory (LSTM)-based model for imitating the behavior of Waymo's self-driving model. The proposed model has been evaluated based on Mean Absolute Error (MAE). The experimental results show that our model outperforms several baseline models in driving action prediction. In addition, a visualization tool is presented for verifying the performance of the model."
                    }
                ]
            },
            "2e6e07f9-cc92-4ae8-bd5f-1794692442e4": {
                "pk": "2e6e07f9-cc92-4ae8-bd5f-1794692442e4",
                "name": "Yufei Wang",
                "collaborators": [
                    "David Held",
                    "Zackory Erickson",
                    "Tianwei Ni",
                    "Mengyue Wu",
                    "Qi Zhao",
                    "Shuchang Lyu",
                    "Lijiang Chen",
                    "Weikang Wan",
                    "Wei Lu",
                    "Xia Wu"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Multi-modal Learning",
                    "Robotics",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Meta-SAC: Auto-tune the Entropy Temperature of Soft Actor-Critic via Metagradient",
                        "abstract": "Exploration-exploitation dilemma has long been a crucial issue in reinforcement learning. In this paper, we propose a new approach to automatically balance between these two. Our method is built upon the Soft Actor-Critic (SAC) algorithm, which uses an \"entropy temperature\" that balances the original task reward and the policy entropy, and hence controls the trade-off between exploitation and exploration. It is empirically shown that SAC is very sensitive to this hyperparameter, and the follow-up work (SAC-v2), which uses constrained optimization for automatic adjustment, has some limitations. The core of our method, namely Meta-SAC, is to use metagradient along with a novel meta objective to automatically tune the entropy temperature in SAC. We show that Meta-SAC achieves promising performances on several of the Mujoco benchmarking tasks, and outperforms SAC-v2 over 10% in one of the most challenging tasks, humanoid-v2."
                    },
                    {
                        "title": "Evaluation of data inconsistency for multi-modal sentiment analysis",
                        "abstract": "Emotion semantic inconsistency is an ubiquitous challenge in multi-modal sentiment analysis (MSA). MSA involves analyzing sentiment expressed across various modalities like text, audio, and videos. Each modality may convey distinct aspects of sentiment, due to subtle and nuanced expression of human beings, leading to inconsistency, which may hinder the prediction of artificial agents. In this work, we introduce a modality conflicting test set and assess the performance of both traditional multi-modal sentiment analysis models and multi-modal large language models (MLLMs). Our findings reveal significant performance degradation across traditional models when confronted with semantically conflicting data and point out the drawbacks of MLLMs when handling multi-modal emotion analysis. Our research presents a new challenge and offer valuable insights for the future development of sentiment analysis systems."
                    },
                    {
                        "title": "Attention-based Feature Decomposition-Reconstruction Network for Scene Text Detection",
                        "abstract": "Recently, scene text detection has been a challenging task. Texts with arbitrary shape or large aspect ratio are usually hard to detect. Previous segmentation-based methods can describe curve text more accurately but suffer from over segmentation and text adhesion. In this paper, we propose attention-based feature decomposition-reconstruction network for scene text detection, which utilizes contextual information and low-level feature to enhance the performance of segmentation-based text detector. In the phase of feature fusion, we introduce cross level attention module to enrich contextual information of text by adding attention mechanism on fused multi-scaled feature. In the phase of probability map generation, a feature decomposition-reconstruction module is proposed to alleviate the over segmentation problem of large aspect ratio text, which decomposes text feature according to their frequency characteristic and then reconstructs it by adding low-level feature. Experiments have been conducted on two public benchmark datasets and results show that our proposed method achieves state-of-the-art performance."
                    },
                    {
                        "title": "Visual Haptic Reasoning: Estimating Contact Forces by Observing Deformable Object Interactions",
                        "abstract": "Robotic manipulation of highly deformable cloth presents a promising opportunity to assist people with several daily tasks, such as washing dishes; folding laundry; or dressing, bathing, and hygiene assistance for individuals with severe motor impairments. In this work, we introduce a formulation that enables a collaborative robot to perform visual haptic reasoning with cloth -- the act of inferring the location and magnitude of applied forces during physical interaction. We present two distinct model representations, trained in physics simulation, that enable haptic reasoning using only visual and robot kinematic observations. We conducted quantitative evaluations of these models in simulation for robot-assisted dressing, bathing, and dish washing tasks, and demonstrate that the trained models can generalize across different tasks with varying interactions, human body sizes, and object shapes. We also present results with a real-world mobile manipulator, which used our simulation-trained models to estimate applied contact forces while performing physically assistive tasks with cloth. Videos can be found at our project webpage."
                    },
                    {
                        "title": "DiffTOP: Differentiable Trajectory Optimization for Deep Reinforcement and Imitation Learning",
                        "abstract": "This paper introduces DiffTOP, which utilizes Differentiable Trajectory OPtimization as the policy representation to generate actions for deep reinforcement and imitation learning. Trajectory optimization is a powerful and widely used algorithm in control, parameterized by a cost and a dynamics function. The key to our approach is to leverage the recent progress in differentiable trajectory optimization, which enables computing the gradients of the loss with respect to the parameters of trajectory optimization. As a result, the cost and dynamics functions of trajectory optimization can be learned end-to-end. DiffTOP addresses the ``objective mismatch'' issue of prior model-based RL algorithms, as the dynamics model in DiffTOP is learned to directly maximize task performance by differentiating the policy gradient loss through the trajectory optimization process. We further benchmark DiffTOP for imitation learning on standard robotic manipulation task suites with high-dimensional sensory observations and compare our method to feed-forward policy classes as well as Energy-Based Models (EBM) and Diffusion. Across 15 model-based RL tasks and 35imitation learning tasks with high-dimensional image and point cloud inputs, DiffTOP outperforms prior state-of-the-art methods in both domains."
                    },
                    {
                        "title": "Regular complete permutation polynomials over quadratic extension fields",
                        "abstract": "Let $r\\geq 3$ be any positive integer which is relatively prime to $p$ and $q^2\\equiv 1 \\pmod r$. Let $\\tau_1, \\tau_2$ be any permutation polynomials over $\\mathbb{F}_{q^2},$ $\\sigma_M$ is an invertible linear map over $\\mathbb{F}_{q^2}$ and $\\sigma=\\tau_1\\circ\\sigma_M\\circ\\tau_2$. In this paper, we prove that, for suitable $\\tau_1, \\tau_2$ and $\\sigma_M$, the map $\\sigma$ could be $r$-regular complete permutation polynomials over quadratic extension fields."
                    },
                    {
                        "title": "UniDual: A Unified Model for Image and Video Understanding",
                        "abstract": "Although a video is effectively a sequence of images, visual perception systems typically model images and videos separately, thus failing to exploit the correlation and the synergy provided by these two media. While a few prior research efforts have explored the benefits of leveraging still-image datasets for video analysis, or vice-versa, most of these attempts have been limited to pretraining a model on one type of visual modality and then adapting it via finetuning on the other modality. In contrast, in this paper we introduce a framework that enables joint training of a unified model on mixed collections of image and video examples spanning different tasks. The key ingredient in our architecture design is a new network block, which we name UniDual. It consists of a shared 2D spatial convolution followed by two parallel point-wise convolutional layers, one devoted to images and the other one used for videos. For video input, the point-wise filtering implements a temporal convolution. For image input, it performs a pixel-wise nonlinear transformation. Repeated stacking of such blocks gives rise to a network where images and videos undergo partially distinct execution pathways, unified by spatial convolutions (capturing commonalities in visual appearance) but separated by point-wise operations (modeling patterns specific to each modality). Extensive experiments on Kinetics and ImageNet demonstrate that our UniDual model jointly trained on these datasets yields substantial accuracy gains for both tasks, compared to 1) training separate models, 2) traditional multi-task learning and 3) the conventional framework of pretraining-followed-by-finetuning. On Kinetics, the UniDual architecture applied to a state-of-the-art video backbone model (R(2+1)D-152) yields an additional video@1 accuracy gain of 1.5%."
                    },
                    {
                        "title": "Learning Visible Connectivity Dynamics for Cloth Smoothing",
                        "abstract": "Robotic manipulation of cloth remains challenging for robotics due to the complex dynamics of the cloth, lack of a low-dimensional state representation, and self-occlusions. In contrast to previous model-based approaches that learn a pixel-based dynamics model or a compressed latent vector dynamics, we propose to learn a particle-based dynamics model from a partial point cloud observation. To overcome the challenges of partial observability, we infer which visible points are connected on the underlying cloth mesh. We then learn a dynamics model over this visible connectivity graph. Compared to previous learning-based approaches, our model poses strong inductive bias with its particle based representation for learning the underlying cloth physics; it is invariant to visual features; and the predictions can be more easily visualized. We show that our method greatly outperforms previous state-of-the-art model-based and model-free reinforcement learning methods in simulation. Furthermore, we demonstrate zero-shot sim-to-real transfer where we deploy the model trained in simulation on a Franka arm and show that the model can successfully smooth different types of cloth from crumpled configurations. Videos can be found on our project website."
                    },
                    {
                        "title": "One Policy to Dress Them All: Learning to Dress People with Diverse Poses and Garments",
                        "abstract": "Robot-assisted dressing could benefit the lives of many people such as older adults and individuals with disabilities. Despite such potential, robot-assisted dressing remains a challenging task for robotics as it involves complex manipulation of deformable cloth in 3D space. Many prior works aim to solve the robot-assisted dressing task, but they make certain assumptions such as a fixed garment and a fixed arm pose that limit their ability to generalize. In this work, we develop a robot-assisted dressing system that is able to dress different garments on people with diverse poses from partial point cloud observations, based on a learned policy. We show that with proper design of the policy architecture and Q function, reinforcement learning (RL) can be used to learn effective policies with partial point cloud observations that work well for dressing diverse garments. We further leverage policy distillation to combine multiple policies trained on different ranges of human arm poses into a single policy that works over a wide range of different arm poses. We conduct comprehensive real-world evaluations of our system with 510 dressing trials in a human study with 17 participants with different arm poses and dressed garments. Our system is able to dress 86% of the length of the participants' arms on average. Videos can be found on our project webpage: https://sites.google.com/view/one-policy-dress."
                    },
                    {
                        "title": "Decomposed Guided Dynamic Filters for Efficient RGB-Guided Depth Completion",
                        "abstract": "RGB-guided depth completion aims at predicting dense depth maps from sparse depth measurements and corresponding RGB images, where how to effectively and efficiently exploit the multi-modal information is a key issue. Guided dynamic filters, which generate spatially-variant depth-wise separable convolutional filters from RGB features to guide depth features, have been proven to be effective in this task. However, the dynamically generated filters require massive model parameters, computational costs and memory footprints when the number of feature channels is large. In this paper, we propose to decompose the guided dynamic filters into a spatially-shared component multiplied by content-adaptive adaptors at each spatial location. Based on the proposed idea, we introduce two decomposition schemes A and B, which decompose the filters by splitting the filter structure and using spatial-wise attention, respectively. The decomposed filters not only maintain the favorable properties of guided dynamic filters as being content-dependent and spatially-variant, but also reduce model parameters and hardware costs, as the learned adaptors are decoupled with the number of feature channels. Extensive experimental results demonstrate that the methods using our schemes outperform state-of-the-art methods on the KITTI dataset, and rank 1st and 2nd on the KITTI benchmark at the time of submission. Meanwhile, they also achieve comparable performance on the NYUv2 dataset. In addition, our proposed methods are general and could be employed as plug-and-play feature fusion blocks in other multi-modal fusion tasks such as RGB-D salient object detection."
                    }
                ]
            },
            "d41055be-b7b9-417d-b455-b7f757c26277": {
                "pk": "d41055be-b7b9-417d-b455-b7f757c26277",
                "name": "Alex Kot",
                "collaborators": [
                    "Rizhao Cai",
                    "Haoliang Li",
                    "Fan Zhang",
                    "Xiaobao Guo",
                    "Xiaojiang Peng",
                    "Olly Styles",
                    "Tanaya Guha",
                    "Victor Sanchez",
                    "Tao Bai",
                    "Jun Zhao"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Adversarial Learning",
                    "Domain Generalization"
                ],
                "publications": [
                    {
                        "title": "Toward domain generalized pruning by scoring out-of-distribution importance",
                        "abstract": "Filter pruning has been widely used for compressing convolutional neural networks to reduce computation costs during the deployment stage. Recent studies have shown that filter pruning techniques can achieve lossless compression of deep neural networks, reducing redundant filters (kernels) without sacrificing accuracy performance. However, the evaluation is done when the training and testing data are from similar environmental conditions (independent and identically distributed), and how the filter pruning techniques would affect the cross-domain generalization (out-of-distribution) performance is largely ignored. We conduct extensive empirical experiments and reveal that although the intra-domain performance could be maintained after filter pruning, the cross-domain performance will decay to a large extent. As scoring a filter's importance is one of the central problems for pruning, we design the importance scoring estimation by using the variance of domain-level risks to consider the pruning risk in the unseen distribution. As such, we can remain more domain generalized filters. The experiments show that under the same pruning ratio, our method can achieve significantly better cross-domain generalization performance than the baseline filter pruning method. For the first attempt, our work sheds light on the joint problem of domain generalization and filter pruning research."
                    },
                    {
                        "title": "MIMIC: Mask Image Pre-training with Mix Contrastive Fine-tuning for Facial Expression Recognition",
                        "abstract": "Cutting-edge research in facial expression recognition (FER) currently favors the utilization of convolutional neural networks (CNNs) backbone which is supervisedly pre-trained on face recognition datasets for feature extraction. However, due to the vast scale of face recognition datasets and the high cost associated with collecting facial labels, this pre-training paradigm incurs significant expenses. Towards this end, we propose to pre-train vision Transformers (ViTs) through a self-supervised approach on a mid-scale general image dataset. In addition, when compared with the domain disparity existing between face datasets and FER datasets, the divergence between general datasets and FER datasets is more pronounced. Therefore, we propose a contrastive fine-tuning approach to effectively mitigate this domain disparity. Specifically, we introduce a novel FER training paradigm named Mask Image pre-training with MIx Contrastive fine-tuning (MIMIC). In the initial phase, we pre-train the ViT via masked image reconstruction on general images. Subsequently, in the fine-tuning stage, we introduce a mix-supervised contrastive learning process, which enhances the model with a more extensive range of positive samples by the mixing strategy. Through extensive experiments conducted on three benchmark datasets, we demonstrate that our MIMIC outperforms the previous training paradigm, showing its capability to learn better representations. Remarkably, the results indicate that the vanilla ViT can achieve impressive performance without the need for intricate, auxiliary-designed modules. Moreover, when scaling up the model size, MIMIC exhibits no performance saturation and is superior to the current state-of-the-art methods."
                    },
                    {
                        "title": "Multi-Camera Trajectory Forecasting: Pedestrian Trajectory Prediction in a Network of Cameras",
                        "abstract": "We introduce the task of multi-camera trajectory forecasting (MCTF), where the future trajectory of an object is predicted in a network of cameras. Prior works consider forecasting trajectories in a single camera view. Our work is the first to consider the challenging scenario of forecasting across multiple non-overlapping camera views. This has wide applicability in tasks such as re-identification and multi-target multi-camera tracking. To facilitate research in this new area, we release the Warwick-NTU Multi-camera Forecasting Database (WNMF), a unique dataset of multi-camera pedestrian trajectories from a network of 15 synchronized cameras. To accurately label this large dataset (600 hours of video footage), we also develop a semi-automated annotation method. An effective MCTF model should proactively anticipate where and when a person will re-appear in the camera network. In this paper, we consider the task of predicting the next camera a pedestrian will re-appear after leaving the view of another camera, and present several baseline approaches for this. The labeled database is available online: https://github.com/olly-styles/Multi-Camera-Trajectory-Forecasting."
                    },
                    {
                        "title": "AI-GAN: Attack-Inspired Generation of Adversarial Examples",
                        "abstract": "Deep neural networks (DNNs) are vulnerable to adversarial examples, which are crafted by adding imperceptible perturbations to inputs. Recently different attacks and strategies have been proposed, but how to generate adversarial examples perceptually realistic and more efficiently remains unsolved. This paper proposes a novel framework called Attack-Inspired GAN (AI-GAN), where a generator, a discriminator, and an attacker are trained jointly. Once trained, it can generate adversarial perturbations efficiently given input images and target classes. Through extensive experiments on several popular datasets \\eg MNIST and CIFAR-10, AI-GAN achieves high attack success rates and reduces generation time significantly in various settings. Moreover, for the first time, AI-GAN successfully scales to complicated datasets \\eg CIFAR-100 with around $90\\%$ success rates among all classes."
                    },
                    {
                        "title": "On Generating Identifiable Virtual Faces",
                        "abstract": "Face anonymization with generative models have become increasingly prevalent since they sanitize private information by generating virtual face images, ensuring both privacy and image utility. Such virtual face images are usually not identifiable after the removal or protection of the original identity. In this paper, we formalize and tackle the problem of generating identifiable virtual face images. Our virtual face images are visually different from the original ones for privacy protection. In addition, they are bound with new virtual identities, which can be directly used for face recognition. We propose an Identifiable Virtual Face Generator (IVFG) to generate the virtual face images. The IVFG projects the latent vectors of the original face images into virtual ones according to a user specific key, based on which the virtual face images are generated. To make the virtual face images identifiable, we propose a multi-task learning objective as well as a triplet styled training strategy to learn the IVFG. We evaluate the performance of our virtual face images using different face recognizers on diffident face image datasets, all of which demonstrate the effectiveness of the IVFG for generate identifiable virtual face images."
                    }
                ]
            },
            "6ffc569e-5f72-4c0d-bcea-4288cd6aceeb": {
                "pk": "6ffc569e-5f72-4c0d-bcea-4288cd6aceeb",
                "name": "Bihan Wen",
                "collaborators": [
                    "Yoram Bresler",
                    "Saiprasad Ravishankar",
                    "Lanqing Guo",
                    "Yanjun Li",
                    "Satoshi Tsutsui",
                    "Chong Wang",
                    "Yuqi Li",
                    "Tao Bai",
                    "Jun Zhao",
                    "Way Tan"
                ],
                "domain": [
                    "Image Processing",
                    "Sparse Representation",
                    "Deep Learning",
                    "Medical Imaging"
                ],
                "publications": [
                    {
                        "title": "FRIST - Flipping and Rotation Invariant Sparsifying Transform Learning and Applications",
                        "abstract": "Features based on sparse representation, especially using the synthesis dictionary model, have been heavily exploited in signal processing and computer vision. However, synthesis dictionary learning typically involves NP-hard sparse coding and expensive learning steps. Recently, sparsifying transform learning received interest for its cheap computation and its optimal updates in the alternating algorithms. In this work, we develop a methodology for learning Flipping and Rotation Invariant Sparsifying Transforms, dubbed FRIST, to better represent natural images that contain textures with various geometrical directions. The proposed alternating FRIST learning algorithm involves efficient optimal updates. We provide a convergence guarantee, and demonstrate the empirical convergence behavior of the proposed FRIST learning approach. Preliminary experiments show the promising performance of FRIST learning for sparse image representation, segmentation, denoising, robust inpainting, and compressed sensing-based magnetic resonance image reconstruction."
                    },
                    {
                        "title": "VIDOSAT: High-dimensional Sparsifying Transform Learning for Online Video Denoising",
                        "abstract": "Techniques exploiting the sparsity of images in a transform domain have been effective for various applications in image and video processing. Transform learning methods involve cheap computations and have been demonstrated to perform well in applications such as image denoising and medical image reconstruction. Recently, we proposed methods for online learning of sparsifying transforms from streaming signals, which enjoy good convergence guarantees, and involve lower computational costs than online synthesis dictionary learning. In this work, we apply online transform learning to video denoising. We present a novel framework for online video denoising based on high-dimensional sparsifying transform learning for spatio-temporal patches. The patches are constructed either from corresponding 2D patches in successive frames or using an online block matching technique. The proposed online video denoising requires little memory, and offers efficient processing. Numerical experiments compare the performance to the proposed video denoising scheme but fixing the transform to be 3D DCT, as well as prior schemes such as dictionary learning-based schemes, and the state-of-the-art VBM3D and VBM4D on several video data sets, demonstrating the promising performance of the proposed methods."
                    },
                    {
                        "title": "The Power of Complementary Regularizers: Image Recovery via Transform Learning and Low-Rank Modeling",
                        "abstract": "Recent works on adaptive sparse and on low-rank signal modeling have demonstrated their usefulness in various image / video processing applications. Patch-based methods exploit local patch sparsity, whereas other works apply low-rankness of grouped patches to exploit image non-local structures. However, using either approach alone usually limits performance in image reconstruction or recovery applications. In this work, we propose a simultaneous sparsity and low-rank model, dubbed STROLLR, to better represent natural images. In order to fully utilize both the local and non-local image properties, we develop an image restoration framework using a transform learning scheme with joint low-rank regularization. The approach owes some of its computational efficiency and good performance to the use of transform learning for adaptive sparse representation rather than the popular synthesis dictionary learning algorithms, which involve approximation of NP-hard sparse coding and expensive learning steps. We demonstrate the proposed framework in various applications to image denoising, inpainting, and compressed sensing based magnetic resonance imaging. Results show promising performance compared to state-of-the-art competing methods."
                    },
                    {
                        "title": "A Set-Theoretic Study of the Relationships of Image Models and Priors for Restoration Problems",
                        "abstract": "Image prior modeling is the key issue in image recovery, computational imaging, compresses sensing, and other inverse problems. Recent algorithms combining multiple effective priors such as the sparse or low-rank models, have demonstrated superior performance in various applications. However, the relationships among the popular image models are unclear, and no theory in general is available to demonstrate their connections. In this paper, we present a theoretical analysis on the image models, to bridge the gap between applications and image prior understanding, including sparsity, group-wise sparsity, joint sparsity, and low-rankness, etc. We systematically study how effective each image model is for image restoration. Furthermore, we relate the denoising performance improvement by combining multiple models, to the image model relationships. Extensive experiments are conducted to compare the denoising results which are consistent with our analysis. On top of the model-based methods, we quantitatively demonstrate the image properties that are inexplicitly exploited by deep learning method, of which can further boost the denoising performance by combining with its complementary image models."
                    },
                    {
                        "title": "Adversarial Purification through Representation Disentanglement",
                        "abstract": "Deep learning models are vulnerable to adversarial examples and make incomprehensible mistakes, which puts a threat on their real-world deployment. Combined with the idea of adversarial training, preprocessing-based defenses are popular and convenient to use because of their task independence and good generalizability. Current defense methods, especially purification, tend to remove ``noise\" by learning and recovering the natural images. However, different from random noise, the adversarial patterns are much easier to be overfitted during model training due to their strong correlation to the images. In this work, we propose a novel adversarial purification scheme by presenting disentanglement of natural images and adversarial perturbations as a preprocessing defense. With extensive experiments, our defense is shown to be generalizable and make significant protection against unseen strong adversarial attacks. It reduces the success rates of state-of-the-art \\textbf{ensemble} attacks from \\textbf{61.7\\%} to \\textbf{14.9\\%} on average, superior to a number of existing methods. Notably, our defense restores the perturbed images perfectly and does not hurt the clean accuracy of backbone models, which is highly desirable in practice."
                    },
                    {
                        "title": "Systematic Analysis and Removal of Circular Artifacts for StyleGAN",
                        "abstract": "StyleGAN is one of the state-of-the-art image generators which is well-known for synthesizing high-resolution and hyper-realistic face images. Though images generated by vanilla StyleGAN model are visually appealing, they sometimes contain prominent circular artifacts which severely degrade the quality of generated images. In this work, we provide a systematic investigation on how those circular artifacts are formed by studying the functionalities of different stages of vanilla StyleGAN architecture, with both mechanism analysis and extensive experiments. The key modules of vanilla StyleGAN that promote such undesired artifacts are highlighted. Our investigation also explains why the artifacts are usually circular, relatively small and rarely split into 2 or more parts. Besides, we propose a simple yet effective solution to remove the prominent circular artifacts for vanilla StyleGAN, by applying a novel pixel-instance normalization (PIN) layer."
                    },
                    {
                        "title": "FINO: Flow-based Joint Image and Noise Model",
                        "abstract": "One of the fundamental challenges in image restoration is denoising, where the objective is to estimate the clean image from its noisy measurements. To tackle such an ill-posed inverse problem, the existing denoising approaches generally focus on exploiting effective natural image priors. The utilization and analysis of the noise model are often ignored, although the noise model can provide complementary information to the denoising algorithms. In this paper, we propose a novel Flow-based joint Image and NOise model (FINO) that distinctly decouples the image and noise in the latent space and losslessly reconstructs them via a series of invertible transformations. We further present a variable swapping strategy to align structural information in images and a noise correlation matrix to constrain the noise based on spatially minimized correlation information. Experimental results demonstrate FINO's capacity to remove both synthetic additive white Gaussian noise (AWGN) and real noise. Furthermore, the generalization of FINO to the removal of spatially variant noise and noise with inaccurate estimation surpasses that of the popular and state-of-the-art methods by large margins."
                    },
                    {
                        "title": "Benchmarking White Blood Cell Classification Under Domain Shift",
                        "abstract": "Recognizing the types of white blood cells (WBCs) in microscopic images of human blood smears is a fundamental task in the fields of pathology and hematology. Although previous studies have made significant contributions to the development of methods and datasets, few papers have investigated benchmarks or baselines that others can easily refer to. For instance, we observed notable variations in the reported accuracies of the same Convolutional Neural Network (CNN) model across different studies, yet no public implementation exists to reproduce these results. In this paper, we establish a benchmark for WBC recognition. Our results indicate that CNN-based models achieve high accuracy when trained and tested under similar imaging conditions. However, their performance drops significantly when tested under different conditions. Moreover, the ResNet classifier, which has been widely employed in previous work, exhibits an unreasonably poor generalization ability under domain shifts due to batch normalization. We investigate this issue and suggest some alternative normalization techniques that can mitigate it. We make fully-reproducible code publicly available\\footnote{\\url{https://github.com/apple2373/wbc-benchmark}}."
                    },
                    {
                        "title": "Benchmarking Adversarial Robustness of Image Shadow Removal with Shadow-adaptive Attacks",
                        "abstract": "Shadow removal is a task aimed at erasing regional shadows present in images and reinstating visually pleasing natural scenes with consistent illumination. While recent deep learning techniques have demonstrated impressive performance in image shadow removal, their robustness against adversarial attacks remains largely unexplored. Furthermore, many existing attack frameworks typically allocate a uniform budget for perturbations across the entire input image, which may not be suitable for attacking shadow images. This is primarily due to the unique characteristic of spatially varying illumination within shadow images. In this paper, we propose a novel approach, called shadow-adaptive adversarial attack. Different from standard adversarial attacks, our attack budget is adjusted based on the pixel intensity in different regions of shadow images. Consequently, the optimized adversarial noise in the shadowed regions becomes visually less perceptible while permitting a greater tolerance for perturbations in non-shadow regions. The proposed shadow-adaptive attacks naturally align with the varying illumination distribution in shadow images, resulting in perturbations that are less conspicuous. Building on this, we conduct a comprehensive empirical evaluation of existing shadow removal methods, subjecting them to various levels of attack on publicly available datasets."
                    },
                    {
                        "title": "SegPoint: Segment Any Point Cloud via Large Language Model",
                        "abstract": "Despite significant progress in 3D point cloud segmentation, existing methods primarily address specific tasks and depend on explicit instructions to identify targets, lacking the capability to infer and understand implicit user intentions in a unified framework. In this work, we propose a model, called SegPoint, that leverages the reasoning capabilities of a multi-modal Large Language Model (LLM) to produce point-wise segmentation masks across a diverse range of tasks: 1) 3D instruction segmentation, 2) 3D referring segmentation, 3) 3D semantic segmentation, and 4) 3D open-vocabulary semantic segmentation. To advance 3D instruction research, we introduce a new benchmark, Instruct3D, designed to evaluate segmentation performance from complex and implicit instructional texts, featuring 2,565 point cloud-instruction pairs. Our experimental results demonstrate that SegPoint achieves competitive performance on established benchmarks such as ScanRefer for referring segmentation and ScanNet for semantic segmentation, while delivering outstanding outcomes on the Instruct3D dataset. To our knowledge, SegPoint is the first model to address these varied segmentation tasks within a single framework, achieving satisfactory performance."
                    },
                    {
                        "title": "Removing Backdoor-Based Watermarks in Neural Networks with Limited Data",
                        "abstract": "Deep neural networks have been widely applied and achieved great success in various fields. As training deep models usually consumes massive data and computational resources, trading the trained deep models is highly demanded and lucrative nowadays. Unfortunately, the naive trading schemes typically involves potential risks related to copyright and trustworthiness issues, e.g., a sold model can be illegally resold to others without further authorization to reap huge profits. To tackle this problem, various watermarking techniques are proposed to protect the model intellectual property, amongst which the backdoor-based watermarking is the most commonly-used one. However, the robustness of these watermarking approaches is not well evaluated under realistic settings, such as limited in-distribution data availability and agnostic of watermarking patterns. In this paper, we benchmark the robustness of watermarking, and propose a novel backdoor-based watermark removal framework using limited data, dubbed WILD. The proposed WILD removes the watermarks of deep models with only a small portion of training data, and the output model can perform the same as models trained from scratch without watermarks injected. In particular, a novel data augmentation method is utilized to mimic the behavior of watermark triggers. Combining with the distribution alignment between the normal and perturbed (e.g., occluded) data in the feature space, our approach generalizes well on all typical types of trigger contents. The experimental results demonstrate that our approach can effectively remove the watermarks without compromising the deep model performance for the original task with the limited access to training data."
                    },
                    {
                        "title": "Reconciliation of Statistical and Spatial Sparsity For Robust Image and Image-Set Classification",
                        "abstract": "Recent image classification algorithms, by learning deep features from large-scale datasets, have achieved significantly better results comparing to the classic feature-based approaches. However, there are still various challenges of image classifications in practice, such as classifying noisy image or image-set queries and training deep image classification models over the limited-scale dataset. Instead of applying generic deep features, the model-based approaches can be more effective and data-efficient for robust image and image-set classification tasks, as various image priors are exploited for modeling the inter- and intra-set data variations while preventing over-fitting. In this work, we propose a novel Joint Statistical and Spatial Sparse representation, dubbed \\textit{J3S}, to model the image or image-set data for classification, by reconciling both their local patch structures and global Gaussian distribution mapped into Riemannian manifold. To the best of our knowledge, no work to date utilized both global statistics and local patch structures jointly via joint sparse representation. We propose to solve the joint sparse coding problem based on the J3S model, by coupling the local and global image representations using joint sparsity. The learned J3S models are used for robust image and image-set classification. Experiments show that the proposed J3S-based image classification scheme outperforms the popular or state-of-the-art competing methods over FMD, UIUC, ETH-80 and YTC databases."
                    },
                    {
                        "title": "REPNP: Plug-and-Play with Deep Reinforcement Learning Prior for Robust Image Restoration",
                        "abstract": "Image restoration schemes based on the pre-trained deep models have received great attention due to their unique flexibility for solving various inverse problems. In particular, the Plug-and-Play (PnP) framework is a popular and powerful tool that can integrate an off-the-shelf deep denoiser for different image restoration tasks with known observation models. However, obtaining the observation model that exactly matches the actual one can be challenging in practice. Thus, the PnP schemes with conventional deep denoisers may fail to generate satisfying results in some real-world image restoration tasks. We argue that the robustness of the PnP framework is largely limited by using the off-the-shelf deep denoisers that are trained by deterministic optimization. To this end, we propose a novel deep reinforcement learning (DRL) based PnP framework, dubbed RePNP, by leveraging a light-weight DRL-based denoiser for robust image restoration tasks. Experimental results demonstrate that the proposed RePNP is robust to the observation model used in the PnP scheme deviating from the actual one. Thus, RePNP can generate more reliable restoration results for image deblurring and super resolution tasks. Compared with several state-of-the-art deep image restoration baselines, RePNP achieves better results subjective to model deviation with fewer model parameters."
                    },
                    {
                        "title": "WBCAtt: A White Blood Cell Dataset Annotated with Detailed Morphological Attributes",
                        "abstract": "The examination of blood samples at a microscopic level plays a fundamental role in clinical diagnostics, influencing a wide range of medical conditions. For instance, an in-depth study of White Blood Cells (WBCs), a crucial component of our blood, is essential for diagnosing blood-related diseases such as leukemia and anemia. While multiple datasets containing WBC images have been proposed, they mostly focus on cell categorization, often lacking the necessary morphological details to explain such categorizations, despite the importance of explainable artificial intelligence (XAI) in medical domains. This paper seeks to address this limitation by introducing comprehensive annotations for WBC images. Through collaboration with pathologists, a thorough literature review, and manual inspection of microscopic images, we have identified 11 morphological attributes associated with the cell and its components (nucleus, cytoplasm, and granules). We then annotated ten thousand WBC images with these attributes. Moreover, we conduct experiments to predict these attributes from images, providing insights beyond basic WBC classification. As the first public dataset to offer such extensive annotations, we also illustrate specific applications that can benefit from our attribute annotations. Overall, our dataset paves the way for interpreting WBC recognition models, further advancing XAI in the fields of pathology and hematology."
                    },
                    {
                        "title": "DP-IQA: Utilizing Diffusion Prior for Blind Image Quality Assessment in the Wild",
                        "abstract": "Blind image quality assessment (IQA) in the wild, which assesses the quality of images with complex authentic distortions and no reference images, presents significant challenges. Given the difficulty in collecting large-scale training data, leveraging limited data to develop a model with strong generalization remains an open problem. Motivated by the robust image perception capabilities of pre-trained text-to-image (T2I) diffusion models, we propose a novel IQA method, diffusion priors-based IQA (DP-IQA), to utilize the T2I model's prior for improved performance and generalization ability. Specifically, we utilize pre-trained Stable Diffusion as the backbone, extracting multi-level features from the denoising U-Net guided by prompt embeddings through a tunable text adapter. Simultaneously, an image adapter compensates for information loss introduced by the lossy pre-trained encoder. Unlike T2I models that require full image distribution modeling, our approach targets image quality assessment, which inherently requires fewer parameters. To improve applicability, we distill the knowledge into a lightweight CNN-based student model, significantly reducing parameters while maintaining or even enhancing generalization performance. Experimental results demonstrate that DP-IQA achieves state-of-the-art performance on various in-the-wild datasets, highlighting the superior generalization capability of T2I priors in blind IQA tasks. To our knowledge, DP-IQA is the first method to apply pre-trained diffusion priors in blind IQA. Codes and checkpoints are available at https://github.com/RomGai/DP-IQA."
                    },
                    {
                        "title": "Transform Learning for Magnetic Resonance Image Reconstruction: From Model-based Learning to Building Neural Networks",
                        "abstract": "Magnetic resonance imaging (MRI) is widely used in clinical practice, but it has been traditionally limited by its slow data acquisition. Recent advances in compressed sensing (CS) techniques for MRI reduce acquisition time while maintaining high image quality. Whereas classical CS assumes the images are sparse in known analytical dictionaries or transform domains, methods using learned image models for reconstruction have become popular. The model could be pre-learned from datasets, or learned simultaneously with the reconstruction, i.e., blind CS (BCS). Besides the well-known synthesis dictionary model, recent advances in transform learning (TL) provide an efficient alternative framework for sparse modeling in MRI. TL-based methods enjoy numerous advantages including exact sparse coding, transform update, and clustering solutions, cheap computation, and convergence guarantees, and provide high-quality results in MRI compared to popular competing methods. This paper provides a review of some recent works in MRI reconstruction from limited data, with focus on the recent TL-based methods. A unified framework for incorporating various TL-based models is presented. We discuss the connections between transform learning and convolutional or filter bank models and corresponding multi-layer extensions, with connections to deep learning. Finally, we discuss recent trends in MRI, open problems, and future directions for the field."
                    }
                ]
            }
        }
    },
    "2409.19734": {
        "paper_data": {
            "title": "T2Vs Meet VLMs: A Scalable Multimodal Dataset for Visual Harmfulness Recognition",
            "url": "http://arxiv.org/abs/2409.19734v2",
            "arxiv_id": "2409.19734",
            "authors": [
                "Chen Yeh",
                "You-Ming Chang",
                "Wei-Chen Chiu",
                "Ning Yu"
            ],
            "abstract": "To address the risks of encountering inappropriate or harmful content, researchers managed to incorporate several harmful contents datasets with machine learning methods to detect harmful concepts. However, existing harmful datasets are curated by the presence of a narrow range of harmful objects, and only cover real harmful content sources. This hinders the generalizability of methods based on such datasets, potentially leading to misjudgments. Therefore, we propose a comprehensive harmful dataset, Visual Harmful Dataset 11K (VHD11K), consisting of 10,000 images and 1,000 videos, crawled from the Internet and generated by 4 generative models, across a total of 10 harmful categories covering a full spectrum of harmful concepts with nontrivial definition. We also propose a novel annotation framework by formulating the annotation process as a multi-agent Visual Question Answering (VQA) task, having 3 different VLMs \"debate\" about whether the given image/video is harmful, and incorporating the in-context learning strategy in the debating process. Therefore, we can ensure that the VLMs consider the context of the given image/video and both sides of the arguments thoroughly before making decisions, further reducing the likelihood of misjudgments in edge cases. Evaluation and experimental results demonstrate that (1) the great alignment between the annotation from our novel annotation framework and those from human, ensuring the reliability of VHD11K; (2) our full-spectrum harmful dataset successfully identifies the inability of existing harmful content detection methods to detect extensive harmful contents and improves the performance of existing harmfulness recognition methods; (3) VHD11K outperforms the baseline dataset, SMID, as evidenced by the superior improvement in harmfulness recognition methods. The complete dataset and code can be found at https://github.com/nctu-eva-lab/VHD11K.",
            "introduction": "   1 Introduction  Nowadays, visual data counts for more than 82% of the Internet traffic[26]. However, this widespread access also increases the risk that underage children encounter harmful or inappropriate contents, highlighting the urgent need for effective recognition methods. Furthermore, the rapid advancement of generative models has made it easier to create and spread harmful contents, underscoring the critical importance of developing methods to detect such synthesized harmful materials.   Currently, most methods for recognizing harmful contents rely on the power of machine learning and neural networks, which perform best with high-quality training datasets. Therefore, having a comprehensive and extensive harmful contents dataset is essential to effectively address this issue using neural networks. However, existing harmful contents datasets have several limitations. For instance, some datasets [15, 33] only cover a narrow range of harmful categories such as \u201chandgun\u201d, \u201cknives\u201d, \u201ccigarettes\u201d, etc., restricting the generalizability of recognition methods. Furthermore, the definition of \u201charmful\u201d is complex and sometimes ambiguous, whereas some datasets [15, 33] focus on detecting harmful objects such as knives or guns without considering the context of the entire image, leading to potential false positive misjudgments in certain cases. Finally, most current harmful contents datasets [15, 8] mainly include images from the real world, overlooking the importance of harmful videos and synthesized harmful content.   To address the aforementioned challenges, we propose a scalable multimodal harmful dataset, Visual Harmful Dataset 11K (VHD11K), and a novel \u201cdebate\u201d annotation framework. The dataset includes 10,000 images and 1,000 videos across 10 categories, which contain real and synthesized samples. For annotation, we first collect raw images and videos from the Internet and 4 generative models, then we structure the annotation process as a multi-agent Visual Question Answering (VQA) task, incorporating pretrained vision-language models (VLMs) as annotators. Thanks to their visual understanding capability, the annotation takes account of the context of the whole image or the entire video through the frames extracted from itself, rather than only the potentially harmful objects, enhancing the reliability of our dataset. Specifically, we employ 3 different VLMs in the roles of a \u201cjudge\u201d, an \u201caffirmative debater\u201d, and a \u201cnegative debater\u201d utilizing the AutoGen framework [47]. These VLMs \u201cdebate\u201d whether a given sample is harmful based on several provided definitions from HarmBench [28] and Safe Latent Diffusion [42] by asking the VLMs a question such as \u201cIs the given image harmful?\u201d. This debating process ensures that VLMs consider both sides of the argument before making decisions, further reducing the likelihood of misjudgments in edge cases. The detailed arguments presented by the affirmative and negative debaters, composed of 10,228,011 words in total, also provide a set of diverse aspects of detecting harmfulness of the given visual content from the context, which are valuable for future studies and are included in our dataset. Furthermore, we use in-context learning [29] to resolve ambiguous cases, ensuring alignment between the annotators and the definition of harmful contents, thereby refining our dataset.   Finally, we distillate the harmful reasons from all the annotations using a Large Language Model (LLM), i.e., LlaMa3, and obtain 10 harmful categories that represent the harmful contents of all images and videos: \u201cViolence and Threats\u201d, \u201cSubstance Misuse\u201d,",
            "references": [
                {
                    "title": "StreamingT2V: Consistent, Dynamic, and Extendable Long Video Generation from Text",
                    "abstract": "Text-to-video diffusion models enable the generation of high-quality videos that follow text instructions, making it easy to create diverse and individual content. However, existing approaches mostly focus on high-quality short video generation (typically 16 or 24 frames), ending up with hard-cuts when naively extended to the case of long video synthesis. To overcome these limitations, we introduce StreamingT2V, an autoregressive approach for long video generation of 80, 240, 600, 1200 or more frames with smooth transitions. The key components are:(i) a short-term memory block called conditional attention module (CAM), which conditions the current generation on the features extracted from the previous chunk via an attentional mechanism, leading to consistent chunk transitions, (ii) a long-term memory block called appearance preservation module, which extracts high-level scene and object features from the first video chunk to prevent the model from forgetting the initial scene, and (iii) a randomized blending approach that enables to apply a video enhancer autoregressively for infinitely long videos without inconsistencies between chunks. Experiments show that StreamingT2V generates high motion amount. In contrast, all competing image-to-video methods are prone to video stagnation when applied naively in an autoregressive manner. Thus, we propose with StreamingT2V a high-quality seamless text-to-long video generator that outperforms competitors with consistency and motion. Our code will be available at: https://github.com/Picsart-AI-Research/StreamingT2V"
                },
                {
                    "title": "HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal",
                    "abstract": "Automated red teaming holds substantial promise for uncovering and mitigating the risks associated with the malicious use of large language models (LLMs), yet the field lacks a standardized evaluation framework to rigorously assess new methods. To address this issue, we introduce HarmBench, a standardized evaluation framework for automated red teaming. We identify several desirable properties previously unaccounted for in red teaming evaluations and systematically design HarmBench to meet these criteria. Using HarmBench, we conduct a large-scale comparison of 18 red teaming methods and 33 target LLMs and defenses, yielding novel insights. We also introduce a highly efficient adversarial training method that greatly enhances LLM robustness across a wide range of attacks, demonstrating how HarmBench enables codevelopment of attacks and defenses. We open source HarmBench at https://github.com/centerforaisafety/HarmBench."
                },
                {
                    "title": "VideoCrafter2: Overcoming Data Limitations for High-Quality Video Diffusion Models",
                    "abstract": "Text-to-video generation aims to produce a video based on a given prompt. Recently, several commercial video models have been able to generate plausible videos with mini-mal noise, excellent details, and high aesthetic scores. However, these models rely on large-scale, well-filtered, high-quality videos that are not accessible to the community. Many existing research works, which train models using the low-quality WebVid-10M dataset, struggle to generate high-quality videos because the models are optimized to fit WebVid-10M. In this work, we explore the training scheme of video models extended from Stable Diffusion and investigate the feasibility of leveraging low-quality videos and synthesized high-quality images to obtain a high-quality video model. We first analyze the connection between the spatial and temporal modules of video models and the distribution shift to low-quality videos. We observe that full training of all modules results in a stronger coupling between spatial and temporal modules than only training temporal modules. Based on this stronger coupling, we shift the distribution to higher quality without motion degradation by finetuning spatial modules with high-quality images, resulting in a generic high-quality video model. Evaluations are conducted to demonstrate the superiority of the proposed method, particularly in picture quality, motion, and concept composition."
                },
                {
                    "title": "HOD: New Harmful Object Detection Benchmarks for Robust Surveillance",
                    "abstract": "Recent multi-media data such as images and videos have been rapidly spread out on various online services such as social network services (SNS). With the explosive growth of online media services, the number of image content that may harm users is also growing exponentially. Therefore, the surveillance of these images is crucial. Thus, most recent online platforms such as Facebook and Instagram have adopted content filtering systems to prevent the prevalence of harmful content and reduce the possible risk of adverse effects on users. Unfortunately, computer vision research on detecting harmful content has not yet attracted attention enough. Users of each platform still manually click the report button to recognize patterns of harmful content they dislike when ex-posed to harmful content. However, the problem with manual reporting is that users are already exposed to harmful content. To address these issues, our research goal in this work is to develop automatic harmful object detection systems for online services. We present a new benchmark dataset for harm-ful object detection. Unlike most related studies focusing on a small subset of object categories, our dataset addresses var-ious categories. Specifically, our proposed dataset contains more than 10,000 images across 6 categories that might be harmful, consisting of not only normal cases but also hard cases that are difficult to detect. Moreover, we have conducted extensive experiments to evaluate the effectiveness of our proposed dataset. We have utilized recent proposed state-of-the-art object detection architectures and shown our proposed dataset can be greatly useful for the realtime harmful object detection task. The codes and datasets are available at https://github.com/poori-nuna/HOD-Benchmark-Dataset."
                },
                {
                    "title": "Hierarchical Spatio-temporal Decoupling for Text-to- Video Generation",
                    "abstract": "Despite diffusion models having shown powerful abilities to generate photorealistic images, generating videos that are realistic and diverse still remains in its infancy. One of the key reasons is that current methods intertwine spatial content and temporal dynamics together, leading to a notably increased complexity of text-to-video generation (T2V). In this work, we propose HiGen, a diffusion model-based method that improves performance by decoupling the spatial and temporal factors of videos from two perspectives, i.e., structure level and content level. At the structure level, we decompose the T2V task into two steps, including spatial reasoning and temporal reasoning, using a unified denoiser. Specifically, we generate spatially coherent priors using text during spatial reasoning and then generate temporally coherent motions from these priors during temporal reasoning. At the content level, we extract two subtle cues from the content of the input video that can express motion and appearance changes, respectively. These two cues then guide the model's training for generating videos, enabling flexible content variations and enhancing temporal stability. Through the decoupled paradigm, HiGen can effectively reduce the complexity of this task and generate realistic videos with semantics accuracy and motion stability. Extensive experiments demonstrate the superior performance of HiGen over the state-of-the-art T2V methods. We have released our source code and models."
                },
                {
                    "title": "SparseCtrl: Adding Sparse Controls to Text-to-Video Diffusion Models",
                    "abstract": "The development of text-to-video (T2V), i.e., generating videos with a given text prompt, has been significantly advanced in recent years. However, relying solely on text prompts often results in ambiguous frame composition due to spatial uncertainty. The research community thus leverages the dense structure signals, e.g., per-frame depth/edge sequences, to enhance controllability, whose collection accordingly increases the burden of inference. In this work, we present SparseCtrl to enable flexible structure control with temporally sparse signals, requiring only one or a few inputs, as shown in Figure 1. It incorporates an additional condition encoder to process these sparse signals while leaving the pre-trained T2V model untouched. The proposed approach is compatible with various modalities, including sketches, depth maps, and RGB images, providing more practical control for video generation and promoting applications such as storyboarding, depth rendering, keyframe animation, and interpolation. Extensive experiments demonstrate the generalization of SparseCtrl on both original and personalized T2V generators. Codes and models will be publicly available at https://guoyww.github.io/projects/SparseCtrl ."
                },
                {
                    "title": "Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets",
                    "abstract": "We present Stable Video Diffusion - a latent video diffusion model for high-resolution, state-of-the-art text-to-video and image-to-video generation. Recently, latent diffusion models trained for 2D image synthesis have been turned into generative video models by inserting temporal layers and finetuning them on small, high-quality video datasets. However, training methods in the literature vary widely, and the field has yet to agree on a unified strategy for curating video data. In this paper, we identify and evaluate three different stages for successful training of video LDMs: text-to-image pretraining, video pretraining, and high-quality video finetuning. Furthermore, we demonstrate the necessity of a well-curated pretraining dataset for generating high-quality videos and present a systematic curation process to train a strong base model, including captioning and filtering strategies. We then explore the impact of finetuning our base model on high-quality data and train a text-to-video model that is competitive with closed-source video generation. We also show that our base model provides a powerful motion representation for downstream tasks such as image-to-video generation and adaptability to camera motion-specific LoRA modules. Finally, we demonstrate that our model provides a strong multi-view 3D-prior and can serve as a base to finetune a multi-view diffusion model that jointly generates multiple views of objects in a feedforward fashion, outperforming image-based methods at a fraction of their compute budget. We release code and model weights at https://github.com/Stability-AI/generative-models ."
                },
                {
                    "title": "mPLUG-OwI2: Revolutionizing Multi-modal Large Language Model with Modality Collaboration",
                    "abstract": "Multi-modal Large Language Models (MLLMs) have demonstrated impressive instruction abilities across various open-ended tasks. However, previous methods primarily fo-cus on enhancing multi-modal capabilities. In this work, we introduce a versatile multi-modal large language model, mPLUG-Owl2, which effectively leverages modality collab-oration to improve performance in both text and multi-modal tasks. mPLUG-Owl2 utilizes a modularized network design, with the language decoder acting as a universal interface for managing different modalities. Specifically, mPLUG-Owl2 incorporates shared functional modules to facilitate modal-ity collaboration and introduces a modality-adaptive module that preserves modality-specific features. Extensive experi-ments reveal that mPLUG-Owl2 is capable of generalizing both text tasks and multi-modal tasks and achieving state-of-the-art performances with a single generic model. Notably, mPLUG-Owl2 is the first MLLM model that demonstrates the modality collaboration phenomenon in both pure-text and multi-modal scenarios, setting a pioneering path in the development of future multi-modal foundation models."
                },
                {
                    "title": "CogVLM: Visual Expert for Pretrained Language Models",
                    "abstract": "We introduce CogVLM, a powerful open-source visual language foundation model. Different from the popular shallow alignment method which maps image features into the input space of language model, CogVLM bridges the gap between the frozen pretrained language model and image encoder by a trainable visual expert module in the attention and FFN layers. As a result, CogVLM enables deep fusion of vision language features without sacrificing any performance on NLP tasks. CogVLM-17B achieves state-of-the-art performance on 10 classic cross-modal benchmarks, including NoCaps, Flicker30k captioning, RefCOCO, RefCOCO+, RefCOCOg, Visual7W, GQA, ScienceQA, VizWiz VQA and TDIUC, and ranks the 2nd on VQAv2, OKVQA, TextVQA, COCO captioning, etc., surpassing or matching PaLI-X 55B. Codes and checkpoints are available at https://github.com/THUDM/CogVLM."
                },
                {
                    "title": "AntifakePrompt: Prompt-Tuned Vision-Language Models are Fake Image Detectors",
                    "abstract": "Deep generative models can create remarkably photorealistic fake images while raising concerns about misinformation and copyright infringement, known as deepfake threats. Deepfake detection technique is developed to distinguish between real and fake images, where the existing methods typically learn classifiers in the image domain or various feature domains. However, the generalizability of deepfake detection against emerging and more advanced generative models remains challenging. In this paper, being inspired by the zero-shot advantages of Vision-Language Models (VLMs), we propose a novel approach using VLMs (e.g. InstructBLIP) and prompt tuning techniques to improve the deepfake detection accuracy over unseen data. We formulate deepfake detection as a visual question answering problem, and tune soft prompts for InstructBLIP to answer the real/fake information of a query image. We conduct full-spectrum experiments on datasets from 3 held-in and 13 held-out generative models, covering modern text-to-image generation, image editing and image attacks. Results demonstrate that (1) the deepfake detection accuracy can be significantly and consistently improved (from 58.8% to 91.31%, in average accuracy over unseen data) using pretrained vision-language models with prompt tuning; (2) our superior performance is at less cost of trainable parameters, resulting in an effective and efficient solution for deepfake detection. Code and models can be found at https://github.com/nctu-eva-lab/AntifakePrompt."
                },
                {
                    "title": "More human than human: measuring ChatGPT political bias",
                    "abstract": "We investigate the political bias of a large language model (LLM), ChatGPT, which has become popular for retrieving factual information and generating content. Although ChatGPT assures that it is impartial, the literature suggests that LLMs exhibit bias involving race, gender, religion, and political orientation. Political bias in LLMs can have adverse political and electoral consequences similar to bias from traditional and social media. Moreover, political bias can be harder to detect and eradicate than gender or racial bias. We propose a novel empirical design to infer whether ChatGPT has political biases by requesting it to impersonate someone from a given side of the political spectrum and comparing these answers with its default. We also propose dose-response, placebo, and profession-politics alignment robustness tests. To reduce concerns about the randomness of the generated text, we collect answers to the same questions 100 times, with question order randomized on each round. We find robust evidence that ChatGPT presents a significant and systematic political bias toward the Democrats in the US, Lula in Brazil, and the Labour Party in the UK. These results translate into real concerns that ChatGPT, and LLMs in general, can extend or even amplify the existing challenges involving political processes posed by the Internet and social media. Our findings have important implications for policymakers, media, politics, and academia stakeholders."
                },
                {
                    "title": "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis",
                    "abstract": "We present SDXL, a latent diffusion model for text-to-image synthesis. Compared to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet backbone: The increase of model parameters is mainly due to more attention blocks and a larger cross-attention context as SDXL uses a second text encoder. We design multiple novel conditioning schemes and train SDXL on multiple aspect ratios. We also introduce a refinement model which is used to improve the visual fidelity of samples generated by SDXL using a post-hoc image-to-image technique. We demonstrate that SDXL shows drastically improved performance compared the previous versions of Stable Diffusion and achieves results competitive with those of black-box state-of-the-art image generators. In the spirit of promoting open research and fostering transparency in large model training and evaluation, we provide access to code and model weights at https://github.com/Stability-AI/generative-models"
                },
                {
                    "title": "LLaVAR: Enhanced Visual Instruction Tuning for Text-Rich Image Understanding",
                    "abstract": "Instruction tuning unlocks the superior capability of Large Language Models (LLM) to interact with humans. Furthermore, recent instruction-following datasets include images as visual inputs, collecting responses for image-based instructions. However, visual instruction-tuned models cannot comprehend textual details within images well. This work enhances the current visual instruction tuning pipeline with text-rich images (e.g., movie posters, book covers, etc.). Specifically, we first use publicly available OCR tools to collect results on 422K text-rich images from the LAION dataset. Moreover, we prompt text-only GPT-4 with recognized texts and image captions to generate 16K conversations, each containing question-answer pairs for text-rich images. By combining our collected data with previous multi-modal instruction-following data, our model, LLaVAR, substantially improves the LLaVA model's capability on text-based VQA datasets (up to 20% accuracy improvement) while achieving an accuracy of 91.42% on ScienceQA. The GPT-4-based instruction-following evaluation also demonstrates the improvement of our model on both natural images and text-rich images. Through qualitative analysis, LLaVAR shows promising interaction (e.g., reasoning, writing, and elaboration) skills with humans based on the latest real-world online content that combines text and images. We make our code/data/models publicly available at https://llavar.github.io/."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning",
                    "abstract": "Large-scale pre-training and instruction tuning have been successful at creating general-purpose language models with broad competence. However, building general-purpose vision-language models is challenging due to the rich input distributions and task diversity resulting from the additional visual input. Although vision-language pretraining has been widely studied, vision-language instruction tuning remains under-explored. In this paper, we conduct a systematic and comprehensive study on vision-language instruction tuning based on the pretrained BLIP-2 models. We gather 26 publicly available datasets, covering a wide variety of tasks and capabilities, and transform them into instruction tuning format. Additionally, we introduce an instruction-aware Query Transformer, which extracts informative features tailored to the given instruction. Trained on 13 held-in datasets, InstructBLIP attains state-of-the-art zero-shot performance across all 13 held-out datasets, substantially outperforming BLIP-2 and larger Flamingo models. Our models also lead to state-of-the-art performance when finetuned on individual downstream tasks (e.g., 90.7% accuracy on ScienceQA questions with image contexts). Furthermore, we qualitatively demonstrate the advantages of InstructBLIP over concurrent multimodal models. All InstructBLIP models are open-sourced at https://github.com/salesforce/LAVIS/tree/main/projects/instructblip."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Models",
                    "abstract": "Text-conditioned image generation models have recently achieved astonishing results in image quality and text alignment and are consequently employed in a fast-growing number of applications. Since they are highly data-driven, relying on billion-sized datasets randomly scraped from the internet, they also suffer, as we demonstrate, from degenerated and biased human behavior. In turn, they may even reinforce such biases. To help combat these undesired side effects, we present safe latent diffusion (SLD). Specifically, to measure the inappropriate degeneration due to unfiltered and imbalanced training sets, we establish a novel image generation test bed-inappropriate image prompts (I2P)-containing dedicated, real-world image-to-text prompts covering concepts such as nudity and violence. As our exhaustive empirical evaluation demonstrates, the introduced SLD removes and suppresses inappropriate image parts during the diffusion process, with no additional training required and no adverse effect on overall image quality or text alignment.11Code available at https://huggingface.co/docs/diffusers/api/pipelines/stable.diffusion.safe"
                },
                {
                    "title": "Scaling Instruction-Finetuned Language Models",
                    "abstract": "Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models."
                },
                {
                    "title": "Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering",
                    "abstract": "When answering a question, humans utilize the information available across different modalities to synthesize a consistent and complete chain of thought (CoT). This process is normally a black box in the case of deep learning models like large-scale language models. Recently, science question benchmarks have been used to diagnose the multi-hop reasoning ability and interpretability of an AI system. However, existing datasets fail to provide annotations for the answers, or are restricted to the textual-only modality, small scales, and limited domain diversity. To this end, we present Science Question Answering (ScienceQA), a new benchmark that consists of ~21k multimodal multiple choice questions with a diverse set of science topics and annotations of their answers with corresponding lectures and explanations. We further design language models to learn to generate lectures and explanations as the chain of thought (CoT) to mimic the multi-hop reasoning process when answering ScienceQA questions. ScienceQA demonstrates the utility of CoT in language models, as CoT improves the question answering performance by 1.20% in few-shot GPT-3 and 3.99% in fine-tuned UnifiedQA. We also explore the upper bound for models to leverage explanations by feeding those in the input; we observe that it improves the few-shot performance of GPT-3 by 18.96%. Our analysis further shows that language models, similar to humans, benefit from explanations to learn from fewer data and achieve the same performance with just 40% of the data. The data and code are available at https://scienceqa.github.io."
                },
                {
                    "title": "Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?",
                    "abstract": "Large language models (LMs) are able to in-context learn\u2014perform a new task via inference alone by conditioning on a few input-label pairs (demonstrations) and making predictions for new inputs. However, there has been little understanding of how the model learns and which aspects of the demonstrations contribute to end task performance. In this paper, we show that ground truth demonstrations are in fact not required\u2014randomly replacing labels in the demonstrations barely hurts performance on a range of classification and multi-choce tasks, consistently over 12 different models including GPT-3. Instead, we find that other aspects of the demonstrations are the key drivers of endtask performance, including the fact that they provide a few examples of (1) the label space, (2) the distribution of the input text, and (3) the overall format of the sequence. Together, our analysis provides a new way of understanding how and why in-context learning works, while opening up new questions about how much can be learned from large language models through inference alone."
                },
                {
                    "title": "Can Machines Help Us Answering Question 16 in Datasheets, and In Turn Reflecting on Inappropriate Content?",
                    "abstract": "This paper contains images and descriptions that are offensive in nature. Large datasets underlying much of current machine learning raise serious issues concerning inappropriate content such as offensive, insulting, threatening, or might otherwise cause anxiety. This calls for increased dataset documentation, e.g., using datasheets. They, among other topics, encourage to reflect on the composition of the datasets. So far, this documentation, however, is done manually and therefore can be tedious and error-prone, especially for large image datasets. Here we ask the arguably \u201ccircular\u201d question of whether a machine can help us reflect on inappropriate content, answering Question 16 in Datasheets. To this end, we propose to use the information stored in pre-trained transformer models to assist us in the documentation process. Specifically, prompt-tuning based on a dataset of socio-moral values steers CLIP to identify potentially inappropriate content, therefore reducing human labor. We then document the inappropriate images found using word clouds, based on captions generated using a vision-language model. The documentations of two popular, large-scale computer vision datasets\u2014ImageNet and OpenImages\u2014produced this way suggest that machines can indeed help dataset creators to answer Question 16 on inappropriate image content."
                },
                {
                    "title": "BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation",
                    "abstract": "Vision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has been largely achieved by scaling up the dataset with noisy image-text pairs collected from the web, which is a suboptimal source of supervision. In this paper, we propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks. BLIP effectively utilizes the noisy web data by bootstrapping the captions, where a captioner generates synthetic captions and a filter removes the noisy ones. We achieve state-of-the-art results on a wide range of vision-language tasks, such as image-text retrieval (+2.7% in average recall@1), image captioning (+2.8% in CIDEr), and VQA (+1.6% in VQA score). BLIP also demonstrates strong generalization ability when directly transferred to video-language tasks in a zero-shot manner. Code, models, and datasets are released at https://github.com/salesforce/BLIP."
                },
                {
                    "title": "OCR-VQA: Visual Question Answering by Reading Text in Images",
                    "abstract": "The problem of answering questions about an image is popularly known as visual question answering (or VQA in short). It is a well-established problem in computer vision. However, none of the VQA methods currently utilize the text often present in the image. These \"texts in images\" provide additional useful cues and facilitate better understanding of the visual content. In this paper, we introduce a novel task of visual question answering by reading text in images, i.e., by optical character recognition or OCR. We refer to this problem as OCR-VQA. To facilitate a systematic way of studying this new problem, we introduce a large-scale dataset, namely OCRVQA-200K. This dataset comprises of 207,572 images of book covers and contains more than 1 million question-answer pairs about these images. We judiciously combine well-established techniques from OCR and VQA domains to present a novel baseline for OCR-VQA-200K. The experimental results and rigorous analysis demonstrate various challenges present in this dataset leaving ample scope for the future research. We are optimistic that this new task along with compiled dataset will open-up many exciting research avenues both for the document image analysis and the VQA communities."
                },
                {
                    "title": "OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge",
                    "abstract": "Visual Question Answering (VQA) in its ideal form lets us study reasoning in the joint space of vision and language and serves as a proxy for the AI task of scene understanding. However, most VQA benchmarks to date are focused on questions such as simple counting, visual attributes, and object detection that do not require reasoning or knowledge beyond what is in the image. In this paper, we address the task of knowledge-based visual question answering and provide a benchmark, called OK-VQA, where the image content is not sufficient to answer the questions, encouraging methods that rely on external knowledge resources. Our new dataset includes more than 14,000 questions that require external knowledge to answer. We show that the performance of the state-of-the-art VQA models degrades drastically in this new setting. Our analysis shows that our knowledge-based VQA task is diverse, difficult, and large compared to previous knowledge-based VQA datasets. We hope that this dataset enables researchers to open up new avenues for research in this domain."
                },
                {
                    "title": "Towards VQA Models That Can Read",
                    "abstract": "Studies have shown that a dominant class of questions asked by visually impaired users on images of their surroundings involves reading text in the image. But today\u2019s VQA models can not read! Our paper takes a first step towards addressing this problem. First, we introduce a new \u201cTextVQA\u201d dataset to facilitate progress on this important problem. Existing datasets either have a small proportion of questions about text (e.g., the VQA dataset) or are too small (e.g., the VizWiz dataset). TextVQA contains 45,336 questions on 28,408 images that require reasoning about text to answer. Second, we introduce a novel model architecture that reads text in the image, reasons about it in the context of the image and the question, and predicts an answer which might be a deduction based on the text and the image or composed of the strings found in the image. Consequently, we call our approach Look, Read, Reason & Answer (LoRRA). We show that LoRRA outperforms existing state-of-the-art VQA models on our TextVQA dataset. We find that the gap between human performance and machine performance is significantly larger on TextVQA than on VQA 2.0, suggesting that TextVQA is well-suited to benchmark progress along directions complementary to VQA 2.0."
                },
                {
                    "title": "The Socio-Moral Image Database (SMID): A novel stimulus set for the study of social, moral and affective processes",
                    "abstract": "A major obstacle for the design of rigorous, reproducible studies in moral psychology is the lack of suitable stimulus sets. Here, we present the Socio-Moral Image Database (SMID), the largest standardized moral stimulus set assembled to date, containing 2,941 freely available photographic images, representing a wide range of morally (and affectively) positive, negative and neutral content. The SMID was validated with over 820,525 individual judgments from 2,716 participants, with normative ratings currently available for all images on affective valence and arousal, moral wrongness, and relevance to each of the five moral values posited by Moral Foundations Theory. We present a thorough analysis of the SMID regarding (1) inter-rater consensus, (2) rating precision, and (3) breadth and variability of moral content. Additionally, we provide recommendations for use aimed at efficient study design and reproducibility, and outline planned extensions to the database. We anticipate that the SMID will serve as a useful resource for psychological, neuroscientific and computational (e.g., natural language processing or computer vision) investigations of social, moral and affective processes. The SMID images, along with associated normative data and additional resources are available at https://osf.io/2rqad/."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "You Only Look Once: Unified, Real-Time Object Detection",
                    "abstract": "We present YOLO, a new approach to object detection. Prior work on object detection repurposes classifiers to perform detection. Instead, we frame object detection as a regression problem to spatially separated bounding boxes and associated class probabilities. A single neural network predicts bounding boxes and class probabilities directly from full images in one evaluation. Since the whole detection pipeline is a single network, it can be optimized end-to-end directly on detection performance. Our unified architecture is extremely fast. Our base YOLO model processes images in real-time at 45 frames per second. A smaller version of the network, Fast YOLO, processes an astounding 155 frames per second while still achieving double the mAP of other real-time detectors. Compared to state-of-the-art detection systems, YOLO makes more localization errors but is less likely to predict false positives on background. Finally, YOLO learns very general representations of objects. It outperforms other detection methods, including DPM and R-CNN, when generalizing from natural images to other domains like artwork."
                },
                {
                    "title": "AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation Framework",
                    "abstract": "This technical report presents AutoGen , 1 a new framework that enables development of LLM applications using multiple agents that can converse with each other to solve tasks. AutoGen agents are customizable, conversable , and seamlessly allow human participation. They can operate in various modes that employ combinations of LLMs, human inputs, and tools. AutoGen \u2019s design offers multiple advantages: a) it gracefully navigates the strong but imperfect generation and reasoning abilities of these LLMs; b) it leverages human understanding and intelligence, while providing valuable automation through conversations between agents; c) it simplifies and unifies the implementation of complex LLM workflows as automated agent chats. We provide many diverse examples of how developers can easily use AutoGen to effectively solve tasks or build applications, ranging from coding, mathematics, operations research, entertainment, online decision-making, question answering, etc."
                },
                {
                    "title": "Evaluating Interfaced LLM Bias",
                    "abstract": "In this research, we comprehensively analyze the potential biases inherent in Large Language Model, utilizing meticulously curated input data to ascertain the extent to which such data sway machine-generated responses to yield prejudiced outcomes. Notwithstanding recent strides in mitigating bias in LLM-based NLP, our \ufb01ndings underscore the continued susceptibility of these models to data-driven bias. We have integrated the PTT NTU board as our primary data source for this investigation. Moreover, our study elucidates that, in certain contexts, machines may manifest biases without supplementary prompts. However, they can be guided toward rendering impartial responses when provided with enhanced contextual nuances."
                },
                {
                    "title": "Aligning Large Multi-Modal Model with Robust Instruction Tuning",
                    "abstract": "Despite the promising progress in multi-modal tasks, current large multi-modal models (LMM) are prone to hallucinating inconsistent descriptions with respect to the associated image and human instructions. This paper addresses this issue by introducing the first large and diverse visual instruction tuning dataset, named Large-scale Robust Visual (LRV)-Instruction . Our dataset consists of 120k visual instructions generated by GPT4, covering 16 vision-and-language tasks with open-ended instructions and answers. Unlike existing studies that primarily focus on positive instruction samples, we design LRV-Instruction to include both positive and negative instructions for more robust visual instruction tuning. Our negative instructions are designed at two semantic levels: (i) Nonexistent Element Manipulation and (ii) Existent Element Manipulation . To efficiently measure the hallucination generated by LMMs, we propose GPT4-Assisted Visual Instruction Evaluation (GAVIE) , a novel approach to evaluate visual instruction tuning without the need for human-annotated groundtruth answers and can adapt to diverse instruction formats. We conduct comprehensive experiments to investigate the hallucination of LMMs. Our results demonstrate that existing LMMs exhibit significant hallucination when presented with our negative instructions, particularly with Existent Element Manipulation instructions. Moreover, by finetuning MiniGPT4 on LRV-Instruction , we successfully mitigate hallucination while improving performance on public datasets using less training data compared to state-of-the-art methods. Additionally, we observed that a balanced ratio of positive and negative instances in the training data leads to a more robust model. Our project link is available at this link."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively recognize and categorize harmful visual content, including both real and synthesized materials, to protect underage children from exposure to inappropriate materials?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing concern of harmful content exposure, particularly for vulnerable populations like children. By developing a comprehensive dataset and robust recognition methods, this research could lead to significant advancements in content moderation technologies, enhancing online safety. Furthermore, it could inspire future research into multimodal content analysis and the ethical implications of generative models, ultimately leading to practical applications in content filtering, parental controls, and automated moderation systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the complexity of defining \"harmful\" content, which can vary based on context and cultural perspectives. Naive approaches may fail due to the ambiguity in harmfulness, leading to false positives or negatives. Additionally, existing datasets are limited in scope, often focusing on specific harmful objects without considering the broader context of images or videos. The integration of real and synthesized content adds another layer of complexity, requiring advanced techniques to ensure accurate recognition across diverse scenarios.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been hindered by the lack of comprehensive datasets that encompass a wide range of harmful categories and contexts. Many existing datasets focus narrowly on specific harmful objects, neglecting the importance of context and the inclusion of synthesized content. Barriers such as the complexity of annotation processes and the need for advanced multimodal understanding have also contributed to the lack of effective solutions. Our approach differs by employing a novel \"debate\" annotation framework using pretrained vision-language models, which allows for a more nuanced understanding of harmfulness by considering multiple perspectives.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves creating the Visual Harmful Dataset 11K (VHD11K), which includes 10,000 images and 1,000 videos across 10 categories of harmful content. We will utilize a multi-agent Visual Question Answering (VQA) task for annotation, employing three different vision-language models in a debating framework to assess harmfulness based on contextual understanding. The expected outcomes include a comprehensive dataset that enhances the reliability of harmful content recognition and a set of distilled harmful categories derived from extensive annotations, which will serve as"
            }
        },
        "author_data": {
            "a20269f1-672b-4406-ba55-fe06bc358487": {
                "pk": "a20269f1-672b-4406-ba55-fe06bc358487",
                "name": "Chen Yeh",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "807626b5-9c17-4d26-9d59-919ed8d02507": {
                "pk": "807626b5-9c17-4d26-9d59-919ed8d02507",
                "name": "You-Ming Chang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "9c466825-5f15-4d8a-8e52-ea0e652e4cbd": {
                "pk": "9c466825-5f15-4d8a-8e52-ea0e652e4cbd",
                "name": "Wei-Chen Chiu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "2e01a2b4-c4ea-4591-8793-ad0086187096": {
                "pk": "2e01a2b4-c4ea-4591-8793-ad0086187096",
                "name": "Ning Yu",
                "collaborators": [
                    "Hongge Xu",
                    "Zuman Zhang",
                    "Timothy Haskins",
                    "Xiaofeng Luo",
                    "Guoying Chen"
                ],
                "domain": [
                    "Machine Learning",
                    "High Energy Physics",
                    "Statistical Physics",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "KNN, An Underestimated Model for Regional Rainfall Forecasting",
                        "abstract": "Regional rainfall forecasting is an important issue in hydrology and meteorology. This paper aims to design an integrated tool by applying various machine learning algorithms, especially the state-of-the-art deep learning algorithms including Deep Neural Network, Wide Neural Network, Deep and Wide Neural Network, Reservoir Computing, Long Short Term Memory, Support Vector Machine, K-Nearest Neighbor for forecasting regional precipitations over different catchments in Upstate New York. Through the experimental results and the comparison among machine learning models including classification and regression, we find that KNN is an outstanding model over other models to handle the uncertainty in the precipitation data. The data normalization methods such as ZScore and MinMax are also evaluated and discussed."
                    },
                    {
                        "title": "Beam energy dependence of $d$ and $\\bar{d}$ productions in Au+Au collisions at RHIC",
                        "abstract": "Light nuclei have much smaller binding energy compared to the temperature of the system created in heavy-ion collisions. Consequently, the distributions of light nuclei can be used to probe the freeze-out properties, such as correlation volume and local baryon density of the medium created in high-energy nuclear collisions. In this paper, we report the results of deuteron and anti-deuteron production in Au+Au collision at $\\sqrt{s_\\mathrm{NN}}$ = 7.7, 11.5, 14.5, 19.6, 27, 39, 62.4, and 200 GeV, measured by STAR at RHIC. The collision energy, centrality and transverse momentum dependence of the coalescence parameter $B_2$ for deuteron and anti-deuteron production are discussed. We find the values of $B_2$ for anti-deuteron are systematically lower than those for deuterons indicating the correlation volume of anti-baryon are larger than that of baryon. In addition, the values of $B_2$ are found to decrease with increasing collision energy and reach a minimum around $\\sqrt{s_\\mathrm{NN}}$ = 20 GeV implying a change of the equation of state of the medium in these collisions."
                    },
                    {
                        "title": "Particle decay from statistical thermal model in high energy nucleus-nucleus collision",
                        "abstract": "In high energy nucleus-nucleus collisions, it is difficult to measure the contributions of resonance strong decay and weak decay to the final measured hadrons as well as the corresponding effects on some physical observables. To provide a reference from statistical thermal model, we performed a systematic analysis for the energy dependence of particle yield and yield ratios in Au + Au collisions. We found that the primary fraction of final hadrons decreases with increasing collision energy and somehow saturates around $\\sqrt{s_\\textrm{NN}}$ = 10 GeV, indicating a limiting temperature in hadronic interactions. The fraction of strong or weak decay for final hadrons show a different energy dependence behavior comparing to the primarily produced hadrons. These energy dependences of various particle yield and yield ratios from strong or weak decay can provide us with baselines for many hadronic observables in high energy nucleus-nucleus collisions."
                    },
                    {
                        "title": "Study the structure of X(3872) from its lineshape",
                        "abstract": "We fit the invariant mass distribution of ${X(3872)}\\rightarrow{J}/{\\psi}\\pi^+\\pi^-$ from LHCb and ${X(3872)}\\rightarrow{D}^{0}\\overline{D}^{0*}$ from Belle using the propagator for S-wave threshold states in effective field theory. In this way, we can directly determine the $Z$ which is the probability of finding the compact component in ${X(3872)}$."
                    },
                    {
                        "title": "The general propagator for S-wave threshold states",
                        "abstract": "We demonstrate that the propagator, derived from an Effective Field Theory (EFT) that incorporates Weinberger's compositeness theorem, provides a more general formula for describing S-wave near-threshold states. By fitting the lineshape using this propagator, we can extract the $Z$ factor for these states and elucidate their structures."
                    }
                ]
            }
        }
    },
    "2309.00359": {
        "paper_data": {
            "title": "Large Content And Behavior Models To Understand, Simulate, And Optimize Content And Behavior",
            "url": "http://arxiv.org/abs/2309.00359v4",
            "arxiv_id": "2309.00359",
            "authors": [
                "Ashmit Khandelwal",
                "Aditya Agrawal",
                "Aanisha Bhattacharyya",
                "Yaman K Singla",
                "Somesh Singh",
                "Uttaran Bhattacharya",
                "Ishita Dasgupta",
                "Stefano Petrangeli",
                "Rajiv Ratn Shah",
                "Changyou Chen",
                "Balaji Krishnamurthy"
            ],
            "abstract": "Shannon and Weaver's seminal information theory divides communication into three levels: technical, semantic, and effectiveness. While the technical level deals with the accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Large Language Models (LLMs), with their wide generalizability, make some progress towards the second level. However, LLMs and other communication models are not conventionally designed for predicting and optimizing communication for desired receiver behaviors and intents. As a result, the effectiveness level remains largely untouched by modern communication systems. In this paper, we introduce the receivers' \"behavior tokens,\" such as shares, likes, clicks, purchases, and retweets, in the LLM's training corpora to optimize content for the receivers and predict their behaviors. Other than showing similar performance to LLMs on content understanding tasks, our trained models show generalization capabilities on the behavior dimension for behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. We show results on all these capabilities using a wide range of tasks on three corpora. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior (https://behavior-in-the-wild.github.io/LCBM).",
            "introduction": "   1 Introduction  Figure 1: Encoding and predicting content (images, videos, and text) and behavior in the language space. Large Content Behavior Models (LCBMs), once trained, can enable a host of different applications, including behavior simulation, content understanding, content-behavior optimization, and content-behavior understanding.              (a)      (b)       (c)      (d)     Figure 2:   Comparison of GPT-3.5, GPT-4, Vicuna-13B, and LCBM-13B on: (a)\u00a0Behavior Simulation accuracy on two types of behaviors: replay value prediction and likes/views prediction. The task is, given the video content and channel information, to predict replay values corresponding to each scene and the ratio of likes to views. (b)\u00a0Content simulation and behavior understanding tasks. The task for content simulation is, given the channel information and scene-level behavior, to predict the scene content. Given information on the video platform and the video content, the task of behavior understanding is to predict and explain the sentiments of the viewers and the commenters. Six evaluators scored the models\u2019 explanations between 0-5 to get the predicted sentiment and explanation scores by comparing the ratings and reasons with the user comments. The annotators did not know which model gave the reasoning. (c)\u00a0Content understanding tasks. We evaluate four tasks: emotion, topic, and persuasion strategy prediction, and action-and-reason understanding. (d)\u00a0Behavior Simulation on in-house Email Marketing Data (R2superscript\ud835\udc452R^{2}italic_R start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT score) and Twitter likes (accuracy), and Content Simulation on Twitter tweet prediction (BLEU/ROUGE scores). It can be noted that on the behavior simulation, content simulation, and behavior understanding tasks, LCBM performs better than 3-shot GPT-3.5 and 10-shot GPT-4 (covering a larger area. On the content understanding tasks, while LCBM outperforms similar-sized Vicuna models, GPT-3.5 performs better. However, we also note that GPT-3.5 and GPT-4 are at least 12 times larger than LCBM-13B. Further, we show the behavior domain adaptation results in Table\u00a04, \u00a010, \u00a010.    Shannon & Weaver (1949), in their seminal paper on communication, includes all of the procedures by which one mind may affect another. This includes all forms of expression, such as words, gestures, speech, pictures, and musical sounds. They mentioned that the broad problem of communication can be studied at three levels: technical, semantic, and effectiveness.   Level A: Technical. How accurately can the symbols of communication be transmitted?   Level B: Semantic. How precisely do the transmitted symbols convey the desired meaning?   Level C: Effectiveness. How well does the received meaning induce the desired conduct in the receiver?   These three levels build on top of each other. Thus, solving the problem at Level C necessarily requires solving the corresponding problems at Levels A and B.   Since the publication of this seminal paper, the tremendous growth in the field of telecommunications, particularly the advent of the Internet and mobile devices, has led to affordable, wide-scale solutions for Level A. With the recent advances in large language models (LLMs) such as BERT (Devlin et\u00a0al., 2018), GPT-3 and 4 (Brown et\u00a0al., 2020; OpenAI, 2023), T5 (Raffel et\u00a0al., 2020), and many more, we have witnessed a significant improvement in the performance of various Natural Language Processing (NLP) tasks. LLMs in zero- or few-shot settings can easily handle tasks such as question answering, summarization, translation, and many more. This has helped us progress towards solving Level B",
            "references": [
                {
                    "title": "Behavior Optimized Image Generation",
                    "abstract": "The last few years have witnessed great success on image generation, which has crossed the acceptance thresholds of aesthetics, making it directly applicable to personal and commercial applications. However, images, especially in marketing and advertising applications, are often created as a means to an end as opposed to just aesthetic concerns. The goal can be increasing sales, getting more clicks, likes, or image sales (in the case of stock businesses). Therefore, the generated images need to perform well on these key performance indicators (KPIs), in addition to being aesthetically good. In this paper, we make the first endeavor to answer the question of\"How can one infuse the knowledge of the end-goal within the image generation process itself to create not just better-looking images but also\"better-performing'' images?''. We propose BoigLLM, an LLM that understands both image content and user behavior. BoigLLM knows how an image should look to get a certain required KPI. We show that BoigLLM outperforms 13x larger models such as GPT-3.5 and GPT-4 in this task, demonstrating that while these state-of-the-art models can understand images, they lack information on how these images perform in the real world. To generate actual pixels of behavior-conditioned images, we train a diffusion-based model (BoigSD) to align with a proposed BoigLLM-defined reward. We show the performance of the overall pipeline on two datasets covering two different behaviors: a stock dataset with the number of forward actions as the KPI and a dataset containing tweets with the total likes as the KPI, denoted as BoigBench. To advance research in the direction of utility-driven image generation and understanding, we release BoigBench, a benchmark dataset containing 168 million enterprise tweets with their media, brand account names, time of post, and total likes."
                },
                {
                    "title": "Planting a SEED of Vision in Large Language Model",
                    "abstract": "We present SEED, an elaborate image tokenizer that empowers Large Language Models (LLMs) with the emergent ability to SEE and Draw at the same time. Research on image tokenizers has previously reached an impasse, as frameworks employing quantized visual tokens have lost prominence due to subpar performance and convergence in multimodal comprehension (compared to BLIP-2, etc.) or generation (compared to Stable Diffusion, etc.). Despite the limitations, we remain confident in its natural capacity to unify visual and textual representations, facilitating scalable multimodal training with LLM's original recipe. In this study, we identify two crucial principles for the architecture and training of SEED that effectively ease subsequent alignment with LLMs. (1) Image tokens should be independent of 2D physical patch positions and instead be produced with a 1D causal dependency, exhibiting intrinsic interdependence that aligns with the left-to-right autoregressive prediction mechanism in LLMs. (2) Image tokens should capture high-level semantics consistent with the degree of semantic abstraction in words, and be optimized for both discriminativeness and reconstruction during the tokenizer training phase. As a result, the off-the-shelf LLM is able to perform both image-to-text and text-to-image generation by incorporating our SEED through efficient LoRA tuning. Comprehensive multimodal pretraining and instruction tuning, which may yield improved results, are reserved for future investigation. This version of SEED was trained in 5.7 days using only 64 V100 GPUs and 5M publicly available image-text pairs. Our preliminary study emphasizes the great potential of discrete visual tokens in versatile multimodal LLMs and the importance of proper image tokenizers in broader research."
                },
                {
                    "title": "Video-LLaMA: An Instruction-tuned Audio-Visual Language Model for Video Understanding",
                    "abstract": "We present Video-LLaMA a multi-modal framework that empowers Large Language Models (LLMs) with the capability of understanding both visual and auditory content in the video. Video-LLaMA bootstraps cross-modal training from the frozen pre-trained visual and audio encoders and the frozen LLMs. Unlike previous works that complement LLMs to process the visual or audio signals only, Video-LLaMA enables video comprehension by tackling two challenges: (1) capturing the temporal changes in visual scenes, (2) integrating audio-visual signals. To counter the first challenge, we propose a Video Q-former to assemble a pre-trained image encoder into our video encoder and introduce a video-to-text generation task to learn video-language correspondence. For the second challenge, we leverage ImageBind, a universal embedding model aligning multiple modalities, as the pre-trained audio encoder and introduce an Audio Q-former on top of ImageBind to learn reasonable auditory query embeddings for the LLM module. To align the output of both visual and audio encoders with LLM's embedding space, we first train Video-LLaMA on massive video/image-caption pairs and then tune our model with visual-instruction datasets of moderate amount but higher quality. We found Video-LLaMA shows the ability to perceive and comprehend video content and generate meaningful responses grounded in the visual and auditory information presented in the videos."
                },
                {
                    "title": "The RefinedWeb Dataset for Falcon LLM: Outperforming Curated Corpora with Web Data, and Web Data Only",
                    "abstract": "Large language models are commonly trained on a mixture of filtered web data and curated high-quality corpora, such as social media conversations, books, or technical papers. This curation process is believed to be necessary to produce performant models with broad zero-shot generalization abilities. However, as larger models requiring pretraining on trillions of tokens are considered, it is unclear how scalable is curation and whether we will run out of unique high-quality data soon. At variance with previous beliefs, we show that properly filtered and deduplicated web data alone can lead to powerful models; even significantly outperforming models from the state-of-the-art trained on The Pile. Despite extensive filtering, the high-quality data we extract from the web is still plentiful, and we are able to obtain five trillion tokens from CommonCrawl. We publicly release an extract of 600 billion tokens from our RefinedWeb dataset, and 1.3/7.5B parameters language models trained on it."
                },
                {
                    "title": "A Video Is Worth 4096 Tokens: Verbalize Videos To Understand Them In Zero Shot",
                    "abstract": "Multimedia content, such as advertisements and story videos, exhibit a rich blend of creativity and multiple modalities. They incorporate elements like text, visuals, audio, and storytelling techniques, employing devices like emotions, symbolism, and slogans to convey meaning. There is a dearth of large annotated training datasets in the multimedia domain hindering the development of supervised learning models with satisfactory performance for real-world applications. On the other hand, the rise of large language models (LLMs) has witnessed remarkable zero-shot performance in various natural language processing (NLP) tasks, such as emotion classification, question-answering, and topic classification. To leverage such advanced techniques to bridge this performance gap in multimedia understanding, we propose verbalizing long videos to generate their descriptions in natural language, followed by performing video-understanding tasks on the generated story as opposed to the original video. Through extensive experiments on fifteen video-understanding tasks, we demonstrate that our method, despite being zero-shot, achieves significantly better results than supervised baselines for video understanding. Furthermore, to alleviate a lack of story understanding benchmarks, we publicly release the first dataset on a crucial task in computational social science on persuasion strategy identification."
                },
                {
                    "title": "Do LLMs Understand User Preferences? Evaluating LLMs On User Rating Prediction",
                    "abstract": "Large Language Models (LLMs) have demonstrated exceptional capabilities in generalizing to new tasks in a zero-shot or few-shot manner. However, the extent to which LLMs can comprehend user preferences based on their previous behavior remains an emerging and still unclear research question. Traditionally, Collaborative Filtering (CF) has been the most effective method for these tasks, predominantly relying on the extensive volume of rating data. In contrast, LLMs typically demand considerably less data while maintaining an exhaustive world knowledge about each item, such as movies or products. In this paper, we conduct a thorough examination of both CF and LLMs within the classic task of user rating prediction, which involves predicting a user's rating for a candidate item based on their past ratings. We investigate various LLMs in different sizes, ranging from 250M to 540B parameters and evaluate their performance in zero-shot, few-shot, and fine-tuning scenarios. We conduct comprehensive analysis to compare between LLMs and strong CF methods, and find that zero-shot LLMs lag behind traditional recommender models that have the access to user interaction data, indicating the importance of user interaction data. However, through fine-tuning, LLMs achieve comparable or even better performance with only a small fraction of the training data, demonstrating their potential through data efficiency."
                },
                {
                    "title": "VideoChat: Chat-Centric Video Understanding",
                    "abstract": "In this paper, we initiate an attempt of developing an end-to-end chat-centric video understanding system, coined as VideoChat. It integrates video foundation models and large language models via a learnable neural interface, excelling in spatiotemporal reasoning, event localization, and causal relationship inference. To instructively tune this system, we build a video-centric instruction dataset, composed of thousands of videos associated with detailed descriptions and conversations. This dataset emphasizes spatiotemporal reasoning and captures causal relationships, providing a valuable asset for training our chat-centric video understanding system. Preliminary qualitative experiments demonstrate the potential of our system across a broad spectrum of video applications, which could serve as a simple prototype system for future research on chat-centric video understanding. Access our code and data at https://github.com/OpenGVLab/Ask-Anything"
                },
                {
                    "title": "MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models",
                    "abstract": "The recent GPT-4 has demonstrated extraordinary multi-modal abilities, such as directly generating websites from handwritten text and identifying humorous elements within images. These features are rarely observed in previous vision-language models. However, the technical details behind GPT-4 continue to remain undisclosed. We believe that the enhanced multi-modal generation capabilities of GPT-4 stem from the utilization of sophisticated large language models (LLM). To examine this phenomenon, we present MiniGPT-4, which aligns a frozen visual encoder with a frozen advanced LLM, Vicuna, using one projection layer. Our work, for the first time, uncovers that properly aligning the visual features with an advanced large language model can possess numerous advanced multi-modal abilities demonstrated by GPT-4, such as detailed image description generation and website creation from hand-drawn drafts. Furthermore, we also observe other emerging capabilities in MiniGPT-4, including writing stories and poems inspired by given images, teaching users how to cook based on food photos, and so on. In our experiment, we found that the model trained on short image caption pairs could produce unnatural language outputs (e.g., repetition and fragmentation). To address this problem, we curate a detailed image description dataset in the second stage to finetune the model, which consequently improves the model's generation reliability and overall usability. Our code, pre-trained model, and collected dataset are available at https://minigpt-4.github.io/."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "Generative Agents: Interactive Simulacra of Human Behavior",
                    "abstract": "Believable proxies of human behavior can empower interactive applications ranging from immersive environments to rehearsal spaces for interpersonal communication to prototyping tools. In this paper, we introduce generative agents: computational software agents that simulate believable human behavior. Generative agents wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations; they remember and reflect on days past as they plan the next day. To enable generative agents, we describe an architecture that extends a large language model to store a complete record of the agent\u2019s experiences using natural language, synthesize those memories over time into higher-level reflections, and retrieve them dynamically to plan behavior. We instantiate generative agents to populate an interactive sandbox environment inspired by The Sims, where end users can interact with a small town of twenty-five agents using natural language. In an evaluation, these generative agents produce believable individual and emergent social behaviors. For example, starting with only a single user-specified notion that one agent wants to throw a Valentine\u2019s Day party, the agents autonomously spread invitations to the party over the next two days, make new acquaintances, ask each other out on dates to the party, and coordinate to show up for the party together at the right time. We demonstrate through ablation that the components of our agent architecture\u2014observation, planning, and reflection\u2014each contribute critically to the believability of agent behavior. By fusing large language models with computational interactive agents, this work introduces architectural and interaction patterns for enabling believable simulations of human behavior."
                },
                {
                    "title": "Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling",
                    "abstract": "How do large language models (LLMs) develop and evolve over the course of training? How do these patterns change as models scale? To answer these questions, we introduce \\textit{Pythia}, a suite of 16 LLMs all trained on public data seen in the exact same order and ranging in size from 70M to 12B parameters. We provide public access to 154 checkpoints for each one of the 16 models, alongside tools to download and reconstruct their exact training dataloaders for further study. We intend \\textit{Pythia} to facilitate research in many areas, and we present several case studies including novel results in memorization, term frequency effects on few-shot performance, and reducing gender bias. We demonstrate that this highly controlled setup can be used to yield novel insights toward LLMs and their training dynamics. Trained models, analysis code, training code, and training data can be found at \\url{https://github.com/EleutherAI/pythia}."
                },
                {
                    "title": "EVA-CLIP: Improved Training Techniques for CLIP at Scale",
                    "abstract": "Contrastive language-image pre-training, CLIP for short, has gained increasing attention for its potential in various scenarios. In this paper, we propose EVA-CLIP, a series of models that significantly improve the efficiency and effectiveness of CLIP training. Our approach incorporates new techniques for representation learning, optimization, and augmentation, enabling EVA-CLIP to achieve superior performance compared to previous CLIP models with the same number of parameters but significantly smaller training costs. Notably, our largest 5.0B-parameter EVA-02-CLIP-E/14+ with only 9 billion seen samples achieves 82.0 zero-shot top-1 accuracy on ImageNet-1K val. A smaller EVA-02-CLIP-L/14+ with only 430 million parameters and 6 billion seen samples achieves 80.4 zero-shot top-1 accuracy on ImageNet-1K val. To facilitate open access and open research, we release the complete suite of EVA-CLIP to the community at https://github.com/baaivision/EVA/tree/master/EVA-CLIP."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Synthesizing Human Gaze Feedback for Improved NLP Performance",
                    "abstract": "Integrating human feedback in models can improve the performance of natural language processing (NLP) models. Feedback can be either explicit (e.g. ranking used in training language models) or implicit (e.g. using human cognitive signals in the form of eyetracking). Prior eye tracking and NLP research reveal that cognitive processes, such as human scanpaths, gleaned from human gaze patterns aid in the understanding and performance of NLP models. However, the collection of real eyetracking data for NLP tasks is challenging due to the requirement of expensive and precise equipment coupled with privacy invasion issues. To address this challenge, we propose ScanTextGAN, a novel model for generating human scanpaths over text. We show that ScanTextGAN-generated scanpaths can approximate meaningful cognitive signals in human gaze patterns. We include synthetically generated scanpaths in four popular NLP tasks spanning six different datasets as proof of concept and show that the models augmented with generated scanpaths improve the performance of all downstream NLP tasks."
                },
                {
                    "title": "Large Language Models are Versatile Decomposers: Decomposing Evidence and Questions for Table-based Reasoning",
                    "abstract": "Table-based reasoning has shown remarkable progress in a wide range of table-based tasks. It is a challenging task, which requires reasoning over both free-form natural language (NL) questions and (semi-)structured tabular data. However, previous table-based reasoning solutions usually suffer from significant performance degradation on ''huge'' evidence (tables). In addition, most existing methods struggle to reason over complex questions since the essential information is scattered in different places. To alleviate the above challenges, we exploit large language models (LLMs) as decomposers for effective table-based reasoning, which (i) decompose huge evidence (a huge table) into sub-evidence (a small table) to mitigate the interference of useless information for table reasoning, and (ii) decompose a complex question into simpler sub-questions for text reasoning. First, we use a powerful LLM to decompose the evidence involved in the current question into the sub-evidence that retains the relevant information and excludes the remaining irrelevant information from the ''huge'' evidence. Second, we propose a novel ''parsing-execution-filling'' strategy to decompose a complex question into simper step-by-step sub-questions by generating intermediate SQL queries as a bridge to produce numerical and logical sub-questions with a powerful LLM. Finally, we leverage the decomposed sub-evidence and sub-questions to get the final answer with a few in-context prompting examples. Extensive experiments on three benchmark datasets (TabFact, WikiTableQuestion, and FetaQA) demonstrate that our method achieves significantly better results than competitive baselines for table-based reasoning. Notably, our method outperforms human performance for the first time on the TabFact dataset. In addition to impressive overall performance, our method also has the advantage of interpretability, where the returned results are to some extent tractable with the generated sub-evidence and sub-questions. For reproducibility, we release our source code and data at: https://github.com/AlibabaResearch/DAMO-ConvAI."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "Robust Speech Recognition via Large-Scale Weak Supervision",
                    "abstract": "We study the capabilities of speech processing systems trained simply to predict large amounts of transcripts of audio on the internet. When scaled to 680,000 hours of multilingual and multitask supervision, the resulting models generalize well to standard benchmarks and are often competitive with prior fully supervised results but in a zero-shot transfer setting without the need for any fine-tuning. When compared to humans, the models approach their accuracy and robustness. We are releasing models and inference code to serve as a foundation for further work on robust speech processing."
                },
                {
                    "title": "Hierarchical Self-supervised Representation Learning for Movie Understanding",
                    "abstract": "Most self-supervised video representation learning approaches focus on action recognition. In contrast, in this paper we focus on self-supervised video learning for movie understanding and propose a novel hierarchical self-supervised pretraining strategy that separately pretrains each level of our hierarchical movie understanding model (based on [37]). Specifically, we propose to pretrain the low-level video backbone using a contrastive learning objective, while pretrain the higher-level video contextualizer using an event mask prediction task, which enables the usage of different data sources for pretraining different levels of the hierarchy. We first show that our self-supervised pre-training strategies are effective and lead to improved performance on all tasks and metrics on VidSitu benchmark [37] (e.g., improving on semantic role prediction from 47% to 61% CIDEr scores). We further demonstrate the effectiveness of our contextualized event features on LVU tasks [54], both when used alone and when combined with instance features, showing their complementarity."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Datasheet for the Pile",
                    "abstract": "This datasheet describes the Pile, a 825 GiB dataset of human-authored text compiled by EleutherAI for use in large-scale language modeling. The Pile is comprised of 22 different text sources, ranging from original scrapes done for this project, to text data made available by the data owners, to third-party scrapes available online."
                },
                {
                    "title": "UniFormer: Unified Transformer for Efficient Spatiotemporal Representation Learning",
                    "abstract": "It is a challenging task to learn rich and multi-scale spatiotemporal semantics from high-dimensional videos, due to large local redundancy and complex global dependency between video frames. The recent advances in this research have been mainly driven by 3D convolutional neural networks and vision transformers. Although 3D convolution can efficiently aggregate local context to suppress local redundancy from a small 3D neighborhood, it lacks the capability to capture global dependency because of the limited receptive field. Alternatively, vision transformers can effectively capture long-range dependency by self-attention mechanism, while having the limitation on reducing local redundancy with blind similarity comparison among all the tokens in each layer. Based on these observations, we propose a novel Unified transFormer (UniFormer) which seamlessly integrates merits of 3D convolution and spatiotemporal self-attention in a concise transformer format, and achieves a preferable balance between computation and accuracy. Different from traditional transformers, our relation aggregator can tackle both spatiotemporal redundancy and dependency, by learning local and global token affinity respectively in shallow and deep layers. We conduct extensive experiments on the popular video benchmarks, e.g., Kinetics-400, Kinetics-600, and Something-Something V1&V2. With only ImageNet-1K pretraining, our UniFormer achieves 82.9%/84.8% top-1 accuracy on Kinetics-400/Kinetics-600, while requiring 10x fewer GFLOPs than other state-of-the-art methods. For Something-Something V1 and V2, our UniFormer achieves new state-of-the-art performances of 60.9% and 71.2% top-1 accuracy respectively. Code is available at https://github.com/Sense-X/UniFormer."
                },
                {
                    "title": "Decision Transformer: Reinforcement Learning via Sequence Modeling",
                    "abstract": "We introduce a framework that abstracts Reinforcement Learning (RL) as a sequence modeling problem. This allows us to draw upon the simplicity and scalability of the Transformer architecture, and associated advances in language modeling such as GPT-x and BERT. In particular, we present Decision Transformer, an architecture that casts the problem of RL as conditional sequence modeling. Unlike prior approaches to RL that fit value functions or compute policy gradients, Decision Transformer simply outputs the optimal actions by leveraging a causally masked Transformer. By conditioning an autoregressive model on the desired return (reward), past states, and actions, our Decision Transformer model can generate future actions that achieve the desired return. Despite its simplicity, Decision Transformer matches or exceeds the performance of state-of-the-art model-free offline RL baselines on Atari, OpenAI Gym, and Key-to-Door tasks."
                },
                {
                    "title": "Towards Long-Form Video Understanding",
                    "abstract": "Our world offers a never-ending stream of visual stimuli, yet today\u2019s vision systems only accurately recognize patterns within a few seconds. These systems understand the present, but fail to contextualize it in past or future events. In this paper, we study long-form video understanding. We introduce a framework for modeling long-form videos and develop evaluation protocols on large-scale datasets. We show that existing state-of-the-art short-term models are limited for long-form tasks. A novel object-centric transformer-based video recognition architecture performs significantly better on 7 diverse tasks. It also outperforms comparable state-of-the-art on the AVA dataset."
                },
                {
                    "title": "Frozen in Time: A Joint Video and Image Encoder for End-to-End Retrieval",
                    "abstract": "Our objective in this work is video-text retrieval \u2013 in particular a joint embedding that enables efficient text-to-video retrieval. The challenges in this area include the design of the visual architecture and the nature of the training data, in that the available large scale video-text training datasets, such as HowTo100M, are noisy and hence competitive performance is achieved only at scale through large amounts of compute.We address both these challenges in this paper. We propose an end-to-end trainable model that is designed to take advantage of both large-scale image and video captioning datasets. Our model is an adaptation and extension of the recent ViT and Timesformer architectures, and consists of attention in both space and time. The model is flexible and can be trained on both image and video text datasets, either independently or in conjunction. It is trained with a curriculum learning schedule that begins by treating images as \u2018frozen\u2019 snapshots of video, and then gradually learns to attend to increasing temporal context when trained on video datasets. We also provide a new video-text pretraining dataset WebVid-2M, comprised of over two million videos with weak captions scraped from the internet. Despite training on datasets that are an order of magnitude smaller, we show that this approach yields state-of-the-art results on standard downstream video-retrieval benchmarks including MSR-VTT, MSVD, DiDeMo and LSMDC."
                },
                {
                    "title": "Conceptual 12M: Pushing Web-Scale Image-Text Pre-Training To Recognize Long-Tail Visual Concepts",
                    "abstract": "The availability of large-scale image captioning and visual question answering datasets has contributed significantly to recent successes in vision-and-language pretraining. However, these datasets are often collected with overrestrictive requirements inherited from their original target tasks (e.g., image caption generation), which limit the resulting dataset scale and diversity. We take a step further in pushing the limits of vision-and-language pretraining data by relaxing the data collection pipeline used in Conceptual Captions 3M (CC3M) [54] and introduce the Conceptual 12M (CC12M), a dataset with 12 million image-text pairs specifically meant to be used for visionand-language pre-training. We perform an analysis of this dataset and benchmark its effectiveness against CC3M on multiple downstream tasks with an emphasis on long-tail visual recognition. Our results clearly illustrate the benefit of scaling up pre-training data for vision-and-language tasks, as indicated by the new state-of-the-art results on both the nocaps and Conceptual Captions benchmarks.1"
                },
                {
                    "title": "Spatiotemporal Contrastive Video Representation Learning",
                    "abstract": "We present a self-supervised Contrastive Video Representation Learning (CVRL) method to learn spatiotemporal visual representations from unlabeled videos. Our representations are learned using a contrastive loss, where two augmented clips from the same short video are pulled together in the embedding space, while clips from different videos are pushed away. We study what makes for good data augmentations for video self-supervised learning and find that both spatial and temporal information are crucial. We carefully design data augmentations involving spatial and temporal cues. Concretely, we propose a temporally consistent spatial augmentation method to impose strong spatial augmentations on each frame of the video while maintaining the temporal consistency across frames. We also propose a sampling-based temporal augmentation method to avoid overly enforcing invariance on clips that are distant in time. On Kinetics-600, a linear classifier trained on the representations learned by CVRL achieves 70.4% top-1 accuracy with a 3D-ResNet-50 (R3D-50) backbone, outperforming ImageNet supervised pre-training by 15.7% and SimCLR unsupervised pre-training by 18.8% using the same inflated R3D-50. The performance of CVRL can be further improved to 72.9% with a larger R3D-152 (2\u00d7 filters) backbone, significantly closing the gap between unsupervised and supervised video representation learning. Our code and models will be available at https://github.com/tensorflow/models/tree/master/official/."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "Social Media Popularity Prediction: A Multiple Feature Fusion Approach with Deep Neural Networks",
                    "abstract": "Social media popularity prediction (SMPD) aims to predict the popularity of the post shared on online social media platforms. This task is crucial for content providers and consumers in a wide range of real-world applications, including multimedia advertising, recommendation system and trend analysis. In this paper, we propose to fuse features from multiple sources by deep neural networks (DNNs) for popularity prediction. Specifically, high-level image and text features are extracted by the advanced pretrained DNN, and numerical features are captured from the metadata of the posts. All of the features are concatenated and fed into a regressor with multiple dense layers. Experiments have demonstrated the effectiveness of the proposed model on the ACM Multimedia Challenge SMPD2019 dataset. We also verify the importance of each feature via univariate test and ablation study, and provide the insights of feature combination for social media popularity prediction."
                },
                {
                    "title": "VideoBERT: A Joint Model for Video and Language Representation Learning",
                    "abstract": "Self-supervised learning has become increasingly important to leverage the abundance of unlabeled data available on platforms like YouTube. Whereas most existing approaches learn low-level representations, we propose a joint visual-linguistic model to learn high-level features without any explicit supervision. In particular, inspired by its recent success in language modeling, we build upon the BERT model to learn bidirectional joint distributions over sequences of visual and linguistic tokens, derived from vector quantization of video data and off-the-shelf speech recognition outputs, respectively. We use VideoBERT in numerous tasks, including action classification and video captioning. We show that it can be applied directly to open-vocabulary classification, and confirm that large amounts of training data and cross-modal information are critical to performance. Furthermore, we outperform the state-of-the-art on video captioning, and quantitative results verify that the model learns high-level semantic features."
                },
                {
                    "title": "Cutting through Content Clutter: How Speech and Image Acts Drive Consumer Sharing of Social Media Brand Messages",
                    "abstract": "Consumer-to-consumer brand message sharing is pivotal for effective social media marketing. Even as companies join social media conversations and generate millions of brand messages, it remains unclear what, how, and when brand messages stand out and prompt sharing by consumers. With a conceptual extension of speech act theory, this study offers a granular assessment of brands\u2019 message intentions (i.e., assertive, expressive, or directive) and the effects on consumer sharing. A text mining study of more than two years of Facebook posts and Twitter tweets by well-known consumer brands empirically demonstrates the impacts of distinct message intentions on consumers\u2019 message sharing. Specifically, the use of rhetorical styles (alliteration and repetitions) and cross-message compositions enhance consumer message sharing. As a further extension, an image-based study demonstrates that the presence of visuals, or so-called image acts, increases the ability to account for message sharing. The findings explicate brand message sharing by consumers and thus offer guidance to content managers for developing more effective conversational strategies in social media marketing."
                },
                {
                    "title": "Conceptual Captions: A Cleaned, Hypernymed, Image Alt-text Dataset For Automatic Image Captioning",
                    "abstract": "We present a new dataset of image caption annotations, Conceptual Captions, which contains an order of magnitude more images than the MS-COCO dataset (Lin et al., 2014) and represents a wider variety of both images and image caption styles. We achieve this by extracting and filtering image caption annotations from billions of webpages. We also present quantitative evaluations of a number of image captioning models and show that a model architecture based on Inception-ResNetv2 (Szegedy et al., 2016) for image-feature extraction and Transformer (Vaswani et al., 2017) for sequence modeling achieves the best performance when trained on the Conceptual Captions dataset."
                },
                {
                    "title": "Retweet Wars: Tweet Popularity Prediction via Dynamic Multimodal Regression",
                    "abstract": "If a picture is worth a thousand words, then images should be utilized together with other available data modalities when predicting the virality of online posts, such as tweets. In this paper, we re-visit the tweet popularity prediction problem by considering all data modalities: tweet language semantics, embedded images, author' social relationships, and the diffusion process of tweets. To model the content of tweets, we propose a joint-embedding neural network that combines visual, textual, and social cues together. Such content features can be either used for prediction directly, or for pre-conditioning a 'dynamics RNN', which models the message propagation process. A novel Poisson regression loss is optimized to train the network. We demonstrate that content based features can be used to improve upon social features and dynamics features via our joint-embedding regression model. Our model outperforms the state-of-the-art on multiple large-scale real-world datasets collected from Twitter."
                },
                {
                    "title": "Automatic Understanding of Image and Video Advertisements",
                    "abstract": "There is more to images than their objective physical content: for example, advertisements are created to persuade a viewer to take a certain action. We propose the novel problem of automatic advertisement understanding. To enable research on this problem, we create two datasets: an image dataset of 64,832 image ads, and a video dataset of 3,477 ads. Our data contains rich annotations encompassing the topic and sentiment of the ads, questions and answers describing what actions the viewer is prompted to take and the reasoning that the ad presents to persuade the viewer (What should I do according to this ad, and why should I do it?), and symbolic references ads make (e.g. a dove symbolizes peace). We also analyze the most common persuasive strategies ads use, and the capabilities that computer vision systems should have to understand these strategies. We present baseline classification results for several prediction tasks, including automatically answering questions about the messages of the ads."
                },
                {
                    "title": "Dense-Captioning Events in Videos",
                    "abstract": "Most natural videos contain numerous events. For example, in a video of a \u201cman playing a piano\u201d, the video might also contain \u201canother man dancing\u201d or \u201ca crowd clapping\u201d. We introduce the task of dense-captioning events, which involves both detecting and describing events in a video. We propose a new model that is able to identify all events in a single pass of the video while simultaneously describing the detected events with natural language. Our model introduces a variant of an existing proposal module that is designed to capture both short as well as long events that span minutes. To capture the dependencies between the events in a video, our model introduces a new captioning module that uses contextual information from past and future events to jointly describe all events. We also introduce ActivityNet Captions, a large-scale benchmark for dense-captioning events. ActivityNet Captions contains 20k videos amounting to 849 video hours with 100k total descriptions, each with its unique start and end time. Finally, we report performances of our model for dense-captioning events, video retrieval and localization."
                },
                {
                    "title": "Keystroke dynamics as signal for shallow syntactic parsing",
                    "abstract": "Keystroke dynamics have been extensively used in psycholinguistic and writing research to gain insights into cognitive processing. But do keystroke logs contain actual signal that can be used to learn better natural language processing models? We postulate that keystroke dynamics contain information about syntactic structure that can inform shallow syntactic parsing. To test this hypothesis, we explore labels derived from keystroke logs as auxiliary task in a multi-task bidirectional Long Short-Term Memory (bi-LSTM). Our results show promising results on two shallow syntactic parsing tasks, chunking and CCG supertagging. Our model is simple, has the advantage that data can come from distinct sources, and produces models that are significantly better than models trained on the text annotations alone."
                },
                {
                    "title": "Multimodal Popularity Prediction of Brand-related Social Media Posts",
                    "abstract": "Brand-related user posts on social networks are growing at a staggering rate, where users express their opinions about brands by sharing multimodal posts. However, while some posts become popular, others are ignored. In this paper, we present an approach for identifying what aspects of posts determine their popularity. We hypothesize that brand-related posts may be popular due to several cues related to factual information, sentiment, vividness and entertainment parameters about the brand. We call the ensemble of cues engagement parameters. In our approach, we propose to use these parameters for predicting brand-related user post popularity. Experiments on a collection of fast food brand-related user posts crawled from Instagram show that: visual and textual features are complementary in predicting the popularity of a post; predicting popularity using our proposed engagement parameters is more accurate than predicting popularity directly from visual and textual features; and our proposed approach makes it possible to understand what drives post popularity in general as well as isolate the brand specific drivers."
                },
                {
                    "title": "Improving sentence compression by learning to predict gaze",
                    "abstract": "We show how eye-tracking corpora can be used to improve sentence compression models, presenting a novel multi-task learning algorithm based on multi-layer LSTMs. We obtain performance competitive with or better than state-of-the-art approaches."
                },
                {
                    "title": "Microsoft COCO Captions: Data Collection and Evaluation Server",
                    "abstract": "In this paper we describe the Microsoft COCO Caption dataset and evaluation server. When completed, the dataset will contain over one and a half million captions describing over 330,000 images. For the training and validation images, five independent human generated captions will be provided. To ensure consistency in evaluation of automatic caption generation algorithms, an evaluation server is used. The evaluation server receives candidate captions and scores them using several popular metrics, including BLEU, METEOR, ROUGE and CIDEr. Instructions for using the evaluation server are provided."
                },
                {
                    "title": "What makes an image popular?",
                    "abstract": "Hundreds of thousands of photographs are uploaded to the internet every minute through various social networking and photo sharing platforms. While some images get millions of views, others are completely ignored. Even from the same users, different photographs receive different number of views. This begs the question: What makes a photograph popular? Can we predict the number of views a photograph will receive even before it is uploaded? These are some of the questions we address in this work. We investigate two key components of an image that affect its popularity, namely the image content and social context. Using a dataset of about 2.3 million images from Flickr, we demonstrate that we can reliably predict the normalized view count of images with a rank correlation of 0.81 using both image content and social cues. In this paper, we show the importance of image cues such as color, gradients, deep learning features and the set of objects present, as well as the importance of various social cues such as number of friends or number of photos uploaded that lead to high or low popularity of images."
                },
                {
                    "title": "Agent-based modeling: A revolution?",
                    "abstract": "A clear consensus among the papers in this Colloquium is that agent-based modeling is a revolutionary development for social science. However, the reasons to expect this revolution lie more in the potential seen in this tool than through realized results. In order for the anticipated revolution to occur, a series of challenges must be met. This paper surveys the challenges suggested by the papers of this session."
                },
                {
                    "title": "A Video Is Worth 4096 Tokens: Verbalize Story Videos To Understand Them In Zero Shot",
                    "abstract": "Multimedia content, such as advertisements and story videos, exhibit a rich blend of creativity and multiple modalities. They incorporate elements like text, visuals, audio, and storytelling techniques, employing devices like emotions, symbolism, and slogans to convey meaning. While previous research in multi-media understanding has focused mainly on videos with specific actions like cooking, there is a dearth of large annotated training datasets, hindering the development of supervised learning models with satisfactory performance for real-world applications. However, the rise of large language models (LLMs) has witnessed remarkable zero-shot performance in various natural language processing (NLP) tasks, such as emotion classification, question-answering, and topic classification. To bridge this performance gap in multimedia understanding, we propose verbalizing story videos to generate their descriptions in natural language and then performing video-understanding tasks on the generated story as opposed to the original video. Through extensive experiments on five video-understanding tasks, we demonstrate that our method, despite being zero-shot, achieves significantly better results than supervised base-lines for video understanding. Further, alleviating a lack of story understanding benchmarks, we publicly release the first dataset on a crucial task in computational social science, persuasion strategy identification."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "The structure and function of communication in society",
                    "abstract": "Turkiye'deki gibi hem siyasal bilgiler fakultesinde dikis tutturamayip iletisim fakultelerine gelen, hem iletisimi hor goren hem de iletisimin olma kosulu olarak \"en az iki kisinin gerektigini\" (gonderici ve alici olmasini) one surecek kadar bilgisiz olanlari yetistirenlerin aksine, ABD'de iletisim alanini kuran ve gelistirenlerin onde gelenleri sosyal bilimciler, ozellikle sosyologlar ve siyaset bilimcileri, olmustur. Nermin Abadan-Unat ve Unsal Oskay gibi bazi istisnalar disinda, hem politika ve iletisim arasinda dogru ve anlamli bag kuramayacak kadar \"yetersizlerin\" hem de bu yetersizliklerini Ingilizceyi iyice ogrenerek ve tembelliklerini yenerek giderme yerine okul odalarinda cayli-kahveli dedikoduyu secenlerin ve iletisim alanina, hic bir olumlu katkida bulunmadiklari gibi, zarar verenlerin aksine, ABD'deki siyaset bilimciler iletisimin gelismesine buyuk katkida bulunmuslardir. Bunlarin basinda gelenlerden biri de H. Lasswell'dir. Harold Lasswell iletisimde sadece bir formul one surmemis, iletisimin yapisi ve toplumdaki islevleri hakkinda bir kuram gelistirmistir. Lasswell'in klasik kuraminin birkac kelimelik formul otesinde bilinmesi ve bu gorusun irdelenmesi gerekir. Lasswell'in \"Structure and function of communication in society\" makalesinden parcalar sunulmasindaki amac, iletisimin gonderici- Mesaj-Alici-Geribesleme cercevesi icinde ele alan basitlikten ve Defleur gibi bilim adamlarinin ultra-tutucu bagnazliklarinin otesinde anlasilmasidir. Bu sunum iletisimin mesaj gonderme-alma mekanik surec algisinin tersine toplumda surekliligi saglamayla iliskisini ortaya koymaktadir. 1 Original kaynak: Lyman Bryson (1948) (ed.) The Communication of Ideas. New york: The Institute for Religious and Social Studies. Kaynak: Schramm, W. ve Roberts, D. F. (1971) The Process and Effects of Mass Communication. Urbana: University of Illinois Press, s. 84-99. Forum"
                },
                {
                    "title": "A Mathematical Theory of Communication",
                    "abstract": "This paper opened the new area the information theory. Before this paper, most people believed that the only way to make the error probability of transmission as small as desired is to reduce the data rate (such as a long repetition scheme). However, surprisingly this paper revealed that it does not need to reduce the data rate for achieving that much of small errors. It proved that we can get some positive data rate that has the same small error probability and also there is an upper bound of the data rate, which means we cannot achieve the data rate with any encoding scheme that has small enough error probability over the upper bound."
                },
                {
                    "title": "Communication models for the study of mass communications",
                    "abstract": "1. Introduction 2. Basic Models 3. Personal Influence, Diffusion and Short-Term Effects of Mass Communication on Individuals 4. Effects of Mass Communication on Culture and Society 5. Audience-Centred Models 6. Media Organisation, Selection and Production 7. Planned Communication 8. New Media and The Information Society 9. International Communication Index"
                }
            ],
            "categories": [
                "cs.CL",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively model and predict content behavior and understanding in the language space using Large Content Behavior Models (LCBMs)?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of content dynamics and user interactions across various platforms. By developing LCBMs that can accurately simulate behavior and understand content, we can enhance applications in content recommendation, marketing strategies, and user engagement analysis. This research could lead to more personalized and effective content delivery, ultimately influencing future research directions in machine learning, natural language processing, and human-computer interaction.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in this problem stem from the complexity of accurately modeling human behavior and content interactions, which involve nuanced emotional and contextual factors. Naive approaches may fail due to their inability to capture the intricate relationships between content features and user responses. Additionally, technical obstacles include the need for large, diverse datasets to train models effectively, as well as the computational resources required to process and analyze this data. Theoretical challenges also arise in defining appropriate metrics for evaluating model performance across different tasks.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on isolated aspects of content or behavior, lacking a comprehensive framework that integrates both. Limitations in existing models, such as their size and scope, have hindered their ability to generalize across different content types and user behaviors. Barriers such as insufficient data, inadequate modeling techniques, and the rapid evolution of content platforms have also contributed to the lack of effective solutions. Our approach aims to bridge these gaps by leveraging LCBMs that are specifically designed to handle the complexities of content behavior modeling.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves training LCBMs on a diverse dataset that includes various forms of content (images, videos, text) and user interactions. We will utilize metrics such as accuracy for behavior simulation tasks and BLEU/ROUGE scores for content simulation tasks. The expected outcomes include improved performance in predicting user behavior, enhanced content understanding, and the ability to generate meaningful insights into viewer sentiments and engagement strategies. By comparing our model's performance against existing benchmarks, we aim to demonstrate its effectiveness and applicability in real-world scenarios."
            }
        },
        "author_data": {
            "6eb883c7-b0b3-46f4-8bd9-b19b1c4ee009": {
                "pk": "6eb883c7-b0b3-46f4-8bd9-b19b1c4ee009",
                "name": "Ashmit Khandelwal",
                "collaborators": [
                    "Sina Malakouti",
                    "Aysan Aghazadeh",
                    "Adriana Kovashka",
                    "Aditya Agrawal",
                    "Aanisha Bhattacharyya",
                    "Yaman Kumar Singla",
                    "Somesh Singh",
                    "Uttaran Bhattacharya",
                    "Ishita Dasgupta",
                    "Stefano Petrangeli",
                    "R. Shah",
                    "Changan Chen",
                    "Balaji Krishnamurthy"
                ],
                "domain": [
                    "Vision Language Models",
                    "Persuasive Media",
                    "Large Language Models",
                    "Reasoning"
                ],
                "publications": [
                    {
                        "title": "Benchmarking VLMs' Reasoning About Persuasive Atypical Images",
                        "abstract": "Vision language models (VLMs) have shown strong zero-shot generalization across various tasks, especially when integrated with large language models (LLMs). However, their ability to comprehend rhetorical and persuasive visual media, such as advertisements, remains understudied. Ads often employ atypical imagery, using surprising object juxtapositions to convey shared properties. For example, Fig. 1 (e) shows a beer with a feather-like texture. This requires advanced reasoning to deduce that this atypical representation signifies the beer's lightness. We introduce three novel tasks, Multi-label Atypicality Classification, Atypicality Statement Retrieval, and Aypical Object Recognition, to benchmark VLMs' understanding of atypicality in persuasive images. We evaluate how well VLMs use atypicality to infer an ad's message and test their reasoning abilities by employing semantically challenging negatives. Finally, we pioneer atypicality-aware verbalization by extracting comprehensive image descriptions sensitive to atypical elements. Our findings reveal that: (1) VLMs lack advanced reasoning capabilities compared to LLMs; (2) simple, effective strategies can extract atypicality-aware information, leading to comprehensive image verbalization; (3) atypicality aids persuasive advertisement understanding. Code and data will be made available."
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 2 [ cs .C L ] 8 S ep 2 02 3"
                    }
                ]
            },
            "03a8ca9c-2d1e-4d38-9079-4345345ea970": {
                "pk": "03a8ca9c-2d1e-4d38-9079-4345345ea970",
                "name": "Aditya Agrawal",
                "collaborators": [
                    "Aanisha Bhattacharyya",
                    "Yaman Kumar Singla",
                    "Somesh Singh",
                    "Uttaran Bhattacharya",
                    "Ishita Dasgupta",
                    "Stefano Petrangeli",
                    "R. Shah",
                    "Changan Chen",
                    "Balaji Krishnamurthy",
                    "Ashmit Khandelwal"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Communication Theory",
                    "Behavior Modeling"
                ],
                "publications": [
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. \u22c6 Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 1 [ cs .C L ] 1 S ep 2 02 3 Published as a conference paper at ICLR 2023"
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 2 [ cs .C L ] 8 S ep 2 02 3"
                    }
                ]
            },
            "cee59623-e6ae-462f-a84e-37524ad8561d": {
                "pk": "cee59623-e6ae-462f-a84e-37524ad8561d",
                "name": "Aanisha Bhattacharyya",
                "collaborators": [
                    "Yaman Kumar Singla",
                    "Somesh Singh",
                    "R. Shah",
                    "Changan Chen",
                    "Balaji Krishnamurthy",
                    "Aditya Agrawal",
                    "Uttaran Bhattacharya",
                    "Ishita Dasgupta",
                    "Stefano Petrangeli",
                    "I. HariniS",
                    "V. Baths",
                    "Ashmit Khandelwal",
                    "Tripti Shukla",
                    "Ayudh Saxena",
                    "Jeevana Kruthi Karnuthala",
                    "Abhinav Bohra",
                    "Bhanu Prakash Reddy Guda",
                    "Abhilasha Sancheti",
                    "Niyati Chhaya"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Multimedia Processing",
                    "Behavioral Analysis"
                ],
                "publications": [
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. \u22c6 Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 1 [ cs .C L ] 1 S ep 2 02 3 Published as a conference paper at ICLR 2023"
                    },
                    {
                        "title": "Long-Term Ad Memorability: Understanding&Generating Memorable Ads",
                        "abstract": "Marketers spend billions of dollars on advertisements, but to what end? At purchase time, if customers cannot recognize the brand for which they saw an ad, the money spent on the ad is essentially wasted. Despite its importance in marketing, until now, there has been no large-scale study on the memorability of ads. All previous memorability studies have been conducted on short-term recall on specific content types like action videos. On the other hand, the advertising industry only cares about long-term memorability, and ads are almost always highly multimodal. Therefore, we release the first memorability dataset, LAMBDA, consisting of 1749 participants and 2205 ads covering 276 brands. Running statistical tests over different participant subpopulations and ad types, we find many interesting insights into what makes an ad memorable, e.g., fast-moving ads are more memorable than those with slower scenes; people who use ad-blockers remember a lower number of ads than those who don't. Next, we present a model, Henry, to predict the memorability of a content. Henry achieves state-of-the-art performance across all prominent literature memorability datasets. It shows strong generalization performance with better results in 0-shot on unseen datasets. Finally, with the intent of memorable ad generation, we present a scalable method to build a high-quality memorable ad generation model by leveraging automatically annotated data. Our approach, SEED (Self rEwarding mEmorability Modeling), starts with a language model trained on LAMBDA as seed data and progressively trains an LLM to generate more memorable ads. We show that the generated advertisements have 44% higher memorability scores than the original ads. We release this large-scale ad dataset, UltraLAMBDA, consisting of 5 million ads. Our code and datasets are available at https://behavior-in-the-wild.github.io/memorability."
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 2 [ cs .C L ] 8 S ep 2 02 3"
                    },
                    {
                        "title": "Aanisha@TamilNLP-ACL2022:Abusive Detection in Tamil",
                        "abstract": "In social media, there are instances where people present their opinions in strong language, resorting to abusive/toxic comments.There are instances of communal hatred, hate-speech, toxicity and bullying. And, in this age of social media, it\u2019s very important to find means to keep check on these toxic comments, as to preserve the mental peace of people in social media.While there are tools, models to detect andpotentially filter these kind of content, developing these kinds of models for the low resource language space is an issue of research.In this paper, the task of abusive comment identification in Tamil language, is seen upon as a multi-class classification problem.There are different pre-processing as well as modelling approaches discussed in this paper.The different approaches are compared on the basis of weighted average accuracy."
                    },
                    {
                        "title": "Videos2Doc: Generating Documents from a Collection of Procedural Videos",
                        "abstract": "The availability of user-generated multi-modal content, including videos and images, in abundance makes it easy for users to use it as a reference and source of information. However, several hours may be required for the consumption of this large corpus of data. Particularly, for authors and content creators abstracting out information from videos to then representing it in a textual format is a tedious task. The challenges are multiplied due to the diversity and the variety introduced when there are several videos associated with a given query or topic of interest. We present, Videos2Doc, a machine learning-based framework for automated document generation from a collection of procedural videos. Videos2Doc enables author-guided document generation for those looking for authoring assistance and an easy consumption experience for those preferring the text or document media over videos. Our proposed interface allows the users to choose several visual and semantic preferences for the output document allowing the generation of custom documents and webpage templates from a given set of inputs. Empirical and qualitative evaluations establish the utility of Videos2Doc as well as the superiority over the current benchmarks. We believe, Videos2Doc will ease the task of making multimedia accessible through automation in conversion to alternate presentation modes."
                    }
                ]
            },
            "291dcb1e-b1b8-44ab-9697-6e23b910fe0f": {
                "pk": "291dcb1e-b1b8-44ab-9697-6e23b910fe0f",
                "name": "Yaman K Singla",
                "collaborators": [
                    "R. Shah",
                    "Balaji Krishnamurthy",
                    "Changyou Chen",
                    "Somesh Singh",
                    "Changan Chen",
                    "Aanisha Bhattacharyya",
                    "I. HariniS",
                    "V. Baths",
                    "Aditya Agrawal",
                    "Uttaran Bhattacharya",
                    "Ishita Dasgupta",
                    "Stefano Petrangeli",
                    "Aanisha Bhattacharya",
                    "Varun Khurana",
                    "Mann Khatri",
                    "P. Kumaraguru",
                    "Jui Shah",
                    "SI Harini",
                    "Shubhankar K. Singh",
                    "Anirudh Pupneja",
                    "Shivaansh Mital",
                    "Cheril Shah",
                    "Manish Bawkar",
                    "Lakshman Prasad Gupta",
                    "Ajit Kumar",
                    "Rushali Gupta",
                    "J. Subramanian",
                    "Zhiqiang Xu",
                    "Mirza Yusuf",
                    "Nora Hollenstein",
                    "R. Kumar",
                    "Ashmit Khandelwal",
                    "Pritish Wadhwa",
                    "G. Satija",
                    "Reshma Sheik",
                    "R. Jha",
                    "Arunim Gupta",
                    "Milan Aggarwal",
                    "Aditya Garg",
                    "Tushar Malyan",
                    "Ayush Bhardwaj",
                    "Debanjan Mahata",
                    "Naveen Agarwal",
                    "Dibya Gautam",
                    "Amardeep Kumar",
                    "Swapnil Parekh",
                    "Anish Acharya",
                    "Sreyan Ghosh",
                    "Sonal Kumar",
                    "S. Umesh"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Multimedia Understanding",
                    "Persuasion Modeling",
                    "Automated Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Teaching Human Behavior Improves Content Understanding Abilities Of LLMs",
                        "abstract": "Communication is defined as\"Who says what to whom with what effect\". A message from a communicator generates downstream receiver effects, also known as behavior. Receiver behavior, being a downstream effect of the message, carries rich signals about it. Even after carrying signals about the message, the behavior data is often ignored while training large language models. We show that training LLMs on receiver behavior can actually help improve their content-understanding abilities. Specifically, we show that training LLMs to predict the receiver behavior of likes and comments improves the LLM's performance on a wide variety of downstream content understanding tasks. We show this performance increase over 46 video and image understanding tasks over 26 benchmark datasets across both 0-shot and fine-tuning settings, outperforming many supervised baselines. Moreover, since receiver behavior, such as likes and comments, is collected by default on the internet and does not need any human annotations to be useful, the performance improvement we get after training on this data is essentially free-lunch. We release the receiver behavior cleaned comments and likes of 750k images and videos collected from multiple platforms along with our instruction-tuning data."
                    },
                    {
                        "title": "Measuring and Improving Persuasiveness of Large Language Models",
                        "abstract": "LLMs are increasingly being used in workflows involving generating content to be consumed by humans (e.g., marketing) and also in directly interacting with humans (e.g., through chatbots). The development of such systems that are capable of generating verifiably persuasive messages presents both opportunities and challenges for society. On the one hand, such systems could positively impact domains like advertising and social good, such as addressing drug addiction, and on the other, they could be misused for spreading misinformation and shaping political opinions. To channel LLMs' impact on society, we need to develop systems to measure and benchmark their persuasiveness. With this motivation, we introduce PersuasionBench and PersuasionArena, the first large-scale benchmark and arena containing a battery of tasks to measure the persuasion ability of generative models automatically. We investigate to what extent LLMs know and leverage linguistic patterns that can help them generate more persuasive language. Our findings indicate that the persuasiveness of LLMs correlates positively with model size, but smaller models can also be made to have a higher persuasiveness than much larger models. Notably, targeted training using synthetic and natural datasets significantly enhances smaller models' persuasive capabilities, challenging scale-dependent assumptions. Our findings carry key implications for both model developers and policymakers. For instance, while the EU AI Act and California's SB-1047 aim to regulate AI models based on the number of floating point operations, we demonstrate that simple metrics like this alone fail to capture the full scope of AI's societal impact. We invite the community to explore and contribute to PersuasionArena and PersuasionBench, available at https://bit.ly/measure-persuasion, to advance our understanding of AI-driven persuasion and its societal implications."
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. \u22c6 Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 1 [ cs .C L ] 1 S ep 2 02 3 Published as a conference paper at ICLR 2023"
                    },
                    {
                        "title": "Long-Term Ad Memorability: Understanding&Generating Memorable Ads",
                        "abstract": "Marketers spend billions of dollars on advertisements, but to what end? At purchase time, if customers cannot recognize the brand for which they saw an ad, the money spent on the ad is essentially wasted. Despite its importance in marketing, until now, there has been no large-scale study on the memorability of ads. All previous memorability studies have been conducted on short-term recall on specific content types like action videos. On the other hand, the advertising industry only cares about long-term memorability, and ads are almost always highly multimodal. Therefore, we release the first memorability dataset, LAMBDA, consisting of 1749 participants and 2205 ads covering 276 brands. Running statistical tests over different participant subpopulations and ad types, we find many interesting insights into what makes an ad memorable, e.g., fast-moving ads are more memorable than those with slower scenes; people who use ad-blockers remember a lower number of ads than those who don't. Next, we present a model, Henry, to predict the memorability of a content. Henry achieves state-of-the-art performance across all prominent literature memorability datasets. It shows strong generalization performance with better results in 0-shot on unseen datasets. Finally, with the intent of memorable ad generation, we present a scalable method to build a high-quality memorable ad generation model by leveraging automatically annotated data. Our approach, SEED (Self rEwarding mEmorability Modeling), starts with a language model trained on LAMBDA as seed data and progressively trains an LLM to generate more memorable ads. We show that the generated advertisements have 44% higher memorability scores than the original ads. We release this large-scale ad dataset, UltraLAMBDA, consisting of 5 million ads. Our code and datasets are available at https://behavior-in-the-wild.github.io/memorability."
                    },
                    {
                        "title": "A Video Is Worth 4096 Tokens: Verbalize Videos To Understand Them In Zero Shot",
                        "abstract": "Multimedia content, such as advertisements and story videos, exhibit a rich blend of creativity and multiple modalities. They incorporate elements like text, visuals, audio, and storytelling techniques, employing devices like emotions, symbolism, and slogans to convey meaning. There is a dearth of large annotated training datasets in the multimedia domain hindering the development of supervised learning models with satisfactory performance for real-world applications. On the other hand, the rise of large language models (LLMs) has witnessed remarkable zero-shot performance in various natural language processing (NLP) tasks, such as emotion classification, question-answering, and topic classification. To leverage such advanced techniques to bridge this performance gap in multimedia understanding, we propose verbalizing long videos to generate their descriptions in natural language, followed by performing video-understanding tasks on the generated story as opposed to the original video. Through extensive experiments on fifteen video-understanding tasks, we demonstrate that our method, despite being zero-shot, achieves significantly better results than supervised baselines for video understanding. Furthermore, to alleviate a lack of story understanding benchmarks, we publicly release the first dataset on a crucial task in computational social science on persuasion strategy identification."
                    },
                    {
                        "title": "H-AES: Towards Automated Essay Scoring for Hindi",
                        "abstract": "The use of Natural Language Processing (NLP) for Automated Essay Scoring (AES) has been well explored in the English language, with benchmark models exhibiting performance comparable to human scorers. However, AES in Hindi and other low-resource languages remains unexplored. In this study, we reproduce and compare state-of-the-art methods for AES in the Hindi domain. We employ classical feature-based Machine Learning (ML) and advanced end-to-end models, including LSTM Networks and Fine-Tuned Transformer Architecture, in our approach and derive results comparable to those in the English language domain. Hindi being a low-resource language, lacks a dedicated essay-scoring corpus. We train and evaluate our models using translated English essays and empirically measure their performance on our own small-scale, real-world Hindi corpus. We follow this up with an in-depth analysis discussing prompt-specific behavior of different language models implemented."
                    },
                    {
                        "title": "Behavior Optimized Image Generation",
                        "abstract": "The last few years have witnessed great success on image generation, which has crossed the acceptance thresholds of aesthetics, making it directly applicable to personal and commercial applications. However, images, especially in marketing and advertising applications, are often created as a means to an end as opposed to just aesthetic concerns. The goal can be increasing sales, getting more clicks, likes, or image sales (in the case of stock businesses). Therefore, the generated images need to perform well on these key performance indicators (KPIs), in addition to being aesthetically good. In this paper, we make the first endeavor to answer the question of\"How can one infuse the knowledge of the end-goal within the image generation process itself to create not just better-looking images but also\"better-performing'' images?''. We propose BoigLLM, an LLM that understands both image content and user behavior. BoigLLM knows how an image should look to get a certain required KPI. We show that BoigLLM outperforms 13x larger models such as GPT-3.5 and GPT-4 in this task, demonstrating that while these state-of-the-art models can understand images, they lack information on how these images perform in the real world. To generate actual pixels of behavior-conditioned images, we train a diffusion-based model (BoigSD) to align with a proposed BoigLLM-defined reward. We show the performance of the overall pipeline on two datasets covering two different behaviors: a stock dataset with the number of forward actions as the KPI and a dataset containing tweets with the total likes as the KPI, denoted as BoigBench. To advance research in the direction of utility-driven image generation and understanding, we release BoigBench, a benchmark dataset containing 168 million enterprise tweets with their media, brand account names, time of post, and total likes."
                    },
                    {
                        "title": "Exploring Graph Neural Networks for Indian Legal Judgment Prediction",
                        "abstract": "The burdensome impact of a skewed judges-to-cases ratio on the judicial system manifests in an overwhelming backlog of pending cases alongside an ongoing influx of new ones. To tackle this issue and expedite the judicial process, the proposition of an automated system capable of suggesting case outcomes based on factual evidence and precedent from past cases gains significance. This research paper centres on developing a graph neural network-based model to address the Legal Judgment Prediction (LJP) problem, recognizing the intrinsic graph structure of judicial cases and making it a binary node classification problem. We explored various embeddings as model features, while nodes such as time nodes and judicial acts were added and pruned to evaluate the model's performance. The study is done while considering the ethical dimension of fairness in these predictions, considering gender and name biases. A link prediction task is also conducted to assess the model's proficiency in anticipating connections between two specified nodes. By harnessing the capabilities of graph neural networks and incorporating fairness analyses, this research aims to contribute insights towards streamlining the adjudication process, enhancing judicial efficiency, and fostering a more equitable legal landscape, ultimately alleviating the strain imposed by mounting case backlogs. Our best-performing model with XLNet pre-trained embeddings as its features gives the macro F1 score of 75% for the LJP task. For link prediction, the same set of features is the best performing giving ROC of more than 80%"
                    },
                    {
                        "title": "Synthesizing Human Gaze Feedback for Improved NLP Performance",
                        "abstract": "Integrating human feedback in models can improve the performance of natural language processing (NLP) models. Feedback can be either explicit (e.g. ranking used in training language models) or implicit (e.g. using human cognitive signals in the form of eyetracking). Prior eye tracking and NLP research reveal that cognitive processes, such as human scanpaths, gleaned from human gaze patterns aid in the understanding and performance of NLP models. However, the collection of real eyetracking data for NLP tasks is challenging due to the requirement of expensive and precise equipment coupled with privacy invasion issues. To address this challenge, we propose ScanTextGAN, a novel model for generating human scanpaths over text. We show that ScanTextGAN-generated scanpaths can approximate meaningful cognitive signals in human gaze patterns. We include synthetically generated scanpaths in four popular NLP tasks spanning six different datasets as proof of concept and show that the models augmented with generated scanpaths improve the performance of all downstream NLP tasks."
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 2 [ cs .C L ] 8 S ep 2 02 3"
                    },
                    {
                        "title": "CiteCaseLAW: Citation Worthiness Detection in Caselaw for Legal Assistive Writing",
                        "abstract": "In legal document writing, one of the key elements is properly citing the case laws and other sources to substantiate claims and arguments. Understanding the legal domain and identifying appropriate citation context or cite-worthy sentences are challenging tasks that demand expensive manual annotation. The presence of jargon, language semantics, and high domain specificity makes legal language complex, making any associated legal task hard for automation. The current work focuses on the problem of citation-worthiness identification. It is designed as the initial step in today's citation recommendation systems to lighten the burden of extracting an adequate set of citation contexts. To accomplish this, we introduce a labeled dataset of 178M sentences for citation-worthiness detection in the legal domain from the Caselaw Access Project (CAP). The performance of various deep learning models was examined on this novel dataset. The domain-specific pre-trained model tends to outperform other models, with an 88% F1-score for the citation-worthiness detection task."
                    },
                    {
                        "title": "A Video Is Worth 4096 Tokens: Verbalize Story Videos To Understand Them In Zero Shot",
                        "abstract": "Multimedia content, such as advertisements and story videos, exhibit a rich blend of creativity and multiple modalities. They incorporate elements like text, visuals, audio, and storytelling techniques, employing devices like emotions, symbolism, and slogans to convey meaning. While previous research in multi-media understanding has focused mainly on videos with specific actions like cooking, there is a dearth of large annotated training datasets, hindering the development of supervised learning models with satisfactory performance for real-world applications. However, the rise of large language models (LLMs) has witnessed remarkable zero-shot performance in various natural language processing (NLP) tasks, such as emotion classification, question-answering, and topic classification. To bridge this performance gap in multimedia understanding, we propose verbalizing story videos to generate their descriptions in natural language and then performing video-understanding tasks on the generated story as opposed to the original video. Through extensive experiments on five video-understanding tasks, we demonstrate that our method, despite being zero-shot, achieves significantly better results than supervised base-lines for video understanding. Further, alleviating a lack of story understanding benchmarks, we publicly release the first dataset on a crucial task in computational social science, persuasion strategy identification."
                    },
                    {
                        "title": "What Do Audio Transformers Hear? Probing Their Representations For Language Delivery & Structure",
                        "abstract": "Transformer models across multiple domains such as natural language processing and speech form an unavoidable part of the tech stack of practitioners and researchers alike. Au-dio transformers that exploit representational learning to train on unlabeled speech have recently been used for tasks from speaker verification to discourse-coherence with much success. However, little is known about what these models learn and represent in the high-dimensional latent space. In this paper, we interpret two such recent state-of-the-art models, wav2vec2.0 and Mockingjay, on linguistic and acoustic features. We probe each of their layers to understand what it is learning and at the same time, we draw a distinction between the two models. By comparing their performance across a wide variety of settings including native, non-native, read and spontaneous speeches, we also show how much these models are able to learn transferable features. Our results show that the models are capable of significantly capturing a wide range of characteristics such as audio, fluency, supraseg-mental pronunciation, and even syntactic and semantic text-based characteristics. For each category of characteristics, we identify a learning pattern for each framework and conclude which model and which layer of that model is better for a specific category of feature to choose for feature extraction for downstream tasks."
                    },
                    {
                        "title": "Persuasion Strategies in Advertisements: Dataset, Modeling, and Baselines",
                        "abstract": "Modeling what makes an advertisement persuasive, i.e., eliciting the desired response from consumer, is critical to the study of propaganda, social psychology, and marketing. Despite its importance, computational modeling of persuasion in computer vision is still in its infancy, primarily due to the lack of benchmark datasets that can provide persuasion-strategy labels associated with ads. Motivated by persuasion literature in social psychology and marketing, we introduce an extensive vocabulary of persuasion strategies and build the first ad image corpus annotated with persuasion strategies. We then formulate the task of persuasion strategy prediction with multi-modal learning, where we design a multi-task attention fusion model that can leverage other ad-understanding tasks to predict persuasion strategies. The dataset also provides image segmentation masks, which labels persuasion strategies in the corresponding ad images on the test split. We publicly release our code and dataset at https://midas-research.github.io/persuasion-advertisements/."
                    },
                    {
                        "title": "LDKP: A Dataset for Identifying Keyphrases from Long Scientific Documents",
                        "abstract": "Identifying keyphrases (KPs) from text documents is a fundamental task in natural language processing and information retrieval. Vast majority of the benchmark datasets for this task are from the scientific domain containing only the document title and abstract information. This limits keyphrase extraction (KPE) and keyphrase generation (KPG) algorithms to identify keyphrases from human-written summaries that are often very short (approx 8 sentences). This presents three challenges for real-world applications: human-written summaries are unavailable for most documents, the documents are almost always long, and a high percentage of KPs are directly found beyond the limited context of title and abstract. Therefore, we release two extensive corpora mapping KPs of ~1.3M and ~100K scientific articles with their fully extracted text and additional metadata including publication venue, year, author, field of study, and citations for facilitating research on this real-world problem."
                    },
                    {
                        "title": "Span Classification with Structured Information for Disfluency Detection in Spoken Utterances",
                        "abstract": "Existing approaches in disfluency detection focus on solving a token-level classification task for identifying and removing disfluencies in text. Moreover, most works focus on leveraging only contextual information captured by the linear sequences in text, thus ignoring the structured information in text which is efficiently captured by dependency trees. In this paper, building on the span classification paradigm of entity recognition, we propose a novel architecture for detecting disfluencies in transcripts from spoken utterances, incorporating both contextual information through transformers and long-distance structured information captured by dependency trees, through graph convolutional networks (GCNs). Experimental results show that our proposed model achieves state-of-the-art results on the widely used English Switchboard for disfluency detection and outperforms prior-art by a significant margin. We make all our codes publicly available on GitHub (https://github.com/Sreyan88/Disfluency-Detection-with-Span-Classification)"
                    },
                    {
                        "title": "What all do audio transformer models hear? Probing Acoustic Representations for Language Delivery and its Structure",
                        "abstract": "In recent times, BERT based transformer models have become an inseparable part of the 'tech stack' of text processing models. Similar progress is being observed in the speech domain with a multitude of models observing state-of-the-art results by using audio transformer models to encode speech. This begs the question of what are these audio transformer models learning. Moreover, although the standard methodology is to choose the last layer embedding for any downstream task, but is it the optimal choice? We try to answer these questions for the two recent audio transformer models, Mockingjay and wave2vec2.0. We compare them on a comprehensive set of language delivery and structure features including audio, fluency and pronunciation features. Additionally, we probe the audio models' understanding of textual surface, syntax, and semantic features and compare them to BERT. We do this over exhaustive settings for native, non-native, synthetic, read and spontaneous speech datasets"
                    }
                ]
            },
            "49e78eb9-f00f-4fa8-8423-411b91a7523f": {
                "pk": "49e78eb9-f00f-4fa8-8423-411b91a7523f",
                "name": "Somesh Singh",
                "collaborators": [
                    "Yaman Kumar Singla",
                    "Balaji Krishnamurthy",
                    "R. Shah",
                    "Changyou Chen",
                    "Aanisha Bhattacharyya",
                    "Changan Chen",
                    "I. HariniS",
                    "V. Baths",
                    "Aditya Agrawal",
                    "Uttaran Bhattacharya",
                    "Ishita Dasgupta",
                    "Stefano Petrangeli",
                    "Swapnil Parekh",
                    "SI Harini",
                    "Ashmit Khandelwal",
                    "J. Li",
                    "Rajaswa Patil",
                    "Swati Agarwal",
                    "\u22c6. HariniSI"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Persuasion",
                    "Adversarial Attacks"
                ],
                "publications": [
                    {
                        "title": "Teaching Human Behavior Improves Content Understanding Abilities Of LLMs",
                        "abstract": "Communication is defined as\"Who says what to whom with what effect\". A message from a communicator generates downstream receiver effects, also known as behavior. Receiver behavior, being a downstream effect of the message, carries rich signals about it. Even after carrying signals about the message, the behavior data is often ignored while training large language models. We show that training LLMs on receiver behavior can actually help improve their content-understanding abilities. Specifically, we show that training LLMs to predict the receiver behavior of likes and comments improves the LLM's performance on a wide variety of downstream content understanding tasks. We show this performance increase over 46 video and image understanding tasks over 26 benchmark datasets across both 0-shot and fine-tuning settings, outperforming many supervised baselines. Moreover, since receiver behavior, such as likes and comments, is collected by default on the internet and does not need any human annotations to be useful, the performance improvement we get after training on this data is essentially free-lunch. We release the receiver behavior cleaned comments and likes of 750k images and videos collected from multiple platforms along with our instruction-tuning data."
                    },
                    {
                        "title": "Measuring and Improving Persuasiveness of Large Language Models",
                        "abstract": "LLMs are increasingly being used in workflows involving generating content to be consumed by humans (e.g., marketing) and also in directly interacting with humans (e.g., through chatbots). The development of such systems that are capable of generating verifiably persuasive messages presents both opportunities and challenges for society. On the one hand, such systems could positively impact domains like advertising and social good, such as addressing drug addiction, and on the other, they could be misused for spreading misinformation and shaping political opinions. To channel LLMs' impact on society, we need to develop systems to measure and benchmark their persuasiveness. With this motivation, we introduce PersuasionBench and PersuasionArena, the first large-scale benchmark and arena containing a battery of tasks to measure the persuasion ability of generative models automatically. We investigate to what extent LLMs know and leverage linguistic patterns that can help them generate more persuasive language. Our findings indicate that the persuasiveness of LLMs correlates positively with model size, but smaller models can also be made to have a higher persuasiveness than much larger models. Notably, targeted training using synthetic and natural datasets significantly enhances smaller models' persuasive capabilities, challenging scale-dependent assumptions. Our findings carry key implications for both model developers and policymakers. For instance, while the EU AI Act and California's SB-1047 aim to regulate AI models based on the number of floating point operations, we demonstrate that simple metrics like this alone fail to capture the full scope of AI's societal impact. We invite the community to explore and contribute to PersuasionArena and PersuasionBench, available at https://bit.ly/measure-persuasion, to advance our understanding of AI-driven persuasion and its societal implications."
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. \u22c6 Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 1 [ cs .C L ] 1 S ep 2 02 3 Published as a conference paper at ICLR 2023"
                    },
                    {
                        "title": "Long-Term Ad Memorability: Understanding&Generating Memorable Ads",
                        "abstract": "Marketers spend billions of dollars on advertisements, but to what end? At purchase time, if customers cannot recognize the brand for which they saw an ad, the money spent on the ad is essentially wasted. Despite its importance in marketing, until now, there has been no large-scale study on the memorability of ads. All previous memorability studies have been conducted on short-term recall on specific content types like action videos. On the other hand, the advertising industry only cares about long-term memorability, and ads are almost always highly multimodal. Therefore, we release the first memorability dataset, LAMBDA, consisting of 1749 participants and 2205 ads covering 276 brands. Running statistical tests over different participant subpopulations and ad types, we find many interesting insights into what makes an ad memorable, e.g., fast-moving ads are more memorable than those with slower scenes; people who use ad-blockers remember a lower number of ads than those who don't. Next, we present a model, Henry, to predict the memorability of a content. Henry achieves state-of-the-art performance across all prominent literature memorability datasets. It shows strong generalization performance with better results in 0-shot on unseen datasets. Finally, with the intent of memorable ad generation, we present a scalable method to build a high-quality memorable ad generation model by leveraging automatically annotated data. Our approach, SEED (Self rEwarding mEmorability Modeling), starts with a language model trained on LAMBDA as seed data and progressively trains an LLM to generate more memorable ads. We show that the generated advertisements have 44% higher memorability scores than the original ads. We release this large-scale ad dataset, UltraLAMBDA, consisting of 5 million ads. Our code and datasets are available at https://behavior-in-the-wild.github.io/memorability."
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 2 [ cs .C L ] 8 S ep 2 02 3"
                    },
                    {
                        "title": "MINIMAL: Mining Models for Universal Adversarial Triggers",
                        "abstract": "It is well known that natural language models are vulnerable to adversarial attacks, which are mostly input-specific in nature. Recently, it has been shown that there also exist input-agnostic attacks in NLP models, called universal adversarial triggers. However, existing methods to craft universal triggers are data intensive. They require large amounts of data samples to generate adversarial triggers, which are typically inaccessible by attackers. For instance, previous works take 3000 data samples per class for the SNLI dataset to generate adversarial triggers. In this paper, we present a novel data-free approach, MINIMAL, to mine input-agnostic adversarial triggers from models. Using the triggers produced with our data-free algorithm, we reduce the accuracy of Stanford Sentiment Treebank\u2019s positive class from 93.6% to 9.6%. Similarly, for the Stanford Natural LanguageInference (SNLI), our single-word trigger reduces the accuracy of the entailment class from 90.95% to less than 0.6%. Despite being completely data-free, we get equivalent accuracy drops as data-dependent methods"
                    },
                    {
                        "title": "Automatic Essay Scoring Systems Are Both Overstable And Oversensitive: Explaining Why And Proposing Defenses",
                        "abstract": "Deep-learning based Automatic Essay Scoring (AES) systems are being actively used in various high-stake applications in education and testing. However, little research has been put to understand and interpret the black-box nature of deep-learning-based scoring algorithms. While previous studies indicate that scoring models can be easily fooled, in this paper, we explore the reason behind their surprising adversarial brittleness. We utilize recent advances in interpretability to find the extent to which features such as coherence, content, vocabulary, and relevance are important for automated scoring mechanisms. We use this to investigate the oversensitivity (i.e., large change in output score with a little change in input essay content) and overstability (i.e., little change in output scores with large changes in input essay content) of AES. Our results indicate that autoscoring models, despite getting trained as \u201cend-to-end\u201d models with rich contextual embeddings such as BERT, behave like bag-of-words models. A few words determine the essay score without the requirement of any context making the model largely overstable. This is in stark contrast to recent probing studies on pre-trained representation learning models, which show that rich linguistic features such as parts-of-speech and morphology are encoded by them. Further, we also find that the models have learnt dataset biases, making them oversensitive. The presence of a few words with high co-occurrence with a certain score class makes the model associate the essay sample with that score. This causes score changes in \u223c95% of samples with an addition of only a few words. To deal with these issues, we propose detection-based protection models that can detect oversensitivity and samples causing overstability with high accuracies. We find that our proposed models are able to detect unusual attribution patterns and flag adversarial samples successfully."
                    },
                    {
                        "title": "BPGC at SemEval-2020 Task 11: Propaganda Detection in News Articles with Multi-Granularity Knowledge Sharing and Linguistic Features Based Ensemble Learning",
                        "abstract": "Propaganda spreads the ideology and beliefs of like-minded people, brainwashing their audiences, and sometimes leading to violence. SemEval 2020 Task-11 aims to design automated systems for news propaganda detection. Task-11 consists of two sub-tasks, namely, Span Identification - given any news article, the system tags those specific fragments which contain at least one propaganda technique; and Technique Classification - correctly classify a given propagandist statement amongst 14 propaganda techniques. For sub-task 1, we use contextual embeddings extracted from pre-trained transformer models to represent the text data at various granularities and propose a multi-granularity knowledge sharing approach. For sub-task 2, we use an ensemble of BERT and logistic regression classifiers with linguistic features. Our results reveal that the linguistic features are the strong indicators for covering minority classes in a highly imbalanced dataset."
                    },
                    {
                        "title": "M EASURING A ND I MPROVING P ERSUASIVENESS O F G ENERATIVE M ODELS",
                        "abstract": "of the generated content. However, our study challenges this scale-dependent assumption. We propose an instruction fine-tuning approach helping to enhance the persuasiveness of smaller language models, enabling them to surpass much larger models (13-100x) such as GPT-3.5 and GPT-4 (OpenAI, 2023). This finding suggests that persuasive capability is not necessarily a function of model scale and can be achieved through targeted training of smaller language models. We also show that persuasive capability developed in one domain ( e.g. twitter) transfers quite well to other domains ( e.g. websites, debates, and argumentation). This finding can potentially help policymakers like in the recent highly debated California bills on AI models and LLMs (Wiener, 2024; Bauer-Kahan, 2024) and the EU AI act (Union, 2024) that aim to decide appropriate standards for the development and use of AI models and datasets. These legislations try to control models above a certain number of floating point operations. Our findings suggest that simple measures like floating point operations or parameter count do not capture the complete picture of the potential societal implications of AI models, particularly with respect to complex issues like digital persuasion. We discuss more ethical challenges of studying persuasion in LLMs in \u00a7F. Our"
                    }
                ]
            },
            "c21817e5-b43e-449c-9f4f-9d0bf87e87d5": {
                "pk": "c21817e5-b43e-449c-9f4f-9d0bf87e87d5",
                "name": "Uttaran Bhattacharya",
                "collaborators": [
                    "Dinesh Manocha",
                    "Aniket Bera",
                    "Ishita Dasgupta",
                    "Stefano Petrangeli",
                    "P. Guhan",
                    "Xiang Chen",
                    "Saayan Mitra",
                    "Gang Wu",
                    "Nicholas Rewkowski",
                    "Trisha Mittal",
                    "Supreeth Narasimhaswamy",
                    "Aditya Agrawal",
                    "Aanisha Bhattacharyya",
                    "Yaman Kumar Singla",
                    "Somesh Singh",
                    "R. Shah",
                    "Changan Chen",
                    "Balaji Krishnamurthy",
                    "Aneesh Bhattacharya",
                    "Viswanathan Swaminathan",
                    "Rohan Chandra",
                    "Akash Maharaj",
                    "Kun Qian",
                    "Sally Fang",
                    "Horia Galatanu",
                    "Manas Garg",
                    "Rachel Hanessian",
                    "Nishant Kapoor",
                    "Ken Russell",
                    "Shivakumar Vaithyanathan",
                    "Yunyao Li",
                    "Somdeb Sarkhel",
                    "Vahid Azizi",
                    "Minh Hoai",
                    "Yizhou Wang",
                    "Ruiyi Zhang",
                    "Haoliang Wang",
                    "Yun Fu",
                    "Manas Paranjape",
                    "Ashmit Khandelwal",
                    "Tanmay Randhavane",
                    "Pooja Kabra",
                    "Kyra Kapsaskis",
                    "Kurt Gray",
                    "Elizabeth Childs",
                    "A. Banerjee",
                    "Niall L. Williams"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Computer Vision",
                    "Generative Models",
                    "Emotion Recognition"
                ],
                "publications": [
                    {
                        "title": "Speech2UnifiedExpressions: Synchronous Synthesis of Co-Speech Affective Face and Body Expressions from Affordable Inputs",
                        "abstract": "We present a multimodal learning-based method to simultaneously synthesize co-speech facial expressions and upper-body gestures for digital characters using RGB video data captured using commodity cameras. Our approach learns from sparse face landmarks and upper-body joints, estimated directly from video data, to generate plausible emotive character motions. Given a speech audio waveform and a token sequence of the speaker\u2019s face landmark motion and body-joint motion computed from a video, our method synthesizes the motion sequences for the speaker\u2019s face landmarks and body joints to match the content and the affect of the speech. We design a generator consisting of a set of encoders to transform all the inputs into a multimodal embedding space capturing their correlations, followed by a pair of decoders to synthesize the desired face and pose motions. To enhance the plausibility of synthesis, we use an adversarial discriminator that learns to differentiate between the face and pose motions computed from the original videos and our synthesized motions based on their affective expressions. To evaluate our approach, we extend the TED Gesture Dataset to include view-normalized, co-speech face landmarks in addition to body gestures. We demonstrate the performance of our method through thorough quantitative and qualitative experiments on multiple evaluation metrics and via a user study. We observe that our method results in low reconstruction error and produces synthesized samples with diverse facial expressions and body gestures for digital characters. The relevant source code and dataset are available at https://github.com/UttaranB127/speech2unified_expressions."
                    },
                    {
                        "title": "Evaluation and Continual Improvement for an Enterprise AI Assistant",
                        "abstract": "The development of conversational AI assistants is an iterative process with many components involved. As such, the evaluation and continual improvement of these assistants is a complex and multifaceted problem. This paper introduces the challenges in evaluating and improving a generative AI assistant for enterprise that is under active development and how we address these challenges. We also share preliminary results and discuss lessons learned."
                    },
                    {
                        "title": "TAME-RD: Text Assisted Replication of Image Multi-Adjustments for Reverse Designing",
                        "abstract": "Given a source and its edited version performed based on human instructions in natural language, how do we extract the underlying edit operations, to automatically replicate similar edits on other images? This is the problem of reverse designing, and we present TAME-RD, a model to solve this problem.TAME-RD automatically learns from the complex interplay of image editing operations and the natural language instructions to learn fully specified edit operations. It predicts both the underlying image edit operations as discrete categories and their corresponding parameter values in the continuous space. We accomplish this by mapping together the contextual information from the natural language text and the structural differences between the corresponding source and edited images using the concept of pre-post effect. We demonstrate the efficiency of our network through quantitative evaluations on multiple datasets. We observe improvements of 6\u2013 10% on various accuracy metrics and 1.01X-4X on the RMSE score and the concordance correlation coefficient for the corresponding parameter values on the benchmark GIER dataset. We also introduce I-MAD, a new two-part dataset: I-MAD-Dense, a collection of approximately 100K source and edited images, together with automatically generated text instructions and annotated"
                    },
                    {
                        "title": "HanDiffuser: Text-to-Image Generation with Realistic Hand Appearances",
                        "abstract": "Text-to-image generative models can generate high-quality humans, but realism is lost when generating hands. Common artifacts include irregular hand poses, shapes, incorrect numbers of fingers, and physically implausible finger orientations. To generate images with realistic hands, we propose a novel diffusion-based architecture called HanDiffuser that achieves realism by injecting hand embeddings in the generative process. HanDiffuser consists of two components: a Text-to-Hand-Params diffusion model to generate SMPL-Body and MANO-Hand parameters from input text prompts, and a Text-Guided Hand-Params-to-Image diffusion model to synthesize images by conditioning on the prompts and hand parameters generated by the previous component. We incorporate multiple aspects of hand representation, including 3D shapes and joint-level finger positions, orientations and articulations, for robust learning and reliable performance during inference. We conduct extensive quantitative and qualitative experiments and perform user studies to demonstrate the efficacy of our method in generating images with high-quality hands. Project page: https://supreethn.github.io/research/handiffuser/index.html"
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. \u22c6 Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 1 [ cs .C L ] 1 S ep 2 02 3 Published as a conference paper at ICLR 2023"
                    },
                    {
                        "title": "Text-to-Hand-Image Generation Using Pose-and Mesh-Guided Diffusion",
                        "abstract": "Text-to-image generative models like Stable Diffusion can generate high-quality humans, but realism is lost when generating hands. The generated hands often have artifacts such as irregular hand poses, shapes, incorrect numbers of fingers, and physically implausible finger orientations. To overcome these issues, we propose to generate high-quality hands by injecting hand structure into a Stable Diffusion model using a ControlNet architecture. Specifically, our method uses rigid 2D hand poses and deformable 3D hand meshes to control the quality of generated hands. Our experiments show that using hand structure information during training enables the generation of realistic hand images. We also extend our method to generate plausible hands as part of full-body images."
                    },
                    {
                        "title": "DanceAnyWay: Synthesizing Mixed-Genre 3D Dance Movements Through Beat Disentanglement",
                        "abstract": "We present DanceAnyWay, a hierarchical generative adversarial learning method to synthesize mixed-genre dance movements of 3D human characters synchronized with music. Our method learns to disentangle the dance movements at the beat frames from the dance movements at all the remaining frames by operating at two hierarchical levels. At the coarser \u201cbeat\u201d level, it encodes the rhythm, pitch, and melody information of the input music via dedicated feature representations only at the beat frames. It leverages them to synthesize the beat poses of the target dance using a sequence-to-sequence learning framework. At the finer \u201cre-pletion\u201d level, our method encodes similar rhythm, pitch, and melody information from all the frames of the input music via dedicated feature representations and couples them with the synthesized beat poses from the coarser level to synthesize the full target dance sequence using an adversarial learning framework. By disentangling the broader dancing styles at the coarser level from the specific dance move-*"
                    },
                    {
                        "title": "VaQuitA: Enhancing Alignment in LLM-Assisted Video Understanding",
                        "abstract": "Recent advancements in language-model-based video understanding have been progressing at a remarkable pace, spurred by the introduction of Large Language Models (LLMs). However, the focus of prior research has been predominantly on devising a projection layer that maps video features to tokens, an approach that is both rudimentary and inefficient. In our study, we introduce a cutting-edge framework, VaQuitA, designed to refine the synergy between video and textual information. At the data level, instead of sampling frames uniformly, we implement a sampling method guided by CLIP-score rankings, which enables a more aligned selection of frames with the given question. At the feature level, we integrate a trainable Video Perceiver alongside a Visual-Query Transformer (abbreviated as VQ-Former), which bolsters the interplay between the input question and the video features. We also discover that incorporating a simple prompt,\"Please be critical\", into the LLM input can substantially enhance its video comprehension capabilities. Our experimental results indicate that VaQuitA consistently sets a new benchmark for zero-shot video question-answering tasks and is adept at producing high-quality, multi-turn video dialogues with users."
                    },
                    {
                        "title": "DanceAnyWay: Synthesizing Beat-Guided 3D Dances with Randomized Temporal Contrastive Learning",
                        "abstract": "We present DanceAnyWay, a generative learning method to synthesize beat-guided dances of 3D human characters synchronized with music. Our method learns to disentangle the dance movements at the beat frames from the dance movements at all the remaining frames by operating at two hierarchical levels. At the coarser \"beat\" level, it encodes the rhythm, pitch, and melody information of the input music via dedicated feature representations only at the beat frames. It leverages them to synthesize the beat poses of the target dances using a sequence-to-sequence learning framework. At the finer \"repletion\" level, our method encodes similar rhythm, pitch, and melody information from all the frames of the input music via dedicated feature representations. It generates the full dance sequences by combining the synthesized beat and repletion poses and enforcing plausibility through an adversarial learning framework. Our training paradigm also enforces fine-grained diversity in the synthesized dances through a randomized temporal contrastive loss, which ensures different segments of the dance sequences have different movements and avoids motion freezing or collapsing to repetitive movements. We evaluate the performance of our approach through extensive experiments on the benchmark AIST++ dataset and observe improvements of about 7%-12% in motion quality metrics and 1.5%-4% in motion diversity metrics over the current baselines, respectively. We also conducted a user study to evaluate the visual quality of our synthesized dances. We noted that, on average, the samples generated by our method were about 9-48% more preferred by the participants and had a 4-27% better five-point Likert-scale score over the best available current baseline in terms of motion quality and synchronization. Our source code and project page are available at https://github.com/aneeshbhattacharya/DanceAnyWay."
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 2 [ cs .C L ] 8 S ep 2 02 3"
                    },
                    {
                        "title": "Learning Gait Emotions Using Affective and Deep Features",
                        "abstract": "We present a novel data-driven algorithm to learn the perceived emotions of individuals based on their walking motion or gaits. Given an RGB video of an individual walking, we extract their walking gait as a sequence of 3D poses. Our goal is to exploit the gait features to learn and model the emotional state of the individual into one of four categorical emotions: happy, sad, angry, or neutral. Our perceived emotion identification approach uses deep features learned using long short-term memory networks (LSTMs) on datasets with labeled emotive gaits. We combine these features with gait-based affective features consisting of posture and movement measures. Our algorithm identifies both the categorical emotions from the gaits and the corresponding values for the dimensional emotion components - valence and arousal. We also introduce and benchmark a dataset called Emotion Walk (EWalk), consisting of videos of gaits of individuals annotated with emotions. We show that our algorithm mapping the combined feature space to the perceived emotional state provides an accuracy of 80.07% on the EWalk dataset, outperforming the current baselines by an absolute 13\u201324%."
                    },
                    {
                        "title": "Show Me What I Like: Detecting User-Specific Video Highlights Using Content-Based Multi-Head Attention",
                        "abstract": "We propose a method to detect individualized highlights for users on given target videos based on their preferred highlight clips marked on previous videos they have watched. Our method explicitly leverages the contents of both the preferred clips and the target videos using pre-trained features for the objects and the human activities. We design a multi-head attention mechanism to adaptively weigh the preferred clips based on their object- and human-activity-based contents, and fuse them using these weights into a single feature representation for each user. We compute similarities between these per-user feature representations and the per-frame features computed from the desired target videos to estimate the user-specific highlight clips from the target videos. We test our method on a large-scale highlight detection dataset containing the annotated highlights of individual users. Compared to current baselines, we observe an absolute improvement of 2-4% in the mean average precision of the detected highlights. We also perform extensive ablation experiments on the number of preferred highlight clips associated with each user as well as on the object- and human-activity-based feature representations to validate that our method is indeed both content-based and user-specific."
                    },
                    {
                        "title": "Speech2AffectiveGestures: Synthesizing Co-Speech Gestures with Generative Adversarial Affective Expression Learning",
                        "abstract": "We present a generative adversarial network to synthesize 3D pose sequences of co-speech upper-body gestures with appropriate affective expressions. Our network consists of two components: a generator to synthesize gestures from a joint embedding space of features encoded from the input speech and the seed poses, and a discriminator to distinguish between the synthesized pose sequences and real 3D pose sequences. We leverage the Mel-frequency cepstral coefficients and the text transcript computed from the input speech in separate encoders in our generator to learn the desired sentiments and the associated affective cues. We design an affective encoder using multi-scale spatial-temporal graph convolutions to transform 3D pose sequences into latent, pose-based affective features. We use our affective encoder in both our generator, where it learns affective features from the seed poses to guide the gesture synthesis, and our discriminator, where it enforces the synthesized gestures to contain the appropriate affective expressions. We perform extensive evaluations on two benchmark datasets for gesture synthesis from the speech, the TED Gesture Dataset and the GENEA Challenge 2020 Dataset. Compared to the best baselines, we improve the mean absolute joint error by 10-33%, the mean acceleration difference by 8-58%, and the Fr\u00e9chet Gesture Distance by 21-34%. We also conduct a user study and observe that compared to the best current baselines, around 15.28% of participants indicated our synthesized gestures appear more plausible, and around 16.32% of participants felt the gestures had more appropriate affective expressions aligned with the speech."
                    },
                    {
                        "title": "Text2Gestures: A Transformer-Based Network for Generating Emotive Body Gestures for Virtual Agents**This work has been supported in part by ARO Grants W911NF1910069 and W911NF1910315, and Intel. Code and additional materials available at: https://gamma.umd.edu/t2g",
                        "abstract": "We present Text2Gestures, a transformer-based learning method to interactively generate emotive full-body gestures for virtual agents aligned with natural language text inputs. Our method generates emotionally expressive gestures by utilizing the relevant biomechanical features for body expressions, also known as affective features. We also consider the intended task corresponding to the text and the target virtual agents' intended gender and handedness in our generation pipeline. We train and evaluate our network on the MPI Emotional Body Expressions Database and observe that our network produces state-of-the-art performance in generating gestures for virtual agents aligned with the text for narration or conversation. Our network can generate these gestures at interactive rates on a commodity GPU. We conduct a web-based user study and observe that around 91% of participants indicated our generated gestures to be at least plausible on a five-point Likert Scale. The emotions perceived by the participants from the gestures are also strongly positively correlated with the corresponding intended emotions, with a minimum Pearson coefficient of 0.77 in the valence dimension."
                    },
                    {
                        "title": "HighlightMe: Detecting Highlights from Human-Centric Videos",
                        "abstract": "We present a domain- and user-preference-agnostic approach to detect highlightable excerpts from human-centric videos. Our method works on the graph-based representation of multiple observable human-centric modalities in the videos, such as poses and faces. We use an autoencoder network equipped with spatial-temporal graph convolutions to detect human activities and interactions based on these modalities. We train our network to map the activity- and interaction-based latent structural representations of the different modalities to per-frame highlight scores based on the representativeness of the frames. We use these scores to compute which frames to highlight and stitch contiguous frames to produce the excerpts. We train our network on the large-scale AVA-Kinetics action dataset and evaluate it on four benchmark video highlight datasets: DSH, TVSum, PHD2, and SumMe. We observe a 4\u201312% improvement in the mean average precision of matching the human-annotated highlights over state-of-the-art methods in these datasets, without requiring any user-provided preferences or dataset-specific fine-tuning."
                    },
                    {
                        "title": "Generating Emotive Gaits for Virtual Agents Using Affect-Based Autoregression",
                        "abstract": "We present a novel autoregression network to generate virtual agents that convey various emotions through their walking styles or gaits. Given the 3D pose sequences of a gait, our network extracts pertinent movement features and affective features from the gait. We use these features to synthesize subsequent gaits such that the virtual agents can express and transition between emotions represented as combinations of happy, sad, angry, and neutral. We incorporate multiple regularizations in the training of our network to simultaneously enforce plausible movements and noticeable emotions on the virtual agents. We also integrate our approach with an AR environment using a Microsoft HoloLens and can generate emotive gaits at interactive rates to increase the social presence. We evaluate how human observers perceive both the naturalness and the emotions from the generated gaits of the virtual agents in a web-based study. Our results indicate around 89% of the users found the naturalness of the gaits satisfactory on a five-point Likert scale, and the emotions they perceived from the virtual agents are statistically similar to the intended emotions of the virtual agents. We also use our network to augment existing gait datasets with emotive gaits and will release this augmented dataset for future research in emotion prediction and emotive gait synthesis. Our project website is available at https://gamma.umd.edu/gen-emotive-gaits/."
                    },
                    {
                        "title": "CMetric: A Driving Behavior Measure using Centrality Functions",
                        "abstract": "We present a new measure, CMetric, to classify driver behaviors using centrality functions. Our formulation combines concepts from computational graph theory and social traffic psychology to quantify and classify the behavior of human drivers. CMetric is used to compute the probability of a vehicle executing a driving style, as well as the intensity used to execute the style. Our approach is designed for realtime autonomous driving applications, where the trajectory of each vehicle or road-agent is extracted from a video. We compute a dynamic geometric graph (DGG) based on the positions and proximity of the road-agents and centrality functions corresponding to closeness and degree. These functions are used to compute the CMetric based on style likelihood and style intensity estimates. Our approach is general and makes no assumption about traffic density, heterogeneity, or how driving behaviors change over time. We present an algorithm to compute CMetric and demonstrate its performance on real-world traffic datasets. To test the accuracy of CMetric, we introduce a new evaluation protocol (called \"Time Deviation Error\") that measures the difference between human prediction and the prediction made by CMetric."
                    },
                    {
                        "title": "Emotions Don't Lie: An Audio-Visual Deepfake Detection Method using Affective Cues",
                        "abstract": "We present a learning-based method for detecting real and fake deepfake multimedia content. To maximize information for learning, we extract and analyze the similarity between the two audio and visual modalities from within the same video. Additionally, we extract and compare affective cues corresponding to perceived emotion from the two modalities within a video to infer whether the input video is \"real\" or \"fake\". We propose a deep learning network, inspired by the Siamese network architecture and the triplet loss. To validate our model, we report the AUC metric on two large-scale deepfake detection datasets, DeepFake-TIMIT Dataset and DFDC. We compare our approach with several SOTA deepfake detection methods and report per-video AUC of 84.4% on the DFDC and 96.6% on the DF-TIMIT datasets, respectively. To the best of our knowledge, ours is the first approach that simultaneously exploits audio and video modalities and also perceived emotions from the two modalities for deepfake detection."
                    }
                ]
            },
            "3f010411-b4c8-4ba6-8c98-831b5369879e": {
                "pk": "3f010411-b4c8-4ba6-8c98-831b5369879e",
                "name": "Ishita Dasgupta",
                "collaborators": [
                    "Uttaran Bhattacharya",
                    "Supreeth Narasimhaswamy",
                    "Xiang Chen",
                    "Saayan Mitra",
                    "Aditya Agrawal",
                    "Aanisha Bhattacharyya",
                    "Yaman Kumar Singla",
                    "Somesh Singh",
                    "Stefano Petrangeli",
                    "R. Shah",
                    "Changan Chen",
                    "Balaji Krishnamurthy",
                    "Minh Hoai",
                    "Ashmit Khandelwal"
                ],
                "domain": [
                    "Generative Models",
                    "Natural Language Processing",
                    "Computer Vision",
                    "Human-Computer Interaction"
                ],
                "publications": [
                    {
                        "title": "HanDiffuser: Text-to-Image Generation with Realistic Hand Appearances",
                        "abstract": "Text-to-image generative models can generate high-quality humans, but realism is lost when generating hands. Common artifacts include irregular hand poses, shapes, incorrect numbers of fingers, and physically implausible finger orientations. To generate images with realistic hands, we propose a novel diffusion-based architecture called HanDiffuser that achieves realism by injecting hand embeddings in the generative process. HanDiffuser consists of two components: a Text-to-Hand-Params diffusion model to generate SMPL-Body and MANO-Hand parameters from input text prompts, and a Text-Guided Hand-Params-to-Image diffusion model to synthesize images by conditioning on the prompts and hand parameters generated by the previous component. We incorporate multiple aspects of hand representation, including 3D shapes and joint-level finger positions, orientations and articulations, for robust learning and reliable performance during inference. We conduct extensive quantitative and qualitative experiments and perform user studies to demonstrate the efficacy of our method in generating images with high-quality hands. Project page: https://supreethn.github.io/research/handiffuser/index.html"
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. \u22c6 Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 1 [ cs .C L ] 1 S ep 2 02 3 Published as a conference paper at ICLR 2023"
                    },
                    {
                        "title": "Text-to-Hand-Image Generation Using Pose-and Mesh-Guided Diffusion",
                        "abstract": "Text-to-image generative models like Stable Diffusion can generate high-quality humans, but realism is lost when generating hands. The generated hands often have artifacts such as irregular hand poses, shapes, incorrect numbers of fingers, and physically implausible finger orientations. To overcome these issues, we propose to generate high-quality hands by injecting hand structure into a Stable Diffusion model using a ControlNet architecture. Specifically, our method uses rigid 2D hand poses and deformable 3D hand meshes to control the quality of generated hands. Our experiments show that using hand structure information during training enables the generation of realistic hand images. We also extend our method to generate plausible hands as part of full-body images."
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 2 [ cs .C L ] 8 S ep 2 02 3"
                    }
                ]
            },
            "bc6fda7f-e8df-4b79-be0a-6ef5710eb079": {
                "pk": "bc6fda7f-e8df-4b79-be0a-6ef5710eb079",
                "name": "Stefano Petrangeli",
                "collaborators": [
                    "Aditya Agrawal",
                    "Aanisha Bhattacharyya",
                    "Yaman Kumar Singla",
                    "Somesh Singh",
                    "Uttaran Bhattacharya",
                    "Ishita Dasgupta",
                    "R. Shah",
                    "Changan Chen",
                    "Balaji Krishnamurthy",
                    "Ashmit Khandelwal"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Communication Theory",
                    "Behavior Modeling"
                ],
                "publications": [
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. \u22c6 Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 1 [ cs .C L ] 1 S ep 2 02 3 Published as a conference paper at ICLR 2023"
                    },
                    {
                        "title": "L ARGE C ONTENT AND B EHAVIOR M ODELS TO U N - DERSTAND , S IMULATE , AND O PTIMIZE C ONTENT AND B EHAVIOR",
                        "abstract": "Shannon, in his seminal paper introducing information theory, divided the communication into three levels: technical, semantic, and effectivenss. While the technical level is concerned with accurate reconstruction of transmitted symbols, the semantic and effectiveness levels deal with the inferred meaning and its effect on the receiver. Thanks to telecommunications, the first level problem has produced great advances like the internet. Large Language Models (LLMs) make some progress towards the second goal, but the third level still remains largely untouched. The third problem deals with predicting and optimizing communication for desired receiver behavior. LLMs, while showing wide generalization capabilities across a wide range of tasks, are unable to solve for this. One reason for the underperformance could be a lack of \u201cbehavior tokens\u201d in LLMs\u2019 training corpora. Behavior tokens define receiver behavior over a communication, such as shares, likes, clicks, purchases, retweets, etc. While preprocessing data for LLM training, behavior tokens are often removed from the corpora as noise. Therefore, in this paper, we make some initial progress towards reintroducing behavior tokens in LLM training. The trained models, other than showing similar performance to LLMs on content understanding tasks, show generalization capabilities on behavior simulation, content simulation, behavior understanding, and behavior domain adaptation. Using a wide range of tasks on two corpora, we show results on all these capabilities. We call these models Large Content and Behavior Models (LCBMs). Further, to spur more research on LCBMs, we release our new Content Behavior Corpus (CBC), a repository containing communicator, message, and corresponding receiver behavior. Equal Contribution. Contact ykumar@adobe.com for questions and suggestions. 1 ar X iv :2 30 9. 00 35 9v 2 [ cs .C L ] 8 S ep 2 02 3"
                    }
                ]
            },
            "907c45cb-83fd-4fd2-8f8b-2be2287a2747": {
                "pk": "907c45cb-83fd-4fd2-8f8b-2be2287a2747",
                "name": "Rajiv Ratn Shah",
                "collaborators": [
                    "Avinash Anand",
                    "Ashish Gudmalwar",
                    "Pankaj Wasnik",
                    "P. Kumaraguru",
                    "Nirmesh Shah",
                    "Ritwik Mishra",
                    "Kritarth Prasad",
                    "Astha Verma",
                    "Roger Zimmermann",
                    "Naman Lal",
                    "Sai Akarsh",
                    "Mayank Kharbanda",
                    "Raghava Mutharaju",
                    "Sreeram Vennam",
                    "Shivam Mhaskar",
                    "Nirmesh J. Shah",
                    "Mohammadi Zaki",
                    "Hitkul Jangra",
                    "Neha Sahipjohn",
                    "Pooja Desur",
                    "Mohit Gupta",
                    "Navya Singla",
                    "Sanjana Sanjeev",
                    "Jatin Kumar",
                    "A. Shivam",
                    "Raj Jaiswal",
                    "Abhishek Dharmadhikari",
                    "Atharva Marathe",
                    "Harsh Popat",
                    "Harshil Mital",
                    "Ashwin R. Nair",
                    "Sidharth Kumar",
                    "Avni Mittal",
                    "Laavanaya Dhawan",
                    "Juhi Krishnamurthy",
                    "Mahisha Ramesh",
                    "Pijush Bhuyan",
                    "Himani",
                    "Shin'ichi Satoh",
                    "Hemant Yadav",
                    "Sunayana Sitaram",
                    "Debnath Kundu",
                    "Atharva Mehta",
                    "Rajesh Kumar",
                    "Apoorv Singh",
                    "Somesh Singh",
                    "I. HariniS",
                    "Yaman Kumar Singla",
                    "V. Baths",
                    "Changyou Chen",
                    "Balaji Krishnamurthy"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Text-to-Speech",
                    "Knowledge Graph",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "VECL-TTS: Voice identity and Emotional style controllable Cross-Lingual Text-to-Speech",
                        "abstract": "Despite the significant advancements in Text-to-Speech (TTS) systems, their full utilization in automatic dubbing remains limited. This task necessitates the extraction of voice identity and emotional style from a reference speech in a source language and subsequently transferring them to a target language using cross-lingual TTS techniques. While previous approaches have mainly concentrated on controlling voice identity within the cross-lingual TTS framework, there has been limited work on incorporating emotion and voice identity together. To this end, we introduce an end-to-end Voice Identity and Emotional Style Controllable Cross-Lingual (VECL) TTS system using multilingual speakers and an emotion embedding network. Moreover, we introduce content and style consistency losses to enhance the quality of synthesized speech further. The proposed system achieved an average relative improvement of 8.83\\% compared to the state-of-the-art (SOTA) methods on a database comprising English and three Indian languages (Hindi, Telugu, and Marathi)."
                    },
                    {
                        "title": "RConE: Rough Cone Embedding for Multi-Hop Logical Query Answering on Multi-Modal Knowledge Graphs",
                        "abstract": "Multi-hop query answering over a Knowledge Graph (KG) involves traversing one or more hops from the start node to answer a query. Path-based and logic-based methods are state-of-the-art for multi-hop question answering. The former is used in link prediction tasks. The latter is for answering complex logical queries. The logical multi-hop querying technique embeds the KG and queries in the same embedding space. The existing work incorporates First Order Logic (FOL) operators, such as conjunction ($\\wedge$), disjunction ($\\vee$), and negation ($\\neg$), in queries. Though current models have most of the building blocks to execute the FOL queries, they cannot use the dense information of multi-modal entities in the case of Multi-Modal Knowledge Graphs (MMKGs). We propose RConE, an embedding method to capture the multi-modal information needed to answer a query. The model first shortlists candidate (multi-modal) entities containing the answer. It then finds the solution (sub-entities) within those entities. Several existing works tackle path-based question-answering in MMKGs. However, to our knowledge, we are the first to introduce logical constructs in querying MMKGs and to answer queries that involve sub-entities of multi-modal entities as the answer. Extensive evaluation of four publicly available MMKGs indicates that RConE outperforms the current state-of-the-art."
                    },
                    {
                        "title": "Multilingual Non-Factoid Question Answering with Silver Answers",
                        "abstract": "Most existing Question Answering Datasets (QuADs) primarily focus on factoid-based short-context Question Answering (QA) in high-resource languages. However, the scope of such datasets for low-resource languages remains limited, with only a few works centered on factoid-based QuADs and none on non-factoid QuADs. Therefore, this work presents MuNfQuAD, a multilingual QuAD with non-factoid questions. It utilizes interrogative sub-headings from BBC news articles as questions and the corresponding paragraphs as silver answers. The dataset comprises over 370K QA pairs across 38 languages, encompassing several low-resource languages, and stands as the largest multilingual QA dataset to date. Based on the manual annotations of 790 QA-pairs from MuNfQuAD (golden set), we observe that 98\\% of questions can be answered using their corresponding silver answer. Our fine-tuned Answer Paragraph Selection (APS) model outperforms the baselines. The APS model attained an accuracy of 80\\% and 72\\%, as well as a macro F1 of 72\\% and 66\\%, on the MuNfQuAD testset and the golden set, respectively. Furthermore, the APS model effectively generalizes certain a language within the golden set, even after being fine-tuned on silver labels."
                    },
                    {
                        "title": "Isometric Neural Machine Translation using Phoneme Count Ratio Reward-based Reinforcement Learning",
                        "abstract": "Traditional Automatic Video Dubbing (AVD) pipeline consists of three key modules, namely, Automatic Speech Recognition (ASR), Neural Machine Translation (NMT), and Text-to-Speech (TTS). Within AVD pipelines, isometric-NMT algorithms are employed to regulate the length of the synthesized output text. This is done to guarantee synchronization with respect to the alignment of video and audio subsequent to the dubbing process. Previous approaches have focused on aligning the number of characters and words in the source and target language texts of Machine Translation models. However, our approach aims to align the number of phonemes instead, as they are closely associated with speech duration. In this paper, we present the development of an isometric NMT system using Reinforcement Learning (RL), with a focus on optimizing the alignment of phoneme counts in the source and target language sentence pairs. To evaluate our models, we propose the Phoneme Count Compliance (PCC) score, which is a measure of length compliance. Our approach demonstrates a substantial improvement of approximately 36% in the PCC score compared to the state-of-the-art models when applied to English-Hindi language pairs. Moreover, we propose a student-teacher architecture within the framework of our RL approach to maintain a trade-off between the phoneme count and translation quality."
                    },
                    {
                        "title": "Put Your Money Where Your Mouth Is: Dataset and Analysis of Real World Habit Building Attempts",
                        "abstract": "The pursuit of habit building is challenging, and most people struggle with it. Research on successful habit formation is mainly based on small human trials focusing on the same habit for all the participants as conducting long-term heterogonous habit studies can be logistically expensive. With the advent of self-help, there has been an increase in online communities and applications that are centered around habit building and logging. Habit building applications can provide large-scale data on real-world habit building attempts and unveil the commonalities among successful ones. We collect public data on stickk.com, which allows users to track progress on habit building attempts called commitments. A commitment can have an external referee, regular check-ins about the progress, and a monetary stake in case of failure. Our data consists of 742,923 users and 397,456 commitments. In addition to the dataset, rooted in theories like Fresh Start Effect, Accountablity, and Loss Aversion, we ask questions about how commitment properties like start date, external accountability, monitory stake, and pursuing multiple habits together affects the odds of success. We found that people tend to start habits on temporal landmarks, but that does not affect the probability of their success. Practices like accountability and stakes are not often used but are strong determents of success. Commitments of 6 to 8 weeks in length, weekly reporting with an external referee, and a monetary amount at stake tend to be most successful. Finally, around 40% of all commitments are attempted simultaneously with other goals. Simultaneous attempts of pursuing commitments may fail early, but if pursued through the initial phase, they are statistically more successful than building one habit at a time."
                    },
                    {
                        "title": "DubWise: Video-Guided Speech Duration Control in Multimodal LLM-based Text-to-Speech for Dubbing",
                        "abstract": "Audio-visual alignment after dubbing is a challenging research problem. To this end, we propose a novel method, DubWise Multi-modal Large Language Model (LLM)-based Text-to-Speech (TTS), which can control the speech duration of synthesized speech in such a way that it aligns well with the speakers lip movements given in the reference video even when the spoken text is different or in a different language. To accomplish this, we propose to utilize cross-modal attention techniques in a pre-trained GPT-based TTS. We combine linguistic tokens from text, speaker identity tokens via a voice cloning network, and video tokens via a proposed duration controller network. We demonstrate the effectiveness of our system on the Lip2Wav-Chemistry and LRS2 datasets. Also, the proposed method achieves improved lip sync and naturalness compared to the SOTAs for the same language but different text (i.e., non-parallel) and the different language, different text (i.e., cross-lingual) scenarios."
                    },
                    {
                        "title": "Multilingual Coreference Resolution in Low-resource South Asian Languages",
                        "abstract": "Coreference resolution involves the task of identifying text spans within a discourse that pertain to the same real-world entity. While this task has been extensively explored in the English language, there has been a notable scarcity of publicly accessible resources and models for coreference resolution in South Asian languages. We introduce a Translated dataset for Multilingual Coreference Resolution (TransMuCoRes) in 31 South Asian languages using off-the-shelf tools for translation and word-alignment. Nearly all of the predicted translations successfully pass a sanity check, and 75% of English references align with their predicted translations. Using multilingual encoders, two off-the-shelf coreference resolution models were trained on a concatenation of TransMuCoRes and a Hindi coreference resolution dataset with manual annotations. The best performing model achieved a score of 64 and 68 for LEA F1 and CoNLL F1, respectively, on our test-split of Hindi golden set. This study is the first to evaluate an end-to-end coreference resolution model on a Hindi golden set. Furthermore, this work underscores the limitations of current coreference evaluation metrics when applied to datasets with split antecedents, advocating for the development of more suitable evaluation metrics."
                    },
                    {
                        "title": "Mathify: Evaluating Large Language Models on Mathematical Problem Solving Tasks",
                        "abstract": "The rapid progress in the field of natural language processing (NLP) systems and the expansion of large language models (LLMs) have opened up numerous opportunities in the field of education and instructional methods. These advancements offer the potential for tailored learning experiences and immediate feedback, all delivered through accessible and cost-effective services. One notable application area for this technological advancement is in the realm of solving mathematical problems. Mathematical problem-solving not only requires the ability to decipher complex problem statements but also the skill to perform precise arithmetic calculations at each step of the problem-solving process. However, the evaluation of the arithmetic capabilities of large language models remains an area that has received relatively little attention. In response, we introduce an extensive mathematics dataset called\"MathQuest\"sourced from the 11th and 12th standard Mathematics NCERT textbooks. This dataset encompasses mathematical challenges of varying complexity and covers a wide range of mathematical concepts. Utilizing this dataset, we conduct fine-tuning experiments with three prominent LLMs: LLaMA-2, WizardMath, and MAmmoTH. These fine-tuned models serve as benchmarks for evaluating their performance on our dataset. Our experiments reveal that among the three models, MAmmoTH-13B emerges as the most proficient, achieving the highest level of competence in solving the presented mathematical problems. Consequently, MAmmoTH-13B establishes itself as a robust and dependable benchmark for addressing NCERT mathematics problems."
                    },
                    {
                        "title": "GeoVQA: A Comprehensive Multimodal Geometry Dataset for Secondary Education",
                        "abstract": "Large language models (LLMs) have exhibited impressive capabilities in human-like reasoning and generation, prompting widespread exploration of their utility in mathematical problem-solving. However, existing research predominantly focuses on text-based mathematical challenges, neglecting those involving geometric concepts. To bridge this gap, our research aims to empower LLMs to tackle geometric problems by integrating image understanding. We investigate the limitations of current Multimodal Large Language Models (MLLMs) in this context, particularly their struggle to accurately interpret complex geometric elements and their mathematical reasoning. To support this endeavor, we introduce GeoVQA, a specialized multimodal math dataset tailored for secondary school education. Furthermore, we conduct experiments to Instruction-tune LLaVA [14] and G-LLaVA [11] models, evaluating their performance against benchmarks such as PGPS9K [21] & Geometry3k [15]. Additionally, we compare the effectiveness of Gemini pro [19] and GPT4 [1] in handling geometric tasks using various prompting techniques, employing metrics like accuracy and human evaluation for meticulous verification. GPT-4 has the best performance on zero-shot inference on our dataset. LLaVA-v1.5-13B model finetuned on our dataset outperforms GPT-4 on PGPS9K benchmark. GitHub: GeoVQA."
                    },
                    {
                        "title": "ExCEDA: Unlocking Attention Paradigms in Extended Duration E-Classrooms by Leveraging Attention-Mechanism Models",
                        "abstract": "Learner engagement wields considerable influence over educational outcomes. However, quantifying engagement within online learning contexts presents a persistent challenge, often exacerbated by the limitations of existing datasets. Existing datasets, with their focus on short-duration videos, hinder the comprehensive assessment of student attention over the extended periods characteristic of typical classroom settings. To address this research gap, we present the ExCEDA -Extended Classroom Engagement Dataset for Assessment. ExCEDA encompasses data from 224 participants, consisting of 6143 images annotated for four crucial affective states: boredom, engagement, confusion, and frustration. Collected during a 50-minute online lecture, ExCEDA uniquely facilitates in-depth analysis of how learner attention fluctuates throughout extended learning periods. We benchmark ExCEDA utilizing state-of-the-art classification models (EfficientNet, Residual Attention Network, and GLAMOR-net). Notably, combining GLAMOR-Net with Facial Action Unit features yields results on par with video-centric datasets. We demonstrate the effectiveness of our approach of combining GLAMOR-Net with Facial features by surpassing the engagement prediction baseline for the EngageNet dataset. Additionally, we introduce CG-ViT, a novel content-guided ViT model, which exhibits substantial performance gains (21.45 %, 9.37 %, 15.18 %, and 13.93 % accuracy increases for engagement, boredom, confusion, and frustration, respectively) over the baseline ViT model on our dataset. We also introduce the concept of dataset personalization that enhanced the predictive accuracy of multi-class classification of Confusion, Boredom, Engagement, and Frustration by 17 %, 15 %, 6 %, and 20 % respectively. Our novel dataset will be made publicly available."
                    },
                    {
                        "title": "JOOCI: a Framework for Learning Comprehensive Speech Representations",
                        "abstract": "Information in speech can be divided into two categories: what is being said (content) and how it is expressed (other). Current state-of-the-art (SOTA) techniques model speech at fixed segments, usually 10-25 ms, using a single embedding. Given the orthogonal nature of other and content information, attempting to optimize both within a single embedding results in suboptimal solutions. This approach divides the models capacity, limiting its ability to build complex hierarchical features effectively. In this work, we present an end-to-end speech representation learning framework designed to jointly optimize the other and content information (JOOCI) in speech. By using separate learnable parameters, JOOCI addresses this optimization challenge by modeling other and content information independently. Our results show that JOOCI consistently outperforms other SOTA models of similar size (100 million parameters) and pre-training data used (960 hours) by a significant margin when evaluated on a range of speech downstream tasks in the SUPERB benchmark, as shown in Table 1."
                    },
                    {
                        "title": "Keystroke Dynamics Against Academic Dishonesty in the Age of LLMs",
                        "abstract": "The transition to online examinations and assignments raises significant concerns about academic integrity. Traditional plagiarism detection systems often struggle to identify instances of intelligent cheating, particularly when students utilize advanced generative AI tools to craft their responses. This study proposes a keystroke dynamics-based method to differentiate between bona fide and assisted writing within academic contexts. To facilitate this, a dataset was developed to capture the keystroke patterns of individuals engaged in writing tasks, both with and without the assistance of generative AI. The detector, trained using a modified TypeNet architecture, achieved accuracies ranging from 74.98% to 85.72% in condition-specific scenarios and from 52.24% to 80.54% in condition-agnostic scenarios. The findings highlight significant differences in keystroke dynamics between genuine and assisted writing. The outcomes of this study enhance our understanding of how users interact with generative AI and have implications for improving the reliability of digital educational platforms."
                    },
                    {
                        "title": "Teaching Human Behavior Improves Content Understanding Abilities Of LLMs",
                        "abstract": "Communication is defined as\"Who says what to whom with what effect\". A message from a communicator generates downstream receiver effects, also known as behavior. Receiver behavior, being a downstream effect of the message, carries rich signals about it. Even after carrying signals about the message, the behavior data is often ignored while training large language models. We show that training LLMs on receiver behavior can actually help improve their content-understanding abilities. Specifically, we show that training LLMs to predict the receiver behavior of likes and comments improves the LLM's performance on a wide variety of downstream content understanding tasks. We show this performance increase over 46 video and image understanding tasks over 26 benchmark datasets across both 0-shot and fine-tuning settings, outperforming many supervised baselines. Moreover, since receiver behavior, such as likes and comments, is collected by default on the internet and does not need any human annotations to be useful, the performance improvement we get after training on this data is essentially free-lunch. We release the receiver behavior cleaned comments and likes of 750k images and videos collected from multiple platforms along with our instruction-tuning data."
                    }
                ]
            },
            "55a8f582-eee2-4681-9c3e-6c4d0d49967a": {
                "pk": "55a8f582-eee2-4681-9c3e-6c4d0d49967a",
                "name": "Changyou Chen",
                "collaborators": [
                    "Yaman Kumar Singla",
                    "R. Shah",
                    "Balaji Krishnamurthy",
                    "Somesh Singh",
                    "I. HariniS",
                    "V. Baths",
                    "Aanisha Bhattacharya",
                    "Varun Khurana",
                    "J. Subramanian",
                    "Zhiqiang Xu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Multimedia Understanding",
                    "Image Generation",
                    "Behavioral Analysis"
                ],
                "publications": [
                    {
                        "title": "Teaching Human Behavior Improves Content Understanding Abilities Of LLMs",
                        "abstract": "Communication is defined as\"Who says what to whom with what effect\". A message from a communicator generates downstream receiver effects, also known as behavior. Receiver behavior, being a downstream effect of the message, carries rich signals about it. Even after carrying signals about the message, the behavior data is often ignored while training large language models. We show that training LLMs on receiver behavior can actually help improve their content-understanding abilities. Specifically, we show that training LLMs to predict the receiver behavior of likes and comments improves the LLM's performance on a wide variety of downstream content understanding tasks. We show this performance increase over 46 video and image understanding tasks over 26 benchmark datasets across both 0-shot and fine-tuning settings, outperforming many supervised baselines. Moreover, since receiver behavior, such as likes and comments, is collected by default on the internet and does not need any human annotations to be useful, the performance improvement we get after training on this data is essentially free-lunch. We release the receiver behavior cleaned comments and likes of 750k images and videos collected from multiple platforms along with our instruction-tuning data."
                    },
                    {
                        "title": "A Video Is Worth 4096 Tokens: Verbalize Videos To Understand Them In Zero Shot",
                        "abstract": "Multimedia content, such as advertisements and story videos, exhibit a rich blend of creativity and multiple modalities. They incorporate elements like text, visuals, audio, and storytelling techniques, employing devices like emotions, symbolism, and slogans to convey meaning. There is a dearth of large annotated training datasets in the multimedia domain hindering the development of supervised learning models with satisfactory performance for real-world applications. On the other hand, the rise of large language models (LLMs) has witnessed remarkable zero-shot performance in various natural language processing (NLP) tasks, such as emotion classification, question-answering, and topic classification. To leverage such advanced techniques to bridge this performance gap in multimedia understanding, we propose verbalizing long videos to generate their descriptions in natural language, followed by performing video-understanding tasks on the generated story as opposed to the original video. Through extensive experiments on fifteen video-understanding tasks, we demonstrate that our method, despite being zero-shot, achieves significantly better results than supervised baselines for video understanding. Furthermore, to alleviate a lack of story understanding benchmarks, we publicly release the first dataset on a crucial task in computational social science on persuasion strategy identification."
                    },
                    {
                        "title": "Behavior Optimized Image Generation",
                        "abstract": "The last few years have witnessed great success on image generation, which has crossed the acceptance thresholds of aesthetics, making it directly applicable to personal and commercial applications. However, images, especially in marketing and advertising applications, are often created as a means to an end as opposed to just aesthetic concerns. The goal can be increasing sales, getting more clicks, likes, or image sales (in the case of stock businesses). Therefore, the generated images need to perform well on these key performance indicators (KPIs), in addition to being aesthetically good. In this paper, we make the first endeavor to answer the question of\"How can one infuse the knowledge of the end-goal within the image generation process itself to create not just better-looking images but also\"better-performing'' images?''. We propose BoigLLM, an LLM that understands both image content and user behavior. BoigLLM knows how an image should look to get a certain required KPI. We show that BoigLLM outperforms 13x larger models such as GPT-3.5 and GPT-4 in this task, demonstrating that while these state-of-the-art models can understand images, they lack information on how these images perform in the real world. To generate actual pixels of behavior-conditioned images, we train a diffusion-based model (BoigSD) to align with a proposed BoigLLM-defined reward. We show the performance of the overall pipeline on two datasets covering two different behaviors: a stock dataset with the number of forward actions as the KPI and a dataset containing tweets with the total likes as the KPI, denoted as BoigBench. To advance research in the direction of utility-driven image generation and understanding, we release BoigBench, a benchmark dataset containing 168 million enterprise tweets with their media, brand account names, time of post, and total likes."
                    }
                ]
            },
            "d40b3dd8-f834-479e-88b7-c8473aeed166": {
                "pk": "d40b3dd8-f834-479e-88b7-c8473aeed166",
                "name": "Balaji Krishnamurthy",
                "collaborators": [
                    "Yaman Kumar Singla",
                    "Somesh Singh",
                    "R. Shah",
                    "Changyou Chen",
                    "I. HariniS",
                    "V. Baths",
                    "SI Harini",
                    "Aanisha Bhattacharya",
                    "Varun Khurana",
                    "J. Subramanian",
                    "Zhiqiang Xu",
                    "\u22c6. HariniSI"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Persuasion",
                    "Multimedia Understanding",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "Teaching Human Behavior Improves Content Understanding Abilities Of LLMs",
                        "abstract": "Communication is defined as\"Who says what to whom with what effect\". A message from a communicator generates downstream receiver effects, also known as behavior. Receiver behavior, being a downstream effect of the message, carries rich signals about it. Even after carrying signals about the message, the behavior data is often ignored while training large language models. We show that training LLMs on receiver behavior can actually help improve their content-understanding abilities. Specifically, we show that training LLMs to predict the receiver behavior of likes and comments improves the LLM's performance on a wide variety of downstream content understanding tasks. We show this performance increase over 46 video and image understanding tasks over 26 benchmark datasets across both 0-shot and fine-tuning settings, outperforming many supervised baselines. Moreover, since receiver behavior, such as likes and comments, is collected by default on the internet and does not need any human annotations to be useful, the performance improvement we get after training on this data is essentially free-lunch. We release the receiver behavior cleaned comments and likes of 750k images and videos collected from multiple platforms along with our instruction-tuning data."
                    },
                    {
                        "title": "Measuring and Improving Persuasiveness of Large Language Models",
                        "abstract": "LLMs are increasingly being used in workflows involving generating content to be consumed by humans (e.g., marketing) and also in directly interacting with humans (e.g., through chatbots). The development of such systems that are capable of generating verifiably persuasive messages presents both opportunities and challenges for society. On the one hand, such systems could positively impact domains like advertising and social good, such as addressing drug addiction, and on the other, they could be misused for spreading misinformation and shaping political opinions. To channel LLMs' impact on society, we need to develop systems to measure and benchmark their persuasiveness. With this motivation, we introduce PersuasionBench and PersuasionArena, the first large-scale benchmark and arena containing a battery of tasks to measure the persuasion ability of generative models automatically. We investigate to what extent LLMs know and leverage linguistic patterns that can help them generate more persuasive language. Our findings indicate that the persuasiveness of LLMs correlates positively with model size, but smaller models can also be made to have a higher persuasiveness than much larger models. Notably, targeted training using synthetic and natural datasets significantly enhances smaller models' persuasive capabilities, challenging scale-dependent assumptions. Our findings carry key implications for both model developers and policymakers. For instance, while the EU AI Act and California's SB-1047 aim to regulate AI models based on the number of floating point operations, we demonstrate that simple metrics like this alone fail to capture the full scope of AI's societal impact. We invite the community to explore and contribute to PersuasionArena and PersuasionBench, available at https://bit.ly/measure-persuasion, to advance our understanding of AI-driven persuasion and its societal implications."
                    },
                    {
                        "title": "A Video Is Worth 4096 Tokens: Verbalize Videos To Understand Them In Zero Shot",
                        "abstract": "Multimedia content, such as advertisements and story videos, exhibit a rich blend of creativity and multiple modalities. They incorporate elements like text, visuals, audio, and storytelling techniques, employing devices like emotions, symbolism, and slogans to convey meaning. There is a dearth of large annotated training datasets in the multimedia domain hindering the development of supervised learning models with satisfactory performance for real-world applications. On the other hand, the rise of large language models (LLMs) has witnessed remarkable zero-shot performance in various natural language processing (NLP) tasks, such as emotion classification, question-answering, and topic classification. To leverage such advanced techniques to bridge this performance gap in multimedia understanding, we propose verbalizing long videos to generate their descriptions in natural language, followed by performing video-understanding tasks on the generated story as opposed to the original video. Through extensive experiments on fifteen video-understanding tasks, we demonstrate that our method, despite being zero-shot, achieves significantly better results than supervised baselines for video understanding. Furthermore, to alleviate a lack of story understanding benchmarks, we publicly release the first dataset on a crucial task in computational social science on persuasion strategy identification."
                    },
                    {
                        "title": "Behavior Optimized Image Generation",
                        "abstract": "The last few years have witnessed great success on image generation, which has crossed the acceptance thresholds of aesthetics, making it directly applicable to personal and commercial applications. However, images, especially in marketing and advertising applications, are often created as a means to an end as opposed to just aesthetic concerns. The goal can be increasing sales, getting more clicks, likes, or image sales (in the case of stock businesses). Therefore, the generated images need to perform well on these key performance indicators (KPIs), in addition to being aesthetically good. In this paper, we make the first endeavor to answer the question of\"How can one infuse the knowledge of the end-goal within the image generation process itself to create not just better-looking images but also\"better-performing'' images?''. We propose BoigLLM, an LLM that understands both image content and user behavior. BoigLLM knows how an image should look to get a certain required KPI. We show that BoigLLM outperforms 13x larger models such as GPT-3.5 and GPT-4 in this task, demonstrating that while these state-of-the-art models can understand images, they lack information on how these images perform in the real world. To generate actual pixels of behavior-conditioned images, we train a diffusion-based model (BoigSD) to align with a proposed BoigLLM-defined reward. We show the performance of the overall pipeline on two datasets covering two different behaviors: a stock dataset with the number of forward actions as the KPI and a dataset containing tweets with the total likes as the KPI, denoted as BoigBench. To advance research in the direction of utility-driven image generation and understanding, we release BoigBench, a benchmark dataset containing 168 million enterprise tweets with their media, brand account names, time of post, and total likes."
                    },
                    {
                        "title": "M EASURING A ND I MPROVING P ERSUASIVENESS O F G ENERATIVE M ODELS",
                        "abstract": "of the generated content. However, our study challenges this scale-dependent assumption. We propose an instruction fine-tuning approach helping to enhance the persuasiveness of smaller language models, enabling them to surpass much larger models (13-100x) such as GPT-3.5 and GPT-4 (OpenAI, 2023). This finding suggests that persuasive capability is not necessarily a function of model scale and can be achieved through targeted training of smaller language models. We also show that persuasive capability developed in one domain ( e.g. twitter) transfers quite well to other domains ( e.g. websites, debates, and argumentation). This finding can potentially help policymakers like in the recent highly debated California bills on AI models and LLMs (Wiener, 2024; Bauer-Kahan, 2024) and the EU AI act (Union, 2024) that aim to decide appropriate standards for the development and use of AI models and datasets. These legislations try to control models above a certain number of floating point operations. Our findings suggest that simple measures like floating point operations or parameter count do not capture the complete picture of the potential societal implications of AI models, particularly with respect to complex issues like digital persuasion. We discuss more ethical challenges of studying persuasion in LLMs in \u00a7F. Our"
                    }
                ]
            }
        }
    },
    "2405.16166": {
        "paper_data": {
            "title": "The Power of Hard Attention Transformers on Data Sequences: A Formal Language Theoretic Perspective",
            "url": "http://arxiv.org/abs/2405.16166v1",
            "arxiv_id": "2405.16166",
            "authors": [
                "Pascal Bergstr\u00e4\u00dfer",
                "Chris K\u00f6cher",
                "Anthony Widjaja Lin",
                "Georg Zetzsche"
            ],
            "abstract": "Formal language theory has recently been successfully employed to unravel the power of transformer encoders. This setting is primarily applicable in Natural Languange Processing (NLP), as a token embedding function (where a bounded number of tokens is admitted) is first applied before feeding the input to the transformer. On certain kinds of data (e.g. time series), we want our transformers to be able to handle \\emph{arbitrary} input sequences of numbers (or tuples thereof) without \\emph{a priori} limiting the values of these numbers. In this paper, we initiate the study of the expressive power of transformer encoders on sequences of data (i.e. tuples of numbers). Our results indicate an increase in expressive power of hard attention transformers over data sequences, in stark contrast to the case of strings. In particular, we prove that Unique Hard Attention Transformers (UHAT) over inputs as data sequences no longer lie within the circuit complexity class $AC^0$ (even without positional encodings), unlike the case of string inputs, but are still within the complexity class $TC^0$ (even with positional encodings). Over strings, UHAT without positional encodings capture only regular languages. In contrast, we show that over data sequences UHAT can capture non-regular properties. Finally, we show that UHAT capture languages definable in an extension of linear temporal logic with unary numeric predicates and arithmetics.",
            "introduction": " Introduction to Circuit Complexity . Springer. [38] Wen, Q., Zhou, T., Zhang, C., Chen, W., Ma, Z., Yan, J., and Sun, L. (2023). Transformers in time series: A survey. In IJCAI , pages 6778\u20136786. ijcai.org. [39] Zhou, H., Zhang, S., Peng, J., Zhang, S., Li, J., Xiong, H., and Zhang, W. (2021). Informer: Beyond efficient transformer for long sequence time-series forecasting. In Thirty-Fifth AAAI Conference on Artificial Intelligence, AAAI 2021, Thirty-Third Conference on Innovative Applica- tions of Artificial Intelligence, IAAI 2021, The Eleventh Symposium on Educational Advances in Artificial Intelligence, EAAI 2021, Virtual Event, February 2-9, 2021 , pages 11106\u201311115. AAAI Press. 12A Example UHAT with real parameters We present here an example UHAT in which two real numbers \u03b1and\u03b2occur (each occurs once in a matrix associated to a particular layer) such that the simple sequence (1,0,e1)(1,0,e2)(1,0,e3), (2) where ei\u2208R3is the i-th unit vector, is accepted if and only if \u03b1\u03b2= 2 and\u03b1=\u03b2. This shows that even if we restrict the input sequence to a particular number Bof bits (e.g. the number of bits to represent the input sequence Eq. (2)), it is not possible to replace the real constants \u03b1and\u03b2by rational numbers without changing the accepted sequences of up to Bbits: The sequence Eq. (2) is only accepted if \u03b1=\u03b2=\u221a 2. And if we change \u03b1or\u03b2in any way, the sequence Eq. (2) will not be accepted anymore. 1. Input layer: (1,0,e1)(1,0,e2)(1,0,e3). 2.Usind standard encoding layer, we multiply the first component in each position by \u03b1. Result: (\u03b1,0,e1)(\u03b1,0,e2)(\u03b1,0,e3). 3.Using attention, we can apply distinct affine transformations to the first three positions. We choose the following affine transformations. The first position is unchanged. The second position is mapped to (1,0,e2), and the third position is mapped to (0, \u03b1,e3), using a matrix that flips the first and second component. Result: (\u03b1,0,e1)(1,0,e2)(0, \u03b1,e3). 4.Using a standard encoding layer, we multiply the first component in each position with \u03b2. Result: (\u03b1\u03b2,0,e1)(\u03b2,0,e2)(0, \u03b1,e3). 5.Finally, by using our result on (LT)2L, we can build further layers so that we accept if and only if (i) the first component of the first position equals 2and (ii) the first component of the second position equals the second component of the third position. Thus, we accept our original input if and only if \u03b1\u03b2= 2and\u03b1=\u03b2. B Omitted definitions and proofs in Section 3 B.1 Descriptional size Letx\u2208Rbe some real number. The (descriptional) size ofxissize(x) = 1 + \u2308log2(|p|+ 1)\u2309+ \u2308log2(|q|+ 1)\u2309ifx=p qis a rational number (where pandqare relatively prime integers) and size(x) = 1 ifxis an irrational number. Note that in the latter case we use the number 1as some placeholder, since we do not have to represent irrational numbers in any algorithm. However, for analysis of the sizes in our constructions we still need some value. Letv\u2208Rdbe a vector. The sizeofvissize(v) =n+Pd i=1size(vi)where v= (v1, . . . , v d)T. LetM\u2208Rm\u00d7nbe a matrix. The sizeofMissize(M) =mn+P 1\u2264i\u2264m,1\u2264j\u2264nsize(aij)where M= (aij)1\u2264i\u2264m,1\u2264j\u2264n. LetA:Rd\u2192Rebe an affine transformation, i.e., we have A(x) =Bx+cwithB\u2208Rd\u00d7eand c\u2208Re. Then the sizeofAissize(A) = size( B) + size( c) + 1 . Now, let p\u2208R[X1, . . . , X n]be a polynomial. Then we have p(X1, . . . , X n) =P 0\u2264r1,...,r n\u2264kcr1,...,r nXr1 1\u00b7\u00b7\u00b7Xrnnfor some numbers k\u2208Nandcr1,...,r n\u2208R. The sizeofp is size(p) =X 0\u2264r1,...,r n\u2264ksize(cr1,...,r n) + size( r1) +\u00b7\u00b7\u00b7+ size( rn) +n . Let\u03b1be a polynomial constraint. We define the sizeof\u03b1inductively on the structure of the formula as follows: 13\u2022 if\u03b1= (p(X1, .",
            "references": [
                {
                    "title": "Transformers as Transducers",
                    "abstract": "We study the sequence-to-sequence mapping capacity of transformers by relating them to finite transducers, and find that they can express surprisingly large classes of transductions. We do so using variants of RASP, a programming language designed to help people\"think like transformers,\"as an intermediate representation. We extend the existing Boolean variant B-RASP to sequence-to-sequence functions and show that it computes exactly the first-order rational functions (such as string rotation). Then, we introduce two new extensions. B-RASP[pos] enables calculations on positions (such as copying the first half of a string) and contains all first-order regular functions. S-RASP adds prefix sum, which enables additional arithmetic operations (such as squaring a string) and contains all first-order polyregular functions. Finally, we show that masked average-hard attention transformers can simulate S-RASP."
                },
                {
                    "title": "iTransformer: Inverted Transformers Are Effective for Time Series Forecasting",
                    "abstract": "The recent boom of linear forecasting models questions the ongoing passion for architectural modifications of Transformer-based forecasters. These forecasters leverage Transformers to model the global dependencies over temporal tokens of time series, with each token formed by multiple variates of the same timestamp. However, Transformers are challenged in forecasting series with larger lookback windows due to performance degradation and computation explosion. Besides, the embedding for each temporal token fuses multiple variates that represent potential delayed events and distinct physical measurements, which may fail in learning variate-centric representations and result in meaningless attention maps. In this work, we reflect on the competent duties of Transformer components and repurpose the Transformer architecture without any modification to the basic components. We propose iTransformer that simply applies the attention and feed-forward network on the inverted dimensions. Specifically, the time points of individual series are embedded into variate tokens which are utilized by the attention mechanism to capture multivariate correlations; meanwhile, the feed-forward network is applied for each variate token to learn nonlinear representations. The iTransformer model achieves state-of-the-art on challenging real-world datasets, which further empowers the Transformer family with promoted performance, generalization ability across different variates, and better utilization of arbitrary lookback windows, making it a nice alternative as the fundamental backbone of time series forecasting. Code is available at this repository: https://github.com/thuml/iTransformer."
                },
                {
                    "title": "Logical Languages Accepted by Transformer Encoders with Hard Attention",
                    "abstract": "We contribute to the study of formal languages that can be recognized by transformer encoders. We focus on two self-attention mechanisms: (1) UHAT (Unique Hard Attention Transformers) and (2) AHAT (Average Hard Attention Transformers). UHAT encoders are known to recognize only languages inside the circuit complexity class ${\\sf AC}^0$, i.e., accepted by a family of poly-sized and depth-bounded boolean circuits with unbounded fan-ins. On the other hand, AHAT encoders can recognize languages outside ${\\sf AC}^0$), but their expressive power still lies within the bigger circuit complexity class ${\\sf TC}^0$, i.e., ${\\sf AC}^0$-circuits extended by majority gates. We first show a negative result that there is an ${\\sf AC}^0$-language that cannot be recognized by an UHAT encoder. On the positive side, we show that UHAT encoders can recognize a rich fragment of ${\\sf AC}^0$-languages, namely, all languages definable in first-order logic with arbitrary unary numerical predicates. This logic, includes, for example, all regular languages from ${\\sf AC}^0$. We then show that AHAT encoders can recognize all languages of our logic even when we enrich it with counting terms. We apply these results to derive new results on the expressive power of UHAT and AHAT up to permutation of letters (a.k.a. Parikh images)."
                },
                {
                    "title": "Stack Attention: Improving the Ability of Transformers to Model Hierarchical Patterns",
                    "abstract": "Attention, specifically scaled dot-product attention, has proven effective for natural language, but it does not have a mechanism for handling hierarchical patterns of arbitrary nesting depth, which limits its ability to recognize certain syntactic structures. To address this shortcoming, we propose stack attention: an attention operator that incorporates stacks, inspired by their theoretical connections to context-free languages (CFLs). We show that stack attention is analogous to standard attention, but with a latent model of syntax that requires no syntactic supervision. We propose two variants: one related to deterministic pushdown automata (PDAs) and one based on nondeterministic PDAs, which allows transformers to recognize arbitrary CFLs. We show that transformers with stack attention are very effective at learning CFLs that standard transformers struggle on, achieving strong results on a CFL with theoretically maximal parsing difficulty. We also show that stack attention is more effective at natural language modeling under a constrained parameter budget, and we include results on machine translation."
                },
                {
                    "title": "Average-Hard Attention Transformers are Constant-Depth Uniform Threshold Circuits",
                    "abstract": "Transformers have emerged as a widely used neural network model for various natural language processing tasks. Previous research explored their relationship with constant-depth threshold circuits, making two assumptions: average-hard attention and logarithmic precision for internal computations relative to input length. Merrill et al. (2022) prove that average-hard attention transformers recognize languages that fall within the complexity class TC0, denoting the set of languages that can be recognized by constant-depth polynomial-size threshold circuits. Likewise, Merrill and Sabharwal (2023) show that log-precision transformers recognize languages within the class of uniform TC0. This shows that both transformer models can be simulated by constant-depth threshold circuits, with the latter being more robust due to generating a uniform circuit family. Our paper shows that the first result can be extended to yield uniform circuits as well."
                },
                {
                    "title": "Skolem Meets Bateman-Horn",
                    "abstract": "The Skolem Problem asks to determine whether a given integer linear recurrence sequence has a zero term. This problem arises across a wide range of topics in computer science, including loop termination, formal languages, automata theory, and control theory, amongst many others. Decidability of the Skolem Problem is notoriously open. The state of the art is a decision procedure for recurrences of order at most 4: an advance achieved some 40 years ago, based on Baker's theorem on linear forms in logarithms of algebraic numbers. A new approach to the Skolem Problem was recently initiated via the notion of a Universal Skolem Set: a set $S$ of positive integers such that it is decidable whether a given non-degenerate linear recurrence sequence has a zero in $S$. Clearly, proving decidability of the Skolem Problem is equivalent to showing that $\\mathbb{N}$ itself is a Universal Skolem Set. The main contribution of the present paper is to construct a Universal Skolem Set that has lower density at least $0.29$. We show moreover that this set has density one subject to the Bateman-Horn conjecture. The latter is a central unifying hypothesis concerning the frequency of prime numbers among the values of systems of polynomials."
                },
                {
                    "title": "Tighter Bounds on the Expressivity of Transformer Encoders",
                    "abstract": "Characterizing neural networks in terms of better-understood formal systems has the potential to yield new insights into the power and limitations of these networks. Doing so for transformers remains an active area of research. Bhattamishra and others have shown that transformer encoders are at least as expressive as a certain kind of counter machine, while Merrill and Sabharwal have shown that fixed-precision transformer encoders recognize only languages in uniform $TC^0$. We connect and strengthen these results by identifying a variant of first-order logic with counting quantifiers that is simultaneously an upper bound for fixed-precision transformer encoders and a lower bound for transformer encoders. This brings us much closer than before to an exact characterization of the languages that transformer encoders recognize."
                },
                {
                    "title": "On the Skolem Problem and the Skolem Conjecture",
                    "abstract": "It is a longstanding open problem whether there is an algorithm to decide the Skolem Problem for linear recurrence sequences (LRS) over the integers, namely whether a given such sequence has a zero term (i.e., whether un = 0 for some n). A major breakthrough in the early 1980s established decidability for LRS of order 4 or less, i.e., for LRS in which every new term depends linearly on the previous four (or fewer) terms. The Skolem Problem for LRS of order 5 or more, in particular, remains a major open challenge to this day. Our main contributions in this paper are as follows: First, we show that the Skolem Problem is decidable for reversible LRS of order 7 or less. (An integer LRS is reversible if its unique extension to a bi-infinite LRS also takes exclusively integer values; a typical example is the classical Fibonacci sequence, whose bi-infinite extension is \u27e8\u2026, 5, \u22123, 2, \u22121, 1, 0, 1, 1, 2, 3, 5, \u2026\u27e9.) Second, assuming the Skolem Conjecture (a central hypothesis in Diophantine analysis, also known as the Exponential Local-Global Principle), we show that the Skolem Problem for LRS of order 5 is decidable, and exhibit a concrete procedure for solving it."
                },
                {
                    "title": "What's Decidable about Discrete Linear Dynamical Systems?",
                    "abstract": ". We survey the state of the art on the algorithmic analysis of discrete linear dynamical systems, focussing in particular on reachability, model-checking, and invariant-generation questions, both unconditionally as well as relative to oracles for the Skolem Problem."
                },
                {
                    "title": "Formal Language Recognition by Hard Attention Transformers: Perspectives from Circuit Complexity",
                    "abstract": "Abstract This paper analyzes three formal models of Transformer encoders that differ in the form of their self-attention mechanism: unique hard attention (UHAT); generalized unique hard attention (GUHAT), which generalizes UHAT; and averaging hard attention (AHAT). We show that UHAT and GUHAT Transformers, viewed as string acceptors, can only recognize formal languages in the complexity class AC0, the class of languages recognizable by families of Boolean circuits of constant depth and polynomial size. This upper bound subsumes Hahn\u2019s (2020) results that GUHAT cannot recognize the DYCK languages or the PARITY language, since those languages are outside AC0 (Furst et al., 1984). In contrast, the non-AC0 languages MAJORITY and DYCK-1 are recognizable by AHAT networks, implying that AHAT can recognize languages that UHAT and GUHAT cannot."
                },
                {
                    "title": "Transformers in Time Series: A Survey",
                    "abstract": "Transformers have achieved superior performances in many tasks in natural language processing and computer vision, which also triggered great interest in the time series community. Among multiple advantages of Transformers, the ability to capture long-range dependencies and interactions is especially attractive for time series modeling, leading to exciting progress in various time series applications. In this paper, we systematically review Transformer schemes for time series modeling by highlighting their strengths as well as limitations. In particular, we examine the development of time series Transformers in two perspectives. From the perspective of network structure, we summarize the adaptations and modifications that have been made to Transformers in order to accommodate the challenges in time series analysis. From the perspective of applications, we categorize time series Transformers based on common tasks including forecasting, anomaly detection, and classification. Empirically, we perform robust analysis, model size analysis, and seasonal-trend decomposition analysis to study how Transformers perform in time series. Finally, we discuss and suggest future directions to provide useful research guidance."
                },
                {
                    "title": "Saturated Transformers are Constant-Depth Threshold Circuits",
                    "abstract": "Abstract Transformers have become a standard neural network architecture for many NLP problems, motivating theoretical analysis of their power in terms of formal languages. Recent work has shown that transformers with hard attention are quite limited in power (Hahn, 2020), as they can be simulated by constant-depth AND/OR circuits (Hao et al., 2022). However, hard attention is a strong assumption, which may complicate the relevance of these results in practice. In this work, we analyze the circuit complexity of transformers with saturated attention: a generalization of hard attention that more closely captures the attention patterns learnable in practical transformers. We first show that saturated transformers transcend the known limitations of hard-attention transformers. We then prove saturated transformers with floating-point values can be simulated by constant-depth threshold circuits, giving the class TC0 as an upper bound on the formal languages they recognize."
                },
                {
                    "title": "Automata modulo theories",
                    "abstract": "Symbolic automata better balances how automata are implemented in practice."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Register Automata with Extrema Constraints, and an Application to Two-Variable Logic",
                    "abstract": "We introduce a model of register automata over infinite trees with extrema constraints. Such an automaton can store elements of a linearly ordered domain in its registers, and can compare those values to the suprema and infima of register values in subtrees. We show that the emptiness problem for these automata is decidable. As an application, we prove decidability of the countable satisfiability problem for two-variable logic in the presence of a tree order, a linear order, and arbitrary atoms that are MSO definable from the tree order. As a consequence, the satisfiability problem for two-variable logic with arbitrary predicates, two of them interpreted by linear orders, is decidable."
                },
                {
                    "title": "Theoretical Limitations of Self-Attention in Neural Sequence Models",
                    "abstract": "Transformers are emerging as the new workhorse of NLP, showing great success across tasks. Unlike LSTMs, transformers process input sequences entirely through self-attention. Previous work has suggested that the computational capabilities of self-attention to process hierarchical structures are limited. In this work, we mathematically investigate the computational power of self-attention to model formal languages. Across both soft and hard attention, we show strong theoretical limitations of the computational abilities of self-attention, finding that it cannot model periodic finite-state languages, nor hierarchical structure, unless the number of layers or heads increases with input length. These limitations seem surprising given the practical success of self-attention and the prominent role assigned to hierarchical structure in linguistics, suggesting that natural language can be approximated well with models that are too weak for the formal languages typically assumed in theoretical linguistics."
                },
                {
                    "title": "Speech-Transformer: A No-Recurrence Sequence-to-Sequence Model for Speech Recognition",
                    "abstract": "Recurrent sequence-to-sequence models using encoder-decoder architecture have made great progress in speech recognition task. However, they suffer from the drawback of slow training speed because the internal recurrence limits the training parallelization. In this paper, we present the Speech-Transformer, a no-recurrence sequence-to-sequence model entirely relies on attention mechanisms to learn the positional dependencies, which can be trained faster with more efficiency. We also propose a 2D-Attention mechanism, which can jointly attend to the time and frequency axes of the 2-dimensional speech inputs, thus providing more expressive representations for the Speech-Transformer. Evaluated on the Wall Street Journal (WSJ) speech recognition dataset, our best model achieves competitive word error rate (WER) of 10.9%, while the whole training process only takes 1.2 days on 1 GPU, significantly faster than the published results of recurrent sequence-to-sequence models."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Querying Graphs with Data",
                    "abstract": "Graph databases have received much attention as of late due to numerous applications in which data is naturally viewed as a graph; these include social networks, RDF and the Semantic Web, biological databases, and many others. There are many proposals for query languages for graph databases that mainly fall into two categories. One views graphs as a particular kind of relational data and uses traditional relational mechanisms for querying. The other concentrates on querying the topology of the graph. These approaches, however, lack the ability to combine data and topology, which would allow queries asking how data changes along paths and patterns enveloping it. In this article, we present a comprehensive study of languages that enable such combination of data and topology querying. These languages come in two flavors. The first follows the standard approach of path queries, which specify how labels of edges change along a path, but now we extend them with ways of specifying how both labels and data change. From the complexity point of view, the right type of formalisms are subclasses of register automata. These, however, are not well suited for querying. To overcome this, we develop several types of extended regular expressions to specify paths with data and study their querying power and complexity. The second approach adopts the popular XML language XPath and extends it from XML documents to graphs. Depending on the exact set of allowed features, we have a family of languages, and our study shows that it includes efficient and highly expressive formalisms for querying both the structure of the data and the data itself."
                },
                {
                    "title": "Symbolic finite state transducers: algorithms and applications",
                    "abstract": "Finite automata and finite transducers are used in a wide range of applications in software engineering, from regular expressions to specification languages. We extend these classic objects with symbolic alphabets represented as parametric theories. Admitting potentially infinite alphabets makes this representation strictly more general and succinct than classical finite transducers and automata over strings. Despite this, the main operations, including composition, checking that a transducer is single-valued, and equivalence checking for single-valued symbolic finite transducers are effective given a decision procedure for the background theory. We provide novel algorithms for these operations and extend composition to symbolic transducers augmented with registers. Our base algorithms are unusual in that they are nonconstructive, therefore, we also supply a separate model generation algorithm that can quickly find counterexamples in the case two symbolic finite transducers are not equivalent. The algorithms give rise to a complete decidable algebra of symbolic transducers. Unlike previous work, we do not need any syntactic restriction of the formulas on the transitions, only a decision procedure. In practice we leverage recent advances in satisfiability modulo theory (SMT) solvers. We demonstrate our techniques on four case studies, covering a wide range of applications. Our techniques can synthesize string pre-images in excess of 8,000 bytes in roughly a minute, and we find that our new encodings significantly outperform previous techniques in succinctness and speed of analysis."
                },
                {
                    "title": "Two-variable logic on data words",
                    "abstract": "In a data word each position carries a label from a finite alphabet and a data value from some infinite domain. This model has been already considered in the realm of semistructured data, timed automata, and extended temporal logics.\n This article shows that satisfiability for the two-variable fragment FO2(\u223c,<,+1) of first-order logic with data equality test \u223c is decidable over finite and infinite data words. Here +1 and < are the usual successor and order predicates, respectively. The satisfiability problem is shown to be at least as hard as reachability in Petri nets. Several extensions of the logic are considered; some remain decidable while some are undecidable."
                },
                {
                    "title": "On the complexity of numerical analysis",
                    "abstract": "We study two quite different approaches to understanding the complexity of fundamental problems in numerical analysis. We show that both hinge on the question of understanding the complexity of the following problem, which we call PosSLP; given a division-free straight-line program producing an integer N, decide whether N > 0. We show that PosSLP lies in the counting hierarchy, and combining our results with work of Tiwari, we show that the Euclidean traveling salesman problem lies in the counting hierarchy - the previous best upper bound for this important problem (in terms of classical complexity classes) being PSPACE"
                },
                {
                    "title": "Finite-memory automata",
                    "abstract": "A model of computation dealing with infinite alphabets is proposed. The model is based on replacing the equality test by unification. It appears to be a natural generalization of the classical Rabin-Scott finite-state automata and possesses many of their properties.<<ETX>>"
                },
                {
                    "title": "Theory of linear and integer programming",
                    "abstract": "Introduction and Preliminaries. Problems, Algorithms, and Complexity. LINEAR ALGEBRA. Linear Algebra and Complexity. LATTICES AND LINEAR DIOPHANTINE EQUATIONS. Theory of Lattices and Linear Diophantine Equations. Algorithms for Linear Diophantine Equations. Diophantine Approximation and Basis Reduction. POLYHEDRA, LINEAR INEQUALITIES, AND LINEAR PROGRAMMING. Fundamental Concepts and Results on Polyhedra, Linear Inequalities, and Linear Programming. The Structure of Polyhedra. Polarity, and Blocking and Anti--Blocking Polyhedra. Sizes and the Theoretical Complexity of Linear Inequalities and Linear Programming. The Simplex Method. Primal--Dual, Elimination, and Relaxation Methods. Khachiyana s Method for Linear Programming. The Ellipsoid Method for Polyhedra More Generally. Further Polynomiality Results in Linear Programming. INTEGER LINEAR PROGRAMMING. Introduction to Integer Linear Programming. Estimates in Integer Linear Programming. The Complexity of Integer Linear Programming. Totally Unimodular Matrices: Fundamental Properties and Examples. Recognizing Total Unimodularity. Further Theory Related to Total Unimodularity. Integral Polyhedra and Total Dual Integrality. Cutting Planes. Further Methods in Integer Linear Programming. References. Indexes."
                },
                {
                    "title": "Constant Depth Reducibility",
                    "abstract": "The purpose of this paper is to study reducibilities that can be computed by combinational logic networks of polynomial size and constant depth containing AND\u2019s, OR\u2019s and NOT\u2019s, with no bound placed on the fan-in of AND-gates and OR-gates. Two such reducibilities are defined, and reductions and equivalences among several common problems such as parity, sorting, integer multiplication, graph connectivity, bipartite matching and network flow are given. Certain problems are shown to be complete, with respect to these reducibilities, in the complexity classes deterministic logarithmic space, nondeterministic logarithmic space, and deterministic polynomial time. New upper bounds on the size-depth (unbounded fan-in) circuit complexity of symmetric Boolean functions are established."
                },
                {
                    "title": "Some NP-complete geometric problems",
                    "abstract": "We show that the STEINER TREE problem and TRAVELING SALESMAN problem for points in the plane are NP-complete when distances are measured either by the rectilinear (Manhattan) metric or by a natural discretized version of the Euclidean metric. Our proofs also indicate that the problems are NP-hard if the distance measure is the (unmodified) Euclidean metric. However, for reasons we discuss, there is some question as to whether these problems, or even the well-solved MINIMUM SPANNING TREE problem, are in NP when the distance measure is the Euclidean metric."
                },
                {
                    "title": "Attention is Turing-Complete",
                    "abstract": "Alternatives to recurrent neural networks, in particular, architectures based on self-attention , are gaining momentum for processing input sequences. In spite of their relevance, the computational properties of such networks have not yet been fully explored. We study the computational power of the Transformer , one of the most paradigmatic architectures ex-emplifying self-attention. We show that the Transformer with hard-attention is Turing complete exclusively based on their capacity to compute and access internal dense representations of the data. Our study also reveals some minimal sets of elements needed to obtain this completeness result."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Automata based verification over linearly ordered data domains",
                    "abstract": "In this paper we work over linearly ordered data domains equipped with finitely many unary predicates and constants. We consider nondeterministic automata processing words and storing finitely many variables ranging over the domain. During a transition, these automata can compare the data values of the current configuration with those of the previous configuration using the linear order, the unary predicates and the constants. \n \nWe show that emptiness for such automata is decidable, both over finite and infinite words, under reasonable computability assumptions on the linear order. \n \nFinally, we show how our automata model can be used for verifying properties of workflow specifications in the presence of an underlying database."
                }
            ],
            "categories": [
                "cs.FL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively utilize transformers for long sequence time-series forecasting while addressing the inherent complexities of circuit complexity in machine learning models?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the research community's understanding of how transformer architectures can be optimized for time-series data, which is prevalent in various fields such as finance, healthcare, and climate science. A successful approach could lead to more accurate forecasting models, enabling better decision-making and resource allocation. Furthermore, it could inspire future research into hybrid models that combine the strengths of transformers with other machine learning techniques, ultimately pushing the boundaries of what is achievable in predictive analytics.\n\n### [Question 3] - Why is it hard?\nThe challenges in this problem stem from the complexity of circuit design in machine learning, particularly when dealing with real-valued parameters and the need for precise transformations. Naive approaches may fail due to the intricate relationships between input sequences and the parameters that govern them, as demonstrated by the requirement that certain conditions (e.g., \u03b1\u03b2=2 and \u03b1=\u03b2) must be met for the model to accept specific sequences. Additionally, the need to balance computational efficiency with model accuracy presents a significant technical obstacle, as does the difficulty in representing irrational numbers within the constraints of traditional algorithms.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often overlooked the specific challenges posed by circuit complexity in the context of transformers and time-series forecasting. Many existing solutions have focused on either improving transformer efficiency or enhancing their predictive capabilities without addressing the underlying mathematical constraints that govern their performance. Barriers such as a lack of comprehensive frameworks for integrating circuit complexity with transformer architectures have hindered progress. Our approach aims to bridge this gap by providing a detailed methodology that incorporates these complexities into the design of forecasting models.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves the following key components: \n1. **Method**: We will develop a transformer-based model that incorporates circuit complexity principles, specifically focusing on the manipulation of real-valued parameters through affine transformations.\n2. **Dataset**: We will utilize a publicly available time-series dataset relevant to our forecasting objectives, ensuring it contains long sequences to test the model's capabilities.\n3. **Metric**: The performance of our model will be evaluated using metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) to assess forecasting accuracy.\n"
            }
        },
        "author_data": {
            "ae583365-2a5c-4773-af52-a931212f68ab": {
                "pk": "ae583365-2a5c-4773-af52-a931212f68ab",
                "name": "Pascal Bergstr\u00e4\u00dfer",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "48d157e2-0e42-4633-a101-2def5f7cc3d5": {
                "pk": "48d157e2-0e42-4633-a101-2def5f7cc3d5",
                "name": "Chris K\u00f6cher",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "07ef2055-1770-4c0c-b14b-939a73e5740e": {
                "pk": "07ef2055-1770-4c0c-b14b-939a73e5740e",
                "name": "Anthony Widjaja Lin",
                "collaborators": [
                    "Matthew Hague",
                    "Daniel Stan",
                    "Zhilin Wu",
                    "Pablo Barcelo",
                    "Sanming Zhou",
                    "Luke Ong",
                    "Philipp R\u00fcmmer",
                    "Michael Benedikt",
                    "Di-De Yen",
                    "Alexander Kozachinskiy"
                ],
                "domain": [
                    "Formal Methods",
                    "Model Checking",
                    "Graph Theory",
                    "String Analysis"
                ],
                "publications": [
                    {
                        "title": "A linear time algorithm for the orbit problem over cyclic groups",
                        "abstract": "The orbit problem is at the heart of symmetry reduction methods for model checking concurrent systems. It asks whether two given configurations in a concurrent system (represented as finite strings over some finite alphabet) are in the same orbit with respect to a given finite permutation group (represented by their generators) acting on this set of configurations by permuting indices. It is known that the problem is in general as hard as the graph isomorphism problem, whose precise complexity (whether it is solvable in polynomial-time) is a long-standing open problem. In this paper, we consider the restriction of the orbit problem when the permutation group is cyclic (i.e. generated by a single permutation), an important restriction of the problem. It is known that this subproblem is solvable in polynomial-time. Our main result is a linear-time algorithm for this subproblem."
                    },
                    {
                        "title": "Decidable models of integer-manipulating programs with recursive parallelism (technical report)",
                        "abstract": "We study safety verification for multithreaded programs with recursive parallelism (i.e. unbounded thread creation and recursion) as well as unbounded integer variables. Since the threads in each program configuration are structured in a hierarchical fashion, our model is state-extended ground-tree rewrite systems equipped with shared unbounded integer counters that can be incremented, decremented, and compared against an integer constant. Since the model is Turing-complete, we propose a decidable underapproximation. First, using a restriction similar to context-bounding, we underapproximate the global control by a weak global control (i.e. DAGs possibly with self-loops), thereby limiting the number of synchronisations between different threads. Second, we bound the number of reversals between non-decrementing and non-incrementing modes of the counters. Under this restriction, we show that reachability becomes NP-complete. In fact, it is poly-time reducible to satisfaction over existential Presburger formulas, which allows one to tap into highly optimised SMT solvers. Our decidable approximation strictly generalises known decidable models including (i) weakly-synchronised ground-tree rewrite systems, and (ii) synchronisation/reversal-bounded concurrent pushdown systems systems with counters. Finally, we show that, when equipped with reversal-bounded counters, relaxing the weak control restriction by the notion of senescence results in undecidability."
                    },
                    {
                        "title": "Detecting Redundant CSS Rules in HTML5 Applications: A Tree-Rewriting Approach",
                        "abstract": "HTML5 applications normally have a large set of CSS (Cascading Style Sheets) rules for data display. Each CSS rule consists of a node selector (given in an XPath-like query language) and a declaration block (assigning values to selected nodes' display attributes). As web applications evolve, maintaining CSS files can easily become problematic. Some CSS rules will be replaced by new ones, but these obsolete (hence redundant) CSS rules often remain in the applications. Not only does this \"bloat\" the applications, but it also significantly increases web browsers' processing time. Most works on detecting redundant CSS rules in HTML5 applications do not consider the dynamic behaviors of HTML5 (specified in JavaScript); in fact, the only proposed method that takes these into account is dynamic analysis (a.k.a. testing), which cannot soundly prove redundancy of CSS rules. In this paper, we introduce an abstraction of HTML5 applications based on monotonic tree-rewriting and study its \"redundancy problem\". We establish the precise complexity of the problem and various subproblems of practical importance (ranging from P to EXP). In particular, our algorithm relies on an efficient reduction to an analysis of symbolic pushdown systems (for which highly optimised solvers are available), which yields a fast method for checking redundancy in practice. We implemented our algorithm and demonstrated its efficacy in detecting redundant CSS rules in HTML5 applications."
                    },
                    {
                        "title": "Regular Model Checking Approach to Knowledge Reasoning over Parameterized Systems (technical report)",
                        "abstract": "We present a general framework for modelling and verifying epistemic properties over parameterized multi-agent systems that communicate by truthful public announcements. In our framework, the number of agents or the amount of certain resources are parameterized (i.e. not known a priori), and the corresponding verification problem asks whether a given epistemic property is true regardless of the instantiation of the parameters. For example, in a muddy children puzzle, one could ask whether each child will eventually find out whether (s)he is muddy, regardless of the number of children.   Our framework is regular model checking (RMC)-based, wherein synchronous finite-state automata (equivalently, monadic second-order logic over words) are used to specify the systems. We propose an extension of public announcement logic as specification language. Of special interests is the addition of the so-called iterated public announcement operators, which are crucial for reasoning about knowledge in parameterized systems. Although the operators make the model checking problem undecidable, we show that this becomes decidable when an appropriate \"disappearance relation\" is given. Further, we show how Angluin's L*-algorithm for learning finite automata can be applied to find a disappearance relation, which is guaranteed to terminate if it is regular. We have implemented the algorithm and apply this to such examples as the Muddy Children Puzzle, the Russian Card Problem, and Large Number Challenge."
                    },
                    {
                        "title": "Monadic Decomposition in Integer Linear Arithmetic (Technical Report)",
                        "abstract": "Monadic decomposability is a notion of variable independence, which asks whether a given formula in a first-order theory is expressible as a Boolean combination of monadic predicates in the theory. Recently, Veanes et al. showed the usefulness of monadic decomposability in the context of SMT (i.e. the input formula is quantifier-free), and found various interesting applications including string analysis. However, checking monadic decomposability is undecidable in general. Decidability for certain theories is known (e.g. Presburger Arithmetic, Tarski's Real-Closed Field), but there are very few results regarding their computational complexity. In this paper, we study monadic decomposability of integer linear arithmetic in the setting of SMT. We show that this decision problem is coNP-complete and, when monadically decomposable, a formula admits a decomposition of exponential size in the worst case. We provide a new application of our results to string constraint solving with length constraints. We then extend our results to variadic decomposability, where predicates could admit multiple free variables (in contrast to monadic decomposability). Finally, we give an application to quantifier elimination in integer linear arithmetic where the variables in a block of quantifiers, if independent, could be eliminated with an exponential (instead of the standard doubly exponential) blow-up."
                    },
                    {
                        "title": "Revisiting the Expressiveness Landscape of Data Graph Queries",
                        "abstract": "The study of graph queries in database theory has spanned more than three decades, resulting in a multitude of proposals for graph query languages. These languages differ in the mechanisms. We can identify three main families of languages, with the canonical representatives being: (1) regular path queries, (2) walk logic, and (3) first-order logic with transitive closure operators. This paper provides a complete picture of the expressive power of these languages in the context of data graphs. Specifically, we consider a graph data model that supports querying over both data and topology. For example, \"Does there exist a path between two different persons in a social network with the same last name?\". We also show that an extension of (1), augmented with transitive closure operators, can unify the expressivity of (1)--(3) without increasing the query evaluation complexity."
                    },
                    {
                        "title": "Logical Languages Accepted by Transformer Encoders with Hard Attention",
                        "abstract": "We contribute to the study of formal languages that can be recognized by transformer encoders. We focus on two self-attention mechanisms: (1) UHAT (Unique Hard Attention Transformers) and (2) AHAT (Average Hard Attention Transformers). UHAT encoders are known to recognize only languages inside the circuit complexity class ${\\sf AC}^0$, i.e., accepted by a family of poly-sized and depth-bounded boolean circuits with unbounded fan-ins. On the other hand, AHAT encoders can recognize languages outside ${\\sf AC}^0$), but their expressive power still lies within the bigger circuit complexity class ${\\sf TC}^0$, i.e., ${\\sf AC}^0$-circuits extended by majority gates. We first show a negative result that there is an ${\\sf AC}^0$-language that cannot be recognized by an UHAT encoder. On the positive side, we show that UHAT encoders can recognize a rich fragment of ${\\sf AC}^0$-languages, namely, all languages definable in first-order logic with arbitrary unary numerical predicates. This logic, includes, for example, all regular languages from ${\\sf AC}^0$. We then show that AHAT encoders can recognize all languages of our logic even when we enrich it with counting terms. We apply these results to derive new results on the expressive power of UHAT and AHAT up to permutation of letters (a.k.a. Parikh images)."
                    },
                    {
                        "title": "Concurrent Stochastic Lossy Channel Games",
                        "abstract": "Concurrent stochastic games are an important formalism for the rational verification of probabilistic multi-agent systems, which involves verifying whether a temporal logic property is satisfied in some or all game-theoretic equilibria of such systems. In this work, we study the rational verification of probabilistic multi-agent systems where agents can cooperate by communicating over unbounded lossy channels. To model such systems, we present concurrent stochastic lossy channel games (CSLCG) and employ an equilibrium concept from cooperative game theory known as the core, which is the most fundamental and widely studied cooperative equilibrium concept. Our main contribution is twofold. First, we show that the rational verification problem is undecidable for systems whose agents have almost-sure LTL objectives. Second, we provide a decidable fragment of such a class of objectives that subsumes almost-sure reachability and safety. Our techniques involve reductions to solving infinite-state zero-sum games with conjunctions of qualitative objectives. To the best of our knowledge, our result represents the first decidability result on the rational verification of stochastic multi-agent systems on infinite arenas."
                    },
                    {
                        "title": "Expressive Path Queries on Graph with Data",
                        "abstract": "Graph data models have recently become popular owing to their applications, e.g., in social networks and the semantic web. Typical navigational query languages over graph databases - such as Conjunctive Regular Path Queries (CRPQs) - cannot express relevant properties of the interaction between the underlying data and the topology. Two languages have been recently proposed to overcome this problem: walk logic (WL) and regular expressions with memory (REM). In this paper, we begin by investigating fundamental properties of WL and REM, i.e., complexity of evaluation problems and expressive power. We first show that the data complexity of WL is nonelementary, which rules out its practicality. On the other hand, while REM has low data complexity, we point out that many natural data/topology properties of graphs expressible in WL cannot be expressed in REM. To this end, we propose register logic, an extension of REM, which we show to be able to express many natural graph properties expressible in WL, while at the same time preserving the elementariness of data complexity of REMs. It is also incomparable to WL in terms of expressive power."
                    },
                    {
                        "title": "A Decision Procedure for Path Feasibility of String Manipulating Programs with Integer Data Type",
                        "abstract": "Strings are widely used in programs, especially in web applications. Integer data type occurs naturally in string-manipulating programs, and is frequently used to refer to lengths of, or positions in, strings. Analysis and testing of string-manipulating programs can be formulated as the path feasibility problem: given a symbolic execution path, does there exist an assignment to the inputs that yields a concrete execution that realizes this path? Such a problem can naturally be reformulated as a string constraint solving problem. Although state-of-the-art string constraint solvers usually provide support for both string and integer data types, they mainly resort to heuristics without completeness guarantees. In this paper, we propose a decision procedure for a class of string-manipulating programs which includes not only a wide range of string operations such as concatenation, replaceAll, reverse, and finite transducers, but also those involving the integer data-type such as length, indexof, and substring. To the best of our knowledge, this represents one of the most expressive string constraint languages that is currently known to be decidable. Our decision procedure is based on a variant of cost register automata. We implement the decision procedure, giving rise to a new solver OSTRICH+. We evaluate the performance of OSTRICH+ on a wide range of existing and new benchmarks. The experimental results show that OSTRICH+ is the first string decision procedure capable of tackling finite transducers and integer constraints, whilst its overall performance is comparable with the state-of-the-art string constraint solvers."
                    }
                ]
            },
            "2f88384e-3912-4599-8493-d3bb075a7d0b": {
                "pk": "2f88384e-3912-4599-8493-d3bb075a7d0b",
                "name": "Georg Zetzsche",
                "collaborators": [
                    "Markus Lohrey",
                    "Dietrich Kuske",
                    "Sylvain Schmitz",
                    "Miko\u0142aj Boja\u0144czyk",
                    "Edon Kelmendi",
                    "Rafa\u0142 Stefa\u0144ski",
                    "P. Buckheister",
                    "Martin Huschenbett",
                    "Laura Ciobanu",
                    "Benjamin Steinberg"
                ],
                "domain": [
                    "Formal Language Theory",
                    "Automata Theory",
                    "Computational Complexity",
                    "Monoid Theory"
                ],
                "publications": [
                    {
                        "title": "On the capabilities of grammars, automata, and transducers controlled by monoids",
                        "abstract": "During the last decades, classical models in language theory have been extended by control mechanisms defined by monoids. We study which monoids cause the extensions of context-free grammars, finite automata, or finite state transducers to exceed the capacity of the original model. Furthermore, we investigate when, in the extended automata model, the nondeterministic variant differs from the deterministic one in capacity. We show that all these conditions are in fact equivalent and present an algebraic characterization. In particular, the open question of whether every language generated by a valence grammar over a finite monoid is context-free is provided with a positive answer."
                    },
                    {
                        "title": "Silent Transitions in Automata with Storage",
                        "abstract": "We consider the computational power of silent transitions in one-way automata with storage. Specifically, we ask which storage mechanisms admit a transformation of a given automaton into one that accepts the same language and reads at least one input symbol in each step.   We study this question using the model of valence automata. Here, a finite automaton is equipped with a storage mechanism that is given by a monoid.   This work presents generalizations of known results on silent transitions. For two classes of monoids, it provides characterizations of those monoids that allow the removal of \\lambda-transitions. Both classes are defined by graph products of copies of the bicyclic monoid and the group of integers. The first class contains pushdown storages as well as the blind counters while the second class contains the blind and the partially blind counters."
                    },
                    {
                        "title": "Computing downward closures for stacked counter automata",
                        "abstract": "The downward closure of a language $L$ of words is the set of all (not necessarily contiguous) subwords of members of $L$. It is well known that the downward closure of any language is regular. Although the downward closure seems to be a promising abstraction, there are only few language classes for which an automaton for the downward closure is known to be computable.   It is shown here that for stacked counter automata, the downward closure is computable. Stacked counter automata are finite automata with a storage mechanism obtained by \\emph{adding blind counters} and \\emph{building stacks}. Hence, they generalize pushdown and blind counter automata.   The class of languages accepted by these automata are precisely those in the hierarchy obtained from the context-free languages by alternating two closure operators: imposing semilinear constraints and taking the algebraic extension. The main tool for computing downward closures is the new concept of Parikh annotations. As a second application of Parikh annotations, it is shown that the hierarchy above is strict at every level."
                    },
                    {
                        "title": "An approach to computing downward closures",
                        "abstract": "The downward closure of a word language is the set of all (not necessarily contiguous) subwords of its members. It is well-known that the downward closure of any language is regular. While the downward closure appears to be a powerful abstraction, algorithms for computing a finite automaton for the downward closure of a given language have been established only for few language classes.   This work presents a simple general method for computing downward closures. For language classes that are closed under rational transductions, it is shown that the computation of downward closures can be reduced to checking a certain unboundedness property.   This result is used to prove that downward closures are computable for (i) every language class with effectively semilinear Parikh images that are closed under rational transductions, (ii) matrix languages, and (iii) indexed languages (equivalently, languages accepted by higher-order pushdown automata of order 2)."
                    },
                    {
                        "title": "The complexity of downward closure comparisons",
                        "abstract": "The downward closure of a language is the set of all (not necessarily contiguous) subwords of its members. It is well-known that the downward closure of every language is regular. Moreover, recent results show that downward closures are computable for quite powerful system models.   One advantage of abstracting a language by its downward closure is that then equivalence and inclusion become decidable. In this work, we study the complexity of these two problems. More precisely, we consider the following decision problems: Given languages $K$ and $L$ from classes $\\mathcal{C}$ and $\\mathcal{D}$, respectively, does the downward closure of $K$ include (equal) that of $L$?   These problems are investigated for finite automata, one-counter automata, context-free grammars, and reversal-bounded counter automata. For each combination, we prove a completeness result either for fixed or for arbitrary alphabets. Moreover, for Petri net languages, we show that both problems are Ackermann-hard and for higher-order pushdown automata of order~$k$, we prove hardness for complements of nondeterministic $k$-fold exponential time."
                    },
                    {
                        "title": "PTL-separability and closures for WQOs on words",
                        "abstract": "We introduce a flexible class of well-quasi-orderings (WQOs) on words that generalizes the ordering of (not necessarily contiguous) subwords. Each such WQO induces a class of piecewise testable languages (PTLs) as Boolean combinations of upward closed sets. In this way, a range of regular language classes arises as PTLs. Moreover, each of the WQOs guarantees regularity of all downward closed sets. We consider two problems. First, we study which (perhaps non-regular) language classes permit a decision procedure to decide whether two given languages are separable by a PTL with respect to a given WQO. Second, we want to effectively compute downward closures with respect to these WQOs. Our first main result that for each of the WQOs, under mild assumptions, both problems reduce to the simultaneous unboundedness problem (SUP) and are thus solvable for many powerful system classes. In the second main result, we apply the framework to show decidability of separability of regular languages by $\\mathcal{B}\\Sigma_1[<, \\mathsf{mod}]$, a fragment of first-order logic with modular predicates."
                    },
                    {
                        "title": "The Emptiness Problem for Valence Automata over Graph Monoids",
                        "abstract": "This work studies which storage mechanisms in automata permit decidability of the emptiness problem. The question is formalized using valence automata, an abstract model of automata in which the storage mechanism is given by a monoid. For each of a variety of storage mechanisms, one can choose a (typically infinite) monoid $M$ such that valence automata over $M$ are equivalent to (one-way) automata with this type of storage. In fact, many important storage mechanisms can be realized by monoids defined by finite graphs, called graph monoids. Examples include pushdown stacks, partially blind counters (which behave like Petri net places), blind counters (which may attain negative values), and combinations thereof.   Hence, we study for which graph monoids the emptiness problem for valence automata is decidable. A particular model realized by graph monoids is that of Petri nets with a pushdown stack. For these, decidability is a long-standing open question and we do not answer it here.   However, if one excludes subgraphs corresponding to this model, a characterization can be achieved. Moreover, we provide a description of those storage mechanisms for which decidability remains open. This leads to a model that naturally generalizes both pushdown Petri nets and the priority multicounter machines introduced by Reinhardt.   The cases that are proven decidable constitute a natural and apparently new extension of Petri nets with decidable reachability. It is finally shown that this model can be combined with another such extension by Atig and Ganty: We present a further decidability result that subsumes both of these Petri net extensions."
                    },
                    {
                        "title": "The Complexity of Knapsack in Graph Groups",
                        "abstract": "Myasnikov et al. have introduced the knapsack problem for arbitrary finitely generated groups. In previous work, the authors proved that for each graph group, the knapsack problem can be solved in $\\mathsf{NP}$. Here, we determine the exact complexity of the problem for every graph group. While the problem is $\\mathsf{TC}^0$-complete for complete graphs, it is $\\mathsf{LogCFL}$-complete for each (non-complete) transitive forest. For every remaining graph, the problem is $\\mathsf{NP}$-complete."
                    },
                    {
                        "title": "Coverability is Undecidable in One-dimensional Pushdown Vector Addition Systems with Resets",
                        "abstract": "We consider the model of pushdown vector addition systems with resets. These consist of vector addition systems that have access to a pushdown stack and have instructions to reset counters. For this model, we study the coverability problem. In the absence of resets, this problem is known to be decidable for one-dimensional pushdown vector addition systems, but decidability is open for general pushdown vector addition systems. Moreover, coverability is known to be decidable for reset vector addition systems without a pushdown stack. We show in this note that the problem is undecidable for one-dimensional pushdown vector addition systems with resets."
                    },
                    {
                        "title": "Knapsack in graph groups, HNN-extensions and amalgamated products",
                        "abstract": "It is shown that the knapsack problem, which was introduced by Myasnikov et al. for arbitrary finitely generated groups, can be solved in NP for graph groups. This result even holds if the group elements are represented in a compressed form by SLPs, which generalizes the classical NP-completeness result of the integer knapsack problem. We also prove general transfer results: NP-membership of the knapsack problem is passed on to finite extensions, HNN-extensions over finite associated subgroups, and amalgamated products with finite identified subgroups."
                    },
                    {
                        "title": "Extensions of $\u03c9$-Regular Languages",
                        "abstract": "We consider extensions of monadic second order logic over $\\omega$-words, which are obtained by adding one language that is not $\\omega$-regular. We show that if the added language $L$ has a neutral letter, then the resulting logic is necessarily undecidable. A corollary is that the $\\omega$-regular languages are the only decidable Boolean-closed full trio over $\\omega$-words."
                    },
                    {
                        "title": "Semilinearity and Context-Freeness of Languages Accepted by Valence Automata",
                        "abstract": "Valence automata are a generalization of various models of automata with storage. Here, each edge carries, in addition to an input word, an element of a monoid. A computation is considered valid if multiplying the monoid elements on the visited edges yields the identity element. By choosing suitable monoids, a variety of automata models can be obtained as special valence automata.   This work is concerned with the accepting power of valence automata. Specifically, we ask for which monoids valence automata can accept only context-free languages or only languages with semilinear Parikh image, respectively.   First, we present a characterization of those graph products (of monoids) for which valence automata accept only context-free languages. Second, we provide a necessary and sufficient condition for a graph product of copies of the bicyclic monoid and the integers to yield only languages with semilinear Parikh image when used as a storage mechanism in valence automata. Third, we show that all languages accepted by valence automata over torsion groups have a semilinear Parikh image."
                    },
                    {
                        "title": "The monoid of queue actions",
                        "abstract": "We investigate the monoid of transformations that are induced by sequences of writing to and reading from a queue storage. We describe this monoid by means of a confluent and terminating semi-Thue system and study some of its basic algebraic properties, e.g., conjugacy. Moreover, we show that while several properties concerning its rational subsets are undecidable, their uniform membership problem is NL-complete. Furthermore, we present an algebraic characterization of this monoid's recognizable subsets. Finally, we prove that it is not Thurston-automatic."
                    },
                    {
                        "title": "Languages ordered by the subword order",
                        "abstract": "We consider a language together with the subword relation, the cover relation, and regular predicates. For such structures, we consider the extension of first-order logic by threshold- and modulo-counting quantifiers. Depending on the language, the used predicates, and the fragment of the logic, we determine four new combinations that yield decidable theories. These results extend earlier ones where only the language of all words without the cover relation and fragments of first-order logic were considered."
                    },
                    {
                        "title": "Slice closures of indexed languages and word equations with counting constraints",
                        "abstract": "Indexed languages are a classical notion in formal language theory. As the language equivalent of second-order pushdown automata, they have received considerable attention in higher-order model checking. Unfortunately, counting properties are notoriously difficult to decide for indexed languages: So far, all results about non-regular counting properties show undecidability.   In this paper, we initiate the study of slice closures of (Parikh images of) indexed languages. A slice is a set of vectors of natural numbers such that membership of $u,u+v,u+w$ implies membership of $u+v+w$. Our main result is that given an indexed language $L$, one can compute a semilinear representation of the smallest slice containing $L$'s Parikh image.   We present two applications. First, one can compute the set of all affine relations satisfied by the Parikh image of an indexed language. In particular, this answers affirmatively a question by Kobayashi: Is it decidable whether in a given indexed language, every word has the same number of $a$'s as $b$'s.   As a second application, we show decidability of (systems of) word equations with rational constraints and a class of counting constraints: These allow us to look for solutions where a counting function (defined by an automaton) is not zero. For example, one can decide whether a word equation with rational constraints has a solution where the number of occurrences of $a$ differs between variables $X$ and $Y$."
                    },
                    {
                        "title": "Rational Subsets and Submonoids of Wreath Products",
                        "abstract": "It is shown that membership in rational subsets of wreath products H \\wr V with H a finite group and V a virtually free group is decidable. On the other hand, it is shown that there exists a fixed finitely generated submonoid in the wreath product Z \\wr Z with an undecidable membership problem."
                    },
                    {
                        "title": "An Approach to Regular Separability in Vector Addition Systems",
                        "abstract": "We study the problem of regular separability of languages of vector addition systems with states (VASS). It asks whether for two given VASS languages K and L, there exists a regular language R that includes K and is disjoint from L. While decidability of the problem in full generality remains an open question, there are several subclasses for which decidability has been shown: It is decidable for (i) one-dimensional VASS, (ii) VASS coverability languages, (iii) languages of integer VASS, and (iv) commutative VASS languages. We propose a general approach to deciding regular separability. We use it to decide regular separability of an arbitrary VASS language from any language in the classes (i), (ii), and (iii). This generalizes all previous results, including (iv)."
                    },
                    {
                        "title": "Priority Downward Closures",
                        "abstract": "When a system sends messages through a lossy channel, then the language encoding all sequences of messages can be abstracted by its downward closure, i.e. the set of all (not necessarily contiguous) subwords. This is useful because even if the system has infinitely many states, its downward closure is a regular language. However, if the channel has congestion control based on priorities assigned to the messages, then we need a finer abstraction: The downward closure with respect to the priority embedding. As for subword-based downward closures, one can also show that these priority downward closures are always regular.   While computing finite automata for the subword-based downward closure is well understood, nothing is known in the case of priorities. We initiate the study of this problem and provide algorithms to compute priority downward closures for regular languages, one-counter languages, and context-free languages."
                    },
                    {
                        "title": "Decidability, Complexity, and Expressiveness of First-Order Logic Over the Subword Ordering",
                        "abstract": "We consider first-order logic over the subword ordering on finite words, where each word is available as a constant. Our first result is that the $\\Sigma_1$ theory is undecidable (already over two letters).   We investigate the decidability border by considering fragments where all but a certain number of variables are alternation bounded, meaning that the variable must always be quantified over languages with a bounded number of letter alternations. We prove that when at most two variables are not alternation bounded, the $\\Sigma_1$ fragment is decidable, and that it becomes undecidable when three variables are not alternation bounded. Regarding higher quantifier alternation depths, we prove that the $\\Sigma_2$ fragment is undecidable already for one variable without alternation bound and that when all variables are alternation bounded, the entire first-order theory is decidable."
                    },
                    {
                        "title": "The complexity of bidirected reachability in valence systems",
                        "abstract": "Reachability problems in infinite-state systems are often subject to extremely high complexity. This motivates the investigation of efficient overapproximations, where we add transitions to obtain a system in which reachability can be decided more efficiently. We consider bidirected infinite-state systems, where for every transition there is a transition with opposite effect.   We study bidirected reachability in the framework of valence systems, an abstract model featuring finitely many control states and an infinite-state storage that is specified by a finite graph. By picking suitable graphs, valence systems can uniformly model counters as in vector addition systems, pushdowns, integer counters, and combinations thereof.   We provide a comprehensive complexity landscape for bidirected reachability and show that the complexity drops (often to polynomial time) from that of general reachability, for almost every storage mechanism where reachability is known to be decidable."
                    }
                ]
            }
        }
    },
    "2311.13647": {
        "paper_data": {
            "title": "Language Model Inversion",
            "url": "http://arxiv.org/abs/2311.13647v1",
            "arxiv_id": "2311.13647",
            "authors": [
                "John X. Morris",
                "Wenting Zhao",
                "Justin T. Chiu",
                "Vitaly Shmatikov",
                "Alexander M. Rush"
            ],
            "abstract": "Language models produce a distribution over the next token; can we use this information to recover the prompt tokens? We consider the problem of language model inversion and show that next-token probabilities contain a surprising amount of information about the preceding text. Often we can recover the text in cases where it is hidden from the user, motivating a method for recovering unknown prompts given only the model's current distribution output. We consider a variety of model access scenarios, and show how even without predictions for every token in the vocabulary we can recover the probability vector through search. On Llama-2 7b, our inversion method reconstructs prompts with a BLEU of $59$ and token-level F1 of $78$ and recovers $27\\%$ of prompts exactly. Code for reproducing all experiments is available at http://github.com/jxmorris12/vec2text.",
            "introduction": "   1 Introduction  Language models are autoregressive, outputting the probability of each next token in a sequence conditioned on the preceeding text. This distribution is used to generate future tokens in the sequence. Can this distribution also be used to reconstruct the prompt?   In most contexts, this question is pointless, since we have already conditioned on this information. However, increasingly language models are being offered \u201cas a service\u201d where the user may have access to the outputs, but not all of the true prompt. In this context, it may be of interest to users to know the prompt and, perhaps, for the service provider to protect it. This goal has been the focus of \u201cjailbreaking\u201d approaches that attempt to use the forward text generation of the model to reveal the prompt.   We formalize this problem of prompt reconstruction as language model inversion, recovering the input prompt conditioned on the language model\u2019s next-token probabilities. Interestingly, work in computer vision has shown that probability predictions of image classifiers retain a surprising amount of detail\u00a0(Dosovitskiy & Brox, 2016), so it is plausible that this also holds for language models. We propose an architecture that predicts prompts by\u201cunrolling\u201d the distribution vector into a sequence that can be processed effectively by a pretrained encoder-decoder language model. This method shows for the first time that language model predictions are mostly invertible: in many cases, we are able to recover very similar inputs to the original, sometimes getting the input text back exactly.   We additionally explore the feasibility of prompt recovery across a spectrum of real-world access patterns: full next-token probability outputs, top-K probabilities, probabilities per token upon request, and discrete sampling. We find that even in the case where we are only able to observe text output from the model (no probabilities), we can recover enough of the probability distribution to reconstruct the prompt.   Our results show that systems that offer text-only access to a language model reveal information about their prompts. With enough queries, we can extract next-token probabilities at a given position, which can be used to reconstruct the input. Unlike text-based jailbreaks, our dense distribution inversion is less inhibited by post-pretraining reinforcement learning techniques such as RLHF to align the model. We also show that our inversion techniques transfer between models of the same family, and are not affected by language model scaling.   Figure 1: Under the assumption that a language model is offered as a service with a hidden prefix prompt that produces next-word probabilities, the system is trained from samples to invert the language model, i.e. to recover the prompt given language model probabilities for the next token.      2 Related work  Inverting deep embeddings. Several lines of work in computer vision have shown that inputs can be approximately reconstructed from the logits of an image classifier (Mahendran & Vedaldi, 2014; Dosovitskiy & Brox, 2016; Teterwak et\u00a0al., 2021) or from a self-supervised representation vector (Bordes et\u00a0al., 2021). Some recent work (Takahashi et\u00a0al., 2023) has shown that outputs of computer vision models may reveal private information when shared in a federated learning setup. There is also work on inverting representations in NLP: Song & Raghunathan",
            "references": [
                {
                    "title": "WizardCoder: Empowering Code Large Language Models with Evol-Instruct",
                    "abstract": "Code Large Language Models (Code LLMs), such as StarCoder, have demonstrated exceptional performance in code-related tasks. However, most existing models are solely pre-trained on extensive raw code data without instruction fine-tuning. In this paper, we introduce WizardCoder, which empowers Code LLMs with complex instruction fine-tuning, by adapting the Evol-Instruct method to the domain of code. Through comprehensive experiments on four prominent code generation benchmarks, namely HumanEval, HumanEval+, MBPP, and DS-1000, we unveil the exceptional capabilities of our model. It surpasses all other open-source Code LLMs by a substantial margin. Moreover, our model even outperforms the largest closed LLMs, Anthropic's Claude and Google's Bard, on HumanEval and HumanEval+. Our code, model weights, and data are public at https://github.com/nlpxucan/WizardLM"
                },
                {
                    "title": "Sentence Embedding Leaks More Information than You Expect: Generative Embedding Inversion Attack to Recover the Whole Sentence",
                    "abstract": "Sentence-level representations are beneficial for various natural language processing tasks. It is commonly believed that vector representations can capture rich linguistic properties. Currently, large language models (LMs) achieve state-of-the-art performance on sentence embedding. However, some recent works suggest that vector representations from LMs can cause information leakage. In this work, we further investigate the information leakage issue and propose a generative embedding inversion attack (GEIA) that aims to reconstruct input sequences based only on their sentence embeddings. Given the black-box access to a language model, we treat sentence embeddings as initial tokens' representations and train or fine-tune a powerful decoder model to decode the whole sequences directly. We conduct extensive experiments to demonstrate that our generative inversion attack outperforms previous embedding inversion attacks in classification metrics and generates coherent and contextually similar sentences as the original inputs."
                },
                {
                    "title": "LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions",
                    "abstract": "Large language models (LLMs) with instruction fine-tuning demonstrate superior generative capabilities. However, these models are resource-intensive. To alleviate this issue, we explore distilling knowledge from instruction-tuned LLMs into much smaller ones. While other similar works have been done, they are often conducted on a limited set of (usually still large) models and are not accompanied by proper evaluations. To this end, we carefully develop a large set of 2.58M instructions based on both existing and newly-generated instructions. In addition to being sizable, we design our instructions to cover a broad set of topics to ensure diversity. Extensive analysis of our instruction dataset confirms its diversity, and we generate responses for these instructions using gpt-3.5-turbo. Leveraging these instructions, we fine-tune a diverse herd of models, collectively referred to as LaMini-LM, which includes models from both the encoder-decoder and decoder-only families, with varying sizes. We evaluate the performance of our models using automatic metrics on 15 different natural language processing (NLP) benchmarks, as well as through human assessment. We also assess the model for hallucination and toxicity, and for the former, we introduce a new benchmark dataset for hallucination-inducing QA. The results demonstrate that our proposed LaMini-LM models are comparable to strong baselines while being much smaller in size."
                },
                {
                    "title": "WizardLM: Empowering Large Language Models to Follow Complex Instructions",
                    "abstract": "Training large language models (LLMs) with open-domain instruction following data brings colossal success. However, manually creating such instruction data is very time-consuming and labor-intensive. Moreover, humans may struggle to produce high-complexity instructions. In this paper, we show an avenue for creating large amounts of instruction data with varying levels of complexity using LLM instead of humans. Starting with an initial set of instructions, we use our proposed Evol-Instruct to rewrite them step by step into more complex instructions. Then, we mix all generated instruction data to fine-tune LLaMA. We call the resulting model WizardLM. Human evaluations on a complexity-balanced test bed and Vicuna's testset show that instructions from Evol-Instruct are superior to human-created ones. By analyzing the human evaluation results of the high complexity part, we demonstrate that outputs from our WizardLM are preferred to outputs from OpenAI ChatGPT. In GPT-4 automatic evaluation, WizardLM achieves more than 90\\% capacity of ChatGPT on 17 out of 29 skills. Even though WizardLM still lags behind ChatGPT in some aspects, our findings suggest that fine-tuning with AI-evolved instructions is a promising direction for enhancing LLMs. Our code and data are public at https://github.com/nlpxucan/WizardLM"
                },
                {
                    "title": "Breaching FedMD: Image Recovery via Paired-Logits Inversion Attack",
                    "abstract": "Federated Learning with Model Distillation (FedMD) is a nascent collaborative learning paradigm, where only output logits of public datasets are transmitted as distilled knowledge, instead of passing on private model parameters that are susceptible to gradient inversion attacks, a known privacy risk in federated learning. In this paper, we found that even though sharing output logits of public datasets is safer than directly sharing gradients, there still exists a sub-stantial risk of data exposure caused by carefully designed malicious attacks. Our study shows that a malicious server can inject a PLI (Paired-Logits Inversion) attack against FedMD and its variants by training an inversion neural network that exploits the confidence gap between the server and client models. Experiments on multiple facial recognition datasets validate that under FedMD-like schemes, by using paired server-client logits of public datasets only, the malicious server is able to reconstruct private images on all tested benchmarks with a high success rate."
                },
                {
                    "title": "Self-Instruct: Aligning Language Models with Self-Generated Instructions",
                    "abstract": "Large \u201cinstruction-tuned\u201d language models (i.e., finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is often limited in quantity, diversity, and creativity, therefore hindering the generality of the tuned model. We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off their own generations. Our pipeline generates instructions, input, and output samples from a language model, then filters invalid or similar ones before using them to finetune the original model. Applying our method to the vanilla GPT3, we demonstrate a 33% absolute improvement over the original model on Super-NaturalInstructions, on par with the performance of InstructGPT-001, which was trained with private user data and human annotations. For further evaluation, we curate a set of expert-written instructions for novel tasks, and show through human evaluation that tuning GPT3 with Self-Instruct outperforms using existing public instruction datasets by a large margin, leaving only a 5% absolute gap behind InstructGPT-001. Self-Instruct provides an almost annotation-free method for aligning pre-trained language models with instructions, and we release our large synthetic dataset to facilitate future studies on instruction tuning."
                },
                {
                    "title": "Unnatural Instructions: Tuning Language Models with (Almost) No Human Labor",
                    "abstract": "Instruction tuning enables pretrained language models to perform new tasks from inference-time natural language descriptions. These approaches rely on vast amounts of human supervision in the form of crowdsourced datasets or user interactions. In this work, we introduce Unnatural Instructions: a large dataset of creative and diverse instructions, collected with virtually no human labor. We collect 64,000 examples by prompting a language model with three seed examples of instructions and eliciting a fourth. This set is then expanded by prompting the model to rephrase each instruction, creating a total of approximately 240,000 examples of instructions, inputs, and outputs. Experiments show that despite containing a fair amount of noise, training on Unnatural Instructions rivals the effectiveness of training on open-source manually-curated datasets, surpassing the performance of models such as T0++ and Tk-Instruct across various benchmarks. These results demonstrate the potential of model-generated data as a cost-effective alternative to crowdsourcing for dataset expansion and diversification."
                },
                {
                    "title": "Scaling Instruction-Finetuned Language Models",
                    "abstract": "Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models."
                },
                {
                    "title": "Text Revealer: Private Text Reconstruction via Model Inversion Attacks against Transformers",
                    "abstract": "Text classification has become widely used in various natural language processing applications like sentiment analysis. Current applications often use large transformer-based language models to classify input texts. However, there is a lack of systematic study on how much private information can be inverted when publishing models. In this paper, we formulate \\emph{Text Revealer} -- the first model inversion attack for text reconstruction against text classification with transformers. Our attacks faithfully reconstruct private texts included in training data with access to the target model. We leverage an external dataset and GPT-2 to generate the target domain-like fluent text, and then perturb its hidden state optimally with the feedback from the target model. Our extensive experiments demonstrate that our attacks are effective for datasets with different text lengths and can reconstruct private texts with accuracy."
                },
                {
                    "title": "Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks",
                    "abstract": "How well can NLP models generalize to a variety of unseen tasks when provided with task instructions? To address this question, we first introduce Super-NaturalInstructions, a benchmark of 1,616 diverse NLP tasks and their expert-written instructions. Our collection covers 76 distinct task types, including but not limited to classification, extraction, infilling, sequence tagging, text rewriting, and text composition. This large and diverse collection of tasks enables rigorous benchmarking of cross-task generalization under instructions\u2014training models to follow instructions on a subset of tasks and evaluating them on the remaining unseen ones.Furthermore, we build Tk-Instruct, a transformer model trained to follow a variety of in-context instructions (plain language task definitions or k-shot examples). Our experiments show that Tk-Instruct outperforms existing instruction-following models such as InstructGPT by over 9% on our benchmark despite being an order of magnitude smaller. We further analyze generalization as a function of various scaling parameters, such as the number of observed tasks, the number of instances per task, and model sizes. We hope our dataset and model facilitate future progress towards more general-purpose NLP models."
                },
                {
                    "title": "High Fidelity Visualization of What Your Self-Supervised Representation Knows About",
                    "abstract": "Discovering what is learned by neural networks remains a challenge. In self-supervised learning, classification is the most common task used to evaluate how good a representation is. However, relying only on such downstream task can limit our understanding of what information is retained in the representation of a given input. In this work, we showcase the use of a Representation Conditional Diffusion Model (RCDM) to visualize in data space the representations learned by self-supervised models. The use of RCDM is motivated by its ability to generate high-quality samples -- on par with state-of-the-art generative models -- while ensuring that the representations of those samples are faithful i.e. close to the one used for conditioning. By using RCDM to analyze self-supervised models, we are able to clearly show visually that i) SSL (backbone) representation are not invariant to the data augmentations they were trained with -- thus debunking an often restated but mistaken belief; ii) SSL post-projector embeddings appear indeed invariant to these data augmentation, along with many other data symmetries; iii) SSL representations appear more robust to small adversarial perturbation of their inputs than representations trained in a supervised manner; and iv) that SSL-trained representations exhibit an inherent structure that can be explored thanks to RCDM visualization and enables image manipulation."
                },
                {
                    "title": "Understanding invariance via feedforward inversion of discriminatively trained classifiers",
                    "abstract": "A discriminatively trained neural net classifier can fit the training data perfectly if all information about its input other than class membership has been discarded prior to the output layer. Surprisingly, past research has discovered that some extraneous visual detail remains in the logit vector. This finding is based on inversion techniques that map deep embeddings back to images. We explore this phenomenon further using a novel synthesis of methods, yielding a feedforward inversion model that produces remarkably high fidelity reconstructions, qualitatively superior to those of past efforts. When applied to an adversarially robust classifier model, the reconstructions contain sufficient local detail and global structure that they might be confused with the original image in a quick glance, and the object category can clearly be gleaned from the reconstruction. Our approach is based on BigGAN (Brock, 2019), with conditioning on logits instead of one-hot class labels. We use our reconstruction model as a tool for exploring the nature of representations, including: the influence of model architecture and training objectives (specifically robust losses), the forms of invariance that networks achieve, representational differences between correctly and incorrectly classified images, and the effects of manipulating logits and images. We believe that our method can inspire future investigations into the nature of information flow in a neural net and can provide diagnostics for improving discriminative models."
                },
                {
                    "title": "Imitation Attacks and Defenses for Black-box Machine Translation Systems",
                    "abstract": "We consider an adversary looking to steal or attack a black-box machine translation (MT) system, either for financial gain or to exploit model errors. We first show that black-box MT systems can be stolen by querying them with monolingual sentences and training models to imitate their outputs. Using simulated experiments, we demonstrate that MT model stealing is possible even when imitation models have different input data or architectures than their victims. Applying these ideas, we train imitation models that reach within 0.6 BLEU of three production MT systems on both high-resource and low-resource language pairs. We then leverage the similarity of our imitation models to transfer adversarial examples to the production systems. We use gradient-based attacks that expose inputs which lead to semantically-incorrect translations, dropped content, and vulgar model outputs. To mitigate these vulnerabilities, we propose a defense that modifies translation outputs in order to misdirect the optimization of imitation models. This defense degrades imitation model BLEU and attack transfer rates at some cost in BLEU and inference speed."
                },
                {
                    "title": "Information Leakage in Embedding Models",
                    "abstract": "Embeddings are functions that map raw input data to low-dimensional vector representations, while preserving important semantic information about the inputs. Pre-training embeddings on a large amount of unlabeled data and fine-tuning them for downstream tasks is now a de facto standard in achieving state of the art learning in many domains. We demonstrate that embeddings, in addition to encoding generic semantics, often also present a vector that leaks sensitive information about the input data. We develop three classes of attacks to systematically study information that might be leaked by embeddings. First, embedding vectors can be inverted to partially recover some of the input data. As an example, we show that our attacks on popular sentence embeddings recover between 50%--70% of the input words (F1 scores of 0.5--0.7). Second, embeddings may reveal sensitive attributes inherent in inputs and independent of the underlying semantic task at hand. Attributes such as authorship of text can be easily extracted by training an inference model on just a handful of labeled embedding vectors. Third, embedding models leak moderate amount of membership information for infrequent training data inputs. We extensively evaluate our attacks on various state-of-the-art embedding models in the text domain. We also propose and evaluate defenses that can prevent the leakage to some extent at a minor cost in utility."
                },
                {
                    "title": "Thieves on Sesame Street! Model Extraction of BERT-based APIs",
                    "abstract": "We study the problem of model extraction in natural language processing, in which an adversary with only query access to a victim model attempts to reconstruct a local copy of that model. Assuming that both the adversary and victim model fine-tune a large pretrained language model such as BERT (Devlin et al. 2019), we show that the adversary does not need any real training data to successfully mount the attack. In fact, the attacker need not even use grammatical or semantically meaningful queries: we show that random sequences of words coupled with task-specific heuristics form effective queries for model extraction on a diverse set of NLP tasks, including natural language inference and question answering. Our work thus highlights an exploit only made feasible by the shift towards transfer learning methods within the NLP community: for a query budget of a few hundred dollars, an attacker can extract a model that performs only slightly worse than the victim model. Finally, we study two defense strategies against model extraction---membership classification and API watermarking---which while successful against naive adversaries, are ineffective against more sophisticated ones."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "A framework for the extraction of Deep Neural Networks by leveraging public data",
                    "abstract": "Machine learning models trained on confidential datasets are increasingly being deployed for profit. Machine Learning as a Service (MLaaS) has made such models easily accessible to end-users. Prior work has developed model extraction attacks, in which an adversary extracts an approximation of MLaaS models by making black-box queries to it. However, none of these works is able to satisfy all the three essential criteria for practical model extraction: (1) the ability to work on deep learning models, (2) the non-requirement of domain knowledge and (3) the ability to work with a limited query budget. We design a model extraction framework that makes use of active learning and large public datasets to satisfy them. We demonstrate that it is possible to use this framework to steal deep classifiers trained on a variety of datasets from image and text domains. By querying a model via black-box access for its top prediction, our framework improves performance on an average over a uniform noise baseline by 4.70x for image tasks and 2.11x for text tasks respectively, while using only 30% (30,000 samples) of the public dataset at its disposal."
                },
                {
                    "title": "The Curious Case of Neural Text Degeneration",
                    "abstract": "Despite considerable advancements with deep neural language models, the enigma of neural text degeneration persists when these models are tested as text generators. The counter-intuitive empirical observation is that even though the use of likelihood as training objective leads to high quality models for a broad range of language understanding tasks, using likelihood as a decoding objective leads to text that is bland and strangely repetitive. \nIn this paper, we reveal surprising distributional differences between human text and machine text. In addition, we find that decoding strategies alone can dramatically effect the quality of machine text, even when generated from exactly the same neural language model. Our findings motivate Nucleus Sampling, a simple but effective method to draw the best out of neural generation. By sampling text from the dynamic nucleus of the probability distribution, which allows for diversity while effectively truncating the less reliable tail of the distribution, the resulting text better demonstrates the quality of human text, yielding enhanced diversity without sacrificing fluency and coherence."
                },
                {
                    "title": "Membership Inference Attacks Against Machine Learning Models",
                    "abstract": "We quantitatively investigate how machine learning models leak information about the individual data records on which they were trained. We focus on the basic membership inference attack: given a data record and black-box access to a model, determine if the record was in the model's training dataset. To perform membership inference against a target model, we make adversarial use of machine learning and train our own inference model to recognize differences in the target model's predictions on the inputs that it trained on versus the inputs that it did not train on. We empirically evaluate our inference techniques on classification models trained by commercial \"machine learning as a service\" providers such as Google and Amazon. Using realistic datasets and classification tasks, including a hospital discharge dataset whose membership is sensitive from the privacy perspective, we show that these models can be vulnerable to membership inference attacks. We then investigate the factors that influence this leakage and evaluate mitigation strategies."
                },
                {
                    "title": "Pointer Sentinel Mixture Models",
                    "abstract": "Recent neural network sequence models with softmax classifiers have achieved their best language modeling performance only with very large hidden states and large vocabularies. Even then they struggle to predict rare or unseen words even if the context makes the prediction unambiguous. We introduce the pointer sentinel mixture architecture for neural sequence models which has the ability to either reproduce a word from the recent context or produce a word from a standard softmax classifier. Our pointer sentinel-LSTM model achieves state of the art language modeling performance on the Penn Treebank (70.9 perplexity) while using far fewer parameters than a standard softmax LSTM. In order to evaluate how well language models can exploit longer contexts and deal with more realistic vocabularies and larger corpora we also introduce the freely available WikiText corpus."
                },
                {
                    "title": "Stealing Machine Learning Models via Prediction APIs",
                    "abstract": "Machine learning (ML) models may be deemed confidential due to their sensitive training data, commercial value, or use in security applications. Increasingly often, confidential ML models are being deployed with publicly accessible query interfaces. ML-as-a-service (\"predictive analytics\") systems are an example: Some allow users to train models on potentially sensitive data and charge others for access on a pay-per-query basis. \nThe tension between model confidentiality and public access motivates our investigation of model extraction attacks. In such attacks, an adversary with black-box access, but no prior knowledge of an ML model's parameters or training data, aims to duplicate the functionality of (i.e., \"steal\") the model. Unlike in classical learning theory settings, ML-as-a-service offerings may accept partial feature vectors as inputs and include confidence values with predictions. Given these practices, we show simple, efficient attacks that extract target ML models with near-perfect fidelity for popular model classes including logistic regression, neural networks, and decision trees. We demonstrate these attacks against the online services of BigML and Amazon Machine Learning. We further show that the natural countermeasure of omitting confidence values from model outputs still admits potentially harmful model extraction attacks. Our results highlight the need for careful ML model deployment and new model extraction countermeasures."
                },
                {
                    "title": "Neural Text Generation from Structured Data with Application to the Biography Domain",
                    "abstract": "This paper introduces a neural model for concept-to-text generation that scales to large, rich domains. We experiment with a new dataset of biographies from Wikipedia that is an order of magnitude larger than existing resources with over 700k samples. The dataset is also vastly more diverse with a 400k vocabulary, compared to a few hundred words for Weathergov or Robocup. Our model builds upon recent work on conditional neural language model for text generation. To deal with the large vocabulary, we extend these models to mix a fixed vocabulary with copy actions that transfer sample-specific words from the input database to the generated output sentence. Our neural model significantly out-performs a classical Kneser-Ney language model adapted to this task by nearly 15 BLEU."
                },
                {
                    "title": "Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures",
                    "abstract": "Machine-learning (ML) algorithms are increasingly utilized in privacy-sensitive applications such as predicting lifestyle choices, making medical diagnoses, and facial recognition. In a model inversion attack, recently introduced in a case study of linear classifiers in personalized medicine by Fredrikson et al., adversarial access to an ML model is abused to learn sensitive genomic information about individuals. Whether model inversion attacks apply to settings outside theirs, however, is unknown. We develop a new class of model inversion attack that exploits confidence values revealed along with predictions. Our new attacks are applicable in a variety of settings, and we explore two in depth: decision trees for lifestyle surveys as used on machine-learning-as-a-service systems and neural networks for facial recognition. In both cases confidence values are revealed to those with the ability to make prediction queries to models. We experimentally show attacks that are able to estimate whether a respondent in a lifestyle survey admitted to cheating on their significant other and, in the other context, show how to recover recognizable images of people's faces given only their name and access to the ML model. We also initiate experimental exploration of natural countermeasures, investigating a privacy-aware decision tree training algorithm that is a simple variant of CART learning, as well as revealing only rounded confidence values. The lesson that emerges is that one can avoid these kinds of MI attacks with negligible degradation to utility."
                },
                {
                    "title": "Inverting Visual Representations with Convolutional Networks",
                    "abstract": "Feature representations, both hand-designed and learned ones, are often hard to analyze and interpret, even when they are extracted from visual data. We propose a new approach to study image representations by inverting them with an up-convolutional neural network. We apply the method to shallow representations (HOG, SIFT, LBP), as well as to deep networks. For shallow representations our approach provides significantly better reconstructions than existing methods, revealing that there is surprisingly rich information contained in these features. Inverting a deep network trained on ImageNet provides several insights into the properties of the feature representation learned by the network. Most strikingly, the colors and the rough contours of an image can be reconstructed from activations in higher network layers and even from the predicted class probabilities."
                },
                {
                    "title": "Understanding deep image representations by inverting them",
                    "abstract": "Image representations, from SIFT and Bag of Visual Words to Convolutional Neural Networks (CNNs), are a crucial component of almost any image understanding system. Nevertheless, our understanding of them remains limited. In this paper we conduct a direct analysis of the visual information contained in representations by asking the following question: given an encoding of an image, to which extent is it possible to reconstruct the image itself? To answer this question we contribute a general framework to invert representations. We show that this method can invert representations such as HOG more accurately than recent alternatives while being applicable to CNNs too. We then use this technique to study the inverse of recent state-of-the-art CNN image representations for the first time. Among our findings, we show that several layers in CNNs retain photographically accurate information about the image, with different degrees of geometric and photometric invariance."
                },
                {
                    "title": "Privacy in Pharmacogenetics: An End-to-End Case Study of Personalized Warfarin Dosing",
                    "abstract": "We initiate the study of privacy in pharmacogenetics, wherein machine learning models are used to guide medical treatments based on a patient's genotype and background. Performing an in-depth case study on privacy in personalized warfarin dosing, we show that suggested models carry privacy risks, in particular because attackers can perform what we call model inversion: an attacker, given the model and some demographic information about a patient, can predict the patient's genetic markers. As differential privacy (DP) is an oft-proposed solution for medical settings such as this, we evaluate its effectiveness for building private versions of pharmacogenetic models. We show that DP mechanisms prevent our model inversion attacks when the privacy budget is carefully selected. We go on to analyze the impact on utility by performing simulated clinical trials with DP dosing models. We find that for privacy budgets effective at preventing attacks, patients would be exposed to increased risk of stroke, bleeding events, and mortality. We conclude that current DP mechanisms do not simultaneously improve genomic privacy while retaining desirable clinical efficacy, highlighting the need for new mechanisms that should be evaluated in situ using the general methodology introduced by our work."
                },
                {
                    "title": "Adversarial learning",
                    "abstract": "Many classification tasks, such as spam filtering, intrusion detection, and terrorism detection, are complicated by an adversary who wishes to avoid detection. Previous work on adversarial classification has made the unrealistic assumption that the attacker has perfect knowledge of the classifier [2]. In this paper, we introduce the adversarial classifier reverse engineering (ACRE) learning problem, the task of learning sufficient information about a classifier to construct adversarial attacks. We present efficient algorithms for reverse engineering linear classifiers with either continuous or Boolean features and demonstrate their effectiveness using real data from the domain of spam filtering."
                },
                {
                    "title": "Bleu: a Method for Automatic Evaluation of Machine Translation",
                    "abstract": "Human evaluations of machine translation are extensive but expensive. Human evaluations can take months to finish and involve human labor that can not be reused. We propose a method of automatic machine translation evaluation that is quick, inexpensive, and language-independent, that correlates highly with human evaluation, and that has little marginal cost per run. We present this method as an automated understudy to skilled human judges which substitutes for them when there is need for quick or frequent evaluations."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "A case study:",
                    "abstract": "Ciprian Radu and Catalin Danaila RKB Bearing Industries Switzerland Using sophisticated CAD 2D/3D tools, in-house developed computing software and FEM analysis allows to improve bearing quality, reliability and delivery time. The first part of this article is a simple introduction to the preliminary stages of 2D drafting, 3D modeling and calculation of the main coefficients for two spherical roller bearings that are to be installed on an oil platform rack and pinion jacking system. The second part of this article will cover in detail the steps of FEM analysis for development and validation of the above application. CAD and FEM analysis improve bearing reliability"
                },
                {
                    "title": "On the Privacy Risk of In-context Learning",
                    "abstract": "Large language models (LLMs) are excellent few-shot learners. They can perform a wide variety of tasks purely based on natural language prompts provided to them. These prompts contain data of a specific downstream task\u2014often the private dataset of a party, e.g., a company that wants to leverage the LLM on their purposes. We show that deploying prompted models presents a significant privacy risk for the data used within the prompt by instantiating a highly effective membership inference attack. We also observe that the privacy risk of prompted models exceeds fine-tuned models at the same utility levels. After identifying the model\u2019s sensitivity to their prompts\u2014in form of a significantly higher prediction confidence on the prompted data\u2014as a cause for the increased risk, we propose ensembling as a mitigation strategy. By aggregating over multiple different versions of a prompted model, membership inference risk can be decreased."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan we reconstruct the input prompt of a language model using its next-token probability outputs, particularly in scenarios where the prompt is not fully accessible?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of prompt reconstruction has significant implications for the research community, particularly in understanding the privacy and security of language models offered as a service. By demonstrating that language model predictions can be inverted, this research could lead to advancements in safeguarding user prompts and enhancing model transparency. Furthermore, it opens avenues for future research on the robustness of language models against information leakage and could inform the development of more secure AI systems. Practical applications may include improved privacy-preserving techniques in AI services and better understanding of model behavior, which is crucial for responsible AI deployment.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of language models and the nature of their probabilistic outputs. Naive approaches may fail because they do not account for the high-dimensional and often noisy nature of the next-token probabilities, which can obscure the original input. Additionally, the task requires overcoming technical obstacles such as accurately interpreting the distribution of probabilities and effectively mapping them back to the original prompt. Theoretical challenges include understanding the limits of information retention in the model's predictions and the potential for ambiguity in the reconstructed prompts.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the outputs of language models without addressing the specific challenge of prompt reconstruction. Limitations in existing solutions include a lack of methodologies that effectively leverage the next-token probabilities for inversion purposes. Barriers such as the complexity of language model architectures and the absence of a formalized approach to this problem have hindered progress. Our approach differs by formalizing the problem of language model inversion and proposing a novel architecture that utilizes the distribution vector to reconstruct prompts, demonstrating that language model predictions are largely invertible.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using a pretrained encoder-decoder language model to \"unroll\" the next-token probability distribution into a sequence that can be processed for prompt reconstruction. We will evaluate our approach using various datasets and metrics, including the accuracy of the reconstructed prompts compared to the original inputs. Expected outcomes include demonstrating the feasibility of prompt recovery across different access patterns (e.g., full next-token probabilities, top-K probabilities"
            }
        },
        "author_data": {
            "94888897-be0d-4a07-8bd8-f3ff51114973": {
                "pk": "94888897-be0d-4a07-8bd8-f3ff51114973",
                "name": "John X. Morris",
                "collaborators": [
                    "Alexander M. Rush",
                    "Vitaly Shmatikov",
                    "Collin Zhang",
                    "Tingwei Zhang",
                    "Eugene Bagdasaryan",
                    "Chandan Singh",
                    "Jianfeng Gao",
                    "Yuntian Deng",
                    "Volodymyr Kuleshov"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Adversarial Machine Learning",
                    "Embedding Techniques",
                    "Prompt Engineering"
                ],
                "publications": [
                    {
                        "title": "Contextual Document Embeddings",
                        "abstract": "Dense document embeddings are central to neural retrieval. The dominant paradigm is to train and construct embeddings by running encoders directly on individual documents. In this work, we argue that these embeddings, while effective, are implicitly out-of-context for targeted use cases of retrieval, and that a contextualized document embedding should take into account both the document and neighboring documents in context - analogous to contextualized word embeddings. We propose two complementary methods for contextualized document embeddings: first, an alternative contrastive learning objective that explicitly incorporates the document neighbors into the intra-batch contextual loss; second, a new contextual architecture that explicitly encodes neighbor document information into the encoded representation. Results show that both methods achieve better performance than biencoders in several settings, with differences especially pronounced out-of-domain. We achieve state-of-the-art results on the MTEB benchmark with no hard negative mining, score distillation, dataset-specific instructions, intra-GPU example-sharing, or extremely large batch sizes. Our method can be applied to improve performance on any contrastive learning dataset and any biencoder."
                    },
                    {
                        "title": "Soft Prompts Go Hard: Steering Visual Language Models with Hidden Meta-Instructions",
                        "abstract": "We introduce a new type of indirect injection attacks against language models that operate on images: hidden ''meta-instructions'' that influence how the model interprets the image and steer the model's outputs to express an adversary-chosen style, sentiment, or point of view. We explain how to create meta-instructions by generating images that act as soft prompts. In contrast to jailbreaking attacks and adversarial examples, outputs produced in response to these images are plausible and based on the visual content of the image, yet also satisfy the adversary's (meta-)objective. We evaluate the efficacy of meta-instructions for multiple visual language models and adversarial meta-objectives, and demonstrate how they can ''unlock'' capabilities of the underlying language models that are unavailable via explicit text instructions. We describe how meta-instruction attacks could cause harm by enabling creation of malicious, self-interpreting content that carries spam, misinformation, and spin. Finally, we discuss defenses."
                    },
                    {
                        "title": "Extracting Prompts by Inverting LLM Outputs",
                        "abstract": "We consider the problem of language model inversion: given outputs of a language model, we seek to extract the prompt that generated these outputs. We develop a new black-box method, output2prompt, that learns to extract prompts without access to the model's logits and without adversarial or jailbreaking queries. In contrast to previous work, output2prompt only needs outputs of normal user queries. To improve memory efficiency, output2prompt employs a new sparse encoding techique. We measure the efficacy of output2prompt on a variety of user and system prompts and demonstrate zero-shot transferability across different LLMs."
                    },
                    {
                        "title": "Tree Prompting: Efficient Task Adaptation without Fine-Tuning",
                        "abstract": "Prompting language models (LMs) is the main interface for applying them to new tasks. However, for smaller LMs, prompting provides low accuracy compared to gradient-based finetuning. Tree Prompting is an approach to prompting which builds a decision tree of prompts, linking multiple LM calls together to solve a task. At inference time, each call to the LM is determined by efficiently routing the outcome of the previous call using the tree. Experiments on classification datasets show that Tree Prompting improves accuracy over competing methods and is competitive with fine-tuning. We also show that variants of Tree Prompting allow inspection of a model's decision-making process."
                    },
                    {
                        "title": "Text Embeddings Reveal (Almost) As Much As Text",
                        "abstract": "How much private information do text embeddings reveal about the original text? We investigate the problem of embedding \\textit{inversion}, reconstructing the full text represented in dense text embeddings. We frame the problem as controlled generation: generating text that, when reembedded, is close to a fixed point in latent space. We find that although a na\\\"ive model conditioned on the embedding performs poorly, a multi-step method that iteratively corrects and re-embeds text is able to recover $92\\%$ of $32\\text{-token}$ text inputs exactly. We train our model to decode text embeddings from two state-of-the-art embedding models, and also show that our model can recover important personal information (full names) from a dataset of clinical notes. Our code is available on Github: \\href{https://github.com/jxmorris12/vec2text}{github.com/jxmorris12/vec2text}."
                    }
                ]
            },
            "0b7ee692-131c-4df4-936e-a9b6778cfabd": {
                "pk": "0b7ee692-131c-4df4-936e-a9b6778cfabd",
                "name": "Wenting Zhao",
                "collaborators": [
                    "Alexander M. Rush",
                    "Shangyi Geng",
                    "Yi Lu",
                    "Jing Nathan Yan",
                    "Songlin Yang",
                    "Justin T. Chiu",
                    "Siyu Ren",
                    "Fei Yuan",
                    "Zhiyong Wu",
                    "Justin T Chiu",
                    "Derek Chen",
                    "Saujas Vaduguru",
                    "Daniel Fried"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Language Models",
                    "Dialogue Systems",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Great Memory, Shallow Reasoning: Limits of kNN-LMs",
                        "abstract": "$K$-nearest neighbor language models ($k$NN-LMs), which integrate retrieval with next-word prediction, have demonstrated strong performance in language modeling as well as downstream NLP benchmarks. These results have led researchers to argue that models trained on poor quality or outdated data could perform well by employing a $k$NN extension that has access to a higher-quality datastore. In this work, we ask whether this improved ability to recall information really translates into downstream abilities. We extensively evaluate $k$NN-LMs on a diverse set of tasks, ranging from sentiment classification and commonsense reasoning to multi-hop reasoning. Results show that $k$NN-LMs excel at memory-intensive tasks, where utilizing the patterns in the input is sufficient for determining the output, but struggle with reasoning tasks that require integrating multiple pieces of information to derive new knowledge. We further demonstrate through oracle experiments and qualitative analysis that even with perfect retrieval, $k$NN-LMs still fail to determine the correct answers, placing an upper bound on their reasoning performance. Code and datastores are released at https://github.com/GSYfate/knnlm-limits/."
                    },
                    {
                        "title": "A Controlled Study on Long Context Extension and Generalization in LLMs",
                        "abstract": "Broad textual understanding and in-context learning require language models that utilize full document contexts. Due to the implementation challenges associated with directly training long-context models, many methods have been proposed for extending models to handle long contexts. However, owing to differences in data and model classes, it has been challenging to compare these approaches, leading to uncertainty as to how to evaluate long-context performance and whether it differs from standard evaluation. We implement a controlled protocol for extension methods with a standardized evaluation, utilizing consistent base models and extension data. Our study yields several insights into long-context behavior. First, we reaffirm the critical role of perplexity as a general-purpose performance indicator even in longer-context tasks. Second, we find that current approximate attention methods systematically underperform across long-context tasks. Finally, we confirm that exact fine-tuning based methods are generally effective within the range of their extension, whereas extrapolation remains challenging. All codebases, models, and checkpoints will be made available open-source, promoting transparency and facilitating further research in this critical area of AI development."
                    },
                    {
                        "title": "Symbolic Planning and Code Generation for Grounded Dialogue",
                        "abstract": "Large language models (LLMs) excel at processing and generating both text and code. However, LLMs have had limited applicability in grounded task-oriented dialogue as they are difficult to steer toward task objectives and fail to handle novel grounding. We present a modular and interpretable grounded dialogue system that addresses these shortcomings by composing LLMs with a symbolic planner and grounded code execution. Our system consists of a reader and planner: the reader leverages an LLM to convert partner utterances into executable code, calling functions that perform grounding. The translated code\u2019s output is stored to track dialogue state, while a symbolic planner determines the next appropriate response. We evaluate our system\u2019s performance on the demanding OneCommon dialogue task, involving collaborative reference resolution on abstract images of scattered dots. Our system substantially outperforms the previous state-of-the-art, including improving task success in human evaluations from 56% to 69% in the most challenging setting."
                    }
                ]
            },
            "0ec2c240-688f-4847-87e4-0123283f33bd": {
                "pk": "0ec2c240-688f-4847-87e4-0123283f33bd",
                "name": "Justin T. Chiu",
                "collaborators": [
                    "Alexander M. Rush",
                    "Wenting Zhao",
                    "Yuntian Deng",
                    "Claire Cardie",
                    "Daniel Fried",
                    "Zhiying Xu",
                    "Francis Y. Yan",
                    "Rachee Singh",
                    "Minlan Yu",
                    "Demi Guo",
                    "Derek Chen",
                    "Saujas Vaduguru",
                    "John X. Morris",
                    "Ramin Zabih",
                    "D. Klein",
                    "F. Obermeyer",
                    "Eli Bingham",
                    "M. Jankowiak",
                    "Neeraj Pradhan",
                    "Noah D. Goodman",
                    "Yoon Kim",
                    "Wenling Shang",
                    "Kihyuk Sohn"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Dialogue Systems",
                    "Explainable AI"
                ],
                "publications": [
                    {
                        "title": "Abductive Commonsense Reasoning Exploiting Mutually Exclusive Explanations",
                        "abstract": "Abductive reasoning aims to find plausible explanations for an event. This style of reasoning is critical for commonsense tasks where there are often multiple plausible explanations. Existing approaches for abductive reasoning in natural language processing (NLP) often rely on manually generated annotations for supervision; however, such annotations can be subjective and biased. Instead of using direct supervision, this work proposes an approach for abductive commonsense reasoning that exploits the fact that only a subset of explanations is correct for a given context. The method uses posterior regularization to enforce a mutual exclusion constraint, encouraging the model to learn the distinction between fluent explanations and plausible ones. We evaluate our approach on a diverse set of abductive reasoning datasets; experimental results show that our approach outperforms or is comparable to directly applying pretrained language models in a zero-shot manner and other knowledge-augmented zero-shot methods."
                    },
                    {
                        "title": "Symbolic Planning and Code Generation for Grounded Dialogue",
                        "abstract": "Large language models (LLMs) excel at processing and generating both text and code. However, LLMs have had limited applicability in grounded task-oriented dialogue as they are difficult to steer toward task objectives and fail to handle novel grounding. We present a modular and interpretable grounded dialogue system that addresses these shortcomings by composing LLMs with a symbolic planner and grounded code execution. Our system consists of a reader and planner: the reader leverages an LLM to convert partner utterances into executable code, calling functions that perform grounding. The translated code\u2019s output is stored to track dialogue state, while a symbolic planner determines the next appropriate response. We evaluate our system\u2019s performance on the demanding OneCommon dialogue task, involving collaborative reference resolution on abstract images of scattered dots. Our system substantially outperforms the previous state-of-the-art, including improving task success in human evaluations from 56% to 69% in the most challenging setting."
                    },
                    {
                        "title": "HOP, UNION, GENERATE: Explainable Multi-hop Reasoning without Rationale Supervision",
                        "abstract": "Explainable multi-hop question answering (QA) not only predicts answers but also identifies rationales, i. e. subsets of input sentences used to derive the answers. This problem has been extensively studied under the supervised setting, where both answer and rationale annotations are given. Because rationale annotations are expensive to collect and not always available, recent efforts have been devoted to developing methods that do not rely on supervision for rationales. However, such methods have limited capacities in modeling interactions between sentences, let alone reasoning across multiple documents. This work proposes a principled, probabilistic approach for training explainable multi-hop QA systems without rationale supervision. Our approach performs multi-hop reasoning by explicitly modeling rationales as sets, enabling the model to capture interactions between documents and sentences within a document. Experimental results show that our approach is more accurate at selecting rationales than the previous methods, while maintaining similar accuracy in predicting answers."
                    },
                    {
                        "title": "Low-Rank Constraints for Fast Inference in Structured Models",
                        "abstract": "Structured distributions, i.e. distributions over combinatorial spaces, are commonly used to learn latent probabilistic representations from observed data. However, scaling these models is bottlenecked by the high computational and memory complexity with respect to the size of the latent representations. Common models such as Hidden Markov Models (HMMs) and Probabilistic Context-Free Grammars (PCFGs) require time and space quadratic and cubic in the number of hidden states respectively. This work demonstrates a simple approach to reduce the computational and memory complexity of a large class of structured models. We show that by viewing the central inference step as a matrix-vector product and using a low-rank constraint, we can trade off model expressivity and speed via the rank. Experiments with neural parameterized structured models for language modeling, polyphonic music modeling, unsupervised grammar induction, and video modeling show that our approach matches the accuracy of standard models at large state spaces while providing practical speedups."
                    },
                    {
                        "title": "Modeling Perspective-Dependent Ambiguity in Grounded Collaborative Dialogue",
                        "abstract": "Errors in reference generation and resolution can occur if a dialogue agent reasons incorrectly about their partner\u2019s perspective. We present, in a collaborative and visually grounded setting, a dialogue planner that infers its partner\u2019s perspective to produce referring expressions that the partner resolves correctly. The dialogue planner models the partner perspective as a latent variable, embedded in a partner model that is used for both model-based planning and incorporating evidence from partner responses. We validate our approach on O NE C OMMON , a challenging dialogue game where players have large differences in their perspectives. In symbolic selfplay, where agents partner with a copy of themselves using symbolic communication, the dialogue planner has promising performance, winning 85.2% of games."
                    },
                    {
                        "title": "Teal: Learning-Accelerated Optimization of WAN Traffic Engineering",
                        "abstract": "The rapid expansion of global cloud wide-area networks (WANs) has posed a challenge for commercial optimization engines to efficiently solve network traffic engineering (TE) problems at scale. Existing acceleration strategies decompose TE optimization into concurrent subproblems but realize limited parallelism due to an inherent tradeoff between run time and allocation performance. We present Teal, a learning-based TE algorithm that leverages the parallel processing power of GPUs to accelerate TE control. First, Teal designs a flow-centric graph neural network (GNN) to capture WAN connectivity and network flows, learning flow features as inputs to downstream allocation. Second, to reduce the problem scale and make learning tractable, Teal employs a multi-agent reinforcement learning (RL) algorithm to independently allocate each traffic demand while optimizing a central TE objective. Finally, Teal fine-tunes allocations with ADMM (Alternating Direction Method of Multipliers), a highly parallelizable optimization algorithm for reducing constraint violations such as overutilized links. We evaluate Teal using traffic matrices from Microsoft's WAN. On a large WAN topology with >1,700 nodes, Teal generates near-optimal flow allocations while running several orders of magnitude faster than the production optimization engine. Compared with other TE acceleration schemes, Teal satisfies 6--32% more traffic demand and yields 197--625\u00d7 speedups."
                    },
                    {
                        "title": "Teal: Learning-Accelerated Optimization of Traffic Engineering",
                        "abstract": "In the last decade, global cloud wide-area networks (WANs) have grown 10 \u00d7 in size due to the deployment of new network sites and datacenters, making it challenging for commercial optimization engines to solve the network traf\ufb01c engineering (TE) problem within the temporal budget of a few minutes. In this work, we show that carefully designed deep learning models are key to accelerating the running time of intra-WAN TE systems for large deployments since deep learning is both massively parallel and it bene\ufb01ts from the wealth of historical traf\ufb01c allocation data from production WANs. However, off-the-shelf deep learning methods fail to perform well on the TE task since they ignore the effects of network connectivity on \ufb02ow allocations. They are also faced with a tractability challenge posed by the large problem scale of TE optimization. Moreover, neural networks do not have mechanisms to readily enforce hard constraints on model outputs (e.g., link capacity constraints). We tackle these challenges by designing a deep learning-based TE system\u2014T EAL . First, T EAL leverages graph neural networks (GNN) to faithfully capture connectivity and model network \ufb02ows. Second, T EAL devises a multi-agent reinforcement learning (RL) algorithm to process individual demands independently in parallel to lower the problem scale. Finally, T EAL reduces link capacity violations and improves solution quality using the alternating direction method of multipliers (ADMM). We evaluate T EAL on traf\ufb01c matrices of a global commercial cloud provider and \ufb01nd that T EAL computes near-optimal traf\ufb01c allocations with a 59 \u00d7 speedup over state-of-the-art TE systems on a WAN topology of over 1,500 nodes."
                    },
                    {
                        "title": "Unsupervised Text Deidentification",
                        "abstract": "Deidentification seeks to anonymize textual data prior to distribution. Automatic deidentification primarily uses supervised named entity recognition from human-labeled data points. We propose an unsupervised deidentification method that masks words that leak personally-identifying information. The approach utilizes a specially trained reidentification model to identify individuals from redacted personal documents. Motivated by K-anonymity based privacy, we generate redactions that ensure a minimum reidentification rank for the correct profile of the document. To evaluate this approach, we consider the task of deidentifying Wikipedia Biographies, and evaluate using an adversarial reidentification metric. Compared to a set of unsupervised baselines, our approach deidentifies documents more completely while removing fewer words. Qualitatively, we see that the approach eliminates many identifying aspects that would fall outside of the common named entity based approach."
                    },
                    {
                        "title": "Reference-Centric Models for Grounded Collaborative Dialogue",
                        "abstract": "We present a grounded neural dialogue model that successfully collaborates with people in a partially-observable reference game. We focus on a setting where two agents each observe an overlapping part of a world context and need to identify and agree on some object they share. Therefore, the agents should pool their information and communicate pragmatically to solve the task. Our dialogue agent accurately grounds referents from the partner\u2019s utterances using a structured reference resolver, conditions on these referents using a recurrent memory, and uses a pragmatic generation procedure to ensure the partner can resolve the references the agent produces. We evaluate on the OneCommon spatial grounding dialogue task (Udagawa and Aizawa 2019), involving a number of dots arranged on a board with continuously varying positions, sizes, and shades. Our agent substantially outperforms the previous state of the art for the task, obtaining a 20% relative improvement in successful task completion in self-play evaluations and a 50% relative improvement in success in human evaluations."
                    },
                    {
                        "title": "Scaling Hidden Markov Language Models",
                        "abstract": "The hidden Markov model (HMM) is a fundamental tool for sequence modeling that cleanly separates the hidden state from the emission structure. However, this separation makes it difficult to fit HMMs to large datasets in modern NLP, and they have fallen out of use due to very poor performance compared to fully observed models. This work revisits the challenge of scaling HMMs to language modeling datasets, taking ideas from recent approaches to neural modeling. We propose methods for scaling HMMs to massive state spaces while maintaining efficient exact inference, a compact parameterization, and effective regularization. Experiments show that this approach leads to models that are more accurate than previous HMM and n-gram-based methods, making progress towards the performance of state-of-the-art neural models."
                    },
                    {
                        "title": "Tensor Variable Elimination for Plated Factor Graphs",
                        "abstract": "A wide class of machine learning algorithms can be reduced to variable elimination on factor graphs. While factor graphs provide a unifying notation for these algorithms, they do not provide a compact way to express repeated structure when compared to plate diagrams for directed graphical models. To exploit efficient tensor algebra in graphs with plates of variables, we generalize undirected factor graphs to plated factor graphs and variable elimination to a tensor variable elimination algorithm that operates directly on plated factor graphs. Moreover, we generalize complexity bounds based on treewidth and characterize the class of plated factor graphs for which inference is tractable. As an application, we integrate tensor variable elimination into the Pyro probabilistic programming language to enable exact inference in discrete latent variable models with repeated structure. We validate our methods with experiments on both directed and undirected graphical models, including applications to polyphonic music modeling, animal movement modeling, and latent sentiment analysis."
                    },
                    {
                        "title": "Latent Alignment and Variational Attention",
                        "abstract": "Neural attention has become central to many state-of-the-art models in natural language processing and related domains. Attention networks are an easy-to-train and effective method for softly simulating alignment; however, the approach does not marginalize over latent alignments in a probabilistic sense. This property makes it difficult to compare attention to other alignment approaches, to compose it with probabilistic models, and to perform posterior inference conditioned on observed data. A related latent approach, hard attention, fixes these issues, but is generally harder to train and less accurate. This work considers variational attention networks, alternatives to soft and hard attention for learning latent variable alignment models, with tighter approximation bounds based on amortized variational inference. We further propose methods for reducing the variance of gradients to make these approaches computationally feasible. Experiments show that for machine translation and visual question answering, inefficient exact latent variable models outperform standard neural attention, but these gains go away when using hard attention based training. On the other hand, variational attention retains most of the performance gain but with training speed comparable to neural attention."
                    },
                    {
                        "title": "2 Background : Latent Alignment and Neural Attention",
                        "abstract": "Neural attention has become central to many state-of-the-art models in natural language processing and related domains. Attention networks are an easy-to-train and effective method for softly simulating alignment; however, the approach does not marginalize over latent alignments in a probabilistic sense. This property makes it difficult to compare attention to other alignment approaches, to compose it with probabilistic models, and to perform posterior inference conditioned on observed data. A related latent approach, hard attention, fixes these issues, but is generally harder to train and less accurate. This work considers variational attention networks, alternatives to soft and hard attention for learning latent variable alignment models, with tighter approximation bounds based on amortized variational inference. We further propose methods for reducing the variance of gradients to make these approaches computationally feasible. Experiments show that for machine translation and visual question answering, inefficient exact latent variable models outperform standard neural attention, but these gains go away when using hard attention based training. On the other hand, variational attention retains most of the performance gain but with training speed comparable to neural attention."
                    },
                    {
                        "title": "Exploring Normalization in Deep Residual Networks with Concatenated Rectified Linear Units",
                        "abstract": "    Deep Residual Networks (ResNets) have recently achieved state-of-the-art results on many challenging computer vision tasks. In this work we analyze the role of Batch Normalization (BatchNorm) layers on ResNets in the hope of improving the current architecture and better incorporating other normalization techniques, such as Normalization Propagation (NormProp), into ResNets. Firstly, we verify that BatchNorm helps distribute representation learning to residual blocks at all layers, as opposed to a plain ResNet without BatchNorm where learning happens mostly in the latter part of the network. We also observe that BatchNorm well regularizes Concatenated ReLU (CReLU) activation scheme on ResNets, whose magnitude of activation grows by preserving both positive and negative responses when going deeper into the network. Secondly, we investigate the use of NormProp as a replacement for BatchNorm in ResNets. Though NormProp theoretically attains the same effect as BatchNorm on generic convolutional neural networks, the identity mapping of ResNets invalidates its theoretical promise and NormProp exhibits a significant performance drop when naively applied. To bridge the gap between BatchNorm and NormProp in ResNets, we propose a simple modification to NormProp and employ the CReLU activation scheme. We experiment on visual object recognition benchmark datasets such as CIFAR-10/100 and ImageNet and demonstrate that 1) the modified NormProp performs better than the original NormProp but is still not comparable to BatchNorm and 2) CReLU improves the performance of ResNets with or without normalizations.   "
                    }
                ]
            },
            "6791d851-e610-4ba0-879c-b2849901044f": {
                "pk": "6791d851-e610-4ba0-879c-b2849901044f",
                "name": "Vitaly Shmatikov",
                "collaborators": [
                    "Eugene Bagdasarian",
                    "Zhen Sun",
                    "R. Schuster",
                    "Tsung-Yin Hsieh",
                    "Ben Nassi",
                    "John X. Morris",
                    "Volodymyr Kuleshov",
                    "Alexander M. Rush",
                    "Rishi Jha",
                    "Tingwei Zhang",
                    "D. Boneh",
                    "Dieter Gollmann",
                    "Dan Halperin",
                    "John Manferdelli",
                    "John Mitchell",
                    "Bennet Yee",
                    "Emily Wenger",
                    "Xiuyu Li",
                    "Ben Y. Zhao",
                    "Jialing Pei",
                    "Congzheng Song",
                    "Eran Tromer",
                    "Tao Yu",
                    "Tal Schuster",
                    "Yoav Meri"
                ],
                "domain": [
                    "Machine Learning",
                    "Security",
                    "Adversarial Attacks",
                    "Data Privacy"
                ],
                "publications": [
                    {
                        "title": "Differential Degradation Vulnerabilities in Censorship Circumvention Systems",
                        "abstract": "Several recently proposed censorship circumvention systems use encrypted network channels of popular applications to hide their communications. For example, a Tor pluggable transport called Snowflake uses the WebRTC data channel, while a system called Protozoa substitutes content in a WebRTC video-call application. By using the same channel as the cover application and (in the case of Protozoa) matching its observable traffic characteristics, these systems aim to resist powerful network-based censors capable of large-scale traffic analysis. Protozoa, in particular, achieves a strong indistinguishability property known as behavioral independence. We demonstrate that this class of systems is generically vulnerable to a new type of active attacks we call\"differential degradation.\"These attacks do not require multi-flow measurements or traffic classification and are thus available to all real-world censors. They exploit the discrepancies between the respective network requirements of the circumvention system and its cover application. We show how a censor can use the minimal application-level information exposed by WebRTC to create network conditions that cause the circumvention system to suffer a much bigger degradation in performance than the cover application. Even when the attack causes no observable differences in network traffic and behavioral independence still holds, the censor can block circumvention at a low cost, without resorting to traffic analysis, and with minimal collateral damage to non-circumvention users. We present effective differential degradation attacks against Snowflake and Protozoa. We explain the root cause of these vulnerabilities, analyze the tradeoffs faced by the designers of circumvention systems, and propose a modified version of Protozoa that resists differential degradation attacks."
                    },
                    {
                        "title": "Mithridates: Boosting Natural Resistance to Backdoor Learning",
                        "abstract": "Machine learning (ML) models trained on data from potentially untrusted sources are vulnerable to poisoning. A small, maliciously crafted subset of the training inputs can cause the model to learn a \u201cbackdoor\u201d task (e.g., misclassify inputs with a certain feature) in addition to its main task. Recent research proposed a variety of backdoor attacks but they remain largely a hypothetical threat, whose efficacy heavily depends on the configuration and training hyperparameters of the target model. At the same time, state-of-the-art defenses require massive changes to the existing ML pipelines and may protect only against some attacks. ML engineers who are not security experts but want to make their models more secure thus face a difficult challenge. They must evaluate dozens of competing defenses proposed in the research literature, choose one of them, completely re-engineer their pipeline as required by the chosen defense, and then repeat the process if the defense disrupts normal model training (while providing theoretical protection against an unknown subset of hypothetical threats). In"
                    },
                    {
                        "title": "(Ab)using Images and Sounds for Indirect Instruction Injection in Multi-Modal LLMs",
                        "abstract": "We demonstrate how images and sounds can be used for indirect prompt and instruction injection in multi-modal LLMs. An attacker generates an adversarial perturbation corresponding to the prompt and blends it into an image or audio recording. When the user asks the (unmodified, benign) model about the perturbed image or audio, the perturbation steers the model to output the attacker-chosen text and/or make the subsequent dialog follow the attacker's instruction. We illustrate this attack with several proof-of-concept examples targeting LLaVa and PandaGPT."
                    },
                    {
                        "title": "Mithridates: Auditing and Boosting Backdoor Resistance of Machine Learning Pipelines",
                        "abstract": "Machine learning (ML) models trained on data from potentially untrusted sources are vulnerable to poisoning. A small, maliciously crafted subset of the training inputs can cause the model to learn a\"backdoor\"task (e.g., misclassify inputs with a certain feature) in addition to its main task. Recent research proposed many hypothetical backdoor attacks whose efficacy heavily depends on the configuration and training hyperparameters of the target model. Given the variety of potential backdoor attacks, ML engineers who are not security experts have no way to measure how vulnerable their current training pipelines are, nor do they have a practical way to compare training configurations so as to pick the more resistant ones. Deploying a defense requires evaluating and choosing from among dozens of research papers and re-engineering the training pipeline. In this paper, we aim to provide ML engineers with pragmatic tools to audit the backdoor resistance of their training pipelines and to compare different training configurations, to help choose one that best balances accuracy and security. First, we propose a universal, attack-agnostic resistance metric based on the minimum number of training inputs that must be compromised before the model learns any backdoor. Second, we design, implement, and evaluate Mithridates a multi-stage approach that integrates backdoor resistance into the training-configuration search. ML developers already rely on hyperparameter search to find configurations that maximize the model's accuracy. Mithridates extends this standard tool to balance accuracy and resistance without disruptive changes to the training pipeline. We show that hyperparameters found by Mithridates increase resistance to multiple types of backdoor attacks by 3-5x with only a slight impact on accuracy. We also discuss extensions to AutoML and federated learning."
                    },
                    {
                        "title": "Hyperparameter Search Is All You Need For Training-Agnostic Backdoor Robustness",
                        "abstract": "Commoditization and broad adoption of machine learning (ML) technologies expose users of these technologies to new security risks. Many models today are based on neural networks. Training and deploying these models for real-world applications involves complex hardware and software pipelines applied to training data from many sources. Models trained on untrusted data are vulnerable to poisoning attacks that introduce unwanted \u201cbackdoor\u201d functionality. Compromising a small fraction of the training data requires few resources from the attacker, but defending against these attacks is a challenge for ML developers. There have been dozens of defenses proposed in the research literature, most of them are expensive to integrate or incompatible with the existing model-training and deployment pipelines. The efficacy of each defense depends on the specifics of the dataset, ML task, model architecture, and training hyperparameters. Integrating and maintaining these defenses requires disruptive changes to the defender\u2019s environment. It is unrealistic to expect engineers in charge of model development and deployment to keep track of the research literature, evaluate every new defense, and repeatedly re-build their pipelines to integrate these defenses. In this paper, we take a pragmatic, developer-centric view and show how practitioners can answer two actionable questions: (1) how robust is my model to backdoor poisoning attacks?, and (2) how can I make it more robust without changing the training pipeline? We focus on the size of the compromised subset of the training data as a universal metric that applies to any poisoning attack and can be controlled by the defender. We use a primitive sub-task, which requires very few poisoned training inputs, to estimate this metric, thus providing a baseline on the model\u2019s robustness to poisoning. Next, we show how to leverage hyperparameter search\u2014a tool that ML developers already extensively use\u2014to balance the model\u2019s accuracy and robustness to poisoning, without any disruptive changes to the training and deployment pipeline. We demonstrate how to use our metric to estimate the robustness of models to backdoor attacks. We then design, implement, and evaluate a multi-stage hyperparameter search method we call Mithridates 1 that strengthens robustness by 3-5x with only a slight impact on the model\u2019s accuracy. We show that the hyperparameters found by our method increase robustness against multiple types of backdoor attacks. Finally, we discuss how to extend our method to AutoML and federated learning settings."
                    },
                    {
                        "title": "Text Embeddings Reveal (Almost) As Much As Text",
                        "abstract": "How much private information do text embeddings reveal about the original text? We investigate the problem of embedding \\textit{inversion}, reconstructing the full text represented in dense text embeddings. We frame the problem as controlled generation: generating text that, when reembedded, is close to a fixed point in latent space. We find that although a na\\\"ive model conditioned on the embedding performs poorly, a multi-step method that iteratively corrects and re-embeds text is able to recover $92\\%$ of $32\\text{-token}$ text inputs exactly. We train our model to decode text embeddings from two state-of-the-art embedding models, and also show that our model can recover important personal information (full names) from a dataset of clinical notes. Our code is available on Github: \\href{https://github.com/jxmorris12/vec2text}{github.com/jxmorris12/vec2text}."
                    },
                    {
                        "title": "Adversarial Illusions in Multi-Modal Embeddings",
                        "abstract": "Multi-modal embeddings encode texts, images, thermal images, sounds, and videos into a single embedding space, aligning representations across different modalities (e.g., associate an image of a dog with a barking sound). In this paper, we show that multi-modal embeddings can be vulnerable to an attack we call\"adversarial illusions.\"Given an image or a sound, an adversary can perturb it to make its embedding close to an arbitrary, adversary-chosen input in another modality. These attacks are cross-modal and targeted: the adversary can align any image or sound with any target of his choice. Adversarial illusions exploit proximity in the embedding space and are thus agnostic to downstream tasks and modalities, enabling a wholesale compromise of current and future tasks, as well as modalities not available to the adversary. Using ImageBind and AudioCLIP embeddings, we demonstrate how adversarially aligned inputs, generated without knowledge of specific downstream tasks, mislead image generation, text generation, zero-shot classification, and audio retrieval. We investigate transferability of illusions across different embeddings and develop a black-box version of our method that we use to demonstrate the first adversarial alignment attack on Amazon's commercial, proprietary Titan embedding. Finally, we analyze countermeasures and evasion attacks."
                    },
                    {
                        "title": "Telepath: A Minecraft-based Covert Communication System",
                        "abstract": "Covert, censorship-resistant communication in the presence of nation-state adversaries requires unobservable channels whose operation is difficult to detect via network-traffic analysis. Traffic substitution, i.e., replacing data transmitted by a \"cover\" application with covert content, takes advantage of already-existing encrypted channels to produce traffic that is statistically indistinguishable from the traffic of the cover application and thus difficult to censor.Online games are a promising platform for building circumvention channels due to their popularity in many censored regions. We show, however, that previously proposed traffic substitution methods cannot be directly applied to games. Their traces, even if statistically similar to game traces, may violate game-specific invariants and are thus easy to detect because they could not have been generated by an actual gameplay.We explain how to identify non-disruptive content whose substitution does not result in client-server inconsistencies and use these ideas to design and implement Telepath, a covert communication system that uses Minecraft as the platform. Telepath takes advantage of (1) Minecraft\u2019s encrypted client-server channel, (2) decentralized architecture that enables individual users to run their own servers, and (3) popularity of \"mods\" that add functionality to Minecraft clients and servers. Telepath runs a Minecraft game but substitutes non-disruptive in-game messages with covert content, without changing the game\u2019s interaction with the network manager.We measure performance of Telepath for Web browsing and audio streaming, and show that network traffic generated by Telepath resists statistical traffic analysis that aims to distinguish it from popular Minecraft bots."
                    },
                    {
                        "title": "Asymmetric Cryptography",
                        "abstract": "and many others for sample slides and materials ..."
                    },
                    {
                        "title": "Data Isotopes for Data Provenance in DNNs",
                        "abstract": "Today, creators of data-hungry deep neural networks (DNNs) scour the Internet for training fodder, leaving users with little control over or knowledge of when their data, and in particular their images, are used to train models. To empower users to counteract unwanted use of their images, we design, implement and evaluate a practical system that enables users to detect if their data was used to train a DNN model for image classification. We show how users can create special images we call isotopes, which introduce ``spurious features'' into DNNs during training. With only query access to a model and no knowledge of the model-training process, nor control of the data labels, a user can apply statistical hypothesis testing to detect if the model learned these spurious features by training on the user's images. Isotopes can be viewed as an application of a particular type of data poisoning. In contrast to backdoors and other poisoning attacks, our purpose is not to cause misclassification but rather to create tell-tale changes in confidence scores output by the model that reveal the presence of isotopes in the training data. Isotopes thus turn DNNs' vulnerability to memorization and spurious correlations into a tool for data provenance. Our results confirm efficacy in multiple image classification settings, detecting and distinguishing between hundreds of isotopes with high accuracy. We further show that our system works on public ML-as-a-service platforms and larger models such as ImageNet, can use physical objects in images instead of digital marks, and remains robust against several adaptive countermeasures."
                    },
                    {
                        "title": "Spinning Language Models: Risks of Propaganda-As-A-Service and Countermeasures",
                        "abstract": "We investigate a new threat to neural sequence-to-sequence (seq2seq) models: training-time attacks that cause models to \u201cspin\u201d their outputs so as to support an adversary-chosen sentiment or point of view\u2014but only when the input contains adversary-chosen trigger words. For example, a spinned 1 summarization model outputs positive summaries of any text that mentions the name of some individual or organization.Model spinning introduces a \u201cmeta-backdoor\u201d into a model. Whereas conventional backdoors cause models to produce incorrect outputs on inputs with the trigger, outputs of spinned models preserve context and maintain standard accuracy metrics, yet also satisfy a meta-task chosen by the adversary.Model spinning enables propaganda-as-a-service, where propaganda is defined as biased speech. An adversary can create customized language models that produce desired spins for chosen triggers, then deploy these models to generate disinformation (a platform attack), or else inject them into ML training pipelines (a supply-chain attack), transferring malicious functionality to downstream models trained by victims.To demonstrate the feasibility of model spinning, we develop a new backdooring technique. It stacks an adversarial meta-task (e.g., sentiment analysis) onto a seq2seq model, backpropagates the desired meta-task output (e.g., positive sentiment) to points in the word-embedding space we call \u201cpseudo-words,\u201d and uses pseudo-words to shift the entire output distribution of the seq2seq model. We evaluate this attack on language generation, summarization, and translation models with different triggers and meta-tasks such as sentiment, toxicity, and entailment. Spinned models largely maintain their accuracy metrics (ROUGE and BLEU) while shifting their outputs to satisfy the adversary\u2019s meta-task. We also show that, in the case of a supply-chain attack, the spin functionality transfers to downstream models.Finally, we propose a black-box, meta-task-independent defense that, given a list of candidate triggers, can detect models that selectively apply spin to inputs with any of these triggers.1We use \u201cspinned\u201d rather than \u201cspun\u201d to match how the word is used in public relations."
                    },
                    {
                        "title": "Spinning Language Models for Propaganda-As-A-Service",
                        "abstract": "\u2014We investigate a new threat to neural sequence-to-sequence (seq2seq) models: training-time attacks that cause models to \u201cspin\u201d their outputs so as to support an adversary-chosen sentiment or point of view, but only when the input contains adversary-chosen trigger words. For example, a spinned summarization model would output positive summaries of any text that mentions the name of some individual or organization. Model spinning enables propaganda-as-a-service. An adversary can create customized language models that produce desired spins for chosen triggers, then deploy them to generate disinformation (a platform attack), or else inject them into ML training pipelines (a supply-chain attack), transferring malicious functionality to downstream models trained by victims. In technical terms, model spinning introduces a \u201cmeta-backdoor\u201d into a model. Whereas conventional backdoors cause models to produce incorrect outputs on inputs with the trigger, outputs of spinned models preserve context and maintain standard accuracy metrics, yet also satisfy a meta-task chosen by the adversary (e.g., positive sentiment). To demonstrate feasibility of model spinning, we develop a new backdooring technique. It stacks the adversarial meta-task (e.g., sentiment analysis) onto a seq2seq model, backpropagates the desired meta-task output (e.g., positive sentiment) to points in the word-embedding space we call \u201cpseudo-words,\u201d and uses pseudo-words to shift the entire output distribution of the seq2seq model. We evaluate this attack on language generation, summarization, and translation models with different triggers and meta-tasks such as sentiment, toxicity, and entailment. Spinned models largely maintain their accuracy metrics (ROUGE and BLEU) while shifting their outputs to satisfy the adversary\u2019s meta-task. We also show that, in the case of a supply-chain attack, the spin functionality transfers to downstream models. Finally, we propose a black-box, meta-task-independent defense"
                    },
                    {
                        "title": "BigFoot: Exploiting and Mitigating Leakage in Encrypted Write-Ahead Logs",
                        "abstract": "Modern databases and data-warehousing systems separate query processing and durable storage. Storage systems have idiosyncratic bugs and security vulnerabilities, thus attacks that compromise only storage are a realistic threat. In this paper, we show that encryption alone is not sufficient to protect databases from compromised storage. Using MongoDB WiredTiger as a concrete example, we demonstrate that sizes of encrypted writes to a durable write-ahead log can reveal sensitive information about the inputs and activities of MongoDB applications. We then design, implement, and evaluate BigFoot, a WAL modification that mitigates size leakage."
                    },
                    {
                        "title": "Spinning Sequence-to-Sequence Models with Meta-Backdoors",
                        "abstract": "We investigate a new threat to neural sequence-to-sequence (seq2seq) models: training-time attacks that cause models to\"spin\"their output and support a certain sentiment when the input contains adversary-chosen trigger words. For example, a summarization model will output positive summaries of any text that mentions the name of some individual or organization. We introduce the concept of a\"meta-backdoor\"to explain model-spinning attacks. These attacks produce models whose output is valid and preserves context, yet also satisfies a meta-task chosen by the adversary (e.g., positive sentiment). Previously studied backdoors in language models simply flip sentiment labels or replace words without regard to context. Their outputs are incorrect on inputs with the trigger. Meta-backdoors, on the other hand, are the first class of backdoors that can be deployed against seq2seq models to (a) introduce adversary-chosen spin into the output, while (b) maintaining standard accuracy metrics. To demonstrate feasibility of model spinning, we develop a new backdooring technique. It stacks the adversarial meta-task (e.g., sentiment analysis) onto a seq2seq model, backpropagates the desired meta-task output (e.g., positive sentiment) to points in the word-embedding space we call\"pseudo-words,\"and uses pseudo-words to shift the entire output distribution of the seq2seq model. Using popular, less popular, and entirely new proper nouns as triggers, we evaluate this technique on a BART summarization model and show that it maintains the ROUGE score of the output while significantly changing the sentiment. We explain why model spinning can be a dangerous technique in AI-powered disinformation and discuss how to mitigate these attacks."
                    },
                    {
                        "title": "De-Anonymizing Text by Fingerprinting Language Generation",
                        "abstract": "Components of machine learning systems are not (yet) perceived as security hotspots. Secure coding practices, such as ensuring that no execution paths depend on confidential inputs, have not yet been adopted by ML developers. We initiate the study of code security of ML systems by investigating how nucleus sampling---a popular approach for generating text, used for applications such as auto-completion---unwittingly leaks texts typed by users. Our main result is that the series of nucleus sizes for many natural English word sequences is a unique fingerprint. We then show how an attacker can infer typed text by measuring these fingerprints via a suitable side channel (e.g., cache access times), explain how this attack could help de-anonymize anonymous texts, and discuss defenses."
                    },
                    {
                        "title": "You Autocomplete Me: Poisoning Vulnerabilities in Neural Code Completion",
                        "abstract": "Code autocompletion is an integral feature of modern code editors and IDEs. The latest generation of autocompleters uses neural language models, trained on public open-source code repositories, to suggest likely (not just statically feasible) completions given the current context.  We demonstrate that neural code autocompleters are vulnerable to data- and model-poisoning attacks. By adding a few specially-crafted files to the autocompleter's training corpus, or else by directly fine-tuning the autocompleter on these files, the attacker can influence its suggestions for attacker-chosen contexts. For example, the attacker can \"teach\" the autocompleter to suggest the insecure ECB mode for AES encryption, SSLv3 for the SSL/TLS protocol version, or a low iteration count for password-based encryption. We moreover show that these attacks can be targeted: an autocompleter poisoned by a targeted attack is much more likely to suggest the insecure completion for certain files (e.g., those from a specific repo).  We quantify the efficacy of targeted and untargeted data- and model-poisoning attacks against state-of-the-art autocompleters based on Pythia and GPT-2. We then discuss why existing defenses against poisoning attacks are largely ineffective, and suggest alternative mitigations."
                    },
                    {
                        "title": "Salvaging Federated Learning by Local Adaptation",
                        "abstract": "Federated learning (FL) is a heavily promoted approach for training ML models on sensitive data, e.g., text typed by users on their smartphones. FL is expressly designed for training on data that are unbalanced and non-iid across the participants. To ensure privacy and integrity of the federated model, latest FL approaches use differential privacy or robust aggregation to limit the influence of \"outlier\" participants.  First, we show that on standard tasks such as next-word prediction, many participants gain no benefit from FL because the federated model is less accurate on their data than the models they can train locally on their own. Second, we show that differential privacy and robust aggregation make this problem worse by further destroying the accuracy of the federated model for many participants.  Then, we evaluate three techniques for local adaptation of federated models: fine-tuning, multi-task learning, and knowledge distillation. We analyze where each technique is applicable and demonstrate that all participants benefit from local adaptation. Participants whose local models are poor obtain big accuracy improvements over conventional FL. Participants whose local models are better than the federated model and who have no incentive to participate in FL today improve less, but sufficiently to make the adapted federated model better than their local models."
                    },
                    {
                        "title": "Humpty Dumpty: Controlling Word Meanings via Corpus Poisoning",
                        "abstract": "Word embeddings, i.e., low-dimensional vector representations such as GloVe and SGNS, encode word \"meaning\" in the sense that distances between words\u2019 vectors correspond to their semantic proximity. This enables transfer learning of semantics for a variety of natural language processing tasks.Word embeddings are typically trained on large public corpora such as Wikipedia or Twitter. We demonstrate that an attacker who can modify the corpus on which the embedding is trained can control the \"meaning\" of new and existing words by changing their locations in the embedding space. We develop an explicit expression over corpus features that serves as a proxy for distance between words and establish a causative relationship between its values and embedding distances. We then show how to use this relationship for two adversarial objectives: (1) make a word a top-ranked neighbor of another word, and (2) move a word from one semantic cluster to another.An attack on the embedding can affect diverse downstream tasks, demonstrating for the first time the power of data poisoning in transfer learning scenarios. We use this attack to manipulate query expansion in information retrieval systems such as resume search, make certain names more or less visible to named entity recognition models, and cause new words to be translated to a particular target word regardless of the language. Finally, we show how the attacker can generate linguistically likely corpus modifications, thus fooling defenses that attempt to filter implausible sentences from the corpus using a language model."
                    }
                ]
            },
            "b8979f67-144f-459d-99ef-4fbbae93fe4a": {
                "pk": "b8979f67-144f-459d-99ef-4fbbae93fe4a",
                "name": "Alexander M. Rush",
                "collaborators": [
                    "John X. Morris",
                    "Chandan Singh",
                    "Jianfeng Gao",
                    "J. Aneja",
                    "Yuntian Deng",
                    "Volodymyr Kuleshov",
                    "Vitaly Shmatikov"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Document Embeddings",
                    "Machine Learning",
                    "Explainable AI"
                ],
                "publications": [
                    {
                        "title": "Contextual Document Embeddings",
                        "abstract": "Dense document embeddings are central to neural retrieval. The dominant paradigm is to train and construct embeddings by running encoders directly on individual documents. In this work, we argue that these embeddings, while effective, are implicitly out-of-context for targeted use cases of retrieval, and that a contextualized document embedding should take into account both the document and neighboring documents in context - analogous to contextualized word embeddings. We propose two complementary methods for contextualized document embeddings: first, an alternative contrastive learning objective that explicitly incorporates the document neighbors into the intra-batch contextual loss; second, a new contextual architecture that explicitly encodes neighbor document information into the encoded representation. Results show that both methods achieve better performance than biencoders in several settings, with differences especially pronounced out-of-domain. We achieve state-of-the-art results on the MTEB benchmark with no hard negative mining, score distillation, dataset-specific instructions, intra-GPU example-sharing, or extremely large batch sizes. Our method can be applied to improve performance on any contrastive learning dataset and any biencoder."
                    },
                    {
                        "title": "Explaining Data Patterns in Natural Language with Language Models",
                        "abstract": "Large language models (LLMs) have displayed an impressive ability to harness natural language to perform complex tasks. We explore whether we can leverage this ability to find and explain patterns in data. Specifically, given a pre-trained LLM and data examples, we apply interpretable autoprompting (iPrompt) to generate a natural language string explaining the data. iPrompt iteratively generates explanations with an LLM and reranks them based on their performance when used as a prompt. Experiments on a wide range of datasets, from synthetic mathematics to natural language understanding, show that iPrompt can yield meaningful insights by accurately finding dataset explanations that are human-interpretable. Moreover, iPrompt is reasonably efficient, as it does not require access to model gradients and works with relatively small models (e.g. ~6 billion parameters rather than >=100 billion). Finally, experiments with scientific datasets show the potential for iPrompt to aid in scientific discovery."
                    },
                    {
                        "title": "Tree Prompting: Efficient Task Adaptation without Fine-Tuning",
                        "abstract": "Prompting language models (LMs) is the main interface for applying them to new tasks. However, for smaller LMs, prompting provides low accuracy compared to gradient-based finetuning. Tree Prompting is an approach to prompting which builds a decision tree of prompts, linking multiple LM calls together to solve a task. At inference time, each call to the LM is determined by efficiently routing the outcome of the previous call using the tree. Experiments on classification datasets show that Tree Prompting improves accuracy over competing methods and is competitive with fine-tuning. We also show that variants of Tree Prompting allow inspection of a model's decision-making process."
                    },
                    {
                        "title": "Text Embeddings Reveal (Almost) As Much As Text",
                        "abstract": "How much private information do text embeddings reveal about the original text? We investigate the problem of embedding \\textit{inversion}, reconstructing the full text represented in dense text embeddings. We frame the problem as controlled generation: generating text that, when reembedded, is close to a fixed point in latent space. We find that although a na\\\"ive model conditioned on the embedding performs poorly, a multi-step method that iteratively corrects and re-embeds text is able to recover $92\\%$ of $32\\text{-token}$ text inputs exactly. We train our model to decode text embeddings from two state-of-the-art embedding models, and also show that our model can recover important personal information (full names) from a dataset of clinical notes. Our code is available on Github: \\href{https://github.com/jxmorris12/vec2text}{github.com/jxmorris12/vec2text}."
                    }
                ]
            }
        }
    },
    "2408.15205": {
        "paper_data": {
            "title": "Leveraging Hallucinations to Reduce Manual Prompt Dependency in Promptable Segmentation",
            "url": "http://arxiv.org/abs/2408.15205v1",
            "arxiv_id": "2408.15205",
            "authors": [
                "Jian Hu",
                "Jiayi Lin",
                "Junchi Yan",
                "Shaogang Gong"
            ],
            "abstract": "Promptable segmentation typically requires instance-specific manual prompts to guide the segmentation of each desired object. To minimize such a need, task-generic promptable segmentation has been introduced, which employs a single task-generic prompt to segment various images of different objects in the same task. Current methods use Multimodal Large Language Models (MLLMs) to reason detailed instance-specific prompts from a task-generic prompt for improving segmentation accuracy. The effectiveness of this segmentation heavily depends on the precision of these derived prompts. However, MLLMs often suffer hallucinations during reasoning, resulting in inaccurate prompting. While existing methods focus on eliminating hallucinations to improve a model, we argue that MLLM hallucinations can reveal valuable contextual insights when leveraged correctly, as they represent pre-trained large-scale knowledge beyond individual images. In this paper, we utilize hallucinations to mine task-related information from images and verify its accuracy for enhancing precision of the generated prompts. Specifically, we introduce an iterative Prompt-Mask Cycle generation framework (ProMaC) with a prompt generator and a mask generator.The prompt generator uses a multi-scale chain of thought prompting, initially exploring hallucinations for extracting extended contextual knowledge on a test image.These hallucinations are then reduced to formulate precise instance-specific prompts, directing the mask generator to produce masks that are consistent with task semantics by mask semantic alignment. The generated masks iteratively induce the prompt generator to focus more on task-relevant image areas and reduce irrelevant hallucinations, resulting jointly in better prompts and masks. Experiments on 5 benchmarks demonstrate the effectiveness of ProMaC. Code given in https://lwpyh.github.io/ProMaC/.",
            "introduction": "   1 Introduction  Current promptable segmentation methods rely on instance-specific manual prompts to guide segmentation, greatly limiting its large-scale application. Recently, a manual-free task-generic promptable segmentation approach was introduced [19]: only a single task-generic prompt is needed for all samples under the same task, e.g., \u201ccamouflaged animal\u201d is a task-generic prompt for all images in a camouflaged object detection task. The model segments task-relevant objects in various images based on this generic prompt, significantly reducing the annotation workload.           (a) Hallucination by co-occurrence prior.   (b) Using hallucinations can benefit accuracte prompt generation.     Figure 1:  (a) During MLLM pretraining, leopards often co-occur with grass. If the lion is masked, the model incorrectly identifies it as a leopard based on the grass. (b) Directly inputting the image into MLLM causes the hidden caterpillar being incorrectly predicted as a leaf. Splitting the image results in interested objects being incomplete or absent, prompting MLLM to induce hallucinations and utilize prior knowledge to predict potential task-related objects within the image. Our visual contrastive reasoning then eliminates the hallucinations and validates the gathered predictions, aiding in the accurate identification of the caterpillar.    A task-generic prompt is both coarse and potentially ambiguous, can result in poor segmentation when directly applied. To address this problem, existing methods [19, 36] utilize the prior knowledge embedded in Multimodal Large Language Models (MLLMs) to infer more detailed, instance-specific prompts, such as bounding boxes or keywords, to guide the segmentation. However, these MLLMs often generate hallucinations due to object co-occurrence priors [66, 63], mistakenly predicting non-existent elements based on the environment as instance-specific prompts (Fig.1(a)). This can mislead segmentation and degrade model performance. While it is common to consider MLLM\u2019s hallucinations as detrimental and should be eradicated [50], this phenomenon actually demonstrates a MLLM\u2019s significant capacity for contextual inference based on prior training. We want to explore MLLM hallucinations as a valuable untapped knowledge resource for scene understanding, critical in complex segmentation scenarios. In practice, when task-related objects are not prominently visible, hallucinations can fill in missing information with plausible predictions based on learned patterns of association. Moreover, they can also extend beyond these familiar patterns, exploring and identifying new relationships within the data that were not explicitly taught during training. This dual ability to replicate and innovate makes hallucinations a valuable asset for enhancing model performance in complex or new situations. This predictive reasoning capacity not only fills perceptual gaps but also enriches the model\u2019s understanding, as hallucinations utilize prior knowledge to replicate and discover new patterns, enhancing insight into the target domain (see Fig.1(b)). Despite the potential benefits, using hallucinations to extract useful information from images to aid task remains unexplored.   In this work, instead of direct eliminating hallucinations, we utilize them as prior knowledge to mine extended task-related information from a given test image, performing scene understanding on the image before segmentation, then systematically reduce irrelevant hallucinations iteratively by visual masking verification, optimizing jointly instance-specific prompts and masks. To this end, we introduce an iterative, training-free Prompt-Mask Cycle Generation method (ProMaC) that refines segmentation through cyclic interactions between a prompt and mask generator (see Fig. 2). The prompt generator uses a multi-scale chain-of-thought prompting mechanism, which utilizes hallucinations to hypothesize and visual masking to verify, thereby creating more accurate instance-specific prompts. We trigger the hallucinatory tendencies of MLLMs,",
            "references": [
                {
                    "title": "Contrastive Region Guidance: Improving Grounding in Vision-Language Models without Training",
                    "abstract": "Highlighting particularly relevant regions of an image can improve the performance of vision-language models (VLMs) on various vision-language (VL) tasks by guiding the model to attend more closely to these regions of interest. For example, VLMs can be given a\"visual prompt\", where visual markers such as bounding boxes delineate key image regions. However, current VLMs that can incorporate visual guidance are either proprietary and expensive or require costly training on curated data that includes visual prompts. We introduce Contrastive Region Guidance (CRG), a training-free guidance method that enables open-source VLMs to respond to visual prompts. CRG contrasts model outputs produced with and without visual prompts, factoring out biases revealed by the model when answering without the information required to produce a correct answer (i.e., the model's prior). CRG achieves substantial improvements in a wide variety of VL tasks: When region annotations are provided, CRG increases absolute accuracy by up to 11.1% on ViP-Bench, a collection of six diverse region-based tasks such as recognition, math, and object relationship reasoning. We also show CRG's applicability to spatial reasoning, with 10% improvement on What'sUp, as well as to compositional generalization -- improving accuracy by 11.5% and 7.5% on two challenging splits from SugarCrepe -- and to image-text alignment for generated images, where we improve by up to 8.4 AUROC and 6.8 F1 points on SeeTRUE. When reference regions are absent, CRG allows us to re-rank proposed regions in referring expression comprehension and phrase grounding benchmarks like RefCOCO/+/g and Flickr30K Entities, with an average gain of 3.2% in accuracy. Our analysis explores alternative masking strategies for CRG, quantifies CRG's probability shift, and evaluates the role of region guidance strength, empirically validating CRG's design choices."
                },
                {
                    "title": "Relax Image-Specific Prompt Requirement in SAM: A Single Generic Prompt for Segmenting Camouflaged Objects",
                    "abstract": "Camouflaged object detection (COD) approaches heavily rely on pixel-level annotated datasets. Weakly-supervised COD (WSCOD) approaches use sparse annotations like scribbles or points to reduce annotation efforts, but this can lead to decreased accuracy. The Segment Anything Model (SAM) shows remarkable segmentation ability with sparse prompts like points. However, manual prompt is not always feasible, as it may not be accessible in real-world application. Additionally, it only provides localization information instead of semantic one, which can intrinsically cause ambiguity in interpreting targets. In this work, we aim to eliminate the need for manual prompt. The key idea is to employ Cross-modal Chains of Thought Prompting (CCTP) to reason visual prompts using the semantic information given by a generic text prompt. To that end, we introduce a test-time instance-wise adaptation mechanism called Generalizable SAM (GenSAM) to automatically generate and optimize visual prompts from the generic task prompt for WSCOD. In particular, CCTP maps a single generic text prompt onto image-specific consensus foreground and background heatmaps using vision-language models, acquiring reliable visual prompts. Moreover, to test-time adapt the visual prompts, we further propose Progressive Mask Generation (PMG) to iteratively reweight the input image, guiding the model to focus on the targeted region in a coarse-to-fine manner. Crucially, all network parameters are fixed, avoiding the need for additional training. Experiments on three benchmarks demonstrate that GenSAM outperforms point supervision approaches and achieves comparable results to scribble supervision ones, solely relying on general task descriptions. Our codes is in https://github.com/jyLin8100/GenSAM."
                },
                {
                    "title": "Sequential Modeling Enables Scalable Learning for Large Vision Models",
                    "abstract": "We introduce a novel sequential modeling approach which enables learning a Large Vision Model (LVM) without making use of any linguistic data. To do this, we define a common format, \u201cvisual sentences \u201d, in which we can represent raw images and videos as well as annotated data sources such as semantic segmentations and depth reconstructions with-out needing any meta-knowledge beyond the pixels. Once this wide variety of visual data (comprising 420 billion to-kens) is represented as sequences, the model can be trained to minimize a cross-entropy loss for next token prediction. By training across various scales of model architecture and data diversity, we provide empirical evidence that our models scale effectively. Many different vision tasks can be solved by designing suitable visual prompts at test time."
                },
                {
                    "title": "ViP-LLaVA: Making Large Multimodal Models Understand Arbitrary Visual Prompts",
                    "abstract": "While existing large vision-language multimodal models focus on whole image understanding, there is a prominent gap in achieving region-specific comprehension. Current approaches that use textual coordinates or spatial encodings often fail to provide a user-friendly interface for visual prompting. To address this challenge, we introduce a novel multimodal model capable of decoding arbitrary (free-form) visual prompts. This allows users to intuitively mark images and interact with the model using natural cues like a \u201cred bounding box\u201d or \u201cpointed arrow\u201d. Our simple design directly overlays visual markers onto the RGB image, eliminating the need for complex region encodings, yet achieves state-of-the-art performance on region-understanding tasks like Visual7W, PointQA, and Visual Commonsense Reasoning benchmark. Furthermore, we present ViP-Bench, a comprehensive benchmark to assess the capability of models in understanding visual prompts across multiple dimensions, enabling future research in this domain. Code, data, and model are publicly available."
                },
                {
                    "title": "Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding",
                    "abstract": "Large Vision-Language Models (LVLMs) have advanced considerably, intertwining visual recognition and language understanding to generate content that is not only coherent but also contextually attuned. Despite their success, LVLMs still suffer from the issue of object hallucinations, where models generate plausible yet incorrect outputs that include objects that do not exist in the images. To mitigate this issue, we introduce Visual Contrastive Decoding (VCD), a simple and training-free method that contrasts output distributions derived from original and distorted visual inputs. The proposed VCD effectively reduces the over-reliance on statistical bias and unimodal priors, two essential causes of object hallucinations. This adjustment ensures the generated content is closely grounded to visual inputs, resulting in contextually accurate outputs. Our experiments show that VCD, without either additional training or the usage of external tools, significantly mitigates the object hallucination issue across different LVLM families. Beyond mitigating object hallucinations, VCD also excels in general LVLM benchmarks, highlighting its wide-ranging applicability."
                },
                {
                    "title": "What's \"up\" with vision-language models? Investigating their struggle with spatial reasoning",
                    "abstract": "Recent vision-language (VL) models are powerful, but can they reliably distinguish\"right\"from\"left\"? We curate three new corpora to quantify model comprehension of such basic spatial relations. These tests isolate spatial reasoning more precisely than existing datasets like VQAv2, e.g., our What'sUp benchmark contains sets of photographs varying only the spatial relations of objects, keeping their identity fixed (see Figure 1: models must comprehend not only the usual case of a dog under a table, but also, the same dog on top of the same table). We evaluate 18 VL models, finding that all perform poorly, e.g., BLIP finetuned on VQAv2, which nears human parity on VQAv2, achieves 56% accuracy on our benchmarks vs. humans at 99%. We conclude by studying causes of this surprising behavior, finding: 1) that popular vision-language pretraining corpora like LAION-2B contain little reliable data for learning spatial relationships; and 2) that basic modeling interventions like up-weighting preposition-containing instances or fine-tuning on our corpora are not sufficient to address the challenges our benchmarks pose. We are hopeful that these corpora will facilitate further research, and we release our data and code at https://github.com/amitakamath/whatsup_vlms."
                },
                {
                    "title": "Set-of-Mark Prompting Unleashes Extraordinary Visual Grounding in GPT-4V",
                    "abstract": "We present Set-of-Mark (SoM), a new visual prompting method, to unleash the visual grounding abilities of large multimodal models (LMMs), such as GPT-4V. As illustrated in Fig. 1 (right), we employ off-the-shelf interactive segmentation models, such as SEEM/SAM, to partition an image into regions at different levels of granularity, and overlay these regions with a set of marks e.g., alphanumerics, masks, boxes. Using the marked image as input, GPT-4V can answer the questions that require visual grounding. We perform a comprehensive empirical study to validate the effectiveness of SoM on a wide range of fine-grained vision and multimodal tasks. For example, our experiments show that GPT-4V with SoM in zero-shot setting outperforms the state-of-the-art fully-finetuned referring expression comprehension and segmentation model on RefCOCOg. Code for SoM prompting is made public at: https://github.com/microsoft/SoM."
                },
                {
                    "title": "Improved Baselines with Visual Instruction Tuning",
                    "abstract": "Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this paper, we present the first systematic study to investigate the design choices of LMMs in a controlled setting under the LLaVA framework. We show that the fully-connected vision-language connector in LLaVA is surprisingly power-ful and data-efficient. With simple modifications to LLa VA, namely, using CLIP- ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~ 1 day on a single 8-AI00 node. Furthermore, we present some early exploration of open problems in LMMs, including scaling to higher resolution inputs, compositional capabilities, and model hallucination, etc. We hope this makes state-of-the-art LMM research more accessible. Code and model will be publicly available."
                },
                {
                    "title": "Analyzing and Mitigating Object Hallucination in Large Vision-Language Models",
                    "abstract": "Large vision-language models (LVLMs) have shown remarkable abilities in understanding visual information with human languages. However, LVLMs still suffer from object hallucination, which is the problem of generating descriptions that include objects that do not actually exist in the images. This can negatively impact many vision-language tasks, such as visual summarization and reasoning. To address this issue, we propose a simple yet powerful algorithm, LVLM Hallucination Revisor (LURE), to post-hoc rectify object hallucination in LVLMs by reconstructing less hallucinatory descriptions. LURE is grounded in a rigorous statistical analysis of the key factors underlying object hallucination, including co-occurrence (the frequent appearance of certain objects alongside others in images), uncertainty (objects with higher uncertainty during LVLM decoding), and object position (hallucination often appears in the later part of the generated text). LURE can also be seamlessly integrated with any LVLMs. We evaluate LURE on six open-source LVLMs, achieving a 23% improvement in general object hallucination evaluation metrics over the previous best approach. In both GPT and human evaluations, LURE consistently ranks at the top. Our data and code are available at https://github.com/YiyangZhou/LURE."
                },
                {
                    "title": "Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Models",
                    "abstract": "While large language models (LLMs) have demonstrated remarkable capabilities across a range of downstream tasks, a significant concern revolves around their propensity to exhibit hallucinations: LLMs occasionally generate content that diverges from the user input, contradicts previously generated context, or misaligns with established world knowledge. This phenomenon poses a substantial challenge to the reliability of LLMs in real-world scenarios. In this paper, we survey recent efforts on the detection, explanation, and mitigation of hallucination, with an emphasis on the unique challenges posed by LLMs. We present taxonomies of the LLM hallucination phenomena and evaluation benchmarks, analyze existing approaches aiming at mitigating LLM hallucination, and discuss potential directions for future research."
                },
                {
                    "title": "Uncertainty-based Heterogeneous Privileged Knowledge Distillation for Recommendation System",
                    "abstract": "In industrial recommendation systems, both data sizes and computational resources vary across different scenarios. For scenarios with limited data, data sparsity can lead to a decrease in model performance. Heterogeneous knowledge distillation-based transfer learning can be used to transfer knowledge from models in data-rich domains. However, in recommendation systems, the target domain possesses specific privileged features that significantly contribute to the model. While existing knowledge distillation methods have not taken these features into consideration, leading to suboptimal transfer weights. To overcome this limitation, we propose a novel algorithm called Uncertainty-based Heterogeneous Privileged Knowledge Distillation (UHPKD). Our method aims to quantify the knowledge of both the source and target domains, which represents the uncertainty of the models. This approach allows us to derive transfer weights based on the knowledge gain, which captures the difference in knowledge between the source and target domains. Experiments conducted on both public and industrial datasets demonstrate the superiority of our UHPKD algorithm compared to other state-of-the-art methods."
                },
                {
                    "title": "Stay on topic with Classifier-Free Guidance",
                    "abstract": "Classifier-Free Guidance (CFG) has recently emerged in text-to-image generation as a lightweight technique to encourage prompt-adherence in generations. In this work, we demonstrate that CFG can be used broadly as an inference-time technique in pure language modeling. We show that CFG (1) improves the performance of Pythia, GPT-2 and LLaMA-family models across an array of tasks: Q\\&A, reasoning, code generation, and machine translation, achieving SOTA on LAMBADA with LLaMA-7B over PaLM-540B; (2) brings improvements equivalent to a model with twice the parameter-count; (3) can stack alongside other inference-time methods like Chain-of-Thought and Self-Consistency, yielding further improvements in difficult tasks; (4) can be used to increase the faithfulness and coherence of assistants in challenging form-driven and content-driven prompts: in a human evaluation we show a 75\\% preference for GPT4All using CFG over baseline."
                },
                {
                    "title": "Amodal Instance Segmentation via Prior-Guided Expansion",
                    "abstract": "Amodal instance segmentation aims to infer the amodal mask, including both the visible part and occluded part of each object instance. Predicting the occluded parts is challenging. Existing methods often produce incomplete amodal boxes and amodal masks, probably due to lacking visual evidences to expand the boxes and masks. To this end, we propose a prior-guided expansion framework, which builds on a two-stage segmentation model (i.e., Mask R-CNN) and performs box-level (resp., pixel-level) expansion for amodal box (resp., mask) prediction, by retrieving regression (resp., flow) transformations from a memory bank of expansion prior. We conduct extensive experiments on KINS, D2SA, and COCOA cls datasets, which show the effectiveness of our method."
                },
                {
                    "title": "Weakly-Supervised Concealed Object Segmentation with SAM-based Pseudo Labeling and Multi-scale Feature Grouping",
                    "abstract": "Weakly-Supervised Concealed Object Segmentation (WSCOS) aims to segment objects well blended with surrounding environments using sparsely-annotated data for model training. It remains a challenging task since (1) it is hard to distinguish concealed objects from the background due to the intrinsic similarity and (2) the sparsely-annotated training data only provide weak supervision for model learning. In this paper, we propose a new WSCOS method to address these two challenges. To tackle the intrinsic similarity challenge, we design a multi-scale feature grouping module that first groups features at different granularities and then aggregates these grouping results. By grouping similar features together, it encourages segmentation coherence, helping obtain complete segmentation results for both single and multiple-object images. For the weak supervision challenge, we utilize the recently-proposed vision foundation model, Segment Anything Model (SAM), and use the provided sparse annotations as prompts to generate segmentation masks, which are used to train the model. To alleviate the impact of low-quality segmentation masks, we further propose a series of strategies, including multi-augmentation result ensemble, entropy-based pixel-level weighting, and entropy-based image-level selection. These strategies help provide more reliable supervision to train the segmentation model. We verify the effectiveness of our method on various WSCOS tasks, and experiments demonstrate that our method achieves state-of-the-art performance on these tasks."
                },
                {
                    "title": "AV-SAM: Segment Anything Model Meets Audio-Visual Localization and Segmentation",
                    "abstract": "Segment Anything Model (SAM) has recently shown its powerful effectiveness in visual segmentation tasks. However, there is less exploration concerning how SAM works on audio-visual tasks, such as visual sound localization and segmentation. In this work, we propose a simple yet effective audio-visual localization and segmentation framework based on the Segment Anything Model, namely AV-SAM, that can generate sounding object masks corresponding to the audio. Specifically, our AV-SAM simply leverages pixel-wise audio-visual fusion across audio features and visual features from the pre-trained image encoder in SAM to aggregate cross-modal representations. Then, the aggregated cross-modal features are fed into the prompt encoder and mask decoder to generate the final audio-visual segmentation masks. We conduct extensive experiments on Flickr-SoundNet and AVSBench datasets. The results demonstrate that the proposed AV-SAM can achieve competitive performance on sounding object localization and segmentation."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "Segment Everything Everywhere All at Once",
                    "abstract": "In this work, we present SEEM, a promptable and interactive model for segmenting everything everywhere all at once in an image, as shown in Fig.1. In SEEM, we propose a novel decoding mechanism that enables diverse prompting for all types of segmentation tasks, aiming at a universal segmentation interface that behaves like large language models (LLMs). More specifically, SEEM is designed with four desiderata: i) Versatility. We introduce a new visual prompt to unify different spatial queries including points, boxes, scribbles and masks, which can further generalize to a different referring image; ii) Compositionality. We learn a joint visual-semantic space between text and visual prompts, which facilitates the dynamic composition of two prompt types required for various segmentation tasks; iii) Interactivity. We further incorporate learnable memory prompts into the decoder to retain segmentation history through mask-guided cross-attention from decoder to image features; and iv) Semantic-awareness. We use a text encoder to encode text queries and mask labels into the same semantic space for open-vocabulary segmentation. We conduct a comprehensive empirical study to validate the effectiveness of SEEM across diverse segmentation tasks. Notably, our single SEEM model achieves competitive performance across interactive segmentation, generic segmentation, referring segmentation, and video object segmentation on 9 datasets with minimum 1/100 supervision. Furthermore, SEEM showcases a remarkable capacity for generalization to novel prompts or their combinations, rendering it a readily universal image segmentation interface."
                },
                {
                    "title": "A Closer Look at the Explainability of Contrastive Language-Image Pre-training",
                    "abstract": "Contrastive language-image pre-training (CLIP) is a powerful vision-language model that has shown great benefits for various tasks. However, we have identified some issues with its explainability, which undermine its credibility and limit the capacity for related tasks. Specifically, we find that CLIP tends to focus on background regions rather than foregrounds, with noisy activations at irrelevant positions on the visualization results. These phenomena conflict with conventional explainability methods based on the class attention map (CAM), where the raw model can highlight the local foreground regions using global supervision without alignment. To address these problems, we take a closer look at its architecture and features. Based on thorough analyses, we find the raw self-attentions link to inconsistent semantic regions, resulting in the opposite visualization. Besides, the noisy activations are owing to redundant features among categories. Building on these insights, we propose the CLIP Surgery for reliable CAM, a method that allows surgery-like modifications to the inference architecture and features, without further fine-tuning as classical CAM methods. This approach significantly improves the explainability of CLIP, surpassing existing methods by large margins. Besides, it enables multimodal visualization and extends the capacity of raw CLIP on open-vocabulary tasks without extra alignment. The code is available at https://github.com/xmed-lab/CLIP_Surgery."
                },
                {
                    "title": "Segment Anything",
                    "abstract": "We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive \u2013 often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images at segment-anything.com to foster research into foundation models for computer vision. We recommend reading the full paper at: arxiv.org/abs/2304.02643."
                },
                {
                    "title": "Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection",
                    "abstract": "In this paper, we present an open-set object detector, called Grounding DINO, by marrying Transformer-based detector DINO with grounded pre-training, which can detect arbitrary objects with human inputs such as category names or referring expressions. The key solution of open-set object detection is introducing language to a closed-set detector for open-set concept generalization. To effectively fuse language and vision modalities, we conceptually divide a closed-set detector into three phases and propose a tight fusion solution, which includes a feature enhancer, a language-guided query selection, and a cross-modality decoder for cross-modality fusion. While previous works mainly evaluate open-set object detection on novel categories, we propose to also perform evaluations on referring expression comprehension for objects specified with attributes. Grounding DINO performs remarkably well on all three settings, including benchmarks on COCO, LVIS, ODinW, and RefCOCO/+/g. Grounding DINO achieves a $52.5$ AP on the COCO detection zero-shot transfer benchmark, i.e., without any training data from COCO. It sets a new record on the ODinW zero-shot benchmark with a mean $26.1$ AP. Code will be available at \\url{https://github.com/IDEA-Research/GroundingDINO}."
                },
                {
                    "title": "ViewCo: Discovering Text-Supervised Segmentation Masks via Multi-View Semantic Consistency",
                    "abstract": "Recently, great success has been made in learning visual representations from text supervision, facilitating the emergence of text-supervised semantic segmentation. However, existing works focus on pixel grouping and cross-modal semantic alignment, while ignoring the correspondence among multiple augmented views of the same image. To overcome such limitation, we propose multi-\\textbf{View} \\textbf{Co}nsistent learning (ViewCo) for text-supervised semantic segmentation. Specifically, we first propose text-to-views consistency modeling to learn correspondence for multiple views of the same input image. Additionally, we propose cross-view segmentation consistency modeling to address the ambiguity issue of text supervision by contrasting the segment features of Siamese visual encoders. The text-to-views consistency benefits the dense assignment of the visual features by encouraging different crops to align with the same text, while the cross-view segmentation consistency modeling provides additional self-supervision, overcoming the limitation of ambiguous text supervision for segmentation masks. Trained with large-scale image-text data, our model can directly segment objects of arbitrary categories in a zero-shot manner. Extensive experiments show that ViewCo outperforms state-of-the-art methods on average by up to 2.9\\%, 1.6\\%, and 2.4\\% mIoU on PASCAL VOC2012, PASCAL Context, and COCO, respectively."
                },
                {
                    "title": "Learning Open-Vocabulary Semantic Segmentation Models From Natural Language Supervision",
                    "abstract": "This paper considers the problem of open-vocabulary semantic segmentation (OVS), that aims to segment objects of arbitrary classes beyond a pre-defined, closed-set categories. The main contributions are as follows: First, we propose a transformer-based model for OVS, termed as OVSegmentor, which only exploits web-crawled imagetext pairs for pre-training without using any mask annotations. OVSegmentor assembles the image pixels into a set of learnable group tokens via a slotattention based binding module, then aligns the group tokens to corresponding caption embeddings. Second, we propose two proxy tasks for training, namely masked entity completion and cross-image mask consistency. The former aims to infer all masked entities in the caption given group tokens, that enables the model to learn fine-grained alignment between visual groups and text entities. The latter enforces consistent mask predictions between images that contain shared entities, encouraging the model to learn visual invariance. Third, we construct CC4M dataset for pre-training by filtering CC12M with frequently appeared entities, which significantly improves training efficiency. Fourth, we perform zero-shot transfer on four benchmark datasets, PASCAL VOC, PASCAL Context, COCO Object, and ADE20K. OVSegmentor achieves superior results over state-of-the-art approaches on PASCAL VOC using only 3% data (4M vs 134M) for pre-training."
                },
                {
                    "title": "Generalized Decoding for Pixel, Image, and Language",
                    "abstract": "We present X-Decoder, a generalized decoding model that can predict pixel-level segmentation and language tokens seamlessly. X-Decoder takes as input two types of queries: (i) generic non-semantic queries and (ii) semantic queries induced from text inputs, to decode different pixel-level and token-level outputs in the same semantic space. With such a novel design, X-Decoder is the first work that provides a unified way to support all types of image segmentation and a variety of vision-language (VL) tasks. Without any pseudo-labeling, our design enables seamless interactions across tasks at different granularities and brings mutual benefits by learning a common and rich pixel-level understanding. After pretraining on a mixed set of a limited amount of segmentation data and millions of image-text pairs, X-Decoder exhibits strong transferability to a wide range of downstream tasks in both zero-shot and finetuning settings. Notably, it achieves (1) state-of-the-art results on open-vocabulary segmentation and referring segmentation on seven datasets; (2) better or competitive finetuned performance to other generalist and specialist models on segmentation and VL tasks; and (3) flexibility for efficient fine-tuning and novel task composition (e.g., referring captioning and image editing shown in Fig. 1). Code, demo, video and visualization are available at: https://x-decoder-vl.github.io."
                },
                {
                    "title": "Self-Instruct: Aligning Language Models with Self-Generated Instructions",
                    "abstract": "Large \u201cinstruction-tuned\u201d language models (i.e., finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is often limited in quantity, diversity, and creativity, therefore hindering the generality of the tuned model. We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off their own generations. Our pipeline generates instructions, input, and output samples from a language model, then filters invalid or similar ones before using them to finetune the original model. Applying our method to the vanilla GPT3, we demonstrate a 33% absolute improvement over the original model on Super-NaturalInstructions, on par with the performance of InstructGPT-001, which was trained with private user data and human annotations. For further evaluation, we curate a set of expert-written instructions for novel tasks, and show through human evaluation that tuning GPT3 with Self-Instruct outperforms using existing public instruction datasets by a large margin, leaving only a 5% absolute gap behind InstructGPT-001. Self-Instruct provides an almost annotation-free method for aligning pre-trained language models with instructions, and we release our large synthetic dataset to facilitate future studies on instruction tuning."
                },
                {
                    "title": "Learning to Generate Text-Grounded Mask for Open-World Semantic Segmentation from Only Image-Text Pairs",
                    "abstract": "We tackle open-world semantic segmentation, which aims at learning to segment arbitrary visual concepts in images, by using only image-text pairs without dense annotations. Existing open-world segmentation methods have shown impressive advances by employing contrastive learning (CL) to learn diverse visual concepts and transferring the learned image-level understanding to the segmentation task. However, these CL-based methods suffer from a train-test discrepancy, since it only considers image-text alignment during training, whereas segmentation requires region-text alignment during testing. In this paper, we proposed a novel Text-grounded Contrastive Learning (TCL) framework that enables a model to directly learn region-text alignment. Our method generates a segmentation mask for a given text, extracts text-grounded image embedding from the masked region, and aligns it with text embedding via TCL. By learning region-text alignment directly, our framework encourages a model to directly improve the quality of generated segmentation masks. In addition, for a rigorous and fair comparison, we present a unified evaluation protocol with widely used 8 semantic segmentation datasets. TCL achieves state-of-the-art zero-shot segmentation performances with large margins in all datasets. Code is available at https://github.com/kakaobrain/tcl."
                },
                {
                    "title": "SegCLIP: Patch Aggregation with Learnable Centers for Open-Vocabulary Semantic Segmentation",
                    "abstract": "Recently, the contrastive language-image pre-training, e.g., CLIP, has demonstrated promising results on various downstream tasks. The pre-trained model can capture enriched visual concepts for images by learning from a large scale of text-image data. However, transferring the learned visual knowledge to open-vocabulary semantic segmentation is still under-explored. In this paper, we propose a CLIP-based model named SegCLIP for the topic of open-vocabulary segmentation in an annotation-free manner. The SegCLIP achieves segmentation based on ViT and the main idea is to gather patches with learnable centers to semantic regions through training on text-image pairs. The gathering operation can dynamically capture the semantic groups, which can be used to generate the final segmentation results. We further propose a reconstruction loss on masked patches and a superpixel-based KL loss with pseudo-labels to enhance the visual representation. Experimental results show that our model achieves comparable or superior segmentation accuracy on the PASCAL VOC 2012 (+0.3% mIoU), PASCAL Context (+2.3% mIoU), and COCO (+2.2% mIoU) compared with baselines. We release the code at https://github.com/ArrowLuo/SegCLIP."
                },
                {
                    "title": "MaPLe: Multi-modal Prompt Learning",
                    "abstract": "Pre-trained vision-language (V-L) models such as CLIP have shown excellent generalization ability to downstream tasks. However, they are sensitive to the choice of input text prompts and require careful selection of prompt templates to perform well. Inspired by the Natural Language Processing (NLP) literature, recent CLIP adaptation approaches learn prompts as the textual inputs to fine-tune CLIP for downstream tasks. We note that using prompting to adapt representations in a single branch of CLIP (language or vision) is sub-optimal since it does not allow the flexibility to dynamically adjust both representation spaces on a downstream task. In this work, we propose Multi-modal Prompt Learning (MaPLe) for both vision and language branches to improve alignment between the vision and language representations. Our design promotes strong coupling between the vision-language prompts to ensure mutual synergy and discourages learning independent uni-modal solutions. Further, we learn separate prompts across different early stages to progressively model the stage-wise feature relationships to allow rich context learning. We evaluate the effectiveness of our approach on three representative tasks of generalization to novel classes, new target datasets and unseen domain shifts. Compared with the state-of-the-art method Co-CoOp, MaPLe exhibits favorable performance and achieves an absolute gain of 3.45% on novel classes and 2.72% on overall harmonic-mean, averaged over 11 diverse image recognition datasets. Our code and pre-trained models are available at https://github.com/muzairkhattak/multimodal-prompt-learning."
                },
                {
                    "title": "Weakly-Supervised Camouflaged Object Detection with Scribble Annotations",
                    "abstract": "Existing camouflaged object detection (COD) methods rely heavily on large-scale datasets with pixel-wise annotations. However, due to the ambiguous boundary, annotating camouflage objects pixel-wisely is very time-consuming and labor-intensive, taking ~60mins to label one image. In this paper, we propose the first weakly-supervised COD method, using scribble annotations as supervision. To achieve this, we first relabel 4,040 images in existing camouflaged object datasets with scribbles, which takes ~10s to label one image. As scribble annotations only describe the primary structure of objects without details, for the network to learn to localize the boundaries of camouflaged objects, we propose a novel consistency loss composed of two parts: a cross-view loss to attain reliable consistency over different images, and an inside-view loss to maintain consistency inside a single prediction map. Besides, we observe that humans use semantic information to segment regions near the boundaries of camouflaged objects. Hence, we further propose a feature-guided loss, which includes visual features directly extracted from images and semantically significant features captured by the model. Finally, we propose a novel network for COD via scribble learning on structural information and semantic relations. Our network has two novel modules: the local-context contrasted (LCC) module, which mimics visual inhibition to enhance image contrast/sharpness and expand the scribbles into potential camouflaged regions, and the logical semantic relation (LSR) module, which analyzes the semantic relation to determine the regions representing the camouflaged object. Experimental results show that our model outperforms relevant SOTA methods on three COD benchmarks with an average improvement of 11.0% on MAE, 3.2% on S-measure, 2.5% on E-measure, and 4.4% on weighted F-measure."
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "Learning Unbiased Transferability for Domain Adaptation by Uncertainty Modeling",
                    "abstract": "Domain adaptation (DA) aims to transfer knowledge learned from a labeled source domain to an unlabeled or a less labeled but related target domain. Ideally, the source and target distributions should be aligned to each other equally to achieve unbiased knowledge transfer. However, due to the significant imbalance between the amount of annotated data in the source and target domains, usually only the target distribution is aligned to the source domain, leading to adapting unnecessary source specific knowledge to the target domain, i.e., biased domain adaptation. To resolve this problem, in this work, we delve into the transferability estimation problem in domain adaptation and propose a non-intrusive Unbiased Transferability Estimation Plug-in (UTEP) by modeling the uncertainty of a discriminator in adversarial-based DA methods to optimize unbiased transfer. We theoretically analyze the effectiveness of the proposed approach to unbiased transferability learning in DA. Furthermore, to alleviate the impact of imbalanced annotated data, we utilize the estimated uncertainty for pseudo label selection of unlabeled samples in the target domain, which helps achieve better marginal and conditional distribution alignments between domains. Extensive experimental results on a high variety of DA benchmark datasets show that the proposed approach can be readily incorporated into various adversarial-based DA methods, achieving state-of-the-art performance."
                },
                {
                    "title": "Exploring Visual Prompts for Adapting Large-Scale Models",
                    "abstract": "We investigate the efficacy of visual prompting to adapt large-scale models in vision. Following the recent approach from prompt tuning and adversarial reprogramming, we learn a single image perturbation such that a frozen model prompted with this perturbation performs a new task. Through comprehensive experiments, we demonstrate that visual prompting is particularly effective for CLIP and robust to distribution shift, achieving performance competitive with standard linear probes. We further analyze properties of the downstream dataset, prompt design, and output transformation in regard to adaptation performance. The surprising effectiveness of visual prompting provides a new perspective on adapting pre-trained models in vision. Code is available at http://hjbahng.github.io/visual_prompting ."
                },
                {
                    "title": "GroupViT: Semantic Segmentation Emerges from Text Supervision",
                    "abstract": "Grouping and recognition are important components of visual scene understanding, e.g., for object detection and semantic segmentation. With end-to-end deep learning systems, grouping of image regions usually happens implicitly via top-down supervision from pixel-level recognition labels. Instead, in this paper, we propose to bring back the grouping mechanism into deep networks, which allows semantic segments to emerge automatically with only text supervision. We propose a hierarchical Grouping Vision Transformer (GroupViT), which goes beyond the regular grid structure representation and learns to group image regions into progressively larger arbitrary-shaped segments. We train GroupViT jointly with a text encoder on a large-scale image-text dataset via contrastive losses. With only text supervision and without any pixel-level annotations, GroupViT learns to group together semantic regions and successfully transfers to the task of semantic segmentation in a zero-shot manner, i.e., without any further fine-tuning. It achieves a zero-shot accuracy of 52.3% mIoU on the PASCAL VOC 2012 and 22.4% mIoU on PASCAL Context datasets, and performs competitively to state-of-the-art transfer-learning methods requiring greater levels of supervision. We open-source our code at https://github.com/NVlabs/GroupViT."
                },
                {
                    "title": "FLAVA: A Foundational Language And Vision Alignment Model",
                    "abstract": "State-of-the-art vision and vision-and-language models rely on large-scale visio-linguistic pretraining for obtaining good performance on a variety of downstream tasks. Generally, such models are often either cross-modal (contrastive) or multi-modal (with earlier fusion) but not both; and they often only target specific modalities or tasks. A promising direction would be to use a single holistic universal model, as a \u201cfoundation\u201d, that targets all modalities at once-a true vision and language foundation model should be good at vision tasks, language tasks, and cross- and multi-modal vision and language tasks. We introduce FLAVA as such a model and demonstrate impressive performance on a wide range of 35 tasks spanning these target modalities."
                },
                {
                    "title": "Cognitive vision inspired object segmentation metric and loss function",
                    "abstract": "Object segmentation (OS) technology is a research hotspot in computer vision, and it has a wide range of applications in many fields. Cognitive vision studies have shown that human vision is highly sensitive to both global information and local details in scenes. To this end, we design a novel, e\ufb00icient, and easy-to-use Enhanced-alignment measure ( E \u03be ) for evaluating the performance of the OS model. E \u03be combines local pixel values with the image-level mean value, jointly evaluate the image-/pixel-level similarity between a segmentation result and a ground-truth (GT) results. Extensive experiments on the four popular benchmarks via five meta-measures, i.e., application ranking, demoting generic, denying noise, human ranking, and recognizing GT, we observe significant relative improvement compared with existing widely-adopted evaluation metrics such as IoU and F \u03b2 . By using the weighted binary cross-entropy loss, the Enhanced-alignment loss, and the weighted IoU loss, we further design a hybrid loss function (Hybrid-E loss ) to guide the network to learn pixel, object-and image-level features. Qualitative and quantitative results show further improvement in terms of accuracy when using our hybrid loss function in three different OS tasks. Source code: https://github.com/GewelsJI/Hybrid-Eloss/."
                },
                {
                    "title": "Domain Adaptive YOLO for One-Stage Cross-Domain Detection",
                    "abstract": "Domain shift is a major challenge for object detectors to generalize well to real world applications. Emerging techniques of domain adaptation for two-stage detectors help to tackle this problem. However, two-stage detectors are not the first choice for industrial applications due to its long time consumption. In this paper, a novel Domain Adaptive YOLO (DA-YOLO) is proposed to improve cross-domain performance for one-stage detectors. Image level features alignment is used to strictly match for local features like texture, and loosely match for global features like illumination. Multi-scale instance level features alignment is presented to reduce instance domain shift effectively , such as variations in object appearance and viewpoint. A consensus regularization to these domain classifiers is employed to help the network generate domain-invariant detections. We evaluate our proposed method on popular datasets like Cityscapes, KITTI, SIM10K and etc.. The results demonstrate significant improvement when tested under different cross-domain scenarios."
                },
                {
                    "title": "MERLOT: Multimodal Neural Script Knowledge Models",
                    "abstract": "As humans, we understand events in the visual world contextually, performing multimodal reasoning across time to make inferences about the past, present, and future. We introduce MERLOT, a model that learns multimodal script knowledge by watching millions of YouTube videos with transcribed speech -- in an entirely label-free, self-supervised manner. By pretraining with a mix of both frame-level (spatial) and video-level (temporal) objectives, our model not only learns to match images to temporally corresponding words, but also to contextualize what is happening globally over time. As a result, MERLOT exhibits strong out-of-the-box representations of temporal commonsense, and achieves state-of-the-art performance on 12 different video QA datasets when finetuned. It also transfers well to the world of static images, allowing models to reason about the dynamic context behind visual scenes. On Visual Commonsense Reasoning, MERLOT answers questions correctly with 80.6% accuracy, outperforming state-of-the-art models of similar size by over 3%, even those that make heavy use of auxiliary supervised data (like object bounding boxes). Ablation analyses demonstrate the complementary importance of: 1) training on videos versus static images; 2) scaling the magnitude and diversity of the pretraining video corpus; and 3) using diverse objectives that encourage full-stack multimodal reasoning, from the recognition to cognition level."
                },
                {
                    "title": "Rich Context Aggregation with Reflection Prior for Glass Surface Detection",
                    "abstract": "Glass surfaces appear everywhere. Their existence can however pose a serious problem to computer vision tasks. Recently, a method is proposed to detect glass surfaces by learning multi-scale contextual information. However, as it is only based on a general context integration operation and does not consider any specific glass surface properties, it gets confused when the images contain objects that are similar to glass surfaces and degenerates in challenging scenes with insufficient contexts. We observe that humans often rely on identifying reflections in order to sense the existence of glass and on locating the boundary in order to deter-mine the extent of the glass. Hence, we propose a model for glass surface detection, which consists of two novel modules: (1) a rich context aggregation module (RCAM) to ex-tract multi-scale boundary features from rich context features for locating glass surface boundaries of different sizes and shapes, and (2) a reflection-based refinement module (RRM) to detect reflection and then incorporate it so as to differentiate glass regions from non-glass regions. In addition, we also propose a challenging dataset consisting of 4,012 glass images with annotations for glass surface detection. Our experiments demonstrate that the proposed model outperforms state-of-the-art methods from relevant fields."
                },
                {
                    "title": "Neural Machine Translation with Monolingual Translation Memory",
                    "abstract": "Prior work has proved that Translation Memory (TM) can boost the performance of Neural Machine Translation (NMT). In contrast to existing work that uses bilingual corpus as TM and employs source-side similarity search for memory retrieval, we propose a new framework that uses monolingual memory and performs learnable memory retrieval in a cross-lingual manner. Our framework has unique advantages. First, the cross-lingual memory retriever allows abundant monolingual data to be TM. Second, the memory retriever and NMT model can be jointly optimized for the ultimate translation goal. Experiments show that the proposed method obtains substantial improvements. Remarkably, it even outperforms strong TM-augmented NMT baselines using bilingual TM. Owning to the ability to leverage monolingual data, our model also demonstrates effectiveness in low-resource and domain adaptation scenarios."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Concealed Object Detection",
                    "abstract": "We present the first systematic study on concealed object detection (COD), which aims to identify objects that are visually embedded in their background. The high intrinsic similarities between the concealed objects and their background make COD far more challenging than traditional object detection/segmentation. To better understand this task, we collect a large-scale dataset, called COD10K, which consists of 10,000 images covering concealed objects in diverse real-world scenarios from 78 object categories. Further, we provide rich annotations including object categories, object boundaries, challenging attributes, object-level labels, and instance-level annotations. Our COD10K is the largest COD dataset to date, with the richest annotations, which enables comprehensive concealed object understanding and can even be used to help progress several other vision tasks, such as detection, segmentation, classification etc. Motivated by how animals hunt in the wild, we also design a simple but strong baseline for COD, termed the Search Identification Network (SINet). Without any bells and whistles, SINet outperforms twelve cutting-edge baselines on all datasets tested, making them robust, general architectures that could serve as catalysts for future research in COD. Finally, we provide some interesting findings, and highlight several potential applications and future directions. To spark research in this new field, our code, dataset, and online demo are available at our project page: http://mmcheng.net/cod."
                },
                {
                    "title": "Conceptual 12M: Pushing Web-Scale Image-Text Pre-Training To Recognize Long-Tail Visual Concepts",
                    "abstract": "The availability of large-scale image captioning and visual question answering datasets has contributed significantly to recent successes in vision-and-language pretraining. However, these datasets are often collected with overrestrictive requirements inherited from their original target tasks (e.g., image caption generation), which limit the resulting dataset scale and diversity. We take a step further in pushing the limits of vision-and-language pretraining data by relaxing the data collection pipeline used in Conceptual Captions 3M (CC3M) [54] and introduce the Conceptual 12M (CC12M), a dataset with 12 million image-text pairs specifically meant to be used for visionand-language pre-training. We perform an analysis of this dataset and benchmark its effectiveness against CC3M on multiple downstream tasks with an emphasis on long-tail visual recognition. Our results clearly illustrate the benefit of scaling up pre-training data for vision-and-language tasks, as indicated by the new state-of-the-art results on both the nocaps and Conceptual Captions benchmarks.1"
                },
                {
                    "title": "Structure-Consistent Weakly Supervised Salient Object Detection with Local Saliency Coherence",
                    "abstract": "Sparse labels have been attracting much attention in recent years. However, the performance gap between weakly supervised and fully supervised salient object detection methods is huge, and most previous weakly supervised works adopt complex training methods with many bells and whistles. In this work, we propose a one-round end-to-end training approach for weakly supervised salient object detection via scribble annotations without pre/post-processing operations or extra supervision data. Since scribble labels fail to offer detailed salient regions, we propose a local coherence loss to propagate the labels to unlabeled regions based on image features and pixel distance, so as to predict integral salient regions with complete object structures. We design a saliency structure consistency loss as self-consistent mechanism to ensure consistent saliency maps are predicted with different scales of the same image as input, which could be viewed as a regularization technique to enhance the model generalization ability. Additionally, we design an aggregation module (AGGM) to better integrate high-level features, low-level features and global context information for the decoder to aggregate various information. Extensive experiments show that our method achieves a new state-of-the-art performance on six benchmarks (e.g. for the ECSSD dataset: F\u03b2 = 0.8995, E\u03be = 0.9079 and MAE = 0.0489), with an average gain of 4.60% for F-measure, 2.05% for E-measure and 1.88% for MAE over the previous best performing method on this task. Source code is available at http://github.com/siyueyu/SCWSSOD."
                },
                {
                    "title": "Weak-shot Fine-grained Classification via Similarity Transfer",
                    "abstract": "Recognizing fine-grained categories remains a challenging task, due to the subtle distinctions among different subordinate categories, which results in the need of abundant annotated samples. To alleviate the data-hungry problem, we consider the problem of learning novel categories from web data with the support of a clean set of base categories, which is referred to as weak-shot learning. Under this setting, we propose to transfer pairwise semantic similarity from base categories to novel categories, because this similarity is highly transferable and beneficial for learning from web data. Specifically, we firstly train a similarity net on clean data, and then employ two simple yet effective strategies to leverage the transferred similarity to denoise web training data. In addition, we apply adversarial loss on similarity net to enhance the transferability of similarity. Comprehensive experiments on three fine-grained datasets demonstrate that we could dramatically facilitate webly supervised learning by a clean set and similarity transfer is effective under this setting."
                },
                {
                    "title": "ToTTo: A Controlled Table-To-Text Generation Dataset",
                    "abstract": "We present ToTTo, an open-domain English table-to-text dataset with over 120,000 training examples that proposes a controlled generation task: given a Wikipedia table and a set of highlighted table cells, produce a one-sentence description. To obtain generated targets that are natural but also faithful to the source table, we introduce a dataset construction process where annotators directly revise existing candidate sentences from Wikipedia. We present systematic analyses of our dataset and annotation process as well as results achieved by several state-of-the-art baselines. While usually fluent, existing methods often hallucinate phrases that are not supported by the table, suggesting that this dataset can serve as a useful research benchmark for high-precision conditional text generation."
                },
                {
                    "title": "Weakly-Supervised Salient Object Detection via Scribble Annotations",
                    "abstract": "Compared with laborious pixel-wise dense labeling, it is much easier to label data by scribbles, which only costs 1~2 seconds to label one image. However, using scribble labels to learn salient object detection has not been explored. In this paper, we propose a weakly-supervised salient object detection model to learn saliency from such annotations. In doing so, we first relabel an existing large-scale salient object detection dataset with scribbles, namely S-DUTS dataset. Since object structure and detail information is not identified by scribbles, directly training with scribble labels will lead to saliency maps of poor boundary localization. To mitigate this problem, we propose an auxiliary edge detection task to localize object edges explicitly, and a gated structure-aware loss to place constraints on the scope of structure to be recovered. Moreover, we design a scribble boosting scheme to iteratively consolidate our scribble annotations, which are then employed as supervision to learn high-quality saliency maps. As existing saliency evaluation metrics neglect to measure structure alignment of the predictions, the saliency map ranking may not comply with human perception. We present a new metric, termed saliency structure measure, as a complementary metric to evaluate sharpness of the prediction. Extensive experiments on six benchmark datasets demonstrate that our method not only outperforms existing weakly-supervised/unsupervised methods, but also is on par with several fully-supervised state-of-the-art models (Our code and data is publicly available at: https://github.com/JingZhang617/Scribble_Saliency)."
                },
                {
                    "title": "Skin Lesion Analysis Toward Melanoma Detection 2018: A Challenge Hosted by the International Skin Imaging Collaboration (ISIC)",
                    "abstract": "This work summarizes the results of the largest skin image analysis challenge in the world, hosted by the International Skin Imaging Collaboration (ISIC), a global partnership that has organized the world's largest public repository of dermoscopic images of skin. The challenge was hosted in 2018 at the Medical Image Computing and Computer Assisted Intervention (MICCAI) conference in Granada, Spain. The dataset included over 12,500 images across 3 tasks. 900 users registered for data download, 115 submitted to the lesion segmentation task, 25 submitted to the lesion attribute detection task, and 159 submitted to the disease classification task. Novel evaluation protocols were established, including a new test for segmentation algorithm performance, and a test for algorithm ability to generalize. Results show that top segmentation algorithms still fail on over 10% of images on average, and algorithms with equal performance on test data can have different abilities to generalize. This is an important consideration for agencies regulating the growing set of machine learning tools in the healthcare domain, and sets a new standard for future public challenges in healthcare."
                },
                {
                    "title": "Conceptual Captions: A Cleaned, Hypernymed, Image Alt-text Dataset For Automatic Image Captioning",
                    "abstract": "We present a new dataset of image caption annotations, Conceptual Captions, which contains an order of magnitude more images than the MS-COCO dataset (Lin et al., 2014) and represents a wider variety of both images and image caption styles. We achieve this by extracting and filtering image caption annotations from billions of webpages. We also present quantitative evaluations of a number of image captioning models and show that a model architecture based on Inception-ResNetv2 (Szegedy et al., 2016) for image-feature extraction and Transformer (Vaswani et al., 2017) for sequence modeling achieves the best performance when trained on the Conceptual Captions dataset."
                },
                {
                    "title": "COCO-Stuff: Thing and Stuff Classes in Context",
                    "abstract": "Semantic classes can be either things (objects with a well-defined shape, e.g. car, person) or stuff (amorphous background regions, e.g. grass, sky). While lots of classification and detection works focus on thing classes, less attention has been given to stuff classes. Nonetheless, stuff classes are important as they allow to explain important aspects of an image, including (1) scene type; (2) which thing classes are likely to be present and their location (through contextual reasoning); (3) physical attributes, material types and geometric properties of the scene. To understand stuff and things in context we introduce COCO-Stuff1, which augments all 164K images of the COCO 2017 dataset with pixel-wise annotations for 91 stuff classes. We introduce an efficient stuff annotation protocol based on superpixels, which leverages the original thing annotations. We quantify the speed versus quality trade-off of our protocol and explore the relation between annotation time and boundary complexity. Furthermore, we use COCO-Stuff to analyze: (a) the importance of stuff and thing classes in terms of their surface cover and how frequently they are mentioned in image captions; (b) the spatial relations between stuff and things, highlighting the rich contextual relations that make our dataset unique; (c) the performance of a modern semantic segmentation method on stuff and thing classes, and whether stuff is easier to segment than things."
                },
                {
                    "title": "Automated Polyp Detection in Colonoscopy Videos Using Shape and Context Information",
                    "abstract": "This paper presents the culmination of our research in designing a system for computer-aided detection (CAD) of polyps in colonoscopy videos. Our system is based on a hybrid context-shape approach, which utilizes context information to remove non-polyp structures and shape information to reliably localize polyps. Specifically, given a colonoscopy image, we first obtain a crude edge map. Second, we remove non-polyp edges from the edge map using our unique feature extraction and edge classification scheme. Third, we localize polyp candidates with probabilistic confidence scores in the refined edge maps using our novel voting scheme. The suggested CAD system has been tested using two public polyp databases, CVC-ColonDB, containing 300 colonoscopy images with a total of 300 polyp instances from 15 unique polyps, and ASU-Mayo database, which is our collection of colonoscopy videos containing 19,400 frames and a total of 5,200 polyp instances from 10 unique polyps. We have evaluated our system using free-response receiver operating characteristic (FROC) analysis. At 0.1 false positives per frame, our system achieves a sensitivity of 88.0% for CVC-ColonDB and a sensitivity of 48% for the ASU-Mayo database. In addition, we have evaluated our system using a new detection latency analysis where latency is defined as the time from the first appearance of a polyp in the colonoscopy video to the time of its first detection by our system. At 0.05 false positives per frame, our system yields a polyp detection latency of 0.3 seconds."
                },
                {
                    "title": "The Role of Context for Object Detection and Semantic Segmentation in the Wild",
                    "abstract": "In this paper we study the role of context in existing state-of-the-art detection and segmentation approaches. Towards this goal, we label every pixel of PASCAL VOC 2010 detection challenge with a semantic category. We believe this data will provide plenty of challenges to the community, as it contains 520 additional classes for semantic segmentation and object detection. Our analysis shows that nearest neighbor based approaches perform poorly on semantic segmentation of contextual classes, showing the variability of PASCAL imagery. Furthermore, improvements of existing contextual models for detection is rather modest. In order to push forward the performance in this difficult scenario, we propose a novel deformable part-based model, which exploits both local context around each candidate detection as well as global context at the level of the scene. We show that this contextual reasoning significantly helps in detecting objects at all scales."
                },
                {
                    "title": "How to Evaluate Foreground Maps",
                    "abstract": "The output of many algorithms in computer-vision is either non-binary maps or binary maps (e.g., salient object detection and object segmentation). Several measures have been suggested to evaluate the accuracy of these foreground maps. In this paper, we show that the most commonly-used measures for evaluating both non-binary maps and binary maps do not always provide a reliable evaluation. This includes the Area-Under-the-Curve measure, the Average-Precision measure, the F\u03b2-measure, and the evaluation measure of the PASCAL VOC segmentation challenge. We start by identifying three causes of inaccurate evaluation. We then propose a new measure that amends these flaws. An appealing property of our measure is being an intuitive generalization of the F\u03b2-measure. Finally we propose four meta-measures to compare the adequacy of evaluation measures. We show via experiments that our novel measure is preferable."
                },
                {
                    "title": "Multi-Weight Partial Domain Adaptation",
                    "abstract": "Domain adaptation (DA) plays an important role in transfer learning. However, when target label space is a subset of source label space, standard DA cannot tackle this issue. Partial domain adaptation focuses on how to transfer knowledge from massive labelled dataset to unlabelled miniature one, which attracts extensive research interest. In this paper, we propose a Multi-Weight Partial Domain Adaptation (MWPDA) to solve the problem. We divide the source domain into two parts: shared classes and outlier classes. MWPDA aims to reduce negative transfer caused by outlier classes when transferring knowledge between domains. Based on OTSU-Algorithm, hard shared-class labels are obtained to decrease weights of outlier classes and increase ones of shared classes. A novel shared-sample classifier is trained for shared-sample weights to distinguish outlier samples. Shared-class weights and shared-sample weights are acted on source classifier and domain discriminator to jointly distinguish outlier classes and samples. This kind of multi-weight mechanism can avoid misalignment to outlier classes and promote classification accuracy. Furthermore, our universal network framework is utilized for both partial domain adaptation and standard domain adaptation issues. Extensive experiments on three benchmark domain adaptation datasets illustrate our method achieves state-of-the-art results."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively utilize hallucinations generated by Multimodal Large Language Models (MLLMs) to enhance task-specific segmentation in complex image scenarios?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it challenges the conventional view of hallucinations in MLLMs as detrimental. By reframing hallucinations as a valuable resource, this research could lead to significant advancements in segmentation techniques, particularly in scenarios where objects are not clearly visible. This could open new avenues for practical applications in fields such as wildlife monitoring, medical imaging, and autonomous driving, where accurate segmentation is essential. Furthermore, it may inspire future research to explore the potential of leveraging other forms of model-generated information, thereby broadening the scope of machine learning applications.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem stem from the inherent ambiguity and co-occurrence biases present in MLLM-generated hallucinations. Naive approaches that simply eliminate hallucinations may overlook their potential benefits, leading to suboptimal segmentation outcomes. The complexities arise from the need to distinguish between useful and irrelevant hallucinations, requiring sophisticated methods to verify and refine the generated prompts. Additionally, the iterative nature of the proposed solution necessitates a robust mechanism for visual masking verification, which adds to the technical and practical obstacles that must be overcome.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on mitigating the negative impacts of hallucinations in MLLMs, often viewing them as errors to be corrected rather than as potential assets. This perspective has limited the exploration of their utility in segmentation tasks. Existing methods have not adequately addressed the dual nature of hallucinations\u2014both as a source of contextual inference and as a means to enhance model performance. Our approach differs by intentionally leveraging hallucinations to extract valuable information, thus filling a significant gap in the literature and providing a novel framework for segmentation.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, the Prompt-Mask Cycle Generation method (ProMaC), involves a cyclic interaction between a prompt generator and a mask generator. The prompt generator employs a multi-scale chain-of-thought prompting mechanism to utilize hallucinations for generating instance-specific prompts. The mask generator iteratively verifies these prompts through visual masking, refining the segmentation process. We will evaluate our approach using a"
            }
        },
        "author_data": {
            "2bf5570b-cf89-4818-be2e-ca87ac88ee6a": {
                "pk": "2bf5570b-cf89-4818-be2e-ca87ac88ee6a",
                "name": "Jian Hu",
                "collaborators": [
                    "Gevorg Stepanyan",
                    "Shixin Liu",
                    "Liuyuan Fan",
                    "Guanwen Fang"
                ],
                "domain": [
                    "Astrophysics",
                    "Stochastic Optimization",
                    "Cosmology",
                    "Black Hole Physics"
                ],
                "publications": [
                    {
                        "title": "The black hole mass-bulge mass correlation: bulges versus pseudo-bulges",
                        "abstract": "We investigate the scaling relations between the supermassive black holes mass (M_bh) and the host bulge mass in elliptical galaxies, classical bulges, and pseudo-bulges. We use two-dimensional image analysis software BUDDA to obtain the structural parameters of 57 galaxies with dynamical M_bh measurement, and determine the bulge K-band luminosities (L_bul,K), stellar masses (M_s), and dynamical masses (M_d). The updated M_bh-L_bul,K, M_bh-M_s, and M_bh-M_d correlations for elliptical galaxies and classical bulges give M_bh~0.006M_s or 0.003M_d. The most tight relationship is log(M_bh/M_sun)=a+b log(M_d/10^11 M_sun), with a=8.46+/-0.05, b=0.90+/-0.06, and intrinsic scatter 0.27 dex. The pseudo-bulges follow their own relations, they harbor an order of magnitude smaller black holes than those in the same massive classical bulges, i.e. M_bh~0.0003M_s or 0.0002M_d. Besides the M_bh-\\sigma (bulge stellar velocity dispersion) relation, these bulge type dependent M_bh-M_bul scaling relations provide information for the growth and coevolution histories of SMBHs and their host bulges. We also find the core elliptical galaxies obey the same M_bh-M_d relation with other normal elliptical galaxies, that is expected in the dissipationless merger scenario."
                    },
                    {
                        "title": "The black hole mass -- stellar velocity dispersion correlation: bulges versus pseudobulges",
                        "abstract": "We investigate the correlation between the supermassive black holes (SMBHs) mass ($M_{\\rm bh}$) and the stellar velocity dispersion ($\\sigma_*$) in two types of host galaxies: the early-type bulges (disk galaxies with classical bulges or elliptical galaxies) and pseudobulges. In the form $\\log (M_{\\rm bh}/{\\rm M_\\odot})=\\alpha+\\beta\\log(\\sigma_*/200 {\\rm km s^{-1}})$, the best-fit results for the 39 early-type bulges are the slope $\\beta=4.06\\pm0.28$ and the normalization $\\alpha=8.28\\pm0.05$; the best-fit results for the 9 pseudobulges are $\\beta=4.5\\pm1.3$, $\\alpha=7.50\\pm0.18$. Both relations have intrinsic scatter in $\\log M_{\\rm bh}$ of $\\lesssim$0.27 dex. The $M_{\\rm bh}$-$\\sigma_*$ relation for pseudobulges is different from the relation in the early-type bulges over the 3$\\sigma$ significance level. The contrasting relations indicate the formation and growth histories of SMBHs depend on their host type. The discrepancy between the slope of the $M_{\\rm bh}$-$\\sigma_*$ relations using different definition of velocity dispersion vanishes in our sample, a uniform slope will constrain the coevolution theories of the SMBHs and their host galaxies more effectively. We also find the slope for the ``core'' elliptical galaxies at the high mass range of the relation appears steeper ($\\beta\\simeq$ 5-6), which may be the imprint of their origin of dissipationless mergers."
                    },
                    {
                        "title": "Exploring the possible evolution of the mass density power-law index of strong gravitational lenses with a model-independent method",
                        "abstract": "In this work, we adopt a cosmological model-independent approach for the first time to test the question of whether the mass density power-law index($\\gamma$) of the strong gravitational lensing system(SGLS) evolves with redshift, and the JLA SNe Ia sample and the quasar sample from Risaliti \\& Lusso (2019) are used to provide the luminosity distances to be calibrated. Our work is based on the flat universe assumption and the cosmic distance duality relation. A reliable data-matching method is used to pair SGLS-SNe and SGLS-quasar. By using the maximum likelihood method to constrain the luminosity distance and $\\gamma$ index, we obtain the likelihood function values for the evolved and non-evolved cases, and then use the Akaike weights and the BIC selection weights to compare the advantages and disadvantages of these two cases. We find that the $\\gamma$ index is slightly more likely to be a non-evolutionary model for $\\gamma=2$ in the case of the currently used samples with low redshift ($z_l<\\sim$0.66). With Akaike weights, the relative probability is 66.3\\% versus 33.7\\% and 69.9\\% versus 30.1\\% for the SGLS+SNe Ia sample and SGLS+quasar sample, respectively, and with BIC selection weights, the relative probability is 87.4\\% versus 12.6\\% and 52.0\\% versus 48.0\\% for the two samples. In the evolving case for the relatively low redshift lens (SGLS+SNe Ia), with redshift 0.0625 to 0.659, $\\gamma= 2.058^{+0.041}_{-0.040}-0.136^{+0.163}_{-0.165}z$. At high redshift (SGLS+quasar ), with redshift 0.0625 to 1.004, $\\gamma= 2.051^{+0.076}_{-0.077}-0.171^{+0.214}_{-0.196}z$. Although not the more likely model, this evolved $\\gamma$ case also fits the data well, with a negative and mild evolution for both low and high redshift samples."
                    },
                    {
                        "title": "Reference-Based Almost Stochastic Dominance Rules with Application in Risk-Averse Optimization",
                        "abstract": "Stochastic dominance is a preference relation of uncertain prospect defined over a class of utility functions. While this utility class represents basic properties of risk aversion, it includes some extreme utility functions rarely characterizing a rational decision maker's preference. In this paper we introduce reference-based almost stochastic dominance (RSD) rules which well balance the general representation of risk aversion and the individualization of the decision maker's risk preference. The key idea is that, in the general utility class, we construct a neighborhood of the decision maker's individual utility function, and represent a preference relation over this neighborhood. The RSD rules reveal the maximum dominance level quantifying the decision maker's robust preference between alternative choices. We also propose RSD constrained stochastic optimization model and develop an approximation algorithm based on Bernstein polynomials. This model is illustrated on a portfolio optimization problem."
                    },
                    {
                        "title": "Managing Distributional Ambiguity in Stochastic Optimization through a Statistical Upper Bound Framework",
                        "abstract": "Stochastic optimization is often hampered by distributional ambiguity, where critical probability distributions are poorly characterized or unknown. Addressing this challenge, we introduce a new framework that targets the minimization of a statistical upper bound for the expected value of uncertain objectives, facilitating more statistically robust decision-making. Central to our approach is the Average Percentile Upper Bound (APUB), a novel construct that simultaneously delivers a statistically rigorous upper bound for the population mean and a meaningful risk metric for the sample mean. The integration of APUB into stochastic optimization not only fortifies the process against distributional ambiguity but also reinforces key data-driven decision-making attributes, such as reliability, consistency, and comprehensibility. Notably, APUB-enriched optimization problems feature tractability, with particular advantages in two-stage stochastic optimization with random recourse. Empirical demonstrations on two-stage product mix and multi-product newsvendor benchmark problems reveal the benefit of the APUB optimization framework, in comparison with conventional techniques such as sample average approximation and distributionally robust optimization."
                    },
                    {
                        "title": "A comparison of cosmological models with high-redshift quasars",
                        "abstract": "The non-linear relationship between the monochromatic X-ray and UV luminosities in quasars offers the possibility of using high-z quasars as standard candles for cosmological testing. In this paper, we use a high-quality catalog of 1598 quasars extending to redshift 6, to compare the flat and uniformly expanding cosmological model, $R_h$ = ct and $\\Lambda$CDM cosmological models which are the most debated. The quasar samples are mainly from the XMM-Newton and the Sloan Digital Sky Survey (SDSS). The final result is that the Akaike Information Criterion favors $\\Lambda$CDM over $R_h$=ct with a relative probability of 86.30% versus 13.70%."
                    }
                ]
            },
            "6c48efad-8004-4cbd-ba33-9d8fb27392bc": {
                "pk": "6c48efad-8004-4cbd-ba33-9d8fb27392bc",
                "name": "Jiayi Lin",
                "collaborators": [
                    "Shaogang Gong",
                    "Yutao Xie",
                    "Lei Zhang",
                    "Hande Dong",
                    "Jian Hu",
                    "Jiayi Zhang",
                    "Chenxin Sun",
                    "Yue Gu",
                    "Qingyu Zhang",
                    "Xiaojiang Du"
                ],
                "domain": [
                    "Vision-Language Models",
                    "Code Understanding",
                    "Machine Learning",
                    "Video Super-Resolution"
                ],
                "publications": [
                    {
                        "title": "GridCLIP: One-Stage Object Detection by Grid-Level CLIP Representation Learning",
                        "abstract": "A vision-language foundation model pretrained on very large-scale image-text paired data has the potential to provide generalizable knowledge representation for downstream visual recognition and detection tasks, especially on supplementing the undersampled categories in downstream model training. Recent studies utilizing CLIP for object detection have shown that a two-stage detector design typically outperforms a one-stage detector, while requiring more expensive training resources and longer inference time. In this work, we propose a one-stage detector GridCLIP that narrows its performance gap to those of two-stage detectors, with approximately 43 and 5 times faster than its two-stage counterpart (ViLD) in the training and test process respectively. GridCLIP learns grid-level representations to adapt to the intrinsic principle of one-stage detection learning by expanding the conventional CLIP image-text holistic mapping to a more fine-grained, grid-text alignment. This differs from the region-text mapping in two-stage detectors that apply CLIP directly by treating regions as images. Specifically, GridCLIP performs Grid-level Alignment to adapt the CLIP image-level representations to grid-level representations by aligning to CLIP category representations to learn the annotated (especially frequent) categories. To learn generalizable visual representations of broader categories, especially undersampled ones, we perform Image-level Alignment during training to propagate broad pre-learned categories in the CLIP image encoder from the image-level to the grid-level representations. Experiments show that the learned CLIP-based grid-level representations boost the performance of undersampled (infrequent and novel) categories, reaching comparable detection performance on the LVIS benchmark."
                    },
                    {
                        "title": "Identify As A Human Does: A Pathfinder of Next-Generation Anti-Cheat Framework for First-Person Shooter Games",
                        "abstract": "The gaming industry has experienced substantial growth, but cheating in online games poses a significant threat to the integrity of the gaming experience. Cheating, particularly in first-person shooter (FPS) games, can lead to substantial losses for the game industry. Existing anti-cheat solutions have limitations, such as client-side hardware constraints, security risks, server-side unreliable methods, and both-sides suffer from a lack of comprehensive real-world datasets. To address these limitations, the paper proposes HAWK, a server-side FPS anti-cheat framework for the popular game CS:GO. HAWK utilizes machine learning techniques to mimic human experts' identification process, leverages novel multi-view features, and it is equipped with a well-defined workflow. The authors evaluate HAWK with the first large and real-world datasets containing multiple cheat types and cheating sophistication, and it exhibits promising efficiency and acceptable overheads, shorter ban times compared to the in-use anti-cheat, a significant reduction in manual labor, and the ability to capture cheaters who evaded official inspections."
                    },
                    {
                        "title": "FDAN: Flow-guided Deformable Alignment Network for Video Super-Resolution",
                        "abstract": "Most Video Super-Resolution (VSR) methods enhance a video reference frame by aligning its neighboring frames and mining information on these frames. Recently, deformable alignment has drawn extensive attention in VSR community for its remarkable performance, which can adaptively align neighboring frames with the reference one. However, we experimentally find that deformable alignment methods still suffer from fast motion due to locally loss-driven offset prediction and lack explicit motion constraints. Hence, we propose a Matching-based Flow Estimation (MFE) module to conduct global semantic feature matching and estimate optical flow as coarse offset for each location. And a Flow-guided Deformable Module (FDM) is proposed to integrate optical flow into deformable convolution. The FDM uses the optical flow to warp the neighboring frames at first. And then, the warped neighboring frames and the reference one are used to predict a set of fine offsets for each coarse offset. In general, we propose an end-to-end deep network called Flow-guided Deformable Alignment Network (FDAN), which reaches the state-of-the-art performance on two benchmark datasets while is still competitive in computation and memory consumption."
                    },
                    {
                        "title": "Scaling Laws Behind Code Understanding Model",
                        "abstract": "The scaling law is becoming a fundamental law in many machine learning areas. That is, test error falls off with the power law when increasing training data, model size, and computing resource. However, whether this law is suitable for the task of code understanding is not well studied, and most current language models for code understanding are about 100M parameters, which are relatively \"small\" compared to large language models. In this paper, we conduct extensive experiments to investigate the scaling law for the code understanding task by varying training data, model size, and computing resource. We validate that the test error of code understanding models falls off with the power law when using larger models, indicating that the scaling law is suitable for the code understanding task. Besides, we apply different scales of models to two downstream code understanding tasks, and find that the performance increases with larger scale of models. Finally, we train a large-scale code understanding model named CoLSBERT with 1.5B parameters on a large dataset using more computing resource, which outperforms previous work by a large margin. We will release our code and the CoLSBERT model when our paper is published."
                    },
                    {
                        "title": "InvSeg: Test-Time Prompt Inversion for Semantic Segmentation",
                        "abstract": "Visual-textual correlations in the attention maps derived from text-to-image diffusion models are proven beneficial to dense visual prediction tasks, e.g., semantic segmentation. However, a significant challenge arises due to the input distributional discrepancy between the context-rich sentences used for image generation and the isolated class names typically employed in semantic segmentation, hindering the diffusion models from capturing accurate visual-textual correlations. To solve this, we propose InvSeg, a test-time prompt inversion method that tackles open-vocabulary semantic segmentation by inverting image-specific visual context into text prompt embedding space, leveraging structure information derived from the diffusion model's reconstruction process to enrich text prompts so as to associate each class with a structure-consistent mask. Specifically, we introduce Contrastive Soft Clustering (CSC) to align derived masks with the image's structure information, softly selecting anchors for each class and calculating weighted distances to push inner-class pixels closer while separating inter-class pixels, thereby ensuring mask distinction and internal consistency. By incorporating sample-specific context, InvSeg learns context-rich text prompts in embedding space and achieves accurate semantic alignment across modalities. Experiments show that InvSeg achieves state-of-the-art performance on the PASCAL VOC and Context datasets. Project page: https://jylin8100.github.io/InvSegProject/."
                    },
                    {
                        "title": "Relax Image-Specific Prompt Requirement in SAM: A Single Generic Prompt for Segmenting Camouflaged Objects",
                        "abstract": "Camouflaged object detection (COD) approaches heavily rely on pixel-level annotated datasets. Weakly-supervised COD (WSCOD) approaches use sparse annotations like scribbles or points to reduce annotation effort, but this can lead to decreased accuracy. The Segment Anything Model (SAM) shows remarkable segmentation ability with sparse prompts like points. However, manual prompt is not always feasible, as it may not be accessible in real-world application. Additionally, it only provides localization information instead of semantic one, which can intrinsically cause ambiguity in interpreting the targets. In this work, we aim to eliminate the need for manual prompt. The key idea is to employ Cross-modal Chains of Thought Prompting (CCTP) to reason visual prompts using the semantic information given by a generic text prompt. To that end, we introduce a test-time adaptation per-instance mechanism called Generalizable SAM (GenSAM) to automatically enerate and optimize visual prompts the generic task prompt for WSCOD. In particular, CCTP maps a single generic text prompt onto image-specific consensus foreground and background heatmaps using vision-language models, acquiring reliable visual prompts. Moreover, to test-time adapt the visual prompts, we further propose Progressive Mask Generation (PMG) to iteratively reweight the input image, guiding the model to focus on the targets in a coarse-to-fine manner. Crucially, all network parameters are fixed, avoiding the need for additional training. Experiments demonstrate the superiority of GenSAM. Experiments on three benchmarks demonstrate that GenSAM outperforms point supervision approaches and achieves comparable results to scribble supervision ones, solely relying on general task descriptions as prompts. our codes is in: https://lwpyh.github.io/GenSAM/."
                    },
                    {
                        "title": "Survey of Code Search Based on Deep Learning",
                        "abstract": "Code writing is repetitive and predictable, inspiring us to develop various code intelligence techniques. This survey focuses on code search, that is, to retrieve code that matches a given query by effectively capturing the semantic similarity between the query and code. Deep learning, being able to extract complex semantics information, has achieved great success in this field. Recently, various deep learning methods, such as graph neural networks and pretraining models, have been applied to code search with significant progress. Deep learning is now the leading paradigm for code search. In this survey, we provide a comprehensive overview of deep learning-based code search. We review the existing deep learning-based code search framework which maps query/code to vectors and measures their similarity. Furthermore, we propose a new taxonomy to illustrate the state-of-the-art deep learning-based code search in a three-steps process: query semantics modeling, code semantics modeling, and matching modeling which involves the deep learning model training. Finally, we suggest potential avenues for future research in this promising field."
                    },
                    {
                        "title": "Toward Exploring the Code Understanding Capabilities of Pre-trained Code Generation Models",
                        "abstract": "Recently, large code generation models trained in a self-supervised manner on extensive unlabeled programming language data have achieved remarkable success. While these models acquire vast amounts of code knowledge, they perform poorly on code understanding tasks, such as code search and clone detection, as they are specifically trained for generation. Pre-training a larger encoder-only architecture model from scratch on massive code data can improve understanding performance. However, this approach is costly and time-consuming, making it suboptimal. In this paper, we pioneer the transfer of knowledge from pre-trained code generation models to code understanding tasks, significantly reducing training costs. We examine effective strategies for enabling decoder-only models to acquire robust code representations. Furthermore, we introduce CL4D, a contrastive learning method designed to enhance the representation capabilities of decoder-only models. Comprehensive experiments demonstrate that our approach achieves state-of-the-art performance in understanding tasks such as code search and clone detection. Our analysis shows that our method effectively reduces the distance between semantically identical samples in the representation space. These findings suggest the potential for unifying code understanding and generation tasks using a decoder-only structured model."
                    }
                ]
            },
            "a4ec536d-4e91-4c28-a9d3-3e18250b9b19": {
                "pk": "a4ec536d-4e91-4c28-a9d3-3e18250b9b19",
                "name": "Junchi Yan",
                "collaborators": [
                    "Hongyuan Zha",
                    "Ning Zhang",
                    "Ruoyu Cheng",
                    "Xue Yang",
                    "Xiaokang Yang",
                    "Xiaoyong Pan",
                    "Haoxuan Wang",
                    "Tianshu Yu",
                    "Jieyi Zhao",
                    "Baoxin Li"
                ],
                "domain": [
                    "Machine Learning",
                    "Deep Learning",
                    "Graph Neural Network",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Attention based convolutional neural network for predicting RNA-protein binding sites",
                        "abstract": "RNA-binding proteins (RBPs) play crucial roles in many biological processes, e.g. gene regulation. Computational identification of RBP binding sites on RNAs are urgently needed. In particular, RBPs bind to RNAs by recognizing sequence motifs. Thus, fast locating those motifs on RNA sequences is crucial and time-efficient for determining whether the RNAs interact with the RBPs or not. In this study, we present an attention based convolutional neural network, iDeepA, to predict RNA-protein binding sites from raw RNA sequences. We first encode RNA sequences into one-hot encoding. Next, we design a deep learning model with a convolutional neural network (CNN) and an attention mechanism, which automatically search for important positions, e.g. binding motifs, to learn discriminant high-level features for predicting RBP binding sites. We evaluate iDeepA on publicly gold-standard RBP binding sites derived from CLIP-seq data. The results demonstrate iDeepA achieves comparable performance with other state-of-the-art methods."
                    },
                    {
                        "title": "Melody Structure Transfer Network: Generating Music with Separable Self-Attention",
                        "abstract": "Symbolic music generation has attracted increasing attention, while most methods focus on generating short piece (mostly less than 8 bars, and up to 32 bars). Generating long music calls for effective expression of the coherent music structure. Despite their success on long sequences, self-attention architectures still have challenge in dealing with long-term music as it requires additional care on the subtle music structure. In this paper, we propose to transfer the structure of training samples for new music generation, and develop a novel separable self-attention based model which enable the learning and transferring of the structure embedding. We show that our transfer model can generate music sequences (up to 100 bars) with interpretable structures, which bears similar structures and composition techniques with the template music from training set. Extensive experiments show its ability of generating music with target structure and well diversity. The generated 3,000 sets of music is uploaded as supplemental material."
                    },
                    {
                        "title": "On Joint Learning for Solving Placement and Routing in Chip Design",
                        "abstract": "For its advantage in GPU acceleration and less dependency on human experts, machine learning has been an emerging tool for solving the placement and routing problems, as two critical steps in modern chip design flow. Being still in its early stage, there are fundamental issues: scalability, reward design, and end-to-end learning paradigm etc. To achieve end-to-end placement learning, we first propose a joint learning method termed by DeepPlace for the placement of macros and standard cells, by the integration of reinforcement learning with a gradient based optimization scheme. To further bridge the placement with the subsequent routing task, we also develop a joint learning approach via reinforcement learning to fulfill both macro placement and routing, which is called DeepPR. One key design in our (reinforcement) learning paradigm involves a multi-view embedding model to encode both global graph level and local node level information of the input macros. Moreover, the random network distillation is devised to encourage exploration. Experiments on public chip design benchmarks show that our method can effectively learn from experience and also provides intermediate placement for the post standard cell placement, within few hours for training."
                    },
                    {
                        "title": "Leveraging Angular Information Between Feature and Classifier for Long-tailed Learning: A Prediction Reformulation Approach",
                        "abstract": "Deep neural networks still struggle on long-tailed image datasets, and one of the reasons is that the imbalance of training data across categories leads to the imbalance of trained model parameters. Motivated by the empirical findings that trained classifiers yield larger weight norms in head classes, we propose to reformulate the recognition probabilities through included angles without re-balancing the classifier weights. Specifically, we calculate the angles between the data feature and the class-wise classifier weights to obtain angle-based prediction results. Inspired by the performance improvement of the predictive form reformulation and the outstanding performance of the widely used two-stage learning framework, we explore the different properties of this angular prediction and propose novel modules to improve the performance of different components in the framework. Our method is able to obtain the best performance among peer methods without pretraining on CIFAR10/100-LT and ImageNet-LT. Source code will be made publicly available."
                    },
                    {
                        "title": "On the Arbitrary-Oriented Object Detection: Classification based Approaches Revisited",
                        "abstract": "Arbitrary-oriented object detection has been a building block for rotation sensitive tasks. We first show that the boundary problem suffered in existing dominant regression-based rotation detectors, is caused by angular periodicity or corner ordering, according to the parameterization protocol. We also show that the root cause is that the ideal predictions can be out of the defined range. Accordingly, we transform the angular prediction task from a regression problem to a classification one. For the resulting circularly distributed angle classification problem, we first devise a Circular Smooth Label technique to handle the periodicity of angle and increase the error tolerance to adjacent angles. To reduce the excessive model parameters by Circular Smooth Label, we further design a Densely Coded Labels, which greatly reduces the length of the encoding. Finally, we further develop an object heading detection module, which can be useful when the exact heading orientation information is needed e.g. for ship and plane heading detection. We release our OHD-SJTU dataset and OHDet detector for heading detection. Extensive experimental results on three large-scale public datasets for aerial images i.e. DOTA, HRSC2016, OHD-SJTU, and face dataset FDDB, as well as scene text dataset ICDAR2015 and MLT, show the effectiveness of our approach."
                    },
                    {
                        "title": "Joint Cuts and Matching of Partitions in One Graph",
                        "abstract": "As two fundamental problems, graph cuts and graph matching have been investigated over decades, resulting in vast literature in these two topics respectively. However the way of jointly applying and solving graph cuts and matching receives few attention. In this paper, we first formalize the problem of simultaneously cutting a graph into two partitions i.e. graph cuts and establishing their correspondence i.e. graph matching. Then we develop an optimization algorithm by updating matching and cutting alternatively, provided with theoretical analysis. The efficacy of our algorithm is verified on both synthetic dataset and real-world images containing similar regions or structures."
                    },
                    {
                        "title": "AlphaRotate: A Rotation Detection Benchmark using TensorFlow",
                        "abstract": "AlphaRotate is an open-source Tensorflow benchmark for performing scalable rotation detection on various datasets. It currently provides more than 18 popular rotation detection models under a single, well-documented API designed for use by both practitioners and researchers. AlphaRotate regards high performance, robustness, sustainability and scalability as the core concept of design, and all models are covered by unit testing, continuous integration, code coverage, maintainability checks, and visual monitoring and analysis. AlphaRotate can be installed from PyPI and is released under the Apache-2.0 License. Source code is available at https://github.com/yangxue0827/RotationDetection."
                    },
                    {
                        "title": "Towards Machine Learning for Placement and Routing in Chip Design: a Methodological Overview",
                        "abstract": "Placement and routing are two indispensable and challenging (NP-hard) tasks in modern chip design flows. Compared with traditional solvers using heuristics or expert-well-designed algorithms, machine learning has shown promising prospects by its data-driven nature, which can be of less reliance on knowledge and priors, and potentially more scalable by its advanced computational paradigms (e.g. deep networks with GPU acceleration). This survey starts with the introduction of basics of placement and routing, with a brief description on classic learning-free solvers. Then we present detailed review on recent advance in machine learning for placement and routing. Finally we discuss the challenges and opportunities for future research."
                    },
                    {
                        "title": "A constrained clustering based approach for matching a collection of feature sets",
                        "abstract": "In this paper, we consider the problem of finding the feature correspondences among a collection of feature sets, by using their point-wise unary features. This is a fundamental problem in computer vision and pattern recognition, which also closely relates to other areas such as operational research. Different from two-set matching which can be transformed to a quadratic assignment programming task that is known NP-hard, inclusion of merely unary attributes leads to a linear assignment problem for matching two feature sets. This problem has been well studied and there are effective polynomial global optimum solvers such as the Hungarian method. However, it becomes ill-posed when the unary attributes are (heavily) corrupted. The global optimal correspondence concerning the best score defined by the attribute affinity/cost between the two sets can be distinct to the ground truth correspondence since the score function is biased by noises. To combat this issue, we devise a method for matching a collection of feature sets by synergetically exploring the information across the sets. In general, our method can be perceived from a (constrained) clustering perspective: in each iteration, it assigns the features of one set to the clusters formed by the rest of feature sets, and updates the cluster centers in turn. Results on both synthetic data and real images suggest the efficacy of our method against state-of-the-arts."
                    },
                    {
                        "title": "Decoupled Learning for Factorial Marked Temporal Point Processes",
                        "abstract": "This paper introduces the factorial marked temporal point process model and presents efficient learning methods. In conventional (multi-dimensional) marked temporal point process models, event is often encoded by a single discrete variable i.e. a marker. In this paper, we describe the factorial marked point processes whereby time-stamped event is factored into multiple markers. Accordingly the size of the infectivity matrix modeling the effect between pairwise markers is in power order w.r.t. the number of the discrete marker space. We propose a decoupled learning method with two learning procedures: i) directly solving the model based on two techniques: Alternating Direction Method of Multipliers and Fast Iterative Shrinkage-Thresholding Algorithm; ii) involving a reformulation that transforms the original problem into a Logistic Regression model for more efficient learning. Moreover, a sparse group regularizer is added to identify the key profile features and event labels. Empirical results on real world datasets demonstrate the efficiency of our decoupled and reformulated method. The source code is available online."
                    },
                    {
                        "title": "Weakly Supervised Audio Source Separation via Spectrum Energy Preserved Wasserstein Learning",
                        "abstract": "Separating audio mixtures into individual instrument tracks has been a long standing challenging task. We introduce a novel weakly supervised audio source separation approach based on deep adversarial learning. Specifically, our loss function adopts the Wasserstein distance which directly measures the distribution distance between the separated sources and the real sources for each individual source. Moreover, a global regularization term is added to fulfill the spectrum energy preservation property regardless separation. Unlike state-of-the-art weakly supervised models which often involve deliberately devised constraints or careful model selection, our approach need little prior model specification on the data, and can be straightforwardly learned in an end-to-end fashion. We show that the proposed method performs competitively on public benchmark against state-of-the-art weakly supervised methods."
                    },
                    {
                        "title": "Generalizing Face Forgery Detection with High-frequency Features",
                        "abstract": "Current face forgery detection methods achieve high accuracy under the within-database scenario where training and testing forgeries are synthesized by the same algorithm. However, few of them gain satisfying performance under the cross-database scenario where training and testing forgeries are synthesized by different algorithms. In this paper, we find that current CNN-based detectors tend to overfit to method-specific color textures and thus fail to generalize. Observing that image noises remove color textures and expose discrepancies between authentic and tampered regions, we propose to utilize the high-frequency noises for face forgery detection. We carefully devise three functional modules to take full advantage of the high-frequency features. The first is the multi-scale high-frequency feature extraction module that extracts high-frequency noises at multiple scales and composes a novel modality. The second is the residual-guided spatial attention module that guides the low-level RGB feature extractor to concentrate more on forgery traces from a new perspective. The last is the cross-modality attention module that leverages the correlation between the two complementary modalities to promote feature learning for each other. Comprehensive evaluations on several benchmark databases corroborate the superior generalization performance of our proposed method."
                    },
                    {
                        "title": "Monocular Visual Analysis for Electronic Line Calling of Tennis Games",
                        "abstract": "Electronic Line Calling is an auxiliary referee system used for tennis matches based on binocular vision technology. While ELC has been widely used, there are still many problems, such as complex installation and maintenance, high cost and etc. We propose a monocular vision technology based ELC method. The method has the following steps. First, locate the tennis ball's trajectory. We propose a multistage tennis ball positioning approach combining background subtraction and color area filtering. Then we propose a bouncing point prediction method by minimizing the fitting loss of the uncertain point. Finally, we find out whether the bouncing point of the ball is out of bounds or not according to the relative position between the bouncing point and the court side line in the two dimensional image. We collected and tagged 394 samples with an accuracy rate of 99.4%, and 81.8% of the 11 samples with bouncing points.The experimental results show that our method is feasible to judge if a ball is out of the court with monocular vision and significantly reduce complex installation and costs of ELC system with binocular vision."
                    },
                    {
                        "title": "IID-GAN: an IID Sampling Perspective for Regularizing Mode Collapse",
                        "abstract": "Despite its success, generative adversarial networks (GANs) still suffer from mode collapse, i.e., the generator can only map latent variables to a partial set of modes in the target distribution. In this paper, we analyze and seek to regularize this issue with an independent and identically distributed (IID) sampling perspective and emphasize that holding the IID property referring to the target distribution for generation can naturally avoid mode collapse. This is based on the basic IID assumption for real data in machine learning. However, though the source samples {z} obey IID, the generations {G(z)} may not necessarily be IID sampling from the target distribution. Based on this observation, considering a necessary condition of IID generation that the inverse samples from target data should also be IID in the source distribution, we propose a new loss to encourage the closeness between inverse samples of real data and the Gaussian source in latent space to regularize the generation to be IID from the target distribution. Experiments on both synthetic and real-world data show the effectiveness of our model."
                    },
                    {
                        "title": "DAAS: Differentiable Architecture and Augmentation Policy Search",
                        "abstract": "Neural architecture search (NAS) has been an active direction of automatic machine learning (Auto-ML), aiming to explore efficient network structures. The searched architecture is evaluated by training on datasets with fixed data augmentation policies. However, recent works on auto-augmentation show that the suited augmentation policies can vary over different structures. Therefore, this work considers the possible coupling between neural architectures and data augmentation and proposes an effective algorithm jointly searching for them. Specifically, 1) for the NAS task, we adopt a single-path based differentiable method with Gumbel-softmax reparameterization strategy due to its memory efficiency; 2) for the auto-augmentation task, we introduce a novel search method based on policy gradient algorithm, which can significantly reduce the computation complexity. Our approach achieves 97.91% accuracy on CIFAR-10 and 76.6% Top-1 accuracy on ImageNet dataset, showing the outstanding performance of our search algorithm."
                    },
                    {
                        "title": "From Quantum Graph Computing to Quantum Graph Learning: A Survey",
                        "abstract": "Quantum computing (QC) is a new computational paradigm whose foundations relate to quantum physics. Notable progress has been made, driving the birth of a series of quantum-based algorithms that take advantage of quantum computational power. In this paper, we provide a targeted survey of the development of QC for graph-related tasks. We first elaborate the correlations between quantum mechanics and graph theory to show that quantum computers are able to generate useful solutions that can not be produced by classical systems efficiently for some problems related to graphs. For its practicability and wide-applicability, we give a brief review of typical graph learning techniques designed for various tasks. Inspired by these powerful methods, we note that advanced quantum algorithms have been proposed for characterizing the graph structures. We give a snapshot of quantum graph learning where expectations serve as a catalyst for subsequent research. We further discuss the challenges of using quantum algorithms in graph learning, and future directions towards more flexible and versatile quantum graph learning solvers."
                    },
                    {
                        "title": "Learning Neural Hamiltonian Dynamics: A Methodological Overview",
                        "abstract": "The past few years have witnessed an increased interest in learning Hamiltonian dynamics in deep learning frameworks. As an inductive bias based on physical laws, Hamiltonian dynamics endow neural networks with accurate long-term prediction, interpretability, and data-efficient learning. However, Hamiltonian dynamics also bring energy conservation or dissipation assumptions on the input data and additional computational overhead. In this paper, we systematically survey recently proposed Hamiltonian neural network models, with a special emphasis on methodologies. In general, we discuss the major contributions of these models, and compare them in four overlapping directions: 1) generalized Hamiltonian system; 2) symplectic integration, 3) generalized input form, and 4) extended problem settings. We also provide an outlook of the fundamental challenges and emerging opportunities in this area."
                    },
                    {
                        "title": "Learning Universe Model for Partial Matching Networks over Multiple Graphs",
                        "abstract": "We consider the general setting for partial matching of two or multiple graphs, in the sense that not necessarily all the nodes in one graph can find their correspondences in another graph and vice versa. We take a universe matching perspective to this ubiquitous problem, whereby each node is either matched into an anchor in a virtual universe graph or regarded as an outlier. Such a universe matching scheme enjoys a few important merits, which have not been adopted in existing learning-based graph matching (GM) literature. First, the subtle logic for inlier matching and outlier detection can be clearly modeled, which is otherwise less convenient to handle in the pairwise matching scheme. Second, it enables end-to-end learning especially for universe level affinity metric learning for inliers matching, and loss design for gathering outliers together. Third, the resulting matching model can easily handle new arriving graphs under online matching, or even the graphs coming from different categories of the training set. To our best knowledge, this is the first deep learning network that can cope with two-graph matching, multiple-graph matching, online matching, and mixture graph matching simultaneously. Extensive experimental results show the state-of-the-art performance of our method in these settings."
                    },
                    {
                        "title": "Geometric Knowledge Distillation: Topology Compression for Graph Neural Networks",
                        "abstract": "We study a new paradigm of knowledge transfer that aims at encoding graph topological information into graph neural networks (GNNs) by distilling knowledge from a teacher GNN model trained on a complete graph to a student GNN model operating on a smaller or sparser graph. To this end, we revisit the connection between thermodynamics and the behavior of GNN, based on which we propose Neural Heat Kernel (NHK) to encapsulate the geometric property of the underlying manifold concerning the architecture of GNNs. A fundamental and principled solution is derived by aligning NHKs on teacher and student models, dubbed as Geometric Knowledge Distillation. We develop non- and parametric instantiations and demonstrate their efficacy in various experimental settings for knowledge distillation regarding different types of privileged topological information and teacher-student schemes."
                    },
                    {
                        "title": "Patch-Level Contrasting without Patch Correspondence for Accurate and Dense Contrastive Representation Learning",
                        "abstract": "We propose ADCLR: A ccurate and D ense Contrastive Representation Learning, a novel self-supervised learning framework for learning accurate and dense vision representation. To extract spatial-sensitive information, ADCLR introduces query patches for contrasting in addition with global contrasting. Compared with previous dense contrasting methods, ADCLR mainly enjoys three merits: i) achieving both global-discriminative and spatial-sensitive representation, ii) model-efficient (no extra parameters in addition to the global contrasting baseline), and iii) correspondence-free and thus simpler to implement. Our approach achieves new state-of-the-art performance for contrastive methods. On classification tasks, for ViT-S, ADCLR achieves 77.5% top-1 accuracy on ImageNet with linear probing, outperforming our baseline (DINO) without our devised techniques as plug-in, by 0.5%. For ViT-B, ADCLR achieves 79.8%, 84.0% accuracy on ImageNet by linear probing and finetune, outperforming iBOT by 0.3%, 0.2% accuracy. For dense tasks, on MS-COCO, ADCLR achieves significant improvements of 44.3% AP on object detection, 39.7% AP on instance segmentation, outperforming previous SOTA method SelfPatch by 2.2% and 1.2%, respectively. On ADE20K, ADCLR outperforms SelfPatch by 1.0% mIoU, 1.2% mAcc on the segme"
                    }
                ]
            },
            "1f0fd04e-fe77-4db4-a70a-ba15af00afe0": {
                "pk": "1f0fd04e-fe77-4db4-a70a-ba15af00afe0",
                "name": "Shaogang Gong",
                "collaborators": [
                    "Xiatian Zhu",
                    "Wei Li",
                    "Jiabo Huang",
                    "Guile Wu",
                    "Tao Xiang",
                    "Qi Dong",
                    "Xu Lan",
                    "Alexander Hudson",
                    "Yu Cao",
                    "Jiayi Lin"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Person Re-Identification",
                    "Transfer Learning"
                ],
                "publications": [
                    {
                        "title": "Transfer Learning for Protein Structure Classification at Low Resolution",
                        "abstract": "Structure determination is key to understanding protein function at a molecular level. Whilst significant advances have been made in predicting structure and function from amino acid sequence, researchers must still rely on expensive, time-consuming analytical methods to visualise detailed protein conformation. In this study, we demonstrate that it is possible to make accurate ($\\geq$80%) predictions of protein class and architecture from structures determined at low ($>$3A) resolution, using a deep convolutional neural network trained on high-resolution ($\\leq$3A) structures represented as 2D matrices. Thus, we provide proof of concept for high-speed, low-cost protein structure classification at low resolution, and a basis for extension to prediction of function. We investigate the impact of the input representation on classification performance, showing that side-chain information may not be necessary for fine-grained structure predictions. Finally, we confirm that high-resolution, low-resolution and NMR-determined structures inhabit a common feature space, and thus provide a theoretical foundation for boosting with single-image super-resolution."
                    },
                    {
                        "title": "Unsupervised Transfer Learning with Self-Supervised Remedy",
                        "abstract": "Generalising deep networks to novel domains without manual labels is challenging to deep learning. This problem is intrinsically difficult due to unpredictable changing nature of imagery data distributions in novel domains. Pre-learned knowledge does not transfer well without making strong assumptions about the learned and the novel domains. Different methods have been studied to address the underlying problem based on different assumptions, e.g. from domain adaptation to zero-shot and few-shot learning. In this work, we address this problem by transfer clustering that aims to learn a discriminative latent space of the unlabelled target data in a novel domain by knowledge transfer from labelled related domains. Specifically, we want to leverage relative (pairwise) imagery information, which is freely available and intrinsic to a target domain, to model the target domain image distribution characteristics as well as the prior-knowledge learned from related labelled domains to enable more discriminative clustering of unlabelled target data. Our method mitigates nontransferrable prior-knowledge by self-supervision, benefiting from both transfer and self-supervised learning. Extensive experiments on four datasets for image clustering tasks reveal the superiority of our model over the state-of-the-art transfer clustering techniques. We further demonstrate its competitive transferability on four zero-shot learning benchmarks."
                    },
                    {
                        "title": "Deep Clustering by Semantic Contrastive Learning",
                        "abstract": "Whilst contrastive learning has recently brought notable benefits to deep clustering of unlabelled images by learning sample-specific discriminative visual features, its potential for explicitly inferring class decision boundaries is less well understood. This is because its instance discrimination strategy is not class sensitive, therefore, the clusters derived on the resulting sample-specific feature space are not optimised for corresponding to meaningful class decision boundaries. In this work, we solve this problem by introducing Semantic Contrastive Learning (SCL). SCL imposes explicitly distance-based cluster structures on unlabelled training data by formulating a semantic (cluster-aware) contrastive learning objective. Moreover, we introduce a clustering consistency condition to be satisfied jointly by both instance visual similarities and cluster decision boundaries, and concurrently optimising both to reason about the hypotheses of semantic ground-truth classes (unknown/unlabelled) on-the-fly by their consensus. This semantic contrastive learning approach to discovering unknown class decision boundaries has considerable advantages to unsupervised learning of object recognition tasks. Extensive experiments show that SCL outperforms state-of-the-art contrastive learning and deep clustering methods on six object recognition benchmarks, especially on the more challenging finer-grained and larger datasets."
                    },
                    {
                        "title": "Few-Shot Image Generation by Conditional Relaxing Diffusion Inversion",
                        "abstract": "In the field of Few-Shot Image Generation (FSIG) using Deep Generative Models (DGMs), accurately estimating the distribution of target domain with minimal samples poses a significant challenge. This requires a method that can both capture the broad diversity and the true characteristics of the target domain distribution. We present Conditional Relaxing Diffusion Inversion (CRDI), an innovative `training-free' approach designed to enhance distribution diversity in synthetic image generation. Distinct from conventional methods, CRDI does not rely on fine-tuning based on only a few samples. Instead, it focuses on reconstructing each target image instance and expanding diversity through few-shot learning. The approach initiates by identifying a Sample-wise Guidance Embedding (SGE) for the diffusion model, which serves a purpose analogous to the explicit latent codes in certain Generative Adversarial Network (GAN) models. Subsequently, the method involves a scheduler that progressively introduces perturbations to the SGE, thereby augmenting diversity. Comprehensive experiments demonstrates that our method surpasses GAN-based reconstruction techniques and equals state-of-the-art (SOTA) FSIG methods in performance. Additionally, it effectively mitigates overfitting and catastrophic forgetting, common drawbacks of fine-tuning approaches."
                    },
                    {
                        "title": "Peer Collaborative Learning for Online Knowledge Distillation",
                        "abstract": "Traditional knowledge distillation uses a two-stage training strategy to transfer knowledge from a high-capacity teacher model to a compact student model, which relies heavily on the pre-trained teacher. Recent online knowledge distillation alleviates this limitation by collaborative learning, mutual learning and online ensembling, following a one-stage end-to-end training fashion. However, collaborative learning and mutual learning fail to construct an online high-capacity teacher, whilst online ensembling ignores the collaboration among branches and its logit summation impedes the further optimisation of the ensemble teacher. In this work, we propose a novel Peer Collaborative Learning method for online knowledge distillation, which integrates online ensembling and network collaboration into a unified framework. Specifically, given a target network, we construct a multi-branch network for training, in which each branch is called a peer. We perform random augmentation multiple times on the inputs to peers and assemble feature representations outputted from peers with an additional classifier as the peer ensemble teacher. This helps to transfer knowledge from a high-capacity teacher to peers, and in turn further optimises the ensemble teacher. Meanwhile, we employ the temporal mean model of each peer as the peer mean teacher to collaboratively transfer knowledge among peers, which helps each peer to learn richer knowledge and facilitates to optimise a more stable model with better generalisation. Extensive experiments on CIFAR-10, CIFAR-100 and ImageNet show that the proposed method significantly improves the generalisation of various backbone networks and outperforms the state-of-the-art methods."
                    },
                    {
                        "title": "Decentralised Learning from Independent Multi-Domain Labels for Person Re-Identification",
                        "abstract": "Deep learning has been successful for many computer vision tasks due to the availability of shared and centralised large-scale training data. However, increasing awareness of privacy concerns poses new challenges to deep learning, especially for human subject related recognition such as person re-identification (Re-ID). In this work, we solve the Re-ID problem by decentralised learning from non-shared private training data distributed at multiple user sites of independent multi-domain label spaces. We propose a novel paradigm called Federated Person Re-Identification (FedReID) to construct a generalisable global model (a central server) by simultaneously learning with multiple privacy-preserved local models (local clients). Specifically, each local client receives global model updates from the server and trains a local model using its local data independent from all the other clients. Then, the central server aggregates transferrable local model updates to construct a generalisable global feature embedding model without accessing local data so to preserve local privacy. This client-server collaborative learning process is iteratively performed under privacy control, enabling FedReID to realise decentralised learning without sharing distributed data nor collecting any centralised data. Extensive experiments on ten Re-ID benchmarks show that FedReID achieves compelling generalisation performance beyond any locally trained models without using shared training data, whilst inherently protects the privacy of each local client. This is uniquely advantageous over contemporary Re-ID methods."
                    },
                    {
                        "title": "GridCLIP: One-Stage Object Detection by Grid-Level CLIP Representation Learning",
                        "abstract": "A vision-language foundation model pretrained on very large-scale image-text paired data has the potential to provide generalizable knowledge representation for downstream visual recognition and detection tasks, especially on supplementing the undersampled categories in downstream model training. Recent studies utilizing CLIP for object detection have shown that a two-stage detector design typically outperforms a one-stage detector, while requiring more expensive training resources and longer inference time. In this work, we propose a one-stage detector GridCLIP that narrows its performance gap to those of two-stage detectors, with approximately 43 and 5 times faster than its two-stage counterpart (ViLD) in the training and test process respectively. GridCLIP learns grid-level representations to adapt to the intrinsic principle of one-stage detection learning by expanding the conventional CLIP image-text holistic mapping to a more fine-grained, grid-text alignment. This differs from the region-text mapping in two-stage detectors that apply CLIP directly by treating regions as images. Specifically, GridCLIP performs Grid-level Alignment to adapt the CLIP image-level representations to grid-level representations by aligning to CLIP category representations to learn the annotated (especially frequent) categories. To learn generalizable visual representations of broader categories, especially undersampled ones, we perform Image-level Alignment during training to propagate broad pre-learned categories in the CLIP image encoder from the image-level to the grid-level representations. Experiments show that the learned CLIP-based grid-level representations boost the performance of undersampled (infrequent and novel) categories, reaching comparable detection performance on the LVIS benchmark."
                    },
                    {
                        "title": "Mitigate Domain Shift by Primary-Auxiliary Objectives Association for Generalizing Person ReID",
                        "abstract": "While deep learning has significantly improved ReID model accuracy under the independent and identical distribution (IID) assumption, it has also become clear that such models degrade notably when applied to an unseen novel domain due to unpredictable/unknown domain shift. Contemporary domain generalization (DG) ReID models struggle in learning domain-invariant representation solely through training on an instance classification objective. We consider that a deep learning model is heavily influenced and therefore biased towards domain-specific characteristics, e.g., background clutter, scale and viewpoint variations, limiting the generalizability of the learned model, and hypothesize that the pedestrians are domain invariant owning they share the same structural characteristics. To enable the ReID model to be less domain-specific from these pure pedestrians, we introduce a method that guides model learning of the primary ReID instance classification objective by a concurrent auxiliary learning objective on weakly labeled pedestrian saliency detection. To solve the problem of conflicting optimization criteria in the model parameter space between the two learning objectives, we introduce a Primary-Auxiliary Objectives Association (PAOA) mechanism to calibrate the loss gradients of the auxiliary task towards the primary learning task gradients. Benefiting from the harmonious multitask learning design, our model can be extended with the recent test-time diagram to form the PAOA+, which performs on-the-fly optimization against the auxiliary objective in order to maximize the model's generative capacity in the test target domain. Experiments demonstrate the superiority of the proposed PAOA model."
                    },
                    {
                        "title": "Semantic Embedding Space for Zero-Shot Action Recognition",
                        "abstract": "The number of categories for action recognition is growing rapidly. It is thus becoming increasingly hard to collect sufficient training data to learn conventional models for each category. This issue may be ameliorated by the increasingly popular 'zero-shot learning' (ZSL) paradigm. In this framework a mapping is constructed between visual features and a human interpretable semantic description of each category, allowing categories to be recognised in the absence of any training data. Existing ZSL studies focus primarily on image data, and attribute-based semantic representations. In this paper, we address zero-shot recognition in contemporary video action recognition tasks, using semantic word vector space as the common space to embed videos and category labels. This is more challenging because the mapping between the semantic space and space-time features of videos containing complex actions is more complex and harder to learn. We demonstrate that a simple self-training and data augmentation strategy can significantly improve the efficacy of this mapping. Experiments on human action datasets including HMDB51 and UCF101 demonstrate that our approach achieves the state-of-the-art zero-shot action recognition performance."
                    },
                    {
                        "title": "Learning a Discriminative Null Space for Person Re-identification",
                        "abstract": "Most existing person re-identification (re-id) methods focus on learning the optimal distance metrics across camera views. Typically a person's appearance is represented using features of thousands of dimensions, whilst only hundreds of training samples are available due to the difficulties in collecting matched training images. With the number of training samples much smaller than the feature dimension, the existing methods thus face the classic small sample size (SSS) problem and have to resort to dimensionality reduction techniques and/or matrix regularisation, which lead to loss of discriminative power. In this work, we propose to overcome the SSS problem in re-id distance metric learning by matching people in a discriminative null space of the training data. In this null space, images of the same person are collapsed into a single point thus minimising the within-class scatter to the extreme and maximising the relative between-class separation simultaneously. Importantly, it has a fixed dimension, a closed-form solution and is very efficient to compute. Extensive experiments carried out on five person re-identification benchmarks including VIPeR, PRID2011, CUHK01, CUHK03 and Market1501 show that such a simple approach beats the state-of-the-art alternatives, often by a big margin."
                    },
                    {
                        "title": "Person Re-Identification by Deep Joint Learning of Multi-Loss Classification",
                        "abstract": "Existing person re-identification (re-id) methods rely mostly on either localised or global feature representation alone. This ignores their joint benefit and mutual complementary effects. In this work, we show the advantages of jointly learning local and global features in a Convolutional Neural Network (CNN) by aiming to discover correlated local and global features in different context. Specifically, we formulate a method for joint learning of local and global feature selection losses designed to optimise person re-id when using only generic matching metrics such as the L2 distance. We design a novel CNN architecture for Jointly Learning Multi-Loss (JLML) of local and global discriminative feature optimisation subject concurrently to the same re-id labelled information. Extensive comparative evaluations demonstrate the advantages of this new JLML model for person re-id over a wide range of state-of-the-art re-id methods on five benchmarks (VIPeR, GRID, CUHK01, CUHK03, Market-1501)."
                    },
                    {
                        "title": "Attribute Recognition by Joint Recurrent Learning of Context and Correlation",
                        "abstract": "Recognising semantic pedestrian attributes in surveillance images is a challenging task for computer vision, particularly when the imaging quality is poor with complex background clutter and uncontrolled viewing conditions, and the number of labelled training data is small. In this work, we formulate a Joint Recurrent Learning (JRL) model for exploring attribute context and correlation in order to improve attribute recognition given small sized training data with poor quality images. The JRL model learns jointly pedestrian attribute correlations in a pedestrian image and in particular their sequential ordering dependencies (latent high-order correlation) in an end-to-end encoder/decoder recurrent network. We demonstrate the performance advantage and robustness of the JRL model over a wide range of state-of-the-art deep models for pedestrian attribute recognition, multi-label image classification, and multi-person image annotation on two largest pedestrian attribute benchmarks PETA and RAP."
                    },
                    {
                        "title": "Class Rectification Hard Mining for Imbalanced Deep Learning",
                        "abstract": "Recognising detailed facial or clothing attributes in images of people is a challenging task for computer vision, especially when the training data are both in very large scale and extremely imbalanced among different attribute classes. To address this problem, we formulate a novel scheme for batch incremental hard sample mining of minority attribute classes from imbalanced large scale training data. We develop an end-to-end deep learning framework capable of avoiding the dominant effect of majority classes by discovering sparsely sampled boundaries of minority classes. This is made possible by introducing a Class Rectification Loss (CRL) regularising algorithm. We demonstrate the advantages and scalability of CRL over existing state-of-the-art attribute recognition and imbalanced data learning models on two large scale imbalanced benchmark datasets, the CelebA facial attribute dataset and the X-Domain clothing attribute dataset."
                    },
                    {
                        "title": "Harmonious Attention Network for Person Re-Identification",
                        "abstract": "Existing person re-identification (re-id) methods either assume the availability of well-aligned person bounding box images as model input or rely on constrained attention selection mechanisms to calibrate misaligned images. They are therefore sub-optimal for re-id matching in arbitrarily aligned person images potentially with large human pose variations and unconstrained auto-detection errors. In this work, we show the advantages of jointly learning attention selection and feature representation in a Convolutional Neural Network (CNN) by maximising the complementary information of different levels of visual attention subject to re-id discriminative learning constraints. Specifically, we formulate a novel Harmonious Attention CNN (HA-CNN) model for joint learning of soft pixel attention and hard regional attention along with simultaneous optimisation of feature representations, dedicated to optimise person re-id in uncontrolled (misaligned) images. Extensive comparative evaluations validate the superiority of this new HA-CNN model for person re-id over a wide variety of state-of-the-art methods on three large-scale benchmarks including CUHK03, Market-1501, and DukeMTMC-ReID."
                    },
                    {
                        "title": "Imbalanced Deep Learning by Minority Class Incremental Rectification",
                        "abstract": "Model learning from class imbalanced training data is a long-standing and significant challenge for machine learning. In particular, existing deep learning methods consider mostly either class balanced data or moderately imbalanced data in model training, and ignore the challenge of learning from significantly imbalanced training data. To address this problem, we formulate a class imbalanced deep learning model based on batch-wise incremental minority (sparsely sampled) class rectification by hard sample mining in majority (frequently sampled) classes during model training. This model is designed to minimise the dominant effect of majority classes by discovering sparsely sampled boundaries of minority classes in an iterative batch-wise learning process. To that end, we introduce a Class Rectification Loss (CRL) function that can be deployed readily in deep network architectures. Extensive experimental evaluations are conducted on three imbalanced person attribute benchmark datasets (CelebA, X-Domain, DeepFashion) and one balanced object category benchmark dataset (CIFAR-100). These experimental results demonstrate the performance advantages and model scalability of the proposed batch-wise incremental minority class rectification model over the existing state-of-the-art models for addressing the problem of imbalanced data learning."
                    },
                    {
                        "title": "Knowledge Distillation by On-the-Fly Native Ensemble",
                        "abstract": "Knowledge distillation is effective to train small and generalisable network models for meeting the low-memory and fast running requirements. Existing offline distillation methods rely on a strong pre-trained teacher, which enables favourable knowledge discovery and transfer but requires a complex two-phase training procedure. Online counterparts address this limitation at the price of lacking a highcapacity teacher. In this work, we present an On-the-fly Native Ensemble (ONE) strategy for one-stage online distillation. Specifically, ONE trains only a single multi-branch network while simultaneously establishing a strong teacher on-the- fly to enhance the learning of target network. Extensive evaluations show that ONE improves the generalisation performance a variety of deep neural networks more significantly than alternative methods on four image classification dataset: CIFAR10, CIFAR100, SVHN, and ImageNet, whilst having the computational efficiency advantages."
                    },
                    {
                        "title": "Person Re-Identification in Identity Regression Space",
                        "abstract": "Most existing person re-identification (re-id) methods are unsuitable for real-world deployment due to two reasons: Unscalability to large population size, and Inadaptability over time. In this work, we present a unified solution to address both problems. Specifically, we propose to construct an Identity Regression Space (IRS) based on embedding different training person identities (classes) and formulate re-id as a regression problem solved by identity regression in the IRS. The IRS approach is characterised by a closed-form solution with high learning efficiency and an inherent incremental learning capability with human-in-the-loop. Extensive experiments on four benchmarking datasets(VIPeR, CUHK01, CUHK03 and Market-1501) show that the IRS model not only outperforms state-of-the-art re-id methods, but also is more scalable to large re-id population size by rapidly updating model and actively selecting informative samples with reduced human labelling effort."
                    },
                    {
                        "title": "Person Search by Multi-Scale Matching",
                        "abstract": "We consider the problem of person search in unconstrained scene images. Existing methods usually focus on improving the person detection accuracy to mitigate negative effects imposed by misalignment, mis-detections, and false alarms resulted from noisy people auto-detection. In contrast to previous studies, we show that sufficiently reliable person instance cropping is achievable by slightly improved state-of-the-art deep learning object detectors (e.g. Faster-RCNN), and the under-studied multi-scale matching problem in person search is a more severe barrier. In this work, we address this multi-scale person search challenge by proposing a Cross-Level Semantic Alignment (CLSA) deep learning approach capable of learning more discriminative identity feature representations in a unified end-to-end model. This is realised by exploiting the in-network feature pyramid structure of a deep neural network enhanced by a novel cross pyramid-level semantic alignment loss function. This favourably eliminates the need for constructing a computationally expensive image pyramid and a complex multi-branch network architecture. Extensive experiments show the modelling advantages and performance superiority of CLSA over the state-of-the-art person search and multi-scale matching methods on two large person search benchmarking datasets: CUHK-SYSU and PRW."
                    },
                    {
                        "title": "Deep Association Learning for Unsupervised Video Person Re-identification",
                        "abstract": "Deep learning methods have started to dominate the research progress of video-based person re-identification (re-id). However, existing methods mostly consider supervised learning, which requires exhaustive manual efforts for labelling cross-view pairwise data. Therefore, they severely lack scalability and practicality in real-world video surveillance applications. In this work, to address the video person re-id task, we formulate a novel Deep Association Learning (DAL) scheme, the first end-to-end deep learning method using none of the identity labels in model initialisation and training. DAL learns a deep re-id matching model by jointly optimising two margin-based association losses in an end-to-end manner, which effectively constrains the association of each frame to the best-matched intra-camera representation and cross-camera representation. Existing standard CNNs can be readily employed within our DAL scheme. Experiment results demonstrate that our proposed DAL significantly outperforms current state-of-the-art unsupervised video person re-id methods on three benchmarks: PRID 2011, iLIDS-VID and MARS."
                    },
                    {
                        "title": "Unsupervised Person Re-identification by Deep Learning Tracklet Association",
                        "abstract": "Mostexistingpersonre-identification(re-id)methods relyon supervised model learning on per-camera-pair manually labelled pairwise training data. This leads to poor scalability in practical re-id deployment due to the lack of exhaustive identity labelling of image positive and negative pairs for every camera pair. In this work, we address this problem by proposing an unsupervised re-id deep learning approach capable of incrementally discovering and exploiting the underlying re-id discriminative information from automatically generated person tracklet data from videos in an end-to-end model optimisation. We formulate a Tracklet Association Unsupervised Deep Learning (TAUDL) framework characterised by jointly learning per-camera (within-camera) tracklet association (labelling) and cross-camera tracklet correlation by maximising the discovery of most likely tracklet relationships across camera views. Extensive experiments demonstrate the superiority of the proposed TAUDL model over the state-of-the-art unsupervised and domain adaptation re- id methods using six person re-id benchmarking datasets."
                    }
                ]
            }
        }
    },
    "2106.16239": {
        "paper_data": {
            "title": "Fixed points of nonnegative neural networks",
            "url": "http://arxiv.org/abs/2106.16239v9",
            "arxiv_id": "2106.16239",
            "authors": [
                "Tomasz J. Piotrowski",
                "Renato L. G. Cavalcante",
                "Mateusz Gabor"
            ],
            "abstract": "We use fixed point theory to analyze nonnegative neural networks, which we define as neural networks that map nonnegative vectors to nonnegative vectors. We first show that nonnegative neural networks with nonnegative weights and biases can be recognized as monotonic and (weakly) scalable mappings within the framework of nonlinear Perron-Frobenius theory. This fact enables us to provide conditions for the existence of fixed points of nonnegative neural networks having inputs and outputs of the same dimension, and these conditions are weaker than those recently obtained using arguments in convex analysis. Furthermore, we prove that the shape of the fixed point set of nonnegative neural networks with nonnegative weights and biases is an interval, which under mild conditions degenerates to a point. These results are then used to obtain the existence of fixed points of more general nonnegative neural networks. From a practical perspective, our results contribute to the understanding of the behavior of autoencoders, and we also offer valuable mathematical machinery for future developments in deep equilibrium models.",
            "introduction": "   1 Introduction  Neural networks consist of multiple layers that are able to extract patterns of input signals for decision-making without any human intervention. This fact has profound consequences in a wide range of wireless communications, image recognition, and speech processing tasks, where neural networks set high standards of performance LeCun et\u00a0al. (2015); Goodfellow et\u00a0al. (2016); Samek et\u00a0al. (2021). Neural networks have also been recently introduced as efficient iterative regularization methods in inverse problems, which opens a vista of new applications, especially in medical imaging Adler and \u00d6ktem (2017); Benning and Burger (2018); Ongie et\u00a0al. (2020).   Our study is specifically focused on nonnegative neural networks, defined as neural networks that map nonnegative vectors to nonnegative vectors. These networks have been widely used in many applications such as spectral analysis, image processing, and text processing Lemme et\u00a0al. (2012); Nguyen et\u00a0al. (2013); Chorowski and Zurada (2015); Hosseini-Asl et\u00a0al. (2015); Zhou et\u00a0al. (2016); Ali and Yangyu (2017); Ayinde and Zurada (2018); Su et\u00a0al. (2018); Chen et\u00a0al. (2019); Yu et\u00a0al. (2019); Palsson et\u00a0al. (2019). Nonnegative networks can be designed to decompose input signals into additive sparse components, leading to the formation of an interpretable hierarchical representation of the input data, mirroring the way the human brain processes sensory inputs Lee and Seung (1999); Shashanka et\u00a0al. (2008); Lemme et\u00a0al. (2012); Nguyen et\u00a0al. (2013); Chorowski and Zurada (2015); Ayinde and Zurada (2018). This principle has been exploited in autoencoders with nonnegative weights, which aims to break down input data into additive features within the encoding layer, which are then additively recombined in the decoding layer to reconstruct the original data Lemme et\u00a0al. (2012); Ayinde and Zurada (2018). In some scenarios, the gains in transparency and interpretation obtained by restricting the weights of the neural network to be nonnegative do not translate into a significant loss of predictive power Nguyen et\u00a0al. (2013); Chorowski and Zurada (2015).   Neural networks with nonnegative weights have also been increasingly gaining attention in recent years because of the resurgence of analog neural networks, which have been demonstrated to be both readily implementable and more energy efficient than their digital counterparts Kendall et\u00a0al. (2020); Paliy et\u00a0al. (2021). In analog networks, the weight parameters are physically encoded in the conductance of programmable resistors. Thus, these analog neural networks are examples of neural networks that are nonnegative because of the laws of physics. Moreover, various approaches can be used to mitigate the loss of predictive power caused by the physical nonnegative constraints Kendall et\u00a0al. (2020). These results are consistent with those obtained with the digital neural networks in Nguyen et\u00a0al. (2013); Chorowski and Zurada (2015).   With the objective of improving our understanding of nonnegative neural networks, we shed light on questions related to the existence of fixed points and the shape of the fixed point set of nonnegative neural networks (with matching input and output spaces). To illustrate the importance of fixed point analysis using concrete examples, let us first consider autoencoders, which can be seen as self-mappings taking the form T:\u2192kkT:{}^{k}\\to{}^{k}italic_T : start_FLOATSUPERSCRIPT italic_k end_FLOATSUPERSCRIPT \u2192 start_FLOATSUPERSCRIPT italic_k end_FLOATSUPERSCRIPT with x\u21a6T1\u2062(T2\u2062(x))maps-to\ud835\udc65subscript\ud835\udc471subscript\ud835\udc472\ud835\udc65x\\mapsto T_{1}(T_{2}(x))italic_x \u21a6 italic_T start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_T start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT",
            "references": [
                {
                    "title": "Positive concave deep equilibrium models",
                    "abstract": "Deep equilibrium (DEQ) models are widely recognized as a memory efficient alternative to standard neural networks, achieving state-of-the-art performance in language modeling and computer vision tasks. These models solve a fixed point equation instead of explicitly computing the output, which sets them apart from standard neural networks. However, existing DEQ models often lack formal guarantees of the existence and uniqueness of the fixed point, and the convergence of the numerical scheme used for computing the fixed point is not formally established. As a result, DEQ models are potentially unstable in practice. To address these drawbacks, we introduce a novel class of DEQ models called positive concave deep equilibrium (pcDEQ) models. Our approach, which is based on nonlinear Perron-Frobenius theory, enforces nonnegative weights and activation functions that are concave on the positive orthant. By imposing these constraints, we can easily ensure the existence and uniqueness of the fixed point without relying on additional complex assumptions commonly found in the DEQ literature, such as those based on monotone operator theory in convex analysis. Furthermore, the fixed point can be computed with the standard fixed point algorithm, and we provide theoretical guarantees of its geometric convergence, which, in particular, simplifies the training process. Experiments demonstrate the competitiveness of our pcDEQ models against other implicit models."
                },
                {
                    "title": "Deep equilibrium models as estimators for continuous latent variables",
                    "abstract": "Principal Component Analysis (PCA) and its exponential family extensions have three components: observations, latents and parameters of a linear transformation. We consider a generalised setting where the canonical parameters of the exponential family are a nonlinear transformation of the latents. We show explicit relationships between particular neural network architectures and the corresponding statistical models. We find that deep equilibrium models -- a recently introduced class of implicit neural networks -- solve maximum a-posteriori (MAP) estimates for the latents and parameters of the transformation. Our analysis provides a systematic way to relate activation functions, dropout, and layer structure, to statistical assumptions about the observations, thus providing foundational principles for unsupervised DEQs. For hierarchical latents, individual neurons can be interpreted as nodes in a deep graphical model. Our DEQ feature maps are end-to-end differentiable, enabling fine-tuning for downstream tasks."
                },
                {
                    "title": "Assessment of Two-Dimensional Materials-Based Technology for Analog Neural Networks",
                    "abstract": "Embedding advanced cognitive capabilities in battery-constrained edge devices requires specialized hardware with new circuit architecture and\u2014in the medium/long term\u2014new device technology. We evaluate the potential of recently investigated devices based on 2-D materials for the realization of analog deep neural networks (DNNs), by comparing the performance of neural networks based on the same circuit architecture using three different device technologies for transistors and analog memories. As a reference result, it is included in the comparison also an implementation on a standard 0.18 $\\mu \\text{m}$ CMOS technology. Our architecture of choice makes use of current-mode analog vector-matrix multipliers (VMMs) based on programmable current mirrors (CMs) consisting of transistors and floating-gate non-volatile memories. We consider experimentally demonstrated transistors and memories based on a monolayer molybdenum disulfide channel and ideal devices based on heterostructures of multilayer\u2013monolayer PtSe2. Following a consistent methodology for device-circuit co-design and optimization, we estimate the layout area, energy efficiency, and throughput as a function of the equivalent number of bits (ENOB), which is strictly correlated with classification accuracy. System-level tradeoffs are apparent: for a small ENOB experimental MoS2 floating-gate devices are already very promising; in our comparison, a larger ENOB (7 bits) is only achieved with CMOS, signaling the necessity to improve linearity and electrostatics of devices with 2-D materials."
                },
                {
                    "title": "The Fixed Point Iteration of Positive Concave Mappings Converges Geometrically if a Fixed Point Exists: Implications to Wireless Systems",
                    "abstract": "We prove that the fixed point iteration of arbitrary positive concave mappings with nonempty fixed point set converges geometrically for any starting point. We also show that positivity is crucial for this result to hold, and the concept of (nonlinear) spectral radius of asymptotic mappings provides us with information about the convergence factor. As a practical implication of the results shown here, we rigorously explain why some power control and load estimation algorithms in wireless networks, which are particular instances of a fixed point iteration, have shown geometric convergence in simulations. These algorithms have been typically derived by considering fixed point iterations of the general class of standard interference mappings, so the possibility of sublinear convergence rate could not be ruled out in previous studies, except in special cases that are often more restrictive than those considered here."
                },
                {
                    "title": "Explaining Deep Neural Networks and Beyond: A Review of Methods and Applications",
                    "abstract": "With the broader and highly successful usage of machine learning (ML) in industry and the sciences, there has been a growing demand for explainable artificial intelligence (XAI). Interpretability and explanation methods for gaining a better understanding of the problem-solving abilities and strategies of nonlinear ML, in particular, deep neural networks, are, therefore, receiving increased attention. In this work, we aim to: 1) provide a timely overview of this active emerging field, with a focus on \u201cpost hoc\u201d explanations, and explain its theoretical foundations; 2) put interpretability algorithms to a test both from a theory and comparative evaluation perspective using extensive simulations; 3) outline best practice aspects, i.e., how to best include interpretation methods into the standard usage of ML; and 4) demonstrate successful usage of XAI in a representative selection of application scenarios. Finally, we discuss challenges and possible future directions of this exciting foundational field of ML."
                },
                {
                    "title": "Deep Equilibrium Architectures for Inverse Problems in Imaging",
                    "abstract": "Recent efforts on solving inverse problems in imaging via deep neural networks use architectures inspired by a fixed number of iterations of an optimization method. The number of iterations is typically quite small due to difficulties in training networks corresponding to more iterations; the resulting solvers cannot be run for more iterations at test time without incurring significant errors. This paper describes an alternative approach corresponding to an infinite number of iterations, yielding a consistent improvement in reconstruction accuracy above state-of-the-art alternatives and where the computational budget can be selected at test time to optimize context-dependent trade-offs between accuracy and computation. The proposed approach leverages ideas from Deep Equilibrium Models, where the fixed-point iteration is constructed to incorporate a known forward model and insights from classical optimization-based reconstruction methods."
                },
                {
                    "title": "Effective Demand Failures and the Limits of Monetary Stabilization Policy",
                    "abstract": "The COVID-19 pandemic presents a challenge for stabilization policy that is different from those resulting from either \u201csupply\u201d or \u201cdemand\u201d shocks that similarly affect all sectors of the economy, owing to the degree to which the necessity of temporarily suspending some (but not all) economic activities disrupts the circular flow of payments, resulting in a failure of what Keynes (1936) calls \u201ceffective demand.\u201d In such a situation, economic activity in many sectors of the economy can be much lower than would maximize welfare (even taking into account the public health constraint), and interest-rate policy cannot eliminate the distortions \u2014 not because of a limit on the extent to which interest rates can be reduced, but because monetary stimulus fails to stimulate demand of the right sorts. Fiscal transfers are instead well-suited to addressing the fundamental problem, and can under certain circumstances achieve a first-best allocation of resources without any need for a monetary policy response."
                },
                {
                    "title": "Training End-to-End Analog Neural Networks with Equilibrium Propagation",
                    "abstract": "We introduce a principled method to train end-to-end analog neural networks by stochastic gradient descent. In these analog neural networks, the weights to be adjusted are implemented by the conductances of programmable resistive devices such as memristors [Chua, 1971], and the nonlinear transfer functions (or `activation functions') are implemented by nonlinear components such as diodes. We show mathematically that a class of analog neural networks (called nonlinear resistive networks) are energy-based models: they possess an energy function as a consequence of Kirchhoff's laws governing electrical circuits. This property enables us to train them using the Equilibrium Propagation framework [Scellier and Bengio, 2017]. Our update rule for each conductance, which is local and relies solely on the voltage drop across the corresponding resistor, is shown to compute the gradient of the loss function. Our numerical simulations, which use the SPICE-based Spectre simulation framework to simulate the dynamics of electrical circuits, demonstrate training on the MNIST classification task, performing comparably or better than equivalent-size software-based neural networks. Our work can guide the development of a new generation of ultra-fast, compact and low-power neural networks supporting on-chip learning."
                },
                {
                    "title": "Multiscale Deep Equilibrium Models",
                    "abstract": "We propose a new class of implicit networks, the multiscale deep equilibrium model (MDEQ), suited to large-scale and highly hierarchical pattern recognition domains. An MDEQ directly solves for and backpropagates through the equilibrium points of multiple feature resolutions simultaneously, using implicit differentiation to avoid storing intermediate states (and thus requiring only O(1) memory consumption). These simultaneously-learned multi-resolution features allow us to train a single model on a diverse set of tasks and loss functions, such as using a single MDEQ to perform both image classification and semantic segmentation. We illustrate the effectiveness of this approach on two large-scale vision tasks: ImageNet classification and semantic segmentation on high-resolution images from the Cityscapes dataset. In both settings, MDEQs are able to match or exceed the performance of recent competitive computer vision models: the first time such performance and scale have been achieved by an implicit deep learning approach. The code and pre-trained models are at this https URL ."
                },
                {
                    "title": "Deep Learning Techniques for Inverse Problems in Imaging",
                    "abstract": "Recent work in machine learning shows that deep neural networks can be used to solve a wide variety of inverse problems arising in computational imaging. We explore the central prevailing themes of this emerging area and present a taxonomy that can be used to categorize different problems and reconstruction methods. Our taxonomy is organized along two central axes: (1) whether or not a forward model is known and to what extent it is used in training and testing, and (2) whether or not the learning is supervised or unsupervised, i.e., whether or not the training relies on access to matched ground truth image and measurement pairs. We also discuss the tradeoffs associated with these different reconstruction approaches, caveats and common failure modes, plus open problems and avenues for future work."
                },
                {
                    "title": "Spectral-Spatial Hyperspectral Unmixing Using Multitask Learning",
                    "abstract": "Hyperspectral unmixing is an important and challenging task in the field of remote sensing which arises when the spatial resolution of sensors is insufficient for the separation of spectrally distinct materials. Hyperspectral images, like other natural images, have highly correlated pixels and it is very desirable to make use of this spatial information. In this paper, a deep learning based method for blind hyperspectral unmixing is presented. The method uses multitask learning through multiple parallel autoencoders to unmix a neighborhood of pixels simultaneously. Operating on image patches instead of single pixels enables the method to take advantage of spatial information in the hyperspectral image. The method is the first in its class to directly utilize the spatial structure of hyperspectral images (HSIs) for the estimation of the spectral signatures of endmembers in the data cube. We evaluate the proposed method using two real HSIs and compare it to seven state-of-the-art methods that either rely only on spectral or both on spectral and spatial information in the HSIs. The proposed method outperforms all the baseline unmixing methods in experiments."
                },
                {
                    "title": "A Note on Decoding Order in User Grouping and Power Optimization for Multi-Cell NOMA With Load Coupling",
                    "abstract": "In this technical note, we present a new theoretical result for multi-cell non-orthogonal multiple access (NOMA). For multi-cell scenarios, a so-called load-coupling model has been proposed earlier to characterize the presence of mutual interference for NOMA, and the optimization process relies on the use of fixed-point iterations across cells. One difficulty here is that the order of decoding for successive interference cancellation (SIC) in NOMA is generally not known a priori. This is because the decoding order in one cell depends on interference, which, in turn, is governed by resource usage in other cells, and vice versa. To achieve convergence, previous works have used workarounds that pose restrictions to NOMA, such that the SIC decoding order remains throughout the fixed-point iterations. As a comment to the previous works, we derive and prove the following result: The convergence is guaranteed, even if the order changes over the iterations. The result not only waives the need of previous workarounds, but also implies that a wide class of optimization problems for multi-cell NOMA is tractable, as long as that for single cell is."
                },
                {
                    "title": "The Structures of SINR Regions for Standard Interference Mappings",
                    "abstract": "In 1995, Yates proposed an axiomatic framework of standard interference mappings and examined the iterative power control algorithm for a system with interference constraints. In 2000s, Boche and Schubert built on a generalization of the theory of standard interference mappings and considered feasibility of the constraints by the interference mapping in their sense. In this paper, we consider the SINR (Signal to Interference and Noise Ratio) region of any continuous and standard interference mapping, i.e., the set of all realizable values of SINR and make clear the structure of the SINR region. We also show the relations between the SINR regions for any continuous and standard interference mapping and its asymptotic mapping. In addition, we discuss an optimization problem with SINR constraints. Under the assumption of feasibility of the problem, we prove that there exists a unique fixed point which is at the same time a unique optimal solution. Furthermore, we show the so-called max min balancing property holds at the optimal solution."
                },
                {
                    "title": "Deep Equilibrium Models",
                    "abstract": "We present a new approach to modeling sequential data: the deep equilibrium model (DEQ). Motivated by an observation that the hidden layers of many existing deep sequence models converge towards some fixed point, we propose the DEQ approach that directly finds these equilibrium points via root-finding. Such a method is equivalent to running an infinite depth (weight-tied) feedforward network, but has the notable advantage that we can analytically backpropagate through the equilibrium point using implicit differentiation. Using this approach, training and prediction in these networks require only constant memory, regardless of the effective \u201cdepth\u201d of the network. We demonstrate how DEQs can be applied to two state-of-the-art deep sequence models: self-attention transformers and trellis networks. On large-scale language modeling tasks, such as the WikiText-103 benchmark, we show that DEQs 1) often improve performance over these state-of-the-art models (for similar parameter counts); 2) have similar computational requirements to existing models; and 3) vastly reduce memory consumption (often the bottleneck for training large sequence models), demonstrating an up-to 88% memory reduction in our experiments. The code is available at https://github.com/locuslab/deq."
                },
                {
                    "title": "Iterative Neural Networks with Bounded Weights",
                    "abstract": "A recent analysis of a model of iterative neural network in Hilbert spaces established fundamental properties of such networks, such as existence of the fixed points sets, convergence analysis, and Lipschitz continuity. Building on these results, we show that under a single mild condition on the weights of the network, one is guaranteed to obtain a neural network converging to its unique fixed point. We provide a bound on the norm of this fixed point in terms of norms of weights and biases of the network. We also show why this model of a feed-forward neural network is not able to accomodate Hopfield networks under our assumption."
                },
                {
                    "title": "Adaptive Neural Signal Detection for Massive MIMO",
                    "abstract": "Traditional symbol detection algorithms either perform poorly or are impractical to implement for Massive Multiple-Input Multiple-Output (MIMO) systems. Recently, several learning-based approaches have achieved promising results on simple channel models (e.g., i.i.d. Gaussian channel coefficients), but as we show, their performance degrades on real-world channels with spatial correlation. We propose MMNet, a deep learning MIMO detection scheme that significantly outperforms existing approaches on realistic channels with the same or lower computational complexity. MMNet \u2019s design builds on the theory of iterative soft-thresholding algorithms, and uses a novel training algorithm that leverages temporal and spectral correlation in real channels to accelerate training. These innovations make it practical to train MMNet online for every realization of the channel. On i.i.d. Gaussian channels, MMNet requires two orders of magnitude fewer operations than existing deep learning schemes but achieves near-optimal performance. On spatially-correlated channels, it achieves the same error rate as the next-best learning scheme (OAMPNet) at 2.5dB lower signal-to-noise ratio (SNR), and with at least $10\\times $ less computational complexity. MMNet is also 4\u20138dB better overall than a classic linear scheme like the minimum mean square error (MMSE) detector."
                },
                {
                    "title": "Weakly Standard Interference Mappings: Existence of Fixed Points and Applications to Power Control in Wireless Networks",
                    "abstract": "We propose novel approaches to identify the existence of fixed points of the so-called weakly standard interference mappings, which include the well-known standard and general interference mappings as particular cases. The approaches are based on the concept of spectral radius of asymptotic mappings, a mathematical tool recently introduced to study the behavior of wireless networks. We show that, for arbitrary weakly standard interference mappings, knowledge of the spectral radius of an associated asymptotic mapping gives a sufficient condition to determine the existence of fixed points or their absence in the positive orthant. If the mapping has a fixed point, we further prove that the set of fixed points has a minimal element that can be easily computed with a simple fixed point algorithm. The theory developed here is applied to the problem of power control for load planning in LTE networks. Unlike previous approaches in the literature, the proposed solution takes into account the limited number of modulation and coding schemes of practical transceivers."
                },
                {
                    "title": "Solving inverse problems using data-driven models",
                    "abstract": "Recent research in inverse problems seeks to develop a mathematically coherent foundation for combining data-driven models, and in particular those based on deep learning, with domain-specific knowledge contained in physical\u2013analytical models. The focus is on solving ill-posed inverse problems that are at the core of many challenging applications in the natural sciences, medicine and life sciences, as well as in engineering and industrial applications. This survey paper aims to give an account of some of the main contributions in data-driven inverse problems."
                },
                {
                    "title": "Connections Between Spectral Properties of Asymptotic Mappings and Solutions to Wireless Network Problems",
                    "abstract": "In this study, we establish connections between asymptotic functions and properties of solutions to important problems in wireless networks. We start by introducing a class of self-mappings (called asymptotic mappings) constructed with asymptotic functions, and we show that spectral properties of these mappings explain the behavior of solutions to some max\u2013min utility optimization problems. For example, in a common family of max\u2013min utility power control problems, we prove that the optimal utility as a function of the power available to transmitters is approximately linear in the low power regime. However, as we move away from this regime, there exists a transition point, easily computed from the spectral radius of an asymptotic mapping, from which gains in utility become increasingly marginal. From these results we derive analogous properties of the transmit energy efficiency. In this study, we also generalize and unify existing approaches for feasibility analysis in wireless networks. Feasibility problems often reduce to determining the existence of the fixed point of a standard interference mapping, and we show that the spectral radius of an asymptotic mapping provides a necessary and sufficient condition for the existence of such a fixed point. We further present a result that determines whether the fixed point satisfies a constraint given in terms of a monotone norm."
                },
                {
                    "title": "Lipschitz Certificates for Layered Network Structures Driven by Averaged Activation Operators",
                    "abstract": "Obtaining sharp Lipschitz constants for feed-forward neural networks is essential to assess their robustness in the face of perturbations of their inputs. We derive such constants in the context of..."
                },
                {
                    "title": "Stacked Nonnegative Sparse Autoencoders for Robust Hyperspectral Unmixing",
                    "abstract": "As an unsupervised learning tool, autoencoder has been widely applied in many fields. In this letter, we propose a new robust unmixing algorithm that is based on stacked nonnegative sparse autoencoders (NNSAEs) for hyperspectral data with outliers and low signal-to-noise ratio. The proposed stacked autoencoders network contains two main steps. In the first step, a series of NNSAE is used to detect the outliers in the data. In the second step, a final autoencoder is performed for unmixing to achieve the endmember signatures and abundance fractions. By taking advantage from nonnegative sparse autoencoding, the proposed approach can well tackle problems with outliers and low noise-signal ratio. The effectiveness of the proposed method is evaluated on both synthetic and real hyperspectral data. In comparison with other unmixing methods, the proposed approach demonstrates competitive performance."
                },
                {
                    "title": "Optimal Control Via Neural Networks: A Convex Approach",
                    "abstract": "Control of complex systems involves both system identification and controller design. Deep neural networks have proven to be successful in many identification tasks, however, from model-based control perspective, these networks are difficult to work with because they are typically nonlinear and nonconvex. Therefore many systems are still identified and controlled based on simple linear models despite their poor representation capability. In this paper we bridge the gap between model accuracy and control tractability faced by neural networks, by explicitly constructing networks that are convex with respect to their inputs. We show that these input convex networks can be trained to obtain accurate models of complex physical systems. In particular, we design input convex recurrent neural networks to capture temporal behavior of dynamical systems. Then optimal controllers can be achieved via solving a convex model predictive control problem. Experiment results demonstrate the good potential of the proposed input convex neural network based approach in a variety of control applications. In particular we show that in the MuJoCo locomotion tasks, we could achieve over 10% higher performance using 5* less time compared with state-of-the-art model-based reinforcement learning method; and in the building HVAC control example, our method achieved up to 20% energy reduction compared with classic linear models."
                },
                {
                    "title": "Learning to Detect",
                    "abstract": "In this paper, we consider multiple-input-multiple-output detection using deep neural networks. We introduce two different deep architectures: a standard fully connected multi-layer network, and a detection network (DetNet), which is specifically designed for the task. The structure of DetNet is obtained by unfolding the iterations of a projected gradient descent algorithm into a network. We compare the accuracy and runtime complexity of the proposed approaches and achieve state-of-the-art performance while maintaining low computational requirements. Furthermore, we manage to train a single network to detect over an entire distribution of channels. Finally, we consider detection with soft outputs and show that the networks can easily be modified to produce soft decisions."
                },
                {
                    "title": "Spectral Radii of Asymptotic Mappings and the Convergence Speed of the Standard Fixed Point Algorithm",
                    "abstract": "Important problems in wireless networks can often be solved by computing fixed points of standard or contractive interference mappings, and the conventional fixed point algorithm is widely used for this purpose. Knowing that the mapping used in the algorithm is not only standard but also contractive (or only contractive) is valuable information because we obtain a guarantee of geometric convergence rate, and the rate is related to a property of the mapping called modulus of contraction. To date, contractive mappings and their moduli of contraction have been identified with case-by-case approaches that can be difficult to generalize. To address this limitation of existing approaches, we show in this study that the spectral radii of asymptotic mappings can be used to identify an important subclass of contractive mappings and also to estimate their moduli of contraction. In addition, if the fixed point algorithm is applied to compute fixed points of positive concave mappings, we show that the spectral radii of asymptotic mappings provide us with simple lower bounds for the estimation error of the iterates. An immediate application of this result proves that a known algorithm for load estimation in wireless networks becomes slower with increasing traffic."
                },
                {
                    "title": "Spectral Normalization for Generative Adversarial Networks",
                    "abstract": "One of the challenges in the study of generative adversarial networks is the instability of its training. In this paper, we propose a novel weight normalization technique called spectral normalization to stabilize the training of the discriminator. Our new normalization technique is computationally light and easy to incorporate into existing implementations. We tested the efficacy of spectral normalization on CIFAR10, STL-10, and ILSVRC2012 dataset, and we experimentally confirmed that spectrally normalized GANs (SN-GANs) is capable of generating images of better or equal quality relative to the previous training stabilization techniques."
                },
                {
                    "title": "Deep Learning of Constrained Autoencoders for Enhanced Understanding of Data",
                    "abstract": "Unsupervised feature extractors are known to perform an efficient and discriminative representation of data. Insight into the mappings they perform and human ability to understand them, however, remain very limited. This is especially prominent when multilayer deep learning architectures are used. This paper demonstrates how to remove these bottlenecks within the architecture of non-negativity constrained autoencoder. It is shown that using both L1 and L2 regularizations that induce non-negativity of weights, most of the weights in the network become constrained to be non-negative, thereby resulting into a more understandable structure with minute deterioration in classification accuracy. Also, this proposed approach extracts features that are more sparse and produces additional output layer sparsification. The method is analyzed for accuracy and feature interpretation on the MNIST data, the NORB normalized uniform object data, and the Reuters text categorization data set."
                },
                {
                    "title": "Modern regularization methods for inverse problems",
                    "abstract": "Regularization methods are a key tool in the solution of inverse problems. They are used to introduce prior knowledge and allow a robust approximation of ill-posed (pseudo-) inverses. In the last two decades interest has shifted from linear to nonlinear regularization methods, even for linear inverse problems. The aim of this paper is to provide a reasonably comprehensive overview of this shift towards modern nonlinear regularization methods, including their analysis, applications and issues for future research. In particular we will discuss variational methods and techniques derived from them, since they have attracted much recent interest and link to other fields, such as image processing and compressed sensing. We further point to developments related to statistical inverse problems, multiscale decompositions and learning theory."
                },
                {
                    "title": "Automatic Modulation Classification Using Deep Learning Based on Sparse Autoencoders With Nonnegativity Constraints",
                    "abstract": "We demonstrate a novel method for the automatic modulation classification based on a deep learning autoencoder network, trained by a nonnegativity constraint algorithm. The learning algorithm aims to constrain the negative weights, learns features that amount to a part-based representation of data, and disentangles a more meaningful hidden structure. The performance of this algorithm is tested on the fourth-order cumulants of the modulated signals. The results indicate that the autoencoder with nonnegativity constraint (ANC) improves the sparsity and minimizes the reconstruction error in comparison with the conventional sparse autoencoder. The classification accuracy of an ANC based deep network shows improved accuracy under limited signal length and fading channel."
                },
                {
                    "title": "Recurrent Inference Machines for Solving Inverse Problems",
                    "abstract": "Much of the recent research on solving iterative inference problems focuses on moving away from hand-chosen inference algorithms and towards learned inference. In the latter, the inference process is unrolled in time and interpreted as a recurrent neural network (RNN) which allows for joint learning of model and inference parameters with back-propagation through time. In this framework, the RNN architecture is directly derived from a hand-chosen inference algorithm, effectively limiting its capabilities. We propose a learning framework, called Recurrent Inference Machines (RIM), in which we turn algorithm construction the other way round: Given data and a task, train an RNN to learn an inference algorithm. Because RNNs are Turing complete [1, 2] they are capable to implement any inference algorithm. The framework allows for an abstraction which removes the need for domain knowledge. We demonstrate in several image restoration experiments that this abstraction is effective, allowing us to achieve state-of-the-art performance on image denoising and super-resolution tasks and superior across-task generalization."
                },
                {
                    "title": "Solving ill-posed inverse problems using iterative deep neural networks",
                    "abstract": "We propose a partially learned approach for the solution of ill-posed inverse problems with not necessarily linear forward operators. The method builds on ideas from classical regularisation theory and recent advances in deep learning to perform learning while making use of prior information about the inverse problem encoded in the forward operator, noise model and a regularising functional. The method results in a gradient-like iterative scheme, where the \u2018gradient\u2019 component is learned using a convolutional network that includes the gradients of the data discrepancy and regulariser as input in each iteration. We present results of such a partially learned gradient scheme on a non-linear tomographic inversion problem with simulated data from both the Sheep-Logan phantom as well as a head CT. The outcome is compared against filtered backprojection and total variation reconstruction and the proposed method provides a 5.4 dB PSNR improvement over the total variation reconstruction while being significantly faster, giving reconstructions of 512\u00d7512 pixel images in about 0.4\u2009s using a single graphics processing unit (GPU)."
                },
                {
                    "title": "Max-Min Utility Optimization in Load Coupled Interference Networks",
                    "abstract": "We propose a novel utility optimization algorithm for wireless networks modeled by systems of nonlinear equations based on the load at the base stations. Unlike previous studies, the algorithm solves a max\u2013min utility optimization problem over the joint space of network load, transmit power, and rates. In more detail, our first main contribution is to show that, in the optimum, users operate at the same rate, base stations are fully loaded, and at least one base station transmits at the maximum power. This characterization of the optimal solution enables a reformulation of the optimization task as a conditional eigenvalue problem associated with a concave mapping that relates the transmit power to the network load. With this reformulation, an efficient iterative solver becomes readily available. Our second main contribution is the derivation of a simple lower bound for conditional eigenvalues of general positive concave mappings. These bounds are of particular interest to network designers, because conditional eigenvalues can often be related to the optimal rates (or the optimal signal-to-interference-noise ratio) of a large class of utility optimization problems, and, in this paper, these bounds are used to derive performance limits of load coupled networks."
                },
                {
                    "title": "Deep Learning of Part-Based Representation of Data Using Sparse Autoencoders With Nonnegativity Constraints",
                    "abstract": "We demonstrate a new deep learning autoencoder network, trained by a nonnegativity constraint algorithm (nonnegativity-constrained autoencoder), that learns features that show part-based representation of data. The learning algorithm is based on constraining negative weights. The performance of the algorithm is assessed based on decomposing data into parts and its prediction performance is tested on three standard image data sets and one text data set. The results indicate that the nonnegativity constraint forces the autoencoder to learn features that amount to a part-based representation of data, while improving sparsity and reconstruction quality in comparison with the traditional sparse autoencoder and nonnegative matrix factorization. It is also shown that this newly acquired representation improves the prediction performance of a deep neural network."
                },
                {
                    "title": "Augmentable Gamma Belief Networks",
                    "abstract": "To infer multilayer deep representations of high-dimensional discrete and nonnegative real vectors, we propose an augmentable gamma belief network (GBN) that factorizes each of its hidden layers into the product of a sparse connection weight matrix and the nonnegative real hidden units of the next layer. The GBN's hidden layers are jointly trained with an upward-downward Gibbs sampler that solves each layer with the same subroutine. The gamma-negative binomial process combined with a layer-wise training strategy allows inferring the width of each layer given a fixed budget on the width of the first layer. Example results illustrate interesting relationships between the width of the first layer and the inferred network structure, and demonstrate that the GBN can add more layers to improve its performance in both unsupervisedly extracting features and predicting heldout data. For exploratory data analysis, we extract trees and subnetworks from the learned deep network to visualize how the very specific factors discovered at the first hidden layer and the increasingly more general factors discovered at deeper hidden layers are related to each other, and we generate synthetic data by propagating random variables through the deep network from the top hidden layer back to the bottom data layer."
                },
                {
                    "title": "Elementary Properties of Positive Concave Mappings With Applications to Network Planning and Optimization",
                    "abstract": "This paper presents novel methods for computing fixed points of positive concave mappings and for characterizing the existence of fixed points. These methods are particularly important in planning and optimization tasks in wireless networks. For example, previous studies have shown that the feasibility of a network design can be quickly evaluated by computing the fixed point of a concave mapping that is constructed based on many environmental and network control parameters such as the position of base stations, channel conditions, and antenna tilts. To address this and more general problems, given a positive concave mapping, we show two alternative but equivalent ways to construct a matrix that is guaranteed to have spectral radius strictly smaller than one if the mapping has a fixed point. This matrix is then used to build a new mapping that preserves the fixed point of the original positive concave mapping. We show that the standard fixed point iterations using the new mapping converges faster than the standard iterations applied to the original concave mapping. As exemplary applications of the proposed methods, we consider the problems of power and load estimation in networks based on the orthogonal frequency division multiple access (OFDMA) technology."
                },
                {
                    "title": "Positive Dynamical Systems in Discrete Time: Theory, Models, and Applications",
                    "abstract": "This book provides a systematic, rigorous and self-contained treatment of positive dynamical systems. A dynamical system is positive when all relevant variables of a systemare nonnegative in a natural way. This is in biology, demography or economics, where the levels of populations or prices of goods are positive. The principle also finds application in electrical engineering, physics and computer sciences."
                },
                {
                    "title": "Toward Energy-Efficient 5G Wireless Communications Technologies",
                    "abstract": "The densification and expansion of wireless networks pose new challenges on energy efficiency. With a drastic increase of infrastructure nodes (e.g. ultra-dense deployment of small cells), the total energy consumption may easily exceed an acceptable level. While most studies focus on the energy radiated by the antennas, the bigger part of the total energy budget is actually consumed by the hardware (e.g., coolers and circuit energy consumption). The ability to shutdown infrastructure nodes (or parts of it) or to adapt the transmission strategy according to the traffic will therefore become an important design aspect of future wireless architectures. Network infrastructure should be regarded as a resource that can be occupied or released on demand. However, the modeling and optimization of such systems are complicated by the potential interference coupling between active nodes. In this article, we give an overview on different aspects of this problem. We show how prior knowledge of traffic patterns can be exploited for optimization. Then, we discuss the framework of interference functions, which has proved a useful tool for various types of coupled systems in the literature. Finally, we introduce different classes of algorithms that have the objective of improving the energy efficiency by adapting the network configuration to the traffic load."
                },
                {
                    "title": "Nonlinear Perron-Frobenius Theory",
                    "abstract": "In the past several decades the classical Perron\u2013Frobenius theory for nonnegative matrices has been extended to obtain remarkably precise and beautiful results for classes of nonlinear maps. This nonlinear Perron\u2013Frobenius theory has found significant uses in computer science, mathematical biology, game theory and the study of dynamical systems. This is the first comprehensive and unified introduction to nonlinear Perron\u2013Frobenius theory suitable for graduate students and researchers entering the field for the first time. It acquaints the reader with recent developments and provides a guide to challenging open problems. To enhance accessibility, the focus is on finite dimensional nonlinear Perron\u2013Frobenius theory, but pointers are provided to infinite dimensional results. Prerequisites are little more than basic real analysis and topology."
                },
                {
                    "title": "Convex Optimization",
                    "abstract": "This textbook is based on lectures given by the authors at MIPT (Moscow), HSE (Moscow), FEFU (Vladivostok), V.I. Vernadsky KFU (Simferopol), ASU (Republic of Adygea), and the University of Grenoble-Alpes (Grenoble, France). First of all, the authors focused on the program of a two-semester course of lectures on convex optimization, which is given to students of MIPT. The first chapter of this book contains the materials of the first semester (\"Fundamentals of convex analysis and optimization\"), the second and third chapters contain the materials of the second semester (\"Numerical methods of convex optimization\"). The textbook has a number of features. First, in contrast to the classic manuals, this book does not provide proofs of all the theorems mentioned. This allowed, on one side, to describe more themes, but on the other side, made the presentation less self-sufficient. The second important point is that part of the material is advanced and is published in the Russian educational literature, apparently for the first time. Third, the accents that are given do not always coincide with the generally accepted accents in the textbooks that are now popular. First of all, we talk about a sufficiently advanced presentation of conic optimization, including robust optimization, as a vivid demonstration of the capabilities of modern convex analysis."
                },
                {
                    "title": "Concave and Convex Interference Functions\u2014General Characterizations and Applications",
                    "abstract": "Many resource allocation problems can be studied within the framework of interference functions. Basic properties of interference functions are non-negativity, scale-invariance, and monotonicity. In this paper, we study interference functions with additional properties, namely convexity, concavity, and log-convexity. Such interference functions occur naturally in various contexts, e.g., adaptive receive strategies, robust power control, or resource allocation over convex utility sets. We show that every convex (resp. concave) interference function can be expressed as a maximum (resp. minimum) over a weighted sum of its arguments. This analytical insight provides a link between the axiomatic interference framework and conventional interference models that are based on the definition of a coupling matrix. We show how the results can be used to derive best-possible convex/concave approximations for general interference functions. The results have further application in the context of feasible sets of multiuser systems. Convex approximations for general feasible sets are derived. Finally, we show how convexity can be exploited to solve the problem of signal-to-interference-plus-noise ratio (SINR)-constrained power minimization with super-linear convergence."
                },
                {
                    "title": "Probabilistic Latent Variable Models as Nonnegative Factorizations",
                    "abstract": "This paper presents a family of probabilistic latent variable models that can be used for analysis of nonnegative data. We show that there are strong ties between nonnegative matrix factorization and this family, and provide some straightforward extensions which can help in dealing with shift invariances, higher-order decompositions and sparsity constraints. We argue through these extensions that the use of this approach allows for rapid development of complex statistical models for analyzing nonnegative data."
                },
                {
                    "title": "Bayesian brain : probabilistic approaches to neural coding",
                    "abstract": "A Bayesian approach can contribute to an understanding of the brain on multiple levels, by giving normative predictions about how an ideal sensory system should combine prior knowledge and observation, by providing mechanistic interpretation of the dynamic functioning of the brain circuit, and by suggesting optimal ways of deciphering experimental data. Bayesian Brain brings together contributions from both experimental and theoretical neuroscientists that examine the brain mechanisms of perception, decision making, and motor control according to the concepts of Bayesian estimation.After an overview of the mathematical concepts, including Bayes' theorem, that are basic to understanding the approaches discussed, contributors discuss how Bayesian concepts can be used for interpretation of such neurobiological data as neural spikes and functional brain imaging. Next, contributors examine the modeling of sensory processing, including the neural coding of information about the outside world. Finally, contributors explore dynamic processes for proper behaviors, including the mathematics of the speed and accuracy of perceptual decisions and neural models of belief propagation."
                },
                {
                    "title": "Extension of order-preserving maps on a cone",
                    "abstract": "We examine the problem of extending, in a natural way, order-preserving maps that are defined on the interior of a closed cone K1 (taking values in another closed cone K2) to the whole of K1. We give conditions, in considerable generality (for cones in both finite- and infinite-dimensional spaces), under which a natural extension exists and is continuous. We also give weaker conditions under which the extension is upper semi-continuous. Maps f defined on the interior of the non-negative cone K in RN, which are both homogeneous of degree 1 and order preserving, are non-expanding in the Thompson metric, and hence continuous. As a corollary of our main results, we deduce that all such maps have a homogeneous order-preserving continuous extension to the whole cone. It follows that such an extension must have at least one eigenvector in K \u2013 {0}. In the case where the cycle time \u03c7(f) of the original map does not exist, such eigenvectors must lie in \u2202K \u2013 {0}. We conclude with some discussions and applications to operator-valued means. We also extend our results to an \u2018intermediate\u2019 situation, which arises in some important application areas, particularly in the construction of diffusions on certain fractals via maps defined on the interior of cones of Dirichlet forms."
                },
                {
                    "title": "Matrix Analysis and Applied Linear Algebra",
                    "abstract": "Preface 1. Linear equations 2. Rectangular systems and echelon forms 3. Matrix algebra 4. Vector spaces 5. Norms, inner products, and orthogonality 6. Determinants 7. Eigenvalues and Eigenvectors 8. Perron-Frobenius theory of nonnegative matrices Index."
                },
                {
                    "title": "Channel Covariance Conversion and Modelling Using Infinite Dimensional Hilbert Spaces",
                    "abstract": "This study considers the channel covariance conversion problem, which consists in estimating the spatial covariance matrix of a wireless channel by exploiting measurements obtained on a different carrier frequency and stationarity properties of the propagation environment across sufficiently close frequency bands. The first contribution given in this study is a modelling framework based on infinite dimensional Hilbert spaces that unifies a plethora of classical and novel covariance models with different degrees of complexity and generality, while still effectively capturing important properties of the propagation environment and of the antenna array. Given this framework, this study addresses the channel covariance conversion problem by proposing two simple yet effective algorithms based on set-theoretic methods that outperform existing model-based approaches both in terms of accuracy and complexity. In particular, the first algorithm is implementable as a simple matrix-vector multiplication. Moreover, in contrast to the aforementioned approaches, both algorithms can be applied to general propagation and array models such as dual-polarized antenna arrays, making them suitable for modern 5 G and beyond systems."
                },
                {
                    "title": "Generalized score matching for non-negative data",
                    "abstract": "Evaluating the joint signi\ufb01cance of covariates is of fundamental importance in a wide range of applications. To this end, p-values are frequently employed and produced by algorithms that are powered by classical large-sample asymptotic theory. It is well known that the conventional p-values in Gaussian linear model are valid even when the dimensionality is a non-vanishing fraction of the sample size, but can break down when the design matrix becomes singular in higher dimensions or when the error distribution deviates from Gaus-sianity. A natural question is when the conventional p-values in generalized linear models become invalid in diverging dimensions. We establish that such a breakdown can occur early in nonlinear models. Our theoretical characterizations are con\ufb01rmed by simulation studies."
                },
                {
                    "title": "Learning Understandable Neural Networks With Nonnegative Weight Constraints",
                    "abstract": "People can understand complex structures if they relate to more isolated yet understandable concepts. Despite this fact, popular pattern recognition tools, such as decision tree or production rule learners, produce only flat models which do not build intermediate data representations. On the other hand, neural networks typically learn hierarchical but opaque models. We show how constraining neurons' weights to be nonnegative improves the interpretability of a network's operation. We analyze the proposed method on large data sets: the MNIST digit recognition data and the Reuters text categorization data. The patterns learned by traditional and constrained network are contrasted to those learned with principal component analysis and nonnegative matrix factorization."
                },
                {
                    "title": "Learning Parts-based Representations with Nonnegative Restricted Boltzmann Machine",
                    "abstract": "The success of any machine learning system depends critically on eective representations of data. In many cases, especially those in vision, it is desirable that a representation scheme uncovers the parts-based, additive nature of the data. Of current representation learning schemes, restricted Boltzmann machines (RBMs) have proved to be highly eective in unsupervised settings. However, when it comes to parts-based discovery, RBMs do not usually produce satisfactory results. We enhance such capacity of RBMs by introducing nonnegativity into the model weights, resulting in a variant called nonnegative restricted Boltzmann machine (NRBM). The NRBM produces not only controllable decomposition of data into interpretable parts but also oers a way to estimate the intrinsic nonlinear dimensionality of data. We demonstrate the capacity of our model on well-known datasets of handwritten digits, faces and documents. The decomposition quality on images is comparable with or better than what produced by the nonnegative matrix factorisation (NMF), and the thematic features uncovered from text are qualitatively interpretable in a similar manner to that of the latent Dirichlet allocation (LDA). However, the learnt features, when used for classication,"
                },
                {
                    "title": "Vector Space Projections: A Numerical Approach to Signal and Image Processing, Neural Nets, and Optics",
                    "abstract": "From the Publisher: \nWritten by two leading authorities in the field, this self-contained volume provides a tutorial on projection methods and how to apply them in science and engineering. It details effective problem-solving strategies, and explores key applications in communication and signal processing, neural networks and pattern recognition, and optics and image processing. This book reviews the fundamentals of vector space theory, covers principles and applications of vector space projections in general, and projections onto convex sets in particular, provides real-world examples solvable on PCs and modest workstations, features more than 100 illustrations, and includes end-of-chapter exercises and references. This useful reference for practicing engineers, scientists, and educators can also be used for graduate-level study in science, mathematics, and engineering. Portions of the book have been used as material in short courses on applications of vector space projections."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively analyze the existence and characteristics of fixed points in nonnegative neural networks to enhance their interpretability and performance?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it can lead to a deeper understanding of nonnegative neural networks, which are increasingly relevant in various applications such as medical imaging and spectral analysis. By addressing the fixed point analysis, we can improve the interpretability of these networks, potentially leading to more transparent AI systems. This research could pave the way for future studies that explore the theoretical foundations of nonnegative networks, ultimately influencing the design of more efficient and effective neural architectures in both digital and analog implementations.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complex mathematical nature of fixed point theory and the nonlinearity inherent in neural networks. Naive approaches may fail because they do not account for the constraints imposed by nonnegativity, which can lead to multiple or no fixed points. Additionally, the interaction between the network architecture and the input data can create intricate landscapes that are difficult to analyze. Overcoming these technical obstacles requires advanced mathematical tools and a deep understanding of both neural network dynamics and fixed point theory.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on the performance of nonnegative neural networks without delving into the underlying fixed point structures. Limitations in mathematical frameworks and a lack of comprehensive methodologies for analyzing fixed points in the context of nonnegative constraints have hindered progress. Additionally, many studies have prioritized empirical results over theoretical insights, leaving a gap in understanding the fundamental properties of these networks. Our approach aims to bridge this gap by providing a systematic analysis of fixed points, which has not been thoroughly explored in prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a detailed mathematical analysis of nonnegative neural networks, focusing on the formulation of fixed point equations and their properties. We will utilize a dataset relevant to applications such as image processing and medical imaging to validate our findings. The metrics for evaluation will include the existence of fixed points, their stability, and the interpretability of the network outputs. We expect to demonstrate that a thorough understanding of fixed points can lead to improved network designs that maintain predictive power while enhancing interpretability, ultimately contributing to the advancement of nonnegative neural networks"
            }
        },
        "author_data": {
            "1498cddc-9c87-454f-9fc6-0cd7ec44aa30": {
                "pk": "1498cddc-9c87-454f-9fc6-0cd7ec44aa30",
                "name": "Tomasz J. Piotrowski",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "fe625f38-7bc0-4b51-96c8-8e1d87334dfb": {
                "pk": "fe625f38-7bc0-4b51-96c8-8e1d87334dfb",
                "name": "Renato L. G. Cavalcante",
                "collaborators": [
                    "Slawomir Stanczak",
                    "Rafail Ismayilov",
                    "S\u0142awomir Sta\u0144czak",
                    "Emmanuel Pollakis",
                    "Jochen Fink",
                    "Lorenzo Miretti",
                    "Tomasz Piotrowski",
                    "Daniel Sch\u00e4ufele",
                    "Yuxiang Shen",
                    "Slawomir Sta\u0144czak"
                ],
                "domain": [
                    "Wireless Networks",
                    "Fixed Point Algorithms",
                    "Optimization",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Spectral radii of asymptotic mappings and the convergence speed of the standard fixed point algorithm",
                        "abstract": "Important problems in wireless networks can often be solved by computing fixed points of standard or contractive interference mappings, and the conventional fixed point algorithm is widely used for this purpose. Knowing that the mapping used in the algorithm is not only standard but also contractive (or only contractive) is valuable information because we obtain a guarantee of geometric convergence rate, and the rate is related to a property of the mapping called modulus of contraction. To date, contractive mappings and their moduli of contraction have been identified with case-by-case approaches that can be difficult to generalize. To address this limitation of existing approaches, we show in this study that the spectral radii of asymptotic mappings can be used to identify an important subclass of contractive mappings and also to estimate their moduli of contraction. In addition, if the fixed point algorithm is applied to compute fixed points of positive concave mappings, we show that the spectral radii of asymptotic mappings provide us with simple lower bounds for the estimation error of the iterates. An immediate application of this result proves that a known algorithm for load estimation in wireless networks becomes slower with increasing traffic."
                    },
                    {
                        "title": "The fixed point iteration of positive concave mappings converges geometrically if a fixed point exists: implications to wireless systems",
                        "abstract": "We prove that the fixed point iteration of arbitrary positive concave mappings with nonempty fixed point set converges geometrically for any starting point. We also show that positivity is crucial for this result to hold, and the concept of (nonlinear) spectral radius of asymptotic mappings provides us with information about the convergence factor. As a practical implication of the results shown here, we rigorously explain why some power control and load estimation algorithms in wireless networks, which are particular instances of the fixed point iteration, have shown geometric convergence in simulations. These algorithms have been typically derived by considering fixed point iterations of the general class of standard interference mappings, so the possibility of sublinear convergence rate could not be ruled out in previous studies, except in special cases that are often more restrictive than those considered here."
                    },
                    {
                        "title": "Tensor Completion for Radio Map Reconstruction using Low Rank and Smoothness",
                        "abstract": "Radio maps are important enablers for many applications in wireless networks, ranging from network planning and optimization to fingerprint based localization. Sampling the complete map is prohibitively expensive in practice, so methods for reconstructing the complete map from a subset of measurements are increasingly gaining attention in the literature. In this paper, we propose two algorithms for this purpose, which build on existing approaches that aim at minimizing the tensor rank while additionally enforcing smoothness of the radio map. Experimental results with synthetic measurements derived via ray tracing show that our algorithms outperform state of the art techniques."
                    },
                    {
                        "title": "Power Estimation in LTE systems with the General Framework of Standard Interference Mappings",
                        "abstract": "We devise novel techniques to obtain the downlink power inducing a given load in long-term evolution (LTE) systems, where we define load as the fraction of resource blocks in the time-frequency grid being requested by users from a given base station. These techniques are particularly important because previous studies have proved that the data rate requirement of users can be satisfied with lower transmit energy if we allow the load to increase. Those studies have also shown that obtaining the power assignment from a desired load profile can be posed as a fixed point problem involving standard interference mappings, but so far the mappings have not been obtained explicitly. One of our main contributions in this study is to close this gap. We derive an interference mapping having as its fixed point the power assignment inducing a desired load, assuming that such an assignment exists. Having this mapping in closed form, we simplify the proof of the aforementioned known results, and we also devise novel iterative algorithms for power computation that have many numerical advantages over previous methods."
                    },
                    {
                        "title": "Elementary Properties of Positive Concave Mappings with Applications to Network Planning and Optimization",
                        "abstract": "This study presents novel methods for computing fixed points of positive concave mappings and for characterizing the existence of fixed points. These methods are particularly important in planning and optimization tasks in wireless networks. For example, previous studies have shown that the feasibility of a network design can be quickly evaluated by computing the fixed point of a concave mapping that is constructed based on many environmental and network control parameters such as the position of base stations, channel conditions, and antenna tilts. To address this and more general problems, given a positive concave mapping, we show two alternative but equivalent ways to construct a matrix that is guaranteed to have spectral radius strictly smaller than one if the mapping has a fixed point. This matrix is then used to build a new mapping that preserves the fixed point of the original positive concave mapping. We show that the standard fixed point iterations using the new mapping converges faster than the standard iterations applied to the original concave mapping. As exemplary applications of the proposed methods, we consider the problems of power and load planning in networks based on the orthogonal frequency division multiple access (OFDMA) technology."
                    },
                    {
                        "title": "Multi-Group Multicast Beamforming by Superiorized Projections onto Convex Sets",
                        "abstract": "In this paper, we propose an iterative algorithm to address the nonconvex multi-group multicast beamforming problem with quality-of-service constraints and per-antenna power constraints. We formulate a convex relaxation of the problem as a semidefinite program in a real Hilbert space, which allows us to approximate a point in the feasible set by iteratively applying a bounded perturbation resilient fixed-point mapping. Inspired by the superiorization methodology, we use this mapping as a basic algorithm, and we add in each iteration a small perturbation with the intent to reduce the objective value and the distance to nonconvex rank-constraint sets. We prove that the sequence of perturbations is bounded, so the algorithm is guaranteed to converge to a feasible point of the relaxed semidefinite program. Simulations show that the proposed approach outperforms existing algorithms in terms of both computation time and approximation gap in many cases."
                    },
                    {
                        "title": "FDD massive MIMO channel spatial covariance conversion using projection methods",
                        "abstract": "Knowledge of second-order statistics of channels (e.g. in the form of covariance matrices) is crucial for the acquisition of downlink channel state information (CSI) in massive MIMO systems operating in the frequency division duplexing (FDD) mode. Current MIMO systems usually obtain downlink covariance information via feedback of the estimated covariance matrix from the user equipment (UE), but in the massive MIMO regime this approach is infeasible because of the unacceptably high training overhead. This paper considers instead the problem of estimating the downlink channel covariance from uplink measurements. We propose two variants of an algorithm based on projection methods in an infinite-dimensional Hilbert space that exploit channel reciprocity properties in the angular domain. The proposed schemes are evaluated via Monte Carlo simulations, and they are shown to outperform current state-of-the art solutions in terms of accuracy and complexity, for typical array geometries and duplex gaps."
                    },
                    {
                        "title": "Downlink channel spatial covariance estimation in realistic FDD massive MIMO systems",
                        "abstract": "The knowledge of the downlink (DL) channel spatial covariance matrix at the BS is of fundamental importance for large-scale array systems operating in frequency division duplexing (FDD) mode. In particular, this knowledge plays a key role in the DL channel state information (CSI) acquisition. In the massive MIMO regime, traditional schemes based on DL pilots are severely limited by the covariance feedback and the DL training overhead. To overcome this problem, many authors have proposed to obtain an estimate of the DL spatial covariance based on uplink (UL) measurements. However, many of these approaches rely on simple channel models, and they are difficult to extend to more complex models that take into account important effects of propagation in 3D environments and of dual-polarized antenna arrays. In this study we propose a novel technique that takes into account the aforementioned effects, in compliance with the requirements of modern 4G and 5G system designs. Numerical simulations show the effectiveness of our approach."
                    },
                    {
                        "title": "Deep Learning Based Hybrid Precoding in Dual-Band Communication Systems",
                        "abstract": "We propose a deep learning-based method that uses spatial and temporal information extracted from the sub-6GHz band to predict/track beams in the millimeter-wave (mmWave) band. In more detail, we consider a dual-band communication system operating in both the sub-6GHz and mmWave bands. The objective is to maximize the achievable mutual information in the mmWave band with a hybrid analog/digital architecture where analog precoders (RF precoders) are taken from a finite codebook. Finding a RF precoder using conventional search methods incurs large signalling overhead, and the signalling scales with the number of RF chains and the resolution of the phase shifters. To overcome the issue of large signalling overhead in the mmWave band, the proposed method exploits the spatiotemporal correlation between sub-6GHz and mmWave bands, and it predicts/tracks the RF precoders in the mmWave band from sub-6GHz channel measurements. The proposed method provides a smaller candidate set so that performing a search over that set significantly reduces the signalling overhead compared with conventional search heuristics. Simulations show that the proposed method can provide reasonable achievable rates while significantly reducing the signalling overhead."
                    },
                    {
                        "title": "Deep Learning Beam Optimization in Millimeter-Wave Communication Systems",
                        "abstract": "We propose a method that combines fixed point algorithms with a neural network to optimize jointly discrete and continuous variables in millimeter-wave communication systems, so that the users' rates are allocated fairly in a well-defined sense. In more detail, the discrete variables include user-access point assignments and the beam configurations, while the continuous variables refer to the power allocation. The beam configuration is predicted from user-related information using a neural network. Given the predicted beam configuration, a fixed point algorithm allocates power and assigns users to access points so that the users achieve the maximum fraction of their interference-free rates. The proposed method predicts the beam configuration in a \"one-shot\" manner, which significantly reduces the complexity of the beam search procedure. Moreover, even if the predicted beam configurations are not optimal, the fixed point algorithm still provides the optimal power allocation and user-access point assignments for the given beam configuration."
                    },
                    {
                        "title": "Superiorized Adaptive Projected Subgradient Method with Application to MIMO Detection",
                        "abstract": "In this paper, we show that the adaptive projected subgradient method (APSM) is bounded perturbation resilient. To illustrate a potential application of this result, we propose a set-theoretic framework for MIMO detection, and we devise algorithms based on a superiorized APSM. Various low-complexity MIMO detection algorithms achieve excellent performance on i.i.d. Gaussian channels, but they typically incur high performance loss if realistic channel models (e.g., correlated channels) are considered. Compared to existing low-complexity iterative detectors such as individually optimal large-MIMO approximate message passing (IO-LAMA), the proposed algorithms can achieve considerably lower symbol error ratios over correlated channels. At the same time, the proposed methods do not require matrix inverses, and their complexity is similar to IO-LAMA."
                    },
                    {
                        "title": "Localization in Dynamic Indoor MIMO-OFDM Wireless Systems using Domain Adaptation",
                        "abstract": "We propose a method for predicting the location of user equipment (UE) using wireless fingerprints in dynamic indoor non-line-of-sight (NLoS) environments. In particular, our method copes with the challenges posed by the drift, birth, and death of scattering clusters resulting from dynamic changes in the wireless environment. Prominent examples of such dynamic wireless environments include factory floors or offices, where the geometry of the environment undergoes changes over time. These changes affect the distribution of wireless fingerprints, demonstrating some similarity between the distributions before and after the change. Consequently, the performance of a location estimator initially designed for a specific environment may degrade significantly when applied after changes have occurred in that environment. To address this limitation, we propose a domain adaptation framework that utilizes neural networks to align the distributions of wireless fingerprints collected both before and after environmental changes. By aligning these distributions, we design an estimator capable of predicting UE locations from their wireless fingerprints in the new environment. Experiments validate the effectiveness of the proposed methods in localizing UEs in dynamic wireless environments."
                    },
                    {
                        "title": "Traffic Demand-Aware Topology Control for Enhanced Energy-Efficiency of Cellular Networks",
                        "abstract": "The service provided by mobile networks operated today is not adapted to spatio-temporal fluctuations in traffic demand, although such fluctuations offer opportunities for energy savings. In particular, significant gains in energy efficiency are realizable by disengaging temporarily redundant hardware components of base stations. We therefore propose a novel optimization framework that considers both the load-dependent energy radiated by the antennas and the remaining forms of energy needed for operating the base stations. The objective is to reduce the energy consumption of mobile networks, while ensuring that the data rate requirements of the users are met throughout the coverage area. Building upon sparse optimization techniques, we develop a majorization-minimization algorithm with the ability to identify energy-efficient network configurations. The iterative algorithm is load-aware, has low computational complexity, and can be implemented in an online fashion to exploit load fluctuations on a short time scale. Simulations show that the algorithm can find network configurations with the energy consumption similar to that obtained with global optimization tools, which cannot be applied to real large networks. Although we consider only one currently deployed cellular technology, the optimization framework is general, potentially applicable to a large class of access technologies."
                    }
                ]
            },
            "b4c07759-6295-4121-941e-fb779aed3891": {
                "pk": "b4c07759-6295-4121-941e-fb779aed3891",
                "name": "Mateusz Gabor",
                "collaborators": [
                    "Rafal Zdunek",
                    "Agnieszka Wylomanska",
                    "Radoslaw Zimroz",
                    "Rafa\u0142 Zdunek",
                    "Tomasz Piotrowski",
                    "Renato L. G. Cavalcante"
                ],
                "domain": [
                    "Machine Learning",
                    "Deep Learning",
                    "Signal Processing",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Reduced storage direct tensor ring decomposition for convolutional neural networks compression",
                        "abstract": "Convolutional neural networks (CNNs) are among the most widely used machine learning models for computer vision tasks, such as image classification. To improve the efficiency of CNNs, many CNNs compressing approaches have been developed. Low-rank methods approximate the original convolutional kernel with a sequence of smaller convolutional kernels, which leads to reduced storage and time complexities. In this study, we propose a novel low-rank CNNs compression method that is based on reduced storage direct tensor ring decomposition (RSDTR). The proposed method offers a higher circular mode permutation flexibility, and it is characterized by large parameter and FLOPS compression rates, while preserving a good classification accuracy of the compressed network. The experiments, performed on the CIFAR-10 and ImageNet datasets, clearly demonstrate the efficiency of RSDTR in comparison to other state-of-the-art CNNs compression approaches."
                    },
                    {
                        "title": "Positive concave deep equilibrium models",
                        "abstract": "Deep equilibrium (DEQ) models are widely recognized as a memory efficient alternative to standard neural networks, achieving state-of-the-art performance in language modeling and computer vision tasks. These models solve a fixed point equation instead of explicitly computing the output, which sets them apart from standard neural networks. However, existing DEQ models often lack formal guarantees of the existence and uniqueness of the fixed point, and the convergence of the numerical scheme used for computing the fixed point is not formally established. As a result, DEQ models are potentially unstable in practice. To address these drawbacks, we introduce a novel class of DEQ models called positive concave deep equilibrium (pcDEQ) models. Our approach, which is based on nonlinear Perron-Frobenius theory, enforces nonnegative weights and activation functions that are concave on the positive orthant. By imposing these constraints, we can easily ensure the existence and uniqueness of the fixed point without relying on additional complex assumptions commonly found in the DEQ literature, such as those based on monotone operator theory in convex analysis. Furthermore, the fixed point can be computed with the standard fixed point algorithm, and we provide theoretical guarantees of its geometric convergence, which, in particular, simplifies the training process. Experiments demonstrate the competitiveness of our pcDEQ models against other implicit models."
                    },
                    {
                        "title": "Bearing damage detection with orthogonal and non-negative low-rank feature extraction",
                        "abstract": "Local damage of bearings can be detected as a weak cyclic and impulsive component in a highly noisy measured signal. A key problem is how to extract the signal of interest (SOI) from the raw signal, i.e., how to identify and design an optimal filter. To tackle this problem, we propose to use stochastic sampled orthogonal non-negative matrix factorization for extracting frequency-based features from a spectrogram of the measured signal. The proposed algorithm finds a selective filter that is tailored to the frequency band of the SOI. We show that our approach outperforms the other state-of-the-art selectors that were previously used in condition monitoring. The efficiency of the proposed method is illustrated using both a simulation study and the following real signals: (a) vibration signal from a test rig in the laboratory and (b) acoustic signal from a belt conveyor."
                    },
                    {
                        "title": "Non-negative matrix underapproximation as optimal frequency band selector",
                        "abstract": "Time-frequency representation (TFR) is often used for non-stationary signal analysis. The most intuitive and interpretable TFR is the spectrogram. Recently, a concept of non-negative matrix factorization (NMF) has been successfully applied to local damage detection in rolling elements of bearings via spectrogram factorization. NMF applied to the spectrogram allows one to find an informative frequency band, which could be further used as a filter characteristic. However, the obtained filter characteristics mostly detect the informative frequency band, which also encompasses a lot of noise. In the case where noise is more problematic, as is the case for acoustic signals from industrial machines, the NMF hardly detects the damage. To solve this problem and obtain more selective filters, which are more robust to noise, we propose the non-negative matrix under-approximation (NMU) as an informative frequency band selector. Due to the more sparse parts-based representation of the NMU compared to NMF, NMU provides more selective filter characteristics, which neglect the non-informative frequency bands related to the noise. In practice, it means that NMU gives a better signal-to-noise ratio for the filtered signal. The efficiency of the proposed approach has been validated on the vibration signal from the test rig and the acoustic signal from an idler."
                    }
                ]
            }
        }
    },
    "2312.04727": {
        "paper_data": {
            "title": "E2ENet: Dynamic Sparse Feature Fusion for Accurate and Efficient 3D Medical Image Segmentation",
            "url": "http://arxiv.org/abs/2312.04727v1",
            "arxiv_id": "2312.04727",
            "authors": [
                "Boqian Wu",
                "Qiao Xiao",
                "Shiwei Liu",
                "Lu Yin",
                "Mykola Pechenizkiy",
                "Decebal Constantin Mocanu",
                "Maurice Van Keulen",
                "Elena Mocanu"
            ],
            "abstract": "Deep neural networks have evolved as the leading approach in 3D medical image segmentation due to their outstanding performance. However, the ever-increasing model size and computation cost of deep neural networks have become the primary barrier to deploying them on real-world resource-limited hardware. In pursuit of improving performance and efficiency, we propose a 3D medical image segmentation model, named Efficient to Efficient Network (E2ENet), incorporating two parametrically and computationally efficient designs. i. Dynamic sparse feature fusion (DSFF) mechanism: it adaptively learns to fuse informative multi-scale features while reducing redundancy. ii. Restricted depth-shift in 3D convolution: it leverages the 3D spatial information while keeping the model and computational complexity as 2D-based methods. We conduct extensive experiments on BTCV, AMOS-CT and Brain Tumor Segmentation Challenge, demonstrating that E2ENet consistently achieves a superior trade-off between accuracy and efficiency than prior arts across various resource constraints. E2ENet achieves comparable accuracy on the large-scale challenge AMOS-CT, while saving over 68\\% parameter count and 29\\% FLOPs in the inference phase, compared with the previous best-performing method. Our code has been made available at: https://github.com/boqian333/E2ENet-Medical.",
            "introduction": "   1 Introduction  Figure 1: A comparison of feature fusion schemes. The purple nodes depict features extracted from the backbone, while the green nodes depict the fused features. In particular, in DiNTS (d), red lines indicate information flow paths determined through neural architecture search techniques. In E2ENet (e), the red lines with different widths represent sparse information flows determined by the DSFF mechanism, allowing for efficient feature fusion. E2ENet is capable of dynamically learning how much of the features to fuse is derived from the backbone.   3D medical image segmentation, plays an essential role in numerous clinical applications, including computer-aided diagnosis (Yu et\u00a0al., 2020b) and image-guided surgery systems (Ronneberger et\u00a0al., 2015), etc. Over the past decade, the rapid development of deep neural networks has achieved tremendous breakthroughs and significantly boosted the progress in this area (Zhou et\u00a0al., 2018; Huang et\u00a0al., 2020; Isensee et\u00a0al., 2021). However, the model sizes and computational costs are also exploding, which deter their deployment in many real-world applications, especially for 3D models where resource consumption scales cubically (Hu et\u00a0al., 2021; Valanarasu & Patel, 2022). This naturally raises a research question: Can we design a 3D medical image segmentation method that trades off accuracy and efficiency better subjected to different resource availability?    Accurately segmenting organs is a challenging task in 3D medical image segmentation due to the variability in size and shape even among the same type of organ, caused by factors such as patient anatomy and disease stage. This diversity amplifies the difficulty to accurately identify and distinguish the boundaries of different organs, leading to potential errors in segmentation. One of the main solutions for accurate medical image segmentation is to favorably leverage the multi-scale features extracted by the backbone network, but this remains a long-standing problem. Pioneering work UNet (Ronneberger et\u00a0al., 2015) utilizes skip connection to propagate detailed information to high-level features. More recently, UNet++ (Zhou et\u00a0al., 2018), CoTr (Xie et\u00a0al., 2021) and DiNTS (He et\u00a0al., 2021) have developed more complex neural networks architectures (e.g. densely skip connections and attention mechanism) or optimization techniques (e.g. neural architecture search (NAS) (Elsken et\u00a0al., 2019)) for cross-scale feature fusion. NAS initially searches for the network topology and operator (e.g., 3x3 Convolution and Maxpool) and subsequently optimize the model weights. However, NAS typically demands a significant amount of computational resources for exploring the network topology and assessing numerous candidate architectures, making it computationally expensive and time-consuming. For instance, C2FNAS (Yu et\u00a0al., 2020a) necessitates nearly one GPU year to discover a single 3D segmentation architecture, DiNTS (He et\u00a0al., 2021) improves the searching efficiency but still requires 5.8 GPU days to search for a single architecture. Compared to NAS approaches, our method does not require costly architecture search time and directly optimizes/searches sparse topology within the pre-defined architecture.   In this paper, we propose Efficient to Efficient Network (E2ENet), a model that can efficiently incorporate both bottom-up and top-down information flows from the backbone network in a dynamic sparse pattern, achieving a much better accuracy-efficiency trade-off. As shown in Figure 1 (e), E2ENet incorporates multi-scale features from the backbone into the final output by gradually fusing the adjacent features, allowing the",
            "references": [
                {
                    "title": "Dynamic Sparse Network for Time Series Classification: Learning What to \"see\"",
                    "abstract": "The receptive field (RF), which determines the region of time series to be ``seen'' and used, is critical to improve the performance for time series classification (TSC). However, the variation of signal scales across and within time series data, makes it challenging to decide on proper RF sizes for TSC. In this paper, we propose a dynamic sparse network (DSN) with sparse connections for TSC, which can learn to cover various RF without cumbersome hyper-parameters tuning. The kernels in each sparse layer are sparse and can be explored under the constraint regions by dynamic sparse training, which makes it possible to reduce the resource cost. The experimental results show that the proposed DSN model can achieve state-of-art performance on both univariate and multivariate TSC datasets with less than 50\\% computational cost compared with recent baseline methods, opening the path towards more accurate resource-aware methods for time series analyses. Our code is publicly available at: https://github.com/QiaoXiao7282/DSN."
                },
                {
                    "title": "Where to Pay Attention in Sparse Training for Feature Selection?",
                    "abstract": "A new line of research for feature selection based on neural networks has recently emerged. Despite its superiority to classical methods, it requires many training iterations to converge and detect informative features. The computational time becomes prohibitively long for datasets with a large number of samples or a very high dimensional feature space. In this paper, we present a new efficient unsupervised method for feature selection based on sparse autoencoders. In particular, we propose a new sparse training algorithm that optimizes a model's sparse topology during training to pay attention to informative features quickly. The attention-based adaptation of the sparse topology enables fast detection of informative features after a few training iterations. We performed extensive experiments on 10 datasets of different types, including image, speech, text, artificial, and biological. They cover a wide range of characteristics, such as low and high-dimensional feature spaces, and few and large training samples. Our proposed approach outperforms the state-of-the-art methods in terms of selecting informative features while reducing training iterations and computational costs substantially. Moreover, the experiments show the robustness of our method in extremely noisy environments."
                },
                {
                    "title": "Extending nnU-Net is all you need",
                    "abstract": "Semantic segmentation is one of the most popular research areas in medical image computing. Perhaps surprisingly, despite its conceptualization dating back to 2018, nnU-Net continues to provide competitive out-of-the-box solutions for a broad variety of segmentation problems and is regularly used as a development framework for challenge-winning algorithms. Here we use nnU-Net to participate in the AMOS2022 challenge, which comes with a unique set of tasks: not only is the dataset one of the largest ever created and boasts 15 target structures, but the competition also requires submitted solutions to handle both MRI and CT scans. Through careful modification of nnU-net's hyperparameters, the addition of residual connections in the encoder and the design of a custom postprocessing strategy, we were able to substantially improve upon the nnU-Net baseline. Our final ensemble achieves Dice scores of 90.13 for Task 1 (CT) and 89.06 for Task 2 (CT+MRI) in a 5-fold cross-validation on the provided training cases."
                },
                {
                    "title": "More ConvNets in the 2020s: Scaling up Kernels Beyond 51x51 using Sparsity",
                    "abstract": "Transformers have quickly shined in the computer vision world since the emergence of Vision Transformers (ViTs). The dominant role of convolutional neural networks (CNNs) seems to be challenged by increasingly effective transformer-based models. Very recently, a couple of advanced convolutional models strike back with large kernels motivated by the local-window attention mechanism, showing appealing performance and efficiency. While one of them, i.e. RepLKNet, impressively manages to scale the kernel size to 31x31 with improved performance, the performance starts to saturate as the kernel size continues growing, compared to the scaling trend of advanced ViTs such as Swin Transformer. In this paper, we explore the possibility of training extreme convolutions larger than 31x31 and test whether the performance gap can be eliminated by strategically enlarging convolutions. This study ends up with a recipe for applying extremely large kernels from the perspective of sparsity, which can smoothly scale up kernels to 61x61 with better performance. Built on this recipe, we propose Sparse Large Kernel Network (SLaK), a pure CNN architecture equipped with sparse factorized 51x51 kernels that can perform on par with or better than state-of-the-art hierarchical Transformers and modern ConvNet architectures like ConvNeXt and RepLKNet, on ImageNet classification as well as a wide range of downstream tasks including semantic segmentation on ADE20K, object detection on PASCAL VOC 2007, and object detection/segmentation on MS COCO."
                },
                {
                    "title": "UNeXt: MLP-based Rapid Medical Image Segmentation Network",
                    "abstract": "UNet and its latest extensions like TransUNet have been the leading medical image segmentation methods in recent years. However, these networks cannot be effectively adopted for rapid image segmentation in point-of-care applications as they are parameter-heavy, computationally complex and slow to use. To this end, we propose UNeXt which is a Convolutional multilayer perceptron (MLP) based network for image segmentation. We design UNeXt in an effective way with an early convolutional stage and a MLP stage in the latent stage. We propose a tokenized MLP block where we efficiently tokenize and project the convolutional features and use MLPs to model the representation. To further boost the performance, we propose shifting the channels of the inputs while feeding in to MLPs so as to focus on learning local dependencies. Using tokenized MLPs in latent space reduces the number of parameters and computational complexity while being able to result in a better representation to help segmentation. The network also consists of skip connections between various levels of encoder and decoder. We test UNeXt on multiple medical image segmentation datasets and show that we reduce the number of parameters by 72x, decrease the computational complexity by 68x, and improve the inference speed by 10x while also obtaining better segmentation performance over the state-of-the-art medical image segmentation architectures. Code is available at https://github.com/jeya-maria-jose/UNeXt-pytorch"
                },
                {
                    "title": "A Data-scalable Transformer for Medical Image Segmentation: Architecture, Model Efficiency, and Benchmark",
                    "abstract": "Transformers have demonstrated remarkable performance in natural language processing and computer vision. However, existing vision Transformers struggle to learn from limited medical data and are unable to generalize on diverse medical image tasks. To tackle these challenges, we present MedFormer, a data-scalable Transformer designed for generalizable 3D medical image segmentation. Our approach incorporates three key elements: a desirable inductive bias, hierarchical modeling with linear-complexity attention, and multi-scale feature fusion that integrates spatial and semantic information globally. MedFormer can learn across tiny- to large-scale data without pre-training. Comprehensive experiments demonstrate MedFormer's potential as a versatile segmentation backbone, outperforming CNNs and vision Transformers on seven public datasets covering multiple modalities (e.g., CT and MRI) and various medical targets (e.g., healthy organs, diseased tissues, and tumors). We provide public access to our models and evaluation pipeline, offering solid baselines and unbiased comparisons to advance a wide range of downstream clinical applications."
                },
                {
                    "title": "Self-Supervised Pre-Training of Swin Transformers for 3D Medical Image Analysis",
                    "abstract": "Vision Transformers (ViT)s have shown great performance in self-supervised learning of global and local representations that can be transferred to downstream applications. Inspired by these results, we introduce a novel self-supervised learning framework with tailored proxy tasks for medical image analysis. Specifically, we propose: (i) a new 3D transformer-based model, dubbed Swin UNEt TRansformers (Swin UNETR), with a hierarchical encoder for self-supervised pretraining; (ii) tailored proxy tasks for learning the underlying pattern of human anatomy. We demonstrate successful pre-training of the proposed model on 5,050 publicly available computed tomography (CT) images from various body organs. The effectiveness of our approach is validated by fine-tuning the pre-trained models on the Beyond the Cranial Vault (BTCV) Segmentation Challenge with 13 abdominal organs and segmentation tasks from the Medical Segmentation Decathlon (MSD) dataset. Our model is currently the state-of-the-art on the public test leaderboards of both MSD11https://decathlon-10.grand-challenge.org/evaluation/challenge/leaderboard/ and BTCV 22https://www.synapse.org/#!Synapse:syn3193805/wiki/217785/ datasets. Code: https://monai.io/research/swin-unetr."
                },
                {
                    "title": "MEST: Accurate and Fast Memory-Economic Sparse Training Framework on the Edge",
                    "abstract": "Recently, a new trend of exploring sparsity for accelerating neural network training has emerged, embracing the paradigm of training on the edge. This paper proposes a novel Memory-Economic Sparse Training (MEST) framework targeting for accurate and fast execution on edge devices. The proposed MEST framework consists of enhancements by Elastic Mutation (EM) and Soft Memory Bound (&S) that ensure superior accuracy at high sparsity ratios. Different from the existing works for sparse training, this current work reveals the importance of sparsity schemes on the performance of sparse training in terms of accuracy as well as training speed on real edge devices. On top of that, the paper proposes to employ data efficiency for further acceleration of sparse training. Our results suggest that unforgettable examples can be identified in-situ even during the dynamic exploration of sparsity masks in the sparse training process, and therefore can be removed for further training speedup on edge devices. Comparing with state-of-the-art (SOTA) works on accuracy, our MEST increases Top-1 accuracy significantly on ImageNet when using the same unstructured sparsity scheme. Systematical evaluation on accuracy, training speed, and memory footprint are conducted, where the proposed MEST framework consistently outperforms representative SOTA works. A reviewer strongly against our work based on his false assumptions and misunderstandings. On top of the previous submission, we employ data efficiency for further acceleration of sparse training. And we explore the impact of model sparsity, sparsity schemes, and sparse training algorithms on the number of removable training examples. Our codes are publicly available at: https://github.com/boone891214/MEST."
                },
                {
                    "title": "UCTransNet: Rethinking the Skip Connections in U-Net from a Channel-wise Perspective with Transformer",
                    "abstract": "Most recent semantic segmentation methods adopt a U-Net framework with an encoder-decoder architecture. It is still challenging for U-Net with a simple skip connection scheme to model the global multi-scale context: 1) Not each skip connection setting is effective due to the issue of incompatible feature sets of encoder and decoder stage, even some skip connection negatively influence the segmentation performance; 2) The original U-Net is worse than the one without any skip connection on some datasets. Based on our findings, we propose a new segmentation framework, named UCTransNet (with a proposed CTrans module in U-Net), from the channel perspective with attention mechanism. Specifically, the CTrans (Channel Transformer) module is an alternate of the U-Net skip connections, which consists of a sub-module to conduct the multi-scale Channel Cross fusion with Transformer (named CCT) and a sub-module Channel-wise Cross-Attention (named CCA) to guide the fused multi-scale channel-wise information to effectively connect to the decoder features for eliminating the ambiguity. Hence, the proposed connection consisting of the CCT and CCA is able to replace the original skip connection to solve the semantic gaps for an accurate automatic medical image segmentation. The experimental results suggest that our UCTransNet produces more precise segmentation performance and achieves consistent improvements over the state-of-the-art for semantic segmentation across different datasets and conventional architectures involving transformer or U-shaped framework. Code: https://github.com/McGregorWwww/UCTransNet."
                },
                {
                    "title": "AS-MLP: An Axial Shifted MLP Architecture for Vision",
                    "abstract": "An Axial Shifted MLP architecture (AS-MLP) is proposed in this paper. Different from MLP-Mixer, where the global spatial feature is encoded for information flow through matrix transposition and one token-mixing MLP, we pay more attention to the local features interaction. By axially shifting channels of the feature map, AS-MLP is able to obtain the information flow from different axial directions, which captures the local dependencies. Such an operation enables us to utilize a pure MLP architecture to achieve the same local receptive field as CNN-like architecture. We can also design the receptive field size and dilation of blocks of AS-MLP, etc, in the same spirit of convolutional neural networks. With the proposed AS-MLP architecture, our model obtains 83.3% Top-1 accuracy with 88M parameters and 15.2 GFLOPs on the ImageNet-1K dataset. Such a simple yet effective architecture outperforms all MLP-based architectures and achieves competitive performance compared to the transformer-based architectures (e.g., Swin Transformer) even with slightly lower FLOPs. In addition, AS-MLP is also the first MLP-based architecture to be applied to the downstream tasks (e.g., object detection and semantic segmentation). The experimental results are also impressive. Our proposed AS-MLP obtains 51.5 mAP on the COCO validation set and 49.5 MS mIoU on the ADE20K dataset, which is competitive compared to the transformer-based architectures. Our AS-MLP establishes a strong baseline of MLP-based architecture. Code is available at https://github.com/svip-lab/AS-MLP."
                },
                {
                    "title": "Sparse Training via Boosting Pruning Plasticity with Neuroregeneration",
                    "abstract": "Works on lottery ticket hypothesis (LTH) and single-shot network pruning (SNIP) have raised a lot of attention currently on post-training pruning (iterative magnitude pruning), and before-training pruning (pruning at initialization). The former method suffers from an extremely large computation cost and the latter usually struggles with insufficient performance. In comparison, during-training pruning, a class of pruning methods that simultaneously enjoys the training/inference efficiency and the comparable performance, temporarily, has been less explored. To better understand during-training pruning, we quantitatively study the effect of pruning throughout training from the perspective of pruning plasticity (the ability of the pruned networks to recover the original performance). Pruning plasticity can help explain several other empirical observations about neural network pruning in literature. We further find that pruning plasticity can be substantially improved by injecting a brain-inspired mechanism called neuroregeneration, i.e., to regenerate the same number of connections as pruned. We design a novel gradual magnitude pruning (GMP) method, named gradual pruning with zero-cost neuroregeneration (\\textbf{GraNet}), that advances state of the art. Perhaps most impressively, its sparse-to-sparse version for the first time boosts the sparse-to-sparse training performance over various dense-to-sparse methods with ResNet-50 on ImageNet without extending the training time. We release all codes in https://github.com/Shiweiliuiiiiiii/GraNet."
                },
                {
                    "title": "CoAtNet: Marrying Convolution and Attention for All Data Sizes",
                    "abstract": "Transformers have attracted increasing interests in computer vision, but they still fall behind state-of-the-art convolutional networks. In this work, we show that while Transformers tend to have larger model capacity, their generalization can be worse than convolutional networks due to the lack of the right inductive bias. To effectively combine the strengths from both architectures, we present CoAtNets(pronounced\"coat\"nets), a family of hybrid models built from two key insights: (1) depthwise Convolution and self-Attention can be naturally unified via simple relative attention; (2) vertically stacking convolution layers and attention layers in a principled way is surprisingly effective in improving generalization, capacity and efficiency. Experiments show that our CoAtNets achieve state-of-the-art performance under different resource constraints across various datasets: Without extra data, CoAtNet achieves 86.0% ImageNet top-1 accuracy; When pre-trained with 13M images from ImageNet-21K, our CoAtNet achieves 88.56% top-1 accuracy, matching ViT-huge pre-trained with 300M images from JFT-300M while using 23x less data; Notably, when we further scale up CoAtNet with JFT-3B, it achieves 90.88% top-1 accuracy on ImageNet, establishing a new state-of-the-art result."
                },
                {
                    "title": "Top-KAST: Top-K Always Sparse Training",
                    "abstract": "Sparse neural networks are becoming increasingly important as the field seeks to improve the performance of existing models by scaling them up, while simultaneously trying to reduce power consumption and computational footprint. Unfortunately, most existing methods for inducing performant sparse models still entail the instantiation of dense parameters, or dense gradients in the backward-pass, during training. For very large models this requirement can be prohibitive. In this work we propose Top-KAST, a method that preserves constant sparsity throughout training (in both the forward and backward-passes). We demonstrate the efficacy of our approach by showing that it performs comparably to or better than previous works when training models on the established ImageNet benchmark, whilst fully maintaining sparsity. In addition to our ImageNet results, we also demonstrate our approach in the domain of language modeling where the current best performing architectures tend to have tens of billions of parameters and scaling up does not yet seem to have saturated performance. Sparse versions of these architectures can be run with significantly fewer resources, making them more widely accessible and applicable. Furthermore, in addition to being effective, our approach is straightforward and can easily be implemented in a wide range of existing machine learning frameworks with only a few additional lines of code. We therefore hope that our contribution will help enable the broader community to explore the potential held by massive models, without incurring massive computational cost."
                },
                {
                    "title": "Pseudo 3D Auto-Correlation Network for Real Image Denoising",
                    "abstract": "The extraction of auto-correlation in images has shown great potential in deep learning networks, such as the self-attention mechanism in the channel domain and the self-similarity mechanism in the spatial domain. However, the realization of the above mechanisms mostly requires complicated module stacking and a large number of convolution calculations, which inevitably increases model complexity and memory cost. Therefore, we propose a pseudo 3D auto-correlation network (P3AN) to explore a more efficient way of capturing contextual information in image de-noising. On the one hand, P3AN uses fast 1D convolution instead of dense connections to realize criss-cross interaction, which requires less computational resources. On the other hand, the operation does not change the feature size and makes it easy to expand. It means that only a simple adaptive fusion is needed to obtain contextual information that includes both the channel domain and the spatial domain. Our method built a pseudo 3D auto-correlation attention block through 1D convolutions and a lightweight 2D structure for more discriminative features. Extensive experiments have been conducted on three synthetic and four real noisy datasets. According to quantitative metrics and visual quality evaluation, the P3AN shows great superiority and surpasses state-of-the-art image denoising methods."
                },
                {
                    "title": "ConViT: improving vision transformers with soft convolutional inductive biases",
                    "abstract": "Convolutional architectures have proven to be extremely successful for vision tasks. Their hard inductive biases enable sample-efficient learning, but come at the cost of a potentially lower performance ceiling. Vision transformers rely on more flexible self-attention layers, and have recently outperformed CNNs for image classification. However, they require costly pre-training on large external datasets or distillation from pre-trained convolutional networks. In this paper, we ask the following question: is it possible to combine the strengths of these two architectures while avoiding their respective limitations? To this end, we introduce gated positional self-attention (GPSA), a form of positional self-attention which can be equipped with a \u2018soft\u2019 convolutional inductive bias. We initialize the GPSA layers to mimic the locality of convolutional layers, then give each attention head the freedom to escape locality by adjusting a gating parameter regulating the attention paid to position versus content information. The resulting convolutional-like ViT architecture, ConViT, outperforms the DeiT (Touvron et al 2020 arXiv:2012.12877) on ImageNet, while offering a much improved sample efficiency. We further investigate the role of locality in learning by first quantifying how it is encouraged in vanilla self-attention layers, then analyzing how it has escaped in GPSA layers. We conclude by presenting various ablations to better understand the success of the ConViT. Our code and models are released publicly at https://github.com/facebookresearch/convit."
                },
                {
                    "title": "UNETR: Transformers for 3D Medical Image Segmentation",
                    "abstract": "Fully Convolutional Neural Networks (FCNNs) with contracting and expanding paths have shown prominence for the majority of medical image segmentation applications since the past decade. In FCNNs, the encoder plays an integral role by learning both global and local features and contextual representations which can be utilized for semantic output prediction by the decoder. Despite their success, the locality of convolutional layers in FCNNs, limits the capability of learning long-range spatial dependencies. Inspired by the recent success of transformers for Natural Language Processing (NLP) in long-range sequence learning, we reformulate the task of volumetric (3D) medical image segmentation as a sequence-to-sequence prediction problem. We introduce a novel architecture, dubbed as UNEt TRansformers (UNETR), that utilizes a transformer as the encoder to learn sequence representations of the input volume and effectively capture the global multi-scale information, while also following the successful \"U-shaped\" network design for the encoder and decoder. The transformer encoder is directly connected to a decoder via skip connections at different resolutions to compute the final semantic segmentation output. We have validated the performance of our method on the Multi Atlas Labeling Beyond The Cranial Vault (BTCV) dataset for multi-organ segmentation and the Medical Segmentation Decathlon (MSD) dataset for brain tumor and spleen segmentation tasks. Our benchmarks demonstrate new state-of-the-art performance on the BTCV leaderboard."
                },
                {
                    "title": "Do We Actually Need Dense Over-Parameterization? In-Time Over-Parameterization in Sparse Training",
                    "abstract": "In this paper, we introduce a new perspective on training deep neural networks capable of state-of-the-art performance without the need for the expensive over-parameterization by proposing the concept of In-Time Over-Parameterization (ITOP) in sparse training. By starting from a random sparse network and continuously exploring sparse connectivities during training, we can perform an Over-Parameterization in the space-time manifold, closing the gap in the expressibility between sparse training and dense training. We further use ITOP to understand the underlying mechanism of Dynamic Sparse Training (DST) and indicate that the benefits of DST come from its ability to consider across time all possible parameters when searching for the optimal sparse connectivity. As long as there are sufficient parameters that have been reliably explored during training, DST can outperform the dense neural network by a large margin. We present a series of experiments to support our conjecture and achieve the state-of-the-art sparse training performance with ResNet-50 on ImageNet. More impressively, our method achieves dominant performance over the overparameterization-based sparse methods at extreme sparsity levels. When trained on CIFAR-100, our method can match the performance of the dense model even at an extreme sparsity (98%). Code can be found this https URL."
                },
                {
                    "title": "Sparsity in Deep Learning: Pruning and growth for efficient inference and training in neural networks",
                    "abstract": "The growing energy and performance costs of deep learning have driven the community to reduce the size of neural networks by selectively pruning components. Similarly to their biological counterparts, sparse networks generalize just as well, if not better than, the original dense networks. Sparsity can reduce the memory footprint of regular networks to fit mobile devices, as well as shorten training time for ever growing networks. In this paper, we survey prior work on sparsity in deep learning and provide an extensive tutorial of sparsification for both inference and training. We describe approaches to remove and add elements of neural networks, different training strategies to achieve model sparsity, and mechanisms to exploit sparsity in practice. Our work distills ideas from more than 300 research papers and provides guidance to practitioners who wish to utilize sparsity today, as well as to researchers whose goal is to push the frontier forward. We include the necessary background on mathematical methods in sparsification, describe phenomena such as early structure adaptation, the intricate relations between sparsity and the training process, and show techniques for achieving acceleration on real hardware. We also define a metric of pruned parameter efficiency that could serve as a baseline for comparison of different sparse networks. We close by speculating on how sparsity can improve future workloads and outline major open problems in the field."
                },
                {
                    "title": "Attention Unet++: A Nested Attention-Aware U-Net for Liver CT Image Segmentation",
                    "abstract": "Liver cancer is one of the cancers with the highest mortality. In order to help doctors diagnose and treat liver lesion, an automatic liver segmentation model is urgently needed due to manually segmentation is time-consuming and error-prone. In this paper, we propose a nested attention-aware segmentation network, named Attention UNet++. Our proposed method has a deep supervised encoder-decoder architecture and a redesigned dense skip connection. Attention UNet++ introduces attention mechanism between nested convolutional blocks so that the features extracted at different levels can be merged with a task-related selection. Besides, due to the introduction of deep supervision, the prediction speed of the pruned network is accelerated at the cost of modest performance degradation. We evaluated proposed model on MICCAI 2017 Liver Tumor Segmentation (LiTS) Challenge Dataset. Attention UNet++ achieved very competitive performance for liver segmentation."
                },
                {
                    "title": "Pruning neural networks without any data by iteratively conserving synaptic flow",
                    "abstract": "Pruning the parameters of deep neural networks has generated intense interest due to potential savings in time, memory and energy both during training and at test time. Recent works have identified, through an expensive sequence of training and pruning cycles, the existence of winning lottery tickets or sparse trainable subnetworks at initialization. This raises a foundational question: can we identify highly sparse trainable subnetworks at initialization, without ever training, or indeed without ever looking at the data? We provide an affirmative answer to this question through theory driven algorithm design. We first mathematically formulate and experimentally verify a conservation law that explains why existing gradient-based pruning algorithms at initialization suffer from layer-collapse, the premature pruning of an entire layer rendering a network untrainable. This theory also elucidates how layer-collapse can be entirely avoided, motivating a novel pruning algorithm Iterative Synaptic Flow Pruning (SynFlow). This algorithm can be interpreted as preserving the total flow of synaptic strengths through the network at initialization subject to a sparsity constraint. Notably, this algorithm makes no reference to the training data and consistently competes with or outperforms existing state-of-the-art pruning algorithms at initialization over a range of models (VGG and ResNet), datasets (CIFAR-10/100 and Tiny ImageNet), and sparsity constraints (up to 99.99 percent). Thus our data-agnostic pruning algorithm challenges the existing paradigm that, at initialization, data must be used to quantify which synapses are important."
                },
                {
                    "title": "UNet 3+: A Full-Scale Connected UNet for Medical Image Segmentation",
                    "abstract": "Recently, a growing interest has been seen in deep learning-based semantic segmentation. UNet, which is one of deep learning networks with an encoder-decoder architecture, is widely used in medical image segmentation. Combining multi-scale features is one of important factors for accurate segmentation. UNet++ was developed as a modified Unet by designing an architecture with nested and dense skip connections. However, it does not explore sufficient information from full scales and there is still a large room for improvement. In this paper, we propose a novel UNet 3+, which takes advantage of full-scale skip connections and deep supervisions. The full-scale skip connections incorporate low-level details with high-level semantics from feature maps in different scales; while the deep supervision learns hierarchical representations from the full-scale aggregated feature maps. The proposed method is especially benefiting for organs that appear at varying scales. In addition to accuracy improvements, the proposed UNet 3+ can reduce the network parameters to improve the computation efficiency. We further propose a hybrid loss function and devise a classification-guided module to enhance the organ boundary and reduce the over-segmentation in a non-organ image, yielding more accurate segmentation results. The effectiveness of the proposed method is demonstrated on two datasets. The code is available at: github.com/ZJUGiveLab/UNet-Version"
                },
                {
                    "title": "Picking Winning Tickets Before Training by Preserving Gradient Flow",
                    "abstract": "Overparameterization has been shown to benefit both the optimization and generalization of neural networks, but large networks are resource hungry at both training and test time. Network pruning can reduce test-time resource requirements, but is typically applied to trained networks and therefore cannot avoid the expensive training process. We aim to prune networks at initialization, thereby saving resources at training time as well. Specifically, we argue that efficient training requires preserving the gradient flow through the network. This leads to a simple but effective pruning criterion we term Gradient Signal Preservation (GraSP). We empirically investigate the effectiveness of the proposed method with extensive experiments on CIFAR-10, CIFAR-100, Tiny-ImageNet and ImageNet, using VGGNet and ResNet architectures. Our method can prune 80% of the weights of a VGG-16 network on ImageNet at initialization, with only a 1.6% drop in top-1 accuracy. Moreover, our method achieves significantly better performance than the baseline at extreme sparsity levels."
                },
                {
                    "title": "C2FNAS: Coarse-to-Fine Neural Architecture Search for 3D Medical Image Segmentation",
                    "abstract": "3D convolution neural networks (CNN) have been proved very successful in parsing organs or tumours in 3D medical images, but it remains sophisticated and time-consuming to choose or design proper 3D networks given different task contexts. Recently, Neural Architecture Search (NAS) is proposed to solve this problem by searching for the best network architecture automatically. However, the inconsistency between search stage and deployment stage often exists in NAS algorithms due to memory constraints and large search space, which could become more serious when applying NAS to some memory and time-consuming tasks, such as 3D medical image segmentation. In this paper, we propose a coarse-to-fine neural architecture search (C2FNAS) to automatically search a 3D segmentation network from scratch without inconsistency on network size or input size. Specifically, we divide the search procedure into two stages: 1) the coarse stage, where we search the macro-level topology of the network, i.e. how each convolution module is connected to other modules; 2) the fine stage, where we search at micro-level for operations in each cell based on previous searched macro-level topology. The coarse-to-fine manner divides the search procedure into two consecutive stages and meanwhile resolves the inconsistency. We evaluate our method on 10 public datasets from Medical Segmentation Decalthon (MSD) challenge, and achieve state-of-the-art performance with the network searched using one dataset, which demonstrates the effectiveness and generalization of our searched models."
                },
                {
                    "title": "Dynamic Convolution: Attention Over Convolution Kernels",
                    "abstract": "Light-weight convolutional neural networks (CNNs) suffer performance degradation as their low computational budgets constrain both the depth (number of convolution layers) and the width (number of channels) of CNNs, resulting in limited representation capability. To address this issue, we present Dynamic Convolution, a new design that increases model complexity without increasing the network depth or width. Instead of using a single convolution kernel per layer, dynamic convolution aggregates multiple parallel convolution kernels dynamically based upon their attentions, which are input dependent. Assembling multiple kernels is not only computationally efficient due to the small kernel size, but also has more representation power since these kernels are aggregated in a non-linear way via attention. By simply using dynamic convolution for the state-of-the-art architecture MobileNetV3-Small, the top-1 accuracy of ImageNet classification is boosted by 2.9% with only 4% additional FLOPs and 2.9 AP gain is achieved on COCO keypoint detection."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Rigging the Lottery: Making All Tickets Winners",
                    "abstract": "Many applications require sparse neural networks due to space or inference time restrictions. There is a large body of work on training dense networks to yield sparse networks for inference, but this limits the size of the largest trainable sparse model to that of the largest trainable dense model. In this paper we introduce a method to train sparse neural networks with a fixed parameter count and a fixed computational cost throughout training, without sacrificing accuracy relative to existing dense-to-sparse training methods. Our method updates the topology of the sparse network during training by using parameter magnitudes and infrequent gradient calculations. We show that this approach requires fewer floating-point operations (FLOPs) to achieve a given level of accuracy compared to prior techniques. We demonstrate state-of-the-art sparse training results on a variety of networks and datasets, including ResNet-50, MobileNets on Imagenet-2012, and RNNs on WikiText-103. Finally, we provide some insights into why allowing the topology to change during the optimization can overcome local minima encountered when the topology remains static. Code used in our work can be found in this http URL."
                },
                {
                    "title": "EfficientDet: Scalable and Efficient Object Detection",
                    "abstract": "Model efficiency has become increasingly important in computer vision. In this paper, we systematically study neural network architecture design choices for object detection and propose several key optimizations to improve efficiency. First, we propose a weighted bi-directional feature pyramid network (BiFPN), which allows easy and fast multi-scale feature fusion; Second, we propose a compound scaling method that uniformly scales the resolution, depth, and width for all backbone, feature network, and box/class prediction networks at the same time. Based on these optimizations and EfficientNet backbones, we have developed a new family of object detectors, called EfficientDet, which consistently achieve much better efficiency than prior art across a wide spectrum of resource constraints. In particular, with single-model and single-scale, our EfficientDetD7 achieves state-of-the-art 52.2 AP on COCO test-dev with 52M parameters and 325B FLOPs, being 4x \u2013 9x smaller and using 13x \u2013 42x fewer FLOPs than previous detector."
                },
                {
                    "title": "Prior-Aware Neural Network for Partially-Supervised Multi-Organ Segmentation",
                    "abstract": "Accurate multi-organ abdominal CT segmentation is essential to many clinical applications such as computer-aided intervention. As data annotation requires massive human labor from experienced radiologists, it is common that training data is usually partially-labeled. However, these background labels can be misleading in multi-organ segmentation since the ``background'' usually contains some other organs of interest. To address the background ambiguity in these partially-labeled datasets, we propose Prior-aware Neural Network (PaNN) via explicitly incorporating anatomical priors on abdominal organ sizes, guiding the training process with domain-specific knowledge. More specifically, PaNN assumes that the average organ size distributions in the abdomen should approximate their empirical distributions, a prior statistics obtained from the fully-labeled dataset. As our objective is difficult to be directly optimized using stochastic gradient descent, it is reformulated as a min-max form and optimized via the stochastic primal-dual gradient algorithm. PaNN achieves state-of-the-art performance on the MICCAI2015 challenge ``Multi-Atlas Labeling Beyond the Cranial Vault'', a competition on organ segmentation in the abdomen. We report an average Dice score of 84.97%, surpassing the prior art by a large margin of 3.27%. Code and models will be made publicly available."
                },
                {
                    "title": "NAS-Unet: Neural Architecture Search for Medical Image Segmentation",
                    "abstract": "Neural architecture search (NAS) has significant progress in improving the accuracy of image classification. Recently, some works attempt to extend NAS to image segmentation which shows preliminary feasibility. However, all of them focus on searching architecture for semantic segmentation in natural scenes. In this paper, we design three types of primitive operation set on search space to automatically find two cell architecture DownSC and UpSC for semantic image segmentation especially medical image segmentation. Inspired by the U-net architecture and its variants successfully applied to various medical image segmentation, we propose NAS-Unet which is stacked by the same number of DownSC and UpSC on a U-like backbone network. The architectures of DownSC and UpSC updated simultaneously by a differential architecture strategy during the search stage. We demonstrate the good segmentation results of the proposed method on Promise12, Chaos, and ultrasound nerve datasets, which collected by magnetic resonance imaging, computed tomography, and ultrasound, respectively. Without any pretraining, our architecture searched on PASCAL VOC2012, attains better performances and much fewer parameters (about 0.8M) than U-net and one of its variants when evaluated on the above three types of medical image datasets."
                },
                {
                    "title": "A large annotated medical image dataset for the development and evaluation of segmentation algorithms",
                    "abstract": "Semantic segmentation of medical images aims to associate a pixel with a label in a medical image without human initialization. The success of semantic segmentation algorithms is contingent on the availability of high-quality imaging data with corresponding labels provided by experts. We sought to create a large collection of annotated medical image datasets of various clinically relevant anatomies available under open source license to facilitate the development of semantic segmentation algorithms. Such a resource would allow: 1) objective assessment of general-purpose segmentation methods through comprehensive benchmarking and 2) open and free access to medical image data for any researcher interested in the problem domain. Through a multi-institutional effort, we generated a large, curated dataset representative of several highly variable segmentation tasks that was used in a crowd-sourced challenge - the Medical Segmentation Decathlon held during the 2018 Medical Image Computing and Computer Aided Interventions Conference in Granada, Spain. Here, we describe these ten labeled image datasets so that these data may be effectively reused by the research community."
                },
                {
                    "title": "Parameter Efficient Training of Deep Convolutional Neural Networks by Dynamic Sparse Reparameterization",
                    "abstract": "Modern deep neural networks are typically highly overparameterized. Pruning techniques are able to remove a significant fraction of network parameters with little loss in accuracy. Recently, techniques based on dynamic reallocation of non-zero parameters have emerged, allowing direct training of sparse networks without having to pre-train a large dense model. Here we present a novel dynamic sparse reparameterization method that addresses the limitations of previous techniques such as high computational cost and the need for manual configuration of the number of free parameters allocated to each layer. We evaluate the performance of dynamic reallocation methods in training deep convolutional networks and show that our method outperforms previous static and dynamic reparameterization methods, yielding the best accuracy for a fixed parameter budget, on par with accuracies obtained by iteratively pruning a pre-trained dense model. We further investigated the mechanisms underlying the superior generalization performance of the resultant sparse networks. We found that neither the structure, nor the initialization of the non-zero parameters were sufficient to explain the superior performance. Rather, effective learning crucially depended on the continuous exploration of the sparse network structure space during training. Our work suggests that exploring structural degrees of freedom during training is more effective than adding extra parameters to the network."
                },
                {
                    "title": "TSM: Temporal Shift Module for Efficient Video Understanding",
                    "abstract": "The explosive growth in video streaming gives rise to challenges on performing video understanding at high accuracy and low computation cost. Conventional 2D CNNs are computationally cheap but cannot capture temporal relationships; 3D CNN based methods can achieve good performance but are computationally intensive, making it expensive to deploy. In this paper, we propose a generic and effective Temporal Shift Module (TSM) that enjoys both high efficiency and high performance. Specifically, it can achieve the performance of 3D CNN but maintain 2D CNN\u2019s complexity. TSM shifts part of the channels along the temporal dimension; thus facilitate information exchanged among neighboring frames. It can be inserted into 2D CNNs to achieve temporal modeling at zero computation and zero parameters. We also extended TSM to online setting, which enables real-time low-latency online video recognition and video object detection. TSM is accurate and efficient: it ranks the first place on the Something-Something leaderboard upon publication; on Jetson Nano and Galaxy Note8, it achieves a low latency of 13ms and 35ms for online video recognition. The code is available at: https://github. com/mit-han-lab/temporal-shift-module."
                },
                {
                    "title": "SNIP: Single-shot Network Pruning based on Connection Sensitivity",
                    "abstract": "Pruning large neural networks while maintaining their performance is often desirable due to the reduced space and time complexity. In existing methods, pruning is done within an iterative optimization procedure with either heuristically designed pruning schedules or additional hyperparameters, undermining their utility. In this work, we present a new approach that prunes a given network once at initialization prior to training. To achieve this, we introduce a saliency criterion based on connection sensitivity that identifies structurally important connections in the network for the given task. This eliminates the need for both pretraining and the complex pruning schedule while making it robust to architecture variations. After pruning, the sparse network is trained in the standard way. Our method obtains extremely sparse networks with virtually the same accuracy as the reference network on the MNIST, CIFAR-10, and Tiny-ImageNet classification tasks and is broadly applicable to various architectures including convolutional, residual and recurrent networks. Unlike existing methods, our approach enables us to demonstrate that the retained connections are indeed relevant to the given task."
                },
                {
                    "title": "Attention U-Net: Learning Where to Look for the Pancreas",
                    "abstract": "We propose a novel attention gate (AG) model for medical imaging that automatically learns to focus on target structures of varying shapes and sizes. Models trained with AGs implicitly learn to suppress irrelevant regions in an input image while highlighting salient features useful for a specific task. This enables us to eliminate the necessity of using explicit external tissue/organ localisation modules of cascaded convolutional neural networks (CNNs). AGs can be easily integrated into standard CNN architectures such as the U-Net model with minimal computational overhead while increasing the model sensitivity and prediction accuracy. The proposed Attention U-Net architecture is evaluated on two large CT abdominal datasets for multi-class image segmentation. Experimental results show that AGs consistently improve the prediction performance of U-Net across different datasets and training sizes while preserving computational efficiency. The code for the proposed architecture is publicly available."
                },
                {
                    "title": "Instance Normalization: The Missing Ingredient for Fast Stylization",
                    "abstract": "It this paper we revisit the fast stylization method introduced in Ulyanov et. al. (2016). We show how a small change in the stylization architecture results in a significant qualitative improvement in the generated images. The change is limited to swapping batch normalization with instance normalization, and to apply the latter both at training and testing times. The resulting method can be used to train high-performance architectures for real-time image generation. The code will is made available on github at this https URL. Full paper can be found at arXiv:1701.02096."
                },
                {
                    "title": "V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation",
                    "abstract": "Convolutional Neural Networks (CNNs) have been recently employed to solve problems from both the computer vision and medical image analysis fields. Despite their popularity, most approaches are only able to process 2D images while most medical data used in clinical practice consists of 3D volumes. In this work we propose an approach to 3D image segmentation based on a volumetric, fully convolutional, neural network. Our CNN is trained end-to-end on MRI volumes depicting prostate, and learns to predict segmentation for the whole volume at once. We introduce a novel objective function, that we optimise during training, based on Dice coefficient. In this way we can deal with situations where there is a strong imbalance between the number of foreground and background voxels. To cope with the limited number of annotated volumes available for training, we augment the data applying random non-linear transformations and histogram matching. We show in our experimental evaluation that our approach achieves good performances on challenging test data while requiring only a fraction of the processing time needed by other previous methods."
                },
                {
                    "title": "nnFormer: Interleaved Transformer for Volumetric Segmentation",
                    "abstract": ". Transformers, the default model of choices in natural language processing, have drawn scant attention from the medical imaging community. Given the ability to exploit long-term dependencies, transformers are promising to help atypical convolutional neural networks (convnets) to overcome its inherent shortcomings of spatial inductive bias. However, most of recently proposed transformer-based segmentation approaches simply treated transformers as assisted modules to help encode global context into convolutional representations without investigating how to optimally combine self-attention (i.e., the core of transformers) with convolution. To address this issue, in this paper, we introduce nnFormer (i.e., n ot-a n other trans Former ), a powerful segmentation model with an interleaved architecture based on empirical combination of self-attention and convolution. In practice, nnFormer learns volumetric representations from 3D local volumes. Compared to the naive voxel-level self-attention implementation, such volume-based operations help to reduce the computational complexity by approximate 98% and 99.5% on Synapse and ACDC datasets, respectively. In comparison to prior-art network con\ufb01gurations, nnFormer achieves tremendous improvements over previous transformer-based methods on two commonly used datasets Synapse and ACDC. For instance, nnFormer outperforms Swin-UNet by over 7 percents on Synapse. Even when compared to nnUNet, currently the best performing fully-convolutional medical segmentation network, nnFormer still provides slightly better performance on Synapse and ACDC. Codes and models are available at https://github.com/282857341/nnFormer ."
                },
                {
                    "title": "Rectifier Nonlinearities Improve Neural Network Acoustic Models",
                    "abstract": "Deep neural network acoustic models produce substantial gains in large vocabulary continuous speech recognition systems. Emerging work with recti\ufb01ed linear (ReL) hidden units demonstrates additional gains in \ufb01nal system performance relative to more commonly used sigmoidal nonlinearities. In this work, we explore the use of deep recti\ufb01er networks as acoustic models for the 300 hour Switchboard conversational speech recognition task. Using simple training procedures without pretraining, networks with recti\ufb01er nonlinearities produce 2% absolute reductions in word error rates over their sigmoidal counterparts. We analyze hidden layer representations to quantify di\ufb00erences in how ReL units encode inputs as compared to sigmoidal units. Finally, we evaluate a variant of the ReL unit with a gradient more amenable to optimization in an attempt to further improve deep recti\ufb01er networks."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan we design a 3D medical image segmentation method that trades off accuracy and efficiency better subjected to different resource availability?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing need for efficient and accurate medical image segmentation methods that can be deployed in real-world clinical settings. By improving the accuracy-efficiency trade-off, this research could lead to advancements in computer-aided diagnosis and image-guided surgery systems, ultimately enhancing patient outcomes. Furthermore, it could inspire future research to explore novel architectures and optimization techniques that balance performance with computational resource constraints, paving the way for more accessible and practical applications in healthcare.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent variability in organ size and shape, which complicates accurate boundary identification in 3D medical images. Naive approaches may fail due to their inability to effectively leverage multi-scale features or adapt to diverse anatomical structures. Additionally, existing methods like neural architecture search (NAS) are computationally expensive and time-consuming, requiring significant resources to explore and optimize network topologies. Overcoming these technical and practical obstacles is essential for developing a viable solution.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the high computational costs associated with complex architectures and optimization techniques, such as NAS, which require extensive resources and time to identify effective models. Additionally, many existing solutions do not adequately address the dynamic nature of feature fusion needed for 3D segmentation. Our approach differs by eliminating the need for costly architecture searches and instead optimizing sparse topology within a predefined architecture, making it more efficient and practical for real-world applications.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Efficient to Efficient Network (E2ENet), focuses on dynamically incorporating both bottom-up and top-down information flows from the backbone network in a sparse pattern. We will utilize a dataset of 3D medical images for segmentation tasks and evaluate our model using metrics such as Dice coefficient and Intersection over Union (IoU) to measure accuracy. The expected outcomes include improved accuracy in organ segmentation while significantly reducing computational costs, thereby achieving a better trade-off between accuracy and efficiency in 3D medical image segmentation."
            }
        },
        "author_data": {
            "4b1ebaa5-4f72-41d2-94ee-50bb048e6add": {
                "pk": "4b1ebaa5-4f72-41d2-94ee-50bb048e6add",
                "name": "Boqian Wu",
                "collaborators": [
                    "Qiao Xiao",
                    "Mykola Pechenizkiy",
                    "Decebal Constantin Mocanu",
                    "Shiwei Liu",
                    "Elena Mocanu",
                    "Lu Yin",
                    "Shunxin Wang",
                    "Nicola Strisciuglio",
                    "Maurice van Keulen",
                    "Yu Zhang"
                ],
                "domain": [
                    "Dynamic Sparse Training",
                    "Deep Learning",
                    "Time Series Classification",
                    "Automatic Speech Recognition"
                ],
                "publications": [
                    {
                        "title": "Dynamic Sparse Training versus Dense Training: The Unexpected Winner in Image Corruption Robustness",
                        "abstract": "It is generally perceived that Dynamic Sparse Training opens the door to a new era of scalability and efficiency for artificial neural networks at, perhaps, some costs in accuracy performance for the classification task. At the same time, Dense Training is widely accepted as being the \"de facto\" approach to train artificial neural networks if one would like to maximize their robustness against image corruption. In this paper, we question this general practice. Consequently, we claim that, contrary to what is commonly thought, the Dynamic Sparse Training methods can consistently outperform Dense Training in terms of robustness accuracy, particularly if the efficiency aspect is not considered as a main objective (i.e., sparsity levels between 10% and up to 50%), without adding (or even reducing) resource cost. We validate our claim on two types of data, images and videos, using several traditional and modern deep learning architectures for computer vision and three widely studied Dynamic Sparse Training algorithms. Our findings reveal a new yet-unknown benefit of Dynamic Sparse Training and open new possibilities in improving deep learning robustness beyond the current state of the art."
                    },
                    {
                        "title": "Dynamic Sparse Network for Time Series Classification: Learning What to \"see''",
                        "abstract": "The receptive field (RF), which determines the region of time series to be ``seen'' and used, is critical to improve the performance for time series classification (TSC). However, the variation of signal scales across and within time series data, makes it challenging to decide on proper RF sizes for TSC. In this paper, we propose a dynamic sparse network (DSN) with sparse connections for TSC, which can learn to cover various RF without cumbersome hyper-parameters tuning. The kernels in each sparse layer are sparse and can be explored under the constraint regions by dynamic sparse training, which makes it possible to reduce the resource cost. The experimental results show that the proposed DSN model can achieve state-of-art performance on both univariate and multivariate TSC datasets with less than 50\\% computational cost compared with recent baseline methods, opening the path towards more accurate resource-aware methods for time series analyses. Our code is publicly available at: https://github.com/QiaoXiao7282/DSN."
                    },
                    {
                        "title": "Are Sparse Neural Networks Better Hard Sample Learners?",
                        "abstract": "While deep learning has demonstrated impressive progress, it remains a daunting challenge to learn from hard samples as these samples are usually noisy and intricate. These hard samples play a crucial role in the optimal performance of deep neural networks. Most research on Sparse Neural Networks (SNNs) has focused on standard training data, leaving gaps in understanding their effectiveness on complex and challenging data. This paper's extensive investigation across scenarios reveals that most SNNs trained on challenging samples can often match or surpass dense models in accuracy at certain sparsity levels, especially with limited data. We observe that layer-wise density ratios tend to play an important role in SNN performance, particularly for methods that train from scratch without pre-trained initialization. These insights enhance our understanding of SNNs' behavior and potential for efficient learning approaches in data-centric AI. Our code is publicly available at: \\url{https://github.com/QiaoXiao7282/hard_sample_learners}."
                    },
                    {
                        "title": "More ConvNets in the 2020s: Scaling up Kernels Beyond 51x51 using Sparsity",
                        "abstract": "Transformers have quickly shined in the computer vision world since the emergence of Vision Transformers (ViTs). The dominant role of convolutional neural networks (CNNs) seems to be challenged by increasingly effective transformer-based models. Very recently, a couple of advanced convolutional models strike back with large kernels motivated by the local-window attention mechanism, showing appealing performance and efficiency. While one of them, i.e. RepLKNet, impressively manages to scale the kernel size to 31x31 with improved performance, the performance starts to saturate as the kernel size continues growing, compared to the scaling trend of advanced ViTs such as Swin Transformer. In this paper, we explore the possibility of training extreme convolutions larger than 31x31 and test whether the performance gap can be eliminated by strategically enlarging convolutions. This study ends up with a recipe for applying extremely large kernels from the perspective of sparsity, which can smoothly scale up kernels to 61x61 with better performance. Built on this recipe, we propose Sparse Large Kernel Network (SLaK), a pure CNN architecture equipped with sparse factorized 51x51 kernels that can perform on par with or better than state-of-the-art hierarchical Transformers and modern ConvNet architectures like ConvNeXt and RepLKNet, on ImageNet classification as well as a wide range of downstream tasks including semantic segmentation on ADE20K, object detection on PASCAL VOC 2007, and object detection/segmentation on MS COCO."
                    },
                    {
                        "title": "Dynamic Data Pruning for Automatic Speech Recognition",
                        "abstract": "The recent success of Automatic Speech Recognition (ASR) is largely attributed to the ever-growing amount of training data. However, this trend has made model training prohibitively costly and imposed computational demands. While data pruning has been proposed to mitigate this issue by identifying a small subset of relevant data, its application in ASR has been barely explored, and existing works often entail significant overhead to achieve meaningful results. To fill this gap, this paper presents the first investigation of dynamic data pruning for ASR, finding that we can reach the full-data performance by dynamically selecting 70% of data. Furthermore, we introduce Dynamic Data Pruning for ASR (DDP-ASR), which offers several fine-grained pruning granularities specifically tailored for speech-related datasets, going beyond the conventional pruning of entire time sequences. Our intensive experiments show that DDP-ASR can save up to 1.6x training time with negligible performance loss."
                    }
                ]
            },
            "0b7678fd-3204-4c2d-b4a5-03d8e36e025e": {
                "pk": "0b7678fd-3204-4c2d-b4a5-03d8e36e025e",
                "name": "Qiao Xiao",
                "collaborators": [
                    "Yu Zhang"
                ],
                "domain": [
                    "Transfer Learning",
                    "Distant Transfer Learning",
                    "Graph-Based Learning"
                ],
                "publications": [
                    {
                        "title": "Distant Transfer Learning via Deep Random Walk",
                        "abstract": "Transfer learning, which is to improve the learning performance in the target domain by leveraging useful knowledge from the source domain, often requires that those two domains are very close, which limits its application scope. Recently, distant transfer learning has been studied to transfer knowledge between two distant or even totally unrelated domains via auxiliary domains that are usually unlabeled as a bridge in the spirit of human transitive inference that it is possible to connect two completely unrelated concepts together through gradual knowledge transfer. In this paper, we study distant transfer learning by proposing a DeEp Random Walk basEd distaNt Transfer (DERWENT) method. Different from existing distant transfer learning models that implicitly identify the path of knowledge transfer between the source and target instances through auxiliary instances, the proposed DERWENT model can explicitly learn such paths via the deep random walk technique. Specifically, based on sequences identified by the random walk technique on a data graph where source and target data have no direct edges, the proposed DERWENT model enforces adjacent data points in a squence to be similar, makes the ending data point be represented by other data points in the same sequence, and considers weighted training losses of source data. Empirical studies on several benchmark datasets demonstrate that the proposed DERWENT algorithm yields the state-of-the-art performance."
                    }
                ]
            },
            "e63bcba1-9319-4a93-b05e-51834c3cecb6": {
                "pk": "e63bcba1-9319-4a93-b05e-51834c3cecb6",
                "name": "Shiwei Liu",
                "collaborators": [
                    "Mykola Pechenizkiy",
                    "Decebal Constantin Mocanu",
                    "Lu Yin",
                    "Zhangyang Wang",
                    "Tianlong Chen",
                    "Yuesong Tian",
                    "Li Shen",
                    "Vlado Menkovski",
                    "Ajay Jaiswal",
                    "Pengxiang Li"
                ],
                "domain": [
                    "Sparse Neural Networks",
                    "Deep Learning",
                    "Model Compression",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Ten Lessons We Have Learned in the New \"Sparseland\": A Short Handbook for Sparse Neural Network Researchers",
                        "abstract": "This article does not propose any novel algorithm or new hardware for sparsity. Instead, it aims to serve the \"common good\" for the increasingly prosperous Sparse Neural Network (SNN) research community. We attempt to summarize some most common confusions in SNNs, that one may come across in various scenarios such as paper review/rebuttal and talks - many drawn from the authors' own bittersweet experiences! We feel that doing so is meaningful and timely, since the focus of SNN research is notably shifting from traditional pruning to more diverse and profound forms of sparsity before, during, and after training. The intricate relationships between their scopes, assumptions, and approaches lead to misunderstandings, for non-experts or even experts in SNNs. In response, we summarize ten Q\\&As of SNNs from many key aspects, including dense vs. sparse, unstructured sparse vs. structured sparse, pruning vs. sparse training, dense-to-sparse training vs. sparse-to-sparse training, static sparsity vs. dynamic sparsity, before-training/during-training vs. post-training sparsity, and many more. We strive to provide proper and generically applicable answers to clarify those confusions to the best extent possible. We hope our summary provides useful general knowledge for people who want to enter and engage with this exciting community; and also provides some \"mind of ease\" convenience for SNN researchers to explain their work in the right contexts. At the very least (and perhaps as this article's most insignificant target functionality), if you are writing/planning to write a paper or rebuttal in the field of SNNs, we hope some of our answers could help you!"
                    },
                    {
                        "title": "On improving deep learning generalization with adaptive sparse connectivity",
                        "abstract": "Large neural networks are very successful in various tasks. However, with limited data, the generalization capabilities of deep neural networks are also very limited. In this paper, we empirically start showing that intrinsically sparse neural networks with adaptive sparse connectivity, which by design have a strict parameter budget during the training phase, have better generalization capabilities than their fully-connected counterparts. Besides this, we propose a new technique to train these sparse models by combining the Sparse Evolutionary Training (SET) procedure with neurons pruning. Operated on MultiLayer Perceptron (MLP) and tested on 15 datasets, our proposed technique zeros out around 50% of the hidden neurons during training, while having a linear number of parameters to optimize with respect to the number of neurons. The results show a competitive classification and generalization performance."
                    },
                    {
                        "title": "Intrinsically Sparse Long Short-Term Memory Networks",
                        "abstract": "Long Short-Term Memory (LSTM) has achieved state-of-the-art performances on a wide range of tasks. Its outstanding performance is guaranteed by the long-term memory ability which matches the sequential data perfectly and the gating structure controlling the information flow. However, LSTMs are prone to be memory-bandwidth limited in realistic applications and need an unbearable period of training and inference time as the model size is ever-increasing. To tackle this problem, various efficient model compression methods have been proposed. Most of them need a big and expensive pre-trained model which is a nightmare for resource-limited devices where the memory budget is strictly limited. To remedy this situation, in this paper, we incorporate the Sparse Evolutionary Training (SET) procedure into LSTM, proposing a novel model dubbed SET-LSTM. Rather than starting with a fully-connected architecture, SET-LSTM has a sparse topology and dramatically fewer parameters in both phases, training and inference. Considering the specific architecture of LSTMs, we replace the LSTM cells and embedding layers with sparse structures and further on, use an evolutionary strategy to adapt the sparse connectivity to the data. Additionally, we find that SET-LSTM can provide many different good combinations of sparse connectivity to substitute the overparameterized optimization problem of dense neural networks. Evaluated on four sentiment analysis classification datasets, the results demonstrate that our proposed model is able to achieve usually better performance than its fully connected counterpart while having less than 4\\% of its parameters."
                    },
                    {
                        "title": "Don't Be So Dense: Sparse-to-Sparse GAN Training Without Sacrificing Performance",
                        "abstract": "Generative adversarial networks (GANs) have received an upsurging interest since being proposed due to the high quality of the generated data. While achieving increasingly impressive results, the resource demands associated with the large model size hinders the usage of GANs in resource-limited scenarios. For inference, the existing model compression techniques can reduce the model complexity with comparable performance. However, the training efficiency of GANs has less been explored due to the fragile training process of GANs. In this paper, we, for the first time, explore the possibility of directly training sparse GAN from scratch without involving any dense or pre-training steps. Even more unconventionally, our proposed method enables directly training sparse unbalanced GANs with an extremely sparse generator from scratch. Instead of training full GANs, we start with sparse GANs and dynamically explore the parameter space spanned over the generator throughout training. Such a sparse-to-sparse training procedure enhances the capacity of the highly sparse generator progressively while sticking to a fixed small parameter budget with appealing training and inference efficiency gains. Extensive experiments with modern GAN architectures validate the effectiveness of our method. Our sparsified GANs, trained from scratch in one single run, are able to outperform the ones learned by expensive iterative pruning and re-training. Perhaps most importantly, we find instead of inheriting parameters from expensive pre-trained GANs, directly training sparse GANs from scratch can be a much more efficient solution. For example, only training with a 80% sparse generator and a 70% sparse discriminator, our method can achieve even better performance than the dense BigGAN."
                    },
                    {
                        "title": "Hierarchical Semantic Segmentation using Psychometric Learning",
                        "abstract": "Assigning meaning to parts of image data is the goal of semantic image segmentation. Machine learning methods, specifically supervised learning is commonly used in a variety of tasks formulated as semantic segmentation. One of the major challenges in the supervised learning approaches is expressing and collecting the rich knowledge that experts have with respect to the meaning present in the image data. Towards this, typically a fixed set of labels is specified and experts are tasked with annotating the pixels, patches or segments in the images with the given labels. In general, however, the set of classes does not fully capture the rich semantic information present in the images. For example, in medical imaging such as histology images, the different parts of cells could be grouped and sub-grouped based on the expertise of the pathologist.   To achieve such a precise semantic representation of the concepts in the image, we need access to the full depth of knowledge of the annotator. In this work, we develop a novel approach to collect segmentation annotations from experts based on psychometric testing. Our method consists of the psychometric testing procedure, active query selection, query enhancement, and a deep metric learning model to achieve a patch-level image embedding that allows for semantic segmentation of images. We show the merits of our method with evaluation on the synthetically generated image, aerial image and histology image."
                    },
                    {
                        "title": "The Emergence of Essential Sparsity in Large Pre-trained Models: The Weights that Matter",
                        "abstract": "Large pre-trained transformers are show-stealer in modern-day deep learning, and it becomes crucial to comprehend the parsimonious patterns that exist within them as they grow in scale. With exploding parameter counts, Lottery Ticket Hypothesis (LTH) and its variants, have lost their pragmatism in sparsifying them due to high computation and memory bottleneck of repetitive train-prune-retrain routine of iterative magnitude pruning (IMP) which worsens with increasing model size. This paper comprehensively studies induced sparse patterns across multiple large pre-trained vision and language transformers. We propose the existence of -- essential sparsity defined with a sharp dropping point beyond which the performance declines much faster w.r.t the rise of sparsity level, when we directly remove weights with the smallest magnitudes in one-shot without re-training. We also find essential sparsity to hold valid for N:M sparsity patterns as well as on modern-scale large language models (Vicuna-7B). We also present an intriguing emerging phenomenon of abrupt sparsification during the pre-training of BERT, i.e., BERT suddenly becomes heavily sparse in pre-training after certain iterations. Moreover, our observations also indicate a counter-intuitive finding that BERT trained with a larger amount of pre-training data tends to have a better ability to condense knowledge in comparatively relatively fewer parameters. Lastly, we investigate the effect of the pre-training loss on essential sparsity and discover that self-supervised learning (SSL) objectives trigger stronger emergent sparsification properties than supervised learning (SL). Our codes are available at \\url{https://github.com/VITA-Group/essential_sparsity}."
                    },
                    {
                        "title": "OwLore: Outlier-weighed Layerwise Sampled Low-Rank Projection for Memory-Efficient LLM Fine-tuning",
                        "abstract": "The rapid advancements in Large Language Models (LLMs) have revolutionized various natural language processing tasks. However, the substantial size of LLMs presents significant challenges in training or fine-tuning. While parameter-efficient approaches such as low-rank adaptation (LoRA) have gained popularity, they often compromise performance compared to full-rank fine-tuning. In this paper, we propose Outlier-weighed Layerwise Sampled Low-Rank Projection (OwLore), a new memory-efficient fine-tuning approach, inspired by the layerwise outlier distribution of LLMs. Unlike LoRA, which adds extra adapters to all layers, OwLore strategically assigns higher sampling probabilities to layers with more outliers, selectively sampling only a few layers and fine-tuning their pre-trained weights. To further increase the number of fine-tuned layers without a proportional rise in memory costs, we incorporate gradient low-rank projection, further boosting the approach's performance. Our extensive experiments across various architectures, including LLaMa2, LLaMa3, and Mistral, demonstrate that OwLore consistently outperforms baseline approaches, including full fine-tuning. Specifically, it achieves up to a 1.1% average accuracy gain on the Commonsense Reasoning benchmark, a 3.0% improvement on MMLU, and a notable 10% boost on MT-Bench, while being more memory efficient. OwLore allows us to fine-tune LLaMa2-7B with only 21GB of memory. Code is available at https://github.com/pixeli99/OwLore."
                    },
                    {
                        "title": "Deuterium-tritium fusion process in strong laser fields: Semiclassical simulation",
                        "abstract": "In this paper, we investigate the deuterium-tritium (DT) fusion process in the presence of strong laser fields with a semiclassical (SC) method. In this model, two nuclei with a given incident kinetic energy that closely approach each other are simulated by tracing the classical Newtonian trajectories in the combined Coulomb repulsive potentials and laser fields. At the nearest position or classical turning point, quantum tunneling through the Coulomb barrier emerges, and its penetrability is estimated with the Wentzel-Kramers-Brillouin formula. Nuclear fusion occurs after the tunneling, and the total fusion cross section takes the Gamow form. We find that the tunneling penetrability can be enhanced dramatically because the nuclei can closely approach each other due to the quiver motion of the charged nuclei driven by the intense laser fields. We then calculate the DT fusion section for a wide range of laser parameters according to various incident nuclei kinetic energies and obtain the phase diagrams for the enhanced DT fusion. We compare our SC results with the quantum results of the Kramers-Henneberger approximation and the Volkov state approximation."
                    },
                    {
                        "title": "Do We Actually Need Dense Over-Parameterization? In-Time Over-Parameterization in Sparse Training",
                        "abstract": "In this paper, we introduce a new perspective on training deep neural networks capable of state-of-the-art performance without the need for the expensive over-parameterization by proposing the concept of In-Time Over-Parameterization (ITOP) in sparse training. By starting from a random sparse network and continuously exploring sparse connectivities during training, we can perform an Over-Parameterization in the space-time manifold, closing the gap in the expressibility between sparse training and dense training. We further use ITOP to understand the underlying mechanism of Dynamic Sparse Training (DST) and indicate that the benefits of DST come from its ability to consider across time all possible parameters when searching for the optimal sparse connectivity. As long as there are sufficient parameters that have been reliably explored during training, DST can outperform the dense neural network by a large margin. We present a series of experiments to support our conjecture and achieve the state-of-the-art sparse training performance with ResNet-50 on ImageNet. More impressively, our method achieves dominant performance over the overparameterization-based sparse methods at extreme sparsity levels. When trained on CIFAR-100, our method can match the performance of the dense model even at an extreme sparsity (98%). Code can be found https://github.com/Shiweiliuiiiiiii/In-Time-Over-Parameterization."
                    },
                    {
                        "title": "Selfish Sparse RNN Training",
                        "abstract": "Sparse neural networks have been widely applied to reduce the computational demands of training and deploying over-parameterized deep neural networks. For inference acceleration, methods that discover a sparse network from a pre-trained dense network (dense-to-sparse training) work effectively. Recently, dynamic sparse training (DST) has been proposed to train sparse neural networks without pre-training a dense model (sparse-to-sparse training), so that the training process can also be accelerated. However, previous sparse-to-sparse methods mainly focus on Multilayer Perceptron Networks (MLPs) and Convolutional Neural Networks (CNNs), failing to match the performance of dense-to-sparse methods in the Recurrent Neural Networks (RNNs) setting. In this paper, we propose an approach to train intrinsically sparse RNNs with a fixed parameter count in one single run, without compromising performance. During training, we allow RNN layers to have a non-uniform redistribution across cell gates for better regularization. Further, we propose SNT-ASGD, a novel variant of the averaged stochastic gradient optimizer, which significantly improves the performance of all sparse training methods for RNNs. Using these strategies, we achieve state-of-the-art sparse training results, better than the dense-to-sparse methods, with various types of RNNs on Penn TreeBank and Wikitext-2 datasets. Our codes are available at https://github.com/Shiweiliuiiiiiii/Selfish-RNN."
                    },
                    {
                        "title": "Phase-sensitive nonclassical properties in quantum metrology with a displaced squeezed vacuum state",
                        "abstract": "We predict that the phase-dependent error distribution of locally unentangled quantum states directly affects quantum parameter estimation accuracy. Therefore, we employ the displaced squeezed vacuum (DSV) state as a probe state and investigate an interesting question of the phase-sensitive nonclassical properties in DSV's metrology. We found that the accuracy limit of parameter estimation is a function of the phase-sensitive parameter $\\phi -\\theta /2$ with a period $\\pi $. We show that when $\\phi -\\theta /2$ $\\in \\left[ k\\pi/2,3k\\pi /4\\right) \\left( k\\in \\mathbb{Z}\\right)$, we can obtain the accuracy of parameter estimation approaching the ultimate quantum limit through using the DSV state with the larger displacement and squeezing strength, whereas $\\phi -\\theta /2$ $\\in \\left(3k\\pi /4,k\\pi \\right] \\left( k\\in \\mathbb{Z}\\right) $, the optimal estimation accuracy can be acquired only when the DSV state degenerates to a squeezed-vacuum state."
                    },
                    {
                        "title": "Optimized Path Planning for USVs under Ocean Currents",
                        "abstract": "Unmanned Surface Vehicles (USVs) in the ocean environment, considering various spatiotemporal factors such as ocean currents and other energy consumption factors. The paper uses Gaussian Process Motion Planning (GPMP2), a Bayesian optimization method that has shown promising results in continuous and nonlinear motion planning algorithms. The proposed work improves GPMP2 by incorporating a new spatiotemporal factor for tracking and predicting ocean currents using a spatiotemporal Bayesian inference. The algorithm is applied to the USV path planning and is shown to optimize for smoothness, obstacle avoidance, and ocean currents in a challenging environment. The work is relevant for practical applications in ocean scenarios where optimal path planning for USVs is essential for minimizing costs and optimizing performance."
                    },
                    {
                        "title": "Full-Rank No More: Low-Rank Weight Training for Modern Speech Recognition Models",
                        "abstract": "This paper investigates the under-explored area of low-rank weight training for large-scale Conformer-based speech recognition models from scratch. Our study demonstrates the viability of this training paradigm for such models, yielding several notable findings. Firstly, we discover that applying a low-rank structure exclusively to the attention modules can unexpectedly enhance performance, even with a significant rank reduction of 12%. In contrast, feed-forward layers present greater challenges, as they begin to exhibit performance degradation with a moderate 50% rank reduction. Furthermore, we find that both initialization and layer-wise rank assignment play critical roles in successful low-rank training. Specifically, employing SVD initialization and linear layer-wise rank mapping significantly boosts the efficacy of low-rank weight training. Building on these insights, we introduce the Low-Rank Speech Model from Scratch (LR-SMS), an approach that achieves performance parity with full-rank training while delivering substantial reductions in parameters count (by at least 2x), and training time speedups (by 1.3x for ASR and 1.15x for AVSR)."
                    },
                    {
                        "title": "Direct numerical simulation of compressible turbulence accelerated by graphics processing unit. Part 1: An open-source high accuracy accelerated computational fluid dynamic software",
                        "abstract": "This paper introduces open-source computational fluid dynamics software named open computational fluid dynamic code for scientific computation with graphics processing unit (GPU) system (OpenCFD-SCU), developed by the authors for direct numerical simulation (DNS) of compressible wall-bounded turbulence. This software is based on the finite difference method and is accelerated by the use of a GPU, which provides an acceleration by a factor of more than 200 compared with central processing unit (CPU) software based on the same algorithm and number of message passing interface (MPI) processes, and the running speed of OpenCFD-SCU with just 512 GPUs exceed that of CPU software with 130\\,000 CPUs. GPU-Stream technology is used to implement overlap of computing and communication, achieving 98.7\\% parallel weak scalability with 24\\,576 GPUs. The software includes a variety of high-precision finite difference schemes, and supports a hybrid finite difference scheme, enabling it to provide both robustness and high precision when simulating complex supersonic and hypersonic flows. When used with the wide range of supercomputers currently available, the software should able to improve the performance of large-scale simulations by up to two orders on the computational scale. Then, OpenCFD-SCU is applied to a validation and verification case of a Mach 2.9 compression ramp with mesh numbers up to 31.2 billion. More challenging cases using hybrid finite schemes are shown in Part 2(Dang, Li et al. 2022). The code is available and supported at \\url{http://developer.hpccube.com/codes/danggl/opencfd-scu.git}."
                    }
                ]
            },
            "e6c5e46b-90d0-43f3-a5ad-78129e36bc2e": {
                "pk": "e6c5e46b-90d0-43f3-a5ad-78129e36bc2e",
                "name": "Lu Yin",
                "collaborators": [
                    "Da Huang"
                ],
                "domain": [
                    "Cosmology",
                    "Dark Energy",
                    "Gravitational Waves",
                    "Astrophysics"
                ],
                "publications": [
                    {
                        "title": "Reducing the $H_0$ Tension with Exponential Acoustic Dark Energy",
                        "abstract": "The Hubble tension arises from different observations between the late-time and early Universe. We explore a new model with dark fluid, called the exponential Acoustic Dark Energy (eADE) model, to relieve the Hubble tension. The eADE model gives an exponential form of the equation of state (EoS) in the acoustic dark energy, which is the first time to explore an exponential form for the EoS. In this model, the gravitational effects from the acoustic oscillations of the model can impact the CMB phenomena at the matter radiation equally epoch. We give the constraints of the eADE model by the current cosmological dataset. The comparison of the phenomena with the standard model can be shown through CMB and matter power spectra. The fitting results of our model have $H_0 = 70.06^{+1.13}_{-1.09}$ in 95$\\%$ C.L. and a smaller best-fit value than $\\Lambda$CDM."
                    },
                    {
                        "title": "Cosmic-Dawn 21-cm Signal from Dynamical Dark Energy",
                        "abstract": "The 21-cm signal is the most important measurement for us to understand physics during cosmic dawn. It is the key for us to understand the expansion history of the Universe and the nature of dark energy. In this paper, we focused on the characteristic 21-cm power spectrum of a special dynamic dark energy - the Interacting Chevallier-Polarski-Linder (ICPL) model - and compared it with those of the $\\Lambda$CDM and CPL models. From the expected noise of HERA, we found more precise experiments in the future can detect the features of interacting dark energy in the 21-cm power spectra. By studying the brightness temperature, we found the ICPL model is closer to the observation of EDGES compared to the $\\Lambda$CDM, thus alleviating the tension between theory and experiments."
                    },
                    {
                        "title": "Does Gauss-Bonnet Inflationary Gravitational Waves satisfy the Pulsar Timing Arrays observations?",
                        "abstract": "The observations from pulsar timing arrays (PTAs), led by the North American Nanohertz Observatory for Gravitational Waves (NANOGrav), have provided opportunities to constrain primordial gravitational waves at low frequencies. In this paper, we analyze the best-fit parameter values for different Gauss-Bonnet Inflationary Gravitational Wave (GB-IGW) models with the PTArcade program, and we compare the results with the observations of NANOGrav, European Pulsar Timing Array (EPTA), Parkes Pulsar Timing Array (PPTA), and International Pulsar Timing Array (IPTA). We find the potential parameter $n$ derived from the GB-IGW model is not necessarily positive. Instead, PTA data suggest the possibility of new parameter ranges. Meanwhile, the other parameters' results also have difference with traditional Cosmic Microwave Background analyses, shows the reason of different GW power spectra expected. Additionally, the GB-IGW models we calculated are closer to the central values of multiple PTA datasets compared to the standard inflation model, making the GB-IGW model more likely to be detected by future space-based gravitational wave observatories."
                    },
                    {
                        "title": "Stochastic Gravitational Waves from Inflaton Decays",
                        "abstract": "Due to the universality of gravitational interactions, it is generally expected that a stochastic gravitational wave (GW) background could form during the reheating period when the inflaton perturbatively decays with the emission of gravitons. Previously, only models in which the inflaton dominantly decays into a pair of light scalar and/or fermion particles were considered in the literature. In the present paper, we focus on the cases with a vector particle pair in the final decay product. The differential decay rates for the three-body gravitational inflaton decays are presented for two typical couplings between the inflaton and vector fields, from which we predict their respective GW frequency spectra. It turns out that, similar to the scalar and fermion cases, the obtained GW spectra is too high in frequency to be observed by the current and near-future GW detection experiments and calls for a new design of high-frequency GW detectors."
                    }
                ]
            },
            "dd894d40-1246-4a1b-b785-7dc8d80ad429": {
                "pk": "dd894d40-1246-4a1b-b785-7dc8d80ad429",
                "name": "Mykola Pechenizkiy",
                "collaborators": [
                    "Vlado Menkovski",
                    "Pratik Gajane",
                    "Danil Provodin",
                    "Maurits Kaptein",
                    "Shiwei Liu",
                    "Akrati Saxena",
                    "Masoud Mansoury",
                    "Iftitahu Ni'mah",
                    "Decebal Constantin Mocanu",
                    "George Fletcher"
                ],
                "domain": [
                    "Machine Learning",
                    "Fairness",
                    "Sentiment Analysis",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Rule-based Emotion Detection on Social Media: Putting Tweets on Plutchik's Wheel",
                        "abstract": "We study sentiment analysis beyond the typical granularity of polarity and instead use Plutchik's wheel of emotions model. We introduce RBEM-Emo as an extension to the Rule-Based Emission Model algorithm to deduce such emotions from human-written messages. We evaluate our approach on two different datasets and compare its performance with the current state-of-the-art techniques for emotion detection, including a recursive auto-encoder. The results of the experimental study suggest that RBEM-Emo is a promising approach advancing the current state-of-the-art in emotion detection."
                    },
                    {
                        "title": "On Formalizing Fairness in Prediction with Machine Learning",
                        "abstract": "Machine learning algorithms for prediction are increasingly being used in critical decisions affecting human lives. Various fairness formalizations, with no firm consensus yet, are employed to prevent such algorithms from systematically discriminating against people based on certain attributes protected by law. The aim of this article is to survey how fairness is formalized in the machine learning literature for the task of prediction and present these formalizations with their corresponding notions of distributive justice from the social sciences literature. We provide theoretical as well as empirical critiques of these notions from the social sciences literature and explain how these critiques limit the suitability of the corresponding fairness formalizations to certain domains. We also suggest two notions of distributive justice which address some of these critiques and discuss avenues for prospective fairness formalizations."
                    },
                    {
                        "title": "Predictive User Modeling with Actionable Attributes",
                        "abstract": "Different machine learning techniques have been proposed and used for modeling individual and group user needs, interests and preferences. In the traditional predictive modeling instances are described by observable variables, called attributes. The goal is to learn a model for predicting the target variable for unseen instances. For example, for marketing purposes a company consider profiling a new user based on her observed web browsing behavior, referral keywords or other relevant information. In many real world applications the values of some attributes are not only observable, but can be actively decided by a decision maker. Furthermore, in some of such applications the decision maker is interested not only to generate accurate predictions, but to maximize the probability of the desired outcome. For example, a direct marketing manager can choose which type of a special offer to send to a client (actionable attribute), hoping that the right choice will result in a positive response with a higher probability. We study how to learn to choose the value of an actionable attribute in order to maximize the probability of a desired outcome in predictive modeling. We emphasize that not all instances are equally sensitive to changes in actions. Accurate choice of an action is critical for those instances, which are on the borderline (e.g. users who do not have a strong opinion one way or the other). We formulate three supervised learning approaches for learning to select the value of an actionable attribute at an instance level. We also introduce a focused training procedure which puts more emphasis on the situations where varying the action is the most likely to take the effect. The proof of concept experimental validation on two real-world case studies in web analytics and e-learning domains highlights the potential of the proposed approaches."
                    },
                    {
                        "title": "Controversy Rules - Discovering Regions Where Classifiers (Dis-)Agree Exceptionally",
                        "abstract": "Finding regions for which there is higher controversy among different classifiers is insightful with regards to the domain and our models. Such evaluation can falsify assumptions, assert some, or also, bring to the attention unknown phenomena. The present work describes an algorithm, which is based on the Exceptional Model Mining framework, and enables that kind of investigations. We explore several public datasets and show the usefulness of this approach in classification tasks. We show in this paper a few interesting observations about those well explored datasets, some of which are general knowledge, and other that as far as we know, were not reported before."
                    },
                    {
                        "title": "Bias Disparity in Collaborative Recommendation: Algorithmic Evaluation and Comparison",
                        "abstract": "Research on fairness in machine learning has been recently extended to recommender systems. One of the factors that may impact fairness is bias disparity, the degree to which a group's preferences on various item categories fail to be reflected in the recommendations they receive. In some cases biases in the original data may be amplified or reversed by the underlying recommendation algorithm. In this paper, we explore how different recommendation algorithms reflect the tradeoff between ranking quality and bias disparity. Our experiments include neighborhood-based, model-based, and trust-aware recommendation algorithms."
                    },
                    {
                        "title": "BSDAR: Beam Search Decoding with Attention Reward in Neural Keyphrase Generation",
                        "abstract": "This study mainly investigates two common decoding problems in neural keyphrase generation: sequence length bias and beam diversity. To tackle the problems, we introduce a beam search decoding strategy based on word-level and ngram-level reward function to constrain and refine Seq2Seq inference at test time. Results show that our simple proposal can overcome the algorithm bias to shorter and nearly identical sequences, resulting in a significant improvement of the decoding performance on generating keyphrases that are present and absent in source text."
                    },
                    {
                        "title": "An Empirical Evaluation of Posterior Sampling for Constrained Reinforcement Learning",
                        "abstract": "We study a posterior sampling approach to efficient exploration in constrained reinforcement learning. Alternatively to existing algorithms, we propose two simple algorithms that are more efficient statistically, simpler to implement and computationally cheaper. The first algorithm is based on a linear formulation of CMDP, and the second algorithm leverages the saddle-point formulation of CMDP. Our empirical results demonstrate that, despite its simplicity, posterior sampling achieves state-of-the-art performance and, in some cases, significantly outperforms optimistic algorithms."
                    },
                    {
                        "title": "On improving deep learning generalization with adaptive sparse connectivity",
                        "abstract": "Large neural networks are very successful in various tasks. However, with limited data, the generalization capabilities of deep neural networks are also very limited. In this paper, we empirically start showing that intrinsically sparse neural networks with adaptive sparse connectivity, which by design have a strict parameter budget during the training phase, have better generalization capabilities than their fully-connected counterparts. Besides this, we propose a new technique to train these sparse models by combining the Sparse Evolutionary Training (SET) procedure with neurons pruning. Operated on MultiLayer Perceptron (MLP) and tested on 15 datasets, our proposed technique zeros out around 50% of the hidden neurons during training, while having a linear number of parameters to optimize with respect to the number of neurons. The results show a competitive classification and generalization performance."
                    },
                    {
                        "title": "ProtoInfoMax: Prototypical Networks with Mutual Information Maximization for Out-of-Domain Detection",
                        "abstract": "The ability to detect Out-of-Domain (OOD) inputs has been a critical requirement in many real-world NLP applications. For example, intent classification in dialogue systems. The reason is that the inclusion of unsupported OOD inputs may lead to catastrophic failure of systems. However, it remains an empirical question whether current methods can tackle such problems reliably in a realistic scenario where zero OOD training data is available. In this study, we propose ProtoInfoMax, a new architecture that extends Prototypical Networks to simultaneously process in-domain and OOD sentences via Mutual Information Maximization (InfoMax) objective. Experimental results show that our proposed method can substantially improve performance up to 20% for OOD detection in low resource settings of text classification. We also show that ProtoInfoMax is less prone to typical overconfidence errors of Neural Networks, leading to more reliable prediction results."
                    },
                    {
                        "title": "The Relationship between the Consistency of Users' Ratings and Recommendation Calibration",
                        "abstract": "Fairness in recommender systems has recently received attention from researchers. Unfair recommendations have negative impact on the effectiveness of recommender systems as it may degrade users' satisfaction, loyalty, and at worst, it can lead to or perpetuate undesirable social dynamics. One of the factors that may impact fairness is calibration, the degree to which users' preferences on various item categories are reflected in the recommendations they receive.   The ability of a recommendation algorithm for generating effective recommendations may depend on the meaningfulness of the input data and the amount of information available in users' profile. In this paper, we aim to explore the relationship between the consistency of users' ratings behavior and the degree of calibrated recommendations they receive. We conduct our analysis on different groups of users based on the consistency of their ratings. Our experimental results on a movie dataset and several recommendation algorithms show that there is a positive correlation between the consistency of users' ratings behavior and the degree of calibration in their recommendations, meaning that user groups with higher inconsistency in their ratings receive less calibrated recommendations."
                    },
                    {
                        "title": "How Fair is Fairness-aware Representative Ranking and Methods for Fair Ranking",
                        "abstract": "Rankings of people and items has been highly used in selection-making, match-making, and recommendation algorithms that have been deployed on ranging of platforms from employment websites to searching tools. The ranking position of a candidate affects the amount of opportunities received by the ranked candidate. It has been observed in several works that the ranking of candidates based on their score can be biased for candidates belonging to the minority community. In recent works, the fairness-aware representative ranking was proposed for computing fairness-aware re-ranking of results. The proposed algorithm achieves the desired distribution of top-ranked results with respect to one or more protected attributes. In this work, we highlight the bias in fairness-aware representative ranking for an individual as well as for a group if the group is sub-active on the platform. We define individual unfairness and group unfairness and propose methods to generate ideal individual and group fair representative ranking if the universal representation ratio is known or unknown. The simulation results show the quantified analysis of fairness in the proposed solutions. The paper is concluded with open challenges and further directions."
                    },
                    {
                        "title": "The Impact of Batch Learning in Stochastic Linear Bandits",
                        "abstract": "We consider a special case of bandit problems, named batched bandits, in which an agent observes batches of responses over a certain time period. Unlike previous work, we consider a more practically relevant batch-centric scenario of batch learning. That is to say, we provide a policy-agnostic regret analysis and demonstrate upper and lower bounds for the regret of a candidate policy. Our main theoretical results show that the impact of batch learning is a multiplicative factor of batch size relative to the regret of online behavior. Primarily, we study two settings of the stochastic linear bandits: bandits with finitely and infinitely many arms. While the regret bounds are the same for both settings, the former setting results hold under milder assumptions. Also, we provide a more robust result for the 2-armed bandit problem as an important insight. Finally, we demonstrate the consistency of theoretical results by conducting empirical experiments and reflect on optimal batch size choice."
                    },
                    {
                        "title": "FAL-CUR: Fair Active Learning using Uncertainty and Representativeness on Fair Clustering",
                        "abstract": "Active Learning (AL) techniques have proven to be highly effective in reducing data labeling costs across a range of machine learning tasks. Nevertheless, one known challenge of these methods is their potential to introduce unfairness towards sensitive attributes. Although recent approaches have focused on enhancing fairness in AL, they tend to reduce the model's accuracy. To address this issue, we propose a novel strategy, named Fair Active Learning using fair Clustering, Uncertainty, and Representativeness (FAL-CUR), to improve fairness in AL. FAL-CUR tackles the fairness problem in AL by combining fair clustering with an acquisition function that determines which samples to query based on their uncertainty and representativeness scores. We evaluate the performance of FAL-CUR on four real-world datasets, and the results demonstrate that FAL-CUR achieves a 15% - 20% improvement in fairness compared to the best state-of-the-art method in terms of equalized odds while maintaining stable accuracy scores. Furthermore, an ablation study highlights the crucial roles of fair clustering in preserving fairness and the acquisition function in stabilizing the accuracy performance."
                    },
                    {
                        "title": "Efficient Exploration in Average-Reward Constrained Reinforcement Learning: Achieving Near-Optimal Regret With Posterior Sampling",
                        "abstract": "We present a new algorithm based on posterior sampling for learning in Constrained Markov Decision Processes (CMDP) in the infinite-horizon undiscounted setting. The algorithm achieves near-optimal regret bounds while being advantageous empirically compared to the existing algorithms. Our main theoretical result is a Bayesian regret bound for each cost component of $\\tilde{O} (DS\\sqrt{AT})$ for any communicating CMDP with $S$ states, $A$ actions, and diameter $D$. This regret bound matches the lower bound in order of time horizon $T$ and is the best-known regret bound for communicating CMDPs achieved by a computationally tractable algorithm. Empirical results show that our posterior sampling algorithm outperforms the existing algorithms for constrained reinforcement learning."
                    },
                    {
                        "title": "One-Shot Federated Learning with Bayesian Pseudocoresets",
                        "abstract": "Optimization-based techniques for federated learning (FL) often come with prohibitive communication cost, as high dimensional model parameters need to be communicated repeatedly between server and clients. In this paper, we follow a Bayesian approach allowing to perform FL with one-shot communication, by solving the global inference problem as a product of local client posteriors. For models with multi-modal likelihoods, such as neural networks, a naive application of this scheme is hampered, since clients will capture different posterior modes, causing a destructive collapse of the posterior on the server side. Consequently, we explore approximate inference in the function-space representation of client posteriors, hence suffering less or not at all from multi-modality. We show that distributed function-space inference is tightly related to learning Bayesian pseudocoresets and develop a tractable Bayesian FL algorithm on this insight. We show that this approach achieves prediction performance competitive to state-of-the-art while showing a striking reduction in communication cost of up to two orders of magnitude. Moreover, due to its Bayesian nature, our method also delivers well-calibrated uncertainty estimates."
                    },
                    {
                        "title": "Intrinsically Sparse Long Short-Term Memory Networks",
                        "abstract": "Long Short-Term Memory (LSTM) has achieved state-of-the-art performances on a wide range of tasks. Its outstanding performance is guaranteed by the long-term memory ability which matches the sequential data perfectly and the gating structure controlling the information flow. However, LSTMs are prone to be memory-bandwidth limited in realistic applications and need an unbearable period of training and inference time as the model size is ever-increasing. To tackle this problem, various efficient model compression methods have been proposed. Most of them need a big and expensive pre-trained model which is a nightmare for resource-limited devices where the memory budget is strictly limited. To remedy this situation, in this paper, we incorporate the Sparse Evolutionary Training (SET) procedure into LSTM, proposing a novel model dubbed SET-LSTM. Rather than starting with a fully-connected architecture, SET-LSTM has a sparse topology and dramatically fewer parameters in both phases, training and inference. Considering the specific architecture of LSTMs, we replace the LSTM cells and embedding layers with sparse structures and further on, use an evolutionary strategy to adapt the sparse connectivity to the data. Additionally, we find that SET-LSTM can provide many different good combinations of sparse connectivity to substitute the overparameterized optimization problem of dense neural networks. Evaluated on four sentiment analysis classification datasets, the results demonstrate that our proposed model is able to achieve usually better performance than its fully connected counterpart while having less than 4\\% of its parameters."
                    },
                    {
                        "title": "NodeSim: Node Similarity based Network Embedding for Diverse Link Prediction",
                        "abstract": "In real-world complex networks, understanding the dynamics of their evolution has been of great interest to the scientific community. Predicting future links is an essential task of social network analysis as the addition or removal of the links over time leads to the network evolution. In a network, links can be categorized as intra-community links if both end nodes of the link belong to the same community, otherwise inter-community links. The existing link-prediction methods have mainly focused on achieving high accuracy for intra-community link prediction. In this work, we propose a network embedding method, called NodeSim, which captures both similarities between the nodes and the community structure while learning the low-dimensional representation of the network. The embedding is learned using the proposed NodeSim random walk, which efficiently explores the diverse neighborhood while keeping the more similar nodes closer in the context of the node. We verify the efficacy of the proposed embedding method over state-of-the-art methods using diverse link prediction. We propose a machine learning model for link prediction that considers both the nodes' embedding and their community information to predict the link between two given nodes. Extensive experimental results on several real-world networks demonstrate the effectiveness of the proposed framework for both inter and intra-community link prediction."
                    },
                    {
                        "title": "Knowledge Elicitation using Deep Metric Learning and Psychometric Testing",
                        "abstract": "Knowledge present in a domain is well expressed as relationships between corresponding concepts. For example, in zoology, animal species form complex hierarchies; in genomics, the different (parts of) molecules are organized in groups and subgroups based on their functions; plants, molecules, and astronomical objects all form complex taxonomies. Nevertheless, when applying supervised machine learning (ML) in such domains, we commonly reduce the complex and rich knowledge to a fixed set of labels, and induce a model shows good generalization performance with respect to these labels. The main reason for such a reductionist approach is the difficulty in eliciting the domain knowledge from the experts. Developing a label structure with sufficient fidelity and providing comprehensive multi-label annotation can be exceedingly labor-intensive in many real-world applications. In this paper, we provide a method for efficient hierarchical knowledge elicitation (HKE) from experts working with high-dimensional data such as images or videos. Our method is based on psychometric testing and active deep metric learning. The developed models embed the high-dimensional data in a metric space where distances are semantically meaningful, and the data can be organized in a hierarchical structure. We provide empirical evidence with a series of experiments on a synthetically generated dataset of simple shapes, and Cifar 10 and Fashion-MNIST benchmarks that our method is indeed successful in uncovering hierarchical structures."
                    },
                    {
                        "title": "Hierarchical Semantic Segmentation using Psychometric Learning",
                        "abstract": "Assigning meaning to parts of image data is the goal of semantic image segmentation. Machine learning methods, specifically supervised learning is commonly used in a variety of tasks formulated as semantic segmentation. One of the major challenges in the supervised learning approaches is expressing and collecting the rich knowledge that experts have with respect to the meaning present in the image data. Towards this, typically a fixed set of labels is specified and experts are tasked with annotating the pixels, patches or segments in the images with the given labels. In general, however, the set of classes does not fully capture the rich semantic information present in the images. For example, in medical imaging such as histology images, the different parts of cells could be grouped and sub-grouped based on the expertise of the pathologist.   To achieve such a precise semantic representation of the concepts in the image, we need access to the full depth of knowledge of the annotator. In this work, we develop a novel approach to collect segmentation annotations from experts based on psychometric testing. Our method consists of the psychometric testing procedure, active query selection, query enhancement, and a deep metric learning model to achieve a patch-level image embedding that allows for semantic segmentation of images. We show the merits of our method with evaluation on the synthetically generated image, aerial image and histology image."
                    },
                    {
                        "title": "Bridging the Performance Gap between FGSM and PGD Adversarial Training",
                        "abstract": "Deep learning achieves state-of-the-art performance in many tasks but exposes to the underlying vulnerability against adversarial examples. Across existing defense techniques, adversarial training with the projected gradient decent attack (adv.PGD) is considered as one of the most effective ways to achieve moderate adversarial robustness. However, adv.PGD requires too much training time since the projected gradient attack (PGD) takes multiple iterations to generate perturbations. On the other hand, adversarial training with the fast gradient sign method (adv.FGSM) takes much less training time since the fast gradient sign method (FGSM) takes one step to generate perturbations but fails to increase adversarial robustness. In this work, we extend adv.FGSM to make it achieve the adversarial robustness of adv.PGD. We demonstrate that the large curvature along FGSM perturbed direction leads to a large difference in performance of adversarial robustness between adv.FGSM and adv.PGD, and therefore propose combining adv.FGSM with a curvature regularization (adv.FGSMR) in order to bridge the performance gap between adv.FGSM and adv.PGD. The experiments show that adv.FGSMR has higher training efficiency than adv.PGD. In addition, it achieves comparable performance of adversarial robustness on MNIST dataset under white-box attack, and it achieves better performance than adv.PGD under white-box attack and effectively defends the transferable adversarial attack on CIFAR-10 dataset."
                    }
                ]
            },
            "0e571055-85b6-4100-95fd-c4a5bec353ef": {
                "pk": "0e571055-85b6-4100-95fd-c4a5bec353ef",
                "name": "Decebal Constantin Mocanu",
                "collaborators": [
                    "Mykola Pechenizkiy",
                    "Ghada Sokar",
                    "Elena Mocanu",
                    "Antonio Liotta",
                    "Shiwei Liu",
                    "Phuong H. Nguyen",
                    "Madeleine Gibescu",
                    "Zahra Atashgahi",
                    "Raymond Veldhuis",
                    "Peter Stone"
                ],
                "domain": [
                    "Deep Learning",
                    "Continual Learning",
                    "Sparse Neural Networks",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "One-Shot Learning using Mixture of Variational Autoencoders: a Generalization Learning approach",
                        "abstract": "Deep learning, even if it is very successful nowadays, traditionally needs very large amounts of labeled data to perform excellent on the classification task. In an attempt to solve this problem, the one-shot learning paradigm, which makes use of just one labeled sample per class and prior knowledge, becomes increasingly important. In this paper, we propose a new one-shot learning method, dubbed MoVAE (Mixture of Variational AutoEncoders), to perform classification. Complementary to prior studies, MoVAE represents a shift of paradigm in comparison with the usual one-shot learning methods, as it does not use any prior knowledge. Instead, it starts from zero knowledge and one labeled sample per class. Afterward, by using unlabeled data and the generalization learning concept (in a way, more as humans do), it is capable to gradually improve by itself its performance. Even more, if there are no unlabeled data available MoVAE can still perform well in one-shot learning classification. We demonstrate empirically the efficiency of our proposed approach on three datasets, i.e. the handwritten digits (MNIST), fashion products (Fashion-MNIST), and handwritten characters (Omniglot), showing that MoVAE outperforms state-of-the-art one-shot learning algorithms."
                    },
                    {
                        "title": "On improving deep learning generalization with adaptive sparse connectivity",
                        "abstract": "Large neural networks are very successful in various tasks. However, with limited data, the generalization capabilities of deep neural networks are also very limited. In this paper, we empirically start showing that intrinsically sparse neural networks with adaptive sparse connectivity, which by design have a strict parameter budget during the training phase, have better generalization capabilities than their fully-connected counterparts. Besides this, we propose a new technique to train these sparse models by combining the Sparse Evolutionary Training (SET) procedure with neurons pruning. Operated on MultiLayer Perceptron (MLP) and tested on 15 datasets, our proposed technique zeros out around 50% of the hidden neurons during training, while having a linear number of parameters to optimize with respect to the number of neurons. The results show a competitive classification and generalization performance."
                    },
                    {
                        "title": "Truly Sparse Neural Networks at Scale",
                        "abstract": "Recently, sparse training methods have started to be established as a de facto approach for training and inference efficiency in artificial neural networks. Yet, this efficiency is just in theory. In practice, everyone uses a binary mask to simulate sparsity since the typical deep learning software and hardware are optimized for dense matrix operations. In this paper, we take an orthogonal approach, and we show that we can train truly sparse neural networks to harvest their full potential. To achieve this goal, we introduce three novel contributions, specially designed for sparse neural networks: (1) a parallel training algorithm and its corresponding sparse implementation from scratch, (2) an activation function with non-trainable parameters to favour the gradient flow, and (3) a hidden neurons importance metric to eliminate redundancies. All in one, we are able to break the record and to train the largest neural network ever trained in terms of representational power -- reaching the bat brain size. The results show that our approach has state-of-the-art performance while opening the path for an environmentally friendly artificial intelligence era."
                    },
                    {
                        "title": "Avoiding Forgetting and Allowing Forward Transfer in Continual Learning via Sparse Networks",
                        "abstract": "Using task-specific components within a neural network in continual learning (CL) is a compelling strategy to address the stability-plasticity dilemma in fixed-capacity models without access to past data. Current methods focus only on selecting a sub-network for a new task that reduces forgetting of past tasks. However, this selection could limit the forward transfer of relevant past knowledge that helps in future learning. Our study reveals that satisfying both objectives jointly is more challenging when a unified classifier is used for all classes of seen tasks-class-Incremental Learning (class-IL)-as it is prone to ambiguities between classes across tasks. Moreover, the challenge increases when the semantic similarity of classes across tasks increases. To address this challenge, we propose a new CL method, named AFAF, that aims to Avoid Forgetting and Allow Forward transfer in class-IL using fix-capacity models. AFAF allocates a sub-network that enables selective transfer of relevant knowledge to a new task while preserving past knowledge, reusing some of the previously allocated components to utilize the fixed-capacity, and addressing class-ambiguities when similarities exist. The experiments show the effectiveness of AFAF in providing models with multiple CL desirable properties, while outperforming state-of-the-art methods on various challenging benchmarks with different semantic similarities."
                    },
                    {
                        "title": "Intrinsically Sparse Long Short-Term Memory Networks",
                        "abstract": "Long Short-Term Memory (LSTM) has achieved state-of-the-art performances on a wide range of tasks. Its outstanding performance is guaranteed by the long-term memory ability which matches the sequential data perfectly and the gating structure controlling the information flow. However, LSTMs are prone to be memory-bandwidth limited in realistic applications and need an unbearable period of training and inference time as the model size is ever-increasing. To tackle this problem, various efficient model compression methods have been proposed. Most of them need a big and expensive pre-trained model which is a nightmare for resource-limited devices where the memory budget is strictly limited. To remedy this situation, in this paper, we incorporate the Sparse Evolutionary Training (SET) procedure into LSTM, proposing a novel model dubbed SET-LSTM. Rather than starting with a fully-connected architecture, SET-LSTM has a sparse topology and dramatically fewer parameters in both phases, training and inference. Considering the specific architecture of LSTMs, we replace the LSTM cells and embedding layers with sparse structures and further on, use an evolutionary strategy to adapt the sparse connectivity to the data. Additionally, we find that SET-LSTM can provide many different good combinations of sparse connectivity to substitute the overparameterized optimization problem of dense neural networks. Evaluated on four sentiment analysis classification datasets, the results demonstrate that our proposed model is able to achieve usually better performance than its fully connected counterpart while having less than 4\\% of its parameters."
                    },
                    {
                        "title": "Learning Invariant Representation for Continual Learning",
                        "abstract": "Continual learning aims to provide intelligent agents that are capable of learning continually a sequence of tasks, building on previously learned knowledge. A key challenge in this learning paradigm is catastrophically forgetting previously learned tasks when the agent faces a new one. Current rehearsal-based methods show their success in mitigating the catastrophic forgetting problem by replaying samples from previous tasks during learning a new one. However, these methods are infeasible when the data of previous tasks is not accessible. In this work, we propose a new pseudo-rehearsal-based method, named learning Invariant Representation for Continual Learning (IRCL), in which class-invariant representation is disentangled from a conditional generative model and jointly used with class-specific representation to learn the sequential tasks. Disentangling the shared invariant representation helps to learn continually a sequence of tasks, while being more robust to forgetting and having better knowledge transfer. We focus on class incremental learning where there is no knowledge about task identity during inference. We empirically evaluate our proposed method on two well-known benchmarks for continual learning: split MNIST and split Fashion MNIST. The experimental results show that our proposed method outperforms regularization-based methods by a big margin and is better than the state-of-the-art pseudo-rehearsal-based method. Finally, we analyze the role of the shared invariant representation in mitigating the forgetting problem especially when the number of replayed samples for each previous task is small."
                    },
                    {
                        "title": "SpaceNet: Make Free Space For Continual Learning",
                        "abstract": "The continual learning (CL) paradigm aims to enable neural networks to learn tasks continually in a sequential fashion. The fundamental challenge in this learning paradigm is catastrophic forgetting previously learned tasks when the model is optimized for a new task, especially when their data is not accessible. Current architectural-based methods aim at alleviating the catastrophic forgetting problem but at the expense of expanding the capacity of the model. Regularization-based methods maintain a fixed model capacity; however, previous studies showed the huge performance degradation of these methods when the task identity is not available during inference (e.g. class incremental learning scenario). In this work, we propose a novel architectural-based method referred as SpaceNet for class incremental learning scenario where we utilize the available fixed capacity of the model intelligently. SpaceNet trains sparse deep neural networks from scratch in an adaptive way that compresses the sparse connections of each task in a compact number of neurons. The adaptive training of the sparse connections results in sparse representations that reduce the interference between the tasks. Experimental results show the robustness of our proposed method against catastrophic forgetting old tasks and the efficiency of SpaceNet in utilizing the available capacity of the model, leaving space for more tasks to be learned. In particular, when SpaceNet is tested on the well-known benchmarks for CL: split MNIST, split Fashion-MNIST, and CIFAR-10/100, it outperforms regularization-based methods by a big performance gap. Moreover, it achieves better performance than architectural-based methods without model expansion and achieved comparable results with rehearsal-based methods, while offering a huge memory reduction."
                    },
                    {
                        "title": "Self-Attention Meta-Learner for Continual Learning",
                        "abstract": "Continual learning aims to provide intelligent agents capable of learning multiple tasks sequentially with neural networks. One of its main challenging, catastrophic forgetting, is caused by the neural networks non-optimal ability to learn in non-stationary distributions. In most settings of the current approaches, the agent starts from randomly initialized parameters and is optimized to master the current task regardless of the usefulness of the learned representation for future tasks. Moreover, each of the future tasks uses all the previously learned knowledge although parts of this knowledge might not be helpful for its learning. These cause interference among tasks, especially when the data of previous tasks is not accessible. In this paper, we propose a new method, named Self-Attention Meta-Learner (SAM), which learns a prior knowledge for continual learning that permits learning a sequence of tasks, while avoiding catastrophic forgetting. SAM incorporates an attention mechanism that learns to select the particular relevant representation for each future task. Each task builds a specific representation branch on top of the selected knowledge, avoiding the interference between tasks. We evaluate the proposed method on the Split CIFAR-10/100 and Split MNIST benchmarks in the task agnostic inference. We empirically show that we can achieve a better performance than several state-of-the-art methods for continual learning by building on the top of selected representation learned by SAM. We also show the role of the meta-attention mechanism in boosting informative features corresponding to the input data and identifying the correct target in the task agnostic inference. Finally, we demonstrate that popular existing continual learning methods gain a performance boost when they adopt SAM as a starting point."
                    },
                    {
                        "title": "A topological insight into restricted Boltzmann machines",
                        "abstract": "Restricted Boltzmann Machines (RBMs) and models derived from them have been successfully used as basic building blocks in deep artificial neural networks for automatic features extraction, unsupervised weights initialization, but also as density estimators. Thus, their generative and discriminative capabilities, but also their computational time are instrumental to a wide range of applications. Our main contribution is to look at RBMs from a topological perspective, bringing insights from network science. Firstly, here we show that RBMs and Gaussian RBMs (GRBMs) are bipartite graphs which naturally have a small-world topology. Secondly, we demonstrate both on synthetic and real-world datasets that by constraining RBMs and GRBMs to a scale-free topology (while still considering local neighborhoods and data distribution), we reduce the number of weights that need to be computed by a few orders of magnitude, at virtually no loss in generative performance. Thirdly, we show that, for a fixed number of weights, our proposed sparse models (which by design have a higher number of hidden neurons) achieve better generative capabilities than standard fully connected RBMs and GRBMs (which by design have a smaller number of hidden neurons), at no additional computational costs."
                    },
                    {
                        "title": "Memory-free Online Change-point Detection: A Novel Neural Network Approach",
                        "abstract": "Change-point detection (CPD), which detects abrupt changes in the data distribution, is recognized as one of the most significant tasks in time series analysis. Despite the extensive literature on offline CPD, unsupervised online CPD still suffers from major challenges, including scalability, hyperparameter tuning, and learning constraints. To mitigate some of these challenges, in this paper, we propose a novel deep learning approach for unsupervised online CPD from multi-dimensional time series, named Adaptive LSTM-Autoencoder Change-Point Detection (ALACPD). ALACPD exploits an LSTM-autoencoder-based neural network to perform unsupervised online CPD. It continuously adapts to the incoming samples without keeping the previously received input, thus being memory-free. We perform an extensive evaluation on several real-world time series CPD benchmarks. We show that ALACPD, on average, ranks first among state-of-the-art CPD algorithms in terms of quality of the time series segmentation, and it is on par with the best performer in terms of the accuracy of the estimated change-points. The implementation of ALACPD is available online on Github\\footnote{\\url{https://github.com/zahraatashgahi/ALACPD}}."
                    },
                    {
                        "title": "Where to Pay Attention in Sparse Training for Feature Selection?",
                        "abstract": "A new line of research for feature selection based on neural networks has recently emerged. Despite its superiority to classical methods, it requires many training iterations to converge and detect informative features. The computational time becomes prohibitively long for datasets with a large number of samples or a very high dimensional feature space. In this paper, we present a new efficient unsupervised method for feature selection based on sparse autoencoders. In particular, we propose a new sparse training algorithm that optimizes a model's sparse topology during training to pay attention to informative features quickly. The attention-based adaptation of the sparse topology enables fast detection of informative features after a few training iterations. We performed extensive experiments on 10 datasets of different types, including image, speech, text, artificial, and biological. They cover a wide range of characteristics, such as low and high-dimensional feature spaces, and few and large training samples. Our proposed approach outperforms the state-of-the-art methods in terms of selecting informative features while reducing training iterations and computational costs substantially. Moreover, the experiments show the robustness of our method in extremely noisy environments."
                    },
                    {
                        "title": "Do We Actually Need Dense Over-Parameterization? In-Time Over-Parameterization in Sparse Training",
                        "abstract": "In this paper, we introduce a new perspective on training deep neural networks capable of state-of-the-art performance without the need for the expensive over-parameterization by proposing the concept of In-Time Over-Parameterization (ITOP) in sparse training. By starting from a random sparse network and continuously exploring sparse connectivities during training, we can perform an Over-Parameterization in the space-time manifold, closing the gap in the expressibility between sparse training and dense training. We further use ITOP to understand the underlying mechanism of Dynamic Sparse Training (DST) and indicate that the benefits of DST come from its ability to consider across time all possible parameters when searching for the optimal sparse connectivity. As long as there are sufficient parameters that have been reliably explored during training, DST can outperform the dense neural network by a large margin. We present a series of experiments to support our conjecture and achieve the state-of-the-art sparse training performance with ResNet-50 on ImageNet. More impressively, our method achieves dominant performance over the overparameterization-based sparse methods at extreme sparsity levels. When trained on CIFAR-100, our method can match the performance of the dense model even at an extreme sparsity (98%). Code can be found https://github.com/Shiweiliuiiiiiii/In-Time-Over-Parameterization."
                    },
                    {
                        "title": "Selfish Sparse RNN Training",
                        "abstract": "Sparse neural networks have been widely applied to reduce the computational demands of training and deploying over-parameterized deep neural networks. For inference acceleration, methods that discover a sparse network from a pre-trained dense network (dense-to-sparse training) work effectively. Recently, dynamic sparse training (DST) has been proposed to train sparse neural networks without pre-training a dense model (sparse-to-sparse training), so that the training process can also be accelerated. However, previous sparse-to-sparse methods mainly focus on Multilayer Perceptron Networks (MLPs) and Convolutional Neural Networks (CNNs), failing to match the performance of dense-to-sparse methods in the Recurrent Neural Networks (RNNs) setting. In this paper, we propose an approach to train intrinsically sparse RNNs with a fixed parameter count in one single run, without compromising performance. During training, we allow RNN layers to have a non-uniform redistribution across cell gates for better regularization. Further, we propose SNT-ASGD, a novel variant of the averaged stochastic gradient optimizer, which significantly improves the performance of all sparse training methods for RNNs. Using these strategies, we achieve state-of-the-art sparse training results, better than the dense-to-sparse methods, with various types of RNNs on Penn TreeBank and Wikitext-2 datasets. Our codes are available at https://github.com/Shiweiliuiiiiiii/Selfish-RNN."
                    },
                    {
                        "title": "Adaptive Sparsity Level during Training for Efficient Time Series Forecasting with Transformers",
                        "abstract": "Efficient time series forecasting has become critical for real-world applications, particularly with deep neural networks (DNNs). Efficiency in DNNs can be achieved through sparse connectivity and reducing the model size. However, finding the sparsity level automatically during training remains challenging due to the heterogeneity in the loss-sparsity tradeoffs across the datasets. In this paper, we propose \\enquote{\\textbf{P}runing with \\textbf{A}daptive \\textbf{S}parsity \\textbf{L}evel} (\\textbf{PALS}), to automatically seek a decent balance between loss and sparsity, all without the need for a predefined sparsity level. PALS draws inspiration from sparse training and during-training methods. It introduces the novel \"expand\" mechanism in training sparse neural networks, allowing the model to dynamically shrink, expand, or remain stable to find a proper sparsity level. In this paper, we focus on achieving efficiency in transformers known for their excellent time series forecasting performance but high computational cost. Nevertheless, PALS can be applied directly to any DNN. To this aim, we demonstrate its effectiveness also on the DLinear model. Experimental results on six benchmark datasets and five state-of-the-art (SOTA) transformer variants show that PALS substantially reduces model size while maintaining comparable performance to the dense model. More interestingly, PALS even outperforms the dense model, in \\textcolor{blue}{12} and \\textcolor{blue}{14} cases out of 30 cases in terms of MSE and MAE loss, respectively, while reducing \\textcolor{blue}{65\\%} parameter count and \\textcolor{blue}{63\\%} FLOPs on average. Our code and supplementary material are available on Github\\footnote{\\tiny \\url{https://github.com/zahraatashgahi/PALS}}."
                    },
                    {
                        "title": "Dynamic Sparse Training for Deep Reinforcement Learning",
                        "abstract": "Deep reinforcement learning (DRL) agents are trained through trial-and-error interactions with the environment. This leads to a long training time for dense neural networks to achieve good performance. Hence, prohibitive computation and memory resources are consumed. Recently, learning efficient DRL agents has received increasing attention. Yet, current methods focus on accelerating inference time. In this paper, we introduce for the first time a dynamic sparse training approach for deep reinforcement learning to accelerate the training process. The proposed approach trains a sparse neural network from scratch and dynamically adapts its topology to the changing data distribution during training. Experiments on continuous control tasks show that our dynamic sparse agents achieve higher performance than the equivalent dense methods, reduce the parameter count and floating-point operations (FLOPs) by 50%, and have a faster learning speed that enables reaching the performance of dense agents with 40-50% reduction in the training steps."
                    },
                    {
                        "title": "Scalable Training of Artificial Neural Networks with Adaptive Sparse Connectivity inspired by Network Science",
                        "abstract": "Through the success of deep learning in various domains, artificial neural networks are currently among the most used artificial intelligence methods. Taking inspiration from the network properties of biological neural networks (e.g. sparsity, scale-freeness), we argue that (contrary to general practice) artificial neural networks, too, should not have fully-connected layers. Here we propose sparse evolutionary training of artificial neural networks, an algorithm which evolves an initial sparse topology (Erd\\H{o}s-R\\'enyi random graph) of two consecutive layers of neurons into a scale-free topology, during learning. Our method replaces artificial neural networks fully-connected layers with sparse ones before training, reducing quadratically the number of parameters, with no decrease in accuracy. We demonstrate our claims on restricted Boltzmann machines, multi-layer perceptrons, and convolutional neural networks for unsupervised and supervised learning on 15 datasets. Our approach has the potential to enable artificial neural networks to scale up beyond what is currently possible."
                    },
                    {
                        "title": "On-line Building Energy Optimization using Deep Reinforcement Learning",
                        "abstract": "Unprecedented high volumes of data are becoming available with the growth of the advanced metering infrastructure. These are expected to benefit planning and operation of the future power system, and to help the customers transition from a passive to an active role. In this paper, we explore for the first time in the smart grid context the benefits of using Deep Reinforcement Learning, a hybrid type of methods that combines Reinforcement Learning with Deep Learning, to perform on-line optimization of schedules for building energy management systems. The learning procedure was explored using two methods, Deep Q-learning and Deep Policy Gradient, both of them being extended to perform multiple actions simultaneously. The proposed approach was validated on the large-scale Pecan Street Inc. database. This highly-dimensional database includes information about photovoltaic power generation, electric vehicles as well as buildings appliances. Moreover, these on-line energy scheduling strategies could be used to provide real-time feedback to consumers to encourage more efficient use of electricity."
                    },
                    {
                        "title": "Estimating 3D Trajectories from 2D Projections via Disjunctive Factored Four-Way Conditional Restricted Boltzmann Machines",
                        "abstract": "Estimation, recognition, and near-future prediction of 3D trajectories based on their two dimensional projections available from one camera source is an exceptionally difficult problem due to uncertainty in the trajectories and environment, high dimensionality of the specific trajectory states, lack of enough labeled data and so on. In this article, we propose a solution to solve this problem based on a novel deep learning model dubbed Disjunctive Factored Four-Way Conditional Restricted Boltzmann Machine (DFFW-CRBM). Our method improves state-of-the-art deep learning techniques for high dimensional time-series modeling by introducing a novel tensor factorization capable of driving forth order Boltzmann machines to considerably lower energy levels, at no computational costs. DFFW-CRBMs are capable of accurately estimating, recognizing, and performing near-future prediction of three-dimensional trajectories from their 2D projections while requiring limited amount of labeled data. We evaluate our method on both simulated and real-world data, showing its effectiveness in predicting and classifying complex ball trajectories and human activities."
                    },
                    {
                        "title": "Predictive No-Reference Assessment of Video Quality",
                        "abstract": "Among the various means to evaluate the quality of video streams, No-Reference (NR) methods have low computation and may be executed on thin clients. Thus, NR algorithms would be perfect candidates in cases of real-time quality assessment, automated quality control and, particularly, in adaptive mobile streaming. Yet, existing NR approaches are often inaccurate, in comparison to Full-Reference (FR) algorithms, especially under lossy network conditions. In this work, we present an NR method that combines machine learning with simple NR metrics to achieve a quality index comparably as accurate as the Video Quality Metric (VQM) Full-Reference algorithm. Our method is tested in an extensive dataset (960 videos), under lossy network conditions and considering nine different machine learning algorithms. Overall, we achieve an over 97% correlation with VQM, while allowing real-time assessment of video quality of experience in realistic streaming scenarios."
                    },
                    {
                        "title": "Novelty Producing Synaptic Plasticity",
                        "abstract": "A learning process with the plasticity property often requires reinforcement signals to guide the process. However, in some tasks (e.g. maze-navigation), it is very difficult (or impossible) to measure the performance of an agent (i.e. a fitness value) to provide reinforcements since the position of the goal is not known. This requires finding the correct behavior among a vast number of possible behaviors without having the knowledge of the reinforcement signals. In these cases, an exhaustive search may be needed. However, this might not be feasible especially when optimizing artificial neural networks in continuous domains. In this work, we introduce novelty producing synaptic plasticity (NPSP), where we evolve synaptic plasticity rules to produce as many novel behaviors as possible to find the behavior that can solve the problem. We evaluate the NPSP on maze-navigation on deceptive maze environments that require complex actions and the achievement of subgoals to complete. Our results show that the search heuristic used with the proposed NPSP is indeed capable of producing much more novel behaviors in comparison with a random search taken as baseline."
                    }
                ]
            },
            "12ea9858-4162-4079-9c39-8d32911cd9c9": {
                "pk": "12ea9858-4162-4079-9c39-8d32911cd9c9",
                "name": "Maurice Van Keulen",
                "collaborators": [
                    "Sylvio Barbon Junior",
                    "Paolo Ceravolo",
                    "Sven Groppe",
                    "Mustafa Jarrar",
                    "Samira Maghool",
                    "Florence S\u00e8des",
                    "Soror Sahri",
                    "Jesper Provoost",
                    "Luc Wismans",
                    "Sander Van der Drift"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Big Data Analytics",
                    "Machine Learning",
                    "Traffic Prediction"
                ],
                "publications": [
                    {
                        "title": "Are Large Language Models the New Interface for Data Pipelines?",
                        "abstract": "A Language Model is a term that encompasses various types of models designed to understand and generate human communication. Large Language Models (LLMs) have gained significant attention due to their ability to process text with human-like fluency and coherence, making them valuable for a wide range of data-related tasks fashioned as pipelines. The capabilities of LLMs in natural language understanding and generation, combined with their scalability, versatility, and state-of-the-art performance, enable innovative applications across various AI-related fields, including eXplainable Artificial Intelligence (XAI), Automated Machine Learning (AutoML), and Knowledge Graphs (KG). Furthermore, we believe these models can extract valuable insights and make data-driven decisions at scale, a practice commonly referred to as Big Data Analytics (BDA). In this position paper, we provide some discussions in the direction of unlocking synergies among these technologies, which can lead to more powerful and intelligent AI solutions, driving improvements in data pipelines across a wide range of applications and domains integrating humans, computers, and knowledge."
                    },
                    {
                        "title": "Short Term Prediction of Parking Area states Using Real Time Data and Machine Learning Techniques",
                        "abstract": "Public road authorities and private mobility service providers need information derived from the current and predicted traffic states to act upon the daily urban system and its spatial and temporal dynamics. In this research, a real-time parking area state (occupancy, in- and outflux) prediction model (up to 60 minutes ahead) has been developed using publicly available historic and real time data sources. Based on a case study in a real-life scenario in the city of Arnhem, a Neural Network-based approach outperforms a Random Forest-based one on all assessed performance measures, although the differences are small. Both are outperforming a naive seasonal random walk model. Although the performance degrades with increasing prediction horizon, the model shows a performance gain of over 150% at a prediction horizon of 60 minutes compared with the naive model. Furthermore, it is shown that predicting the in- and outflux is a far more difficult task (i.e. performance gains of 30%) which needs more training data, not based exclusively on occupancy rate. However, the performance of predicting in- and outflux is less sensitive to the prediction horizon. In addition, it is shown that real-time information of current occupancy rate is the independent variable with the highest contribution to the performance, although time, traffic flow and weather variables also deliver a significant contribution. During real-time deployment, the model performs three times better than the naive model on average. As a result, it can provide valuable information for proactive traffic management as well as mobility service providers."
                    }
                ]
            },
            "0c0525b7-9425-452e-8cb0-a76420326888": {
                "pk": "0c0525b7-9425-452e-8cb0-a76420326888",
                "name": "Elena Mocanu",
                "collaborators": [
                    "Decebal Constantin Mocanu",
                    "Mykola Pechenizkiy",
                    "Phuong H. Nguyen",
                    "Madeleine Gibescu",
                    "Ghada Sokar",
                    "Antonio Liotta",
                    "Peter Stone",
                    "Qiao Xiao",
                    "Boqian Wu",
                    "Shiwei Liu"
                ],
                "domain": [
                    "Deep Learning",
                    "One-Shot Learning",
                    "Reinforcement Learning",
                    "Sparse Neural Networks"
                ],
                "publications": [
                    {
                        "title": "One-Shot Learning using Mixture of Variational Autoencoders: a Generalization Learning approach",
                        "abstract": "Deep learning, even if it is very successful nowadays, traditionally needs very large amounts of labeled data to perform excellent on the classification task. In an attempt to solve this problem, the one-shot learning paradigm, which makes use of just one labeled sample per class and prior knowledge, becomes increasingly important. In this paper, we propose a new one-shot learning method, dubbed MoVAE (Mixture of Variational AutoEncoders), to perform classification. Complementary to prior studies, MoVAE represents a shift of paradigm in comparison with the usual one-shot learning methods, as it does not use any prior knowledge. Instead, it starts from zero knowledge and one labeled sample per class. Afterward, by using unlabeled data and the generalization learning concept (in a way, more as humans do), it is capable to gradually improve by itself its performance. Even more, if there are no unlabeled data available MoVAE can still perform well in one-shot learning classification. We demonstrate empirically the efficiency of our proposed approach on three datasets, i.e. the handwritten digits (MNIST), fashion products (Fashion-MNIST), and handwritten characters (Omniglot), showing that MoVAE outperforms state-of-the-art one-shot learning algorithms."
                    },
                    {
                        "title": "Energy Disaggregation for Real-Time Building Flexibility Detection",
                        "abstract": "Energy is a limited resource which has to be managed wisely, taking into account both supply-demand matching and capacity constraints in the distribution grid. One aspect of the smart energy management at the building level is given by the problem of real-time detection of flexible demand available. In this paper we propose the use of energy disaggregation techniques to perform this task. Firstly, we investigate the use of existing classification methods to perform energy disaggregation. A comparison is performed between four classifiers, namely Naive Bayes, k-Nearest Neighbors, Support Vector Machine and AdaBoost. Secondly, we propose the use of Restricted Boltzmann Machine to automatically perform feature extraction. The extracted features are then used as inputs to the four classifiers and consequently shown to improve their accuracy. The efficiency of our approach is demonstrated on a real database consisting of detailed appliance-level measurements with high temporal resolution, which has been used for energy disaggregation in previous studies, namely the REDD. The results show robustness and good generalization capabilities to newly presented buildings with at least 96% accuracy."
                    },
                    {
                        "title": "A topological insight into restricted Boltzmann machines",
                        "abstract": "Restricted Boltzmann Machines (RBMs) and models derived from them have been successfully used as basic building blocks in deep artificial neural networks for automatic features extraction, unsupervised weights initialization, but also as density estimators. Thus, their generative and discriminative capabilities, but also their computational time are instrumental to a wide range of applications. Our main contribution is to look at RBMs from a topological perspective, bringing insights from network science. Firstly, here we show that RBMs and Gaussian RBMs (GRBMs) are bipartite graphs which naturally have a small-world topology. Secondly, we demonstrate both on synthetic and real-world datasets that by constraining RBMs and GRBMs to a scale-free topology (while still considering local neighborhoods and data distribution), we reduce the number of weights that need to be computed by a few orders of magnitude, at virtually no loss in generative performance. Thirdly, we show that, for a fixed number of weights, our proposed sparse models (which by design have a higher number of hidden neurons) achieve better generative capabilities than standard fully connected RBMs and GRBMs (which by design have a smaller number of hidden neurons), at no additional computational costs."
                    },
                    {
                        "title": "Scalable Training of Artificial Neural Networks with Adaptive Sparse Connectivity inspired by Network Science",
                        "abstract": "Through the success of deep learning in various domains, artificial neural networks are currently among the most used artificial intelligence methods. Taking inspiration from the network properties of biological neural networks (e.g. sparsity, scale-freeness), we argue that (contrary to general practice) artificial neural networks, too, should not have fully-connected layers. Here we propose sparse evolutionary training of artificial neural networks, an algorithm which evolves an initial sparse topology (Erd\\H{o}s-R\\'enyi random graph) of two consecutive layers of neurons into a scale-free topology, during learning. Our method replaces artificial neural networks fully-connected layers with sparse ones before training, reducing quadratically the number of parameters, with no decrease in accuracy. We demonstrate our claims on restricted Boltzmann machines, multi-layer perceptrons, and convolutional neural networks for unsupervised and supervised learning on 15 datasets. Our approach has the potential to enable artificial neural networks to scale up beyond what is currently possible."
                    },
                    {
                        "title": "On-line Building Energy Optimization using Deep Reinforcement Learning",
                        "abstract": "Unprecedented high volumes of data are becoming available with the growth of the advanced metering infrastructure. These are expected to benefit planning and operation of the future power system, and to help the customers transition from a passive to an active role. In this paper, we explore for the first time in the smart grid context the benefits of using Deep Reinforcement Learning, a hybrid type of methods that combines Reinforcement Learning with Deep Learning, to perform on-line optimization of schedules for building energy management systems. The learning procedure was explored using two methods, Deep Q-learning and Deep Policy Gradient, both of them being extended to perform multiple actions simultaneously. The proposed approach was validated on the large-scale Pecan Street Inc. database. This highly-dimensional database includes information about photovoltaic power generation, electric vehicles as well as buildings appliances. Moreover, these on-line energy scheduling strategies could be used to provide real-time feedback to consumers to encourage more efficient use of electricity."
                    },
                    {
                        "title": "Dynamic Sparse Training for Deep Reinforcement Learning",
                        "abstract": "Deep reinforcement learning (DRL) agents are trained through trial-and-error interactions with the environment. This leads to a long training time for dense neural networks to achieve good performance. Hence, prohibitive computation and memory resources are consumed. Recently, learning efficient DRL agents has received increasing attention. Yet, current methods focus on accelerating inference time. In this paper, we introduce for the first time a dynamic sparse training approach for deep reinforcement learning to accelerate the training process. The proposed approach trains a sparse neural network from scratch and dynamically adapts its topology to the changing data distribution during training. Experiments on continuous control tasks show that our dynamic sparse agents achieve higher performance than the equivalent dense methods, reduce the parameter count and floating-point operations (FLOPs) by 50%, and have a faster learning speed that enables reaching the performance of dense agents with 40-50% reduction in the training steps."
                    },
                    {
                        "title": "Dynamic Sparse Network for Time Series Classification: Learning What to \"see''",
                        "abstract": "The receptive field (RF), which determines the region of time series to be ``seen'' and used, is critical to improve the performance for time series classification (TSC). However, the variation of signal scales across and within time series data, makes it challenging to decide on proper RF sizes for TSC. In this paper, we propose a dynamic sparse network (DSN) with sparse connections for TSC, which can learn to cover various RF without cumbersome hyper-parameters tuning. The kernels in each sparse layer are sparse and can be explored under the constraint regions by dynamic sparse training, which makes it possible to reduce the resource cost. The experimental results show that the proposed DSN model can achieve state-of-art performance on both univariate and multivariate TSC datasets with less than 50\\% computational cost compared with recent baseline methods, opening the path towards more accurate resource-aware methods for time series analyses. Our code is publicly available at: https://github.com/QiaoXiao7282/DSN."
                    },
                    {
                        "title": "Sparse Training Theory for Scalable and Efficient Agents",
                        "abstract": "A fundamental task for artificial intelligence is learning. Deep Neural Networks have proven to cope perfectly with all learning paradigms, i.e. supervised, unsupervised, and reinforcement learning. Nevertheless, traditional deep learning approaches make use of cloud computing facilities and do not scale well to autonomous agents with low computational resources. Even in the cloud, they suffer from computational and memory limitations, and they cannot be used to model adequately large physical worlds for agents which assume networks with billions of neurons. These issues are addressed in the last few years by the emerging topic of sparse training, which trains sparse networks from scratch. This paper discusses sparse training state-of-the-art, its challenges and limitations while introducing a couple of new theoretical research directions which has the potential of alleviating sparse training limitations to push deep learning scalability well beyond its current boundaries. Nevertheless, the theoretical advancements impact in complex multi-agents settings is discussed from a real-world perspective, using the smart grid case study."
                    },
                    {
                        "title": "Quick and Robust Feature Selection: the Strength of Energy-efficient Sparse Training for Autoencoders",
                        "abstract": "Major complications arise from the recent increase in the amount of high-dimensional data, including high computational costs and memory requirements. Feature selection, which identifies the most relevant and informative attributes of a dataset, has been introduced as a solution to this problem. Most of the existing feature selection methods are computationally inefficient; inefficient algorithms lead to high energy consumption, which is not desirable for devices with limited computational and energy resources. In this paper, a novel and flexible method for unsupervised feature selection is proposed. This method, named QuickSelection, introduces the strength of the neuron in sparse neural networks as a criterion to measure the feature importance. This criterion, blended with sparsely connected denoising autoencoders trained with the sparse evolutionary training procedure, derives the importance of all input features simultaneously. We implement QuickSelection in a purely sparse manner as opposed to the typical approach of using a binary mask over connections to simulate sparsity. It results in a considerable speed increase and memory reduction. When tested on several benchmark datasets, including five low-dimensional and three high-dimensional datasets, the proposed method is able to achieve the best trade-off of classification and clustering accuracy, running time, and maximum memory usage, among widely used approaches for feature selection. Besides, our proposed method requires the least amount of energy among the state-of-the-art autoencoder-based feature selection methods."
                    },
                    {
                        "title": "Automatic Noise Filtering with Dynamic Sparse Training in Deep Reinforcement Learning",
                        "abstract": "Tomorrow's robots will need to distinguish useful information from noise when performing different tasks. A household robot for instance may continuously receive a plethora of information about the home, but needs to focus on just a small subset to successfully execute its current chore. Filtering distracting inputs that contain irrelevant data has received little attention in the reinforcement learning literature. To start resolving this, we formulate a problem setting in reinforcement learning called the $\\textit{extremely noisy environment}$ (ENE), where up to $99\\%$ of the input features are pure noise. Agents need to detect which features provide task-relevant information about the state of the environment. Consequently, we propose a new method termed $\\textit{Automatic Noise Filtering}$ (ANF), which uses the principles of dynamic sparse training in synergy with various deep reinforcement learning algorithms. The sparse input layer learns to focus its connectivity on task-relevant features, such that ANF-SAC and ANF-TD3 outperform standard SAC and TD3 by a large margin, while using up to $95\\%$ fewer weights. Furthermore, we devise a transfer learning setting for ENEs, by permuting all features of the environment after 1M timesteps to simulate the fact that other information sources can become relevant as the world evolves. Again, ANF surpasses the baselines in final performance and sample complexity. Our code is available at https://github.com/bramgrooten/automatic-noise-filtering"
                    },
                    {
                        "title": "Dynamic Sparse Training versus Dense Training: The Unexpected Winner in Image Corruption Robustness",
                        "abstract": "It is generally perceived that Dynamic Sparse Training opens the door to a new era of scalability and efficiency for artificial neural networks at, perhaps, some costs in accuracy performance for the classification task. At the same time, Dense Training is widely accepted as being the \"de facto\" approach to train artificial neural networks if one would like to maximize their robustness against image corruption. In this paper, we question this general practice. Consequently, we claim that, contrary to what is commonly thought, the Dynamic Sparse Training methods can consistently outperform Dense Training in terms of robustness accuracy, particularly if the efficiency aspect is not considered as a main objective (i.e., sparsity levels between 10% and up to 50%), without adding (or even reducing) resource cost. We validate our claim on two types of data, images and videos, using several traditional and modern deep learning architectures for computer vision and three widely studied Dynamic Sparse Training algorithms. Our findings reveal a new yet-unknown benefit of Dynamic Sparse Training and open new possibilities in improving deep learning robustness beyond the current state of the art."
                    },
                    {
                        "title": "Deep Ensembling with No Overhead for either Training or Testing: The All-Round Blessings of Dynamic Sparsity",
                        "abstract": "The success of deep ensembles on improving predictive performance, uncertainty estimation, and out-of-distribution robustness has been extensively studied in the machine learning literature. Albeit the promising results, naively training multiple deep neural networks and combining their predictions at inference leads to prohibitive computational costs and memory requirements. Recently proposed efficient ensemble approaches reach the performance of the traditional deep ensembles with significantly lower costs. However, the training resources required by these approaches are still at least the same as training a single dense model. In this work, we draw a unique connection between sparse neural network training and deep ensembles, yielding a novel efficient ensemble learning framework called FreeTickets. Instead of training multiple dense networks and averaging them, we directly train sparse subnetworks from scratch and extract diverse yet accurate subnetworks during this efficient, sparse-to-sparse training. Our framework, FreeTickets, is defined as the ensemble of these relatively cheap sparse subnetworks. Despite being an ensemble method, FreeTickets has even fewer parameters and training FLOPs than a single dense model. This seemingly counter-intuitive outcome is due to the ultra training/inference efficiency of dynamic sparse training. FreeTickets surpasses the dense baseline in all the following criteria: prediction accuracy, uncertainty estimation, out-of-distribution (OoD) robustness, as well as efficiency for both training and inference. Impressively, FreeTickets outperforms the naive deep ensemble with ResNet50 on ImageNet using around only 1/5 of the training FLOPs required by the latter. We have released our source code at https://github.com/VITA-Group/FreeTickets."
                    }
                ]
            }
        }
    },
    "2402.07437": {
        "paper_data": {
            "title": "Learning Optimal Tax Design in Nonatomic Congestion Games",
            "url": "http://arxiv.org/abs/2402.07437v1",
            "arxiv_id": "2402.07437",
            "authors": [
                "Qiwen Cui",
                "Maryam Fazel",
                "Simon S. Du"
            ],
            "abstract": "We study how to learn the optimal tax design to maximize the efficiency in nonatomic congestion games. It is known that self-interested behavior among the players can damage the system's efficiency. Tax mechanisms is a common method to alleviate this issue and induce socially optimal behavior. In this work, we take the initial step for learning the optimal tax that can minimize the social cost with \\emph{equilibrium feedback}, i.e., the tax designer can only observe the equilibrium state under the enforced tax. Existing algorithms are not applicable due to the exponentially large tax function space, nonexistence of the gradient, and nonconvexity of the objective. To tackle these challenges, our algorithm leverages several novel components: (1) piece-wise linear tax to approximate the optimal tax; (2) an extra linear term to guarantee a strongly convex potential function; (3) efficient subroutine to find the ``boundary'' tax. The algorithm can find an $\\epsilon$-optimal tax with $O(\\beta F^2/\\epsilon)$ sample complexity, where $\\beta$ is the smoothness of the cost function and $F$ is the number of facilities.",
            "introduction": " Introduction to nonparametric regression . John Wiley & Sons, 2005. Heinrich Von Stackelberg. Market structure and equilibrium . Springer Science & Business Media, 2010. John Glen Wardrop. Road paper. some theoretical aspects of r oad tra\ufb03c research. Proceedings of the institution of civil engineers , 1(3):325\u2013362, 1952. Geng Zhao, Banghua Zhu, Jiantao Jiao, and Michael Jordan. On line learning in stackelberg games with an omniscient follower. In International Conference on Machine Learning , pages 42304\u201342316. PMLR, 2023. Han Zhong, Zhuoran Yang, Zhaoran Wang, and Michael I Jordan. Can reinforcement learning \ufb01nd stackelberg-nash equilibria in general-sum markov games w ith myopic followers? arXiv preprint arXiv:2112.13521 , 2021. 13A Basics about Congestion Games Lemma 4. If strategy xis an\u01eb-NE in a congestion game, then xis an\u01eb-minimizer of the corresponding potential function \u03a6(\u00b7). Proof. Letx\u2217= argminx\u2208X\u03a6(x)andy=\u03c6(x). First, we show that \u2207i,a\u03a6(x) =\u2207i,a/summationdisplay f/integraldisplayyf 0cf(u)du =\u2207i,a/summationdisplay f\u2208a/integraldisplayyf 0cf(u)du =/summationdisplay f\u2208acf(yf)\u2207i,ayf =/summationdisplay f\u2208acf(yf)\u2207i,a/summationdisplay i\u2032,a\u2032:f\u2208a\u2032xi\u2032,a\u2032 =/summationdisplay f\u2208acf(yf). Then we have \u03a6(x)\u2212\u03a6(x\u2217)\u2264 /a\\}b\u2207acketle{tx\u2212x\u2217,\u2207\u03a6(x)/a\\}b\u2207acket\u2207i}ht (Convexity) \u2264/summationdisplay i\u2208[m]/a\\}b\u2207acketle{txi\u2212x\u2217 i,\u2207i\u03a6(x)/a\\}b\u2207acket\u2207i}ht \u2264/summationdisplay i\u2208[m]\uf8ee \uf8f0/summationdisplay a\u2208Aixi,a/summationdisplay f\u2208acf(yf)\u2212min a\u2208Aiwi/summationdisplay f\u2208acf(yf)\uf8f9 \uf8fb \u2264/summationdisplay i\u2208[m]/summationdisplay a\u2208Aixi,a\u01eb =/summationdisplay i\u2208[m]wi\u01eb =\u01eb. Lemma 1. Under Assumption 1,\u03a6repais convex. If y\u2217= argminy\u03a6repa(y), then for any x\u2208\u03c6\u22121(y\u2217),xis a Nash equilibrium. Proof. Fory1,y2\u2208 Y, we have \u03a6repa(y1)+\u03a6repa(y2)\u22122\u03a6repa(y1+y2 2) =/summationdisplay f\uf8ee \uf8f0/integraldisplayy1 f 0cf(u)du+/integraldisplayy2 f 0cf(u)du\u22122/integraldisplayy1 f+y2 f 2 0cf(u)du\uf8f9 \uf8fb. Now we show that/integraltexty1 f 0cf(u)du+/integraltexty2 f 0cf(u)du\u22122/integraltexty1 f+y2 f 2 0cf(u)duis nonnegative for all f\u2208 F. W.l.o.g., we assumey1 f\u2264y2 fand we have /integraldisplayy1 f 0cf(u)du+/integraldisplayy2 f 0cf(u)du\u22122/integraldisplayy1 f+y2 f 2 0cf(u)du=/integraldisplayy2 f y1 f+y2 f 2cf(u)du\u2212/integraldisplayy1 f+y2 f 2 y1 fcf(u)du =/integraldisplayy1 f+y2 f 2 y1 f/bracketleftBigg cf(u+y2 f\u2212y1 f 2)\u2212cf(u)/bracketrightBigg du 14\u22650, where the last step is from Assumption 1(monotonicity). As a result, \u03a6repais convex. Lety\u2217= argminy\u03a6repa(y)andx\u2208\u03c6\u22121(y\u2217). If there exists x\u2032\u2208 Xsuch that \u03a6(x\u2032)<\u03a6(x), then we have\u03a6repa(\u03c6(x\u2032))<\u03a6repa(y\u2217), which contradicts the de\ufb01nition of y\u2217. As a result, xis the minimizer of \u03a6(\u00b7), which means xis a Nash equilibrium. Lemma 5. The Nash load under tax \u03c4:y(\u03c4) = argminy\u2208Y\u03a6(y;\u03c4)is not continuous w.r.t. \u03c4. In addition, the social welfare \u03a8(y(\u03c4))is not convex w.r.t. \u03c4. Proof. For the \ufb01rst part, we construct a congestion game with two fac ilitiesf1,f2, one commodity with action set {f1,f2}, and constant cost c1= 1,c2= 1\u2212\u01ebwith\u01eb >0. Then for tax \u03c4= 0, we have y(\u03c4) = [0,1]. For constant tax \u03c41= 0,\u03c42= 2\u01eb, we have y(\u03c4) = [1,0]. As\u01ebcan be arbitrarily small, y(\u03c4)is not continuous w.r.t.\u03c4. For the second part, we construct a congestion game with two f acilitiesf1,f2, one commodity with action set{f1,f2}, and cost function c1= 1,c2(u) =\u221auforu\u2208[0,1]. We apply constant tax \u03c4:\u03c41=t,\u03c42= 0 fort\u2208[\u22121,0]. The Nash equilibrium under tax \u03c4isy(\u03c4) = [1\u2212(1+t)2,(1+t)2]. Then the social cost is \u03a8(y(\u03c4)) = 1\u2212t(1+t)2, which is not convex on [\u22121,0]. B Missing Proofs in Section 5 Lemma 2. If the subgradient of the cost function cfis lower bounded by \u01eb >0for allf\u2208 F, then the potential function \u03a6repa(y)is\u01eb-strongly convex. However, \u03a6(x)is not necessarily strongly convex. Proof. First, by the de\ufb01nition of the potential function \u03a6, it is easy to show that \u2207\u03a6(y) = [cf(yf)]f\u2208F. For y1,y2\u2208 Y, we have (\u2207\u03a6(y1)\u2212\u2207\u03a6(y2))\u22a4(y1\u2212y2) =/summationdisplay f\u2208F(cf(y1 f)\u2212cf(y2 f))(y1 f\u2212y2 f)\u2265/summationdisplay f\u2208F\u01eb(y1 f\u2212y2 f)2=\u01eb/vextenddouble/vextenddoubley1\u2212y2/vextenddouble/vextenddouble2 2, which implies \u03a6(\u00b7)is a\u01eb-strongly convex function. For the second argument, we only need to construct a congesti on game such that there exists two strategy x1,x2\u2208 Xsuch that for \u03c6(tx1+(1\u2212t)x2)is a constant for t\u2208[0,1], which implies the potential function \u03a6(tx1+(1\u2212t)x2) = \u03a6repa(\u03c6(tx1+(1\u2212t)x2))is a constant w.r.t. t. However, a strongly convex function cannot be a constant on a line, which implies \u03a6is not strongly convex. We construct a congestion game with three facilities f1,f2,f3, three actions a1={f1},a2={f2},a3= {f3}and three commodities with action set {a1,a2},{a2,a3},{a3,a1}. Strategy x1:x1 1= [1,0,0],x1 2= [0,1,0],x1 3= [0,0,1]andx2:x2 1= [0,1,0],x2 2= [0,0,1],x2 3= [1,0,0]. Thentx1+ (1\u2212t)x2is a feasible strategy and we have \u03c6(tx1+(1\u2212t)x2) = [1,1,1]for allt\u2208[0,1]. Lemma 3. If two taxes \u03c4and\u02d9\u03c4only di\ufb00er in facility fand the Nash loads yand\u02d9yare di\ufb00erent, then yf/\\e}atio\\slash= \u02d9yf. Proof. For simplicity, we consider the equivalent tax-free case th at we have two costs c,\u02d9cwith subgradient lower bounded by \u01eband they only",
            "references": [
                {
                    "title": "Polynomial Convergence of Bandit No-Regret Dynamics in Congestion Games",
                    "abstract": "We introduce an online learning algorithm in the bandit feedback model that, once adopted by all agents of a congestion game, results in game-dynamics that converge to an $\\epsilon$-approximate Nash Equilibrium in a polynomial number of rounds with respect to $1/\\epsilon$, the number of players and the number of available resources. The proposed algorithm also guarantees sublinear regret to any agent adopting it. As a result, our work answers an open question from arXiv:2206.01880 and extends the recent results of arXiv:2306.15543 to the bandit feedback model. We additionally establish that our online learning algorithm can be implemented in polynomial time for the important special case of Network Congestion Games on Directed Acyclic Graphs (DAG) by constructing an exact $1$-barycentric spanner for DAGs."
                },
                {
                    "title": "Semi Bandit Dynamics in Congestion Games: Convergence to Nash Equilibrium and No-Regret Guarantees",
                    "abstract": "In this work, we introduce a new variant of online gradient descent, which provably converges to Nash Equilibria and simultaneously attains sublinear regret for the class of congestion games in the semi-bandit feedback setting. Our proposed method admits convergence rates depending only polynomially on the number of players and the number of facilities, but not on the size of the action set, which can be exponentially large in terms of the number of facilities. Moreover, the running time of our method has polynomial-time dependence on the implicit description of the game. As a result, our work answers an open question from (Du et. al, 2022)."
                },
                {
                    "title": "Taming the Exponential Action Set: Sublinear Regret and Fast Convergence to Nash Equilibrium in Online Congestion Games",
                    "abstract": "The congestion game is a powerful model that encompasses a range of engineering systems such as traffic networks and resource allocation. It describes the behavior of a group of agents who share a common set of $F$ facilities and take actions as subsets with $k$ facilities. In this work, we study the online formulation of congestion games, where agents participate in the game repeatedly and observe feedback with randomness. We propose CongestEXP, a decentralized algorithm that applies the classic exponential weights method. By maintaining weights on the facility level, the regret bound of CongestEXP avoids the exponential dependence on the size of possible facility sets, i.e., $\\binom{F}{k} \\approx F^k$, and scales only linearly with $F$. Specifically, we show that CongestEXP attains a regret upper bound of $O(kF\\sqrt{T})$ for every individual player, where $T$ is the time horizon. On the other hand, exploiting the exponential growth of weights enables CongestEXP to achieve a fast convergence rate. If a strict Nash equilibrium exists, we show that CongestEXP can converge to the strict Nash policy almost exponentially fast in $O(F\\exp(-t^{1-\\alpha}))$, where $t$ is the number of iterations and $\\alpha \\in (1/2, 1)$."
                },
                {
                    "title": "Online Learning in Stackelberg Games with an Omniscient Follower",
                    "abstract": "We study the problem of online learning in a two-player decentralized cooperative Stackelberg game. In each round, the leader first takes an action, followed by the follower who takes their action after observing the leader's move. The goal of the leader is to learn to minimize the cumulative regret based on the history of interactions. Differing from the traditional formulation of repeated Stackelberg games, we assume the follower is omniscient, with full knowledge of the true reward, and that they always best-respond to the leader's actions. We analyze the sample complexity of regret minimization in this repeated Stackelberg game. We show that depending on the reward structure, the existence of the omniscient follower may change the sample complexity drastically, from constant to exponential, even for linear cooperative Stackelberg games. This poses unique challenges for the learning process of the leader and the subsequent regret analysis."
                },
                {
                    "title": "Offline congestion games: How feedback type affects data coverage requirement",
                    "abstract": "This paper investigates when one can efficiently recover an approximate Nash Equilibrium (NE) in offline congestion games. The existing dataset coverage assumption in offline general-sum games inevitably incurs a dependency on the number of actions, which can be exponentially large in congestion games. We consider three different types of feedback with decreasing revealed information. Starting from the facility-level (a.k.a., semi-bandit) feedback, we propose a novel one-unit deviation coverage condition and give a pessimism-type algorithm that can recover an approximate NE. For the agent-level (a.k.a., bandit) feedback setting, interestingly, we show the one-unit deviation coverage condition is not sufficient. On the other hand, we convert the game to multi-agent linear bandits and show that with a generalized data coverage assumption in offline linear bandits, we can efficiently recover the approximate NE. Lastly, we consider a novel type of feedback, the game-level feedback where only the total reward from all agents is revealed. Again, we show the coverage assumption for the agent-level feedback setting is insufficient in the game-level feedback setting, and with a stronger version of the data coverage assumption for linear bandits, we can recover an approximate NE. Together, our results constitute the first study of offline congestion games and imply formal separations between different types of feedback."
                },
                {
                    "title": "Differentiable Bilevel Programming for Stackelberg Congestion Games",
                    "abstract": "In a Stackelberg congestion game (SCG), a leader aims to maximize their own gain by anticipating and manipulating the equilibrium state at which the followers settle by playing a congestion game. Often formulated as bilevel programs, large-scale SCGs are well known for their intractability and complexity. Here, we attempt to tackle this computational challenge by marrying traditional methodologies with the latest differentiable programming techniques in machine learning. The core idea centers on replacing the lower-level equilibrium problem with a smooth evolution trajectory defined by the imitative logit dynamic (ILD), which we prove converges to the equilibrium of the congestion game under mild conditions. Building upon this theoretical foundation, we propose two new local search algorithms for SCGs. The first is a gradient descent algorithm that obtains the derivatives by unrolling ILD via differentiable programming. Thanks to the smoothness of ILD, the algorithm promises both efficiency and scalability. The second algorithm adds a heuristic twist by cutting short the followers' evolution trajectory. Behaviorally, this means that, instead of anticipating the followers' best response at equilibrium, the leader seeks to approximate that response by only looking ahead a limited number of steps. Our numerical experiments are carried out over various instances of classic SCG applications, ranging from toy benchmarks to large-scale real-world examples. The results show the proposed algorithms are reliable and scalable local solvers that deliver high-quality solutions with greater regularity and significantly less computational effort compared to the many incumbents included in our study."
                },
                {
                    "title": "Learning in Congestion Games with Bandit Feedback",
                    "abstract": "In this paper, we investigate Nash-regret minimization in congestion games, a class of games with benign theoretical structure and broad real-world applications. We first propose a centralized algorithm based on the optimism in the face of uncertainty principle for congestion games with (semi-)bandit feedback, and obtain finite-sample guarantees. Then we propose a decentralized algorithm via a novel combination of the Frank-Wolfe method and G-optimal design. By exploiting the structure of the congestion game, we show the sample complexity of both algorithms depends only polynomially on the number of players and the number of facilities, but not the size of the action set, which can be exponentially large in terms of the number of facilities. We further define a new problem class, Markov congestion games, which allows us to model the non-stationarity in congestion games. We propose a centralized algorithm for Markov congestion games, whose sample complexity again has only polynomial dependence on all relevant problem parameters, but not the size of the action set."
                },
                {
                    "title": "Can Reinforcement Learning Find Stackelberg-Nash Equilibria in General-Sum Markov Games with Myopic Followers?",
                    "abstract": "We study multi-player general-sum Markov games with one of the players designated as the leader and the other players regarded as followers. In particular, we focus on the class of games where the followers are myopic, i.e., they aim to maximize their instantaneous rewards. For such a game, our goal is to find a Stackelberg-Nash equilibrium (SNE), which is a policy pair $(\\pi^*, \\nu^*)$ such that (i) $\\pi^*$ is the optimal policy for the leader when the followers always play their best response, and (ii) $\\nu^*$ is the best response policy of the followers, which is a Nash equilibrium of the followers' game induced by $\\pi^*$. We develop sample-efficient reinforcement learning (RL) algorithms for solving for an SNE in both online and offline settings. Our algorithms are optimistic and pessimistic variants of least-squares value iteration, and they are readily able to incorporate function approximation tools in the setting of large state spaces. Furthermore, for the case with linear function approximation, we prove that our algorithms achieve sublinear regret and suboptimality under online and offline setups respectively. To the best of our knowledge, we establish the first provably efficient RL algorithms for solving for SNEs in general-sum Markov games with myopic followers."
                },
                {
                    "title": "Inducing Equilibria via Incentives: Simultaneous Design-and-Play Ensures Global Convergence",
                    "abstract": "To regulate a social system comprised of self-interested agents, economic incentives are often required to induce a desirable outcome. This incentive design problem naturally possesses a bilevel structure, in which a designer modifies the rewards of the agents with incentives while anticipating the response of the agents, who play a non-cooperative game that converges to an equilibrium. The existing bilevel optimization algorithms raise a dilemma when applied to this problem: anticipating how incentives affect the agents at equilibrium requires solving the equilibrium problem repeatedly, which is computationally inefficient; bypassing the time-consuming step of equilibrium-finding can reduce the computational cost, but may lead the designer to a sub-optimal solution. To address such a dilemma, we propose a method that tackles the designer's and agents' problems simultaneously in a single loop. Specifically, at each iteration, both the designer and the agents only move one step. Nevertheless, we allow the designer to gradually learn the overall influence of the incentives on the agents, which guarantees optimality after convergence. The convergence rate of the proposed scheme is also established for a broad class of games."
                },
                {
                    "title": "In Congestion Games, Taxes Achieve Optimal Approximation",
                    "abstract": "We consider the problem of minimizing social cost in atomic congestion games and show, perhaps surprisingly, that efficiently computed taxation mechanisms yield the same performance achievable by the best polynomial time algorithm, even when the latter has full control over the players' actions. It follows that no other tractable approach geared at incentivizing desirable system behavior can improve upon this result, regardless of whether it is based on taxations, coordination mechanisms, information provision, or any other principle. Three technical contributions underpin this conclusion. First, we show that computing the minimum social cost is NP-hard to approximate within a given factor depending solely on the admissible resource costs. Second, we design a tractable taxation mechanism whose efficiency (price of anarchy) matches this hardness factor, and thus is optimal. As these results extend to coarse correlated equilibria, any no-regret algorithm inherits the same performances, allowing us to devise polynomial time algorithms with optimal approximation."
                },
                {
                    "title": "Sample-Efficient Learning of Stackelberg Equilibria in General-Sum Games",
                    "abstract": "Real world applications such as economics and policy making often involve solving multi-agent games with two unique features: (1) The agents are inherently asymmetric and partitioned into leaders and followers; (2) The agents have different reward functions, thus the game is general-sum. The majority of existing results in this field focuses on either symmetric solution concepts (e.g. Nash equilibrium) or zero-sum games. It remains open how to learn the Stackelberg equilibrium -- an asymmetric analog of the Nash equilibrium -- in general-sum games efficiently from noisy samples. This paper initiates the theoretical study of sample-efficient learning of the Stackelberg equilibrium, in the bandit feedback setting where we only observe noisy samples of the reward. We consider three representative two-player general-sum games: bandit games, bandit-reinforcement learning (bandit-RL) games, and linear bandit games. In all these games, we identify a fundamental gap between the exact value of the Stackelberg equilibrium and its estimated version using finitely many noisy samples, which can not be closed information-theoretically regardless of the algorithm. We then establish sharp positive results on sample-efficient learning of Stackelberg equilibrium with value optimal up to the gap identified above, with matching lower bounds in the dependency on the gap, error tolerance, and the size of the action spaces. Overall, our results unveil unique challenges in learning Stackelberg equilibria under noisy bandit feedback, which we hope could shed light on future research on this topic."
                },
                {
                    "title": "End-to-End Learning and Intervention in Games",
                    "abstract": "In a social system, the self-interest of agents can be detrimental to the collective good, sometimes leading to social dilemmas. To resolve such a conflict, a central designer may intervene by either redesigning the system or incentivizing the agents to change their behaviors. To be effective, the designer must anticipate how the agents react to the intervention, which is dictated by their often unknown payoff functions. Therefore, learning about the agents is a prerequisite for intervention. In this paper, we provide a unified framework for learning and intervention in games. We cast the equilibria of games as individual layers and integrate them into an end-to-end optimization framework. To enable the backward propagation through the equilibria of games, we propose two approaches, respectively based on explicit and implicit differentiation. Specifically, we cast the equilibria as the solutions to variational inequalities (VIs). The explicit approach unrolls the projection method for solving VIs, while the implicit approach exploits the sensitivity of the solutions to VIs. At the core of both approaches is the differentiation through a projection operator. Moreover, we establish the correctness of both approaches and identify the conditions under which one approach is more desirable than the other. The analytical results are validated using several real-world problems."
                },
                {
                    "title": "Bandit Algorithms",
                    "abstract": "sets of environments and policies respectively and ` : E \u00d7\u03a0\u2192 [0, 1] a bounded loss function. Given a policy \u03c0 let `(\u03c0) = (`(\u03bd1, \u03c0), . . . , `(\u03bdN , \u03c0)) be the loss vector resulting from policy \u03c0. Define S = {`(\u03c0) : \u03c0 \u2208 \u03a0} and \u03bb(S) = {x \u2208 cl(S) : y 6< x for all y \u2208 S} , where y 6< x is defined to mean it is not true that yi \u2264 xi for all i with strict inequality for at least one i. Prove that if \u03bb(S) \u2286 S and S is convex, then for each x \u2208 \u03bb(S) there exists a prior q \u2208 P(E) and policy \u03c0\u2217 such that `(\u03c0) = x and \u2211 \u03bd\u2208E q(\u03bd)`(\u03bd, \u03c0\u2217) = min \u03c0\u2208\u03a0 \u2211"
                },
                {
                    "title": "Optimal Taxes in Atomic Congestion Games",
                    "abstract": "How can we design mechanisms to promote efficient use of shared resources? Here, we answer this question in relation to the well-studied class of atomic congestion games, used to model a variety of problems, including traffic routing. Within this context, a methodology for designing tolling mechanisms that minimize the system inefficiency (price of anarchy) exploiting solely local information is so far missing in spite of the scientific interest. In this article, we resolve this problem through a tractable linear programming formulation that applies to and beyond polynomial congestion games. When specializing our approach to the polynomial case, we obtain tight values for the optimal price of anarchy and corresponding tolls, uncovering an unexpected link with load balancing games. We also derive optimal tolling mechanisms that are constant with the congestion level, generalizing the results of Caragiannis et al. [8] to polynomial congestion games and beyond. Finally, we apply our techniques to compute the efficiency of the marginal cost mechanism. Surprisingly, optimal tolling mechanism using only local information perform closely to existing mechanism that utilize global information, e.g., Bil\u00f2 and Vinci [6], while the marginal cost mechanism, known to be optimal in the continuous-flow model, has lower efficiency than that encountered levying no toll. All results are tight for pure Nash equilibria and extend to coarse correlated equilibria."
                },
                {
                    "title": "Learning Optimal Strategies to Commit To",
                    "abstract": "Over the past decades, various theories and algorithms have been developed under the framework of Stackelberg games and part of these innovations have been fielded under the scenarios of national security defenses and wildlife protections. However, one of the remaining difficulties in the literature is that most of theoretical works assume full information of the payoff matrices, while in applications, the leader often has no prior knowledge about the follower\u2019s payoff matrix, but may gain information about the follower\u2019s utility function through repeated interactions. In this paper, we study the problem of learning the optimal leader strategy in Stackelberg (security) games and develop novel algorithms as well as new hardness results."
                },
                {
                    "title": "Solving linear programs in the current matrix multiplication time",
                    "abstract": "This paper shows how to solve linear programs of the form minAx=b,x\u22650 c\u22a4x with n variables in time O*((n\u03c9+n2.5\u2212\u03b1/2+n2+1/6) log(n/\u03b4)) where \u03c9 is the exponent of matrix multiplication, \u03b1 is the dual exponent of matrix multiplication, and \u03b4 is the relative accuracy. For the current value of \u03c9\u223c2.37 and \u03b1\u223c0.31, our algorithm takes O*(n\u03c9 log(n/\u03b4)) time. When \u03c9 = 2, our algorithm takes O*(n2+1/6 log(n/\u03b4)) time. Our algorithm utilizes several new concepts that we believe may be of independent interest: (1) We define a stochastic central path method. (2) We show how to maintain a projection matrix \u221aW A\u22a4(AWA\u22a4)\u22121A \u221aW in sub-quadratic time under \u21132 multiplicative changes in the diagonal matrix W."
                },
                {
                    "title": "Dynamic Taxes for Polynomial Congestion Games",
                    "abstract": "We consider the efficiency of taxation in congestion games with polynomial latency functions along the line of research initiated by [Caragiannis et al., ACM Transactions on Algorithms, 2010] who focused on both pure and mixed Nash equilibria in games with affine latencies only. By exploiting the primal-dual method [Bilo, Proceedings of the 10th Workshop on Approximation and Online Algorithms, 2012], we obtain interesting upper bounds with respect to a variety of different solution concepts ranging from approximate pure Nash equilibria up to approximate coarse correlated equilibria, and including also approximate one-round walks starting from the empty state. Our findings show a high beneficial effect of taxation which increases more than linearly with the degree of the latency functions. In some cases, a tight relationship with some well-studied polynomials in Combinatorics and Number Theory, such as the Touchard and the Geometric polynomials, arises. In these cases we can also show matching lower bounds, albeit under mild assumptions; interestingly, our upper bounds are derived by exploiting the combinatorial definition of these polynomials, while our lower bounds are constructed by relying on their analytical characterization."
                },
                {
                    "title": "Online Learning of Nash Equilibria in Congestion Games",
                    "abstract": "We study the repeated, nonatomic congestion game, in which multiple populations of players share resources and make, at each iteration, a decentralized decision on which resources to utilize. We investigate the following question: given a model of how individual players update their strategies, does the resulting dynamics of strategy profiles converge to the set of Nash equilibria of the one-shot game? We consider in particular a model in which players update their strategies using algorithms with sublinear discounted regret. We show that the resulting sequence of strategy profiles converges to the set of Nash equilibria in the sense of Cesaro means. However, convergence of the actual sequence is not guaranteed in general. We show that it can be guaranteed for a class of algorithms with a sublinear discounted regret and which satisfy an additional condition. We call such algorithms AREP (approximate replicator) algorithms, as they can be interpreted as a discrete-time approximation of the replicator equat..."
                },
                {
                    "title": "Learning Optimal Commitment to Overcome Insecurity",
                    "abstract": "Game-theoretic algorithms for physical security have made an impressive real-world impact. These algorithms compute an optimal strategy for the defender to commit to in a Stackelberg game, where the attacker observes the defender's strategy and best-responds. In order to build the game model, though, the payoffs of potential attackers for various outcomes must be estimated; inaccurate estimates can lead to significant inefficiencies. We design an algorithm that optimizes the defender's strategy with no prior information, by observing the attacker's responses to randomized deployments of resources and learning his priorities. In contrast to previous work, our algorithm requires a number of queries that is polynomial in the representation of the game."
                },
                {
                    "title": "Generalized Efficiency Bounds in Distributed Resource Allocation",
                    "abstract": "Game theory is emerging as a popular tool for distributed control of multiagent systems. To take advantage of these game theoretic tools, the interactions of the autonomous agents must be designed within a game-theoretic environment. A central component of this game-theoretic design is the assignment of a local utility function to each agent. One promising approach to utility design is assigning each agent a utility function according to the agent's Shapley value. This method frequently results in games that possess many desirable features, such as the existence of pure Nash equilibria with near-optimal efficiency. In this paper, we explore the relationship between the Shapley value utility design and the resulting efficiency of both pure Nash equilibria and coarse correlated equilibria. To study this relationship, we introduce a simple class of resource allocation problems. Within this class, we derive an explicit relationship between the structure of the resource allocation problem and the efficiency of the resulting equilibria. Lastly, we derive a bicriteria bound for this class of resource allocation problems-a bound on the value of the optimal allocation relative to the value of an equilibrium allocation with additional agents."
                },
                {
                    "title": "Introduction to Nonparametric Regression",
                    "abstract": "The prerequisites vary in level and depth as topics progress, ranging from probability (sometimes measure-theoretic) and Markov chains to statistical inference (at the level of Bickel and Doksum 1977) and simulation methods. Yet, depending on one\u2019s background and interests, there is some useful material that is accessible to theoreticians and practitioners interested in HMM. There are no exercises at the end of chapters, but the monograph is enriched with a good literature review and complementary material at the end of many chapters. Thus, it is suitable for independent study, and many of the chapters can be easily incorporated into advanced courses on stochastic processes or computational statistics. Even with multiple authors (in addition to those listed, contributors include R. Douc, C. P. Robert, G. Fort, and S. Boucheron), the style is uniform and commendable. Very few typographical errors were noticed in the main text. The flawed index (with mismatched page numbers) has been rectified by the authors (see http://www.tsi.enst.fr/~cappe/ihmm/errata.html). This monograph is a valuable resource. It provides a good literature review, an excellent account of the state of the art research on the necessary theory and algorithms, and ample illustrations of numerous applications of HMM. It goes much beyond the earlier resources on HMMs\u2014the basic introduction to the discrete valued HMM by MacDonald and Zucchini (1997) and the probabilistic account with engineering emphasis by Elliott, Aggoun, and Moore (1995). Theoreticians will benefit from the rigorous development of the theoretical underpinnings, and those with application-oriented interest will like the clear presentations of the algorithms and applications. The HMM methodology is here to stay and will grow more rapidly than ever, and this monograph will serve as a standard reference in that context. The back cover states that \u201cit will be useful for researchers and practitioners in areas such as statistics, signal processing, communications engineering, control theory, econometrics, finance. . . .\u201d This claim is fully justified, and I anticipate this work to serve well many Technometrics readers in the coming years."
                },
                {
                    "title": "Taxes for linear atomic congestion games",
                    "abstract": "We study congestion games where players aim to access a set of resources. Each player has a set of possible strategies and each resource has a function associating the latency it incurs to the players using it. Players are non--cooperative and each wishes to follow a strategy that minimizes her own latency with no regard to the global optimum. Previous work has studied the impact of this selfish behavior on system performance. In this article, we study the question of how much the performance can be improved if players are forced to pay taxes for using resources. Our objective is to extend the original game so that selfish behavior does not deteriorate performance. We consider atomic congestion games with linear latency functions and present both negative and positive results. Our negative results show that optimal system performance cannot be achieved even in very simple games. On the positive side, we show that there are ways to assign taxes that can improve the performance of linear congestion games by forcing players to follow strategies where the total latency suffered is within a factor of 2 of the minimum possible; this result is shown to be tight. Furthermore, even in cases where in the absence of taxes the system behavior may be very poor, we show that the total disutility of players (latency plus taxes) is not much larger than the optimal total latency. Besides existential results, we show how to compute taxes in time polynomial in the size of the game by solving convex quadratic programs. Similar questions have been extensively studied in the model of non-atomic congestion games. To the best of our knowledge, this is the first study of the efficiency of taxes in atomic congestion games."
                },
                {
                    "title": "Studies in the Economics of Transportation.",
                    "abstract": "THE CAPABILITIES AND EFFICIENT OPERATION OF SEVERAL TYPES OF TRANSPORTATION SYSTEMS ARE STUDIED. FOR HIGHWAY TRAFFIC THE MAIN EMPHASIS IS ON CONGESTION AND ITS CONSEQUENCES. QUEUING MODELS ARE USED TO DETERMINE FLUCTUATIONS IN WAITING TIME FOR TWO CASES, THE FLOW OF CARS THROUGH AN INTERSECTION AND THE PASSING OF SLOWER CARS BY FASTER CARS ON A TWO-LANE HIGHWAY. THE RESULTS OF THESE STUDIES ARE INCORPORATED INTO A NON-LINEAR MODEL OF HIGHWAY TRAFFIC BEHAVIOR ON A ROAD NETWORK. THE CAPACITIES ARE EXPRESSED BY THE RELATIONSHIP BETWEEN TRAFFIC FLOWS ON ALL ROUTES AND THE COSTS ENCOUNTERED ON EACH AS A RESULT OF THESE FLOWS. HIGHWAY TRAFFIC DEMANDS ARE THE RESULT OF A GREAT MANY INDIVIDUAL DECISIONS ABOUT DESTINATION, ROUTES, AND PREFERRED SPEEDS. THE ANALOGY BETWEEN HIGHWAY TRAFFIC EQUILIBRIUM AND THE MARKET RESPONSE MODELS USED BY ECONOMISTS IS GIVEN. IF THE INDIVIDUALLY INCURRED COST OF TRANSPORTATION IS REGARDED AS THE \"PRICE\" IN THE TRANSPORTATION \"MARKET,\" THE ECONOMIST'S DEMAND CURVE BECOMES APPLICABLE. FOR A SINGLE ROAD IT GIVES THE FLOW OF TRAFFIC DEMANDED FOR ANY GIVEN COST ENCOUNTERED ON THAT ROAD. GENERALLY, THE HIGHER THE COST, THE SMALLER THE FLOW. CONSIDERING A ROAD NETWORK THE DEMAND FUNCTION GIVES THE FLOWS ON EACH ROUTE IN RESPONSE TO A SET OF TRANSPORTATION COSTS ALONG THESE ROUTES. EQUILIBRIUM EXISTS IF THE FLOWS ON ALL ROADS RESULTING FROM THESE COSTS ARE THE SAME AS THE FLOWS THAT PRODUCE THESE SAME COSTS. THIS CONCEPT IS APPLIED TO A STUDY OF TRAFFIC EQUILIBRIUM AND USE OF THE ANALYSIS IN THE PREDICTION OF TRAFFIC FLOWS IS DISCUSSED. THE ANALYSIS OF DEMAND AND OF EQUILIBRIUM RECOGNIZES THAT FREEDOM OF CHOICE IS PART OF THE SERVICE RENDERED BY THE ROAD NETWORK. TOLLS FROM THE USERS OF CONGESTED ROADS AT RATES THAT WOULD MEASURE THE COST TO OTHERS CAUSED BY THE AVERAGE ROAD USER WOULD ENCOURAGE BETTER USE OF THE HIGHWAY SYSTEM. THERE ARE OBSERVATIONS ON HOW CLOSELY MAXIMUM EFFICIENCY CAN BE APPROACHED BY PROPER CHOICE OF RATES ON PRESENT TOLL ROADS AND BY OTHER WAYS OF PENALIZING ADDITIONS TO TRAFFIC CONGESTION. BECAUSE OF THE NONLINEAR RELATIONSHIP BETWEEN USE AND COST, SUCH PRICING DOES NOT NECESSARILY PRODUCE REVENUES EQUAL TO THE TOTAL COST OF OPERATING AND FINANCING THE INDIVIDUAL FACILITY. SOME OBSERVATIONS ON THE QUESTION OF CRITERIA FOR EXTENSIONS OR IMPROVEMENTS OF THE NETWORK TO RELIEVE CONGESTION ARE GIVEN. /AUTHOR/"
                },
                {
                    "title": "Wealth and Welfare",
                    "abstract": "THOUGH the tide of this book adds \" of the Punjab \"-\u2014and must thu^ limit its circulation, espedaUy in Britain, whose invindble ignorance of India persists, de^ite all ui^ency of need to mitij^te it, as our writer peculiarly tends and helps to do\u2014it is espmally reviewed here as a real and solid contribution to economics, and of the kind most urgently needed by its students in this country, as weU as India. ECONOMICS, for Mr. Caivert, is not merely, or eveai mainly, the study of \u2022\" the Market,\" ruled by its gods of *' supply and demand.\" Nor does it turn upon \" Manufactures and Commerce,\" as for Adam Smith; nor yet on \" Capital,\" as since Ricardo ; and yet as little on Marx's turning of those arguments outside in, towards the apotheosis of \" Labour,\" meaning essentially the urban and mechanical proletariat, with resultant exasperation accordingly from Socialism since his publication, to Bolshevism to-day. WHAT Mr. Caivert has done has been to recover for himself the essential truths of economics as seen by Vauban and the Physiocrais, from whom Adam Smith arose, and most unhappily departed, instead of developing as he should have done. In a word, his economics is ^mdammtally Rural economics: and from his first prefatory page he stoutly carri^ his war with the conventional urban economics into England as well as India\u2014 citing for the former also Mr. CoUett\u2014\"The ignorance of the urban majority in this country on agricultural matters is so colossal and so genuine -as almost to deserve respect.\""
                },
                {
                    "title": "Algorithmic Game Theory",
                    "abstract": "Online mechanisms extend the methods of mechanism design to dynamic environments with multiple agents and private information. Decisions must be made as information about types is revealed online and without knowledge of the future in the sense of online algorithms. We \ufb01rst consider single-valued preference domains and characterize the space of decision policies that can be truthfully implemented in a dominant strategy equilibrium. Working in a model-free environment we present truthful auctions for domains with ex-piring items and limited-supply items. Turning to a more general preference domain, and assuming the existence of a probabilistic model for agent types, we de\ufb01ne a dynamic Vickrey-Clarke-Groves mechanism that is e\ufb03cient and Bayes-Nash incentive compatible. We close with some thoughts about future research directions in this area."
                },
                {
                    "title": "Selfish routing and the price of anarchy",
                    "abstract": "Selfish routing is a classical mathematical model of how self-interested users might route traffic through a congested network. The outcome of selfish routing is generally inefficient, in that it fails to optimize natural objective functions. The price of anarchy is a quantitative measure of this inefficiency. We survey recent work that analyzes the price of anarchy of selfish routing. We also describe related results on bounding the worst-possible severity of a phenomenon called Braess's Paradox, and on three techniques for reducing the price of anarchy of selfish routing. This survey concentrates on the contributions of the author's PhD thesis, but also discusses several more recent results in the area."
                },
                {
                    "title": "Congestion Toll Pricing of Tra c Networks",
                    "abstract": "This paper concerns tolling methodologies for tra c networks which ensure that the resultant equilibrium ows are system optimal. A nonnegative vector is de ned to be a valid toll vector, if the set of tolled user equilibrium solutions is a subset of the set of untolled system optimal solutions. The problem of characterizing the toll set T , which is the set of all valid toll vectors, is studied. Descriptions and characterizations of T are given for the cases when either the cost map is strictly monotonic or is a ne monotonic. In the latter case, the cost map is of the form Qv + c, where Q is a not necessarily symmetric matrix and Q+Q is positive semide nite. The results are illustrated with several examples. 1 Support from the NSF grant CCR-9400216 is acknowledged. 2 Current address: Dept. of Mathematics, Link\u007foping University, Link\u007foping, Sweden"
                },
                {
                    "title": "Potential Games",
                    "abstract": "We define and discuss several notions of potential functions for games in strategic form. We characterize games that have a potential function, and we present a variety of applications ."
                }
            ],
            "categories": [
                "cs.GT",
                "cs.AI",
                "cs.LG",
                "cs.MA"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively model and solve congestion games in the context of nonparametric regression to find Nash equilibria under varying tax structures?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the understanding of strategic interactions in networked systems, such as transportation and communication networks. By addressing congestion games, we can improve resource allocation, optimize traffic flow, and enhance decision-making processes in various applications. This research could lead to significant advancements in both theoretical frameworks and practical implementations, influencing future studies in game theory, machine learning, and economic modeling.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the inherent complexity of congestion games, where players' strategies are interdependent and can lead to multiple equilibria. Naive approaches may fail due to the non-convex nature of the potential functions involved, which complicates the optimization process. Additionally, the presence of discontinuities in Nash loads with respect to tax changes introduces further technical obstacles, making it difficult to derive stable and reliable solutions.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often overlooked the intricate relationship between tax structures and Nash equilibria in congestion games, leading to gaps in understanding. Existing solutions may have focused on simplified models that do not capture the full complexity of real-world scenarios. Our approach differs by integrating nonparametric regression techniques to better model the underlying dynamics of congestion games, thereby providing a more comprehensive framework for analysis.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves using nonparametric regression to model the cost functions in congestion games, allowing for a more flexible representation of player strategies. We will utilize a dataset derived from real-world traffic patterns to validate our model, measuring performance through metrics such as social welfare and equilibrium stability. The expected outcomes include the identification of Nash equilibria under various tax scenarios, providing insights into optimal resource allocation and strategic decision-making in congested environments."
            }
        },
        "author_data": {
            "b45e8ddf-255f-4109-a717-ffb0ab62af6b": {
                "pk": "b45e8ddf-255f-4109-a717-ffb0ab62af6b",
                "name": "Qiwen Cui",
                "collaborators": [
                    "Simon S. Du",
                    "Zhihan Xiong",
                    "Maryam Fazel",
                    "Lin F. Yang",
                    "Haozhe Jiang",
                    "Yan Dai",
                    "Kaiqing Zhang",
                    "Xinqi Wang",
                    "Minghan Yang",
                    "Dong Xu"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Multi-Agent Systems",
                    "Sample Complexity",
                    "Game Theory"
                ],
                "publications": [
                    {
                        "title": "Is Plug-in Solver Sample-Efficient for Feature-based Reinforcement Learning?",
                        "abstract": "It is believed that a model-based approach for reinforcement learning (RL) is the key to reduce sample complexity. However, the understanding of the sample optimality of model-based RL is still largely missing, even for the linear case. This work considers sample complexity of finding an $\\epsilon$-optimal policy in a Markov decision process (MDP) that admits a linear additive feature representation, given only access to a generative model. We solve this problem via a plug-in solver approach, which builds an empirical model and plans in this empirical model via an arbitrary plug-in solver. We prove that under the anchor-state assumption, which implies implicit non-negativity in the feature space, the minimax sample complexity of finding an $\\epsilon$-optimal policy in a $\\gamma$-discounted MDP is $O(K/(1-\\gamma)^3\\epsilon^2)$, which only depends on the dimensionality $K$ of the feature space and has no dependence on the state or action space. We further extend our results to a relaxed setting where anchor-states may not exist and show that a plug-in approach can be sample efficient as well, providing a flexible approach to design model-based algorithms for RL."
                    },
                    {
                        "title": "Minimax Sample Complexity for Turn-based Stochastic Game",
                        "abstract": "The empirical success of Multi-agent reinforcement learning is encouraging, while few theoretical guarantees have been revealed. In this work, we prove that the plug-in solver approach, probably the most natural reinforcement learning algorithm, achieves minimax sample complexity for turn-based stochastic game (TBSG). Specifically, we plan in an empirical TBSG by utilizing a `simulator' that allows sampling from arbitrary state-action pair. We show that the empirical Nash equilibrium strategy is an approximate Nash equilibrium strategy in the true TBSG and give both problem-dependent and problem-independent bound. We develop absorbing TBSG and reward perturbation techniques to tackle the complex statistical dependence. The key idea is artificially introducing a suboptimality gap in TBSG and then the Nash equilibrium strategy lies in a finite set."
                    },
                    {
                        "title": "When is Offline Two-Player Zero-Sum Markov Game Solvable?",
                        "abstract": "We study what dataset assumption permits solving offline two-player zero-sum Markov games. In stark contrast to the offline single-agent Markov decision process, we show that the single strategy concentration assumption is insufficient for learning the Nash equilibrium (NE) strategy in offline two-player zero-sum Markov games. On the other hand, we propose a new assumption named unilateral concentration and design a pessimism-type algorithm that is provably efficient under this assumption. In addition, we show that the unilateral concentration assumption is necessary for learning an NE strategy. Furthermore, our algorithm can achieve minimax sample complexity without any modification for two widely studied settings: dataset with uniform concentration assumption and turn-based Markov games. Our work serves as an important initial step towards understanding offline multi-agent reinforcement learning."
                    },
                    {
                        "title": "Provably Efficient Offline Multi-agent Reinforcement Learning via Strategy-wise Bonus",
                        "abstract": "This paper considers offline multi-agent reinforcement learning. We propose the strategy-wise concentration principle which directly builds a confidence interval for the joint strategy, in contrast to the point-wise concentration principle that builds a confidence interval for each point in the joint action space. For two-player zero-sum Markov games, by exploiting the convexity of the strategy-wise bonus, we propose a computationally efficient algorithm whose sample complexity enjoys a better dependency on the number of actions than the prior methods based on the point-wise bonus. Furthermore, for offline multi-agent general-sum Markov games, based on the strategy-wise bonus and a novel surrogate function, we give the first algorithm whose sample complexity only scales $\\sum_{i=1}^mA_i$ where $A_i$ is the action size of the $i$-th player and $m$ is the number of players. In sharp contrast, the sample complexity of methods based on the point-wise bonus would scale with the size of the joint action space $\\Pi_{i=1}^m A_i$ due to the curse of multiagents. Lastly, all of our algorithms can naturally take a pre-specified strategy class $\\Pi$ as input and output a strategy that is close to the best strategy in $\\Pi$. In this setting, the sample complexity only scales with $\\log |\\Pi|$ instead of $\\sum_{i=1}^mA_i$."
                    },
                    {
                        "title": "Refined Sample Complexity for Markov Games with Independent Linear Function Approximation",
                        "abstract": "Markov Games (MG) is an important model for Multi-Agent Reinforcement Learning (MARL). It was long believed that the \"curse of multi-agents\" (i.e., the algorithmic performance drops exponentially with the number of agents) is unavoidable until several recent works (Daskalakis et al., 2023; Cui et al., 2023; Wang et al., 2023). While these works resolved the curse of multi-agents, when the state spaces are prohibitively large and (linear) function approximations are deployed, they either had a slower convergence rate of $O(T^{-1/4})$ or brought a polynomial dependency on the number of actions $A_{\\max}$ -- which is avoidable in single-agent cases even when the loss functions can arbitrarily vary with time. This paper first refines the AVLPR framework by Wang et al. (2023), with an insight of designing *data-dependent* (i.e., stochastic) pessimistic estimation of the sub-optimality gap, allowing a broader choice of plug-in algorithms. When specialized to MGs with independent linear function approximations, we propose novel *action-dependent bonuses* to cover occasionally extreme estimation errors. With the help of state-of-the-art techniques from the single-agent RL literature, we give the first algorithm that tackles the curse of multi-agents, attains the optimal $O(T^{-1/2})$ convergence rate, and avoids $\\text{poly}(A_{\\max})$ dependency simultaneously."
                    },
                    {
                        "title": "Breaking the Curse of Multiagents in a Large State Space: RL in Markov Games with Independent Linear Function Approximation",
                        "abstract": "We propose a new model, independent linear Markov game, for multi-agent reinforcement learning with a large state space and a large number of agents. This is a class of Markov games with independent linear function approximation, where each agent has its own function approximation for the state-action value functions that are marginalized by other players' policies. We design new algorithms for learning the Markov coarse correlated equilibria (CCE) and Markov correlated equilibria (CE) with sample complexity bounds that only scale polynomially with each agent's own function class complexity, thus breaking the curse of multiagents. In contrast, existing works for Markov games with function approximation have sample complexity bounds scale with the size of the \\emph{joint action space} when specialized to the canonical tabular Markov game setting, which is exponentially large in the number of agents. Our algorithms rely on two key technical innovations: (1) utilizing policy replay to tackle non-stationarity incurred by multiple agents and the use of function approximation; (2) separating learning Markov equilibria and exploration in the Markov games, which allows us to use the full-information no-regret learning oracle instead of the stronger bandit-feedback no-regret learning oracle used in the tabular setting. Furthermore, we propose an iterative-best-response type algorithm that can learn pure Markov Nash equilibria in independent linear Markov potential games. In the tabular case, by adapting the policy replay mechanism for independent linear Markov games, we propose an algorithm with $\\widetilde{O}(\\epsilon^{-2})$ sample complexity to learn Markov CCE, which improves the state-of-the-art result $\\widetilde{O}(\\epsilon^{-3})$ in Daskalakis et al. 2022, where $\\epsilon$ is the desired accuracy, and also significantly improves other problem parameters."
                    },
                    {
                        "title": "On Gap-dependent Bounds for Offline Reinforcement Learning",
                        "abstract": "This paper presents a systematic study on gap-dependent sample complexity in offline reinforcement learning. Prior work showed when the density ratio between an optimal policy and the behavior policy is upper bounded (the optimal policy coverage assumption), then the agent can achieve an $O\\left(\\frac{1}{\\epsilon^2}\\right)$ rate, which is also minimax optimal. We show under the optimal policy coverage assumption, the rate can be improved to $O\\left(\\frac{1}{\\epsilon}\\right)$ when there is a positive sub-optimality gap in the optimal $Q$-function. Furthermore, we show when the visitation probabilities of the behavior policy are uniformly lower bounded for states where an optimal policy's visitation probabilities are positive (the uniform optimal policy coverage assumption), the sample complexity of identifying an optimal policy is independent of $\\frac{1}{\\epsilon}$. Lastly, we present nearly-matching lower bounds to complement our gap-dependent upper bounds."
                    },
                    {
                        "title": "Learning in Congestion Games with Bandit Feedback",
                        "abstract": "In this paper, we investigate Nash-regret minimization in congestion games, a class of games with benign theoretical structure and broad real-world applications. We first propose a centralized algorithm based on the optimism in the face of uncertainty principle for congestion games with (semi-)bandit feedback, and obtain finite-sample guarantees. Then we propose a decentralized algorithm via a novel combination of the Frank-Wolfe method and G-optimal design. By exploiting the structure of the congestion game, we show the sample complexity of both algorithms depends only polynomially on the number of players and the number of facilities, but not the size of the action set, which can be exponentially large in terms of the number of facilities. We further define a new problem class, Markov congestion games, which allows us to model the non-stationarity in congestion games. We propose a centralized algorithm for Markov congestion games, whose sample complexity again has only polynomial dependence on all relevant problem parameters, but not the size of the action set."
                    },
                    {
                        "title": "NG+ : A Multi-Step Matrix-Product Natural Gradient Method for Deep Learning",
                        "abstract": "In this paper, a novel second-order method called NG+ is proposed. By following the rule ``the shape of the gradient equals the shape of the parameter\", we define a generalized fisher information matrix (GFIM) using the products of gradients in the matrix form rather than the traditional vectorization. Then, our generalized natural gradient direction is simply the inverse of the GFIM multiplies the gradient in the matrix form. Moreover, the GFIM and its inverse keeps the same for multiple steps so that the computational cost can be controlled and is comparable with the first-order methods. A global convergence is established under some mild conditions and a regret bound is also given for the online learning setting. Numerical results on image classification with ResNet50, quantum chemistry modeling with Schnet, neural machine translation with Transformer and recommendation system with DLRM illustrate that GN+ is competitive with the state-of-the-art methods."
                    },
                    {
                        "title": "$\\mathbf{(N,K)}$-Puzzle: A Cost-Efficient Testbed for Benchmarking Reinforcement Learning Algorithms in Generative Language Model",
                        "abstract": "Recent advances in reinforcement learning (RL) algorithms aim to enhance the performance of language models at scale. Yet, there is a noticeable absence of a cost-effective and standardized testbed tailored to evaluating and comparing these algorithms. To bridge this gap, we present a generalized version of the 24-Puzzle: the $(N,K)$-Puzzle, which challenges language models to reach a target value $K$ with $N$ integers. We evaluate the effectiveness of established RL algorithms such as Proximal Policy Optimization (PPO), alongside novel approaches like Identity Policy Optimization (IPO) and Direct Policy Optimization (DPO)."
                    },
                    {
                        "title": "Near-Optimal Randomized Exploration for Tabular Markov Decision Processes",
                        "abstract": "We study algorithms using randomized value functions for exploration in reinforcement learning. This type of algorithms enjoys appealing empirical performance. We show that when we use 1) a single random seed in each episode, and 2) a Bernstein-type magnitude of noise, we obtain a worst-case $\\widetilde{O}\\left(H\\sqrt{SAT}\\right)$ regret bound for episodic time-inhomogeneous Markov Decision Process where $S$ is the size of state space, $A$ is the size of action space, $H$ is the planning horizon and $T$ is the number of interactions. This bound polynomially improves all existing bounds for algorithms based on randomized value functions, and for the first time, matches the $\\Omega\\left(H\\sqrt{SAT}\\right)$ lower bound up to logarithmic factors. Our result highlights that randomized exploration can be near-optimal, which was previously achieved only by optimistic algorithms. To achieve the desired result, we develop 1) a new clipping operation to ensure both the probability of being optimistic and the probability of being pessimistic are lower bounded by a constant, and 2) a new recursive formula for the absolute value of estimation errors to analyze the regret."
                    },
                    {
                        "title": "Offline congestion games: How feedback type affects data coverage requirement",
                        "abstract": "This paper investigates when one can efficiently recover an approximate Nash Equilibrium (NE) in offline congestion games. The existing dataset coverage assumption in offline general-sum games inevitably incurs a dependency on the number of actions, which can be exponentially large in congestion games. We consider three different types of feedback with decreasing revealed information. Starting from the facility-level (a.k.a., semi-bandit) feedback, we propose a novel one-unit deviation coverage condition and give a pessimism-type algorithm that can recover an approximate NE. For the agent-level (a.k.a., bandit) feedback setting, interestingly, we show the one-unit deviation coverage condition is not sufficient. On the other hand, we convert the game to multi-agent linear bandits and show that with a generalized data coverage assumption in offline linear bandits, we can efficiently recover the approximate NE. Lastly, we consider a novel type of feedback, the game-level feedback where only the total reward from all agents is revealed. Again, we show the coverage assumption for the agent-level feedback setting is insufficient in the game-level feedback setting, and with a stronger version of the data coverage assumption for linear bandits, we can recover an approximate NE. Together, our results constitute the first study of offline congestion games and imply formal separations between different types of feedback."
                    },
                    {
                        "title": "A Black-box Approach for Non-stationary Multi-agent Reinforcement Learning",
                        "abstract": "We investigate learning the equilibria in non-stationary multi-agent systems and address the challenges that differentiate multi-agent learning from single-agent learning. Specifically, we focus on games with bandit feedback, where testing an equilibrium can result in substantial regret even when the gap to be tested is small, and the existence of multiple optimal solutions (equilibria) in stationary games poses extra challenges. To overcome these obstacles, we propose a versatile black-box approach applicable to a broad spectrum of problems, such as general-sum games, potential games, and Markov games, when equipped with appropriate learning and testing oracles for stationary environments. Our algorithms can achieve $\\widetilde{O}\\left(\\Delta^{1/4}T^{3/4}\\right)$ regret when the degree of nonstationarity, as measured by total variation $\\Delta$, is known, and $\\widetilde{O}\\left(\\Delta^{1/5}T^{4/5}\\right)$ regret when $\\Delta$ is unknown, where $T$ is the number of rounds. Meanwhile, our algorithm inherits the favorable dependence on number of agents from the oracles. As a side contribution that may be independent of interest, we show how to test for various types of equilibria by a black-box reduction to single-agent learning, which includes Nash equilibria, correlated equilibria, and coarse correlated equilibria."
                    },
                    {
                        "title": "Free from Bellman Completeness: Trajectory Stitching via Model-based Return-conditioned Supervised Learning",
                        "abstract": "Off-policy dynamic programming (DP) techniques such as $Q$-learning have proven to be important in sequential decision-making problems. In the presence of function approximation, however, these techniques often diverge due to the absence of Bellman completeness in the function classes considered, a crucial condition for the success of DP-based methods. In this paper, we show how off-policy learning techniques based on return-conditioned supervised learning (RCSL) are able to circumvent these challenges of Bellman completeness, converging under significantly more relaxed assumptions inherited from supervised learning. We prove there exists a natural environment in which if one uses two-layer multilayer perceptron as the function approximator, the layer width needs to grow linearly with the state space size to satisfy Bellman completeness while a constant layer width is enough for RCSL. These findings take a step towards explaining the superior empirical performance of RCSL methods compared to DP-based methods in environments with near-optimal datasets. Furthermore, in order to learn from sub-optimal datasets, we propose a simple framework called MBRCSL, granting RCSL methods the ability of dynamic programming to stitch together segments from distinct trajectories. MBRCSL leverages learned dynamics models and forward sampling to accomplish trajectory stitching while avoiding the need for Bellman completeness that plagues all dynamic programming algorithms. We propose both theoretical analysis and experimental evaluation to back these claims, outperforming state-of-the-art model-free and model-based offline RL algorithms across several simulated robotics problems."
                    },
                    {
                        "title": "Randomized Exploration for Reinforcement Learning with General Value Function Approximation",
                        "abstract": "We propose a model-free reinforcement learning algorithm inspired by the popular randomized least squares value iteration (RLSVI) algorithm as well as the optimism principle. Unlike existing upper-confidence-bound (UCB) based approaches, which are often computationally intractable, our algorithm drives exploration by simply perturbing the training data with judiciously chosen i.i.d. scalar noises. To attain optimistic value function estimation without resorting to a UCB-style bonus, we introduce an optimistic reward sampling procedure. When the value functions can be represented by a function class $\\mathcal{F}$, our algorithm achieves a worst-case regret bound of $\\widetilde{O}(\\mathrm{poly}(d_EH)\\sqrt{T})$ where $T$ is the time elapsed, $H$ is the planning horizon and $d_E$ is the $\\textit{eluder dimension}$ of $\\mathcal{F}$. In the linear setting, our algorithm reduces to LSVI-PHE, a variant of RLSVI, that enjoys an $\\widetilde{\\mathcal{O}}(\\sqrt{d^3H^3T})$ regret. We complement the theory with an empirical evaluation across known difficult exploration tasks."
                    }
                ]
            },
            "3cdaf786-45e1-4ad6-b7a0-c383d983d43c": {
                "pk": "3cdaf786-45e1-4ad6-b7a0-c383d983d43c",
                "name": "Maryam Fazel",
                "collaborators": [
                    "Omid Sadeghi",
                    "Reza Eghbali",
                    "Zaid Harchaoui",
                    "Amin Jalali",
                    "Lin Xiao",
                    "Vincent Roulet",
                    "Siddhartha Srinivasa",
                    "Krishnamurthy Dvijotham",
                    "Emanuel Todorov",
                    "Dmitriy Drusvyatskiy"
                ],
                "domain": [
                    "Quantum Computing",
                    "Online Optimization",
                    "Convex Optimization",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Computational approach to quantum encoder design for purity optimization",
                        "abstract": "In this paper, we address the problem of designing a quantum encoder that maximizes the minimum output purity of a given decohering channel, where the minimum is taken over all possible pure inputs. This problem is cast as a max-min optimization problem with a rank constraint on an appropriately defined matrix variable. The problem is computationally very hard because it is non-convex with respect to both the objective function (output purity) and the rank constraint. Despite this difficulty, we provide a tractable computational algorithm that produces the exact optimal solution for codespace of dimension two. Moreover, this algorithm is easily extended to cover the general class of codespaces, in which case the solution is suboptimal in the sense that the suboptimized output purity serves as a lower bound of the exact optimal purity. The algorithm consists of a sequence of semidefinite programmings and can be performed easily. Two typical quantum error channels are investigated to illustrate the effectiveness of our method."
                    },
                    {
                        "title": "Decomposable Norm Minimization with Proximal-Gradient Homotopy Algorithm",
                        "abstract": "We study the convergence rate of the proximal-gradient homotopy algorithm applied to norm-regularized linear least squares problems, for a general class of norms. The homotopy algorithm reduces the regularization parameter in a series of steps, and uses a proximal-gradient algorithm to solve the problem at each step. Proximal-gradient algorithm has a linear rate of convergence given that the objective function is strongly convex, and the gradient of the smooth component of the objective function is Lipschitz continuous. In many applications, the objective function in this type of problem is not strongly convex, especially when the problem is high-dimensional and regularizers are chosen that induce sparsity or low-dimensionality. We show that if the linear sampling matrix satisfies certain assumptions and the regularizing norm is decomposable, proximal-gradient homotopy algorithm converges with a \\emph{linear rate} even though the objective function is not strongly convex. Our result generalizes results on the linear convergence of homotopy algorithm for $l_1$-regularized least squares problems. Numerical experiments are presented that support the theoretical convergence rate analysis."
                    },
                    {
                        "title": "Worst Case Competitive Analysis of Online Algorithms for Conic Optimization",
                        "abstract": "Online optimization covers problems such as online resource allocation, online bipartite matching, adwords (a central problem in e-commerce and advertising), and adwords with separable concave returns. We analyze the worst case competitive ratio of two primal-dual algorithms for a class of online convex (conic) optimization problems that contains the previous examples as special cases defined on the positive orthant. We derive a sufficient condition on the objective function that guarantees a constant worst case competitive ratio (greater than or equal to $\\frac{1}{2}$) for monotone objective functions. We provide new examples of online problems on the positive orthant and the positive semidefinite cone that satisfy the sufficient condition. We show how smoothing can improve the competitive ratio of these algorithms, and in particular for separable functions, we show that the optimal smoothing can be derived by solving a convex optimization problem. This result allows us to directly optimize the competitive ratio bound over a class of smoothing functions, and hence design effective smoothing customized for a given cost function."
                    },
                    {
                        "title": "No-Regret Online Prediction with Strategic Experts",
                        "abstract": "We study a generalization of the online binary prediction with expert advice framework where at each round, the learner is allowed to pick $m\\geq 1$ experts from a pool of $K$ experts and the overall utility is a modular or submodular function of the chosen experts. We focus on the setting in which experts act strategically and aim to maximize their influence on the algorithm's predictions by potentially misreporting their beliefs about the events. Among others, this setting finds applications in forecasting competitions where the learner seeks not only to make predictions by aggregating different forecasters but also to rank them according to their relative performance. Our goal is to design algorithms that satisfy the following two requirements: 1) $\\textit{Incentive-compatible}$: Incentivize the experts to report their beliefs truthfully, and 2) $\\textit{No-regret}$: Achieve sublinear regret with respect to the true beliefs of the best fixed set of $m$ experts in hindsight. Prior works have studied this framework when $m=1$ and provided incentive-compatible no-regret algorithms for the problem. We first show that a simple reduction of our problem to the $m=1$ setting is neither efficient nor effective. Then, we provide algorithms that utilize the specific structure of the utility functions to achieve the two desired goals."
                    },
                    {
                        "title": "Online Continuous DR-Submodular Maximization with Long-Term Budget Constraints",
                        "abstract": "In this paper, we study a class of online optimization problems with long-term budget constraints where the objective functions are not necessarily concave (nor convex) but they instead satisfy the Diminishing Returns (DR) property. Specifically, a sequence of monotone DR-submodular objective functions $\\{f_t(x)\\}_{t=1}^T$ and monotone linear budget functions $\\{\\langle p_t,x \\rangle \\}_{t=1}^T$ arrive over time and assuming a total targeted budget $B_T$, the goal is to choose points $x_t$ at each time $t\\in\\{1,\\dots,T\\}$, without knowing $f_t$ and $p_t$ on that step, to achieve sub-linear regret bound while the total budget violation $\\sum_{t=1}^T \\langle p_t,x_t \\rangle -B_T$ is sub-linear as well. Prior work has shown that achieving sub-linear regret is impossible if the budget functions are chosen adversarially. Therefore, we modify the notion of regret by comparing the agent against a $(1-\\frac{1}{e})$-approximation to the best fixed decision in hindsight which satisfies the budget constraint proportionally over any window of length $W$. We propose the Online Saddle Point Hybrid Gradient (OSPHG) algorithm to solve this class of online problems. For $W=T$, we recover the aforementioned impossibility result. However, when $W=o(T)$, we show that it is possible to obtain sub-linear bounds for both the $(1-\\frac{1}{e})$-regret and the total budget violation."
                    },
                    {
                        "title": "Fast First-Order Methods for Monotone Strongly DR-Submodular Maximization",
                        "abstract": "Continuous DR-submodular functions are a class of functions that satisfy the Diminishing Returns (DR) property, which implies that they are concave along non-negative directions. Existing works have studied monotone continuous DR-submodular maximization subject to a convex constraint and have proposed efficient algorithms with approximation guarantees. However, in many applications, e.g., computing the stability number of a graph and mean-field inference for probabilistic log-submodular models, the DR-submodular function has the additional property of being \\emph{strongly} concave along non-negative directions that could be utilized for obtaining faster convergence rates. In this paper, we first introduce and characterize the class of \\emph{strongly DR-submodular} functions and show how such a property implies strong concavity along non-negative directions. Then, we study $L$-smooth monotone strongly DR-submodular functions that have bounded curvature, and we show how to exploit such additional structure to obtain algorithms with improved approximation guarantees and faster convergence rates for the maximization problem. In particular, we propose the SDRFW algorithm that matches the provably optimal $1-\\frac{c}{e}$ approximation ratio after only $\\lceil\\frac{L}{\\mu}\\rceil$ iterations, where $c\\in[0,1]$ and $\\mu\\geq 0$ are the curvature and the strong DR-submodularity parameter. Furthermore, we study the Projected Gradient Ascent (PGA) method for this problem and provide a refined analysis of the algorithm with an improved $\\frac{1}{1+c}$ approximation ratio and a linear convergence rate. Given that both algorithms require knowledge of the smoothness parameter $L$, we provide a \\emph{novel} characterization of $L$ for DR-submodular functions showing that in many cases, computing $L$ could be formulated as a convex problem, i.e., a geometric program, that could be solved efficiently."
                    },
                    {
                        "title": "Competitive Algorithms for Online Budget-Constrained Continuous DR-Submodular Problems",
                        "abstract": "In this paper, we study a certain class of online optimization problems, where the goal is to maximize a function that is not necessarily concave and satisfies the Diminishing Returns (DR) property under budget constraints. We analyze a primal-dual algorithm, called the Generalized Sequential algorithm, and we obtain the first bound on the competitive ratio of online monotone DR-submodular function maximization subject to linear packing constraints which matches the known tight bound in the special case of linear objective function."
                    },
                    {
                        "title": "Dual Approximation Policy Optimization",
                        "abstract": "We propose Dual Approximation Policy Optimization (DAPO), a framework that incorporates general function approximation into policy mirror descent methods. In contrast to the popular approach of using the $L_2$-norm to measure function approximation errors, DAPO uses the dual Bregman divergence induced by the mirror map for policy projection. This duality framework has both theoretical and practical implications: not only does it achieve fast linear convergence with general function approximation, but it also includes several well-known practical methods as special cases, immediately providing strong convergence guarantees."
                    },
                    {
                        "title": "On the Convergence of the Iterative Linear Exponential Quadratic Gaussian Algorithm to Stationary Points",
                        "abstract": "A classical method for risk-sensitive nonlinear control is the iterative linear exponential quadratic Gaussian algorithm. We present its convergence analysis from a first-order optimization viewpoint. We identify the objective that the algorithm actually minimizes and we show how the addition of a proximal term guarantees convergence to a stationary point."
                    },
                    {
                        "title": "A Simplified Approach to Recovery Conditions for Low Rank Matrices",
                        "abstract": "Recovering sparse vectors and low-rank matrices from noisy linear measurements has been the focus of much recent research. Various reconstruction algorithms have been studied, including $\\ell_1$ and nuclear norm minimization as well as $\\ell_p$ minimization with $p<1$. These algorithms are known to succeed if certain conditions on the measurement map are satisfied. Proofs of robust recovery for matrices have so far been much more involved than in the vector case.   In this paper, we show how several robust classes of recovery conditions can be extended from vectors to matrices in a simple and transparent way, leading to the best known restricted isometry and nullspace conditions for matrix recovery. Our results rely on the ability to \"vectorize\" matrices through the use of a key singular value inequality."
                    },
                    {
                        "title": "Convex Structured Controller Design",
                        "abstract": "We consider the problem of synthesizing optimal linear feedback policies subject to arbitrary convex constraints on the feedback matrix. This is known to be a hard problem in the usual formulations ($\\Htwo,\\Hinf,\\LQR$) and previous works have focused on characterizing classes of structural constraints that allow efficient solution through convex optimization or dynamic programming techniques. In this paper, we propose a new control objective and show that this formulation makes the problem of computing optimal linear feedback matrices convex under arbitrary convex constraints on the feedback matrix. This allows us to solve problems in decentralized control (sparsity in the feedback matrices), control with delays and variable impedance control. Although the control objective is nonstandard, we present theoretical and empirical evidence that it agrees well with standard notions of control. We also present an extension to nonlinear control affine systems. We present numerical experiments validating our approach."
                    },
                    {
                        "title": "Universal Convexification via Risk-Aversion",
                        "abstract": "We develop a framework for convexifying a fairly general class of optimization problems. Under additional assumptions, we analyze the suboptimality of the solution to the convexified problem relative to the original nonconvex problem and prove additive approximation guarantees. We then develop algorithms based on stochastic gradient methods to solve the resulting optimization problems and show bounds on convergence rates. %We show a simple application of this framework to supervised learning, where one can perform integration explicitly and can use standard (non-stochastic) optimization algorithms with better convergence guarantees. We then extend this framework to apply to a general class of discrete-time dynamical systems. In this context, our convexification approach falls under the well-studied paradigm of risk-sensitive Markov Decision Processes. We derive the first known model-based and model-free policy gradient optimization algorithms with guaranteed convergence to the optimal solution. Finally, we present numerical results validating our formulation in different applications."
                    },
                    {
                        "title": "Variational Gram Functions: Convex Analysis and Optimization",
                        "abstract": "We propose a new class of convex penalty functions, called \\emph{variational Gram functions} (VGFs), that can promote pairwise relations, such as orthogonality, among a set of vectors in a vector space. These functions can serve as regularizers in convex optimization problems arising from hierarchical classification, multitask learning, and estimating vectors with disjoint supports, among other applications. We study convexity for VGFs, and give efficient characterizations for their convex conjugates, subdifferentials, and proximal operators. We discuss efficient optimization algorithms for regularized loss minimization problems where the loss admits a common, yet simple, variational representation and the regularizer is a VGF. These algorithms enjoy a simple kernel trick, an efficient line search, as well as computational advantages over first order methods based on the subdifferential or proximal maps. We also establish a general representer theorem for such learning problems. Lastly, numerical experiments on a hierarchical classification problem are presented to demonstrate the effectiveness of VGFs and the associated optimization algorithms."
                    },
                    {
                        "title": "Relative Density and Exact Recovery in Heterogeneous Stochastic Block Models",
                        "abstract": "The Stochastic Block Model (SBM) is a widely used random graph model for networks with communities. Despite the recent burst of interest in recovering communities in the SBM from statistical and computational points of view, there are still gaps in understanding the fundamental information theoretic and computational limits of recovery. In this paper, we consider the SBM in its full generality, where there is no restriction on the number and sizes of communities or how they grow with the number of nodes, as well as on the connection probabilities inside or across communities. This generality allows us to move past the artifacts of homogenous SBM, and understand the right parameters (such as the relative densities of communities) that define the various recovery thresholds. We outline the implications of our generalizations via a set of illustrative examples. For instance, $\\log n$ is considered to be the standard lower bound on the cluster size for exact recovery via convex methods, for homogenous SBM. We show that it is possible, in the right circumstances (when sizes are spread and the smaller the cluster, the denser), to recover very small clusters (up to $\\sqrt{\\log n}$ size), if there are just a few of them (at most polylogarithmic in $n$)."
                    },
                    {
                        "title": "An optimal first order method based on optimal quadratic averaging",
                        "abstract": "In a recent paper, Bubeck, Lee, and Singh introduced a new first order method for minimizing smooth strongly convex functions. Their geometric descent algorithm, largely inspired by the ellipsoid method, enjoys the optimal linear rate of convergence. We show that the same iterate sequence is generated by a scheme that in each iteration computes an optimal average of quadratic lower-models of the function. Indeed, the minimum of the averaged quadratic approaches the true minimum at an optimal rate. This intuitive viewpoint reveals clear connections to the original fast-gradient methods and cutting plane ideas, and leads to limited-memory extensions with improved performance."
                    },
                    {
                        "title": "Fair Packing and Covering on a Relative Scale",
                        "abstract": "Fair resource allocation is a fundamental optimization problem with applications in operations research, networking, and economic and game theory. Research in these areas has led to the general acceptance of a class of $\\alpha$-fair utility functions parameterized by $\\alpha \\in [0, \\infty]$. We consider $\\alpha$-fair packing -- the problem of maximizing $\\alpha$-fair utilities under positive linear constraints -- and provide a simple first-order method for solving it with relative-error guarantees. The method has a significantly lower convergence time than the state of the art, and to analyze it, we leverage the Approximate Duality Gap Technique, which provides an intuitive interpretation of the convergence argument. Finally, we introduce a natural counterpart of $\\alpha$-fairness for minimization problems and motivate its usage in the context of fair task allocation. This generalization yields $\\alpha$-fair covering problems, for which we provide the first near-linear-time solvers with relative-error guarantees."
                    },
                    {
                        "title": "New Computational and Statistical Aspects of Regularized Regression with Application to Rare Feature Selection and Aggregation",
                        "abstract": "Prior knowledge on properties of a target model often come as discrete or combinatorial descriptions. This work provides a unified computational framework for defining norms that promote such structures. More specifically, we develop associated tools for optimization involving such norms given only the orthogonal projection oracle onto the non-convex set of desired models. As an example, we study a norm, which we term the doubly-sparse norm, for promoting vectors with few nonzero entries taking only a few distinct values. We further discuss how the K-means algorithm can serve as the underlying projection oracle in this case and how it can be efficiently represented as a quadratically constrained quadratic program. Our motivation for the study of this norm is regularized regression in the presence of rare features which poses a challenge to various methods within high-dimensional statistics, and in machine learning in general. The proposed estimation procedure is designed to perform automatic feature selection and aggregation for which we develop statistical bounds. The bounds are general and offer a statistical framework for norm-based regularization. The bounds rely on novel geometric quantities on which we attempt to elaborate as well."
                    },
                    {
                        "title": "Escaping from saddle points on Riemannian manifolds",
                        "abstract": "We consider minimizing a nonconvex, smooth function $f$ on a Riemannian manifold $\\mathcal{M}$. We show that a perturbed version of Riemannian gradient descent algorithm converges to a second-order stationary point (and hence is able to escape saddle points on the manifold). The rate of convergence depends as $1/\\epsilon^2$ on the accuracy $\\epsilon$, which matches a rate known only for unconstrained smooth minimization. The convergence rate depends polylogarithmically on the manifold dimension $d$, hence is almost dimension-free. The rate also has a polynomial dependence on the parameters describing the curvature of the manifold and the smoothness of the function. While the unconstrained problem (Euclidean setting) is well-studied, our result is the first to prove such a rate for nonconvex, manifold-constrained problems."
                    },
                    {
                        "title": "Complexity Bounds of Iterative Linear Quadratic Optimization Algorithms for Discrete Time Nonlinear Control",
                        "abstract": "A classical approach for solving discrete time nonlinear control on a finite horizon consists in repeatedly minimizing linear quadratic approximations of the original problem around current candidate solutions. While widely popular in many domains, such an approach has mainly been analyzed locally. We observe that global convergence guarantees can be ensured provided that the linearized discrete time dynamics are surjective and costs on the state variables are strongly convex. We present how the surjectivity of the linearized dynamics can be ensured by appropriate discretization schemes given the existence of a feedback linearization scheme. We present complexity bounds of algorithms based on linear quadratic approximations through the lens of generalized Gauss-Newton methods. Our analysis uncovers several convergence phases for regularized generalized Gauss-Newton algorithms."
                    },
                    {
                        "title": "Flat minima generalize for low-rank matrix recovery",
                        "abstract": "Empirical evidence suggests that for a variety of overparameterized nonlinear models, most notably in neural network training, the growth of the loss around a minimizer strongly impacts its performance. Flat minima -- those around which the loss grows slowly -- appear to generalize well. This work takes a step towards understanding this phenomenon by focusing on the simplest class of overparameterized nonlinear models: those arising in low-rank matrix recovery. We analyze overparameterized matrix and bilinear sensing, robust PCA, covariance matrix estimation, and single hidden layer neural networks with quadratic activation functions. In all cases, we show that flat minima, measured by the trace of the Hessian, exactly recover the ground truth under standard statistical assumptions. For matrix completion, we establish weak recovery, although empirical evidence suggests exact recovery holds here as well. We conclude with synthetic experiments that illustrate our findings and discuss the effect of depth on flat solutions."
                    }
                ]
            },
            "3187acee-693e-4cfe-b35b-d6611484b2a6": {
                "pk": "3187acee-693e-4cfe-b35b-d6611484b2a6",
                "name": "Simon S. Du",
                "collaborators": [
                    "Surbhi Goel",
                    "Sivaraman Balakrishnan",
                    "Aarti Singh",
                    "Jason D. Lee",
                    "Yining Wang",
                    "Yi Wu",
                    "Zihan Zhang",
                    "Xiangyang Ji",
                    "Tian Ye"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Optimization",
                    "Statistical Learning"
                ],
                "publications": [
                    {
                        "title": "Improved Learning of One-hidden-layer Convolutional Neural Networks with Overlaps",
                        "abstract": "We propose a new algorithm to learn a one-hidden-layer convolutional neural network where both the convolutional weights and the outputs weights are parameters to be learned. Our algorithm works for a general class of (potentially overlapping) patches, including commonly used structures for computer vision tasks. Our algorithm draws ideas from (1) isotonic regression for learning neural networks and (2) landscape analysis of non-convex matrix factorization problems. We believe these findings may inspire further development in designing provable algorithms for learning neural networks and other complex models."
                    },
                    {
                        "title": "Computationally Efficient Robust Estimation of Sparse Functionals",
                        "abstract": "Many conventional statistical procedures are extremely sensitive to seemingly minor deviations from modeling assumptions. This problem is exacerbated in modern high-dimensional settings, where the problem dimension can grow with and possibly exceed the sample size. We consider the problem of robust estimation of sparse functionals, and provide a computationally and statistically efficient algorithm in the high-dimensional setting. Our theory identifies a unified set of deterministic conditions under which our algorithm guarantees accurate recovery. By further establishing that these deterministic conditions hold with high-probability for a wide range of statistical models, our theory applies to many problems of considerable interest including sparse mean and covariance estimation; sparse linear regression; and sparse generalized linear models."
                    },
                    {
                        "title": "On the Power of Over-parametrization in Neural Networks with Quadratic Activation",
                        "abstract": "We provide new theoretical insights on why over-parametrization is effective in learning neural networks. For a $k$ hidden node shallow network with quadratic activation and $n$ training data points, we show as long as $ k \\ge \\sqrt{2n}$, over-parametrization enables local search algorithms to find a \\emph{globally} optimal solution for general smooth and convex loss functions. Further, despite that the number of parameters may exceed the sample size, using theory of Rademacher complexity, we show with weight decay, the solution also generalizes well if the data is sampled from a regular distribution such as Gaussian. To prove when $k\\ge \\sqrt{2n}$, the loss function has benign landscape properties, we adopt an idea from smoothed analysis, which may have other applications in studying loss surfaces of neural networks."
                    },
                    {
                        "title": "Near-Linear Time Local Polynomial Nonparametric Estimation with Box Kernels",
                        "abstract": "Local polynomial regression (Fan and Gijbels 1996) is an important class of methods for nonparametric density estimation and regression problems. However, straightforward implementation of local polynomial regression has quadratic time complexity which hinders its applicability in large-scale data analysis. In this paper, we significantly accelerate the computation of local polynomial estimates by novel applications of multi-dimensional binary indexed trees (Fenwick 1994). Both time and space complexity of our proposed algorithm is nearly linear in the number of input data points. Simulation results confirm the efficiency and effectiveness of our proposed approach."
                    },
                    {
                        "title": "Horizon-Free Reinforcement Learning in Polynomial Time: the Power of Stationary Policies",
                        "abstract": "This paper gives the first polynomial-time algorithm for tabular Markov Decision Processes (MDP) that enjoys a regret bound \\emph{independent on the planning horizon}. Specifically, we consider tabular MDP with $S$ states, $A$ actions, a planning horizon $H$, total reward bounded by $1$, and the agent plays for $K$ episodes. We design an algorithm that achieves an $O\\left(\\mathrm{poly}(S,A,\\log K)\\sqrt{K}\\right)$ regret in contrast to existing bounds which either has an additional $\\mathrm{polylog}(H)$ dependency~\\citep{zhang2020reinforcement} or has an exponential dependency on $S$~\\citep{li2021settling}. Our result relies on a sequence of new structural lemmas establishing the approximation power, stability, and concentration property of stationary policies, which can have applications in other problems related to Markov chains."
                    },
                    {
                        "title": "Global Convergence of Gradient Descent for Asymmetric Low-Rank Matrix Factorization",
                        "abstract": "We study the asymmetric low-rank factorization problem: \\[\\min_{\\mathbf{U} \\in \\mathbb{R}^{m \\times d}, \\mathbf{V} \\in \\mathbb{R}^{n \\times d}} \\frac{1}{2}\\|\\mathbf{U}\\mathbf{V}^\\top -\\mathbf{\\Sigma}\\|_F^2\\] where $\\mathbf{\\Sigma}$ is a given matrix of size $m \\times n$ and rank $d$. This is a canonical problem that admits two difficulties in optimization: 1) non-convexity and 2) non-smoothness (due to unbalancedness of $\\mathbf{U}$ and $\\mathbf{V}$). This is also a prototype for more complex problems such as asymmetric matrix sensing and matrix completion. Despite being non-convex and non-smooth, it has been observed empirically that the randomly initialized gradient descent algorithm can solve this problem in polynomial time. Existing theories to explain this phenomenon all require artificial modifications of the algorithm, such as adding noise in each iteration and adding a balancing regularizer to balance the $\\mathbf{U}$ and $\\mathbf{V}$.   This paper presents the first proof that shows randomly initialized gradient descent converges to a global minimum of the asymmetric low-rank factorization problem with a polynomial rate. For the proof, we develop 1) a new symmetrization technique to capture the magnitudes of the symmetry and asymmetry, and 2) a quantitative perturbation analysis to approximate matrix derivatives. We believe both are useful for other related non-convex problems."
                    }
                ]
            }
        }
    },
    "2410.02629": {
        "paper_data": {
            "title": "Estimating Generalization Performance Along the Trajectory of Proximal SGD in Robust Regression",
            "url": "http://arxiv.org/abs/2410.02629v1",
            "arxiv_id": "2410.02629",
            "authors": [
                "Kai Tan",
                "Pierre C. Bellec"
            ],
            "abstract": "This paper studies the generalization performance of iterates obtained by Gradient Descent (GD), Stochastic Gradient Descent (SGD) and their proximal variants in high-dimensional robust regression problems. The number of features is comparable to the sample size and errors may be heavy-tailed. We introduce estimators that precisely track the generalization error of the iterates along the trajectory of the iterative algorithm. These estimators are provably consistent under suitable conditions. The results are illustrated through several examples, including Huber regression, pseudo-Huber regression, and their penalized variants with non-smooth regularizer. We provide explicit generalization error estimates for iterates generated from GD and SGD, or from proximal SGD in the presence of a non-smooth regularizer. The proposed risk estimates serve as effective proxies for the actual generalization error, allowing us to determine the optimal stopping iteration that minimizes the generalization error. Extensive simulations confirm the effectiveness of the proposed generalization error estimates.",
            "introduction": "   1 Introduction  Consider the linear model:    \ud835\udc9a=\ud835\udc7f\u2062\ud835\udc83\u2217+\ud835\udf3a,\ud835\udc9a\ud835\udc7fsuperscript\ud835\udc83\ud835\udf3a\\bm{y}=\\bm{X}\\bm{b}^{*}+{\\bm{\\varepsilon}},bold_italic_y = bold_italic_X bold_italic_b start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT + bold_italic_\u03b5 ,  (1)   where \ud835\udc9a\u2208\u211dn\ud835\udc9asuperscript\u211d\ud835\udc5b\\bm{y}\\in{\\mathbb{R}}^{n}bold_italic_y \u2208 blackboard_R start_POSTSUPERSCRIPT italic_n end_POSTSUPERSCRIPT is the response vector, \ud835\udc7f\u2208\u211dn\u00d7p\ud835\udc7fsuperscript\u211d\ud835\udc5b\ud835\udc5d\\bm{X}\\in{\\mathbb{R}}^{n\\times p}bold_italic_X \u2208 blackboard_R start_POSTSUPERSCRIPT italic_n \u00d7 italic_p end_POSTSUPERSCRIPT is the design matrix, \ud835\udc83\u2217\u2208\u211dpsuperscript\ud835\udc83superscript\u211d\ud835\udc5d\\bm{b}^{*}\\in{\\mathbb{R}}^{p}bold_italic_b start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT \u2208 blackboard_R start_POSTSUPERSCRIPT italic_p end_POSTSUPERSCRIPT is the unknown regression vector, and \ud835\udf3a\u2208\u211dn\ud835\udf3asuperscript\u211d\ud835\udc5b{\\bm{\\varepsilon}}\\in{\\mathbb{R}}^{n}bold_italic_\u03b5 \u2208 blackboard_R start_POSTSUPERSCRIPT italic_n end_POSTSUPERSCRIPT is the noise vector that we assume independent of \ud835\udc7f\ud835\udc7f\\bm{X}bold_italic_X. The entries of \ud835\udf3a\ud835\udf3a{\\bm{\\varepsilon}}bold_italic_\u03b5 may be heavy-tailed, for instance our working assumptions allow for infinite second moment.   For the estimation of \ud835\udc83\u2217superscript\ud835\udc83\\bm{b}^{*}bold_italic_b start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT, we consider the following regularized optimization problem    \ud835\udc83^\u2208arg\u2062min\ud835\udc83\u2208\u211dp1n\u2062\u2211i=1n\u03c1\u2062(yi\u2212\ud835\udc99i\u22a4\u2062\ud835\udc83)+g\u2062(\ud835\udc83),^\ud835\udc83subscriptargmin\ud835\udc83superscript\u211d\ud835\udc5d1\ud835\udc5bsuperscriptsubscript\ud835\udc561\ud835\udc5b\ud835\udf0csubscript\ud835\udc66\ud835\udc56superscriptsubscript\ud835\udc99\ud835\udc56top\ud835\udc83\ud835\udc54\ud835\udc83{\\widehat{\\bm{b}}}\\in\\mathop{\\rm arg\\,min}_{\\bm{b}\\in{\\mathbb{R}}^{p}}\\frac{1}% {n}\\sum_{i=1}^{n}\\rho(y_{i}-\\bm{x}_{i}^{\\top}\\bm{b})+g(\\bm{b}),over^ start_ARG bold_italic_b end_ARG \u2208 start_BIGOP roman_arg roman_min end_BIGOP start_POSTSUBSCRIPT bold_italic_b \u2208 blackboard_R start_POSTSUPERSCRIPT italic_p end_POSTSUPERSCRIPT end_POSTSUBSCRIPT divide start_ARG 1 end_ARG start_ARG italic_n end_ARG \u2211 start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_n end_POSTSUPERSCRIPT italic_\u03c1 ( italic_y start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT - bold_italic_x start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT start_POSTSUPERSCRIPT \u22a4 end_POSTSUPERSCRIPT bold_italic_b ) + italic_g ( bold_italic_b ) ,  (2)   where \u03c1:\u211d\u2192\u211d:\ud835\udf0c\u2192\u211d\u211d\\rho:{\\mathbb{R}}\\to{\\mathbb{R}}italic_\u03c1 : blackboard_R \u2192 blackboard_R is a data-fitting loss and g:\u211dp\u2192\u211d:\ud835\udc54\u2192superscript\u211d\ud835\udc5d\u211dg:{\\mathbb{R}}^{p}\\to{\\mathbb{R}}italic_g : blackboard_R start_POSTSUPERSCRIPT italic_p end_POSTSUPERSCRIPT \u2192 blackboard_R is a regularization function. In the present robust regression setting, typical examples of \u03c1\ud835\udf0c\\rhoitalic_\u03c1 include the Huber [14] loss \u03c1\u2062(r;\u03b4)=\u03b42\u2062\u222b0|r/\u03b4|min\u2061(1,x)\u2062d\u2062x\ud835\udf0c\ud835\udc5f\ud835\udeffsuperscript\ud835\udeff2superscriptsubscript0\ud835\udc5f\ud835\udeff1\ud835\udc65d\ud835\udc65\\rho(r;\\delta)=\\delta^{2}\\int_{0}^{|r/\\delta|}\\min(1,x)\\,\\text{d}xitalic_\u03c1 ( italic_r ; italic_\u03b4 ) = italic_\u03b4 start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT \u222b start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT | italic_r / italic_\u03b4 | end_POSTSUPERSCRIPT roman_min ( 1 , italic_x ) d italic_x, the Pseudo-Huber loss \u03c1\u2062(r;\u03b4)=\u03b42\u2062(1+(r/\u03b4)2\u22121)\ud835\udf0c\ud835\udc5f\ud835\udeffsuperscript\ud835\udeff21superscript\ud835\udc5f\ud835\udeff21\\rho(r;\\delta)=\\delta^{2}(\\sqrt{1+(r/\\delta)^{2}}-1)italic_\u03c1 ( italic_r ; italic_\u03b4 ) = italic_\u03b4 start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ( square-root start_ARG 1 + ( italic_r / italic_\u03b4 ) start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT end_ARG - 1 ) or other Lipschitz loss functions to combat the possible heavy-tails of the additive noise. Typical examples of penalty functions include the L1/Lasso [24] penalty g\u2062(\ud835\udc83)=\u03bb\u2062\u2225\ud835\udc83\u22251\ud835\udc54\ud835\udc83\ud835\udf06subscriptdelimited-\u2225\u2225\ud835\udc831g(\\bm{b})=\\lambda\\lVert\\bm{b}\\rVert_{1}italic_g ( bold_italic_b ) = italic_\u03bb \u2225 bold_italic_b \u2225 start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT, group-Lasso penalty [26] for grouped variables, or their non-convex variants including for instance SCAD [12] or MCP [27].   In order to solve the optimization problem (2), practitioners resort to iterative algorithms, for instance gradient descent, accelerated gradient descent, stochastic gradient descent, and the corresponding proximal methods [19] in the presence of a non-smooth regularizer. Let the algorithm starts with some initializer \ud835\udc83^1\u2208\u211dpsuperscript^\ud835\udc831superscript\u211d\ud835\udc5d{\\widehat{\\bm{b}}}^{1}\\in{\\mathbb{R}}^{p}over^ start_ARG bold_italic_b end_ARG start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT \u2208 blackboard_R start_POSTSUPERSCRIPT italic_p end_POSTSUPERSCRIPT (typically \ud835\udc83^1=\ud835\udfcesuperscript^\ud835\udc8310{\\widehat{\\bm{b}}}^{1}=\\mathbf{0}over^ start_ARG bold_italic_b end_ARG start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT = bold_0) followed by consecutive iterates \ud835\udc83^2,\ud835\udc83^3,\u2026superscript^\ud835\udc832superscript^\ud835\udc833\u2026{\\widehat{\\bm{b}}}^{2},{\\widehat{\\bm{b}}}^{3},\\dotsover^ start_ARG bold_italic_b end_ARG start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT , over^ start_ARG bold_italic_b end_ARG start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT , \u2026, where \ud835\udc83^tsuperscript^\ud835\udc83\ud835\udc61{\\widehat{\\bm{b}}}^{t}over^ start_ARG bold_italic_b end_ARG start_POSTSUPERSCRIPT italic_t end_POSTSUPERSCRIPT is typically obtained, for gradient descent and its variants as will be detailed below, from \ud835\udc83^t\u22121superscript^\ud835\udc83\ud835\udc611{\\widehat{\\bm{b}}}^{t-1}over^ start_ARG bold_italic_b end_ARG start_POSTSUPERSCRIPT italic_t - 1 end_POSTSUPERSCRIPT and by applying an additive correction involving the gradient of the objective function. Our goal of this paper is to quantify the predictive performance of each iterates \ud835\udc83^tsuperscript^\ud835\udc83\ud835\udc61{\\widehat{\\bm{b}}}^{t}over^ start_ARG bold_italic_b end_ARG start_POSTSUPERSCRIPT italic_t end_POSTSUPERSCRIPT.   We assume throughout that the covariance \ud835\udd3c\u2062[\ud835\udc99i\u2062\ud835\udc99i\u22a4]=\ud835\udeba\ud835\udd3cdelimited-[]subscript\ud835\udc99\ud835\udc56superscriptsubscript\ud835\udc99\ud835\udc56top\ud835\udeba{\\mathbb{E}}[\\bm{x}_{i}\\bm{x}_{i}^{\\top}]=\\bm{\\Sigma}blackboard_E [ bold_italic_x start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT bold_italic_x start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT start_POSTSUPERSCRIPT \u22a4 end_POSTSUPERSCRIPT ] = bold_\u03a3 of the feature vectors is finite. We measure the prediction performance using the out-of-sample error    \ud835\udd3c\u2062[(\ud835\udc99n\u2062e\u2062w\u22a4\u2062\ud835\udc83^t\u2212\ud835\udc99n\u2062e\u2062w\u22a4\u2062\ud835\udc83\u2217)2\u2223(\ud835\udc99i,yi)i\u2208[n]]=\u2016\ud835\udeba1/2\u2062(\ud835\udc83^t\u2212\ud835\udc83\u2217)\u20162\ud835\udd3cdelimited-[]conditionalsuperscriptsuperscriptsubscript\ud835\udc99\ud835\udc5b\ud835\udc52\ud835\udc64topsuperscript^\ud835\udc83\ud835\udc61superscriptsubscript\ud835\udc99\ud835\udc5b\ud835\udc52\ud835\udc64topsuperscript\ud835\udc832subscriptsubscript\ud835\udc99\ud835\udc56subscript\ud835\udc66\ud835\udc56\ud835\udc56delimited-[]\ud835\udc5bsuperscriptnormsuperscript\ud835\udeba12superscript^\ud835\udc83\ud835\udc61superscript\ud835\udc832{\\mathbb{E}}\\Bigl{[}\\Bigl{(}\\bm{x}_{new}^{\\top}{\\widehat{\\bm{b}}}^{t}-\\bm{x}_{% new}^{\\top}\\bm{b}^{*}\\Bigr{)}^{2}\\mid(\\bm{x}_{i},y_{i})_{i\\in[n]}\\Bigr{]}=\\|% \\bm{\\Sigma}^{1/2}({\\widehat{\\bm{b}}}^{t}-\\bm{b}^{*})\\|^{2}blackboard_E [ ( bold_italic_x start_POSTSUBSCRIPT italic_n italic_e italic_w end_POSTSUBSCRIPT start_POSTSUPERSCRIPT \u22a4 end_POSTSUPERSCRIPT over^ start_ARG bold_italic_b end_ARG start_POSTSUPERSCRIPT italic_t end_POSTSUPERSCRIPT - bold_italic_x start_POSTSUBSCRIPT italic_n italic_e italic_w end_POSTSUBSCRIPT start_POSTSUPERSCRIPT \u22a4 end_POSTSUPERSCRIPT bold_italic_b start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT ) start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT \u2223 ( bold_italic_x start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT",
            "references": [
                {
                    "title": "Uncertainty quantification for iterative algorithms in linear models with application to early stopping",
                    "abstract": "This paper investigates the iterates $\\hbb^1,\\dots,\\hbb^T$ obtained from iterative algorithms in high-dimensional linear regression problems, in the regime where the feature dimension $p$ is comparable with the sample size $n$, i.e., $p \\asymp n$. The analysis and proposed estimators are applicable to Gradient Descent (GD), proximal GD and their accelerated variants such as Fast Iterative Soft-Thresholding (FISTA). The paper proposes novel estimators for the generalization error of the iterate $\\hbb^t$ for any fixed iteration $t$ along the trajectory. These estimators are proved to be $\\sqrt n$-consistent under Gaussian designs. Applications to early-stopping are provided: when the generalization error of the iterates is a U-shape function of the iteration $t$, the estimates allow to select from the data an iteration $\\hat t$ that achieves the smallest generalization error along the trajectory. Additionally, we provide a technique for developing debiasing corrections and valid confidence intervals for the components of the true coefficient vector from the iterate $\\hbb^t$ at any finite iteration $t$. Extensive simulations on synthetic data illustrate the theoretical results."
                },
                {
                    "title": "Failures and Successes of Cross-Validation for Early-Stopped Gradient Descent",
                    "abstract": "We analyze the statistical properties of generalized cross-validation (GCV) and leave-one-out cross-validation (LOOCV) applied to early-stopped gradient descent (GD) in high-dimensional least squares regression. We prove that GCV is generically inconsistent as an estimator of the prediction risk of early-stopped GD, even for a well-specified linear model with isotropic features. In contrast, we show that LOOCV converges uniformly along the GD trajectory to the prediction risk. Our theory requires only mild assumptions on the data distribution and does not require the underlying regression function to be linear. Furthermore, by leveraging the individual LOOCV errors, we construct consistent estimators for the entire prediction error distribution along the GD trajectory and consistent estimators for a wide class of error functionals. This in particular enables the construction of pathwise prediction intervals based on GD iterates that have asymptotically correct nominal coverage conditional on the training data."
                },
                {
                    "title": "Hyperparameter tuning via trajectory predictions: Stochastic prox-linear methods in matrix sensing",
                    "abstract": "Motivated by the desire to understand stochastic algorithms for nonconvex optimization that are robust to their hyperparameter choices, we analyze a mini-batched prox-linear iterative algorithm for the problem of recovering an unknown rank-1 matrix from rank-1 Gaussian measurements corrupted by noise. We derive a deterministic recursion that predicts the error of this method and show, using a non-asymptotic framework, that this prediction is accurate for any batch-size and a large range of step-sizes. In particular, our analysis reveals that this method, though stochastic, converges linearly from a local initialization with a fixed step-size to a statistical error floor. Our analysis also exposes how the batch-size, step-size, and noise level affect the (linear) convergence rate and the eventual statistical estimation error, and we demonstrate how to use our deterministic predictions to perform hyperparameter tuning (e.g. step-size and batch-size selection) without ever running the method. On a technical level, our analysis is enabled in part by showing that the fluctuations of the empirical iterates around our deterministic predictions scale with the error of the previous iterate."
                },
                {
                    "title": "Iterative Approximate Cross-Validation",
                    "abstract": "Cross-validation (CV) is one of the most popular tools for assessing and selecting predictive models. However, standard CV suffers from high computational cost when the number of folds is large. Recently, under the empirical risk minimization (ERM) framework, a line of works proposed efficient methods to approximate CV based on the solution of the ERM problem trained on the full dataset. However, in large-scale problems, it can be hard to obtain the exact solution of the ERM problem, either due to limited computational resources or due to early stopping as a way of preventing overfitting. In this paper, we propose a new paradigm to efficiently approximate CV when the ERM problem is solved via an iterative first-order algorithm, without running until convergence. Our new method extends existing guarantees for CV approximation to hold along the whole trajectory of the algorithm, including at convergence, thus generalizing existing CV approximation methods. Finally, we illustrate the accuracy and computational efficiency of our method through a range of empirical studies."
                },
                {
                    "title": "Rigorous dynamical mean field theory for stochastic gradient descent methods",
                    "abstract": "We prove closed-form equations for the exact high-dimensional asymptotics of a family of first order gradient-based methods, learning an estimator (e.g. M-estimator, shallow neural network, ...) from observations on Gaussian data with empirical risk minimization. This includes widely used algorithms such as stochastic gradient descent (SGD) or Nesterov acceleration. The obtained equations match those resulting from the discretization of dynamical mean-field theory (DMFT) equations from statistical physics when applied to gradient flow. Our proof method allows us to give an explicit description of how memory kernels build up in the effective dynamics, and to include non-separable update functions, allowing datasets with non-identity covariance matrices. Finally, we provide numerical implementations of the equations for SGD with generic extensive batch-size and with constant learning rates."
                },
                {
                    "title": "Noise covariance estimation in multi-task high-dimensional linear models",
                    "abstract": "This paper studies the multi-task high-dimensional linear regression models where the noise among different tasks is correlated, in the moderately high dimensional regime where sample size $n$ and dimension $p$ are of the same order. Our goal is to estimate the covariance matrix of the noise random vectors, or equivalently the correlation of the noise variables on any pair of two tasks. Treating the regression coefficients as a nuisance parameter, we leverage the multi-task elastic-net and multi-task lasso estimators to estimate the nuisance. By precisely understanding the bias of the squared residual matrix and by correcting this bias, we develop a novel estimator of the noise covariance that converges in Frobenius norm at the rate $n^{-1/2}$ when the covariates are Gaussian. This novel estimator is efficiently computable. Under suitable conditions, the proposed estimator of the noise covariance attains the same rate of convergence as the\"oracle\"estimator that knows in advance the regression coefficients of the multi-task model. The Frobenius error bounds obtained in this paper also illustrate the advantage of this new estimator compared to a method-of-moments estimator that does not attempt to estimate the nuisance. As a byproduct of our techniques, we obtain an estimate of the generalization error of the multi-task elastic-net and multi-task lasso estimators. Extensive simulation studies are carried out to illustrate the numerical performance of the proposed method."
                },
                {
                    "title": "The high-dimensional asymptotics of first order methods with random data",
                    "abstract": "We study a class of deterministic flows in ${\\mathbb R}^{d\\times k}$, parametrized by a random matrix ${\\boldsymbol X}\\in {\\mathbb R}^{n\\times d}$ with i.i.d. centered subgaussian entries. We characterize the asymptotic behavior of these flows over bounded time horizons, in the high-dimensional limit in which $n,d\\to\\infty$ with $k$ fixed and converging aspect ratios $n/d\\to\\delta$. The asymptotic characterization we prove is in terms of a system of a nonlinear stochastic process in $k$ dimensions, whose parameters are determined by a fixed point condition. This type of characterization is known in physics as dynamical mean field theory. Rigorous results of this type have been obtained in the past for a few spin glass models. Our proof is based on time discretization and a reduction to certain iterative schemes known as approximate message passing (AMP) algorithms, as opposed to earlier work that was based on large deviations theory and stochastic processes theory. The new approach allows for a more elementary proof and implies that the high-dimensional behavior of the flow is universal with respect to the distribution of the entries of ${\\boldsymbol X}$. As specific applications, we obtain high-dimensional characterizations of gradient flow in some classical models from statistics and machine learning, under a random design assumption."
                },
                {
                    "title": "Sharp global convergence guarantees for iterative nonconvex optimization: A Gaussian process perspective",
                    "abstract": "We consider a general class of regression models with normally distributed covariates, and the associated nonconvex problem of fitting these models from data. We develop a general recipe for analyzing the convergence of iterative algorithms for this task from a random initialization. In particular, provided each iteration can be written as the solution to a convex optimization problem satisfying some natural conditions, we leverage Gaussian comparison theorems to derive a deterministic sequence that provides sharp upper and lower bounds on the error of the algorithm with sample-splitting. Crucially, this deterministic sequence accurately captures both the convergence rate of the algorithm and the eventual error floor in the finite-sample regime, and is distinct from the commonly used\"population\"sequence that results from taking the infinite-sample limit. We apply our general framework to derive several concrete consequences for parameter estimation in popular statistical models including phase retrieval and mixtures of regressions. Provided the sample size scales near-linearly in the dimension, we show sharp global convergence rates for both higher-order algorithms based on alternating updates and first-order algorithms based on subgradient descent. These corollaries, in turn, yield multiple consequences, including: (a) Proof that higher-order algorithms can converge significantly faster than their first-order counterparts (and sometimes super-linearly), even if the two share the same population update and (b) Intricacies in super-linear convergence behavior for higher-order algorithms, which can be nonstandard (e.g., with exponent 3/2) and sensitive to the noise level in the problem. We complement these results with extensive numerical experiments, which show excellent agreement with our theoretical predictions."
                },
                {
                    "title": "Derivatives and residual distribution of regularized M-estimators with application to adaptive tuning",
                    "abstract": "This paper studies M-estimators with gradient-Lipschitz loss function regularized with convex penalty in linear models with Gaussian design matrix and arbitrary noise distribution. A practical example is the robust M-estimator constructed with the Huber loss and the Elastic-Net penalty and the noise distribution has heavy-tails. Our main contributions are three-fold. (i) We provide general formulae for the derivatives of regularized M-estimators \u03b2 y , X ) where differentiation is taken with respect to both y and X ; this reveals a simple differentiability structure shared by all convex regularized M-estimators. (ii) Using these derivatives, we characterize the distribution of the residual r i = y i \u2212 x \u22a4 i (cid:98) \u03b2 in the intermediate high-dimensional regime where dimension and sample size are of the same order. (iii) Motivated by the distribution of the residuals, we propose a novel adaptive criterion to select tuning parameters of regularized M-estimators. The criterion ap-proximates the out-of-sample error up to an additive constant independent of the estimator, so that minimizing the criterion provides a proxy for minimizing the out-of-sample error. The proposed adaptive criterion does not require the knowledge of the noise distribution or of the covariance of the design. Simulated data confirms the theoretical findings, regarding both the distribution of the residuals and the success of the criterion as a proxy of the out-of-sample error. Finally our results reveal new relationships between the derivatives of (cid:98) \u03b2 ( y , X ) and the effective degrees of freedom of the M-estimator, which are of independent interest. High-dimensional statistics, Residual distribution. Data Generation Process. Simulation data are generated from a linear model y = X\u03b2 \u2217 + \u03b5 with anisotropic Gaussian design \u03a3 and heavy-tail noise vector \u03b5 . The design matrix X has n = 1001 rows and p = 1000 columns. Each row of X is i.i.d. N ( 0 , \u03a3 ) , with the same \u03a3 across all repetitions, generated once by \u03a3 = R \u22a4 R / (2 p ) with R \u2208 R 2 p \u00d7 p being a Rademacher matrix with i.i.d. entries P ( R ij = \u00b1 1) = 12 . The true signal vector \u03b2 \u2217 \u2208 R p has its first 100 coordinates set to p 1 / 2 / 100 = \u221a 10 / 10 and the rest 900 coordinates set to 0 . The noise vector \u03b5 \u2208 R n has i.i.d. entries from the t-distribution with 2 degrees of freedom (so that Var[ \u03b5 i ] = \u221e , i.e., \u03b5 i is heavy-tailed)."
                },
                {
                    "title": "SGD in the Large: Average-case Analysis, Asymptotics, and Stepsize Criticality",
                    "abstract": "We propose a new framework, inspired by random matrix theory, for analyzing the dynamics of stochastic gradient descent (SGD) when both number of samples and dimensions are large. This framework applies to any fixed stepsize and the finite sum setting. Using this new framework, we show that the dynamics of SGD on a least squares problem with random data become deterministic in the large sample and dimensional limit. Furthermore, the limiting dynamics are governed by a Volterra integral equation. This model predicts that SGD undergoes a phase transition at an explicitly given critical stepsize that ultimately affects its convergence rate, which we also verify experimentally. Finally, when input data is isotropic, we provide explicit expressions for the dynamics and average-case convergence rates (i.e., the complexity of an algorithm averaged over all possible inputs). These rates show significant improvement over the worst-case complexities."
                },
                {
                    "title": "Out-of-sample error estimate for robust M-estimators with convex penalty",
                    "abstract": "A generic out-of-sample error estimate is proposed for robust $M$-estimators regularized with a convex penalty in high-dimensional linear regression where $(X,y)$ is observed and $p,n$ are of the same order. If $\\psi$ is the derivative of the robust data-fitting loss $\\rho$, the estimate depends on the observed data only through the quantities $\\hat\\psi = \\psi(y-X\\hat\\beta)$, $X^\\top \\hat\\psi$ and the derivatives $(\\partial/\\partial y) \\hat\\psi$ and $(\\partial/\\partial y) X\\hat\\beta$ for fixed $X$. \nThe out-of-sample error estimate enjoys a relative error of order $n^{-1/2}$ in a linear model with Gaussian covariates and independent noise, either non-asymptotically when $p/n\\le \\gamma$ or asymptotically in the high-dimensional asymptotic regime $p/n\\to\\gamma'\\in(0,\\infty)$. General differentiable loss functions $\\rho$ are allowed provided that $\\psi=\\rho'$ is 1-Lipschitz. The validity of the out-of-sample error estimate holds either under a strong convexity assumption, or for the $\\ell_1$-penalized Huber M-estimator if the number of corrupted observations and sparsity of the true $\\beta$ are bounded from above by $s_*n$ for some small enough constant $s_*\\in(0,1)$ independent of $n,p$. \nFor the square loss and in the absence of corruption in the response, the results additionally yield $n^{-1/2}$-consistent estimates of the noise variance and of the generalization error. This generalizes, to arbitrary convex penalty, estimates that were previously known for the Lasso."
                },
                {
                    "title": "The Lasso with general Gaussian designs with applications to hypothesis testing",
                    "abstract": "The Lasso is a method for high-dimensional regression, which is now commonly used when the number of covariates $p$ is of the same order or larger than the number of observations $n$. Classical asymptotic normality theory is not applicable for this model due to two fundamental reasons: $(1)$ The regularized risk is non-smooth; $(2)$ The distance between the estimator $\\bf \\widehat{\\theta}$ and the true parameters vector $\\bf \\theta^\\star$ cannot be neglected. As a consequence, standard perturbative arguments that are the traditional basis for asymptotic normality fail. \nOn the other hand, the Lasso estimator can be precisely characterized in the regime in which both $n$ and $p$ are large, while $n/p$ is of order one. This characterization was first obtained in the case of standard Gaussian designs, and subsequently generalized to other high-dimensional estimation procedures. Here we extend the same characterization to Gaussian correlated designs with non-singular covariance structure. This characterization is expressed in terms of a simpler ``fixed design'' model. We establish non-asymptotic bounds on the distance between distributions of various quantities in the two models, which hold uniformly over signals $\\bf \\theta^\\star$ in a suitable sparsity class, and values of the regularization parameter. \nAs applications, we study the distribution of the debiased Lasso, and show that a degrees-of-freedom correction is necessary for computing valid confidence intervals."
                },
                {
                    "title": "A scalable estimate of the out\u2010of\u2010sample prediction error via approximate leave\u2010one\u2010out cross\u2010validation",
                    "abstract": "The paper considers the problem of out\u2010of\u2010sample risk estimation under the high dimensional settings where standard techniques such as K\u2010fold cross\u2010validation suffer from large biases. Motivated by the low bias of the leave\u2010one\u2010out cross\u2010validation method, we propose a computationally efficient closed form approximate leave\u2010one\u2010out formula ALO for a large class of regularized estimators. Given the regularized estimate, calculating ALO requires a minor computational overhead. With minor assumptions about the data\u2010generating process, we obtain a finite sample upper bound for the difference between leave\u2010one\u2010out cross\u2010validation and approximate leave\u2010one\u2010out cross\u2010validation, |LO\u2212ALO|. Our theoretical analysis illustrates that |LO\u2212ALO|\u21920 with overwhelming probability, when n,p\u2192\u221e, where the dimension p of the feature vectors may be comparable with or even greater than the number of observations, n. Despite the high dimensionality of the problem, our theoretical results do not require any sparsity assumption on the vector of regression coefficients. Our extensive numerical experiments show that |LO\u2212ALO| decreases as n and p increase, revealing the excellent finite sample performance of approximate leave\u2010one\u2010out cross\u2010validation. We further illustrate the usefulness of our proposed out\u2010of\u2010sample risk estimation method by an example of real recordings from spatially sensitive neurons (grid cells) in the medial entorhinal cortex of a rat."
                },
                {
                    "title": "The estimation error of general first order methods",
                    "abstract": "Modern large-scale statistical models require to estimate thousands to millions of parameters. This is often accomplished by iterative algorithms such as gradient descent, projected gradient descent or their accelerated versions. What are the fundamental limits to these approaches? This question is well understood from an optimization viewpoint when the underlying objective is convex. Work in this area characterizes the gap to global optimality as a function of the number of iterations. However, these results have only indirect implications in terms of the gap to statistical optimality. \nHere we consider two families of high-dimensional estimation problems: high-dimensional regression and low-rank matrix estimation, and introduce a class of `general first order methods' that aim at efficiently estimating the underlying parameters. This class of algorithms is broad enough to include classical first order optimization (for convex and non-convex objectives), but also other types of algorithms. Under a random design assumption, we derive lower bounds on the estimation error that hold in the high-dimensional asymptotics in which both the number of observations and the number of parameters diverge. These lower bounds are optimal in the sense that there exist algorithms whose estimation error matches the lower bounds up to asymptotically negligible terms. We illustrate our general results through applications to sparse phase retrieval and sparse principal component analysis."
                },
                {
                    "title": "The distribution of the Lasso: Uniform control over sparse balls and adaptive parameter tuning",
                    "abstract": "The Lasso is a popular regression method for high-dimensional problems in which the number of parameters $\\theta_1,\\dots,\\theta_N$, is larger than the number $n$ of samples: $N>n$. A useful heuristics relates the statistical properties of the Lasso estimator to that of a simple soft-thresholding denoiser,in a denoising problem in which the parameters $(\\theta_i)_{i\\le N}$ are observed in Gaussian noise, with a carefully tuned variance. Earlier work confirmed this picture in the limit $n,N\\to\\infty$, pointwise in the parameters $\\theta$, and in the value of the regularization parameter. \nHere, we consider a standard random design model and prove exponential concentration of its empirical distribution around the prediction provided by the Gaussian denoising model. Crucially, our results are uniform with respect to $\\theta$ belonging to $\\ell_q$ balls, $q\\in [0,1]$, and with respect to the regularization parameter. This allows to derive sharp results for the performances of various data-driven procedures to tune the regularization. \nOur proofs make use of Gaussian comparison inequalities, and in particular of a version of Gordon's minimax theorem developed by Thrampoulidis, Oymak, and Hassibi, which controls the optimum value of the Lasso optimization problem. Crucially, we prove a stability property of the minimizer in Wasserstein distance, that allows to characterize properties of the minimizer itself."
                },
                {
                    "title": "Estimating LASSO Risk and Noise Level",
                    "abstract": "We study the fundamental problems of variance and risk estimation in high dimensional statistical modeling. In particular, we consider the problem of learning a coefficient vector \u03b80 \u2208 \u211dp from noisy linear observations y = X\u03b80 + w \u2208 \u211dn (p > n) and the popular estimation procedure of solving the l1-penalized least squares objective known as the LASSO or Basis Pursuit DeNoising (BPDN). In this context, we develop new estimators for the l2 estimation risk $\\|\\hat{\\theta}-\\theta_0\\|_2$ and the variance of the noise when distributions of \u03b80 and w are unknown. These can be used to select the regularization parameter optimally. Our approach combines Stein's unbiased risk estimate [Ste81] and the recent results of [BM12a][BM12b] on the analysis of approximate message passing and the risk of LASSO. \n \nWe establish high-dimensional consistency of our estimators for sequences of matrices X of increasing dimensions, with independent Gaussian entries. We establish validity for a broader class of Gaussian designs, conditional on a certain conjecture from statistical physics. \n \nTo the best of our knowledge, this result is the first that provides an asymptotically consistent risk estimator for the LASSO solely based on data. In addition, we demonstrate through simulations that our variance estimation outperforms several existing methods in the literature."
                },
                {
                    "title": "Proximal Algorithms",
                    "abstract": "This monograph is about a class of optimization algorithms called proximal algorithms. Much like Newton's method is a standard tool for solving unconstrained smooth optimization problems of modest size, proximal algorithms can be viewed as an analogous tool for nonsmooth, constrained, large-scale, or distributed versions of these problems. They are very generally applicable, but are especially well-suited to problems of substantial recent interest involving large or high-dimensional datasets. Proximal methods sit at a higher level of abstraction than classical algorithms like Newton's method: the base operation is evaluating the proximal operator of a function, which itself involves solving a small convex optimization problem. These subproblems, which generalize the problem of projecting a point onto a convex set, often admit closed-form solutions or can be solved very quickly with standard or simple specialized methods. Here, we discuss the many different interpretations of proximal operators and algorithms, describe their connections to many other topics in optimization and applied mathematics, survey some popular algorithms, and provide a large number of examples of proximal operators that commonly arise in practice."
                },
                {
                    "title": "Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers",
                    "abstract": "Many problems of recent interest in statistics and machine learning can be posed in the framework of convex optimization. Due to the explosion in size and complexity of modern datasets, it is increasingly important to be able to solve problems with a very large number of features or training examples. As a result, both the decentralized collection or storage of these datasets as well as accompanying distributed solution methods are either necessary or at least highly desirable. In this review, we argue that the alternating direction method of multipliers is well suited to distributed convex optimization, and in particular to large-scale problems arising in statistics, machine learning, and related areas. The method was developed in the 1970s, with roots in the 1950s, and is equivalent or closely related to many other algorithms, such as dual decomposition, the method of multipliers, Douglas\u2013Rachford splitting, Spingarn's method of partial inverses, Dykstra's alternating projections, Bregman iterative algorithms for l1 problems, proximal methods, and others. After briefly surveying the theory and history of the algorithm, we discuss applications to a wide variety of statistical and machine learning problems of recent interest, including the lasso, sparse logistic regression, basis pursuit, covariance selection, support vector machines, and many others. We also discuss general distributed optimization, extensions to the nonconvex setting, and efficient implementation, including some details on distributed MPI and Hadoop MapReduce implementations."
                },
                {
                    "title": "Nearly unbiased variable selection under minimax concave penalty",
                    "abstract": "We propose MC+, a fast, continuous, nearly unbiased and accurate method of penalized variable selection in high-dimensional linear regression. The LASSO is fast and continuous, but biased. The bias of the LASSO may prevent consistent variable selection. Subset selection is unbiased but computationally costly. The MC+ has two elements: a minimax concave penalty (MCP) and a penalized linear unbiased selection (PLUS) algorithm. The MCP provides the convexity of the penalized loss in sparse regions to the greatest extent given certain thresholds for variable selection and unbiasedness. The PLUS computes multiple exact local minimizers of a possibly nonconvex penalized loss function in a certain main branch of the graph of critical points of the penalized loss. Its output is a continuous piecewise linear path encompassing from the origin for infinite penalty to a least squares solution for zero penalty. We prove that at a universal penalty level, the MC+ has high probability of matching the signs of the unknowns, and thus correct selection, without assuming the strong irrepresentable condition required by the LASSO. This selection consistency applies to the case of p \u00bb n, and is proved to hold for exactly the MC+ solution among possibly many local minimizers. We prove that the MC+ attains certain minimax convergence rates in probability for the estimation of regression coefficients in l r balls. We use the SURE method to derive degrees of freedom and C p -type risk estimates for general penalized LSE, including the LASSO and MC+ estimators, and prove their unbiasedness. Based on the estimated degrees of freedom, we propose an estimator of the noise level for proper choice of the penalty level. For full rank designs and general sub-quadratic penalties, we provide necessary and sufficient conditions for the continuity of the penalized LSE. Simulation results overwhelmingly support our claim of superior variable selection properties and demonstrate the computational efficiency of the proposed method."
                },
                {
                    "title": "Model selection and estimation in regression with grouped variables",
                    "abstract": "Summary.\u2002 We consider the problem of selecting grouped variables (factors) for accurate prediction in regression. Such a problem arises naturally in many practical situations with the multifactor analysis\u2010of\u2010variance problem as the most important and well\u2010known example. Instead of selecting factors by stepwise backward elimination, we focus on the accuracy of estimation and consider extensions of the lasso, the LARS algorithm and the non\u2010negative garrotte for factor selection. The lasso, the LARS algorithm and the non\u2010negative garrotte are recently proposed regression methods that can be used to select individual variables. We study and propose efficient algorithms for the extensions of these methods for factor selection and show that these extensions give superior performance to the traditional stepwise backward elimination method in factor selection problems. We study the similarities and the differences between these methods. Simulations and real examples are used to illustrate the methods."
                },
                {
                    "title": "Variable Selection via Nonconcave Penalized Likelihood and its Oracle Properties",
                    "abstract": "Variable selection is fundamental to high-dimensional statistical modeling, including nonparametric regression. Many approaches in use are stepwise selection procedures, which can be computationally expensive and ignore stochastic errors in the variable selection process. In this article, penalized likelihood approaches are proposed to handle these kinds of problems. The proposed methods select variables and estimate coefficients simultaneously. Hence they enable us to construct confidence intervals for estimated parameters. The proposed approaches are distinguished from others in that the penalty functions are symmetric, nonconcave on (0, \u221e), and have singularities at the origin to produce sparse solutions. Furthermore, the penalty functions should be bounded by a constant to reduce bias and satisfy certain conditions to yield continuous solutions. A new algorithm is proposed for optimizing penalized likelihood functions. The proposed ideas are widely applicable. They are readily applied to a variety of parametric models such as generalized linear models and robust regression models. They can also be applied easily to nonparametric modeling by using wavelets and splines. Rates of convergence of the proposed penalized likelihood estimators are established. Furthermore, with proper choice of regularization parameters, we show that the proposed estimators perform as well as the oracle procedure in variable selection; namely, they work as well as if the correct submodel were known. Our simulation shows that the newly proposed methods compare favorably with other variable selection techniques. Furthermore, the standard error formulas are tested to be accurate enough for practical applications."
                },
                {
                    "title": "A Comparison of GCV and GML for Choosing the Smoothing Parameter in the Generalized Spline Smoothing Problem",
                    "abstract": "On considere des procedures de lissage spline etudiees en 1978 et 1983 et leur extension a la resolution d'equations d'operateurs lineaires avec donnees bruitees"
                },
                {
                    "title": "Regression Shrinkage and Selection via the Lasso",
                    "abstract": "SUMMARY We propose a new method for estimation in linear models. The 'lasso' minimizes the residual sum of squares subject to the sum of the absolute value of the coefficients being less than a constant. Because of the nature of this constraint it tends to produce some coefficients that are exactly 0 and hence gives interpretable models. Our simulation studies suggest that the lasso enjoys some of the favourable properties of both subset selection and ridge regression. It produces interpretable models like subset selection and exhibits the stability of ridge regression. There is also an interesting relationship with recent work in adaptive function estimation by Donoho and Johnstone. The lasso idea is quite general and can be applied in a variety of statistical models: extensions to generalized regression models and tree-based models are briefly described."
                }
            ],
            "categories": [
                "math.ST",
                "cs.LG",
                "stat.ME",
                "stat.TH"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively quantify the predictive performance of iterates in robust regression models with heavy-tailed noise?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses the limitations of existing regression techniques in handling heavy-tailed noise, which is common in real-world data. By improving the understanding of predictive performance in iterative algorithms, this research could lead to more robust statistical methods, enhancing the reliability of predictions in various applications such as finance, healthcare, and social sciences. Furthermore, it could inspire future research to explore new loss functions and regularization techniques that are better suited for complex data distributions.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the complexities of heavy-tailed distributions, which can lead to significant deviations in predictive performance. Naive approaches, such as standard regression techniques, may fail because they do not account for the influence of outliers or the non-standard behavior of the noise. Additionally, the iterative nature of the optimization process introduces further complications, as each iteration's performance can be influenced by the previous estimates, making it difficult to establish a clear relationship between the iterates and their predictive accuracy. Technical obstacles include the need for robust convergence criteria and the development of effective metrics to evaluate performance under these conditions.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on traditional regression methods that assume normally distributed errors, neglecting the implications of heavy-tailed noise. Existing solutions may lack the necessary robustness or fail to provide a comprehensive framework for evaluating iterative performance. Barriers to solving this problem include a limited understanding of the behavior of iterative algorithms in the presence of heavy-tailed noise and the absence of suitable metrics for performance evaluation. Our approach differs by explicitly addressing these gaps, focusing on the predictive performance of each iterate and incorporating robust loss functions that are designed to handle heavy-tailed distributions.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves formulating a robust regression optimization problem that incorporates a suitable loss function (e.g., Huber or Pseudo-Huber loss) and a regularization term (e.g., L1 or group-Lasso penalty). We will utilize a dataset that exhibits heavy-tailed noise characteristics and apply iterative algorithms such as gradient descent to estimate the regression coefficients. The key metric for evaluation will be the out-of-sample error, specifically measuring the"
            }
        },
        "author_data": {
            "173c21a2-f5cf-410e-ad15-902a883bac73": {
                "pk": "173c21a2-f5cf-410e-ad15-902a883bac73",
                "name": "Kai Tan",
                "collaborators": [
                    "Pierre C. Bellec",
                    "Chengqing Li",
                    "Gabriel Romon",
                    "Pierre C Bellec"
                ],
                "domain": [
                    "Statistical Learning",
                    "High-Dimensional Data",
                    "Mathematical Modeling",
                    "Regression Analysis"
                ],
                "publications": [
                    {
                        "title": "The Graph Structure of Baker's maps Implemented on a Computer",
                        "abstract": "The complex dynamics of baker's map and its variants in an infinite-precision mathematical domain have been extensively analyzed in the past five decades. However, their real structure implemented in a finite-precision computer remains unclear. This paper gives an explicit formulation for the quantized baker's map and its extension into higher dimensions. Our study reveals certain properties, such as the in-degree distribution in the state-mapping network approaching a constant with increasing precision, and a consistent maximum in-degree across various levels of fixed-point arithmetic precision. We also observe a fractal pattern in baker's map functional graph as precision increases, characterized by fractal dimensions. We then thoroughly examine the structural nuances of functional graphs created by the higher-dimensional baker's map (HDBM) in both fixed-point and floating-point arithmetic domains. An interesting aspect of our study is the use of interval arithmetic to establish a relationship between the HDBM's functional graphs across these two computational domains. A particularly intriguing discovery is the emergence of a semi-fractal pattern within the functional graph of a specific baker's map variant, observed as the precision is incrementally increased. The insights gained from our research offer a foundational understanding for the dynamic analysis and application of baker's map and its variants in various domains."
                    },
                    {
                        "title": "Multinomial Logistic Regression: Asymptotic Normality on Null Covariates in High-Dimensions",
                        "abstract": "This paper investigates the asymptotic distribution of the maximum-likelihood estimate (MLE) in multinomial logistic models in the high-dimensional regime where dimension and sample size are of the same order. While classical large-sample theory provides asymptotic normality of the MLE under certain conditions, such classical results are expected to fail in high-dimensions as documented for the binary logistic case in the seminal work of Sur and Cand\\`es [2019]. We address this issue in classification problems with 3 or more classes, by developing asymptotic normality and asymptotic chi-square results for the multinomial logistic MLE (also known as cross-entropy minimizer) on null covariates. Our theory leads to a new methodology to test the significance of a given feature. Extensive simulation studies on synthetic data corroborate these asymptotic results and confirm the validity of proposed p-values for testing the significance of a given feature."
                    },
                    {
                        "title": "Uncertainty quantification for iterative algorithms in linear models with application to early stopping",
                        "abstract": "This paper investigates the iterates $\\hbb^1,\\dots,\\hbb^T$ obtained from iterative algorithms in high-dimensional linear regression problems, in the regime where the feature dimension $p$ is comparable with the sample size $n$, i.e., $p \\asymp n$. The analysis and proposed estimators are applicable to Gradient Descent (GD), proximal GD and their accelerated variants such as Fast Iterative Soft-Thresholding (FISTA). The paper proposes novel estimators for the generalization error of the iterate $\\hbb^t$ for any fixed iteration $t$ along the trajectory. These estimators are proved to be $\\sqrt n$-consistent under Gaussian designs. Applications to early-stopping are provided: when the generalization error of the iterates is a U-shape function of the iteration $t$, the estimates allow to select from the data an iteration $\\hat t$ that achieves the smallest generalization error along the trajectory. Additionally, we provide a technique for developing debiasing corrections and valid confidence intervals for the components of the true coefficient vector from the iterate $\\hbb^t$ at any finite iteration $t$. Extensive simulations on synthetic data illustrate the theoretical results."
                    },
                    {
                        "title": "Noise Covariance Estimation in Multi-Task High-dimensional Linear Models",
                        "abstract": "This paper studies the multi-task high-dimensional linear regression models where the noise among different tasks is correlated, in the moderately high dimensional regime where sample size $n$ and dimension $p$ are of the same order. Our goal is to estimate the covariance matrix of the noise random vectors, or equivalently the correlation of the noise variables on any pair of two tasks. Treating the regression coefficients as a nuisance parameter, we leverage the multi-task elastic-net and multi-task lasso estimators to estimate the nuisance. By precisely understanding the bias of the squared residual matrix and by correcting this bias, we develop a novel estimator of the noise covariance that converges in Frobenius norm at the rate $n^{-1/2}$ when the covariates are Gaussian. This novel estimator is efficiently computable.   Under suitable conditions, the proposed estimator of the noise covariance attains the same rate of convergence as the \"oracle\" estimator that knows in advance the regression coefficients of the multi-task model. The Frobenius error bounds obtained in this paper also illustrate the advantage of this new estimator compared to a method-of-moments estimator that does not attempt to estimate the nuisance.   As a byproduct of our techniques, we obtain an estimate of the generalization error of the multi-task elastic-net and multi-task lasso estimators. Extensive simulation studies are carried out to illustrate the numerical performance of the proposed method."
                    }
                ]
            },
            "614ed7c8-b002-4e90-8e13-3198ce1a8c14": {
                "pk": "614ed7c8-b002-4e90-8e13-3198ce1a8c14",
                "name": "Pierre C. Bellec",
                "collaborators": [
                    "Alexandre B. Tsybakov",
                    "Guillaume Lecu\u00e9"
                ],
                "domain": [
                    "Statistical Learning",
                    "High-Dimensional Statistics",
                    "Model Selection",
                    "Regression Analysis"
                ],
                "publications": [
                    {
                        "title": "Aggregation of supports along the Lasso path",
                        "abstract": "In linear regression with fixed design, we propose two procedures that aggregate a data-driven collection of supports. The collection is a subset of the $2^p$ possible supports and both its cardinality and its elements can depend on the data. The procedures satisfy oracle inequalities with no assumption on the design matrix. Then we use these procedures to aggregate the supports that appear on the regularization path of the Lasso in order to construct an estimator that mimics the best Lasso estimator. If the restricted eigenvalue condition on the design matrix is satisfied, then this estimator achieves optimal prediction bounds. Finally, we discuss the computational cost of these procedures."
                    },
                    {
                        "title": "Optimal exponential bounds for aggregation of density estimators",
                        "abstract": "We consider the problem of model selection type aggregation in the context of density estimation. We first show that empirical risk minimization is sub-optimal for this problem and it shares this property with the exponential weights aggregate, empirical risk minimization over the convex hull of the dictionary functions, and all selectors. Using a penalty inspired by recent works on the $Q$-aggregation procedure, we derive a sharp oracle inequality in deviation under a simple boundedness assumption and we show that the rate is optimal in a minimax sense. Unlike the procedures based on exponential weights, this estimator is fully adaptive under the uniform prior. In particular, its construction does not rely on the sup-norm of the unknown density. By providing lower bounds with exponential tails, we show that the deviation term appearing in the sharp oracle inequalities cannot be improved."
                    },
                    {
                        "title": "Sharp oracle inequalities for Least Squares estimators in shape restricted regression",
                        "abstract": "The performance of Least Squares (LS) estimators is studied in isotonic, unimodal and convex regression. Our results have the form of sharp oracle inequalities that account for the model misspecification error. In isotonic and unimodal regression, the LS estimator achieves the nonparametric rate $n^{-2/3}$ as well as a parametric rate of order $k/n$ up to logarithmic factors, where $k$ is the number of constant pieces of the true parameter.   In univariate convex regression, the LS estimator satisfies an adaptive risk bound of order $q/n$ up to logarithmic factors, where $q$ is the number of affine pieces of the true regression function. This adaptive risk bound holds for any design points. While Guntuboyina and Sen (2013) established that the nonparametric rate of convex regression is of order $n^{-4/5}$ for equispaced design points, we show that the nonparametric rate of convex regression can be as slow as $n^{-2/3}$ for some worst-case design points. This phenomenon can be explained as follows: Although convexity brings more structure than unimodality, for some worst-case design points this extra structure is uninformative and the nonparametric rates of unimodal regression and convex regression are both $n^{-2/3}$."
                    },
                    {
                        "title": "Optimal bounds for aggregation of affine estimators",
                        "abstract": "We study the problem of aggregation of estimators when the estimators are not independent of the data used for aggregation and no sample splitting is allowed. If the estimators are deterministic vectors, it is well known that the minimax rate of aggregation is of order $\\log(M)$, where $M$ is the number of estimators to aggregate. It is proved that for affine estimators, the minimax rate of aggregation is unchanged: it is possible to handle the linear dependence between the affine estimators and the data used for aggregation at no extra cost. The minimax rate is not impacted either by the variance of the affine estimators, or any other measure of their statistical complexity. The minimax rate is attained with a penalized procedure over the convex hull of the estimators, for a penalty that is inspired from the $Q$-aggregation procedure. The results follow from the interplay between the penalty, strong convexity and concentration."
                    },
                    {
                        "title": "Adaptive confidence sets in shape restricted regression",
                        "abstract": "A simple construction of adaptive confidence sets is proposed in isotonic, convex and unimodal regression. In univariate isotonic regression, the proposed confidence set enjoys uniform coverage over all non-decreasing regression functions. Furthermore, the diameter of the proposed confidence set automatically adapts to the unknown number of pieces of the true parameter, in the sense that the diameter is bounded from above by the minimax risk over the class of $k$-piecewise constant functions. The diameter of the confidence set is a simple increasing function of the number of jumps of the isotonic least-squares estimate.   A similar construction is proposed in convex regression where the true regression function is convex and piecewise affine. Here, the confidence set enjoys uniform coverage and its diameter automatically adapt to the number of affine pieces of the true regression function. The diameter of the confidence set is an increasing function of the number of affine pieces of the convex least-squares estimate.   We explain how to extend this technique to a non-convex set by proposing a similar adaptive confidence set in unimodal regression. The confidence set automatically adapts to the number of jumps of the true unimodal regression function and its diameter is an increasing function of the number of jumps of the unimodal least-squares estimate."
                    },
                    {
                        "title": "Bounds on the prediction error of penalized least squares estimators with convex penalty",
                        "abstract": "This paper considers the penalized least squares estimator with arbitrary convex penalty. When the observation noise is Gaussian, we show that the prediction error is a subgaussian random variable concentrated around its median. We apply this concentration property to derive sharp oracle inequalities for the prediction error of the LASSO, the group LASSO and the SLOPE estimators, both in probability and in expectation. In contrast to the previous work on the LASSO type methods, our oracle inequalities in probability are obtained at any confidence level for estimators with tuning parameters that do not depend on the confidence level. This is also the reason why we are able to establish sparsity oracle bounds in expectation for the LASSO type estimators, while the previously known techniques did not allow for the control of the expected risk. In addition, we show that the concentration rate in the oracle inequalities is better than it was commonly understood before."
                    },
                    {
                        "title": "Sharp oracle bounds for monotone and convex regression through aggregation",
                        "abstract": "We derive oracle inequalities for the problems of isotonic and convex regression using the combination of $Q$-aggregation procedure and sparsity pattern aggregation. This improves upon the previous results including the oracle inequalities for the constrained least squares estimator. One of the improvements is that our oracle inequalities are sharp, i.e., with leading constant 1. It allows us to obtain bounds for the minimax regret thus accounting for model misspecification, which was not possible based on the previous results. Another improvement is that we obtain oracle inequalities both with high probability and in expectation."
                    },
                    {
                        "title": "Towards the study of least squares estimators with convex penalty",
                        "abstract": "Penalized least squares estimation is a popular technique in high-dimensional statistics. It includes such methods as the LASSO, the group LASSO, and the nuclear norm penalized least squares. The existing theory of these methods is not fully satisfying since it allows one to prove oracle inequalities with fixed high probability only for the estimators depending on this probability. Furthermore, the control of compatibility factors appearing in the oracle bounds is often not explicit. Some very recent developments suggest that the theory of oracle inequalities can be revised in an improved way. In this paper, we provide an overview of ideas and tools leading to such an improved theory. We show that, along with overcoming the disadvantages mentioned above, the methodology extends to the hilbertian framework and it applies to a large class of convex penalties. This paper is partly expository. In particular, we provide adapted proofs of some results from other recent work."
                    },
                    {
                        "title": "Slope meets Lasso: improved oracle bounds and optimality",
                        "abstract": "We show that two polynomial time methods, a Lasso estimator with adaptively chosen tuning parameter and a Slope estimator, adaptively achieve the exact minimax prediction and $\\ell_2$ estimation rate $(s/n)\\log (p/s)$ in high-dimensional linear regression on the class of $s$-sparse target vectors in $\\mathbb R^p$. This is done under the Restricted Eigenvalue (RE) condition for the Lasso and under a slightly more constraining assumption on the design for the Slope. The main results have the form of sharp oracle inequalities accounting for the model misspecification error. The minimax optimal bounds are also obtained for the $\\ell_q$ estimation errors with $1\\le q\\le 2$ when the model is well-specified. The results are non-asymptotic, and hold both in probability and in expectation. The assumptions that we impose on the design are satisfied with high probability for a large class of random matrices with independent and possibly anisotropically distributed rows. We give a comparative analysis of conditions, under which oracle bounds for the Lasso and Slope estimators can be obtained. In particular, we show that several known conditions, such as the RE condition and the sparse eigenvalue condition are equivalent if the $\\ell_2$-norms of regressors are uniformly bounded."
                    }
                ]
            }
        }
    },
    "2305.14122": {
        "paper_data": {
            "title": "Transferring Learning Trajectories of Neural Networks",
            "url": "http://arxiv.org/abs/2305.14122v2",
            "arxiv_id": "2305.14122",
            "authors": [
                "Daiki Chijiwa"
            ],
            "abstract": "Training deep neural networks (DNNs) is computationally expensive, which is problematic especially when performing duplicated or similar training runs in model ensemble or fine-tuning pre-trained models, for example. Once we have trained one DNN on some dataset, we have its learning trajectory (i.e., a sequence of intermediate parameters during training) which may potentially contain useful information for learning the dataset. However, there has been no attempt to utilize such information of a given learning trajectory for another training. In this paper, we formulate the problem of \"transferring\" a given learning trajectory from one initial parameter to another one (learning transfer problem) and derive the first algorithm to approximately solve it by matching gradients successively along the trajectory via permutation symmetry. We empirically show that the transferred parameters achieve non-trivial accuracy before any direct training, and can be trained significantly faster than training from scratch.",
            "introduction": "   1 Introduction  Enormous computational cost is a major issue in deep learning, especially in training large-scale neural networks (NNs). Their highly non-convex objective and high-dimensional parameters make their training difficult and inefficient. Toward a better understanding of training processes of NNs, their loss landscapes\u00a0(Hochreiter & Schmidhuber, 1997; Choromanska et\u00a0al., 2015) have been actively studied from viewpoints of optimization\u00a0(Haeffele & Vidal, 2017; Li & Yuan, 2017; Yun et\u00a0al., 2018) and geometry\u00a0(Freeman & Bruna, 2017; Simsek et\u00a0al., 2021). One of the geometric approaches to loss landscapes is mode connectivity\u00a0(Garipov et\u00a0al., 2018; Draxler et\u00a0al., 2018), which shows the existence of low-loss curves between any two optimal solutions trained with different random initializations or data ordering. This indicates a surprising connection between seemingly different independent trainings.   Linear mode connectivity (LMC), a special case of mode connectivity, focuses on whether or not two optimal solutions are connected by a low-loss linear path, which is originally studied in relation to neural network pruning\u00a0(Frankle et\u00a0al., 2020). It is known that the solutions trained from the same initialization (and data ordering in the early phase) tend to be linearly mode connected\u00a0(Nagarajan & Kolter, 2019; Frankle et\u00a0al., 2020), but otherwise they cannot be linearly connected in general. However, Entezari et\u00a0al. (2021) observed that even two solutions trained from different random initializations can be linearly connected by an appropriate permutation symmetry. Ainsworth et\u00a0al. (2023) developed an efficient method to find such permutations and confirmed the same phenomena with modern NN architectures. These observations strength the expectation on some sort of similarity between two independent training runs even from different initializations, via permutation symmetry.   In this paper, motivated by these observations, we make the first attempt to leverage such similarity between independent training processes for efficient training. In particular, we introduce a novel problem called learning transfer problem, which aims to reduce training costs for seemingly duplicated training runs on the same dataset, such as model ensemble or knowledge distillation, by transferring a learning trajectory for one initial parameter to another one without actual training. The problem statement is informally stated as follows:   Learning transfer problem (informal).  Suppose that a source learning trajectory (\u03b810,\u22ef,\u03b81T)superscriptsubscript\ud835\udf0310normal-\u22efsuperscriptsubscript\ud835\udf031\ud835\udc47(\\theta_{1}^{0},\\cdots,\\theta_{1}^{T})( italic_\u03b8 start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT , \u22ef , italic_\u03b8 start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_T end_POSTSUPERSCRIPT ) is given for some initial parameter \u03b810superscriptsubscript\ud835\udf0310\\theta_{1}^{0}italic_\u03b8 start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT. Given another initial parameter \u03b820superscriptsubscript\ud835\udf0320\\theta_{2}^{0}italic_\u03b8 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT, called target initialization, how can we synthesize the learning trajectory (\u03b820,\u22ef,\u03b82T)superscriptsubscript\ud835\udf0320normal-\u22efsuperscriptsubscript\ud835\udf032\ud835\udc47(\\theta_{2}^{0},\\cdots,\\theta_{2}^{T})( italic_\u03b8 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT , \u22ef , italic_\u03b8 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_T end_POSTSUPERSCRIPT ) for \u03b820superscriptsubscript\ud835\udf0320\\theta_{2}^{0}italic_\u03b8 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT efficiently?    To tackle this problem, as illustrated in Figure\u00a02, we take an approach to transform the source trajectory (\u03b810,\u22ef,\u03b81T)superscriptsubscript\ud835\udf0310\u22efsuperscriptsubscript\ud835\udf031\ud835\udc47(\\theta_{1}^{0},\\cdots,\\theta_{1}^{T})( italic_\u03b8 start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT , \u22ef , italic_\u03b8 start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_T end_POSTSUPERSCRIPT ) by an appropriate permutation symmetry \u03c0\ud835\udf0b\\piitalic_\u03c0 as in the previous works of LMC. In Section\u00a03, we formulate the learning transfer problem as a non-linear optimization problem for \u03c0\ud835\udf0b\\piitalic_\u03c0. We also investigate how much the source trajectory for \u03b810superscriptsubscript\ud835\udf0310\\theta_{1}^{0}italic_\u03b8 start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT can be transformed close to the target trajectory for \u03b820superscriptsubscript\ud835\udf0320\\theta_{2}^{0}italic_\u03b8 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 0",
            "references": [
                {
                    "title": "On the Variance of Neural Network Training with respect to Test Sets and Distributions",
                    "abstract": "Typical neural network trainings have substantial variance in test-set performance between repeated runs, impeding hyperparameter comparison and training reproducibility. In this work we present the following results towards understanding this variation. (1) Despite having significant variance on their test-sets, we demonstrate that standard CIFAR-10 and ImageNet trainings have little variance in performance on the underlying test-distributions from which their test-sets are sampled. (2) We show that these trainings make approximately independent errors on their test-sets. That is, the event that a trained network makes an error on one particular example does not affect its chances of making errors on other examples, relative to their average rates over repeated runs of training with the same hyperparameters. (3) We prove that the variance of neural network trainings on their test-sets is a downstream consequence of the class-calibration property discovered by Jiang et al. (2021). Our analysis yields a simple formula which accurately predicts variance for the binary classification case. (4) We conduct preliminary studies of data augmentation, learning rate, finetuning instability and distribution-shift through the lens of variance between runs."
                },
                {
                    "title": "Knowledge Distillation for Efficient Sequences of Training Runs",
                    "abstract": "In many practical scenarios -- like hyperparameter search or continual retraining with new data -- related training runs are performed many times in sequence. Current practice is to train each of these models independently from scratch. We study the problem of exploiting the computation invested in previous runs to reduce the cost of future runs using knowledge distillation (KD). We find that augmenting future runs with KD from previous runs dramatically reduces the time necessary to train these models, even taking into account the overhead of KD. We improve on these results with two strategies that reduce the overhead of KD by 80-90% with minimal effect on accuracy and vast pareto-improvements in overall cost. We conclude that KD is a promising avenue for reducing the cost of the expensive preparatory work that precedes training final models in practice."
                },
                {
                    "title": "Editing Models with Task Arithmetic",
                    "abstract": "Changing how pre-trained models behave -- e.g., improving their performance on a downstream task or mitigating biases learned during pre-training -- is a common practice when developing machine learning systems. In this work, we propose a new paradigm for steering the behavior of neural networks, centered around \\textit{task vectors}. A task vector specifies a direction in the weight space of a pre-trained model, such that movement in that direction improves performance on the task. We build task vectors by subtracting the weights of a pre-trained model from the weights of the same model after fine-tuning on a task. We show that these task vectors can be modified and combined together through arithmetic operations such as negation and addition, and the behavior of the resulting model is steered accordingly. Negating a task vector decreases performance on the target task, with little change in model behavior on control tasks. Moreover, adding task vectors together can improve performance on multiple tasks at once. Finally, when tasks are linked by an analogy relationship of the form ``A is to B as C is to D\", combining task vectors from three of the tasks can improve performance on the fourth, even when no data from the fourth task is used for training. Overall, our experiments with several models, modalities and tasks show that task arithmetic is a simple, efficient and effective way of editing models."
                },
                {
                    "title": "Mechanistic Mode Connectivity",
                    "abstract": "We study neural network loss landscapes through the lens of mode connectivity, the observation that minimizers of neural networks retrieved via training on a dataset are connected via simple paths of low loss. Specifically, we ask the following question: are minimizers that rely on different mechanisms for making their predictions connected via simple paths of low loss? We provide a definition of mechanistic similarity as shared invariances to input transformations and demonstrate that lack of linear connectivity between two models implies they use dissimilar mechanisms for making their predictions. Relevant to practice, this result helps us demonstrate that naive fine-tuning on a downstream dataset can fail to alter a model's mechanisms, e.g., fine-tuning can fail to eliminate a model's reliance on spurious attributes. Our analysis also motivates a method for targeted alteration of a model's mechanisms, named connectivity-based fine-tuning (CBFT), which we analyze using several synthetic datasets for the task of reducing a model's reliance on spurious attributes."
                },
                {
                    "title": "REPAIR: REnormalizing Permuted Activations for Interpolation Repair",
                    "abstract": "In this paper we look into the conjecture of Entezari et al. (2021) which states that if the permutation invariance of neural networks is taken into account, then there is likely no loss barrier to the linear interpolation between SGD solutions. First, we observe that neuron alignment methods alone are insufficient to establish low-barrier linear connectivity between SGD solutions due to a phenomenon we call variance collapse: interpolated deep networks suffer a collapse in the variance of their activations, causing poor performance. Next, we propose REPAIR (REnormalizing Permuted Activations for Interpolation Repair) which mitigates variance collapse by rescaling the preactivations of such interpolated networks. We explore the interaction between our method and the choice of normalization layer, network width, and depth, and demonstrate that using REPAIR on top of neuron alignment methods leads to 60%-100% relative barrier reduction across a wide variety of architecture families and tasks. In particular, we report a 74% barrier reduction for ResNet50 on ImageNet and 90% barrier reduction for ResNet18 on CIFAR10."
                },
                {
                    "title": "Plateau in Monotonic Linear Interpolation - A \"Biased\" View of Loss Landscape for Deep Networks",
                    "abstract": "Monotonic linear interpolation (MLI) - on the line connecting a random initialization with the minimizer it converges to, the loss and accuracy are monotonic - is a phenomenon that is commonly observed in the training of neural networks. Such a phenomenon may seem to suggest that optimization of neural networks is easy. In this paper, we show that the MLI property is not necessarily related to the hardness of optimization problems, and empirical observations on MLI for deep neural networks depend heavily on biases. In particular, we show that interpolating both weights and biases linearly leads to very different influences on the final output, and when different classes have different last-layer biases on a deep network, there will be a long plateau in both the loss and accuracy interpolation (which existing theory of MLI cannot explain). We also show how the last-layer biases for different classes can be different even on a perfectly balanced dataset using a simple model. Empirically we demonstrate that similar intuitions hold on practical networks and realistic datasets."
                },
                {
                    "title": "Random initialisations performing above chance and how to find them",
                    "abstract": "Neural networks trained with stochastic gradient descent (SGD) starting from different random initialisations typically find functionally very similar solutions, raising the question of whether there are meaningful differences between different SGD solutions. Entezari et al.\\ recently conjectured that despite different initialisations, the solutions found by SGD lie in the same loss valley after taking into account the permutation invariance of neural networks. Concretely, they hypothesise that any two solutions found by SGD can be permuted such that the linear interpolation between their parameters forms a path without significant increases in loss. Here, we use a simple but powerful algorithm to find such permutations that allows us to obtain direct empirical evidence that the hypothesis is true in fully connected networks. Strikingly, we find that two networks already live in the same loss valley at the time of initialisation and averaging their random, but suitably permuted initialisation performs significantly above chance. In contrast, for convolutional architectures, our evidence suggests that the hypothesis does not hold. Especially in a large learning rate regime, SGD seems to discover diverse modes."
                },
                {
                    "title": "Git Re-Basin: Merging Models modulo Permutation Symmetries",
                    "abstract": "The success of deep learning is due in large part to our ability to solve certain massive non-convex optimization problems with relative ease. Though non-convex optimization is NP-hard, simple algorithms -- often variants of stochastic gradient descent -- exhibit surprising effectiveness in fitting large neural networks in practice. We argue that neural network loss landscapes often contain (nearly) a single basin after accounting for all possible permutation symmetries of hidden units a la Entezari et al. 2021. We introduce three algorithms to permute the units of one model to bring them into alignment with a reference model in order to merge the two models in weight space. This transformation produces a functionally equivalent set of weights that lie in an approximately convex basin near the reference model. Experimentally, we demonstrate the single basin phenomenon across a variety of model architectures and datasets, including the first (to our knowledge) demonstration of zero-barrier linear mode connectivity between independently trained ResNet models on CIFAR-10. Additionally, we identify intriguing phenomena relating model width and training time to mode connectivity. Finally, we discuss shortcomings of the linear mode connectivity hypothesis, including a counterexample to the single basin theory."
                },
                {
                    "title": "Patching open-vocabulary models by interpolating weights",
                    "abstract": "Open-vocabulary models like CLIP achieve high accuracy across many image classification tasks. However, there are still settings where their zero-shot performance is far from optimal. We study model patching, where the goal is to improve accuracy on specific tasks without degrading accuracy on tasks where performance is already adequate. Towards this goal, we introduce PAINT, a patching method that uses interpolations between the weights of a model before fine-tuning and the weights after fine-tuning on a task to be patched. On nine tasks where zero-shot CLIP performs poorly, PAINT increases accuracy by 15 to 60 percentage points while preserving accuracy on ImageNet within one percentage point of the zero-shot model. PAINT also allows a single model to be patched on multiple tasks and improves with model scale. Furthermore, we identify cases of broad transfer, where patching on one task increases accuracy on other tasks even when the tasks have disjoint classes. Finally, we investigate applications beyond common benchmarks such as counting or reducing the impact of typographic attacks on CLIP. Our findings demonstrate that it is possible to expand the set of tasks on which open-vocabulary models achieve high accuracy without re-training them from scratch."
                },
                {
                    "title": "Branch-Train-Merge: Embarrassingly Parallel Training of Expert Language Models",
                    "abstract": "We present Branch-Train-Merge (BTM), a communication-efficient algorithm for embarrassingly parallel training of large language models (LLMs). We show it is possible to independently train subparts of a new class of LLMs on different subsets of the data, eliminating the massive multi-node synchronization currently required to train LLMs. BTM learns a set of independent expert LMs (ELMs), each specialized to a different textual domain, such as scientific or legal text. These ELMs can be added and removed to update data coverage, ensembled to generalize to new domains, or averaged to collapse back to a single LM for efficient inference. New ELMs are learned by branching from (mixtures of) ELMs in the current set, further training the parameters on data for the new domain, and then merging the resulting model back into the set for future use. Experiments show that BTM improves in- and out-of-domain perplexities as compared to GPT-style Transformer LMs, when controlling for training cost. Through extensive analysis, we show that these results are robust to different ELM initialization schemes, but require expert domain specialization; LM ensembles with random data splits do not perform well. We also present a study of scaling BTM into a new corpus of 64 domains (192B whitespace-separated tokens in total); the resulting LM (22.4B total parameters) performs as well as a Transformer LM trained with 2.5 times more compute. These gains grow with the number of domains, suggesting more aggressive parallelism could be used to efficiently train larger models in future work."
                },
                {
                    "title": "Linear Connectivity Reveals Generalization Strategies",
                    "abstract": "It is widely accepted in the mode connectivity literature that when two neural networks are trained similarly on the same data, they are connected by a path through parameter space over which test set accuracy is maintained. Under some circumstances, including transfer learning from pretrained models, these paths are presumed to be linear. In contrast to existing results, we find that among text classifiers (trained on MNLI, QQP, and CoLA), some pairs of finetuned models have large barriers of increasing loss on the linear paths between them. On each task, we find distinct clusters of models which are linearly connected on the test loss surface, but are disconnected from models outside the cluster -- models that occupy separate basins on the surface. By measuring performance on specially-crafted diagnostic datasets, we find that these clusters correspond to different generalization strategies: one cluster behaves like a bag of words model under domain shift, while another cluster uses syntactic heuristics. Our work demonstrates how the geometry of the loss surface can guide models towards different heuristic functions."
                },
                {
                    "title": "Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time",
                    "abstract": "The conventional recipe for maximizing model accuracy is to (1) train multiple models with various hyperparameters and (2) pick the individual model which performs best on a held-out validation set, discarding the remainder. In this paper, we revisit the second step of this procedure in the context of fine-tuning large pre-trained models, where fine-tuned models often appear to lie in a single low error basin. We show that averaging the weights of multiple models fine-tuned with different hyperparameter configurations often improves accuracy and robustness. Unlike a conventional ensemble, we may average many models without incurring any additional inference or memory costs -- we call the results\"model soups.\"When fine-tuning large pre-trained models such as CLIP, ALIGN, and a ViT-G pre-trained on JFT, our soup recipe provides significant improvements over the best model in a hyperparameter sweep on ImageNet. The resulting ViT-G model, which attains 90.94% top-1 accuracy on ImageNet, achieved a new state of the art. Furthermore, we show that the model soup approach extends to multiple image classification and natural language processing tasks, improves out-of-distribution performance, and improves zero-shot performance on new downstream tasks. Finally, we analytically relate the performance similarity of weight-averaging and logit-ensembling to flatness of the loss and confidence of the predictions, and validate this relation empirically. Code is available at https://github.com/mlfoundations/model-soups."
                },
                {
                    "title": "Editing a classifier by rewriting its prediction rules",
                    "abstract": "We present a methodology for modifying the behavior of a classifier by directly rewriting its prediction rules. Our approach requires virtually no additional data collection and can be applied to a variety of settings, including adapting a model to new environments, and modifying it to ignore spurious features. Our code is available at https://github.com/MadryLab/EditingClassifiers ."
                },
                {
                    "title": "Merging Models with Fisher-Weighted Averaging",
                    "abstract": "Averaging the parameters of models that have the same architecture and initialization can provide a means of combining their respective capabilities. In this paper, we take the perspective that this\"merging\"operation can be seen as choosing parameters that approximately maximize the joint likelihood of the posteriors of the models' parameters. Computing a simple average of the models' parameters therefore corresponds to making an isotropic Gaussian approximation to their posteriors. We develop an alternative merging procedure based on the Laplace approximation where we approximate each model's posterior as a Gaussian distribution whose precision matrix corresponds to its Fisher information. We first show that our\"Fisher merging\"technique provides a performance boost in settings where simple parameter averaging is currently used -- specifically, robust fine-tuning and model ensembling. Then, we compare merging to standard gradient-based transfer learning and demonstrate that merging enables a fundamentally different method for transferring capabilities across models. Specifically, we show that Fisher merging is competitive with gradient-based transfer learning approaches (while being significantly cheaper) in intermediate-task training and domain-adaptive pre-training. We also show that our merging procedure makes it possible to combine models in previously unexplored ways. We release our code to facilitate future research into methods for merging models."
                },
                {
                    "title": "The Role of Permutation Invariance in Linear Mode Connectivity of Neural Networks",
                    "abstract": "In this paper, we conjecture that if the permutation invariance of neural networks is taken into account, SGD solutions will likely have no barrier in the linear interpolation between them. Although it is a bold conjecture, we show how extensive empirical attempts fall short of refuting it. We further provide a preliminary theoretical result to support our conjecture. Our conjecture has implications for lottery ticket hypothesis, distributed training, and ensemble methods."
                },
                {
                    "title": "What can linear interpolation of neural network loss landscapes tell us?",
                    "abstract": "Studying neural network loss landscapes provides insights into the nature of the underlying optimization problems. Unfortunately, loss landscapes are notoriously difficult to visualize in a human-comprehensible fashion. One common way to address this problem is to plot linear slices of the landscape, for example from the initial state of the network to the final state after optimization. On the basis of this analysis, prior work has drawn broader conclusions about the difficulty of the optimization problem. In this paper, we put inferences of this kind to the test, systematically evaluating how linear interpolation and final performance vary when altering the data, choice of initialization, and other optimizer and architecture design choices. Further, we use linear interpolation to study the role played by individual layers and substructures of the network. We find that certain layers are more sensitive to the choice of initialization, but that the shape of the linear path is not indicative of the changes in test accuracy of the model. Our results cast doubt on the broader intuition that the presence or absence of barriers when interpolating necessarily relates to the success of optimization."
                },
                {
                    "title": "Geometry of the Loss Landscape in Overparameterized Neural Networks: Symmetries and Invariances",
                    "abstract": "We study how permutation symmetries in overparameterized multi-layer neural networks generate `symmetry-induced' critical points. Assuming a network with $ L $ layers of minimal widths $ r_1^*, \\ldots, r_{L-1}^* $ reaches a zero-loss minimum at $ r_1^*! \\cdots r_{L-1}^*! $ isolated points that are permutations of one another, we show that adding one extra neuron to each layer is sufficient to connect all these previously discrete minima into a single manifold. For a two-layer overparameterized network of width $ r^*+ h =: m $ we explicitly describe the manifold of global minima: it consists of $ T(r^*, m) $ affine subspaces of dimension at least $ h $ that are connected to one another. For a network of width $m$, we identify the number $G(r,m)$ of affine subspaces containing only symmetry-induced critical points that are related to the critical points of a smaller network of width $r<r^*$. Via a combinatorial analysis, we derive closed-form formulas for $ T $ and $ G $ and show that the number of symmetry-induced critical subspaces dominates the number of affine subspaces forming the global minima manifold in the mildly overparameterized regime (small $ h $) and vice versa in the vastly overparameterized regime ($h \\gg r^*$). Our results provide new insights into the minimization of the non-convex loss function of overparameterized neural networks."
                },
                {
                    "title": "See through Gradients: Image Batch Recovery via GradInversion",
                    "abstract": "Training deep neural networks requires gradient estimation from data batches to update parameters. Gradients per parameter are averaged over a set of data and this has been presumed to be safe for privacy-preserving training in joint, collaborative, and federated learning applications. Prior work only showed the possibility of recovering input data given gradients under very restrictive conditions \u2013 a single input point, or a network with no non-linearities, or a small 32 \u00d7 32 px input batch. Therefore, averaging gradients over larger batches was thought to be safe. In this work, we introduce GradInversion, using which input images from a larger batch (8 \u2013 48 images) can also be recovered for large networks such as ResNets (50 layers), on complex datasets such as ImageNet (1000 classes, 224 \u00d7 224 px). We formulate an optimization task that converts random noise into natural images, matching gradients while regularizing image fidelity. We also propose an algorithm for target class label recovery given gradients. We further propose a group consistency regularization framework, where multiple agents starting from different random seeds work together to find an enhanced reconstruction of the original data batch. We show that gradients encode a surprisingly large amount of information, such that all the individual images can be recovered with high fidelity via GradInversion, even for complex datasets, deep networks, and large batch sizes."
                },
                {
                    "title": "Nondeterminism and Instability in Neural Network Optimization",
                    "abstract": "Nondeterminism in neural network optimization produces uncertainty in performance, making small improvements difficult to discern from run-to-run variability. While uncertainty can be reduced by training multiple model copies, doing so is time-consuming, costly, and harms reproducibility. In this work, we establish an experimental protocol for understanding the effect of optimization nondeterminism on model diversity, allowing us to isolate the effects of a variety of sources of nondeterminism. Surprisingly, we find that all sources of nondeterminism have similar effects on measures of model diversity. To explain this intriguing fact, we identify the instability of model training, taken as an end-to-end procedure, as the key determinant. We show that even one-bit changes in initial parameters result in models converging to vastly different values. Last, we propose two approaches for reducing the effects of instability on run-to-run variability."
                },
                {
                    "title": "Loss Surface Simplexes for Mode Connecting Volumes and Fast Ensembling",
                    "abstract": "With a better understanding of the loss surfaces for multilayer networks, we can build more robust and accurate training procedures. Recently it was discovered that independently trained SGD solutions can be connected along one-dimensional paths of near-constant training loss. In this paper, we show that there are mode-connecting simplicial complexes that form multi-dimensional manifolds of low loss, connecting many independently trained models. Inspired by this discovery, we show how to efficiently build simplicial complexes for fast ensembling, outperforming independently trained deep ensembles in accuracy, calibration, and robustness to dataset shift. Notably, our approach only requires a few training epochs to discover a low-loss simplex, starting from a pre-trained solution. Code is available at https://github.com/g-benton/loss-surface-simplexes."
                },
                {
                    "title": "Learning Neural Network Subspaces",
                    "abstract": "Recent observations have advanced our understanding of the neural network optimization landscape, revealing the existence of (1) paths of high accuracy containing diverse solutions and (2) wider minima offering improved performance. Previous methods observing diverse paths require multiple training runs. In contrast we aim to leverage both property (1) and (2) with a single method and in a single training run. With a similar computational cost as training one model, we learn lines, curves, and simplexes of high-accuracy neural networks. These neural network subspaces contain diverse solutions that can be ensembled, approaching the ensemble performance of independently trained networks without the training cost. Moreover, using the subspace midpoint boosts accuracy, calibration, and robustness to label noise, outperforming Stochastic Weight Averaging."
                },
                {
                    "title": "On the Reproducibility of Neural Network Predictions",
                    "abstract": "Standard training techniques for neural networks involve multiple sources of randomness, e.g., initialization, mini-batch ordering and in some cases data augmentation. Given that neural networks are heavily over-parameterized in practice, such randomness can cause {\\em churn} -- for the same input, disagreements between predictions of the two models independently trained by the same algorithm, contributing to the `reproducibility challenges' in modern machine learning. In this paper, we study this problem of churn, identify factors that cause it, and propose two simple means of mitigating it. We first demonstrate that churn is indeed an issue, even for standard image classification tasks (CIFAR and ImageNet), and study the role of the different sources of training randomness that cause churn. By analyzing the relationship between churn and prediction confidences, we pursue an approach with two components for churn reduction. First, we propose using \\emph{minimum entropy regularizers} to increase prediction confidences. Second, \\changes{we present a novel variant of co-distillation approach~\\citep{anil2018large} to increase model agreement and reduce churn}. We present empirical results showing the effectiveness of both techniques in reducing churn while improving the accuracy of the underlying model."
                },
                {
                    "title": "Revisiting \"Qualitatively Characterizing Neural Network Optimization Problems\"",
                    "abstract": "We revisit and extend the experiments of Goodfellow et al. (2014), who showed that - for then state-of-the-art networks - \"the objective function has a simple, approximately convex shape\" along the linear path between initialization and the trained weights. We do not find this to be the case for modern networks on CIFAR-10 and ImageNet. Instead, although loss is roughly monotonically non-increasing along this path, it remains high until close to the optimum. In addition, training quickly becomes linearly separated from the optimum by loss barriers. We conclude that, although Goodfellow et al.'s findings describe the \"relatively easy to optimize\" MNIST setting, behavior is qualitatively different in modern settings."
                },
                {
                    "title": "Linear Mode Connectivity in Multitask and Continual Learning",
                    "abstract": "Continual (sequential) training and multitask (simultaneous) training are often attempting to solve the same overall objective: to find a solution that performs well on all considered tasks. The main difference is in the training regimes, where continual learning can only have access to one task at a time, which for neural networks typically leads to catastrophic forgetting. That is, the solution found for a subsequent task does not perform well on the previous ones anymore. However, the relationship between the different minima that the two training regimes arrive at is not well understood. What sets them apart? Is there a local structure that could explain the difference in performance achieved by the two different schemes? Motivated by recent work showing that different minima of the same task are typically connected by very simple curves of low error, we investigate whether multitask and continual solutions are similarly connected. We empirically find that indeed such connectivity can be reliably achieved and, more interestingly, it can be done by a linear path, conditioned on having the same initialization for both. We thoroughly analyze this observation and discuss its significance for the continual learning process. Furthermore, we exploit this finding to propose an effective algorithm that constrains the sequentially learned minima to behave as the multitask solution. We show that our method outperforms several state of the art continual learning algorithms on various vision benchmarks."
                },
                {
                    "title": "What is being transferred in transfer learning?",
                    "abstract": "One desired capability for machines is the ability to transfer their knowledge of one domain to another where data is (usually) scarce. Despite ample adaptation of transfer learning in various deep learning applications, we yet do not understand what enables a successful transfer and which part of the network is responsible for that. In this paper, we provide new tools and analyses to address these fundamental questions. Through a series of analyses on transferring to block-shuffled images, we separate the effect of feature reuse from learning low-level statistics of data and show that some benefit of transfer learning comes from the latter. We present that when training from pre-trained weights, the model stays in the same basin in the loss landscape and different instances of such model are similar in feature space and close in parameter space."
                },
                {
                    "title": "Dataset Condensation with Gradient Matching",
                    "abstract": "Efficient training of deep neural networks is an increasingly important problem in the era of sophisticated architectures and large-scale datasets. This paper proposes a training set synthesis technique, called Dataset Condensation, that learns to produce a small set of informative samples for training deep neural networks from scratch in a small fraction of the required computational cost on the original data while achieving comparable results. We rigorously evaluate its performance in several computer vision benchmarks and show that it significantly outperforms the state-of-the-art methods. Finally we show promising applications of our method in continual learning and domain adaptation."
                },
                {
                    "title": "Editable Neural Networks",
                    "abstract": "These days deep neural networks are ubiquitously used in a wide range of tasks, from image classification and machine translation to face identification and self-driving cars. In many applications, a single model error can lead to devastating financial, reputational and even life-threatening consequences. Therefore, it is crucially important to correct model mistakes quickly as they appear. In this work, we investigate the problem of neural network editing - how one can efficiently patch a mistake of the model on a particular sample, without influencing the model behavior on other samples. Namely, we propose Editable Training, a model-agnostic training technique that encourages fast editing of the trained model. We empirically demonstrate the effectiveness of this method on large-scale image classification and machine translation tasks."
                },
                {
                    "title": "Linear Mode Connectivity and the Lottery Ticket Hypothesis",
                    "abstract": "We study whether a neural network optimizes to the same, linearly connected minimum under different samples of SGD noise (e.g., random data order and augmentation). We find that standard vision models become stable to SGD noise in this way early in training. From then on, the outcome of optimization is determined to a linearly connected region. We use this technique to study iterative magnitude pruning (IMP), the procedure used by work on the lottery ticket hypothesis to identify subnetworks that could have trained in isolation to full accuracy. We find that these subnetworks only reach full accuracy when they are stable to SGD noise, which either occurs at initialization for small-scale settings (MNIST) or early in training for large-scale settings (ResNet-50 and Inception-v3 on ImageNet)."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "On Warm-Starting Neural Network Training",
                    "abstract": "In many real-world deployments of machine learning systems, data arrive piecemeal. These learning scenarios may be passive, where data arrive incrementally due to structural properties of the problem (e.g., daily financial data) or active, where samples are selected according to a measure of their quality (e.g., experimental design). In both of these cases, we are building a sequence of models that incorporate an increasing amount of data. We would like each of these models in the sequence to be performant and take advantage of all the data that are available to that point. Conventional intuition suggests that when solving a sequence of related optimization problems of this form, it should be possible to initialize using the solution of the previous iterate---to \"warm start\" the optimization rather than initialize from scratch---and see reductions in wall-clock time. However, in practice this warm-starting seems to yield poorer generalization performance than models that have fresh random initializations, even though the final training losses are similar. While it appears that some hyperparameter settings allow a practitioner to close this generalization gap, they seem to only do so in regimes that damage the wall-clock gains of the warm start. Nevertheless, it is highly desirable to be able to warm-start neural network training, as it would dramatically reduce the resource usage associated with the construction of performant deep learning systems. In this work, we take a closer look at this empirical phenomenon and try to understand when and how it occurs. We also provide a surprisingly simple trick that overcomes this pathology in several important situations, and present experiments that elucidate some of its properties."
                },
                {
                    "title": "Model Fusion via Optimal Transport",
                    "abstract": "Combining different models is a widely used paradigm in machine learning applications. While the most common approach is to form an ensemble of models and average their individual predictions, this approach is often rendered infeasible by given resource constraints in terms of memory and computation, which grow linearly with the number of models. We present a layer-wise model fusion algorithm for neural networks that utilizes optimal transport to (soft-) align neurons across the models before averaging their associated parameters. We show that this can successfully yield \"one-shot\" knowledge transfer (i.e, without requiring any retraining) between neural networks trained on heterogeneous non-i.i.d. data. In both i.i.d. and non-i.i.d. settings, we illustrate that our approach significantly outperforms vanilla averaging, as well as how it can serve as an efficient replacement for the ensemble with moderate fine-tuning, for standard convolutional networks (like VGG11), residual networks (like ResNet18), and multi-layer perceptrons on CIFAR10 and MNIST. Finally, our approach also provides a principled way to combine the parameters of neural networks with different widths, and we explore its application for model compression. Code is available under the following link https://github.com/modelfusion."
                },
                {
                    "title": "Deep Ensembles: A Loss Landscape Perspective",
                    "abstract": "Deep ensembles have been empirically shown to be a promising approach for improving accuracy, uncertainty and out-of-distribution robustness of deep learning models. While deep ensembles were theoretically motivated by the bootstrap, non-bootstrap ensembles trained with just random initialization also perform well in practice, which suggests that there could be other explanations for why deep ensembles work well. Bayesian neural networks, which learn distributions over the parameters of the network, are theoretically well-motivated by Bayesian principles, but do not perform as well as deep ensembles in practice, particularly under dataset shift. One possible explanation for this gap between theory and practice is that popular scalable variational Bayesian methods tend to focus on a single mode, whereas deep ensembles tend to explore diverse modes in function space. We investigate this hypothesis by building on recent work on understanding the loss landscape of neural networks and adding our own exploration to measure the similarity of functions in the space of predictions. Our results show that random initializations explore entirely different modes, while functions along an optimization trajectory or sampled from the subspace thereof cluster within a single mode predictions-wise, while often deviating significantly in the weight space. Developing the concept of the diversity--accuracy plane, we show that the decorrelation power of random initializations is unmatched by popular subspace sampling methods. Finally, we evaluate the relative effects of ensembling, subspace based methods and ensembles of subspace based methods, and the experimental results validate our hypothesis."
                },
                {
                    "title": "Uniform convergence may be unable to explain generalization in deep learning",
                    "abstract": "Aimed at explaining the surprisingly good generalization behavior of overparameterized deep networks, recent works have developed a variety of generalization bounds for deep learning, all based on the fundamental learning-theoretic technique of uniform convergence. While it is well-known that many of these existing bounds are numerically large, through numerous experiments, we bring to light a more concerning aspect of these bounds: in practice, these bounds can {\\em increase} with the training dataset size. Guided by our observations, we then present examples of overparameterized linear classifiers and neural networks trained by gradient descent (GD) where uniform convergence provably cannot ``explain generalization'' -- even if we take into account the implicit bias of GD {\\em to the fullest extent possible}. More precisely, even if we consider only the set of classifiers output by GD, which have test errors less than some small $\\epsilon$ in our settings, we show that applying (two-sided) uniform convergence on this set of classifiers will yield only a vacuous generalization guarantee larger than $1-\\epsilon$. Through these findings, we cast doubt on the power of uniform convergence-based generalization bounds to provide a complete picture of why overparameterized deep networks generalize well."
                },
                {
                    "title": "Essentially No Barriers in Neural Network Energy Landscape",
                    "abstract": "Training neural networks involves finding minima of a high-dimensional non-convex loss function. Knowledge of the structure of this energy landscape is sparse. Relaxing from linear interpolations, we construct continuous paths between minima of recent neural network architectures on CIFAR10 and CIFAR100. Surprisingly, the paths are essentially flat in both the training and test landscapes. This implies that neural networks have enough capacity for structural changes, or that these changes are small between minima. Also, each minimum has at least one vanishing Hessian eigenvalue in addition to those resulting from trivial invariance."
                },
                {
                    "title": "Loss Surfaces, Mode Connectivity, and Fast Ensembling of DNNs",
                    "abstract": "The loss functions of deep neural networks are complex and their geometric properties are not well understood. We show that the optima of these complex loss functions are in fact connected by simple curves over which training and test accuracy are nearly constant. We introduce a training procedure to discover these high-accuracy pathways between modes. Inspired by this new geometric insight, we also propose a new ensembling method entitled Fast Geometric Ensembling (FGE). Using FGE we can train high-performing ensembles in the time required to train a single model. We achieve improved performance compared to the recent state-of-the-art Snapshot Ensembles, on CIFAR-10, CIFAR-100, and ImageNet."
                },
                {
                    "title": "Visualizing the Loss Landscape of Neural Nets",
                    "abstract": "Neural network training relies on our ability to find \"good\" minimizers of highly non-convex loss functions. It is well-known that certain network architecture designs (e.g., skip connections) produce loss functions that train easier, and well-chosen training parameters (batch size, learning rate, optimizer) produce minimizers that generalize better. However, the reasons for these differences, and their effects on the underlying loss landscape, are not well understood. In this paper, we explore the structure of neural loss functions, and the effect of loss landscapes on generalization, using a range of visualization methods. First, we introduce a simple \"filter normalization\" method that helps us visualize loss function curvature and make meaningful side-by-side comparisons between loss functions. Then, using a variety of visualizations, we explore how network architecture affects the loss landscape, and how training parameters affect the shape of minimizers."
                },
                {
                    "title": "Global optimality conditions for deep neural networks",
                    "abstract": "We study the error landscape of deep linear and nonlinear neural networks with the squared error loss. Minimizing the loss of a deep linear neural network is a nonconvex problem, and despite recent progress, our understanding of this loss surface is still incomplete. For deep linear networks, we present necessary and sufficient conditions for a critical point of the risk function to be a global minimum. Surprisingly, our conditions provide an efficiently checkable test for global optimality, while such tests are typically intractable in nonconvex optimization. We further extend these results to deep nonlinear neural networks and prove similar sufficient conditions for global optimality, albeit in a more limited function space setting."
                },
                {
                    "title": "Global Optimality in Neural Network Training",
                    "abstract": "The past few years have seen a dramatic increase in the performance of recognition systems thanks to the introduction of deep networks for representation learning. However, the mathematical reasons for this success remain elusive. A key issue is that the neural network training problem is nonconvex, hence optimization algorithms may not return a global minima. This paper provides sufficient conditions to guarantee that local minima are globally optimal and that a local descent strategy can reach a global minima from any initialization. Our conditions require both the network output and the regularization to be positively homogeneous functions of the network parameters, with the regularization being designed to control the network size. Our results apply to networks with one hidden layer, where size is measured by the number of neurons in the hidden layer, and multiple deep subnetworks connected in parallel, where size is measured by the number of subnetworks."
                },
                {
                    "title": "Convergence Analysis of Two-layer Neural Networks with ReLU Activation",
                    "abstract": "In recent years, stochastic gradient descent (SGD) based techniques has become the standard tools for training neural networks. However, formal theoretical understanding of why SGD can train neural networks in practice is largely missing. \nIn this paper, we make progress on understanding this mystery by providing a convergence analysis for SGD on a rich subset of two-layer feedforward networks with ReLU activations. This subset is characterized by a special structure called \"identity mapping\". We prove that, if input follows from Gaussian distribution, with standard $O(1/\\sqrt{d})$ initialization of the weights, SGD converges to the global minimum in polynomial number of steps. Unlike normal vanilla networks, the \"identity mapping\" makes our network asymmetric and thus the global minimum is unique. To complement our theory, we are also able to show experimentally that multi-layer networks with this mapping have better performance compared with normal vanilla networks. \nOur convergence theorem differs from traditional non-convex optimization techniques. We show that SGD converges to optimal in \"two phases\": In phase I, the gradient points to the wrong direction, however, a potential function $g$ gradually decreases. Then in phase II, SGD enters a nice one point convex region and converges. We also show that the identity mapping is necessary for convergence, as it moves the initial point to a better place for optimization. Experiment verifies our claims."
                },
                {
                    "title": "Snapshot Ensembles: Train 1, get M for free",
                    "abstract": "Ensembles of neural networks are known to be much more robust and accurate than individual networks. However, training multiple deep networks for model averaging is computationally expensive. In this paper, we propose a method to obtain the seemingly contradictory goal of ensembling multiple neural networks at no additional training cost. We achieve this goal by training a single neural network, converging to several local minima along its optimization path and saving the model parameters. To obtain repeated rapid convergence, we leverage recent work on cyclic learning rate schedules. The resulting technique, which we refer to as Snapshot Ensembling, is simple, yet surprisingly effective. We show in a series of experiments that our approach is compatible with diverse network architectures and learning tasks. It consistently yields lower error rates than state-of-the-art single models at no additional training cost, and compares favorably with traditional network ensembles. On CIFAR-10 and CIFAR-100 our DenseNet Snapshot Ensembles obtain error rates of 3.4% and 17.4% respectively."
                },
                {
                    "title": "Topology and Geometry of Half-Rectified Network Optimization",
                    "abstract": "The loss surface of deep neural networks has recently attracted interest in the optimization and machine learning communities as a prime example of high-dimensional non-convex problem. Some insights were recently gained using spin glass models and mean-field approximations, but at the expense of strongly simplifying the nonlinear nature of the model. \nIn this work, we do not make any such assumption and study conditions on the data distribution and model architecture that prevent the existence of bad local minima. Our theoretical work quantifies and formalizes two important \\emph{folklore} facts: (i) the landscape of deep linear networks has a radically different topology from that of deep half-rectified ones, and (ii) that the energy landscape in the non-linear case is fundamentally controlled by the interplay between the smoothness of the data distribution and model over-parametrization. Our main theoretical contribution is to prove that half-rectified single layer networks are asymptotically connected, and we provide explicit bounds that reveal the aforementioned interplay. \nThe conditioning of gradient descent is the next challenge we address. We study this question through the geometry of the level sets, and we introduce an algorithm to efficiently estimate the regularity of such sets on large-scale networks. Our empirical results show that these level sets remain connected throughout all the learning phase, suggesting a near convex behavior, but they become exponentially more curvy as the energy level decays, in accordance to what is observed in practice with very low curvature attractors."
                },
                {
                    "title": "On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima",
                    "abstract": "The stochastic gradient descent (SGD) method and its variants are algorithms of choice for many Deep Learning tasks. These methods operate in a small-batch regime wherein a fraction of the training data, say $32$-$512$ data points, is sampled to compute an approximation to the gradient. It has been observed in practice that when using a larger batch there is a degradation in the quality of the model, as measured by its ability to generalize. We investigate the cause for this generalization drop in the large-batch regime and present numerical evidence that supports the view that large-batch methods tend to converge to sharp minimizers of the training and testing functions - and as is well known, sharp minima lead to poorer generalization. In contrast, small-batch methods consistently converge to flat minimizers, and our experiments support a commonly held view that this is due to the inherent noise in the gradient estimation. We discuss several strategies to attempt to help large-batch methods eliminate this generalization gap."
                },
                {
                    "title": "SGDR: Stochastic Gradient Descent with Warm Restarts",
                    "abstract": "Restart techniques are common in gradient-free optimization to deal with multimodal functions. Partial warm restarts are also gaining popularity in gradient-based optimization to improve the rate of convergence in accelerated gradient schemes to deal with ill-conditioned functions. In this paper, we propose a simple warm restart technique for stochastic gradient descent to improve its anytime performance when training deep neural networks. We empirically study its performance on the CIFAR-10 and CIFAR-100 datasets, where we demonstrate new state-of-the-art results at 3.14% and 16.21%, respectively. We also demonstrate its advantages on a dataset of EEG recordings and on a downsampled version of the ImageNet dataset. Our source code is available at this https URL"
                },
                {
                    "title": "Gradient Descent Only Converges to Minimizers",
                    "abstract": "We show that gradient descent converges to a local minimizer, almost surely with random initialization. This is proved by applying the Stable Manifold Theorem from dynamical systems theory."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Distilling the Knowledge in a Neural Network",
                    "abstract": "A very simple way to improve the performance of almost any machine learning algorithm is to train many different models on the same data and then to average their predictions. Unfortunately, making predictions using a whole ensemble of models is cumbersome and may be too computationally expensive to allow deployment to a large number of users, especially if the individual models are large neural nets. Caruana and his collaborators have shown that it is possible to compress the knowledge in an ensemble into a single model which is much easier to deploy and we develop this approach further using a different compression technique. We achieve some surprising results on MNIST and we show that we can significantly improve the acoustic model of a heavily used commercial system by distilling the knowledge in an ensemble of models into a single model. We also introduce a new type of ensemble composed of one or more full models and many specialist models which learn to distinguish fine-grained classes that the full models confuse. Unlike a mixture of experts, these specialist models can be trained rapidly and in parallel."
                },
                {
                    "title": "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift",
                    "abstract": "Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.82% top-5 test error, exceeding the accuracy of human raters."
                },
                {
                    "title": "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification",
                    "abstract": "Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on the learnable activation and advanced initialization, we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66% [33]). To our knowledge, our result is the first to surpass the reported human-level performance (5.1%, [26]) on this dataset."
                },
                {
                    "title": "Adam: A Method for Stochastic Optimization",
                    "abstract": "We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
                },
                {
                    "title": "Qualitatively characterizing neural network optimization problems",
                    "abstract": "Training neural networks involves solving large-scale non-convex optimization problems. This task has long been believed to be extremely difficult, with fear of local minima and other obstacles motivating a variety of schemes to improve optimization, such as unsupervised pretraining. However, modern neural networks are able to achieve negligible training error on complex tasks, using only direct training with stochastic gradient descent. We introduce a simple analysis technique to look for evidence that such networks are overcoming local optima. We find that, in fact, on a straight path from initialization to solution, a variety of state of the art neural networks never encounter any significant obstacles."
                },
                {
                    "title": "The Loss Surfaces of Multilayer Networks",
                    "abstract": "We study the connection between the highly non-convex loss function of a simple model of the fully-connected feed-forward neural network and the Hamiltonian of the spherical spin-glass model under the assumptions of: i) variable independence, ii) redundancy in network parametrization, and iii) uniformity. These assumptions enable us to explain the complexity of the fully decoupled neural network through the prism of the results from random matrix theory. We show that for large-size decoupled networks the lowest critical values of the random loss function form a layered structure and they are located in a well-defined band lower-bounded by the global minimum. The number of local minima outside that band diminishes exponentially with the size of the network. We empirically verify that the mathematical model exhibits similar behavior as the computer simulations, despite the presence of high dependencies in real networks. We conjecture that both simulated annealing and SGD converge to the band of low critical points, and that all critical points found there are local minima of high quality measured by the test error. This emphasizes a major difference between largeand small-size networks where for the latter poor quality local minima have nonzero probability of being recovered. Finally, we prove that recovering the global minimum becomes harder as the network size increases and that it is in practice irrelevant as global minimum often leads to overfitting."
                },
                {
                    "title": "3D Object Representations for Fine-Grained Categorization",
                    "abstract": "While 3D object representations are being revived in the context of multi-view object class detection and scene understanding, they have not yet attained wide-spread use in fine-grained categorization. State-of-the-art approaches achieve remarkable performance when training data is plentiful, but they are typically tied to flat, 2D representations that model objects as a collection of unconnected views, limiting their ability to generalize across viewpoints. In this paper, we therefore lift two state-of-the-art 2D object representations to 3D, on the level of both local feature appearance and location. In extensive experiments on existing and newly proposed datasets, we show our 3D object representations outperform their state-of-the-art 2D counterparts for fine-grained categorization and demonstrate their efficacy for estimating 3D geometry from images via ultra-wide baseline matching and 3D reconstruction."
                },
                {
                    "title": "The Caltech-UCSD Birds-200-2011 Dataset",
                    "abstract": "CUB-200-2011 is an extended version of CUB-200 [7], a challenging dataset of 200 bird species. The extended version roughly doubles the number of images per category and adds new part localization annotations. All images are annotated with bounding boxes, part locations, and at- tribute labels. Images and annotations were filtered by mul- tiple users of Mechanical Turk. We introduce benchmarks and baseline experiments for multi-class categorization and part localization."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "On Monotonic Linear Interpolation of Neural Network Parameters",
                    "abstract": "Linearly interpolating between initial neural network parameters and converged parameters after training with SGD typically leads to a monotonic decrease in the training objective. This Monotonic Linear Interpolation (MLI) property, \ufb01rst observed by Goodfellow et al. [11], persists in spite of the non-convex objectives and highly non-linear training dynamics of neural networks. Extending on this work, we show that this property holds under varying network architectures, optimizers, and learning problems. We evaluate several possible hypotheses for this property that, to our knowledge, have not yet been explored. Additionally, we show that networks violating this property can be produced systematically, by forcing the weights to move far from initialization. The MLI property raises important questions about the loss landscape geometry of neural nets and highlights the need to further study its global properties."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Flat Minima",
                    "abstract": "We present a new algorithm for finding low-complexity neural networks with high generalization capability. The algorithm searches for a flat minimum of the error function. A flat minimum is a large connected region in weight space where the error remains approximately constant. An MDL-based, Bayesian argument suggests that flat minima correspond to simple networks and low expected overfitting. The argument is based on a Gibbs algorithm variant and a novel way of splitting generalization error into underfitting and overfitting error. Unlike many previous approaches, ours does not require gaussian assumptions and does not depend on a good weight prior. Instead we have a prior over input output functions, thus taking into account net architecture and training set. Although our algorithm requires the computation of second-order derivatives, it has backpropagation's order of complexity. Automatically, it effectively prunes units, weights, and input lines. Various experiments with feedforward and recurrent nets are described. In an application to stock market prediction, flat minimum search outperforms conventional backprop, weight decay, and optimal brain surgeon/optimal brain damage."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we efficiently synthesize a learning trajectory for a neural network initialized with different parameters by leveraging the similarities observed in independent training processes?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the learning transfer problem has significant implications for the research community as it addresses the computational inefficiencies associated with training large-scale neural networks. By enabling the transfer of learning trajectories between different initializations, this research could lead to more efficient training methods, reducing the time and resources required for model ensemble and knowledge distillation. This advancement could foster further exploration into optimization techniques and enhance the understanding of neural network dynamics, ultimately leading to practical applications in various domains such as computer vision, natural language processing, and beyond.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe learning transfer problem is challenging due to the highly non-convex nature of neural network loss landscapes and the complexity of finding appropriate permutation symmetries that connect different training trajectories. Naive approaches may fail because they do not account for the intricate relationships between different training runs, which can vary significantly based on initialization and data ordering. Additionally, the need to formulate the problem as a non-linear optimization task introduces technical obstacles, as it requires a deep understanding of both the geometry of the loss landscape and the dynamics of the training process.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on understanding loss landscapes and mode connectivity without directly addressing the transfer of learning trajectories between different initializations. Limitations in existing solutions include a lack of methods to effectively leverage permutation symmetries for trajectory synthesis and insufficient exploration of the connections between independent training runs. This research differs by explicitly formulating the learning transfer problem and proposing a novel approach to transform source trajectories using permutation symmetries, thereby filling a critical gap in the literature.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves formulating the learning transfer problem as a non-linear optimization task aimed at finding an appropriate permutation symmetry that connects the source learning trajectory to the target initialization. The approach will utilize datasets relevant to the specific neural network tasks and will evaluate the effectiveness of the synthesized trajectories using metrics such as training loss and convergence speed. The expected outcomes include a demonstration of reduced training costs and improved efficiency in achieving similar performance levels across different initializations, thereby validating the potential of learning transfer in neural network training."
            }
        },
        "author_data": {
            "6aa19061-6c3a-4b0e-9feb-afb42dfa2a43": {
                "pk": "6aa19061-6c3a-4b0e-9feb-afb42dfa2a43",
                "name": "Daiki Chijiwa",
                "collaborators": [
                    "Shin'ya Yamaguchi",
                    "Sekitoshi Kanai",
                    "Atsutoshi Kumagai",
                    "H. Kashima",
                    "Yasutoshi Ida",
                    "Kazuki Adachi",
                    "Hikari Otsuka",
                    "\u00c1ngel L\u00f3pez Garc\u00eda-Arias",
                    "Yasuyuki Okoshi",
                    "Kazushi Kawamura",
                    "Thiem Van Chu",
                    "Daichi Fujiki",
                    "Susumu Takeuchi",
                    "Masato Motomura",
                    "Masanori Yamada",
                    "Tomoya Yamashita",
                    "Kenji Umakoshi",
                    "T. Inoue"
                ],
                "domain": [
                    "Transfer Learning",
                    "Few-Shot Learning",
                    "Neural Network Pruning",
                    "Data Augmentation"
                ],
                "publications": [
                    {
                        "title": "Adaptive Random Feature Regularization on Fine-tuning Deep Neural Networks",
                        "abstract": "While fine-tuning is a de facto standard method for training deep neural networks, it still suffers from overfitting when using small target datasets. Previous methods improve fine-tuning performance by maintaining knowledge of the source datasets or introducing regularization terms such as contrastive loss. However, these methods require auxiliary source information (e.g., source labels or datasets) or heavy additional computations. In this paper, we propose a simple method called adaptive random feature regularization (AdaRand). AdaRand helps the feature extractors of training models to adaptively change the distribution of feature vectors for downstream classification tasks without auxiliary source information and with reasonable computation costs. To this end, AdaRand minimizes the gap between feature vectors and random reference vectors that are sampled from class conditional Gaussian distributions. Furthermore, AdaRand dynamically updates the conditional distribution to follow the currently updated feature extractors and balance the distance between classes in feature spaces. Our experiments show that AdaRand outperforms the other fine-tuning regularization requiring auxiliary source information and heavy computation costs."
                    },
                    {
                        "title": "Partial Search in a Frozen Network is Enough to Find a Strong Lottery Ticket",
                        "abstract": "Randomly initialized dense networks contain subnetworks that achieve high accuracy without weight learning -- strong lottery tickets (SLTs). Recently, Gadhikar et al. (2023) demonstrated that SLTs can also be found within a randomly pruned source network, thus reducing the SLT search space. However, this limits the search to SLTs that are even sparser than the source, leading to worse accuracy due to unintentionally high sparsity. This paper proposes a method that reduces the SLT search space by an arbitrary ratio independent of the desired SLT sparsity. A random subset of the initial weights is excluded from the search space by freezing it -- i.e., by either permanently pruning them or locking them as a fixed part of the SLT. In addition to reducing search space, the proposed random freezing can also provide the benefit of reducing the model size for inference. Furthermore, experimental results show that the proposed method finds SLTs with better accuracy-to-model size trade-off than the SLTs obtained from dense or randomly pruned source networks. In particular, the SLTs found in Frozen ResNets on image classification using ImageNet significantly improve the accuracy-to-search space and accuracy-to-model size trade-offs over SLTs within dense (non-freezing) or sparse (non-locking) random networks."
                    },
                    {
                        "title": "Regularizing Neural Networks with Meta-Learning Generative Models",
                        "abstract": "This paper investigates methods for improving generative data augmentation for deep learning. Generative data augmentation leverages the synthetic samples produced by generative models as an additional dataset for classification with small dataset settings. A key challenge of generative data augmentation is that the synthetic data contain uninformative samples that degrade accuracy. This is because the synthetic samples do not perfectly represent class categories in real data and uniform sampling does not necessarily provide useful samples for tasks. In this paper, we present a novel strategy for generative data augmentation called meta generative regularization (MGR). To avoid the degradation of generative data augmentation, MGR utilizes synthetic samples in the regularization term for feature extractors instead of in the loss function, e.g., cross-entropy. These synthetic samples are dynamically determined to minimize the validation losses through meta-learning. We observed that MGR can avoid the performance degradation of na\\\"ive generative data augmentation and boost the baselines. Experiments on six datasets showed that MGR is effective particularly when datasets are smaller and stably outperforms baselines."
                    },
                    {
                        "title": "Toward Data Efficient Model Merging between Different Datasets without Performance Degradation",
                        "abstract": "Model merging is attracting attention as a novel method for creating a new model by combining the weights of different trained models. While previous studies reported that model merging works well for models trained on a single dataset with different random seeds, model merging between different datasets remains unsolved. In this paper, we attempt to reveal the difficulty in merging such models trained on different datasets and alleviate it. Our empirical analyses show that, in contrast to the single-dataset scenarios, dataset information needs to be accessed to achieve high accuracy when merging models trained on different datasets. However, the requirement to use full datasets not only incurs significant computational costs but also becomes a major limitation when integrating models developed and shared by others. To address this, we demonstrate that dataset reduction techniques, such as coreset selection and dataset condensation, effectively reduce the data requirement for model merging. In our experiments with SPLIT-CIFAR10 model merging, the accuracy is significantly improved by $31%$ when using the full dataset and $24%$ when using the sampled subset compared with not using the dataset."
                    },
                    {
                        "title": "Transfer Learning with Pre-trained Conditional Generative Models",
                        "abstract": "Transfer learning is crucial in training deep neural networks on new target tasks. Current transfer learning methods always assume at least one of (i) source and target task label spaces overlap, (ii) source datasets are available, and (iii) target network architectures are consistent with source ones. However, holding these assumptions is difficult in practical settings because the target task rarely has the same labels as the source task, the source dataset access is restricted due to storage costs and privacy, and the target architecture is often specialized to each task. To transfer source knowledge without these assumptions, we propose a transfer learning method that uses deep generative models and is composed of the following two stages: pseudo pre-training (PP) and pseudo semi-supervised learning (P-SSL). PP trains a target architecture with an artificial dataset synthesized by using conditional source generative models. P-SSL applies SSL algorithms to labeled target data and unlabeled pseudo samples, which are generated by cascading the source classifier and generative models to condition them with target samples. Our experimental results indicate that our method can outperform the baselines of scratch training and knowledge distillation."
                    },
                    {
                        "title": "Meta-ticket: Finding optimal subnetworks for few-shot learning within randomly initialized neural networks",
                        "abstract": "Few-shot learning for neural networks (NNs) is an important problem that aims to train NNs with a few data. The main challenge is how to avoid overfitting since over-parameterized NNs can easily overfit to such small dataset. Previous work (e.g. MAML by Finn et al. 2017) tackles this challenge by meta-learning, which learns how to learn from a few data by using various tasks. On the other hand, one conventional approach to avoid overfitting is restricting hypothesis spaces by endowing sparse NN structures like convolution layers in computer vision. However, although such manually-designed sparse structures are sample-efficient for sufficiently large datasets, they are still insufficient for few-shot learning. Then the following questions naturally arise: (1) Can we find sparse structures effective for few-shot learning by meta-learning? (2) What benefits will it bring in terms of meta-generalization? In this work, we propose a novel meta-learning approach, called Meta-ticket, to find optimal sparse subnetworks for few-shot learning within randomly initialized NNs. We empirically validated that Meta-ticket successfully discover sparse subnetworks that can learn specialized features for each given task. Due to this task-wise adaptation ability, Meta-ticket achieves superior meta-generalization compared to MAML-based methods especially with large NNs. The code is available at: https://github.com/dchiji-ntt/meta-ticket"
                    },
                    {
                        "title": "Pruning Randomly Initialized Neural Networks with Iterative Randomization",
                        "abstract": "Pruning the weights of randomly initialized neural networks plays an important role in the context of lottery ticket hypothesis. Ramanujan et al. (2020) empirically showed that only pruning the weights can achieve remarkable performance instead of optimizing the weight values. However, to achieve the same level of performance as the weight optimization, the pruning approach requires more parameters in the networks before pruning and thus more memory space. To overcome this parameter inefficiency, we introduce a novel framework to prune randomly initialized neural networks with iteratively randomizing weight values (IteRand). Theoretically, we prove an approximation theorem in our framework, which indicates that the randomizing operations are provably effective to reduce the required number of the parameters. We also empirically demonstrate the parameter efficiency in multiple experiments on CIFAR-10 and ImageNet. The code is available at: https://github.com/dchiji-ntt/iterand"
                    }
                ]
            }
        }
    },
    "2310.16228": {
        "paper_data": {
            "title": "On the Foundations of Shortcut Learning",
            "url": "http://arxiv.org/abs/2310.16228v2",
            "arxiv_id": "2310.16228",
            "authors": [
                "Katherine L. Hermann",
                "Hossein Mobahi",
                "Thomas Fel",
                "Michael C. Mozer"
            ],
            "abstract": "Deep-learning models can extract a rich assortment of features from data. Which features a model uses depends not only on \\emph{predictivity} -- how reliably a feature indicates training-set labels -- but also on \\emph{availability} -- how easily the feature can be extracted from inputs. The literature on shortcut learning has noted examples in which models privilege one feature over another, for example texture over shape and image backgrounds over foreground objects. Here, we test hypotheses about which input properties are more available to a model, and systematically study how predictivity and availability interact to shape models' feature use. We construct a minimal, explicit generative framework for synthesizing classification datasets with two latent features that vary in predictivity and in factors we hypothesize to relate to availability, and we quantify a model's shortcut bias -- its over-reliance on the shortcut (more available, less predictive) feature at the expense of the core (less available, more predictive) feature. We find that linear models are relatively unbiased, but introducing a single hidden layer with ReLU or Tanh units yields a bias. Our empirical findings are consistent with a theoretical account based on Neural Tangent Kernels. Finally, we study how models used in practice trade off predictivity and availability in naturalistic datasets, discovering availability manipulations which increase models' degree of shortcut bias. Taken together, these findings suggest that the propensity to learn shortcut features is a fundamental characteristic of deep nonlinear architectures warranting systematic study given its role in shaping how models solve tasks.",
            "introduction": "   1 Introduction  Natural data domains provide a rich, high-dimensional input from which deep-learning models can extract a variety of candidate features. During training, models determine which features to rely on. Following training, the chosen features determine how models generalize. A challenge for machine learning arises when models come to rely on spurious or shortcut features instead of the core or defining features of a domain (Arjovsky et\u00a0al., 2019; McCoy et\u00a0al., 2019; Lapuschkin et\u00a0al., 2019; Geirhos et\u00a0al., 2020; Singla & Feizi, 2022). Shortcut \u201ccheat\u201d features, which are correlated with core \u201ctrue\u201d features in the training set, obtain good performance on the training set as well as on an iid test set, but poor generalization on out-of-distribution inputs. For instance, ImageNet-trained CNNs classify primarily according to an object\u2019s texture (Baker et\u00a0al., 2018; Geirhos et\u00a0al., 2018a; Hermann et\u00a0al., 2020), whereas people define and classify solid objects by shape (e.g., Landau et\u00a0al., 1988). Focusing on texture leads to reliable classification on many images but might result in misclassification of, say, a hairless cat, which has wrinkly skin more like that of an elephant. The terms \u201cspurious\u201d and \u201cshortcut\u201d are largely synonymous in the literature, although the former often refers to features that arise unintentionally in a poorly constructed dataset, and the latter to features easily latched onto by a model. In addition to a preference for texture over shape, other common shortcut features include a propensity to classify based on image backgrounds rather than foreground objects (Beery et\u00a0al., 2018; Sagawa et\u00a0al., 2020a; Xiao et\u00a0al., 2020; Moayeri et\u00a0al., 2022), or based on individual diagnostic pixels rather than higher-order image content (Malhotra et\u00a0al., 2020).   The literature examining feature use has often focused on predictivity\u2014how well a feature indicates the target output. Anomalies have been identified in which networks come to rely systematically on one feature over another when the features are equally predictive, or even when the preferred feature has lower predictivity than the non-preferred feature (Beery et\u00a0al., 2018; P\u00e9rez et\u00a0al., 2018; Tachet et\u00a0al., 2018; Arjovsky et\u00a0al., 2019; McCoy et\u00a0al., 2019; Hermann & Lampinen, 2020; Shah et\u00a0al., 2020; Nagarajan et\u00a0al., 2020; Pezeshki et\u00a0al., 2021; Fel et\u00a0al., 2023). Although we lack a general understanding of the cause of such preference anomalies, several specific cases have been identified. For example, features that are linearly related to classification labels are preferred by models over features that require nonlinear transforms (Hermann & Lampinen, 2020; Shah et\u00a0al., 2020). Another factor leading to anomalous feature preferences is the redundancy of representation, e.g., the size of the pixel footprint in an image (Sagawa et\u00a0al., 2020a; Wolff & Wolff, 2022; Tartaglini et\u00a0al., 2022).   Because predictivity alone is insufficient to explain feature reliance, here we explicitly introduce the notion of availability to refer to the factors that influence the likelihood that a model will use a feature more so than a purely statistical account would predict. A more-available feature is easier for the model to extract and leverage. Past research has systemtically manipulated predictivity; in the present work, we systematically manipulate both predictivity and availability to better understand their interaction and to characterize conditions giving rise to shortcut learning. Our contributions are:   \u2022  We define quantitative measures of predictivity and availability using a generative framework that allows",
            "references": [
                {
                    "title": "Spatial-frequency channels, shape bias, and adversarial robustness",
                    "abstract": "What spatial frequency information do humans and neural networks use to recognize objects? In neuroscience, critical band masking is an established tool that can reveal the frequency-selective filters used for object recognition. Critical band masking measures the sensitivity of recognition performance to noise added at each spatial frequency. Existing critical band masking studies show that humans recognize periodic patterns (gratings) and letters by means of a spatial-frequency filter (or\"channel'') that has a frequency bandwidth of one octave (doubling of frequency). Here, we introduce critical band masking as a task for network-human comparison and test 14 humans and 76 neural networks on 16-way ImageNet categorization in the presence of narrowband noise. We find that humans recognize objects in natural images using the same one-octave-wide channel that they use for letters and gratings, making it a canonical feature of human object recognition. On the other hand, the neural network channel, across various architectures and training strategies, is 2-4 times as wide as the human channel. In other words, networks are vulnerable to high and low frequency noise that does not affect human performance. Adversarial and augmented-image training are commonly used to increase network robustness and shape bias. Does this training align network and human object recognition channels? Three network channel properties (bandwidth, center frequency, peak noise sensitivity) correlate strongly with shape bias (53% variance explained) and with robustness of adversarially-trained networks (74% variance explained). Adversarial training increases robustness but expands the channel bandwidth even further away from the human bandwidth. Thus, critical band masking reveals that the network channel is more than twice as wide as the human channel, and that adversarial training only increases this difference."
                },
                {
                    "title": "Towards Last-layer Retraining for Group Robustness with Fewer Annotations",
                    "abstract": "Empirical risk minimization (ERM) of neural networks is prone to over-reliance on spurious correlations and poor generalization on minority groups. The recent deep feature reweighting (DFR) technique achieves state-of-the-art group robustness via simple last-layer retraining, but it requires held-out group and class annotations to construct a group-balanced reweighting dataset. In this work, we examine this impractical requirement and find that last-layer retraining can be surprisingly effective with no group annotations (other than for model selection) and only a handful of class annotations. We first show that last-layer retraining can greatly improve worst-group accuracy even when the reweighting dataset has only a small proportion of worst-group data. This implies a\"free lunch\"where holding out a subset of training data to retrain the last layer can substantially outperform ERM on the entire dataset with no additional data or annotations. To further improve group robustness, we introduce a lightweight method called selective last-layer finetuning (SELF), which constructs the reweighting dataset using misclassifications or disagreements. Our empirical and theoretical results present the first evidence that model disagreement upsamples worst-group data, enabling SELF to nearly match DFR on four well-established benchmarks across vision and language tasks with no group annotations and less than 3% of the held-out class annotations. Our code is available at https://github.com/tmlabonte/last-layer-retraining."
                },
                {
                    "title": "Don't blame Dataset Shift! Shortcut Learning due to Gradients and Cross Entropy",
                    "abstract": "Common explanations for shortcut learning assume that the shortcut improves prediction under the training distribution but not in the test distribution. Thus, models trained via the typical gradient-based optimization of cross-entropy, which we call default-ERM, utilize the shortcut. However, even when the stable feature determines the label in the training distribution and the shortcut does not provide any additional information, like in perception tasks, default-ERM still exhibits shortcut learning. Why are such solutions preferred when the loss for default-ERM can be driven to zero using the stable feature alone? By studying a linear perception task, we show that default-ERM's preference for maximizing the margin leads to models that depend more on the shortcut than the stable feature, even without overparameterization. This insight suggests that default-ERM's implicit inductive bias towards max-margin is unsuitable for perception tasks. Instead, we develop an inductive bias toward uniform margins and show that this bias guarantees dependence only on the perfect stable feature in the linear perception task. We develop loss functions that encourage uniform-margin solutions, called margin control (MARG-CTRL). MARG-CTRL mitigates shortcut learning on a variety of vision and language tasks, showing that better inductive biases can remove the need for expensive two-stage shortcut-mitigating methods in perception tasks."
                },
                {
                    "title": "A Holistic Approach to Unifying Automatic Concept Extraction and Concept Importance Estimation",
                    "abstract": "In recent years, concept-based approaches have emerged as some of the most promising explainability methods to help us interpret the decisions of Artificial Neural Networks (ANNs). These methods seek to discover intelligible visual 'concepts' buried within the complex patterns of ANN activations in two key steps: (1) concept extraction followed by (2) importance estimation. While these two steps are shared across methods, they all differ in their specific implementations. Here, we introduce a unifying theoretical framework that comprehensively defines and clarifies these two steps. This framework offers several advantages as it allows us: (i) to propose new evaluation metrics for comparing different concept extraction approaches; (ii) to leverage modern attribution methods and evaluation metrics to extend and systematically evaluate state-of-the-art concept-based approaches and importance estimation techniques; (iii) to derive theoretical guarantees regarding the optimality of such methods. We further leverage our framework to try to tackle a crucial question in explainability: how to efficiently identify clusters of data points that are classified based on a similar shared strategy. To illustrate these findings and to highlight the main strategies of a model, we introduce a visual representation called the strategic cluster graph. Finally, we present https://serre-lab.github.io/Lens, a dedicated website that offers a complete compilation of these visualizations for all classes of the ImageNet dataset."
                },
                {
                    "title": "Enhancing Representation Learning with Deep Classifiers in Presence of Shortcut",
                    "abstract": "A deep neural classifier trained on an upstream task can be leveraged to boost the performance of another classifier in a related downstream task through the representations learned in hidden layers. However, presence of shortcuts (easy-to-learn features) in the upstream task can considerably impair the versatility of intermediate representations and, in turn, the downstream performance. In this paper, we propose a method to improve the representations learned by deep neural image classifiers in spite of a shortcut in upstream data. In our method, the upstream classification objective is augmented with a type of adversarial training where an auxiliary network, so called lens, fools the classifier by exploiting the shortcut in reconstructing images. Empirical comparisons in self-supervised and transfer learning problems with three shortcut-biased datasets suggest the advantages of our method in terms of downstream performance and/or training time."
                },
                {
                    "title": "Sifer: Overcoming simplicity bias in deep networks using a feature sieve",
                    "abstract": "Simplicity bias is the concerning tendency of deep networks to over-depend on simple, weakly predictive features, to the exclusion of stronger, more complex features. This is exacerbated in real-world applications by limited training data and spurious feature-label correlations, leading to biased, incorrect predictions. We propose a direct, interventional method for addressing simplicity bias in DNNs, which we call the feature sieve. We aim to automatically identify and suppress easily-computable spurious features in lower layers of the network, thereby allowing the higher network levels to extract and utilize richer, more meaningful representations. We provide concrete evidence of this differential suppression&enhancement of relevant features on both controlled datasets and real-world images, and report substantial gains on many real-world debiasing benchmarks (11.4% relative gain on Imagenet-A; 3.2% on BAR, etc). Crucially, we do not depend on prior knowledge of spurious attributes or features, and in fact outperform many baselines that explicitly incorporate such information. We believe that our feature sieve work opens up exciting new research directions in automated adversarial feature extraction and representation learning for deep networks."
                },
                {
                    "title": "Last Layer Re-Training is Sufficient for Robustness to Spurious Correlations",
                    "abstract": "Neural network classifiers can largely rely on simple spurious features, such as backgrounds, to make predictions. However, even in these cases, we show that they still often learn core features associated with the desired attributes of the data, contrary to recent findings. Inspired by this insight, we demonstrate that simple last layer retraining can match or outperform state-of-the-art approaches on spurious correlation benchmarks, but with profoundly lower complexity and computational expenses. Moreover, we show that last layer retraining on large ImageNet-trained models can also significantly reduce reliance on background and texture information, improving robustness to covariate shift, after only minutes of training on a single GPU."
                },
                {
                    "title": "A Developmentally-Inspired Examination of Shape versus Texture Bias in Machines",
                    "abstract": "Early in development, children learn to extend novel category labels to objects with the same shape, a phenomenon known as the shape bias. Inspired by these findings, Geirhos et al. (2019) examined whether deep neural networks show a shape or texture bias by constructing images with conflicting shape and texture cues. They found that convolutional neural networks strongly preferred to classify familiar objects based on texture as opposed to shape, suggesting a texture bias. However, there are a number of differences between how the networks were tested in this study versus how children are typically tested. In this work, we re-examine the inductive biases of neural networks by adapting the stimuli and procedure from Geirhos et al. (2019) to more closely follow the developmental paradigm and test on a wide range of pre-trained neural networks. Across three experiments, we find that deep neural networks exhibit a preference for shape rather than texture when tested under conditions that more closely replicate the developmental procedure."
                },
                {
                    "title": "A Comprehensive Study of Image Classification Model Sensitivity to Foregrounds, Backgrounds, and Visual Attributes",
                    "abstract": "While datasets with single-label supervision have propelled rapid advances in image classification, additional annotations are necessary in order to quantitatively assess how models make predictions. To this end, for a subset of ImageNet samples, we collect segmentation masks for the entire object and 18 informative attributes. We call this dataset RIVAL10 (RIch Visual Attributes with Localization), consisting of roughly 26k instances over 10 classes. Using RIVAL10, we evaluate the sensitivity of a broad set of models to noise corruptions in foregrounds, backgrounds and attributes. In our analysis, we consider diverse state-of-the-art architectures (ResNets, Transformers) and training procedures (CLIP, SimCLR, DeiT, Adversarial Training). We find that, somewhat surprisingly, in ResNets, adversarial training makes models more sensitive to the background compared to foreground than standard training. Similarly, contrastively-trained models also have lower relative foreground sensitivity in both transformers and ResNets. Lastly, we observe intriguing adaptive abilities of transformers to increase relative foreground sensitivity as corruption level increases. Using saliency methods, we automatically discover spurious features that drive the background sensitivity of models and assess alignment of saliency maps with foregrounds. Finally, we quantitatively study the attribution problem for neural features by comparing feature saliency with ground-truth localization of semantic attributes."
                },
                {
                    "title": "Signal Strength and Noise Drive Feature Preference in CNN Image Classifiers",
                    "abstract": "Feature preference in Convolutional Neural Network (CNN) image classifiers is integral to their decision making process, and while the topic has been well studied, it is still not understood at a fundamental level. We test a range of task relevant feature attributes (including shape, texture, and color) with varying degrees of signal and noise in highly controlled CNN image classification experiments using synthetic datasets to determine feature preferences. We find that CNNs will prefer features with stronger signal strength and lower noise irrespective of whether the feature is texture, shape, or color. This provides guidance for a predictive model for task relevant feature preferences, demonstrates pathways for bias in machine models that can be avoided with careful controls on experimental setup, and suggests that comparisons between how humans and machines prefer task relevant features in vision classification tasks should be revisited. Code to reproduce experiments in this paper can be found at \\url{https://github.com/mwolff31/signal_preference}."
                },
                {
                    "title": "Look at the Variance! Efficient Black-box Explanations with Sobol-based Sensitivity Analysis",
                    "abstract": "We describe a novel attribution method which is grounded in Sensitivity Analysis and uses Sobol indices. Beyond modeling the individual contributions of image regions, Sobol indices provide an efficient way to capture higher-order interactions between image regions and their contributions to a neural network's prediction through the lens of variance. We describe an approach that makes the computation of these indices efficient for high-dimensional problems by using perturbation masks coupled with efficient estimators to handle the high dimensionality of images. Importantly, we show that the proposed method leads to favorable scores on standard benchmarks for vision (and language models) while drastically reducing the computing time compared to other black-box methods -- even surpassing the accuracy of state-of-the-art white-box methods which require access to internal representations. Our code is freely available: https://github.com/fel-thomas/Sobol-Attribution-Method"
                },
                {
                    "title": "Salient ImageNet: How to discover spurious features in Deep Learning?",
                    "abstract": "we identify spurious or core neural features (penultimate layer neurons of a robust model) via limited human supervision (e.g., using top 5 activating images per feature). We then show that these neural feature annotations generalize extremely well to many more images without any human supervision. We use the activation maps for these neural features as the soft masks to highlight spurious or core visual features. Using this methodology, we introduce the Salient Imagenet dataset containing core and spurious masks for a large set of samples from Imagenet. Using this dataset, we show that several popular Imagenet models rely heavily on various spurious features in their predictions, in-dicating the standard accuracy alone is not suf\ufb01cient to fully assess model perfor-1"
                },
                {
                    "title": "Can contrastive learning avoid shortcut solutions?",
                    "abstract": "The generalization of representations learned via contrastive learning depends crucially on what features of the data are extracted. However, we observe that the contrastive loss does not always sufficiently guide which features are extracted, a behavior that can negatively impact the performance on downstream tasks via \"shortcuts\", i.e., by inadvertently suppressing important predictive features. We find that feature extraction is influenced by the difficulty of the so-called instance discrimination task (i.e., the task of discriminating pairs of similar points from pairs of dissimilar ones). Although harder pairs improve the representation of some features, the improvement comes at the cost of suppressing previously well represented features. In response, we propose implicit feature modification (IFM), a method for altering positive and negative samples in order to guide contrastive models towards capturing a wider variety of predictive features. Empirically, we observe that IFM reduces feature suppression, and as a result improves performance on vision and medical imaging tasks. The code is available at: https://github.com/joshr17/IFM."
                },
                {
                    "title": "Evading the Simplicity Bias: Training a Diverse Set of Models Discovers Solutions with Superior OOD Generalization",
                    "abstract": "Neural networks trained with SGD were recently shown to rely preferentially on linearly-predictive features and can ignore complex, equally-predictive ones. This simplicity bias can explain their lack of robustness out of distribution (OOD). The more complex the task to learn, the more likely it is that statistical artifacts (i.e. selection biases, spurious correlations) are simpler than the mechanisms to learn. We demonstrate that the simplicity bias can be mitigated and OOD generalization improved. We train a set of similar models to fit the data in different ways using a penalty on the alignment of their input gradients. We show theoretically and empirically that this induces the learning of more com-plex predictive patterns. OOD generalization fundamentally requires information beyond i. i.d. examples, such as multiple training environ-ments, counterfactual examples, or other side information. Our approach shows that we can defer this requirement to an independent model selection stage. We obtain SOTA re-sults in visual recognition on biased data and generalization across visual domains. The method - the first to evade the simplicity bias - highlights the need for a better under-standing and control of inductive biases in deep learning."
                },
                {
                    "title": "Characterizing and Improving the Robustness of Self-Supervised Learning through Background Augmentations",
                    "abstract": "Recent progress in self-supervised learning has demonstrated promising results in multiple visual tasks. An important ingredient in high-performing self-supervised methods is the use of data augmentation by training models to place different augmented views of the same image nearby in embedding space. However, commonly used augmentation pipelines treat images holistically, ignoring the semantic relevance of parts of an image-e.g. a subject vs. a background-which can lead to the learning of spurious correlations. Our work addresses this problem by investigating a class of simple, yet highly effective\"background augmentations\", which encourage models to focus on semantically-relevant content by discouraging them from focusing on image backgrounds. Through a systematic investigation, we show that background augmentations lead to substantial improvements in performance across a spectrum of state-of-the-art self-supervised methods (MoCo-v2, BYOL, SwAV) on a variety of tasks, e.g. $\\sim$+1-2% gains on ImageNet, enabling performance on par with the supervised baseline. Further, we find the improvement in limited-labels settings is even larger (up to 4.2%). Background augmentations also improve robustness to a number of distribution shifts, including natural adversarial examples, ImageNet-9, adversarial attacks, ImageNet-Renditions. We also make progress in completely unsupervised saliency detection, in the process of generating saliency masks used for background augmentations."
                },
                {
                    "title": "Gradient Starvation: A Learning Proclivity in Neural Networks",
                    "abstract": "We identify and formalize a fundamental gradient descent phenomenon resulting in a learning proclivity in over-parameterized neural networks. Gradient Starvation arises when cross-entropy loss is minimized by capturing only a subset of features relevant for the task, despite the presence of other predictive features that fail to be discovered. This work provides a theoretical explanation for the emergence of such feature imbalance in neural networks. Using tools from Dynamical Systems theory, we identify simple properties of learning dynamics during gradient descent that lead to this imbalance, and prove that such a situation can be expected given certain statistical structure in training data. Based on our proposed formalism, we develop guarantees for a novel regularization method aimed at decoupling feature learning dynamics, improving accuracy and robustness in cases hindered by gradient starvation. We illustrate our findings with simple and real-world out-of-distribution (OOD) generalization experiments."
                },
                {
                    "title": "Understanding the Failure Modes of Out-of-Distribution Generalization",
                    "abstract": "Empirical studies suggest that machine learning models often rely on features, such as the background, that may be spuriously correlated with the label only during training time, resulting in poor accuracy during test-time. In this work, we identify the fundamental factors that give rise to this behavior, by explaining why models fail this way {\\em even} in easy-to-learn tasks where one would expect these models to succeed. In particular, through a theoretical study of gradient-descent-trained linear classifiers on some easy-to-learn tasks, we uncover two complementary failure modes. These modes arise from how spurious correlations induce two kinds of skews in the data: one geometric in nature, and another, statistical in nature. Finally, we construct natural modifications of image classification datasets to understand when these failure modes can arise in practice. We also design experiments to isolate the two failure modes when training modern neural networks on these datasets."
                },
                {
                    "title": "Deep Equals Shallow for ReLU Networks in Kernel Regimes",
                    "abstract": "Deep networks are often considered to be more expressive than shallow ones in terms of approximation. Indeed, certain functions can be approximated by deep networks provably more efficiently than by shallow ones, however, no tractable algorithms are known for learning such deep models. Separately, a recent line of work has shown that deep networks trained with gradient descent may behave like (tractable) kernel methods in a certain over-parameterized regime, where the kernel is determined by the architecture and initialization, and this paper focuses on approximation for such kernels. We show that for ReLU activations, the kernels derived from deep fully-connected networks have essentially the same approximation properties as their \"shallow\" two-layer counterpart, namely the same eigenvalue decay for the corresponding integral operator. This highlights the limitations of the kernel framework for understanding the benefits of such deep architectures. Our main theoretical result relies on characterizing such eigenvalue decays through differentiability properties of the kernel function, which also easily applies to the study of other kernels defined on the sphere."
                },
                {
                    "title": "What shapes feature representations? Exploring datasets, architectures, and training",
                    "abstract": "In naturalistic learning problems, a model's input contains a wide range of features, some useful for the task at hand, and others not. Of the useful features, which ones does the model use? Of the task-irrelevant features, which ones does the model represent? Answers to these questions are important for understanding the basis of models' decisions, as well as for building models that learn versatile, adaptable representations useful beyond the original training task. We study these questions using synthetic datasets in which the task-relevance of input features can be controlled directly. We find that when two features redundantly predict the labels, the model preferentially represents one, and its preference reflects what was most linearly decodable from the untrained model. Over training, task-relevant features are enhanced, and task-irrelevant features are partially suppressed. Interestingly, in some cases, an easier, weakly predictive feature can suppress a more strongly predictive, but more difficult one. Additionally, models trained to recognize both easy and hard features learn representations most similar to models that use only the easy feature. Further, easy features lead to more consistent representations across model runs than do hard features. Finally, models have greater representational similarity to an untrained model than to models trained on a different task. Our results highlight the complex processes that determine which features a model represents."
                },
                {
                    "title": "Noise or Signal: The Role of Image Backgrounds in Object Recognition",
                    "abstract": "We assess the tendency of state-of-the-art object recognition models to depend on signals from image backgrounds. We create a toolkit for disentangling foreground and background signal on ImageNet images, and find that (a) models can achieve non-trivial accuracy by relying on the background alone, (b) models often misclassify images even in the presence of correctly classified foregrounds--up to 87.5% of the time with adversarially chosen backgrounds, and (c) more accurate models tend to depend on backgrounds less. Our analysis of backgrounds brings us closer to understanding which correlations machine learning models use, and how they determine models' out of distribution performance."
                },
                {
                    "title": "The Pitfalls of Simplicity Bias in Neural Networks",
                    "abstract": "Several works have proposed Simplicity Bias (SB)---the tendency of standard training procedures such as Stochastic Gradient Descent (SGD) to find simple models---to justify why neural networks generalize well [Arpit et al. 2017, Nakkiran et al. 2019, Valle-Perez et al. 2019]. However, the precise notion of simplicity remains vague. Furthermore, previous settings that use SB to justify why neural networks generalize well do not simultaneously capture the brittleness of neural networks---a widely observed phenomenon in practice [Goodfellow et al. 2014, Jo and Bengio 2017]. To this end, we introduce a collection of piecewise-linear and image-based datasets that (a) naturally incorporate a precise notion of simplicity and (b) capture the subtleties of neural networks trained on real datasets. Through theory and experiments on these datasets, we show that SB of SGD and variants is extreme: neural networks rely exclusively on the simplest feature and remain invariant to all predictive complex features. Consequently, the extreme nature of SB explains why seemingly benign distribution shifts and small adversarial perturbations significantly degrade model performance. Moreover, contrary to conventional wisdom, SB can also hurt generalization on the same data distribution, as SB persists even when the simplest feature has less predictive power than the more complex features. We also demonstrate that common approaches for improving generalization and robustness---ensembles and adversarial training---do not mitigate SB and its shortcomings. Given the central role played by SB in generalization and robustness, we hope that the datasets and methods in this paper serve as an effective testbed to evaluate novel algorithmic approaches aimed at avoiding the pitfalls of extreme SB."
                },
                {
                    "title": "An Investigation of Why Overparameterization Exacerbates Spurious Correlations",
                    "abstract": "We study why overparameterization -- increasing model size well beyond the point of zero training error -- can hurt test error on minority groups despite improving average test error when there are spurious correlations in the data. Through simulations and experiments on two image datasets, we identify two key properties of the training data that drive this behavior: the proportions of majority versus minority groups, and the signal-to-noise ratio of the spurious correlations. We then analyze a linear setting and theoretically show how the inductive bias of models towards \"memorizing\" fewer examples can cause overparameterization to hurt. Our analysis leads to a counterintuitive approach of subsampling the majority group, which empirically achieves low minority error in the overparameterized regime, even though the standard approach of upweighting the minority fails. Overall, our results suggest a tension between using overparameterized models versus using all the training data for achieving low worst-group error."
                },
                {
                    "title": "Distributionally Robust Neural Networks",
                    "abstract": "Overparameterized neural networks trained to minimize average loss can be highly accurate on average on an i.i.d. test set, yet consistently fail on atypical groups of the data (e.g., by learning spurious correlations that do not hold at test time). Distributionally robust optimization (DRO) provides an approach for learning models that instead minimize worst-case training loss over a set of pre-defined groups. We find, however, that naively applying DRO to overparameterized neural networks fails: these models can perfectly fit the training data, and any model with vanishing average training loss will also already have vanishing worst-case training loss. Instead, the poor worst-case performance of these models arises from poor generalization on some groups. As a solution, we show that increased regularization---e.g., stronger-than-typical weight decay or early stopping---allows DRO models to achieve substantially higher worst-group accuracies, with 10% to 40% improvements over standard models on a natural language inference task and two image tasks, while maintaining high average accuracies. Our results suggest that regularization is critical for worst-group performance in the overparameterized regime, even if it is not needed for average performance. Finally, we introduce and provide convergence guarantees for a stochastic optimizer for this group DRO setting, underpinning the empirical study above."
                },
                {
                    "title": "Automatic Shortcut Removal for Self-Supervised Representation Learning",
                    "abstract": "In self-supervised visual representation learning, a feature extractor is trained on a \"pretext task\" for which labels can be generated cheaply, without human annotation. A central challenge in this approach is that the feature extractor quickly learns to exploit low-level visual features such as color aberrations or watermarks and then fails to learn useful semantic representations. Much work has gone into identifying such \"shortcut\" features and hand-designing schemes to reduce their effect. Here, we propose a general framework for mitigating the effect shortcut features. Our key assumption is that those features which are the first to be exploited for solving the pretext task may also be the most vulnerable to an adversary trained to make the task harder. We show that this assumption holds across common pretext tasks and datasets by training a \"lens\" network to make small image changes that maximally reduce performance in the pretext task. Representations learned with the modified images outperform those learned without in all tested cases. Additionally, the modifications made by the lens reveal how the choice of pretext task and dataset affects the features learned by self-supervision."
                },
                {
                    "title": "The Origins and Prevalence of Texture Bias in Convolutional Neural Networks",
                    "abstract": "Recent work has indicated that, unlike humans, ImageNet-trained CNNs tend to classify images by texture rather than by shape. How pervasive is this bias, and where does it come from? We find that, when trained on datasets of images with conflicting shape and texture, CNNs learn to classify by shape at least as easily as by texture. What factors, then, produce the texture bias in CNNs trained on ImageNet? Different unsupervised training objectives and different architectures have small but significant and largely independent effects on the level of texture bias. However, all objectives and architectures still lead to models that make texture-based classification decisions a majority of the time, even if shape information is decodable from their hidden representations. The effect of data augmentation is much larger. By taking less aggressive random crops at training time and applying simple, naturalistic augmentation (color distortion, noise, and blur), we train models that classify ambiguous images by shape a majority of the time, and outperform baselines on out-of-distribution test sets. Our results indicate that apparent differences in the way humans and ImageNet-trained CNNs process images may arise not from differences in their internal workings, but from differences in the data that they see."
                },
                {
                    "title": "Invariant Risk Minimization",
                    "abstract": "We introduce Invariant Risk Minimization (IRM), a learning paradigm to estimate invariant correlations across multiple training distributions. To achieve this goal, IRM learns a data representation such that the optimal classifier, on top of that data representation, matches for all training distributions. Through theory and experiments, we show how the invariances learned by IRM relate to the causal structures governing the data and enable out-of-distribution generalization."
                },
                {
                    "title": "A Fourier Perspective on Model Robustness in Computer Vision",
                    "abstract": "Achieving robustness to distributional shift is a longstanding and challenging goal of computer vision. Data augmentation is a commonly used approach for improving robustness, however robustness gains are typically not uniform across corruption types. Indeed increasing performance in the presence of random noise is often met with reduced performance on other corruptions such as contrast change. Understanding when and why these sorts of trade-offs occur is a crucial step towards mitigating them. Towards this end, we investigate recently observed trade-offs caused by Gaussian data augmentation and adversarial training. We find that both methods improve robustness to corruptions that are concentrated in the high frequency domain while reducing robustness to corruptions that are concentrated in the low frequency domain. This suggests that one way to mitigate these trade-offs via data augmentation is to use a more diverse set of augmentations. Towards this end we observe that AutoAugment, a recently proposed data augmentation policy optimized for clean accuracy, achieves state-of-the-art robustness on the CIFAR-10-C benchmark."
                },
                {
                    "title": "Implicit Regularization in Deep Matrix Factorization",
                    "abstract": "Efforts to understand the generalization mystery in deep learning have led to the belief that gradient-based optimization induces a form of implicit regularization, a bias towards models of low \"complexity.\" We study the implicit regularization of gradient descent over deep linear neural networks for matrix completion and sensing, a model referred to as deep matrix factorization. Our first finding, supported by theory and experiments, is that adding depth to a matrix factorization enhances an implicit tendency towards low-rank solutions, oftentimes leading to more accurate recovery. Secondly, we present theoretical and empirical arguments questioning a nascent view by which implicit regularization in matrix factorization can be captured using simple mathematical norms. Our results point to the possibility that the language of standard regularizers may not be rich enough to fully encompass the implicit regularization brought forth by gradient-based optimization."
                },
                {
                    "title": "Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference",
                    "abstract": "A machine learning system can score well on a given test set by relying on heuristics that are effective for frequent example types but break down in more challenging cases. We study this issue within natural language inference (NLI), the task of determining whether one sentence entails another. We hypothesize that statistical NLI models may adopt three fallible syntactic heuristics: the lexical overlap heuristic, the subsequence heuristic, and the constituent heuristic. To determine whether models have adopted these heuristics, we introduce a controlled evaluation set called HANS (Heuristic Analysis for NLI Systems), which contains many examples where the heuristics fail. We find that models trained on MNLI, including BERT, a state-of-the-art model, perform very poorly on HANS, suggesting that they have indeed adopted these heuristics. We conclude that there is substantial room for improvement in NLI systems, and that the HANS dataset can motivate and measure progress in this area."
                },
                {
                    "title": "Deep convolutional networks do not classify based on global object shape",
                    "abstract": "Deep convolutional networks (DCNNs) are achieving previously unseen performance in object classification, raising questions about whether DCNNs operate similarly to human vision. In biological vision, shape is arguably the most important cue for recognition. We tested the role of shape information in DCNNs trained to recognize objects. In Experiment 1, we presented a trained DCNN with object silhouettes that preserved overall shape but were filled with surface texture taken from other objects. Shape cues appeared to play some role in the classification of artifacts, but little or none for animals. In Experiments 2\u20134, DCNNs showed no ability to classify glass figurines or outlines but correctly classified some silhouettes. Aspects of these results led us to hypothesize that DCNNs do not distinguish object\u2019s bounding contours from other edges, and that DCNNs access some local shape features, but not global shape. In Experiment 5, we tested this hypothesis with displays that preserved local features but disrupted global shape, and vice versa. With disrupted global shape, which reduced human accuracy to 28%, DCNNs gave the same classification labels as with ordinary shapes. Conversely, local contour changes eliminated accurate DCNN classification but caused no difficulty for human observers. These results provide evidence that DCNNs have access to some local shape information in the form of local edge relations, but they have no access to global object shapes."
                },
                {
                    "title": "ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness",
                    "abstract": "Convolutional Neural Networks (CNNs) are commonly thought to recognise objects by learning increasingly complex representations of object shapes. Some recent studies suggest a more important role of image textures. We here put these conflicting hypotheses to a quantitative test by evaluating CNNs and human observers on images with a texture-shape cue conflict. We show that ImageNet-trained CNNs are strongly biased towards recognising textures rather than shapes, which is in stark contrast to human behavioural evidence and reveals fundamentally different classification strategies. We then demonstrate that the same standard architecture (ResNet-50) that learns a texture-based representation on ImageNet is able to learn a shape-based representation instead when trained on \"Stylized-ImageNet\", a stylized version of ImageNet. This provides a much better fit for human behavioural performance in our well-controlled psychophysical lab setting (nine experiments totalling 48,560 psychophysical trials across 97 observers) and comes with a number of unexpected emergent benefits such as improved object detection performance and previously unseen robustness towards a wide range of image distortions, highlighting advantages of a shape-based representation."
                },
                {
                    "title": "On the Learning Dynamics of Deep Neural Networks",
                    "abstract": "While a lot of progress has been made in recent years, the dynamics of learning in deep nonlinear neural networks remain to this day largely misunderstood. In this work, we study the case of binary classification and prove various properties of learning in such networks under strong assumptions such as linear separability of the data. Extending existing results from the linear case, we confirm empirical observations by proving that the classification error also follows a sigmoidal shape in nonlinear architectures. We show that given proper initialization, learning expounds parallel independent modes and that certain regions of parameter space might lead to failed training. We also demonstrate that input norm and features' frequency in the dataset lead to distinct convergence speeds which might shed some light on the generalization capabilities of deep neural networks. We provide a comparison between the dynamics of learning with cross-entropy and hinge losses, which could prove useful to understand recent progress in the training of generative adversarial networks. Finally, we identify a phenomenon that we baptize \\textit{gradient starvation} where the most frequent features in a dataset prevent the learning of other less frequent but equally informative features."
                },
                {
                    "title": "On the Structural Sensitivity of Deep Convolutional Networks to the Directions of Fourier Basis Functions",
                    "abstract": "Data-agnostic quasi-imperceptible perturbations on inputs are known to degrade recognition accuracy of deep convolutional networks severely. This phenomenon is considered to be a potential security issue. Moreover, some results on statistical generalization guarantees indicate that the phenomena can be a key to improve the networks' generalization. However, the characteristics of the shared directions of such harmful perturbations remain unknown. Our primal finding is that convolutional networks are sensitive to the directions of Fourier basis functions. We derived the property by specializing a hypothesis of the cause of the sensitivity, known as the linearity of neural networks, to convolutional networks and empirically validated it. As a byproduct of the analysis, we propose an algorithm to create shift-invariant universal adversarial perturbations available in black-box settings."
                },
                {
                    "title": "Generalisation in humans and deep neural networks",
                    "abstract": "We compare the robustness of humans and current convolutional deep neural networks (DNNs) on object recognition under twelve different types of image degradations. First, using three well known DNNs (ResNet-152, VGG-19, GoogLeNet) we find the human visual system to be more robust to nearly all of the tested image manipulations, and we observe progressively diverging classification error-patterns between humans and DNNs when the signal gets weaker. Secondly, we show that DNNs trained directly on distorted images consistently surpass human performance on the exact distortion types they were trained on, yet they display extremely poor generalisation abilities when tested on other distortion types. For example, training on salt-and-pepper noise does not imply robustness on uniform white noise and vice versa. Thus, changes in the noise distribution between training and testing constitutes a crucial challenge to deep learning vision systems that can be systematically addressed in a lifelong machine learning approach. Our new dataset consisting of 83K carefully measured human psychophysical trials provide a useful reference for lifelong robustness against image degradations set by the human visual system."
                },
                {
                    "title": "On the Spectral Bias of Neural Networks",
                    "abstract": "Neural networks are known to be a class of highly expressive functions able to fit even random input-output mappings with $100\\%$ accuracy. In this work, we present properties of neural networks that complement this aspect of expressivity. By using tools from Fourier analysis, we show that deep ReLU networks are biased towards low frequency functions, meaning that they cannot have local fluctuations without affecting their global behavior. Intuitively, this property is in line with the observation that over-parameterized networks find simple patterns that generalize across data samples. We also investigate how the shape of the data manifold affects expressivity by showing evidence that learning high frequencies gets \\emph{easier} with increasing manifold complexity, and present a theoretical understanding of this behavior. Finally, we study the robustness of the frequency components with respect to parameter perturbation, to develop the intuition that the parameters must be finely tuned to express high frequency functions."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "RISE: Randomized Input Sampling for Explanation of Black-box Models",
                    "abstract": "Deep neural networks are being used increasingly to automate data analysis and decision making, yet their decision-making process is largely unclear and is difficult to explain to the end users. In this paper, we address the problem of Explainable AI for deep neural networks that take images as input and output a class probability. We propose an approach called RISE that generates an importance map indicating how salient each pixel is for the model's prediction. In contrast to white-box approaches that estimate pixel importance using gradients or other internal network state, RISE works on black-box models. It estimates importance empirically by probing the model with randomly masked versions of the input image and obtaining the corresponding outputs. We compare our approach to state-of-the-art importance extraction methods using both an automatic deletion/insertion metric and a pointing metric based on human-annotated object segments. Extensive experiments on several benchmark datasets show that our approach matches or exceeds the performance of other methods, including white-box approaches. \nProject page: this http URL"
                },
                {
                    "title": "Places: A 10 Million Image Database for Scene Recognition",
                    "abstract": "The rise of multi-million-item dataset initiatives has enabled data-hungry machine learning algorithms to reach near-human semantic classification performance at tasks such as visual object and scene recognition. Here we describe the Places Database, a repository of 10 million scene photographs, labeled with scene semantic categories, comprising a large and diverse list of the types of environments encountered in the world. Using the state-of-the-art Convolutional Neural Networks (CNNs), we provide scene classification CNNs (Places-CNNs) as baselines, that significantly outperform the previous approaches. Visualization of the CNNs trained on Places shows that object detectors emerge as an intermediate representation of scene classification. With its high-coverage and high-diversity of exemplars, the Places Database along with the Places-CNNs offer a novel resource to guide future progress on scene recognition problems."
                },
                {
                    "title": "Deep learning generalizes because the parameter-function map is biased towards simple functions",
                    "abstract": "Deep neural networks (DNNs) generalize remarkably well without explicit regularization even in the strongly over-parametrized regime where classical learning theory would instead predict that they would severely overfit. While many proposals for some kind of implicit regularization have been made to rationalise this success, there is no consensus for the fundamental reason why DNNs do not strongly overfit. In this paper, we provide a new explanation. By applying a very general probability-complexity bound recently derived from algorithmic information theory (AIT), we argue that the parameter-function map of many DNNs should be exponentially biased towards simple functions. We then provide clear evidence for this strong simplicity bias in a model DNN for Boolean functions, as well as in much larger fully connected and convolutional networks applied to CIFAR10 and MNIST. As the target functions in many real problems are expected to be highly structured, this intrinsic simplicity bias helps explain why deep networks generalize well on real world problems. This picture also facilitates a novel PAC-Bayes approach where the prior is taken over the DNN input-output function space, rather than the more conventional prior over parameter space. If we assume that the training algorithm samples parameters close to uniformly within the zero-error region then the PAC-Bayes theorem can be used to guarantee good expected generalization for target functions producing high-likelihood training sets. By exploiting recently discovered connections between DNNs and Gaussian processes to estimate the marginal likelihood, we produce relatively tight generalization PAC-Bayes error bounds which correlate well with the true error on realistic datasets such as MNIST and CIFAR10 and for architectures including convolutional and fully connected networks."
                },
                {
                    "title": "Annotation Artifacts in Natural Language Inference Data",
                    "abstract": "Large-scale datasets for natural language inference are created by presenting crowd workers with a sentence (premise), and asking them to generate three new sentences (hypotheses) that it entails, contradicts, or is logically neutral with respect to. We show that, in a significant portion of such data, this protocol leaves clues that make it possible to identify the label by looking only at the hypothesis, without observing the premise. Specifically, we show that a simple text categorization model can correctly classify the hypothesis alone in about 67% of SNLI (Bowman et. al, 2015) and 53% of MultiNLI (Williams et. al, 2017). Our analysis reveals that specific linguistic phenomena such as negation and vagueness are highly correlated with certain inference classes. Our findings suggest that the success of natural language inference models to date has been overestimated, and that the task remains a hard open problem."
                },
                {
                    "title": "Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification",
                    "abstract": "Recent studies demonstrate that machine learning algorithms can discriminate based on classes like race and gender. In this work, we present an approach to evaluate bias present in automated facial analysis algorithms and datasets with respect to phenotypic subgroups. Using the dermatologist approved Fitzpatrick Skin Type classi\ufb01cation system, we characterize the gender and skin type distribution of two facial analysis benchmarks, IJB-A and Adience. We \ufb01nd that these datasets are overwhelm-ingly composed of lighter-skinned subjects (79 . 6% for IJB-A and 86 . 2% for Adience) and introduce a new facial analysis dataset which is balanced by gender and skin type. We evaluate 3 commercial gender classi\ufb01cation systems using our dataset and show that darker-skinned females are the most misclassi\ufb01ed group (with error rates of up to 34 . 7%). The maximum error rate for lighter-skinned males is 0 . 8%. The substantial disparities in the accuracy of classifying darker females, lighter females, darker males, and lighter males in gender classi\ufb01cation systems require urgent attention if commercial companies are to build genuinely fair, transparent and accountable facial analysis algorithms."
                },
                {
                    "title": "Measuring the tendency of CNNs to Learn Surface Statistical Regularities",
                    "abstract": "Deep CNNs are known to exhibit the following peculiarity: on the one hand they generalize extremely well to a test set, while on the other hand they are extremely sensitive to so-called adversarial perturbations. The extreme sensitivity of high performance CNNs to adversarial examples casts serious doubt that these networks are learning high level abstractions in the dataset. We are concerned with the following question: How can a deep CNN that does not learn any high level semantics of the dataset manage to generalize so well? The goal of this article is to measure the tendency of CNNs to learn surface statistical regularities of the dataset. To this end, we use Fourier filtering to construct datasets which share the exact same high level abstractions but exhibit qualitatively different surface statistical regularities. For the SVHN and CIFAR-10 datasets, we present two Fourier filtered variants: a low frequency variant and a randomly filtered variant. Each of the Fourier filtering schemes is tuned to preserve the recognizability of the objects. Our main finding is that CNNs exhibit a tendency to latch onto the Fourier image statistics of the training dataset, sometimes exhibiting up to a 28% generalization gap across the various test sets. Moreover, we observe that significantly increasing the depth of a network has a very marginal impact on closing the aforementioned generalization gap. Thus we provide quantitative evidence supporting the hypothesis that deep CNNs tend to learn surface statistical regularities in the dataset rather than higher-level abstract concepts."
                },
                {
                    "title": "Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraints",
                    "abstract": "Language is increasingly being used to de-fine rich visual recognition problems with supporting image collections sourced from the web. Structured prediction models are used in these tasks to take advantage of correlations between co-occurring labels and visual input but risk inadvertently encoding social biases found in web corpora. In this work, we study data and models associated with multilabel object classification and visual semantic role labeling. We find that (a) datasets for these tasks contain significant gender bias and (b) models trained on these datasets further amplify existing bias. For example, the activity cooking is over 33% more likely to involve females than males in a training set, and a trained model further amplifies the disparity to 68% at test time. We propose to inject corpus-level constraints for calibrating existing structured prediction models and design an algorithm based on Lagrangian relaxation for collective inference. Our method results in almost no performance loss for the underlying recognition task but decreases the magnitude of bias amplification by 47.5% and 40.5% for multilabel classification and visual semantic role labeling, respectively\u3002"
                },
                {
                    "title": "SmoothGrad: removing noise by adding noise",
                    "abstract": "Explaining the output of a deep network remains a challenge. In the case of an image classifier, one type of explanation is to identify pixels that strongly influence the final decision. A starting point for this strategy is the gradient of the class score function with respect to the input image. This gradient can be interpreted as a sensitivity map, and there are several techniques that elaborate on this basic idea. This paper makes two contributions: it introduces SmoothGrad, a simple method that can help visually sharpen gradient-based sensitivity maps, and it discusses lessons in the visualization of these maps. We publish the code for our experiments and a website with our results."
                },
                {
                    "title": "Generating captions without looking beyond objects",
                    "abstract": "This paper explores new evaluation perspectives for image captioning and introduces a noun translation task that achieves comparative image caption generation performance by translating from a set of nouns to captions. This implies that in image captioning, all word categories other than nouns can be evoked by a powerful language model without sacrificing performance on n-gram precision. The paper also investigates lower and upper bounds of how much individual word categories in the captions contribute to the final BLEU score. A large possible improvement exists for nouns, verbs, and prepositions."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "In Search of the Real Inductive Bias: On the Role of Implicit Regularization in Deep Learning",
                    "abstract": "We present experiments demonstrating that some other form of capacity control, different from network size, plays a central role in learning multilayer feed-forward networks. We argue, partially through analogy to matrix factorization, that this is an inductive bias that can help shed light on deep learning."
                },
                {
                    "title": "Deep Learning Face Attributes in the Wild",
                    "abstract": "Predicting face attributes in the wild is challenging due to complex face variations. We propose a novel deep learning framework for attribute prediction in the wild. It cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently. LNet is pre-trained by massive general object categories for face localization, while ANet is pre-trained by massive face identities for attribute prediction. This framework not only outperforms the state-of-the-art with a large margin, but also reveals valuable facts on learning face representation. (1) It shows how the performances of face localization (LNet) and attribute prediction (ANet) can be improved by different pre-training strategies. (2) It reveals that although the filters of LNet are fine-tuned only with image-level attribute tags, their response maps over entire images have strong indication of face locations. This fact enables training LNet for face localization with only image-level annotations, but without face bounding boxes or landmarks, which are required by all attribute recognition works. (3) It also demonstrates that the high-level hidden neurons of ANet automatically discover semantic concepts after pre-training with massive face identities, and such concepts are significantly enriched after fine-tuning with attribute tags. Each attribute can be well explained with a sparse linear combination of these concepts."
                },
                {
                    "title": "The Caltech-UCSD Birds-200-2011 Dataset",
                    "abstract": "CUB-200-2011 is an extended version of CUB-200 [7], a challenging dataset of 200 bird species. The extended version roughly doubles the number of images per category and adds new part localization annotations. All images are annotated with bounding boxes, part locations, and at- tribute labels. Images and annotations were filtered by mul- tiple users of Mechanical Turk. We introduce benchmarks and baseline experiments for multi-class categorization and part localization."
                },
                {
                    "title": "Kernel Methods for Deep Learning",
                    "abstract": "We introduce a new family of positive-definite kernel functions that mimic the computation in large, multilayer neural nets. These kernel functions can be used in shallow architectures, such as support vector machines (SVMs), or in deep kernel-based architectures that we call multilayer kernel machines (MKMs). We evaluate SVMs and MKMs with these kernel functions on problems designed to illustrate the advantages of deep architectures. On several problems, we obtain better results than previous, leading benchmarks from both SVMs with Gaussian kernels as well as deep belief nets."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow do the concepts of predictivity and availability interact to influence shortcut learning in machine learning models, and what conditions lead to the reliance on spurious features over core features?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the fundamental issue of model generalization in machine learning. Understanding the interplay between predictivity and availability can lead to the development of more robust models that rely on core features, thereby improving their performance on out-of-distribution inputs. This research could advance knowledge in feature selection and representation learning, ultimately leading to practical applications in various domains, such as computer vision and natural language processing, where reliable model predictions are essential.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of feature interactions and the difficulty in quantifying both predictivity and availability. Naive approaches may fail because they often consider only the statistical correlation of features without accounting for their availability, which can lead to misleading conclusions about feature importance. Additionally, the redundancy of representation and the nonlinear relationships between features and labels introduce further complications that need to be addressed to fully understand shortcut learning.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on predictivity without adequately considering the role of availability, leading to a gap in understanding how these factors interact. Barriers include a lack of comprehensive frameworks to quantify both aspects and insufficient experimental manipulation of these variables. Our approach differs by systematically manipulating both predictivity and availability, allowing for a more nuanced exploration of their effects on feature reliance and shortcut learning.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves defining quantitative measures of predictivity and availability within a generative framework. We will conduct experiments using diverse datasets to manipulate these measures and observe their effects on model behavior. The expected outcomes include a clearer understanding of the conditions that lead to shortcut learning, as well as insights into how to design models that prioritize core features over spurious ones, ultimately enhancing generalization capabilities."
            }
        },
        "author_data": {
            "2830176e-d8a1-453d-b241-81020ed4ef57": {
                "pk": "2830176e-d8a1-453d-b241-81020ed4ef57",
                "name": "Katherine L. Hermann",
                "collaborators": [
                    "Andrew Kyle Lampinen",
                    "Thomas Fel",
                    "Louis B\u00e9thune",
                    "Thomas Serre",
                    "Stephanie C. Y. Chan"
                ],
                "domain": [
                    "Representation Learning",
                    "Deep Learning",
                    "Interpretability",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Understanding Visual Feature Reliance through the Lens of Complexity",
                        "abstract": "Recent studies suggest that deep learning models inductive bias towards favoring simpler features may be one of the sources of shortcut learning. Yet, there has been limited focus on understanding the complexity of the myriad features that models learn. In this work, we introduce a new metric for quantifying feature complexity, based on $\\mathscr{V}$-information and capturing whether a feature requires complex computational transformations to be extracted. Using this $\\mathscr{V}$-information metric, we analyze the complexities of 10,000 features, represented as directions in the penultimate layer, that were extracted from a standard ImageNet-trained vision model. Our study addresses four key questions: First, we ask what features look like as a function of complexity and find a spectrum of simple to complex features present within the model. Second, we ask when features are learned during training. We find that simpler features dominate early in training, and more complex features emerge gradually. Third, we investigate where within the network simple and complex features flow, and find that simpler features tend to bypass the visual hierarchy via residual connections. Fourth, we explore the connection between features complexity and their importance in driving the networks decision. We find that complex features tend to be less important. Surprisingly, important features become accessible at earlier layers during training, like a sedimentation process, allowing the model to build upon these foundational elements."
                    },
                    {
                        "title": "Learned feature representations are biased by complexity, learning order, position, and more",
                        "abstract": "Representation learning, and interpreting learned representations, are key areas of focus in machine learning and neuroscience. Both fields generally use representations as a means to understand or improve a system's computations. In this work, however, we explore surprising dissociations between representation and computation that may pose challenges for such efforts. We create datasets in which we attempt to match the computational role that different features play, while manipulating other properties of the features or the data. We train various deep learning architectures to compute these multiple abstract features about their inputs. We find that their learned feature representations are systematically biased towards representing some features more strongly than others, depending upon extraneous properties such as feature complexity, the order in which features are learned, and the distribution of features over the inputs. For example, features that are simpler to compute or learned first tend to be represented more strongly and densely than features that are more complex or learned later, even if all features are learned equally well. We also explore how these biases are affected by architectures, optimizers, and training regimes (e.g., in transformers, features decoded earlier in the output sequence also tend to be represented more strongly). Our results help to characterize the inductive biases of gradient-based representation learning. We then illustrate the downstream effects of these biases on various commonly-used methods for analyzing or intervening on representations. These results highlight a key challenge for interpretability $-$ or for comparing the representations of models and brains $-$ disentangling extraneous biases from the computationally important aspects of a system's internal representations."
                    }
                ]
            },
            "19fc38e8-dbca-432c-b30d-aba9f9fc5f91": {
                "pk": "19fc38e8-dbca-432c-b30d-aba9f9fc5f91",
                "name": "Hossein Mobahi",
                "collaborators": [
                    "Yann N. Dauphin",
                    "Atish Agarwala"
                ],
                "domain": [
                    "Deep Learning",
                    "Generalization",
                    "Hessian Analysis",
                    "Regularization"
                ],
                "publications": [
                    {
                        "title": "Neglected Hessian component explains mysteries in Sharpness regularization",
                        "abstract": "Recent work has shown that methods like SAM which either explicitly or implicitly penalize second order information can improve generalization in deep learning. Seemingly similar methods like weight noise and gradient penalties often fail to provide such benefits. We show that these differences can be explained by the structure of the Hessian of the loss. First, we show that a common decomposition of the Hessian can be quantitatively interpreted as separating the feature exploitation from feature exploration. The feature exploration, which can be described by the Nonlinear Modeling Error matrix (NME), is commonly neglected in the literature since it vanishes at interpolation. Our work shows that the NME is in fact important as it can explain why gradient penalties are sensitive to the choice of activation function. Using this insight we design interventions to improve performance. We also provide evidence that challenges the long held equivalence of weight noise and gradient penalties. This equivalence relies on the assumption that the NME can be ignored, which we find does not hold for modern networks since they involve significant feature learning. We find that regularizing feature exploitation but not feature exploration yields performance similar to gradient penalties."
                    }
                ]
            },
            "3065e42f-8dfa-4f60-8a7d-c5959321a045": {
                "pk": "3065e42f-8dfa-4f60-8a7d-c5959321a045",
                "name": "Thomas Fel",
                "collaborators": [
                    "Christopher Hamblin",
                    "Srijani Saha",
                    "Talia Konkle",
                    "George A. Alvarez"
                ],
                "domain": [
                    "Neural Network Interpretability",
                    "Feature Visualization",
                    "Computer Vision",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Feature Accentuation: Revealing 'What' Features Respond to in Natural Images",
                        "abstract": "Efforts to decode neural network vision models necessitate a comprehensive grasp of both the spatial and semantic facets governing feature responses within images. Most research has primarily centered around attribution methods, which provide explanations in the form of heatmaps, showing where the model directs its attention for a given feature. However, grasping 'where' alone falls short, as numerous studies have highlighted the limitations of those methods and the necessity to understand 'what' the model has recognized at the focal point of its attention. In parallel, 'Feature visualization' offers another avenue for interpreting neural network features. This approach synthesizes an optimal image through gradient ascent, providing clearer insights into 'what' features respond to. However, feature visualizations only provide one global explanation per feature; they do not explain why features activate for particular images. In this work, we introduce a new method to the interpretability tool-kit, 'feature accentuation', which is capable of conveying both where and what in arbitrary input images induces a feature's response. At its core, feature accentuation is image-seeded (rather than noise-seeded) feature visualization. We find a particular combination of parameterization, augmentation, and regularization yields naturalistic visualizations that resemble the seed image and target feature simultaneously. Furthermore, we validate these accentuations are processed along a natural circuit by the model. We make our precise implementation of feature accentuation available to the community as the Faccent library, an extension of Lucent."
                    }
                ]
            },
            "cf7f40d6-9250-4e7e-ac74-e9e2a35cbb8d": {
                "pk": "cf7f40d6-9250-4e7e-ac74-e9e2a35cbb8d",
                "name": "Michael C. Mozer",
                "collaborators": [
                    "R. Dodier",
                    "Marc Anderson",
                    "Lucky Vidmar",
                    "Robert F Cruickshank"
                ],
                "domain": [
                    "Crystallization",
                    "Chemical Engineering",
                    "Process Engineering"
                ],
                "publications": [
                    {
                        "title": "The Neural Network House: An Overview",
                        "abstract": "A process and apparatus for obtaining crystals of a particular substance from a flowing solution thereof, by supplying the seeding crystals to one nth portion of the saturated solution and supplying one nth portion of the solution to the crystalliser further downstream at each of (n-1) places. The supersaturation of the solution is maintained at the required value by keeping the static pressure so low that self-vaporisation occurs, whereas this static pressure is caused to drop gradually in the crystalliser from its inlet end toward its discharge end."
                    }
                ]
            }
        }
    },
    "2406.11813": {
        "paper_data": {
            "title": "How Do Large Language Models Acquire Factual Knowledge During Pretraining?",
            "url": "http://arxiv.org/abs/2406.11813v1",
            "arxiv_id": "2406.11813",
            "authors": [
                "Hoyeon Chang",
                "Jinho Park",
                "Seonghyeon Ye",
                "Sohee Yang",
                "Youngkyung Seo",
                "Du-Seong Chang",
                "Minjoon Seo"
            ],
            "abstract": "Despite the recent observation that large language models (LLMs) can store substantial factual knowledge, there is a limited understanding of the mechanisms of how they acquire factual knowledge through pretraining. This work addresses this gap by studying how LLMs acquire factual knowledge during pretraining. The findings reveal several important insights into the dynamics of factual knowledge acquisition during pretraining. First, counterintuitively, we observe that pretraining on more data shows no significant improvement in the model's capability to acquire and maintain factual knowledge. Next, there is a power-law relationship between training steps and forgetting of memorization and generalization of factual knowledge, and LLMs trained with duplicated training data exhibit faster forgetting. Third, training LLMs with larger batch sizes can enhance the models' robustness to forgetting. Overall, our observations suggest that factual knowledge acquisition in LLM pretraining occurs by progressively increasing the probability of factual knowledge presented in the pretraining data at each step. However, this increase is diluted by subsequent forgetting. Based on this interpretation, we demonstrate that we can provide plausible explanations for recently observed behaviors of LLMs, such as the poor performance of LLMs on long-tail knowledge and the benefits of deduplicating the pretraining corpus.",
            "introduction": "   1 Introduction  Recent studies on LLMs have shown their ability to capture substantial factual knowledge from the pretraining data [12, 29, 33]. Unfortunately, little is understood about the mechanisms of how LLMs acquire factual knowledge during pretraining. In this work, we make an initial attempt to understand the dynamics of factual knowledge acquisition in LLM pretraining. We study three important yet unanswered research questions:   RQ1.  How is factual knowledge acquired during LLM pretraining and how are LLMs affected by the training data at each training step?    RQ2.  How is the effectivity of factual knowledge acquisition affected by training conditions?    RQ3.  How is the acquired factual knowledge forgotten, and how is the trend affected by training conditions?      To answer the research questions, we analyze how LLMs acquire and retain factual knowledge in terms of memorization and generalization by varying the following training conditions: knowledge injection scenarios, pretraining stages, model sizes, and training batch sizes. Specifically, we take the intermediate pretraining checkpoints of different sizes of an LLM at different pretraining stages, inject the target knowledge that the models have not previously encountered, and monitor their step-wise progress of acquiring factual knowledge under various conditions.   Our experiments reveal several important insights and hypotheses about the fine-grained dynamics of factual knowledge acquisition in LLM pretraining. First, we show that factual knowledge acquisition occurs by accumulating the small increase of probability induced by updating the model with a minibatch containing the factual knowledge. Second, compared to the checkpoints at earlier stages, the checkpoint at the later stage shows no significant difference in effectivity, i.e., no significant improvement in the ability to acquire memorization and generalization immediately. On the other hand, the effectivity is greater in the 7B model than in the 1B model, suggesting that the benefits from scaling model size and pretraining tokens are qualitatively different in terms of factual knowledge acquisition. Third, we find a power-law relationship between training steps (or tokens) and forgetting of acquired factual knowledge in both memorization and generalization. Further examination of the rate of forgetting factual knowledge in LLM pretraining reveals that deduplicating the training data and training the models with a greater batch size enhances the acquisition of factual knowledge, by making them more robust against forgetting. Based on our understanding of the dynamics of factual knowledge acquisition, we demonstrate that the recently observed behaviors, including the improvement of LLMs\u2019 performance with more training data, the failure to acquire long-tail knowledge [20, 27], and the importance of dataset deduplication [23, 44] can be explained.   Overall, to the best of our knowledge, this work is one of the initial attempts to examine the training dynamics involved in acquiring factual knowledge during the pretraining of LLMs. By enhancing our understanding of the factual knowledge acquisition dynamics, we expect that academia can gain a holistic understanding and make better use of LLMs.     2 Related Work  Recently, there has been a surge in interest in LLMs [9, 11, 15, 17, 42]. [17] and [21] reported that the performance of LLMs adheres to a scaling law, correlating positively with both the model size and the size of the pretraining corpus. Extensive studies have examined the knowledge encoded in the parameters of LLMs [29,",
            "references": [
                {
                    "title": "OLMo: Accelerating the Science of Language Models",
                    "abstract": "Language models (LMs) have become ubiquitous in both NLP research and in commercial product offerings. As their commercial importance has surged, the most powerful models have become closed off, gated behind proprietary interfaces, with important details of their training data, architectures, and development undisclosed. Given the importance of these details in scientifically studying these models, including their biases and potential risks, we believe it is essential for the research community to have access to powerful, truly open LMs. To this end, we have built OLMo, a competitive, truly Open Language Model, to enable the scientific study of language models. Unlike most prior efforts that have only released model weights and inference code, we release OLMo alongside open training data and training and evaluation code. We hope this release will empower the open research community and inspire a new wave of innovation."
                },
                {
                    "title": "Dolma: an Open Corpus of Three Trillion Tokens for Language Model Pretraining Research",
                    "abstract": "Information about pretraining corpora used to train the current best-performing language models is seldom discussed: commercial models rarely detail their data, and even open models are often released without accompanying training data or recipes to reproduce them. As a result, it is challenging to conduct and advance scientific research on language modeling, such as understanding how training data impacts model capabilities and limitations. To facilitate scientific research on language model pretraining, we curate and release Dolma, a three-trillion-token English corpus, built from a diverse mixture of web content, scientific papers, code, public-domain books, social media, and encyclopedic materials. We extensively document Dolma, including its design principles, details about its construction, and a summary of its contents. We present analyses and experimental results on intermediate states of Dolma to share what we have learned about important data curation practices. Finally, we open-source our data curation toolkit to enable reproduction of our work as well as support further research in large-scale data curation."
                },
                {
                    "title": "Critical Data Size of Language Models from a Grokking Perspective",
                    "abstract": "We explore the critical data size in language models, a threshold that marks a fundamental shift from quick memorization to slow generalization. We formalize the phase transition under the grokking configuration into the Data Efficiency Hypothesis and identify data insufficiency, sufficiency, and surplus regimes in language models training dynamics. We develop a grokking configuration to reproduce grokking on simplistic language models stably by rescaling initialization and weight decay. We show that generalization occurs only when language models reach a critical size. We analyze grokking across sample-wise and model-wise, verifying the proposed data efficiency hypothesis. Our experiments reveal smoother phase transitions occurring at the critical dataset size for language datasets. As the model size increases, this critical point also becomes larger, indicating that larger models require more data. Our results deepen the understanding of language model training, offering a novel perspective on the role of data in the learning mechanism of language models."
                },
                {
                    "title": "Physics of Language Models: Part 3.1, Knowledge Storage and Extraction",
                    "abstract": "Large language models (LLMs) can store a vast amount of world knowledge, often extractable via question-answering (e.g.,\"What is Abraham Lincoln's birthday?\"). However, do they answer such questions based on exposure to similar questions during training (i.e., cheating), or by genuinely learning to extract knowledge from sources like Wikipedia? In this paper, we investigate this issue using a controlled biography dataset. We find a strong correlation between the model's ability to extract knowledge and various diversity measures of the training data. $\\textbf{Essentially}$, for knowledge to be reliably extracted, it must be sufficiently augmented (e.g., through paraphrasing, sentence shuffling, translations) $\\textit{during pretraining}$. Without such augmentation, knowledge may be memorized but not extractable, leading to 0% accuracy, regardless of subsequent instruction fine-tuning. To understand why this occurs, we employ (nearly) linear probing to demonstrate a strong connection between the observed correlation and how the model internally encodes knowledge -- whether it is linearly encoded in the hidden embeddings of entity names or distributed across other token embeddings in the training text. This paper provides $\\textbf{several key recommendations for LLM pretraining in the industry}$: (1) rewrite the pretraining data -- using small, auxiliary models -- to provide knowledge augmentation, and (2) incorporate more instruction-finetuning data into the pretraining stage before it becomes too late."
                },
                {
                    "title": "Physics of Language Models: Part 3.2, Knowledge Manipulation",
                    "abstract": "Language models can store vast factual knowledge, yet their ability to flexibly use this knowledge for downstream tasks (e.g., via instruction finetuning) remains questionable. This paper investigates four fundamental knowledge manipulation tasks: retrieval (e.g.,\"What is person A's attribute X?\"), classification (e.g.,\"Is A's attribute X even or odd?\"), comparison (e.g.,\"Is A greater than B in attribute X?\"), and inverse search (e.g.,\"Which person's attribute X equals T?\"). We show that language models excel in knowledge retrieval but struggle even in the simplest classification or comparison tasks unless Chain of Thoughts (CoTs) are employed during both training and inference. Moreover, their performance in inverse knowledge search is virtually 0%, regardless of the prompts. Our primary contribution is a controlled, synthetic experiment that confirms these weaknesses are inherent to language models: they cannot efficiently manipulate knowledge from pre-training data, even when such knowledge is perfectly stored in the models, despite adequate training and sufficient model size. Our findings also apply to modern pretrained language models such as GPT-4, thus giving rise to many Turing tests to distinguish Humans from contemporary AIs."
                },
                {
                    "title": "D4: Improving LLM Pretraining via Document De-Duplication and Diversification",
                    "abstract": "Over recent years, an increasing amount of compute and data has been poured into training large language models (LLMs), usually by doing one-pass learning on as many tokens as possible randomly selected from large-scale web corpora. While training on ever-larger portions of the internet leads to consistent performance improvements, the size of these improvements diminishes with scale, and there has been little work exploring the effect of data selection on pre-training and downstream performance beyond simple de-duplication methods such as MinHash. Here, we show that careful data selection (on top of de-duplicated data) via pre-trained model embeddings can speed up training (20% efficiency gains) and improves average downstream accuracy on 16 NLP tasks (up to 2%) at the 6.7B model scale. Furthermore, we show that repeating data intelligently consistently outperforms baseline training (while repeating random data performs worse than baseline training). Our results indicate that clever data selection can significantly improve LLM pre-training, calls into question the common practice of training for a single epoch on as much data as possible, and demonstrates a path to keep improving our models past the limits of randomly sampling web data."
                },
                {
                    "title": "An Empirical Study of Catastrophic Forgetting in Large Language Models During Continual Fine-tuning",
                    "abstract": "Catastrophic forgetting (CF) is a phenomenon that occurs in machine learning when a model forgets previously learned information while acquiring new knowledge. As large language models (LLMs) have demonstrated remarkable performance, it is intriguing to investigate whether CF exists during the continual instruction tuning of LLMs. This study empirically evaluates the forgetting phenomenon in LLMs' knowledge during continual instruction tuning from the perspectives of domain knowledge, reasoning, and reading comprehension. The experiments reveal that catastrophic forgetting is generally observed in LLMs ranging from 1b to 7b parameters. Moreover, as the model scale increases, the severity of forgetting intensifies. Comparing the decoder-only model BLOOMZ with the encoder-decoder model mT0, BLOOMZ exhibits less forgetting and retains more knowledge. Interestingly, we also observe that LLMs can mitigate language biases, such as gender bias, during continual fine-tuning. Furthermore, our findings indicate that ALPACA maintains more knowledge and capacity compared to LLAMA during continual fine-tuning, suggesting that general instruction tuning can help alleviate the forgetting phenomenon in LLMs during subsequent fine-tuning processes."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "To Repeat or Not To Repeat: Insights from Scaling LLM under Token-Crisis",
                    "abstract": "Recent research has highlighted the importance of dataset size in scaling language models. However, large language models (LLMs) are notoriously token-hungry during pre-training, and high-quality text data on the web is approaching its scaling limit for LLMs. To further enhance LLMs, a straightforward approach is to repeat the pre-training data for additional epochs. In this study, we empirically investigate three key aspects under this approach. First, we explore the consequences of repeating pre-training data, revealing that the model is susceptible to overfitting, leading to multi-epoch degradation. Second, we examine the key factors contributing to multi-epoch degradation, finding that significant factors include dataset size, model parameters, and training objectives, while less influential factors consist of dataset quality and model FLOPs. Finally, we explore whether widely used regularization can alleviate multi-epoch degradation. Most regularization techniques do not yield significant improvements, except for dropout, which demonstrates remarkable effectiveness but requires careful tuning when scaling up the model size. Additionally, we discover that leveraging mixture-of-experts (MoE) enables cost-effective and efficient hyper-parameter tuning for computationally intensive dense LLMs with comparable trainable parameters, potentially impacting efficient LLM development on a broader scale."
                },
                {
                    "title": "StarCoder: may the source be with you!",
                    "abstract": "The BigCode community, an open-scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder and StarCoderBase: 15.5B parameter models with 8K context length, infilling capabilities and fast large-batch inference enabled by multi-query attention. StarCoderBase is trained on 1 trillion tokens sourced from The Stack, a large collection of permissively licensed GitHub repositories with inspection tools and an opt-out process. We fine-tuned StarCoderBase on 35B Python tokens, resulting in the creation of StarCoder. We perform the most comprehensive evaluation of Code LLMs to date and show that StarCoderBase outperforms every open Code LLM that supports multiple programming languages and matches or outperforms the OpenAI code-cushman-001 model. Furthermore, StarCoder outperforms every model that is fine-tuned on Python, can be prompted to achieve 40\\% pass@1 on HumanEval, and still retains its performance on other programming languages. We take several important steps towards a safe open-access model release, including an improved PII redaction pipeline and a novel attribution tracing tool, and make the StarCoder models publicly available under a more commercially viable version of the Open Responsible AI Model license."
                },
                {
                    "title": "Emergent and Predictable Memorization in Large Language Models",
                    "abstract": "Memorization, or the tendency of large language models (LLMs) to output entire sequences from their training data verbatim, is a key concern for safely deploying language models. In particular, it is vital to minimize a model's memorization of sensitive datapoints such as those containing personal identifiable information (PII). The prevalence of such undesirable memorization can pose issues for model trainers, and may even require discarding an otherwise functional model. We therefore seek to predict which sequences will be memorized before a large model's full train-time by extrapolating the memorization behavior of lower-compute trial runs. We measure memorization of the Pythia model suite and plot scaling laws for forecasting memorization, allowing us to provide equi-compute recommendations to maximize the reliability (recall) of such predictions. We additionally provide further novel discoveries on the distribution of memorization scores across models and data. We release all code and data necessary to reproduce the results in this paper at https://github.com/EleutherAI/pythia"
                },
                {
                    "title": "Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling",
                    "abstract": "How do large language models (LLMs) develop and evolve over the course of training? How do these patterns change as models scale? To answer these questions, we introduce \\textit{Pythia}, a suite of 16 LLMs all trained on public data seen in the exact same order and ranging in size from 70M to 12B parameters. We provide public access to 154 checkpoints for each one of the 16 models, alongside tools to download and reconstruct their exact training dataloaders for further study. We intend \\textit{Pythia} to facilitate research in many areas, and we present several case studies including novel results in memorization, term frequency effects on few-shot performance, and reducing gender bias. We demonstrate that this highly controlled setup can be used to yield novel insights toward LLMs and their training dynamics. Trained models, analysis code, training code, and training data can be found at \\url{https://github.com/EleutherAI/pythia}."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "The BigScience ROOTS Corpus: A 1.6TB Composite Multilingual Dataset",
                    "abstract": "As language models grow ever larger, the need for large-scale high-quality text datasets has never been more pressing, especially in multilingual settings. The BigScience workshop, a 1-year international and multidisciplinary initiative, was formed with the goal of researching and training large language models as a values-driven undertaking, putting issues of ethics, harm, and governance in the foreground. This paper documents the data creation and curation efforts undertaken by BigScience to assemble the Responsible Open-science Open-collaboration Text Sources (ROOTS) corpus, a 1.6TB dataset spanning 59 languages that was used to train the 176-billion-parameter BigScience Large Open-science Open-access Multilingual (BLOOM) language model. We further release a large initial subset of the corpus and analyses thereof, and hope to empower large-scale monolingual and multilingual modeling projects with both the data and the processing tools, as well as stimulate research around this large multilingual corpus."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memories",
                    "abstract": "Despite their impressive performance on diverse tasks, large language models (LMs) still struggle with tasks requiring rich world knowledge, implying the difficulty of encoding a wealth of world knowledge in their parameters. This paper aims to understand LMs\u2019 strengths and limitations in memorizing factual knowledge, by conducting large-scale knowledge probing experiments on two open-domain entity-centric QA datasets: PopQA, our new dataset with 14k questions about long-tail entities, and EntityQuestions, a widely used open-domain QA dataset. We find that LMs struggle with less popular factual knowledge, and that retrieval augmentation helps significantly in these cases. Scaling, on the other hand, mainly improves memorization of popular knowledge, and fails to appreciably improve memorization of factual knowledge in the tail. Based on those findings, we devise a new method for retrieval-augmentation that improves performance and reduces inference costs by only retrieving non-parametric memories when necessary."
                },
                {
                    "title": "Large Language Models Struggle to Learn Long-Tail Knowledge",
                    "abstract": "The Internet contains a wealth of knowledge -- from the birthdays of historical figures to tutorials on how to code -- all of which may be learned by language models. However, while certain pieces of information are ubiquitous on the web, others appear extremely rarely. In this paper, we study the relationship between the knowledge memorized by large language models and the information in pre-training datasets scraped from the web. In particular, we show that a language model's ability to answer a fact-based question relates to how many documents associated with that question were seen during pre-training. We identify these relevant documents by entity linking pre-training datasets and counting documents that contain the same entities as a given question-answer pair. Our results demonstrate strong correlational and causal relationships between accuracy and relevant document count for numerous question answering datasets (e.g., TriviaQA), pre-training corpora (e.g., ROOTS), and model sizes (e.g., 176B parameters). Moreover, while larger models are better at learning long-tail knowledge, we estimate that today's models must be scaled by many orders of magnitude to reach competitive QA performance on questions with little support in the pre-training data. Finally, we show that retrieval-augmentation can reduce the dependence on relevant pre-training information, presenting a promising approach for capturing the long-tail."
                },
                {
                    "title": "Noise-Robust De-Duplication at Scale",
                    "abstract": "Identifying near duplicates within large, noisy text corpora has a myriad of applications that range from de-duplicating training datasets, reducing privacy risk, and evaluating test set leakage, to identifying reproduced news articles and literature within large corpora. Across these diverse applications, the overwhelming majority of work relies on N-grams. Limited efforts have been made to evaluate how well N-gram methods perform, in part because it is unclear how one could create an unbiased evaluation dataset for a massive corpus. This study uses the unique timeliness of historical news wires to create a 27,210 document dataset, with 122,876 positive duplicate pairs, for studying noise-robust de-duplication. The time-sensitivity of news makes comprehensive hand labelling feasible - despite the massive overall size of the corpus - as duplicates occur within a narrow date range. The study then develops and evaluates a range of de-duplication methods: hashing and N-gram overlap (which predominate in the literature), a contrastively trained bi-encoder, and a re-rank style approach combining a bi- and cross-encoder. The neural approaches significantly outperform hashing and N-gram overlap. We show that the bi-encoder scales well, de-duplicating a 10 million article corpus on a single GPU card in a matter of hours. We also apply our pre-trained model to the RealNews and patent portions of C4 (Colossal Clean Crawled Corpus), illustrating that a neural approach can identify many near duplicates missed by hashing, in the presence of various types of noise. The public release of our NEWS-COPY de-duplication dataset, codebase, and the pre-trained models will facilitate further research and applications."
                },
                {
                    "title": "Measuring Causal Effects of Data Statistics on Language Model's 'Factual' Predictions",
                    "abstract": "Large amounts of training data are one of the major reasons for the high performance of state-of-the-art NLP models. But what exactly in the training data causes a model to make a certain prediction? We seek to answer this question by providing a language for describing how training data influences predictions, through a causal framework. Importantly, our framework bypasses the need to retrain expensive models and allows us to estimate causal effects based on observational data alone. Addressing the problem of extracting factual knowledge from pretrained language models (PLMs), we focus on simple data statistics such as co-occurrence counts and show that these statistics do influence the predictions of PLMs, suggesting that such models rely on shallow heuristics. Our causal framework and our results demonstrate the importance of studying datasets and the benefits of causality for understanding NLP models."
                },
                {
                    "title": "Emergent Abilities of Large Language Models",
                    "abstract": "Scaling up language models has been shown to predictably improve performance and sample efficiency on a wide range of downstream tasks. This paper instead discusses an unpredictable phenomenon that we refer to as emergent abilities of large language models. We consider an ability to be emergent if it is not present in smaller models but is present in larger models. Thus, emergent abilities cannot be predicted simply by extrapolating the performance of smaller models. The existence of such emergence implies that additional scaling could further expand the range of capabilities of language models."
                },
                {
                    "title": "Memorization Without Overfitting: Analyzing the Training Dynamics of Large Language Models",
                    "abstract": "Despite their wide adoption, the underlying training and memorization dynamics of very large language models is not well understood. We empirically study exact memorization in causal and masked language modeling, across model sizes and throughout the training process. We measure the effects of dataset size, learning rate, and model size on memorization, finding that larger language models memorize training data faster across all settings. Surprisingly, we show that larger models can memorize a larger portion of the data before over-fitting and tend to forget less throughout the training process. We also analyze the memorization dynamics of different parts of speech and find that models memorize nouns and numbers first; we hypothesize and provide empirical evidence that nouns and numbers act as a unique identifier for memorizing individual training examples. Together, these findings present another piece of the broader puzzle of trying to understand what actually improves as models get bigger."
                },
                {
                    "title": "Entity Cloze By Date: What LMs Know About Unseen Entities",
                    "abstract": "Language models (LMs) are typically trained once on a large-scale corpus and used for years without being updated. However, in a dynamic world, new entities constantly arise. We propose a framework to analyze what LMs can infer about new entities that did not exist when the LMs were pretrained. We derive a dataset of entities indexed by their origination date and paired with their English Wikipedia articles, from which we can find sentences about each entity. We evaluate LMs' perplexity on masked spans within these sentences. We show that models more informed about the entities, such as those with access to a textual definition of them, achieve lower perplexity on this benchmark. Our experimental results demonstrate that making inferences about new entities remains difficult for LMs. Given its wide coverage on entity knowledge and temporal indexing, our dataset can be used to evaluate LMs and techniques designed to modify or extend their knowledge. Our automatic data collection pipeline can be easily used to continually update our benchmark."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "How Pre-trained Language Models Capture Factual Knowledge? A Causal-Inspired Analysis",
                    "abstract": "Recently, there has been a trend to investigate the factual knowledge captured by Pre-trained Language Models (PLMs). Many works show the PLMs\u2019 ability to fill in the missing factual words in cloze-style prompts such as \u201dDante was born in [MASK].\u201d However, it is still a mystery how PLMs generate the results correctly: relying on effective clues or shortcut patterns? We try to answer this question by a causal-inspired analysis that quantitatively measures and evaluates the word-level patterns that PLMs depend on to generate the missing words. We check the words that have three typical associations with the missing words: knowledge-dependent, positionally close, and highly co-occurred. Our analysis shows: (1) PLMs generate the missing factual words more by the positionally close and highly co-occurred words than the knowledge-dependent words; (2) the dependence on the knowledge-dependent words is more effective than the positionally close and highly co-occurred words. Accordingly, we conclude that the PLMs capture the factual knowledge ineffectively because of depending on the inadequate associations."
                },
                {
                    "title": "Training Compute-Optimal Large Language Models",
                    "abstract": "We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over 400 language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, Chinchilla, that uses the same compute budget as Gopher but with 70B parameters and 4$\\times$ more more data. Chinchilla uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that Chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, Chinchilla reaches a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, greater than a 7% improvement over Gopher."
                },
                {
                    "title": "Quantifying Memorization Across Neural Language Models",
                    "abstract": "Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations."
                },
                {
                    "title": "Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets",
                    "abstract": "In this paper we propose to study generalization of neural networks on small algorithmically generated datasets. In this setting, questions about data efficiency, memorization, generalization, and speed of learning can be studied in great detail. In some situations we show that neural networks learn through a process of\"grokking\"a pattern in the data, improving generalization performance from random chance level to perfect generalization, and that this improvement in generalization can happen well past the point of overfitting. We also study generalization as a function of dataset size and find that smaller datasets require increasing amounts of optimization for generalization. We argue that these datasets provide a fertile ground for studying a poorly understood aspect of deep learning: generalization of overparametrized neural networks beyond memorization of the finite training dataset."
                },
                {
                    "title": "Deduplicating Training Data Makes Language Models Better",
                    "abstract": "We find that existing language modeling datasets contain many near-duplicate examples and long repetitive substrings.As a result, over 1% of the unprompted output of language models trained on these datasets is copied verbatim from the training data.We develop two tools that allow us to deduplicate training datasets\u2014for example removing from C4 a single 61 word English sentence that is repeated over 60,000 times.Deduplication allows us to train models that emit memorized text ten times less frequently and require fewer training steps to achieve the same or better accuracy.We can also reduce train-test overlap, which affects over 4% of the validation set of standard datasets, thus allowing for more accurate evaluation.Code for deduplication is released at https://github.com/google-research/deduplicate-text-datasets."
                },
                {
                    "title": "Memorisation versus Generalisation in Pre-trained Language Models",
                    "abstract": "State-of-the-art pre-trained language models have been shown to memorise facts and perform well with limited amounts of training data. To gain a better understanding of how these models learn, we study their generalisation and memorisation capabilities in noisy and low-resource scenarios. We find that the training of these models is almost unaffected by label noise and that it is possible to reach near-optimal results even on extremely noisy datasets. However, our experiments also show that they mainly learn from high-frequency patterns and largely fail when tested on low-resource tasks such as few-shot learning and rare entity recognition. To mitigate such limitations, we propose an extension based on prototypical networks that improves performance in low-resource named entity recognition tasks."
                },
                {
                    "title": "Analyzing Commonsense Emergence in Few-shot Knowledge Models",
                    "abstract": "Recently, commonsense knowledge models - pretrained language models (LM) fine-tuned on knowledge graph (KG) tuples - showed that considerable amounts of commonsense knowledge can be encoded in the parameters of large language models. However, as parallel studies show that LMs are poor hypothesizers of declarative commonsense relationships on their own, it remains unclear whether this knowledge is learned during pretraining or from fine-tuning on KG examples. To investigate this question, we train commonsense knowledge models in few-shot settings to study the emergence of their commonsense representation abilities. Our results show that commonsense knowledge models can rapidly adapt from limited examples, indicating that KG fine-tuning serves to learn an interface to encoded knowledge learned during pretraining. Importantly, our analysis of absolute, angular, and distributional parameter changes during few-shot fine-tuning provides novel insights into how this interface is learned."
                },
                {
                    "title": "Anatomy of Catastrophic Forgetting: Hidden Representations and Task Semantics",
                    "abstract": "A central challenge in developing versatile machine learning systems is catastrophic forgetting: a model trained on tasks in sequence will suffer significant performance drops on earlier tasks. Despite the ubiquity of catastrophic forgetting, there is limited understanding of the underlying process and its causes. In this paper, we address this important knowledge gap, investigating how forgetting affects representations in neural network models. Through representational analysis techniques, we find that deeper layers are disproportionately the source of forgetting. Supporting this, a study of methods to mitigate forgetting illustrates that they act to stabilize deeper layers. These insights enable the development of an analytic argument and empirical picture relating the degree of forgetting to representational similarity between tasks. Consistent with this picture, we observe maximal forgetting occurs for task sequences with intermediate similarity. We perform empirical studies on the standard split CIFAR-10 setup and also introduce a novel CIFAR-100 based task approximating realistic input distribution shift."
                },
                {
                    "title": "The Surprising Simplicity of the Early-Time Learning Dynamics of Neural Networks",
                    "abstract": "Modern neural networks are often regarded as complex black-box functions whose behavior is difficult to understand owing to their nonlinear dependence on the data and the nonconvexity in their loss landscapes. In this work, we show that these common perceptions can be completely false in the early phase of learning. In particular, we formally prove that, for a class of well-behaved input distributions, the early-time learning dynamics of a two-layer fully-connected neural network can be mimicked by training a simple linear model on the inputs. We additionally argue that this surprising simplicity can persist in networks with more layers and with convolutional architecture, which we verify empirically. Key to our analysis is to bound the spectral norm of the difference between the Neural Tangent Kernel (NTK) at initialization and an affine transform of the data kernel; however, unlike many previous results utilizing the NTK, we do not require the network to have disproportionately large width, and the network is allowed to escape the kernel regime later in training."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "How Much Knowledge Can You Pack into the Parameters of a Language Model?",
                    "abstract": "It has recently been observed that neural language models trained on unstructured text can implicitly store and retrieve knowledge using natural language queries. In this short paper, we measure the practical utility of this approach by fine-tuning pre-trained models to answer questions without access to any external context or knowledge. We show that this approach scales surprisingly well with model size and outperforms models that explicitly look up knowledge on the open-domain variants of Natural Questions and WebQuestions. To facilitate reproducibility and future work, we release our code and trained models."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "Language Models as Knowledge Bases?",
                    "abstract": "Recent progress in pretraining language models on large textual corpora led to a surge of improvements for downstream NLP tasks. Whilst learning linguistic knowledge, these models may also be storing relational knowledge present in the training data, and may be able to answer queries structured as \u201cfill-in-the-blank\u201d cloze statements. Language models have many advantages over structured knowledge bases: they require no schema engineering, allow practitioners to query about an open class of relations, are easy to extend to more data, and require no human supervision to train. We present an in-depth analysis of the relational knowledge already present (without fine-tuning) in a wide range of state-of-the-art pretrained language models. We find that (i) without fine-tuning, BERT contains relational knowledge competitive with traditional NLP methods that have some access to oracle knowledge, (ii) BERT also does remarkably well on open-domain question answering against a supervised baseline, and (iii) certain types of factual knowledge are learned much more readily than others by standard language model pretraining approaches. The surprisingly strong ability of these models to recall factual knowledge without any fine-tuning demonstrates their potential as unsupervised open-domain QA systems. The code to reproduce our analysis is available at https://github.com/facebookresearch/LAMA."
                },
                {
                    "title": "Does learning require memorization? a short tale about a long tail",
                    "abstract": "State-of-the-art results on image recognition tasks are achieved using over-parameterized learning algorithms that (nearly) perfectly fit the training set and are known to fit well even random labels. This tendency to memorize seemingly useless training data labels is not explained by existing theoretical analyses. Memorization of the training data also presents significant privacy risks when the training data contains sensitive personal information and thus it is important to understand whether such memorization is necessary for accurate learning. We provide a simple conceptual explanation and a theoretical model demonstrating that for natural data distributions memorization of labels is necessary for achieving close-to-optimal generalization error. The model is motivated and supported by the results of several recent empirical works. In our model, data is sampled from a mixture of subpopulations and the frequencies of these subpopulations are chosen from some prior. The model allows to quantify the effect of not fitting the training data on the generalization performance of the learned classifier and demonstrates that memorization is necessary whenever frequencies are long-tailed. Image and text data are known to follow such distributions and therefore our results establish a formal link between these empirical phenomena. Our results also have concrete implications for the cost of ensuring differential privacy in learning."
                },
                {
                    "title": "A Closer Look at Memorization in Deep Networks",
                    "abstract": "We examine the role of memorization in deep learning, drawing connections to capacity, generalization, and adversarial robustness. While deep networks are capable of memorizing noise data, our results suggest that they tend to prioritize learning simple patterns first. In our experiments, we expose qualitative differences in gradient-based optimization of deep neural networks (DNNs) on noise vs. real data. We also demonstrate that for appropriately tuned explicit regularization (e.g., dropout) we can degrade DNN training performance on noise datasets without compromising generalization on real data. Our analysis suggests that the notions of effective capacity which are dataset independent are unlikely to explain the generalization performance of deep networks when trained with gradient based methods because training data itself plays an important role in determining the degree of memorization."
                },
                {
                    "title": "Are Emergent Abilities of Large Language Models a Mirage?",
                    "abstract": "Recent work claims that large language models display emergent abilities, abilities not present in smaller-scale models that are present in larger-scale models. What makes emergent abilities intriguing is two-fold: their sharpness, transitioning seemingly instantaneously from not present to present, and their unpredictability, appearing at seemingly unforeseeable model scales. Here, we present an alternative explanation for emergent abilities: that for a particular task and model family, when analyzing fixed model outputs, emergent abilities appear due to the researcher's choice of metric rather than due to fundamental changes in model behavior with scale. Specifically, nonlinear or discontinuous metrics produce apparent emergent abilities, whereas linear or continuous metrics produce smooth, continuous predictable changes in model performance. We present our alternative explanation in a simple mathematical model, then test it in three complementary ways: we (1) make, test and confirm three predictions on the effect of metric choice using the InstructGPT/GPT-3 family on tasks with claimed emergent abilities; (2) make, test and confirm two predictions about metric choices in a meta-analysis of emergent abilities on BIG-Bench; and (3) show to choose metrics to produce never-before-seen seemingly emergent abilities in multiple vision tasks across diverse deep networks. Via all three analyses, we provide evidence that alleged emergent abilities evaporate with different metrics or with better statistics, and may not be a fundamental property of scaling AI models."
                },
                {
                    "title": "Tracing Knowledge in Language Models Back to the Training Data",
                    "abstract": "Neural language models (LMs) have been shown to memorize a great deal of factual knowledge. But when an LM generates an assertion, it is often dif\ufb01cult to determine where it learned this information and whether it is true. In this paper, we introduce a new benchmark for fact tracing : tracing language models\u2019 assertions back to the training examples that provided evidence for those predictions. Prior work has suggested that dataset-level in-\ufb02uence methods might offer an effective framework for tracing predictions back to training data. However, such methods have not been evaluated for fact tracing, and researchers primarily have studied them through qualitative analysis or as a data cleaning technique for classi\ufb01cation/regression tasks. We present the \ufb01rst experiments that evaluate in\ufb02uence meth-ods for fact tracing, using well-understood information retrieval (IR) metrics. We compare two popular families of in\ufb02uence meth-ods \u2013 gradient -based and embedding -based \u2013 and show that neither can fact-trace reliably; indeed, both methods fail to outperform an IR baseline (BM25) that does not even access the LM. We explore why this occurs (e.g., gradient saturation) and demonstrate that existing in\ufb02u-ence methods must be improved signi\ufb01cantly before they can reliably attribute factual predictions in LMs. 1"
                },
                {
                    "title": "Emergent Structures and Training Dynamics in Large Language Models",
                    "abstract": "Large language models have achieved success on a number of downstream tasks, particularly in a few and zero-shot manner. As a consequence, researchers have been investigating both the kind of information these networks learn and how such information can be encoded in the parameters of the model. We survey the literature on changes in the network during training, drawing from work outside of NLP when necessary, and on learned representations of linguistic features in large language models. We note in particular the lack of sufficient research on the emergence of functional units, subsections of the network where related functions are grouped or organised, within large language models and motivate future work that grounds the study of language models in an analysis of their changing internal structure during training time."
                },
                {
                    "title": "Investigating Learning Dynamics of BERT Fine-Tuning",
                    "abstract": "The recently introduced pre-trained language model BERT advances the state-of-the-art on many NLP tasks through the fine-tuning approach, but few studies investigate how the fine-tuning process improves the model performance on downstream tasks. In this paper, we inspect the learning dynamics of BERT fine-tuning with two indicators. We use JS divergence to detect the change of the attention mode and use SVCCA distance to examine the change to the feature extraction mode during BERT fine-tuning. We conclude that BERT fine-tuning mainly changes the attention mode of the last layers and modifies the feature extraction mode of the intermediate and last layers. Moreover, we analyze the consistency of BERT fine-tuning between different random seeds and different datasets. In summary, we provide a distinctive understanding of the learning dynamics of BERT fine-tuning, which sheds some light on improving the fine-tuning results."
                }
            ],
            "categories": [
                "cs.CL",
                "I.2.7"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow is factual knowledge acquired during LLM pretraining and how are LLMs affected by the training data at each training step?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding the dynamics of factual knowledge acquisition in LLMs is crucial for the research community as it can lead to improved model design and training strategies. By elucidating how LLMs acquire and retain knowledge, researchers can develop more effective training protocols, enhance model performance, and address issues such as knowledge retention and forgetting. This knowledge can also inform practical applications in various fields, including natural language processing, information retrieval, and AI ethics, ultimately advancing the state of the art in machine learning.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of LLM architectures and the vast amount of data they are trained on. Naive approaches may fail because they do not account for the intricate interactions between model parameters and training data, nor do they consider the varying effects of different training conditions. Additionally, the theoretical understanding of how knowledge is encoded and forgotten in LLMs is still limited, making it difficult to design experiments that accurately capture these dynamics. Overcoming these obstacles requires sophisticated experimental designs and analytical methods to isolate the effects of various training parameters.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the performance of LLMs rather than the underlying mechanisms of knowledge acquisition. There has been a lack of detailed studies that systematically investigate the dynamics of factual knowledge acquisition during pretraining. Barriers such as the complexity of LLMs, the scale of data, and the difficulty in measuring knowledge retention have hindered progress. This work differs from prior studies by specifically examining the step-wise progress of knowledge acquisition under varied training conditions, providing a more granular understanding of the training dynamics involved.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves analyzing intermediate pretraining checkpoints of LLMs of different sizes at various stages, injecting target knowledge, and monitoring the step-wise acquisition of factual knowledge. Key components include varying training conditions such as knowledge injection scenarios, model sizes, and training batch sizes. The expected outcomes include insights into the accumulation of factual knowledge, the relationship between model size and knowledge acquisition effectiveness, and the dynamics of knowledge forgetting. This approach aims to provide a comprehensive understanding of how"
            }
        },
        "author_data": {
            "92992ef1-6541-4d01-ba7c-9cbd27f2f368": {
                "pk": "92992ef1-6541-4d01-ba7c-9cbd27f2f368",
                "name": "Hoyeon Chang",
                "collaborators": [
                    "Hyunji Lee",
                    "Minjoon Seo",
                    "Jaeyoung Kim",
                    "Hanseok Oh",
                    "Sohee Yang",
                    "Vlad Karpukhin",
                    "Yi Lu",
                    "Seongyun Lee",
                    "Geewook Kim",
                    "Jiyeon Kim"
                ],
                "domain": [
                    "Generative Models",
                    "Multimodal Learning",
                    "Safety in AI",
                    "Document Retrieval"
                ],
                "publications": [
                    {
                        "title": "Nonparametric Decoding for Generative Retrieval",
                        "abstract": "The generative retrieval model depends solely on the information encoded in its model parameters without external memory, its information capacity is limited and fixed. To overcome the limitation, we propose Nonparametric Decoding (Np Decoding) which can be applied to existing generative retrieval models. Np Decoding uses nonparametric contextualized vocab embeddings (external memory) rather than vanilla vocab embeddings as decoder vocab embeddings. By leveraging the contextualized vocab embeddings, the generative retrieval model is able to utilize both the parametric and nonparametric space. Evaluation over 9 datasets (8 single-hop and 1 multi-hop) in the document retrieval task shows that applying Np Decoding to generative retrieval models significantly improves the performance. We also show that Np Decoding is data- and parameter-efficient, and shows high performance in the zero-shot setting."
                    },
                    {
                        "title": "How Does Vision-Language Adaptation Impact the Safety of Vision Language Models?",
                        "abstract": "Vision-Language adaptation (VL adaptation) transforms Large Language Models (LLMs) into Large Vision-Language Models (LVLMs) for multimodal tasks, but this process often compromises the inherent safety capabilities embedded in the original LLMs. Despite potential harmfulness due to weakened safety measures, in-depth analysis on the effects of VL adaptation on safety remains under-explored. This study examines how VL adaptation influences safety and evaluates the impact of safety fine-tuning methods. Our analysis reveals that safety degradation occurs during VL adaptation, even when the training data is safe. While safety tuning techniques like supervised fine-tuning with safety datasets or reinforcement learning from human feedback mitigate some risks, they still lead to safety degradation and a reduction in helpfulness due to over-rejection issues. Further analysis of internal model weights suggests that VL adaptation may impact certain safety-related layers, potentially lowering overall safety levels. Additionally, our findings demonstrate that the objectives of VL adaptation and safety tuning are divergent, which often results in their simultaneous application being suboptimal. To address this, we suggest the weight merging approach as an optimal solution effectively reducing safety degradation while maintaining helpfulness. These insights help guide the development of more reliable and secure LVLMs for real-world applications."
                    }
                ]
            },
            "3166571f-a508-4d00-a9b0-aa457bfd92de": {
                "pk": "3166571f-a508-4d00-a9b0-aa457bfd92de",
                "name": "Jinho Park",
                "collaborators": [
                    "Doyoung Kim",
                    "Jongwon Lee",
                    "Minjoon Seo",
                    "Se Young Chun",
                    "Mingoo Seok"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Robotics",
                    "Machine Learning",
                    "Cognitive Science"
                ],
                "publications": [
                    {
                        "title": "How language models extrapolate outside the training data: A case study in Textualized Gridworld",
                        "abstract": "Language models' ability to extrapolate learned behaviors to novel, more complex environments beyond their training scope is highly unknown. This study introduces a path planning task in a textualized Gridworld to probe language models' extrapolation capabilities. We show that conventional approaches, including next-token prediction and Chain of Thought (CoT) fine-tuning, fail to generalize in larger, unseen environments. Inspired by human cognition and dual-process theory, we propose language models should construct cognitive maps before interaction. Our research demonstrates that autoregressive generation of cognitive maps and planning sequences enhances planning capabilities in extrapolated environments. Unlike CoT, we find that cognitive maps cannot be obtained through simple prompting, necessitating additional training schemes for integration. Our findings in Gridworld offer insights into training language models with improved reasoning and adaptability, potentially advancing more human-like cognition and opening avenues for enhancing model generalization across diverse, complex tasks."
                    },
                    {
                        "title": "UL-VIO: Ultra-lightweight Visual-Inertial Odometry with Noise Robust Test-time Adaptation",
                        "abstract": "Data-driven visual-inertial odometry (VIO) has received highlights for its performance since VIOs are a crucial compartment in autonomous robots. However, their deployment on resource-constrained devices is non-trivial since large network parameters should be accommodated in the device memory. Furthermore, these networks may risk failure post-deployment due to environmental distribution shifts at test time. In light of this, we propose UL-VIO -- an ultra-lightweight (<1M) VIO network capable of test-time adaptation (TTA) based on visual-inertial consistency. Specifically, we perform model compression to the network while preserving the low-level encoder part, including all BatchNorm parameters for resource-efficient test-time adaptation. It achieves 36X smaller network size than state-of-the-art with a minute increase in error -- 1% on the KITTI dataset. For test-time adaptation, we propose to use the inertia-referred network outputs as pseudo labels and update the BatchNorm parameter for lightweight yet effective adaptation. To the best of our knowledge, this is the first work to perform noise-robust TTA on VIO. Experimental results on the KITTI, EuRoC, and Marulan datasets demonstrate the effectiveness of our resource-efficient adaptation method under diverse TTA scenarios with dynamic domain shifts."
                    }
                ]
            },
            "454ac27c-a094-4db0-8062-dcfcf6ab5977": {
                "pk": "454ac27c-a094-4db0-8062-dcfcf6ab5977",
                "name": "Seonghyeon Ye",
                "collaborators": [
                    "Minjoon Seo",
                    "Joel Jang",
                    "Doyoung Kim",
                    "Seungone Kim",
                    "Sohee Yang",
                    "Joongbo Shin",
                    "Hyeonbin Hwang",
                    "Janghoon Han",
                    "Jiseon Kim",
                    "Alice Oh"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Continual Learning",
                    "Instruction Tuning",
                    "Prompt Engineering"
                ],
                "publications": [
                    {
                        "title": "Efficient Contrastive Learning via Novel Data Augmentation and Curriculum Learning",
                        "abstract": "We introduce EfficientCL, a memory-efficient continual pretraining method that applies contrastive learning with novel data augmentation and curriculum learning. For data augmentation, we stack two types of operation sequentially: cutoff and PCA jittering. While pretraining steps proceed, we apply curriculum learning by incrementing the augmentation degree for each difficulty step. After data augmentation is finished, contrastive learning is applied on projected embeddings of original and augmented examples. When finetuned on GLUE benchmark, our model outperforms baseline models, especially for sentence-level tasks. Additionally, this improvement is capable with only 70% of computational memory compared to the baseline model."
                    },
                    {
                        "title": "Can Large Language Models Truly Understand Prompts? A Case Study with Negated Prompts",
                        "abstract": "Previous work has shown that there exists a scaling law between the size of Language Models (LMs) and their zero-shot performance on different downstream NLP tasks. In this work, we show that this phenomenon does not hold when evaluating large LMs on tasks with negated prompts, but instead shows an inverse scaling law. We evaluate 9 different tasks with negated prompts on (1) pretrained LMs (OPT & GPT-3) of varying sizes (125M - 175B), (2) LMs further pretrained to generalize to novel prompts (InstructGPT), (3) LMs provided with few-shot examples, and (4) LMs fine-tuned specifically on negated prompts; all LM types perform worse on negated prompts as they scale and show a huge performance gap between the human performance when comparing the average score on both original and negated prompts. By highlighting a critical limitation of existing LMs and methods, we urge the community to develop new approaches of developing LMs that actually follow the given instructions. We provide the code and the datasets to explore negated prompts at https://github.com/joeljang/negated-prompts-for-llms"
                    },
                    {
                        "title": "Efficiently Enhancing Zero-Shot Performance of Instruction Following Model via Retrieval of Soft Prompt",
                        "abstract": "Enhancing the zero-shot performance of instruction-following models requires heavy computation, either by scaling the total number of training datasets or the model size. In this work, we explore how retrieval of soft prompts obtained through prompt tuning can efficiently assist hard prompts in zero-shot task generalization. Specifically, we train soft prompt embeddings for each prompt through prompt tuning, store the samples of the training instances mapped with the prompt embeddings, and retrieve the corresponding prompt embedding of the training instance closest to the query instance during inference. While only adding 0.007% additional parameters, retrieval of soft prompt enhances the performance of T0 on unseen tasks by outperforming it on 10 out of 11 datasets as well as improving the mean accuracy of T0 on BIG-bench benchmark by 2.39% points. Also, we report an interesting finding that retrieving source embeddings trained on similar answer choice formats is more important than those on similar task types."
                    },
                    {
                        "title": "Guess the Instruction! Flipped Learning Makes Language Models Stronger Zero-Shot Learners",
                        "abstract": "Meta-training, which fine-tunes the language model (LM) on various downstream tasks by maximizing the likelihood of the target label given the task instruction and input instance, has improved the zero-shot task generalization performance. However, meta-trained LMs still struggle to generalize to challenging tasks containing novel labels unseen during meta-training. In this paper, we propose Flipped Learning, an alternative method of meta-training which trains the LM to generate the task instruction given the input instance and label. During inference, the LM trained with Flipped Learning, referred to as Flipped, selects the label option that is most likely to generate the task instruction. On 14 tasks of the BIG-bench benchmark, the 11B-sized Flipped outperforms zero-shot T0-11B and even a 16 times larger 3-shot GPT-3 (175B) on average by 8.4% and 9.7% points, respectively. Flipped gives particularly large improvements on tasks with unseen labels, outperforming T0-11B by up to +20% average F1 score. This indicates that the strong task generalization of Flipped comes from improved generalization to novel labels. We release our code at https://github.com/seonghyeonye/Flipped-Learning."
                    },
                    {
                        "title": "Self-Explore: Enhancing Mathematical Reasoning in Language Models with Fine-grained Rewards",
                        "abstract": "Training on large amounts of rationales (i.e., CoT Fine-tuning) is effective at improving the reasoning capabilities of large language models (LLMs). However, acquiring human-authored rationales or augmenting rationales from proprietary models is costly and not scalable. In this paper, we study the problem of whether LLMs could self-improve their reasoning capabilities. To this end, we propose Self-Explore, where the LLM is tasked to explore the first wrong step (i.e., the first pit) within the rationale and use such signals as fine-grained rewards for further improvement. On the GSM8K and MATH test set, Self-Explore achieves 11.57% and 2.89% improvement on average across three LLMs compared to supervised fine-tuning (SFT). Our code is available at https://github.com/hbin0701/Self-Explore."
                    },
                    {
                        "title": "Investigating the Effectiveness of Task-Agnostic Prefix Prompt for Instruction Following",
                        "abstract": "In this paper, we present our finding that prepending a Task-Agnostic Prefix Prompt (TAPP) to the input improves the instruction-following ability of various Large Language Models (LLMs) during inference. TAPP is different from canonical prompts for LLMs in that it is a fixed prompt prepended to the beginning of every input regardless of the target task for zero-shot generalization. We observe that both base LLMs (i.e. not fine-tuned to follow instructions) and instruction-tuned models benefit from TAPP, resulting in 34.58% and 12.26% improvement on average, respectively. This implies that the instruction-following ability of LLMs can be improved during inference time with a fixed prompt constructed with simple heuristics. We hypothesize that TAPP assists language models to better estimate the output distribution by focusing more on the instruction of the target task during inference. In other words, such ability does not seem to be sufficiently activated in not only base LLMs but also many instruction-fine-tuned LLMs. All experiments are reproducible from https://github.com/seonghyeonye/TAPP."
                    },
                    {
                        "title": "Dimensional Emotion Detection from Categorical Emotion",
                        "abstract": "We present a model to predict fine-grained emotions along the continuous dimensions of valence, arousal, and dominance (VAD) with a corpus with categorical emotion annotations. Our model is trained by minimizing the EMD (Earth Mover's Distance) loss between the predicted VAD score distribution and the categorical emotion distributions sorted along VAD, and it can simultaneously classify the emotion categories and predict the VAD scores for a given sentence. We use pre-trained RoBERTa-Large and fine-tune on three different corpora with categorical labels and evaluate on EmoBank corpus with VAD scores. We show that our approach reaches comparable performance to that of the state-of-the-art classifiers in categorical emotion classification and shows significant positive correlations with the ground truth VAD scores. Also, further training with supervision of VAD labels leads to improved performance especially when dataset is small. We also present examples of predictions of appropriate emotion words that are not part of the original annotations."
                    },
                    {
                        "title": "INSTRUCTIR: A Benchmark for Instruction Following of Information Retrieval Models",
                        "abstract": "Despite the critical need to align search targets with users' intention, retrievers often only prioritize query information without delving into the users' intended search context. Enhancing the capability of retrievers to understand intentions and preferences of users, akin to language model instructions, has the potential to yield more aligned search targets. Prior studies restrict the application of instructions in information retrieval to a task description format, neglecting the broader context of diverse and evolving search scenarios. Furthermore, the prevailing benchmarks utilized for evaluation lack explicit tailoring to assess instruction-following ability, thereby hindering progress in this field. In response to these limitations, we propose a novel benchmark,INSTRUCTIR, specifically designed to evaluate instruction-following ability in information retrieval tasks. Our approach focuses on user-aligned instructions tailored to each query instance, reflecting the diverse characteristics inherent in real-world search scenarios. Through experimental analysis, we observe that retrievers fine-tuned to follow task-style instructions, such as INSTRUCTOR, can underperform compared to their non-instruction-tuned counterparts. This underscores potential overfitting issues inherent in constructing retrievers trained on existing instruction-aware retrieval datasets."
                    },
                    {
                        "title": "Improving Probability-based Prompt Selection Through Unified Evaluation and Analysis",
                        "abstract": "Previous works in prompt engineering for large language models have introduced different gradient-free probability-based prompt selection methods that aim to choose the optimal prompt among the candidates for a given task but have failed to provide a comprehensive and fair comparison between each other. In this paper, we propose a unified framework to interpret and evaluate the existing probability-based prompt selection methods by performing extensive experiments on 13 common and diverse NLP tasks. We find that each of the existing methods can be interpreted as some variant of the method that maximizes mutual information between the input and the predicted output (MI). Utilizing this finding, we develop several other combinatorial variants of MI and increase the effectiveness of the oracle prompt selection method from 87.79% to 94.98%, measured as the ratio of the performance of the selected prompt to that of the optimal oracle prompt. Furthermore, considering that all the methods rely on the output probability distribution of the model that might be biased, we propose a novel calibration method called Calibration by Marginalization (CBM) that is orthogonal to the existing methods and helps increase the prompt selection effectiveness of the best method to 96.85%, achieving 99.44% of the oracle prompt F1 without calibration."
                    },
                    {
                        "title": "Instruction Matters: A Simple yet Effective Task Selection for Optimized Instruction Tuning of Specific Tasks",
                        "abstract": "Instruction tuning has been proven effective in enhancing zero-shot generalization across various tasks and in improving the performance of specific tasks. For task-specific improvements, strategically selecting and training on related tasks that provide meaningful supervision is crucial, as this approach enhances efficiency and prevents performance degradation from learning irrelevant tasks. In this light, we introduce a simple yet effective task selection method that leverages instruction information alone to identify relevant tasks, optimizing instruction tuning for specific tasks. Our method is significantly more efficient than traditional approaches, which require complex measurements of pairwise transferability between tasks or the creation of data samples for the target task. Additionally, by aligning the model with the unique instructional template style of the meta-dataset, we enhance its ability to granularly discern relevant tasks, leading to improved overall performance. Experimental results demonstrate that training on a small set of tasks, chosen solely based on the instructions, results in substantial improvements in performance on benchmarks such as P3, Big-Bench, NIV2, and Big-Bench Hard. Significantly, these improvements surpass those achieved by prior task selection methods, highlighting the superiority of our approach."
                    },
                    {
                        "title": "FLASK: Fine-grained Language Model Evaluation based on Alignment Skill Sets",
                        "abstract": "Evaluation of Large Language Models (LLMs) is challenging because instruction-following necessitates alignment with human values and the required set of skills varies depending on the instruction. However, previous studies have mainly focused on coarse-grained evaluation (i.e. overall preference-based evaluation), which limits interpretability since it does not consider the nature of user instructions that require instance-wise skill composition. In this paper, we introduce FLASK (Fine-grained Language Model Evaluation based on Alignment Skill Sets), a fine-grained evaluation protocol for both human-based and model-based evaluation which decomposes coarse-level scoring to a skill set-level scoring for each instruction. We experimentally observe that the fine-graininess of evaluation is crucial for attaining a holistic view of model performance and increasing the reliability of the evaluation. Using FLASK, we compare multiple open-source and proprietary LLMs and observe a high correlation between model-based and human-based evaluations. We publicly release the evaluation data and code implementation at https://github.com/kaistAI/FLASK."
                    },
                    {
                        "title": "Carpe Diem: On the Evaluation of World Knowledge in Lifelong Language Models",
                        "abstract": "The dynamic nature of knowledge in an ever-changing world presents challenges for language models trained on static data; the model in the real world often requires not only acquiring new knowledge but also overwriting outdated information into updated ones. To study the ability of language models for these time-dependent dynamics in human language, we introduce a novel task, EvolvingQA, a temporally evolving question-answering benchmark designed for training and evaluating LMs on an evolving Wikipedia database. The construction of EvolvingQA is automated with our pipeline using large language models. We uncover that existing continual learning baselines suffer from updating and removing outdated knowledge. Our analysis suggests that models fail to rectify knowledge due to small weight gradients. In addition, we elucidate that language models particularly struggle to reflect the change of numerical or temporal information. Our work aims to model the dynamic nature of real-world information, suggesting faithful evaluations of the evolution-adaptability of language models."
                    },
                    {
                        "title": "Towards Continual Knowledge Learning of Language Models",
                        "abstract": "Large Language Models (LMs) are known to encode world knowledge in their parameters as they pretrain on a vast amount of web corpus, which is often utilized for performing knowledge-dependent downstream tasks such as question answering, fact-checking, and open dialogue. In real-world scenarios, the world knowledge stored in the LMs can quickly become outdated as the world changes, but it is non-trivial to avoid catastrophic forgetting and reliably acquire new knowledge while preserving invariant knowledge. To push the community towards better maintenance of ever-changing LMs, we formulate a new continual learning (CL) problem called Continual Knowledge Learning (CKL). We construct a new benchmark and metric to quantify the retention of time-invariant world knowledge, the update of outdated knowledge, and the acquisition of new knowledge. We adopt applicable recent methods from literature to create several strong baselines. Through extensive experiments, we find that CKL exhibits unique challenges that are not addressed in previous CL setups, where parameter expansion is necessary to reliably retain and learn knowledge simultaneously. By highlighting the critical causes of knowledge forgetting, we show that CKL is a challenging and important problem that helps us better understand and train ever-changing LMs. The benchmark datasets, evaluation script, and baseline code to reproduce our results are available at https://github.com/joeljang/continual-knowledge-learning."
                    },
                    {
                        "title": "TemporalWiki: A Lifelong Benchmark for Training and Evaluating Ever-Evolving Language Models",
                        "abstract": "Language Models (LMs) become outdated as the world changes; they often fail to perform tasks requiring recent factual information which was absent or different during training, a phenomenon called temporal misalignment. This is especially a challenging problem because the research community still lacks a coherent dataset for assessing the adaptability of LMs to frequently-updated knowledge corpus such as Wikipedia. To this end, we introduce TemporalWiki, a lifelong benchmark for ever-evolving LMs that utilizes the difference between consecutive snapshots of English Wikipedia and English Wikidata for training and evaluation, respectively. The benchmark hence allows researchers to periodically track an LM's ability to retain previous knowledge and acquire updated/new knowledge at each point in time. We also find that training an LM on the diff data through continual learning methods achieves similar or better perplexity than on the entire snapshot in our benchmark with 12 times less computational cost, which verifies that factual knowledge in LMs can be safely updated with minimal training data via continual learning. The dataset and the code are available at https://github.com/joeljang/temporalwiki."
                    },
                    {
                        "title": "Exploring the Benefits of Training Expert Language Models over Instruction Tuning",
                        "abstract": "Recently, Language Models (LMs) instruction-tuned on multiple tasks, also known as multitask-prompted fine-tuning (MT), have shown the capability to generalize to unseen tasks. Previous work has shown that scaling the number of training tasks is the key component in making stronger MT LMs. In this work, we report an unexpected finding that an expert LM fine-tuned on just a single task can outperform an MT LM trained with 300+ different tasks on 11 different unseen datasets and on 13 datasets of the BIG-bench benchmark by a mean accuracy of 3.20% and 1.29%, respectively. This finding casts doubt on the previously held belief that simply scaling the number of tasks makes stronger MT LMs. Leveraging this finding, we further show that this distributed approach of training a separate expert LM per training task instead of a single MT LM for zero-shot inference possesses many benefits including (1) avoiding negative task transfer that often occurs during instruction tuning, (2) being able to continually learn new tasks without having to re-train on previous tasks to avoid catastrophic forgetting, and (3) showing compositional capabilities when merging individual experts together. The code is available at https://github.com/joeljang/ELM."
                    },
                    {
                        "title": "The CoT Collection: Improving Zero-shot and Few-shot Learning of Language Models via Chain-of-Thought Fine-Tuning",
                        "abstract": "Language models (LMs) with less than 100B parameters are known to perform poorly on chain-of-thought (CoT) reasoning in contrast to large LMs when solving unseen tasks. In this work, we aim to equip smaller LMs with the step-by-step reasoning capability by instruction tuning with CoT rationales. In order to achieve this goal, we first introduce a new instruction-tuning dataset called the CoT Collection, which augments the existing Flan Collection (including only 9 CoT tasks) with additional 1.84 million rationales across 1,060 tasks. We show that CoT fine-tuning Flan-T5 (3B & 11B) with CoT Collection enables smaller LMs to have better CoT capabilities on unseen tasks. On the BIG-Bench-Hard (BBH) benchmark, we report an average improvement of +4.34% (Flan-T5 3B) and +2.60% (Flan-T5 11B), in terms of zero-shot task accuracy. Furthermore, we show that instruction tuning with CoT Collection allows LMs to possess stronger few-shot learning capabilities on 4 domain-specific tasks, resulting in an improvement of +2.24% (Flan-T5 3B) and +2.37% (Flan-T5 11B), even outperforming ChatGPT utilizing demonstrations until the max length by a +13.98% margin. Our code, the CoT Collection data, and model checkpoints are publicly available."
                    }
                ]
            },
            "dea79145-9103-4582-b5aa-b9b11c3d04d2": {
                "pk": "dea79145-9103-4582-b5aa-b9b11c3d04d2",
                "name": "Sohee Yang",
                "collaborators": [
                    "Minjoon Seo",
                    "Joel Jang",
                    "Hyunji Lee",
                    "Seonghyeon Ye",
                    "Hanseok Oh",
                    "Mor Geva",
                    "Sang-Woo Lee",
                    "Joongbo Shin",
                    "Janghoon Han",
                    "Gyeonghun Kim"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Information Retrieval",
                    "Knowledge Representation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Is Retriever Merely an Approximator of Reader?",
                        "abstract": "The state of the art in open-domain question answering (QA) relies on an efficient retriever that drastically reduces the search space for the expensive reader. A rather overlooked question in the community is the relationship between the retriever and the reader, and in particular, if the whole purpose of the retriever is just a fast approximation for the reader. Our empirical evidence indicates that the answer is no, and that the reader and the retriever are complementary to each other even in terms of accuracy only. We make a careful conjecture that the architectural constraint of the retriever, which has been originally intended for enabling approximate search, seems to also make the model more robust in large-scale search. We then propose to distill the reader into the retriever so that the retriever absorbs the strength of the reader while keeping its own benefit. Experimental results show that our method can enhance the document recall rate as well as the end-to-end QA accuracy of off-the-shelf retrievers in open-domain QA tasks."
                    },
                    {
                        "title": "Designing a Minimal Retrieve-and-Read System for Open-Domain Question Answering",
                        "abstract": "In open-domain question answering (QA), retrieve-and-read mechanism has the inherent benefit of interpretability and the easiness of adding, removing, or editing knowledge compared to the parametric approaches of closed-book QA models. However, it is also known to suffer from its large storage footprint due to its document corpus and index. Here, we discuss several orthogonal strategies to drastically reduce the footprint of a retrieve-and-read open-domain QA system by up to 160x. Our results indicate that retrieve-and-read can be a viable option even in a highly constrained serving environment such as edge devices, as we show that it can achieve better accuracy than a purely parametric model with comparable docker-level system size."
                    },
                    {
                        "title": "Generative Multi-hop Retrieval",
                        "abstract": "A common practice for text retrieval is to use an encoder to map the documents and the query to a common vector space and perform a nearest neighbor search (NNS); multi-hop retrieval also often adopts the same paradigm, usually with a modification of iteratively reformulating the query vector so that it can retrieve different documents at each hop. However, such a bi-encoder approach has limitations in multi-hop settings; (1) the reformulated query gets longer as the number of hops increases, which further tightens the embedding bottleneck of the query vector, and (2) it is prone to error propagation. In this paper, we focus on alleviating these limitations in multi-hop settings by formulating the problem in a fully generative way. We propose an encoder-decoder model that performs multi-hop retrieval by simply generating the entire text sequences of the retrieval targets, which means the query and the documents interact in the language model's parametric space rather than L2 or inner product space as in the bi-encoder approach. Our approach, Generative Multi-hop Retrieval(GMR), consistently achieves comparable or higher performance than bi-encoder models in five datasets while demonstrating superior GPU memory and storage footprint."
                    },
                    {
                        "title": "Do Large Language Models Latently Perform Multi-Hop Reasoning?",
                        "abstract": "We study whether Large Language Models (LLMs) latently perform multi-hop reasoning with complex prompts such as \"The mother of the singer of 'Superstition' is\". We look for evidence of a latent reasoning pathway where an LLM (1) latently identifies \"the singer of 'Superstition'\" as Stevie Wonder, the bridge entity, and (2) uses its knowledge of Stevie Wonder's mother to complete the prompt. We analyze these two hops individually and consider their co-occurrence as indicative of latent multi-hop reasoning. For the first hop, we test if changing the prompt to indirectly mention the bridge entity instead of any other entity increases the LLM's internal recall of the bridge entity. For the second hop, we test if increasing this recall causes the LLM to better utilize what it knows about the bridge entity. We find strong evidence of latent multi-hop reasoning for the prompts of certain relation types, with the reasoning pathway used in more than 80% of the prompts. However, the utilization is highly contextual, varying across different types of prompts. Also, on average, the evidence for the second hop and the full multi-hop traversal is rather moderate and only substantial for the first hop. Moreover, we find a clear scaling trend with increasing model size for the first hop of reasoning but not for the second hop. Our experimental findings suggest potential challenges and opportunities for future development and applications of LLMs."
                    },
                    {
                        "title": "Spatial Dependency Parsing for Semi-Structured Document Information Extraction",
                        "abstract": "Information Extraction (IE) for semi-structured document images is often approached as a sequence tagging problem by classifying each recognized input token into one of the IOB (Inside, Outside, and Beginning) categories. However, such problem setup has two inherent limitations that (1) it cannot easily handle complex spatial relationships and (2) it is not suitable for highly structured information, which are nevertheless frequently observed in real-world document images. To tackle these issues, we first formulate the IE task as spatial dependency parsing problem that focuses on the relationship among text tokens in the documents. Under this setup, we then propose SPADE (SPAtial DEpendency parser) that models highly complex spatial relationships and an arbitrary number of information layers in the documents in an end-to-end manner. We evaluate it on various kinds of documents such as receipts, name cards, forms, and invoices, and show that it achieves a similar or better performance compared to strong baselines including BERT-based IOB taggger."
                    },
                    {
                        "title": "Efficient Dialogue State Tracking by Selectively Overwriting Memory",
                        "abstract": "Recent works in dialogue state tracking (DST) focus on an open vocabulary-based setting to resolve scalability and generalization issues of the predefined ontology-based approaches. However, they are inefficient in that they predict the dialogue state at every turn from scratch. Here, we consider dialogue state as an explicit fixed-sized memory and propose a selectively overwriting mechanism for more efficient DST. This mechanism consists of two steps: (1) predicting state operation on each of the memory slots, and (2) overwriting the memory with new values, of which only a few are generated according to the predicted state operations. Our method decomposes DST into two sub-tasks and guides the decoder to focus only on one of the tasks, thus reducing the burden of the decoder. This enhances the effectiveness of training and DST performance. Our SOM-DST (Selectively Overwriting Memory for Dialogue State Tracking) model achieves state-of-the-art joint goal accuracy with 51.72% in MultiWOZ 2.0 and 53.01% in MultiWOZ 2.1 in an open vocabulary-based DST setting. In addition, we analyze the accuracy gaps between the current and the ground truth-given situations and suggest that it is a promising direction to improve state operation prediction to boost the DST performance."
                    },
                    {
                        "title": "Hopping Too Late: Exploring the Limitations of Large Language Models on Multi-Hop Queries",
                        "abstract": "Large language models (LLMs) can solve complex multi-step problems, but little is known about how these computations are implemented internally. Motivated by this, we study how LLMs answer multi-hop queries such as \"The spouse of the performer of Imagine is\". These queries require two information extraction steps: a latent one for resolving the first hop (\"the performer of Imagine\") into the bridge entity (John Lennon), and another for resolving the second hop (\"the spouse of John Lennon\") into the target entity (Yoko Ono). Understanding how the latent step is computed internally is key to understanding the overall computation. By carefully analyzing the internal computations of transformer-based LLMs, we discover that the bridge entity is resolved in the early layers of the model. Then, only after this resolution, the two-hop query is solved in the later layers. Because the second hop commences in later layers, there could be cases where these layers no longer encode the necessary knowledge for correctly predicting the answer. Motivated by this, we propose a novel \"back-patching\" analysis method whereby a hidden representation from a later layer is patched back to an earlier layer. We find that in up to 66% of previously incorrect cases there exists a back-patch that results in the correct generation of the answer, showing that the later layers indeed sometimes lack the needed functionality. Overall, our methods and findings open further opportunities for understanding and improving latent reasoning in transformer-based LLMs."
                    },
                    {
                        "title": "Improving Probability-based Prompt Selection Through Unified Evaluation and Analysis",
                        "abstract": "Previous works in prompt engineering for large language models have introduced different gradient-free probability-based prompt selection methods that aim to choose the optimal prompt among the candidates for a given task but have failed to provide a comprehensive and fair comparison between each other. In this paper, we propose a unified framework to interpret and evaluate the existing probability-based prompt selection methods by performing extensive experiments on 13 common and diverse NLP tasks. We find that each of the existing methods can be interpreted as some variant of the method that maximizes mutual information between the input and the predicted output (MI). Utilizing this finding, we develop several other combinatorial variants of MI and increase the effectiveness of the oracle prompt selection method from 87.79% to 94.98%, measured as the ratio of the performance of the selected prompt to that of the optimal oracle prompt. Furthermore, considering that all the methods rely on the output probability distribution of the model that might be biased, we propose a novel calibration method called Calibration by Marginalization (CBM) that is orthogonal to the existing methods and helps increase the prompt selection effectiveness of the best method to 96.85%, achieving 99.44% of the oracle prompt F1 without calibration."
                    },
                    {
                        "title": "Investigating the Effectiveness of Task-Agnostic Prefix Prompt for Instruction Following",
                        "abstract": "In this paper, we present our finding that prepending a Task-Agnostic Prefix Prompt (TAPP) to the input improves the instruction-following ability of various Large Language Models (LLMs) during inference. TAPP is different from canonical prompts for LLMs in that it is a fixed prompt prepended to the beginning of every input regardless of the target task for zero-shot generalization. We observe that both base LLMs (i.e. not fine-tuned to follow instructions) and instruction-tuned models benefit from TAPP, resulting in 34.58% and 12.26% improvement on average, respectively. This implies that the instruction-following ability of LLMs can be improved during inference time with a fixed prompt constructed with simple heuristics. We hypothesize that TAPP assists language models to better estimate the output distribution by focusing more on the instruction of the target task during inference. In other words, such ability does not seem to be sufficiently activated in not only base LLMs but also many instruction-fine-tuned LLMs. All experiments are reproducible from https://github.com/seonghyeonye/TAPP."
                    },
                    {
                        "title": "Exploring the Practicality of Generative Retrieval on Dynamic Corpora",
                        "abstract": "Benchmarking the performance of information retrieval (IR) is mostly conducted with a fixed set of documents (static corpora). However, in realistic scenarios, this is rarely the case and the documents to be retrieved are constantly updated and added. In this paper, we focus on Generative Retrievals (GR), which apply autoregressive language models to IR problems, and explore their adaptability and robustness in dynamic scenarios. We also conduct an extensive evaluation of computational and memory efficiency, crucial factors for real-world deployment of IR systems handling vast and ever-changing document collections. Our results on the StreamingQA benchmark demonstrate that GR is more adaptable to evolving knowledge (4-11%), robust in learning knowledge with temporal information, and efficient in terms of inference FLOPs (x2), indexing time (x6), and storage footprint (x4) compared to Dual Encoders (DE), which are commonly used in retrieval systems. Our paper highlights the potential of GR for future use in practical IR systems within dynamic environments."
                    },
                    {
                        "title": "Large-Scale Answerer in Questioner's Mind for Visual Dialog Question Generation",
                        "abstract": "Answerer in Questioner's Mind (AQM) is an information-theoretic framework that has been recently proposed for task-oriented dialog systems. AQM benefits from asking a question that would maximize the information gain when it is asked. However, due to its intrinsic nature of explicitly calculating the information gain, AQM has a limitation when the solution space is very large. To address this, we propose AQM+ that can deal with a large-scale problem and ask a question that is more coherent to the current context of the dialog. We evaluate our method on GuessWhich, a challenging task-oriented visual dialog problem, where the number of candidate classes is near 10K. Our experimental results and ablation studies show that AQM+ outperforms the state-of-the-art models by a remarkable margin with a reasonable approximation. In particular, the proposed AQM+ reduces more than 60% of error as the dialog proceeds, while the comparative algorithms diminish the error by less than 6%. Based on our results, we argue that AQM+ is a general task-oriented dialog algorithm that can be applied for non-yes-or-no responses."
                    },
                    {
                        "title": "Knowledge Unlearning for Mitigating Privacy Risks in Language Models",
                        "abstract": "Pretrained Language Models (LMs) memorize a vast amount of knowledge during initial pretraining, including information that may violate the privacy of personal lives and identities. Previous work addressing privacy issues for language models has mostly focused on data preprocessing and differential privacy methods, both requiring re-training the underlying LM. We propose knowledge unlearning as an alternative method to reduce privacy risks for LMs post hoc. We show that simply performing gradient ascent on target token sequences is effective at forgetting them with little to no degradation of general language modeling performances for larger LMs; it sometimes even substantially improves the underlying LM with just a few iterations. We also find that sequential unlearning is better than trying to unlearn all the data at once and that unlearning is highly dependent on which kind of data (domain) is forgotten. By showing comparisons with a previous data preprocessing method and a decoding method known to mitigate privacy risks for LMs, we show that unlearning can give a stronger empirical privacy guarantee in scenarios where the data vulnerable to extraction attacks are known a priori while being much more efficient and robust. We release the code and dataset needed to replicate our results at https://github.com/joeljang/knowledge-unlearning."
                    },
                    {
                        "title": "Nonparametric Decoding for Generative Retrieval",
                        "abstract": "The generative retrieval model depends solely on the information encoded in its model parameters without external memory, its information capacity is limited and fixed. To overcome the limitation, we propose Nonparametric Decoding (Np Decoding) which can be applied to existing generative retrieval models. Np Decoding uses nonparametric contextualized vocab embeddings (external memory) rather than vanilla vocab embeddings as decoder vocab embeddings. By leveraging the contextualized vocab embeddings, the generative retrieval model is able to utilize both the parametric and nonparametric space. Evaluation over 9 datasets (8 single-hop and 1 multi-hop) in the document retrieval task shows that applying Np Decoding to generative retrieval models significantly improves the performance. We also show that Np Decoding is data- and parameter-efficient, and shows high performance in the zero-shot setting."
                    },
                    {
                        "title": "Towards Continual Knowledge Learning of Language Models",
                        "abstract": "Large Language Models (LMs) are known to encode world knowledge in their parameters as they pretrain on a vast amount of web corpus, which is often utilized for performing knowledge-dependent downstream tasks such as question answering, fact-checking, and open dialogue. In real-world scenarios, the world knowledge stored in the LMs can quickly become outdated as the world changes, but it is non-trivial to avoid catastrophic forgetting and reliably acquire new knowledge while preserving invariant knowledge. To push the community towards better maintenance of ever-changing LMs, we formulate a new continual learning (CL) problem called Continual Knowledge Learning (CKL). We construct a new benchmark and metric to quantify the retention of time-invariant world knowledge, the update of outdated knowledge, and the acquisition of new knowledge. We adopt applicable recent methods from literature to create several strong baselines. Through extensive experiments, we find that CKL exhibits unique challenges that are not addressed in previous CL setups, where parameter expansion is necessary to reliably retain and learn knowledge simultaneously. By highlighting the critical causes of knowledge forgetting, we show that CKL is a challenging and important problem that helps us better understand and train ever-changing LMs. The benchmark datasets, evaluation script, and baseline code to reproduce our results are available at https://github.com/joeljang/continual-knowledge-learning."
                    },
                    {
                        "title": "TemporalWiki: A Lifelong Benchmark for Training and Evaluating Ever-Evolving Language Models",
                        "abstract": "Language Models (LMs) become outdated as the world changes; they often fail to perform tasks requiring recent factual information which was absent or different during training, a phenomenon called temporal misalignment. This is especially a challenging problem because the research community still lacks a coherent dataset for assessing the adaptability of LMs to frequently-updated knowledge corpus such as Wikipedia. To this end, we introduce TemporalWiki, a lifelong benchmark for ever-evolving LMs that utilizes the difference between consecutive snapshots of English Wikipedia and English Wikidata for training and evaluation, respectively. The benchmark hence allows researchers to periodically track an LM's ability to retain previous knowledge and acquire updated/new knowledge at each point in time. We also find that training an LM on the diff data through continual learning methods achieves similar or better perplexity than on the entire snapshot in our benchmark with 12 times less computational cost, which verifies that factual knowledge in LMs can be safely updated with minimal training data via continual learning. The dataset and the code are available at https://github.com/joeljang/temporalwiki."
                    }
                ]
            },
            "e3c0f594-4aa3-4d95-99d2-3dee827da6c8": {
                "pk": "e3c0f594-4aa3-4d95-99d2-3dee827da6c8",
                "name": "Youngkyung Seo",
                "collaborators": [
                    "Yoonseok Heo",
                    "Jun-Seok Koh",
                    "Du-Seong Chang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Dialog Systems",
                    "Memory Management"
                ],
                "publications": [
                    {
                        "title": "Efficient and Accurate Memorable Conversation Model using DPO based on sLLM",
                        "abstract": "In multi-session dialog system, it is essential to continuously update the memory as the session progresses. Simply accumulating memory can make it difficult to focus on the content of the conversation for inference due to the limited input sentence size. Therefore, efficient and accurate conversation model that is capable of managing memory to reflect the conversation history continuously is necessary. This paper presents a conversation model that efficiently manages memory as sessions progress and incorporates this into the model to reflect the conversation history accurately with 3 methodologies: SFT, DPO and DPO with SFT model. Our model using DPO algorithm shows an improvement about 0.0591 of BERTScore in memory accuracy, and the rate of responses reflecting the memory increased as well. Also, response generation performance enhanced about 4.292 in fluency, 3.935 in coherence, and 2.896 in consistency. This paper describes a training method that yields better performance than models with more than twice the parameter size, even when the model size is smaller. Thus, our model demonstrates efficiency not only in terms of accuracy but also in resource utilization."
                    }
                ]
            },
            "eeb918db-10ed-43b5-af4d-f25bc4e08788": {
                "pk": "eeb918db-10ed-43b5-af4d-f25bc4e08788",
                "name": "Du-Seong Chang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "1b62902c-7714-4ff6-8a50-dc3543222901": {
                "pk": "1b62902c-7714-4ff6-8a50-dc3543222901",
                "name": "Minjoon Seo",
                "collaborators": [
                    "Hannaneh Hajishirzi",
                    "Sewon Min",
                    "Ali Farhadi",
                    "Joel Jang",
                    "Wonseok Hwang",
                    "Seunghyun Park",
                    "Sohee Yang",
                    "Geewook Kim",
                    "Sungdong Kim",
                    "Jinyeong Yim"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Question Answering",
                    "Machine Learning",
                    "Semantic Parsing"
                ],
                "publications": [
                    {
                        "title": "Is Retriever Merely an Approximator of Reader?",
                        "abstract": "The state of the art in open-domain question answering (QA) relies on an efficient retriever that drastically reduces the search space for the expensive reader. A rather overlooked question in the community is the relationship between the retriever and the reader, and in particular, if the whole purpose of the retriever is just a fast approximation for the reader. Our empirical evidence indicates that the answer is no, and that the reader and the retriever are complementary to each other even in terms of accuracy only. We make a careful conjecture that the architectural constraint of the retriever, which has been originally intended for enabling approximate search, seems to also make the model more robust in large-scale search. We then propose to distill the reader into the retriever so that the retriever absorbs the strength of the reader while keeping its own benefit. Experimental results show that our method can enhance the document recall rate as well as the end-to-end QA accuracy of off-the-shelf retrievers in open-domain QA tasks."
                    },
                    {
                        "title": "Designing a Minimal Retrieve-and-Read System for Open-Domain Question Answering",
                        "abstract": "In open-domain question answering (QA), retrieve-and-read mechanism has the inherent benefit of interpretability and the easiness of adding, removing, or editing knowledge compared to the parametric approaches of closed-book QA models. However, it is also known to suffer from its large storage footprint due to its document corpus and index. Here, we discuss several orthogonal strategies to drastically reduce the footprint of a retrieve-and-read open-domain QA system by up to 160x. Our results indicate that retrieve-and-read can be a viable option even in a highly constrained serving environment such as edge devices, as we show that it can achieve better accuracy than a purely parametric model with comparable docker-level system size."
                    },
                    {
                        "title": "On Efficient Language and Vision Assistants for Visually-Situated Natural Language Understanding: What Matters in Reading and Reasoning",
                        "abstract": "Recent advancements in language and vision assistants have showcased impressive capabilities but suffer from a lack of transparency, limiting broader research and reproducibility. While open-source models handle general image tasks effectively, they face challenges with the high computational demands of complex visually-situated text understanding. Such tasks often require increased token inputs and large vision modules to harness high-resolution information. Striking a balance between model size and data importance remains an open question. This study aims to redefine the design of vision-language models by identifying key components and creating efficient models with constrained inference costs. By strategically formulating datasets, optimizing vision modules, and enhancing supervision techniques, we achieve significant improvements in inference throughput while maintaining high performance. Extensive experiments across models ranging from 160M to 13B parameters offer insights into model optimization. We will fully open-source our codebase, models, and datasets at https://github.com/naver-ai/elva."
                    },
                    {
                        "title": "Rethinking the Role of Proxy Rewards in Language Model Alignment",
                        "abstract": "Learning from human feedback via proxy reward modeling has been studied to align Large Language Models (LLMs) with human values. However, achieving reliable training through that proxy reward model (RM) is not a trivial problem, and its behavior remained as a black-box. In this paper, we study the role of proxy rewards in the LLM alignment via `reverse reward engineering' by composing interpretable features as a white-box reward function. We aim to replicate the ground truth (gold) reward signal by achieving a monotonic relationship between the proxy and gold reward signals after training the model using the proxy reward in reinforcement learning (RL). Our findings indicate that successfully emulating the gold reward requires generating responses that are relevant with enough length to open-ended questions, while also ensuring response consistency in closed-ended questions. Furthermore, resulting models optimizing our devised white-box reward show competitive performances with strong open-source RMs in alignment benchmarks. We highlight its potential usage as a simple but strong reward baseline for the LLM alignment, not requiring explicit human feedback dataset and RM training. Our code is available at https://github.com/naver-ai/rethinking-proxy-reward."
                    },
                    {
                        "title": "Query-Reduction Networks for Question Answering",
                        "abstract": "In this paper, we study the problem of question answering when reasoning over multiple facts is required. We propose Query-Reduction Network (QRN), a variant of Recurrent Neural Network (RNN) that effectively handles both short-term (local) and long-term (global) sequential dependencies to reason over multiple facts. QRN considers the context sentences as a sequence of state-changing triggers, and reduces the original query to a more informed query as it observes each trigger (context sentence) through time. Our experiments show that QRN produces the state-of-the-art results in bAbI QA and dialog tasks, and in a real goal-oriented dialog dataset. In addition, QRN formulation allows parallelization on RNN's time axis, saving an order of magnitude in time complexity for training and inference."
                    },
                    {
                        "title": "Bidirectional Attention Flow for Machine Comprehension",
                        "abstract": "Machine comprehension (MC), answering a query about a given context paragraph, requires modeling complex interactions between the context and the query. Recently, attention mechanisms have been successfully extended to MC. Typically these methods use attention to focus on a small portion of the context and summarize it with a fixed-size vector, couple attentions temporally, and/or often form a uni-directional attention. In this paper we introduce the Bi-Directional Attention Flow (BIDAF) network, a multi-stage hierarchical process that represents the context at different levels of granularity and uses bi-directional attention flow mechanism to obtain a query-aware context representation without early summarization. Our experimental evaluations show that our model achieves the state-of-the-art results in Stanford Question Answering Dataset (SQuAD) and CNN/DailyMail cloze test."
                    },
                    {
                        "title": "Neural Speed Reading via Skim-RNN",
                        "abstract": "Inspired by the principles of speed reading, we introduce Skim-RNN, a recurrent neural network (RNN) that dynamically decides to update only a small fraction of the hidden state for relatively unimportant input tokens. Skim-RNN gives computational advantage over an RNN that always updates the entire hidden state. Skim-RNN uses the same input and output interfaces as a standard RNN and can be easily used instead of RNNs in existing models. In our experiments, we show that Skim-RNN can achieve significantly reduced computational cost without losing accuracy compared to standard RNNs across five different natural language tasks. In addition, we demonstrate that the trade-off between accuracy and speed of Skim-RNN can be dynamically controlled during inference time in a stable manner. Our analysis also shows that Skim-RNN running on a single CPU offers lower latency compared to standard RNNs on GPUs."
                    },
                    {
                        "title": "Contextualized Sparse Representations for Real-Time Open-Domain Question Answering",
                        "abstract": "Open-domain question answering can be formulated as a phrase retrieval problem, in which we can expect huge scalability and speed benefit but often suffer from low accuracy due to the limitation of existing phrase representation models. In this paper, we aim to improve the quality of each phrase embedding by augmenting it with a contextualized sparse representation (Sparc). Unlike previous sparse vectors that are term-frequency-based (e.g., tf-idf) or directly learned (only few thousand dimensions), we leverage rectified self-attention to indirectly learn sparse vectors in n-gram vocabulary space. By augmenting the previous phrase retrieval model (Seo et al., 2019) with Sparc, we show 4%+ improvement in CuratedTREC and SQuAD-Open. Our CuratedTREC score is even better than the best known retrieve & read model with at least 45x faster inference speed."
                    },
                    {
                        "title": "Gradient Ascent Post-training Enhances Language Model Generalization",
                        "abstract": "In this work, we empirically show that updating pretrained LMs (350M, 1.3B, 2.7B) with just a few steps of Gradient Ascent Post-training (GAP) on random, unlabeled text corpora enhances its zero-shot generalization capabilities across diverse NLP tasks. Specifically, we show that GAP can allow LMs to become comparable to 2-3x times larger LMs across 12 different NLP tasks. We also show that applying GAP on out-of-distribution corpora leads to the most reliable performance improvements. Our findings indicate that GAP can be a promising method for improving the generalization capability of LMs without any task-specific fine-tuning."
                    },
                    {
                        "title": "Question Answering through Transfer Learning from Large Fine-grained Supervision Data",
                        "abstract": "We show that the task of question answering (QA) can significantly benefit from the transfer learning of models trained on a different large, fine-grained QA dataset. We achieve the state of the art in two well-studied QA datasets, WikiQA and SemEval-2016 (Task 3A), through a basic transfer learning technique from SQuAD. For WikiQA, our model outperforms the previous best model by more than 8%. We demonstrate that finer supervision provides better guidance for learning lexical and syntactic information than coarser supervision, through quantitative results and visual analysis. We also show that a similar transfer learning procedure achieves the state of the art on an entailment task."
                    },
                    {
                        "title": "A Comprehensive Exploration on WikiSQL with Table-Aware Word Contextualization",
                        "abstract": "We present SQLova, the first Natural-language-to-SQL (NL2SQL) model to achieve human performance in WikiSQL dataset. We revisit and discuss diverse popular methods in NL2SQL literature, take a full advantage of BERT {Devlin et al., 2018) through an effective table contextualization method, and coherently combine them, outperforming the previous state of the art by 8.2% and 2.5% in logical form and execution accuracy, respectively. We particularly note that BERT with a seq2seq decoder leads to a poor performance in the task, indicating the importance of a careful design when using such large pretrained models. We also provide a comprehensive analysis on the dataset and our model, which can be helpful for designing future NL2SQL datsets and models. We especially show that our model's performance is near the upper bound in WikiSQL, where we observe that a large portion of the evaluation errors are due to wrong annotations, and our model is already exceeding human performance by 1.3% in execution accuracy."
                    },
                    {
                        "title": "ViSeRet: A simple yet effective approach to moment retrieval via fine-grained video segmentation",
                        "abstract": "Video-text retrieval has many real-world applications such as media analytics, surveillance, and robotics. This paper presents the 1st place solution to the video retrieval track of the ICCV VALUE Challenge 2021. We present a simple yet effective approach to jointly tackle two video-text retrieval tasks (video retrieval and video corpus moment retrieval) by leveraging the model trained only on the video retrieval task. In addition, we create an ensemble model that achieves the new state-of-the-art performance on all four datasets (TVr, How2r, YouCook2r, and VATEXr) presented in the VALUE Challenge."
                    },
                    {
                        "title": "Zero-Shot Relation Extraction via Reading Comprehension",
                        "abstract": "We show that relation extraction can be reduced to answering simple reading comprehension questions, by associating one or more natural-language questions with each relation slot. This reduction has several advantages: we can (1) learn relation-extraction models by extending recent neural reading-comprehension techniques, (2) build very large training sets for those models by combining relation-specific crowd-sourced questions with distant supervision, and even (3) do zero-shot learning by extracting new relation types that are only specified at test-time, for which we have no labeled training examples. Experiments on a Wikipedia slot-filling task demonstrate that the approach can generalize to new questions for known relation types with high accuracy, and that zero-shot generalization to unseen relation types is possible, at lower accuracy levels, setting the bar for future work on this task."
                    },
                    {
                        "title": "Mixture Content Selection for Diverse Sequence Generation",
                        "abstract": "Generating diverse sequences is important in many NLP applications such as question generation or summarization that exhibit semantically one-to-many relationships between source and the target sequences. We present a method to explicitly separate diversification from generation using a general plug-and-play module (called SELECTOR) that wraps around and guides an existing encoder-decoder model. The diversification stage uses a mixture of experts to sample different binary masks on the source sequence for diverse content selection. The generation stage uses a standard encoder-decoder model given each selected content from the source sequence. Due to the non-differentiable nature of discrete sampling and the lack of ground truth labels for binary mask, we leverage a proxy for ground truth mask and adopt stochastic hard-EM for training. In question generation (SQuAD) and abstractive summarization (CNN-DM), our method demonstrates significant improvements in accuracy, diversity and training efficiency, including state-of-the-art top-1 accuracy in both datasets, 6% gain in top-5 accuracy, and 3.7 times faster training over a state of the art model. Our code is publicly available at https://github.com/clovaai/FocusSeq2Seq."
                    },
                    {
                        "title": "Semi-Structured Query Grounding for Document-Oriented Databases with Deep Retrieval and Its Application to Receipt and POI Matching",
                        "abstract": "Semi-structured query systems for document-oriented databases have many real applications. One particular application that we are interested in is matching each financial receipt image with its corresponding place of interest (POI, e.g., restaurant) in the nationwide database. The problem is especially challenging in the real production environment where many similar or incomplete entries exist in the database and queries are noisy (e.g., errors in optical character recognition). In this work, we aim to address practical challenges when using embedding-based retrieval for the query grounding problem in semi-structured data. Leveraging recent advancements in deep language encoding for retrieval, we conduct extensive experiments to find the most effective combination of modules for the embedding and retrieval of both query and database entries without any manually engineered component. The proposed model significantly outperforms the conventional manual pattern-based model while requiring much less development and maintenance cost. We also discuss some core observations in our experiments, which could be helpful for practitioners working on a similar problem in other domains."
                    },
                    {
                        "title": "Prompt Injection: Parameterization of Fixed Inputs",
                        "abstract": "Recent works have shown that attaching prompts to the input is effective at conditioning Language Models (LM) to perform specific tasks. However, prompts are always included in the input text during inference, thus incurring substantial computational and memory overhead. Also, there is currently no straightforward method of utilizing prompts that are longer than the maximum input length of the LMs without incurring additional costs during inference. We propose Prompt Injection (PI), a novel formulation of injecting the prompt into the parameters of an LM to be an efficient alternative to attaching fixed prompts to the input. We show that in scenarios with long fixed prompts, PI can be up to 280 times more efficient in terms of total FLOPs than previous approaches. We further explore methodologies for PI and show promising results in persona-dependent conversation, semantic parsing, and zero-shot learning with task instructions. Through these explorations, we show that PI can be a promising direction for conditioning language models, especially in scenarios with long and fixed prompts."
                    },
                    {
                        "title": "Can Large Language Models Truly Understand Prompts? A Case Study with Negated Prompts",
                        "abstract": "Previous work has shown that there exists a scaling law between the size of Language Models (LMs) and their zero-shot performance on different downstream NLP tasks. In this work, we show that this phenomenon does not hold when evaluating large LMs on tasks with negated prompts, but instead shows an inverse scaling law. We evaluate 9 different tasks with negated prompts on (1) pretrained LMs (OPT & GPT-3) of varying sizes (125M - 175B), (2) LMs further pretrained to generalize to novel prompts (InstructGPT), (3) LMs provided with few-shot examples, and (4) LMs fine-tuned specifically on negated prompts; all LM types perform worse on negated prompts as they scale and show a huge performance gap between the human performance when comparing the average score on both original and negated prompts. By highlighting a critical limitation of existing LMs and methods, we urge the community to develop new approaches of developing LMs that actually follow the given instructions. We provide the code and the datasets to explore negated prompts at https://github.com/joeljang/negated-prompts-for-llms"
                    },
                    {
                        "title": "A Bayesian Approach To Analysing Training Data Attribution In Deep Learning",
                        "abstract": "Training data attribution (TDA) techniques find influential training data for the model's prediction on the test data of interest. They approximate the impact of down- or up-weighting a particular training sample. While conceptually useful, they are hardly applicable to deep models in practice, particularly because of their sensitivity to different model initialisation. In this paper, we introduce a Bayesian perspective on the TDA task, where the learned model is treated as a Bayesian posterior and the TDA estimates as random variables. From this novel viewpoint, we observe that the influence of an individual training sample is often overshadowed by the noise stemming from model initialisation and SGD batch composition. Based on this observation, we argue that TDA can only be reliably used for explaining deep model predictions that are consistently influenced by certain training data, independent of other noise factors. Our experiments demonstrate the rarity of such noise-independent training-test data pairs but confirm their existence. We recommend that future researchers and practitioners trust TDA estimates only in such cases. Further, we find a disagreement between ground truth and estimated TDA distributions and encourage future work to study this gap. Code is provided at https://github.com/ElisaNguyen/bayesian-tda."
                    },
                    {
                        "title": "How language models extrapolate outside the training data: A case study in Textualized Gridworld",
                        "abstract": "Language models' ability to extrapolate learned behaviors to novel, more complex environments beyond their training scope is highly unknown. This study introduces a path planning task in a textualized Gridworld to probe language models' extrapolation capabilities. We show that conventional approaches, including next-token prediction and Chain of Thought (CoT) fine-tuning, fail to generalize in larger, unseen environments. Inspired by human cognition and dual-process theory, we propose language models should construct cognitive maps before interaction. Our research demonstrates that autoregressive generation of cognitive maps and planning sequences enhances planning capabilities in extrapolated environments. Unlike CoT, we find that cognitive maps cannot be obtained through simple prompting, necessitating additional training schemes for integration. Our findings in Gridworld offer insights into training language models with improved reasoning and adaptability, potentially advancing more human-like cognition and opening avenues for enhancing model generalization across diverse, complex tasks."
                    },
                    {
                        "title": "Syntactic Question Abstraction and Retrieval for Data-Scarce Semantic Parsing",
                        "abstract": "Deep learning approaches to semantic parsing require a large amount of labeled data, but annotating complex logical forms is costly. Here, we propose Syntactic Question Abstraction and Retrieval (SQAR), a method to build a neural semantic parser that translates a natural language (NL) query to a SQL logical form (LF) with less than 1,000 annotated examples. SQAR first retrieves a logical pattern from the train data by computing the similarity between NL queries and then grounds a lexical information on the retrieved pattern in order to generate the final LF. We validate SQAR by training models using various small subsets of WikiSQL train data achieving up to 4.9% higher LF accuracy compared to the previous state-of-the-art models on WikiSQL test set. We also show that by using query-similarity to retrieve logical pattern, SQAR can leverage a paraphrasing dataset achieving up to 5.9% higher LF accuracy compared to the case where SQAR is trained by using only WikiSQL data. In contrast to a simple pattern classification approach, SQAR can generate unseen logical patterns upon the addition of new examples without re-training the model. We also discuss an ideal way to create cost efficient and robust train datasets when the data distribution can be approximated under a data-hungry setting."
                    }
                ]
            }
        }
    },
    "2305.10790": {
        "paper_data": {
            "title": "Listen, Think, and Understand",
            "url": "http://arxiv.org/abs/2305.10790v3",
            "arxiv_id": "2305.10790",
            "authors": [
                "Yuan Gong",
                "Hongyin Luo",
                "Alexander H. Liu",
                "Leonid Karlinsky",
                "James Glass"
            ],
            "abstract": "The ability of artificial intelligence (AI) systems to perceive and comprehend audio signals is crucial for many applications. Although significant progress has been made in this area since the development of AudioSet, most existing models are designed to map audio inputs to pre-defined, discrete sound label sets. In contrast, humans possess the ability to not only classify sounds into general categories, but also to listen to the finer details of the sounds, explain the reason for the predictions, think about what the sound infers, and understand the scene and what action needs to be taken, if any. Such capabilities beyond perception are not yet present in existing audio models. On the other hand, modern large language models (LLMs) exhibit emerging reasoning ability but they lack audio perception capabilities. Therefore, we ask the question: can we build a model that has both audio perception and a reasoning ability?   In this paper, we propose a new audio foundation model, called LTU (Listen, Think, and Understand). To train LTU, we created a new OpenAQA-5M dataset consisting of 1.9 million closed-ended and 3.7 million open-ended, diverse (audio, question, answer) tuples, and have used an autoregressive training framework with a perception-to-understanding curriculum. LTU demonstrates strong performance and generalization ability on conventional audio tasks such as classification and captioning. More importantly, it exhibits emerging audio reasoning and comprehension abilities that are absent in existing audio models. To the best of our knowledge, LTU is one of the first multimodal large language models that focus on general audio (rather than just speech) understanding.",
            "introduction": "   1 Introduction  As we go about our daily lives, we are surrounded by complex mixtures of audio signals. Our cognitive abilities enable us not only to perceive and identify these sounds but also to comprehend their implicit meaning. For instance, upon hearing a clock chime six times, we probably assume it\u2019s 6 o\u2019clock; when hearing a train whistle we might infer the arrival or departure of a train; when encountering unfamiliar animal sounds in the wilderness, we possess the ability to assess potential danger by discerning certain audio characteristics. Additionally, we can also grasp the emotional atmosphere conveyed by audio cues. We anticipate that future artificial intelligence systems will have similar abilities to perceive and comprehend an auditory environment.    Thanks to the release of the large-scale audio corpus AudioSet\u00a0(Gemmeke et\u00a0al., 2017), significant advances have been achieved in general audio event recognition over the past five years. Notably, the mean Average Precision (mAP) for audio tagging has improved significantly, increasing from 31.4\u00a0(Gemmeke et\u00a0al., 2017) to 47.3\u00a0(Huang et\u00a0al., 2022), demonstrating the increasingly strong capabilities of deep learning models to perceive sounds. However, models trained with discrete sound label sets (e.g., the 527-category AudioSet), while being strong in perception, possess very limited reasoning and understanding capabilities. For example, the model may recognize a clock chime 6 times, but not know that it indicates a time of 6 o\u2019clock. On the other hand, we find that modern large language models (LLMs), such as GPT\u00a0(Brown et\u00a0al., 2020; OpenAI, 2023) and LLaMA\u00a0(Touvron et\u00a0al., 2023) possess strong audio knowledge and reasoning abilities without any in-domain audio training or fine-tuning, e.g., ChatGPT outputs \u201cdog bark is short, repetitive, and sharp woof or bark sound\u201d when we ask it the question \u201cwhat does a dog bark sound like?\u201d However, even the most advanced LLMs are not yet able to perceive general audio events.   The potential synergy between conventional audio models and LLMs in sound perception and reasoning motivates us to integrate them into a single model that is able to Listen to, Think about, and Understand the sound environment; we call this new model LTU. Specifically, the LTU framework effectively integrates a high-performant audio perception model (AST\u00a0(Gong et\u00a0al., 2021b)) with an open-source LLM (LLaMA\u00a0(Touvron et\u00a0al., 2023)). Since the data required to train LTU did not exist, we created a new dataset called OpenAQA-5M that combines 8 mainstream audio datasets, and formats all data as (audio, question, answer) tuples. Notably, OpenAQA-5M includes 3.7 million free-form open-ended and very diverse audio question-answer (AQA) pairs generated by a new GPT-assisted method called Audio Instruction Generation (AIG).   Performance-wise, on audio classification tasks, LTU outperforms the conventional audio-text CLAP model\u00a0(Elizalde et\u00a0al., 2023) on all 8 benchmarks with an average relative improvement of 23.6%. In addition, unlike CLAP, LTU does not require a predefined label set during inference. On captioning tasks, LTU is comparable with SOTA specialized models. Of greater interest, LTU exhibits emerging audio reasoning and comprehension abilities (shown in Figure\u00a01 and Table\u00a06) that are absent in existing audio models. It can answer open-ended questions about audio with an instruction following and factual correctness rate of 82.9%. With these capabilities,",
            "references": [
                {
                    "title": "On Decoder-Only Architecture For Speech-to-Text and Large Language Model Integration",
                    "abstract": "Large language models (LLMs) have achieved remarkable success in the field of natural language processing, enabling better human-computer interaction using natural language. However, the seamless integration of speech signals into LLMs has not been explored well. The \u201cdecoder-only\u201c architecture has also not been well studied for speech processing tasks. In this research, we introduce Speech-LLaMA, a novel approach that effectively incorporates acoustic information into text-based large language models. Our method leverages Connectionist Temporal Classification and a simple audio encoder to map the compressed acoustic features to the continuous semantic space of the LLM. In addition, we further probe the decoder-only architecture for speech-to-text tasks by training a smaller scale randomly initialized speech-LLaMA model from speech-text paired data alone. We conduct experiments on multilingual speech-to-text translation tasks and demonstrate a significant improvement over strong baselines, highlighting the potential advantages of decoder-only models for speech-to-text conversion."
                },
                {
                    "title": "Kosmos-2: Grounding Multimodal Large Language Models to the World",
                    "abstract": "We introduce Kosmos-2, a Multimodal Large Language Model (MLLM), enabling new capabilities of perceiving object descriptions (e.g., bounding boxes) and grounding text to the visual world. Specifically, we represent refer expressions as links in Markdown, i.e., ``[text span](bounding boxes)'', where object descriptions are sequences of location tokens. Together with multimodal corpora, we construct large-scale data of grounded image-text pairs (called GrIT) to train the model. In addition to the existing capabilities of MLLMs (e.g., perceiving general modalities, following instructions, and performing in-context learning), Kosmos-2 integrates the grounding capability into downstream applications. We evaluate Kosmos-2 on a wide range of tasks, including (i) multimodal grounding, such as referring expression comprehension, and phrase grounding, (ii) multimodal referring, such as referring expression generation, (iii) perception-language tasks, and (iv) language understanding and generation. This work lays out the foundation for the development of Embodiment AI and sheds light on the big convergence of language, multimodal perception, action, and world modeling, which is a key step toward artificial general intelligence. Code and pretrained models are available at https://aka.ms/kosmos-2."
                },
                {
                    "title": "CLAP Learning Audio Concepts from Natural Language Supervision",
                    "abstract": "Mainstream machine listening models are trained to learn audio concepts under the paradigm of one class label to many recordings focusing on one task. Learning under such restricted supervision limits the flexibility of models because they require labeled audio for training and can only predict the predefined categories. Instead, we propose to learn audio concepts from natural language supervision. We call our approach Contrastive Language-Audio Pretraining (CLAP), which connects language and audio by using two encoders and a contrastive learning objective, bringing audio and text descriptions into a joint multimodal space. We trained CLAP with 128k audio and text pairs and evaluated it on 16 downstream tasks across 7 domains, such as classification of sound events, scenes, music, and speech. CLAP establishes state-of-the-art (SoTA) in Zero-Shot performance. Also, we evaluated CLAP\u2019s audio encoder in a supervised learning setup and achieved SoTA in 5 tasks. The Zero-Shot capability removes the need of training with class labeled audio, enables flexible class prediction at inference time, and generalizes well in multiple downstream tasks. Code is available at: https://github.com/microsoft/CLAP."
                },
                {
                    "title": "Dataset Balancing Can Hurt Model Performance",
                    "abstract": "Machine learning from training data with a skewed distribution of examples per class can lead to models that favor performance on common classes at the expense of performance on rare ones. AudioSet has a very wide range of priors over its 527 sound event classes. Classification performance on AudioSet is usually evaluated by a simple average over per-class metrics, meaning that performance on rare classes is equal in importance to the performance on common ones. Several recent papers have used dataset balancing techniques to improve performance on AudioSet. We find, however, that while balancing improves performance on the public AudioSet evaluation data it simultaneously hurts performance on an unpublished evaluation set collected under the same conditions. By varying the degree of balancing, we show that its benefits are fragile and depend on the evaluation set. We also do not find evidence indicating that balancing improves rare class performance relative to common classes. We therefore caution against blind application of balancing, as well as against paying too much attention to small improvements on a public evaluation set."
                },
                {
                    "title": "Pengi: An Audio Language Model for Audio Tasks",
                    "abstract": "In the domain of audio processing, Transfer Learning has facilitated the rise of Self-Supervised Learning and Zero-Shot Learning techniques. These approaches have led to the development of versatile models capable of tackling a wide array of tasks, while delivering state-of-the-art performance. However, current models inherently lack the capacity to produce the requisite language for open-ended tasks, such as Audio Captioning or Audio Question&Answering. We introduce Pengi, a novel Audio Language Model that leverages Transfer Learning by framing all audio tasks as text-generation tasks. It takes as input, an audio recording, and text, and generates free-form text as output. The input audio is represented as a sequence of continuous embeddings by an audio encoder. A text encoder does the same for the corresponding text input. Both sequences are combined as a prefix to prompt a pre-trained frozen language model. The unified architecture of Pengi enables open-ended tasks and close-ended tasks without any additional fine-tuning or task-specific extensions. When evaluated on 22 downstream tasks, our approach yields state-of-the-art performance in several of them. Our results show that connecting language models with audio models is a major step towards general-purpose audio understanding"
                },
                {
                    "title": "SpeechGPT: Empowering Large Language Models with Intrinsic Cross-Modal Conversational Abilities",
                    "abstract": "Multi-modal large language models are regarded as a crucial step towards Artificial General Intelligence (AGI) and have garnered significant interest with the emergence of ChatGPT. However, current speech-language models typically adopt the cascade paradigm, preventing inter-modal knowledge transfer. In this paper, we propose SpeechGPT, a large language model with intrinsic cross-modal conversational abilities, capable of perceiving and generating multi-model content. With discrete speech representations, we first construct SpeechInstruct, a large-scale cross-modal speech instruction dataset. Additionally, we employ a three-stage training strategy that includes modality-adaptation pre-training, cross-modal instruction fine-tuning, and chain-of-modality instruction fine-tuning. The experimental results demonstrate that SpeechGPT has an impressive capacity to follow multi-modal human instructions and highlight the potential of handling multiple modalities with one model. Demos are shown in https://0nutation.github.io/SpeechGPT.github.io/."
                },
                {
                    "title": "AudioGPT: Understanding and Generating Speech, Music, Sound, and Talking Head",
                    "abstract": "Large language models (LLMs) have exhibited remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. Despite the recent success, current LLMs are not capable of processing complex audio information or conducting spoken conversations (like Siri or Alexa). In this work, we propose a multi-modal AI system named AudioGPT, which complements LLMs (i.e., ChatGPT) with 1) foundation models to process complex audio information and solve numerous understanding and generation tasks; and 2) the input/output interface (ASR, TTS) to support spoken dialogue. With an increasing demand to evaluate multi-modal LLMs of human intention understanding and cooperation with foundation models, we outline the principles and processes and test AudioGPT in terms of consistency, capability, and robustness. Experimental results demonstrate the capabilities of AudioGPT in solving 16 AI tasks with speech, music, sound, and talking head understanding and generation in multi-round dialogues, which empower humans to create rich and diverse audio content with unprecedented ease. Code can be found in https://github.com/AIGC-Audio/AudioGPT"
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "Instruction Tuning with GPT-4",
                    "abstract": "Prior work has shown that finetuning large language models (LLMs) using machine-generated instruction-following data enables such models to achieve remarkable zero-shot capabilities on new tasks, and no human-written instructions are needed. In this paper, we present the first attempt to use GPT-4 to generate instruction-following data for LLM finetuning. Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by previous state-of-the-art models. We also collect feedback and comparison data from GPT-4 to enable a comprehensive evaluation and reward model training. We make our data generated using GPT-4 as well as our codebase publicly available."
                },
                {
                    "title": "WavCaps: A ChatGPT-Assisted Weakly-Labelled Audio Captioning Dataset for Audio-Language Multimodal Research",
                    "abstract": "The advancement of audio-language (AL) multimodal learning tasks has been significant in recent years, yet the limited size of existing audio-language datasets poses challenges for researchers due to the costly and time-consuming collection process. To address this data scarcity issue, we introduce WavCaps, the first large-scale weakly-labelled audio captioning dataset, comprising approximately 400 k audio clips with paired captions. We sourced audio clips and their raw descriptions from web sources and a sound event detection dataset. However, the online-harvested raw descriptions are highly noisy and unsuitable for direct use in tasks such as automated audio captioning. To overcome this issue, we propose a three-stage processing pipeline for filtering noisy data and generating high-quality captions, where ChatGPT, a large language model, is leveraged to filter and transform raw descriptions automatically. We conduct a comprehensive analysis of the characteristics of WavCaps dataset and evaluate it on multiple downstream audio-language multimodal learning tasks. The systems trained on WavCaps outperform previous state-of-the-art (SOTA) models by a significant margin. Our aspiration is for the WavCaps dataset we have proposed to facilitate research in audio-language multimodal learning and demonstrate the potential of utilizing large language models (LLMs) to enhance academic research."
                },
                {
                    "title": "LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention",
                    "abstract": "We present LLaMA-Adapter, a lightweight adaption method to efficiently fine-tune LLaMA into an instruction-following model. Using 52K self-instruct demonstrations, LLaMA-Adapter only introduces 1.2M learnable parameters upon the frozen LLaMA 7B model, and costs less than one hour for fine-tuning on 8 A100 GPUs. Specifically, we adopt a set of learnable adaption prompts, and prepend them to the word tokens at higher transformer layers. Then, a zero-initialized attention mechanism with zero gating is proposed, which adaptively injects the new instructional cues into LLaMA, while effectively preserves its pre-trained knowledge. With our efficient training, LLaMA-Adapter can generate high-quality responses, comparable to Alpaca with fully fine-tuned 7B parameters. Besides language commands, our approach can be simply extended to multi-modal instructions for learning image-conditioned LLaMA model, which achieves superior reasoning performance on ScienceQA and COCO Caption benchmarks. Furthermore, we also evaluate the zero-initialized attention mechanism for fine-tuning other pre-trained models (ViT, RoBERTa) on traditional vision and language tasks, demonstrating the superior generalization capacity of our approach. Code is released at https://github.com/OpenGVLab/LLaMA-Adapter."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "PaLM-E: An Embodied Multimodal Language Model",
                    "abstract": "Large language models excel at a wide range of complex tasks. However, enabling general inference in the real world, e.g., for robotics problems, raises the challenge of grounding. We propose embodied language models to directly incorporate real-world continuous sensor modalities into language models and thereby establish the link between words and percepts. Input to our embodied language model are multi-modal sentences that interleave visual, continuous state estimation, and textual input encodings. We train these encodings end-to-end, in conjunction with a pre-trained large language model, for multiple embodied tasks including sequential robotic manipulation planning, visual question answering, and captioning. Our evaluations show that PaLM-E, a single large embodied multimodal model, can address a variety of embodied reasoning tasks, from a variety of observation modalities, on multiple embodiments, and further, exhibits positive transfer: the model benefits from diverse joint training across internet-scale language, vision, and visual-language domains. Our largest model, PaLM-E-562B with 562B parameters, in addition to being trained on robotics tasks, is a visual-language generalist with state-of-the-art performance on OK-VQA, and retains generalist language capabilities with increasing scale."
                },
                {
                    "title": "SpeechPrompt v2: Prompt Tuning for Speech Classification Tasks",
                    "abstract": "Prompt tuning is a technology that tunes a small set of parameters to steer a pre-trained language model (LM) to directly generate the output for downstream tasks. Recently, prompt tuning has demonstrated its storage and computation efficiency in both natural language processing (NLP) and speech processing fields. These advantages have also revealed prompt tuning as a candidate approach to serving pre-trained LM for multiple tasks in a unified manner. For speech processing, SpeechPrompt shows its high parameter efficiency and competitive performance on a few speech classification tasks. However, whether SpeechPrompt is capable of serving a large number of tasks is unanswered. In this work, we propose SpeechPrompt v2, a prompt tuning framework capable of performing a wide variety of speech classification tasks, covering multiple languages and prosody-related tasks. The experiment result shows that SpeechPrompt v2 achieves performance on par with prior works with less than 0.15M trainable parameters in a unified framework."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Language Is Not All You Need: Aligning Perception with Language Models",
                    "abstract": "A big convergence of language, multimodal perception, action, and world modeling is a key step toward artificial general intelligence. In this work, we introduce Kosmos-1, a Multimodal Large Language Model (MLLM) that can perceive general modalities, learn in context (i.e., few-shot), and follow instructions (i.e., zero-shot). Specifically, we train Kosmos-1 from scratch on web-scale multimodal corpora, including arbitrarily interleaved text and images, image-caption pairs, and text data. We evaluate various settings, including zero-shot, few-shot, and multimodal chain-of-thought prompting, on a wide range of tasks without any gradient updates or finetuning. Experimental results show that Kosmos-1 achieves impressive performance on (i) language understanding, generation, and even OCR-free NLP (directly fed with document images), (ii) perception-language tasks, including multimodal dialogue, image captioning, visual question answering, and (iii) vision tasks, such as image recognition with descriptions (specifying classification via text instructions). We also show that MLLMs can benefit from cross-modal transfer, i.e., transfer knowledge from language to multimodal, and from multimodal to language. In addition, we introduce a dataset of Raven IQ test, which diagnoses the nonverbal reasoning capability of MLLMs."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "Large-Scale Contrastive Language-Audio Pretraining with Feature Fusion and Keyword-to-Caption Augmentation",
                    "abstract": "Contrastive learning has shown remarkable success in the field of multimodal representation learning. In this paper, we propose a pipeline of contrastive language-audio pretraining to develop an audio representation by combining audio data with natural language descriptions. To accomplish this target, we first release LAION-Audio-630K, a large collection of 633,526 audio-text pairs from different data sources. Second, we construct a contrastive language-audio pretraining model by considering different audio encoders and text encoders. We incorporate the feature fusion mechanism and keyword-to-caption augmentation into the model design to further enable the model to process audio inputs of variable lengths and enhance the performance. Third, we perform comprehensive experiments to evaluate our model across three tasks: text-to-audio retrieval, zero-shot audio classification, and supervised audio classification. The results demonstrate that our model achieves superior performance in text-to-audio retrieval task. In audio classification tasks, the model achieves state-of-the-art performance in the zero-shot setting and is able to obtain performance comparable to models\u2019 results in the non-zero-shot setting. LAION-Audio-630K1 and the proposed model 2 are both available to the public."
                },
                {
                    "title": "Zero-Shot Audio Classification Using Synthesised Classifiers and Pre-Trained Models",
                    "abstract": "Audio classification equips a machine with the feature of recognising the source of an audio sample. Different from the conventional setting, by using zero-shot learning, an audio classifier can work for new audio sources that are not appearing during training. However, current zero-shot audio classification methods have no sufficient capability in retrieving the discriminative information from seen-class audio samples, and hence, lead to very limited performance for transferring knowledge in representing audio features. To this end, we propose an approach using multiple synthesised classifiers and pre-trained models in order to jointly optimise several phantom discriminative classifiers on the audio features generated through pre-trained ResNet models. Our experimental results, based on the ESC-50 dataset, validate the effectiveness of the proposed approach, compared with the state-of-the-art approaches."
                },
                {
                    "title": "Contrastive Audio-Visual Masked Autoencoder",
                    "abstract": "In this paper, we first extend the recent Masked Auto-Encoder (MAE) model from a single modality to audio-visual multi-modalities. Subsequently, we propose the Contrastive Audio-Visual Masked Auto-Encoder (CAV-MAE) by combining contrastive learning and masked data modeling, two major self-supervised learning frameworks, to learn a joint and coordinated audio-visual representation. Our experiments show that the contrastive audio-visual correspondence learning objective not only enables the model to perform audio-visual retrieval tasks, but also helps the model learn a better joint representation. As a result, our fully self-supervised pretrained CAV-MAE achieves a new SOTA accuracy of 65.9% on VGGSound, and is comparable with the previous best supervised pretrained model on AudioSet in the audio-visual event classification task. Code and pretrained models are at https://github.com/yuangongnd/cav-mae."
                },
                {
                    "title": "An Exploration of Prompt Tuning on Generative Spoken Language Model for Speech Processing Tasks",
                    "abstract": "Speech representations learned from Self-supervised learning (SSL) models have been found beneficial for various speech processing tasks. However, utilizing SSL representations usually requires fine-tuning the pre-trained models or designing task-specific downstream models and loss functions, causing much memory usage and human labor. On the other hand, prompting in Natural Language Processing (NLP) is an efficient and widely used technique to leverage pre-trained language models (LMs). Nevertheless, such a paradigm is little studied in the speech community. We report in this paper the first exploration of the prompt tuning paradigm for speech processing tasks based on Generative Spoken Language Model (GSLM). Experiment results show that the prompt tuning technique achieves competitive performance in speech classification tasks with fewer trainable parameters than fine-tuning specialized downstream models. We further study the technique in challenging sequence generation tasks. Prompt tuning also demonstrates its potential, while the limitation and possible research directions are discussed in this paper."
                },
                {
                    "title": "Masked Autoencoders that Listen",
                    "abstract": "This paper studies a simple extension of image-based Masked Autoencoders (MAE) to self-supervised representation learning from audio spectrograms. Following the Transformer encoder-decoder design in MAE, our Audio-MAE first encodes audio spectrogram patches with a high masking ratio, feeding only the non-masked tokens through encoder layers. The decoder then re-orders and decodes the encoded context padded with mask tokens, in order to reconstruct the input spectrogram. We find it beneficial to incorporate local window attention in the decoder, as audio spectrograms are highly correlated in local time and frequency bands. We then fine-tune the encoder with a lower masking ratio on target datasets. Empirically, Audio-MAE sets new state-of-the-art performance on six audio and speech classification tasks, outperforming other recent models that use external supervised pre-training. The code and models will be at https://github.com/facebookresearch/AudioMAE."
                },
                {
                    "title": "Vocalsound: A Dataset for Improving Human Vocal Sounds Recognition",
                    "abstract": "Recognizing human non-speech vocalizations is an important task and has broad applications such as automatic sound transcription and health condition monitoring. However, existing datasets have a relatively small number of vocal sound samples or noisy labels. As a consequence, state-of-the-art audio event classification models may not perform well in detecting human vocal sounds. To support research on building robust and accurate vocal sound recognition, we have created a VocalSound dataset consisting of over 21,000 crowdsourced recordings of laughter, sighs, coughs, throat clearing, sneezes, and sniffs from 3,365 unique subjects. Experiments show that the vocal sound recognition performance of a model can be significantly improved by 41.9% by adding VocalSound dataset to an existing dataset as training material. In addition, different from previous datasets, the VocalSound dataset contains meta information such as speaker age, gender, native language, country, and health condition."
                },
                {
                    "title": "Flamingo: a Visual Language Model for Few-Shot Learning",
                    "abstract": "Building models that can be rapidly adapted to novel tasks using only a handful of annotated examples is an open challenge for multimodal machine learning research. We introduce Flamingo, a family of Visual Language Models (VLM) with this ability. We propose key architectural innovations to: (i) bridge powerful pretrained vision-only and language-only models, (ii) handle sequences of arbitrarily interleaved visual and textual data, and (iii) seamlessly ingest images or videos as inputs. Thanks to their flexibility, Flamingo models can be trained on large-scale multimodal web corpora containing arbitrarily interleaved text and images, which is key to endow them with in-context few-shot learning capabilities. We perform a thorough evaluation of our models, exploring and measuring their ability to rapidly adapt to a variety of image and video tasks. These include open-ended tasks such as visual question-answering, where the model is prompted with a question which it has to answer; captioning tasks, which evaluate the ability to describe a scene or an event; and close-ended tasks such as multiple-choice visual question-answering. For tasks lying anywhere on this spectrum, a single Flamingo model can achieve a new state of the art with few-shot learning, simply by prompting the model with task-specific examples. On numerous benchmarks, Flamingo outperforms models fine-tuned on thousands of times more task-specific data."
                },
                {
                    "title": "WAVPROMPT: Towards Few-Shot Spoken Language Understanding with Frozen Language Models",
                    "abstract": "Large-scale auto-regressive language models pretrained on massive text have demonstrated their impressive ability to perform new natural language tasks with only a few text examples, without the need for fine-tuning. Recent studies further show that such a few-shot learning ability can be extended to the text-image setting by training an encoder to encode the images into embeddings functioning like the text embeddings of the language model. Interested in exploring the possibility of transferring the few-shot learning ability to the audio-text setting, we propose a novel speech understanding framework, WavPrompt, where we finetune a wav2vec model to generate a sequence of audio embeddings understood by the language model. We show that WavPrompt is a few-shot learner that can perform speech understanding tasks better than a naive text baseline. We conduct detailed ablation studies on different components and hyperparameters to empirically identify the best model configuration. In addition, we conduct a non-speech understanding experiment to show WavPrompt can extract more information than just the transcriptions. Code is available at https://github.com/Hertin/WavPrompt"
                },
                {
                    "title": "HTS-AT: A Hierarchical Token-Semantic Audio Transformer for Sound Classification and Detection",
                    "abstract": "Audio classification is an important task of mapping audio samples into their corresponding labels. Recently, the transformer model with self-attention mechanisms has been adopted in this field. However, existing audio transformers require large GPU memories and long training time, meanwhile relying on pretrained vision models to achieve high performance, which limits the model\u2019s scalability in audio tasks. To combat these problems, we introduce HTS-AT: an audio transformer with a hierarchical structure to reduce the model size and training time. It is further combined with a token-semantic module to map final outputs into class featuremaps, thus enabling the model for the audio event detection (i.e. localization in time). We evaluate HTS-AT on three datasets of audio classification where it achieves new state-of-the-art (SOTA) results on AudioSet and ESC50, and equals the SOTA on Speech Command V2. It also achieves better performance in event localization than the previous CNN-based models. Moreover, HTS-AT requires only 35% model parameters and 15% training time of the previous audio transformer. These results demonstrate the high performance and high efficiency of HTS-AT."
                },
                {
                    "title": "Unsupervised Audio-Caption Aligning Learns Correspondences Between Individual Sound Events and Textual Phrases",
                    "abstract": "We investigate unsupervised learning of correspondences between sound events and textual phrases through aligning audio clips with textual captions describing the content of a whole audio clip. We align originally unaligned and unannotated audio clips and their captions by scoring the similarities between audio frames and words, as encoded by modality-specific encoders and using a ranking-loss criterion to optimize the model. After training, we obtain clip-caption similarity by averaging frame-word similarities and estimate event-phrase correspondences by calculating frame-phrase similarities. We evaluate the method with two cross-modal tasks: audio-caption retrieval, and phrase-based sound event detection (SED). Experimental results show that the proposed method can globally associate audio clips with captions as well as locally learn correspondences between individual sound events and textual phrases in an unsupervised manner."
                },
                {
                    "title": "Audio Captioning Transformer",
                    "abstract": "Audio captioning aims to automatically generate a natural language description of an audio clip. Most captioning models follow an encoder-decoder architecture, where the decoder predicts words based on the audio features extracted by the encoder. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are often used as the audio encoder. However, CNNs can be limited in modelling temporal relationships among the time frames in an audio signal, while RNNs can be limited in modelling the long-range dependencies among the time frames. In this paper, we propose an Audio Captioning Transformer (ACT), which is a full Transformer network based on an encoder-decoder architecture and is totally convolution-free. The proposed method has a better ability to model the global information within an audio signal as well as capture temporal relationships between audio events. We evaluate our model on AudioCaps, which is the largest audio captioning dataset publicly available. Our model shows competitive performance compared to other state-of-the-art approaches."
                },
                {
                    "title": "Audioclip: Extending Clip to Image, Text and Audio",
                    "abstract": "The rapidly evolving field of sound classification has greatly benefited from the methods of other domains. Today, the trend is to fuse domain-specific tasks and approaches together, which provides the community with new outstanding models.We present AudioCLIP \u2013 an extension of the CLIP model that handles audio in addition to text and images. Utilizing the AudioSet dataset, our proposed model incorporates the ESResNeXt audio-model into the CLIP framework, thus enabling it to perform multimodal classification and keeping CLIP\u2019s zero-shot capabilities.AudioCLIP achieves new state-of-the-art results in the Environmental Sound Classification (ESC) task and out-performs others by reaching accuracies of 97.15 % on ESC-50 and 90.07 % on UrbanSound8K. Further, it sets new baselines in the zero-shot ESC-task on the same datasets (69.40 % and 68.78 %, respectively).We also asses the influence of different training setups on the final performance of the proposed model. For the sake of reproducibility, our code is published."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "NAAQA: A Neural Architecture for Acoustic Question Answering",
                    "abstract": "The goal of the Acoustic Question Answering (AQA) task is to answer a free-form text question about the content of an acoustic scene. It was inspired by the Visual Question Answering (VQA) task. In this paper, based on the previously introduced CLEAR dataset, we propose a new benchmark for AQA, namely CLEAR2, that emphasizes the specific challenges of acoustic inputs. These include handling of variable duration scenes, and scenes built with elementary sounds that differ between training and test set. We also introduce NAAQA, a neural architecture that leverages specific properties of acoustic inputs. The use of 1D convolutions in time and frequency to process 2D spectro-temporal representations of acoustic content shows promising results and enables reductions in model complexity. We show that time coordinate maps augment temporal localization capabilities which enhance performance of the network by <inline-formula><tex-math notation=\"LaTeX\">$\\sim$</tex-math><alternatives><mml:math><mml:mo>\u223c</mml:mo></mml:math><inline-graphic xlink:href=\"abdelnour-ieq1-3194311.gif\"/></alternatives></inline-formula>17 percentage points. On the other hand, frequency coordinate maps have little influence on this task. NAAQA achieves 79.5% of accuracy on the AQA task with <inline-formula><tex-math notation=\"LaTeX\">$\\sim$</tex-math><alternatives><mml:math><mml:mo>\u223c</mml:mo></mml:math><inline-graphic xlink:href=\"abdelnour-ieq2-3194311.gif\"/></alternatives></inline-formula>four times fewer parameters than the previously explored VQA model. We evaluate the performance of NAAQA on an independent data set reconstructed from DAQA. We also test the addition of a MALiMo module in our model on both CLEAR2 and DAQA. We provide a detailed analysis of the results for the different question types. We release the code to produce CLEAR2 as well as NAAQA to foster research in this newly emerging machine learning task."
                },
                {
                    "title": "The Benefit of Temporally-Strong Labels in Audio Event Classification",
                    "abstract": "To reveal the importance of temporal precision in ground truth audio event labels, we collected precise (\u223c0.1 sec resolution) \"strong\" labels for a portion of the AudioSet dataset. We devised a temporally-strong evaluation set (including explicit negatives of varying difficulty) and a small strong-labeled training subset of 67k clips (compared to the original dataset\u2019s 1.8M clips labeled at 10 sec resolution). We show that fine-tuning with a mix of weak- and strongly-labeled data can substantially improve classifier performance, even when evaluated using only the original weak labels. For a ResNet-50 architecture, d\u2032 on the strong evaluation data including explicit negatives improves from 1.13 to 1.39. The new labels are available as an update to AudioSet."
                },
                {
                    "title": "Zero-Shot Single-Microphone Sound Classification and Localization in a Building Via the Synthesis of Unseen Features",
                    "abstract": "In this paper, we propose a learning-based approach to identify the type and position of sounds using a single microphone in a real-world building. We attempt to treat this problem as a joint classification problem in which we predict the exact positions of sounds while classifying the types that are assumed to be from pre-defined types of sounds. The most problematic issue is that while the types are readily classified under supervised learning frameworks with one-hot encoded labels, it is difficult to predict the exact positions of the sound from unseen positions during training. To address this potential discrepancy, we formulate the position identification problem as a zero-shot learning problem inspired by the human ability to perceive new concepts from previously learned concepts. We extract feature representations from audio data and vectorize the type and position of the sound source as \u2018type/position-aware attributes,\u2019 instead of labeling each class with a simple one-hot vector. We then train a promising generative model to bridge the extracted features and the attributes by learning the class-invariant structure to transfer the knowledge from seen to unseen classes through their attributes; generative adversarial networks are conditioned on the class-embeddings. Our proposed methods are evaluated on an indoor noise dataset, SNU-B36-EX, a real-world dataset collected inside a building."
                },
                {
                    "title": "The Power of Scale for Parameter-Efficient Prompt Tuning",
                    "abstract": "In this work, we explore \u201cprompt tuning,\u201d a simple yet effective mechanism for learning \u201csoft prompts\u201d to condition frozen language models to perform specific downstream tasks. Unlike the discrete text prompts used by GPT-3, soft prompts are learned through backpropagation and can be tuned to incorporate signals from any number of labeled examples. Our end-to-end learned approach outperforms GPT-3\u2019s few-shot learning by a large margin. More remarkably, through ablations on model size using T5, we show that prompt tuning becomes more competitive with scale: as models exceed billions of parameters, our method \u201ccloses the gap\u201d and matches the strong performance of model tuning (where all model weights are tuned). This finding is especially relevant because large models are costly to share and serve and the ability to reuse one frozen model for multiple downstream tasks can ease this burden. Our method can be seen as a simplification of the recently proposed \u201cprefix tuning\u201d of Li and Liang (2021) and we provide a comparison to this and other similar approaches. Finally, we show that conditioning a frozen model with soft prompts confers benefits in robustness to domain transfer and enables efficient \u201cprompt ensembling.\u201d We release code and model checkpoints to reproduce our experiments."
                },
                {
                    "title": "AST: Audio Spectrogram Transformer",
                    "abstract": "In the past decade, convolutional neural networks (CNNs) have been widely adopted as the main building block for end-to-end audio classification models, which aim to learn a direct mapping from audio spectrograms to corresponding labels. To better capture long-range global context, a recent trend is to add a self-attention mechanism on top of the CNN, forming a CNN-attention hybrid model. However, it is unclear whether the reliance on a CNN is necessary, and if neural networks purely based on attention are sufficient to obtain good performance in audio classification. In this paper, we answer the question by introducing the Audio Spectrogram Transformer (AST), the first convolution-free, purely attention-based model for audio classification. We evaluate AST on various audio classification benchmarks, where it achieves new state-of-the-art results of 0.485 mAP on AudioSet, 95.6% accuracy on ESC-50, and 98.1% accuracy on Speech Commands V2."
                },
                {
                    "title": "Investigating Local and Global Information for Automated Audio Captioning with Transfer Learning",
                    "abstract": "Automated audio captioning (AAC) aims at generating summarizing descriptions for audio clips. Multitudinous concepts are described in an audio caption, ranging from local information such as sound events to global information like acoustic scenery. Currently, the mainstream paradigm for AAC is the end-to-end encoder-decoder architecture, expecting the encoder to learn all levels of concepts embedded in the audio automatically. This paper first proposes a topic model for audio descriptions, comprehensively analyzing the hierarchical audio topics that are commonly covered. We then explore a transfer learning scheme to access local and global information. Two source tasks are identified to respectively represent local and global information, being Audio Tagging (AT) and Acoustic Scene Classification (ASC). Experiments are conducted on the AAC benchmark dataset Clotho and Audiocaps, amounting to a vast increase in all eight metrics with topic transfer learning. Further, it is discovered that local information and abstract representation learning are more crucial to AAC than global information and temporal relationship learning."
                },
                {
                    "title": "PSLA: Improving Audio Tagging With Pretraining, Sampling, Labeling, and Aggregation",
                    "abstract": "Audio tagging is an active research area and has a wide range of applications. Since the release of AudioSet, great progress has been made in advancing model performance, which mostly comes from the development of novel model architectures and attention modules. However, we find that appropriate training techniques are equally important for building audio tagging models with AudioSet, but have not received the attention they deserve. To fill the gap, in this work, we present PSLA, a collection of model agnostic training techniques that can noticeably boost the model accuracy including ImageNet pretraining, balanced sampling, data augmentation, label enhancement, model aggregation. While many of these techniques have been previously explored, we conduct a thorough investigation on their design choices and combine them together. By training an EfficientNet with pretraining, balanced sampling, data augmentation, and model aggregation, we obtain a single model (with 13.6 M parameters) and an ensemble model that achieve mean average precision (mAP) scores of 0.444 and 0.474 on AudioSet, respectively, outperforming the previous best system of 0.439 with 81 M parameters. In addition, our model also achieves a new state-of-the-art mAP of 0.567 on FSD50K. We also investigate the impact of label enhancement on the model performance."
                },
                {
                    "title": "On Generative Spoken Language Modeling from Raw Audio",
                    "abstract": "Abstract We introduce Generative Spoken Language Modeling, the task of learning the acoustic and linguistic characteristics of a language from raw audio (no text, no labels), and a set of metrics to automatically evaluate the learned representations at acoustic and linguistic levels for both encoding and generation. We set up baseline systems consisting of a discrete speech encoder (returning pseudo-text units), a generative language model (trained on pseudo- text), and a speech decoder (generating a waveform from pseudo-text) all trained without supervision and validate the proposed metrics with human evaluation. Across 3 speech encoders (CPC, wav2vec 2.0, HuBERT), we find that the number of discrete units (50, 100, or 200) matters in a task-dependent and encoder- dependent way, and that some combinations approach text-based systems.1"
                },
                {
                    "title": "Audio Captioning using Pre-Trained Large-Scale Language Model Guided by Audio-based Similar Caption Retrieval",
                    "abstract": "The goal of audio captioning is to translate input audio into its description using natural language. One of the problems in audio captioning is the lack of training data due to the difficulty in collecting audio-caption pairs by crawling the web. In this study, to overcome this problem, we propose to use a pre-trained large-scale language model. Since an audio input cannot be directly inputted into such a language model, we utilize guidance captions retrieved from a training dataset based on similarities that may exist in different audio. Then, the caption of the audio input is generated by using a pre-trained language model while referring to the guidance captions. Experimental results show that (i) the proposed method has succeeded to use a pre-trained language model for audio captioning, and (ii) the oracle performance of the pre-trained model-based caption generator was clearly better than that of the conventional method trained from scratch."
                },
                {
                    "title": "Zero-Shot Audio Classification with Factored Linear and Nonlinear Acoustic-Semantic Projections",
                    "abstract": "In this paper, we study zero-shot learning in audio classification through factored linear and nonlinear acoustic-semantic projections between audio instances and sound classes. Zero-shot learning in audio classification refers to classification problems that aim at recognizing audio instances of sound classes, which have no available training data but only semantic side information. In this paper, we address zero-shot learning by employing factored linear and nonlinear acoustic-semantic projections. We develop factored linear projections by applying rank decomposition to a bilinear model, and use nonlinear activation functions, such as tanh, to model the non-linearity between acoustic embeddings and semantic embeddings. Compared with the prior bilinear model, experimental results show that the proposed projection methods are effective for improving classification performance of zero-shot learning in audio classification."
                },
                {
                    "title": "Zero-Shot Audio Classification Via Semantic Embeddings",
                    "abstract": "In this paper, we study zero-shot learning in audio classification via semantic embeddings extracted from textual labels and sentence descriptions of sound classes. Our goal is to obtain a classifier that is capable of recognizing audio instances of sound classes that have no available training samples, but only semantic side information. We employ a bilinear compatibility framework to learn an acoustic-semantic projection between intermediate-level representations of audio instances and sound classes, i.e., acoustic embeddings and semantic embeddings. We use VGGish to extract deep acoustic embeddings from audio clips, and pre-trained language models (Word2Vec, GloVe, BERT) to generate either label embeddings from textual labels or sentence embeddings from sentence descriptions of sound classes. Audio classification is performed by a linear compatibility function that measures how compatible an acoustic embedding and a semantic embedding are. We evaluate the proposed method on a small balanced dataset ESC-50 and a large-scale unbalanced audio subset of AudioSet. The experimental results show that classification performance is significantly improved by involving sound classes that are semantically close to the test classes in training. Meanwhile, we demonstrate that both label embeddings and sentence embeddings are useful for zero-shot learning. Classification performance is improved by concatenating label/sentence embeddings generated with different language models. With their hybrid concatenations, the results are improved further."
                },
                {
                    "title": "FSD50K: An Open Dataset of Human-Labeled Sound Events",
                    "abstract": "Most existing datasets for sound event recognition (SER) are relatively small and/or domain-specific, with the exception of AudioSet, based on over 2 M tracks from YouTube videos and encompassing over 500 sound classes. However, AudioSet is not an open dataset as its official release consists of pre-computed audio features. Downloading the original audio tracks can be problematic due to YouTube videos gradually disappearing and usage rights issues. To provide an alternative benchmark dataset and thus foster SER research, we introduce FSD50K, an open dataset containing over 51 k audio clips totalling over 100 h of audio manually labeled using 200 classes drawn from the AudioSet Ontology. The audio clips are licensed under Creative Commons licenses, making the dataset freely distributable (including waveforms). We provide a detailed description of the FSD50K creation process, tailored to the particularities of Freesound data, including challenges encountered and solutions adopted. We include a comprehensive dataset characterization along with discussion of limitations and key factors to allow its audio-informed usage. Finally, we conduct sound event classification experiments to provide baseline systems as well as insight on the main factors to consider when splitting Freesound audio data for SER. Our goal is to develop a dataset to be widely adopted by the community as a new open benchmark for SER research."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Vggsound: A Large-Scale Audio-Visual Dataset",
                    "abstract": "Our goal is to collect a large-scale audio-visual dataset with low label noise from videos \u2018in the wild\u2019 using computer vision techniques. The resulting dataset can be used for training and evaluating audio recognition models. We make three contributions. First, we propose a scalable pipeline based on computer vision techniques to create an audio dataset from open-source media. Our pipeline involves obtaining videos from YouTube; using image classification algorithms to localize audio-visual correspondence; and filtering out ambient noise using audio verification. Second, we use this pipeline to curate the VGGSound dataset consisting of more than 200k videos for 300 audio classes. Third, we investigate various Convolutional Neural Network (CNN) architectures and aggregation approaches to establish audio recognition baselines for our new dataset. Compared to existing audio datasets, VGGSound ensures audio-visual correspondence and is collected under unconstrained conditions. Code and the dataset are available at http://www.robots.ox.ac.uk/~vgg/data/vggsound/."
                },
                {
                    "title": "PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition",
                    "abstract": "Audio pattern recognition is an important research topic in the machine learning area, and includes several tasks such as audio tagging, acoustic scene classification, music classification, speech emotion classification and sound event detection. Recently, neural networks have been applied to tackle audio pattern recognition problems. However, previous systems are built on specific datasets with limited durations. Recently, in computer vision and natural language processing, systems pretrained on large-scale datasets have generalized well to several tasks. However, there is limited research on pretraining systems on large-scale datasets for audio pattern recognition. In this paper, we propose pretrained audio neural networks (PANNs) trained on the large-scale AudioSet dataset. These PANNs are transferred to other audio related tasks. We investigate the performance and computational complexity of PANNs modeled by a variety of convolutional neural networks. We propose an architecture called Wavegram-Logmel-CNN using both log-mel spectrogram and waveform as input feature. Our best PANN system achieves a state-of-the-art mean average precision (mAP) of 0.439 on AudioSet tagging, outperforming the best previous system of 0.392. We transfer PANNs to six audio pattern recognition tasks, and demonstrate state-of-the-art performance in several of those tasks. We have released the source code and pretrained models of PANNs: https://github.com/qiuqiangkong/audioset_tagging_cnn."
                },
                {
                    "title": "Sound Event Detection of Weakly Labelled Data With CNN-Transformer and Automatic Threshold Optimization",
                    "abstract": "Sound event detection (SED) is a task to detect sound events in an audio recording. One challenge of the SED task is that many datasets such as the Detection and Classification of Acoustic Scenes and Events (DCASE) datasets are weakly labelled. That is, there are only audio tags for each audio clip without the onset and offset times of sound events. We compare segment-wise and clip-wise training for SED that is lacking in previous works. We propose a convolutional neural network transformer (CNN-Transfomer) for audio tagging and SED, and show that CNN-Transformer performs similarly to a convolutional recurrent neural network (CRNN). Another challenge of SED is that thresholds are required for detecting sound events. Previous works set thresholds empirically, and are not an optimal approaches. To solve this problem, we propose an automatic threshold optimization method. The first stage is to optimize the system with respect to metrics that do not depend on thresholds, such as mean average precision (mAP). The second stage is to optimize the thresholds with respect to metrics that depends on those thresholds. Our proposed automatic threshold optimization system achieves a state-of-the-art audio tagging F1 of 0.646, outperforming that without threshold optimization of 0.629, and a sound event detection F1 of 0.584, outperforming that without threshold optimization of 0.564."
                },
                {
                    "title": "Clotho: an Audio Captioning Dataset",
                    "abstract": "Audio captioning is the novel task of general audio content description using free text. It is an intermodal translation task (not speech-to-text), where a system accepts as an input an audio signal and outputs the textual description (i.e. the caption) of that signal. In this paper we present Clotho, a dataset for audio captioning consisting of 4981 audio samples of 15 to 30 seconds duration and 24 905 captions of eight to 20 words length, and a baseline method to provide initial results. Clotho is built with focus on audio content and caption diversity, and the splits of the data are not hampering the training or evaluation of methods. All sounds are from the Freesound platform, and captions are crowdsourced using Amazon Mechanical Turk and annotators from English speaking countries. Unique words, named entities, and speech transcription are removed with post-processing. Clotho is freely available online 1."
                },
                {
                    "title": "A Deep Residual Network for Large-Scale Acoustic Scene Analysis",
                    "abstract": "Many of the recent advances in audio event detection, particularly on the AudioSet data set, have focused on improving performance using the released embeddings produced by a pretrained model. In this work, we instead study the task of training a multi-label event classifier directly from the audio recordings of AudioSet. Using the audio recordings, not only are we able to reproduce results from prior work, we have also confirmed improvements of other proposed additions, such as an attention module. Moreover, by training the embedding network jointly with the additions, we achieve an mAP of 0.392 and an AUC of 0.971, surpassing the state of the art without transfer learning from a large data set. We also analyze the output activations of the network and find that the models are able to localize audio events when a finer time resolution is needed."
                },
                {
                    "title": "CTRL: A Conditional Transformer Language Model for Controllable Generation",
                    "abstract": "Large-scale language models show promising text generation capabilities, but users cannot easily control particular aspects of the generated text. We release CTRL, a 1.63 billion-parameter conditional transformer language model, trained to condition on control codes that govern style, content, and task-specific behavior. Control codes were derived from structure that naturally co-occurs with raw text, preserving the advantages of unsupervised learning while providing more explicit control over text generation. These codes also allow CTRL to predict which parts of the training data are most likely given a sequence. This provides a potential method for analyzing large amounts of data via model-based source attribution. We have released multiple full-sized, pretrained versions of CTRL at this https URL."
                },
                {
                    "title": "Crowdsourcing a Dataset of Audio Captions",
                    "abstract": "Audio captioning is a novel field of multi-modal translation and it is the task of creating a textual description of the content of an audio signal (e.g. \"people talking in a big room\"). The creation of a dataset for this task requires a considerable amount of work, rendering the crowdsourcing a very attractive option. In this paper we present a three steps based framework for crowdsourcing an audio captioning dataset, based on concepts and practises followed for the creation of widely used image captioning and machine translations datasets. During the first step initial captions are gathered. A grammatically corrected and/or rephrased version of each initial caption is obtained in second step. Finally, the initial and edited captions are rated, keeping the top ones for the produced dataset. We objectively evaluate the impact of our framework during the process of creating an audio captioning dataset, in terms of diversity and amount of typographical errors in the obtained captions. The obtained results show that the resulting dataset has less typographical errors than the initial captions, and on average each sound in the produced dataset has captions with a Jaccard similarity of 0.24, roughly equivalent to two ten-word captions having in common four words with the same root, indicating that the captions are dissimilar while they still contain some of the same information."
                },
                {
                    "title": "AudioCaps: Generating Captions for Audios in The Wild",
                    "abstract": "We explore the problem of Audio Captioning: generating natural language description for any kind of audio in the wild, which has been surprisingly unexplored in previous research. We contribute a large-scale dataset of 46K audio clips with human-written text pairs collected via crowdsourcing on the AudioSet dataset. Our thorough empirical studies not only show that our collected captions are indeed faithful to audio inputs but also discover what forms of audio representation and captioning models are effective for the audio captioning. From extensive experiments, we also propose two novel components that help improve audio captioning performance: the top-down multi-scale encoder and aligned semantic attention."
                },
                {
                    "title": "Zero-Shot Audio Classification Based On Class Label Embeddings",
                    "abstract": "This paper proposes a zero-shot learning approach for audio classification based on the textual information about class labels without any audio samples from target classes. We propose an audio classification system built on the bilinear model, which takes audio feature embeddings and semantic class label embeddings as input, and measures the compatibility between an audio feature embedding and a class label embedding. We use VGGish to extract audio feature embeddings from audio recordings. We treat textual labels as semantic side information of audio classes, and use Word2Vec to generate class label embeddings. Results on the ESC-50 dataset show that the proposed system can perform zero-shot audio classification with small training dataset. It can achieve accuracy (26 % on average) better than random guess (10 %) on each audio category. Particularly, it reaches up to 39.7 % for the category of natural audio classes."
                },
                {
                    "title": "Look, Listen, and Learn More: Design Choices for Deep Audio Embeddings",
                    "abstract": "A considerable challenge in applying deep learning to audio classification is the scarcity of labeled data. An increasingly popular solution is to learn deep audio embeddings from large audio collections and use them to train shallow classifiers using small labeled datasets. Look, Listen, and Learn (L3-Net) is an embedding trained through self-supervised learning of audio-visual correspondence in videos as opposed to other embeddings requiring labeled data. This framework has the potential to produce powerful out-of-the-box embeddings for downstream audio classification tasks, but has a number of unexplained design choices that may impact the embeddings\u2019 behavior. In this paper we investigate how L3-Net design choices impact the performance of downstream audio classifiers trained with these embeddings. We show that audio-informed choices of input representation are important, and that using sufficient data for training the embedding is key. Surprisingly, we find that matching the content for training the embedding to the downstream task is not beneficial. Finally, we show that our best variant of the L3-Net embedding outperforms both the VGGish and SoundNet embeddings, while having fewer parameters and being trained on less data. Our implementation of the L3-Net embedding model as well as pre-trained models are made freely available online."
                },
                {
                    "title": "Weakly Labelled AudioSet Tagging With Attention Neural Networks",
                    "abstract": "Audio tagging is the task of predicting the presence or absence of sound classes within an audio clip. Previous work in audio tagging focused on relatively small datasets limited to recognizing a small number of sound classes. We investigate audio tagging on AudioSet, which is a dataset consisting of over 2 million audio clips and 527 classes. AudioSet is weakly labelled, in that only the presence or absence of sound classes is known for each clip, whereas the onset and offset times are unknown. To address the weakly labelled audio tagging problem, we propose attention neural networks as a way to attend the most salient parts of an audio clip. We bridge the connection between attention neural networks and multiple instance learning (MIL) methods, and propose decision-level and feature-level attention neural networks for audio tagging. We investigate attention neural networks modeled by different functions, depths, and widths. Experiments on AudioSet show that the feature-level attention neural network achieves a state-of-the-art mean average precision of 0.369, outperforming the best MIL method of 0.317 and Google's deep neural network baseline of 0.314. In addition, we discover that the audio tagging performance on AudioSet-embedding features has a weak correlation with the number of training samples and the quality of labels of each sound class."
                },
                {
                    "title": "CLEAR: A Dataset for Compositional Language and Elementary Acoustic Reasoning",
                    "abstract": "We introduce the task of acoustic question answering (AQA) in the area of acoustic reasoning. In this task an agent learns to answer questions on the basis of acoustic context. In order to promote research in this area, we propose a data generation paradigm adapted from CLEVR (Johnson et al. 2017). We generate acoustic scenes by leveraging a bank elementary sounds. We also provide a number of functional programs that can be used to compose questions and answers that exploit the relationships between the attributes of the elementary sounds in each scene. We provide AQA datasets of various sizes as well as the data generation code. As a preliminary experiment to validate our data, we report the accuracy of current state of the art visual question answering models when they are applied to the AQA task without modifications. Although there is a plethora of question answering tasks based on text, image or video data, to our knowledge, we are the first to propose answering questions directly on audio streams. We hope this contribution will facilitate the development of research in the area."
                },
                {
                    "title": "A Comparison of Five Multiple Instance Learning Pooling Functions for Sound Event Detection with Weak Labeling",
                    "abstract": "Sound event detection (SED) entails two subtasks: recognizing what types of sound events are present in an audio stream (audio tagging), and pinpointing their onset and offset times (localization). In the popular multiple instance learning (MIL) framework for SED with weak labeling, an important component is the pooling function. This paper compares five types of pooling functions both theoretically and experimentally, with special focus on their performance of localization. Although the attention pooling function is currently receiving the most attention, we find the linear softmax pooling function to perform the best among the five. Using this pooling function, we build a neural network called TALNet. It is the first system to reach state-of-the-art audio tagging performance on Audio Set, while exhibiting strong localization performance on the DCASE 2017 challenge at the same time."
                },
                {
                    "title": "Hierarchical Neural Story Generation",
                    "abstract": "We explore story generation: creative systems that can build coherent and fluent passages of text about a topic. We collect a large dataset of 300K human-written stories paired with writing prompts from an online forum. Our dataset enables hierarchical story generation, where the model first generates a premise, and then transforms it into a passage of text. We gain further improvements with a novel form of model fusion that improves the relevance of the story to the prompt, and adding a new gated multi-scale self-attention mechanism to model long-range context. Experiments show large improvements over strong baselines on both automated and human evaluations. Human judges prefer stories generated by our approach to those from a strong non-hierarchical model by a factor of two to one."
                },
                {
                    "title": "Multi-level attention model for weakly supervised audio classification",
                    "abstract": "In this paper, we propose a multi-level attention model for the \nweakly labelled audio classification problem. The objective of audio \nclassification is to predict the presence or the absence of sound \nevents in an audio clip. Recently, Google published a large scale \nweakly labelled AudioSet dataset containing 2 million audio clips \nwith only the presence or the absence labels of the sound events, \nwithout the onset and offset time of the sound events. Previously \nproposed attention models only applied a single attention module \non the last layer of a neural network which limited the capacity of \nthe attention model. In this paper, we propose a multi-level attention \nmodel which consists of multiple attention modules applied on the \nintermediate neural network layers. The outputs of these attention \nmodules are concatenated to a vector followed by a fully connected \nlayer to obtain the final prediction of each class. Experiments show \nthat the proposed multi-attention attention model achieves a state-of-the-art \nmean average precision (mAP) of 0.360, outperforming \nthe single attention model and the Google baseline system of 0.327 \nand 0.314, respectively."
                },
                {
                    "title": "DCASE2017 Challenge Setup: Tasks, Datasets and Baseline System",
                    "abstract": "DCASE 2017 Challenge consists of four tasks: acoustic scene classification , detection of rare sound events, sound event detection in real-life audio, and large-scale weakly supervised sound event detection for smart cars. This paper presents the setup of these tasks: task definition, dataset, experimental setup, and baseline system results on the development dataset. The baseline systems for all tasks rely on the same implementation using multilayer perceptron and log mel-energies, but differ in the structure of the output layer and the decision making process, as well as the evaluation of system output using task specific metrics."
                },
                {
                    "title": "Knowledge Transfer from Weakly Labeled Audio Using Convolutional Neural Network for Sound Events and Scenes",
                    "abstract": "In this work we propose approaches to effectively transfer knowledge from weakly labeled web audio data. We first describe a convolutional neural network (CNN) based framework for sound event detection and classification using weakly labeled audio data. Our model trains efficiently from audios of variable lengths; hence, it is well suited for transfer learning. We then propose methods to learn representations using this model which can be effectively used for solving the target task. We study both transductive and inductive transfer learning tasks, showing the effectiveness of our methods for both domain and task adaptation. We show that the learned representations using the proposed CNN model generalizes well enough to reach human level accuracy on ESC-50 sound events dataset and sets state of art results on this dataset. We further use them for acoustic scene classification task and once again show that our proposed approaches suit well for this task as well. We also show that our methods are helpful in capturing semantic meanings and relations as well. Moreover, in this process we also set state-of-art results on Audioset dataset using balanced training set."
                },
                {
                    "title": "Audio Set Classification with Attention Model: A Probabilistic Perspective",
                    "abstract": "This paper investigates the Audio Set classification. Audio Set is a large scale weakly labelled dataset (WLD) of audio clips. In WLD only the presence of a label is known, without knowing the happening time of the labels. We propose an attention model to solve this WLD problem and explain the attention model from a novel probabilistic perspective. Each audio clip in Audio Set consists of a collection of features. We call each feature as an instance and the collection as a bag following the terminology in multiple instance learning. In the attention model, each instance in the bag has a trainable probability measure for each class. The classification of the bag is the expectation of the classification output of the instances in the bag with respect to the learned probability measure. Experiments show that the proposed attention model achieves a mAP of 0.327 on Audio Set, outperforming the Google's baseline of 0.314."
                },
                {
                    "title": "Reducing Model Complexity for DNN Based Large-Scale Audio Classification",
                    "abstract": "Audio classification is the task of identifying the sound categories that are associated with a given audio signal. This paper presents an investigation on large-scale audio classification based on the recently released AudioSet database. AudioSet comprises 2 millions of audio samples from YouTube, which are human-annotated with 527 sound category labels. Audio classification experiments with the balanced training set and the evaluation set of AudioSet are carried out by applying different types of neural network models. The classification performance and the model complexity of these models are compared and analyzed. While the CNN models show better performance than MLP and RNN, its model complexity is relatively high and undesirable for practical use. We propose two different strategies that aim at constructing low-dimensional embedding feature extractors and hence reducing the number of model parameters. It is shown that the simplified CNN model has only 1/22 model parameters of the original model, with only a slight degradation of performance."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Audio Set: An ontology and human-labeled dataset for audio events",
                    "abstract": "Audio event recognition, the human-like ability to identify and relate sounds from audio, is a nascent problem in machine perception. Comparable problems such as object detection in images have reaped enormous benefits from comprehensive datasets - principally ImageNet. This paper describes the creation of Audio Set, a large-scale dataset of manually-annotated audio events that endeavors to bridge the gap in data availability between image and audio research. Using a carefully structured hierarchical ontology of 632 audio classes guided by the literature and manual curation, we collect data from human labelers to probe the presence of specific audio classes in 10 second segments of YouTube videos. Segments are proposed for labeling using searches based on metadata, context (e.g., links), and content analysis. The result is a dataset of unprecedented breadth and size that will, we hope, substantially stimulate the development of high-performance audio event recognizers."
                },
                {
                    "title": "TUT database for acoustic scene classification and sound event detection",
                    "abstract": "We introduce TUT Acoustic Scenes 2016 database for environmental sound research, consisting of binaural recordings from 15 different acoustic environments. A subset of this database, called TUT Sound Events 2016, contains annotations for individual sound events, specifically created for sound event detection. TUT Sound Events 2016 consists of residential area and home environments, and is manually annotated to mark onset, offset and label of sound events. In this paper we present the recording and annotation procedure, the database content, a recommended cross-validation setup and performance of supervised acoustic scene classification system and event detection baseline system using mel frequency cepstral coefficients and Gaussian mixture models. The database is publicly released to provide support for algorithm development and common ground for comparison of different techniques."
                },
                {
                    "title": "ESC: Dataset for Environmental Sound Classification",
                    "abstract": "One of the obstacles in research activities concentrating on environmental sound classification is the scarcity of suitable and publicly available datasets. This paper tries to address that issue by presenting a new annotated collection of 2000 short clips comprising 50 classes of various common sound events, and an abundant unified compilation of 250000 unlabeled auditory excerpts extracted from recordings available through the Freesound project. The paper also provides an evaluation of human accuracy in classifying environmental sounds and compares it to the performance of selected baseline classifiers using features derived from mel-frequency cepstral coefficients and zero-crossing rate."
                },
                {
                    "title": "A study of instrument-wise onset detection in Beijing Opera percussion ensembles",
                    "abstract": "Note onset detection and instrument recognition are two of the most investigated tasks in Music Information Retrieval (MIR). Various detection methods have been proposed in previous research for western music, with less focus on other music cultures of the world. In this paper, we focus on onset detection for percussion instruments in Beijing Opera, a major genre of Chinese traditional music. A dataset of individual audio samples of four primary percussion instruments is used to obtain the spectral bases for each instrument. With these bases, we separate the input percussion ensemble recordings into its spectral sources and their activations using a Non-negative Matrix Factorization (NMF) based algorithm. A simple onset detection conducted on each NMF activation presents satisfactory overall detection rates, and provides us valuable implications and suggestions for future development of drum transcription and percussion pattern analysis in Beijing Opera."
                },
                {
                    "title": "An Empirical Investigation of Catastrophic Forgeting in Gradient-Based Neural Networks",
                    "abstract": "Catastrophic forgetting is a problem faced by many machine learning models and algorithms. When trained on one task, then trained on a second task, many machine learning models \"forget\" how to perform the first task. This is widely believed to be a serious problem for neural networks. Here, we investigate the extent to which the catastrophic forgetting problem occurs for modern neural networks, comparing both established and recent gradient-based training algorithms and activation functions. We also examine the effect of the relationship between the first task and the second task on catastrophic forgetting. We find that it is always best to train using the dropout algorithm--the dropout algorithm is consistently best at adapting to the new task, remembering the old task, and has the best tradeoff curve between these two extremes. We find that different tasks and relationships between tasks result in very different rankings of activation function performance. This suggests the choice of activation function should always be cross-validated."
                },
                {
                    "title": "Freesound technical demo",
                    "abstract": "Freesound is an online collaborative sound database where people with diverse interests share recorded sound samples under Creative Commons licenses. It was started in 2005 and it is being maintained to support diverse research projects and as a service to the overall research and artistic community. In this demo we want to introduce Freesound to the multimedia community and show its potential as a research resource. We begin by describing some general aspects of Freesound, its architecture and functionalities, and then explain potential usages that this framework has for research applications."
                },
                {
                    "title": "Improving Audio-Language Learning with MixGen and Multi-Level Test-Time Augmentation",
                    "abstract": "In this paper, we propose two novel augmentation methods 1) audio-language MixGen (AL-MixGen) and 2) multi-level test-time augmentation (Multi-TTA) for audio-language learning. Inspired by MixGen, which is originally applied to vision-language learning, we introduce an augmentation method for the audio-language domain. We also explore the impact of test-time augmentations and present Multi-TTA which generalizes test-time augmentation over multiple layers of a deep learning model. Incorporating AL-MixGen and Multi-TTA into the baseline achieves 47.5 SPIDEr on audio cap-tioning, which is an +18.2% over the baseline and outperforms the state-of-the-art approach with a 5x smaller model. In audio-text retrieval, the proposed methods surpass the baseline performance as well."
                },
                {
                    "title": "Audio Captioning Based on Transformer and Pre-Trained CNN",
                    "abstract": "Automated audio captioning is the task that generates text description of a piece of audio. This paper proposes a solution of automated audio captioning based on a combination of pre-trained CNN layers and a sequence-to-sequence architecture based on Transformer. The pre-trained CNN layers are adopted from a CNN based neural network for acoustic event tagging, which makes the latent variable resulted more ef\ufb01cient on generating captions. Transformer decoder is used in the sequence-to-sequence architecture as a consequence of comparing the performance of the more classical LSTM layers. The proposed system achieves a SPIDEr score of 0.227 for the DCASE challenge 2020 Task 6 with data augmentation and label smoothing applied."
                },
                {
                    "title": "A CRNN-GRU Based Reinforcement Learning Approach to Audio Captioning",
                    "abstract": "Audio captioning aims at generating a natural sentence to describe the content in an audio clip. This paper proposes the use of a powerful CRNN encoder combined with a GRU decoder to tackle this multi-modal task. In addition to standard cross-entropy, reinforcement learning is also investigated for generating richer and more accurate captions. Our approach significantly improves against the baseline model on all shown metrics achieving a relative improvement of at least 34%. Results indicate that our proposed CRNNGRU model with reinforcement learning achieves a SPIDEr of 0.190 on the Clotho evaluation set. With data augmentation, the performance is further boosted to 0.223. In the DCASE challenge Task 6 we ranked fourth based on SPIDEr, second on 5 metrics including BLEU, ROUGE-L and METEOR, without ensemble or data augmentation while maintaining a small model size (only 5 Million parameters)."
                }
            ],
            "categories": [
                "eess.AS",
                "cs.SD"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively integrate audio perception models with large language models to enable a system that can listen to, think about, and understand audio environments?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning, particularly in audio understanding and reasoning. By developing a model that combines audio perception with reasoning capabilities, we can enhance the ability of AI systems to interpret complex auditory environments, leading to practical applications in areas such as autonomous navigation, human-computer interaction, and assistive technologies. This research could pave the way for future studies that explore multimodal learning and the integration of different AI capabilities, ultimately contributing to more intelligent and context-aware systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of audio signals and the need for a model to not only recognize sounds but also understand their context and implications. Naive approaches may fail because they typically focus on classification without addressing the reasoning aspect, which is essential for comprehending audio events. Additionally, the lack of existing datasets that combine audio with open-ended questions and answers presents a significant obstacle. The integration of high-performing audio models with language models also requires sophisticated methodologies to ensure effective communication between the two components.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either audio event recognition or language understanding in isolation, leading to a gap in integrated models that can handle both tasks simultaneously. Existing solutions often rely on predefined label sets, limiting their flexibility and reasoning capabilities. Barriers such as the absence of comprehensive datasets that link audio with contextual questions and answers have hindered progress. Our approach differs by creating the OpenAQA-5M dataset and developing the LTU model, which allows for open-ended reasoning about audio without the constraints of traditional label sets.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the integration of the AST audio perception model with the LLaMA language model within the LTU framework. We will utilize the newly created OpenAQA-5M dataset, which consists of 3.7 million audio question-answer pairs, to train the model. The performance will be evaluated using metrics such as accuracy in audio classification and the correctness rate of open-ended question answering. We expect LTU to outperform existing models in audio classification tasks and demonstrate significant reasoning capabilities, achieving an"
            }
        },
        "author_data": {
            "d89b3ac2-6189-4abc-bb29-61de50d03d78": {
                "pk": "d89b3ac2-6189-4abc-bb29-61de50d03d78",
                "name": "Yuan Gong",
                "collaborators": [
                    "James R. Glass",
                    "James Glass",
                    "Andrew Rouditchenko",
                    "Leonid Karlinsky",
                    "Alexander H. Liu",
                    "Hongyin Luo",
                    "Sameer Khurana",
                    "Yu-An Chung",
                    "Hildegard Kuehne",
                    "Rog\u00e9rio Feris",
                    "Nauman Dawalatabad",
                    "Tianhua Zhang",
                    "Yung-Sung Chuang",
                    "Xixin Wu",
                    "Yoon Kim",
                    "Jian Yang",
                    "C. Poellabauer",
                    "Samuel Thomas",
                    "Saurabhchand Bhati",
                    "Liming Wang",
                    "Marco Vilela",
                    "Katerina Placek",
                    "Brian Tracey",
                    "Yishu Gong",
                    "A. Premasiri",
                    "Fernando Vieira",
                    "Jiaxin Ge",
                    "Mingye Gao",
                    "Helen Meng",
                    "D. Fox",
                    "H. Meng",
                    "R. Au",
                    "David F. Harwath",
                    "Hilde Kuehne",
                    "I. Chu",
                    "Peng Chang",
                    "Jingbo Yu",
                    "Cheng-I Lai",
                    "Jacob Huber",
                    "Mitchell MacKnight"
                ],
                "domain": [
                    "Audio Processing",
                    "Speech Recognition",
                    "Machine Learning",
                    "Multi-modal Learning"
                ],
                "publications": [
                    {
                        "title": "Whisper-Flamingo: Integrating Visual Features into Whisper for Audio-Visual Speech Recognition and Translation",
                        "abstract": "Audio-Visual Speech Recognition (AVSR) uses lip-based video to improve performance in noise. Since videos are harder to obtain than audio, the video training data of AVSR models is usually limited to a few thousand hours. In contrast, speech models such as Whisper are trained with hundreds of thousands of hours of data, and thus learn a better speech-to-text decoder. The huge training data difference motivates us to adapt Whisper to handle video inputs. Inspired by Flamingo which injects visual features into language models, we propose Whisper-Flamingo which integrates visual features into the Whisper speech recognition and translation model with gated cross attention. Our audio-visual Whisper-Flamingo outperforms audio-only Whisper on English speech recognition and En-X translation for 6 languages in noisy conditions. Moreover, Whisper-Flamingo is a versatile model and conducts all of these tasks using one set of parameters, while prior methods are trained separately on each language."
                    },
                    {
                        "title": "DASS: Distilled Audio State Space Models Are Stronger and More Duration-Scalable Learners",
                        "abstract": "State-space models (SSMs) have emerged as an alternative to Transformers for audio modeling due to their high computational efficiency with long inputs. While recent efforts on Audio SSMs have reported encouraging results, two main limitations remain: First, in 10-second short audio tagging tasks, Audio SSMs still underperform compared to Transformer-based models such as Audio Spectrogram Transformer (AST). Second, although Audio SSMs theoretically support long audio inputs, their actual performance with long audio has not been thoroughly evaluated. To address these limitations, in this paper, 1) We applied knowledge distillation in audio space model training, resulting in a model called Knowledge Distilled Audio SSM (DASS). To the best of our knowledge, it is the first SSM that outperforms the Transformers on AudioSet and achieves an mAP of 47.6; and 2) We designed a new test called Audio Needle In A Haystack (Audio NIAH). We find that DASS, trained with only 10-second audio clips, can retrieve sound events in audio recordings up to 2.5 hours long, while the AST model fails when the input is just 50 seconds, demonstrating SSMs are indeed more duration scalable."
                    },
                    {
                        "title": "Automatic Prediction of Amyotrophic Lateral Sclerosis Progression using Longitudinal Speech Transformer",
                        "abstract": "Automatic prediction of amyotrophic lateral sclerosis (ALS) disease progression provides a more efficient and objective alternative than manual approaches. We propose ALS longitudinal speech transformer (ALST), a neural network-based automatic predictor of ALS disease progression from longitudinal speech recordings of ALS patients. By taking advantage of high-quality pretrained speech features and longitudinal information in the recordings, our best model achieves 91.0\\% AUC, improving upon the previous best model by 5.6\\% relative on the ALS TDI dataset. Careful analysis reveals that ALST is capable of fine-grained and interpretable predictions of ALS progression, especially for distinguishing between rarer and more severe cases. Code is publicly available."
                    },
                    {
                        "title": "Natural Language Embedded Programs for Hybrid Language Symbolic Reasoning",
                        "abstract": "How can we perform computations over natural language representations to solve tasks that require symbolic and numeric reasoning? We propose natural language embedded programs (NLEP) as a unifying framework for addressing math/symbolic reasoning, natural language understanding, and instruction following tasks. Our approach prompts a language model to generate full Python programs that define functions over data structures which contain natural language representations of structured knowledge. A Python interpreter then executes the generated code and prints the output. Despite using a task-general prompt, we find that this approach can improve upon strong baselines across a range of different tasks including math and symbolic reasoning, text classification, question answering, and instruction following. We found that the generated programs are interpretable since they outline the exact reasoning process followed by the program interpreter."
                    },
                    {
                        "title": "Whisper-AT: Noise-Robust Automatic Speech Recognizers are Also Strong General Audio Event Taggers",
                        "abstract": "In this paper, we focus on Whisper, a recent automatic speech recognition model trained with a massive 680k hour labeled speech corpus recorded in diverse conditions. We first show an interesting finding that while Whisper is very robust against real-world background sounds (e.g., music), its audio representation is actually not noise-invariant, but is instead highly correlated to non-speech sounds, indicating that Whisper recognizes speech conditioned on the noise type. With this finding, we build a unified audio tagging and speech recognition model Whisper-AT by freezing the backbone of Whisper, and training a lightweight audio tagging model on top of it. With<1% extra computational cost, Whisper-AT can recognize audio events, in addition to spoken text, in a single forward pass."
                    },
                    {
                        "title": "SAIL: Search-Augmented Instruction Learning",
                        "abstract": "Large language models (LLMs) have been significantly improved by instruction fine-tuning, but still lack transparency and the ability to utilize up-to-date knowledge and information. In this work, we propose search-augmented instruction learning (SAIL), which grounds the language generation and instruction following abilities on complex search results generated by in-house and external search engines. With an instruction tuning corpus, we collect search results for each training case from different search APIs and domains, and construct a new search-grounded training set containing \\textit{(instruction, grounding information, response)} triplets. We then fine-tune the LLaMA-7B model on the constructed training set. Since the collected results contain unrelated and disputing languages, the model needs to learn to ground on trustworthy search results, filter out distracting passages, and generate the target response. The search result-denoising process entails explicit trustworthy information selection and multi-hop reasoning, since the retrieved passages might be informative but not contain the instruction-following answer. Experiments show that the fine-tuned SAIL-7B model has a strong instruction-following ability, and it performs significantly better on transparency-sensitive tasks, including open-ended question answering and fact checking."
                    },
                    {
                        "title": "Joint Audio and Speech Understanding",
                        "abstract": "Humans are surrounded by audio signals that include both speech and non-speech sounds. The recognition and understanding of speech and non-speech audio events, along with a profound comprehension of the relationship between them, constitute fundamental cognitive capabilities. For the first time, we build a machine learning model, called LTU-AS, that has a conceptually similar universal audio perception and advanced reasoning ability. Specifically, by integrating Whisper [1] as a perception module and LLaMA [2] as a reasoning module, LTU-AS can simultaneously recognize and jointly understand spoken text, speech paralinguistics, and non-speech audio events - almost everything perceivable from audio signals."
                    },
                    {
                        "title": "Detecting Dementia from Long Neuropsychological Interviews",
                        "abstract": ","
                    },
                    {
                        "title": "Contrastive Audio-Visual Masked Autoencoder",
                        "abstract": "In this paper, we first extend the recent Masked Auto-Encoder (MAE) model from a single modality to audio-visual multi-modalities. Subsequently, we propose the Contrastive Audio-Visual Masked Auto-Encoder (CAV-MAE) by combining contrastive learning and masked data modeling, two major self-supervised learning frameworks, to learn a joint and coordinated audio-visual representation. Our experiments show that the contrastive audio-visual correspondence learning objective not only enables the model to perform audio-visual retrieval tasks, but also helps the model learn a better joint representation. As a result, our fully self-supervised pretrained CAV-MAE achieves a new SOTA accuracy of 65.9% on VGGSound, and is comparable with the previous best supervised pretrained model on AudioSet in the audio-visual event classification task. Code and pretrained models are at https://github.com/yuangongnd/cav-mae."
                    },
                    {
                        "title": "UAVM: A Unified Model for Audio-Visual Learning",
                        "abstract": "\u2014Conventional audio-visual models have independent audio and video branches. We design a uni\ufb01ed model for audio and video processing called Uni\ufb01ed Audio-Visual Model (UAVM). In this paper, we describe UAVM, report its new state-of-the-art audio-visual event classi\ufb01cation accuracy of 65.8% on VGGSound, and describe the intriguing properties of the model."
                    },
                    {
                        "title": "Transformer-Based Multi-Aspect Multi-Granularity Non-Native English Speaker Pronunciation Assessment",
                        "abstract": "Automatic pronunciation assessment is an important technology to help self-directed language learners. While pronunciation quality has multiple aspects including accuracy, fluency, completeness, and prosody, previous efforts typically only model one aspect (e.g., accuracy) at one granularity (e.g., at the phoneme-level). In this work, we explore modeling multi-aspect pronunciation assessment at multiple granularities. Specifically, we train a Goodness Of Pronunciation feature-based Transformer (GOPT) with multi-task learning. Experiments show that GOPT achieves the best results on speechocean762 with a public automatic speech recognition (ASR) acoustic model trained on Librispeech."
                    },
                    {
                        "title": "CMKD: CNN/Transformer-Based Cross-Model Knowledge Distillation for Audio Classification",
                        "abstract": "Audio classification is an active research area with a wide range of applications. Over the past decade, convolutional neural networks (CNNs) have been the de-facto standard building block for end-to-end audio classification models. Recently, neural networks based solely on self-attention mechanisms such as the Audio Spectrogram Transformer (AST) have been shown to outperform CNNs. In this paper, we find an intriguing interaction between the two very different models - CNN and AST models are good teachers for each other. When we use either of them as the teacher and train the other model as the student via knowledge distillation (KD), the performance of the student model noticeably improves, and in many cases, is better than the teacher model. In our experiments with this CNN/Transformer Cross-Model Knowledge Distillation (CMKD) method we achieve new state-of-the-art performance on FSD50K, AudioSet, and ESC-50."
                    },
                    {
                        "title": "UAVM: Towards Unifying Audio and Visual Models",
                        "abstract": "Conventional audio-visual models have independent audio and video branches. In this work, we <italic>unify</italic> the audio and visual branches by designing a <underline>U</underline>nified <underline>A</underline>udio-<underline>V</underline>isual <underline>M</underline>odel (UAVM). The UAVM achieves a new state-of-the-art audio-visual event classification accuracy of 65.8% on VGGSound. More interestingly, we also find a few intriguing properties of UAVM that the modality-independent counterparts do not have. Code at <uri>github.com/yuangongnd/uavm</uri>."
                    },
                    {
                        "title": "Vocalsound: A Dataset for Improving Human Vocal Sounds Recognition",
                        "abstract": "Recognizing human non-speech vocalizations is an important task and has broad applications such as automatic sound transcription and health condition monitoring. However, existing datasets have a relatively small number of vocal sound samples or noisy labels. As a consequence, state-of-the-art audio event classification models may not perform well in detecting human vocal sounds. To support research on building robust and accurate vocal sound recognition, we have created a VocalSound dataset consisting of over 21,000 crowdsourced recordings of laughter, sighs, coughs, throat clearing, sneezes, and sniffs from 3,365 unique subjects. Experiments show that the vocal sound recognition performance of a model can be significantly improved by 41.9% by adding VocalSound dataset to an existing dataset as training material. In addition, different from previous datasets, the VocalSound dataset contains meta information such as speaker age, gender, native language, country, and health condition."
                    },
                    {
                        "title": "AST: Audio Spectrogram Transformer",
                        "abstract": "In the past decade, convolutional neural networks (CNNs) have been widely adopted as the main building block for end-to-end audio classification models, which aim to learn a direct mapping from audio spectrograms to corresponding labels. To better capture long-range global context, a recent trend is to add a self-attention mechanism on top of the CNN, forming a CNN-attention hybrid model. However, it is unclear whether the reliance on a CNN is necessary, and if neural networks purely based on attention are sufficient to obtain good performance in audio classification. In this paper, we answer the question by introducing the Audio Spectrogram Transformer (AST), the first convolution-free, purely attention-based model for audio classification. We evaluate AST on various audio classification benchmarks, where it achieves new state-of-the-art results of 0.485 mAP on AudioSet, 95.6% accuracy on ESC-50, and 98.1% accuracy on Speech Commands V2."
                    },
                    {
                        "title": "SSAST: Self-Supervised Audio Spectrogram Transformer",
                        "abstract": "Recently, neural networks based purely on self-attention, such as the Vision Transformer (ViT), have been shown to outperform deep learning models constructed with convolutional neural networks (CNNs) on various vision tasks, thus extending the success of Transformers, which were originally developed for language processing, to the vision domain. A recent study showed that a similar methodology can also be applied to the audio domain. Specifically, the Audio Spectrogram Transformer (AST) achieves state-of-the-art results on various audio classification benchmarks. However, pure Transformer models tend to require more training data compared to CNNs, and the success of the AST relies on supervised pretraining that requires a large amount of labeled data and a complex training pipeline, thus limiting the practical usage of AST. This paper focuses on audio and speech classification, and aims to reduce the need for large amounts of labeled data for the AST by leveraging self-supervised learning using unlabeled data. Specifically, we propose to pretrain the AST model with joint discriminative and generative masked spectrogram patch modeling (MSPM) using unlabeled audio from AudioSet and Librispeech. We evaluate our pretrained models on both audio and speech classification tasks including audio event classification, keyword spotting, emotion recognition, and speaker identification. The proposed self-supervised framework significantly boosts AST performance on all tasks, with an average improvement of 60.9%, leading to similar or even better results than a supervised pretrained AST. To the best of our knowledge, it is the first patch-based self-supervised learning framework in the audio and speech domain, and also the first self-supervised learning framework for AST."
                    },
                    {
                        "title": "PSLA: Improving Audio Tagging With Pretraining, Sampling, Labeling, and Aggregation",
                        "abstract": "Audio tagging is an active research area and has a wide range of applications. Since the release of AudioSet, great progress has been made in advancing model performance, which mostly comes from the development of novel model architectures and attention modules. However, we find that appropriate training techniques are equally important for building audio tagging models with AudioSet, but have not received the attention they deserve. To fill the gap, in this work, we present PSLA, a collection of model agnostic training techniques that can noticeably boost the model accuracy including ImageNet pretraining, balanced sampling, data augmentation, label enhancement, model aggregation. While many of these techniques have been previously explored, we conduct a thorough investigation on their design choices and combine them together. By training an EfficientNet with pretraining, balanced sampling, data augmentation, and model aggregation, we obtain a single model (with 13.6 M parameters) and an ensemble model that achieve mean average precision (mAP) scores of 0.444 and 0.474 on AudioSet, respectively, outperforming the previous best system of 0.439 with 81 M parameters. In addition, our model also achieves a new state-of-the-art mAP of 0.567 on FSD50K. We also investigate the impact of label enhancement on the model performance."
                    },
                    {
                        "title": "Detecting Replay Attacks Using Multi-Channel Audio: A Neural Network-Based Method",
                        "abstract": "With the rapidly growing number of security-sensitive systems that use voice as the primary input, it becomes increasingly important to address these systems\u2019 potential vulnerability to replay attacks. Previous efforts to address this concern have focused primarily on single-channel audio. In this paper, we introduce a novel neural network-based replay attack detection model that further leverages spatial information of multi-channel audio and is able to significantly improve the replay attack detection performance."
                    },
                    {
                        "title": "ReMASC: Realistic Replay Attack Corpus for Voice Controlled Systems",
                        "abstract": "This paper introduces a new database of voice recordings with the goal of supporting research on vulnerabilities and protection of voice-controlled systems (VCSs). In contrast to prior efforts, the proposed database contains both genuine voice commands and replayed recordings of such commands, collected in realistic VCSs usage scenarios and using modern voice assistant development kits. Specifically, the database contains recordings from four systems (each with a different microphone array) in a variety of environmental conditions with different forms of background noise and relative positions between speaker and device. To the best of our knowledge, this is the first publicly available database that has been specifically designed for the protection of state-of-the-art voice-controlled systems against various replay attacks in various conditions and environments."
                    }
                ]
            },
            "a8c00595-3a10-4589-82f8-44c8a8503107": {
                "pk": "a8c00595-3a10-4589-82f8-44c8a8503107",
                "name": "Hongyin Luo",
                "collaborators": [
                    "James R. Glass",
                    "James Glass",
                    "Tianhua Zhang",
                    "Xixin Wu",
                    "Yung-Sung Chuang",
                    "Yoon Kim",
                    "Leonid Karlinsky",
                    "Helen M. Meng",
                    "Jiaxin Ge",
                    "Yuan Gong",
                    "Junmo Kang",
                    "David Cox",
                    "Rog\u00e9rio Feris",
                    "Alan Ritter",
                    "Kun Li",
                    "Wei Fang",
                    "D. Fox",
                    "H. Meng",
                    "Zhen Wang",
                    "Jacob A. Hansen",
                    "Rameswar Panda",
                    "Philip Schroeder",
                    "Nathaniel Morgan",
                    "Luc Gaitskell",
                    "Thomas Hartvigsen",
                    "Yada Zhu",
                    "Mingye Gao",
                    "Helen Meng",
                    "Liping Tang",
                    "Siyuan Qian",
                    "Yulu Gan",
                    "Jie Fu",
                    "Shanghang Zhang",
                    "Alexander H. Liu",
                    "Yujia Xie",
                    "Pengcheng He",
                    "Rumen Dangovski",
                    "Yang Zhang",
                    "Shiyu Chang",
                    "M. Soljavci'c",
                    "Shang-Wen Li",
                    "Wen-tau Yih"
                ],
                "domain": [
                    "Large Language Models",
                    "Natural Language Processing",
                    "Knowledge Graphs",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Self-MoE: Towards Compositional Large Language Models with Self-Specialized Experts",
                        "abstract": "We present Self-MoE, an approach that transforms a monolithic LLM into a compositional, modular system of self-specialized experts, named MiXSE (MiXture of Self-specialized Experts). Our approach leverages self-specialization, which constructs expert modules using self-generated synthetic data, each equipping a shared base LLM with distinct domain-specific capabilities, activated via self-optimized routing. This allows for dynamic and capability-specific handling of various target tasks, enhancing overall capabilities, without extensive human-labeled data and added parameters. Our empirical results reveal that specializing LLMs may exhibit potential trade-offs in performances on non-specialized tasks. On the other hand, our Self-MoE demonstrates substantial improvements (6.5%p on average) over the base LLM across diverse benchmarks such as knowledge, reasoning, math, and coding. It also consistently outperforms other methods, including instance merging and weight merging, while offering better flexibility and interpretability by design with semantic experts and routing. Our findings highlight the critical role of modularity, the applicability of Self-MoE to multiple base LLMs, and the potential of self-improvement in achieving efficient, scalable, and adaptable systems."
                    },
                    {
                        "title": "THREAD: Thinking Deeper with Recursive Spawning",
                        "abstract": "Large language models (LLMs) have shown impressive capabilities across diverse settings, but still struggle as the length and complexity of the context increases. To address this challenge, we propose Thinking Recursively and Dynamically (ThReaD). THREAD frames model generation as a thread of execution that, based on the context, can run to completion or dynamically spawn new threads. By spawning, threads can offload work (e.g., thinking, retrieving information) to child threads, which only return tokens needed for the parent thread to do its work. In effect, this enables the model to adapt, as needed, the amount of intermediate work used to produce tokens. We apply THREAD in the settings of LLM task solving and question answering, where the dynamic threading allows the model to recursively decompose the given task or question into progressively simpler sub-problems that can be solved by separate child threads. We test THREAD, implemented using a few-shot learning approach, on diverse benchmarks for agent tasks and data-grounded question answering. THREAD achieves state-of-the-art performance with GPT-4 and GPT-3.5 on these benchmarks, including ALFWorld, TextCraft, and WebShop, along with two new benchmarks, DataCommons QA and MIMIC-III ICU QA. In addition, THREAD outperforms existing frameworks by 10% to 50% absolute points with smaller models, including Llama-3-8b and CodeLlama-7b."
                    },
                    {
                        "title": "Decoding on Graphs: Faithful and Sound Reasoning on Knowledge Graphs through Generation of Well-Formed Chains",
                        "abstract": "Knowledge Graphs (KGs) can serve as reliable knowledge sources for question answering (QA) due to their structured representation of knowledge. Existing research on the utilization of KG for large language models (LLMs) prevalently relies on subgraph retriever or iterative prompting, overlooking the potential synergy of LLMs' step-wise reasoning capabilities and KGs' structural nature. In this paper, we present DoG (Decoding on Graphs), a novel framework that facilitates a deep synergy between LLMs and KGs. We first define a concept, well-formed chain, which consists of a sequence of interrelated fact triplets on the KGs, starting from question entities and leading to answers. We argue that this concept can serve as a principle for making faithful and sound reasoning for KGQA. To enable LLMs to generate well-formed chains, we propose graph-aware constrained decoding, in which a constraint derived from the topology of the KG regulates the decoding process of the LLMs. This constrained decoding method ensures the generation of well-formed chains while making full use of the step-wise reasoning capabilities of LLMs. Based on the above, DoG, a training-free approach, is able to provide faithful and sound reasoning trajectories grounded on the KGs. Experiments across various KGQA tasks with different background KGs demonstrate that DoG achieves superior and robust performance. DoG also shows general applicability with various open-source LLMs."
                    },
                    {
                        "title": "Adaptive Query Rewriting: Aligning Rewriters through Marginal Probability of Conversational Answers",
                        "abstract": "Query rewriting is a crucial technique for passage retrieval in open-domain conversational question answering (CQA). It decontexualizes conversational queries into self-contained questions suitable for off-the-shelf retrievers. Existing methods attempt to incorporate retriever's preference during the training of rewriting models. However, these approaches typically rely on extensive annotations such as in-domain rewrites and/or relevant passage labels, limiting the models' generalization and adaptation capabilities. In this paper, we introduce AdaQR ($\\textbf{Ada}$ptive $\\textbf{Q}$uery $\\textbf{R}$ewriting), a framework for training query rewriting models with limited rewrite annotations from seed datasets and completely no passage label. Our approach begins by fine-tuning compact large language models using only ~$10\\%$ of rewrite annotations from the seed dataset training split. The models are then utilized to generate rewrite candidates for each query instance. A novel approach is then proposed to assess retriever's preference for these candidates by the probability of answers conditioned on the conversational query by marginalizing the Top-$K$ passages. This serves as the reward for optimizing the rewriter further using Direct Preference Optimization (DPO), a process free of rewrite and retrieval annotations. Experimental results on four open-domain CQA datasets demonstrate that AdaQR not only enhances the in-domain capabilities of the rewriter with limited annotation requirement, but also adapts effectively to out-of-domain datasets."
                    },
                    {
                        "title": "Interpretable Unified Language Checking",
                        "abstract": "Despite recent concerns about undesirable behaviors generated by large language models (LLMs), including non-factual, biased, and hateful language, we find LLMs are inherent multi-task language checkers based on their latent representations of natural and social knowledge. We present an interpretable, unified, language checking (UniLC) method for both human and machine-generated language that aims to check if language input is factual and fair. While fairness and fact-checking tasks have been handled separately with dedicated models, we find that LLMs can achieve high performance on a combination of fact-checking, stereotype detection, and hate speech detection tasks with a simple, few-shot, unified set of prompts. With the ``1/2-shot'' multi-task language checking method proposed in this work, the GPT3.5-turbo model outperforms fully supervised baselines on several language tasks. The simple approach and results suggest that based on strong latent knowledge representations, an LLM can be an adaptive and explainable tool for detecting misinformation, stereotypes, and hate speech."
                    },
                    {
                        "title": "Entailment as Robust Self-Learner",
                        "abstract": "Entailment has been recognized as an important metric for evaluating natural language understanding (NLU) models, and recent studies have found that entailment pretraining benefits weakly supervised fine-tuning. In this work, we design a prompting strategy that formulates a number of different NLU tasks as contextual entailment. This approach improves the zero-shot adaptation of pretrained entailment models. Secondly, we notice that self-training entailment-based models with unlabeled data can significantly improve the adaptation performance on downstream tasks. To achieve more stable improvement, we propose the Simple Pseudo-Label Editing (SimPLE) algorithm for better pseudo-labeling quality in self-training. We also found that both pretrained entailment-based models and the self-trained models are robust against adversarial evaluation data. Experiments on binary and multi-class classification tasks show that SimPLE leads to more robust self-training results, indicating that the self-trained entailment models are more efficient and trustworthy than large language models on language understanding tasks."
                    },
                    {
                        "title": "Self-Specialization: Uncovering Latent Expertise within Large Language Models",
                        "abstract": "Recent works have demonstrated the effectiveness of self-alignment in which a large language model is aligned to follow general instructions using instructional data generated from the model itself starting from a handful of human-written seeds. Instead of general alignment, in this work, we focus on self-alignment for expert domain specialization (e.g., biomedicine, finance). As a preliminary, we quantitively show the marginal effect that generic instruction-following training has on downstream expert domains' performance. To remedy this, we propose self-specialization - allowing for effective model specialization while achieving cross-task generalization by leveraging only a few labeled seeds. Self-specialization offers a data- and parameter-efficient way of\"carving out\"an expert model out of a generalist pre-trained LLM. Exploring a variety of popular open large models as a base for specialization, our experimental results in both biomedical and financial domains show that our self-specialized models outperform their base models by a large margin, and even larger models that are generally instruction-tuned or that have been adapted to the target domain by other means."
                    },
                    {
                        "title": "Natural Language Embedded Programs for Hybrid Language Symbolic Reasoning",
                        "abstract": "How can we perform computations over natural language representations to solve tasks that require symbolic and numeric reasoning? We propose natural language embedded programs (NLEP) as a unifying framework for addressing math/symbolic reasoning, natural language understanding, and instruction following tasks. Our approach prompts a language model to generate full Python programs that define functions over data structures which contain natural language representations of structured knowledge. A Python interpreter then executes the generated code and prints the output. Despite using a task-general prompt, we find that this approach can improve upon strong baselines across a range of different tasks including math and symbolic reasoning, text classification, question answering, and instruction following. We found that the generated programs are interpretable since they outline the exact reasoning process followed by the program interpreter."
                    },
                    {
                        "title": "ConvRGX: Recognition, Generation, and Extraction for Self-trained Conversational Question Answering",
                        "abstract": "Collecting and constructing human-annotated corpora for training conversational question-answering (CQA) models has recently been shown to be inefficient and costly. To solve this problem, previous works have proposed training QA models with automatically generated QA data. In this work, we extend earlier studies on QA synthesis, and propose an efficient QA data generation algorithm under conversational settings. Our model recognizes potential dialogue topics, generates corresponding questions, and extracts answers from grounding passages. To improve the quality of generated QAs and downstream self-training of CQA models, we propose dropout and agreement-based QA selection methods. We conduct experiments on both data augmentation and domain adaptation settings. Experiments on the QuAC and Doc2Dial tasks show that the proposed method can significantly improve the quality of generated QA data, and also improves the accuracy of self-trained CQA models based on the constructed training corpora."
                    },
                    {
                        "title": "SAIL: Search-Augmented Instruction Learning",
                        "abstract": "Large language models (LLMs) have been significantly improved by instruction fine-tuning, but still lack transparency and the ability to utilize up-to-date knowledge and information. In this work, we propose search-augmented instruction learning (SAIL), which grounds the language generation and instruction following abilities on complex search results generated by in-house and external search engines. With an instruction tuning corpus, we collect search results for each training case from different search APIs and domains, and construct a new search-grounded training set containing \\textit{(instruction, grounding information, response)} triplets. We then fine-tune the LLaMA-7B model on the constructed training set. Since the collected results contain unrelated and disputing languages, the model needs to learn to ground on trustworthy search results, filter out distracting passages, and generate the target response. The search result-denoising process entails explicit trustworthy information selection and multi-hop reasoning, since the retrieved passages might be informative but not contain the instruction-following answer. Experiments show that the fine-tuned SAIL-7B model has a strong instruction-following ability, and it performs significantly better on transparency-sensitive tasks, including open-ended question answering and fact checking."
                    },
                    {
                        "title": "Chain of Thought Prompt Tuning in Vision Language Models",
                        "abstract": "Language-Image Pre-training has demonstrated promising results on zero-shot and few-shot downstream tasks by prompting visual models with natural language prompts. However, most recent studies only use a single prompt for tuning, neglecting the inherent step-to-step cognitive reasoning process that humans conduct in complex task settings, for example, when processing images from unfamiliar domains. Chain of Thought is a simple and effective approximation to human reasoning process and has been proven useful for natural language processing (NLP) tasks. Based on this cognitive intuition, we believe that conducting effective reasoning is also an important problem in visual tasks, and a chain of thought could be a solution to this problem. In this work, we propose a novel chain of thought prompt tuning for vision-language modeling. Extensive experiments show that our method not only generalizes better in image classification tasks, has greater transferability beyond a single dataset, and has stronger domain generalization performance, but also performs much better in imagetext retrieval and visual question answering, which require more reasoning capabilities. We are the first to successfully adapt chain-of-thought prompting that combines visual and textual embeddings. We will release our codes"
                    },
                    {
                        "title": "Logic Against Bias: Textual Entailment Mitigates Stereotypical Sentence Reasoning",
                        "abstract": "Due to their similarity-based learning objectives, pretrained sentence encoders often internalize stereotypical assumptions that reflect the social biases that exist within their training corpora. In this paper, we describe several kinds of stereotypes concerning different communities that are present in popular sentence representation models, including pretrained next sentence prediction and contrastive sentence representation models. We compare such models to textual entailment models that learn language logic for a variety of downstream language understanding tasks. By comparing strong pretrained models based on text similarity with textual entailment learning, we conclude that the explicit logic learning with textual entailment can significantly reduce bias and improve the recognition of social communities, without an explicit de-biasing process."
                    },
                    {
                        "title": "Joint Audio and Speech Understanding",
                        "abstract": "Humans are surrounded by audio signals that include both speech and non-speech sounds. The recognition and understanding of speech and non-speech audio events, along with a profound comprehension of the relationship between them, constitute fundamental cognitive capabilities. For the first time, we build a machine learning model, called LTU-AS, that has a conceptually similar universal audio perception and advanced reasoning ability. Specifically, by integrating Whisper [1] as a perception module and LLaMA [2] as a reasoning module, LTU-AS can simultaneously recognize and jointly understand spoken text, speech paralinguistics, and non-speech audio events - almost everything perceivable from audio signals."
                    },
                    {
                        "title": "DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models",
                        "abstract": "Despite their impressive capabilities, large language models (LLMs) are prone to hallucinations, i.e., generating content that deviates from facts seen during pretraining. We propose a simple decoding strategy for reducing hallucinations with pretrained LLMs that does not require conditioning on retrieved external knowledge nor additional fine-tuning. Our approach obtains the next-token distribution by contrasting the differences in logits obtained from projecting the later layers versus earlier layers to the vocabulary space, exploiting the fact that factual knowledge in an LLMs has generally been shown to be localized to particular transformer layers. We find that this Decoding by Contrasting Layers (DoLa) approach is able to better surface factual knowledge and reduce the generation of incorrect facts. DoLa consistently improves the truthfulness across multiple choices tasks and open-ended generation tasks, for example improving the performance of LLaMA family models on TruthfulQA by 12-17% absolute points, demonstrating its potential in making LLMs reliably generate truthful facts."
                    },
                    {
                        "title": "DiffCSE: Difference-based Contrastive Learning for Sentence Embeddings",
                        "abstract": "We propose DiffCSE, an unsupervised contrastive learning framework for learning sentence embeddings. DiffCSE learns sentence embeddings that are sensitive to the difference between the original sentence and an edited sentence, where the edited sentence is obtained by stochastically masking out the original sentence and then sampling from a masked language model. We show that DiffSCE is an instance of equivariant contrastive learning, which generalizes contrastive learning and learns representations that are insensitive to certain types of augmentations and sensitive to other \u201charmful\u201d types of augmentations. Our experiments show that DiffCSE achieves state-of-the-art results among unsupervised sentence representation learning methods, outperforming unsupervised SimCSE by 2.3 absolute points on semantic textual similarity tasks."
                    }
                ]
            },
            "69fdd58c-7615-48c7-b4b3-a29c9e17d296": {
                "pk": "69fdd58c-7615-48c7-b4b3-a29c9e17d296",
                "name": "Alexander H. Liu",
                "collaborators": [
                    "James R. Glass",
                    "Andrew Rouditchenko",
                    "Hung-yi Lee",
                    "Yuan Gong",
                    "Cheng-I Lai",
                    "Heng-Jui Chang",
                    "James Glass",
                    "Michael Auli",
                    "Wei-Ning Hsu",
                    "Leonid Karlinsky",
                    "David F. Harwath",
                    "Hilde Kuehne",
                    "Yang Zhang",
                    "Shiyu Chang",
                    "Yung-Sung Chuang",
                    "Kaizhi Qian",
                    "David Cox",
                    "Shun-Po Chuang",
                    "Tzu-Wei Sung",
                    "Hongyin Luo",
                    "Alexei Baevskiv",
                    "Yi-Long Liou",
                    "Jui-Yang Hsu",
                    "Chen-Sheng Chen",
                    "Tsung-Te Liu",
                    "Samuel J. Yang",
                    "Alexei Baevski",
                    "Kevin Duarte",
                    "Brian Chen",
                    "Nina Shvetsova",
                    "Samuel Thomas",
                    "M. Shah",
                    "Yi Liao",
                    "Sameer Khurana",
                    "Erica Cooper",
                    "Yiyuan Liao",
                    "J. Yamagishi",
                    "SouYoung Jin",
                    "A. Oliva",
                    "Yu-An Chung",
                    "Lin-Shan Lee",
                    "Tao Tu",
                    "Yuan-Jui Chen"
                ],
                "domain": [
                    "Audio Processing",
                    "Speech Recognition",
                    "Self-Supervised Learning",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Joint Audio and Speech Understanding",
                        "abstract": "Humans are surrounded by audio signals that include both speech and non-speech sounds. The recognition and understanding of speech and non-speech audio events, along with a profound comprehension of the relationship between them, constitute fundamental cognitive capabilities. For the first time, we build a machine learning model, called LTU-AS, that has a conceptually similar universal audio perception and advanced reasoning ability. Specifically, by integrating Whisper [1] as a perception module and LLaMA [2] as a reasoning module, LTU-AS can simultaneously recognize and jointly understand spoken text, speech paralinguistics, and non-speech audio events - almost everything perceivable from audio signals."
                    },
                    {
                        "title": "Self-supervised Fine-tuning for Improved Content Representations by Speaker-invariant Clustering",
                        "abstract": "Self-supervised speech representation models have succeeded in various tasks, but improving them for content-related problems using unlabeled data is challenging. We propose speaker-invariant clustering (Spin), a novel self-supervised learning method that clusters speech representations and performs swapped prediction between the original and speaker-perturbed utterances. Spin disentangles speaker information and preserves content representations with just 45 minutes of fine-tuning on a single GPU. Spin improves pre-trained networks and outperforms prior methods in speech recognition and acoustic unit discovery."
                    },
                    {
                        "title": "DinoSR: Self-Distillation and Online Clustering for Self-supervised Speech Representation Learning",
                        "abstract": "In this paper, we introduce self-distillation and online clustering for self-supervised speech representation learning (DinoSR) which combines masked language modeling, self-distillation, and online clustering. We show that these concepts complement each other and result in a strong representation learning model for speech. DinoSR first extracts contextualized embeddings from the input audio with a teacher network, then runs an online clustering system on the embeddings to yield a machine-discovered phone inventory, and finally uses the discretized tokens to guide a student network. We show that DinoSR surpasses previous state-of-the-art performance in several downstream tasks, and provide a detailed analysis of the model and the learned discrete units."
                    },
                    {
                        "title": "Simple and Effective Unsupervised Speech Synthesis",
                        "abstract": "We introduce the first unsupervised speech synthesis system based on a simple, yet effective recipe. The framework lever-ages recent work in unsupervised speech recognition as well as existing neural-based speech synthesis. Using only unlabeled speech audio and unlabeled text as well as a lexicon, our method enables speech synthesis without the need for a human-labeled corpus. Experiments demonstrate the unsupervised system can synthesize speech similar to a supervised counterpart in terms of naturalness and intelligibility measured by human evaluation."
                    },
                    {
                        "title": "Contrastive Audio-Visual Masked Autoencoder",
                        "abstract": "In this paper, we first extend the recent Masked Auto-Encoder (MAE) model from a single modality to audio-visual multi-modalities. Subsequently, we propose the Contrastive Audio-Visual Masked Auto-Encoder (CAV-MAE) by combining contrastive learning and masked data modeling, two major self-supervised learning frameworks, to learn a joint and coordinated audio-visual representation. Our experiments show that the contrastive audio-visual correspondence learning objective not only enables the model to perform audio-visual retrieval tasks, but also helps the model learn a better joint representation. As a result, our fully self-supervised pretrained CAV-MAE achieves a new SOTA accuracy of 65.9% on VGGSound, and is comparable with the previous best supervised pretrained model on AudioSet in the audio-visual event classification task. Code and pretrained models are at https://github.com/yuangongnd/cav-mae."
                    },
                    {
                        "title": "A Fully Integrated 1.7mW Attention-Based Automatic Speech Recognition Processor",
                        "abstract": "This brief presents a low-power attention-based automatic speech recognition (ASR) processor achieving real-time recognition capability. The proposed attention window algorithm, compact end-to-end neural-network topology, and efficient computation dataflow effectively minimize the hardware complexity and power consumption, enabling a fully integrated low-power ASR processor solution without the necessity of any off-chip memory resource. The proposed design techniques reduced 98.9% weight memory and 92.1% power consumption with minimal degradation of 2.24% in recognition accuracy. The proposed ASR processor operates at 100MHz with 1.7mW at 0.9V, demonstrating 2x and 1.68x performance improvements in speed and power, respectively, compared to the previous ASR designs that require additional supports of off-chip memory or external decoder."
                    },
                    {
                        "title": "Masked Autoencoders As The Unified Learners For Pre-Trained Sentence Representation",
                        "abstract": "Despite the progresses on pre-trained language models, there is a lack of unified frameworks for pre-trained sentence representation. As such, it calls for different pre-training methods for specific scenarios, and the pre-trained models are likely to be limited by their universality and representation quality. In this work, we extend the recently proposed MAE style pre-training strategy, RetroMAE, such that it may effectively support a wide variety of sentence representation tasks. The extended framework consists of two stages, with RetroMAE conducted throughout the process. The first stage performs RetroMAE over generic corpora, like Wikipedia, BookCorpus, etc., from which the base model is learned. The second stage takes place on domain-specific data, e.g., MS MARCO and NLI, where the base model is continuingly trained based on RetroMAE and contrastive learning. The pre-training outputs at the two stages may serve different applications, whose effectiveness are verified with comprehensive experiments. Concretely, the base model are proved to be effective for zero-shot retrieval, with remarkable performances achieved on BEIR benchmark. The continuingly pre-trained models further benefit more downstream tasks, including the domain-specific dense retrieval on MS MARCO, Natural Questions, and the sentence embeddings' quality for standard STS and transfer tasks in SentEval. The empirical insights of this work may inspire the future design of sentence representation pre-training. Our pre-trained models and source code will be released to the public communities."
                    },
                    {
                        "title": "UAVM: A Unified Model for Audio-Visual Learning",
                        "abstract": "\u2014Conventional audio-visual models have independent audio and video branches. We design a uni\ufb01ed model for audio and video processing called Uni\ufb01ed Audio-Visual Model (UAVM). In this paper, we describe UAVM, report its new state-of-the-art audio-visual event classi\ufb01cation accuracy of 65.8% on VGGSound, and describe the intriguing properties of the model."
                    },
                    {
                        "title": "UAVM: Towards Unifying Audio and Visual Models",
                        "abstract": "Conventional audio-visual models have independent audio and video branches. In this work, we <italic>unify</italic> the audio and visual branches by designing a <underline>U</underline>nified <underline>A</underline>udio-<underline>V</underline>isual <underline>M</underline>odel (UAVM). The UAVM achieves a new state-of-the-art audio-visual event classification accuracy of 65.8% on VGGSound. More interestingly, we also find a few intriguing properties of UAVM that the modality-independent counterparts do not have. Code at <uri>github.com/yuangongnd/uavm</uri>."
                    },
                    {
                        "title": "Towards End-to-End Unsupervised Speech Recognition",
                        "abstract": "Unsupervised speech recognition has shown great potential to make Automatic Speech Recognition (ASR) systems accessible to every language. However, existing methods still heavily rely on hand-crafted pre-processing. Similar to the trend of making supervised speech recognition end-to-end, we introduce wav2vec-U 2.0 which does away with all audio-side pre-processing and improves accuracy through better architecture. In addition, we introduce an auxiliary self-supervised objective that ties model predictions back to the input. Experiments show that wav2vec-U 2.0 improves unsupervised recognition results across different languages while being conceptually simpler."
                    },
                    {
                        "title": "Routing with Self-Attention for Multimodal Capsule Networks",
                        "abstract": "The task of multimodal learning has seen a growing interest recently as it allows for training neural architectures based on different modalities such as vision, text, and audio. One challenge in training such models is that they need to jointly learn semantic concepts and their relationships across different input representations. Capsule networks have been shown to perform well in context of capturing the relation between low-level input features and higher-level concepts. However, capsules have so far mainly been used only in small-scale fully supervised settings due to the resource demand of conventional routing algorithms. We present a new multimodal capsule network that allows us to leverage the strength of capsules in the context of a multimodal learning framework on large amounts of video data. To adapt the capsules to large-scale input data, we propose a novel routing by self-attention mechanism that selects relevant capsules which are then used to generate a final joint multimodal feature representation. This allows not only for robust training with noisy video data, but also to scale up the size of the capsule network compared to traditional routing methods while still being computationally efficient. We evaluate the proposed architecture by pretraining it on a large-scale multimodal video dataset and applying it on four datasets in two challenging downstream tasks. Results show that the proposed multimodal capsule network is not only able to improve results compared to other routing techniques, but also achieves competitive performance on the task of multimodal learning."
                    },
                    {
                        "title": "PARP: Prune, Adjust and Re-Prune for Self-Supervised Speech Recognition",
                        "abstract": "Self-supervised speech representation learning (speech SSL) has demonstrated the benefit of scale in learning rich representations for Automatic Speech Recognition (ASR) with limited paired data, such as wav2vec 2.0. We investigate the existence of sparse subnetworks in pre-trained speech SSL models that achieve even better low-resource ASR results. However, directly applying widely adopted pruning methods such as the Lottery Ticket Hypothesis (LTH) is suboptimal in the computational cost needed. Moreover, we show that the discovered subnetworks yield minimal performance gain compared to the original dense network. We present Prune-Adjust-Re-Prune (PARP), which discovers and finetunes subnetworks for much better performance, while only requiring a single downstream ASR finetuning run. PARP is inspired by our surprising observation that subnetworks pruned for pre-training tasks need merely a slight adjustment to achieve a sizeable performance boost in downstream ASR tasks. Extensive experiments on low-resource ASR verify (1) sparse subnetworks exist in mono-lingual/multi-lingual pre-trained speech SSL, and (2) the computational advantage and performance gain of PARP over baseline pruning methods. In particular, on the 10min Librispeech split without LM decoding, PARP discovers subnetworks from wav2vec 2.0 with an absolute 10.9%/12.6% WER decrease compared to the full model. We further demonstrate the effectiveness of PARP via: cross-lingual pruning without any phone recognition degradation, the discovery of a multi-lingual subnetwork for 10 spoken languages in 1 finetuning run, and its applicability to pre-trained BERT/XLNet for natural language tasks."
                    },
                    {
                        "title": "Improving Automatic Speech Recognition and Speech Translation via Word Embedding Prediction",
                        "abstract": "In this article, we target speech translation (ST). We propose lightweight approaches that generally improve either ASR or end-to-end ST models. We leverage continuous representations of words, known as word embeddings, to improve ASR in cascaded systems as well as end-to-end ST models. The benefit of using word embedding is that word embedding can be obtained easily by training on pure textual data, which alleviates data scarcity issue. Also, word embedding provides additional contextual information to speech models. We motivate to distill the knowledge from word embedding into speech models. In ASR, we use word embeddings as a regularizer to reduce the WER, and further propose a novel decoding method to fuse the semantic relations among words for further improvement. In the end-to-end ST model, we propose leveraging word embeddings as an intermediate representation to enhance translation performance. Our analysis shows that it is possible to map speech signals to semantic space, which motivates future work on applying the proposed methods in spoken language processing tasks."
                    },
                    {
                        "title": "On the Interplay between Sparsity, Naturalness, Intelligibility, and Prosody in Speech Synthesis",
                        "abstract": "Are end-to-end text-to-speech (TTS) models over-parametrized? To what extent can these models be pruned, and what happens to their synthesis capabilities? This work serves as a starting point to explore pruning both spectrogram prediction networks and vocoders. We thoroughly investigate the tradeoffs between sparsity and its subsequent effects on synthetic speech. Additionally, we explore several aspects of TTS pruning: amount of finetuning data versus sparsity, TTS-Augmentation to utilize unspoken text, and combining knowledge distillation and pruning. Our findings suggest that not only are end-to-end TTS models highly prunable, but also, perhaps surprisingly, pruned TTS models can produce synthetic speech with equal or higher naturalness and intelligibility, with similar prosody. All of our experiments are conducted on publicly available models, and findings in this work are backed by large-scale subjective tests and objective measures. Code and 200 pruned models are made available to facilitate future research on efficiency in TTS1."
                    },
                    {
                        "title": "Cross-Modal Discrete Representation Learning",
                        "abstract": "In contrast to recent advances focusing on high-level representation learning across modalities, in this work we present a self-supervised learning framework that is able to learn a representation that captures finer levels of granularity across different modalities such as concepts or events represented by visual objects or spoken words. Our framework relies on a discretized embedding space created via vector quantization that is shared across different modalities. Beyond the shared embedding space, we propose a Cross-Modal Code Matching objective that forces the representations from different views (modalities) to have a similar distribution over the discrete embedding space such that cross-modal objects/actions localization can be performed without direct supervision. We show that the proposed discretized multi-modal fine-grained representation (e.g., pixel/word/frame) can complement high-level summary representations (e.g., video/sentence/waveform) for improved performance on cross-modal retrieval tasks. We also observe that the discretized representation uses individual clusters to represent the same semantic concept across modalities."
                    },
                    {
                        "title": "Non-Autoregressive Predictive Coding for Learning Speech Representations from Local Dependencies",
                        "abstract": "Self-supervised speech representations have been shown to be effective in a variety of speech applications. However, existing representation learning methods generally rely on the autoregressive model and/or observed global dependencies while generating the representation. In this work, we propose Non-Autoregressive Predictive Coding (NPC), a self-supervised method, to learn a speech representation in a non-autoregressive manner by relying only on local dependencies of speech. NPC has a conceptually simple objective and can be implemented easily with the introduced Masked Convolution Blocks. NPC offers a significant speedup for inference since it is parallelizable in time and has a fixed inference time for each time step regardless of the input sequence length. We discuss and verify the effectiveness of NPC by theoretically and empirically comparing it with other methods. We show that the NPC representation is comparable to other methods in speech experiments on phonetic and speaker classification while being more efficient."
                    },
                    {
                        "title": "End-to-End Whispered Speech Recognition with Frequency-Weighted Approaches and Pseudo Whisper Pre-training",
                        "abstract": "Whispering is an important mode of human speech, but no end-to-end recognition results for it were reported yet, probably due to the scarcity of available whispered speech data. In this paper, we present several approaches for end-to-end (E2E) recognition of whispered speech considering the special characteristics of whispered speech and the scarcity of data. This includes a frequency-weighted SpecAugment policy and a frequency-divided CNN feature extractor for better capturing the high-frequency structures of whispered speech, and a layer-wise transfer learning approach to pre-train a model with normal or normal-to-whispered converted speech then fine-tune it with whispered speech to bridge the gap between whispered and normal speech. We achieve an overall relative reduction of 19.8% in PER and 44.4% in CER on a relatively small whispered TIMIT corpus. The results indicate as long as we have a good E2E model pre-trained on normal or pseudo-whispered speech, a relatively small set of whispered speech may suffice to obtain a reasonably good E2E whispered speech recognizer."
                    },
                    {
                        "title": "Worse WER, but Better BLEU? Leveraging Word Embedding as Intermediate in Multitask End-to-End Speech Translation",
                        "abstract": "Speech translation (ST) aims to learn transformations from speech in the source language to the text in the target language. Previous works show that multitask learning improves the ST performance, in which the recognition decoder generates the text of the source language, and the translation decoder obtains the final translations based on the output of the recognition decoder. Because whether the output of the recognition decoder has the correct semantics is more critical than its accuracy, we propose to improve the multitask ST model by utilizing word embedding as the intermediate."
                    },
                    {
                        "title": "Semi-supervised Learning for Multi-speaker Text-to-speech Synthesis Using Discrete Speech Representation",
                        "abstract": "Recently, end-to-end multi-speaker text-to-speech (TTS) systems gain success in the situation where a lot of high-quality speech plus their corresponding transcriptions are available. However, laborious paired data collection processes prevent many institutes from building multi-speaker TTS systems of great performance. In this work, we propose a semi-supervised learning approach for multi-speaker TTS. A multi-speaker TTS model can learn from the untranscribed audio via the proposed encoder-decoder framework with discrete speech representation. The experiment results demonstrate that with only an hour of paired speech data, no matter the paired data is from multiple speakers or a single speaker, the proposed model can generate intelligible speech in different voices. We found the model can benefit from the proposed semi-supervised learning approach even when part of the unpaired speech data is noisy. In addition, our analysis reveals that different speaker characteristics of the paired data have an impact on the effectiveness of semi-supervised TTS."
                    }
                ]
            },
            "620557a3-0317-4629-b4ce-6423ed4c93ae": {
                "pk": "620557a3-0317-4629-b4ce-6423ed4c93ae",
                "name": "Leonid Karlinsky",
                "collaborators": [
                    "Rog\u00e9rio Feris",
                    "James Glass",
                    "Rameswar Panda",
                    "Sivan Doveh",
                    "Roei Herzig",
                    "Donghyun Kim",
                    "David Cox",
                    "Hildegard Kuehne",
                    "Assaf Arbelle",
                    "Aude Oliva",
                    "R. Feris",
                    "Junmo Kang",
                    "Hongyin Luo",
                    "Jacob A. Hansen",
                    "Alan Ritter",
                    "M. J. Mirza",
                    "Wei Lin",
                    "Trevor Darrell",
                    "Andrew Rouditchenko",
                    "Yuan Gong",
                    "Samuel Thomas",
                    "Paola Cascante-Bonilla",
                    "Zexue He",
                    "Julian McAuley",
                    "Dmitry Krotov",
                    "Zhen Wang",
                    "Mengjie Zhao",
                    "Zhuoyuan Mao",
                    "Paul Gavrikov",
                    "Michael Dorkenwald",
                    "Shiqi Yang",
                    "Saurav Jha",
                    "Hiromi Wakaki",
                    "Yuki Mitsufuji",
                    "Horst Possegger",
                    "Nasim Borazjanizadeh",
                    "James Seale Smith",
                    "Lazar Valkov",
                    "Shaunak Halbe",
                    "V. Gutta",
                    "Z. Kira",
                    "Saurabhchand Bhati",
                    "Aaron K. Baughman",
                    "Stephen Hammer",
                    "Rahul Agarwal",
                    "Gozde Akay",
                    "Eduardo Morales",
                    "Tony Johnson",
                    "Brandon Huang",
                    "Chancharik Mitra",
                    "Matt Stallone",
                    "Vaibhav Saxena",
                    "Bridget McGinn",
                    "Tim Bula",
                    "Mayank Mishra",
                    "Adriana Meza Soria",
                    "Gaoyuan Zhang",
                    "Aditya Prasad",
                    "Yikang Shen",
                    "Saptha Surendran",
                    "Shanmukha C. Guttula",
                    "Hima Patel",
                    "Parameswaran Selvam",
                    "Xuan-Hong Dang",
                    "Y. Koyfman",
                    "Atin Sood",
                    "Nirmit Desai",
                    "David D. Cox",
                    "Ruchir Puri",
                    "Irene Huang",
                    "V. Butoi",
                    "Hilde Kuhene",
                    "Trevor Darrel",
                    "Chuang Gan",
                    "Runqian Wang",
                    "Soumya Ghosh",
                    "Diego Antognini",
                    "Howard Zhong",
                    "Samarth Mishra",
                    "SouYoung Jin",
                    "Venkatesh Saligrama",
                    "Ligong Han",
                    "Seung-Jun Han",
                    "Shivchander Sudalairaj",
                    "Charlotte Loh",
                    "Rumen Dangovski",
                    "Fei Deng",
                    "Pulkit Agrawal",
                    "Dimitris N. Metaxas",
                    "Tsui-Wei Weng",
                    "Akash Srivastava",
                    "Yada Zhu",
                    "Sameer Khurana",
                    "Hilde Kuehne",
                    "David F. Harwath",
                    "Brian Kingsbury",
                    "Sivan Harary",
                    "Amit Alfassy",
                    "R. Giryes",
                    "S. Ullman"
                ],
                "domain": [
                    "Large Language Models",
                    "Vision-Language Models",
                    "Compositional Reasoning",
                    "Continual Learning"
                ],
                "publications": [
                    {
                        "title": "CAMELoT: Towards Large Language Models with Training-Free Consolidated Associative Memory",
                        "abstract": "Large Language Models (LLMs) struggle to handle long input sequences due to high memory and runtime costs. Memory-augmented models have emerged as a promising solution to this problem, but current methods are hindered by limited memory capacity and require costly re-training to integrate with a new LLM. In this work, we introduce an associative memory module which can be coupled to any pre-trained (frozen) attention-based LLM without re-training, enabling it to handle arbitrarily long input sequences. Unlike previous methods, our associative memory module consolidates representations of individual tokens into a non-parametric distribution model, dynamically managed by properly balancing the novelty and recency of the incoming data. By retrieving information from this consolidated associative memory, the base LLM can achieve significant (up to 29.7% on Arxiv) perplexity reduction in long-context modeling compared to other baselines evaluated on standard benchmarks. This architecture, which we call CAMELoT (Consolidated Associative Memory Enhanced Long Transformer), demonstrates superior performance even with a tiny context window of 128 tokens, and also enables improved in-context learning with a much larger set of demonstrations."
                    },
                    {
                        "title": "Self-MoE: Towards Compositional Large Language Models with Self-Specialized Experts",
                        "abstract": "We present Self-MoE, an approach that transforms a monolithic LLM into a compositional, modular system of self-specialized experts, named MiXSE (MiXture of Self-specialized Experts). Our approach leverages self-specialization, which constructs expert modules using self-generated synthetic data, each equipping a shared base LLM with distinct domain-specific capabilities, activated via self-optimized routing. This allows for dynamic and capability-specific handling of various target tasks, enhancing overall capabilities, without extensive human-labeled data and added parameters. Our empirical results reveal that specializing LLMs may exhibit potential trade-offs in performances on non-specialized tasks. On the other hand, our Self-MoE demonstrates substantial improvements (6.5%p on average) over the base LLM across diverse benchmarks such as knowledge, reasoning, math, and coding. It also consistently outperforms other methods, including instance merging and weight merging, while offering better flexibility and interpretability by design with semantic experts and routing. Our findings highlight the critical role of modularity, the applicability of Self-MoE to multiple base LLMs, and the potential of self-improvement in achieving efficient, scalable, and adaptable systems."
                    },
                    {
                        "title": "GLOV: Guided Large Language Models as Implicit Optimizers for Vision Language Models",
                        "abstract": "In this work, we propose a novel method (GLOV) enabling Large Language Models (LLMs) to act as implicit Optimizers for Vision-Langugage Models (VLMs) to enhance downstream vision tasks. Our GLOV meta-prompts an LLM with the downstream task description, querying it for suitable VLM prompts (e.g., for zero-shot classification with CLIP). These prompts are ranked according to a purity measure obtained through a fitness function. In each respective optimization step, the ranked prompts are fed as in-context examples (with their accuracies) to equip the LLM with the knowledge of the type of text prompts preferred by the downstream VLM. Furthermore, we also explicitly steer the LLM generation process in each optimization step by specifically adding an offset difference vector of the embeddings from the positive and negative solutions found by the LLM, in previous optimization steps, to the intermediate layer of the network for the next generation step. This offset vector steers the LLM generation toward the type of language preferred by the downstream VLM, resulting in enhanced performance on the downstream vision tasks. We comprehensively evaluate our GLOV on 16 diverse datasets using two families of VLMs, i.e., dual-encoder (e.g., CLIP) and encoder-decoder (e.g., LLaVa) models -- showing that the discovered solutions can enhance the recognition performance by up to 15.0% and 57.5% (3.8% and 21.6% on average) for these models."
                    },
                    {
                        "title": "Navigating the Labyrinth: Evaluating and Enhancing LLMs' Ability to Reason About Search Problems",
                        "abstract": "Recently, Large Language Models (LLMs) attained impressive performance in math and reasoning benchmarks. However, they still often struggle with logic problems and puzzles that are relatively easy for humans. To further investigate this, we introduce a new benchmark, SearchBench, containing 11 unique search problem types, each equipped with automated pipelines to generate an arbitrary number of instances and analyze the feasibility, correctness, and optimality of LLM-generated solutions. We show that even the most advanced LLMs fail to solve these problems end-to-end in text, e.g. GPT4 solves only 1.4%. SearchBench problems require considering multiple pathways to the solution as well as backtracking, posing a significant challenge to auto-regressive models. Instructing LLMs to generate code that solves the problem helps, but only slightly, e.g., GPT4's performance rises to 11.7%. In this work, we show that in-context learning with A* algorithm implementations enhances performance. The full potential of this promoting approach emerges when combined with our proposed Multi-Stage-Multi-Try method, which breaks down the algorithm implementation into two stages and verifies the first stage against unit tests, raising GPT-4's performance above 57%."
                    },
                    {
                        "title": "Whisper-Flamingo: Integrating Visual Features into Whisper for Audio-Visual Speech Recognition and Translation",
                        "abstract": "Audio-Visual Speech Recognition (AVSR) uses lip-based video to improve performance in noise. Since videos are harder to obtain than audio, the video training data of AVSR models is usually limited to a few thousand hours. In contrast, speech models such as Whisper are trained with hundreds of thousands of hours of data, and thus learn a better speech-to-text decoder. The huge training data difference motivates us to adapt Whisper to handle video inputs. Inspired by Flamingo which injects visual features into language models, we propose Whisper-Flamingo which integrates visual features into the Whisper speech recognition and translation model with gated cross attention. Our audio-visual Whisper-Flamingo outperforms audio-only Whisper on English speech recognition and En-X translation for 6 languages in noisy conditions. Moreover, Whisper-Flamingo is a versatile model and conducts all of these tasks using one set of parameters, while prior methods are trained separately on each language."
                    },
                    {
                        "title": "Adaptive Memory Replay for Continual Learning",
                        "abstract": "Foundation Models (FMs) have become the hallmark of modern AI, however, these models are trained on massive data, leading to financially expensive training. Updating FMs as new data becomes available is important, however, can lead to \u2018catastrophic forgetting\u2019, where models underperform on tasks related to data sub-populations observed too long ago. This continual learning (CL) phenomenon has been extensively studied, but primarily in a setting where only a small amount of past data can be stored. We advocate for a new paradigm where memory is abundant, allowing us to keep all previous data, but computational resources are limited. In this setting, traditional replay-based CL approaches are outperformed by a simple baseline which replays past data selected uniformly at random [33], indicating that this setting necessitates a new approach. We address this by introducing a framework of adaptive memory replay for continual learning, where sampling of past data is phrased as a multi-armed bandit problem. We utilize Bolzmann sampling to derive a method which dynamically selects past data for training conditioned on the current task, assuming full data access and emphasizing training efficiency. Through extensive evaluations on both vision and language pre-training tasks, we demonstrate the effectiveness of our approach, which maintains high performance while reducing forgetting by up to 10% at no training efficiency cost."
                    },
                    {
                        "title": "DASS: Distilled Audio State Space Models Are Stronger and More Duration-Scalable Learners",
                        "abstract": "State-space models (SSMs) have emerged as an alternative to Transformers for audio modeling due to their high computational efficiency with long inputs. While recent efforts on Audio SSMs have reported encouraging results, two main limitations remain: First, in 10-second short audio tagging tasks, Audio SSMs still underperform compared to Transformer-based models such as Audio Spectrogram Transformer (AST). Second, although Audio SSMs theoretically support long audio inputs, their actual performance with long audio has not been thoroughly evaluated. To address these limitations, in this paper, 1) We applied knowledge distillation in audio space model training, resulting in a model called Knowledge Distilled Audio SSM (DASS). To the best of our knowledge, it is the first SSM that outperforms the Transformers on AudioSet and achieves an mAP of 47.6; and 2) We designed a new test called Audio Needle In A Haystack (Audio NIAH). We find that DASS, trained with only 10-second audio clips, can retrieve sound events in audio recordings up to 2.5 hours long, while the AST model fails when the input is just 50 seconds, demonstrating SSMs are indeed more duration scalable."
                    },
                    {
                        "title": "Large Scale Generative AI Text Applied to Sports and Music",
                        "abstract": "We address the problem of scaling up the production of media content, including commentary and personalized news stories, for large-scale sports and music events worldwide. Our approach relies on generative AI models to transform a large volume of multimodal data (e.g., videos, articles, real-time scoring feeds, statistics, and fact sheets) into coherent and fluent text. Based on this approach, we introduce, for the first time, an AI commentary system, which was deployed to produce automated narrations for highlight packages at the 2023 US Open, Wimbledon, and Masters tournaments. In the same vein, our solution was extended to create personalized content for ESPN Fantasy Football and stories about music artists for the Grammy awards. These applications were built using a common software architecture achieved a 15x speed improvement with an average Rouge-L of 82.00 and perplexity of 6.6. Our work was successfully deployed at the aforementioned events, supporting 90 million fans around the world with 8 billion page views, continuously pushing the bounds on what is possible at the intersection of sports, entertainment, and AI."
                    },
                    {
                        "title": "Multimodal Task Vectors Enable Many-Shot Multimodal In-Context Learning",
                        "abstract": "The recent success of interleaved Large Multimodal Models (LMMs) in few-shot learning suggests that in-context learning (ICL) with many examples can be promising for learning new tasks. However, this many-shot multimodal ICL setting has one crucial problem: it is fundamentally limited by the model's context length set at pretraining. The problem is especially prominent in the multimodal domain, which processes both text and images, requiring additional tokens. This motivates the need for a multimodal method to compress many shots into fewer tokens without finetuning. In this work, we enable LMMs to perform multimodal, many-shot in-context learning by leveraging Multimodal Task Vectors (MTV)--compact implicit representations of in-context examples compressed in the model's attention heads. Specifically, we first demonstrate the existence of such MTV in LMMs and then leverage these extracted MTV to enable many-shot in-context learning for various vision-and-language tasks. Our experiments suggest that MTV can scale in performance with the number of compressed shots and generalize to similar out-of-domain tasks without additional context length for inference."
                    },
                    {
                        "title": "Scaling Granite Code Models to 128K Context",
                        "abstract": "This paper introduces long-context Granite code models that support effective context windows of up to 128K tokens. Our solution for scaling context length of Granite 3B/8B code models from 2K/4K to 128K consists of a light-weight continual pretraining by gradually increasing its RoPE base frequency with repository-level file packing and length-upsampled long-context data. Additionally, we also release instruction-tuned models with long-context support which are derived by further finetuning the long context base models on a mix of permissively licensed short and long-context instruction-response pairs. While comparing to the original short-context Granite code models, our long-context models achieve significant improvements on long-context tasks without any noticeable performance degradation on regular code completion benchmarks (e.g., HumanEval). We release all our long-context Granite code models under an Apache 2.0 license for both research and commercial use."
                    },
                    {
                        "title": "ConMe: Rethinking Evaluation of Compositional Reasoning for Modern VLMs",
                        "abstract": "Compositional Reasoning (CR) entails grasping the significance of attributes, relations, and word order. Recent Vision-Language Models (VLMs), comprising a visual encoder and a Large Language Model (LLM) decoder, have demonstrated remarkable proficiency in such reasoning tasks. This prompts a crucial question: have VLMs effectively tackled the CR challenge? We conjecture that existing CR benchmarks may not adequately push the boundaries of modern VLMs due to the reliance on an LLM-only negative text generation pipeline. Consequently, the negatives produced either appear as outliers from the natural language distribution learned by VLMs' LLM decoders or as improbable within the corresponding image context. To address these limitations, we introduce ConMe -- a compositional reasoning benchmark and a novel data generation pipeline leveraging VLMs to produce `hard CR Q&A'. Through a new concept of VLMs conversing with each other to collaboratively expose their weaknesses, our pipeline autonomously generates, evaluates, and selects challenging compositional reasoning questions, establishing a robust CR benchmark, also subsequently validated manually. Our benchmark provokes a noteworthy, up to 33%, decrease in CR performance compared to preceding benchmarks, reinstating the CR challenge even for state-of-the-art VLMs."
                    },
                    {
                        "title": "Trans-LoRA: towards data-free Transferable Parameter Efficient Finetuning",
                        "abstract": "Low-rank adapters (LoRA) and their variants are popular parameter-efficient fine-tuning (PEFT) techniques that closely match full model fine-tune performance while requiring only a small number of additional parameters. These additional LoRA parameters are specific to the base model being adapted. When the base model needs to be deprecated and replaced with a new one, all the associated LoRA modules need to be re-trained. Such re-training requires access to the data used to train the LoRA for the original base model. This is especially problematic for commercial cloud applications where the LoRA modules and the base models are hosted by service providers who may not be allowed to host proprietary client task data. To address this challenge, we propose $\\textit{Trans-LoRA}$ -- a novel method for lossless, nearly data-free transfer of LoRAs across base models. Our approach relies on synthetic data to transfer LoRA modules. Using large language models, we design a synthetic data generator to approximate the data-generating process of the $\\textit{observed}$ task data subset. Training on the resulting synthetic dataset transfers LoRA modules to new models. We show the effectiveness of our approach using both LLama and Gemma model families. Our approach achieves lossless (mostly improved) LoRA transfer between models within and across different base model families, and even between different PEFT methods, on a wide variety of tasks."
                    },
                    {
                        "title": "Learning Human Action Recognition Representations Without Real Humans",
                        "abstract": "Pre-training on massive video datasets has become essential to achieve high action recognition performance on smaller downstream datasets. However, most large-scale video datasets contain images of people and hence are accompanied with issues related to privacy, ethics, and data protection, often preventing them from being publicly shared for reproducible research. Existing work has attempted to alleviate these problems by blurring faces, downsampling videos, or training on synthetic data. On the other hand, analysis on the transferability of privacy-preserving pre-trained models to downstream tasks has been limited. In this work, we study this problem by first asking the question: can we pre-train models for human action recognition with data that does not include real humans? To this end, we present, for the first time, a benchmark that leverages real-world videos with humans removed and synthetic data containing virtual humans to pre-train a model. We then evaluate the transferability of the representation learned on this data to a diverse set of downstream action recognition benchmarks. Furthermore, we propose a novel pre-training strategy, called Privacy-Preserving MAE-Align, to effectively combine synthetic data and human-removed real data. Our approach outperforms previous baselines by up to 5% and closes the performance gap between human and no-human action recognition representations on downstream tasks, for both linear probing and fine-tuning. Our benchmark, code, and models are available at https://github.com/howardzh01/PPMA ."
                    },
                    {
                        "title": "Constructive Assimilation: Boosting Contrastive Learning Performance through View Generation Strategies",
                        "abstract": "Transformations based on domain expertise (expert transformations), such as random-resized-crop and color-jitter, have proven critical to the success of contrastive learning techniques such as SimCLR. Recently, several attempts have been made to replace such domain-specific, human-designed transformations with generated views that are learned. However for imagery data, so far none of these view-generation methods has been able to outperform expert transformations. In this work, we tackle a different question: instead of replacing expert transformations with generated views, can we constructively assimilate generated views with expert transformations? We answer this question in the affirmative and propose a view generation method and a simple, effective assimilation method that together improve the state-of-the-art by up to ~3.6% on three different datasets. Importantly, we conduct a detailed empirical study that systematically analyzes a range of view generation and assimilation methods and provides a holistic picture of the efficacy of learned views in contrastive representation learning."
                    },
                    {
                        "title": "Self-Specialization: Uncovering Latent Expertise within Large Language Models",
                        "abstract": "Recent works have demonstrated the effectiveness of self-alignment in which a large language model is aligned to follow general instructions using instructional data generated from the model itself starting from a handful of human-written seeds. Instead of general alignment, in this work, we focus on self-alignment for expert domain specialization (e.g., biomedicine, finance). As a preliminary, we quantitively show the marginal effect that generic instruction-following training has on downstream expert domains' performance. To remedy this, we propose self-specialization - allowing for effective model specialization while achieving cross-task generalization by leveraging only a few labeled seeds. Self-specialization offers a data- and parameter-efficient way of\"carving out\"an expert model out of a generalist pre-trained LLM. Exploring a variety of popular open large models as a base for specialization, our experimental results in both biomedical and financial domains show that our self-specialized models outperform their base models by a large margin, and even larger models that are generally instruction-tuned or that have been adapted to the target domain by other means."
                    },
                    {
                        "title": "Comparison of Multilingual Self-Supervised and Weakly-Supervised Speech Pre-Training for Adaptation to Unseen Languages",
                        "abstract": "Recent models such as XLS-R and Whisper have made multilingual speech technologies more accessible by pre-training on audio from around 100 spoken languages each. However, there are thousands of spoken languages worldwide, and adapting to new languages is an important problem. In this work, we aim to understand which model adapts better to languages unseen during pre-training. We fine-tune both models on 13 unseen languages and 18 seen languages. Our results show that the number of hours seen per language and language family during pre-training is predictive of how the models compare, despite the significant differences in the pre-training methods."
                    },
                    {
                        "title": "Dense and Aligned Captions (DAC) Promote Compositional Reasoning in VL Models",
                        "abstract": "Vision and Language (VL) models offer an effective method for aligning representation spaces of images and text, leading to numerous applications such as cross-modal retrieval, visual question answering, captioning, and more. However, the aligned image-text spaces learned by all the popular VL models are still suffering from the so-called `object bias' - their representations behave as `bags of nouns', mostly ignoring or downsizing the attributes, relations, and states of objects described/appearing in texts/images. Although some great attempts at fixing these `compositional reasoning' issues were proposed in the recent literature, the problem is still far from being solved. In this paper, we uncover two factors limiting the VL models' compositional reasoning performance. These two factors are properties of the paired VL dataset used for finetuning and pre-training the VL model: (i) the caption quality, or in other words `image-alignment', of the texts; and (ii) the `density' of the captions in the sense of mentioning all the details appearing on the image. We propose a fine-tuning approach for automatically treating these factors leveraging a standard VL dataset (CC3M). Applied to CLIP, we demonstrate its significant compositional reasoning performance increase of up to $\\sim27\\%$ over the base model, up to $\\sim20\\%$ over the strongest baseline, and by $6.7\\%$ on average."
                    },
                    {
                        "title": "3VL: using Trees to teach Vision&Language models compositional concepts",
                        "abstract": "Vision-Language models (VLMs) have proved effective at aligning image and text representations, producing superior zero-shot results when transferred to many downstream tasks. However, these representations suffer some key shortcomings in Compositional Language Concepts (CLC) understanding such as recognizing objects' attributes, states, and relations between different objects. Moreover, VLMs typically have poor interpretability, making it challenging to debug and mitigate compositional-understanding failures. In this work, we introduce the Tree-augmented Vision-Language (3VL) model architecture and training technique accompanied by our proposed Anchor inference method and Differential Relevance (DiRe) interpretability tool. By expanding the text of an arbitrary image-text pair into a hierarchical tree structure using language analysis tools, 3VL allows inducing this structure into the visual representation learned by the model, enhancing its interpretability and compositional reasoning. Additionally, we show how Anchor, a simple technique for text unification, can be employed to filter nuisance factors while increasing CLC understanding performance, e.g., on the fundamental VL-Checklist benchmark. We also exhibit how DiRe, which performs a differential comparison between VLM relevancy maps, enables us to generate compelling visualizations of the reasons for a model's success or failure."
                    },
                    {
                        "title": "Going Beyond Nouns With Vision & Language Models Using Synthetic Data",
                        "abstract": "Large-scale pre-trained Vision & Language (VL) models have shown remarkable performance in many applications, enabling replacing a fixed set of supported classes with zero-shot open vocabulary reasoning over (almost arbitrary) natural language prompts. However, recent works have uncovered a fundamental weakness of these models. For example, their difficulty to understand Visual Language Concepts (VLC) that go \u2018beyond nouns\u2019 such as the meaning of non-object words (e.g., attributes, actions, relations, states, etc.), or difficulty in performing compositional reasoning such as understanding the significance of the order of the words in a sentence. In this work, we investigate to which extent purely synthetic data could be leveraged to teach these models to overcome such shortcomings without compromising their zero-shot capabilities. We contribute Synthetic Visual Concepts (SyViC) - a million-scale synthetic dataset and data generation codebase allowing to generate additional suitable data to improve VLC understanding and compositional reasoning of VL models. Additionally, we propose a general VL finetuning strategy for effectively leveraging SyViC towards achieving these improvements. Our extensive experiments and ablations on VL-Checklist, Winoground, and ARO benchmarks demonstrate that it is possible to adapt strong pre-trained VL models with synthetic data significantly enhancing their VLC understanding (e.g. by 9.9% on ARO and 4.3% on VL-Checklist) with under 1% drop in their zero-shot accuracy."
                    }
                ]
            },
            "7451d1dc-3356-4852-bd52-ad0af55bf8ad": {
                "pk": "7451d1dc-3356-4852-bd52-ad0af55bf8ad",
                "name": "James Glass",
                "collaborators": [
                    "Andrew Rouditchenko",
                    "Yuan Gong",
                    "Alexander H. Liu",
                    "Samuel Thomas",
                    "Leonid Karlinsky",
                    "Sameer Khurana",
                    "R. Feris",
                    "Hilde Kuehne",
                    "Hildegard Kuehne",
                    "Rog\u00e9rio Feris",
                    "Sung-Lin Yeh",
                    "David F. Harwath",
                    "Brian Kingsbury",
                    "Brian Chen",
                    "Nina Shvetsova",
                    "D. Kondermann",
                    "Shih-Fu Chang",
                    "Heng-Jui Chang"
                ],
                "domain": [
                    "Audio-Visual Speech Recognition",
                    "Self-Supervised Learning",
                    "Multimodal Learning",
                    "Speech Processing"
                ],
                "publications": [
                    {
                        "title": "Whisper-Flamingo: Integrating Visual Features into Whisper for Audio-Visual Speech Recognition and Translation",
                        "abstract": "Audio-Visual Speech Recognition (AVSR) uses lip-based video to improve performance in noise. Since videos are harder to obtain than audio, the video training data of AVSR models is usually limited to a few thousand hours. In contrast, speech models such as Whisper are trained with hundreds of thousands of hours of data, and thus learn a better speech-to-text decoder. The huge training data difference motivates us to adapt Whisper to handle video inputs. Inspired by Flamingo which injects visual features into language models, we propose Whisper-Flamingo which integrates visual features into the Whisper speech recognition and translation model with gated cross attention. Our audio-visual Whisper-Flamingo outperforms audio-only Whisper on English speech recognition and En-X translation for 6 languages in noisy conditions. Moreover, Whisper-Flamingo is a versatile model and conducts all of these tasks using one set of parameters, while prior methods are trained separately on each language."
                    },
                    {
                        "title": "Revisiting Self-supervised Learning of Speech Representation from a Mutual Information Perspective",
                        "abstract": "Existing studies on self-supervised speech representation learning have focused on developing new training methods and applying pre-trained models for different applications. However, the quality of these models is often measured by the performance of different downstream tasks. How well the representations access the information of interest is less studied. In this work, we take a closer look into existing self-supervised methods of speech from an information-theoretic perspective. We aim to develop metrics using mutual information to help practical problems such as model design and selection. We use linear probes to estimate the mutual information between the target information and learned representations, showing another insight into the accessibility to the target information from speech representations. Further, we explore the potential of evaluating representations in a self-supervised fashion, where we estimate the mutual information between different parts of the data without using any labels. Finally, we show that both supervised and unsupervised measures echo the performance of the models on layer-wise linear probing and speech recognition."
                    },
                    {
                        "title": "Comparison of Multilingual Self-Supervised and Weakly-Supervised Speech Pre-Training for Adaptation to Unseen Languages",
                        "abstract": "Recent models such as XLS-R and Whisper have made multilingual speech technologies more accessible by pre-training on audio from around 100 spoken languages each. However, there are thousands of spoken languages worldwide, and adapting to new languages is an important problem. In this work, we aim to understand which model adapts better to languages unseen during pre-training. We fine-tune both models on 13 unseen languages and 18 seen languages. Our results show that the number of hours seen per language and language family during pre-training is predictive of how the models compare, despite the significant differences in the pre-training methods."
                    },
                    {
                        "title": "What, When, and Where? Self-Supervised Spatio- Temporal Grounding in Untrimmed Multi-Action Videos from Narrated Instructions",
                        "abstract": "Spatio-temporal grounding describes the task of localizing events in space and time, e.g., in video data, based on verbal descriptions only. Models for this task are usually trained with human-annotated sentences and bounding box supervision. This work addresses this task from a multimodal supervision perspective, proposing a framework for spatio-temporal action grounding trained on loose video and subtitle supervision only, without human annotation. To this end, we combine local representation learning, which focuses on leveraging fine-grained spatial information, with a global representation encoding that captures higher-level representations and incorporates both in a joint approach. To evaluate this challenging task in a real-life setting, a new benchmark dataset is proposed, providing dense spatio-temporal grounding annotations in long, untrimmed, multi-action instructional videos for over 5K events. We evaluate the proposed approach and other methods on the proposed and standard downstream tasks, showing that our method improves over current baselines in various settings, including spatial, temporal, and untrimmed multi-action spatio-temporal grounding."
                    },
                    {
                        "title": "Whisper-AT: Noise-Robust Automatic Speech Recognizers are Also Strong General Audio Event Taggers",
                        "abstract": "In this paper, we focus on Whisper, a recent automatic speech recognition model trained with a massive 680k hour labeled speech corpus recorded in diverse conditions. We first show an interesting finding that while Whisper is very robust against real-world background sounds (e.g., music), its audio representation is actually not noise-invariant, but is instead highly correlated to non-speech sounds, indicating that Whisper recognizes speech conditioned on the noise type. With this finding, we build a unified audio tagging and speech recognition model Whisper-AT by freezing the backbone of Whisper, and training a lightweight audio tagging model on top of it. With<1% extra computational cost, Whisper-AT can recognize audio events, in addition to spoken text, in a single forward pass."
                    },
                    {
                        "title": "Self-supervised Fine-tuning for Improved Content Representations by Speaker-invariant Clustering",
                        "abstract": "Self-supervised speech representation models have succeeded in various tasks, but improving them for content-related problems using unlabeled data is challenging. We propose speaker-invariant clustering (Spin), a novel self-supervised learning method that clusters speech representations and performs swapped prediction between the original and speaker-perturbed utterances. Spin disentangles speaker information and preserves content representations with just 45 minutes of fine-tuning on a single GPU. Spin improves pre-trained networks and outperforms prior methods in speech recognition and acoustic unit discovery."
                    },
                    {
                        "title": "UAVM: A Unified Model for Audio-Visual Learning",
                        "abstract": "\u2014Conventional audio-visual models have independent audio and video branches. We design a uni\ufb01ed model for audio and video processing called Uni\ufb01ed Audio-Visual Model (UAVM). In this paper, we describe UAVM, report its new state-of-the-art audio-visual event classi\ufb01cation accuracy of 65.8% on VGGSound, and describe the intriguing properties of the model."
                    },
                    {
                        "title": "UAVM: Towards Unifying Audio and Visual Models",
                        "abstract": "Conventional audio-visual models have independent audio and video branches. In this work, we <italic>unify</italic> the audio and visual branches by designing a <underline>U</underline>nified <underline>A</underline>udio-<underline>V</underline>isual <underline>M</underline>odel (UAVM). The UAVM achieves a new state-of-the-art audio-visual event classification accuracy of 65.8% on VGGSound. More interestingly, we also find a few intriguing properties of UAVM that the modality-independent counterparts do not have. Code at <uri>github.com/yuangongnd/uavm</uri>."
                    },
                    {
                        "title": "Qualitative analysis: viewing, consumption and usability of svod",
                        "abstract": "Qualitative Analysis: Viewing, Consumption and Usability of SVOD During the last twenty years, media industries negotiated and navigated epic shifts in distribution and screen technologies that have had implications for all aspects of producing and viewing television globally. This thesis examines Subscription-Video-On-Demand (SVOD) service using Netflix as one of the main examples that has seen a dramatic increase in recent years compared to the traditional way of viewing content through television. Netflix is a global internet television network provider, offering TV and movie content with unlimited viewing on any Internet-connected screen plus a no-commitment monthly fee. Using the background of SVOD within the context of the media landscape, the digital age is introduced by defining the reality of it. Also, discussed is the shift by users in respect to the Internet and self-scheduling of viewing content as well as characteristics of digital storytelling. SVOD, using Netflix as an example, is examined by paying attention to its development and service. With the aid of focus group sessions and an online survey, a qualitative analysis is utilised to elaborate and gain insight into the phenomenon of SVOD usability, technology and binge-watching of content.  The combination SVOD sites and technology has begun to change the way television has been encompassed within our society. Television is no longer seen as solely anchored to the home and can be viewed as a portable medium, anytime, anywhere. The focus group is one qualitative research method that can be used to explore such issues and that is why it has been selected for my thesis research. By gathering people into focus groups and creating environmental conditions for more spontaneous expression and facilitating the interaction of everybody. Watching online film and television over the internet, forces the breakdown of the traditional value chain and viewing methods as well as consumption of it. The structure and role of SVOD within our society, has allowed viewers of online television programmes to alter time shifting and self-scheduling habits. Netflix has fast become popular around the globe for its simple approach to providing SVOD content. The phenomenon of binge-watching television programmes and movies has transcended from DVD's. Netflix has aggressively pushed the limitations of viewing to new frontiers. SVOD customers stream hundreds of millions of hours of online video each month, conversely, TV programme fans take to the social media-verse to chat about TV shows or movies.  This research aims to narrow this research gap surrounding SVOD. The results will be used to add value to the research of this growing and changing phenomenon of online TV consumption with SVOD. By way of qualitative research methods using a thematic analysis with such factors as: Inductive, Theory, Hypothesis, Observation and Confirmation. The theoretical framework has been based on organising key issues in data and grouped under themes highlighting important relations within the research questions. Applying a thematic analysis was vital as it served to frame key topics that involved specific descriptions in relation to the research question. Thus, by using the principles of the thematic analysis we were able to analyse the vast and complex nature of SVOD while answering the initial objectives. Using core qualitative aspects, valuable data can be extracted from the data which was necessary in order to achieve a valid outcome of discovery for the questions sought. Thematic analysis is an approach to use that allows researchers to identify emergent topics not explicitly stated in survey questions carried out."
                    }
                ]
            }
        }
    },
    "2403.11391": {
        "paper_data": {
            "title": "Investigating the Benefits of Projection Head for Representation Learning",
            "url": "http://arxiv.org/abs/2403.11391v1",
            "arxiv_id": "2403.11391",
            "authors": [
                "Yihao Xue",
                "Eric Gan",
                "Jiayi Ni",
                "Siddharth Joshi",
                "Baharan Mirzasoleiman"
            ],
            "abstract": "An effective technique for obtaining high-quality representations is adding a projection head on top of the encoder during training, then discarding it and using the pre-projection representations. Despite its proven practical effectiveness, the reason behind the success of this technique is poorly understood. The pre-projection representations are not directly optimized by the loss function, raising the question: what makes them better? In this work, we provide a rigorous theoretical answer to this question. We start by examining linear models trained with self-supervised contrastive loss. We reveal that the implicit bias of training algorithms leads to layer-wise progressive feature weighting, where features become increasingly unequal as we go deeper into the layers. Consequently, lower layers tend to have more normalized and less specialized representations. We theoretically characterize scenarios where such representations are more beneficial, highlighting the intricate interplay between data augmentation and input features. Additionally, we demonstrate that introducing non-linearity into the network allows lower layers to learn features that are completely absent in higher layers. Finally, we show how this mechanism improves the robustness in supervised contrastive learning and supervised learning. We empirically validate our results through various experiments on CIFAR-10/100, UrbanCars and shifted versions of ImageNet. We also introduce a potential alternative to projection head, which offers a more interpretable and controllable design.",
            "introduction": "   1 Introduction  Representation learning has been the subject of extensive study in the last decade (Chen et\u00a0al., 2020; Zbontar et\u00a0al., 2021; Khosla et\u00a0al., 2020; Ermolov et\u00a0al., 2021). Despite the great progress, learning representations that generalize well across various domains has remained challenging. Among the existing techniques, contrastive self-supervised learning has gained a lot of attention, due to its ability to learn robust representations that generalize better across various domains. Contrastive learning (CL) learns representations by maximizing the agreement between different augmented views of the same examples and minimizing that of different examples (Chen et\u00a0al., 2020). Pre-training with CL is often essential before transfering models to new or shifted domains (Hendrycks et\u00a0al., 2019).   A key factor that enables contrastive learning to learn robust representations is projection head, a shallow MLP that is used during pretraining and is discarded afterwards (Chen et\u00a0al., 2020), which has become a standard for learning high-quality representations (He et\u00a0al., 2020; Yeh et\u00a0al., 2022; Bardes et\u00a0al., 2022; Chuang et\u00a0al., 2020; Grill et\u00a0al., 2020; Garrido et\u00a0al., 2022; Zbontar et\u00a0al., 2021; Ermolov et\u00a0al., 2021; Joshi & Mirzasoleiman, 2023). It is particularly advantageous when there is a misalignment between training and downstream objectives (Bordes et\u00a0al., 2023). The benefit of projection head extends beyond self-supervised learning to other representation learning methods, including supervised contrastive learning (Khosla et\u00a0al., 2020). However, the mechanism by which the projection head improves the generalizability and robustness of representations is poorly understood.   Theoretically analyzing the effect of projection head for CL is challenging, as one needs to understand feature learning both before and after the projection head and compare them. A few recent studies (Tian et\u00a0al., 2021; Wang et\u00a0al., 2021; Wen & Li, 2022) have focused on non-contrastive self-supervised methods and demonstrated that the projection head can mitigate the dimensional collapse problem, where all inputs are mapped to the same representation. However, such results do not extend to contrastive methods due to its inherently different learning mechanism. Besides, they do not provide an understanding of the projection head\u2019s role in enhancing robustness under misalignment between pretraining and downstream objectives.   In this work, we analyze the effect of projection head on the quality and robustness of representations learned by CL, and extend our results to supervised CL (SCL) and supervised learning (SL). First, we theoretically show that linear models progressively assign weights to features as they operate through the layers. Thus, in deeper layers the features are represented more unequally and the representations are more specialized toward the pretraining objective. Moreover, non-linear activations exacerbate this effect allowing lower layers to learn features that are entirely absent in the outputs of the projection head. We demonstrate that projection head provably improves the robustness and generalizability of the representations, when data augmentation harms useful features of the pretraining data, or when features relevant to the downstream task are too weak or too strong in the pretraining data. Finally, we extend our results beyond CL to SCL and SL. In this setting, we reveal that lower layers can learn subclass-level features that are not represented in the final layer, demonstrating how representations before the final representation layer can significantly reduce class/neural collapse, a problem previously observed in the final representations.    We conduct extensive experiments to confirm our theoretical",
            "references": [
                {
                    "title": "Unraveling Projection Heads in Contrastive Learning: Insights from Expansion and Shrinkage",
                    "abstract": "We investigate the role of projection heads, also known as projectors, within the encoder-projector framework (e.g., SimCLR) used in contrastive learning. We aim to demystify the observed phenomenon where representations learned before projectors outperform those learned after -- measured using the downstream linear classification accuracy, even when the projectors themselves are linear. In this paper, we make two significant contributions towards this aim. Firstly, through empirical and theoretical analysis, we identify two crucial effects -- expansion and shrinkage -- induced by the contrastive loss on the projectors. In essence, contrastive loss either expands or shrinks the signal direction in the representations learned by an encoder, depending on factors such as the augmentation strength, the temperature used in contrastive loss, etc. Secondly, drawing inspiration from the expansion and shrinkage phenomenon, we propose a family of linear transformations to accurately model the projector's behavior. This enables us to precisely characterize the downstream linear classification accuracy in the high-dimensional asymptotic limit. Our findings reveal that linear projectors operating in the shrinkage (or expansion) regime hinder (or improve) the downstream classification accuracy. This provides the first theoretical explanation as to why (linear) projectors impact the downstream performance of learned representations. Our theoretical findings are further corroborated by extensive experiments on both synthetic data and real image data."
                },
                {
                    "title": "Which Features are Learnt by Contrastive Learning? On the Role of Simplicity Bias in Class Collapse and Feature Suppression",
                    "abstract": "Contrastive learning (CL) has emerged as a powerful technique for representation learning, with or without label supervision. However, supervised CL is prone to collapsing representations of subclasses within a class by not capturing all their features, and unsupervised CL may suppress harder class-relevant features by focusing on learning easy class-irrelevant features; both significantly compromise representation quality. Yet, there is no theoretical understanding of \\textit{class collapse} or \\textit{feature suppression} at \\textit{test} time. We provide the first unified theoretically rigorous framework to determine \\textit{which} features are learnt by CL. Our analysis indicate that, perhaps surprisingly, bias of (stochastic) gradient descent towards finding simpler solutions is a key factor in collapsing subclass representations and suppressing harder class-relevant features. Moreover, we present increasing embedding dimensionality and improving the quality of data augmentations as two theoretically motivated solutions to {feature suppression}. We also provide the first theoretical explanation for why employing supervised and unsupervised CL together yields higher-quality representations, even when using commonly-used stochastic gradient methods."
                },
                {
                    "title": "Data-Efficient Contrastive Self-supervised Learning: Most Beneficial Examples for Supervised Learning Contribute the Least",
                    "abstract": "Self-supervised learning (SSL) learns high-quality representations from large pools of unlabeled training data. As datasets grow larger, it becomes crucial to identify the examples that contribute the most to learning such representations. This enables efficient SSL by reducing the volume of data required. Nevertheless, quantifying the value of examples for SSL has remained an open question. In this work, we address this problem for the first time, by proving that examples that contribute the most to contrastive SSL are those that have the most similar augmentations to other examples, in expectation. We provide rigorous guarantees for the generalization performance of contrastive learning on such subsets. Through extensive experiments, we show that we can safely exclude 20% of examples from CIFAR100 and 40% from STL10 and TinyImageNet, without affecting downstream task performance. In general, subsets selected by our method outperform random subsets by over 3% across these datasets. Interestingly, we also discover the subsets that contribute the most to contrastive learning are those that contribute the least to supervised learning. Code available at https://github.com/bigml-cs-ucla/sas-data-efficient-contrastive-learning."
                },
                {
                    "title": "Understanding Multimodal Contrastive Learning and Incorporating Unpaired Data",
                    "abstract": "Language-supervised vision models have recently attracted great attention in computer vision. A common approach to build such models is to use contrastive learning on paired data across the two modalities, as exemplified by Contrastive Language-Image Pre-Training (CLIP). In this paper, under linear representation settings, (i) we initiate the investigation of a general class of nonlinear loss functions for multimodal contrastive learning (MMCL) including CLIP loss and show its connection to singular value decomposition (SVD). Namely, we show that each step of loss minimization by gradient descent can be seen as performing SVD on a contrastive cross-covariance matrix. Based on this insight, (ii) we analyze the performance of MMCL. We quantitatively show that the feature learning ability of MMCL can be better than that of unimodal contrastive learning applied to each modality even under the presence of wrongly matched pairs. This characterizes the robustness of MMCL to noisy data. Furthermore, when we have access to additional unpaired data, (iii) we propose a new MMCL loss that incorporates additional unpaired datasets. We show that the algorithm can detect the ground-truth pairs and improve performance by fully exploiting unpaired datasets. The performance of the proposed algorithm was verified by numerical experiments."
                },
                {
                    "title": "You Only Need a Good Embeddings Extractor to Fix Spurious Correlations",
                    "abstract": "Spurious correlations in training data often lead to robustness issues since models learn to use them as shortcuts. For example, when predicting whether an object is a cow, a model might learn to rely on its green background, so it would do poorly on a cow on a sandy background. A standard dataset for measuring state-of-the-art on methods mitigating this problem is Waterbirds. The best method (Group Distributionally Robust Optimization - GroupDRO) currently achieves 89\\% worst group accuracy and standard training from scratch on raw images only gets 72\\%. GroupDRO requires training a model in an end-to-end manner with subgroup labels. In this paper, we show that we can achieve up to 90\\% accuracy without using any sub-group information in the training set by simply using embeddings from a large pre-trained vision model extractor and training a linear classifier on top of it. With experiments on a wide range of pre-trained models and pre-training datasets, we show that the capacity of the pre-training model and the size of the pre-training dataset matters. Our experiments reveal that high capacity vision transformers perform better compared to high capacity convolutional neural networks, and larger pre-training dataset leads to better worst-group accuracy on the spurious correlation dataset."
                },
                {
                    "title": "A Whac-A-Mole Dilemma: Shortcuts Come in Multiples Where Mitigating One Amplifies Others",
                    "abstract": "Machine learning models have been found to learn shortcuts\u2014unintended decision rules that are unable to generalize\u2014undermining models' reliability. Previous works address this problem under the tenuous assumption that only a single shortcut exists in the training data. Real-world images are rife with multiple visual cues from background to texture. Key to advancing the reliability of vision systems is understanding whether existing methods can overcome multiple shortcuts or struggle in a Whac-A-Mole game, i.e., where mitigating one shortcut amplifies reliance on others. To address this shortcoming, we propose two benchmarks: 1) UrbanCars, a dataset with precisely controlled spurious cues, and 2) ImageNet-W, an evaluation set based on ImageNet for watermark, a shortcut we discovered affects nearly every modern vision model. Along with texture and background, ImageNet-W allows us to study multiple shortcuts emerging from training on natural images. We find computer vision models, including large foundation models\u2014regardless of training set, architecture, and supervision\u2014struggle when multiple shortcuts are present. Even methods explicitly designed to combat shortcuts struggle in a Whac-A-Mole dilemma. To tackle this challenge, we propose Last Layer Ensemble, a simple-yet-effective method to mitigate multiple shortcuts without Whac-A-Mole behavior. Our results surface multi-shortcut mitigation as an overlooked challenge critical to advancing the reliability of vision systems. The datasets and code are released: https://github.com/facebookresearch/Whac-A-Mole."
                },
                {
                    "title": "A Theoretical Study of Inductive Biases in Contrastive Learning",
                    "abstract": "Understanding self-supervised learning is important but challenging. Previous theoretical works study the role of pretraining losses, and view neural networks as general black boxes. However, the recent work of Saunshi et al. argues that the model architecture -- a component largely ignored by previous works -- also has significant influences on the downstream performance of self-supervised learning. In this work, we provide the first theoretical analysis of self-supervised learning that incorporates the effect of inductive biases originating from the model class. In particular, we focus on contrastive learning -- a popular self-supervised learning method that is widely used in the vision domain. We show that when the model has limited capacity, contrastive representations would recover certain special clustering structures that are compatible with the model architecture, but ignore many other clustering structures in the data distribution. As a result, our theory can capture the more realistic setting where contrastive representations have much lower dimensionality than the number of clusters in the data distribution. We instantiate our theory on several synthetic data distributions, and provide empirical evidence to support the theory."
                },
                {
                    "title": "Are All Losses Created Equal: A Neural Collapse Perspective",
                    "abstract": "While cross entropy (CE) is the most commonly used loss to train deep neural networks for classification tasks, many alternative losses have been developed to obtain better empirical performance. Among them, which one is the best to use is still a mystery, because there seem to be multiple factors affecting the answer, such as properties of the dataset, the choice of network architecture, and so on. This paper studies the choice of loss function by examining the last-layer features of deep networks, drawing inspiration from a recent line work showing that the global optimal solution of CE and mean-square-error (MSE) losses exhibits a Neural Collapse phenomenon. That is, for sufficiently large networks trained until convergence, (i) all features of the same class collapse to the corresponding class mean and (ii) the means associated with different classes are in a configuration where their pairwise distances are all equal and maximized. We extend such results and show through global solution and landscape analyses that a broad family of loss functions including commonly used label smoothing (LS) and focal loss (FL) exhibits Neural Collapse. Hence, all relevant losses(i.e., CE, LS, FL, MSE) produce equivalent features on training data. Based on the unconstrained feature model assumption, we provide either the global landscape analysis for LS loss or the local landscape analysis for FL loss and show that the (only!) global minimizers are neural collapse solutions, while all other critical points are strict saddles whose Hessian exhibit negative curvature directions either in the global scope for LS loss or in the local scope for FL loss near the optimal solution. The experiments further show that Neural Collapse features obtained from all relevant losses lead to largely identical performance on test data as well, provided that the network is sufficiently large and trained until convergence."
                },
                {
                    "title": "Guillotine Regularization: Why removing layers is needed to improve generalization in Self-Supervised Learning",
                    "abstract": "One unexpected technique that emerged in recent years consists in training a Deep Network (DN) with a Self-Supervised Learning (SSL) method, and using this network on downstream tasks but with its last few projector layers entirely removed. This trick of throwing away the projector is actually critical for SSL methods to display competitive performances on ImageNet for which more than 30 percentage points can be gained that way. This is a little vexing, as one would hope that the network layer at which invariance is explicitly enforced by the SSL criterion during training (the last projector layer) should be the one to use for best generalization performance downstream. But it seems not to be, and this study sheds some light on why. This trick, which we name Guillotine Regularization (GR), is in fact a generically applicable method that has been used to improve generalization performance in transfer learning scenarios. In this work, we identify the underlying reasons behind its success and show that the optimal layer to use might change significantly depending on the training setup, the data or the downstream task. Lastly, we give some insights on how to reduce the need for a projector in SSL by aligning the pretext SSL task and the downstream task."
                },
                {
                    "title": "On the duality between contrastive and non-contrastive self-supervised learning",
                    "abstract": "Recent approaches in self-supervised learning of image representations can be categorized into different families of methods and, in particular, can be divided into contrastive and non-contrastive approaches. While differences between the two families have been thoroughly discussed to motivate new approaches, we focus more on the theoretical similarities between them. By designing contrastive and covariance based non-contrastive criteria that can be related algebraically and shown to be equivalent under limited assumptions, we show how close those families can be. We further study popular methods and introduce variations of them, allowing us to relate this theoretical result to current practices and show the influence (or lack thereof) of design choices on downstream performance. Motivated by our equivalence result, we investigate the low performance of SimCLR and show how it can match VICReg's with careful hyperparameter tuning, improving significantly over known baselines. We also challenge the popular assumption that non-contrastive methods need large output dimensions. Our theoretical and quantitative results suggest that the numerical gaps between contrastive and non-contrastive methods in certain regimes can be closed given better network design choices and hyperparameter tuning. The evidence shows that unifying different SOTA methods is an important direction to build a better understanding of self-supervised learning."
                },
                {
                    "title": "The Mechanism of Prediction Head in Non-contrastive Self-supervised Learning",
                    "abstract": "Recently the surprising discovery of the Bootstrap Your Own Latent (BYOL) method by Grill et al. shows the negative term in contrastive loss can be removed if we add the so-called prediction head to the network. This initiated the research of non-contrastive self-supervised learning. It is mysterious why even when there exist trivial collapsed global optimal solutions, neural networks trained by (stochastic) gradient descent can still learn competitive representations. This phenomenon is a typical example of implicit bias in deep learning and remains little understood. In this work, we present our empirical and theoretical discoveries on non-contrastive self-supervised learning. Empirically, we find that when the prediction head is initialized as an identity matrix with only its off-diagonal entries being trainable, the network can learn competitive representations even though the trivial optima still exist in the training objective. Theoretically, we present a framework to understand the behavior of the trainable, but identity-initialized prediction head. Under a simple setting, we characterized the substitution effect and acceleration effect of the prediction head. The substitution effect happens when learning the stronger features in some neurons can substitute for learning these features in other neurons through updating the prediction head. And the acceleration effect happens when the substituted features can accelerate the learning of other weaker features to prevent them from being ignored. These two effects enable the neural networks to learn all the features rather than focus only on learning the stronger features, which is likely the cause of the dimensional collapse phenomenon. To the best of our knowledge, this is also the first end-to-end optimization guarantee for non-contrastive methods using nonlinear neural networks with a trainable prediction head and normalization."
                },
                {
                    "title": "Perfectly Balanced: Improving Transfer and Robustness of Supervised Contrastive Learning",
                    "abstract": "An ideal learned representation should display transferability and robustness. Supervised contrastive learning (SupCon) is a promising method for training accurate models, but produces representations that do not capture these properties due to class collapse -- when all points in a class map to the same representation. Recent work suggests that\"spreading out\"these representations improves them, but the precise mechanism is poorly understood. We argue that creating spread alone is insufficient for better representations, since spread is invariant to permutations within classes. Instead, both the correct degree of spread and a mechanism for breaking this invariance are necessary. We first prove that adding a weighted class-conditional InfoNCE loss to SupCon controls the degree of spread. Next, we study three mechanisms to break permutation invariance: using a constrained encoder, adding a class-conditional autoencoder, and using data augmentation. We show that the latter two encourage clustering of latent subclasses under more realistic conditions than the former. Using these insights, we show that adding a properly-weighted class-conditional InfoNCE loss and a class-conditional autoencoder to SupCon achieves 11.1 points of lift on coarse-to-fine transfer across 5 standard datasets and 4.7 points on worst-group robustness on 3 datasets, setting state-of-the-art on CelebA by 11.5 points."
                },
                {
                    "title": "Last Layer Re-Training is Sufficient for Robustness to Spurious Correlations",
                    "abstract": "Neural network classifiers can largely rely on simple spurious features, such as backgrounds, to make predictions. However, even in these cases, we show that they still often learn core features associated with the desired attributes of the data, contrary to recent findings. Inspired by this insight, we demonstrate that simple last layer retraining can match or outperform state-of-the-art approaches on spurious correlation benchmarks, but with profoundly lower complexity and computational expenses. Moreover, we show that last layer retraining on large ImageNet-trained models can also significantly reduce reliance on background and texture information, improving robustness to covariate shift, after only minutes of training on a single GPU."
                },
                {
                    "title": "On the Optimization Landscape of Neural Collapse under MSE Loss: Global Optimality with Unconstrained Features",
                    "abstract": "When training deep neural networks for classification tasks, an intriguing empirical phenomenon has been widely observed in the last-layer classifiers and features, where (i) the class means and the last-layer classifiers all collapse to the vertices of a Simplex Equiangular Tight Frame (ETF) up to scaling, and (ii) cross-example within-class variability of last-layer activations collapses to zero. This phenomenon is called Neural Collapse (NC), which seems to take place regardless of the choice of loss functions. In this work, we justify NC under the mean squared error (MSE) loss, where recent empirical evidence shows that it performs comparably or even better than the de-facto cross-entropy loss. Under a simplified unconstrained feature model, we provide the first global landscape analysis for vanilla nonconvex MSE loss and show that the (only!) global minimizers are neural collapse solutions, while all other critical points are strict saddles whose Hessian exhibit negative curvature directions. Furthermore, we justify the usage of rescaled MSE loss by probing the optimization landscape around the NC solutions, showing that the landscape can be improved by tuning the rescaling hyperparameters. Finally, our theoretical findings are experimentally verified on practical network architectures."
                },
                {
                    "title": "Understanding Contrastive Learning Requires Incorporating Inductive Biases",
                    "abstract": "Contrastive learning is a popular form of self-supervised learning that encourages augmentations (views) of the same input to have more similar representations compared to augmentations of different inputs. Recent attempts to theoretically explain the success of contrastive learning on downstream classification tasks prove guarantees depending on properties of {\\em augmentations} and the value of {\\em contrastive loss} of representations. We demonstrate that such analyses, that ignore {\\em inductive biases} of the function class and training algorithm, cannot adequately explain the success of contrastive learning, even {\\em provably} leading to vacuous guarantees in some settings. Extensive experiments on image and text domains highlight the ubiquity of this problem -- different function classes and algorithms behave very differently on downstream tasks, despite having the same augmentations and contrastive losses. Theoretical analysis is presented for the class of linear representations, where incorporating inductive biases of the function class allows contrastive learning to work with less stringent conditions compared to prior analyses."
                },
                {
                    "title": "Limitations of Neural Collapse for Understanding Generalization in Deep Learning",
                    "abstract": "The recent work of Papyan, Han,&Donoho (2020) presented an intriguing\"Neural Collapse\"phenomenon, showing a structural property of interpolating classifiers in the late stage of training. This opened a rich area of exploration studying this phenomenon. Our motivation is to study the upper limits of this research program: How far will understanding Neural Collapse take us in understanding deep learning? First, we investigate its role in generalization. We refine the Neural Collapse conjecture into two separate conjectures: collapse on the train set (an optimization property) and collapse on the test distribution (a generalization property). We find that while Neural Collapse often occurs on the train set, it does not occur on the test set. We thus conclude that Neural Collapse is primarily an optimization phenomenon, with as-yet-unclear connections to generalization. Second, we investigate the role of Neural Collapse in feature learning. We show simple, realistic experiments where training longer leads to worse last-layer features, as measured by transfer-performance on a downstream task. This suggests that neural collapse is not always desirable for representation learning, as previously claimed. Finally, we give preliminary evidence of a\"cascading collapse\"phenomenon, wherein some form of Neural Collapse occurs not only for the last layer, but in earlier layers as well. We hope our work encourages the community to continue the rich line of Neural Collapse research, while also considering its inherent limitations."
                },
                {
                    "title": "Domain-Adjusted Regression or: ERM May Already Learn Features Sufficient for Out-of-Distribution Generalization",
                    "abstract": "A common explanation for the failure of deep networks to generalize out-of-distribution is that they fail to recover the \u201ccorrect\u201d features. We challenge this notion with a simple experiment which suggests that ERM already learns suf\ufb01cient features and that the current bottleneck is not feature learning, but robust regression . Our \ufb01ndings also imply that given a small amount of data from the target distribution, retraining only the last linear layer will give excellent performance. We therefore argue that devising simpler methods for learning predictors on existing features is a promising direction for future research. Towards this end, we introduce Domain-Adjusted Regression (DARE), a convex objective for learning a linear predictor that is provably robust under a new model of distribution shift. Rather than learning one function, DARE performs a domain-speci\ufb01c adjustment to unify the domains in a canonical latent space and learns to predict in this space. Under a natural model, we prove that the DARE solution is the minimax-optimal predictor for a constrained set of test distributions. Further, we provide the \ufb01rst \ufb01nite-environment convergence guarantee to the minimax risk, improving over existing analyses which only yield minimax predictors after an environment threshold. Evaluated on \ufb01netuned features, we \ufb01nd that DARE compares favorably to prior methods, consistently achieving equal or better performance."
                },
                {
                    "title": "Investigating Why Contrastive Learning Benefits Robustness Against Label Noise",
                    "abstract": "Self-supervised Contrastive Learning (CL) has been recently shown to be very effective in preventing deep networks from overfitting noisy labels. Despite its empirical success, the theoretical understanding of the effect of contrastive learning on boosting robustness is very limited. In this work, we rigorously prove that the representation matrix learned by contrastive learning boosts robustness, by having: (i) one prominent singular value corresponding to each sub-class in the data, and significantly smaller remaining singular values; and (ii) {a large alignment between the prominent singular vectors and the clean labels of each sub-class. The above properties enable a linear layer trained on such representations to effectively learn the clean labels without overfitting the noise.} We further show that the low-rank structure of the Jacobian of deep networks pre-trained with contrastive learning allows them to achieve a superior performance initially, when fine-tuned on noisy labels. Finally, we demonstrate that the initial robustness provided by contrastive learning enables robust training methods to achieve state-of-the-art performance under extreme noise levels, e.g., an average of 27.18\\% and 15.58\\% increase in accuracy on CIFAR-10 and CIFAR-100 with 80\\% symmetric noisy labels, and 4.11\\% increase in accuracy on WebVision."
                },
                {
                    "title": "Understanding Dimensional Collapse in Contrastive Self-supervised Learning",
                    "abstract": "Self-supervised visual representation learning aims to learn useful representations without relying on human annotations. Joint embedding approach bases on maximizing the agreement between embedding vectors from different views of the same image. Various methods have been proposed to solve the collapsing problem where all embedding vectors collapse to a trivial constant solution. Among these methods, contrastive learning prevents collapse via negative sample pairs. It has been shown that non-contrastive methods suffer from a lesser collapse problem of a different nature: dimensional collapse, whereby the embedding vectors end up spanning a lower-dimensional subspace instead of the entire available embedding space. Here, we show that dimensional collapse also happens in contrastive learning. In this paper, we shed light on the dynamics at play in contrastive learning that leads to dimensional collapse. Inspired by our theory, we propose a novel contrastive learning method, called DirectCLR, which directly optimizes the representation space without relying on an explicit trainable projector. Experiments show that DirectCLR outperforms SimCLR with a trainable linear projector on ImageNet."
                },
                {
                    "title": "Self-supervised Learning is More Robust to Dataset Imbalance",
                    "abstract": "Self-supervised learning (SSL) is a scalable way to learn general visual representations since it learns without labels. However, large-scale unlabeled datasets in the wild often have long-tailed label distributions, where we know little about the behavior of SSL. In this work, we systematically investigate self-supervised learning under dataset imbalance. First, we find out via extensive experiments that off-the-shelf self-supervised representations are already more robust to class imbalance than supervised representations. The performance gap between balanced and imbalanced pre-training with SSL is significantly smaller than the gap with supervised learning, across sample sizes, for both in-domain and, especially, out-of-domain evaluation. Second, towards understanding the robustness of SSL, we hypothesize that SSL learns richer features from frequent data: it may learn label-irrelevant-but-transferable features that help classify the rare classes and downstream tasks. In contrast, supervised learning has no incentive to learn features irrelevant to the labels from frequent examples. We validate this hypothesis with semi-synthetic experiments and theoretical analyses on a simplified setting. Third, inspired by the theoretical insights, we devise a re-weighted regularization technique that consistently improves the SSL representation quality on imbalanced datasets with several evaluation criteria, closing the small gap between balanced and imbalanced datasets with the same number of examples."
                },
                {
                    "title": "Towards Demystifying Representation Learning with Non-contrastive Self-supervision",
                    "abstract": "Non-contrastive methods of self-supervised learning (such as BYOL and SimSiam) learn representations by minimizing the distance between two views of the same image. These approaches have achieved remarkable performance in practice, but the theoretical understanding lags behind. Tian et al. 2021 explained why the representation does not collapse to zero, however, how the feature is learned still remains mysterious. In our work, we prove in a linear network, non-contrastive methods learn a desirable projection matrix and also reduce the sample complexity on downstream tasks. Our analysis suggests that weight decay acts as an implicit threshold that discards the features with high variance under data augmentations, and keeps the features with low variance. Inspired by our theory, we design a simpler and more computationally efficient algorithm DirectCopy by removing the eigen-decomposition step in the original DirectPred algorithm in Tian et al. 2021. Our experiments show that DirectCopy rivals or even outperforms DirectPred on STL-10, CIFAR-10, CIFAR-100, and ImageNet."
                },
                {
                    "title": "The Power of Contrast for Feature Learning: A Theoretical Analysis",
                    "abstract": "Contrastive learning has achieved state-of-the-art performance in various self-supervised learning tasks and even outperforms its supervised counterpart. Despite its empirical success, theoretical understanding of the superiority of contrastive learning is still limited. In this paper, under linear representation settings, (i) we provably show that contrastive learning outperforms the standard autoencoders and generative adversarial networks, two classical generative unsupervised learning methods, for both feature recovery and in-domain downstream tasks; (ii) we also illustrate the impact of labeled data in supervised contrastive learning. This provides theoretical support for recent findings that contrastive learning with labels improves the performance of learned representations in the in-domain downstream task, but it can harm the performance in transfer learning. We verify our theory with numerical experiments."
                },
                {
                    "title": "Provable Guarantees for Self-Supervised Deep Learning with Spectral Contrastive Loss",
                    "abstract": "Recent works in self-supervised learning have advanced the state-of-the-art by relying on the contrastive learning paradigm, which learns representations by pushing positive pairs, or similar examples from the same class, closer together while keeping negative pairs far apart. Despite the empirical successes, theoretical foundations are limited -- prior analyses assume conditional independence of the positive pairs given the same class label, but recent empirical applications use heavily correlated positive pairs (i.e., data augmentations of the same image). Our work analyzes contrastive learning without assuming conditional independence of positive pairs using a novel concept of the augmentation graph on data. Edges in this graph connect augmentations of the same data, and ground-truth classes naturally form connected sub-graphs. We propose a loss that performs spectral decomposition on the population augmentation graph and can be succinctly written as a contrastive learning objective on neural net representations. Minimizing this objective leads to features with provable accuracy guarantees under linear probe evaluation. By standard generalization bounds, these accuracy guarantees also hold when minimizing the training contrastive loss. Empirically, the features learned by our objective can match or outperform several strong baselines on benchmark vision datasets. In all, this work provides the first provable analysis for contrastive learning where guarantees for linear probe evaluation can apply to realistic empirical settings."
                },
                {
                    "title": "Neural Collapse Under MSE Loss: Proximity to and Dynamics on the Central Path",
                    "abstract": "The recently discovered Neural Collapse (NC) phenomenon occurs pervasively in today's deep net training paradigm of driving cross-entropy (CE) loss towards zero. During NC, last-layer features collapse to their class-means, both classifiers and class-means collapse to the same Simplex Equiangular Tight Frame, and classifier behavior collapses to the nearest-class-mean decision rule. Recent works demonstrated that deep nets trained with mean squared error (MSE) loss perform comparably to those trained with CE. As a preliminary, we empirically establish that NC emerges in such MSE-trained deep nets as well through experiments on three canonical networks and five benchmark datasets. We provide, in a Google Colab notebook, PyTorch code for reproducing MSE-NC and CE-NC: at https://colab.research.google.com/github/neuralcollapse/neuralcollapse/blob/main/neuralcollapse.ipynb. The analytically-tractable MSE loss offers more mathematical opportunities than the hard-to-analyze CE loss, inspiring us to leverage MSE loss towards the theoretical investigation of NC. We develop three main contributions: (I) We show a new decomposition of the MSE loss into (A) terms directly interpretable through the lens of NC and which assume the last-layer classifier is exactly the least-squares classifier; and (B) a term capturing the deviation from this least-squares classifier. (II) We exhibit experiments on canonical datasets and networks demonstrating that term-(B) is negligible during training. This motivates us to introduce a new theoretical construct: the central path, where the linear classifier stays MSE-optimal for feature activations throughout the dynamics. (III) By studying renormalized gradient flow along the central path, we derive exact dynamics that predict NC."
                },
                {
                    "title": "A Geometric Analysis of Neural Collapse with Unconstrained Features",
                    "abstract": "We provide the first global optimization landscape analysis of $Neural\\;Collapse$ -- an intriguing empirical phenomenon that arises in the last-layer classifiers and features of neural networks during the terminal phase of training. As recently reported by Papyan et al., this phenomenon implies that ($i$) the class means and the last-layer classifiers all collapse to the vertices of a Simplex Equiangular Tight Frame (ETF) up to scaling, and ($ii$) cross-example within-class variability of last-layer activations collapses to zero. We study the problem based on a simplified $unconstrained\\;feature\\;model$, which isolates the topmost layers from the classifier of the neural network. In this context, we show that the classical cross-entropy loss with weight decay has a benign global landscape, in the sense that the only global minimizers are the Simplex ETFs while all other critical points are strict saddles whose Hessian exhibit negative curvature directions. In contrast to existing landscape analysis for deep neural networks which is often disconnected from practice, our analysis of the simplified model not only does it explain what kind of features are learned in the last layer, but it also shows why they can be efficiently optimized in the simplified settings, matching the empirical observations in practical deep network architectures. These findings could have profound implications for optimization, generalization, and robustness of broad interests. For example, our experiments demonstrate that one may set the feature dimension equal to the number of classes and fix the last-layer classifier to be a Simplex ETF for network training, which reduces memory cost by over $20\\%$ on ResNet18 without sacrificing the generalization performance."
                },
                {
                    "title": "Barlow Twins: Self-Supervised Learning via Redundancy Reduction",
                    "abstract": "Self-supervised learning (SSL) is rapidly closing the gap with supervised methods on large computer vision benchmarks. A successful approach to SSL is to learn embeddings which are invariant to distortions of the input sample. However, a recurring issue with this approach is the existence of trivial constant solutions. Most current methods avoid such solutions by careful implementation details. We propose an objective function that naturally avoids collapse by measuring the cross-correlation matrix between the outputs of two identical networks fed with distorted versions of a sample, and making it as close to the identity matrix as possible. This causes the embedding vectors of distorted versions of a sample to be similar, while minimizing the redundancy between the components of these vectors. The method is called Barlow Twins, owing to neuroscientist H. Barlow's redundancy-reduction principle applied to a pair of identical networks. Barlow Twins does not require large batches nor asymmetry between the network twins such as a predictor network, gradient stopping, or a moving average on the weight updates. Intriguingly it benefits from very high-dimensional output vectors. Barlow Twins outperforms previous methods on ImageNet for semi-supervised classification in the low-data regime, and is on par with current state of the art for ImageNet classification with a linear classifier head, and for transfer tasks of classification and object detection."
                },
                {
                    "title": "Understanding self-supervised Learning Dynamics without Contrastive Pairs",
                    "abstract": "While contrastive approaches of self-supervised learning (SSL) learn representations by minimizing the distance between two augmented views of the same data point (positive pairs) and maximizing views from different data points (negative pairs), recent \\emph{non-contrastive} SSL (e.g., BYOL and SimSiam) show remarkable performance {\\it without} negative pairs, with an extra learnable predictor and a stop-gradient operation. A fundamental question arises: why do these methods not collapse into trivial representations? We answer this question via a simple theoretical study and propose a novel approach, DirectPred, that \\emph{directly} sets the linear predictor based on the statistics of its inputs, without gradient training. On ImageNet, it performs comparably with more complex two-layer non-linear predictors that employ BatchNorm and outperforms a linear predictor by $2.5\\%$ in 300-epoch training (and $5\\%$ in 60-epoch). DirectPred is motivated by our theoretical study of the nonlinear learning dynamics of non-contrastive SSL in simple linear networks. Our study yields conceptual insights into how non-contrastive SSL methods learn, how they avoid representational collapse, and how multiple factors, like predictor networks, stop-gradients, exponential moving averages, and weight decay all come into play. Our simple theory recapitulates the results of real-world ablation studies in both STL-10 and ImageNet. Code is released https://github.com/facebookresearch/luckmatters/tree/master/ssl."
                },
                {
                    "title": "Exploring deep neural networks via layer-peeled model: Minority collapse in imbalanced training",
                    "abstract": "Significance The remarkable development of deep learning over the past decade relies heavily on sophisticated heuristics and tricks. To better exploit its potential in the coming decade, perhaps a rigorous framework for reasoning about deep learning is needed, which, however, is not easy to build due to the intricate details of neural networks. For near-term purposes, a practical alternative is to develop a mathematically tractable surrogate model, yet maintaining many characteristics of neural networks. This paper proposes a model of this kind that we term the Layer-Peeled Model. The effectiveness of this model is evidenced by, among others, its ability to reproduce a known empirical pattern and to predict a hitherto-unknown phenomenon when training deep-learning models on imbalanced datasets. In this paper, we introduce the Layer-Peeled Model, a nonconvex, yet analytically tractable, optimization program, in a quest to better understand deep neural networks that are trained for a sufficiently long time. As the name suggests, this model is derived by isolating the topmost layer from the remainder of the neural network, followed by imposing certain constraints separately on the two parts of the network. We demonstrate that the Layer-Peeled Model, albeit simple, inherits many characteristics of well-trained neural networks, thereby offering an effective tool for explaining and predicting common empirical patterns of deep-learning training. First, when working on class-balanced datasets, we prove that any solution to this model forms a simplex equiangular tight frame, which, in part, explains the recently discovered phenomenon of neural collapse [V. Papyan, X. Y. Han, D. L. Donoho, Proc. Natl. Acad. Sci. U.S.A. 117, 24652\u201324663 (2020)]. More importantly, when moving to the imbalanced case, our analysis of the Layer-Peeled Model reveals a hitherto-unknown phenomenon that we term Minority Collapse, which fundamentally limits the performance of deep-learning models on the minority classes. In addition, we use the Layer-Peeled Model to gain insights into how to mitigate Minority Collapse. Interestingly, this phenomenon is first predicted by the Layer-Peeled Model before being confirmed by our computational experiments."
                },
                {
                    "title": "WILDS: A Benchmark of in-the-Wild Distribution Shifts",
                    "abstract": "Distribution shifts -- where the training distribution differs from the test distribution -- can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity in the real-world deployments, these distribution shifts are under-represented in the datasets widely used in the ML community today. To address this gap, we present WILDS, a curated benchmark of 10 datasets reflecting a diverse range of distribution shifts that naturally arise in real-world applications, such as shifts across hospitals for tumor identification; across camera traps for wildlife monitoring; and across time and location in satellite imaging and poverty mapping. On each dataset, we show that standard training yields substantially lower out-of-distribution than in-distribution performance. This gap remains even with models trained by existing methods for tackling distribution shifts, underscoring the need for new methods for training models that are more robust to the types of distribution shifts that arise in practice. To facilitate method development, we provide an open-source package that automates dataset loading, contains default model architectures and hyperparameters, and standardizes evaluations. Code and leaderboards are available at https://wilds.stanford.edu."
                },
                {
                    "title": "Prevalence of neural collapse during the terminal phase of deep learning training",
                    "abstract": "Significance Modern deep neural networks for image classification have achieved superhuman performance. Yet, the complex details of trained networks have forced most practitioners and researchers to regard them as black boxes with little that could be understood. This paper considers in detail a now-standard training methodology: driving the cross-entropy loss to zero, continuing long after the classification error is already zero. Applying this methodology to an authoritative collection of standard deepnets and datasets, we observe the emergence of a simple and highly symmetric geometry of the deepnet features and of the deepnet classifier, and we document important benefits that the geometry conveys\u2014thereby helping us understand an important component of the modern deep learning training paradigm. Modern practice for training classification deepnets involves a terminal phase of training (TPT), which begins at the epoch where training error first vanishes. During TPT, the training error stays effectively zero, while training loss is pushed toward zero. Direct measurements of TPT, for three prototypical deepnet architectures and across seven canonical classification datasets, expose a pervasive inductive bias we call neural collapse (NC), involving four deeply interconnected phenomena. (NC1) Cross-example within-class variability of last-layer training activations collapses to zero, as the individual activations themselves collapse to their class means. (NC2) The class means collapse to the vertices of a simplex equiangular tight frame (ETF). (NC3) Up to rescaling, the last-layer classifiers collapse to the class means or in other words, to the simplex ETF (i.e., to a self-dual configuration). (NC4) For a given activation, the classifier\u2019s decision collapses to simply choosing whichever class has the closest train class mean (i.e., the nearest class center [NCC] decision rule). The symmetric and very simple geometry induced by the TPT confers important benefits, including better generalization performance, better robustness, and better interpretability."
                },
                {
                    "title": "Whitening for Self-Supervised Representation Learning",
                    "abstract": "Recent literature on self-supervised learning is based on the contrastive loss, where image instances which share the same semantic content (\"positives\") are contrasted with instances extracted from other images (\"negatives\"). However, in order for the learning to be effective, a lot of negatives should be compared with a positive pair. This is not only computationally demanding, but it also requires that the positive and the negative representations are kept consistent with each other over a long training period. In this paper we propose a different direction and a new loss function for self-supervised learning which is based on the whitening of the latent-space features. The whitening operation has a \"scattering\" effect on the batch samples, which compensates the lack of a large number of negatives, avoiding degenerate solutions where all the sample representations collapse to a single point. We empirically show that our loss accelerates self-supervised training and the learned representations are much more effective for downstream tasks than previously published work."
                },
                {
                    "title": "Measuring Robustness to Natural Distribution Shifts in Image Classification",
                    "abstract": "We study how robust current ImageNet models are to distribution shifts arising from natural variations in datasets. Most research on robustness focuses on synthetic image perturbations (noise, simulated weather artifacts, adversarial examples, etc.), which leaves open how robustness on synthetic distribution shift relates to distribution shift arising in real data. Informed by an evaluation of 204 ImageNet models in 213 different test conditions, we find that there is often little to no transfer of robustness from current synthetic to natural distribution shift. Moreover, most current techniques provide no robustness to the natural distribution shifts in our testbed. The main exception is training on larger and more diverse datasets, which in multiple cases increases robustness, but is still far from closing the performance gaps. Our results indicate that distribution shifts arising in real data are currently an open research problem. We provide our testbed and data as a resource for future work at this https URL ."
                },
                {
                    "title": "The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization",
                    "abstract": "We introduce four new real-world distribution shift datasets consisting of changes in image style, image blurriness, geographic location, camera operation, and more. With our new datasets, we take stock of previously proposed methods for improving out-of-distribution robustness and put them to the test. We find that using larger models and artificial data augmentations can improve robustness on real-world distribution shifts, contrary to claims in prior work. We find improvements in artificial robustness benchmarks can transfer to real-world distribution shifts, contrary to claims in prior work. Motivated by our observation that data augmentations can help with real-world distribution shifts, we also introduce a new data augmentation method which advances the state-of-the-art and outperforms models pre-trained with 1000\u00d7 more labeled data. Overall we find that some methods consistently help with distribution shifts in texture and local image statistics, but these methods do not help with some other distribution shifts like geographic changes. Our results show that future research must study multiple distribution shifts simultaneously, as we demonstrate that no evaluated method consistently improves robustness."
                },
                {
                    "title": "Noise or Signal: The Role of Image Backgrounds in Object Recognition",
                    "abstract": "We assess the tendency of state-of-the-art object recognition models to depend on signals from image backgrounds. We create a toolkit for disentangling foreground and background signal on ImageNet images, and find that (a) models can achieve non-trivial accuracy by relying on the background alone, (b) models often misclassify images even in the presence of correctly classified foregrounds--up to 87.5% of the time with adversarially chosen backgrounds, and (c) more accurate models tend to depend on backgrounds less. Our analysis of backgrounds brings us closer to understanding which correlations machine learning models use, and how they determine models' out of distribution performance."
                },
                {
                    "title": "Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning",
                    "abstract": "We introduce Bootstrap Your Own Latent (BYOL), a new approach to self-supervised image representation learning. BYOL relies on two neural networks, referred to as online and target networks, that interact and learn from each other. From an augmented view of an image, we train the online network to predict the target network representation of the same image under a different augmented view. At the same time, we update the target network with a slow-moving average of the online network. While state-of-the art methods rely on negative pairs, BYOL achieves a new state of the art without them. BYOL reaches $74.3\\%$ top-1 classification accuracy on ImageNet using a linear evaluation with a ResNet-50 architecture and $79.6\\%$ with a larger ResNet. We show that BYOL performs on par or better than the current state of the art on both transfer and semi-supervised benchmarks. Our implementation and pretrained models are given on GitHub."
                },
                {
                    "title": "An Investigation of Why Overparameterization Exacerbates Spurious Correlations",
                    "abstract": "We study why overparameterization -- increasing model size well beyond the point of zero training error -- can hurt test error on minority groups despite improving average test error when there are spurious correlations in the data. Through simulations and experiments on two image datasets, we identify two key properties of the training data that drive this behavior: the proportions of majority versus minority groups, and the signal-to-noise ratio of the spurious correlations. We then analyze a linear setting and theoretically show how the inductive bias of models towards \"memorizing\" fewer examples can cause overparameterization to hurt. Our analysis leads to a counterintuitive approach of subsampling the majority group, which empirically achieves low minority error in the overparameterized regime, even though the standard approach of upweighting the minority fails. Overall, our results suggest a tension between using overparameterized models versus using all the training data for achieving low worst-group error."
                },
                {
                    "title": "Supervised Contrastive Learning",
                    "abstract": "Cross entropy is the most widely used loss function for supervised training of image classification models. In this paper, we propose a novel training methodology that consistently outperforms cross entropy on supervised learning tasks across different architectures and data augmentations. We modify the batch contrastive loss, which has recently been shown to be very effective at learning powerful representations in the self-supervised setting. We are thus able to leverage label information more effectively than cross entropy. Clusters of points belonging to the same class are pulled together in embedding space, while simultaneously pushing apart clusters of samples from different classes. In addition to this, we leverage key ingredients such as large batch sizes and normalized embeddings, which have been shown to benefit self-supervised learning. On both ResNet-50 and ResNet-200, we outperform cross entropy by over 1%, setting a new state of the art number of 78.8% among methods that use AutoAugment data augmentation. The loss also shows clear benefits for robustness to natural corruptions on standard benchmarks on both calibration and accuracy. Compared to cross entropy, our supervised contrastive loss is more stable to hyperparameter settings such as optimizers or data augmentations."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization",
                    "abstract": "Overparameterized neural networks can be highly accurate on average on an i.i.d. test set yet consistently fail on atypical groups of the data (e.g., by learning spurious correlations that hold on average but not in such groups). Distributionally robust optimization (DRO) allows us to learn models that instead minimize the worst-case training loss over a set of pre-defined groups. However, we find that naively applying group DRO to overparameterized neural networks fails: these models can perfectly fit the training data, and any model with vanishing average training loss also already has vanishing worst-case training loss. Instead, the poor worst-case performance arises from poor generalization on some groups. By coupling group DRO models with increased regularization---a stronger-than-typical L2 penalty or early stopping---we achieve substantially higher worst-group accuracies, with 10-40 percentage point improvements on a natural language inference task and two image tasks, while maintaining high average accuracies. Our results suggest that regularization is important for worst-group generalization in the overparameterized regime, even if it is not needed for average generalization. Finally, we introduce a stochastic optimization algorithm, with convergence guarantees, to efficiently train group DRO models."
                },
                {
                    "title": "Momentum Contrast for Unsupervised Visual Representation Learning",
                    "abstract": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a moving-averaged encoder. This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning. MoCo provides competitive results under the common linear protocol on ImageNet classification. More importantly, the representations learned by MoCo transfer well to downstream tasks. MoCo can outperform its supervised pre-training counterpart in 7 detection/segmentation tasks on PASCAL VOC, COCO, and other datasets, sometimes surpassing it by large margins. This suggests that the gap between unsupervised and supervised representation learning has been largely closed in many vision tasks."
                },
                {
                    "title": "Natural Adversarial Examples",
                    "abstract": "We introduce two challenging datasets that reliably cause machine learning model performance to substantially degrade. The datasets are collected with a simple adversarial filtration technique to create datasets with limited spurious cues. Our datasets\u2019 real-world, unmodified examples transfer to various unseen models reliably, demonstrating that computer vision models have shared weaknesses. The first dataset is called IMAGENET-A and is like the ImageNet test set, but it is far more challenging for existing models. We also curate an adversarial out-of-distribution detection dataset called IMAGENET-O, which is the first out-of-distribution detection dataset created for ImageNet models. On IMAGENET-A a DenseNet-121 obtains around 2% accuracy, an accuracy drop of approximately 90%, and its out-of-distribution detection performance on IMAGENET-O is near random chance levels. We find that existing data augmentation techniques hardly boost performance, and using other public training datasets provides improvements that are limited. However, we find that improvements to computer vision architectures provide a promising path towards robust models."
                },
                {
                    "title": "Do Image Classifiers Generalize Across Time?",
                    "abstract": "Vision models notoriously flicker when applied to videos: they correctly recognize objects in some frames, but fail on perceptually similar, nearby frames. In this work, we systematically analyze the robustness of image classifiers to such temporal perturbations in videos. To do so, we construct two new datasets, ImageNet-Vid-Robust and YTBB-Robust, containing a total of 57,897 images grouped into 3,139 sets of perceptually similar images. Our datasets were derived from ImageNet-Vid and Youtube-BB, respectively, and thoroughly re-annotated by human experts for image similarity. We evaluate a diverse array of classifiers pre-trained on ImageNet and show a median classification accuracy drop of 16 and 10 points, respectively, on our two datasets. Additionally, we evaluate three detection models and show that natural perturbations induce both classification as well as localization errors, leading to a median drop in detection mAP of 14 points. Our analysis demonstrates that perturbations occurring naturally in videos pose a substantial and realistic challenge to deploying convolutional neural networks in environments that require both reliable and low-latency predictions."
                },
                {
                    "title": "Learning Representations by Maximizing Mutual Information Across Views",
                    "abstract": "We propose an approach to self-supervised representation learning based on maximizing mutual information between features extracted from multiple views of a shared context. For example, one could produce multiple views of a local spatio-temporal context by observing it from different locations (e.g., camera positions within a scene), and via different modalities (e.g., tactile, auditory, or visual). Or, an ImageNet image could provide a context from which one produces multiple views by repeatedly applying data augmentation. Maximizing mutual information between features extracted from these views requires capturing information about high-level factors whose influence spans multiple views -- e.g., presence of certain objects or occurrence of certain events. \nFollowing our proposed approach, we develop a model which learns image representations that significantly outperform prior methods on the tasks we consider. Most notably, using self-supervised learning, our model learns representations which achieve 68.1% accuracy on ImageNet using standard linear evaluation. This beats prior results by over 12% and concurrent results by 7%. When we extend our model to use mixture-based representations, segmentation behaviour emerges as a natural side-effect. Our code is available online: this https URL."
                },
                {
                    "title": "Adversarial Examples Are Not Bugs, They Are Features",
                    "abstract": "Adversarial examples have attracted significant attention in machine learning, but the reasons for their existence and pervasiveness remain unclear. We demonstrate that adversarial examples can be directly attributed to the presence of non-robust features: features derived from patterns in the data distribution that are highly predictive, yet brittle and incomprehensible to humans. After capturing these features within a theoretical framework, we establish their widespread existence in standard datasets. Finally, we present a simple setting where we can rigorously tie the phenomena we observe in practice to a misalignment between the (human-specified) notion of robustness and the inherent geometry of the data."
                },
                {
                    "title": "Unsupervised Embedding Learning via Invariant and Spreading Instance Feature",
                    "abstract": "This paper studies the unsupervised embedding learning problem, which requires an effective similarity measurement between samples in low-dimensional embedding space. Motivated by the positive concentrated and negative separated properties observed from category-wise supervised learning, we propose to utilize the instance-wise supervision to approximate these properties, which aims at learning data augmentation invariant and instance spread-out features. To achieve this goal, we propose a novel instance based softmax embedding method, which directly optimizes the `real' instance features on top of the softmax function. It achieves significantly faster learning speed and higher accuracy than all existing methods. The proposed method performs well for both seen and unseen testing categories with cosine similarity. It also achieves competitive performance even without pre-trained network over samples from fine-grained categories."
                },
                {
                    "title": "Do ImageNet Classifiers Generalize to ImageNet?",
                    "abstract": "We build new test sets for the CIFAR-10 and ImageNet datasets. Both benchmarks have been the focus of intense research for almost a decade, raising the danger of overfitting to excessively re-used test sets. By closely following the original dataset creation processes, we test to what extent current classification models generalize to new data. We evaluate a broad range of models and find accuracy drops of 3% - 15% on CIFAR-10 and 11% - 14% on ImageNet. However, accuracy gains on the original test sets translate to larger gains on the new test sets. Our results suggest that the accuracy drops are not caused by adaptivity, but by the models' inability to generalize to slightly \"harder\" images than those found in the original test sets."
                },
                {
                    "title": "Using Pre-Training Can Improve Model Robustness and Uncertainty",
                    "abstract": "He et al. (2018) have called into question the utility of pre-training by showing that training from scratch can often yield similar performance to pre-training. We show that although pre-training may not improve performance on traditional classification metrics, it improves model robustness and uncertainty estimates. Through extensive experiments on adversarial examples, label corruption, class imbalance, out-of-distribution detection, and confidence calibration, we demonstrate large gains from pre-training and complementary effects with task-specific methods. We introduce adversarial pre-training and show approximately a 10% absolute improvement over the previous state-of-the-art in adversarial robustness. In some cases, using pre-training without task-specific methods also surpasses the state-of-the-art, highlighting the need for pre-training when evaluating future methods on robustness and uncertainty tasks."
                },
                {
                    "title": "Revisiting Self-Supervised Visual Representation Learning",
                    "abstract": "Unsupervised visual representation learning remains a largely unsolved problem in computer vision research. Among a big body of recently proposed approaches for unsupervised learning of visual representations, a class of self-supervised techniques achieves superior performance on many challenging benchmarks. A large number of the pretext tasks for self-supervised learning have been studied, but other important aspects, such as the choice of convolutional neural networks (CNN), has not received equal attention. Therefore, we revisit numerous previously proposed self-supervised models, conduct a thorough large scale study and, as a result, uncover multiple crucial insights. We challenge a number of common practices in self-supervised visual representation learning and observe that standard recipes for CNN design do not always translate to self-supervised representation learning. As part of our study, we drastically boost the performance of previously proposed techniques and outperform previously published state-of-the-art results by a large margin. We will release the code for reproducing our experiments when the anonymity requirements are lifted."
                },
                {
                    "title": "ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness",
                    "abstract": "Convolutional Neural Networks (CNNs) are commonly thought to recognise objects by learning increasingly complex representations of object shapes. Some recent studies suggest a more important role of image textures. We here put these conflicting hypotheses to a quantitative test by evaluating CNNs and human observers on images with a texture-shape cue conflict. We show that ImageNet-trained CNNs are strongly biased towards recognising textures rather than shapes, which is in stark contrast to human behavioural evidence and reveals fundamentally different classification strategies. We then demonstrate that the same standard architecture (ResNet-50) that learns a texture-based representation on ImageNet is able to learn a shape-based representation instead when trained on \"Stylized-ImageNet\", a stylized version of ImageNet. This provides a much better fit for human behavioural performance in our well-controlled psychophysical lab setting (nine experiments totalling 48,560 psychophysical trials across 97 observers) and comes with a number of unexpected emergent benefits such as improved object detection performance and previously unseen robustness towards a wide range of image distortions, highlighting advantages of a shape-based representation."
                },
                {
                    "title": "Representation Learning with Contrastive Predictive Coding",
                    "abstract": "While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments."
                },
                {
                    "title": "On the Optimization of Deep Networks: Implicit Acceleration by Overparameterization",
                    "abstract": "Conventional wisdom in deep learning states that increasing depth improves expressiveness but complicates optimization. This paper suggests that, sometimes, increasing depth can speed up optimization. The effect of depth on optimization is decoupled from expressiveness by focusing on settings where additional layers amount to overparameterization - linear neural networks, a well-studied model. Theoretical analysis, as well as experiments, show that here depth acts as a preconditioner which may accelerate convergence. Even on simple convex problems such as linear regression with $\\ell_p$ loss, $p>2$, gradient descent can benefit from transitioning to a non-convex overparameterized objective, more than it would from some common acceleration schemes. We also prove that it is mathematically impossible to obtain the acceleration effect of overparametrization via gradients of any regularizer."
                },
                {
                    "title": "The Implicit Bias of Gradient Descent on Separable Data",
                    "abstract": "We examine gradient descent on unregularized logistic regression problems, with homogeneous linear predictors on linearly separable datasets. We show the predictor converges to the direction of the max-margin (hard margin SVM) solution. The result also generalizes to other monotone decreasing loss functions with an infimum at infinity, to multi-class problems, and to training a weight layer in a deep network in a certain restricted setting. Furthermore, we show this convergence is very slow, and only logarithmic in the convergence of the loss itself. This can help explain the benefit of continuing to optimize the logistic or cross-entropy loss even after the training error is zero and the training loss is extremely small, and, as we show, even if the validation loss increases. Our methodology can also aid in understanding implicit regularization n more complex models and with other optimization methods."
                },
                {
                    "title": "Implicit Regularization in Matrix Factorization",
                    "abstract": "We study implicit regularization when optimizing an underdetermined quadratic objective over a matrix $X$ with gradient descent on a factorization of X. We conjecture and provide empirical and theoretical evidence that with small enough step sizes and initialization close enough to the origin, gradient descent on a full dimensional factorization converges to the minimum nuclear norm solution."
                },
                {
                    "title": "Object Recognition with and without Objects",
                    "abstract": "While recent deep neural networks have achieved a promising performance on object recognition, they rely implicitly on the visual contents of the whole image. In this paper, we train deep neural networks on the foreground (object) and background (context) regions of images respectively. Considering human recognition in the same situations, networks trained on the pure background without objects achieves highly reasonable recognition performance that beats humans by a large margin if only given context. However, humans still outperform networks with pure object available, which indicates networks and human beings have different mechanisms in understanding an image. Furthermore, we straightforwardly combine multiple trained networks to explore different visual cues learned by different networks. Experiments show that useful visual hints can be explicitly learned separately and then combined to achieve higher performance, which verifies the advantages of the proposed framework."
                },
                {
                    "title": "In Search of the Real Inductive Bias: On the Role of Implicit Regularization in Deep Learning",
                    "abstract": "We present experiments demonstrating that some other form of capacity control, different from network size, plays a central role in learning multilayer feed-forward networks. We argue, partially through analogy to matrix factorization, that this is an inductive bias that can help shed light on deep learning."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Generalization Performance of Support Vector Machines and Other Pattern Classifiers",
                    "abstract": "The aim of this chapter is to summarise results that have been obtained for high conndence generalization error bounds for the Support Vector Machine (SVM) and other pattern classiiers related to the SVM. As a by-product of the analysis we argue that the margin and number of support vectors are both estimators of the degree to which the distribution generating the inputs assists identiication of the target hyperplane. 1.1 Introduction Generalization analysis of pattern classiiers is concerned with determining the factors that aaect the accuracy of a pattern classiier. Such an analysis requires assumptions to be made about how the data used to train the classiier was gathered and how subsequent data will be generated. One of the most popular assumptions originally championed by Vapnik and Chervonenkis 12] is to assume that the training and testing data are both generated according to the same probability distribution. The distribution can be viewed as a model of the natural processes which give rise to the observed phenomenon. Since it is usually more diicult to estimate the distribution than to learn the classiication function, it is important that no assumptions are made about the distribution, resulting in a so-called distribution-free analysis. We will consider bounds on the generalization error, that is the probability of"
                },
                {
                    "title": "Project and Probe: Sample-Efficient Adaptation by Interpolating Orthogonal Features",
                    "abstract": "Transfer learning with a small amount of target data is an effective and common approach to adapting a pre-trained model to distribution shifts. In some situations, target data labels may be expensive to obtain, so we may only have access to a limited number of target data points. To make the most of a very small target dataset, we propose a lightweight, sample-efficient approach that learns a diverse set of features and adapts to a target distribution by interpolating these features. Our approach, P ROJECT AND P ROBE (P RO 2 ), first learns a linear projection that maps a pre-trained embedding onto orthogonal directions while being predictive of labels in the source dataset. The goal of this step is to learn a variety of predictive features, so that at least some of them remain useful after distribution shift. P RO 2 then learns a linear classifier on top of these projected features using a small target dataset. Theoretically, we find that P RO 2 results in more sample-efficient generalization by inducing a favorable bias-variance tradeoff. Our experiments on four datasets, with multiple distribution shift settings for each, show that P RO 2 improves performance by 5-15% when given limited target data compared to prior methods such as standard linear probing."
                },
                {
                    "title": "Towards Mitigating Spurious Correlations in the Wild: A Benchmark & a more Realistic Dataset",
                    "abstract": "Deep neural networks often exploit non-predictive features that are spuriously correlated with class labels, leading to poor performance on groups of examples without such features. Despite the growing body of recent works on remedying spurious correlations, the lack of a standardized benchmark hinders reproducible evaluation and comparison of the proposed solutions. To address this, we present S PU C O , a python package with modular implementations of state-of-the-art solutions enabling easy and reproducible evaluation of current methods. Using S PU C O , we demonstrate the limitations of existing datasets and evaluation schemes in validating the learning of predictive features over spurious ones. To overcome these limitations, we propose two new vision datasets: (1) S PU C O MNIST, a synthetic dataset that enables simulating the effect of real world data properties e.g. difficulty of learning spurious feature, as well as noise in the labels and features; (2) S PU C O A NIMALS , a large-scale dataset curated from ImageNet that captures spurious correlations in the wild much more closely than existing datasets. These contributions highlight the shortcomings of current methods and provide a direction for future research in tackling spurious correlations. S PU C O , containing the benchmark and datasets, can be found at https://github.com/BigML-CS-UCLA/SpuCo , with detailed documentation available at https://spuco.readthedocs.io/en/latest/ ."
                },
                {
                    "title": "ObjectNet: A large-scale bias-controlled dataset for pushing the limits of object recognition models",
                    "abstract": "We collect a large real-world test set, ObjectNet, for object recognition with controls where object backgrounds, rotations, and imaging viewpoints are random. Most scientific experiments have controls, confounds which are removed from the data, to ensure that subjects cannot perform a task by exploiting trivial correlations in the data. Historically, large machine learning and computer vision datasets have lacked such controls. This has resulted in models that must be fine-tuned for new datasets and perform better on datasets than in real-world applications. When tested on ObjectNet, object detectors show a 40-45% drop in performance, with respect to their performance on other benchmarks, due to the controls for biases. Controls make ObjectNet robust to fine-tuning showing only small performance increases. We develop a highly automated platform that enables gathering datasets with controls by crowdsourcing image capturing and annotation. ObjectNet is the same size as the ImageNet test set (50,000 images), and by design does not come paired with a training set in order to encourage generalization. The dataset is both easier than ImageNet (objects are largely centered and unoccluded) and harder (due to the controls). Although we focus on object recognition here, data with controls can be gathered at scale using automated tools throughout machine learning to generate datasets that exercise models in new ways thus providing valuable feedback to researchers. This work opens up new avenues for research in generalizable, robust, and more human-like computer vision and in creating datasets where results are predictive of real-world performance."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "The mnist database of handwritten digits",
                    "abstract": "Disclosed is an improved articulated bar flail having shearing edges for efficiently shredding materials. An improved shredder cylinder is disclosed with a plurality of these flails circumferentially spaced and pivotally attached to the periphery of a rotatable shaft. Also disclosed is an improved shredder apparatus which has a pair of these shredder cylinders mounted to rotate about spaced parallel axes which cooperates with a conveyer apparatus which has a pair of inclined converging conveyer belts with one of the belts mounted to move with respect to the other belt to allow the transport of articles of various sizes therethrough."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nWhat is the mechanism by which the projection head in contrastive self-supervised learning improves the generalizability and robustness of learned representations?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it will deepen our understanding of representation learning, particularly in contrastive methods. By elucidating the role of the projection head, future research can be directed towards optimizing representation learning techniques, leading to more robust models that generalize better across various domains. This understanding could also pave the way for practical applications in fields such as computer vision and natural language processing, where robust representations are essential for performance in real-world tasks.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to analyze feature learning both before and after the projection head, which requires a nuanced understanding of the interaction between layers in neural networks. Naive approaches may fail because they do not account for the complexities introduced by non-linear activations and the varying importance of features across layers. Additionally, the inherent differences in learning mechanisms between contrastive and non-contrastive methods complicate the theoretical analysis, making it difficult to draw general conclusions about the projection head's impact.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on non-contrastive self-supervised methods, leaving a gap in understanding the projection head's role in contrastive learning. The lack of theoretical frameworks that adequately address the unique learning dynamics of contrastive methods has been a barrier to progress. Additionally, existing studies have not explored the projection head's effects on robustness under misalignment between pretraining and downstream objectives, which is a critical aspect of this problem. Our approach differs by providing a comprehensive theoretical analysis and extending findings to supervised contrastive learning and supervised learning, addressing these gaps.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a theoretical analysis of the projection head's impact on representation quality and robustness in contrastive learning. We will utilize a combination of linear and non-linear models to demonstrate how the projection head improves robustness when data augmentation affects useful features. The dataset will include various augmented views of the same examples, and we will measure the generalizability of representations using metrics such as classification accuracy and robustness to feature perturbations. We expect to show that the projection head significantly enhances the learned representations, reducing class"
            }
        },
        "author_data": {
            "bf9798fc-958c-4dcb-a24a-c5968f18dd29": {
                "pk": "bf9798fc-958c-4dcb-a24a-c5968f18dd29",
                "name": "Yihao Xue",
                "collaborators": [
                    "Baharan Mirzasoleiman",
                    "Kyle Whitecross",
                    "Yu Yang",
                    "Rui Yang",
                    "Xiaohan Chen",
                    "Siddharth Joshi",
                    "Zidong Wang",
                    "Ali Payani",
                    "S. Joshi",
                    "Wenhan Yang"
                ],
                "domain": [
                    "Contrastive Learning",
                    "Fault Diagnosis",
                    "Federated Learning",
                    "Robustness"
                ],
                "publications": [
                    {
                        "title": "Understanding the Robustness of Multi-modal Contrastive Learning to Distribution Shift",
                        "abstract": "Recently, multimodal contrastive learning (MMCL) approaches, such as CLIP, have achieved a remarkable success in learning representations that are robust against distribution shift and generalize to new domains. Despite the empirical success, the mechanism behind learning such generalizable representations is not understood. In this work, we rigorously analyze this problem and uncover two mechanisms behind MMCL's robustness: \\emph{intra-class contrasting}, which allows the model to learn features with a high variance, and \\emph{inter-class feature sharing}, where annotated details in one class help learning other classes better. Both mechanisms prevent spurious features that are over-represented in the training data to overshadow the generalizable core features. This yields superior zero-shot classification accuracy under distribution shift. Furthermore, we theoretically demonstrate the benefits of using rich captions on robustness and explore the effect of annotating different types of details in the captions. We validate our theoretical findings through experiments, including a well-designed synthetic experiment and an experiment involving training CLIP models on MSCOCO/Conceptual Captions and evaluating them on shifted ImageNets."
                    },
                    {
                        "title": "A Novel Local Binary Temporal Convolutional Neural Network for Bearing Fault Diagnosis",
                        "abstract": "In bearing fault diagnosis, the faulty data are generally limited due to the high cost of fault signal collection. Considering the excessive parameters in the traditional convolutional neural network (CNN), such a limited data issue can cause overfitting problem during the model training, eventually resulting in poor fault diagnosis performance. To resolve the overfitting issue and elevate the diagnostic accuracy of the conventional methods, a novel fault diagnosis method based on local binary temporal CNN (LBTCNN) is proposed in this article. In the proposed LBTCNN, a novel temporal module with dilated causal convolution for deep feature extraction is proposed to increase model depth under limited model parameters, and a local binary convolution (LBC) layer is adopted to reduce the computational parameters. To evaluate the effectiveness of the proposed method, several experiments under different scenarios such as limited samples and different noise levels are conducted on two datasets, including the rolling bearing accelerated life test dataset of Xi\u2019an Jiaotong University and Changxing Sumyoung Technology (XJTU-SY), and the motor bearing dataset of Case Western Reserve University (CWRU). The comparison results demonstrate that the LBTCNN method is superior over six other prominent fault diagnosis approaches under different bearing operation stages, different training samples, and different signal-to-noise ratios (SNRs)."
                    },
                    {
                        "title": "Few-shot Adaptation to Distribution Shifts By Mixing Source and Target Embeddings",
                        "abstract": "Pretrained machine learning models need to be adapted to distribution shifts when deployed in new target environments. When obtaining labeled data from the target distribution is expensive, few-shot adaptation with only a few examples from the target distribution becomes essential. In this work, we propose MixPro, a lightweight and highly data-efficient approach for few-shot adaptation. MixPro first generates a relatively large dataset by mixing (linearly combining) pre-trained embeddings of large source data with those of the few target examples. This process preserves important features of both source and target distributions, while mitigating the specific noise in the small target data. Then, it trains a linear classifier on the mixed embeddings to effectively adapts the model to the target distribution without overfitting the small target data. Theoretically, we demonstrate the advantages of MixPro over previous methods. Our experiments, conducted across various model architectures on 8 datasets featuring different types of distribution shifts, reveal that MixPro can outperform baselines by up to 7\\%, with only 2-4 target examples."
                    },
                    {
                        "title": "Towards Mitigating Spurious Correlations in the Wild: A Benchmark & a more Realistic Dataset",
                        "abstract": "Deep neural networks often exploit non-predictive features that are spuriously correlated with class labels, leading to poor performance on groups of examples without such features. Despite the growing body of recent works on remedying spurious correlations, the lack of a standardized benchmark hinders reproducible evaluation and comparison of the proposed solutions. To address this, we present S PU C O , a python package with modular implementations of state-of-the-art solutions enabling easy and reproducible evaluation of current methods. Using S PU C O , we demonstrate the limitations of existing datasets and evaluation schemes in validating the learning of predictive features over spurious ones. To overcome these limitations, we propose two new vision datasets: (1) S PU C O MNIST, a synthetic dataset that enables simulating the effect of real world data properties e.g. difficulty of learning spurious feature, as well as noise in the labels and features; (2) S PU C O A NIMALS , a large-scale dataset curated from ImageNet that captures spurious correlations in the wild much more closely than existing datasets. These contributions highlight the shortcomings of current methods and provide a direction for future research in tackling spurious correlations. S PU C O , containing the benchmark and datasets, can be found at https://github.com/BigML-CS-UCLA/SpuCo , with detailed documentation available at https://spuco.readthedocs.io/en/latest/ ."
                    },
                    {
                        "title": "Eliminating Spurious Correlations from Pre-trained Models via Data Mixing",
                        "abstract": "Machine learning models pre-trained on large datasets have achieved remarkable convergence and robustness properties. However, these models often exploit spurious correlations between certain attributes and labels, which are prevalent in the majority of examples within speci\ufb01c categories but are not predictive of these categories in general. The learned spurious correlations may persist even after \ufb01ne-tuning on new data, which degrades models\u2019 performance on examples that do not exhibit the spurious correlation. In this work, we propose a simple and highly effective method to eliminate spurious correlations from pre-trained models. The key idea of our method is to leverage a small set of examples with spurious attributes, and balance the spurious attributes across all classes via data mixing. We theoretically con\ufb01rm the effectiveness of our method, and empirically demonstrate its state-of-the-art performance on various vision and NLP tasks, including eliminating spurious correlations from pre-trained ResNet50 on Waterbirds and CelebA, adversarially pre-trained ResNet50 on ImageNet, and BERT pre-trained on CivilComments."
                    },
                    {
                        "title": "Deep Transfer Learning for Bearing Fault Diagnosis: A Systematic Review Since 2016",
                        "abstract": "The traditional deep learning-based bearing fault diagnosis approaches assume that the training and test data follow the same distribution. This assumption, however, is not always true for the bearing data collected in practical scenarios, leading to a significant decline in fault diagnosis performance. In order to satisfy this assumption, the transfer learning concept is introduced in deep learning by transferring the knowledge learned from other data or models. Due to the excellent capability of feature learning and domain transfer, deep transfer learning methods have gained widespread attention in bearing fault diagnosis in recent years. This review presents a comprehensive review of the development of deep transfer learning-based bearing fault diagnosis approaches since 2016. In this review, a novel taxonomy of deep transfer learning-based bearing fault diagnosis methods is proposed from the perspective of target domain data properties divided by labels, machines, and faults. By covering the whole life cycle of deep transfer learning-based fault diagnosis and discussing the research challenges and opportunities, this review provides a systematic guideline for researchers and practitioners to efficiently identify suitable deep transfer learning models based on the actual problems encountered in bearing fault diagnosis."
                    },
                    {
                        "title": "Towards Mitigating more Challenging Spurious Correlations: A Benchmark&New Datasets",
                        "abstract": "Deep neural networks often exploit non-predictive features that are spuriously correlated with class labels, leading to poor performance on groups of examples without such features. Despite the growing body of recent works on remedying spurious correlations, the lack of a standardized benchmark hinders reproducible evaluation and comparison of the proposed solutions. To address this, we present SpuCo, a python package with modular implementations of state-of-the-art solutions enabling easy and reproducible evaluation of current methods. Using SpuCo, we demonstrate the limitations of existing datasets and evaluation schemes in validating the learning of predictive features over spurious ones. To overcome these limitations, we propose two new vision datasets: (1) SpuCoMNIST, a synthetic dataset that enables simulating the effect of real world data properties e.g. difficulty of learning spurious feature, as well as noise in the labels and features; (2) SpuCoAnimals, a large-scale dataset curated from ImageNet that captures spurious correlations in the wild much more closely than existing datasets. These contributions highlight the shortcomings of current methods and provide a direction for future research in tackling spurious correlations. SpuCo, containing the benchmark and datasets, can be found at https://github.com/BigML-CS-UCLA/SpuCo, with detailed documentation available at https://spuco.readthedocs.io/en/latest/."
                    },
                    {
                        "title": "Which Features are Learnt by Contrastive Learning? On the Role of Simplicity Bias in Class Collapse and Feature Suppression",
                        "abstract": "Contrastive learning (CL) has emerged as a powerful technique for representation learning, with or without label supervision. However, supervised CL is prone to collapsing representations of subclasses within a class by not capturing all their features, and unsupervised CL may suppress harder class-relevant features by focusing on learning easy class-irrelevant features; both significantly compromise representation quality. Yet, there is no theoretical understanding of \\textit{class collapse} or \\textit{feature suppression} at \\textit{test} time. We provide the first unified theoretically rigorous framework to determine \\textit{which} features are learnt by CL. Our analysis indicate that, perhaps surprisingly, bias of (stochastic) gradient descent towards finding simpler solutions is a key factor in collapsing subclass representations and suppressing harder class-relevant features. Moreover, we present increasing embedding dimensionality and improving the quality of data augmentations as two theoretically motivated solutions to {feature suppression}. We also provide the first theoretical explanation for why employing supervised and unsupervised CL together yields higher-quality representations, even when using commonly-used stochastic gradient methods."
                    },
                    {
                        "title": "Weighted Multi-view Zero-shot Learning Prototype Shift Model in Fault Diagnosis",
                        "abstract": "Different fault severities in the mechanical fault diagnosis field contribute to different fault classes\u2019 distribution and prototype position. It brings a challenge to mechanical fault diagnosis. This paper proposes a weighted multi-view hierarchical, zero-shot learning model to address the problem. The multi-view inputs are proposed to locate the positions of fault prototypes accurately. To precisely utilize multi-view input information, an unsupervised weighting strategy is proposed to balance positive and negative views. Hierarchical classification is utilized to gradually change this classification boundary or prototype position of the model. The effectiveness of the proposed method is verified through comparative experiments, and the results indicate that the model can classify mechanical fault across different fault severity problems. Compared with other classical zero-shot learning models, the results show that the proposed model performs better in mechanical fault diagnosis than the existing methods."
                    },
                    {
                        "title": "Investigating the Impact of Model Width and Density on Generalization in Presence of Label Noise",
                        "abstract": "Increasing the size of overparameterized neural networks has been a key in achieving state-of-the-art performance. This is captured by the double descent phenomenon, where the test loss follows a decreasing-increasing-decreasing pattern (or sometimes monotonically decreasing) as model width increases. However, the effect of label noise on the test loss curve has not been fully explored. In this work, we uncover an intriguing phenomenon where label noise leads to a \\textit{final ascent} in the originally observed double descent curve. Specifically, under a sufficiently large noise-to-sample-size ratio, optimal generalization is achieved at intermediate widths. Through theoretical analysis, we attribute this phenomenon to the shape transition of test loss variance induced by label noise. Furthermore, we extend the final ascent phenomenon to model density and provide the first theoretical characterization showing that reducing density by randomly dropping trainable parameters improves generalization under label noise. We also thoroughly examine the roles of regularization and sample size. Surprisingly, we find that larger $\\ell_2$ regularization and robust learning methods against label noise exacerbate the final ascent. We confirm the validity of our findings through extensive experiments on ReLu networks trained on MNIST, ResNets/ViTs trained on CIFAR-10/100, and InceptionResNet-v2 trained on Stanford Cars with real-world noisy labels."
                    },
                    {
                        "title": "Investigating Why Contrastive Learning Benefits Robustness Against Label Noise",
                        "abstract": "Self-supervised Contrastive Learning (CL) has been recently shown to be very effective in preventing deep networks from overfitting noisy labels. Despite its empirical success, the theoretical understanding of the effect of contrastive learning on boosting robustness is very limited. In this work, we rigorously prove that the representation matrix learned by contrastive learning boosts robustness, by having: (i) one prominent singular value corresponding to each sub-class in the data, and significantly smaller remaining singular values; and (ii) {a large alignment between the prominent singular vectors and the clean labels of each sub-class. The above properties enable a linear layer trained on such representations to effectively learn the clean labels without overfitting the noise.} We further show that the low-rank structure of the Jacobian of deep networks pre-trained with contrastive learning allows them to achieve a superior performance initially, when fine-tuned on noisy labels. Finally, we demonstrate that the initial robustness provided by contrastive learning enables robust training methods to achieve state-of-the-art performance under extreme noise levels, e.g., an average of 27.18\\% and 15.58\\% increase in accuracy on CIFAR-10 and CIFAR-100 with 80\\% symmetric noisy labels, and 4.11\\% increase in accuracy on WebVision."
                    },
                    {
                        "title": "Superior generalization of smaller models in the presence of significant label noise",
                        "abstract": "The bene\ufb01ts of over-parameterization in achieving superior generalization performance have been shown in several recent studies, justifying the trend of using larger models in practice. In the context of robust learning however, the effect of neural network size has not been well studied. In this work, we \ufb01nd that in the presence of a substantial fraction of mislabeled examples, increasing the network size beyond some point can be harmful. In particular, the originally monotonic or \u2018double descent\u2019 test loss curve (w.r.t. network width) turns into a U-shaped or a double U-shaped curve when label noise increases, suggesting that the best generalization is achieved by some model with intermediate size. We observe that when network size is controlled by density through random pruning, similar test loss behaviour is observed. We also take a closer look into both phenomenon through bias-variance decomposition and theoretically characterize how label noise shapes the variance term. Similar behavior of the test loss can be observed even when state-of-the-art robust methods are applied, indicating that limiting the network size could further boost existing methods. Finally, we empirically examine the effect of network size on the smoothness of learned functions, and \ufb01nd that the originally negative correlation between size and smoothness is \ufb02ipped by label noise."
                    },
                    {
                        "title": "T HE F INAL A SCENT : W HEN B IGGER M ODELS G ENERALIZE W ORSE ON N OISY -L ABELED D ATA",
                        "abstract": ","
                    },
                    {
                        "title": "2022 T HE F INAL A SCENT : W HEN B IGGER M ODELS G ENERALIZE W ORSE ON N OISY -L ABELED D ATA",
                        "abstract": ","
                    },
                    {
                        "title": "Towards Understanding the In\ufb02uence of Individual Clients in Federated Learning",
                        "abstract": "Federated learning allows mobile clients to jointly train a global model without sending their private data to a central server. Extensive works have studied the performance guarantee of the global model, however, it is still unclear how each individual client in\ufb02uences the collaborative training process. In this work, we de\ufb01ned a new notion, called Fed-In\ufb02uence , to quantify this in\ufb02uence over the model parameters, and proposed an effective and ef\ufb01cient algorithm to estimate this metric. In particular, our design satis\ufb01es several desirable properties: (1) it requires neither retraining nor re-tracing, adding only linear computational overhead to clients and the server; (2) it strictly maintains the tenets of federated learning, without revealing any client\u2019s local private data; and (3) it works well on both convex and non-convex loss functions, and does not require the \ufb01nal model to be optimal. Empirical results on a synthetic dataset and the FEMNIST dataset demonstrate that our estimation method can approximate Fed-In\ufb02uence with small bias. Further, we show an application of Fed-In\ufb02uence in model debugging."
                    },
                    {
                        "title": "Successive Over Relaxation Recurrent Confidence Inference Network Based on Linear Extrapolation",
                        "abstract": "It is very important to be able to deduce an unknown conclusion from one or several known premises when solving a logical inference. The existing inference methods or models have certain logical inference abilities. However, because of the diversity of the forms of problems and the complexity of the derivation process, the scope of applying these methods is limited; this means the inference results are not ideal. Therefore, this paper proposes a new neural network model to solve the logic inference problem found in calculus. By using the successive over relaxation (SOR) method and the principle of recurrent confidence, the recurrent confidence inference network (RCI-Net) is built to solve the inference problem. The network simulates the solving process of the inference problem. Based on the known premise of this problem, it is calculated step by step so that the result of the calculation becomes gradually closer to the answer. At the same time, to make RCI-Net have stronger logical inference ability, our team uses the half mean squared error (HMSE) to construct the loss function of the model, improving the training efficiency of the model and preventing training collapse caused by the loss value exceeding the system\u2019s value range. Our team takes Sudoku reasoning problem as an example to carry out experiments. The results show that when the number of prompts of the reasoning problem is 17, the accuracy of the test set model can reach 99.67%, which is 3.07% higher than the existing models. It proves that the algorithm has better effect than the existing methods in solving logical reasoning problems."
                    }
                ]
            },
            "d7324691-88dc-40b6-b58f-b5fced2bc4b8": {
                "pk": "d7324691-88dc-40b6-b58f-b5fced2bc4b8",
                "name": "Eric Gan",
                "collaborators": [
                    "Baharan Mirzasoleiman",
                    "Dang Nguyen",
                    "Paymon Haddad",
                    "Lauren Watson",
                    "Mohan Dantam",
                    "Rik Sarkar",
                    "Yu Yang",
                    "G. Dziugaite",
                    "Yihao Xue",
                    "S. Joshi"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Generalization",
                    "Contrastive Learning"
                ],
                "publications": [
                    {
                        "title": "Make the Most of Your Data: Changing the Training Data Distribution to Improve In-distribution Generalization Performance",
                        "abstract": "Can we modify the training data distribution to encourage the underlying optimization method toward finding solutions with superior generalization performance on in-distribution data? In this work, we approach this question for the first time by comparing the inductive bias of gradient descent (GD) with that of sharpness-aware minimization (SAM). By studying a two-layer CNN, we prove that SAM learns easy and difficult features more uniformly, particularly in early epochs. That is, SAM is less susceptible to simplicity bias compared to GD. Based on this observation, we propose USEFUL, an algorithm that clusters examples based on the network output early in training and upsamples examples with no easy features to alleviate the pitfalls of the simplicity bias. We show empirically that modifying the training data distribution in this way effectively improves the generalization performance on the original data distribution when training with (S)GD by mimicking the training dynamics of SAM. Notably, we demonstrate that our method can be combined with SAM and existing data augmentation strategies to achieve, to the best of our knowledge, state-of-the-art performance for training ResNet18 on CIFAR10, STL10, CINIC10, Tiny-ImageNet; ResNet34 on CIFAR100; and VGG19 and DenseNet121 on CIFAR10."
                    },
                    {
                        "title": "Inference and Interference: The Role of Clipping, Pruning and Loss Landscapes in Differentially Private Stochastic Gradient Descent",
                        "abstract": "Differentially private stochastic gradient descent (DP-SGD) is known to have poorer training and test performance on large neural networks, compared to ordinary stochastic gradient descent (SGD). In this paper, we perform a detailed study and comparison of the two processes and unveil several new insights. By comparing the behavior of the two processes separately in early and late epochs, we find that while DP-SGD makes slower progress in early stages, it is the behavior in the later stages that determines the end result. This separate analysis of the clipping and noise addition steps of DP-SGD shows that while noise introduces errors to the process, gradient descent can recover from these errors when it is not clipped, and clipping appears to have a larger impact than noise. These effects are amplified in higher dimensions (large neural networks), where the loss basin occupies a lower dimensional space. We argue theoretically and using extensive experiments that magnitude pruning can be a suitable dimension reduction technique in this regard, and find that heavy pruning can improve the test accuracy of DPSGD."
                    },
                    {
                        "title": "Identifying Spurious Biases Early in Training through the Lens of Simplicity Bias",
                        "abstract": "Neural networks trained with (stochastic) gradient descent have an inductive bias towards learning simpler solutions. This makes them highly prone to learning spurious correlations in the training data, that may not hold at test time. In this work, we provide the first theoretical analysis of the effect of simplicity bias on learning spurious correlations. Notably, we show that examples with spurious features are provably separable based on the model's output early in training. We further illustrate that if spurious features have a small enough noise-to-signal ratio, the network's output on the majority of examples is almost exclusively determined by the spurious features, leading to poor worst-group test accuracy. Finally, we propose SPARE, which identifies spurious correlations early in training and utilizes importance sampling to alleviate their effect. Empirically, we demonstrate that SPARE outperforms state-of-the-art methods by up to 21.1% in worst-group accuracy, while being up to 12x faster. We also show that SPARE is a highly effective but lightweight method to discover spurious correlations."
                    },
                    {
                        "title": "Which Features are Learnt by Contrastive Learning? On the Role of Simplicity Bias in Class Collapse and Feature Suppression",
                        "abstract": "Contrastive learning (CL) has emerged as a powerful technique for representation learning, with or without label supervision. However, supervised CL is prone to collapsing representations of subclasses within a class by not capturing all their features, and unsupervised CL may suppress harder class-relevant features by focusing on learning easy class-irrelevant features; both significantly compromise representation quality. Yet, there is no theoretical understanding of \\textit{class collapse} or \\textit{feature suppression} at \\textit{test} time. We provide the first unified theoretically rigorous framework to determine \\textit{which} features are learnt by CL. Our analysis indicate that, perhaps surprisingly, bias of (stochastic) gradient descent towards finding simpler solutions is a key factor in collapsing subclass representations and suppressing harder class-relevant features. Moreover, we present increasing embedding dimensionality and improving the quality of data augmentations as two theoretically motivated solutions to {feature suppression}. We also provide the first theoretical explanation for why employing supervised and unsupervised CL together yields higher-quality representations, even when using commonly-used stochastic gradient methods."
                    },
                    {
                        "title": "Seeing is Believing: Brain-Inspired Modular Training for Mechanistic Interpretability",
                        "abstract": "We introduce Brain-Inspired Modular Training (BIMT), a method for making neural networks more modular and interpretable. Inspired by brains, BIMT embeds neurons in a geometric space and augments the loss function with a cost proportional to the length of each neuron connection. We demonstrate that BIMT discovers useful modular neural networks for many simple tasks, revealing compositional structures in symbolic formulas, interpretable decision boundaries and features for classification, and mathematical structure in algorithmic datasets. The ability to directly see modules with the naked eye can complement current mechanistic interpretability strategies such as probes, interventions or staring at all weights."
                    }
                ]
            },
            "46465755-3058-4a95-8a36-d6496e897efb": {
                "pk": "46465755-3058-4a95-8a36-d6496e897efb",
                "name": "Jiayi Ni",
                "collaborators": [
                    "Siddharth Joshi",
                    "Baharan Mirzasoleiman"
                ],
                "domain": [
                    "Dataset Distillation",
                    "Self-Supervised Learning",
                    "Knowledge Distillation",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Dataset Distillation via Knowledge Distillation: Towards Efficient Self-Supervised Pre-Training of Deep Networks",
                        "abstract": "Dataset distillation (DD) generates small synthetic datasets that can efficiently train deep networks with a limited amount of memory and compute. Despite the success of DD methods for supervised learning, DD for self-supervised pre-training of deep models has remained unaddressed. Pre-training on unlabeled data is crucial for efficiently generalizing to downstream tasks with limited labeled data. In this work, we propose the first effective DD method for SSL pre-training. First, we show, theoretically and empirically, that naive application of supervised DD methods to SSL fails, due to the high variance of the SSL gradient. Then, we address this issue by relying on insights from knowledge distillation (KD) literature. Specifically, we train a small student model to match the representations of a larger teacher model trained with SSL. Then, we generate a small synthetic dataset by matching the training trajectories of the student models. As the KD objective has considerably lower variance than SSL, our approach can generate synthetic datasets that can successfully pre-train high-quality encoders. Through extensive experiments, we show that our distilled sets lead to up to 13% higher accuracy than prior work, on a variety of downstream tasks, in the presence of limited labeled data."
                    }
                ]
            },
            "062be91c-1082-4ea0-af87-e862787bfef0": {
                "pk": "062be91c-1082-4ea0-af87-e862787bfef0",
                "name": "Siddharth Joshi",
                "collaborators": [
                    "Baharan Mirzasoleiman",
                    "Arnav Jain",
                    "Ali Payani",
                    "Yihao Xue",
                    "Jiayi Ni",
                    "Dang Nguyen",
                    "Yu Yang",
                    "Wenhan Yang"
                ],
                "domain": [
                    "Dataset Distillation",
                    "Contrastive Learning",
                    "Multimodal Learning",
                    "Spurious Correlations"
                ],
                "publications": [
                    {
                        "title": "Dataset Distillation via Knowledge Distillation: Towards Efficient Self-Supervised Pre-Training of Deep Networks",
                        "abstract": "Dataset distillation (DD) generates small synthetic datasets that can efficiently train deep networks with a limited amount of memory and compute. Despite the success of DD methods for supervised learning, DD for self-supervised pre-training of deep models has remained unaddressed. Pre-training on unlabeled data is crucial for efficiently generalizing to downstream tasks with limited labeled data. In this work, we propose the first effective DD method for SSL pre-training. First, we show, theoretically and empirically, that naive application of supervised DD methods to SSL fails, due to the high variance of the SSL gradient. Then, we address this issue by relying on insights from knowledge distillation (KD) literature. Specifically, we train a small student model to match the representations of a larger teacher model trained with SSL. Then, we generate a small synthetic dataset by matching the training trajectories of the student models. As the KD objective has considerably lower variance than SSL, our approach can generate synthetic datasets that can successfully pre-train high-quality encoders. Through extensive experiments, we show that our distilled sets lead to up to 13% higher accuracy than prior work, on a variety of downstream tasks, in the presence of limited labeled data."
                    },
                    {
                        "title": "Data-Efficient Contrastive Language-Image Pretraining: Prioritizing Data Quality over Quantity",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) on large-scale image-caption datasets learns representations that can achieve remarkable zero-shot generalization. However, such models require a massive amount of pre-training data. Improving the quality of the pre-training data has been shown to be much more effective in improving CLIP's performance than increasing its volume. Nevertheless, finding small subsets of training data that provably generalize the best has remained an open question. In this work, we propose the first theoretically rigorous data selection method for CLIP. We show that subsets that closely preserve the cross-covariance of the images and captions of the full data provably achieve a superior generalization performance. Our extensive experiments on ConceptualCaptions3M and ConceptualCaptions12M demonstrate that subsets found by \\method\\ achieve over 2.7x and 1.4x the accuracy of the next best baseline on ImageNet and its shifted versions. Moreover, we show that our subsets obtain 1.5x the average accuracy across 11 downstream datasets, of the next best baseline. The code is available at: https://github.com/BigML-CS-UCLA/clipcov-data-efficient-clip."
                    },
                    {
                        "title": "Understanding the Robustness of Multi-modal Contrastive Learning to Distribution Shift",
                        "abstract": "Recently, multimodal contrastive learning (MMCL) approaches, such as CLIP, have achieved a remarkable success in learning representations that are robust against distribution shift and generalize to new domains. Despite the empirical success, the mechanism behind learning such generalizable representations is not understood. In this work, we rigorously analyze this problem and uncover two mechanisms behind MMCL's robustness: \\emph{intra-class contrasting}, which allows the model to learn features with a high variance, and \\emph{inter-class feature sharing}, where annotated details in one class help learning other classes better. Both mechanisms prevent spurious features that are over-represented in the training data to overshadow the generalizable core features. This yields superior zero-shot classification accuracy under distribution shift. Furthermore, we theoretically demonstrate the benefits of using rich captions on robustness and explore the effect of annotating different types of details in the captions. We validate our theoretical findings through experiments, including a well-designed synthetic experiment and an experiment involving training CLIP models on MSCOCO/Conceptual Captions and evaluating them on shifted ImageNets."
                    },
                    {
                        "title": "Towards Mitigating more Challenging Spurious Correlations: A Benchmark&New Datasets",
                        "abstract": "Deep neural networks often exploit non-predictive features that are spuriously correlated with class labels, leading to poor performance on groups of examples without such features. Despite the growing body of recent works on remedying spurious correlations, the lack of a standardized benchmark hinders reproducible evaluation and comparison of the proposed solutions. To address this, we present SpuCo, a python package with modular implementations of state-of-the-art solutions enabling easy and reproducible evaluation of current methods. Using SpuCo, we demonstrate the limitations of existing datasets and evaluation schemes in validating the learning of predictive features over spurious ones. To overcome these limitations, we propose two new vision datasets: (1) SpuCoMNIST, a synthetic dataset that enables simulating the effect of real world data properties e.g. difficulty of learning spurious feature, as well as noise in the labels and features; (2) SpuCoAnimals, a large-scale dataset curated from ImageNet that captures spurious correlations in the wild much more closely than existing datasets. These contributions highlight the shortcomings of current methods and provide a direction for future research in tackling spurious correlations. SpuCo, containing the benchmark and datasets, can be found at https://github.com/BigML-CS-UCLA/SpuCo, with detailed documentation available at https://spuco.readthedocs.io/en/latest/."
                    },
                    {
                        "title": "Data-E \ufffd cient Contrastive Language-Image Pretraining: Prioritizing Data Quality over Quantity",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) on large-scale image-caption datasets learns representations that can achieve remarkable zero-shot generalization. However, such models require a massive amount of pre-training data. Improving the quality of the pre-training data has been shown to be much more e \u21b5 ective in improving CLIP\u2019s performance than increasing its volume. Nevertheless, \ufb01nding small subsets of training data that provably generalize the best has remained an open question. In this work, we propose the \ufb01rst theoretically rigorous data selection method for CLIP. We show that subsets that closely preserve the cross-covariance of the images and captions of the full data provably achieve a superior generalization performance. Our extensive experiments on ConceptualCaptions3M and ConceptualCap-tions12M demonstrate that subsets found by ClipCov achieve over 2.7x and 1.4x the accuracy of the next best baseline on ImageNet and its shifted versions. Moreover, we show that our subsets obtain 1.5x the average accuracy across 11 downstream datasets, of the next best baseline. The code is available at: https://github.com/BigML-CS-UCLA/ clipcov-data-efficient-clip ."
                    }
                ]
            },
            "7e9d2938-da6c-4cc9-857f-598d0f172097": {
                "pk": "7e9d2938-da6c-4cc9-857f-598d0f172097",
                "name": "Baharan Mirzasoleiman",
                "collaborators": [
                    "Yu Yang",
                    "Dang Nguyen",
                    "Wenhan Yang",
                    "Siddharth Joshi",
                    "Yihao Xue",
                    "S. Joshi",
                    "Ali Payani",
                    "Eric Gan",
                    "Jingdong Gao",
                    "Rathul Anand"
                ],
                "domain": [
                    "Self-Supervised Learning",
                    "Data Efficiency",
                    "Contrastive Learning",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "Mini-batch Coresets for Memory-efficient Training of Large Language Models",
                        "abstract": "Training with larger mini-batches improves the convergence rate and can yield superior performance. However, training with large mini-batches becomes prohibitive for Large Language Models (LLMs), due to the large GPU memory requirement. To address this problem, an effective approach is finding small mini-batch coresets that closely match the gradient of larger mini-batches. However, this approach becomes infeasible and ineffective for LLMs, due to the highly imbalanced nature of the sources in language data, use of the Adam optimizer, and the very large gradient dimensionality of LLMs. In this work, we address the above challenges by proposing Coresets for Training LLMs (CoLM). First, we show that mini-batch coresets found by gradient matching do not contain representative examples of the small sources w.h.p., and thus including all examples of the small sources in the mini-batch coresets is crucial for optimal performance. Second, we normalize the gradients by their historical exponential to find mini-batch coresets for training with Adam. Finally, we leverage zeroth-order methods to find smooth gradient of the last V -projection matrix and sparsify it to keep the dimensions with the largest normalized gradient magnitude. We apply CoLM to fine-tuning Phi-2, Phi-3, and Zephyr with LoRA on MathInstruct and SuperGLUE benchmark. Remarkably, CoLM reduces the memory requirement of fine-tuning by 2x and even outperforms training with 4x larger mini-batches. Notably, CoLM easily stack with existing memory-efficient training methods, such as LoRA."
                    },
                    {
                        "title": "Dataset Distillation via Knowledge Distillation: Towards Efficient Self-Supervised Pre-Training of Deep Networks",
                        "abstract": "Dataset distillation (DD) generates small synthetic datasets that can efficiently train deep networks with a limited amount of memory and compute. Despite the success of DD methods for supervised learning, DD for self-supervised pre-training of deep models has remained unaddressed. Pre-training on unlabeled data is crucial for efficiently generalizing to downstream tasks with limited labeled data. In this work, we propose the first effective DD method for SSL pre-training. First, we show, theoretically and empirically, that naive application of supervised DD methods to SSL fails, due to the high variance of the SSL gradient. Then, we address this issue by relying on insights from knowledge distillation (KD) literature. Specifically, we train a small student model to match the representations of a larger teacher model trained with SSL. Then, we generate a small synthetic dataset by matching the training trajectories of the student models. As the KD objective has considerably lower variance than SSL, our approach can generate synthetic datasets that can successfully pre-train high-quality encoders. Through extensive experiments, we show that our distilled sets lead to up to 13% higher accuracy than prior work, on a variety of downstream tasks, in the presence of limited labeled data."
                    },
                    {
                        "title": "Data-Efficient Contrastive Language-Image Pretraining: Prioritizing Data Quality over Quantity",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) on large-scale image-caption datasets learns representations that can achieve remarkable zero-shot generalization. However, such models require a massive amount of pre-training data. Improving the quality of the pre-training data has been shown to be much more effective in improving CLIP's performance than increasing its volume. Nevertheless, finding small subsets of training data that provably generalize the best has remained an open question. In this work, we propose the first theoretically rigorous data selection method for CLIP. We show that subsets that closely preserve the cross-covariance of the images and captions of the full data provably achieve a superior generalization performance. Our extensive experiments on ConceptualCaptions3M and ConceptualCaptions12M demonstrate that subsets found by \\method\\ achieve over 2.7x and 1.4x the accuracy of the next best baseline on ImageNet and its shifted versions. Moreover, we show that our subsets obtain 1.5x the average accuracy across 11 downstream datasets, of the next best baseline. The code is available at: https://github.com/BigML-CS-UCLA/clipcov-data-efficient-clip."
                    },
                    {
                        "title": "A Deep Learning Framework for Wireless Radiation Field Reconstruction and Channel Prediction",
                        "abstract": "We present NeWRF, a deep learning framework for predicting wireless channels. Wireless channel prediction is a long-standing problem in the wireless community and is a key technology for improving the coverage of wireless network deployments. Today, a wireless deployment is evaluated by a site survey which is a cumbersome process requiring an experienced engineer to perform extensive channel measurements. To reduce the cost of site surveys, we develop NeWRF, which is based on recent advances in Neural Radiance Fields (NeRF). NeWRF trains a neural network model with a sparse set of channel measurements, and predicts the wireless channel accurately at any location in the site. We introduce a series of techniques that integrate wireless propagation properties into the NeRF framework to account for the fundamental differences between the behavior of light and wireless signals. We conduct extensive evaluations of our framework and show that our approach can accurately predict channels at unvisited locations with significantly lower measurement density than prior state-of-the-art"
                    },
                    {
                        "title": "Make the Most of Your Data: Changing the Training Data Distribution to Improve In-distribution Generalization Performance",
                        "abstract": "Can we modify the training data distribution to encourage the underlying optimization method toward finding solutions with superior generalization performance on in-distribution data? In this work, we approach this question for the first time by comparing the inductive bias of gradient descent (GD) with that of sharpness-aware minimization (SAM). By studying a two-layer CNN, we prove that SAM learns easy and difficult features more uniformly, particularly in early epochs. That is, SAM is less susceptible to simplicity bias compared to GD. Based on this observation, we propose USEFUL, an algorithm that clusters examples based on the network output early in training and upsamples examples with no easy features to alleviate the pitfalls of the simplicity bias. We show empirically that modifying the training data distribution in this way effectively improves the generalization performance on the original data distribution when training with (S)GD by mimicking the training dynamics of SAM. Notably, we demonstrate that our method can be combined with SAM and existing data augmentation strategies to achieve, to the best of our knowledge, state-of-the-art performance for training ResNet18 on CIFAR10, STL10, CINIC10, Tiny-ImageNet; ResNet34 on CIFAR100; and VGG19 and DenseNet121 on CIFAR10."
                    },
                    {
                        "title": "SmallToLarge (S2L): Scalable Data Selection for Fine-tuning Large Language Models by Summarizing Training Trajectories of Small Models",
                        "abstract": "Despite the effectiveness of data selection for large language models (LLMs) during pretraining and instruction fine-tuning phases, improving data efficiency in supervised fine-tuning (SFT) for specialized domains poses significant challenges due to the complexity of fine-tuning data. To bridge this gap, we introduce an effective and scalable data selection method for SFT, SmallToLarge (S2L), which leverages training trajectories from small models to guide the data selection for larger models. We demonstrate through extensive experiments that S2L significantly improves data efficiency in SFT for mathematical problem-solving, reducing the training data to just 11% of the original MathInstruct dataset (Yue et al., 2023) to match full dataset performance while outperforming state-of-the-art data selection algorithms by an average of 4.7% across 6 in- and out-domain evaluation datasets. Remarkably, selecting only 50K data for SFT, S2L achieves a 32.7% accuracy on the most challenging MATH (Hendrycks et al., 2021) benchmark, improving Phi-2 (Li et al., 2023b) by 16.6%. In clinical text summarization on the MIMIC-III dataset (Johnson et al., 2016), S2L again outperforms training on the full dataset using only 50% of the data. Notably, S2L can perform data selection using a reference model 40x smaller than the target model, proportionally reducing the cost of data selection."
                    },
                    {
                        "title": "Understanding the Robustness of Multi-modal Contrastive Learning to Distribution Shift",
                        "abstract": "Recently, multimodal contrastive learning (MMCL) approaches, such as CLIP, have achieved a remarkable success in learning representations that are robust against distribution shift and generalize to new domains. Despite the empirical success, the mechanism behind learning such generalizable representations is not understood. In this work, we rigorously analyze this problem and uncover two mechanisms behind MMCL's robustness: \\emph{intra-class contrasting}, which allows the model to learn features with a high variance, and \\emph{inter-class feature sharing}, where annotated details in one class help learning other classes better. Both mechanisms prevent spurious features that are over-represented in the training data to overshadow the generalizable core features. This yields superior zero-shot classification accuracy under distribution shift. Furthermore, we theoretically demonstrate the benefits of using rich captions on robustness and explore the effect of annotating different types of details in the captions. We validate our theoretical findings through experiments, including a well-designed synthetic experiment and an experiment involving training CLIP models on MSCOCO/Conceptual Captions and evaluating them on shifted ImageNets."
                    },
                    {
                        "title": "Ordering for Non-Replacement SGD",
                        "abstract": "One approach for reducing run time and improving efficiency of machine learning is to reduce the convergence rate of the optimization algorithm used. Shuffling is an algorithm technique that is widely used in machine learning, but it only started to gain attention theoretically in recent years. With different convergence rates developed for random shuffling and incremental gradient descent, we seek to find an ordering that can improve the convergence rates for the non-replacement form of the algorithm. Based on existing bounds of the distance between the optimal and current iterate, we derive an upper bound that is dependent on the gradients at the beginning of the epoch. Through analysis of the bound, we are able to develop optimal orderings for constant and decreasing step sizes for strongly convex and convex functions. We further test and verify our results through experiments on synthesis and real data sets. In addition, we are able to combine the ordering with mini-batch and further apply it to more complex neural networks, which show promising results."
                    },
                    {
                        "title": "Inference and Interference: The Role of Clipping, Pruning and Loss Landscapes in Differentially Private Stochastic Gradient Descent",
                        "abstract": "Differentially private stochastic gradient descent (DP-SGD) is known to have poorer training and test performance on large neural networks, compared to ordinary stochastic gradient descent (SGD). In this paper, we perform a detailed study and comparison of the two processes and unveil several new insights. By comparing the behavior of the two processes separately in early and late epochs, we find that while DP-SGD makes slower progress in early stages, it is the behavior in the later stages that determines the end result. This separate analysis of the clipping and noise addition steps of DP-SGD shows that while noise introduces errors to the process, gradient descent can recover from these errors when it is not clipped, and clipping appears to have a larger impact than noise. These effects are amplified in higher dimensions (large neural networks), where the loss basin occupies a lower dimensional space. We argue theoretically and using extensive experiments that magnitude pruning can be a suitable dimension reduction technique in this regard, and find that heavy pruning can improve the test accuracy of DPSGD."
                    },
                    {
                        "title": "Data-Efficient Contrastive Self-supervised Learning: Most Beneficial Examples for Supervised Learning Contribute the Least",
                        "abstract": "Self-supervised learning (SSL) learns high-quality representations from large pools of unlabeled training data. As datasets grow larger, it becomes crucial to identify the examples that contribute the most to learning such representations. This enables efficient SSL by reducing the volume of data required. Nevertheless, quantifying the value of examples for SSL has remained an open question. In this work, we address this problem for the first time, by proving that examples that contribute the most to contrastive SSL are those that have the most similar augmentations to other examples, in expectation. We provide rigorous guarantees for the generalization performance of contrastive learning on such subsets. Through extensive experiments, we show that we can safely exclude 20% of examples from CIFAR100 and 40% from STL10 and TinyImageNet, without affecting downstream task performance. In general, subsets selected by our method outperform random subsets by over 3% across these datasets. Interestingly, we also discover the subsets that contribute the most to contrastive learning are those that contribute the least to supervised learning. Code available at https://github.com/bigml-cs-ucla/sas-data-efficient-contrastive-learning."
                    },
                    {
                        "title": "Few-shot Adaptation to Distribution Shifts By Mixing Source and Target Embeddings",
                        "abstract": "Pretrained machine learning models need to be adapted to distribution shifts when deployed in new target environments. When obtaining labeled data from the target distribution is expensive, few-shot adaptation with only a few examples from the target distribution becomes essential. In this work, we propose MixPro, a lightweight and highly data-efficient approach for few-shot adaptation. MixPro first generates a relatively large dataset by mixing (linearly combining) pre-trained embeddings of large source data with those of the few target examples. This process preserves important features of both source and target distributions, while mitigating the specific noise in the small target data. Then, it trains a linear classifier on the mixed embeddings to effectively adapts the model to the target distribution without overfitting the small target data. Theoretically, we demonstrate the advantages of MixPro over previous methods. Our experiments, conducted across various model architectures on 8 datasets featuring different types of distribution shifts, reveal that MixPro can outperform baselines by up to 7\\%, with only 2-4 target examples."
                    },
                    {
                        "title": "Data Distillation Can Be Like Vodka: Distilling More Times For Better Quality",
                        "abstract": "Dataset distillation aims to minimize the time and memory needed for training deep networks on large datasets, by creating a small set of synthetic images that has a similar generalization performance to that of the full dataset. However, current dataset distillation techniques fall short, showing a notable performance gap when compared to training on the original data. In this work, we are the first to argue that using just one synthetic subset for distillation will not yield optimal generalization performance. This is because the training dynamics of deep networks drastically change during the training. Hence, multiple synthetic subsets are required to capture the training dynamics at different phases of training. To address this issue, we propose Progressive Dataset Distillation (PDD). PDD synthesizes multiple small sets of synthetic images, each conditioned on the previous sets, and trains the model on the cumulative union of these subsets without requiring additional training time. Our extensive experiments show that PDD can effectively improve the performance of existing dataset distillation methods by up to 4.3%. In addition, our method for the first time enable generating considerably larger synthetic datasets."
                    },
                    {
                        "title": "Better Safe than Sorry: Pre-training CLIP against Targeted Data Poisoning and Backdoor Attacks",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) on large image-caption datasets has achieved remarkable success in zero-shot classification and enabled transferability to new domains. However, CLIP is extremely more vulnerable to targeted data poisoning and backdoor attacks, compared to supervised learning. Perhaps surprisingly, poisoning 0.0001% of CLIP pre-training data is enough to make targeted data poisoning attacks successful. This is four orders of magnitude smaller than what is required to poison supervised models. Despite this vulnerability, existing methods are very limited in defending CLIP models during pre-training. In this work, we propose a strong defense, SAFECLIP, to safely pre-train CLIP against targeted data poisoning and backdoor attacks. SAFECLIP warms up the model by applying unimodal contrastive learning (CL) on image and text modalities separately. Then, it divides the data into safe and risky sets, by applying a Gaussian Mixture Model to the cosine similarity of image-caption pair representations. SAFECLIP pre-trains the model by applying the CLIP loss to the safe set and applying unimodal CL to image and text modalities of the risky set separately. By gradually increasing the size of the safe set during pre-training, SAFECLIP effectively breaks targeted data poisoning and backdoor attacks without harming the CLIP performance. Our extensive experiments on CC3M, Visual Genome, and MSCOCO demonstrate that SAFECLIP significantly reduces the success rate of targeted data poisoning attacks from 93.75% to 0% and that of various backdoor attacks from up to 100% to 0%, without harming CLIP's performance."
                    },
                    {
                        "title": "Data-Efficient Contrastive Self-supervised Learning: Easy Examples Contribute the Most",
                        "abstract": "Self-supervised learning (SSL) learns high-quality representations from large pools of unlabeled training data. As datasets grow larger, it becomes crucial to identify the examples that contribute the most to learning such representations. This enables ef\ufb01cient SSL by reducing the volume of data required for learning high-quality representations. Nevertheless, quantifying the value of examples for SSL has remained an open question. In this work, we address this for the \ufb01rst time, by proving that examples that contribute the most to contrastive SSL are those that have the most similar augmentations to other examples, in expectation. We provide rigorous guarantees for the generalization performance of SSL on such subsets. Empirically, we discover, perhaps surprisingly, the subsets that contribute the most to SSL are those that contribute the least to supervised learning. Through extensive experiments, we show that our subsets outperform random sub-sets by more than 3% on CIFAR100, CIFAR10, and STL10. Interestingly, we also \ufb01nd that we can safely exclude 20% of examples from CIFAR100 and 40% from STL10, without affecting down-stream task performance."
                    },
                    {
                        "title": "NeSSA: Near-Storage Data Selection for Accelerated Machine Learning Training",
                        "abstract": "Large-scale machine learning (ML) models rely on extremely large datasets to learn their exponentially growing number of parameters. While these models achieve unprecedented success, the increase in training time and hardware resources required is unsustainable. Further, we find that as dataset sizes increase, data movement becomes a significant component of overall training time. We propose NeSSA, a novel SmartSSD+GPU training architecture to intelligently select important subsets of large datasets near-storage, such that training on the subset mimics training on the full dataset with a very small loss in accuracy. To the best of our knowledge, this is the first work to propose such a near-storage data selection model for efficient ML training. We have evaluated our method for the CIFAR-10, SVHN, CINIC-10, CIFAR-100, TinyImageNet, and ImageNet-100 datasets. We also test across ResNet-20, ResNet-18, and ResNet-50 models."
                    },
                    {
                        "title": "High Probability Bounds for Stochastic Continuous Submodular Maximization",
                        "abstract": "We consider maximization of stochastic monotone continuous submodular functions (CSF) with a diminishing return property. Existing algorithms only guarantee the performance \\textit{in expectation}, and do not bound the probability of getting a bad solution. This implies that for a particular run of the algorithms, the solution may be much worse than the provided guarantee in expectation. In this paper, we first empirically verify that this is indeed the case. Then, we provide the first \\textit{high-probability} analysis of the existing methods for stochastic CSF maximization, namely PGA, boosted PGA, SCG, and SCG++. Finally, we provide an improved high-probability bound for SCG, under slightly stronger assumptions, with a better convergence rate than that of the expected solution. Through extensive experiments on non-concave quadratic programming (NQP) and optimal budget allocation, we confirm the validity of our bounds and show that even in the worst-case, PGA converges to $OPT/2$, and boosted PGA, SCG, SCG++ converge to $(1 - 1/e)OPT$, but at a slower rate than that of the expected solution."
                    },
                    {
                        "title": "Mitigating Spurious Correlations in Multi-modal Models during Fine-tuning",
                        "abstract": "Spurious correlations that degrade model generalization or lead the model to be right for the wrong reasons are one of the main robustness concerns for real-world deployments. However, mitigating these correlations during pre-training for large-scale models can be costly and impractical, particularly for those without access to high-performance computing resources. This paper proposes a novel approach to address spurious correlations during fine-tuning for a given domain of interest. With a focus on multi-modal models (e.g., CLIP), the proposed method leverages different modalities in these models to detect and explicitly set apart spurious attributes from the affected class, achieved through a multi-modal contrastive loss function that expresses spurious relationships through language. Our experimental results and in-depth visualizations on CLIP show that such an intervention can effectively i) improve the model's accuracy when spurious attributes are not present, and ii) directs the model's activation maps towards the actual class rather than the spurious attribute when present. In particular, on the Waterbirds dataset, our algorithm achieved a worst-group accuracy 23% higher than ERM on CLIP with a ResNet-50 backbone, and 32% higher on CLIP with a ViT backbone, while maintaining the same average accuracy as ERM."
                    },
                    {
                        "title": "Towards Mitigating Spurious Correlations in the Wild: A Benchmark & a more Realistic Dataset",
                        "abstract": "Deep neural networks often exploit non-predictive features that are spuriously correlated with class labels, leading to poor performance on groups of examples without such features. Despite the growing body of recent works on remedying spurious correlations, the lack of a standardized benchmark hinders reproducible evaluation and comparison of the proposed solutions. To address this, we present S PU C O , a python package with modular implementations of state-of-the-art solutions enabling easy and reproducible evaluation of current methods. Using S PU C O , we demonstrate the limitations of existing datasets and evaluation schemes in validating the learning of predictive features over spurious ones. To overcome these limitations, we propose two new vision datasets: (1) S PU C O MNIST, a synthetic dataset that enables simulating the effect of real world data properties e.g. difficulty of learning spurious feature, as well as noise in the labels and features; (2) S PU C O A NIMALS , a large-scale dataset curated from ImageNet that captures spurious correlations in the wild much more closely than existing datasets. These contributions highlight the shortcomings of current methods and provide a direction for future research in tackling spurious correlations. S PU C O , containing the benchmark and datasets, can be found at https://github.com/BigML-CS-UCLA/SpuCo , with detailed documentation available at https://spuco.readthedocs.io/en/latest/ ."
                    }
                ]
            }
        }
    },
    "2407.12034": {
        "paper_data": {
            "title": "Understanding Transformers via N-gram Statistics",
            "url": "http://arxiv.org/abs/2407.12034v1",
            "arxiv_id": "2407.12034",
            "authors": [
                "Timothy Nguyen"
            ],
            "abstract": "Transformer based large-language models (LLMs) display extreme proficiency with language yet a precise understanding of how they work remains elusive. One way of demystifying transformer predictions would be to describe how they depend on their context in terms of simple template functions. This paper takes a first step in this direction by considering families of functions (i.e. rules) formed out of simple N-gram based statistics of the training data. By studying how well these rulesets approximate transformer predictions, we obtain a variety of novel discoveries: a simple method to detect overfitting during training without using a holdout set, a quantitative measure of how transformers progress from learning simple to more complex statistical rules over the course of training, a model-variance criterion governing when transformer predictions tend to be described by N-gram rules, and insights into how well transformers can be approximated by N-gram rulesets in the limit where these rulesets become increasingly complex. In this latter direction, we find that for 78% of LLM next-token distributions on TinyStories, their top-1 predictions agree with those provided by our N-gram rulesets.",
            "introduction": "   1 Introduction  This paper is an attempt to answer the following   Question: How does a transformer-based large language model (LLM) make use of its context when predicting the next token?   Our approach proceeds via studying the statistical properties of training data. This is perhaps the most natural place to start even though it is not exhaustive (e.g. it does not include in-context learning (Brown et\u00a0al.,, 2020)). The reasons to understand LLM behavior in terms of the statistics of their training data are plenty. First, the functional form of how LLMs use their training data is not well-understood (though there has been progress on understanding memorization (Nasr et\u00a0al.,, 2023; Carlini et\u00a0al.,, 2023)). Second, the over-reliance of LLMs on training data statistics leads to brittleness (e.g. the \u201creversal curse\" (Berglund et\u00a0al.,, 2024)) and the perpetuation of dataset biases (Gallegos et\u00a0al.,, 2024). Understanding the nature of this statistical dependence can lead to improved and more informed dataset curation and training methods. Finally, in various scenarios, the performance of LLMs on downstream tasks are found to be correlated with frequency of relevant training data (Razeghi et\u00a0al.,, 2022; Elazar et\u00a0al.,, 2023; Kandpal et\u00a0al.,, 2023; Kang and Choi,, 2023). A better understanding of this phenomenon would allow better steering of models towards desired performance levels.   We can think of the complexity of an LLM next token prediction (regarded as a probability distribution over tokens) along two axes: form and selection. Form refers to the functional form of the prediction as a function of the context, e.g. whether the prediction is some explicit function of associated training data statistics (see Figure 1). Selection refers to which functional form, chosen from a set of functional templates, suitably describes the transformer prediction (supposing the choice set is sufficiently rich). As a first nontrivial step, one might hope that an approximate model for an LLM is that each of its next token predictions can be roughly described by simple statistical rules from the context (simple form) even if the mechanism for its rule selection remains hidden (complex selection)111It is important to emphasize that we seek a descriptive approximation of a transformer, rather than an explanatory one. A description merely requires that we can provide a post-hoc, per-instance approximation of transformer predictions in terms of an available rule; an explanation means we provide reasons for and thus can predict in advance why and when a particular rule approximates transformer predictions. Hence, we make the distinction between form (description) and selection (explanation).. This paper is an attempt to see how far this perspective can be pushed, and fortuitously we obtain additional insights for understanding LLM behavior along the way. The statistical rules we consider, which are based on N\ud835\udc41Nitalic_N-grams, are defined in Section 4, with Figure 1 showing some examples.   Figure 1: Illustration of rule approximation. Given a context, different N\ud835\udc41Nitalic_N-gram based rules formed out of the context will yield different next-token predictive distributions. In the above example, the context consists of three tokens. The first rule uses all three tokens of the context and makes a prediction based on the corresponding 4444-gram rule derived from the training",
            "references": [
                {
                    "title": "The Evolution of Statistical Induction Heads: In-Context Learning Markov Chains",
                    "abstract": "Large language models have the ability to generate text that mimics patterns in their inputs. We introduce a simple Markov Chain sequence modeling task in order to study how this in-context learning (ICL) capability emerges. In our setting, each example is sampled from a Markov chain drawn from a prior distribution over Markov chains. Transformers trained on this task form \\emph{statistical induction heads} which compute accurate next-token probabilities given the bigram statistics of the context. During the course of training, models pass through multiple phases: after an initial stage in which predictions are uniform, they learn to sub-optimally predict using in-context single-token statistics (unigrams); then, there is a rapid phase transition to the correct in-context bigram solution. We conduct an empirical and theoretical investigation of this multi-phase process, showing how successful learning results from the interaction between the transformer's layers, and uncovering evidence that the presence of the simpler unigram solution may delay formation of the final bigram solution. We examine how learning is affected by varying the prior distribution over Markov chains, and consider the generalization of our in-context learning of Markov chains (ICL-MC) task to $n$-grams for $n>2$."
                },
                {
                    "title": "Infini-gram: Scaling Unbounded n-gram Language Models to a Trillion Tokens",
                    "abstract": "Are $n$-gram language models still relevant in this era of neural large language models (LLMs)? Our answer is yes, and we showcase their values in both text analysis and improving neural LLMs. This was done by modernizing $n$-gram LMs in two aspects. First, we train them at the same data scale as neural LLMs -- 5 trillion tokens. This is the largest $n$-gram LM ever built. Second, existing $n$-gram LMs use small $n$ which hinders their performance; we instead allow $n$ to be arbitrarily large, by introducing a new $\\infty$-gram LM with backoff. Instead of pre-computing $n$-gram count tables (which would be very expensive), we develop an engine named infini-gram -- powered by suffix arrays -- that can compute $\\infty$-gram (as well as $n$-gram with arbitrary $n$) probabilities with millisecond-level latency. The $\\infty$-gram framework and infini-gram engine enable us to conduct many novel and interesting analyses of human-written and machine-generated text: we find that the $\\infty$-gram LM has fairly high accuracy for next-token prediction (47%), and can complement neural LLMs to greatly reduce their perplexity. When analyzing machine-generated text, we also observe irregularities in the machine--$\\infty$-gram agreement level with respect to the suffix length, which indicates deficiencies in neural LLM pretraining and the positional embeddings of Transformers."
                },
                {
                    "title": "Zoology: Measuring and Improving Recall in Efficient Language Models",
                    "abstract": "Attention-free language models that combine gating and convolutions are growing in popularity due to their efficiency and increasingly competitive performance. To better understand these architectures, we pretrain a suite of 17 attention and\"gated-convolution\"language models, finding that SoTA gated-convolution architectures still underperform attention by up to 2.1 perplexity points on the Pile. In fine-grained analysis, we find 82% of the gap is explained by each model's ability to recall information that is previously mentioned in-context, e.g.\"Hakuna Matata means no worries Hakuna Matata it means no\"$\\rightarrow$\"??\". On this task, termed\"associative recall\", we find that attention outperforms gated-convolutions by a large margin: a 70M parameter attention model outperforms a 1.4 billion parameter gated-convolution model on associative recall. This is surprising because prior work shows gated convolutions can perfectly solve synthetic tests for AR capability. To close the gap between synthetics and real language, we develop a new formalization of the task called multi-query associative recall (MQAR) that better reflects actual language. We perform an empirical and theoretical study of MQAR that elucidates differences in the parameter-efficiency of attention and gated-convolution recall. Informed by our analysis, we evaluate simple convolution-attention hybrids and show that hybrids with input-dependent sparse attention patterns can close 97.4% of the gap to attention, while maintaining sub-quadratic scaling. Our code is accessible at: https://github.com/HazyResearch/zoology."
                },
                {
                    "title": "Scalable Extraction of Training Data from (Production) Language Models",
                    "abstract": "This paper studies extractable memorization: training data that an adversary can efficiently extract by querying a machine learning model without prior knowledge of the training dataset. We show an adversary can extract gigabytes of training data from open-source language models like Pythia or GPT-Neo, semi-open models like LLaMA or Falcon, and closed models like ChatGPT. Existing techniques from the literature suffice to attack unaligned models; in order to attack the aligned ChatGPT, we develop a new divergence attack that causes the model to diverge from its chatbot-style generations and emit training data at a rate 150x higher than when behaving properly. Our methods show practical attacks can recover far more data than previously thought, and reveal that current alignment techniques do not eliminate memorization."
                },
                {
                    "title": "Impact of Co-occurrence on Factual Knowledge of Large Language Models",
                    "abstract": "Large language models (LLMs) often make factually incorrect responses despite their success in various applications. In this paper, we hypothesize that relying heavily on simple co-occurrence statistics of the pre-training corpora is one of the main factors that cause factual errors. Our results reveal that LLMs are vulnerable to the co-occurrence bias, defined as preferring frequently co-occurred words over the correct answer. Consequently, LLMs struggle to recall facts whose subject and object rarely co-occur in the pre-training dataset although they are seen during finetuning. We show that co-occurrence bias remains despite scaling up model sizes or finetuning. Therefore, we suggest finetuning on a debiased dataset to mitigate the bias by filtering out biased samples whose subject-object co-occurrence count is high. Although debiased finetuning allows LLMs to memorize rare facts in the training set, it is not effective in recalling rare facts unseen during finetuning. Further research in mitigation will help build reliable language models by preventing potential errors. The code is available at \\url{https://github.com/CheongWoong/impact_of_cooccurrence}."
                },
                {
                    "title": "The Reversal Curse: LLMs trained on \"A is B\" fail to learn \"B is A\"",
                    "abstract": "We expose a surprising failure of generalization in auto-regressive large language models (LLMs). If a model is trained on a sentence of the form\"A is B\", it will not automatically generalize to the reverse direction\"B is A\". This is the Reversal Curse. For instance, if a model is trained on\"Valentina Tereshkova was the first woman to travel to space\", it will not automatically be able to answer the question,\"Who was the first woman to travel to space?\". Moreover, the likelihood of the correct answer (\"Valentina Tershkova\") will not be higher than for a random name. Thus, models do not generalize a prevalent pattern in their training set: if\"A is B\"occurs,\"B is A\"is more likely to occur. It is worth noting, however, that if\"A is B\"appears in-context, models can deduce the reverse relationship. We provide evidence for the Reversal Curse by finetuning GPT-3 and Llama-1 on fictitious statements such as\"Uriah Hawthorne is the composer of Abyssal Melodies\"and showing that they fail to correctly answer\"Who composed Abyssal Melodies?\". The Reversal Curse is robust across model sizes and model families and is not alleviated by data augmentation. We also evaluate ChatGPT (GPT-3.5 and GPT-4) on questions about real-world celebrities, such as\"Who is Tom Cruise's mother? [A: Mary Lee Pfeiffer]\"and the reverse\"Who is Mary Lee Pfeiffer's son?\". GPT-4 correctly answers questions like the former 79% of the time, compared to 33% for the latter. Code available at: https://github.com/lukasberglund/reversal_curse."
                },
                {
                    "title": "Sudden Drops in the Loss: Syntax Acquisition, Phase Transitions, and Simplicity Bias in MLMs",
                    "abstract": "Most interpretability research in NLP focuses on understanding the behavior and features of a fully trained model. However, certain insights into model behavior may only be accessible by observing the trajectory of the training process. We present a case study of syntax acquisition in masked language models (MLMs) that demonstrates how analyzing the evolution of interpretable artifacts throughout training deepens our understanding of emergent behavior. In particular, we study Syntactic Attention Structure (SAS), a naturally emerging property of MLMs wherein specific Transformer heads tend to focus on specific syntactic relations. We identify a brief window in pretraining when models abruptly acquire SAS, concurrent with a steep drop in loss. This breakthrough precipitates the subsequent acquisition of linguistic capabilities. We then examine the causal role of SAS by manipulating SAS during training, and demonstrate that SAS is necessary for the development of grammatical capabilities. We further find that SAS competes with other beneficial traits during training, and that briefly suppressing SAS improves model quality. These findings offer an interpretation of a real-world example of both simplicity bias and breakthrough training dynamics."
                },
                {
                    "title": "Neurons in Large Language Models: Dead, N-gram, Positional",
                    "abstract": "We analyze a family of large language models in such a lightweight manner that can be done on a single GPU. Specifically, we focus on the OPT family of models ranging from 125m to 66b parameters and rely only on whether an FFN neuron is activated or not. First, we find that the early part of the network is sparse and represents many discrete features. Here, many neurons (more than 70% in some layers of the 66b model) are\"dead\", i.e. they never activate on a large collection of diverse data. At the same time, many of the alive neurons are reserved for discrete features and act as token and n-gram detectors. Interestingly, their corresponding FFN updates not only promote next token candidates as could be expected, but also explicitly focus on removing the information about triggering them tokens, i.e., current input. To the best of our knowledge, this is the first example of mechanisms specialized at removing (rather than adding) information from the residual stream. With scale, models become more sparse in a sense that they have more dead neurons and token detectors. Finally, some neurons are positional: them being activated or not depends largely (or solely) on position and less so (or not at all) on textual data. We find that smaller models have sets of neurons acting as position range indicators while larger models operate in a less explicit manner."
                },
                {
                    "title": "Bias and Fairness in Large Language Models: A Survey",
                    "abstract": "Abstract Rapid advancements of large language models (LLMs) have enabled the processing, understanding, and generation of human-like text, with increasing integration into systems that touch our social sphere. Despite this success, these models can learn, perpetuate, and amplify harmful social biases. In this article, we present a comprehensive survey of bias evaluation and mitigation techniques for LLMs. We first consolidate, formalize, and expand notions of social bias and fairness in natural language processing, defining distinct facets of harm and introducing several desiderata to operationalize fairness for LLMs. We then unify the literature by proposing three intuitive taxonomies, two for bias evaluation, namely, metrics and datasets, and one for mitigation. Our first taxonomy of metrics for bias evaluation disambiguates the relationship between metrics and evaluation datasets, and organizes metrics by the different levels at which they operate in a model: embeddings, probabilities, and generated text. Our second taxonomy of datasets for bias evaluation categorizes datasets by their structure as counterfactual inputs or prompts, and identifies the targeted harms and social groups; we also release a consolidation of publicly available datasets for improved access. Our third taxonomy of techniques for bias mitigation classifies methods by their intervention during pre-processing, in-training, intra-processing, and post-processing, with granular subcategories that elucidate research trends. Finally, we identify open problems and challenges for future work. Synthesizing a wide range of recent research, we aim to provide a clear guide of the existing literature that empowers researchers and practitioners to better understand and prevent the propagation of bias in LLMs."
                },
                {
                    "title": "TinyStories: How Small Can Language Models Be and Still Speak Coherent English?",
                    "abstract": "Language models (LMs) are powerful tools for natural language processing, but they often struggle to produce coherent and fluent text when they are small. Models with around 125M parameters such as GPT-Neo (small) or GPT-2 (small) can rarely generate coherent and consistent English text beyond a few words even after extensive training. This raises the question of whether the emergence of the ability to produce coherent English text only occurs at larger scales (with hundreds of millions of parameters or more) and complex architectures (with many layers of global attention). In this work, we introduce TinyStories, a synthetic dataset of short stories that only contain words that a typical 3 to 4-year-olds usually understand, generated by GPT-3.5 and GPT-4. We show that TinyStories can be used to train and evaluate LMs that are much smaller than the state-of-the-art models (below 10 million total parameters), or have much simpler architectures (with only one transformer block), yet still produce fluent and consistent stories with several paragraphs that are diverse and have almost perfect grammar, and demonstrate reasoning capabilities. We also introduce a new paradigm for the evaluation of language models: We suggest a framework which uses GPT-4 to grade the content generated by these models as if those were stories written by students and graded by a (human) teacher. This new paradigm overcomes the flaws of standard benchmarks which often requires the model's output to be very structures, and moreover provides a multidimensional score for the model, providing scores for different capabilities such as grammar, creativity and consistency. We hope that TinyStories can facilitate the development, analysis and research of LMs, especially for low-resource or specialized domains, and shed light on the emergence of language capabilities in LMs."
                },
                {
                    "title": "Large Language Models Struggle to Learn Long-Tail Knowledge",
                    "abstract": "The Internet contains a wealth of knowledge -- from the birthdays of historical figures to tutorials on how to code -- all of which may be learned by language models. However, while certain pieces of information are ubiquitous on the web, others appear extremely rarely. In this paper, we study the relationship between the knowledge memorized by large language models and the information in pre-training datasets scraped from the web. In particular, we show that a language model's ability to answer a fact-based question relates to how many documents associated with that question were seen during pre-training. We identify these relevant documents by entity linking pre-training datasets and counting documents that contain the same entities as a given question-answer pair. Our results demonstrate strong correlational and causal relationships between accuracy and relevant document count for numerous question answering datasets (e.g., TriviaQA), pre-training corpora (e.g., ROOTS), and model sizes (e.g., 176B parameters). Moreover, while larger models are better at learning long-tail knowledge, we estimate that today's models must be scaled by many orders of magnitude to reach competitive QA performance on questions with little support in the pre-training data. Finally, we show that retrieval-augmentation can reduce the dependence on relevant pre-training information, presenting a promising approach for capturing the long-tail."
                },
                {
                    "title": "Measuring Causal Effects of Data Statistics on Language Model's 'Factual' Predictions",
                    "abstract": "Large amounts of training data are one of the major reasons for the high performance of state-of-the-art NLP models. But what exactly in the training data causes a model to make a certain prediction? We seek to answer this question by providing a language for describing how training data influences predictions, through a causal framework. Importantly, our framework bypasses the need to retrain expensive models and allows us to estimate causal effects based on observational data alone. Addressing the problem of extracting factual knowledge from pretrained language models (PLMs), we focus on simple data statistics such as co-occurrence counts and show that these statistics do influence the predictions of PLMs, suggesting that such models rely on shallow heuristics. Our causal framework and our results demonstrate the importance of studying datasets and the benefits of causality for understanding NLP models."
                },
                {
                    "title": "Training Compute-Optimal Large Language Models",
                    "abstract": "We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over 400 language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, Chinchilla, that uses the same compute budget as Gopher but with 70B parameters and 4$\\times$ more more data. Chinchilla uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that Chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, Chinchilla reaches a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, greater than a 7% improvement over Gopher."
                },
                {
                    "title": "Quantifying Memorization Across Neural Language Models",
                    "abstract": "Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations."
                },
                {
                    "title": "The Grammar-Learning Trajectories of Neural Language Models",
                    "abstract": "The learning trajectories of linguistic phenomena in humans provide insight into linguistic representation, beyond what can be gleaned from inspecting the behavior of an adult speaker. To apply a similar approach to analyze neural language models (NLM), it is first necessary to establish that different models are similar enough in the generalizations they make. In this paper, we show that NLMs with different initialization, architecture, and training data acquire linguistic phenomena in a similar order, despite their different end performance. These findings suggest that there is some mutual inductive bias that underlies these models\u2019 learning of linguistic phenomena. Taking inspiration from psycholinguistics, we argue that studying this inductive bias is an opportunity to study the linguistic representation implicit in NLMs.Leveraging these findings, we compare the relative performance on different phenomena at varying learning stages with simpler reference models. Results suggest that NLMs exhibit consistent \u201cdevelopmental\u201d stages. Moreover, we find the learning trajectory to be approximately one-dimensional: given an NLM with a certain overall performance, it is possible to predict what linguistic generalizations it has already acquired.Initial analysis of these stages presents phenomena clusters (notably morphological ones), whose performance progresses in unison, suggesting a potential link between the generalizations behind them."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Large Language Models in Machine Translation",
                    "abstract": "Systems, methods, and computer program products for machine translation are provided. In some implementations a system is provided. The system includes a language model including a collection of n-grams from a corpus, each n-gram having a corresponding relative frequency in the corpus and an order n corresponding to a number of tokens in the n-gram, each n-gram corresponding to a backoff n-gram having an order of n-1 and a collection of backoff scores, each backoff score associated with an n-gram, the backoff score determined as a function of a backoff factor and a relative frequency of a corresponding backoff n-gram in the corpus."
                },
                {
                    "title": "Rule Extraction from Recurrent Neural Networks: ATaxonomy and Review",
                    "abstract": "Rule extraction (RE) from recurrent neural networks (RNNs) refers to finding models of the underlying RNN, typically in the form of finite state machines, that mimic the network to a satisfactory degree while having the advantage of being more transparent. RE from RNNs can be argued to allow a deeper and more profound form of analysis of RNNs than other, more or less ad hoc methods. RE may give us understanding of RNNs in the intermediate levels between quite abstract theoretical knowledge of RNNs as a class of computing devices and quantitative performance evaluations of RNN instantiations. The development of techniques for extraction of rules from RNNs has been an active field since the early 1990s. This article reviews the progress of this development and analyzes it in detail. In order to structure the survey and evaluate the techniques, a taxonomy specifically designed for this purpose has been developed. Moreover, important open research issues are identified that, if addressed properly, possibly can give the field a significant push forward."
                },
                {
                    "title": "A bit of progress in language modeling",
                    "abstract": "In the past several years, a number of different language modeling improvements over simple trigram models have been found, including caching, higher-order n -grams, skipping, interpolated Kneser?Ney smoothing, and clustering. We present explorations of variations on, or of the limits of, each of these techniques, including showing that sentence mixture models may have more potential. While all of these techniques have been studied separately, they have rarely been studied in combination. We compare a combination of all techniques together to a Katz smoothed trigram model with no count cutoffs. We achieve perplexity reductions between 38 and 50% (1 bit of entropy), depending on training data size, as well as a word error rate reduction of 8.9%. Our perplexity reductions are perhaps the highest reported compared to a fair baseline."
                },
                {
                    "title": "Improved backing-off for M-gram language modeling",
                    "abstract": "In stochastic language modeling, backing-off is a widely used method to cope with the sparse data problem. In case of unseen events this method backs off to a less specific distribution. In this paper we propose to use distributions which are especially optimized for the task of backing-off. Two different theoretical derivations lead to distributions which are quite different from the probability distributions that are usually used for backing-off. Experiments show an improvement of about 10% in terms of perplexity and 5% in terms of word error rate."
                },
                {
                    "title": "Impact of Pretraining Term Frequencies on Few-Shot Numerical Reasoning",
                    "abstract": ","
                },
                {
                    "title": "The Connectionist Scientist Game: Rule Extraction and Refinement in a Neural Network",
                    "abstract": "Clayton McMillan, Michael C. Mozer, & Paul SmolenskyCU-CS-530-91 May 1991Department of Computer Science andInstitute of Cognitive ScienceUniversity of ColoradoBoulder, CO 80309-0430 USA(303) 492-4103(303) 492-2844 Faxmozer@cs.colorado.eduTo appear in: Proceedings of the Thirteenth Annual Conference ofthe Cognitive Science Society, Hillsdale, NJ: Erlbaum. 1991."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does a transformer-based large language model (LLM) make use of its context when predicting the next token?\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding how LLMs utilize their training data statistics for next-token prediction is crucial for several reasons. First, it addresses the gap in knowledge regarding the functional form of LLM predictions, which is essential for improving model robustness and reducing brittleness, such as the \"reversal curse.\" Second, insights gained from this research can inform better dataset curation and training methods, ultimately leading to more reliable and fair AI systems. Additionally, correlating LLM performance on downstream tasks with the frequency of relevant training data can enhance model steering towards desired outcomes, thereby advancing both theoretical understanding and practical applications in natural language processing.\n\n**[Question 3] - Why is it hard?**  \nThe complexity of LLM next-token prediction arises from two axes: form and selection. The challenge lies in accurately modeling the functional form of predictions based on context, as well as determining which statistical rules best describe the transformer\u2019s behavior. Naive approaches may fail because they might overlook the intricate relationships between context and prediction, leading to oversimplified models that do not capture the underlying dynamics. Additionally, the hidden mechanisms of rule selection add another layer of complexity, making it difficult to derive a comprehensive understanding of LLM behavior.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either the memorization aspects of LLMs or the performance metrics on downstream tasks without adequately addressing the statistical dependencies in next-token prediction. Barriers include a lack of comprehensive frameworks to analyze the interplay between context and prediction, as well as insufficient exploration of the statistical rules that govern LLM behavior. This paper aims to fill these gaps by providing a descriptive approximation of transformer predictions through N-gram based statistical rules, which differs from prior work that may not have considered such a focused approach.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves analyzing the statistical properties of training data to approximate LLM next-token predictions using N-gram based rules. The dataset will consist of various text corpora used to train the LLM, and the evaluation metric will focus on the accuracy of next-token predictions based on the derived statistical rules. The expected outcomes include a clearer understanding of how context influences predictions, insights into the nature of statistical"
            }
        },
        "author_data": {
            "e4073c51-3706-4630-b780-fb4d5966650f": {
                "pk": "e4073c51-3706-4630-b780-fb4d5966650f",
                "name": "Timothy Nguyen",
                "collaborators": [
                    "Jaehoon Lee",
                    "Oleg Evnin",
                    "Roman Novak",
                    "Lechao Xiao",
                    "Zhourong Chen",
                    "Jiefeng Chen",
                    "Dilan Gorur",
                    "Arslan Chaudhry"
                ],
                "domain": [
                    "Mathematical Physics",
                    "Quantum Field Theory",
                    "Machine Learning",
                    "Differential Geometry"
                ],
                "publications": [
                    {
                        "title": "Anisotropic Function Spaces and Elliptic Boundary Value Problems",
                        "abstract": "In this paper, we study anisotropic Bessel potential and Besov spaces, where the anisotropy measures the extra amount of regularity in certain directions. Some basic properties of these spaces are established along with applications to elliptic boundary value problems."
                    },
                    {
                        "title": "Lagrangian correspondences and Donaldson's TQFT construction of the Seiberg-Witten invariants of 3-manifolds",
                        "abstract": "Using Morse-Bott techniques adapted to the gauge-theoretic setting, we show that the limiting boundary values of the space of finite energy monopoles on a connected 3-manifold with at least two cylindrical ends provides an immersed Lagrangian submanifold of the vortex moduli space at infinity. By studying the signed intersections of such Lagrangians, we supply the analytic details of Donaldson's TQFT construction of the Seiberg-Witten invariants of a closed 3-manifold."
                    },
                    {
                        "title": "A Response to Economics as Gauge Theory",
                        "abstract": "We provide an analysis of the recent work by Malaney-Weinstein on \"Economics as Gauge Theory\" presented on November 10, 2021 at the Money and Banking Workshop hosted by University of Chicago. In particular, we distill the technical mathematics used in their work into a form more suitable to a wider audience. Furthermore, we resolve the conjectures posed by Malaney-Weinstein, revealing that they provide no discernible value for the calculation of index numbers or rates of inflation. Our conclusion is that the main contribution of the Malaney-Weinstein work is that it provides a striking example of how to obscure simple concepts through an uneconomical use of gauge theory."
                    },
                    {
                        "title": "The Seiberg-Witten Equations on Manifolds with Boundary I: The Space of Monopoles and Their Boundary Values",
                        "abstract": "In this paper, we study the Seiberg-Witten equations on a compact 3-manifold with boundary. Solutions to these equations are called monopoles. Under some simple topological assumptions, we show that the solution space of all monopoles is a Banach manifold in suitable function space topologies. We then prove that the restriction of the space of monopoles to the boundary is a submersion onto a Lagrangian submanifold of the space of connections and spinors on the boundary. Both these spaces are infinite dimensional, even modulo gauge, since no boundary conditions are specified for the Seiberg-Witten equations on the 3-manifold. We study the analytic properties of these monopole spaces with an eye towards developing a monopole Floer theory for 3-manifolds with boundary, which we pursue in Part II."
                    },
                    {
                        "title": "The Seiberg-Witten Equations on Manifolds with Boundary II: Lagrangian Boundary Conditions for a Floer Theory",
                        "abstract": "In this paper, we study the Seiberg-Witten equations on the product R x Y, where Y is a compact 3-manifold with boundary. Following the approach of Salamon and Wehrheim in the instanton case, we impose Lagrangian boundary conditions for the Seiberg- Witten equations. The resulting equations we obtain constitute a nonlinear, nonlocal boundary value problem. We establish regularity, compactness, and Fredholm properties for the Seiberg- Witten equations supplied with Lagrangian boundary conditions arising from the monopole spaces studied in [20]. This work therefore serves as an analytic foundation for the construction of a monopole Floer theory for 3-manifolds with boundary."
                    },
                    {
                        "title": "Quantization of the Nonlinear Sigma Model Revisited",
                        "abstract": "We revisit the subject of perturbatively quantizing the nonlinear sigma model in two dimensions from a rigorous, mathematical point of view. Our main contribution is to make precise the cohomological problem of eliminating potential anomalies that may arise when trying to preserve symmetries under quantization. The symmetries we consider are twofold: (i) diffeomorphism covariance for a general target manifold; (ii) a transitive group of isometries when the target manifold is a homogeneous space. We show that there are no anomalies in case (i) and that (ii) is also anomaly-free under additional assumptions on the target homogeneous space, in agreement with the work of Friedan. We carry out some explicit computations for the $O(N)$-model. Finally, we show how a suitable notion of the renormalization group establishes the Ricci flow as the one loop renormalization group flow of the nonlinear sigma model."
                    },
                    {
                        "title": "Wilson Loop Area Law for 2D Yang-Mills in Generalized Axial Gauge",
                        "abstract": "We prove that Wilson loop expectation values for arbitrary simple closed contours obey an area law up to second order in perturbative two-dimensional Yang-Mills theory. Our analysis occurs within a general family of axial-like gauges, which include and interpolate between holomorphic gauge and the Wu-Mandelstam-Liebrandt light cone gauge. Our methods make use of the homotopy invariance properties of iterated integrals of closed one-forms, which allows us to evaluate the nontrivial integrals occurring at second order. We close with a discussion on complex gauge-fixing and deformation of integration cycles for holomorphic path integrals to shed light on some of the quantum field-theoretic underpinnings of our results."
                    },
                    {
                        "title": "Stochastic Feynman Rules for Yang-Mills Theory on the Plane",
                        "abstract": "We analyze quantum Yang-Mills theory on $\\mathbb{R}^2$ using a novel discretization method based on an algebraic analogue of stochastic calculus. Such an analogue involves working with \"Gaussian\" free fields whose covariance matrix is indefinite rather than positive definite. Specifically, we work with Lie-algebra valued fields on a lattice and exploit an approximate gauge-invariance that is restored when taking the continuum limit. This analysis is applied to show the equivalence between Wilson loop expectations computed using partial axial-gauge, complete axial-gauge, and the Migdal-Witten lattice formulation. As a consequence, we obtain intriguing Lie-theoretic identities involving heat kernels and iterated integrals."
                    },
                    {
                        "title": "The Perturbative Approach to Path Integrals: A Succinct Mathematical Treatment",
                        "abstract": "We study finite-dimensional integrals in a way that elucidates the mathematical meaning behind the formal manipulations of path integrals occurring in quantum field theory. This involves a proper understanding of how Wick's theorem allows one to evaluate integrals perturbatively, i.e., as a series expansion in a formal parameter irrespective of convergence properties. We establish invariance properties of such a Wick expansion under coordinate changes and the action of a Lie group of symmetries, and we use this to study essential features of path integral manipulations, including coordinate changes, Ward identities, Schwinger-Dyson equations, Faddeev-Popov gauge-fixing, and eliminating fields by their equation of motion. We also discuss the asymptotic nature of the Wick expansion and the implications this has for defining path integrals perturbatively and nonperturbatively."
                    },
                    {
                        "title": "Quantum Yang-Mills Theory in Two Dimensions: Exact versus Perturbative",
                        "abstract": "The standard Feynman diagrammatic approach to quantum field theories assumes that perturbation theory approximates the full quantum theory at small coupling even when a mathematically rigorous construction of the latter is absent. On the other hand, two-dimensional Yang-Mills theory is a rare (if not the only) example of a nonabelian (pure) gauge theory whose full quantum theory has a rigorous construction. Indeed, the theory can be formulated via a lattice approximation, from which Wilson loop expecation values in the continuum limit can be described in terms of heat kernels on the gauge group. It is therefore fundamental to investigate how the exact answer for 2D Yang-Mills compares with that of the continuum perturbative approach, which a priori are unrelated. In this paper, we provide a mathematically rigorous formulation of the perturbative quantization of 2D Yang-Mills, and we consider perturbative Wilson loop expectation values on $\\mathbb{R}^2$ and $S^2$ in Coulomb gauge, holomorphic gauge, and axial gauge (on $\\mathbb{R}^2$). We show the following equivalences and nonequivalences between these gauges: (i) Coulomb and holomorphic gauge are equivalent and are independent of the choice of gauge-fixing metric; (ii) both are inequivalent with axial-gauge. Additionally, we show that the asymptotics of exact lattice Wilson loop expectations on $S^2$ agree with perturbatively computed expectations in holomorphic gauge for simple closed curves to all orders. However, as a consequence of (ii), this result is necessarily false on $\\mathbb{R}^2$. Our work therefore presents fundamental progress in the analysis of how continuum perturbation theory succeeds or fails in capturing the asymptotics of the continuum limit of the lattice theory."
                    },
                    {
                        "title": "On discrete features of the wave equation in singular pp-wave backgrounds",
                        "abstract": "We analyze the wave equation in families of pp-wave geometries developing strong localized scale-invariant singularities in certain limits. For both cases of well-localized pp-waves and the so-called null-cosmologies, we observe an intriguing discrete dependence of the existence of a singular limit on the normalization of the pp-wave profile. We also find restrictive matching conditions relating the geometries before and after the singularity (if a singular limit for the solutions of the wave equation with initial conditions specified away from the near-singular region is assumed to exist)."
                    },
                    {
                        "title": "Dataset Distillation with Infinitely Wide Convolutional Networks",
                        "abstract": "The effectiveness of machine learning algorithms arises from being able to extract useful features from large amounts of data. As model and dataset sizes increase, dataset distillation methods that compress large datasets into significantly smaller yet highly performant ones will become valuable in terms of training efficiency and useful feature extraction. To that end, we apply a novel distributed kernel based meta-learning framework to achieve state-of-the-art results for dataset distillation using infinitely wide convolutional neural networks. For instance, using only 10 datapoints (0.02% of original dataset), we obtain over 65% test accuracy on CIFAR-10 image classification task, a dramatic improvement over the previous best test accuracy of 40%. Our state-of-the-art results extend across many other settings for MNIST, Fashion-MNIST, CIFAR-10, CIFAR-100, and SVHN. Furthermore, we perform some preliminary analyses of our distilled datasets to shed light on how they differ from naturally occurring data."
                    },
                    {
                        "title": "Dataset Meta-Learning from Kernel Ridge-Regression",
                        "abstract": "One of the most fundamental aspects of any machine learning algorithm is the training data used by the algorithm. We introduce the novel concept of $\\epsilon$-approximation of datasets, obtaining datasets which are much smaller than or are significant corruptions of the original training data while maintaining similar model performance. We introduce a meta-learning algorithm called Kernel Inducing Points (KIP) for obtaining such remarkable datasets, inspired by the recent developments in the correspondence between infinitely-wide neural networks and kernel ridge-regression (KRR). For KRR tasks, we demonstrate that KIP can compress datasets by one or two orders of magnitude, significantly improving previous dataset distillation and subset selection methods while obtaining state of the art results for MNIST and CIFAR-10 classification. Furthermore, our KIP-learned datasets are transferable to the training of finite-width neural networks even beyond the lazy-training regime, which leads to state of the art results for neural network dataset distillation with potential applications to privacy-preservation."
                    },
                    {
                        "title": "Is forgetting less a good inductive bias for forward transfer?",
                        "abstract": "One of the main motivations of studying continual learning is that the problem setting allows a model to accrue knowledge from past tasks to learn new tasks more efficiently. However, recent studies suggest that the key metric that continual learning algorithms optimize, reduction in catastrophic forgetting, does not correlate well with the forward transfer of knowledge. We believe that the conclusion previous works reached is due to the way they measure forward transfer. We argue that the measure of forward transfer to a task should not be affected by the restrictions placed on the continual learner in order to preserve knowledge of previous tasks. Instead, forward transfer should be measured by how easy it is to learn a new task given a set of representations produced by continual learning on previous tasks. Under this notion of forward transfer, we evaluate different continual learning algorithms on a variety of image classification benchmarks. Our results indicate that less forgetful representations lead to a better forward transfer suggesting a strong correlation between retaining past information and learning efficiency on new tasks. Further, we found less forgetful representations to be more diverse and discriminative compared to their forgetful counterparts."
                    }
                ]
            }
        }
    },
    "2405.15118": {
        "paper_data": {
            "title": "GS-Hider: Hiding Messages into 3D Gaussian Splatting",
            "url": "http://arxiv.org/abs/2405.15118v2",
            "arxiv_id": "2405.15118",
            "authors": [
                "Xuanyu Zhang",
                "Jiarui Meng",
                "Runyi Li",
                "Zhipei Xu",
                "Yongbing Zhang",
                "Jian Zhang"
            ],
            "abstract": "3D Gaussian Splatting (3DGS) has already become the emerging research focus in the fields of 3D scene reconstruction and novel view synthesis. Given that training a 3DGS requires a significant amount of time and computational cost, it is crucial to protect the copyright, integrity, and privacy of such 3D assets. Steganography, as a crucial technique for encrypted transmission and copyright protection, has been extensively studied. However, it still lacks profound exploration targeted at 3DGS. Unlike its predecessor NeRF, 3DGS possesses two distinct features: 1) explicit 3D representation; and 2) real-time rendering speeds. These characteristics result in the 3DGS point cloud files being public and transparent, with each Gaussian point having a clear physical significance. Therefore, ensuring the security and fidelity of the original 3D scene while embedding information into the 3DGS point cloud files is an extremely challenging task. To solve the above-mentioned issue, we first propose a steganography framework for 3DGS, dubbed GS-Hider, which can embed 3D scenes and images into original GS point clouds in an invisible manner and accurately extract the hidden messages. Specifically, we design a coupled secured feature attribute to replace the original 3DGS's spherical harmonics coefficients and then use a scene decoder and a message decoder to disentangle the original RGB scene and the hidden message. Extensive experiments demonstrated that the proposed GS-Hider can effectively conceal multimodal messages without compromising rendering quality and possesses exceptional security, robustness, capacity, and flexibility. Our project is available at: https://xuanyuzhang21.github.io/project/gshider.",
            "introduction": "   1 Introduction  As a frontier in computer vision and graphics, 3D scene reconstruction and novel view synthesis are crucial in fields such as movie production, game engines, virtual reality, and autonomous driving. Specifically, thanks to its high fidelity and fast rendering speeds, 3D Gaussian Splatting (3DGS)\u00a0[20] has become a mainstream approach for 3D rendering. Considering that rendering a 3DGS is extremely costly, protecting the copyright and privacy of 3D assets should be a priority. As a widely studied technique in copyright protection, digital watermarking, and encrypted communication, steganography aims to hide messages like audio, images, and bits into digital content in an invisible manner. In its reveal process, it is only possible for the receivers with pre-defined revealing operations to reconstruct secret information from the container. Therefore, a natural idea arises: Can we design a steganography method tailored for 3DGS to protect the copyright and privacy of 3D scenes?   Previous research on 3D steganography has already attracted significant attention. Classical 3D steganography methods\u00a0[37, 39, 51] often use Fourier and wavelet transforms to embed messages into explicit 3D representations such as meshes and point clouds. During extraction, the receiver must obtain the complete 3D representation. Considering that sometimes only a few views of a 3D scene are publicly available online, the 3D-to-2D watermarking mechanism\u00a0[56] is designed via a deep encoder-decoder framework, enabling to extract the copyright embedded in the mesh from any 2D perspective. Recently, steganography methods for implicit representations, such as NeRF\u00a0[34], have emerged. These methods modify the weights of the NeRF\u00a0[23] or replace the color representation of the NeRF\u00a0[32], ensuring that each rendered view contains hidden copyright information.   However, the methods mentioned above do not effectively apply to 3DGS due to its unique properties. First, since 3DGS is an explicit 3D representation where the attribute of each point has clear physical meanings, we cannot treat it as an implicit representation like NeRF, where the message can be directly and seamlessly embedded into the model weights via optimization\u00a0[23]. This approach could disrupt the fidelity of the rendered RGB views and compromise the integrity of the embedded message. Meanwhile, it is difficult for a single message decoder to faithfully memorize large-capacity hidden messages, such as an entire 3D scene. Second, since 3DGS is a real-time renderable 3D representation, users might upload the entire 3DGS point cloud file online and allow others to render it. Thus, the various attributes of the 3DGS are transparent and public, making it difficult to effectively conceal information by simply adding an attribute. Such approaches could lead to significant security concerns. Third, similar to image steganography\u00a0[60, 18], we aspire for 3DGS steganography to conceal versatile messages (such as 3D scenes, and images) and to explore the capacity limits. Hiding multiple messages allows various users to extract distinct information from the same 3D scene, thereby enhancing the adaptability and engagement of the transmission process. The application scenario of the proposed GS-Hider is presented in Fig.\u00a01.   Figure 1: Application scenario of the proposed GS-Hider. The 3DGS trainer (Alice) requires the training views of the original and hidden scenes to train our GS-Hider, comprising a",
            "references": [
                {
                    "title": "WateRF: Robust Watermarks in Radiance Fields for Protection of Copyrights",
                    "abstract": "The advances in the Neural Radiance Fields (NeRF) research offer extensive applications in diverse domains, but protecting their copyrights has not yet been researched in depth. Recently, NeRF watermarking has been considered one of the pivotal solutions for safely deploying NeRF-based 3D representations. However, existing methods are designed to apply only to implicit or explicit NeRF representations. In this work, we introduce an innovative watermarking method that can be employed in both representations of NeRF. This is achieved by fine-tuning NeRF to embed binary messages in the rendering process. In detail, we propose utilizing the discrete wavelet transform in the NeRF space for watermarking. Furthermore, we adopt a deferred back-propagation technique and introduce a combination with the patch-wise loss to improve rendering quality and bit accuracy with minimum trade-offs. We evaluate our method in three different aspects: capacity, invisibility, and robustness of the embedded watermarks in the 2D-rendered images. Our method achieves state-of-the-art performance with faster training speed over the compared state-of-the-art methods. Project page: https://kuai-lab.github.io/cvpr2024waterf/"
                },
                {
                    "title": "V2A-Mark: Versatile Deep Visual-Audio Watermarking for Manipulation Localization and Copyright Protection",
                    "abstract": "AI-generated video has revolutionized short video production, filmmaking, and personalized media, making video local editing an essential tool. However, this progress also blurs the line between reality and fiction, posing challenges in multimedia forensics. To solve this urgent issue, V2A-Mark is proposed to address the limitations of current video tampering forensics, such as poor generalizability, singular function, and single modality focus. Combining the fragility of video-into-video steganography with deep robust watermarking, our method can embed invisible visual-audio localization watermarks and copyright watermarks into the original video frames and audio, enabling precise manipulation localization and copyright protection. We also design a temporal alignment and fusion module and degradation prompt learning to enhance the localization accuracy and decoding robustness. Meanwhile, we introduce a sample-level audio localization method and a cross-modal copyright extraction mechanism to couple the information of audio and video frames. The effectiveness of V2A-Mark has been verified on a visual-audio tampering dataset, emphasizing its superiority in localization precision and copyright accuracy, crucial for the sustainable development of video editing in the AIGC video era."
                },
                {
                    "title": "Proactive Detection of Voice Cloning with Localized Watermarking",
                    "abstract": "In the rapidly evolving field of speech generative models, there is a pressing need to ensure audio authenticity against the risks of voice cloning. We present AudioSeal, the first audio watermarking technique designed specifically for localized detection of AI-generated speech. AudioSeal employs a generator/detector architecture trained jointly with a localization loss to enable localized watermark detection up to the sample level, and a novel perceptual loss inspired by auditory masking, that enables AudioSeal to achieve better imperceptibility. AudioSeal achieves state-of-the-art performance in terms of robustness to real life audio manipulations and imperceptibility based on automatic and human evaluation metrics. Additionally, AudioSeal is designed with a fast, single-pass detector, that significantly surpasses existing models in speed - achieving detection up to two orders of magnitude faster, making it ideal for large-scale and real-time applications."
                },
                {
                    "title": "IPR-NeRF: Ownership Verification meets Neural Radiance Field",
                    "abstract": "Neural Radiance Field (NeRF) models have gained significant attention in the computer vision community in the recent past with state-of-the-art visual quality and produced impressive demonstrations. Since then, technopreneurs have sought to leverage NeRF models into a profitable business. Therefore, NeRF models make it worth the risk of plagiarizers illegally copying, re-distributing, or misusing those models. This paper proposes a comprehensive intellectual property (IP) protection framework for the NeRF model in both black-box and white-box settings, namely IPR-NeRF. In the black-box setting, a diffusion-based solution is introduced to embed and extract the watermark via a two-stage optimization process. In the white-box setting, a designated digital signature is embedded into the weights of the NeRF model by adopting the sign loss objective. Our extensive experiments demonstrate that not only does our approach maintain the fidelity (\\ie, the rendering quality) of IPR-NeRF models, but it is also robust against both ambiguity and removal attacks compared to prior arts."
                },
                {
                    "title": "Noise-NeRF: Hide Information in Neural Radiance Fields using Trainable Noise",
                    "abstract": "Neural Radiance Field (NeRF) has been proposed as an innovative advancement in 3D reconstruction techniques. However, little research has been conducted on the issues of information confidentiality and security to NeRF, such as steganography. Existing NeRF steganography solutions have shortcomings in low steganography quality, model weight damage, and limited amount of steganographic information. This paper proposes Noise-NeRF, a novel NeRF steganography method employing Adaptive Pixel Selection strategy and Pixel Perturbation strategy to improve the quality and efficiency of steganography via trainable noise. Extensive experiments validate the state-of-the-art performances of Noise-NeRF on both steganography quality and rendering quality, as well as effectiveness in super-resolution image steganography."
                },
                {
                    "title": "Spacetime Gaussian Feature Splatting for Real-Time Dynamic View Synthesis",
                    "abstract": "Novel view synthesis of dynamic scenes has been an intriguing yet challenging problem. Despite recent advancements, simultaneously achieving high-resolution photorealistic results, real-time rendering, and compact storage remains a formidable task. To address these challenges, we propose Spacetime Gaussian Feature Splatting as a novel dynamic scene representation, composed of three pivotal components. First, we formulate expressive Spacetime Gaussians by enhancing 3D Gaussians with temporal opacity and parametric motion/rotation. This enables Space-time Gaussians to capture static, dynamic, as well as transient content within a scene. Second, we introduce splatted feature rendering, which replaces spherical harmonics with neural features. These features facilitate the modeling of view- and time-dependent appearance while maintaining small size. Third, we leverage the guidance of training error and coarse depth to sample new Gaussians in areas that are challenging to converge with existing pipelines. Experiments on several established real-world datasets demonstrate that our method achieves state-of-the-art rendering quality and speed, while retaining compact storage. At 8K resolution, our lite-version model can render at 60 FPS on an Nvidia RTX 4090 GPU."
                },
                {
                    "title": "EditGuard: Versatile Image Watermarking for Tamper Localization and Copyright Protection",
                    "abstract": "In the era of AI-generated content (AIGC), malicious tampering poses imminent threats to copyright integrity and information security. Current deep image watermarking, while widely accepted for safeguarding visual content, can only protect copyright and ensure traceability. They fall short in localizing increasingly realistic image tampering, potentially leading to trust crises, privacy violations, and legal disputes. To solve this challenge, we propose an innovative proactive forensics framework EditGuard, to unify copyright protection and tamper-agnostic localization, especially for AIGC-based editing methods. It can offer a meticulous embedding of imperceptible watermarks and precise decoding of tampered areas and copyright in-formation. Leveraging our observed fragility and locality of image-into-image steganography, the realization of Edit-Guard can be converted into a united image-bit steganography issue, thus completely decoupling the training process from the tampering types. Extensive experiments verify that our EditGuard balances the tamper localization accuracy, copyright recovery precision, and generalizability to various AIGC-based tampering methods, especially for image forgery that is difficult for the naked eye to detect."
                },
                {
                    "title": "SMERF: Streamable Memory Efficient Radiance Fields for Real-Time Large-Scene Exploration",
                    "abstract": "\n Recent techniques for real-time view synthesis have rapidly advanced in fidelity and speed, and modern methods are capable of rendering near-photorealistic scenes at interactive frame rates. At the same time, a tension has arisen between explicit scene representations amenable to rasterization and neural fields built on ray marching, with state-of-the-art instances of the latter surpassing the former in quality while being prohibitively expensive for real-time applications. We introduce SMERF, a view synthesis approach that achieves state-of-the-art accuracy among real-time methods on large scenes with footprints up to 300 m\n 2\n at a volumetric resolution of 3.5 mm\n 3\n . Our method is built upon two primary contributions: a hierarchical model partitioning scheme, which increases model capacity while constraining compute and memory consumption, and a distillation training strategy that simultaneously yields high fidelity and internal consistency. Our method enables full six degrees of freedom navigation in a web browser and renders in real-time on commodity smartphones and laptops. Extensive experiments show that our method exceeds the state-of-the-art in real-time novel view synthesis by 0.78 dB on standard benchmarks and 1.78 dB on large scenes, renders frames three orders of magnitude faster than state-of-the-art radiance field models, and achieves real-time performance across a wide variety of commodity devices, including smartphones. We encourage readers to explore these models interactively at our project website: https://smerf-3d.github.io.\n"
                },
                {
                    "title": "Gaussian Splatting SLAM",
                    "abstract": "We present the first application of 3D Gaussian Splatting in monocular SLAM, the most fundamental but the hardest setup for Visual SLAM. Our method, which runs live at 3fps, utilises Gaussians as the only 3D representation, unifying the required representation for accurate, efficient tracking, mapping, and high-quality rendering. Designed for challenging monocular settings, our approach is seamlessly extendable to RGB-D SLAM when an external depth sensor is available. Several innovations are required to continuously reconstruct 3D scenes with high fidelity from a live camera. First, to move beyond the original 3DGS algorithm, which requires accurate poses from an offline Structure from Motion (SfM) system, we formulate camera tracking for 3DGS using direct optimisation against the 3D Gaussians, and show that this enables fast and robust tracking with a wide basin of convergence. Second, by utilising the explicit nature of the Gaussians, we introduce geometric verification and regularisation to handle the ambiguities occurring in incremental 3D dense reconstruction. Finally, we introduce afull SLAM system which not only achieves state-of-the-art results in novel view synthesis and trajectory estimation but also reconstruction of tiny and even transparent objects."
                },
                {
                    "title": "Hiding Functions within Functions: Steganography by Implicit Neural Representations",
                    "abstract": "Deep steganography utilizes the powerful capabilities of deep neural networks to embed and extract messages, but its reliance on an additional message extractor limits its practical use due to the added suspicion it can raise from steganalyzers. To address this problem, we propose StegaINR, which utilizes Implicit Neural Representation (INR) to implement steganography. StegaINR embeds a secret function into a stego function, which serves as both the message extractor and the stego media for secure transmission on a public channel. Recipients need only use a shared key to recover the secret function from the stego function, allowing them to obtain the secret message. Our approach makes use of continuous functions, enabling it to handle various types of messages. To our knowledge, this is the first work to introduce INR into steganography. We performed evaluations on image and climate data to test our method in different deployment contexts."
                },
                {
                    "title": "Detecting Voice Cloning Attacks via Timbre Watermarking",
                    "abstract": "Nowadays, it is common to release audio content to the public. However, with the rise of voice cloning technology, attackers have the potential to easily impersonate a specific person by utilizing his publicly released audio without any permission. Therefore, it becomes significant to detect any potential misuse of the released audio content and protect its timbre from being impersonated. To this end, we introduce a novel concept,\"Timbre Watermarking\", which embeds watermark information into the target individual's speech, eventually defeating the voice cloning attacks. To ensure the watermark is robust to the voice cloning model's learning process, we design an end-to-end voice cloning-resistant detection framework. The core idea of our solution is to embed and extract the watermark in the frequency domain in a temporally invariant manner. To acquire generalization across different voice cloning attacks, we modulate their shared process and integrate it into our framework as a distortion layer. Experiments demonstrate that the proposed timbre watermarking can defend against different voice cloning attacks, exhibit strong resistance against various adaptive attacks (e.g., reconstruction-based removal attacks, watermark overwriting attacks), and achieve practicality in real-world services such as PaddleSpeech, Voice-Cloning-App, and so-vits-svc. In addition, ablation studies are also conducted to verify the effectiveness of our design. Some audio samples are available at https://timbrewatermarking.github.io/samples."
                },
                {
                    "title": "Feature 3DGS: Supercharging 3D Gaussian Splatting to Enable Distilled Feature Fields",
                    "abstract": "3D scene representations have gained immense popularity in recent years. Methods that use Neural Radiance fields are versatile for traditional tasks such as novel view synthesis. In recent times, some work has emerged that aims to extend the functionality of NeRF beyond view synthesis, for semantically aware tasks such as editing and segmentation using 3D feature field distillation from 2D foundation models. However, these methods have two major limitations: (a) they are limited by the rendering speed of NeRF pipelines, and (b) implicitly represented feature fields suffer from continuity artifacts reducing feature quality. Recently, 3D Gaussian Splatting has shown state-of-the-art performance on real-time radiance field rendering. In this work, we go one step further: in addition to radiance field rendering, we enable 3D Gaussian splatting on arbitrary-dimension semantic features via 2D foundation model distillation. This translation is not straightforward: naively incorporating feature fields in the 3DGS framework encounters significant challenges, notably the disparities in spatial resolution and channel consistency between RGB images and feature maps. We propose architectural and training changes to efficiently avert this problem. Our proposed method is general, and our experiments showcase novel view semantic segmentation, language-guided editing and segment anything through learning feature fields from state-of-the-art 2D foundation models such as SAM and CLIP-LSeg. Across experiments, our distillation method is able to provide comparable or better results, while being significantly faster to both train and render. Additionally, to the best of our knowledge, we are the first method to enable point and bounding-box prompting for radiance field manipulation, by leveraging the SAM model. Project website at: https://feature-3dgs.github.io/."
                },
                {
                    "title": "SplaTAM: Splat, Track & Map 3D Gaussians for Dense RGB-D SLAM",
                    "abstract": "Dense simultaneous localization and mapping (SLAM) is crucial for robotics and augmented reality applications. However, current methods are often hampered by the non-volumetric or implicit way they represent a scene. This work introduces SplaTAM, an approach that, for the first time, leverages explicit volumetric representations, i.e., 3D Gaussians, to enable high-fidelity reconstruction from a single unposed RGB-D camera, surpassing the capabilities of existing methods. SplaTAM employs a simple online tracking and mapping system tailored to the underlying Gaussian representation. It utilizes a silhouette mask to elegantly capture the presence of scene density. This combination enables several benefits over prior representations, including fast rendering and dense optimization, quickly determining if areas have been previously mapped, and structured map expansion by adding more Gaussians. Extensive experiments show that SplaTAM achieves up to 2 x superior performance in camera pose estimation, map construction, and novel-view synthesis over existing methods, paving the way for more immersive high-fidelity SLAM applications."
                },
                {
                    "title": "Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering",
                    "abstract": "Neural rendering methods have significantly advanced photo-realistic 3D scene rendering in various academic and industrial applications. The recent 3D Gaussian Splatting method has achieved the state-of-the-art rendering quality and speed combining the benefits of both primitive-based representations and volumetric representations. However, it often leads to heavily redundant Gaussians that try to fit every training view, neglecting the underlying scene ge-ometry. Consequently, the resulting model becomes less robust to significant view changes, texture-less area and lighting effects. We introduce Scaffold-GS, which uses an-chor points to distribute local 3D Gaussians, and predicts their attributes on-the-fly based on viewing direction and distance within the view frustum. Anchor growing and pruning strategies are developed based on the importance of neural Gaussians to reliably improve the scene cover-age. We show that our method effectively reduces redun-dant Gaussians while delivering high-quality rendering. We also demonstrates an enhanced capability to accommodate scenes with varying levels-of-detail and view-dependent ob-servations, without sacrificing the rendering speed. Project page: https://city-super.github.iolscaffold-gsl."
                },
                {
                    "title": "Mip-Splatting: Alias-Free 3D Gaussian Splatting",
                    "abstract": "Recently, 3D Gaussian Splatting has demonstrated impressive novel view synthesis results, reaching high fidelity and efficiency. However, strong artifacts can be observed when changing the sampling rate, e.g., by changing focal length or camera distance. We find that the source for this phenomenon can be attributed to the lack of 3D frequency constraints and the usage of a 2D dilation filter. To address this problem, we introduce a 3D smoothing filter to constrains the size of the 3D Gaussian primitives based on the maximal sampling frequency induced by the input views. It eliminates high-frequency artifacts when zooming in. Moreover, replacing 2D dilation with a 2D Mip filter, which simulates a 2D box filter, effectively mitigates aliasing and dilation issues. Our evaluation, including scenarios such a training on single-scale images and testing on multiple scales, validates the effectiveness of our approach."
                },
                {
                    "title": "GaussianEditor: Swift and Controllable 3D Editing with Gaussian Splatting",
                    "abstract": "3D editing plays a crucial role in many areas such as gaming and virtual reality. Traditional 3D editing methods, which rely on representations like meshes and point clouds, often fall short in realistically depicting complex scenes. On the other hand, methods based on implicit 3D representations, like Neural Radiance Field (NeRF), render complex scenes effectively but suffer from slow processing speeds and limited control over specific scene areas. In response to these challenges, our paper presents GaussianEditor, the first 3D editing algorithm based on Gaussian Splatting (GS), a novel 3D representation. GaussianEditor enhances precision and control in editing through our proposed Gaussian semantic tracing, which traces the editing target throughout the training process. Additionally, we propose Hierarchical Gaussian splatting (HGS) to achieve stabilized and fine results under stochastic generative guidance from 2D diffusion models. We also develop editing strategies for efficient object removal and integration, a challenging task for existing methods. Our comprehensive experiments demonstrate GaussianEditor's superior control, effective, and efficient performance, marking a significant advancement in 3D editing."
                },
                {
                    "title": "Towards Function Space Mesh Watermarking: Protecting the Copyright of Signed Distance Fields",
                    "abstract": "The signed distance field (SDF) represents 3D geometries in continuous function space. Due to its continuous nature, explicit 3D models (e.g., meshes) can be extracted from it at arbitrary resolution, which means losing the SDF is equivalent to losing the mesh. Recent research has shown meshes can also be extracted from SDF-enhanced neural radiance fields (NeRF). Such a signal raises an alarm that any implicit neural representation with SDF enhancement can extract the original mesh, which indicates identifying the SDF's intellectual property becomes an urgent issue. This paper proposes FuncMark, a robust and invisible watermarking method to protect the copyright of signed distance fields by leveraging analytic on-surface deformations to embed binary watermark messages. Such deformation can survive isosurfacing and thus be inherited by the extracted meshes for further watermark message decoding. Our method can recover the message with high-resolution meshes extracted from SDFs and detect the watermark even when mesh vertices are extremely sparse. Furthermore, our method is robust even when various distortions (including remeshing) are encountered. Extensive experiments demonstrate that our \\tool significantly outperforms state-of-the-art approaches and the message is still detectable even when only 50 vertex samples are given."
                },
                {
                    "title": "GaussianDreamer: Fast Generation from Text to 3D Gaussians by Bridging 2D and 3D Diffusion Models",
                    "abstract": "In recent times, the generation of 3D assets from text prompts has shown impressive results. Both 2D and 3D diffusion models can help generate decent 3D objects based on prompts. 3D diffusion models have good 3D consistency, but their quality and generalization are limited as trainable 3D data is expensive and hard to obtain. 2D diffusion models enjoy strong abilities of generalization and fine generation, but 3D consistency is hard to guarantee. This paper attempts to bridge the power from the two types of diffusion models via the recent explicit and efficient 3D Gaussian splatting representation. A fast 3D object gener-ation framework, named as GaussianDreamer, is proposed, where the 3D diffusion model provides priors for initial-ization and the 2D diffusion model enriches the geometry and appearance. Operations of noisy point growing and color perturbation are introduced to enhance the initialized Gaussians. Our GaussianDreamer can generate a high-quality 3D instance or 3D avatar within 15 minutes on one GPU, much faster than previous methods, while the generated instances can be directly rendered in real time. Demos and code are available at https://taoranyi.com/gaussiandreamer/."
                },
                {
                    "title": "4D Gaussian Splatting for Real-Time Dynamic Scene Rendering",
                    "abstract": "Representing and rendering dynamic scenes has been an important but challenging task. Especially, to accurately model complex motions, high efficiency is usually hard to guarantee. To achieve real-time dynamic scene rendering while also enjoying high training and storage efficiency, we propose 4D Gaussian Splatting (4D-GS) as a holistic representation for dynamic scenes rather than applying 3D-GS for each individual frame. In 4D-GS, a novel explicit representation containing both 3D Gaussians and 4D neural voxels is proposed. A decomposed neural voxel encoding algorithm inspired by HexPlane is proposed to efficiently build Gaussian features from 4D neural voxels and then a lightweight MLP is applied to predict Gaussian deformations at novel timestamps. Our 4D-GS method achieves real-time rendering under high resolutions, 82 FPS at an 800x800 resolution on an RTX 3090 GPU while maintaining comparable or better quality than previous state- of-the-art methods. More demos and code are available at https://guanjunwu.github.io/4dgs/."
                },
                {
                    "title": "DreamGaussian: Generative Gaussian Splatting for Efficient 3D Content Creation",
                    "abstract": "Recent advances in 3D content creation mostly leverage optimization-based 3D generation via score distillation sampling (SDS). Though promising results have been exhibited, these methods often suffer from slow per-sample optimization, limiting their practical usage. In this paper, we propose DreamGaussian, a novel 3D content generation framework that achieves both efficiency and quality simultaneously. Our key insight is to design a generative 3D Gaussian Splatting model with companioned mesh extraction and texture refinement in UV space. In contrast to the occupancy pruning used in Neural Radiance Fields, we demonstrate that the progressive densification of 3D Gaussians converges significantly faster for 3D generative tasks. To further enhance the texture quality and facilitate downstream applications, we introduce an efficient algorithm to convert 3D Gaussians into textured meshes and apply a fine-tuning stage to refine the details. Extensive experiments demonstrate the superior efficiency and competitive generation quality of our proposed approach. Notably, DreamGaussian produces high-quality textured meshes in just 2 minutes from a single-view image, achieving approximately 10 times acceleration compared to existing methods."
                },
                {
                    "title": "Deformable 3D Gaussians for High-Fidelity Monocular Dynamic Scene Reconstruction",
                    "abstract": "Implicit neural representation has paved the way for new approaches to dynamic scene reconstruction. Nonetheless, cutting-edge dynamic neural rendering methods rely heavily on these implicit representations, which frequently struggle to capture the intricate details of objects in the scene. Furthermore, implicit methods have difficulty achieving real-time rendering in general dynamic scenes, limiting their use in a variety of tasks. To address the issues, we propose a deformable 3D Gaussians splatting method that reconstructs scenes using 3D Gaussians and learns them in canonical space with a deformation field to model monocular dynamic scenes. We also introduce an annealing smoothing training mechanism with no extra overhead, which can mitigate the impact of inaccurate poses on the smoothness of time interpolation tasks in real-world scenes. Through a differential Gaussian rasterizer, the deformable 3D Gaussians not only achieve higher rendering quality but also real-time rendering speed. Experiments show that our method outperforms existing methods significantly in terms of both rendering quality and speed, making it well-suited for tasks such as novel-view synthesis, time interpolation, and real-time rendering. Our code is available at https://github.com/ingra14m/Deformable-3D-Gaussians."
                },
                {
                    "title": "MarkNerf: Watermarking for Neural Radiance Field",
                    "abstract": "A watermarking algorithm is proposed in this paper to address the copyright protection issue of implicit 3D models. The algorithm involves embedding watermarks into the images in the training set through an embedding network, and subsequently utilizing the NeRF model for 3D modeling. A copyright verifier is employed to generate a backdoor image by providing a secret perspective as input to the neural radiation field. Subsequently, a watermark extractor is devised using the hyperparameterization method of the neural network to extract the embedded watermark image from that perspective. In a black box scenario, if there is a suspicion that the 3D model has been used without authorization, the verifier can extract watermarks from a secret perspective to verify network copyright. Experimental results demonstrate that the proposed algorithm effectively safeguards the copyright of 3D models. Furthermore, the extracted watermarks exhibit favorable visual effects and demonstrate robust resistance against various types of noise attacks."
                },
                {
                    "title": "Dynamic 3D Gaussians: Tracking by Persistent Dynamic View Synthesis",
                    "abstract": "We present a method that simultaneously addresses the tasks of dynamic scene novel-view synthesis and six degree-of-freedom (6-DOF) tracking of all dense scene elements. We follow an analysis-by-synthesis framework, inspired by recent work that models scenes as a collection of 3D Gaussians which are optimized to reconstruct input images via differentiable rendering. To model dynamic scenes, we allow Gaussians to move and rotate over time while enforcing that they have persistent color, opacity, and size. By regularizing Gaussians\u2019 motion and rotation with local-rigidity constraints, we show that our Dynamic 3D Gaussians correctly model the same area of physical space over time, including the rotation of that space. Dense 6-DOF tracking and dynamic reconstruction emerges naturally from persistent dynamic view synthesis, without requiring any correspondence or flow as input. We demonstrate a large number of downstream applications enabled by our representation, including first-person view synthesis, dynamic compositional scene synthesis, and 4D video editing.111. Project Website: dynamic3dgaussians.github.io"
                },
                {
                    "title": "3D Gaussian Splatting for Real-Time Radiance Field Rendering",
                    "abstract": "Radiance Field methods have recently revolutionized novel-view synthesis of scenes captured with multiple photos or videos. However, achieving high visual quality still requires neural networks that are costly to train and render, while recent faster methods inevitably trade off speed for quality. For unbounded and complete scenes (rather than isolated objects) and 1080p resolution rendering, no current method can achieve real-time display rates. We introduce three key elements that allow us to achieve state-of-the-art visual quality while maintaining competitive training times and importantly allow high-quality real-time (\u2265 30 fps) novel-view synthesis at 1080p resolution. First, starting from sparse points produced during camera calibration, we represent the scene with 3D Gaussians that preserve desirable properties of continuous volumetric radiance fields for scene optimization while avoiding unnecessary computation in empty space; Second, we perform interleaved optimization/density control of the 3D Gaussians, notably optimizing anisotropic covariance to achieve an accurate representation of the scene; Third, we develop a fast visibility-aware rendering algorithm that supports anisotropic splatting and both accelerates training and allows realtime rendering. We demonstrate state-of-the-art visual quality and real-time rendering on several established datasets."
                },
                {
                    "title": "CopyRNeRF: Protecting the CopyRight of Neural Radiance Fields",
                    "abstract": "Neural Radiance Fields (NeRF) have the potential to be a major representation of media. Since training a NeRF has never been an easy task, the protection of its model copyright should be a priority. In this paper, by analyzing the pros and cons of possible copyright protection solutions, we propose to protect the copyright of NeRF models by replacing the original color representation in NeRF with a watermarked color representation. Then, a distortion-resistant rendering scheme is designed to guarantee robust message extraction in 2D renderings of NeRF. Our proposed method can directly protect the copyright of NeRF models while maintaining high rendering quality and bit accuracy when compared among optional solutions. Project page: https://luo-ziyuan.github.io/copyrnerf."
                },
                {
                    "title": "CRoSS: Diffusion Model Makes Controllable, Robust and Secure Image Steganography",
                    "abstract": "Current image steganography techniques are mainly focused on cover-based methods, which commonly have the risk of leaking secret images and poor robustness against degraded container images. Inspired by recent developments in diffusion models, we discovered that two properties of diffusion models, the ability to achieve translation between two images without training, and robustness to noisy data, can be used to improve security and natural robustness in image steganography tasks. For the choice of diffusion model, we selected Stable Diffusion, a type of conditional diffusion model, and fully utilized the latest tools from open-source communities, such as LoRAs and ControlNets, to improve the controllability and diversity of container images. In summary, we propose a novel image steganography framework, named Controllable, Robust and Secure Image Steganography (CRoSS), which has significant advantages in controllability, robustness, and security compared to cover-based image steganography methods. These benefits are obtained without additional training. To our knowledge, this is the first work to introduce diffusion models to the field of image steganography. In the experimental section, we conducted detailed experiments to demonstrate the advantages of our proposed CRoSS framework in controllability, robustness, and security."
                },
                {
                    "title": "Large-Capacity and Flexible Video Steganography via Invertible Neural Network",
                    "abstract": "Video steganography is the art of unobtrusively concealing secret data in a cover video and then recovering the secret data through a decoding protocol at the receiver end. Although several attempts have been made, most of them are limited to low-capacity and fixed steganography. To rectify these weaknesses, we propose a Large-capacity and Flexible Video Steganography Network (LF-VSN) in this paper. For large-capacity, we present a reversible pipeline to perform multiple videos hiding and recovering through a single invertible neural network (INN). Our method can hide/recover 7 secret videos in/from 1 cover video with promising performance. For flexibility, we propose a key-controllable scheme, enabling different receivers to recover particular secret videos from the same cover video through specific keys. Moreover, we further improve the flexibility by proposing a scalable strategy in multiple videos hiding, which can hide variable numbers of secret videos in a cover video with a single model and a single training session. Extensive experiments demonstrate that with the significant improvement of the video steganography performance, our proposed LF-VSN has high security, large hiding capacity, and flexibility. The source code is available at https://github.com/MC-E/LF-VSN."
                },
                {
                    "title": "Zip-NeRF: Anti-Aliased Grid-Based Neural Radiance Fields",
                    "abstract": "Neural Radiance Field training can be accelerated through the use of grid-based representations in NeRF\u2019s learned mapping from spatial coordinates to colors and volumetric density. However, these grid-based approaches lack an explicit understanding of scale and therefore often introduce aliasing, usually in the form of jaggies or missing scene content. Anti-aliasing has previously been addressed by mip-NeRF 360, which reasons about sub-volumes along a cone rather than points along a ray, but this approach is not natively compatible with current grid-based techniques. We show how ideas from rendering and signal processing can be used to construct a technique that combines mip-NeRF 360 and grid-based models such as Instant NGP to yield error rates that are 8% \u2013 77% lower than either prior technique, and that trains 24\u00d7 faster than mip-NeRF 360."
                },
                {
                    "title": "The Stable Signature: Rooting Watermarks in Latent Diffusion Models",
                    "abstract": "Generative image modeling enables a wide range of applications but raises ethical concerns about responsible deployment. We introduce an active content tracing method combining image watermarking and Latent Diffusion Models. The goal is for all generated images to conceal an invisible watermark allowing for future detection and/or identification. The method quickly fine-tunes the latent decoder of the image generator, conditioned on a binary signature. A pre-trained watermark extractor recovers the hidden signature from any generated image and a statistical test then determines whether it comes from the generative model. We evaluate the invisibility and robustness of the watermarks on a variety of generation tasks, showing that the Stable Signature is robust to image modifications. For instance, it detects the origin of an image generated from a text prompt, then cropped to keep 10% of the content, with 90+% accuracy at a false positive rate below 10\u22126."
                },
                {
                    "title": "MERF: Memory-Efficient Radiance Fields for Real-time View Synthesis in Unbounded Scenes",
                    "abstract": "Neural radiance fields enable state-of-the-art photorealistic view synthesis. However, existing radiance field representations are either too compute-intensive for real-time rendering or require too much memory to scale to large scenes. We present a Memory-Efficient Radiance Field (MERF) representation that achieves real-time rendering of large-scale scenes in a browser. MERF reduces the memory consumption of prior sparse volumetric radiance fields using a combination of a sparse feature grid and high-resolution 2D feature planes. To support large-scale unbounded scenes, we introduce a novel contraction function that maps scene coordinates into a bounded volume while still allowing for efficient ray-box intersection. We design a lossless procedure for baking the parameterization used during training into a model that achieves real-time rendering while still preserving the photorealistic view synthesis quality of a volumetric radiance field."
                },
                {
                    "title": "K-Planes: Explicit Radiance Fields in Space, Time, and Appearance",
                    "abstract": "We introduce k-planes, a white-box model for radiance fields in arbitrary dimensions. Our model uses planes to represent a d-dimensional scene, providing a seamless way to go from static (d = 3) to dynamic (d= 4) scenes. This planar factorization makes adding dimension-specific priors easy, e.g. temporal smoothness and multi-resolution spatial structure, and induces a natural decomposition of static and dynamic components of a scene. We use a linear feature decoder with a learned color basis that yields similar performance as a nonlinear black-box MLP decoder. Across a range of synthetic and real, static and dynamic, fixed and varying appearance scenes, k-planes yields competitive and often state-of-the-art recon- struction fidelity with low memory usage, achieving 1000x compression over a full 4D grid, and fast optimization with a pure PyTorch implementation. For video results and code, please see sarafridov.github.io/K-Planes."
                },
                {
                    "title": "HexPlane: A Fast Representation for Dynamic Scenes",
                    "abstract": "Modeling and re-rendering dynamic 3D scenes is a challenging task in 3D vision. Prior approaches build on NeRF and rely on implicit representations. This is slow since it requires many MLP evaluations, constraining real-world applications. We show that dynamic 3D scenes can be explicitly represented by six planes of learned features, leading to an elegant solution we call HexPlane. A HexPlane computes features for points in spacetime by fusing vectors extracted from each plane, which is highly efficient. Pairing a HexPlane with a tiny MLP to regress output colors and training via volume rendering gives impressive results for novel view synthesis on dynamic scenes, matching the image quality of prior work but reducing training time by more than 100\u00d7. Extensive ablations confirm our HexPlane design and show that it is robust to different feature fusion mechanisms, coordinate systems, and decoding mechanisms. HexPlane is a simple and effective solution for representing 4D volumes, and we hope they can broadly contribute to modeling spacetime for dynamic 3D scenes.11Project page: https://caoang327.github.io/HexPlane."
                },
                {
                    "title": "DeAR: A Deep-learning-based Audio Re-recording Resilient Watermarking",
                    "abstract": "Audio watermarking is widely used for leaking source tracing. The robustness of the watermark determines the traceability of the algorithm. With the development of digital technology, audio re-recording (AR) has become an efficient and covert means to steal secrets. AR process could drastically destroy the watermark signal while preserving the original information. This puts forward a new requirement for audio watermarking at this stage, that is, to be robust to AR distortions. Unfortunately, none of the existing algorithms can effectively resist AR attacks due to the complexity of the AR process. To address this limitation, this paper proposes DeAR, a deep-learning-based audio re-recording resistant watermarking. Inspired by DNN-based image watermarking, we pioneer a deep learning framework for audio carriers, based on which the watermark signal can be effectively embedded and extracted. Meanwhile, in order to resist the AR attack, we delicately analyze the distortions that occurred in the AR process and design the corresponding distortion layer to cooperate with the proposed watermarking framework. Extensive experiments show that the proposed algorithm can resist not only common electronic channel distortions but also AR distortions. Under the premise of high-quality embedding (SNR=25.86dB), in the case of a common re-recording distance (20cm), the algorithm can effectively achieve an average bit recovery accuracy of 98.55%."
                },
                {
                    "title": "StegaNeRF: Embedding Invisible Information within Neural Radiance Fields",
                    "abstract": "Recent advancements in neural rendering have paved the way for a future marked by the widespread distribution of visual data through the sharing of Neural Radiance Field (NeRF) model weights. However, while established techniques exist for embedding ownership or copyright information within conventional visual data such as images and videos, the challenges posed by the emerging NeRF format have remained unaddressed. In this paper, we introduce StegaNeRF, an innovative approach for steganographic information embedding within NeRF renderings. We have meticulously developed an optimization framework that enables precise retrieval of hidden information from images generated by NeRF, while ensuring the original visual quality of the rendered images to remain intact. Through rigorous experimentation, we assess the efficacy of our methodology across various potential deployment scenarios. Furthermore, we delve into the insights gleaned from our analysis. StegaNeRF represents an initial foray into the intriguing realm of infusing NeRF renderings with customizable, imperceptible, and recoverable information, all while minimizing any discernible impact on the rendered images. For more details, please visit our project page: https://xggnet.github.io/StegaNeRF/"
                },
                {
                    "title": "Neural Point Catacaustics for Novel-View Synthesis of Reflections",
                    "abstract": "View-dependent effects such as reflections pose a substantial challenge for image-based and neural rendering algorithms. Above all, curved reflectors are particularly hard, as they lead to highly non-linear reflection flows as the camera moves. We introduce a new point-based representation to compute Neural Point Catacaustics allowing novel-view synthesis of scenes with curved reflectors, from a set of casually-captured input photos. At the core of our method is a neural warp field that models catacaustic trajectories of reflections, so complex specular effects can be rendered using efficient point splatting in conjunction with a neural renderer. One of our key contributions is the explicit representation of reflections with a reflection point cloud which is displaced by the neural warp field, and a primary point cloud which is optimized to represent the rest of the scene. After a short manual annotation step, our approach allows interactive high-quality renderings of novel views with accurate reflection flow. Additionally, the explicit representation of reflection flow supports several forms of scene manipulation in captured scenes, such as reflection editing, cloning of specular objects, reflection tracking across views, and comfortable stereo viewing. We provide the source code and other supplemental material on https://repo-sam.inria.fr/fungraph/neural_catacaustics/"
                },
                {
                    "title": "Robust Invertible Image Steganography",
                    "abstract": "Image steganography aims to hide secret images into a container image, where the secret is hidden from human vision and can be restored when necessary. Previous image steganography methods are limited in hiding capacity and robustness, commonly vulnerable to distortion on container images such as Gaussian noise, Poisson noise, and lossy compression. This paper presents a novelflow-basedframe-work for robust invertible image steganography, dubbed as RIIS. A conditional normalizing flow is introduced to model the distribution of the redundant high-frequency component with the condition of the container image. Moreover, a well-designed container enhancement module (CEM) also contributes to the robust reconstruction. To regulate the net-work parameters for different distortion levels, a distortion-guided modulation (DGM) is implemented over flow-based blocks to make it a one-size-fits-all model. In terms of both clean and distorted image steganography, extensive experi-ments reveal that the proposed RIIS efficiently improves the robustness while maintaining imperceptibility and capacity. As far as we know, we are the first to propose a learning-based scheme to enhance the robustness of image steganog-raphy in the literature. The guarantee of steganography ro-bustness significantly broadens the application of steganog-raphy in real-world applications."
                },
                {
                    "title": "DeVRF: Fast Deformable Voxel Radiance Fields for Dynamic Scenes",
                    "abstract": "Modeling dynamic scenes is important for many applications such as virtual reality and telepresence. Despite achieving unprecedented fidelity for novel view synthesis in dynamic scenes, existing methods based on Neural Radiance Fields (NeRF) suffer from slow convergence (i.e., model training time measured in days). In this paper, we present DeVRF, a novel representation to accelerate learning dynamic radiance fields. The core of DeVRF is to model both the 3D canonical space and 4D deformation field of a dynamic, non-rigid scene with explicit and discrete voxel-based representations. However, it is quite challenging to train such a representation which has a large number of model parameters, often resulting in overfitting issues. To overcome this challenge, we devise a novel static-to-dynamic learning paradigm together with a new data capture setup that is convenient to deploy in practice. This paradigm unlocks efficient learning of deformable radiance fields via utilizing the 3D volumetric canonical space learnt from multi-view static images to ease the learning of 4D voxel deformation field with only few-view dynamic sequences. To further improve the efficiency of our DeVRF and its synthesized novel view's quality, we conduct thorough explorations and identify a set of strategies. We evaluate DeVRF on both synthetic and real-world dynamic scenes with different types of deformation. Experiments demonstrate that DeVRF achieves two orders of magnitude speedup (100x faster) with on-par high-fidelity results compared to the previous state-of-the-art approaches. The code and dataset will be released in https://github.com/showlab/DeVRF."
                },
                {
                    "title": "Fast Dynamic Radiance Fields with Time-Aware Neural Voxels",
                    "abstract": "Neural radiance fields (NeRF) have shown great success in modeling 3D scenes and synthesizing novel-view images. However, most previous NeRF methods take much time to optimize one single scene. Explicit data structures, e.g. voxel features, show great potential to accelerate the training process. However, voxel features face two big challenges to be applied to dynamic scenes, i.e. modeling temporal information and capturing different scales of point motions. We propose a radiance field framework by representing scenes with time-aware voxel features, named as TiNeuVox. A tiny coordinate deformation network is introduced to model coarse motion trajectories and temporal information is further enhanced in the radiance network. A multi-distance interpolation method is proposed and applied on voxel features to model both small and large motions. Our framework significantly accelerates the optimization of dynamic radiance fields while maintaining high rendering quality. Empirical evaluation is performed on both synthetic and real scenes. Our TiNeuVox completes training with only 8 minutes and 8-MB storage cost while showing similar or even better rendering performance than previous dynamic NeRF methods. Code is available at https://github.com/hustvl/TiNeuVox."
                },
                {
                    "title": "SqueezeNeRF: Further factorized FastNeRF for memory-efficient inference",
                    "abstract": "Neural Radiance Fields (NeRF) has emerged as the state-of-the-art method for novel view generation of complex scenes, but is very slow during inference. Recently, there have been multiple works on speeding up NeRF inference, but the state of the art methods for real-time NeRF inference rely on caching the neural network output, which occupies several gigabytes of disk space that limits their real-world applicability. As caching the neural network of original NeRF network is not feasible, Garbin et al. proposed \"FastNeRF\" which factorizes the problem into 2 sub-networks - one which depends only on the 3D coordinate of a sample point and one which depends only on the 2D camera viewing direction. Although this factorization enables them to reduce the cache size and perform inference at over 200 frames per second, the memory overhead is still substantial. In this work, we propose SqueezeNeRF, which is more than 60 times memory-efficient than the sparse cache of FastNeRF and is still able to render at more than 190 frames per second on a high spec GPU during inference."
                },
                {
                    "title": "Instant neural graphics primitives with a multiresolution hash encoding",
                    "abstract": "Neural graphics primitives, parameterized by fully connected neural networks, can be costly to train and evaluate. We reduce this cost with a versatile new input encoding that permits the use of a smaller network without sacrificing quality, thus significantly reducing the number of floating point and memory access operations: a small neural network is augmented by a multiresolution hash table of trainable feature vectors whose values are optimized through stochastic gradient descent. The multiresolution structure allows the network to disambiguate hash collisions, making for a simple architecture that is trivial to parallelize on modern GPUs. We leverage this parallelism by implementing the whole system using fully-fused CUDA kernels with a focus on minimizing wasted bandwidth and compute operations. We achieve a combined speedup of several orders of magnitude, enabling training of high-quality neural graphics primitives in a matter of seconds, and rendering in tens of milliseconds at a resolution of 1920\u00d71080."
                },
                {
                    "title": "Plenoxels: Radiance Fields without Neural Networks",
                    "abstract": "We introduce Plenoxels (plenoptic voxels), a systemfor photorealistic view synthesis. Plenoxels represent a scene as a sparse 3D grid with spherical harmonics. This representation can be optimized from calibrated images via gradient methods and regularization without any neural components. On standard, benchmark tasks, Plenoxels are optimized two orders of magnitude faster than Neural Radiance Fields with no loss in visual quality. For video and code, please see https://alexyu.net/plenoxels."
                },
                {
                    "title": "NeRF-SR: High Quality Neural Radiance Fields using Supersampling",
                    "abstract": "We present NeRF-SR, a solution for high-resolution (HR) novel view synthesis with mostly low-resolution (LR) inputs. Our method is built upon Neural Radiance Fields (NeRF) that predicts per-point density and color with a multi-layer perceptron. While producing images at arbitrary scales, NeRF struggles with resolutions that go beyond observed images. Our key insight is that NeRF benefits from 3D consistency, which means an observed pixel absorbs information from nearby views. We first exploit it by a super-sampling strategy that shoots multiple rays at each image pixel, which further enforces multi-view constraint at a sub-pixel level. Then, we show that NeRF-SR can further boost the performance of super-sampling by a refinement network that leverages the estimated depth at hand to hallucinate details from related patches on only one HR reference image. Experiment results demonstrate that NeRF-SR generates high-quality results for novel view synthesis at HR on both synthetic and real-world datasets without any external information. Project page: https://cwchenwang.github.io/NeRF-SR"
                },
                {
                    "title": "Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields",
                    "abstract": "Though neural radiance fields (NeRF) have demon-strated impressive view synthesis results on objects and small bounded regions of space, they struggle on \u201cun-bounded\u201d scenes, where the camera may point in any di-rection and content may exist at any distance. In this set-ting, existing NeRF-like models often produce blurry or low-resolution renderings (due to the unbalanced detail and scale of nearby and distant objects), are slow to train, and may exhibit artifacts due to the inherent ambiguity of the task of reconstructing a large scene from a small set of images. We present an extension of mip-NeRF (a NeRF variant that addresses sampling and aliasing) that uses a non-linear scene parameterization, online distillation, and a novel distortion-based regularizer to overcome the chal-lenges presented by unbounded scenes. Our model, which we dub \u201cmip-NeRF 360\u201d as we target scenes in which the camera rotates 360 degrees around a point, reduces mean-squared error by 57% compared to mip-NeRF, and is able to produce realistic synthesized views and detailed depth maps for highly intricate, unbounded real-world scenes."
                },
                {
                    "title": "Direct Voxel Grid Optimization: Super-fast Convergence for Radiance Fields Reconstruction",
                    "abstract": "We present a super-fast convergence approach to reconstructing the per-scene radiance field from a set of images that capture the scene with known poses. This task, which is often applied to novel view synthesis, is recently revolution-ized by Neural Radiance Field (NeRF) for its state-of-the-art quality and fiexibility. However, NeRF and its variants require a lengthy training time ranging from hours to days for a single scene. In contrast, our approach achieves NeRF-comparable quality and converges rapidly from scratch in less than 15 minutes with a single GPU. We adopt a representation consisting of a density voxel grid for scene geometry and a feature voxel grid with a shallow network for complex view-dependent appearance. Modeling with explicit and discretized volume representations is not new, but we propose two simple yet non-trivial techniques that contribute to fast convergence speed and high-quality output. First, we introduce the post-activation interpolation on voxel density, which is capable of producing sharp surfaces in lower grid resolution. Second, direct voxel density optimization is prone to suboptimal geometry solutions, so we robustify the optimization process by imposing several priors. Finally, evaluation on five inward-facing benchmarks shows that our method matches, if not surpasses, NeRF's quality, yet it only takes about 15 minutes to train from scratch for a new scene. Code: https://github.com/sunset1995/DirectVoxGO."
                },
                {
                    "title": "HiNet: Deep Image Hiding by Invertible Network",
                    "abstract": "Image hiding aims to hide a secret image into a cover image in an imperceptible way, and then recover the secret image perfectly at the receiver end. Capacity, invisibility and security are three primary challenges in image hiding task. This paper proposes a novel invertible neural network (INN) based framework, HiNet, to simultaneously overcome the three challenges in image hiding. For large capacity, we propose an inverse learning mechanism by simultaneously learning the image concealing and revealing processes. Our method is able to achieve the concealing of a full-size secret image into a cover image with the same size. For high invisibility, instead of pixel domain hiding, we propose to hide the secret information in wavelet domain. Furthermore, we propose a new low-frequency wavelet loss to constrain that secret information is hidden in high-frequency wavelet subbands, which significantly improves the hiding security. Experimental results show that our HiNet significantly outperforms other state-of-the-art image hiding methods, with more than 10 dB PSNR improvement in secret image recovery on ImageNet, COCO and DIV2K datasets. Codes are available at https://github.com/TomTomTommi/HiNet."
                },
                {
                    "title": "Point\u2010Based Neural Rendering with Per\u2010View Optimization",
                    "abstract": "There has recently been great interest in neural rendering methods. Some approaches use 3D geometry reconstructed with Multi\u2010View Stereo (MVS) but cannot recover from the errors of this process, while others directly learn a volumetric neural representation, but suffer from expensive training and inference. We introduce a general approach that is initialized with MVS, but allows further optimization of scene properties in the space of input views, including depth and reprojected features, resulting in improved novel\u2010view synthesis. A key element of our approach is our new differentiable point\u2010based pipeline, based on bi\u2010directional Elliptical Weighted Average splatting, a probabilistic depth test and effective camera selection. We use these elements together in our neural renderer, that outperforms all previous methods both in quality and speed in almost all scenes we tested. Our pipeline can be applied to multi\u2010view harmonization and stylization in addition to novel\u2010view synthesis."
                },
                {
                    "title": "Deep 3D-to-2D Watermarking: Embedding Messages in 3D Meshes and Extracting Them from 2D Renderings",
                    "abstract": "Digital watermarking is widely used for copyright protection. Traditional 3D watermarking approaches or commercial software are typically designed to embed messages into 3D meshes, and later retrieve the messages directly from distorted/undistorted watermarked 3D meshes. However, in many cases, users only have access to rendered 2D images instead of 3D meshes. Unfortunately, retrieving messages from 2D renderings of 3D meshes is still challenging and underexplored. We introduce a novel end-to-end learning framework to solve this problem through: 1) an encoder to covertly embed messages in both mesh geometry and textures; 2) a differentiable renderer to render watermarked 3D objects from different camera angles and under varied lighting conditions; 3) a decoder to recover the messages from 2D rendered images. From our experiments, we show that our model can learn to embed information visually imperceptible to humans, and to retrieve the embedded information from 2D renderings that undergo 3D distortions. In addition, we demonstrate that our method can also work with other renderers, such as ray tracers and real-time renderers with and without fine-tuning."
                },
                {
                    "title": "DVMark: A Deep Multiscale Framework for Video Watermarking",
                    "abstract": "Video watermarking embeds a message into a cover video in an imperceptible manner, which can be retrieved even if the video undergoes certain modifications or distortions. Traditional watermarking methods are often manually designed for particular types of distortions and thus cannot simultaneously handle a broad spectrum of distortions. To this end, we propose a robust deep learning-based solution for video watermarking that is end-to-end trainable. Our model consists of a novel multiscale design where the watermarks are distributed across multiple spatial-temporal scales. Extensive evaluations on a wide variety of distortions show that our method outperforms traditional video watermarking methods as well as deep image watermarking models by a large margin. We further demonstrate the practicality of our method on a realistic video-editing application."
                },
                {
                    "title": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields",
                    "abstract": "The rendering procedure used by neural radiance fields (NeRF) samples a scene with a single ray per pixel and may therefore produce renderings that are excessively blurred or aliased when training or testing images observe scene content at different resolutions. The straightforward solution of supersampling by rendering with multiple rays per pixel is impractical for NeRF, because rendering each ray requires querying a multilayer perceptron hundreds of times. Our solution, which we call \"mip-NeRF\" (\u00e0 la \"mipmap\"), extends NeRF to represent the scene at a continuously-valued scale. By efficiently rendering anti-aliased conical frustums instead of rays, mip-NeRF reduces objectionable aliasing artifacts and significantly improves NeRF\u2019s ability to represent fine details, while also being 7% faster than NeRF and half the size. Compared to NeRF, mip-NeRF reduces average error rates by 17% on the dataset presented with NeRF and by 60% on a challenging multiscale variant of that dataset that we present. Mip-NeRF is also able to match the accuracy of a brute-force supersampled NeRF on our multiscale dataset while being 22\u00d7 faster."
                },
                {
                    "title": "Deep Model Intellectual Property Protection via Deep Watermarking",
                    "abstract": "Despite the tremendous success, deep neural networks are exposed to serious IP infringement risks. Given a target deep model, if the attacker knows its full information, it can be easily stolen by fine-tuning. Even if only its output is accessible, a surrogate model can be trained through student-teacher learning by generating many input-output training pairs. Therefore, deep model IP protection is important and necessary. However, it is still seriously under-researched. In this work, we propose a new model watermarking framework for protecting deep networks trained for low-level computer vision or image processing tasks. Specifically, a special task-agnostic barrier is added after the target model, which embeds a unified and invisible watermark into its outputs. When the attacker trains one surrogate model by using the input-output pairs of the barrier target model, the hidden watermark will be learned and extracted afterwards. To enable watermarks from binary bits to high-resolution images, a deep invisible watermarking mechanism is designed. By jointly training the target model and watermark embedding, the extra barrier can even be absorbed into the target model. Through extensive experiments, we demonstrate the robustness of the proposed framework, which can resist attacks with different network structures and objective functions."
                },
                {
                    "title": "D-NeRF: Neural Radiance Fields for Dynamic Scenes",
                    "abstract": "Neural rendering techniques combining machine learning with geometric reasoning have arisen as one of the most promising approaches for synthesizing novel views of a scene from a sparse set of images. Among these, stands out the Neural radiance fields (NeRF) [31], which trains a deep network to map 5D input coordinates (representing spatial location and viewing direction) into a volume density and view-dependent emitted radiance. However, despite achieving an unprecedented level of photorealism on the generated images, NeRF is only applicable to static scenes, where the same spatial location can be queried from different images. In this paper we introduce D-NeRF, a method that extends neural radiance fields to a dynamic domain, allowing to reconstruct and render novel images of objects under rigid and non-rigid motions from a single camera moving around the scene. For this purpose we consider time as an additional input to the system, and split the learning process in two main stages: one that encodes the scene into a canonical space and another that maps this canonical representation into the deformed scene at a particular time. Both mappings are simultaneously learned using fully-connected networks. Once the networks are trained, D-NeRF can render novel images, controlling both the camera view and the time variable, and thus, the object movement. We demonstrate the effectiveness of our approach on scenes with objects under rigid, articulated and non-rigid motions. Code, model weights and the dynamic scenes dataset will be available at [1]."
                },
                {
                    "title": "Hiding Images within Images",
                    "abstract": "We present a system to hide a full color image inside another of the same size with minimal quality loss to either image. Deep neural networks are simultaneously trained to create the hiding and revealing processes and are designed to specifically work as a pair. The system is trained on images drawn randomly from the ImageNet database, and works well on natural images from a wide variety of sources. Beyond demonstrating the successful application of deep learning to hiding images, we examine how the result is achieved and apply numerous transformations to analyze if image quality in the host and hidden image can be maintained. These transformation range from simple image manipulations to sophisticated machine learning-based adversaries. Two extensions to the basic system are presented that mitigate the possibility of discovering the content of the hidden image. With these extensions, not only can the hidden information be kept secure, but the system can be used to hide even more than a single image. Applications for this technology include image authentication, digital watermarks, finding exact regions of image manipulation, and storing meta-information about image rendering and content."
                },
                {
                    "title": "Differentiable surface splatting for point-based geometry processing",
                    "abstract": "We propose Differentiable Surface Splatting (DSS), a high-fidelity differentiable renderer for point clouds. Gradients for point locations and normals are carefully designed to handle discontinuities of the rendering function. Regularization terms are introduced to ensure uniform distribution of the points on the underlying surface. We demonstrate applications of DSS to inverse rendering for geometry synthesis and denoising, where large scale topological changes, as well as small scale detail modifications, are accurately and robustly handled without requiring explicit connectivity, outperforming state-of-the-art techniques. The data and code are at https://github.com/yifita/DSS."
                },
                {
                    "title": "Structure-from-Motion Revisited",
                    "abstract": "Incremental Structure-from-Motion is a prevalent strategy for 3D reconstruction from unordered image collections. While incremental reconstruction systems have tremendously advanced in all regards, robustness, accuracy, completeness, and scalability remain the key problems towards building a truly general-purpose pipeline. We propose a new SfM technique that improves upon the state of the art to make a further step towards this ultimate goal. The full reconstruction pipeline is released to the public as an open-source implementation."
                },
                {
                    "title": "StegExpose - A Tool for Detecting LSB Steganography",
                    "abstract": "Steganalysis tools play an important part in saving time and providing new angles of attack for forensic analysts. StegExpose is a solution designed for use in the real world, and is able to analyse images for LSB steganography in bulk using proven attacks in a time efficient manner. When steganalytic methods are combined intelligently, they are able generate even more accurate results. This is the prime focus of StegExpose."
                },
                {
                    "title": "3D ShapeNets: A deep representation for volumetric shapes",
                    "abstract": "3D shape is a crucial but heavily underutilized cue in today's computer vision systems, mostly due to the lack of a good generic shape representation. With the recent availability of inexpensive 2.5D depth sensors (e.g. Microsoft Kinect), it is becoming increasingly important to have a powerful 3D shape representation in the loop. Apart from category recognition, recovering full 3D shapes from view-based 2.5D depth maps is also a critical part of visual understanding. To this end, we propose to represent a geometric 3D shape as a probability distribution of binary variables on a 3D voxel grid, using a Convolutional Deep Belief Network. Our model, 3D ShapeNets, learns the distribution of complex 3D shapes across different object categories and arbitrary poses from raw CAD data, and discovers hierarchical compositional part representation automatically. It naturally supports joint object recognition and shape completion from 2.5D depth maps, and it enables active object recognition through view planning. To train our 3D deep learning model, we construct ModelNet - a large-scale 3D CAD model dataset. Extensive experiments show that our 3D deep representation enables significant performance improvement over the-state-of-the-arts in a variety of tasks."
                },
                {
                    "title": "Hide and Seek: An Introduction to Steganography",
                    "abstract": "Although people have hidden secrets in plain sight-now called steganography-throughout the ages, the recent growth in computational power and technology has propelled it to the forefront of today's security techniques. Essentially, the information-hiding process in a steganographic system starts by identifying a cover medium's redundant bits (those that can be modified without destroying that medium's integrity). The embedding process creates a stego medium by replacing these redundant bits with data from the hidden message. This article discusses existing steganographic systems and presents recent research in detecting them via statistical steganalysis. Here, we present recent research and discuss the practical application of detection algorithms and the mechanisms for getting around them."
                },
                {
                    "title": "A Frequency\u2010Domain Approach to Watermarking 3D Shapes",
                    "abstract": "This paper presents a robust watermarking algorithm with informed detection for 3D polygonal meshes. The algorithm is based on our previous algorithm [ 22] that employs mesh\u2010spectral analysis to modify mesh shapes in their transformed domain. This paper presents extensions to our previous algorithm so that (1) much larger meshes can be watermarked within a reasonable time, and that (2) the watermark is robust against connectivity alteration (e.g., mesh simplification), and that (3) the watermark is robust against attacks that combine similarity transformation with such other attacks as cropping, mesh simplification, and smoothing. Experiment showed that our new watermarks are resistant against mesh simplification and remeshing combined with resection, similarity transformation, and other operations.."
                },
                {
                    "title": "Robust mesh watermarking",
                    "abstract": "We describe a robust method for watermarking triangle meshes. Watermarking provides a mechanism for copyright protection of digital media by embedding information identifying the owner in the data. The bulk of the research on digital watermarks has focused on media such as images, video, audio, and text. Robust watermarks must be able to survive a variety of \u201cattacks\u201d, including resizing, cropping, and filtering. For resilience to such attacks, recent watermarking schemes employ a \u201cspread-spectrum\u201d approach \u2013 they transform the document to the frequency domain and perturb the coefficients of the perceptually most significant basis functions. We extend this spread-spectrum approach to work for the robust watermarking of arbitrary triangle meshes. Generalizing spread spectrum techniques to surfaces presents two major challenges. First, arbitrary surfaces lack a natural parametrization for frequency-based decomposition. Our solution is to construct a set of scalar basis function over the mesh vertices using multiresolution analysis. The watermark perturbs vertices along the direction of the surface normal, weighted by the basis functions. The second challenge is that simplification and other attacks may modify the connectivity of the mesh. We use an optimization technique to resample an attacked mesh using the original mesh connectivity. Results show that our watermarks are resistant to common mesh operations such as translation, rotation, scaling, cropping, smoothing, simplification, and resampling, as well as malicious attacks such as the insertion of noise, modification of low-order bits, or even insertion of other watermarks. CR Categories: I.3.5 [Computer Graphics]: Computational Geometry and Object Modeling\u2014Surface Representations."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nCan we design a steganography method tailored for 3D Gaussian Splatting (3DGS) to protect the copyright and privacy of 3D scenes?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses the growing need for copyright protection and privacy in 3D assets, particularly in fields like virtual reality, gaming, and autonomous driving. A successful steganography method for 3DGS could lead to advancements in secure digital content sharing, enabling creators to protect their intellectual property while still allowing for the dissemination of 3D scenes. This could foster innovation in 3D content creation and distribution, ultimately influencing future research in digital rights management and secure communications.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the unique properties of 3DGS as an explicit 3D representation. Naive approaches that treat 3DGS like implicit representations (e.g., NeRF) may fail because they could disrupt the fidelity of rendered views and compromise the integrity of the hidden message. Additionally, the need to conceal information in a way that does not expose it to public scrutiny is complicated by the transparent nature of 3DGS attributes. Furthermore, effectively encoding large-capacity messages while maintaining the quality of the 3D scene presents significant technical and practical obstacles.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research on 3D steganography has primarily focused on explicit representations like meshes and point clouds or on implicit representations like NeRF, which do not translate well to 3DGS. The limitations of existing methods include their inability to handle the real-time rendering requirements of 3DGS and the challenges of embedding messages without compromising the visual fidelity of the 3D scene. Barriers such as the lack of tailored methodologies for 3DGS and the complexity of hiding multiple messages simultaneously have prevented effective solutions. Our approach aims to address these gaps by developing a dedicated steganography method that respects the unique characteristics of 3DGS.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology, GS-Hider, involves training a model using views of both the original and hidden 3D scenes to embed messages into the 3DGS representation. We will utilize a dataset of 3D scenes and employ metrics such as visual fidelity and message"
            }
        },
        "author_data": {
            "38be0fa6-41f2-459b-b65b-80f8e4cb9af7": {
                "pk": "38be0fa6-41f2-459b-b65b-80f8e4cb9af7",
                "name": "Xuanyu Zhang",
                "collaborators": [
                    "Qing Yang",
                    "Dongliang Xu",
                    "Jian Zhang",
                    "Bingbing Li",
                    "Jiwen Yu",
                    "Youmin Xu",
                    "Yongbing Zhang",
                    "Ruiqin Xiong",
                    "Qilin Sun"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Knowledge Graph",
                    "Image Processing",
                    "Hyperspectral Imaging"
                ],
                "publications": [
                    {
                        "title": "Self-QA: Unsupervised Knowledge Guided Language Model Alignment",
                        "abstract": "Large-scale language models like ChatGPT and GPT-4 have gained attention for their impressive conversational and generative capabilities. However, the creation of supervised paired question-answering data for instruction tuning presents formidable challenges. This endeavor necessitates substantial human effort for data annotation and wrestles with issues concerning data quality, diversity, accuracy, and other related factors. To overcome these obstacles, we introduce an innovative framework named Self-QA, which replaces the traditional practice of human-written instruction seeds with a vast amount of unsupervised knowledge, enabling the model to generate a larger quantity of correct and domain-specific instruction data. The effectiveness of our proposed method is demonstrated through experiments conducted on unsupervised corpora from various domains."
                    },
                    {
                        "title": "TranS: Transition-based Knowledge Graph Embedding with Synthetic Relation Representation",
                        "abstract": "Knowledge graph embedding (KGE) aims to learn continuous vectors of relations and entities in knowledge graph. Recently, transition-based KGE methods have achieved promising performance, where the single relation vector learns to translate head entity to tail entity. However, this scoring pattern is not suitable for complex scenarios where the same entity pair has different relations. Previous models usually focus on the improvement of entity representation for 1-to-N, N-to-1 and N-to-N relations, but ignore the single relation vector. In this paper, we propose a novel transition-based method, TranS, for knowledge graph embedding. The single relation vector in traditional scoring patterns is replaced with synthetic relation representation, which can solve these issues effectively and efficiently. Experiments on a large knowledge graph dataset, ogbl-wikikg2, show that our model achieves state-of-the-art results."
                    },
                    {
                        "title": "XuanYuan 2.0: A Large Chinese Financial Chat Model with Hundreds of Billions Parameters",
                        "abstract": "In recent years, pre-trained language models have undergone rapid development with the emergence of large-scale models. However, there is a lack of open-sourced chat models specifically designed for the Chinese language, especially in the field of Chinese finance, at the scale of hundreds of billions. To address this gap, we introduce XuanYuan 2.0, the largest Chinese chat model to date, built upon the BLOOM-176B architecture. Additionally, we propose a novel training method called hybrid-tuning to mitigate catastrophic forgetting. By combining general-domain with domain-specific knowledge and integrating the stages of pre-training and fine-tuning, XuanYuan 2.0 is capable of providing accurate and contextually appropriate responses in the Chinese financial domain."
                    },
                    {
                        "title": "CGCE: A Chinese Generative Chat Evaluation Benchmark for General and Financial Domains",
                        "abstract": "Generative chat models, such as ChatGPT and GPT-4, have revolutionized natural language generation (NLG) by incorporating instructions and human feedback to achieve significant performance improvements. However, the lack of standardized evaluation benchmarks for chat models, particularly for Chinese and domain-specific models, hinders their assessment and progress. To address this gap, we introduce the Chinese Generative Chat Evaluation (CGCE) benchmark, focusing on general and financial domains. The CGCE benchmark encompasses diverse tasks, including 200 questions in the general domain and 150 specific professional questions in the financial domain. Manual scoring evaluates factors such as accuracy, coherence, expression clarity, and completeness. The CGCE benchmark provides researchers with a standardized framework to assess and compare Chinese generative chat models, fostering advancements in NLG research."
                    },
                    {
                        "title": "CRoSS: Diffusion Model Makes Controllable, Robust and Secure Image Steganography",
                        "abstract": "Current image steganography techniques are mainly focused on cover-based methods, which commonly have the risk of leaking secret images and poor robustness against degraded container images. Inspired by recent developments in diffusion models, we discovered that two properties of diffusion models, the ability to achieve translation between two images without training, and robustness to noisy data, can be used to improve security and natural robustness in image steganography tasks. For the choice of diffusion model, we selected Stable Diffusion, a type of conditional diffusion model, and fully utilized the latest tools from open-source communities, such as LoRAs and ControlNets, to improve the controllability and diversity of container images. In summary, we propose a novel image steganography framework, named Controllable, Robust and Secure Image Steganography (CRoSS), which has significant advantages in controllability, robustness, and security compared to cover-based image steganography methods. These benefits are obtained without additional training. To our knowledge, this is the first work to introduce diffusion models to the field of image steganography. In the experimental section, we conducted detailed experiments to demonstrate the advantages of our proposed CRoSS framework in controllability, robustness, and security."
                    },
                    {
                        "title": "HerosNet: Hyperspectral Explicable Reconstruction and Optimal Sampling Deep Network for Snapshot Compressive Imaging",
                        "abstract": "Hyperspectral imaging is an essential imaging modality for a wide range of applications, especially in remote sensing, agriculture, and medicine. Inspired by existing hyperspectral cameras that are either slow, expensive, or bulky, reconstructing hyperspectral images (HSIs) from a low-budget snapshot measurement has drawn wide attention. By mapping a truncated numerical optimization algorithm into a network with a fixed number of phases, recent deep unfolding networks (DUNs) for spectral snapshot compressive sensing (SCI) have achieved remarkable success. However, DUNs are far from reaching the scope of industrial applications limited by the lack of cross-phase feature interaction and adaptive parameter adjustment. In this paper, we propose a novel Hyperspectral Explicable Reconstruction and Optimal Sampling deep Network for SCI, dubbed HerosNet, which includes several phases under the ISTA-unfolding framework. Each phase can flexibly simulate the sensing matrix and contextually adjust the step size in the gradient descent step, and hierarchically fuse and interact the hidden states of previous phases to effectively recover current HSI frames in the proximal mapping step. Simultaneously, a hardware-friendly optimal binary mask is learned end-to-end to further improve the reconstruction performance. Finally, our HerosNet is validated to outperform the state-of-the-art methods on both simulation and real datasets by large margins. The source code is available at https://github.com/jianzhangcs/HerosNet."
                    }
                ]
            },
            "856e667e-31f9-438e-99be-5800922546d7": {
                "pk": "856e667e-31f9-438e-99be-5800922546d7",
                "name": "Jiarui Meng",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "1d90c6a3-8974-48a4-9832-dfdb90a0c4f9": {
                "pk": "1d90c6a3-8974-48a4-9832-dfdb90a0c4f9",
                "name": "Runyi Li",
                "collaborators": [
                    "Jian Zhang",
                    "Xuanyu Zhang",
                    "Weiqi Li",
                    "Zhipei Xu",
                    "Jiwen Yu",
                    "Youmin Xu",
                    "Xuhan Sheng",
                    "Yongbing Zhang",
                    "Zecheng Tang",
                    "Qing Huang"
                ],
                "domain": [
                    "Image Processing",
                    "Generative AI",
                    "Watermarking",
                    "Forensics"
                ],
                "publications": [
                    {
                        "title": "Diffusion Denoising for Low-Dose-CT Model",
                        "abstract": "Low-dose Computed Tomography (LDCT) reconstruction is an important task in medical image analysis. Recent years have seen many deep learning based methods, proved to be effective in this area. However, these methods mostly follow a supervised architecture, which needs paired CT image of full dose and quarter dose, and the solution is highly dependent on specific measurements. In this work, we introduce Denoising Diffusion LDCT Model, dubbed as DDLM, generating noise-free CT image using conditioned sampling. DDLM uses pretrained model, and need no training nor tuning process, thus our proposal is in unsupervised manner. Experiments on LDCT images have shown comparable performance of DDLM using less inference time, surpassing other state-of-the-art methods, proving both accurate and efficient. Implementation code will be set to public soon."
                    },
                    {
                        "title": "OmniSSR: Zero-shot Omnidirectional Image Super-Resolution using Stable Diffusion Model",
                        "abstract": "Omnidirectional images (ODIs) are commonly used in real-world visual tasks, and high-resolution ODIs help improve the performance of related visual tasks. Most existing super-resolution methods for ODIs use end-to-end learning strategies, resulting in inferior realness of generated images and a lack of effective out-of-domain generalization capabilities in training methods. Image generation methods represented by diffusion model provide strong priors for visual tasks and have been proven to be effectively applied to image restoration tasks. Leveraging the image priors of the Stable Diffusion (SD) model, we achieve omnidirectional image super-resolution with both fidelity and realness, dubbed as OmniSSR. Firstly, we transform the equirectangular projection (ERP) images into tangent projection (TP) images, whose distribution approximates the planar image domain. Then, we use SD to iteratively sample initial high-resolution results. At each denoising iteration, we further correct and update the initial results using the proposed Octadecaplex Tangent Information Interaction (OTII) and Gradient Decomposition (GD) technique to ensure better consistency. Finally, the TP images are transformed back to obtain the final high-resolution results. Our method is zero-shot, requiring no training or fine-tuning. Experiments of our method on two benchmark datasets demonstrate the effectiveness of our proposed method."
                    },
                    {
                        "title": "Protect-Your-IP: Scalable Source-Tracing and Attribution against Personalized Generation",
                        "abstract": "With the advent of personalized generation models, users can more readily create images resembling existing content, heightening the risk of violating portrait rights and intellectual property (IP). Traditional post-hoc detection and source-tracing methods for AI-generated content (AIGC) employ proactive watermark approaches; however, these are less effective against personalized generation models. Moreover, attribution techniques for AIGC rely on passive detection but often struggle to differentiate AIGC from authentic images, presenting a substantial challenge. Integrating these two processes into a cohesive framework not only meets the practical demands for protection and forensics but also improves the effectiveness of attribution tasks. Inspired by this insight, we propose a unified approach for image copyright source-tracing and attribution, introducing an innovative watermarking-attribution method that blends proactive and passive strategies. We embed copyright watermarks into protected images and train a watermark decoder to retrieve copyright information from the outputs of personalized models, using this watermark as an initial step for confirming if an image is AIGC-generated. To pinpoint specific generation techniques, we utilize powerful visual backbone networks for classification. Additionally, we implement an incremental learning strategy to adeptly attribute new personalized models without losing prior knowledge, thereby enhancing the model's adaptability to novel generation methods. We have conducted experiments using various celebrity portrait series sourced online, and the results affirm the efficacy of our method in source-tracing and attribution tasks, as well as its robustness against knowledge forgetting."
                    },
                    {
                        "title": "EditGuard: Versatile Image Watermarking for Tamper Localization and Copyright Protection",
                        "abstract": "In the era where AI-generated content (AIGC) models can produce stunning and lifelike images, the lingering shadow of unauthorized reproductions and malicious tampering poses imminent threats to copyright integrity and information security. Current image watermarking methods, while widely accepted for safeguarding visual content, can only protect copyright and ensure traceability. They fall short in localizing increasingly realistic image tampering, potentially leading to trust crises, privacy violations, and legal disputes. To solve this challenge, we propose an innovative proactive forensics framework EditGuard, to unify copyright protection and tamper-agnostic localization, especially for AIGC-based editing methods. It can offer a meticulous embedding of imperceptible watermarks and precise decoding of tampered areas and copyright information. Leveraging our observed fragility and locality of image-into-image steganography, the realization of EditGuard can be converted into a united image-bit steganography issue, thus completely decoupling the training process from the tampering types. Extensive experiments demonstrate that our EditGuard balances the tamper localization accuracy, copyright recovery precision, and generalizability to various AIGC-based tampering methods, especially for image forgery that is difficult for the naked eye to detect. The project page is available at https://xuanyuzhang21.github.io/project/editguard/."
                    },
                    {
                        "title": "V2A-Mark: Versatile Deep Visual-Audio Watermarking for Manipulation Localization and Copyright Protection",
                        "abstract": "AI-generated video has revolutionized short video production, filmmaking, and personalized media, making video local editing an essential tool. However, this progress also blurs the line between reality and fiction, posing challenges in multimedia forensics. To solve this urgent issue, V2A-Mark is proposed to address the limitations of current video tampering forensics, such as poor generalizability, singular function, and single modality focus. Combining the fragility of video-into-video steganography with deep robust watermarking, our method can embed invisible visual-audio localization watermarks and copyright watermarks into the original video frames and audio, enabling precise manipulation localization and copyright protection. We also design a temporal alignment and fusion module and degradation prompt learning to enhance the localization accuracy and decoding robustness. Meanwhile, we introduce a sample-level audio localization method and a cross-modal copyright extraction mechanism to couple the information of audio and video frames. The effectiveness of V2A-Mark has been verified on a visual-audio tampering dataset, emphasizing its superiority in localization precision and copyright accuracy, crucial for the sustainable development of video editing in the AIGC video era."
                    },
                    {
                        "title": "FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models",
                        "abstract": "The rapid development of generative AI is a double-edged sword, which not only facilitates content creation but also makes image manipulation easier and more difficult to detect. Although current image forgery detection and localization (IFDL) methods are generally effective, they tend to face two challenges: \\textbf{1)} black-box nature with unknown detection principle, \\textbf{2)} limited generalization across diverse tampering methods (e.g., Photoshop, DeepFake, AIGC-Editing). To address these issues, we propose the explainable IFDL task and design FakeShield, a multi-modal framework capable of evaluating image authenticity, generating tampered region masks, and providing a judgment basis based on pixel-level and image-level tampering clues. Additionally, we leverage GPT-4o to enhance existing IFDL datasets, creating the Multi-Modal Tamper Description dataSet (MMTD-Set) for training FakeShield's tampering analysis capabilities. Meanwhile, we incorporate a Domain Tag-guided Explainable Forgery Detection Module (DTE-FDM) and a Multi-modal Forgery Localization Module (MFLM) to address various types of tamper detection interpretation and achieve forgery localization guided by detailed textual descriptions. Extensive experiments demonstrate that FakeShield effectively detects and localizes various tampering techniques, offering an explainable and superior solution compared to previous IFDL methods."
                    }
                ]
            },
            "374c03b8-8340-49cc-bdf2-f49c766cd327": {
                "pk": "374c03b8-8340-49cc-bdf2-f49c766cd327",
                "name": "Zhipei Xu",
                "collaborators": [
                    "Runyi Li",
                    "Xuanyu Zhang",
                    "Jian Zhang",
                    "Yongbing Zhang",
                    "Youmin Xu",
                    "Jiwen Yu",
                    "Weiqi Li",
                    "Zecheng Tang",
                    "Qing Huang",
                    "Yirong Chen"
                ],
                "domain": [
                    "AI-generated content",
                    "Multimedia forensics",
                    "Generative AI",
                    "Large language models"
                ],
                "publications": [
                    {
                        "title": "Protect-Your-IP: Scalable Source-Tracing and Attribution against Personalized Generation",
                        "abstract": "With the advent of personalized generation models, users can more readily create images resembling existing content, heightening the risk of violating portrait rights and intellectual property (IP). Traditional post-hoc detection and source-tracing methods for AI-generated content (AIGC) employ proactive watermark approaches; however, these are less effective against personalized generation models. Moreover, attribution techniques for AIGC rely on passive detection but often struggle to differentiate AIGC from authentic images, presenting a substantial challenge. Integrating these two processes into a cohesive framework not only meets the practical demands for protection and forensics but also improves the effectiveness of attribution tasks. Inspired by this insight, we propose a unified approach for image copyright source-tracing and attribution, introducing an innovative watermarking-attribution method that blends proactive and passive strategies. We embed copyright watermarks into protected images and train a watermark decoder to retrieve copyright information from the outputs of personalized models, using this watermark as an initial step for confirming if an image is AIGC-generated. To pinpoint specific generation techniques, we utilize powerful visual backbone networks for classification. Additionally, we implement an incremental learning strategy to adeptly attribute new personalized models without losing prior knowledge, thereby enhancing the model's adaptability to novel generation methods. We have conducted experiments using various celebrity portrait series sourced online, and the results affirm the efficacy of our method in source-tracing and attribution tasks, as well as its robustness against knowledge forgetting."
                    },
                    {
                        "title": "V2A-Mark: Versatile Deep Visual-Audio Watermarking for Manipulation Localization and Copyright Protection",
                        "abstract": "AI-generated video has revolutionized short video production, filmmaking, and personalized media, making video local editing an essential tool. However, this progress also blurs the line between reality and fiction, posing challenges in multimedia forensics. To solve this urgent issue, V2A-Mark is proposed to address the limitations of current video tampering forensics, such as poor generalizability, singular function, and single modality focus. Combining the fragility of video-into-video steganography with deep robust watermarking, our method can embed invisible visual-audio localization watermarks and copyright watermarks into the original video frames and audio, enabling precise manipulation localization and copyright protection. We also design a temporal alignment and fusion module and degradation prompt learning to enhance the localization accuracy and decoding robustness. Meanwhile, we introduce a sample-level audio localization method and a cross-modal copyright extraction mechanism to couple the information of audio and video frames. The effectiveness of V2A-Mark has been verified on a visual-audio tampering dataset, emphasizing its superiority in localization precision and copyright accuracy, crucial for the sustainable development of video editing in the AIGC video era."
                    },
                    {
                        "title": "FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models",
                        "abstract": "The rapid development of generative AI is a double-edged sword, which not only facilitates content creation but also makes image manipulation easier and more difficult to detect. Although current image forgery detection and localization (IFDL) methods are generally effective, they tend to face two challenges: \\textbf{1)} black-box nature with unknown detection principle, \\textbf{2)} limited generalization across diverse tampering methods (e.g., Photoshop, DeepFake, AIGC-Editing). To address these issues, we propose the explainable IFDL task and design FakeShield, a multi-modal framework capable of evaluating image authenticity, generating tampered region masks, and providing a judgment basis based on pixel-level and image-level tampering clues. Additionally, we leverage GPT-4o to enhance existing IFDL datasets, creating the Multi-Modal Tamper Description dataSet (MMTD-Set) for training FakeShield's tampering analysis capabilities. Meanwhile, we incorporate a Domain Tag-guided Explainable Forgery Detection Module (DTE-FDM) and a Multi-modal Forgery Localization Module (MFLM) to address various types of tamper detection interpretation and achieve forgery localization guided by detailed textual descriptions. Extensive experiments demonstrate that FakeShield effectively detects and localizes various tampering techniques, offering an explainable and superior solution compared to previous IFDL methods."
                    },
                    {
                        "title": "BianQue: Balancing the Questioning and Suggestion Ability of Health LLMs with Multi-turn Health Conversations Polished by ChatGPT",
                        "abstract": "Large language models (LLMs) have performed well in providing general and extensive health suggestions in single-turn conversations, exemplified by systems such as ChatGPT, ChatGLM, ChatDoctor, DoctorGLM, and etc. However, the limited information provided by users during single turn results in inadequate personalization and targeting of the generated suggestions, which requires users to independently select the useful part. It is mainly caused by the missing ability to engage in multi-turn questioning. In real-world medical consultations, doctors usually employ a series of iterative inquiries to comprehend the patient's condition thoroughly, enabling them to provide effective and personalized suggestions subsequently, which can be defined as chain of questioning (CoQ) for LLMs. To improve the CoQ of LLMs, we propose BianQue, a ChatGLM-based LLM finetuned with the self-constructed health conversation dataset BianQueCorpus that is consist of multiple turns of questioning and health suggestions polished by ChatGPT. Experimental results demonstrate that the proposed BianQue can simultaneously balance the capabilities of both questioning and health suggestions, which will help promote the research and application of LLMs in the field of proactive health."
                    }
                ]
            },
            "954d9e01-8602-4152-8e9f-de306314842d": {
                "pk": "954d9e01-8602-4152-8e9f-de306314842d",
                "name": "Yongbing Zhang",
                "collaborators": [
                    "Jian Zhang",
                    "Ye Zhang",
                    "Linghan Cai",
                    "Xuanyu Zhang",
                    "Guofang Wang",
                    "Ziyue Wang",
                    "Yifeng Wang",
                    "Jiayu Li",
                    "Zhengxiang Chen",
                    "Jih-Hsin Cheng"
                ],
                "domain": [
                    "Geometric Analysis",
                    "Machine Learning",
                    "Image Segmentation",
                    "Differential Geometry"
                ],
                "publications": [
                    {
                        "title": "A monotonicity formula and Type-II singularities for the mean curvature flow",
                        "abstract": "In this paper, we introduce a monotonicity formula for the mean curvature flow. We also apply this monotonicity formula to study the asymptotic behavior of eternal solutions."
                    },
                    {
                        "title": "Graham-Witten's conformal invariant for closed four dimensional submanifolds",
                        "abstract": "It was proved by Graham and Witten in 1999 that conformal invariants of submanifolds can be obtained via volume renormalization of minimal surfaces in conformally compact Einstein manifolds. The conformal invariant of a submanifold $\\Sigma$ is contained in the volume expansion of the minimal surface which is asymptotic to $\\Sigma$ when the minimal surface approaches the conformaly infinity. In the paper we give the explicit expression of Graham-Witten's conformal invariant for closed four dimensional submanifolds and find critical points of the conformal invariant in the case of Euclidean ambient spaces."
                    },
                    {
                        "title": "The Sasaki-Ricci flow on Sasakian 3-spheres",
                        "abstract": "We show that on a Sasakian 3-sphere the Sasaki-Ricci flow initiating from a Sasakian metric of positive transverse scalar curvature converges to a gradient Sasaki- Ricci soliton. We also show the existence and uniqueness of gradient Sasaki-Ricci soliton on each Sasakian 3-sphere."
                    },
                    {
                        "title": "Lagrangian F-stability of closed Lagrangian self-shrinkers",
                        "abstract": "In this paper, we study the Lagrangian F-stability of closed Lagrangian self-shrinkers immersed in complex Euclidean space. We show that any closed Lagrangian self-shrinker with first Betti number greater than one is Lagrangian F-unstable. In particular, any two-dimensional embedded closed Lagrangian self-shrinker is Lagrangian F-unstable. For a closed Lagrangian self-shrinker with first Betti number equal to one, we show that Lagrangian F-stability is equivalent to Hamiltonian F-stability. We also characterize Hamiltonian F-stability of a closed Lagrangian self-shrinker by its spectral property of the twisted Laplacian."
                    },
                    {
                        "title": "Stabilities of homothetically shrinking Yang-Mills solitons",
                        "abstract": "In this paper we introduce entropy-stability and F-stability for homothetically shrinking Yang-Mills solitons, employing entropy and second variation of $\\mathcal{F}$-functional respectively. For a homothetically shrinking soliton which does not descend, we prove that entropy-stability implies F-stability. These stabilities have connections with the study of Type-I singularities of the Yang-Mills flow. Two byproducts are also included: We show that the Yang-Mills flow in dimension four cannot develop a Type-I singularity; and we obtain a gap theorem for homothetically shrinking solitons."
                    },
                    {
                        "title": "A conformal integral invariant on Riemannian foliations",
                        "abstract": "Let $M$ be a closed manifold which admits a foliation structure $\\mathcal{F}$ of codimension $q\\geq 2$ and a bundle-like metric $g_0$. Let $[g_0]_B$ be the space of bundle-like metrics which differ from $g_0$ only along the horizontal directions by a multiple of a positive basic function. Assume $Y$ is a transverse conformal vector field and the mean curvature of the leaves of $(M,\\mathcal{F},g_0)$ vanishes. We show that the integral $\\int_MY(R^T_{g^T})d\\mu_g$ is independent of the choice of $g\\in [g_0]_B$, where $g^T$ is the transverse metric induced by $g$ and $R^T$ is the transverse scalar curvature. Moreover if $q\\geq 3$, we have $\\int_MY(R^T_{g^T})d\\mu_g=0$ for any $g\\in [g_0]_B$. However there exist codimension 2 minimal Riemannian foliations $(M,\\mathcal{F},g)$ and transverse conformal vector fields $Y$ such that $\\int_MY(R^T_{g^T})d\\mu_g\\neq 0$. Therefore, it is a nontrivial obstruction for the transverse Yamabe problem on minimal Riemannian foliation of codimension 2."
                    },
                    {
                        "title": "SEINE: Structure Encoding and Interaction Network for Nuclei Instance Segmentation",
                        "abstract": "Nuclei instance segmentation in histopathological images is of great importance for biological analysis and cancer diagnosis but remains challenging for two reasons. (1) Similar visual presentation of intranuclear and extranuclear regions of chromophobe nuclei often causes under-segmentation, and (2) current methods lack the exploration of nuclei structure, resulting in fragmented instance predictions. To address these problems, this paper proposes a structure encoding and interaction network, termed SEINE, which develops the structure modeling scheme of nuclei and exploits the structure similarity between nuclei to improve the integrality of each segmented instance. Concretely, SEINE introduces a contour-based structure encoding (SE) that considers the correlation between nuclei structure and semantics, realizing a reasonable representation of the nuclei structure. Based on the encoding, we propose a structure-guided attention (SGA) module that takes the clear nuclei as prototypes to enhance the structure learning for the fuzzy nuclei. To strengthen the structural learning ability, a semantic feature fusion (SFF) is presented to boost the semantic consistency of semantic and structure branches. Furthermore, a position enhancement (PE) method is applied to suppress incorrect nuclei boundary predictions. Extensive experiments demonstrate the superiority of our approaches, and SEINE achieves state-of-the-art (SOTA) performance on four datasets. The code is available at https://github.com/zhangye-zoe/SEINE."
                    },
                    {
                        "title": "Invariant surface area functionals and singular Yamabe problem in 3-dimensional CR geometry",
                        "abstract": "We express two CR invariant surface area elements in terms of quantities in pseudohermitian geometry. We deduce the Euler-Lagrange equations of the associated energy functionals. Many solutions are given and discussed. In relation to the singular CR Yamabe problem, we show that one of the energy functionals appears as the coefficient (up to a constant multiple) of the log term in the associated volume renormalization."
                    },
                    {
                        "title": "PUERT: Probabilistic Under-sampling and Explicable Reconstruction Network for CS-MRI",
                        "abstract": "Compressed Sensing MRI (CS-MRI) aims at reconstructing de-aliased images from sub-Nyquist sampling k-space data to accelerate MR Imaging, thus presenting two basic issues, i.e., where to sample and how to reconstruct. To deal with both problems simultaneously, we propose a novel end-to-end Probabilistic Under-sampling and Explicable Reconstruction neTwork, dubbed PUERT, to jointly optimize the sampling pattern and the reconstruction network. Instead of learning a deterministic mask, the proposed sampling subnet explores an optimal probabilistic sub-sampling pattern, which describes independent Bernoulli random variables at each possible sampling point, thus retaining robustness and stochastics for a more reliable CS reconstruction. A dynamic gradient estimation strategy is further introduced to gradually approximate the binarization function in backward propagation, which efficiently preserves the gradient information and further improves the reconstruction quality. Moreover, in our reconstruction subnet, we adopt a model-based network design scheme with high efficiency and interpretability, which is shown to assist in further exploitation for the sampling subnet. Extensive experiments on two widely used MRI datasets demonstrate that our proposed PUERT not only achieves state-of-the-art results in terms of both quantitative metrics and visual quality but also yields a sub-sampling pattern and a reconstruction model that are both customized to training data."
                    },
                    {
                        "title": "On the Mathematical Understanding of ResNet with Feynman Path Integral",
                        "abstract": "In this paper, we aim to understand Residual Network (ResNet) in a scientifically sound way by providing a bridge between ResNet and Feynman path integral. In particular, we prove that the effect of residual block is equivalent to partial differential equation, and the ResNet transforming process can be equivalently converted to Feynman path integral. These conclusions greatly help us mathematically understand the advantage of ResNet in addressing the gradient vanishing issue. More importantly, our analyses offer a path integral view of ResNet, and demonstrate that the output of certain network can be obtained by adding contributions of all paths. Moreover, the contribution of each path is proportional to e^{-S}, where S is the action given by time integral of Lagrangian L. This lays the solid foundation in the understanding of ResNet, and provides insights in the future design of convolutional neural network architecture. Based on these results, we have designed the network using partial differential operators, which further validates our theoritical analyses."
                    },
                    {
                        "title": "Freshness-Optimal Caching for Information Updating Systems with Limited Cache Storage Capacity",
                        "abstract": "In this paper, we investigate a cache updating system with a server containing $N$ files, $K$ relays and $M$ users. The server keeps the freshest versions of the files which are updated with fixed rates. Each relay can download the fresh files from the server in a certain period of time. Each user can get the fresh files from any relay as long as the relay has stored the fresh versions of the requested files. Due to the limited storage capacity and updating capacity of each relay, different cache designs will lead to different average freshness of all updating files at users. In order to keep the average freshness as large as possible in the cache updating system, we formulate an average freshness-optimal cache updating problem (AFOCUP) to obtain an optimal cache scheme. However, because of the nonlinearity of the AFOCUP, it is difficult to seek out the optimal cache scheme. As a result, an linear approximate model is suggested by distributing the total update rates completely in accordance with the number of files in the relay in advance. Then we utilize the greedy algorithm to search the optimal cache scheme that is satisfied with the limited storage capacity of each relay. Finally, some numerical examples are provided to illustrate the performance of the approximate solution."
                    },
                    {
                        "title": "HerosNet: Hyperspectral Explicable Reconstruction and Optimal Sampling Deep Network for Snapshot Compressive Imaging",
                        "abstract": "Hyperspectral imaging is an essential imaging modality for a wide range of applications, especially in remote sensing, agriculture, and medicine. Inspired by existing hyperspectral cameras that are either slow, expensive, or bulky, reconstructing hyperspectral images (HSIs) from a low-budget snapshot measurement has drawn wide attention. By mapping a truncated numerical optimization algorithm into a network with a fixed number of phases, recent deep unfolding networks (DUNs) for spectral snapshot compressive sensing (SCI) have achieved remarkable success. However, DUNs are far from reaching the scope of industrial applications limited by the lack of cross-phase feature interaction and adaptive parameter adjustment. In this paper, we propose a novel Hyperspectral Explicable Reconstruction and Optimal Sampling deep Network for SCI, dubbed HerosNet, which includes several phases under the ISTA-unfolding framework. Each phase can flexibly simulate the sensing matrix and contextually adjust the step size in the gradient descent step, and hierarchically fuse and interact the hidden states of previous phases to effectively recover current HSI frames in the proximal mapping step. Simultaneously, a hardware-friendly optimal binary mask is learned end-to-end to further improve the reconstruction performance. Finally, our HerosNet is validated to outperform the state-of-the-art methods on both simulation and real datasets by large margins. The source code is available at https://github.com/jianzhangcs/HerosNet."
                    },
                    {
                        "title": "Self-Supervised Scalable Deep Compressed Sensing",
                        "abstract": "Compressed sensing (CS) is a promising tool for reducing sampling costs. Current deep neural network (NN)-based CS methods face the challenges of collecting labeled measurement-ground truth (GT) data and generalizing to real applications. This paper proposes a novel $\\mathbf{S}$elf-supervised s$\\mathbf{C}$alable deep CS method, comprising a deep $\\mathbf{L}$earning scheme called $\\mathbf{SCL}$ and a family of $\\mathbf{Net}$works named $\\mathbf{SCNet}$, which does not require GT and can handle arbitrary sampling ratios and matrices once trained on a partial measurement set. Our SCL contains a dual-domain loss and a four-stage recovery strategy. The former encourages a cross-consistency on two measurement parts and a sampling-reconstruction cycle-consistency regarding arbitrary ratios and matrices to maximize data/information utilization. The latter can progressively leverage common signal prior in external measurements and internal characteristics of test samples and learned NNs to improve accuracy. SCNet combines both the explicit guidance from optimization algorithms with implicit regularization from advanced NN blocks to learn a collaborative signal representation. Our theoretical analyses and experiments on simulated and real captured data, covering 1-/2-/3-D natural and scientific signals, demonstrate the effectiveness, superior performance, flexibility, and generalization ability of our method over existing self-supervised methods and its significant potential in competing against state-of-the-art supervised methods. Code is available at https://github.com/Guaishou74851/SCNet."
                    },
                    {
                        "title": "Protect-Your-IP: Scalable Source-Tracing and Attribution against Personalized Generation",
                        "abstract": "With the advent of personalized generation models, users can more readily create images resembling existing content, heightening the risk of violating portrait rights and intellectual property (IP). Traditional post-hoc detection and source-tracing methods for AI-generated content (AIGC) employ proactive watermark approaches; however, these are less effective against personalized generation models. Moreover, attribution techniques for AIGC rely on passive detection but often struggle to differentiate AIGC from authentic images, presenting a substantial challenge. Integrating these two processes into a cohesive framework not only meets the practical demands for protection and forensics but also improves the effectiveness of attribution tasks. Inspired by this insight, we propose a unified approach for image copyright source-tracing and attribution, introducing an innovative watermarking-attribution method that blends proactive and passive strategies. We embed copyright watermarks into protected images and train a watermark decoder to retrieve copyright information from the outputs of personalized models, using this watermark as an initial step for confirming if an image is AIGC-generated. To pinpoint specific generation techniques, we utilize powerful visual backbone networks for classification. Additionally, we implement an incremental learning strategy to adeptly attribute new personalized models without losing prior knowledge, thereby enhancing the model's adaptability to novel generation methods. We have conducted experiments using various celebrity portrait series sourced online, and the results affirm the efficacy of our method in source-tracing and attribution tasks, as well as its robustness against knowledge forgetting."
                    },
                    {
                        "title": "Rethinking Attention-Based Multiple Instance Learning for Whole-Slide Pathological Image Classification: An Instance Attribute Viewpoint",
                        "abstract": "Multiple instance learning (MIL) is a robust paradigm for whole-slide pathological image (WSI) analysis, processing gigapixel-resolution images with slide-level labels. As pioneering efforts, attention-based MIL (ABMIL) and its variants are increasingly becoming popular due to the characteristics of simultaneously handling clinical diagnosis and tumor localization. However, the attention mechanism exhibits limitations in discriminating between instances, which often misclassifies tissues and potentially impairs MIL performance. This paper proposes an Attribute-Driven MIL (AttriMIL) framework to address these issues. Concretely, we dissect the calculation process of ABMIL and present an attribute scoring mechanism that measures the contribution of each instance to bag prediction effectively, quantifying instance attributes. Based on attribute quantification, we develop a spatial attribute constraint and an attribute ranking constraint to model instance correlations within and across slides, respectively. These constraints encourage the network to capture the spatial correlation and semantic similarity of instances, improving the ability of AttriMIL to distinguish tissue types and identify challenging instances. Additionally, AttriMIL employs a histopathology adaptive backbone that maximizes the pre-trained model's feature extraction capability for collecting pathological features. Extensive experiments on three public benchmarks demonstrate that our AttriMIL outperforms existing state-of-the-art frameworks across multiple evaluation metrics. The implementation code is available at https://github.com/MedCAI/AttriMIL."
                    },
                    {
                        "title": "Dynamic Pseudo Label Optimization in Point-Supervised Nuclei Segmentation",
                        "abstract": "Deep learning has achieved impressive results in nuclei segmentation, but the massive requirement for pixel-wise labels remains a significant challenge. To alleviate the annotation burden, existing methods generate pseudo masks for model training using point labels. However, the generated masks are inevitably different from the ground truth, and these dissimilarities are not handled reasonably during the network training, resulting in the subpar performance of the segmentation model. To tackle this issue, we propose a framework named DoNuSeg, enabling \\textbf{D}ynamic pseudo label \\textbf{O}ptimization in point-supervised \\textbf{Nu}clei \\textbf{Seg}mentation. Specifically, DoNuSeg takes advantage of class activation maps (CAMs) to adaptively capture regions with semantics similar to annotated points. To leverage semantic diversity in the hierarchical feature levels, we design a dynamic selection module to choose the optimal one among CAMs from different encoder blocks as pseudo masks. Meanwhile, a CAM-guided contrastive module is proposed to further enhance the accuracy of pseudo masks. In addition to exploiting the semantic information provided by CAMs, we consider location priors inherent to point labels, developing a task-decoupled structure for effectively differentiating nuclei. Extensive experiments demonstrate that DoNuSeg outperforms state-of-the-art point-supervised methods. The code is available at https://github.com/shinning0821/MICCAI24-DoNuSeg."
                    },
                    {
                        "title": "AugDiff: Diffusion based Feature Augmentation for Multiple Instance Learning in Whole Slide Image",
                        "abstract": "Multiple Instance Learning (MIL), a powerful strategy for weakly supervised learning, is able to perform various prediction tasks on gigapixel Whole Slide Images (WSIs). However, the tens of thousands of patches in WSIs usually incur a vast computational burden for image augmentation, limiting the MIL model's improvement in performance. Currently, the feature augmentation-based MIL framework is a promising solution, while existing methods such as Mixup often produce unrealistic features. To explore a more efficient and practical augmentation method, we introduce the Diffusion Model (DM) into MIL for the first time and propose a feature augmentation framework called AugDiff. Specifically, we employ the generation diversity of DM to improve the quality of feature augmentation and the step-by-step generation property to control the retention of semantic information. We conduct extensive experiments over three distinct cancer datasets, two different feature extractors, and three prevalent MIL algorithms to evaluate the performance of AugDiff. Ablation study and visualization further verify the effectiveness. Moreover, we highlight AugDiff's higher-quality augmented feature over image augmentation and its superiority over self-supervised learning. The generalization over external datasets indicates its broader applications."
                    }
                ]
            },
            "0b57914d-ba99-4941-8c74-e551f22b63d9": {
                "pk": "0b57914d-ba99-4941-8c74-e551f22b63d9",
                "name": "Jian Zhang",
                "collaborators": [
                    "Naihuan Jing",
                    "Caiqing Jian",
                    "Xinyu Cheng",
                    "Lihui Wang",
                    "Jianqing Fan",
                    "Mengxue Bai",
                    "Shuihui Zhu",
                    "Chaomei Chen",
                    "Wenming Zou",
                    "Ioannis Mitliagkas"
                ],
                "domain": [
                    "Modal Logic",
                    "Path Planning",
                    "Quantum Chemistry",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "On the Relational Translation Method for Propositional Modal Logics",
                        "abstract": "One way of proving theorems in modal logics is translating them into the predicate calculus and then using conventional resolution-style theorem provers. This approach has been regarded as inappropriate in practice, because the resulting formulas are too lengthy and it is impossible to show the non-theoremhood of modal formulas. In this paper, we demonstrate the practical feasibility of the (relational) translation method. Using a state-of-the-art theorem prover for first-order predicate logic, we proved many benchmark theorems available from the modal logic literature. We show the invalidity of propositional modal and temporal logic formulas, using model generators or satisfiability testers for the classical logic. Many satisfiable formulas are found to have very small models. Finally, several different approaches are compared."
                    },
                    {
                        "title": "AI based Algorithms of Path Planning, Navigation and Control for Mobile Ground Robots and UAVs",
                        "abstract": "As the demands of autonomous mobile robots are increasing in recent years, the requirement of the path planning/navigation algorithm should not be content with the ability to reach the target without any collisions, but also should try to achieve possible optimal or suboptimal path from the initial position to the target according to the robot's constrains in practice. This report investigates path planning and control strategies for mobile robots with machine learning techniques, including ground mobile robots and flying UAVs.   In this report, the hybrid reactive collision-free navigation problem under an unknown static environment is investigated firstly. By combining both the reactive navigation and Q-learning method, we intend to keep the good characteristics of reactive navigation algorithm and Q-learning and overcome the shortcomings of only relying on one of them. The proposed method is then extended into 3D environments. The performance of the mentioned strategies are verified by extensive computer simulations, and good results are obtained. Furthermore, the more challenging dynamic environment situation is taken into our consideration. We tackled this problem by developing a new path planning method that utilizes the integrated environment representation and reinforcement learning. Our novel approach enables to find the optimal path to the target efficiently and avoid collisions in a cluttered environment with steady and moving obstacles. The performance of these methods is compared with other different aspects."
                    },
                    {
                        "title": "Scalar Coupling Constant Prediction Using Graph Embedding Local Attention Encoder",
                        "abstract": "Scalar coupling constant (SCC) plays a key role in the analysis of three-dimensional structure of organic matter, however, the traditional SCC prediction using quantum mechanical calculations is very time-consuming. To calculate SCC efficiently and accurately, we proposed a graph embedding local self-attention encoder (GELAE) model, in which, a novel invariant structure representation of the coupling system in terms of bond length, bond angle and dihedral angle was presented firstly, and then a local self-attention module embedded with the adjacent matrix of a graph was designed to extract effectively the features of coupling systems, finally, with a modified classification loss function, the SCC was predicted. To validate the superiority of the proposed method, we conducted a series of comparison experiments using different structure representations, different attention modules, and different losses. The experimental results demonstrate that, compared to the traditional chemical bond structure representations, the rotation and translation invariant structure representations proposed in this work can improve the SCC prediction accuracy; with the graph embedded local self-attention, the mean absolute error (MAE) of the prediction model in the validation set decreases from 0.1603 Hz to 0.1067 Hz; using the classification based loss function instead of the scaled regression loss, the MAE of the predicted SCC can be decreased to 0.0963 HZ, which is close to the quantum chemistry standard on CHAMPS dataset."
                    },
                    {
                        "title": "Sieve empirical likelihood ratio tests for nonparametric functions",
                        "abstract": "Generalized likelihood ratio statistics have been proposed in Fan, Zhang and Zhang [Ann. Statist. 29 (2001) 153-193] as a generally applicable method for testing nonparametric hypotheses about nonparametric functions. The likelihood ratio statistics are constructed based on the assumption that the distributions of stochastic errors are in a certain parametric family. We extend their work to the case where the error distribution is completely unspecified via newly proposed sieve empirical likelihood ratio (SELR) tests. The approach is also applied to test conditional estimating equations on the distributions of stochastic errors. It is shown that the proposed SELR statistics follow asymptotically rescaled \\chi^2-distributions, with the scale constants and the degrees of freedom being independent of the nuisance parameters. This demonstrates that the Wilks phenomenon observed in Fan, Zhang and Zhang [Ann. Statist. 29 (2001) 153-193] continues to hold under more relaxed models and a larger class of techniques. The asymptotic power of the proposed test is also derived, which achieves the optimal rate for nonparametric hypothesis testing. The proposed approach has two advantages over the generalized likelihood ratio method: it requires one only to specify some conditional estimating equations rather than the entire distribution of the stochastic error, and the procedure adapts automatically to the unknown error distribution including heteroscedasticity. A simulation study is conducted to evaluate our proposed procedure empirically."
                    },
                    {
                        "title": "Stability of Big Solitons in a Competitive Power Nonlinear Schr\u00f6dinger Equation",
                        "abstract": "By introducing and solving two correlative constrained variational problems as well as spectrum analysis, an approach to fix soliton frequency from the prescribed mass for nonlinear Schr\\\"odinger equations is found, and an open problem in normalized solutions is answered. Then existence and orbital stability of big solitons depending on frequencies for nonlinear Schr\\\"odinger equation with competitive power nonlinearity is proved for the first time. In addition multi-solitons of the equation with different speeds are constructed by stable big solitons."
                    },
                    {
                        "title": "The monotonicity conjecture and stability of solitons for the Cubic-Quintic NLS on R^3",
                        "abstract": "In this paper, we prove stability or instability of solitons for the cubic-quintic nonlinear Schrodinger equation at every frequency. The monotonicity conjecture raised by Killip, Oh, Pocovnicu and Visan is resolved. We introduce and solve a new cross-constrained variational problem. Then the uniqueness of the energy minimizers is proposed and shown. According to a spectral approach and variational arguments, we develop a set of geometric analysis methods. Correspondences between the soliton frequency and the prescribed mass are established. Classification of normalized solutions is given."
                    },
                    {
                        "title": "Multiparameter quantum Pfaffians",
                        "abstract": "The multiparameter quantum Pfaffian of the $(p, \\lambda)$-quantum group is introduced and studied together with the quantum determinant, and an identity relating the two invariants is given. Generalization to the multiparameter hyper-Pfaffian and relationship with the quantum minors are also considered."
                    },
                    {
                        "title": "Dynamical quantum determinants and Pfaffians",
                        "abstract": "We introduce the dynamical quantum Pfaffian on the dynamical quantum general linear group and prove its fundamental transformation identity. Hyper quantum dynamical Pfaffian is also introduced and formulas connecting them are given."
                    },
                    {
                        "title": "Collaboration in an Open Data eScience: A Case Study of Sloan Digital Sky Survey",
                        "abstract": "Current science and technology has produced more and more publically accessible scientific data. However, little is known about how the open data trend impacts a scientific community, specifically in terms of its collaboration behaviors. This paper aims to enhance our understanding of the dynamics of scientific collaboration in the open data eScience environment via a case study of co-author networks of an active and highly cited open data project, called Sloan Digital Sky Survey. We visualized the co-authoring networks and measured their properties over time at three levels: author, institution, and country levels. We compared these measurements to a random network model and also compared results across the three levels. The study found that 1) the collaboration networks of the SDSS community transformed from random networks to small-world networks; 2) the number of author-level collaboration instances has not changed much over time, while the number of collaboration instances at the other two levels has increased over time; 3) pairwise institutional collaboration become common in recent years. The open data trend may have both positive and negative impacts on scientific collaboration."
                    },
                    {
                        "title": "Multiplicity and concentration behavior of solutions to the critical Kirchhoff type problem",
                        "abstract": "In this paper, we study the multiplicity and concentration of the positive solutions to the following critical Kirchhoff type problem: \\begin{equation*}   -\\left(\\varepsilon^2 a+\\varepsilon b\\int_{\\R^3}|\\nabla u|^2\\mathrm{d} x\\right)\\Delta u + V(x) u = f(u)+u^5\\ \\ {\\rm in } \\ \\ \\R^3, \\end{equation*} where $\\varepsilon$ is a small positive parameter, $a$, $b$ are positive constants, $V \\in C(\\mathbb{R}^3)$ is a positive potential, $f \\in C^1(\\R^+, \\R)$ is a subcritical nonlinear term, $u^5$ is a pure critical nonlinearity. When $\\varepsilon>0$ small, we establish the relationship between the number of positive solutions and the profile of the potential $V$. The exponential decay at infinity of the solution is also obtained. In particular, we show that each solution concentrates around a local strict minima of $V$ as $\\varepsilon \\rightarrow 0$."
                    },
                    {
                        "title": "YellowFin and the Art of Momentum Tuning",
                        "abstract": "Hyperparameter tuning is one of the most time-consuming workloads in deep learning. State-of-the-art optimizers, such as AdaGrad, RMSProp and Adam, reduce this labor by adaptively tuning an individual learning rate for each variable. Recently researchers have shown renewed interest in simpler methods like momentum SGD as they may yield better test metrics. Motivated by this trend, we ask: can simple adaptive methods based on SGD perform as well or better? We revisit the momentum SGD algorithm and show that hand-tuning a single learning rate and momentum makes it competitive with Adam. We then analyze its robustness to learning rate misspecification and objective curvature variation. Based on these insights, we design YellowFin, an automatic tuner for momentum and learning rate in SGD. YellowFin optionally uses a negative-feedback loop to compensate for the momentum dynamics in asynchronous settings on the fly. We empirically show that YellowFin can converge in fewer iterations than Adam on ResNets and LSTMs for image recognition, language modeling and constituency parsing, with a speedup of up to 3.28x in synchronous and up to 2.69x in asynchronous settings."
                    }
                ]
            }
        }
    },
    "2406.12763": {
        "paper_data": {
            "title": "Implicit Bias of Mirror Flow on Separable Data",
            "url": "http://arxiv.org/abs/2406.12763v2",
            "arxiv_id": "2406.12763",
            "authors": [
                "Scott Pesme",
                "Radu-Alexandru Dragomir",
                "Nicolas Flammarion"
            ],
            "abstract": "We examine the continuous-time counterpart of mirror descent, namely mirror flow, on classification problems which are linearly separable. Such problems are minimised `at infinity' and have many possible solutions; we study which solution is preferred by the algorithm depending on the mirror potential. For exponential tailed losses and under mild assumptions on the potential, we show that the iterates converge in direction towards a $\\phi_\\infty$-maximum margin classifier. The function $\\phi_\\infty$ is the $\\textit{horizon function}$ of the mirror potential and characterises its shape `at infinity'. When the potential is separable, a simple formula allows to compute this function. We analyse several examples of potentials and provide numerical experiments highlighting our results.",
            "introduction": "   1 Introduction  Heavily over-parametrised yet barely regularised neural networks can easily perfectly fit a noisy training set while still performing very well on unseen data\u00a0(Zhang et\u00a0al., 2017). This statistical phenomenon is surprising since it is known that there exists interpolating solutions which have terrible generalisation performances\u00a0(Liu et\u00a0al., 2020). To understand this benign overfitting, it is essential to take into account the training algorithm. If overfitting is indeed harmless, it must be because the optimisation process has steered us towards a solution with favorable generalisation properties.   From this simple observation, a major line of work studying the implicit regularisation of gradient methods has emerged. These results show that the recovered solution enjoys some type of low norm property in the infinite space of zero-loss solutions. Gradient descent (and its variations) has therefore been analysed in various settings, the simplest and most emblematic being that of gradient descent for least-squares regression: it converges towards the solution which has the lowest \u21132subscript\u21132\\ell_{2}roman_\u2113 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT distance from the initialisation\u00a0(Lemaire, 1996). In the classification setting with linearly separable data, iterates of gradient methods must diverge to infinity to minimise the loss. Therefore, the directional convergence of the iterates is considered and Soudry et\u00a0al. (2018) show that gradient descent selects the \u21132subscript\u21132\\ell_{2}roman_\u2113 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT-max-margin solution amongst all classifiers.   Going beyond linear settings, it has been observed that an underlying mirror-descent structure very recurrently emerges when analysing gradient descent in a range of non-linear parametrisations\u00a0(Woodworth et\u00a0al., 2020; Azulay et\u00a0al., 2021). Providing convergence and implicit regularisation results for mirror descent has therefore gained significant importance.   In this context, for linear regression, Gunasekar et\u00a0al. (2018) show that the iterates converge to the solution that has minimal Bregman distance to the initial point. Turning towards the classification setting, an apparent gap emerges as there is still no clear understanding of what happens: Can directional convergence be characterised in terms of a max-margin problem? If so, what is the associated norm? Quite surprisingly, this question remains largely unanswered, as it is only understood for L\ud835\udc3fLitalic_L-homogeneous potentials\u00a0(Sun et\u00a0al., 2023). Our paper bridges this gap by formally characterising the implicit bias of mirror descent for separable classification problems.    1.1 Informal statement of the main result  For a separable dataset (xi,yi)i\u2208[n]subscriptsubscript\ud835\udc65\ud835\udc56subscript\ud835\udc66\ud835\udc56\ud835\udc56delimited-[]\ud835\udc5b(x_{i},y_{i})_{i\\in[n]}( italic_x start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT , italic_y start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) start_POSTSUBSCRIPT italic_i \u2208 [ italic_n ] end_POSTSUBSCRIPT, we study the mirror flow d\u2062\u2207\u03d5\u2062(\u03b2t)=\u2212\u2207L\u2062(\u03b2t)\u2062d\u2062td\u2207italic-\u03d5subscript\ud835\udefd\ud835\udc61\u2207\ud835\udc3fsubscript\ud835\udefd\ud835\udc61d\ud835\udc61\\mathrm{d}\\nabla\\phi(\\beta_{t})=-\\nabla L(\\beta_{t})\\mathrm{d}troman_d \u2207 italic_\u03d5 ( italic_\u03b2 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) = - \u2207 italic_L ( italic_\u03b2 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) roman_d italic_t with potential \u03d5:\u211dd\u2192\u211d:italic-\u03d5\u2192superscript\u211d\ud835\udc51\u211d\\phi:\\mathbb{R}^{d}\\to\\mathbb{R}italic_\u03d5 : blackboard_R start_POSTSUPERSCRIPT italic_d end_POSTSUPERSCRIPT \u2192 blackboard_R and an exponential tailed classification loss L\ud835\udc3fLitalic_L. We prove that \u03b2tsubscript\ud835\udefd\ud835\udc61\\beta_{t}italic_\u03b2 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT converges in direction towards the solution of the \u03d5\u221esubscriptitalic-\u03d5\\phi_{\\infty}italic_\u03d5 start_POSTSUBSCRIPT \u221e end_POSTSUBSCRIPT-maximum margin solution where the (asymmetric) norm \u03d5\u221esubscriptitalic-\u03d5\\phi_{\\infty}italic_\u03d5 start_POSTSUBSCRIPT \u221e end_POSTSUBSCRIPT captures the shape of the potential \u03d5italic-\u03d5\\phiitalic_\u03d5 \u2018at infinity\u2019 (see Figure 2 for an intuitive illustration).    Theorem 1 (Main result, Informal).   There exists a horizon function \u03d5\u221esubscriptitalic-\u03d5\\phi_{\\infty}italic_\u03d5 start_POSTSUBSCRIPT \u221e end_POSTSUBSCRIPT such that for any separable dataset, the normalised mirror flow iterates \u03b2\u00aft\u2254\u03b2t/\u2016\u03b2t\u2016\u2254subscript\u00af\ud835\udefd\ud835\udc61subscript\ud835\udefd\ud835\udc61normsubscript\ud835\udefd\ud835\udc61{\\bar{\\beta}}_{t}\\coloneqq\\beta_{t}/\\|\\beta_{t}\\|over\u00af start_ARG italic_\u03b2 end_ARG start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT \u2254 italic_\u03b2 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT / \u2225 italic_\u03b2 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT \u2225 converge and satisfy:    limt\u2192\u221e\u03b2\u00aftis",
            "references": [
                {
                    "title": "Leveraging Continuous Time to Understand Momentum When Training Diagonal Linear Networks",
                    "abstract": "In this work, we investigate the effect of momentum on the optimisation trajectory of gradient descent. We leverage a continuous-time approach in the analysis of momentum gradient descent with step size $\\gamma$ and momentum parameter $\\beta$ that allows us to identify an intrinsic quantity $\\lambda = \\frac{ \\gamma }{ (1 - \\beta)^2 }$ which uniquely defines the optimisation path and provides a simple acceleration rule. When training a $2$-layer diagonal linear network in an overparametrised regression setting, we characterise the recovered solution through an implicit regularisation problem. We then prove that small values of $\\lambda$ help to recover sparse solutions. Finally, we give similar but weaker results for stochastic momentum gradient descent. We provide numerical experiments which support our claims."
                },
                {
                    "title": "A Unified Approach to Controlling Implicit Regularization via Mirror Descent",
                    "abstract": "Inspired by the remarkable success of large neural networks, there has been significant interest in understanding the generalization performance of over-parameterized models. Substantial efforts have been invested in characterizing how optimization algorithms impact generalization through their\"preferred\"solutions, a phenomenon commonly referred to as implicit regularization. In particular, it has been argued that gradient descent (GD) induces an implicit $\\ell_2$-norm regularization in regression and classification problems. However, the implicit regularization of different algorithms are confined to either a specific geometry or a particular class of learning problems, indicating a gap in a general approach for controlling the implicit regularization. To address this, we present a unified approach using mirror descent (MD), a notable generalization of GD, to control implicit regularization in both regression and classification settings. More specifically, we show that MD with the general class of homogeneous potential functions converges in direction to a generalized maximum-margin solution for linear classification problems, thereby answering a long-standing question in the classification setting. Further, we show that MD can be implemented efficiently and enjoys fast convergence under suitable conditions. Through comprehensive experiments, we demonstrate that MD is a versatile method to produce learned models with different regularizers, which in turn have different generalization performances."
                },
                {
                    "title": "Saddle-to-Saddle Dynamics in Diagonal Linear Networks",
                    "abstract": "\n In this paper we fully describe the trajectory of gradient flow over two-layer diagonal linear networks for the regression setting in the limit of vanishing initialisation. We show that the limiting flow successively jumps from a saddle of the training loss to another until reaching the minimum \n \n \n \n \n \u2113\n 1\n \n \n \n -norm solution. We explicitly characterise the visited saddles as well as the jump times through a recursive algorithm reminiscent of the LARS algorithm used for computing the Lasso path. Starting from the zero vector, coordinates are successively activated until the minimum \n \n \n \n \n \u2113\n 1\n \n \n \n -norm solution is recovered, revealing an incremental learning. Our proof leverages a convenient arc-length time-reparametrisation which enables to keep track of the transitions between the jumps. Our analysis requires negligible assumptions on the data, applies to both under and overparametrised settings and covers complex cases where there is no monotonicity of the number of active coordinates. We provide numerical experiments to support our findings."
                },
                {
                    "title": "(S)GD over Diagonal Linear Networks: Implicit bias, Large Stepsizes and Edge of Stability",
                    "abstract": "In this paper, we investigate the impact of stochasticity and large stepsizes on the implicit regularisation of gradient descent (GD) and stochastic gradient descent (SGD) over diagonal linear networks. We prove the convergence of GD and SGD with macroscopic stepsizes in an overparametrised regression setting and characterise their solutions through an implicit regularisation problem. Our crisp characterisation leads to qualitative insights about the impact of stochasticity and stepsizes on the recovered solution. Specifically, we show that large stepsizes consistently benefit SGD for sparse regression problems, while they can hinder the recovery of sparse solutions for GD. These effects are magnified for stepsizes in a tight window just below the divergence threshold, in the\"edge of stability\"regime. Our findings are supported by experimental results."
                },
                {
                    "title": "Implicit Bias of Gradient Descent on Reparametrized Models: On Equivalence to Mirror Descent",
                    "abstract": "As part of the effort to understand implicit bias of gradient descent in overparametrized models, several results have shown how the training trajectory on the overparametrized model can be understood as mirror descent on a different objective. The main result here is a characterization of this phenomenon under a notion termed commuting parametrization, which encompasses all the previous results in this setting. It is shown that gradient flow with any commuting parametrization is equivalent to continuous mirror descent with a related Legendre function. Conversely, continuous mirror descent with any Legendre function can be viewed as gradient flow with a related commuting parametrization. The latter result relies upon Nash's embedding theorem."
                },
                {
                    "title": "Mirror Descent Maximizes Generalized Margin and Can Be Implemented Efficiently",
                    "abstract": "Driven by the empirical success and wide use of deep neural networks, understanding the generalization performance of overparameterized models has become an increasingly popular question. To this end, there has been substantial effort to characterize the implicit bias of the optimization algorithms used, such as gradient descent (GD), and the structural properties of their preferred solutions. This paper answers an open question in this literature: For the classification setting, what solution does mirror descent (MD) converge to? Specifically, motivated by its efficient implementation, we consider the family of mirror descent algorithms with potential function chosen as the $p$-th power of the $\\ell_p$-norm, which is an important generalization of GD. We call this algorithm $p$-$\\textsf{GD}$. For this family, we characterize the solutions it obtains and show that it converges in direction to a generalized maximum-margin solution with respect to the $\\ell_p$-norm for linearly separable classification. While the MD update rule is in general expensive to compute and perhaps not suitable for deep learning, $p$-$\\textsf{GD}$ is fully parallelizable in the same manner as SGD and can be used to train deep neural networks with virtually no additional computational overhead. Using comprehensive experiments with both linear and deep neural network models, we demonstrate that $p$-$\\textsf{GD}$ can noticeably affect the structure and the generalization performance of the learned models."
                },
                {
                    "title": "Convex Analysis at Infinity: An Introduction to Astral Space",
                    "abstract": "Not all convex functions on $\\mathbb{R}^n$ have finite minimizers; some can only be minimized by a sequence as it heads to infinity. In this work, we aim to develop a theory for understanding such minimizers at infinity. We study astral space, a compact extension of $\\mathbb{R}^n$ to which such points at infinity have been added. Astral space is constructed to be as small as possible while still ensuring that all linear functions can be continuously extended to the new space. Although astral space includes all of $\\mathbb{R}^n$, it is not a vector space, nor even a metric space. However, it is sufficiently well-structured to allow useful and meaningful extensions of concepts of convexity, conjugacy, and subdifferentials. We develop these concepts and analyze various properties of convex functions on astral space, including the detailed structure of their minimizers, exact characterizations of continuity, and convergence of descent algorithms."
                },
                {
                    "title": "Implicit Bias of SGD for Diagonal Linear Networks: a Provable Benefit of Stochasticity",
                    "abstract": "Understanding the implicit bias of training algorithms is of crucial importance in order to explain the success of overparametrised neural networks. In this paper, we study the dynamics of stochastic gradient descent over diagonal linear networks through its continuous time version, namely stochastic gradient flow. We explicitly characterise the solution chosen by the stochastic flow and prove that it always enjoys better generalisation properties than that of gradient flow. Quite surprisingly, we show that the convergence speed of the training loss controls the magnitude of the biasing effect: the slower the convergence, the better the bias. To fully complete our analysis, we provide convergence guarantees for the dynamics. We also give experimental results which support our theoretical claims. Our findings highlight the fact that structured noise can induce better generalisation and they help explain the greater performances observed in practice of stochastic gradient descent over gradient descent."
                },
                {
                    "title": "Implicit Regularization in Matrix Sensing via Mirror Descent",
                    "abstract": "We study discrete-time mirror descent applied to the unregularized empirical risk in matrix sensing. In both the general case of rectangular matrices and the particular case of positive semidefinite matrices, a simple potential-based analysis in terms of the Bregman divergence allows us to establish convergence of mirror descent -- with different choices of the mirror maps -- to a matrix that, among all global minimizers of the empirical risk, minimizes a quantity explicitly related to the nuclear norm, the Frobenius norm, and the von Neumann entropy. In both cases, this characterization implies that mirror descent, a first-order algorithm minimizing the unregularized empirical risk, recovers low-rank matrices under the same set of assumptions that are sufficient to guarantee recovery for nuclear-norm minimization. When the sensing matrices are symmetric and commute, we show that gradient descent with full-rank factorized parametrization is a first-order approximation to mirror descent, in which case we obtain an explicit characterization of the implicit bias of gradient flow as a by-product."
                },
                {
                    "title": "On the Implicit Bias of Initialization Shape: Beyond Infinitesimal Mirror Descent",
                    "abstract": "Recent work has highlighted the role of initialization scale in determining the structure of the solutions that gradient methods converge to. In particular, it was shown that large initialization leads to the neural tangent kernel regime solution, whereas small initialization leads to so called\"rich regimes\". However, the initialization structure is richer than the overall scale alone and involves relative magnitudes of different weights and layers in the network. Here we show that these relative scales, which we refer to as initialization shape, play an important role in determining the learned model. We develop a novel technique for deriving the inductive bias of gradient-flow and use it to obtain closed-form implicit regularizers for multiple cases of interest."
                },
                {
                    "title": "A Unifying View on Implicit Bias in Training Linear Neural Networks",
                    "abstract": "We study the implicit bias of gradient flow (i.e., gradient descent with infinitesimal step size) on linear neural network training. We propose a tensor formulation of neural networks that includes fully-connected, diagonal, and convolutional networks as special cases, and investigate the linear version of the formulation called linear tensor networks. For $L$-layer linear tensor networks that are orthogonally decomposable, we show that gradient flow on separable classification finds a stationary point of the $\\ell_{2/L}$ max-margin problem in a \"transformed\" input space defined by the network. For underdetermined regression, we prove that gradient flow finds a global minimum which minimizes a norm-like function that interpolates between weighted $\\ell_1$ and $\\ell_2$ norms in the transformed input space. Our theorems subsume existing results in the literature while removing most of the convergence assumptions. We also provide experiments that corroborate our analysis."
                },
                {
                    "title": "Winnowing with Gradient Descent",
                    "abstract": "We rewrite the non-negative weights wi of multiplicative updates by \u03c1 2 i and replace the multiplicative updates by a gradient descent step w.r.t. the new parameter \u03c1i. We do this for the Winnow update, the Hedge update, and the unnnormalized and normalized exponentiated gradient (EG) updates for linear regression. When the original weights wi are scaled to sum to one (as done for Hedge and normalized EG), then in the corresponding reparameterized update, the \u03c1i parameters are now divided by 1/\u2016\u03c1\u20162 after the gradient descent step. We show that the reparameterizations closely track the original multiplicative updates by proving the same online regret bounds in each case (albeit in some cases with slightly worse constants). In particular, our work implies that the typical logarithmic dependence in the number of features of the multiplicative updates for sparse targets can also be achieved with the reparameterized gradient descent updates. As a side, our work exhibits a simple linear 2-layer neural net that when trained with gradient descent can provable solve a certain sparse linear problem (known has the Hadamard problem) with exponentially fewer examples than any kernel method. The talk will stay at a high level. In particular we will discuss the implication of these results on the power of neural nets versus kernel methods and on the influence of the training algorithm for neural nets. We conclude with a number of fundamental open problems. Most of the presented work work was done jointly with Ehsan Amid."
                },
                {
                    "title": "Implicit Bias in Deep Linear Classification: Initialization Scale vs Training Accuracy",
                    "abstract": "We provide a detailed asymptotic study of gradient flow trajectories and their implicit optimization bias when minimizing the exponential loss over \"diagonal linear networks\". This is the simplest model displaying a transition between \"kernel\" and non-kernel (\"rich\" or \"active\") regimes. We show how the transition is controlled by the relationship between the initialization scale and how accurately we minimize the training loss. Our results indicate that some limit behaviors of gradient descent only kick in at ridiculous training accuracies (well beyond $10^{-100}$). Moreover, the implicit bias at reasonable initialization scales and training accuracies is more complex and not captured by these limits."
                },
                {
                    "title": "Gradient descent follows the regularization path for general losses",
                    "abstract": "Recent work across many machine learning disciplines has highlighted that standard descent methods, even without explicit regularization, do not merely minimize the training error, but also exhibit an implicit bias. This bias is typically towards a certain regularized solution, and relies upon the details of the learning process, for instance the use of the cross-entropy loss. In this work, we show that for empirical risk minimization over linear predictors with arbitrary convex, strictly decreasing losses, if the risk does not attain its infimum, then the gradient-descent path and the algorithm-independent regularization path converge to the same direction (whenever either converges to a direction). Using this result, we provide a justification for the widely-used exponentially-tailed losses (such as the exponential loss or the logistic loss): while this convergence to a direction for exponentially-tailed losses is necessarily to the maximum-margin direction, other losses such as polynomially-tailed losses may induce convergence to a direction with a poor margin."
                },
                {
                    "title": "Reparameterizing Mirror Descent as Gradient Descent",
                    "abstract": "Most of the recent successful applications of neural networks have been based on training with gradient descent updates. However, for some small networks, other mirror descent updates learn provably more efficiently when the target is sparse. We present a general framework for casting a mirror descent update as a gradient descent update on a different set of parameters. In some cases, the mirror descent reparameterization can be described as training a modified network with standard backpropagation. The reparameterization framework is versatile and covers a wide range of mirror descent updates, even cases where the domain is constrained. Our construction for the reparameterization argument is done for the continuous versions of the updates. Finding general criteria for the discrete versions to closely track their continuous counterparts remains an interesting open problem."
                },
                {
                    "title": "Implicit Bias of Gradient Descent for Wide Two-layer Neural Networks Trained with the Logistic Loss",
                    "abstract": "Neural networks trained to minimize the logistic (a.k.a. cross-entropy) loss with gradient-based methods are observed to perform well in many supervised classification tasks. Towards understanding this phenomenon, we analyze the training and generalization behavior of infinitely wide two-layer neural networks with homogeneous activations. We show that the limits of the gradient flow on exponentially tailed losses can be fully characterized as a max-margin classifier in a certain non-Hilbertian space of functions. In presence of hidden low-dimensional structures, the resulting margin is independent of the ambiant dimension, which leads to strong generalization bounds. In contrast, training only the output layer implicitly solves a kernel support vector machine, which a priori does not enjoy such an adaptivity. Our analysis of training is non-quantitative in terms of running time but we prove computational guarantees in simplified settings by showing equivalences with online mirror descent. Finally, numerical experiments suggest that our analysis describes well the practical behavior of two-layer neural networks with ReLU activation and confirm the statistical benefits of this implicit bias."
                },
                {
                    "title": "Implicit Regularization for Optimal Sparse Recovery",
                    "abstract": "We investigate implicit regularization schemes for gradient descent methods applied to unpenalized least squares regression to solve the problem of reconstructing a sparse signal from an underdetermined system of linear measurements under the restricted isometry assumption. For a given parametrization yielding a non-convex optimization problem, we show that prescribed choices of initialization, step size and stopping time yield a statistically and computationally optimal algorithm that achieves the minimax rate with the same cost required to read the data up to poly-logarithmic factors. Beyond minimax optimality, we show that our algorithm adapts to instance difficulty and yields a dimension-independent rate when the signal-to-noise ratio is high enough. Key to the computational efficiency of our method is an increasing step size scheme that adapts to refined estimates of the true solution. We validate our findings with numerical experiments and compare our algorithm against explicit $\\ell_{1}$ penalization. Going from hard instances to easy ones, our algorithm is seen to undergo a phase transition, eventually matching least squares with an oracle knowledge of the true support."
                },
                {
                    "title": "Kernel and Rich Regimes in Overparametrized Models",
                    "abstract": "A recent line of work studies overparametrized neural networks in the \"kernel regime,\" i.e. when the network behaves during training as a kernelized linear predictor, and thus training with gradient descent has the effect of finding the minimum RKHS norm solution. This stands in contrast to other studies which demonstrate how gradient descent on overparametrized multilayer networks can induce rich implicit biases that are not RKHS norms. Building on an observation by Chizat and Bach, we show how the scale of the initialization controls the transition between the \"kernel\" (aka lazy) and \"rich\" (aka active) regimes and affects generalization properties in multilayer homogeneous models. We also highlight an interesting role for the width of a model in the case that the predictor is not identically zero at initialization. We provide a complete and detailed analysis for a family of simple depth-$D$ models that already exhibit an interesting and meaningful transition between the kernel and rich regimes, and we also demonstrate this transition empirically for more complex matrix factorization models and multilayer non-linear networks."
                },
                {
                    "title": "Gradient Descent Maximizes the Margin of Homogeneous Neural Networks",
                    "abstract": "In this paper, we study the implicit regularization of the gradient descent algorithm in homogeneous neural networks, including fully-connected and convolutional neural networks with ReLU or LeakyReLU activations. In particular, we study the gradient descent or gradient flow (i.e., gradient descent with infinitesimal step size) optimizing the logistic loss or cross-entropy loss of any homogeneous model (possibly non-smooth), and show that if the training loss decreases below a certain threshold, then we can define a smoothed version of the normalized margin which increases over time. We also formulate a natural constrained optimization problem related to margin maximization, and prove that both the normalized margin and its smoothed version converge to the objective value at a KKT point of the optimization problem. Our results generalize the previous results for logistic regression with one-layer or multi-layer linear networks, and provide more quantitative convergence results with weaker assumptions than previous results for homogeneous smooth neural networks. We conduct several experiments to justify our theoretical finding on MNIST and CIFAR-10 datasets. Finally, as margin is closely related to robustness, we discuss potential benefits of training longer for improving the robustness of the model."
                },
                {
                    "title": "Characterizing the implicit bias via a primal-dual analysis",
                    "abstract": "This paper shows that the implicit bias of gradient descent on linearly separable data is exactly characterized by the optimal solution of a dual optimization problem given by a smoothed margin, even for general losses. This is in contrast to prior results, which are often tailored to exponentially-tailed losses. For the exponential loss specifically, with $n$ training examples and $t$ gradient descent steps, our dual analysis further allows us to prove an $O(\\ln(n)/\\ln(t))$ convergence rate to the $\\ell_2$ maximum margin direction, when a constant step size is used. This rate is tight in both $n$ and $t$, which has not been presented by prior work. On the other hand, with a properly chosen but aggressive step size schedule, we prove $O(1/t)$ rates for both $\\ell_2$ margin maximization and implicit bias, whereas prior work (including all first-order methods for the general hard-margin linear SVM problem) proved $\\widetilde{O}(1/\\sqrt{t})$ margin rates, or $O(1/t)$ margin rates to a suboptimal margin, with an implied (slower) bias rate. Our key observations include that gradient descent on the primal variable naturally induces a mirror descent update on the dual variable, and that the dual objective in this setting is smooth enough to give a faster rate."
                },
                {
                    "title": "Bad Global Minima Exist and SGD Can Reach Them",
                    "abstract": "Several recent works have aimed to explain why severely overparameterized models, generalize well when trained by Stochastic Gradient Descent (SGD). The emergent consensus explanation has two parts: the first is that there are \"no bad local minima\", while the second is that SGD performs implicit regularization by having a bias towards low complexity models. We revisit both of these ideas in the context of image classification with common deep neural network architectures. Our first finding is that there exist bad global minima, i.e., models that fit the training set perfectly, yet have poor generalization. Our second finding is that given only unlabeled training data, we can easily construct initializations that will cause SGD to quickly converge to such bad global minima. For example, on CIFAR, CINIC10, and (Restricted) ImageNet, this can be achieved by starting SGD at a model derived by fitting random labels on the training data: while subsequent SGD training (with the correct labels) will reach zero training error, the resulting model will exhibit a test accuracy degradation of up to 40% compared to training from a random initialization. Finally, we show that regularization seems to provide SGD with an escape route: once heuristics such as data augmentation are used, starting from a complex model (adversarial initialization) has no effect on the test accuracy."
                },
                {
                    "title": "Lexicographic and Depth-Sensitive Margins in Homogeneous and Non-Homogeneous Deep Models",
                    "abstract": "With an eye toward understanding complexity control in deep learning, we study how infinitesimal regularization or gradient descent optimization lead to margin maximizing solutions in both homogeneous and non-homogeneous models, extending previous work that focused on infinitesimal regularization only in homogeneous models. To this end we study the limit of loss minimization with a diverging norm constraint (the \"constrained path\"), relate it to the limit of a \"margin path\" and characterize the resulting solution. For non-homogeneous ensemble models, which output is a sum of homogeneous sub-models, we show that this solution discards the shallowest sub-models if they are unnecessary. For homogeneous models, we show convergence to a \"lexicographic max-margin solution\", and provide conditions under which max-margin solutions are also attained as the limit of unconstrained gradient descent."
                },
                {
                    "title": "Stochastic Gradient Descent on Separable Data: Exact Convergence with a Fixed Learning Rate",
                    "abstract": "Stochastic Gradient Descent (SGD) is a central tool in machine learning. We prove that SGD converges to zero loss, even with a fixed (non-vanishing) learning rate --- in the special case of homogeneous linear classifiers with smooth monotone loss functions, optimized on linearly separable data. Previous works assumed either a vanishing learning rate, iterate averaging, or loss assumptions that do not hold for monotone loss functions used for classification, such as the logistic loss. We prove our result on a fixed dataset, both for sampling with or without replacement. Furthermore, for logistic loss (and similar exponentially-tailed losses), we prove that with SGD the weight vector converges in direction to the $L_2$ max margin vector as $O(1/\\log(t))$ for almost all separable datasets, and the loss converges as $O(1/t)$ --- similarly to gradient descent. Lastly, we examine the case of a fixed learning rate proportional to the minibatch size. We prove that in this case, the asymptotic convergence rate of SGD (with replacement) does not depend on the minibatch size in terms of epochs, if the support vectors span the data. These results may suggest an explanation to similar behaviors observed in deep networks, when trained with SGD."
                },
                {
                    "title": "Risk and parameter convergence of logistic regression",
                    "abstract": "Gradient descent, when applied to the task of logistic regression, outputs iterates which are biased to follow a unique ray defined by the data. The direction of this ray is the maximum margin predictor of a maximal linearly separable subset of the data; the gradient descent iterates converge to this ray in direction at the rate $\\mathcal{O}(\\ln\\ln t / \\ln t)$. The ray does not pass through the origin in general, and its offset is the bounded global optimum of the risk over the remaining data; gradient descent recovers this offset at a rate $\\mathcal{O}((\\ln t)^2 / \\sqrt{t})$."
                },
                {
                    "title": "Convergence of Gradient Descent on Separable Data",
                    "abstract": "We provide a detailed study on the implicit bias of gradient descent when optimizing loss functions with strictly monotone tails, such as the logistic loss, over separable datasets. We look at two basic questions: (a) what are the conditions on the tail of the loss function under which gradient descent converges in the direction of the $L_2$ maximum-margin separator? (b) how does the rate of margin convergence depend on the tail of the loss function and the choice of the step size? We show that for a large family of super-polynomial tailed losses, gradient descent iterates on linear networks of any depth converge in the direction of $L_2$ maximum-margin solution, while this does not hold for losses with heavier tails. Within this family, for simple linear models we show that the optimal rates with fixed step size is indeed obtained for the commonly used exponentially tailed losses such as logistic loss. However, with a fixed step size the optimal convergence rate is extremely slow as $1/\\log(t)$, as also proved in Soudry et al. (2018). For linear models with exponential loss, we further prove that the convergence rate could be improved to $\\log (t) /\\sqrt{t}$ by using aggressive step sizes that compensates for the rapidly vanishing gradients. Numerical results suggest this method might be useful for deep networks."
                },
                {
                    "title": "Characterizing Implicit Bias in Terms of Optimization Geometry",
                    "abstract": "We study the implicit bias of generic optimization methods, such as mirror descent, natural gradient descent, and steepest descent with respect to different potentials and norms, when optimizing underdetermined linear regression or separable linear classification problems. We explore the question of whether the specific global minimum (among the many possible global minima) reached by an algorithm can be characterized in terms of the potential or norm of the optimization geometry, and independently of hyperparameter choices such as step-size and momentum."
                },
                {
                    "title": "The Implicit Bias of Gradient Descent on Separable Data",
                    "abstract": "We examine gradient descent on unregularized logistic regression problems, with homogeneous linear predictors on linearly separable datasets. We show the predictor converges to the direction of the max-margin (hard margin SVM) solution. The result also generalizes to other monotone decreasing loss functions with an infimum at infinity, to multi-class problems, and to training a weight layer in a deep network in a certain restricted setting. Furthermore, we show this convergence is very slow, and only logarithmic in the convergence of the loss itself. This can help explain the benefit of continuing to optimize the logistic or cross-entropy loss even after the training error is zero and the training loss is extremely small, and, as we show, even if the validation loss increases. Our methodology can also aid in understanding implicit regularization n more complex models and with other optimization methods."
                },
                {
                    "title": "Implicit Regularization in Matrix Factorization",
                    "abstract": "We study implicit regularization when optimizing an underdetermined quadratic objective over a matrix $X$ with gradient descent on a factorization of X. We conjecture and provide empirical and theoretical evidence that with small enough step sizes and initialization close enough to the origin, gradient descent on a full dimensional factorization converges to the minimum nuclear norm solution."
                },
                {
                    "title": "A Descent Lemma Beyond Lipschitz Gradient Continuity: First-Order Methods Revisited and Applications",
                    "abstract": "The proximal gradient and its variants is one of the most attractive first-order algorithm for minimizing the sum of two convex functions, with one being nonsmooth. However, it requires the differentiable part of the objective to have a Lipschitz continuous gradient, thus precluding its use in many applications. In this paper we introduce a framework which allows to circumvent the intricate question of Lipschitz continuity of gradients by using an elegant and easy to check convexity condition which captures the geometry of the constraints. This condition translates into a new descent lemma which in turn leads to a natural derivation of the proximal-gradient scheme with Bregman distances. We then identify a new notion of asymmetry measure for Bregman distances, which is central in determining the relevant step-size. These novelties allow to prove a global sublinear rate of convergence, and as a by-product, global pointwise convergence is obtained. This provides a new path to a broad spectrum of problems arising in key applications which were, until now, considered as out of reach via proximal gradient methods. We illustrate this potential by showing how our results can be applied to build new and simple schemes for Poisson inverse problems."
                },
                {
                    "title": "Understanding deep learning requires rethinking generalization",
                    "abstract": "Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family, or to the regularization techniques used during training. \nThrough extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization, and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. \nWe interpret our experimental findings by comparison with traditional models."
                },
                {
                    "title": "THE HEAVY BALL WITH FRICTION METHOD, I. THE CONTINUOUS DYNAMICAL SYSTEM: GLOBAL EXPLORATION OF THE LOCAL MINIMA OF A REAL-VALUED FUNCTION BY ASYMPTOTIC ANALYSIS OF A DISSIPATIVE DYNAMICAL SYSTEM",
                    "abstract": "Let H be a real Hilbert space and \u03a6:H \u21a6 R a continuously differentiable function, whose gradient is Lipschitz continuous on bounded sets. We study the nonlinear dissipative dynamical system: ${\\ddot x}(t)+\\lambda{\\dot x}(t)+\\nabla\\Phi(x(t))=0, \\lambda>0, plus Cauchy data, mainly in view of the unconstrained minimization of the function \u03a6. New results concerning the convergence of a solution to a critical point are given in various situations, including when \u03a6 is convex (possibly with multiple minima) or is a Morse function (the critical point being then generically a local minimum); a counterexample shows that, without peculiar assumptions, a trajectory may not converge. By following the trajectories, we obtain a method for exploring local minima of \u03a6. A singular perturbation analysis links our results with those concerning gradient systems."
                },
                {
                    "title": "Tame Topology and O-minimal Structures",
                    "abstract": "1. Some elementary results 2. Semialgebraic sets 3. Cell decomposition 4. Definable invariants: Dimension and Euler characteristic 5. The Vapnik-Chernovenkis property in o-minimal structures 6. Point-set topology in o-minimal structures 7. Smoothness 8. Triangulation 9. Trivialization 10. Definable spaces and quotients."
                },
                {
                    "title": "Geometric categories and o-minimal structures",
                    "abstract": "The theory of subanalytic sets is an excellent tool in various analytic-geometric contexts; see, for example, Bierstone and Milman [1]. Regrettably, certain \u201cnice\u201d sets\u2014like { (x, x) : x > 0 } for positive irrational r, and { (x, e\u22121/x) : x > 0 }\u2014are not subanalytic (at the origin) in R. Here we make available an extension of the category of subanalytic sets that has these sets among its objects, and that behaves much like the category of subanalytic sets. The possibility of doing this emerged in 1991 when Wilkie [27] proved that the real exponential field is \u201cmodel complete\u201d, followed soon by work of Ressayre, Macintyre, Marker and the authors; see [21], [5], [7] and [19]. However, there are two obstructions to the use by geometers of this development: (i) while the proofs in these articles make essential use of model theory, many results are also stated there (efficiently, but unnecessarily) in model-theoretic terms; (ii) the results of these papers apply directly only to the cartesian spaces R, and not to arbitrary real analytic manifolds. Consequently, in order to carry out our goal we recast here some results in those papers\u2014as well as many of their consequences\u2014in more familiar terms, with emphasis on results of a geometric nature, and allowing arbitrary (real analytic) manifolds as ambient spaces. We thank W. Schmid and K. Vilonen for their suggestion that this would be a useful undertaking; indeed, they gave us a \u201cwish list\u201d (inspired by Chapters 8 and 9 of Kashiwara and Schapira [12]; see also \u00a710 of [22]) which strongly influenced the form and content of this paper. We axiomatize in Section 1 the notion of \u201cbehaving like the category of subanalytic sets\u201d by introducing the notion of \u201canalytic-geometric category\u201d. (The category Can of subanalytic sets is the \u201csmallest\u201d analytic-geometric category.) We also state in Section 1 a number of properties shared by all analytic-geometric categories. Proofs of the more difficult results of this nature, like the Whitney-stratifiability of sets and maps in such a category, often involve the use of charts to reduce to the case of subsets of R. For subsets of R, there already exists the theory of \u201co-minimal structures on the real field\u201d (defined in Section 2); this subject is developed in detail in [4] and is an abstraction of"
                },
                {
                    "title": "On the spectral bias of two-layer linear networks",
                    "abstract": "This paper studies the behaviour of two-layer fully connected networks with linear activations trained with gradient flow on the square loss. We show how the optimization process carries an implicit bias on the parameters that depends on the scale of its initialization. The main result of the paper is a variational characterization of the loss minimizers retrieved by the gradient flow for a specific initialization shape. This characterization reveals that, in the small scale initialization regime, the linear neural network\u2019s hidden layer is biased toward having a low-rank structure. To complement our results, we showcase a hidden mirror flow that tracks the dynamics of the singular values of the weights matrices and describe their time evolution. We support our findings with numerical experiments illustrating the phenomena."
                },
                {
                    "title": "A short geometric proof that Hausdorff limits are definable in any o-minimal structure",
                    "abstract": "The aim of this note is to give yet another proof of the following theorem: given an arbitrary o-minimal structure on the ordered field of real numbers R and any definable family A of definable nonempty compact subsets of R, then the closure of A in the sense of the Hausdorff metric (or, equivalently, in the Vietoris topology) is a definable family. In particular, any limit in the sense of the Hausdorff metric of a convergent sequence of subsets of a definable family is definable in the same o-minimal structure. The original proofs by Brocker [1], Marker and Steinhorn [7], Pillay [11] (see also van den Dries [15]) were based on model theory. Lion and Speissegger [6] gave a geometric proof of the theorem. Our proof below is based on the idea of Lipschitz cell decompositions."
                },
                {
                    "title": "On the graph convergence of subdifferentials of convex functions",
                    "abstract": "This paper provides another proof of the Attouch Theorem relating the epigraphical limit of sequences of convex functions to the set limit of the graphs of the subdifferentials."
                },
                {
                    "title": "An asymptotical variational principle associated with the steepest descent method for a convex function.",
                    "abstract": "Let X be a real Hilbert space endowed with inner product \u3008., .\u3009 and associated norm \u2016.\u2016, and let f be a proper closed convex function on X. The paper considers the problem of minimizing f , that is, of finding infX f and some element in the optimal set S := Argmin f , this set assumed being non empty. Letting \u2202f denote the subdifferential operator associated with f , we focus on the continuous steepest descent method associated with f , i.e., the differential inclusion"
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG",
                "math.OC"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan we formally characterize the implicit bias of mirror descent in separable classification problems?\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding the implicit bias of mirror descent in separable classification problems is crucial for advancing the theoretical foundations of machine learning. By addressing this question, we can provide insights into how optimization algorithms influence generalization, which could lead to the development of more effective training methods. This research could also inspire future studies on implicit regularization in non-linear settings, ultimately enhancing the performance of machine learning models in practical applications.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of characterizing the convergence behavior of mirror descent in non-linear settings. Naive approaches may fail because they do not account for the intricate relationships between the optimization process and the geometry of the solution space. Additionally, the lack of a clear understanding of the max-margin problem in the context of mirror descent adds a layer of theoretical complexity that must be navigated.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on linear settings or specific types of potentials, leaving a gap in understanding the behavior of mirror descent in more general cases. Barriers such as the lack of appropriate mathematical tools to analyze non-linear dynamics and the complexity of the underlying optimization landscape have hindered progress. Our approach differs by providing a formal characterization of the implicit bias in separable classification, addressing the previously unexplored aspects of mirror descent.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves studying the mirror flow dynamics defined by the equation \\( d\\nabla\\phi(\\beta_t) = -\\nabla L(\\beta_t) dt \\) with an exponential-tailed classification loss \\( L \\). We will analyze the convergence of the iterates \\( \\beta_t \\) towards the \\( \\phi_{\\infty} \\)-maximum margin solution for separable datasets. The expected outcome is a formal characterization of the implicit bias of mirror descent, providing insights into the asymptotic behavior of the optimization process and its implications for generalization in classification tasks."
            }
        },
        "author_data": {
            "15b438a6-bd21-45cd-9ad6-b767fbccddc0": {
                "pk": "15b438a6-bd21-45cd-9ad6-b767fbccddc0",
                "name": "Scott Pesme",
                "collaborators": [
                    "Nicolas Flammarion",
                    "Aymeric Dieuleveut",
                    "Loucas Pillaud-Vivien",
                    "Mathieu Even",
                    "Suriya Gunasekar",
                    "Hristo Papazov"
                ],
                "domain": [
                    "Robust Regression",
                    "Stochastic Optimization",
                    "Implicit Regularization",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Online Robust Regression via SGD on the l1 loss",
                        "abstract": "We consider the robust linear regression problem in the online setting where we have access to the data in a streaming manner, one data point after the other. More specifically, for a true parameter $\\theta^*$, we consider the corrupted Gaussian linear model $y = \\langle x , \\ \\theta^* \\rangle + \\varepsilon + b$ where the adversarial noise $b$ can take any value with probability $\\eta$ and equals zero otherwise. We consider this adversary to be oblivious (i.e., $b$ independent of the data) since this is the only contamination model under which consistency is possible. Current algorithms rely on having the whole data at hand in order to identify and remove the outliers. In contrast, we show in this work that stochastic gradient descent on the $\\ell_1$ loss converges to the true parameter vector at a $\\tilde{O}( 1 / (1 - \\eta)^2 n )$ rate which is independent of the values of the contaminated measurements. Our proof relies on the elegant smoothing of the non-smooth $\\ell_1$ loss by the Gaussian data and a classical non-asymptotic analysis of Polyak-Ruppert averaged SGD. In addition, we provide experimental evidence of the efficiency of this simple and highly scalable algorithm."
                    },
                    {
                        "title": "Saddle-to-Saddle Dynamics in Diagonal Linear Networks",
                        "abstract": "In this paper we fully describe the trajectory of gradient flow over diagonal linear networks in the limit of vanishing initialisation. We show that the limiting flow successively jumps from a saddle of the training loss to another until reaching the minimum $\\ell_1$-norm solution. This saddle-to-saddle dynamics translates to an incremental learning process as each saddle corresponds to the minimiser of the loss constrained to an active set outside of which the coordinates must be zero. We explicitly characterise the visited saddles as well as the jumping times through a recursive algorithm reminiscent of the LARS algorithm used for computing the Lasso path. Our proof leverages a convenient arc-length time-reparametrisation which enables to keep track of the heteroclinic transitions between the jumps. Our analysis requires negligible assumptions on the data, applies to both under and overparametrised settings and covers complex cases where there is no monotonicity of the number of active coordinates. We provide numerical experiments to support our findings."
                    },
                    {
                        "title": "On Convergence-Diagnostic based Step Sizes for Stochastic Gradient Descent",
                        "abstract": "Constant step-size Stochastic Gradient Descent exhibits two phases: a transient phase during which iterates make fast progress towards the optimum, followed by a stationary phase during which iterates oscillate around the optimal point. In this paper, we show that efficiently detecting this transition and appropriately decreasing the step size can lead to fast convergence rates. We analyse the classical statistical test proposed by Pflug (1983), based on the inner product between consecutive stochastic gradients. Even in the simple case where the objective function is quadratic we show that this test cannot lead to an adequate convergence diagnostic. We then propose a novel and simple statistical procedure that accurately detects stationarity and we provide experimental results showing state-of-the-art performance on synthetic and real-world datasets."
                    },
                    {
                        "title": "Implicit Bias of SGD for Diagonal Linear Networks: a Provable Benefit of Stochasticity",
                        "abstract": "Understanding the implicit bias of training algorithms is of crucial importance in order to explain the success of overparametrised neural networks. In this paper, we study the dynamics of stochastic gradient descent over diagonal linear networks through its continuous time version, namely stochastic gradient flow. We explicitly characterise the solution chosen by the stochastic flow and prove that it always enjoys better generalisation properties than that of gradient flow. Quite surprisingly, we show that the convergence speed of the training loss controls the magnitude of the biasing effect: the slower the convergence, the better the bias. To fully complete our analysis, we provide convergence guarantees for the dynamics. We also give experimental results which support our theoretical claims. Our findings highlight the fact that structured noise can induce better generalisation and they help explain the greater performances observed in practice of stochastic gradient descent over gradient descent."
                    },
                    {
                        "title": "(S)GD over Diagonal Linear Networks: Implicit Regularisation, Large Stepsizes and Edge of Stability",
                        "abstract": "In this paper, we investigate the impact of stochasticity and large stepsizes on the implicit regularisation of gradient descent (GD) and stochastic gradient descent (SGD) over diagonal linear networks. We prove the convergence of GD and SGD with macroscopic stepsizes in an overparametrised regression setting and characterise their solutions through an implicit regularisation problem. Our crisp characterisation leads to qualitative insights about the impact of stochasticity and stepsizes on the recovered solution. Specifically, we show that large stepsizes consistently benefit SGD for sparse regression problems, while they can hinder the recovery of sparse solutions for GD. These effects are magnified for stepsizes in a tight window just below the divergence threshold, in the \"edge of stability\" regime. Our findings are supported by experimental results."
                    },
                    {
                        "title": "Leveraging Continuous Time to Understand Momentum When Training Diagonal Linear Networks",
                        "abstract": "In this work, we investigate the effect of momentum on the optimisation trajectory of gradient descent. We leverage a continuous-time approach in the analysis of momentum gradient descent with step size $\\gamma$ and momentum parameter $\\beta$ that allows us to identify an intrinsic quantity $\\lambda = \\frac{ \\gamma }{ (1 - \\beta)^2 }$ which uniquely defines the optimisation path and provides a simple acceleration rule. When training a $2$-layer diagonal linear network in an overparametrised regression setting, we characterise the recovered solution through an implicit regularisation problem. We then prove that small values of $\\lambda$ help to recover sparse solutions. Finally, we give similar but weaker results for stochastic momentum gradient descent. We provide numerical experiments which support our claims."
                    }
                ]
            },
            "a4b9452a-159b-4f47-aba8-1bfd6cf02603": {
                "pk": "a4b9452a-159b-4f47-aba8-1bfd6cf02603",
                "name": "Radu-Alexandru Dragomir",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "30830446-20f4-4af6-8c9b-2b94900a14d3": {
                "pk": "30830446-20f4-4af6-8c9b-2b94900a14d3",
                "name": "Nicolas Flammarion",
                "collaborators": [
                    "Etienne Boursier",
                    "Francis Bach",
                    "Scott Pesme",
                    "Maksym Andriushchenko",
                    "Aditya Varre",
                    "Aymeric Dieuleveut",
                    "Yeshwanth Cherapanamjeri",
                    "Peter L. Bartlett",
                    "Mikhail Konobeev",
                    "O\u011fuz Kaan Y\u00fcksel"
                ],
                "domain": [
                    "Optimization",
                    "Machine Learning",
                    "Stochastic Methods",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Stochastic Composite Least-Squares Regression with convergence rate O(1/n)",
                        "abstract": "We consider the minimization of composite objective functions composed of the expectation of quadratic functions and an arbitrary convex function. We study the stochastic dual averaging algorithm with a constant step-size, showing that it leads to a convergence rate of O(1/n) without strong convexity assumptions. This thus extends earlier results on least-squares regression with the Euclidean geometry to (a) all convex regularizers and constraints, and (b) all geome-tries represented by a Bregman divergence. This is achieved by a new proof technique that relates stochastic and deterministic recursions."
                    },
                    {
                        "title": "Accelerated SGD for Non-Strongly-Convex Least Squares",
                        "abstract": "We consider stochastic approximation for the least squares regression problem in the non-strongly convex setting. We present the first practical algorithm that achieves the optimal prediction error rates in terms of dependence on the noise of the problem, as $O(d/t)$ while accelerating the forgetting of the initial conditions to $O(d/t^2)$. Our new algorithm is based on a simple modification of the accelerated gradient descent. We provide convergence results for both the averaged and the last iterate of the algorithm. In order to describe the tightness of these new bounds, we present a matching lower bound in the noiseless setting and thus show the optimality of our algorithm."
                    },
                    {
                        "title": "Online Robust Regression via SGD on the l1 loss",
                        "abstract": "We consider the robust linear regression problem in the online setting where we have access to the data in a streaming manner, one data point after the other. More specifically, for a true parameter $\\theta^*$, we consider the corrupted Gaussian linear model $y = \\langle x , \\ \\theta^* \\rangle + \\varepsilon + b$ where the adversarial noise $b$ can take any value with probability $\\eta$ and equals zero otherwise. We consider this adversary to be oblivious (i.e., $b$ independent of the data) since this is the only contamination model under which consistency is possible. Current algorithms rely on having the whole data at hand in order to identify and remove the outliers. In contrast, we show in this work that stochastic gradient descent on the $\\ell_1$ loss converges to the true parameter vector at a $\\tilde{O}( 1 / (1 - \\eta)^2 n )$ rate which is independent of the values of the contaminated measurements. Our proof relies on the elegant smoothing of the non-smooth $\\ell_1$ loss by the Gaussian data and a classical non-asymptotic analysis of Polyak-Ruppert averaged SGD. In addition, we provide experimental evidence of the efficiency of this simple and highly scalable algorithm."
                    },
                    {
                        "title": "Towards Understanding Sharpness-Aware Minimization",
                        "abstract": "Sharpness-Aware Minimization (SAM) is a recent training method that relies on worst-case weight perturbations which significantly improves generalization in various settings. We argue that the existing justifications for the success of SAM which are based on a PAC-Bayes generalization bound and the idea of convergence to flat minima are incomplete. Moreover, there are no explanations for the success of using $m$-sharpness in SAM which has been shown as essential for generalization. To better understand this aspect of SAM, we theoretically analyze its implicit bias for diagonal linear networks. We prove that SAM always chooses a solution that enjoys better generalization properties than standard gradient descent for a certain class of problems, and this effect is amplified by using $m$-sharpness. We further study the properties of the implicit bias on non-linear networks empirically, where we show that fine-tuning a standard model with SAM can lead to significant generalization improvements. Finally, we provide convergence results of SAM for non-convex objectives when used with stochastic gradients. We illustrate these results empirically for deep networks and discuss their relation to the generalization behavior of SAM. The code of our experiments is available at https://github.com/tml-epfl/understanding-sam."
                    },
                    {
                        "title": "Penalising the biases in norm regularisation enforces sparsity",
                        "abstract": "Controlling the parameters' norm often yields good generalisation when training neural networks. Beyond simple intuitions, the relation between regularising parameters' norm and obtained estimators remains theoretically misunderstood. For one hidden ReLU layer networks with unidimensional data, this work shows the parameters' norm required to represent a function is given by the total variation of its second derivative, weighted by a $\\sqrt{1+x^2}$ factor. Notably, this weighting factor disappears when the norm of bias terms is not regularised. The presence of this additional weighting factor is of utmost significance as it is shown to enforce the uniqueness and sparsity (in the number of kinks) of the minimal norm interpolator. Conversely, omitting the bias' norm allows for non-sparse solutions. Penalising the bias terms in the regularisation, either explicitly or implicitly, thus leads to sparse estimators."
                    },
                    {
                        "title": "Saddle-to-Saddle Dynamics in Diagonal Linear Networks",
                        "abstract": "In this paper we fully describe the trajectory of gradient flow over diagonal linear networks in the limit of vanishing initialisation. We show that the limiting flow successively jumps from a saddle of the training loss to another until reaching the minimum $\\ell_1$-norm solution. This saddle-to-saddle dynamics translates to an incremental learning process as each saddle corresponds to the minimiser of the loss constrained to an active set outside of which the coordinates must be zero. We explicitly characterise the visited saddles as well as the jumping times through a recursive algorithm reminiscent of the LARS algorithm used for computing the Lasso path. Our proof leverages a convenient arc-length time-reparametrisation which enables to keep track of the heteroclinic transitions between the jumps. Our analysis requires negligible assumptions on the data, applies to both under and overparametrised settings and covers complex cases where there is no monotonicity of the number of active coordinates. We provide numerical experiments to support our findings."
                    },
                    {
                        "title": "Early alignment in two-layer networks training is a two-edged sword",
                        "abstract": "Training neural networks with first order optimisation methods is at the core of the empirical success of deep learning. The scale of initialisation is a crucial factor, as small initialisations are generally associated to a feature learning regime, for which gradient descent is implicitly biased towards simple solutions. This work provides a general and quantitative description of the early alignment phase, originally introduced by Maennel et al. (2018) . For small initialisation and one hidden ReLU layer networks, the early stage of the training dynamics leads to an alignment of the neurons towards key directions. This alignment induces a sparse representation of the network, which is directly related to the implicit bias of gradient flow at convergence. This sparsity inducing alignment however comes at the expense of difficulties in minimising the training objective: we also provide a simple data example for which overparameterised networks fail to converge towards global minima and only converge to a spurious stationary point instead."
                    },
                    {
                        "title": "From Averaging to Acceleration, There is Only a Step-size",
                        "abstract": "We show that accelerated gradient descent, averaged gradient descent and the heavy-ball method for non-strongly-convex problems may be reformulated as constant parameter second-order difference equation algorithms, where stability of the system is equivalent to convergence at rate O(1/n 2), where n is the number of iterations. We provide a detailed analysis of the eigenvalues of the corresponding linear dynamical system , showing various oscillatory and non-oscillatory behaviors, together with a sharp stability result with explicit constants. We also consider the situation where noisy gradients are available, where we extend our general convergence result, which suggests an alternative algorithm (i.e., with different step sizes) that exhibits the good aspects of both averaging and acceleration."
                    },
                    {
                        "title": "Harder, Better, Faster, Stronger Convergence Rates for Least-Squares Regression",
                        "abstract": "We consider the optimization of a quadratic objective function whose gradients are only accessible through a stochastic oracle that returns the gradient at any given point plus a zero-mean finite variance random error. We present the first algorithm that achieves jointly the optimal prediction error rates for least-squares regression, both in terms of forgetting of initial conditions in O(1/n 2), and in terms of dependence on the noise and dimension d of the problem, as O(d/n). Our new algorithm is based on averaged accelerated regularized gradient descent, and may also be analyzed through finer assumptions on initial conditions and the Hessian matrix, leading to dimension-free quantities that may still be small while the \" optimal \" terms above are large. In order to characterize the tightness of these new bounds, we consider an application to non-parametric regression and use the known lower bounds on the statistical performance (without computational limits), which happen to match our bounds obtained from a single pass on the data and thus show optimality of our algorithm in a wide variety of particular trade-offs between bias and variance."
                    },
                    {
                        "title": "Fast Mean Estimation with Sub-Gaussian Rates",
                        "abstract": "We propose an estimator for the mean of a random vector in $\\mathbb{R}^d$ that can be computed in time $O(n^4+n^2d)$ for $n$ i.i.d.~samples and that has error bounds matching the sub-Gaussian case. The only assumptions we make about the data distribution are that it has finite mean and covariance; in particular, we make no assumptions about higher-order moments. Like the polynomial time estimator introduced by Hopkins, 2018, which is based on the sum-of-squares hierarchy, our estimator achieves optimal statistical efficiency in this challenging setting, but it has a significantly faster runtime and a simpler analysis."
                    },
                    {
                        "title": "Trace norm regularization for multi-task learning with scarce data",
                        "abstract": "Multi-task learning leverages structural similarities between multiple tasks to learn despite very few samples. Motivated by the recent success of neural networks applied to data-scarce tasks, we consider a linear low-dimensional shared representation model. Despite an extensive literature, existing theoretical results either guarantee weak estimation rates or require a large number of samples per task. This work provides the first estimation error bound for the trace norm regularized estimator when the number of samples per task is small. The advantages of trace norm regularization for learning data-scarce tasks extend to meta-learning and are confirmed empirically on synthetic datasets."
                    },
                    {
                        "title": "Understanding and Improving Fast Adversarial Training",
                        "abstract": "A recent line of work focused on making adversarial training computationally efficient for deep learning models. In particular, Wong et al. (2020) showed that $\\ell_\\infty$-adversarial training with fast gradient sign method (FGSM) can fail due to a phenomenon called \"catastrophic overfitting\", when the model quickly loses its robustness over a single epoch of training. We show that adding a random step to FGSM, as proposed in Wong et al. (2020), does not prevent catastrophic overfitting, and that randomness is not important per se -- its main role being simply to reduce the magnitude of the perturbation. Moreover, we show that catastrophic overfitting is not inherent to deep and overparametrized networks, but can occur in a single-layer convolutional network with a few filters. In an extreme case, even a single filter can make the network highly non-linear locally, which is the main reason why FGSM training fails. Based on this observation, we propose a new regularization method, GradAlign, that prevents catastrophic overfitting by explicitly maximizing the gradient alignment inside the perturbation set and improves the quality of the FGSM solution. As a result, GradAlign allows to successfully apply FGSM training also for larger $\\ell_\\infty$-perturbations and reduce the gap to multi-step adversarial training. The code of our experiments is available at https://github.com/tml-epfl/understanding-fast-adv-training."
                    },
                    {
                        "title": "Simplicity bias and optimization threshold in two-layer ReLU networks",
                        "abstract": "Understanding generalization of overparametrized neural networks remains a fundamental challenge in machine learning. Most of the literature mostly studies generalization from an interpolation point of view, taking convergence of parameters towards a global minimum of the training loss for granted. While overparametrized architectures indeed interpolated the data for typical classification tasks, this interpolation paradigm does not seem valid anymore for more complex tasks such as in-context learning or diffusion. Instead for such tasks, it has been empirically observed that the trained models goes from global minima to spurious local minima of the training loss as the number of training samples becomes larger than some level we call optimization threshold. While the former yields a poor generalization to the true population loss, the latter was observed to actually correspond to the minimiser of this true loss. This paper explores theoretically this phenomenon in the context of two-layer ReLU networks. We demonstrate that, despite overparametrization, networks often converge toward simpler solutions rather than interpolating the training data, which can lead to a drastic improvement on the test loss with respect to interpolating solutions. Our analysis relies on the so called early alignment phase, during which neurons align towards specific directions. This directional alignment, which occurs in the early stage of training, leads to a simplicity bias, wherein the network approximates the ground truth model without converging to the global minimum of the training loss. Our results suggest that this bias, resulting in an optimization threshold from which interpolation is not reached anymore, is beneficial and enhances the generalization of trained models."
                    },
                    {
                        "title": "First-order ANIL provably learns representations despite overparametrization",
                        "abstract": "Due to its empirical success in few-shot classification and reinforcement learning, meta-learning has recently received significant interest. Meta-learning methods leverage data from previous tasks to learn a new task in a sample-efficient manner. In particular, model-agnostic methods look for initialization points from which gradient descent quickly adapts to any new task. Although it has been empirically suggested that such methods perform well by learning shared representations during pretraining, there is limited theoretical evidence of such behavior. More importantly, it has not been shown that these methods still learn a shared structure, despite architectural misspecifications. In this direction, this work shows, in the limit of an infinite number of tasks, that first-order ANIL with a linear two-layer network architecture successfully learns linear shared representations. This result even holds with overparametrization; having a width larger than the dimension of the shared representations results in an asymptotically low-rank solution. The learned solution then yields a good adaptation performance on any new task after a single gradient step. Overall, this illustrates how well model-agnostic methods such as first-order ANIL can learn shared representations."
                    },
                    {
                        "title": "Does Refusal Training in LLMs Generalize to the Past Tense?",
                        "abstract": "Refusal training is widely used to prevent LLMs from generating harmful, undesirable, or illegal outputs. We reveal a curious generalization gap in the current refusal training approaches: simply reformulating a harmful request in the past tense (e.g., \"How to make a Molotov cocktail?\" to \"How did people make a Molotov cocktail?\") is often sufficient to jailbreak many state-of-the-art LLMs. We systematically evaluate this method on Llama-3 8B, Claude-3.5 Sonnet, GPT-3.5 Turbo, Gemma-2 9B, Phi-3-Mini, GPT-4o mini, GPT-4o, o1-mini, o1-preview, and R2D2 models using GPT-3.5 Turbo as a reformulation model. For example, the success rate of this simple attack on GPT-4o increases from 1% using direct requests to 88% using 20 past tense reformulation attempts on harmful requests from JailbreakBench with GPT-4 as a jailbreak judge. Interestingly, we also find that reformulations in the future tense are less effective, suggesting that refusal guardrails tend to consider past historical questions more benign than hypothetical future questions. Moreover, our experiments on fine-tuning GPT-3.5 Turbo show that defending against past reformulations is feasible when past tense examples are explicitly included in the fine-tuning data. Overall, our findings highlight that the widely used alignment techniques -- such as SFT, RLHF, and adversarial training -- employed to align the studied models can be brittle and do not always generalize as intended. We provide code and jailbreak artifacts at https://github.com/tml-epfl/llm-past-tense."
                    },
                    {
                        "title": "Optimal Rates of Statistical Seriation",
                        "abstract": "Given a matrix the seriation problem consists in permuting its rows in such way that all its columns have the same shape, for example, they are monotone increasing. We propose a statistical approach to this problem where the matrix of interest is observed with noise and study the corresponding minimax rate of estimation of the matrices. Specifically, when the columns are either unimodal or monotone, we show that the least squares estimator is optimal up to logarithmic factors and adapts to matrices with a certain natural structure. Finally, we propose a computationally efficient estimator in the monotonic case and study its performance both theoretically and experimentally. Our work is at the intersection of shape constrained estimation and recent work that involves permutation learning, such as graph denoising and ranking."
                    },
                    {
                        "title": "Escaping from saddle points on Riemannian manifolds",
                        "abstract": "We consider minimizing a nonconvex, smooth function $f$ on a Riemannian manifold $\\mathcal{M}$. We show that a perturbed version of Riemannian gradient descent algorithm converges to a second-order stationary point (and hence is able to escape saddle points on the manifold). The rate of convergence depends as $1/\\epsilon^2$ on the accuracy $\\epsilon$, which matches a rate known only for unconstrained smooth minimization. The convergence rate depends polylogarithmically on the manifold dimension $d$, hence is almost dimension-free. The rate also has a polynomial dependence on the parameters describing the curvature of the manifold and the smoothness of the function. While the unconstrained problem (Euclidean setting) is well-studied, our result is the first to prove such a rate for nonconvex, manifold-constrained problems."
                    },
                    {
                        "title": "Last iterate convergence of SGD for Least-Squares in the Interpolation regime",
                        "abstract": "Motivated by the recent successes of neural networks that have the ability to fit the data perfectly and generalize well, we study the noiseless model in the fundamental least-squares setup. We assume that an optimum predictor fits perfectly inputs and outputs $\\langle \\theta_* , \\phi(X) \\rangle = Y$, where $\\phi(X)$ stands for a possibly infinite dimensional non-linear feature map. To solve this problem, we consider the estimator given by the last iterate of stochastic gradient descent (SGD) with constant step-size. In this context, our contribution is two fold: (i) from a (stochastic) optimization perspective, we exhibit an archetypal problem where we can show explicitly the convergence of SGD final iterate for a non-strongly convex problem with constant step-size whereas usual results use some form of average and (ii) from a statistical perspective, we give explicit non-asymptotic convergence rates in the over-parameterized setting and leverage a fine-grained parameterization of the problem to exhibit polynomial rates that can be faster than $O(1/T)$. The link with reproducing kernel Hilbert spaces is established."
                    },
                    {
                        "title": "Robust Discriminative Clustering with Sparse Regularizers",
                        "abstract": "Clustering high-dimensional data often requires some form of dimensionality reduction, where clustered variables are separated from \"noise-looking\" variables. We cast this problem as finding a low-dimensional projection of the data which is well-clustered. This yields a one-dimensional projection in the simplest situation with two clusters, and extends naturally to a multi-label scenario for more than two clusters. In this paper, (a) we first show that this joint clustering and dimension reduction formulation is equivalent to previously proposed discriminative clustering frameworks, thus leading to convex relaxations of the problem, (b) we propose a novel sparse extension, which is still cast as a convex relaxation and allows estimation in higher dimensions, (c) we propose a natural extension for the multi-label scenario, (d) we provide a new theoretical analysis of the performance of these formulations with a simple probabilistic model, leading to scalings over the form $d=O(\\sqrt{n})$ for the affine invariant case and $d=O(n)$ for the sparse case, where $n$ is the number of examples and $d$ the ambient dimension, and finally, (e) we propose an efficient iterative algorithm with running-time complexity proportional to $O(nd^2)$, improving on earlier algorithms which had quadratic complexity in the number of examples."
                    },
                    {
                        "title": "On Convergence-Diagnostic based Step Sizes for Stochastic Gradient Descent",
                        "abstract": "Constant step-size Stochastic Gradient Descent exhibits two phases: a transient phase during which iterates make fast progress towards the optimum, followed by a stationary phase during which iterates oscillate around the optimal point. In this paper, we show that efficiently detecting this transition and appropriately decreasing the step size can lead to fast convergence rates. We analyse the classical statistical test proposed by Pflug (1983), based on the inner product between consecutive stochastic gradients. Even in the simple case where the objective function is quadratic we show that this test cannot lead to an adequate convergence diagnostic. We then propose a novel and simple statistical procedure that accurately detects stationarity and we provide experimental results showing state-of-the-art performance on synthetic and real-world datasets."
                    }
                ]
            }
        }
    },
    "2404.13628": {
        "paper_data": {
            "title": "Mixture of LoRA Experts",
            "url": "http://arxiv.org/abs/2404.13628v1",
            "arxiv_id": "2404.13628",
            "authors": [
                "Xun Wu",
                "Shaohan Huang",
                "Furu Wei"
            ],
            "abstract": "LoRA has gained widespread acceptance in the fine-tuning of large pre-trained models to cater to a diverse array of downstream tasks, showcasing notable effectiveness and efficiency, thereby solidifying its position as one of the most prevalent fine-tuning techniques. Due to the modular nature of LoRA's plug-and-play plugins, researchers have delved into the amalgamation of multiple LoRAs to empower models to excel across various downstream tasks. Nonetheless, extant approaches for LoRA fusion grapple with inherent challenges. Direct arithmetic merging may result in the loss of the original pre-trained model's generative capabilities or the distinct identity of LoRAs, thereby yielding suboptimal outcomes. On the other hand, Reference tuning-based fusion exhibits limitations concerning the requisite flexibility for the effective combination of multiple LoRAs. In response to these challenges, this paper introduces the Mixture of LoRA Experts (MoLE) approach, which harnesses hierarchical control and unfettered branch selection. The MoLE approach not only achieves superior LoRA fusion performance in comparison to direct arithmetic merging but also retains the crucial flexibility for combining LoRAs effectively. Extensive experimental evaluations conducted in both the Natural Language Processing (NLP) and Vision & Language (V&L) domains substantiate the efficacy of MoLE.",
            "introduction": "   1 Introduction  Figure 1: Workflow of MoLE. In the training phase, MoLE predicts weights for multiple LoRAs. In the inference phase, MoLE can allocate weights to multiple LoRAs, or, without altering the gating weights, achieve a more flexible LoRA composition by masking out undesired LoRAs and recalculating and distributing weights proportionally.   Recent advances in deep learning have been driven by large-scale pre-trained models such as OPT\u00a0(Zhang et\u00a0al., 2022), LLaMA\u00a0(Touvron et\u00a0al., 2023) in the Natural Language Processing\u00a0(NLP) domain and CLIP\u00a0(Radford et\u00a0al., 2021a), DALL\u00b7E 2\u00a0(Ramesh et\u00a0al., 2022) in the Vision & Language\u00a0(V&L) domain. These models show outstanding performance across various tasks when fine-tuned on down-stream datasets, but their increasing size entails significant computational costs for full fine-tuning.       (a) (b) (c)   Figure 2: Overview of LoRA composition methods:\u00a0(a) Linear arithmetic composition (Eq.2), which commonly applies the same composition weight \ud835\udc7eisubscript\ud835\udc7e\ud835\udc56\\bm{W}_{i}bold_italic_W start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT to all layers of the it\u2062hsuperscript\ud835\udc56\ud835\udc61\u210ei^{th}italic_i start_POSTSUPERSCRIPT italic_t italic_h end_POSTSUPERSCRIPT LoRA. (b) Reference tuning-based composition involves retraining a large model by integrating outputs from multiple LoRAs using manually-crafted mask information. (c) Our MoLE, which learns a distribution \u03a5jsuperscript\u03a5\ud835\udc57\\Upsilon^{j}roman_\u03a5 start_POSTSUPERSCRIPT italic_j end_POSTSUPERSCRIPT for the jt\u2062hsuperscript\ud835\udc57\ud835\udc61\u210ej^{th}italic_j start_POSTSUPERSCRIPT italic_t italic_h end_POSTSUPERSCRIPT layer of LoRAs to determine the composition weight \ud835\udc7eijsubscriptsuperscript\ud835\udc7e\ud835\udc57\ud835\udc56\\bm{W}^{j}_{i}bold_italic_W start_POSTSUPERSCRIPT italic_j end_POSTSUPERSCRIPT start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT.   To mitigate this, LoRA\u00a0(Hu et\u00a0al., 2021) is introduced. By freezing the pretrained model weights and injecting trainable rank decomposition matrices, LoRA is proven to be an effective fine-tuning methodology in scenarios with constrained computational resources\u00a0(Lester et\u00a0al., 2021; An et\u00a0al., 2022).   While LoRA serves as plug-and-play plugins for pre-trained models, recent initiatives explores the composition of separate trained LoRAs to achieve joint generation of learned characteristics\u00a0(Huang et\u00a0al., 2023; Zhang et\u00a0al., 2023; Ruiz et\u00a0al., 2023). However, these efforts may encounter several challenges. As shown in Figure\u00a02 (a), linear arithmetic composition\u00a0(Zhang et\u00a0al., 2023; Huang et\u00a0al., 2023; Han et\u00a0al., 2023) composes trained LoRAs directly. However, composing multiple LoRAs (typically \u2265\\geq\u2265 3) can impair the generative performance of pre-trained models. To mitigate this, weight normalization is applied prior to the composition, but may erase the unique characteristics of individual trained LoRAs as the composing weight of each LoRA is reduced (refer to Observation 1 in \u00a7\u00a03.1). Another approach, as depicted in Figure\u00a02 (b), known as reference tuning-based composition\u00a0(Gu et\u00a0al., 2023), is tailored for the V&L domain and achieves superior performance. However, it is limited in terms of LoRA flexibility due to the utilization of manually-designed masks and involves substantial training costs, necessitating a full model retraining. In light of the above situation, an important question arises:     How can multiple trained LoRAs be composed dynamically and efficiently, while preserving  all their individual characteristics?    To address that issues, we introduce Mixture of LoRA Experts\u00a0(MoLE). Recognizing that individual layers of a trained LoRA exhibit distinct characteristics, which collectively define the overall characteristic of the trained LoRA (refer to Observation 2 in \u00a7\u00a03.1), MoLE involves modulating the weights of different trained LoRAs within each layer, which we refer to as \u201chierarchical weight contro\u201d. As shown in Figure\u00a02\u00a0(c), MoLE views each layer of trained LoRAs as a individual expert and incorporates a gating function within each layer to learn the optimal composition weights based on a specified domain objective. This dynamically enhances desirable",
            "references": [
                {
                    "title": "LoraHub: Efficient Cross-Task Generalization via Dynamic LoRA Composition",
                    "abstract": "Low-rank adaptations (LoRA) are often employed to fine-tune large language models (LLMs) for new tasks. This paper investigates LoRA composability for cross-task generalization and introduces LoraHub, a simple framework devised for the purposive assembly of LoRA modules trained on diverse given tasks, with the objective of achieving adaptable performance on unseen tasks. With just a few examples from a new task, LoraHub can fluidly combine multiple LoRA modules, eliminating the need for human expertise and assumptions. Notably, the composition requires neither additional model parameters nor gradients. Empirical results on the Big-Bench Hard benchmark suggest that LoraHub, while not surpassing the performance of in-context learning, offers a notable performance-efficiency trade-off in few-shot scenarios by employing a significantly reduced number of tokens per example during inference. Notably, LoraHub establishes a better upper bound compared to in-context learning when paired with different demonstration examples, demonstrating its potential for future development. Our vision is to establish a platform for LoRA modules, empowering users to share their trained LoRA modules. This collaborative approach facilitates the seamless application of LoRA modules to novel tasks, contributing to an adaptive ecosystem. Our code is available at https://github.com/sail-sg/lorahub, and all the pre-trained LoRA modules are released at https://huggingface.co/lorahub."
                },
                {
                    "title": "Composing Parameter-Efficient Modules with Arithmetic Operations",
                    "abstract": "As an efficient alternative to conventional full finetuning, parameter-efficient finetuning (PEFT) is becoming the prevailing method to adapt pretrained language models. In PEFT, a lightweight module is learned on each dataset while the underlying pretrained language model remains unchanged, resulting in multiple compact modules representing diverse skills when applied to various domains and tasks. In this paper, we propose to compose these parameter-efficient modules through linear arithmetic operations in the weight space, thereby integrating different module capabilities. Specifically, we first define addition and negation operators for the module, and then further compose these two basic operators to perform flexible arithmetic. Our approach requires \\emph{no additional training} and enables highly flexible module composition. We apply different arithmetic operations to compose the parameter-efficient modules for (1) distribution generalization, (2) multi-tasking, (3) unlearning, and (4) domain transfer. Additionally, we extend our approach to detoxify Alpaca-LoRA, the latest instruction-tuned large language model based on LLaMA. Empirical results demonstrate that our approach produces new and effective parameter-efficient modules that significantly outperform existing ones across all settings."
                },
                {
                    "title": "Mix-of-Show: Decentralized Low-Rank Adaptation for Multi-Concept Customization of Diffusion Models",
                    "abstract": "Public large-scale text-to-image diffusion models, such as Stable Diffusion, have gained significant attention from the community. These models can be easily customized for new concepts using low-rank adaptations (LoRAs). However, the utilization of multiple concept LoRAs to jointly support multiple customized concepts presents a challenge. We refer to this scenario as decentralized multi-concept customization, which involves single-client concept tuning and center-node concept fusion. In this paper, we propose a new framework called Mix-of-Show that addresses the challenges of decentralized multi-concept customization, including concept conflicts resulting from existing single-client LoRA tuning and identity loss during model fusion. Mix-of-Show adopts an embedding-decomposed LoRA (ED-LoRA) for single-client tuning and gradient fusion for the center node to preserve the in-domain essence of single concepts and support theoretically limitless concept fusion. Additionally, we introduce regionally controllable sampling, which extends spatially controllable sampling (e.g., ControlNet and T2I-Adaptor) to address attribute binding and missing object problems in multi-concept sampling. Extensive experiments demonstrate that Mix-of-Show is capable of composing multiple customized concepts with high fidelity, including characters, objects, and scenes."
                },
                {
                    "title": "SVDiff: Compact Parameter Space for Diffusion Fine-Tuning",
                    "abstract": "Diffusion models have achieved remarkable success in text-to-image generation, enabling the creation of high-quality images from text prompts or other modalities. However, existing methods for customizing these models are limited by handling multiple personalized subjects and the risk of overfitting. Moreover, their large number of parameters is inefficient for model storage. In this paper, we propose a novel approach to address these limitations in existing text-to-image diffusion models for personalization. Our method involves fine-tuning the singular values of the weight matrices, leading to a compact and efficient parameter space that reduces the risk of overfitting and language-drifting. We also propose a Cut-Mix-Unmix data-augmentation technique to enhance the quality of multi-subject image generation and a simple text-based image editing framework. Our proposed SVDiff method has a significantly smaller model size compared to existing methods (\u22482,200 times fewer parameters compared with vanilla DreamBooth), making it more practical for real-world applications."
                },
                {
                    "title": "P+: Extended Textual Conditioning in Text-to-Image Generation",
                    "abstract": "We introduce an Extended Textual Conditioning space in text-to-image models, referred to as $P+$. This space consists of multiple textual conditions, derived from per-layer prompts, each corresponding to a layer of the denoising U-net of the diffusion model. We show that the extended space provides greater disentangling and control over image synthesis. We further introduce Extended Textual Inversion (XTI), where the images are inverted into $P+$, and represented by per-layer tokens. We show that XTI is more expressive and precise, and converges faster than the original Textual Inversion (TI) space. The extended inversion method does not involve any noticeable trade-off between reconstruction and editability and induces more regular inversions. We conduct a series of extensive experiments to analyze and understand the properties of the new space, and to showcase the effectiveness of our method for personalizing text-to-image models. Furthermore, we utilize the unique properties of this space to achieve previously unattainable results in object-style mixing using text-to-image models. Project page: https://prompt-plus.github.io"
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Multi-Concept Customization of Text-to-Image Diffusion",
                    "abstract": "While generative models produce high-quality images of concepts learned from a large-scale database, a user often wishes to synthesize instantiations of their own concepts (for example, their family, pets, or items). Can we teach a model to quickly acquire a new concept, given a few examples? Furthermore, can we compose multiple new concepts together? We propose Custom Diffusion, an efficient method for augmenting existing text-to-image models. We find that only optimizing a few parameters in the text-to-image conditioning mechanism is sufficiently powerful to represent new concepts while enabling fast tuning (~ 6 minutes). Additionally, we can jointly train for multiple concepts or combine multiple fine-tuned models into one via closed-form constrained optimization. Our fine-tuned model generates variations of multiple new concepts and seamlessly composes them with existing concepts in novel settings. Our method outperforms or performs on par with several baselines and concurrent works in both qualitative and quantitative evaluations, while being memory and computationally efficient."
                },
                {
                    "title": "Scaling Instruction-Finetuned Language Models",
                    "abstract": "Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models."
                },
                {
                    "title": "DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation",
                    "abstract": "Large text-to-image models achieved a remarkable leap in the evolution of AI, enabling high-quality and diverse synthesis of images from a given text prompt. However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts. In this work, we present a new approach for \u201cpersonalization\u201d of text-to-image diffusion models. Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject. Once the subject is embedded in the output domain of the model, the unique identifier can be used to synthesize novel photorealistic images of the subject contextualized in different scenes. By leveraging the semantic prior embedded in the model with a new autogenous class-specific prior preservation loss, our technique enables synthesizing the subject in diverse scenes, poses, views and lighting conditions that do not appear in the reference images. We apply our technique to several previously-unassailable tasks, including subject recontextualization, text-guided view synthesis, and artistic rendering, all while preserving the subject's key features. We also provide a new dataset and evaluation protocol for this new task of subject-driven generation. Project page: https://dreambooth.github.io/"
                },
                {
                    "title": "An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion",
                    "abstract": "Text-to-image models offer unprecedented freedom to guide creation through natural language. Yet, it is unclear how such freedom can be exercised to generate images of specific unique concepts, modify their appearance, or compose them in new roles and novel scenes. In other words, we ask: how can we use language-guided models to turn our cat into a painting, or imagine a new product based on our favorite toy? Here we present a simple approach that allows such creative freedom. Using only 3-5 images of a user-provided concept, like an object or a style, we learn to represent it through new\"words\"in the embedding space of a frozen text-to-image model. These\"words\"can be composed into natural language sentences, guiding personalized creation in an intuitive way. Notably, we find evidence that a single word embedding is sufficient for capturing unique and varied concepts. We compare our approach to a wide range of baselines, and demonstrate that it can more faithfully portray the concepts across a range of applications and tasks. Our code, data and new words will be available at: https://textual-inversion.github.io"
                },
                {
                    "title": "MoEC: Mixture of Expert Clusters",
                    "abstract": "Sparsely Mixture of Experts (MoE) has received great interest due to its promising scaling capability with affordable computational overhead. MoE models convert dense layers into sparse experts, and utilize a gated routing network to make experts conditionally activated. However, as the number of experts grows, MoE with outrageous parameters suffers from overfitting and sparse data allocation. Such problems are especially severe on tasks with limited data, thus hindering the progress towards improving performance by scaling up. We verify that there exists a performance upper bound of scaling up sparse MoE. In this work, we propose Mixture of Expert Clusters \u2014 a general approach to enable expert layers to learn more diverse and appropriate knowledge by imposing variance-based constraints on the routing stage. Given this, we could further propose a cluster-level expert dropout strategy specifically designed for the expert cluster structure. Our experiments reveal that MoEC could improve performance on machine translation and natural language understanding tasks. MoEC plays a positive role in mitigating overfitting and sparse data allocation problems, thus fully releasing the potential of large-scale sparse models."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
                    "abstract": "Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples."
                },
                {
                    "title": "Input-Tuning: Adapting Unfamiliar Inputs to Frozen Pretrained Models",
                    "abstract": "Recently the prompt-tuning paradigm has attracted significant attention. By only tuning continuous prompts with a frozen pre-trained language model (PLM), prompt-tuning takes a step towards deploying a shared frozen PLM to serve numerous downstream tasks. Although prompt-tuning shows good performance on certain natural language understanding (NLU) tasks, its effectiveness on natural language generation (NLG) tasks is still under-explored. In this paper, we argue that one of the factors hindering the development of prompt-tuning on NLG tasks is the unfamiliar inputs (i.e., inputs are linguistically different from the pretraining corpus). For example, our preliminary exploration reveals a large performance gap between prompt-tuning and fine-tuning when unfamiliar inputs occur frequently in NLG tasks. This motivates us to propose input-tuning, which fine-tunes both the continuous prompts and the input representations, leading to a more effective way to adapt unfamiliar inputs to frozen PLMs. Our proposed input-tuning is conceptually simple and empirically powerful. Experimental results on seven NLG tasks demonstrate that input-tuning is significantly and consistently better than prompt-tuning. Furthermore, on three of these tasks, input-tuning can achieve a comparable or even better performance than fine-tuning."
                },
                {
                    "title": "VL-ADAPTER: Parameter-Efficient Transfer Learning for Vision-and-Language Tasks",
                    "abstract": "Recently, fine-tuning language models pre-trained on large text corpora have provided huge improvements on vision-and-language (V&L) tasks as well as on pure language tasks. However, fine-tuning the entire parameter set of pre-trained models becomes impractical since the model size is growing rapidly. Hence, in this paper, we introduce adapter-based parameter-efficient transfer learning techniques to V&L models such as VL-BART and VL-T5. We evaluate our methods in a unified multi-task setup on both image-text and video-text benchmarks. For the image-text tasks, we use four diverse V&L datasets: VQAv2, GQA, NLVR2, and MSCOCO image captioning. For video-text tasks, we use TVQA, How2QA, TVC, and YC2C. With careful training and thorough experiments, we benchmark three popular adapter-based methods (Adapter, Hyperformer, Compacter) against the standard full fine-tuning and the recently proposed prompt-tuning approach. We also enhance the efficiency and performance of adapters by sharing their weights to attain knowledge across tasks. Our results demonstrate that training the adapter with the weight-sharing technique (4.18% of total parameters for image-text tasks and 3.39% for video-text tasks) can match the performance of fine-tuning the entire model. Lastly, we present a comprehensive analysis including the combination of adapter and task-specific prompts and the impact of V&L pre-training on adapters. 11The code for our CVPR 2022 paper is available at: https://github.com/ylsung/VL_adapter."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "The Power of Scale for Parameter-Efficient Prompt Tuning",
                    "abstract": "In this work, we explore \u201cprompt tuning,\u201d a simple yet effective mechanism for learning \u201csoft prompts\u201d to condition frozen language models to perform specific downstream tasks. Unlike the discrete text prompts used by GPT-3, soft prompts are learned through backpropagation and can be tuned to incorporate signals from any number of labeled examples. Our end-to-end learned approach outperforms GPT-3\u2019s few-shot learning by a large margin. More remarkably, through ablations on model size using T5, we show that prompt tuning becomes more competitive with scale: as models exceed billions of parameters, our method \u201ccloses the gap\u201d and matches the strong performance of model tuning (where all model weights are tuned). This finding is especially relevant because large models are costly to share and serve and the ability to reuse one frozen model for multiple downstream tasks can ease this burden. Our method can be seen as a simplification of the recently proposed \u201cprefix tuning\u201d of Li and Liang (2021) and we provide a comparison to this and other similar approaches. Finally, we show that conditioning a frozen model with soft prompts confers benefits in robustness to domain transfer and enables efficient \u201cprompt ensembling.\u201d We release code and model checkpoints to reproduce our experiments."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Adversarial NLI: A New Benchmark for Natural Language Understanding",
                    "abstract": "We introduce a new large-scale NLI benchmark dataset, collected via an iterative, adversarial human-and-model-in-the-loop procedure. We show that training models on this new dataset leads to state-of-the-art performance on a variety of popular NLI benchmarks, while posing a more difficult challenge with its new test set. Our analysis sheds light on the shortcomings of current state-of-the-art models, and shows that non-expert annotators are successful at finding their weaknesses. The data collection method can be applied in a never-ending learning scenario, becoming a moving target for NLU, rather than a static benchmark that will quickly saturate."
                },
                {
                    "title": "Know What You Don\u2019t Know: Unanswerable Questions for SQuAD",
                    "abstract": "Extractive reading comprehension systems can often locate the correct answer to a question in a context document, but they also tend to make unreliable guesses on questions for which the correct answer is not stated in the context. Existing datasets either focus exclusively on answerable questions, or use automatically generated unanswerable questions that are easy to identify. To address these weaknesses, we present SQuADRUn, a new dataset that combines the existing Stanford Question Answering Dataset (SQuAD) with over 50,000 unanswerable questions written adversarially by crowdworkers to look similar to answerable ones. To do well on SQuADRUn, systems must not only answer questions when possible, but also determine when no answer is supported by the paragraph and abstain from answering. SQuADRUn is a challenging natural language understanding task for existing models: a strong neural system that gets 86% F1 on SQuAD achieves only 66% F1 on SQuADRUn. We release SQuADRUn to the community as the successor to SQuAD."
                },
                {
                    "title": "BigBench: towards an industry standard benchmark for big data analytics",
                    "abstract": "There is a tremendous interest in big data by academia, industry and a large user base. Several commercial and open source providers unleashed a variety of products to support big data storage and processing. As these products mature, there is a need to evaluate and compare the performance of these systems.\n In this paper, we present BigBench, an end-to-end big data benchmark proposal. The underlying business model of BigBench is a product retailer. The proposal covers a data model and synthetic data generator that addresses the variety, velocity and volume aspects of big data systems containing structured, semi-structured and unstructured data. The structured part of the BigBench data model is adopted from the TPC-DS benchmark, which is enriched with semi-structured and unstructured data components. The semi-structured part captures registered and guest user clicks on the retailer's website. The unstructured data captures product reviews submitted online. The data generator designed for BigBench provides scalable volumes of raw data based on a scale factor. The BigBench workload is designed around a set of queries against the data model. From a business prospective, the queries cover the different categories of big data analytics proposed by McKinsey. From a technical prospective, the queries are designed to span three different dimensions based on data sources, query processing types and analytic techniques.\n We illustrate the feasibility of BigBench by implementing it on the Teradata Aster Database. The test includes generating and loading a 200 Gigabyte BigBench data set and testing the workload by executing the BigBench queries (written using Teradata Aster SQL-MR) and reporting their response times."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.LG",
                "cs.MM"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can multiple trained LoRAs be composed dynamically and efficiently, while preserving all their individual characteristics?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of existing methods for composing multiple LoRAs, which often lead to a loss of unique characteristics and require substantial computational resources. By enabling dynamic and efficient composition of LoRAs, this research could significantly enhance the performance of large-scale pre-trained models in various applications, leading to advancements in Natural Language Processing and Vision & Language tasks. This could pave the way for more flexible and resource-efficient fine-tuning methodologies, ultimately influencing future research directions and practical applications in machine learning.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities of preserving the unique characteristics of individual LoRAs while composing them. Naive approaches, such as linear arithmetic composition, can degrade the generative performance of pre-trained models when combining multiple LoRAs, especially when the number exceeds three. Additionally, reference tuning-based methods require full model retraining and rely on manually-designed masks, which limits flexibility and increases computational costs. Overcoming these technical and practical obstacles requires innovative methodologies that can dynamically adjust composition weights without erasing the distinct features of each LoRA.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either linear composition methods or reference tuning-based approaches, both of which have significant limitations. The linear methods often lead to performance degradation when composing multiple LoRAs, while reference tuning methods are inflexible and computationally expensive due to their reliance on manual masks and full model retraining. These barriers have prevented the development of a more dynamic and efficient solution. The proposed Mixture of LoRA Experts (MoLE) approach differs by introducing hierarchical weight control and a gating function that allows for optimal composition weights to be learned based on domain objectives, thus improving upon prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology, Mixture of LoRA Experts (MoLE), involves modulating the weights of different trained LoRAs within each layer, treating each layer as an individual expert. The approach incorporates a gating function to learn optimal composition weights dynamically. The dataset used will consist of various pre-trained models in the NLP and V&L domains, and the performance will"
            }
        },
        "author_data": {
            "238ac45c-de93-4f2d-ba72-13b197452ba9": {
                "pk": "238ac45c-de93-4f2d-ba72-13b197452ba9",
                "name": "Xun Wu",
                "collaborators": [
                    "Shaohan Huang",
                    "Furu Wei",
                    "Wenhui Wang"
                ],
                "domain": [
                    "Machine Learning",
                    "Generative Models",
                    "Multimodal Learning",
                    "Sparse Mixtures of Experts"
                ],
                "publications": [
                    {
                        "title": "Multi-Head Mixture-of-Experts",
                        "abstract": "Sparse Mixtures of Experts (SMoE) scales model capacity without significant increases in training and inference costs, but exhibits the following two issues: (1) Low expert activation, where only a small subset of experts are activated for optimization. (2) Lacking fine-grained analytical capabilities for multiple semantic concepts within individual tokens. We propose Multi-Head Mixture-of-Experts (MH-MoE), which employs a multi-head mechanism to split each token into multiple sub-tokens. These sub-tokens are then assigned to and processed by a diverse set of experts in parallel, and seamlessly reintegrated into the original token form. The multi-head mechanism enables the model to collectively attend to information from various representation spaces within different experts, while significantly enhances expert activation, thus deepens context understanding and alleviate overfitting. Moreover, our MH-MoE is straightforward to implement and decouples from other SMoE optimization methods, making it easy to integrate with other SMoE models for enhanced performance. Extensive experimental results across three tasks: English-focused language modeling, Multi-lingual language modeling and Masked multi-modality modeling tasks, demonstrate the effectiveness of MH-MoE."
                    },
                    {
                        "title": "Multimodal Large Language Model is a Human-Aligned Annotator for Text-to-Image Generation",
                        "abstract": "Recent studies have demonstrated the exceptional potentials of leveraging human preference datasets to refine text-to-image generative models, enhancing the alignment between generated images and textual prompts. Despite these advances, current human preference datasets are either prohibitively expensive to construct or suffer from a lack of diversity in preference dimensions, resulting in limited applicability for instruction tuning in open-source text-to-image generative models and hinder further exploration. To address these challenges and promote the alignment of generative models through instruction tuning, we leverage multimodal large language models to create VisionPrefer, a high-quality and fine-grained preference dataset that captures multiple preference aspects. We aggregate feedback from AI annotators across four aspects: prompt-following, aesthetic, fidelity, and harmlessness to construct VisionPrefer. To validate the effectiveness of VisionPrefer, we train a reward model VP-Score over VisionPrefer to guide the training of text-to-image generative models and the preference prediction accuracy of VP-Score is comparable to human annotators. Furthermore, we use two reinforcement learning methods to supervised fine-tune generative models to evaluate the performance of VisionPrefer, and extensive experimental results demonstrate that VisionPrefer significantly improves text-image alignment in compositional image generation across diverse aspects, e.g., aesthetic, and generalizes better than previous human-preference metrics across various image distributions. Moreover, VisionPrefer indicates that the integration of AI-generated synthetic data as a supervisory signal is a promising avenue for achieving improved alignment with human preferences in vision generative models."
                    }
                ]
            },
            "8b646f49-0391-4dbf-ab96-8a72f709bdbf": {
                "pk": "8b646f49-0391-4dbf-ab96-8a72f709bdbf",
                "name": "Shaohan Huang",
                "collaborators": [
                    "Furu Wei",
                    "Weiwei Deng",
                    "Qi Zhang",
                    "Haizhen Huang",
                    "Zihan Zhang",
                    "Feng Sun",
                    "Li Dong",
                    "Shuming Ma",
                    "Wenhui Wang",
                    "Yuxuan Liu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Multimodal Learning",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Text Diffusion with Reinforced Conditioning",
                        "abstract": "Diffusion models have demonstrated exceptional capability in generating high-quality images, videos, and audio. Due to their adaptiveness in iterative refinement, they provide a strong potential for achieving better non-autoregressive sequence generation. However, existing text diffusion models still fall short in their performance due to a challenge in handling the discreteness of language. This paper thoroughly analyzes text diffusion models and uncovers two significant limitations: degradation of self-conditioning during training and misalignment between training and sampling. Motivated by our findings, we propose a novel Text Diffusion model called TReC, which mitigates the degradation with Reinforced Conditioning and the misalignment by Time-Aware Variance Scaling. Our extensive experiments demonstrate the competitiveness of TReC against autoregressive, non-autoregressive, and diffusion baselines. Moreover, qualitative analysis shows its advanced ability to fully utilize the diffusion process in refining samples."
                    },
                    {
                        "title": "You Only Cache Once: Decoder-Decoder Architectures for Language Models",
                        "abstract": "We introduce a decoder-decoder architecture, YOCO, for large language models, which only caches key-value pairs once. It consists of two components, i.e., a cross-decoder stacked upon a self-decoder. The self-decoder efficiently encodes global key-value (KV) caches that are reused by the cross-decoder via cross-attention. The overall model behaves like a decoder-only Transformer, although YOCO only caches once. The design substantially reduces GPU memory demands, yet retains global attention capability. Additionally, the computation flow enables prefilling to early exit without changing the final output, thereby significantly speeding up the prefill stage. Experimental results demonstrate that YOCO achieves favorable performance compared to Transformer in various settings of scaling up model size and number of training tokens. We also extend YOCO to 1M context length with near-perfect needle retrieval accuracy. The profiling results show that YOCO improves inference memory, prefill latency, and throughput by orders of magnitude across context lengths and model sizes. Code is available at https://aka.ms/YOCO."
                    },
                    {
                        "title": "Adapting Large Language Models via Reading Comprehension",
                        "abstract": "We explore how continued pre-training on domain-specific corpora influences large language models, revealing that training on the raw corpora endows the model with domain knowledge, but drastically hurts its prompting ability for question answering. Taken inspiration from human learning via reading comprehension\u2014practice after reading improves the ability to answer questions based on the learned knowledge\u2014we propose a simple method for transforming raw corpora into reading comprehension texts. Each raw text is enriched with a series of tasks related to its content. Our method, highly scalable and applicable to any pre-training corpora, consistently enhances performance across various tasks in three different domains: biomedicine, finance"
                    },
                    {
                        "title": "ResLoRA: Identity Residual Mapping in Low-Rank Adaption",
                        "abstract": "As one of the most popular parameter-efficient fine-tuning (PEFT) methods, low-rank adaptation (LoRA) is commonly applied to fine-tune large language models (LLMs). However, updating the weights of LoRA blocks effectively and expeditiously is challenging due to the long calculation path in the original model. To address this, we propose ResLoRA, an improved framework of LoRA. By adding residual paths during training and using merging approaches to eliminate these extra paths during inference, our method can achieve better results in fewer training steps without any extra trainable parameters or inference cost compared to LoRA. The experiments on NLG, NLU, and text-to-image tasks demonstrate the effectiveness of our method. To the best of our knowledge, ResLoRA is the first work that combines the residual path with LoRA. The code of our method is available at https://github.com/microsoft/LMOps/tree/main/reslora ."
                    },
                    {
                        "title": "Synthetic Data (Almost) from Scratch: Generalized Instruction Tuning for Language Models",
                        "abstract": "We introduce Generalized Instruction Tuning (called GLAN), a general and scalable method for instruction tuning of Large Language Models (LLMs). Unlike prior work that relies on seed examples or existing datasets to construct instruction tuning data, GLAN exclusively utilizes a pre-curated taxonomy of human knowledge and capabilities as input and generates large-scale synthetic instruction data across all disciplines. Specifically, inspired by the systematic structure in human education system, we build the taxonomy by decomposing human knowledge and capabilities to various fields, sub-fields and ultimately, distinct disciplines semi-automatically, facilitated by LLMs. Subsequently, we generate a comprehensive list of subjects for every discipline and proceed to design a syllabus tailored to each subject, again utilizing LLMs. With the fine-grained key concepts detailed in every class session of the syllabus, we are able to generate diverse instructions with a broad coverage across the entire spectrum of human knowledge and skills. Extensive experiments on large language models (e.g., Mistral) demonstrate that GLAN excels in multiple dimensions from mathematical reasoning, coding, academic exams, logical reasoning to general instruction following without using task-specific training data of these tasks. In addition, GLAN allows for easy customization and new fields or skills can be added by simply incorporating a new node into our taxonomy."
                    },
                    {
                        "title": "The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits",
                        "abstract": "Recent research, such as BitNet, is paving the way for a new era of 1-bit Large Language Models (LLMs). In this work, we introduce a 1-bit LLM variant, namely BitNet b1.58, in which every single parameter (or weight) of the LLM is ternary {-1, 0, 1}. It matches the full-precision (i.e., FP16 or BF16) Transformer LLM with the same model size and training tokens in terms of both perplexity and end-task performance, while being significantly more cost-effective in terms of latency, memory, throughput, and energy consumption. More profoundly, the 1.58-bit LLM defines a new scaling law and recipe for training new generations of LLMs that are both high-performance and cost-effective. Furthermore, it enables a new computation paradigm and opens the door for designing specific hardware optimized for 1-bit LLMs."
                    },
                    {
                        "title": "MoRA: High-Rank Updating for Parameter-Efficient Fine-Tuning",
                        "abstract": "Low-rank adaptation is a popular parameter-efficient fine-tuning method for large language models. In this paper, we analyze the impact of low-rank updating, as implemented in LoRA. Our findings suggest that the low-rank updating mechanism may limit the ability of LLMs to effectively learn and memorize new knowledge. Inspired by this observation, we propose a new method called MoRA, which employs a square matrix to achieve high-rank updating while maintaining the same number of trainable parameters. To achieve it, we introduce the corresponding non-parameter operators to reduce the input dimension and increase the output dimension for the square matrix. Furthermore, these operators ensure that the weight can be merged back into LLMs, which makes our method can be deployed like LoRA. We perform a comprehensive evaluation of our method across five tasks: instruction tuning, mathematical reasoning, continual pretraining, memory and pretraining. Our method outperforms LoRA on memory-intensive tasks and achieves comparable performance on other tasks."
                    },
                    {
                        "title": "HD-Eval: Aligning Large Language Model Evaluators Through Hierarchical Criteria Decomposition",
                        "abstract": "Large language models (LLMs) have emerged as a promising alternative to expensive human evaluations. However, the alignment and coverage of LLM-based evaluations are often limited by the scope and potential bias of the evaluation prompts and criteria. To address this challenge, we propose HD-Eval, a novel framework that iteratively aligns LLM-based evaluators with human preference via Hierarchical Criteria Decomposition. HD-Eval inherits the essence from the evaluation mindset of human experts and enhances the alignment of LLM-based evaluators by decomposing a given evaluation task into finer-grained criteria, aggregating them according to estimated human preferences, pruning insignificant criteria with attribution, and further decomposing significant criteria. By integrating these steps within an iterative alignment training process, we obtain a hierarchical decomposition of criteria that comprehensively captures aspects of natural language at multiple levels of granularity. Implemented as a white box, the human preference-guided aggregator is efficient to train and more explainable than relying solely on prompting, and its independence from model parameters makes it applicable to closed-source LLMs. Extensive experiments on three evaluation domains demonstrate the superiority of HD-Eval in further aligning state-of-the-art evaluators and providing deeper insights into the explanation of evaluation results and the task itself."
                    },
                    {
                        "title": "Se2: Sequential Example Selection for In-Context Learning",
                        "abstract": "The remarkable capability of large language models (LLMs) for in-context learning (ICL) needs to be activated by demonstration examples. Prior work has extensively explored the selection of examples for ICL, predominantly following the\"select then organize\"paradigm, such approaches often neglect the internal relationships between examples and exist an inconsistency between the training and inference. In this paper, we formulate the problem as a $Se$quential $Se$lection problem and introduce $Se^2$, a sequential-aware method that leverages the LLM's feedback on varying context, aiding in capturing inter-relationships and sequential information among examples, significantly enriching the contextuality and relevance of ICL prompts. Meanwhile, we utilize beam search to seek and construct example sequences, enhancing both quality and diversity. Extensive experiments across 23 NLP tasks from 8 distinct categories illustrate that $Se^2$ markedly surpasses competitive baselines and achieves 42\\% relative improvement over random selection. Further in-depth analysis shows the effectiveness of proposed strategies, highlighting $Se^2$'s exceptional stability and adaptability across various scenarios. Code available at https://github.com/microsoft/LMOps."
                    },
                    {
                        "title": "Improving Domain Adaptation through Extended-Text Reading Comprehension",
                        "abstract": "To enhance the domain-specific capabilities of large language models, continued pre-training on a domain-specific corpus is a prevalent method. Recent work demonstrates that adapting models using reading comprehension data formatted by regex-based patterns can significantly improve performance on domain-specific tasks. However, regex-based patterns are incapable of parsing raw corpora using domain-specific knowledge. Furthermore, the question and answer pairs are extracted directly from the corpus in predefined formats offers limited context. To address this limitation, we improve reading comprehension via LLM and clustering. LLM focuses on leveraging domain knowledge within the corpus to refine comprehension stage, while clustering supplies relevant knowledge by extending the context to enrich reading stage. Additionally, our method incorporates parameter-efficient fine-tuning to improve the efficiency of domain adaptation. In comparison to AdaptLLM, our method achieves an improvement exceeding 5% in domain-specific tasks. Our code will available at https://github.com/microsoft/LMOps."
                    },
                    {
                        "title": "Multi-Head Mixture-of-Experts",
                        "abstract": "Sparse Mixtures of Experts (SMoE) scales model capacity without significant increases in training and inference costs, but exhibits the following two issues: (1) Low expert activation, where only a small subset of experts are activated for optimization. (2) Lacking fine-grained analytical capabilities for multiple semantic concepts within individual tokens. We propose Multi-Head Mixture-of-Experts (MH-MoE), which employs a multi-head mechanism to split each token into multiple sub-tokens. These sub-tokens are then assigned to and processed by a diverse set of experts in parallel, and seamlessly reintegrated into the original token form. The multi-head mechanism enables the model to collectively attend to information from various representation spaces within different experts, while significantly enhances expert activation, thus deepens context understanding and alleviate overfitting. Moreover, our MH-MoE is straightforward to implement and decouples from other SMoE optimization methods, making it easy to integrate with other SMoE models for enhanced performance. Extensive experimental results across three tasks: English-focused language modeling, Multi-lingual language modeling and Masked multi-modality modeling tasks, demonstrate the effectiveness of MH-MoE."
                    },
                    {
                        "title": "Multimodal Large Language Model is a Human-Aligned Annotator for Text-to-Image Generation",
                        "abstract": "Recent studies have demonstrated the exceptional potentials of leveraging human preference datasets to refine text-to-image generative models, enhancing the alignment between generated images and textual prompts. Despite these advances, current human preference datasets are either prohibitively expensive to construct or suffer from a lack of diversity in preference dimensions, resulting in limited applicability for instruction tuning in open-source text-to-image generative models and hinder further exploration. To address these challenges and promote the alignment of generative models through instruction tuning, we leverage multimodal large language models to create VisionPrefer, a high-quality and fine-grained preference dataset that captures multiple preference aspects. We aggregate feedback from AI annotators across four aspects: prompt-following, aesthetic, fidelity, and harmlessness to construct VisionPrefer. To validate the effectiveness of VisionPrefer, we train a reward model VP-Score over VisionPrefer to guide the training of text-to-image generative models and the preference prediction accuracy of VP-Score is comparable to human annotators. Furthermore, we use two reinforcement learning methods to supervised fine-tune generative models to evaluate the performance of VisionPrefer, and extensive experimental results demonstrate that VisionPrefer significantly improves text-image alignment in compositional image generation across diverse aspects, e.g., aesthetic, and generalizes better than previous human-preference metrics across various image distributions. Moreover, VisionPrefer indicates that the integration of AI-generated synthetic data as a supervisory signal is a promising avenue for achieving improved alignment with human preferences in vision generative models."
                    },
                    {
                        "title": "Improving Log-Based Anomaly Detection by Pre-Training Hierarchical Transformers",
                        "abstract": "Pre-trained models, such as BERT, have resulted in significant pre-trained models, such as BERT, have resulted in significant improvements in many natural language processing (NLP) applications. However, due to differences in word distribution and domain data distribution, applying NLP advancements to log analysis directly faces some performance challenges. This paper studies how to adapt the recently introduced pre-trained language model BERT for log analysis. In this work, we propose a pre-trained log representation model with hierarchical bidirectional encoder transformers (namely, HilBERT). Unlike previous work, which used raw text as pre-training data, we parse logs into templates before using the log templates to pre-train HilBERT. We also design a hierarchical transformers model to capture log template sequence-level information. We use log-based anomaly detection for downstream tasks and fine-tune our model with different log data. Our experiments demonstrate that HilBERT outperforms other baseline techniques on unstable log data. While BERT obtains performance comparable to that of previous state-of-the-art models, HilBERT can significantly address the problem of log instability and achieve accurate and robust results."
                    },
                    {
                        "title": "Pre-training Language Model as a Multi-perspective Course Learner",
                        "abstract": "ELECTRA, the generator-discriminator pre-training framework, has achieved impressive semantic construction capability among various downstream tasks. Despite the convincing performance, ELECTRA still faces the challenges of monotonous training and deficient interaction. Generator with only masked language modeling (MLM) leads to biased learning and label imbalance for discriminator, decreasing learning efficiency; no explicit feedback loop from discriminator to generator results in the chasm between these two components, underutilizing the course learning. In this study, a multi-perspective course learning (MCL) method is proposed to fetch a many degrees and visual angles for sample-efficient pre-training, and to fully leverage the relationship between generator and discriminator. Concretely, three self-supervision courses are designed to alleviate inherent flaws of MLM and balance the label in a multi-perspective way. Besides, two self-correction courses are proposed to bridge the chasm between the two encoders by creating a\"correction notebook\"for secondary-supervision. Moreover, a course soups trial is conducted to solve the\"tug-of-war\"dynamics problem of MCL, evolving a stronger pre-trained model. Experimental results show that our method significantly improves ELECTRA's average performance by 2.8% and 3.2% absolute points respectively on GLUE and SQuAD 2.0 benchmarks, and overshadows recent advanced ELECTRA-style models under the same settings. The pre-trained MCL model is available at https://huggingface.co/McmanusChen/MCL-base."
                    },
                    {
                        "title": "Retentive Network: A Successor to Transformer for Large Language Models",
                        "abstract": "In this work, we propose Retentive Network (RetNet) as a foundation architecture for large language models, simultaneously achieving training parallelism, low-cost inference, and good performance. We theoretically derive the connection between recurrence and attention. Then we propose the retention mechanism for sequence modeling, which supports three computation paradigms, i.e., parallel, recurrent, and chunkwise recurrent. Specifically, the parallel representation allows for training parallelism. The recurrent representation enables low-cost $O(1)$ inference, which improves decoding throughput, latency, and GPU memory without sacrificing performance. The chunkwise recurrent representation facilitates efficient long-sequence modeling with linear complexity, where each chunk is encoded parallelly while recurrently summarizing the chunks. Experimental results on language modeling show that RetNet achieves favorable scaling results, parallel training, low-cost deployment, and efficient inference. The intriguing properties make RetNet a strong successor to Transformer for large language models. Code will be available at https://aka.ms/retnet."
                    },
                    {
                        "title": "Calibrating LLM-Based Evaluator",
                        "abstract": "Recent advancements in large language models (LLMs) and their emergent capabilities make LLM a promising reference-free evaluator on the quality of natural language generation, and a competent alternative to human evaluation. However, hindered by the closed-source or high computational demand to host and tune, there is a lack of practice to further calibrate an off-the-shelf LLM-based evaluator towards better human alignment. In this work, we propose AutoCalibrate, a multi-stage, gradient-free approach to automatically calibrate and align an LLM-based evaluator toward human preference. Instead of explicitly modeling human preferences, we first implicitly encompass them within a set of human labels. Then, an initial set of scoring criteria is drafted by the language model itself, leveraging in-context learning on different few-shot examples. To further calibrate this set of criteria, we select the best performers and re-draft them with self-refinement. Our experiments on multiple text quality evaluation datasets illustrate a significant improvement in correlation with expert evaluation through calibration. Our comprehensive qualitative analysis conveys insightful intuitions and observations on the essence of effective scoring criteria."
                    },
                    {
                        "title": "BitNet: Scaling 1-bit Transformers for Large Language Models",
                        "abstract": "The increasing size of large language models has posed challenges for deployment and raised concerns about environmental impact due to high energy consumption. In this work, we introduce BitNet, a scalable and stable 1-bit Transformer architecture designed for large language models. Specifically, we introduce BitLinear as a drop-in replacement of the nn.Linear layer in order to train 1-bit weights from scratch. Experimental results on language modeling show that BitNet achieves competitive performance while substantially reducing memory footprint and energy consumption, compared to state-of-the-art 8-bit quantization methods and FP16 Transformer baselines. Furthermore, BitNet exhibits a scaling law akin to full-precision Transformers, suggesting its potential for effective scaling to even larger language models while maintaining efficiency and performance benefits."
                    },
                    {
                        "title": "Kosmos-2: Grounding Multimodal Large Language Models to the World",
                        "abstract": "We introduce Kosmos-2, a Multimodal Large Language Model (MLLM), enabling new capabilities of perceiving object descriptions (e.g., bounding boxes) and grounding text to the visual world. Specifically, we represent refer expressions as links in Markdown, i.e., ``[text span](bounding boxes)'', where object descriptions are sequences of location tokens. Together with multimodal corpora, we construct large-scale data of grounded image-text pairs (called GrIT) to train the model. In addition to the existing capabilities of MLLMs (e.g., perceiving general modalities, following instructions, and performing in-context learning), Kosmos-2 integrates the grounding capability into downstream applications. We evaluate Kosmos-2 on a wide range of tasks, including (i) multimodal grounding, such as referring expression comprehension, and phrase grounding, (ii) multimodal referring, such as referring expression generation, (iii) perception-language tasks, and (iv) language understanding and generation. This work lays out the foundation for the development of Embodiment AI and sheds light on the big convergence of language, multimodal perception, action, and world modeling, which is a key step toward artificial general intelligence. Code and pretrained models are available at https://aka.ms/kosmos-2."
                    },
                    {
                        "title": "Kosmos-G: Generating Images in Context with Multimodal Large Language Models",
                        "abstract": "Recent advancements in subject-driven image generation have made significant strides. However, current methods still fall short in diverse application scenarios, as they require test-time tuning and cannot accept interleaved multi-image and text input. These limitations keep them far from the ultimate goal of\"image as a foreign language in image generation.\"This paper presents Kosmos-G, a model that leverages the advanced multimodal perception capabilities of Multimodal Large Language Models (MLLMs) to tackle the aforementioned challenge. Our approach aligns the output space of MLLM with CLIP using the textual modality as an anchor and performs compositional instruction tuning on curated data. Kosmos-G demonstrates an impressive capability of zero-shot subject-driven generation with interleaved multi-image and text input. Notably, the score distillation instruction tuning requires no modifications to the image decoder. This allows for a seamless substitution of CLIP and effortless integration with a myriad of U-Net techniques ranging from fine-grained controls to personalized image decoder variants. We posit Kosmos-G as an initial attempt towards the goal of\"image as a foreign language in image generation.\"The code can be found at https://aka.ms/Kosmos-G"
                    }
                ]
            },
            "7f621808-cf6c-46b4-a2c1-c79ad4438851": {
                "pk": "7f621808-cf6c-46b4-a2c1-c79ad4438851",
                "name": "Furu Wei",
                "collaborators": [
                    "Shaohan Huang",
                    "Li Dong",
                    "Weiwei Deng",
                    "Qi Zhang",
                    "Zihan Zhang",
                    "Haizhen Huang",
                    "Feng Sun",
                    "Yuxian Gu",
                    "Shuming Ma",
                    "Yuxuan Liu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Knowledge Distillation",
                    "Instruction Tuning"
                ],
                "publications": [
                    {
                        "title": "Text Diffusion with Reinforced Conditioning",
                        "abstract": "Diffusion models have demonstrated exceptional capability in generating high-quality images, videos, and audio. Due to their adaptiveness in iterative refinement, they provide a strong potential for achieving better non-autoregressive sequence generation. However, existing text diffusion models still fall short in their performance due to a challenge in handling the discreteness of language. This paper thoroughly analyzes text diffusion models and uncovers two significant limitations: degradation of self-conditioning during training and misalignment between training and sampling. Motivated by our findings, we propose a novel Text Diffusion model called TReC, which mitigates the degradation with Reinforced Conditioning and the misalignment by Time-Aware Variance Scaling. Our extensive experiments demonstrate the competitiveness of TReC against autoregressive, non-autoregressive, and diffusion baselines. Moreover, qualitative analysis shows its advanced ability to fully utilize the diffusion process in refining samples."
                    },
                    {
                        "title": "You Only Cache Once: Decoder-Decoder Architectures for Language Models",
                        "abstract": "We introduce a decoder-decoder architecture, YOCO, for large language models, which only caches key-value pairs once. It consists of two components, i.e., a cross-decoder stacked upon a self-decoder. The self-decoder efficiently encodes global key-value (KV) caches that are reused by the cross-decoder via cross-attention. The overall model behaves like a decoder-only Transformer, although YOCO only caches once. The design substantially reduces GPU memory demands, yet retains global attention capability. Additionally, the computation flow enables prefilling to early exit without changing the final output, thereby significantly speeding up the prefill stage. Experimental results demonstrate that YOCO achieves favorable performance compared to Transformer in various settings of scaling up model size and number of training tokens. We also extend YOCO to 1M context length with near-perfect needle retrieval accuracy. The profiling results show that YOCO improves inference memory, prefill latency, and throughput by orders of magnitude across context lengths and model sizes. Code is available at https://aka.ms/YOCO."
                    },
                    {
                        "title": "Direct Preference Knowledge Distillation for Large Language Models",
                        "abstract": "In the field of large language models (LLMs), Knowledge Distillation (KD) is a critical technique for transferring capabilities from teacher models to student models. However, existing KD methods face limitations and challenges in distillation of LLMs, including efficiency and insufficient measurement capabilities of traditional KL divergence. It is shown that LLMs can serve as an implicit reward function, which we define as a supplement to KL divergence. In this work, we propose Direct Preference Knowledge Distillation (DPKD) for LLMs. DPKD utilizes distribution divergence to represent the preference loss and implicit reward function. We re-formulate KD of LLMs into two stages: first optimizing and objective consisting of implicit reward and reverse KL divergence and then improving the preference probability of teacher outputs over student outputs. We conducted experiments and analysis on various datasets with LLM parameters ranging from 120M to 13B and demonstrate the broad applicability and effectiveness of our DPKD approach. Meanwhile, we prove the value and effectiveness of the introduced implicit reward and output preference in KD through experiments and theoretical analysis. The DPKD method outperforms the baseline method in both output response precision and exact match percentage. Code and data are available at https://aka.ms/dpkd."
                    },
                    {
                        "title": "Adapting Large Language Models via Reading Comprehension",
                        "abstract": "We explore how continued pre-training on domain-specific corpora influences large language models, revealing that training on the raw corpora endows the model with domain knowledge, but drastically hurts its prompting ability for question answering. Taken inspiration from human learning via reading comprehension\u2014practice after reading improves the ability to answer questions based on the learned knowledge\u2014we propose a simple method for transforming raw corpora into reading comprehension texts. Each raw text is enriched with a series of tasks related to its content. Our method, highly scalable and applicable to any pre-training corpora, consistently enhances performance across various tasks in three different domains: biomedicine, finance"
                    },
                    {
                        "title": "ResLoRA: Identity Residual Mapping in Low-Rank Adaption",
                        "abstract": "As one of the most popular parameter-efficient fine-tuning (PEFT) methods, low-rank adaptation (LoRA) is commonly applied to fine-tune large language models (LLMs). However, updating the weights of LoRA blocks effectively and expeditiously is challenging due to the long calculation path in the original model. To address this, we propose ResLoRA, an improved framework of LoRA. By adding residual paths during training and using merging approaches to eliminate these extra paths during inference, our method can achieve better results in fewer training steps without any extra trainable parameters or inference cost compared to LoRA. The experiments on NLG, NLU, and text-to-image tasks demonstrate the effectiveness of our method. To the best of our knowledge, ResLoRA is the first work that combines the residual path with LoRA. The code of our method is available at https://github.com/microsoft/LMOps/tree/main/reslora ."
                    },
                    {
                        "title": "Synthetic Data (Almost) from Scratch: Generalized Instruction Tuning for Language Models",
                        "abstract": "We introduce Generalized Instruction Tuning (called GLAN), a general and scalable method for instruction tuning of Large Language Models (LLMs). Unlike prior work that relies on seed examples or existing datasets to construct instruction tuning data, GLAN exclusively utilizes a pre-curated taxonomy of human knowledge and capabilities as input and generates large-scale synthetic instruction data across all disciplines. Specifically, inspired by the systematic structure in human education system, we build the taxonomy by decomposing human knowledge and capabilities to various fields, sub-fields and ultimately, distinct disciplines semi-automatically, facilitated by LLMs. Subsequently, we generate a comprehensive list of subjects for every discipline and proceed to design a syllabus tailored to each subject, again utilizing LLMs. With the fine-grained key concepts detailed in every class session of the syllabus, we are able to generate diverse instructions with a broad coverage across the entire spectrum of human knowledge and skills. Extensive experiments on large language models (e.g., Mistral) demonstrate that GLAN excels in multiple dimensions from mathematical reasoning, coding, academic exams, logical reasoning to general instruction following without using task-specific training data of these tasks. In addition, GLAN allows for easy customization and new fields or skills can be added by simply incorporating a new node into our taxonomy."
                    },
                    {
                        "title": "Instruction Pre-Training: Language Models are Supervised Multitask Learners",
                        "abstract": "Unsupervised multitask pre-training has been the critical method behind the recent success of language models (LMs). However, supervised multitask learning still holds significant promise, as scaling it in the post-training stage trends towards better generalization. In this paper, we explore supervised multitask pre-training by proposing Instruction Pre-Training, a framework that scalably augments massive raw corpora with instruction-response pairs to pre-train LMs. The instruction-response pairs are generated by an efficient instruction synthesizer built on open-source models. In our experiments, we synthesize 200M instruction-response pairs covering 40+ task categories to verify the effectiveness of Instruction Pre-Training. In pre-training from scratch, Instruction Pre-Training not only consistently enhances pre-trained base models but also benefits more from further instruction tuning. In continual pre-training, Instruction Pre-Training enables Llama3-8B to be comparable to or even outperform Llama3-70B. Our model, code, and data are available at https://github.com/microsoft/LMOps."
                    },
                    {
                        "title": "MoRA: High-Rank Updating for Parameter-Efficient Fine-Tuning",
                        "abstract": "Low-rank adaptation is a popular parameter-efficient fine-tuning method for large language models. In this paper, we analyze the impact of low-rank updating, as implemented in LoRA. Our findings suggest that the low-rank updating mechanism may limit the ability of LLMs to effectively learn and memorize new knowledge. Inspired by this observation, we propose a new method called MoRA, which employs a square matrix to achieve high-rank updating while maintaining the same number of trainable parameters. To achieve it, we introduce the corresponding non-parameter operators to reduce the input dimension and increase the output dimension for the square matrix. Furthermore, these operators ensure that the weight can be merged back into LLMs, which makes our method can be deployed like LoRA. We perform a comprehensive evaluation of our method across five tasks: instruction tuning, mathematical reasoning, continual pretraining, memory and pretraining. Our method outperforms LoRA on memory-intensive tasks and achieves comparable performance on other tasks."
                    },
                    {
                        "title": "HD-Eval: Aligning Large Language Model Evaluators Through Hierarchical Criteria Decomposition",
                        "abstract": "Large language models (LLMs) have emerged as a promising alternative to expensive human evaluations. However, the alignment and coverage of LLM-based evaluations are often limited by the scope and potential bias of the evaluation prompts and criteria. To address this challenge, we propose HD-Eval, a novel framework that iteratively aligns LLM-based evaluators with human preference via Hierarchical Criteria Decomposition. HD-Eval inherits the essence from the evaluation mindset of human experts and enhances the alignment of LLM-based evaluators by decomposing a given evaluation task into finer-grained criteria, aggregating them according to estimated human preferences, pruning insignificant criteria with attribution, and further decomposing significant criteria. By integrating these steps within an iterative alignment training process, we obtain a hierarchical decomposition of criteria that comprehensively captures aspects of natural language at multiple levels of granularity. Implemented as a white box, the human preference-guided aggregator is efficient to train and more explainable than relying solely on prompting, and its independence from model parameters makes it applicable to closed-source LLMs. Extensive experiments on three evaluation domains demonstrate the superiority of HD-Eval in further aligning state-of-the-art evaluators and providing deeper insights into the explanation of evaluation results and the task itself."
                    },
                    {
                        "title": "Se2: Sequential Example Selection for In-Context Learning",
                        "abstract": "The remarkable capability of large language models (LLMs) for in-context learning (ICL) needs to be activated by demonstration examples. Prior work has extensively explored the selection of examples for ICL, predominantly following the\"select then organize\"paradigm, such approaches often neglect the internal relationships between examples and exist an inconsistency between the training and inference. In this paper, we formulate the problem as a $Se$quential $Se$lection problem and introduce $Se^2$, a sequential-aware method that leverages the LLM's feedback on varying context, aiding in capturing inter-relationships and sequential information among examples, significantly enriching the contextuality and relevance of ICL prompts. Meanwhile, we utilize beam search to seek and construct example sequences, enhancing both quality and diversity. Extensive experiments across 23 NLP tasks from 8 distinct categories illustrate that $Se^2$ markedly surpasses competitive baselines and achieves 42\\% relative improvement over random selection. Further in-depth analysis shows the effectiveness of proposed strategies, highlighting $Se^2$'s exceptional stability and adaptability across various scenarios. Code available at https://github.com/microsoft/LMOps."
                    },
                    {
                        "title": "Improving Domain Adaptation through Extended-Text Reading Comprehension",
                        "abstract": "To enhance the domain-specific capabilities of large language models, continued pre-training on a domain-specific corpus is a prevalent method. Recent work demonstrates that adapting models using reading comprehension data formatted by regex-based patterns can significantly improve performance on domain-specific tasks. However, regex-based patterns are incapable of parsing raw corpora using domain-specific knowledge. Furthermore, the question and answer pairs are extracted directly from the corpus in predefined formats offers limited context. To address this limitation, we improve reading comprehension via LLM and clustering. LLM focuses on leveraging domain knowledge within the corpus to refine comprehension stage, while clustering supplies relevant knowledge by extending the context to enrich reading stage. Additionally, our method incorporates parameter-efficient fine-tuning to improve the efficiency of domain adaptation. In comparison to AdaptLLM, our method achieves an improvement exceeding 5% in domain-specific tasks. Our code will available at https://github.com/microsoft/LMOps."
                    },
                    {
                        "title": "Towards Optimal Learning of Language Models",
                        "abstract": "This work studies the general principles of improving the learning of language models (LMs), which aims at reducing the necessary training steps for achieving superior performance. Specifically, we present a theory for the optimal learning of LMs. We first propose an objective that optimizes LM learning by maximizing the data compression ratio in an\"LM-training-as-lossless-compression\"view. Then, we derive a theorem, named Learning Law, to reveal the properties of the dynamics in the optimal learning process under our objective. The theorem is then validated by experiments on a linear classification and a real-world language modeling task. Finally, we empirically verify that the optimal learning of LMs essentially stems from the improvement of the coefficients in the scaling law of LMs, indicating great promise and significance for designing practical learning acceleration methods. Our code can be found at https://aka.ms/LearningLaw."
                    },
                    {
                        "title": "Multi-Head Mixture-of-Experts",
                        "abstract": "Sparse Mixtures of Experts (SMoE) scales model capacity without significant increases in training and inference costs, but exhibits the following two issues: (1) Low expert activation, where only a small subset of experts are activated for optimization. (2) Lacking fine-grained analytical capabilities for multiple semantic concepts within individual tokens. We propose Multi-Head Mixture-of-Experts (MH-MoE), which employs a multi-head mechanism to split each token into multiple sub-tokens. These sub-tokens are then assigned to and processed by a diverse set of experts in parallel, and seamlessly reintegrated into the original token form. The multi-head mechanism enables the model to collectively attend to information from various representation spaces within different experts, while significantly enhances expert activation, thus deepens context understanding and alleviate overfitting. Moreover, our MH-MoE is straightforward to implement and decouples from other SMoE optimization methods, making it easy to integrate with other SMoE models for enhanced performance. Extensive experimental results across three tasks: English-focused language modeling, Multi-lingual language modeling and Masked multi-modality modeling tasks, demonstrate the effectiveness of MH-MoE."
                    },
                    {
                        "title": "Semi-Parametric Retrieval via Binary Token Index",
                        "abstract": "The landscape of information retrieval has broadened from search services to a critical component in various advanced applications, where indexing efficiency, cost-effectiveness, and freshness are increasingly important yet remain less explored. To address these demands, we introduce Semi-parametric Vocabulary Disentangled Retrieval (SVDR). SVDR is a novel semi-parametric retrieval framework that supports two types of indexes: an embedding-based index for high effectiveness, akin to existing neural retrieval methods; and a binary token index that allows for quick and cost-effective setup, resembling traditional term-based retrieval. In our evaluation on three open-domain question answering benchmarks with the entire Wikipedia as the retrieval corpus, SVDR consistently demonstrates superiority. It achieves a 3% higher top-1 retrieval accuracy compared to the dense retriever DPR when using an embedding-based index and an 9% higher top-1 accuracy compared to BM25 when using a binary token index. Specifically, the adoption of a binary token index reduces index preparation time from 30 GPU hours to just 2 CPU hours and storage size from 31 GB to 2 GB, achieving a 90% reduction compared to an embedding-based index."
                    },
                    {
                        "title": "Multimodal Large Language Model is a Human-Aligned Annotator for Text-to-Image Generation",
                        "abstract": "Recent studies have demonstrated the exceptional potentials of leveraging human preference datasets to refine text-to-image generative models, enhancing the alignment between generated images and textual prompts. Despite these advances, current human preference datasets are either prohibitively expensive to construct or suffer from a lack of diversity in preference dimensions, resulting in limited applicability for instruction tuning in open-source text-to-image generative models and hinder further exploration. To address these challenges and promote the alignment of generative models through instruction tuning, we leverage multimodal large language models to create VisionPrefer, a high-quality and fine-grained preference dataset that captures multiple preference aspects. We aggregate feedback from AI annotators across four aspects: prompt-following, aesthetic, fidelity, and harmlessness to construct VisionPrefer. To validate the effectiveness of VisionPrefer, we train a reward model VP-Score over VisionPrefer to guide the training of text-to-image generative models and the preference prediction accuracy of VP-Score is comparable to human annotators. Furthermore, we use two reinforcement learning methods to supervised fine-tune generative models to evaluate the performance of VisionPrefer, and extensive experimental results demonstrate that VisionPrefer significantly improves text-image alignment in compositional image generation across diverse aspects, e.g., aesthetic, and generalizes better than previous human-preference metrics across various image distributions. Moreover, VisionPrefer indicates that the integration of AI-generated synthetic data as a supervisory signal is a promising avenue for achieving improved alignment with human preferences in vision generative models."
                    },
                    {
                        "title": "BitNet: Scaling 1-bit Transformers for Large Language Models",
                        "abstract": "The increasing size of large language models has posed challenges for deployment and raised concerns about environmental impact due to high energy consumption. In this work, we introduce BitNet, a scalable and stable 1-bit Transformer architecture designed for large language models. Specifically, we introduce BitLinear as a drop-in replacement of the nn.Linear layer in order to train 1-bit weights from scratch. Experimental results on language modeling show that BitNet achieves competitive performance while substantially reducing memory footprint and energy consumption, compared to state-of-the-art 8-bit quantization methods and FP16 Transformer baselines. Furthermore, BitNet exhibits a scaling law akin to full-precision Transformers, suggesting its potential for effective scaling to even larger language models while maintaining efficiency and performance benefits."
                    },
                    {
                        "title": "Tuna: Instruction Tuning using Feedback from Large Language Models",
                        "abstract": "Instruction tuning of open-source large language models (LLMs) like LLaMA, using direct outputs from more powerful LLMs such as Instruct-GPT and GPT-4, has proven to be a cost-effective way to align model behaviors with human preferences. However, the instruction-tuned model has only seen one response per instruction, lacking the knowledge of potentially better responses. In this paper, we propose finetuning an instruction-tuned LLM using our novel \\textit{probabilistic ranking} and \\textit{contextual ranking} approaches to increase the likelihood of generating better responses. Probabilistic ranking enables the instruction-tuned model to inherit the relative rankings of high-quality and low-quality responses from the teacher LLM. On the other hand, learning with contextual ranking allows the model to refine its own response distribution using the contextual understanding ability of stronger LLMs. Furthermore, we apply probabilistic ranking and contextual ranking sequentially to the instruction-tuned LLM. The resulting model, which we call \\textbf{Tuna}, consistently improves the performance on Super Natural Instructions (119 test tasks), LMentry (25 test tasks), Vicuna QA, and can even obtain better results than several strong reinforcement learning baselines. Our code and data are available at \\url{ https://github.com/microsoft/LMOps}."
                    },
                    {
                        "title": "Kosmos-G: Generating Images in Context with Multimodal Large Language Models",
                        "abstract": "Recent advancements in subject-driven image generation have made significant strides. However, current methods still fall short in diverse application scenarios, as they require test-time tuning and cannot accept interleaved multi-image and text input. These limitations keep them far from the ultimate goal of\"image as a foreign language in image generation.\"This paper presents Kosmos-G, a model that leverages the advanced multimodal perception capabilities of Multimodal Large Language Models (MLLMs) to tackle the aforementioned challenge. Our approach aligns the output space of MLLM with CLIP using the textual modality as an anchor and performs compositional instruction tuning on curated data. Kosmos-G demonstrates an impressive capability of zero-shot subject-driven generation with interleaved multi-image and text input. Notably, the score distillation instruction tuning requires no modifications to the image decoder. This allows for a seamless substitution of CLIP and effortless integration with a myriad of U-Net techniques ranging from fine-grained controls to personalized image decoder variants. We posit Kosmos-G as an initial attempt towards the goal of\"image as a foreign language in image generation.\"The code can be found at https://aka.ms/Kosmos-G"
                    },
                    {
                        "title": "Auto-ICL: In-Context Learning without Human Supervision",
                        "abstract": "With in-context learning ability, the performance of large language models can be significantly boosted when provided with appropriate context. However, existing in-context learning methods mainly rely on human-provided contexts, such as labeled examples and explicit instructions. Writing context by humans is labor-intensive on various tasks and limits the model to tasks manageable by humans. To overcome these limitations, we propose Automatic In-Context Learning framework that enables the model to autonomously generate examples and instructions for problem-solving. With experiments across various models and datasets, results show that model-generated contexts outperform human-annotated contexts, including Few-Shot and Few-Shot-CoT methods, and surpass existing self-generated context methods like Zero-CoT and Auto-CoT."
                    }
                ]
            }
        }
    },
    "2402.12550": {
        "paper_data": {
            "title": "Multilinear Mixture of Experts: Scalable Expert Specialization through Factorization",
            "url": "http://arxiv.org/abs/2402.12550v4",
            "arxiv_id": "2402.12550",
            "authors": [
                "James Oldfield",
                "Markos Georgopoulos",
                "Grigorios G. Chrysos",
                "Christos Tzelepis",
                "Yannis Panagakis",
                "Mihalis A. Nicolaou",
                "Jiankang Deng",
                "Ioannis Patras"
            ],
            "abstract": "The Mixture of Experts (MoE) paradigm provides a powerful way to decompose dense layers into smaller, modular computations often more amenable to human interpretation, debugging, and editability. However, a major challenge lies in the computational cost of scaling the number of experts high enough to achieve fine-grained specialization. In this paper, we propose the Multilinear Mixture of Experts ($\\mu$MoE) layer to address this, focusing on vision models. $\\mu$MoE layers enable scalable expert specialization by performing an implicit computation on prohibitively large weight tensors entirely in factorized form. Consequently, $\\mu$MoEs (1) avoid the restrictively high inference-time costs of dense MoEs, yet (2) do not inherit the training issues of the popular sparse MoEs' discrete (non-differentiable) expert routing. We present both qualitative and quantitative evidence that scaling $\\mu$MoE layers when fine-tuning foundation models for vision tasks leads to more specialized experts at the class-level, further enabling manual bias correction in CelebA attribute classification. Finally, we show qualitative results demonstrating the expert specialism achieved when pre-training large GPT2 and MLP-Mixer models with parameter-matched $\\mu$MoE blocks at every layer, maintaining comparable accuracy. Our code is available at: https://github.com/james-oldfield/muMoE.",
            "introduction": "   1 Introduction  The Mixture of Experts (MoE) architecture [1] has reemerged as a powerful class of conditional computation, playing the pivotal role in scaling up recent large language [2, 3, 4, 5], vision [6], and multi-modal models [7]. MoEs apply different subsets of layers (referred to as \u2018experts\u2019) for each input, in contrast to the traditional approach of using the same single layer for all inputs. This provides a form of input-conditional computation [8, 9, 10, 11] that is expressive yet efficient. However, through their substantial performance gains, an important emergent property of MoEs is frequently underutilized: the innate tendency of experts to specialize in distinct subtasks. Indeed, the foundational work of Jacobs et\u00a0al. [12] on MoEs describes this property, highlighting how implementing a particular function with modular building blocks (experts) often leads to subcomputations that are easier to understand individually than their dense layer counterparts\u2013with larger expert counts allowing for more fine-grained specialization.   Independent of model performance, a successful decomposition of the layer\u2019s functionality into human-comprehensible subtasks offers many significant benefits. Firstly, the mechanisms through which a network produces an output are more interpretable: the output is a sum of modular components, each contributing individual functionality. Yet, the value of interpretable computation extends beyond just transparency [13] and explainability [14]. An important corollary of successful task decomposition amongst experts is that layers are easier to debug and edit. Biased or unsafe behaviors can be better localized to specific experts\u2019 subcomputation, facilitating manual correction or surgery in a way that minimally affects the other functionality of the network. Addressing such behaviors is particularly crucial in the context of foundation models; being often fine-tuned as black boxes pre-trained on unknown, potentially imbalanced data distributions. Furthermore, there is evidence that traditional fairness techniques are less effective in large-scale models [15, 16]. However, to achieve fine-grained expert specialism at the class level (or more granular still), one needs the ability to significantly scale up the number of experts. When using only a small expert count, each expert is forced to process and generalize across multiple distinct semantic concepts, hindering specialization. Conversely, a large expert count means each can specialize to a more specific set of semantically similar inputs. Alas, the dominating \u2018sparse\u2019 MoE paradigm of selecting only the top-K\ud835\udc3eKitalic_K experts [17] is not only parameter-inefficient for large expert counts, but also has several well-known issues due to its discrete expert routing\u2013often leading to training instability and difficulties in scaling the total expert count, amongst other challenges [18, 19].      Table 1: Benefits of the proposed \u03bc\ud835\udf07\\muitalic_\u03bcMoEs\u2019 model form over existing MoEs.          Parameter- FLOPs-    Differentiable efficient efficient     Dense MoE [1]  \\faSmileO \\faFrownO \\faFrownO   Sparse MoE [17]  \\faFrownO \\faFrownO \\faSmileO    \ud835\udf41\ud835\udf41\\bm{\\mu}bold_italic_\u03bcMoE (ours)  \\faSmileO \\faSmileO \\faSmileO         In this paper, we propose the Multilinear Mixture of Experts (\u03bc\ud835\udf07\\muitalic_\u03bcMoE) layer to address these issues. \u03bc\ud835\udf07\\muitalic_\u03bcMoEs are designed to scale gracefully to dense operations involving tens of thousands of experts at once through implicit computations on a factorized form of the experts\u2019 weights. Furthermore, in contrast to the dominant sparse MoEs\u2019 [17] non-differentiable nature, \u03bc\ud835\udf07\\muitalic_\u03bcMoEs are differentiable by design, and thus do not inherit the associated training issues. We summarize the benefits of \u03bc\ud835\udf07\\muitalic_\u03bcMoEs\u2019 model form over existing MoEs in Section\u00a01. Crucially, we show evidence that scaling up the number of \u03bc\ud835\udf07\\muitalic_\u03bcMoE experts leads to increased expert",
            "references": [
                {
                    "title": "Weight-based Decomposition: A Case for Bilinear MLPs",
                    "abstract": "Gated Linear Units (GLUs) have become a common building block in modern foundation models. Bilinear layers drop the non-linearity in the\"gate\"but still have comparable performance to other GLUs. An attractive quality of bilinear layers is that they can be fully expressed in terms of a third-order tensor and linear operations. Leveraging this, we develop a method to decompose the bilinear tensor into a set of sparsely interacting eigenvectors that show promising interpretability properties in preliminary experiments for shallow image classifiers (MNIST) and small language models (Tiny Stories). Since the decomposition is fully equivalent to the model's original computations, bilinear layers may be an interpretability-friendly architecture that helps connect features to the model weights. Application of our method may not be limited to pretrained bilinear models since we find that language models such as TinyLlama-1.1B can be finetuned into bilinear variants."
                },
                {
                    "title": "The Truth is in There: Improving Reasoning in Language Models with Layer-Selective Rank Reduction",
                    "abstract": "Transformer-based Large Language Models (LLMs) have become a fixture in modern machine learning. Correspondingly, significant resources are allocated towards research that aims to further advance this technology, typically resulting in models of increasing size that are trained on increasing amounts of data. This work, however, demonstrates the surprising result that it is often possible to significantly improve the performance of LLMs by selectively removing higher-order components of their weight matrices. This simple intervention, which we call LAyer-SElective Rank reduction (LASER), can be done on a model after training has completed, and requires no additional parameters or data. We show extensive experiments demonstrating the generality of this finding across language models and datasets, and provide in-depth analyses offering insights into both when LASER is effective and the mechanism by which it operates."
                },
                {
                    "title": "Factorized Dynamic Fully-Connected Layers for Neural Networks",
                    "abstract": "The design of neural network layers plays a crucial role in determining the efficiency and performance of various computer vision tasks. However, most existing layers compromise between fast feature extraction and reasoning abilities, resulting in suboptimal outcomes. In this paper, we propose a novel and efficient operator for representation learning that can dynamically adjust to the underlying data structure. We introduce a general Dynamic Fully-Connected (DFC) layer, a non-linear extension of a Fully-Connected layer that has a learnable receptive field, is instance-adaptive, and spatially aware. We propose to use CP decomposition to reduce the complexity of the DFC layer without compromising its expressivity. Then, we leverage Summed Area Tables and Modulation to create an adaptive receptive field that can process the input with constant complexity. We evaluate the effectiveness of our method on image classification and other downstream vision tasks using both hierarchical and isotropic architectures. Our results demonstrate that our method outperforms other commonly used layers by a significant margin while keeping a fixed computational budget, therefore establishing a new strategy to efficiently design neural architectures that can capture the multi-scale features of the input without increasing complexity."
                },
                {
                    "title": "From Sparse to Soft Mixtures of Experts",
                    "abstract": "Sparse mixture of expert architectures (MoEs) scale model capacity without significant increases in training or inference costs. Despite their success, MoEs suffer from a number of issues: training instability, token dropping, inability to scale the number of experts, or ineffective finetuning. In this work, we propose Soft MoE, a fully-differentiable sparse Transformer that addresses these challenges, while maintaining the benefits of MoEs. Soft MoE performs an implicit soft assignment by passing different weighted combinations of all input tokens to each expert. As in other MoEs, experts in Soft MoE only process a subset of the (combined) tokens, enabling larger model capacity (and performance) at lower inference cost. In the context of visual recognition, Soft MoE greatly outperforms dense Transformers (ViTs) and popular MoEs (Tokens Choice and Experts Choice). Furthermore, Soft MoE scales well: Soft MoE Huge/14 with 128 experts in 16 MoE layers has over 40x more parameters than ViT Huge/14, with only 2% increased inference time, and substantially better quality."
                },
                {
                    "title": "Task Arithmetic in the Tangent Space: Improved Editing of Pre-Trained Models",
                    "abstract": "Task arithmetic has recently emerged as a cost-effective and scalable approach to edit pre-trained models directly in weight space: By adding the fine-tuned weights of different tasks, the model's performance can be improved on these tasks, while negating them leads to task forgetting. Yet, our understanding of the effectiveness of task arithmetic and its underlying principles remains limited. We present a comprehensive study of task arithmetic in vision-language models and show that weight disentanglement is the crucial factor that makes it effective. This property arises during pre-training and manifests when distinct directions in weight space govern separate, localized regions in function space associated with the tasks. Notably, we show that fine-tuning models in their tangent space by linearizing them amplifies weight disentanglement. This leads to substantial performance improvements across multiple task arithmetic benchmarks and diverse models. Building on these findings, we provide theoretical and empirical analyses of the neural tangent kernel (NTK) of these models and establish a compelling link between task arithmetic and the spatial localization of the NTK eigenfunctions. Overall, our work uncovers novel insights into the fundamental mechanisms of task arithmetic and offers a more reliable and effective approach to edit pre-trained models through the NTK linearization."
                },
                {
                    "title": "A technical note on bilinear layers for interpretability",
                    "abstract": "The ability of neural networks to represent more features than neurons makes interpreting them challenging. This phenomenon, known as superposition, has spurred efforts to find architectures that are more interpretable than standard multilayer perceptrons (MLPs) with elementwise activation functions. In this note, I examine bilinear layers, which are a type of MLP layer that are mathematically much easier to analyze while simultaneously performing better than standard MLPs. Although they are nonlinear functions of their input, I demonstrate that bilinear layers can be expressed using only linear operations and third order tensors. We can integrate this expression for bilinear layers into a mathematical framework for transformer circuits, which was previously limited to attention-only transformers. These results suggest that bilinear layers are easier to analyze mathematically than current architectures and thus may lend themselves to deeper safety insights by allowing us to talk more formally about circuits in neural networks. Additionally, bilinear layers may offer an alternative path for mechanistic interpretability through understanding the mechanisms of feature construction instead of enumerating a (potentially exponentially) large number of features in large models."
                },
                {
                    "title": "Last-Layer Fairness Fine-tuning is Simple and Effective for Neural Networks",
                    "abstract": "As machine learning has been deployed ubiquitously across applications in modern data science, algorithmic fairness has become a great concern. Among them, imposing fairness constraints during learning, i.e. in-processing fair training, has been a popular type of training method because they don't require accessing sensitive attributes during test time in contrast to post-processing methods. While this has been extensively studied in classical machine learning models, their impact on deep neural networks remains unclear. Recent research has shown that adding fairness constraints to the objective function leads to severe over-fitting to fairness criteria in large models, and how to solve this challenge is an important open question. To tackle this, we leverage the wisdom and power of pre-training and fine-tuning and develop a simple but novel framework to train fair neural networks in an efficient and inexpensive way -- last-layer fine-tuning alone can effectively promote fairness in deep neural networks. This framework offers valuable insights into representation learning for training fair neural networks."
                },
                {
                    "title": "Editing Models with Task Arithmetic",
                    "abstract": "Changing how pre-trained models behave -- e.g., improving their performance on a downstream task or mitigating biases learned during pre-training -- is a common practice when developing machine learning systems. In this work, we propose a new paradigm for steering the behavior of neural networks, centered around \\textit{task vectors}. A task vector specifies a direction in the weight space of a pre-trained model, such that movement in that direction improves performance on the task. We build task vectors by subtracting the weights of a pre-trained model from the weights of the same model after fine-tuning on a task. We show that these task vectors can be modified and combined together through arithmetic operations such as negation and addition, and the behavior of the resulting model is steered accordingly. Negating a task vector decreases performance on the target task, with little change in model behavior on control tasks. Moreover, adding task vectors together can improve performance on multiple tasks at once. Finally, when tasks are linked by an analogy relationship of the form ``A is to B as C is to D\", combining task vectors from three of the tasks can improve performance on the fourth, even when no data from the fourth task is used for training. Overall, our experiments with several models, modalities and tasks show that task arithmetic is a simple, efficient and effective way of editing models."
                },
                {
                    "title": "MegaBlocks: Efficient Sparse Training with Mixture-of-Experts",
                    "abstract": "We present MegaBlocks, a system for efficient Mixture-of-Experts (MoE) training on GPUs. Our system is motivated by the limitations of current frameworks, which restrict the dynamic routing in MoE layers to satisfy the constraints of existing software and hardware. These formulations force a tradeoff between model quality and hardware efficiency, as users must choose between dropping tokens from the computation or wasting computation and memory on padding. To address these limitations, we reformulate MoE computation in terms of block-sparse operations and develop new block-sparse GPU kernels that efficiently handle the dynamism present in MoEs. Our approach never drops tokens and maps efficiently to modern hardware, enabling end-to-end training speedups of up to 40% over MoEs trained with the state-of-the-art Tutel library and 2.4x over DNNs trained with the highly-optimized Megatron-LM framework."
                },
                {
                    "title": "Toy Models of Superposition",
                    "abstract": "Neural networks often pack many unrelated concepts into a single neuron \u2013 a puzzling phenomenon known as 'polysemanticity' which makes interpretability much more challenging. This paper provides a toy model where polysemanticity can be fully understood, arising as a result of models storing additional sparse features in \"superposition.\" We demonstrate the existence of a phase change, a surprising connection to the geometry of uniform polytopes, and evidence of a link to adversarial examples. We also discuss potential implications for mechanistic interpretability. We recommend reading this paper as an HTML article. It would be very convenient if the individual neurons of artificial neural networks corresponded to cleanly interpretable features of the input. For example, in an \u201cideal\u201d ImageNet classifier, each neuron would fire only in the presence of a specific visual feature, such as the color red, a leftfacing curve, or a dog snout. Empirically, in models we have studied, some of the neurons do cleanly map to features. But it isn't always the case that features correspond so cleanly to neurons, especially in large language models where it actually seems rare for neurons to correspond to clean features. This brings up many questions. Why is it that neurons sometimes align with features and sometimes don't? Why do some models and tasks have many of these clean neurons, while they're vanishingly rare in others? In this paper, we use toy models \u2014 small ReLU networks trained on synthetic data with sparse input features \u2014 to investigate how and when models represent more features than they have dimensions. We call this phenomenon superposition. When features are sparse, superposition allows compression beyond what a linear model would do, at the cost of \"interference\" that requires nonlinear filtering. Consider a toy model where we train an embedding of five features of varying importance in two dimensions, add a ReLU afterwards for filtering, and vary the sparsity of the features. With dense features, the model learns to represent an orthogonal basis of the most important two features (similar to what Principal Component Analysis might give us), and the other three features are not represented. But if we make the features sparse, this changes: 1 This figure and a few others can be reproduced using the toy model framework Colab notebook in our Github repo Not only can models store additional features in superposition by tolerating some interference, but we'll show that, at least in certain limited cases, models can perform computation while in superposition. (In particular, we'll show that models can put simple circuits computing the absolute value function in superposition.) This leads us to hypothesize that the neural networks we observe in practice are in some sense noisily simulating larger, highly sparse networks. In other words, it's possible that models we train can be thought of as doing \u201cthe same thing as\u201d an imagined muchlarger model, representing the exact same features but with no interference. Feature superposition isn't a novel idea. A number of previous interpretability papers have speculated about it , and it's very closely related to the long-studied topic of compressed sensing in mathematics , as well as the ideas of distributed, dense, and population codes in neuroscience and deep learning . What, then, is the contribution of this paper? For interpretability researchers, our main contribution is providing a direct demonstration that superposition occurs in artificial neural networks given a relatively natural setup, suggesting this may also occur in practice. We offer a theory of when and why this occurs, revealing a phase diagram for superposition. We also discover that, at least in our toy model, superposition exhibits complex geometric structure. But our results may also be of broader interest. We find preliminary evidence that superposition may be linked to adversarial examples and grokking, and might also suggest a theory for the performance of mixture of experts models. More broadly, the toy model we investigate has unexpectedly rich structure, exhibiting phase changes, a geometric structure based on uniform polytopes, \"energy level\"-like jumps during training, and a phenomenon which is qualitatively similar to the fractional quantum Hall effect in physics. We originally investigated the subject to gain understanding of cleanly-interpretable neurons in larger models, but we've found these toy models to be surprisingly interesting in their own right. [1, 2] [3] [4] [5] KEY RESULTS FROM OUR TOY MODELS In our toy models, we are able to demonstrate that: Superposition is a real, observed phenomenon. Both monosemantic and polysemantic neurons can form. At least some kinds of computation can be performed in superposition. Whether features are stored in superposition is governed by a phase change. Superposition organizes features into geometric structures such as digons, triangles, pentagons, and tetrahedrons. Our toy models are simple ReLU networks, so it seems fair to say that neural networks exhibit these properties in at least some regimes, but it's very unclear what to generalize to real networks. Definitions and Motivation: Features, Directions, and Superposition In our work, we often think of neural networks as having features of the input represented as directions in activation space. This isn't a trivial claim. It isn't obvious what kind of structure we should expect neural network representations to have. When we say something like \"word embeddings have a gender direction\" or \"vision models have curve detector neurons\", one is implicitly making strong claims about the structure of network representations. Despite this, we believe this kind of \"linear representation hypothesis\" is supported both by significant empirical findings and theoretical arguments. One might think of this as two separate properties, which we'll explore in more detail shortly: Decomposability: Network representations can be described in terms of independently understandable features. Linearity: Features are represented by direction. If we hope to reverse engineer neural networks, we need a property like decomposability. Decomposability is what allows us to reason about the model without fitting the whole thing in our heads! But it's not enough for things to be decomposable: we need to be able to access the decomposition somehow. In order to do this, we need to identify the individual features within a representation. In a linear representation, this corresponds to determining which directions in activation space correspond to which independent features of the input. Sometimes, identifying feature directions is very easy because features seem to correspond to neurons. For example, many neurons in the early layers of InceptionV1 clearly correspond to features (e.g. curve detector neurons ). Why is it that we sometimes get this extremely helpful property, but in other cases don't? We hypothesize that there are really two countervailing forces driving this: [6] Privileged Basis: Only some representations have a privileged basis which encourages features to align with basis directions (i.e. to correspond to neurons). Superposition: Linear representations can represent more features than dimensions, using a strategy we call superposition. This can be seen as neural networks simulating larger networks. This pushes features away from corresponding to neurons. Superposition has been hypothesized in previous work  . However, we're not aware of feature superposition having been unambiguously demonstrated to occur in neural networks before ( demonstrates a closely related phenomenon of model superposition). The goal of this paper is to change that, demonstrating superposition and exploring how it interacts with privileged bases. If superposition occurs in networks, it deeply influences what approaches to interpretability research make sense, so unambiguous demonstration seems important. [1, 2] [7] The goal of this section will be to motivate these ideas and unpack them in detail. It's worth noting that many of the ideas in this section have close connections to ideas in other lines of interpretability research (especially disentanglement), neuroscience (distributed representations, population codes, etc), compressed sensing, and many other lines of work. This section will focus on articulating our perspective on the problem. We'll discuss these other lines of work in detail in Related Work."
                },
                {
                    "title": "Patching open-vocabulary models by interpolating weights",
                    "abstract": "Open-vocabulary models like CLIP achieve high accuracy across many image classification tasks. However, there are still settings where their zero-shot performance is far from optimal. We study model patching, where the goal is to improve accuracy on specific tasks without degrading accuracy on tasks where performance is already adequate. Towards this goal, we introduce PAINT, a patching method that uses interpolations between the weights of a model before fine-tuning and the weights after fine-tuning on a task to be patched. On nine tasks where zero-shot CLIP performs poorly, PAINT increases accuracy by 15 to 60 percentage points while preserving accuracy on ImageNet within one percentage point of the zero-shot model. PAINT also allows a single model to be patched on multiple tasks and improves with model scale. Furthermore, we identify cases of broad transfer, where patching on one task increases accuracy on other tasks even when the tasks have disjoint classes. Finally, we investigate applications beyond common benchmarks such as counting or reducing the impact of typographic attacks on CLIP. Our findings demonstrate that it is possible to expand the set of tasks on which open-vocabulary models achieve high accuracy without re-training them from scratch."
                },
                {
                    "title": "Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networks",
                    "abstract": "The last decade of machine learning has seen drastic increases in scale and capabilities. Deep neural networks (DNNs) are increasingly being deployed in the real world. However, they are difficult to analyze, raising concerns about using them without a rigorous understanding of how they function. Effective tools for interpreting them will be important for building more trustworthy AI by helping to identify problems, fix bugs, and improve basic understanding. In particular, \u201cinner\u201d interpretability techniques, which focus on explaining the internal components of DNNs, are well-suited for developing a mechanistic understanding, guiding manual modifications, and reverse engineering solutions. Much recent work has focused on DNN interpretability, and rapid progress has thus far made a thorough systematization of methods difficult. In this survey, we review over 300 works with a focus on inner interpretability tools. We introduce a taxonomy that classifies methods by what part of the network they help to explain (weights, neurons, subnetworks, or latent representations) and whether they are implemented during (intrinsic) or after (post hoc) training. To our knowledge, we are also the first to survey a number of connections between interpretability research and work in adversarial robustness, continual learning, modularity, network compression, and studying the human visual system. We discuss key challenges and argue that the status quo in interpretability research is largely unproductive. Finally, we highlight the importance of future work that emphasizes diagnostics, debugging, adversaries, and benchmarking in order to make interpretability tools more useful to engineers in practical applications."
                },
                {
                    "title": "Distilling Model Failures as Directions in Latent Space",
                    "abstract": "Existing methods for isolating hard subpopulations and spurious correlations in datasets often require human intervention. This can make these methods labor-intensive and dataset-specific. To address these shortcomings, we present a scalable method for automatically distilling a model's failure modes. Specifically, we harness linear classifiers to identify consistent error patterns, and, in turn, induce a natural representation of these failure modes as directions within the feature space. We demonstrate that this framework allows us to discover and automatically caption challenging subpopulations within the training dataset. Moreover, by combining our framework with off-the-shelf diffusion models, we can generate images that are especially challenging for the analyzed model, and thus can be used to perform synthetic data augmentation that helps remedy the model's failure modes. Code available at https://github.com/MadryLab/failure-directions"
                },
                {
                    "title": "Multimodal Contrastive Learning with LIMoE: the Language-Image Mixture of Experts",
                    "abstract": "Large sparsely-activated models have obtained excellent performance in multiple domains. However, such models are typically trained on a single modality at a time. We present the Language-Image MoE, LIMoE, a sparse mixture of experts model capable of multimodal learning. LIMoE accepts both images and text simultaneously, while being trained using a contrastive loss. MoEs are a natural fit for a multimodal backbone, since expert layers can learn an appropriate partitioning of modalities. However, new challenges arise; in particular, training stability and balanced expert utilization, for which we propose an entropy-based regularization scheme. Across multiple scales, we demonstrate remarkable performance improvement over dense models of equivalent computational cost. LIMoE-L/16 trained comparably to CLIP-L/14 achieves 78.6% zero-shot ImageNet accuracy (vs. 76.2%), and when further scaled to H/14 (with additional data) it achieves 84.1%, comparable to state-of-the-art methods which use larger custom per-modality backbones and pre-training schemes. We analyse the quantitative and qualitative behavior of LIMoE, and demonstrate phenomena such as differing treatment of the modalities and the organic emergence of modality-specific experts."
                },
                {
                    "title": "Interpretable Mixture of Experts",
                    "abstract": "The need for reliable model explanations is prominent for many machine learning applications, particularly for tabular and time-series data as their use cases often involve high-stakes decision making. Towards this goal, we introduce a novel interpretable modeling framework, Interpretable Mixture of Experts (IME), that yields high accuracy, comparable to `black-box' Deep Neural Networks (DNNs) in many cases, along with useful interpretability capabilities. IME consists of an assignment module and a mixture of experts, with each sample being assigned to a single expert for prediction. We introduce multiple options for IME based on the assignment and experts being interpretable. When the experts are chosen to be interpretable such as linear models, IME yields an inherently-interpretable architecture where the explanations produced by IME are the exact descriptions of how the prediction is computed. In addition to constituting a standalone inherently-interpretable architecture, IME has the premise of being integrated with existing DNNs to offer interpretability to a subset of samples while maintaining the accuracy of the DNNs. Through extensive experiments on 15 tabular and time-series datasets, IME is demonstrated to be more accurate than single interpretable models and perform comparably with existing state-of-the-art DNNs in accuracy. On most datasets, IME even outperforms DNNs, while providing faithful explanations. Lastly, IME's explanations are compared to commonly-used post-hoc explanations methods through a user study -- participants are able to better predict the model behavior when given IME explanations, while finding IME's explanations more faithful and trustworthy."
                },
                {
                    "title": "Sparsely-gated Mixture-of-Expert Layers for CNN Interpretability",
                    "abstract": "Sparsely-gated Mixture of Expert (MoE) layers have been recently successfully applied for scaling large transformers, especially for language modeling tasks. An intriguing side effect of sparse MoE layers is that they convey inherent interpretability to a model via natural expert specialization. In this work, we apply sparse MoE layers to CNNs for computer vision tasks and analyze the resulting effect on model interpretability. To stabilize MoE training, we present both soft and hard constraint-based approaches. With hard constraints, the weights of certain experts are allowed to become zero, while soft constraints balance the contribution of experts with an additional auxiliary loss. As a result, soft constraints handle expert utilization better and support the expert specialization process, while hard constraints maintain more generalized experts and increase overall model performance. Our findings demonstrate that experts can implicitly focus on individual sub-domains of the input space. For example, experts trained for CIFAR-100 image classification specialize in recognizing different domains such as flowers or animals without previous data clustering. Experiments with RetinaNet and the COCO dataset further indicate that object detection experts can also specialize in detecting objects of distinct sizes."
                },
                {
                    "title": "Sparsely Activated Mixture-of-Experts are Robust Multi-Task Learners",
                    "abstract": "Traditional multi-task learning (MTL) methods use dense networks that use the same set of shared weights across several different tasks. This often creates interference where two or more tasks compete to pull model parameters in different directions. In this work, we study whether sparsely activated Mixture-of-Experts (MoE) improve multi-task learning by specializing some weights for learning shared representations and using the others for learning task-specific information. To this end, we devise task-aware gating functions to route examples from different tasks to specialized experts which share subsets of network weights conditioned on the task. This results in a sparsely activated multi-task model with a large number of parameters, but with the same computational cost as that of a dense model. We demonstrate such sparse networks to improve multi-task learning along three key dimensions: (i) transfer to low-resource tasks from related tasks in the training mixture; (ii) sample-efficient generalization to tasks not seen during training by making use of task-aware routing from seen related tasks; (iii) robustness to the addition of unrelated tasks by avoiding catastrophic forgetting of existing tasks."
                },
                {
                    "title": "Parameter-Efficient Mixture-of-Experts Architecture for Pre-trained Language Models",
                    "abstract": "Recently, Mixture-of-Experts (short as MoE) architecture has achieved remarkable success in increasing the model capacity of large-scale language models. However, MoE requires incorporating significantly more parameters than the base model being extended. In this paper, we propose building a parameter-efficient MoE architecture by sharing information across experts. We adopt matrix product operator (MPO, a tensor decomposition from quantum many-body physics) to reconstruct the parameter matrix in the expert layer and increase model capacity for pre-trained language models by sharing parameters of the central tensor (containing the core information) among different experts while enabling the specificity through the auxiliary tensors (complementing the central tensor) of different experts. To address the unbalanced optimization issue, we further design the gradient mask strategy for the MPO-based MoE architecture. Extensive experiments based on T5 and GPT-2 show improved performance and efficiency of the pre-trained language model (27.2x reduction in total parameters for the superior model performance, compared with the Switch Transformers). Our code is publicly available at https://github.com/RUCAIBox/MPO/MPOE."
                },
                {
                    "title": "Locating and Editing Factual Associations in GPT",
                    "abstract": "We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's factual predictions. This reveals a distinct set of steps in middle-layer feed-forward modules that mediate factual predictions while processing subject tokens. To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update specific factual associations using Rank-One Model Editing (ROME). We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another. Our results confirm an important role for mid-layer feed-forward modules in storing factual associations and suggest that direct manipulation of computational mechanisms may be a feasible approach for model editing. The code, dataset, visualizations, and an interactive demo notebook are available at https://rome.baulab.info/"
                },
                {
                    "title": "GLaM: Efficient Scaling of Language Models with Mixture-of-Experts",
                    "abstract": "Scaling language models with more data, compute and parameters has driven significant progress in natural language processing. For example, thanks to scaling, GPT-3 was able to achieve strong results on in-context learning tasks. However, training these large dense models requires significant amounts of computing resources. In this paper, we propose and develop a family of language models named GLaM (Generalist Language Model), which uses a sparsely activated mixture-of-experts architecture to scale the model capacity while also incurring substantially less training cost compared to dense variants. The largest GLaM has 1.2 trillion parameters, which is approximately 7x larger than GPT-3. It consumes only 1/3 of the energy used to train GPT-3 and requires half of the computation flops for inference, while still achieving better overall zero-shot and one-shot performance across 29 NLP tasks."
                },
                {
                    "title": "Quantifying Local Specialization in Deep Neural Networks",
                    "abstract": "A neural network is locally specialized to the extent that parts of its computational graph (i.e. structure) can be abstractly represented as performing some comprehensible sub-task relevant to the overall task (i.e. functionality). Are modern deep neural networks locally specialized? How can this be quantified? In this paper, we consider the problem of taking a neural network whose neurons are partitioned into clusters, and quantifying how functionally specialized the clusters are. We propose two proxies for this: importance, which reflects how crucial sets of neurons are to network performance; and coherence, which reflects how consistently their neurons associate with features of the inputs. To measure these proxies, we develop a set of statistical methods based on techniques conventionally used to interpret individual neurons. We apply the proxies to partitionings generated by spectrally clustering a graph representation of the network's neurons with edges determined either by network weights or correlations of activations. We show that these partitionings, even ones based only on weights (i.e. strictly from non-runtime analysis), reveal groups of neurons that are important and coherent. These results suggest that graph-based partitioning can reveal local specialization and that statistical methods can be used to automatedly screen for sets of neurons that can be understood abstractly."
                },
                {
                    "title": "Poly-NL: Linear Complexity Non-local Layers With 3rd Order Polynomials",
                    "abstract": "Spatial self-attention layers, in the form of Non-Local blocks, introduce long-range dependencies in Convolutional Neural Networks by computing pairwise similarities among all possible positions. Such pairwise functions underpin the effectiveness of non-local layers, but also determine a complexity that scales quadratically with respect to the input size both in space and time. This is a severely limiting factor that practically hinders the applicability of non-local blocks to even moderately sized inputs. Previous works focused on reducing the complexity by modifying the underlying matrix operations, however in this work we aim to retain full expressiveness of non-local layers while keeping complexity linear. We overcome the efficiency limitation of non-local blocks by framing them as special cases of 3rd order polynomial functions. This fact enables us to formulate novel fast Non-Local blocks, capable of reducing the complexity from quadratic to linear with no loss in performance, by replacing any direct computation of pairwise similarities with element-wise multiplications. The proposed method, which we dub as \"Poly-NL\", is competitive with state-of-the-art performance across image recognition, instance segmentation, and face detection tasks, while having considerably less computational overhead."
                },
                {
                    "title": "Robust fine-tuning of zero-shot models",
                    "abstract": "Large pre-trained models such as CLIP or ALIGN offer consistent accuracy across a range of data distributions when performing zero-shot inference (i.e., without fine-tuning on a specific dataset). Although existing fine-tuning methods substantially improve accuracy on a given target distribution, they often reduce robustness to distribution shifts. We address this tension by introducing a simple and effective method for improving robustness while fine-tuning: ensembling the weights of the zero-shot and fine-tuned models (WiSE-FT). Compared to standard fine-tuning, WiSE-FT provides large accuracy improvements under distribution shift, while preserving high accuracy on the target distribution. On ImageNet and five derived distribution shifts, WiSE-FT improves accuracy under distribution shift by 4 to 6 percentage points (pp) over prior work while increasing ImageNet accuracy by 1.6 pp. WiSE-FT achieves similarly large robustness gains (2 to 23 pp) on a diverse set of six further distribution shifts, and accuracy gains of 0.8 to 3.3 pp compared to standard fine-tuning on commonly used transfer learning datasets. These improvements come at no additional computational cost during fine-tuning or inference."
                },
                {
                    "title": "DEMix Layers: Disentangling Domains for Modular Language Modeling",
                    "abstract": "We introduce a new domain expert mixture (DEMix) layer that enables conditioning a language model (LM) on the domain of the input text. A DEMix layer includes a collection of expert feedforward networks, each specialized to a domain, that makes the LM modular: experts can be mixed, added, or removed after initial training. Extensive experiments with autoregressive transformer LMs (up to 1.3B parameters) show that DEMix layers reduce test-time perplexity (especially for out-of-domain data), increase training efficiency, and enable rapid adaptation. Mixing experts during inference, using a parameter-free weighted ensemble, enables better generalization to heterogeneous or unseen domains. We also show it is possible to add experts to adapt to new domains without forgetting older ones, and remove experts to restrict access to unwanted domains. Overall, these results demonstrate benefits of domain modularity in language models."
                },
                {
                    "title": "Go Wider Instead of Deeper",
                    "abstract": "More transformer blocks with residual connections have recently achieved impressive results on various tasks. To achieve better performance with fewer trainable parameters, recent methods are proposed to go shallower by parameter sharing or model compressing along with the depth. However, weak modeling capacity limits their performance. Contrastively, going wider by inducing more trainable matrixes and parameters would produce a huge model requiring advanced parallelism to train and inference. \n \n \n In this paper, we propose a parameter-efficient framework, going wider instead of deeper. Specially, following existing works, we adapt parameter sharing to compress along depth. But, such deployment would limit the performance. To maximize modeling capacity, we scale along model width by replacing feed-forward network (FFN) with mixture-of-experts (MoE). Across transformer blocks, instead of sharing normalization layers, we propose to use individual layernorms to transform various semantic representations in a more parameter-efficient way. To evaluate our plug-and-run framework, we design WideNet and conduct comprehensive experiments on popular computer vision and natural language processing benchmarks. On ImageNet-1K, our best model outperforms Vision Transformer (ViT) by 1.5% with 0.72 times trainable parameters. Using 0.46 times and 0.13 times parameters, our WideNet can still surpass ViT and ViT-MoE by 0.8% and 2.1%, respectively. On four natural language processing datasets, WideNet outperforms ALBERT by 1.8% on average and surpass BERT using factorized embedding parameterization by 0.8% with fewer parameters."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Scaling Vision with Sparse Mixture of Experts",
                    "abstract": "Sparsely-gated Mixture of Experts networks (MoEs) have demonstrated excellent scalability in Natural Language Processing. In Computer Vision, however, almost all performant networks are\"dense\", that is, every input is processed by every parameter. We present a Vision MoE (V-MoE), a sparse version of the Vision Transformer, that is scalable and competitive with the largest dense networks. When applied to image recognition, V-MoE matches the performance of state-of-the-art networks, while requiring as little as half of the compute at inference time. Further, we propose an extension to the routing algorithm that can prioritize subsets of each input across the entire batch, leading to adaptive per-image compute. This allows V-MoE to trade-off performance and compute smoothly at test-time. Finally, we demonstrate the potential of V-MoE to scale vision models, and train a 15B parameter model that attains 90.35% on ImageNet."
                },
                {
                    "title": "Counterfactual Interventions Reveal the Causal Effect of Relative Clause Representations on Agreement Prediction",
                    "abstract": "When language models process syntactically complex sentences, do they use their representations of syntax in a manner that is consistent with the grammar of the language? We propose AlterRep, an intervention-based method to address this question. For any linguistic feature of a given sentence, AlterRep generates counterfactual representations by altering how the feature is encoded, while leaving in- tact all other aspects of the original representation. By measuring the change in a model\u2019s word prediction behavior when these counterfactual representations are substituted for the original ones, we can draw conclusions about the causal effect of the linguistic feature in question on the model\u2019s behavior. We apply this method to study how BERT models of different sizes process relative clauses (RCs). We find that BERT variants use RC boundary information during word prediction in a manner that is consistent with the rules of English grammar; this RC boundary information generalizes to a considerable extent across different RC types, suggesting that BERT represents RCs as an abstract linguistic category."
                },
                {
                    "title": "MLP-Mixer: An all-MLP Architecture for Vision",
                    "abstract": "Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e.\"mixing\"the per-location features), and one with MLPs applied across patches (i.e.\"mixing\"spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "An Interpretability Illusion for BERT",
                    "abstract": "We describe an\"interpretability illusion\"that arises when analyzing the BERT model. Activations of individual neurons in the network may spuriously appear to encode a single, simple concept, when in fact they are encoding something far more complex. The same effect holds for linear combinations of activations. We trace the source of this illusion to geometric properties of BERT's embedding space as well as the fact that common text corpora represent only narrow slices of possible English sentences. We provide a taxonomy of model-learned concepts and discuss methodological implications for interpretability research, especially the importance of testing hypotheses on multiple data sets."
                },
                {
                    "title": "BASE Layers: Simplifying Training of Large, Sparse Models",
                    "abstract": "We introduce a new balanced assignment of experts (BASE) layer for large language models that greatly simplifies existing high capacity sparse layers. Sparse layers can dramatically improve the efficiency of training and inference by routing each token to specialized expert modules that contain only a small fraction of the model parameters. However, it can be difficult to learn balanced routing functions that make full use of the available experts; existing approaches typically use routing heuristics or auxiliary expert-balancing loss functions. In contrast, we formulate token-to-expert allocation as a linear assignment problem, allowing an optimal assignment in which each expert receives an equal number of tokens. This optimal assignment scheme improves efficiency by guaranteeing balanced compute loads, and also simplifies training by not requiring any new hyperparameters or auxiliary losses. Code is publicly released at https://github.com/pytorch/fairseq/"
                },
                {
                    "title": "Revisiting Dynamic Convolution via Matrix Decomposition",
                    "abstract": "Recent research in dynamic convolution shows substantial performance boost for efficient CNNs, due to the adaptive aggregation of K static convolution kernels.It has two limitations: (a) it increases the number of convolutional weights by K-times, and (b) the joint optimization of dynamic attention and static convolution kernels is challenging. In this paper, we revisit it from a new perspective of matrix decomposition and reveal the key issue is that dynamic convolution applies dynamic attentions over channel groups after projecting into a higher dimensional intermediate space. To address this issue, we propose dynamic channel fusion to replace dynamic attentions over channel groups. Dynamic channel fusion not only enables significant dimension reduction of the intermediate space, but also mitigates the joint optimization difficulty. As a result, our method is easier to train and requires significantly fewer parameters without sacrificing accuracy."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Technical Challenges for Training Fair Neural Networks",
                    "abstract": "As machine learning algorithms have been widely deployed across applications, many concerns have been raised over the fairness of their predictions, especially in high stakes settings (such as facial recognition and medical imaging). To respond to these concerns, the community has proposed and formalized various notions of fairness as well as methods for rectifying unfair behavior. While fairness constraints have been studied extensively for classical models, the effectiveness of methods for imposing fairness on deep neural networks is unclear. In this paper, we observe that these large models overfit to fairness objectives, and produce a range of unintended and undesirable consequences. We conduct our experiments on both facial recognition and automated medical diagnosis datasets using state-of-the-art architectures."
                },
                {
                    "title": "Dynamic Neural Networks: A Survey",
                    "abstract": "Dynamic neural network is an emerging research topic in deep learning. Compared to static models which have fixed computational graphs and parameters at the inference stage, dynamic networks can adapt their structures or parameters to different inputs, leading to notable advantages in terms of accuracy, computational efficiency, adaptiveness, etc. In this survey, we comprehensively review this rapidly developing area by dividing dynamic networks into three main categories: 1) sample-wise dynamic models that process each sample with data-dependent architectures or parameters; 2) spatial-wise dynamic networks that conduct adaptive computation with respect to different spatial locations of image data; and 3) temporal-wise dynamic models that perform adaptive inference along the temporal dimension for sequential data such as videos and texts. The important research problems of dynamic networks, e.g., architecture design, decision making scheme, optimization technique and applications, are reviewed systematically. Finally, we discuss the open problems in this field together with interesting future research directions."
                },
                {
                    "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity",
                    "abstract": "In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with outrageous numbers of parameters -- but a constant computational cost. However, despite several notable successes of MoE, widespread adoption has been hindered by complexity, communication costs and training instability -- we address these with the Switch Transformer. We simplify the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs. Our proposed training techniques help wrangle the instabilities and we show large sparse models may be trained, for the first time, with lower precision (bfloat16) formats. We design models based off T5-Base and T5-Large to obtain up to 7x increases in pre-training speed with the same computational resources. These improvements extend into multilingual settings where we measure gains over the mT5-Base version across all 101 languages. Finally, we advance the current scale of language models by pre-training up to trillion parameter models on the\"Colossal Clean Crawled Corpus\"and achieve a 4x speedup over the T5-XXL model."
                },
                {
                    "title": "Amnesic Probing: Behavioral Explanation with Amnesic Counterfactuals",
                    "abstract": "Abstract A growing body of work makes use of probing in order to investigate the working of neural models, often considered black boxes. Recently, an ongoing debate emerged surrounding the limitations of the probing paradigm. In this work, we point out the inability to infer behavioral conclusions from probing results, and offer an alternative method that focuses on how the information is being used, rather than on what information is encoded. Our method, Amnesic Probing, follows the intuition that the utility of a property for a given task can be assessed by measuring the influence of a causal intervention that removes it from the representation. Equipped with this new analysis tool, we can ask questions that were not possible before, for example, is part-of-speech information important for word prediction? We perform a series of analyses on BERT to answer these types of questions. Our findings demonstrate that conventional probing performance is not correlated to task importance, and we call for increased scrutiny of claims that draw behavioral or causal conclusions from probing results.1"
                },
                {
                    "title": "Multilinear Latent Conditioning for Generating Unseen Attribute Combinations",
                    "abstract": "Deep generative models rely on their inductive bias to facilitate generalization, especially for problems with high dimensional data, like images. However, empirical studies have shown that variational autoencoders (VAE) and generative adversarial networks (GAN) lack the generalization ability that occurs naturally in human perception. For example, humans can visualize a woman smiling after only seeing a smiling man. On the contrary, the standard conditional VAE (cVAE) is unable to generate unseen attribute combinations. To this end, we extend cVAE by introducing a multilinear latent conditioning framework that captures the multiplicative interactions between the attributes. We implement two variants of our model and demonstrate their efficacy on MNIST, Fashion-MNIST and CelebA. Altogether, we design a novel conditioning framework that can be used with any architecture to synthesize unseen attribute combinations."
                },
                {
                    "title": "GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding",
                    "abstract": "Neural network scaling has been critical for improving the model quality in many real-world machine learning applications with vast amounts of training data and compute. Although this trend of scaling is affirmed to be a sure-fire approach for better model quality, there are challenges on the path such as the computation cost, ease of programming, and efficient implementation on parallel devices. GShard is a module composed of a set of lightweight annotation APIs and an extension to the XLA compiler. It provides an elegant way to express a wide range of parallel computation patterns with minimal changes to the existing model code. GShard enabled us to scale up multilingual neural machine translation Transformer model with Sparsely-Gated Mixture-of-Experts beyond 600 billion parameters using automatic sharding. We demonstrate that such a giant model can efficiently be trained on 2048 TPU v3 accelerators in 4 days to achieve far superior quality for translation from 100 languages to English compared to the prior art."
                },
                {
                    "title": "Deep Polynomial Neural Networks",
                    "abstract": "Deep convolutional neural networks (DCNNs) are currently the method of choice both for generative, as well as for discriminative learning in computer vision and machine learning. The success of DCNNs can be attributed to the careful selection of their building blocks (e.g., residual blocks, rectifiers, sophisticated normalization schemes, to mention but a few). In this paper, we propose <inline-formula><tex-math notation=\"LaTeX\">$\\Pi$</tex-math><alternatives><mml:math><mml:mi>\u03a0</mml:mi></mml:math><inline-graphic xlink:href=\"chrysos-ieq1-3058891.gif\"/></alternatives></inline-formula>-Nets, a new class of function approximators based on polynomial expansions. <inline-formula><tex-math notation=\"LaTeX\">$\\Pi$</tex-math><alternatives><mml:math><mml:mi>\u03a0</mml:mi></mml:math><inline-graphic xlink:href=\"chrysos-ieq2-3058891.gif\"/></alternatives></inline-formula>-Nets are polynomial neural networks, i.e., the output is a high-order polynomial of the input. The unknown parameters, which are naturally represented by high-order tensors, are estimated through a collective tensor factorization with factors sharing. We introduce three tensor decompositions that significantly reduce the number of parameters and show how they can be efficiently implemented by hierarchical neural networks. We empirically demonstrate that <inline-formula><tex-math notation=\"LaTeX\">$\\Pi$</tex-math><alternatives><mml:math><mml:mi>\u03a0</mml:mi></mml:math><inline-graphic xlink:href=\"chrysos-ieq3-3058891.gif\"/></alternatives></inline-formula>-Nets are very expressive and they even produce good results without the use of non-linear activation functions in a large battery of tasks and signals, i.e., images, graphs, and audio. When used in conjunction with activation functions, <inline-formula><tex-math notation=\"LaTeX\">$\\Pi$</tex-math><alternatives><mml:math><mml:mi>\u03a0</mml:mi></mml:math><inline-graphic xlink:href=\"chrysos-ieq4-3058891.gif\"/></alternatives></inline-formula>-Nets produce state-of-the-art results in three challenging tasks, i.e., image generation, face verification and 3D mesh representation learning. The source code is available at <uri>https://github.com/grigorisg9gr/polynomial_nets</uri>."
                },
                {
                    "title": "TESA: Tensor Element Self-Attention via Matricization",
                    "abstract": "Representation learning is a fundamental part of modern computer vision, where abstract representations of data are encoded as tensors optimized to solve problems like image segmentation and inpainting. Recently, self-attention in the form of Non-Local Block has emerged as a powerful technique to enrich features, by capturing complex interdependencies in feature tensors. However, standard self-attention approaches leverage only spatial relationships, drawing similarities between vectors and overlooking correlations between channels. In this paper, we introduce a new method, called Tensor Element Self-Attention (TESA) that generalizes such work to capture interdependencies along all dimensions of the tensor using matricization. An order R tensor produces R results, one for each dimension. The results are then fused to produce an enriched output which encapsulates similarity among tensor elements. Additionally, we analyze self-attention mathematically, providing new perspectives on how it adjusts the singular values of the input feature tensor. With these new insights, we present experimental results demonstrating how TESA can benefit diverse problems including classification and instance segmentation. By simply adding a TESA module to existing networks, we substantially improve competitive baselines and set new state-of-the-art results for image inpainting on Celeb and low light raw-to-rgb image translation on SID."
                },
                {
                    "title": "Fairness without Demographics through Adversarially Reweighted Learning",
                    "abstract": "Much of the previous machine learning (ML) fairness literature assumes that protected features such as race and sex are present in the dataset, and relies upon them to mitigate fairness concerns. However, in practice factors like privacy and regulation often preclude the collection of protected features, or their use for training or inference, severely limiting the applicability of traditional fairness research. Therefore we ask: How can we train a ML model to improve fairness when we do not even know the protected group memberships? In this work we address this problem by proposing Adversarially Reweighted Learning (ARL). In particular, we hypothesize that non-protected features and task labels are valuable for identifying fairness issues, and can be used to co-train an adversarial reweighting approach for improving fairness. Our results show that ARL improves Rawlsian Max-Min fairness, with significant AUC improvements for worst-case protected groups in multiple datasets,outperforming state-of-the-art alternatives."
                },
                {
                    "title": "P\u2013nets: Deep Polynomial Neural Networks",
                    "abstract": "Deep Convolutional Neural Networks (DCNNs) is currently the method of choice both for generative, as well as for discriminative learning in computer vision and machine learning. The success of DCNNs can be attributed to the careful selection of their building blocks (e.g., residual blocks, rectifiers, sophisticated normalization schemes, to mention but a few). In this paper, we propose $\\Pi$-Nets, a new class of DCNNs. $\\Pi$-Nets are polynomial neural networks, i.e., the output is a high-order polynomial of the input. $\\Pi$-Nets can be implemented using special kind of skip connections and their parameters can be represented via high-order tensors. We empirically demonstrate that $\\Pi$-Nets have better representation power than standard DCNNs and they even produce good results without the use of non-linear activation functions in a large battery of tasks and signals, i.e., images, graphs, and audio. When used in conjunction with activation functions, $\\Pi$-Nets produce state-of-the-art results in challenging tasks, such as image generation. Lastly, our framework elucidates why recent generative models, such as StyleGAN, improve upon their predecessors, e.g., ProGAN."
                },
                {
                    "title": "Dynamic Convolution: Attention Over Convolution Kernels",
                    "abstract": "Light-weight convolutional neural networks (CNNs) suffer performance degradation as their low computational budgets constrain both the depth (number of convolution layers) and the width (number of channels) of CNNs, resulting in limited representation capability. To address this issue, we present Dynamic Convolution, a new design that increases model complexity without increasing the network depth or width. Instead of using a single convolution kernel per layer, dynamic convolution aggregates multiple parallel convolution kernels dynamically based upon their attentions, which are input dependent. Assembling multiple kernels is not only computationally efficient due to the small kernel size, but also has more representation power since these kernels are aggregated in a non-linear way via attention. By simply using dynamic convolution for the state-of-the-art architecture MobileNetV3-Small, the top-1 accuracy of ImageNet classification is boosted by 2.9% with only 4% additional FLOPs and 2.9 AP gain is achieved on COCO keypoint detection."
                },
                {
                    "title": "Towards Fairness in Visual Recognition: Effective Strategies for Bias Mitigation",
                    "abstract": "Computer vision models learn to perform a task by capturing relevant statistics from training data. It has been shown that models learn spurious age, gender, and race correlations when trained for seemingly unrelated tasks like activity recognition or image captioning. Various mitigation techniques have been presented to prevent models from utilizing or learning such biases. However, there has been little systematic comparison between these techniques. We design a simple but surprisingly effective visual recognition benchmark for studying bias mitigation. Using this benchmark, we provide a thorough analysis of a wide range of techniques. We highlight the shortcomings of popular adversarial training approaches for bias mitigation, propose a simple but similarly effective alternative to the inference-time Reducing Bias Amplification method of Zhao et al., and design a domain-independent training technique that outperforms all other methods. Finally, we validate our findings on the attribute classification task in the CelebA dataset, where attribute presence is known to be correlated with the gender of people in the image, and demonstrate that the proposed technique is effective at mitigating real-world gender bias."
                },
                {
                    "title": "Mitigate Bias in Face Recognition using Skewness-Aware Reinforcement Learning",
                    "abstract": "Racial equality is an important theme of international human rights law, but it has been largely obscured when the overall face recognition accuracy is pursued blindly. More facts indicate racial bias indeed degrades the fairness of recognition system and the error rates on non-Caucasians are usually much higher than Caucasians. To encourage fairness, we introduce the idea of adaptive margin to learn balanced performance for different races based on large margin losses. A reinforcement learning based race balance network (RL-RBN) is proposed. We formulate the process of finding the optimal margins for non-Caucasians as a Markov decision process and employ deep Q-learning to learn policies for an agent to select appropriate margin by approximating the Q-value function. Guided by the agent, the skewness of feature scatter between races can be reduced. Besides, we provide two ethnicity aware training datasets, called BUPT-Globalface and BUPT-Balancedface dataset, which can be utilized to study racial bias from both data and algorithm aspects. Extensive experiments on RFW database show that RL-RBN successfully mitigates racial bias and learns more balanced performance for different races."
                },
                {
                    "title": "Adaptively Sparse Transformers",
                    "abstract": "Attention mechanisms have become ubiquitous in NLP. Recent architectures, notably the Transformer, learn powerful context-aware word representations through layered, multi-headed attention. The multiple heads learn diverse types of word relationships. However, with standard softmax attention, all attention heads are dense, assigning a non-zero weight to all context words. In this work, we introduce the adaptively sparse Transformer, wherein attention heads have flexible, context-dependent sparsity patterns. This sparsity is accomplished by replacing softmax with alpha-entmax: a differentiable generalization of softmax that allows low-scoring words to receive precisely zero weight. Moreover, we derive a method to automatically learn the alpha parameter \u2013 which controls the shape and sparsity of alpha-entmax \u2013 allowing attention heads to choose between focused or spread-out behavior. Our adaptively sparse Transformer improves interpretability and head diversity when compared to softmax Transformers on machine translation datasets. Findings of the quantitative and qualitative analysis of our approach include that heads in different layers learn different sparsity preferences and tend to be more diverse in their attention distributions than softmax Transformers. Furthermore, at no cost in accuracy, sparsity in attention heads helps to uncover different head specializations."
                },
                {
                    "title": "Factorized Higher-Order CNNs With an Application to Spatio-Temporal Emotion Estimation",
                    "abstract": "Training deep neural networks with spatio-temporal (i.e., 3D) or multidimensional convolutions of higher-order is computationally challenging due to millions of unknown parameters across dozens of layers. To alleviate this, one approach is to apply low-rank tensor decompositions to convolution kernels in order to compress the network and reduce its number of parameters. Alternatively, new convolutional blocks, such as MobileNet, can be directly designed for efficiency. In this paper, we unify these two approaches by proposing a tensor factorization framework for efficient multidimensional (separable) convolutions of higher-order. Interestingly, the proposed framework enables a novel higher-order transduction, allowing to train a network on a given domain (e.g., 2D images or N-dimensional data in general) and using transduction to generalize to higher-order data such as videos (or (N+K)--dimensional data in general), capturing for instance temporal dynamics while preserving the learnt spatial information. We apply the proposed methodology, coined CP-Higher-Order Convolution (HO-CPConv), to spatio-temporal facial emotion analysis. Most existing facial affect models focus on static imagery and discard all temporal information. This is due to the above-mentioned burden of training 3D convolutional nets and the lack of large bodies of video data annotated by experts. We address both issues with our proposed framework. Initial training is first done on static imagery before using transduction to generalize to the temporal domain. We demonstrate superior performance on three challenging large scale affect estimation datasets, AffectNet, SEWA, and AFEW-VA."
                },
                {
                    "title": "Sparse Sequence-to-Sequence Models",
                    "abstract": "Sequence-to-sequence models are a powerful workhorse of NLP. Most variants employ a softmax transformation in both their attention mechanism and output layer, leading to dense alignments and strictly positive output probabilities. This density is wasteful, making models less interpretable and assigning probability mass to many implausible outputs. In this paper, we propose sparse sequence-to-sequence models, rooted in a new family of \\alpha-entmax transformations, which includes softmax and sparsemax as particular cases, and is sparse for any \\alpha > 1. We provide fast algorithms to evaluate these transformations and their gradients, which scale well for large vocabulary sizes. Our models are able to produce sparse alignments and to assign nonzero probability to a short list of plausible outputs, sometimes rendering beam search exact. Experiments on morphological inflection and machine translation reveal consistent gains over dense models."
                },
                {
                    "title": "Incremental multi-domain learning with network latent tensor factorization",
                    "abstract": "The prominence of deep learning, large amount of annotated data and increasingly powerful hardware made it possible to reach remarkable performance for supervised classification tasks, in many cases saturating the training sets. However the resulting models are specialized to a single very specific task and domain. Adapting the learned classification to new domains is a hard problem due to at least three reasons: (1) the new domains and the tasks might be drastically different; (2) there might be very limited amount of annotated data on the new domain and (3) full training of a new model for each new task is prohibitive in terms of computation and memory, due to the sheer number of parameters of deep CNNs. In this paper, we present a method to learn new-domains and tasks incrementally, building on prior knowledge from already learned tasks and without catastrophic forgetting. We do so by jointly parametrizing weights across layers using low-rank Tucker structure. The core is task agnostic while a set of task specific factors are learnt on each new domain. We show that leveraging tensor structure enables better performance than simply using matrix operations. Joint tensor modelling also naturally leverages correlations across different layers. Compared with previous methods which have focused on adapting each layer separately, our approach results in more compact representations for each new task/domain. We apply the proposed method to the 10 datasets of the Visual Decathlon Challenge and show that our method offers on average about 7.5\u00d7 reduction in number of parameters and competitive performance in terms of both classification accuracy and Decathlon score."
                },
                {
                    "title": "CondConv: Conditionally Parameterized Convolutions for Efficient Inference",
                    "abstract": "Convolutional layers are one of the basic building blocks of modern deep neural networks. One fundamental assumption is that convolutional kernels should be shared for all examples in a dataset. We propose conditionally parameterized convolutions (CondConv), which learn specialized convolutional kernels for each example. Replacing normal convolutions with CondConv enables us to increase the size and capacity of a network, while maintaining efficient inference. We demonstrate that scaling networks with CondConv improves the performance and inference cost trade-off of several existing convolutional neural network architectures on both classification and detection tasks. On ImageNet classification, our CondConv approach applied to EfficientNet-B0 achieves state-ofthe-art performance of 78.3% accuracy with only 413M multiply-adds. Code and checkpoints for the CondConv Tensorflow layer and CondConv-EfficientNet models are available at: https://github.com/tensorflow/tpu/tree/master/ models/official/efficientnet/condconv."
                },
                {
                    "title": "Datasheets for datasets",
                    "abstract": "Documentation to facilitate communication between dataset creators and consumers."
                },
                {
                    "title": "Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification",
                    "abstract": "Recent studies demonstrate that machine learning algorithms can discriminate based on classes like race and gender. In this work, we present an approach to evaluate bias present in automated facial analysis algorithms and datasets with respect to phenotypic subgroups. Using the dermatologist approved Fitzpatrick Skin Type classi\ufb01cation system, we characterize the gender and skin type distribution of two facial analysis benchmarks, IJB-A and Adience. We \ufb01nd that these datasets are overwhelm-ingly composed of lighter-skinned subjects (79 . 6% for IJB-A and 86 . 2% for Adience) and introduce a new facial analysis dataset which is balanced by gender and skin type. We evaluate 3 commercial gender classi\ufb01cation systems using our dataset and show that darker-skinned females are the most misclassi\ufb01ed group (with error rates of up to 34 . 7%). The maximum error rate for lighter-skinned males is 0 . 8%. The substantial disparities in the accuracy of classifying darker females, lighter females, darker males, and lighter males in gender classi\ufb01cation systems require urgent attention if commercial companies are to build genuinely fair, transparent and accountable facial analysis algorithms."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Tensor Contraction Layers for Parsimonious Deep Nets",
                    "abstract": "Tensors offer a natural representation for many kinds of data frequently encountered in machine learning. Images, for example, are naturally represented as third order tensors, where the modes correspond to height, width, and channels. In particular, tensor decompositions are noted for their ability to discover multi-dimensional dependencies and produce compact low-rank approximations of data. In this paper, we explore the use of tensor contractions as neural network layers and investigate several ways to apply them to activation tensors. Specifically, we propose the Tensor Contraction Layer (TCL), the first attempt to incorporate tensor contractions as end-to-end trainable neural network layers. Applied to existing networks, TCLs reduce the dimensionality of the activation tensors and thus the number of model parameters. We evaluate the TCL on the task of image recognition, augmenting popular networks (AlexNet, VGG). The resulting models are trainable end-to-end. We evaluate TCL's performance on the task of image recognition, using the CIFAR100 and ImageNet datasets, studying the effect of parameter reduction via tensor contraction on performance. We demonstrate significant model compression without significant impact on the accuracy and, in some cases, improved performance."
                },
                {
                    "title": "Exponential Machines",
                    "abstract": "Modeling interactions between features improves the performance of machine learning solutions in many domains (e.g. recommender systems or sentiment analysis). In this paper, we introduce Exponential Machines (ExM), a predictor that models all interactions of every order. The key idea is to represent an exponentially large tensor of parameters in a factorized format called Tensor Train (TT). The Tensor Train format regularizes the model and lets you control the number of underlying parameters. To train the model, we develop a stochastic Riemannian optimization procedure, which allows us to fit tensors with 2^160 entries. We show that the model achieves state-of-the-art performance on synthetic data with high-order interactions and that it works on par with high-order factorization machines on a recommender system dataset MovieLens 100K."
                },
                {
                    "title": "Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer",
                    "abstract": "The capacity of a neural network to absorb information is limited by its number of parameters. Conditional computation, where parts of the network are active on a per-example basis, has been proposed in theory as a way of dramatically increasing model capacity without a proportional increase in computation. In practice, however, there are significant algorithmic and performance challenges. In this work, we address these challenges and finally realize the promise of conditional computation, achieving greater than 1000x improvements in model capacity with only minor losses in computational efficiency on modern GPU clusters. We introduce a Sparsely-Gated Mixture-of-Experts layer (MoE), consisting of up to thousands of feed-forward sub-networks. A trainable gating network determines a sparse combination of these experts to use for each example. We apply the MoE to the tasks of language modeling and machine translation, where model capacity is critical for absorbing the vast quantities of knowledge available in the training corpora. We present model architectures in which a MoE with up to 137 billion parameters is applied convolutionally between stacked LSTM layers. On large language modeling and machine translation benchmarks, these models achieve significantly better results than state-of-the-art at lower computational cost."
                },
                {
                    "title": "Ultimate tensorization: compressing convolutional and FC layers alike",
                    "abstract": "Convolutional neural networks excel in image recognition tasks, but this comes at the cost of high computational and memory complexity. To tackle this problem, [1] developed a tensor factorization framework to compress fully-connected layers. In this paper, we focus on compressing convolutional layers. We show that while the direct application of the tensor framework [1] to the 4-dimensional kernel of convolution does compress the layer, we can do better. We reshape the convolutional kernel into a tensor of higher order and factorize it. We combine the proposed approach with the previous work to compress both convolutional and fully-connected layers of a network and achieve 80x network compression rate with 1.1% accuracy drop on the CIFAR-10 dataset."
                },
                {
                    "title": "Equality of Opportunity in Supervised Learning",
                    "abstract": "We propose a criterion for discrimination against a specified sensitive attribute in supervised learning, where the goal is to predict some target based on available features. Assuming data about the predictor, target, and membership in the protected group are available, we show how to optimally adjust any learned predictor so as to remove discrimination according to our definition. Our framework also improves incentives by shifting the cost of poor classification from disadvantaged groups to the decision maker, who can respond by improving the classification accuracy. We enourage readers to consult the more complete manuscript on the arXiv."
                },
                {
                    "title": "HyperNetworks",
                    "abstract": "This work explores hypernetworks: an approach of using one network, also known as a hypernetwork, to generate the weights for another network. We apply hypernetworks to generate adaptive weights for recurrent networks. In this case, hypernetworks can be viewed as a relaxed form of weight-sharing across layers. In our implementation, hypernetworks are are trained jointly with the main network in an end-to-end fashion. Our main result is that hypernetworks can generate non-shared weights for LSTM and achieve state-of-the-art results on a variety of sequence modelling tasks including character-level language modelling, handwriting generation and neural machine translation, challenging the weight-sharing paradigm for recurrent networks."
                },
                {
                    "title": "Tensor Ring Decomposition",
                    "abstract": "Tensor networks have in recent years emerged as the powerful tools for solving the large-scale optimization problems. One of the most popular tensor network is tensor train (TT) decomposition that acts as the building blocks for the complicated tensor networks. However, the TT decomposition highly depends on permutations of tensor dimensions, due to its strictly sequential multilinear products over latent cores, which leads to difficulties in finding the optimal TT representation. In this paper, we introduce a fundamental tensor decomposition model to represent a large dimensional tensor by a circular multilinear products over a sequence of low dimensional cores, which can be graphically interpreted as a cyclic interconnection of 3rd-order tensors, and thus termed as tensor ring (TR) decomposition. The key advantage of TR model is the circular dimensional permutation invariance which is gained by employing the trace operation and treating the latent cores equivalently. TR model can be viewed as a linear combination of TT decompositions, thus obtaining the powerful and generalized representation abilities. For optimization of latent cores, we present four different algorithms based on the sequential SVDs, ALS scheme, and block-wise ALS techniques. Furthermore, the mathematical properties of TR model are investigated, which shows that the basic multilinear algebra can be performed efficiently by using TR representaions and the classical tensor decompositions can be conveniently transformed into the TR representation. Finally, the experiments on both synthetic signals and real-world datasets were conducted to evaluate the performance of different algorithms."
                },
                {
                    "title": "The mythos of model interpretability",
                    "abstract": "In machine learning, the concept of interpretability is both important and slippery."
                },
                {
                    "title": "Deep Multi-task Representation Learning: A Tensor Factorisation Approach",
                    "abstract": "Most contemporary multi-task learning methods assume linear models. This setting is considered shallow in the era of deep learning. In this paper, we present a new deep multi-task representation learning framework that learns cross-task sharing structure at every layer in a deep network. Our approach is based on generalising the matrix factorisation techniques explicitly or implicitly used by many conventional MTL algorithms to tensor factorisation, to realise automatic learning of end-to-end knowledge sharing in deep networks. This is in contrast to existing deep learning approaches that need a user-defined multi-task sharing strategy. Our approach applies to both homogeneous and heterogeneous MTL. Experiments demonstrate the efficacy of our deep multi-task representation learning in terms of both higher accuracy and fewer design choices."
                },
                {
                    "title": "\u201cWhy Should I Trust You?\u201d: Explaining the Predictions of Any Classifier",
                    "abstract": "Despite widespread adoption, machine learning models remain mostly black boxes. Understanding the reasons behind predictions is, however, quite important in assessing trust, which is fundamental if one plans to take action based on a prediction, or when choosing whether to deploy a new model. Such understanding also provides insights into the model, which can be used to transform an untrustworthy model or prediction into a trustworthy one. In this work, we propose LIME, a novel explanation technique that explains the predictions of any classifier in an interpretable and faithful manner, by learning an interpretable model locally varound the prediction. We also propose a method to explain models by presenting representative individual predictions and their explanations in a non-redundant way, framing the task as a submodular optimization problem. We demonstrate the flexibility of these methods by explaining different models for text (e.g. random forests) and image classification (e.g. neural networks). We show the utility of explanations via novel experiments, both simulated and with human subjects, on various scenarios that require trust: deciding if one should trust a prediction, choosing between models, improving an untrustworthy classifier, and identifying why a classifier should not be trusted."
                },
                {
                    "title": "Conditional Computation in Neural Networks for faster models",
                    "abstract": "Deep learning has become the state-of-art tool in many applications, but the evaluation and training of deep models can be time-consuming and computationally expensive. The conditional computation approach has been proposed to tackle this problem (Bengio et al., 2013; Davis & Arel, 2013). It operates by selectively activating only parts of the network at a time. In this paper, we use reinforcement learning as a tool to optimize conditional computation policies. More specifically, we cast the problem of learning activation-dependent policies for dropping out blocks of units as a reinforcement learning problem. We propose a learning scheme motivated by computation speed, capturing the idea of wanting to have parsimonious activations while maintaining prediction accuracy. We apply a policy gradient algorithm for learning policies that optimize this loss function and propose a regularization mechanism that encourages diversification of the dropout policy. We present encouraging empirical results showing that this approach improves the speed of computation without impacting the quality of the approximation."
                },
                {
                    "title": "Tensorizing Neural Networks",
                    "abstract": "Deep neural networks currently demonstrate state-of-the-art performance in several domains. At the same time, models of this class are very demanding in terms of computational resources. In particular, a large amount of memory is required by commonly used fully-connected layers, making it hard to use the models on low-end devices and stopping the further increase of the model size. In this paper we convert the dense weight matrices of the fully-connected layers to the Tensor Train [17] format such that the number of parameters is reduced by a huge factor and at the same time the expressive power of the layer is preserved. In particular, for the Very Deep VGG networks [21] we report the compression factor of the dense weight matrix of a fully-connected layer up to 200000 times leading to the compression factor of the whole network up to 7 times."
                },
                {
                    "title": "Deep Learning Face Attributes in the Wild",
                    "abstract": "Predicting face attributes in the wild is challenging due to complex face variations. We propose a novel deep learning framework for attribute prediction in the wild. It cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently. LNet is pre-trained by massive general object categories for face localization, while ANet is pre-trained by massive face identities for attribute prediction. This framework not only outperforms the state-of-the-art with a large margin, but also reveals valuable facts on learning face representation. (1) It shows how the performances of face localization (LNet) and attribute prediction (ANet) can be improved by different pre-training strategies. (2) It reveals that although the filters of LNet are fine-tuned only with image-level attribute tags, their response maps over entire images have strong indication of face locations. This fact enables training LNet for face localization with only image-level annotations, but without face bounding boxes or landmarks, which are required by all attribute recognition works. (3) It also demonstrates that the high-level hidden neurons of ANet automatically discover semantic concepts after pre-training with massive face identities, and such concepts are significantly enriched after fine-tuning with attribute tags. Each attribute can be well explained with a sparse linear combination of these concepts."
                },
                {
                    "title": "Learning Factored Representations in a Deep Mixture of Experts",
                    "abstract": "Mixtures of Experts combine the outputs of several \"expert\" networks, each of which specializes in a different part of the input space. This is achieved by training a \"gating\" network that maps each input to a distribution over the experts. Such models show promise for building larger networks that are still cheap to compute at test time, and more parallelizable at training time. In this this work, we extend the Mixture of Experts to a stacked model, the Deep Mixture of Experts, with multiple sets of gating and experts. This exponentially increases the number of effective experts by associating each input with a combination of experts at each layer, yet maintains a modest model size. On a randomly translated version of the MNIST dataset, we find that the Deep Mixture of Experts automatically learns to develop location-dependent (\"where\") experts at the first layer, and class-specific (\"what\") experts at the second layer. In addition, we see that the different combinations are in use when the model is applied to a dataset of speech monophones. These demonstrate effective use of all expert combinations."
                },
                {
                    "title": "Low-Rank Approximations for Conditional Feedforward Computation in Deep Neural Networks",
                    "abstract": "Scalability properties of deep neural networks raise key research questions, particularly as the problems considered become larger and more challenging. This paper expands on the idea of conditional computation introduced by Bengio, et. al., where the nodes of a deep network are augmented by a set of gating units that determine when a node should be calculated. By factorizing the weight matrix into a low-rank approximation, an estimation of the sign of the pre-nonlinearity activation can be efficiently obtained. For networks using rectified-linear hidden units, this implies that the computation of a hidden unit with an estimated negative pre-nonlinearity can be ommitted altogether, as its value will become zero when nonlinearity is applied. For sparse neural networks, this can result in considerable speed gains. Experimental results using the MNIST and SVHN data sets with a fully-connected deep neural network demonstrate the performance robustness of the proposed scheme with respect to the error introduced by the conditional computation process."
                },
                {
                    "title": "Tensor-Train Decomposition",
                    "abstract": "A simple nonrecursive form of the tensor decomposition in $d$ dimensions is presented. It does not inherently suffer from the curse of dimensionality, it has asymptotically the same number of parameters as the canonical decomposition, but it is stable and its computation is based on low-rank approximation of auxiliary unfolding matrices. The new form gives a clear and convenient way to implement all basic operations efficiently. A fast rounding procedure is presented, as well as basic linear algebra operations. Examples showing the benefits of the decomposition are given, and the efficiency is demonstrated by the computation of the smallest eigenvalue of a 19-dimensional operator."
                },
                {
                    "title": "Tensor Decompositions and Applications",
                    "abstract": "This survey provides an overview of higher-order tensor decompositions, their applications, and available software. A tensor is a multidimensional or $N$-way array. Decompositions of higher-order tensors (i.e., $N$-way arrays with $N \\geq 3$) have applications in psycho-metrics, chemometrics, signal processing, numerical linear algebra, computer vision, numerical analysis, data mining, neuroscience, graph analysis, and elsewhere. Two particular tensor decompositions can be considered to be higher-order extensions of the matrix singular value decomposition: CANDECOMP/PARAFAC (CP) decomposes a tensor as a sum of rank-one tensors, and the Tucker decomposition is a higher-order form of principal component analysis. There are many other tensor decompositions, including INDSCAL, PARAFAC2, CANDELINC, DEDICOM, and PARATUCK2 as well as nonnegative variants of all of the above. The N-way Toolbox, Tensor Toolbox, and Multilinear Engine are examples of software packages for working with tensors."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "On the Best Rank-1 and Rank-(R1 , R2, ... , RN) Approximation of Higher-Order Tensors",
                    "abstract": "In this paper we discuss a multilinear generalization of the best rank-R approximation problem for matrices, namely, the approximation of a given higher-order tensor, in an optimal least-squares sense, by a tensor that has prespecified column rank value, row rank value, etc. For matrices, the solution is conceptually obtained by truncation of the singular value decomposition (SVD); however, this approach does not have a straightforward multilinear counterpart. We discuss higher-order generalizations of the power method and the orthogonal iteration method."
                },
                {
                    "title": "Hierarchical Mixtures of Experts and the EM Algorithm",
                    "abstract": "We present a tree-structured architecture for supervised learning. The statistical model underlying the architecture is a hierarchical mixture model in which both the mixture coefficients and the mixture components are generalized linear models (GLIM's). Learning is treated as a maximum likelihood problem; in particular, we present an Expectation-Maximization (EM) algorithm for adjusting the parameters of the architecture. We also develop an on-line learning algorithm in which the parameters are updated incrementally. Comparative simulation results are presented in the robot dynamics domain."
                },
                {
                    "title": "Adaptive Mixtures of Local Experts",
                    "abstract": "We present a new supervised learning procedure for systems composed of many separate networks, each of which learns to handle a subset of the complete set of training cases. The new procedure can be viewed either as a modular version of a multilayer supervised network, or as an associative version of competitive learning. It therefore provides a new link between these two apparently different approaches. We demonstrate that the learning procedure divides up a vowel discrimination task into appropriate subtasks, each of which can be solved by a very simple expert network."
                },
                {
                    "title": "Task Decomposition Through Competition in a Modular Connectionist Architecture: The What and Where Vision Tasks",
                    "abstract": "A novel modular connectionist architecture is presented in which the networks composing the architecture compete to learn the training patterns. An outcome of the competition is that different networks learn different training patterns and, thus, learn to compute different functions. The architecture performs task decomposition in the sense that it learns to partition a task into two or more functionally independent tasks and allocates distinct networks to learn each task. In addition, the architecture tends to allocate to each task the network whose topology is most appropriate to that task. The architecture's performance on \u201cwhat\u201d and \u201cwhere\u201d vision tasks is presented and compared with the performance of two multilayer networks. Finally, it is noted that function decomposition is an underconstrained problem, and, thus, different modular architectures may decompose a function in different ways. A desirable decomposition can be achieved if the architecture is suitably restricted in the types of functions that it can compute. Finding appropriate restrictions is possible through the application of domain knowledge. A strength of the modular architecture is that its structure is well suited for incorporating domain knowledge."
                },
                {
                    "title": "Einops: Clear and Reliable Tensor Manipulations with Einstein-like Notation",
                    "abstract": "Tensor computations underlie modern scienti\ufb01c computing and deep learning. A number of tensor frameworks emerged varying in execution model, hardware support, memory management, model de\ufb01nition, etc. However, tensor operations in all frameworks follow the same paradigm. Recent neural network architectures demonstrate demand for higher expressiveness of tensor operations. The current paradigm is not suited to write readable, reliable, or easy-to-modify code for multidimensional tensor manipulations. Moreover, some commonly used operations do not provide suf\ufb01cient checks and can break a tensor structure. These mistakes are elusive as no tools or tests can detect them. Independently, API discrepancies complicate code transfer between frameworks. We propose einops notation: a uniform and generic way to manipulate tensor structure, that signi\ufb01cantly improves code readability and \ufb02exibility by focusing on the structure of input and output tensors. We implement einops notation in a Python package that ef\ufb01ciently supports multiple widely used frameworks and provides framework-independent minimalist API for tensor manipulations."
                },
                {
                    "title": "Models with Conditional Computation Learn Suboptimal Solutions",
                    "abstract": "Sparsely-activated neural networks with conditional computation learn to route their inputs through different subnetworks, providing a strong structural prior and reducing computational costs. Despite their possible benefits, models with learned routing often underperform their parameter-matched densely-activated counterparts as well as models that use non-learned heuristic routing strategies. In this paper, we hypothesize that these shortcomings stem from the gradient estimation techniques used to train sparsely-activated models with non-differentiable discrete routing decisions. To test this hypothesis, we evaluate the performance of sparsely-activated models trained with various gradient estimation techniques in three settings where a high-quality heuristic routing strategy can be designed. Our experiments reveal that learned routing reaches substantially worse solutions than heuristic routing in various settings. As a first step towards remedying this gap, we demonstrate that supervising the routing decision on a small fraction of the examples is sufficient to help the model to learn better routing strategies. Our results shed light on the difficulties of learning effective routing and set the stage for future work on conditional computation mechanisms and training techniques."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively scale the Mixture of Experts (MoE) architecture to utilize a large number of experts while maintaining differentiability and improving interpretability and specialization in model outputs?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of conditional computation in large-scale models. By enabling fine-grained expert specialization, we can enhance model interpretability, making it easier to debug and correct biased behaviors. This has significant implications for the development of foundation models, particularly in ensuring fairness and safety in AI applications. Addressing this question could lead to practical applications in various domains, including natural language processing and computer vision, where understanding model decisions is essential.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to balance a large number of experts with the complexities of training stability and parameter efficiency. Naive approaches, such as simply increasing the number of experts, may lead to issues with discrete expert routing, resulting in training instability and inefficiencies. Additionally, achieving effective specialization without compromising the model's overall performance requires overcoming technical obstacles related to differentiability and the management of expert interactions.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on sparse MoE architectures, which suffer from non-differentiable routing mechanisms that hinder scalability and stability. Existing solutions have not adequately addressed the need for a large number of experts to specialize effectively without introducing training difficulties. Our approach, the Multilinear Mixture of Experts (\u03bcMoE), differs by providing a differentiable framework that allows for implicit computations on a factorized form of expert weights, thus overcoming the limitations of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the \u03bcMoE layer, which allows for the efficient scaling of tens of thousands of experts through differentiable operations. We will evaluate our approach using benchmark datasets relevant to large language and vision models, measuring performance through metrics such as accuracy, interpretability, and training stability. The expected outcomes include improved model performance with increased expert specialization, enhanced interpretability of model decisions, and a more robust framework for addressing biases in AI systems."
            }
        },
        "author_data": {
            "c992eed4-acf0-4a88-9c68-41a99a73bbcf": {
                "pk": "c992eed4-acf0-4a88-9c68-41a99a73bbcf",
                "name": "James Oldfield",
                "collaborators": [
                    "Yannis Panagakis",
                    "Mihalis A. Nicolaou",
                    "Ioannis Patras",
                    "Christos Tzelepis",
                    "Markos Georgopoulos",
                    "Jean Kossaifi",
                    "Grigorios G. Chrysos",
                    "Anima Anandkumar",
                    "Stefanos Zafeiriou",
                    "Georgios Tzimiropoulos"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Image Synthesis",
                    "Representation Learning",
                    "Data Augmentation"
                ],
                "publications": [
                    {
                        "title": "PandA: Unsupervised Learning of Parts and Appearances in the Feature Maps of GANs",
                        "abstract": "Recent advances in the understanding of Generative Adversarial Networks (GANs) have led to remarkable progress in visual editing and synthesis tasks, capitalizing on the rich semantics that are embedded in the latent spaces of pre-trained GANs. However, existing methods are often tailored to specific GAN architectures and are limited to either discovering global semantic directions that do not facilitate localized control, or require some form of supervision through manually provided regions or segmentation masks. In this light, we present an architecture-agnostic approach that jointly discovers factors representing spatial parts and their appearances in an entirely unsupervised fashion. These factors are obtained by applying a semi-nonnegative tensor factorization on the feature maps, which in turn enables context-aware local image editing with pixel-level control. In addition, we show that the discovered appearance factors correspond to saliency maps that localize concepts of interest, without using any labels. Experiments on a wide range of GAN architectures and datasets show that, in comparison to the state of the art, our method is far more efficient in terms of training time and, most importantly, provides much more accurate localized control. Our code is available at: https://github.com/james-oldfield/PandA."
                    },
                    {
                        "title": "Adversarial Learning of Disentangled and Generalizable Representations for Visual Attributes",
                        "abstract": "Recently, a multitude of methods for image-to-image translation have demonstrated impressive results on problems such as multi-domain or multi-attribute transfer. The vast majority of such works leverages the strengths of adversarial learning and deep convolutional autoencoders to achieve realistic results by well-capturing the target data distribution. Nevertheless, the most prominent representatives of this class of methods do not facilitate semantic structure in the latent space, and usually rely on binary domain labels for test-time transfer. This leads to rigid models, unable to capture the variance of each domain label. In this light, we propose a novel adversarial learning method that (i) facilitates the emergence of latent structure by semantically disentangling sources of variation, and (ii) encourages learning generalizable, continuous, and transferable latent codes that enable flexible attribute mixing. This is achieved by introducing a novel loss function that encourages representations to result in uniformly distributed class posteriors for disentangled attributes. In tandem with an algorithm for inducing generalizable properties, the resulting representations can be utilized for a variety of tasks such as intensity-preserving multi-attribute image translation and synthesis, without requiring labelled test data. We demonstrate the merits of the proposed method by a set of qualitative and quantitative experiments on popular databases such as MultiPIE, RaFD, and BU-3DFE, where our method outperforms other, state-of-the-art methods in tasks such as intensity-preserving multi-attribute transfer and synthesis."
                    },
                    {
                        "title": "Tensor Component Analysis for Interpreting the Latent Space of GANs",
                        "abstract": "This paper addresses the problem of finding interpretable directions in the latent space of pre-trained Generative Adversarial Networks (GANs) to facilitate controllable image synthesis. Such interpretable directions correspond to transformations that can affect both the style and geometry of the synthetic images. However, existing approaches that utilise linear techniques to find these transformations often fail to provide an intuitive way to separate these two sources of variation. To address this, we propose to a) perform a multilinear decomposition of the tensor of intermediate representations, and b) use a tensor-based regression to map directions found using this decomposition to the latent space. Our scheme allows for both linear edits corresponding to the individual modes of the tensor, and non-linear ones that model the multiplicative interactions between them. We show experimentally that we can utilise the former to better separate style- from geometry-based transformations, and the latter to generate an extended set of possible transformations in comparison to prior works. We demonstrate our approach's efficacy both quantitatively and qualitatively compared to the current state-of-the-art."
                    },
                    {
                        "title": "Parts of Speech-Grounded Subspaces in Vision-Language Models",
                        "abstract": "Latent image representations arising from vision-language models have proved immensely useful for a variety of downstream tasks. However, their utility is limited by their entanglement with respect to different visual attributes. For instance, recent work has shown that CLIP image representations are often biased toward specific visual properties (such as objects or actions) in an unpredictable manner. In this paper, we propose to separate representations of the different visual modalities in CLIP's joint vision-language space by leveraging the association between parts of speech and specific visual modes of variation (e.g. nouns relate to objects, adjectives describe appearance). This is achieved by formulating an appropriate component analysis model that learns subspaces capturing variability corresponding to a specific part of speech, while jointly minimising variability to the rest. Such a subspace yields disentangled representations of the different visual properties of an image or text in closed form while respecting the underlying geometry of the manifold on which the representations lie. What's more, we show the proposed model additionally facilitates learning subspaces corresponding to specific visual appearances (e.g. artists' painting styles), which enables the selective removal of entire visual themes from CLIP-based text-to-image synthesis. We validate the model both qualitatively, by visualising the subspace projections with a text-to-image model and by preventing the imitation of artists' styles, and quantitatively, through class invariance metrics and improvements to baseline zero-shot classification."
                    },
                    {
                        "title": "Tensor Methods in Computer Vision and Deep Learning",
                        "abstract": "Tensors, or multidimensional arrays, are data structures that can naturally represent visual data of multiple dimensions. Inherently able to efficiently capture structured, latent semantic spaces and high-order interactions, tensors have a long history of applications in a wide span of computer vision problems. With the advent of the deep learning paradigm shift in computer vision, tensors have become even more fundamental. Indeed, essential ingredients in modern deep learning architectures, such as convolutions and attention mechanisms, can readily be considered as tensor mappings. In effect, tensor methods are increasingly finding significant applications in deep learning, including the design of memory and compute efficient network architectures, improving robustness to random noise and adversarial attacks, and aiding the theoretical understanding of deep networks.   This article provides an in-depth and practical review of tensors and tensor methods in the context of representation learning and deep learning, with a particular focus on visual data analysis and computer vision applications. Concretely, besides fundamental work in tensor-based visual data analysis methods, we focus on recent developments that have brought on a gradual increase of tensor methods, especially in deep learning architectures, and their implications in computer vision applications. To further enable the newcomer to grasp such concepts quickly, we provide companion Python notebooks, covering key aspects of the paper and implementing them, step-by-step with TensorLy."
                    },
                    {
                        "title": "ContraCLIP: Interpretable GAN generation driven by pairs of contrasting sentences",
                        "abstract": "This work addresses the problem of discovering non-linear interpretable paths in the latent space of pre-trained GANs in a model-agnostic manner. In the proposed method, the discovery is driven by a set of pairs of natural language sentences with contrasting semantics, named semantic dipoles, that serve as the limits of the interpretation that we require by the trainable latent paths to encode. By using the pre-trained CLIP encoder, the sentences are projected into the vision-language space, where they serve as dipoles, and where RBF-based warping functions define a set of non-linear directional paths, one for each semantic dipole, allowing in this way traversals from one semantic pole to the other. By defining an objective that discovers paths in the latent space of GANs that generate changes along the desired paths in the vision-language embedding space, we provide an intuitive way of controlling the underlying generative factors and address some of the limitations of the state-of-the-art works, namely, that a) they are typically tailored to specific GAN architectures (i.e., StyleGAN), b) they disregard the relative position of the manipulated and the original image in the image embedding and the relative position of the image and the text embeddings, and c) they lead to abrupt image manipulations and quickly arrive at regions of low density and, thus, low image quality, providing limited control of the generative factors. We provide extensive qualitative and quantitative results that demonstrate our claims with two pre-trained GANs, and make the code and the pre-trained models publicly available at: https://github.com/chi0tzp/ContraCLIP"
                    },
                    {
                        "title": "Enhancing Facial Data Diversity with Style-based Face Aging",
                        "abstract": "A significant limiting factor in training fair classifiers relates to the presence of dataset bias. In particular, face datasets are typically biased in terms of attributes such as gender, age, and race. If not mitigated, bias leads to algorithms that exhibit unfair behaviour towards such groups. In this work, we address the problem of increasing the diversity of face datasets with respect to age. Concretely, we propose a novel, generative style-based architecture for data augmentation that captures fine-grained aging patterns by conditioning on multi-resolution age-discriminative representations. By evaluating on several age-annotated datasets in both single- and cross-database experiments, we show that the proposed method outperforms state-of-the-art algorithms for age transfer, especially in the case of age groups that lie in the tails of the label distribution. We further show significantly increased diversity in the augmented datasets, outperforming all compared methods according to established metrics."
                    },
                    {
                        "title": "Cluster-guided Image Synthesis with Unconditional Models",
                        "abstract": "Generative Adversarial Networks (GANs) are the driving force behind the state-of-the-art in image generation. Despite their ability to synthesize high-resolution photo-realistic images, generating content with on-demand conditioning of different granularity remains a challenge. This challenge is usually tackled by annotating massive datasets with the attributes of interest, a laborious task that is not always a viable option. Therefore, it is vital to introduce control into the generation process of unsupervised generative models. In this work, we focus on controllable image generation by leveraging GANs that are well-trained in an unsupervised fashion. To this end, we discover that the representation space of intermediate layers of the generator forms a number of clusters that separate the data according to semantically meaningful attributes (e.g., hair color and pose). By conditioning on the cluster assignments, the proposed method is able to control the semantic class of the generated image. Our approach enables sampling from each cluster by Implicit Maximum Likelihood Estimation (IMLE). We showcase the efficacy of our approach on faces (CelebA-HQ and FFHQ), animals (Imagenet) and objects (LSUN) using different pre-trained generative models. The results highlight the ability of our approach to condition image generation on attributes like gender, pose and hair style on faces, as well as a variety of features on different object classes."
                    }
                ]
            },
            "63ef86d4-fb47-400d-8f73-527086b1cf55": {
                "pk": "63ef86d4-fb47-400d-8f73-527086b1cf55",
                "name": "Markos Georgopoulos",
                "collaborators": [
                    "Yannis Panagakis",
                    "Maja Pantic",
                    "Matthias Nie\u00dfner",
                    "Grigorios G Chrysos",
                    "James Oldfield",
                    "Tobias Kirschstein",
                    "Simon Giebenhain",
                    "Martin R\u00fcnz",
                    "Lourdes Agapito",
                    "Mihalis A. Nicolaou"
                ],
                "domain": [
                    "Computer Vision",
                    "Generative Models",
                    "Deep Learning",
                    "Fairness in AI"
                ],
                "publications": [
                    {
                        "title": "Investigating Bias in Deep Face Analysis: The KANFace Dataset and Empirical Study",
                        "abstract": "Deep learning-based methods have pushed the limits of the state-of-the-art in face analysis. However, despite their success, these models have raised concerns regarding their bias towards certain demographics. This bias is inflicted both by limited diversity across demographics in the training set, as well as the design of the algorithms. In this work, we investigate the demographic bias of deep learning models in face recognition, age estimation, gender recognition and kinship verification. To this end, we introduce the most comprehensive, large-scale dataset of facial images and videos to date. It consists of 40K still images and 44K sequences (14.5M video frames in total) captured in unconstrained, real-world conditions from 1,045 subjects. The data are manually annotated in terms of identity, exact age, gender and kinship. The performance of state-of-the-art models is scrutinized and demographic bias is exposed by conducting a series of experiments. Lastly, a method to debias network embeddings is introduced and tested on the proposed benchmarks."
                    },
                    {
                        "title": "Modeling of Facial Aging and Kinship: A Survey",
                        "abstract": "Computational facial models that capture properties of facial cues related to aging and kinship increasingly attract the attention of the research community, enabling the development of reliable methods for age progression, age estimation, age-invariant facial characterization, and kinship verification from visual data. In this paper, we review recent advances in modeling of facial aging and kinship. In particular, we provide an up-to date, complete list of available annotated datasets and an in-depth analysis of geometric, hand-crafted, and learned facial representations that are used for facial aging and kinship characterization. Moreover, evaluation protocols and metrics are reviewed and notable experimental results for each surveyed task are analyzed. This survey allows us to identify challenges and discuss future research directions for the development of robust facial models in real-world conditions."
                    },
                    {
                        "title": "Multilinear Latent Conditioning for Generating Unseen Attribute Combinations",
                        "abstract": "Deep generative models rely on their inductive bias to facilitate generalization, especially for problems with high dimensional data, like images. However, empirical studies have shown that variational autoencoders (VAE) and generative adversarial networks (GAN) lack the generalization ability that occurs naturally in human perception. For example, humans can visualize a woman smiling after only seeing a smiling man. On the contrary, the standard conditional VAE (cVAE) is unable to generate unseen attribute combinations. To this end, we extend cVAE by introducing a multilinear latent conditioning framework that captures the multiplicative interactions between the attributes. We implement two variants of our model and demonstrate their efficacy on MNIST, Fashion-MNIST and CelebA. Altogether, we design a novel conditioning framework that can be used with any architecture to synthesize unseen attribute combinations."
                    },
                    {
                        "title": "CoPE: Conditional image generation using Polynomial Expansions",
                        "abstract": "Generative modeling has evolved to a notable field of machine learning. Deep polynomial neural networks (PNNs) have demonstrated impressive results in unsupervised image generation, where the task is to map an input vector (i.e., noise) to a synthesized image. However, the success of PNNs has not been replicated in conditional generation tasks, such as super-resolution. Existing PNNs focus on single-variable polynomial expansions which do not fare well to two-variable inputs, i.e., the noise variable and the conditional variable. In this work, we introduce a general framework, called CoPE, that enables a polynomial expansion of two input variables and captures their auto- and cross-correlations. We exhibit how CoPE can be trivially augmented to accept an arbitrary number of input variables. CoPE is evaluated in five tasks (class-conditional generation, inverse problems, edges-to-image translation, image-to-image translation, attribute-guided generation) involving eight datasets. The thorough evaluation suggests that CoPE can be useful for tackling diverse conditional generation tasks. The source code of CoPE is available at \\url{https://github.com/grigorisg9gr/polynomial_nets_for_conditional_generation}."
                    },
                    {
                        "title": "Cluster-guided Image Synthesis with Unconditional Models",
                        "abstract": "Generative Adversarial Networks (GANs) are the driving force behind the state-of-the-art in image generation. Despite their ability to synthesize high-resolution photo-realistic images, generating content with on-demand conditioning of different granularity remains a challenge. This challenge is usually tackled by annotating massive datasets with the attributes of interest, a laborious task that is not always a viable option. Therefore, it is vital to introduce control into the generation process of unsupervised generative models. In this work, we focus on controllable image generation by leveraging GANs that are well-trained in an unsupervised fashion. To this end, we discover that the representation space of intermediate layers of the generator forms a number of clusters that separate the data according to semantically meaningful attributes (e.g., hair color and pose). By conditioning on the cluster assignments, the proposed method is able to control the semantic class of the generated image. Our approach enables sampling from each cluster by Implicit Maximum Likelihood Estimation (IMLE). We showcase the efficacy of our approach on faces (CelebA-HQ and FFHQ), animals (Imagenet) and objects (LSUN) using different pre-trained generative models. The results highlight the ability of our approach to condition image generation on attributes like gender, pose and hair style on faces, as well as a variety of features on different object classes."
                    },
                    {
                        "title": "Enhancing Facial Data Diversity with Style-based Face Aging",
                        "abstract": "A significant limiting factor in training fair classifiers relates to the presence of dataset bias. In particular, face datasets are typically biased in terms of attributes such as gender, age, and race. If not mitigated, bias leads to algorithms that exhibit unfair behaviour towards such groups. In this work, we address the problem of increasing the diversity of face datasets with respect to age. Concretely, we propose a novel, generative style-based architecture for data augmentation that captures fine-grained aging patterns by conditioning on multi-resolution age-discriminative representations. By evaluating on several age-annotated datasets in both single- and cross-database experiments, we show that the proposed method outperforms state-of-the-art algorithms for age transfer, especially in the case of age groups that lie in the tails of the label distribution. We further show significantly increased diversity in the augmented datasets, outperforming all compared methods according to established metrics."
                    },
                    {
                        "title": "Multilinear Operator Networks",
                        "abstract": "Despite the remarkable capabilities of deep neural networks in image recognition, the dependence on activation functions remains a largely unexplored area and has yet to be eliminated. On the other hand, Polynomial Networks is a class of models that does not require activation functions, but have yet to perform on par with modern architectures. In this work, we aim close this gap and propose MONet, which relies solely on multilinear operators. The core layer of MONet, called Mu-Layer, captures multiplicative interactions of the elements of the input token. MONet captures high-degree interactions of the input elements and we demonstrate the efficacy of our approach on a series of image recognition and scientific computing benchmarks. The proposed model outperforms prior polynomial networks and performs on par with modern architectures. We believe that MONet can inspire further research on models that use entirely multilinear operations."
                    },
                    {
                        "title": "Leveraging the Context through Multi-Round Interactions for Jailbreaking Attacks",
                        "abstract": "Large Language Models (LLMs) are susceptible to Jailbreaking attacks, which aim to extract harmful information by subtly modifying the attack query. As defense mechanisms evolve, directly obtaining harmful information becomes increasingly challenging for Jailbreaking attacks. In this work, inspired from Chomsky's transformational-generative grammar theory and human practices of indirect context to elicit harmful information, we focus on a new attack form, called Contextual Interaction Attack. We contend that the prior context\\u2014the information preceding the attack query\\u2014plays a pivotal role in enabling strong Jailbreaking attacks. Specifically, we propose a first multi-turn approach that leverages benign preliminary questions to interact with the LLM. Due to the autoregressive nature of LLMs, which use previous conversation rounds as context during generation, we guide the model's question-response pair to construct a context that is semantically aligned with the attack query to execute the attack. We conduct experiments on seven different LLMs and demonstrate the efficacy of this attack, which is black-box and can also transfer across LLMs. We believe this can lead to further developments and understanding of security in LLMs."
                    },
                    {
                        "title": "GGHead: Fast and Generalizable 3D Gaussian Heads",
                        "abstract": "Learning 3D head priors from large 2D image collections is an important step towards high-quality 3D-aware human modeling. A core requirement is an efficient architecture that scales well to large-scale datasets and large image resolutions. Unfortunately, existing 3D GANs struggle to scale to generate samples at high resolutions due to their relatively slow train and render speeds, and typically have to rely on 2D superresolution networks at the expense of global 3D consistency. To address these challenges, we propose Generative Gaussian Heads (GGHead), which adopts the recent 3D Gaussian Splatting representation within a 3D GAN framework. To generate a 3D representation, we employ a powerful 2D CNN generator to predict Gaussian attributes in the UV space of a template head mesh. This way, GGHead exploits the regularity of the template's UV layout, substantially facilitating the challenging task of predicting an unstructured set of 3D Gaussians. We further improve the geometric fidelity of the generated 3D representations with a novel total variation loss on rendered UV coordinates. Intuitively, this regularization encourages that neighboring rendered pixels should stem from neighboring Gaussians in the template's UV space. Taken together, our pipeline can efficiently generate 3D heads trained only from single-view 2D image observations. Our proposed framework matches the quality of existing 3D head GANs on FFHQ while being both substantially faster and fully 3D consistent. As a result, we demonstrate real-time generation and rendering of high-quality 3D-consistent heads at $1024^2$ resolution for the first time. Project Website: https://tobias-kirschstein.github.io/gghead"
                    },
                    {
                        "title": "Augmenting Deep Classifiers with Polynomial Neural Networks",
                        "abstract": "Deep neural networks have been the driving force behind the success in classification tasks, e.g., object and audio recognition. Impressive results and generalization have been achieved by a variety of recently proposed architectures, the majority of which are seemingly disconnected. In this work, we cast the study of deep classifiers under a unifying framework. In particular, we express state-of-the-art architectures (e.g., residual and non-local networks) in the form of different degree polynomials of the input. Our framework provides insights on the inductive biases of each model and enables natural extensions building upon their polynomial nature. The efficacy of the proposed models is evaluated on standard image and audio classification benchmarks. The expressivity of the proposed models is highlighted both in terms of increased model performance as well as model compression. Lastly, the extensions allowed by this taxonomy showcase benefits in the presence of limited data and long-tailed data distributions. We expect this taxonomy to provide links between existing domain-specific architectures. The source code is available at \\url{https://github.com/grigorisg9gr/polynomials-for-augmenting-NNs}."
                    },
                    {
                        "title": "Tensor Component Analysis for Interpreting the Latent Space of GANs",
                        "abstract": "This paper addresses the problem of finding interpretable directions in the latent space of pre-trained Generative Adversarial Networks (GANs) to facilitate controllable image synthesis. Such interpretable directions correspond to transformations that can affect both the style and geometry of the synthetic images. However, existing approaches that utilise linear techniques to find these transformations often fail to provide an intuitive way to separate these two sources of variation. To address this, we propose to a) perform a multilinear decomposition of the tensor of intermediate representations, and b) use a tensor-based regression to map directions found using this decomposition to the latent space. Our scheme allows for both linear edits corresponding to the individual modes of the tensor, and non-linear ones that model the multiplicative interactions between them. We show experimentally that we can utilise the former to better separate style- from geometry-based transformations, and the latter to generate an extended set of possible transformations in comparison to prior works. We demonstrate our approach's efficacy both quantitatively and qualitatively compared to the current state-of-the-art."
                    },
                    {
                        "title": "Learning Neural Parametric Head Models",
                        "abstract": "We propose a novel 3D morphable model for complete human heads based on hybrid neural fields. At the core of our model lies a neural parametric representation that disentangles identity and expressions in disjoint latent spaces. To this end, we capture a person's identity in a canonical space as a signed distance field (SDF), and model facial expressions with a neural deformation field. In addition, our representation achieves high-fidelity local detail by introducing an ensemble of local fields centered around facial anchor points. To facilitate generalization, we train our model on a newly-captured dataset of over 5200 head scans from 255 different identities using a custom high-end 3D scanning setup. Our dataset significantly exceeds comparable existing datasets, both with respect to quality and completeness of geometry, averaging around 3.5M mesh faces per scan. Finally, we demonstrate that our approach outperforms state-of-the-art methods in terms of fitting error and reconstruction quality."
                    },
                    {
                        "title": "MonoNPHM: Dynamic Head Reconstruction from Monocular Videos",
                        "abstract": "We present Monocular Neural Parametric Head Models (MonoNPHM) for dynamic 3D head reconstructions from monocular RGB videos. To this end, we propose a latent appearance space that parameterizes a texture field on top of a neural parametric model. We constrain predicted color values to be correlated with the underlying geometry such that gradients from RGB effectively influence latent geometry codes during inverse rendering. To increase the representational capacity of our expression space, we augment our backward deformation field with hyper-dimensions, thus improving color and geometry representation in topologically challenging expressions. Using MonoNPHM as a learned prior, we approach the task of 3D head reconstruction using signed distance field based volumetric rendering. By numerically inverting our backward deformation field, we incorporated a landmark loss using facial anchor points that are closely tied to our canonical geometry representation. To evaluate the task of dynamic face reconstruction from monocular RGB videos we record 20 challenging Kinect sequences under casual conditions. MonoNPHM outperforms all baselines with a significant margin, and makes an important step towards easily accessible neural parametric face models through RGB tracking."
                    },
                    {
                        "title": "HumanRF: High-Fidelity Neural Radiance Fields for Humans in Motion",
                        "abstract": "Representing human performance at high-fidelity is an essential building block in diverse applications, such as film production, computer games or videoconferencing. To close the gap to production-level quality, we introduce HumanRF, a 4D dynamic neural scene representation that captures full-body appearance in motion from multi-view video input, and enables playback from novel, unseen viewpoints. Our novel representation acts as a dynamic video encoding that captures fine details at high compression rates by factorizing space-time into a temporal matrix-vector decomposition. This allows us to obtain temporally coherent reconstructions of human actors for long sequences, while representing high-resolution details even in the context of challenging motion. While most research focuses on synthesizing at resolutions of 4MP or lower, we address the challenge of operating at 12MP. To this end, we introduce ActorsHQ, a novel multi-view dataset that provides 12MP footage from 160 cameras for 16 sequences with high-fidelity, per-frame mesh reconstructions. We demonstrate challenges that emerge from using such high-resolution data and show that our newly introduced HumanRF effectively leverages this data, making a significant step towards production-level quality novel view synthesis."
                    }
                ]
            },
            "d344260c-ee20-413a-ba92-250e85c33a68": {
                "pk": "d344260c-ee20-413a-ba92-250e85c33a68",
                "name": "Grigorios G. Chrysos",
                "collaborators": [
                    "Stefanos Zafeiriou",
                    "Volkan Cevher",
                    "Yannis Panagakis",
                    "Aggelina Chatziagapi",
                    "Dimitris Samaras",
                    "Yixin Cheng",
                    "Markos Georgopoulos",
                    "Jean Kossaifi",
                    "Paolo Favaro",
                    "Thomas Pethick"
                ],
                "domain": [
                    "Computer Vision",
                    "Generative Models",
                    "Neural Networks",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Robust Conditional Generative Adversarial Networks",
                        "abstract": "Conditional generative adversarial networks (cGAN) have led to large improvements in the task of conditional image generation, which lies at the heart of computer vision. The major focus so far has been on performance improvement, while there has been little effort in making cGAN more robust to noise. The regression (of the generator) might lead to arbitrarily large errors in the output, which makes cGAN unreliable for real-world applications. In this work, we introduce a novel conditional GAN model, called RoCGAN, which leverages structure in the target space of the model to address the issue. Our model augments the generator with an unsupervised pathway, which promotes the outputs of the generator to span the target manifold even in the presence of intense noise. We prove that RoCGAN share similar theoretical properties as GAN and experimentally verify that our model outperforms existing state-of-the-art cGAN architectures by a large margin in a variety of domains including images from natural scenes and faces."
                    },
                    {
                        "title": "Deep Face Deblurring",
                        "abstract": "Blind deblurring consists a long studied task, however the outcomes of generic methods are not effective in real world blurred images. Domain-specific methods for deblurring targeted object categories, e.g. text or faces, frequently outperform their generic counterparts, hence they are attracting an increasing amount of attention. In this work, we develop such a domain-specific method to tackle deblurring of human faces, henceforth referred to as face deblurring. Studying faces is of tremendous significance in computer vision, however face deblurring has yet to demonstrate some convincing results. This can be partly attributed to the combination of i) poor texture and ii) highly structure shape that yield the contour/gradient priors (that are typically used) sub-optimal. In our work instead of making assumptions over the prior, we adopt a learning approach by inserting weak supervision that exploits the well-documented structure of the face. Namely, we utilise a deep network to perform the deblurring and employ a face alignment technique to pre-process each face. We additionally surpass the requirement of the deep network for thousands training samples, by introducing an efficient framework that allows the generation of a large dataset. We utilised this framework to create 2MF2, a dataset of over two million frames. We conducted experiments with real world blurred facial images and report that our method returns a result close to the sharp natural latent image."
                    },
                    {
                        "title": "Visual Data Augmentation through Learning",
                        "abstract": "The rapid progress in machine learning methods has been empowered by i) huge datasets that have been collected and annotated, ii) improved engineering (e.g. data pre-processing/normalization). The existing datasets typically include several million samples, which constitutes their extension a colossal task. In addition, the state-of-the-art data-driven methods demand a vast amount of data, hence a standard engineering trick employed is artificial data augmentation for instance by adding into the data cropped and (affinely) transformed images. However, this approach does not correspond to any change in the natural 3D scene.   We propose instead to perform data augmentation through learning realistic local transformations. We learn a forward and an inverse transformation that maps an image from the high-dimensional space of pixel intensities to a latent space which varies (approximately) linearly with the latent space of a realistically transformed version of the image. Such transformed images can be considered two successive frames in a video. Next, we utilize these transformations to learn a linear model that modifies the latent spaces and then use the inverse transformation to synthesize a new image. We argue that the this procedure produces powerful invariant representations. We perform both qualitative and quantitative experiments that demonstrate our proposed method creates new realistic images."
                    },
                    {
                        "title": "Motion deblurring of faces",
                        "abstract": "Face analysis is a core part of computer vision, in which remarkable progress has been observed in the past decades. Current methods achieve recognition and tracking with invariance to fundamental modes of variation such as illumination, 3D pose, expressions. Notwithstanding, a much less standing mode of variation is motion deblurring, which however presents substantial challenges in face analysis. Recent approaches either make oversimplifying assumptions, e.g. in cases of joint optimization with other tasks, or fail to preserve the highly structured shape/identity information. Therefore, we propose a data-driven method that encourages identity preservation. The proposed model includes two parallel streams (sub-networks): the first deblurs the image, the second implicitly extracts and projects the identity of both the sharp and the blurred image in similar subspaces. We devise a method for creating realistic motion blur by averaging a variable number of frames to train our model. The averaged images originate from a 2MF2 dataset with 10 million facial frames, which we introduce for the task. Considering deblurring as an intermediate step, we utilize the deblurred outputs to conduct a thorough experimentation on high-level face analysis tasks, i.e. landmark localization and face verification. The experimental evaluation demonstrates the superiority of our method."
                    },
                    {
                        "title": "Revisiting adversarial training for the worst-performing class",
                        "abstract": "Despite progress in adversarial training (AT), there is a substantial gap between the top-performing and worst-performing classes in many datasets. For example, on CIFAR10, the accuracies for the best and worst classes are 74% and 23%, respectively. We argue that this gap can be reduced by explicitly optimizing for the worst-performing class, resulting in a min-max-max optimization formulation. Our method, called class focused online learning (CFOL), includes high probability convergence guarantees for the worst class loss and can be easily integrated into existing training setups with minimal computational overhead. We demonstrate an improvement to 32% in the worst class accuracy on CIFAR10, and we observe consistent behavior across CIFAR100 and STL10. Our study highlights the importance of moving beyond average accuracy, which is particularly important in safety-critical applications."
                    },
                    {
                        "title": "MI-NeRF: Learning a Single Face NeRF from Multiple Identities",
                        "abstract": "In this work, we introduce a method that learns a single dynamic neural radiance field (NeRF) from monocular talking face videos of multiple identities. NeRFs have shown remarkable results in modeling the 4D dynamics and appearance of human faces. However, they require per-identity optimization. Although recent approaches have proposed techniques to reduce the training and rendering time, increasing the number of identities can be expensive. We introduce MI-NeRF (multi-identity NeRF), a single unified network that models complex non-rigid facial motion for multiple identities, using only monocular videos of arbitrary length. The core premise in our method is to learn the non-linear interactions between identity and non-identity specific information with a multiplicative module. By training on multiple videos simultaneously, MI-NeRF not only reduces the total training time compared to standard single-identity NeRFs, but also demonstrates robustness in synthesizing novel expressions for any input identity. We present results for both facial expression transfer and talking face video synthesis. Our method can be further personalized for a target identity given only a short video."
                    },
                    {
                        "title": "MIGS: Multi-Identity Gaussian Splatting via Tensor Decomposition",
                        "abstract": "We introduce MIGS (Multi-Identity Gaussian Splatting), a novel method that learns a single neural representation for multiple identities, using only monocular videos. Recent 3D Gaussian Splatting (3DGS) approaches for human avatars require per-identity optimization. However, learning a multi-identity representation presents advantages in robustly animating humans under arbitrary poses. We propose to construct a high-order tensor that combines all the learnable 3DGS parameters for all the training identities. By assuming a low-rank structure and factorizing the tensor, we model the complex rigid and non-rigid deformations of multiple subjects in a unified network, significantly reducing the total number of parameters. Our proposed approach leverages information from all the training identities and enables robust animation under challenging unseen poses, outperforming existing approaches. It can also be extended to learn unseen identities."
                    },
                    {
                        "title": "Self-Supervised Neural Architecture Search for Imbalanced Datasets",
                        "abstract": "Neural Architecture Search (NAS) provides state-of-the-art results when trained on well-curated datasets with annotated labels. However, annotating data or even having balanced number of samples can be a luxury for practitioners from different scientific fields, e.g., in the medical domain. To that end, we propose a NAS-based framework that bears the threefold contributions: (a) we focus on the self-supervised scenario, i.e., where no labels are required to determine the architecture, and (b) we assume the datasets are imbalanced, (c) we design each component to be able to run on a resource constrained setup, i.e., on a single GPU (e.g. Google Colab). Our components build on top of recent developments in self-supervised learning~\\citep{zbontar2021barlow}, self-supervised NAS~\\citep{kaplan2020self} and extend them for the case of imbalanced datasets. We conduct experiments on an (artificially) imbalanced version of CIFAR-10 and we demonstrate our proposed method outperforms standard neural networks, while using $27\\times$ less parameters. To validate our assumption on a naturally imbalanced dataset, we also conduct experiments on ChestMNIST and COVID-19 X-ray. The results demonstrate how the proposed method can be used in imbalanced datasets, while it can be fully run on a single GPU. Code is available \\href{https://github.com/TimofeevAlex/ssnas_imbalanced}{here}."
                    },
                    {
                        "title": "Multilinear Operator Networks",
                        "abstract": "Despite the remarkable capabilities of deep neural networks in image recognition, the dependence on activation functions remains a largely unexplored area and has yet to be eliminated. On the other hand, Polynomial Networks is a class of models that does not require activation functions, but have yet to perform on par with modern architectures. In this work, we aim close this gap and propose MONet, which relies solely on multilinear operators. The core layer of MONet, called Mu-Layer, captures multiplicative interactions of the elements of the input token. MONet captures high-degree interactions of the input elements and we demonstrate the efficacy of our approach on a series of image recognition and scientific computing benchmarks. The proposed model outperforms prior polynomial networks and performs on par with modern architectures. We believe that MONet can inspire further research on models that use entirely multilinear operations."
                    },
                    {
                        "title": "Leveraging the Context through Multi-Round Interactions for Jailbreaking Attacks",
                        "abstract": "Large Language Models (LLMs) are susceptible to Jailbreaking attacks, which aim to extract harmful information by subtly modifying the attack query. As defense mechanisms evolve, directly obtaining harmful information becomes increasingly challenging for Jailbreaking attacks. In this work, inspired from Chomsky's transformational-generative grammar theory and human practices of indirect context to elicit harmful information, we focus on a new attack form, called Contextual Interaction Attack. We contend that the prior context\\u2014the information preceding the attack query\\u2014plays a pivotal role in enabling strong Jailbreaking attacks. Specifically, we propose a first multi-turn approach that leverages benign preliminary questions to interact with the LLM. Due to the autoregressive nature of LLMs, which use previous conversation rounds as context during generation, we guide the model's question-response pair to construct a context that is semantically aligned with the attack query to execute the attack. We conduct experiments on seven different LLMs and demonstrate the efficacy of this attack, which is black-box and can also transfer across LLMs. We believe this can lead to further developments and understanding of security in LLMs."
                    },
                    {
                        "title": "A Comprehensive Performance Evaluation of Deformable Face Tracking \"In-the-Wild\"",
                        "abstract": "Recently, technologies such as face detection, facial landmark localisation and face recognition and verification have matured enough to provide effective and efficient solutions for imagery captured under arbitrary conditions (referred to as \"in-the-wild\"). This is partially attributed to the fact that comprehensive \"in-the-wild\" benchmarks have been developed for face detection, landmark localisation and recognition/verification. A very important technology that has not been thoroughly evaluated yet is deformable face tracking \"in-the-wild\". Until now, the performance has mainly been assessed qualitatively by visually assessing the result of a deformable face tracking technology on short videos. In this paper, we perform the first, to the best of our knowledge, thorough evaluation of state-of-the-art deformable face tracking pipelines using the recently introduced 300VW benchmark. We evaluate many different architectures focusing mainly on the task of on-line deformable face tracking. In particular, we compare the following general strategies: (a) generic face detection plus generic facial landmark localisation, (b) generic model free tracking plus generic facial landmark localisation, as well as (c) hybrid approaches using state-of-the-art face detection, model free tracking and facial landmark localisation technologies. Our evaluation reveals future avenues for further research on the topic."
                    },
                    {
                        "title": "$\u03a0-$nets: Deep Polynomial Neural Networks",
                        "abstract": "Deep Convolutional Neural Networks (DCNNs) is currently the method of choice both for generative, as well as for discriminative learning in computer vision and machine learning. The success of DCNNs can be attributed to the careful selection of their building blocks (e.g., residual blocks, rectifiers, sophisticated normalization schemes, to mention but a few). In this paper, we propose $\\Pi$-Nets, a new class of DCNNs. $\\Pi$-Nets are polynomial neural networks, i.e., the output is a high-order polynomial of the input. $\\Pi$-Nets can be implemented using special kind of skip connections and their parameters can be represented via high-order tensors. We empirically demonstrate that $\\Pi$-Nets have better representation power than standard DCNNs and they even produce good results without the use of non-linear activation functions in a large battery of tasks and signals, i.e., images, graphs, and audio. When used in conjunction with activation functions, $\\Pi$-Nets produce state-of-the-art results in challenging tasks, such as image generation. Lastly, our framework elucidates why recent generative models, such as StyleGAN, improve upon their predecessors, e.g., ProGAN."
                    }
                ]
            },
            "f2f2df62-1418-4cee-b607-8b2da437fdae": {
                "pk": "f2f2df62-1418-4cee-b607-8b2da437fdae",
                "name": "Christos Tzelepis",
                "collaborators": [
                    "Ioannis Patras",
                    "Georgios Tzimiropoulos",
                    "Stella Bounareli",
                    "Vasileios Argyriou",
                    "James Oldfield",
                    "Ting-Ting Xie",
                    "Vasileios Mezaris",
                    "Nicu Sebe",
                    "Yannis Panagakis",
                    "Mihalis A. Nicolaou"
                ],
                "domain": [
                    "Generative Models",
                    "Video Analysis",
                    "Uncertainty Quantification",
                    "Fairness in AI"
                ],
                "publications": [
                    {
                        "title": "Uncertainty Propagation in Convolutional Neural Networks: Technical Report",
                        "abstract": "In this technical report we study the problem of propagation of uncertainty (in terms of variances of given uni-variate normal random variables) through typical building blocks of a Convolutional Neural Network (CNN). These include layers that perform linear operations, such as 2D convolutions, fully-connected, and average pooling layers, as well as layers that act non-linearly on their input, such as the Rectified Linear Unit (ReLU). Finally, we discuss the sigmoid function, for which we give approximations of its first- and second-order moments, as well as the binary cross-entropy loss function, for which we approximate its expected value under normal random inputs."
                    },
                    {
                        "title": "Learning to detect video events from zero or very few video examples",
                        "abstract": "In this work we deal with the problem of high-level event detection in video. Specifically, we study the challenging problems of i) learning to detect video events from solely a textual description of the event, without using any positive video examples, and ii) additionally exploiting very few positive training samples together with a small number of ``related'' videos. For learning only from an event's textual description, we first identify a general learning framework and then study the impact of different design choices for various stages of this framework. For additionally learning from example videos, when true positive training samples are scarce, we employ an extension of the Support Vector Machine that allows us to exploit ``related'' event videos by automatically introducing different weights for subsets of the videos in the overall training set. Experimental evaluations performed on the large-scale TRECVID MED 2014 video dataset provide insight on the effectiveness of the proposed methods."
                    },
                    {
                        "title": "Linear Maximum Margin Classifier for Learning from Uncertain Data",
                        "abstract": "In this paper, we propose a maximum margin classifier that deals with uncertainty in data input. More specifically, we reformulate the SVM framework such that each training example can be modeled by a multi-dimensional Gaussian distribution described by its mean vector and its covariance matrix -- the latter modeling the uncertainty. We address the classification problem and define a cost function that is the expected value of the classical SVM cost when data samples are drawn from the multi-dimensional Gaussian distributions that form the set of the training examples. Our formulation approximates the classical SVM formulation when the training examples are isotropic Gaussians with variance tending to zero. We arrive at a convex optimization problem, which we solve efficiently in the primal form using a stochastic gradient descent approach. The resulting classifier, which we name SVM with Gaussian Sample Uncertainty (SVM-GSU), is tested on synthetic data and five publicly available and popular datasets; namely, the MNIST, WDBC, DEAP, TV News Channel Commercial Detection, and TRECVID MED datasets. Experimental results verify the effectiveness of the proposed method."
                    },
                    {
                        "title": "WarpedGANSpace: Finding non-linear RBF paths in GAN latent space",
                        "abstract": "This work addresses the problem of discovering, in an unsupervised manner, interpretable paths in the latent space of pretrained GANs, so as to provide an intuitive and easy way of controlling the underlying generative factors. In doing so, it addresses some of the limitations of the state-of-the-art works, namely, a) that they discover directions that are independent of the latent code, i.e., paths that are linear, and b) that their evaluation relies either on visual inspection or on laborious human labeling. More specifically, we propose to learn non-linear warpings on the latent space, each one parametrized by a set of RBF-based latent space warping functions, and where each warping gives rise to a family of non-linear paths via the gradient of the function. Building on the work of Voynov and Babenko, that discovers linear paths, we optimize the trainable parameters of the set of RBFs, so as that images that are generated by codes along different paths, are easily distinguishable by a discriminator network. This leads to easily distinguishable image transformations, such as pose and facial expressions in facial images. We show that linear paths can be derived as a special case of our method, and show experimentally that non-linear paths in the latent space lead to steeper, more disentangled and interpretable changes in the image space than in state-of-the art methods, both qualitatively and quantitatively. We make the code and the pretrained models publicly available at: https://github.com/chi0tzp/WarpedGANSpace."
                    },
                    {
                        "title": "ContraCLIP: Interpretable GAN generation driven by pairs of contrasting sentences",
                        "abstract": "This work addresses the problem of discovering non-linear interpretable paths in the latent space of pre-trained GANs in a model-agnostic manner. In the proposed method, the discovery is driven by a set of pairs of natural language sentences with contrasting semantics, named semantic dipoles, that serve as the limits of the interpretation that we require by the trainable latent paths to encode. By using the pre-trained CLIP encoder, the sentences are projected into the vision-language space, where they serve as dipoles, and where RBF-based warping functions define a set of non-linear directional paths, one for each semantic dipole, allowing in this way traversals from one semantic pole to the other. By defining an objective that discovers paths in the latent space of GANs that generate changes along the desired paths in the vision-language embedding space, we provide an intuitive way of controlling the underlying generative factors and address some of the limitations of the state-of-the-art works, namely, that a) they are typically tailored to specific GAN architectures (i.e., StyleGAN), b) they disregard the relative position of the manipulated and the original image in the image embedding and the relative position of the image and the text embeddings, and c) they lead to abrupt image manipulations and quickly arrive at regions of low density and, thus, low image quality, providing limited control of the generative factors. We provide extensive qualitative and quantitative results that demonstrate our claims with two pre-trained GANs, and make the code and the pre-trained models publicly available at: https://github.com/chi0tzp/ContraCLIP"
                    },
                    {
                        "title": "Attribute-preserving Face Dataset Anonymization via Latent Code Optimization",
                        "abstract": "This work addresses the problem of anonymizing the identity of faces in a dataset of images, such that the privacy of those depicted is not violated, while at the same time the dataset is useful for downstream task such as for training machine learning models. To the best of our knowledge, we are the first to explicitly address this issue and deal with two major drawbacks of the existing state-of-the-art approaches, namely that they (i) require the costly training of additional, purpose-trained neural networks, and/or (ii) fail to retain the facial attributes of the original images in the anonymized counterparts, the preservation of which is of paramount importance for their use in downstream tasks. We accordingly present a task-agnostic anonymization procedure that directly optimizes the images' latent representation in the latent space of a pre-trained GAN. By optimizing the latent codes directly, we ensure both that the identity is of a desired distance away from the original (with an identity obfuscation loss), whilst preserving the facial attributes (using a novel feature-matching loss in FaRL's deep feature space). We demonstrate through a series of both qualitative and quantitative experiments that our method is capable of anonymizing the identity of the images whilst -- crucially -- better-preserving the facial attributes. We make the code and the pre-trained models publicly available at: https://github.com/chi0tzp/FALCO."
                    },
                    {
                        "title": "Boundary Uncertainty in a Single-Stage Temporal Action Localization Network",
                        "abstract": "In this paper, we address the problem of temporal action localization with a single stage neural network. In the proposed architecture we model the boundary predictions as uni-variate Gaussian distributions in order to model their uncertainties, which is the first in this area to the best of our knowledge. We use two uncertainty-aware boundary regression losses: first, the Kullback-Leibler divergence between the ground truth location of the boundary and the Gaussian modeling the prediction of the boundary and second, the expectation of the $\\ell_1$ loss under the same Gaussian. We show that with both uncertainty modeling approaches improve the detection performance by more than $1.5\\%$ in mAP@tIoU=0.5 and that the proposed simple one-stage network performs closely to more complex one and two stage networks."
                    },
                    {
                        "title": "Are CLIP features all you need for Universal Synthetic Image Origin Attribution?",
                        "abstract": "The steady improvement of Diffusion Models for visual synthesis has given rise to many new and interesting use cases of synthetic images but also has raised concerns about their potential abuse, which poses significant societal threats. To address this, fake images need to be detected and attributed to their source model, and given the frequent release of new generators, realistic applications need to consider an Open-Set scenario where some models are unseen at training time. Existing forensic techniques are either limited to Closed-Set settings or to GAN-generated images, relying on fragile frequency-based \"fingerprint\" features. By contrast, we propose a simple yet effective framework that incorporates features from large pre-trained foundation models to perform Open-Set origin attribution of synthetic images produced by various generative models, including Diffusion Models. We show that our method leads to remarkable attribution performance, even in the low-data regime, exceeding the performance of existing methods and generalizes better on images obtained from a diverse set of architectures. We make the code publicly available at: https://github.com/ciodar/UniversalAttribution."
                    },
                    {
                        "title": "Temporal Action Localization with Variance-Aware Networks",
                        "abstract": "This work addresses the problem of temporal action localization with Variance-Aware Networks (VAN), i.e., DNNs that use second-order statistics in the input and/or the output of regression tasks. We first propose a network (VANp) that when presented with the second-order statistics of the input, i.e., each sample has a mean and a variance, it propagates the mean and the variance throughout the network to deliver outputs with second order statistics. In this framework, both the input and the output could be interpreted as Gaussians. To do so, we derive differentiable analytic solutions, or reasonable approximations, to propagate across commonly used NN layers. To train the network, we define a differentiable loss based on the KL-divergence between the predicted Gaussian and a Gaussian around the ground truth action borders, and use standard back-propagation. Importantly, the variances propagation in VANp does not require any additional parameters, and during testing, does not require any additional computations either. In action localization, the means and the variances of the input are computed at pooling operations, that are typically used to bring arbitrarily long videos to a vector with fixed dimensions. Second, we propose two alternative formulations that augment the first (respectively, the last) layer of a regression network with additional parameters so as to take in the input (respectively, predict in the output) both means and variances. Results in the action localization problem show that the incorporation of second order statistics improves over the baseline network, and that VANp surpasses the accuracy of virtually all other two-stage networks without involving any additional parameters."
                    },
                    {
                        "title": "Improving Fairness using Vision-Language Driven Image Augmentation",
                        "abstract": "Fairness is crucial when training a deep-learning discriminative model, especially in the facial domain. Models tend to correlate specific characteristics (such as age and skin color) with unrelated attributes (downstream tasks), resulting in biases which do not correspond to reality. It is common knowledge that these correlations are present in the data and are then transferred to the models during training. This paper proposes a method to mitigate these correlations to improve fairness. To do so, we learn interpretable and meaningful paths lying in the semantic space of a pre-trained diffusion model (DiffAE) -- such paths being supervised by contrastive text dipoles. That is, we learn to edit protected characteristics (age and skin color). These paths are then applied to augment images to improve the fairness of a given dataset. We test the proposed method on CelebA-HQ and UTKFace on several downstream tasks with age and skin color as protected characteristics. As a proxy for fairness, we compute the difference in accuracy with respect to the protected characteristics. Quantitative results show how the augmented images help the model improve the overall accuracy, the aforementioned metric, and the disparity of equal opportunity. Code is available at: https://github.com/Moreno98/Vision-Language-Bias-Control."
                    },
                    {
                        "title": "Few-Shot Action Localization without Knowing Boundaries",
                        "abstract": "Learning to localize actions in long, cluttered, and untrimmed videos is a hard task, that in the literature has typically been addressed assuming the availability of large amounts of annotated training samples for each class -- either in a fully-supervised setting, where action boundaries are known, or in a weakly-supervised setting, where only class labels are known for each video. In this paper, we go a step further and show that it is possible to learn to localize actions in untrimmed videos when a) only one/few trimmed examples of the target action are available at test time, and b) when a large collection of videos with only class label annotation (some trimmed and some weakly annotated untrimmed ones) are available for training; with no overlap between the classes used during training and testing. To do so, we propose a network that learns to estimate Temporal Similarity Matrices (TSMs) that model a fine-grained similarity pattern between pairs of videos (trimmed or untrimmed), and uses them to generate Temporal Class Activation Maps (TCAMs) for seen or unseen classes. The TCAMs serve as temporal attention mechanisms to extract video-level representations of untrimmed videos, and to temporally localize actions at test time. To the best of our knowledge, we are the first to propose a weakly-supervised, one/few-shot action localization network that can be trained in an end-to-end fashion. Experimental results on THUMOS14 and ActivityNet1.2 datasets, show that our method achieves performance comparable or better to state-of-the-art fully-supervised, few-shot learning methods."
                    },
                    {
                        "title": "HyperReenact: One-Shot Reenactment via Jointly Learning to Refine and Retarget Faces",
                        "abstract": "In this paper, we present our method for neural face reenactment, called HyperReenact, that aims to generate realistic talking head images of a source identity, driven by a target facial pose. Existing state-of-the-art face reenactment methods train controllable generative models that learn to synthesize realistic facial images, yet producing reenacted faces that are prone to significant visual artifacts, especially under the challenging condition of extreme head pose changes, or requiring expensive few-shot fine-tuning to better preserve the source identity characteristics. We propose to address these limitations by leveraging the photorealistic generation ability and the disentangled properties of a pretrained StyleGAN2 generator, by first inverting the real images into its latent space and then using a hypernetwork to perform: (i) refinement of the source identity characteristics and (ii) facial pose re-targeting, eliminating this way the dependence on external editing methods that typically produce artifacts. Our method operates under the one-shot setting (i.e., using a single source frame) and allows for cross-subject reenactment, without requiring any subject-specific fine-tuning. We compare our method both quantitatively and qualitatively against several state-of-the-art techniques on the standard benchmarks of VoxCeleb1 and VoxCeleb2, demonstrating the superiority of our approach in producing artifact-free images, exhibiting remarkable robustness even under extreme head pose changes. We make the code and the pretrained models publicly available at: https://github.com/StelaBou/HyperReenact ."
                    },
                    {
                        "title": "One-shot Neural Face Reenactment via Finding Directions in GAN's Latent Space",
                        "abstract": "In this paper, we present our framework for neural face/head reenactment whose goal is to transfer the 3D head orientation and expression of a target face to a source face. Previous methods focus on learning embedding networks for identity and head pose/expression disentanglement which proves to be a rather hard task, degrading the quality of the generated images. We take a different approach, bypassing the training of such networks, by using (fine-tuned) pre-trained GANs which have been shown capable of producing high-quality facial images. Because GANs are characterized by weak controllability, the core of our approach is a method to discover which directions in latent GAN space are responsible for controlling head pose and expression variations. We present a simple pipeline to learn such directions with the aid of a 3D shape model which, by construction, inherently captures disentangled directions for head pose, identity, and expression. Moreover, we show that by embedding real images in the GAN latent space, our method can be successfully used for the reenactment of real-world faces. Our method features several favorable properties including using a single source image (one-shot) and enabling cross-person reenactment. Extensive qualitative and quantitative results show that our approach typically produces reenacted faces of notably higher quality than those produced by state-of-the-art methods for the standard benchmarks of VoxCeleb1 & 2."
                    },
                    {
                        "title": "MM2Latent: Text-to-facial image generation and editing in GANs with multimodal assistance",
                        "abstract": "Generating human portraits is a hot topic in the image generation area, e.g. mask-to-face generation and text-to-face generation. However, these unimodal generation methods lack controllability in image generation. Controllability can be enhanced by exploring the advantages and complementarities of various modalities. For instance, we can utilize the advantages of text in controlling diverse attributes and masks in controlling spatial locations. Current state-of-the-art methods in multimodal generation face limitations due to their reliance on extensive hyperparameters, manual operations during the inference stage, substantial computational demands during training and inference, or inability to edit real images. In this paper, we propose a practical framework - MM2Latent - for multimodal image generation and editing. We use StyleGAN2 as our image generator, FaRL for text encoding, and train an autoencoders for spatial modalities like mask, sketch and 3DMM. We propose a strategy that involves training a mapping network to map the multimodal input into the w latent space of StyleGAN. The proposed framework 1) eliminates hyperparameters and manual operations in the inference stage, 2) ensures fast inference speeds, and 3) enables the editing of real images. Extensive experiments demonstrate that our method exhibits superior performance in multimodal image generation, surpassing recent GAN- and diffusion-based methods. Also, it proves effective in multimodal image editing and is faster than GAN- and diffusion-based methods. We make the code publicly available at: https://github.com/Open-Debin/MM2Latent"
                    },
                    {
                        "title": "DiffusionAct: Controllable Diffusion Autoencoder for One-shot Face Reenactment",
                        "abstract": "Video-driven neural face reenactment aims to synthesize realistic facial images that successfully preserve the identity and appearance of a source face, while transferring the target head pose and facial expressions. Existing GAN-based methods suffer from either distortions and visual artifacts or poor reconstruction quality, i.e., the background and several important appearance details, such as hair style/color, glasses and accessories, are not faithfully reconstructed. Recent advances in Diffusion Probabilistic Models (DPMs) enable the generation of high-quality realistic images. To this end, in this paper we present DiffusionAct, a novel method that leverages the photo-realistic image generation of diffusion models to perform neural face reenactment. Specifically, we propose to control the semantic space of a Diffusion Autoencoder (DiffAE), in order to edit the facial pose of the input images, defined as the head pose orientation and the facial expressions. Our method allows one-shot, self, and cross-subject reenactment, without requiring subject-specific fine-tuning. We compare against state-of-the-art GAN-, StyleGAN2-, and diffusion-based methods, showing better or on-par reenactment performance."
                    },
                    {
                        "title": "Ensembled Cold-Diffusion Restorations for Unsupervised Anomaly Detection",
                        "abstract": "Unsupervised Anomaly Detection (UAD) methods aim to identify anomalies in test samples comparing them with a normative distribution learned from a dataset known to be anomaly-free. Approaches based on generative models offer interpretability by generating anomaly-free versions of test images, but are typically unable to identify subtle anomalies. Alternatively, approaches using feature modelling or self-supervised methods, such as the ones relying on synthetically generated anomalies, do not provide out-of-the-box interpretability. In this work, we present a novel method that combines the strengths of both strategies: a generative cold-diffusion pipeline (i.e., a diffusion-like pipeline which uses corruptions not based on noise) that is trained with the objective of turning synthetically-corrupted images back to their normal, original appearance. To support our pipeline we introduce a novel synthetic anomaly generation procedure, called DAG, and a novel anomaly score which ensembles restorations conditioned with different degrees of abnormality. Our method surpasses the prior state-of-the art for unsupervised anomaly detection in three different Brain MRI datasets."
                    },
                    {
                        "title": "DnS: Distill-and-Select for Efficient and Accurate Video Indexing and Retrieval",
                        "abstract": "In this paper, we address the problem of high performance and computationally efficient content-based video retrieval in large-scale datasets. Current methods typically propose either: (i) fine-grained approaches employing spatio-temporal representations and similarity calculations, achieving high performance at a high computational cost or (ii) coarse-grained approaches representing/indexing videos as global vectors, where the spatio-temporal structure is lost, providing low performance but also having low computational cost. In this work, we propose a Knowledge Distillation framework, called Distill-and-Select (DnS), that starting from a well-performing fine-grained Teacher Network learns: a) Student Networks at different retrieval performance and computational efficiency trade-offs and b) a Selector Network that at test time rapidly directs samples to the appropriate student to maintain both high retrieval performance and high computational efficiency. We train several students with different architectures and arrive at different trade-offs of performance and efficiency, i.e., speed and storage requirements, including fine-grained students that store/index videos using binary representations. Importantly, the proposed scheme allows Knowledge Distillation in large, unlabelled datasets -- this leads to good students. We evaluate DnS on five public datasets on three different video retrieval tasks and demonstrate a) that our students achieve state-of-the-art performance in several cases and b) that the DnS framework provides an excellent trade-off between retrieval performance, computational speed, and storage space. In specific configurations, the proposed method achieves similar mAP with the teacher but is 20 times faster and requires 240 times less storage space. The collected dataset and implementation are publicly available: https://github.com/mever-team/distill-and-select."
                    },
                    {
                        "title": "StyleMask: Disentangling the Style Space of StyleGAN2 for Neural Face Reenactment",
                        "abstract": "In this paper we address the problem of neural face reenactment, where, given a pair of a source and a target facial image, we need to transfer the target's pose (defined as the head pose and its facial expressions) to the source image, by preserving at the same time the source's identity characteristics (e.g., facial shape, hair style, etc), even in the challenging case where the source and the target faces belong to different identities. In doing so, we address some of the limitations of the state-of-the-art works, namely, a) that they depend on paired training data (i.e., source and target faces have the same identity), b) that they rely on labeled data during inference, and c) that they do not preserve identity in large head pose changes. More specifically, we propose a framework that, using unpaired randomly generated facial images, learns to disentangle the identity characteristics of the face from its pose by incorporating the recently introduced style space $\\mathcal{S}$ of StyleGAN2, a latent representation space that exhibits remarkable disentanglement properties. By capitalizing on this, we learn to successfully mix a pair of source and target style codes using supervision from a 3D model. The resulting latent code, that is subsequently used for reenactment, consists of latent units corresponding to the facial pose of the target only and of units corresponding to the identity of the source only, leading to notable improvement in the reenactment performance compared to recent state-of-the-art methods. In comparison to state of the art, we quantitatively and qualitatively show that the proposed method produces higher quality results even on extreme pose variations. Finally, we report results on real images by first embedding them on the latent space of the pretrained generator. We make the code and pretrained models publicly available at: https://github.com/StelaBou/StyleMask"
                    },
                    {
                        "title": "PandA: Unsupervised Learning of Parts and Appearances in the Feature Maps of GANs",
                        "abstract": "Recent advances in the understanding of Generative Adversarial Networks (GANs) have led to remarkable progress in visual editing and synthesis tasks, capitalizing on the rich semantics that are embedded in the latent spaces of pre-trained GANs. However, existing methods are often tailored to specific GAN architectures and are limited to either discovering global semantic directions that do not facilitate localized control, or require some form of supervision through manually provided regions or segmentation masks. In this light, we present an architecture-agnostic approach that jointly discovers factors representing spatial parts and their appearances in an entirely unsupervised fashion. These factors are obtained by applying a semi-nonnegative tensor factorization on the feature maps, which in turn enables context-aware local image editing with pixel-level control. In addition, we show that the discovered appearance factors correspond to saliency maps that localize concepts of interest, without using any labels. Experiments on a wide range of GAN architectures and datasets show that, in comparison to the state of the art, our method is far more efficient in terms of training time and, most importantly, provides much more accurate localized control. Our code is available at: https://github.com/james-oldfield/PandA."
                    },
                    {
                        "title": "Parts of Speech-Grounded Subspaces in Vision-Language Models",
                        "abstract": "Latent image representations arising from vision-language models have proved immensely useful for a variety of downstream tasks. However, their utility is limited by their entanglement with respect to different visual attributes. For instance, recent work has shown that CLIP image representations are often biased toward specific visual properties (such as objects or actions) in an unpredictable manner. In this paper, we propose to separate representations of the different visual modalities in CLIP's joint vision-language space by leveraging the association between parts of speech and specific visual modes of variation (e.g. nouns relate to objects, adjectives describe appearance). This is achieved by formulating an appropriate component analysis model that learns subspaces capturing variability corresponding to a specific part of speech, while jointly minimising variability to the rest. Such a subspace yields disentangled representations of the different visual properties of an image or text in closed form while respecting the underlying geometry of the manifold on which the representations lie. What's more, we show the proposed model additionally facilitates learning subspaces corresponding to specific visual appearances (e.g. artists' painting styles), which enables the selective removal of entire visual themes from CLIP-based text-to-image synthesis. We validate the model both qualitatively, by visualising the subspace projections with a text-to-image model and by preventing the imitation of artists' styles, and quantitatively, through class invariance metrics and improvements to baseline zero-shot classification."
                    }
                ]
            },
            "322a790b-ea6d-4ff3-9295-b8acce5caa72": {
                "pk": "322a790b-ea6d-4ff3-9295-b8acce5caa72",
                "name": "Yannis Panagakis",
                "collaborators": [
                    "Stefanos Zafeiriou",
                    "Maja Pantic",
                    "Markos Georgopoulos",
                    "Vicky Kouni",
                    "Grigorios Chrysos",
                    "Niannan Xue",
                    "Triantafyllos Kefalas",
                    "Mehdi Bahri",
                    "Mihalis A. Nicolaou",
                    "Jean Kossaifi"
                ],
                "domain": [
                    "Generative Models",
                    "Compressed Sensing",
                    "Deep Learning",
                    "Image Processing"
                ],
                "publications": [
                    {
                        "title": "DECONET: an Unfolding Network for Analysis-based Compressed Sensing with Generalization Error Bounds",
                        "abstract": "We present a new deep unfolding network for analysis-sparsity-based Compressed Sensing. The proposed network coined Decoding Network (DECONET) jointly learns a decoder that reconstructs vectors from their incomplete, noisy measurements and a redundant sparsifying analysis operator, which is shared across the layers of DECONET. Moreover, we formulate the hypothesis class of DECONET and estimate its associated Rademacher complexity. Then, we use this estimate to deliver meaningful upper bounds for the generalization error of DECONET. Finally, the validity of our theoretical results is assessed and comparisons to state-of-the-art unfolding networks are made, on both synthetic and real-world datasets. Experimental results indicate that our proposed network outperforms the baselines, consistently for all datasets, and its behaviour complies with our theoretical findings."
                    },
                    {
                        "title": "Generalization analysis of an unfolding network for analysis-based Compressed Sensing",
                        "abstract": "Unfolding networks have shown promising results in the Compressed Sensing (CS) field. Yet, the investigation of their generalization ability is still in its infancy. In this paper, we perform generalization analysis of a state-of-the-art ADMM-based unfolding network, which jointly learns a decoder for CS and a sparsifying redundant analysis operator. To this end, we first impose a structural constraint on the learnable sparsifier, which parametrizes the network's hypothesis class. For the latter, we estimate its Rademacher complexity. With this estimate in hand, we deliver generalization error bounds for the examined network. Finally, the validity of our theory is assessed and numerical comparisons to a state-of-the-art unfolding network are made, on synthetic and real-world datasets. Our experimental results demonstrate that our proposed framework complies with our theoretical findings and outperforms the baseline, consistently for all datasets."
                    },
                    {
                        "title": "GAGAN: Geometry-Aware Generative Adversarial Networks",
                        "abstract": "Deep generative models learned through adversarial training have become increasingly popular for their ability to generate naturalistic image textures. However, aside from their texture, the visual appearance of objects is significantly influenced by their shape geometry; information which is not taken into account by existing generative models. This paper introduces the Geometry-Aware Generative Adversarial Networks (GAGAN) for incorporating geometric information into the image generation process. Specifically, in GAGAN the generator samples latent variables from the probability space of a statistical shape model. By mapping the output of the generator to a canonical coordinate frame through a differentiable geometric transformation, we enforce the geometry of the objects and add an implicit connection from the prior to the generated object. Experimental results on face generation indicate that the GAGAN can generate realistic images of faces with arbitrary facial attributes such as facial expression, pose, and morphology, that are of better quality than current GAN-based methods. Our method can be used to augment any existing GAN architecture and improve the quality of the images generated."
                    },
                    {
                        "title": "PolyGAN: High-Order Polynomial Generators",
                        "abstract": "Generative Adversarial Networks (GANs) have become the gold standard when it comes to learning generative models for high-dimensional distributions. Since their advent, numerous variations of GANs have been introduced in the literature, primarily focusing on utilization of novel loss functions, optimization/regularization strategies and network architectures. In this paper, we turn our attention to the generator and investigate the use of high-order polynomials as an alternative class of universal function approximators. Concretely, we propose PolyGAN, where we model the data generator by means of a high-order polynomial whose unknown parameters are naturally represented by high-order tensors. We introduce two tensor decompositions that significantly reduce the number of parameters and show how they can be efficiently implemented by hierarchical neural networks that only employ linear/convolutional blocks. We exhibit for the first time that by using our approach a GAN generator can approximate the data distribution without using any activation functions. Thorough experimental evaluation on both synthetic and real data (images and 3D point clouds) demonstrates the merits of PolyGAN against the state of the art."
                    },
                    {
                        "title": "Investigating Bias in Deep Face Analysis: The KANFace Dataset and Empirical Study",
                        "abstract": "Deep learning-based methods have pushed the limits of the state-of-the-art in face analysis. However, despite their success, these models have raised concerns regarding their bias towards certain demographics. This bias is inflicted both by limited diversity across demographics in the training set, as well as the design of the algorithms. In this work, we investigate the demographic bias of deep learning models in face recognition, age estimation, gender recognition and kinship verification. To this end, we introduce the most comprehensive, large-scale dataset of facial images and videos to date. It consists of 40K still images and 44K sequences (14.5M video frames in total) captured in unconstrained, real-world conditions from 1,045 subjects. The data are manually annotated in terms of identity, exact age, gender and kinship. The performance of state-of-the-art models is scrutinized and demographic bias is exposed by conducting a series of experiments. Lastly, a method to debias network embeddings is introduced and tested on the proposed benchmarks."
                    },
                    {
                        "title": "Informed Non-convex Robust Principal Component Analysis with Features",
                        "abstract": "We revisit the problem of robust principal component analysis with features acting as prior side information. To this aim, a novel, elegant, non-convex optimization approach is proposed to decompose a given observation matrix into a low-rank core and the corresponding sparse residual. Rigorous theoretical analysis of the proposed algorithm results in exact recovery guarantees with low computational complexity. Aptly designed synthetic experiments demonstrate that our method is the first to wholly harness the power of non-convexity over convexity in terms of both recoverability and speed. That is, the proposed non-convex approach is more accurate and faster compared to the best available algorithms for the problem under study. Two real-world applications, namely image classification and face denoising further exemplify the practical superiority of the proposed method."
                    },
                    {
                        "title": "Modeling of Facial Aging and Kinship: A Survey",
                        "abstract": "Computational facial models that capture properties of facial cues related to aging and kinship increasingly attract the attention of the research community, enabling the development of reliable methods for age progression, age estimation, age-invariant facial characterization, and kinship verification from visual data. In this paper, we review recent advances in modeling of facial aging and kinship. In particular, we provide an up-to date, complete list of available annotated datasets and an in-depth analysis of geometric, hand-crafted, and learned facial representations that are used for facial aging and kinship characterization. Moreover, evaluation protocols and metrics are reviewed and notable experimental results for each surveyed task are analyzed. This survey allows us to identify challenges and discuss future research directions for the development of robust facial models in real-world conditions."
                    },
                    {
                        "title": "Multilinear Latent Conditioning for Generating Unseen Attribute Combinations",
                        "abstract": "Deep generative models rely on their inductive bias to facilitate generalization, especially for problems with high dimensional data, like images. However, empirical studies have shown that variational autoencoders (VAE) and generative adversarial networks (GAN) lack the generalization ability that occurs naturally in human perception. For example, humans can visualize a woman smiling after only seeing a smiling man. On the contrary, the standard conditional VAE (cVAE) is unable to generate unseen attribute combinations. To this end, we extend cVAE by introducing a multilinear latent conditioning framework that captures the multiplicative interactions between the attributes. We implement two variants of our model and demonstrate their efficacy on MNIST, Fashion-MNIST and CelebA. Altogether, we design a novel conditioning framework that can be used with any architecture to synthesize unseen attribute combinations."
                    },
                    {
                        "title": "Side Information in Robust Principal Component Analysis: Algorithms and Applications",
                        "abstract": "Robust Principal Component Analysis (RPCA) aims at recovering a low-rank subspace from grossly corrupted high-dimensional (often visual) data and is a cornerstone in many machine learning and computer vision applications. Even though RPCA has been shown to be very successful in solving many rank minimisation problems, there are still cases where degenerate or suboptimal solutions are obtained. This is likely to be remedied by taking into account of domain-dependent prior knowledge. In this paper, we propose two models for the RPCA problem with the aid of side information on the low-rank structure of the data. The versatility of the proposed methods is demonstrated by applying them to four applications, namely background subtraction, facial image denoising, face and facial expression recognition. Experimental results on synthetic and five real world datasets indicate the robustness and effectiveness of the proposed methods on these application domains, largely outperforming six previous approaches."
                    },
                    {
                        "title": "CoPE: Conditional image generation using Polynomial Expansions",
                        "abstract": "Generative modeling has evolved to a notable field of machine learning. Deep polynomial neural networks (PNNs) have demonstrated impressive results in unsupervised image generation, where the task is to map an input vector (i.e., noise) to a synthesized image. However, the success of PNNs has not been replicated in conditional generation tasks, such as super-resolution. Existing PNNs focus on single-variable polynomial expansions which do not fare well to two-variable inputs, i.e., the noise variable and the conditional variable. In this work, we introduce a general framework, called CoPE, that enables a polynomial expansion of two input variables and captures their auto- and cross-correlations. We exhibit how CoPE can be trivially augmented to accept an arbitrary number of input variables. CoPE is evaluated in five tasks (class-conditional generation, inverse problems, edges-to-image translation, image-to-image translation, attribute-guided generation) involving eight datasets. The thorough evaluation suggests that CoPE can be useful for tackling diverse conditional generation tasks. The source code of CoPE is available at \\url{https://github.com/grigorisg9gr/polynomial_nets_for_conditional_generation}."
                    },
                    {
                        "title": "Audio-visual video-to-speech synthesis with synthesized input audio",
                        "abstract": "Video-to-speech synthesis involves reconstructing the speech signal of a speaker from a silent video. The implicit assumption of this task is that the sound signal is either missing or contains a high amount of noise/corruption such that it is not useful for processing. Previous works in the literature either use video inputs only or employ both video and audio inputs during training, and discard the input audio pathway during inference. In this work we investigate the effect of using video and audio inputs for video-to-speech synthesis during both training and inference. In particular, we use pre-trained video-to-speech models to synthesize the missing speech signals and then train an audio-visual-to-speech synthesis model, using both the silent video and the synthesized speech as inputs, to predict the final reconstructed speech. Our experiments demonstrate that this approach is successful with both raw waveforms and mel spectrograms as target outputs."
                    },
                    {
                        "title": "Robust Kronecker Component Analysis",
                        "abstract": "Dictionary learning and component analysis models are fundamental for learning compact representations that are relevant to a given task (feature extraction, dimensionality reduction, denoising, etc.). The model complexity is encoded by means of specific structure, such as sparsity, low-rankness, or nonnegativity. Unfortunately, approaches like K-SVD - that learn dictionaries for sparse coding via Singular Value Decomposition (SVD) - are hard to scale to high-volume and high-dimensional visual data, and fragile in the presence of outliers. Conversely, robust component analysis methods such as the Robust Principal Component Analysis (RPCA) are able to recover low-complexity (e.g., low-rank) representations from data corrupted with noise of unknown magnitude and support, but do not provide a dictionary that respects the structure of the data (e.g., images), and also involve expensive computations. In this paper, we propose a novel Kronecker-decomposable component analysis model, coined as Robust Kronecker Component Analysis (RKCA), that combines ideas from sparse dictionary learning and robust component analysis. RKCA has several appealing properties, including robustness to gross corruption; it can be used for low-rank modeling, and leverages separability to solve significantly smaller problems. We design an efficient learning algorithm by drawing links with a restricted form of tensor factorization, and analyze its optimality and low-rankness properties. The effectiveness of the proposed approach is demonstrated on real-world applications, namely background subtraction and image denoising and completion, by performing a thorough comparison with the current state of the art."
                    },
                    {
                        "title": "Visual Data Augmentation through Learning",
                        "abstract": "The rapid progress in machine learning methods has been empowered by i) huge datasets that have been collected and annotated, ii) improved engineering (e.g. data pre-processing/normalization). The existing datasets typically include several million samples, which constitutes their extension a colossal task. In addition, the state-of-the-art data-driven methods demand a vast amount of data, hence a standard engineering trick employed is artificial data augmentation for instance by adding into the data cropped and (affinely) transformed images. However, this approach does not correspond to any change in the natural 3D scene.   We propose instead to perform data augmentation through learning realistic local transformations. We learn a forward and an inverse transformation that maps an image from the high-dimensional space of pixel intensities to a latent space which varies (approximately) linearly with the latent space of a realistically transformed version of the image. Such transformed images can be considered two successive frames in a video. Next, we utilize these transformations to learn a linear model that modifies the latent spaces and then use the inverse transformation to synthesize a new image. We argue that the this procedure produces powerful invariant representations. We perform both qualitative and quantitative experiments that demonstrate our proposed method creates new realistic images."
                    },
                    {
                        "title": "Adversarial Learning of Disentangled and Generalizable Representations for Visual Attributes",
                        "abstract": "Recently, a multitude of methods for image-to-image translation have demonstrated impressive results on problems such as multi-domain or multi-attribute transfer. The vast majority of such works leverages the strengths of adversarial learning and deep convolutional autoencoders to achieve realistic results by well-capturing the target data distribution. Nevertheless, the most prominent representatives of this class of methods do not facilitate semantic structure in the latent space, and usually rely on binary domain labels for test-time transfer. This leads to rigid models, unable to capture the variance of each domain label. In this light, we propose a novel adversarial learning method that (i) facilitates the emergence of latent structure by semantically disentangling sources of variation, and (ii) encourages learning generalizable, continuous, and transferable latent codes that enable flexible attribute mixing. This is achieved by introducing a novel loss function that encourages representations to result in uniformly distributed class posteriors for disentangled attributes. In tandem with an algorithm for inducing generalizable properties, the resulting representations can be utilized for a variety of tasks such as intensity-preserving multi-attribute image translation and synthesis, without requiring labelled test data. We demonstrate the merits of the proposed method by a set of qualitative and quantitative experiments on popular databases such as MultiPIE, RaFD, and BU-3DFE, where our method outperforms other, state-of-the-art methods in tasks such as intensity-preserving multi-attribute transfer and synthesis."
                    },
                    {
                        "title": "Non-Negative Matrix Factorizations for Multiplex Network Analysis",
                        "abstract": "Networks have been a general tool for representing, analyzing, and modeling relational data arising in several domains. One of the most important aspect of network analysis is community detection or network clustering. Until recently, the major focus have been on discovering community structure in single (i.e., monoplex) networks. However, with the advent of relational data with multiple modalities, multiplex networks, i.e., networks composed of multiple layers representing different aspects of relations, have emerged. Consequently, community detection in multiplex network, i.e., detecting clusters of nodes shared by all layers, has become a new challenge. In this paper, we propose Network Fusion for Composite Community Extraction (NF-CCE), a new class of algorithms, based on four different non-negative matrix factorization models, capable of extracting composite communities in multiplex networks. Each algorithm works in two steps: first, it finds a non-negative, low-dimensional feature representation of each network layer; then, it fuses the feature representation of layers into a common non-negative, low-dimensional feature representation via collective factorization. The composite clusters are extracted from the common feature representation. We demonstrate the superior performance of our algorithms over the state-of-the-art methods on various types of multiplex networks, including biological, social, economic, citation, phone communication, and brain multiplex networks."
                    },
                    {
                        "title": "Robust Kronecker-Decomposable Component Analysis for Low-Rank Modeling",
                        "abstract": "Dictionary learning and component analysis are part of one of the most well-studied and active research fields, at the intersection of signal and image processing, computer vision, and statistical machine learning. In dictionary learning, the current methods of choice are arguably K-SVD and its variants, which learn a dictionary (i.e., a decomposition) for sparse coding via Singular Value Decomposition. In robust component analysis, leading methods derive from Principal Component Pursuit (PCP), which recovers a low-rank matrix from sparse corruptions of unknown magnitude and support. However, K-SVD is sensitive to the presence of noise and outliers in the training set. Additionally, PCP does not provide a dictionary that respects the structure of the data (e.g., images), and requires expensive SVD computations when solved by convex relaxation. In this paper, we introduce a new robust decomposition of images by combining ideas from sparse dictionary learning and PCP. We propose a novel Kronecker-decomposable component analysis which is robust to gross corruption, can be used for low-rank modeling, and leverages separability to solve significantly smaller problems. We design an efficient learning algorithm by drawing links with a restricted form of tensor factorization. The effectiveness of the proposed approach is demonstrated on real-world applications, namely background subtraction and image denoising, by performing a thorough comparison with the current state of the art."
                    },
                    {
                        "title": "Fast Multilevel Algorithms for Compressive Principle Component Pursuit",
                        "abstract": "Recovering a low-rank matrix from highly corrupted measurements arises in compressed sensing of structured high-dimensional signals (e.g., videos and hyperspectral images among others). Robust principal component analysis (RPCA), solved via principal component pursuit (PCP), recovers a low-rank matrix from sparse corruptions that are of unknown value and support by decomposing the observation matrix into two terms: a low-rank matrix and a sparse one, accounting for sparse noise and outliers. In the more general setting, where only a fraction of the data matrix has been observed, low-rank matrix recovery is achieved by solving the compressive principle component pursuit (CPCP). Both PCP and CPCP are well-studied convex programs, and numerous iterative algorithms have been proposed for their optimisation. Nevertheless, these algorithms involve singular value decomposition (SVD) at each iteration, which renders their applicability challenging in the case of massive data. In this paper, we propose a multilevel approach for the solution of PCP and CPCP problems. The core principle behind our algorithm is to apply SVD in models of lower-dimensionality than the original one and then lift its solution to the original problem dimension. We show that the proposed algorithms are easy to implement, converge at the same rate but with much lower iteration cost. Numerical experiments on numerous synthetic and real problems indicate that the proposed multilevel algorithms are several times faster than their original counterparts, namely PCP and CPCP."
                    },
                    {
                        "title": "Unsupervised Discovery of Semantic Concepts in Satellite Imagery with Style-based Wavelet-driven Generative Models",
                        "abstract": "In recent years, considerable advancements have been made in the area of Generative Adversarial Networks (GANs), particularly with the advent of style-based architectures that address many key shortcomings - both in terms of modeling capabilities and network interpretability. Despite these improvements, the adoption of such approaches in the domain of satellite imagery is not straightforward. Typical vision datasets used in generative tasks are well-aligned and annotated, and exhibit limited variability. In contrast, satellite imagery exhibits great spatial and spectral variability, wide presence of fine, high-frequency details, while the tedious nature of annotating satellite imagery leads to annotation scarcity - further motivating developments in unsupervised learning. In this light, we present the first pre-trained style- and wavelet-based GAN model that can readily synthesize a wide gamut of realistic satellite images in a variety of settings and conditions - while also preserving high-frequency information. Furthermore, we show that by analyzing the intermediate activations of our network, one can discover a multitude of interpretable semantic directions that facilitate the guided synthesis of satellite images in terms of high-level concepts (e.g., urbanization) without using any form of supervision. Via a set of qualitative and quantitative experiments we demonstrate the efficacy of our framework, in terms of suitability for downstream tasks (e.g., data augmentation), quality of synthetic imagery, as well as generalization capabilities to unseen datasets."
                    },
                    {
                        "title": "Scale Equivariant Graph Metanetworks",
                        "abstract": "This paper pertains to an emerging machine learning paradigm: learning higher-order functions, i.e. functions whose inputs are functions themselves, $\\textit{particularly when these inputs are Neural Networks (NNs)}$. With the growing interest in architectures that process NNs, a recurring design principle has permeated the field: adhering to the permutation symmetries arising from the connectionist structure of NNs. $\\textit{However, are these the sole symmetries present in NN parameterizations}$? Zooming into most practical activation functions (e.g. sine, ReLU, tanh) answers this question negatively and gives rise to intriguing new symmetries, which we collectively refer to as $\\textit{scaling symmetries}$, that is, non-zero scalar multiplications and divisions of weights and biases. In this work, we propose $\\textit{Scale Equivariant Graph MetaNetworks - ScaleGMNs}$, a framework that adapts the Graph Metanetwork (message-passing) paradigm by incorporating scaling symmetries and thus rendering neuron and edge representations equivariant to valid scalings. We introduce novel building blocks, of independent technical interest, that allow for equivariance or invariance with respect to individual scalar multipliers or their product and use them in all components of ScaleGMN. Furthermore, we prove that, under certain expressivity conditions, ScaleGMN can simulate the forward and backward pass of any input feedforward neural network. Experimental results demonstrate that our method advances the state-of-the-art performance for several datasets and activation functions, highlighting the power of scaling symmetries as an inductive bias for NN processing."
                    },
                    {
                        "title": "Large-scale unsupervised audio pre-training for video-to-speech synthesis",
                        "abstract": "Video-to-speech synthesis is the task of reconstructing the speech signal from a silent video of a speaker. Most established approaches to date involve a two-step process, whereby an intermediate representation from the video, such as a spectrogram, is extracted first and then passed to a vocoder to produce the raw audio. Some recent work has focused on end-to-end synthesis, whereby the generation of raw audio and any intermediate representations is performed jointly. All such approaches involve training on data from almost exclusively audio-visual datasets, i.e. every audio sample has a corresponding video sample. This precludes the use of abundant audio-only datasets which may not have a corresponding visual modality (e.g. audiobooks, radio podcasts, speech recognition datasets etc.), as well as audio-only architectures that have been developed by the audio machine learning community over the years. In this paper we propose to train encoder-decoder models on more than 3,500 hours of audio data at 24kHz, and then use the pre-trained decoders to initialize the audio decoders for the video-to-speech synthesis task. The pre-training step uses audio samples only and does not require labels or corresponding samples from other modalities (visual, text). We demonstrate that this pre-training step improves the reconstructed speech and that it is an unexplored way to improve the quality of the generator in a cross-modal task while only requiring samples from one of the modalities. We conduct experiments using both raw audio and mel spectrograms as target outputs and benchmark our models with existing work."
                    }
                ]
            },
            "1661bf11-a5ce-4ab9-807f-8ada383cdfb3": {
                "pk": "1661bf11-a5ce-4ab9-807f-8ada383cdfb3",
                "name": "Mihalis A. Nicolaou",
                "collaborators": [
                    "Stefanos Zafeiriou",
                    "Maja Pantic",
                    "Charles Ringer"
                ],
                "domain": [
                    "Probabilistic Modeling",
                    "Component Analysis",
                    "Deep Learning",
                    "Video Analysis"
                ],
                "publications": [
                    {
                        "title": "A Unified Framework for Probabilistic Component Analysis",
                        "abstract": "We present a unifying framework which reduces the construction of probabilistic component analysis techniques to a mere selection of the latent neighbourhood, thus providing an elegant and principled framework for creating novel component analysis models as well as constructing probabilistic equivalents of deterministic component analysis methods. Under our framework, we unify many very popular and well-studied component analysis algorithms, such as Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Locality Preserving Projections (LPP) and Slow Feature Analysis (SFA), some of which have no probabilistic equivalents in literature thus far. We firstly define the Markov Random Fields (MRFs) which encapsulate the latent connectivity of the aforementioned component analysis techniques; subsequently, we show that the projection directions produced by all PCA, LDA, LPP and SFA are also produced by the Maximum Likelihood (ML) solution of a single joint probability density function, composed by selecting one of the defined MRF priors while utilising a simple observation model. Furthermore, we propose novel Expectation Maximization (EM) algorithms, exploiting the proposed joint PDF, while we generalize the proposed methodologies to arbitrary connectivities via parameterizable MRF products. Theoretical analysis and experiments on both simulated and real world data show the usefulness of the proposed framework, by deriving methods which well outperform state-of-the-art equivalents."
                    },
                    {
                        "title": "Deep Unsupervised Multi-View Detection of Video Game Stream Highlights",
                        "abstract": "We consider the problem of automatic highlight-detection in video game streams. Currently, the vast majority of highlight-detection systems for games are triggered by the occurrence of hard-coded game events (e.g., score change, end-game), while most advanced tools and techniques are based on detection of highlights via visual analysis of game footage. We argue that in the context of game streaming, events that may constitute highlights are not only dependent on game footage, but also on social signals that are conveyed by the streamer during the play session (e.g., when interacting with viewers, or when commenting and reacting to the game). In this light, we present a multi-view unsupervised deep learning methodology for novelty-based highlight detection. The method jointly analyses both game footage and social signals such as the players facial expressions and speech, and shows promising results for generating highlights on streams of popular games such as Player Unknown's Battlegrounds."
                    }
                ]
            },
            "2266255e-640a-497a-9890-8c80b47db3f3": {
                "pk": "2266255e-640a-497a-9890-8c80b47db3f3",
                "name": "Jiankang Deng",
                "collaborators": [
                    "Stefanos Zafeiriou",
                    "Ismail Elezi",
                    "Zhipeng Du",
                    "Miaojing Shi",
                    "Yuxiang Zhou",
                    "Niannan Xue",
                    "Jia Guo",
                    "Yunqi Miao",
                    "Jungong Han",
                    "Michail Tarasiou"
                ],
                "domain": [
                    "Computer Vision",
                    "3D Reconstruction",
                    "Deep Learning",
                    "Facial Recognition"
                ],
                "publications": [
                    {
                        "title": "Dense 3D Face Decoding over 2500FPS: Joint Texture & Shape Convolutional Mesh Decoders",
                        "abstract": "3D Morphable Models (3DMMs) are statistical models that represent facial texture and shape variations using a set of linear bases and more particular Principal Component Analysis (PCA). 3DMMs were used as statistical priors for reconstructing 3D faces from images by solving non-linear least square optimization problems. Recently, 3DMMs were used as generative models for training non-linear mappings (\\ie, regressors) from image to the parameters of the models via Deep Convolutional Neural Networks (DCNNs). Nevertheless, all of the above methods use either fully connected layers or 2D convolutions on parametric unwrapped UV spaces leading to large networks with many parameters. In this paper, we present the first, to the best of our knowledge, non-linear 3DMMs by learning joint texture and shape auto-encoders using direct mesh convolutions. We demonstrate how these auto-encoders can be used to train very light-weight models that perform Coloured Mesh Decoding (CMD) in-the-wild at a speed of over 2500 FPS."
                    },
                    {
                        "title": "Informed Non-convex Robust Principal Component Analysis with Features",
                        "abstract": "We revisit the problem of robust principal component analysis with features acting as prior side information. To this aim, a novel, elegant, non-convex optimization approach is proposed to decompose a given observation matrix into a low-rank core and the corresponding sparse residual. Rigorous theoretical analysis of the proposed algorithm results in exact recovery guarantees with low computational complexity. Aptly designed synthetic experiments demonstrate that our method is the first to wholly harness the power of non-convexity over convexity in terms of both recoverability and speed. That is, the proposed non-convex approach is more accurate and faster compared to the best available algorithms for the problem under study. Two real-world applications, namely image classification and face denoising further exemplify the practical superiority of the proposed method."
                    },
                    {
                        "title": "Stacked Dense U-Nets with Dual Transformers for Robust Face Alignment",
                        "abstract": "Facial landmark localisation in images captured in-the-wild is an important and challenging problem. The current state-of-the-art revolves around certain kinds of Deep Convolutional Neural Networks (DCNNs) such as stacked U-Nets and Hourglass networks. In this work, we innovatively propose stacked dense U-Nets for this task. We design a novel scale aggregation network topology structure and a channel aggregation building block to improve the model's capacity without sacrificing the computational complexity and model size. With the assistance of deformable convolutions inside the stacked dense U-Nets and coherent loss for outside data transformation, our model obtains the ability to be spatially invariant to arbitrary input face images. Extensive experiments on many in-the-wild datasets, validate the robustness of the proposed method under extreme poses, exaggerated expressions and heavy occlusions. Finally, we show that accurate 3D face alignment can assist pose-invariant face recognition where we achieve a new state-of-the-art accuracy on CFP-FP."
                    },
                    {
                        "title": "Perspective Reconstruction of Human Faces by Joint Mesh and Landmark Regression",
                        "abstract": "Even though 3D face reconstruction has achieved impressive progress, most orthogonal projection-based face reconstruction methods can not achieve accurate and consistent reconstruction results when the face is very close to the camera due to the distortion under the perspective projection. In this paper, we propose to simultaneously reconstruct 3D face mesh in the world space and predict 2D face landmarks on the image plane to address the problem of perspective 3D face reconstruction. Based on the predicted 3D vertices and 2D landmarks, the 6DoF (6 Degrees of Freedom) face pose can be easily estimated by the PnP solver to represent perspective projection. Our approach achieves 1st place on the leader-board of the ECCV 2022 WCPA challenge and our model is visually robust under different identities, expressions and poses. The training code and models are released to facilitate future research."
                    },
                    {
                        "title": "Joint Multi-view Face Alignment in the Wild",
                        "abstract": "The de facto algorithm for facial landmark estimation involves running a face detector with a subsequent deformable model fitting on the bounding box. This encompasses two basic problems: i) the detection and deformable fitting steps are performed independently, while the detector might not provide best-suited initialisation for the fitting step, ii) the face appearance varies hugely across different poses, which makes the deformable face fitting very challenging and thus distinct models have to be used (\\eg, one for profile and one for frontal faces). In this work, we propose the first, to the best of our knowledge, joint multi-view convolutional network to handle large pose variations across faces in-the-wild, and elegantly bridge face detection and facial landmark localisation tasks. Existing joint face detection and landmark localisation methods focus only on a very small set of landmarks. By contrast, our method can detect and align a large number of landmarks for semi-frontal (68 landmarks) and profile (39 landmarks) faces. We evaluate our model on a plethora of datasets including standard static image datasets such as IBUG, 300W, COFW, and the latest Menpo Benchmark for both semi-frontal and profile faces. Significant improvement over state-of-the-art methods on deformable face tracking is witnessed on 300VW benchmark. We also demonstrate state-of-the-art results for face detection on FDDB and MALF datasets."
                    },
                    {
                        "title": "Confidence-guided Centroids for Unsupervised Person Re-Identification",
                        "abstract": "Unsupervised person re-identification (ReID) aims to train a feature extractor for identity retrieval without exploiting identity labels. Due to the blind trust in imperfect clustering results, the learning is inevitably misled by unreliable pseudo labels. Albeit the pseudo label refinement has been investigated by previous works, they generally leverage auxiliary information such as camera IDs and body part predictions. This work explores the internal characteristics of clusters to refine pseudo labels. To this end, Confidence-Guided Centroids (CGC) are proposed to provide reliable cluster-wise prototypes for feature learning. Since samples with high confidence are exclusively involved in the formation of centroids, the identity information of low-confidence samples, i.e., boundary samples, are NOT likely to contribute to the corresponding centroid. Given the new centroids, current learning scheme, where samples are enforced to learn from their assigned centroids solely, is unwise. To remedy the situation, we propose to use Confidence-Guided pseudo Label (CGL), which enables samples to approach not only the originally assigned centroid but other centroids that are potentially embedded with their identity information. Empowered by confidence-guided centroids and labels, our method yields comparable performance with, or even outperforms, state-of-the-art pseudo label refinement works that largely leverage auxiliary information."
                    },
                    {
                        "title": "Domain-General Crowd Counting in Unseen Scenarios",
                        "abstract": "Domain shift across crowd data severely hinders crowd counting models to generalize to unseen scenarios. Although domain adaptive crowd counting approaches close this gap to a certain extent, they are still dependent on the target domain data to adapt (e.g. finetune) their models to the specific domain. In this paper, we aim to train a model based on a single source domain which can generalize well on any unseen domain. This falls into the realm of domain generalization that remains unexplored in crowd counting. We first introduce a dynamic sub-domain division scheme which divides the source domain into multiple sub-domains such that we can initiate a meta-learning framework for domain generalization. The sub-domain division is dynamically refined during the meta-learning. Next, in order to disentangle domain-invariant information from domain-specific information in image features, we design the domain-invariant and -specific crowd memory modules to re-encode image features. Two types of losses, i.e. feature reconstruction and orthogonal losses, are devised to enable this disentanglement. Extensive experiments on several standard crowd counting benchmarks i.e. SHA, SHB, QNRF, and NWPU, show the strong generalizability of our method."
                    },
                    {
                        "title": "Rethinking the Domain Gap in Near-infrared Face Recognition",
                        "abstract": "Heterogeneous face recognition (HFR) involves the intricate task of matching face images across the visual domains of visible (VIS) and near-infrared (NIR). While much of the existing literature on HFR identifies the domain gap as a primary challenge and directs efforts towards bridging it at either the input or feature level, our work deviates from this trend. We observe that large neural networks, unlike their smaller counterparts, when pre-trained on large scale homogeneous VIS data, demonstrate exceptional zero-shot performance in HFR, suggesting that the domain gap might be less pronounced than previously believed. By approaching the HFR problem as one of low-data fine-tuning, we introduce a straightforward framework: comprehensive pre-training, succeeded by a regularized fine-tuning strategy, that matches or surpasses the current state-of-the-art on four publicly available benchmarks. Corresponding codes can be found at https://github.com/michaeltrs/RethinkNIRVIS."
                    },
                    {
                        "title": "Boosting Object Detection with Zero-Shot Day-Night Domain Adaptation",
                        "abstract": "Detecting objects in low-light scenarios presents a persistent challenge, as detectors trained on well-lit data exhibit significant performance degradation on low-light data due to low visibility. Previous methods mitigate this issue by exploring image enhancement or object detection techniques with real low-light image datasets. However, the progress is impeded by the inherent difficulties about collecting and annotating low-light images. To address this challenge, we propose to boost low-light object detection with zero-shot day-night domain adaptation, which aims to generalize a detector from well-lit scenarios to low-light ones without requiring real low-light data. Revisiting Retinex theory in the low-level vision, we first design a reflectance representation learning module to learn Retinex-based illumination invariance in images with a carefully designed illumination invariance reinforcement strategy. Next, an interchange-redecomposition-coherence procedure is introduced to improve over the vanilla Retinex image decomposition process by performing two sequential image decompositions and introducing a redecomposition cohering loss. Extensive experiments on ExDark, DARK FACE, and CODaN datasets show strong low-light generalizability of our method. Our code is available at https://github.com/ZPDu/DAI-Net."
                    },
                    {
                        "title": "G3DR: Generative 3D Reconstruction in ImageNet",
                        "abstract": "We introduce a novel 3D generative method, Generative 3D Reconstruction (G3DR) in ImageNet, capable of generating diverse and high-quality 3D objects from single images, addressing the limitations of existing methods. At the heart of our framework is a novel depth regularization technique that enables the generation of scenes with high-geometric fidelity. G3DR also leverages a pretrained language-vision model, such as CLIP, to enable reconstruction in novel views and improve the visual realism of generations. Additionally, G3DR designs a simple but effective sampling procedure to further improve the quality of generations. G3DR offers diverse and efficient 3D asset generation based on class or text conditioning. Despite its simplicity, G3DR is able to beat state-of-theart methods, improving over them by up to 22% in perceptual metrics and 90% in geometry scores, while needing only half of the training time. Code is available at https://github.com/preddy5/G3DR"
                    },
                    {
                        "title": "$V_kD:$ Improving Knowledge Distillation using Orthogonal Projections",
                        "abstract": "Knowledge distillation is an effective method for training small and efficient deep learning models. However, the efficacy of a single method can degenerate when transferring to other tasks, modalities, or even other architectures. To address this limitation, we propose a novel constrained feature distillation method. This method is derived from a small set of core principles, which results in two emerging components: an orthogonal projection and a task-specific normalisation. Equipped with both of these components, our transformer models can outperform all previous methods on ImageNet and reach up to a 4.4% relative improvement over the previous state-of-the-art methods. To further demonstrate the generality of our method, we apply it to object detection and image generation, whereby we obtain consistent and substantial performance improvements over state-of-the-art. Code and models are publicly available: https://github.com/roymiles/vkd"
                    },
                    {
                        "title": "WaveFace: Authentic Face Restoration with Efficient Frequency Recovery",
                        "abstract": "Although diffusion models are rising as a powerful solution for blind face restoration, they are criticized for two problems: 1) slow training and inference speed, and 2) failure in preserving identity and recovering fine-grained facial details. In this work, we propose WaveFace to solve the problems in the frequency domain, where low- and high-frequency components decomposed by wavelet transformation are considered individually to maximize authenticity as well as efficiency. The diffusion model is applied to recover the low-frequency component only, which presents general information of the original image but 1/16 in size. To preserve the original identity, the generation is conditioned on the low-frequency component of low-quality images at each denoising step. Meanwhile, high-frequency components at multiple decomposition levels are handled by a unified network, which recovers complex facial details in a single step. Evaluations on four benchmark datasets show that: 1) WaveFace outperforms state-of-the-art methods in authenticity, especially in terms of identity preservation, and 2) authentic images are restored with the efficiency 10x faster than existing diffusion model-based BFR methods."
                    },
                    {
                        "title": "Deep Active Learning: A Reality Check",
                        "abstract": "We conduct a comprehensive evaluation of state-of-the-art deep active learning methods. Surprisingly, under general settings, no single-model method decisively outperforms entropy-based active learning, and some even fall short of random sampling. We delve into overlooked aspects like starting budget, budget step, and pretraining's impact, revealing their significance in achieving superior results. Additionally, we extend our evaluation to other tasks, exploring the active learning effectiveness in combination with semi-supervised learning, and object detection. Our experiments provide valuable insights and concrete recommendations for future active learning studies. By uncovering the limitations of current methods and understanding the impact of different experimental settings, we aim to inspire more efficient training of deep learning models in real-world scenarios with limited annotation budgets. This work contributes to advancing active learning's efficacy in deep learning and empowers researchers to make informed decisions when applying active learning to their tasks."
                    },
                    {
                        "title": "Improving face generation quality and prompt following with synthetic captions",
                        "abstract": "Recent advancements in text-to-image generation using diffusion models have significantly improved the quality of generated images and expanded the ability to depict a wide range of objects. However, ensuring that these models adhere closely to the text prompts remains a considerable challenge. This issue is particularly pronounced when trying to generate photorealistic images of humans. Without significant prompt engineering efforts models often produce unrealistic images and typically fail to incorporate the full extent of the prompt information. This limitation can be largely attributed to the nature of captions accompanying the images used in training large scale diffusion models, which typically prioritize contextual information over details related to the person's appearance. In this paper we address this issue by introducing a training-free pipeline designed to generate accurate appearance descriptions from images of people. We apply this method to create approximately 250,000 captions for publicly available face datasets. We then use these synthetic captions to fine-tune a text-to-image diffusion model. Our results demonstrate that this approach significantly improves the model's ability to generate high-quality, realistic human faces and enhances adherence to the given prompts, compared to the baseline model. We share our synthetic captions, pretrained checkpoints and training code."
                    },
                    {
                        "title": "VeLoRA: Memory Efficient Training using Rank-1 Sub-Token Projections",
                        "abstract": "Large language models (LLMs) have recently emerged as powerful tools for tackling many language-processing tasks. Despite their success, training and fine-tuning these models is still far too computationally and memory intensive. In this paper, we identify and characterise the important components needed for effective model convergence using gradient descent. In doing so we find that the intermediate activations used to implement backpropagation can be excessively compressed without incurring any degradation in performance. This result leads us to a cheap and memory-efficient algorithm for both fine-tuning and pre-training LLMs. The proposed algorithm simply divides the tokens up into smaller sub-tokens before projecting them onto a fixed 1-dimensional subspace during the forward pass. These features are then coarsely reconstructed during the backward pass to implement the update rules. We confirm the effectiveness of our algorithm as being complimentary to many state-of-the-art PEFT methods on the VTAB-1k fine-tuning benchmark. Furthermore, we outperform QLoRA for fine-tuning LLaMA and show competitive performance against other memory-efficient pre-training methods on the large-scale C4 dataset."
                    },
                    {
                        "title": "From Attention to Activation: Unravelling the Enigmas of Large Language Models",
                        "abstract": "We study two strange phenomena in auto-regressive Transformers: (1) the dominance of the first token in attention heads; (2) the occurrence of large outlier activations in the hidden states. We find that popular large language models, such as Llama attend maximally to the first token in 98% of attention heads, a behaviour we attribute to the softmax function. To mitigate this issue, we propose a reformulation of softmax to softmax-1. Furthermore, we identify adaptive optimisers, e.g. Adam, as the primary contributor to the large outlier activations and introduce OrthoAdam, a novel optimiser that utilises orthogonal matrices to transform gradients, to address this issue. Finally, not only do our methods prevent these phenomena from occurring, but additionally, they enable Transformers to sustain their performance when quantised using basic algorithms, something that standard methods are unable to do. In summary, our methods reduce the attention proportion on the first token from 65% to 3.3%, the activation kurtosis in the hidden states from 1657 to 3.1, and perplexity penalty under 4-bit weight quantisation from 3565 to 0.3."
                    },
                    {
                        "title": "OSTeC: One-Shot Texture Completion",
                        "abstract": "The last few years have witnessed the great success of non-linear generative models in synthesizing high-quality photorealistic face images. Many recent 3D facial texture reconstruction and pose manipulation from a single image approaches still rely on large and clean face datasets to train image-to-image Generative Adversarial Networks (GANs). Yet the collection of such a large scale high-resolution 3D texture dataset is still very costly and difficult to maintain age/ethnicity balance. Moreover, regression-based approaches suffer from generalization to the in-the-wild conditions and are unable to fine-tune to a target-image. In this work, we propose an unsupervised approach for one-shot 3D facial texture completion that does not require large-scale texture datasets, but rather harnesses the knowledge stored in 2D face generators. The proposed approach rotates an input image in 3D and fill-in the unseen regions by reconstructing the rotated image in a 2D face generator, based on the visible parts. Finally, we stitch the most visible textures at different angles in the UV image-plane. Further, we frontalize the target image by projecting the completed texture into the generator. The qualitative and quantitative experiments demonstrate that the completed UV textures and frontalized images are of high quality, resembles the original identity, can be used to train a texture GAN model for 3DMM fitting and improve pose-invariant face recognition."
                    },
                    {
                        "title": "Redesigning Multi-Scale Neural Network for Crowd Counting",
                        "abstract": "Perspective distortions and crowd variations make crowd counting a challenging task in computer vision. To tackle it, many previous works have used multi-scale architecture in deep neural networks (DNNs). Multi-scale branches can be either directly merged (e.g. by concatenation) or merged through the guidance of proxies (e.g. attentions) in the DNNs. Despite their prevalence, these combination methods are not sophisticated enough to deal with the per-pixel performance discrepancy over multi-scale density maps. In this work, we redesign the multi-scale neural network by introducing a hierarchical mixture of density experts, which hierarchically merges multi-scale density maps for crowd counting. Within the hierarchical structure, an expert competition and collaboration scheme is presented to encourage contributions from all scales; pixel-wise soft gating nets are introduced to provide pixel-wise soft weights for scale combinations in different hierarchies. The network is optimized using both the crowd density map and the local counting map, where the latter is obtained by local integration on the former. Optimizing both can be problematic because of their potential conflicts. We introduce a new relative local counting loss based on relative count differences among hard-predicted local regions in an image, which proves to be complementary to the conventional absolute error loss on the density map. Experiments show that our method achieves the state-of-the-art performance on five public datasets, i.e. ShanghaiTech, UCF_CC_50, JHU-CROWD++, NWPU-Crowd and Trancos."
                    }
                ]
            },
            "2b679640-624a-472b-b094-4a569243d63f": {
                "pk": "2b679640-624a-472b-b094-4a569243d63f",
                "name": "Ioannis Patras",
                "collaborators": [
                    "Giorgos Kordopatis-Zilos",
                    "Symeon Papadopoulos",
                    "Ioannis Kompatsiaris",
                    "Ioannis Maniadis Metaxas",
                    "Georgios Tzimiropoulos",
                    "Christos Tzelepis",
                    "Petar Palasek",
                    "Chen Feng",
                    "Georgios Zoumpourlis",
                    "Aria Ahmadi"
                ],
                "domain": [
                    "Video Retrieval",
                    "Deep Learning",
                    "Self-Supervised Learning",
                    "Action Recognition"
                ],
                "publications": [
                    {
                        "title": "FIVR: Fine-grained Incident Video Retrieval",
                        "abstract": "This paper introduces the problem of Fine-grained Incident Video Retrieval (FIVR). Given a query video, the objective is to retrieve all associated videos, considering several types of associations that range from duplicate videos to videos from the same incident. FIVR offers a single framework that contains several retrieval tasks as special cases. To address the benchmarking needs of all such tasks, we construct and present a large-scale annotated video dataset, which we call FIVR-200K, and it comprises 225,960 videos. To create the dataset, we devise a process for the collection of YouTube videos based on major news events from recent years crawled from Wikipedia and deploy a retrieval pipeline for the automatic selection of query videos based on their estimated suitability as benchmarks. We also devise a protocol for the annotation of the dataset with respect to the four types of video associations defined by FIVR. Finally, we report the results of an experimental study on the dataset comparing five state-of-the-art methods developed based on a variety of visual descriptors, highlighting the challenges of the current problem."
                    },
                    {
                        "title": "DivClust: Controlling Diversity in Deep Clustering",
                        "abstract": "Clustering has been a major research topic in the field of machine learning, one to which Deep Learning has recently been applied with significant success. However, an aspect of clustering that is not addressed by existing deep clustering methods, is that of efficiently producing multiple, diverse partitionings for a given dataset. This is particularly important, as a diverse set of base clusterings are necessary for consensus clustering, which has been found to produce better and more robust results than relying on a single clustering. To address this gap, we propose DivClust, a diversity controlling loss that can be incorporated into existing deep clustering frameworks to produce multiple clusterings with the desired degree of diversity. We conduct experiments with multiple datasets and deep clustering frameworks and show that: a) our method effectively controls diversity across frameworks and datasets with very small additional computational cost, b) the sets of clusterings learned by DivClust include solutions that significantly outperform single-clustering baselines, and c) using an off-the-shelf consensus clustering algorithm, DivClust produces consensus clustering solutions that consistently outperform single-clustering baselines, effectively improving the performance of the base deep clustering framework."
                    },
                    {
                        "title": "ViSiL: Fine-grained Spatio-Temporal Video Similarity Learning",
                        "abstract": "In this paper we introduce ViSiL, a Video Similarity Learning architecture that considers fine-grained Spatio-Temporal relations between pairs of videos -- such relations are typically lost in previous video retrieval approaches that embed the whole frame or even the whole video into a vector descriptor before the similarity estimation. By contrast, our Convolutional Neural Network (CNN)-based approach is trained to calculate video-to-video similarity from refined frame-to-frame similarity matrices, so as to consider both intra- and inter-frame relations. In the proposed method, pairwise frame similarity is estimated by applying Tensor Dot (TD) followed by Chamfer Similarity (CS) on regional CNN frame features - this avoids feature aggregation before the similarity calculation between frames. Subsequently, the similarity matrix between all video frames is fed to a four-layer CNN, and then summarized using Chamfer Similarity (CS) into a video-to-video similarity score -- this avoids feature aggregation before the similarity calculation between videos and captures the temporal similarity patterns between matching frame sequences. We train the proposed network using a triplet loss scheme and evaluate it on five public benchmark datasets on four different video retrieval problems where we demonstrate large improvements in comparison to the state of the art. The implementation of ViSiL is publicly available."
                    },
                    {
                        "title": "Efficient Unsupervised Visual Representation Learning with Explicit Cluster Balancing",
                        "abstract": "Self-supervised learning has recently emerged as the preeminent pretraining paradigm across and between modalities, with remarkable results. In the image domain specifically, group (or cluster) discrimination has been one of the most successful methods. However, such frameworks need to guard against heavily imbalanced cluster assignments to prevent collapse to trivial solutions. Existing works typically solve this by reweighing cluster assignments to promote balance, or with offline operations (e.g. regular re-clustering) that prevent collapse. However, the former typically requires large batch sizes, which leads to increased resource requirements, and the latter introduces scalability issues with regard to large datasets. In this work, we propose ExCB, a framework that tackles this problem with a novel cluster balancing method. ExCB estimates the relative size of the clusters across batches and balances them by adjusting cluster assignments, proportionately to their relative size and in an online manner. Thereby, it overcomes previous methods' dependence on large batch sizes and is fully online, and therefore scalable to any dataset. We conduct extensive experiments to evaluate our approach and demonstrate that ExCB: a) achieves state-of-the-art results with significantly reduced resource requirements compared to previous works, b) is fully online, and therefore scalable to large datasets, and c) is stable and effective even with very small batch sizes."
                    },
                    {
                        "title": "Semi-supervised Fisher vector network",
                        "abstract": "In this work we explore how the architecture proposed in [8], which expresses the processing steps of the classical Fisher vector pipeline approaches, i.e. dimensionality reduction by principal component analysis (PCA) projection, Gaussian mixture model (GMM) and Fisher vector descriptor extraction as network layers, can be modified into a hybrid network that combines the benefits of both unsupervised and supervised training methods, resulting in a model that learns a semi-supervised Fisher vector descriptor of the input data. We evaluate the proposed model at image classification and action recognition problems and show how the model's classification performance improves as the amount of unlabeled data increases during training."
                    },
                    {
                        "title": "Uncertainty Propagation in Convolutional Neural Networks: Technical Report",
                        "abstract": "In this technical report we study the problem of propagation of uncertainty (in terms of variances of given uni-variate normal random variables) through typical building blocks of a Convolutional Neural Network (CNN). These include layers that perform linear operations, such as 2D convolutions, fully-connected, and average pooling layers, as well as layers that act non-linearly on their input, such as the Rectified Linear Unit (ReLU). Finally, we discuss the sigmoid function, for which we give approximations of its first- and second-order moments, as well as the binary cross-entropy loss function, for which we approximate its expected value under normal random inputs."
                    },
                    {
                        "title": "Unsupervised convolutional neural networks for motion estimation",
                        "abstract": "Traditional methods for motion estimation estimate the motion field F between a pair of images as the one that minimizes a predesigned cost function. In this paper, we propose a direct method and train a Convolutional Neural Network (CNN) that when, at test time, is given a pair of images as input it produces a dense motion field F at its output layer. In the absence of large datasets with ground truth motion that would allow classical supervised training, we propose to train the network in an unsupervised manner. The proposed cost function that is optimized during training, is based on the classical optical flow constraint. The latter is differentiable with respect to the motion field and, therefore, allows backpropagation of the error to previous layers of the network. Our method is tested on both synthetic and real image sequences and performs similarly to the state-of-the-art methods."
                    },
                    {
                        "title": "Discriminative convolutional Fisher vector network for action recognition",
                        "abstract": "In this work we propose a novel neural network architecture for the problem of human action recognition in videos. The proposed architecture expresses the processing steps of classical Fisher vector approaches, that is dimensionality reduction by principal component analysis (PCA) projection, Gaussian mixture model (GMM) and Fisher vector descriptor extraction, as network layers. By contrast to other methods where these steps are performed consecutively and the corresponding parameters are learned in an unsupervised manner, having them defined as a single neural network allows us to refine the whole model discriminatively in an end to end fashion. Furthermore, we show that the proposed architecture can be used as a replacement for the fully connected layers in popular convolutional networks achieving a comparable classification performance, or even significantly surpassing the performance of similar architectures while reducing the total number of trainable parameters by a factor of 5. We show that our method achieves significant improvements in comparison to the classical chain."
                    },
                    {
                        "title": "Adaptive Soft Contrastive Learning",
                        "abstract": "Self-supervised learning has recently achieved great success in representation learning without human annotations. The dominant method -- that is contrastive learning, is generally based on instance discrimination tasks, i.e., individual samples are treated as independent categories. However, presuming all the samples are different contradicts the natural grouping of similar samples in common visual datasets, e.g., multiple views of the same dog. To bridge the gap, this paper proposes an adaptive method that introduces soft inter-sample relations, namely Adaptive Soft Contrastive Learning (ASCL). More specifically, ASCL transforms the original instance discrimination task into a multi-instance soft discrimination task, and adaptively introduces inter-sample relations. As an effective and concise plug-in module for existing self-supervised learning frameworks, ASCL achieves the best performance on several benchmarks in terms of both performance and efficiency. Code is available at https://github.com/MrChenFeng/ASCL_ICPR2022."
                    },
                    {
                        "title": "Capsule Network based Contrastive Learning of Unsupervised Visual Representations",
                        "abstract": "Capsule Networks have shown tremendous advancement in the past decade, outperforming the traditional CNNs in various task due to it's equivariant properties. With the use of vector I/O which provides information of both magnitude and direction of an object or it's part, there lies an enormous possibility of using Capsule Networks in unsupervised learning environment for visual representation tasks such as multi class image classification. In this paper, we propose Contrastive Capsule (CoCa) Model which is a Siamese style Capsule Network using Contrastive loss with our novel architecture, training and testing algorithm. We evaluate the model on unsupervised image classification CIFAR-10 dataset and achieve a top-1 test accuracy of 70.50% and top-5 test accuracy of 98.10%. Due to our efficient architecture our model has 31 times less parameters and 71 times less FLOPs than the current SOTA in both supervised and unsupervised learning."
                    },
                    {
                        "title": "Mirror, mirror on the wall, tell me, is the error small?",
                        "abstract": "Do object part localization methods produce bilaterally symmetric results on mirror images? Surprisingly not, even though state of the art methods augment the training set with mirrored images. In this paper we take a closer look into this issue. We first introduce the concept of mirrorability as the ability of a model to produce symmetric results in mirrored images and introduce a corresponding measure, namely the \\textit{mirror error} that is defined as the difference between the detection result on an image and the mirror of the detection result on its mirror image. We evaluate the mirrorability of several state of the art algorithms in two of the most intensively studied problems, namely human pose estimation and face alignment. Our experiments lead to several interesting findings: 1) Surprisingly, most of state of the art methods struggle to preserve the mirror symmetry, despite the fact that they do have very similar overall performance on the original and mirror images; 2) the low mirrorability is not caused by training or testing sample bias - all algorithms are trained on both the original images and their mirrored versions; 3) the mirror error is strongly correlated to the localization/alignment error (with correlation coefficients around 0.7). Since the mirror error is calculated without knowledge of the ground truth, we show two interesting applications - in the first it is used to guide the selection of difficult samples and in the second to give feedback in a popular Cascaded Pose Regression method for face alignment."
                    },
                    {
                        "title": "Motor Imagery Decoding Using Ensemble Curriculum Learning and Collaborative Training",
                        "abstract": "In this work, we study the problem of cross-subject motor imagery (MI) decoding from electroencephalography (EEG) data. Multi-subject EEG datasets present several kinds of domain shifts due to various inter-individual differences (e.g. brain anatomy, personality and cognitive profile). These domain shifts render multi-subject training a challenging task and also impede robust cross-subject generalization. Inspired by the importance of domain generalization techniques for tackling such issues, we propose a two-stage model ensemble architecture built with multiple feature extractors (first stage) and a shared classifier (second stage), which we train end-to-end with two novel loss terms. The first loss applies curriculum learning, forcing each feature extractor to specialize to a subset of the training subjects and promoting feature diversity. The second loss is an intra-ensemble distillation objective that allows collaborative exchange of knowledge between the models of the ensemble. We compare our method against several state-of-the-art techniques, conducting subject-independent experiments on two large MI datasets, namely PhysioNet and OpenBMI. Our algorithm outperforms all of the methods in both 5-fold cross-validation and leave-one-subject-out evaluation settings, using a substantially lower number of trainable parameters. We demonstrate that our model ensembling approach combining the powers of curriculum learning and collaborative training, leads to high learning capacity and robust performance. Our work addresses the issue of domain shifts in multi-subject EEG datasets, paving the way for calibration-free brain-computer interfaces. We make our code publicly available at: https://github.com/gzoumpourlis/Ensemble-MI"
                    },
                    {
                        "title": "MaskCon: Masked Contrastive Learning for Coarse-Labelled Dataset",
                        "abstract": "Deep learning has achieved great success in recent years with the aid of advanced neural network structures and large-scale human-annotated datasets. However, it is often costly and difficult to accurately and efficiently annotate large-scale datasets, especially for some specialized domains where fine-grained labels are required. In this setting, coarse labels are much easier to acquire as they do not require expert knowledge. In this work, we propose a contrastive learning method, called $\\textbf{Mask}$ed $\\textbf{Con}$trastive learning~($\\textbf{MaskCon}$) to address the under-explored problem setting, where we learn with a coarse-labelled dataset in order to address a finer labelling problem. More specifically, within the contrastive learning framework, for each sample our method generates soft-labels with the aid of coarse labels against other samples and another augmented view of the sample in question. By contrast to self-supervised contrastive learning where only the sample's augmentations are considered hard positives, and in supervised contrastive learning where only samples with the same coarse labels are considered hard positives, we propose soft labels based on sample distances, that are masked by the coarse labels. This allows us to utilize both inter-sample relations and coarse labels. We demonstrate that our method can obtain as special cases many existing state-of-the-art works and that it provides tighter bounds on the generalization error. Experimentally, our method achieves significant improvement over the current state-of-the-art in various datasets, including CIFAR10, CIFAR100, ImageNet-1K, Standford Online Products and Stanford Cars196 datasets. Code and annotations are available at https://github.com/MrChenFeng/MaskCon_CVPR2023."
                    },
                    {
                        "title": "Prompting Visual-Language Models for Dynamic Facial Expression Recognition",
                        "abstract": "This paper presents a novel visual-language model called DFER-CLIP, which is based on the CLIP model and designed for in-the-wild Dynamic Facial Expression Recognition (DFER). Specifically, the proposed DFER-CLIP consists of a visual part and a textual part. For the visual part, based on the CLIP image encoder, a temporal model consisting of several Transformer encoders is introduced for extracting temporal facial expression features, and the final feature embedding is obtained as a learnable \"class\" token. For the textual part, we use as inputs textual descriptions of the facial behaviour that is related to the classes (facial expressions) that we are interested in recognising -- those descriptions are generated using large language models, like ChatGPT. This, in contrast to works that use only the class names and more accurately captures the relationship between them. Alongside the textual description, we introduce a learnable token which helps the model learn relevant context information for each expression during training. Extensive experiments demonstrate the effectiveness of the proposed method and show that our DFER-CLIP also achieves state-of-the-art results compared with the current supervised DFER methods on the DFEW, FERV39k, and MAFW benchmarks. Code is publicly available at https://github.com/zengqunzhao/DFER-CLIP."
                    },
                    {
                        "title": "Self-Supervised Facial Representation Learning with Facial Region Awareness",
                        "abstract": "Self-supervised pre-training has been proved to be effective in learning transferable representations that benefit various visual tasks. This paper asks this question: can self-supervised pre-training learn general facial representations for various facial analysis tasks? Recent efforts toward this goal are limited to treating each face image as a whole, i.e., learning consistent facial representations at the image-level, which overlooks the consistency of local facial representations (i.e., facial regions like eyes, nose, etc). In this work, we make a first attempt to propose a novel self-supervised facial representation learning framework to learn consistent global and local facial representations, Facial Region Awareness (FRA). Specifically, we explicitly enforce the consistency of facial regions by matching the local facial representations across views, which are extracted with learned heatmaps highlighting the facial regions. Inspired by the mask prediction in supervised semantic segmentation, we obtain the heatmaps via cosine similarity between the per-pixel projection of feature maps and facial mask embeddings computed from learnable positional embeddings, which leverage the attention mechanism to globally look up the facial image for facial regions. To learn such heatmaps, we formulate the learning of facial mask embeddings as a deep clustering problem by assigning the pixel features from the feature maps to them. The transfer learning results on facial classification and regression tasks show that our FRA outperforms previous pre-trained models and more importantly, using ResNet as the unified backbone for various tasks, our FRA achieves comparable or even better performance compared with SOTA methods in facial analysis tasks."
                    },
                    {
                        "title": "DnS: Distill-and-Select for Efficient and Accurate Video Indexing and Retrieval",
                        "abstract": "In this paper, we address the problem of high performance and computationally efficient content-based video retrieval in large-scale datasets. Current methods typically propose either: (i) fine-grained approaches employing spatio-temporal representations and similarity calculations, achieving high performance at a high computational cost or (ii) coarse-grained approaches representing/indexing videos as global vectors, where the spatio-temporal structure is lost, providing low performance but also having low computational cost. In this work, we propose a Knowledge Distillation framework, called Distill-and-Select (DnS), that starting from a well-performing fine-grained Teacher Network learns: a) Student Networks at different retrieval performance and computational efficiency trade-offs and b) a Selector Network that at test time rapidly directs samples to the appropriate student to maintain both high retrieval performance and high computational efficiency. We train several students with different architectures and arrive at different trade-offs of performance and efficiency, i.e., speed and storage requirements, including fine-grained students that store/index videos using binary representations. Importantly, the proposed scheme allows Knowledge Distillation in large, unlabelled datasets -- this leads to good students. We evaluate DnS on five public datasets on three different video retrieval tasks and demonstrate a) that our students achieve state-of-the-art performance in several cases and b) that the DnS framework provides an excellent trade-off between retrieval performance, computational speed, and storage space. In specific configurations, the proposed method achieves similar mAP with the teacher but is 20 times faster and requires 240 times less storage space. The collected dataset and implementation are publicly available: https://github.com/mever-team/distill-and-select."
                    },
                    {
                        "title": "Aligned Unsupervised Pretraining of Object Detectors with Self-training",
                        "abstract": "The unsupervised pretraining of object detectors has recently become a key component of object detector training, as it leads to improved performance and faster convergence during the supervised fine-tuning stage. Existing unsupervised pretraining methods, however, typically rely on low-level information to define proposals that are used to train the detector. Furthermore, in the absence of class labels for these proposals, an auxiliary loss is used to add high-level semantics. This results in complex pipelines and a task gap between the pretraining and the downstream task. We propose a framework that mitigates this issue and consists of three simple yet key ingredients: (i) richer initial proposals that do encode high-level semantics, (ii) class pseudo-labeling through clustering, that enables pretraining using a standard object detection training pipeline, (iii) self-training to iteratively improve and enrich the object proposals. Once the pretraining and downstream tasks are aligned, a simple detection pipeline without further bells and whistles can be directly used for pretraining and, in fact, results in state-of-the-art performance on both the full and low data regimes, across detector architectures and datasets, by significant margins. We further show that our pretraining strategy is also capable of pretraining from scratch (including the backbone) and works on complex images like COCO, paving the path for unsupervised representation learning using object detection directly as a pretext task."
                    },
                    {
                        "title": "Learning to detect video events from zero or very few video examples",
                        "abstract": "In this work we deal with the problem of high-level event detection in video. Specifically, we study the challenging problems of i) learning to detect video events from solely a textual description of the event, without using any positive video examples, and ii) additionally exploiting very few positive training samples together with a small number of ``related'' videos. For learning only from an event's textual description, we first identify a general learning framework and then study the impact of different design choices for various stages of this framework. For additionally learning from example videos, when true positive training samples are scarce, we employ an extension of the Support Vector Machine that allows us to exploit ``related'' event videos by automatically introducing different weights for subsets of the videos in the overall training set. Experimental evaluations performed on the large-scale TRECVID MED 2014 video dataset provide insight on the effectiveness of the proposed methods."
                    },
                    {
                        "title": "TARN: Temporal Attentive Relation Network for Few-Shot and Zero-Shot Action Recognition",
                        "abstract": "In this paper we propose a novel Temporal Attentive Relation Network (TARN) for the problems of few-shot and zero-shot action recognition. At the heart of our network is a meta-learning approach that learns to compare representations of variable temporal length, that is, either two videos of different length (in the case of few-shot action recognition) or a video and a semantic representation such as word vector (in the case of zero-shot action recognition). By contrast to other works in few-shot and zero-shot action recognition, we a) utilise attention mechanisms so as to perform temporal alignment, and b) learn a deep-distance measure on the aligned representations at video segment level. We adopt an episode-based training scheme and train our network in an end-to-end manner. The proposed method does not require any fine-tuning in the target domain or maintaining additional representations as is the case of memory networks. Experimental results show that the proposed architecture outperforms the state of the art in few-shot action recognition, and achieves competitive results in zero-shot action recognition."
                    }
                ]
            }
        }
    },
    "2311.16424": {
        "paper_data": {
            "title": "Manifold Preserving Guided Diffusion",
            "url": "http://arxiv.org/abs/2311.16424v1",
            "arxiv_id": "2311.16424",
            "authors": [
                "Yutong He",
                "Naoki Murata",
                "Chieh-Hsin Lai",
                "Yuhta Takida",
                "Toshimitsu Uesaka",
                "Dongjun Kim",
                "Wei-Hsiang Liao",
                "Yuki Mitsufuji",
                "J. Zico Kolter",
                "Ruslan Salakhutdinov",
                "Stefano Ermon"
            ],
            "abstract": "Despite the recent advancements, conditional image generation still faces challenges of cost, generalizability, and the need for task-specific training. In this paper, we propose Manifold Preserving Guided Diffusion (MPGD), a training-free conditional generation framework that leverages pretrained diffusion models and off-the-shelf neural networks with minimal additional inference cost for a broad range of tasks. Specifically, we leverage the manifold hypothesis to refine the guided diffusion steps and introduce a shortcut algorithm in the process. We then propose two methods for on-manifold training-free guidance using pre-trained autoencoders and demonstrate that our shortcut inherently preserves the manifolds when applied to latent diffusion models. Our experiments show that MPGD is efficient and effective for solving a variety of conditional generation applications in low-compute settings, and can consistently offer up to 3.8x speed-ups with the same number of diffusion steps while maintaining high sample quality compared to the baselines.",
            "introduction": " introduction of the diffusion model, there were some conclusion, the image spaces of\u2202E \u2202x0\u22a4and\u2202D \u2202z0are identical. Theorem 2 If an autoencoder with encoder Eand decoder Dis a perfect autoencoder for the support X \u2282 M of the data distribution, then \u2207x0L(D(E(x0));y) =\u2202L \u2202D\u2202D \u2202E\u2202E \u2202x0|t\u2208 Tx0M. Proof. As stated in Shao et al. (2018), given the assumption of a perfect autoencoder, for any z0\u2208 Rk, the Jacobian\u2202D \u2202z0maps the tangent space Tz0Zatz0to the tangent space Tx0Mof the data manifold at x0=D(z0). In other words, the range of the Jacobian\u2202D \u2202z0lies within the tangent space at x0. Since Z=Rk, its tangent spaces are isomorphic to Rk. Therefore, for any vector vz\u2208Rk, the vector\u2202D \u2202z0vzlies in the tangent space at x0. This means that when taking the gradient of the loss function with respect to z0and applying the Jacobian\u2202D \u2202z0to the gradient, the resulting vector\u2202D \u2202z0\u2202D \u2202z0\u22a4\u2202L \u2202x0\u22a4also lies in the tangent space at x0. Finally, by Lemma 4, since\u2202E \u2202x0\u22a4and \u2202D \u2202z0share the same range, \u2207x0L(D(E(x0)), y) =\u2202L \u2202D\u2202D \u2202E\u2202E \u2202x0|t= (\u2202E \u2202x0\u22a4\u2202D \u2202z0\u22a4\u2202L \u2202x0\u22a4)\u22a4lies in the tangent space at x0. B.4 T HEORETICAL ANALYSIS ON MPGD-Z In this section, we provide the theoretical analysis and proof for algorithm MPGD-Z as a proposition to Theorem 2. Proposition 3. If an autoencoder with encoder Eand decoder Dis a perfect autoencoder for X \u2282 M , then D(\u2207z0L(D(z0));y) =D(\u2202L \u2202D\u2202D \u2202z0)\u2208 Tx0M. Proof. Similar to Theorem 2, by Lemma 4\u2202D \u2202z0= (\u2202E \u2202x0)\u22a4. Therefore, \u2207z0L(D(z0));y) = \u2202L \u2202D\u2202D \u2202z0=\u2202L \u2202D(\u2202E \u2202x0)\u22a4\u2208 Tz0Z \u2282 Z . Because we have linear manifold assumption, the updated z0|t=z0|t+\u2207z0L(D(z0));y)is also on Z. Since Dis surjective to M,D(z0+\u2207z0L(D(z0));y) is on the data manifold. 18Preprint Hence the update rules for MPGD-Z is on-manifold. Notice that in practice the autoencoder can exhibit reconstruction error. To mitigate this problem, we add the inference time reconstruction error x0|t\u2212D(E(x0|t))back to the guided clean data estimation after the update. In other words, we use the empirical update rule z0|t=E(x0|t) \u2206x0|t=x0|t\u2212D(z0|t) z0|t=z0|t\u2212ct\u2207z0|tL(D(z0|t);y) x0|t=D(z0|t) + \u2206 x0|t This empirical update rule can be viewed as adding a weighted regularization term \u03bbrec\u2225x0|t\u2212 SG(D(E(x0|t)))\u22252to the guidance loss where \u03bbrecis a scalar weight and SG (D(E(x0|t)))is the reconstructed clean data estimation that is fixed before any guidance update (SG here denotes \u201dstop gradient\u201d). B.5 T HEORETICAL ANALYSIS ON MPGD-LDM In this section, we provide the theoretical analysis for algorithm MPGD-LDM with perfect autoen- coder assumption. Proposition 4. If an autoencoder with encoder Eand decoder Dis a perfect autoencoder for X \u2282 M , then a guided latent diffusion sampling described in Algorithm 2 can generate a sample x0\u2208 M . Proof. Since we have a perfect autoencoder, the latent space is exactly Rk. As a result, the latent diffusion process will not move the latent sample out of the latent space. Because Dis surjective to the manifold, D(z0)is on the data manifold. C E MPIRICAL VERIFICATION OF THE MANIFOLD PRESERVING ABILITIES 0.0 0.2 0.4 0.6 0.8 1.0 Diffusion Time Step0.0000.0250.0500.0750.1000.1250.1500.175Deviation from the manifoldDPS MPGD w/o Proj. MPGD-AE (0.50.3) MPGD-AE (0.50) Figure 8: We analyze the deviation from the manifold throughout the diffusion process for different EXPERIMENTS We provide diverse additional qualitative RESULTS IN THE background of the image is predominantly white, prominent Gaussian noise-like patterns remain in the final Results of FaceID guidance generation with Stable Diffusion and different optimization algorithms. In particular, non-linear conjugate gradient method improves the guidance quality while maintaining the fidelity, suggesting a promising direction for future investigation. Input Reference FreeDoM w/ time-traveling (Baseline)FreeDoM w/o time-traveling (Baseline)MPGD-LDM w/o time-traveling (Ours) Text prompt: \u201ca cat wearing glasses\u201d Figure 19: Additional qualitative examples of style guidance Stable Diffusion generation. F L IMITATIONS F.1 F",
            "references": [
                {
                    "title": "Adapt and Diffuse: Sample-adaptive Reconstruction via Latent Diffusion Models",
                    "abstract": "Inverse problems arise in a multitude of applications, where the goal is to recover a clean signal from noisy and possibly (non)linear observations. The difficulty of a reconstruction problem depends on multiple factors, such as the structure of the ground truth signal, the severity of the degradation and the complex interactions between the above. This results in natural sample-by-sample variation in the difficulty of a reconstruction task, which is often overlooked by contemporary techniques. Our key observation is that most existing inverse problem solvers lack the ability to adapt their compute power to the difficulty of the reconstruction task, resulting in subpar performance and wasteful resource allocation. We propose a novel method that we call severity encoding, to estimate the degradation severity of noisy, degraded signals in the latent space of an autoencoder. We show that the estimated severity has strong correlation with the true corruption level and can give useful hints at the difficulty of reconstruction problems on a sample-by-sample basis. Furthermore, we propose a reconstruction method based on latent diffusion models that leverages the predicted degradation severities to fine-tune the reverse diffusion sampling trajectory and thus achieve sample-adaptive inference times. Our framework acts as a wrapper that can be combined with any latent diffusion-based baseline solver, imbuing it with sample-adaptivity and acceleration. We perform numerical experiments on both linear and nonlinear inverse problems and demonstrate that our technique greatly improves the performance of the baseline solver and achieves up to 10\u00d7 acceleration in mean sampling speed."
                },
                {
                    "title": "Solving Inverse Problems with Latent Diffusion Models via Hard Data Consistency",
                    "abstract": "Diffusion models have recently emerged as powerful generative priors for solving inverse problems. However, training diffusion models in the pixel space are both data-intensive and computationally demanding, which restricts their applicability as priors for high-dimensional real-world data such as medical images. Latent diffusion models, which operate in a much lower-dimensional space, offer a solution to these challenges. However, incorporating latent diffusion models to solve inverse problems remains a challenging problem due to the nonlinearity of the encoder and decoder. To address these issues, we propose \\textit{ReSample}, an algorithm that can solve general inverse problems with pre-trained latent diffusion models. Our algorithm incorporates data consistency by solving an optimization problem during the reverse sampling process, a concept that we term as hard data consistency. Upon solving this optimization problem, we propose a novel resampling scheme to map the measurement-consistent sample back onto the noisy data manifold and theoretically demonstrate its benefits. Lastly, we apply our algorithm to solve a wide range of linear and nonlinear inverse problems in both natural and medical images, demonstrating that our approach outperforms existing state-of-the-art approaches, including those based on pixel-space diffusion models."
                },
                {
                    "title": "Solving Linear Inverse Problems Provably via Posterior Sampling with Latent Diffusion Models",
                    "abstract": "We present the first framework to solve linear inverse problems leveraging pre-trained latent diffusion models. Previously proposed algorithms (such as DPS and DDRM) only apply to pixel-space diffusion models. We theoretically analyze our algorithm showing provable sample recovery in a linear model setting. The algorithmic insight obtained from our analysis extends to more general settings often considered in practice. Experimentally, we outperform previously proposed posterior sampling algorithms in a wide variety of problems including random inpainting, block inpainting, denoising, deblurring, destriping, and super-resolution."
                },
                {
                    "title": "Lifting Architectural Constraints of Injective Flows",
                    "abstract": "Normalizing Flows explicitly maximize a full-dimensional likelihood on the training data. However, real data is typically only supported on a lower-dimensional manifold leading the model to expend significant compute on modeling noise. Injective Flows fix this by jointly learning a manifold and the distribution on it. So far, they have been limited by restrictive architectures and/or high computational cost. We lift both constraints by a new efficient estimator for the maximum likelihood loss, compatible with free-form bottleneck architectures. We further show that naively learning both the data manifold and the distribution on it can lead to divergent solutions, and use this insight to motivate a stable maximum likelihood training objective. We perform extensive experiments on toy, tabular and image data, demonstrating the competitive performance of the resulting model."
                },
                {
                    "title": "Which Models have Perceptually-Aligned Gradients? An Explanation via Off-Manifold Robustness",
                    "abstract": "One of the remarkable properties of robust computer vision models is that their input-gradients are often aligned with human perception, referred to in the literature as perceptually-aligned gradients (PAGs). Despite only being trained for classification, PAGs cause robust models to have rudimentary generative capabilities, including image generation, denoising, and in-painting. However, the underlying mechanisms behind these phenomena remain unknown. In this work, we provide a first explanation of PAGs via \\emph{off-manifold robustness}, which states that models must be more robust off- the data manifold than they are on-manifold. We first demonstrate theoretically that off-manifold robustness leads input gradients to lie approximately on the data manifold, explaining their perceptual alignment. We then show that Bayes optimal models satisfy off-manifold robustness, and confirm the same empirically for robust models trained via gradient norm regularization, randomized smoothing, and adversarial training with projected gradient descent. Quantifying the perceptual alignment of model gradients via their similarity with the gradients of generative models, we show that off-manifold robustness correlates well with perceptual alignment. Finally, based on the levels of on- and off-manifold robustness, we identify three different regimes of robustness that affect both perceptual alignment and model accuracy: weak robustness, bayes-aligned robustness, and excessive robustness. Code is available at \\url{https://github.com/tml-tuebingen/pags}."
                },
                {
                    "title": "A Variational Perspective on Solving Inverse Problems with Diffusion Models",
                    "abstract": "Diffusion models have emerged as a key pillar of foundation models in visual domains. One of their critical applications is to universally solve different downstream inverse tasks via a single diffusion prior without re-training for each task. Most inverse tasks can be formulated as inferring a posterior distribution over data (e.g., a full image) given a measurement (e.g., a masked image). This is however challenging in diffusion models since the nonlinear and iterative nature of the diffusion process renders the posterior intractable. To cope with this challenge, we propose a variational approach that by design seeks to approximate the true posterior distribution. We show that our approach naturally leads to regularization by denoising diffusion process (RED-Diff) where denoisers at different timesteps concurrently impose different structural constraints over the image. To gauge the contribution of denoisers from different timesteps, we propose a weighting mechanism based on signal-to-noise-ratio (SNR). Our approach provides a new variational perspective for solving inverse problems with diffusion models, allowing us to formulate sampling as stochastic optimization, where one can simply apply off-the-shelf solvers with lightweight iterates. Our experiments for image restoration tasks such as inpainting and superresolution demonstrate the strengths of our method compared with state-of-the-art sampling-based diffusion models."
                },
                {
                    "title": "End-to-End Diffusion Latent Optimization Improves Classifier Guidance",
                    "abstract": "Classifier guidance\u2014using the gradients of an image classifier to steer the generations of a diffusion model\u2014has the potential to dramatically expand the creative control over image generation and editing. However, currently classifier guidance requires either training new noise-aware models to obtain accurate gradients or using a one-step denoising approximation of the final generation, which leads to misaligned gradients and sub-optimal control. We highlight this approximation\u2019s shortcomings and propose a novel guidance method: Direct Optimization of Diffusion Latents (DOODL), which enables plug-and-play guidance by optimizing diffusion latents w.r.t. the gradients of a pre-trained classifier on the true generated pixels, using an invertible diffusion process to achieve memory-efficient backpropagation. Showcasing the potential of more precise guidance, DOODL outperforms one-step classifier guidance on computational and human evaluation metrics across different forms of guidance: using CLIP guidance to improve generations of complex prompts from DrawBench, using fine-grained visual classifiers to expand the vocabulary of Stable Diffusion, enabling image-conditioned generation with a CLIP visual encoder, and improving image aesthetics using an aesthetic scoring network."
                },
                {
                    "title": "FreeDoM: Training-Free Energy-Guided Conditional Diffusion Model",
                    "abstract": "Recently, conditional diffusion models have gained popularity in numerous applications due to their exceptional generation ability. However, many existing methods are training-required. They need to train a time-dependent classifier or a condition-dependent score estimator, which increases the cost of constructing conditional diffusion models and is inconvenient to transfer across different conditions. Some current works aim to overcome this limitation by proposing training-free solutions, but most can only be applied to a specific category of tasks and not to more general conditions. In this work, we propose a training-Free conditional Diffusion Model (FreeDoM) used for various conditions. Specifically, we leverage off-the-shelf pretrained networks, such as a face detection model, to construct time-independent energy functions, which guide the generation process without requiring training. Furthermore, because the construction of the energy function is very flexible and adaptable to various conditions, our proposed FreeDoM has a broader range of applications than existing training-free methods. FreeDoM is advantageous in its simplicity, effectiveness, and low cost. Experiments demonstrate that FreeDoM is effective for various conditions and suitable for diffusion models of diverse data domains, including image and latent code domains. Code is available at https://github.com/vvictoryuki/FreeDoM."
                },
                {
                    "title": "Decomposed Diffusion Sampler for Accelerating Large-Scale Inverse Problems",
                    "abstract": "Krylov subspace, which is generated by multiplying a given vector by the matrix of a linear transformation and its successive powers, has been extensively studied in classical optimization literature to design algorithms that converge quickly for large linear inverse problems. For example, the conjugate gradient method (CG), one of the most popular Krylov subspace methods, is based on the idea of minimizing the residual error in the Krylov subspace. However, with the recent advancement of high-performance diffusion solvers for inverse problems, it is not clear how classical wisdom can be synergistically combined with modern diffusion models. In this study, we propose a novel and efficient diffusion sampling strategy that synergistically combines the diffusion sampling and Krylov subspace methods. Specifically, we prove that if the tangent space at a denoised sample by Tweedie's formula forms a Krylov subspace, then the CG initialized with the denoised data ensures the data consistency update to remain in the tangent space. This negates the need to compute the manifold-constrained gradient (MCG), leading to a more efficient diffusion sampling method. Our method is applicable regardless of the parametrization and setting (i.e., VE, VP). Notably, we achieve state-of-the-art reconstruction quality on challenging real-world medical inverse imaging problems, including multi-coil MRI reconstruction and 3D CT reconstruction. Moreover, our proposed method achieves more than 80 times faster inference time than the previous state-of-the-art method. Code is available at https://github.com/HJ-harry/DDS"
                },
                {
                    "title": "T2I-Adapter: Learning Adapters to Dig out More Controllable Ability for Text-to-Image Diffusion Models",
                    "abstract": "The incredible generative ability of large-scale text-to-image (T2I) models has demonstrated strong power of learning complex structures and meaningful semantics. However, relying solely on text prompts cannot fully take advantage of the knowledge learned by the model, especially when flexible and accurate controlling (e.g., structure and color) is needed. In this paper, we aim to ``dig out\" the capabilities that T2I models have implicitly learned, and then explicitly use them to control the generation more granularly. Specifically, we propose to learn low-cost T2I-Adapters to align internal knowledge in T2I models with external control signals, while freezing the original large T2I models. In this way, we can train various adapters according to different conditions, achieving rich control and editing effects in the color and structure of the generation results. Further, the proposed T2I-Adapters have attractive properties of practical value, such as composability and generalization ability. Extensive experiments demonstrate that our T2I-Adapter has promising generation quality and a wide range of applications. Our code is available at https://github.com/TencentARC/T2I-Adapter."
                },
                {
                    "title": "Universal Guidance for Diffusion Models",
                    "abstract": "Typical diffusion models are trained to accept a particular form of conditioning, most commonly text, and cannot be conditioned on other modalities without retraining. In this work, we propose a universal guidance algorithm that enables diffusion models to be controlled by arbitrary guidance modalities without the need to retrain any use-specific components. We show that our algorithm successfully generates quality images with guidance functions including segmentation, face recognition, object detection, and classifier signals. Code is available at github.com/arpitbansal297/Universal-Guided-Diffusion."
                },
                {
                    "title": "Adding Conditional Control to Text-to-Image Diffusion Models",
                    "abstract": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models."
                },
                {
                    "title": "GLIGEN: Open-Set Grounded Text-to-Image Generation",
                    "abstract": "Large-scale text-to-image diffusion models have made amazing advances. However, the status quo is to use text input alone, which can impede controllability. In this work, we propose GLIGEN, Grounded-Language-to-Image Generation, a novel approach that builds upon and extends the functionality of existing pre-trained text-to-image diffusion models by enabling them to also be conditioned on grounding inputs. To preserve the vast concept knowledge of the pre-trained model, we freeze all of its weights and inject the grounding information into new trainable layers via a gated mechanism. Our model achieves open-world grounded text2img generation with caption and bounding box condition inputs, and the grounding ability generalizes well to novel spatial configurations and concepts. GLIGEN's zero-shot performance on COCO and LVIS outperforms existing supervised layout-to-image baselines by a large margin."
                },
                {
                    "title": "Your diffusion model secretly knows the dimension of the data manifold",
                    "abstract": "In this work, we propose a novel framework for estimating the dimension of the data manifold using a trained diffusion model. A diffusion model approximates the score function i.e. the gradient of the log density of a noise-corrupted version of the target distribution for varying levels of corruption. We prove that, if the data concentrates around a manifold embedded in the high-dimensional ambient space, then as the level of corruption decreases, the score function points towards the manifold, as this direction becomes the direction of maximal likelihood increase. Therefore, for small levels of corruption, the diffusion model provides us with access to an approximation of the normal bundle of the data manifold. This allows us to estimate the dimension of the tangent space, thus, the intrinsic dimension of the data manifold. To the best of our knowledge, our method is the first estimator of the data manifold dimension based on diffusion models and it outperforms well established statistical estimators in controlled experiments on both Euclidean and image data."
                },
                {
                    "title": "Zero-Shot Image Restoration Using Denoising Diffusion Null-Space Model",
                    "abstract": "Most existing Image Restoration (IR) models are task-specific, which can not be generalized to different degradation operators. In this work, we propose the Denoising Diffusion Null-Space Model (DDNM), a novel zero-shot framework for arbitrary linear IR problems, including but not limited to image super-resolution, colorization, inpainting, compressed sensing, and deblurring. DDNM only needs a pre-trained off-the-shelf diffusion model as the generative prior, without any extra training or network modifications. By refining only the null-space contents during the reverse diffusion process, we can yield diverse results satisfying both data consistency and realness. We further propose an enhanced and robust version, dubbed DDNM+, to support noisy restoration and improve restoration quality for hard tasks. Our experiments on several IR tasks reveal that DDNM outperforms other state-of-the-art zero-shot IR methods. We also demonstrate that DDNM+ can solve complex real-world applications, e.g., old photo restoration."
                },
                {
                    "title": "Diffusion Posterior Sampling for General Noisy Inverse Problems",
                    "abstract": "Diffusion models have been recently studied as powerful generative inverse problem solvers, owing to their high quality reconstructions and the ease of combining existing iterative solvers. However, most works focus on solving simple linear inverse problems in noiseless settings, which significantly under-represents the complexity of real-world problems. In this work, we extend diffusion solvers to efficiently handle general noisy (non)linear inverse problems via approximation of the posterior sampling. Interestingly, the resulting posterior sampling scheme is a blended version of diffusion sampling with the manifold constrained gradient without a strict measurement consistency projection step, yielding a more desirable generative path in noisy settings compared to the previous studies. Our method demonstrates that diffusion models can incorporate various measurement noise statistics such as Gaussian and Poisson, and also efficiently handle noisy nonlinear inverse problems such as Fourier phase retrieval and non-uniform deblurring. Code available at https://github.com/DPS2022/diffusion-posterior-sampling"
                },
                {
                    "title": "DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation",
                    "abstract": "Large text-to-image models achieved a remarkable leap in the evolution of AI, enabling high-quality and diverse synthesis of images from a given text prompt. However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts. In this work, we present a new approach for \u201cpersonalization\u201d of text-to-image diffusion models. Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject. Once the subject is embedded in the output domain of the model, the unique identifier can be used to synthesize novel photorealistic images of the subject contextualized in different scenes. By leveraging the semantic prior embedded in the model with a new autogenous class-specific prior preservation loss, our technique enables synthesizing the subject in diverse scenes, poses, views and lighting conditions that do not appear in the reference images. We apply our technique to several previously-unassailable tasks, including subject recontextualization, text-guided view synthesis, and artistic rendering, all while preserving the subject's key features. We also provide a new dataset and evaluation protocol for this new task of subject-driven generation. Project page: https://dreambooth.github.io/"
                },
                {
                    "title": "An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion",
                    "abstract": "Text-to-image models offer unprecedented freedom to guide creation through natural language. Yet, it is unclear how such freedom can be exercised to generate images of specific unique concepts, modify their appearance, or compose them in new roles and novel scenes. In other words, we ask: how can we use language-guided models to turn our cat into a painting, or imagine a new product based on our favorite toy? Here we present a simple approach that allows such creative freedom. Using only 3-5 images of a user-provided concept, like an object or a style, we learn to represent it through new\"words\"in the embedding space of a frozen text-to-image model. These\"words\"can be composed into natural language sentences, guiding personalized creation in an intuitive way. Notably, we find evidence that a single word embedding is sufficient for capturing unique and varied concepts. We compare our approach to a wide range of baselines, and demonstrate that it can more faithfully portray the concepts across a range of applications and tasks. Our code, data and new words will be available at: https://textual-inversion.github.io"
                },
                {
                    "title": "Scaling Autoregressive Models for Content-Rich Text-to-Image Generation",
                    "abstract": "We present the Pathways Autoregressive Text-to-Image (Parti) model, which generates high-fidelity photorealistic images and supports content-rich synthesis involving complex compositions and world knowledge. Parti treats text-to-image generation as a sequence-to-sequence modeling problem, akin to machine translation, with sequences of image tokens as the target outputs rather than text tokens in another language. This strategy can naturally tap into the rich body of prior work on large language models, which have seen continued advances in capabilities and performance through scaling data and model sizes. Our approach is simple: First, Parti uses a Transformer-based image tokenizer, ViT-VQGAN, to encode images as sequences of discrete tokens. Second, we achieve consistent quality improvements by scaling the encoder-decoder Transformer model up to 20B parameters, with a new state-of-the-art zero-shot FID score of 7.23 and finetuned FID score of 3.22 on MS-COCO. Our detailed analysis on Localized Narratives as well as PartiPrompts (P2), a new holistic benchmark of over 1600 English prompts, demonstrate the effectiveness of Parti across a wide variety of categories and difficulty aspects. We also explore and highlight limitations of our models in order to define and exemplify key areas of focus for further improvements. See https://parti.research.google/ for high-resolution images."
                },
                {
                    "title": "Diffusion models as plug-and-play priors",
                    "abstract": "We consider the problem of inferring high-dimensional data $\\mathbf{x}$ in a model that consists of a prior $p(\\mathbf{x})$ and an auxiliary differentiable constraint $c(\\mathbf{x},\\mathbf{y})$ on $x$ given some additional information $\\mathbf{y}$. In this paper, the prior is an independently trained denoising diffusion generative model. The auxiliary constraint is expected to have a differentiable form, but can come from diverse sources. The possibility of such inference turns diffusion models into plug-and-play modules, thereby allowing a range of potential applications in adapting models to new domains and tasks, such as conditional generation or image segmentation. The structure of diffusion models allows us to perform approximate inference by iterating differentiation through the fixed denoising network enriched with different amounts of noise at each step. Considering many noised versions of $\\mathbf{x}$ in evaluation of its fitness is a novel search mechanism that may lead to new algorithms for solving combinatorial optimization problems."
                },
                {
                    "title": "The Manifold Hypothesis for Gradient-Based Explanations",
                    "abstract": "When do gradient-based explanation algorithms provide perceptually-aligned explanations? We propose a criterion: the feature attributions need to be aligned with the tangent space of the data manifold. To provide evidence for this hypothesis, we introduce a framework based on variational autoencoders that allows to estimate and generate image manifolds. Through experiments across a range of different datasets \u2013 MNIST, EMNIST, CIFAR10, X-ray pneumonia and Diabetic Retinopathy detection \u2013 we demonstrate that the more a feature attribution is aligned with the tangent space of the data, the more perceptually-aligned it tends to be. We then show that the attributions provided by popular post-hoc methods such as Integrated Gradients and SmoothGrad are more strongly aligned with the data manifold than the raw gradient. Adversarial training also improves the alignment of model gradients with the data manifold. As a consequence, we suggest that explanation algorithms should actively strive to align their explanations with the data manifold. An extended version of this paper is available at https://arxiv.org/abs/2206.07387."
                },
                {
                    "title": "Improving Diffusion Models for Inverse Problems using Manifold Constraints",
                    "abstract": "Recently, diffusion models have been used to solve various inverse problems in an unsupervised manner with appropriate modifications to the sampling process. However, the current solvers, which recursively apply a reverse diffusion step followed by a projection-based measurement consistency step, often produce suboptimal results. By studying the generative sampling path, here we show that current solvers throw the sample path off the data manifold, and hence the error accumulates. To address this, we propose an additional correction term inspired by the manifold constraint, which can be used synergistically with the previous solvers to make the iterations close to the manifold. The proposed manifold constraint is straightforward to implement within a few lines of code, yet boosts the performance by a surprisingly large margin. With extensive experiments, we show that our method is superior to the previous methods both theoretically and empirically, producing promising results in many applications such as image inpainting, colorization, and sparse-view computed tomography. Code available https://github.com/HJ-harry/MCG_diffusion"
                },
                {
                    "title": "Denoising Diffusion Restoration Models",
                    "abstract": "Many interesting tasks in image restoration can be cast as linear inverse problems. A recent family of approaches for solving these problems uses stochastic algorithms that sample from the posterior distribution of natural images given the measurements. However, efficient solutions often require problem-specific supervised training to model the posterior, whereas unsupervised methods that are not problem-specific typically rely on inefficient iterative methods. This work addresses these issues by introducing Denoising Diffusion Restoration Models (DDRM), an efficient, unsupervised posterior sampling method. Motivated by variational inference, DDRM takes advantage of a pre-trained denoising diffusion generative model for solving any linear inverse problem. We demonstrate DDRM's versatility on several image datasets for super-resolution, deblurring, inpainting, and colorization under various amounts of measurement noise. DDRM outperforms the current leading unsupervised methods on the diverse ImageNet dataset in reconstruction quality, perceptual quality, and runtime, being 5x faster than the nearest competitor. DDRM also generalizes well for natural images out of the distribution of the observed ImageNet training set."
                },
                {
                    "title": "RePaint: Inpainting using Denoising Diffusion Probabilistic Models",
                    "abstract": "Free-form inpainting is the task of adding new content to an image in the regions specified by an arbitrary binary mask. Most existing approaches train for a certain distribution of masks, which limits their generalization capabilities to unseen mask types. Furthermore, training with pixel-wise and perceptual losses often leads to simple textural extensions towards the missing areas instead of semantically meaningful generation. In this work, we propose RePaint: A Denoising Diffusion Probabilistic Model (DDPM) based inpainting approach that is applicable to even extreme masks. We employ a pretrained unconditional DDPM as the generative prior. To condition the generation process, we only alter the reverse diffusion iterations by sampling the unmasked regions using the given image infor-mation. Since this technique does not modify or condition the original DDPM network itself, the model produces high-quality and diverse output images for any inpainting form. We validate our method for both faces and general-purpose image inpainting using standard and extreme masks. Re-Paint outperforms state-of-the-art Autoregressive, and GAN approaches for at least five out of six mask distributions. Github Repository: git.io/RePaint"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models",
                    "abstract": "Diffusion models have recently been shown to generate high-quality synthetic images, especially when paired with a guidance technique to trade off diversity for fidelity. We explore diffusion models for the problem of text-conditional image synthesis and compare two different guidance strategies: CLIP guidance and classifier-free guidance. We find that the latter is preferred by human evaluators for both photorealism and caption similarity, and often produces photorealistic samples. Samples from a 3.5 billion parameter text-conditional diffusion model using classifier-free guidance are favored by human evaluators to those from DALL-E, even when the latter uses expensive CLIP reranking. Additionally, we find that our models can be fine-tuned to perform image inpainting, enabling powerful text-driven image editing. We train a smaller model on a filtered dataset and release the code and weights at https://github.com/openai/glide-text2im."
                },
                {
                    "title": "ILVR: Conditioning Method for Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) have shown remarkable performance in unconditional image generation. However, due to the stochasticity of the generative process in DDPM, it is challenging to generate images with the desired semantics. In this work, we propose Iterative Latent Variable Refinement (ILVR), a method to guide the generative process in DDPM to generate high-quality images based on a given reference image. Here, the refinement of the generative process in DDPM enables a single DDPM to sample images from various sets directed by the reference image. The proposed ILVR method generates high-quality images while controlling the generation. The controllability of our method allows adaptation of a single DDPM without any additional learning in various image generation tasks, such as generation from various downsampling factors, multi-domain image translation, paint-to-image, and editing with scribbles."
                },
                {
                    "title": "SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations",
                    "abstract": "Guided image synthesis enables everyday users to create and edit photo-realistic images with minimum effort. The key challenge is balancing faithfulness to the user input (e.g., hand-drawn colored strokes) and realism of the synthesized image. Existing GAN-based methods attempt to achieve such balance using either conditional GANs or GAN inversions, which are challenging and often require additional training data or loss functions for individual applications. To address these issues, we introduce a new image synthesis and editing method, Stochastic Differential Editing (SDEdit), based on a diffusion model generative prior, which synthesizes realistic images by iteratively denoising through a stochastic differential equation (SDE). Given an input image with user guide of any type, SDEdit first adds noise to the input, then subsequently denoises the resulting image through the SDE prior to increase its realism. SDEdit does not require task-specific training or inversions and can naturally achieve the balance between realism and faithfulness. SDEdit significantly outperforms state-of-the-art GAN-based methods by up to 98.09% on realism and 91.72% on overall satisfaction scores, according to a human perception study, on multiple tasks, including stroke-based image synthesis and editing as well as image compositing."
                },
                {
                    "title": "SNIPS: Solving Noisy Inverse Problems Stochastically",
                    "abstract": "In this work we introduce a novel stochastic algorithm dubbed SNIPS, which draws samples from the posterior distribution of any linear inverse problem, where the observation is assumed to be contaminated by additive white Gaussian noise. Our solution incorporates ideas from Langevin dynamics and Newton's method, and exploits a pre-trained minimum mean squared error (MMSE) Gaussian denoiser. The proposed approach relies on an intricate derivation of the posterior score function that includes a singular value decomposition (SVD) of the degradation operator, in order to obtain a tractable iterative algorithm for the desired sampling. Due to its stochasticity, the algorithm can produce multiple high perceptual quality samples for the same noisy observation. We demonstrate the abilities of the proposed paradigm for image deblurring, super-resolution, and compressive sensing. We show that the samples produced are sharp, detailed and consistent with the given measurements, and their diversity exposes the inherent uncertainty in the inverse problem being solved."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "Taming Transformers for High-Resolution Image Synthesis",
                    "abstract": "Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks. In contrast to CNNs, they contain no inductive bias that prioritizes local interactions. This makes them expressive, but also computationally infeasible for long sequences, such as high-resolution images. We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images. We show how to (i) use CNNs to learn a context-rich vocabulary of image constituents, and in turn (ii) utilize transformers to efficiently model their composition within high-resolution images. Our approach is readily applied to conditional synthesis tasks, where both non-spatial information, such as object classes, and spatial information, such as segmentations, can control the generated image. In particular, we present the first results on semantically-guided synthesis of megapixel images with transformers. Project page at https://git.io/JLlvY."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "Fairwashing Explanations with Off-Manifold Detergent",
                    "abstract": "Explanation methods promise to make black-box classifiers more transparent. As a result, it is hoped that they can act as proof for a sensible, fair and trustworthy decision-making process of the algorithm and thereby increase its acceptance by the end-users. In this paper, we show both theoretically and experimentally that these hopes are presently unfounded. Specifically, we show that, for any classifier $g$, one can always construct another classifier $\\tilde{g}$ which has the same behavior on the data (same train, validation, and test error) but has arbitrarily manipulated explanation maps. We derive this statement theoretically using differential geometry and demonstrate it experimentally for various explanation methods, architectures, and datasets. Motivated by our theoretical insights, we then propose a modification of existing explanation methods which makes them significantly more robust."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Foundations of Data Science",
                    "abstract": "Computer science as an academic discipline began in the 1960\u2019s. Emphasis was on programming languages, compilers, operating systems, and the mathematical theory that supported these areas. Courses in theoretical computer science covered finite automata, regular expressions, context-free languages, and computability. In the 1970\u2019s, the study of algorithms was added as an important component of theory. The emphasis was on making computers useful. Today, a fundamental change is taking place and the focus is more on applications. There are many reasons for this change. The merging of computing and communications has played an important role. The enhanced ability to observe, collect, and store data in the natural sciences, in commerce, and in other fields calls for a change in our understanding of data and how to handle it in the modern setting. The emergence of the web and social networks as central aspects of daily life presents both opportunities and challenges for theory."
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "Semantic Image Synthesis With Spatially-Adaptive Normalization",
                    "abstract": "We propose spatially-adaptive normalization, a simple but effective layer for synthesizing photorealistic images given an input semantic layout. Previous methods directly feed the semantic layout as input to the network, forcing the network to memorize the information throughout all the layers. Instead, we propose using the input layout for modulating the activations in normalization layers through a spatially-adaptive, learned affine transformation. Experiments on several challenging datasets demonstrate the superiority of our method compared to existing approaches, regarding both visual fidelity and alignment with input layouts. Finally, our model allows users to easily control the style and content of image synthesis results as well as create multi-modal results. Code is available upon publication."
                },
                {
                    "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                    "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                },
                {
                    "title": "ArcFace: Additive Angular Margin Loss for Deep Face Recognition",
                    "abstract": "One of the main challenges in feature learning using Deep Convolutional Neural Networks (DCNNs) for large-scale face recognition is the design of appropriate loss functions that can enhance the discriminative power. Centre loss penalises the distance between deep features and their corresponding class centres in the Euclidean space to achieve intra-class compactness. SphereFace assumes that the linear transformation matrix in the last fully connected layer can be used as a representation of the class centres in the angular space and therefore penalises the angles between deep features and their corresponding weights in a multiplicative way. Recently, a popular line of research is to incorporate margins in well-established loss functions in order to maximise face class separability. In this paper, we propose an Additive Angular Margin Loss (ArcFace) to obtain highly discriminative features for face recognition. The proposed ArcFace has a clear geometric interpretation due to its exact correspondence to geodesic distance on a hypersphere. We present arguably the most extensive experimental evaluation against all recent state-of-the-art face recognition methods on ten face recognition benchmarks which includes a new large-scale image database with trillions of pairs and a large-scale video dataset. We show that ArcFace consistently outperforms the state of the art and can be easily implemented with negligible computational overhead. To facilitate future research, the code has been made available."
                },
                {
                    "title": "The Unreasonable Effectiveness of Deep Features as a Perceptual Metric",
                    "abstract": "While it is nearly effortless for humans to quickly assess the perceptual similarity between two images, the underlying processes are thought to be quite complex. Despite this, the most widely used perceptual metrics today, such as PSNR and SSIM, are simple, shallow functions, and fail to account for many nuances of human perception. Recently, the deep learning community has found that features of the VGG network trained on ImageNet classification has been remarkably useful as a training loss for image synthesis. But how perceptual are these so-called \"perceptual losses\"? What elements are critical for their success? To answer these questions, we introduce a new dataset of human perceptual similarity judgments. We systematically evaluate deep features across different architectures and tasks and compare them with classic metrics. We find that deep features outperform all previous metrics by large margins on our dataset. More surprisingly, this result is not restricted to ImageNet-trained VGG features, but holds across different deep architectures and levels of supervision (supervised, self-supervised, or even unsupervised). Our results suggest that perceptual similarity is an emergent property shared across deep visual representations."
                },
                {
                    "title": "Demystifying MMD GANs",
                    "abstract": "We investigate the training and performance of generative adversarial networks using the Maximum Mean Discrepancy (MMD) as critic, termed MMD GANs. As our main theoretical contribution, we clarify the situation with bias in GAN loss functions raised by recent work: we show that gradient estimators used in the optimization process for both MMD GANs and Wasserstein GANs are unbiased, but learning a discriminator based on samples leads to biased gradients for the generator parameters. We also discuss the issue of kernel choice for the MMD critic, and characterize the kernel corresponding to the energy distance used for the Cramer GAN critic. Being an integral probability metric, the MMD benefits from training strategies recently developed for Wasserstein GANs. In experiments, the MMD GAN is able to employ a smaller critic network than the Wasserstein GAN, resulting in a simpler and faster-training algorithm with matching performance. We also propose an improved measure of GAN convergence, the Kernel Inception Distance, and show how to use it to dynamically adapt learning rates during GAN training."
                },
                {
                    "title": "The Riemannian Geometry of Deep Generative Models",
                    "abstract": "Deep generative models learn a mapping from a low-dimensional latent space to a high-dimensional data space. Under certain regularity conditions, these models parameterize nonlinear manifolds in the data space. In this paper, we investigate the Riemannian geometry of these generated manifolds. First, we develop efficient algorithms for computing geodesic curves, which provide an intrinsic notion of distance between points on the manifold. Second, we develop an algorithm for parallel translation of a tangent vector along a path on the manifold. We show how parallel translation can be used to generate analogies, i.e., to transport a change in one data point into a semantically similar change of another data point. Our experiments on real image data show that the manifolds learned by deep generative models, while nonlinear, are surprisingly close to zero curvature. The practical implication is that linear paths in the latent space closely approximate geodesics on the generated manifold."
                },
                {
                    "title": "Neural Discrete Representation Learning",
                    "abstract": "Learning useful representations without supervision remains a key challenge in machine learning. In this paper, we propose a simple yet powerful generative model that learns such discrete representations. Our model, the Vector Quantised-Variational AutoEncoder (VQ-VAE), differs from VAEs in two key ways: the encoder network outputs discrete, rather than continuous, codes; and the prior is learnt rather than static. In order to learn a discrete latent representation, we incorporate ideas from vector quantisation (VQ). Using the VQ method allows the model to circumvent issues of \"posterior collapse\" -- where the latents are ignored when they are paired with a powerful autoregressive decoder -- typically observed in the VAE framework. Pairing these representations with an autoregressive prior, the model can generate high quality images, videos, and speech as well as doing high quality speaker conversion and unsupervised learning of phonemes, providing further evidence of the utility of the learnt representations."
                },
                {
                    "title": "Image-to-Image Translation with Conditional Adversarial Networks",
                    "abstract": "We investigate conditional adversarial networks as a general-purpose solution to image-to-image translation problems. These networks not only learn the mapping from input image to output image, but also learn a loss function to train this mapping. This makes it possible to apply the same generic approach to problems that traditionally would require very different loss formulations. We demonstrate that this approach is effective at synthesizing photos from label maps, reconstructing objects from edge maps, and colorizing images, among other tasks. Moreover, since the release of the pix2pix software associated with this paper, hundreds of twitter users have posted their own artistic experiments using our system. As a community, we no longer hand-engineer our mapping functions, and this work suggests we can achieve reasonable results without handengineering our loss functions either."
                },
                {
                    "title": "Large-scale Classification of Fine-Art Paintings: Learning The Right Metric on The Right Feature",
                    "abstract": "In the past few years, the number of fine-art collections that are digitized and publicly available has been growing rapidly. With the availability of such large collections of digitized artworks comes the need to develop multimedia systems to archive and retrieve this pool of data. Measuring the visual similarity between artistic items is an essential step for such multimedia systems, which can benefit more high-level multimedia tasks. In order to model this similarity between paintings, we should extract the appropriate visual features for paintings and find out the best approach to learn the similarity metric based on these features. We investigate a comprehensive list of visual features and metric learning approaches to learn an optimized similarity measure between paintings. We develop a machine that is able to make aesthetic-related semantic-level judgments, such as predicting a painting's style, genre, and artist, as well as providing similarity measures optimized based on the knowledge available in the domain of art historical interpretation. Our experiments show the value of using this similarity measure for the aforementioned prediction tasks."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "Tweedie\u2019s Formula and Selection Bias",
                    "abstract": "We suppose that the statistician observes some large number of estimates zi, each with its own unobserved expectation parameter \u03bci. The largest few of the zi\u2019s are likely to substantially overestimate their corresponding \u03bci\u2019s, this being an example of selection bias, or regression to the mean. Tweedie\u2019s formula, first reported by Robbins in 1956, offers a simple empirical Bayes approach for correcting selection bias. This article investigates its merits and limitations. In addition to the methodology, Tweedie\u2019s formula raises more general questions concerning empirical Bayes theory, discussed here as \u201crelevance\u201d and \u201cempirical Bayes information.\u201d There is a close connection between applications of the formula and James\u2013Stein estimation."
                },
                {
                    "title": "Bayesian Learning via Stochastic Gradient Langevin Dynamics",
                    "abstract": "In this paper we propose a new framework for learning from large scale datasets based on iterative learning from small mini-batches. By adding the right amount of noise to a standard stochastic gradient optimization algorithm we show that the iterates will converge to samples from the true posterior distribution as we anneal the stepsize. This seamless transition between optimization and Bayesian posterior sampling provides an inbuilt protection against overfitting. We also propose a practical method for Monte Carlo estimates of posterior statistics which monitors a \"sampling threshold\" and collects samples after it has been surpassed. We apply the method to three models: a mixture of Gaussians, logistic regression and ICA with natural gradients."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Loss-Guided Diffusion Models for Plug-and-Play Controllable Generation",
                    "abstract": "We consider guiding denoising diffusion models with general differentiable loss functions in a plug-and-play fashion, enabling controllable generation without additional training. This paradigm, termed Loss-Guided Diffusion (LGD), can easily be integrated into all diffusion models and leverage various efficient samplers. Despite the benefits, the resulting guidance term is, unfortunately, an intractable integral and needs to be approximated. Existing methods compute the guidance term based on a point estimate. However, we show that such approaches have significant errors over the scale of the approximations. To address this issue, we propose a Monte Carlo method that uses multiple samples from a suitable distribution to reduce bias. Our method is effective in various synthetic and real-world settings, including image super-resolution, text or label-conditional image generation, and controllable motion synthesis. Notably, we show how our method can be applied to control a pretrained motion diffusion model to follow certain paths and avoid obstacles that are proven challenging to prior methods."
                },
                {
                    "title": "A SURVEY OF NONLINEAR CONJUGATE GRADIENT METHODS",
                    "abstract": "This paper reviews the development of dierent versions of nonlinear conjugate gradient methods, with special attention given to global convergence properties."
                },
                {
                    "title": "ADAPTIVE ESTIMATION OF A QUADRATIC FUNCTIONAL BY MODEL SELECTION",
                    "abstract": "We consider the problem of estimating (cid:1) s (cid:1) 2 when s belongs to some separable Hilbert space and one observes the Gaussian process Y (cid:2) t (cid:3) = (cid:5) s(cid:4)t (cid:6) + \u03c3L (cid:2) t (cid:3) , for all t \u2208 (cid:1) , where L is some Gaussian isonormal process. This framework allows us in particular to consider the classical \u201cGaussian sequence model\u201d for which (cid:1) = l 2 (cid:2) (cid:2) \u2217 (cid:3) and L (cid:2) t (cid:3) = (cid:1) \u03bb \u2265 1 t \u03bb \u03b5 \u03bb , where (cid:2) \u03b5 \u03bb (cid:3) \u03bb \u2265 1 is a sequence of i.i.d. standard normal variables. Our approach consists in considering some at most countable families of \ufb01nite-dimensional linear subspaces of (cid:1) (the models ) and then using model selection via some con- veniently penalized least squares criterion to build new estimators of (cid:1) s (cid:1) 2 . We prove a general nonasymptotic risk bound which allows us to show that such penalized estimators are adaptive on a variety of collections of sets for the parameter s , depending on the family of models from which they are built. In particular, in the context of the Gaussian sequence model, a convenient choice of the family of models allows de\ufb01ning estimators which are adaptive over collections of hyperrectangles, ellipsoids, l p -bodies or Besov bodies. We take special care to describe the conditions under which the penalized estimator is ef\ufb01cient when the level of noise \u03c3 tends to zero. Our construction is an alternative to the one by Efro\u00a8\u0131movich and Low for hyperrectangles and provides new results otherwise."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we improve the performance and fidelity of diffusion models in generating high-quality samples while ensuring that the generated samples remain on the data manifold?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the field of generative models, particularly in applications such as image synthesis, where maintaining the integrity of the data manifold is essential for producing realistic outputs. A successful approach could lead to significant improvements in the quality of generated images, enhancing their applicability in various domains such as art, design, and virtual reality. Furthermore, it could inspire future research into more robust generative techniques and better understanding of the underlying data structures.\n\n### [Question 3] - Why is it hard?\nThe challenges in addressing this problem stem from the inherent complexity of ensuring that generated samples do not deviate from the data manifold during the diffusion process. Naive approaches may fail because they do not account for the intricate relationships between the latent space and the data manifold, leading to potential distortions in the generated samples. Technical obstacles include the need for precise gradient calculations and the management of reconstruction errors, which can complicate the optimization process. Additionally, ensuring that the autoencoder remains a perfect representation of the data distribution adds another layer of difficulty.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often overlooked the importance of maintaining manifold constraints during the generation process, leading to gaps in understanding how to effectively guide diffusion models. Existing solutions may have focused on improving sample quality without adequately addressing the manifold preservation aspect. Barriers such as the lack of robust theoretical frameworks and empirical validation of manifold-preserving techniques have hindered progress. Our approach differs by explicitly incorporating manifold constraints into the optimization process, leveraging theoretical insights from perfect autoencoders to guide the diffusion sampling effectively.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves the use of a guided latent diffusion sampling algorithm that incorporates a perfect autoencoder assumption. We will utilize a dataset of high-quality images and employ metrics such as fidelity and diversity to evaluate the generated samples. The expected outcomes include improved sample quality that adheres to the data manifold, as well as a reduction in reconstruction errors through the integration of a weighted regularization term. We anticipate that our approach will demonstrate superior performance compared to existing methods, particularly in maintaining the integrity of the generated outputs."
            }
        },
        "author_data": {
            "08c9a7ba-d666-4e5f-9848-a636c19c31d0": {
                "pk": "08c9a7ba-d666-4e5f-9848-a636c19c31d0",
                "name": "Yutong He",
                "collaborators": [
                    "Yuki Mitsufuji",
                    "Naoki Murata",
                    "Alexander Robey",
                    "Yiding Jiang",
                    "J. Williams",
                    "George J. Pappas",
                    "Hamed Hassani",
                    "Ruslan Salakhutdinov",
                    "J. Z. Kolter",
                    "AI Sony"
                ],
                "domain": [
                    "Text-to-Image Generation",
                    "Prompt Engineering",
                    "Visual-Language Processing",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Automated Black-box Prompt Engineering for Personalized Text-to-Image Generation",
                        "abstract": "Prompt engineering is effective for controlling the output of text-to-image (T2I) generative models, but it is also laborious due to the need for manually crafted prompts. This challenge has spurred the development of algorithms for automated prompt generation. However, these methods often struggle with transferability across T2I models, require white-box access to the underlying model, and produce non-intuitive prompts. In this work, we introduce PRISM, an algorithm that automatically identifies human-interpretable and transferable prompts that can effectively generate desired concepts given only black-box access to T2I models. Inspired by large language model (LLM) jailbreaking, PRISM leverages the in-context learning ability of LLMs to iteratively refine the candidate prompts distribution for given reference images. Our experiments demonstrate the versatility and effectiveness of PRISM in generating accurate prompts for objects, styles and images across multiple T2I models, including Stable Diffusion, DALL-E, and Midjourney."
                    },
                    {
                        "title": "Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion",
                        "abstract": "Consistency Models (CM) (Song et al., 2023) accelerate score-based diffusion model sampling at the cost of sample quality but lack a natural way to trade-off quality for speed. To address this limitation, we propose Consistency Trajectory Model (CTM), a generalization encompassing CM and score-based models as special cases. CTM trains a single neural network that can -- in a single forward pass -- output scores (i.e., gradients of log-density) and enables unrestricted traversal between any initial and final time along the Probability Flow Ordinary Differential Equation (ODE) in a diffusion process. CTM enables the efficient combination of adversarial training and denoising score matching loss to enhance performance and achieves new state-of-the-art FIDs for single-step diffusion model sampling on CIFAR-10 (FID 1.73) and ImageNet at 64x64 resolution (FID 1.92). CTM also enables a new family of sampling schemes, both deterministic and stochastic, involving long jumps along the ODE solution trajectories. It consistently improves sample quality as computational budgets increase, avoiding the degradation seen in CM. Furthermore, unlike CM, CTM's access to the score function can streamline the adoption of established controllable/conditional generation methods from the diffusion community. This access also enables the computation of likelihood. The code is available at https://github.com/sony/ctm."
                    },
                    {
                        "title": "Towards reporting bias in visual-language datasets: bimodal augmentation by decoupling object-attribute association",
                        "abstract": "Reporting bias arises when people assume that some knowledge is universally understood and hence, do not necessitate explicit elaboration. In this paper, we focus on the wide existence of reporting bias in visual-language datasets, embodied as the object-attribute association, which can subsequentially degrade models trained on them. To mitigate this bias, we propose a bimodal augmentation (BiAug) approach through object-attribute decoupling to flexibly synthesize visual-language examples with a rich array of object-attribute pairing and construct cross-modal hard negatives. We employ large language models (LLMs) in conjunction with a grounding object detector to extract target objects. Subsequently, the LLM generates a detailed attribute description for each object and produces a corresponding hard negative counterpart. An inpainting model is then used to create images based on these detailed object descriptions. By doing so, the synthesized examples explicitly complement omitted objects and attributes to learn, and the hard negative pairs steer the model to distinguish object attributes. Our experiments demonstrated that BiAug is superior in object-attribute understanding. In addition, BiAug also improves the performance on zero-shot retrieval tasks on general benchmarks like MSCOCO and Flickr30K. BiAug refines the way of collecting text-image datasets. Mitigating the reporting bias helps models achieve a deeper understanding of visual-language phenomena, expanding beyond mere frequent patterns to encompass the richness and diversity of real-world scenarios."
                    }
                ]
            },
            "3e98e198-7b40-49b1-b33f-51e9b01ff1b5": {
                "pk": "3e98e198-7b40-49b1-b33f-51e9b01ff1b5",
                "name": "Naoki Murata",
                "collaborators": [
                    "Yuki Mitsufuji",
                    "Chieh-Hsin Lai",
                    "Yuhta Takida",
                    "Toshimitsu Uesaka",
                    "Wei-Hsiang Liao",
                    "Stefano Ermon",
                    "Takashi Shibuya",
                    "Dongjun Kim",
                    "Yukara Ikemiya",
                    "Yixiao Zhang"
                ],
                "domain": [
                    "Generative Models",
                    "Diffusion Models",
                    "Contrastive Learning",
                    "Music Generation"
                ],
                "publications": [
                    {
                        "title": "Weighted Point Cloud Embedding for Multimodal Contrastive Learning Toward Optimal Similarity Metric",
                        "abstract": "In typical multimodal contrastive learning, such as CLIP, encoders produce one point in the latent representation space for each input. However, one-point representation has difficulty in capturing the relationship and the similarity structure of a huge amount of instances in the real world. For richer classes of the similarity, we propose the use of weighted point clouds, namely, sets of pairs of weight and vector, as representations of instances. In this work, we theoretically show the benefit of our proposed method through a new understanding of the contrastive loss of CLIP, which we call symmetric InfoNCE. We clarify that the optimal similarity that minimizes symmetric InfoNCE is the pointwise mutual information, and show an upper bound of excess risk on downstream classification tasks of representations that achieve the optimal similarity. In addition, we show that our proposed similarity based on weighted point clouds consistently achieves the optimal similarity. To verify the effectiveness of our proposed method, we demonstrate pretraining of text-image representation models and classification tasks on common benchmarks."
                    },
                    {
                        "title": "PaGoDA: Progressive Growing of a One-Step Generator from a Low-Resolution Diffusion Teacher",
                        "abstract": "The diffusion model performs remarkable in generating high-dimensional content but is computationally intensive, especially during training. We propose Progressive Growing of Diffusion Autoencoder (PaGoDA), a novel pipeline that reduces the training costs through three stages: training diffusion on downsampled data, distilling the pretrained diffusion, and progressive super-resolution. With the proposed pipeline, PaGoDA achieves a $64\\times$ reduced cost in training its diffusion model on 8x downsampled data; while at the inference, with the single-step, it performs state-of-the-art on ImageNet across all resolutions from 64x64 to 512x512, and text-to-image. PaGoDA's pipeline can be applied directly in the latent space, adding compression alongside the pre-trained autoencoder in Latent Diffusion Models (e.g., Stable Diffusion). The code is available at https://github.com/sony/pagoda."
                    },
                    {
                        "title": "Instruct-MusicGen: Unlocking Text-to-Music Editing for Music Language Models via Instruction Tuning",
                        "abstract": "Recent advances in text-to-music editing, which employ text queries to modify music (e.g.\\ by changing its style or adjusting instrumental components), present unique challenges and opportunities for AI-assisted music creation. Previous approaches in this domain have been constrained by the necessity to train specific editing models from scratch, which is both resource-intensive and inefficient; other research uses large language models to predict edited music, resulting in imprecise audio reconstruction. To Combine the strengths and address these limitations, we introduce Instruct-MusicGen, a novel approach that finetunes a pretrained MusicGen model to efficiently follow editing instructions such as adding, removing, or separating stems. Our approach involves a modification of the original MusicGen architecture by incorporating a text fusion module and an audio fusion module, which allow the model to process instruction texts and audio inputs concurrently and yield the desired edited music. Remarkably, Instruct-MusicGen only introduces 8% new parameters to the original MusicGen model and only trains for 5K steps, yet it achieves superior performance across all tasks compared to existing baselines, and demonstrates performance comparable to the models trained for specific tasks. This advancement not only enhances the efficiency of text-to-music editing but also broadens the applicability of music language models in dynamic music production environments."
                    },
                    {
                        "title": "Mitigating Embedding Collapse in Diffusion Models for Categorical Data",
                        "abstract": "Latent diffusion models have enabled continuous-state diffusion models to handle a variety of datasets, including categorical data. However, most methods rely on fixed pretrained embeddings, limiting the benefits of joint training with the diffusion model. While jointly learning the embedding (via reconstruction loss) and the latent diffusion model (via score matching loss) could enhance performance, our analysis shows that end-to-end training risks embedding collapse, degrading generation quality. To address this issue, we introduce CATDM, a continuous diffusion framework within the embedding space that stabilizes training. We propose a novel objective combining the joint embedding-diffusion variational lower bound with a Consistency-Matching (CM) regularizer, alongside a shifted cosine noise schedule and random dropping strategy. The CM regularizer ensures the recovery of the true data distribution. Experiments on benchmarks show that CATDM mitigates embedding collapse, yielding superior results on FFHQ, LSUN Churches, and LSUN Bedrooms. In particular, CATDM achieves an FID of 6.81 on ImageNet $256\\times256$ with 50 steps. It outperforms non-autoregressive models in machine translation and is on a par with previous methods in text generation."
                    },
                    {
                        "title": "Automated Black-box Prompt Engineering for Personalized Text-to-Image Generation",
                        "abstract": "Prompt engineering is effective for controlling the output of text-to-image (T2I) generative models, but it is also laborious due to the need for manually crafted prompts. This challenge has spurred the development of algorithms for automated prompt generation. However, these methods often struggle with transferability across T2I models, require white-box access to the underlying model, and produce non-intuitive prompts. In this work, we introduce PRISM, an algorithm that automatically identifies human-interpretable and transferable prompts that can effectively generate desired concepts given only black-box access to T2I models. Inspired by large language model (LLM) jailbreaking, PRISM leverages the in-context learning ability of LLMs to iteratively refine the candidate prompts distribution for given reference images. Our experiments demonstrate the versatility and effectiveness of PRISM in generating accurate prompts for objects, styles and images across multiple T2I models, including Stable Diffusion, DALL-E, and Midjourney."
                    },
                    {
                        "title": "MoLA: Motion Generation and Editing with Latent Diffusion Enhanced by Adversarial Training",
                        "abstract": "In motion generation, controllability as well as generation quality and speed is becoming more and more important. There are various motion editing tasks, such as in-betweening, upper body editing, and path-following, but existing methods perform motion editing with a data-space diffusion model, which is slow in inference compared to a latent diffusion model. In this paper, we propose MoLA, which provides fast and high-quality motion generation and also can deal with multiple editing tasks in a single framework. For high-quality and fast generation, we employ a variational autoencoder and latent diffusion model, and improve the performance with adversarial training. In addition, we apply a training-free guided generation framework to achieve various editing tasks with motion control inputs. We quantitatively show the effectiveness of adversarial learning in text-to-motion generation, and demonstrate the applicability of our editing framework to multiple editing tasks in the motion domain."
                    },
                    {
                        "title": "GenWarp: Single Image to Novel Views with Semantic-Preserving Generative Warping",
                        "abstract": "Generating novel views from a single image remains a challenging task due to the complexity of 3D scenes and the limited diversity in the existing multi-view datasets to train a model on. Recent research combining large-scale text-to-image (T2I) models with monocular depth estimation (MDE) has shown promise in handling in-the-wild images. In these methods, an input view is geometrically warped to novel views with estimated depth maps, then the warped image is inpainted by T2I models. However, they struggle with noisy depth maps and loss of semantic details when warping an input view to novel viewpoints. In this paper, we propose a novel approach for single-shot novel view synthesis, a semantic-preserving generative warping framework that enables T2I generative models to learn where to warp and where to generate, through augmenting cross-view attention with self-attention. Our approach addresses the limitations of existing methods by conditioning the generative model on source view images and incorporating geometric warping signals. Qualitative and quantitative evaluations demonstrate that our model outperforms existing methods in both in-domain and out-of-domain scenarios. Project page is available at https://GenWarp-NVS.github.io/."
                    },
                    {
                        "title": "Human-Feedback Efficient Reinforcement Learning for Online Diffusion Model Finetuning",
                        "abstract": "Controllable generation through Stable Diffusion (SD) fine-tuning aims to improve fidelity, safety, and alignment with human guidance. Existing reinforcement learning from human feedback methods usually rely on predefined heuristic reward functions or pretrained reward models built on large-scale datasets, limiting their applicability to scenarios where collecting such data is costly or difficult. To effectively and efficiently utilize human feedback, we develop a framework, HERO, which leverages online human feedback collected on the fly during model learning. Specifically, HERO features two key mechanisms: (1) Feedback-Aligned Representation Learning, an online training method that captures human feedback and provides informative learning signals for fine-tuning, and (2) Feedback-Guided Image Generation, which involves generating images from SD's refined initialization samples, enabling faster convergence towards the evaluator's intent. We demonstrate that HERO is 4x more efficient in online feedback for body part anomaly correction compared to the best existing method. Additionally, experiments show that HERO can effectively handle tasks like reasoning, counting, personalization, and reducing NSFW content with only 0.5K online feedback."
                    },
                    {
                        "title": "MusicMagus: Zero-Shot Text-to-Music Editing via Diffusion Models",
                        "abstract": "Recent advances in text-to-music generation models have opened new avenues in musical creativity. However, the task of editing these generated music remains a significant challenge. This paper introduces a novel approach to edit music generated by such models, enabling the modification of specific attributes, such as genre, mood, and instrument, while maintaining other aspects unchanged. Our method transforms text editing to the latent space manipulation, and adds an additional constraint to enforce consistency. It seamlessly integrates with existing pretrained text-to-music diffusion models without requiring additional training. Experimental results demonstrate superior performance over both zero-shot and certain supervised baselines in style and timbre transfer evaluations. We also show the practical applicability of our approach in real-world music editing scenarios."
                    },
                    {
                        "title": "G2D2: Gradient-guided Discrete Diffusion for image inverse problem solving",
                        "abstract": "Recent literature has effectively utilized diffusion models trained on continuous variables as priors for solving inverse problems. Notably, discrete diffusion models with discrete latent codes have shown strong performance, particularly in modalities suited for discrete compressed representations, such as image and motion generation. However, their discrete and non-differentiable nature has limited their application to inverse problems formulated in continuous spaces. This paper presents a novel method for addressing linear inverse problems by leveraging image-generation models based on discrete diffusion as priors. We overcome these limitations by approximating the true posterior distribution with a variational distribution constructed from categorical distributions and continuous relaxation techniques. Furthermore, we employ a star-shaped noise process to mitigate the drawbacks of traditional discrete diffusion models with absorbing states, demonstrating that our method performs comparably to continuous diffusion techniques. To the best of our knowledge, this is the first approach to use discrete diffusion model-based priors for solving image inverse problems."
                    },
                    {
                        "title": "SAN: Inducing Metrizability of GAN with Discriminative Normalized Linear Layer",
                        "abstract": "Generative adversarial networks (GANs) learn a target probability distribution by optimizing a generator and a discriminator with minimax objectives. This paper addresses the question of whether such optimization actually provides the generator with gradients that make its distribution close to the target distribution. We derive metrizable conditions, sufficient conditions for the discriminator to serve as the distance between the distributions by connecting the GAN formulation with the concept of sliced optimal transport. Furthermore, by leveraging these theoretical results, we propose a novel GAN training scheme, called slicing adversarial network (SAN). With only simple modifications, a broad class of existing GANs can be converted to SANs. Experiments on synthetic and image datasets support our theoretical results and the SAN's effectiveness as compared to usual GANs. Furthermore, we also apply SAN to StyleGAN-XL, which leads to state-of-the-art FID score amongst GANs for class conditional generation on ImageNet 256$\\times$256. Our implementation is available on https://ytakida.github.io/san."
                    },
                    {
                        "title": "Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion",
                        "abstract": "Consistency Models (CM) (Song et al., 2023) accelerate score-based diffusion model sampling at the cost of sample quality but lack a natural way to trade-off quality for speed. To address this limitation, we propose Consistency Trajectory Model (CTM), a generalization encompassing CM and score-based models as special cases. CTM trains a single neural network that can -- in a single forward pass -- output scores (i.e., gradients of log-density) and enables unrestricted traversal between any initial and final time along the Probability Flow Ordinary Differential Equation (ODE) in a diffusion process. CTM enables the efficient combination of adversarial training and denoising score matching loss to enhance performance and achieves new state-of-the-art FIDs for single-step diffusion model sampling on CIFAR-10 (FID 1.73) and ImageNet at 64x64 resolution (FID 1.92). CTM also enables a new family of sampling schemes, both deterministic and stochastic, involving long jumps along the ODE solution trajectories. It consistently improves sample quality as computational budgets increase, avoiding the degradation seen in CM. Furthermore, unlike CM, CTM's access to the score function can streamline the adoption of established controllable/conditional generation methods from the diffusion community. This access also enables the computation of likelihood. The code is available at https://github.com/sony/ctm."
                    },
                    {
                        "title": "HQ-VAE: Hierarchical Discrete Representation Learning with Variational Bayes",
                        "abstract": "Vector quantization (VQ) is a technique to deterministically learn features with discrete codebook representations. It is commonly performed with a variational autoencoding model, VQ-VAE, which can be further extended to hierarchical structures for making high-fidelity reconstructions. However, such hierarchical extensions of VQ-VAE often suffer from the codebook/layer collapse issue, where the codebook is not efficiently used to express the data, and hence degrades reconstruction accuracy. To mitigate this problem, we propose a novel unified framework to stochastically learn hierarchical discrete representation on the basis of the variational Bayes framework, called hierarchically quantized variational autoencoder (HQ-VAE). HQ-VAE naturally generalizes the hierarchical variants of VQ-VAE, such as VQ-VAE-2 and residual-quantized VAE (RQ-VAE), and provides them with a Bayesian training scheme. Our comprehensive experiments on image datasets show that HQ-VAE enhances codebook usage and improves reconstruction performance. We also validated HQ-VAE in terms of its applicability to a different modality with an audio dataset."
                    },
                    {
                        "title": "On the Equivalence of Consistency-Type Models: Consistency Models, Consistent Diffusion Models, and Fokker-Planck Regularization",
                        "abstract": "The emergence of various notions of ``consistency'' in diffusion models has garnered considerable attention and helped achieve improved sample quality, likelihood estimation, and accelerated sampling. Although similar concepts have been proposed in the literature, the precise relationships among them remain unclear. In this study, we establish theoretical connections between three recent ``consistency'' notions designed to enhance diffusion models for distinct objectives. Our insights offer the potential for a more comprehensive and encompassing framework for consistency-type models."
                    },
                    {
                        "title": "Adversarially Slicing Generative Networks: Discriminator Slices Feature for One-Dimensional Optimal Transport",
                        "abstract": "Generative adversarial networks (GANs) learn a target probability distribution by optimizing a generator and a discriminator with minimax objectives. This paper addresses the question of whether such optimization actually provides the generator with gradients that make its distribution close to the target distribution. We derive suf\ufb01cient conditions for the discriminator to serve as the distance between the distributions by connecting the GAN formulation with the concept of sliced optimal transport. Furthermore, by leveraging these theoretical results, we propose a novel GAN training scheme, called adversarially slicing generative network (ASGN). With only simple modi\ufb01cations, the ASGN is applicable to a broad class of existing GANs. Experiments on synthetic and image datasets support our theoretical results and the ASGN\u2019s effectiveness as compared to usual GANs."
                    },
                    {
                        "title": "Improving Score-based Diffusion Models by Enforcing the Underlying Score Fokker-Planck Equation",
                        "abstract": "Score-based generative models learn a family of noise-conditional score functions corresponding to the data density perturbed with increasingly large amounts of noise. These perturbed data densities are tied together by the Fokker-Planck equation (FPE), a partial differential equation (PDE) governing the spatial-temporal evolution of a density undergoing a diffusion process. In this work, we derive a corresponding equation, called the score FPE that characterizes the noise-conditional scores of the perturbed data densities (i.e., their gradients). Surprisingly, despite impressive empirical performance, we observe that scores learned via denoising score matching (DSM) do not satisfy the underlying score FPE. We prove that satisfying the FPE is desirable as it improves the likelihood and the degree of conservativity. Hence, we propose to regularize the DSM objective to enforce satisfaction of the score FPE, and we show the effectiveness of this approach across various datasets."
                    },
                    {
                        "title": "VRDMG: Vocal Restoration via Diffusion Posterior Sampling with Multiple Guidance",
                        "abstract": "Restoring degraded music signals is essential to enhance audio quality for downstream music manipulation. Recent diffusion-based music restoration methods have demonstrated impressive performance, and among them, diffusion posterior sampling (DPS) stands out given its intrinsic properties, making it versatile across various restoration tasks. In this paper, we identify that there are potential issues which will degrade current DPS-based methods\u2019 performance and introduce the way to mitigate the issues inspired by diverse diffusion guidance techniques including the RePaint (RP) strategy and the Pseudoinverse-Guided Diffusion Models (\u03a0GDM). We demonstrate our methods for the vocal declipping and bandwidth extension tasks under various levels of distortion and cutoff frequency, respectively. In both tasks, our methods outperform the current DPS-based music restoration benchmarks. We refer to http://carlosholivan.github.io/demos/audio-restoration-2023.html for examples of the restored audio samples."
                    },
                    {
                        "title": "On the Language Encoder of Contrastive Cross-modal Models",
                        "abstract": "Contrastive cross-modal models such as CLIP and CLAP aid various vision-language (VL) and audio-language (AL) tasks. However, there has been limited investigation of and improvement in their language encoder, which is the central component of encoding natural language descriptions of image/audio into vector representations. We extensively evaluate how unsupervised and supervised sentence embedding training affect language encoder quality and cross-modal task performance. In VL pretraining, we found that sentence embedding training language encoder quality and aids in cross-modal tasks, improving contrastive VL models such as CyCLIP. In contrast, AL pretraining benefits less from sentence embedding training, which may result from the limited amount of pretraining data. We analyze the representation spaces to understand the strengths of sentence embedding training, and find that it improves text-space uniformity, at the cost of decreased cross-modal alignment."
                    },
                    {
                        "title": "GibbsDDRM: A Partially Collapsed Gibbs Sampler for Solving Blind Inverse Problems with Denoising Diffusion Restoration",
                        "abstract": "Pre-trained diffusion models have been successfully used as priors in a variety of linear inverse problems, where the goal is to reconstruct a signal from noisy linear measurements. However, existing approaches require knowledge of the linear operator. In this paper, we propose GibbsDDRM, an extension of Denoising Diffusion Restoration Models (DDRM) to a blind setting in which the linear measurement operator is unknown. GibbsDDRM constructs a joint distribution of the data, measurements, and linear operator by using a pre-trained diffusion model for the data prior, and it solves the problem by posterior sampling with an efficient variant of a Gibbs sampler. The proposed method is problem-agnostic, meaning that a pre-trained diffusion model can be applied to various inverse problems without fine-tuning. In experiments, it achieved high performance on both blind image deblurring and vocal dereverberation tasks, despite the use of simple generic priors for the underlying linear operators."
                    }
                ]
            },
            "accf92e0-f0b3-44a4-a840-fbfa2ab8a9e6": {
                "pk": "accf92e0-f0b3-44a4-a840-fbfa2ab8a9e6",
                "name": "Chieh-Hsin Lai",
                "collaborators": [
                    "Yuki Mitsufuji",
                    "Yuhta Takida",
                    "Naoki Murata",
                    "Toshimitsu Uesaka",
                    "Wei-Hsiang Liao",
                    "Stefano Ermon",
                    "Takashi Shibuya",
                    "Dongjun Kim",
                    "Woosung Choi",
                    "Koichi Saito"
                ],
                "domain": [
                    "Generative Models",
                    "Diffusion Models",
                    "Music Processing",
                    "Contrastive Learning"
                ],
                "publications": [
                    {
                        "title": "Weighted Point Cloud Embedding for Multimodal Contrastive Learning Toward Optimal Similarity Metric",
                        "abstract": "In typical multimodal contrastive learning, such as CLIP, encoders produce one point in the latent representation space for each input. However, one-point representation has difficulty in capturing the relationship and the similarity structure of a huge amount of instances in the real world. For richer classes of the similarity, we propose the use of weighted point clouds, namely, sets of pairs of weight and vector, as representations of instances. In this work, we theoretically show the benefit of our proposed method through a new understanding of the contrastive loss of CLIP, which we call symmetric InfoNCE. We clarify that the optimal similarity that minimizes symmetric InfoNCE is the pointwise mutual information, and show an upper bound of excess risk on downstream classification tasks of representations that achieve the optimal similarity. In addition, we show that our proposed similarity based on weighted point clouds consistently achieves the optimal similarity. To verify the effectiveness of our proposed method, we demonstrate pretraining of text-image representation models and classification tasks on common benchmarks."
                    },
                    {
                        "title": "DisMix: Disentangling Mixtures of Musical Instruments for Source-level Pitch and Timbre Manipulation",
                        "abstract": "Existing work on pitch and timbre disentanglement has been mostly focused on single-instrument music audio, excluding the cases where multiple instruments are presented. To fill the gap, we propose DisMix, a generative framework in which the pitch and timbre representations act as modular building blocks for constructing the melody and instrument of a source, and the collection of which forms a set of per-instrument latent representations underlying the observed mixture. By manipulating the representations, our model samples mixtures with novel combinations of pitch and timbre of the constituent instruments. We can jointly learn the disentangled pitch-timbre representations and a latent diffusion transformer that reconstructs the mixture conditioned on the set of source-level representations. We evaluate the model using both a simple dataset of isolated chords and a realistic four-part chorales in the style of J.S. Bach, identify the key components for the success of disentanglement, and demonstrate the application of mixture transformation based on source-level attribute manipulation."
                    },
                    {
                        "title": "PaGoDA: Progressive Growing of a One-Step Generator from a Low-Resolution Diffusion Teacher",
                        "abstract": "The diffusion model performs remarkable in generating high-dimensional content but is computationally intensive, especially during training. We propose Progressive Growing of Diffusion Autoencoder (PaGoDA), a novel pipeline that reduces the training costs through three stages: training diffusion on downsampled data, distilling the pretrained diffusion, and progressive super-resolution. With the proposed pipeline, PaGoDA achieves a $64\\times$ reduced cost in training its diffusion model on 8x downsampled data; while at the inference, with the single-step, it performs state-of-the-art on ImageNet across all resolutions from 64x64 to 512x512, and text-to-image. PaGoDA's pipeline can be applied directly in the latent space, adding compression alongside the pre-trained autoencoder in Latent Diffusion Models (e.g., Stable Diffusion). The code is available at https://github.com/sony/pagoda."
                    },
                    {
                        "title": "$\\textit{Jump Your Steps}$: Optimizing Sampling Schedule of Discrete Diffusion Models",
                        "abstract": "Diffusion models have seen notable success in continuous domains, leading to the development of discrete diffusion models (DDMs) for discrete variables. Despite recent advances, DDMs face the challenge of slow sampling speeds. While parallel sampling methods like $\\tau$-leaping accelerate this process, they introduce $\\textit{Compounding Decoding Error}$ (CDE), where discrepancies arise between the true distribution and the approximation from parallel token generation, leading to degraded sample quality. In this work, we present $\\textit{Jump Your Steps}$ (JYS), a novel approach that optimizes the allocation of discrete sampling timesteps by minimizing CDE without extra computational cost. More precisely, we derive a practical upper bound on CDE and propose an efficient algorithm for searching for the optimal sampling schedule. Extensive experiments across image, music, and text generation show that JYS significantly improves sampling quality, establishing it as a versatile framework for enhancing DDM performance for fast sampling."
                    },
                    {
                        "title": "Latent Diffusion Bridges for Unsupervised Musical Audio Timbre Transfer",
                        "abstract": "Music timbre transfer is a challenging task that involves modifying the timbral characteristics of an audio signal while preserving its melodic structure. In this paper, we propose a novel method based on dual diffusion bridges, trained using the CocoChorales Dataset, which consists of unpaired monophonic single-instrument audio data. Each diffusion model is trained on a specific instrument with a Gaussian prior. During inference, a model is designated as the source model to map the input audio to its corresponding Gaussian prior, and another model is designated as the target model to reconstruct the target audio from this Gaussian prior, thereby facilitating timbre transfer. We compare our approach against existing unsupervised timbre transfer models such as VAEGAN and Gaussian Flow Bridges (GFB). Experimental results demonstrate that our method achieves both better Fr\\'echet Audio Distance (FAD) and melody preservation, as reflected by lower pitch distances (DPD) compared to VAEGAN and GFB. Additionally, we discover that the noise level from the Gaussian prior, $\\sigma$, can be adjusted to control the degree of melody preservation and amount of timbre transferred."
                    },
                    {
                        "title": "SoundCTM: Uniting Score-based and Consistency Models for Text-to-Sound Generation",
                        "abstract": "Sound content is an indispensable element for multimedia works such as video games, music, and films. Recent high-quality diffusion-based sound generation models can serve as valuable tools for the creators. However, despite producing high-quality sounds, these models often suffer from slow inference speeds. This drawback burdens creators, who typically refine their sounds through trial and error to align them with their artistic intentions. To address this issue, we introduce Sound Consistency Trajectory Models (SoundCTM). Our model enables flexible transitioning between high-quality 1-step sound generation and superior sound quality through multi-step generation. This allows creators to initially control sounds with 1-step samples before refining them through multi-step generation. While CTM fundamentally achieves flexible 1-step and multi-step generation, its impressive performance heavily depends on an additional pretrained feature extractor and an adversarial loss, which are expensive to train and not always available in other domains. Thus, we reframe CTM's training framework and introduce a novel feature distance by utilizing the teacher's network for a distillation loss. Additionally, while distilling classifier-free guided trajectories, we train conditional and unconditional student models simultaneously and interpolate between these models during inference. We also propose training-free controllable frameworks for SoundCTM, leveraging its flexible sampling capability. SoundCTM achieves both promising 1-step and multi-step real-time sound generation without using any extra off-the-shelf networks. Furthermore, we demonstrate SoundCTM's capability of controllable sound generation in a training-free manner. Our codes, pretrained models, and audio samples are available at https://github.com/sony/soundctm."
                    },
                    {
                        "title": "Bellman Diffusion: Generative Modeling as Learning a Linear Operator in the Distribution Space",
                        "abstract": "Deep Generative Models (DGMs), including Energy-Based Models (EBMs) and Score-based Generative Models (SGMs), have advanced high-fidelity data generation and complex continuous distribution approximation. However, their application in Markov Decision Processes (MDPs), particularly in distributional Reinforcement Learning (RL), remains underexplored, with conventional histogram-based methods dominating the field. This paper rigorously highlights that this application gap is caused by the nonlinearity of modern DGMs, which conflicts with the linearity required by the Bellman equation in MDPs. For instance, EBMs involve nonlinear operations such as exponentiating energy functions and normalizing constants. To address this, we introduce Bellman Diffusion, a novel DGM framework that maintains linearity in MDPs through gradient and scalar field modeling. With divergence-based training techniques to optimize neural network proxies and a new type of stochastic differential equation (SDE) for sampling, Bellman Diffusion is guaranteed to converge to the target distribution. Our empirical results show that Bellman Diffusion achieves accurate field estimations and is a capable image generator, converging 1.5x faster than the traditional histogram-based baseline in distributional RL tasks. This work enables the effective integration of DGMs into MDP applications, unlocking new avenues for advanced decision-making frameworks."
                    },
                    {
                        "title": "A Survey on Diffusion Models for Inverse Problems",
                        "abstract": "Diffusion models have become increasingly popular for generative modeling due to their ability to generate high-quality samples. This has unlocked exciting new possibilities for solving inverse problems, especially in image restoration and reconstruction, by treating diffusion models as unsupervised priors. This survey provides a comprehensive overview of methods that utilize pre-trained diffusion models to solve inverse problems without requiring further training. We introduce taxonomies to categorize these methods based on both the problems they address and the techniques they employ. We analyze the connections between different approaches, offering insights into their practical implementation and highlighting important considerations. We further discuss specific challenges and potential solutions associated with using latent diffusion models for inverse problems. This work aims to be a valuable resource for those interested in learning about the intersection of diffusion models and inverse problems."
                    },
                    {
                        "title": "GenWarp: Single Image to Novel Views with Semantic-Preserving Generative Warping",
                        "abstract": "Generating novel views from a single image remains a challenging task due to the complexity of 3D scenes and the limited diversity in the existing multi-view datasets to train a model on. Recent research combining large-scale text-to-image (T2I) models with monocular depth estimation (MDE) has shown promise in handling in-the-wild images. In these methods, an input view is geometrically warped to novel views with estimated depth maps, then the warped image is inpainted by T2I models. However, they struggle with noisy depth maps and loss of semantic details when warping an input view to novel viewpoints. In this paper, we propose a novel approach for single-shot novel view synthesis, a semantic-preserving generative warping framework that enables T2I generative models to learn where to warp and where to generate, through augmenting cross-view attention with self-attention. Our approach addresses the limitations of existing methods by conditioning the generative model on source view images and incorporating geometric warping signals. Qualitative and quantitative evaluations demonstrate that our model outperforms existing methods in both in-domain and out-of-domain scenarios. Project page is available at https://GenWarp-NVS.github.io/."
                    },
                    {
                        "title": "Human-Feedback Efficient Reinforcement Learning for Online Diffusion Model Finetuning",
                        "abstract": "Controllable generation through Stable Diffusion (SD) fine-tuning aims to improve fidelity, safety, and alignment with human guidance. Existing reinforcement learning from human feedback methods usually rely on predefined heuristic reward functions or pretrained reward models built on large-scale datasets, limiting their applicability to scenarios where collecting such data is costly or difficult. To effectively and efficiently utilize human feedback, we develop a framework, HERO, which leverages online human feedback collected on the fly during model learning. Specifically, HERO features two key mechanisms: (1) Feedback-Aligned Representation Learning, an online training method that captures human feedback and provides informative learning signals for fine-tuning, and (2) Feedback-Guided Image Generation, which involves generating images from SD's refined initialization samples, enabling faster convergence towards the evaluator's intent. We demonstrate that HERO is 4x more efficient in online feedback for body part anomaly correction compared to the best existing method. Additionally, experiments show that HERO can effectively handle tasks like reasoning, counting, personalization, and reducing NSFW content with only 0.5K online feedback."
                    },
                    {
                        "title": "G2D2: Gradient-guided Discrete Diffusion for image inverse problem solving",
                        "abstract": "Recent literature has effectively utilized diffusion models trained on continuous variables as priors for solving inverse problems. Notably, discrete diffusion models with discrete latent codes have shown strong performance, particularly in modalities suited for discrete compressed representations, such as image and motion generation. However, their discrete and non-differentiable nature has limited their application to inverse problems formulated in continuous spaces. This paper presents a novel method for addressing linear inverse problems by leveraging image-generation models based on discrete diffusion as priors. We overcome these limitations by approximating the true posterior distribution with a variational distribution constructed from categorical distributions and continuous relaxation techniques. Furthermore, we employ a star-shaped noise process to mitigate the drawbacks of traditional discrete diffusion models with absorbing states, demonstrating that our method performs comparably to continuous diffusion techniques. To the best of our knowledge, this is the first approach to use discrete diffusion model-based priors for solving image inverse problems."
                    },
                    {
                        "title": "The Sound Demixing Challenge 2023 - Music Demixing Track",
                        "abstract": "This paper summarizes the music demixing (MDX) track of the Sound Demixing Challenge (SDX'23). We provide a summary of the challenge setup and introduce the task of robust music source separation (MSS), i.e., training MSS models in the presence of errors in the training data. We propose a formalization of the errors that can occur in the design of a training dataset for MSS systems and introduce two new datasets that simulate such errors: SDXDB23_LabelNoise and SDXDB23_Bleeding. We describe the methods that achieved the highest scores in the competition. Moreover, we present a direct comparison with the previous edition of the challenge (the Music Demixing Challenge 2021): the best performing system achieved an improvement of over 1.6dB in signal-to-distortion ratio over the winner of the previous competition, when evaluated on MDXDB21. Besides relying on the signal-to-distortion ratio as objective metric, we also performed a listening test with renowned producers and musicians to study the perceptual quality of the systems and report here the results. Finally, we provide our insights into the organization of the competition and our prospects for future editions."
                    },
                    {
                        "title": "SAN: Inducing Metrizability of GAN with Discriminative Normalized Linear Layer",
                        "abstract": "Generative adversarial networks (GANs) learn a target probability distribution by optimizing a generator and a discriminator with minimax objectives. This paper addresses the question of whether such optimization actually provides the generator with gradients that make its distribution close to the target distribution. We derive metrizable conditions, sufficient conditions for the discriminator to serve as the distance between the distributions by connecting the GAN formulation with the concept of sliced optimal transport. Furthermore, by leveraging these theoretical results, we propose a novel GAN training scheme, called slicing adversarial network (SAN). With only simple modifications, a broad class of existing GANs can be converted to SANs. Experiments on synthetic and image datasets support our theoretical results and the SAN's effectiveness as compared to usual GANs. Furthermore, we also apply SAN to StyleGAN-XL, which leads to state-of-the-art FID score amongst GANs for class conditional generation on ImageNet 256$\\times$256. Our implementation is available on https://ytakida.github.io/san."
                    },
                    {
                        "title": "Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion",
                        "abstract": "Consistency Models (CM) (Song et al., 2023) accelerate score-based diffusion model sampling at the cost of sample quality but lack a natural way to trade-off quality for speed. To address this limitation, we propose Consistency Trajectory Model (CTM), a generalization encompassing CM and score-based models as special cases. CTM trains a single neural network that can -- in a single forward pass -- output scores (i.e., gradients of log-density) and enables unrestricted traversal between any initial and final time along the Probability Flow Ordinary Differential Equation (ODE) in a diffusion process. CTM enables the efficient combination of adversarial training and denoising score matching loss to enhance performance and achieves new state-of-the-art FIDs for single-step diffusion model sampling on CIFAR-10 (FID 1.73) and ImageNet at 64x64 resolution (FID 1.92). CTM also enables a new family of sampling schemes, both deterministic and stochastic, involving long jumps along the ODE solution trajectories. It consistently improves sample quality as computational budgets increase, avoiding the degradation seen in CM. Furthermore, unlike CM, CTM's access to the score function can streamline the adoption of established controllable/conditional generation methods from the diffusion community. This access also enables the computation of likelihood. The code is available at https://github.com/sony/ctm."
                    },
                    {
                        "title": "HQ-VAE: Hierarchical Discrete Representation Learning with Variational Bayes",
                        "abstract": "Vector quantization (VQ) is a technique to deterministically learn features with discrete codebook representations. It is commonly performed with a variational autoencoding model, VQ-VAE, which can be further extended to hierarchical structures for making high-fidelity reconstructions. However, such hierarchical extensions of VQ-VAE often suffer from the codebook/layer collapse issue, where the codebook is not efficiently used to express the data, and hence degrades reconstruction accuracy. To mitigate this problem, we propose a novel unified framework to stochastically learn hierarchical discrete representation on the basis of the variational Bayes framework, called hierarchically quantized variational autoencoder (HQ-VAE). HQ-VAE naturally generalizes the hierarchical variants of VQ-VAE, such as VQ-VAE-2 and residual-quantized VAE (RQ-VAE), and provides them with a Bayesian training scheme. Our comprehensive experiments on image datasets show that HQ-VAE enhances codebook usage and improves reconstruction performance. We also validated HQ-VAE in terms of its applicability to a different modality with an audio dataset."
                    },
                    {
                        "title": "On the Equivalence of Consistency-Type Models: Consistency Models, Consistent Diffusion Models, and Fokker-Planck Regularization",
                        "abstract": "The emergence of various notions of ``consistency'' in diffusion models has garnered considerable attention and helped achieve improved sample quality, likelihood estimation, and accelerated sampling. Although similar concepts have been proposed in the literature, the precise relationships among them remain unclear. In this study, we establish theoretical connections between three recent ``consistency'' notions designed to enhance diffusion models for distinct objectives. Our insights offer the potential for a more comprehensive and encompassing framework for consistency-type models."
                    },
                    {
                        "title": "Adversarially Slicing Generative Networks: Discriminator Slices Feature for One-Dimensional Optimal Transport",
                        "abstract": "Generative adversarial networks (GANs) learn a target probability distribution by optimizing a generator and a discriminator with minimax objectives. This paper addresses the question of whether such optimization actually provides the generator with gradients that make its distribution close to the target distribution. We derive suf\ufb01cient conditions for the discriminator to serve as the distance between the distributions by connecting the GAN formulation with the concept of sliced optimal transport. Furthermore, by leveraging these theoretical results, we propose a novel GAN training scheme, called adversarially slicing generative network (ASGN). With only simple modi\ufb01cations, the ASGN is applicable to a broad class of existing GANs. Experiments on synthetic and image datasets support our theoretical results and the ASGN\u2019s effectiveness as compared to usual GANs."
                    },
                    {
                        "title": "Improving Score-based Diffusion Models by Enforcing the Underlying Score Fokker-Planck Equation",
                        "abstract": "Score-based generative models learn a family of noise-conditional score functions corresponding to the data density perturbed with increasingly large amounts of noise. These perturbed data densities are tied together by the Fokker-Planck equation (FPE), a partial differential equation (PDE) governing the spatial-temporal evolution of a density undergoing a diffusion process. In this work, we derive a corresponding equation, called the score FPE that characterizes the noise-conditional scores of the perturbed data densities (i.e., their gradients). Surprisingly, despite impressive empirical performance, we observe that scores learned via denoising score matching (DSM) do not satisfy the underlying score FPE. We prove that satisfying the FPE is desirable as it improves the likelihood and the degree of conservativity. Hence, we propose to regularize the DSM objective to enforce satisfaction of the score FPE, and we show the effectiveness of this approach across various datasets."
                    },
                    {
                        "title": "VRDMG: Vocal Restoration via Diffusion Posterior Sampling with Multiple Guidance",
                        "abstract": "Restoring degraded music signals is essential to enhance audio quality for downstream music manipulation. Recent diffusion-based music restoration methods have demonstrated impressive performance, and among them, diffusion posterior sampling (DPS) stands out given its intrinsic properties, making it versatile across various restoration tasks. In this paper, we identify that there are potential issues which will degrade current DPS-based methods\u2019 performance and introduce the way to mitigate the issues inspired by diverse diffusion guidance techniques including the RePaint (RP) strategy and the Pseudoinverse-Guided Diffusion Models (\u03a0GDM). We demonstrate our methods for the vocal declipping and bandwidth extension tasks under various levels of distortion and cutoff frequency, respectively. In both tasks, our methods outperform the current DPS-based music restoration benchmarks. We refer to http://carlosholivan.github.io/demos/audio-restoration-2023.html for examples of the restored audio samples."
                    }
                ]
            },
            "ad752d90-27be-4927-87bb-829b32f6d096": {
                "pk": "ad752d90-27be-4927-87bb-829b32f6d096",
                "name": "Yuhta Takida",
                "collaborators": [
                    "Yuki Mitsufuji",
                    "Chieh-Hsin Lai",
                    "Naoki Murata",
                    "Toshimitsu Uesaka",
                    "Stefano Ermon",
                    "Wei-Hsiang Liao",
                    "Takashi Shibuya",
                    "Woosung Choi",
                    "Koichi Saito",
                    "Dongjun Kim"
                ],
                "domain": [
                    "Generative Models",
                    "Diffusion Models",
                    "Contrastive Learning",
                    "Audio Processing"
                ],
                "publications": [
                    {
                        "title": "Weighted Point Cloud Embedding for Multimodal Contrastive Learning Toward Optimal Similarity Metric",
                        "abstract": "In typical multimodal contrastive learning, such as CLIP, encoders produce one point in the latent representation space for each input. However, one-point representation has difficulty in capturing the relationship and the similarity structure of a huge amount of instances in the real world. For richer classes of the similarity, we propose the use of weighted point clouds, namely, sets of pairs of weight and vector, as representations of instances. In this work, we theoretically show the benefit of our proposed method through a new understanding of the contrastive loss of CLIP, which we call symmetric InfoNCE. We clarify that the optimal similarity that minimizes symmetric InfoNCE is the pointwise mutual information, and show an upper bound of excess risk on downstream classification tasks of representations that achieve the optimal similarity. In addition, we show that our proposed similarity based on weighted point clouds consistently achieves the optimal similarity. To verify the effectiveness of our proposed method, we demonstrate pretraining of text-image representation models and classification tasks on common benchmarks."
                    },
                    {
                        "title": "DisMix: Disentangling Mixtures of Musical Instruments for Source-level Pitch and Timbre Manipulation",
                        "abstract": "Existing work on pitch and timbre disentanglement has been mostly focused on single-instrument music audio, excluding the cases where multiple instruments are presented. To fill the gap, we propose DisMix, a generative framework in which the pitch and timbre representations act as modular building blocks for constructing the melody and instrument of a source, and the collection of which forms a set of per-instrument latent representations underlying the observed mixture. By manipulating the representations, our model samples mixtures with novel combinations of pitch and timbre of the constituent instruments. We can jointly learn the disentangled pitch-timbre representations and a latent diffusion transformer that reconstructs the mixture conditioned on the set of source-level representations. We evaluate the model using both a simple dataset of isolated chords and a realistic four-part chorales in the style of J.S. Bach, identify the key components for the success of disentanglement, and demonstrate the application of mixture transformation based on source-level attribute manipulation."
                    },
                    {
                        "title": "PaGoDA: Progressive Growing of a One-Step Generator from a Low-Resolution Diffusion Teacher",
                        "abstract": "The diffusion model performs remarkable in generating high-dimensional content but is computationally intensive, especially during training. We propose Progressive Growing of Diffusion Autoencoder (PaGoDA), a novel pipeline that reduces the training costs through three stages: training diffusion on downsampled data, distilling the pretrained diffusion, and progressive super-resolution. With the proposed pipeline, PaGoDA achieves a $64\\times$ reduced cost in training its diffusion model on 8x downsampled data; while at the inference, with the single-step, it performs state-of-the-art on ImageNet across all resolutions from 64x64 to 512x512, and text-to-image. PaGoDA's pipeline can be applied directly in the latent space, adding compression alongside the pre-trained autoencoder in Latent Diffusion Models (e.g., Stable Diffusion). The code is available at https://github.com/sony/pagoda."
                    },
                    {
                        "title": "$\\textit{Jump Your Steps}$: Optimizing Sampling Schedule of Discrete Diffusion Models",
                        "abstract": "Diffusion models have seen notable success in continuous domains, leading to the development of discrete diffusion models (DDMs) for discrete variables. Despite recent advances, DDMs face the challenge of slow sampling speeds. While parallel sampling methods like $\\tau$-leaping accelerate this process, they introduce $\\textit{Compounding Decoding Error}$ (CDE), where discrepancies arise between the true distribution and the approximation from parallel token generation, leading to degraded sample quality. In this work, we present $\\textit{Jump Your Steps}$ (JYS), a novel approach that optimizes the allocation of discrete sampling timesteps by minimizing CDE without extra computational cost. More precisely, we derive a practical upper bound on CDE and propose an efficient algorithm for searching for the optimal sampling schedule. Extensive experiments across image, music, and text generation show that JYS significantly improves sampling quality, establishing it as a versatile framework for enhancing DDM performance for fast sampling."
                    },
                    {
                        "title": "Mitigating Embedding Collapse in Diffusion Models for Categorical Data",
                        "abstract": "Latent diffusion models have enabled continuous-state diffusion models to handle a variety of datasets, including categorical data. However, most methods rely on fixed pretrained embeddings, limiting the benefits of joint training with the diffusion model. While jointly learning the embedding (via reconstruction loss) and the latent diffusion model (via score matching loss) could enhance performance, our analysis shows that end-to-end training risks embedding collapse, degrading generation quality. To address this issue, we introduce CATDM, a continuous diffusion framework within the embedding space that stabilizes training. We propose a novel objective combining the joint embedding-diffusion variational lower bound with a Consistency-Matching (CM) regularizer, alongside a shifted cosine noise schedule and random dropping strategy. The CM regularizer ensures the recovery of the true data distribution. Experiments on benchmarks show that CATDM mitigates embedding collapse, yielding superior results on FFHQ, LSUN Churches, and LSUN Bedrooms. In particular, CATDM achieves an FID of 6.81 on ImageNet $256\\times256$ with 50 steps. It outperforms non-autoregressive models in machine translation and is on a par with previous methods in text generation."
                    },
                    {
                        "title": "SoundCTM: Uniting Score-based and Consistency Models for Text-to-Sound Generation",
                        "abstract": "Sound content is an indispensable element for multimedia works such as video games, music, and films. Recent high-quality diffusion-based sound generation models can serve as valuable tools for the creators. However, despite producing high-quality sounds, these models often suffer from slow inference speeds. This drawback burdens creators, who typically refine their sounds through trial and error to align them with their artistic intentions. To address this issue, we introduce Sound Consistency Trajectory Models (SoundCTM). Our model enables flexible transitioning between high-quality 1-step sound generation and superior sound quality through multi-step generation. This allows creators to initially control sounds with 1-step samples before refining them through multi-step generation. While CTM fundamentally achieves flexible 1-step and multi-step generation, its impressive performance heavily depends on an additional pretrained feature extractor and an adversarial loss, which are expensive to train and not always available in other domains. Thus, we reframe CTM's training framework and introduce a novel feature distance by utilizing the teacher's network for a distillation loss. Additionally, while distilling classifier-free guided trajectories, we train conditional and unconditional student models simultaneously and interpolate between these models during inference. We also propose training-free controllable frameworks for SoundCTM, leveraging its flexible sampling capability. SoundCTM achieves both promising 1-step and multi-step real-time sound generation without using any extra off-the-shelf networks. Furthermore, we demonstrate SoundCTM's capability of controllable sound generation in a training-free manner. Our codes, pretrained models, and audio samples are available at https://github.com/sony/soundctm."
                    },
                    {
                        "title": "Distillation of Discrete Diffusion through Dimensional Correlations",
                        "abstract": "Diffusion models have demonstrated exceptional performances in various fields of generative modeling. While they often outperform competitors including VAEs and GANs in sample quality and diversity, they suffer from slow sampling speed due to their iterative nature. Recently, distillation techniques and consistency models are mitigating this issue in continuous domains, but discrete diffusion models have some specific challenges towards faster generation. Most notably, in the current literature, correlations between different dimensions (pixels, locations) are ignored, both by its modeling and loss functions, due to computational limitations. In this paper, we propose\"mixture\"models in discrete diffusion that are capable of treating dimensional correlations while remaining scalable, and we provide a set of loss functions for distilling the iterations of existing models. Two primary theoretical insights underpin our approach: first, that dimensionally independent models can well approximate the data distribution if they are allowed to conduct many sampling steps, and second, that our loss functions enables mixture models to distill such many-step conventional models into just a few steps by learning the dimensional correlations. We empirically demonstrate that our proposed method for discrete diffusions work in practice, by distilling a continuous-time discrete diffusion model pretrained on the CIFAR-10 dataset."
                    },
                    {
                        "title": "VRVQ: Variable Bitrate Residual Vector Quantization for Audio Compression",
                        "abstract": "Recent state-of-the-art neural audio compression models have progressively adopted residual vector quantization (RVQ). Despite this success, these models employ a fixed number of codebooks per frame, which can be suboptimal in terms of rate-distortion tradeoff, particularly in scenarios with simple input audio, such as silence. To address this limitation, we propose variable bitrate RVQ (VRVQ) for audio codecs, which allows for more efficient coding by adapting the number of codebooks used per frame. Furthermore, we propose a gradient estimation method for the non-differentiable masking operation that transforms from the importance map to the binary importance mask, improving model training via a straight-through estimator. We demonstrate that the proposed training framework achieves superior results compared to the baseline method and shows further improvement when applied to the current state-of-the-art codec."
                    },
                    {
                        "title": "MoLA: Motion Generation and Editing with Latent Diffusion Enhanced by Adversarial Training",
                        "abstract": "In motion generation, controllability as well as generation quality and speed is becoming more and more important. There are various motion editing tasks, such as in-betweening, upper body editing, and path-following, but existing methods perform motion editing with a data-space diffusion model, which is slow in inference compared to a latent diffusion model. In this paper, we propose MoLA, which provides fast and high-quality motion generation and also can deal with multiple editing tasks in a single framework. For high-quality and fast generation, we employ a variational autoencoder and latent diffusion model, and improve the performance with adversarial training. In addition, we apply a training-free guided generation framework to achieve various editing tasks with motion control inputs. We quantitatively show the effectiveness of adversarial learning in text-to-motion generation, and demonstrate the applicability of our editing framework to multiple editing tasks in the motion domain."
                    },
                    {
                        "title": "G2D2: Gradient-guided Discrete Diffusion for image inverse problem solving",
                        "abstract": "Recent literature has effectively utilized diffusion models trained on continuous variables as priors for solving inverse problems. Notably, discrete diffusion models with discrete latent codes have shown strong performance, particularly in modalities suited for discrete compressed representations, such as image and motion generation. However, their discrete and non-differentiable nature has limited their application to inverse problems formulated in continuous spaces. This paper presents a novel method for addressing linear inverse problems by leveraging image-generation models based on discrete diffusion as priors. We overcome these limitations by approximating the true posterior distribution with a variational distribution constructed from categorical distributions and continuous relaxation techniques. Furthermore, we employ a star-shaped noise process to mitigate the drawbacks of traditional discrete diffusion models with absorbing states, demonstrating that our method performs comparably to continuous diffusion techniques. To the best of our knowledge, this is the first approach to use discrete diffusion model-based priors for solving image inverse problems."
                    },
                    {
                        "title": "SAN: Inducing Metrizability of GAN with Discriminative Normalized Linear Layer",
                        "abstract": "Generative adversarial networks (GANs) learn a target probability distribution by optimizing a generator and a discriminator with minimax objectives. This paper addresses the question of whether such optimization actually provides the generator with gradients that make its distribution close to the target distribution. We derive metrizable conditions, sufficient conditions for the discriminator to serve as the distance between the distributions by connecting the GAN formulation with the concept of sliced optimal transport. Furthermore, by leveraging these theoretical results, we propose a novel GAN training scheme, called slicing adversarial network (SAN). With only simple modifications, a broad class of existing GANs can be converted to SANs. Experiments on synthetic and image datasets support our theoretical results and the SAN's effectiveness as compared to usual GANs. Furthermore, we also apply SAN to StyleGAN-XL, which leads to state-of-the-art FID score amongst GANs for class conditional generation on ImageNet 256$\\times$256. Our implementation is available on https://ytakida.github.io/san."
                    },
                    {
                        "title": "Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion",
                        "abstract": "Consistency Models (CM) (Song et al., 2023) accelerate score-based diffusion model sampling at the cost of sample quality but lack a natural way to trade-off quality for speed. To address this limitation, we propose Consistency Trajectory Model (CTM), a generalization encompassing CM and score-based models as special cases. CTM trains a single neural network that can -- in a single forward pass -- output scores (i.e., gradients of log-density) and enables unrestricted traversal between any initial and final time along the Probability Flow Ordinary Differential Equation (ODE) in a diffusion process. CTM enables the efficient combination of adversarial training and denoising score matching loss to enhance performance and achieves new state-of-the-art FIDs for single-step diffusion model sampling on CIFAR-10 (FID 1.73) and ImageNet at 64x64 resolution (FID 1.92). CTM also enables a new family of sampling schemes, both deterministic and stochastic, involving long jumps along the ODE solution trajectories. It consistently improves sample quality as computational budgets increase, avoiding the degradation seen in CM. Furthermore, unlike CM, CTM's access to the score function can streamline the adoption of established controllable/conditional generation methods from the diffusion community. This access also enables the computation of likelihood. The code is available at https://github.com/sony/ctm."
                    },
                    {
                        "title": "HQ-VAE: Hierarchical Discrete Representation Learning with Variational Bayes",
                        "abstract": "Vector quantization (VQ) is a technique to deterministically learn features with discrete codebook representations. It is commonly performed with a variational autoencoding model, VQ-VAE, which can be further extended to hierarchical structures for making high-fidelity reconstructions. However, such hierarchical extensions of VQ-VAE often suffer from the codebook/layer collapse issue, where the codebook is not efficiently used to express the data, and hence degrades reconstruction accuracy. To mitigate this problem, we propose a novel unified framework to stochastically learn hierarchical discrete representation on the basis of the variational Bayes framework, called hierarchically quantized variational autoencoder (HQ-VAE). HQ-VAE naturally generalizes the hierarchical variants of VQ-VAE, such as VQ-VAE-2 and residual-quantized VAE (RQ-VAE), and provides them with a Bayesian training scheme. Our comprehensive experiments on image datasets show that HQ-VAE enhances codebook usage and improves reconstruction performance. We also validated HQ-VAE in terms of its applicability to a different modality with an audio dataset."
                    },
                    {
                        "title": "On the Equivalence of Consistency-Type Models: Consistency Models, Consistent Diffusion Models, and Fokker-Planck Regularization",
                        "abstract": "The emergence of various notions of ``consistency'' in diffusion models has garnered considerable attention and helped achieve improved sample quality, likelihood estimation, and accelerated sampling. Although similar concepts have been proposed in the literature, the precise relationships among them remain unclear. In this study, we establish theoretical connections between three recent ``consistency'' notions designed to enhance diffusion models for distinct objectives. Our insights offer the potential for a more comprehensive and encompassing framework for consistency-type models."
                    },
                    {
                        "title": "BIGVSAN: Enhancing Gan-Based Neural Vocoders with Slicing Adversarial Network",
                        "abstract": "Generative adversarial network (GAN)-based vocoders have been intensively studied because they can synthesize high-fidelity audio waveforms faster than real-time. However, it has been reported that most GANs fail to obtain the optimal projection for discriminating between real and fake data in the feature space. In the literature, it has been demonstrated that slicing adversarial network (SAN), an improved GAN training framework that can find the optimal projection, is effective in the image generation task. In this paper, we investigate the effectiveness of SAN in the vocoding task. For this purpose, we propose a scheme to modify least-squares GAN, which most GAN-based vocoders adopt, so that their loss functions satisfy the requirements of SAN. Through our experiments, we demonstrate that SAN can improve the performance of GAN-based vocoders, including BigVGAN, with small modifications. Our code is available at https://github.com/sony/bigvsan."
                    },
                    {
                        "title": "Adversarially Slicing Generative Networks: Discriminator Slices Feature for One-Dimensional Optimal Transport",
                        "abstract": "Generative adversarial networks (GANs) learn a target probability distribution by optimizing a generator and a discriminator with minimax objectives. This paper addresses the question of whether such optimization actually provides the generator with gradients that make its distribution close to the target distribution. We derive suf\ufb01cient conditions for the discriminator to serve as the distance between the distributions by connecting the GAN formulation with the concept of sliced optimal transport. Furthermore, by leveraging these theoretical results, we propose a novel GAN training scheme, called adversarially slicing generative network (ASGN). With only simple modi\ufb01cations, the ASGN is applicable to a broad class of existing GANs. Experiments on synthetic and image datasets support our theoretical results and the ASGN\u2019s effectiveness as compared to usual GANs."
                    },
                    {
                        "title": "Improving Score-based Diffusion Models by Enforcing the Underlying Score Fokker-Planck Equation",
                        "abstract": "Score-based generative models learn a family of noise-conditional score functions corresponding to the data density perturbed with increasingly large amounts of noise. These perturbed data densities are tied together by the Fokker-Planck equation (FPE), a partial differential equation (PDE) governing the spatial-temporal evolution of a density undergoing a diffusion process. In this work, we derive a corresponding equation, called the score FPE that characterizes the noise-conditional scores of the perturbed data densities (i.e., their gradients). Surprisingly, despite impressive empirical performance, we observe that scores learned via denoising score matching (DSM) do not satisfy the underlying score FPE. We prove that satisfying the FPE is desirable as it improves the likelihood and the degree of conservativity. Hence, we propose to regularize the DSM objective to enforce satisfaction of the score FPE, and we show the effectiveness of this approach across various datasets."
                    },
                    {
                        "title": "On the Language Encoder of Contrastive Cross-modal Models",
                        "abstract": "Contrastive cross-modal models such as CLIP and CLAP aid various vision-language (VL) and audio-language (AL) tasks. However, there has been limited investigation of and improvement in their language encoder, which is the central component of encoding natural language descriptions of image/audio into vector representations. We extensively evaluate how unsupervised and supervised sentence embedding training affect language encoder quality and cross-modal task performance. In VL pretraining, we found that sentence embedding training language encoder quality and aids in cross-modal tasks, improving contrastive VL models such as CyCLIP. In contrast, AL pretraining benefits less from sentence embedding training, which may result from the limited amount of pretraining data. We analyze the representation spaces to understand the strengths of sentence embedding training, and find that it improves text-space uniformity, at the cost of decreased cross-modal alignment."
                    },
                    {
                        "title": "GibbsDDRM: A Partially Collapsed Gibbs Sampler for Solving Blind Inverse Problems with Denoising Diffusion Restoration",
                        "abstract": "Pre-trained diffusion models have been successfully used as priors in a variety of linear inverse problems, where the goal is to reconstruct a signal from noisy linear measurements. However, existing approaches require knowledge of the linear operator. In this paper, we propose GibbsDDRM, an extension of Denoising Diffusion Restoration Models (DDRM) to a blind setting in which the linear measurement operator is unknown. GibbsDDRM constructs a joint distribution of the data, measurements, and linear operator by using a pre-trained diffusion model for the data prior, and it solves the problem by posterior sampling with an efficient variant of a Gibbs sampler. The proposed method is problem-agnostic, meaning that a pre-trained diffusion model can be applied to various inverse problems without fine-tuning. In experiments, it achieved high performance on both blind image deblurring and vocal dereverberation tasks, despite the use of simple generic priors for the underlying linear operators."
                    }
                ]
            },
            "8a9edfdd-4a65-4157-8d5a-4ac13181efad": {
                "pk": "8a9edfdd-4a65-4157-8d5a-4ac13181efad",
                "name": "Toshimitsu Uesaka",
                "collaborators": [
                    "Yuki Mitsufuji",
                    "Yuhta Takida",
                    "Naoki Murata",
                    "Chieh-Hsin Lai",
                    "Stefano Ermon",
                    "Wei-Hsiang Liao",
                    "Takashi Shibuya",
                    "Shusuke Takahashi",
                    "Koichi Saito",
                    "Ryosuke Sawata"
                ],
                "domain": [
                    "Generative Models",
                    "Diffusion Models",
                    "Music Processing",
                    "Contrastive Learning"
                ],
                "publications": [
                    {
                        "title": "Weighted Point Cloud Embedding for Multimodal Contrastive Learning Toward Optimal Similarity Metric",
                        "abstract": "In typical multimodal contrastive learning, such as CLIP, encoders produce one point in the latent representation space for each input. However, one-point representation has difficulty in capturing the relationship and the similarity structure of a huge amount of instances in the real world. For richer classes of the similarity, we propose the use of weighted point clouds, namely, sets of pairs of weight and vector, as representations of instances. In this work, we theoretically show the benefit of our proposed method through a new understanding of the contrastive loss of CLIP, which we call symmetric InfoNCE. We clarify that the optimal similarity that minimizes symmetric InfoNCE is the pointwise mutual information, and show an upper bound of excess risk on downstream classification tasks of representations that achieve the optimal similarity. In addition, we show that our proposed similarity based on weighted point clouds consistently achieves the optimal similarity. To verify the effectiveness of our proposed method, we demonstrate pretraining of text-image representation models and classification tasks on common benchmarks."
                    },
                    {
                        "title": "DisMix: Disentangling Mixtures of Musical Instruments for Source-level Pitch and Timbre Manipulation",
                        "abstract": "Existing work on pitch and timbre disentanglement has been mostly focused on single-instrument music audio, excluding the cases where multiple instruments are presented. To fill the gap, we propose DisMix, a generative framework in which the pitch and timbre representations act as modular building blocks for constructing the melody and instrument of a source, and the collection of which forms a set of per-instrument latent representations underlying the observed mixture. By manipulating the representations, our model samples mixtures with novel combinations of pitch and timbre of the constituent instruments. We can jointly learn the disentangled pitch-timbre representations and a latent diffusion transformer that reconstructs the mixture conditioned on the set of source-level representations. We evaluate the model using both a simple dataset of isolated chords and a realistic four-part chorales in the style of J.S. Bach, identify the key components for the success of disentanglement, and demonstrate the application of mixture transformation based on source-level attribute manipulation."
                    },
                    {
                        "title": "PaGoDA: Progressive Growing of a One-Step Generator from a Low-Resolution Diffusion Teacher",
                        "abstract": "The diffusion model performs remarkable in generating high-dimensional content but is computationally intensive, especially during training. We propose Progressive Growing of Diffusion Autoencoder (PaGoDA), a novel pipeline that reduces the training costs through three stages: training diffusion on downsampled data, distilling the pretrained diffusion, and progressive super-resolution. With the proposed pipeline, PaGoDA achieves a $64\\times$ reduced cost in training its diffusion model on 8x downsampled data; while at the inference, with the single-step, it performs state-of-the-art on ImageNet across all resolutions from 64x64 to 512x512, and text-to-image. PaGoDA's pipeline can be applied directly in the latent space, adding compression alongside the pre-trained autoencoder in Latent Diffusion Models (e.g., Stable Diffusion). The code is available at https://github.com/sony/pagoda."
                    },
                    {
                        "title": "Mitigating Embedding Collapse in Diffusion Models for Categorical Data",
                        "abstract": "Latent diffusion models have enabled continuous-state diffusion models to handle a variety of datasets, including categorical data. However, most methods rely on fixed pretrained embeddings, limiting the benefits of joint training with the diffusion model. While jointly learning the embedding (via reconstruction loss) and the latent diffusion model (via score matching loss) could enhance performance, our analysis shows that end-to-end training risks embedding collapse, degrading generation quality. To address this issue, we introduce CATDM, a continuous diffusion framework within the embedding space that stabilizes training. We propose a novel objective combining the joint embedding-diffusion variational lower bound with a Consistency-Matching (CM) regularizer, alongside a shifted cosine noise schedule and random dropping strategy. The CM regularizer ensures the recovery of the true data distribution. Experiments on benchmarks show that CATDM mitigates embedding collapse, yielding superior results on FFHQ, LSUN Churches, and LSUN Bedrooms. In particular, CATDM achieves an FID of 6.81 on ImageNet $256\\times256$ with 50 steps. It outperforms non-autoregressive models in machine translation and is on a par with previous methods in text generation."
                    },
                    {
                        "title": "G2D2: Gradient-guided Discrete Diffusion for image inverse problem solving",
                        "abstract": "Recent literature has effectively utilized diffusion models trained on continuous variables as priors for solving inverse problems. Notably, discrete diffusion models with discrete latent codes have shown strong performance, particularly in modalities suited for discrete compressed representations, such as image and motion generation. However, their discrete and non-differentiable nature has limited their application to inverse problems formulated in continuous spaces. This paper presents a novel method for addressing linear inverse problems by leveraging image-generation models based on discrete diffusion as priors. We overcome these limitations by approximating the true posterior distribution with a variational distribution constructed from categorical distributions and continuous relaxation techniques. Furthermore, we employ a star-shaped noise process to mitigate the drawbacks of traditional discrete diffusion models with absorbing states, demonstrating that our method performs comparably to continuous diffusion techniques. To the best of our knowledge, this is the first approach to use discrete diffusion model-based priors for solving image inverse problems."
                    },
                    {
                        "title": "SAN: Inducing Metrizability of GAN with Discriminative Normalized Linear Layer",
                        "abstract": "Generative adversarial networks (GANs) learn a target probability distribution by optimizing a generator and a discriminator with minimax objectives. This paper addresses the question of whether such optimization actually provides the generator with gradients that make its distribution close to the target distribution. We derive metrizable conditions, sufficient conditions for the discriminator to serve as the distance between the distributions by connecting the GAN formulation with the concept of sliced optimal transport. Furthermore, by leveraging these theoretical results, we propose a novel GAN training scheme, called slicing adversarial network (SAN). With only simple modifications, a broad class of existing GANs can be converted to SANs. Experiments on synthetic and image datasets support our theoretical results and the SAN's effectiveness as compared to usual GANs. Furthermore, we also apply SAN to StyleGAN-XL, which leads to state-of-the-art FID score amongst GANs for class conditional generation on ImageNet 256$\\times$256. Our implementation is available on https://ytakida.github.io/san."
                    },
                    {
                        "title": "Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion",
                        "abstract": "Consistency Models (CM) (Song et al., 2023) accelerate score-based diffusion model sampling at the cost of sample quality but lack a natural way to trade-off quality for speed. To address this limitation, we propose Consistency Trajectory Model (CTM), a generalization encompassing CM and score-based models as special cases. CTM trains a single neural network that can -- in a single forward pass -- output scores (i.e., gradients of log-density) and enables unrestricted traversal between any initial and final time along the Probability Flow Ordinary Differential Equation (ODE) in a diffusion process. CTM enables the efficient combination of adversarial training and denoising score matching loss to enhance performance and achieves new state-of-the-art FIDs for single-step diffusion model sampling on CIFAR-10 (FID 1.73) and ImageNet at 64x64 resolution (FID 1.92). CTM also enables a new family of sampling schemes, both deterministic and stochastic, involving long jumps along the ODE solution trajectories. It consistently improves sample quality as computational budgets increase, avoiding the degradation seen in CM. Furthermore, unlike CM, CTM's access to the score function can streamline the adoption of established controllable/conditional generation methods from the diffusion community. This access also enables the computation of likelihood. The code is available at https://github.com/sony/ctm."
                    },
                    {
                        "title": "HQ-VAE: Hierarchical Discrete Representation Learning with Variational Bayes",
                        "abstract": "Vector quantization (VQ) is a technique to deterministically learn features with discrete codebook representations. It is commonly performed with a variational autoencoding model, VQ-VAE, which can be further extended to hierarchical structures for making high-fidelity reconstructions. However, such hierarchical extensions of VQ-VAE often suffer from the codebook/layer collapse issue, where the codebook is not efficiently used to express the data, and hence degrades reconstruction accuracy. To mitigate this problem, we propose a novel unified framework to stochastically learn hierarchical discrete representation on the basis of the variational Bayes framework, called hierarchically quantized variational autoencoder (HQ-VAE). HQ-VAE naturally generalizes the hierarchical variants of VQ-VAE, such as VQ-VAE-2 and residual-quantized VAE (RQ-VAE), and provides them with a Bayesian training scheme. Our comprehensive experiments on image datasets show that HQ-VAE enhances codebook usage and improves reconstruction performance. We also validated HQ-VAE in terms of its applicability to a different modality with an audio dataset."
                    },
                    {
                        "title": "On the Equivalence of Consistency-Type Models: Consistency Models, Consistent Diffusion Models, and Fokker-Planck Regularization",
                        "abstract": "The emergence of various notions of ``consistency'' in diffusion models has garnered considerable attention and helped achieve improved sample quality, likelihood estimation, and accelerated sampling. Although similar concepts have been proposed in the literature, the precise relationships among them remain unclear. In this study, we establish theoretical connections between three recent ``consistency'' notions designed to enhance diffusion models for distinct objectives. Our insights offer the potential for a more comprehensive and encompassing framework for consistency-type models."
                    },
                    {
                        "title": "Adversarially Slicing Generative Networks: Discriminator Slices Feature for One-Dimensional Optimal Transport",
                        "abstract": "Generative adversarial networks (GANs) learn a target probability distribution by optimizing a generator and a discriminator with minimax objectives. This paper addresses the question of whether such optimization actually provides the generator with gradients that make its distribution close to the target distribution. We derive suf\ufb01cient conditions for the discriminator to serve as the distance between the distributions by connecting the GAN formulation with the concept of sliced optimal transport. Furthermore, by leveraging these theoretical results, we propose a novel GAN training scheme, called adversarially slicing generative network (ASGN). With only simple modi\ufb01cations, the ASGN is applicable to a broad class of existing GANs. Experiments on synthetic and image datasets support our theoretical results and the ASGN\u2019s effectiveness as compared to usual GANs."
                    },
                    {
                        "title": "Improving Score-based Diffusion Models by Enforcing the Underlying Score Fokker-Planck Equation",
                        "abstract": "Score-based generative models learn a family of noise-conditional score functions corresponding to the data density perturbed with increasingly large amounts of noise. These perturbed data densities are tied together by the Fokker-Planck equation (FPE), a partial differential equation (PDE) governing the spatial-temporal evolution of a density undergoing a diffusion process. In this work, we derive a corresponding equation, called the score FPE that characterizes the noise-conditional scores of the perturbed data densities (i.e., their gradients). Surprisingly, despite impressive empirical performance, we observe that scores learned via denoising score matching (DSM) do not satisfy the underlying score FPE. We prove that satisfying the FPE is desirable as it improves the likelihood and the degree of conservativity. Hence, we propose to regularize the DSM objective to enforce satisfaction of the score FPE, and we show the effectiveness of this approach across various datasets."
                    },
                    {
                        "title": "GibbsDDRM: A Partially Collapsed Gibbs Sampler for Solving Blind Inverse Problems with Denoising Diffusion Restoration",
                        "abstract": "Pre-trained diffusion models have been successfully used as priors in a variety of linear inverse problems, where the goal is to reconstruct a signal from noisy linear measurements. However, existing approaches require knowledge of the linear operator. In this paper, we propose GibbsDDRM, an extension of Denoising Diffusion Restoration Models (DDRM) to a blind setting in which the linear measurement operator is unknown. GibbsDDRM constructs a joint distribution of the data, measurements, and linear operator by using a pre-trained diffusion model for the data prior, and it solves the problem by posterior sampling with an efficient variant of a Gibbs sampler. The proposed method is problem-agnostic, meaning that a pre-trained diffusion model can be applied to various inverse problems without fine-tuning. In experiments, it achieved high performance on both blind image deblurring and vocal dereverberation tasks, despite the use of simple generic priors for the underlying linear operators."
                    },
                    {
                        "title": "A Versatile Diffusion-based Generative Refiner for Speech Enhancement",
                        "abstract": "Although deep neural network (DNN)-based speech enhancement (SE) methods outperform the previous non-DNN-based ones, they often degrade the perceptual quality of generated outputs. To tackle this problem, we introduce a DNN-based generative refiner, Diffiner, aiming to improve perceptual speech quality pre-processed by an SE method. We train a diffusion-based generative model by utilizing a dataset consisting of clean speech only. Then, our refiner effectively mixes clean parts newly generated via denoising diffusion restoration into the degraded and distorted parts caused by a preceding SE method, resulting in refined speech. Once our refiner is trained on a set of clean speech, it can be applied to various SE methods without additional training specialized for each SE module. Therefore, our refiner can be a versatile post-processing module w.r.t. SE methods and has high potential in terms of modularity. Experimental results show that our method improved perceptual speech quality regardless of the preceding SE methods used."
                    },
                    {
                        "title": "Unsupervised Vocal Dereverberation with Diffusion-Based Generative Models",
                        "abstract": "Removing reverb from reverberant music is a necessary technique to clean up audio for downstream music manipulations. Reverberation of music contains two categories, natural reverb, and artificial reverb. Artificial reverb has a wider diversity than natural reverb due to its various parameter setups and reverberation types. However, recent supervised dereverberation methods may fail because they rely on sufficiently diverse and numerous pairs of reverberant observations and retrieved data for training in order to be generalizable to unseen observations during inference. To resolve these problems, we propose an unsupervised method that can remove a general kind of artificial reverb for music without requiring pairs of data for training. The proposed method is based on diffusion models, where it initializes the unknown reverberation operator with a conventional signal processing technique and simultaneously refines the estimate with the help of diffusion models. We show through objective and perceptual evaluations that our method outperforms the current leading vocal dereverberation benchmarks."
                    },
                    {
                        "title": "SQ-VAE: Variational Bayes on Discrete Representation with Self-annealed Stochastic Quantization",
                        "abstract": "One noted issue of vector-quantized variational autoencoder (VQ-VAE) is that the learned discrete representation uses only a fraction of the full capacity of the codebook, also known as codebook collapse. We hypothesize that the training scheme of VQ-VAE, which involves some carefully designed heuristics, underlies this issue. In this paper, we propose a new training scheme that extends the standard VAE via novel stochastic dequantization and quantization, called stochastically quantized variational autoencoder (SQ-VAE). In SQ-VAE, we observe a trend that the quantization is stochastic at the initial stage of the training but gradually converges toward a deterministic quantization, which we call self-annealing. Our experiments show that SQ-VAE improves codebook utilization without using common heuristics. Furthermore, we empirically show that SQ-VAE is superior to VAE and VQ-VAE in vision- and speech-related tasks."
                    },
                    {
                        "title": "FP-Diffusion: Improving Score-based Diffusion Models by Enforcing the Underlying Score Fokker-Planck Equation",
                        "abstract": "Score-based generative models (SGMs) learn a family of noise-conditional score functions corresponding to the data density perturbed with increasingly large amounts of noise. These perturbed data densities are linked together by the Fokker-Planck equation (FPE), a partial differential equation (PDE) governing the spatial-temporal evolution of a density undergoing a diffusion process. In this work, we derive a corresponding equation called the score FPE that characterizes the noise-conditional scores of the perturbed data densities (i.e., their gradients). Surprisingly, despite the impressive empirical performance, we observe that scores learned through denoising score matching (DSM) fail to fulfill the underlying score FPE, which is an inherent self-consistency property of the ground truth score. We prove that satisfying the score FPE is desirable as it improves the likelihood and the degree of conservativity. Hence, we propose to regularize the DSM objective to enforce satisfaction of the score FPE, and we show the effectiveness of this approach across various datasets."
                    },
                    {
                        "title": "Diffroll: Diffusion-Based Generative Music Transcription with Unsupervised Pretraining Capability",
                        "abstract": "In this paper we propose a novel generative approach, DiffRoll, to tackle automatic music transcription (AMT). Instead of treating AMT as a discriminative task in which the model is trained to convert spectrograms into piano rolls, we think of it as a conditional generative task where we train our model to generate realistic looking piano rolls from pure Gaussian noise conditioned on spectrograms. This new AMT formulation enables DiffRoll to transcribe, generate and even inpaint music. Due to the classifier-free nature, DiffRoll is also able to be trained on unpaired datasets where only piano rolls are available. Our experiments show that DiffRoll outperforms its discriminative counterpart by 19 percentage points (ppt.) and our ablation studies also indicate that it outperforms similar existing methods by 4.8 ppt.Source code and demonstration are available at https://sony.github.io/DiffRoll/."
                    },
                    {
                        "title": "Regularizing Score-based Models with Score Fokker-Planck Equations",
                        "abstract": "Score-based generative models learn a family of noise-conditional score functions corresponding to the data density perturbed with increasingly large amounts of noise. These perturbed data densities are tied together by the Fokker-Planck equation (FPE), a PDE governing the spatial-temporal evolution of a density undergoing a diffusion process. In this work, we derive a corresponding equation characterizing the noise-conditional scores of the perturbed data densities (i.e., their gradients), termed the score FPE . Surprisingly, despite impressive empirical performance, we observe that scores learned via denoising score matching (DSM) do not satisfy the underlying score FPE. We mathematically analyze three implications of satisfying the score FPE and a potential explanation for why the score FPE is not ful\ufb01lled in practice. At last, we propose to regularize the DSM objective to enforce satisfaction of the score FPE, and show its effectiveness on synthetic data and MNIST."
                    },
                    {
                        "title": "A VERSATILE DIFFUSION-BASED GENERATIVE REFINER FOR SPEECH ENHANCEMENT",
                        "abstract": "Although deep neural network (DNN)-based speech enhancement (SE) methods outperform the previous non-DNN-based ones, they often degrade the perceptual quality of generated outputs. To tackle this problem, We introduce a DNN-based generative re\ufb01ner aiming to improve perceptual speech quality pre-processed by an SE method. As the re\ufb01ner, we train a diffusion-based generative model by utilizing a dataset consisting of clean speech only. Then, the model replaces the degraded and distorted parts caused by a preceding SE method with newly generated clean parts by denoising diffusion restoration. Once our re\ufb01ner is trained on a set of clean speech, it can be applied to various SE methods without additional training specialized for each SE module. Therefore, our re\ufb01ner can be a versatile post-processing module w.r.t. SE methods and has high potential in terms of modularity. Experimental results show that our method improved perceptual speech quality regardless of the preceding SE methods used."
                    }
                ]
            },
            "35be5a1c-9394-461d-9fa7-63028b424f8a": {
                "pk": "35be5a1c-9394-461d-9fa7-63028b424f8a",
                "name": "Dongjun Kim",
                "collaborators": [
                    "Il-Chul Moon",
                    "Wanmo Kang",
                    "Seung-Jae Shin",
                    "Kyungwoo Song",
                    "Weonyoung Joo",
                    "Yeongmin Kim",
                    "Byeonghu Na",
                    "Tae-Sub Yun",
                    "J. Bae",
                    "Seungjae Shin"
                ],
                "domain": [
                    "Generative Models",
                    "Diffusion Models",
                    "Simulation",
                    "Bayesian Inference"
                ],
                "publications": [
                    {
                        "title": "Training Unbiased Diffusion Models From Biased Dataset",
                        "abstract": "With significant advancements in diffusion models, addressing the potential risks of dataset bias becomes increasingly important. Since generated outputs directly suffer from dataset bias, mitigating latent bias becomes a key factor in improving sample quality and proportion. This paper proposes time-dependent importance reweighting to mitigate the bias for the diffusion models. We demonstrate that the time-dependent density ratio becomes more precise than previous approaches, thereby minimizing error propagation in generative learning. While directly applying it to score-matching is intractable, we discover that using the time-dependent density ratio both for reweighting and score correction can lead to a tractable form of the objective function to regenerate the unbiased data density. Furthermore, we theoretically establish a connection with traditional score-matching, and we demonstrate its convergence to an unbiased distribution. The experimental evidence supports the usefulness of the proposed method, which outperforms baselines including time-independent importance reweighting on CIFAR-10, CIFAR-100, FFHQ, and CelebA with various bias settings. Our code is available at https://github.com/alsdudrla10/TIW-DSM."
                    },
                    {
                        "title": "Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion",
                        "abstract": "Consistency Models (CM) (Song et al., 2023) accelerate score-based diffusion model sampling at the cost of sample quality but lack a natural way to trade-off quality for speed. To address this limitation, we propose Consistency Trajectory Model (CTM), a generalization encompassing CM and score-based models as special cases. CTM trains a single neural network that can -- in a single forward pass -- output scores (i.e., gradients of log-density) and enables unrestricted traversal between any initial and final time along the Probability Flow Ordinary Differential Equation (ODE) in a diffusion process. CTM enables the efficient combination of adversarial training and denoising score matching loss to enhance performance and achieves new state-of-the-art FIDs for single-step diffusion model sampling on CIFAR-10 (FID 1.73) and ImageNet at 64x64 resolution (FID 1.92). CTM also enables a new family of sampling schemes, both deterministic and stochastic, involving long jumps along the ODE solution trajectories. It consistently improves sample quality as computational budgets increase, avoiding the degradation seen in CM. Furthermore, unlike CM, CTM's access to the score function can streamline the adoption of established controllable/conditional generation methods from the diffusion community. This access also enables the computation of likelihood. The code is available at https://github.com/sony/ctm."
                    },
                    {
                        "title": "Agent-Based Model for Urban Administration: A Case Study of Bridge Construction and its Traffic Dispersion Effect",
                        "abstract": ": From the viewpoint of urban administration, simulation is regarded as a policy tool that provides administrators with information about the current urban situation and enables them to verify the effectiveness of urban policies. This study proposes a traffic simulation model for a real city named Sejong in South Korea. Our proposed model employs agent-based simulation with the city-level real data, which mainly focuses on describing the movement behavior of individuals using urban traffics in the real city. By aggregating the agents\u2019 decisions and interactions during the movement, the proposed model can discover a demand for the city\u2019s transportation system. To do this, this study validated the proposed model so that the modeled traffic system was similar to the real one, and then we conducted a case study to compare and analyze the effects of traffic dispersion led by the upcoming bridge construction in the real city. The case study showed that the proposed model can provide policy evaluation on the optimal location of the bridge construction considering the city traffic flow. Furthermore, the case study presented that the agent-based modeling enables micro-level analysis on the city traffic flow to understand on the policy implications."
                    },
                    {
                        "title": "Automatic Calibration Framework of Agent-based Models for Dynamic and Heterogeneous Parameters",
                        "abstract": "Agent-based models (ABMs) highlight the importance of simulation validation, such as qualitative face validation and quantitative empirical validation. In particular, we focused on quantitative validation by adjusting simulation input parameters of the ABM. This study introduces an automatic calibration framework that combines the suggested dynamic and heterogeneous calibration methods. Specifically, the dynamic calibration fits the simulation results to the real-world data by automatically capturing suitable simulation time to adjust the simulation parameters. Meanwhile, the heterogeneous calibration reduces the distributional discrepancy between individuals in the simulation and the real world by adjusting agent related parameters cluster-wisely."
                    },
                    {
                        "title": "Refining Generative Process with Discriminator Guidance in Score-based Diffusion Models",
                        "abstract": "The proposed method, Discriminator Guidance, aims to improve sample generation of pre-trained diffusion models. The approach introduces a discriminator that gives explicit supervision to a denoising sample path whether it is realistic or not. Unlike GANs, our approach does not require joint training of score and discriminator networks. Instead, we train the discriminator after score training, making discriminator training stable and fast to converge. In sample generation, we add an auxiliary term to the pre-trained score to deceive the discriminator. This term corrects the model score to the data score at the optimal discriminator, which implies that the discriminator helps better score estimation in a complementary way. Using our algorithm, we achive state-of-the-art results on ImageNet 256x256 with FID 1.83 and recall 0.64, similar to the validation data's FID (1.68) and recall (0.66). We release the code at https://github.com/alsdudrla10/DG."
                    },
                    {
                        "title": "Maximum Likelihood Training of Implicit Nonlinear Diffusion Models",
                        "abstract": "Whereas diverse variations of diffusion models exist, extending the linear diffusion into a nonlinear diffusion process is investigated by very few works. The nonlinearity effect has been hardly understood, but intuitively, there would be promising diffusion patterns to efficiently train the generative distribution towards the data distribution. This paper introduces a data-adaptive nonlinear diffusion process for score-based diffusion models. The proposed Implicit Nonlinear Diffusion Model (INDM) learns by combining a normalizing flow and a diffusion process. Specifically, INDM implicitly constructs a nonlinear diffusion on the \\textit{data space} by leveraging a linear diffusion on the \\textit{latent space} through a flow network. This flow network is key to forming a nonlinear diffusion, as the nonlinearity depends on the flow network. This flexible nonlinearity improves the learning curve of INDM to nearly Maximum Likelihood Estimation (MLE) against the non-MLE curve of DDPM++, which turns out to be an inflexible version of INDM with the flow fixed as an identity mapping. Also, the discretization of INDM shows the sampling robustness. In experiments, INDM achieves the state-of-the-art FID of 1.75 on CelebA. We release our code at \\url{https://github.com/byeonghu-na/INDM}."
                    },
                    {
                        "title": "Score Matching Model for Unbounded Data Score",
                        "abstract": "Recent advance in diffusion models incorporates the Stochastic Differential Equation (SDE), which brings the state-of-the art performance on image generation tasks. This paper improves such diffusion models by analyzing the model at the zero diffusion time. In real datasets, the score function diverges as the diffusion time ( t ) decreases to zero, and this observation leads an argument that the score estimation fails at t = 0 with any neural network structure. Subsequently, we introduce Unbounded Diffusion Model (UDM) that resolves the score diverging problem with an easily applicable modi\ufb01cation to any diffusion models. Additionally, we introduce a new SDE that overcomes the theoretic and practical limitations of Variance Exploding SDE. On top of that, the introduced Soft Truncation method improves the sample quality by mitigating the loss scale issue that happens at t = 0 . We further provide a theoretic result of the proposed method to uncover the behind mechanism of the diffusion models."
                    },
                    {
                        "title": "Soft Truncation: A Universal Training Technique of Score-based Diffusion Model for High Precision Score Estimation",
                        "abstract": "Recent advances in diffusion models bring state-of-the-art performance on image generation tasks. However, empirical results from previous research in diffusion models imply an inverse correlation between density estimation and sample generation performances. This paper investigates with sufficient empirical evidence that such inverse correlation happens because density estimation is significantly contributed by small diffusion time, whereas sample generation mainly depends on large diffusion time. However, training a score network well across the entire diffusion time is demanding because the loss scale is significantly imbalanced at each diffusion time. For successful training, therefore, we introduce Soft Truncation, a universally applicable training technique for diffusion models, that softens the fixed and static truncation hyperparameter into a random variable. In experiments, Soft Truncation achieves state-of-the-art performance on CIFAR-10, CelebA, CelebA-HQ 256x256, and STL-10 datasets."
                    },
                    {
                        "title": "Posterior-Aided Regularization for Likelihood-Free Inference",
                        "abstract": "The recent development of likelihood-free inference aims training a \ufb02exible density estimator for the target posterior with a set of input-output pairs from simulation . Given the diversity of simulation structures, it is dif\ufb01cult to \ufb01nd a single uni\ufb01ed inference method for each simulation model. This paper proposes a universally applicable regularization technique, called Posterior-Aided Regular-ization (PAR), which is applicable to learning the density estimator, regardless of the model structure. Particularly, PAR solves the mode collapse problem that arises as the output dimension of the simulation increases. PAR resolves this posterior mode degeneracy through a mixture of 1) the reverse KL divergence with the mode seeking property; and 2) the mutual information for the high quality representation on likelihood. Because of the estimation intractability of PAR, we provide a uni\ufb01ed estimation method of PAR to estimate both reverse KL term and mutual information term with a single neural network. Af-terwards, we theoretically prove the asymptotic convergence of the regularized optimal solution to the unregularized optimal solution as the regularization magnitude converges to zero. Additionally, we empirically show that past sequential neural likelihood inferences in conjunction with PAR present the statistically signi\ufb01cant gains on diverse simulation tasks."
                    },
                    {
                        "title": "Neural Posterior Regularization for Likelihood-Free Inference",
                        "abstract": "A simulation is useful when the phenomenon of interest is either expensive to regenerate or irreproducible with the same context. Recently, Bayesian inference on the distribution of the simulation input parameter has been implemented sequentially to minimize the required simulation budget for the task of simulation validation to the real-world. However, the Bayesian inference is still challenging when the ground-truth posterior is multi-modal with a high-dimensional simulation output. This paper introduces a regularization technique, namely Neural Posterior Regularization (NPR), which enforces the model to explore the input parameter space e \ufb00 ectively. Afterward, we provide the closed-form solution of the regularized optimization that enables analyzing the e \ufb00 ect of the regularization. We empirically validate that NPR attains the statistically signi\ufb01cant gain on benchmark performances for diverse simulation tasks."
                    },
                    {
                        "title": "Adversarial Likelihood-Free Inference on Black-Box Generator",
                        "abstract": "Generative Adversarial Network (GAN) can be viewed as an implicit estimator of a data distribution, and this perspective motivates using the adversarial concept in the true input parameter estimation of black-box generators. While previous works on likelihood-free inference introduces an implicit proposal distribution on the generator input, this paper analyzes theoretic limitations of the proposal distribution approach. On top of that, we introduce a new algorithm, Adversarial Likelihood-Free Inference (ALFI), to mitigate the analyzed limitations, so ALFI is able to find the posterior distribution on the input parameter for black-box generative models. We experimented ALFI with diverse simulation models as well as pre-trained statistical models, and we identified that ALFI achieves the best parameter estimation accuracy with a limited simulation budget."
                    },
                    {
                        "title": "Generalized Gumbel-Softmax Gradient Estimator for Various Discrete Random Variables",
                        "abstract": "Estimating the gradients of stochastic nodes is one of the crucial research questions in the deep generative modeling community, which enables the gradient descent optimization on neural network parameters. This estimation problem becomes further complex when we regard the stochastic nodes to be discrete because pathwise derivative techniques cannot be applied. Hence, the stochastic gradient estimation of discrete distributions requires either a score function method or continuous relaxation of the discrete random variables. This paper proposes a general version of the Gumbel-Softmax estimator with continuous relaxation, and this estimator is able to relax the discreteness of probability distributions including more diverse types, other than categorical and Bernoulli. In detail, we utilize the truncation of discrete random variables and the Gumbel-Softmax trick with a linear transformation for the relaxed reparameterization. The proposed approach enables the relaxed discrete random variable to be reparameterized and to backpropagated through a large scale stochastic computational graph. Our experiments consist of (1) synthetic data analyses, which show the efficacy of our methods; and (2) applications on VAE and topic model, which demonstrate the value of the proposed estimation in practices."
                    },
                    {
                        "title": "Generalized Gumbel-Softmax Gradient Estimator for Generic Discrete Random Variables",
                        "abstract": "Estimating the gradients of stochastic nodes in stochastic computational graphs is one of the crucial research questions in the deep generative modeling community, which enables the gradient descent optimization on neural network parameters. Stochastic gradient estimators of discrete random variables are widely explored, for example, Gumbel-Softmax reparameterization trick for Bernoulli and categorical distributions. Meanwhile, other discrete distribution cases such as the Poisson, geometric, binomial, multinomial, negative binomial, etc. have not been explored. This paper proposes a generalized version of the Gumbel-Softmax estimator, which is able to reparameterize generic discrete distributions, not restricted to the Bernoulli and the categorical. The proposed estimator utilizes the truncation of discrete random variables, the Gumbel-Softmax trick, and a special form of linear transformation. Our experiments consist of (1) synthetic examples and applications on VAE, which show the efficacy of our methods; and (2) topic models, which demonstrate the value of the proposed estimation in practice."
                    }
                ]
            },
            "7dae60b5-6d85-423a-bd4b-c4cf3f7f46f0": {
                "pk": "7dae60b5-6d85-423a-bd4b-c4cf3f7f46f0",
                "name": "Wei-Hsiang Liao",
                "collaborators": [
                    "Yuki Mitsufuji",
                    "Chieh-Hsin Lai",
                    "Naoki Murata",
                    "K. Cheuk",
                    "Woosung Choi",
                    "Yuhta Takida",
                    "Marco A. Mart'inez-Ram'irez",
                    "Toshimitsu Uesaka",
                    "Yukara Ikemiya",
                    "Simon Dixon"
                ],
                "domain": [
                    "Music Generation",
                    "Audio Processing",
                    "Deep Learning",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "DisMix: Disentangling Mixtures of Musical Instruments for Source-level Pitch and Timbre Manipulation",
                        "abstract": "Existing work on pitch and timbre disentanglement has been mostly focused on single-instrument music audio, excluding the cases where multiple instruments are presented. To fill the gap, we propose DisMix, a generative framework in which the pitch and timbre representations act as modular building blocks for constructing the melody and instrument of a source, and the collection of which forms a set of per-instrument latent representations underlying the observed mixture. By manipulating the representations, our model samples mixtures with novel combinations of pitch and timbre of the constituent instruments. We can jointly learn the disentangled pitch-timbre representations and a latent diffusion transformer that reconstructs the mixture conditioned on the set of source-level representations. We evaluate the model using both a simple dataset of isolated chords and a realistic four-part chorales in the style of J.S. Bach, identify the key components for the success of disentanglement, and demonstrate the application of mixture transformation based on source-level attribute manipulation."
                    },
                    {
                        "title": "PaGoDA: Progressive Growing of a One-Step Generator from a Low-Resolution Diffusion Teacher",
                        "abstract": "The diffusion model performs remarkable in generating high-dimensional content but is computationally intensive, especially during training. We propose Progressive Growing of Diffusion Autoencoder (PaGoDA), a novel pipeline that reduces the training costs through three stages: training diffusion on downsampled data, distilling the pretrained diffusion, and progressive super-resolution. With the proposed pipeline, PaGoDA achieves a $64\\times$ reduced cost in training its diffusion model on 8x downsampled data; while at the inference, with the single-step, it performs state-of-the-art on ImageNet across all resolutions from 64x64 to 512x512, and text-to-image. PaGoDA's pipeline can be applied directly in the latent space, adding compression alongside the pre-trained autoencoder in Latent Diffusion Models (e.g., Stable Diffusion). The code is available at https://github.com/sony/pagoda."
                    },
                    {
                        "title": "Instruct-MusicGen: Unlocking Text-to-Music Editing for Music Language Models via Instruction Tuning",
                        "abstract": "Recent advances in text-to-music editing, which employ text queries to modify music (e.g.\\ by changing its style or adjusting instrumental components), present unique challenges and opportunities for AI-assisted music creation. Previous approaches in this domain have been constrained by the necessity to train specific editing models from scratch, which is both resource-intensive and inefficient; other research uses large language models to predict edited music, resulting in imprecise audio reconstruction. To Combine the strengths and address these limitations, we introduce Instruct-MusicGen, a novel approach that finetunes a pretrained MusicGen model to efficiently follow editing instructions such as adding, removing, or separating stems. Our approach involves a modification of the original MusicGen architecture by incorporating a text fusion module and an audio fusion module, which allow the model to process instruction texts and audio inputs concurrently and yield the desired edited music. Remarkably, Instruct-MusicGen only introduces 8% new parameters to the original MusicGen model and only trains for 5K steps, yet it achieves superior performance across all tasks compared to existing baselines, and demonstrates performance comparable to the models trained for specific tasks. This advancement not only enhances the efficiency of text-to-music editing but also broadens the applicability of music language models in dynamic music production environments."
                    },
                    {
                        "title": "SilentCipher: Deep Audio Watermarking",
                        "abstract": "In the realm of audio watermarking, it is challenging to simultaneously encode imperceptible messages while enhancing the message capacity and robustness. Although recent advancements in deep learning-based methods bolster the message capacity and robustness over traditional methods, the encoded messages introduce audible artefacts that restricts their usage in professional settings. In this study, we introduce three key innovations. Firstly, our work is the first deep learning-based model to integrate psychoacoustic model based thresholding to achieve imperceptible watermarks. Secondly, we introduce psuedo-differentiable compression layers, enhancing the robustness of our watermarking algorithm. Lastly, we introduce a method to eliminate the need for perceptual losses, enabling us to achieve SOTA in both robustness as well as imperceptible watermarking. Our contributions lead us to SilentCipher, a model enabling users to encode messages within audio signals sampled at 44.1kHz."
                    },
                    {
                        "title": "Towards Assessing Data Replication in Music Generation with Music Similarity Metrics on Raw Audio",
                        "abstract": "Recent advancements in music generation are raising multiple concerns about the implications of AI in creative music processes, current business models and impacts related to intellectual property management. A relevant discussion and related technical challenge is the potential replication and plagiarism of the training set in AI-generated music, which could lead to misuse of data and intellectual property rights violations. To tackle this issue, we present the Music Replication Assessment (MiRA) tool: a model-independent open evaluation method based on diverse audio music similarity metrics to assess data replication. We evaluate the ability of five metrics to identify exact replication by conducting a controlled replication experiment in different music genres using synthetic samples. Our results show that the proposed methodology can estimate exact data replication with a proportion higher than 10%. By introducing the MiRA tool, we intend to encourage the open evaluation of music-generative models by researchers, developers, and users concerning data replication, highlighting the importance of the ethical, social, legal, and economic consequences. Code and examples are available for reproducibility purposes."
                    },
                    {
                        "title": "Latent Diffusion Bridges for Unsupervised Musical Audio Timbre Transfer",
                        "abstract": "Music timbre transfer is a challenging task that involves modifying the timbral characteristics of an audio signal while preserving its melodic structure. In this paper, we propose a novel method based on dual diffusion bridges, trained using the CocoChorales Dataset, which consists of unpaired monophonic single-instrument audio data. Each diffusion model is trained on a specific instrument with a Gaussian prior. During inference, a model is designated as the source model to map the input audio to its corresponding Gaussian prior, and another model is designated as the target model to reconstruct the target audio from this Gaussian prior, thereby facilitating timbre transfer. We compare our approach against existing unsupervised timbre transfer models such as VAEGAN and Gaussian Flow Bridges (GFB). Experimental results demonstrate that our method achieves both better Fr\\'echet Audio Distance (FAD) and melody preservation, as reflected by lower pitch distances (DPD) compared to VAEGAN and GFB. Additionally, we discover that the noise level from the Gaussian prior, $\\sigma$, can be adjusted to control the degree of melody preservation and amount of timbre transferred."
                    },
                    {
                        "title": "Improving Unsupervised Clean-to-Rendered Guitar Tone Transformation Using GANs and Integrated Unaligned Clean Data",
                        "abstract": "Recent years have seen increasing interest in applying deep learning methods to the modeling of guitar amplifiers or effect pedals. Existing methods are mainly based on the supervised approach, requiring temporally-aligned data pairs of unprocessed and rendered audio. However, this approach does not scale well, due to the complicated process involved in creating the data pairs. A very recent work done by Wright et al. has explored the potential of leveraging unpaired data for training, using a generative adversarial network (GAN)-based framework. This paper extends their work by using more advanced discriminators in the GAN, and using more unpaired data for training. Specifically, drawing inspiration from recent advancements in neural vocoders, we employ in our GAN-based model for guitar amplifier modeling two sets of discriminators, one based on multi-scale discriminator (MSD) and the other multi-period discriminator (MPD). Moreover, we experiment with adding unprocessed audio signals that do not have the corresponding rendered audio of a target tone to the training data, to see how much the GAN model benefits from the unpaired data. Our experiments show that the proposed two extensions contribute to the modeling of both low-gain and high-gain guitar amplifiers."
                    },
                    {
                        "title": "GRAFX: An Open-Source Library for Audio Processing Graphs in PyTorch",
                        "abstract": "We present GRAFX, an open-source library designed for handling audio processing graphs in PyTorch. Along with various library functionalities, we describe technical details on the efficient parallel computation of input graphs, signals, and processor parameters in GPU. Then, we show its example use under a music mixing scenario, where parameters of every differentiable processor in a large graph are optimized via gradient descent. The code is available at https://github.com/sh-lee97/grafx."
                    },
                    {
                        "title": "OpenMU: Your Swiss Army Knife for Music Understanding",
                        "abstract": "We present OpenMU-Bench, a large-scale benchmark suite for addressing the data scarcity issue in training multimodal language models to understand music. To construct OpenMU-Bench, we leveraged existing datasets and bootstrapped new annotations. OpenMU-Bench also broadens the scope of music understanding by including lyrics understanding and music tool usage. Using OpenMU-Bench, we trained our music understanding model, OpenMU, with extensive ablations, demonstrating that OpenMU outperforms baseline models such as MU-Llama. Both OpenMU and OpenMU-Bench are open-sourced to facilitate future research in music understanding and to enhance creative music production efficiency."
                    },
                    {
                        "title": "LOCKEY: A Novel Approach to Model Authentication and Deepfake Tracking",
                        "abstract": "This paper presents a novel approach to deter unauthorized deepfakes and enable user tracking in generative models, even when the user has full access to the model parameters, by integrating key-based model authentication with watermarking techniques. Our method involves providing users with model parameters accompanied by a unique, user-specific key. During inference, the model is conditioned upon the key along with the standard input. A valid key results in the expected output, while an invalid key triggers a degraded output, thereby enforcing key-based model authentication. For user tracking, the model embeds the user's unique key as a watermark within the generated content, facilitating the identification of the user's ID. We demonstrate the effectiveness of our approach on two types of models, audio codecs and vocoders, utilizing the SilentCipher watermarking method. Additionally, we assess the robustness of the embedded watermarks against various distortions, validating their reliability in various scenarios."
                    },
                    {
                        "title": "VRVQ: Variable Bitrate Residual Vector Quantization for Audio Compression",
                        "abstract": "Recent state-of-the-art neural audio compression models have progressively adopted residual vector quantization (RVQ). Despite this success, these models employ a fixed number of codebooks per frame, which can be suboptimal in terms of rate-distortion tradeoff, particularly in scenarios with simple input audio, such as silence. To address this limitation, we propose variable bitrate RVQ (VRVQ) for audio codecs, which allows for more efficient coding by adapting the number of codebooks used per frame. Furthermore, we propose a gradient estimation method for the non-differentiable masking operation that transforms from the importance map to the binary importance mask, improving model training via a straight-through estimator. We demonstrate that the proposed training framework achieves superior results compared to the baseline method and shows further improvement when applied to the current state-of-the-art codec."
                    },
                    {
                        "title": "MR-MT3: Memory Retaining Multi-Track Music Transcription to Mitigate Instrument Leakage",
                        "abstract": "This paper presents enhancements to the MT3 model, a state-of-the-art (SOTA) token-based multi-instrument automatic music transcription (AMT) model. Despite SOTA performance, MT3 has the issue of instrument leakage, where transcriptions are fragmented across different instruments. To mitigate this, we propose MR-MT3, with enhancements including a memory retention mechanism, prior token sampling, and token shuffling are proposed. These methods are evaluated on the Slakh2100 dataset, demonstrating improved onset F1 scores and reduced instrument leakage. In addition to the conventional multi-instrument transcription F1 score, new metrics such as the instrument leakage ratio and the instrument detection F1 score are introduced for a more comprehensive assessment of transcription quality. The study also explores the issue of domain overfitting by evaluating MT3 on single-instrument monophonic datasets such as ComMU and NSynth. The findings, along with the source code, are shared to facilitate future work aimed at refining token-based multi-instrument AMT models."
                    },
                    {
                        "title": "Searching For Music Mixing Graphs: A Pruning Approach",
                        "abstract": "Music mixing is compositional -- experts combine multiple audio processors to achieve a cohesive mix from dry source tracks. We propose a method to reverse engineer this process from the input and output audio. First, we create a mixing console that applies all available processors to every chain. Then, after the initial console parameter optimization, we alternate between removing redundant processors and fine-tuning. We achieve this through differentiable implementation of both processors and pruning. Consequently, we find a sparse mixing graph that achieves nearly identical matching quality of the full mixing console. We apply this procedure to dry-mix pairs from various datasets and collect graphs that also can be used to train neural networks for music mixing applications."
                    },
                    {
                        "title": "Human-Feedback Efficient Reinforcement Learning for Online Diffusion Model Finetuning",
                        "abstract": "Controllable generation through Stable Diffusion (SD) fine-tuning aims to improve fidelity, safety, and alignment with human guidance. Existing reinforcement learning from human feedback methods usually rely on predefined heuristic reward functions or pretrained reward models built on large-scale datasets, limiting their applicability to scenarios where collecting such data is costly or difficult. To effectively and efficiently utilize human feedback, we develop a framework, HERO, which leverages online human feedback collected on the fly during model learning. Specifically, HERO features two key mechanisms: (1) Feedback-Aligned Representation Learning, an online training method that captures human feedback and provides informative learning signals for fine-tuning, and (2) Feedback-Guided Image Generation, which involves generating images from SD's refined initialization samples, enabling faster convergence towards the evaluator's intent. We demonstrate that HERO is 4x more efficient in online feedback for body part anomaly correction compared to the best existing method. Additionally, experiments show that HERO can effectively handle tasks like reasoning, counting, personalization, and reducing NSFW content with only 0.5K online feedback."
                    },
                    {
                        "title": "MusicMagus: Zero-Shot Text-to-Music Editing via Diffusion Models",
                        "abstract": "Recent advances in text-to-music generation models have opened new avenues in musical creativity. However, the task of editing these generated music remains a significant challenge. This paper introduces a novel approach to edit music generated by such models, enabling the modification of specific attributes, such as genre, mood, and instrument, while maintaining other aspects unchanged. Our method transforms text editing to the latent space manipulation, and adds an additional constraint to enforce consistency. It seamlessly integrates with existing pretrained text-to-music diffusion models without requiring additional training. Experimental results demonstrate superior performance over both zero-shot and certain supervised baselines in style and timbre transfer evaluations. We also show the practical applicability of our approach in real-world music editing scenarios."
                    },
                    {
                        "title": "Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion",
                        "abstract": "Consistency Models (CM) (Song et al., 2023) accelerate score-based diffusion model sampling at the cost of sample quality but lack a natural way to trade-off quality for speed. To address this limitation, we propose Consistency Trajectory Model (CTM), a generalization encompassing CM and score-based models as special cases. CTM trains a single neural network that can -- in a single forward pass -- output scores (i.e., gradients of log-density) and enables unrestricted traversal between any initial and final time along the Probability Flow Ordinary Differential Equation (ODE) in a diffusion process. CTM enables the efficient combination of adversarial training and denoising score matching loss to enhance performance and achieves new state-of-the-art FIDs for single-step diffusion model sampling on CIFAR-10 (FID 1.73) and ImageNet at 64x64 resolution (FID 1.92). CTM also enables a new family of sampling schemes, both deterministic and stochastic, involving long jumps along the ODE solution trajectories. It consistently improves sample quality as computational budgets increase, avoiding the degradation seen in CM. Furthermore, unlike CM, CTM's access to the score function can streamline the adoption of established controllable/conditional generation methods from the diffusion community. This access also enables the computation of likelihood. The code is available at https://github.com/sony/ctm."
                    },
                    {
                        "title": "HQ-VAE: Hierarchical Discrete Representation Learning with Variational Bayes",
                        "abstract": "Vector quantization (VQ) is a technique to deterministically learn features with discrete codebook representations. It is commonly performed with a variational autoencoding model, VQ-VAE, which can be further extended to hierarchical structures for making high-fidelity reconstructions. However, such hierarchical extensions of VQ-VAE often suffer from the codebook/layer collapse issue, where the codebook is not efficiently used to express the data, and hence degrades reconstruction accuracy. To mitigate this problem, we propose a novel unified framework to stochastically learn hierarchical discrete representation on the basis of the variational Bayes framework, called hierarchically quantized variational autoencoder (HQ-VAE). HQ-VAE naturally generalizes the hierarchical variants of VQ-VAE, such as VQ-VAE-2 and residual-quantized VAE (RQ-VAE), and provides them with a Bayesian training scheme. Our comprehensive experiments on image datasets show that HQ-VAE enhances codebook usage and improves reconstruction performance. We also validated HQ-VAE in terms of its applicability to a different modality with an audio dataset."
                    },
                    {
                        "title": "Timbre-Trap: A Low-Resource Framework for Instrument-Agnostic Music Transcription",
                        "abstract": "In recent years, research on music transcription has focused mainly on architecture design and instrument-specific data acquisition. With the lack of availability of diverse datasets, progress is often limited to solo-instrument tasks such as piano transcription. Several works have explored multi-instrument transcription as a means to bolster the performance of models on low-resource tasks, but these methods face the same data availability issues. We propose Timbre-Trap, a novel framework which unifies music transcription and audio reconstruction by exploiting the strong separability between pitch and timbre. We train a single autoencoder to simultaneously estimate pitch salience and reconstruct complex spectral coefficients, selecting between either output during the decoding stage via a simple switch mechanism. In this way, the model learns to produce coefficients corresponding to timbre-less audio, which can be interpreted as pitch salience. We demonstrate that the framework leads to performance comparable to state-of-the-art instrument-agnostic transcription methods, while only requiring a small amount of annotated data."
                    },
                    {
                        "title": "VRDMG: Vocal Restoration via Diffusion Posterior Sampling with Multiple Guidance",
                        "abstract": "Restoring degraded music signals is essential to enhance audio quality for downstream music manipulation. Recent diffusion-based music restoration methods have demonstrated impressive performance, and among them, diffusion posterior sampling (DPS) stands out given its intrinsic properties, making it versatile across various restoration tasks. In this paper, we identify that there are potential issues which will degrade current DPS-based methods\u2019 performance and introduce the way to mitigate the issues inspired by diverse diffusion guidance techniques including the RePaint (RP) strategy and the Pseudoinverse-Guided Diffusion Models (\u03a0GDM). We demonstrate our methods for the vocal declipping and bandwidth extension tasks under various levels of distortion and cutoff frequency, respectively. In both tasks, our methods outperform the current DPS-based music restoration benchmarks. We refer to http://carlosholivan.github.io/demos/audio-restoration-2023.html for examples of the restored audio samples."
                    }
                ]
            },
            "7148b270-86a5-42b7-9efb-364f8025bfc1": {
                "pk": "7148b270-86a5-42b7-9efb-364f8025bfc1",
                "name": "Yuki Mitsufuji",
                "collaborators": [
                    "Yuhta Takida",
                    "Mengjie Zhao",
                    "Chieh-Hsin Lai",
                    "Naoki Murata",
                    "Wei-Hsiang Liao",
                    "Takashi Shibuya",
                    "Shiqi Yang",
                    "Hiromi Wakaki",
                    "Toshimitsu Uesaka",
                    "Masato Ishii"
                ],
                "domain": [
                    "Generative Models",
                    "Audio-Visual Processing",
                    "Contrastive Learning",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Weighted Point Cloud Embedding for Multimodal Contrastive Learning Toward Optimal Similarity Metric",
                        "abstract": "In typical multimodal contrastive learning, such as CLIP, encoders produce one point in the latent representation space for each input. However, one-point representation has difficulty in capturing the relationship and the similarity structure of a huge amount of instances in the real world. For richer classes of the similarity, we propose the use of weighted point clouds, namely, sets of pairs of weight and vector, as representations of instances. In this work, we theoretically show the benefit of our proposed method through a new understanding of the contrastive loss of CLIP, which we call symmetric InfoNCE. We clarify that the optimal similarity that minimizes symmetric InfoNCE is the pointwise mutual information, and show an upper bound of excess risk on downstream classification tasks of representations that achieve the optimal similarity. In addition, we show that our proposed similarity based on weighted point clouds consistently achieves the optimal similarity. To verify the effectiveness of our proposed method, we demonstrate pretraining of text-image representation models and classification tasks on common benchmarks."
                    },
                    {
                        "title": "DisMix: Disentangling Mixtures of Musical Instruments for Source-level Pitch and Timbre Manipulation",
                        "abstract": "Existing work on pitch and timbre disentanglement has been mostly focused on single-instrument music audio, excluding the cases where multiple instruments are presented. To fill the gap, we propose DisMix, a generative framework in which the pitch and timbre representations act as modular building blocks for constructing the melody and instrument of a source, and the collection of which forms a set of per-instrument latent representations underlying the observed mixture. By manipulating the representations, our model samples mixtures with novel combinations of pitch and timbre of the constituent instruments. We can jointly learn the disentangled pitch-timbre representations and a latent diffusion transformer that reconstructs the mixture conditioned on the set of source-level representations. We evaluate the model using both a simple dataset of isolated chords and a realistic four-part chorales in the style of J.S. Bach, identify the key components for the success of disentanglement, and demonstrate the application of mixture transformation based on source-level attribute manipulation."
                    },
                    {
                        "title": "Discriminator-Guided Cooperative Diffusion for Joint Audio and Video Generation",
                        "abstract": "In this study, we aim to construct an audio-video generative model with minimal computational cost by leveraging pre-trained single-modal generative models for audio and video. To achieve this, we propose a novel method that guides each single-modal model to cooperatively generate well-aligned samples across modalities. Specifically, given two pre-trained base diffusion models, we train a lightweight joint guidance module to adjust scores separately estimated by the base models to match the score of joint distribution over audio and video. We theoretically show that this guidance can be computed through the gradient of the optimal discriminator distinguishing real audio-video pairs from fake ones independently generated by the base models. On the basis of this analysis, we construct the joint guidance module by training this discriminator. Additionally, we adopt a loss function to make the gradient of the discriminator work as a noise estimator, as in standard diffusion models, stabilizing the gradient of the discriminator. Empirical evaluations on several benchmark datasets demonstrate that our method improves both single-modal fidelity and multi-modal alignment with a relatively small number of parameters."
                    },
                    {
                        "title": "PaGoDA: Progressive Growing of a One-Step Generator from a Low-Resolution Diffusion Teacher",
                        "abstract": "The diffusion model performs remarkable in generating high-dimensional content but is computationally intensive, especially during training. We propose Progressive Growing of Diffusion Autoencoder (PaGoDA), a novel pipeline that reduces the training costs through three stages: training diffusion on downsampled data, distilling the pretrained diffusion, and progressive super-resolution. With the proposed pipeline, PaGoDA achieves a $64\\times$ reduced cost in training its diffusion model on 8x downsampled data; while at the inference, with the single-step, it performs state-of-the-art on ImageNet across all resolutions from 64x64 to 512x512, and text-to-image. PaGoDA's pipeline can be applied directly in the latent space, adding compression alongside the pre-trained autoencoder in Latent Diffusion Models (e.g., Stable Diffusion). The code is available at https://github.com/sony/pagoda."
                    },
                    {
                        "title": "GLOV: Guided Large Language Models as Implicit Optimizers for Vision Language Models",
                        "abstract": "In this work, we propose a novel method (GLOV) enabling Large Language Models (LLMs) to act as implicit Optimizers for Vision-Langugage Models (VLMs) to enhance downstream vision tasks. Our GLOV meta-prompts an LLM with the downstream task description, querying it for suitable VLM prompts (e.g., for zero-shot classification with CLIP). These prompts are ranked according to a purity measure obtained through a fitness function. In each respective optimization step, the ranked prompts are fed as in-context examples (with their accuracies) to equip the LLM with the knowledge of the type of text prompts preferred by the downstream VLM. Furthermore, we also explicitly steer the LLM generation process in each optimization step by specifically adding an offset difference vector of the embeddings from the positive and negative solutions found by the LLM, in previous optimization steps, to the intermediate layer of the network for the next generation step. This offset vector steers the LLM generation toward the type of language preferred by the downstream VLM, resulting in enhanced performance on the downstream vision tasks. We comprehensively evaluate our GLOV on 16 diverse datasets using two families of VLMs, i.e., dual-encoder (e.g., CLIP) and encoder-decoder (e.g., LLaVa) models -- showing that the discovered solutions can enhance the recognition performance by up to 15.0% and 57.5% (3.8% and 21.6% on average) for these models."
                    },
                    {
                        "title": "$\\textit{Jump Your Steps}$: Optimizing Sampling Schedule of Discrete Diffusion Models",
                        "abstract": "Diffusion models have seen notable success in continuous domains, leading to the development of discrete diffusion models (DDMs) for discrete variables. Despite recent advances, DDMs face the challenge of slow sampling speeds. While parallel sampling methods like $\\tau$-leaping accelerate this process, they introduce $\\textit{Compounding Decoding Error}$ (CDE), where discrepancies arise between the true distribution and the approximation from parallel token generation, leading to degraded sample quality. In this work, we present $\\textit{Jump Your Steps}$ (JYS), a novel approach that optimizes the allocation of discrete sampling timesteps by minimizing CDE without extra computational cost. More precisely, we derive a practical upper bound on CDE and propose an efficient algorithm for searching for the optimal sampling schedule. Extensive experiments across image, music, and text generation show that JYS significantly improves sampling quality, establishing it as a versatile framework for enhancing DDM performance for fast sampling."
                    },
                    {
                        "title": "A Simple but Strong Baseline for Sounding Video Generation: Effective Adaptation of Audio and Video Diffusion Models for Joint Generation",
                        "abstract": "In this work, we build a simple but strong baseline for sounding video generation. Given base diffusion models for audio and video, we integrate them with additional modules into a single model and train it to make the model jointly generate audio and video. To enhance alignment between audio-video pairs, we introduce two novel mechanisms in our model. The first one is timestep adjustment, which provides different timestep information to each base model. It is designed to align how samples are generated along with timesteps across modalities. The second one is a new design of the additional modules, termed Cross-Modal Conditioning as Positional Encoding (CMC-PE). In CMC-PE, cross-modal information is embedded as if it represents temporal position information, and the embeddings are fed into the model like positional encoding. Compared with the popular cross-attention mechanism, CMC-PE provides a better inductive bias for temporal alignment in the generated data. Experimental results validate the effectiveness of the two newly introduced mechanisms and also demonstrate that our method outperforms existing methods."
                    },
                    {
                        "title": "Instruct-MusicGen: Unlocking Text-to-Music Editing for Music Language Models via Instruction Tuning",
                        "abstract": "Recent advances in text-to-music editing, which employ text queries to modify music (e.g.\\ by changing its style or adjusting instrumental components), present unique challenges and opportunities for AI-assisted music creation. Previous approaches in this domain have been constrained by the necessity to train specific editing models from scratch, which is both resource-intensive and inefficient; other research uses large language models to predict edited music, resulting in imprecise audio reconstruction. To Combine the strengths and address these limitations, we introduce Instruct-MusicGen, a novel approach that finetunes a pretrained MusicGen model to efficiently follow editing instructions such as adding, removing, or separating stems. Our approach involves a modification of the original MusicGen architecture by incorporating a text fusion module and an audio fusion module, which allow the model to process instruction texts and audio inputs concurrently and yield the desired edited music. Remarkably, Instruct-MusicGen only introduces 8% new parameters to the original MusicGen model and only trains for 5K steps, yet it achieves superior performance across all tasks compared to existing baselines, and demonstrates performance comparable to the models trained for specific tasks. This advancement not only enhances the efficiency of text-to-music editing but also broadens the applicability of music language models in dynamic music production environments."
                    },
                    {
                        "title": "Mitigating Embedding Collapse in Diffusion Models for Categorical Data",
                        "abstract": "Latent diffusion models have enabled continuous-state diffusion models to handle a variety of datasets, including categorical data. However, most methods rely on fixed pretrained embeddings, limiting the benefits of joint training with the diffusion model. While jointly learning the embedding (via reconstruction loss) and the latent diffusion model (via score matching loss) could enhance performance, our analysis shows that end-to-end training risks embedding collapse, degrading generation quality. To address this issue, we introduce CATDM, a continuous diffusion framework within the embedding space that stabilizes training. We propose a novel objective combining the joint embedding-diffusion variational lower bound with a Consistency-Matching (CM) regularizer, alongside a shifted cosine noise schedule and random dropping strategy. The CM regularizer ensures the recovery of the true data distribution. Experiments on benchmarks show that CATDM mitigates embedding collapse, yielding superior results on FFHQ, LSUN Churches, and LSUN Bedrooms. In particular, CATDM achieves an FID of 6.81 on ImageNet $256\\times256$ with 50 steps. It outperforms non-autoregressive models in machine translation and is on a par with previous methods in text generation."
                    },
                    {
                        "title": "ComperDial: Commonsense Persona-grounded Dialogue Dataset and Benchmark",
                        "abstract": "We propose a new benchmark, ComperDial, which facilitates the training and evaluation of evaluation metrics for open-domain dialogue systems. ComperDial consists of human-scored responses for 10,395 dialogue turns in 1,485 conversations collected from 99 dialogue agents submitted to the Commonsense Persona-grounded Dialogue (CPD) challenge. As a result, for any dialogue, our benchmark includes multiple diverse responses with variety of characteristics to ensure more robust evaluation of learned dialogue metrics. In addition to single-turn response scores, ComperDial also contains dialogue-level human-annotated scores, enabling joint assessment of multi-turn model responses throughout a dialogue. Finally, building off ComperDial, we devise a new automatic evaluation metric to measure the general similarity of model-generated dialogues to human conversations. Our experimental results demonstrate that our novel metric, CPDScore is more correlated with human judgments than existing metrics. We release both ComperDial and CPDScore to the community to accelerate development of automatic evaluation metrics for open-domain dialogue systems."
                    },
                    {
                        "title": "Mining Your Own Secrets: Diffusion Classifier Scores for Continual Personalization of Text-to-Image Diffusion Models",
                        "abstract": "Personalized text-to-image diffusion models have grown popular for their ability to efficiently acquire a new concept from user-defined text descriptions and a few images. However, in the real world, a user may wish to personalize a model on multiple concepts but one at a time, with no access to the data from previous concepts due to storage/privacy concerns. When faced with this continual learning (CL) setup, most personalization methods fail to find a balance between acquiring new concepts and retaining previous ones -- a challenge that continual personalization (CP) aims to solve. Inspired by the successful CL methods that rely on class-specific information for regularization, we resort to the inherent class-conditioned density estimates, also known as diffusion classifier (DC) scores, for continual personalization of text-to-image diffusion models. Namely, we propose using DC scores for regularizing the parameter-space and function-space of text-to-image diffusion models, to achieve continual personalization. Using several diverse evaluation setups, datasets, and metrics, we show that our proposed regularization-based CP methods outperform the state-of-the-art C-LoRA, and other baselines. Finally, by operating in the replay-free CL setup and on low-rank adapters, our method incurs zero storage and parameter overhead, respectively, over the state-of-the-art."
                    },
                    {
                        "title": "Automated Black-box Prompt Engineering for Personalized Text-to-Image Generation",
                        "abstract": "Prompt engineering is effective for controlling the output of text-to-image (T2I) generative models, but it is also laborious due to the need for manually crafted prompts. This challenge has spurred the development of algorithms for automated prompt generation. However, these methods often struggle with transferability across T2I models, require white-box access to the underlying model, and produce non-intuitive prompts. In this work, we introduce PRISM, an algorithm that automatically identifies human-interpretable and transferable prompts that can effectively generate desired concepts given only black-box access to T2I models. Inspired by large language model (LLM) jailbreaking, PRISM leverages the in-context learning ability of LLMs to iteratively refine the candidate prompts distribution for given reference images. Our experiments demonstrate the versatility and effectiveness of PRISM in generating accurate prompts for objects, styles and images across multiple T2I models, including Stable Diffusion, DALL-E, and Midjourney."
                    },
                    {
                        "title": "Improving Unsupervised Clean-to-Rendered Guitar Tone Transformation Using GANs and Integrated Unaligned Clean Data",
                        "abstract": "Recent years have seen increasing interest in applying deep learning methods to the modeling of guitar amplifiers or effect pedals. Existing methods are mainly based on the supervised approach, requiring temporally-aligned data pairs of unprocessed and rendered audio. However, this approach does not scale well, due to the complicated process involved in creating the data pairs. A very recent work done by Wright et al. has explored the potential of leveraging unpaired data for training, using a generative adversarial network (GAN)-based framework. This paper extends their work by using more advanced discriminators in the GAN, and using more unpaired data for training. Specifically, drawing inspiration from recent advancements in neural vocoders, we employ in our GAN-based model for guitar amplifier modeling two sets of discriminators, one based on multi-scale discriminator (MSD) and the other multi-period discriminator (MPD). Moreover, we experiment with adding unprocessed audio signals that do not have the corresponding rendered audio of a target tone to the training data, to see how much the GAN model benefits from the unpaired data. Our experiments show that the proposed two extensions contribute to the modeling of both low-gain and high-gain guitar amplifiers."
                    },
                    {
                        "title": "SoundCTM: Uniting Score-based and Consistency Models for Text-to-Sound Generation",
                        "abstract": "Sound content is an indispensable element for multimedia works such as video games, music, and films. Recent high-quality diffusion-based sound generation models can serve as valuable tools for the creators. However, despite producing high-quality sounds, these models often suffer from slow inference speeds. This drawback burdens creators, who typically refine their sounds through trial and error to align them with their artistic intentions. To address this issue, we introduce Sound Consistency Trajectory Models (SoundCTM). Our model enables flexible transitioning between high-quality 1-step sound generation and superior sound quality through multi-step generation. This allows creators to initially control sounds with 1-step samples before refining them through multi-step generation. While CTM fundamentally achieves flexible 1-step and multi-step generation, its impressive performance heavily depends on an additional pretrained feature extractor and an adversarial loss, which are expensive to train and not always available in other domains. Thus, we reframe CTM's training framework and introduce a novel feature distance by utilizing the teacher's network for a distillation loss. Additionally, while distilling classifier-free guided trajectories, we train conditional and unconditional student models simultaneously and interpolate between these models during inference. We also propose training-free controllable frameworks for SoundCTM, leveraging its flexible sampling capability. SoundCTM achieves both promising 1-step and multi-step real-time sound generation without using any extra off-the-shelf networks. Furthermore, we demonstrate SoundCTM's capability of controllable sound generation in a training-free manner. Our codes, pretrained models, and audio samples are available at https://github.com/sony/soundctm."
                    },
                    {
                        "title": "Visual Echoes: A Simple Unified Transformer for Audio-Visual Generation",
                        "abstract": "In recent years, with the realistic generation results and a wide range of personalized applications, diffusion-based generative models gain huge attention in both visual and audio generation areas. Compared to the considerable advancements of text2image or text2audio generation, research in audio2visual or visual2audio generation has been relatively slow. The recent audio-visual generation methods usually resort to huge large language model or composable diffusion models. Instead of designing another giant model for audio-visual generation, in this paper we take a step back showing a simple and lightweight generative transformer, which is not fully investigated in multi-modal generation, can achieve excellent results on image2audio generation. The transformer operates in the discrete audio and visual Vector-Quantized GAN space, and is trained in the mask denoising manner. After training, the classifier-free guidance could be deployed off-the-shelf achieving better performance, without any extra training or modification. Since the transformer model is modality symmetrical, it could also be directly deployed for audio2image generation and co-generation. In the experiments, we show that our simple method surpasses recent image2audio generation methods. Generated audio samples can be found at https://docs.google.com/presentation/d/1ZtC0SeblKkut4XJcRaDsSTuCRIXB3ypxmSi7HTY3IyQ/"
                    },
                    {
                        "title": "OpenMU: Your Swiss Army Knife for Music Understanding",
                        "abstract": "We present OpenMU-Bench, a large-scale benchmark suite for addressing the data scarcity issue in training multimodal language models to understand music. To construct OpenMU-Bench, we leveraged existing datasets and bootstrapped new annotations. OpenMU-Bench also broadens the scope of music understanding by including lyrics understanding and music tool usage. Using OpenMU-Bench, we trained our music understanding model, OpenMU, with extensive ablations, demonstrating that OpenMU outperforms baseline models such as MU-Llama. Both OpenMU and OpenMU-Bench are open-sourced to facilitate future research in music understanding and to enhance creative music production efficiency."
                    },
                    {
                        "title": "Bellman Diffusion: Generative Modeling as Learning a Linear Operator in the Distribution Space",
                        "abstract": "Deep Generative Models (DGMs), including Energy-Based Models (EBMs) and Score-based Generative Models (SGMs), have advanced high-fidelity data generation and complex continuous distribution approximation. However, their application in Markov Decision Processes (MDPs), particularly in distributional Reinforcement Learning (RL), remains underexplored, with conventional histogram-based methods dominating the field. This paper rigorously highlights that this application gap is caused by the nonlinearity of modern DGMs, which conflicts with the linearity required by the Bellman equation in MDPs. For instance, EBMs involve nonlinear operations such as exponentiating energy functions and normalizing constants. To address this, we introduce Bellman Diffusion, a novel DGM framework that maintains linearity in MDPs through gradient and scalar field modeling. With divergence-based training techniques to optimize neural network proxies and a new type of stochastic differential equation (SDE) for sampling, Bellman Diffusion is guaranteed to converge to the target distribution. Our empirical results show that Bellman Diffusion achieves accurate field estimations and is a capable image generator, converging 1.5x faster than the traditional histogram-based baseline in distributional RL tasks. This work enables the effective integration of DGMs into MDP applications, unlocking new avenues for advanced decision-making frameworks."
                    },
                    {
                        "title": "Distillation of Discrete Diffusion through Dimensional Correlations",
                        "abstract": "Diffusion models have demonstrated exceptional performances in various fields of generative modeling. While they often outperform competitors including VAEs and GANs in sample quality and diversity, they suffer from slow sampling speed due to their iterative nature. Recently, distillation techniques and consistency models are mitigating this issue in continuous domains, but discrete diffusion models have some specific challenges towards faster generation. Most notably, in the current literature, correlations between different dimensions (pixels, locations) are ignored, both by its modeling and loss functions, due to computational limitations. In this paper, we propose\"mixture\"models in discrete diffusion that are capable of treating dimensional correlations while remaining scalable, and we provide a set of loss functions for distilling the iterations of existing models. Two primary theoretical insights underpin our approach: first, that dimensionally independent models can well approximate the data distribution if they are allowed to conduct many sampling steps, and second, that our loss functions enables mixture models to distill such many-step conventional models into just a few steps by learning the dimensional correlations. We empirically demonstrate that our proposed method for discrete diffusions work in practice, by distilling a continuous-time discrete diffusion model pretrained on the CIFAR-10 dataset."
                    },
                    {
                        "title": "DiffuCOMET: Contextual Commonsense Knowledge Diffusion",
                        "abstract": "Inferring contextually-relevant and diverse commonsense to understand narratives remains challenging for knowledge models. In this work, we develop a series of knowledge models, DiffuCOMET, that leverage diffusion to learn to reconstruct the implicit semantic connections between narrative contexts and relevant commonsense knowledge. Across multiple diffusion steps, our method progressively refines a representation of commonsense facts that is anchored to a narrative, producing contextually-relevant and diverse commonsense inferences for an input context. To evaluate DiffuCOMET, we introduce new metrics for commonsense inference that more closely measure knowledge diversity and contextual relevance. Our results on two different benchmarks, ComFact and WebNLG+, show that knowledge generated by DiffuCOMET achieves a better trade-off between commonsense diversity, contextual relevance and alignment to known gold references, compared to baseline knowledge models."
                    },
                    {
                        "title": "SpecMaskGIT: Masked Generative Modeling of Audio Spectrograms for Efficient Audio Synthesis and Beyond",
                        "abstract": "Recent advances in generative models that iteratively synthesize audio clips sparked great success to text-to-audio synthesis (TTA), but with the cost of slow synthesis speed and heavy computation. Although there have been attempts to accelerate the iterative procedure, high-quality TTA systems remain inefficient due to hundreds of iterations required in the inference phase and large amount of model parameters. To address the challenges, we propose SpecMaskGIT, a light-weighted, efficient yet effective TTA model based on the masked generative modeling of spectrograms. First, SpecMaskGIT synthesizes a realistic 10s audio clip by less than 16 iterations, an order-of-magnitude less than previous iterative TTA methods. As a discrete model, SpecMaskGIT outperforms larger VQ-Diffusion and auto-regressive models in the TTA benchmark, while being real-time with only 4 CPU cores or even 30x faster with a GPU. Next, built upon a latent space of Mel-spectrogram, SpecMaskGIT has a wider range of applications (e.g., the zero-shot bandwidth extension) than similar methods built on the latent wave domain. Moreover, we interpret SpecMaskGIT as a generative extension to previous discriminative audio masked Transformers, and shed light on its audio representation learning potential. We hope our work inspires the exploration of masked audio modeling toward further diverse scenarios."
                    }
                ]
            },
            "a3c33109-1954-438c-bd8b-b780972d445f": {
                "pk": "a3c33109-1954-438c-bd8b-b780972d445f",
                "name": "J. Zico Kolter",
                "collaborators": [
                    "Zhengyang Geng",
                    "Dylan Sam",
                    "Yiding Jiang",
                    "Ashwini Pokle",
                    "Samuel Sokota",
                    "J. Williams",
                    "Rattana Pukdee",
                    "William Luo",
                    "Justin Lin",
                    "Weijian Luo"
                ],
                "domain": [
                    "Generative Models",
                    "Reinforcement Learning",
                    "Deep Learning",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "Consistency Models Made Easy",
                        "abstract": "Consistency models (CMs) offer faster sampling than traditional diffusion models, but their training is resource-intensive. For example, as of 2024, training a state-of-the-art CM on CIFAR-10 takes one week on 8 GPUs. In this work, we propose an effective scheme for training CMs that largely improves the efficiency of building such models. Specifically, by expressing CM trajectories via a particular differential equation, we argue that diffusion models can be viewed as a special case of CMs. We can thus fine-tune a consistency model starting from a pretrained diffusion model and progressively approximate the full consistency condition to stronger degrees over the training process. Our resulting method, which we term Easy Consistency Tuning (ECT), achieves vastly reduced training times while improving upon the quality of previous methods: for example, ECT achieves a 2-step FID of 2.73 on CIFAR10 within 1 hour on a single A100 GPU, matching Consistency Distillation trained for hundreds of GPU hours. Owing to this computational efficiency, we investigate the scaling laws of CMs under ECT, showing that they obey the classic power law scaling, hinting at their ability to improve efficiency and performance at larger scales. Our code (https://github.com/locuslab/ect) is publicly available, making CMs more accessible to the broader community."
                    },
                    {
                        "title": "One-Step Diffusion Distillation through Score Implicit Matching",
                        "abstract": "Despite their strong performances on many generative tasks, diffusion models require a large number of sampling steps in order to generate realistic samples. This has motivated the community to develop effective methods to distill pre-trained diffusion models into more efficient models, but these methods still typically require few-step inference or perform substantially worse than the underlying model. In this paper, we present Score Implicit Matching (SIM) a new approach to distilling pre-trained diffusion models into single-step generator models, while maintaining almost the same sample generation ability as the original model as well as being data-free with no need of training samples for distillation. The method rests upon the fact that, although the traditional score-based loss is intractable to minimize for generator models, under certain conditions we can efficiently compute the gradients for a wide class of score-based divergences between a diffusion model and a generator. SIM shows strong empirical performances for one-step generators: on the CIFAR10 dataset, it achieves an FID of 2.06 for unconditional generation and 1.96 for class-conditional generation. Moreover, by applying SIM to a leading transformer-based diffusion model, we distill a single-step generator for text-to-image (T2I) generation that attains an aesthetic score of 6.42 with no performance decline over the original multi-step counterpart, clearly outperforming the other one-step generators including SDXL-TURBO of 5.33, SDXL-LIGHTNING of 5.34 and HYPER-SDXL of 5.85. We will release this industry-ready one-step transformer-based T2I generator along with this paper."
                    },
                    {
                        "title": "Computing Low-Entropy Couplings for Large-Support Distributions",
                        "abstract": "Minimum-entropy coupling (MEC) -- the process of finding a joint distribution with minimum entropy for given marginals -- has applications in areas such as causality and steganography. However, existing algorithms are either computationally intractable for large-support distributions or limited to specific distribution types and sensitive to hyperparameter choices. This work addresses these limitations by unifying a prior family of iterative MEC (IMEC) approaches into a generalized partition-based formalism. From this framework, we derive a novel IMEC algorithm called ARIMEC, capable of handling arbitrary discrete distributions, and introduce a method to make IMEC robust to suboptimal hyperparameter settings. These innovations facilitate the application of IMEC to high-throughput steganography with language models, among other settings. Our codebase is available at https://github.com/ssokota/mec ."
                    },
                    {
                        "title": "Automated Black-box Prompt Engineering for Personalized Text-to-Image Generation",
                        "abstract": "Prompt engineering is effective for controlling the output of text-to-image (T2I) generative models, but it is also laborious due to the need for manually crafted prompts. This challenge has spurred the development of algorithms for automated prompt generation. However, these methods often struggle with transferability across T2I models, require white-box access to the underlying model, and produce non-intuitive prompts. In this work, we introduce PRISM, an algorithm that automatically identifies human-interpretable and transferable prompts that can effectively generate desired concepts given only black-box access to T2I models. Inspired by large language model (LLM) jailbreaking, PRISM leverages the in-context learning ability of LLMs to iteratively refine the candidate prompts distribution for given reference images. Our experiments demonstrate the versatility and effectiveness of PRISM in generating accurate prompts for objects, styles and images across multiple T2I models, including Stable Diffusion, DALL-E, and Midjourney."
                    },
                    {
                        "title": "FUSE-ing Language Models: Zero-Shot Adapter Discovery for Prompt Optimization Across Tokenizers",
                        "abstract": "The widespread use of large language models has resulted in a multitude of tokenizers and embedding spaces, making knowledge transfer in prompt discovery tasks difficult. In this work, we propose FUSE (Flexible Unification of Semantic Embeddings), an inexpensive approach to approximating an adapter layer that maps from one model's textual embedding space to another, even across different tokenizers. We introduce a third-order tensor-based representation of a model's embedding space that aligns semantic embeddings that have been split apart by different tokenizers, and use this representation to derive an approximation of the gradient of one model's outputs with respect to another model's embedding space. We show the efficacy of our approach via multi-objective optimization over vision-language and causal language models for image captioning and sentiment-based image captioning."
                    },
                    {
                        "title": "Massive Activations in Large Language Models",
                        "abstract": "We observe an empirical phenomenon in Large Language Models (LLMs) -- very few activations exhibit significantly larger values than others (e.g., 100,000 times larger). We call them massive activations. First, we demonstrate the widespread existence of massive activations across various LLMs and characterize their locations. Second, we find their values largely stay constant regardless of the input, and they function as indispensable bias terms in LLMs. Third, these massive activations lead to the concentration of attention probabilities to their corresponding tokens, and further, implicit bias terms in the self-attention output. Last, we also study massive activations in Vision Transformers. Code is available at https://github.com/locuslab/massive-activations."
                    },
                    {
                        "title": "Bayesian Neural Networks with Domain Knowledge Priors",
                        "abstract": "Bayesian neural networks (BNNs) have recently gained popularity due to their ability to quantify model uncertainty. However, specifying a prior for BNNs that captures relevant domain knowledge is often extremely challenging. In this work, we propose a framework for integrating general forms of domain knowledge (i.e., any knowledge that can be represented by a loss function) into a BNN prior through variational inference, while enabling computationally efficient posterior inference and sampling. Specifically, our approach results in a prior over neural network weights that assigns high probability mass to models that better align with our domain knowledge, leading to posterior samples that also exhibit this behavior. We show that BNNs using our proposed domain knowledge priors outperform those with standard priors (e.g., isotropic Gaussian, Gaussian process), successfully incorporating diverse types of prior information such as fairness, physics rules, and healthcare knowledge and achieving better predictive performance. We also present techniques for transferring the learned priors across different model architectures, demonstrating their broad utility across various settings."
                    },
                    {
                        "title": "Transformers to SSMs: Distilling Quadratic Knowledge to Subquadratic Models",
                        "abstract": "Transformer architectures have become a dominant paradigm for domains like language modeling but suffer in many inference settings due to their quadratic-time self-attention. Recently proposed subquadratic architectures, such as Mamba, have shown promise, but have been pretrained with substantially less computational resources than the strongest Transformer models. In this work, we present a method that is able to distill a pretrained Transformer architecture into alternative architectures such as state space models (SSMs). The key idea to our approach is that we can view both Transformers and SSMs as applying different forms of mixing matrices over the token sequences. We can thus progressively distill the Transformer architecture by matching different degrees of granularity in the SSM: first matching the mixing matrices themselves, then the hidden units at each block, and finally the end-to-end predictions. Our method, called MOHAWK, is able to distill a Mamba-2 variant based on the Phi-1.5 architecture (Phi-Mamba) using only 3B tokens and a hybrid version (Hybrid Phi-Mamba) using 5B tokens. Despite using less than 1% of the training data typically used to train models from scratch, Phi-Mamba boasts substantially stronger performance compared to all past open-source non-Transformer models. MOHAWK allows models like SSMs to leverage computational resources invested in training Transformer-based architectures, highlighting a new avenue for building such models."
                    },
                    {
                        "title": "AcceleratedLiNGAM: Learning Causal DAGs at the speed of GPUs",
                        "abstract": "Existing causal discovery methods based on combinatorial optimization or search are slow, prohibiting their application on large-scale datasets. In response, more recent methods attempt to address this limitation by formulating causal discovery as structure learning with continuous optimization but such approaches thus far provide no statistical guarantees. In this paper, we show that by efficiently parallelizing existing causal discovery methods, we can in fact scale them to thousands of dimensions, making them practical for substantially larger-scale problems. In particular, we parallelize the LiNGAM method, which is quadratic in the number of variables, obtaining up to a 32-fold speed-up on benchmark datasets when compared with existing sequential implementations. Specifically, we focus on the causal ordering subprocedure in DirectLiNGAM and implement GPU kernels to accelerate it. This allows us to apply DirectLiNGAM to causal inference on large-scale gene expression data with genetic interventions yielding competitive results compared with specialized continuous optimization methods, and Var-LiNGAM for causal discovery on U.S. stock data."
                    },
                    {
                        "title": "Learning with Explanation Constraints",
                        "abstract": "As larger deep learning models are hard to interpret, there has been a recent focus on generating explanations of these black-box models. In contrast, we may have apriori explanations of how models should behave. In this paper, we formalize this notion as learning from explanation constraints and provide a learning theoretic framework to analyze how such explanations can improve the learning of our models. One may naturally ask,\"When would these explanations be helpful?\"Our first key contribution addresses this question via a class of models that satisfies these explanation constraints in expectation over new data. We provide a characterization of the benefits of these models (in terms of the reduction of their Rademacher complexities) for a canonical class of explanations given by gradient information in the settings of both linear models and two layer neural networks. In addition, we provide an algorithmic solution for our framework, via a variational approximation that achieves better performance and satisfies these constraints more frequently, when compared to simpler augmented Lagrangian methods to incorporate these explanations. We demonstrate the benefits of our approach over a large array of synthetic and real-world experiments."
                    },
                    {
                        "title": "Sinkhorn-Flow: Predicting Probability Mass Flow in Dynamical Systems Using Optimal Transport",
                        "abstract": "Predicting how distributions over discrete variables vary over time is a common task in time series forecasting. But whereas most approaches focus on merely predicting the distribution at subsequent time steps, a crucial piece of information in many settings is to determine how this probability mass flows between the different elements over time. We propose a new approach to predicting such mass flow over time using optimal transport. Specifically, we propose a generic approach to predicting transport matrices in end-to-end deep learning systems, replacing the standard softmax operation with Sinkhorn iterations. We apply our approach to the task of predicting how communities will evolve over time in social network settings, and show that the approach improves substantially over alternative prediction methods. We specifically highlight results on the task of predicting faction evolution in Ukrainian parliamentary voting."
                    },
                    {
                        "title": "One-Step Diffusion Distillation via Deep Equilibrium Models",
                        "abstract": "Diffusion models excel at producing high-quality samples but naively require hundreds of iterations, prompting multiple attempts to distill the generation process into a faster network. However, many existing approaches suffer from a variety of challenges: the process for distillation training can be complex, often requiring multiple training stages, and the resulting models perform poorly when utilized in single-step generative applications. In this paper, we introduce a simple yet effective means of distilling diffusion models directly from initial noise to the resulting image. Of particular importance to our approach is to leverage a new Deep Equilibrium (DEQ) model as the distilled architecture: the Generative Equilibrium Transformer (GET). Our method enables fully offline training with just noise/image pairs from the diffusion model while achieving superior performance compared to existing one-step methods on comparable training budgets. We demonstrate that the DEQ architecture is crucial to this capability, as GET matches a $5\\times$ larger ViT in terms of FID scores while striking a critical balance of computational cost and image quality. Code, checkpoints, and datasets are available."
                    },
                    {
                        "title": "Abstracting Imperfect Information Away from Two-Player Zero-Sum Games",
                        "abstract": "In their seminal work, Nayyar et al. (2013) showed that imperfect information can be abstracted away from common-payoff games by having players publicly announce their policies as they play. This insight underpins sound solvers and decision-time planning algorithms for common-payoff games. Unfortunately, a naive application of the same insight to two-player zero-sum games fails because Nash equilibria of the game with public policy announcements may not correspond to Nash equilibria of the original game. As a consequence, existing sound decision-time planning algorithms require complicated additional mechanisms that have unappealing properties. The main contribution of this work is showing that certain regularized equilibria do not possess the aforementioned non-correspondence problem -- thus, computing them can be treated as perfect-information problems. Because these regularized equilibria can be made arbitrarily close to Nash equilibria, our result opens the door to a new perspective to solving two-player zero-sum games and yields a simplified framework for decision-time planning in two-player zero-sum games, void of the unappealing properties that plague existing decision-time planning approaches."
                    },
                    {
                        "title": "On the Importance of Exploration for Generalization in Reinforcement Learning",
                        "abstract": "Existing approaches for improving generalization in deep reinforcement learning (RL) have mostly focused on representation learning, neglecting RL-specific aspects such as exploration. We hypothesize that the agent's exploration strategy plays a key role in its ability to generalize to new environments. Through a series of experiments in a tabular contextual MDP, we show that exploration is helpful not only for efficiently finding the optimal policy for the training environments but also for acquiring knowledge that helps decision making in unseen environments. Based on these observations, we propose EDE: Exploration via Distributional Ensemble, a method that encourages exploration of states with high epistemic uncertainty through an ensemble of Q-value distributions. Our algorithm is the first value-based approach to achieve state-of-the-art on both Procgen and Crafter, two benchmarks for generalization in RL with high-dimensional observations. The open-sourced implementation can be found at https://github.com/facebookresearch/ede ."
                    },
                    {
                        "title": "Mimetic Initialization of Self-Attention Layers",
                        "abstract": "It is notoriously difficult to train Transformers on small datasets; typically, large pre-trained models are instead used as the starting point. We explore the weights of such pre-trained Transformers (particularly for vision) to attempt to find reasons for this discrepancy. Surprisingly, we find that simply initializing the weights of self-attention layers so that they\"look\"more like their pre-trained counterparts allows us to train vanilla Transformers faster and to higher final accuracies, particularly on vision tasks such as CIFAR-10 and ImageNet classification, where we see gains in accuracy of over 5% and 4%, respectively. Our initialization scheme is closed form, learning-free, and very simple: we set the product of the query and key weights to be approximately the identity, and the product of the value and projection weights to approximately the negative identity. As this mimics the patterns we saw in pre-trained Transformers, we call the technique\"mimetic initialization\"."
                    },
                    {
                        "title": "Universal and Transferable Adversarial Attacks on Aligned Language Models",
                        "abstract": "Because\"out-of-the-box\"large language models are capable of generating a great deal of objectionable content, recent work has focused on aligning these models in an attempt to prevent undesirable generation. While there has been some success at circumventing these measures -- so-called\"jailbreaks\"against LLMs -- these attacks have required significant human ingenuity and are brittle in practice. In this paper, we propose a simple and effective attack method that causes aligned language models to generate objectionable behaviors. Specifically, our approach finds a suffix that, when attached to a wide range of queries for an LLM to produce objectionable content, aims to maximize the probability that the model produces an affirmative response (rather than refusing to answer). However, instead of relying on manual engineering, our approach automatically produces these adversarial suffixes by a combination of greedy and gradient-based search techniques, and also improves over past automatic prompt generation methods. Surprisingly, we find that the adversarial prompts generated by our approach are quite transferable, including to black-box, publicly released LLMs. Specifically, we train an adversarial attack suffix on multiple prompts (i.e., queries asking for many different types of objectionable content), as well as multiple models (in our case, Vicuna-7B and 13B). When doing so, the resulting attack suffix is able to induce objectionable content in the public interfaces to ChatGPT, Bard, and Claude, as well as open source LLMs such as LLaMA-2-Chat, Pythia, Falcon, and others. In total, this work significantly advances the state-of-the-art in adversarial attacks against aligned language models, raising important questions about how such systems can be prevented from producing objectionable information. Code is available at github.com/llm-attacks/llm-attacks."
                    },
                    {
                        "title": "Practical Critic Gradient based Actor Critic for On-Policy Reinforcement Learning",
                        "abstract": "On-policy reinforcement learning algorithms have been shown to be remarkably efficient at learning policies for continuous control robotics tasks. They are highly parallelizable and hence have benefited tremendously from the recent rise in GPU based parallel simulators. The most widely used on-policy reinforcement learning algorithm is proximal policy optimization (PPO) which was introduced in 2017 and was designed for a somewhat different setting with CPU based serial or less parallelizable simulators. However, suprisingly, it has maintained dominance even on tasks based on the highly parallelizable simulators of today. In this paper, we show that a different class of on-policy algorithms based on estimating the policy gradient using the critic-action gradients are better suited when using highly parallelizable simulators. The primary issues for these algorithms arise from the lack of diversity of the on-policy experiences used for the updates and the instabilities arising from the interaction between the biased critic gradients and the rapidly changing policy distribution. We address the former by simply increasing the number of parallel simulation runs (thanks to the GPU based simulators) along with an appropriate schedule on the policy entropy to ensure diversity of samples. We address the latter by adding a policy averaging step and value averaging step (as in off-policy methods). With these modifications, we observe that the critic gradient based on-policy method (CGAC) consistently achieves higher episode returns compared with existing baselines. Furthermore, in environments with high dimensional action space, CGAC also trains much faster (in wall-clock time) than the corresponding baselines."
                    },
                    {
                        "title": "Projected Off-Policy Q-Learning (POP-QL) for Stabilizing Offline Reinforcement Learning",
                        "abstract": "A key problem in off-policy Reinforcement Learning (RL) is the mismatch, or distribution shift, between the dataset and the distribution over states and actions visited by the learned policy. This problem is exacerbated in the fully offline setting. The main approach to correct this shift has been through importance sampling, which leads to high-variance gradients. Other approaches, such as conservatism or behavior-regularization, regularize the policy at the cost of performance. In this paper, we propose a new approach for stable off-policy Q-Learning. Our method, Projected Off-Policy Q-Learning (POP-QL), is a novel actor-critic algorithm that simultaneously reweights off-policy samples and constrains the policy to prevent divergence and reduce value-approximation error. In our experiments, POP-QL not only shows competitive performance on standard benchmarks, but also out-performs competing methods in tasks where the data-collection policy is significantly sub-optimal."
                    },
                    {
                        "title": "On the Joint Interaction of Models, Data, and Features",
                        "abstract": "Learning features from data is one of the defining characteristics of deep learning, but our theoretical understanding of the role features play in deep learning is still rudimentary. To address this gap, we introduce a new tool, the interaction tensor, for empirically analyzing the interaction between data and model through features. With the interaction tensor, we make several key observations about how features are distributed in data and how models with different random seeds learn different features. Based on these observations, we propose a conceptual framework for feature learning. Under this framework, the expected accuracy for a single hypothesis and agreement for a pair of hypotheses can both be derived in closed-form. We demonstrate that the proposed framework can explain empirically observed phenomena, including the recently discovered Generalization Disagreement Equality (GDE) that allows for estimating the generalization error with only unlabeled data. Further, our theory also provides explicit construction of natural data distributions that break the GDE. Thus, we believe this work provides valuable new insight into our understanding of feature learning."
                    }
                ]
            },
            "e7fe89ea-8d42-4a0e-9406-d952050714ff": {
                "pk": "e7fe89ea-8d42-4a0e-9406-d952050714ff",
                "name": "Ruslan Salakhutdinov",
                "collaborators": [
                    "Jing Yu Koh",
                    "Daniel Fried",
                    "Chen Henry Wu",
                    "Aditi Raghunathan",
                    "Robert Lo",
                    "Lawrence Jang",
                    "Vikram Duvvur",
                    "Ming Chong Lim",
                    "Po-Yu Huang",
                    "Graham Neubig"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Autonomous Agents",
                    "Computer Vision",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Adversarial Attacks on Multimodal Agents",
                        "abstract": "Vision-enabled language models (VLMs) are now used to build autonomous multimodal agents capable of taking actions in real environments. In this paper, we show that multimodal agents raise new safety risks, even though attacking agents is more challenging than prior attacks due to limited access to and knowledge about the environment. Our attacks use adversarial text strings to guide gradient-based perturbation over one trigger image in the environment: (1) our captioner attack attacks white-box captioners if they are used to process images into captions as additional inputs to the VLM; (2) our CLIP attack attacks a set of CLIP models jointly, which can transfer to proprietary VLMs. To evaluate the attacks, we curated VisualWebArena-Adv, a set of adversarial tasks based on VisualWebArena, an environment for web-based multimodal agent tasks. Within an L-infinity norm of $16/256$ on a single image, the captioner attack can make a captioner-augmented GPT-4V agent execute the adversarial goals with a 75% success rate. When we remove the captioner or use GPT-4V to generate its own captions, the CLIP attack can achieve success rates of 21% and 43%, respectively. Experiments on agents based on other VLMs, such as Gemini-1.5, Claude-3, and GPT-4o, show interesting differences in their robustness. Further analysis reveals several key factors contributing to the attack's success, and we also discuss the implications for defenses as well. Project page: https://chenwu.io/attack-agent Code and data: https://github.com/ChenWu98/agent-attack"
                    },
                    {
                        "title": "VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks",
                        "abstract": "Autonomous agents capable of planning, reasoning, and executing actions on the web offer a promising avenue for automating computer tasks. However, the majority of existing benchmarks primarily focus on text-based agents, neglecting many natural tasks that require visual information to effectively solve. Given that most computer interfaces cater to human perception, visual information often augments textual data in ways that text-only models struggle to harness effectively. To bridge this gap, we introduce VisualWebArena, a benchmark designed to assess the performance of multimodal web agents on realistic \\textit{visually grounded tasks}. VisualWebArena comprises of a set of diverse and complex web-based tasks that evaluate various capabilities of autonomous multimodal agents. To perform on this benchmark, agents need to accurately process image-text inputs, interpret natural language instructions, and execute actions on websites to accomplish user-defined objectives. We conduct an extensive evaluation of state-of-the-art LLM-based autonomous agents, including several multimodal models. Through extensive quantitative and qualitative analysis, we identify several limitations of text-only LLM agents, and reveal gaps in the capabilities of state-of-the-art multimodal language agents. VisualWebArena provides a framework for evaluating multimodal autonomous language agents, and offers insights towards building stronger autonomous agents for the web. Our code, baseline models, and data is publicly available at https://jykoh.com/vwa."
                    },
                    {
                        "title": "Automated Black-box Prompt Engineering for Personalized Text-to-Image Generation",
                        "abstract": "Prompt engineering is effective for controlling the output of text-to-image (T2I) generative models, but it is also laborious due to the need for manually crafted prompts. This challenge has spurred the development of algorithms for automated prompt generation. However, these methods often struggle with transferability across T2I models, require white-box access to the underlying model, and produce non-intuitive prompts. In this work, we introduce PRISM, an algorithm that automatically identifies human-interpretable and transferable prompts that can effectively generate desired concepts given only black-box access to T2I models. Inspired by large language model (LLM) jailbreaking, PRISM leverages the in-context learning ability of LLMs to iteratively refine the candidate prompts distribution for given reference images. Our experiments demonstrate the versatility and effectiveness of PRISM in generating accurate prompts for objects, styles and images across multiple T2I models, including Stable Diffusion, DALL-E, and Midjourney."
                    },
                    {
                        "title": "Tree Search for Language Model Agents",
                        "abstract": "Autonomous agents powered by language models (LMs) have demonstrated promise in their ability to perform decision-making tasks such as web automation. However, a key limitation remains: LMs, primarily optimized for natural language understanding and generation, struggle with multi-step reasoning, planning, and using environmental feedback when attempting to solve realistic computer tasks. Towards addressing this, we propose an inference-time search algorithm for LM agents to explicitly perform exploration and multi-step planning in interactive web environments. Our approach is a form of best-first tree search that operates within the actual environment space, and is complementary with most existing state-of-the-art agents. It is the first tree search algorithm for LM agents that shows effectiveness on realistic web tasks. On the challenging VisualWebArena benchmark, applying our search algorithm on top of a GPT-4o agent yields a 39.7% relative increase in success rate compared to the same baseline without search, setting a state-of-the-art success rate of 26.4%. On WebArena, search also yields a 28.0% relative improvement over a baseline agent, setting a competitive success rate of 19.2%. Our experiments highlight the effectiveness of search for web agents, and we demonstrate that performance scales with increased test-time compute. We conduct a thorough analysis of our results to highlight improvements from search, limitations, and promising directions for future work. Our code and models are publicly released at https://jykoh.com/search-agents."
                    },
                    {
                        "title": "OmniACT: A Dataset and Benchmark for Enabling Multimodal Generalist Autonomous Agents for Desktop and Web",
                        "abstract": "For decades, human-computer interaction has fundamentally been manual. Even today, almost all productive work done on the computer necessitates human input at every step. Autonomous virtual agents represent an exciting step in automating many of these menial tasks. Virtual agents would empower users with limited technical proficiency to harness the full possibilities of computer systems. They could also enable the efficient streamlining of numerous computer tasks, ranging from calendar management to complex travel bookings, with minimal human intervention. In this paper, we introduce OmniACT, the first-of-a-kind dataset and benchmark for assessing an agent's capability to generate executable programs to accomplish computer tasks. Our scope extends beyond traditional web automation, covering a diverse range of desktop applications. The dataset consists of fundamental tasks such as\"Play the next song\", as well as longer horizon tasks such as\"Send an email to John Doe mentioning the time and place to meet\". Specifically, given a pair of screen image and a visually-grounded natural language task, the goal is to generate a script capable of fully executing the task. We run several strong baseline language model agents on our benchmark. The strongest baseline, GPT-4, performs the best on our benchmark However, its performance level still reaches only 15% of the human proficiency in generating executable scripts capable of completing the task, demonstrating the challenge of our task for conventional web agents. Our benchmark provides a platform to measure and evaluate the progress of language model agents in automating computer tasks and motivates future work towards building multimodal models that bridge large language models and the visual grounding of computer screens."
                    }
                ]
            },
            "b82363a8-8d8d-4ce2-80be-789ebf52e5fa": {
                "pk": "b82363a8-8d8d-4ce2-80be-789ebf52e5fa",
                "name": "Stefano Ermon",
                "collaborators": [
                    "Minkai Xu",
                    "J. Leskovec",
                    "Chenlin Meng",
                    "Ling Yang",
                    "Bin Cui",
                    "Zhaochen Yu",
                    "K. Azizzadenesheli",
                    "Tailin Wu",
                    "Charles Marx",
                    "Hongxu Ma"
                ],
                "domain": [
                    "Diffusion Models",
                    "Machine Learning",
                    "Natural Language Processing",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "Mastering Text-to-Image Diffusion: Recaptioning, Planning, and Generating with Multimodal LLMs",
                        "abstract": "Diffusion models have exhibit exceptional performance in text-to-image generation and editing. However, existing methods often face challenges when handling complex text prompts that involve multiple objects with multiple attributes and relationships. In this paper, we propose a brand new training-free text-to-image generation/editing framework, namely Recaption, Plan and Generate (RPG), harnessing the powerful chain-of-thought reasoning ability of multimodal LLMs to enhance the compositionality of text-to-image diffusion models. Our approach employs the MLLM as a global planner to decompose the process of generating complex images into multiple simpler generation tasks within subregions. We propose complementary regional diffusion to enable region-wise compositional generation. Furthermore, we integrate text-guided image generation and editing within the proposed RPG in a closed-loop fashion, thereby enhancing generalization ability. Extensive experiments demonstrate our RPG outperforms state-of-the-art text-to-image diffusion models, including DALL-E 3 and SDXL, particularly in multi-category object composition and text-image semantic alignment. Notably, our RPG framework exhibits wide compatibility with various MLLM architectures (e.g., MiniGPT-4) and diffusion backbones (e.g., ControlNet). Our code is available at: https://github.com/YangLing0818/RPG-DiffusionMaster"
                    },
                    {
                        "title": "Equivariant Graph Neural Operator for Modeling 3D Dynamics",
                        "abstract": "Modeling the complex three-dimensional (3D) dynamics of relational systems is an important problem in the natural sciences, with applications ranging from molecular simulations to particle mechanics. Machine learning methods have achieved good success by learning graph neural networks to model spatial interactions. However, these approaches do not faithfully capture temporal correlations since they only model next-step predictions. In this work, we propose Equivariant Graph Neural Operator (EGNO), a novel and principled method that directly models dynamics as trajectories instead of just next-step prediction. Different from existing methods, EGNO explicitly learns the temporal evolution of 3D dynamics where we formulate the dynamics as a function over time and learn neural operators to approximate it. To capture the temporal correlations while keeping the intrinsic SE(3)-equivariance, we develop equivariant temporal convolutions parameterized in the Fourier space and build EGNO by stacking the Fourier layers over equivariant networks. EGNO is the first operator learning framework that is capable of modeling solution dynamics functions over time while retaining 3D equivariance. Comprehensive experiments in multiple domains, including particle simulations, human motion capture, and molecular dynamics, demonstrate the significantly superior performance of EGNO against existing methods, thanks to the equivariant temporal modeling. Our code is available at https://github.com/MinkaiXu/egno."
                    },
                    {
                        "title": "Segment Any Change",
                        "abstract": "Visual foundation models have achieved remarkable results in zero-shot image classification and segmentation, but zero-shot change detection remains an open problem. In this paper, we propose the segment any change models (AnyChange), a new type of change detection model that supports zero-shot prediction and generalization on unseen change types and data distributions. AnyChange is built on the segment anything model (SAM) via our training-free adaptation method, bitemporal latent matching. By revealing and exploiting intra-image and inter-image semantic similarities in SAM's latent space, bitemporal latent matching endows SAM with zero-shot change detection capabilities in a training-free way. We also propose a point query mechanism to enable AnyChange's zero-shot object-centric change detection capability. We perform extensive experiments to confirm the effectiveness of AnyChange for zero-shot change detection. AnyChange sets a new record on the SECOND benchmark for unsupervised change detection, exceeding the previous SOTA by up to 4.4% F$_1$ score, and achieving comparable accuracy with negligible manual annotations (1 pixel per image) for supervised change detection."
                    },
                    {
                        "title": "Uncertainty Quantification for Forward and Inverse Problems of PDEs via Latent Global Evolution",
                        "abstract": "Deep learning-based surrogate models have demonstrated remarkable advantages over classical solvers in terms of speed, often achieving speedups of 10 to 1000 times over traditional partial differential equation (PDE) solvers. However, a significant challenge hindering their widespread adoption in both scientific and industrial domains is the lack of understanding about their prediction uncertainties, particularly in scenarios that involve critical decision making. To address this limitation, we propose a method that integrates efficient and precise uncertainty quantification into a deep learning-based surrogate model. Our method, termed Latent Evolution of PDEs with Uncertainty Quantification (LE-PDE-UQ), endows deep learning-based surrogate models with robust and efficient uncertainty quantification capabilities for both forward and inverse problems. LE-PDE-UQ leverages latent vectors within a latent space to evolve both the system's state and its corresponding uncertainty estimation. The latent vectors are decoded to provide predictions for the system's state as well as estimates of its uncertainty. In extensive experiments, we demonstrate the accurate uncertainty quantification performance of our approach, surpassing that of strong baselines including deep ensembles, Bayesian neural network layers, and dropout. Our method excels at propagating uncertainty over extended auto-regressive rollouts, making it suitable for scenarios involving long-term predictions. Our code is available at: https://github.com/AI4Science-WestlakeU/le-pde-uq."
                    },
                    {
                        "title": "Calibrated Probabilistic Forecasts for Arbitrary Sequences",
                        "abstract": "Real-world data streams can change unpredictably due to distribution shifts, feedback loops and adversarial actors, which challenges the validity of forecasts. We present a forecasting framework ensuring valid uncertainty estimates regardless of how data evolves. Leveraging the concept of Blackwell approachability from game theory, we introduce a forecasting framework that guarantees calibrated uncertainties for outcomes in any compact space (e.g., classification or bounded regression). We extend this framework to recalibrate existing forecasters, guaranteeing accurate uncertainties without sacrificing predictive performance. We implement both general-purpose gradient-based algorithms and algorithms optimized for popular special cases of our framework. Empirically, our algorithms improve calibration and downstream decision-making for energy systems."
                    },
                    {
                        "title": "TorchSpatial: A Location Encoding Framework and Benchmark for Spatial Representation Learning",
                        "abstract": "Spatial representation learning (SRL) aims at learning general-purpose neural network representations from various types of spatial data (e.g., points, polylines, polygons, networks, images, etc.) in their native formats. Learning good spatial representations is a fundamental problem for various downstream applications such as species distribution modeling, weather forecasting, trajectory generation, geographic question answering, etc. Even though SRL has become the foundation of almost all geospatial artificial intelligence (GeoAI) research, we have not yet seen significant efforts to develop an extensive deep learning framework and benchmark to support SRL model development and evaluation. To fill this gap, we propose TorchSpatial, a learning framework and benchmark for location (point) encoding, which is one of the most fundamental data types of spatial representation learning. TorchSpatial contains three key components: 1) a unified location encoding framework that consolidates 15 commonly recognized location encoders, ensuring scalability and reproducibility of the implementations; 2) the LocBench benchmark tasks encompassing 7 geo-aware image classification and 4 geo-aware image regression datasets; 3) a comprehensive suite of evaluation metrics to quantify geo-aware models' overall performance as well as their geographic bias, with a novel Geo-Bias Score metric. Finally, we provide a detailed analysis and insights into the model performance and geographic bias of different location encoders. We believe TorchSpatial will foster future advancement of spatial representation learning and spatial fairness in GeoAI research. The TorchSpatial model framework, LocBench, and Geo-Bias Score evaluation framework are available at https://github.com/seai-lab/TorchSpatial."
                    },
                    {
                        "title": "On the Scalability of Diffusion-based Text-to-Image Generation",
                        "abstract": "Scaling up model and data size has been quite successful for the evolution of LLMs. However, the scaling law for the diffusion based text-to-image (T2I) models is not fully explored. It is also unclear how to efficiently scale the model for better performance at reduced cost. The different training settings and expensive training cost make a fair model comparison extremely difficult. In this work, we empirically study the scaling properties of diffusion based T2I models by performing extensive and rigours ablations on scaling both denoising backbones and training set, including training scaled UNet and Transformer variants ranging from 0.4B to 4B parameters on datasets upto 600M images. For model scaling, we find the location and amount of cross attention distinguishes the performance of existing UNet designs. And increasing the transformer blocks is more parameter-efficient for improving text-image alignment than increasing channel numbers. We then identify an efficient UNet variant, which is 45% smaller and 28% faster than SDXL's UNet. On the data scaling side, we show the quality and diversity of the training set matters more than simply dataset size. Increasing caption density and diversity improves text-image alignment performance and the learning efficiency. Finally, we provide scaling functions to predict the text-image alignment performance as functions of the scale of model size, compute and dataset size."
                    },
                    {
                        "title": "Contextualized Diffusion Models for Text-Guided Image and Video Generation",
                        "abstract": "Conditional diffusion models have exhibited superior performance in high-fidelity text-guided visual generation and editing. Nevertheless, prevailing text-guided visual diffusion models primarily focus on incorporating text-visual relationships exclusively into the reverse process, often disregarding their relevance in the forward process. This inconsistency between forward and reverse processes may limit the precise conveyance of textual semantics in visual synthesis results. To address this issue, we propose a novel and general contextualized diffusion model (ContextDiff) by incorporating the cross-modal context encompassing interactions and alignments between text condition and visual sample into forward and reverse processes. We propagate this context to all timesteps in the two processes to adapt their trajectories, thereby facilitating cross-modal conditional modeling. We generalize our contextualized diffusion to both DDPMs and DDIMs with theoretical derivations, and demonstrate the effectiveness of our model in evaluations with two challenging tasks: text-to-image generation, and text-to-video editing. In each task, our ContextDiff achieves new state-of-the-art performance, significantly enhancing the semantic alignment between text condition and generated samples, as evidenced by quantitative and qualitative evaluations. Our code is available at https://github.com/YangLing0818/ContextDiff"
                    },
                    {
                        "title": "Consistency Flow Matching: Defining Straight Flows with Velocity Consistency",
                        "abstract": "Flow matching (FM) is a general framework for defining probability paths via Ordinary Differential Equations (ODEs) to transform between noise and data samples. Recent approaches attempt to straighten these flow trajectories to generate high-quality samples with fewer function evaluations, typically through iterative rectification methods or optimal transport solutions. In this paper, we introduce Consistency Flow Matching (Consistency-FM), a novel FM method that explicitly enforces self-consistency in the velocity field. Consistency-FM directly defines straight flows starting from different times to the same endpoint, imposing constraints on their velocity values. Additionally, we propose a multi-segment training approach for Consistency-FM to enhance expressiveness, achieving a better trade-off between sampling quality and speed. Preliminary experiments demonstrate that our Consistency-FM significantly improves training efficiency by converging 4.4x faster than consistency models and 1.7x faster than rectified flow models while achieving better generation quality. Our code is available at: https://github.com/YangLing0818/consistency_flow_matching"
                    },
                    {
                        "title": "Aligning Target-Aware Molecule Diffusion Models with Exact Energy Optimization",
                        "abstract": "Generating ligand molecules for specific protein targets, known as structure-based drug design, is a fundamental problem in therapeutics development and biological discovery. Recently, target-aware generative models, especially diffusion models, have shown great promise in modeling protein-ligand interactions and generating candidate drugs. However, existing models primarily focus on learning the chemical distribution of all drug candidates, which lacks effective steerability on the chemical quality of model generations. In this paper, we propose a novel and general alignment framework to align pretrained target diffusion models with preferred functional properties, named AliDiff. AliDiff shifts the target-conditioned chemical distribution towards regions with higher binding affinity and structural rationality, specified by user-defined reward functions, via the preference optimization approach. To avoid the overfitting problem in common preference optimization objectives, we further develop an improved Exact Energy Preference Optimization method to yield an exact and efficient alignment of the diffusion models, and provide the closed-form expression for the converged distribution. Empirical studies on the CrossDocked2020 benchmark show that AliDiff can generate molecules with state-of-the-art binding energies with up to -7.07 Avg. Vina Score, while maintaining strong molecular properties. Code is available at https://github.com/MinkaiXu/AliDiff."
                    },
                    {
                        "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                        "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                    },
                    {
                        "title": "GEO-Bench: Toward Foundation Models for Earth Monitoring",
                        "abstract": "Recent progress in self-supervision has shown that pre-training large neural networks on vast amounts of unsupervised data can lead to substantial increases in generalization to downstream tasks. Such models, recently coined foundation models, have been transformational to the field of natural language processing. Variants have also been proposed for image data, but their applicability to remote sensing tasks is limited. To stimulate the development of foundation models for Earth monitoring, we propose a benchmark comprised of six classification and six segmentation tasks, which were carefully curated and adapted to be both relevant to the field and well-suited for model evaluation. We accompany this benchmark with a robust methodology for evaluating models and reporting aggregated results to enable a reliable assessment of progress. Finally, we report results for 20 baselines to gain information about the performance of existing models. We believe that this benchmark will be a driver of progress across a variety of Earth monitoring tasks."
                    },
                    {
                        "title": "Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems",
                        "abstract": "Advances in artificial intelligence (AI) are fueling a new paradigm of discoveries in natural sciences. Today, AI has started to advance natural sciences by improving, accelerating, and enabling our understanding of natural phenomena at a wide range of spatial and temporal scales, giving rise to a new area of research known as AI for science (AI4Science). Being an emerging research paradigm, AI4Science is unique in that it is an enormous and highly interdisciplinary area. Thus, a unified and technical treatment of this field is needed yet challenging. This work aims to provide a technically thorough account of a subarea of AI4Science; namely, AI for quantum, atomistic, and continuum systems. These areas aim at understanding the physical world from the subatomic (wavefunctions and electron density), atomic (molecules, proteins, materials, and interactions), to macro (fluids, climate, and subsurface) scales and form an important subarea of AI4Science. A unique advantage of focusing on these areas is that they largely share a common set of challenges, thereby allowing a unified and foundational treatment. A key common challenge is how to capture physics first principles, especially symmetries, in natural systems by deep learning methods. We provide an in-depth yet intuitive account of techniques to achieve equivariance to symmetry transformations. We also discuss other common technical challenges, including explainability, out-of-distribution generalization, knowledge transfer with foundation and large language models, and uncertainty quantification. To facilitate learning and education, we provide categorized lists of resources that we found to be useful. We strive to be thorough and unified and hope this initial effort may trigger more community interests and efforts to further advance AI4Science."
                    },
                    {
                        "title": "But Are You Sure? An Uncertainty-Aware Perspective on Explainable AI",
                        "abstract": "Although black-box models can accurately predict outcomes such as weather patterns, they often lack transparency, making it challenging to extract meaningful insights (such as which atmospheric conditions signal future rainfall). Model explanations attempt to identify the essential features of a model, but these explanations can be inconsistent: two near-optimal models may admit vastly different explanations. In this paper, we propose a solution to this problem by constructing uncertainty sets for explanations of the optimal model(s) in both frequentist and Bayesian settings. Our uncertainty sets are guaranteed to include the explanation of the optimal model with high probability, even though this model is unknown. We demonstrate the effectiveness of our approach in both synthetic and real-world experiments, illustrating how our uncertainty sets can be used to calibrate trust in model explanations."
                    },
                    {
                        "title": "SSIF: Learning Continuous Image Representation for Spatial-Spectral Super-Resolution",
                        "abstract": "Existing digital sensors capture images at fixed spatial and spectral resolutions (e.g., RGB, multispectral, and hyperspectral images), and each combination requires bespoke machine learning models. Neural Implicit Functions partially overcome the spatial resolution challenge by representing an image in a resolution-independent way. However, they still operate at fixed, pre-defined spectral resolutions. To address this challenge, we propose Spatial-Spectral Implicit Function (SSIF), a neural implicit model that represents an image as a function of both continuous pixel coordinates in the spatial domain and continuous wavelengths in the spectral domain. We empirically demonstrate the effectiveness of SSIF on two challenging spatio-spectral super-resolution benchmarks. We observe that SSIF consistently outperforms state-of-the-art baselines even when the baselines are allowed to train separate models at each spectral resolution. We show that SSIF generalizes well to both unseen spatial resolutions and spectral resolutions. Moreover, SSIF can generate high-resolution images that improve the performance of downstream tasks (e.g., land use classification) by 1.7%-7%."
                    },
                    {
                        "title": "MERMAIDE: Learning to Align Learners using Model-Based Meta-Learning",
                        "abstract": "We study how a principal can efficiently and effectively intervene on the rewards of a previously unseen learning agent in order to induce desirable outcomes. This is relevant to many real-world settings like auctions or taxation, where the principal may not know the learning behavior nor the rewards of real people. Moreover, the principal should be few-shot adaptable and minimize the number of interventions, because interventions are often costly. We introduce MERMAIDE, a model-based meta-learning framework to train a principal that can quickly adapt to out-of-distribution agents with different learning strategies and reward functions. We validate this approach step-by-step. First, in a Stackelberg setting with a best-response agent, we show that meta-learning enables quick convergence to the theoretically known Stackelberg equilibrium at test time, although noisy observations severely increase the sample complexity. We then show that our model-based meta-learning approach is cost-effective in intervening on bandit agents with unseen explore-exploit strategies. Finally, we outperform baselines that use either meta-learning or agent behavior modeling, in both $0$-shot and $K=1$-shot settings with partial agent information."
                    },
                    {
                        "title": "MADiff: Offline Multi-agent Learning with Diffusion Models",
                        "abstract": "Diffusion model (DM) recently achieved huge success in various scenarios including offline reinforcement learning, where the diffusion planner learn to generate desired trajectories during online evaluations. However, despite the effectiveness in single-agent learning, it remains unclear how DMs can operate in multi-agent problems, where agents can hardly complete teamwork without good coordination by independently modeling each agent's trajectories. In this paper, we propose MADiff, a novel generative multi-agent learning framework to tackle this problem. MADiff is realized with an attention-based diffusion model to model the complex coordination among behaviors of multiple agents. To the best of our knowledge, MADiff is the first diffusion-based multi-agent learning framework, which behaves as both a decentralized policy and a centralized controller. During decentralized executions, MADiff simultaneously performs teammate modeling, and the centralized controller can also be applied in multi-agent trajectory predictions. Our experiments show the superior performance of MADiff compared to baseline algorithms in a wide range of multi-agent learning tasks, which emphasizes the effectiveness of MADiff in modeling complex multi-agent interactions. Our code is available at https://github.com/zbzhu99/madiff."
                    },
                    {
                        "title": "Holistic Evaluation of Text-To-Image Models",
                        "abstract": "The stunning qualitative improvement of recent text-to-image models has led to their widespread attention and adoption. However, we lack a comprehensive quantitative understanding of their capabilities and risks. To fill this gap, we introduce a new benchmark, Holistic Evaluation of Text-to-Image Models (HEIM). Whereas previous evaluations focus mostly on text-image alignment and image quality, we identify 12 aspects, including text-image alignment, image quality, aesthetics, originality, reasoning, knowledge, bias, toxicity, fairness, robustness, multilinguality, and efficiency. We curate 62 scenarios encompassing these aspects and evaluate 26 state-of-the-art text-to-image models on this benchmark. Our results reveal that no single model excels in all aspects, with different models demonstrating different strengths. We release the generated images and human evaluation results for full transparency at https://crfm.stanford.edu/heim/v1.1.0 and the code at https://github.com/stanford-crfm/helm, which is integrated with the HELM codebase."
                    },
                    {
                        "title": "Understanding and Adopting Rational Behavior by Bellman Score Estimation",
                        "abstract": "We are interested in solving a class of problems that seek to understand and adopt rational behavior from demonstrations. We may broadly classify these problems into four categories of reward identi\ufb01cation, counterfactual analysis, behavior imitation"
                    }
                ]
            }
        }
    },
    "2405.12601": {
        "paper_data": {
            "title": "FFAM: Feature Factorization Activation Map for Explanation of 3D Detectors",
            "url": "http://arxiv.org/abs/2405.12601v1",
            "arxiv_id": "2405.12601",
            "authors": [
                "Shuai Liu",
                "Boyang Li",
                "Zhiyu Fang",
                "Mingyue Cui",
                "Kai Huang"
            ],
            "abstract": "LiDAR-based 3D object detection has made impressive progress recently, yet most existing models are black-box, lacking interpretability. Previous explanation approaches primarily focus on analyzing image-based models and are not readily applicable to LiDAR-based 3D detectors. In this paper, we propose a feature factorization activation map (FFAM) to generate high-quality visual explanations for 3D detectors. FFAM employs non-negative matrix factorization to generate concept activation maps and subsequently aggregates these maps to obtain a global visual explanation. To achieve object-specific visual explanations, we refine the global visual explanation using the feature gradient of a target object. Additionally, we introduce a voxel upsampling strategy to align the scale between the activation map and input point cloud. We qualitatively and quantitatively analyze FFAM with multiple detectors on several datasets. Experimental results validate the high-quality visual explanations produced by FFAM. The Code will be available at \\url{https://github.com/Say2L/FFAM.git}.",
            "introduction": "   1 Introduction  In recent years, there has been rapid development in LiDAR-based 3D object detection [35, 31, 33, 37, 11], making it widely utilized in autonomous driving, industrial automation, and robot navigation. However, existing detection methods predominantly rely on deep neural networks with highly nonlinear and complex structures. Essentially, these models can be considered as \"black box\" systems. Such opaque modeling techniques hinder users from fully trusting the detection models, particularly in sensitive and high-risk domains. Consequently, understanding the decision-making process of these inherently opaque models is urgently needed.   Visual explanation methods [19, 1, 32, 14, 15] have gained widespread adoption for analyzing models based on deep neural networks. These methods generate saliency maps that highlight the crucial elements influencing the model\u2019s decision within the input map. Perturbation-based [14, 15], class activation map (CAM)-based [40, 19, 1], and gradient-based [27, 23, 24] methods are the three main categories of visual explanation methods. However, these methods primarily focus on image-based models and are not directly applicable to point cloud-based models. The pioneering work in analyzing 3D detectors is OccAM [18], which extends D-RISE [15] to perturb point clouds. As a perturbation-based approach, OccAM first randomly samples numerous sub-point clouds and measures the change in model predictions. However, the large number of inference calculations makes OccAM computationally intensive, and the sampling number easily impacts the quality of generated saliency maps.   Interpreting 3D detectors presents three key challenges. First, point clouds are inherently three-dimensional (3D). It is essential to generate corresponding 3D saliency maps for accurate interpretation. However, existing methods, such as popular CAM-based techniques, primarily utilize activation maps from the network\u2019s last layer to generate 2D saliency maps. Second, the explanation method for 3D detectors should provide detailed explanations for individual objects of interest. Yet, most existing methods yield class-specific saliency maps, which means they cannot focus on explaining a specific detection object. Lastly, point clouds are sparsely distributed in 3D space, rendering linear interpolation employed by many image-based explanation methods ineffective.   Figure 1: Visualization of FFAM outputs. (a) global concept activation map and (b) object-specific activation map.    To address the aforementioned challenges, this paper introduces a feature factorization activation map (FFAM) to obtain visual explanations for 3D detectors. Specifically, to solve the first challenge, FFAM leverages the 3D feature maps within the 3D backbone [35] of detectors, rather than relying on the bird\u2019s eye view (BEV) feature maps from the last layer. Drawing inspiration from DFF [3], we employ non-negative matrix factorization (NMF) to uncover latent semantic concepts within these 3D feature maps. Typically, point features with effective detection clues in 3D detectors contain richer semantic concepts. Thus, we aggregate the concept activation maps to generate a global concept activation map that highlights important points, as shown in Figure\u00a01(a). To address the second challenge of obtaining object-specific saliency maps, we utilize the gradients of the 3D feature map, generated by an object-specific loss, to refine the global concept activation map. This process is illustrated in Figure\u00a01(b), showcasing the desired effect. To tackle the final challenge, we introduce a voxel upsampling strategy to sample values from sparse",
            "references": [
                {
                    "title": "DCDet: Dynamic Cross-based 3D Object Detector",
                    "abstract": "Recently, significant progress has been made in the research of 3D object detection. However, most prior studies have focused on the utilization of center-based or anchor-based label assignment schemes. Alternative label assignment strategies remain unexplored in 3D object detection. We find that the center-based label assignment often fails to generate sufficient positive samples for training, while the anchor-based label assignment tends to encounter an imbalanced issue when handling objects with different scales. To solve these issues, we introduce a dynamic cross label assignment (DCLA) scheme, which dynamically assigns positive samples for each object from a cross-shaped region, thus providing sufficient and balanced positive samples for training. Furthermore, to address the challenge of accurately regressing objects with varying scales, we put forth a rotation-weighted Intersection over Union (RWIoU) metric to replace the widely used L1 metric in regression loss. Extensive experiments demonstrate the generality and effectiveness of our DCLA and RWIoU-based regression loss. The Code is available at https://github.com/Say2L/DCDet.git."
                },
                {
                    "title": "HEDNet: A Hierarchical Encoder-Decoder Network for 3D Object Detection in Point Clouds",
                    "abstract": "3D object detection in point clouds is important for autonomous driving systems. A primary challenge in 3D object detection stems from the sparse distribution of points within the 3D scene. Existing high-performance methods typically employ 3D sparse convolutional neural networks with small kernels to extract features. To reduce computational costs, these methods resort to submanifold sparse convolutions, which prevent the information exchange among spatially disconnected features. Some recent approaches have attempted to address this problem by introducing large-kernel convolutions or self-attention mechanisms, but they either achieve limited accuracy improvements or incur excessive computational costs. We propose HEDNet, a hierarchical encoder-decoder network for 3D object detection, which leverages encoder-decoder blocks to capture long-range dependencies among features in the spatial space, particularly for large and distant objects. We conducted extensive experiments on the Waymo Open and nuScenes datasets. HEDNet achieved superior detection accuracy on both datasets than previous state-of-the-art methods with competitive efficiency. The code is available at https://github.com/zhanggang001/HEDNet."
                },
                {
                    "title": "G-CAME: Gaussian-Class Activation Mapping Explainer for Object Detectors",
                    "abstract": "Nowadays, deep neural networks for object detection in images are very prevalent. However, due to the complexity of these networks, users find it hard to understand why these objects are detected by models. We proposed Gaussian Class Activation Mapping Explainer (G-CAME), which generates a saliency map as the explanation for object detection models. G-CAME can be considered a CAM-based method that uses the activation maps of selected layers combined with the Gaussian kernel to highlight the important regions in the image for the predicted box. Compared with other Region-based methods, G-CAME can transcend time constraints as it takes a very short time to explain an object. We also evaluated our method qualitatively and quantitatively with YOLOX on the MS-COCO 2017 dataset and guided to apply G-CAME into the two-stage Faster-RCNN model."
                },
                {
                    "title": "ODAM: Gradient-based instance-specific visual explanations for object detection",
                    "abstract": "We propose the gradient-weighted Object Detector Activation Maps (ODAM), a visualized explanation technique for interpreting the predictions of object detectors. Utilizing the gradients of detector targets flowing into the intermediate feature maps, ODAM produces heat maps that show the influence of regions on the detector's decision for each predicted attribute. Compared to previous works classification activation maps (CAM), ODAM generates instance-specific explanations rather than class-specific ones. We show that ODAM is applicable to both one-stage detectors and two-stage detectors with different types of detector backbones and heads, and produces higher-quality visual explanations than the state-of-the-art both effectively and efficiently. We next propose a training scheme, Odam-Train, to improve the explanation ability on object discrimination of the detector through encouraging consistency between explanations for detections on the same object, and distinct explanations for detections on different objects. Based on the heat maps produced by ODAM with Odam-Train, we propose Odam-NMS, which considers the information of the model's explanation for each prediction to distinguish the duplicate detected objects. We present a detailed analysis of the visualized explanations of detectors and carry out extensive experiments to validate the effectiveness of the proposed ODAM."
                },
                {
                    "title": "VoxelNeXt: Fully Sparse VoxelNet for 3D Object Detection and Tracking",
                    "abstract": "3D object detectors usually rely on hand-crafted proxies, e.g., anchors or centers, and translate well-studied 2D frameworks to 3D. Thus, sparse voxel features need to be densified and processed by dense prediction heads, which inevitably costs extra computation. In this paper, we instead propose VoxelNext for fully sparse 3D object detection. Our core insight is to predict objects directly based on sparse voxel features, without relying on hand-crafted proxies. Our strong sparse convolutional network VoxelNeXt detects and tracks 3D objects through voxel features entirely. It is an elegant and efficient framework, with no need for sparse-to-dense conversion or NMS post-processing. Our method achieves a better speed-accuracy trade-off than other mainframe detectors on the nuScenes dataset. For the first time, we show that a fully sparse voxel-based representation works decently for LIDAR 3D object detection and tracking. Extensive experiments on nuScenes, Waymo, and Argoverse2 benchmarks validate the effectiveness of our approach. Without bells and whistles, our model outperforms all existing LIDAR methods on the nuScenes tracking test benchmark. Code and models are available at github.com/dvlab-research/VoxelNeXt."
                },
                {
                    "title": "DSVT: Dynamic Sparse Voxel Transformer with Rotated Sets",
                    "abstract": "Designing an efficient yet deployment-friendly 3D backbone to handle sparse point clouds is a fundamental problem in 3D perception. Compared with the customized sparse convolution, the attention mechanism in Transformers is more appropriate for flexibly modeling long-range relationships and is easier to be deployed in real-world applications. However, due to the sparse characteristics of point clouds, it is non-trivial to apply a standard transformer on sparse points. In this paper, we present Dynamic Sparse Voxel Transformer (DSVT), a single-stride window-based voxel Transformer backbone for outdoor 3D perception. In order to efficiently process sparse points in parallel, we propose Dynamic Sparse Window Attention, which partitions a series of local regions in each window according to its sparsity and then computes the features of all regions in a fully parallel manner. To allow the cross-set connection, we design a rotated set partitioning strategy that alternates between two partitioning configurations in consecutive self-attention layers. To support effective downsampling and better encode geometric information, we also propose an attention-style 3D pooling module on sparse points, which is powerful and deployment-friendly without utilizing any customized CUDA operations. Our model achieves state-of-the-art performance with a broad range of 3D perception tasks. More importantly, DSVT can be easily deployed by TensorRT with real-time inference speed (27Hz). Code will be available at https://github.com/Haiyang-W/DSVT."
                },
                {
                    "title": "OccAM's Laser: Occlusion-based Attribution Maps for 3D Object Detectors on LiDAR Data",
                    "abstract": "While 3D object detection in LiDAR point clouds is well-established in academia and industry, the explainability of these models is a largely unexplored field. In this paper, we propose a method to generate attribution maps for the detected objects in order to better understand the behavior of such models. These maps indicate the importance of each 3D point in predicting the specific objects. Our method works with black-box models: We do not require any prior knowledge of the architecture nor access to the model's internals, like parameters, activations or gradients. Our efficient perturbation-based approach empirically estimates the importance of each point by testing the model with randomly generated subsets of the input point cloud. Our sub-sampling strategy takes into account the special characteristics of LiDAR data, such as the depth-dependent point density. We show a detailed evaluation of the attribution maps and demonstrate that they are interpretable and highly informative. Furthermore, we compare the attribution maps of recent 3D object detection architectures to provide insights into their decision-making processes."
                },
                {
                    "title": "Visualizing Global Explanations of Point Cloud DNNs",
                    "abstract": "So far, few researchers have targeted the explainability of point cloud neural networks. Part of the explainability methods are not directly applicable to those networks due to the structural specifics. In this work, we show that Activation Maximization (AM) with traditional pixel-wise regularizations fails to generate human-perceptible global explanations for point cloud networks. We propose new generative model-based AM approaches to clearly outline the global explanations and enhance their comprehensibility. Additionally, we propose a composite evaluation metric to address the limitations of existing evaluating methods, which simultaneously takes into account activation value, diversity and perceptibility. Extensive experiments demonstrate that our generative-based AM approaches outperform regularization-based ones both qualitatively and quantitatively. To the best of our knowledge, this is the first work investigating global explainability of point cloud networks. Our code is available at: https://github.com/Explain3D/PointCloudAM."
                },
                {
                    "title": "LayerCAM: Exploring Hierarchical Class Activation Maps for Localization",
                    "abstract": "The class activation maps are generated from the final convolutional layer of CNN. They can highlight discriminative object regions for the class of interest. These discovered object regions have been widely used for weakly-supervised tasks. However, due to the small spatial resolution of the final convolutional layer, such class activation maps often locate coarse regions of the target objects, limiting the performance of weakly-supervised tasks that need pixel-accurate object locations. Thus, we aim to generate more fine-grained object localization information from the class activation maps to locate the target objects more accurately. In this paper, by rethinking the relationships between the feature maps and their corresponding gradients, we propose a simple yet effective method, called LayerCAM. It can produce reliable class activation maps for different layers of CNN. This property enables us to collect object localization information from coarse (rough spatial localization) to fine (precise fine-grained details) levels. We further integrate them into a high-quality class activation map, where the object-related pixels can be better highlighted. To evaluate the quality of the class activation maps produced by LayerCAM, we apply them to weakly-supervised object localization and semantic segmentation. Experiments demonstrate that the class activation maps generated by our method are more effective and reliable than those by the existing attention methods. The code will be made publicly available."
                },
                {
                    "title": "Voxel R-CNN: Towards High Performance Voxel-based 3D Object Detection",
                    "abstract": "Recent advances on 3D object detection heavily rely on how the 3D data are represented, i.e., voxel-based or point-based representation. Many existing high performance 3D detectors are point-based because this structure can better retain precise point positions. Nevertheless, point-level features lead to high computation overheads due to unordered storage. In contrast, the voxel-based structure is better suited for feature extraction but often yields lower accuracy because the input data are divided into grids. In this paper, we take a slightly different viewpoint --- we find that precise positioning of raw points is not essential for high performance 3D object detection and that the coarse voxel granularity can also offer sufficient detection accuracy. Bearing this view in mind, we devise a simple but effective voxel-based framework, named Voxel R-CNN. By taking full advantage of voxel features in a two-stage approach, our method achieves comparable detection accuracy with state-of-the-art point-based models, but at a fraction of the computation cost. Voxel R-CNN consists of a 3D backbone network, a 2D bird-eye-view (BEV) Region Proposal Network, and a detect head. A voxel RoI pooling is devised to extract RoI features directly from voxel features for further refinement. Extensive experiments are conducted on the widely used KITTI Dataset and the more recent Waymo Open Dataset. Our results show that compared to existing voxel-based methods, Voxel R-CNN delivers a higher detection accuracy while maintaining a real-time frame processing rate, i.e., at a speed of 25 FPS on an NVIDIA RTX 2080 Ti GPU. The code is available at https://github.com/djiajunustc/Voxel-R-CNN."
                },
                {
                    "title": "Center-based 3D Object Detection and Tracking",
                    "abstract": "Three-dimensional objects are commonly represented as 3D boxes in a point-cloud. This representation mimics the well-studied image-based 2D bounding-box detection but comes with additional challenges. Objects in a 3D world do not follow any particular orientation, and box-based detectors have difficulties enumerating all orientations or fitting an axis-aligned bounding box to rotated objects. In this paper, we instead propose to represent, detect, and track 3D objects as points. Our framework, CenterPoint, first detects centers of objects using a keypoint detector and regresses to other attributes, including 3D size, 3D orientation, and velocity. In a second stage, it refines these estimates using additional point features on the object. In CenterPoint, 3D object tracking simplifies to greedy closest-point matching. The resulting detection and tracking algorithm is simple, efficient, and effective. CenterPoint achieved state-of-the-art performance on the nuScenes benchmark for both 3D detection and tracking, with 65.5 NDS and 63.8 AMOTA for a single model. On the Waymo Open Dataset, Center-Point outperforms all previous single model methods by a large margin and ranks first among all Lidar-only submissions. The code and pretrained models are available at https://github.com/tianweiy/CenterPoint."
                },
                {
                    "title": "Black-box Explanation of Object Detectors via Saliency Maps",
                    "abstract": "We propose D-RISE, a method for generating visual explanations for the predictions of object detectors. Utilizing the proposed similarity metric that accounts for both localization and categorization aspects of object detection allows our method to produce saliency maps that show image areas that most affect the prediction. D-RISE can be considered \"black-box\" in the software testing sense, as it only needs access to the inputs and outputs of an object detector. Compared to gradient-based methods, D-RISE is more general and agnostic to the particular type of object detector being tested, and does not need knowledge of the inner workings of the model. We show that D-RISE can be easily applied to different object detectors including one-stage detectors such as YOLOv3 and two-stage detectors such as Faster-RCNN. We present a detailed analysis of the generated visual explanations to highlight the utilization of context and possible biases learned by object detectors."
                },
                {
                    "title": "3D Point Cloud Feature Explanations Using Gradient-Based Methods",
                    "abstract": "Explainability is an important factor to drive user trust in the use of neural networks for tasks with material impact. However, most of the work done in this area focuses on image analysis and does not take into account 3D data. We extend the saliency methods that have been shown to work on image data to deal with 3D data. We analyse the features in point clouds and voxel spaces and show that edges and corners in 3D data are deemed as important features while planar surfaces are deemed less important. The approach is model-agnostic and can provide useful information about learnt features. Driven by the insight that 3D data is inherently sparse, we visualise the features learnt by a voxel-based classification network and show that these features are also sparse and can be pruned relatively easily, leading to more efficient neural networks. Our results show that the Voxception-ResNet model can be pruned down to 5% of its parameters with negligible loss in accuracy."
                },
                {
                    "title": "Ablation-CAM: Visual Explanations for Deep Convolutional Network via Gradient-free Localization",
                    "abstract": "In response to recent criticism of gradient-based visualization techniques, we propose a new methodology to generate visual explanations for deep Convolutional Neural Networks (CNN) - based models. Our approach \u2013 Ablation-based Class Activation Mapping (Ablation CAM) uses ablation analysis to determine the importance (weights) of individual feature map units w.r.t. class. Further, this is used to produce a coarse localization map highlighting the important regions in the image for predicting the concept. Our objective and subjective evaluations show that this gradient-free approach works better than state-of-the-art Grad-CAM technique. Moreover, further experiments are carried out to show that Ablation-CAM is class discriminative as well as can be used to evaluate trust in a model."
                },
                {
                    "title": "PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection",
                    "abstract": "We present a novel and high-performance 3D object detection framework, named PointVoxel-RCNN (PV-RCNN), for accurate 3D object detection from point clouds. Our proposed method deeply integrates both 3D voxel Convolutional Neural Network (CNN) and PointNet-based set abstraction to learn more discriminative point cloud features. It takes advantages of efficient learning and high-quality proposals of the 3D voxel CNN and the flexible receptive fields of the PointNet-based networks. Specifically, the proposed framework summarizes the 3D scene with a 3D voxel CNN into a small set of keypoints via a novel voxel set abstraction module to save follow-up computations and also to encode representative scene features. Given the high-quality 3D proposals generated by the voxel CNN, the RoI-grid pooling is proposed to abstract proposal-specific features from the keypoints to the RoI-grid points via keypoint set abstraction. Compared with conventional pooling operations, the RoI-grid feature points encode much richer context information for accurately estimating object confidences and locations. Extensive experiments on both the KITTI dataset and the Waymo Open dataset show that our proposed PV-RCNN surpasses state-of-the-art 3D detection methods with remarkable margins."
                },
                {
                    "title": "Scalability in Perception for Autonomous Driving: Waymo Open Dataset",
                    "abstract": "The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data. Existing self-driving datasets are limited in the scale and variation of the environments they capture, even though generalization within and between operating regions is crucial to the over-all viability of the technology. In an effort to help align the research community\u2019s contributions with real-world self-driving problems, we introduce a new large scale, high quality, diverse dataset. Our new dataset consists of 1150 scenes that each span 20 seconds, consisting of well synchronized and calibrated high quality LiDAR and camera data captured across a range of urban and suburban geographies. It is 15x more diverse than the largest camera+LiDAR dataset available based on our proposed diversity metric. We exhaustively annotated this data with 2D (camera image) and 3D (LiDAR) bounding boxes, with consistent identifiers across frames. Finally, we provide strong baselines for 2D as well as 3D detection and tracking tasks. We further study the effects of dataset size and generalization across geographies on 3D detection methods. Find data, code and more up-to-date information at http://www.waymo.com/open."
                },
                {
                    "title": "Score-CAM: Score-Weighted Visual Explanations for Convolutional Neural Networks",
                    "abstract": "Recently, increasing attention has been drawn to the internal mechanisms of convolutional neural networks, and the reason why the network makes specific decisions. In this paper, we develop a novel post-hoc visual explanation method called Score-CAM based on class activation mapping. Unlike previous class activation mapping based approaches, Score-CAM gets rid of the dependence on gradients by obtaining the weight of each activation map through its forward passing score on target class, the final result is obtained by a linear combination of weights and activation maps. We demonstrate that Score-CAM achieves better visual performance and fairness for interpreting the decision making process. Our approach outperforms previous methods on both recognition and localization tasks, it also passes the sanity check. We also indicate its application as debugging tools. The implementation is available1."
                },
                {
                    "title": "Towards Interpretable Object Detection by Unfolding Latent Structures",
                    "abstract": "This paper first proposes a method of formulating model interpretability in visual understanding tasks based on the idea of unfolding latent structures. It then presents a case study in object detection using popular two-stage region-based convolutional network (i.e., R-CNN) detection systems. The proposed method focuses on weakly-supervised extractive rationale generation, that is learning to unfold latent discriminative part configurations of object instances automatically and simultaneously in detection without using any supervision for part configurations. It utilizes a top-down hierarchical and compositional grammar model embedded in a directed acyclic AND-OR Graph (AOG) to explore and unfold the space of latent part configurations of regions of interest (RoIs). It presents an AOGParsing operator that seamlessly integrates with the RoIPooling/RoIAlign operator widely used in R-CNN and is trained end-to-end. In object detection, a bounding box is interpreted by the best parse tree derived from the AOG on-the-fly, which is treated as the qualitatively extractive rationale generated for interpreting detection. In experiments, Faster R-CNN is used to test the proposed method on the PASCAL VOC 2007 and the COCO 2017 object detection datasets. The experimental results show that the proposed method can compute promising latent structures without hurting the performance. The code and pretrained models are available at https://github.com/iVMCL/iRCNN."
                },
                {
                    "title": "Interpretable and Fine-Grained Visual Explanations for Convolutional Neural Networks",
                    "abstract": "To verify and validate networks, it is essential to gain insight into their decisions, limitations as well as possible shortcomings of training data. In this work, we propose a post-hoc, optimization based visual explanation method, which highlights the evidence in the input image for a specific prediction. Our approach is based on a novel technique to defend against adversarial evidence (i.e. faulty evidence due to artefacts) by filtering gradients during optimization. The defense does not depend on human-tuned parameters. It enables explanations which are both fine-grained and preserve the characteristics of images, such as edges and colors. The explanations are interpretable, suited for visualizing detailed evidence and can be tested as they are valid model inputs. We qualitatively and quantitatively evaluate our approach on a multitude of models and datasets."
                },
                {
                    "title": "Full-Gradient Representation for Neural Network Visualization",
                    "abstract": "We introduce a new tool for interpreting neural nets, namely full-gradients, which decomposes the neural net response into input sensitivity and per-neuron sensitivity components. This is the first proposed representation which satisfies two key properties: completeness and weak dependence, which provably cannot be satisfied by any saliency map-based interpretability method. Using full-gradients, we also propose an approximate saliency map representation for convolutional nets dubbed FullGrad, obtained by aggregating the full-gradient components. We experimentally evaluate the usefulness of FullGrad in explaining model behaviour with two quantitative tests: pixel perturbation and remove-and-retrain. Our experiments reveal that our method explains model behavior correctly, and more comprehensively than other methods in the literature. Visual inspection also reveals that our saliency maps are sharper and more tightly confined to object regions than other methods."
                },
                {
                    "title": "PointPillars: Fast Encoders for Object Detection From Point Clouds",
                    "abstract": "Object detection in point clouds is an important aspect of many robotics applications such as autonomous driving. In this paper, we consider the problem of encoding a point cloud into a format appropriate for a downstream detection pipeline. Recent literature suggests two types of encoders; fixed encoders tend to be fast but sacrifice accuracy, while encoders that are learned from data are more accurate, but slower. In this work, we propose PointPillars, a novel encoder which utilizes PointNets to learn a representation of point clouds organized in vertical columns (pillars). While the encoded features can be used with any standard 2D convolutional detection architecture, we further propose a lean downstream network. Extensive experimentation shows that PointPillars outperforms previous encoders with respect to both speed and accuracy by a large margin. Despite only using lidar, our full detection pipeline significantly outperforms the state of the art, even among fusion methods, with respect to both the 3D and bird\u2019s eye view KITTI benchmarks. This detection performance is achieved while running at 62 Hz: a 2 - 4 fold runtime improvement. A faster version of our method matches the state of the art at 105 Hz. These benchmarks suggest that PointPillars is an appropriate encoding for object detection in point clouds."
                },
                {
                    "title": "PointRCNN: 3D Object Proposal Generation and Detection From Point Cloud",
                    "abstract": "In this paper, we propose PointRCNN for 3D object detection from raw point cloud. The whole framework is composed of two stages: stage-1 for the bottom-up 3D proposal generation and stage-2 for refining proposals in the canonical coordinates to obtain the final detection results. Instead of generating proposals from RGB image or projecting point cloud to bird's view or voxels as previous methods do, our stage-1 sub-network directly generates a small number of high-quality 3D proposals from point cloud in a bottom-up manner via segmenting the point cloud of the whole scene into foreground points and background. The stage-2 sub-network transforms the pooled points of each proposal to canonical coordinates to learn better local spatial features, which is combined with global semantic features of each point learned in stage-1 for accurate box refinement and confidence prediction. Extensive experiments on the 3D detection benchmark of KITTI dataset show that our proposed architecture outperforms state-of-the-art methods with remarkable margins by using only point cloud as input. The code is available at https://github.com/sshaoshuai/PointRCNN."
                },
                {
                    "title": "PointCloud Saliency Maps",
                    "abstract": "3D point-cloud recognition with PointNet and its variants has received remarkable progress. A missing ingredient, however, is the ability to automatically evaluate point-wise importance w.r.t.\\! classification performance, which is usually reflected by a saliency map. A saliency map is an important tool as it allows one to perform further processes on point-cloud data. In this paper, we propose a novel way of characterizing critical points and segments to build point-cloud saliency maps. Our method assigns each point a score reflecting its contribution to the model-recognition loss. The saliency map explicitly explains which points are the key for model recognition. Furthermore, aggregations of highly-scored points indicate important segments/subsets in a point-cloud. Our motivation for constructing a saliency map is by point dropping, which is a non-differentiable operator. To overcome this issue, we approximate point-dropping with a differentiable procedure of shifting points towards the cloud centroid. Consequently, each saliency score can be efficiently measured by the corresponding gradient of the loss w.r.t the point under the spherical coordinates. Extensive evaluations on several state-of-the-art point-cloud recognition models, including PointNet, PointNet++ and DGCNN, demonstrate the veracity and generality of our proposed saliency map. Code for experiments is released on \\url{https://github.com/tianzheng4/PointCloud-Saliency-Maps}"
                },
                {
                    "title": "SECOND: Sparsely Embedded Convolutional Detection",
                    "abstract": "LiDAR-based or RGB-D-based object detection is used in numerous applications, ranging from autonomous driving to robot vision. Voxel-based 3D convolutional networks have been used for some time to enhance the retention of information when processing point cloud LiDAR data. However, problems remain, including a slow inference speed and low orientation estimation performance. We therefore investigate an improved sparse convolution method for such networks, which significantly increases the speed of both training and inference. We also introduce a new form of angle loss regression to improve the orientation estimation performance and a new data augmentation approach that can enhance the convergence speed and performance. The proposed network produces state-of-the-art results on the KITTI 3D object detection benchmarks while maintaining a fast inference speed."
                },
                {
                    "title": "RISE: Randomized Input Sampling for Explanation of Black-box Models",
                    "abstract": "Deep neural networks are being used increasingly to automate data analysis and decision making, yet their decision-making process is largely unclear and is difficult to explain to the end users. In this paper, we address the problem of Explainable AI for deep neural networks that take images as input and output a class probability. We propose an approach called RISE that generates an importance map indicating how salient each pixel is for the model's prediction. In contrast to white-box approaches that estimate pixel importance using gradients or other internal network state, RISE works on black-box models. It estimates importance empirically by probing the model with randomly masked versions of the input image and obtaining the corresponding outputs. We compare our approach to state-of-the-art importance extraction methods using both an automatic deletion/insertion metric and a pointing metric based on human-annotated object segments. Extensive experiments on several benchmark datasets show that our approach matches or exceeds the performance of other methods, including white-box approaches. \nProject page: this http URL"
                },
                {
                    "title": "Visual Explanation by Interpretation: Improving Visual Feedback Capabilities of Deep Neural Networks",
                    "abstract": "Learning-based representations have become the defacto means to address computer vision tasks. Despite their massive adoption, the amount of work aiming at understanding the internal representations learned by these models is rather limited. Existing methods aimed at model interpretation either require exhaustive manual inspection of visualizations, or link internal network activations with external \"possibly useful\" annotated concepts. We propose an intermediate scheme in which, given a pretrained model, we automatically identify internal features relevant for the set of classes considered by the model, without requiring additional annotations. We interpret the model through average visualizations of these features. Then, at test time, we explain the network prediction by accompanying the predicted class label with supporting heatmap visualizations derived from the identified relevant features. In addition, we propose a method to address the artifacts introduced by strided operations in deconvnet-based visualizations. Our evaluation on the MNIST, ILSVRC 12 and Fashion 144k datasets quantitatively shows that the proposed method is able to identify relevant internal features for the classes of interest while improving the quality of the produced visualizations."
                },
                {
                    "title": "Grad-CAM++: Generalized Gradient-Based Visual Explanations for Deep Convolutional Networks",
                    "abstract": "Over the last decade, Convolutional Neural Network (CNN) models have been highly successful in solving complex vision based problems. However, deep models are perceived as \"black box\" methods considering the lack of understanding of their internal functioning. There has been a significant recent interest to develop explainable deep learning models, and this paper is an effort in this direction. Building on a recently proposed method called Grad-CAM, we propose Grad-CAM++ to provide better visual explanations of CNN model predictions (when compared to Grad-CAM), in terms of better localization of objects as well as explaining occurrences of multiple objects of a class in a single image. We provide a mathematical explanation for the proposed method, Grad-CAM++, which uses a weighted combination of the positive partial derivatives of the last convolutional layer feature maps with respect to a specific class score as weights to generate a visual explanation for the class label under consideration. Our extensive experiments and evaluations, both subjective and objective, on standard datasets showed that Grad-CAM++ indeed provides better visual explanations for a given CNN architecture when compared to Grad-CAM."
                },
                {
                    "title": "SmoothGrad: removing noise by adding noise",
                    "abstract": "Explaining the output of a deep network remains a challenge. In the case of an image classifier, one type of explanation is to identify pixels that strongly influence the final decision. A starting point for this strategy is the gradient of the class score function with respect to the input image. This gradient can be interpreted as a sensitivity map, and there are several techniques that elaborate on this basic idea. This paper makes two contributions: it introduces SmoothGrad, a simple method that can help visually sharpen gradient-based sensitivity maps, and it discusses lessons in the visualization of these maps. We publish the code for our experiments and a website with our results."
                },
                {
                    "title": "PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space",
                    "abstract": "Few prior works study deep learning on point sets. PointNet by Qi et al. is a pioneer in this direction. However, by design PointNet does not capture local structures induced by the metric space points live in, limiting its ability to recognize fine-grained patterns and generalizability to complex scenes. In this work, we introduce a hierarchical neural network that applies PointNet recursively on a nested partitioning of the input point set. By exploiting metric space distances, our network is able to learn local features with increasing contextual scales. With further observation that point sets are usually sampled with varying densities, which results in greatly decreased performance for networks trained on uniform densities, we propose novel set learning layers to adaptively combine features from multiple scales. Experiments show that our network called PointNet++ is able to learn deep point set features efficiently and robustly. In particular, results significantly better than state-of-the-art have been obtained on challenging benchmarks of 3D point clouds."
                },
                {
                    "title": "Real Time Image Saliency for Black Box Classifiers",
                    "abstract": "In this work we develop a fast saliency detection method that can be applied to any differentiable image classifier. We train a masking model to manipulate the scores of the classifier by masking salient parts of the input image. Our model generalises well to unseen images and requires a single forward pass to perform saliency detection, therefore suitable for use in real-time systems. We test our approach on CIFAR-10 and ImageNet datasets and show that the produced saliency maps are easily interpretable, sharp, and free of artifacts. We suggest a new metric for saliency and test our method on the ImageNet object localisation task. We achieve results outperforming other weakly supervised methods."
                },
                {
                    "title": "Axiomatic Attribution for Deep Networks",
                    "abstract": "We study the problem of attributing the prediction of a deep network to its input features, a problem previously studied by several other works. We identify two fundamental axioms\u2014 Sensitivity and Implementation Invariance that attribution methods ought to satisfy. We show that they are not satisfied by most known attribution methods, which we consider to be a fundamental weakness of those methods. We use the axioms to guide the design of a new attribution method called Integrated Gradients. Our method requires no modification to the original network and is extremely simple to implement; it just needs a few calls to the standard gradient operator. We apply this method to a couple of image models, a couple of text models and a chemistry model, demonstrating its ability to debug networks, to extract rules from a network, and to enable users to engage with models better."
                },
                {
                    "title": "Learning Deep Features for Discriminative Localization",
                    "abstract": "In this work, we revisit the global average pooling layer proposed in [13], and shed light on how it explicitly enables the convolutional neural network (CNN) to have remarkable localization ability despite being trained on imagelevel labels. While this technique was previously proposed as a means for regularizing training, we find that it actually builds a generic localizable deep representation that exposes the implicit attention of CNNs on an image. Despite the apparent simplicity of global average pooling, we are able to achieve 37.1% top-5 error for object localization on ILSVRC 2014 without training on any bounding box annotation. We demonstrate in a variety of experiments that our network is able to localize the discriminative image regions despite just being trained for solving classification task1."
                },
                {
                    "title": "Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps",
                    "abstract": "This paper addresses the visualisation of image classification models, learnt using deep Convolutional Networks (ConvNets). We consider two visualisation techniques, based on computing the gradient of the class score with respect to the input image. The first one generates an image, which maximises the class score [Erhan et al., 2009], thus visualising the notion of the class, captured by a ConvNet. The second technique computes a class saliency map, specific to a given image and class. We show that such maps can be employed for weakly supervised object segmentation using classification ConvNets. Finally, we establish the connection between the gradient-based ConvNet visualisation methods and deconvolutional networks [Zeiler et al., 2013]."
                },
                {
                    "title": "Are we ready for autonomous driving? The KITTI vision benchmark suite",
                    "abstract": "Today, visual recognition systems are still rarely employed in robotics applications. Perhaps one of the main reasons for this is the lack of demanding benchmarks that mimic such scenarios. In this paper, we take advantage of our autonomous driving platform to develop novel challenging benchmarks for the tasks of stereo, optical flow, visual odometry/SLAM and 3D object detection. Our recording platform is equipped with four high resolution video cameras, a Velodyne laser scanner and a state-of-the-art localization system. Our benchmarks comprise 389 stereo and optical flow image pairs, stereo visual odometry sequences of 39.2 km length, and more than 200k 3D object annotations captured in cluttered scenarios (up to 15 cars and 30 pedestrians are visible per image). Results from state-of-the-art algorithms reveal that methods ranking high on established datasets such as Middlebury perform below average when being moved outside the laboratory to the real world. Our goal is to reduce this bias by providing challenging benchmarks with novel difficulties to the computer vision community. Our benchmarks are available online at: www.cvlibs.net/datasets/kitti."
                },
                {
                    "title": "CluB: Cluster Meets BEV for LiDAR-Based 3D Object Detection",
                    "abstract": "Currently, LiDAR-based 3D detectors are broadly categorized into two groups, namely, BEV-based detectors and cluster-based detectors. BEV-based detectors capture the contextual information from the Bird\u2019s Eye View (BEV) and fill their center voxels via feature diffusion with a stack of convolution layers, which, however, weakens the capability of presenting an object with the center point. On the other hand, cluster-based detectors exploit the voting mechanism and aggregate the foreground points into object-centric clusters for further prediction. In this paper, we explore how to effectively combine these two complementary representations into a unified framework. Specifically, we propose a new 3D object detection framework, referred to as CluB, which incorporates an auxiliary cluster-based branch into the BEV-based detector by enriching the object representation at both feature and query levels. Technically, CluB is comprised of two steps. First, we construct a cluster feature diffusion module to establish the association between cluster features and BEV features in a subtle and adaptive fashion. Based on that, an imitation loss is introduced to distill object-centric knowledge from the cluster features to the BEV features. Second, we design a cluster query generation module to leverage the voting centers directly from the cluster branch, thus enriching the diversity of object queries. Meanwhile, a direction loss is employed to encourage a more accurate voting center for each cluster. Extensive experiments are conducted on Waymo and nuScenes datasets, and our CluB achieves state-of-the-art performance on both benchmarks."
                },
                {
                    "title": "Visualizing Higher-Layer Features of a Deep Network",
                    "abstract": "Deep architectures have demonstrated state-of-the-art results in a variety of settings, especially with vision datasets. Beyond the model definitions and the quantitative analyses, there is a need for qualitative comparisons of the solutions learned by various deep architectures. The goal of this paper is to find good qualitative interpretations of high level features represented by such models. To this end, we contrast and compare several techniques applied on Stacked Denoising Autoencoders and Deep Belief Networks, trained on several vision datasets. We show that, perhaps counter-intuitively, such interpretation is possible at the unit level, that it is simple to accomplish and that the results are consistent across various techniques. We hope that such techniques will allow researchers in deep architectures to understand more of how and why deep architectures work."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively generate interpretable visual explanations for 3D object detection models using point clouds, addressing the limitations of existing methods that primarily focus on image-based models?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for enhancing the trustworthiness and transparency of 3D object detection models, particularly in high-stakes applications like autonomous driving and robotics. By providing interpretable explanations, we can facilitate better understanding and validation of model decisions, which is essential for user acceptance and safety. This research could lead to advancements in the field of explainable AI, influencing future studies to prioritize interpretability in complex models and potentially leading to more robust and reliable applications in various domains.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of point clouds, which are three-dimensional and sparsely distributed. Existing methods often generate 2D saliency maps that do not capture the 3D nature of the data, leading to inadequate interpretations. Additionally, many current techniques focus on class-specific explanations rather than providing insights into individual objects, which is necessary for practical applications. The computational intensity of perturbation-based methods, like OccAM, further complicates the task, as they require extensive inference calculations that can be resource-prohibitive.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily concentrated on image-based models, leaving a gap in the development of effective explanation methods for 3D detectors. Existing solutions often lack the capability to generate 3D saliency maps or focus on specific objects, limiting their applicability. The barriers include the complexity of point cloud data and the inadequacy of linear interpolation methods used in image-based approaches. Our approach differs by leveraging 3D feature maps and employing non-negative matrix factorization to uncover latent concepts, thus providing a more tailored and effective solution for interpreting 3D object detection models.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the use of a feature factorization activation map (FFAM) to generate visual explanations for 3D detectors. We will utilize 3D feature maps from the detector's backbone and apply non-negative matrix factorization to extract latent semantic concepts. The methodology includes generating a global concept activation map and refining it using gradients from an object-specific loss to create object-specific saliency maps."
            }
        },
        "author_data": {
            "d61a23b4-673f-425b-a584-de3435b81391": {
                "pk": "d61a23b4-673f-425b-a584-de3435b81391",
                "name": "Shuai Liu",
                "collaborators": [
                    "Andrew Eberhard",
                    "Yousong Luo",
                    "Jingbin Liu",
                    "Xinyang Gu",
                    "Jonathan May",
                    "Zhenhuan Liu",
                    "Jie Yang",
                    "Wei Liu",
                    "Mengye Lu",
                    "Gaocheng Liu"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Optimization",
                    "Machine Translation",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Style Transfer with Multi-iteration Preference Optimization",
                        "abstract": "Numerous recent techniques for text style transfer characterize their approaches as variants of reinforcement learning and preference optimization. In this work, we consider the relationship between these approaches and a class of optimization approaches developed primarily for (non-neural) statistical machine translation, formerly known as `tuning'. Inspired by these techniques from the past, we improve upon established preference optimization approaches, incorporating multiple iterations of exploration and optimization, and choosing contrastive examples by following a `hope' vs `fear' sampling strategy. Cognizant of the difference between machine translation and style transfer, however, we further tailor our framework with a new pseudo-parallel generation method and a dynamic weighted reward aggregation method to tackle the lack of parallel data and the need for a multi-objective reward. We evaluate our model on two commonly used text style transfer datasets. Through automatic and human evaluation results we show the effectiveness and the superiority of our model compared to state-of-the-art baselines."
                    },
                    {
                        "title": "T-Code: Simple Temporal Latent Code for Efficient Dynamic View Synthesis",
                        "abstract": "Novel view synthesis for dynamic scenes is one of the spotlights in computer vision. The key to efficient dynamic view synthesis is to find a compact representation to store the information across time. Though existing methods achieve fast dynamic view synthesis by tensor decomposition or hash grid feature concatenation, their mixed representations ignore the structural difference between time domain and spatial domain, resulting in sub-optimal computation and storage cost. This paper presents T-Code, the efficient decoupled latent code for the time dimension only. The decomposed feature design enables customizing modules to cater for different scenarios with individual specialty and yielding desired results at lower cost. Based on T-Code, we propose our highly compact hybrid neural graphics primitives (HybridNGP) for multi-camera setting and deformation neural graphics primitives with T-Code (DNGP-T) for monocular scenario. Experiments show that HybridNGP delivers high fidelity results at top processing speed with much less storage consumption, while DNGP-T achieves state-of-the-art quality and high training speed for monocular reconstruction."
                    },
                    {
                        "title": "The U-Lagrangian of a prox-regular function",
                        "abstract": "When restricted to a subspace, a nonsmooth function can be differentiable. It is known that for a nonsmooth convex function f and a point x, the Euclidean space can be decomposed into two subspaces: U, over which a special Lagrangian can be defined and has nice smooth properties and V, the orthogonal complement subspace of U. In this paper we generalize the definition of UV-decomposition and U-Lagrangian to the context of nonconvex functions, specifically that of a prox-regular function."
                    },
                    {
                        "title": "A Novel Distance Matric: Generalized Relative Entropy",
                        "abstract": "Information entropy and its extension, which are important generalization of entropy, have been applied in many research domains today. In this paper, a novel generalized relative entropy is constructed to avoid some defects of traditional relative entropy. We presented the structure of generalized relative entropy after the discussion of defects in relative entropy. Moreover, some properties of the provided generalized relative entropy is presented and proved. The provided generalized relative entropy is proved to have a finite range and is a finite distance metric."
                    },
                    {
                        "title": "On Partial Smoothness, Tilt Stability and the $\\mathcal{VU}$--Decomposition",
                        "abstract": "Under the assumption of prox-regularity and the presence of a tilt stable local minimum we are able to show that a $\\mathcal{VU}$ like decomposition gives rise to the existence of a smooth manifold on which the function in question coincides locally with a smooth function."
                    },
                    {
                        "title": "A version of bundle method with linear programming",
                        "abstract": "Bundle methods have been intensively studied for solving both convex and nonconvex optimization problems. In most of the bundle methods developed thus far, at least one quadratic programming (QP) subproblem needs to be solved in each iteration. In this paper, we exploit the feasibility of developing a bundle algorithm that only solves linear subproblems. We start from minimization of a convex function and show that the sequence of major iterations converge to a minimizer. For nonconvex functions we consider functions that are locally Lipschitz continuous and prox-regular on a bounded level set, and minimize the cutting-plane model over a trust region with infinity norm. The para-convexity of such functions allows us to use the locally convexified model and its convexity properties. Under some conditions and assumptions, we study the convergence of the proposed algorithm through the outer semicontinuity of the proximal mapping. Encouraging results of preliminary numerical experiments on standard test sets are provided."
                    },
                    {
                        "title": "Reinforcement learning with world model",
                        "abstract": "Nowadays, model-free reinforcement learning algorithms have achieved remarkable performance on many decision making and control tasks, but high sample complexity and low sample efficiency still hinder the wide use of model-free reinforcement learning algorithms. In this paper, we argue that if we intend to design an intelligent agent that learns fast and transfers well, the agent must be able to reflect key elements of intelligence, like intuition, Memory, PredictionandCuriosity. We propose an agent framework that integrates off-policy reinforcement learning with world model learning, so as to embody the important features of intelligence in our algorithm design. We adopt the state-of-art model-free reinforcement learning algorithm, Soft Actor-Critic, as the agent intuition, and world model learning through RNN to endow the agent with memory, curiosity, and the ability to predict. We show that these ideas can work collaboratively with each other and our agent (RMC) can give new state-of-art results while maintaining sample efficiency and training stability. Moreover, our agent framework can be easily extended from MDP to POMDP problems without performance loss."
                    },
                    {
                        "title": "Holographic s-wave and p-wave Josephson junction with backreaction",
                        "abstract": "In this paper, we study the holographic models of s-wave and p-wave Josephoson junction away from probe limit in (3+1)-dimensional spacetime, respectively. With the backreaction of the matter, we obtained the anisotropic black hole solution with the condensation of matter fields. We observe that the critical temperature of Josephoson junction decreases with increasing backreaction. In addition to this, the tunneling current and condenstion of Josephoson junction become smaller as backreaction grows larger, but the relationship between current and phase difference still holds for sine function. Moreover, condenstion of Josephoson junction deceases with increasing width of junction exponentially."
                    },
                    {
                        "title": "Policy Optimization Reinforcement Learning with Entropy Regularization",
                        "abstract": "Entropy regularization is an important idea in reinforcement learning, with great success in recent algorithms like Soft Q Network (SQN) and Soft Actor-Critic (SAC1). In this work, we extend this idea into the on-policy realm. We propose the soft policy gradient theorem (SPGT) for on-policy maximum entropy reinforcement learning. With SPGT, a series of new policy optimization algorithms are derived, such as SPG, SA2C, SA3C, SDDPG, STRPO, SPPO, SIMPALA and so on. We find that SDDPG is equivalent to SAC1. For policy gradient, the policy network is often represented as a Gaussian distribution with a global action variance, which damages the representation capacity. We introduce a local action variance for policy network and find it can work collaboratively with the idea of entropy regularization. Our method outperforms prior works on a range of benchmark tasks. Furthermore, our method can be easily extended to large scale experiment with great stability and parallelism."
                    }
                ]
            },
            "0a58a636-28b6-432a-a19a-f93a2470b63f": {
                "pk": "0a58a636-28b6-432a-a19a-f93a2470b63f",
                "name": "Boyang Li",
                "collaborators": [
                    "Yiyu Shi",
                    "Haoxin Li",
                    "Yuan Liu",
                    "Tong Wang",
                    "Dejing Dou",
                    "Xu Guo",
                    "Devaansh Gupta",
                    "Hanwang Zhang",
                    "Ping Chen",
                    "Beth Cardier"
                ],
                "domain": [
                    "Computer Vision",
                    "Natural Language Processing",
                    "Explainable AI",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "New method for coherent imaging using incompatible sources",
                        "abstract": "Non-invasive imaging plays a crucial role in the early detection, diagnosis, and treatment of numerous medical conditions. This article discusses recent advancements in coherent imaging techniques.The article begins by discussing the evolution of CDI technology, focusing on its higher resolution capabilities for detailed analysis and the use of white light in modern systems. Subsequently, advancements in CDI techniques are examined, such as imaging for white matter tract analysis. And its applications in real-time visualization of internal structures is highlighted, along with its application to various samples. Finally introduce a method to use coherent imaging method on incompatible non-coherent sources."
                    },
                    {
                        "title": "A Training Data Recipe to Accelerate A* Search with Language Models",
                        "abstract": "Combining Large Language Models (LLMs) with heuristic search algorithms like A* holds the promise of enhanced LLM reasoning and scalable inference. To accelerate training and reduce computational demands, we investigate the coreset selection problem for the training data of LLM heuristic learning. Few methods to learn the heuristic functions consider the interaction between the search algorithm and the machine learning model. In this work, we empirically disentangle the requirements of A* search algorithm from the requirements of the LLM to generalise on this task. Surprisingly, we find an overlap between their requirements; A* requires more accurate predictions on search nodes near the goal, and LLMs need the same set of nodes for effective generalisation. With these insights, we derive a data-selection distribution for learning LLM-based heuristics. On three classical planning domains, maze navigation, Sokoban and sliding tile puzzles, our technique reduces the number of iterations required to find the solutions by up to 15x, with a wall-clock speed-up of search up to 5x. The codebase is at https://github.com/devaansh100/a_star."
                    },
                    {
                        "title": "Mitigating and Evaluating Static Bias of Action Representations in the Background and the Foreground",
                        "abstract": "In video action recognition, shortcut static features can interfere with the learning of motion features, resulting in poor out-of-distribution (OOD) generalization. The video background is clearly a source of static bias, but the video foreground, such as the clothing of the actor, can also provide static bias. In this paper, we empirically verify the existence of foreground static bias by creating test videos with conflicting signals from the static and moving portions of the video. To tackle this issue, we propose a simple yet effective technique, StillMix, to learn robust action representations. Specifically, StillMix identifies bias-inducing video frames using a 2D reference network and mixes them with videos for training, serving as effective bias suppression even when we cannot explicitly extract the source of bias within each video frame or enumerate types of bias. Finally, to precisely evaluate static bias, we synthesize two new benchmarks, SCUBA for static cues in the background, and SCUFO for static cues in the foreground. With extensive experiments, we demonstrate that StillMix mitigates both types of static bias and improves video representations for downstream applications. Code is available at https://github.com/lihaoxin05/StillMix."
                    },
                    {
                        "title": "Predicting the Quality of Short Narratives from Social Media",
                        "abstract": "An important and difficult challenge in building computational models for narratives is the automatic evaluation of narrative quality. Quality evaluation connects narrative understanding and generation as generation systems need to evaluate their own products. To circumvent difficulties in acquiring annotations, we employ upvotes in social media as an approximate measure for story quality. We collected 54,484 answers from a crowd-powered question-and-answer website, Quora, and then used active learning to build a classifier that labeled 28,320 answers as stories. To predict the number of upvotes without the use of social network features, we create neural networks that model textual regions and the interdependence among regions, which serve as strong benchmarks for future research. To our best knowledge, this is the first large-scale study for automatic evaluation of narrative quality."
                    },
                    {
                        "title": "Annotating High-Level Structures of Short Stories and Personal Anecdotes",
                        "abstract": "Stories are a vital form of communication in human culture; they are employed daily to persuade, to elicit sympathy, or to convey a message. Computational understanding of human narratives, especially high-level narrative structures, remain limited to date. Multiple literary theories for narrative structures exist, but operationalization of the theories has remained a challenge. We developed an annotation scheme by consolidating and extending existing narratological theories, including Labov and Waletsky's (1967) functional categorization scheme and Freytag's (1863) pyramid of dramatic tension, and present 360 annotated short stories collected from online sources. In the future, this research will support an approach that enables systems to intelligently sustain complex communications with humans."
                    },
                    {
                        "title": "A Neural Multi-sequence Alignment TeCHnique (NeuMATCH)",
                        "abstract": "The alignment of heterogeneous sequential data (video to text) is an important and challenging problem. Standard techniques for this task, including Dynamic Time Warping (DTW) and Conditional Random Fields (CRFs), suffer from inherent drawbacks. Mainly, the Markov assumption implies that, given the immediate past, future alignment decisions are independent of further history. The separation between similarity computation and alignment decision also prevents end-to-end training. In this paper, we propose an end-to-end neural architecture where alignment actions are implemented as moving data between stacks of Long Short-term Memory (LSTM) blocks. This flexible architecture supports a large variety of alignment tasks, including one-to-one, one-to-many, skipping unmatched elements, and (with extensions) non-monotonic alignment. Extensive experiments on semi-synthetic and real datasets show that our algorithm outperforms state-of-the-art baselines."
                    },
                    {
                        "title": "Privacy-Aware Cost-Effective Scheduling Considering Non-Schedulable Appliances in Smart Home",
                        "abstract": "A Smart Home provides integrating and electronic information services to help residential users manage their energy usage and bill cost but also exposes users to significant privacy risks due to fine-grained information collected by smart meters. Taking account of users' privacy concerns, this paper focuses on cost-effective runtime scheduling designed for schedulable and non-schedulable appliances. To alleviate the influence of operation uncertainties introduced by non-schedulable appliances, we formulate the problem by minimizing the expected sum of electricity cost under the worst privacy situation. Inventing the iterative alternative algorithm, we effectively schedule the appliances and rechargeable battery in a cost-effective way while satisfying users' privacy requirement. Experimental evaluation based on real-world data demonstrates the effectiveness of the proposed algorithm."
                    },
                    {
                        "title": "Searching for Stage-wise Neural Graphs In the Limit",
                        "abstract": "Search space is a key consideration for neural architecture search. Recently, Xie et al. (2019) found that randomly generated networks from the same distribution perform similarly, which suggests we should search for random graph distributions instead of graphs. We propose graphon as a new search space. A graphon is the limit of Cauchy sequence of graphs and a scale-free probabilistic distribution, from which graphs of different number of nodes can be drawn. By utilizing properties of the graphon space and the associated cut-distance metric, we develop theoretically motivated techniques that search for and scale up small-capacity stage-wise graphs found on small datasets to large-capacity graphs that can handle ImageNet. The scaled stage-wise graphs outperform DenseNet and randomly wired Watts-Strogatz networks, indicating the benefits of graphon theory in NAS applications."
                    },
                    {
                        "title": "GridPE: Unifying Positional Encoding in Transformers with a Grid Cell-Inspired Framework",
                        "abstract": "Understanding spatial location and relationships is a fundamental capability for modern artificial intelligence systems. Insights from human spatial cognition provide valuable guidance in this domain. Neuroscientific discoveries have highlighted the role of grid cells as a fundamental neural component for spatial representation, including distance computation, path integration, and scale discernment. In this paper, we introduce a novel positional encoding scheme inspired by Fourier analysis and the latest findings in computational neuroscience regarding grid cells. Assuming that grid cells encode spatial position through a summation of Fourier basis functions, we demonstrate the translational invariance of the grid representation during inner product calculations. Additionally, we derive an optimal grid scale ratio for multi-dimensional Euclidean spaces based on principles of biological efficiency. Utilizing these computational principles, we have developed a Grid-cell inspired Positional Encoding technique, termed GridPE, for encoding locations within high-dimensional spaces. We integrated GridPE into the Pyramid Vision Transformer architecture. Our theoretical analysis shows that GridPE provides a unifying framework for positional encoding in arbitrary high-dimensional spaces. Experimental results demonstrate that GridPE significantly enhances the performance of transformers, underscoring the importance of incorporating neuroscientific insights into the design of artificial intelligence systems."
                    },
                    {
                        "title": "An Empirical Study on the Relation between Network Interpretability and Adversarial Robustness",
                        "abstract": "Deep neural networks (DNNs) have had many successes, but they suffer from two major issues: (1) a vulnerability to adversarial examples and (2) a tendency to elude human interpretation. Interestingly, recent empirical and theoretical evidence suggests these two seemingly disparate issues are actually connected. In particular, robust models tend to provide more interpretable gradients than non-robust models. However, whether this relationship works in the opposite direction remains obscure. With this paper, we seek empirical answers to the following question: can models acquire adversarial robustness when they are trained to have interpretable gradients? We introduce a theoretically inspired technique called Interpretation Regularization (IR), which encourages a model's gradients to (1) match the direction of interpretable target salience maps and (2) have small magnitude. To assess model performance and tease apart factors that contribute to adversarial robustness, we conduct extensive experiments on MNIST and CIFAR-10 with both $\\ell_2$ and $\\ell_\\infty$ attacks. We demonstrate that training the networks to have interpretable gradients improves their robustness to adversarial perturbations. Applying the network interpretation technique SmoothGrad yields additional performance gains, especially in cross-norm attacks and under heavy perturbations. The results indicate that the interpretability of the model gradients is a crucial factor for adversarial robustness. Code for the experiments can be found at https://github.com/a1noack/interp_regularization."
                    },
                    {
                        "title": "Training on Synthetic Data Beats Real Data in Multimodal Relation Extraction",
                        "abstract": "The task of multimodal relation extraction has attracted significant research attention, but progress is constrained by the scarcity of available training data. One natural thought is to extend existing datasets with cross-modal generative models. In this paper, we consider a novel problem setting, where only unimodal data, either text or image, are available during training. We aim to train a multimodal classifier from synthetic data that perform well on real multimodal test data. However, training with synthetic data suffers from two obstacles: lack of data diversity and label information loss. To alleviate the issues, we propose Mutual Information-aware Multimodal Iterated Relational dAta GEneration (MI2RAGE), which applies Chained Cross-modal Generation (CCG) to promote diversity in the generated data and exploits a teacher network to select valuable training samples with high mutual information with the ground-truth labels. Comparing our method to direct training on synthetic data, we observed a significant improvement of 24.06% F1 with synthetic text and 26.42% F1 with synthetic images. Notably, our best model trained on completely synthetic images outperforms prior state-of-the-art models trained on real multimodal data by a margin of 3.76% in F1. Our codebase will be made available upon acceptance."
                    },
                    {
                        "title": "Energy-recycling Blockchain with Proof-of-Deep-Learning",
                        "abstract": "An enormous amount of energy is wasted in Proofof-Work (PoW) mechanisms adopted by popular blockchain applications (e.g., PoW-based cryptocurrencies), because miners must conduct a large amount of computation. Owing to this, one serious rising concern is that the energy waste not only dilutes the value of the blockchain but also hinders its further application. In this paper, we propose a novel blockchain design that fully recycles the energy required for facilitating and maintaining it, which is re-invested to the computation of deep learning. We realize this by proposing Proof-of-Deep-Learning (PoDL) such that a valid proof for a new block can be generated if and only if a proper deep learning model is produced. We present a proof-of-concept design of PoDL that is compatible with the majority of the cryptocurrencies that are based on hash-based PoW mechanisms. Our benchmark and simulation results show that the proposed design is feasible for various popular cryptocurrencies such as Bitcoin, Bitcoin Cash, and Litecoin."
                    },
                    {
                        "title": "Real-Time Adversarial Attacks",
                        "abstract": "In recent years, many efforts have demonstrated that modern machine learning algorithms are vulnerable to adversarial attacks, where small, but carefully crafted, perturbations on the input can make them fail. While these attack methods are very effective, they only focus on scenarios where the target model takes static input, i.e., an attacker can observe the entire original sample and then add a perturbation at any point of the sample. These attack approaches are not applicable to situations where the target model takes streaming input, i.e., an attacker is only able to observe past data points and add perturbations to the remaining (unobserved) data points of the input. In this paper, we propose a real-time adversarial attack scheme for machine learning models with streaming inputs."
                    },
                    {
                        "title": "Proof of Learning (PoLe): Empowering Machine Learning with Consensus Building on Blockchains",
                        "abstract": "The progress of deep learning (DL), especially the recent development of automatic design of networks, has brought unprecedented performance gains at heavy computational cost. On the other hand, blockchain systems routinely perform a huge amount of computation that does not achieve practical purposes in order to build Proof-of-Work (PoW) consensus from decentralized participants. In this paper, we propose a new consensus mechanism, Proof of Learning (PoLe), which directs the computation spent for consensus toward optimization of neural networks (NN). In our mechanism, the training/testing data are released to the entire blockchain network (BCN) and the consensus nodes train NN models on the data, which serves as the proof of learning. When the consensus on the BCN considers a NN model to be valid, a new block is appended to the blockchain. We experimentally compare the PoLe protocol with Proof of Work (PoW) and show that PoLe can achieve a more stable block generation rate, which leads to more efficient transaction processing. We also introduce a novel cheating prevention mechanism, Secure Mapping Layer (SML), which can be straightforwardly implemented as a linear NN layer. Empirical evaluation shows that SML can detect cheating nodes at small cost to the predictive performance."
                    },
                    {
                        "title": "Improving the Sample Efficiency of Prompt Tuning with Domain Adaptation",
                        "abstract": "Prompt tuning, or the conditioning of a frozen pretrained language model (PLM) with soft prompts learned from data, has demonstrated impressive performance on a wide range of NLP tasks. However, prompt tuning requires a large training dataset to be effective and is outperformed by finetuning the entire PLM in data-scarce regimes. Previous work (Gu et al., 2022, Vu et al., 2022) proposed to transfer soft prompts pretrained on the source domain to the target domain. In this paper, we explore domain adaptation for prompt tuning, a problem setting where unlabeled data from the target domain are available during pretraining. We propose bOosting Prompt TunIng with doMain Adaptation (OPTIMA), which regularizes the decision boundary to be smooth around regions where source and target data distributions are similar. Extensive experiments demonstrate that OPTIMA significantly enhances the transferability and sample-efficiency of prompt tuning compared to strong baselines. Moreover, in few-shot settings, OPTIMA exceeds full-model tuning by a large margin."
                    },
                    {
                        "title": "Movie Box Office Prediction With Self-Supervised and Visually Grounded Pretraining",
                        "abstract": "Investments in movie production are associated with a high level of risk as movie revenues have long-tailed and bimodal distributions. Accurate prediction of box-office revenue may mitigate the uncertainty and encourage investment. However, learning effective representations for actors, directors, and user-generated content-related keywords remains a challenging open problem. In this work, we investigate the effects of self-supervised pretraining and propose visual grounding of content keywords in objects from movie posters as a pertaining objective. Experiments on a large dataset of 35,794 movies demonstrate significant benefits of self-supervised training and visual grounding. In particular, visual grounding pretraining substantially improves learning on movies with content keywords and achieves 14.5% relative performance gains compared to a finetuned BERT model with identical architecture."
                    },
                    {
                        "title": "May I Ask a Follow-up Question? Understanding the Benefits of Conversations in Neural Network Explainability",
                        "abstract": "Research in explainable AI (XAI) aims to provide insights into the decision-making process of opaque AI models. To date, most XAI methods offer one-off and static explanations, which cannot cater to the diverse backgrounds and understanding levels of users. With this paper, we investigate if free-form conversations can enhance users' comprehension of static explanations, improve acceptance and trust in the explanation methods, and facilitate human-AI collaboration. Participants are presented with static explanations, followed by a conversation with a human expert regarding the explanations. We measure the effect of the conversation on participants' ability to choose, from three machine learning models, the most accurate one based on explanations and their self-reported comprehension, acceptance, and trust. Empirical results show that conversations significantly improve comprehension, acceptance, trust, and collaboration. Our findings highlight the importance of customized model explanations in the format of free-form conversations and provide insights for the future design of conversational explanations."
                    }
                ]
            },
            "a013d2a0-7ce7-421f-aa59-14cbf82255f7": {
                "pk": "a013d2a0-7ce7-421f-aa59-14cbf82255f7",
                "name": "Zhiyu Fang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "8fbcf710-bf31-4642-befb-437028b7ca76": {
                "pk": "8fbcf710-bf31-4642-befb-437028b7ca76",
                "name": "Mingyue Cui",
                "collaborators": [
                    "Daosong Hu",
                    "Ruomeng Wang",
                    "Liang Zhao",
                    "Song Ding",
                    "Kai Huang"
                ],
                "domain": [
                    "Medical Imaging",
                    "3D Reconstruction",
                    "Deep Learning",
                    "Vascular Modeling"
                ],
                "publications": [
                    {
                        "title": "4D-CAT: Synthesis of 4D Coronary Artery Trees from Systole and Diastole",
                        "abstract": "The three-dimensional vascular model reconstructed from CT images is widely used in medical diagnosis. At different phases, the beating of the heart can cause deformation of vessels, resulting in different vascular imaging states and false positive diagnostic results. The 4D model can simulate a complete cardiac cycle. Due to the dose limitation of contrast agent injection in patients, it is valuable to synthesize a 4D coronary artery trees through finite phases imaging. In this paper, we propose a method for generating a 4D coronary artery trees, which maps the systole to the diastole through deformation field prediction, interpolates on the timeline, and the motion trajectory of points are obtained. Specifically, the centerline is used to represent vessels and to infer deformation fields using cube-based sorting and neural networks. Adjacent vessel points are aggregated and interpolated based on the deformation field of the centerline point to obtain displacement vectors of different phases. Finally, the proposed method is validated through experiments to achieve the registration of non-rigid vascular points and the generation of 4D coronary trees."
                    }
                ]
            },
            "f1bab3ca-0d6c-4080-8a61-68a0bdb4cd60": {
                "pk": "f1bab3ca-0d6c-4080-8a61-68a0bdb4cd60",
                "name": "Kai Huang",
                "collaborators": [
                    "Izabela Szlufarska",
                    "Felix Rech",
                    "Ingo Rehberg",
                    "Philipp Ramming",
                    "Manuel Baur",
                    "Andreas Zippelius",
                    "Yongqi Liang",
                    "Thomas M\u00fcller",
                    "Simeon V\u00f6lkel",
                    "Andrea Fortini"
                ],
                "domain": [
                    "Granular Materials",
                    "Acoustics",
                    "Molecular Dynamics",
                    "Stochastic Optimization"
                ],
                "publications": [
                    {
                        "title": "Impulse response of the Bayreuth Festspielhaus",
                        "abstract": "The Bayreuth Festspielhaus is well known for its architecture because its design is heavily influenced by composer Richard Wagner. Due to the special acoustic design, the reverberation time (i.e., time scale for the sound pressure level to decay $60$~dB) is larger than usual opera houses. Using hand-claps and smart phone recordings, I measured the impulse response of the Bayreuth Festspielhaus in the auditorium, on the stage, as well as in the orchestra pit. The measured reverberation time shows quantitative agreement with the literature values within a certain frequency range, demonstrating the possibility of using this approach to monitor room acoustics."
                    },
                    {
                        "title": "Internal and surface waves in vibrofluidized granular materials: Role of cohesion",
                        "abstract": "Wave phenomena in vibrofluidized dry and partially wet granular materials confined in a quasi-two-dimensional geometry are investigated with numerical simulations considering individual particles as hard spheres. Short ranged cohesive interactions arising from the formation of liquid bridges between adjacent particles are modeled by changing the velocity dependent coefficient of restitution. Such a change effectively suppresses the formation of surface waves, in agreement with previous experimental observations. The difference in pattern creation arises from the suppressed momentum transfer due to wetting and it can be quantitatively understood from an analysis of binary impacts."
                    },
                    {
                        "title": "$1/f$ noise on the brink of wet granular melting",
                        "abstract": "The collective behavior of a two-dimensional wet granular cluster under horizontal swirling motions is investigated experimentally. Depending on the balance between the energy injection and dissipation, the cluster evolves into various nonequilibrium stationary states with strong internal structure fluctuations with time. Quantitative characterizations of the fluctuations with the bond orientational order parameter $q_{\\rm 6}$ reveal power spectra of the form $f^{\\alpha}$ with the exponent $\\alpha$ closely related to the stationary states of the system. In particular, $1/f$ type of noise with $\\alpha\\approx-1$ emerges as melting starts from the free surface of the cluster, suggesting the possibility of using $1/f$ noise as an indicator for phase transitions in systems driven far from thermodynamic equilibrium."
                    },
                    {
                        "title": "Period tripling causes rotating spirals in agitated wet granular layers",
                        "abstract": "Pattern formation of a thin layer of vertically agitated wet granular matter is investigated experimentally. Rotating spirals with three arms, which correspond to the kinks between regions with different colliding phases, are the dominating pattern. This preferred number of arms corresponds to period tripling of the agitated granular layer, unlike predominantly subharmonic Faraday crispations in dry granular matter. The chirality of the spatiotemporal pattern corresponds to the rotation direction of the spirals."
                    },
                    {
                        "title": "Clustering and melting in a wet granular monolayer",
                        "abstract": "We investigate experimentally the collective behavior of a wet granular monolayer under vertical vibrations. The spherical particles are partially wet such that there are short-ranged attractive interactions between adjacent particles. As the vibration strength increases, clustering, reorganizing and melting regimes are identified subsequently through a characterization with the bond-orientational order parameters and the mean kinetic energy of the particles. The melting transition is found to be a continuous process starting from the defects inside the crystal."
                    },
                    {
                        "title": "Green-Kubo relation for friction at liquid-solid interfaces",
                        "abstract": "We have developed a Green-Kubo relation that enables accurate calculations of friction at solid-liquid interfaces directly from equilibrium molecular dynamics (MD) simulations and that provides a pathway to bypass the time-scale limitations of typical nonequilibrium MD simulations. The theory has been validated for a number of different interfaces and it is demonstrated that the liquid-solid slip is an intrinsic property of an interface. Because of the high numerical efficiency of our method, it can be used in the design of interfaces for applications in aqueous environments, such as nano- and microfluidics."
                    },
                    {
                        "title": "Dynamics of wet granular hexagons",
                        "abstract": "The collective behavior of vibrated hexagonal disks confined in a monolayer is investigated experimentally. Due to the broken circular symmetry, hexagons prefer to rotate upon sufficiently strong driving. Due to the formation of liquid bridges, short-ranged cohesive interactions are introduced upon wetting. Consequently, a nonequilibrium stationary state with the rotating disks self-organized in a hexagonal structure arises. The bond length of the hexagonal structure is slightly smaller than the circumdiameter of a hexagon, indicating geometric frustration. This investigation provides an example where the collective behavior of granular matter is tuned by the shape of individual particles."
                    },
                    {
                        "title": "Density-wave fronts on the brink of wet granular condensation",
                        "abstract": "Density-wave fronts in a vibrofluidized wet granular layer undergoing a gas-liquid-like transition are investigated experimentally. The threshold of the instability is governed by the amplitude of the vertical vibrations. Fronts, which are curved into a spiral shape, propagate coherently along the circular rim of the container with leading edges. They are stable beyond a critical distance from the container center. Based on an analysis of the emerging time and length scales, we propose a model for the pattern formation by considering the competition between the time scale for the condensation of wet granular particles from a gas-like state and that of the energy injection resisting this process."
                    },
                    {
                        "title": "Hasse principle violation for algebraic families of del Pezzo surfaces of degree 4 and hyperelliptic curves of genus congruent to 1 modulo 4",
                        "abstract": "Let g be a positive integer congruent to 1 modulo 4 and K be an arbitrary number field. We construct infinitely many explicit one-parameter algebraic families of degree 4 del Pezzo surfaces and of genus g hyperelliptic curves such that each K-member of the families violates the Hasse principle. In particular, we obtain algebraic families of non-trivial 2-torsion elements in the Tate-Shafarevich group of elliptic curves over K. These Hasse principle violations are explained by the Brauer-Manin obstruction."
                    },
                    {
                        "title": "Influence of the liquid film thickness on the coefficient of restitution for wet particles",
                        "abstract": "The normal coefficient of restitution (COR) for a spherical particle bouncing on a wet plane is investigated experimentally and compared with a model characterizing the energy loss at impact. For fixed ratios of liquid film thickness $\\delta$ to particle diameter $D$, the wet COR is always found to decay linearly with St$^{-1}$, where St, the Stokes number, measures the particle inertia with respect to the viscous force of the liquid. Such a dependency suggests a convenient way of predicting the wet COR with two fit parameters: A critical COR at infinitely large St and a critical St at zero COR. We characterize the dependency of the two parameters on $\\delta/D$ and compare it with a model considering the energy loss from the inertia and the viscosity of the wetting liquid. This investigation suggests an analytical prediction of the COR for wet particles."
                    },
                    {
                        "title": "Dynamics of wetting explored with inkjet printing",
                        "abstract": "An inkjet printer head, which is capable of depositing liquid droplets with a resolution of $22$ picoliters and high repeatability, is employed to investigate the wetting dynamics of drops printed on a horizontal plane as well as on a granular monolayer. For a sessile drop on a horizontal plane, we characterize the contact angle hysteresis, drop volume and contact line dynamics from side view images. We show that the evaporation rate scales with the dimension of the contact line instead of the surface area of the drop. We demonstrate that the system evolves into a closed cycle upon repeating the depositing-evaporating process, owing to the high repeatability of the printing facility. Finally, we extend the investigation to a granular monolayer in order to explore the interplay between liquid deposition and granular particles."
                    },
                    {
                        "title": "Radar for projectile impact on granular media",
                        "abstract": "From the prevention of natural disasters such as landslide and avalanches, to the enhancement of energy efficiencies in chemical and civil engineering industries, understanding the collective dynamics of granular materials is a fundamental question that are closely related to our daily lives. Using a recently developed multi-static radar system operating at $10$\\,GHz (X-band), we explore the possibility of tracking a projectile moving inside a granular medium, focusing on possible sources of uncertainties in the detection and reconstruction processes. On the one hand, particle tracking with continuous wave radar provides an extremely high temporal resolution. On the other hand, there are still challenges in obtaining tracer trajectories accurately. We show that some of the challenges can be resolved through a correction of the IQ mismatch in the raw signals obtained. Consequently, the tracer trajectories can be obtained with sub-millimeter spatial resolution. Such an advance can not only shed light on radar particle tracking, but also on a wide range of scenarios where issues relevant to IQ mismatch arise."
                    },
                    {
                        "title": "Friction and slip at solid/liquid interface in vibrational systems",
                        "abstract": "Molecular dynamics simulations have been performed to study frictional slip and its influence on energy dissipation and momentum transfer at atomically smooth solid/water interfaces. By modifying surface chemistry, we investigate the relationship between slip and the mechanical response of a vibrating solid for both hydrophilic and hydrophobic surfaces. We discover physical phenomena that emerge at high frequencies and that have significant contributions to energy dissipation. A new analytical model is developed to describe mechanical response of the resonators in this high frequency regime, which is relevant in such applications as MEMS-based biosensors. We find a linear relationship between the slip length and the ratio of the damping rate shift to resonant frequency shift, which provides a new way to obtain information about slip length from experiments."
                    },
                    {
                        "title": "Effect of interfaces on the nearby Brownian motion",
                        "abstract": "Near-boundary Brownian motion is a classic hydrodynamic problem of great importance in a variety of fields, from biophysics to micro-/nanofluidics. However, due to challenges in experimental measurements of near-boundary dynamics, the effect of interfaces on Brownian motion has remained elusive. Here, we report a computational study of this effect using microsecond-long large-scale molecular dynamics simulations and our newly developed Green-Kubo relation for friction at the liquid-solid interface. Our computer experiment unambiguously reveals that the t^(-3/2) long-time decay of the velocity autocorrelation function of a Brownian particle in bulk liquid is replaced by a t^(-5/2) decay near a boundary. We discover a general breakdown of traditional no-slip boundary condition at short time scales and we show that this breakdown has a profound impact on the near-boundary Brownian motion. Our results demonstrate the potential of Brownian-particle based micro-/nano-sonar to probe the local wettability of liquid-solid interfaces."
                    },
                    {
                        "title": "Correction of IQ mismatch for a particle tracking radar",
                        "abstract": "For a better understanding of granular flow problems such as silo blockage, avalanche triggering, mixing and segregation, it is essential to have a `microscopic' view of individual particles. In order to cope with the difficulty arising from the opacity of granular materials, such as sands, powders and grains, a small scale bi-static radar system operating at $10$\\,GHz (X-band) was recently introduced to trace a sub-centimeter particle in three dimensions. Similar to a moving target indicator radar, the relative movement of the tracer with respect to each of the three receiving antennae is obtained via comparing the phase shift of the electromagnetic wave traveling through the target area with an IQ-Mixer. From the azimuth and tilt angles of the receiving antennae obtained in the calibration, the target trajectory in a three-dimensional Cartesian system is reconstructed. Using a free-falling sphere as a test case, we discuss the accuracy of this radar system and possible ways to enhance it by IQ mismatch corrections."
                    },
                    {
                        "title": "Role of defects in the onset of wall-induced granular convection",
                        "abstract": "We investigate the onset of the wall-induced convection in vertically vibrated granular matter by means of experiments and two-dimensional computer simulations. In both simulations and experiments we find that the wall-induced convection occurs inside the bouncing bed region of the parameter space in which the granular bed behaves like a bouncing ball. A good agreement between experiments and simulations is found for the peak vibration acceleration at which convection starts. By comparing the results of simulations initialised with and without defects, we find that the onset of convection occurs at lower vibration strengths in the presence of defects. Furthermore, we find that the convection of granular particles initialised in a perfect hexagonal lattice is related to the nucleation of defects and the process is described by an Arrhenius law."
                    },
                    {
                        "title": "A Multi-stage Stochastic Programming Approach for Network Capacity Expansion with Multiple Sources of Capacity",
                        "abstract": "In networks, there are often more than one source of capacity. The capacities can be permanently or temporarily owned by the decision maker. Depending on the nature of sources, we identify the permanent capacity, spot market capacity and contract capacity. We use a scenario tree to model the uncertainty, and build a multi-stage stochastic integer program that can incorporate multiple sources and multiple types of capacities in a general network. We propose two solution methodologies for the problem. Firstly, we design an asymptotically convergent approximation algorithm. Secondly, we design a cutting plane algorithm based on Benders decomposition to find tight bounds for the problem. The numerical experiments show superb performance of the proposed algorithms compared with commercial software."
                    }
                ]
            }
        }
    },
    "2407.09941": {
        "paper_data": {
            "title": "Hydra: Bidirectional State Space Models Through Generalized Matrix Mixers",
            "url": "http://arxiv.org/abs/2407.09941v1",
            "arxiv_id": "2407.09941",
            "authors": [
                "Sukjun Hwang",
                "Aakash Lahoti",
                "Tri Dao",
                "Albert Gu"
            ],
            "abstract": "A wide array of sequence models are built on a framework modeled after Transformers, comprising alternating sequence mixer and channel mixer layers. This paper studies a unifying matrix mixer view of sequence mixers that can be conceptualized as a linear map on the input sequence. This framework encompasses a broad range of well-known sequence models, including the self-attention of Transformers as well as recent strong alternatives such as structured state space models (SSMs), and allows understanding downstream characteristics such as efficiency and expressivity through properties of their structured matrix class. We identify a key axis of matrix parameterizations termed sequence alignment, which increases the flexibility and performance of matrix mixers, providing insights into the strong performance of Transformers and recent SSMs such as Mamba. Furthermore, the matrix mixer framework offers a systematic approach to developing sequence mixers with desired properties, allowing us to develop several new sub-quadratic sequence models. In particular, we propose a natural bidirectional extension of the Mamba model (Hydra), parameterized as a quasiseparable matrix mixer, which demonstrates superior performance over other sequence models including Transformers on non-causal tasks. As a drop-in replacement for attention layers, Hydra outperforms BERT by 0.8 points on the GLUE benchmark and ViT by 2% Top-1 accuracy on ImageNet.",
            "introduction": " Introduction of sequence alignment. Within this framework, we define sequence alignment as a novel attribute, which induces sequence models with essential features of data-dependence andextendability . We introduce Sequence Aligned Matrices (SAM) that exhibit these properties, demonstrating their superior performance in downstream tasks compared to non-aligned counterparts (Section 2.2). (iii)Exploration of structured matrix parameterizations. Through the matrix mixer framework, we systematically explore and categorize a broad spectrum of existing sequence models (Section 2.3). Motivated by sequence alignment, we also present new sequence models with underexplored structured matrix configurations, such as Vandermonde and Cauchy matrices (Section 2.4). Building on our matrix mixer framework, we introduce a novel sequence model Hydra , which employs a quasiseparable matrix mixer (Section 3). Quasiseparable matrices are a fundamental matrix structure with several important properties, making them ideal for sequence modeling. For example, they generalize both the low-rank matrix mixers found in linear attention models and the semiseparable matrices utilized in state space models. In fact, quasiseparable matrix mixers can be seen as a natural bidirectional extension of semiseparable matrices, addressing a major limitation in state space models by making their use in non-causal settings possible. Additionally, Hydra maintains the strong performance and linear-time computational efficiency of SSMs, thanks to structured matrix multiplication algorithms. Unlike prior attempts to make SSMs bidirectional by ad-hoc methods, we provide PyTorch codes in Figure 6, Figure 7, Fig- ure 8, Figure 9, Figure 10, and Figure 11. Our primary focus on this ablation is the comparison of expressivity between different Matrix Mixers. Therefore, our implementations do not necessarily adopt algorithms with efficient computational complexities for simplicity. The abbreviation distands for data-independent and ddrepresents data-dependent, in which ddvariants are equipped with the SAM property. As the Quasiseparable variant is equivalent to the Hydra blocks, its implementation can be found in the Python file provided in the supplementary. Further specific details are described below. Dense. While extending Dense matrices to incorporate the SAM attribute is not straightforward, implementing the vanilla Dense mixers ( i.e.without SAM) is extremely simple: they are equivalent to MLP-Mixer [39], employing a mixer matrix of R\ud835\udc3f\u00d7\ud835\udc3f. Toeplitz. Toeplitz matrix mixers without SAM properties are equivalent to convolution operations [18, 17], formulated withR2\ud835\udc3f\u22121parameters. We also present a Toeplitz matrix mixer with SAM properties, thus data-dependent and can handle sequences of arbitrary lengths. The core idea is fairly similar to the quasiseparable matrix mixer \u2013 Hydra \u2013 that the Toeplitz matrix mixer integrates outputs from two separate forward and reverse sequence convolutions. Specifically, for \ud835\udc56\u22080,...\ud835\udc3f\u22121, each token \ud835\udc65\ud835\udc56generates two convolution parameters \ud835\udc5a\u2212\ud835\udc56and\ud835\udc5a\ud835\udc56. Using all\ud835\udc5aparameters, a data-dependent dynamic convolutional weight is generated as {\ud835\udc5a\u2212\ud835\udc3f+1,\ud835\udc5a\u2212\ud835\udc3f+2,...,\ud835\udc5a\u22121,\ud835\udc5a0,\ud835\udc5a1,...\ud835\udc5a\ud835\udc3f\u22122,\ud835\udc5a\ud835\udc3f\u22121}. Vandermonde. In Section 2.4, we introduced the single-headed sequence aligned Vandermonde matrix mixer. Leveraging the definition, we present multi-headed implementation by a seamless extension. Cauchy. The constant for preventing the denominator approaching zero is initialized to 0.0and0.5for the diand the dd variants, respectively. Definition E.1 (Cauchy Matrix) .Given q,k\u2208R\ud835\udc3f, a matrix Mis Cauchy if each(\ud835\udc56,\ud835\udc57)-entry\ud835\udc5a\ud835\udc56\ud835\udc57satisfies\ud835\udc5a\ud835\udc56\ud835\udc57=1 \ud835\udc5e\ud835\udc56\u2212\ud835\udc58\ud835\udc57;\ud835\udc5e\ud835\udc56\u2212 \ud835\udc58\ud835\udc57\u22600. 19class Dense (nn. Module ): def __init__ ( self , d_model , max_seq_len , # max_seq_len is necessary for Dense . expand =2, headdim =128 , device =None , dtype =None , ): factory_kwargs = {\" device \": device , \" dtype \": dtype } super (). __init__ () self . d_model = d_model self . max_seq_len = max_seq_len self . expand = expand self . d_inner = self . expand * self .",
            "references": [
                {
                    "title": "Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality",
                    "abstract": "While Transformers have been the main architecture behind deep learning's success in language modeling, state-space models (SSMs) such as Mamba have recently been shown to match or outperform Transformers at small to medium scale. We show that these families of models are actually quite closely related, and develop a rich framework of theoretical connections between SSMs and variants of attention, connected through various decompositions of a well-studied class of structured semiseparable matrices. Our state space duality (SSD) framework allows us to design a new architecture (Mamba-2) whose core layer is an a refinement of Mamba's selective SSM that is 2-8X faster, while continuing to be competitive with Transformers on language modeling."
                },
                {
                    "title": "Caduceus: Bi-Directional Equivariant Long-Range DNA Sequence Modeling",
                    "abstract": "Large-scale sequence modeling has sparked rapid advances that now extend into biology and genomics. However, modeling genomic sequences introduces challenges such as the need to model long-range token interactions, the effects of upstream and downstream regions of the genome, and the reverse complementarity (RC) of DNA. Here, we propose an architecture motivated by these challenges that builds off the long-range Mamba block, and extends it to a BiMamba component that supports bi-directionality, and to a MambaDNA block that additionally supports RC equivariance. We use MambaDNA as the basis of Caduceus, the first family of RC equivariant bi-directional long-range DNA language models, and we introduce pre-training and fine-tuning strategies that yield Caduceus DNA foundation models. Caduceus outperforms previous long-range models on downstream benchmarks; on a challenging long-range variant effect prediction task, Caduceus exceeds the performance of 10x larger models that do not leverage bi-directionality or equivariance."
                },
                {
                    "title": "Simple linear attention language models balance the recall-throughput tradeoff",
                    "abstract": "Recent work has shown that attention-based language models excel at recall, the ability to ground generations in tokens previously seen in context. However, the efficiency of attention-based models is bottle-necked during inference by the KV-cache's aggressive memory consumption. In this work, we explore whether we can improve language model efficiency (e.g. by reducing memory consumption) without compromising on recall. By applying experiments and theory to a broad set of architectures, we identify a key tradeoff between a model's state size and recall ability. We show that efficient alternatives to attention (e.g. H3, Mamba, RWKV) maintain a fixed-size recurrent state, but struggle at recall. We propose BASED a simple architecture combining linear and sliding window attention. By varying BASED window size and linear attention feature dimension, we can dial the state size and traverse the pareto frontier of the recall-memory tradeoff curve, recovering the full quality of attention on one end and the small state size of attention-alternatives on the other. We train language models up to 1.3b parameters and show that BASED matches the strongest sub-quadratic models (e.g. Mamba) in perplexity and outperforms them on real-world recall-intensive tasks by 6.22 accuracy points. Implementations of linear attention are often less efficient than optimized standard attention implementations. To make BASED competitive, we develop IO-aware algorithms that enable 24x higher throughput on language generation than FlashAttention-2, when generating 1024 tokens using 1.3b parameter models. Code for this work is provided at: https://github.com/HazyResearch/based."
                },
                {
                    "title": "Repeat After Me: Transformers are Better than State Space Models at Copying",
                    "abstract": "Transformers are the dominant architecture for sequence modeling, but there is growing interest in models that use a fixed-size latent state that does not depend on the sequence length, which we refer to as\"generalized state space models\"(GSSMs). In this paper we show that while GSSMs are promising in terms of inference-time efficiency, they are limited compared to transformer models on tasks that require copying from the input context. We start with a theoretical analysis of the simple task of string copying and prove that a two layer transformer can copy strings of exponential length while GSSMs are fundamentally limited by their fixed-size latent state. Empirically, we find that transformers outperform GSSMs in terms of efficiency and generalization on synthetic tasks that require copying the context. Finally, we evaluate pretrained large language models and find that transformer models dramatically outperform state space models at copying and retrieving information from context. Taken together, these results suggest a fundamental gap between transformers and GSSMs on tasks of practical interest."
                },
                {
                    "title": "Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model",
                    "abstract": "Recently the state space models (SSMs) with efficient hardware-aware designs, i.e., the Mamba deep learning model, have shown great potential for long sequence modeling. Meanwhile building efficient and generic vision backbones purely upon SSMs is an appealing direction. However, representing visual data is challenging for SSMs due to the position-sensitivity of visual data and the requirement of global context for visual understanding. In this paper, we show that the reliance on self-attention for visual representation learning is not necessary and propose a new generic vision backbone with bidirectional Mamba blocks (Vim), which marks the image sequences with position embeddings and compresses the visual representation with bidirectional state space models. On ImageNet classification, COCO object detection, and ADE20k semantic segmentation tasks, Vim achieves higher performance compared to well-established vision transformers like DeiT, while also demonstrating significantly improved computation&memory efficiency. For example, Vim is 2.8$\\times$ faster than DeiT and saves 86.8% GPU memory when performing batch inference to extract features on images with a resolution of 1248$\\times$1248. The results demonstrate that Vim is capable of overcoming the computation&memory constraints on performing Transformer-style understanding for high-resolution images and it has great potential to be the next-generation backbone for vision foundation models. Code is available at https://github.com/hustvl/Vim."
                },
                {
                    "title": "MosaicBERT: A Bidirectional Encoder Optimized for Fast Pretraining",
                    "abstract": "Although BERT-style encoder models are heavily used in NLP research, many researchers do not pretrain their own BERTs from scratch due to the high cost of training. In the past half-decade since BERT first rose to prominence, many advances have been made with other transformer architectures and training configurations that have yet to be systematically incorporated into BERT. Here, we introduce MosaicBERT, a BERT-style encoder architecture and training recipe that is empirically optimized for fast pretraining. This efficient architecture incorporates FlashAttention, Attention with Linear Biases (ALiBi), Gated Linear Units (GLU), a module to dynamically remove padded tokens, and low precision LayerNorm into the classic transformer encoder block. The training recipe includes a 30% masking ratio for the Masked Language Modeling (MLM) objective, bfloat16 precision, and vocabulary size optimized for GPU throughput, in addition to best-practices from RoBERTa and other encoder models. When pretrained from scratch on the C4 dataset, this base model achieves a downstream average GLUE (dev) score of 79.6 in 1.13 hours on 8 A100 80 GB GPUs at a cost of roughly $20. We plot extensive accuracy vs. pretraining speed Pareto curves and show that MosaicBERT base and large are consistently Pareto optimal when compared to a competitive BERT base and large. This empirical speed up in pretraining enables researchers and engineers to pretrain custom BERT-style models at low cost instead of finetune on existing generic models. We open source our model weights and code."
                },
                {
                    "title": "Zoology: Measuring and Improving Recall in Efficient Language Models",
                    "abstract": "Attention-free language models that combine gating and convolutions are growing in popularity due to their efficiency and increasingly competitive performance. To better understand these architectures, we pretrain a suite of 17 attention and\"gated-convolution\"language models, finding that SoTA gated-convolution architectures still underperform attention by up to 2.1 perplexity points on the Pile. In fine-grained analysis, we find 82% of the gap is explained by each model's ability to recall information that is previously mentioned in-context, e.g.\"Hakuna Matata means no worries Hakuna Matata it means no\"$\\rightarrow$\"??\". On this task, termed\"associative recall\", we find that attention outperforms gated-convolutions by a large margin: a 70M parameter attention model outperforms a 1.4 billion parameter gated-convolution model on associative recall. This is surprising because prior work shows gated convolutions can perfectly solve synthetic tests for AR capability. To close the gap between synthetics and real language, we develop a new formalization of the task called multi-query associative recall (MQAR) that better reflects actual language. We perform an empirical and theoretical study of MQAR that elucidates differences in the parameter-efficiency of attention and gated-convolution recall. Informed by our analysis, we evaluate simple convolution-attention hybrids and show that hybrids with input-dependent sparse attention patterns can close 97.4% of the gap to attention, while maintaining sub-quadratic scaling. Our code is accessible at: https://github.com/HazyResearch/zoology."
                },
                {
                    "title": "Mamba: Linear-Time Sequence Modeling with Selective State Spaces",
                    "abstract": "Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5$\\times$ higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation."
                },
                {
                    "title": "Monarch Mixer: A Simple Sub-Quadratic GEMM-Based Architecture",
                    "abstract": "Machine learning models are increasingly being scaled in both sequence length and model dimension to reach longer contexts and better performance. However, existing architectures such as Transformers scale quadratically along both these axes. We ask: are there performant architectures that can scale sub-quadratically along sequence length and model dimension? We introduce Monarch Mixer (M2), a new architecture that uses the same sub-quadratic primitive along both sequence length and model dimension: Monarch matrices, a simple class of expressive structured matrices that captures many linear transforms, achieves high hardware efficiency on GPUs, and scales sub-quadratically. As a proof of concept, we explore the performance of M2 in three domains: non-causal BERT-style language modeling, ViT-style image classification, and causal GPT-style language modeling. For non-causal BERT-style modeling, M2 matches BERT-base and BERT-large in downstream GLUE quality with up to 27% fewer parameters, and achieves up to 9.1$\\times$ higher throughput at sequence length 4K. On ImageNet, M2 outperforms ViT-b by 1% in accuracy, with only half the parameters. Causal GPT-style models introduce a technical challenge: enforcing causality via masking introduces a quadratic bottleneck. To alleviate this bottleneck, we develop a novel theoretical view of Monarch matrices based on multivariate polynomial evaluation and interpolation, which lets us parameterize M2 to be causal while remaining sub-quadratic. Using this parameterization, M2 matches GPT-style Transformers at 360M parameters in pretraining perplexity on The PILE--showing for the first time that it may be possible to match Transformer quality without attention or MLPs."
                },
                {
                    "title": "Retentive Network: A Successor to Transformer for Large Language Models",
                    "abstract": "In this work, we propose Retentive Network (RetNet) as a foundation architecture for large language models, simultaneously achieving training parallelism, low-cost inference, and good performance. We theoretically derive the connection between recurrence and attention. Then we propose the retention mechanism for sequence modeling, which supports three computation paradigms, i.e., parallel, recurrent, and chunkwise recurrent. Specifically, the parallel representation allows for training parallelism. The recurrent representation enables low-cost $O(1)$ inference, which improves decoding throughput, latency, and GPU memory without sacrificing performance. The chunkwise recurrent representation facilitates efficient long-sequence modeling with linear complexity, where each chunk is encoded parallelly while recurrently summarizing the chunks. Experimental results on language modeling show that RetNet achieves favorable scaling results, parallel training, low-cost deployment, and efficient inference. The intriguing properties make RetNet a strong successor to Transformer for large language models. Code will be available at https://aka.ms/retnet."
                },
                {
                    "title": "Multi-Head State Space Model for Speech Recognition",
                    "abstract": "State space models (SSMs) have recently shown promising results on small-scale sequence and language modelling tasks, rivalling and outperforming many attention-based approaches. In this paper, we propose a multi-head state space (MH-SSM) architecture equipped with special gating mechanisms, where parallel heads are taught to learn local and global temporal dynamics on sequence data. As a drop-in replacement for multi-head attention in transformer encoders, this new model significantly outperforms the transformer transducer on the LibriSpeech speech recognition corpus. Furthermore, we augment the transformer block with MH-SSMs layers, referred to as the Stateformer, achieving state-of-the-art performance on the LibriSpeech task, with word error rates of 1.76\\%/4.37\\% on the development and 1.91\\%/4.36\\% on the test sets without using an external language model."
                },
                {
                    "title": "Toeplitz Neural Network for Sequence Modeling",
                    "abstract": "Sequence modeling has important applications in natural language processing and computer vision. Recently, the transformer-based models have shown strong performance on various sequence modeling tasks, which rely on attention to capture pairwise token relations, and position embedding to inject positional information. While showing good performance, the transformer models are inefficient to scale to long input sequences, mainly due to the quadratic space-time complexity of attention. To overcome this inefficiency, we propose to model sequences with a relative position encoded Toeplitz matrix and use a Toeplitz matrix-vector production trick to reduce the space-time complexity of the sequence modeling to log linear. A lightweight sub-network called relative position encoder is proposed to generate relative position coefficients with a fixed budget of parameters, enabling the proposed Toeplitz neural network to deal with varying sequence lengths. In addition, despite being trained on 512-token sequences, our model can extrapolate input sequence length up to 14K tokens in inference with consistent performance. Extensive experiments on autoregressive and bidirectional language modeling, image modeling, and the challenging Long-Range Arena benchmark show that our method achieves better performance than its competitors in most downstream tasks while being significantly faster. The code is available at https://github.com/OpenNLPLab/Tnn."
                },
                {
                    "title": "Resurrecting Recurrent Neural Networks for Long Sequences",
                    "abstract": "Recurrent Neural Networks (RNNs) offer fast inference on long sequences but are hard to optimize and slow to train. Deep state-space models (SSMs) have recently been shown to perform remarkably well on long sequence modeling tasks, and have the added benefits of fast parallelizable training and RNN-like fast inference. However, while SSMs are superficially similar to RNNs, there are important differences that make it unclear where their performance boost over RNNs comes from. In this paper, we show that careful design of deep RNNs using standard signal propagation arguments can recover the impressive performance of deep SSMs on long-range reasoning tasks, while also matching their training speed. To achieve this, we analyze and ablate a series of changes to standard RNNs including linearizing and diagonalizing the recurrence, using better parameterizations and initializations, and ensuring proper normalization of the forward pass. Our results provide new insights on the origins of the impressive performance of deep SSMs, while also introducing an RNN block called the Linear Recurrent Unit that matches both their performance on the Long Range Arena benchmark and their computational efficiency."
                },
                {
                    "title": "Hyena Hierarchy: Towards Larger Convolutional Language Models",
                    "abstract": "Recent advances in deep learning have relied heavily on the use of large Transformers due to their ability to learn at scale. However, the core building block of Transformers, the attention operator, exhibits quadratic cost in sequence length, limiting the amount of context accessible. Existing subquadratic methods based on low-rank and sparse approximations need to be combined with dense attention layers to match Transformers, indicating a gap in capability. In this work, we propose Hyena, a subquadratic drop-in replacement for attention constructed by interleaving implicitly parametrized long convolutions and data-controlled gating. In recall and reasoning tasks on sequences of thousands to hundreds of thousands of tokens, Hyena improves accuracy by more than 50 points over operators relying on state-spaces and other implicit and explicit methods, matching attention-based models. We set a new state-of-the-art for dense-attention-free architectures on language modeling in standard datasets (WikiText103 and The Pile), reaching Transformer quality with a 20% reduction in training compute required at sequence length 2K. Hyena operators are twice as fast as highly optimized attention at sequence length 8K, and 100x faster at sequence length 64K."
                },
                {
                    "title": "Exact computations with quasiseparable matrices",
                    "abstract": "Quasiseparable matrices are a class of rank-structured matrices widely used in numerical linear algebra and of growing interest in computer algebra, with applications in e.g. the linearization of polynomial matrices. Various representation formats exist for these matrices that have rarely been compared. We show how the most central formats SSS and HSS can be adapted to symbolic computation, where the exact rank replaces threshold based numerical ranks. We clarify their links and compare them with the Bruhat format. To this end, we state their space and time cost estimates based on fast matrix multiplication, and compare them, with their leading constants. The comparison is supported by software experiments. We make further progresses for the Bruhat format, for which we give a generation algorithm, following a Crout elimination scheme, which specializes into fast algorithms for the construction from a sparse matrix or from the sum of Bruhat representations."
                },
                {
                    "title": "Hungry Hungry Hippos: Towards Language Modeling with State Space Models",
                    "abstract": "State space models (SSMs) have demonstrated state-of-the-art sequence modeling performance in some modalities, but underperform attention in language modeling. Moreover, despite scaling nearly linearly in sequence length instead of quadratically, SSMs are still slower than Transformers due to poor hardware utilization. In this paper, we make progress on understanding the expressivity gap between SSMs and attention in language modeling, and on reducing the hardware barrier between SSMs and attention. First, we use synthetic language modeling tasks to understand the gap between SSMs and attention. We find that existing SSMs struggle with two capabilities: recalling earlier tokens in the sequence and comparing tokens across the sequence. To understand the impact on language modeling, we propose a new SSM layer, H3, that is explicitly designed for these abilities. H3 matches attention on the synthetic languages and comes within 0.4 PPL of Transformers on OpenWebText. Furthermore, a hybrid 125M-parameter H3-attention model that retains two attention layers surprisingly outperforms Transformers on OpenWebText by 1.0 PPL. Next, to improve the efficiency of training SSMs on modern hardware, we propose FlashConv. FlashConv uses a fused block FFT algorithm to improve efficiency on sequences up to 8K, and introduces a novel state passing algorithm that exploits the recurrent properties of SSMs to scale to longer sequences. FlashConv yields 2$\\times$ speedup on the long-range arena benchmark and allows hybrid language models to generate text 2.4$\\times$ faster than Transformers. Using FlashConv, we scale hybrid H3-attention language models up to 2.7B parameters on the Pile and find promising initial results, achieving lower perplexity than Transformers and outperforming Transformers in zero- and few-shot learning on a majority of tasks in the SuperGLUE benchmark."
                },
                {
                    "title": "Pretraining Without Attention",
                    "abstract": "Transformers have been essential to pretraining success in NLP. While other architectures have been used, downstream accuracy is either significantly worse, or requires attention layers to match standard benchmarks such as GLUE. This work explores pretraining without attention by using recent advances in sequence routing based on state-space models (SSMs). Our proposed model, Bidirectional Gated SSM (BiGS), combines SSM layers with a multiplicative gating architecture that has been effective in simplified sequence modeling architectures. The model learns static layers that do not consider pair-wise interactions. Even so, BiGS is able to match BERT pretraining accuracy on GLUE and can be extended to long-form pretraining of 4096 tokens without approximation. Analysis shows that while the models have similar average accuracy, the approach has different inductive biases than BERT in terms of interactions and syntactic representations. All models from this work are available at https://github.com/jxiw/BiGS."
                },
                {
                    "title": "The Devil in Linear Transformer",
                    "abstract": "Linear transformers aim to reduce the quadratic space-time complexity of vanilla transformers. However, they usually suffer from degraded performances on various tasks and corpus. In this paper, we examine existing kernel-based linear transformers and identify two key issues that lead to such performance gaps: 1) unbounded gradients in the attention computation adversely impact the convergence of linear transformer models; 2) attention dilution which trivially distributes attention scores over long sequences while neglecting neighbouring structures. To address these issues, we first identify that the scaling of attention matrices is the devil in unbounded gradients, which turns out unnecessary in linear attention as we show theoretically and empirically. To this end, we propose a new linear attention that replaces the scaling operation with a normalization to stabilize gradients. For the issue of attention dilution, we leverage a diagonal attention to confine attention to only neighbouring tokens in early layers. Benefiting from the stable gradients and improved attention, our new linear transformer model, transNormer, demonstrates superior performance on text classification and language modeling tasks, as well as on the challenging Long-Range Arena benchmark, surpassing vanilla transformer and existing linear variants by a clear margin while being significantly more space-time efficient. The code is available at https://github.com/OpenNLPLab/Transnormer ."
                },
                {
                    "title": "Long Range Language Modeling via Gated State Spaces",
                    "abstract": "State space models have shown to be effective at modeling long range dependencies, specially on sequence classification tasks. In this work we focus on autoregressive sequence modeling over English books, Github source code and ArXiv mathematics articles. Based on recent developments around the effectiveness of gated activation functions, we propose a new layer named Gated State Space (GSS) and show that it trains significantly faster than the diagonal version of S4 (i.e. DSS) on TPUs, is fairly competitive with several well-tuned Transformer-based baselines and exhibits zero-shot generalization to longer inputs while being straightforward to implement. Finally, we show that leveraging self-attention to model local dependencies improves the performance of GSS even further."
                },
                {
                    "title": "On the Parameterization and Initialization of Diagonal State Space Models",
                    "abstract": "State space models (SSM) have recently been shown to be very effective as a deep learning layer as a promising alternative to sequence models such as RNNs, CNNs, or Transformers. The first version to show this potential was the S4 model, which is particularly effective on tasks involving long-range dependencies by using a prescribed state matrix called the HiPPO matrix. While this has an interpretable mathematical mechanism for modeling long dependencies, it introduces a custom representation and algorithm that can be difficult to implement. On the other hand, a recent variant of S4 called DSS showed that restricting the state matrix to be fully diagonal can still preserve the performance of the original model when using a specific initialization based on approximating S4's matrix. This work seeks to systematically understand how to parameterize and initialize such diagonal state space models. While it follows from classical results that almost all SSMs have an equivalent diagonal form, we show that the initialization is critical for performance. We explain why DSS works mathematically, by showing that the diagonal restriction of S4's matrix surprisingly recovers the same kernel in the limit of infinite state dimension. We also systematically describe various design choices in parameterizing and computing diagonal SSMs, and perform a controlled empirical study ablating the effects of these choices. Our final model S4D is a simple diagonal version of S4 whose kernel computation requires just 2 lines of code and performs comparably to S4 in almost all settings, with state-of-the-art results for image, audio, and medical time-series domains, and averaging 85\\% on the Long Range Arena benchmark."
                },
                {
                    "title": "Monarch: Expressive Structured Matrices for Efficient and Accurate Training",
                    "abstract": "Large neural networks excel in many domains, but they are expensive to train and fine-tune. A popular approach to reduce their compute or memory requirements is to replace dense weight matrices with structured ones (e.g., sparse, low-rank, Fourier transform). These methods have not seen widespread adoption (1) in end-to-end training due to unfavorable efficiency--quality tradeoffs, and (2) in dense-to-sparse fine-tuning due to lack of tractable algorithms to approximate a given dense weight matrix. To address these issues, we propose a class of matrices (Monarch) that is hardware-efficient (they are parameterized as products of two block-diagonal matrices for better hardware utilization) and expressive (they can represent many commonly used transforms). Surprisingly, the problem of approximating a dense weight matrix with a Monarch matrix, though nonconvex, has an analytical optimal solution. These properties of Monarch matrices unlock new ways to train and fine-tune sparse and dense models. We empirically validate that Monarch can achieve favorable accuracy-efficiency tradeoffs in several end-to-end sparse training applications: speeding up ViT and GPT-2 training on ImageNet classification and Wikitext-103 language modeling by 2x with comparable model quality, and reducing the error on PDE solving and MRI reconstruction tasks by 40%. In sparse-to-dense training, with a simple technique called\"reverse sparsification,\"Monarch matrices serve as a useful intermediate representation to speed up GPT-2 pretraining on OpenWebText by 2x without quality drop. The same technique brings 23% faster BERT pretraining than even the very optimized implementation from Nvidia that set the MLPerf 1.1 record. In dense-to-sparse fine-tuning, as a proof-of-concept, our Monarch approximation algorithm speeds up BERT fine-tuning on GLUE by 1.7x with comparable accuracy."
                },
                {
                    "title": "It's Raw! Audio Generation with State-Space Models",
                    "abstract": "Developing architectures suitable for modeling raw audio is a challenging problem due to the high sampling rates of audio waveforms. Standard sequence modeling approaches like RNNs and CNNs have previously been tailored to fit the demands of audio, but the resultant architectures make undesirable computational tradeoffs and struggle to model waveforms effectively. We propose SaShiMi, a new multi-scale architecture for waveform modeling built around the recently introduced S4 model for long sequence modeling. We identify that S4 can be unstable during autoregressive generation, and provide a simple improvement to its parameterization by drawing connections to Hurwitz matrices. SaShiMi yields state-of-the-art performance for unconditional waveform generation in the autoregressive setting. Additionally, SaShiMi improves non-autoregressive generation performance when used as the backbone architecture for a diffusion model. Compared to prior architectures in the autoregressive generation setting, SaShiMi generates piano and speech waveforms which humans find more musical and coherent respectively, e.g. 2x better mean opinion scores than WaveNet on an unconditional speech generation task. On a music generation task, SaShiMi outperforms WaveNet on density estimation and speed at both training and inference even when using 3x fewer parameters. Code can be found at https://github.com/HazyResearch/state-spaces and samples at https://hazyresearch.stanford.edu/sashimi-examples."
                },
                {
                    "title": "A ConvNet for the 2020s",
                    "abstract": "The \u201cRoaring 20s\u201d of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually \u201cmodernize\u201d a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets."
                },
                {
                    "title": "Efficiently Modeling Long Sequences with Structured State Spaces",
                    "abstract": "A central goal of sequence modeling is designing a single principled model that can address sequence data across a range of modalities and tasks, particularly on long-range dependencies. Although conventional models including RNNs, CNNs, and Transformers have specialized variants for capturing long dependencies, they still struggle to scale to very long sequences of $10000$ or more steps. A promising recent approach proposed modeling sequences by simulating the fundamental state space model (SSM) \\( x'(t) = Ax(t) + Bu(t), y(t) = Cx(t) + Du(t) \\), and showed that for appropriate choices of the state matrix \\( A \\), this system could handle long-range dependencies mathematically and empirically. However, this method has prohibitive computation and memory requirements, rendering it infeasible as a general sequence modeling solution. We propose the Structured State Space sequence model (S4) based on a new parameterization for the SSM, and show that it can be computed much more efficiently than prior approaches while preserving their theoretical strengths. Our technique involves conditioning \\( A \\) with a low-rank correction, allowing it to be diagonalized stably and reducing the SSM to the well-studied computation of a Cauchy kernel. S4 achieves strong empirical results across a diverse range of established benchmarks, including (i) 91\\% accuracy on sequential CIFAR-10 with no data augmentation or auxiliary losses, on par with a larger 2-D ResNet, (ii) substantially closing the gap to Transformers on image and language modeling tasks, while performing generation $60\\times$ faster (iii) SoTA on every task from the Long Range Arena benchmark, including solving the challenging Path-X task of length 16k that all prior work fails on, while being as efficient as all competitors."
                },
                {
                    "title": "FNet: Mixing Tokens with Fourier Transforms",
                    "abstract": "We show that Transformer encoder architectures can be sped up, with limited accuracy costs, by replacing the self-attention sublayers with simple linear transformations that \u201cmix\u201d input tokens. Most surprisingly, we find that replacing the self-attention sublayer in a Transformer encoder with a standard, unparameterized Fourier Transform achieves 92-97% of the accuracy of BERT counterparts on the GLUE benchmark, but trains 80% faster on GPUs and 70% faster on TPUs at standard 512 input lengths. At longer input lengths, our FNet model is significantly faster: when compared to the \u201cefficient Transformers\u201d on the Long Range Arena benchmark, FNet matches the accuracy of the most accurate models, while outpacing the fastest models across all sequence lengths on GPUs (and across relatively shorter lengths on TPUs). Finally, FNet has a light memory footprint and is particularly efficient at smaller model sizes; for a fixed speed and accuracy budget, small FNet models outperform Transformer counterparts."
                },
                {
                    "title": "MLP-Mixer: An all-MLP Architecture for Vision",
                    "abstract": "Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e.\"mixing\"the per-location features), and one with MLPs applied across patches (i.e.\"mixing\"spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers."
                },
                {
                    "title": "How to Train BERT with an Academic Budget",
                    "abstract": "While large language models a la BERT are used ubiquitously in NLP, pretraining them is considered a luxury that only a few well-funded industry labs can afford. How can one train such models with a more modest budget? We present a recipe for pretraining a masked language model in 24 hours using a single low-end deep learning server. We demonstrate that through a combination of software optimizations, design choices, and hyperparameter tuning, it is possible to produce models that are competitive with BERT-base on GLUE tasks at a fraction of the original pretraining cost."
                },
                {
                    "title": "CKConv: Continuous Kernel Convolution For Sequential Data",
                    "abstract": "Conventional neural architectures for sequential data present important limitations. Recurrent networks suffer from exploding and vanishing gradients, small effective memory horizons, and must be trained sequentially. Convolutional networks are unable to handle sequences of unknown size and their memory horizon must be defined a priori. In this work, we show that all these problems can be solved by formulating convolutional kernels in CNNs as continuous functions. The resulting Continuous Kernel Convolution (CKConv) allows us to model arbitrarily long sequences in a parallel manner, within a single operation, and without relying on any form of recurrence. We show that Continuous Kernel Convolutional Networks (CKCNNs) obtain state-of-the-art results in multiple datasets, e.g., permuted MNIST, and, thanks to their continuous nature, are able to handle non-uniformly sampled datasets and irregularly-sampled data natively. CKCNNs match or perform better than neural ODEs designed for these purposes in a faster and simpler manner."
                },
                {
                    "title": "Training data-efficient image transformers & distillation through attention",
                    "abstract": "Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption. In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data. More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention",
                    "abstract": "Transformers achieve remarkable performance in several tasks but due to their quadratic complexity, with respect to the input's length, they are prohibitively slow for very long sequences. To address this limitation, we express the self-attention as a linear dot-product of kernel feature maps and make use of the associativity property of matrix products to reduce the complexity from $\\mathcal{O}\\left(N^2\\right)$ to $\\mathcal{O}\\left(N\\right)$, where $N$ is the sequence length. We show that this formulation permits an iterative implementation that dramatically accelerates autoregressive transformers and reveals their relationship to recurrent neural networks. Our linear transformers achieve similar performance to vanilla transformers and they are up to 4000x faster on autoregressive prediction of very long sequences."
                },
                {
                    "title": "Linformer: Self-Attention with Linear Complexity",
                    "abstract": "Large transformer models have shown extraordinary success in achieving state-of-the-art results in many natural language processing applications. However, training and deploying these models can be prohibitively costly for long sequences, as the standard self-attention mechanism of the Transformer uses $O(n^2)$ time and space with respect to sequence length. In this paper, we demonstrate that the self-attention mechanism can be approximated by a low-rank matrix. We further exploit this finding to propose a new self-attention mechanism, which reduces the overall self-attention complexity from $O(n^2)$ to $O(n)$ in both time and space. The resulting linear transformer, the \\textit{Linformer}, performs on par with standard Transformer models, while being much more memory- and time-efficient."
                },
                {
                    "title": "Kaleidoscope: An Efficient, Learnable Representation For All Structured Linear Maps",
                    "abstract": "Modern neural network architectures use structured linear transformations, such as low-rank matrices, sparse matrices, permutations, and the Fourier transform, to improve inference speed and reduce memory usage compared to general linear maps. However, choosing which of the myriad structured transformations to use (and its associated parameterization) is a laborious task that requires trading off speed, space, and accuracy. We consider a different approach: we introduce a family of matrices called kaleidoscope matrices (K-matrices) that provably capture any structured matrix with near-optimal space (parameter) and time (arithmetic operation) complexity. We empirically validate that K-matrices can be automatically learned within end-to-end pipelines to replace hand-crafted procedures, in order to improve model quality. For example, replacing channel shuffles in ShuffleNet improves classification accuracy on ImageNet by up to 5%. Learnable K-matrices can also simplify hand-engineered pipelines---we replace filter bank feature computation in speech data preprocessing with a kaleidoscope layer, resulting in only 0.4% loss in accuracy on the TIMIT speech recognition task. K-matrices can also capture latent structure in models: for a challenging permuted image classification task, adding a K-matrix to a standard convolutional architecture can enable learning the latent permutation and improve accuracy by over 8 points. We provide a practically efficient implementation of our approach, and use K-matrices in a Transformer network to attain 36% faster end-to-end inference speed on a language translation task."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "Learning Fast Algorithms for Linear Transforms Using Butterfly Factorizations",
                    "abstract": "Fast linear transforms are ubiquitous in machine learning, including the discrete Fourier transform, discrete cosine transform, and other structured transformations such as convolutions. All of these transforms can be represented by dense matrix-vector multiplication, yet each has a specialized and highly efficient (subquadratic) algorithm. We ask to what extent hand-crafting these algorithms and implementations is necessary, what structural priors they encode, and how much knowledge is required to automatically learn a fast algorithm for a provided structured transform. Motivated by a characterization of fast matrix-vector multiplication as products of sparse matrices, we introduce a parameterization of divide-and-conquer methods that is capable of representing a large class of transforms. This generic formulation can automatically learn an efficient algorithm for many important transforms; for example, it recovers the O(N log N) Cooley-Tukey FFT algorithm to machine precision, for dimensions N up to 1024. Furthermore, our method can be incorporated as a lightweight replacement of generic matrices in machine learning pipelines to learn efficient and compressible transformations. On a standard task of compressing a single hidden-layer network, our method exceeds the classification accuracy of unconstrained matrices on CIFAR-10 by 3.9 points-the first time a structured approach has done so-with 4\u00d7 faster inference speed and 40\u00d7 fewer parameters."
                },
                {
                    "title": "GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding",
                    "abstract": "Human ability to understand language is general, flexible, and robust. In contrast, most NLU models above the word level are designed for a specific task and struggle with out-of-domain data. If we aspire to develop models with understanding beyond the detection of superficial correspondences between inputs and outputs, then it is critical to develop a unified model that can execute a range of linguistic tasks across different domains. To facilitate research in this direction, we present the General Language Understanding Evaluation (GLUE, gluebenchmark.com): a benchmark of nine diverse NLU tasks, an auxiliary dataset for probing models for understanding of specific linguistic phenomena, and an online platform for evaluating and comparing models. For some benchmark tasks, training data is plentiful, but for others it is limited or does not match the genre of the test set. GLUE thus favors models that can represent linguistic knowledge in a way that facilitates sample-efficient learning and effective knowledge-transfer across tasks. While none of the datasets in GLUE were created from scratch for the benchmark, four of them feature privately-held test data, which is used to ensure that the benchmark is used fairly. We evaluate baselines that use ELMo (Peters et al., 2018), a powerful transfer learning technique, as well as state-of-the-art sentence representation models. The best models still achieve fairly low absolute scores. Analysis with our diagnostic dataset yields similarly weak performance over all phenomena tested, with some exceptions."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Neural Machine Translation by Jointly Learning to Align and Translate",
                    "abstract": "Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively incorporate sequence alignment properties into sequence models to enhance their performance in downstream tasks?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses the limitations of existing sequence models that often overlook data-dependence and extendability. By introducing Sequence Aligned Matrices (SAM) and novel structured matrix configurations, this research could lead to significant advancements in sequence modeling techniques, improving their expressivity and efficiency. The implications extend to various practical applications, such as natural language processing, bioinformatics, and time-series analysis, where enhanced sequence models can lead to better predictions and insights.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the inherent complexities of integrating sequence alignment properties into existing models. Naive approaches may fail because they do not account for the data-dependent nature of sequences, leading to suboptimal performance. Additionally, technical obstacles include the need for efficient structured matrix multiplication algorithms and the difficulty in generalizing existing models to incorporate bidirectional processing without sacrificing computational efficiency. The theoretical understanding of quasiseparable matrices and their application in non-causal settings also presents a significant hurdle.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has largely focused on either data-independent models or simplistic adaptations of existing frameworks, which do not fully leverage the potential of sequence alignment. Barriers include a lack of comprehensive exploration of structured matrix parameterizations and the absence of effective methodologies for integrating SAM properties into sequence models. This research differs by systematically categorizing existing models and introducing new configurations, such as the Hydra model, which provides a robust framework for bidirectional processing and efficient computation.\n\n### [Question 5] - What are the key components of my approach and results?\nThe proposed methodology involves the development of Sequence Aligned Matrices (SAM) and the introduction of the Hydra model, which utilizes quasiseparable matrix mixers. The approach will be evaluated using various datasets relevant to sequence modeling tasks, with performance metrics focusing on expressivity and computational efficiency. Expected outcomes include demonstrating the superior performance of SAM-equipped models in comparison to traditional non-aligned models, as well as providing practical implementations in PyTorch to facilitate further research and application in the field."
            }
        },
        "author_data": {
            "0d444b4d-43d3-4fd0-af1d-35ef62ce1b59": {
                "pk": "0d444b4d-43d3-4fd0-af1d-35ef62ce1b59",
                "name": "Sukjun Hwang",
                "collaborators": [
                    "Seoung Wug Oh",
                    "Seon Joo Kim",
                    "Miran Heo",
                    "Joon-Young Lee",
                    "Hanjung Kim",
                    "Jaehyun Kang",
                    "Jeongseok Hyun",
                    "Su Ho Han",
                    "Yeonchool Park",
                    "Hyunwoo Kim"
                ],
                "domain": [
                    "Video Instance Segmentation",
                    "Object Detection",
                    "Deep Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Single-shot Path Integrated Panoptic Segmentation",
                        "abstract": "Panoptic segmentation, which is a novel task of unifying instance segmentation and semantic segmentation, has attracted a lot of attention lately. However, most of the previous methods are composed of multiple pathways with each pathway specialized to a designated segmentation task. In this paper, we propose to resolve panoptic segmentation in single-shot by integrating the execution flows. With the integrated pathway, a unified feature map called Panoptic-Feature is generated, which includes the information of both things and stuffs. Panoptic-Feature becomes more sophisticated by auxiliary problems that guide to cluster pixels that belong to the same instance and differentiate between objects of different classes. A collection of convolutional filters, where each filter represents either a thing or stuff, is applied to Panoptic-Feature at once, materializing the single-shot panoptic segmentation. Taking the advantages of both top-down and bottom-up approaches, our method, named SPINet, enjoys high efficiency and accuracy on major panoptic segmentation benchmarks: COCO and Cityscapes."
                    },
                    {
                        "title": "Video Instance Segmentation using Inter-Frame Communication Transformers",
                        "abstract": "We propose a novel end-to-end solution for video instance segmentation (VIS) based on transformers. Recently, the per-clip pipeline shows superior performance over per-frame methods leveraging richer information from multiple frames. However, previous per-clip models require heavy computation and memory usage to achieve frame-to-frame communications, limiting practicality. In this work, we propose Inter-frame Communication Transformers (IFC), which significantly reduces the overhead for information-passing between frames by efficiently encoding the context within the input clip. Specifically, we propose to utilize concise memory tokens as a mean of conveying information as well as summarizing each frame scene. The features of each frame are enriched and correlated with other frames through exchange of information between the precisely encoded memory tokens. We validate our method on the latest benchmark sets and achieved the state-of-the-art performance (AP 44.6 on YouTube-VIS 2019 val set using the offline inference) while having a considerably fast runtime (89.4 FPS). Our method can also be applied to near-online inference for processing a video in real-time with only a small delay. The code will be made available."
                    },
                    {
                        "title": "Cannot See the Forest for the Trees: Aggregating Multiple Viewpoints to Better Classify Objects in Videos",
                        "abstract": "Recently, both long-tailed recognition and object tracking have made great advances individually. TAO benchmark presented a mixture of the two, long-tailed object tracking, in order to further reflect the aspect of the real-world. To date, existing solutions have adopted detectors showing robustness in long-tailed distributions, which derive per-frame results. Then, they used tracking algorithms that combine the temporally independent detections to finalize tracklets. However, as the approaches did not take temporal changes in scenes into account, inconsistent classification results in videos led to low overall performance. In this paper, we present a set classifier that improves accuracy of classifying tracklets by aggregating information from multiple viewpoints contained in a tracklet. To cope with sparse annotations in videos, we further propose augmentation of tracklets that can maximize data efficiency. The set classifier is plug-and-playable to existing object trackers, and highly improves the performance of long-tailed object tracking. By simply attaching our method to QDTrack on top of ResNet-101, we achieve the new state-of-the-art, 19.9% and 15.7% TrackAP_50 on TAO validation and test sets, respectively."
                    },
                    {
                        "title": "VITA: Video Instance Segmentation via Object Token Association",
                        "abstract": "We introduce a novel paradigm for offline Video Instance Segmentation (VIS), based on the hypothesis that explicit object-oriented information can be a strong clue for understanding the context of the entire sequence. To this end, we propose VITA, a simple structure built on top of an off-the-shelf Transformer-based image instance segmentation model. Specifically, we use an image object detector as a means of distilling object-specific contexts into object tokens. VITA accomplishes video-level understanding by associating frame-level object tokens without using spatio-temporal backbone features. By effectively building relationships between objects using the condensed information, VITA achieves the state-of-the-art on VIS benchmarks with a ResNet-50 backbone: 49.8 AP, 45.7 AP on YouTube-VIS 2019 & 2021, and 19.6 AP on OVIS. Moreover, thanks to its object token-based structure that is disjoint from the backbone features, VITA shows several practical advantages that previous offline VIS methods have not explored - handling long and high-resolution videos with a common GPU, and freezing a frame-level detector trained on image domain. Code is available at https://github.com/sukjunhwang/VITA."
                    },
                    {
                        "title": "VISAGE: Video Instance Segmentation with Appearance-Guided Enhancement",
                        "abstract": "In recent years, online Video Instance Segmentation (VIS) methods have shown remarkable advancement with their powerful query-based detectors. Utilizing the output queries of the detector at the frame-level, these methods achieve high accuracy on challenging benchmarks. However, our observations demonstrate that these methods heavily rely on location information, which often causes incorrect associations between objects. This paper presents that a key axis of object matching in trackers is appearance information, which becomes greatly instructive under conditions where positional cues are insufficient for distinguishing their identities. Therefore, we suggest a simple yet powerful extension to object decoders that explicitly extract embeddings from backbone features and drive queries to capture the appearances of objects, which greatly enhances instance association accuracy. Furthermore, recognizing the limitations of existing benchmarks in fully evaluating appearance awareness, we have constructed a synthetic dataset to rigorously validate our method. By effectively resolving the over-reliance on location information, we achieve state-of-the-art results on YouTube-VIS 2019/2021 and Occluded VIS (OVIS). Code is available at https://github.com/KimHanjung/VISAGE."
                    },
                    {
                        "title": "A Generalized Framework for Video Instance Segmentation",
                        "abstract": "The handling of long videos with complex and occluded sequences has recently emerged as a new challenge in the video instance segmentation (VIS) community. However, existing methods have limitations in addressing this challenge. We argue that the biggest bottleneck in current approaches is the discrepancy between training and inference. To effectively bridge this gap, we propose a Generalized framework for VIS, namely GenVIS, that achieves state-of-the-art performance on challenging benchmarks without designing complicated architectures or requiring extra post-processing. The key contribution of GenVIS is the learning strategy, which includes a query-based training pipeline for sequential learning with a novel target label assignment. Additionally, we introduce a memory that effectively acquires information from previous states. Thanks to the new perspective, which focuses on building relationships between separate frames or clips, GenVIS can be flexibly executed in both online and semi-online manner. We evaluate our approach on popular VIS benchmarks, achieving state-of-the-art results on YouTube-VIS 2019/2021/2022 and Occluded VIS (OVIS). Notably, we greatly outperform the state-of-the-art on the long VIS benchmark (OVIS), improving 5.6 AP with ResNet-50 backbone. Code is available at https://github.com/miranheo/GenVIS."
                    },
                    {
                        "title": "VISOLO: Grid-Based Space-Time Aggregation for Efficient Online Video Instance Segmentation",
                        "abstract": "For online video instance segmentation (VIS), fully utilizing the information from previous frames in an efficient manner is essential for real-time applications. Most previous methods follow a two-stage approach requiring additional computations such as RPN and RoIAlign, and do not fully exploit the available information in the video for all subtasks in VIS. In this paper, we propose a novel single-stage framework for online VIS built based on the grid structured feature representation. The grid-based features allow us to employ fully convolutional networks for real-time processing, and also to easily reuse and share features within different components. We also introduce cooperatively operating modules that aggregate information from available frames, in order to enrich the features for all subtasks in VIS. Our design fully takes advantage of previous information in a grid form for all tasks in VIS in an efficient way, and we achieved the new state-of-the-art accuracy (38.6 AP and 36.9 AP) and speed (40.0 FPS) on YouTube-VIS 2019 and 2021 datasets among online VIS methods. The code is available at https://github.com/SuHoHan95/VISOLO."
                    }
                ]
            },
            "7e815c8a-1cf3-4603-8e74-5b0999d8fd5c": {
                "pk": "7e815c8a-1cf3-4603-8e74-5b0999d8fd5c",
                "name": "Aakash Lahoti",
                "collaborators": [
                    "Spandan Senapati",
                    "Ketan Rajawat",
                    "Alec Koppel",
                    "Stefani Karp",
                    "Ezra Winston",
                    "Aarti Singh",
                    "Yuanzhi Li"
                ],
                "domain": [
                    "Optimization",
                    "Machine Learning",
                    "Convolutional Neural Networks",
                    "Statistical Learning"
                ],
                "publications": [
                    {
                        "title": "Sharpened Lazy Incremental Quasi-Newton Method",
                        "abstract": "The problem of minimizing the sum of $n$ functions in $d$ dimensions is ubiquitous in machine learning and statistics. In many applications where the number of observations $n$ is large, it is necessary to use incremental or stochastic methods, as their per-iteration cost is independent of $n$. Of these, Quasi-Newton (QN) methods strike a balance between the per-iteration cost and the convergence rate. Specifically, they exhibit a superlinear rate with $O(d^2)$ cost in contrast to the linear rate of first-order methods with $O(d)$ cost and the quadratic rate of second-order methods with $O(d^3)$ cost. However, existing incremental methods have notable shortcomings: Incremental Quasi-Newton (IQN) only exhibits asymptotic superlinear convergence. In contrast, Incremental Greedy BFGS (IGS) offers explicit superlinear convergence but suffers from poor empirical performance and has a per-iteration cost of $O(d^3)$. To address these issues, we introduce the Sharpened Lazy Incremental Quasi-Newton Method (SLIQN) that achieves the best of both worlds: an explicit superlinear convergence rate, and superior empirical performance at a per-iteration $O(d^2)$ cost. SLIQN features two key changes: first, it incorporates a hybrid strategy of using both classic and greedy BFGS updates, allowing it to empirically outperform both IQN and IGS. Second, it employs a clever constant multiplicative factor along with a lazy propagation strategy, which enables it to have a cost of $O(d^2)$. Additionally, our experiments demonstrate the superiority of SLIQN over other incremental and stochastic Quasi-Newton variants and establish its competitiveness with second-order incremental methods."
                    },
                    {
                        "title": "Role of Locality and Weight Sharing in Image-Based Tasks: A Sample Complexity Separation between CNNs, LCNs, and FCNs",
                        "abstract": "Vision tasks are characterized by the properties of locality and translation invariance. The superior performance of convolutional neural networks (CNNs) on these tasks is widely attributed to the inductive bias of locality and weight sharing baked into their architecture. Existing attempts to quantify the statistical benefits of these biases in CNNs over locally connected convolutional neural networks (LCNs) and fully connected neural networks (FCNs) fall into one of the following categories: either they disregard the optimizer and only provide uniform convergence upper bounds with no separating lower bounds, or they consider simplistic tasks that do not truly mirror the locality and translation invariance as found in real-world vision tasks. To address these deficiencies, we introduce the Dynamic Signal Distribution (DSD) classification task that models an image as consisting of $k$ patches, each of dimension $d$, and the label is determined by a $d$-sparse signal vector that can freely appear in any one of the $k$ patches. On this task, for any orthogonally equivariant algorithm like gradient descent, we prove that CNNs require $\\tilde{O}(k+d)$ samples, whereas LCNs require $\\Omega(kd)$ samples, establishing the statistical advantages of weight sharing in translation invariant tasks. Furthermore, LCNs need $\\tilde{O}(k(k+d))$ samples, compared to $\\Omega(k^2d)$ samples for FCNs, showcasing the benefits of locality in local tasks. Additionally, we develop information theoretic tools for analyzing randomized algorithms, which may be of interest for statistical research."
                    }
                ]
            },
            "cda04de3-5356-4064-93bc-45aa080993a4": {
                "pk": "cda04de3-5356-4064-93bc-45aa080993a4",
                "name": "Tri Dao",
                "collaborators": [
                    "Albert Gu",
                    "Christopher R\u00e9",
                    "Jian Zhang",
                    "Avner May",
                    "Atri Rudra",
                    "Christopher De Sa",
                    "Matthew Eichhorn",
                    "Govinda M Kamath",
                    "Vasilis Syrgkanis",
                    "Lester Mackey"
                ],
                "domain": [
                    "Transformers",
                    "Attention Mechanism",
                    "Kernel Methods",
                    "Model Compression"
                ],
                "publications": [
                    {
                        "title": "FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning",
                        "abstract": "Scaling Transformers to longer sequence lengths has been a major problem in the last several years, promising to improve performance in language modeling and high-resolution image understanding, as well as to unlock new applications in code, audio, and video generation. The attention layer is the main bottleneck in scaling to longer sequences, as its runtime and memory increase quadratically in the sequence length. FlashAttention exploits the asymmetric GPU memory hierarchy to bring significant memory saving (linear instead of quadratic) and runtime speedup (2-4$\\times$ compared to optimized baselines), with no approximation. However, FlashAttention is still not nearly as fast as optimized matrix-multiply (GEMM) operations, reaching only 25-40\\% of the theoretical maximum FLOPs/s. We observe that the inefficiency is due to suboptimal work partitioning between different thread blocks and warps on the GPU, causing either low-occupancy or unnecessary shared memory reads/writes. We propose FlashAttention-2, with better work partitioning to address these issues. In particular, we (1) tweak the algorithm to reduce the number of non-matmul FLOPs (2) parallelize the attention computation, even for a single head, across different thread blocks to increase occupancy, and (3) within each thread block, distribute the work between warps to reduce communication through shared memory. These yield around 2$\\times$ speedup compared to FlashAttention, reaching 50-73\\% of the theoretical maximum FLOPs/s on A100 and getting close to the efficiency of GEMM operations. We empirically validate that when used end-to-end to train GPT-style models, FlashAttention-2 reaches training speed of up to 225 TFLOPs/s per A100 GPU (72\\% model FLOPs utilization)."
                    },
                    {
                        "title": "Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality",
                        "abstract": "While Transformers have been the main architecture behind deep learning's success in language modeling, state-space models (SSMs) such as Mamba have recently been shown to match or outperform Transformers at small to medium scale. We show that these families of models are actually quite closely related, and develop a rich framework of theoretical connections between SSMs and variants of attention, connected through various decompositions of a well-studied class of structured semiseparable matrices. Our state space duality (SSD) framework allows us to design a new architecture (Mamba-2) whose core layer is an a refinement of Mamba's selective SSM that is 2-8X faster, while continuing to be competitive with Transformers on language modeling."
                    },
                    {
                        "title": "Mamba: Linear-Time Sequence Modeling with Selective State Spaces",
                        "abstract": "Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5$\\times$ higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation."
                    },
                    {
                        "title": "Gaussian Quadrature for Kernel Features",
                        "abstract": "Kernel methods have recently attracted resurgent interest, showing performance competitive with deep neural networks in tasks such as speech recognition. The random Fourier features map is a technique commonly used to scale up kernel machines, but employing the randomized feature map means that $O(\\epsilon^{-2})$ samples are required to achieve an approximation error of at most $\\epsilon$. We investigate some alternative schemes for constructing feature maps that are deterministic, rather than random, by approximating the kernel in the frequency domain using Gaussian quadrature. We show that deterministic feature maps can be constructed, for any $\\gamma > 0$, to achieve error $\\epsilon$ with $O(e^{e^\\gamma} + \\epsilon^{-1/\\gamma})$ samples as $\\epsilon$ goes to 0. Our method works particularly well with sparse ANOVA kernels, which are inspired by the convolutional layer of CNNs. We validate our methods on datasets in different domains, such as MNIST and TIMIT, showing that deterministic features are faster to generate and achieve accuracy comparable to the state-of-the-art kernel methods based on random Fourier features."
                    },
                    {
                        "title": "Low-Precision Random Fourier Features for Memory-Constrained Kernel Approximation",
                        "abstract": "We investigate how to train kernel approximation methods that generalize well under a memory budget. Building on recent theoretical work, we define a measure of kernel approximation error which we find to be more predictive of the empirical generalization performance of kernel approximation methods than conventional metrics. An important consequence of this definition is that a kernel approximation matrix must be high rank to attain close approximation. Because storing a high-rank approximation is memory intensive, we propose using a low-precision quantization of random Fourier features (LP-RFFs) to build a high-rank approximation under a memory budget. Theoretically, we show quantization has a negligible effect on generalization performance in important settings. Empirically, we demonstrate across four benchmark datasets that LP-RFFs can match the performance of full-precision RFFs and the Nystr\\\"{o}m method, with 3x-10x and 50x-460x less memory, respectively."
                    },
                    {
                        "title": "On the Downstream Performance of Compressed Word Embeddings",
                        "abstract": "Compressing word embeddings is important for deploying NLP models in memory-constrained settings. However, understanding what makes compressed embeddings perform well on downstream tasks is challenging---existing measures of compression quality often fail to distinguish between embeddings that perform well and those that do not. We thus propose the eigenspace overlap score as a new measure. We relate the eigenspace overlap score to downstream performance by developing generalization bounds for the compressed embeddings in terms of this score, in the context of linear and logistic regression. We then show that we can lower bound the eigenspace overlap score for a simple uniform quantization compression method, helping to explain the strong empirical performance of this method. Finally, we show that by using the eigenspace overlap score as a selection criterion between embeddings drawn from a representative set we compressed, we can efficiently identify the better performing embedding with up to $2\\times$ lower selection error rates than the next best measure of compression quality, and avoid the cost of training a model for each task of interest."
                    },
                    {
                        "title": "Learning Fast Algorithms for Linear Transforms Using Butterfly Factorizations",
                        "abstract": "Fast linear transforms are ubiquitous in machine learning, including the discrete Fourier transform, discrete cosine transform, and other structured transformations such as convolutions. All of these transforms can be represented by dense matrix-vector multiplication, yet each has a specialized and highly efficient (subquadratic) algorithm. We ask to what extent hand-crafting these algorithms and implementations is necessary, what structural priors they encode, and how much knowledge is required to automatically learn a fast algorithm for a provided structured transform. Motivated by a characterization of fast matrix-vector multiplication as products of sparse matrices, we introduce a parameterization of divide-and-conquer methods that is capable of representing a large class of transforms. This generic formulation can automatically learn an efficient algorithm for many important transforms; for example, it recovers the $O(N \\log N)$ Cooley-Tukey FFT algorithm to machine precision, for dimensions $N$ up to $1024$. Furthermore, our method can be incorporated as a lightweight replacement of generic matrices in machine learning pipelines to learn efficient and compressible transformations. On a standard task of compressing a single hidden-layer network, our method exceeds the classification accuracy of unconstrained matrices on CIFAR-10 by 3.9 points -- the first time a structured approach has done so -- with 4X faster inference speed and 40X fewer parameters."
                    },
                    {
                        "title": "Knowledge Distillation as Semiparametric Inference",
                        "abstract": "A popular approach to model compression is to train an inexpensive student model to mimic the class probabilities of a highly accurate but cumbersome teacher model. Surprisingly, this two-step knowledge distillation process often leads to higher accuracy than training the student directly on labeled data. To explain and enhance this phenomenon, we cast knowledge distillation as a semiparametric inference problem with the optimal student model as the target, the unknown Bayes class probabilities as nuisance, and the teacher probabilities as a plug-in nuisance estimate. By adapting modern semiparametric tools, we derive new guarantees for the prediction error of standard distillation and develop two enhancements -- cross-fitting and loss correction -- to mitigate the impact of teacher overfitting and underfitting on student performance. We validate our findings empirically on both tabular and image data and observe consistent improvements from our knowledge distillation enhancements."
                    },
                    {
                        "title": "HiPPO: Recurrent Memory with Optimal Polynomial Projections",
                        "abstract": "A central problem in learning from sequential data is representing cumulative history in an incremental fashion as more data is processed. We introduce a general framework (HiPPO) for the online compression of continuous signals and discrete time series by projection onto polynomial bases. Given a measure that specifies the importance of each time step in the past, HiPPO produces an optimal solution to a natural online function approximation problem. As special cases, our framework yields a short derivation of the recent Legendre Memory Unit (LMU) from first principles, and generalizes the ubiquitous gating mechanism of recurrent neural networks such as GRUs. This formal framework yields a new memory update mechanism (HiPPO-LegS) that scales through time to remember all history, avoiding priors on the timescale. HiPPO-LegS enjoys the theoretical benefits of timescale robustness, fast updates, and bounded gradients. By incorporating the memory dynamics into recurrent neural networks, HiPPO RNNs can empirically capture complex temporal dependencies. On the benchmark permuted MNIST dataset, HiPPO-LegS sets a new state-of-the-art accuracy of 98.3%. Finally, on a novel trajectory classification task testing robustness to out-of-distribution timescales and missing data, HiPPO-LegS outperforms RNN and neural ODE baselines by 25-40% accuracy."
                    },
                    {
                        "title": "FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision",
                        "abstract": "Attention, as a core layer of the ubiquitous Transformer architecture, is the bottleneck for large language models and long-context applications. FlashAttention elaborated an approach to speed up attention on GPUs through minimizing memory reads/writes. However, it has yet to take advantage of new capabilities present in recent hardware, with FlashAttention-2 achieving only 35% utilization on the H100 GPU. We develop three main techniques to speed up attention on Hopper GPUs: exploiting asynchrony of the Tensor Cores and TMA to (1) overlap overall computation and data movement via warp-specialization and (2) interleave block-wise matmul and softmax operations, and (3) block quantization and incoherent processing that leverages hardware support for FP8 low-precision. We demonstrate that our method, FlashAttention-3, achieves speedup on H100 GPUs by 1.5-2.0$\\times$ with FP16 reaching up to 740 TFLOPs/s (75% utilization), and with FP8 reaching close to 1.2 PFLOPs/s. We validate that FP8 FlashAttention-3 achieves 2.6$\\times$ lower numerical error than a baseline FP8 attention."
                    }
                ]
            },
            "58872768-e667-40ce-967f-9d43d891bf0b": {
                "pk": "58872768-e667-40ce-967f-9d43d891bf0b",
                "name": "Albert Gu",
                "collaborators": [
                    "Christopher R\u00e9",
                    "Tri Dao",
                    "Karan Goel",
                    "Atri Rudra",
                    "Ines Chami",
                    "Ankit Gupta",
                    "Isys Johnson",
                    "Dat Nguyen",
                    "Jonathan Berant",
                    "Anupam Gupta"
                ],
                "domain": [
                    "Deep Learning",
                    "State Space Models",
                    "Sequence Modeling",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality",
                        "abstract": "While Transformers have been the main architecture behind deep learning's success in language modeling, state-space models (SSMs) such as Mamba have recently been shown to match or outperform Transformers at small to medium scale. We show that these families of models are actually quite closely related, and develop a rich framework of theoretical connections between SSMs and variants of attention, connected through various decompositions of a well-studied class of structured semiseparable matrices. Our state space duality (SSD) framework allows us to design a new architecture (Mamba-2) whose core layer is an a refinement of Mamba's selective SSM that is 2-8X faster, while continuing to be competitive with Transformers on language modeling."
                    },
                    {
                        "title": "Mamba: Linear-Time Sequence Modeling with Selective State Spaces",
                        "abstract": "Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5$\\times$ higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation."
                    },
                    {
                        "title": "HoroPCA: Hyperbolic Dimensionality Reduction via Horospherical Projections",
                        "abstract": "This paper studies Principal Component Analysis (PCA) for data lying in hyperbolic spaces. Given directions, PCA relies on: (1) a parameterization of subspaces spanned by these directions, (2) a method of projection onto subspaces that preserves information in these directions, and (3) an objective to optimize, namely the variance explained by projections. We generalize each of these concepts to the hyperbolic space and propose HoroPCA, a method for hyperbolic dimensionality reduction. By focusing on the core problem of extracting principal directions, HoroPCA theoretically better preserves information in the original data such as distances, compared to previous generalizations of PCA. Empirically, we validate that HoroPCA outperforms existing dimensionality reduction methods, significantly reducing error in distance preservation. As a data whitening method, it improves downstream classification by up to 3.9% compared to methods that don't use whitening. Finally, we show that HoroPCA can be used to visualize hyperbolic data in two dimensions."
                    },
                    {
                        "title": "Diagonal State Spaces are as Effective as Structured State Spaces",
                        "abstract": "Modeling long range dependencies in sequential data is a fundamental step towards attaining human-level performance in many modalities such as text, vision, audio and video. While attention-based models are a popular and effective choice in modeling short-range interactions, their performance on tasks requiring long range reasoning has been largely inadequate. In an exciting result, Gu et al. (ICLR 2022) proposed the $\\textit{Structured State Space}$ (S4) architecture delivering large gains over state-of-the-art models on several long-range tasks across various modalities. The core proposition of S4 is the parameterization of state matrices via a diagonal plus low rank structure, allowing efficient computation. In this work, we show that one can match the performance of S4 even without the low rank correction and thus assuming the state matrices to be diagonal. Our $\\textit{Diagonal State Space}$ (DSS) model matches the performance of S4 on Long Range Arena tasks, speech classification on Speech Commands dataset, while being conceptually simpler and straightforward to implement."
                    },
                    {
                        "title": "On the Parameterization and Initialization of Diagonal State Space Models",
                        "abstract": "State space models (SSM) have recently been shown to be very effective as a deep learning layer as a promising alternative to sequence models such as RNNs, CNNs, or Transformers. The first version to show this potential was the S4 model, which is particularly effective on tasks involving long-range dependencies by using a prescribed state matrix called the HiPPO matrix. While this has an interpretable mathematical mechanism for modeling long dependencies, it introduces a custom representation and algorithm that can be difficult to implement. On the other hand, a recent variant of S4 called DSS showed that restricting the state matrix to be fully diagonal can still preserve the performance of the original model when using a specific initialization based on approximating S4's matrix. This work seeks to systematically understand how to parameterize and initialize such diagonal state space models. While it follows from classical results that almost all SSMs have an equivalent diagonal form, we show that the initialization is critical for performance. We explain why DSS works mathematically, by showing that the diagonal restriction of S4's matrix surprisingly recovers the same kernel in the limit of infinite state dimension. We also systematically describe various design choices in parameterizing and computing diagonal SSMs, and perform a controlled empirical study ablating the effects of these choices. Our final model S4D is a simple diagonal version of S4 whose kernel computation requires just 2 lines of code and performs comparably to S4 in almost all settings, with state-of-the-art results for image, audio, and medical time-series domains, and averaging 85\\% on the Long Range Arena benchmark."
                    },
                    {
                        "title": "Efficiently Modeling Long Sequences with Structured State Spaces",
                        "abstract": "A central goal of sequence modeling is designing a single principled model that can address sequence data across a range of modalities and tasks, particularly on long-range dependencies. Although conventional models including RNNs, CNNs, and Transformers have specialized variants for capturing long dependencies, they still struggle to scale to very long sequences of $10000$ or more steps. A promising recent approach proposed modeling sequences by simulating the fundamental state space model (SSM) \\( x'(t) = Ax(t) + Bu(t), y(t) = Cx(t) + Du(t) \\), and showed that for appropriate choices of the state matrix \\( A \\), this system could handle long-range dependencies mathematically and empirically. However, this method has prohibitive computation and memory requirements, rendering it infeasible as a general sequence modeling solution. We propose the Structured State Space sequence model (S4) based on a new parameterization for the SSM, and show that it can be computed much more efficiently than prior approaches while preserving their theoretical strengths. Our technique involves conditioning \\( A \\) with a low-rank correction, allowing it to be diagonalized stably and reducing the SSM to the well-studied computation of a Cauchy kernel. S4 achieves strong empirical results across a diverse range of established benchmarks, including (i) 91\\% accuracy on sequential CIFAR-10 with no data augmentation or auxiliary losses, on par with a larger 2-D ResNet, (ii) substantially closing the gap to Transformers on image and language modeling tasks, while performing generation $60\\times$ faster (iii) SoTA on every task from the Long Range Arena benchmark, including solving the challenging Path-X task of length 16k that all prior work fails on, while being as efficient as all competitors."
                    },
                    {
                        "title": "The Power of Deferral: Maintaining a Constant-Competitive Steiner Tree Online",
                        "abstract": "In the online Steiner tree problem, a sequence of points is revealed one-by-one: when a point arrives, we only have time to add a single edge connecting this point to the previous ones, and we want to minimize the total length of edges added. For two decades, we know that the greedy algorithm maintains a tree whose cost is O(log n) times the Steiner tree cost, and this is best possible. But suppose, in addition to the new edge we add, we can change a single edge from the previous set of edges: can we do much better? Can we maintain a tree that is constant-competitive?   We answer this question in the affirmative. We give a primal-dual algorithm, and a novel dual-based analysis, that makes only a single swap per step (in addition to adding the edge connecting the new point to the previous ones), and such that the tree's cost is only a constant times the optimal cost.   Previous results for this problem gave an algorithm that performed an amortized constant number of swaps: for each n, the number of swaps in the first n steps was O(n). We also give a simpler tight analysis for this amortized case."
                    },
                    {
                        "title": "Representation Tradeoffs for Hyperbolic Embeddings",
                        "abstract": "Hyperbolic embeddings offer excellent quality with few dimensions when embedding hierarchical data structures like synonym or type hierarchies. Given a tree, we give a combinatorial construction that embeds the tree in hyperbolic space with arbitrarily low distortion without using optimization. On WordNet, our combinatorial embedding obtains a mean-average-precision of 0.989 with only two dimensions, while Nickel et al.'s recent construction obtains 0.87 using 200 dimensions. We provide upper and lower bounds that allow us to characterize the precision-dimensionality tradeoff inherent in any hyperbolic embedding. To embed general metric spaces, we propose a hyperbolic generalization of multidimensional scaling (h-MDS). We show how to perform exact recovery of hyperbolic points from distances, provide a perturbation analysis, and give a recovery result that allows us to reduce dimensionality. The h-MDS approach offers consistently low distortion even with few dimensions across several datasets. Finally, we extract lessons from the algorithms and theory above to design a PyTorch-based implementation that can handle incomplete information and is scalable."
                    },
                    {
                        "title": "HiPPO: Recurrent Memory with Optimal Polynomial Projections",
                        "abstract": "A central problem in learning from sequential data is representing cumulative history in an incremental fashion as more data is processed. We introduce a general framework (HiPPO) for the online compression of continuous signals and discrete time series by projection onto polynomial bases. Given a measure that specifies the importance of each time step in the past, HiPPO produces an optimal solution to a natural online function approximation problem. As special cases, our framework yields a short derivation of the recent Legendre Memory Unit (LMU) from first principles, and generalizes the ubiquitous gating mechanism of recurrent neural networks such as GRUs. This formal framework yields a new memory update mechanism (HiPPO-LegS) that scales through time to remember all history, avoiding priors on the timescale. HiPPO-LegS enjoys the theoretical benefits of timescale robustness, fast updates, and bounded gradients. By incorporating the memory dynamics into recurrent neural networks, HiPPO RNNs can empirically capture complex temporal dependencies. On the benchmark permuted MNIST dataset, HiPPO-LegS sets a new state-of-the-art accuracy of 98.3%. Finally, on a novel trajectory classification task testing robustness to out-of-distribution timescales and missing data, HiPPO-LegS outperforms RNN and neural ODE baselines by 25-40% accuracy."
                    },
                    {
                        "title": "How to Train Your HiPPO: State Space Models with Generalized Orthogonal Basis Projections",
                        "abstract": "Linear time-invariant state space models (SSM) are a classical model from engineering and statistics, that have recently been shown to be very promising in machine learning through the Structured State Space sequence model (S4). A core component of S4 involves initializing the SSM state matrix to a particular matrix called a HiPPO matrix, which was empirically important for S4's ability to handle long sequences. However, the specific matrix that S4 uses was actually derived in previous work for a particular time-varying dynamical system, and the use of this matrix as a time-invariant SSM had no known mathematical interpretation. Consequently, the theoretical mechanism by which S4 models long-range dependencies actually remains unexplained. We derive a more general and intuitive formulation of the HiPPO framework, which provides a simple mathematical interpretation of S4 as a decomposition onto exponentially-warped Legendre polynomials, explaining its ability to capture long dependencies. Our generalization introduces a theoretically rich class of SSMs that also lets us derive more intuitive S4 variants for other bases such as the Fourier basis, and explains other aspects of training S4, such as how to initialize the important timescale parameter. These insights improve S4's performance to 86% on the Long Range Arena benchmark, with 96% on the most difficult Path-X task."
                    },
                    {
                        "title": "Model Patching: Closing the Subgroup Performance Gap with Data Augmentation",
                        "abstract": "Classifiers in machine learning are often brittle when deployed. Particularly concerning are models with inconsistent performance on specific subgroups of a class, e.g., exhibiting disparities in skin cancer classification in the presence or absence of a spurious bandage. To mitigate these performance differences, we introduce model patching, a two-stage framework for improving robustness that encourages the model to be invariant to subgroup differences, and focus on class information shared by subgroups. Model patching first models subgroup features within a class and learns semantic transformations between them, and then trains a classifier with data augmentations that deliberately manipulate subgroup features. We instantiate model patching with CAMEL, which (1) uses a CycleGAN to learn the intra-class, inter-subgroup augmentations, and (2) balances subgroup performance using a theoretically-motivated subgroup consistency regularizer, accompanied by a new robust objective. We demonstrate CAMEL's effectiveness on 3 benchmark datasets, with reductions in robust error of up to 33% relative to the best baseline. Lastly, CAMEL successfully patches a model that fails due to spurious features on a real-world skin cancer dataset."
                    },
                    {
                        "title": "Learning Fast Algorithms for Linear Transforms Using Butterfly Factorizations",
                        "abstract": "Fast linear transforms are ubiquitous in machine learning, including the discrete Fourier transform, discrete cosine transform, and other structured transformations such as convolutions. All of these transforms can be represented by dense matrix-vector multiplication, yet each has a specialized and highly efficient (subquadratic) algorithm. We ask to what extent hand-crafting these algorithms and implementations is necessary, what structural priors they encode, and how much knowledge is required to automatically learn a fast algorithm for a provided structured transform. Motivated by a characterization of fast matrix-vector multiplication as products of sparse matrices, we introduce a parameterization of divide-and-conquer methods that is capable of representing a large class of transforms. This generic formulation can automatically learn an efficient algorithm for many important transforms; for example, it recovers the $O(N \\log N)$ Cooley-Tukey FFT algorithm to machine precision, for dimensions $N$ up to $1024$. Furthermore, our method can be incorporated as a lightweight replacement of generic matrices in machine learning pipelines to learn efficient and compressible transformations. On a standard task of compressing a single hidden-layer network, our method exceeds the classification accuracy of unconstrained matrices on CIFAR-10 by 3.9 points -- the first time a structured approach has done so -- with 4X faster inference speed and 40X fewer parameters."
                    },
                    {
                        "title": "From Trees to Continuous Embeddings and Back: Hyperbolic Hierarchical Clustering",
                        "abstract": "Similarity-based Hierarchical Clustering (HC) is a classical unsupervised machine learning algorithm that has traditionally been solved with heuristic algorithms like Average-Linkage. Recently, Dasgupta reframed HC as a discrete optimization problem by introducing a global cost function measuring the quality of a given tree. In this work, we provide the first continuous relaxation of Dasgupta's discrete optimization problem with provable quality guarantees. The key idea of our method, HypHC, is showing a direct correspondence from discrete trees to continuous representations (via the hyperbolic embeddings of their leaf nodes) and back (via a decoding algorithm that maps leaf embeddings to a dendrogram), allowing us to search the space of discrete binary trees with continuous optimization. Building on analogies between trees and hyperbolic space, we derive a continuous analogue for the notion of lowest common ancestor, which leads to a continuous relaxation of Dasgupta's discrete objective. We can show that after decoding, the global minimizer of our continuous relaxation yields a discrete tree with a (1 + epsilon)-factor approximation for Dasgupta's optimal tree, where epsilon can be made arbitrarily small and controls optimization challenges. We experimentally evaluate HypHC on a variety of HC benchmarks and find that even approximate solutions found with gradient descent have superior clustering quality than agglomerative heuristics or other gradient based algorithms. Finally, we highlight the flexibility of HypHC using end-to-end training in a downstream classification task."
                    },
                    {
                        "title": "It's Raw! Audio Generation with State-Space Models",
                        "abstract": "Developing architectures suitable for modeling raw audio is a challenging problem due to the high sampling rates of audio waveforms. Standard sequence modeling approaches like RNNs and CNNs have previously been tailored to fit the demands of audio, but the resultant architectures make undesirable computational tradeoffs and struggle to model waveforms effectively. We propose SaShiMi, a new multi-scale architecture for waveform modeling built around the recently introduced S4 model for long sequence modeling. We identify that S4 can be unstable during autoregressive generation, and provide a simple improvement to its parameterization by drawing connections to Hurwitz matrices. SaShiMi yields state-of-the-art performance for unconditional waveform generation in the autoregressive setting. Additionally, SaShiMi improves non-autoregressive generation performance when used as the backbone architecture for a diffusion model. Compared to prior architectures in the autoregressive generation setting, SaShiMi generates piano and speech waveforms which humans find more musical and coherent respectively, e.g. 2x better mean opinion scores than WaveNet on an unconditional speech generation task. On a music generation task, SaShiMi outperforms WaveNet on density estimation and speed at both training and inference even when using 3x fewer parameters. Code can be found at https://github.com/HazyResearch/state-spaces and samples at https://hazyresearch.stanford.edu/sashimi-examples."
                    },
                    {
                        "title": "Improving the Gating Mechanism of Recurrent Neural Networks",
                        "abstract": "Gating mechanisms are widely used in neural network models, where they allow gradients to backpropagate more easily through depth or time. However, their saturation property introduces problems of its own. For example, in recurrent models these gates need to have outputs near 1 to propagate information over long time-delays, which requires them to operate in their saturation regime and hinders gradient-based learning of the gate mechanism. We address this problem by deriving two synergistic modifications to the standard gating mechanism that are easy to implement, introduce no additional hyperparameters, and improve learnability of the gates when they are close to saturation. We show how these changes are related to and improve on alternative recently proposed gating mechanisms such as chrono initialization and Ordered Neurons. Empirically, our simple gating mechanisms robustly improve the performance of recurrent models on a range of applications, including synthetic memorization tasks, sequential image classification, language modeling, and reinforcement learning, particularly when long-term dependencies are involved."
                    },
                    {
                        "title": "Combining Recurrent, Convolutional, and Continuous-time Models with Linear State-Space Layers",
                        "abstract": "Recurrent neural networks (RNNs), temporal convolutions, and neural differential equations (NDEs) are popular families of deep learning models for time-series data, each with unique strengths and tradeoffs in modeling power and computational efficiency. We introduce a simple sequence model inspired by control systems that generalizes these approaches while addressing their shortcomings. The Linear State-Space Layer (LSSL) maps a sequence $u \\mapsto y$ by simply simulating a linear continuous-time state-space representation $\\dot{x} = Ax + Bu, y = Cx + Du$. Theoretically, we show that LSSL models are closely related to the three aforementioned families of models and inherit their strengths. For example, they generalize convolutions to continuous-time, explain common RNN heuristics, and share features of NDEs such as time-scale adaptation. We then incorporate and generalize recent theory on continuous-time memorization to introduce a trainable subset of structured matrices $A$ that endow LSSLs with long-range memory. Empirically, stacking LSSL layers into a simple deep neural network obtains state-of-the-art results across time series benchmarks for long dependencies in sequential image classification, real-world healthcare regression tasks, and speech. On a difficult speech classification task with length-16000 sequences, LSSL outperforms prior approaches by 24 accuracy points, and even outperforms baselines that use hand-crafted features on 100x shorter sequences."
                    }
                ]
            }
        }
    },
    "2405.11891": {
        "paper_data": {
            "title": "Unveiling and Manipulating Prompt Influence in Large Language Models",
            "url": "http://arxiv.org/abs/2405.11891v1",
            "arxiv_id": "2405.11891",
            "authors": [
                "Zijian Feng",
                "Hanzhang Zhou",
                "Zixiao Zhu",
                "Junlang Qian",
                "Kezhi Mao"
            ],
            "abstract": "Prompts play a crucial role in guiding the responses of Large Language Models (LLMs). However, the intricate role of individual tokens in prompts, known as input saliency, in shaping the responses remains largely underexplored. Existing saliency methods either misalign with LLM generation objectives or rely heavily on linearity assumptions, leading to potential inaccuracies. To address this, we propose Token Distribution Dynamics (TDD), a \\textcolor{black}{simple yet effective} approach to unveil and manipulate the role of prompts in generating LLM outputs. TDD leverages the robust interpreting capabilities of the language model head (LM head) to assess input saliency. It projects input tokens into the embedding space and then estimates their significance based on distribution dynamics over the vocabulary. We introduce three TDD variants: forward, backward, and bidirectional, each offering unique insights into token relevance. Extensive experiments reveal that the TDD surpasses state-of-the-art baselines with a big margin in elucidating the causal relationships between prompts and LLM outputs. Beyond mere interpretation, we apply TDD to two prompt manipulation tasks for controlled text generation: zero-shot toxic language suppression and sentiment steering. Empirical results underscore TDD's proficiency in identifying both toxic and sentimental cues in prompts, subsequently mitigating toxicity or modulating sentiment in the generated content.",
            "introduction": "   1 Introduction  The formulation of prompts significantly shapes the textual responses of large language models (LLMs). Comprehending the influence of individual input tokens in prompts, i.e., input saliency, can augment our insight into LLM interpretability and foster the development of refined prompting strategies to modulate LLM outputs. However, input saliency in LLMs and its bearing on prompt-based generation control are rarely explored.   Although various saliency methods ranging from perturbation-based (Feng et\u00a0al., 2018; Prabhakaran et\u00a0al., 2019), gradient-based (Chefer et\u00a0al., 2021; Ali et\u00a0al., 2022), to vector-based (Modarressi et\u00a0al., 2022; Ferrando et\u00a0al., 2022) for Transformer-based language models have been devised to gauge input token significance, their adaptation to LLMs presents certain limitations. First, these methods are chiefly designed for text classification using masked language models such as BERT (Devlin et\u00a0al., 2019) and RoBERTa (Liu et\u00a0al., 2019). Their saliency explanations, limited to class labels, do not align with the objectives of autoregressive LLMs, which aim to explain token generation across the whole vocabulary. Second, many of these methods rely heavily on linearity assumptions to approximate the intricate non-linear behaviors in language models. For instance, vector-based approaches (Modarressi et\u00a0al., 2022; Ferrando et\u00a0al., 2022) often assume linearly combined attention weights across layers, while gradient-based ones (Wallace et\u00a0al., 2019; Chefer et\u00a0al., 2021; Ali et\u00a0al., 2022) employ local linear approximations, referencing Taylor\u2019s expansion theorem. Such assumptions can lead to significant inaccuracies, especially as the number of LLM layers increases. These limitations can hinder the understanding of how LLMs respond to prompts and hence the effective use of prompts for desired outputs.   Unlike previous methods that use attention or gradient as a medium to compute saliency, in this study, we propose adopting token distributions as a means to estimate token saliency. Recent studies on GPT2 show that token representations can be visualized as distributions across the vocabulary, termed token distributions, through the LM head (Nostalgebraist, 2020; Geva et\u00a0al., 2021; 2022; Dar et\u00a0al., 2023). We expand this notion to other contemporary LLMs such as Pythia (Biderman et\u00a0al., 2023) and LLaMA2 (Touvron et\u00a0al., 2023), emphasizing the relationship between token distributions and token contributions. Our findings indicate that the LM head acts as an effective interpreter for LLMs, skillfully decoding ambiguous input token representations and mapping them to the embedding space. These projected token distributions within the embedding space are interpretable and comprehensible to humans, and align with model predictions. We argue that variations in token distributions across layers can be attributed to token saliency.   To unveil token saliency in prompts for LLM generation, we introduce the Token Distribution Dynamics (TDD) approach. In alignment with the generative objectives of LLMs, our methodology offers contrastive explanations (Yin & Neubig, 2022), elucidating why LLMs prioritize one token over others in the entire vocabulary. We present three variants: TDD-forward, TDD-backward, and TDD-bidirectional. TDD-forward capitalizes on token distribution dynamics throughout the progression of learning and prediction, while TDD-backward evaluates the token importance using backward token dynamics. In parallel, TDD-bidirectional integrates insights from both directional perspectives. Despite its apparent simplicity, TDD serves as an exceptionally potent tool. Comprehensive experiments show that TDD markedly outperforms advanced saliency methods in discerning the causal relationships",
            "references": [
                {
                    "title": "Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignment",
                    "abstract": "Ensuring alignment, which refers to making models behave in accordance with human intentions [1,2], has become a critical task before deploying large language models (LLMs) in real-world applications. For instance, OpenAI devoted six months to iteratively aligning GPT-4 before its release [3]. However, a major challenge faced by practitioners is the lack of clear guidance on evaluating whether LLM outputs align with social norms, values, and regulations. This obstacle hinders systematic iteration and deployment of LLMs. To address this issue, this paper presents a comprehensive survey of key dimensions that are crucial to consider when assessing LLM trustworthiness. The survey covers seven major categories of LLM trustworthiness: reliability, safety, fairness, resistance to misuse, explainability and reasoning, adherence to social norms, and robustness. Each major category is further divided into several sub-categories, resulting in a total of 29 sub-categories. Additionally, a subset of 8 sub-categories is selected for further investigation, where corresponding measurement studies are designed and conducted on several widely-used LLMs. The measurement results indicate that, in general, more aligned models tend to perform better in terms of overall trustworthiness. However, the effectiveness of alignment varies across the different trustworthiness categories considered. This highlights the importance of conducting more fine-grained analyses, testing, and making continuous improvements on LLM alignment. By shedding light on these key dimensions of LLM trustworthiness, this paper aims to provide valuable insights and guidance to practitioners in the field. Understanding and addressing these concerns will be crucial in achieving reliable and ethically sound deployment of LLMs in various applications."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Models",
                    "abstract": "Generative Pre-trained Transformer (GPT) models have exhibited exciting progress in their capabilities, capturing the interest of practitioners and the public alike. Yet, while the literature on the trustworthiness of GPT models remains limited, practitioners have proposed employing capable GPT models for sensitive applications such as healthcare and finance -- where mistakes can be costly. To this end, this work proposes a comprehensive trustworthiness evaluation for large language models with a focus on GPT-4 and GPT-3.5, considering diverse perspectives -- including toxicity, stereotype bias, adversarial robustness, out-of-distribution robustness, robustness on adversarial demonstrations, privacy, machine ethics, and fairness. Based on our evaluations, we discover previously unpublished vulnerabilities to trustworthiness threats. For instance, we find that GPT models can be easily misled to generate toxic and biased outputs and leak private information in both training data and conversation history. We also find that although GPT-4 is usually more trustworthy than GPT-3.5 on standard benchmarks, GPT-4 is more vulnerable given jailbreaking system or user prompts, potentially because GPT-4 follows (misleading) instructions more precisely. Our work illustrates a comprehensive trustworthiness evaluation of GPT models and sheds light on the trustworthiness gaps. Our benchmark is publicly available at https://decodingtrust.github.io/ ; our dataset can be previewed at https://huggingface.co/datasets/AI-Secure/DecodingTrust ; a concise version of this work is at https://openreview.net/pdf?id=kaHpo8OZw2 ."
                },
                {
                    "title": "DecompX: Explaining Transformers Decisions by Propagating Token Decomposition",
                    "abstract": "An emerging solution for explaining Transformer-based models is to use vector-based analysis on how the representations are formed. However, providing a faithful vector-based explanation for a multi-layer model could be challenging in three aspects: (1) Incorporating all components into the analysis, (2) Aggregating the layer dynamics to determine the information flow and mixture throughout the entire model, and (3) Identifying the connection between the vector-based analysis and the model\u2019s predictions. In this paper, we present DecompX to tackle these challenges. DecompX is based on the construction of decomposed token representations and their successive propagation throughout the model without mixing them in between layers. Additionally, our proposal provides multiple advantages over existing solutions for its inclusion of all encoder components (especially nonlinear feed-forward networks) and the classification head. The former allows acquiring precise vectors while the latter transforms the decomposition into meaningful prediction-based values, eliminating the need for norm- or summation-based vector aggregation. According to the standard faithfulness evaluations, DecompX consistently outperforms existing gradient-based and vector-based approaches on various datasets.Our code is available at https://github.com/mohsenfayyaz/DecompX."
                },
                {
                    "title": "Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling",
                    "abstract": "How do large language models (LLMs) develop and evolve over the course of training? How do these patterns change as models scale? To answer these questions, we introduce \\textit{Pythia}, a suite of 16 LLMs all trained on public data seen in the exact same order and ranging in size from 70M to 12B parameters. We provide public access to 154 checkpoints for each one of the 16 models, alongside tools to download and reconstruct their exact training dataloaders for further study. We intend \\textit{Pythia} to facilitate research in many areas, and we present several case studies including novel results in memorization, term frequency effects on few-shot performance, and reducing gender bias. We demonstrate that this highly controlled setup can be used to yield novel insights toward LLMs and their training dynamics. Trained models, analysis code, training code, and training data can be found at \\url{https://github.com/EleutherAI/pythia}."
                },
                {
                    "title": "AtMan: Understanding Transformer Predictions Through Memory Efficient Attention Manipulation",
                    "abstract": "Generative transformer models have become increasingly complex, with large numbers of parameters and the ability to process multiple input modalities. Current methods for explaining their predictions are resource-intensive. Most crucially, they require prohibitively large amounts of extra memory, since they rely on backpropagation which allocates almost twice as much GPU memory as the forward pass. This makes it difficult, if not impossible, to use them in production. We present AtMan that provides explanations of generative transformer models at almost no extra cost. Specifically, AtMan is a modality-agnostic perturbation method that manipulates the attention mechanisms of transformers to produce relevance maps for the input with respect to the output prediction. Instead of using backpropagation, AtMan applies a parallelizable token-based search method based on cosine similarity neighborhood in the embedding space. Our exhaustive experiments on text and image-text benchmarks demonstrate that AtMan outperforms current state-of-the-art gradient-based methods on several metrics while being computationally efficient. As such, AtMan is suitable for use in large model inference deployments."
                },
                {
                    "title": "BLOOM: A 176B-Parameter Open-Access Multilingual Language Model",
                    "abstract": "Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are frequently kept from the public. As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model designed and built thanks to a collaboration of hundreds of researchers. BLOOM is a decoder-only Transformer language model that was trained on the ROOTS corpus, a dataset comprising hundreds of sources in 46 natural and 13 programming languages (59 in total). We find that BLOOM achieves competitive performance on a wide variety of benchmarks, with stronger results after undergoing multitask prompted finetuning. To facilitate future research and applications using LLMs, we publicly release our models and code under the Responsible AI License."
                },
                {
                    "title": "Analyzing Transformers in Embedding Space",
                    "abstract": "Understanding Transformer-based models has attracted significant attention, as they lie at the heart of recent technological advances across machine learning. While most interpretability methods rely on running models over inputs, recent work has shown that a zero-pass approach, where parameters are interpreted directly without a forward/backward pass is feasible for some Transformer parameters, and for two-layer attention networks. In this work, we present a theoretical analysis where all parameters of a trained Transformer are interpreted by projecting them into the embedding space, that is, the space of vocabulary items they operate on. We derive a simple theoretical framework to support our arguments and provide ample evidence for its validity. First, an empirical analysis showing that parameters of both pretrained and fine-tuned models can be interpreted in embedding space. Second, we present two applications of our framework: (a) aligning the parameters of different models that share a vocabulary, and (b) constructing a classifier without training by \u201ctranslating\u201d the parameters of a fine-tuned classifier to parameters of a different model that was only pretrained. Overall, our findings open the door to interpretation methods that, at least in part, abstract away from model specifics and operate in the embedding space only."
                },
                {
                    "title": "GlobEnc: Quantifying Global Token Attribution by Incorporating the Whole Encoder Layer in Transformers",
                    "abstract": "There has been a growing interest in interpreting the underlying dynamics of Transformers. While self-attention patterns were initially deemed as the primary option, recent studies have shown that integrating other components can yield more accurate explanations. This paper introduces a novel token attribution analysis method that incorporates all the components in the encoder block and aggregates this throughout layers. Through extensive quantitative and qualitative experiments, we demonstrate that our method can produce faithful and meaningful global token attributions. Our experiments reveal that incorporating almost every encoder component results in increasingly more accurate analysis in both local (single layer) and global (the whole model) settings. Our global attribution analysis significantly outperforms previous methods on various tasks regarding correlation with gradient-based saliency scores. Our code is freely available at https://github.com/mohsenfayyaz/GlobEnc."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space",
                    "abstract": "Transformer-based language models (LMs) are at the core of modern NLP, but their internal prediction construction process is opaque and largely not understood. In this work, we make a substantial step towards unveiling this underlying prediction process, by reverse-engineering the operation of the feed-forward network (FFN) layers, one of the building blocks of transformer models. We view the token representation as a changing distribution over the vocabulary, and the output from each FFN layer as an additive update to that distribution. Then, we analyze the FFN updates in the vocabulary space, showing that each update can be decomposed to sub-updates corresponding to single FFN parameter vectors, each promoting concepts that are often human-interpretable. We then leverage these findings for controlling LM predictions, where we reduce the toxicity of GPT2 by almost 50%, and for improving computation efficiency with a simple early exit rule, saving 20% of computation on average."
                },
                {
                    "title": "Measuring the Mixing of Contextual Information in the Transformer",
                    "abstract": "The Transformer architecture aggregates input information through the self-attention mechanism, but there is no clear understanding of how this information is mixed across the entire model. Additionally, recent works have demonstrated that attention weights alone are not enough to describe the flow of information. In this paper, we consider the whole attention block \u2013multi-head attention, residual connection, and layer normalization\u2013 and define a metric to measure token-to-token interactions within each layer. Then, we aggregate layer-wise interpretations to provide input attribution scores for model predictions. Experimentally, we show that our method, ALTI (Aggregation of Layer-wise Token-to-token Interactions), provides more faithful explanations and increased robustness than gradient-based methods."
                },
                {
                    "title": "Interpreting Language Models with Contrastive Explanations",
                    "abstract": "Model interpretability methods are often used to explain NLP model decisions on tasks such as text classification, where the output space is relatively small. However, when applied to language generation, where the output space often consists of tens of thousands of tokens, these methods are unable to provide informative explanations. Language models must consider various features to predict a token, such as its part of speech, number, tense, or semantics.Existing explanation methods conflate evidence for all these features into a single explanation, which is less interpretable for human understanding.To disentangle the different decisions in language modeling, we focus on explaining language models contrastively: we look for salient input tokens that explain why the model predicted one token instead of another. We demonstrate that contrastive explanations are quantifiably better than non-contrastive explanations in verifying major grammatical phenomena, and that they significantly improve contrastive model simulatability for human observers. We also identify groups of contrastive decisions where the model uses similar evidence, and we are able to characterize what input tokens models use during various language generation decisions."
                },
                {
                    "title": "XAI for Transformers: Better Explanations through Conservative Propagation",
                    "abstract": "Transformers have become an important workhorse of machine learning, with numerous applications. This necessitates the development of reliable methods for increasing their transparency. Multiple interpretability methods, often based on gradient information, have been proposed. We show that the gradient in a Transformer reflects the function only locally, and thus fails to reliably identify the contribution of input features to the prediction. We identify Attention Heads and LayerNorm as main reasons for such unreliable explanations and propose a more stable way for propagation through these layers. Our proposal, which can be seen as a proper extension of the well-established LRP method to Transformers, is shown both theoretically and empirically to overcome the deficiency of a simple gradient-based approach, and achieves state-of-the-art explanation performance on a broad range of Transformer models and datasets."
                },
                {
                    "title": "Transductive Learning for Unsupervised Text Style Transfer",
                    "abstract": "Unsupervised style transfer models are mainly based on an inductive learning approach, which represents the style as embeddings, decoder parameters, or discriminator parameters and directly applies these general rules to the test cases. However, the lacking of parallel corpus hinders the ability of these inductive learning methods on this task. As a result, it is likely to cause severe inconsistent style expressions, like \u2018the salad is rude\u2019. To tackle this problem, we propose a novel transductive learning approach in this paper, based on a retrieval-based context-aware style representation. Specifically, an attentional encoder-decoder with a retriever framework is utilized. It involves top-K relevant sentences in the target style in the transfer process. In this way, we can learn a context-aware style embedding to alleviate the above inconsistency problem. In this paper, both sparse (BM25) and dense retrieval functions (MIPS) are used, and two objective functions are designed to facilitate joint learning. Experimental results show that our method outperforms several strong baselines. The proposed transductive learning approach is general and effective to the task of unsupervised style transfer, and we will apply it to the other two typical methods in the future."
                },
                {
                    "title": "A Recipe for Arbitrary Text Style Transfer with Large Language Models",
                    "abstract": "In this paper, we leverage large language models (LLMs) to perform zero-shot text style transfer. We present a prompting method that we call augmented zero-shot learning, which frames style transfer as a sentence rewriting task and requires only a natural language instruction, without model fine-tuning or exemplars in the target style. Augmented zero-shot learning is simple and demonstrates promising results not just on standard style transfer tasks such as sentiment, but also on arbitrary transformations such as \u2018make this melodramatic\u2019 or \u2018insert a metaphor.\u2019"
                },
                {
                    "title": "Knowledgeable Prompt-tuning: Incorporating Knowledge into Prompt Verbalizer for Text Classification",
                    "abstract": "Tuning pre-trained language models (PLMs) with task-specific prompts has been a promising approach for text classification. Particularly, previous studies suggest that prompt-tuning has remarkable superiority in the low-data scenario over the generic fine-tuning methods with extra classifiers. The core idea of prompt-tuning is to insert text pieces, i.e., template, to the input and transform a classification problem into a masked language modeling problem, where a crucial step is to construct a projection, i.e., verbalizer, between a label space and a label word space. A verbalizer is usually handcrafted or searched by gradient descent, which may lack coverage and bring considerable bias and high variances to the results. In this work, we focus on incorporating external knowledge into the verbalizer, forming a knowledgeable prompttuning (KPT), to improve and stabilize prompttuning. Specifically, we expand the label word space of the verbalizer using external knowledge bases (KBs) and refine the expanded label word space with the PLM itself before predicting with the expanded label word space. Extensive experiments on zero and few-shot text classification tasks demonstrate the effectiveness of knowledgeable prompt-tuning."
                },
                {
                    "title": "DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Experts",
                    "abstract": "Despite recent advances in natural language generation, it remains challenging to control attributes of generated text. We propose DExperts: Decoding-time Experts, a decoding-time method for controlled text generation that combines a pretrained language model with \u201cexpert\u201d LMs and/or \u201canti-expert\u201d LMs in a product of experts. Intuitively, under the ensemble, tokens only get high probability if they are considered likely by the experts, and unlikely by the anti-experts. We apply DExperts to language detoxification and sentiment-controlled generation, where we outperform existing controllable generation methods on both automatic and human evaluations. Moreover, because DExperts operates only on the output of the pretrained LM, it is effective with (anti-)experts of smaller size, including when operating on GPT-3. Our work highlights the promise of tuning small LMs on text with (un)desirable attributes for efficient decoding-time steering."
                },
                {
                    "title": "Generic Attention-model Explainability for Interpreting Bi-Modal and Encoder-Decoder Transformers",
                    "abstract": "Transformers are increasingly dominating multi-modal reasoning tasks, such as visual question answering, achieving state-of-the-art results thanks to their ability to contextualize information using the self-attention and co-attention mechanisms. These attention modules also play a role in other computer vision tasks including object detection and image segmentation. Unlike Transformers that only use self-attention, Transformers with co-attention require to consider multiple attention maps in parallel in order to highlight the information that is relevant to the prediction in the model\u2019s input. In this work, we propose the first method to explain prediction by any Transformer-based architecture, including bi-modal Transformers and Transformers with co-attentions. We provide generic solutions and apply these to the three most commonly used of these architectures: (i) pure self-attention, (ii) self-attention combined with co-attention, and (iii) encoder-decoder attention. We show that our method is superior to all existing methods which are adapted from single modality explainability. Our code is available at: https://github.com/hila-chefer/Transformer-MM-Explainability."
                },
                {
                    "title": "Contrastive Explanations for Model Interpretability",
                    "abstract": "Contrastive explanations clarify why an event occurred in contrast to another. They are inherently intuitive to humans to both produce and comprehend. We propose a method to produce contrastive explanations in the latent space, via a projection of the input representation, such that only the features that differentiate two potential decisions are captured. Our modification allows model behavior to consider only contrastive reasoning, and uncover which aspects of the input are useful for and against particular decisions. Our contrastive explanations can additionally answer for which label, and against which alternative label, is a given input feature useful. We produce contrastive explanations via both high-level abstract concept attribution and low-level input token/span attribution for two NLP classification benchmarks. Our findings demonstrate the ability of label-contrastive explanations to provide fine-grained interpretability of model decisions."
                },
                {
                    "title": "Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLP",
                    "abstract": "Abstract \u26a0 This paper contains prompts and model outputs that are offensive in nature. When trained on large, unfiltered crawls from the Internet, language models pick up and reproduce all kinds of undesirable biases that can be found in the data: They often generate racist, sexist, violent, or otherwise toxic language. As large models require millions of training examples to achieve good performance, it is difficult to completely prevent them from being exposed to such content. In this paper, we first demonstrate a surprising finding: Pretrained language models recognize, to a considerable degree, their undesirable biases and the toxicity of the content they produce. We refer to this capability as self-diagnosis. Based on this finding, we then propose a decoding algorithm that, given only a textual description of the undesired behavior, reduces the probability of a language model producing problematic text. We refer to this approach as self-debiasing. Self-debiasing does not rely on manually curated word lists, nor does it require any training data or changes to the model\u2019s parameters. While we by no means eliminate the issue of language models generating biased text, we believe our approach to be an important step in this direction.1"
                },
                {
                    "title": "Transformer Feed-Forward Layers Are Key-Value Memories",
                    "abstract": "Feed-forward layers constitute two-thirds of a transformer model\u2019s parameters, yet their role in the network remains under-explored. We show that feed-forward layers in transformer-based language models operate as key-value memories, where each key correlates with textual patterns in the training examples, and each value induces a distribution over the output vocabulary. Our experiments show that the learned patterns are human-interpretable, and that lower layers tend to capture shallow patterns, while upper layers learn more semantic ones. The values complement the keys\u2019 input patterns by inducing output distributions that concentrate probability mass on tokens likely to appear immediately after each pattern, particularly in the upper layers. Finally, we demonstrate that the output of a feed-forward layer is a composition of its memories, which is subsequently refined throughout the model\u2019s layers via residual connections to produce the final output distribution."
                },
                {
                    "title": "Reformulating Unsupervised Style Transfer as Paraphrase Generation",
                    "abstract": "Modern NLP defines the task of style transfer as modifying the style of a given sentence without appreciably changing its semantics, which implies that the outputs of style transfer systems should be paraphrases of their inputs. However, many existing systems purportedly designed for style transfer inherently warp the input's meaning through attribute transfer, which changes semantic properties such as sentiment. In this paper, we reformulate unsupervised style transfer as a paraphrase generation problem, and present a simple methodology based on fine-tuning pretrained language models on automatically generated paraphrase data. Despite its simplicity, our method significantly outperforms state-of-the-art style transfer systems on both human and automatic evaluations. We also survey 23 style transfer papers and discover that existing automatic metrics can be easily gamed and propose fixed variants. Finally, we pivot to a more real-world style transfer setting by collecting a large dataset of 15M sentences in 11 diverse styles, which we use for an in-depth analysis of our system."
                },
                {
                    "title": "A Diagnostic Study of Explainability Techniques for Text Classification",
                    "abstract": "Recent developments in machine learning have introduced models that approach human performance at the cost of increased architectural complexity. Efforts to make the rationales behind the models' predictions transparent have inspired an abundance of new explainability techniques. Provided with an already trained model, they compute saliency scores for the words of an input instance. However, there exists no definitive guide on (i) how to choose such a technique given a particular application task and model architecture, and (ii) the benefits and drawbacks of using each such technique. In this paper, we develop a comprehensive list of diagnostic properties for evaluating existing explainability techniques. We then employ the proposed list to compare a set of diverse explainability techniques on downstream text classification tasks and neural network architectures. We also compare the saliency scores assigned by the explainability techniques with human annotations of salient input regions to find relations between a model's performance and the agreement of its rationales with human ones. Overall, we find that the gradient-based explanations perform best across tasks and model architectures, and we present further insights into the properties of the reviewed explainability techniques."
                },
                {
                    "title": "RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models",
                    "abstract": "Pretrained neural language models (LMs) are prone to generating racist, sexist, or otherwise toxic language which hinders their safe deployment. We investigate the extent to which pretrained LMs can be prompted to generate toxic language, and the effectiveness of controllable text generation algorithms at preventing such toxic degeneration. We create and release RealToxicityPrompts, a dataset of 100K naturally occurring, sentence-level prompts derived from a large corpus of English web text, paired with toxicity scores from a widely-used toxicity classifier. Using RealToxicityPrompts, we find that pretrained LMs can degenerate into toxic text even from seemingly innocuous prompts. We empirically assess several controllable generation methods, and find that while data- or compute-intensive methods (e.g., adaptive pretraining on non-toxic data) are more effective at steering away from toxicity than simpler solutions (e.g., banning \u201cbad\u201d words), no current method is failsafe against neural toxic degeneration. To pinpoint the potential cause of such persistent toxic degeneration, we analyze two web text corpora used to pretrain several LMs (including GPT-2; Radford et. al, 2019), and find a significant amount of offensive, factually unreliable, and otherwise toxic content. Our work provides a test bed for evaluating toxic generations by LMs and stresses the need for better data selection processes for pretraining."
                },
                {
                    "title": "BLiMP: The Benchmark of Linguistic Minimal Pairs for English",
                    "abstract": "Abstract We introduce The Benchmark of Linguistic Minimal Pairs (BLiMP),1 a challenge set for evaluating the linguistic knowledge of language models (LMs) on major grammatical phenomena in English. BLiMP consists of 67 individual datasets, each containing 1,000 minimal pairs\u2014that is, pairs of minimally different sentences that contrast in grammatical acceptability and isolate specific phenomenon in syntax, morphology, or semantics. We generate the data according to linguist-crafted grammar templates, and human aggregate agreement with the labels is 96.4%. We evaluate n-gram, LSTM, and Transformer (GPT-2 and Transformer-XL) LMs by observing whether they assign a higher probability to the acceptable sentence in each minimal pair. We find that state-of-the-art models identify morphological contrasts related to agreement reliably, but they struggle with some subtle semantic and syntactic phenomena, such as negative polarity items and extraction islands."
                },
                {
                    "title": "Perturbation Sensitivity Analysis to Detect Unintended Model Biases",
                    "abstract": "Data-driven statistical Natural Language Processing (NLP) techniques leverage large amounts of language data to build models that can understand language. However, most language data reflect the public discourse at the time the data was produced, and hence NLP models are susceptible to learning incidental associations around named referents at a particular point in time, in addition to general linguistic meaning. An NLP system designed to model notions such as sentiment and toxicity should ideally produce scores that are independent of the identity of such entities mentioned in text and their social associations. For example, in a general purpose sentiment analysis system, a phrase such as I hate Katy Perry should be interpreted as having the same sentiment as I hate Taylor Swift. Based on this idea, we propose a generic evaluation framework, Perturbation Sensitivity Analysis, which detects unintended model biases related to named entities, and requires no new annotations or corpora. We demonstrate the utility of this analysis by employing it on two different NLP models \u2014 a sentiment model and a toxicity model \u2014 applied on online comments in English language from four different genres."
                },
                {
                    "title": "AllenNLP Interpret: A Framework for Explaining Predictions of NLP Models",
                    "abstract": "Neural NLP models are increasingly accurate but are imperfect and opaque\u2014they break in counterintuitive ways and leave end users puzzled at their behavior. Model interpretation methods ameliorate this opacity by providing explanations for specific model predictions. Unfortunately, existing interpretation codebases make it difficult to apply these methods to new models and tasks, which hinders adoption for practitioners and burdens interpretability researchers. We introduce AllenNLP Interpret, a flexible framework for interpreting NLP models. The toolkit provides interpretation primitives (e.g., input gradients) for any AllenNLP model and task, a suite of built-in interpretation methods, and a library of front-end visualization components. We demonstrate the toolkit\u2019s flexibility and utility by implementing live demos for five interpretation methods (e.g., saliency maps and adversarial attacks) on a variety of models and tasks (e.g., masked language modeling using BERT and reading comprehension using BiDAF). These demos, alongside our code and tutorials, are available at https://allennlp.org/interpret."
                },
                {
                    "title": "RoBERTa: A Robustly Optimized BERT Pretraining Approach",
                    "abstract": "Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code."
                },
                {
                    "title": "Unsupervised Text Style Transfer using Language Models as Discriminators",
                    "abstract": "Binary classifiers are often employed as discriminators in GAN-based unsupervised style transfer systems to ensure that transferred sentences are similar to sentences in the target domain. One difficulty with this approach is that the error signal provided by the discriminator can be unstable and is sometimes insufficient to train the generator to produce fluent language. In this paper, we propose a new technique that uses a target domain language model as the discriminator, providing richer and more stable token-level feedback during the learning process. We train the generator to minimize the negative log likelihood (NLL) of generated sentences, evaluated by the language model. By using a continuous approximation of discrete sampling under the generator, our model can be trained using back-propagation in an end- to-end fashion. Moreover, our empirical results show that when using a language model as a structured discriminator, it is possible to forgo adversarial steps during training, making the process more stable. We compare our model with previous work using convolutional neural networks (CNNs) as discriminators and show that our approach leads to improved performance on three tasks: word substitution decipherment, sentiment modification, and related language translation."
                },
                {
                    "title": "Pathologies of Neural Models Make Interpretations Difficult",
                    "abstract": "One way to interpret neural model predictions is to highlight the most important input features\u2014for example, a heatmap visualization over the words in an input sentence. In existing interpretation methods for NLP, a word\u2019s importance is determined by either input perturbation\u2014measuring the decrease in model confidence when that word is removed\u2014or by the gradient with respect to that word. To understand the limitations of these methods, we use input reduction, which iteratively removes the least important word from the input. This exposes pathological behaviors of neural models: the remaining words appear nonsensical to humans and are not the ones determined as important by interpretation methods. As we confirm with human experiments, the reduced examples lack information to support the prediction of any label, but models still make the same predictions with high confidence. To explain these counterintuitive results, we draw connections to adversarial examples and confidence calibration: pathological behaviors reveal difficulties in interpreting neural models trained with maximum likelihood. To mitigate their deficiencies, we fine-tune the models by encouraging high entropy outputs on reduced examples. Fine-tuned models become more interpretable under input reduction, without accuracy loss on regular examples."
                },
                {
                    "title": "A Unified Approach to Interpreting Model Predictions",
                    "abstract": "Understanding why a model makes a certain prediction can be as crucial as the prediction's accuracy in many applications. However, the highest accuracy for large modern datasets is often achieved by complex models that even experts struggle to interpret, such as ensemble or deep learning models, creating a tension between accuracy and interpretability. In response, various methods have recently been proposed to help users interpret the predictions of complex models, but it is often unclear how these methods are related and when one method is preferable over another. To address this problem, we present a unified framework for interpreting predictions, SHAP (SHapley Additive exPlanations). SHAP assigns each feature an importance value for a particular prediction. Its novel components include: (1) the identification of a new class of additive feature importance measures, and (2) theoretical results showing there is a unique solution in this class with a set of desirable properties. The new class unifies six existing methods, notable because several recent methods in the class lack the proposed desirable properties. Based on insights from this unification, we present new methods that show improved computational performance and/or better consistency with human intuition than previous approaches."
                },
                {
                    "title": "Learning Important Features Through Propagating Activation Differences",
                    "abstract": "The purported \"black box\" nature of neural networks is a barrier to adoption in applications where interpretability is essential. Here we present DeepLIFT (Deep Learning Important FeaTures), a method for decomposing the output prediction of a neural network on a specific input by backpropagating the contributions of all neurons in the network to every feature of the input. DeepLIFT compares the activation of each neuron to its 'reference activation' and assigns contribution scores according to the difference. By optionally giving separate consideration to positive and negative contributions, DeepLIFT can also reveal dependencies which are missed by other approaches. Scores can be computed efficiently in a single backward pass. We apply DeepLIFT to models trained on MNIST and simulated genomic data, and show significant advantages over gradient-based methods. Video tutorial: http://goo.gl/qKb7pL, code: http://goo.gl/RM8jvH."
                },
                {
                    "title": "Axiomatic Attribution for Deep Networks",
                    "abstract": "We study the problem of attributing the prediction of a deep network to its input features, a problem previously studied by several other works. We identify two fundamental axioms\u2014 Sensitivity and Implementation Invariance that attribution methods ought to satisfy. We show that they are not satisfied by most known attribution methods, which we consider to be a fundamental weakness of those methods. We use the axioms to guide the design of a new attribution method called Integrated Gradients. Our method requires no modification to the original network and is extremely simple to implement; it just needs a few calls to the standard gradient operator. We apply this method to a couple of image models, a couple of text models and a chemistry model, demonstrating its ability to debug networks, to extract rules from a network, and to enable users to engage with models better."
                },
                {
                    "title": "Understanding Neural Networks through Representation Erasure",
                    "abstract": "While neural networks have been successfully applied to many natural language processing tasks, they come at the cost of interpretability. In this paper, we propose a general methodology to analyze and interpret decisions from a neural model by observing the effects on the model of erasing various parts of the representation, such as input word-vector dimensions, intermediate hidden units, or input words. We present several approaches to analyzing the effects of such erasure, from computing the relative difference in evaluation metrics, to using reinforcement learning to erase the minimum set of input words in order to flip a neural model's decision. In a comprehensive analysis of multiple NLP tasks, including linguistic feature classification, sentence-level sentiment analysis, and document level sentiment aspect prediction, we show that the proposed methodology not only offers clear explanations about neural model decisions, but also provides a way to conduct error analysis on neural models."
                },
                {
                    "title": "Character-level Convolutional Networks for Text Classification",
                    "abstract": "This article offers an empirical exploration on the use of character-level convolutional networks (ConvNets) for text classification. We constructed several large-scale datasets to show that character-level convolutional networks could achieve state-of-the-art or competitive results. Comparisons are offered against traditional models such as bag of words, n-grams and their TFIDF variants, and deep learning models such as word-based ConvNets and recurrent neural networks."
                },
                {
                    "title": "On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation",
                    "abstract": "Understanding and interpreting classification decisions of automated image classification systems is of high value in many applications, as it allows to verify the reasoning of the system and provides additional information to the human expert. Although machine learning methods are solving very successfully a plethora of tasks, they have in most cases the disadvantage of acting as a black box, not providing any information about what made them arrive at a particular decision. This work proposes a general solution to the problem of understanding classification decisions by pixel-wise decomposition of nonlinear classifiers. We introduce a methodology that allows to visualize the contributions of single pixels to predictions for kernel-based classifiers over Bag of Words features and for multilayered neural networks. These pixel contributions can be visualized as heatmaps and are provided to a human expert who can intuitively not only verify the validity of the classification decision, but also focus further analysis on regions of potential interest. We evaluate our method for classifiers trained on PASCAL VOC 2009 images, synthetic image data containing geometric shapes, the MNIST handwritten digits data set and for the pre-trained ImageNet model available as part of the Caffe open source package."
                },
                {
                    "title": "Extraction of Salient Sentences from Labelled Documents",
                    "abstract": "We present a hierarchical convolutional document model with an architecture designed to support introspection of the document structure. Using this model, we show how to use visualisation techniques from the computer vision literature to identify and extract topic-relevant sentences. We also introduce a new scalable evaluation technique for automatic sentence extraction systems that avoids the need for time consuming human annotation of validation data."
                },
                {
                    "title": "Neural Machine Translation by Jointly Learning to Align and Translate",
                    "abstract": "Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition."
                },
                {
                    "title": "Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps",
                    "abstract": "This paper addresses the visualisation of image classification models, learnt using deep Convolutional Networks (ConvNets). We consider two visualisation techniques, based on computing the gradient of the class score with respect to the input image. The first one generates an image, which maximises the class score [Erhan et al., 2009], thus visualising the notion of the class, captured by a ConvNet. The second technique computes a class saliency map, specific to a given image and class. We show that such maps can be employed for weakly supervised object segmentation using classification ConvNets. Finally, we establish the connection between the gradient-based ConvNet visualisation methods and deconvolutional networks [Zeiler et al., 2013]."
                },
                {
                    "title": "Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank",
                    "abstract": "Semantic word spaces have been very useful but cannot express the meaning of longer phrases in a principled way. Further progress towards understanding compositionality in tasks such as sentiment detection requires richer supervised training and evaluation resources and more powerful models of composition. To remedy this, we introduce a Sentiment Treebank. It includes fine grained sentiment labels for 215,154 phrases in the parse trees of 11,855 sentences and presents new challenges for sentiment compositionality. To address them, we introduce the Recursive Neural Tensor Network. When trained on the new treebank, this model outperforms all previous methods on several metrics. It pushes the state of the art in single sentence positive/negative classification from 80% up to 85.4%. The accuracy of predicting fine-grained sentiment labels for all phrases reaches 80.7%, an improvement of 9.7% over bag of features baselines. Lastly, it is the only model that can accurately capture the effects of negation and its scope at various tree levels for both positive and negative phrases."
                },
                {
                    "title": "SenticNet: A Publicly Available Semantic Resource for Opinion Mining",
                    "abstract": "Today millions of web-users express their opinions about many topics through blogs, wikis, fora, chats and social networks. For sectors such as e-commerce and e-tourism, it is very useful to automatically analyze the huge amount of social information available on the Web, but the extremely unstructured nature of these contents makes it a difficult task. SenticNet is a publicly available resource for opinion mining built exploiting AI and Semantic Web techniques. It uses dimensionality reduction to infer the polarity of common sense concepts and hence provide a public resource for mining opinions from natural language text at a semantic, rather than just syntactic, level."
                },
                {
                    "title": "Contrastive Explanation",
                    "abstract": "According to a causal model of explanation, we explain phenomena by giving their causes or, where the phenomena are themselves causal regularities, we explain them by giving a mechanism linking cause and effect. If we explain why smoking causes cancer, we do not give the cause of this causal connection, but we do give the causal mechanism that makes it. The claim that to explain is to give a cause is not only natural and plausible, but it also avoids many of the objections to other accounts of explanation, such as the views that to explain is to give a reason to believe the phenomenon occurred, to somehow make the phenomenon familiar, or to give a Deductive-Nomological argument. Unlike the reason for belief account, a causal model makes a clear distinction between understanding why a phenomenon occurs and merely knowing that it does, and the model does so in a way that makes understanding unmysterious and objective. Understanding is not some sort of super-knowledge, but simply more knowledge: knowledge of the phenomenon and knowledge of its causal history. A causal model makes it clear how something can explain without itself being explained, and so avoids the regress of whys, since we can know a phenomenon's cause without knowing the cause of the cause. It also accounts for legitimate self-evidencing explanations, explanations where the phenomenon is an essential part of the evidence that the explanation is correct, so the explanation can not supply a non-circular reason for believing the phenomenon occurred. There is no barrier to knowing a cause through its effects and also knowing that it is their cause. The speed of recession of a star explains its observed red-shift, even though the shift is an essential part of the evidence for its speed of recession. The model also avoids the most serious objection to the familiarity view, which is that some phenomena are familiar yet not understood, since a phenomenon can be perfectly familiar, such as the blueness of the sky or the fact that the same side of the moon always faces the earth, even if we do not know its cause. Finally, a causal model avoids many of the objections to the Deductive-Nomological model. Ordinary explanations do not have to meet the requirements of the Deductive-Nomological model, because one does not need to give a law to give a cause, and one does not need to know a law to have good reason to believe that a cause is a cause. As for the notorious over-permissiveness of the Deductive-Nomological model, the reason recession explains red-shift but not conversely, is simply that causes explain effects but not conversely, and the reason a conjunction of laws does not explain its conjuncts is that conjunctions do not cause their conjuncts."
                },
                {
                    "title": "Tailor: A Soft-Prompt-Based Approach to Attribute-Based Controlled Text Generation",
                    "abstract": "Attribute-based Controlled Text Generation (CTG) refers to generating sentences that satisfy desirable attributes (e.g., emotions and topics). Existing work usually utilize fine-tuning or resort to extra attribute classifiers, yet suffer from increases in storage and inference time. To address these concerns, we explore attribute-based CTG in a parameter-efficient manner. In short, the proposed Tailor represents each attribute as a pre-trained continuous vector i.e., single-attribute prompt), which guides the generation of a fixed pre-trained language model (PLM) to satisfy a pre-specified attribute. These prompts can be simply concatenated as a whole for multi-attribute CTG without any re-training. Nevertheless, this may raise problems of fluency downgrading and position sensitivity. To solve this, Tailor provides two solutions to enhance the combination. The former contains a multi-attribute prompt mask and a re-indexing position sequence to bridge the gap between the training (one single-attribute prompt for each task) and the testing stage (concatenating two prompts). The latter introduces a trainable prompt connector to further enhance the combinations. Experiments demonstrate that, only requiring 0.08% extra training parameters of the GPT-2, Tailor can achieve effective and general improvements on eleven attribute-specific generation tasks."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively measure and interpret input saliency in large language models (LLMs) to enhance prompt-based generation control?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the interpretability of LLMs, which can lead to more effective prompting strategies and improved model outputs. By understanding input saliency, researchers can develop better tools for controlling LLM behavior, which has significant implications for various applications, including natural language processing, human-computer interaction, and AI ethics. This research could pave the way for future studies that explore the nuances of LLM responses, ultimately contributing to the development of more reliable and user-friendly AI systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem stem from the complexity of LLM architectures and their non-linear behaviors. Existing saliency methods are primarily designed for text classification tasks and rely on linearity assumptions that do not hold true for autoregressive models. This can lead to inaccuracies in measuring token significance. Additionally, the intricate relationships between input tokens and model predictions complicate the task of isolating and interpreting the effects of individual tokens on the generated output. Overcoming these technical and theoretical obstacles is essential for developing a robust saliency measurement approach.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has focused on saliency methods that are not well-suited for LLMs, often targeting masked language models and failing to account for the unique generative objectives of autoregressive models. Limitations in existing approaches, such as reliance on linear approximations and a narrow focus on class labels, have hindered progress in understanding token saliency in LLMs. Our approach differs by utilizing token distributions to estimate saliency, providing a more accurate and interpretable framework for analyzing LLM behavior, which has not been adequately addressed in prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nWe propose the Token Distribution Dynamics (TDD) approach, which includes three variants: TDD-forward, TDD-backward, and TDD-bidirectional. This methodology leverages token distributions to estimate saliency and provides contrastive explanations for token prioritization in LLM outputs. We will evaluate our approach using contemporary LLMs such as Pythia and LLaMA2, employing metrics that assess the accuracy of saliency interpretations against model predictions."
            }
        },
        "author_data": {
            "5ddbdb53-8f95-4d33-abf3-c29d73165742": {
                "pk": "5ddbdb53-8f95-4d33-abf3-c29d73165742",
                "name": "Zijian Feng",
                "collaborators": [
                    "Hanzhang Zhou",
                    "Zixiao Zhu",
                    "Kezhi Mao",
                    "Junlang Qian",
                    "Mao Kezhi",
                    "Hui Lu",
                    "Zhengming Ma",
                    "Zengrong Zhan",
                    "Jiajing Guo"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Few-shot Learning",
                    "Controllable Text Generation",
                    "Manifold Learning"
                ],
                "publications": [
                    {
                        "title": "EDEntail: An Entailment-based Few-shot Text Classification with Extensional Definition",
                        "abstract": "Few-shot text classification has seen significant advancements, particularly with entailment-based methods, which typically use either class labels or intensional definitions of class labels in hypotheses for label semantics expression. In this paper, we propose EDEntail, a method that employs extensional definition (EDef) of class labels in hypotheses, aiming to express the semantics of class labels more explicitly. To achieve the above goal, we develop an algo-rithm to gather and select extensional descriptive words of class labels and then order and format them into a sequence to form hypotheses. Our method has been evaluated and compared with state-of-the-art models on five classification datasets. The results demonstrate that our approach surpasses the supervised-learning methods and prompt-based methods under the few-shot setting, which underlines the potential of using an extensional definition of class labels for entailment-based few-shot text classification. Our code is available at https: //github.com/MidiyaZhu/EDEntail ."
                    },
                    {
                        "title": "Logit Separability-Driven Samples and Multiple Class-Related Words Selection for Advancing In-Context Learning",
                        "abstract": "Effective organization of in-context learning (ICL) demonstrations is key to improving the quality of large language model (LLM) responses. To create better sample-label pairs that instruct LLM understanding, we introduce logit separability, a criterion to assess the clarity of both samples and class-related words at the logit level. This facilitates the optimization of sample and label selection, enhancing the precision of information provided in ICL demonstrations. Additionally, we find that incorporating multiple class-related words for each sample, rather than relying on a single class name, improves performance by offering a broader range of label information. Building on these insights, we propose LICL, a logit separability-based method that jointly organizes samples and integrates multiple class-related words into each sample-label pair. Evaluations across seven classification datasets show that this approach significantly improves ICL performance by providing clearer instructions and richer label information."
                    },
                    {
                        "title": "FreeCtrl: Constructing Control Centers with Feedforward Layers for Learning-Free Controllable Text Generation",
                        "abstract": "Controllable text generation (CTG) seeks to craft texts adhering to specific attributes, traditionally employing learning-based techniques such as training, fine-tuning, or prefix-tuning with attribute-specific datasets. These approaches, while effective, demand extensive computational and data resources. In contrast, some proposed learning-free alternatives circumvent learning but often yield inferior results, exemplifying the fundamental machine learning trade-off between computational expense and model efficacy. To overcome these limitations, we propose FreeCtrl, a learning-free approach that dynamically adjusts the weights of selected feedforward neural network (FFN) vectors to steer the outputs of large language models (LLMs). FreeCtrl hinges on the principle that the weights of different FFN vectors influence the likelihood of different tokens appearing in the output. By identifying and adaptively adjusting the weights of attribute-related FFN vectors, FreeCtrl can control the output likelihood of attribute keywords in the generated content. Extensive experiments on single- and multi-attribute control reveal that the learning-free FreeCtrl outperforms other learning-free and learning-based methods, successfully resolving the dilemma between learning costs and model performance."
                    },
                    {
                        "title": "PromptExplainer: Explaining Language Models through Prompt-based Learning",
                        "abstract": "Pretrained language models have become workhorses for various natural language processing (NLP) tasks, sparking a growing demand for enhanced interpretability and transparency. However, prevailing explanation methods, such as attention-based and gradient-based strategies, largely rely on linear approximations, potentially causing inaccuracies such as accentuating irrelevant input tokens. To mitigate the issue, we develop PromptExplainer, a novel method for explaining language models through prompt-based learning. PromptExplainer aligns the explanation process with the masked language modeling (MLM) task of pretrained language models and leverages the prompt-based learning framework for explanation generation. It disentangles token representations into the explainable embedding space using the MLM head and extracts discriminative features with a verbalizer to generate class-dependent explanations. Extensive experiments demonstrate that PromptExplainer significantly outperforms state-of-the-art explanation methods."
                    },
                    {
                        "title": "UniBias: Unveiling and Mitigating LLM Bias through Internal Attention and FFN Manipulation",
                        "abstract": "Large language models (LLMs) have demonstrated impressive capabilities in various tasks using the in-context learning (ICL) paradigm. However, their effectiveness is often compromised by inherent bias, leading to prompt brittleness, i.e., sensitivity to design settings such as example selection, order, and prompt formatting. Previous studies have addressed LLM bias through external adjustment of model outputs, but the internal mechanisms that lead to such bias remain unexplored. Our work delves into these mechanisms, particularly investigating how feedforward neural networks (FFNs) and attention heads result in the bias of LLMs. By Interpreting the contribution of individual FFN vectors and attention heads, we identify the biased LLM components that skew LLMs' prediction toward specific labels. To mitigate these biases, we introduce UniBias, an inference-only method that effectively identifies and eliminates biased FFN vectors and attention heads. Extensive experiments across 12 NLP datasets demonstrate that UniBias significantly enhances ICL performance and alleviates prompt brittleness of LLMs."
                    },
                    {
                        "title": "LLMs Learn Task Heuristics from Demonstrations: A Heuristic-Driven Prompting Strategy for Document-Level Event Argument Extraction",
                        "abstract": "In this study, we investigate in-context learning (ICL) in document-level event argument extraction (EAE) to alleviate the dependency on large-scale labeled data for this task. We introduce the Heuristic-Driven Link-of-Analogy (HD-LoA) prompting to address the challenge of example selection and to develop a prompting strategy tailored for EAE. Specifically, we hypothesize and validate that LLMs learn task-specific heuristics from demonstrations via ICL. Building upon this hypothesis, we introduce an explicit heuristic-driven demonstration construction approach, which transforms the haphazard example selection process into a methodical method that emphasizes task heuristics. Additionally, inspired by the analogical reasoning of human, we propose the link-of-analogy prompting, which enables LLMs to process new situations by drawing analogies to known situations, enhancing their performance on unseen classes beyond limited ICL examples. Experiments show that our method outperforms existing prompting methods and few-shot supervised learning methods on document-level EAE datasets. Additionally, the HD-LoA prompting shows effectiveness in diverse tasks like sentiment analysis and natural language inference, demonstrating its broad adaptability."
                    },
                    {
                        "title": "Closed Boundary Learning for Classification Tasks with the Universum Class",
                        "abstract": ","
                    },
                    {
                        "title": "Manifold Learning Based on Straight-Like Geodesics and Local Coordinates",
                        "abstract": "In this article, a manifold learning algorithm based on straight-like geodesics and local coordinates is proposed, called SGLC-ML for short. The contribution and innovation of SGLC-ML lie in that; first, SGLC-ML divides the manifold data into a number of straight-like geodesics, instead of a number of local areas like many manifold learning algorithms do. Figuratively speaking, SGLC-ML covers manifold data set with a sparse net woven with threads (straight-like geodesics), while other manifold learning algorithms with a tight roof made of titles (local areas). Second, SGLC-ML maps all straight-like geodesics into straight lines of a low-dimensional Euclidean space. All these straight lines start from the same point and extend along the same coordinate axis. These straight lines are exactly the local coordinates of straight-like geodesics as described in the mathematical definition of the manifold. With the help of local coordinates, dimensionality reduction can be divided into two relatively simple processes: calculation and alignment of local coordinates. However, many manifold learning algorithms seem to ignore the advantages of local coordinates. The experimental results between SGLC-ML and other state-of-the-art algorithms are presented to verify the good performance of SGLC-ML."
                    }
                ]
            },
            "e5b6f947-c8d0-4870-826a-5b1486bd723b": {
                "pk": "e5b6f947-c8d0-4870-826a-5b1486bd723b",
                "name": "Hanzhang Zhou",
                "collaborators": [
                    "Zijian Feng",
                    "Kezhi Mao",
                    "Zixiao Zhu",
                    "Junlang Qian",
                    "Mao Kezhi",
                    "Hui Lu",
                    "Zeju Li",
                    "Jinhua Yu",
                    "Yuanyuan Wang",
                    "Haowei Yang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Few-shot Learning",
                    "Machine Learning",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "EDEntail: An Entailment-based Few-shot Text Classification with Extensional Definition",
                        "abstract": "Few-shot text classification has seen significant advancements, particularly with entailment-based methods, which typically use either class labels or intensional definitions of class labels in hypotheses for label semantics expression. In this paper, we propose EDEntail, a method that employs extensional definition (EDef) of class labels in hypotheses, aiming to express the semantics of class labels more explicitly. To achieve the above goal, we develop an algo-rithm to gather and select extensional descriptive words of class labels and then order and format them into a sequence to form hypotheses. Our method has been evaluated and compared with state-of-the-art models on five classification datasets. The results demonstrate that our approach surpasses the supervised-learning methods and prompt-based methods under the few-shot setting, which underlines the potential of using an extensional definition of class labels for entailment-based few-shot text classification. Our code is available at https: //github.com/MidiyaZhu/EDEntail ."
                    },
                    {
                        "title": "Logit Separability-Driven Samples and Multiple Class-Related Words Selection for Advancing In-Context Learning",
                        "abstract": "Effective organization of in-context learning (ICL) demonstrations is key to improving the quality of large language model (LLM) responses. To create better sample-label pairs that instruct LLM understanding, we introduce logit separability, a criterion to assess the clarity of both samples and class-related words at the logit level. This facilitates the optimization of sample and label selection, enhancing the precision of information provided in ICL demonstrations. Additionally, we find that incorporating multiple class-related words for each sample, rather than relying on a single class name, improves performance by offering a broader range of label information. Building on these insights, we propose LICL, a logit separability-based method that jointly organizes samples and integrates multiple class-related words into each sample-label pair. Evaluations across seven classification datasets show that this approach significantly improves ICL performance by providing clearer instructions and richer label information."
                    },
                    {
                        "title": "FreeCtrl: Constructing Control Centers with Feedforward Layers for Learning-Free Controllable Text Generation",
                        "abstract": "Controllable text generation (CTG) seeks to craft texts adhering to specific attributes, traditionally employing learning-based techniques such as training, fine-tuning, or prefix-tuning with attribute-specific datasets. These approaches, while effective, demand extensive computational and data resources. In contrast, some proposed learning-free alternatives circumvent learning but often yield inferior results, exemplifying the fundamental machine learning trade-off between computational expense and model efficacy. To overcome these limitations, we propose FreeCtrl, a learning-free approach that dynamically adjusts the weights of selected feedforward neural network (FFN) vectors to steer the outputs of large language models (LLMs). FreeCtrl hinges on the principle that the weights of different FFN vectors influence the likelihood of different tokens appearing in the output. By identifying and adaptively adjusting the weights of attribute-related FFN vectors, FreeCtrl can control the output likelihood of attribute keywords in the generated content. Extensive experiments on single- and multi-attribute control reveal that the learning-free FreeCtrl outperforms other learning-free and learning-based methods, successfully resolving the dilemma between learning costs and model performance."
                    },
                    {
                        "title": "PromptExplainer: Explaining Language Models through Prompt-based Learning",
                        "abstract": "Pretrained language models have become workhorses for various natural language processing (NLP) tasks, sparking a growing demand for enhanced interpretability and transparency. However, prevailing explanation methods, such as attention-based and gradient-based strategies, largely rely on linear approximations, potentially causing inaccuracies such as accentuating irrelevant input tokens. To mitigate the issue, we develop PromptExplainer, a novel method for explaining language models through prompt-based learning. PromptExplainer aligns the explanation process with the masked language modeling (MLM) task of pretrained language models and leverages the prompt-based learning framework for explanation generation. It disentangles token representations into the explainable embedding space using the MLM head and extracts discriminative features with a verbalizer to generate class-dependent explanations. Extensive experiments demonstrate that PromptExplainer significantly outperforms state-of-the-art explanation methods."
                    },
                    {
                        "title": "UniBias: Unveiling and Mitigating LLM Bias through Internal Attention and FFN Manipulation",
                        "abstract": "Large language models (LLMs) have demonstrated impressive capabilities in various tasks using the in-context learning (ICL) paradigm. However, their effectiveness is often compromised by inherent bias, leading to prompt brittleness, i.e., sensitivity to design settings such as example selection, order, and prompt formatting. Previous studies have addressed LLM bias through external adjustment of model outputs, but the internal mechanisms that lead to such bias remain unexplored. Our work delves into these mechanisms, particularly investigating how feedforward neural networks (FFNs) and attention heads result in the bias of LLMs. By Interpreting the contribution of individual FFN vectors and attention heads, we identify the biased LLM components that skew LLMs' prediction toward specific labels. To mitigate these biases, we introduce UniBias, an inference-only method that effectively identifies and eliminates biased FFN vectors and attention heads. Extensive experiments across 12 NLP datasets demonstrate that UniBias significantly enhances ICL performance and alleviates prompt brittleness of LLMs."
                    },
                    {
                        "title": "LLMs Learn Task Heuristics from Demonstrations: A Heuristic-Driven Prompting Strategy for Document-Level Event Argument Extraction",
                        "abstract": "In this study, we investigate in-context learning (ICL) in document-level event argument extraction (EAE) to alleviate the dependency on large-scale labeled data for this task. We introduce the Heuristic-Driven Link-of-Analogy (HD-LoA) prompting to address the challenge of example selection and to develop a prompting strategy tailored for EAE. Specifically, we hypothesize and validate that LLMs learn task-specific heuristics from demonstrations via ICL. Building upon this hypothesis, we introduce an explicit heuristic-driven demonstration construction approach, which transforms the haphazard example selection process into a methodical method that emphasizes task heuristics. Additionally, inspired by the analogical reasoning of human, we propose the link-of-analogy prompting, which enables LLMs to process new situations by drawing analogies to known situations, enhancing their performance on unseen classes beyond limited ICL examples. Experiments show that our method outperforms existing prompting methods and few-shot supervised learning methods on document-level EAE datasets. Additionally, the HD-LoA prompting shows effectiveness in diverse tasks like sentiment analysis and natural language inference, demonstrating its broad adaptability."
                    },
                    {
                        "title": "Closed Boundary Learning for Classification Tasks with the Universum Class",
                        "abstract": ","
                    },
                    {
                        "title": "Document-Level Event Argument Extraction by Leveraging Redundant Information and Closed Boundary Loss",
                        "abstract": "In document-level event argument extraction, an argument is likely to appear multiple times in different expressions in the document. The redundancy of arguments underlying multiple sentences is beneficial but is often overlooked. In addition, in event argument extraction, most entities are regarded as class \u201cothers\u201d, i.e. Universum class, which is defined as a collection of samples that do not belong to any class of interest. Universum class is composed of heterogeneous entities without typical common features. Classifiers trained by cross entropy loss could easily misclassify the Universum class because of their open decision boundary. In this paper, to make use of redundant event information underlying a document, we build an entity coreference graph with the graph2token module to produce a comprehensive and coreference-aware representation for every entity and then build an entity summary graph to merge the multiple extraction results. To better classify Universum class, we propose a new loss function to build classifiers with closed boundaries. Experimental results show that our model outperforms the previous state-of-the-art models by 3.35% in F1-score."
                    },
                    {
                        "title": "DeepVolume: Brain Structure and Spatial Connection-Aware Network for Brain MRI Super-Resolution",
                        "abstract": "Thin-section magnetic resonance imaging (MRI) can provide higher resolution anatomical structures and more precise clinical information than thick-section images. However, thin-section MRI is not always available due to the imaging cost issue. In multicenter retrospective studies, a large number of data are often in thick-section manner with different section thickness. The lack of thin-section data and the difference in section thickness bring considerable difficulties in the study based on the image big data. In this article, we introduce DeepVolume, a two-step deep learning architecture to address the challenge of accurate thin-section MR image reconstruction. The first stage is the brain structure-aware network, in which the thick-section MR images in axial and sagittal planes are fused by a multitask 3-D U-net with prior knowledge of brain volume segmentation, which encourages the reconstruction result to have correct brain structure. The second stage is the spatial connection-aware network, in which the preliminary reconstruction results are adjusted slice-by-slice by a recurrent convolutional network embedding convolutional long short-term memory (LSTM) block, which enhances the precision of the reconstruction by utilizing the previously unassessed sagittal information. We used 305 paired brain MRI samples with thickness of 1.0 mm and 6.5 mm in this article. Extensive experiments illustrate that DeepVolume can produce the state-of-the-art reconstruction results by embedding more anatomical knowledge. Furthermore, considering DeepVolume as an intermediate step, the practical and clinical value of our method is validated by applying the brain volume estimation and voxel-based morphometry. The results show that DeepVolume can provide much more reliable brain volume estimation in the normalized space based on the thick-section MR images compared with the traditional solutions."
                    }
                ]
            },
            "eef75229-fdd6-4258-a393-f633f8870f71": {
                "pk": "eef75229-fdd6-4258-a393-f633f8870f71",
                "name": "Zixiao Zhu",
                "collaborators": [
                    "Zijian Feng",
                    "Hanzhang Zhou",
                    "Kezhi Mao",
                    "Junlang Qian",
                    "Mao Kezhi",
                    "Hui Lu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Few-shot Learning",
                    "Controllable Text Generation",
                    "Interpretability"
                ],
                "publications": [
                    {
                        "title": "EDEntail: An Entailment-based Few-shot Text Classification with Extensional Definition",
                        "abstract": "Few-shot text classification has seen significant advancements, particularly with entailment-based methods, which typically use either class labels or intensional definitions of class labels in hypotheses for label semantics expression. In this paper, we propose EDEntail, a method that employs extensional definition (EDef) of class labels in hypotheses, aiming to express the semantics of class labels more explicitly. To achieve the above goal, we develop an algo-rithm to gather and select extensional descriptive words of class labels and then order and format them into a sequence to form hypotheses. Our method has been evaluated and compared with state-of-the-art models on five classification datasets. The results demonstrate that our approach surpasses the supervised-learning methods and prompt-based methods under the few-shot setting, which underlines the potential of using an extensional definition of class labels for entailment-based few-shot text classification. Our code is available at https: //github.com/MidiyaZhu/EDEntail ."
                    },
                    {
                        "title": "Logit Separability-Driven Samples and Multiple Class-Related Words Selection for Advancing In-Context Learning",
                        "abstract": "Effective organization of in-context learning (ICL) demonstrations is key to improving the quality of large language model (LLM) responses. To create better sample-label pairs that instruct LLM understanding, we introduce logit separability, a criterion to assess the clarity of both samples and class-related words at the logit level. This facilitates the optimization of sample and label selection, enhancing the precision of information provided in ICL demonstrations. Additionally, we find that incorporating multiple class-related words for each sample, rather than relying on a single class name, improves performance by offering a broader range of label information. Building on these insights, we propose LICL, a logit separability-based method that jointly organizes samples and integrates multiple class-related words into each sample-label pair. Evaluations across seven classification datasets show that this approach significantly improves ICL performance by providing clearer instructions and richer label information."
                    },
                    {
                        "title": "FreeCtrl: Constructing Control Centers with Feedforward Layers for Learning-Free Controllable Text Generation",
                        "abstract": "Controllable text generation (CTG) seeks to craft texts adhering to specific attributes, traditionally employing learning-based techniques such as training, fine-tuning, or prefix-tuning with attribute-specific datasets. These approaches, while effective, demand extensive computational and data resources. In contrast, some proposed learning-free alternatives circumvent learning but often yield inferior results, exemplifying the fundamental machine learning trade-off between computational expense and model efficacy. To overcome these limitations, we propose FreeCtrl, a learning-free approach that dynamically adjusts the weights of selected feedforward neural network (FFN) vectors to steer the outputs of large language models (LLMs). FreeCtrl hinges on the principle that the weights of different FFN vectors influence the likelihood of different tokens appearing in the output. By identifying and adaptively adjusting the weights of attribute-related FFN vectors, FreeCtrl can control the output likelihood of attribute keywords in the generated content. Extensive experiments on single- and multi-attribute control reveal that the learning-free FreeCtrl outperforms other learning-free and learning-based methods, successfully resolving the dilemma between learning costs and model performance."
                    },
                    {
                        "title": "PromptExplainer: Explaining Language Models through Prompt-based Learning",
                        "abstract": "Pretrained language models have become workhorses for various natural language processing (NLP) tasks, sparking a growing demand for enhanced interpretability and transparency. However, prevailing explanation methods, such as attention-based and gradient-based strategies, largely rely on linear approximations, potentially causing inaccuracies such as accentuating irrelevant input tokens. To mitigate the issue, we develop PromptExplainer, a novel method for explaining language models through prompt-based learning. PromptExplainer aligns the explanation process with the masked language modeling (MLM) task of pretrained language models and leverages the prompt-based learning framework for explanation generation. It disentangles token representations into the explainable embedding space using the MLM head and extracts discriminative features with a verbalizer to generate class-dependent explanations. Extensive experiments demonstrate that PromptExplainer significantly outperforms state-of-the-art explanation methods."
                    },
                    {
                        "title": "UniBias: Unveiling and Mitigating LLM Bias through Internal Attention and FFN Manipulation",
                        "abstract": "Large language models (LLMs) have demonstrated impressive capabilities in various tasks using the in-context learning (ICL) paradigm. However, their effectiveness is often compromised by inherent bias, leading to prompt brittleness, i.e., sensitivity to design settings such as example selection, order, and prompt formatting. Previous studies have addressed LLM bias through external adjustment of model outputs, but the internal mechanisms that lead to such bias remain unexplored. Our work delves into these mechanisms, particularly investigating how feedforward neural networks (FFNs) and attention heads result in the bias of LLMs. By Interpreting the contribution of individual FFN vectors and attention heads, we identify the biased LLM components that skew LLMs' prediction toward specific labels. To mitigate these biases, we introduce UniBias, an inference-only method that effectively identifies and eliminates biased FFN vectors and attention heads. Extensive experiments across 12 NLP datasets demonstrate that UniBias significantly enhances ICL performance and alleviates prompt brittleness of LLMs."
                    },
                    {
                        "title": "LLMs Learn Task Heuristics from Demonstrations: A Heuristic-Driven Prompting Strategy for Document-Level Event Argument Extraction",
                        "abstract": "In this study, we investigate in-context learning (ICL) in document-level event argument extraction (EAE) to alleviate the dependency on large-scale labeled data for this task. We introduce the Heuristic-Driven Link-of-Analogy (HD-LoA) prompting to address the challenge of example selection and to develop a prompting strategy tailored for EAE. Specifically, we hypothesize and validate that LLMs learn task-specific heuristics from demonstrations via ICL. Building upon this hypothesis, we introduce an explicit heuristic-driven demonstration construction approach, which transforms the haphazard example selection process into a methodical method that emphasizes task heuristics. Additionally, inspired by the analogical reasoning of human, we propose the link-of-analogy prompting, which enables LLMs to process new situations by drawing analogies to known situations, enhancing their performance on unseen classes beyond limited ICL examples. Experiments show that our method outperforms existing prompting methods and few-shot supervised learning methods on document-level EAE datasets. Additionally, the HD-LoA prompting shows effectiveness in diverse tasks like sentiment analysis and natural language inference, demonstrating its broad adaptability."
                    }
                ]
            },
            "baf11a6e-967c-43d3-af0f-bdae846411e4": {
                "pk": "baf11a6e-967c-43d3-af0f-bdae846411e4",
                "name": "Junlang Qian",
                "collaborators": [
                    "Zixiao Zhu",
                    "Zijian Feng",
                    "Hanzhang Zhou",
                    "Kezhi Mao",
                    "Mao Kezhi",
                    "Hui Lu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Few-shot Learning",
                    "In-context Learning",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "EDEntail: An Entailment-based Few-shot Text Classification with Extensional Definition",
                        "abstract": "Few-shot text classification has seen significant advancements, particularly with entailment-based methods, which typically use either class labels or intensional definitions of class labels in hypotheses for label semantics expression. In this paper, we propose EDEntail, a method that employs extensional definition (EDef) of class labels in hypotheses, aiming to express the semantics of class labels more explicitly. To achieve the above goal, we develop an algo-rithm to gather and select extensional descriptive words of class labels and then order and format them into a sequence to form hypotheses. Our method has been evaluated and compared with state-of-the-art models on five classification datasets. The results demonstrate that our approach surpasses the supervised-learning methods and prompt-based methods under the few-shot setting, which underlines the potential of using an extensional definition of class labels for entailment-based few-shot text classification. Our code is available at https: //github.com/MidiyaZhu/EDEntail ."
                    },
                    {
                        "title": "Logit Separability-Driven Samples and Multiple Class-Related Words Selection for Advancing In-Context Learning",
                        "abstract": "Effective organization of in-context learning (ICL) demonstrations is key to improving the quality of large language model (LLM) responses. To create better sample-label pairs that instruct LLM understanding, we introduce logit separability, a criterion to assess the clarity of both samples and class-related words at the logit level. This facilitates the optimization of sample and label selection, enhancing the precision of information provided in ICL demonstrations. Additionally, we find that incorporating multiple class-related words for each sample, rather than relying on a single class name, improves performance by offering a broader range of label information. Building on these insights, we propose LICL, a logit separability-based method that jointly organizes samples and integrates multiple class-related words into each sample-label pair. Evaluations across seven classification datasets show that this approach significantly improves ICL performance by providing clearer instructions and richer label information."
                    },
                    {
                        "title": "UniBias: Unveiling and Mitigating LLM Bias through Internal Attention and FFN Manipulation",
                        "abstract": "Large language models (LLMs) have demonstrated impressive capabilities in various tasks using the in-context learning (ICL) paradigm. However, their effectiveness is often compromised by inherent bias, leading to prompt brittleness, i.e., sensitivity to design settings such as example selection, order, and prompt formatting. Previous studies have addressed LLM bias through external adjustment of model outputs, but the internal mechanisms that lead to such bias remain unexplored. Our work delves into these mechanisms, particularly investigating how feedforward neural networks (FFNs) and attention heads result in the bias of LLMs. By Interpreting the contribution of individual FFN vectors and attention heads, we identify the biased LLM components that skew LLMs' prediction toward specific labels. To mitigate these biases, we introduce UniBias, an inference-only method that effectively identifies and eliminates biased FFN vectors and attention heads. Extensive experiments across 12 NLP datasets demonstrate that UniBias significantly enhances ICL performance and alleviates prompt brittleness of LLMs."
                    },
                    {
                        "title": "LLMs Learn Task Heuristics from Demonstrations: A Heuristic-Driven Prompting Strategy for Document-Level Event Argument Extraction",
                        "abstract": "In this study, we investigate in-context learning (ICL) in document-level event argument extraction (EAE) to alleviate the dependency on large-scale labeled data for this task. We introduce the Heuristic-Driven Link-of-Analogy (HD-LoA) prompting to address the challenge of example selection and to develop a prompting strategy tailored for EAE. Specifically, we hypothesize and validate that LLMs learn task-specific heuristics from demonstrations via ICL. Building upon this hypothesis, we introduce an explicit heuristic-driven demonstration construction approach, which transforms the haphazard example selection process into a methodical method that emphasizes task heuristics. Additionally, inspired by the analogical reasoning of human, we propose the link-of-analogy prompting, which enables LLMs to process new situations by drawing analogies to known situations, enhancing their performance on unseen classes beyond limited ICL examples. Experiments show that our method outperforms existing prompting methods and few-shot supervised learning methods on document-level EAE datasets. Additionally, the HD-LoA prompting shows effectiveness in diverse tasks like sentiment analysis and natural language inference, demonstrating its broad adaptability."
                    }
                ]
            },
            "edefd064-5164-4d18-b723-e286f8255381": {
                "pk": "edefd064-5164-4d18-b723-e286f8255381",
                "name": "Kezhi Mao",
                "collaborators": [
                    "Hanzhang Zhou",
                    "Zijian Feng",
                    "Zixiao Zhu",
                    "Junlang Qian",
                    "Hui Lu",
                    "Erik Cambria",
                    "G. Huang",
                    "L. L. C. Kasun",
                    "Hongming Zhou",
                    "C. Vong"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Few-shot Learning",
                    "Controllable Text Generation",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "EDEntail: An Entailment-based Few-shot Text Classification with Extensional Definition",
                        "abstract": "Few-shot text classification has seen significant advancements, particularly with entailment-based methods, which typically use either class labels or intensional definitions of class labels in hypotheses for label semantics expression. In this paper, we propose EDEntail, a method that employs extensional definition (EDef) of class labels in hypotheses, aiming to express the semantics of class labels more explicitly. To achieve the above goal, we develop an algo-rithm to gather and select extensional descriptive words of class labels and then order and format them into a sequence to form hypotheses. Our method has been evaluated and compared with state-of-the-art models on five classification datasets. The results demonstrate that our approach surpasses the supervised-learning methods and prompt-based methods under the few-shot setting, which underlines the potential of using an extensional definition of class labels for entailment-based few-shot text classification. Our code is available at https: //github.com/MidiyaZhu/EDEntail ."
                    },
                    {
                        "title": "FreeCtrl: Constructing Control Centers with Feedforward Layers for Learning-Free Controllable Text Generation",
                        "abstract": "Controllable text generation (CTG) seeks to craft texts adhering to specific attributes, traditionally employing learning-based techniques such as training, fine-tuning, or prefix-tuning with attribute-specific datasets. These approaches, while effective, demand extensive computational and data resources. In contrast, some proposed learning-free alternatives circumvent learning but often yield inferior results, exemplifying the fundamental machine learning trade-off between computational expense and model efficacy. To overcome these limitations, we propose FreeCtrl, a learning-free approach that dynamically adjusts the weights of selected feedforward neural network (FFN) vectors to steer the outputs of large language models (LLMs). FreeCtrl hinges on the principle that the weights of different FFN vectors influence the likelihood of different tokens appearing in the output. By identifying and adaptively adjusting the weights of attribute-related FFN vectors, FreeCtrl can control the output likelihood of attribute keywords in the generated content. Extensive experiments on single- and multi-attribute control reveal that the learning-free FreeCtrl outperforms other learning-free and learning-based methods, successfully resolving the dilemma between learning costs and model performance."
                    },
                    {
                        "title": "PromptExplainer: Explaining Language Models through Prompt-based Learning",
                        "abstract": "Pretrained language models have become workhorses for various natural language processing (NLP) tasks, sparking a growing demand for enhanced interpretability and transparency. However, prevailing explanation methods, such as attention-based and gradient-based strategies, largely rely on linear approximations, potentially causing inaccuracies such as accentuating irrelevant input tokens. To mitigate the issue, we develop PromptExplainer, a novel method for explaining language models through prompt-based learning. PromptExplainer aligns the explanation process with the masked language modeling (MLM) task of pretrained language models and leverages the prompt-based learning framework for explanation generation. It disentangles token representations into the explainable embedding space using the MLM head and extracts discriminative features with a verbalizer to generate class-dependent explanations. Extensive experiments demonstrate that PromptExplainer significantly outperforms state-of-the-art explanation methods."
                    },
                    {
                        "title": "UniBias: Unveiling and Mitigating LLM Bias through Internal Attention and FFN Manipulation",
                        "abstract": "Large language models (LLMs) have demonstrated impressive capabilities in various tasks using the in-context learning (ICL) paradigm. However, their effectiveness is often compromised by inherent bias, leading to prompt brittleness, i.e., sensitivity to design settings such as example selection, order, and prompt formatting. Previous studies have addressed LLM bias through external adjustment of model outputs, but the internal mechanisms that lead to such bias remain unexplored. Our work delves into these mechanisms, particularly investigating how feedforward neural networks (FFNs) and attention heads result in the bias of LLMs. By Interpreting the contribution of individual FFN vectors and attention heads, we identify the biased LLM components that skew LLMs' prediction toward specific labels. To mitigate these biases, we introduce UniBias, an inference-only method that effectively identifies and eliminates biased FFN vectors and attention heads. Extensive experiments across 12 NLP datasets demonstrate that UniBias significantly enhances ICL performance and alleviates prompt brittleness of LLMs."
                    },
                    {
                        "title": "LLMs Learn Task Heuristics from Demonstrations: A Heuristic-Driven Prompting Strategy for Document-Level Event Argument Extraction",
                        "abstract": "In this study, we investigate in-context learning (ICL) in document-level event argument extraction (EAE) to alleviate the dependency on large-scale labeled data for this task. We introduce the Heuristic-Driven Link-of-Analogy (HD-LoA) prompting to address the challenge of example selection and to develop a prompting strategy tailored for EAE. Specifically, we hypothesize and validate that LLMs learn task-specific heuristics from demonstrations via ICL. Building upon this hypothesis, we introduce an explicit heuristic-driven demonstration construction approach, which transforms the haphazard example selection process into a methodical method that emphasizes task heuristics. Additionally, inspired by the analogical reasoning of human, we propose the link-of-analogy prompting, which enables LLMs to process new situations by drawing analogies to known situations, enhancing their performance on unseen classes beyond limited ICL examples. Experiments show that our method outperforms existing prompting methods and few-shot supervised learning methods on document-level EAE datasets. Additionally, the HD-LoA prompting shows effectiveness in diverse tasks like sentiment analysis and natural language inference, demonstrating its broad adaptability."
                    },
                    {
                        "title": "Closed Boundary Learning for Classification Tasks with the Universum Class",
                        "abstract": ","
                    },
                    {
                        "title": "Document-Level Event Argument Extraction by Leveraging Redundant Information and Closed Boundary Loss",
                        "abstract": "In document-level event argument extraction, an argument is likely to appear multiple times in different expressions in the document. The redundancy of arguments underlying multiple sentences is beneficial but is often overlooked. In addition, in event argument extraction, most entities are regarded as class \u201cothers\u201d, i.e. Universum class, which is defined as a collection of samples that do not belong to any class of interest. Universum class is composed of heterogeneous entities without typical common features. Classifiers trained by cross entropy loss could easily misclassify the Universum class because of their open decision boundary. In this paper, to make use of redundant event information underlying a document, we build an entity coreference graph with the graph2token module to produce a comprehensive and coreference-aware representation for every entity and then build an entity summary graph to merge the multiple extraction results. To better classify Universum class, we propose a new loss function to build classifiers with closed boundaries. Experimental results show that our model outperforms the previous state-of-the-art models by 3.35% in F1-score."
                    },
                    {
                        "title": "Extreme Learning Machines",
                        "abstract": "The Random Neural Network (RNN) is a spiking model for neuronal ensembles that has been used for both modeling and understanding the behaviour of natural systems such as the somatosensory system of rats, and learning based engineering applications in computer vision, image compression, packet network routing and Software Defined Networks, and for combinatorial optimisation. It is a recurrent network with a highly efficient and fast exact solution based on the product form, and has a fast O(n3) recurrent network learning algorithm. It has also been used for Reinforcement Learning in the design of the patented Cognitive Packet Network routing system. This talk will discuss our recent work in using the RNN for Deep and Extreme Learning. In particular we will show how dense clusters of soma to soma interactions can be modeled with the RNN and demonstrate the results obtained with such multiple layer deep networks that are applied to a variety of instances of classification problems, as well as to the dynamic control of task assignment in the Cloud."
                    }
                ]
            }
        }
    },
    "2310.02391": {
        "paper_data": {
            "title": "SE(3)-Stochastic Flow Matching for Protein Backbone Generation",
            "url": "http://arxiv.org/abs/2310.02391v4",
            "arxiv_id": "2310.02391",
            "authors": [
                "Avishek Joey Bose",
                "Tara Akhound-Sadegh",
                "Guillaume Huguet",
                "Kilian Fatras",
                "Jarrid Rector-Brooks",
                "Cheng-Hao Liu",
                "Andrei Cristian Nica",
                "Maksym Korablyov",
                "Michael Bronstein",
                "Alexander Tong"
            ],
            "abstract": "The computational design of novel protein structures has the potential to impact numerous scientific disciplines greatly. Toward this goal, we introduce FoldFlow, a series of novel generative models of increasing modeling power based on the flow-matching paradigm over $3\\mathrm{D}$ rigid motions -- i.e. the group $\\text{SE}(3)$ -- enabling accurate modeling of protein backbones. We first introduce FoldFlow-Base, a simulation-free approach to learning deterministic continuous-time dynamics and matching invariant target distributions on $\\text{SE}(3)$. We next accelerate training by incorporating Riemannian optimal transport to create FoldFlow-OT, leading to the construction of both more simple and stable flows. Finally, we design FoldFlow-SFM, coupling both Riemannian OT and simulation-free training to learn stochastic continuous-time dynamics over $\\text{SE}(3)$. Our family of FoldFlow, generative models offers several key advantages over previous approaches to the generative modeling of proteins: they are more stable and faster to train than diffusion-based approaches, and our models enjoy the ability to map any invariant source distribution to any invariant target distribution over $\\text{SE}(3)$. Empirically, we validate FoldFlow, on protein backbone generation of up to $300$ amino acids leading to high-quality designable, diverse, and novel samples.",
            "introduction": "   1 Introduction  Proteins are one of the basic building blocks of life. Their complex geometric structure enables specific inter-molecular interactions that allow for crucial functions within organisms, such as acting as catalysts in chemical reactions, transporters for molecules, and providing immune responses. Normally, such functions arise as a result of evolution. With the emergence of computational techniques, it has become possible to rationally design novel proteins with desired structures that program their functions. Such methods are now seen as the future of drug design and can lead to solutions to long-standing global health challenges. Some recent examples include rationally designed protein binders for receptors related to influenza\u00a0(Strauch et\u00a0al., 2017), COVID-19\u00a0(Cao et\u00a0al., 2020a; Gainza et\u00a0al., 2023), and cancer\u00a0(Silva et\u00a0al., 2019).              Method  Any source   OT   Stochastic   Score free      RFDiffusion \u2717 \u2717 \u2713 \u2717   FrameDiff \u2717 \u2717 \u2713 \u2717   FoldFlow-Base \u2713 \u2717 \u2717 \u2713   FoldFlow-OT \u2713 \u2713 \u2717 \u2713   FoldFlow-SFM \u2713 \u2713*{}^{*}start_FLOATSUPERSCRIPT * end_FLOATSUPERSCRIPT \u2713 \u2713       Figure 1: Left: Conditional probability paths learned by FoldFlow-Base (left), FoldFlow-OT (mid), and FoldFlow-SFM (right). We visualize the rotation trajectory of a single residue by the action of SO\u2062(3)SO3\\mathrm{SO(3)}roman_SO ( 3 ) on its homogenous space \ud835\udd4a2superscript\ud835\udd4a2\\mathbb{S}^{2}blackboard_S start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT. Right: Table with the properties of each model: whether they can map from a general source distribution, perform optimal transport, are stochastic, or require the score of the \u2110\u2062\ud835\udca2SO\u2062(3)\u2110subscript\ud835\udca2SO3\\mathcal{IG}_{\\mathrm{SO(3)}}caligraphic_I caligraphic_G start_POSTSUBSCRIPT roman_SO ( 3 ) end_POSTSUBSCRIPT density.   In protein engineering, the term de novo design refers to a setting when a new protein is designed to satisfy pre-specified structural and functional properties\u00a0(Huang et\u00a0al., 2016). Chemically, a protein is a sequence of amino acids (residues) linked into a chain that folds into a complex 3D structure under the influence of electrostatic forces. The protein backbone can be seen as N\ud835\udc41Nitalic_N rigid bodies (corresponding to N\ud835\udc41Nitalic_N residues) that contain four heavy atoms N\u2212C\u03b1\u2212C\u2212ONsubscriptC\ud835\udefcCO\\text{N}-\\text{C}_{\\alpha}-\\text{C}-\\text{O}N - C start_POSTSUBSCRIPT italic_\u03b1 end_POSTSUBSCRIPT - C - O. Mathematically, each residue can be associated with a frame that adheres to the symmetries of orientation-preserving rigid transformations (3D rotations and translations), forming the special Euclidean group SE\u2062(3)SE3\\mathrm{SE(3)}roman_SE ( 3 )\u00a0(Jumper et\u00a0al., 2021); the entire protein backbone is described by the group product SE\u2062(3)NSEsuperscript3\ud835\udc41\\mathrm{SE(3)}^{\\scriptscriptstyle N}roman_SE ( 3 ) start_POSTSUPERSCRIPT italic_N end_POSTSUPERSCRIPT. The problem of protein design can be formulated as sampling from the distribution over this group, which is a perspective used in our paper. Recently, generative models have been generalized to Riemannian manifolds\u00a0(Mathieu & Nickel, 2020; De\u00a0Bortoli et\u00a0al., 2022). However, they are not purpose-built to exploit the rich geometric structure of SE\u2062(3)NSEsuperscript3\ud835\udc41\\mathrm{SE(3)}^{\\scriptscriptstyle N}roman_SE ( 3 ) start_POSTSUPERSCRIPT italic_N end_POSTSUPERSCRIPT. Furthermore, several approaches require numerically expensive steps like simulating a Stochastic Differential Equation (SDE) during training or using the Riemannian divergence in the objective\u00a0(Huang et\u00a0al., 2022; Leach et\u00a0al., 2022; Ben-Hamu et\u00a0al., 2022).   Our approach. We introduce FoldFlow, a family of continuous normalizing flows (CNFs) tailored for distributions on SE\u2062(3)NSEsuperscript3\ud835\udc41\\mathrm{SE(3)}^{\\scriptscriptstyle N}roman_SE ( 3 ) start_POSTSUPERSCRIPT italic_N end_POSTSUPERSCRIPT (fig.\u00a01). We use the framework of Conditional Flow Matching (CFM), a simulation-free approach to learning CNFs by directly regressing time-dependent vector fields that generate probability paths \u00a0(Lipman et\u00a0al., 2022; Tong et\u00a0al., 2023b). In particular, we introduce three new CFM-based models that learn SE\u2062(3)NSEsuperscript3\ud835\udc41\\mathrm{SE(3)}^{\\scriptscriptstyle N}roman_SE ( 3 ) start_POSTSUPERSCRIPT italic_N end_POSTSUPERSCRIPT-invariant distributions to generate protein backbones. In contrast to the previous SE\u2062(3)SE3\\mathrm{SE(3)}roman_SE ( 3 ) diffusion approach of Yim et\u00a0al. (2023b), our FoldFlow is able to start from an",
            "references": [
                {
                    "title": "Harmonic Self-Conditioned Flow Matching for Multi-Ligand Docking and Binding Site Design",
                    "abstract": "A significant amount of protein function requires binding small molecules, including enzymatic catalysis. As such, designing binding pockets for small molecules has several impactful applications ranging from drug synthesis to energy storage. Towards this goal, we first develop HarmonicFlow, an improved generative process over 3D protein-ligand binding structures based on our self-conditioned flow matching objective. FlowSite extends this flow model to jointly generate a protein pocket's discrete residue types and the molecule's binding 3D structure. We show that HarmonicFlow improves upon the state-of-the-art generative processes for docking in simplicity, generality, and performance. Enabled by this structure modeling, FlowSite designs binding sites substantially better than baseline approaches and provides the first general solution for binding site design."
                },
                {
                    "title": "Fast protein backbone generation with SE(3) flow matching",
                    "abstract": "We present FrameFlow, a method for fast protein backbone generation using SE(3) flow matching. Specifically, we adapt FrameDiff, a state-of-the-art diffusion model, to the flow-matching generative modeling paradigm. We show how flow matching can be applied on SE(3) and propose modifications during training to effectively learn the vector field. Compared to FrameDiff, FrameFlow requires five times fewer sampling timesteps while achieving two fold better designability. The ability to generate high quality protein samples at a fraction of the cost of previous methods paves the way towards more efficient generative models in de novo protein design."
                },
                {
                    "title": "Protein generation with evolutionary diffusion: sequence is all you need",
                    "abstract": "Deep generative models are increasingly powerful tools for the in silico design of novel proteins. Recently, a family of generative models called diffusion models has demonstrated the ability to generate biologically plausible proteins that are dissimilar to any actual proteins seen in nature, enabling unprecedented capability and control in de novo protein design. However, current state-of-the-art models generate protein structures, which limits the scope of their training data and restricts generations to a small and biased subset of protein design space. Here, we introduce a general-purpose diffusion framework, EvoDiff, that combines evolutionary-scale data with the distinct conditioning capabilities of diffusion models for controllable protein generation in sequence space. EvoDiff generates high-fidelity, diverse, and structurally-plausible proteins that cover natural sequence and functional space. Critically, EvoDiff can generate proteins inaccessible to structure-based models, such as those with disordered regions, while maintaining the ability to design scaffolds for functional structural motifs, demonstrating the universality of our sequence-based formulation. We envision that EvoDiff will expand capabilities in protein engineering beyond the structure-function paradigm toward programmable, sequence-first design."
                },
                {
                    "title": "SE(3) Equivariant Augmented Coupling Flows",
                    "abstract": "Coupling normalizing flows allow for fast sampling and density evaluation, making them the tool of choice for probabilistic modeling of physical systems. However, the standard coupling architecture precludes endowing flows that operate on the Cartesian coordinates of atoms with the SE(3) and permutation invariances of physical systems. This work proposes a coupling flow that preserves SE(3) and permutation equivariance by performing coordinate splits along additional augmented dimensions. At each layer, the flow maps atoms' positions into learned SE(3) invariant bases, where we apply standard flow transformations, such as monotonic rational-quadratic splines, before returning to the original basis. Crucially, our flow preserves fast sampling and density evaluation, and may be used to produce unbiased estimates of expectations with respect to the target distribution via importance sampling. When trained on the DW4, LJ13, and QM9-positional datasets, our flow is competitive with equivariant continuous normalizing flows, while allowing sampling more than an order of magnitude faster. Moreover, to the best of our knowledge, we are the first to learn the full Boltzmann distribution of alanine dipeptide by only modeling the Cartesian positions of its atoms. Lastly, we demonstrate that our flow can be trained to approximately sample from the Boltzmann distribution of the DW4 and LJ13 particle systems using only their energy functions."
                },
                {
                    "title": "OpenFold: Retraining AlphaFold2 yields new insights into its learning mechanisms and capacity for generalization",
                    "abstract": "AlphaFold2 revolutionized structural biology with the ability to predict protein structures with exceptionally high accuracy. Its implementation, however, lacks the code and data required to train new models. These are necessary to (i) tackle new tasks, like protein-ligand complex structure prediction, (ii) investigate the process by which the model learns, which remains poorly understood, and (iii) assess the model\u2019s generalization capacity to unseen regions of fold space. Here we report OpenFold, a fast, memory-efficient, and trainable implementation of AlphaFold2. We train OpenFold from scratch, fully matching the accuracy of AlphaFold2. Having established parity, we assess OpenFold\u2019s capacity to generalize across fold space by retraining it using carefully designed datasets. We find that OpenFold is remarkably robust at generalizing despite extreme reductions in training set size and diversity, including near-complete elisions of classes of secondary structure elements. By analyzing intermediate structures produced by OpenFold during training, we also gain surprising insights into the manner in which the model learns to fold proteins, discovering that spatial dimensions are learned sequentially. Taken together, our studies demonstrate the power and utility of OpenFold, which we believe will prove to be a crucial new resource for the protein modeling community."
                },
                {
                    "title": "Simulation-free Schr\u00f6dinger bridges via score and flow matching",
                    "abstract": "We present simulation-free score and flow matching ([SF]$^2$M), a simulation-free objective for inferring stochastic dynamics given unpaired samples drawn from arbitrary source and target distributions. Our method generalizes both the score-matching loss used in the training of diffusion models and the recently proposed flow matching loss used in the training of continuous normalizing flows. [SF]$^2$M interprets continuous-time stochastic generative modeling as a Schr\\\"odinger bridge problem. It relies on static entropy-regularized optimal transport, or a minibatch approximation, to efficiently learn the SB without simulating the learned stochastic process. We find that [SF]$^2$M is more efficient and gives more accurate solutions to the SB problem than simulation-based methods from prior work. Finally, we apply [SF]$^2$M to the problem of learning cell dynamics from snapshot data. Notably, [SF]$^2$M is the first method to accurately model cell dynamics in high dimensions and can recover known gene regulatory networks from simulated data. Our code is available in the TorchCFM package at https://github.com/atong01/conditional-flow-matching."
                },
                {
                    "title": "Equivariant flow matching",
                    "abstract": "Normalizing flows are a class of deep generative models that are especially interesting for modeling probability distributions in physics, where the exact likelihood of flows allows reweighting to known target energy functions and computing unbiased observables. For instance, Boltzmann generators tackle the long-standing sampling problem in statistical physics by training flows to produce equilibrium samples of many-body systems such as small molecules and proteins. To build effective models for such systems, it is crucial to incorporate the symmetries of the target energy into the model, which can be achieved by equivariant continuous normalizing flows (CNFs). However, CNFs can be computationally expensive to train and generate samples from, which has hampered their scalability and practical application. In this paper, we introduce equivariant flow matching, a new training objective for equivariant CNFs that is based on the recently proposed optimal transport flow matching. Equivariant flow matching exploits the physical symmetries of the target energy for efficient, simulation-free training of equivariant CNFs. We demonstrate the effectiveness of flow matching on rotation and permutation invariant many-particle systems and a small molecule, alanine dipeptide, where for the first time we obtain a Boltzmann generator with significant sampling efficiency without relying on tailored internal coordinate featurization. Our results show that the equivariant flow matching objective yields flows with shorter integration paths, improved sampling efficiency, and higher scalability compared to existing methods."
                },
                {
                    "title": "Towards Predicting Equilibrium Distributions for Molecular Systems with Deep Learning",
                    "abstract": "Advances in deep learning have greatly improved structure prediction of molecules. However, many macroscopic observations that are important for real-world applications are not functions of a single molecular structure, but rather determined from the equilibrium distribution of structures. Traditional methods for obtaining these distributions, such as molecular dynamics simulation, are computationally expensive and often intractable. In this paper, we introduce a novel deep learning framework, called Distributional Graphormer (DiG), in an attempt to predict the equilibrium distribution of molecular systems. Inspired by the annealing process in thermodynamics, DiG employs deep neural networks to transform a simple distribution towards the equilibrium distribution, conditioned on a descriptor of a molecular system, such as a chemical graph or a protein sequence. This framework enables efficient generation of diverse conformations and provides estimations of state densities. We demonstrate the performance of DiG on several molecular tasks, including protein conformation sampling, ligand structure sampling, catalyst-adsorbate sampling, and property-guided structure generation. DiG presents a significant advancement in methodology for statistically understanding molecular systems, opening up new research opportunities in molecular science."
                },
                {
                    "title": "Normalizing flows for lattice gauge theory in arbitrary space-time dimension",
                    "abstract": "Applications of normalizing flows to the sampling of field configurations in lattice gauge theory have so far been explored almost exclusively in two space-time dimensions. We report new algorithmic developments of gauge-equivariant flow architectures facilitating the generalization to higher-dimensional lattice geometries. Specifically, we discuss masked autoregressive transformations with tractable and unbiased Jacobian determinants, a key ingredient for scalable and asymptotically exact flow-based sampling algorithms. For concreteness, results from a proof-of-principle application to SU(3) lattice gauge theory in four space-time dimensions are reported."
                },
                {
                    "title": "Multisample Flow Matching: Straightening Flows with Minibatch Couplings",
                    "abstract": "Simulation-free methods for training continuous-time generative models construct probability paths that go between noise distributions and individual data samples. Recent works, such as Flow Matching, derived paths that are optimal for each data sample. However, these algorithms rely on independent data and noise samples, and do not exploit underlying structure in the data distribution for constructing probability paths. We propose Multisample Flow Matching, a more general framework that uses non-trivial couplings between data and noise samples while satisfying the correct marginal constraints. At very small overhead costs, this generalization allows us to (i) reduce gradient variance during training, (ii) obtain straighter flows for the learned vector field, which allows us to generate high-quality samples using fewer function evaluations, and (iii) obtain transport maps with lower cost in high dimensions, which has applications beyond generative modeling. Importantly, we do so in a completely simulation-free manner with a simple minimization objective. We show that our proposed methods improve sample consistency on downsampled ImageNet data sets, and lead to better low-cost sample generation."
                },
                {
                    "title": "Diffusion Schr\u00f6dinger Bridge Matching",
                    "abstract": "Solving transport problems, i.e. finding a map transporting one given distribution to another, has numerous applications in machine learning. Novel mass transport methods motivated by generative modeling have recently been proposed, e.g. Denoising Diffusion Models (DDMs) and Flow Matching Models (FMMs) implement such a transport through a Stochastic Differential Equation (SDE) or an Ordinary Differential Equation (ODE). However, while it is desirable in many applications to approximate the deterministic dynamic Optimal Transport (OT) map which admits attractive properties, DDMs and FMMs are not guaranteed to provide transports close to the OT map. In contrast, Schr\\\"odinger bridges (SBs) compute stochastic dynamic mappings which recover entropy-regularized versions of OT. Unfortunately, existing numerical methods approximating SBs either scale poorly with dimension or accumulate errors across iterations. In this work, we introduce Iterative Markovian Fitting (IMF), a new methodology for solving SB problems, and Diffusion Schr\\\"odinger Bridge Matching (DSBM), a novel numerical algorithm for computing IMF iterates. DSBM significantly improves over previous SB numerics and recovers as special/limiting cases various recent transport methods. We demonstrate the performance of DSBM on a variety of problems."
                },
                {
                    "title": "EDGI: Equivariant Diffusion for Planning with Embodied Agents",
                    "abstract": "Embodied agents operate in a structured world, often solving tasks with spatial, temporal, and permutation symmetries. Most algorithms for planning and model-based reinforcement learning (MBRL) do not take this rich geometric structure into account, leading to sample inefficiency and poor generalization. We introduce the Equivariant Diffuser for Generating Interactions (EDGI), an algorithm for MBRL and planning that is equivariant with respect to the product of the spatial symmetry group SE(3), the discrete-time translation group Z, and the object permutation group Sn. EDGI follows the Diffuser framework (Janner et al., 2022) in treating both learning a world model and planning in it as a conditional generative modeling problem, training a diffusion model on an offline trajectory dataset. We introduce a new SE(3)xZxSn-equivariant diffusion model that supports multiple representations. We integrate this model in a planning loop, where conditioning and classifier guidance let us softly break the symmetry for specific tasks as needed. On object manipulation and navigation tasks, EDGI is substantially more sample efficient and generalizes better across the symmetry group than non-equivariant models."
                },
                {
                    "title": "Stochastic Interpolants: A Unifying Framework for Flows and Diffusions",
                    "abstract": "A class of generative models that unifies flow-based and diffusion-based methods is introduced. These models extend the framework proposed in Albergo&Vanden-Eijnden (2023), enabling the use of a broad class of continuous-time stochastic processes called `stochastic interpolants' to bridge any two arbitrary probability density functions exactly in finite time. These interpolants are built by combining data from the two prescribed densities with an additional latent variable that shapes the bridge in a flexible way. The time-dependent probability density function of the stochastic interpolant is shown to satisfy a first-order transport equation as well as a family of forward and backward Fokker-Planck equations with tunable diffusion. Upon consideration of the time evolution of an individual sample, this viewpoint immediately leads to both deterministic and stochastic generative models based on probability flow equations or stochastic differential equations with an adjustable level of noise. The drift coefficients entering these models are time-dependent velocity fields characterized as the unique minimizers of simple quadratic objective functions, one of which is a new objective for the score of the interpolant density. Remarkably, we show that minimization of these quadratic objectives leads to control of the likelihood for any of our generative models built upon stochastic dynamics. By contrast, we establish that generative models based upon a deterministic dynamics must, in addition, control the Fisher divergence between the target and the model. We also construct estimators for the likelihood and the cross-entropy of interpolant-based generative models, discuss connections with other stochastic bridges, and demonstrate that such models recover the Schr\\\"odinger bridge between the two target densities when explicitly optimizing over the interpolant."
                },
                {
                    "title": "Aligned Diffusion Schr\u00f6dinger Bridges",
                    "abstract": "Diffusion Schr\\\"odinger bridges (DSB) have recently emerged as a powerful framework for recovering stochastic dynamics via their marginal observations at different time points. Despite numerous successful applications, existing algorithms for solving DSBs have so far failed to utilize the structure of aligned data, which naturally arises in many biological phenomena. In this paper, we propose a novel algorithmic framework that, for the first time, solves DSBs while respecting the data alignment. Our approach hinges on a combination of two decades-old ideas: The classical Schr\\\"odinger bridge theory and Doob's $h$-transform. Compared to prior methods, our approach leads to a simpler training procedure with lower variance, which we further augment with principled regularization schemes. This ultimately leads to sizeable improvements across experiments on synthetic and real data, including the tasks of predicting conformational changes in proteins and temporal evolution of cellular differentiation processes."
                },
                {
                    "title": "I2SB: Image-to-Image Schr\u00f6dinger Bridge",
                    "abstract": "We propose Image-to-Image Schr\\\"odinger Bridge (I$^2$SB), a new class of conditional diffusion models that directly learn the nonlinear diffusion processes between two given distributions. These diffusion bridges are particularly useful for image restoration, as the degraded images are structurally informative priors for reconstructing the clean images. I$^2$SB belongs to a tractable class of Schr\\\"odinger bridge, the nonlinear extension to score-based models, whose marginal distributions can be computed analytically given boundary pairs. This results in a simulation-free framework for nonlinear diffusions, where the I$^2$SB training becomes scalable by adopting practical techniques used in standard diffusion models. We validate I$^2$SB in solving various image restoration tasks, including inpainting, super-resolution, deblurring, and JPEG restoration on ImageNet 256x256 and show that I$^2$SB surpasses standard conditional diffusion models with more interpretable generative processes. Moreover, I$^2$SB matches the performance of inverse methods that additionally require the knowledge of the corruption operators. Our work opens up new algorithmic opportunities for developing efficient nonlinear diffusion models on a large scale. scale. Project page and codes: https://i2sb.github.io/"
                },
                {
                    "title": "SE(3) diffusion model with application to protein backbone generation",
                    "abstract": "The design of novel protein structures remains a challenge in protein engineering for applications across biomedicine and chemistry. In this line of work, a diffusion model over rigid bodies in 3D (referred to as frames) has shown success in generating novel, functional protein backbones that have not been observed in nature. However, there exists no principled methodological framework for diffusion on SE(3), the space of orientation preserving rigid motions in R3, that operates on frames and confers the group invariance. We address these shortcomings by developing theoretical foundations of SE(3) invariant diffusion models on multiple frames followed by a novel framework, FrameDiff, for learning the SE(3) equivariant score over multiple frames. We apply FrameDiff on monomer backbone generation and find it can generate designable monomers up to 500 amino acids without relying on a pretrained protein structure prediction network that has been integral to previous methods. We find our samples are capable of generalizing beyond any known protein structure."
                },
                {
                    "title": "Improving and generalizing flow-based generative models with minibatch optimal transport",
                    "abstract": "Continuous normalizing flows (CNFs) are an attractive generative modeling technique, but they have been held back by limitations in their simulation-based maximum likelihood training. We introduce the generalized conditional flow matching (CFM) technique, a family of simulation-free training objectives for CNFs. CFM features a stable regression objective like that used to train the stochastic flow in diffusion models but enjoys the efficient inference of deterministic flow models. In contrast to both diffusion models and prior CNF training algorithms, CFM does not require the source distribution to be Gaussian or require evaluation of its density. A variant of our objective is optimal transport CFM (OT-CFM), which creates simpler flows that are more stable to train and lead to faster inference, as evaluated in our experiments. Furthermore, we show that when the true OT plan is available, our OT-CFM method approximates dynamic OT. Training CNFs with CFM improves results on a variety of conditional and unconditional generation tasks, such as inferring single cell dynamics, unsupervised image translation, and Schr\\\"odinger bridge inference."
                },
                {
                    "title": "Generating Novel, Designable, and Diverse Protein Structures by Equivariantly Diffusing Oriented Residue Clouds",
                    "abstract": "Proteins power a vast array of functional processes in living cells. The capability to create new proteins with designed structures and functions would thus enable the engineering of cellular behavior and development of protein-based therapeutics and materials. Structure-based protein design aims to find structures that are designable (can be realized by a protein sequence), novel (have dissimilar geometry from natural proteins), and diverse (span a wide range of geometries). While advances in protein structure prediction have made it possible to predict structures of novel protein sequences, the combinatorially large space of sequences and structures limits the practicality of search-based methods. Generative models provide a compelling alternative, by implicitly learning the low-dimensional structure of complex data distributions. Here, we leverage recent advances in denoising diffusion probabilistic models and equivariant neural networks to develop Genie, a generative model of protein structures that performs discrete-time diffusion using a cloud of oriented reference frames in 3D space. Through in silico evaluations, we demonstrate that Genie generates protein backbones that are more designable, novel, and diverse than existing models. This indicates that Genie is capturing key aspects of the distribution of protein structure space and facilitates protein design with high success rates. Code for generating new proteins and training new versions of Genie is available at https://github.com/aqlaboratory/genie."
                },
                {
                    "title": "Unsupervised Protein-Ligand Binding Energy Prediction via Neural Euler's Rotation Equation",
                    "abstract": "Protein-ligand binding prediction is a fundamental problem in AI-driven drug discovery. Prior work focused on supervised learning methods using a large set of binding affinity data for small molecules, but it is hard to apply the same strategy to other drug classes like antibodies as labelled data is limited. In this paper, we explore unsupervised approaches and reformulate binding energy prediction as a generative modeling task. Specifically, we train an energy-based model on a set of unlabelled protein-ligand complexes using SE(3) denoising score matching and interpret its log-likelihood as binding affinity. Our key contribution is a new equivariant rotation prediction network called Neural Euler's Rotation Equations (NERE) for SE(3) score matching. It predicts a rotation by modeling the force and torque between protein and ligand atoms, where the force is defined as the gradient of an energy function with respect to atom coordinates. We evaluate NERE on protein-ligand and antibody-antigen binding affinity prediction benchmarks. Our model outperforms all unsupervised baselines (physics-based and statistical potentials) and matches supervised learning methods in the antibody case."
                },
                {
                    "title": "A high-level programming language for generative protein design",
                    "abstract": "Combining a basic set of building blocks into more complex forms is a universal design principle. Most protein designs have proceeded from a manual bottom-up approach using parts created by nature, but top-down design of proteins is fundamentally hard due to biological complexity. We demonstrate how the modularity and programmability long sought for protein design can be realized through generative artificial intelligence. Advanced protein language models demonstrate emergent learning of atomic resolution structure and protein design principles. We leverage these developments to enable the programmable design of de novo protein sequences and structures of high complexity. First, we describe a high-level programming language based on modular building blocks that allows a designer to easily compose a set of desired properties. We then develop an energy-based generative model, built on atomic resolution structure prediction with a language model, that realizes all-atom structure designs that have the programmed properties. Designing a diverse set of specifications, including constraints on atomic coordinates, secondary structure, symmetry, and multimerization, demonstrates the generality and controllability of the approach. Enumerating constraints at increasing levels of hierarchical complexity shows that the approach can access a combinatorially large design space."
                },
                {
                    "title": "Language models generalize beyond natural proteins",
                    "abstract": "Learning the design patterns of proteins from sequences across evolution may have promise toward generative protein design. However it is unknown whether language models, trained on sequences of natural proteins, will be capable of more than memorization of existing protein families. Here we show that language models generalize beyond natural proteins to generate de novo proteins. We focus on two protein design tasks: fixed backbone design where the structure is specified, and unconstrained generation where the structure is sampled from the model. Remarkably although the models are trained only on sequences, we find that they are capable of designing structure. A total of 228 generated proteins are evaluated experimentally with high overall success rates (152/228 or 67%) in producing a soluble and monomeric species by size exclusion chromatography. Out of 152 experimentally successful designs, 35 have no significant sequence match to known natural proteins. Of the remaining 117, sequence identity to the nearest sequence match is at median 27%, below 20% for 6 designs, and as low as 18% for 3 designs. For fixed backbone design, the language model generates successful designs for each of eight experimentally evaluated artificially created fixed backbone targets. For unconstrained generation, sampled proteins cover diverse topologies and secondary structure compositions, and have high experimental success rate (71/129 or 55%). The designs reflect deep patterns linking sequence and structure, including motifs that occur in related natural structures, and motifs that are not observed in similar structural contexts in known protein families. The results show that language models, though only trained on sequences, learn a deep grammar that enables the design of protein structure, extending beyond natural proteins."
                },
                {
                    "title": "Equivariant 3D-Conditional Diffusion Models for Molecular Linker Design",
                    "abstract": "Fragment-based drug discovery has been an effective paradigm in early-stage drug development. An open challenge in this area is designing linkers between disconnected molecular fragments of interest to obtain chemically-relevant candidate drug molecules. In this work, we propose DiffLinker, an E(3)-equivariant 3D-conditional diffusion model for molecular linker design. Given a set of disconnected fragments, our model places missing atoms in between and designs a molecule incorporating all the initial fragments. Unlike previous approaches that are only able to connect pairs of molecular fragments, our method can link an arbitrary number of fragments. Additionally, the model automatically determines the number of atoms in the linker and its attachment points to the input fragments. We demonstrate that DiffLinker outperforms other methods on the standard datasets generating more diverse and synthetically-accessible molecules. Besides, we experimentally test our method in real-world applications, showing that it can successfully generate valid linkers conditioned on target protein pockets."
                },
                {
                    "title": "Flow Matching for Generative Modeling",
                    "abstract": "We introduce a new paradigm for generative modeling built on Continuous Normalizing Flows (CNFs), allowing us to train CNFs at unprecedented scale. Specifically, we present the notion of Flow Matching (FM), a simulation-free approach for training CNFs based on regressing vector fields of fixed conditional probability paths. Flow Matching is compatible with a general family of Gaussian probability paths for transforming between noise and data samples -- which subsumes existing diffusion paths as specific instances. Interestingly, we find that employing FM with diffusion paths results in a more robust and stable alternative for training diffusion models. Furthermore, Flow Matching opens the door to training CNFs with other, non-diffusion probability paths. An instance of particular interest is using Optimal Transport (OT) displacement interpolation to define the conditional probability paths. These paths are more efficient than diffusion paths, provide faster training and sampling, and result in better generalization. Training CNFs using Flow Matching on ImageNet leads to consistently better performance than alternative diffusion-based methods in terms of both likelihood and sample quality, and allows fast and reliable sample generation using off-the-shelf numerical ODE solvers."
                },
                {
                    "title": "Building Normalizing Flows with Stochastic Interpolants",
                    "abstract": "A generative model based on a continuous-time normalizing flow between any pair of base and target probability densities is proposed. The velocity field of this flow is inferred from the probability current of a time-dependent density that interpolates between the base and the target in finite time. Unlike conventional normalizing flow inference methods based the maximum likelihood principle, which require costly backpropagation through ODE solvers, our interpolant approach leads to a simple quadratic loss for the velocity itself which is expressed in terms of expectations that are readily amenable to empirical estimation. The flow can be used to generate samples from either the base or target, and to estimate the likelihood at any time along the interpolant. In addition, the flow can be optimized to minimize the path length of the interpolant density, thereby paving the way for building optimal transport maps. In situations where the base is a Gaussian density, we also show that the velocity of our normalizing flow can also be used to construct a diffusion model to sample the target as well as estimate its score. However, our approach shows that we can bypass this diffusion completely and work at the level of the probability flow with greater simplicity, opening an avenue for methods based solely on ordinary differential equations as an alternative to those based on stochastic differential equations. Benchmarking on density estimation tasks illustrates that the learned flow can match and surpass conventional continuous flows at a fraction of the cost, and compares well with diffusions on image generation on CIFAR-10 and ImageNet $32\\times32$. The method scales ab-initio ODE flows to previously unreachable image resolutions, demonstrated up to $128\\times128$."
                },
                {
                    "title": "Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow",
                    "abstract": "We present rectified flow, a surprisingly simple approach to learning (neural) ordinary differential equation (ODE) models to transport between two empirically observed distributions \\pi_0 and \\pi_1, hence providing a unified solution to generative modeling and domain transfer, among various other tasks involving distribution transport. The idea of rectified flow is to learn the ODE to follow the straight paths connecting the points drawn from \\pi_0 and \\pi_1 as much as possible. This is achieved by solving a straightforward nonlinear least squares optimization problem, which can be easily scaled to large models without introducing extra parameters beyond standard supervised learning. The straight paths are special and preferred because they are the shortest paths between two points, and can be simulated exactly without time discretization and hence yield computationally efficient models. We show that the procedure of learning a rectified flow from data, called rectification, turns an arbitrary coupling of \\pi_0 and \\pi_1 to a new deterministic coupling with provably non-increasing convex transport costs. In addition, recursively applying rectification allows us to obtain a sequence of flows with increasingly straight paths, which can be simulated accurately with coarse time discretization in the inference phase. In empirical studies, we show that rectified flow performs superbly on image generation, image-to-image translation, and domain adaptation. In particular, on image generation and translation, our method yields nearly straight flows that give high quality results even with a single Euler discretization step."
                },
                {
                    "title": "Uni-Fold: An Open-Source Platform for Developing Protein Folding Models beyond AlphaFold",
                    "abstract": "Recent breakthroughs on protein structure prediction, namely AlphaFold, have led to unprecedented new possibilities in related areas. However, the lack of training utilities in its current open-source code hinders the community from further developing or adapting the model. Here we present Uni-Fold as a thoroughly open-source platform for developing protein folding models beyond AlphaFold. We reimplemented AlphaFold and AlphaFold-Multimer in the PyTorch framework, and reproduced their from-scratch training processes with equivalent or better accuracy. Based on various optimizations, Uni-Fold achieves about 2.2 times training acceleration compared with AlphaFold under similar hardware configuration. On a benchmark of recently released multimeric protein structures, Uni-Fold outperforms AlphaFold-Multimer by approximately 2% on the TM-Score. Uni-Fold is currently the only open-source repository that supports both training and inference of multimeric protein models. The source code, model parameters, test data, and web server of Uni-Fold are publicly available3."
                },
                {
                    "title": "Riemannian Diffusion Models",
                    "abstract": "Diffusion models are recent state-of-the-art methods for image generation and likelihood estimation. In this work, we generalize continuous-time diffusion models to arbitrary Riemannian manifolds and derive a variational framework for likelihood estimation. Computationally, we propose new methods for computing the Riemannian divergence which is needed in the likelihood estimation. Moreover, in generalizing the Euclidean case, we prove that maximizing this variational lower-bound is equivalent to Riemannian score matching. Empirically, we demonstrate the expressive power of Riemannian diffusion models on a wide spectrum of smooth manifolds, such as spheres, tori, hyperboloids, and orthogonal groups. Our proposed method achieves new state-of-the-art likelihoods on all benchmarks."
                },
                {
                    "title": "High-resolution de novo structure prediction from primary sequence",
                    "abstract": "Recent breakthroughs have used deep learning to exploit evolutionary information in multiple sequence alignments (MSAs) to accurately predict protein structures. However, MSAs of homologous proteins are not always available, such as with orphan proteins or fast-evolving proteins like antibodies, and a protein typically folds in a natural setting from its primary amino acid sequence into its three-dimensional structure, suggesting that evolutionary information and MSAs should not be necessary to predict a protein\u2019s folded form. Here, we introduce OmegaFold, the first computational method to successfully predict high-resolution protein structure from a single primary sequence alone. Using a new combination of a protein language model that allows us to make predictions from single sequences and a geometry-inspired transformer model trained on protein structures, OmegaFold outperforms RoseTTAFold and achieves similar prediction accuracy to AlphaFold2 on recently released structures. OmegaFold enables accurate predictions on orphan proteins that do not belong to any functionally characterized protein family and antibodies that tend to have noisy MSAs due to fast evolution. Our study fills a much-encountered gap in structure prediction and brings us a step closer to understanding protein folding in nature."
                },
                {
                    "title": "Matching Normalizing Flows and Probability Paths on Manifolds",
                    "abstract": "Continuous Normalizing Flows (CNFs) are a class of generative models that transform a prior distribution to a model distribution by solving an ordinary differential equation (ODE). We propose to train CNFs on manifolds by minimizing probability path divergence (PPD), a novel family of divergences between the probability density path generated by the CNF and a target probability density path. PPD is formulated using a logarithmic mass conservation formula which is a linear first order partial differential equation relating the log target probabilities and the CNF's defining vector field. PPD has several key benefits over existing methods: it sidesteps the need to solve an ODE per iteration, readily applies to manifold data, scales to high dimensions, and is compatible with a large family of target paths interpolating pure noise and data in finite time. Theoretically, PPD is shown to bound classical probability divergences. Empirically, we show that CNFs learned by minimizing PPD achieve state-of-the-art results in likelihoods and sample quality on existing low-dimensional manifold benchmarks, and is the first example of a generative model to scale to moderately high dimensional manifolds."
                },
                {
                    "title": "Robust deep learning based protein sequence design using ProteinMPNN",
                    "abstract": "While deep learning has revolutionized protein structure prediction, almost all experimentally characterized de novo protein designs have been generated using physically based approaches such as Rosetta. Here we describe a deep learning based protein sequence design method, ProteinMPNN, with outstanding performance in both in silico and experimental tests. The amino acid sequence at different positions can be coupled between single or multiple chains, enabling application to a wide range of current protein design challenges. On native protein backbones, ProteinMPNN has a sequence recovery of 52.4%, compared to 32.9% for Rosetta. Incorporation of noise during training improves sequence recovery on protein structure models, and produces sequences which more robustly encode their structures as assessed using structure prediction algorithms. We demonstrate the broad utility and high accuracy of ProteinMPNN using X-ray crystallography, cryoEM and functional studies by rescuing previously failed designs, made using Rosetta or AlphaFold, of protein monomers, cyclic homo-oligomers, tetrahedral nanoparticles, and target binding proteins. One-sentence summary A deep learning based protein sequence design method is described that is widely applicable to current design challenges and shows outstanding performance in both in silico and experimental tests."
                },
                {
                    "title": "Protein Structure and Sequence Generation with Equivariant Denoising Diffusion Probabilistic Models",
                    "abstract": "Proteins are macromolecules that mediate a significant fraction of the cellular processes that underlie life. An important task in bioengineering is designing proteins with specific 3D structures and chemical properties which enable targeted functions. To this end, we introduce a generative model of both protein structure and sequence that can operate at significantly larger scales than previous molecular generative modeling approaches. The model is learned entirely from experimental data and conditions its generation on a compact specification of protein topology to produce a full-atom backbone configuration as well as sequence and side-chain predictions. We demonstrate the quality of the model via qualitative and quantitative analysis of its samples. Videos of sampling trajectories are available at https://nanand2.github.io/proteins ."
                },
                {
                    "title": "Equivariant Diffusion for Molecule Generation in 3D",
                    "abstract": "This work introduces a diffusion model for molecule generation in 3D that is equivariant to Euclidean transformations. Our E(3) Equivariant Diffusion Model (EDM) learns to denoise a diffusion process with an equivariant network that jointly operates on both continuous (atom coordinates) and categorical features (atom types). In addition, we provide a probabilistic analysis which admits likelihood computation of molecules using our model. Experimentally, the proposed method significantly outperforms previous 3D molecular generative methods regarding the quality of generated samples and efficiency at training time."
                },
                {
                    "title": "GeoDiff: a Geometric Diffusion Model for Molecular Conformation Generation",
                    "abstract": "Predicting molecular conformations from molecular graphs is a fundamental problem in cheminformatics and drug discovery. Recently, significant progress has been achieved with machine learning approaches, especially with deep generative models. Inspired by the diffusion process in classical non-equilibrium thermodynamics where heated particles will diffuse from original states to a noise distribution, in this paper, we propose a novel generative model named GeoDiff for molecular conformation prediction. GeoDiff treats each atom as a particle and learns to directly reverse the diffusion process (i.e., transforming from a noise distribution to stable conformations) as a Markov chain. Modeling such a generation process is however very challenging as the likelihood of conformations should be roto-translational invariant. We theoretically show that Markov chains evolving with equivariant Markov kernels can induce an invariant distribution by design, and further propose building blocks for the Markov kernels to preserve the desirable equivariance property. The whole framework can be efficiently trained in an end-to-end fashion by optimizing a weighted variational lower bound to the (conditional) likelihood. Experiments on multiple benchmarks show that GeoDiff is superior or comparable to existing state-of-the-art approaches, especially on large molecules."
                },
                {
                    "title": "Riemannian Score-Based Generative Modeling",
                    "abstract": "Score-based generative models (SGMs) are a powerful class of generative models that exhibit remarkable empirical performance. Score-based generative modelling (SGM) consists of a ``noising'' stage, whereby a diffusion is used to gradually add Gaussian noise to data, and a generative model, which entails a ``denoising'' process defined by approximating the time-reversal of the diffusion. Existing SGMs assume that data is supported on a Euclidean space, i.e. a manifold with flat geometry. In many domains such as robotics, geoscience or protein modelling, data is often naturally described by distributions living on Riemannian manifolds and current SGM techniques are not appropriate. We introduce here Riemannian Score-based Generative Models (RSGMs), a class of generative models extending SGMs to Riemannian manifolds. We demonstrate our approach on a variety of manifolds, and in particular with earth and climate science spherical data."
                },
                {
                    "title": "Bridge Simulation and Metric Estimation on Lie Groups and Homogeneous Spaces",
                    "abstract": "We present schemes for simulating Brownian bridges on complete and connected Lie groups and homogeneous spaces. We use this to construct an estimation scheme for recovering an unknown left- or right-invariant Riemannian metric on the Lie group from samples. We subsequently show how pushing forward the distributions generated by Brownian motions on the group results in distributions on homogeneous spaces that exhibit non-trivial covariance structure. The pushforward measure gives rise to new parametric families of distributions on commonly occurring spaces such as spheres and symmetric positive tensors. We extend the estimation scheme to fit these distributions to homogeneous space-valued data. We demonstrate both the simulation schemes and estimation procedures on Lie groups and homogenous spaces, including $\\SPD(3) = \\GL_+(3)/\\SO(3)$ and $\\mathbb S^2 = \\SO(3)/\\SO(2)$."
                },
                {
                    "title": "AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models",
                    "abstract": "Abstract The AlphaFold Protein Structure Database (AlphaFold DB, https://alphafold.ebi.ac.uk) is an openly accessible, extensive database of high-accuracy protein-structure predictions. Powered by AlphaFold v2.0 of DeepMind, it has enabled an unprecedented expansion of the structural coverage of the known protein-sequence space. AlphaFold DB provides programmatic access to and interactive visualization of predicted atomic coordinates, per-residue and pairwise model-confidence estimates and predicted aligned errors. The initial release of AlphaFold DB contains over 360,000 predicted structures across 21 model-organism proteomes, which will soon be expanded to cover most of the (over 100 million) representative sequences from the UniRef90 data set."
                },
                {
                    "title": "Equivariant Finite Normalizing Flows",
                    "abstract": "Generative modelling seeks to uncover the underlying factors that give rise to observed data that can often be modeled as the natural symmetries that manifest themselves through in-variances and equivariances to certain transformation laws. However, current approaches to representing these symmetries are couched in the formalism of continuous normalizing \ufb02ows that require the construction of equivariant vector \ufb01elds\u2014inhibiting their simple application to conventional higher dimensional generative modelling domains like natural images. In this paper, we focus on building equivariant normalizing \ufb02ows using discrete layers. We \ufb01rst theoretically prove the existence of an equivariant map for compact groups whose actions are on compact spaces. We further introduce three new equivariant \ufb02ows: G -Residual Flows, G -Coupling Flows, and G -Inverse Autoregressive Flows that elevate classical Residual, Coupling, and Inverse Autoregressive Flows with equivariant maps to a prescribed group G . Our construction of G -Residual Flows are also universal, in the sense that we prove an G -equivariant di\ufb00eomorphism can be exactly mapped by a G -residual \ufb02ow. Finally, we complement our theoretical insights with demonstrative experiments\u2014for the \ufb01rst time\u2014 on image datasets like CIFAR-10 and show G -Equivariant Finite Normalizing \ufb02ows lead to increased data e\ufb03ciency, faster convergence, and improved likelihood estimates."
                },
                {
                    "title": "Directed Evolution: Methodologies and Applications.",
                    "abstract": "Directed evolution aims to expedite the natural evolution process of biological molecules and systems in a test tube through iterative rounds of gene diversifications and library screening/selection. It has become one of the most powerful and widespread tools for engineering improved or novel functions in proteins, metabolic pathways, and even whole genomes. This review describes the commonly used gene diversification strategies, screening/selection methods, and recently developed continuous evolution strategies for directed evolution. Moreover, we highlight some representative applications of directed evolution in engineering nucleic acids, proteins, pathways, genetic circuits, viruses, and whole cells. Finally, we discuss the challenges and future perspectives in directed evolution."
                },
                {
                    "title": "Equivariant Manifold Flows",
                    "abstract": "Tractably modelling distributions over manifolds has long been an important goal in the natural sciences. Recent work has focused on developing general machine learning models to learn such distributions. However, for many applications these distributions must respect manifold symmetries -- a trait which most previous models disregard. In this paper, we lay the theoretical foundations for learning symmetry-invariant distributions on arbitrary manifolds via equivariant manifold flows. We demonstrate the utility of our approach by using it to learn gauge invariant densities over $SU(n)$ in the context of quantum field theory."
                },
                {
                    "title": "Riemannian Convex Potential Maps",
                    "abstract": "Modeling distributions on Riemannian manifolds is a crucial component in understanding non-Euclidean data that arises, e.g., in physics and geology. The budding approaches in this space are limited by representational and computational tradeoffs. We propose and study a class of flows that uses convex potentials from Riemannian optimal transport. These are universal and can model distributions on any compact Riemannian manifold without requiring domain knowledge of the manifold to be integrated into the architecture. We demonstrate that these flows can model standard distributions on spheres, and tori, on synthetic and geological data. Our source code is freely available online at http://github.com/facebookresearch/rcpm"
                },
                {
                    "title": "Diffusion Schr\u00f6dinger Bridge with Applications to Score-Based Generative Modeling",
                    "abstract": "Progressively applying Gaussian noise transforms complex data distributions to approximately Gaussian. Reversing this dynamic defines a generative model. When the forward noising process is given by a Stochastic Differential Equation (SDE), Song et al. (2021) demonstrate how the time inhomogeneous drift of the associated reverse-time SDE may be estimated using score-matching. A limitation of this approach is that the forward-time SDE must be run for a sufficiently long time for the final distribution to be approximately Gaussian. In contrast, solving the Schr\\\"odinger Bridge problem (SB), i.e. an entropy-regularized optimal transport problem on path spaces, yields diffusions which generate samples from the data distribution in finite time. We present Diffusion SB (DSB), an original approximation of the Iterative Proportional Fitting (IPF) procedure to solve the SB problem, and provide theoretical analysis along with generative modeling experiments. The first DSB iteration recovers the methodology proposed by Song et al. (2021), with the flexibility of using shorter time intervals, as subsequent DSB iterations reduce the discrepancy between the final-time marginal of the forward (resp. backward) SDE with respect to the prior (resp. data) distribution. Beyond generative modeling, DSB offers a widely applicable computational optimal transport tool as the continuous state-space analogue of the popular Sinkhorn algorithm (Cuturi, 2013)."
                },
                {
                    "title": "E(n) Equivariant Normalizing Flows",
                    "abstract": "This paper introduces a generative model equivariant to Euclidean symmetries: E(n) Equivariant Normalizing Flows (E-NFs). To construct E-NFs, we take the discriminative E(n) graph neural networks and integrate them as a differential equation to obtain an invertible equivariant function: a continuous-time normalizing flow. We demonstrate that E-NFs considerably outperform baselines and existing methods from the literature on particle systems such as DW4 and LJ13, and on molecules from QM9 in terms of log-likelihood. To the best of our knowledge, this is the first flow that jointly generates molecule features and positions in 3D."
                },
                {
                    "title": "Unbalanced minibatch Optimal Transport; applications to Domain Adaptation",
                    "abstract": "Optimal transport distances have found many applications in machine learning for their capacity to compare non-parametric probability distributions. Yet their algorithmic complexity generally prevents their direct use on large scale datasets. Among the possible strategies to alleviate this issue, practitioners can rely on computing estimates of these distances over subsets of data, {\\em i.e.} minibatches. While computationally appealing, we highlight in this paper some limits of this strategy, arguing it can lead to undesirable smoothing effects. As an alternative, we suggest that the same minibatch strategy coupled with unbalanced optimal transport can yield more robust behavior. We discuss the associated theoretical properties, such as unbiased estimators, existence of gradients and concentration bounds. Our experimental study shows that in challenging problems associated to domain adaptation, the use of unbalanced optimal transport leads to significantly better results, competing with or surpassing recent baselines."
                },
                {
                    "title": "Manifold Density Estimation via Generalized Dequantization",
                    "abstract": "Density estimation is an important technique for characterizing distributions given observations. Much existing research on density estimation has focused on cases wherein the data lies in a Euclidean space. However, some kinds of data are not well-modeled by supposing that their underlying geometry is Euclidean. Instead, it can be useful to model such data as lying on a {\\it manifold} with some known structure. For instance, some kinds of data may be known to lie on the surface of a sphere. We study the problem of estimating densities on manifolds. We propose a method, inspired by the literature on\"dequantization,\"which we interpret through the lens of a coordinate transformation of an ambient Euclidean space and a smooth manifold of interest. Using methods from normalizing flows, we apply this method to the dequantization of smooth manifold structures in order to model densities on the sphere, tori, and the orthogonal group."
                },
                {
                    "title": "Minibatch optimal transport distances; analysis and applications",
                    "abstract": "Optimal transport distances have become a classic tool to compare probability distributions and have found many applications in machine learning. Yet, despite recent algorithmic developments, their complexity prevents their direct use on large scale datasets. To overcome this challenge, a common workaround is to compute these distances on minibatches i.e. to average the outcome of several smaller optimal transport problems. We propose in this paper an extended analysis of this practice, which effects were previously studied in restricted cases. We first consider a large variety of Optimal Transport kernels. We notably argue that the minibatch strategy comes with appealing properties such as unbiased estimators, gradients and a concentration bound around the expectation, but also with limits: the minibatch OT is not a distance. To recover some of the lost distance axioms, we introduce a debiased minibatch OT function and study its statistical and optimisation properties. Along with this theoretical analysis, we also conduct empirical experiments on gradient flows, generative adversarial networks (GANs) or color transfer that highlight the practical interest of this strategy."
                },
                {
                    "title": "Same Same But DifferNet: Semi-Supervised Defect Detection with Normalizing Flows",
                    "abstract": "The detection of manufacturing errors is crucial in fabrication processes to ensure product quality and safety standards. Since many defects occur very rarely and their characteristics are mostly unknown a priori, their detection is still an open research question. To this end, we propose DifferNet: It leverages the descriptiveness of features extracted by convolutional neural networks to estimate their density using normalizing flows. Normalizing flows are well-suited to deal with low dimensional data distributions. However, they struggle with the high dimensionality of images. Therefore, we employ a multi-scale feature extractor which enables the normalizing flow to assign meaningful likelihoods to the images. Based on these likelihoods we develop a scoring function that indicates defects. Moreover, propagating the score back to the image enables pixel-wise localization. To achieve a high robustness and performance we exploit multiple transformations in training and evaluation. In contrast to most other methods, ours does not require a large number of training samples and performs well with as low as 16 images. We demonstrate the superior performance over existing approaches on the challenging and newly proposed MVTec AD [4] and Magnetic Tile Defects [14] datasets."
                },
                {
                    "title": "Sampling using SU(N) gauge equivariant flows",
                    "abstract": "We develop a flow-based sampling algorithm for $SU(N)$ lattice gauge theories that is gauge-invariant by construction. Our key contribution is constructing a class of flows on an $SU(N)$ variable (or on a $U(N)$ variable by a simple alternative) that respect matrix conjugation symmetry. We apply this technique to sample distributions of single $SU(N)$ variables and to construct flow-based samplers for $SU(2)$ and $SU(3)$ lattice gauge theory in two dimensions."
                },
                {
                    "title": "De novo design of picomolar SARS-CoV-2 miniprotein inhibitors",
                    "abstract": "We used two approaches to design proteins with shape and chemical complementarity to the receptor binding domain (RBD) of SARS-CoV-2 Spike protein near the binding site for the human ACE2 receptor. Scaffolds were built around an ACE2 helix that interacts with the RBD, or de novo designed scaffolds were docked against the RBD to identify new binding modes. In both cases, designed sequences were optimized first in silico and then experimentally for target binding, folding and stability. Nine designs bound the RBD with affinities ranging from 100pM to 10nM, and blocked bona fide SARS-CoV-2 infection of Vero E6 cells with IC50 values ranging from 35 pM to 35 nM; the most potent of these \u2014 56 and 64 residue hyperstable proteins made using the second approach \u2014 are roughly six times more potent on a per mass basis (IC50 ~ 0.23 ng/ml) than the best monoclonal antibodies reported thus far. Cryo-electron microscopy structures of the SARS-CoV-2 spike ectodomain trimer in complex with the two most potent minibinders show that the structures of the designs and their binding interactions with the RBD are nearly identical to the computational models, and that all three RBDs in a single Spike protein can be engaged simultaneously. These hyperstable minibinders provide promising starting points for new SARS-CoV-2 therapeutics, and illustrate the power of computational protein design for rapidly generating potential therapeutic candidates against pandemic threats."
                },
                {
                    "title": "Riemannian Continuous Normalizing Flows",
                    "abstract": "Normalizing flows have shown great promise for modelling flexible probability distributions in a computationally tractable way. However, whilst data is often naturally described on Riemannian manifolds such as spheres, torii, and hyperbolic spaces, most normalizing flows implicitly assume a flat geometry, making them either misspecified or ill-suited in these situations. To overcome this problem, we introduce Riemannian continuous normalizing flows, a model which admits the parametrization of flexible probability measures on smooth manifolds by defining flows as the solution to ordinary differential equations. We show that this approach can lead to substantial improvements on both synthetic and real-world data when compared to standard flows or previously introduced projected flows."
                },
                {
                    "title": "Equivariant Flows: exact likelihood generative learning for symmetric densities",
                    "abstract": "Normalizing flows are exact-likelihood generative neural networks which approximately transform samples from a simple prior distribution to samples of the probability distribution of interest. Recent work showed that such generative models can be utilized in statistical mechanics to sample equilibrium states of many-body systems in physics and chemistry. To scale and generalize these results, it is essential that the natural symmetries in the probability density - in physics defined by the invariances of the target potential - are built into the flow. We provide a theoretical sufficient criterion showing that the distribution generated by equivariant normalizing flows is invariant with respect to these symmetries by design. Furthermore, we propose building blocks for flows which preserve symmetries which are usually found in physical/chemical many-body particle systems. Using benchmark systems motivated from molecular physics, we demonstrate that those symmetry preserving flows can provide better generalization capabilities and sampling efficiency."
                },
                {
                    "title": "Equivariant flow-based sampling for lattice gauge theory",
                    "abstract": "We define a class of machine-learned flow-based sampling algorithms for lattice gauge theories that are gauge invariant by construction. We demonstrate the application of this framework to U(1) gauge theory in two spacetime dimensions, and find that, at small bare coupling, the approach is orders of magnitude more efficient at sampling topological quantities than more traditional sampling procedures such as hybrid Monte\u00a0Carlo and heat bath."
                },
                {
                    "title": "TrajectoryNet: A Dynamic Optimal Transport Network for Modeling Cellular Dynamics",
                    "abstract": "It is increasingly common to encounter data from dynamic processes captured by static cross-sectional measurements over time, particularly in biomedical settings. Recent attempts to model individual trajectories from this data use optimal transport to create pairwise matchings between time points. However, these methods cannot model continuous dynamics and non-linear paths that entities can take in these systems. To address this issue, we establish a link between continuous normalizing flows and dynamic optimal transport, that allows us to model the expected paths of points over time. Continuous normalizing flows are generally under constrained, as they are allowed to take an arbitrary path from the source to the target distribution. We present TrajectoryNet, which controls the continuous paths taken between distributions to produce dynamic optimal transport. We show how this is particularly applicable for studying cellular dynamics in data from single-cell RNA sequencing (scRNA-seq) technologies, and that TrajectoryNet improves upon recently proposed static optimal transport-based models that can be used for interpolating cellular distributions."
                },
                {
                    "title": "How to train your neural ODE",
                    "abstract": "Training neural ODEs on large datasets has not been tractable due to the necessity of allowing the adaptive numerical ODE solver to refine its step size to very small values. In practice this leads to dynamics equivalent to many hundreds or even thousands of layers. In this paper, we overcome this apparent difficulty by introducing a theoretically-grounded combination of both optimal transport and stability regularizations which encourage neural ODEs to prefer simpler dynamics out of all the dynamics that solve a problem well. Simpler dynamics lead to faster convergence and to fewer discretizations of the solver, considerably decreasing wall-clock time without loss in performance. Our approach allows us to train neural ODE based generative models to the same performance as the unregularized dynamics in just over a day on one GPU, whereas unregularized dynamics can take up to 4-6 days of training time on multiple GPUs. This brings neural ODEs significantly closer to practical relevance in large-scale applications."
                },
                {
                    "title": "Learning with minibatch Wasserstein : asymptotic and gradient properties",
                    "abstract": "Optimal transport distances are powerful tools to compare probability distributions and have found many applications in machine learning. Yet their algorithmic complexity prevents their direct use on large scale datasets. To overcome this challenge, practitioners compute these distances on minibatches {\\em i.e.} they average the outcome of several smaller optimal transport problems. We propose in this paper an analysis of this practice, which effects are not well understood so far. We notably argue that it is equivalent to an implicit regularization of the original problem, with appealing properties such as unbiased estimators, gradients and a concentration bound around the expectation, but also with defects such as loss of distance property. Along with this theoretical analysis, we also conduct empirical experiments on gradient flows, GANs or color transfer that highlight the practical interest of this strategy."
                },
                {
                    "title": "Computational Optimal Transport",
                    "abstract": "Optimal transport (OT) theory can be informally described using the words of the French mathematician Gaspard Monge (1746-1818): A worker with a shovel in hand has to move a large pile of sand lying on a construction site. The goal of the worker is to erect with all that sand a target pile with a prescribed shape (for example, that of a giant sand castle). Naturally, the worker wishes to minimize her total effort, quantified for instance as the total distance or time spent carrying shovelfuls of sand. Mathematicians interested in OT cast that problem as that of comparing two probability distributions, two different piles of sand of the same volume. They consider all of the many possible ways to morph, transport or reshape the first pile into the second, and associate a \"global\" cost to every such transport, using the \"local\" consideration of how much it costs to move a grain of sand from one place to another. Recent years have witnessed the spread of OT in several fields, thanks to the emergence of approximate solvers that can scale to sizes and dimensions that are relevant to data sciences. Thanks to this newfound scalability, OT is being increasingly used to unlock various problems in imaging sciences (such as color or texture processing), computer vision and graphics (for shape manipulation) or machine learning (for regression, classification and density fitting). This short book reviews OT with a bias toward numerical methods and their applications in data sciences, and sheds lights on the theoretical properties of OT that make it particularly useful for some of these applications."
                },
                {
                    "title": "Improving GANs Using Optimal Transport",
                    "abstract": "We present Optimal Transport GAN (OT-GAN), a variant of generative adversarial nets minimizing a new metric measuring the distance between the generator distribution and the data distribution. This metric, which we call mini-batch energy distance, combines optimal transport in primal form with an energy distance defined in an adversarially learned feature space, resulting in a highly discriminative distance function with unbiased mini-batch gradients. Experimentally we show OT-GAN to be highly stable when trained with large mini-batches, and we present state-of-the-art results on several popular benchmark problems for image generation."
                },
                {
                    "title": "Learning Generative Models with Sinkhorn Divergences",
                    "abstract": "The ability to compare two degenerate probability distributions (i.e. two probability distributions supported on two distinct low-dimensional manifolds living in a much higher-dimensional space) is a crucial problem arising in the estimation of generative models for high-dimensional observations such as those arising in computer vision or natural language. It is known that optimal transport metrics can represent a cure for this problem, since they were specifically designed as an alternative to information divergences to handle such problematic scenarios. Unfortunately, training generative machines using OT raises formidable computational and statistical challenges, because of (i) the computational burden of evaluating OT losses, (ii) the instability and lack of smoothness of these losses, (iii) the difficulty to estimate robustly these losses and their gradients in high dimension. This paper presents the first tractable computational method to train large scale generative models using an optimal transport loss, and tackles these three issues by relying on two key ideas: (a) entropic smoothing, which turns the original OT loss into one that can be computed using Sinkhorn fixed point iterations; (b) algorithmic (automatic) differentiation of these iterations. These two approximations result in a robust and differentiable approximation of the OT loss with streamlined GPU execution. Entropic smoothing generates a family of losses interpolating between Wasserstein (OT) and Maximum Mean Discrepancy (MMD), thus allowing to find a sweet spot leveraging the geometry of OT and the favorable high-dimensional sample complexity of MMD which comes with unbiased gradient estimates. The resulting computational architecture complements nicely standard deep network generative models by a stack of extra layers implementing the loss function."
                },
                {
                    "title": "Wasserstein GAN",
                    "abstract": "We introduce a new algorithm named WGAN, an alternative to traditional GAN training. In this new model, we show that we can improve the stability of learning, get rid of problems like mode collapse, and provide meaningful learning curves useful for debugging and hyperparameter searches. Furthermore, we show that the corresponding optimization problem is sound, and provide extensive theoretical work highlighting the deep connections to other distances between distributions."
                },
                {
                    "title": "Density estimation using Real NVP",
                    "abstract": "Unsupervised learning of probabilistic models is a central yet challenging problem in machine learning. Specifically, designing models with tractable learning, sampling, inference and evaluation is crucial in solving this task. We extend the space of such models using real-valued non-volume preserving (real NVP) transformations, a set of powerful invertible and learnable transformations, resulting in an unsupervised learning algorithm with exact log-likelihood computation, exact sampling, exact inference of latent variables, and an interpretable latent space. We demonstrate its ability to model natural images on four datasets through sampling, log-likelihood evaluation and latent variable manipulations."
                },
                {
                    "title": "Improved Inverse Scaling and Squaring Algorithms for the Matrix Logarithm",
                    "abstract": "A popular method for computing the matrix logarithm is the\ninverse scaling and squaring method,\nwhich essentially carries out the steps of the scaling and squaring\nmethod for the matrix exponential in reverse order.\nHere we make several improvements to the method,\nputting its development on a par with our recent version\n[\\emph{SIAM J. Matrix Anal.\\ Appl.}, 31 (2009), pp.\\ 970--989]\nof the scaling and squaring method for the exponential.\nIn particular,\nwe introduce backward error analysis to replace the previous forward\nerror analysis;\nobtain backward error bounds in terms of the quantities\n$\\|A^p\\|^{1/p}$, for several small integer $p$, instead of $\\|A\\|$;\nand use special techniques to compute the argument of the\nPad\\'e approximant more accurately.\nWe derive one algorithm\nthat employs a Schur decomposition,\nand thereby works with triangular matrices,\nand another that requires only matrix multiplications and the solution\nof multiple right-hand side linear systems.\nNumerical experiments show the new algorithms to be generally faster and more\naccurate than their existing counterparts\nand suggest that the Schur-based method is the method of choice for computing\nthe matrix logarithm."
                },
                {
                    "title": "Computational Design of Proteins Targeting the Conserved Stem Region of Influenza Hemagglutinin",
                    "abstract": "Proteins can be designed that bind to specific patches on target proteins to alter their subsequent interactions. We describe a general computational method for designing proteins that bind a surface patch of interest on a target macromolecule. Favorable interactions between disembodied amino acid residues and the target surface are identified and used to anchor de novo designed interfaces. The method was used to design proteins that bind a conserved surface patch on the stem of the influenza hemagglutinin (HA) from the 1918 H1N1 pandemic virus. After affinity maturation, two of the designed proteins, HB36 and HB80, bind H1 and H5 HAs with low nanomolar affinity. Further, HB80 inhibits the HA fusogenic conformational changes induced at low pH. The crystal structure of HB36 in complex with 1918/H1 HA revealed that the actual binding interface is nearly identical to that in the computational design model. Such designed binding proteins may be useful for both diagnostics and therapeutics."
                },
                {
                    "title": "Nanosecond to microsecond protein dynamics probed by magnetic relaxation dispersion of buried water molecules.",
                    "abstract": "Large-scale protein conformational motions on nanosecond-microsecond time scales are important for many biological processes, but remain largely unexplored because of methodological limitations. NMR relaxation methods can access these time scales if protein tumbling is prevented, but the isotropy required for high-resolution solution NMR is then lost. However, if the immobilized protein molecules are randomly oriented, the water 2H and 17O spins relax as in a solution of freely tumbling protein molecules, with the crucial difference that they now sample motions on all time scales up to approximately 100 micros. In particular, the exchange rates of internal water molecules can be determined directly from the 2H or 17O magnetic relaxation dispersion (MRD) profile. This possibility opens up a new window for characterizing the motions of individual internal water molecules as well as the large-scale protein conformational fluctuations that govern the exchange rates of structural water molecules. We introduce and validate this new NMR method by presenting and analyzing an extensive set of 2H and 17O MRD data from cross-linked gels of two model proteins: bovine pancreatic trypsin inhibitor and ubiquitin. We determine residence times and order parameters of four internal water molecules in these proteins and show that they are quantitatively consistent with the information available from crystallography and solution MRD. We also show how slow motions of side-chains bearing labile hydrogens can be monitored by the same approach. Proteins of any size can be studied at physiological hydration levels with this method."
                },
                {
                    "title": "A User's Guide to Measure-Theoretic Probability",
                    "abstract": "Use of Tables, and Applications), with 236 pages of text, 110 pages of tables; and a CD-ROM with software (written in Java) that implements the methods presented in the book. According to the Preface, the authors were motivated to write the book in part because of a consulting project. They \u201cnoticed that the recommended methods [for estimating proportions] are dif\u008e cult to handle for practitioners and, moreover, they are inaccurate and yield often useless results.\u201d I have a problem with these words all appearing in the same sentence. The methods that are at times useless (e.g., an interval based on a normal approximation) are not dif\u008e cult to handle, and methods that can be dif\u008e cult to handle (e.g., an exact con\u008e dence interval without the right software) do not yield useless results. At any rate, the authors decided to basically start from scratch and develop their own theory of estimation. This theory is based on a geometric approach and seems closely related to \u008e ducial intervals as developed in several articles by Neyman. One important feature of the authors\u2019 method is that it allows one to specify bounds on the possible parameter values. This of course can also be done within the existing paradigms; a frequentist can use a constrained maximum likelihood estimator, and a Bayesian can specify a prior with support on only part of the parameter space. I think the authors are more than a little bit presumptuous when they say (p. xiv) that \u201cthe handbook represents a \u008e rst step toward developing statistics as a (natural) science, and further steps must follow.\u201d Later (on p. xvii), we learn that \u201cit may also be looked upon as a \u008e rst step in reconciling classical statistics with Bayes statistics.\u201d Apparently this reconciliation will require the Bayesians to make all the concessions, given the authors\u2019 reverence for Jerzy Neyman. They even display Neyman\u2019s portrait in the user interface for their software. Besides con\u008e dence intervals (which the authors sometimes call \u201cmeasurement intervals\u201d for reasons that are unclear to me), the authors also discuss \u201cprediction regions.\u201d Interestingly, a prediction region does not require any data; rather, it is like an acceptance region for a null hypothesis on the Bernoulli parameter p. Again, we can specify bounds on the parameter values. I do not care much for the terminology, because it suggests that we are predicting a future result rather than testing a hypothesis. The book is a tough read. The writing is hard to follow and somewhat overblown and confrontational, and the notation is thick. There are parts that I think are incorrect; for example, in discussing the exact con\u008e dence interval for a Bernoulli parameter in Section 3.4.1, the claim is made that an inferior interval estimator is obtained if one uses only the sample sum when the sample of individual outcomes is known. In other words, a suf\u008e cient statistic is claimed to be insuf\u008e cient. This error is of no consequence, however, because the authors\u2019 own tables and software require only the sample sum. In the end, I believe that an engineer or scientist will not be making a serious mistake if he or she uses the accompanying software to obtain a con\u008e dence interval. I base this belief on the fact that the few examples I tried yielded results quite close to Bayes credibility intervals based on a uniform prior. It is indeed interesting to have software that can output a critical region for testing a Bernoulli parameter with an interval constraint on its value. However, as Tietjen (1986, p. 37) says, \u201c\u008e ducial intervals are understood by only a few and used by fewer still.\u201d My guess is that that also will be the fate of this work."
                },
                {
                    "title": "Topics in Optimal Transportation",
                    "abstract": "Introduction The Kantorovich duality Geometry of optimal transportation Brenier's polar factorization theorem The Monge-Ampere equation Displacement interpolation and displacement convexity Geometric and Gaussian inequalities The metric side of optimal transportation A differential point of view on optimal transportation Entropy production and transportation inequalities Problems Bibliography Table of short statements Index."
                },
                {
                    "title": "Kinematic Dexterity of Robotic Mechanisms",
                    "abstract": "In this article we develop a mathematical theory for optimizing the kinematic dexterity of robotic mechanisms and obtain a collection of analytical tools for robot design. The performance criteria we consider are workspace volume and dexterity; by the latter we mean the ability to move and apply forces in arbitrary directions as easily as possible. Clearly, dexterity and workspace volume are intrinsic to a mechanism, so that any mathematical formulation of these properties must necessarily be independent of the particular coordinate representation of the kinematics. By regarding the forward kinematics of a mechanism as defining a mapping between Riemannian manifolds, we apply the coordinate-free language of differential geometry to define natural measures of kinematic dexterity and workspace volume. This approach takes into account the geometric and topolog ical structures of the joint and workspaces. We show that the functional associated with harmonic mapping theory provides a natural measure of the kinematic dexterity of a mechan ism. Optimal designs among the basic classes of mechanisms are determined as extrema of this measure. We also examine the qualitative connections between kinematic dexterity and workspace volume."
                },
                {
                    "title": "Polar Factorization and Monotone Rearrangement of Vector-Valued Functions",
                    "abstract": "Given a probability space (X, p) and a bounded domain R in Rd equipped with the Lebesgue measure 1 . I (normalized so that 10 I = I ), it is shown (under additional technical assumptions on X and Q) that for every vector-valued function u E Lp(X, p; Rd) there is a unique \u201cpolar factorization\u201d u = V$.s, where $ is a convex function defined on R and s is a measure-preserving mapping from (X, p) into (Q, I . I), provided that u is nondegenerate, in the sense that p(u-\u2019(E)) = 0 for each Lebesgue negligible subset E of Rd. Through this result, the concepts of polar factorization of real matrices, Helmholtz decomposition of vector fields, and nondecreasing rearrangements of real-valued functions are unified. The Monge-Amgre equation is involved in the polar factorization and the proof relies on the study of an appropriate \u201cMonge-Kantorovich\u201d problem."
                },
                {
                    "title": "Learning Diffusion Bridges on Constrained Domains",
                    "abstract": "Diffusion models have achieved promising results on generative learning recently. However, because diffusion processes are most naturally applied on the unconstrained Euclidean space R d , key challenges arise for developing diffusion based models for learning data on constrained and structured domains. We present a simple and unified framework to achieve this that can be easily adopted to various types of domains, including product spaces of any type (be it bounded/unbounded, continuous/discrete, categorical/ordinal"
                },
                {
                    "title": "POT: Python Optimal Transport",
                    "abstract": "Optimal transport has recently been reintroduced to the machine learning community thanks in part to novel e\ufb03cient optimization procedures allowing for medium to large scale applications. We propose a Python toolbox that implements several key optimal transport ideas for the machine learning community. The toolbox contains implementations of a number of founding works of OT for machine learning such as Sinkhorn algorithm and Wasserstein barycenters, but also provides generic solvers that can be used for conducting novel fundamental research. This toolbox, named POT for Python Optimal Transport, is open source with an MIT license."
                },
                {
                    "title": "Isotropic Distributions for 3-Dimension Rotations and One-Sample Bayes Inference",
                    "abstract": ". . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 CHAPTER"
                },
                {
                    "title": "Structure quality and target parameters",
                    "abstract": "Because protein crystals typically diffract to resolutions of 2\u20133\u00a0A for structures of interest, diffraction data must be supplemented by additional information to be used as parameters for modelling the crystal content. Which information to use, and for what kind of model, depends on the questions to be resolved by the modelled structure. The most obvious choice for target parameters are bond lengths and angles derived from atomic resolution structures. This article describes a set of bond and angle target parameters derived from small-molecule structures in the Cambridge Structural Database and their application to the refinement of protein structure models. Other potential sources of target parameters are also briefly discussed."
                },
                {
                    "title": "OPTIMAL CALCULATION OF ODF WITH THE CANONICAL NORMAL DISTRIBUTION ON THE ROTATION GROUP",
                    "abstract": "Macroscopic physical properties of most polycrystalline materials are controlled by orientation distribution of their grains. The orientation distribution function (ODF) of a polycrystal is seldom if ever determined directly from an experiment. Usually experimental data are represented by a set of pole figures (PFs), these latter are some integral projections of the ODF. The main problem of quantitative texture analysis is to recover ODF from its corresponding PFs. With any set of PFs the solution of this problem is non-unique. That is why some assumptions about ODF structure are necessary. We consider ODF as superposition of the canonical normal distribution (CND) on the rotation group SO(3)."
                },
                {
                    "title": "Normal Distribution on the Rotation Group So(3)",
                    "abstract": "We study the normal distribution on the rotation group SO(3). If we take as the normal \ndistribution on the rotation group the distribution defined by the central limit theorem in \nParthasarathy (1964) rather than the distribution with density analogous to the normal \ndistribution in Eucledian space, then its density will be different from the usual (1/2\u03c0\u03c3) \nexp\u2061(\u2212(x\u2212m)2/2\u03c32) one. Nevertheless, many properties of this distribution will be \nanalogous to the normal distribution in the Eucledian space. It is possible to obtain \nexplicit expressions for density of normal distribution only for special cases. One of these \ncases is the circular normal distribution."
                },
                {
                    "title": "Making antibodies by phage display technology.",
                    "abstract": "Antibody fragments of predetermined binding specificity have recently been constructed from repertoires of antibody V genes, bypassing hybridoma technology and even immunization. The V gene repertoires are harvested from populations of lymphocytes, or assembled in vitro, and cloned for display of associated heavy and light chain variable domains on the surface of filamentous bacteriophage. Rare phage are selected from the repertoire by binding to antigen; soluble antibody fragments are expressed from infected bacteria; and the affinity of binding of selected antibodies is improved by mutation. The process mimics immune selection, and antibodies with many different binding specificities have been isolated from the same phage repertoire. Thus human antibody fragments have been isolated with specificities against both foreign and self antigens, including haptens, carbohydrates, secreted and cell surface proteins, viral coat proteins, and intracellular antigens from the lumen of the endoplasmic reticulum and the nucleus. Such antibodies have potential as reagents for research and in therapy."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively sample from the distribution of protein backbones represented in the special Euclidean group SE(3)\u207f to design novel proteins with specified structural and functional properties?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of protein engineering, as it can lead to the rational design of proteins that address significant global health challenges, such as developing targeted therapies for diseases like cancer and viral infections. By improving our ability to design proteins, this research could enhance drug discovery processes, leading to more effective treatments and a deeper understanding of protein functions. Furthermore, it could inspire future research in computational biology and bioinformatics, fostering innovations in related fields.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complex geometric structure of protein backbones, which are represented in a high-dimensional space governed by the symmetries of SE(3)\u207f. Naive approaches may fail due to the intricacies of sampling from this manifold, as they often overlook the underlying geometric properties and may require computationally expensive simulations or numerical methods that are not efficient. Additionally, the need to accurately model the interactions and folding dynamics of proteins adds layers of complexity that must be addressed.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has not fully leveraged the rich geometric structure of SE(3)\u207f, often relying on generalized generative models that do not specifically cater to the unique properties of protein backbones. Barriers such as the reliance on expensive numerical methods, like simulating Stochastic Differential Equations (SDEs), have hindered progress. Our approach, FoldFlow, differs by utilizing Conditional Flow Matching (CFM) to directly learn from the data without requiring these costly simulations, thus providing a more efficient and effective solution.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of FoldFlow, a family of continuous normalizing flows (CNFs) specifically designed for distributions on SE(3)\u207f. We will employ Conditional Flow Matching (CFM) to learn time-dependent vector fields that generate probability paths for protein backbones. The dataset will consist of known protein structures, and we will evaluate our models using metrics that assess the quality of the generated protein backbones in terms of structural fidelity and functional relevance. We expect our"
            }
        },
        "author_data": {
            "7fbbbd8e-f776-4a32-950f-764248e1e490": {
                "pk": "7fbbbd8e-f776-4a32-950f-764248e1e490",
                "name": "Avishek Joey Bose",
                "collaborators": [
                    "Michael M. Bronstein",
                    "Alexander Tong",
                    "Gauthier Gidel",
                    "Jarrid Rector-Brooks",
                    "Cheng-Hao Liu",
                    "Sarthak Mittal",
                    "Y. Bengio",
                    "Pablo Lemos",
                    "Tara Akhound-Sadegh",
                    "Quentin Bertrand"
                ],
                "domain": [
                    "Generative Modeling",
                    "Protein Design",
                    "Deep Learning",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Steering Masked Discrete Diffusion Models via Discrete Denoising Posterior Prediction",
                        "abstract": "Generative modeling of discrete data underlies important applications spanning text-based agents like ChatGPT to the design of the very building blocks of life in protein sequences. However, application domains need to exert control over the generated data by steering the generative process - typically via RLHF - to satisfy a specified property, reward, or affinity metric. In this paper, we study the problem of steering Masked Diffusion Models (MDMs), a recent class of discrete diffusion models that offer a compelling alternative to traditional autoregressive models. We introduce Discrete Denoising Posterior Prediction (DDPP), a novel framework that casts the task of steering pre-trained MDMs as a problem of probabilistic inference by learning to sample from a target Bayesian posterior. Our DDPP framework leads to a family of three novel objectives that are all simulation-free, and thus scalable while applying to general non-differentiable reward functions. Empirically, we instantiate DDPP by steering MDMs to perform class-conditional pixel-level image modeling, RLHF-based alignment of MDMs using text-based rewards, and finetuning protein language models to generate more diverse secondary structures and shorter proteins. We substantiate our designs via wet-lab validation, where we observe transient expression of reward-optimized protein sequences."
                    },
                    {
                        "title": "Sequence-Augmented SE(3)-Flow Matching For Conditional Protein Backbone Generation",
                        "abstract": "Proteins are essential for almost all biological processes and derive their diverse functions from complex 3D structures, which are in turn determined by their amino acid sequences. In this paper, we exploit the rich biological inductive bias of amino acid sequences and introduce FoldFlow-2, a novel sequence-conditioned SE(3)-equivariant flow matching model for protein structure generation. FoldFlow-2 presents substantial new architectural features over the previous FoldFlow family of models including a protein large language model to encode sequence, a new multi-modal fusion trunk that combines structure and sequence representations, and a geometric transformer based decoder. To increase diversity and novelty of generated samples -- crucial for de-novo drug design -- we train FoldFlow-2 at scale on a new dataset that is an order of magnitude larger than PDB datasets of prior works, containing both known proteins in PDB and high-quality synthetic structures achieved through filtering. We further demonstrate the ability to align FoldFlow-2 to arbitrary rewards, e.g. increasing secondary structures diversity, by introducing a Reinforced Finetuning (ReFT) objective. We empirically observe that FoldFlow-2 outperforms previous state-of-the-art protein structure-based generative models, improving over RFDiffusion in terms of unconditional generation across all metrics including designability, diversity, and novelty across all protein lengths, as well as exhibiting generalization on the task of equilibrium conformation sampling. Finally, we demonstrate that a fine-tuned FoldFlow-2 makes progress on challenging conditional design tasks such as designing scaffolds for the VHH nanobody."
                    },
                    {
                        "title": "Self-Consuming Generative Models with Curated Data Provably Optimize Human Preferences",
                        "abstract": "The rapid progress in generative models has resulted in impressive leaps in generation quality, blurring the lines between synthetic and real data. Web-scale datasets are now prone to the inevitable contamination by synthetic data, directly impacting the training of future generated models. Already, some theoretical results on self-consuming generative models (a.k.a., iterative retraining) have emerged in the literature, showcasing that either model collapse or stability could be possible depending on the fraction of generated data used at each retraining step. However, in practice, synthetic data is often subject to human feedback and curated by users before being used and uploaded online. For instance, many interfaces of popular text-to-image generative models, such as Stable Diffusion or Midjourney, produce several variations of an image for a given query which can eventually be curated by the users. In this paper, we theoretically study the impact of data curation on iterated retraining of generative models and show that it can be seen as an \\emph{implicit preference optimization mechanism}. However, unlike standard preference optimization, the generative model does not have access to the reward function or negative samples needed for pairwise comparisons. Moreover, our study doesn't require access to the density function, only to samples. We prove that, if the data is curated according to a reward model, then the expected reward of the iterative retraining procedure is maximized. We further provide theoretical results on the stability of the retraining loop when using a positive fraction of real data at each step. Finally, we conduct illustrative experiments on both synthetic datasets and on CIFAR10 showing that such a procedure amplifies biases of the reward model."
                    },
                    {
                        "title": "Metric Flow Matching for Smooth Interpolations on the Data Manifold",
                        "abstract": "Matching objectives underpin the success of modern generative models and rely on constructing conditional paths that transform a source distribution into a target distribution. Despite being a fundamental building block, conditional paths have been designed principally under the assumption of Euclidean geometry, resulting in straight interpolations. However, this can be particularly restrictive for tasks such as trajectory inference, where straight paths might lie outside the data manifold, thus failing to capture the underlying dynamics giving rise to the observed marginals. In this paper, we propose Metric Flow Matching (MFM), a novel simulation-free framework for conditional flow matching where interpolants are approximate geodesics learned by minimizing the kinetic energy of a data-induced Riemannian metric. This way, the generative model matches vector fields on the data manifold, which corresponds to lower uncertainty and more meaningful interpolations. We prescribe general metrics to instantiate MFM, independent of the task, and test it on a suite of challenging problems including LiDAR navigation, unpaired image translation, and modeling cellular dynamics. We observe that MFM outperforms the Euclidean baselines, particularly achieving SOTA on single-cell trajectory prediction."
                    },
                    {
                        "title": "Fisher Flow Matching for Generative Modeling over Discrete Data",
                        "abstract": "Generative modeling over discrete data has recently seen numerous success stories, with applications spanning language modeling, biological sequence design, and graph-structured molecular data. The predominant generative modeling paradigm for discrete data is still autoregressive, with more recent alternatives based on diffusion or flow-matching falling short of their impressive performance in continuous data settings, such as image or video generation. In this work, we introduce Fisher-Flow, a novel flow-matching model for discrete data. Fisher-Flow takes a manifestly geometric perspective by considering categorical distributions over discrete data as points residing on a statistical manifold equipped with its natural Riemannian metric: the $\\textit{Fisher-Rao metric}$. As a result, we demonstrate discrete data itself can be continuously reparameterised to points on the positive orthant of the $d$-hypersphere $\\mathbb{S}^d_+$, which allows us to define flows that map any source distribution to target in a principled manner by transporting mass along (closed-form) geodesics of $\\mathbb{S}^d_+$. Furthermore, the learned flows in Fisher-Flow can be further bootstrapped by leveraging Riemannian optimal transport leading to improved training dynamics. We prove that the gradient flow induced by Fisher-Flow is optimal in reducing the forward KL divergence. We evaluate Fisher-Flow on an array of synthetic and diverse real-world benchmarks, including designing DNA Promoter, and DNA Enhancer sequences. Empirically, we find that Fisher-Flow improves over prior diffusion and flow-matching models on these benchmarks."
                    },
                    {
                        "title": "Predicting Drug Effects from High-Dimensional, Asymmetric Drug Datasets by Using Graph Neural Networks: A Comprehensive Analysis of Multitarget Drug Effect Prediction",
                        "abstract": "Graph neural networks (GNNs) have emerged as one of the most effective ML techniques for drug effect prediction from drug molecular graphs. Despite having immense potential, GNN models lack performance when using datasets that contain high-dimensional, asymmetrically co-occurrent drug effects as targets with complex correlations between them. Training individual learning models for each drug effect and incorporating every prediction result for a wide spectrum of drug effects are impractical. Therefore, an opportunity exists to address this challenge as multitarget prediction problems and predict all drug effects at a time. We developed standard and hybrid GNNs to perform two separate tasks: multiregression for continuous values and multilabel classification for categorical values contained in our datasets. Because multilabel classification makes the target data even more sparse and introduces asymmetric label co-occurrence, learning these models becomes difficult and heavily impacts the GNN's performance. To address these challenges, we propose a new data oversampling technique to improve multilabel classification performances on all the given imbalanced molecular graph datasets. Using the technique, we improve the data imbalance ratio of the drug effects while protecting the datasets' integrity. Finally, we evaluate the multilabel classification performance of the best-performing hybrid GNN model on all the oversampled datasets obtained from the proposed oversampling technique. In all the evaluation metrics (i.e., precision, recall, and F1 score), this model significantly outperforms other ML models, including GNN models when they are trained on the original datasets or oversampled datasets with MLSMOTE, which is a well-known oversampling technique."
                    },
                    {
                        "title": "Iterated Denoising Energy Matching for Sampling from Boltzmann Densities",
                        "abstract": "Efficiently generating statistically independent samples from an unnormalized probability distribution, such as equilibrium samples of many-body systems, is a foundational problem in science. In this paper, we propose Iterated Denoising Energy Matching (iDEM), an iterative algorithm that uses a novel stochastic score matching objective leveraging solely the energy function and its gradient -- and no data samples -- to train a diffusion-based sampler. Specifically, iDEM alternates between (I) sampling regions of high model density from a diffusion-based sampler and (II) using these samples in our stochastic matching objective to further improve the sampler. iDEM is scalable to high dimensions as the inner matching objective, is simulation-free, and requires no MCMC samples. Moreover, by leveraging the fast mode mixing behavior of diffusion, iDEM smooths out the energy landscape enabling efficient exploration and learning of an amortized sampler. We evaluate iDEM on a suite of tasks ranging from standard synthetic energy functions to invariant $n$-body particle systems. We show that the proposed approach achieves state-of-the-art performance on all metrics and trains $2-5\\times$ faster, which allows it to be the first method to train using energy on the challenging $55$-particle Lennard-Jones system."
                    },
                    {
                        "title": "Comparative Study of Large Language Model Architectures on Frontier",
                        "abstract": "Large language models (LLMs) have garnered significant attention in both the AI community and beyond. Among these, the Generative Pre-trained Transformer (GPT) has emerged as the dominant architecture, spawning numerous variants. However, these variants have undergone pre-training under diverse conditions, including variations in input data, data preprocessing, and training methodologies, resulting in a lack of controlled comparative studies. Here we meticulously examine two prominent open-sourced GPT architectures, GPT-NeoX and LLaMA, leveraging the computational power of Frontier, the world\u2019s first Exascale supercomputer. Employing the same materials science text corpus and a comprehensive end-to-end pipeline, we conduct a comparative analysis of their training and downstream performance. Our efforts culminate in achieving state-of-the-art performance on a challenging materials science benchmark. Furthermore, we investigate the computation and energy efficiency, and propose a computationally efficient method for architecture design. To our knowledge, these pre-trained models represent the largest available for materials science. Our findings provide practical guidance for building LLMs on HPC platforms."
                    },
                    {
                        "title": "On the Stability of Iterative Retraining of Generative Models on their own Data",
                        "abstract": "Deep generative models have made tremendous progress in modeling complex data, often exhibiting generation quality that surpasses a typical human's ability to discern the authenticity of samples. Undeniably, a key driver of this success is enabled by the massive amounts of web-scale data consumed by these models. Due to these models' striking performance and ease of availability, the web will inevitably be increasingly populated with synthetic content. Such a fact directly implies that future iterations of generative models will be trained on both clean and artificially generated data from past models. In this paper, we develop a framework to rigorously study the impact of training generative models on mixed datasets -- from classical training on real data to self-consuming generative models trained on purely synthetic data. We first prove the stability of iterative training under the condition that the initial generative models approximate the data distribution well enough and the proportion of clean training data (w.r.t. synthetic data) is large enough. We empirically validate our theory on both synthetic and natural images by iteratively training normalizing flows and state-of-the-art diffusion models on CIFAR10 and FFHQ."
                    },
                    {
                        "title": "Feature Likelihood Divergence: Evaluating the Generalization of Generative Models Using Samples",
                        "abstract": "The past few years have seen impressive progress in the development of deep generative models capable of producing high-dimensional, complex, and photo-realistic data. However, current methods for evaluating such models remain incomplete: standard likelihood-based metrics do not always apply and rarely correlate with perceptual fidelity, while sample-based metrics, such as FID, are insensitive to overfitting, i.e., inability to generalize beyond the training set. To address these limitations, we propose a new metric called the Feature Likelihood Divergence (FLD), a parametric sample-based metric that uses density estimation to provide a comprehensive trichotomic evaluation accounting for novelty (i.e., different from the training samples), fidelity, and diversity of generated samples. We empirically demonstrate the ability of FLD to identify overfitting problem cases, even when previously proposed metrics fail. We also extensively evaluate FLD on various image datasets and model classes, demonstrating its ability to match intuitions of previous metrics like FID while offering a more comprehensive evaluation of generative models. Code is available at https://github.com/marcojira/fld."
                    }
                ]
            },
            "3311bb3a-0e11-421b-b904-06208e2163ac": {
                "pk": "3311bb3a-0e11-421b-b904-06208e2163ac",
                "name": "Tara Akhound-Sadegh",
                "collaborators": [
                    "Siamak Ravanbakhsh",
                    "Pablo Lemos",
                    "Cheng-Hao Liu",
                    "Jarrid Rector-Brooks",
                    "Alexander Tong",
                    "A. Bose",
                    "Laurence Perreault-Levasseur",
                    "Johannes Brandstetter",
                    "Max Welling",
                    "Guillaume Huguet"
                ],
                "domain": [
                    "Protein Structure Prediction",
                    "Energy-Based Models",
                    "Physics-Informed Neural Networks",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Sequence-Augmented SE(3)-Flow Matching For Conditional Protein Backbone Generation",
                        "abstract": "Proteins are essential for almost all biological processes and derive their diverse functions from complex 3D structures, which are in turn determined by their amino acid sequences. In this paper, we exploit the rich biological inductive bias of amino acid sequences and introduce FoldFlow-2, a novel sequence-conditioned SE(3)-equivariant flow matching model for protein structure generation. FoldFlow-2 presents substantial new architectural features over the previous FoldFlow family of models including a protein large language model to encode sequence, a new multi-modal fusion trunk that combines structure and sequence representations, and a geometric transformer based decoder. To increase diversity and novelty of generated samples -- crucial for de-novo drug design -- we train FoldFlow-2 at scale on a new dataset that is an order of magnitude larger than PDB datasets of prior works, containing both known proteins in PDB and high-quality synthetic structures achieved through filtering. We further demonstrate the ability to align FoldFlow-2 to arbitrary rewards, e.g. increasing secondary structures diversity, by introducing a Reinforced Finetuning (ReFT) objective. We empirically observe that FoldFlow-2 outperforms previous state-of-the-art protein structure-based generative models, improving over RFDiffusion in terms of unconditional generation across all metrics including designability, diversity, and novelty across all protein lengths, as well as exhibiting generalization on the task of equilibrium conformation sampling. Finally, we demonstrate that a fine-tuned FoldFlow-2 makes progress on challenging conditional design tasks such as designing scaffolds for the VHH nanobody."
                    },
                    {
                        "title": "Iterated Denoising Energy Matching for Sampling from Boltzmann Densities",
                        "abstract": "Efficiently generating statistically independent samples from an unnormalized probability distribution, such as equilibrium samples of many-body systems, is a foundational problem in science. In this paper, we propose Iterated Denoising Energy Matching (iDEM), an iterative algorithm that uses a novel stochastic score matching objective leveraging solely the energy function and its gradient -- and no data samples -- to train a diffusion-based sampler. Specifically, iDEM alternates between (I) sampling regions of high model density from a diffusion-based sampler and (II) using these samples in our stochastic matching objective to further improve the sampler. iDEM is scalable to high dimensions as the inner matching objective, is simulation-free, and requires no MCMC samples. Moreover, by leveraging the fast mode mixing behavior of diffusion, iDEM smooths out the energy landscape enabling efficient exploration and learning of an amortized sampler. We evaluate iDEM on a suite of tasks ranging from standard synthetic energy functions to invariant $n$-body particle systems. We show that the proposed approach achieves state-of-the-art performance on all metrics and trains $2-5\\times$ faster, which allows it to be the first method to train using energy on the challenging $55$-particle Lennard-Jones system."
                    },
                    {
                        "title": "Sampling from Energy-based Policies using Diffusion",
                        "abstract": "Energy-based policies offer a flexible framework for modeling complex, multimodal behaviors in reinforcement learning (RL). In maximum entropy RL, the optimal policy is a Boltzmann distribution derived from the soft Q-function, but direct sampling from this distribution in continuous action spaces is computationally intractable. As a result, existing methods typically use simpler parametric distributions, like Gaussians, for policy representation - limiting their ability to capture the full complexity of multimodal action distributions. In this paper, we introduce a diffusion-based approach for sampling from energy-based policies, where the negative Q-function defines the energy function. Based on this approach, we propose an actor-critic method called Diffusion Q-Sampling (DQS) that enables more expressive policy representations, allowing stable learning in diverse environments. We show that our approach enhances exploration and captures multimodal behavior in continuous control tasks, addressing key limitations of existing methods."
                    },
                    {
                        "title": "Lie Point Symmetry and Physics Informed Networks",
                        "abstract": "Symmetries have been leveraged to improve the generalization of neural networks through different mechanisms from data augmentation to equivariant architectures. However, despite their potential, their integration into neural solvers for partial differential equations (PDEs) remains largely unexplored. We explore the integration of PDE symmetries, known as Lie point symmetries, in a major family of neural solvers known as physics-informed neural networks (PINNs). We propose a loss function that informs the network about Lie point symmetries in the same way that PINN models try to enforce the underlying PDE through a loss function. Intuitively, our symmetry loss ensures that the infinitesimal generators of the Lie group conserve the PDE solutions. Effectively, this means that once the network learns a solution, it also learns the neighbouring solutions generated by Lie point symmetries. Empirical evaluations indicate that the inductive bias introduced by the Lie point symmetries of the PDEs greatly boosts the sample efficiency of PINNs."
                    },
                    {
                        "title": "Physics-Informed Transformer Networks",
                        "abstract": "Physics-informed neural networks (PINNs) have been recognized as a viable alternative to conventional numerical solvers for Partial Differential Equations (PDEs). The main appeal of PINNs is that since they directly enforce the PDE equation, one does not require access to costly ground truth solutions for training the model. However, a key challenge is their limited generalization across varied initial conditions. Addressing this, our study presents a novel Physics-Informed Transformer (PIT) model for learning the solution operator for PDEs. Using the attention mechanism, PIT learns to leverage the relationships between its initial condition and query points, resulting in a significant improvement in generalization. Moreover, in contrast to existing physics-informed networks, our model is invariant to the discretization of the input domain, providing great flexibility in problem specification and training. We validated our proposed method on the 1D Burgers\u2019 and the 2D Heat equations, demonstrating notable improvement over standard PINN models for operator learning with negligible computational overhead."
                    },
                    {
                        "title": "Lie Point Symmetry and Physics Informed Neural Networks",
                        "abstract": "Physics-informed neural networks (PINNs) are computationally ef\ufb01cient alternatives to traditional partial differential equation (PDE) solvers. However, their reliability is dependent on the accuracy of the trained neural network. In this work, we introduce a mechanism for leveraging the symmetries of a given PDE to improve PINN performance. In particular, we propose a loss function that informs the network about Lie point symmetries, similar to how traditional PINN models try to enforce the underlying PDE. Intuitively, our symmetry loss ensures that in\ufb01nitesimal generators of the Lie group preserve solutions of the PDE. Effectively, this means that once the network learns a solution, it also learns the neighbouring solutions generated by Lie point symmetries. Our results con\ufb01rm that Lie point symmetries of the respective PDEs are an effective inductive bias for PINNs and can lead to a signi\ufb01cant increase in sample ef\ufb01ciency."
                    }
                ]
            },
            "1603ad29-dac8-4382-a4ea-385f7b235af1": {
                "pk": "1603ad29-dac8-4382-a4ea-385f7b235af1",
                "name": "Guillaume Huguet",
                "collaborators": [
                    "Alexander Tong",
                    "Smita Krishnaswamy",
                    "Guy Wolf",
                    "Chen Liu",
                    "Danqi Liao",
                    "Yanlei Zhang",
                    "Kilian Fatras",
                    "Ke Xu",
                    "Liangbo L. Shen",
                    "Zilong Wang"
                ],
                "domain": [
                    "Generative Modeling",
                    "Protein Structure Prediction",
                    "Medical Imaging",
                    "Single-Cell Analysis"
                ],
                "publications": [
                    {
                        "title": "Sequence-Augmented SE(3)-Flow Matching For Conditional Protein Backbone Generation",
                        "abstract": "Proteins are essential for almost all biological processes and derive their diverse functions from complex 3D structures, which are in turn determined by their amino acid sequences. In this paper, we exploit the rich biological inductive bias of amino acid sequences and introduce FoldFlow-2, a novel sequence-conditioned SE(3)-equivariant flow matching model for protein structure generation. FoldFlow-2 presents substantial new architectural features over the previous FoldFlow family of models including a protein large language model to encode sequence, a new multi-modal fusion trunk that combines structure and sequence representations, and a geometric transformer based decoder. To increase diversity and novelty of generated samples -- crucial for de-novo drug design -- we train FoldFlow-2 at scale on a new dataset that is an order of magnitude larger than PDB datasets of prior works, containing both known proteins in PDB and high-quality synthetic structures achieved through filtering. We further demonstrate the ability to align FoldFlow-2 to arbitrary rewards, e.g. increasing secondary structures diversity, by introducing a Reinforced Finetuning (ReFT) objective. We empirically observe that FoldFlow-2 outperforms previous state-of-the-art protein structure-based generative models, improving over RFDiffusion in terms of unconditional generation across all metrics including designability, diversity, and novelty across all protein lengths, as well as exhibiting generalization on the task of equilibrium conformation sampling. Finally, we demonstrate that a fine-tuned FoldFlow-2 makes progress on challenging conditional design tasks such as designing scaffolds for the VHH nanobody."
                    },
                    {
                        "title": "ImageFlowNet: Forecasting Multiscale Trajectories of Disease Progression with Irregularly-Sampled Longitudinal Medical Images",
                        "abstract": "The forecasting of disease progression from images is a holy grail for clinical decision making. However, this task is complicated by the inherent high dimensionality, temporal sparsity and sampling irregularity in longitudinal image acquisitions. Existing methods often rely on extracting hand-crafted features and performing time-series analysis in this vector space, leading to a loss of rich spatial information within the images. To overcome these challenges, we introduce ImageFlowNet, a novel framework that learns latent-space flow fields that evolve multiscale representations in joint embedding spaces using neural ODEs and SDEs to model disease progression in the image domain. Notably, Image-FlowNet learns multiscale joint representation spaces by combining cohorts of patients together so that information can be transferred between the patient samples. The dynamics then provide plausible trajectories of progression, with the SDE providing alternative trajectories from the same starting point. We provide theoretical insights that support our formulation of ODEs, and motivate our regu-larizations involving high-level visual features, latent space organization, and trajectory smoothness. We then demonstrate ImageFlowNet\u2019s effectiveness through empirical evaluations on three longitudinal medical image datasets depicting progression in retinal geographic"
                    },
                    {
                        "title": "Geometry-Aware Generative Autoencoders for Warped Riemannian Metric Learning and Generative Modeling on Data Manifolds",
                        "abstract": "Rapid growth of high-dimensional datasets in fields such as single-cell RNA sequencing and spatial genomics has led to unprecedented opportunities for scientific discovery, but it also presents unique computational and statistical challenges. Traditional methods struggle with geometry-aware data generation, interpolation along meaningful trajectories, and transporting populations via feasible paths. To address these issues, we introduce Geometry-Aware Generative Autoencoder (GAGA), a novel framework that combines extensible manifold learning with generative modeling. GAGA constructs a neural network embedding space that respects the intrinsic geometries discovered by manifold learning and learns a novel warped Riemannian metric on the data space. This warped metric is derived from both the points on the data manifold and negative samples off the manifold, allowing it to characterize a meaningful geometry across the entire latent space. Using this metric, GAGA can uniformly sample points on the manifold, generate points along geodesics, and interpolate between populations across the learned manifold using geodesic-guided flows. GAGA shows competitive performance in simulated and real-world datasets, including a 30% improvement over the state-of-the-art methods in single-cell population-level trajectory inference."
                    },
                    {
                        "title": "ImageFlowNet: Forecasting Multiscale Image-Level Trajectories of Disease Progression with Irregularly-Sampled Longitudinal Medical Images",
                        "abstract": "Advances in medical imaging technologies have enabled the collection of longitudinal images, which involve repeated scanning of the same patients over time, to monitor disease progression. However, predictive modeling of such data remains challenging due to high dimensionality, irregular sampling, and data sparsity. To address these issues, we propose ImageFlowNet, a novel model designed to forecast disease trajectories from initial images while preserving spatial details. ImageFlowNet first learns multiscale joint representation spaces across patients and time points, then optimizes deterministic or stochastic flow fields within these spaces using a position-parameterized neural ODE/SDE framework. The model leverages a UNet architecture to create robust multiscale representations and mitigates data scarcity by combining knowledge from all patients. We provide theoretical insights that support our formulation of ODEs, and motivate our regularizations involving high-level visual features, latent space organization, and trajectory smoothness. We validate ImageFlowNet on three longitudinal medical image datasets depicting progression in geographic atrophy, multiple sclerosis, and glioblastoma, demonstrating its ability to effectively forecast disease progression and outperform existing methods. Our contributions include the development of ImageFlowNet, its theoretical underpinnings, and empirical validation on real-world datasets. The official implementation is available at https://github.com/KrishnaswamyLab/ImageFlowNet."
                    },
                    {
                        "title": "Assessing Neural Network Representations During Training Using Noise-Resilient Diffusion Spectral Entropy",
                        "abstract": "Entropy and mutual information in neural networks provide rich information on the learning process, but they have proven difficult to compute reliably in high dimensions. Indeed, in noisy and high-dimensional data, traditional estimates in ambient dimensions approach a fixed entropy and are prohibitively hard to compute. To address these issues, we leverage data geometry to access the underlying manifold and reliably compute these information-theoretic measures. Specifically, we define diffusion spectral entropy (DSE) in neural representations of a dataset as well as diffusion spectral mutual information (DSMI) between different variables representing data. First, we show that they form noise-resistant measures of intrinsic dimensionality and relationship strength in high-dimensional simulated data that outperform classic Shannon entropy, nonparametric estimation, and mutual information neural estimation (MINE). We then study the evolution of representations in classification networks with supervised learning, self-supervision, or overfitting. We observe that (1) DSE of neural representations increases during training; (2) DSMI with the class label increases during generalizable learning but stays stagnant during overfitting; (3) DSMI with the input signal shows differing trends: on MNIST it increases, while on CIFAR-10 and STL-10 it decreases. Finally, we show that DSE can be used to guide better network initialization and that DSMI can be used to predict downstream classification accuracy across 962 models on ImageNet."
                    },
                    {
                        "title": "Geodesic Sinkhorn: optimal transport for high-dimensional datasets",
                        "abstract": ". Understanding the dynamics and reactions of cells from population snapshots is a major challenge in single-cell transcriptomics. Here, we present Geodesic Sinkhorn, a method for interpolating populations along a data manifold that leverages existing kernels developed for single-cell dimensionality reduction and visualization methods. Our Geodesic Sinkhorn method uses a heat-geodesic ground distance that, as compared to Euclidean ground distances, is more accurate for interpolating single-cell dynamics on a wide variety of datasets and signi\ufb01cantly speeds up the computation for sparse kernels. We \ufb01rst apply Geodesic Sinkhorn to 10 single-cell transcriptomics time series interpolation datasets as a drop-in replacement for existing interpolation methods where it outperforms on all datasets, showing its e\ufb00ectiveness in modeling cell dynamics. Second, we show how to e\ufb03ciently approximate the operator with polynomial kernels allowing us to improve scaling to large datasets. Finally, we de\ufb01ne the conditional Wasserstein-average treatment e\ufb00ect and show how it can elucidate the treatment e\ufb00ect on single-cell populations on a drug screen."
                    },
                    {
                        "title": "Embedding Signals on Knowledge Graphs with Unbalanced Diffusion Earth Mover's Distance",
                        "abstract": "In modern relational machine learning it is common to encounter large graphs that arise via interactions or similarities between observations in many domains. Further"
                    },
                    {
                        "title": "Geometry-Aware Autoencoders for Metric Learning and Generative Modeling on Data Manifolds",
                        "abstract": "Non-linear dimensionality reduction methods have proven successful at learning low-dimensional representations of high-dimensional point clouds on or near data manifolds. However, existing methods are not easily extensible\u2014 for large datasets, it is prohibitively expensive to add new points to these embeddings. As a result, it is very difficult to use existing embeddings generatively, to sample new points on and along these manifolds. In this paper, we propose GAGA (geometry-aware generative autoencoders) a framework which merges the power of generative deep learning with non-linear manifold learning by: 1) learning generalizable geometry-aware neural network embeddings based on non-linear dimensionality reduction methods like PHATE and diffusion maps, 2) deriving a non-euclidean pullback metric on the data space to generate points faithfully along data manifold geodesics, and 3) learning a flow on the manifold that allows us to transport populations. We provide illustration on easily-interpretable synthetic datasets and showcase results on simulated and real single cell datasets. We show that the geodesic-based generation can be especially important for scientific datasets where the manifold represents a state space and geodesics can represent dynamics of entities over this space."
                    },
                    {
                        "title": "Simulation-Free Schr\u00a8odinger Bridges via Score and Flow Matching",
                        "abstract": "We present simulation-free score and flow matching ([SF] 2 M), a simulation-free objective for inferring stochastic dynamics given unpaired source and target samples drawn from arbitrary distributions. Our method generalizes both the score-matching loss used in the training of diffusion models and the recently proposed flow matching loss used in the training of continuous normalizing flows. [SF] 2 M interprets continuous-time stochastic generative modeling as a Schr \u00a8 odinger bridge (SB) problem. It relies on static entropy-regularized optimal transport, or a minibatch approximation, to efficiently learn the SB without simulating the learned stochastic process. We find that [SF] 2 M is more efficient and gives more accurate solutions to the SB problem than simulation-based methods from prior work. Finally, we apply [SF] 2 M to the problem of learning cell dynamics from snapshot data. Notably, [SF] 2 M is the first method to accurately model cell dynamics in high dimensions and can recover known gene regulatory networks from simulated data. Code: https://github.com/atong01/ conditional-flow-matching ."
                    }
                ]
            },
            "d9029a8f-bb43-4a61-878f-45e93dacfd1a": {
                "pk": "d9029a8f-bb43-4a61-878f-45e93dacfd1a",
                "name": "Kilian Fatras",
                "collaborators": [
                    "N. Courty",
                    "R\u00e9mi Flamary",
                    "Alexander Tong",
                    "Jarrid Rector-Brooks",
                    "Alexia Jolicoeur-Martineau",
                    "Nikolay Malkin",
                    "Yanlei Zhang",
                    "G. Huguet",
                    "Guy Wolf",
                    "Y. Bengio"
                ],
                "domain": [
                    "Optimal Transport",
                    "Generative Models",
                    "Adversarial Learning",
                    "Domain Adaptation"
                ],
                "publications": [
                    {
                        "title": "Sequence-Augmented SE(3)-Flow Matching For Conditional Protein Backbone Generation",
                        "abstract": "Proteins are essential for almost all biological processes and derive their diverse functions from complex 3D structures, which are in turn determined by their amino acid sequences. In this paper, we exploit the rich biological inductive bias of amino acid sequences and introduce FoldFlow-2, a novel sequence-conditioned SE(3)-equivariant flow matching model for protein structure generation. FoldFlow-2 presents substantial new architectural features over the previous FoldFlow family of models including a protein large language model to encode sequence, a new multi-modal fusion trunk that combines structure and sequence representations, and a geometric transformer based decoder. To increase diversity and novelty of generated samples -- crucial for de-novo drug design -- we train FoldFlow-2 at scale on a new dataset that is an order of magnitude larger than PDB datasets of prior works, containing both known proteins in PDB and high-quality synthetic structures achieved through filtering. We further demonstrate the ability to align FoldFlow-2 to arbitrary rewards, e.g. increasing secondary structures diversity, by introducing a Reinforced Finetuning (ReFT) objective. We empirically observe that FoldFlow-2 outperforms previous state-of-the-art protein structure-based generative models, improving over RFDiffusion in terms of unconditional generation across all metrics including designability, diversity, and novelty across all protein lengths, as well as exhibiting generalization on the task of equilibrium conformation sampling. Finally, we demonstrate that a fine-tuned FoldFlow-2 makes progress on challenging conditional design tasks such as designing scaffolds for the VHH nanobody."
                    },
                    {
                        "title": "Generating and Imputing Tabular Data via Diffusion and Flow-based Gradient-Boosted Trees",
                        "abstract": "Tabular data is hard to acquire and is subject to missing values. This paper introduces a novel approach for generating and imputing mixed-type (continuous and categorical) tabular data utilizing score-based diffusion and conditional flow matching. In contrast to prior methods that rely on neural networks to learn the score function or the vector field, we adopt XGBoost, a widely used Gradient-Boosted Tree (GBT) technique. To test our method, we build one of the most extensive benchmarks for tabular data generation and imputation, containing 27 diverse datasets and 9 metrics. Through empirical evaluation across the benchmark, we demonstrate that our approach outperforms deep-learning generation methods in data generation tasks and remains competitive in data imputation. Notably, it can be trained in parallel using CPUs without requiring a GPU. Our Python and R code is available at https://github.com/SamsungSAILMontreal/ForestDiffusion."
                    },
                    {
                        "title": "PopulAtion Parameter Averaging (PAPA)",
                        "abstract": "Ensemble methods combine the predictions of multiple models to improve performance, but they require significantly higher computation costs at inference time. To avoid these costs, multiple neural networks can be combined into one by averaging their weights. However, this usually performs significantly worse than ensembling. Weight averaging is only beneficial when different enough to benefit from combining them, but similar enough to average well. Based on this idea, we propose PopulAtion Parameter Averaging (PAPA): a method that combines the generality of ensembling with the efficiency of weight averaging. PAPA leverages a population of diverse models (trained on different data orders, augmentations, and regularizations) while slowly pushing the weights of the networks toward the population average of the weights. We also propose PAPA variants (PAPA-all, and PAPA-2) that average weights rarely rather than continuously; all methods increase generalization, but PAPA tends to perform best. PAPA reduces the performance gap between averaging and ensembling, increasing the average accuracy of a population of models by up to 0.8% on CIFAR-10, 1.9% on CIFAR-100, and 1.6% on ImageNet when compared to training independent (non-averaged) models."
                    },
                    {
                        "title": "Simulation-free Schr\u00f6dinger bridges via score and flow matching",
                        "abstract": "We present simulation-free score and flow matching ([SF]$^2$M), a simulation-free objective for inferring stochastic dynamics given unpaired samples drawn from arbitrary source and target distributions. Our method generalizes both the score-matching loss used in the training of diffusion models and the recently proposed flow matching loss used in the training of continuous normalizing flows. [SF]$^2$M interprets continuous-time stochastic generative modeling as a Schr\\\"odinger bridge problem. It relies on static entropy-regularized optimal transport, or a minibatch approximation, to efficiently learn the SB without simulating the learned stochastic process. We find that [SF]$^2$M is more efficient and gives more accurate solutions to the SB problem than simulation-based methods from prior work. Finally, we apply [SF]$^2$M to the problem of learning cell dynamics from snapshot data. Notably, [SF]$^2$M is the first method to accurately model cell dynamics in high dimensions and can recover known gene regulatory networks from simulated data. Our code is available in the TorchCFM package at https://github.com/atong01/conditional-flow-matching."
                    },
                    {
                        "title": "Improving and generalizing flow-based generative models with minibatch optimal transport",
                        "abstract": "Continuous normalizing flows (CNFs) are an attractive generative modeling technique, but they have been held back by limitations in their simulation-based maximum likelihood training. We introduce the generalized conditional flow matching (CFM) technique, a family of simulation-free training objectives for CNFs. CFM features a stable regression objective like that used to train the stochastic flow in diffusion models but enjoys the efficient inference of deterministic flow models. In contrast to both diffusion models and prior CNF training algorithms, CFM does not require the source distribution to be Gaussian or require evaluation of its density. A variant of our objective is optimal transport CFM (OT-CFM), which creates simpler flows that are more stable to train and lead to faster inference, as evaluated in our experiments. Furthermore, we show that when the true OT plan is available, our OT-CFM method approximates dynamic OT. Training CNFs with CFM improves results on a variety of conditional and unconditional generation tasks, such as inferring single cell dynamics, unsupervised image translation, and Schr\\\"odinger bridge inference."
                    },
                    {
                        "title": "Unbalanced Optimal Transport meets Sliced-Wasserstein",
                        "abstract": "Optimal transport (OT) has emerged as a powerful framework to compare probability measures, a fundamental task in many statistical and machine learning problems. Substantial advances have been made over the last decade in designing OT variants which are either computationally and statistically more efficient, or more robust to the measures and datasets to compare. Among them, sliced OT distances have been extensively used to mitigate optimal transport's cubic algorithmic complexity and curse of dimensionality. In parallel, unbalanced OT was designed to allow comparisons of more general positive measures, while being more robust to outliers. In this paper, we propose to combine these two concepts, namely slicing and unbalanced OT, to develop a general framework for efficiently comparing positive measures. We propose two new loss functions based on the idea of slicing unbalanced OT, and study their induced topology and statistical properties. We then develop a fast Frank-Wolfe-type algorithm to compute these loss functions, and show that the resulting methodology is modular as it encompasses and extends prior related work. We finally conduct an empirical analysis of our loss functions and methodology on both synthetic and real datasets, to illustrate their relevance and applicability."
                    },
                    {
                        "title": "Conditional Flow Matching: Simulation-Free Dynamic Optimal Transport",
                        "abstract": "Continuous normalizing \ufb02ows (CNFs) are an attractive generative modeling technique, but they have thus far been held back by limitations in their simulation-based maximum likelihood training. In this paper, we introduce a new technique called conditional \ufb02ow matching (CFM), a simulation-free training objective for CNFs. CFM features a stable regression objective like that used to train the stochastic \ufb02ow in diffusion models but enjoys the ef\ufb01cient inference of deterministic \ufb02ow models. In contrast to both diffusion models and prior CNF training algorithms, our CFM objec-tive does not require the source distribution to be Gaussian or require evaluation of its density. Based on this new objective, we also introduce optimal transport CFM (OT-CFM), which creates simpler \ufb02ows that are more stable to train and lead to faster inference, as evaluated in our experiments. Training CNFs with CFM improves results on a variety of conditional and unconditional generation tasks such as inferring single cell dynamics, unsupervised image translation, and Schr \u00a8 odinger bridge inference. Code is available at https://github.com/atong01/ conditional-flow-matching ."
                    },
                    {
                        "title": "Diffusion models with location-scale noise",
                        "abstract": "Diffusion Models (DMs) are powerful generative models that add Gaussian noise to the data and learn to remove it. We wanted to determine which noise distribution (Gaussian or non-Gaussian) led to better generated data in DMs. Since DMs do not work by design with non-Gaussian noise, we built a framework that allows reversing a diffusion process with non-Gaussian location-scale noise. We use that framework to show that the Gaussian distribution performs the best over a wide range of other distributions (Laplace, Uniform, t, Generalized-Gaussian)."
                    },
                    {
                        "title": "No Wrong Turns: The Simple Geometry Of Neural Networks Optimization Paths",
                        "abstract": "Understanding the optimization dynamics of neural networks is necessary for closing the gap between theory and practice. Stochastic first-order optimization algorithms are known to efficiently locate favorable minima in deep neural networks. This efficiency, however, contrasts with the non-convex and seemingly complex structure of neural loss landscapes. In this study, we delve into the fundamental geometric properties of sampled gradients along optimization paths. We focus on two key quantities, which appear in the restricted secant inequality and error bound. Both hold high significance for first-order optimization. Our analysis reveals that these quantities exhibit predictable, consistent behavior throughout training, despite the stochasticity induced by sampling minibatches. Our findings suggest that not only do optimization trajectories never encounter significant obstacles, but they also maintain stable dynamics during the majority of training. These observed properties are sufficiently expressive to theoretically guarantee linear convergence and prescribe learning rate schedules mirroring empirical practices. We conduct our experiments on image classification, semantic segmentation and language modeling across different batch sizes, network architectures, datasets, optimizers, and initialization seeds. We discuss the impact of each factor. Our work provides novel insights into the properties of neural network loss functions, and opens the door to theoretical frameworks more relevant to prevalent practice."
                    },
                    {
                        "title": "Optimal transport meets noisy label robust loss and MixUp regularization for domain adaptation",
                        "abstract": "It is common in computer vision to be confronted with domain shift: images which have the same class but different acquisition conditions. In domain adaptation (DA), one wants to classify unlabeled target images using source labeled images. Unfortunately, deep neural networks trained on a source training set perform poorly on target images which do not belong to the training domain. One strategy to improve these performances is to align the source and target image distributions in an embedded space using optimal transport (OT). However OT can cause negative transfer, i.e. aligning samples with different labels, which leads to overfitting especially in the presence of label shift between domains. In this work, we mitigate negative alignment by explaining it as a noisy label assignment to target images. We then mitigate its effect by appropriate regularization. We propose to couple the MixUp regularization \\citep{zhang2018mixup} with a loss that is robust to noisy labels in order to improve domain adaptation performance. We show in an extensive ablation study that a combination of the two techniques is critical to achieve improved performance. Finally, we evaluate our method, called \\textsc{mixunbot}, on several benchmarks and real-world DA problems."
                    },
                    {
                        "title": "A Reproducible and Realistic Evaluation of Partial Domain Adaptation Methods",
                        "abstract": "Unsupervised Domain Adaptation (UDA) aims at classifying unlabeled target images leveraging source labeled ones. In this work, we consider the Partial Domain Adaptation (PDA) variant, where we have extra source classes not present in the target domain. Most successful algorithms use model selection strategies that rely on target labels to find the best hyper-parameters and/or models along training. However, these strategies violate the main assumption in PDA: only unlabeled target domain samples are available. Moreover, there are also inconsistencies in the experimental settings - architecture, hyper-parameter tuning, number of runs - yielding unfair comparisons. The main goal of this work is to provide a realistic evaluation of PDA methods with the different model selection strategies under a consistent evaluation protocol. We evaluate 7 representative PDA algorithms on 2 different real-world datasets using 7 different model selection strategies. Our two main findings are: (i) without target labels for model selection, the accuracy of the methods decreases up to 30 percentage points; (ii) only one method and model selection pair performs well on both datasets. Experiments were performed with our PyTorch framework, BenchmarkPDA, which we open source."
                    },
                    {
                        "title": "On making optimal transport robust to all outliers",
                        "abstract": "Optimal transport (OT) is known to be sensitive against outliers because of its marginal constraints. Outlier robust OT variants have been proposed based on the de\ufb01nition that outliers are samples which are expensive to move. In this paper, we show that this de\ufb01nition is restricted by considering the case where outliers are closer to the target measure than clean samples. We show that outlier robust OT fully transports these outliers leading to poor performances in practice. To tackle these outliers, we propose to detect them by relying on a classi\ufb01er trained with adversarial training to classify source and target samples. A sample is then considered as an outlier if the prediction from the classi\ufb01er is different from its assigned label. To decrease the in\ufb02uence of these outliers in the transport problem, we propose to either remove them from the problem or to increase the cost of moving them by using the classi\ufb01er prediction. We show that we successfully detect these outliers and that they do not in\ufb02uence the transport problem on several experiments such as gradient \ufb02ows, generative models and label propagation."
                    },
                    {
                        "title": "Minibatch optimal transport distances; analysis and applications",
                        "abstract": "Optimal transport distances have become a classic tool to compare probability distributions and have found many applications in machine learning. Yet, despite recent algorithmic developments, their complexity prevents their direct use on large scale datasets. To overcome this challenge, a common workaround is to compute these distances on minibatches i.e. to average the outcome of several smaller optimal transport problems. We propose in this paper an extended analysis of this practice, which effects were previously studied in restricted cases. We first consider a large variety of Optimal Transport kernels. We notably argue that the minibatch strategy comes with appealing properties such as unbiased estimators, gradients and a concentration bound around the expectation, but also with limits: the minibatch OT is not a distance. To recover some of the lost distance axioms, we introduce a debiased minibatch OT function and study its statistical and optimisation properties. Along with this theoretical analysis, we also conduct empirical experiments on gradient flows, generative adversarial networks (GANs) or color transfer that highlight the practical interest of this strategy."
                    },
                    {
                        "title": "Unbalanced minibatch Optimal Transport; applications to Domain Adaptation",
                        "abstract": "Optimal transport distances have found many applications in machine learning for their capacity to compare non-parametric probability distributions. Yet their algorithmic complexity generally prevents their direct use on large scale datasets. Among the possible strategies to alleviate this issue, practitioners can rely on computing estimates of these distances over subsets of data, {\\em i.e.} minibatches. While computationally appealing, we highlight in this paper some limits of this strategy, arguing it can lead to undesirable smoothing effects. As an alternative, we suggest that the same minibatch strategy coupled with unbalanced optimal transport can yield more robust behavior. We discuss the associated theoretical properties, such as unbiased estimators, existence of gradients and concentration bounds. Our experimental study shows that in challenging problems associated to domain adaptation, the use of unbalanced optimal transport leads to significantly better results, competing with or surpassing recent baselines."
                    },
                    {
                        "title": "Generating Natural Adversarial Remote Sensing Images",
                        "abstract": "Over the last years, remote sensing image (RSI) analysis has started resorting to using deep neural networks to solve most of the commonly faced problems, such as detection, land cover classification, or segmentation. As far as critical decision-making can be based upon the results of RSI analysis, it is important to clearly identify and understand potential security threats occurring in those machine learning algorithms. Notably, it has recently been found that neural networks are particularly sensitive to carefully designed attacks, generally crafted given the full knowledge of the considered deep network. In this article, we consider the more realistic but challenging case where one wants to generate such attacks in the case of a black-box neural network. In this case, only the prediction score of the network is accessible, on a specific dataset. Examples that lure away the network\u2019s prediction, while being perceptually similar to real images, are called natural or unrestricted adversarial examples. We present an original method to generate such examples based on a variant of the Wasserstein generative adversarial network. We demonstrate its effectiveness on natural adversarial hyperspectral image generation and image modification for fooling a state-of-the-art detector. Among others, we also conduct a perceptual evaluation with human annotators to better assess the effectiveness of the proposed method. Our code is available for the community: https://github.com/PythonOT/ARWGAN."
                    },
                    {
                        "title": "Generating Natural Adversarial Hyperspectral examples with a modified Wasserstein GAN",
                        "abstract": "Adversarial examples are a hot topic due to their abilities to fool a classifier's prediction. There are two strategies to create such examples, one uses the attacked classifier's gradients, while the other only requires access to the clas-sifier's prediction. This is particularly appealing when the classifier is not full known (black box model). In this paper, we present a new method which is able to generate natural adversarial examples from the true data following the second paradigm. Based on Generative Adversarial Networks (GANs) [5], it reweights the true data empirical distribution to encourage the classifier to generate ad-versarial examples. We provide a proof of concept of our method by generating adversarial hyperspectral signatures on a remote sensing dataset."
                    },
                    {
                        "title": "Pushing the right boundaries matters! Wasserstein Adversarial Training for Label Noise",
                        "abstract": "Noisy labels often occur in vision datasets, especially when they are issued from crowdsourcing or Web scraping. In this paper, we propose a new regularization method which enables one to learn robust classifiers in presence of noisy data. To achieve this goal, we augment the virtual adversarial loss with a Wasserstein distance. This distance allows us to take into account specific relations between classes by leveraging on the geometric properties of this optimal transport distance. Notably, we encode the class similarities in the ground cost that is used to compute the Wasserstein distance. As a consequence, we can promote smoothness between classes that are very dissimilar, while keeping the classification decision function sufficiently complex for similar classes. While designing this ground cost can be left as a problem-specific modeling task, we show in this paper that using the semantic relations between classes names already leads to good results.Our proposed Wasserstein Adversarial Training (WAT) outperforms state of the art on four datasets corrupted with noisy labels: three classical benchmarks and one real case in remote sensing image semantic segmentation."
                    },
                    {
                        "title": "Learning with minibatch Wasserstein : asymptotic and gradient properties",
                        "abstract": "Optimal transport distances are powerful tools to compare probability distributions and have found many applications in machine learning. Yet their algorithmic complexity prevents their direct use on large scale datasets. To overcome this challenge, practitioners compute these distances on minibatches {\\em i.e.} they average the outcome of several smaller optimal transport problems. We propose in this paper an analysis of this practice, which effects are not well understood so far. We notably argue that it is equivalent to an implicit regularization of the original problem, with appealing properties such as unbiased estimators, gradients and a concentration bound around the expectation, but also with defects such as loss of distance property. Along with this theoretical analysis, we also conduct empirical experiments on gradient flows, GANs or color transfer that highlight the practical interest of this strategy."
                    }
                ]
            },
            "46c4cdf2-1345-4aa0-a64e-f8fa04fa52b4": {
                "pk": "46c4cdf2-1345-4aa0-a64e-f8fa04fa52b4",
                "name": "Jarrid Rector-Brooks",
                "collaborators": [
                    "Y. Bengio",
                    "Nikolay Malkin",
                    "Moksh Jain",
                    "Emmanuel Bengio",
                    "Cheng-Hao Liu",
                    "Alexander Tong",
                    "Maksym Korablyov",
                    "Sarthak Mittal",
                    "Pablo Lemos",
                    "Michael M. Bronstein"
                ],
                "domain": [
                    "Generative Modeling",
                    "Protein Design",
                    "Reinforcement Learning",
                    "Active Learning"
                ],
                "publications": [
                    {
                        "title": "Steering Masked Discrete Diffusion Models via Discrete Denoising Posterior Prediction",
                        "abstract": "Generative modeling of discrete data underlies important applications spanning text-based agents like ChatGPT to the design of the very building blocks of life in protein sequences. However, application domains need to exert control over the generated data by steering the generative process - typically via RLHF - to satisfy a specified property, reward, or affinity metric. In this paper, we study the problem of steering Masked Diffusion Models (MDMs), a recent class of discrete diffusion models that offer a compelling alternative to traditional autoregressive models. We introduce Discrete Denoising Posterior Prediction (DDPP), a novel framework that casts the task of steering pre-trained MDMs as a problem of probabilistic inference by learning to sample from a target Bayesian posterior. Our DDPP framework leads to a family of three novel objectives that are all simulation-free, and thus scalable while applying to general non-differentiable reward functions. Empirically, we instantiate DDPP by steering MDMs to perform class-conditional pixel-level image modeling, RLHF-based alignment of MDMs using text-based rewards, and finetuning protein language models to generate more diverse secondary structures and shorter proteins. We substantiate our designs via wet-lab validation, where we observe transient expression of reward-optimized protein sequences."
                    },
                    {
                        "title": "Sequence-Augmented SE(3)-Flow Matching For Conditional Protein Backbone Generation",
                        "abstract": "Proteins are essential for almost all biological processes and derive their diverse functions from complex 3D structures, which are in turn determined by their amino acid sequences. In this paper, we exploit the rich biological inductive bias of amino acid sequences and introduce FoldFlow-2, a novel sequence-conditioned SE(3)-equivariant flow matching model for protein structure generation. FoldFlow-2 presents substantial new architectural features over the previous FoldFlow family of models including a protein large language model to encode sequence, a new multi-modal fusion trunk that combines structure and sequence representations, and a geometric transformer based decoder. To increase diversity and novelty of generated samples -- crucial for de-novo drug design -- we train FoldFlow-2 at scale on a new dataset that is an order of magnitude larger than PDB datasets of prior works, containing both known proteins in PDB and high-quality synthetic structures achieved through filtering. We further demonstrate the ability to align FoldFlow-2 to arbitrary rewards, e.g. increasing secondary structures diversity, by introducing a Reinforced Finetuning (ReFT) objective. We empirically observe that FoldFlow-2 outperforms previous state-of-the-art protein structure-based generative models, improving over RFDiffusion in terms of unconditional generation across all metrics including designability, diversity, and novelty across all protein lengths, as well as exhibiting generalization on the task of equilibrium conformation sampling. Finally, we demonstrate that a fine-tuned FoldFlow-2 makes progress on challenging conditional design tasks such as designing scaffolds for the VHH nanobody."
                    },
                    {
                        "title": "Learning Conditional Policies for Crystal Design Using Offline Reinforcement Learning",
                        "abstract": "Navigating through the exponentially large chemical space to search for desirable materials is an extremely challenging task in material discovery. Recent developments in generative and geometric deep learning have shown..."
                    },
                    {
                        "title": "Improved off-policy training of diffusion samplers",
                        "abstract": "We study the problem of training diffusion models to sample from a distribution with a given unnormalized density or energy function. We benchmark several diffusion-structured inference methods, including simulation-based variational approaches and off-policy methods (continuous generative flow networks). Our results shed light on the relative advantages of existing algorithms while bringing into question some claims from past work. We also propose a novel exploration strategy for off-policy methods, based on local search in the target space with the use of a replay buffer, and show that it improves the quality of samples on a variety of target distributions. Our code for the sampling methods and benchmarks studied is made public at https://github.com/GFNOrg/gfn-diffusion as a base for future work on diffusion models for amortized inference."
                    },
                    {
                        "title": "Iterated Denoising Energy Matching for Sampling from Boltzmann Densities",
                        "abstract": "Efficiently generating statistically independent samples from an unnormalized probability distribution, such as equilibrium samples of many-body systems, is a foundational problem in science. In this paper, we propose Iterated Denoising Energy Matching (iDEM), an iterative algorithm that uses a novel stochastic score matching objective leveraging solely the energy function and its gradient -- and no data samples -- to train a diffusion-based sampler. Specifically, iDEM alternates between (I) sampling regions of high model density from a diffusion-based sampler and (II) using these samples in our stochastic matching objective to further improve the sampler. iDEM is scalable to high dimensions as the inner matching objective, is simulation-free, and requires no MCMC samples. Moreover, by leveraging the fast mode mixing behavior of diffusion, iDEM smooths out the energy landscape enabling efficient exploration and learning of an amortized sampler. We evaluate iDEM on a suite of tasks ranging from standard synthetic energy functions to invariant $n$-body particle systems. We show that the proposed approach achieves state-of-the-art performance on all metrics and trains $2-5\\times$ faster, which allows it to be the first method to train using energy on the challenging $55$-particle Lennard-Jones system."
                    },
                    {
                        "title": "Adaptive teachers for amortized samplers",
                        "abstract": "Amortized inference is the task of training a parametric model, such as a neural network, to approximate a distribution with a given unnormalized density where exact sampling is intractable. When sampling is implemented as a sequential decision-making process, reinforcement learning (RL) methods, such as generative flow networks, can be used to train the sampling policy. Off-policy RL training facilitates the discovery of diverse, high-reward candidates, but existing methods still face challenges in efficient exploration. We propose to use an adaptive training distribution (the Teacher) to guide the training of the primary amortized sampler (the Student) by prioritizing high-loss regions. The Teacher, an auxiliary behavior model, is trained to sample high-error regions of the Student and can generalize across unexplored modes, thereby enhancing mode coverage by providing an efficient training curriculum. We validate the effectiveness of this approach in a synthetic environment designed to present an exploration challenge, two diffusion-based sampling tasks, and four biochemical discovery tasks demonstrating its ability to improve sample efficiency and mode coverage."
                    },
                    {
                        "title": "Generative Active Learning for the Search of Small-molecule Protein Binders",
                        "abstract": "Despite substantial progress in machine learning for scientific discovery in recent years, truly de novo design of small molecules which exhibit a property of interest remains a significant challenge. We introduce LambdaZero, a generative active learning approach to search for synthesizable molecules. Powered by deep reinforcement learning, LambdaZero learns to search over the vast space of molecules to discover candidates with a desired property. We apply LambdaZero with molecular docking to design novel small molecules that inhibit the enzyme soluble Epoxide Hydrolase 2 (sEH), while enforcing constraints on synthesizability and drug-likeliness. LambdaZero provides an exponential speedup in terms of the number of calls to the expensive molecular docking oracle, and LambdaZero de novo designed molecules reach docking scores that would otherwise require the virtual screening of a hundred billion molecules. Importantly, LambdaZero discovers novel scaffolds of synthesizable, drug-like inhibitors for sEH. In in vitro experimental validation, a series of ligands from a generated quinazoline-based scaffold were synthesized, and the lead inhibitor N-(4,6-di(pyrrolidin-1-yl)quinazolin-2-yl)-N-methylbenzamide (UM0152893) displayed sub-micromolar enzyme inhibition of sEH."
                    },
                    {
                        "title": "Amortizing intractable inference in diffusion models for vision, language, and control",
                        "abstract": "Diffusion models have emerged as effective distribution estimators in vision, language, and reinforcement learning, but their use as priors in downstream tasks poses an intractable posterior inference problem. This paper studies amortized sampling of the posterior over data, $\\mathbf{x}\\sim p^{\\rm post}(\\mathbf{x})\\propto p(\\mathbf{x})r(\\mathbf{x})$, in a model that consists of a diffusion generative model prior $p(\\mathbf{x})$ and a black-box constraint or likelihood function $r(\\mathbf{x})$. We state and prove the asymptotic correctness of a data-free learning objective, relative trajectory balance, for training a diffusion model that samples from this posterior, a problem that existing methods solve only approximately or in restricted cases. Relative trajectory balance arises from the generative flow network perspective on diffusion models, which allows the use of deep reinforcement learning techniques to improve mode coverage. Experiments illustrate the broad potential of unbiased inference of arbitrary posteriors under diffusion priors: in vision (classifier guidance), language (infilling under a discrete diffusion LLM), and multimodal data (text-to-image generation). Beyond generative modeling, we apply relative trajectory balance to the problem of continuous control with a score-based behavior prior, achieving state-of-the-art results on benchmarks in offline reinforcement learning."
                    },
                    {
                        "title": "Thompson sampling for improved exploration in GFlowNets",
                        "abstract": "Generative flow networks (GFlowNets) are amortized variational inference algorithms that treat sampling from a distribution over compositional objects as a sequential decision-making problem with a learnable action policy. Unlike other algorithms for hierarchical sampling that optimize a variational bound, GFlowNet algorithms can stably run off-policy, which can be advantageous for discovering modes of the target distribution. Despite this flexibility in the choice of behaviour policy, the optimal way of efficiently selecting trajectories for training has not yet been systematically explored. In this paper, we view the choice of trajectories for training as an active learning problem and approach it using Bayesian techniques inspired by methods for multi-armed bandits. The proposed algorithm, Thompson sampling GFlowNets (TS-GFN), maintains an approximate posterior distribution over policies and samples trajectories from this posterior for training. We show in two domains that TS-GFN yields improved exploration and thus faster convergence to the target distribution than the off-policy exploration strategies used in past work."
                    },
                    {
                        "title": "Improving and generalizing flow-based generative models with minibatch optimal transport",
                        "abstract": "Continuous normalizing flows (CNFs) are an attractive generative modeling technique, but they have been held back by limitations in their simulation-based maximum likelihood training. We introduce the generalized conditional flow matching (CFM) technique, a family of simulation-free training objectives for CNFs. CFM features a stable regression objective like that used to train the stochastic flow in diffusion models but enjoys the efficient inference of deterministic flow models. In contrast to both diffusion models and prior CNF training algorithms, CFM does not require the source distribution to be Gaussian or require evaluation of its density. A variant of our objective is optimal transport CFM (OT-CFM), which creates simpler flows that are more stable to train and lead to faster inference, as evaluated in our experiments. Furthermore, we show that when the true OT plan is available, our OT-CFM method approximates dynamic OT. Training CNFs with CFM improves results on a variety of conditional and unconditional generation tasks, such as inferring single cell dynamics, unsupervised image translation, and Schr\\\"odinger bridge inference."
                    },
                    {
                        "title": "Conditional Flow Matching: Simulation-Free Dynamic Optimal Transport",
                        "abstract": "Continuous normalizing \ufb02ows (CNFs) are an attractive generative modeling technique, but they have thus far been held back by limitations in their simulation-based maximum likelihood training. In this paper, we introduce a new technique called conditional \ufb02ow matching (CFM), a simulation-free training objective for CNFs. CFM features a stable regression objective like that used to train the stochastic \ufb02ow in diffusion models but enjoys the ef\ufb01cient inference of deterministic \ufb02ow models. In contrast to both diffusion models and prior CNF training algorithms, our CFM objec-tive does not require the source distribution to be Gaussian or require evaluation of its density. Based on this new objective, we also introduce optimal transport CFM (OT-CFM), which creates simpler \ufb02ows that are more stable to train and lead to faster inference, as evaluated in our experiments. Training CNFs with CFM improves results on a variety of conditional and unconditional generation tasks such as inferring single cell dynamics, unsupervised image translation, and Schr \u00a8 odinger bridge inference. Code is available at https://github.com/atong01/ conditional-flow-matching ."
                    },
                    {
                        "title": "Learning GFlowNets from partial episodes for improved convergence and stability",
                        "abstract": "Generative flow networks (GFlowNets) are a family of algorithms for training a sequential sampler of discrete objects under an unnormalized target density and have been successfully used for various probabilistic modeling tasks. Existing training objectives for GFlowNets are either local to states or transitions, or propagate a reward signal over an entire sampling trajectory. We argue that these alternatives represent opposite ends of a gradient bias-variance tradeoff and propose a way to exploit this tradeoff to mitigate its harmful effects. Inspired by the TD($\\lambda$) algorithm in reinforcement learning, we introduce subtrajectory balance or SubTB($\\lambda$), a GFlowNet training objective that can learn from partial action subsequences of varying lengths. We show that SubTB($\\lambda$) accelerates sampler convergence in previously studied and new environments and enables training GFlowNets in environments with longer action sequences and sparser reward landscapes than what was possible before. We also perform a comparative analysis of stochastic gradient dynamics, shedding light on the bias-variance tradeoff in GFlowNet training and the advantages of subtrajectory balance."
                    },
                    {
                        "title": "RECOVER: sequential model optimization platform for combination drug repurposing identifies novel synergistic compounds in vitro",
                        "abstract": "For large libraries of small molecules, exhaustive combinatorial chemical screens become infeasible to perform when considering a range of disease models, assay conditions, and dose ranges. Deep learning models have achieved state of the art results in silico for the prediction of synergy scores. However, databases of drug combinations are biased towards synergistic agents and these results do not necessarily generalise out of distribution. We employ a sequential model optimization search utilising a deep learning model to quickly discover synergistic drug combinations active against a cancer cell line, requiring substantially less screening than an exhaustive evaluation. Our small scale wet lab experiments only account for evaluation of ~5% of the total search space. After only 3 rounds of ML-guided in vitro experimentation (including a calibration round), we find that the set of drug pairs queried is enriched for highly synergistic combinations; two additional rounds of ML-guided experiments were performed to ensure reproducibility of trends. Remarkably, we rediscover drug combinations later confirmed to be under study within clinical trials. Moreover, we find that drug embeddings generated using only structural information begin to reflect mechanisms of action. Prior in silico benchmarking suggests we can enrich search queries by a factor of ~5-10x for highly synergistic drug combinations by using sequential rounds of evaluation when compared to random selection, or by a factor of>3x when using a pretrained model selecting all drug combinations at a single time point."
                    },
                    {
                        "title": "Biological Sequence Design with GFlowNets",
                        "abstract": "Design of de novo biological sequences with desired properties, like protein and DNA sequences, often involves an active loop with several rounds of molecule ideation and expensive wet-lab evaluations. These experiments can consist of multiple stages, with increasing levels of precision and cost of evaluation, where candidates are filtered. This makes the diversity of proposed candidates a key consideration in the ideation phase. In this work, we propose an active learning algorithm leveraging epistemic uncertainty estimation and the recently proposed GFlowNets as a generator of diverse candidate solutions, with the objective to obtain a diverse batch of useful (as defined by some utility function, for example, the predicted anti-microbial activity of a peptide) and informative candidates after each round. We also propose a scheme to incorporate existing labeled datasets of candidates, in addition to a reward function, to speed up learning in GFlowNets. We present empirical results on several biological sequence design tasks, and we find that our method generates more diverse and novel batches with high scoring candidates compared to existing approaches."
                    },
                    {
                        "title": "Multi-Objective GFlowNets",
                        "abstract": "We study the problem of generating diverse candidates in the context of Multi-Objective Optimization. In many applications of machine learning such as drug discovery and material design, the goal is to generate candidates which simultaneously optimize a set of potentially conflicting objectives. Moreover, these objectives are often imperfect evaluations of some underlying property of interest, making it important to generate diverse candidates to have multiple options for expensive downstream evaluations. We propose Multi-Objective GFlowNets (MOGFNs), a novel method for generating diverse Pareto optimal solutions, based on GFlowNets. We introduce two variants of MOGFNs: MOGFN-PC, which models a family of independent sub-problems defined by a scalarization function, with reward-conditional GFlowNets, and MOGFN-AL, which solves a sequence of sub-problems defined by an acquisition function in an active learning loop. Our experiments on wide variety of synthetic and benchmark tasks demonstrate advantages of the proposed methods in terms of the Pareto performance and importantly, improved candidate diversity, which is the main contribution of this work."
                    },
                    {
                        "title": "DEUP: Direct Epistemic Uncertainty Prediction",
                        "abstract": "Epistemic Uncertainty is a measure of the lack of knowledge of a learner which diminishes with more evidence. While existing work focuses on using the variance of the Bayesian posterior due to parameter uncertainty as a measure of epistemic uncertainty, we argue that this does not capture the part of lack of knowledge induced by model misspecification. We discuss how the excess risk, which is the gap between the generalization error of a predictor and the Bayes predictor, is a sound measure of epistemic uncertainty which captures the effect of model misspecification. We thus propose a principled framework for directly estimating the excess risk by learning a secondary predictor for the generalization error and subtracting an estimate of aleatoric uncertainty, i.e., intrinsic unpredictability. We discuss the merits of this novel measure of epistemic uncertainty, and highlight how it differs from variance-based measures of epistemic uncertainty and addresses its major pitfall. Our framework, Direct Epistemic Uncertainty Prediction (DEUP) is particularly interesting in interactive learning environments, where the learner is allowed to acquire novel examples in each round. Through a wide set of experiments, we illustrate how existing methods in sequential model optimization can be improved with epistemic uncertainty estimates from DEUP, and how DEUP can be used to drive exploration in reinforcement learning. We also evaluate the quality of uncertainty estimates from DEUP for probabilistic image classification and predicting synergies of drug combinations."
                    },
                    {
                        "title": "Revisiting Projection-Free Optimization for Strongly Convex Constraint Sets",
                        "abstract": "We revisit the Frank-Wolfe (FW) optimization under strongly convex constraint sets. We provide a faster convergence rate for FW without line search, showing that a previously overlooked variant of FW is indeed faster than the standard variant. With line search, we show that FW can converge to the global optimum, even for smooth functions that are not convex, but are quasi-convex and locally-Lipschitz. We also show that, for the general case of (smooth) non-convex functions, FW with line search converges with high probability to a stationary point at a rate of O(1/t), as long as the constraint set is strongly convex\u2014one of the fastest convergence rates in non-convex optimization."
                    }
                ]
            },
            "f793ad89-2487-44f9-8b43-9ca6cb39ea20": {
                "pk": "f793ad89-2487-44f9-8b43-9ca6cb39ea20",
                "name": "Cheng-Hao Liu",
                "collaborators": [
                    "Jarrid Rector-Brooks",
                    "Y. Bengio",
                    "Alexander Tong",
                    "A. Bose",
                    "Sarthak Mittal",
                    "Michael M. Bronstein",
                    "Pablo Lemos",
                    "Tara Akhound-Sadegh",
                    "Mohsin Hasan",
                    "Zhangzhi Peng"
                ],
                "domain": [
                    "Generative Modeling",
                    "Protein Structure Prediction",
                    "Drug Discovery",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Steering Masked Discrete Diffusion Models via Discrete Denoising Posterior Prediction",
                        "abstract": "Generative modeling of discrete data underlies important applications spanning text-based agents like ChatGPT to the design of the very building blocks of life in protein sequences. However, application domains need to exert control over the generated data by steering the generative process - typically via RLHF - to satisfy a specified property, reward, or affinity metric. In this paper, we study the problem of steering Masked Diffusion Models (MDMs), a recent class of discrete diffusion models that offer a compelling alternative to traditional autoregressive models. We introduce Discrete Denoising Posterior Prediction (DDPP), a novel framework that casts the task of steering pre-trained MDMs as a problem of probabilistic inference by learning to sample from a target Bayesian posterior. Our DDPP framework leads to a family of three novel objectives that are all simulation-free, and thus scalable while applying to general non-differentiable reward functions. Empirically, we instantiate DDPP by steering MDMs to perform class-conditional pixel-level image modeling, RLHF-based alignment of MDMs using text-based rewards, and finetuning protein language models to generate more diverse secondary structures and shorter proteins. We substantiate our designs via wet-lab validation, where we observe transient expression of reward-optimized protein sequences."
                    },
                    {
                        "title": "Sequence-Augmented SE(3)-Flow Matching For Conditional Protein Backbone Generation",
                        "abstract": "Proteins are essential for almost all biological processes and derive their diverse functions from complex 3D structures, which are in turn determined by their amino acid sequences. In this paper, we exploit the rich biological inductive bias of amino acid sequences and introduce FoldFlow-2, a novel sequence-conditioned SE(3)-equivariant flow matching model for protein structure generation. FoldFlow-2 presents substantial new architectural features over the previous FoldFlow family of models including a protein large language model to encode sequence, a new multi-modal fusion trunk that combines structure and sequence representations, and a geometric transformer based decoder. To increase diversity and novelty of generated samples -- crucial for de-novo drug design -- we train FoldFlow-2 at scale on a new dataset that is an order of magnitude larger than PDB datasets of prior works, containing both known proteins in PDB and high-quality synthetic structures achieved through filtering. We further demonstrate the ability to align FoldFlow-2 to arbitrary rewards, e.g. increasing secondary structures diversity, by introducing a Reinforced Finetuning (ReFT) objective. We empirically observe that FoldFlow-2 outperforms previous state-of-the-art protein structure-based generative models, improving over RFDiffusion in terms of unconditional generation across all metrics including designability, diversity, and novelty across all protein lengths, as well as exhibiting generalization on the task of equilibrium conformation sampling. Finally, we demonstrate that a fine-tuned FoldFlow-2 makes progress on challenging conditional design tasks such as designing scaffolds for the VHH nanobody."
                    },
                    {
                        "title": "Iterated Denoising Energy Matching for Sampling from Boltzmann Densities",
                        "abstract": "Efficiently generating statistically independent samples from an unnormalized probability distribution, such as equilibrium samples of many-body systems, is a foundational problem in science. In this paper, we propose Iterated Denoising Energy Matching (iDEM), an iterative algorithm that uses a novel stochastic score matching objective leveraging solely the energy function and its gradient -- and no data samples -- to train a diffusion-based sampler. Specifically, iDEM alternates between (I) sampling regions of high model density from a diffusion-based sampler and (II) using these samples in our stochastic matching objective to further improve the sampler. iDEM is scalable to high dimensions as the inner matching objective, is simulation-free, and requires no MCMC samples. Moreover, by leveraging the fast mode mixing behavior of diffusion, iDEM smooths out the energy landscape enabling efficient exploration and learning of an amortized sampler. We evaluate iDEM on a suite of tasks ranging from standard synthetic energy functions to invariant $n$-body particle systems. We show that the proposed approach achieves state-of-the-art performance on all metrics and trains $2-5\\times$ faster, which allows it to be the first method to train using energy on the challenging $55$-particle Lennard-Jones system."
                    },
                    {
                        "title": "A community effort in SARS-CoV-2 drug discovery",
                        "abstract": "The COVID\u201019 pandemic continues to pose a substantial threat to human lives and is likely to do so for years to come. Despite the availability of vaccines, searching for efficient small\u2010molecule drugs that are widely available, including in low\u2010 and middle\u2010income countries, is an ongoing challenge. In this work, we report the results of an open science community effort, the \u201cBillion molecules against COVID\u201019 challenge\u201d, to identify small\u2010molecule inhibitors against SARS\u2010CoV\u20102 or relevant human receptors. Participating teams used a wide variety of computational methods to screen a minimum of 1 billion virtual molecules against 6 protein targets. Overall, 31 teams participated, and they suggested a total of 639,024 molecules, which were subsequently ranked to find \u2018consensus compounds\u2019. The organizing team coordinated with various contract research organizations (CROs) and collaborating institutions to synthesize and test 878 compounds for biological activity against proteases (Nsp5, Nsp3, TMPRSS2), nucleocapsid N, RdRP (only the Nsp12 domain), and (alpha) spike protein S. Overall, 27 compounds with weak inhibition/binding were experimentally identified by binding\u2010, cleavage\u2010, and/or viral suppression assays and are presented here. Open science approaches such as the one presented here contribute to the knowledge base of future drug discovery efforts in finding better SARS\u2010CoV\u20102 treatments."
                    }
                ]
            },
            "2ae12888-292c-484e-8f68-a0a20a48f479": {
                "pk": "2ae12888-292c-484e-8f68-a0a20a48f479",
                "name": "Andrei Cristian Nica",
                "collaborators": [
                    "A. Florea",
                    "Maksym Korablyov",
                    "Moksh Jain",
                    "Emmanuel Bengio",
                    "Y. Bengio",
                    "Cheng-Hao Liu",
                    "Jarrid Rector-Brooks",
                    "Mihai Tr\u0103sc\u0103u",
                    "Alexandru Sorici",
                    "A. V. D. Sloot"
                ],
                "domain": [
                    "Machine Learning",
                    "Reinforcement Learning",
                    "Drug Discovery",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Generative Active Learning for the Search of Small-molecule Protein Binders",
                        "abstract": "Despite substantial progress in machine learning for scientific discovery in recent years, truly de novo design of small molecules which exhibit a property of interest remains a significant challenge. We introduce LambdaZero, a generative active learning approach to search for synthesizable molecules. Powered by deep reinforcement learning, LambdaZero learns to search over the vast space of molecules to discover candidates with a desired property. We apply LambdaZero with molecular docking to design novel small molecules that inhibit the enzyme soluble Epoxide Hydrolase 2 (sEH), while enforcing constraints on synthesizability and drug-likeliness. LambdaZero provides an exponential speedup in terms of the number of calls to the expensive molecular docking oracle, and LambdaZero de novo designed molecules reach docking scores that would otherwise require the virtual screening of a hundred billion molecules. Importantly, LambdaZero discovers novel scaffolds of synthesizable, drug-like inhibitors for sEH. In in vitro experimental validation, a series of ligands from a generated quinazoline-based scaffold were synthesized, and the lead inhibitor N-(4,6-di(pyrrolidin-1-yl)quinazolin-2-yl)-N-methylbenzamide (UM0152893) displayed sub-micromolar enzyme inhibition of sEH."
                    },
                    {
                        "title": "Order from Chaos: Leveraging the Order in Time as Intrinsic Reward",
                        "abstract": "Reinforcement learning (RL) agents often struggle with exploration problems where the environment is complex and the reward signal is sparse or delayed. In this paper, we propose a novel intrinsic reward mechanism that leverages the order of time to guide agents toward less chaotic policies. Our approach involves training a model to predict the correct order of a shuffled sequence of observations, which enables us to introduce the \u201corderability\u201d score. This score captures the extent to which observations from a trajectory are uniquely ordered in time, and we hypothesize that it is a helpful metric for assessing the learning progress of reinforcement learning agents. By incorporating the orderability score as an intrinsic reward, we aim to encourage agents to explore their environment more effectively and achieve faster and more consistent reward maximization. In our experiments, we demonstrate that agents trained with the orderability intrinsic reward outperform baseline methods on challenging exploration tasks, highlighting the potential of our approach. By shedding light on the importance of time\u2019s order in RL, we provide a fresh perspective on the challenge of exploration and pave the way for future research in this area."
                    },
                    {
                        "title": "E VALUATING G ENERALIZATION IN GF LOW N ETS FOR M OLECULE D ESIGN",
                        "abstract": "Deep learning bears promise for drug discovery problems such as de novo molecular design. Generating data to train such models is a costly and time-consuming process, given the need for wet-lab experiments or expensive simulations. This problem is compounded by the notorious data-hungriness of machine learning algorithms. In small molecule generation the recently proposed GFlowNet method has shown good performance in generating diverse high-scoring candidates, and has the interesting advantage of being an off-policy offline method. Finding an appropriate generalization evaluation metric for such models, one predictive of the desired search performance (i.e. finding high-scoring diverse candidates), will help guide online data collection for such an algorithm. In this work, we develop techniques for evaluating GFlowNet performance on a test set, and identify the most promising metric for predicting generalization. We present empirical results on several small-molecule design tasks in drug discovery, for several GFlowNet training setups, and we find a metric strongly correlated with diverse high-scoring batch generation. This metric should be used to identify the best generative model from which to sample batches of molecules to be evaluated."
                    },
                    {
                        "title": "Learning GFlowNets from partial episodes for improved convergence and stability",
                        "abstract": "Generative flow networks (GFlowNets) are a family of algorithms for training a sequential sampler of discrete objects under an unnormalized target density and have been successfully used for various probabilistic modeling tasks. Existing training objectives for GFlowNets are either local to states or transitions, or propagate a reward signal over an entire sampling trajectory. We argue that these alternatives represent opposite ends of a gradient bias-variance tradeoff and propose a way to exploit this tradeoff to mitigate its harmful effects. Inspired by the TD($\\lambda$) algorithm in reinforcement learning, we introduce subtrajectory balance or SubTB($\\lambda$), a GFlowNet training objective that can learn from partial action subsequences of varying lengths. We show that SubTB($\\lambda$) accelerates sampler convergence in previously studied and new environments and enables training GFlowNets in environments with longer action sequences and sparser reward landscapes than what was possible before. We also perform a comparative analysis of stochastic gradient dynamics, shedding light on the bias-variance tradeoff in GFlowNet training and the advantages of subtrajectory balance."
                    },
                    {
                        "title": "The Paradox of Choice: Using Attention in Hierarchical Reinforcement Learning",
                        "abstract": "Decision-making AI agents are often faced with two important challenges: the depth of the planning horizon, and the branching factor due to having many choices. Hierarchical reinforcement learning methods aim to solve the first problem, by providing shortcuts that skip over multiple time steps. To cope with the breadth, it is desirable to restrict the agent's attention at each step to a reasonable number of possible choices. The concept of affordances (Gibson, 1977) suggests that only certain actions are feasible in certain states. In this work, we model\"affordances\"through an attention mechanism that limits the available choices of temporally extended options. We present an online, model-free algorithm to learn affordances that can be used to further learn subgoal options. We investigate the role of hard versus soft attention in training data collection, abstract value learning in long-horizon tasks, and handling a growing number of choices. We identify and empirically illustrate the settings in which the paradox of choice arises, i.e. when having fewer but more meaningful choices improves the learning speed and performance of a reinforcement learning agent."
                    },
                    {
                        "title": "Collecting and Processing a Self-Driving Dataset in the UPB Campus",
                        "abstract": "Although there is a diversity of publicly available datasets for autonomous driving, from small-scale to larger collections with thousands of miles of driving, we consider that the process of collecting and processing them is often overlooked in the literature. From a data-driven perspective, quality of a dataset has proven as important as quantity especially when evaluating self-driving technologies where safety is crucial. In this paper, we provide a guideline going through all the steps from configuring the hardware setup to obtaining a clean dataset. We describe the data collection scenario design, the hardware and software employed in the process, the challenges that must be considered, data filtering and validation stage. This work stems from our experience in collecting the UPB campus driving dataset released together with this work. It is our belief that having a clean and efficient process of collecting a small but meaningful dataset has the potential to improve benchmarking autonomous driving solutions, capturing local environment particularities."
                    },
                    {
                        "title": "End-to-end models for self-driving cars on UPB campus roads",
                        "abstract": "In autonomous driving, making the vehicle trajectory planning a learnable process (so called end-to-end driving), given visual input and car motion measurements alone is an idea with longstanding tradition. In a recent project we investigate the potential to incorporate the steering predictions of an end-to-end model operating under constraints of a single monocular camera input, into the planning pipeline of existing open-source frameworks for autonomous driving. We describe a process to design and train such a model given a specific geographical area with its own particularities in terms of road network, traffic and driving conditions. We investigate data augmentation techniques which improve the performance of the model. Quality and performance analysis of the model is performed based on a collected dataset in the UPB campus. The evaluation is carried out using a simulation pipeline which applies transformations to the frames in test videos such that they reflect decisions made by the model in earlier steps. This brings the system evaluation closer to out of distribution testing. Results indicate that the proposed models offers a good performance for the collected dataset."
                    },
                    {
                        "title": "Learning to Maximize Return in a Stag Hunt Collaborative Scenario through Deep Reinforcement Learning",
                        "abstract": "In this paper we present a deep reinforcement learning approach for learning to play a time extended social dilemma game in a simulated environment. Agents face different types of adversaries with different levels of commitment to a collaborative strategy. Our method builds on recent advances in policy gradient training using deep neural networks. We investigate multiple stochastic gradient algorithms such as Reinforce or Actor Critic with auxiliary tasks for faster convergence."
                    }
                ]
            },
            "656492d3-5442-4f5c-ae08-8438e74b8285": {
                "pk": "656492d3-5442-4f5c-ae08-8438e74b8285",
                "name": "Maksym Korablyov",
                "collaborators": [
                    "Y. Bengio",
                    "Moksh Jain",
                    "Cheng-Hao Liu",
                    "Emmanuel Bengio",
                    "Jarrid Rector-Brooks",
                    "A. Pavlova",
                    "L. Bassit",
                    "A. Nica",
                    "Stanislaw Jastrzebski",
                    "Marwin H. S. Segler"
                ],
                "domain": [
                    "Drug Discovery",
                    "Molecular Design",
                    "Machine Learning",
                    "Biophysics"
                ],
                "publications": [
                    {
                        "title": "Biophysics-Guided Lead Discovery of HBV Capsid Assembly Modifiers",
                        "abstract": "Hepatitis B virus (HBV) is the leading cause of chronic liver pathologies worldwide. HBV nucleocapsid, a key structural component, is formed through the self-assembly of the capsid protein units. Therefore, interfering with the self-assembly process is a promising approach for the development of novel antiviral agents. Applied to HBV, this approach has led to several classes of capsid assembly modulators (CAMs). Here, we report structurally novel CAMs with moderate activity and low toxicity, discovered through a biophysics-guided approach combining docking, molecular dynamics simulations, and a series of assays with a particular emphasis on biophysical experiments. Several of the identified compounds induce the formation of aberrant capsids and inhibit HBV DNA replication in vitro, suggesting that they possess modest capsid assembly modulation effects. The synergistic computational and experimental approaches provided key insights that facilitated the identification of compounds with promising activities. The discovery of preclinical CAMs presents opportunities for subsequent optimization efforts, thereby opening new avenues for HBV inhibition."
                    },
                    {
                        "title": "Generative Active Learning for the Search of Small-molecule Protein Binders",
                        "abstract": "Despite substantial progress in machine learning for scientific discovery in recent years, truly de novo design of small molecules which exhibit a property of interest remains a significant challenge. We introduce LambdaZero, a generative active learning approach to search for synthesizable molecules. Powered by deep reinforcement learning, LambdaZero learns to search over the vast space of molecules to discover candidates with a desired property. We apply LambdaZero with molecular docking to design novel small molecules that inhibit the enzyme soluble Epoxide Hydrolase 2 (sEH), while enforcing constraints on synthesizability and drug-likeliness. LambdaZero provides an exponential speedup in terms of the number of calls to the expensive molecular docking oracle, and LambdaZero de novo designed molecules reach docking scores that would otherwise require the virtual screening of a hundred billion molecules. Importantly, LambdaZero discovers novel scaffolds of synthesizable, drug-like inhibitors for sEH. In in vitro experimental validation, a series of ligands from a generated quinazoline-based scaffold were synthesized, and the lead inhibitor N-(4,6-di(pyrrolidin-1-yl)quinazolin-2-yl)-N-methylbenzamide (UM0152893) displayed sub-micromolar enzyme inhibition of sEH."
                    },
                    {
                        "title": "Thompson sampling for improved exploration in GFlowNets",
                        "abstract": "Generative flow networks (GFlowNets) are amortized variational inference algorithms that treat sampling from a distribution over compositional objects as a sequential decision-making problem with a learnable action policy. Unlike other algorithms for hierarchical sampling that optimize a variational bound, GFlowNet algorithms can stably run off-policy, which can be advantageous for discovering modes of the target distribution. Despite this flexibility in the choice of behaviour policy, the optimal way of efficiently selecting trajectories for training has not yet been systematically explored. In this paper, we view the choice of trajectories for training as an active learning problem and approach it using Bayesian techniques inspired by methods for multi-armed bandits. The proposed algorithm, Thompson sampling GFlowNets (TS-GFN), maintains an approximate posterior distribution over policies and samples trajectories from this posterior for training. We show in two domains that TS-GFN yields improved exploration and thus faster convergence to the target distribution than the off-policy exploration strategies used in past work."
                    },
                    {
                        "title": "A community effort in SARS-CoV-2 drug discovery",
                        "abstract": "The COVID\u201019 pandemic continues to pose a substantial threat to human lives and is likely to do so for years to come. Despite the availability of vaccines, searching for efficient small\u2010molecule drugs that are widely available, including in low\u2010 and middle\u2010income countries, is an ongoing challenge. In this work, we report the results of an open science community effort, the \u201cBillion molecules against COVID\u201019 challenge\u201d, to identify small\u2010molecule inhibitors against SARS\u2010CoV\u20102 or relevant human receptors. Participating teams used a wide variety of computational methods to screen a minimum of 1 billion virtual molecules against 6 protein targets. Overall, 31 teams participated, and they suggested a total of 639,024 molecules, which were subsequently ranked to find \u2018consensus compounds\u2019. The organizing team coordinated with various contract research organizations (CROs) and collaborating institutions to synthesize and test 878 compounds for biological activity against proteases (Nsp5, Nsp3, TMPRSS2), nucleocapsid N, RdRP (only the Nsp12 domain), and (alpha) spike protein S. Overall, 27 compounds with weak inhibition/binding were experimentally identified by binding\u2010, cleavage\u2010, and/or viral suppression assays and are presented here. Open science approaches such as the one presented here contribute to the knowledge base of future drug discovery efforts in finding better SARS\u2010CoV\u20102 treatments."
                    },
                    {
                        "title": "The Mechanism of Action of Hepatitis B Virus Capsid Assembly Modulators Can Be Predicted from Binding to Early Assembly Intermediates.",
                        "abstract": "Interfering with the self-assembly of virus nucleocapsids is a promising approach for the development of novel antiviral agents. Applied to hepatitis B virus (HBV), this approach has led to several classes of capsid assembly modulators (CAMs) that target the virus by either accelerating nucleocapsid assembly or misdirecting it into noncapsid-like particles, thereby inhibiting the HBV replication cycle. Here, we have assessed the structures of early nucleocapsid assembly intermediates, bound with and without CAMs, using molecular dynamics simulations. We find that distinct conformations of the intermediates are induced depending on whether the bound CAM accelerates or misdirects assembly. Specifically, the assembly intermediates with bound misdirecting CAMs appear to be flattened relative to those with bound accelerators. Finally, the potency of CAMs within the same class was studied. We find that an increased number of contacts with the capsid protein and favorable binding energies inferred from free energy perturbation calculations are indicative of increased potency."
                    },
                    {
                        "title": "RetroGNN: Fast Estimation of Synthesizability for Virtual Screening and De Novo Design by Learning from Slow Retrosynthesis Software",
                        "abstract": "De novo molecule design algorithms often result in chemically unfeasible or synthetically inaccessible molecules. A natural idea to mitigate this problem is to bias these algorithms toward more easily synthesizable molecules using a proxy score for synthetic accessibility. However, using currently available proxies can still result in highly unrealistic compounds. Here, we propose a novel approach, RetroGNN, to estimate synthesizability. First, we search for routes using synthesis planning software for a large number of random molecules. This information is then used to train a graph neural network to predict the outcome of the synthesis planner given the target molecule, in which the regression task can be used as a synthesizability scorer. We highlight how RetroGNN can be used in generative molecule-discovery pipelines together with other scoring functions. We evaluate our approach on several QSAR-based molecule design benchmarks, for which we find synthesizable molecules with state-of-the-art scores. Compared to the virtual screening of 5 million existing molecules from the ZINC database, using RetroGNNScore with a simple fragment-based de novo design algorithm finds molecules predicted to be more likely to possess the desired activity exponentially faster, while maintaining good druglike properties and being easier to synthesize. Importantly, our deep neural network can successfully filter out hard to synthesize molecules while achieving a 105 times speedup over using retrosynthesis planning software."
                    },
                    {
                        "title": "E VALUATING G ENERALIZATION IN GF LOW N ETS FOR M OLECULE D ESIGN",
                        "abstract": "Deep learning bears promise for drug discovery problems such as de novo molecular design. Generating data to train such models is a costly and time-consuming process, given the need for wet-lab experiments or expensive simulations. This problem is compounded by the notorious data-hungriness of machine learning algorithms. In small molecule generation the recently proposed GFlowNet method has shown good performance in generating diverse high-scoring candidates, and has the interesting advantage of being an off-policy offline method. Finding an appropriate generalization evaluation metric for such models, one predictive of the desired search performance (i.e. finding high-scoring diverse candidates), will help guide online data collection for such an algorithm. In this work, we develop techniques for evaluating GFlowNet performance on a test set, and identify the most promising metric for predicting generalization. We present empirical results on several small-molecule design tasks in drug discovery, for several GFlowNet training setups, and we find a metric strongly correlated with diverse high-scoring batch generation. This metric should be used to identify the best generative model from which to sample batches of molecules to be evaluated."
                    },
                    {
                        "title": "Learning GFlowNets from partial episodes for improved convergence and stability",
                        "abstract": "Generative flow networks (GFlowNets) are a family of algorithms for training a sequential sampler of discrete objects under an unnormalized target density and have been successfully used for various probabilistic modeling tasks. Existing training objectives for GFlowNets are either local to states or transitions, or propagate a reward signal over an entire sampling trajectory. We argue that these alternatives represent opposite ends of a gradient bias-variance tradeoff and propose a way to exploit this tradeoff to mitigate its harmful effects. Inspired by the TD($\\lambda$) algorithm in reinforcement learning, we introduce subtrajectory balance or SubTB($\\lambda$), a GFlowNet training objective that can learn from partial action subsequences of varying lengths. We show that SubTB($\\lambda$) accelerates sampler convergence in previously studied and new environments and enables training GFlowNets in environments with longer action sequences and sparser reward landscapes than what was possible before. We also perform a comparative analysis of stochastic gradient dynamics, shedding light on the bias-variance tradeoff in GFlowNet training and the advantages of subtrajectory balance."
                    },
                    {
                        "title": "RECOVER: sequential model optimization platform for combination drug repurposing identifies novel synergistic compounds in vitro",
                        "abstract": "For large libraries of small molecules, exhaustive combinatorial chemical screens become infeasible to perform when considering a range of disease models, assay conditions, and dose ranges. Deep learning models have achieved state of the art results in silico for the prediction of synergy scores. However, databases of drug combinations are biased towards synergistic agents and these results do not necessarily generalise out of distribution. We employ a sequential model optimization search utilising a deep learning model to quickly discover synergistic drug combinations active against a cancer cell line, requiring substantially less screening than an exhaustive evaluation. Our small scale wet lab experiments only account for evaluation of ~5% of the total search space. After only 3 rounds of ML-guided in vitro experimentation (including a calibration round), we find that the set of drug pairs queried is enriched for highly synergistic combinations; two additional rounds of ML-guided experiments were performed to ensure reproducibility of trends. Remarkably, we rediscover drug combinations later confirmed to be under study within clinical trials. Moreover, we find that drug embeddings generated using only structural information begin to reflect mechanisms of action. Prior in silico benchmarking suggests we can enrich search queries by a factor of ~5-10x for highly synergistic drug combinations by using sequential rounds of evaluation when compared to random selection, or by a factor of>3x when using a pretrained model selecting all drug combinations at a single time point."
                    },
                    {
                        "title": "Properties of Minimizing Entropy",
                        "abstract": "Compact data representations are one approach for improving generalization of learned functions. We explicitly illustrate the relationship between entropy and cardinality, both measures of compactness, including how gradient descent on the former reduces the latter. Whereas entropy is distribution sensitive, cardinality is not. We propose a third compactness measure that is a compromise between the two: expected cardinality, or the expected number of unique states in any finite number of draws, which is more meaningful than standard cardinality as it discounts states with negligible probability mass. We show that minimizing entropy also minimizes expected cardinality."
                    },
                    {
                        "title": "DEUP: Direct Epistemic Uncertainty Prediction",
                        "abstract": "Epistemic Uncertainty is a measure of the lack of knowledge of a learner which diminishes with more evidence. While existing work focuses on using the variance of the Bayesian posterior due to parameter uncertainty as a measure of epistemic uncertainty, we argue that this does not capture the part of lack of knowledge induced by model misspecification. We discuss how the excess risk, which is the gap between the generalization error of a predictor and the Bayes predictor, is a sound measure of epistemic uncertainty which captures the effect of model misspecification. We thus propose a principled framework for directly estimating the excess risk by learning a secondary predictor for the generalization error and subtracting an estimate of aleatoric uncertainty, i.e., intrinsic unpredictability. We discuss the merits of this novel measure of epistemic uncertainty, and highlight how it differs from variance-based measures of epistemic uncertainty and addresses its major pitfall. Our framework, Direct Epistemic Uncertainty Prediction (DEUP) is particularly interesting in interactive learning environments, where the learner is allowed to acquire novel examples in each round. Through a wide set of experiments, we illustrate how existing methods in sequential model optimization can be improved with epistemic uncertainty estimates from DEUP, and how DEUP can be used to drive exploration in reinforcement learning. We also evaluate the quality of uncertainty estimates from DEUP for probabilistic image classification and predicting synergies of drug combinations."
                    },
                    {
                        "title": "Flow Network based Generative Models for Non-Iterative Diverse Candidate Generation",
                        "abstract": "This paper is about the problem of learning a stochastic policy for generating an object (like a molecular graph) from a sequence of actions, such that the probability of generating an object is proportional to a given positive reward for that object. Whereas standard return maximization tends to converge to a single return-maximizing sequence, there are cases where we would like to sample a diverse set of high-return solutions. These arise, for example, in black-box function optimization when few rounds are possible, each with large batches of queries, where the batches should be diverse, e.g., in the design of new molecules. One can also see this as a problem of approximately converting an energy function to a generative distribution. While MCMC methods can achieve that, they are expensive and generally only perform local exploration. Instead, training a generative policy amortizes the cost of search during training and yields to fast generation. Using insights from Temporal Difference learning, we propose GFlowNet, based on a view of the generative process as a flow network, making it possible to handle the tricky case where different trajectories can yield the same final state, e.g., there are many ways to sequentially add atoms to generate some molecular graph. We cast the set of trajectories as a flow and convert the flow consistency equations into a learning objective, akin to the casting of the Bellman equations into Temporal Difference methods. We prove that any global minimum of the proposed objectives yields a policy which samples from the desired distribution, and demonstrate the improved performance and diversity of GFlowNet on a simple domain where there are many modes to the reward function, and on a molecule synthesis task."
                    },
                    {
                        "title": "Mechanism of action of HBV capsid assembly modulators predicted from binding to early assembly intermediates",
                        "abstract": "Interfering with the self-assembly of virus nucleocapsids is a promising approach for the development of novel antiviral agents. Applied to hepatitis B virus (HBV), this approach has led to several classes of capsid assembly modulators (CAMs) that target the virus by either accelerating nucleocapsid assembly or misdirecting it into non-capsid-like particles. Here, we have assessed the structures of early nucleocapsid assembly intermediates, with and without bound CAMs, using molecular dynamics simulations. We find that distinct conformations of the intermediates are induced depending on whether the bound CAM accelerates or misdirects assembly; these structures are predictive of the final assembly. We also selected non-capsid-like structures from our simulations for virtual screening, resulting in the discovery of several compounds with moderate anti-viral activity and low toxicity. Cryo-electron microscopy and capsid melting experiments suggest that our compounds possess a novel mechanism for assembly modulation, potentially opening new avenues for HBV inhibition."
                    },
                    {
                        "title": "RetroGNN: Approximating Retrosynthesis by Graph Neural Networks for De Novo Drug Design",
                        "abstract": "De novo molecule generation often results in chemically unfeasible molecules. A natural idea to mitigate this problem is to bias the search process towards more easily synthesizable molecules using a proxy for synthetic accessibility. However, using currently available proxies still results in highly unrealistic compounds. We investigate the feasibility of training deep graph neural networks to approximate the outputs of a retrosynthesis planning software, and their use to bias the search process. We evaluate our method on a benchmark involving searching for drug-like molecules with antibiotic properties. Compared to enumerating over five million existing molecules from the ZINC database, our approach finds molecules predicted to be more likely to be antibiotics while maintaining good drug-like properties and being easily synthesizable. Importantly, our deep neural network can successfully filter out hard to synthesize molecules while achieving a $10^5$ times speed-up over using the retrosynthesis planning software."
                    },
                    {
                        "title": "Capsule networks for low-data transfer learning",
                        "abstract": "We propose a capsule network-based architecture for generalizing learning to new data with few examples. Using both generative and non-generative capsule networks with intermediate routing, we are able to generalize to new information over 25 times faster than a similar convolutional neural network. We train the networks on the multiMNIST dataset lacking one digit. After the networks reach their maximum accuracy, we inject 1-100 examples of the missing digit into the training set, and measure the number of batches needed to return to a comparable level of accuracy. We then discuss the improvement in low-data transfer learning that capsule networks bring, and propose future directions for capsule research."
                    }
                ]
            },
            "513c5641-817a-46a3-a859-b659f6216382": {
                "pk": "513c5641-817a-46a3-a859-b659f6216382",
                "name": "Michael Bronstein",
                "collaborators": [
                    "Jarrid Rector-Brooks",
                    "Cheng-Hao Liu",
                    "Y. Bengio",
                    "Alexander Tong",
                    "A. Bose",
                    "Mohsin Hasan",
                    "Zhangzhi Peng",
                    "Zachary Quinn",
                    "Sarthak Mittal",
                    "Nouha Dziri"
                ],
                "domain": [
                    "Generative Modeling",
                    "Protein Structure Prediction",
                    "Graph Neural Network",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Steering Masked Discrete Diffusion Models via Discrete Denoising Posterior Prediction",
                        "abstract": "Generative modeling of discrete data underlies important applications spanning text-based agents like ChatGPT to the design of the very building blocks of life in protein sequences. However, application domains need to exert control over the generated data by steering the generative process - typically via RLHF - to satisfy a specified property, reward, or affinity metric. In this paper, we study the problem of steering Masked Diffusion Models (MDMs), a recent class of discrete diffusion models that offer a compelling alternative to traditional autoregressive models. We introduce Discrete Denoising Posterior Prediction (DDPP), a novel framework that casts the task of steering pre-trained MDMs as a problem of probabilistic inference by learning to sample from a target Bayesian posterior. Our DDPP framework leads to a family of three novel objectives that are all simulation-free, and thus scalable while applying to general non-differentiable reward functions. Empirically, we instantiate DDPP by steering MDMs to perform class-conditional pixel-level image modeling, RLHF-based alignment of MDMs using text-based rewards, and finetuning protein language models to generate more diverse secondary structures and shorter proteins. We substantiate our designs via wet-lab validation, where we observe transient expression of reward-optimized protein sequences."
                    },
                    {
                        "title": "Sequence-Augmented SE(3)-Flow Matching For Conditional Protein Backbone Generation",
                        "abstract": "Proteins are essential for almost all biological processes and derive their diverse functions from complex 3D structures, which are in turn determined by their amino acid sequences. In this paper, we exploit the rich biological inductive bias of amino acid sequences and introduce FoldFlow-2, a novel sequence-conditioned SE(3)-equivariant flow matching model for protein structure generation. FoldFlow-2 presents substantial new architectural features over the previous FoldFlow family of models including a protein large language model to encode sequence, a new multi-modal fusion trunk that combines structure and sequence representations, and a geometric transformer based decoder. To increase diversity and novelty of generated samples -- crucial for de-novo drug design -- we train FoldFlow-2 at scale on a new dataset that is an order of magnitude larger than PDB datasets of prior works, containing both known proteins in PDB and high-quality synthetic structures achieved through filtering. We further demonstrate the ability to align FoldFlow-2 to arbitrary rewards, e.g. increasing secondary structures diversity, by introducing a Reinforced Finetuning (ReFT) objective. We empirically observe that FoldFlow-2 outperforms previous state-of-the-art protein structure-based generative models, improving over RFDiffusion in terms of unconditional generation across all metrics including designability, diversity, and novelty across all protein lengths, as well as exhibiting generalization on the task of equilibrium conformation sampling. Finally, we demonstrate that a fine-tuned FoldFlow-2 makes progress on challenging conditional design tasks such as designing scaffolds for the VHH nanobody."
                    },
                    {
                        "title": "Generative Active Learning for the Search of Small-molecule Protein Binders",
                        "abstract": "Despite substantial progress in machine learning for scientific discovery in recent years, truly de novo design of small molecules which exhibit a property of interest remains a significant challenge. We introduce LambdaZero, a generative active learning approach to search for synthesizable molecules. Powered by deep reinforcement learning, LambdaZero learns to search over the vast space of molecules to discover candidates with a desired property. We apply LambdaZero with molecular docking to design novel small molecules that inhibit the enzyme soluble Epoxide Hydrolase 2 (sEH), while enforcing constraints on synthesizability and drug-likeliness. LambdaZero provides an exponential speedup in terms of the number of calls to the expensive molecular docking oracle, and LambdaZero de novo designed molecules reach docking scores that would otherwise require the virtual screening of a hundred billion molecules. Importantly, LambdaZero discovers novel scaffolds of synthesizable, drug-like inhibitors for sEH. In in vitro experimental validation, a series of ligands from a generated quinazoline-based scaffold were synthesized, and the lead inhibitor N-(4,6-di(pyrrolidin-1-yl)quinazolin-2-yl)-N-methylbenzamide (UM0152893) displayed sub-micromolar enzyme inhibition of sEH."
                    },
                    {
                        "title": "Using machine learning to decode animal communication",
                        "abstract": "Description New methods promise transformative insights and conservation benefits The past few years have seen a surge of interest in using machine learning (ML) methods for studying the behavior of nonhuman animals (hereafter \u201canimals\u201d) (1). A topic that has attracted particular attention is the decoding of animal communication systems using deep learning and other approaches (2). Now is the time to tackle challenges concerning data availability, model validation, and research ethics, and to embrace opportunities for building collaborations across disciplines and initiatives."
                    },
                    {
                        "title": "A Hitchhiker's Guide to Geometric GNNs for 3D Atomic Systems",
                        "abstract": "Recent advances in computational modelling of atomic systems, spanning molecules, proteins, and materials, represent them as geometric graphs with atoms embedded as nodes in 3D Euclidean space. In these graphs, the geometric attributes transform according to the inherent physical symmetries of 3D atomic systems, including rotations and translations in Euclidean space, as well as node permutations. In recent years, Geometric Graph Neural Networks have emerged as the preferred machine learning architecture powering applications ranging from protein structure prediction to molecular simulations and material generation. Their specificity lies in the inductive biases they leverage - such as physical symmetries and chemical properties - to learn informative representations of these geometric graphs. In this opinionated paper, we provide a comprehensive and self-contained overview of the field of Geometric GNNs for 3D atomic systems. We cover fundamental background material and introduce a pedagogical taxonomy of Geometric GNN architectures: (1) invariant networks, (2) equivariant networks in Cartesian basis, (3) equivariant networks in spherical basis, and (4) unconstrained networks. Additionally, we outline key datasets and application areas and suggest future research directions. The objective of this work is to present a structured perspective on the field, making it accessible to newcomers and aiding practitioners in gaining an intuition for its mathematical abstractions."
                    }
                ]
            },
            "4c1016ba-018a-417c-888b-980eb091bb27": {
                "pk": "4c1016ba-018a-417c-888b-980eb091bb27",
                "name": "Alexander Tong",
                "collaborators": [
                    "A. Bose",
                    "Guillaume Huguet",
                    "Jarrid Rector-Brooks",
                    "Cheng-Hao Liu",
                    "Michael M. Bronstein",
                    "Y. Bengio",
                    "Guy Wolf",
                    "Sarthak Mittal",
                    "Kilian Fatras",
                    "Pablo Lemos"
                ],
                "domain": [
                    "Generative Modeling",
                    "Protein Structure",
                    "Machine Learning",
                    "Stochastic Dynamics"
                ],
                "publications": [
                    {
                        "title": "Steering Masked Discrete Diffusion Models via Discrete Denoising Posterior Prediction",
                        "abstract": "Generative modeling of discrete data underlies important applications spanning text-based agents like ChatGPT to the design of the very building blocks of life in protein sequences. However, application domains need to exert control over the generated data by steering the generative process - typically via RLHF - to satisfy a specified property, reward, or affinity metric. In this paper, we study the problem of steering Masked Diffusion Models (MDMs), a recent class of discrete diffusion models that offer a compelling alternative to traditional autoregressive models. We introduce Discrete Denoising Posterior Prediction (DDPP), a novel framework that casts the task of steering pre-trained MDMs as a problem of probabilistic inference by learning to sample from a target Bayesian posterior. Our DDPP framework leads to a family of three novel objectives that are all simulation-free, and thus scalable while applying to general non-differentiable reward functions. Empirically, we instantiate DDPP by steering MDMs to perform class-conditional pixel-level image modeling, RLHF-based alignment of MDMs using text-based rewards, and finetuning protein language models to generate more diverse secondary structures and shorter proteins. We substantiate our designs via wet-lab validation, where we observe transient expression of reward-optimized protein sequences."
                    },
                    {
                        "title": "Sequence-Augmented SE(3)-Flow Matching For Conditional Protein Backbone Generation",
                        "abstract": "Proteins are essential for almost all biological processes and derive their diverse functions from complex 3D structures, which are in turn determined by their amino acid sequences. In this paper, we exploit the rich biological inductive bias of amino acid sequences and introduce FoldFlow-2, a novel sequence-conditioned SE(3)-equivariant flow matching model for protein structure generation. FoldFlow-2 presents substantial new architectural features over the previous FoldFlow family of models including a protein large language model to encode sequence, a new multi-modal fusion trunk that combines structure and sequence representations, and a geometric transformer based decoder. To increase diversity and novelty of generated samples -- crucial for de-novo drug design -- we train FoldFlow-2 at scale on a new dataset that is an order of magnitude larger than PDB datasets of prior works, containing both known proteins in PDB and high-quality synthetic structures achieved through filtering. We further demonstrate the ability to align FoldFlow-2 to arbitrary rewards, e.g. increasing secondary structures diversity, by introducing a Reinforced Finetuning (ReFT) objective. We empirically observe that FoldFlow-2 outperforms previous state-of-the-art protein structure-based generative models, improving over RFDiffusion in terms of unconditional generation across all metrics including designability, diversity, and novelty across all protein lengths, as well as exhibiting generalization on the task of equilibrium conformation sampling. Finally, we demonstrate that a fine-tuned FoldFlow-2 makes progress on challenging conditional design tasks such as designing scaffolds for the VHH nanobody."
                    },
                    {
                        "title": "Metric Flow Matching for Smooth Interpolations on the Data Manifold",
                        "abstract": "Matching objectives underpin the success of modern generative models and rely on constructing conditional paths that transform a source distribution into a target distribution. Despite being a fundamental building block, conditional paths have been designed principally under the assumption of Euclidean geometry, resulting in straight interpolations. However, this can be particularly restrictive for tasks such as trajectory inference, where straight paths might lie outside the data manifold, thus failing to capture the underlying dynamics giving rise to the observed marginals. In this paper, we propose Metric Flow Matching (MFM), a novel simulation-free framework for conditional flow matching where interpolants are approximate geodesics learned by minimizing the kinetic energy of a data-induced Riemannian metric. This way, the generative model matches vector fields on the data manifold, which corresponds to lower uncertainty and more meaningful interpolations. We prescribe general metrics to instantiate MFM, independent of the task, and test it on a suite of challenging problems including LiDAR navigation, unpaired image translation, and modeling cellular dynamics. We observe that MFM outperforms the Euclidean baselines, particularly achieving SOTA on single-cell trajectory prediction."
                    },
                    {
                        "title": "Iterated Denoising Energy Matching for Sampling from Boltzmann Densities",
                        "abstract": "Efficiently generating statistically independent samples from an unnormalized probability distribution, such as equilibrium samples of many-body systems, is a foundational problem in science. In this paper, we propose Iterated Denoising Energy Matching (iDEM), an iterative algorithm that uses a novel stochastic score matching objective leveraging solely the energy function and its gradient -- and no data samples -- to train a diffusion-based sampler. Specifically, iDEM alternates between (I) sampling regions of high model density from a diffusion-based sampler and (II) using these samples in our stochastic matching objective to further improve the sampler. iDEM is scalable to high dimensions as the inner matching objective, is simulation-free, and requires no MCMC samples. Moreover, by leveraging the fast mode mixing behavior of diffusion, iDEM smooths out the energy landscape enabling efficient exploration and learning of an amortized sampler. We evaluate iDEM on a suite of tasks ranging from standard synthetic energy functions to invariant $n$-body particle systems. We show that the proposed approach achieves state-of-the-art performance on all metrics and trains $2-5\\times$ faster, which allows it to be the first method to train using energy on the challenging $55$-particle Lennard-Jones system."
                    },
                    {
                        "title": "Assessing Neural Network Representations During Training Using Noise-Resilient Diffusion Spectral Entropy",
                        "abstract": "Entropy and mutual information in neural networks provide rich information on the learning process, but they have proven difficult to compute reliably in high dimensions. Indeed, in noisy and high-dimensional data, traditional estimates in ambient dimensions approach a fixed entropy and are prohibitively hard to compute. To address these issues, we leverage data geometry to access the underlying manifold and reliably compute these information-theoretic measures. Specifically, we define diffusion spectral entropy (DSE) in neural representations of a dataset as well as diffusion spectral mutual information (DSMI) between different variables representing data. First, we show that they form noise-resistant measures of intrinsic dimensionality and relationship strength in high-dimensional simulated data that outperform classic Shannon entropy, nonparametric estimation, and mutual information neural estimation (MINE). We then study the evolution of representations in classification networks with supervised learning, self-supervision, or overfitting. We observe that (1) DSE of neural representations increases during training; (2) DSMI with the class label increases during generalizable learning but stays stagnant during overfitting; (3) DSMI with the input signal shows differing trends: on MNIST it increases, while on CIFAR-10 and STL-10 it decreases. Finally, we show that DSE can be used to guide better network initialization and that DSMI can be used to predict downstream classification accuracy across 962 models on ImageNet."
                    },
                    {
                        "title": "Geodesic Sinkhorn: optimal transport for high-dimensional datasets",
                        "abstract": ". Understanding the dynamics and reactions of cells from population snapshots is a major challenge in single-cell transcriptomics. Here, we present Geodesic Sinkhorn, a method for interpolating populations along a data manifold that leverages existing kernels developed for single-cell dimensionality reduction and visualization methods. Our Geodesic Sinkhorn method uses a heat-geodesic ground distance that, as compared to Euclidean ground distances, is more accurate for interpolating single-cell dynamics on a wide variety of datasets and signi\ufb01cantly speeds up the computation for sparse kernels. We \ufb01rst apply Geodesic Sinkhorn to 10 single-cell transcriptomics time series interpolation datasets as a drop-in replacement for existing interpolation methods where it outperforms on all datasets, showing its e\ufb00ectiveness in modeling cell dynamics. Second, we show how to e\ufb03ciently approximate the operator with polynomial kernels allowing us to improve scaling to large datasets. Finally, we de\ufb01ne the conditional Wasserstein-average treatment e\ufb00ect and show how it can elucidate the treatment e\ufb00ect on single-cell populations on a drug screen."
                    },
                    {
                        "title": "Simulation-Free Schr\u00a8odinger Bridges via Score and Flow Matching",
                        "abstract": "We present simulation-free score and flow matching ([SF] 2 M), a simulation-free objective for inferring stochastic dynamics given unpaired source and target samples drawn from arbitrary distributions. Our method generalizes both the score-matching loss used in the training of diffusion models and the recently proposed flow matching loss used in the training of continuous normalizing flows. [SF] 2 M interprets continuous-time stochastic generative modeling as a Schr \u00a8 odinger bridge (SB) problem. It relies on static entropy-regularized optimal transport, or a minibatch approximation, to efficiently learn the SB without simulating the learned stochastic process. We find that [SF] 2 M is more efficient and gives more accurate solutions to the SB problem than simulation-based methods from prior work. Finally, we apply [SF] 2 M to the problem of learning cell dynamics from snapshot data. Notably, [SF] 2 M is the first method to accurately model cell dynamics in high dimensions and can recover known gene regulatory networks from simulated data. Code: https://github.com/atong01/ conditional-flow-matching ."
                    }
                ]
            }
        }
    },
    "2407.19474": {
        "paper_data": {
            "title": "Visual Riddles: a Commonsense and World Knowledge Challenge for Large Vision and Language Models",
            "url": "http://arxiv.org/abs/2407.19474v1",
            "arxiv_id": "2407.19474",
            "authors": [
                "Nitzan Bitton-Guetta",
                "Aviv Slobodkin",
                "Aviya Maimon",
                "Eliya Habba",
                "Royi Rassin",
                "Yonatan Bitton",
                "Idan Szpektor",
                "Amir Globerson",
                "Yuval Elovici"
            ],
            "abstract": "Imagine observing someone scratching their arm; to understand why, additional context would be necessary. However, spotting a mosquito nearby would immediately offer a likely explanation for the person's discomfort, thereby alleviating the need for further information. This example illustrates how subtle visual cues can challenge our cognitive skills and demonstrates the complexity of interpreting visual scenarios. To study these skills, we present Visual Riddles, a benchmark aimed to test vision and language models on visual riddles requiring commonsense and world knowledge. The benchmark comprises 400 visual riddles, each featuring a unique image created by a variety of text-to-image models, question, ground-truth answer, textual hint, and attribution. Human evaluation reveals that existing models lag significantly behind human performance, which is at 82\\% accuracy, with Gemini-Pro-1.5 leading with 40\\% accuracy. Our benchmark comes with automatic evaluation tasks to make assessment scalable. These findings underscore the potential of Visual Riddles as a valuable resource for enhancing vision and language models' capabilities in interpreting complex visual scenarios.",
            "introduction": "   1 Introduction  Humans intuitively utilize commonsense reasoning to interpret complex elements in visual scenes, a skill that current vision and language models frequently lack. For instance, the simple act of a person scratching their arm gains added significance when a mosquito is present on a nearby nightstand, as depicted in\u00a0Fig.\u00a01. While humans easily recognize such contextual nuances, existing image-understanding models struggle to integrate visual cues with world knowledge stemming from cultural aspects, life-experiences, and physical or social knowledge\u00a0[1, 2, 3, 4]. This gap has spurred the development of various benchmarks like VisIT-Bench [5] and Encyclopedic VQA [6], which rely on pre-existing images to formulate challenging questions. While effective, this approach risks using images seen during pre-training of large vision-language models and restricts the scope of scenarios to those already captured, potentially limiting creativity and challenge variety.   To address these shortcoming, we introduce Visual Riddles, a benchmark comprising 400 visual riddles, each featuring a question and a synthetic image generated specifically for the challenge. The process of creating a riddle involves designing a scenario with embedded clues that appear as natural elements, then translating this scenario into both an image and a question (\u00a72). For instance, in the top example of\u00a0Fig.\u00a01, a seemingly inconspicuous mosquito acts as a crucial clue for the source of the person\u2019s discomfort. This method, going from scenario to image, and not the other way around, not only enhances creative liberty but also broadens the range of everyday situations explored, challenging both humans and models in their interpretative and commonsense reasoning abilities. The benchmark also incorporates textual hints and attributions (see \u00a0Fig.\u00a01) to direct focus to clues and provide external sources of verification for answers, thus expanding the challenge\u2019s scope and depth.   Our benchmark\u2019s main task (\u00a73) involves solving riddles in an open-ended visual question answering (VQA) format, which takes as input an image and a question, and expects a free-text answer. This setup evaluates the ability to detect subtle visual cues and apply commonsense reasoning or world knowledge to formulate answers. Additionally, we investigate the impact of including hints and attributions in the input to enhance comprehension.   Yet, as in every QA benchmark, evaluation is a key challenge in this setting. To facilitate scalable research, we introduce two additional tasks. The first is a multiple-choice version of the main task, also for the hint- and attribution-assisted variants, allowing for easy automatic scoring based on accuracy. The second task assesses the ability of models to determine the accuracy of open-ended responses in two settings: reference-free, where models evaluate responses based only on the image and the question, and reference-based, where the correct answer is also provided. This task suggests auto-raters to evaluate our riddles, and aim to advance research on such automatic evaluation methods.   Experimental results (\u00a74) reveal a significant gap in performance between humans and state-of-the-art vision language models, with the top-performing model, Gemini-Pro-1.5 [7], only achieving 40% accuracy versus humans\u2019 82%. Surprisingly, the multiple-choice format proved almost as difficult, with performances slightly superior to those in the open-ended task, but with notably high performances when auxiliary data is included. Automated tests, both reference-free and reference-based, show that Gemini-Pro-1.5, the top auto-rater,",
            "references": [
                {
                    "title": "Vibe-Eval: A hard evaluation suite for measuring progress of multimodal language models",
                    "abstract": "We introduce Vibe-Eval: a new open benchmark and framework for evaluating multimodal chat models. Vibe-Eval consists of 269 visual understanding prompts, including 100 of hard difficulty, complete with gold-standard responses authored by experts. Vibe-Eval is open-ended and challenging with dual objectives: (i) vibe checking multimodal chat models for day-to-day tasks and (ii) rigorously testing and probing the capabilities of present frontier models. Notably, our hard set contains>50% questions that all frontier models answer incorrectly. We explore the nuances of designing, evaluating, and ranking models on ultra challenging prompts. We also discuss trade-offs between human and automatic evaluation, and show that automatic model evaluation using Reka Core roughly correlates to human judgment. We offer free API access for the purpose of lightweight evaluation and plan to conduct formal human evaluations for public models that perform well on the Vibe-Eval's automatic scores. We release the evaluation code and data, see https://github.com/reka-ai/reka-vibe-eval"
                },
                {
                    "title": "MileBench: Benchmarking MLLMs in Long Context",
                    "abstract": "Despite the advancements and impressive performance of Multimodal Large Language Models (MLLMs) on benchmarks, their effectiveness in real-world, long-context, and multi-image tasks is unclear due to the benchmarks' limited scope. Existing benchmarks often focus on single-image and short-text samples, and when assessing multi-image tasks, they either limit the image count or focus on specific task (e.g time-series captioning), potentially obscuring the performance challenges of MLLMs. To address these limitations, we introduce MileBench, a pioneering benchmark designed to test the MultImodal Long-contExt capabilities of MLLMs. This benchmark comprises not only multimodal long contexts, but also multiple tasks requiring both comprehension and generation. We establish two distinct evaluation sets, diagnostic and realistic, to systematically assess MLLMs' long-context adaptation capacity and their ability to complete tasks in long-context scenarios. Our experimental results, obtained from testing 22 models, revealed that while the closed-source GPT-4o outperforms others, most open-source MLLMs struggle in long-context situations. Interestingly, the performance gap tends to widen with an increase in the number of images. We strongly encourage an intensification of research efforts towards enhancing MLLMs' long-context capabilities, especially in scenarios involving multiple images."
                },
                {
                    "title": "BLINK: Multimodal Large Language Models Can See but Not Perceive",
                    "abstract": "We introduce Blink, a new benchmark for multimodal language models (LLMs) that focuses on core visual perception abilities not found in other evaluations. Most of the Blink tasks can be solved by humans\"within a blink\"(e.g., relative depth estimation, visual correspondence, forensics detection, and multi-view reasoning). However, we find these perception-demanding tasks cast significant challenges for current multimodal LLMs because they resist mediation through natural language. Blink reformats 14 classic computer vision tasks into 3,807 multiple-choice questions, paired with single or multiple images and visual prompting. While humans get 95.70% accuracy on average, Blink is surprisingly challenging for existing multimodal LLMs: even the best-performing GPT-4V and Gemini achieve accuracies of 51.26% and 45.72%, only 13.17% and 7.63% higher than random guessing, indicating that such perception abilities have not\"emerged\"yet in recent multimodal LLMs. Our analysis also highlights that specialist CV models could solve these problems much better, suggesting potential pathways for future improvements. We believe Blink will stimulate the community to help multimodal LLMs catch up with human-level visual perception."
                },
                {
                    "title": "Multi-Review Fusion-in-Context",
                    "abstract": "Grounded text generation, encompassing tasks such as long-form question-answering and summarization, necessitates both content selection and content consolidation. Current end-to-end methods are difficult to control and interpret due to their opaqueness. Accordingly, recent works have proposed a modular approach, with separate components for each step. Specifically, we focus on the second subtask, of generating coherent text given pre-selected content in a multi-document setting. Concretely, we formalize Fusion-in-Context (FiC) as a standalone task, whose input consists of source texts with highlighted spans of targeted content. A model then needs to generate a coherent passage that includes all and only the target information. Our work includes the development of a curated dataset of 1000 instances in the reviews domain, alongside a novel evaluation framework for assessing the faithfulness and coverage of highlights, which strongly correlate to human judgment. Several baseline models exhibit promising outcomes and provide insightful analyses. This study lays the groundwork for further exploration of modular text generation in the multi-document setting, offering potential improvements in the quality and reliability of generated content. Our benchmark, FuseReviews, including the dataset, evaluation framework, and designated leaderboard, can be found at https://fusereviews.github.io/."
                },
                {
                    "title": "AVIBench: Towards Evaluating the Robustness of Large Vision-Language Model on Adversarial Visual-Instructions",
                    "abstract": "Large Vision-Language Models (LVLMs) have shown significant progress in well responding to visual-instructions from users. However, these instructions, encompassing images and text, are susceptible to both intentional and inadvertent attacks. Despite the critical importance of LVLMs' robustness against such threats, current research in this area remains limited. To bridge this gap, we introduce AVIBench, a framework designed to analyze the robustness of LVLMs when facing various adversarial visual-instructions (AVIs), including four types of image-based AVIs, ten types of text-based AVIs, and nine types of content bias AVIs (such as gender, violence, cultural, and racial biases, among others). We generate 260K AVIs encompassing five categories of multimodal capabilities (nine tasks) and content bias. We then conduct a comprehensive evaluation involving 14 open-source LVLMs to assess their performance. AVIBench also serves as a convenient tool for practitioners to evaluate the robustness of LVLMs against AVIs. Our findings and extensive experimental results shed light on the vulnerabilities of LVLMs, and highlight that inherent biases exist even in advanced closed-source LVLMs like GeminiProVision and GPT-4V. This underscores the importance of enhancing the robustness, security, and fairness of LVLMs. The source code and benchmark will be made publicly available."
                },
                {
                    "title": "Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context",
                    "abstract": "In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content."
                },
                {
                    "title": "LCM-LoRA: A Universal Stable-Diffusion Acceleration Module",
                    "abstract": "Latent Consistency Models (LCMs) have achieved impressive performance in accelerating text-to-image generative tasks, producing high-quality images with minimal inference steps. LCMs are distilled from pre-trained latent diffusion models (LDMs), requiring only ~32 A100 GPU training hours. This report further extends LCMs' potential in two aspects: First, by applying LoRA distillation to Stable-Diffusion models including SD-V1.5, SSD-1B, and SDXL, we have expanded LCM's scope to larger models with significantly less memory consumption, achieving superior image generation quality. Second, we identify the LoRA parameters obtained through LCM distillation as a universal Stable-Diffusion acceleration module, named LCM-LoRA. LCM-LoRA can be directly plugged into various Stable-Diffusion fine-tuned models or LoRAs without training, thus representing a universally applicable accelerator for diverse image generation tasks. Compared with previous numerical PF-ODE solvers such as DDIM, DPM-Solver, LCM-LoRA can be viewed as a plug-in neural PF-ODE solver that possesses strong generalization abilities. Project page: https://github.com/luosiallen/latent-consistency-model."
                },
                {
                    "title": "VisIT-Bench: A Benchmark for Vision-Language Instruction Following Inspired by Real-World Use",
                    "abstract": "We introduce VisIT-Bench (Visual InsTruction Benchmark), a benchmark for evaluation of instruction-following vision-language models for real-world use. Our starting point is curating 70 'instruction families' that we envision instruction tuned vision-language models should be able to address. Extending beyond evaluations like VQAv2 and COCO, tasks range from basic recognition to game playing and creative generation. Following curation, our dataset comprises 592 test queries, each with a human-authored instruction-conditioned caption. These descriptions surface instruction-specific factors, e.g., for an instruction asking about the accessibility of a storefront for wheelchair users, the instruction-conditioned caption describes ramps/potential obstacles. These descriptions enable 1) collecting human-verified reference outputs for each instance; and 2) automatic evaluation of candidate multimodal generations using a text-only LLM, aligning with human judgment. We quantify quality gaps between models and references using both human and automatic evaluations; e.g., the top-performing instruction-following model wins against the GPT-4 reference in just 27% of the comparison. VisIT-Bench is dynamic to participate, practitioners simply submit their model's response on the project website; Data, code and leaderboard is available at visit-bench.github.io."
                },
                {
                    "title": "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis",
                    "abstract": "We present SDXL, a latent diffusion model for text-to-image synthesis. Compared to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet backbone: The increase of model parameters is mainly due to more attention blocks and a larger cross-attention context as SDXL uses a second text encoder. We design multiple novel conditioning schemes and train SDXL on multiple aspect ratios. We also introduce a refinement model which is used to improve the visual fidelity of samples generated by SDXL using a post-hoc image-to-image technique. We demonstrate that SDXL shows drastically improved performance compared the previous versions of Stable Diffusion and achieves results competitive with those of black-box state-of-the-art image generators. In the spirit of promoting open research and fostering transparency in large model training and evaluation, we provide access to code and model weights at https://github.com/Stability-AI/generative-models"
                },
                {
                    "title": "What You See is What You Read? Improving Text-Image Alignment Evaluation",
                    "abstract": "Automatically determining whether a text and a corresponding image are semantically aligned is a significant challenge for vision-language models, with applications in generative text-to-image and image-to-text tasks. In this work, we study methods for automatic text-image alignment evaluation. We first introduce SeeTRUE: a comprehensive evaluation set, spanning multiple datasets from both text-to-image and image-to-text generation tasks, with human judgements for whether a given text-image pair is semantically aligned. We then describe two automatic methods to determine alignment: the first involving a pipeline based on question generation and visual question answering models, and the second employing an end-to-end classification approach by finetuning multimodal pretrained models. Both methods surpass prior approaches in various text-image alignment tasks, with significant improvements in challenging cases that involve complex composition or unnatural images. Finally, we demonstrate how our approaches can localize specific misalignments between an image and a given text, and how they can be used to automatically re-rank candidates in text-to-image generation."
                },
                {
                    "title": "TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering",
                    "abstract": "Despite thousands of researchers, engineers, and artists actively working on improving text-to-image generation models, systems often fail to produce images that accurately align with the text inputs. We introduce TIFA (Text-to-Image Faithfulness evaluation with question Answering), an automatic evaluation metric that measures the faithfulness of a generated image to its text input via visual question answering (VQA). Specifically, given a text input, we automatically generate several question-answer pairs using a language model. We calculate image faithfulness by checking whether existing VQA models can answer these questions using the generated image. TIFA is a reference-free metric that allows for fine-grained and interpretable evaluations of generated images. TIFA also has better correlations with human judgments than existing metrics. Based on this approach, we introduce TIFA v1.0, a benchmark consisting of 4K diverse text inputs and 25K questions across 12 categories (object, counting, etc.). We present a comprehensive evaluation of existing text-to-image models using TIFA v1.0 and highlight the limitations and challenges of current models. For instance, we find that current text-to-image models, despite doing well on color and material, still struggle in counting, spatial relations, and composing multiple objects. We hope our benchmark will help carefully measure the research progress in text-to-image synthesis and provide valuable insights for further research. 1"
                },
                {
                    "title": "Breaking Common Sense: WHOOPS! A Vision-and-Language Benchmark of Synthetic and Compositional Images",
                    "abstract": "Weird, unusual, and uncanny images pique the curiosity of observers because they challenge commonsense. For example, an image released during the 2022 world cup depicts the famous soccer stars Lionel Messi and Cristiano Ronaldo playing chess, which playfully violates our expectation that their competition should occur on the football field.1 Humans can easily recognize and interpret these unconventional images, but can AI models do the same? We introduce WHOOPS!, a new dataset and benchmark for visual commonsense. The dataset is comprised of purposefully commonsense-defying images created by designers using publicly-available image generation tools like Midjourney. We consider several tasks posed over the dataset. In addition to image captioning, cross-modal matching, and visual question answering, we introduce a difficult explanation generation task, where models must identify and explain why a given image is unusual. Our results show that state-of-the-art models such as GPT3 and BLIP2 still lag behind human performance on WHOOPS!. We hope our dataset will inspire the development of AI models with stronger visual commonsense reasoning abilities. 2"
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "VASR: Visual Analogies of Situation Recognition",
                    "abstract": "A core process in human cognition is analogical mapping: the ability to identify a similar relational structure between different situations.\nWe introduce a novel task, Visual Analogies of Situation Recognition, adapting the classical word-analogy task into the visual domain. Given a triplet of images, the task is to select an image candidate B' that completes the analogy (A to A' is like B to what?). Unlike previous work on visual analogy that focused on simple image transformations, we tackle complex analogies requiring understanding of scenes. \n\nWe leverage situation recognition annotations and the CLIP model to generate a large set of 500k candidate analogies. Crowdsourced annotations for a sample of the data indicate that humans agree with the dataset label ~80% of the time (chance level 25%). Furthermore, we use human annotations to create a gold-standard dataset of 3,820 validated analogies.\nOur experiments demonstrate that state-of-the-art models do well when distractors are chosen randomly (~86%), but struggle with carefully chosen distractors (~53%, compared to 90% human accuracy). We hope our dataset will encourage the development of new analogy-making models. Website: https://vasr-dataset.github.io/"
                },
                {
                    "title": "Do Androids Laugh at Electric Sheep? Humor \u201cUnderstanding\u201d Benchmarks from The New Yorker Caption Contest",
                    "abstract": "Large neural networks can now generate jokes, but do they really \u201cunderstand\u201d humor? We challenge AI models with three tasks derived from the New Yorker Cartoon Caption Contest: matching a joke to a cartoon, identifying a winning caption, and explaining why a winning caption is funny. These tasks encapsulate progressively more sophisticated aspects of \u201cunderstanding\u201d a cartoon; key elements are the complex, often surprising relationships between images and captions and the frequent inclusion of indirect and playful allusions to human experience and culture. We investigate both multimodal and language-only models: the former are challenged with the cartoon images directly, while the latter are given multifaceted descriptions of the visual scene to simulate human-level visual understanding. We find that both types of models struggle at all three tasks. For example, our best multimodal models fall 30 accuracy points behind human performance on the matching task, and, even when provided ground-truth visual scene descriptors, human-authored explanations are preferred head-to-head over the best machine-authored ones (few-shot GPT-4) in more than 2/3 of cases. We release models, code, leaderboard, and corpus, which includes newly-gathered annotations describing the image\u2019s locations/entities, what\u2019s unusual in the scene, and an explanation of the joke."
                },
                {
                    "title": "WinoGAViL: Gamified Association Benchmark to Challenge Vision-and-Language Models",
                    "abstract": "While vision-and-language models perform well on tasks such as visual question answering, they struggle when it comes to basic human commonsense reasoning skills. In this work, we introduce WinoGAViL: an online game of vision-and-language associations (e.g., between werewolves and a full moon), used as a dynamic evaluation benchmark. Inspired by the popular card game Codenames, a spymaster gives a textual cue related to several visual candidates, and another player tries to identify them. Human players are rewarded for creating associations that are challenging for a rival AI model but still solvable by other human players. We use the game to collect 3.5K instances, finding that they are intuitive for humans (>90% Jaccard index) but challenging for state-of-the-art AI models, where the best model (ViLT) achieves a score of 52%, succeeding mostly where the cue is visually salient. Our analysis as well as the feedback we collect from players indicate that the collected associations require diverse reasoning skills, including general knowledge, common sense, abstraction, and more. We release the dataset, the code and the interactive game, allowing future data collection that can be used to develop models with better association abilities."
                },
                {
                    "title": "Flamingo: a Visual Language Model for Few-Shot Learning",
                    "abstract": "Building models that can be rapidly adapted to novel tasks using only a handful of annotated examples is an open challenge for multimodal machine learning research. We introduce Flamingo, a family of Visual Language Models (VLM) with this ability. We propose key architectural innovations to: (i) bridge powerful pretrained vision-only and language-only models, (ii) handle sequences of arbitrarily interleaved visual and textual data, and (iii) seamlessly ingest images or videos as inputs. Thanks to their flexibility, Flamingo models can be trained on large-scale multimodal web corpora containing arbitrarily interleaved text and images, which is key to endow them with in-context few-shot learning capabilities. We perform a thorough evaluation of our models, exploring and measuring their ability to rapidly adapt to a variety of image and video tasks. These include open-ended tasks such as visual question-answering, where the model is prompted with a question which it has to answer; captioning tasks, which evaluate the ability to describe a scene or an event; and close-ended tasks such as multiple-choice visual question-answering. For tasks lying anywhere on this spectrum, a single Flamingo model can achieve a new state of the art with few-shot learning, simply by prompting the model with task-specific examples. On numerous benchmarks, Flamingo outperforms models fine-tuned on thousands of times more task-specific data."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "SummaC: Re-Visiting NLI-based Models for Inconsistency Detection in Summarization",
                    "abstract": "In the summarization domain, a key requirement for summaries is to be factually consistent with the input document. Previous work has found that natural language inference (NLI) models do not perform competitively when applied to inconsistency detection. In this work, we revisit the use of NLI for inconsistency detection, finding that past work suffered from a mismatch in input granularity between NLI datasets (sentence-level), and inconsistency detection (document level). We provide a highly effective and light-weight method called SummaCConv that enables NLI models to be successfully used for this task by segmenting documents into sentence units and aggregating scores between pairs of sentences. We furthermore introduce a new benchmark called SummaC (Summary Consistency) which consists of six large inconsistency detection datasets. On this dataset, SummaCConv obtains state-of-the-art results with a balanced accuracy of 74.4%, a 5% improvement compared with prior work."
                },
                {
                    "title": "CommonsenseQA 2.0: Exposing the Limits of AI through Gamification",
                    "abstract": "Constructing benchmarks that test the abilities of modern natural language understanding models is difficult - pre-trained language models exploit artifacts in benchmarks to achieve human parity, but still fail on adversarial examples and make errors that demonstrate a lack of common sense. In this work, we propose gamification as a framework for data construction. The goal of players in the game is to compose questions that mislead a rival AI while using specific phrases for extra points. The game environment leads to enhanced user engagement and simultaneously gives the game designer control over the collected data, allowing us to collect high-quality data at scale. Using our method we create CommonsenseQA 2.0, which includes 14,343 yes/no questions, and demonstrate its difficulty for models that are orders-of-magnitude larger than the AI used in the game itself. Our best baseline, the T5-based Unicorn with 11B parameters achieves an accuracy of 70.2%, substantially higher than GPT-3 (52.9%) in a few-shot inference setup. Both score well below human performance which is at 94.1%."
                },
                {
                    "title": "CLIPScore: A Reference-free Evaluation Metric for Image Captioning",
                    "abstract": "Image captioning has conventionally relied on reference-based automatic evaluations, where machine captions are compared against captions written by humans. This is in contrast to the reference-free manner in which humans assess caption quality. In this paper, we report the surprising empirical finding that CLIP (Radford et al., 2021), a cross-modal model pretrained on 400M image+caption pairs from the web, can be used for robust automatic evaluation of image captioning without the need for references. Experiments spanning several corpora demonstrate that our new reference-free metric, CLIPScore, achieves the highest correlation with human judgements, outperforming existing reference-based metrics like CIDEr and SPICE. Information gain experiments demonstrate that CLIPScore, with its tight focus on image-text compatibility, is complementary to existing reference-based metrics that emphasize text-text similarities. Thus, we also present a reference-augmented version, RefCLIPScore, which achieves even higher correlation. Beyond literal description tasks, several case studies reveal domains where CLIPScore performs well (clip-art images, alt-text rating), but also where it is relatively weaker in comparison to reference-based metrics, e.g., news captions that require richer contextual knowledge."
                },
                {
                    "title": "Q^{2}: Evaluating Factual Consistency in Knowledge-Grounded Dialogues via Question Generation and Question Answering",
                    "abstract": "Neural knowledge-grounded generative models for dialogue often produce content that is factually inconsistent with the knowledge they rely on, making them unreliable and limiting their applicability. Inspired by recent work on evaluating factual consistency in abstractive summarization, we propose an automatic evaluation metric for factual consistency in knowledge-grounded dialogue using automatic question generation and question answering. Our metric, denoted Q^2, compares answer spans using natural language inference (NLI), instead of token-based matching as done in previous work. To foster proper evaluation, we curate a novel dataset of dialogue system outputs for the Wizard-of-Wikipedia dataset, manually annotated for factual consistency. We perform a thorough meta-evaluation of Q^2 against other metrics using this dataset and two others, where it consistently shows higher correlation with human judgements."
                },
                {
                    "title": "Zero-Shot Text-to-Image Generation",
                    "abstract": "Text-to-image generation has traditionally focused on finding better modeling assumptions for training on a fixed dataset. These assumptions might involve complex architectures, auxiliary losses, or side information such as object part labels or segmentation masks supplied during training. We describe a simple approach for this task based on a transformer that autoregressively models the text and image tokens as a single stream of data. With sufficient data and scale, our approach is competitive with previous domain-specific models when evaluated in a zero-shot fashion."
                },
                {
                    "title": "COMET: A Neural Framework for MT Evaluation",
                    "abstract": "We present COMET, a neural framework for training multilingual machine translation evaluation models which obtains new state-of-the-art levels of correlation with human judgements. Our framework leverages recent breakthroughs in cross-lingual pretrained language modeling resulting in highly multilingual and adaptable MT evaluation models that exploit information from both the source input and a target-language reference translation in order to more accurately predict MT quality. To showcase our framework, we train three models with different types of human judgements: Direct Assessments, Human-mediated Translation Edit Rate and Multidimensional Quality Metrics. Our models achieve new state-of-the-art performance on the WMT 2019 Metrics shared task and demonstrate robustness to high-performing systems."
                },
                {
                    "title": "BERTScore: Evaluating Text Generation with BERT",
                    "abstract": "We propose BERTScore, an automatic evaluation metric for text generation. Analogously to common metrics, BERTScore computes a similarity score for each token in the candidate sentence with each token in the reference sentence. However, instead of exact matches, we compute token similarity using contextual embeddings. We evaluate using the outputs of 363 machine translation and image captioning systems. BERTScore correlates better with human judgments and provides stronger model selection performance than existing metrics. Finally, we use an adversarial paraphrase detection task to show that BERTScore is more robust to challenging examples when compared to existing metrics."
                },
                {
                    "title": "From Recognition to Cognition: Visual Commonsense Reasoning",
                    "abstract": "Visual understanding goes well beyond object recognition. With one glance at an image, we can effortlessly imagine the world beyond the pixels: for instance, we can infer people's actions, goals, and mental states. While this task is easy for humans, it is tremendously difficult for today's vision systems, requiring higher-order cognition and commonsense reasoning about the world. We formalize this task as Visual Commonsense Reasoning. Given a challenging question about an image, a machine must answer correctly and then provide a rationale justifying its answer. Next, we introduce a new dataset, VCR, consisting of 290k multiple choice QA problems derived from 110k movie scenes. The key recipe for generating non-trivial and high-quality problems at scale is Adversarial Matching, a new approach to transform rich annotations into multiple choice questions with minimal bias. Experimental results show that while humans find VCR easy (over 90% accuracy), state-of-the-art vision models struggle (~45%). To move towards cognition-level understanding, we present a new reasoning engine, Recognition to Cognition Networks (R2C), that models the necessary layered inferences for grounding, contextualization, and reasoning. R2C helps narrow the gap between humans and machines (~65%); still, the challenge is far from solved, and we provide analysis that suggests avenues for future work."
                },
                {
                    "title": "Learning Common Sense through Visual Abstraction",
                    "abstract": "Common sense is essential for building intelligent machines. While some commonsense knowledge is explicitly stated in human-generated text and can be learnt by mining the web, much of it is unwritten. It is often unnecessary and even unnatural to write about commonsense facts. While unwritten, this commonsense knowledge is not unseen! The visual world around us is full of structure modeled by commonsense knowledge. Can machines learn common sense simply by observing our visual world? Unfortunately, this requires automatic and accurate detection of objects, their attributes, poses, and interactions between objects, which remain challenging problems. Our key insight is that while visual common sense is depicted in visual content, it is the semantic features that are relevant and not low-level pixel information. In other words, photorealism is not necessary to learn common sense. We explore the use of human-generated abstract scenes made from clipart for learning common sense. In particular, we reason about the plausibility of an interaction or relation between a pair of nouns by measuring the similarity of the relation and nouns with other relations and nouns we have seen in abstract scenes. We show that the commonsense knowledge we learn is complementary to what can be learnt from sources of text."
                },
                {
                    "title": "ROUGE: A Package for Automatic Evaluation of Summaries",
                    "abstract": "ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST."
                },
                {
                    "title": "Bleu: a Method for Automatic Evaluation of Machine Translation",
                    "abstract": "Human evaluations of machine translation are extensive but expensive. Human evaluations can take months to finish and involve human labor that can not be reused. We propose a method of automatic machine translation evaluation that is quick, inexpensive, and language-independent, that correlates highly with human evaluation, and that has little marginal cost per run. We present this method as an automated understudy to skilled human judges which substitutes for them when there is need for quick or frequent evaluations."
                },
                {
                    "title": "TRUE: Re-evaluating Factual Consistency Evaluation",
                    "abstract": "Grounded text generation systems often generate text that contains factual inconsistencies, hindering their real-world applicability. Automatic factual consistency evaluation may help alleviate this limitation by accelerating evaluation cycles, filtering inconsistent outputs and augmenting training data. While attracting increasing attention, such evaluation metrics are usually developed and evaluated in silo for a single task or dataset, slowing their adoption. Moreover, previous meta-evaluation protocols focused on system-level correlations with human annotations, which leave the example-level accuracy of such metrics unclear.In this work, we introduce TRUE: a comprehensive survey and assessment of factual consistency metrics on a standardized collection of existing texts from diverse tasks, manually annotated for factual consistency. Our standardization enables an example-level meta-evaluation protocol that is more actionable and interpretable than previously reported correlations, yielding clearer quality measures. Across diverse state-of-the-art metrics and 11 datasets we find that large-scale NLI and question generation-and-answering-based approaches achieve strong and complementary results. We recommend those methods as a starting point for model and metric developers, and hope TRUE will foster progress towards even better evaluation methods."
                },
                {
                    "title": "Content Analysis: An Introduction to Its Methodology",
                    "abstract": "History Conceptual Foundations Uses and Kinds of Inference The Logic of Content Analysis Designs Unitizing Sampling Recording Data Languages Constructs for Inference Analytical Techniques The Use of Computers Reliability Validity A Practical Guide"
                }
            ],
            "categories": [
                "cs.CV",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we enhance the commonsense reasoning capabilities of vision-language models to interpret complex visual scenes that involve subtle contextual cues?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning, particularly in vision and language integration. By improving commonsense reasoning in models, we can create systems that better understand and interact with the world, leading to more intelligent applications in areas such as autonomous systems, human-computer interaction, and content generation. This research could pave the way for future benchmarks and methodologies that challenge existing models, ultimately driving innovation and deeper understanding in AI.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of integrating visual cues with commonsense knowledge, which requires models to not only recognize objects but also understand their contextual significance. Naive approaches may fail because they often rely on surface-level features without grasping the underlying relationships and implications of those features. Additionally, the need for models to interpret synthetic images and riddles introduces technical obstacles in generating and evaluating responses accurately, as well as theoretical challenges in defining what constitutes effective commonsense reasoning.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on existing images and questions, limiting the scope of scenarios and creativity in testing commonsense reasoning. Existing benchmarks often do not account for the dynamic nature of real-world situations and the subtleties involved in human interpretation. Barriers such as the reliance on pre-trained models and the lack of innovative methodologies for generating diverse and challenging scenarios have hindered progress. Our approach differs by introducing a novel benchmark, Visual Riddles, which emphasizes scenario creation and the integration of hints and attributions, thus expanding the challenge beyond traditional methods.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves creating a benchmark of 400 visual riddles, each consisting of a synthetic image and an open-ended question designed to test commonsense reasoning. We will utilize a diverse dataset of scenarios and evaluate model performance using metrics such as accuracy in both open-ended and multiple-choice formats. Expected outcomes include a clearer understanding of the gaps in current vision-language models, insights into the effectiveness of hints and attributions, and the development of automated evaluation methods to assess model responses, ultimately revealing the extent of the performance gap between humans and state-of-the-art models."
            }
        },
        "author_data": {
            "2ee811f3-bcd4-4bd1-9e66-1c3c14e43e15": {
                "pk": "2ee811f3-bcd4-4bd1-9e66-1c3c14e43e15",
                "name": "Nitzan Bitton-Guetta",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "26adf4bf-1aa0-4026-957a-daa82ec7433d": {
                "pk": "26adf4bf-1aa0-4026-957a-daa82ec7433d",
                "name": "Aviv Slobodkin",
                "collaborators": [
                    "Ido Dagan",
                    "Eran Hirsch",
                    "Ori Shapira",
                    "Leshem Choshen",
                    "Omri Abend",
                    "Avi Caciularu",
                    "Ran Levy",
                    "Paul Roit",
                    "Ori Ernst",
                    "Omer Goldman"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Text Summarization",
                    "Semantic Understanding",
                    "Machine Translation"
                ],
                "publications": [
                    {
                        "title": "Mediators in Determining what Processing BERT Performs First",
                        "abstract": "Probing neural models for the ability to perform downstream tasks using their activation patterns is often used to localize what parts of the network specialize in performing what tasks. However, little work addressed potential mediating factors in such comparisons. As a test-case mediating factor, we consider the prediction's context length, namely the length of the span whose processing is minimally required to perform the prediction. We show that not controlling for context length may lead to contradictory conclusions as to the localization patterns of the network, depending on the distribution of the probing dataset. Indeed, when probing BERT with seven tasks, we find that it is possible to get 196 different rankings between them when manipulating the distribution of context lengths in the probing dataset. We conclude by presenting best practices for conducting such comparisons in the future."
                    },
                    {
                        "title": "Semantics-aware Attention Improves Neural Machine Translation",
                        "abstract": "The integration of syntactic structures into Transformer machine translation has shown positive results, but to our knowledge, no work has attempted to do so with semantic structures. In this work we propose two novel parameter-free methods for injecting semantic information into Transformers, both rely on semantics-aware masking of (some of) the attention heads. One such method operates on the encoder, through a Scene-Aware Self-Attention (SASA) head. Another on the decoder, through a Scene-Aware Cross-Attention (SACrA) head. We show a consistent improvement over the vanilla Transformer and syntax-aware models for four language pairs. We further show an additional gain when using both semantic and syntactic structures in some language pairs."
                    },
                    {
                        "title": "Dont Add, dont Miss: Effective Content Preserving Generation from Pre-Selected Text Spans",
                        "abstract": "The recently introduced Controlled Text Reduction (CTR) task isolates the text generation step within typical summarization-style tasks. It does so by challenging models to generate coherent text conforming to pre-selected content within the input text (``highlights''). This framing enables increased modularity in summarization-like tasks, allowing to couple a single CTR model with various content-selection setups and modules. However, there are currently no reliable CTR models, while the performance of the existing baseline for the task is mediocre, falling short of practical utility. Here, we address this gap by introducing a high-quality, open-source CTR model that tackles two prior key limitations: inadequate enforcement of the content-preservation constraint, and suboptimal silver training data. Addressing these, we amplify the content-preservation constraint in both training, via RL, and inference, via a controlled decoding strategy. Further, we substantially improve the silver training data quality via GPT-4 distillation. Overall, pairing the distilled dataset with the highlight-adherence strategies yields marked gains over the current baseline, of up to 30 ROUGE-L points, providing a reliable CTR model for downstream use."
                    },
                    {
                        "title": "Multi-Review Fusion-in-Context",
                        "abstract": "Grounded text generation, encompassing tasks such as long-form question-answering and summarization, necessitates both content selection and content consolidation. Current end-to-end methods are difficult to control and interpret due to their opaqueness. Accordingly, recent works have proposed a modular approach, with separate components for each step. Specifically, we focus on the second subtask, of generating coherent text given pre-selected content in a multi-document setting. Concretely, we formalize Fusion-in-Context (FiC) as a standalone task, whose input consists of source texts with highlighted spans of targeted content. A model then needs to generate a coherent passage that includes all and only the target information. Our work includes the development of a curated dataset of 1000 instances in the reviews domain, alongside a novel evaluation framework for assessing the faithfulness and coverage of highlights, which strongly correlate to human judgment. Several baseline models exhibit promising outcomes and provide insightful analyses. This study lays the groundwork for further exploration of modular text generation in the multi-document setting, offering potential improvements in the quality and reliability of generated content. Our benchmark, FuseReviews, including the dataset, evaluation framework, and designated leaderboard, can be found at https://fusereviews.github.io/."
                    },
                    {
                        "title": "Controlled Text Reduction",
                        "abstract": "Producing a reduced version of a source text, as in generic or focused summarization, inherently involves two distinct subtasks: deciding on targeted content and generating a coherent text conveying it. While some popular approaches address summarization as a single end-to-end task, prominent works support decomposed modeling for individual subtasks. Further, semi-automated text reduction is also very appealing, where users may identify targeted content while models would generate a corresponding coherent summary.   In this paper, we focus on the second subtask, of generating coherent text given pre-selected content. Concretely, we formalize \\textit{Controlled Text Reduction} as a standalone task, whose input is a source text with marked spans of targeted content (\"highlighting\"). A model then needs to generate a coherent text that includes all and only the target information. We advocate the potential of such models, both for modular fully-automatic summarization, as well as for semi-automated human-in-the-loop use cases. Facilitating proper research, we crowdsource high-quality dev and test datasets for the task. Further, we automatically generate a larger \"silver\" training dataset from available summarization benchmarks, leveraging a pretrained summary-source alignment model. Finally, employing these datasets, we present a supervised baseline model, showing promising results and insightful analyses."
                    },
                    {
                        "title": "SummHelper: Collaborative Human-Computer Summarization",
                        "abstract": "Current approaches for text summarization are predominantly automatic, with rather limited space for human intervention and control over the process. In this paper, we introduce SummHelper, a 2-phase summarization assistant designed to foster human-machine collaboration. The initial phase involves content selection, where the system recommends potential content, allowing users to accept, modify, or introduce additional selections. The subsequent phase, content consolidation, involves SummHelper generating a coherent summary from these selections, which users can then refine using visual mappings between the summary and the source text. Small-scale user studies reveal the effectiveness of our application, with participants being especially appreciative of the balance between automated guidance and opportunities for personal input."
                    },
                    {
                        "title": "The Curious Case of Hallucinatory (Un)answerability: Finding Truths in the Hidden States of Over-Confident Large Language Models",
                        "abstract": "Large language models (LLMs) have been shown to possess impressive capabilities, while also raising crucial concerns about the faithfulness of their responses. A primary issue arising in this context is the management of (un)answerable queries by LLMs, which often results in hallucinatory behavior due to overconfidence. In this paper, we explore the behavior of LLMs when presented with (un)answerable queries. We ask: do models represent the fact that the question is (un)answerable when generating a hallucinatory answer? Our results show strong indications that such models encode the answerability of an input query, with the representation of the first decoded token often being a strong indicator. These findings shed new light on the spatial organization within the latent representations of LLMs, unveiling previously unexplored facets of these models. Moreover, they pave the way for the development of improved decoding techniques with better adherence to factual generation, particularly in scenarios where query (un)answerability is a concern."
                    },
                    {
                        "title": "Attribute First, then Generate: Locally-attributable Grounded Text Generation",
                        "abstract": "Recent efforts to address hallucinations in Large Language Models (LLMs) have focused on attributed text generation, which supplements generated texts with citations of supporting sources for post-generation fact-checking and corrections. Yet, these citations often point to entire documents or paragraphs, burdening users with extensive verification work. In this paper, we introduce a locally-attributable text generation approach, prioritizing concise attributions. Our method, named \"Attribute First, then Generate\", breaks down the conventional end-to-end generation process into three intuitive steps: content selection, sentence planning, and sequential sentence generation. By initially identifying relevant source segments (\"select first\") and then conditioning the generation process on them (\"then generate\"), we ensure these segments also act as the output's fine-grained attributions (\"select\" becomes \"attribute\"). Tested on Multi-document Summarization and Long-form Question-answering, our method not only yields more concise citations than the baselines but also maintains - and in some cases enhances - both generation quality and attribution accuracy. Furthermore, it significantly reduces the time required for fact verification by human assessors."
                    },
                    {
                        "title": "Explicating the Implicit: Argument Detection Beyond Sentence Boundaries",
                        "abstract": "Detecting semantic arguments of a predicate word has been conventionally modeled as a sentence-level task. The typical reader, however, perfectly interprets predicate-argument relations in a much wider context than just the sentence where the predicate was evoked. In this work, we reformulate the problem of argument detection through textual entailment to capture semantic relations across sentence boundaries. We propose a method that tests whether some semantic relation can be inferred from a full passage by first encoding it into a simple and standalone proposition and then testing for entailment against the passage. Our method does not require direct supervision, which is generally absent due to dataset scarcity, but instead builds on existing NLI and sentence-level SRL resources. Such a method can potentially explicate pragmatically understood relations into a set of explicit sentences. We demonstrate it on a recent document-level benchmark, outperforming some supervised methods and contemporary language models."
                    },
                    {
                        "title": "Is It Really Long Context if All You Need Is Retrieval? Towards Genuinely Difficult Long Context NLP",
                        "abstract": "Improvements in language models' capabilities have pushed their applications towards longer contexts, making long-context evaluation and development an active research area. However, many disparate use-cases are grouped together under the umbrella term of \"long-context\", defined simply by the total length of the model's input, including - for example - Needle-in-a-Haystack tasks, book summarization, and information aggregation. Given their varied difficulty, in this position paper we argue that conflating different tasks by their context length is unproductive. As a community, we require a more precise vocabulary to understand what makes long-context tasks similar or different. We propose to unpack the taxonomy of long-context based on the properties that make them more difficult with longer contexts. We propose two orthogonal axes of difficulty: (I) Diffusion: How hard is it to find the necessary information in the context? (II) Scope: How much necessary information is there to find? We survey the literature on long-context, provide justification for this taxonomy as an informative descriptor, and situate the literature with respect to it. We conclude that the most difficult and interesting settings, whose necessary information is very long and highly diffused within the input, is severely under-explored. By using a descriptive vocabulary and discussing the relevant properties of difficulty in long-context, we can implement more informed research in this area. We call for a careful design of tasks and benchmarks with distinctly long context, taking into account the characteristics that make it qualitatively different from shorter context."
                    },
                    {
                        "title": "The Power of Summary-Source Alignments",
                        "abstract": "Multi-document summarization (MDS) is a challenging task, often decomposed to subtasks of salience and redundancy detection, followed by text generation. In this context, alignment of corresponding sentences between a reference summary and its source documents has been leveraged to generate training data for some of the component tasks. Yet, this enabling alignment step has usually been applied heuristically on the sentence level on a limited number of subtasks. In this paper, we propose extending the summary-source alignment framework by (1) applying it at the more fine-grained proposition span level, (2) annotating alignment manually in a multi-document setup, and (3) revealing the great potential of summary-source alignments to yield several datasets for at least six different tasks. Specifically, for each of the tasks, we release a manually annotated test set that was derived automatically from the alignment annotation. We also release development and train sets in the same way, but from automatically derived alignments. Using the datasets, each task is demonstrated with baseline models and corresponding evaluation metrics to spur future research on this broad challenge."
                    }
                ]
            },
            "ce84cc2e-998e-4b92-919f-a63bd51d9d67": {
                "pk": "ce84cc2e-998e-4b92-919f-a63bd51d9d67",
                "name": "Aviya Maimon",
                "collaborators": [
                    "Reut Tsarfaty",
                    "Omer Goldman",
                    "Alon Jacovi",
                    "Aviv Slobodkin",
                    "Ido Dagan"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Coherence Assessment",
                    "Generative AI",
                    "Long-context Evaluation"
                ],
                "publications": [
                    {
                        "title": "A Novel Computational and Modeling Foundation for Automatic Coherence Assessment",
                        "abstract": "Coherence is an essential property of well-written texts, that refers to the way textual units relate to one another. In the era of generative AI, coherence assessment is essential for many NLP tasks; summarization, generation, long-form question-answering, and more. However, in NLP {coherence} is an ill-defined notion, not having a formal definition or evaluation metrics, that would allow for large-scale automatic and systematic coherence assessment. To bridge this gap, in this work we employ the formal linguistic definition of \\citet{Reinhart:1980} of what makes a discourse coherent, consisting of three conditions -- {\\em cohesion, consistency} and {\\em relevance} -- and formalize these conditions as respective computational tasks. We hypothesize that (i) a model trained on all of these tasks will learn the features required for coherence detection, and that (ii) a joint model for all tasks will exceed the performance of models trained on each task individually. On two benchmarks for coherence scoring rated by humans, one containing 500 automatically-generated short stories and another containing 4k real-world texts, our experiments confirm that jointly training on the proposed tasks leads to better performance on each task compared with task-specific models, and to better performance on assessing coherence overall, compared with strong baselines. We conclude that the formal and computational setup of coherence as proposed here provides a solid foundation for advanced methods of large-scale automatic assessment of coherence."
                    },
                    {
                        "title": "CoheSentia: A Novel Benchmark of Incremental versus Holistic Assessment of Coherence in Generated Texts",
                        "abstract": "Coherence is a linguistic term that refers to the relations between small textual units (sentences, propositions), which make the text logically consistent and meaningful to the reader. With the advances of generative foundational models in NLP, there is a pressing need to automatically assess the human-perceived coherence of automatically generated texts. Up until now, little work has been done on explicitly assessing the coherence of generated texts and analyzing the factors contributing to (in)coherence. Previous work on the topic used other tasks, e.g., sentence reordering, as proxies of coherence, rather than approaching coherence detection heads on. In this paper, we introduce {\\sc CoheSentia}, a novel benchmark of human-perceived coherence of automatically generated texts. Our annotation protocol reflects two perspectives; one is global, assigning a single coherence score, and the other is incremental, scoring sentence by sentence. The incremental method produces an (in)coherence score for each text fragment and also pinpoints reasons for incoherence at that point. Our benchmark contains 500 automatically-generated and human-annotated paragraphs, each annotated in both methods, by multiple raters. Our analysis shows that the inter-annotator agreement in the incremental mode is higher than in the holistic alternative, and our experiments show that standard LMs fine-tuned for coherence detection show varied performance on the different factors contributing to (in)coherence. All in all, these models yield unsatisfactory performance, emphasizing the need for developing more reliable methods for coherence assessment."
                    },
                    {
                        "title": "Is It Really Long Context if All You Need Is Retrieval? Towards Genuinely Difficult Long Context NLP",
                        "abstract": "Improvements in language models' capabilities have pushed their applications towards longer contexts, making long-context evaluation and development an active research area. However, many disparate use-cases are grouped together under the umbrella term of \"long-context\", defined simply by the total length of the model's input, including - for example - Needle-in-a-Haystack tasks, book summarization, and information aggregation. Given their varied difficulty, in this position paper we argue that conflating different tasks by their context length is unproductive. As a community, we require a more precise vocabulary to understand what makes long-context tasks similar or different. We propose to unpack the taxonomy of long-context based on the properties that make them more difficult with longer contexts. We propose two orthogonal axes of difficulty: (I) Diffusion: How hard is it to find the necessary information in the context? (II) Scope: How much necessary information is there to find? We survey the literature on long-context, provide justification for this taxonomy as an informative descriptor, and situate the literature with respect to it. We conclude that the most difficult and interesting settings, whose necessary information is very long and highly diffused within the input, is severely under-explored. By using a descriptive vocabulary and discussing the relevant properties of difficulty in long-context, we can implement more informed research in this area. We call for a careful design of tasks and benchmarks with distinctly long context, taking into account the characteristics that make it qualitatively different from shorter context."
                    }
                ]
            },
            "09628d55-a0ef-4ca9-9763-6d94a672fde5": {
                "pk": "09628d55-a0ef-4ca9-9763-6d94a672fde5",
                "name": "Eliya Habba",
                "collaborators": [
                    "Renana Keydar",
                    "Dan Bareket",
                    "Gabriel Stanovsky"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Legal Informatics",
                    "Sentiment Analysis"
                ],
                "publications": [
                    {
                        "title": "The Perfect Victim: Computational Analysis of Judicial Attitudes towards Victims of Sexual Violence",
                        "abstract": "We develop computational models to analyze court statements in order to assess judicial attitudes toward victims of sexual violence in the Israeli court system. The study examines the resonance of \"rape myths\" in the criminal justice system's response to sex crimes, in particular in judicial assessment of victim's credibility. We begin by formulating an ontology for evaluating judicial attitudes toward victim's credibility, with eight ordinal labels and binary categorizations. Second, we curate a manually annotated dataset for judicial assessments of victim's credibility in the Hebrew language, as well as a model that can extract credibility labels from court cases. The dataset consists of 855 verdict decision documents in sexual assault cases from 1990-2021, annotated with the help of legal experts and trained law students. The model uses a combined approach of syntactic and latent structures to find sentences that convey the judge's attitude towards the victim and classify them according to the credibility label set. Our ontology, data, and models will be made available upon request, in the hope they spur future progress in this judicial important task."
                    }
                ]
            },
            "c8d235b0-9618-491b-89a0-17167d14145c": {
                "pk": "c8d235b0-9618-491b-89a0-17167d14145c",
                "name": "Royi Rassin",
                "collaborators": [
                    "Yoav Goldberg",
                    "Shauli Ravfogel",
                    "Eran Hirsch",
                    "Gal Chechik",
                    "Reut Tsarfaty",
                    "Daniel Glickman",
                    "Yaron Fairstein",
                    "Oren Kalinsky",
                    "Guy Kushilevitz",
                    "Nachshon Cohen"
                ],
                "domain": [
                    "Text-to-Image Generation",
                    "Natural Language Processing",
                    "Machine Learning",
                    "Copyright Compliance"
                ],
                "publications": [
                    {
                        "title": "DALLE-2 is Seeing Double: Flaws in Word-to-Concept Mapping in Text2Image Models",
                        "abstract": "We study the way DALLE-2 maps symbols (words) in the prompt to their references (entities or properties of entities in the generated image). We show that in stark contrast to the way human process language, DALLE-2 does not follow the constraint that each word has a single role in the interpretation, and sometimes re-use the same symbol for different purposes. We collect a set of stimuli that reflect the phenomenon: we show that DALLE-2 depicts both senses of nouns with multiple senses at once; and that a given word can modify the properties of two distinct entities in the image, or can be depicted as one object and also modify the properties of another object, creating a semantic leakage of properties between entities. Taken together, our study highlights the differences between DALLE-2 and human language processing and opens an avenue for future study on the inductive biases of text-to-image models."
                    },
                    {
                        "title": "Conjunct Resolution in the Face of Verbal Omissions",
                        "abstract": "Verbal omissions are complex syntactic phenomena in VP coordination structures. They occur when verbs and (some of) their arguments are omitted from subsequent clauses after being explicitly stated in an initial clause. Recovering these omitted elements is necessary for accurate interpretation of the sentence, and while humans easily and intuitively fill in the missing information, state-of-the-art models continue to struggle with this task. Previous work is limited to small-scale datasets, synthetic data creation methods, and to resolution methods in the dependency-graph level. In this work we propose a conjunct resolution task that operates directly on the text and makes use of a split-and-rephrase paradigm in order to recover the missing elements in the coordination structure. To this end, we first formulate a pragmatic framework of verbal omissions which describes the different types of omissions, and develop an automatic scalable collection method. Based on this method, we curate a large dataset, containing over 10K examples of naturally-occurring verbal omissions with crowd-sourced annotations of the resolved conjuncts. We train various neural baselines for this task, and show that while our best method obtains decent performance, it leaves ample space for improvement. We propose our dataset, metrics and models as a starting point for future research on this topic."
                    },
                    {
                        "title": "Linguistic Binding in Diffusion Models: Enhancing Attribute Correspondence through Attention Map Alignment",
                        "abstract": "Text-conditioned image generation models often generate incorrect associations between entities and their visual attributes. This reflects an impaired mapping between linguistic binding of entities and modifiers in the prompt and visual binding of the corresponding elements in the generated image. As one notable example, a query like \"a pink sunflower and a yellow flamingo\" may incorrectly produce an image of a yellow sunflower and a pink flamingo. To remedy this issue, we propose SynGen, an approach which first syntactically analyses the prompt to identify entities and their modifiers, and then uses a novel loss function that encourages the cross-attention maps to agree with the linguistic binding reflected by the syntax. Specifically, we encourage large overlap between attention maps of entities and their modifiers, and small overlap with other entities and modifier words. The loss is optimized during inference, without retraining or fine-tuning the model. Human evaluation on three datasets, including one new and challenging set, demonstrate significant improvements of SynGen compared with current state of the art methods. This work highlights how making use of sentence structure during inference can efficiently and substantially improve the faithfulness of text-to-image generation."
                    },
                    {
                        "title": "Evaluating D-MERIT of Partial-annotation on Information Retrieval",
                        "abstract": "Retrieval models are often evaluated on partially-annotated datasets. Each query is mapped to a few relevant texts and the remaining corpus is assumed to be irrelevant. As a result, models that successfully retrieve false negatives are punished in evaluation. Unfortunately, completely annotating all texts for every query is not resource efficient. In this work, we show that using partially-annotated datasets in evaluation can paint a distorted picture. We curate D-MERIT, a passage retrieval evaluation set from Wikipedia, aspiring to contain all relevant passages for each query. Queries describe a group (e.g., \"journals about linguistics\") and relevant passages are evidence that entities belong to the group (e.g., a passage indicating that \"Language\" is a journal about linguistics). We show that evaluating on a dataset containing annotations for only a subset of the relevant passages might result in misleading ranking of the retrieval systems and that as more relevant texts are included in the evaluation set, the rankings converge. We propose our dataset as a resource for evaluation and our study as a recommendation for balance between resource-efficiency and reliable evaluation when annotating evaluation sets for text retrieval."
                    },
                    {
                        "title": "Make It Count: Text-to-Image Generation with an Accurate Number of Objects",
                        "abstract": "Despite the unprecedented success of text-to-image diffusion models, controlling the number of depicted objects using text is surprisingly hard. This is important for various applications from technical documents, to children's books to illustrating cooking recipes. Generating object-correct counts is fundamentally challenging because the generative model needs to keep a sense of separate identity for every instance of the object, even if several objects look identical or overlap, and then carry out a global computation implicitly during generation. It is still unknown if such representations exist. To address count-correct generation, we first identify features within the diffusion model that can carry the object identity information. We then use them to separate and count instances of objects during the denoising process and detect over-generation and under-generation. We fix the latter by training a model that predicts both the shape and location of a missing object, based on the layout of existing ones, and show how it can be used to guide denoising with correct object count. Our approach, CountGen, does not depend on external source to determine object layout, but rather uses the prior from the diffusion model itself, creating prompt-dependent and seed-dependent layouts. Evaluated on two benchmark datasets, we find that CountGen strongly outperforms the count-accuracy of existing baselines."
                    },
                    {
                        "title": "DreamSync: Aligning Text-to-Image Generation with Image Understanding Feedback",
                        "abstract": "Despite their wide-spread success, Text-to-Image models (T2I) still struggle to produce images that are both aesthetically pleasing and faithful to the user's input text. We introduce DreamSync, a model-agnostic training algorithm by design that improves T2I models to be faithful to the text input. DreamSync builds off a recent insight from TIFA's evaluation framework -- that large vision-language models (VLMs) can effectively identify the fine-grained discrepancies between generated images and the text inputs. DreamSync uses this insight to train T2I models without any labeled data; it improves T2I models using its own generations. First, it prompts the model to generate several candidate images for a given input text. Then, it uses two VLMs to select the best generation: a Visual Question Answering model that measures the alignment of generated images to the text, and another that measures the generation's aesthetic quality. After selection, we use LoRA to iteratively finetune the T2I model to guide its generation towards the selected best generations. DreamSync does not need any additional human annotation. model architecture changes, or reinforcement learning. Despite its simplicity, DreamSync improves both the semantic alignment and aesthetic appeal of two diffusion-based T2I models, evidenced by multiple benchmarks (+1.7% on TIFA, +2.9% on DSG1K, +3.4% on VILA aesthetic) and human evaluation."
                    },
                    {
                        "title": "How Many Van Goghs Does It Take to Van Gogh? Finding the Imitation Threshold",
                        "abstract": "Text-to-image models are trained using large datasets collected by scraping image-text pairs from the internet. These datasets often include private, copyrighted, and licensed material. Training models on such datasets enables them to generate images with such content, which might violate copyright laws and individual privacy. This phenomenon is termed imitation -- generation of images with content that has recognizable similarity to its training images. In this work we study the relationship between a concept's frequency in the training dataset and the ability of a model to imitate it. We seek to determine the point at which a model was trained on enough instances to imitate a concept -- the imitation threshold. We posit this question as a new problem: Finding the Imitation Threshold (FIT) and propose an efficient approach that estimates the imitation threshold without incurring the colossal cost of training multiple models from scratch. We experiment with two domains -- human faces and art styles -- for which we create four datasets, and evaluate three text-to-image models which were trained on two pretraining datasets. Our results reveal that the imitation threshold of these models is in the range of 200-600 images, depending on the domain and the model. The imitation threshold can provide an empirical basis for copyright violation claims and acts as a guiding principle for text-to-image model developers that aim to comply with copyright and privacy laws. We release the code and data at \\url{https://github.com/vsahil/MIMETIC-2.git} and the project's website is hosted at \\url{https://how-many-van-goghs-does-it-take.github.io}."
                    }
                ]
            },
            "ecf4fb87-052c-40fb-86c9-80da7291030b": {
                "pk": "ecf4fb87-052c-40fb-86c9-80da7291030b",
                "name": "Yonatan Bitton",
                "collaborators": [
                    "Idan Szpektor",
                    "Roy Schwartz",
                    "Gabriel Stanovsky",
                    "Ron Yosef",
                    "Michal Yarom",
                    "Dafna Shahaf",
                    "Roee Aharoni",
                    "Hritik Bansal",
                    "Michael Elhadad",
                    "Joao Magalhaes"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Visual Question Answering",
                    "Natural Language Processing",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Automatic Generation of Contrast Sets from Scene Graphs: Probing the Compositional Consistency of GQA",
                        "abstract": "Recent works have shown that supervised models often exploit data artifacts to achieve good test scores while their performance severely degrades on samples outside their training distribution. Contrast sets (Gardneret al., 2020) quantify this phenomenon by perturbing test samples in a minimal way such that the output label is modified. While most contrast sets were created manually, requiring intensive annotation effort, we present a novel method which leverages rich semantic input representation to automatically generate contrast sets for the visual question answering task. Our method computes the answer of perturbed questions, thus vastly reducing annotation cost and enabling thorough evaluation of models' performance on various semantic aspects (e.g., spatial or relational reasoning). We demonstrate the effectiveness of our approach on the GQA dataset and its semantic scene graph image representation. We find that, despite GQA's compositionality and carefully balanced label distribution, two high-performing models drop 13-17% in accuracy compared to the original test set. Finally, we show that our automatic perturbation can be applied to the training set to mitigate the degradation in performance, opening the door to more robust models."
                    },
                    {
                        "title": "Data Efficient Masked Language Modeling for Vision and Language",
                        "abstract": "Masked language modeling (MLM) is one of the key sub-tasks in vision-language pretraining. In the cross-modal setting, tokens in the sentence are masked at random, and the model predicts the masked tokens given the image and the text. In this paper, we observe several key disadvantages of MLM in this setting. First, as captions tend to be short, in a third of the sentences no token is sampled. Second, the majority of masked tokens are stop-words and punctuation, leading to under-utilization of the image. We investigate a range of alternative masking strategies specific to the cross-modal setting that address these shortcomings, aiming for better fusion of text and image in the learned representation. When pre-training the LXMERT model, our alternative masking strategies consistently improve over the original masking strategy on three downstream tasks, especially in low resource settings. Further, our pre-training approach substantially outperforms the baseline model on a prompt-based probing task designed to elicit image objects. These results and our analysis indicate that our method allows for better utilization of the training data."
                    },
                    {
                        "title": "IRFL: Image Recognition of Figurative Language",
                        "abstract": "Figures of speech such as metaphors, similes, and idioms are integral parts of human communication. They are ubiquitous in many forms of discourse, allowing people to convey complex, abstract ideas and evoke emotion. As figurative forms are often conveyed through multiple modalities (e.g., both text and images), understanding multimodal figurative language is an important AI challenge, weaving together profound vision, language, commonsense and cultural knowledge. In this work, we develop the Image Recognition of Figurative Language (IRFL) dataset. We leverage human annotation and an automatic pipeline we created to generate a multimodal dataset, and introduce two novel tasks as a benchmark for multimodal figurative language understanding. We experimented with state-of-the-art vision and language models and found that the best (22%) performed substantially worse than humans (97%). We release our dataset, benchmark, and code, in hopes of driving the development of models that can better understand figurative language."
                    },
                    {
                        "title": "Read, Look or Listen? What's Needed for Solving a Multimodal Dataset",
                        "abstract": "The prevalence of large-scale multimodal datasets presents unique challenges in assessing dataset quality. We propose a two-step method to analyze multimodal datasets, which leverages a small seed of human annotation to map each multimodal instance to the modalities required to process it. Our method sheds light on the importance of different modalities in datasets, as well as the relationship between them. We apply our approach to TVQA, a video question-answering dataset, and discover that most questions can be answered using a single modality, without a substantial bias towards any specific modality. Moreover, we find that more than 70% of the questions are solvable using several different single-modality strategies, e.g., by either looking at the video or listening to the audio, highlighting the limited integration of multiple modalities in TVQA. We leverage our annotation and analyze the MERLOT Reserve, finding that it struggles with image-based questions compared to text and audio, but also with auditory speaker identification. Based on our observations, we introduce a new test set that necessitates multiple modalities, observing a dramatic drop in model performance. Our methodology provides valuable insights into multimodal datasets and highlights the need for the development of more robust models."
                    },
                    {
                        "title": "ParallelPARC: A Scalable Pipeline for Generating Natural-Language Analogies",
                        "abstract": "Analogy-making is central to human cognition, allowing us to adapt to novel situations -- an ability that current AI systems still lack. Most analogy datasets today focus on simple analogies (e.g., word analogies); datasets including complex types of analogies are typically manually curated and very small. We believe that this holds back progress in computational analogy. In this work, we design a data generation pipeline, ParallelPARC (Parallel Paragraph Creator) leveraging state-of-the-art Large Language Models (LLMs) to create complex, paragraph-based analogies, as well as distractors, both simple and challenging. We demonstrate our pipeline and create ProPara-Logy, a dataset of analogies between scientific processes. We publish a gold-set, validated by humans, and a silver-set, generated automatically. We test LLMs' and humans' analogy recognition in binary and multiple-choice settings, and found that humans outperform the best models (~13% gap) after a light supervision. We demonstrate that our silver-set is useful for training models. Lastly, we show challenging distractors confuse LLMs, but not humans. We hope our pipeline will encourage research in this emerging field."
                    },
                    {
                        "title": "VASR: Visual Analogies of Situation Recognition",
                        "abstract": "A core process in human cognition is analogical mapping: the ability to identify a similar relational structure between different situations. We introduce a novel task, Visual Analogies of Situation Recognition, adapting the classical word-analogy task into the visual domain. Given a triplet of images, the task is to select an image candidate B' that completes the analogy (A to A' is like B to what?). Unlike previous work on visual analogy that focused on simple image transformations, we tackle complex analogies requiring understanding of scenes.   We leverage situation recognition annotations and the CLIP model to generate a large set of 500k candidate analogies. Crowdsourced annotations for a sample of the data indicate that humans agree with the dataset label ~80% of the time (chance level 25%). Furthermore, we use human annotations to create a gold-standard dataset of 3,820 validated analogies. Our experiments demonstrate that state-of-the-art models do well when distractors are chosen randomly (~86%), but struggle with carefully chosen distractors (~53%, compared to 90% human accuracy). We hope our dataset will encourage the development of new analogy-making models. Website: https://vasr-dataset.github.io/"
                    },
                    {
                        "title": "Mismatch Quest: Visual and Textual Feedback for Image-Text Misalignment",
                        "abstract": "While existing image-text alignment models reach high quality binary assessments, they fall short of pinpointing the exact source of misalignment. In this paper, we present a method to provide detailed textual and visual explanation of detected misalignments between text-image pairs. We leverage large language models and visual grounding models to automatically construct a training set that holds plausible misaligned captions for a given image and corresponding textual explanations and visual indicators. We also publish a new human curated test set comprising ground-truth textual and visual misalignment annotations. Empirical results show that fine-tuning vision language models on our training set enables them to articulate misalignments and visually indicate them within images, outperforming strong baselines both on the binary alignment classification and the explanation generation tasks. Our method code and human curated test set are available at: https://mismatch-quest.github.io/"
                    },
                    {
                        "title": "Contrastive Sequential-Diffusion Learning: An approach to Multi-Scene Instructional Video Synthesis",
                        "abstract": "Action-centric sequence descriptions like recipe instructions and do-it-yourself projects include non-linear patterns in which the next step may require to be visually consistent not on the immediate previous step but on earlier steps. Current video synthesis approaches fail to generate consistent multi-scene videos for such task descriptions. We propose a contrastive sequential video diffusion method that selects the most suitable previously generated scene to guide and condition the denoising process of the next scene. The result is a multi-scene video that is grounded in the scene descriptions and coherent w.r.t the scenes that require consistent visualisation. Our experiments with real-world data demonstrate the practicality and improved consistency of our model compared to prior work."
                    },
                    {
                        "title": "Breaking Common Sense: WHOOPS! A Vision-and-Language Benchmark of Synthetic and Compositional Images",
                        "abstract": "Weird, unusual, and uncanny images pique the curiosity of observers because they challenge commonsense. For example, an image released during the 2022 world cup depicts the famous soccer stars Lionel Messi and Cristiano Ronaldo playing chess, which playfully violates our expectation that their competition should occur on the football field. Humans can easily recognize and interpret these unconventional images, but can AI models do the same? We introduce WHOOPS!, a new dataset and benchmark for visual commonsense. The dataset is comprised of purposefully commonsense-defying images created by designers using publicly-available image generation tools like Midjourney. We consider several tasks posed over the dataset. In addition to image captioning, cross-modal matching, and visual question answering, we introduce a difficult explanation generation task, where models must identify and explain why a given image is unusual. Our results show that state-of-the-art models such as GPT3 and BLIP2 still lag behind human performance on WHOOPS!. We hope our dataset will inspire the development of AI models with stronger visual commonsense reasoning abilities. Data, models and code are available at the project website: whoops-benchmark.github.io"
                    },
                    {
                        "title": "q2d: Turning Questions into Dialogs to Teach Models How to Search",
                        "abstract": "One of the exciting capabilities of recent language models for dialog is their ability to independently search for relevant information to ground a given dialog response. However, obtaining training data to teach models how to issue search queries is time and resource consuming. In this work, we propose q2d: an automatic data generation pipeline that generates information-seeking dialogs from questions. We prompt a large language model (PaLM) to create conversational versions of question answering datasets, and use it to improve query generation models that communicate with external search APIs to ground dialog responses. Unlike previous approaches which relied on human written dialogs with search queries, our method allows to automatically generate query-based grounded dialogs with better control and scale. Our experiments demonstrate that: (1) For query generation on the QReCC dataset, models trained on our synthetically-generated data achieve 90%--97% of the performance of models trained on the human-generated data; (2) We can successfully generate data for training dialog models in new domains without any existing dialog data as demonstrated on the multi-hop MuSiQue and Bamboogle QA datasets. (3) We perform a thorough analysis of the generated dialogs showing that humans find them of high quality and struggle to distinguish them from human-written dialogs."
                    },
                    {
                        "title": "Video-STaR: Self-Training Enables Video Instruction Tuning with Any Supervision",
                        "abstract": "The performance of Large Vision Language Models (LVLMs) is dependent on the size and quality of their training datasets. Existing video instruction tuning datasets lack diversity as they are derived by prompting large language models with video captions to generate question-answer pairs, and are therefore mostly descriptive. Meanwhile, many labeled video datasets with diverse labels and supervision exist - however, we find that their integration into LVLMs is non-trivial. Herein, we present Video Self-Training with augmented Reasoning (Video-STaR), the first video self-training approach. Video-STaR allows the utilization of any labeled video dataset for video instruction tuning. In Video-STaR, an LVLM cycles between instruction generation and finetuning, which we show (I) improves general video understanding and (II) adapts LVLMs to novel downstream tasks with existing supervision. During generation, an LVLM is prompted to propose an answer. The answers are then filtered only to those that contain the original video labels, and the LVLM is then re-trained on the generated dataset. By only training on generated answers that contain the correct video labels, Video-STaR utilizes these existing video labels as weak supervision for video instruction tuning. Our results demonstrate that Video-STaR-enhanced LVLMs exhibit improved performance in (I) general video QA, where TempCompass performance improved by 10%, and (II) on downstream tasks, where Video-STaR improved Kinetics700-QA accuracy by 20% and action quality assessment on FineDiving by 15%."
                    },
                    {
                        "title": "WinoGAViL: Gamified Association Benchmark to Challenge Vision-and-Language Models",
                        "abstract": "While vision-and-language models perform well on tasks such as visual question answering, they struggle when it comes to basic human commonsense reasoning skills. In this work, we introduce WinoGAViL: an online game of vision-and-language associations (e.g., between werewolves and a full moon), used as a dynamic evaluation benchmark. Inspired by the popular card game Codenames, a spymaster gives a textual cue related to several visual candidates, and another player tries to identify them. Human players are rewarded for creating associations that are challenging for a rival AI model but still solvable by other human players. We use the game to collect 3.5K instances, finding that they are intuitive for humans (>90% Jaccard index) but challenging for state-of-the-art AI models, where the best model (ViLT) achieves a score of 52%, succeeding mostly where the cue is visually salient. Our analysis as well as the feedback we collect from players indicate that the collected associations require diverse reasoning skills, including general knowledge, common sense, abstraction, and more. We release the dataset, the code and the interactive game, allowing future data collection that can be used to develop models with better association abilities."
                    },
                    {
                        "title": "What You See is What You Read? Improving Text-Image Alignment Evaluation",
                        "abstract": "Automatically determining whether a text and a corresponding image are semantically aligned is a significant challenge for vision-language models, with applications in generative text-to-image and image-to-text tasks. In this work, we study methods for automatic text-image alignment evaluation. We first introduce SeeTRUE: a comprehensive evaluation set, spanning multiple datasets from both text-to-image and image-to-text generation tasks, with human judgements for whether a given text-image pair is semantically aligned. We then describe two automatic methods to determine alignment: the first involving a pipeline based on question generation and visual question answering models, and the second employing an end-to-end classification approach by finetuning multimodal pretrained models. Both methods surpass prior approaches in various text-image alignment tasks, with significant improvements in challenging cases that involve complex composition or unnatural images. Finally, we demonstrate how our approaches can localize specific misalignments between an image and a given text, and how they can be used to automatically re-rank candidates in text-to-image generation."
                    },
                    {
                        "title": "VisIT-Bench: A Benchmark for Vision-Language Instruction Following Inspired by Real-World Use",
                        "abstract": "We introduce VisIT-Bench (Visual InsTruction Benchmark), a benchmark for evaluation of instruction-following vision-language models for real-world use. Our starting point is curating 70 'instruction families' that we envision instruction tuned vision-language models should be able to address. Extending beyond evaluations like VQAv2 and COCO, tasks range from basic recognition to game playing and creative generation. Following curation, our dataset comprises 592 test queries, each with a human-authored instruction-conditioned caption. These descriptions surface instruction-specific factors, e.g., for an instruction asking about the accessibility of a storefront for wheelchair users, the instruction-conditioned caption describes ramps/potential obstacles. These descriptions enable 1) collecting human-verified reference outputs for each instance; and 2) automatic evaluation of candidate multimodal generations using a text-only LLM, aligning with human judgment. We quantify quality gaps between models and references using both human and automatic evaluations; e.g., the top-performing instruction-following model wins against the GPT-4 reference in just 27% of the comparison. VisIT-Bench is dynamic to participate, practitioners simply submit their model's response on the project website; Data, code and leaderboard is available at visit-bench.github.io."
                    },
                    {
                        "title": "Transferring Visual Attributes from Natural Language to Verified Image Generation",
                        "abstract": "Text to image generation methods (T2I) are widely popular in generating art and other creative artifacts. While visual hallucinations can be a positive factor in scenarios where creativity is appreciated, such artifacts are poorly suited for cases where the generated image needs to be grounded in complex natural language without explicit visual elements. In this paper, we propose to strengthen the consistency property of T2I methods in the presence of natural complex language, which often breaks the limits of T2I methods by including non-visual information, and textual elements that require knowledge for accurate generation. To address these phenomena, we propose a Natural Language to Verified Image generation approach (NL2VI) that converts a natural prompt into a visual prompt, which is more suitable for image generation. A T2I model then generates an image for the visual prompt, which is then verified with VQA algorithms. Experimentally, aligning natural prompts with image generation can improve the consistency of the generated images by up to 11% over the state of the art. Moreover, improvements can generalize to challenging domains like cooking and DIY tasks, where the correctness of the generated image is crucial to illustrate actions."
                    },
                    {
                        "title": "VideoCon: Robust Video-Language Alignment via Contrast Captions",
                        "abstract": "Despite being (pre)trained on a massive amount of data, state-of-the-art video-language alignment models are not robust to semantically-plausible contrastive changes in the video captions. Our work addresses this by identifying a broad spectrum of contrast misalignments, such as replacing entities, actions, and flipping event order, which alignment models should be robust against. To this end, we introduce the VideoCon, a video-language alignment dataset constructed by a large language model that generates plausible contrast video captions and explanations for differences between original and contrast video captions. Then, a generative video-language model is finetuned with VideoCon to assess video-language entailment and generate explanations. Our VideoCon-based alignment model significantly outperforms current models. It exhibits a 12-point increase in AUC for the video-language alignment task on human-generated contrast captions. Finally, our model sets new state of the art zero-shot performance in temporally-extensive video-language tasks such as text-to-video retrieval (SSv2-Temporal) and video question answering (ATP-Hard). Moreover, our model shows superior performance on novel videos and human-crafted captions and explanations. Our code and data are available at https://github.com/Hritikbansal/videocon."
                    },
                    {
                        "title": "TALC: Time-Aligned Captions for Multi-Scene Text-to-Video Generation",
                        "abstract": "Recent advances in diffusion-based generative modeling have led to the development of text-to-video (T2V) models that can generate high-quality videos conditioned on a text prompt. Most of these T2V models often produce single-scene video clips that depict an entity performing a particular action (e.g., 'a red panda climbing a tree'). However, it is pertinent to generate multi-scene videos since they are ubiquitous in the real-world (e.g., 'a red panda climbing a tree' followed by 'the red panda sleeps on the top of the tree'). To generate multi-scene videos from a pretrained T2V model, we introduce Time-Aligned Captions (TALC) framework. Specifically, we enhance the text-conditioning mechanism in the T2V architecture to recognize the temporal alignment between the video scenes and scene descriptions. As a result, we show that the pretrained T2V model can generate multi-scene videos that adhere to the multi-scene text descriptions and be visually consistent (e.g., w.r.t entity and background). Our TALC-finetuned model outperforms the baseline methods on multi-scene video-text data by 15.5 points on aggregated score, averaging visual consistency and text adherence using human evaluation. The project website is https://talc-mst2v.github.io/."
                    },
                    {
                        "title": "NL-Eye: Abductive NLI for Images",
                        "abstract": "Will a Visual Language Model (VLM)-based bot warn us about slipping if it detects a wet floor? Recent VLMs have demonstrated impressive capabilities, yet their ability to infer outcomes and causes remains underexplored. To address this, we introduce NL-Eye, a benchmark designed to assess VLMs' visual abductive reasoning skills. NL-Eye adapts the abductive Natural Language Inference (NLI) task to the visual domain, requiring models to evaluate the plausibility of hypothesis images based on a premise image and explain their decisions. NL-Eye consists of 350 carefully curated triplet examples (1,050 images) spanning diverse reasoning categories: physical, functional, logical, emotional, cultural, and social. The data curation process involved two steps - writing textual descriptions and generating images using text-to-image models, both requiring substantial human involvement to ensure high-quality and challenging scenes. Our experiments show that VLMs struggle significantly on NL-Eye, often performing at random baseline levels, while humans excel in both plausibility prediction and explanation quality. This demonstrates a deficiency in the abductive reasoning capabilities of modern VLMs. NL-Eye represents a crucial step toward developing VLMs capable of robust multimodal reasoning for real-world applications, including accident-prevention bots and generated video verification."
                    },
                    {
                        "title": "A Chain-of-Thought Is as Strong as Its Weakest Link: A Benchmark for Verifiers of Reasoning Chains",
                        "abstract": "Prompting language models to provide step-by-step answers (e.g., \"Chain-of-Thought\") is the prominent approach for complex reasoning tasks, where more accurate reasoning chains typically improve downstream task performance. Recent literature discusses automatic methods to verify reasoning to evaluate and improve their correctness. However, no fine-grained step-level datasets are available to enable thorough evaluation of such verification methods, hindering progress in this direction. We introduce REVEAL: Reasoning Verification Evaluation, a dataset to benchmark automatic verifiers of complex Chain-of-Thought reasoning in open-domain question-answering settings. REVEAL includes comprehensive labels for the relevance, attribution to evidence passages, and logical correctness of each reasoning step in a language model's answer, across a variety of datasets and state-of-the-art language models. Evaluation on REVEAL shows that verifiers struggle at verifying reasoning chains - in particular, verifying logical correctness and detecting contradictions. Available at https://reveal-dataset.github.io/ ."
                    }
                ]
            },
            "2f8a21b5-13bd-4df7-b201-99a15cd990fb": {
                "pk": "2f8a21b5-13bd-4df7-b201-99a15cd990fb",
                "name": "Idan Szpektor",
                "collaborators": [
                    "Roee Aharoni",
                    "Yonatan Bitton",
                    "Roi Reichart",
                    "Jonathan Herzig",
                    "Matan Eyal",
                    "Reut Tsarfaty",
                    "Leshem Choshen",
                    "Omri Abend",
                    "Eric Malmi",
                    "Orgad Keller"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Dialogue Systems",
                    "Multilingual Models",
                    "Knowledge Representation"
                ],
                "publications": [
                    {
                        "title": "The Yahoo Query Treebank, V. 1.0",
                        "abstract": "A description and annotation guidelines for the Yahoo Webscope release of Query Treebank, Version 1.0, May 2016."
                    },
                    {
                        "title": "DiscoFuse: A Large-Scale Dataset for Discourse-Based Sentence Fusion",
                        "abstract": "Sentence fusion is the task of joining several independent sentences into a single coherent text. Current datasets for sentence fusion are small and insufficient for training modern neural models. In this paper, we propose a method for automatically-generating fusion examples from raw text and present DiscoFuse, a large scale dataset for discourse-based sentence fusion. We author a set of rules for identifying a diverse set of discourse phenomena in raw text, and decomposing the text into two independent sentences. We apply our approach on two document collections: Wikipedia and Sports articles, yielding 60 million fusion examples annotated with discourse information required to reconstruct the fused text. We develop a sequence-to-sequence model on DiscoFuse and thoroughly analyze its strengths and weaknesses with respect to the various discourse phenomena, using both automatic as well as human evaluation. Finally, we conduct transfer learning experiments with WebSplit, a recent dataset for text simplification. We show that pretraining on DiscoFuse substantially improves performance on WebSplit when viewed as a sentence fusion task."
                    },
                    {
                        "title": "A Dataset for Sentence Retrieval for Open-Ended Dialogues",
                        "abstract": "We address the task of sentence retrieval for open-ended dialogues. The goal is to retrieve sentences from a document corpus that contain information useful for generating the next turn in a given dialogue. Prior work on dialogue-based retrieval focused on specific types of dialogues: either conversational QA or conversational search. To address a broader scope of this task where any type of dialogue can be used, we constructed a dataset that includes open-ended dialogues from Reddit, candidate sentences from Wikipedia for each dialogue and human annotations for the sentences. We report the performance of several retrieval baselines, including neural retrieval models, over the dataset. To adapt neural models to the types of dialogues in the dataset, we explored an approach to induce a large-scale weakly supervised training data from Reddit. Using this training set significantly improved the performance over training on the MS MARCO dataset."
                    },
                    {
                        "title": "Constructing Benchmarks and Interventions for Combating Hallucinations in LLMs",
                        "abstract": "Large language models (LLMs) are prone to hallucinations, which sparked a widespread effort to detect and prevent them. Recent work attempts to mitigate hallucinations by intervening in the model's generation, typically computing representative vectors of hallucinations vs. grounded generations, for steering the model's hidden states away from a hallucinatory state. However, common studies employ different setups and do not properly separate different possible causes of hallucinations, making interventions misguided. In this work, we introduce a method for categorizing examples based on the model's prior knowledge, named WACK. We construct WACK benchmarks that support interventions in two settings: open-book and closed-book question answering. Using the benchmarks, we perform an extensive investigation of the effect of different choices for intervention, such as the intervened components, and how often and how strongly to intervene. We find that intervention success varies depending on the component, with the attention blocks performing well and the residual stream proving detrimental to language modeling capabilities. We also show that interventions can benefit from representative vectors collected before, rather than after, a hallucination occurs. Finally, we introduce a new dynamic intervention, which intervenes only if needed, and thus is more robust than standard static interventions. The code is available at https://github.com/technion-cs-nlp/hallucination-mitigation ."
                    },
                    {
                        "title": "A Joint Named-Entity Recognizer for Heterogeneous Tag-sets Using a Tag Hierarchy",
                        "abstract": "We study a variant of domain adaptation for named-entity recognition where multiple, heterogeneously tagged training sets are available. Furthermore, the test tag-set is not identical to any individual training tag-set. Yet, the relations between all tags are provided in a tag hierarchy, covering the test tags as a combination of training tags. This setting occurs when various datasets are created using different annotation schemes. This is also the case of extending a tag-set with a new tag by annotating only the new tag in a new dataset. We propose to use the given tag hierarchy to jointly learn a neural network that shares its tagging layer among all tag-sets. We compare this model to combining independent models and to a model based on the multitasking approach. Our experiments show the benefit of the tag-hierarchy model, especially when facing non-trivial consolidation of tag-sets."
                    },
                    {
                        "title": "Semantically Driven Sentence Fusion: Modeling and Evaluation",
                        "abstract": "Sentence fusion is the task of joining related sentences into coherent text. Current training and evaluation schemes for this task are based on single reference ground-truths and do not account for valid fusion variants. We show that this hinders models from robustly capturing the semantic relationship between input sentences. To alleviate this, we present an approach in which ground-truth solutions are automatically expanded into multiple references via curated equivalence classes of connective phrases. We apply this method to a large-scale dataset and use the augmented dataset for both model training and evaluation. To improve the learning of semantic representation using multiple references, we enrich the model with auxiliary discourse classification tasks under a multi-tasking framework. Our experiments highlight the improvements of our approach over state-of-the-art models."
                    },
                    {
                        "title": "Multilingual Sequence-to-Sequence Models for Hebrew NLP",
                        "abstract": "Recent work attributes progress in NLP to large language models (LMs) with increased model size and large quantities of pretraining data. Despite this, current state-of-the-art LMs for Hebrew are both under-parameterized and under-trained compared to LMs in other languages. Additionally, previous work on pretrained Hebrew LMs focused on encoder-only models. While the encoder-only architecture is beneficial for classification tasks, it does not cater well for sub-word prediction tasks, such as Named Entity Recognition, when considering the morphologically rich nature of Hebrew. In this paper we argue that sequence-to-sequence generative architectures are more suitable for LLMs in the case of morphologically rich languages (MRLs) such as Hebrew. We demonstrate that by casting tasks in the Hebrew NLP pipeline as text-to-text tasks, we can leverage powerful multilingual, pretrained sequence-to-sequence models as mT5, eliminating the need for a specialized, morpheme-based, separately fine-tuned decoder. Using this approach, our experiments show substantial improvements over previously published results on existing Hebrew NLP benchmarks. These results suggest that multilingual sequence-to-sequence models present a promising building block for NLP for MRLs."
                    },
                    {
                        "title": "Contrastive Sequential-Diffusion Learning: An approach to Multi-Scene Instructional Video Synthesis",
                        "abstract": "Action-centric sequence descriptions like recipe instructions and do-it-yourself projects include non-linear patterns in which the next step may require to be visually consistent not on the immediate previous step but on earlier steps. Current video synthesis approaches fail to generate consistent multi-scene videos for such task descriptions. We propose a contrastive sequential video diffusion method that selects the most suitable previously generated scene to guide and condition the denoising process of the next scene. The result is a multi-scene video that is grounded in the scene descriptions and coherent w.r.t the scenes that require consistent visualisation. Our experiments with real-world data demonstrate the practicality and improved consistency of our model compared to prior work."
                    },
                    {
                        "title": "Beneath the Surface of Consistency: Exploring Cross-lingual Knowledge Representation Sharing in LLMs",
                        "abstract": "The veracity of a factoid is largely independent of the language it is written in. However, language models are inconsistent in their ability to answer the same factual question across languages. This raises questions about how LLMs represent a given fact across languages. We explore multilingual factual knowledge through two aspects: the model's ability to answer a query consistently across languages, and the ability to ''store'' answers in a shared representation for several languages. We propose a methodology to measure the extent of representation sharing across languages by repurposing knowledge editing methods. We examine LLMs with various multilingual configurations using a new multilingual dataset. We reveal that high consistency does not necessarily imply shared representation, particularly for languages with different scripts. Moreover, we find that script similarity is a dominant factor in representation sharing. Finally, we observe that if LLMs could fully share knowledge across languages, their accuracy in their best-performing language could benefit an increase of up to 150\\% on average. These findings highlight the need for improved multilingual knowledge representation in LLMs and suggest a path for the development of more robust and consistent multilingual LLMs."
                    },
                    {
                        "title": "On the Robustness of Dialogue History Representation in Conversational Question Answering: A Comprehensive Study and a New Prompt-based Method",
                        "abstract": "Most works on modeling the conversation history in Conversational Question Answering (CQA) report a single main result on a common CQA benchmark. While existing models show impressive results on CQA leaderboards, it remains unclear whether they are robust to shifts in setting (sometimes to more realistic ones), training data size (e.g. from large to small sets) and domain. In this work, we design and conduct the first large-scale robustness study of history modeling approaches for CQA. We find that high benchmark scores do not necessarily translate to strong robustness, and that various methods can perform extremely differently under different settings. Equipped with the insights from our study, we design a novel prompt-based history modeling approach, and demonstrate its strong robustness across various settings. Our approach is inspired by existing methods that highlight historic answers in the passage. However, instead of highlighting by modifying the passage token embeddings, we add textual prompts directly in the passage text. Our approach is simple, easy-to-plug into practically any model, and highly effective, thus we recommend it as a starting point for future model developers. We also hope that our study and insights will raise awareness to the importance of robustness-focused evaluation, in addition to obtaining high leaderboard scores, leading to better CQA systems."
                    },
                    {
                        "title": "Video-STaR: Self-Training Enables Video Instruction Tuning with Any Supervision",
                        "abstract": "The performance of Large Vision Language Models (LVLMs) is dependent on the size and quality of their training datasets. Existing video instruction tuning datasets lack diversity as they are derived by prompting large language models with video captions to generate question-answer pairs, and are therefore mostly descriptive. Meanwhile, many labeled video datasets with diverse labels and supervision exist - however, we find that their integration into LVLMs is non-trivial. Herein, we present Video Self-Training with augmented Reasoning (Video-STaR), the first video self-training approach. Video-STaR allows the utilization of any labeled video dataset for video instruction tuning. In Video-STaR, an LVLM cycles between instruction generation and finetuning, which we show (I) improves general video understanding and (II) adapts LVLMs to novel downstream tasks with existing supervision. During generation, an LVLM is prompted to propose an answer. The answers are then filtered only to those that contain the original video labels, and the LVLM is then re-trained on the generated dataset. By only training on generated answers that contain the correct video labels, Video-STaR utilizes these existing video labels as weak supervision for video instruction tuning. Our results demonstrate that Video-STaR-enhanced LVLMs exhibit improved performance in (I) general video QA, where TempCompass performance improved by 10%, and (II) on downstream tasks, where Video-STaR improved Kinetics700-QA accuracy by 20% and action quality assessment on FineDiving by 15%."
                    },
                    {
                        "title": "$Q^{2}$: Evaluating Factual Consistency in Knowledge-Grounded Dialogues via Question Generation and Question Answering",
                        "abstract": "Neural knowledge-grounded generative models for dialogue often produce content that is factually inconsistent with the knowledge they rely on, making them unreliable and limiting their applicability. Inspired by recent work on evaluating factual consistency in abstractive summarization, we propose an automatic evaluation metric for factual consistency in knowledge-grounded dialogue using automatic question generation and question answering. Our metric, denoted $Q^2$, compares answer spans using natural language inference (NLI), instead of token-based matching as done in previous work. To foster proper evaluation, we curate a novel dataset of dialogue system outputs for the Wizard-of-Wikipedia dataset, manually annotated for factual consistency. We perform a thorough meta-evaluation of $Q^2$ against other metrics using this dataset and two others, where it consistently shows higher correlation with human judgements."
                    },
                    {
                        "title": "All You May Need for VQA are Image Captions",
                        "abstract": "Visual Question Answering (VQA) has benefited from increasingly sophisticated models, but has not enjoyed the same level of engagement in terms of data creation. In this paper, we propose a method that automatically derives VQA examples at volume, by leveraging the abundance of existing image-caption annotations combined with neural models for textual question generation. We show that the resulting data is of high-quality. VQA models trained on our data improve state-of-the-art zero-shot accuracy by double digits and achieve a level of robustness that lacks in the same model trained on human-annotated VQA data."
                    },
                    {
                        "title": "DisentQA: Disentangling Parametric and Contextual Knowledge with Counterfactual Question Answering",
                        "abstract": "Question answering models commonly have access to two sources of \"knowledge\" during inference time: (1) parametric knowledge - the factual knowledge encoded in the model weights, and (2) contextual knowledge - external knowledge (e.g., a Wikipedia passage) given to the model to generate a grounded answer. Having these two sources of knowledge entangled together is a core issue for generative QA models as it is unclear whether the answer stems from the given non-parametric knowledge or not. This unclarity has implications on issues of trust, interpretability and factuality. In this work, we propose a new paradigm in which QA models are trained to disentangle the two sources of knowledge. Using counterfactual data augmentation, we introduce a model that predicts two answers for a given question: one based on given contextual knowledge and one based on parametric knowledge. Our experiments on the Natural Questions dataset show that this approach improves the performance of QA models by making them more robust to knowledge conflicts between the two knowledge sources, while generating useful disentangled answers."
                    },
                    {
                        "title": "Multilingual Instruction Tuning With Just a Pinch of Multilinguality",
                        "abstract": "As instruction-tuned large language models (LLMs) gain global adoption, their ability to follow instructions in multiple languages becomes increasingly crucial. In this work, we investigate how multilinguality during instruction tuning of a multilingual LLM affects instruction-following across languages from the pre-training corpus. We first show that many languages transfer some instruction-following capabilities to other languages from even monolingual tuning. Furthermore, we find that only 40 multilingual examples integrated in an English tuning set substantially improve multilingual instruction-following, both in seen and unseen languages during tuning. In general, we observe that models tuned on multilingual mixtures exhibit comparable or superior performance in multiple languages compared to monolingually tuned models, despite training on 10x fewer examples in those languages. Finally, we find that diversifying the instruction tuning set with even just 2-4 languages significantly improves cross-lingual generalization. Our results suggest that building massively multilingual instruction-tuned models can be done with only a very small set of multilingual instruction-responses."
                    },
                    {
                        "title": "What's the best place for an AI conference, Vancouver or ______: Why completing comparative questions is difficult",
                        "abstract": "Although large neural language models (LMs) like BERT can be finetuned to yield state-of-the-art results on many NLP tasks, it is often unclear what these models actually learn. Here we study using such LMs to fill in entities in human-authored comparative questions, like ``Which country is older, India or ______?'' -- i.e., we study the ability of neural LMs to ask (not answer) reasonable questions. We show that accuracy in this fill-in-the-blank task is well-correlated with human judgements of whether a question is reasonable, and that these models can be trained to achieve nearly human-level performance in completing comparative questions in three different subdomains. However, analysis shows that what they learn fails to model any sort of broad notion of which entities are semantically comparable or similar -- instead the trained models are very domain-specific, and performance is highly correlated with co-occurrences between specific entities observed in the training set. This is true both for models that are pretrained on general text corpora, as well as models trained on a large corpus of comparison questions. Our study thus reinforces recent results on the difficulty of making claims about a deep model's world knowledge or linguistic competence based on performance on specific benchmark problems. We make our evaluation datasets publicly available to foster future research on complex understanding and reasoning in such models at standards of human interaction."
                    },
                    {
                        "title": "TrueTeacher: Learning Factual Consistency Evaluation with Large Language Models",
                        "abstract": "Factual consistency evaluation is often conducted using Natural Language Inference (NLI) models, yet these models exhibit limited success in evaluating summaries. Previous work improved such models with synthetic training data. However, the data is typically based on perturbed human-written summaries, which often differ in their characteristics from real model-generated summaries and have limited coverage of possible factual errors. Alternatively, large language models (LLMs) have recently shown promising results in directly evaluating generative tasks, but are too computationally expensive for practical use. Motivated by these limitations, we introduce TrueTeacher, a method for generating synthetic data by annotating diverse model-generated summaries using a LLM. Unlike prior work, TrueTeacher does not rely on human-written summaries, and is multilingual by nature. Experiments on the TRUE benchmark show that a student model trained using our data, substantially outperforms both the state-of-the-art model with similar capacity, and the LLM teacher. In a systematic study, we compare TrueTeacher to existing synthetic data generation methods and demonstrate its superiority and robustness to domain-shift. We also show that our method generalizes to multilingual scenarios. Lastly, we release our large scale synthetic dataset (1.4M examples), generated using TrueTeacher, and a checkpoint trained on this data."
                    },
                    {
                        "title": "Transferring Visual Attributes from Natural Language to Verified Image Generation",
                        "abstract": "Text to image generation methods (T2I) are widely popular in generating art and other creative artifacts. While visual hallucinations can be a positive factor in scenarios where creativity is appreciated, such artifacts are poorly suited for cases where the generated image needs to be grounded in complex natural language without explicit visual elements. In this paper, we propose to strengthen the consistency property of T2I methods in the presence of natural complex language, which often breaks the limits of T2I methods by including non-visual information, and textual elements that require knowledge for accurate generation. To address these phenomena, we propose a Natural Language to Verified Image generation approach (NL2VI) that converts a natural prompt into a visual prompt, which is more suitable for image generation. A T2I model then generates an image for the visual prompt, which is then verified with VQA algorithms. Experimentally, aligning natural prompts with image generation can improve the consistency of the generated images by up to 11% over the state of the art. Moreover, improvements can generalize to challenging domains like cooking and DIY tasks, where the correctness of the generated image is crucial to illustrate actions."
                    },
                    {
                        "title": "VideoCon: Robust Video-Language Alignment via Contrast Captions",
                        "abstract": "Despite being (pre)trained on a massive amount of data, state-of-the-art video-language alignment models are not robust to semantically-plausible contrastive changes in the video captions. Our work addresses this by identifying a broad spectrum of contrast misalignments, such as replacing entities, actions, and flipping event order, which alignment models should be robust against. To this end, we introduce the VideoCon, a video-language alignment dataset constructed by a large language model that generates plausible contrast video captions and explanations for differences between original and contrast video captions. Then, a generative video-language model is finetuned with VideoCon to assess video-language entailment and generate explanations. Our VideoCon-based alignment model significantly outperforms current models. It exhibits a 12-point increase in AUC for the video-language alignment task on human-generated contrast captions. Finally, our model sets new state of the art zero-shot performance in temporally-extensive video-language tasks such as text-to-video retrieval (SSv2-Temporal) and video question answering (ATP-Hard). Moreover, our model shows superior performance on novel videos and human-crafted captions and explanations. Our code and data are available at https://github.com/Hritikbansal/videocon."
                    },
                    {
                        "title": "Unpacking Tokenization: Evaluating Text Compression and its Correlation with Model Performance",
                        "abstract": "Despite it being the cornerstone of BPE, the most common tokenization algorithm, the importance of compression in the tokenization process is still unclear. In this paper, we argue for the theoretical importance of compression, that can be viewed as 0-gram language modeling where equal probability is assigned to all tokens. We also demonstrate the empirical importance of compression for downstream success of pre-trained language models. We control the compression ability of several BPE tokenizers by varying the amount of documents available during their training: from 1 million documents to a character-based tokenizer equivalent to no training data at all. We then pre-train English language models based on those tokenizers and fine-tune them over several tasks. We show that there is a correlation between tokenizers' compression and models' downstream performance, suggesting that compression is a reliable intrinsic indicator of tokenization quality. These correlations are more pronounced for generation tasks (over classification) or for smaller models (over large ones). We replicated a representative part of our experiments on Turkish and found similar results, confirming that our results hold for languages with typological characteristics dissimilar to English. We conclude that building better compressing tokenizers is a fruitful avenue for further research and for improving overall model performance."
                    }
                ]
            },
            "fc0d3d2e-0b78-41c1-b076-71357e1c89eb": {
                "pk": "fc0d3d2e-0b78-41c1-b076-71357e1c89eb",
                "name": "Amir Globerson",
                "collaborators": [
                    "Alon Brutzkus",
                    "Trevor Darrell",
                    "Amir Bar",
                    "Gal Chechik",
                    "Naftali Tishby",
                    "Nir Rosenfeld",
                    "Roei Herzig",
                    "Nataly Brukhim",
                    "Tommi S. Jaakkola",
                    "Yoav Wald"
                ],
                "domain": [
                    "Machine Learning",
                    "Graphical Models",
                    "Deep Learning",
                    "Structured Prediction"
                ],
                "publications": [
                    {
                        "title": "The Minimum Information Principle for Discriminative Learning",
                        "abstract": "Exponential models of distributions are widely used in machine learning for classiffication and modelling. It is well known that they can be interpreted as maximum entropy models under empirical expectation constraints. In this work, we argue that for classiffication tasks, mutual information is a more suitable information theoretic measure to be optimized. We show how the principle of minimum mutual information generalizes that of maximum entropy, and provides a comprehensive framework for building discriminative classiffiers. A game theoretic interpretation of our approach is then given, and several generalization bounds provided. We present iterative algorithms for solving the minimum information problem and its convex dual, and demonstrate their performance on various classiffication tasks. The results show that minimum information classiffiers outperform the corresponding maximum entropy models."
                    },
                    {
                        "title": "Predict and Constrain: Modeling Cardinality in Deep Structured Prediction",
                        "abstract": "Many machine learning problems require the prediction of multi-dimensional labels. Such structured prediction models can benefit from modeling dependencies between labels. Recently, several deep learning approaches to structured prediction have been proposed. Here we focus on capturing cardinality constraints in such models. Namely, constraining the number of non-zero labels that the model outputs. Such constraints have proven very useful in previous structured prediction approaches, but it is a challenge to introduce them into a deep learning framework. Here we show how to do this via a novel deep architecture. Our approach outperforms strong baselines, achieving state-of-the-art results on multi-label classification benchmarks."
                    },
                    {
                        "title": "Convergent Propagation Algorithms via Oriented Trees",
                        "abstract": "Inference problems in graphical models are often approximated by casting them as constrained optimization problems. Message passing algorithms, such as belief propagation, have previously been suggested as methods for solving these optimization problems. However, there are few convergence guarantees for such algorithms, and the algorithms are therefore not guaranteed to solve the corresponding optimization problem. Here we present an oriented tree decomposition algorithm that is guaranteed to converge to the global optimum of the Tree-Reweighted (TRW) variational problem. Our algorithm performs local updates in the convex dual of the TRW problem - an unconstrained generalized geometric program. Primal updates, also local, correspond to oriented reparametrization operations that leave the distribution intact."
                    },
                    {
                        "title": "Globally Optimal Gradient Descent for a ConvNet with Gaussian Inputs",
                        "abstract": "Deep learning models are often successfully trained using gradient descent, despite the worst case hardness of the underlying non-convex optimization problem. The key question is then under what conditions can one prove that optimization will succeed. Here we provide a strong result of this kind. We consider a neural net with one hidden layer and a convolutional structure with no overlap and a ReLU activation function. For this architecture we show that learning is NP-complete in the general case, but that when the input distribution is Gaussian, gradient descent converges to the global optimum in polynomial time. To the best of our knowledge, this is the first global optimality guarantee of gradient descent on a convolutional neural network with ReLU activations."
                    },
                    {
                        "title": "Robust Conditional Probabilities",
                        "abstract": "Conditional probabilities are a core concept in machine learning. For example, optimal prediction of a label $Y$ given an input $X$ corresponds to maximizing the conditional probability of $Y$ given $X$. A common approach to inference tasks is learning a model of conditional probabilities. However, these models are often based on strong assumptions (e.g., log-linear models), and hence their estimate of conditional probabilities is not robust and is highly dependent on the validity of their assumptions.   Here we propose a framework for reasoning about conditional probabilities without assuming anything about the underlying distributions, except knowledge of their second order marginals, which can be estimated from data. We show how this setting leads to guaranteed bounds on conditional probabilities, which can be calculated efficiently in a variety of settings, including structured-prediction. Finally, we apply them to semi-supervised deep learning, obtaining results competitive with variational autoencoders."
                    },
                    {
                        "title": "Why do Larger Models Generalize Better? A Theoretical Perspective via the XOR Problem",
                        "abstract": "Empirical evidence suggests that neural networks with ReLU activations generalize better with over-parameterization. However, there is currently no theoretical analysis that explains this observation. In this work, we provide theoretical and empirical evidence that, in certain cases, overparameterized convolutional networks generalize better than small networks because of an interplay between weight clustering and feature exploration at initialization. We demonstrate this theoretically for a 3-layer convolutional neural network with max-pooling, in a novel setting which extends the XOR problem. We show that this interplay implies that with overparamterization, gradient descent converges to global minima with better generalization performance compared to global minima of small networks. Empirically, we demonstrate these phenomena for a 3-layer convolutional neural network in the MNIST task."
                    },
                    {
                        "title": "What Cannot be Learned with Bethe Approximations",
                        "abstract": "We address the problem of learning the parameters in graphical models when inference is intractable. A common strategy in this case is to replace the partition function with its Bethe approximation. We show that there exists a regime of empirical marginals where such Bethe learning will fail. By failure we mean that the empirical marginals cannot be recovered from the approximated maximum likelihood parameters (i.e., moment matching is not achieved). We provide several conditions on empirical marginals that yield outer and inner bounds on the set of Bethe learnable marginals. An interesting implication of our results is that there exists a large class of marginals that cannot be obtained as stable fixed points of belief propagation. Taken together our results provide a novel approach to analyzing learning with Bethe approximations and highlight when it can be expected to work or fail."
                    },
                    {
                        "title": "Semi-Supervised Learning with Competitive Infection Models",
                        "abstract": "The goal in semi-supervised learning is to effectively combine labeled and unlabeled data. One way to do this is by encouraging smoothness across edges in a graph whose nodes correspond to input examples. In many graph-based methods, labels can be thought of as propagating over the graph, where the underlying propagation mechanism is based on random walks or on averaging dynamics. While theoretically elegant, these dynamics suffer from several drawbacks which can hurt predictive performance.   Our goal in this work is to explore alternative mechanisms for propagating labels. In particular, we propose a method based on dynamic infection processes, where unlabeled nodes can be \"infected\" with the label of their already infected neighbors. Our algorithm is efficient and scalable, and an analysis of the underlying optimization objective reveals a surprising relation to other Laplacian approaches. We conclude with a thorough set of experiments across multiple benchmarks and various learning settings."
                    },
                    {
                        "title": "An Optimization and Generalization Analysis for Max-Pooling Networks",
                        "abstract": "Max-Pooling operations are a core component of deep learning architectures. In particular, they are part of most convolutional architectures used in machine vision, since pooling is a natural approach to pattern detection problems. However, these architectures are not well understood from a theoretical perspective. For example, we do not understand when they can be globally optimized, and what is the effect of over-parameterization on generalization. Here we perform a theoretical analysis of a convolutional max-pooling architecture, proving that it can be globally optimized, and can generalize well even for highly over-parameterized models. Our analysis focuses on a data generating distribution inspired by pattern detection problem, where a \"discriminative\" pattern needs to be detected among \"spurious\" patterns. We empirically validate that CNNs significantly outperform fully connected networks in our setting, as predicted by our theoretical results."
                    },
                    {
                        "title": "Finding Visual Task Vectors",
                        "abstract": "Visual Prompting is a technique for teaching models to perform a visual task via in-context examples, without any additional training. In this work, we analyze the activations of MAE-VQGAN, a recent Visual Prompting model, and find task vectors, activations that encode task-specific information. Equipped with this insight, we demonstrate that it is possible to identify the task vectors and use them to guide the network towards performing different tasks without providing any input-output examples. To find task vectors, we compute the average intermediate activations per task and use the REINFORCE algorithm to search for the subset of task vectors. The resulting task vectors guide the model towards performing a task better than the original model without the need for input-output examples."
                    },
                    {
                        "title": "Discriminative Learning via Semidefinite Probabilistic Models",
                        "abstract": "Discriminative linear models are a popular tool in machine learning. These can be generally divided into two types: The first is linear classifiers, such as support vector machines, which are well studied and provide state-of-the-art results. One shortcoming of these models is that their output (known as the 'margin') is not calibrated, and cannot be translated naturally into a distribution over the labels. Thus, it is difficult to incorporate such models as components of larger systems, unlike probabilistic based approaches. The second type of approach constructs class conditional distributions using a nonlinearity (e.g. log-linear models), but is occasionally worse in terms of classification error. We propose a supervised learning method which combines the best of both approaches. Specifically, our method provides a distribution over the labels, which is a linear function of the model parameters. As a consequence, differences between probabilities are linear functions, a property which most probabilistic models (e.g. log-linear) do not have.   Our model assumes that classes correspond to linear subspaces (rather than to half spaces). Using a relaxed projection operator, we construct a measure which evaluates the degree to which a given vector 'belongs' to a subspace, resulting in a distribution over labels. Interestingly, this view is closely related to similar concepts in quantum detection theory. The resulting models can be trained either to maximize the margin or to optimize average likelihood measures. The corresponding optimization problems are semidefinite programs which can be solved efficiently. We illustrate the performance of our algorithm on real world datasets, and show that it outperforms 2nd order kernel methods."
                    },
                    {
                        "title": "Optimal Tagging with Markov Chain Optimization",
                        "abstract": "Many information systems use tags and keywords to describe and annotate content. These allow for efficient organization and categorization of items, as well as facilitate relevant search queries. As such, the selected set of tags for an item can have a considerable effect on the volume of traffic that eventually reaches an item. In settings where tags are chosen by an item's creator, who in turn is interested in maximizing traffic, a principled approach for choosing tags can prove valuable. In this paper we introduce the problem of optimal tagging, where the task is to choose a subset of tags for a new item such that the probability of a browsing user reaching that item is maximized. We formulate the problem by modeling traffic using a Markov chain, and asking how transitions in this chain should be modified to maximize traffic into a certain state of interest. The resulting optimization problem involves maximizing a certain function over subsets, under a cardinality constraint. We show that the optimization problem is NP-hard, but nonetheless has a simple (1-1/e)-approximation via a simple greedy algorithm. Furthermore, the structure of the problem allows for an efficient implementation of the greedy step.To demonstrate the effectiveness of our method, we perform experiments on three tagging datasets, and show that the greedy algorithm outperforms other baselines."
                    },
                    {
                        "title": "Towards Understanding Learning in Neural Networks with Linear Teachers",
                        "abstract": "Can a neural network minimizing cross-entropy learn linearly separable data? Despite progress in the theory of deep learning, this question remains unsolved. Here we prove that SGD globally optimizes this learning problem for a two-layer network with Leaky ReLU activations. The learned network can in principle be very complex. However, empirical evidence suggests that it often turns out to be approximately linear. We provide theoretical support for this phenomenon by proving that if network weights converge to two weight clusters, this will imply an approximately linear decision boundary. Finally, we show a condition on the optimization that leads to weight clustering. We provide empirical results that validate our theoretical analysis."
                    },
                    {
                        "title": "Gaussian Robust Classification",
                        "abstract": "Supervised learning is all about the ability to generalize knowledge. Specifically, the goal of the learning is to train a classifier using training data, in such a way that it will be capable of classifying new unseen data correctly. In order to acheive this goal, it is important to carefully design the learner, so it will not overfit the training data. The later can is done usually by adding a regularization term. The statistical learning theory explains the success of this method by claiming that it restricts the complexity of the learned model. This explanation, however, is rather abstract and does not have a geometric intuition. The generalization error of a classifier may be thought of as correlated with its robustness to perturbations of the data: a classifier that copes with disturbance is expected to generalize well. Indeed, Xu et al. [2009] have shown that the SVM formulation is equivalent to a robust optimization (RO) formulation, in which an adversary displaces the training and testing points within a ball of pre-determined radius. In this work we explore a different kind of robustness, namely changing each data point with a Gaussian cloud centered at the sample. Loss is evaluated as the expectation of an underlying loss function on the cloud. This setup fits the fact that in many applications, the data is sampled along with noise. We develop an RO framework, in which the adversary chooses the covariance of the noise. In our algorithm named GURU, the tuning parameter is a spectral bound on the noise, thus it can be estimated using physical or applicative considerations. Our experiments show that this framework performs as well as SVM and even slightly better in some cases. Generalizations for Mercer kernels and for the multiclass case are presented as well. We also show that our framework may be further generalized, using the technique of convex perspective functions."
                    },
                    {
                        "title": "Learning Canonical Representations for Scene Graph to Image Generation",
                        "abstract": "Generating realistic images of complex visual scenes becomes challenging when one wishes to control the structure of the generated images. Previous approaches showed that scenes with few entities can be controlled using scene graphs, but this approach struggles as the complexity of the graph (the number of objects and edges) increases. In this work, we show that one limitation of current methods is their inability to capture semantic equivalence in graphs. We present a novel model that addresses these issues by learning canonical graph representations from the data, resulting in improved image generation for complex visual scenes. Our model demonstrates improved empirical performance on large scene graphs, robustness to noise in the input scene graph, and generalization on semantically equivalent graphs. Finally, we show improved performance of the model on three different benchmarks: Visual Genome, COCO, and CLEVR."
                    },
                    {
                        "title": "Compositional Video Synthesis with Action Graphs",
                        "abstract": "Videos of actions are complex signals containing rich compositional structure in space and time. Current video generation methods lack the ability to condition the generation on multiple coordinated and potentially simultaneous timed actions. To address this challenge, we propose to represent the actions in a graph structure called Action Graph and present the new ``Action Graph To Video'' synthesis task. Our generative model for this task (AG2Vid) disentangles motion and appearance features, and by incorporating a scheduling mechanism for actions facilitates a timely and coordinated video generation. We train and evaluate AG2Vid on the CATER and Something-Something V2 datasets, and show that the resulting videos have better visual quality and semantic consistency compared to baselines. Finally, our model demonstrates zero-shot abilities by synthesizing novel compositions of the learned actions. For code and pretrained models, see the project page https://roeiherz.github.io/AG2Video"
                    },
                    {
                        "title": "Visual Prompting via Image Inpainting",
                        "abstract": "How does one adapt a pre-trained visual model to novel downstream tasks without task-specific finetuning or any model modification? Inspired by prompting in NLP, this paper investigates visual prompting: given input-output image example(s) of a new task at test time and a new input image, the goal is to automatically produce the output image, consistent with the given examples. We show that posing this problem as simple image inpainting - literally just filling in a hole in a concatenated visual prompt image - turns out to be surprisingly effective, provided that the inpainting algorithm has been trained on the right data. We train masked auto-encoders on a new dataset that we curated - 88k unlabeled figures from academic papers sources on Arxiv. We apply visual prompting to these pretrained models and demonstrate results on various downstream image-to-image tasks, including foreground segmentation, single object detection, colorization, edge detection, etc."
                    },
                    {
                        "title": "Sufficient Dimensionality Reduction with Irrelevant Statistics",
                        "abstract": "The problem of finding a reduced dimensionality representation of categorical variables while preserving their most relevant characteristics is fundamental for the analysis of complex data. Specifically, given a co-occurrence matrix of two variables, one often seeks a compact representation of one variable which preserves information about the other variable. We have recently introduced ``Sufficient Dimensionality Reduction' [GT-2003], a method that extracts continuous reduced dimensional features whose measurements (i.e., expectation values) capture maximal mutual information among the variables. However, such measurements often capture information that is irrelevant for a given task. Widely known examples are illumination conditions, which are irrelevant as features for face recognition, writing style which is irrelevant as a feature for content classification, and intonation which is irrelevant as a feature for speech recognition. Such irrelevance cannot be deduced apriori, since it depends on the details of the task, and is thus inherently ill defined in the purely unsupervised case. Separating relevant from irrelevant features can be achieved using additional side data that contains such irrelevant structures. This approach was taken in [CT-2002], extending the information bottleneck method, which uses clustering to compress the data. Here we use this side-information framework to identify features whose measurements are maximally informative for the original data set, but carry as little information as possible on a side data set. In statistical terms this can be understood as extracting statistics which are maximally sufficient for the original dataset, while simultaneously maximally ancillary for the side dataset. We formulate this tradeoff as a constrained optimization problem and characterize its solutions. We then derive a gradient descent algorithm for this problem, which is based on the Generalized Iterative Scaling method for finding maximum entropy distributions. The method is demonstrated on synthetic data, as well as on real face recognition datasets, and is shown to outperform standard methods such as oriented PCA."
                    },
                    {
                        "title": "Covering Uncommon Ground: Gap-Focused Question Generation for Answer Assessment",
                        "abstract": "Human communication often involves information gaps between the interlocutors. For example, in an educational dialogue, a student often provides an answer that is incomplete, and there is a gap between this answer and the perfect one expected by the teacher. Successful dialogue then hinges on the teacher asking about this gap in an effective manner, thus creating a rich and interactive educational experience. We focus on the problem of generating such gap-focused questions (GFQs) automatically. We define the task, highlight key desired aspects of a good GFQ, and propose a model that satisfies these. Finally, we provide an evaluation by human annotators of our generated questions compared against human generated ones, demonstrating competitive performance."
                    },
                    {
                        "title": "Convexifying the Bethe Free Energy",
                        "abstract": "The introduction of loopy belief propagation (LBP) revitalized the application of graphical models in many domains. Many recent works present improvements on the basic LBP algorithm in an attempt to overcome convergence and local optima problems. Notable among these are convexified free energy approximations that lead to inference procedures with provable convergence and quality properties. However, empirically LBP still outperforms most of its convex variants in a variety of settings, as we also demonstrate here. Motivated by this fact we seek convexified free energies that directly approximate the Bethe free energy. We show that the proposed approximations compare favorably with state-of-the art convex free energy approximations."
                    }
                ]
            },
            "6656a439-ed94-4e85-917c-263b2906d612": {
                "pk": "6656a439-ed94-4e85-917c-263b2906d612",
                "name": "Yuval Elovici",
                "collaborators": [
                    "Mordechai Guri",
                    "Lior Rokach",
                    "Ziv Katzir",
                    "Eitan Menahem",
                    "Yuval Shahar",
                    "Yisroel Mirsky",
                    "Asaf Shabtai",
                    "Jaidip Kotak",
                    "Tom Mahler",
                    "Naor Kalbo"
                ],
                "domain": [
                    "Cybersecurity",
                    "Machine Learning",
                    "IoT Security",
                    "Risk Assessment"
                ],
                "publications": [
                    {
                        "title": "A New Methodology for Information Security Risk Assessment for Medical Devices and Its Evaluation",
                        "abstract": "As technology advances towards more connected and digital environments, medical devices are becoming increasingly connected to hospital networks and to the Internet, which exposes them, and thus the patients using them, to new cybersecurity threats. Currently, there is a lack of a methodology dedicated to information security risk assessment for medical devices.   In this study, we present the Threat identification, ontology-based Likelihood, severity Decomposition, and Risk integration (TLDR) methodology for information security risk assessment for medical devices. The TLDR methodology uses the following steps: (1) identifying the potentially vulnerable components of medical devices, in this case, four different medical imaging devices (MIDs); (2) identifying the potential attacks, in this case, 23 potential attacks on MIDs; (3) mapping the discovered attacks into a known attack ontology - in this case, the Common Attack Pattern Enumeration and Classifications (CAPECs); (4) estimating the likelihood of the mapped CAPECs in the medical domain with the assistance of a panel of senior healthcare Information Security Experts (ISEs); (5) computing the CAPEC-based likelihood estimates of each attack; (6) decomposing each attack into several severity aspects and assigning them weights; (7) assessing the magnitude of the impact of each of the severity aspects for each attack with the assistance of a panel of senior Medical Experts (MEs); (8) computing the composite severity assessments for each attack; and finally, (9) integrating the likelihood and severity of each attack into its risk, and thus prioritizing it. The details of steps six to eight are beyond the scope of the current study; in the current study, we had replaced them by a single step that included asking the panel of MEs [in this case, radiologists], to assess the overall severity for each attack and use it as its severity..."
                    },
                    {
                        "title": "Reaction to New Security Threat Class",
                        "abstract": "Each new identified security threat class triggers new research and development efforts by the scientific and professional communities. In this study, we investigate the rate at which the scientific and professional communities react to new identified threat classes as it is reflected in the number of patents, scientific articles and professional publications over a long period of time. The following threat classes were studied: Phishing; SQL Injection; BotNet; Distributed Denial of Service; and Advanced Persistent Threat. Our findings suggest that in most cases it takes a year for the scientific community and more than two years for industry to react to a new threat class with patents. Since new products follow patents, it is reasonable to expect that there will be a window of approximately two to three years in which no effective product is available to cope with the new threat class."
                    },
                    {
                        "title": "The Security of IP-based Video Surveillance Systems",
                        "abstract": "IP-based Surveillance systems protect industrial facilities, railways, gas stations, and even one's own home. Therefore, unauthorized access to these systems has serious security implications. In this survey, we analyze the system's (1) threat agents, (2) attack goals, (3) practical attacks, (4) possible attack outcomes, and (5) provide example attack vectors."
                    },
                    {
                        "title": "Data Mining of Online Genealogy Datasets for Revealing Lifespan Patterns in Human Population",
                        "abstract": "Online genealogy datasets contain extensive information about millions of people and their past and present family connections. This vast amount of data can assist in identifying various patterns in human population. In this study, we present methods and algorithms which can assist in identifying variations in lifespan distributions of human population in the past centuries, in detecting social and genetic features which correlate with human lifespan, and in constructing predictive models of human lifespan based on various features which can easily be extracted from genealogy datasets.   We have evaluated the presented methods and algorithms on a large online genealogy dataset with over a million profiles and over 9 million connections, all of which were collected from the WikiTree website. Our findings indicate that significant but small positive correlations exist between the parents' lifespan and their children's lifespan. Additionally, we found slightly higher and significant correlations between the lifespans of spouses. We also discovered a very small positive and significant correlation between longevity and reproductive success in males, and a small and significant negative correlation between longevity and reproductive success in females. Moreover, our machine learning algorithms presented better than random classification results in predicting which people who outlive the age of 50 will also outlive the age of 80.   We believe that this study will be the first of many studies which utilize the wealth of data on human populations, existing in online genealogy datasets, to better understand factors which influence human lifespan. Understanding these factors can assist scientists in providing solutions for successful aging."
                    },
                    {
                        "title": "Why Blocking Targeted Adversarial Perturbations Impairs the Ability to Learn",
                        "abstract": "Despite their accuracy, neural network-based classifiers are still prone to manipulation through adversarial perturbations. Those perturbations are designed to be misclassified by the neural network, while being perceptually identical to some valid input. The vast majority of attack methods rely on white-box conditions, where the attacker has full knowledge of the attacked network's parameters. This allows the attacker to calculate the network's loss gradient with respect to some valid input and use this gradient in order to create an adversarial example. The task of blocking white-box attacks has proven difficult to solve. While a large number of defense methods have been suggested, they have had limited success. In this work we examine this difficulty and try to understand it. We systematically explore the abilities and limitations of defensive distillation, one of the most promising defense mechanisms against adversarial perturbations suggested so far in order to understand the defense challenge. We show that contrary to commonly held belief, the ability to bypass defensive distillation is not dependent on an attack's level of sophistication. In fact, simple approaches, such as the Targeted Gradient Sign Method, are capable of effectively bypassing defensive distillation. We prove that defensive distillation is highly effective against non-targeted attacks but is unsuitable for targeted attacks. This discovery leads us to realize that targeted attacks leverage the same input gradient that allows a network to be trained. This implies that blocking them will require losing the network's ability to learn, presenting an impossible tradeoff to the research community."
                    },
                    {
                        "title": "IoT Device Identification Using Deep Learning",
                        "abstract": "The growing use of IoT devices in organizations has increased the number of attack vectors available to attackers due to the less secure nature of the devices. The widely adopted bring your own device (BYOD) policy which allows an employee to bring any IoT device into the workplace and attach it to an organization's network also increases the risk of attacks. In order to address this threat, organizations often implement security policies in which only the connection of white-listed IoT devices is permitted. To monitor adherence to such policies and protect their networks, organizations must be able to identify the IoT devices connected to their networks and, more specifically, to identify connected IoT devices that are not on the white-list (unknown devices). In this study, we applied deep learning on network traffic to automatically identify IoT devices connected to the network. In contrast to previous work, our approach does not require that complex feature engineering be applied on the network traffic, since we represent the communication behavior of IoT devices using small images built from the IoT devices network traffic payloads. In our experiments, we trained a multiclass classifier on a publicly available dataset, successfully identifying 10 different IoT devices and the traffic of smartphones and computers, with over 99% accuracy. We also trained multiclass classifiers to detect unauthorized IoT devices connected to the network, achieving over 99% overall average detection accuracy."
                    },
                    {
                        "title": "Who's Afraid of Adversarial Transferability?",
                        "abstract": "Adversarial transferability, namely the ability of adversarial perturbations to simultaneously fool multiple learning models, has long been the \"big bad wolf\" of adversarial machine learning. Successful transferability-based attacks requiring no prior knowledge of the attacked model's parameters or training data have been demonstrated numerous times in the past, implying that machine learning models pose an inherent security threat to real-life systems. However, all of the research performed in this area regarded transferability as a probabilistic property and attempted to estimate the percentage of adversarial examples that are likely to mislead a target model given some predefined evaluation set. As a result, those studies ignored the fact that real-life adversaries are often highly sensitive to the cost of a failed attack. We argue that overlooking this sensitivity has led to an exaggerated perception of the transferability threat, when in fact real-life transferability-based attacks are quite unlikely. By combining theoretical reasoning with a series of empirical results, we show that it is practically impossible to predict whether a given adversarial example is transferable to a specific target model in a black-box setting, hence questioning the validity of adversarial transferability as a real-life attack tool for adversaries that are sensitive to the cost of a failed attack."
                    },
                    {
                        "title": "IoT Device Identification Based on Network Communication Analysis Using Deep Learning",
                        "abstract": "Attack vectors for adversaries have increased in organizations because of the growing use of less secure IoT devices. The risk of attacks on an organization's network has also increased due to the bring your own device (BYOD) policy which permits employees to bring IoT devices onto the premises and attach them to the organization's network. To tackle this threat and protect their networks, organizations generally implement security policies in which only white listed IoT devices are allowed on the organization's network. To monitor compliance with such policies, it has become essential to distinguish IoT devices permitted within an organization's network from non white listed (unknown) IoT devices. In this research, deep learning is applied to network communication for the automated identification of IoT devices permitted on the network. In contrast to existing methods, the proposed approach does not require complex feature engineering of the network communication, because the 'communication behavior' of IoT devices is represented as small images which are generated from the device's network communication payload. The proposed approach is applicable for any IoT device, regardless of the protocol used for communication. As our approach relies on the network communication payload, it is also applicable for the IoT devices behind a network address translation (NAT) enabled router. In this study, we trained various classifiers on a publicly accessible dataset to identify IoT devices in different scenarios, including the identification of known and unknown IoT devices, achieving over 99% overall average detection accuracy."
                    },
                    {
                        "title": "LLMCloudHunter: Harnessing LLMs for Automated Extraction of Detection Rules from Cloud-Based CTI",
                        "abstract": "As the number and sophistication of cyber attacks have increased, threat hunting has become a critical aspect of active security, enabling proactive detection and mitigation of threats before they cause significant harm. Open-source cyber threat intelligence (OS-CTI) is a valuable resource for threat hunters, however, it often comes in unstructured formats that require further manual analysis. Previous studies aimed at automating OSCTI analysis are limited since (1) they failed to provide actionable outputs, (2) they did not take advantage of images present in OSCTI sources, and (3) they focused on on-premises environments, overlooking the growing importance of cloud environments. To address these gaps, we propose LLMCloudHunter, a novel framework that leverages large language models (LLMs) to automatically generate generic-signature detection rule candidates from textual and visual OSCTI data. We evaluated the quality of the rules generated by the proposed framework using 12 annotated real-world cloud threat reports. The results show that our framework achieved a precision of 92% and recall of 98% for the task of accurately extracting API calls made by the threat actor and a precision of 99% with a recall of 98% for IoCs. Additionally, 99.18% of the generated detection rule candidates were successfully compiled and converted into Splunk queries."
                    },
                    {
                        "title": "dr0wned - Cyber-Physical Attack with Additive Manufacturing",
                        "abstract": "Additive manufacturing (AM), or 3D printing, is an emerging manufacturing technology that is expected to have far-reaching socioeconomic, environmental, and geopolitical implications. As use of this technology increases, it will become more common to produce functional parts, including components for safety-critical systems. AM's dependence on computerization raises the concern that the manufactured part's quality can be compromised by sabotage. This paper demonstrates the validity of this concern, as we present the very first full chain of attack involving AM, beginning with a cyber attack aimed at compromising a benign AM component, continuing with malicious modification of a manufactured object's blueprint, leading to the sabotage of the manufactured functional part, and resulting in the physical destruction of a cyber-physical system that employs this part. The contributions of this paper are as follows. We propose a systematic approach to identify opportunities for an attack involving AM that enables an adversary to achieve his/her goals. Then we propose a methodology to assess the level of difficulty of an attack, thus enabling differentiation between possible attack chains. Finally, to demonstrate the experimental proof for the entire attack chain, we sabotage the 3D printed propeller of a quadcopter UAV, causing the quadcopter to literally fall from the sky."
                    },
                    {
                        "title": "USBee: Air-Gap Covert-Channel via Electromagnetic Emission from USB",
                        "abstract": "In recent years researchers have demonstrated how attackers could use USB connectors implanted with RF transmitters to exfiltrate data from secure, and even air-gapped, computers (e.g., COTTONMOUTH in the leaked NSA ANT catalog). Such methods require a hardware modification of the USB plug or device, in which a dedicated RF transmitter is embedded. In this paper we present USBee, a software that can utilize an unmodified USB device connected to a computer as a RF transmitter. We demonstrate how a software can intentionally generate controlled electromagnetic emissions from the data bus of a USB connector. We also show that the emitted RF signals can be controlled and modulated with arbitrary binary data. We implement a prototype of USBee, and discuss its design and implementation details including signal generation and modulation. We evaluate the transmitter by building a receiver and demodulator using GNU Radio. Our evaluation shows that USBee can be used for transmitting binary data to a nearby receiver at a bandwidth of 20 to 80 BPS (bytes per second)."
                    },
                    {
                        "title": "Detecting Adversarial Perturbations Through Spatial Behavior in Activation Spaces",
                        "abstract": "Neural network based classifiers are still prone to manipulation through adversarial perturbations. State of the art attacks can overcome most of the defense or detection mechanisms suggested so far, and adversaries have the upper hand in this arms race. Adversarial examples are designed to resemble the normal input from which they were constructed, while triggering an incorrect classification. This basic design goal leads to a characteristic spatial behavior within the context of Activation Spaces, a term coined by the authors to refer to the hyperspaces formed by the activation values of the network's layers. Within the output of the first layers of the network, an adversarial example is likely to resemble normal instances of the source class, while in the final layers such examples will diverge towards the adversary's target class. The steps below enable us to leverage this inherent shift from one class to another in order to form a novel adversarial example detector. We construct Euclidian spaces out of the activation values of each of the deep neural network layers. Then, we induce a set of k-nearest neighbor classifiers (k-NN), one per activation space of each neural network layer, using the non-adversarial examples. We leverage those classifiers to produce a sequence of class labels for each nonperturbed input sample and estimate the a priori probability for a class label change between one activation space and another. During the detection phase we compute a sequence of classification labels for each input using the trained classifiers. We then estimate the likelihood of those classification sequences and show that adversarial sequences are far less likely than normal ones. We evaluated our detection method against the state of the art C&W attack method, using two image classification datasets (MNIST, CIFAR-10) reaching an AUC 0f 0.95 for the CIFAR-10 dataset."
                    },
                    {
                        "title": "Personal Information Leakage During Password Recovery of Internet Services",
                        "abstract": "In this paper we examine the standard password recovery process of large Internet services such as Gmail, Facebook, and Twitter. Although most of these services try to maintain user privacy, with regard to registration information and other personal information provided by the user, we demonstrate that personal information can still be obtained by unauthorized individuals or attackers. This information includes the full (or partial) email address, phone number, friends list, address, etc. We examine different scenarios and demonstrate how the details revealed in the password recovery process can be used to deduct more focused information about users."
                    },
                    {
                        "title": "9-1-1 DDoS: Threat, Analysis and Mitigation",
                        "abstract": "The 911 emergency service belongs to one of the 16 critical infrastructure sectors in the United States. Distributed denial of service (DDoS) attacks launched from a mobile phone botnet pose a significant threat to the availability of this vital service. In this paper we show how attackers can exploit the cellular network protocols in order to launch an anonymized DDoS attack on 911. The current FCC regulations require that all emergency calls be immediately routed regardless of the caller's identifiers (e.g., IMSI and IMEI). A rootkit placed within the baseband firmware of a mobile phone can mask and randomize all cellular identifiers, causing the device to have no genuine identification within the cellular network. Such anonymized phones can issue repeated emergency calls that cannot be blocked by the network or the emergency call centers, technically or legally. We explore the 911 infrastructure and discuss why it is susceptible to this kind of attack. We then implement different forms of the attack and test our implementation on a small cellular network. Finally, we simulate and analyze anonymous attacks on a model of current 911 infrastructure in order to measure the severity of their impact. We found that with less than 6K bots (or $100K hardware), attackers can block emergency services in an entire state (e.g., North Carolina) for days. We believe that this paper will assist the respective organizations, lawmakers, and security professionals in understanding the scope of this issue in order to prevent possible 911-DDoS attacks in the future."
                    },
                    {
                        "title": "PowerHammer: Exfiltrating Data from Air-Gapped Computers through Power Lines",
                        "abstract": "In this paper we provide an implementation, evaluation, and analysis of PowerHammer, a malware (bridgeware [1]) that uses power lines to exfiltrate data from air-gapped computers. In this case, a malicious code running on a compromised computer can control the power consumption of the system by intentionally regulating the CPU utilization. Data is modulated, encoded, and transmitted on top of the current flow fluctuations, and then it is conducted and propagated through the power lines. This phenomena is known as a 'conducted emission'. We present two versions of the attack. Line level powerhammering: In this attack, the attacker taps the in-home power lines1 that are directly attached to the electrical outlet. Phase level power-hammering: In this attack, the attacker taps the power lines at the phase level, in the main electrical service panel. In both versions of the attack, the attacker measures the emission conducted and then decodes the exfiltrated data. We describe the adversarial attack model and present modulations and encoding schemes along with a transmission protocol. We evaluate the covert channel in different scenarios and discuss signal-to-noise (SNR), signal processing, and forms of interference. We also present a set of defensive countermeasures. Our results show that binary data can be covertly exfiltrated from air-gapped computers through the power lines at bit rates of 1000 bit/sec for the line level power-hammering attack and 10 bit/sec for the phase level power-hammering attack."
                    },
                    {
                        "title": "Temporal Pattern Discovery for Accurate Sepsis Diagnosis in ICU Patients",
                        "abstract": "Sepsis is a condition caused by the body's overwhelming and life-threatening response to infection, which can lead to tissue damage, organ failure, and finally death. Common signs and symptoms include fever, increased heart rate, increased breathing rate, and confusion. Sepsis is difficult to predict, diagnose, and treat. Patients who develop sepsis have an increased risk of complications and death and face higher health care costs and longer hospitalization. Today, sepsis is one of the leading causes of mortality among populations in intensive care units (ICUs). In this paper, we look at the problem of early detection of sepsis by using temporal data mining. We focus on the use of knowledge-based temporal abstraction to create meaningful interval-based abstractions, and on time-interval mining to discover frequent interval-based patterns. We used 2,560 cases derived from the MIMIC-III database. We found that the distribution of the temporal patterns whose frequency is above 10% discovered in the records of septic patients during the last 6 and 12 hours before onset of sepsis is significantly different from that distribution within a similar period, during an equivalent time window during hospitalization, in the records of non-septic patients. This discovery is encouraging for the purpose of performing an early diagnosis of sepsis using the discovered patterns as constructed features."
                    },
                    {
                        "title": "Incremental Deployment of Network Monitors Based on Group Betweenness Centrality",
                        "abstract": "In many applications we are required to increase the deployment of a distributed monitoring system on an evolving network. In this paper we present a new method for finding candidate locations for additional deployment in the network. This method is based on the Group Betweenness Centrality (GBC) measure that is used to estimate the influence of a group of nodes over the information flow in the network. The new method assists in finding the location of k additional monitors in the evolving network, such that the portion of additional traffic covered is at least (1-1/e) of the optimal."
                    },
                    {
                        "title": "Combining One-Class Classifiers via Meta-Learning",
                        "abstract": "Selecting the best classifier among the available ones is a difficult task, especially when only instances of one class exist. In this work we examine the notion of combining one-class classifiers as an alternative for selecting the best classifier. In particular, we propose two new one-class classification performance measures to weigh classifiers and show that a simple ensemble that implements these measures can outperform the most popular one-class ensembles. Furthermore, we propose a new one-class ensemble scheme, TUPSO, which uses meta-learning to combine one-class classifiers. Our experiments demonstrate the superiority of TUPSO over all other tested ensembles and show that the TUPSO performance is statistically indistinguishable from that of the hypothetical best classifier."
                    },
                    {
                        "title": "Securing Your Transactions: Detecting Anomalous Patterns In XML Documents",
                        "abstract": "XML transactions are used in many information systems to store data and interact with other systems. Abnormal transactions, the result of either an on-going cyber attack or the actions of a benign user, can potentially harm the interacting systems and therefore they are regarded as a threat. In this paper we address the problem of anomaly detection and localization in XML transactions using machine learning techniques. We present a new XML anomaly detection framework, XML-AD. Within this framework, an automatic method for extracting features from XML transactions was developed as well as a practical method for transforming XML features into vectors of fixed dimensionality. With these two methods in place, the XML-AD framework makes it possible to utilize general learning algorithms for anomaly detection. Central to the functioning of the framework is a novel multi-univariate anomaly detection algorithm, ADIFA. The framework was evaluated on four XML transactions datasets, captured from real information systems, in which it achieved over 89% true positive detection rate with less than a 0.2% false positive rate."
                    },
                    {
                        "title": "ACTIDS: An Active Strategy For Detecting And Localizing Network Attacks",
                        "abstract": "In this work we investigate a new approach for detecting attacks which aim to degrade the network's Quality of Service (QoS). To this end, a new network-based intrusion detection system (NIDS) is proposed. Most contemporary NIDSs take a passive approach by solely monitoring the network's production traffic. This paper explores a complementary approach in which distributed agents actively send out periodic probes. The probes are continuously monitored to detect anomalous behavior of the network. The proposed approach takes away much of the variability of the network's production traffic that makes it so difficult to classify. This enables the NIDS to detect more subtle attacks which would not be detected using the passive approach alone. Furthermore, the active probing approach allows the NIDS to be effectively trained using only examples of the network's normal states, hence enabling an effective detection of zero-day attacks. Using realistic experiments, we show that an NIDS which also leverages the active approach is considerably more effective in detecting attacks which aim to degrade the network's QoS when compared to an NIDS which relies solely on the passive approach. Lastly, we show that the false positives rate remains very low even in the face of Byzantine faults."
                    }
                ]
            }
        }
    },
    "2405.20724": {
        "paper_data": {
            "title": "Learning on Large Graphs using Intersecting Communities",
            "url": "http://arxiv.org/abs/2405.20724v1",
            "arxiv_id": "2405.20724",
            "authors": [
                "Ben Finkelshtein",
                "\u0130smail \u0130lkan Ceylan",
                "Michael Bronstein",
                "Ron Levie"
            ],
            "abstract": "Message Passing Neural Networks (MPNNs) are a staple of graph machine learning. MPNNs iteratively update each node's representation in an input graph by aggregating messages from the node's neighbors, which necessitates a memory complexity of the order of the number of graph edges. This complexity might quickly become prohibitive for large graphs provided they are not very sparse. In this paper, we propose a novel approach to alleviate this problem by approximating the input graph as an intersecting community graph (ICG) -- a combination of intersecting cliques. The key insight is that the number of communities required to approximate a graph does not depend on the graph size. We develop a new constructive version of the Weak Graph Regularity Lemma to efficiently construct an approximating ICG for any input graph. We then devise an efficient graph learning algorithm operating directly on ICG in linear memory and time with respect to the number of nodes (rather than edges). This offers a new and fundamentally different pipeline for learning on very large non-sparse graphs, whose applicability is demonstrated empirically on node classification tasks and spatio-temporal data processing.",
            "introduction": "   1 Introduction  The vast majority of graph neural networks (GNNs) learn representations of graphs based on the message passing paradigm\u00a0[21], where every node representation is synchronously updated based on an aggregate of messages flowing from its neighbors. This mode of operation hence assumes that the set of all graph edges are loaded in the memory, which leads to a memory complexity bottleneck. Specifically, considering a graph with N\ud835\udc41Nitalic_N nodes and E\ud835\udc38Eitalic_E edges, message-passing leads to a memory complexity that is linear in the number of edges, which quickly becomes prohibitive for large graphs especially when E\u226bNmuch-greater-than\ud835\udc38\ud835\udc41E\\gg Nitalic_E \u226b italic_N. This limitation motivated a body of work with the ultimate goal of making the graph machine learning pipeline amenable to large graphs \u00a0[22, 12, 13, 7, 57].   In this work, we take a drastically different approach to alleviate this problem and design a novel graph machine learning pipeline for learning over large graphs with memory complexity linear in the number of nodes. At the heart of our approach lies a graph approximation result which informally states the following: every undirected graph with node features can be represented by a linear combination of intersecting communities (i.e., cliques) such that the number of communities required to achieve a certain approximation error is independent of the graph size for dense graphs. Intuitively, this results allows us to utilize an \u201capproximating graph\u201d, which we call intersecting community graph (ICG), for learning over large graphs. Based on this insight \u2014 radically departing from the standard message-passing paradigm \u2014 we propose a new class of neural networks acting on the ICG and on the set of nodes from the original input graph, which leads to an algorithm with memory and runtime complexity that is linear in the number nodes.   Our analysis breaks the traditional dichotomy of approaches that are appropriate either for small and dense graphs or for large and sparse graphs. To our knowledge, we present the first rigorously motivated approach allowing to efficiently handle graphs that are both large and dense. The main advantage lies in being able to process a very large non-sparse graph without excessive memory and runtime requirements. The computation of ICG requires linear time in the number of edges, but this is an offline pre-computation that needs to be performed once. After constructing the ICG, learning to solve the task on the graph \u2014 the most computation demanding aspect \u2014 is very efficient, since the architecture search and hyper-parameter optimization is linear in the number of nodes.   Context.  The focus of our study is graph-signals: undirected graphs with node features, where the underlying graph is given and fixed. This is arguably the most common learning setup for large graphs, including tasks such as semi-supervised (transductive) node classification. As we allow varying features, our approach is also suitable for spatiotemporal tasks on graphs, where a single graph is given as a fixed domain on which a dynamic process manifests as time-varying node features. Overall, our approach paves the way for carrying out many important learning tasks on very large and relatively dense graphs such as social networks that typically have",
            "references": [
                {
                    "title": "A graphon-signal analysis of graph neural networks",
                    "abstract": "We present an approach for analyzing message passing graph neural networks (MPNNs) based on an extension of graphon analysis to a so called graphon-signal analysis. A MPNN is a function that takes a graph and a signal on the graph (a graph-signal) and returns some value. Since the input space of MPNNs is non-Euclidean, i.e., graphs can be of any size and topology, properties such as generalization are less well understood for MPNNs than for Euclidean neural networks. We claim that one important missing ingredient in past work is a meaningful notion of graph-signal similarity measure, that endows the space of inputs to MPNNs with a regular structure. We present such a similarity measure, called the graphon-signal cut distance, which makes the space of all graph-signals a dense subset of a compact metric space -- the graphon-signal space. Informally, two deterministic graph-signals are close in cut distance if they ``look like'' they were sampled from the same random graph-signal model. Hence, our cut distance is a natural notion of graph-signal similarity, which allows comparing any pair of graph-signals of any size and topology. We prove that MPNNs are Lipschitz continuous functions over the graphon-signal metric space. We then give two applications of this result: 1) a generalization bound for MPNNs, and, 2) the stability of MPNNs to subsampling of graph-signals. Our results apply to any regular enough MPNN on any distribution of graph-signals, making the analysis rather universal."
                },
                {
                    "title": "A critical look at the evaluation of GNNs under heterophily: are we really making progress?",
                    "abstract": "Node classification is a classical graph machine learning task on which Graph Neural Networks (GNNs) have recently achieved strong results. However, it is often believed that standard GNNs only work well for homophilous graphs, i.e., graphs where edges tend to connect nodes of the same class. Graphs without this property are called heterophilous, and it is typically assumed that specialized methods are required to achieve strong performance on such graphs. In this work, we challenge this assumption. First, we show that the standard datasets used for evaluating heterophily-specific models have serious drawbacks, making results obtained by using them unreliable. The most significant of these drawbacks is the presence of a large number of duplicate nodes in the datasets Squirrel and Chameleon, which leads to train-test data leakage. We show that removing duplicate nodes strongly affects GNN performance on these datasets. Then, we propose a set of heterophilous graphs of varying properties that we believe can serve as a better benchmark for evaluating the performance of GNNs under heterophily. We show that standard GNNs achieve strong results on these heterophilous graphs, almost always outperforming specialized models. Our datasets and the code for reproducing our experiments are available at https://github.com/yandex-research/heterophilous-graphs"
                },
                {
                    "title": "Taming Local Effects in Graph-based Spatiotemporal Forecasting",
                    "abstract": "Spatiotemporal graph neural networks have shown to be effective in time series forecasting applications, achieving better performance than standard univariate predictors in several settings. These architectures take advantage of a graph structure and relational inductive biases to learn a single (global) inductive model to predict any number of the input time series, each associated with a graph node. Despite the gain achieved in computational and data efficiency w.r.t. fitting a set of local models, relying on a single global model can be a limitation whenever some of the time series are generated by a different spatiotemporal stochastic process. The main objective of this paper is to understand the interplay between globality and locality in graph-based spatiotemporal forecasting, while contextually proposing a methodological framework to rationalize the practice of including trainable node embeddings in such architectures. We ascribe to trainable node embeddings the role of amortizing the learning of specialized components. Moreover, embeddings allow for 1) effectively combining the advantages of shared message-passing layers with node-specific parameters and 2) efficiently transferring the learned model to new node sets. Supported by strong empirical evidence, we provide insights and guidelines for specializing graph-based models to the dynamics of each time series and show how this aspect plays a crucial role in obtaining accurate predictions."
                },
                {
                    "title": "Unveiling the Sampling Density in Non-Uniform Geometric Graphs",
                    "abstract": "A powerful framework for studying graphs is to consider them as geometric graphs: nodes are randomly sampled from an underlying metric space, and any pair of nodes is connected if their distance is less than a specified neighborhood radius. Currently, the literature mostly focuses on uniform sampling and constant neighborhood radius. However, real-world graphs are likely to be better represented by a model in which the sampling density and the neighborhood radius can both vary over the latent space. For instance, in a social network communities can be modeled as densely sampled areas, and hubs as nodes with larger neighborhood radius. In this work, we first perform a rigorous mathematical analysis of this (more general) class of models, including derivations of the resulting graph shift operators. The key insight is that graph shift operators should be corrected in order to avoid potential distortions introduced by the non-uniform sampling. Then, we develop methods to estimate the unknown sampling density in a self-supervised fashion. Finally, we present exemplary applications in which the learnt density is used to 1) correct the graph shift operator and improve performance on a variety of tasks, 2) improve pooling, and 3) extract knowledge from networks. Our experimental findings support our theory and provide strong evidence for our model."
                },
                {
                    "title": "Finding Global Homophily in Graph Neural Networks When Meeting Heterophily",
                    "abstract": "We investigate graph neural networks on graphs with heterophily. Some existing methods amplify a node's neighborhood with multi-hop neighbors to include more nodes with homophily. However, it is a significant challenge to set personalized neighborhood sizes for different nodes. Further, for other homophilous nodes excluded in the neighborhood, they are ignored for information aggregation. To address these problems, we propose two models GloGNN and GloGNN++, which generate a node's embedding by aggregating information from global nodes in the graph. In each layer, both models learn a coefficient matrix to capture the correlations between nodes, based on which neighborhood aggregation is performed. The coefficient matrix allows signed values and is derived from an optimization problem that has a closed-form solution. We further accelerate neighborhood aggregation and derive a linear time complexity. We theoretically explain the models' effectiveness by proving that both the coefficient matrix and the generated node embedding matrix have the desired grouping effect. We conduct extensive experiments to compare our models against 11 other competitors on 15 benchmark datasets in a wide range of domains, scales and graph heterophilies. Experimental results show that our methods achieve superior performance and are also very efficient."
                },
                {
                    "title": "Understanding over-squashing and bottlenecks on graphs via curvature",
                    "abstract": "Most graph neural networks (GNNs) use the message passing paradigm, in which node features are propagated on the input graph. Recent works pointed to the distortion of information flowing from distant nodes as a factor limiting the efficiency of message passing for tasks relying on long-distance interactions. This phenomenon, referred to as 'over-squashing', has been heuristically attributed to graph bottlenecks where the number of $k$-hop neighbors grows rapidly with $k$. We provide a precise description of the over-squashing phenomenon in GNNs and analyze how it arises from bottlenecks in the graph. For this purpose, we introduce a new edge-based combinatorial curvature and prove that negatively curved edges are responsible for the over-squashing issue. We also propose and experimentally test a curvature-based graph rewiring method to alleviate the over-squashing."
                },
                {
                    "title": "Large Scale Learning on Non-Homophilous Graphs: New Benchmarks and Strong Simple Methods",
                    "abstract": "Many widely used datasets for graph machine learning tasks have generally been homophilous, where nodes with similar labels connect to each other. Recently, new Graph Neural Networks (GNNs) have been developed that move beyond the homophily regime; however, their evaluation has often been conducted on small graphs with limited application domains. We collect and introduce diverse non-homophilous datasets from a variety of application areas that have up to 384x more nodes and 1398x more edges than prior datasets. We further show that existing scalable graph learning and graph minibatching techniques lead to performance degradation on these non-homophilous datasets, thus highlighting the need for further work on scalable non-homophilous methods. To address these concerns, we introduce LINKX -- a strong simple method that admits straightforward minibatch training and inference. Extensive experimental results with representative simple methods and GNNs across our proposed datasets show that LINKX achieves state-of-the-art performance for learning on non-homophilous graphs. Our codes and data are available at https://github.com/CUAI/Non-Homophily-Large-Scale."
                },
                {
                    "title": "SLAPS: Self-Supervision Improves Structure Learning for Graph Neural Networks",
                    "abstract": "Graph neural networks (GNNs) work well when the graph structure is provided. However, this structure may not always be available in real-world applications. One solution to this problem is to infer a task-specific latent structure and then apply a GNN to the inferred graph. Unfortunately, the space of possible graph structures grows super-exponentially with the number of nodes and so the task-specific supervision may be insufficient for learning both the structure and the GNN parameters. In this work, we propose the Simultaneous Learning of Adjacency and GNN Parameters with Self-supervision, or SLAPS, a method that provides more supervision for inferring a graph structure through self-supervision. A comprehensive experimental study demonstrates that SLAPS scales to large graphs with hundreds of thousands of nodes and outperforms several models that have been proposed to learn a task-specific graph structure on established benchmarks."
                },
                {
                    "title": "Adaptive Graph Convolutional Recurrent Network for Traffic Forecasting",
                    "abstract": "Modeling complex spatial and temporal correlations in the correlated time series data is indispensable for understanding the traffic dynamics and predicting the future status of an evolving traffic system. Recent works focus on designing complicated graph neural network architectures to capture shared patterns with the help of pre-defined graphs. In this paper, we argue that learning node-specific patterns is essential for traffic forecasting while the pre-defined graph is avoidable. To this end, we propose two adaptive modules for enhancing Graph Convolutional Network (GCN) with new capabilities: 1) a Node Adaptive Parameter Learning (NAPL) module to capture node-specific patterns; 2) a Data Adaptive Graph Generation (DAGG) module to infer the inter-dependencies among different traffic series automatically. We further propose an Adaptive Graph Convolutional Recurrent Network (AGCRN) to capture fine-grained spatial and temporal correlations in traffic series automatically based on the two modules and recurrent networks. Our experiments on two real-world traffic datasets show AGCRN outperforms state-of-the-art by a significant margin without pre-defined graphs about spatial connections."
                },
                {
                    "title": "Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs",
                    "abstract": "We investigate the representation power of graph neural networks in the semi-supervised node classification task under heterophily or low homophily, i.e., in networks where connected nodes may have different class labels and dissimilar features. Many popular GNNs fail to generalize to this setting, and are even outperformed by models that ignore the graph structure (e.g., multilayer perceptrons). Motivated by this limitation, we identify a set of key designs -- ego- and neighbor-embedding separation, higher-order neighborhoods, and combination of intermediate representations -- that boost learning from the graph structure under heterophily. We combine them into a graph neural network, H2GCN, which we use as the base method to empirically evaluate the effectiveness of the identified designs. Going beyond the traditional benchmarks with strong homophily, our empirical analysis shows that the identified designs increase the accuracy of GNNs by up to 40% and 27% over models without them on synthetic and real networks with heterophily, respectively, and yield competitive performance under homophily."
                },
                {
                    "title": "Adaptive Universal Generalized PageRank Graph Neural Network",
                    "abstract": "In many important graph data processing applications the acquired information includes both node features and observations of the graph topology. Graph neural networks (GNNs) are designed to exploit both sources of evidence but they do not optimally trade-off their utility and integrate them in a manner that is also universal. Here, universality refers to independence on homophily or heterophily graph assumptions. We address these issues by introducing a new Generalized PageRank (GPR) GNN architecture that adaptively learns the GPR weights so as to jointly optimize node feature and topological information extraction, regardless of the extent to which the node labels are homophilic or heterophilic. Learned GPR weights automatically adjust to the node label pattern, irrelevant on the type of initialization, and thereby guarantee excellent learning performance for label patterns that are usually hard to handle. Furthermore, they allow one to avoid feature over-smoothing, a process which renders feature information nondiscriminative, without requiring the network to be shallow. Our accompanying theoretical analysis of the GPR-GNN method is facilitated by novel synthetic benchmark datasets generated by the so-called contextual stochastic block model. We also compare the performance of our GNN architecture with that of several state-of-the-art GNNs on the problem of node-classification, using well-known benchmark homophilic and heterophilic datasets. The results demonstrate that GPR-GNN offers significant performance improvement compared to existing techniques on both synthetic and benchmark data."
                },
                {
                    "title": "Pointer Graph Networks",
                    "abstract": "Graph neural networks (GNNs) are typically applied to static graphs that are assumed to be known upfront. This static input structure is often informed purely by insight of the machine learning practitioner, and might not be optimal for the actual task the GNN is solving. In absence of reliable domain expertise, one might resort to inferring the latent graph structure, which is often difficult due to the vast search space of possible graphs. Here we introduce Pointer Graph Networks (PGNs) which augment sets or graphs with additional inferred edges for improved model expressivity. PGNs allow each node to dynamically point to another node, followed by message passing over these pointers. The sparsity of this adaptable graph structure makes learning tractable while still being sufficiently expressive to simulate complex algorithms. Critically, the pointing mechanism is directly supervised to model long-term sequences of operations on classical data structures, incorporating useful structural inductive biases from theoretical computer science. Qualitatively, we demonstrate that PGNs can learn parallelisable variants of pointer-based data structures, namely disjoint set unions and link/cut trees. PGNs generalise out-of-distribution to 5x larger test inputs on dynamic graph connectivity tasks, outperforming unrestricted GNNs and Deep Sets."
                },
                {
                    "title": "On the Bottleneck of Graph Neural Networks and its Practical Implications",
                    "abstract": "Graph neural networks (GNNs) were shown to effectively learn from highly structured data containing elements (nodes) with relationships (edges) between them. GNN variants differ in how each node in the graph absorbs the information flowing from its neighbor nodes. In this paper, we highlight an inherent problem in GNNs: the mechanism of propagating information between neighbors creates a bottleneck when every node aggregates messages from its neighbors. This bottleneck causes the over-squashing of exponentially-growing information into fixed-size vectors. As a result, the graph fails to propagate messages flowing from distant nodes and performs poorly when the prediction task depends on long-range information. We demonstrate that the bottleneck hinders popular GNNs from fitting the training data. We show that GNNs that absorb incoming edges equally, like GCN and GIN, are more susceptible to over-squashing than other GNN types. We further show that existing, extensively-tuned, GNN-based models suffer from over-squashing and that breaking the bottleneck improves state-of-the-art results without any hyperparameter tuning or additional weights."
                },
                {
                    "title": "SIGN: Scalable Inception Graph Neural Networks",
                    "abstract": "Graph representation learning has recently been applied to a broad spectrum of problems ranging from computer graphics and chemistry to high energy physics and social media. The popularity of graph neural networks has sparked interest, both in academia and in industry, in developing methods that scale to very large graphs such as Facebook or Twitter social networks. In most of these approaches, the computational cost is alleviated by a sampling strategy retaining a subset of node neighbors or subgraphs at training time. In this paper we propose a new, efficient and scalable graph deep learning architecture which sidesteps the need for graph sampling by using graph convolutional filters of different size that are amenable to efficient precomputation, allowing extremely fast training and inference. Our architecture allows using different local graph operators (e.g. motif-induced adjacency matrices or Personalized Page Rank diffusion matrix) to best suit the task at hand. We conduct extensive experimental evaluation on various open benchmarks and show that our approach is competitive with other state-of-the-art architectures, while requiring a fraction of the training and inference time."
                },
                {
                    "title": "Graphon Signal Processing",
                    "abstract": "Graphons are infinite-dimensional objects that represent the limit of convergent sequences of graphs as their number of nodes goes to infinity. This paper derives a theory of graphon signal processing centered on the notions of graphon Fourier transform and linear shift invariant graphon filters, the graphon counterparts of the graph Fourier transform and graph filters. It is shown that for convergent sequences of graphs and associated graph signals: (i) the graph Fourier transform converges to the graphon Fourier transform when the graphon signal is bandlimited; (ii) the spectral and vertex responses of graph filters converge to the spectral and vertex responses of graphon filters with the same coefficients. These theorems imply that for graphs that belong to certain families, i.e., that are part of sequences that converge to a certain graphon, graph Fourier analysis and graph filter design have well defined limits. In turn, these facts extend applicability of graph signal processing to graphs with large number of nodes \u2014 since signal processing pipelines designed for limit graphons can be applied to finite graphs \u2014 and to dynamic graphs \u2014 since we can relate the result of SP pipelines designed for different graphs from the same convergent graph sequence."
                },
                {
                    "title": "Ripple Walk Training: A Subgraph-based Training Framework for Large and Deep Graph Neural Network",
                    "abstract": "Graph neural networks (GNNs) have achieved outstanding performance in learning graph-structured data and various tasks. However, many current GNNs suffer from three common problems when facing large-size graphs or using a deeper structure: neighbors explosion, node dependence, and oversmoothing. Such problems attribute to the data structures of the graph itself or the designing of the multi-layers GNNs framework, and can lead to low training efficiency and high space complexity. To deal with these problems, in this paper, we propose a general subgraph-based training framework, namely Ripple Walk Training (RWT), for deep and large graph neural networks. RWT samples subgraphs from the full graph to constitute a mini-batch, and the full GNN is updated based on the mini-batch gradient. We analyze the high-quality sub graphs to train GNNs in a theoretical way. A novel sampling method Ripple Walk Sampler works for sampling these high-quality subgraphs to constitute the mini-batch, which considers both the randomness and connectivity of the graph-structured data. Extensive experiments on different sizes of graphs demonstrate the effectiveness and efficiency of RWT in training various GNNs (GCN & GAT). Our code is released in the https://github.com/anonymous2review/RippleWalk."
                },
                {
                    "title": "Geom-GCN: Geometric Graph Convolutional Networks",
                    "abstract": "Message-passing neural networks (MPNNs) have been successfully applied in a wide variety of applications in the real world. However, two fundamental weaknesses of MPNNs' aggregators limit their ability to represent graph-structured data: losing the structural information of nodes in neighborhoods and lacking the ability to capture long-range dependencies in disassortative graphs. Few studies have noticed the weaknesses from different perspectives. From the observations on classical neural network and network geometry, we propose a novel geometric aggregation scheme for graph neural networks to overcome the two weaknesses. The behind basic idea is the aggregation on a graph can benefit from a continuous space underlying the graph. The proposed aggregation scheme is permutation-invariant and consists of three modules, node embedding, structural neighborhood, and bi-level aggregation. We also present an implementation of the scheme in graph convolutional networks, termed Geom-GCN, to perform transductive learning on graphs. Experimental results show the proposed Geom-GCN achieved state-of-the-art performance on a wide range of open datasets of graphs."
                },
                {
                    "title": "Differentiable Graph Module (DGM) for Graph Convolutional Networks",
                    "abstract": "Graph deep learning has recently emerged as a powerful ML concept allowing to generalize successful deep neural architectures to non-euclidean structured data. Such methods have shown promising results on a broad spectrum of applications ranging from social science, biomedicine, and particle physics to computer vision, graphics, and chemistry. One of the limitations of the majority of current graph neural network architectures is that they are often restricted to the transductive setting and rely on the assumption that the underlying graph is known and fixed. Often, this assumption is not true since the graph may be noisy, or partially and even completely unknown. In such cases, it would be helpful to infer the graph directly from the data, especially in inductive settings where some nodes were not present in the graph at training time. Furthermore, learning a graph may become an end in itself, as the inferred structure may provide complementary insights next to the downstream task. In this paper, we introduce Differentiable Graph Module (DGM), a learnable function that predicts edge probabilities in the graph which are optimal for the downstream task. DGM can be combined with convolutional graph neural network layers and trained in an end-to-end fashion. We provide an extensive evaluation of applications from the domains of healthcare (disease prediction), brain imaging (age prediction), computer graphics (3D point cloud segmentation), and computer vision (zero-shot learning). We show that our model provides a significant improvement over baselines both in transductive and inductive settings and achieves state-of-the-art results."
                },
                {
                    "title": "Structural Deep Clustering Network",
                    "abstract": "Clustering is a fundamental task in data analysis. Recently, deep clustering, which derives inspiration primarily from deep learning approaches, achieves state-of-the-art performance and has attracted considerable attention. Current deep clustering methods usually boost the clustering results by means of the powerful representation ability of deep learning, e.g., autoencoder, suggesting that learning an effective representation for clustering is a crucial requirement. The strength of deep clustering methods is to extract the useful representations from the data itself, rather than the structure of data, which receives scarce attention in representation learning. Motivated by the great success of Graph Convolutional Network (GCN) in encoding the graph structure, we propose a Structural Deep Clustering Network (SDCN) to integrate the structural information into deep clustering. Specifically, we design a delivery operator to transfer the representations learned by autoencoder to the corresponding GCN layer, and a dual self-supervised mechanism to unify these two different deep neural architectures and guide the update of the whole model. In this way, the multiple structures of data, from low-order to high-order, are naturally combined with the multiple representations learned by autoencoder. Furthermore, we theoretically analyze the delivery operator, i.e., with the delivery operator, GCN improves the autoencoder-specific representation as a high-order graph regularization constraint and autoencoder helps alleviate the over-smoothing problem in GCN. Through comprehensive experiments, we demonstrate that our propose model can consistently perform better over the state-of-the-art techniques."
                },
                {
                    "title": "Graph Transformer Networks",
                    "abstract": "Graph neural networks (GNNs) have been widely used in representation learning on graphs and achieved state-of-the-art performance in tasks such as node classification and link prediction. However, most existing GNNs are designed to learn node representations on the fixed and homogeneous graphs. The limitations especially become problematic when learning representations on a misspecified graph or a heterogeneous graph that consists of various types of nodes and edges. In this paper, we propose Graph Transformer Networks (GTNs) that are capable of generating new graph structures, which involve identifying useful connections between unconnected nodes on the original graph, while learning effective node representation on the new graphs in an end-to-end fashion. Graph Transformer layer, a core layer of GTNs, learns a soft selection of edge types and composite relations for generating useful multi-hop connections so-call meta-paths. Our experiments show that GTNs learn new graph structures, based on data and tasks without domain knowledge, and yield powerful node representation via convolution on the new graphs. Without domain-specific graph preprocessing, GTNs achieved the best performance in all three benchmark node classification tasks against the state-of-the-art methods that require pre-defined meta-paths from domain knowledge."
                },
                {
                    "title": "Diffusion Improves Graph Learning",
                    "abstract": "Graph convolution is the core of most Graph Neural Networks (GNNs) and usually approximated by message passing between direct (one-hop) neighbors. In this work, we remove the restriction of using only the direct neighbors by introducing a powerful, yet spatially localized graph convolution: Graph diffusion convolution (GDC). GDC leverages generalized graph diffusion, examples of which are the heat kernel and personalized PageRank. It alleviates the problem of noisy and often arbitrarily defined edges in real graphs. We show that GDC is closely related to spectral-based models and thus combines the strengths of both spatial (message passing) and spectral methods. We demonstrate that replacing message passing with graph diffusion convolution consistently leads to significant performance improvements across a wide range of models on both supervised and unsupervised tasks and a variety of datasets. Furthermore, GDC is not limited to GNNs but can trivially be combined with any graph-based model or algorithm (e.g. spectral clustering) without requiring any changes to the latter or affecting its computational complexity. Our implementation is available online."
                },
                {
                    "title": "Multi-scale Attributed Node Embedding",
                    "abstract": "\n We present network embedding algorithms that capture information about a node from the local distribution over node attributes around it, as observed over random walks following an approach similar to Skip-gram. Observations from neighbourhoods of different sizes are either pooled (AE) or encoded distinctly in a multi-scale approach (MUSAE). Capturing attribute-neighbourhood relationships over multiple scales is useful for a range of applications, including latent feature identification across disconnected networks with similar features. We prove theoretically that matrices of node-feature pointwise mutual information are implicitly factorized by the embeddings. Experiments show that our algorithms are computationally efficient and outperform comparable models on social networks and web graphs."
                },
                {
                    "title": "GraphSAINT: Graph Sampling Based Inductive Learning Method",
                    "abstract": "Graph Convolutional Networks (GCNs) are powerful models for learning representations of attributed this http URL scale GCNs to large graphs, state-of-the-art methods use various layer sampling techniques to alleviate the \"neighbor explosion\" problem during minibatch training. Here we proposeGraphSAINT, a graph sampling based inductive learning method that improves training efficiency in a fundamentally different way. By a change of perspective, GraphSAINT constructs minibatches by sampling the training graph, rather than the nodes or edges across GCN layers. Each iteration, a complete GCN is built from the properly sampled subgraph. Thus, we ensure fixed number of well-connected nodes in all layers. We further propose normalization technique to eliminate bias, and sampling algorithms for variance reduction. Importantly, we can decouple the sampling process from the forward and backward propagation of training, and extend GraphSAINT with other graph samplers and GCN variants. Comparing with strong baselines using layer sampling, GraphSAINT demonstrates superior performance in both accuracy and training time on four large graphs."
                },
                {
                    "title": "Spectral Clustering with Graph Neural Networks for Graph Pooling",
                    "abstract": "Spectral clustering (SC) is a popular clustering technique to find strongly connected communities on a graph. SC can be used in Graph Neural Networks (GNNs) to implement pooling operations that aggregate nodes belonging to the same cluster. However, the eigendecomposition of the Laplacian is expensive and, since clustering results are graph-specific, pooling methods based on SC must perform a new optimization for each new sample. In this paper, we propose a graph clustering approach that addresses these limitations of SC. We formulate a continuous relaxation of the normalized minCUT problem and train a GNN to compute cluster assignments that minimize this objective. Our GNN-based implementation is differentiable, does not require to compute the spectral decomposition, and learns a clustering function that can be quickly evaluated on out-of-sample graphs. From the proposed clustering method, we design a graph pooling operator that overcomes some important limitations of state-of-the-art graph pooling techniques and achieves the best performance in several supervised and unsupervised tasks."
                },
                {
                    "title": "Attributed Graph Clustering: A Deep Attentional Embedding Approach",
                    "abstract": "Graph clustering is a fundamental task which discovers communities or groups in networks. Recent studies have mostly focused on developing deep learning approaches to learn a compact graph embedding, upon which classic clustering methods like k-means or spectral clustering algorithms are applied. These two-step frameworks are difficult to manipulate and usually lead to suboptimal performance, mainly because the graph embedding is not goal-directed, i.e., designed for the specific clustering task. In this paper, we propose a goal-directed deep learning approach, Deep Attentional Embedded Graph Clustering (DAEGC for short). Our method focuses on attributed graphs to sufficiently explore the two sides of information in graphs. By employing an attention network to capture the importance of the neighboring nodes to a target node, our DAEGC algorithm encodes the topological structure and node content in a graph to a compact representation, on which an inner product decoder is trained to reconstruct the graph structure. Furthermore, soft labels from the graph embedding itself are generated to supervise a self-training graph clustering process, which iteratively refines the clustering results. The self-training process is jointly learned and optimized with the graph embedding in a unified framework, to mutually benefit both components. Experimental results compared with state-of-the-art algorithms demonstrate the superiority of our method."
                },
                {
                    "title": "Graph WaveNet for Deep Spatial-Temporal Graph Modeling",
                    "abstract": "Spatial-temporal graph modeling is an important task to analyze the spatial relations and temporal trends of components in a system. Existing approaches mostly capture the spatial dependency on a fixed graph structure, assuming that the underlying relation between entities is pre-determined. However, the explicit graph structure (relation) does not necessarily reflect the true dependency and genuine relation may be missing due to the incomplete connections in the data. Furthermore, existing methods are ineffective to capture the temporal trends as the RNNs or CNNs employed in these methods cannot capture long-range temporal sequences. To overcome these limitations, we propose in this paper a novel graph neural network architecture, {Graph WaveNet}, for spatial-temporal graph modeling. By developing a novel adaptive dependency matrix and learn it through node embedding, our model can precisely capture the hidden spatial dependency in the data. With a stacked dilated 1D convolution component whose receptive field grows exponentially as the number of layers increases, Graph WaveNet is able to handle very long sequences. These two components are integrated seamlessly in a unified framework and the whole framework is learned in an end-to-end manner. Experimental results on two public traffic network datasets, METR-LA and PEMS-BAY, demonstrate the superior performance of our algorithm."
                },
                {
                    "title": "A Flexible Generative Framework for Graph-based Semi-supervised Learning",
                    "abstract": "We consider a family of problems that are concerned about making predictions for the majority of unlabeled, graph-structured data samples based on a small proportion of labeled samples. Relational information among the data samples, often encoded in the graph/network structure, is shown to be helpful for these semi-supervised learning tasks. However, conventional graph-based regularization methods and recent graph neural networks do not fully leverage the interrelations between the features, the graph, and the labels. In this work, we propose a flexible generative framework for graph-based semi-supervised learning, which approaches the joint distribution of the node features, labels, and the graph structure. Borrowing insights from random graph models in network science literature, this joint distribution can be instantiated using various distribution families. For the inference of missing labels, we exploit recent advances of scalable variational inference techniques to approximate the Bayesian posterior. We conduct thorough experiments on benchmark datasets for graph-based semi-supervised learning. Results show that the proposed methods outperform the state-of-the-art models in most settings."
                },
                {
                    "title": "Cluster-GCN: An Efficient Algorithm for Training Deep and Large Graph Convolutional Networks",
                    "abstract": "Graph convolutional network (GCN) has been successfully applied to many graph-based applications; however, training a large-scale GCN remains challenging. Current SGD-based algorithms suffer from either a high computational cost that exponentially grows with number of GCN layers, or a large space requirement for keeping the entire graph and the embedding of each node in memory. In this paper, we propose Cluster-GCN, a novel GCN algorithm that is suitable for SGD-based training by exploiting the graph clustering structure. Cluster-GCN works as the following: at each step, it samples a block of nodes that associate with a dense subgraph identified by a graph clustering algorithm, and restricts the neighborhood search within this subgraph. This simple but effective strategy leads to significantly improved memory and computational efficiency while being able to achieve comparable test accuracy with previous algorithms. To test the scalability of our algorithm, we create a new Amazon2M data with 2 million nodes and 61 million edges which is more than 5 times larger than the previous largest publicly available dataset (Reddit). For training a 3-layer GCN on this data, Cluster-GCN is faster than the previous state-of-the-art VR-GCN (1523 seconds vs 1961 seconds) and using much less memory (2.2GB vs 11.2GB). Furthermore, for training 4 layer GCN on this data, our algorithm can finish in around 36 minutes while all the existing GCN training algorithms fail to train due to the out-of-memory issue. Furthermore, Cluster-GCN allows us to train much deeper GCN without much time and memory overhead, which leads to improved prediction accuracy---using a 5-layer Cluster-GCN, we achieve state-of-the-art test F1 score 99.36 on the PPI dataset, while the previous best result was 98.71 by~\\citezhang2018gaan."
                },
                {
                    "title": "Adversarial Examples on Graph Data: Deep Insights into Attack and Defense",
                    "abstract": "Graph deep learning models, such as graph convolutional networks (GCN) achieve remarkable performance for tasks on graph data. Similar to other types of deep models, graph deep learning models often suffer from adversarial attacks. However, compared with non-graph data, the discrete features, graph connections and different definitions of imperceptible perturbations bring unique challenges and opportunities for the adversarial attacks and defenses for graph data. In this paper, we propose both attack and defense techniques. For attack, we show that the discreteness problem could easily be resolved by introducing integrated gradients which could accurately reflect the effect of perturbing certain features or edges while still benefiting from the parallel computations. For defense, we observe that the adversarially manipulated graph for the targeted attack differs from normal graphs statistically. Based on this observation, we propose a defense approach which inspects the graph and recovers the potential adversarial perturbations. Our experiments on a number of datasets show the effectiveness of the proposed methods."
                },
                {
                    "title": "Hierarchical Graph Representation Learning with Differentiable Pooling",
                    "abstract": "Recently, graph neural networks (GNNs) have revolutionized the field of graph representation learning through effectively learned node embeddings, and achieved state-of-the-art results in tasks such as node classification and link prediction. However, current GNN methods are inherently flat and do not learn hierarchical representations of graphs---a limitation that is especially problematic for the task of graph classification, where the goal is to predict the label associated with an entire graph. Here we propose DiffPool, a differentiable graph pooling module that can generate hierarchical representations of graphs and can be combined with various graph neural network architectures in an end-to-end fashion. DiffPool learns a differentiable soft cluster assignment for nodes at each layer of a deep GNN, mapping nodes to a set of clusters, which then form the coarsened input for the next GNN layer. Our experimental results show that combining existing GNN methods with DiffPool yields an average improvement of 5-10% accuracy on graph classification benchmarks, compared to all existing pooling approaches, achieving a new state-of-the-art on four out of five benchmark datasets."
                },
                {
                    "title": "On Estimation and Inference in Latent Structure Random Graphs",
                    "abstract": "We define a latent structure model (LSM) random graph as a random dot product graph (RDPG) in which the latent position distribution incorporates both probabilistic and geometric constraints, delineated by a family of underlying distributions on some fixed Euclidean space, and a structural support submanifold from which the latent positions for the graph are drawn. For a one-dimensional latent structure model with known structural support, we show how spectral estimates of the latent positions of an RDPG can be used for efficient estimation of the paramaters of the LSM. We describe how to estimate or learn the structural support in cases where it is unknown, with an illustrative focus on graphs with latent positions along the Hardy-Weinberg curve. Finally, we use the latent structure model formulation to test bilateral homology in the Drosophila connectome."
                },
                {
                    "title": "Neural Relational Inference for Interacting Systems",
                    "abstract": "Interacting systems are prevalent in nature, from dynamical systems in physics to complex societal dynamics. The interplay of components can give rise to complex behavior, which can often be explained using a simple model of the system's constituent parts. In this work, we introduce the neural relational inference (NRI) model: an unsupervised model that learns to infer interactions while simultaneously learning the dynamics purely from observational data. Our model takes the form of a variational auto-encoder, in which the latent code represents the underlying interaction graph and the reconstruction is based on graph neural networks. In experiments on simulated physical systems, we show that our NRI model can accurately recover ground-truth interactions in an unsupervised manner. We further demonstrate that we can find an interpretable structure and predict complex dynamics in real motion capture and sports tracking data."
                },
                {
                    "title": "FastGCN: Fast Learning with Graph Convolutional Networks via Importance Sampling",
                    "abstract": "The graph convolutional networks (GCN) recently proposed by Kipf and Welling are an effective graph model for semi-supervised learning. This model, however, was originally designed to be learned with the presence of both training and test data. Moreover, the recursive neighborhood expansion across layers poses time and memory challenges for training with large, dense graphs. To relax the requirement of simultaneous availability of test data, we interpret graph convolutions as integral transforms of embedding functions under probability measures. Such an interpretation allows for the use of Monte Carlo approaches to consistently estimate the integrals, which in turn leads to a batched training scheme as we propose in this work---FastGCN. Enhanced with importance sampling, FastGCN not only is efficient for training but also generalizes well for inference. We show a comprehensive set of experiments to demonstrate its effectiveness compared with GCN and related models. In particular, training is orders of magnitude more efficient while predictions remain comparably accurate."
                },
                {
                    "title": "Dynamic Graph CNN for Learning on Point Clouds",
                    "abstract": "Point clouds provide a flexible geometric representation suitable for countless applications in computer graphics; they also comprise the raw output of most 3D data acquisition devices. While hand-designed features on point clouds have long been proposed in graphics and vision, however, the recent overwhelming success of convolutional neural networks (CNNs) for image analysis suggests the value of adapting insight from CNN to the point cloud world. Point clouds inherently lack topological information, so designing a model to recover topology can enrich the representation power of point clouds. To this end, we propose a new neural network module dubbed EdgeConv suitable for CNN-based high-level tasks on point clouds, including classification and segmentation. EdgeConv acts on graphs dynamically computed in each layer of the network. It is differentiable and can be plugged into existing architectures. Compared to existing modules operating in extrinsic space or treating each point independently, EdgeConv has several appealing properties: It incorporates local neighborhood information; it can be stacked applied to learn global shape properties; and in multi-layer systems affinity in feature space captures semantic characteristics over potentially long distances in the original embedding. We show the performance of our model on standard benchmarks, including ModelNet40, ShapeNetPart, and S3DIS."
                },
                {
                    "title": "Graph Attention Networks",
                    "abstract": "We present graph attention networks (GATs), novel neural network architectures that operate on graph-structured data, leveraging masked self-attentional layers to address the shortcomings of prior methods based on graph convolutions or their approximations. By stacking layers in which nodes are able to attend over their neighborhoods' features, we enable (implicitly) specifying different weights to different nodes in a neighborhood, without requiring any kind of costly matrix operation (such as inversion) or depending on knowing the graph structure upfront. In this way, we address several key challenges of spectral-based graph neural networks simultaneously, and make our model readily applicable to inductive as well as transductive problems. Our GAT models have achieved or matched state-of-the-art results across four established transductive and inductive graph benchmarks: the Cora, Citeseer and Pubmed citation network datasets, as well as a protein-protein interaction dataset (wherein test graphs remain unseen during training)."
                },
                {
                    "title": "Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting",
                    "abstract": "Spatiotemporal forecasting has various applications in neuroscience, climate and transportation domain. Traffic forecasting is one canonical example of such learning task. The task is challenging due to (1) complex spatial dependency on road networks, (2) non-linear temporal dynamics with changing road conditions and (3) inherent difficulty of long-term forecasting. To address these challenges, we propose to model the traffic flow as a diffusion process on a directed graph and introduce Diffusion Convolutional Recurrent Neural Network (DCRNN), a deep learning framework for traffic forecasting that incorporates both spatial and temporal dependency in the traffic flow. Specifically, DCRNN captures the spatial dependency using bidirectional random walks on the graph, and the temporal dependency using the encoder-decoder architecture with scheduled sampling. We evaluate the framework on two real-world large scale road network traffic datasets and observe consistent improvement of 12% - 15% over state-of-the-art baselines."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Inductive Representation Learning on Large Graphs",
                    "abstract": "Low-dimensional embeddings of nodes in large graphs have proved extremely useful in a variety of prediction tasks, from content recommendation to identifying protein functions. However, most existing approaches require that all nodes in the graph are present during training of the embeddings; these previous approaches are inherently transductive and do not naturally generalize to unseen nodes. Here we present GraphSAGE, a general, inductive framework that leverages node feature information (e.g., text attributes) to efficiently generate node embeddings for previously unseen data. Instead of training individual embeddings for each node, we learn a function that generates embeddings by sampling and aggregating features from a node's local neighborhood. Our algorithm outperforms strong baselines on three inductive node-classification benchmarks: we classify the category of unseen nodes in evolving information graphs based on citation and Reddit post data, and we show that our algorithm generalizes to completely unseen graphs using a multi-graph dataset of protein-protein interactions."
                },
                {
                    "title": "Neural Message Passing for Quantum Chemistry",
                    "abstract": "Supervised learning on molecules has incredible potential to be useful in chemistry, drug discovery, and materials science. Luckily, several promising and closely related neural network models invariant to molecular symmetries have already been described in the literature. These models learn a message passing algorithm and aggregation procedure to compute a function of their entire input graph. At this point, the next step is to find a particularly effective variant of this general approach and apply it to chemical prediction benchmarks until we either solve them or reach the limits of the approach. In this paper, we reformulate existing models into a single common framework we call Message Passing Neural Networks (MPNNs) and explore additional novel variations within this framework. Using MPNNs we demonstrate state of the art results on an important molecular property prediction benchmark; these results are strong enough that we believe future work should focus on datasets with larger molecules or more accurate ground truth labels."
                },
                {
                    "title": "Variational Graph Auto-Encoders",
                    "abstract": "We introduce the variational graph auto-encoder (VGAE), a framework for unsupervised learning on graph-structured data based on the variational auto-encoder (VAE). This model makes use of latent variables and is capable of learning interpretable latent representations for undirected graphs. We demonstrate this model using a graph convolutional network (GCN) encoder and a simple inner product decoder. Our model achieves competitive results on a link prediction task in citation networks. In contrast to most existing models for unsupervised learning on graph-structured data and link prediction, our model can naturally incorporate node features, which significantly improves predictive performance on a number of benchmark datasets."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Overlapping community detection at scale: a nonnegative matrix factorization approach",
                    "abstract": "Network communities represent basic structures for understanding the organization of real-world networks. A community (also referred to as a module or a cluster) is typically thought of as a group of nodes with more connections amongst its members than between its members and the remainder of the network. Communities in networks also overlap as nodes belong to multiple clusters at once. Due to the difficulties in evaluating the detected communities and the lack of scalable algorithms, the task of overlapping community detection in large networks largely remains an open problem. In this paper we present BIGCLAM (Cluster Affiliation Model for Big Networks), an overlapping community detection method that scales to large networks of millions of nodes and edges. We build on a novel observation that overlaps between communities are densely connected. This is in sharp contrast with present community detection methods which implicitly assume that overlaps between communities are sparsely connected and thus cannot properly extract overlapping communities in networks. In this paper, we develop a model-based community detection algorithm that can detect densely overlapping, hierarchically nested as well as non-overlapping communities in massive networks. We evaluate our algorithm on 6 large social, collaboration and information networks with ground-truth community information. Experiments show state of the art performance both in terms of the quality of detected communities as well as in speed and scalability of our algorithm."
                },
                {
                    "title": "Structure and Overlaps of Communities in Networks",
                    "abstract": "One of the main organizing principles in real-world social, information and technological networks is that of network communities, where sets of nodes organize into densely linked clusters. Even though detection of such communities is of great interest, understanding the structure communities in large networks remains relatively limited. Due to unavailability of labeled ground-truth data it is practically impossible to evaluate and compare different models and notions of communities on a large scale. \nIn this paper we identify 6 large social, collaboration, and information networks where nodes explicitly state their community memberships. We define ground-truth communities by using these explicit memberships. We then empirically study how such ground-truth communities emerge in networks and how they overlap. We observe some surprising phenomena. First, ground-truth communities contain high-degree hub nodes that reside in community overlaps and link to most of the members of the community. Second, the overlaps of communities are more densely connected than the non-overlapping parts of communities, in contrast to the conventional wisdom that community overlaps are more sparsely connected than the communities themselves. \nExisting models of network communities do not capture dense community overlaps. We present the Community-Affiliation Graph Model (AGM), a conceptual model of network community structure, which reliably captures the overall structure of networks as well as the overlapping nature of network communities."
                },
                {
                    "title": "Numerical Methods for Large Eigenvalue Problems",
                    "abstract": "Preface to the Classics Edition Preface 1. Background in matrix theory and linear algebra 2. Sparse matrices 3. Perturbation theory and error analysis 4. The tools of spectral approximation 5. Subspace iteration 6. Krylov subspace methods 7. Filtering and restarting techniques 8. Preconditioning techniques 9. Non-standard eigenvalue problems 10. Origins of matrix eigenvalue problems References Index."
                },
                {
                    "title": "Mixed Membership Stochastic Blockmodels",
                    "abstract": "Observations consisting of measurements on relationships for pairs of objects arise in many settings, such as protein interaction and gene regulatory networks, collections of author-recipient email, and social networks. Analyzing such data with probabilisic models can be delicate because the simple exchangeability assumptions underlying many boilerplate models no longer hold. In this paper, we describe a latent variable model of such data called the mixed membership stochastic blockmodel. This model extends blockmodels for relational data to ones which capture mixed membership latent relational structure, thus providing an object-specific low-dimensional representation. We develop a general variational inference algorithm for fast approximate posterior inference. We explore applications to social and protein interaction networks."
                },
                {
                    "title": "Approximating the cut-norm via Grothendieck's inequality",
                    "abstract": "The <i>cut-norm</i> ||A||<inf>C</inf> of a real matrix A=(a<inf>ij</inf>)<inf>i\u2208 R,j\u2208S</inf> is the maximum, over all I \u2282 R, J \u2282 S of the quantity | \u03a3<inf>i \u2208 I, j \u2208 J</inf> a<inf>ij</inf>|. This concept plays a major role in the design of efficient approximation algorithms for dense graph and matrix problems. Here we show that the problem of approximating the cut-norm of a given real matrix is MAX SNP hard, and provide an efficient approximation algorithm. This algorithm finds, for a given matrix A=(a<inf>ij</inf>)<inf>i \u2208 R, j \u2208 S</inf>, two subsets I \u2282 R and J \u2282 S, such that | \u03a3<inf>i \u2208 I, j \u2208 J</inf> a<inf>ij</inf>| \u2265 \u03c1 ||A||<inf>C</inf>, where \u03c1 > 0 is an absolute constant satisfying $\u03c1 > 0. 56. The algorithm combines semidefinite programming with a rounding technique based on Grothendieck's Inequality. We present three known proofs of Grothendieck's inequality, with the necessary modifications which emphasize their algorithmic aspects. These proofs contain rounding techniques which go beyond the random hyperplane rounding of Goemans and Williamson [12], allowing us to transfer various algorithms for dense graph and matrix problems to the sparse case."
                },
                {
                    "title": "Computing the norm \u2225A\u2225\u221e,1 is NP-hard",
                    "abstract": "It is proved that computing the subordinate matrix norm \u2225A\u2225\u221e1 is NP-hard, Even more, existence of a polynomial-time algorithm for computing this norm with relative accuracy less than 1/(4n2 ), where n is matrix size, implies P = NP."
                },
                {
                    "title": "The algorithmic aspects of the regularity lemma",
                    "abstract": "The regularity lemma of Szemeredi (1978) is a result that asserts that every graph can be partitioned in a certain regular way. This result has numerous applications, but its known proof is not algorithmic. The authors first demonstrate the computational difficulty of finding a regular partition; they show that deciding if a given partition of an input graph satisfies the properties guaranteed by the lemma is co-NP-complete. However, they also prove that despite this difficulty the lemma can be made constructive; they show how to obtain, for any input graph, a partition with the properties guaranteed by the lemma, efficiently. The desired partition, for an n-vertex graph, can be found in time O(M(n)), where M(n)=O(n/sup 2.376/) is the time needed to multiply two n by n matrices with 0,1-entries over the integers. The algorithm can be parallelized and implemented in NC/sup 1/.<<ETX>>"
                },
                {
                    "title": "Regular Partitions of Graphs",
                    "abstract": "Abstract : A crucial lemma in recent work of the author (showing that k-term arithmetic progression-free sets of integers must have density zero) stated (approximately) that any large bipartite graph can be decomposed into relatively few 'nearly regular' bipartite subgraphs. In this note the author generalizes this result to arbitrary graphs, at the same time strengthening and simplifying the original bipartite result."
                },
                {
                    "title": "Large Networks and Graph Limits",
                    "abstract": "The book Large Networks and Graph Limits, xiv + 475 pp., published in late 2012, comprises five parts, the first an illuminating introduction and the last a tantalizing taste of how the scope of the theory developed in its pages might be extended to other combinatorial structures than graphs. The three central parts treat in depth the topics of graph algebras, limits for sequences of dense graphs (this constitutes the most substantial part, occupying nearly half the book) and limits for sequences of bounded degree graphs. Primarily the book is aimed at graduate students and research mathematicians interested in graph theory and its application to networks (for example, the internet and networks in social science, biology, statistical physics and engineering). There are 23 chapters and an appendix, the latter conveniently giving necessary background from areas of mathematics outside mainstream graph theory. A bibliography collects together the extensive research in this area up to 2012, and a subject, author and notation index facilitate navigation of the book. The author maintains a webpage for corrections and supplementary material. Indeed, via the author\u2019s homepage the reader can freely access the many papers he has written with collaborators on the topic of graph homomorphisms and graph limits. The book synthesizes much of the material in these papers, with some revision in"
                }
            ],
            "categories": [
                "cs.LG",
                "cs.SI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we efficiently learn representations of large and dense graphs while overcoming the memory complexity bottleneck associated with traditional message-passing paradigms in graph neural networks?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant limitation in current graph neural network methodologies, enabling the analysis of large and dense graphs that are prevalent in real-world applications such as social networks and spatiotemporal data. By developing a method that operates with memory complexity linear in the number of nodes, this research could lead to advancements in various domains, including semi-supervised learning and dynamic graph analysis. The implications of this work could inspire future research to explore new architectures and algorithms that leverage the proposed intersecting community graph (ICG) approach, ultimately enhancing our understanding of graph-based learning.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of large graphs, where traditional message-passing methods require storing and processing all edges, leading to prohibitive memory usage. Naive approaches may fail because they do not account for the dense nature of the graphs, resulting in inefficient computations and inability to scale. Additionally, the theoretical understanding of graph representations and the need for effective approximation methods pose significant obstacles. Overcoming these challenges requires innovative strategies to represent and process graph data without succumbing to memory limitations.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either small and dense graphs or large and sparse graphs, often neglecting the unique challenges posed by large dense graphs. Existing solutions have been limited by their reliance on the message-passing paradigm, which inherently leads to high memory complexity. Barriers such as a lack of understanding of effective graph approximations and the absence of methodologies that can efficiently handle both size and density have prevented progress. Our approach differs by introducing the concept of the intersecting community graph (ICG), which allows for efficient learning without the memory constraints of traditional methods.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves constructing an intersecting community graph (ICG) that approximates the original graph while maintaining linear memory complexity in relation to the number of nodes. We will utilize a dataset of large and dense graphs, focusing on tasks such as semi-supervised node classification. The performance will be evaluated using metrics such as classification"
            }
        },
        "author_data": {
            "40a0a80a-1e16-4a94-9310-d70ecc970a3a": {
                "pk": "40a0a80a-1e16-4a94-9310-d70ecc970a3a",
                "name": "Ben Finkelshtein",
                "collaborators": [
                    "Chaim Baskin",
                    "\u0130smail \u0130lkan Ceylan",
                    "Haggai Maron",
                    "Nadav Dym",
                    "Xingyue Huang",
                    "Michael Bronstein",
                    "Evgenii Zheltonozhskii",
                    "Uri Alon",
                    "Itay Eilat",
                    "Nir Rosenfeld"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Adversarial Machine Learning",
                    "Strategic Classification",
                    "3D Learning"
                ],
                "publications": [
                    {
                        "title": "A Simple and Universal Rotation Equivariant Point-cloud Network",
                        "abstract": "Equivariance to permutations and rigid motions is an important inductive bias for various 3D learning problems. Recently it has been shown that the equivariant Tensor Field Network architecture is universal -- it can approximate any equivariant function. In this paper we suggest a much simpler architecture, prove that it enjoys the same universality guarantees and evaluate its performance on Modelnet40. The code to reproduce our experiments is available at \\url{https://github.com/simpleinvariance/UniversalNetwork}"
                    },
                    {
                        "title": "Cooperative Graph Neural Networks",
                        "abstract": "Graph neural networks are popular architectures for graph machine learning, based on iterative computation of node representations of an input graph through a series of invariant transformations. A large class of graph neural networks follow a standard message-passing paradigm: at every layer, each node state is updated based on an aggregate of messages from its neighborhood. In this work, we propose a novel framework for training graph neural networks, where every node is viewed as a player that can choose to either 'listen', 'broadcast', 'listen and broadcast', or to 'isolate'. The standard message propagation scheme can then be viewed as a special case of this framework where every node 'listens and broadcasts' to all neighbors. Our approach offers a more flexible and dynamic message-passing paradigm, where each node can determine its own strategy based on their state, effectively exploring the graph topology while learning. We provide a theoretical analysis of the new message-passing scheme which is further supported by an extensive empirical analysis on a synthetic dataset and on real-world datasets."
                    },
                    {
                        "title": "Single-Node Attacks for Fooling Graph Neural Networks",
                        "abstract": "Graph neural networks (GNNs) have shown broad applicability in a variety of domains. These domains, e.g., social networks and product recommendations, are fertile ground for malicious users and behavior. In this paper, we show that GNNs are vulnerable to the extremely limited (and thus quite realistic) scenarios of a single-node adversarial attack, where the perturbed node cannot be chosen by the attacker. That is, an attacker can force the GNN to classify any target node to a chosen label, by only slightly perturbing the features or the neighbor list of another single arbitrary node in the graph, even when not being able to select that specific attacker node. When the adversary is allowed to select the attacker node, these attacks are even more effective. We demonstrate empirically that our attack is effective across various common GNN types (e.g., GCN, GraphSAGE, GAT, GIN) and robustly optimized GNNs (e.g., Robust GCN, SM GCN, GAL, LAT-GCN), outperforming previous attacks across different real-world datasets both in a targeted and non-targeted attacks. Our code is available at https://github.com/benfinkelshtein/SINGLE ."
                    },
                    {
                        "title": "Strategic Classification with Graph Neural Networks",
                        "abstract": "Strategic classification studies learning in settings where users can modify their features to obtain favorable predictions. Most current works focus on simple classifiers that trigger independent user responses. Here we examine the implications of learning with more elaborate models that break the independence assumption. Motivated by the idea that applications of strategic classification are often social in nature, we focus on \\emph{graph neural networks}, which make use of social relations between users to improve predictions. Using a graph for learning introduces inter-user dependencies in prediction; our key point is that strategic users can exploit these to promote their goals. As we show through analysis and simulation, this can work either against the system -- or for it. Based on this, we propose a differentiable framework for strategically-robust learning of graph-based classifiers. Experiments on several real networked datasets demonstrate the utility of our approach."
                    },
                    {
                        "title": "Almost Surely Asymptotically Constant Graph Neural Networks",
                        "abstract": "We present a new angle on the expressive power of graph neural networks (GNNs) by studying how the predictions of a GNN probabilistic classifier evolve as we apply it on larger graphs drawn from some random graph model. We show that the output converges to a constant function, which upper-bounds what these classifiers can uniformly express. This strong convergence phenomenon applies to a very wide class of GNNs, including state of the art models, with aggregates including mean and the attention-based mechanism of graph transformers. Our results apply to a broad class of random graph models, including sparse and dense variants of the Erd\\H{o}s-R\\'enyi model, the stochastic block model, and the Barab\\'asi-Albert model. We empirically validate these findings, observing that the convergence phenomenon appears not only on random graphs but also on some real-world graphs."
                    }
                ]
            },
            "cbe088b8-c857-4680-9be4-7ca55780f22a": {
                "pk": "cbe088b8-c857-4680-9be4-7ca55780f22a",
                "name": "\u0130smail \u0130lkan Ceylan",
                "collaborators": [
                    "Rafael Pe\u00f1aloza",
                    "Diego Calvanese",
                    "Marco Montali",
                    "Ario Santoso"
                ],
                "domain": [
                    "Ontology",
                    "Uncertainty",
                    "Dynamic Systems",
                    "Knowledge Representation"
                ],
                "publications": [
                    {
                        "title": "Dynamic Bayesian Ontology Languages",
                        "abstract": "Many formalisms combining ontology languages with uncertainty, usually in the form of probabilities, have been studied over the years. Most of these formalisms, however, assume that the probabilistic structure of the knowledge remains static over time. We present a general approach for extending ontology languages to handle time-evolving uncertainty represented by a dynamic Bayesian network. We show how reasoning in the original language and dynamic Bayesian inferences can be exploited for effective reasoning in our framework."
                    },
                    {
                        "title": "Adding Context to Knowledge and Action Bases",
                        "abstract": "Knowledge and Action Bases (KABs) have been recently proposed as a formal framework to capture the dynamics of systems which manipulate Description Logic (DL) Knowledge Bases (KBs) through action execution. In this work, we enrich the KAB setting with contextual information, making use of different context dimensions. On the one hand, context is determined by the environment using context-changing actions that make use of the current state of the KB and the current context. On the other hand, it affects the set of TBox assertions that are relevant at each time point, and that have to be considered when processing queries posed over the KAB. Here we extend to our enriched setting the results on verification of rich temporal properties expressed in mu-calculus, which had been established for standard KABs. Specifically, we show that under a run-boundedness condition, verification stays decidable."
                    }
                ]
            },
            "96c29ae4-eda5-4eb0-8bce-2c918b8d8873": {
                "pk": "96c29ae4-eda5-4eb0-8bce-2c918b8d8873",
                "name": "Michael Bronstein",
                "collaborators": [
                    "Or Litany",
                    "Alex Bronstein",
                    "Ameesh Makadia",
                    "Emanuele Rodol\u00e0",
                    "Daniel Cremers"
                ],
                "domain": [
                    "3D Reconstruction",
                    "Shape Completion",
                    "Non-Rigid Deformation",
                    "Graph Convolutional Networks"
                ],
                "publications": [
                    {
                        "title": "Deformable Shape Completion with Graph Convolutional Autoencoders",
                        "abstract": "The availability of affordable and portable depth sensors has made scanning objects and people simpler than ever. However, dealing with occlusions and missing parts is still a significant challenge. The problem of reconstructing a (possibly non-rigidly moving) 3D object from a single or multiple partial scans has received increasing attention in recent years. In this work, we propose a novel learning-based method for the completion of partial shapes. Unlike the majority of existing approaches, our method focuses on objects that can undergo non-rigid deformations. The core of our method is a variational autoencoder with graph convolutional operations that learns a latent space for complete realistic shapes. At inference, we optimize to find the representation in this latent space that best fits the generated shape to the known partial input. The completed shape exhibits a realistic appearance on the unknown part. We show promising results towards the completion of synthetic and real scans of human body and face meshes exhibiting different styles of articulation and partiality."
                    },
                    {
                        "title": "Non-Rigid Puzzles",
                        "abstract": "Shape correspondence is a fundamental problem in computer graphics and vision, with applications in various problems including animation, texture mapping, robotic vision, medical imaging, archaeology and many more. In settings where the shapes are allowed to undergo non-rigid deformations and only partial views are available, the problem becomes very challenging. To this end, we present a non-rigid multi-part shape matching algorithm. We assume to be given a reference shape and its multiple parts undergoing a non-rigid deformation. Each of these query parts can be additionally contaminated by clutter, may overlap with other parts, and there might be missing parts or redundant ones. Our method simultaneously solves for the segmentation of the reference model, and for a dense correspondence to (subsets of) the parts. Experimental results on synthetic as well as real scans demonstrate the effectiveness of our method in dealing with this challenging matching scenario."
                    }
                ]
            },
            "d8b1b311-f17b-4c42-a868-b3633f855786": {
                "pk": "d8b1b311-f17b-4c42-a868-b3633f855786",
                "name": "Ron Levie",
                "collaborators": [
                    "Haim Avron",
                    "Nir Sochen",
                    "Gitta Kutyniok",
                    "Daniel Zilberg",
                    "Efrat Krimer Avraham",
                    "Elvin Isufi"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Signal Processing",
                    "Wavelet Transform",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "A graphon-signal analysis of graph neural networks",
                        "abstract": "We present an approach for analyzing message passing graph neural networks (MPNNs) based on an extension of graphon analysis to a so called graphon-signal analysis. A MPNN is a function that takes a graph and a signal on the graph (a graph-signal) and returns some value. Since the input space of MPNNs is non-Euclidean, i.e., graphs can be of any size and topology, properties such as generalization are less well understood for MPNNs than for Euclidean neural networks. We claim that one important missing ingredient in past work is a meaningful notion of graph-signal similarity measure, that endows the space of inputs to MPNNs with a regular structure. We present such a similarity measure, called the graphon-signal cut distance, which makes the space of all graph-signals a dense subset of a compact metric space -- the graphon-signal space. Informally, two deterministic graph-signals are close in cut distance if they ``look like'' they were sampled from the same random graph-signal model. Hence, our cut distance is a natural notion of graph-signal similarity, which allows comparing any pair of graph-signals of any size and topology. We prove that MPNNs are Lipschitz continuous functions over the graphon-signal metric space. We then give two applications of this result: 1) a generalization bound for MPNNs, and, 2) the stability of MPNNs to subsampling of graph-signals. Our results apply to any regular enough MPNN on any distribution of graph-signals, making the analysis rather universal."
                    },
                    {
                        "title": "Randomized Signal Processing with Continuous Frames",
                        "abstract": "This paper focuses on signal processing tasks in which the signal is transformed from the signal space to a higher dimensional coefficient space (also called phase space) using a continuous frame, processed in the coefficient space, and synthesized to an output signal. We show how to approximate such methods, termed phase space signal processing methods, using a Monte Carlo method. As opposed to standard discretizations of continuous frames, based on sampling discrete frames from the continuous system, the proposed Monte Carlo method is directly a quadrature approximation of the continuous frame. We show that the Monte Carlo method allows working with highly redundant continuous frames, since the number of samples required for a certain accuracy is proportional to the dimension of the signal space, and not to the dimension of the phase space. Moreover, even though the continuous frame is highly redundant, the Monte Carlo samples are spread uniformly, and hence represent the coefficient space more faithfully than standard frame discretizations."
                    },
                    {
                        "title": "Uncertainty principles and optimally sparse wavelet transforms",
                        "abstract": "In this paper we introduce a new localization framework for wavelet transforms, such as the 1D wavelet transform and the Shearlet transform. Our goal is to design nonadaptive window functions that promote sparsity in some sense. For that, we introduce a framework for analyzing localization aspects of window functions. Our localization theory diverges from the conventional theory in two ways. First, we distinguish between the group generators, and the operators that measure localization (called observables). Second, we define the uncertainty of a signal transform based on a window as a whole, instead of defining the uncertainty of an individual window. We show that the uncertainty of a window function, in the signal space, is closely related to the localization of the reproducing kernel of the wavelet transform, in phase space. As a result, we show that using uncertainty minimizing window functions, results in representations which are optimally sparse in some sense."
                    },
                    {
                        "title": "Randomized Continuous Frames in Time-Frequency Analysis",
                        "abstract": "Recently, a Monte Carlo approach was proposed for processing highly redundant continuous frames. In this paper we present and analyze applications of this new theory. The computational complexity of the Monte Carlo method relies on the continuous frame being so called linear volume discretizable (LVD). The LVD property means that the number of samples in the coefficient space required by the Monte Carlo method is proportional to the resolution of the discrete signal. We show in this paper that the continuous wavelet transform (CWT) and the localizing time-frequency transform (LTFT) are LVD. The LTFT is a time-frequency representation based on a 3D time-frequency space with a richer class of time-frequency atoms than classical time-frequency transforms like the short time Fourier transform (STFT) and the CWT. Our analysis proves that performing signal processing with the LTFT has the same asymptotic complexity as signal processing with the STFT and CWT (based on FFT), even though the coefficient space of the LTFT is higher dimensional."
                    },
                    {
                        "title": "A Wavelet Plancherel Theory with Application to Multipliers and Sparse Approximations",
                        "abstract": "We introduce an extension of continuous wavelet theory that enables an efficient implementation of multiplicative operators in the coefficient space. In the new theory, the signal space is embedded in a larger abstract signal space -- the so called window-signal space. There is a canonical extension of the wavelet transform to an isometric isomorphism between the window-signal space and the coefficient space. Hence, the new framework is called a wavelet-Plancherel theory, and the extended wavelet transform is called the wavelet-Plancherel transform. Since the wavelet-Plancherel transform is an isometric isomorphism, any operation in the coefficient space can be pulled-back to an operation in the window-signal space. It is then possible to improve the computational complexity of methods that involve a multiplicative operator in the coefficient space, by performing all computations directly in the window-signal space. As one example application, we show how continuous wavelet multipliers (also called Calder\\'{o}n-Toeplitz Operators), with polynomial symbols, can be implemented with linear complexity in the resolution of the 1D signal. As another example, we develop a framework for efficiently computing greedy sparse approximations to signals based on elements of continuous wavelet systems."
                    },
                    {
                        "title": "PieClam: A Universal Graph Autoencoder Based on Overlapping Inclusive and Exclusive Communities",
                        "abstract": "We propose PieClam (Prior Inclusive Exclusive Cluster Affiliation Model): a probabilistic graph model for representing any graph as overlapping generalized communities. Our method can be interpreted as a graph autoencoder: nodes are embedded into a code space by an algorithm that maximizes the log-likelihood of the decoded graph, given the input graph. PieClam is a community affiliation model that extends well-known methods like BigClam in two main manners. First, instead of the decoder being defined via pairwise interactions between the nodes in the code space, we also incorporate a learned prior on the distribution of nodes in the code space, turning our method into a graph generative model. Secondly, we generalize the notion of communities by allowing not only sets of nodes with strong connectivity, which we call inclusive communities, but also sets of nodes with strong disconnection, which we call exclusive communities. To model both types of communities, we propose a new type of decoder based the Lorentz inner product, which we prove to be much more expressive than standard decoders based on standard inner products or norm distances. By introducing a new graph similarity measure, that we call the log cut distance, we show that PieClam is a universal autoencoder, able to uniformly approximately reconstruct any graph. Our method is shown to obtain competitive performance in graph anomaly detection benchmarks."
                    },
                    {
                        "title": "Wavelet Design with Optimally Localized Ambiguity Function: a Variational Approach",
                        "abstract": "In this paper, we design mother wavelets for the 1D continuous wavelet transform with some optimality properties. An optimal mother wavelet here is one that has an ambiguity function with minimal spread in the continuous coefficient space (also called phase space). Since the ambiguity function is the reproducing kernel of the coefficient space, optimal windows lead to phase space representations which are \"optimally sharp.\" Namely, the wavelet coefficients have minimal correlations with each other. Such a construction also promotes sparsity in phase space. The spread of the ambiguity function is modeled as the sum of variances along the axes in phase space. In order to optimize the mother wavelet directly as a 1D signal, we pull-back the variances, defined on the 2D phase space, to the so called window-signal space. This is done using the recently developed wavelet-Plancharel theory. The approach allows formulating the optimization problem of the 2D ambiguity function as a minimization problem of the 1D mother wavelet. The resulting 1D formulation is more efficient and does not involve complicated constraints on the 2D ambiguity function. We optimize the mother wavelet using gradient descent, which yields a locally optimal mother wavelet."
                    },
                    {
                        "title": "On the Transferability of Spectral Graph Filters",
                        "abstract": "This paper focuses on spectral filters on graphs, namely filters defined as elementwise multiplication in the frequency domain of a graph. In many graph signal processing settings, it is important to transfer a filter from one graph to another. One example is in graph convolutional neural networks (ConvNets), where the dataset consists of signals defined on many different graphs, and the learned filters should generalize to signals on new graphs, not present in the training set. A necessary condition for transferability (the ability to transfer filters) is stability. Namely, given a graph filter, if we add a small perturbation to the graph, then the filter on the perturbed graph is a small perturbation of the original filter. It is a common misconception that spectral filters are not stable, and this paper aims at debunking this mistake. We introduce a space of filters, called the Cayley smoothness space, that contains the filters of state-of-the-art spectral filtering methods, and whose filters can approximate any generic spectral filter. For filters in this space, the perturbation in the filter is bounded by a constant times the perturbation in the graph, and filters in the Cayley smoothness space are thus termed linearly stable. By combining stability with the known property of equivariance, we prove that graph spectral filters are transferable."
                    },
                    {
                        "title": "Quasi Monte Carlo Time-Frequency Analysis",
                        "abstract": "We study signal processing tasks in which the signal is mapped via some generalized time-frequency transform to a higher dimensional time-frequency space, processed there, and synthesized to an output signal. We show how to approximate such methods using a quasi-Monte Carlo (QMC) approach. We consider cases where the time-frequency representation is redundant, having feature axes in addition to the time and frequency axes. The proposed QMC method allows sampling both efficiently and evenly such redundant time-frequency representations. Indeed, 1) the number of samples required for a certain accuracy is log-linear in the resolution of the signal space, and depends only weakly on the dimension of the redundant time-frequency space, and 2) the quasi-random samples have low discrepancy, so they are spread evenly in the redundant time-frequency space. One example of such redundant representation is the localizing time-frequency transform (LTFT), where the time-frequency plane is enhanced by a third axis. This higher dimensional time-frequency space improves the quality of some time-frequency signal processing tasks, like the phase vocoder (an audio signal processing effect). Since the computational complexity of the QMC is log-linear in the resolution of the signal space, this higher dimensional time-frequency space does not degrade the computation complexity of the proposed QMC method. The proposed QMC method is more efficient than standard Monte Carlo methods, since the deterministic QMC sample points are optimally spread in the time-frequency space, while random samples are not."
                    }
                ]
            }
        }
    },
    "2403.17329": {
        "paper_data": {
            "title": "Deep Support Vectors",
            "url": "http://arxiv.org/abs/2403.17329v2",
            "arxiv_id": "2403.17329",
            "authors": [
                "Junhoo Lee",
                "Hyunho Lee",
                "Kyomin Hwang",
                "Nojun Kwak"
            ],
            "abstract": "Deep learning has achieved tremendous success. \\nj{However,} unlike SVMs, which provide direct decision criteria and can be trained with a small dataset, it still has significant weaknesses due to its requirement for massive datasets during training and the black-box characteristics on decision criteria. \\nj{This paper addresses} these issues by identifying support vectors in deep learning models. To this end, we propose the DeepKKT condition, an adaptation of the traditional Karush-Kuhn-Tucker (KKT) condition for deep learning models, and confirm that generated Deep Support Vectors (DSVs) using this condition exhibit properties similar to traditional support vectors. This allows us to apply our method to few-shot dataset distillation problems and alleviate the black-box characteristics of deep learning models. Additionally, we demonstrate that the DeepKKT condition can transform conventional classification models into generative models with high fidelity, particularly as latent \\jh{generative} models using class labels as latent variables. We validate the effectiveness of DSVs \\nj{using common datasets (ImageNet, CIFAR10 \\nj{and} CIFAR100) on the general architectures (ResNet and ConvNet)}, proving their practical applicability. (See Fig.~\\ref{fig:generated})",
            "introduction": "   1 Introduction  Although deep learning has gained enormous success, it requires huge amounts of data for training, and its black-box characteristics regarding decision criteria result in a lack of reliability. For example, CLIP [24] needs 400 million image pairs for training and Stable Diffusion XL (SDXL) [25] requires 5 billion images. This implies only a small number of groups can train foundation models from scratch. Also, the black-box nature makes it hard to anticipate the model\u2019s performance in different environments. For example, suppose we are to classify pictures of deer and most training deer images contain antlers. For the test images taken in winter, the performance will be worse as deer shed their antlers in winter. As modern deep learning models do not provide any decision criterion i.e., black box, we cannot determine whether the domain of the model has shifted, or if the model is biased in advance, thus cannot anticipate the performance drop in this case.   Interestingly, these problems do not occur in previous state-of-the-art, support vector machines (SVMs), which require substantially less data, enabling almost anyone to train a model from scratch. Also, as it encodes the decision boundary explicitly, SVM can reconstruct the support vectors from the training dataset using the KKT condition. Since it is a white box, one can anticipate the test\u2019s performance in advance. In the deer classification example, if the model\u2019s support vectors of deer have prominent antlers, using that SVM is not appropriate for photos taken in winter.    In this paper, we tackle the natural limitations of deep learning \u2013 the need for large data and black-box characteristics \u2013 by extracting SVM features in deep learning models. In doing so, we introduce the DeepKKT condition for deep models, which corresponds to the KKT condition in traditional SVMs. By either selecting deep support vectors (DSVs) from training data or generating them from already trained deep learning models, we show DSVs can play a similar role to conventional support vectors. Like support vectors can reconstruct SVM, we can reconstruct the deep models from scratch only with DSVs. Also, we show that DSVs encode the decision criterion visually, providing a global explanation for the trained deep model. Expanding beyond conventional support vectors, DSVs suggest that a trained deep classification model can also function as a latent generative model by utilizing logits as latent variables and applying DeepKKT conditions.   To this end, we generalize the KKT condition and define the DeepKKT condition considering that the data handled by a deep model is high-dimensional and multi-class. We demonstrate that the selected data points (selected DSVs) among the training data satisfying the DeepKKT condition are closer to the decision boundary than other training data, as evidenced by comparing entropy values. Also, we show that the calculated Lagrangian multiplier can reveal the level of uncertainty of the model for the sample in question. Additionally, we demonstrate that the DSVs outperform existing algorithms in the few-shot dataset distillation setting, where only a portion of the training set is used, indicating that DSVs exhibit characteristics similar to SVMs. Moreover, we confirm that modifying existing images using information obtained from",
            "references": [
                {
                    "title": "Generator Born from Classifier",
                    "abstract": "In this paper, we make a bold attempt toward an ambitious task: given a pre-trained classifier, we aim to reconstruct an image generator, without relying on any data samples. From a black-box perspective, this challenge seems intractable, since it inevitably involves identifying the inverse function for a classifier, which is, by nature, an information extraction process. As such, we resort to leveraging the knowledge encapsulated within the parameters of the neural network. Grounded on the theory of Maximum-Margin Bias of gradient descent, we propose a novel learning paradigm, in which the generator is trained to ensure that the convergence conditions of the network parameters are satisfied over the generated distribution of the samples. Empirical validation from various image generation tasks substantiates the efficacy of our strategy."
                },
                {
                    "title": "Good regularity creates large learning rate implicit biases: edge of stability, balancing, and catapult",
                    "abstract": "Large learning rates, when applied to gradient descent for nonconvex optimization, yield various implicit biases including the edge of stability (Cohen et al., 2021), balancing (Wang et al., 2022), and catapult (Lewkowycz et al., 2020). These phenomena cannot be well explained by classical optimization theory. Though significant theoretical progress has been made in understanding these implicit biases, it remains unclear for which objective functions would they be more likely. This paper provides an initial step in answering this question and also shows that these implicit biases are in fact various tips of the same iceberg. To establish these results, we develop a global convergence theory under large learning rates, for a family of nonconvex functions without globally Lipschitz continuous gradient, which was typically assumed in existing convergence analysis. Specifically, these phenomena are more likely to occur when the optimization objective function has good regularity. This regularity, together with gradient descent using a large learning rate that favors flatter regions, results in these nontrivial dynamical behaviors. Another corollary is the first non-asymptotic convergence rate bound for large-learning-rate gradient descent optimization of nonconvex functions. Although our theory only applies to specific functions so far, the possibility of extrapolating it to neural networks is also experimentally validated, for which different choices of loss, activation functions, and other techniques such as batch normalization can all affect regularity significantly and lead to very different training dynamics."
                },
                {
                    "title": "Dataset Distillation via Factorization",
                    "abstract": "In this paper, we study \\xw{dataset distillation (DD)}, from a novel perspective and introduce a \\emph{dataset factorization} approach, termed \\emph{HaBa}, which is a plug-and-play strategy portable to any existing DD baseline. Unlike conventional DD approaches that aim to produce distilled and representative samples, \\emph{HaBa} explores decomposing a dataset into two components: data \\emph{Ha}llucination networks and \\emph{Ba}ses, where the latter is fed into the former to reconstruct image samples. The flexible combinations between bases and hallucination networks, therefore, equip the distilled data with exponential informativeness gain, which largely increase the representation capability of distilled datasets. To furthermore increase the data efficiency of compression results, we further introduce a pair of adversarial contrastive constraints on the resultant hallucination networks and bases, which increase the diversity of generated images and inject more discriminant information into the factorization. Extensive comparisons and experiments demonstrate that our method can yield significant improvement on downstream classification tasks compared with previous state of the arts, while reducing the total number of compressed parameters by up to 65\\%. Moreover, distilled datasets by our approach also achieve \\textasciitilde10\\% higher accuracy than baseline methods in cross-architecture generalization. Our code is available \\href{https://github.com/Huage001/DatasetFactorization}{here}."
                },
                {
                    "title": "Reconstructing Training Data from Trained Neural Networks",
                    "abstract": "Understanding to what extent neural networks memorize training data is an intriguing question with practical and theoretical implications. In this paper we show that in some cases a significant fraction of the training data can in fact be reconstructed from the parameters of a trained neural network classifier. We propose a novel reconstruction scheme that stems from recent theoretical results about the implicit bias in training neural networks with gradient-based methods. To the best of our knowledge, our results are the first to show that reconstructing a large portion of the actual training samples from a trained neural network classifier is generally possible. This has negative implications on privacy, as it can be used as an attack for revealing sensitive training data. We demonstrate our method for binary MLP classifiers on a few standard computer vision datasets."
                },
                {
                    "title": "Dataset Distillation by Matching Training Trajectories",
                    "abstract": "Dataset distillation is the task of synthesizing a small dataset such that a model trained on the synthetic set will match the test accuracy of the model trained on the full dataset. In this paper, we propose a new formulation that optimizes our distilled data to guide networks to a similar state as those trained on real data across many training steps. Given a network, we train it for several iterations on our distilled data and optimize the distilled data with respect to the distance between the synthetically trained parameters and the parameters trained on real data. To efficiently obtain the initial and target network parameters for large-scale datasets, we pre-compute and store training trajectories of expert networks trained on the real dataset. Our method handily outperforms existing methods and also allows us to distill higher-resolution visual data."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "On the Equivalence between Neural Network and Support Vector Machine",
                    "abstract": "Recent research shows that the dynamics of an infinitely wide neural network (NN) trained by gradient descent can be characterized by Neural Tangent Kernel (NTK) \\citep{jacot2018neural}. Under the squared loss, the infinite-width NN trained by gradient descent with an infinitely small learning rate is equivalent to kernel regression with NTK \\citep{arora2019exact}. However, the equivalence is only known for ridge regression currently \\citep{arora2019harnessing}, while the equivalence between NN and other kernel machines (KMs), e.g. support vector machine (SVM), remains unknown. Therefore, in this work, we propose to establish the equivalence between NN and SVM, and specifically, the infinitely wide NN trained by soft margin loss and the standard soft margin SVM with NTK trained by subgradient descent. Our main theoretical results include establishing the equivalences between NNs and a broad family of $\\ell_2$ regularized KMs with finite-width bounds, which cannot be handled by prior work, and showing that every finite-width NN trained by such regularized loss functions is approximately a KM. Furthermore, we demonstrate our theory can enable three practical applications, including (i) \\textit{non-vacuous} generalization bound of NN via the corresponding KM; (ii) \\textit{non-trivial} robustness certificate for the infinite-width NN (while existing robustness verification methods would provide vacuous bounds); (iii) intrinsically more robust infinite-width NNs than those from previous kernel regression. Our code for the experiments is available at \\url{https://github.com/leslie-CH/equiv-nn-svm}."
                },
                {
                    "title": "Dataset Condensation with Distribution Matching",
                    "abstract": "Computational cost of training state-of-the-art deep models in many learning problems is rapidly increasing due to more sophisticated models and larger datasets. A recent promising direction for reducing training cost is dataset condensation that aims to replace the original large training set with a significantly smaller learned synthetic set while preserving the original information. While training deep models on the small set of condensed images can be extremely fast, their synthesis remains computationally expensive due to the complex bi-level optimization and secondorder derivative computation. In this work, we propose a simple yet effective method that synthesizes condensed images by matching feature distributions of the synthetic and original training images in many sampled embedding spaces. Our method significantly reduces the synthesis cost while achieving comparable or better performance. Thanks to its efficiency, we apply our method to more realistic and larger datasets with sophisticated neural architectures and obtain a significant performance boost1. We also show promising practical benefits of our method in continual learning and neural architecture search."
                },
                {
                    "title": "See through Gradients: Image Batch Recovery via GradInversion",
                    "abstract": "Training deep neural networks requires gradient estimation from data batches to update parameters. Gradients per parameter are averaged over a set of data and this has been presumed to be safe for privacy-preserving training in joint, collaborative, and federated learning applications. Prior work only showed the possibility of recovering input data given gradients under very restrictive conditions \u2013 a single input point, or a network with no non-linearities, or a small 32 \u00d7 32 px input batch. Therefore, averaging gradients over larger batches was thought to be safe. In this work, we introduce GradInversion, using which input images from a larger batch (8 \u2013 48 images) can also be recovered for large networks such as ResNets (50 layers), on complex datasets such as ImageNet (1000 classes, 224 \u00d7 224 px). We formulate an optimization task that converts random noise into natural images, matching gradients while regularizing image fidelity. We also propose an algorithm for target class label recovery given gradients. We further propose a group consistency regularization framework, where multiple agents starting from different random seeds work together to find an enhanced reconstruction of the original data batch. We show that gradients encode a surprisingly large amount of information, such that all the individual images can be recovered with high fidelity via GradInversion, even for complex datasets, deep networks, and large batch sizes."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Understanding deep learning (still) requires rethinking generalization",
                    "abstract": "Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small gap between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family or to the regularization techniques used during training. Through extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. We interpret our experimental findings by comparison with traditional models. We supplement this republication with a new section at the end summarizing recent progresses in the field since the original version of this paper."
                },
                {
                    "title": "Dataset Condensation with Differentiable Siamese Augmentation",
                    "abstract": "In many machine learning problems, large-scale datasets have become the de-facto standard to train state-of-the-art deep networks at the price of heavy computation load. In this paper, we focus on condensing large training sets into significantly smaller synthetic sets which can be used to train deep neural networks from scratch with minimum drop in performance. Inspired from the recent training set synthesis methods, we propose Differentiable Siamese Augmentation that enables effective use of data augmentation to synthesize more informative synthetic images and thus achieves better performance when training networks with augmentations. Experiments on multiple image classification benchmarks demonstrate that the proposed method obtains substantial gains over the state-of-the-art, 7% improvements on CIFAR10 and CIFAR100 datasets. We show with only less than 1% data that our method achieves 99.6%, 94.9%, 88.5%, 71.5% relative performance on MNIST, FashionMNIST, SVHN, CIFAR10 respectively. We also explore the use of our method in continual learning and neural architecture search, and show promising results."
                },
                {
                    "title": "Sharpness-Aware Minimization for Efficiently Improving Generalization",
                    "abstract": "In today's heavily overparameterized models, the value of the training loss provides few guarantees on model generalization ability. Indeed, optimizing only the training loss value, as is commonly done, can easily lead to suboptimal model quality. Motivated by the connection between geometry of the loss landscape and generalization---including a generalization bound that we prove here---we introduce a novel, effective procedure for instead simultaneously minimizing loss value and loss sharpness. In particular, our procedure, Sharpness-Aware Minimization (SAM), seeks parameters that lie in neighborhoods having uniformly low loss; this formulation results in a min-max optimization problem on which gradient descent can be performed efficiently. We present empirical results showing that SAM improves model generalization across a variety of benchmark datasets (e.g., CIFAR-{10, 100}, ImageNet, finetuning tasks) and models, yielding novel state-of-the-art performance for several. Additionally, we find that SAM natively provides robustness to label noise on par with that provided by state-of-the-art procedures that specifically target learning with noisy labels."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Dataset Condensation with Gradient Matching",
                    "abstract": "Efficient training of deep neural networks is an increasingly important problem in the era of sophisticated architectures and large-scale datasets. This paper proposes a training set synthesis technique, called Dataset Condensation, that learns to produce a small set of informative samples for training deep neural networks from scratch in a small fraction of the required computational cost on the original data while achieving comparable results. We rigorously evaluate its performance in several computer vision benchmarks and show that it significantly outperforms the state-of-the-art methods. Finally we show promising applications of our method in continual learning and domain adaptation."
                },
                {
                    "title": "Totally Deep Support Vector Machines",
                    "abstract": "Support vector machines (SVMs) have been successful in solving many computer vision tasks including image and video category recognition especially for small and mid-scale training problems. The principle of these non-parametric models is to learn hyperplanes that separate data belonging to different classes while maximizing their margins. However, SVMs constrain the learned hyperplanes to lie in the span of support vectors, fixed/taken from training data, and this reduces their representational power and may lead to limited generalization performances. In this paper, we relax this constraint and allow the support vectors to be learned (instead of being fixed/taken from training data) in order to better fit a given classification task. Our approach, referred to as deep total variation support vector machines, is parametric and relies on a novel deep architecture that learns not only the SVM and the kernel parameters but also the support vectors, resulting into highly effective classifiers. We also show (under a particular setting of the activation functions in this deep architecture) that a large class of kernels and their combinations can be learned. Experiments conducted on the challenging task of skeleton-based action recognition show the outperformance of our deep total variation SVMs w.r.t different baselines as well as the related work."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "Gradient Descent Maximizes the Margin of Homogeneous Neural Networks",
                    "abstract": "In this paper, we study the implicit regularization of the gradient descent algorithm in homogeneous neural networks, including fully-connected and convolutional neural networks with ReLU or LeakyReLU activations. In particular, we study the gradient descent or gradient flow (i.e., gradient descent with infinitesimal step size) optimizing the logistic loss or cross-entropy loss of any homogeneous model (possibly non-smooth), and show that if the training loss decreases below a certain threshold, then we can define a smoothed version of the normalized margin which increases over time. We also formulate a natural constrained optimization problem related to margin maximization, and prove that both the normalized margin and its smoothed version converge to the objective value at a KKT point of the optimization problem. Our results generalize the previous results for logistic regression with one-layer or multi-layer linear networks, and provide more quantitative convergence results with weaker assumptions than previous results for homogeneous smooth neural networks. We conduct several experiments to justify our theoretical finding on MNIST and CIFAR-10 datasets. Finally, as margin is closely related to robustness, we discuss potential benefits of training longer for improving the robustness of the model."
                },
                {
                    "title": "SVM-based Deep Stacking Networks",
                    "abstract": "The deep network model, with the majority built on neural networks, has been proved to be a powerful framework to represent complex data for high performance machine learning. In recent years, more and more studies turn to nonneural network approaches to build diverse deep structures, and the Deep Stacking Network (DSN) model is one of such approaches that uses stacked easy-to-learn blocks to build a parameter-training-parallelizable deep network. In this paper, we propose a novel SVM-based Deep Stacking Network (SVM-DSN), which uses the DSN architecture to organize linear SVM classifiers for deep learning. A BP-like layer tuning scheme is also proposed to ensure holistic and local optimizations of stacked SVMs simultaneously. Some good math properties of SVM, such as the convex optimization, is introduced into the DSN framework by our model. From a global view, SVM-DSN can iteratively extract data representations layer by layer as a deep neural network but with parallelizability, and from a local view, each stacked SVM can converge to its optimal solution and obtain the support vectors, which compared with neural networks could lead to interesting improvements in anti-saturation and interpretability. Experimental results on both image and text data sets demonstrate the excellent performances of SVM-DSN compared with some competitive benchmark models."
                },
                {
                    "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                    "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                },
                {
                    "title": "Dataset Distillation",
                    "abstract": "Model distillation aims to distill the knowledge of a complex model into a simpler one. In this paper, we consider an alternative formulation called dataset distillation: we keep the model fixed and instead attempt to distill the knowledge from a large training dataset into a small one. The idea is to synthesize a small number of data points that do not need to come from the correct data distribution, but will, when given to the learning algorithm as training data, approximate the model trained on the original data. For example, we show that it is possible to compress 60,000 MNIST training images into just 10 synthetic distilled images (one per class) and achieve close to original performance with only a few gradient descent steps, given a fixed network initialization. We evaluate our method in various initialization settings and with different learning objectives. Experiments on multiple datasets show the advantage of our approach compared to alternative methods."
                },
                {
                    "title": "An Empirical Study of Example Forgetting during Deep Neural Network Learning",
                    "abstract": "Inspired by the phenomenon of catastrophic forgetting, we investigate the learning dynamics of neural networks as they train on single classification tasks. Our goal is to understand whether a related phenomenon occurs when data does not undergo a clear distributional shift. We define a \u201cforgetting event\u201d to have occurred when an individual training example transitions from being classified correctly to incorrectly over the course of learning. Across several benchmark data sets, we find that: (i) certain examples are forgotten with high frequency, and some not at all; (ii) a data set\u2019s (un)forgettable examples generalize across neural architectures; and (iii) based on forgetting dynamics, a significant fraction of examples can be omitted from the training data set while still maintaining state-of-the-art generalization performance."
                },
                {
                    "title": "StarGAN: Unified Generative Adversarial Networks for Multi-domain Image-to-Image Translation",
                    "abstract": "Recent studies have shown remarkable success in image-to-image translation for two domains. However, existing approaches have limited scalability and robustness in handling more than two domains, since different models should be built independently for every pair of image domains. To address this limitation, we propose StarGAN, a novel and scalable approach that can perform image-to-image translations for multiple domains using only a single model. Such a unified model architecture of StarGAN allows simultaneous training of multiple datasets with different domains within a single network. This leads to StarGAN's superior quality of translated images compared to existing models as well as the novel capability of flexibly translating an input image to any desired target domain. We empirically demonstrate the effectiveness of our approach on a facial attribute transfer and a facial expression synthesis tasks."
                },
                {
                    "title": "The Implicit Bias of Gradient Descent on Separable Data",
                    "abstract": "We examine gradient descent on unregularized logistic regression problems, with homogeneous linear predictors on linearly separable datasets. We show the predictor converges to the direction of the max-margin (hard margin SVM) solution. The result also generalizes to other monotone decreasing loss functions with an infimum at infinity, to multi-class problems, and to training a weight layer in a deep network in a certain restricted setting. Furthermore, we show this convergence is very slow, and only logarithmic in the convergence of the loss itself. This can help explain the benefit of continuing to optimize the logistic or cross-entropy loss even after the training error is zero and the training loss is extremely small, and, as we show, even if the validation loss increases. Our methodology can also aid in understanding implicit regularization n more complex models and with other optimization methods."
                },
                {
                    "title": "Universal Adversarial Perturbations",
                    "abstract": "Given a state-of-the-art deep neural network classifier, we show the existence of a universal (image-agnostic) and very small perturbation vector that causes natural images to be misclassified with high probability. We propose a systematic algorithm for computing universal perturbations, and show that state-of-the-art deep neural networks are highly vulnerable to such perturbations, albeit being quasi-imperceptible to the human eye. We further empirically analyze these universal perturbations and show, in particular, that they generalize very well across neural networks. The surprising existence of universal perturbations reveals important geometric correlations among the high-dimensional decision boundary of classifiers. It further outlines potential security breaches with the existence of single directions in the input space that adversaries can possibly exploit to break a classifier on most natural images."
                },
                {
                    "title": "Federated Optimization: Distributed Machine Learning for On-Device Intelligence",
                    "abstract": "We introduce a new and increasingly relevant setting for distributed optimization in machine learning, where the data defining the optimization are unevenly distributed over an extremely large number of nodes. The goal is to train a high-quality centralized model. We refer to this setting as Federated Optimization. In this setting, communication efficiency is of the utmost importance and minimizing the number of rounds of communication is the principal goal. \nA motivating example arises when we keep the training data locally on users' mobile devices instead of logging it to a data center for training. In federated optimziation, the devices are used as compute nodes performing computation on their local data in order to update a global model. We suppose that we have extremely large number of devices in the network --- as many as the number of users of a given service, each of which has only a tiny fraction of the total data available. In particular, we expect the number of data points available locally to be much smaller than the number of devices. Additionally, since different users generate data with different patterns, it is reasonable to assume that no device has a representative sample of the overall distribution. \nWe show that existing algorithms are not suitable for this setting, and propose a new algorithm which shows encouraging experimental results for sparse convex problems. This work also sets a path for future research needed in the context of \\federated optimization."
                },
                {
                    "title": "Communication-Efficient Learning of Deep Networks from Decentralized Data",
                    "abstract": "Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data center and training there using conventional approaches. We advocate an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. \nWe present a practical method for the federated learning of deep networks based on iterative model averaging, and conduct an extensive empirical evaluation, considering five different model architectures and four datasets. These experiments demonstrate the approach is robust to the unbalanced and non-IID data distributions that are a defining characteristic of this setting. Communication costs are the principal constraint, and we show a reduction in required communication rounds by 10-100x as compared to synchronized stochastic gradient descent."
                },
                {
                    "title": "Federated Learning of Deep Networks using Model Averaging",
                    "abstract": "Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data-center and training there using conventional approaches. We advocate an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. \nWe present a practical method for the federated learning of deep networks that proves robust to the unbalanced and non-IID data distributions that naturally arise. This method allows high-quality models to be trained in relatively few rounds of communication, the principal constraint for federated learning. The key insight is that despite the non-convex loss functions we optimize, parameter averaging over updates from multiple clients produces surprisingly good results, for example decreasing the communication needed to train an LSTM language model by two orders of magnitude."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Deep Learning Face Attributes in the Wild",
                    "abstract": "Predicting face attributes in the wild is challenging due to complex face variations. We propose a novel deep learning framework for attribute prediction in the wild. It cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently. LNet is pre-trained by massive general object categories for face localization, while ANet is pre-trained by massive face identities for attribute prediction. This framework not only outperforms the state-of-the-art with a large margin, but also reveals valuable facts on learning face representation. (1) It shows how the performances of face localization (LNet) and attribute prediction (ANet) can be improved by different pre-training strategies. (2) It reveals that although the filters of LNet are fine-tuned only with image-level attribute tags, their response maps over entire images have strong indication of face locations. This fact enables training LNet for face localization with only image-level annotations, but without face bounding boxes or landmarks, which are required by all attribute recognition works. (3) It also demonstrates that the high-level hidden neurons of ANet automatically discover semantic concepts after pre-training with massive face identities, and such concepts are significantly enriched after fine-tuning with attribute tags. Each attribute can be well explained with a sparse linear combination of these concepts."
                },
                {
                    "title": "Understanding deep image representations by inverting them",
                    "abstract": "Image representations, from SIFT and Bag of Visual Words to Convolutional Neural Networks (CNNs), are a crucial component of almost any image understanding system. Nevertheless, our understanding of them remains limited. In this paper we conduct a direct analysis of the visual information contained in representations by asking the following question: given an encoding of an image, to which extent is it possible to reconstruct the image itself? To answer this question we contribute a general framework to invert representations. We show that this method can invert representations such as HOG more accurately than recent alternatives while being applicable to CNNs too. We then use this technique to study the inverse of recent state-of-the-art CNN image representations for the first time. Among our findings, we show that several layers in CNNs retain photographically accurate information about the image, with different degrees of geometric and photometric invariance."
                },
                {
                    "title": "Deep Learning using Linear Support Vector Machines",
                    "abstract": "Recently, fully-connected and convolutional neural networks have been trained to achieve state-of-the-art performance on a wide variety of tasks such as speech recognition, image classification, natural language processing, and bioinformatics. For classification tasks, most of these \"deep learning\" models employ the softmax activation function for prediction and minimize cross-entropy loss. In this paper, we demonstrate a small but consistent advantage of replacing the softmax layer with a linear support vector machine. Learning minimizes a margin-based loss instead of the cross-entropy loss. While there have been various combinations of neural nets and SVMs in prior art, our results using L2-SVMs show that by simply replacing softmax with linear SVMs gives significant gains on popular deep learning datasets MNIST, CIFAR-10, and the ICML 2013 Representation Learning Workshop's face expression recognition challenge."
                },
                {
                    "title": "The MNIST Database of Handwritten Digit Images for Machine Learning Research [Best of the Web]",
                    "abstract": "In this issue, \u201cBest of the Web\u201d presents the modified National Institute of Standards and Technology (MNIST) resources, consisting of a collection of handwritten digit images used extensively in optical character recognition and machine learning research."
                },
                {
                    "title": "Super-Samples from Kernel Herding",
                    "abstract": "We extend the herding algorithm to continuous spaces by using the kernel trick. The resulting \"kernel herding\" algorithm is an infinite memory deterministic process that learns to approximate a PDF with a collection of samples. We show that kernel herding decreases the error of expectations of functions in the Hilbert space at a rate O(1/T) which is much faster than the usual O(1/pT) for iid random samples. We illustrate kernel herding by approximating Bayesian predictive distributions."
                },
                {
                    "title": "Support vector machines",
                    "abstract": "Support vector machines (SVMs) are a family of machine learning methods, originally introduced for the problem of classification and later generalized to various other situations. They are based on principles of statistical learning theory and convex optimization, and are currently used in various domains of application, including bioinformatics, text categorization, and computer vision. Copyright \u00a9 2009 John Wiley & Sons, Inc."
                },
                {
                    "title": "Reading Digits in Natural Images with Unsupervised Feature Learning",
                    "abstract": "Detecting and reading text from natural images is a hard computer vision task that is central to a variety of emerging applications. Related problems like document character recognition have been widely studied by computer vision and machine learning researchers and are virtually solved for practical applications like reading handwritten digits. Reliably recognizing characters in more complex scenes like photographs, however, is far more difficult: the best existing methods lag well behind human performance on the same tasks. In this paper we attack the problem of recognizing digits in a real application using unsupervised feature learning methods: reading house numbers from street level photos. To this end, we introduce a new benchmark dataset for research use containing over 600,000 labeled digits cropped from Street View images. We then demonstrate the difficulty of recognizing these digits when the problem is approached with hand-designed features. Finally, we employ variants of two recently proposed unsupervised feature learning methods and find that they are convincingly superior on our benchmarks."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we address the limitations of deep learning models, specifically their reliance on large datasets and their black-box nature, by integrating features from support vector machines (SVMs)?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it could democratize access to effective machine learning models, allowing more researchers and practitioners to train models without the need for vast amounts of data. By bridging the gap between deep learning and SVMs, this research could lead to more interpretable models, enhancing trust and reliability in AI systems. Furthermore, it could pave the way for advancements in few-shot learning and model generalization, ultimately leading to practical applications in various fields such as healthcare, finance, and autonomous systems.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the inherent complexities of deep learning architectures, which are designed to handle high-dimensional data and multi-class classification. Naive approaches may fail because they do not account for the intricate relationships between data points in deep models, nor do they effectively extract meaningful features akin to support vectors. Additionally, the black-box nature of deep learning makes it difficult to interpret decision boundaries and assess model performance in varying conditions, complicating the integration of SVM features.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has largely focused on either deep learning or traditional machine learning methods like SVMs, often treating them as separate paradigms. Limitations in existing solutions include a lack of methodologies to extract and utilize features from deep models in a way that mirrors the interpretability of SVMs. Barriers such as the complexity of high-dimensional data and the absence of a generalized framework for integrating the KKT condition into deep learning have hindered progress. Our approach differs by introducing the DeepKKT condition, which allows for the extraction of deep support vectors (DSVs) and their application in model reconstruction and performance assessment.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves defining the DeepKKT condition for deep learning models, which parallels the KKT condition in SVMs. We will utilize datasets that allow for the selection of deep support vectors (DSVs) from both training data and pre-trained models. The evaluation metrics will include entropy values to assess the proximity of DSVs to the decision boundary and the calculated Lagrangian multipliers to gauge model uncertainty."
            }
        },
        "author_data": {
            "c7855cdc-6003-4fea-b8d4-37f5282ebfe8": {
                "pk": "c7855cdc-6003-4fea-b8d4-37f5282ebfe8",
                "name": "Junhoo Lee",
                "collaborators": [
                    "Nojun Kwak",
                    "Hyunho Lee",
                    "Yearim Kim",
                    "Hojun Lee",
                    "Suyoung Kim",
                    "Jaeyoung Yoo"
                ],
                "domain": [
                    "Dataset Distillation",
                    "Meta-Learning",
                    "Coreset Selection",
                    "Object Detection"
                ],
                "publications": [
                    {
                        "title": "Practical Dataset Distillation Based on Deep Support Vectors",
                        "abstract": "Conventional dataset distillation requires significant computational resources and assumes access to the entire dataset, an assumption impractical as it presumes all data resides on a central server. In this paper, we focus on dataset distillation in practical scenarios with access to only a fraction of the entire dataset. We introduce a novel distillation method that augments the conventional process by incorporating general model knowledge via the addition of Deep KKT (DKKT) loss. In practical settings, our approach showed improved performance compared to the baseline distribution matching distillation method on the CIFAR-10 dataset. Additionally, we present experimental evidence that Deep Support Vectors (DSVs) offer unique information to the original distillation, and their integration results in enhanced performance."
                    },
                    {
                        "title": "Any-Way Meta Learning",
                        "abstract": "Although meta-learning seems promising performance in the realm of rapid adaptability, it is constrained by fixed cardinality. When faced with tasks of varying cardinalities that were unseen during training, the model lacks its ability. In this paper, we address and resolve this challenge by harnessing `label equivalence' emerged from stochastic numeric label assignments during episodic task sampling. Questioning what defines ``true\" meta-learning, we introduce the ``any-way\" learning paradigm, an innovative model training approach that liberates model from fixed cardinality constraints. Surprisingly, this model not only matches but often outperforms traditional fixed-way models in terms of performance, convergence speed, and stability. This disrupts established notions about domain generalization. Furthermore, we argue that the inherent label equivalence naturally lacks semantic information. To bridge this semantic information gap arising from label equivalence, we further propose a mechanism for infusing semantic class information into the model. This would enhance the model's comprehension and functionality. Experiments conducted on renowned architectures like MAML and ProtoNet affirm the effectiveness of our method."
                    },
                    {
                        "title": "Coreset Selection for Object Detection",
                        "abstract": "Coreset selection is a method for selecting a small, representative subset of an entire dataset. It has been primarily researched in image classification, assuming there is only one object per image. However, coreset selection for object detection is more challenging as an image can contain multiple objects. As a result, much research has yet to be done on this topic. Therefore, we introduce a new approach, Coreset Selection for Object Detection (CSOD). CSOD generates imagewise and classwise representative feature vectors for multiple objects of the same class within each image. Subsequently, we adopt submodular optimization for considering both representativeness and diversity and utilize the representative vectors in the submodular optimization process to select a subset. When we evaluated CSOD on the Pascal VOC dataset, CSOD outperformed random selection by +6.4%p in AP$_{50}$ when selecting 200 images."
                    }
                ]
            },
            "bde33b43-4855-4c43-a173-8f5b7ee6294d": {
                "pk": "bde33b43-4855-4c43-a173-8f5b7ee6294d",
                "name": "Hyunho Lee",
                "collaborators": [
                    "Wenwen Li",
                    "Junhoo Lee",
                    "Nojun Kwak",
                    "Hojoon Lee",
                    "Sizhe Wang",
                    "Chia-Yu Hsu",
                    "Samantha T. Arundel",
                    "Yearim Kim",
                    "Seonghoon Jeong",
                    "Yunhui Jo"
                ],
                "domain": [
                    "Deep Learning",
                    "Remote Sensing",
                    "Reinforcement Learning",
                    "Geospatial AI"
                ],
                "publications": [
                    {
                        "title": "Improving Interpretability of Deep Active Learning for Flood Inundation Mapping Through Class Ambiguity Indices Using Multi-spectral Satellite Imagery",
                        "abstract": "Flood inundation mapping is a critical task for responding to the increasing risk of flooding linked to global warming. Significant advancements of deep learning in recent years have triggered its extensive applications, including flood inundation mapping. To cope with the time-consuming and labor-intensive data labeling process in supervised learning, deep active learning strategies are one of the feasible approaches. However, there remains limited exploration into the interpretability of how deep active learning strategies operate, with a specific focus on flood inundation mapping in the field of remote sensing. In this study, we introduce a novel framework of Interpretable Deep Active Learning for Flood inundation Mapping (IDAL-FIM), specifically in terms of class ambiguity of multi-spectral satellite images. In the experiments, we utilize Sen1Floods11 dataset, and adopt U-Net with MC-dropout. In addition, we employ five acquisition functions, which are the random, K-means, BALD, entropy, and margin acquisition functions. Based on the experimental results, we demonstrate that two proposed class ambiguity indices are effective variables to interpret the deep active learning by establishing statistically significant correlation with the predictive uncertainty of the deep learning model at the tile level. Then, we illustrate the behaviors of deep active learning through visualizing two-dimensional density plots and providing interpretations regarding the operation of deep active learning, in flood inundation mapping."
                    },
                    {
                        "title": "Practical Dataset Distillation Based on Deep Support Vectors",
                        "abstract": "Conventional dataset distillation requires significant computational resources and assumes access to the entire dataset, an assumption impractical as it presumes all data resides on a central server. In this paper, we focus on dataset distillation in practical scenarios with access to only a fraction of the entire dataset. We introduce a novel distillation method that augments the conventional process by incorporating general model knowledge via the addition of Deep KKT (DKKT) loss. In practical settings, our approach showed improved performance compared to the baseline distribution matching distillation method on the CIFAR-10 dataset. Additionally, we present experimental evidence that Deep Support Vectors (DSVs) offer unique information to the original distillation, and their integration results in enhanced performance."
                    },
                    {
                        "title": "Any-Way Meta Learning",
                        "abstract": "Although meta-learning seems promising performance in the realm of rapid adaptability, it is constrained by fixed cardinality. When faced with tasks of varying cardinalities that were unseen during training, the model lacks its ability. In this paper, we address and resolve this challenge by harnessing `label equivalence' emerged from stochastic numeric label assignments during episodic task sampling. Questioning what defines ``true\" meta-learning, we introduce the ``any-way\" learning paradigm, an innovative model training approach that liberates model from fixed cardinality constraints. Surprisingly, this model not only matches but often outperforms traditional fixed-way models in terms of performance, convergence speed, and stability. This disrupts established notions about domain generalization. Furthermore, we argue that the inherent label equivalence naturally lacks semantic information. To bridge this semantic information gap arising from label equivalence, we further propose a mechanism for infusing semantic class information into the model. This would enhance the model's comprehension and functionality. Experiments conducted on renowned architectures like MAML and ProtoNet affirm the effectiveness of our method."
                    },
                    {
                        "title": "Study of quality assurance regulations for linear accelerators in Korea: A comparison study between the current status in Korea and the international guidelines",
                        "abstract": "Quality assurance (QA) for medical linear accelerators is indispensable for appropriate cancer treatment. Some international organizations and western advanced countries provide QA guidelines for linear accelerators. Currently, QA regulations for linear accelerators in Korean hospitals specify a system in which each hospital stipulates its independent hospital-based protocols for QA procedures (HP_QAPs) and conducts QA based on these HP_QAPs while regulatory authorities verify whether items under these HP_QAPs have been performed. However, because this regulatory method cannot guarantee the quality of universal treatment, and QA items with tolerance criteria are different in many hospitals, the presentation of standardized QA items and tolerance criteria is essential. In this study, QA items in HP_QAPs from various hospitals and those presented by international organizations. Concordance rates between QA items for linear accelerators that were presented by the aforementioned organizations and those currently being implemented in Korean hospitals were shown to exhibit a daily QA of 50%, a weekly QA of 22%, a monthly QA of 43%, and an annual QA of 65%, and the overall concordance rates of all QA items were approximately 48%. In comparison between QA items being implemented in Korean hospitals and those being implemented in western advanced countries, concordance rates were shown to exhibit a daily QA of 50%, a weekly QA of 33%, a monthly QA of 60%, and an annual QA of 67%, and the overall concordance rate of all QA items were approximately 57%. The results of this study indicate that the HP_QAPs currently implemented by Korean hospitals as QA standards for linear accelerators used in radiation therapy do not meet international standards. To solve this problem, it is necessary to develop national standardized QA items and procedures for linear accelerators."
                    },
                    {
                        "title": "On the Importance of Feature Decorrelation for Unsupervised Representation Learning in Reinforcement Learning",
                        "abstract": "Recently, unsupervised representation learning (URL) has improved the sample efficiency of Reinforcement Learning (RL) by pretraining a model from a large unlabeled dataset. The underlying principle of these methods is to learn temporally predictive representations by predicting future states in the latent space. However, an important challenge of this approach is the representational collapse, where the subspace of the latent representations collapses into a low-dimensional manifold. To address this issue, we propose a novel URL framework that causally predicts future states while increasing the dimension of the latent manifold by decorrelating the features in the latent space. Through extensive empirical studies, we demonstrate that our framework effectively learns predictive representations without collapse, which significantly improves the sample efficiency of state-of-the-art URL methods on the Atari 100k benchmark. The code is available at https://github.com/dojeon-ai/SimTPR."
                    },
                    {
                        "title": "JORLDY: a fully customizable open source framework for reinforcement learning",
                        "abstract": "Recently, Reinforcement Learning (RL) has been actively researched in both academic and industrial fields. However, there exist only a few RL frameworks which are developed for researchers or students who want to study RL. In response, we propose an open-source RL framework \"Join Our Reinforcement Learning framework for Developing Yours\" (JORLDY). JORLDY provides more than 20 widely used RL algorithms which are implemented with Pytorch. Also, JORLDY supports multiple RL environments which include OpenAI gym, Unity ML-Agents, Mujoco, Super Mario Bros and Procgen. Moreover, the algorithmic components such as agent, network, environment can be freely customized, so that the users can easily modify and append algorithmic components. We expect that JORLDY will support various RL research and contribute further advance the field of RL. The source code of JORLDY is provided on the following Github: https://github.com/kakaoenterprise/JORLDY"
                    },
                    {
                        "title": "Assessment of a new GeoAI foundation model for flood inundation mapping",
                        "abstract": "Vision foundation models are a new frontier in Geospatial Artificial Intelligence (GeoAI), an interdisciplinary research area that applies and extends AI for geospatial problem solving and geographic knowledge discovery, because of their potential to enable powerful image analysis by learning and extracting important image features from vast amounts of geospatial data. This paper evaluates the performance of the first-of-its-kind geospatial foundation model, IBM-NASA's Prithvi, to support a crucial geospatial analysis task: flood inundation mapping. This model is compared with convolutional neural network and vision transformer-based architectures in terms of mapping accuracy for flooded areas. A benchmark dataset, Sen1Floods11, is used in the experiments, and the models' predictability, generalizability, and transferability are evaluated based on both a test dataset and a dataset that is completely unseen by the model. Results show the good transferability of the Prithvi model, highlighting its performance advantages in segmenting flooded areas in previously unseen regions. The findings also indicate areas for improvement for the Prithvi model in terms of adopting multi-scale representation learning, developing more end-to-end pipelines for high-level image analysis tasks, and offering more flexibility in terms of input data bands."
                    },
                    {
                        "title": "Segment Anything Model Can Not Segment Anything: Assessing AI Foundation Model's Generalizability in Permafrost Mapping",
                        "abstract": "This paper assesses trending AI foundation models, especially emerging computer vision foundation models and their performance in natural landscape feature segmentation. While the term foundation model has quickly garnered interest from the geospatial domain, its definition remains vague. Hence, this paper will first introduce AI foundation models and their defining characteristics. Built upon the tremendous success achieved by Large Language Models (LLMs) as the foundation models for language tasks, this paper discusses the challenges of building foundation models for geospatial artificial intelligence (GeoAI) vision tasks. To evaluate the performance of large AI vision models, especially Meta's Segment Anything Model (SAM), we implemented different instance segmentation pipelines that minimize the changes to SAM to leverage its power as a foundation model. A series of prompt strategies was developed to test SAM's performance regarding its theoretical upper bound of predictive accuracy, zero-shot performance, and domain adaptability through fine-tuning. The analysis used two permafrost feature datasets, ice-wedge polygons and retrogressive thaw slumps because (1) these landform features are more challenging to segment than manmade features due to their complicated formation mechanisms, diverse forms, and vague boundaries; (2) their presence and changes are important indicators for Arctic warming and climate change. The results show that although promising, SAM still has room for improvement to support AI-augmented terrain mapping. The spatial and domain generalizability of this finding is further validated using a more general dataset EuroCrop for agricultural field mapping. Finally, we discuss future research directions that strengthen SAM's applicability in challenging geospatial domains."
                    }
                ]
            },
            "2f247b5b-d7af-4c62-8e7e-17141e687768": {
                "pk": "2f247b5b-d7af-4c62-8e7e-17141e687768",
                "name": "Kyomin Hwang",
                "collaborators": [
                    "Nojun Kwak",
                    "Suyoung Kim",
                    "JunHoo Lee",
                    "Inseop Chung",
                    "Jayeon Yoo"
                ],
                "domain": [
                    "Large Models",
                    "Continual Learning",
                    "Adversarial Attacks",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Do not think pink elephant!",
                        "abstract": "Large Models (LMs) have heightened expectations for the potential of general AI as they are akin to human intelligence. This paper shows that recent large models such as Stable Diffusion and DALL-E3 also share the vulnerability of human intelligence, namely the \"white bear phenomenon\". We investigate the causes of the white bear phenomenon by analyzing their representation space. Based on this analysis, we propose a simple prompt-based attack method, which generates figures prohibited by the LM provider's policy. To counter these attacks, we introduce prompt-based defense strategies inspired by cognitive therapy techniques, successfully mitigating attacks by up to 48.22\\%."
                    },
                    {
                        "title": "Mitigating the Bias in the Model for Continual Test-Time Adaptation",
                        "abstract": "Continual Test-Time Adaptation (CTA) is a challenging task that aims to adapt a source pre-trained model to continually changing target domains. In the CTA setting, a model does not know when the target domain changes, thus facing a drastic change in the distribution of streaming inputs during the test-time. The key challenge is to keep adapting the model to the continually changing target domains in an online manner. We find that a model shows highly biased predictions as it constantly adapts to the chaining distribution of the target data. It predicts certain classes more often than other classes, making inaccurate over-confident predictions. This paper mitigates this issue to improve performance in the CTA scenario. To alleviate the bias issue, we make class-wise exponential moving average target prototypes with reliable target samples and exploit them to cluster the target features class-wisely. Moreover, we aim to align the target distributions to the source distribution by anchoring the target feature to its corresponding source prototype. With extensive experiments, our proposed method achieves noteworthy performance gain when applied on top of existing CTA methods without substantial adaptation time overhead."
                    }
                ]
            },
            "1f0e17a6-f20d-4e94-9b8a-cdfb1e67596a": {
                "pk": "1f0e17a6-f20d-4e94-9b8a-cdfb1e67596a",
                "name": "Nojun Kwak",
                "collaborators": [
                    "SeongUk Park",
                    "Simyung Chang",
                    "John Yang",
                    "Chaerin Kong",
                    "Jaeyoung Yoo",
                    "Minsung Hyun",
                    "Junyoung Choi",
                    "Seonhoon Kim",
                    "Seohyeong Jeong",
                    "Sungheon Park"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Computer Vision",
                    "Reinforcement Learning",
                    "Knowledge Distillation"
                ],
                "publications": [
                    {
                        "title": "Learning Dynamic BERT via Trainable Gate Variables and a Bi-modal Regularizer",
                        "abstract": "The BERT model has shown significant success on various natural language processing tasks. However, due to the heavy model size and high computational cost, the model suffers from high latency, which is fatal to its deployments on resource-limited devices. To tackle this problem, we propose a dynamic inference method on BERT via trainable gate variables applied on input tokens and a regularizer that has a bi-modal property. Our method shows reduced computational cost on the GLUE dataset with a minimal performance drop. Moreover, the model adjusts with a trade-off between performance and computational cost with the user-specified hyperparameter."
                    },
                    {
                        "title": "3D Human Pose Estimation with Relational Networks",
                        "abstract": "In this paper, we propose a novel 3D human pose estimation algorithm from a single image based on neural networks. We adopted the structure of the relational networks in order to capture the relations among different body parts. In our method, each pair of different body parts generates features, and the average of the features from all the pairs are used for 3D pose estimation. In addition, we propose a dropout method that can be used in relational modules, which inherently imposes robustness to the occlusions. The proposed network achieves state-of-the-art performance for 3D pose estimation in Human 3.6M dataset, and it effectively produces plausible results even in the existence of missing joints."
                    },
                    {
                        "title": "Image Enhancement by Recurrently-trained Super-resolution Network",
                        "abstract": "We introduce a new learning strategy for image enhancement by recurrently training the same simple superresolution (SR) network multiple times. After initially training an SR network by using pairs of a corrupted low resolution (LR) image and an original image, the proposed method makes use of the trained SR network to generate new high resolution (HR) images with a doubled resolution from the original uncorrupted images. Then, the new HR images are downscaled to the original resolution, which work as target images for the SR network in the next stage. The newly generated HR images by the repeatedly trained SR network show better image quality and this strategy of training LR to mimic new HR can lead to a more efficient SR network. Up to a certain point, by repeating this process multiple times, better and better images are obtained. This recurrent leaning strategy for SR can be a good solution for downsizing convolution networks and making a more efficient SR network. To measure the enhanced image quality, for the first time in this area of super-resolution and image enhancement, we use VIQET MOS score which reflects human visual quality more accurately than the conventional MSE measure."
                    },
                    {
                        "title": "Local-Selective Feature Distillation for Single Image Super-Resolution",
                        "abstract": "Recent improvements in convolutional neural network (CNN)-based single image super-resolution (SISR) methods rely heavily on fabricating network architectures, rather than finding a suitable training algorithm other than simply minimizing the regression loss. Adapting knowledge distillation (KD) can open a way for bringing further improvement for SISR, and it is also beneficial in terms of model efficiency. KD is a model compression method that improves the performance of Deep Neural Networks (DNNs) without using additional parameters for testing. It is getting the limelight recently for its competence at providing a better capacity-performance tradeoff. In this paper, we propose a novel feature distillation (FD) method which is suitable for SISR. We show the limitations of the existing FitNet-based FD method that it suffers in the SISR task, and propose to modify the existing FD algorithm to focus on local feature information. In addition, we propose a teacher-student-difference-based soft feature attention method that selectively focuses on specific pixel locations to extract feature information. We call our method local-selective feature distillation (LSFD) and verify that our method outperforms conventional FD methods in SISR problems."
                    },
                    {
                        "title": "Backdoor Attacks in Federated Learning by Rare Embeddings and Gradient Ensembling",
                        "abstract": "Recent advances in federated learning have demonstrated its promising capability to learn on decentralized datasets. However, a considerable amount of work has raised concerns due to the potential risks of adversaries participating in the framework to poison the global model for an adversarial purpose. This paper investigates the feasibility of model poisoning for backdoor attacks through rare word embeddings of NLP models. In text classification, less than 1% of adversary clients suffices to manipulate the model output without any drop in the performance on clean sentences. For a less complex dataset, a mere 0.1% of adversary clients is enough to poison the global model effectively. We also propose a technique specialized in the federated learning scheme called Gradient Ensemble, which enhances the backdoor performance in all our experimental settings."
                    },
                    {
                        "title": "Conservative Generator, Progressive Discriminator: Coordination of Adversaries in Few-shot Incremental Image Synthesis",
                        "abstract": "The capacity to learn incrementally from an online stream of data is an envied trait of human learners, as deep neural networks typically suffer from catastrophic forgetting and stability-plasticity dilemma. Several works have previously explored incremental few-shot learning, a task with greater challenges due to data constraint, mostly in classification setting with mild success. In this work, we study the underrepresented task of generative incremental few-shot learning. To effectively handle the inherent challenges of incremental learning and few-shot learning, we propose a novel framework named ConPro that leverages the two-player nature of GANs. Specifically, we design a conservative generator that preserves past knowledge in parameter and compute efficient manner, and a progressive discriminator that learns to reason semantic distances between past and present task samples, minimizing overfitting with few data points and pursuing good forward transfer. We present experiments to validate the effectiveness of ConPro."
                    },
                    {
                        "title": "Analyzing Multimodal Objectives Through the Lens of Generative Diffusion Guidance",
                        "abstract": "Recent years have witnessed astonishing advances in the field of multimodal representation learning, with contrastive learning being the cornerstone for major breakthroughs. Latest works delivered further improvements by incorporating different objectives such as masked modeling and captioning into the frameworks, but our understanding on how these objectives facilitate learning remains vastly incomplete. In this paper, we leverage the fact that classifier-guided diffusion models generate images that reflect the semantic signals provided by the classifier to study the characteristics of multimodal learning objectives. Specifically, we compare contrastive, matching and captioning loss in terms of their semantic signals, and introduce a simple baseline that not only supports our analyses but also improves the quality of generative guidance in a straightforward manner."
                    },
                    {
                        "title": "FEED: Feature-level Ensemble for Knowledge Distillation",
                        "abstract": "Knowledge Distillation (KD) aims to transfer knowledge in a teacher-student framework, by providing the predictions of the teacher network to the student network in the training stage to help the student network generalize better. It can use either a teacher with high capacity or {an} ensemble of multiple teachers. However, the latter is not convenient when one wants to use feature-map-based distillation methods. For a solution, this paper proposes a versatile and powerful training algorithm named FEature-level Ensemble for knowledge Distillation (FEED), which aims to transfer the ensemble knowledge using multiple teacher networks. We introduce a couple of training algorithms that transfer ensemble knowledge to the student at the feature map level. Among the feature-map-based distillation methods, using several non-linear transformations in parallel for transferring the knowledge of the multiple teacher{s} helps the student find more generalized solutions. We name this method as parallel FEED, andexperimental results on CIFAR-100 and ImageNet show that our method has clear performance enhancements, without introducing any additional parameters or computations at test time. We also show the experimental results of sequentially feeding teacher's information to the student, hence the name sequential FEED, and discuss the lessons obtained. Additionally, the empirical results on measuring the reconstruction errors at the feature map give hints for the enhancements."
                    },
                    {
                        "title": "Unpriortized Autoencoder For Image Generation",
                        "abstract": "In this paper, we treat the image generation task using an autoencoder, a representative latent model. Unlike many studies regularizing the latent variable's distribution by assuming a manually specified prior, we approach the image generation task using an autoencoder by directly estimating the latent distribution. To this end, we introduce 'latent density estimator' which captures latent distribution explicitly and propose its structure. Through experiments, we show that our generative model generates images with the improved visual quality compared to previous autoencoder-based generative models."
                    },
                    {
                        "title": "Disentangling Options with Hellinger Distance Regularizer",
                        "abstract": "In reinforcement learning (RL), temporal abstraction still remains as an important and unsolved problem. The options framework provided clues to temporal abstraction in the RL, and the option-critic architecture elegantly solved the two problems of finding options and learning RL agents in an end-to-end manner. However, it is necessary to examine whether the options learned through this method play a mutually exclusive role. In this paper, we propose a Hellinger distance regularizer, a method for disentangling options. In addition, we will shed light on various indicators from the statistical point of view to compare with the options learned through the existing option-critic architecture."
                    },
                    {
                        "title": "Do not think pink elephant!",
                        "abstract": "Large Models (LMs) have heightened expectations for the potential of general AI as they are akin to human intelligence. This paper shows that recent large models such as Stable Diffusion and DALL-E3 also share the vulnerability of human intelligence, namely the \"white bear phenomenon\". We investigate the causes of the white bear phenomenon by analyzing their representation space. Based on this analysis, we propose a simple prompt-based attack method, which generates figures prohibited by the LM provider's policy. To counter these attacks, we introduce prompt-based defense strategies inspired by cognitive therapy techniques, successfully mitigating attacks by up to 48.22\\%."
                    },
                    {
                        "title": "Practical Dataset Distillation Based on Deep Support Vectors",
                        "abstract": "Conventional dataset distillation requires significant computational resources and assumes access to the entire dataset, an assumption impractical as it presumes all data resides on a central server. In this paper, we focus on dataset distillation in practical scenarios with access to only a fraction of the entire dataset. We introduce a novel distillation method that augments the conventional process by incorporating general model knowledge via the addition of Deep KKT (DKKT) loss. In practical settings, our approach showed improved performance compared to the baseline distribution matching distillation method on the CIFAR-10 dataset. Additionally, we present experimental evidence that Deep Support Vectors (DSVs) offer unique information to the original distillation, and their integration results in enhanced performance."
                    },
                    {
                        "title": "Broadcasting Convolutional Network for Visual Relational Reasoning",
                        "abstract": "In this paper, we propose the Broadcasting Convolutional Network (BCN) that extracts key object features from the global field of an entire input image and recognizes their relationship with local features. BCN is a simple network module that collects effective spatial features, embeds location information and broadcasts them to the entire feature maps. We further introduce the Multi-Relational Network (multiRN) that improves the existing Relation Network (RN) by utilizing the BCN module. In pixel-based relation reasoning problems, with the help of BCN, multiRN extends the concept of `pairwise relations' in conventional RNs to `multiwise relations' by relating each object with multiple objects at once. This yields in O(n) complexity for n objects, which is a vast computational gain from RNs that take O(n^2). Through experiments, multiRN has achieved a state-of-the-art performance on CLEVR dataset, which proves the usability of BCN on relation reasoning problems."
                    },
                    {
                        "title": "Image Restoration by Estimating Frequency Distribution of Local Patches",
                        "abstract": "In this paper, we propose a method to solve the image restoration problem, which tries to restore the details of a corrupted image, especially due to the loss caused by JPEG compression. We have treated an image in the frequency domain to explicitly restore the frequency components lost during image compression. In doing so, the distribution in the frequency domain is learned using the cross entropy loss. Unlike recent approaches, we have reconstructed the details of an image without using the scheme of adversarial training. Rather, the image restoration problem is treated as a classification problem to determine the frequency coefficient for each frequency band in an image patch. In this paper, we show that the proposed method effectively restores a JPEG-compressed image with more detailed high frequency components, making the restored image more vivid."
                    },
                    {
                        "title": "Semantic Sentence Matching with Densely-connected Recurrent and Co-attentive Information",
                        "abstract": "Sentence matching is widely used in various natural language tasks such as natural language inference, paraphrase identification, and question answering. For these tasks, understanding logical and semantic relationship between two sentences is required but it is yet challenging. Although attention mechanism is useful to capture the semantic relationship and to properly align the elements of two sentences, previous methods of attention mechanism simply use a summation operation which does not retain original features enough. Inspired by DenseNet, a densely connected convolutional network, we propose a densely-connected co-attentive recurrent neural network, each layer of which uses concatenated information of attentive features as well as hidden features of all the preceding recurrent layers. It enables preserving the original and the co-attentive feature information from the bottommost word embedding layer to the uppermost recurrent layer. To alleviate the problem of an ever-increasing size of feature vectors due to dense concatenation operations, we also propose to use an autoencoder after dense concatenation. We evaluate our proposed architecture on highly competitive benchmark datasets related to sentence matching. Experimental results show that our architecture, which retains recurrent and attentive features, achieves state-of-the-art performances for most of the tasks."
                    },
                    {
                        "title": "StackNet: Stacking Parameters for Continual learning",
                        "abstract": "Training a neural network for a classification task typically assumes that the data to train are given from the beginning. However, in the real world, additional data accumulate gradually and the model requires additional training without accessing the old training data. This usually leads to the catastrophic forgetting problem which is inevitable for the traditional training methodology of neural networks. In this paper, we propose a continual learning method that is able to learn additional tasks while retaining the performance of previously learned tasks by stacking parameters. Composed of two complementary components, the index module and the StackNet, our method estimates the index of the corresponding task for an input sample with the index module and utilizes a particular portion of StackNet with this index. The StackNet guarantees no degradation in the performance of the previously learned tasks and the index module shows high confidence in finding the origin of an input sample. Compared to the previous work of PackNet, our method is competitive and highly intuitive."
                    },
                    {
                        "title": "Textbook Question Answering with Multi-modal Context Graph Understanding and Self-supervised Open-set Comprehension",
                        "abstract": "In this work, we introduce a novel algorithm for solving the textbook question answering (TQA) task which describes more realistic QA problems compared to other recent tasks. We mainly focus on two related issues with analysis of the TQA dataset. First, solving the TQA problems requires to comprehend multi-modal contexts in complicated input data. To tackle this issue of extracting knowledge features from long text lessons and merging them with visual features, we establish a context graph from texts and images, and propose a new module f-GCN based on graph convolutional networks (GCN). Second, scientific terms are not spread over the chapters and subjects are split in the TQA dataset. To overcome this so called \"out-of-domain\" issue, before learning QA problems, we introduce a novel self-supervised open-set learning process without any annotations. The experimental results show that our model significantly outperforms prior state-of-the-art methods. Moreover, ablation studies validate that both methods of incorporating f-GCN for extracting knowledge from multi-modal contexts and our newly proposed self-supervised learning process are effective for TQA problems."
                    },
                    {
                        "title": "Sym-parameterized Dynamic Inference for Mixed-Domain Image Translation",
                        "abstract": "Recent advances in image-to-image translation have led to some ways to generate multiple domain images through a single network. However, there is still a limit in creating an image of a target domain without a dataset on it. We propose a method that expands the concept of `multi-domain' from data to the loss area and learns the combined characteristics of each domain to dynamically infer translations of images in mixed domains. First, we introduce Sym-parameter and its learning method for variously mixed losses while synchronizing them with input conditions. Then, we propose Sym-parameterized Generative Network (SGN) which is empirically confirmed of learning mixed characteristics of various data and losses, and translating images to any mixed-domain without ground truths, such as 30% Van Gogh and 20% Monet and 40% snowy."
                    },
                    {
                        "title": "Genetic-Gated Networks for Deep Reinforcement",
                        "abstract": "We introduce the Genetic-Gated Networks (G2Ns), simple neural networks that combine a gate vector composed of binary genetic genes in the hidden layer(s) of networks. Our method can take both advantages of gradient-free optimization and gradient-based optimization methods, of which the former is effective for problems with multiple local minima, while the latter can quickly find local minima. In addition, multiple chromosomes can define different models, making it easy to construct multiple models and can be effectively applied to problems that require multiple models. We show that this G2N can be applied to typical reinforcement learning algorithms to achieve a large improvement in sample efficiency and performance."
                    },
                    {
                        "title": "Task-oriented Design through Deep Reinforcement Learning",
                        "abstract": "We propose a new low-cost machine-learning-based methodology which assists designers in reducing the gap between the problem and the solution in the design process. Our work applies reinforcement learning (RL) to find the optimal task-oriented design solution through the construction of the design action for each task. For this task-oriented design, the 3D design process in product design is assigned to an action space in Deep RL, and the desired 3D model is obtained by training each design action according to the task. By showing that this method achieves satisfactory design even when applied to a task pursuing multiple goals, we suggest the direction of how machine learning can contribute to the design process. Also, we have validated with product designers that this methodology can assist the creative part in the process of design."
                    }
                ]
            }
        }
    },
    "2405.18822": {
        "paper_data": {
            "title": "Toxicity Detection for Free",
            "url": "http://arxiv.org/abs/2405.18822v1",
            "arxiv_id": "2405.18822",
            "authors": [
                "Zhanhao Hu",
                "Julien Piet",
                "Geng Zhao",
                "Jiantao Jiao",
                "David Wagner"
            ],
            "abstract": "Current LLMs are generally aligned to follow safety requirements and tend to refuse toxic prompts. However, LLMs can fail to refuse toxic prompts or be overcautious and refuse benign examples. In addition, state-of-the-art toxicity detectors have low TPRs at low FPR, incurring high costs in real-world applications where toxic examples are rare. In this paper, we explore Moderation Using LLM Introspection (MULI), which detects toxic prompts using the information extracted directly from LLMs themselves. We found significant gaps between benign and toxic prompts in the distribution of alternative refusal responses and in the distribution of the first response token's logits. These gaps can be used to detect toxicities: We show that a toy model based on the logits of specific starting tokens gets reliable performance, while requiring no training or additional computational cost. We build a more robust detector using a sparse logistic regression model on the first response token logits, which greatly exceeds SOTA detectors under multiple metrics.",
            "introduction": "   1 Introduction  Significant progress has been made in recent large language models. LLMs acquire substantial knowledge from wide text corpora, demonstrating a remarkable ability to provide high-quality responses to various prompts. They are widely used in downstream tasks such as chatbots\u00a0[15, 4] and general tool use\u00a0[20, 6]. However, LLMs raise serious safety concerns. For instance, malicious users could ask LLMs to write phishing emails or provide instructions on how to commit a crime\u00a0[26, 9].   Current LLMs have incorporated safety alignment\u00a0[24, 21] in their training phase to alleviate security concerns. Consequently, they are generally tuned to decline to answer toxic prompts. However, alignment is not perfect, and many models can be either overcautious (which is frustrating for benign users) or too-easily deceived (e.g., by jailbreak attacks) [25, 13, 18, 14]. One approach is to supplement alignment tuning with a toxicity detector\u00a0[10, 2, 1, 3], a classifier that is designed to detect toxic, harmful, or inappropriate prompts to the LLM. By querying the detector for every prompt, LLM vendors can immediately stop generating responses whenever they detect toxic content. These detectors are usually based on an additional LLM that is finetuned on toxic and benign data.   Figure 1:  Pipeline of MULI. Left: Existing methods use a separate LLM as a toxicity detector, thus having up to a 2x overhead. Right: We leverage the original LLM\u2019s first token logits to detect toxicity using sparse logistic regression, incurring negligible overhead.    Current detectors are imperfect and make mistakes. In real-world applications, toxic examples are rare and most prompts are benign, so test data exhibits high class imbalance: even small False Positive Rates (FPR) can cause many false alarms in this scenario \u00a0[5]. Unfortunately, state-of-the-art content moderation classifiers and toxicity detectors are not able to achieve high True Positive Rates (TPR) and very low FPRs, and they struggle with some inputs.   Existing detectors also impose extra costs. At training time, one must collect a comprehensive dataset of toxic and benign examples for fine-tuning such a model. At test time, LLM providers must also query a separate toxicity detection model, which increases the cost of LLM serving and can incur additional latency. Some detectors require seeing both the entire input to the LLM and the entire output, which is incompatible with providing streaming responses; in practice, providers deal with this by applying the detector only to the input (which in current schemes leads to missing some toxic responses) or applying the detector once the entire output has been generated and attempting to erase the output if it is toxic (but by then, the output may already have been displayed to the user or returned to the API client, so it is arguably too late).   In this work, we propose a new approach to toxicity detection, Moderation Using LLM Introspection (MULI), that addresses these shortcomings. We simultaneously achieve better detection performance than existing detectors and eliminate extra costs. Our scheme, MULI, is based on examining the output of the model being queried (Figure\u00a01). This avoids the need to apply a separate detection model; and achieves good performance without needing the output, so we can proactively block prompts that are toxic or would lead",
            "references": [
                {
                    "title": "Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations",
                    "abstract": "We introduce Llama Guard, an LLM-based input-output safeguard model geared towards Human-AI conversation use cases. Our model incorporates a safety risk taxonomy, a valuable tool for categorizing a specific set of safety risks found in LLM prompts (i.e., prompt classification). This taxonomy is also instrumental in classifying the responses generated by LLMs to these prompts, a process we refer to as response classification. For the purpose of both prompt and response classification, we have meticulously gathered a dataset of high quality. Llama Guard, a Llama2-7b model that is instruction-tuned on our collected dataset, albeit low in volume, demonstrates strong performance on existing benchmarks such as the OpenAI Moderation Evaluation dataset and ToxicChat, where its performance matches or exceeds that of currently available content moderation tools. Llama Guard functions as a language model, carrying out multi-class classification and generating binary decision scores. Furthermore, the instruction fine-tuning of Llama Guard allows for the customization of tasks and the adaptation of output formats. This feature enhances the model's capabilities, such as enabling the adjustment of taxonomy categories to align with specific use cases, and facilitating zero-shot or few-shot prompting with diverse taxonomies at the input. We are making Llama Guard model weights available and we encourage researchers to further develop and adapt them to meet the evolving needs of the community for AI safety."
                },
                {
                    "title": "Mistral 7B",
                    "abstract": "We introduce Mistral 7B v0.1, a 7-billion-parameter language model engineered for superior performance and efficiency. Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation. Our model leverages grouped-query attention (GQA) for faster inference, coupled with sliding window attention (SWA) to effectively handle sequences of arbitrary length with a reduced inference cost. We also provide a model fine-tuned to follow instructions, Mistral 7B -- Instruct, that surpasses the Llama 2 13B -- Chat model both on human and automated benchmarks. Our models are released under the Apache 2.0 license."
                },
                {
                    "title": "Large Language Model Alignment: A Survey",
                    "abstract": "Recent years have witnessed remarkable progress made in large language models (LLMs). Such advancements, while garnering significant attention, have concurrently elicited various concerns. The potential of these models is undeniably vast; however, they may yield texts that are imprecise, misleading, or even detrimental. Consequently, it becomes paramount to employ alignment techniques to ensure these models to exhibit behaviors consistent with human values. This survey endeavors to furnish an extensive exploration of alignment methodologies designed for LLMs, in conjunction with the extant capability research in this domain. Adopting the lens of AI alignment, we categorize the prevailing methods and emergent proposals for the alignment of LLMs into outer and inner alignment. We also probe into salient issues including the models' interpretability, and potential vulnerabilities to adversarial attacks. To assess LLM alignment, we present a wide variety of benchmarks and evaluation methodologies. After discussing the state of alignment research for LLMs, we finally cast a vision toward the future, contemplating the promising avenues of research that lie ahead. Our aspiration for this survey extends beyond merely spurring research interests in this realm. We also envision bridging the gap between the AI alignment research community and the researchers engrossed in the capability exploration of LLMs for both capable and safe LLMs."
                },
                {
                    "title": "LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset",
                    "abstract": "Studying how people interact with large language models (LLMs) in real-world scenarios is increasingly important due to their widespread use in various applications. In this paper, we introduce LMSYS-Chat-1M, a large-scale dataset containing one million real-world conversations with 25 state-of-the-art LLMs. This dataset is collected from 210K unique IP addresses in the wild on our Vicuna demo and Chatbot Arena website. We offer an overview of the dataset's content, including its curation process, basic statistics, and topic distribution, highlighting its diversity, originality, and scale. We demonstrate its versatility through four use cases: developing content moderation models that perform similarly to GPT-4, building a safety benchmark, training instruction-following models that perform similarly to Vicuna, and creating challenging benchmark questions. We believe that this dataset will serve as a valuable resource for understanding and advancing LLM capabilities. The dataset is publicly available at https://huggingface.co/datasets/lmsys/lmsys-chat-1m."
                },
                {
                    "title": "Universal and Transferable Adversarial Attacks on Aligned Language Models",
                    "abstract": "Because\"out-of-the-box\"large language models are capable of generating a great deal of objectionable content, recent work has focused on aligning these models in an attempt to prevent undesirable generation. While there has been some success at circumventing these measures -- so-called\"jailbreaks\"against LLMs -- these attacks have required significant human ingenuity and are brittle in practice. In this paper, we propose a simple and effective attack method that causes aligned language models to generate objectionable behaviors. Specifically, our approach finds a suffix that, when attached to a wide range of queries for an LLM to produce objectionable content, aims to maximize the probability that the model produces an affirmative response (rather than refusing to answer). However, instead of relying on manual engineering, our approach automatically produces these adversarial suffixes by a combination of greedy and gradient-based search techniques, and also improves over past automatic prompt generation methods. Surprisingly, we find that the adversarial prompts generated by our approach are quite transferable, including to black-box, publicly released LLMs. Specifically, we train an adversarial attack suffix on multiple prompts (i.e., queries asking for many different types of objectionable content), as well as multiple models (in our case, Vicuna-7B and 13B). When doing so, the resulting attack suffix is able to induce objectionable content in the public interfaces to ChatGPT, Bard, and Claude, as well as open source LLMs such as LLaMA-2-Chat, Pythia, Falcon, and others. In total, this work significantly advances the state-of-the-art in adversarial attacks against aligned language models, raising important questions about how such systems can be prevented from producing objectionable information. Code is available at github.com/llm-attacks/llm-attacks."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Latent Jailbreak: A Benchmark for Evaluating Text Safety and Output Robustness of Large Language Models",
                    "abstract": "Considerable research efforts have been devoted to ensuring that large language models (LLMs) align with human values and generate safe text. However, an excessive focus on sensitivity to certain topics can compromise the model's robustness in following instructions, thereby impacting its overall performance in completing tasks. Previous benchmarks for jailbreaking LLMs have primarily focused on evaluating the safety of the models without considering their robustness. In this paper, we propose a benchmark that assesses both the safety and robustness of LLMs, emphasizing the need for a balanced approach. To comprehensively study text safety and output robustness, we introduce a latent jailbreak prompt dataset, each involving malicious instruction embedding. Specifically, we instruct the model to complete a regular task, such as translation, with the text to be translated containing malicious instructions. To further analyze safety and robustness, we design a hierarchical annotation framework. We present a systematic analysis of the safety and robustness of LLMs regarding the position of explicit normal instructions, word replacements (verbs in explicit normal instructions, target groups in malicious instructions, cue words for explicit normal instructions), and instruction replacements (different explicit normal instructions). Our results demonstrate that current LLMs not only prioritize certain instruction verbs but also exhibit varying jailbreak rates for different instruction verbs in explicit normal instructions. Code and data are available at https://github.com/qiuhuachuan/latent-jailbreak."
                },
                {
                    "title": "Jailbroken: How Does LLM Safety Training Fail?",
                    "abstract": "Large language models trained for safety and harmlessness remain susceptible to adversarial misuse, as evidenced by the prevalence of\"jailbreak\"attacks on early releases of ChatGPT that elicit undesired behavior. Going beyond recognition of the issue, we investigate why such attacks succeed and how they can be created. We hypothesize two failure modes of safety training: competing objectives and mismatched generalization. Competing objectives arise when a model's capabilities and safety goals conflict, while mismatched generalization occurs when safety training fails to generalize to a domain for which capabilities exist. We use these failure modes to guide jailbreak design and then evaluate state-of-the-art models, including OpenAI's GPT-4 and Anthropic's Claude v1.3, against both existing and newly designed attacks. We find that vulnerabilities persist despite the extensive red-teaming and safety-training efforts behind these models. Notably, new attacks utilizing our failure modes succeed on every prompt in a collection of unsafe requests from the models' red-teaming evaluation sets and outperform existing ad hoc jailbreaks. Our analysis emphasizes the need for safety-capability parity -- that safety mechanisms should be as sophisticated as the underlying model -- and argues against the idea that scaling alone can resolve these safety failure modes."
                },
                {
                    "title": "From ChatGPT to ThreatGPT: Impact of Generative AI in Cybersecurity and Privacy",
                    "abstract": "Undoubtedly, the evolution of Generative AI (GenAI) models has been the highlight of digital transformation in the year 2022. As the different GenAI models like ChatGPT and Google Bard continue to foster their complexity and capability, it\u2019s critical to understand its consequences from a cybersecurity perspective. Several instances recently have demonstrated the use of GenAI tools in both the defensive and offensive side of cybersecurity, and focusing on the social, ethical and privacy implications this technology possesses. This research paper highlights the limitations, challenges, potential risks, and opportunities of GenAI in the domain of cybersecurity and privacy. The work presents the vulnerabilities of ChatGPT, which can be exploited by malicious users to exfiltrate malicious information bypassing the ethical constraints on the model. This paper demonstrates successful example attacks like Jailbreaks, reverse psychology, and prompt injection attacks on the ChatGPT. The paper also investigates how cyber offenders can use the GenAI tools in developing cyber attacks, and explore the scenarios where ChatGPT can be used by adversaries to create social engineering attacks, phishing attacks, automated hacking, attack payload generation, malware creation, and polymorphic malware. This paper then examines defense techniques and uses GenAI tools to improve security measures, including cyber defense automation, reporting, threat intelligence, secure code generation and detection, attack identification, developing ethical guidelines, incidence response plans, and malware detection. We will also discuss the social, legal, and ethical implications of ChatGPT. In conclusion, the paper highlights open challenges and future directions to make this GenAI secure, safe, trustworthy, and ethical as the community understands its cybersecurity impacts."
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "Large Language Models as Tool Makers",
                    "abstract": "Recent research has highlighted the potential of large language models (LLMs) to improve their problem-solving capabilities with the aid of suitable external tools. In our work, we further advance this concept by introducing a closed-loop framework, referred to as LLMs A s Tool Makers (LATM), where LLMs create their own reusable tools for problem-solving. Our approach consists of two phases: 1) tool making: an LLM acts as the tool maker that crafts tools for a set of tasks. 2) tool using: another LLM acts as the tool user, which applies the tool built by the tool maker for problem-solving. On the problem-solving server side, tool-making enables continual tool generation and caching as new requests emerge. This framework enables subsequent requests to access cached tools via their corresponding APIs, enhancing the efficiency of task resolution. Recognizing that tool-making requires more sophisticated capabilities, we assign this task to a powerful, albeit resource-intensive, model. Conversely, the simpler tool-using phase is delegated to a lightweight model. This strategic division of labor allows the once-off cost of tool-making to be spread over multiple instances of tool-using, significantly reducing average costs while maintaining strong performance. Furthermore, our method offers a functional cache through the caching and reuse of tools, which stores the functionality of a class of requests instead of the natural language responses from LLMs, thus extending the applicability of the conventional cache mechanism. We evaluate our approach across various complex reasoning tasks, including Big-Bench tasks. With GPT-4 as the tool maker and GPT-3.5 as the tool user, LATM demonstrates performance equivalent to using GPT-4 for both roles, but with a significantly reduced inference cost."
                },
                {
                    "title": "Jailbreaking ChatGPT via Prompt Engineering: An Empirical Study",
                    "abstract": "Large Language Models (LLMs), like ChatGPT, have demonstrated vast potential but also introduce challenges related to content constraints and potential misuse. Our study investigates three key research questions: (1) the number of different prompt types that can jailbreak LLMs, (2) the effectiveness of jailbreak prompts in circumventing LLM constraints, and (3) the resilience of ChatGPT against these jailbreak prompts. Initially, we develop a classification model to analyze the distribution of existing prompts, identifying ten distinct patterns and three categories of jailbreak prompts. Subsequently, we assess the jailbreak capability of prompts with ChatGPT versions 3.5 and 4.0, utilizing a dataset of 3,120 jailbreak questions across eight prohibited scenarios. Finally, we evaluate the resistance of ChatGPT against jailbreak prompts, finding that the prompts can consistently evade the restrictions in 40 use-case scenarios. The study underscores the importance of prompt structures in jailbreaking LLMs and discusses the challenges of robust jailbreak prompt generation and prevention."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Toolformer: Language Models Can Teach Themselves to Use Tools",
                    "abstract": "Language models (LMs) exhibit remarkable abilities to solve new tasks from just a few examples or textual instructions, especially at scale. They also, paradoxically, struggle with basic functionality, such as arithmetic or factual lookup, where much simpler and smaller models excel. In this paper, we show that LMs can teach themselves to use external tools via simple APIs and achieve the best of both worlds. We introduce Toolformer, a model trained to decide which APIs to call, when to call them, what arguments to pass, and how to best incorporate the results into future token prediction. This is done in a self-supervised way, requiring nothing more than a handful of demonstrations for each API. We incorporate a range of tools, including a calculator, a Q\\&A system, two different search engines, a translation system, and a calendar. Toolformer achieves substantially improved zero-shot performance across a variety of downstream tasks, often competitive with much larger models, without sacrificing its core language modeling abilities."
                },
                {
                    "title": "Scaling Instruction-Finetuned Language Models",
                    "abstract": "Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Ethical and social risks of harm from Language Models",
                    "abstract": "This paper aims to help structure the risk landscape associated with large-scale Language Models (LMs). In order to foster advances in responsible innovation, an in-depth understanding of the potential risks posed by these models is needed. A wide range of established and anticipated risks are analysed in detail, drawing on multidisciplinary expertise and literature from computer science, linguistics, and social sciences. We outline six specific risk areas: I. Discrimination, Exclusion and Toxicity, II. Information Hazards, III. Misinformation Harms, V. Malicious Uses, V. Human-Computer Interaction Harms, VI. Automation, Access, and Environmental Harms. The first area concerns the perpetuation of stereotypes, unfair discrimination, exclusionary norms, toxic language, and lower performance by social group for LMs. The second focuses on risks from private data leaks or LMs correctly inferring sensitive information. The third addresses risks arising from poor, false or misleading information including in sensitive domains, and knock-on risks such as the erosion of trust in shared information. The fourth considers risks from actors who try to use LMs to cause harm. The fifth focuses on risks specific to LLMs used to underpin conversational agents that interact with human users, including unsafe use, manipulation or deception. The sixth discusses the risk of environmental harm, job automation, and other challenges that may have a disparate effect on different social groups or communities. In total, we review 21 risks in-depth. We discuss the points of origin of different risks and point to potential mitigation approaches. Lastly, we discuss organisational responsibilities in implementing mitigations, and the role of collaboration and participation. We highlight directions for further research, particularly on expanding the toolkit for assessing and evaluating the outlined risks in LMs."
                },
                {
                    "title": "What\u2019s in the Box? An Analysis of Undesirable Content in the Common Crawl Corpus",
                    "abstract": "Whereas much of the success of the current generation of neural language models has been driven by increasingly large training corpora, relatively little research has been dedicated to analyzing these massive sources of textual data. In this exploratory analysis, we delve deeper into the Common Crawl, a colossal web corpus that is extensively used for training language models. We find that it contains a significant amount of undesirable content, including hate speech and sexually explicit content, even after filtering procedures. We discuss the potential impacts of this content on language models and conclude with future research directions and a more mindful approach to corpus collection and analysis."
                },
                {
                    "title": "Aligning Large Language Models with Human: A Survey",
                    "abstract": "Large Language Models (LLMs) trained on extensive textual corpora have emerged as leading solutions for a broad array of Natural Language Processing (NLP) tasks. Despite their notable performance, these models are prone to certain limitations such as mis-understanding human instructions, generating potentially biased content, or factually incorrect (hallucinated) information. Hence, aligning LLMs with human expectations has become an active area of interest within the research community. This survey presents a comprehensive overview of these alignment technologies, including the following aspects. (1) Data collection : the methods for effectively collecting high-quality instructions for LLM alignment, including the use of NLP benchmarks, human annotations, and leveraging strong LLMs. (2) Training methodologies : a detailed review of the prevailing training methods employed for LLM alignment. Our exploration encompasses Supervised Fine-tuning, both Online and Of\ufb02ine human preference training, along with parameter-ef\ufb01cient training mechanisms. (3) Model Evaluation : the methods for evaluating the effectiveness of these human-aligned LLMs, presenting a multifaceted approach towards their assessment. In conclusion, we collate and distill our \ufb01ndings, shedding light on several promising future research avenues in the \ufb01eld. This survey, therefore, serves as a valuable resource for anyone invested in understanding and advancing the alignment of LLMs to better suit human-oriented tasks and expectations. An associated GitHub link collecting the latest papers is available at"
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve toxicity detection in large language models (LLMs) to reduce false positives and enhance safety without incurring additional computational costs?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the pressing safety concerns associated with LLMs, which are increasingly integrated into various applications. By enhancing toxicity detection, we can ensure that LLMs provide safer interactions, thereby fostering user trust and expanding their practical applications. This research could lead to advancements in content moderation techniques, influencing future studies on LLM safety and alignment, and potentially setting new standards for responsible AI deployment.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in improving toxicity detection stem from the high class imbalance in real-world data, where toxic prompts are rare compared to benign ones. This imbalance complicates the achievement of high True Positive Rates (TPR) and low False Positive Rates (FPR). Naive approaches may fail because they do not account for the nuances of language and context, leading to either excessive caution or vulnerability to manipulation (e.g., jailbreak attacks). Additionally, existing methods often require separate models that introduce latency and additional costs, making it difficult to implement effective real-time moderation.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the reliance on separate toxicity detection models, which not only increase computational overhead but also struggle with high FPRs and TPRs due to class imbalance. Barriers such as the need for comprehensive datasets for training and the inability to detect toxicity in real-time outputs have hindered progress. Our approach, MULI, differs by leveraging the original LLM\u2019s first token logits for toxicity detection, eliminating the need for a separate model and addressing the shortcomings of prior methods.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Moderation Using LLM Introspection (MULI), involves using sparse logistic regression on the first token logits of the original LLM to detect toxicity. We will evaluate our approach using a dataset that includes both toxic and benign examples, measuring performance through metrics such as True Positive Rate (TPR) and False Positive Rate (FPR). The expected outcomes include improved detection performance compared to existing methods, reduced computational overhead, and the ability to proactively block toxic prompts before they lead to harmful outputs."
            }
        },
        "author_data": {
            "143b24e1-7ca1-4090-b626-8f2aa3c0bcef": {
                "pk": "143b24e1-7ca1-4090-b626-8f2aa3c0bcef",
                "name": "Zhanhao Hu",
                "collaborators": [
                    "Xiaolin Hu",
                    "Bo Zhang",
                    "Xiaopei Zhu",
                    "Xiao Li",
                    "Siyuan Huang",
                    "Jianmin Li",
                    "Qiongxiu Li",
                    "Tao Wang",
                    "Jun Zhu",
                    "Yuqiu Liu"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Spiking Neural Networks",
                    "Privacy in Machine Learning",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "An STDP-Based Supervised Learning Algorithm for Spiking Neural Networks",
                        "abstract": "Compared with rate-based artificial neural networks, Spiking Neural Networks (SNN) provide a more biological plausible model for the brain. But how they perform supervised learning remains elusive. Inspired by recent works of Bengio et al., we propose a supervised learning algorithm based on Spike-Timing Dependent Plasticity (STDP) for a hierarchical SNN consisting of Leaky Integrate-and-fire (LIF) neurons. A time window is designed for the presynaptic neuron and only the spikes in this window take part in the STDP updating process. The model is trained on the MNIST dataset. The classification accuracy approach that of a Multilayer Perceptron (MLP) with similar architecture trained by the standard back-propagation algorithm."
                    },
                    {
                        "title": "Amplification trojan network: Attack deep neural networks by amplifying their inherent weakness",
                        "abstract": "Recent works found that deep neural networks (DNNs) can be fooled by adversarial examples, which are crafted by adding adversarial noise on clean inputs. The accuracy of DNNs on adversarial examples will decrease as the magnitude of the adversarial noise increase. In this study, we show that DNNs can be also fooled when the noise is very small under certain circumstances. This new type of attack is called Amplification Trojan Attack (ATAttack). Specifically, we use a trojan network to transform the inputs before sending them to the target DNN. This trojan network serves as an amplifier to amplify the inherent weakness of the target DNN. The target DNN, which is infected by the trojan network, performs normally on clean data while being more vulnerable to adversarial examples. Since it only transforms the inputs, the trojan network can hide in DNN-based pipelines, e.g. by infecting the pre-processing procedure of the inputs before sending them to the DNNs. This new type of threat should be considered in developing safe DNNs."
                    },
                    {
                        "title": "Infrared Invisible Clothing:Hiding from Infrared Detectors at Multiple Angles in Real World",
                        "abstract": "Thermal infrared imaging is widely used in body temperature measurement, security monitoring, and so on, but its safety research attracted attention only in recent years. We proposed the infrared adversarial clothing, which could fool infrared pedestrian detectors at different angles. We simulated the process from cloth to clothing in the digital world and then designed the adversarial \"QR code\" pattern. The core of our method is to design a basic pattern that can be expanded periodically, and make the pattern after random cropping and deformation still have an adversarial effect, then we can process the flat cloth with an adversarial pattern into any 3D clothes. The results showed that the optimized \"QR code\" pattern lowered the Average Precision (AP) of YOLOv3 by 87.7%, while the random \"QR code\" pattern and blank pattern lowered the AP of YOLOv3 by 57.9% and 30.1%, respectively, in the digital world. We then manufactured an adversarial shirt with a new material: aerogel. Physical-world experiments showed that the adversarial \"QR code\" pattern clothing lowered the AP of YOLOv3 by 64.6%, while the random \"QR code\" pattern clothing and fully heat-insulated clothing lowered the AP of YOLOv3 by 28.3% and 22.8%, respectively. We used the model ensemble technique to improve the attack transferability to unseen models."
                    },
                    {
                        "title": "On the Privacy Effect of Data Enhancement via the Lens of Memorization",
                        "abstract": "Machine learning poses severe privacy concerns as it has been shown that the learned models can reveal sensitive information about their training data. Many works have investigated the effect of widely adopted data augmentation and adversarial training techniques, termed data enhancement in the paper, on the privacy leakage of machine learning models. Such privacy effects are often measured by membership inference attacks (MIAs), which aim to identify whether a particular example belongs to the training set or not. We propose to investigate privacy from a new perspective called memorization. Through the lens of memorization, we find that previously deployed MIAs produce misleading results as they are less likely to identify samples with higher privacy risks as members compared to samples with low privacy risks. To solve this problem, we deploy a recent attack that can capture individual samples' memorization degrees for evaluation. Through extensive experiments, we unveil several findings about the connections between three essential properties of machine learning models, including privacy, generalization gap, and adversarial robustness. We demonstrate that the generalization gap and privacy leakage are less correlated than those of the previous results. Moreover, there is not necessarily a trade-off between adversarial robustness and privacy as stronger adversarial robustness does not make the model more susceptible to privacy attacks."
                    },
                    {
                        "title": "Infrared Adversarial Car Stickers",
                        "abstract": "Infrared physical adversarial examples are of great significance for studying the security of infrared AI systems that are widely used in our lives such as autonomous driving. Previous infrared physical attacks mainly focused on 2D infrared pedestrian detection which may not fully manifest its destructiveness to AI systems. In this work, we propose a physical attack method against infrared detectors based on 3D modeling, which is applied to a real car. The goal is to design a set of infrared adversarial stickers to make cars invisible to infrared detectors at various viewing angles, distances, and scenes. We build a 3D infrared car model with real infrared characteristics and propose an infrared adversarial pattern generation method based on 3D mesh shadow. We propose a 3D control points-based mesh smoothing algorithm and use a set of smoothness loss functions to enhance the smoothness of adversarial meshes and facilitate the sticker implementation. Besides, We designed the aluminum stickers and conducted physical experiments on two real Mercedes-Benz A200L cars. Our adversarial stickers hid the cars from Faster RCNN, an object detector, at various viewing angles, distances, and scenes. The attack success rate (ASR) was 91.49% for real cars. In comparison, the ASRs of random stickers and no sticker were only 6.21% and 0.66%, respectively. In addition, the ASRs of the designed stickers against six unseen object detectors such as YOLOv3 and Deformable DETR were between 73.35%-95.80%, showing good transferability of the attack performance across detectors."
                    },
                    {
                        "title": "Adversarial Texture for Fooling Person Detectors in the Physical World",
                        "abstract": "Nowadays, cameras equipped with AI systems can capture and analyze images to detect people automatically. However, the AI system can make mistakes when receiving deliberately designed patterns in the real world, i.e., physical adversarial examples. Prior works have shown that it is possible to print adversarial patches on clothes to evade DNN-based person detectors. However, these adversarial examples could have catastrophic drops in the attack success rate when the viewing angle (i.e., the camera's angle towards the object) changes. To perform a multi-angle attack, we propose Adversarial Texture (AdvTexture). AdvTexture can cover clothes with arbitrary shapes so that people wearing such clothes can hide from person detectors from different viewing angles. We propose a generative method, named Toroidal-Cropping-based Expandable Generative Attack (TC-EGA), to craft AdvTexture with repetitive structures. We printed several pieces of cloth with AdvTexure and then made T-shirts, skirts, and dresses in the physical world. Experiments showed that these clothes could fool person detectors in the physical world."
                    },
                    {
                        "title": "Physically Realizable Natural-Looking Clothing Textures Evade Person Detectors via 3D Modeling",
                        "abstract": "Recent works have proposed to craft adversarial clothes for evading person detectors, while they are either only effective at limited viewing angles or very conspicuous to humans. We aim to craft adversarial texture for clothes based on 3D modeling, an idea that has been used to craft rigid adversarial objects such as a 3D-printed turtle. Unlike rigid objects, humans and clothes are non-rigid, leading to difficulties in physical realization. In order to craft natural-looking adversarial clothes that can evade person detectors at multiple viewing angles, we propose adversarial camouflage textures (AdvCaT) that resemble one kind of the typical textures of daily clothes, camouflage textures. We leverage the Voronoi diagram and Gumbel-softmax trick to parameterize the camouflage textures and optimize the parameters via 3D modeling. Moreover, we propose an efficient augmentation pipeline on 3D meshes combining topologically plausible projection (TopoProj) and Thin Plate Spline (TPS) to narrow the gap between digital and real-world objects. We printed the developed 3D texture pieces on fabric materials and tailored them into T-shirts and trousers. Experiments show high attack success rates of these clothes against multiple detectors."
                    },
                    {
                        "title": "Language-Driven Anchors for Zero-Shot Adversarial Robustness",
                        "abstract": "Deep Neural Networks (DNNs) are known to be susceptible to adversarial attacks. Previous researches mainly focus on improving adversarial robustness in the fully supervised setting, leaving the challenging domain of zero-shot adversarial robustness an open question. In this work, we investigate this domain by leveraging the recent advances in large vision-language models, such as CLIP, to introduce zero-shot adversarial robustness to DNNs. We propose LAAT, a Language-driven, Anchor-based Adversarial Training strategy. LAAT utilizes the features of a text encoder for each category as fixed anchors (normalized feature embeddings) for each category, which are then employed for adversarial training. By leveraging the semantic consistency of the text encoders, LAAT aims to enhance the adversarial robustness of the image model on novel categories. However, naively using text encoders leads to poor results. Through analysis, we identified the issue to be the high cosine similarity between text encoders. We then design an expansion algorithm and an alignment cross-entropy loss to alleviate the problem. Our experimental results demonstrated that LAAT significantly improves zero-shot adversarial robustness over state-of-the-art methods. LAAT has the potential to enhance adversarial robustness by large-scale multimodal models, especially when labeled data is unavailable during training."
                    },
                    {
                        "title": "Perfect Gradient Inversion in Federated Learning: A New Paradigm from the Hidden Subset Sum Problem",
                        "abstract": "Federated Learning (FL) has emerged as a popular paradigm for collaborative learning among multiple parties. It is considered privacy-friendly because local data remains on personal devices, and only intermediate parameters -- such as gradients or model updates -- are shared. Although gradient inversion is widely viewed as a common attack method in FL, analytical research on reconstructing input training samples from shared gradients remains limited and is typically confined to constrained settings like small batch sizes. In this paper, we aim to overcome these limitations by addressing the problem from a cryptographic perspective. We mathematically formulate the input reconstruction problem using the gradient information shared in FL as the Hidden Subset Sum Problem (HSSP), an extension of the well-known NP-complete Subset Sum Problem (SSP). Leveraging this formulation allows us to achieve perfect input reconstruction, thereby mitigating issues such as dependence on label diversity and underperformance with large batch sizes that hinder existing empirical gradient inversion attacks. Moreover, our analysis provides insights into why empirical input reconstruction attacks degrade with larger batch sizes. By modeling the problem as HSSP, we demonstrate that the batch size \\( B \\) significantly affects attack complexity, with time complexity reaching \\( \\mathcal{O}(B^9) \\). We further show that applying secure data aggregation techniques -- such as homomorphic encryption and secure multiparty computation -- provides a strong defense by increasing the time complexity to \\( \\mathcal{O}(N^9 B^9) \\), where \\( N \\) is the number of local clients in FL. To the best of our knowledge, this is the first work to rigorously analyze privacy issues in FL by modeling them as HSSP, providing a concrete analytical foundation for further exploration and development of defense strategies."
                    }
                ]
            },
            "b6050a89-1429-4d3a-8d9a-4b88f9d29aa0": {
                "pk": "b6050a89-1429-4d3a-8d9a-4b88f9d29aa0",
                "name": "Julien Piet",
                "collaborators": [
                    "Chawin Sitawarin",
                    "David Wagner",
                    "Sizhe Chen",
                    "Jaiden Fairoze",
                    "Nicholas Weaver",
                    "Vivian Fang",
                    "Norman Mu",
                    "Samuel Pfrommer",
                    "Brendon G. Anderson",
                    "Somayeh Sojoudi"
                ],
                "domain": [
                    "Cryptocurrency",
                    "Large Language Models",
                    "Adversarial Robustness",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Extracting Godl [sic] from the Salt Mines: Ethereum Miners Extracting Value",
                        "abstract": "Cryptocurrency miners have great latitude in deciding which transactions they accept, including their own, and the order in which they accept them. Ethereum miners in particular use this flexibility to collect MEV-Miner Extractable Value-by structuring transactions to extract additional revenue. Ethereum also contains numerous bots that attempt to obtain MEV based on public-but-not-yet-confirmed transactions. Private relays shelter operations from these selfsame bots by directly submitting transactions to mining pools.   In this work, we develop an algorithm to detect MEV exploitation present in previously mined blocks. We use our implementation of the detector to analyze MEV usage and profit redistribution, finding that miners make the lion's share of the profits, rather than independent users of the private relays. More specifically, (i) 73% of private transactions hide trading activity or re-distribute miner rewards, and 87.6% of MEV collection is accomplished with privately submitted transactions, (ii) our algorithm finds more than $6M worth of MEV profit in a period of 12 days, two thirds of which go directly to miners, and (iii) MEV represents 9.2% of miners' profit from transaction fees.   Furthermore, in those 12 days, we also identify four blocks that contain enough MEV profits to make time-bandit forking attacks economically viable for large miners, undermining the security and stability of Ethereum as a whole."
                    },
                    {
                        "title": "Mark My Words: Analyzing and Evaluating Language Model Watermarks",
                        "abstract": "The capabilities of large language models have grown significantly in recent years and so too have concerns about their misuse. It is important to be able to distinguish machine-generated text from human-authored content. Prior works have proposed numerous schemes to watermark text, which would benefit from a systematic evaluation framework. This work focuses on LLM output watermarking techniques - as opposed to image or model watermarks - and proposes Mark My Words, a comprehensive benchmark for them under different natural language tasks. We focus on three main metrics: quality, size (i.e., the number of tokens needed to detect a watermark), and tamper resistance (i.e., the ability to detect a watermark after perturbing marked text). Current watermarking techniques are nearly practical enough for real-world use: Kirchenbauer et al. [33]'s scheme can watermark models like Llama 2 7B-chat or Mistral-7B-Instruct with no perceivable loss in quality on natural language tasks, the watermark can be detected with fewer than 100 tokens, and their scheme offers good tamper resistance to simple perturbations. However, they struggle to efficiently watermark code generations. We publicly release our benchmark (https://github.com/wagner-group/MarkMyWords)."
                    },
                    {
                        "title": "StruQ: Defending Against Prompt Injection with Structured Queries",
                        "abstract": "Recent advances in Large Language Models (LLMs) enable exciting LLM-integrated applications, which perform text-based tasks by utilizing their advanced language understanding capabilities. However, as LLMs have improved, so have the attacks against them. Prompt injection attacks are an important threat: they trick the model into deviating from the original application's instructions and instead follow user directives. These attacks rely on the LLM's ability to follow instructions and inability to separate prompts and user data. We introduce structured queries, a general approach to tackle this problem. Structured queries separate prompts and data into two channels. We implement a system that supports structured queries. This system is made of (1) a secure front-end that formats a prompt and user data into a special format, and (2) a specially trained LLM that can produce high-quality outputs from these inputs. The LLM is trained using a novel fine-tuning strategy: we convert a base (non-instruction-tuned) LLM to a structured instruction-tuned model that will only follow instructions in the prompt portion of a query. To do so, we augment standard instruction tuning datasets with examples that also include instructions in the data portion of the query, and fine-tune the model to ignore these. Our system significantly improves resistance to prompt injection attacks, with little or no impact on utility. Our code is released at https://github.com/Sizhe-Chen/StruQ."
                    },
                    {
                        "title": "Asymmetric Certified Robustness via Feature-Convex Neural Networks",
                        "abstract": "Recent works have introduced input-convex neural networks (ICNNs) as learning models with advantageous training, inference, and generalization properties linked to their convex structure. In this paper, we propose a novel feature-convex neural network architecture as the composition of an ICNN with a Lipschitz feature map in order to achieve adversarial robustness. We consider the asymmetric binary classification setting with one \"sensitive\" class, and for this class we prove deterministic, closed-form, and easily-computable certified robust radii for arbitrary $\\ell_p$-norms. We theoretically justify the use of these models by characterizing their decision region geometry, extending the universal approximation theorem for ICNN regression to the classification setting, and proving a lower bound on the probability that such models perfectly fit even unstructured uniformly distributed data in sufficiently high dimensions. Experiments on Malimg malware classification and subsets of MNIST, Fashion-MNIST, and CIFAR-10 datasets show that feature-convex classifiers attain state-of-the-art certified $\\ell_1$-radii as well as substantial $\\ell_2$- and $\\ell_{\\infty}$-radii while being far more computationally efficient than any competitive baseline."
                    },
                    {
                        "title": "Jatmo: Prompt Injection Defense by Task-Specific Finetuning",
                        "abstract": "Large Language Models (LLMs) are attracting significant research attention due to their instruction-following abilities, allowing users and developers to leverage LLMs for a variety of tasks. However, LLMs are vulnerable to prompt-injection attacks: a class of attacks that hijack the model's instruction-following abilities, changing responses to prompts to undesired, possibly malicious ones. In this work, we introduce Jatmo, a method for generating task-specific models resilient to prompt-injection attacks. Jatmo leverages the fact that LLMs can only follow instructions once they have undergone instruction tuning. It harnesses a teacher instruction-tuned model to generate a task-specific dataset, which is then used to fine-tune a base model (i.e., a non-instruction-tuned model). Jatmo only needs a task prompt and a dataset of inputs for the task: it uses the teacher model to generate outputs. For situations with no pre-existing datasets, Jatmo can use a single example, or in some cases none at all, to produce a fully synthetic dataset. Our experiments on seven tasks show that Jatmo models provide similar quality of outputs on their specific task as standard LLMs, while being resilient to prompt injections. The best attacks succeeded in less than 0.5% of cases against our models, versus 87% success rate against GPT-3.5-Turbo. We release Jatmo at https://github.com/wagner-group/prompt-injection-defense."
                    }
                ]
            },
            "8cdb8d37-27d0-4e1c-b625-6d8448c5f46f": {
                "pk": "8cdb8d37-27d0-4e1c-b625-6d8448c5f46f",
                "name": "Geng Zhao",
                "collaborators": [
                    "Itai Ashlagi",
                    "Mark Braverman",
                    "Christian Borgs",
                    "Amin Saberi",
                    "Clayton Thomas",
                    "Banghua Zhu",
                    "Jiantao Jiao",
                    "Michael I. Jordan",
                    "Hannah Li",
                    "Ramesh Johari"
                ],
                "domain": [
                    "Random Graphs",
                    "Market Design",
                    "Online Learning",
                    "Game Theory"
                ],
                "publications": [
                    {
                        "title": "Locality via Global Ties: Stability of the 2-Core Against Misspecification",
                        "abstract": "For many random graph models, the analysis of a related birth process suggests local sampling algorithms for the size of, e.g., the giant connected component, the $k$-core, the size and probability of an epidemic outbreak, etc. In this paper, we study the question of when these algorithms are robust against misspecification of the graph model, for the special case of the 2-core. We show that, for locally converging graphs with bounded average degrees, under a weak notion of expansion, a local sampling algorithm provides robust estimates for the size of both the 2-core and its largest component. Our weak notion of expansion generalizes the classical definition of expansion, while holding for many well-studied random graph models.   Our method involves a two-step sprinkling argument. In the first step, we use sprinkling to establish the existence of a non-empty $2$-core inside the giant, while in the second, we use this non-empty $2$-core as seed for a second sprinkling argument to establish that the giant contains a linear sized $2$-core. The second step is based on a novel coloring scheme for the vertices in the tree-part. Our algorithmic results follow from the structural properties for the $2$-core established in the course of our sprinkling arguments.   The run-time of our local algorithm is constant independent of the graph size, with the value of the constant depending on the desired asymptotic accuracy $\\epsilon$. But given the existential nature of local limits, our arguments do not give any bound on the functional dependence of this constant on $\\epsilon$, nor do they give a bound on how large the graph has to be for the asymptotic additive error bound $\\epsilon$ to hold."
                    },
                    {
                        "title": "Welfare Distribution in Two-sided Random Matching Markets",
                        "abstract": "We study the welfare structure in two-sided large random matching markets. In the model, each agent has a latent personal score for every agent on the other side of the market and her preferences follow a logit model based on these scores. Under a contiguity condition, we provide a tight description of stable outcomes. First, we identify an intrinsic fitness for each agent that represents her relative competitiveness in the market, independent of the realized stable outcome. The intrinsic fitness values correspond to scaling coefficients needed to make a latent mutual matrix bi-stochastic, where the latent scores can be interpreted as a-priori probabilities of a pair being matched. Second, in every stable (or even approximately stable) matching, the welfare or the ranks of the agents on each side of the market, when scaled by their intrinsic fitness, have an approximately exponential empirical distribution. Moreover, the average welfare of agents on one side of the market is sufficient to determine the average on the other side. Overall, each agent's welfare is determined by a global parameter, her intrinsic fitness, and an extrinsic factor with exponential distribution across the population."
                    },
                    {
                        "title": "Tiered Random Matching Markets: Rank is Proportional to Popularity",
                        "abstract": "We study the stable marriage problem in two-sided markets with randomly generated preferences. We consider agents on each side divided into a constant number of \"soft tiers\", which intuitively indicate the quality of the agent. Specifically, every agent within a tier has the same public score, and agents on each side have preferences independently generated proportionally to the public scores of the other side.   We compute the expected average rank which agents in each tier have for their partners in the men-optimal stable matching, and prove concentration results for the average rank in asymptotically large markets. Furthermore, we show that despite having a significant effect on ranks, public scores do not strongly influence the probability of an agent matching to a given tier of the other side. This generalizes results of [Pittel 1989] which correspond to uniform preferences. The results quantitatively demonstrate the effect of competition due to the heterogeneous attractiveness of agents in the market, and we give the first explicit calculations of rank beyond uniform markets."
                    },
                    {
                        "title": "Online Learning in Stackelberg Games with an Omniscient Follower",
                        "abstract": "We study the problem of online learning in a two-player decentralized cooperative Stackelberg game. In each round, the leader first takes an action, followed by the follower who takes their action after observing the leader's move. The goal of the leader is to learn to minimize the cumulative regret based on the history of interactions. Differing from the traditional formulation of repeated Stackelberg games, we assume the follower is omniscient, with full knowledge of the true reward, and that they always best-respond to the leader's actions. We analyze the sample complexity of regret minimization in this repeated Stackelberg game. We show that depending on the reward structure, the existence of the omniscient follower may change the sample complexity drastically, from constant to exponential, even for linear cooperative Stackelberg games. This poses unique challenges for the learning process of the leader and the subsequent regret analysis."
                    },
                    {
                        "title": "Interference, Bias, and Variance in Two-Sided Marketplace Experimentation: Guidance for Platforms",
                        "abstract": "Two-sided marketplace platforms often run experiments to test the effect of an intervention before launching it platform-wide. A typical approach is to randomize individuals into the treatment group, which receives the intervention, and the control group, which does not. The platform then compares the performance in the two groups to estimate the effect if the intervention were launched to everyone. We focus on two common experiment types, where the platform randomizes individuals either on the supply side or on the demand side. The resulting estimates of the treatment effect in these experiments are typically biased: because individuals in the market compete with each other, individuals in the treatment group affect those in the control group and vice versa, creating interference. We develop a simple tractable market model to study bias and variance in these experiments with interference. We focus on two choices available to the platform: (1) Which side of the platform should it randomize on (supply or demand)? (2) What proportion of individuals should be allocated to treatment? We find that both choices affect the bias and variance of the resulting estimators but in different ways. The bias-optimal choice of experiment type depends on the relative amounts of supply and demand in the market, and we discuss how a platform can use market data to select the experiment type. Importantly, we find in many circumstances, choosing the bias-optimal experiment type has little effect on variance. On the other hand, the choice of treatment proportion can induce a bias-variance tradeoff, where the bias-minimizing proportion increases variance. We discuss how a platform can navigate this tradeoff and best choose the treatment proportion, using a combination of modeling as well as contextual knowledge about the market, the risk of the intervention, and reasonable effect sizes of the intervention."
                    }
                ]
            },
            "c0d4f4f5-6afa-4345-a664-1d672397e796": {
                "pk": "c0d4f4f5-6afa-4345-a664-1d672397e796",
                "name": "Jiantao Jiao",
                "collaborators": [
                    "Yanjun Han",
                    "Tsachy Weissman",
                    "Kartik Venkat",
                    "Thomas Courtade",
                    "Weihao Gao",
                    "Banghua Zhu",
                    "Jacob Steinhardt",
                    "Lin Zhang",
                    "Robert Nowak",
                    "Rajarshi Mukherjee"
                ],
                "domain": [
                    "Statistical Estimation",
                    "Information Theory",
                    "Machine Learning",
                    "Bias Analysis"
                ],
                "publications": [
                    {
                        "title": "Bias Correction with Jackknife, Bootstrap, and Taylor Series",
                        "abstract": "We analyze bias correction methods using jackknife, bootstrap, and Taylor series. We focus on the binomial model, and consider the problem of bias correction for estimating $f(p)$, where $f \\in C[0,1]$ is arbitrary. We characterize the supremum norm of the bias of general jackknife and bootstrap estimators for any continuous functions, and demonstrate the in delete-$d$ jackknife, different values of $d$ may lead to drastically different behaviors in jackknife. We show that in the binomial model, iterating the bootstrap bias correction infinitely many times may lead to divergence of bias and variance, and demonstrate that the bias properties of the bootstrap bias corrected estimator after $r-1$ rounds are of the same order as that of the $r$-jackknife estimator if a bounded coefficients condition is satisfied."
                    },
                    {
                        "title": "An Extremal Inequality for Long Markov Chains",
                        "abstract": "Let $X,Y$ be jointly Gaussian vectors, and consider random variables $U,V$ that satisfy the Markov constraint $U-X-Y-V$. We prove an extremal inequality relating the mutual informations between all ${4 \\choose 2}$ pairs of random variables from the set $(U,X,Y,V)$. As a first application, we show that the rate region for the two-encoder quadratic Gaussian source coding problem follows as an immediate corollary of the the extremal inequality. In a second application, we establish the rate region for a vector-Gaussian source coding problem where L\\\"{o}wner-John ellipsoids are approximated based on rate-constrained descriptions of the data."
                    },
                    {
                        "title": "Adaptive Estimation of Shannon Entropy",
                        "abstract": "We consider estimating the Shannon entropy of a discrete distribution $P$ from $n$ i.i.d. samples. Recently, Jiao, Venkat, Han, and Weissman, and Wu and Yang constructed approximation theoretic estimators that achieve the minimax $L_2$ rates in estimating entropy. Their estimators are consistent given $n \\gg \\frac{S}{\\ln S}$ samples, where $S$ is the alphabet size, and it is the best possible sample complexity. In contrast, the Maximum Likelihood Estimator (MLE), which is the empirical entropy, requires $n\\gg S$ samples.   In the present paper we significantly refine the minimax results of existing work. To alleviate the pessimism of minimaxity, we adopt the adaptive estimation framework, and show that the minimax rate-optimal estimator in Jiao, Venkat, Han, and Weissman achieves the minimax rates simultaneously over a nested sequence of subsets of distributions $P$, without knowing the alphabet size $S$ or which subset $P$ lies in. In other words, their estimator is adaptive with respect to this nested sequence of the parameter space, which is characterized by the entropy of the distribution. We also characterize the maximum risk of the MLE over this nested sequence, and show, for every subset in the sequence, that the performance of the minimax rate-optimal estimator with $n$ samples is essentially that of the MLE with $n\\ln n$ samples, thereby further substantiating the generality of the phenomenon identified by Jiao, Venkat, Han, and Weissman."
                    },
                    {
                        "title": "Generalizations of Maximal Inequalities to Arbitrary Selection Rules",
                        "abstract": "We present a generalization of the maximal inequalities that upper bound the expectation of the maximum of $n$ jointly distributed random variables. We control the expectation of a randomly selected random variable from $n$ jointly distributed random variables, and present bounds that are at least as tight as the classical maximal inequalities, and much tighter when the distribution of selection index is near deterministic. A new family of information theoretic measures were introduced in the process, which may be of independent interest."
                    },
                    {
                        "title": "Dependence Measures Bounding the Exploration Bias for General Measurements",
                        "abstract": "We propose a framework to analyze and quantify the bias in adaptive data analysis. It generalizes that proposed by Russo and Zou'15, applying to measurements whose moment generating function exists, measurements with a finite $p$-norm, and measurements in general Orlicz spaces. We introduce a new class of dependence measures which retain key properties of mutual information while more effectively quantifying the exploration bias for heavy tailed distributions. We provide examples of cases where our bounds are nearly tight in situations where the original framework of Russo and Zou'15 does not apply."
                    },
                    {
                        "title": "Does Dirichlet Prior Smoothing Solve the Shannon Entropy Estimation Problem?",
                        "abstract": "The Dirichlet prior is widely used in estimating discrete distributions and functionals of discrete distributions. In terms of Shannon entropy estimation, one approach is to plug-in the Dirichlet prior smoothed distribution into the entropy functional, while the other one is to calculate the Bayes estimator for entropy under the Dirichlet prior for squared error, which is the conditional expectation. We show that in general they do \\emph{not} improve over the maximum likelihood estimator, which plugs-in the empirical distribution into the entropy functional. No matter how we tune the parameters in the Dirichlet prior, this approach cannot achieve the minimax rates in entropy estimation, as recently characterized by Jiao, Venkat, Han, and Weissman, and Wu and Yang. The performance of the minimax rate-optimal estimator with $n$ samples is essentially \\emph{at least} as good as that of the Dirichlet smoothed entropy estimators with $n\\ln n$ samples.   We harness the theory of approximation using positive linear operators for analyzing the bias of plug-in estimators for general functionals under arbitrary statistical models, thereby further consolidating the interplay between these two fields, which was thoroughly developed and exploited by Jiao, Venkat, Han, and Weissman. We establish new results in approximation theory, and apply them to analyze the bias of the Dirichlet prior smoothed plug-in entropy estimator. This interplay between bias analysis and approximation theory is of relevance and consequence far beyond the specific problem setting in this paper."
                    },
                    {
                        "title": "Minimax Estimation of the $L_1$ Distance",
                        "abstract": "We consider the problem of estimating the $L_1$ distance between two discrete probability measures $P$ and $Q$ from empirical data in a nonasymptotic and large alphabet setting. When $Q$ is known and one obtains $n$ samples from $P$, we show that for every $Q$, the minimax rate-optimal estimator with $n$ samples achieves performance comparable to that of the maximum likelihood estimator (MLE) with $n\\ln n$ samples. When both $P$ and $Q$ are unknown, we construct minimax rate-optimal estimators whose worst case performance is essentially that of the known $Q$ case with $Q$ being uniform, implying that $Q$ being uniform is essentially the most difficult case. The \\emph{effective sample size enlargement} phenomenon, identified in Jiao \\emph{et al.} (2015), holds both in the known $Q$ case for every $Q$ and the $Q$ unknown case. However, the construction of optimal estimators for $\\|P-Q\\|_1$ requires new techniques and insights beyond the approximation-based method of functional estimation in Jiao \\emph{et al.} (2015)."
                    },
                    {
                        "title": "Relations between Information and Estimation in Discrete-Time L\u00e9vy Channels",
                        "abstract": "Fundamental relations between information and estimation have been established in the literature for the discrete-time Gaussian and Poisson channels. In this work, we demonstrate that such relations hold for a much larger class of observation models. We introduce the natural family of discrete-time L\\'evy channels where the distribution of the output conditioned on the input is infinitely divisible. For L\\'evy channels, we establish new representations relating the mutual information between the channel input and output to an optimal expected estimation loss, thereby unifying and considerably extending results from the Gaussian and Poisson settings. We demonstrate the richness of our results by working out two examples of L\\'evy channels, namely the gamma channel and the negative binomial channel, with corresponding relations between information and estimation. Extensions to the setting of mismatched estimation are also presented."
                    },
                    {
                        "title": "The Nearest Neighbor Information Estimator is Adaptively Near Minimax Rate-Optimal",
                        "abstract": "We analyze the Kozachenko--Leonenko (KL) nearest neighbor estimator for the differential entropy. We obtain the first uniform upper bound on its performance over H\\\"older balls on a torus without assuming any conditions on how close the density could be from zero. Accompanying a new minimax lower bound over the H\\\"older ball, we show that the KL estimator is achieving the minimax rates up to logarithmic factors without cognizance of the smoothness parameter $s$ of the H\\\"older ball for $s\\in (0,2]$ and arbitrary dimension $d$, rendering it the first estimator that provably satisfies this property."
                    },
                    {
                        "title": "When does the Tukey median work?",
                        "abstract": "We analyze the performance of the Tukey median estimator under total variation (TV) distance corruptions. Previous results show that under Huber's additive corruption model, the breakdown point is 1/3 for high-dimensional halfspace-symmetric distributions. We show that under TV corruptions, the breakdown point reduces to 1/4 for the same set of distributions. We also show that a certain projection algorithm can attain the optimal breakdown point of 1/2. Both the Tukey median estimator and the projection algorithm achieve sample complexity linear in dimension."
                    },
                    {
                        "title": "Local moment matching: A unified methodology for symmetric functional estimation and distribution estimation under Wasserstein distance",
                        "abstract": "We present \\emph{Local Moment Matching (LMM)}, a unified methodology for symmetric functional estimation and distribution estimation under Wasserstein distance. We construct an efficiently computable estimator that achieves the minimax rates in estimating the distribution up to permutation, and show that the plug-in approach of our unlabeled distribution estimator is \"universal\" in estimating symmetric functionals of discrete distributions. Instead of doing best polynomial approximation explicitly as in existing literature of functional estimation, the plug-in approach conducts polynomial approximation implicitly and attains the optimal sample complexity for the entropy, power sum and support size functionals."
                    },
                    {
                        "title": "Beyond Maximum Likelihood: from Theory to Practice",
                        "abstract": "Maximum likelihood is the most widely used statistical estimation technique. Recent work by the authors introduced a general methodology for the construction of estimators for functionals in parametric models, and demonstrated improvements - both in theory and in practice - over the maximum likelihood estimator (MLE), particularly in high dimensional scenarios involving parameter dimension comparable to or larger than the number of samples. This approach to estimation, building on results from approximation theory, is shown to yield minimax rate-optimal estimators for a wide class of functionals, implementable with modest computational requirements. In a nutshell, a message of this recent work is that, for a wide class of functionals, the performance of these essentially optimal estimators with $n$ samples is comparable to that of the MLE with $n \\ln n$ samples.   In the present paper, we highlight the applicability of the aforementioned methodology to statistical problems beyond functional estimation, and show that it can yield substantial gains. For example, we demonstrate that for learning tree-structured graphical models, our approach achieves a significant reduction of the required data size compared with the classical Chow--Liu algorithm, which is an implementation of the MLE, to achieve the same accuracy. The key step in improving the Chow--Liu algorithm is to replace the empirical mutual information with the estimator for mutual information proposed by the authors. Further, applying the same replacement approach to classical Bayesian network classification, the resulting classifiers uniformly outperform the previous classifiers on 26 widely used datasets."
                    },
                    {
                        "title": "Mutual Information, Relative Entropy and Estimation Error in Semi-martingale Channels",
                        "abstract": "Fundamental relations between information and estimation have been established in the literature for the continuous-time Gaussian and Poisson channels, in a long line of work starting from the classical representation theorems by Duncan and Kabanov respectively. In this work, we demonstrate that such relations hold for a much larger family of continuous-time channels. We introduce the family of semi-martingale channels where the channel output is a semi-martingale stochastic process, and the channel input modulates the characteristics of the semi-martingale. For these channels, which includes as a special case the continuous time Gaussian and Poisson models, we establish new representations relating the mutual information between the channel input and output to an optimal causal filtering loss, thereby unifying and considerably extending results from the Gaussian and Poisson settings. Extensions to the setting of mismatched estimation are also presented where the relative entropy between the laws governing the output of the channel under two different input distributions is equal to the cumulative difference between the estimation loss incurred by using the mismatched and optimal causal filters respectively. The main tool underlying these results is the Doob--Meyer decomposition of a class of likelihood ratio sub-martingales. The results in this work can be viewed as the continuous-time analogues of recent generalizations for relations between information and estimation for discrete-time L\\'evy channels."
                    },
                    {
                        "title": "Minimax-Optimal Bounds for Detectors Based on Estimated Prior Probabilities",
                        "abstract": "In many signal detection and classification problems, we have knowledge of the distribution under each hypothesis, but not the prior probabilities. This paper is aimed at providing theory to quantify the performance of detection via estimating prior probabilities from either labeled or unlabeled training data. The error or {\\em risk} is considered as a function of the prior probabilities. We show that the risk function is locally Lipschitz in the vicinity of the true prior probabilities, and the error of detectors based on estimated prior probabilities depends on the behavior of the risk function in this locality. In general, we show that the error of detectors based on the Maximum Likelihood Estimate (MLE) of the prior probabilities converges to the Bayes error at a rate of $n^{-1/2}$, where $n$ is the number of training data. If the behavior of the risk function is more favorable, then detectors based on the MLE have errors converging to the corresponding Bayes errors at optimal rates of the form $n^{-(1+\\alpha)/2}$, where $\\alpha>0$ is a parameter governing the behavior of the risk function with a typical value $\\alpha = 1$. The limit $\\alpha \\rightarrow \\infty$ corresponds to a situation where the risk function is flat near the true probabilities, and thus insensitive to small errors in the MLE; in this case the error of the detector based on the MLE converges to the Bayes error exponentially fast with $n$. We show the bounds are achievable no matter given labeled or unlabeled training data and are minimax-optimal in labeled case."
                    },
                    {
                        "title": "Minimax Estimation of Functionals of Discrete Distributions",
                        "abstract": "We propose a general methodology for the construction and analysis of minimax estimators for a wide class of functionals of finite dimensional parameters, and elaborate on the case of discrete distributions, where the alphabet size $S$ is unknown and may be comparable with the number of observations $n$. We treat the respective regions where the functional is \"nonsmooth\" and \"smooth\" separately. In the \"nonsmooth\" regime, we apply an unbiased estimator for the best polynomial approximation of the functional whereas, in the \"smooth\" regime, we apply a bias-corrected Maximum Likelihood Estimator (MLE). We illustrate the merit of this approach by thoroughly analyzing two important cases: the entropy $H(P) = \\sum_{i = 1}^S -p_i \\ln p_i$ and $F_\\alpha(P) = \\sum_{i = 1}^S p_i^\\alpha,\\alpha>0$. We obtain the minimax $L_2$ rates for estimating these functionals. In particular, we demonstrate that our estimator achieves the optimal sample complexity $n \\asymp S/\\ln S$ for entropy estimation. We also show that the sample complexity for estimating $F_\\alpha(P),0<\\alpha<1$ is $n\\asymp S^{1/\\alpha}/ \\ln S$, which can be achieved by our estimator but not the MLE. For $1<\\alpha<3/2$, we show the minimax $L_2$ rate for estimating $F_\\alpha(P)$ is $(n\\ln n)^{-2(\\alpha-1)}$ regardless of the alphabet size, while the $L_2$ rate for the MLE is $n^{-2(\\alpha-1)}$. For all the above cases, the behavior of the minimax rate-optimal estimators with $n$ samples is essentially that of the MLE with $n\\ln n$ samples. We highlight the practical advantages of our schemes for entropy and mutual information estimation. We demonstrate that our approach reduces running time and boosts the accuracy compared to existing various approaches. Moreover, we show that the mutual information estimator induced by our methodology leads to significant performance boosts over the Chow--Liu algorithm in learning graphical models."
                    },
                    {
                        "title": "Maximum Likelihood Estimation of Functionals of Discrete Distributions",
                        "abstract": "We consider the problem of estimating functionals of discrete distributions, and focus on tight nonasymptotic analysis of the worst case squared error risk of widely used estimators. We apply concentration inequalities to analyze the random fluctuation of these estimators around their expectations, and the theory of approximation using positive linear operators to analyze the deviation of their expectations from the true functional, namely their \\emph{bias}.   We characterize the worst case squared error risk incurred by the Maximum Likelihood Estimator (MLE) in estimating the Shannon entropy $H(P) = \\sum_{i = 1}^S -p_i \\ln p_i$, and $F_\\alpha(P) = \\sum_{i = 1}^S p_i^\\alpha,\\alpha>0$, up to multiplicative constants, for any alphabet size $S\\leq \\infty$ and sample size $n$ for which the risk may vanish. As a corollary, for Shannon entropy estimation, we show that it is necessary and sufficient to have $n \\gg S$ observations for the MLE to be consistent. In addition, we establish that it is necessary and sufficient to consider $n \\gg S^{1/\\alpha}$ samples for the MLE to consistently estimate $F_\\alpha(P), 0<\\alpha<1$. The minimax rate-optimal estimators for both problems require $S/\\ln S$ and $S^{1/\\alpha}/\\ln S$ samples, which implies that the MLE has a strictly sub-optimal sample complexity. When $1<\\alpha<3/2$, we show that the worst-case squared error rate of convergence for the MLE is $n^{-2(\\alpha-1)}$ for infinite alphabet size, while the minimax squared error rate is $(n\\ln n)^{-2(\\alpha-1)}$. When $\\alpha\\geq 3/2$, the MLE achieves the minimax optimal rate $n^{-1}$ regardless of the alphabet size.   As an application of the general theory, we analyze the Dirichlet prior smoothing techniques for Shannon entropy estimation. We show that no matter how we tune the parameters in the Dirichlet prior, this technique cannot achieve the minimax rates in entropy estimation."
                    },
                    {
                        "title": "Minimax Estimation of Discrete Distributions under $\\ell_1$ Loss",
                        "abstract": "We analyze the problem of discrete distribution estimation under $\\ell_1$ loss. We provide non-asymptotic upper and lower bounds on the maximum risk of the empirical distribution (the maximum likelihood estimator), and the minimax risk in regimes where the alphabet size $S$ may grow with the number of observations $n$. We show that among distributions with bounded entropy $H$, the asymptotic maximum risk for the empirical distribution is $2H/\\ln n$, while the asymptotic minimax risk is $H/\\ln n$. Moreover, Moreover, we show that a hard-thresholding estimator oblivious to the unknown upper bound $H$, is asymptotically minimax. However, if we constrain the estimates to lie in the simplex of probability distributions, then the asymptotic minimax risk is again $2H/\\ln n$. We draw connections between our work and the literature on density estimation, entropy estimation, total variation distance ($\\ell_1$ divergence) estimation, joint distribution estimation in stochastic processes, normal mean estimation, and adaptive estimation."
                    },
                    {
                        "title": "Minimax Rate-Optimal Estimation of Divergences between Discrete Distributions",
                        "abstract": "We study the minimax estimation of $\\alpha$-divergences between discrete distributions for integer $\\alpha\\ge 1$, which include the Kullback--Leibler divergence and the $\\chi^2$-divergences as special examples. Dropping the usual theoretical tricks to acquire independence, we construct the first minimax rate-optimal estimator which does not require any Poissonization, sample splitting, or explicit construction of approximating polynomials. The estimator uses a hybrid approach which solves a problem-independent linear program based on moment matching in the non-smooth regime, and applies a problem-dependent bias-corrected plug-in estimator in the smooth regime, with a soft decision boundary between these regimes."
                    },
                    {
                        "title": "On Estimation of $L_{r}$-Norms in Gaussian White Noise Models",
                        "abstract": "We provide a complete picture of asymptotically minimax estimation of $L_r$-norms (for any $r\\ge 1$) of the mean in Gaussian white noise model over Nikolskii-Besov spaces. In this regard, we complement the work of Lepski, Nemirovski and Spokoiny (1999), who considered the cases of $r=1$ (with poly-logarithmic gap between upper and lower bounds) and $r$ even (with asymptotically sharp upper and lower bounds) over H\\\"{o}lder spaces. We additionally consider the case of asymptotically adaptive minimax estimation and demonstrate a difference between even and non-even $r$ in terms of an investigator's ability to produce asymptotically adaptive minimax estimators without paying a penalty."
                    },
                    {
                        "title": "Approximate Profile Maximum Likelihood",
                        "abstract": "We propose an efficient algorithm for approximate computation of the profile maximum likelihood (PML), a variant of maximum likelihood maximizing the probability of observing a sufficient statistic rather than the empirical sample. The PML has appealing theoretical properties, but is difficult to compute exactly. Inspired by observations gleaned from exactly solvable cases, we look for an approximate PML solution, which, intuitively, clumps comparably frequent symbols into one symbol. This amounts to lower-bounding a certain matrix permanent by summing over a subgroup of the symmetric group rather than the whole group during the computation. We extensively experiment with the approximate solution, and find the empirical performance of our approach is competitive and sometimes significantly better than state-of-the-art performance for various estimation problems."
                    }
                ]
            },
            "37751769-7c19-4baf-a69c-f8de04f1ab13": {
                "pk": "37751769-7c19-4baf-a69c-f8de04f1ab13",
                "name": "David Wagner",
                "collaborators": [
                    "Nicholas Carlini",
                    "Nora Weickgenannt",
                    "Dirk H. Rischke"
                ],
                "domain": [
                    "Quantum Hydrodynamics",
                    "Adversarial Machine Learning",
                    "Spin Dynamics",
                    "Nonlocal Collision Theory"
                ],
                "publications": [
                    {
                        "title": "Resummed spin hydrodynamics from quantum kinetic theory",
                        "abstract": "In this work, the equations of dissipative relativistic spin hydrodynamics based on quantum kinetic theory are derived. Employing the inverse-Reynolds dominance (IReD) approach, a resummation scheme based on a power counting in Knudsen and inverse Reynolds numbers is constructed, leading to hydrodynamic equations that are accurate to second order. It is found that the spin dynamics can be characterized by eleven equations: six of them describe the evolution of the components of the spin potential, while the remaining five provide the equation of motion of a dissipative irreducible rank-two tensor. For a simple truncation, the first- and second-order transport coefficients are computed explicitly."
                    },
                    {
                        "title": "Defensive Distillation is Not Robust to Adversarial Examples",
                        "abstract": "We show that defensive distillation is not secure: it is no more resistant to targeted misclassification attacks than unprotected neural networks."
                    },
                    {
                        "title": "MagNet and \"Efficient Defenses Against Adversarial Attacks\" are Not Robust to Adversarial Examples",
                        "abstract": "MagNet and \"Efficient Defenses...\" were recently proposed as a defense to adversarial examples. We find that we can construct adversarial examples that defeat these defenses with only a slight increase in distortion."
                    },
                    {
                        "title": "Lorentz-covariant nonlocal collision term for spin-1/2 particles",
                        "abstract": "We revisit the derivation of the nonlocal collision term in the Boltzmann equation for spin-1/2 particles, using both the Wigner-function approach by de Groot, van Leeuwen, and van Weert, and the Kadanoff-Baym equation in $T$-matrix approximation. Contrary to previous calculations, our results maintain full Lorentz covariance of the nonlocal collision term."
                    }
                ]
            }
        }
    },
    "2309.08045": {
        "paper_data": {
            "title": "Traveling Waves Encode the Recent Past and Enhance Sequence Learning",
            "url": "http://arxiv.org/abs/2309.08045v2",
            "arxiv_id": "2309.08045",
            "authors": [
                "T. Anderson Keller",
                "Lyle Muller",
                "Terrence Sejnowski",
                "Max Welling"
            ],
            "abstract": "Traveling waves of neural activity have been observed throughout the brain at a diversity of regions and scales; however, their precise computational role is still debated. One physically inspired hypothesis suggests that the cortical sheet may act like a wave-propagating system capable of invertibly storing a short-term memory of sequential stimuli through induced waves traveling across the cortical surface, and indeed many experimental results from neuroscience correlate wave activity with memory tasks. To date, however, the computational implications of this idea have remained hypothetical due to the lack of a simple recurrent neural network architecture capable of exhibiting such waves. In this work, we introduce a model to fill this gap, which we denote the Wave-RNN (wRNN), and demonstrate how such an architecture indeed efficiently encodes the recent past through a suite of synthetic memory tasks where wRNNs learn faster and reach significantly lower error than wave-free counterparts. We further explore the implications of this memory storage system on more complex sequence modeling tasks such as sequential image classification and find that wave-based models not only again outperform comparable wave-free RNNs while using significantly fewer parameters, but additionally perform comparably to more complex gated architectures such as LSTMs and GRUs.",
            "introduction": "   1 Introduction   Since the earliest neural recordings (Caton, 1875; Beck, 1890), neural oscillations and the spatial organization of neural activity have persisted as topics of great interest in the neuroscience community. Consequently, a plethora of hypothesized potential functions for these widely observed phenomena have been put forth in the literature. For example: Pitts & McCulloch (1947) propose that alpha oscillations perform \u2018cortical scanning\u2019 akin to radar; Milner (1974) suggests that synchrony may serve to segregate the visual scene and \u2018bind\u2019 individual features into cohesive objects; (Engel et\u00a0al., 2001) suggest brain-wide oscillations serve as \u2018top-down\u2019 stored knowledge and contextual influence for local processing; Raghavachari et\u00a0al. (2001) show theta is consistent with a gating mechanism for human working memory; Buzs\u00e1ki et\u00a0al. (2013) posit that oscillations form a hierarchical system that offers a syntactical structure for spike traffic; and Liebe et\u00a0al. (2012) & de Mooij-van Malsen et\u00a0al. (2023) implicate oscillatory coherence with the transfer of information on working memory tasks.   Recently, the advents of high density multi-electrode arrays and high resolution imaging have led to the discovery that many of the oscillations observed in the brain are better described as traveling waves of activity rather than precise zero-lag synchrony (Muller et\u00a0al., 2018). For example, alpha and theta oscillations have been measured to precisely correspond to traveling waves in both the cortex and hippocampus of humans (Lubenov & Siapas, 2009; Lozano-Soldevilla & VanRullen, 2019; Zhang et\u00a0al., 2018). Furthermore, it has become increasingly clear that wave-like dynamics are prevalent throughout the brain, from local (Muller et\u00a0al., 2014) to global (Muller et\u00a0al., 2016) scales, and across virtually all brain regions measured (Townsend et\u00a0al., 2015; Gu et\u00a0al., 2021b; Pang et\u00a0al., 2023). Research has correlated wave activity with a range of functional roles rivaling oscillations in diversity and number, including: perceptual awareness (Davis et\u00a0al., 2020); attentional scanning (Fries, 2023); information transfer (Rubino et\u00a0al., 2006); motor control sequencing and topographic organization (Takahashi et\u00a0al., 2011); integrating visual sensory signals with saccadic eye movements (Zanos et\u00a0al., 2015); and coordinating and reconfiguring distinct functional regions (Xu et\u00a0al., 2023).   Most relevant to the present study, significant work has shown correlations between wave dynamics and memory. For example: King & Wyart (2021) provide evidence that traveling waves propagating from posterior to anterior cortex serve to encode sequential stimuli in an overlapping \u2018multiplexed\u2019 manner; Sauseng et\u00a0al. (2002) show that traveling theta oscillations propagate from anterior to posterior regions during retrieval from long-term memory; and Zabeh et\u00a0al. (2023) show traveling beta waves in frontal and parietal lobes encode memory of recent rewards. As described by Muller et\u00a0al. (2018), one way to understand the relation between traveling waves and memory comes from an analogy to physical wave systems. Specifically, non-dissipative wave-propagating systems can be shown to contain all information about past disturbances in a \u2018time-reversible\u2019 manner. In contrast, in a static wave-free \u2018bump\u2019 system, onset time of stimuli are tied to position, and therefore information about the sequential order of multiple inputs at the same position is lost. This is visualized in Figure 1, and is demonstrated experimentally in Section 3. Although an imperfect analogy to the brain, such ideas have inspired neuroscientists to suggest that",
            "references": [
                {
                    "title": "Resurrecting Recurrent Neural Networks for Long Sequences",
                    "abstract": "Recurrent Neural Networks (RNNs) offer fast inference on long sequences but are hard to optimize and slow to train. Deep state-space models (SSMs) have recently been shown to perform remarkably well on long sequence modeling tasks, and have the added benefits of fast parallelizable training and RNN-like fast inference. However, while SSMs are superficially similar to RNNs, there are important differences that make it unclear where their performance boost over RNNs comes from. In this paper, we show that careful design of deep RNNs using standard signal propagation arguments can recover the impressive performance of deep SSMs on long-range reasoning tasks, while also matching their training speed. To achieve this, we analyze and ablate a series of changes to standard RNNs including linearizing and diagonalizing the recurrence, using better parameterizations and initializations, and ensuring proper normalization of the forward pass. Our results provide new insights on the origins of the impressive performance of deep SSMs, while also introducing an RNN block called the Linear Recurrent Unit that matches both their performance on the Long Range Arena benchmark and their computational efficiency."
                },
                {
                    "title": "Hyena Hierarchy: Towards Larger Convolutional Language Models",
                    "abstract": "Recent advances in deep learning have relied heavily on the use of large Transformers due to their ability to learn at scale. However, the core building block of Transformers, the attention operator, exhibits quadratic cost in sequence length, limiting the amount of context accessible. Existing subquadratic methods based on low-rank and sparse approximations need to be combined with dense attention layers to match Transformers, indicating a gap in capability. In this work, we propose Hyena, a subquadratic drop-in replacement for attention constructed by interleaving implicitly parametrized long convolutions and data-controlled gating. In recall and reasoning tasks on sequences of thousands to hundreds of thousands of tokens, Hyena improves accuracy by more than 50 points over operators relying on state-spaces and other implicit and explicit methods, matching attention-based models. We set a new state-of-the-art for dense-attention-free architectures on language modeling in standard datasets (WikiText103 and The Pile), reaching Transformer quality with a 20% reduction in training compute required at sequence length 2K. Hyena operators are twice as fast as highly optimized attention at sequence length 8K, and 100x faster at sequence length 64K."
                },
                {
                    "title": "Efficiently Modeling Long Sequences with Structured State Spaces",
                    "abstract": "A central goal of sequence modeling is designing a single principled model that can address sequence data across a range of modalities and tasks, particularly on long-range dependencies. Although conventional models including RNNs, CNNs, and Transformers have specialized variants for capturing long dependencies, they still struggle to scale to very long sequences of $10000$ or more steps. A promising recent approach proposed modeling sequences by simulating the fundamental state space model (SSM) \\( x'(t) = Ax(t) + Bu(t), y(t) = Cx(t) + Du(t) \\), and showed that for appropriate choices of the state matrix \\( A \\), this system could handle long-range dependencies mathematically and empirically. However, this method has prohibitive computation and memory requirements, rendering it infeasible as a general sequence modeling solution. We propose the Structured State Space sequence model (S4) based on a new parameterization for the SSM, and show that it can be computed much more efficiently than prior approaches while preserving their theoretical strengths. Our technique involves conditioning \\( A \\) with a low-rank correction, allowing it to be diagonalized stably and reducing the SSM to the well-studied computation of a Cauchy kernel. S4 achieves strong empirical results across a diverse range of established benchmarks, including (i) 91\\% accuracy on sequential CIFAR-10 with no data augmentation or auxiliary losses, on par with a larger 2-D ResNet, (ii) substantially closing the gap to Transformers on image and language modeling tasks, while performing generation $60\\times$ faster (iii) SoTA on every task from the Long Range Arena benchmark, including solving the challenging Path-X task of length 16k that all prior work fails on, while being as efficient as all competitors."
                },
                {
                    "title": "Combining Recurrent, Convolutional, and Continuous-time Models with Linear State-Space Layers",
                    "abstract": "Recurrent neural networks (RNNs), temporal convolutions, and neural differential equations (NDEs) are popular families of deep learning models for time-series data, each with unique strengths and tradeoffs in modeling power and computational efficiency. We introduce a simple sequence model inspired by control systems that generalizes these approaches while addressing their shortcomings. The Linear State-Space Layer (LSSL) maps a sequence $u \\mapsto y$ by simply simulating a linear continuous-time state-space representation $\\dot{x} = Ax + Bu, y = Cx + Du$. Theoretically, we show that LSSL models are closely related to the three aforementioned families of models and inherit their strengths. For example, they generalize convolutions to continuous-time, explain common RNN heuristics, and share features of NDEs such as time-scale adaptation. We then incorporate and generalize recent theory on continuous-time memorization to introduce a trainable subset of structured matrices $A$ that endow LSSLs with long-range memory. Empirically, stacking LSSL layers into a simple deep neural network obtains state-of-the-art results across time series benchmarks for long dependencies in sequential image classification, real-world healthcare regression tasks, and speech. On a difficult speech classification task with length-16000 sequences, LSSL outperforms prior approaches by 24 accuracy points, and even outperforms baselines that use hand-crafted features on 100x shorter sequences."
                },
                {
                    "title": "Modeling Category-Selective Cortical Regions with Topographic Variational Autoencoders",
                    "abstract": "Category-selectivity in the brain describes the observation that certain spatially localized areas of the cerebral cortex tend to respond robustly and selectively to stimuli from specific limited categories. One of the most well known examples of category-selectivity is the Fusiform Face Area (FFA), an area of the inferior temporal cortex in primates which responds preferentially to images of faces when compared with objects or other generic stimuli. In this work, we leverage the newly introduced Topographic Variational Autoencoder to model the emergence of such localized category-selectivity in an unsupervised manner. Experimentally, we demonstrate our model yields spatially dense neural clusters selective to faces, bodies, and places through visualized maps of Cohen's d metric. We compare our model with related supervised approaches, namely the Topographic Deep Artificial Neural Network (TDANN) of Lee et al., and discuss both theoretical and empirical similarities. Finally, we show preliminary results suggesting that our model yields a nested spatial hierarchy of increasingly abstract categories, analogous to observations from the human ventral temporal cortex."
                },
                {
                    "title": "FlexConv: Continuous Kernel Convolutions with Differentiable Kernel Sizes",
                    "abstract": "When designing Convolutional Neural Networks (CNNs), one must select the size\\break of the convolutional kernels before training. Recent works show CNNs benefit from different kernel sizes at different layers, but exploring all possible combinations is unfeasible in practice. A more efficient approach is to learn the kernel size during training. However, existing works that learn the kernel size have a limited bandwidth. These approaches scale kernels by dilation, and thus the detail they can describe is limited. In this work, we propose FlexConv, a novel convolutional operation with which high bandwidth convolutional kernels of learnable kernel size can be learned at a fixed parameter cost. FlexNets model long-term dependencies without the use of pooling, achieve state-of-the-art performance on several sequential datasets, outperform recent works with learned kernel sizes, and are competitive with much deeper ResNets on image benchmark datasets. Additionally, FlexNets can be deployed at higher resolutions than those seen during training. To avoid aliasing, we propose a novel kernel parameterization with which the frequency of the kernels can be analytically controlled. Our novel kernel parameterization shows higher descriptive power and faster convergence speed than existing parameterizations. This leads to important improvements in classification accuracy."
                },
                {
                    "title": "Long Expressive Memory for Sequence Modeling",
                    "abstract": "We propose a novel method called Long Expressive Memory (LEM) for learning long-term sequential dependencies. LEM is gradient-based, it can efficiently process sequential tasks with very long-term dependencies, and it is sufficiently expressive to be able to learn complicated input-output maps. To derive LEM, we consider a system of multiscale ordinary differential equations, as well as a suitable time-discretization of this system. For LEM, we derive rigorous bounds to show the mitigation of the exploding and vanishing gradients problem, a well-known challenge for gradient-based recurrent sequential learning methods. We also prove that LEM can approximate a large class of dynamical systems to high accuracy. Our empirical results, ranging from image and time-series classification through dynamical systems prediction to speech recognition and language modeling, demonstrate that LEM outperforms state-of-the-art recurrent neural networks, gated recurrent units, and long short-term memory models."
                },
                {
                    "title": "Predictive Coding with Topographic Variational Autoencoders",
                    "abstract": "Predictive coding is a model of visual processing which suggests that the brain is a generative model of input, with prediction error serving as a signal for both learning and attention. In this work, we show how the equivariant capsules learned by a Topographic Variational Autoen-coder can be extended to fit within the predictive coding framework by treating the slow rolling of capsule activations as the forward prediction operator. We demonstrate quantitatively that such an extension leads to improved sequence modeling compared with both topographic and non-topographic baselines, and that the resulting forward predictions are qualitatively more coherent with the provided partial input transformations."
                },
                {
                    "title": "Topographic VAEs learn Equivariant Capsules",
                    "abstract": "In this work we seek to bridge the concepts of topographic organization and equivariance in neural networks. To accomplish this, we introduce the Topographic VAE: a novel method for efficiently training deep generative models with topographically organized latent variables. We show that such a model indeed learns to organize its activations according to salient characteristics such as digit class, width, and style on MNIST. Furthermore, through topographic organization over time (i.e. temporal coherence), we demonstrate how predefined latent space transformation operators can be encouraged for observed transformed input sequences -- a primitive form of unsupervised learned equivariance. We demonstrate that this model successfully learns sets of approximately equivariant features (i.e.\"capsules\") directly from sequences and achieves higher likelihood on correspondingly transforming test sequences. Equivariance is verified quantitatively by measuring the approximate commutativity of the inference network and the sequence transformations. Finally, we demonstrate approximate equivariance to complex transformations, expanding upon the capabilities of existing group equivariant neural networks."
                },
                {
                    "title": "A deep convolutional neural network that is invariant to time rescaling",
                    "abstract": "Human learners can readily understand speech, or a melody, when it is presented slower or faster than usual. Although deep convolutional neural networks (CNNs) are extremely powerful in extracting information from time series, they require explicit training to generalize to different time scales. This paper presents a deep CNN that incorporates a temporal representation inspired by recent findings from neuroscience. In the mammalian brain, time is represented by populations of neurons with temporal receptive fields. Critically, the peaks of the receptive fields form a geometric series, such that the population codes a set of temporal basis functions over log time. Because memory for the recent past is a function of log time, rescaling the input results in translation of the memory. The Scale-Invariant Temporal History Convolution network (SITHCon) builds a convolutional layer over this logarithmically-distributed temporal memory. A max-pool operation results in a network that is invariant to rescalings of time modulo edge effects. We compare performance of SITHCon to a Temporal Convolution Network (TCN). Although both networks can learn classification and regression problems on both univariate and multivariate time series f(t), only SITHCon generalizes to rescalings f(at). This property, inspired by findings from contemporary neuroscience and consistent with findings from cognitive psychology, may enable networks that learn with fewer training examples, fewer weights and that generalize more robustly to out of sample data."
                },
                {
                    "title": "DeepSITH: Efficient Learning via Decomposition of What and When Across Time Scales",
                    "abstract": "Extracting temporal relationships over a range of scales is a hallmark of human perception and cognition -- and thus it is a critical feature of machine learning applied to real-world problems. Neural networks are either plagued by the exploding/vanishing gradient problem in recurrent neural networks (RNNs) or must adjust their parameters to learn the relevant time scales (e.g., in LSTMs). This paper introduces DeepSITH, a network comprising biologically-inspired Scale-Invariant Temporal History (SITH) modules in series with dense connections between layers. SITH modules respond to their inputs with a geometrically-spaced set of time constants, enabling the DeepSITH network to learn problems along a continuum of time-scales. We compare DeepSITH to LSTMs and other recent RNNs on several time series prediction and decoding tasks. DeepSITH achieves state-of-the-art performance on these problems."
                },
                {
                    "title": "UnICORNN: A recurrent model for learning very long time dependencies",
                    "abstract": "The design of recurrent neural networks (RNNs) to accurately process sequential inputs with long-time dependencies is very challenging on account of the exploding and vanishing gradient problem. To overcome this, we propose a novel RNN architecture which is based on a structure preserving discretization of a Hamiltonian system of second-order ordinary differential equations that models networks of oscillators. The resulting RNN is fast, invertible (in time), memory efficient and we derive rigorous bounds on the hidden state gradients to prove the mitigation of the exploding and vanishing gradient problem. A suite of experiments are presented to demonstrate that the proposed RNN provides state of the art performance on a variety of learning tasks with (very) long-time dependencies."
                },
                {
                    "title": "Parallelizing Legendre Memory Unit Training",
                    "abstract": "Recently, a new recurrent neural network (RNN) named the Legendre Memory Unit (LMU) was proposed and shown to achieve state-of-the-art performance on several benchmark datasets. Here we leverage the linear time-invariant (LTI) memory component of the LMU to construct a simplified variant that can be parallelized during training (and yet executed as an RNN during inference), thus overcoming a well known limitation of training RNNs on GPUs. We show that this reformulation that aids parallelizing, which can be applied generally to any deep network whose recurrent components are linear, makes training up to 200 times faster. Second, to validate its utility, we compare its performance against the original LMU and a variety of published LSTM and transformer networks on seven benchmarks, ranging from psMNIST to sentiment analysis to machine translation. We demonstrate that our models exhibit superior performance on all datasets, often using fewer parameters. For instance, our LMU sets a new state-of-the-art result on psMNIST, and uses half the parameters while outperforming DistilBERT and LSTM models on IMDB sentiment analysis."
                },
                {
                    "title": "Coupled Oscillatory Recurrent Neural Network (coRNN): An accurate and (gradient) stable architecture for learning long time dependencies",
                    "abstract": "Circuits of biological neurons, such as in the functional parts of the brain can be modeled as networks of coupled oscillators. Inspired by the ability of these systems to express a rich set of outputs while keeping (gradients of) state variables bounded, we propose a novel architecture for recurrent neural networks. Our proposed RNN is based on a time-discretization of a system of second-order ordinary differential equations, modeling networks of controlled nonlinear oscillators. We prove precise bounds on the gradients of the hidden states, leading to the mitigation of the exploding and vanishing gradient problem for this RNN. Experiments show that the proposed RNN is comparable in performance to the state of the art on a variety of benchmarks, demonstrating the potential of this architecture to provide stable and accurate RNNs for processing complex sequential data."
                },
                {
                    "title": "Brain activity fluctuations propagate as waves traversing the cortical hierarchy",
                    "abstract": "The brain exhibits highly organized patterns of spontaneous activity as measured by resting-state fMRI fluctuations that are being widely used to assess the brain\u2019s functional connectivity. Some evidence suggests that spatiotemporally coherent waves are a core feature of spontaneous activity that shapes functional connectivity, though this has been difficult to establish using fMRI given the temporal constraints of the hemodynamic signal. Here we investigated the structure of spontaneous waves in human fMRI and monkey electrocorticography. In both species, we found clear, repeatable, and directionally constrained activity waves coursed along a spatial axis approximately representing cortical hierarchical organization. These cortical propagations were closely associated with activity changes in distinct subcortical structures, particularly those related to arousal regulation, and modulated across different states of vigilance. The findings demonstrate a neural origin of spatiotemporal fMRI wave propagation at rest and link it to the principal gradient of resting-state fMRI connectivity."
                },
                {
                    "title": "HiPPO: Recurrent Memory with Optimal Polynomial Projections",
                    "abstract": "A central problem in learning from sequential data is representing cumulative history in an incremental fashion as more data is processed. We introduce a general framework (HiPPO) for the online compression of continuous signals and discrete time series by projection onto polynomial bases. Given a measure that specifies the importance of each time step in the past, HiPPO produces an optimal solution to a natural online function approximation problem. As special cases, our framework yields a short derivation of the recent Legendre Memory Unit (LMU) from first principles, and generalizes the ubiquitous gating mechanism of recurrent neural networks such as GRUs. This formal framework yields a new memory update mechanism (HiPPO-LegS) that scales through time to remember all history, avoiding priors on the timescale. HiPPO-LegS enjoys the theoretical benefits of timescale robustness, fast updates, and bounded gradients. By incorporating the memory dynamics into recurrent neural networks, HiPPO RNNs can empirically capture complex temporal dependencies. On the benchmark permuted MNIST dataset, HiPPO-LegS sets a new state-of-the-art accuracy of 98.3%. Finally, on a novel trajectory classification task testing robustness to out-of-distribution timescales and missing data, HiPPO-LegS outperforms RNN and neural ODE baselines by 25-40% accuracy."
                },
                {
                    "title": "Topographic deep artificial neural networks reproduce the hallmarks of the primate inferior temporal cortex face processing network",
                    "abstract": "A salient characteristic of monkey inferior temporal (IT) cortex is the IT face processing network. Its hallmarks include: \u201cface neurons\u201d that respond more to faces than non-face objects, strong spatial clustering of those neurons in foci at each IT anatomical level (\u201cface patches\u201d), and the preferential interconnection of those foci. While some deep artificial neural networks (ANNs) are good predictors of IT neuronal responses, including face neurons, they do not explain those face network hallmarks. Here we ask if they might be explained with a simple, metabolically motivated addition to current ANN ventral stream models. Specifically, we designed and successfully trained topographic deep ANNs (TDANNs) to solve real-world visual recognition tasks (as in prior work), but, in addition, we also optimized each network to minimize a proxy for neuronal wiring length within its IT layers. We report that after this dual optimization, the model IT layers of TDANNs reproduce the hallmarks of the IT face network: the presence of face neurons, clusters of face neurons that quantitatively match those found in IT face patches, connectivity between those patches, and the emergence of face viewpoint invariance along the network hierarchy. We find that these phenomena emerge for a range of naturalistic experience, but not for highly unnatural training. Taken together, these results show that the IT face processing network could be a consequence of a basic hierarchical anatomy along the ventral stream, selection pressure on the visual system to accomplish general object categorization, and selection pressure to minimize axonal wiring length."
                },
                {
                    "title": "Lipschitz Recurrent Neural Networks",
                    "abstract": "Viewing recurrent neural networks (RNNs) as continuous-time dynamical systems, we propose a recurrent unit that describes the hidden state's evolution with two parts: a well-understood linear component plus a Lipschitz nonlinearity. This particular functional form facilitates stability analysis of the long-term behavior of the recurrent unit using tools from nonlinear systems theory. In turn, this enables architectural design decisions before experimentation. Sufficient conditions for global stability of the recurrent unit are obtained, motivating a novel scheme for constructing hidden-to-hidden matrices. Our experiments demonstrate that the Lipschitz RNN can outperform existing recurrent units on a range of benchmark tasks, including computer vision, language modeling and speech prediction tasks. Finally, through Hessian-based analysis we demonstrate that our Lipschitz recurrent unit is more robust with respect to input and parameter perturbations as compared to other continuous-time RNNs."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "The Human Brain Encodes a Chronicle of Visual Events at Each Instant of Time Through the Multiplexing of Traveling Waves",
                    "abstract": "The human brain continuously processes streams of visual input. Yet, a single image typically triggers neural responses that extend beyond 1s. To understand how the brain encodes and maintains successive images, we analyzed with electroencephalography the brain activity of human subjects while they watched \u223c5000 visual stimuli presented in fast sequences. First, we confirm that each stimulus can be decoded from brain activity for \u223c1s, and we demonstrate that the brain simultaneously represents multiple images at each time instant. Second, we source localize the corresponding brain responses in the expected visual hierarchy and show that distinct brain regions represent, at each time instant, different snapshots of past stimulations. Third, we propose a simple framework to further characterize the dynamical system of these traveling waves. Our results show that a chain of neural circuits, which each consist of (1) a hidden maintenance mechanism and (2) an observable update mechanism, accounts for the dynamics of macroscopic brain representations elicited by visual sequences. Together, these results detail a simple architecture explaining how successive visual events and their respective timings can be simultaneously represented in the brain. SIGNIFICANCE STATEMENT Our retinas are continuously bombarded with a rich flux of visual input. Yet, how our brain continuously processes such visual streams is a major challenge to neuroscience. Here, we developed techniques to decode and track, from human brain activity, multiple images flashed in rapid succession. Our results show that the brain simultaneously represents multiple successive images at each time instant by multiplexing them along a neural cascade. Dynamical modeling shows that these results can be explained by a hierarchy of neural assemblies that continuously propagate multiple visual contents. Overall, this study sheds new light on the biological basis of our visual experience."
                },
                {
                    "title": "Brain-Like Object Recognition with High-Performing Shallow Recurrent ANNs",
                    "abstract": "Deep convolutional artificial neural networks (ANNs) are the leading class of candidate models of the mechanisms of visual processing in the primate ventral stream. While initially inspired by brain anatomy, over the past years, these ANNs have evolved from a simple eight-layer architecture in AlexNet to extremely deep and branching architectures, demonstrating increasingly better object categorization performance, yet bringing into question how brain-like they still are. In particular, typical deep models from the machine learning community are often hard to map onto the brain's anatomy due to their vast number of layers and missing biologically-important connections, such as recurrence. Here we demonstrate that better anatomical alignment to the brain and high performance on machine learning as well as neuroscience measures do not have to be in contradiction. We developed CORnet-S, a shallow ANN with four anatomically mapped areas and recurrent connectivity, guided by Brain-Score, a new large-scale composite of neural and behavioral benchmarks for quantifying the functional fidelity of models of the primate ventral visual stream. Despite being significantly shallower than most models, CORnet-S is the top model on Brain-Score and outperforms similarly compact models on ImageNet. Moreover, our extensive analyses of CORnet-S circuitry variants reveal that recurrence is the main predictive factor of both Brain-Score and ImageNet top-1 performance. Finally, we report that the temporal evolution of the CORnet-S \"IT\" neural population resembles the actual monkey IT population dynamics. Taken together, these results establish CORnet-S, a compact, recurrent ANN, as the current best model of the primate ventral visual stream."
                },
                {
                    "title": "AntisymmetricRNN: A Dynamical System View on Recurrent Neural Networks",
                    "abstract": "Recurrent neural networks have gained widespread use in modeling sequential data. Learning long-term dependencies using these models remains difficult though, due to exploding or vanishing gradients. In this paper, we draw connections between recurrent networks and ordinary differential equations. A special form of recurrent networks called the AntisymmetricRNN is proposed under this theoretical framework, which is able to capture long-term dependencies thanks to the stability property of its underlying differential equation. Existing approaches to improving RNN trainability often incur significant computation overhead. In comparison, AntisymmetricRNN achieves the same goal by design. We showcase the advantage of this new architecture through extensive simulations and experiments. AntisymmetricRNN exhibits much more predictable dynamics. It outperforms regular LSTM models on tasks requiring long-term memory and matches the performance on tasks where short-term dependencies dominate despite being much simpler."
                },
                {
                    "title": "Cheap Orthogonal Constraints in Neural Networks: A Simple Parametrization of the Orthogonal and Unitary Group",
                    "abstract": "We introduce a novel approach to perform first-order optimization with orthogonal and unitary constraints. This approach is based on a parametrization stemming from Lie group theory through the exponential map. The parametrization transforms the constrained optimization problem into an unconstrained one over a Euclidean space, for which common first-order optimization methods can be used. The theoretical results presented are general enough to cover the special orthogonal group, the unitary group and, in general, any connected compact Lie group. We discuss how this and other parametrizations can be computed efficiently through an implementation trick, making numerically complex parametrizations usable at a negligible runtime cost in neural networks. In particular, we apply our results to RNNs with orthogonal recurrent weights, yielding a new architecture called expRNN. We demonstrate how our method constitutes a more robust approach to optimization with orthogonal constraints, showing faster, accurate, and more stable convergence in several tasks designed to test RNNs."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "Independently Recurrent Neural Network (IndRNN): Building A Longer and Deeper RNN",
                    "abstract": "Recurrent neural networks (RNNs) have been widely used for processing sequential data. However, RNNs are commonly difficult to train due to the well-known gradient vanishing and exploding problems and hard to learn long-term patterns. Long short-term memory (LSTM) and gated recurrent unit (GRU) were developed to address these problems, but the use of hyperbolic tangent and the sigmoid action functions results in gradient decay over layers. Consequently, construction of an efficiently trainable deep network is challenging. In addition, all the neurons in an RNN layer are entangled together and their behaviour is hard to interpret. To address these problems, a new type of RNN, referred to as independently recurrent neural network (IndRNN), is proposed in this paper, where neurons in the same layer are independent of each other and they are connected across layers. We have shown that an IndRNN can be easily regulated to prevent the gradient exploding and vanishing problems while allowing the network to learn long-term dependencies. Moreover, an IndRNN can work with non-saturated activation functions such as relu (rectified linear unit) and be still trained robustly. Multiple IndRNNs can be stacked to construct a network that is deeper than the existing RNNs. Experimental results have shown that the proposed IndRNN is able to process very long sequences (over 5000 time steps), can be used to construct very deep networks (21 layers used in the experiment) and still be trained robustly. Better performances have been achieved on various tasks by using IndRNNs compared with the traditional RNN and LSTM."
                },
                {
                    "title": "Can recurrent neural networks warp time?",
                    "abstract": "Successful recurrent models such as long short-term memories (LSTMs) and gated recurrent units (GRUs) use ad hoc gating mechanisms. Empirically these models have been found to improve the learning of medium to long term temporal dependencies and to help with vanishing gradient issues. We prove that learnable gates in a recurrent model formally provide quasi- invariance to general time transformations in the input data. We recover part of the LSTM architecture from a simple axiomatic approach. This result leads to a new way of initializing gate biases in LSTMs and GRUs. Ex- perimentally, this new chrono initialization is shown to greatly improve learning of long term dependencies, with minimal implementation effort."
                },
                {
                    "title": "Using Fast Weights to Attend to the Recent Past",
                    "abstract": "Until recently, research on artificial neural networks was largely restricted to systems with only two types of variable: Neural activities that represent the current or recent input and weights that learn to capture regularities among inputs, outputs and payoffs. There is no good reason for this restriction. Synapses have dynamics at many different time-scales and this suggests that artificial neural networks might benefit from variables that change slower than activities but much faster than the standard weights. These ``fast weights'' can be used to store temporary memories of the recent past and they provide a neurally plausible way of implementing the type of attention to the past that has recently proven helpful in sequence-to-sequence models. By using fast weights we can avoid the need to store copies of neural activity patterns."
                },
                {
                    "title": "Rotating waves during human sleep spindles organize global patterns of activity that repeat precisely through the night",
                    "abstract": "During sleep, the thalamus generates a characteristic pattern of transient, 11-15 Hz sleep spindle oscillations, which synchronize the cortex through large-scale thalamocortical loops. Spindles have been increasingly demonstrated to be critical for sleep-dependent consolidation of memory, but the specific neural mechanism for this process remains unclear. We show here that cortical spindles are spatiotemporally organized into circular wave-like patterns, organizing neuronal activity over tens of milliseconds, within the timescale for storing memories in large-scale networks across the cortex via spike-time dependent plasticity. These circular patterns repeat over hours of sleep with millisecond temporal precision, allowing reinforcement of the activity patterns through hundreds of reverberations. These results provide a novel mechanistic account for how global sleep oscillations and synaptic plasticity could strengthen networks distributed across the cortex to store coherent and integrated memories. DOI: http://dx.doi.org/10.7554/eLife.17267.001"
                },
                {
                    "title": "Wave-Based Turing Machine: Time Reversal and Information Erasing.",
                    "abstract": "The investigation of dynamical systems has revealed a deep-rooted difference between waves and objects regarding temporal reversibility and particlelike objects. In nondissipative chaos, the dynamic of waves always remains time reversible, unlike that of particles. Here, we explore the dynamics of a wave-particle entity. It consists in a drop bouncing on a vibrated liquid bath, self-propelled and piloted by the surface waves it generates. This walker, in which there is an information exchange between the particle and the wave, can be analyzed in terms of a Turing machine with waves as the information repository. The experiments reveal that in this system, the drop can read information backwards while erasing it. The drop can thus backtrack on its previous trajectory. A transient temporal reversibility, restricted to the drop motion, is obtained in spite of the system being both dissipative and chaotic."
                },
                {
                    "title": "One-shot Learning with Memory-Augmented Neural Networks",
                    "abstract": "Despite recent breakthroughs in the applications of deep neural networks, one setting that presents a persistent challenge is that of \u201cone-shot learning.\u201d Traditional gradient-based networks require a lot of data to learn, often through extensive iterative training. When new data is encountered, the models must inefficiently relearn their parameters to adequately incorporate the new information without catastrophic interference. Architectures with augmented memory capacities, such as Neural Turing Machines (NTMs), offer the ability to quickly encode and retrieve new information, and hence can potentially obviate the downsides of conventional models. Here, we demonstrate the ability of a memory-augmented neural network to rapidly assimilate new data, and leverage this data to make accurate predictions after only a few samples. We also introduce a new method for accessing an external memory that focuses on memory content, unlike previous methods that additionally use memory locationbased focusing mechanisms."
                },
                {
                    "title": "Recurrent Orthogonal Networks and Long-Memory Tasks",
                    "abstract": "Although RNNs have been shown to be powerful tools for processing sequential data, finding architectures or optimization strategies that allow them to model very long term dependencies is still an active area of research. In this work, we carefully analyze two synthetic datasets originally outlined in (Hochreiter and Schmidhuber, 1997) which are used to evaluate the ability of RNNs to store information over many time steps. We explicitly construct RNN solutions to these problems, and using these constructions, illuminate both the problems themselves and the way in which RNNs store different types of information in their hidden states. These constructions furthermore explain the success of recent methods that specify unitary initializations or constraints on the transition matrices."
                },
                {
                    "title": "Neural GPUs Learn Algorithms",
                    "abstract": "Learning an algorithm from examples is a fundamental problem that has been widely studied. Recently it has been addressed using neural networks, in particular by Neural Turing Machines (NTMs). These are fully differentiable computers that use backpropagation to learn their own programming. Despite their appeal NTMs have a weakness that is caused by their sequential nature: they are not parallel and are are hard to train due to their large depth when unfolded. \nWe present a neural network architecture to address this problem: the Neural GPU. It is based on a type of convolutional gated recurrent unit and, like the NTM, is computationally universal. Unlike the NTM, the Neural GPU is highly parallel which makes it easier to train and efficient to run. \nAn essential property of algorithms is their ability to handle inputs of arbitrary size. We show that the Neural GPU can be trained on short instances of an algorithmic task and successfully generalize to long instances. We verified it on a number of tasks including long addition and long multiplication of numbers represented in binary. We train the Neural GPU on numbers with upto 20 bits and observe no errors whatsoever while testing it, even on much longer numbers. \nTo achieve these results we introduce a technique for training deep recurrent networks: parameter sharing relaxation. We also found a small amount of dropout and gradient noise to have a large positive effect on learning and generalization."
                },
                {
                    "title": "Unitary Evolution Recurrent Neural Networks",
                    "abstract": "Recurrent neural networks (RNNs) are notoriously difficult to train. When the eigenvalues of the hidden to hidden weight matrix deviate from absolute value 1, optimization becomes difficult due to the well studied issue of vanishing and exploding gradients, especially when trying to learn long-term dependencies. To circumvent this problem, we propose a new architecture that learns a unitary weight matrix, with eigenvalues of absolute value exactly 1. The challenge we address is that of parametrizing unitary matrices in a way that does not require expensive computations (such as eigendecomposition) after each weight update. We construct an expressive unitary weight matrix by composing several structured matrices that act as building blocks with parameters to be learned. Optimization with this parameterization becomes feasible only when considering hidden states in the complex domain. We demonstrate the potential of this architecture by achieving state of the art results in several hard tasks involving very long-term dependencies."
                },
                {
                    "title": "Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting",
                    "abstract": "The goal of precipitation nowcasting is to predict the future rainfall intensity in a local region over a relatively short period of time. Very few previous studies have examined this crucial and challenging weather forecasting problem from the machine learning perspective. In this paper, we formulate precipitation nowcasting as a spatiotemporal sequence forecasting problem in which both the input and the prediction target are spatiotemporal sequences. By extending the fully connected LSTM (FC-LSTM) to have convolutional structures in both the input-to-state and state-to-state transitions, we propose the convolutional LSTM (ConvLSTM) and use it to build an end-to-end trainable model for the precipitation nowcasting problem. Experiments show that our ConvLSTM network captures spatiotemporal correlations better and consistently outperforms FC-LSTM and the state-of-the-art operational ROVER algorithm for precipitation nowcasting."
                },
                {
                    "title": "A Simple Way to Initialize Recurrent Networks of Rectified Linear Units",
                    "abstract": "Learning long term dependencies in recurrent networks is difficult due to vanishing and exploding gradients. To overcome this difficulty, researchers have developed sophisticated optimization techniques and network architectures. In this paper, we propose a simpler solution that use recurrent neural networks composed of rectified linear units. Key to our solution is the use of the identity matrix or its scaled version to initialize the recurrent weight matrix. We find that our solution is comparable to LSTM on our four benchmarks: two toy problems involving long-range temporal structures, a large language modeling problem and a benchmark speech recognition problem."
                },
                {
                    "title": "Emergence of Complex Wave Patterns in Primate Cerebral Cortex",
                    "abstract": "Slow brain rhythms are attributed to near-simultaneous (synchronous) changes in activity in neuron populations in the brain. Because they are slow and widespread, synchronous rhythms have not been considered crucial for information processing in the waking state. Here we adapted methods from turbulence physics to analyze \u03b4-band (1\u20134 Hz) rhythms in local field potential (LFP) activity, in multielectrode recordings from cerebral cortex in anesthetized marmoset monkeys. We found that synchrony contributes only a small fraction (less than one-fourth) to the local spatiotemporal structure of \u03b4-band signals. Rather, \u03b4-band activity is dominated by propagating plane waves and spatiotemporal structures, which we call complex waves. Complex waves are manifest at submillimeter spatial scales, and millisecond-range temporal scales. We show that complex waves can be characterized by their relation to phase singularities within local nerve cell networks. We validate the biological relevance of complex waves by showing that nerve cell spike rates are higher in presence of complex waves than in the presence of synchrony and that there are nonrandom patterns of evolution from one type of complex wave to another. We conclude that slow brain rhythms predominantly indicate spatiotemporally organized activity in local nerve cell circuits, not synchronous activity within and across brain regions."
                },
                {
                    "title": "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification",
                    "abstract": "Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on the learnable activation and advanced initialization, we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66% [33]). To our knowledge, our result is the first to surpass the reported human-level performance (5.1%, [26]) on this dataset."
                },
                {
                    "title": "Neural Turing Machines",
                    "abstract": "We extend the capabilities of neural networks by coupling them to external memory resources, which they can interact with by attentional processes. The combined system is analogous to a Turing Machine or Von Neumann architecture but is differentiable end-toend, allowing it to be efficiently trained with gradient descent. Preliminary results demonstrate that Neural Turing Machines can infer simple algorithms such as copying, sorting, and associative recall from input and output examples."
                },
                {
                    "title": "Deep Neural Networks Rival the Representation of Primate IT Cortex for Core Visual Object Recognition",
                    "abstract": "The primate visual system achieves remarkable visual object recognition performance even in brief presentations, and under changes to object exemplar, geometric transformations, and background variation (a.k.a. core visual object recognition). This remarkable performance is mediated by the representation formed in inferior temporal (IT) cortex. In parallel, recent advances in machine learning have led to ever higher performing models of object recognition using artificial deep neural networks (DNNs). It remains unclear, however, whether the representational performance of DNNs rivals that of the brain. To accurately produce such a comparison, a major difficulty has been a unifying metric that accounts for experimental limitations, such as the amount of noise, the number of neural recording sites, and the number of trials, and computational limitations, such as the complexity of the decoding classifier and the number of classifier training examples. In this work, we perform a direct comparison that corrects for these experimental limitations and computational considerations. As part of our methodology, we propose an extension of \u201ckernel analysis\u201d that measures the generalization accuracy as a function of representational complexity. Our evaluations show that, unlike previous bio-inspired models, the latest DNNs rival the representational performance of IT cortex on this visual object recognition task. Furthermore, we show that models that perform well on measures of representational performance also perform well on measures of representational similarity to IT, and on measures of predicting individual IT multi-unit responses. Whether these DNNs rely on computational mechanisms similar to the primate visual system is yet to be determined, but, unlike all previous bio-inspired models, that possibility cannot be ruled out merely on representational performance grounds."
                },
                {
                    "title": "Propagating Waves in Human Motor Cortex",
                    "abstract": "Previous studies in non-human primates (NHPs) have shown that beta oscillations (15\u201330 Hz) of local field potentials (LFPs) in the arm/hand areas of primary motor cortex (MI) propagate as traveling waves across the cortex. These waves exhibited two stereotypical features across animals and tasks: (1) The waves propagated in two dominant modal directions roughly 180\u00b0 apart, and (2) their propagation speed ranged from 10 to 35 cm/s. It is, however, unknown if such cortical waves occur in the human motor cortex. This study shows that the two properties of propagating beta waves are present in MI of a tetraplegic human patient while he was instructed to perform an instruction delay center-out task using a cursor controlled by the chin. Moreover, we show that beta waves are sustained and have similar properties whether the subject was engaged in the task or at rest. The directions of the successive sustained waves both in the human subject and a NHP subject tended to switch from one dominant mode to the other, and at least in the NHP subject the estimated distance traveled between successive waves traveling into and out of the central sulcus is consistent with the hypothesis of wave reflection between the border of motor and somatosensory cortices. Further, we show that the occurrence of the beta waves is not uniquely tied to periods of increased power in the beta frequency band. These results demonstrate that traveling beta waves in MI are a general phenomenon occurring in human as well as NHPs. Consistent with the NHP data, the dominant directions of the beta LFP waves in human aligned to the proximal to distal gradient of joint representations in MI somatotopy. This consistent finding of wave propagation may imply the existence of a hardwired organization of motor cortex that mediates this spatiotemporal pattern."
                },
                {
                    "title": "Gating of Human Theta Oscillations by a Working Memory Task",
                    "abstract": "Electrode grids on the cortical surface of epileptic patients provide a unique opportunity to observe brain activity with high temporal\u2013spatial resolution and high signal-to-noise ratio during a cognitive task. Previous work showed that large-amplitude theta frequency oscillations occurred intermittently during a maze navigation task, but it was unclear whether theta related to the spatial or working memory components of the task. To determine whether theta occurs during a nonspatial task, we made recordings while subjects performed the Sternberg working memory task. Our results show event-related theta and reveal a new phenomenon, the cognitive \u201cgating\u201d of a brain oscillation: at many cortical sites, the amplitude of theta oscillations increased dramatically at the start of the trial, continued through all phases of the trial, including the delay period, and decreased sharply at the end. Gating could be seen in individual trials and varying the duration of the trial systematically varied the period of gating. These results suggest that theta oscillations could have an important role in organizing multi-item working memory."
                },
                {
                    "title": "Long Short-Term Memory",
                    "abstract": "Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, LSTM can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units. Multiplicative gate units learn to open and close access to the constant error flow. LSTM is local in space and time; its computational complexity per time step and weight is O. 1. Our experiments with artificial data involve local, distributed, real-valued, and noisy pattern representations. In comparisons with real-time recurrent learning, back propagation through time, recurrent cascade correlation, Elman nets, and neural sequence chunking, LSTM leads to many more successful runs, and learns much faster. LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms."
                },
                {
                    "title": "Movement characterization with the spatiotemporal Fourier transform of low-light-level images.",
                    "abstract": "The spatiotemporal Fourier transform of a series of frames is proved to be an attractive technique for object velocity determination. The main disadvantage of the spatiotemporal transform methods is that they require a long processing time even when fast transforms are used. We show that the spatiotemporal Fourier-transform technique, which has been applied to standard-light-intensity images, can also be used with low-light-level images. The result is an important processing time reduction; thus when fast processing is required, it would be advantageous to detect in low-light-level conditions with intensity filters. The technique behavior for low-light-level images is checked with computer-simulated and experimental data."
                },
                {
                    "title": "Finding Structure in Time",
                    "abstract": "Time underlies many interesting human behaviors. Thus, the question of how to represent time in connectionist models is very important. One approach is to represent time implicitly by its effects on processing rather than explicitly (as in a spatial representation). The current report develops a proposal along these lines first described by Jordan (1986) which involves the use of recurrent links in order to provide networks with a dynamic memory. In this approach, hidden unit patterns are fed back to themselves; the internal representations which develop thus reflect task demands in the context of prior internal states. A set of simulations is reported which range from relatively simple problems (temporal version of XOR) to discovering syntactic/semantic features for words. The networks are able to learn interesting internal representations which incorporate task demands with memory demands; indeed, in this approach the notion of memory is inextricably bound up with task processing. These representations reveal a rich structure, which allows them to be highly context-dependent while also expressing generalizations across classes of items. These representations suggest a method for representing lexical categories and the type/token distinction."
                },
                {
                    "title": "Recognition and velocity computation of large moving objects in images",
                    "abstract": "Two one-dimensional time sequences are generated from the projections of the two-dimensional sequence on the x and y axes. Then the two-dimensional fast Fourier transform for the generated time sequences is computed. A peak in the spectrum for the selected spatial frequency is detected. The temporal frequency at which the peak is detected gives an estimate of the velocity of the moving object. Analytical formulations for large moving objects in a time sequence with zero background are presented. An algorithm is given for velocity formulation. >"
                },
                {
                    "title": "Neural Wave Machines: Learning Spatiotemporally Structured Representations with Locally Coupled Oscillatory Recurrent Neural Networks",
                    "abstract": "Traveling waves have been measured at a diversity of regions and scales in the brain, however a consensus as to their computational purpose has yet to be reached. An intriguing hypothesis is that traveling waves serve to structure neural representations both in space and time, thereby acting as an inductive bias towards natural data. In this work, we investigate this hypothesis by introducing the Neural Wave Machine (NWM) \u2013 a locally coupled oscillatory recurrent neural network capable of exhibiting traveling waves in its hidden state. After training on simple dynamic sequences, we show that this model indeed learns static spatial structure such as topographic organization, and further uses complex spatiotemporal structure such as traveling waves to encode observed transformations. To measure the computational implications of this structure, we use a suite of sequence classification and physical dynamics modeling tasks to show that the NWM is both more parameter efficient, and is able to forecast future trajectories of simple physical dynamical systems more accurately than existing state of the art counterparts. We conclude with a discussion of how this model may allow for novel investigations of the computational hypotheses surrounding traveling waves which were previously challenging or impossible."
                },
                {
                    "title": "Hippocampus as a generative circuit for predictive coding of future sequences",
                    "abstract": "The interaction between hippocampus and cortex is key to memory formation and representation learning. Based on anatomical wiring and transmission delays, we proposed a self-supervised recurrent neural network (PredRAE) with a predictive reconstruction loss to account for the cognitive functions of hippocampus. This framework extends predictive coding in the time axis and incorporates predictive features in Bayes \ufb01lters for temporal prediction. In simulations, we were able to reproduce characteristic place cell features such as one-shot plasticity, localized spatial representation and replay, which marks the trace of memory formation. The simulated place cells also exhibited precise spike timing, evidenced by phase precession. Trained on MNIST sequences, PredRAE learned the underlying temporal dependencies and a spontaneous representation of digit labels and rotation dynamics in the linear transformation of its hidden unit activities. Such learning is robust against 16 different image corruptions. Inspired by the brain circuit, the simple and concise framework has great potential to approximate human performance with its capacity to robustly disentangle representations and generalize temporal dynamics."
                },
                {
                    "title": "Improving the Gating Mechanism of Recurrent Neural Networks",
                    "abstract": "Section 4 briefly introduced chrono initialization (Tallec & Ollivier, 2018) and the ON-LSTM (Shen et al., 2018), closely related methods that modify the gating mechanism of LSTMs. We provide more detailed discussion on these in Sections B.3 and B.4 respectively. Section B.1 has a more thorough overview of related work on recurrent neural networks that address long-term dependencies or saturating gates."
                }
            ],
            "categories": [
                "cs.NE",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow do traveling waves of neural activity in the brain contribute to memory encoding and retrieval processes?\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding the role of traveling waves in memory processes has significant implications for the neuroscience community. It could reshape our understanding of how information is encoded and retrieved in the brain, potentially leading to new insights into cognitive functions and disorders. This research could pave the way for innovative approaches in neuropsychology, cognitive rehabilitation, and the development of brain-computer interfaces, ultimately advancing both theoretical knowledge and practical applications in neuroscience.\n\n**[Question 3] - Why is it hard?**  \nThe complexity of neural dynamics poses significant challenges in isolating the specific contributions of traveling waves to memory processes. Naive approaches may fail because they often overlook the intricate interactions between different types of neural oscillations and the spatial-temporal dynamics of wave propagation. Additionally, technical obstacles such as the limitations of current imaging techniques and the difficulty in accurately modeling the brain's multi-scale activity further complicate the investigation of this phenomenon.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on static oscillatory patterns rather than the dynamic nature of traveling waves, leading to a limited understanding of their functional roles. Gaps in methodologies, such as the lack of high-resolution imaging techniques in earlier studies, have hindered the exploration of wave dynamics. Furthermore, existing theories have not adequately integrated the concept of wave propagation with memory encoding, which is a key aspect of this research. My approach will leverage advanced imaging technologies and a novel analytical framework to directly investigate the relationship between traveling waves and memory.\n\n**[Question 5] - What are the key components of my approach and results?**  \nMy proposed methodology involves using high-density multi-electrode arrays to record neural activity while participants engage in memory tasks. I will analyze the resulting data to identify traveling wave patterns and their correlation with memory performance metrics, such as recall accuracy and response times. The expected outcomes include a clearer understanding of how traveling waves facilitate memory encoding and retrieval, potentially leading to the development of new models of cognitive processing in the brain."
            }
        },
        "author_data": {
            "04e50528-414f-4635-8d23-fb8b853c4141": {
                "pk": "04e50528-414f-4635-8d23-fb8b853c4141",
                "name": "T. Anderson Keller",
                "collaborators": [
                    "Xavier Suau",
                    "L. Zappella",
                    "Max Welling",
                    "M. Welling",
                    "Federico Danieli",
                    "Arno Blaas",
                    "Chen Huang",
                    "Jason Ramapuram",
                    "Dan Busbridge",
                    "Yue Song"
                ],
                "domain": [
                    "Self-Supervised Learning",
                    "Representation Learning",
                    "Neural Networks",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "DUET: 2D Structured and Approximately Equivariant Representations",
                        "abstract": "Multiview Self-Supervised Learning (MSSL) is based on learning invariances with respect to a set of input transformations. However, invariance partially or totally removes transformation-related information from the representations, which might harm performance for specific downstream tasks that require such information. We propose 2D strUctured and EquivarianT representations (coined DUET), which are 2d representations organized in a matrix structure, and equivariant with respect to transformations acting on the input data. DUET representations maintain information about an input transformation, while remaining semantically expressive. Compared to SimCLR (Chen et al., 2020) (unstructured and invariant) and ESSL (Dangovski et al., 2022) (unstructured and equivariant), the structured and equivariant nature of DUET representations enables controlled generation with lower reconstruction error, while controllability is not possible with SimCLR or ESSL. DUET also achieves higher accuracy for several discriminative tasks, and improves transfer learning."
                    },
                    {
                        "title": "Flow Factorized Representation Learning",
                        "abstract": "A prominent goal of representation learning research is to achieve representations which are factorized in a useful manner with respect to the ground truth factors of variation. The fields of disentangled and equivariant representation learning have approached this ideal from a range of complimentary perspectives; however, to date, most approaches have proven to either be ill-specified or insufficiently flexible to effectively separate all realistic factors of interest in a learned latent space. In this work, we propose an alternative viewpoint on such structured representation learning which we call Flow Factorized Representation Learning, and demonstrate it to learn both more efficient and more usefully structured representations than existing frameworks. Specifically, we introduce a generative model which specifies a distinct set of latent probability paths that define different input transformations. Each latent flow is generated by the gradient field of a learned potential following dynamic optimal transport. Our novel setup brings new understandings to both \\textit{disentanglement} and \\textit{equivariance}. We show that our model achieves higher likelihoods on standard representation learning benchmarks while simultaneously being closer to approximately equivariant models. Furthermore, we demonstrate that the transformations learned by our model are flexibly composable and can also extrapolate to new data, implying a degree of robustness and generalizability approaching the ultimate goal of usefully factorized representation learning."
                    },
                    {
                        "title": "Neural Wave Machines: Learning Spatiotemporally Structured Representations with Locally Coupled Oscillatory Recurrent Neural Networks",
                        "abstract": "Traveling waves have been measured at a diversity of regions and scales in the brain, however a consensus as to their computational purpose has yet to be reached. An intriguing hypothesis is that traveling waves serve to structure neural representations both in space and time, thereby acting as an inductive bias towards natural data. In this work, we investigate this hypothesis by introducing the Neural Wave Machine (NWM) \u2013 a locally coupled oscillatory recurrent neural network capable of exhibiting traveling waves in its hidden state. After training on simple dynamic sequences, we show that this model indeed learns static spatial structure such as topographic organization, and further uses complex spatiotemporal structure such as traveling waves to encode observed transformations. To measure the computational implications of this structure, we use a suite of sequence classification and physical dynamics modeling tasks to show that the NWM is both more parameter efficient, and is able to forecast future trajectories of simple physical dynamical systems more accurately than existing state of the art counterparts. We conclude with a discussion of how this model may allow for novel investigations of the computational hypotheses surrounding traveling waves which were previously challenging or impossible."
                    },
                    {
                        "title": "Homomorphic Self-Supervised Learning",
                        "abstract": "In this work, we observe that many existing self-supervised learning algorithms can be both unified and generalized when seen through the lens of equivariant representations. Specifically, we introduce a general framework we call Homomorphic Self-Supervised Learning, and theoretically show how it may subsume the use of input-augmentations provided an augmentation-homomorphic feature extractor. We validate this theory experimentally for simple augmentations, demonstrate how the framework fails when representational structure is removed, and further empirically explore how the parameters of this framework relate to those of traditional augmentation-based self-supervised learning. We conclude with a discussion of the potential benefits afforded by this new perspective on self-supervised learning."
                    },
                    {
                        "title": "Predictive Coding with Topographic Variational Autoencoders",
                        "abstract": "Predictive coding is a model of visual processing which suggests that the brain is a generative model of input, with prediction error serving as a signal for both learning and attention. In this work, we show how the equivariant capsules learned by a Topographic Variational Autoen-coder can be extended to fit within the predictive coding framework by treating the slow rolling of capsule activations as the forward prediction operator. We demonstrate quantitatively that such an extension leads to improved sequence modeling compared with both topographic and non-topographic baselines, and that the resulting forward predictions are qualitatively more coherent with the provided partial input transformations."
                    },
                    {
                        "title": "Topographic VAEs learn Equivariant Capsules",
                        "abstract": "In this work we seek to bridge the concepts of topographic organization and equivariance in neural networks. To accomplish this, we introduce the Topographic VAE: a novel method for efficiently training deep generative models with topographically organized latent variables. We show that such a model indeed learns to organize its activations according to salient characteristics such as digit class, width, and style on MNIST. Furthermore, through topographic organization over time (i.e. temporal coherence), we demonstrate how predefined latent space transformation operators can be encouraged for observed transformed input sequences -- a primitive form of unsupervised learned equivariance. We demonstrate that this model successfully learns sets of approximately equivariant features (i.e.\"capsules\") directly from sequences and achieves higher likelihood on correspondingly transforming test sequences. Equivariance is verified quantitatively by measuring the approximate commutativity of the inference network and the sequence transformations. Finally, we demonstrate approximate equivariance to complex transformations, expanding upon the capabilities of existing group equivariant neural networks."
                    },
                    {
                        "title": "Developing Fine-Grained Actigraphies for Rheumatoid Arthritis Patients from a Single Accelerometer Using Machine Learning",
                        "abstract": "In addition to routine clinical examination, unobtrusive and physical monitoring of Rheumatoid Arthritis (RA) patients provides an important source of information to enable understanding the impact of the disease on quality of life. Besides an increase in sedentary behaviour, pain in RA can negatively impact simple physical activities such as getting out of bed and standing up from a chair. The objective of this work is to develop a method that can generate fine-grained actigraphies to capture the impact of the disease on the daily activities of patients. A processing methodology is presented to automatically tag activity accelerometer data from a cohort of moderate-to-severe RA patients. A study of procesing methods based on machine learning and deep learning is provided. Thirty subjects, 10 RA patients and 20 healthy control subjects, were recruited in the study. A single tri-axial accelerometer was attached to the position of the fifth lumbar vertebra (L5) of each subject with a tag prediction granularity of 3 s. The proposed method is capable of handling unbalanced datasets from tagged data while accounting for long-duration activities such as sitting and lying, as well as short transitions such as sit-to-stand or lying-to-sit. The methodology also includes a novel mechanism for automatically applying a threshold to predictions by their confidence levels, in addition to a logical filter to correct for infeasible sequences of activities. Performance tests showed that the method was able to achieve around 95% accuracy and 81% F-score. The produced actigraphies can be helpful to generate objective RA disease-specific markers of patient mobility in-between clinical site visits."
                    },
                    {
                        "title": "Comparison and Fine-grained Analysis of Sequence Encoders for Natural Language Processing",
                        "abstract": "Most machine learning algorithms require a fixed length input to be able to perform commonly desired tasks such as classification, clustering, and regression. For natural language processing, the inherently unbounded and recursive nature of the input poses a unique challenge when deriving such fixed length representations. Although today there is a general consensus on how to generate fixed length representations of individual words which preserve their meaning, the same cannot be said for sequences of words in sentences, paragraphs, or documents. In this work, we study the encoders commonly used to generate fixed length representations of natural language sequences, and analyze their effectiveness across a variety of high and low level tasks including sentence classification and question answering. Additionally, we propose novel improvements to the existing Skip-Thought and End-to-End Memory Network architectures and study their performance on both the original and auxiliary tasks. Ultimately, we show that the setting in which the encoders are trained, and the corpus used for training, have a greater influence of the final learned representation than the underlying sequence encoders themselves."
                    },
                    {
                        "title": "Designing Policy Recommendations to Reduce Home Abandonment in Mexico",
                        "abstract": "Infonavit, the largest provider of mortgages in Mexico, assists working families to obtain low-interest rate housing solutions. An increasingly prevalent problem is home abandonment: when a homeowner decides to leave their property and forego their investment. A major causal factor of this outcome is a mismatch between the homeowner's needs, in terms of access to services and employment, and the location characteristics of the home. This paper describes our collaboration with Infonavit to reduce home abandonment at two levels: develop policy recommendations for targeted improvements in location characteristics, and develop a decision-support tool to assist the homeowner in the home location decision. Using 20 years of mortgage history data combined with surveys, census, and location information, we develop a model to predict the probability of home abandonment based on both individual and location characteristics. The model is used to develop a tool that provides Infonavit the ability to give advice to Mexican workers when they apply for a loan, evaluate and improve the locations of new housing developments, and provide data-driven recommendations to the federal government to influence local development initiatives and infrastructure investments. The result is improving economic outcomes for the citizens of Mexico by pre-emptively identifying at-risk home mortgages, thereby allowing them to be altered or remedied before they result in abandonment."
                    },
                    {
                        "title": "Experimental realization of a nonlinear acoustic lens with a tunable focus",
                        "abstract": "We realize a nonlinear acoustic lens composed of a two-dimensional array of sphere chains interfaced with water. The chains are able to support solitary waves which, when interfaced with a linear medium, transmit compact pulses with minimal oscillations. When focused, the lens is able to produce compact pressure pulses of high amplitude, the \u201csound bullets.\u201d We demonstrate that the focal point can be controlled via pre-compression of the individual chains, as this changes the wave speed within them. The experimental results agree well both spatially and temporally with analytical predictions over a range of focus locations."
                    },
                    {
                        "title": "Use of waveguides for polarized neutron studies at the micrometric scale",
                        "abstract": "A BSTRACT : We present the use of planar waveguides to produce neutron microbeams in order to investigate magnetic micro-structures with a micrometric spatial resolution. We report experimental results on such measurements on a polarized neutron reflectometer using a nonmagnetic waveguide structure to probe an amorphous magnetic microwire. The spin-resolved transmission of a neutron microbeam through a microwire was studied. The one dimensional mapping of the neutron precession across the wire was measured."
                    },
                    {
                        "title": "Scientific Review: The Triple Axis Spin-Echo Spectrometer TRISP at the FRM II",
                        "abstract": "TRISP is a novel high-resolution neutron spectrometer, combining the triple axis and neutron resonance spin echo (NRSE) techniques. The design of TRISP is optimized for the study of intrinsic linewidths of elementary excitations (phonons, magnons) with an energy resolution in the \u03bceV region over a broad range of momentum and energy transfers [1]. Compared to conventional triple axis spectrometers (TAS), this corresponds to an improvement of the energy resolution of one to two orders of magnitude."
                    }
                ]
            },
            "44284057-5606-4242-8106-258750b2046e": {
                "pk": "44284057-5606-4242-8106-258750b2046e",
                "name": "Lyle Muller",
                "collaborators": [
                    "T. Sejnowski",
                    "Christopher Steward",
                    "Roberto C. Budzinski",
                    "Zachary W. Davis",
                    "Alexandra Busch",
                    "John Reynolds",
                    "J. Martinez-Trujillo",
                    "T. A. Keller",
                    "Max Welling",
                    "Gabriel B. Benigno"
                ],
                "domain": [
                    "Neuroscience",
                    "Neural Dynamics",
                    "Machine Learning",
                    "Brain Stimulation"
                ],
                "publications": [
                    {
                        "title": "Horizontal cortical connections shape intrinsic traveling waves into feature-selective motifs that regulate perceptual sensitivity",
                        "abstract": "Intrinsic, ongoing fluctuations of cortical activity form traveling waves that modulate the gain of sensory-evoked responses and perceptual sensitivity. Several lines of evidence suggest that intrinsic traveling waves (iTWs) may arise, in part, from the coordination of synaptic activity through the recurrent horizontal connectivity within cortical areas, which include long range patchy connections that link neurons with shared feature preferences. In a spiking network model with anatomical topology that incorporates feature-selective patchy connections, which we call the Balanced Patchy Network (BPN), we observe repeated iTWs, which we refer to as motifs. In the model, motifs stem from fluctuations in the excitability of like-tuned neurons that result from shifts in E/I balance as action potentials traverse these patchy connections. To test if feature-selective motifs occur in vivo, we examined data previously recorded using multielectrode arrays in Area MT of marmosets trained to perform a threshold visual detection task. Using a newly developed method for comparing the similarity of wave patterns we found that some iTWs can be grouped into motifs. As predicted by the BPN, many of these motifs are feature selective, exhibiting direction-selective modulations in ongoing spiking activity. Further, motifs modulate the gain of the response evoked by a target and perceptual sensitivity to the target if the target matches the preference of the motif. These results provide evidence that iTWs are shaped by the patterns of horizontal fiber projections in the cortex and that patchy connections enable iTWs to regulate neural and perceptual sensitivity in a feature selective manner."
                    },
                    {
                        "title": "Focal control of non-invasive deep brain stimulation using multipolar temporal interference",
                        "abstract": "Temporal interference (TI) is a method of non-invasive brain stimulation using transcutaneous electrodes that allows the targeting and modulation of deeper brain structures, not normally associated with non-invasive simulation, while avoiding unwanted stimulation of shallower cortical structures. The properties of TI have been previously demonstrated, however, the problem of decoupling stimulation focality from stimulation intensity has not been addressed. In this paper, we provide a possible novel solution, multipolar TI (mTI), which allows increased independent control over both the size of the stimulated region and the stimulation intensity. The mTI method uses multiple carrier frequencies to create multiple overlapping amplitude-modulated envelopes, rather than using one envelope as in standard TI. The study presents an explanation of the concept of mTI along with experimental data gathered from Rhesus macaques and mice. We improved the focality at depth in anesthetized mice and monkeys, and using the new focality in awake monkeys, evoked targeted activity at depth in the superior colliculus. The mTI method could be an interesting and potentially useful new tool alongside other forms of non-invasive brain stimulation. Teaser: Multipolar Temporal Interference Stimulation can produce a more focal brain stimulation at depth compared to Temporal Interference."
                    },
                    {
                        "title": "Task-specific topographical maps of neural activity in the primate lateral prefrontal cortex",
                        "abstract": "Neurons in the primate lateral prefrontal cortex (LPFC) flexibly adapt their activity to support a wide range of cognitive tasks. Whether and how the topography of LPFC neural activity changes as a function of task is unclear. In the present study, we address this issue by characterizing the functional topography of LPFC neural activity in awake behaving macaques performing three distinct cognitive tasks. We recorded from chronically implanted multi-electrode arrays and show that the topography of LPFC activity is stable within a task, but adaptive across tasks. The topography of neural activity exhibits a spatial scale compatible with that of cortical columns and prior anatomical tracing work on afferent LPFC inputs. Our findings show that LPFC maps of neural population activity are stable for a specific task, providing robust neural codes that support task specialization. Moreover, the variability in functional topographies across tasks indicates activity landscapes can adapt, providing flexibility to LPFC neural codes."
                    },
                    {
                        "title": "A Translaminar Spacetime Code Supports Touch-Evoked Traveling Waves",
                        "abstract": "Linking sensory-evoked traveling waves to underlying circuit patterns is critical to understanding the neural basis of sensory perception. To form this link, we performed simultaneous electrophysiology and two-photon calcium imaging through transparent NeuroGrids and mapped touch-evoked cortical traveling waves and their underlying microcircuit dynamics. In awake mice, both passive and active whisker touch elicited traveling waves within and across barrels, with a fast early component followed by a variable late wave that lasted hundreds of milliseconds post-stimulus. Strikingly, late-wave dynamics were modulated by stimulus value and correlated with task performance. Mechanistically, the late wave component was i) modulated by motor feedback, ii) complemented by a sparse ensemble pattern across layer 2/3, which a balanced-state network model reconciled via inhibitory stabilization, and iii) aligned to regenerative Layer-5 apical dendritic Ca2+ events. Our results reveal a translaminar spacetime pattern organized by cortical feedback in the sensory cortex that supports touch-evoked traveling waves. GRAPHICAL ABSTRACT AND HIGHLIGHTS Whisker touch evokes both early- and late-traveling waves in the barrel cortex over 100\u2019s of milliseconds Reward reinforcement modulates wave dynamics Late wave emergence coincides with network sparsity in L23 and time-locked L5 dendritic Ca2+ spikes Experimental and computational results link motor feedback to distinct translaminar spacetime patterns"
                    },
                    {
                        "title": "A Spacetime Perspective on Dynamical Computation in Neural Information Processing Systems",
                        "abstract": "There is now substantial evidence for traveling waves and other structured spatiotemporal recurrent neural dynamics in cortical structures; but these observations have typically been difficult to reconcile with notions of topographically organized selectivity and feedforward receptive fields. We introduce a new 'spacetime' perspective on neural computation in which structured selectivity and dynamics are not contradictory but instead are complimentary. We show that spatiotemporal dynamics may be a mechanism by which natural neural systems encode approximate visual, temporal, and abstract symmetries of the world as conserved quantities, thereby enabling improved generalization and long-term working memory."
                    },
                    {
                        "title": "Transformers and cortical waves: encoders for pulling in context across time",
                        "abstract": "The capabilities of transformer networks such as ChatGPT and other large language models (LLMs) have captured the world's attention. The crucial computational mechanism underlying their performance relies on transforming a complete input sequence - for example, all the words in a sentence - into a long 'encoding vector' that allows transformers to learn long-range temporal dependencies in naturalistic sequences. Specifically, 'self-attention' applied to this encoding vector enhances temporal context in transformers by computing associations between pairs of words in the input sequence. We suggest that waves of neural activity traveling across single cortical areas, or multiple regions on the whole-brain scale, could implement a similar encoding principle. By encapsulating recent input history into a single spatial pattern at each moment in time, cortical waves may enable a temporal context to be extracted from sequences of sensory inputs, the same computational principle as that used in transformers."
                    },
                    {
                        "title": "Visual Cortical Traveling Waves: From Spontaneous Spiking Populations to Stimulus-Evoked Models of Short-Term Prediction",
                        "abstract": "Thanks to recent advances in neurotechnology, waves of activity sweeping across entire cortical regions are now routinely observed. Moreover, these waves have been found to impact neural responses as well as perception, and the responses themselves are found to be structured as traveling waves. How exactly do these waves arise? Do they confer any computational advantages? These traveling waves represent an opportunity for an expanded theory of neural computation, in which their dynamic local network activity may complement the moment-to-moment variability of our sensory experience. This thesis aims to help uncover the origin and role of traveling waves in the visual cortex through three Works. In Work 1, by simulating a network of conductance-based spiking neurons with realistically large network size and synaptic density, distance-dependent horizontal axonal time delays were found to be important for the widespread emergence of spontaneous traveling waves consistent with those in vivo. Furthermore, these waves were found to be a dynamic mechanism of gain modulation that may explain the in-vivo result of their modulation of perception. In Work 2, the Kuramoto oscillator model was formulated in the complex domain to study a network possessing distance-dependent time delays. Like in Work 1, these delays produced traveling waves, and the eigenspectrum of the complex-valued delayed matrix, containing a delay operator, provided an analytical explanation of them. In Work 3, the model from Work 2 was adapted into a recurrent neural network for the task of forecasting the frames of videos, with the question of how such a biologically constrained model may be useful in visual computation. We found that the wave activity emergent in this network was helpful, as they were tightly linked with high forecast performance, and shuffle controls revealed simultaneous abolishment of both the waves and performance. All together, these works shed light on the possible origins and uses of traveling waves in the visual cortex. In particular, time delays profoundly shape the spatiotemporal dynamics into traveling waves. This was confirmed numerically (Work 1) and analytically (Work 2). In Work 3, these waves were found to aid in the dynamic computation of visual forecasting."
                    },
                    {
                        "title": "Invited Session I: 30 years of normalization in the visual system: Spatiotemporal fluctuations in E/I balance & their impact on perception.",
                        "abstract": "The Normalization Model has helped illuminate our understanding of neural computations in a wide variety of contexts. Here, I will describe our recent efforts to understand how intrinsic fluctuations in the ratio of excitatory to inhibitory conductances play out in space and time, and how these fluctuations impact perception. We find that spontaneous cortical activity is organized into traveling waves that traverse visual cortex several times per second. Recording in Area MT of the common marmoset, we find that these intrinsic traveling waves (iTWs) regulate both the gain of the stimulus-evoked spiking response and the monkey's perceptual sensitivity. A large-scale spiking network model recapitulates the properties of iTWs measured in vivo and explains the observed fluctuations in response gain as resulting from momentary fluctuations in E/I balance. The model predicts that iTWs are sparse, in the sense that only a small fraction of the neural population participates in any individual iTW. As a result, iTWs can occur without inducing correlated variability, which we have found, in separate experiments, can impair sensory discrimination. We thus refer to the model as the sparse-wave model of iTWs. Taken together, these findings lead to the conclusion that traveling waves strongly modulate neural and perceptual sensitivity."
                    },
                    {
                        "title": "Dynamical properties and mechanisms of metastability: a perspective in neuroscience",
                        "abstract": "Metastability, characterized by a variability of regimes in time, is a ubiquitous type of neural dynamics. It has been formulated in many different ways in the neuroscience literature, however, which may cause some confusion. In this Perspective, we discuss metastability from the point of view of dynamical systems theory. We extract from the literature a very simple but general definition through the concept of metastable regimes as long-lived but transient epochs of activity with unique dynamical properties. This definition serves as an umbrella term that encompasses formulations from other works, and readily connects to concepts from dynamical systems theory. This allows us to examine general dynamical properties of metastable regimes, propose in a didactic manner several dynamics-based mechanisms that generate them, and discuss a theoretical tool to characterize them quantitatively. This perspective leads to insights that help to address issues debated in the literature and also suggest pathways for future research."
                    },
                    {
                        "title": "Algebraic approach to spike-time neural codes in the hippocampus.",
                        "abstract": "Although temporal coding through spike-time patterns has long been of interest in neuroscience, the specific structures that could be useful for spike-time codes remain highly unclear. Here, we introduce an analytical approach, using techniques from discrete mathematics, to study spike-time codes. As an initial example, we focus on the phenomenon of \"phase precession\" in the rodent hippocampus. During navigation and learning on a physical track, specific cells in a rodent's brain form a highly structured pattern relative to the oscillation of population activity in this region. Studies of phase precession largely focus on its role in precisely ordering spike times for synaptic plasticity, as the role of phase precession in memory formation is well established. Comparatively less attention has been paid to the fact that phase precession represents one of the best candidates for a spike-time neural code. The precise nature of this code remains an open question. Here, we derive an analytical expression for a function mapping points in physical space to complex-valued spikes by representing individual spike times as complex numbers. The properties of this function make explicit a specific relationship between past and future in spike patterns of the hippocampus. Importantly, this mathematical approach generalizes beyond the specific phenomenon studied here, providing a technique to study the neural codes within precise spike-time sequences found during sensory coding and motor behavior. We then introduce a spike-based decoding algorithm, based on this function, that successfully decodes a simulated animal's trajectory using only the animal's initial position and a pattern of spike times. This decoder is robust to noise in spike times and works on a timescale almost an order of magnitude shorter than typically used with decoders that work on average firing rate. These results illustrate the utility of a discrete approach, based on the structure and symmetries in spike patterns across finite sets of cells, to provide insight into the structure and function of neural systems."
                    },
                    {
                        "title": "Image segmentation with traveling waves in an exactly solvable recurrent neural network",
                        "abstract": "We study image segmentation using spatiotemporal dynamics in a recurrent neural network where the state of each unit is given by a complex number. We show that this network generates sophisticated spatiotemporal dynamics that can effectively divide an image into groups according to a scene's structural characteristics. Using an exact solution of the recurrent network's dynamics, we present a precise description of the mechanism underlying object segmentation in this network, providing a clear mathematical interpretation of how the network performs this task. We then demonstrate a simple algorithm for object segmentation that generalizes across inputs ranging from simple geometric objects in grayscale images to natural images. Object segmentation across all images is accomplished with one recurrent neural network that has a single, fixed set of weights. This demonstrates the expressive potential of recurrent neural networks when constructed using a mathematical approach that brings together their structure, dynamics, and computation."
                    },
                    {
                        "title": "Theory of transient chimeras in finite Sakaguchi-Kuramoto networks",
                        "abstract": "Chimera states are a phenomenon in which order and disorder can co-exist within a network that is fully homogeneous. Precisely how transient chimeras emerge in finite networks of Kuramoto oscillators with phase-lag remains unclear. Utilizing an operator-based framework to study nonlinear oscillator networks at finite scale, we reveal the spatiotemporal impact of the adjacency matrix eigenvectors on the Sakaguchi-Kuramoto dynamics. We identify a specific condition for the emergence of transient chimeras in these finite networks: the eigenvectors of the network adjacency matrix create a combination of a zero phase-offset mode and low spatial frequency waves traveling in opposite directions. This combination of eigenvectors leads directly to the coherent and incoherent clusters in the chimera. This approach provides two specific analytical predictions: (1) a precise formula predicting the combination of connectivity and phase-lag that creates transient chimeras, (2) a mathematical procedure for rewiring arbitrary networks to produce transient chimeras."
                    },
                    {
                        "title": "Predicting Stellar Mass Accretion: An Optimized Echo State Network Approach in Time Series Modeling",
                        "abstract": "Modeling the dynamics of the formation and evolution of protostellar disks as well as the history of stellar mass accretion typically involve the numerical solution of complex systems of coupled differential equations. The resulting mass accretion history of protostars is known to be highly episodic due to recurrent instabilities and also exhibits short timescale flickering. By leveraging the strong predictive abilities of neural networks, we extract some of the critical temporal dynamics experienced during the mass accretion including periods of instability. Particularly, we utilize a novel form of the Echo-State Neural Network (ESN), which has been shown to efficiently deal with data having inherent nonlinearity. We introduce the use of Optimized-ESN (Opt-ESN) to make model-independent time series forecasting of mass accretion rate in the evolution of protostellar disks. We apply the network to multiple hydrodynamic simulations with different initial conditions and exhibiting a variety of temporal dynamics to demonstrate the predictability of the Opt-ESN model. The model is trained on simulation data of $\\sim 1-2$ Myr, and achieves predictions with a low normalized mean square error ($\\sim 10^{-5}$ to $10^{-3}$) for forecasts ranging between 100 and 3800 yr. This result shows the promise of the application of machine learning based models to time-domain astronomy."
                    }
                ]
            },
            "1ad442cd-5375-4f71-9eee-38c768a7d982": {
                "pk": "1ad442cd-5375-4f71-9eee-38c768a7d982",
                "name": "Terrence Sejnowski",
                "collaborators": [
                    "T. Bartol",
                    "M. Samavat",
                    "L. Muller",
                    "Kristen M. Harris",
                    "H. Siegelmann",
                    "\u2020. KristenMHarris",
                    "Cailey Bromer",
                    "Dusten D. Hubbard",
                    "Dakota C. Hanka",
                    "M. Kuwajima"
                ],
                "domain": [
                    "Neuroscience",
                    "Synaptic Plasticity",
                    "Information Theory",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Long-Term Potentiation Produces a Sustained Expansion of Synaptic Information Storage Capacity in Adult Rat Hippocampus",
                        "abstract": "Long-term potentiation (LTP) has become a standard model for investigating synaptic mechanisms of learning and memory. Increasingly, it is of interest to understand how LTP affects the synaptic information storage capacity of the targeted population of synapses. Here, structural synaptic plasticity during LTP was explored using three-dimensional reconstruction from serial section electron microscopy. Storage capacity was assessed by applying a new analytical approach, Shannon information theory, to delineate the number of functionally distinguishable synaptic strengths. LTP was induced by delta-burst stimulation of perforant pathway inputs to the middle molecular layer of hippocampal dentate granule cells in adult rats. Spine head volumes were measured as predictors of synaptic strength and compared between LTP and control hemispheres at 30 min and 2 hr after the induction of LTP. Synapses from the same axon onto the same dendrite were used to determine the precision of synaptic plasticity based on the similarity of their physical dimensions. Shannon entropy was measured by exploiting the frequency of spine heads in functionally distinguishable sizes to assess the degree to which LTP altered the number of bits of information storage. Outcomes from these analyses reveal that LTP expanded storage capacity; the distribution of spine head volumes was increased from 2 bits in controls to 3 bits at 30 min and 2.7 bits at 2 hr after the induction of LTP. Furthermore, the distribution of spine head volumes was more uniform across the increased number of functionally distinguishable sizes following LTP, thus achieving more efficient use of coding space across the population of synapses. Significance Establishing relationships between structure, function, and information storage capacity provides a new approach to assessing network strength from structural measurements. Long term potentiation (LTP) is a standard model for investigating synaptic mechanisms of learning and memory. Information is a retrievable quantity that is being stored in synapses as synaptic strength and is correlated with multiple structural components of synaptic strength. Structural synaptic plasticity was measured in 3D reconstructions from serial section electron microscopy of spine head volume, as a proxy for synapse strength, at 30 min and 2 hr after LTP induction. Outcomes indicate that LTP enhances information storage capacity for at least 2 hr by increasing the precision of the synaptic structure and expanding the range of synapse sizes."
                    },
                    {
                        "title": "A spatial model of autophosphorylation of CaMKII in a glutamatergic spine suggests a network-driven kinetic mechanism for bistable changes in synaptic strength",
                        "abstract": "Long-term potentiation (LTP) is a biochemical process in excitatory glutamatergic synapses in the Central Nervous System (CNS). It is initiated by a bout of synaptic activation that is strong enough to contribute to production of an action potential in the axon of the postsynaptic neuron, and it results in an increase in the size of postsynaptic depolarization during subsequent activity. The first step leading to LTP is activation and autophosphorylation of an abundant postsynaptic enzyme, Ca2+/calmodulin-dependent protein kinase II (CaMKII). We use simulation of activation of CaMKII holoenzymes in a realistic spatial model of a spine synapse, created in MCell4, to test three hypotheses about how the autophosphorylation response of CaMKII is shaped during a repeated high-frequency stimulus. First, the simulation results indicate that autophosphorylation of CaMKII does not constitute a bistable switch under biologically realistic conditions. Instead, prolonged autophosphorylation of CaMKII may contribute to a biochemical \u201ckinetic proof-reading\u201d mechanism that controls induction of synaptic plasticity. Second, concentration of CaMKII near the postsynaptic membrane increases the local concentration of kinase activity. However, neither localization nor \u201cCa2+-calmodulin-trapping (CaM-trapping)\u201d increase the proportion of autophosphorylated subunits in holoenzymes after a complex stimulus, as previously hypothesized. Finally, we show that, as hypothesized, the amplitude of autophosphorylation in the first 30 seconds after a stimulus is extremely sensitive to the level and location of PP1 activity when PP1 is present in biologically accurate amounts. We further show that prolonged steric hindrance of dephosphorylation of CaMKII, caused by CaM-trapping, can increase the amplitude of autophosphorylation after a complex stimulus. These simulation results sharpen our quantitative understanding of the early events leading to LTP at excitatory synapses. Author Summary Neurons in the brain are interconnected in an organized fashion by synapses that transmit neuronal activity from one neuron to another. Most of the billions of neurons in the brain have about 10,000 synapses spread over the neuronal membrane. Information is stored in the brain when the ability of specific synapses to pass along neuronal activity is strengthened resulting in formation of new networks. The increase in strength of a synapse is tightly controlled by the frequency and amplitude of its activity, and by neurohormonal signals, which, in combination, can cause long-lasting biochemical changes at the synapse that underlie learning and memory. Defects in these biochemical pathways cause mental and neurological diseases. To develop treatments, we need to understand the precise choreography of these critical biochemical changes. However, the tiny size of the synaptic compartment makes precise measurements of the biochemical reactions impossible. We have used computer simulation techniques and information gathered from experiments on purified synaptic proteins to simulate, within a single synapse, the choreography of the first biochemical step in synaptic strengthening: activation of the enzyme Ca2+/ calmodulin-dependent protein kinase II. Our results provide insights that can be used in future studies to develop treatments for neuronal diseases."
                    },
                    {
                        "title": "Synaptic Information Storage Capacity Measured With Information Theory",
                        "abstract": "Abstract Variation in the strength of synapses can be quantified by measuring the anatomical properties of synapses. Quantifying precision of synaptic plasticity is fundamental to understanding information storage and retrieval in neural circuits. Synapses from the same axon onto the same dendrite have a common history of coactivation, making them ideal candidates for determining the precision of synaptic plasticity based on the similarity of their physical dimensions. Here, the precision and amount of information stored in synapse dimensions were quantified with Shannon information theory, expanding prior analysis that used signal detection theory (Bartol et al., 2015). The two methods were compared using dendritic spine head volumes in the middle of the stratum radiatum of hippocampal area CA1 as well-defined measures of synaptic strength. Information theory delineated the number of distinguishable synaptic strengths based on nonoverlapping bins of dendritic spine head volumes. Shannon entropy was applied to measure synaptic information storage capacity (SISC) and resulted in a lower bound of 4.1 bits and upper bound of 4.59 bits of information based on 24 distinguishable sizes. We further compared the distribution of distinguishable sizes and a uniform distribution using Kullback-Leibler divergence and discovered that there was a nearly uniform distribution of spine head volumes across the sizes, suggesting optimal use of the distinguishable values. Thus, SISC provides a new analytical measure that can be generalized to probe synaptic strengths and capacity for plasticity in different brain regions of different species and among animals raised in different conditions or during learning. How brain diseases and disorders affect the precision of synaptic plasticity can also be probed."
                    },
                    {
                        "title": "A Spacetime Perspective on Dynamical Computation in Neural Information Processing Systems",
                        "abstract": "There is now substantial evidence for traveling waves and other structured spatiotemporal recurrent neural dynamics in cortical structures; but these observations have typically been difficult to reconcile with notions of topographically organized selectivity and feedforward receptive fields. We introduce a new 'spacetime' perspective on neural computation in which structured selectivity and dynamics are not contradictory but instead are complimentary. We show that spatiotemporal dynamics may be a mechanism by which natural neural systems encode approximate visual, temporal, and abstract symmetries of the world as conserved quantities, thereby enabling improved generalization and long-term working memory."
                    },
                    {
                        "title": "Transformers and cortical waves: encoders for pulling in context across time",
                        "abstract": "The capabilities of transformer networks such as ChatGPT and other large language models (LLMs) have captured the world's attention. The crucial computational mechanism underlying their performance relies on transforming a complete input sequence - for example, all the words in a sentence - into a long 'encoding vector' that allows transformers to learn long-range temporal dependencies in naturalistic sequences. Specifically, 'self-attention' applied to this encoding vector enhances temporal context in transformers by computing associations between pairs of words in the input sequence. We suggest that waves of neural activity traveling across single cortical areas, or multiple regions on the whole-brain scale, could implement a similar encoding principle. By encapsulating recent input history into a single spatial pattern at each moment in time, cortical waves may enable a temporal context to be extracted from sequences of sensory inputs, the same computational principle as that used in transformers."
                    },
                    {
                        "title": "Hidden Traveling Waves bind Working Memory Variables in Recurrent Neural Networks",
                        "abstract": "Traveling waves are a fundamental phenomenon in the brain, playing a crucial role in short-term information storage. In this study, we leverage the concept of traveling wave dynamics within a neural lattice to formulate a theoretical model of neural working memory, study its properties, and its real world implications in AI. The proposed model diverges from traditional approaches, which assume information storage in static, register-like locations updated by interference. Instead, the model stores data as waves that is updated by the wave's boundary conditions. We rigorously examine the model's capabilities in representing and learning state histories, which are vital for learning history-dependent dynamical systems. The findings reveal that the model reliably stores external information and enhances the learning process by addressing the diminishing gradient problem. To understand the model's real-world applicability, we explore two cases: linear boundary condition (LBC) and non-linear, self-attention-driven boundary condition (SBC). The model with the linear boundary condition results in a shift matrix plus low-rank matrix currently used in H3 state space RNN. Further, our experiments with LBC reveal that this matrix is effectively learned by Recurrent Neural Networks (RNNs) through backpropagation when modeling history-dependent dynamical systems. Conversely, the SBC parallels the autoregressive loop of an attention-only transformer with the context vector representing the wave substrate. Collectively, our findings suggest the broader relevance of traveling waves in AI and its potential in advancing neural network architectures."
                    },
                    {
                        "title": "Visual Cortical Traveling Waves: From Spontaneous Spiking Populations to Stimulus-Evoked Models of Short-Term Prediction",
                        "abstract": "Thanks to recent advances in neurotechnology, waves of activity sweeping across entire cortical regions are now routinely observed. Moreover, these waves have been found to impact neural responses as well as perception, and the responses themselves are found to be structured as traveling waves. How exactly do these waves arise? Do they confer any computational advantages? These traveling waves represent an opportunity for an expanded theory of neural computation, in which their dynamic local network activity may complement the moment-to-moment variability of our sensory experience. This thesis aims to help uncover the origin and role of traveling waves in the visual cortex through three Works. In Work 1, by simulating a network of conductance-based spiking neurons with realistically large network size and synaptic density, distance-dependent horizontal axonal time delays were found to be important for the widespread emergence of spontaneous traveling waves consistent with those in vivo. Furthermore, these waves were found to be a dynamic mechanism of gain modulation that may explain the in-vivo result of their modulation of perception. In Work 2, the Kuramoto oscillator model was formulated in the complex domain to study a network possessing distance-dependent time delays. Like in Work 1, these delays produced traveling waves, and the eigenspectrum of the complex-valued delayed matrix, containing a delay operator, provided an analytical explanation of them. In Work 3, the model from Work 2 was adapted into a recurrent neural network for the task of forecasting the frames of videos, with the question of how such a biologically constrained model may be useful in visual computation. We found that the wave activity emergent in this network was helpful, as they were tightly linked with high forecast performance, and shuffle controls revealed simultaneous abolishment of both the waves and performance. All together, these works shed light on the possible origins and uses of traveling waves in the visual cortex. In particular, time delays profoundly shape the spatiotemporal dynamics into traveling waves. This was confirmed numerically (Work 1) and analytically (Work 2). In Work 3, these waves were found to aid in the dynamic computation of visual forecasting."
                    },
                    {
                        "title": "Using Shannon Information to Probe the Precision of Synaptic Strengths",
                        "abstract": "Synapses between neurons control the the strengths of neuronal communication in neural circuits and their strengths are in turn dynamically regulated by experience. Because dendritic spine head volumes are highly correlated synaptic strength [1]"
                    },
                    {
                        "title": "Neuromodulators generate multiple context-relevant behaviors in a recurrent neural network by shifting activity hypertubes",
                        "abstract": "1 Mood, arousal, and other internal states can drastically alter behavior, even in identical external 2 circumstances \u2014 a cold glass of water when you are thirsty is much more desirable than when 3 you are sated. Neuromodulators are critical controllers of such neural states, with dysfunctions 4 linked to various neuropsychiatric disorders. Although biological aspects of neuromodulation have 5 been well studied, the computational principles underlying how large-scale neuromodulation of dis-6 tributed neural populations shifts brain states remain unclear. We use recurrent neural networks 7 to model how synaptic weight modulation \u2014 an important function of neuromodulators \u2014 can 8 achieve nuanced alterations in neural computation, even in a highly simplified form. We find that 9 under structural constraints like those in brains, this provides a fundamental mechanism that can 10 increase the computational capability and flexibility of a neural network by enabling overlapping 11 storage of synaptic memories able to generate diverse, even diametrically opposed, behaviors. Our"
                    },
                    {
                        "title": "Exploring The Precision of Real Intelligence at Synapse Resolution",
                        "abstract": "Synapses are the fundamental units of storage of information in neural circuits and their structure and strength are adjusted through synaptic plasticity. Hence, exploring different aspects of synaptic plasticity processes in the hippocampus is crucial to understanding mechanisms of learning and memory, improving artificial intelligence algorithms, and neuromorphic computers. The scope of this manuscript is to explore the precision of synaptic plasticity. Here we measured the precision of multiple synaptic features (Spine head volume, post synaptic density area, spine neck diameter, spine neck length and number of docked vesicles). Results show synaptic plasticity is highly precise and sub cellular resources such as mitochondria may have impact on it. ."
                    },
                    {
                        "title": "A Unifying Information-Theoretic Framework for Independent Component Analysis",
                        "abstract": "-we show that different theories recently proposed for independent component analysis (ICA) lead to the same iterative learning algorithm for blind separation of mixed independent sources. We review those theories and suggest that information theory can be used to unify several lines of research. Pearlmutter and Parra [I] and Cardoso [2] showed that the infomax approach of Bell and Sejnowski [3] and the maximum likelihood estimation approach are equivalent. We show that negentropy maximization also has equivalent properties, and therefore, all three approaches yield the same learning rule for a fixed nonlinearity. Girolami and Fyfe [4] have shown that the nonlinear principal component analysis (PCA) algorithm of Karhunen and Joutsensalo [5] and Oja [6] can also be viewed from information-theoretic principles since it minimizes the sum of squares of the fourth-order marginal cumulants, and therefore, approximately minimizes the mutual information [7]. Lambert [8] has proposed different Bussgang cost functions for multichannel blind deconvolution. We show how the Bussgang property relates to the infomax principle. Finally, we discuss convergence and stability as well as future research issues in blind source separation. @ 2000 Elsevier Science Ltd. All rights reserved."
                    }
                ]
            },
            "0e5cce41-e26d-4a1d-9d24-46ef72865871": {
                "pk": "0e5cce41-e26d-4a1d-9d24-46ef72865871",
                "name": "Max Welling",
                "collaborators": [
                    "Yue Song",
                    "T. A. Keller",
                    "Thomas Anderson Keller",
                    "Yisong Yue",
                    "Pietro Perona",
                    "L. Muller",
                    "T. Sejnowski",
                    "N. Sebe",
                    "Jakub M. Tomczak",
                    "Maximilian Ilse"
                ],
                "domain": [
                    "Representation Learning",
                    "Neural Networks",
                    "Medical Imaging",
                    "Equivariance"
                ],
                "publications": [
                    {
                        "title": "Unsupervised Representation Learning from Sparse Transformation Analysis",
                        "abstract": "There is a vast literature on representation learning based on principles such as coding efficiency, statistical independence, causality, controllability, or symmetry. In this paper we propose to learn representations from sequence data by factorizing the transformations of the latent variables into sparse components. Input data are first encoded as distributions of latent activations and subsequently transformed using a probability flow model, before being decoded to predict a future input state. The flow model is decomposed into a number of rotational (divergence-free) vector fields and a number of potential flow (curl-free) fields. Our sparsity prior encourages only a small number of these fields to be active at any instant and infers the speed with which the probability flows along these fields. Training this model is completely unsupervised using a standard variational objective and results in a new form of disentangled representations where the input is not only represented by a combination of independent factors, but also by a combination of independent transformation primitives given by the learned flow fields. When viewing the transformations as symmetries one may interpret this as learning approximately equivariant representations. Empirically we demonstrate that this model achieves state of the art in terms of both data likelihood and unsupervised approximate equivariance errors on datasets composed of sequence transformations."
                    },
                    {
                        "title": "A Spacetime Perspective on Dynamical Computation in Neural Information Processing Systems",
                        "abstract": "There is now substantial evidence for traveling waves and other structured spatiotemporal recurrent neural dynamics in cortical structures; but these observations have typically been difficult to reconcile with notions of topographically organized selectivity and feedforward receptive fields. We introduce a new 'spacetime' perspective on neural computation in which structured selectivity and dynamics are not contradictory but instead are complimentary. We show that spatiotemporal dynamics may be a mechanism by which natural neural systems encode approximate visual, temporal, and abstract symmetries of the world as conserved quantities, thereby enabling improved generalization and long-term working memory."
                    },
                    {
                        "title": "Flow Factorized Representation Learning",
                        "abstract": "A prominent goal of representation learning research is to achieve representations which are factorized in a useful manner with respect to the ground truth factors of variation. The fields of disentangled and equivariant representation learning have approached this ideal from a range of complimentary perspectives; however, to date, most approaches have proven to either be ill-specified or insufficiently flexible to effectively separate all realistic factors of interest in a learned latent space. In this work, we propose an alternative viewpoint on such structured representation learning which we call Flow Factorized Representation Learning, and demonstrate it to learn both more efficient and more usefully structured representations than existing frameworks. Specifically, we introduce a generative model which specifies a distinct set of latent probability paths that define different input transformations. Each latent flow is generated by the gradient field of a learned potential following dynamic optimal transport. Our novel setup brings new understandings to both \\textit{disentanglement} and \\textit{equivariance}. We show that our model achieves higher likelihoods on standard representation learning benchmarks while simultaneously being closer to approximately equivariant models. Furthermore, we demonstrate that the transformations learned by our model are flexibly composable and can also extrapolate to new data, implying a degree of robustness and generalizability approaching the ultimate goal of usefully factorized representation learning."
                    },
                    {
                        "title": "Topographic VAEs learn Equivariant Capsules",
                        "abstract": "In this work we seek to bridge the concepts of topographic organization and equivariance in neural networks. To accomplish this, we introduce the Topographic VAE: a novel method for efficiently training deep generative models with topographically organized latent variables. We show that such a model indeed learns to organize its activations according to salient characteristics such as digit class, width, and style on MNIST. Furthermore, through topographic organization over time (i.e. temporal coherence), we demonstrate how predefined latent space transformation operators can be encouraged for observed transformed input sequences -- a primitive form of unsupervised learned equivariance. We demonstrate that this model successfully learns sets of approximately equivariant features (i.e.\"capsules\") directly from sequences and achieves higher likelihood on correspondingly transforming test sequences. Equivariance is verified quantitatively by measuring the approximate commutativity of the inference network and the sequence transformations. Finally, we demonstrate approximate equivariance to complex transformations, expanding upon the capabilities of existing group equivariant neural networks."
                    },
                    {
                        "title": "UvA-DARE (Digital Academic Repository) Deep Learning with Order-invariant Operator for Multi-instance Histopathology Classification",
                        "abstract": "The computer-aided analysis of medical scans is a longstanding goal in the medical imaging \ufb01eld. Currently, deep learning has became a dominant methodology for supporting pathologists and radiologist. Deep learning algorithms have been successfully applied to digital pathology and radiology, nevertheless, there are still practical issues that prevent these tools to be widely used in practice. The main obstacles are low number of available cases and large size of images (a.k.a. the small n, large p problem in machine learning), and a very limited access to annotation at a pixel level that can lead to severe over\ufb01tting and large computational requirements. We propose to handle these issues by introducing a framework that processes a medical image as a collection of small patches using a single, shared neural network. The \ufb01nal diagnosis is provided by combining scores of individual patches using a permutation-invariant operator (combination). In machine learning community such approach is called a multi-instance learning (MIL)."
                    }
                ]
            }
        }
    },
    "2409.19422": {
        "paper_data": {
            "title": "Identifiable Shared Component Analysis of Unpaired Multimodal Mixtures",
            "url": "http://arxiv.org/abs/2409.19422v2",
            "arxiv_id": "2409.19422",
            "authors": [
                "Subash Timilsina",
                "Sagar Shrestha",
                "Xiao Fu"
            ],
            "abstract": "A core task in multi-modal learning is to integrate information from multiple feature spaces (e.g., text and audio), offering modality-invariant essential representations of data. Recent research showed that, classical tools such as {\\it canonical correlation analysis} (CCA) provably identify the shared components up to minor ambiguities, when samples in each modality are generated from a linear mixture of shared and private components. Such identifiability results were obtained under the condition that the cross-modality samples are aligned/paired according to their shared information. This work takes a step further, investigating shared component identifiability from multi-modal linear mixtures where cross-modality samples are unaligned. A distribution divergence minimization-based loss is proposed, under which a suite of sufficient conditions ensuring identifiability of the shared components are derived. Our conditions are based on cross-modality distribution discrepancy characterization and density-preserving transform removal, which are much milder than existing studies relying on independent component analysis. More relaxed conditions are also provided via adding reasonable structural constraints, motivated by available side information in various applications. The identifiability claims are thoroughly validated using synthetic and real-world data.",
            "introduction": "   1 Introduction  The same data entities can often be represented in different feature spaces (e.g., audio, text and image), due to the variety of sensing modalities. Learning common latent components of data from multiple modalities is well-motivated in representation learning. The shared components are considered modality-invariant essential representations of data, which can often enhance performance of downstream tasks by shedding modality-specific noise [1, 2, 3, 4] and avoiding over-fitting [5, 6, 7].   A prominent theoretical aspect of shared component learning lies in identifiability of the components of interest. The literature posed an intriguing theoretical question [1, 2, 8]: If every modality of data is represented by a linear mixture of shared and private components with an unknown mixing system, are the shared components identifiable (up to acceptable ambiguities)? Such component identification problems are often nontrivial due to the ill-posed nature of any linear mixture model (see, e.g., [9, 10, 11, 12, 13, 14]). Interestingly, the work [1] showed that using the classical canonical correlation analysis (CCA) provably find the shared components up to rotation and scaling. In fact, shared component identification from multimodal/multiview linear mixtures were considered in various contexts (see, e.g., [15, 16, 17, 18]), although some of these works did not model private components. The identifiability results in [1, 2] were generalized to nonlinear mixture models as well [4, 19]. The shared component identification perspective was also related to the success of representation learning in self-supervised learning (SSL) [6, 5, 7].   Nonetheless, the treatment in [1, 2] and the related works [17, 16, 15] all assumed that the cross-modality data are aligned (i.e., paired) according to their shared components. In many applications, such as cross-language information retrieval [20, 21, 22], domain adaptation [23, 24, 25], and biological data translation [26, 27], aligned cross-modality data are hard to acquire, if not outright unavailable. A natural question is: When the multimodal linear mixtures are unaligned, can the shared latent components still be provably identified under reasonably mild conditions?   Existing Studies. Theoretical characteristics of unaligned multimodal learning were studied under various settings. The work [28] considered a case where one modality is a linear transform of another modality, and showed that the linear transformation is potentially identifiable. The recent work [29] extended this model to a nonlinear transform setting. However, these works did not consider latent component models\u2014yet the latter are more versatile in many ways, e.g., facilitating one-to-many translations [30, 31]. The work [32] considered unaligned mixtures of shared and private components, but the assumptions (e.g., the availability of a large amount of modalities) to ensure identifiability may not be easy to satisfy. The most related work is perhaps [8]. But their approach also relied on somewhat stringent assumptions, e.g., that all the latent components are element-wise statistically independent with at most one component being Gaussian. This is because their procedure had to invoke the classical independent component analysis (ICA) [33].   Contributions. In this work, we provide a suite of sufficient conditions under which the shared components can be provably identified from unaligned multimodal linear mixtures up to reasonable ambiguities. The model and identification problem are referred to as unaligned shared component analysis (unaligned SCA) in the sequel.   (i) An Identifiable Learning Loss for",
            "references": [
                {
                    "title": "Towards Identifiable Unsupervised Domain Translation: A Diversified Distribution Matching Approach",
                    "abstract": "Unsupervised domain translation (UDT) aims to find functions that convert samples from one domain (e.g., sketches) to another domain (e.g., photos) without changing the high-level semantic meaning (also referred to as ``content''). The translation functions are often sought by probability distribution matching of the transformed source domain and target domain. CycleGAN stands as arguably the most representative approach among this line of work. However, it was noticed in the literature that CycleGAN and variants could fail to identify the desired translation functions and produce content-misaligned translations. This limitation arises due to the presence of multiple translation functions -- referred to as ``measure-preserving automorphism\"(MPA) -- in the solution space of the learning criteria. Despite awareness of such identifiability issues, solutions have remained elusive. This study delves into the core identifiability inquiry and introduces an MPA elimination theory. Our analysis shows that MPA is unlikely to exist, if multiple pairs of diverse cross-domain conditional distributions are matched by the learning function. Our theory leads to a UDT learner using distribution matching over auxiliary variable-induced subsets of the domains -- other than over the entire data domains as in the classical approaches. The proposed framework is the first to rigorously establish translation identifiability under reasonable UDT settings, to our best knowledge. Experiments corroborate with our theoretical claims."
                },
                {
                    "title": "Revisiting Deep Generalized Canonical Correlation Analysis",
                    "abstract": "Canonical correlation analysis (CCA) is a classic statistical method for discovering latent co-variation that underpins two or more observed random vectors. Several extensions and variations of CCA have been proposed that have strengthened our capabilities in terms of revealing common random factors from multiview datasets. In this work, we first revisit the most recent deterministic extensions of deep CCA and highlight the strengths and limitations of these state-of-the-art methods. Some methods allow trivial solutions, while others can miss weak common factors. Others overload the problem by also seeking to reveal what is not common among the views \u2013 i.e., the private components that are needed to fully reconstruct each view. The latter tends to overload the problem and its computational and sample complexities. Aiming to improve upon these limitations, we design a novel and efficient formulation that alleviates some of the current restrictions. The main idea is to model the private components as conditionally independent given the common ones, which enables the proposed compact formulation. In addition, we also provide a sufficient condition for identifying the common random factors. Judicious experiments with synthetic and real datasets showcase the validity of our claims and the effectiveness of the proposed approach."
                },
                {
                    "title": "Unbalancedness in Neural Monge Maps Improves Unpaired Domain Translation",
                    "abstract": "In optimal transport (OT), a Monge map is known as a mapping that transports a source distribution to a target distribution in the most cost-efficient way. Recently, multiple neural estimators for Monge maps have been developed and applied in diverse unpaired domain translation tasks, e.g. in single-cell biology and computer vision. However, the classic OT framework enforces mass conservation, which makes it prone to outliers and limits its applicability in real-world scenarios. The latter can be particularly harmful in OT domain translation tasks, where the relative position of a sample within a distribution is explicitly taken into account. While unbalanced OT tackles this challenge in the discrete setting, its integration into neural Monge map estimators has received limited attention. We propose a theoretically grounded method to incorporate unbalancedness into any Monge map estimator. We improve existing estimators to model cell trajectories over time and to predict cellular responses to perturbations. Moreover, our approach seamlessly integrates with the OT flow matching (OT-FM) framework. While we show that OT-FM performs competitively in image translation, we further improve performance by incorporating unbalancedness (UOT-FM), which better preserves relevant features. We hence establish UOT-FM as a principled method for unpaired image translation."
                },
                {
                    "title": "Self-Supervised Disentanglement by Leveraging Structure in Data Augmentations",
                    "abstract": "Self-supervised representation learning often uses data augmentations to induce some invariance to\"style\"attributes of the data. However, with downstream tasks generally unknown at training time, it is difficult to deduce a priori which attributes of the data are indeed\"style\"and can be safely discarded. To deal with this, current approaches try to retain some style information by tuning the degree of invariance to some particular task, such as ImageNet object classification. However, prior work has shown that such task-specific tuning can lead to significant performance degradation on other tasks that rely on the discarded style. To address this, we introduce a more principled approach that seeks to disentangle style features rather than discard them. The key idea is to add multiple style embedding spaces where: (i) each is invariant to all-but-one augmentation; and (ii) joint entropy is maximized. We formalize our structured data-augmentation procedure from a causal latent-variable-model perspective, and prove identifiability of both content and individual style variables. We empirically demonstrate the benefits of our approach on both synthetic and real-world data."
                },
                {
                    "title": "Unpaired Multi-Domain Causal Representation Learning",
                    "abstract": "The goal of causal representation learning is to find a representation of data that consists of causally related latent variables. We consider a setup where one has access to data from multiple domains that potentially share a causal representation. Crucially, observations in different domains are assumed to be unpaired, that is, we only observe the marginal distribution in each domain but not their joint distribution. In this paper, we give sufficient conditions for identifiability of the joint distribution and the shared causal graph in a linear setup. Identifiability holds if we can uniquely recover the joint distribution and the shared causal representation from the marginal distributions in each domain. We transform our identifiability results into a practical method to recover the shared latent causal graph."
                },
                {
                    "title": "Polytopic Matrix Factorization: Determinant Maximization Based Criterion and Identifiability",
                    "abstract": "We introduce Polytopic Matrix Factorization (PMF) as a novel data decomposition approach. In this new framework, we model input data as unknown linear transformations of some latent vectors drawn from a polytope. In this sense, the article considers a semi-structured data model, in which the input matrix is modeled as the product of a full column rank matrix and a matrix containing samples from a polytope as its column vectors. The choice of polytope reflects the presumed features of the latent components and their mutual relationships. As the factorization criterion, we propose the determinant maximization (Det-Max) for the sample autocorrelation matrix of the latent vectors. We introduce a sufficient condition for identifiability, which requires that the convex hull of the latent vectors contains the maximum volume inscribed ellipsoid of the polytope with a particular tightness constraint. Based on the Det-Max criterion and the proposed identifiability condition, we show that all polytopes that satisfy a particular symmetry restriction qualify for the PMF framework. Having infinitely many polytope choices provides a form of flexibility in characterizing latent vectors. In particular, it is possible to define latent vectors with heterogeneous features, enabling the assignment of attributes such as nonnegativity and sparsity at the subvector level. The article offers examples illustrating the connection between polytope choices and the corresponding feature representations."
                },
                {
                    "title": "Shared Independent Component Analysis for Multi-Subject Neuroimaging",
                    "abstract": "We consider shared response modeling, a multi-view learning problem where one wants to identify common components from multiple datasets or views. We introduce Shared Independent Component Analysis (ShICA) that models each view as a linear transform of shared independent components contaminated by additive Gaussian noise. We show that this model is identifiable if the components are either non-Gaussian or have enough diversity in noise variances. We then show that in some cases multi-set canonical correlation analysis can recover the correct unmixing matrices, but that even a small amount of sampling noise makes Multiset CCA fail. To solve this problem, we propose to use joint diagonalization after Multiset CCA, leading to a new approach called ShICA-J. We show via simulations that ShICA-J leads to improved results while being very fast to fit. While ShICA-J is based on second-order statistics, we further propose to leverage non-Gaussianity of the components using a maximum-likelihood method, ShICA-ML, that is both more accurate and more costly. Further, ShICA comes with a principled method for shared components estimation. Finally, we provide empirical evidence on fMRI and MEG datasets that ShICA yields more accurate estimation of the components than alternatives."
                },
                {
                    "title": "Unsupervised Real-World Super-Resolution: A Domain Adaptation Perspective",
                    "abstract": "Most existing convolution neural network (CNN) based super-resolution (SR) methods generate their paired training dataset by artificially synthesizing low-resolution (LR) images from the high-resolution (HR) ones. However, this dataset preparation strategy harms the application of these CNNs in real-world scenarios due to the inherent domain gap between the training and testing data. A popular attempts towards the challenge is unpaired generative adversarial networks, which generate \"real\" LR counterparts from real HR images using image-to-image translation and then perform super-resolution from \"real\" LR\u2192SR. Despite great progress, it is still difficult to synthesize perfect \"real\" LR images for super-resolution. In this paper, we firstly consider the real-world SR problem from the traditional domain adaptation perspective. We propose a novel unpaired SR training framework based on feature distribution alignment, with which we can obtain degradation-indistinguishable feature maps and then map them to HR images. In order to generate better SR images for target LR domain, we introduce several regularization losses to force the aligned feature to locate around the target domain. Our experiments indicate that our SR network obtains the state-of-the-art performance over both blind and unpaired SR methods on diverse datasets."
                },
                {
                    "title": "Understanding Latent Correlation-Based Multiview Learning and Self-Supervision: An Identifiability Perspective",
                    "abstract": "Multiple views of data, both naturally acquired (e.g., image and audio) and artificially produced (e.g., via adding different noise to data samples), have proven useful in enhancing representation learning. Natural views are often handled by multiview analysis tools, e.g., (deep) canonical correlation analysis [(D)CCA], while the artificial ones are frequently used in self-supervised learning (SSL) paradigms, e.g., BYOL and Barlow Twins. Both types of approaches often involve learning neural feature extractors such that the embeddings of data exhibit high cross-view correlations. Although intuitive, the effectiveness of correlation-based neural embedding is mostly empirically validated. This work aims to understand latent correlation maximization-based deep multiview learning from a latent component identification viewpoint. An intuitive generative model of multiview data is adopted, where the views are different nonlinear mixtures of shared and private components. Since the shared components are view/distortion-invariant, representing the data using such components is believed to reveal the identity of the samples effectively and robustly. Under this model, latent correlation maximization is shown to guarantee the extraction of the shared components across views (up to certain ambiguities). In addition, it is further shown that the private information in each view can be provably disentangled from the shared using proper regularization design. A finite sample analysis, which has been rare in nonlinear mixture identifiability study, is also presented. The theoretical results and newly designed regularization are tested on a series of tasks."
                },
                {
                    "title": "Self-Supervised Learning with Data Augmentations Provably Isolates Content from Style",
                    "abstract": "Self-supervised representation learning has shown remarkable success in a number of domains. A common practice is to perform data augmentation via hand-crafted transformations intended to leave the semantics of the data invariant. We seek to understand the empirical success of this approach from a theoretical perspective. We formulate the augmentation process as a latent variable model by postulating a partition of the latent representation into a content component, which is assumed invariant to augmentation, and a style component, which is allowed to change. Unlike prior work on disentanglement and independent component analysis, we allow for both nontrivial statistical and causal dependencies in the latent space. We study the identifiability of the latent representation based on pairs of views of the observations and prove sufficient conditions that allow us to identify the invariant content partition up to an invertible mapping in both generative and discriminative settings. We find numerical simulations with dependent latent variables are consistent with our theory. Lastly, we introduce Causal3DIdent, a dataset of high-dimensional, visually complex images with rich causal dependencies, which we use to study the effect of data augmentations performed in practice."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Nonlinear Multiview Analysis: Identifiability and Neural Network-Assisted Implementation",
                    "abstract": "Multiview analysis aims at extracting shared latent components from data samples that are acquired in different domains, e.g., image, text, and audio. Classic multiview analysis, e.g., canonical correlation analysis (CCA), tackles this problem via matching the linearly transformed views in a certain latent domain. More recently, powerful nonlinear learning tools such as kernel methods and neural networks are utilized for enhancing the classic CCA. However, unlike linear CCA whose theoretical aspects are clearly understood, nonlinear CCA approaches are largely intuition-driven. In particular, it is unclear under what conditions the shared latent components across the views can be identified\u2014while identifiability plays an essential role in many applications. In this work, we revisit nonlinear multiview analysis and address both the theoretical and computational aspects. Our work leverages a useful nonlinear model, namely, the post-nonlinear model, from the nonlinear mixture separation literature. Combining with multiview data, we take a nonlinear multiview mixture learning viewpoint, which is a natural extension of the classic generative models for linear CCA. From there, we derive a learning criterion. We show that minimizing this criterion leads to identification of the latent shared components up to certain ambiguities, under reasonable conditions. Our derivation and formulation also offer new insights and interpretations to existing deep neural network-based CCA formulations. On the computation side, we propose an effective algorithm with simple and scalable update rules. A series of simulations and real-data experiments corroborate our theoretical analysis."
                },
                {
                    "title": "Cell-Edge Detection via Selective Cooperation and Generalized Canonical Correlation",
                    "abstract": "Improving the uplink quality of service for users located around the boundaries between cells is a key challenge in cellular systems. Existing approaches relying on power control throttle the rates of cell-center users, while multi-user detection requires accurate channel estimates for the cell-edge users, which is another challenge due to their low received signal-to-noise ratio (SNR). Utilizing the fact that cell-edge user signals are weak but common (received at roughly equal power) at different base stations (BSs), this paper establishes a connection between cell-edge user detection and generalized canonical correlation analysis (GCCA). It puts forth a GCCA-based method that leverages selective BS cooperation to recover the cell-edge user signal subspace even at low SNR. The cell-edge user signals can then be extracted from the resulting mixture via algebraic signal processing techniques. The paper includes theoretical analysis showing why GCCA recovers the correct subspace containing the cell-edge user signals under mild conditions. The proposed method can also identify the number of cell-edge users in the system, i.e., the common subspace dimension. Simulations reveal significant performance improvement relative to various multiuser detection techniques. Cell-edge detection performance is further studied as a function of how many / which BSs are selected, and it is shown that using the closest three BS is always the best choice."
                },
                {
                    "title": "Generalized Canonical Correlation Analysis: A Subspace Intersection Approach",
                    "abstract": "Generalized Canonical Correlation Analysis (GCCA) is an important tool that finds numerous applications in data mining, machine learning, and artificial intelligence. It aims at finding \u2018common\u2019 random variables that are strongly correlated across multiple feature representations (views) of the same set of entities. CCA and to a lesser extent GCCA have been studied from the statistical and algorithmic points of view, but not as much from the standpoint of linear algebra. This paper offers a fresh algebraic perspective on GCCA based on a (bi-)linear generative model that naturally captures its essence. It is shown that from a linear algebra point of view, GCCA is tantamount to subspace intersection; and conditions under which the common subspace of the different views is identifiable are provided. A novel GCCA algorithm is proposed based on subspace intersection, which scales up to handle large GCCA tasks. Synthetic as well as real data experiments are provided to showcase the effectiveness of the proposed approach."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "Kernel of CycleGAN as a Principle homogeneous space",
                    "abstract": "Unpaired image-to-image translation has attracted significant interest due to the invention of CycleGAN, a method which utilizes a combination of adversarial and cycle consistency losses to avoid the need for paired data. It is known that the CycleGAN problem might admit multiple solutions, and our goal in this paper is to analyze the space of exact solutions and to give perturbation bounds for approximate solutions. We show theoretically that the exact solution space is invariant with respect to automorphisms of the underlying probability spaces, and, furthermore, that the group of automorphisms acts freely and transitively on the space of exact solutions. We examine the case of zero `pure' CycleGAN loss first in its generality, and, subsequently, expand our analysis to approximate solutions for `extended' CycleGAN loss where identity loss term is included. In order to demonstrate that these results are applicable, we show that under mild conditions nontrivial smooth automorphisms exist. Furthermore, we provide empirical evidence that neural networks can learn these automorphisms with unexpected and unwanted results. We conclude that finding optimal solutions to the CycleGAN loss does not necessarily lead to the envisioned result in image-to-image translation tasks and that underlying hidden symmetries can render the result utterly useless."
                },
                {
                    "title": "StarGAN v2: Diverse Image Synthesis for Multiple Domains",
                    "abstract": "A good image-to-image translation model should learn a mapping between different visual domains while satisfying the following properties: 1) diversity of generated images and 2) scalability over multiple domains. Existing methods address either of the issues, having limited diversity or multiple models for all domains. We propose StarGAN v2, a single framework that tackles both and shows significantly improved results over the baselines. Experiments on CelebA-HQ and a new animal faces dataset (AFHQ) validate our superiority in terms of visual quality, diversity, and scalability. To better assess image-to-image translation models, we release AFHQ, high-quality animal faces with large inter- and intra-domain differences. The code, pretrained models, and dataset are available at https://github.com/clovaai/stargan-v2."
                },
                {
                    "title": "The Incomplete Rosetta Stone problem: Identifiability results for Multi-view Nonlinear ICA",
                    "abstract": "We consider the problem of recovering a common latent source with independent components from multiple views. This applies to settings in which a variable is measured with multiple experimental modalities, and where the goal is to synthesize the disparate measurements into a single unified representation. We consider the case that the observed views are a nonlinear mixing of component-wise corruptions of the sources. When the views are considered separately, this reduces to nonlinear Independent Component Analysis (ICA) for which it is provably impossible to undo the mixing. We present novel identifiability proofs that this is possible when the multiple views are considered jointly, showing that the mixing can theoretically be undone using function approximators such as deep neural networks. In contrast to known identifiability results for nonlinear ICA, we prove that independent latent sources with arbitrary mixing can be recovered as long as multiple, sufficiently different noisy views are available."
                },
                {
                    "title": "Bridging Theory and Algorithm for Domain Adaptation",
                    "abstract": "This paper addresses the problem of unsupervised domain adaption from theoretical and algorithmic perspectives. Existing domain adaptation theories naturally imply minimax optimization algorithms, which connect well with the domain adaptation methods based on adversarial learning. However, several disconnections still exist and form the gap between theory and algorithm. We extend previous theories (Mansour et al., 2009c; Ben-David et al., 2010) to multiclass classification in domain adaptation, where classifiers based on the scoring functions and margin loss are standard choices in algorithm design. We introduce Margin Disparity Discrepancy, a novel measurement with rigorous generalization bounds, tailored to the distribution comparison with the asymmetric margin loss, and to the minimax optimization for easier training. Our theory can be seamlessly transformed into an adversarial learning algorithm for domain adaptation, successfully bridging the gap between theory and algorithm. A series of empirical studies show that our algorithm achieves the state of the art accuracies on challenging domain adaptation tasks."
                },
                {
                    "title": "Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examples",
                    "abstract": "Few-shot classification refers to learning a classifier for new classes given only a few examples. While a plethora of models have emerged to tackle it, we find the procedure and datasets that are used to assess their progress lacking. To address this limitation, we propose Meta-Dataset: a new benchmark for training and evaluating models that is large-scale, consists of diverse datasets, and presents more realistic tasks. We experiment with popular baselines and meta-learners on Meta-Dataset, along with a competitive method that we propose. We analyze performance as a function of various characteristics of test tasks and examine the models' ability to leverage diverse training sources for improving their generalization. We also propose a new set of baselines for quantifying the benefit of meta-learning in Meta-Dataset. Our extensive experimentation has uncovered important research challenges and we hope to inspire work in these directions."
                },
                {
                    "title": "Joint profiling of chromatin accessibility and gene expression in thousands of single cells",
                    "abstract": "Single-cell chromatin and RNA analysis Single-cell analyses have begun to provide insight into the differences among and within the individual cells that make up a tissue or organism. However, technological barriers owing to the small amount of material present in each single cell have prevented parallel analyses. Cao et al. present sci-CAR, a pooled barcode method that jointly analyzes both the RNA transcripts and chromatin profiles of single cells. By applying sci-CAR to lung adenocarcinoma cells and mouse kidney tissue, the authors demonstrate precision in assessing expression and genome accessibility at a genome-wide scale. The approach provides an improvement over bulk analysis, which can be confounded by differing cellular subgroups. Science, this issue p. 1380 A technique termed sci-CAR can assess both chromatin accessibility and RNA transcription at the single-cell level. Although we can increasingly measure transcription, chromatin, methylation, and other aspects of molecular biology at single-cell resolution, most assays survey only one aspect of cellular biology. Here we describe sci-CAR, a combinatorial indexing\u2013based coassay that jointly profiles chromatin accessibility and mRNA (CAR) in each of thousands of single cells. As a proof of concept, we apply sci-CAR to 4825 cells, including a time series of dexamethasone treatment, as well as to 11,296 cells from the adult mouse kidney. With the resulting data, we compare the pseudotemporal dynamics of chromatin accessibility and gene expression, reconstruct the chromatin accessibility profiles of cell types defined by RNA profiles, and link cis-regulatory sites to their target genes on the basis of the covariance of chromatin accessibility and transcription across large numbers of single cells."
                },
                {
                    "title": "Unsupervised Alignment of Embeddings with Wasserstein Procrustes",
                    "abstract": "We consider the task of aligning two sets of points in high dimension, which has many applications in natural language processing and computer vision. As an example, it was recently shown that it is possible to infer a bilingual lexicon, without supervised data, by aligning word embeddings trained on monolingual data. These recent advances are based on adversarial training to learn the mapping between the two embeddings. In this paper, we propose to use an alternative formulation, based on the joint estimation of an orthogonal matrix and a permutation matrix. While this problem is not convex, we propose to initialize our optimization algorithm by using a convex relaxation, traditionally considered for the graph isomorphism problem. We propose a stochastic algorithm to minimize our cost function on large scale problems. Finally, we evaluate our method on the problem of unsupervised word translation, by aligning word embeddings trained on monolingual data. On this task, our method obtains state of the art results, while requiring less computational resources than competing approaches."
                },
                {
                    "title": "MAGAN: Aligning Biological Manifolds",
                    "abstract": "It is increasingly common in many types of natural and physical systems (especially biological systems) to have different types of measurements performed on the same underlying system. In such settings, it is important to align the manifolds arising from each measurement in order to integrate such data and gain an improved picture of the system. We tackle this problem using generative adversarial networks (GANs). Recently, GANs have been utilized to try to find correspondences between sets of samples. However, these GANs are not explicitly designed for proper alignment of manifolds. We present a new GAN called the Manifold-Aligning GAN (MAGAN) that aligns two manifolds such that related points in each measurement space are aligned together. We demonstrate applications of MAGAN in single-cell biology in integrating two different measurement types together. In our demonstrated examples, cells from the same tissue are measured with both genomic (single-cell RNA-sequencing) and proteomic (mass cytometry) technologies. We show that the MAGAN successfully aligns them such that known correlations between measured markers are improved compared to other recently proposed models."
                },
                {
                    "title": "StarGAN: Unified Generative Adversarial Networks for Multi-domain Image-to-Image Translation",
                    "abstract": "Recent studies have shown remarkable success in image-to-image translation for two domains. However, existing approaches have limited scalability and robustness in handling more than two domains, since different models should be built independently for every pair of image domains. To address this limitation, we propose StarGAN, a novel and scalable approach that can perform image-to-image translations for multiple domains using only a single model. Such a unified model architecture of StarGAN allows simultaneous training of multiple datasets with different domains within a single network. This leads to StarGAN's superior quality of translated images compared to existing models as well as the novel capability of flexibly translating an input image to any desired target domain. We empirically demonstrate the effectiveness of our approach on a facial attribute transfer and a facial expression synthesis tasks."
                },
                {
                    "title": "Unsupervised Machine Translation Using Monolingual Corpora Only",
                    "abstract": "Machine translation has recently achieved impressive performance thanks to recent advances in deep learning and the availability of large-scale parallel corpora. There have been numerous attempts to extend these successes to low-resource language pairs, yet requiring tens of thousands of parallel sentences. In this work, we take this research direction to the extreme and investigate whether it is possible to learn to translate even without any parallel data. We propose a model that takes sentences from monolingual corpora in two different languages and maps them into the same latent space. By learning to reconstruct in both languages from this shared feature space, the model effectively learns to translate without using any labeled data. We demonstrate our model on two widely used datasets and two language pairs, reporting BLEU scores of 32.8 and 15.1 on the Multi30k and WMT English-French datasets, without using even a single parallel sentence at training time."
                },
                {
                    "title": "Word Translation Without Parallel Data",
                    "abstract": "State-of-the-art methods for learning cross-lingual word embeddings have relied on bilingual dictionaries or parallel corpora. Recent studies showed that the need for parallel data supervision can be alleviated with character-level information. While these methods showed encouraging results, they are not on par with their supervised counterparts and are limited to pairs of languages sharing a common alphabet. In this work, we show that we can build a bilingual dictionary between two languages without using any parallel corpora, by aligning monolingual word embedding spaces in an unsupervised way. Without using any character information, our model even outperforms existing supervised methods on cross-lingual tasks for some language pairs. Our experiments demonstrate that our method works very well also for distant language pairs, like English-Russian or English-Chinese. We finally describe experiments on the English-Esperanto low-resource language pair, on which there only exists a limited amount of parallel data, to show the potential impact of our method in fully unsupervised machine translation. Our code, embeddings and dictionaries are publicly available."
                },
                {
                    "title": "Estimating Mixed Memberships With Sharp Eigenvector Deviations",
                    "abstract": "Abstract We consider the problem of estimating community memberships of nodes in a network, where every node is associated with a vector determining its degree of membership in each community. Existing provably consistent algorithms often require strong assumptions about the population, are computationally expensive, and only provide an overall error bound for the whole community membership matrix. This article provides uniform rates of convergence for the inferred community membership vector of each node in a network generated from the mixed membership stochastic blockmodel (MMSB); to our knowledge, this is the first work to establish per-node rates for overlapping community detection in networks. We achieve this by establishing sharp row-wise eigenvector deviation bounds for MMSB. Based on the simplex structure inherent in the eigen-decomposition of the population matrix, we build on established corner-finding algorithms from the optimization community to infer the community membership vectors. Our results hold over a broad parameter regime where the average degree only grows poly-logarithmically with the number of nodes. Using experiments with simulated and real datasets, we show that our method achieves better error with lower variability over competing methods, and processes real world networks of up to 100,000 nodes within tens of seconds. Supplementary materials for this article are available online."
                },
                {
                    "title": "The Role of Minimal Complexity Functions in Unsupervised Learning of Semantic Mappings",
                    "abstract": "We discuss the feasibility of the following learning problem: given unmatched samples from two domains and nothing else, learn a mapping between the two, which preserves semantics. Due to the lack of paired samples and without any definition of the semantic information, the problem might seem ill-posed. Specifically, in typical cases, it seems possible to build infinitely many alternative mappings from every target mapping. This apparent ambiguity stands in sharp contrast to the recent empirical success in solving this problem. \nWe identify the abstract notion of aligning two domains in a semantic way with concrete terms of minimal relative complexity. A theoretical framework for measuring the complexity of compositions of functions is developed in order to show that it is reasonable to expect the minimal complexity mapping to be unique. The measured complexity used is directly related to the depth of the neural networks being learned and a semantically aligned mapping could then be captured simply by learning using architectures that are not much bigger than the minimal architecture. \nVarious predictions are made based on the hypothesis that semantic alignment can be captured by the minimal mapping. These are verified extensively. In addition, a new mapping algorithm is proposed and shown to lead to better mapping results."
                },
                {
                    "title": "Deep Hashing Network for Unsupervised Domain Adaptation",
                    "abstract": "In recent years, deep neural networks have emerged as a dominant machine learning tool for a wide variety of application domains. However, training a deep neural network requires a large amount of labeled data, which is an expensive process in terms of time, labor and human expertise. Domain adaptation or transfer learning algorithms address this challenge by leveraging labeled data in a different, but related source domain, to develop a model for the target domain. Further, the explosive growth of digital data has posed a fundamental challenge concerning its storage and retrieval. Due to its storage and retrieval efficiency, recent years have witnessed a wide application of hashing in a variety of computer vision applications. In this paper, we first introduce a new dataset, Office-Home, to evaluate domain adaptation algorithms. The dataset contains images of a variety of everyday objects from multiple domains. We then propose a novel deep learning framework that can exploit labeled source data and unlabeled target data to learn informative hash codes, to accurately classify unseen target data. To the best of our knowledge, this is the first research effort to exploit the feature learning capabilities of deep neural networks to learn representative hash codes to address the domain adaptation problem. Our extensive empirical studies on multiple transfer tasks corroborate the usefulness of the framework in learning efficient hash codes which outperform existing competitive baselines for unsupervised domain adaptation."
                },
                {
                    "title": "Conditional Adversarial Domain Adaptation",
                    "abstract": "Adversarial learning has been embedded into deep networks to learn disentangled and transferable representations for domain adaptation. Existing adversarial domain adaptation methods may struggle to align different domains of multimodal distributions that are native in classification problems. In this paper, we present conditional adversarial domain adaptation, a principled framework that conditions the adversarial adaptation models on discriminative information conveyed in the classifier predictions. Conditional domain adversarial networks (CDANs) are designed with two novel conditioning strategies: multilinear conditioning that captures the cross-covariance between feature representations and classifier predictions to improve the discriminability, and entropy conditioning that controls the uncertainty of classifier predictions to guarantee the transferability. Experiments testify that the proposed approach exceeds the state-of-the-art results on five benchmark datasets."
                },
                {
                    "title": "Deep Clustering via Joint Convolutional Autoencoder Embedding and Relative Entropy Minimization",
                    "abstract": "In this paper, we propose a new clustering model, called DEeP Embedded Regularized ClusTering (DEPICT), which efficiently maps data into a discriminative embedding subspace and precisely predicts cluster assignments. DEPICT generally consists of a multinomial logistic regression function stacked on top of a multi-layer convolutional autoencoder. We define a clustering objective function using relative entropy (KL divergence) minimization, regularized by a prior for the frequency of cluster assignments. An alternating strategy is then derived to optimize the objective by updating parameters and estimating cluster assignments. Furthermore, we employ the reconstruction loss functions in our autoencoder, as a data-dependent regularization term, to prevent the deep embedding function from overfitting. In order to benefit from end-to-end optimization and eliminate the necessity for layer-wise pre-training, we introduce a joint learning framework to minimize the unified clustering and reconstruction loss functions together and train all network layers simultaneously. Experimental results indicate the superiority and faster running time of DEPICT in real-world clustering tasks, where no labeled data is available for hyper-parameter tuning."
                },
                {
                    "title": "Adversarial Discriminative Domain Adaptation",
                    "abstract": "Adversarial learning methods are a promising approach to training robust deep networks, and can generate complex samples across diverse domains. They can also improve recognition despite the presence of domain shift or dataset bias: recent adversarial approaches to unsupervised domain adaptation reduce the difference between the training and test domain distributions and thus improve generalization performance. However, while generative adversarial networks (GANs) show compelling visualizations, they are not optimal on discriminative tasks and can be limited to smaller shifts. On the other hand, discriminative approaches can handle larger domain shifts, but impose tied weights on the model and do not exploit a GAN-based loss. In this work, we first outline a novel generalized framework for adversarial adaptation, which subsumes recent state-of-the-art approaches as special cases, and use this generalized view to better relate prior approaches. We then propose a previously unexplored instance of our general framework which combines discriminative modeling, untied weight sharing, and a GAN loss, which we call Adversarial Discriminative Domain Adaptation (ADDA). We show that ADDA is more effective yet considerably simpler than competing domain-adversarial methods, and demonstrate the promise of our approach by exceeding state-of-the-art unsupervised adaptation results on standard domain adaptation tasks as well as a difficult cross-modality object classification task."
                },
                {
                    "title": "FastText.zip: Compressing text classification models",
                    "abstract": "We consider the problem of producing compact architectures for text classification, such that the full model fits in a limited amount of memory. After considering different solutions inspired by the hashing literature, we propose a method built upon product quantization to store the word embeddings. While the original technique leads to a loss in accuracy, we adapt this method to circumvent the quantization artifacts. As a result, our approach produces a text classifier, derived from the fastText approach, which at test time requires only a fraction of the memory compared to the original one, without noticeably sacrificing the quality in terms of classification accuracy. Our experiments carried out on several benchmarks show that our approach typically requires two orders of magnitude less memory than fastText while being only slightly inferior with respect to accuracy. As a result, it outperforms the state of the art by a good margin in terms of the compromise between memory usage and accuracy."
                },
                {
                    "title": "Deep Variational Canonical Correlation Analysis",
                    "abstract": "We present deep variational canonical correlation analysis (VCCA), a deep multi-view learning model that extends the latent variable model interpretation of linear CCA to nonlinear observation models parameterized by deep neural networks. We derive variational lower bounds of the data likelihood by parameterizing the posterior probability of the latent variables from the view that is available at test time. We also propose a variant of VCCA called VCCA-private that can, in addition to the \"common variables\" underlying both views, extract the \"private variables\" within each view, and disentangles the shared and private information for multi-view data without hard supervision. Experimental results on real-world datasets show that our methods are competitive across domains."
                },
                {
                    "title": "Matching Networks for One Shot Learning",
                    "abstract": "Learning from a few examples remains a key challenge in machine learning. Despite recent advances in important domains such as vision and language, the standard supervised deep learning paradigm does not offer a satisfactory solution for learning new concepts rapidly from little data. In this work, we employ ideas from metric learning based on deep neural features and from recent advances that augment neural networks with external memories. Our framework learns a network that maps a small labelled support set and an unlabelled example to its label, obviating the need for fine-tuning to adapt to new class types. We then define one-shot learning problems on vision (using Omniglot, ImageNet) and language tasks. Our algorithm improves one-shot accuracy on ImageNet from 87.6% to 93.2% and from 88.0% to 93.8% on Omniglot compared to competing approaches. We also demonstrate the usefulness of the same model on language modeling by introducing a one-shot task on the Penn Treebank."
                },
                {
                    "title": "Deep Transfer Learning with Joint Adaptation Networks",
                    "abstract": "Deep networks have been successfully applied to learn transferable features for adapting models from a source domain to a different target domain. In this paper, we present joint adaptation networks (JAN), which learn a transfer network by aligning the joint distributions of multiple domain-specific layers across domains based on a joint maximum mean discrepancy (JMMD) criterion. Adversarial training strategy is adopted to maximize JMMD such that the distributions of the source and target domains are made more distinguishable. Learning can be performed by stochastic gradient descent with the gradients computed by back-propagation in linear-time. Experiments testify that our model yields state of the art results on standard datasets."
                },
                {
                    "title": "Unsupervised Deep Embedding for Clustering Analysis",
                    "abstract": "Clustering is central to many data-driven application domains and has been studied extensively in terms of distance functions and grouping algorithms. Relatively little work has focused on learning representations for clustering. In this paper, we propose Deep Embedded Clustering (DEC), a method that simultaneously learns feature representations and cluster assignments using deep neural networks. DEC learns a mapping from the data space to a lower-dimensional feature space in which it iteratively optimizes a clustering objective. Our experimental evaluations on image and text corpora show significant improvement over state-of-the-art methods."
                },
                {
                    "title": "Complete Dictionary Recovery Over the Sphere I: Overview and the Geometric Picture",
                    "abstract": "We consider the problem of recovering a complete (i.e., square and invertible) matrix <inline-formula> <tex-math notation=\"LaTeX\">$ A_{0}$ </tex-math></inline-formula>, from <inline-formula> <tex-math notation=\"LaTeX\">$ Y \\in \\mathbb R ^{n \\times p}$ </tex-math></inline-formula> with <inline-formula> <tex-math notation=\"LaTeX\">$ Y = A_{0} X_{0}$ </tex-math></inline-formula>, provided <inline-formula> <tex-math notation=\"LaTeX\">$ X_{0}$ </tex-math></inline-formula> is sufficiently sparse. This recovery problem is central to theoretical understanding of dictionary learning, which seeks a sparse representation for a collection of input signals and finds numerous applications in modern signal processing and machine learning. We give the first efficient algorithm that provably recovers <inline-formula> <tex-math notation=\"LaTeX\">$ A_{0}$ </tex-math></inline-formula> when <inline-formula> <tex-math notation=\"LaTeX\">$ X_{0}$ </tex-math></inline-formula> has <inline-formula> <tex-math notation=\"LaTeX\">$O \\left ({ n }\\right )$ </tex-math></inline-formula> nonzeros per column, under suitable probability model for <inline-formula> <tex-math notation=\"LaTeX\">$ X_{0}$ </tex-math></inline-formula>. In contrast, prior results based on efficient algorithms either only guarantee recovery when <inline-formula> <tex-math notation=\"LaTeX\">$ X_{0}$ </tex-math></inline-formula> has <inline-formula> <tex-math notation=\"LaTeX\">$O(\\sqrt {n})$ </tex-math></inline-formula> zeros per column, or require multiple rounds of semidefinite programming relaxation to work when <inline-formula> <tex-math notation=\"LaTeX\">$ X_{0}$ </tex-math></inline-formula> has <inline-formula> <tex-math notation=\"LaTeX\">$O(n)$ </tex-math></inline-formula> nonzeros per column. Our algorithmic pipeline centers around solving a certain nonconvex optimization problem with a spherical constraint. In this paper, we provide a geometric characterization of the objective landscape. In particular, we show that the problem is highly structured with high probability: 1) there are no \u201cspurious\u201d local minimizers and 2) around all saddle points the objective has a negative directional curvature. This distinctive structure makes the problem amenable to efficient optimization algorithms. In a companion paper, we design a second-order trust-region algorithm over the sphere that provably converges to a local minimizer from arbitrary initializations, despite the presence of saddle points."
                },
                {
                    "title": "Multiview LSA: Representation Learning via Generalized CCA",
                    "abstract": "Multiview LSA (MVLSA) is a generalization of Latent Semantic Analysis (LSA) that supports the fusion of arbitrary views of data and relies on Generalized Canonical Correlation Analysis (GCCA). We present an algorithm for fast approximate computation of GCCA, which when coupled with methods for handling missing values, is general enough to approximate some recent algorithms for inducing vector representations of words. Experiments across a comprehensive collection of test-sets show our approach to be competitive with the state of the art."
                },
                {
                    "title": "Learning Transferable Features with Deep Adaptation Networks",
                    "abstract": "Recent studies reveal that a deep neural network can learn transferable features which generalize well to novel tasks for domain adaptation. However, as deep features eventually transition from general to specific along the network, the feature transferability drops significantly in higher layers with increasing domain discrepancy. Hence, it is important to formally reduce the dataset bias and enhance the transferability in task-specific layers. In this paper, we propose a new Deep Adaptation Network (DAN) architecture, which generalizes deep convolutional neural network to the domain adaptation scenario. In DAN, hidden representations of all task-specific layers are embedded in a reproducing kernel Hilbert space where the mean embeddings of different domain distributions can be explicitly matched. The domain discrepancy is further reduced using an optimal multikernel selection method for mean embedding matching. DAN can learn transferable features with statistical guarantees, and can scale linearly by unbiased estimate of kernel embedding. Extensive empirical evidence shows that the proposed architecture yields state-of-the-art image classification error rates on standard domain adaptation benchmarks."
                },
                {
                    "title": "Adam: A Method for Stochastic Optimization",
                    "abstract": "We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
                },
                {
                    "title": "Unsupervised Domain Adaptation by Backpropagation",
                    "abstract": "Top-performing deep architectures are trained on massive amounts of labeled data. In the absence of labeled data for a certain task, domain adaptation often provides an attractive option given that labeled data of similar nature but from a different domain (e.g. synthetic images) are available. Here, we propose a new approach to domain adaptation in deep architectures that can be trained on large amount of labeled data from the source domain and large amount of unlabeled data from the target domain (no labeled target-domain data is necessary). \nAs the training progresses, the approach promotes the emergence of \"deep\" features that are (i) discriminative for the main learning task on the source domain and (ii) invariant with respect to the shift between the domains. We show that this adaptation behaviour can be achieved in almost any feed-forward model by augmenting it with few standard layers and a simple new gradient reversal layer. The resulting augmented architecture can be trained using standard backpropagation. \nOverall, the approach can be implemented with little effort using any of the deep-learning packages. The method performs very well in a series of image classification experiments, achieving adaptation effect in the presence of big domain shifts and outperforming previous state-of-the-art on Office datasets."
                },
                {
                    "title": "Generative adversarial networks",
                    "abstract": "Generative adversarial networks are a kind of artificial intelligence algorithm designed to solve the generative modeling problem. The goal of a generative model is to study a collection of training examples and learn the probability distribution that generated them. Generative Adversarial Networks (GANs) are then able to generate more examples from the estimated probability distribution. Generative models based on deep learning are common, but GANs are among the most successful generative models (especially in terms of their ability to generate realistic high-resolution images). GANs have been successfully applied to a wide variety of tasks (mostly in research settings) but continue to present unique challenges and research opportunities because they are based on game theory while most other approaches to generative modeling are based on optimization."
                },
                {
                    "title": "The Why and How of Nonnegative Matrix Factorization",
                    "abstract": "Nonnegative matrix factorization (NMF) has become a widely used tool for the analysis of high-dimensional data as it automatically extracts sparse and meaningful features from a set of nonnegative data vectors. We first illustrate this property of NMF on three applications, in image processing, text mining and hyperspectral imaging --this is the why. Then we address the problem of solving NMF, which is NP-hard in general. We review some standard NMF algorithms, and also present a recent subclass of NMF problems, referred to as near-separable NMF, that can be solved efficiently (that is, in polynomial time), even in the presence of noise --this is the how. Finally, we briefly describe some problems in mathematics and computer science closely related to NMF via the nonnegative rank."
                },
                {
                    "title": "A Class of Bounded Component Analysis Algorithms for the Separation of Both Independent and Dependent Sources",
                    "abstract": "Bounded Component Analysis (BCA) is a recent approach which enables the separation of both dependent and independent signals from their mixtures. In this approach, under the practical source boundedness assumption, the widely used statistical independence assumption is replaced by a more generic domain separability assumption. This article introduces a geometric framework for the development of Bounded Component Analysis algorithms. Two main geometric objects related to the separator output samples, namely Principal Hyper-Ellipsoid and Bounding Hyper-Rectangle, are introduced. The maximization of the volume ratio of these objects, and its extensions, are introduced as relevant optimization problems for Bounded Component Analysis. The article also provides corresponding iterative algorithms for both real and complex sources. The numerical examples illustrate the potential advantage of the proposed BCA framework in terms of correlated source separation capability as well as performance improvement for short data records."
                },
                {
                    "title": "Independent Vector Analysis: Identification Conditions and Performance Bounds",
                    "abstract": "Recently, an extension of independent component analysis (ICA) from one to multiple datasets, termed independent vector analysis (IVA), has been a subject of significant research interest. IVA has also been shown to be a generalization of Hotelling's canonical correlation analysis. In this paper, we provide the identification conditions for a general IVA formulation, which accounts for linear, nonlinear, and sample-to-sample dependencies. The identification conditions are a generalization of previous results for ICA and for IVA when samples are independently and identically distributed. Furthermore, a principal aim of IVA is identification of dependent sources between datasets. Thus, we provide additional conditions for when the arbitrary ordering of the estimated sources can be common across datasets. Performance bounds in terms of the Crame\u0301r-Rao lower bound are also provided for demixing matrices and interference to source ratio. The performance of two IVA algorithms are compared to the theoretical bounds."
                },
                {
                    "title": "A Practical Algorithm for Topic Modeling with Provable Guarantees",
                    "abstract": "Topic models provide a useful method for dimensionality reduction and exploratory data analysis in large text corpora. Most approaches to topic model inference have been based on a maximum likelihood objective. Efficient algorithms exist that approximate this objective, but they have no provable guarantees. Recently, algorithms have been introduced that provide provable bounds, but these algorithms are not practical because they are inefficient and not robust to violations of model assumptions. In this paper we present an algorithm for topic model inference that is both provable and practical. The algorithm produces results comparable to the best MCMC implementations while running orders of magnitude faster."
                },
                {
                    "title": "A Kernel Two-Sample Test",
                    "abstract": "We propose a framework for analyzing and comparing distributions, which we use to construct statistical tests to determine if two samples are drawn from different distributions. Our test statistic is the largest difference in expectations over functions in the unit ball of a reproducing kernel Hilbert space (RKHS), and is called the maximum mean discrepancy (MMD).We present two distribution free tests based on large deviation bounds for the MMD, and a third test based on the asymptotic distribution of this statistic. The MMD can be computed in quadratic time, although efficient linear time approximations are available. Our statistic is an instance of an integral probability metric, and various classical metrics on distributions are obtained when alternative function classes are used in place of an RKHS. We apply our two-sample tests to a variety of problems, including attribute matching for databases using the Hungarian marriage method, where they perform strongly. Excellent performance is also obtained when comparing distributions over graphs, for which these are the first such tests."
                },
                {
                    "title": "Handbook of Blind Source Separation: Independent Component Analysis and Applications",
                    "abstract": "Edited by the people who were forerunners in creating the field, together with contributions from 34 leading international experts, this handbook provides the definitive reference on Blind Source Separation, giving a broad and comprehensive description of all the core principles and methods, numerical algorithms and major applications in the fields of telecommunications, biomedical engineering and audio, acoustic and speech processing. Going beyond a machine learning perspective, the book reflects recent results in signal processing and numerical analysis, and includes topics such as optimization criteria, mathematical tools, the design of numerical algorithms, convolutive mixtures, and time frequency approaches. This Handbook is an ideal reference for university researchers, RD algebraic identification of under-determined mixtures, time-frequency methods, Bayesian approaches, blind identification under non negativity approaches, semi-blind methods for communicationsShows the applications of the methods to key application areas such as telecommunications, biomedical engineering, speech, acoustic, audio and music processing, while also giving a general method for developing applications"
                },
                {
                    "title": "Joint Blind Source Separation by Multiset Canonical Correlation Analysis",
                    "abstract": "In this paper, we introduce a simple and effective scheme to achieve joint blind source separation (BSS) of multiple datasets using multiset canonical correlation analysis (M-CCA) [J. R. Kettenring, \"Canonical analysis of several sets of variables\", Biometrika, vol. 58, pp. 433-451, 1971]. We first propose a generative model of joint BSS based on the correlation of latent sources within and between datasets. We specify source separability conditions, and show that, when the conditions are satisfied, the group of corresponding sources from each dataset can be jointly extracted by M-CCA through maximization of correlation among the extracted sources. We compare source separation performance of the M-CCA scheme with other joint BSS methods and demonstrate the superior performance of the M-CCA scheme in achieving joint BSS for a large number of datasets, group of corresponding sources with heterogeneous correlation values, and complex-valued sources with circular and non-circular distributions. We apply M-CCA to analysis of functional magnetic resonance imaging (fMRI) data from multiple subjects and show its utility in estimating meaningful brain activations from a visuomotor task."
                },
                {
                    "title": "Minimum Volume Simplex Analysis: A Fast Algorithm to Unmix Hyperspectral Data",
                    "abstract": "This paper presents a new method of minimum volume class for hyperspectral unmixing, termed minimum volume simplex analysis (MVSA). The underlying mixing model is linear; i.e., the mixed hyperspectral vectors are modeled by a linear mixture of the end-member signatures weighted by the correspondent abundance fractions. MVSA approaches hyperspectral unmixing by fitting a minimum volume simplex to the hyperspectral data, constraining the abundance fractions to belong to the probability simplex. The resulting optimization problem is solved by implementing a sequence of quadratically constrained subproblems. In a final step, the hard constraint on the abundance fractions is replaced with a hinge type loss function to account for outliers and noise. We illustrate the state-of-the-art performance of the MVSA algorithm in unmixing simulated data sets. We are mainly concerned with the realistic scenario in which the pure pixel assumption (i.e., there exists at least one pure pixel per end member) is not fulfilled. In these conditions, the MVSA yields much better performance than the pure pixel based algorithms."
                },
                {
                    "title": "Analysis of Representations for Domain Adaptation",
                    "abstract": "Discriminative learning methods for classification perform well when training and test data are drawn from the same distribution. In many situations, though, we have labeled training data for a source domain, and we wish to learn a classifier which performs well on a target domain with a different distribution. Under what conditions can we adapt a classifier trained on the source domain for use in the target domain? Intuitively, a good feature representation is a crucial factor in the success of domain adaptation. We formalize this intuition theoretically with a generalization bound for domain adaption. Our theory illustrates the tradeoffs inherent in designing a representation for domain adaptation and gives a new justification for a recently proposed model. It also points toward a promising new model for domain adaptation: one which explicitly minimizes the difference between the source and target domains, while at the same time maximizing the margin of the training set."
                },
                {
                    "title": "Super-symmetric decomposition of the fourth-order cumulant tensor. Blind identification of more sources than sensors",
                    "abstract": "Ideas for higher-order array processing are introduced, focusing on fourth-order cumulant statistics. They are expressed in an index-free formalism allowing the exploitation of all their symmetry properties. It is shown that, when dealing with 4-index quantities, symmetries are related to rank properties. The rich symmetry structure yields a class of identification algorithms. An algebraic technique for blind identification is included, and a few others are briefly indicated.<<ETX>>"
                },
                {
                    "title": "Multi-domain image generation and translation with identifiability guarantees",
                    "abstract": "Multi-domain image generation and unpaired image-to-to-image translation are two important and related computer vision problems. The common technique for the two tasks is the learning of a joint distribution from multiple marginal distributions. However, it is well known that there can be infinitely many joint distributions that can derive the same marginals. Hence, it is necessary to formulate suitable constraints to address this highly ill-posed problem. Inspired by the recent advances in nonlinear Independent Component Analysis (ICA) theory, we propose a new method to learn the joint distribution from the marginals by enforcing a specific type of minimal change across domains. We report one of the first results connecting multi-domain generative models to identifiability and shows why identifiability is essential and how to achieve it theoretically and practically. We apply our method to five multi-domain image generation and six image-toimage translation tasks. The superior performance of our model supports our theory and demonstrates the effectiveness of our method. The training code are available at https://github.com/Mid-Push/i-stylegan."
                },
                {
                    "title": "Rectifier Nonlinearities Improve Neural Network Acoustic Models",
                    "abstract": "Deep neural network acoustic models produce substantial gains in large vocabulary continuous speech recognition systems. Emerging work with recti\ufb01ed linear (ReL) hidden units demonstrates additional gains in \ufb01nal system performance relative to more commonly used sigmoidal nonlinearities. In this work, we explore the use of deep recti\ufb01er networks as acoustic models for the 300 hour Switchboard conversational speech recognition task. Using simple training procedures without pretraining, networks with recti\ufb01er nonlinearities produce 2% absolute reductions in word error rates over their sigmoidal counterparts. We analyze hidden layer representations to quantify di\ufb00erences in how ReL units encode inputs as compared to sigmoidal units. Finally, we evaluate a variant of the ReL unit with a gradient more amenable to optimization in an attempt to further improve deep recti\ufb01er networks."
                },
                {
                    "title": "A Probabilistic Interpretation of Canonical Correlation Analysis",
                    "abstract": "We give a probabilistic interpretation of canonical correlation (CCA) analysis as a latent variable model for two Gaussian random vectors. Our interpretation is similar to the probabilistic interpretation of principal component analysis (Tipping and Bishop, 1999, Roweis, 1998). In addition, we cast Fisher linear discriminant analysis (LDA) within the CCA framework."
                },
                {
                    "title": "SIGNAL PROCESSING Independent component analysis , A new concept ? *",
                    "abstract": "The independent component analysis (ICA) of a random vector consists of searching for a linear transformation that minimizes the statistical dependence between its components. In order to define suitable search criteria, the expansion of mutual information is utilized as a function of cumulants of increasing orders. An efficient algorithm is proposed, which allows the computation of the ICA of a data matrix within a polynomial time. The concept of lCA may actually be seen as an extension of the principal component analysis (PCA), which can only impose independence up to the second order and, consequently, defines directions that are orthogonal. Potential applications of ICA include data analysis and compression, Bayesian detection, localization of sources, and blind identification and deconvolution. Zusammenfassung Die Analyse unabhfingiger Komponenten (ICA) eines Vektors beruht auf der Suche nach einer linearen Transformation, die die statistische Abh~ingigkeit zwischen den Komponenten minimiert. Zur Definition geeigneter Such-Kriterien wird die Entwicklung gemeinsamer Information als Funktion von Kumulanten steigender Ordnung genutzt. Es wird ein effizienter Algorithmus vorgeschlagen, der die Berechnung der ICA ffir Datenmatrizen innerhalb einer polynomischen Zeit erlaubt. Das Konzept der ICA kann eigentlich als Erweiterung der 'Principal Component Analysis' (PCA) betrachtet werden, die nur die Unabh~ingigkeit bis zur zweiten Ordnung erzwingen kann und deshalb Richtungen definiert, die orthogonal sind. Potentielle Anwendungen der ICA beinhalten Daten-Analyse und Kompression, Bayes-Detektion, Quellenlokalisierung und blinde Identifikation und Entfaltung. R~sum~ L'Analyse en Composantes Ind6pendantes (ICA) d'un vecteur al6atoire consiste en la recherche d'une transformation lin6aire qui minimise la d6pendance statistique entre ses composantes. Afin de d6finir des crit6res d'optimisation appropribs, on utilise un d6veloppement en s6rie de l'information mutuelle en fonction de cumulants d'ordre croissant. On propose ensuite un algorithme pratique permettant le calcul de I'ICA d'une matrice de donn6es en un temps polynomial. Le concept d'ICA peut 6tre vu en r~alitb comme une extension de l'Analyse en Composantes Principales (PCA) qui, elle, ne peut imposer l'ind6pendance qu'au second ordre et d6finit par cons6quent des directions orthogonales. Les applications potentielles de I'ICA incluent l'analyse et la compression de donn6es, la d&ection bayesienne, la localisation de sources, et l'identification et la d6convolution aveugles. + This work was supported in part by the DRET. 0165-1684/94/$7.00 \u00a9 1994 Elsevier Science B.V. All rights reserved SSDI 0165-1684(93)E0093-Z 288 P. Comon / Signal Processing 36 (1994) 28~314"
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan the shared latent components of unaligned multimodal linear mixtures be provably identified under reasonably mild conditions?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of representation learning, particularly in scenarios where aligned multimodal data is difficult to obtain. By identifying shared components in unaligned data, we can enhance the performance of various downstream tasks, such as cross-language information retrieval and domain adaptation. This research could lead to significant advancements in self-supervised learning and improve the robustness of machine learning models by effectively utilizing diverse data sources, ultimately fostering new practical applications across multiple domains.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the ill-posed nature of linear mixture models, which complicates the identification of shared components. Naive approaches may fail because they often assume aligned data, which is not always available in real-world applications. Additionally, the complexities arise from the need to distinguish between shared and private components without clear alignment, requiring sophisticated methods to ensure accurate identification under mild conditions. Technical obstacles include the need for robust algorithms that can handle the ambiguities inherent in unaligned data.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on aligned multimodal data, neglecting the complexities introduced by unaligned mixtures. Existing studies either relied on stringent assumptions that are difficult to satisfy or did not consider latent component models, which are more versatile. The limitations of prior work, such as the reliance on independent component analysis (ICA) with strict independence assumptions, have hindered progress. Our approach differs by providing a suite of sufficient conditions for identifying shared components in unaligned settings, addressing the gaps left by earlier studies.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, termed unaligned shared component analysis (unaligned SCA), involves developing an identifiable learning loss that facilitates the identification of shared components from unaligned multimodal linear mixtures. We will utilize a diverse dataset comprising various modalities and evaluate our approach using metrics that assess the accuracy of shared component identification. The expected outcomes include a set of sufficient conditions under which shared components can be identified, contributing to the theoretical understanding of unaligned multimodal learning and providing practical tools for real-world applications."
            }
        },
        "author_data": {
            "7e0294dc-6692-426a-b8a2-f361e1a0f506": {
                "pk": "7e0294dc-6692-426a-b8a2-f361e1a0f506",
                "name": "Subash Timilsina",
                "collaborators": [
                    "Sagar Shrestha",
                    "Xiao Fu",
                    "Abhish Khanal",
                    "Deepak Chand",
                    "Prakash Chaudhary",
                    "Sanjeeb Prasad Panday",
                    "Aman Shakya",
                    "Rom Kant Pandey"
                ],
                "domain": [
                    "Signal Processing",
                    "Sound Localization",
                    "Radio Map Estimation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Quantized Radio Map Estimation Using Tensor and Deep Generative Models",
                        "abstract": "Spectrum cartography (SC), also known as radio map estimation (RME), aims at crafting multi-domain (e.g., frequency and space) radio power propagation maps from limited sensor measurements. While early methods often lacked theoretical support, recent works have demonstrated that radio maps can be provably recovered using low-dimensional models -- such as the block-term tensor decomposition (BTD) model and certain deep generative models (DGMs) -- of the high-dimensional multi-domain radio signals. However, these existing provable SC approaches assume that sensors send real-valued (full-resolution) measurements to the fusion center, which is unrealistic. This work puts forth a quantized SC framework that generalizes the BTD and DGM-based SC to scenarios where heavily quantized sensor measurements are used. A maximum likelihood estimation (MLE)-based SC framework under a Gaussian quantizer is proposed. Recoverability of the radio map using the MLE criterion are characterized under realistic conditions, e.g., imperfect radio map modeling and noisy measurements. Simulations and real-data experiments are used to showcase the effectiveness of the proposed approach."
                    },
                    {
                        "title": "Search Disaster Victims using Sound Source Localization",
                        "abstract": "Sound Source Localization (SSL) are used to estimate the position of sound sources. Various methods have been used for detecting sound and its localization. This paper presents a system for stationary sound source localization by cubical microphone array consisting of eight microphones placed on four vertical adjacent faces which is mounted on three wheel omni-directional drive for the inspection and monitoring of the disaster victims in disaster areas. The proposed method localizes sound source on a 3D space by grid search method using Generalized Cross Correlation Phase Transform (GCC-PHAT) which is robust when operating in real life scenario where there is lack of visibility. The computed azimuth and elevation angle of victimized human voice are fed to embedded omni-directional drive system which navigates the vehicle automatically towards the stationary sound source."
                    }
                ]
            },
            "18ebd364-01b4-4bf8-943d-da02e2a4bcb5": {
                "pk": "18ebd364-01b4-4bf8-943d-da02e2a4bcb5",
                "name": "Sagar Shrestha",
                "collaborators": [
                    "Xiao Fu",
                    "Mingyi Hong",
                    "Subash Timilsina"
                ],
                "domain": [
                    "Federated Learning",
                    "Deep Learning",
                    "Signal Processing",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Communication-Efficient Federated Linear and Deep Generalized Canonical Correlation Analysis",
                        "abstract": "Classic and deep generalized canonical correlation analysis (GCCA) algorithms seek low-dimensional common representations of data entities from multiple ``views'' (e.g., audio and image) using linear transformations and neural networks, respectively. When the views are acquired and stored at different computing agents (e.g., organizations and edge devices) and data sharing is undesired due to privacy or communication cost considerations, federated learning-based GCCA is well-motivated. In federated learning, the views are kept locally at the agents and only derived, limited information exchange with a central server is allowed. However, applying existing GCCA algorithms onto such federated learning settings may incur prohibitively high communication overhead. This work puts forth a communication-efficient federated learning framework for both linear and deep GCCA under the maximum variance (MAX-VAR) formulation. The overhead issue is addressed by aggressively compressing (via quantization) the exchanging information between the computing agents and a central controller. Compared to the unquantized version, our empirical study shows that the proposed algorithm enjoys a substantial reduction of communication overheads with virtually no loss in accuracy and convergence speed. Rigorous convergence analyses are also presented, which is a nontrivial effort. Generic federated optimization results do not cover the special problem structure of GCCA. Our result shows that the proposed algorithms for both linear and deep GCCA converge to critical points at a sublinear rate, even under heavy quantization and stochastic approximations. In addition, in the linear MAX-VAR case, the quantized algorithm approaches a global optimum in a geometric rate under reasonable conditions. Synthetic and real-data experiments are used to showcase the effectiveness of the proposed approach."
                    },
                    {
                        "title": "Towards Identifiable Unsupervised Domain Translation: A Diversified Distribution Matching Approach",
                        "abstract": "Unsupervised domain translation (UDT) aims to find functions that convert samples from one domain (e.g., sketches) to another domain (e.g., photos) without changing the high-level semantic meaning (also referred to as ``content''). The translation functions are often sought by probability distribution matching of the transformed source domain and target domain. CycleGAN stands as arguably the most representative approach among this line of work. However, it was noticed in the literature that CycleGAN and variants could fail to identify the desired translation functions and produce content-misaligned translations. This limitation arises due to the presence of multiple translation functions -- referred to as ``measure-preserving automorphism\" (MPA) -- in the solution space of the learning criteria. Despite awareness of such identifiability issues, solutions have remained elusive. This study delves into the core identifiability inquiry and introduces an MPA elimination theory. Our analysis shows that MPA is unlikely to exist, if multiple pairs of diverse cross-domain conditional distributions are matched by the learning function. Our theory leads to a UDT learner using distribution matching over auxiliary variable-induced subsets of the domains -- other than over the entire data domains as in the classical approaches. The proposed framework is the first to rigorously establish translation identifiability under reasonable UDT settings, to our best knowledge. Experiments corroborate with our theoretical claims."
                    },
                    {
                        "title": "Deep Spectrum Cartography: Completing Radio Map Tensors Using Learned Neural Models",
                        "abstract": "The spectrum cartography (SC) technique constructs multi-domain (e.g., frequency, space, and time) radio frequency (RF) maps from limited measurements, which can be viewed as an ill-posed tensor completion problem. Model-based cartography techniques often rely on handcrafted priors (e.g., sparsity, smoothness and low-rank structures) for the completion task. Such priors may be inadequate to capture the essence of complex wireless environments -- especially when severe shadowing happens. To circumvent such challenges, offline-trained deep neural models of radio maps were considered for SC, as deep neural networks (DNNs) are able to \"learn\" intricate underlying structures from data. However, such deep learning (DL)-based SC approaches encounter serious challenges in both off-line model learning (training) and completion (generalization), possibly because the latent state space for generating the radio maps is prohibitively large. In this work, an emitter radio map disaggregation-based approach is proposed, under which only individual emitters' radio maps are modeled by DNNs. This way, the learning and generalization challenges can both be substantially alleviated. Using the learned DNNs, a fast nonnegative matrix factorization-based two-stage SC method and a performance-enhanced iterative optimization algorithm are proposed. Theoretical aspects -- such as recoverability of the radio tensor, sample complexity, and noise robustness -- under the proposed framework are characterized, and such theoretical properties have been elusive in the context of DL-based radio tensor completion. Experiments using synthetic and real-data from indoor and heavily shadowed environments are employed to showcase the effectiveness of the proposed methods."
                    },
                    {
                        "title": "Optimal Solutions for Joint Beamforming and Antenna Selection: From Branch and Bound to Graph Neural Imitation Learning",
                        "abstract": "This work revisits the joint beamforming (BF) and antenna selection (AS) problem, as well as its robust beamforming (RBF) version under imperfect channel state information (CSI). Such problems arise due to various reasons, e.g., the costly nature of the radio frequency (RF) chains and energy/resource-saving considerations. The joint (R)BF\\&AS problem is a mixed integer and nonlinear program, and thus finding {\\it optimal solutions} is often costly, if not outright impossible. The vast majority of the prior works tackled these problems using techniques such as continuous approximations, greedy methods, and supervised machine learning -- yet these approaches do not ensure optimality or even feasibility of the solutions. The main contribution of this work is threefold. First, an effective {\\it branch and bound} (B\\&B) framework for solving the problems of interest is proposed. Leveraging existing BF and RBF solvers, it is shown that the B\\&B framework guarantees global optimality of the considered problems. Second, to expedite the potentially costly B\\&B algorithm, a machine learning (ML)-based scheme is proposed to help skip intermediate states of the B\\&B search tree. The learning model features a {\\it graph neural network} (GNN)-based design that is resilient to a commonly encountered challenge in wireless communications, namely, the change of problem size (e.g., the number of users) across the training and test stages. Third, comprehensive performance characterizations are presented, showing that the GNN-based method retains the global optimality of B\\&B with provably reduced complexity, under reasonable conditions. Numerical simulations also show that the ML-based acceleration can often achieve an order-of-magnitude speedup relative to B\\&B."
                    },
                    {
                        "title": "Quantized Radio Map Estimation Using Tensor and Deep Generative Models",
                        "abstract": "Spectrum cartography (SC), also known as radio map estimation (RME), aims at crafting multi-domain (e.g., frequency and space) radio power propagation maps from limited sensor measurements. While early methods often lacked theoretical support, recent works have demonstrated that radio maps can be provably recovered using low-dimensional models -- such as the block-term tensor decomposition (BTD) model and certain deep generative models (DGMs) -- of the high-dimensional multi-domain radio signals. However, these existing provable SC approaches assume that sensors send real-valued (full-resolution) measurements to the fusion center, which is unrealistic. This work puts forth a quantized SC framework that generalizes the BTD and DGM-based SC to scenarios where heavily quantized sensor measurements are used. A maximum likelihood estimation (MLE)-based SC framework under a Gaussian quantizer is proposed. Recoverability of the radio map using the MLE criterion are characterized under realistic conditions, e.g., imperfect radio map modeling and noisy measurements. Simulations and real-data experiments are used to showcase the effectiveness of the proposed approach."
                    }
                ]
            },
            "be2cd4ec-05c0-4ad0-baa4-d63c62f60642": {
                "pk": "be2cd4ec-05c0-4ad0-baa4-d63c62f60642",
                "name": "Xiao Fu",
                "collaborators": [
                    "Qi Lyu",
                    "Shahana Ibrahim",
                    "Sagar Shrestha",
                    "Guijun Zhang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Unsupervised Learning",
                    "Graph Analysis"
                ],
                "publications": [
                    {
                        "title": "A More Efficient Chinese Named Entity Recognition base on BERT and Syntactic Analysis",
                        "abstract": "We propose a new Named entity recognition (NER) method to effectively make use of the results of Part-of-speech (POS) tagging, Chinese word segmentation (CWS) and parsing while avoiding NER error caused by POS tagging error. This paper first uses Stanford natural language process (NLP) tool to annotate large-scale untagged data so as to reduce the dependence on the tagged data; then a new NLP model, g-BERT model, is designed to compress Bidirectional Encoder Representations from Transformers (BERT) model in order to reduce calculation quantity; finally, the model is evaluated based on Chinese NER dataset. The experimental results show that the calculation quantity in g-BERT model is reduced by 60% and performance improves by 2% with Test F1 to 96.5 compared with that in BERT model."
                    },
                    {
                        "title": "Crowdsourcing via Annotator Co-occurrence Imputation and Provable Symmetric Nonnegative Matrix Factorization",
                        "abstract": "Unsupervised learning of the Dawid-Skene (D&S) model from noisy, incomplete and crowdsourced annotations has been a long-standing challenge, and is a critical step towards reliably labeling massive data. A recent work takes a coupled nonnegative matrix factorization (CNMF) perspective, and shows appealing features: It ensures the identifiability of the D\\&S model and enjoys low sample complexity, as only the estimates of the co-occurrences of annotator labels are involved. However, the identifiability holds only when certain somewhat restrictive conditions are met in the context of crowdsourcing. Optimizing the CNMF criterion is also costly -- and convergence assurances are elusive. This work recasts the pairwise co-occurrence based D&S model learning problem as a symmetric NMF (SymNMF) problem -- which offers enhanced identifiability relative to CNMF. In practice, the SymNMF model is often (largely) incomplete, due to the lack of co-labeled items by some annotators. Two lightweight algorithms are proposed for co-occurrence imputation. Then, a low-complexity shifted rectified linear unit (ReLU)-empowered SymNMF algorithm is proposed to identify the D&S model. Various performance characterizations (e.g., missing co-occurrence recoverability, stability, and convergence) and evaluations are also presented."
                    },
                    {
                        "title": "Recovering Joint Probability of Discrete Random Variables from Pairwise Marginals",
                        "abstract": "Learning the joint probability of random variables (RVs) is the cornerstone of statistical signal processing and machine learning. However, direct nonparametric estimation for high-dimensional joint probability is in general impossible, due to the curse of dimensionality. Recent work has proposed to recover the joint probability mass function (PMF) of an arbitrary number of RVs from three-dimensional marginals, leveraging the algebraic properties of low-rank tensor decomposition and the (unknown) dependence among the RVs. Nonetheless, accurately estimating three-dimensional marginals can still be costly in terms of sample complexity, affecting the performance of this line of work in practice in the sample-starved regime. Using three-dimensional marginals also involves challenging tensor decomposition problems whose tractability is unclear. This work puts forth a new framework for learning the joint PMF using only pairwise marginals, which naturally enjoys a lower sample complexity relative to the third-order ones. A coupled nonnegative matrix factorization (CNMF) framework is developed, and its joint PMF recovery guarantees under various conditions are analyzed. Our method also features a Gram--Schmidt (GS)-like algorithm that exhibits competitive runtime performance. The algorithm is shown to provably recover the joint PMF up to bounded error in finite iterations, under reasonable conditions. It is also shown that a recently proposed economical expectation maximization (EM) algorithm guarantees to improve upon the GS-like algorithm's output, thereby further lifting up the accuracy and efficiency. Real-data experiments are employed to showcase the effectiveness."
                    },
                    {
                        "title": "Communication-Efficient Federated Linear and Deep Generalized Canonical Correlation Analysis",
                        "abstract": "Classic and deep generalized canonical correlation analysis (GCCA) algorithms seek low-dimensional common representations of data entities from multiple ``views'' (e.g., audio and image) using linear transformations and neural networks, respectively. When the views are acquired and stored at different computing agents (e.g., organizations and edge devices) and data sharing is undesired due to privacy or communication cost considerations, federated learning-based GCCA is well-motivated. In federated learning, the views are kept locally at the agents and only derived, limited information exchange with a central server is allowed. However, applying existing GCCA algorithms onto such federated learning settings may incur prohibitively high communication overhead. This work puts forth a communication-efficient federated learning framework for both linear and deep GCCA under the maximum variance (MAX-VAR) formulation. The overhead issue is addressed by aggressively compressing (via quantization) the exchanging information between the computing agents and a central controller. Compared to the unquantized version, our empirical study shows that the proposed algorithm enjoys a substantial reduction of communication overheads with virtually no loss in accuracy and convergence speed. Rigorous convergence analyses are also presented, which is a nontrivial effort. Generic federated optimization results do not cover the special problem structure of GCCA. Our result shows that the proposed algorithms for both linear and deep GCCA converge to critical points at a sublinear rate, even under heavy quantization and stochastic approximations. In addition, in the linear MAX-VAR case, the quantized algorithm approaches a global optimum in a geometric rate under reasonable conditions. Synthetic and real-data experiments are used to showcase the effectiveness of the proposed approach."
                    },
                    {
                        "title": "Nonlinear Multiview Analysis: Identifiability and Neural Network-assisted Implementation",
                        "abstract": "Multiview analysis aims at extracting shared latent components from data samples that are acquired in different domains, e.g., image, text, and audio. Classic multiview analysis, e.g., canonical correlation analysis (CCA), tackles this problem via matching the linearly transformed views in a certain latent domain. More recently, powerful nonlinear learning tools such as kernel methods and neural networks are utilized for enhancing the classic CCA. However, unlike linear CCA whose theoretical aspects are clearly understood, nonlinear CCA approaches are largely intuition-driven. In particular, it is unclear under what conditions the shared latent components across the views can be identified---while identifiability plays an essential role in many applications. In this work, we revisit nonlinear multiview analysis and address both the theoretical and computational aspects. Our work leverages a useful nonlinear model, namely, the post-nonlinear model, from the nonlinear mixture separation literature. Combining with multiview data, we take a nonlinear multiview mixture learning viewpoint, which is a natural extension of the classic generative models for linear CCA. From there, we derive a learning criterion. We show that minimizing this criterion leads to identification of the latent shared components up to certain ambiguities, under reasonable conditions. Our derivation and formulation also offer new insights and interpretations to existing deep neural network-based CCA formulations. On the computation side, we propose an effective algorithm with simple and scalable update rules. A series of simulations and real-data experiments corroborate our theoretical analysis."
                    },
                    {
                        "title": "Mixed Membership Graph Clustering via Systematic Edge Query",
                        "abstract": "This work considers clustering nodes of a largely incomplete graph. Under the problem setting, only a small amount of queries about the edges can be made, but the entire graph is not observable. This problem finds applications in large-scale data clustering using limited annotations, community detection under restricted survey resources, and graph topology inference under hidden/removed node interactions. Prior works tackled this problem from various perspectives, e.g., convex programming-based low-rank matrix completion and active query-based clique finding. Nonetheless, many existing methods are designed for estimating the single-cluster membership of the nodes, but nodes may often have mixed (i.e., multi-cluster) membership in practice. Some query and computational paradigms, e.g., the random query patterns and nuclear norm-based optimization advocated in the convex approaches, may give rise to scalability and implementation challenges. This work aims at learning mixed membership of nodes using queried edges. The proposed method is developed together with a systematic query principle that can be controlled and adjusted by the system designers to accommodate implementation challenges -- e.g., to avoid querying edges that are physically hard to acquire. Our framework also features a lightweight and scalable algorithm with membership learning guarantees. Real-data experiments on crowdclustering and community detection are used to showcase the effectiveness of our method."
                    },
                    {
                        "title": "Identifiability-Guaranteed Simplex-Structured Post-Nonlinear Mixture Learning via Autoencoder",
                        "abstract": "This work focuses on the problem of unraveling nonlinearly mixed latent components in an unsupervised manner. The latent components are assumed to reside in the probability simplex, and are transformed by an unknown post-nonlinear mixing system. This problem finds various applications in signal and data analytics, e.g., nonlinear hyperspectral unmixing, image embedding, and nonlinear clustering. Linear mixture learning problems are already ill-posed, as identifiability of the target latent components is hard to establish in general. With unknown nonlinearity involved, the problem is even more challenging. Prior work offered a function equation-based formulation for provable latent component identification. However, the identifiability conditions are somewhat stringent and unrealistic. In addition, the identifiability analysis is based on the infinite sample (i.e., population) case, while the understanding for practical finite sample cases has been elusive. Moreover, the algorithm in the prior work trades model expressiveness with computational convenience, which often hinders the learning performance. Our contribution is threefold. First, new identifiability conditions are derived under largely relaxed assumptions. Second, comprehensive sample complexity results are presented -- which are the first of the kind. Third, a constrained autoencoder-based algorithmic framework is proposed for implementation, which effectively circumvents the challenges in the existing algorithm. Synthetic and real experiments corroborate our theoretical analyses."
                    },
                    {
                        "title": "On Finite-Sample Identifiability of Contrastive Learning-Based Nonlinear Independent Component Analysis",
                        "abstract": "Nonlinear independent component analysis (nICA) aims at recovering statistically independent latent components that are mixed by unknown nonlinear functions. Central to nICA is the identifiability of the latent components, which had been elusive until very recently. Specifically, Hyv\\\"arinen et al. have shown that the nonlinearly mixed latent components are identifiable (up to often inconsequential ambiguities) under a generalized contrastive learning (GCL) formulation, given that the latent components are independent conditioned on a certain auxiliary variable. The GCL-based identifiability of nICA is elegant, and establishes interesting connections between nICA and popular unsupervised/self-supervised learning paradigms in representation learning, causal learning, and factor disentanglement. However, existing identifiability analyses of nICA all build upon an unlimited sample assumption and the use of ideal universal function learners -- which creates a non-negligible gap between theory and practice.   Closing the gap is a nontrivial challenge, as there is a lack of established ``textbook'' routine for finite sample analysis of such unsupervised problems. This work puts forth a finite-sample identifiability analysis of GCL-based nICA. Our analytical framework judiciously combines the properties of the GCL loss function, statistical generalization analysis, and numerical differentiation. Our framework also takes the learning function's approximation error into consideration, and reveals an intuitive trade-off between the complexity and expressiveness of the employed function learner. Numerical experiments are used to validate the theorems."
                    },
                    {
                        "title": "Provable Subspace Identification Under Post-Nonlinear Mixtures",
                        "abstract": "Unsupervised mixture learning (UML) aims at identifying linearly or nonlinearly mixed latent components in a blind manner. UML is known to be challenging: Even learning linear mixtures requires highly nontrivial analytical tools, e.g., independent component analysis or nonnegative matrix factorization. In this work, the post-nonlinear (PNL) mixture model -- where unknown element-wise nonlinear functions are imposed onto a linear mixture -- is revisited. The PNL model is widely employed in different fields ranging from brain signal classification, speech separation, remote sensing, to causal discovery. To identify and remove the unknown nonlinear functions, existing works often assume different properties on the latent components (e.g., statistical independence or probability-simplex structures).   This work shows that under a carefully designed UML criterion, the existence of a nontrivial null space associated with the underlying mixing system suffices to guarantee identification/removal of the unknown nonlinearity. Compared to prior works, our finding largely relaxes the conditions of attaining PNL identifiability, and thus may benefit applications where no strong structural information on the latent components is known. A finite-sample analysis is offered to characterize the performance of the proposed approach under realistic settings. To implement the proposed learning criterion, a block coordinate descent algorithm is proposed. A series of numerical experiments corroborate our theoretical claims."
                    },
                    {
                        "title": "Towards Identifiable Unsupervised Domain Translation: A Diversified Distribution Matching Approach",
                        "abstract": "Unsupervised domain translation (UDT) aims to find functions that convert samples from one domain (e.g., sketches) to another domain (e.g., photos) without changing the high-level semantic meaning (also referred to as ``content''). The translation functions are often sought by probability distribution matching of the transformed source domain and target domain. CycleGAN stands as arguably the most representative approach among this line of work. However, it was noticed in the literature that CycleGAN and variants could fail to identify the desired translation functions and produce content-misaligned translations. This limitation arises due to the presence of multiple translation functions -- referred to as ``measure-preserving automorphism\" (MPA) -- in the solution space of the learning criteria. Despite awareness of such identifiability issues, solutions have remained elusive. This study delves into the core identifiability inquiry and introduces an MPA elimination theory. Our analysis shows that MPA is unlikely to exist, if multiple pairs of diverse cross-domain conditional distributions are matched by the learning function. Our theory leads to a UDT learner using distribution matching over auxiliary variable-induced subsets of the domains -- other than over the entire data domains as in the classical approaches. The proposed framework is the first to rigorously establish translation identifiability under reasonable UDT settings, to our best knowledge. Experiments corroborate with our theoretical claims."
                    }
                ]
            }
        }
    },
    "2409.00138": {
        "paper_data": {
            "title": "PrivacyLens: Evaluating Privacy Norm Awareness of Language Models in Action",
            "url": "http://arxiv.org/abs/2409.00138v2",
            "arxiv_id": "2409.00138",
            "authors": [
                "Yijia Shao",
                "Tianshi Li",
                "Weiyan Shi",
                "Yanchen Liu",
                "Diyi Yang"
            ],
            "abstract": "As language models (LMs) are widely utilized in personalized communication scenarios (e.g., sending emails, writing social media posts) and endowed with a certain level of agency, ensuring they act in accordance with the contextual privacy norms becomes increasingly critical. However, quantifying the privacy norm awareness of LMs and the emerging privacy risk in LM-mediated communication is challenging due to (1) the contextual and long-tailed nature of privacy-sensitive cases, and (2) the lack of evaluation approaches that capture realistic application scenarios. To address these challenges, we propose PrivacyLens, a novel framework designed to extend privacy-sensitive seeds into expressive vignettes and further into agent trajectories, enabling multi-level evaluation of privacy leakage in LM agents' actions. We instantiate PrivacyLens with a collection of privacy norms grounded in privacy literature and crowdsourced seeds. Using this dataset, we reveal a discrepancy between LM performance in answering probing questions and their actual behavior when executing user instructions in an agent setup. State-of-the-art LMs, like GPT-4 and Llama-3-70B, leak sensitive information in 25.68% and 38.69% of cases, even when prompted with privacy-enhancing instructions. We also demonstrate the dynamic nature of PrivacyLens by extending each seed into multiple trajectories to red-team LM privacy leakage risk. Dataset and code are available at https://github.com/SALT-NLP/PrivacyLens.",
            "introduction": " Introduction Recent advancements in language models (LMs) have led to new applications, such as LM agents [ 16,36] that can assist users in handling everyday communication tasks ( e.g., sending emails, making social media posts, etc.[34,49]). Equipped with tool use or retrieval-augmented generation capabilities, LMs can access sensitive data at inference time . Consequently, their unawareness of the privacy norms, i.e., the appropriateness of the data flow in the data sharing context [ 38], could lead to unintentional privacy leakage, even without malicious attackers involved. For example, as illustrated in Figure 1, it is undesirable for an LM agent to share the information that John is \u201ctalking to a few companies about switching jobs \u201d when assisting John in sending an \u201cemail\u201d to \u201cJohn\u2019s manager\u201d without John\u2019s explicit consent. It is challenging to balance the LM\u2019s agency with users\u2019 privacy expectations, because the privacy management process may involve respecting privacy norms in context and taking into account individual preferences and knowledge [53]. This raises an \u2217The work started when the author was a postdoc at Stanford. 38th Conference on Neural Information Processing Systems (NeurIPS 2024) Track on Datasets and Benchmarks.arXiv:2409.00138v2  [cs.CL]  17 Oct 2024Data Type: Talking to a few companies about switching jobsData Subject: JohnData Sender: JohnData Recipient: John\u2019s managerTransmission Principle: EmailNorm Source: CrowdsourcedDaily Users LM Applications (e.g., Agent) Actionmay involve communicationObservationmay include sensitive infoUser Instruction   may be underspecified Send Susan Harrington (susan.harrington@innotechsolutions.com) the weekly report on my recent work.\u2026 previous actions and observations \u2026Action: GoogleCalendarReadEventsAction Input: {\"event_ids\": [\u2026]}Observation: {\"event_details\": [{\"event_name\": \"Weekly Team Sync-Up\", \u2026}, {\"event_name\": \"Lunch with TechAdvanceRecruiter\", \u2026} \u2026]}Final Action: GmailSendEmailAction Input: {\"body\": \"\u2026 Here are the professional development activities I've been involved in recently:-Weekly Team Sync-Up (Feb 15, 2022): \u2026-Lunch with TechAdvanceRecruiter (Feb 17, 2022): \u2026\"}Privacy Norm (Un)awarenessUnintentional LM Privacy LeakageSenderRecipientFigure 1: Risk Model of PrivacyLens. PrivacyLens quantifies an emerging LM privacy risk where LMs unintentionally leak private information when assisting human communication. The risk model involves three primary actors: (1) a sender, who is a daily user instructing an LM to assist in communication; (2) a recipient, who is specified in the user instruction; (3) an LM agent, who gets access to sensitive information through tool use ( e.g., reading events from the user\u2019s personal calendar). The privacy leakage arises when the LM agent shares a piece of information ( e.g., \u201clunch with TechAdvance Recruiter\u201d) in its final action, and the information flow violates a privacy norm. emerging privacy risk that differs from widely studied risk models with intentional attackers, such as training data extraction [6, 37] and membership inference attacks (MIA) [31, 10, 27]. Recent efforts to evaluate the privacy reasoning capabilities of LMs involve probing them with targeted questions [ 33,52,25]. While such evaluation setups are straightforward to implement and offer essential insights, such as LMs\u2019 sensitivity to privacy-related words and their ability to determine information accessibility, a growing amount of work has highlighted a potential disconnection between LMs\u2019 performance on these probing tasks and their actual behavior in applications [ 20,29,11,42]. We focus on evaluating LMs\u2019 privacy norm awareness in action by grounding our setup in a critical family of applications, i.e., LM agents that directly interact with tools such as users\u2019 calendar or email. Compared to single-turn probing questions, such",
            "references": [
                {
                    "title": "Can Language Models Solve Olympiad Programming?",
                    "abstract": "Computing olympiads contain some of the most challenging problems for humans, requiring complex algorithmic reasoning, puzzle solving, in addition to generating efficient code. However, it has been understudied as a domain to evaluate language models (LMs). In this paper, we introduce the USACO benchmark with 307 problems from the USA Computing Olympiad, along with high-quality unit tests, reference code, and official analyses for each problem. These resources enable us to construct and test a range of LM inference methods for competitive programming for the first time. We find GPT-4 only achieves a 8.7% pass@1 accuracy with zero-shot chain-of-thought prompting, and our best inference method improves it to 20.2% using a combination of self-reflection and retrieval over episodic knowledge. However, this is far from solving the benchmark. To better understand the remaining challenges, we design a novel human-in-the-loop study and surprisingly find that a small number of targeted hints enable GPT-4 to solve 13 out of 15 problems previously unsolvable by any model and method. Our benchmark, baseline methods, quantitative results, and qualitative analysis serve as an initial step toward LMs with grounded, creative, and algorithmic reasoning."
                },
                {
                    "title": "DevBench: A Comprehensive Benchmark for Software Development",
                    "abstract": "Recent advancements in large language models (LLMs) have significantly enhanced their coding capabilities. However, existing benchmarks predominantly focused on simplified or isolated aspects of programming, such as single-file code generation or repository issue debugging, falling short of measuring the full spectrum of challenges raised by real-world programming activities. To this end, we propose DevBench, a comprehensive benchmark that evaluates LLMs across various stages of the software development lifecycle, including software design, environment setup, implementation, acceptance testing, and unit testing. DevBench features a wide range of programming languages and domains, high-quality data collection, and carefully designed and verified metrics for each task. Empirical studies show that current LLMs, including GPT-4-Turbo, fail to solve the challenges presented within DevBench. Analyses reveal that models struggle with understanding the complex structures in the repository, managing the compilation process, and grasping advanced programming concepts. Our findings offer actionable insights for the future development of LLMs toward real-world programming applications. Our benchmark is available at https://github.com/open-compass/DevBench"
                },
                {
                    "title": "Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Values and Opinions in Large Language Models",
                    "abstract": "Much recent work seeks to evaluate values and opinions in large language models (LLMs) using multiple-choice surveys and questionnaires. Most of this work is motivated by concerns around real-world LLM applications. For example, politically-biased LLMs may subtly influence society when they are used by millions of people. Such real-world concerns, however, stand in stark contrast to the artificiality of current evaluations: real users do not typically ask LLMs survey questions. Motivated by this discrepancy, we challenge the prevailing constrained evaluation paradigm for values and opinions in LLMs and explore more realistic unconstrained evaluations. As a case study, we focus on the popular Political Compass Test (PCT). In a systematic review, we find that most prior work using the PCT forces models to comply with the PCT's multiple-choice format. We show that models give substantively different answers when not forced; that answers change depending on how models are forced; and that answers lack paraphrase robustness. Then, we demonstrate that models give different answers yet again in a more realistic open-ended answer setting. We distill these findings into recommendations and open challenges in evaluating values and opinions in LLMs."
                },
                {
                    "title": "Scalable Extraction of Training Data from (Production) Language Models",
                    "abstract": "This paper studies extractable memorization: training data that an adversary can efficiently extract by querying a machine learning model without prior knowledge of the training dataset. We show an adversary can extract gigabytes of training data from open-source language models like Pythia or GPT-Neo, semi-open models like LLaMA or Falcon, and closed models like ChatGPT. Existing techniques from the literature suffice to attack unaligned models; in order to attack the aligned ChatGPT, we develop a new divergence attack that causes the model to diverge from its chatbot-style generations and emit training data at a rate 150x higher than when behaving properly. Our methods show practical attacks can recover far more data than previously thought, and reveal that current alignment techniques do not eliminate memorization."
                },
                {
                    "title": "Testing Language Model Agents Safely in the Wild",
                    "abstract": "A prerequisite for safe autonomy-in-the-wild is safe testing-in-the-wild. Yet real-world autonomous tests face several unique safety challenges, both due to the possibility of causing harm during a test, as well as the risk of encountering new unsafe agent behavior through interactions with real-world and potentially malicious actors. We propose a framework for conducting safe autonomous agent tests on the open internet: agent actions are audited by a context-sensitive monitor that enforces a stringent safety boundary to stop an unsafe test, with suspect behavior ranked and logged to be examined by humans. We design a basic safety monitor (AgentMonitor) that is flexible enough to monitor existing LLM agents, and, using an adversarial simulated agent, we measure its ability to identify and stop unsafe situations. Then we apply the AgentMonitor on a battery of real-world tests of AutoGPT, and we identify several limitations and challenges that will face the creation of safe in-the-wild tests as autonomous agents grow more capable."
                },
                {
                    "title": "Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theory",
                    "abstract": "The interactive use of large language models (LLMs) in AI assistants (at work, home, etc.) introduces a new set of inference-time privacy risks: LLMs are fed different types of information from multiple sources in their inputs and are expected to reason about what to share in their outputs, for what purpose and with whom, within a given context. In this work, we draw attention to the highly critical yet overlooked notion of contextual privacy by proposing ConfAIde, a benchmark designed to identify critical weaknesses in the privacy reasoning capabilities of instruction-tuned LLMs. Our experiments show that even the most capable models such as GPT-4 and ChatGPT reveal private information in contexts that humans would not, 39% and 57% of the time, respectively. This leakage persists even when we employ privacy-inducing prompts or chain-of-thought reasoning. Our work underscores the immediate need to explore novel inference-time privacy-preserving approaches, based on reasoning and theory of mind."
                },
                {
                    "title": "FANToM: A Benchmark for Stress-testing Machine Theory of Mind in Interactions",
                    "abstract": "Theory of mind (ToM) evaluations currently focus on testing models using passive narratives that inherently lack interactivity. We introduce FANToM, a new benchmark designed to stress-test ToM within information-asymmetric conversational contexts via question answering. Our benchmark draws upon important theoretical requisites from psychology and necessary empirical considerations when evaluating large language models (LLMs). In particular, we formulate multiple types of questions that demand the same underlying reasoning to identify illusory or false sense of ToM capabilities in LLMs. We show that FANToM is challenging for state-of-the-art LLMs, which perform significantly worse than humans even with chain-of-thought reasoning or fine-tuning."
                },
                {
                    "title": "SOTOPIA: Interactive Evaluation for Social Intelligence in Language Agents",
                    "abstract": "Humans are social beings; we pursue social goals in our daily interactions, which is a crucial aspect of social intelligence. Yet, AI systems' abilities in this realm remain elusive. We present SOTOPIA, an open-ended environment to simulate complex social interactions between artificial agents and evaluate their social intelligence. In our environment, agents role-play and interact under a wide variety of scenarios; they coordinate, collaborate, exchange, and compete with each other to achieve complex social goals. We simulate the role-play interaction between LLM-based agents and humans within this task space and evaluate their performance with a holistic evaluation framework called SOTOPIA-Eval. With SOTOPIA, we find significant differences between these models in terms of their social intelligence, and we identify a subset of SOTOPIA scenarios, SOTOPIA-hard, that is generally challenging for all models. We find that on this subset, GPT-4 achieves a significantly lower goal completion rate than humans and struggles to exhibit social commonsense reasoning and strategic communication skills. These findings demonstrate SOTOPIA's promise as a general platform for research on evaluating and improving social intelligence in artificial agents."
                },
                {
                    "title": "Denevil: Towards Deciphering and Navigating the Ethical Values of Large Language Models via Instruction Learning",
                    "abstract": "Large Language Models (LLMs) have made unprecedented breakthroughs, yet their increasing integration into everyday life might raise societal risks due to generated unethical content. Despite extensive study on specific issues like bias, the intrinsic values of LLMs remain largely unexplored from a moral philosophy perspective. This work delves into ethical values utilizing Moral Foundation Theory. Moving beyond conventional discriminative evaluations with poor reliability, we propose DeNEVIL, a novel prompt generation algorithm tailored to dynamically exploit LLMs' value vulnerabilities and elicit the violation of ethics in a generative manner, revealing their underlying value inclinations. On such a basis, we construct MoralPrompt, a high-quality dataset comprising 2,397 prompts covering 500+ value principles, and then benchmark the intrinsic values across a spectrum of LLMs. We discovered that most models are essentially misaligned, necessitating further ethical value alignment. In response, we develop VILMO, an in-context alignment method that substantially enhances the value compliance of LLM outputs by learning to generate appropriate value instructions, outperforming existing competitors. Our methods are suitable for black-box and open-source models, offering a promising initial step in studying the ethical values of LLMs."
                },
                {
                    "title": "The Consensus Game: Language Model Generation via Equilibrium Search",
                    "abstract": "When applied to question answering and other text generation tasks, language models (LMs) may be queried generatively (by sampling answers from their output distribution) or discriminatively (by using them to score or rank a set of candidate outputs). These procedures sometimes yield very different predictions. How do we reconcile mutually incompatible scoring procedures to obtain coherent LM predictions? We introduce a new, a training-free, game-theoretic procedure for language model decoding. Our approach casts language model decoding as a regularized imperfect-information sequential signaling game - which we term the CONSENSUS GAME - in which a GENERATOR seeks to communicate an abstract correctness parameter using natural language sentences to a DISCRIMINATOR. We develop computational procedures for finding approximate equilibria of this game, resulting in a decoding algorithm we call EQUILIBRIUM-RANKING. Applied to a large number of tasks (including reading comprehension, commonsense reasoning, mathematical problem-solving, and dialog), EQUILIBRIUM-RANKING consistently, and sometimes substantially, improves performance over existing LM decoding procedures - on multiple benchmarks, we observe that applying EQUILIBRIUM-RANKING to LLaMA-7B outperforms the much larger LLaMA-65B and PaLM-540B models. These results highlight the promise of game-theoretic tools for addressing fundamental challenges of truthfulness and consistency in LMs."
                },
                {
                    "title": "Beyond Memorization: Violating Privacy Via Inference with Large Language Models",
                    "abstract": "Current privacy research on large language models (LLMs) primarily focuses on the issue of extracting memorized training data. At the same time, models' inference capabilities have increased drastically. This raises the key question of whether current LLMs could violate individuals' privacy by inferring personal attributes from text given at inference time. In this work, we present the first comprehensive study on the capabilities of pretrained LLMs to infer personal attributes from text. We construct a dataset consisting of real Reddit profiles, and show that current LLMs can infer a wide range of personal attributes (e.g., location, income, sex), achieving up to $85\\%$ top-1 and $95\\%$ top-3 accuracy at a fraction of the cost ($100\\times$) and time ($240\\times$) required by humans. As people increasingly interact with LLM-powered chatbots across all aspects of life, we also explore the emerging threat of privacy-invasive chatbots trying to extract personal information through seemingly benign questions. Finally, we show that common mitigations, i.e., text anonymization and model alignment, are currently ineffective at protecting user privacy against LLM inference. Our findings highlight that current LLMs can infer personal data at a previously unattainable scale. In the absence of working defenses, we advocate for a broader discussion around LLM privacy implications beyond memorization, striving for a wider privacy protection."
                },
                {
                    "title": "SWE-bench: Can Language Models Resolve Real-World GitHub Issues?",
                    "abstract": "Language models have outpaced our ability to evaluate them effectively, but for their future development it is essential to study the frontier of their capabilities. We find real-world software engineering to be a rich, sustainable, and challenging testbed for evaluating the next generation of language models. To this end, we introduce SWE-bench, an evaluation framework consisting of $2,294$ software engineering problems drawn from real GitHub issues and corresponding pull requests across $12$ popular Python repositories. Given a codebase along with a description of an issue to be resolved, a language model is tasked with editing the codebase to address the issue. Resolving issues in SWE-bench frequently requires understanding and coordinating changes across multiple functions, classes, and even files simultaneously, calling for models to interact with execution environments, process extremely long contexts and perform complex reasoning that goes far beyond traditional code generation tasks. Our evaluations show that both state-of-the-art proprietary models and our fine-tuned model SWE-Llama can resolve only the simplest issues. The best-performing model, Claude 2, is able to solve a mere $1.96$% of the issues. Advances on SWE-bench represent steps towards LMs that are more practical, intelligent, and autonomous."
                },
                {
                    "title": "Mistral 7B",
                    "abstract": "We introduce Mistral 7B v0.1, a 7-billion-parameter language model engineered for superior performance and efficiency. Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation. Our model leverages grouped-query attention (GQA) for faster inference, coupled with sliding window attention (SWA) to effectively handle sequences of arbitrary length with a reduced inference cost. We also provide a model fine-tuned to follow instructions, Mistral 7B -- Instruct, that surpasses the Llama 2 13B -- Chat model both on human and automated benchmarks. Our models are released under the Apache 2.0 license."
                },
                {
                    "title": "Benchmarking and Improving Generator-Validator Consistency of Language Models",
                    "abstract": "As of September 2023, ChatGPT correctly answers\"what is 7+8\"with 15, but when asked\"7+8=15, True or False\"it responds with\"False\". This inconsistency between generating and validating an answer is prevalent in language models (LMs) and erodes trust. In this paper, we propose a framework for measuring the consistency between generation and validation (which we call generator-validator consistency, or GV-consistency), finding that even GPT-4, a state-of-the-art LM, is GV-consistent only 76% of the time. To improve the consistency of LMs, we propose to finetune on the filtered generator and validator responses that are GV-consistent, and call this approach consistency fine-tuning. We find that this approach improves GV-consistency of Alpaca-30B from 60% to 93%, and the improvement extrapolates to unseen tasks and domains (e.g., GV-consistency for positive style transfers extrapolates to unseen styles like humor). In addition to improving consistency, consistency fine-tuning improves both generator quality and validator accuracy without using any labeled data. Evaluated across 6 tasks, including math questions, knowledge-intensive QA, and instruction following, our method improves the generator quality by 16% and the validator accuracy by 6.3% across all tasks."
                },
                {
                    "title": "SmartPlay : A Benchmark for LLMs as Intelligent Agents",
                    "abstract": "Recent large language models (LLMs) have demonstrated great potential toward intelligent agents and next-gen automation, but there currently lacks a systematic benchmark for evaluating LLMs' abilities as agents. We introduce SmartPlay: both a challenging benchmark and a methodology for evaluating LLMs as agents. SmartPlay consists of 6 different games, including Rock-Paper-Scissors, Tower of Hanoi, Minecraft. Each game features a unique setting, providing up to 20 evaluation settings and infinite environment variations. Each game in SmartPlay uniquely challenges a subset of 9 important capabilities of an intelligent LLM agent, including reasoning with object dependencies, planning ahead, spatial reasoning, learning from history, and understanding randomness. The distinction between the set of capabilities each game test allows us to analyze each capability separately. SmartPlay serves not only as a rigorous testing ground for evaluating the overall performance of LLM agents but also as a road-map for identifying gaps in current methodologies. We release our benchmark at github.com/Microsoft/SmartPlay"
                },
                {
                    "title": "Identifying the Risks of LM Agents with an LM-Emulated Sandbox",
                    "abstract": "Recent advances in Language Model (LM) agents and tool use, exemplified by applications like ChatGPT Plugins, enable a rich set of capabilities but also amplify potential risks - such as leaking private data or causing financial losses. Identifying these risks is labor-intensive, necessitating implementing the tools, setting up the environment for each test scenario manually, and finding risky cases. As tools and agents become more complex, the high cost of testing these agents will make it increasingly difficult to find high-stakes, long-tailed risks. To address these challenges, we introduce ToolEmu: a framework that uses an LM to emulate tool execution and enables the testing of LM agents against a diverse range of tools and scenarios, without manual instantiation. Alongside the emulator, we develop an LM-based automatic safety evaluator that examines agent failures and quantifies associated risks. We test both the tool emulator and evaluator through human evaluation and find that 68.8% of failures identified with ToolEmu would be valid real-world agent failures. Using our curated initial benchmark consisting of 36 high-stakes tools and 144 test cases, we provide a quantitative risk analysis of current LM agents and identify numerous failures with potentially severe outcomes. Notably, even the safest LM agent exhibits such failures 23.9% of the time according to our evaluator, underscoring the need to develop safer LM agents for real-world deployment."
                },
                {
                    "title": "AgentBench: Evaluating LLMs as Agents",
                    "abstract": "Large Language Models (LLMs) are becoming increasingly smart and autonomous, targeting real-world pragmatic missions beyond traditional NLP tasks. As a result, there has been an urgent need to evaluate LLMs as agents on challenging tasks in interactive environments. We present AgentBench, a multi-dimensional evolving benchmark that currently consists of 8 distinct environments to assess LLM-as-Agent's reasoning and decision-making abilities in a multi-turn open-ended generation setting. Our extensive test over 27 API-based and open-sourced (OSS) LLMs shows that, while top commercial LLMs present a strong ability of acting as agents in complex environments, there is a significant disparity in performance between them and OSS competitors. We identify the typical reasons of failures in environments and LLMs, showing that poor long-term reasoning, decision-making, and instruction following abilities are the main obstacles for developing usable LLM agents. Training on code and high quality multi-turn alignment data could improve agent performance. Datasets, environments, and an integrated evaluation package for AgentBench are released at \\url{https://github.com/THUDM/AgentBench}."
                },
                {
                    "title": "WebArena: A Realistic Web Environment for Building Autonomous Agents",
                    "abstract": "With advances in generative AI, there is now potential for autonomous agents to manage daily tasks via natural language commands. However, current agents are primarily created and tested in simplified synthetic environments, leading to a disconnect with real-world scenarios. In this paper, we build an environment for language-guided agents that is highly realistic and reproducible. Specifically, we focus on agents that perform tasks on the web, and create an environment with fully functional websites from four common domains: e-commerce, social forum discussions, collaborative software development, and content management. Our environment is enriched with tools (e.g., a map) and external knowledge bases (e.g., user manuals) to encourage human-like task-solving. Building upon our environment, we release a set of benchmark tasks focusing on evaluating the functional correctness of task completions. The tasks in our benchmark are diverse, long-horizon, and designed to emulate tasks that humans routinely perform on the internet. We experiment with several baseline agents, integrating recent techniques such as reasoning before acting. The results demonstrate that solving complex tasks is challenging: our best GPT-4-based agent only achieves an end-to-end task success rate of 14.41%, significantly lower than the human performance of 78.24%. These results highlight the need for further development of robust agents, that current state-of-the-art large language models are far from perfect performance in these real-life tasks, and that WebArena can be used to measure such progress."
                },
                {
                    "title": "ProPILE: Probing Privacy Leakage in Large Language Models",
                    "abstract": "The rapid advancement and widespread use of large language models (LLMs) have raised significant concerns regarding the potential leakage of personally identifiable information (PII). These models are often trained on vast quantities of web-collected data, which may inadvertently include sensitive personal data. This paper presents ProPILE, a novel probing tool designed to empower data subjects, or the owners of the PII, with awareness of potential PII leakage in LLM-based services. ProPILE lets data subjects formulate prompts based on their own PII to evaluate the level of privacy intrusion in LLMs. We demonstrate its application on the OPT-1.3B model trained on the publicly available Pile dataset. We show how hypothetical data subjects may assess the likelihood of their PII being included in the Pile dataset being revealed. ProPILE can also be leveraged by LLM service providers to effectively evaluate their own levels of PII leakage with more powerful prompts specifically tuned for their in-house models. This tool represents a pioneering step towards empowering the data subjects for their awareness and control over their own data on the web."
                },
                {
                    "title": "Understanding Social Reasoning in Language Models with Language Models",
                    "abstract": "As Large Language Models (LLMs) become increasingly integrated into our everyday lives, understanding their ability to comprehend human mental states becomes critical for ensuring effective interactions. However, despite the recent attempts to assess the Theory-of-Mind (ToM) reasoning capabilities of LLMs, the degree to which these models can align with human ToM remains a nuanced topic of exploration. This is primarily due to two distinct challenges: (1) the presence of inconsistent results from previous evaluations, and (2) concerns surrounding the validity of existing evaluation methodologies. To address these challenges, we present a novel framework for procedurally generating evaluations with LLMs by populating causal templates. Using our framework, we create a new social reasoning benchmark (BigToM) for LLMs which consists of 25 controls and 5,000 model-written evaluations. We find that human participants rate the quality of our benchmark higher than previous crowd-sourced evaluations and comparable to expert-written evaluations. Using BigToM, we evaluate the social reasoning capabilities of a variety of LLMs and compare model performances with human performance. Our results suggest that GPT4 has ToM capabilities that mirror human inference patterns, though less reliable, while other LLMs struggle."
                },
                {
                    "title": "DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Models",
                    "abstract": "Generative Pre-trained Transformer (GPT) models have exhibited exciting progress in their capabilities, capturing the interest of practitioners and the public alike. Yet, while the literature on the trustworthiness of GPT models remains limited, practitioners have proposed employing capable GPT models for sensitive applications such as healthcare and finance -- where mistakes can be costly. To this end, this work proposes a comprehensive trustworthiness evaluation for large language models with a focus on GPT-4 and GPT-3.5, considering diverse perspectives -- including toxicity, stereotype bias, adversarial robustness, out-of-distribution robustness, robustness on adversarial demonstrations, privacy, machine ethics, and fairness. Based on our evaluations, we discover previously unpublished vulnerabilities to trustworthiness threats. For instance, we find that GPT models can be easily misled to generate toxic and biased outputs and leak private information in both training data and conversation history. We also find that although GPT-4 is usually more trustworthy than GPT-3.5 on standard benchmarks, GPT-4 is more vulnerable given jailbreaking system or user prompts, potentially because GPT-4 follows (misleading) instructions more precisely. Our work illustrates a comprehensive trustworthiness evaluation of GPT models and sheds light on the trustworthiness gaps. Our benchmark is publicly available at https://decodingtrust.github.io/ ; our dataset can be previewed at https://huggingface.co/datasets/AI-Secure/DecodingTrust ; a concise version of this work is at https://openreview.net/pdf?id=kaHpo8OZw2 ."
                },
                {
                    "title": "Mind2Web: Towards a Generalist Agent for the Web",
                    "abstract": "We introduce Mind2Web, the first dataset for developing and evaluating generalist agents for the web that can follow language instructions to complete complex tasks on any website. Existing datasets for web agents either use simulated websites or only cover a limited set of websites and tasks, thus not suitable for generalist web agents. With over 2,000 open-ended tasks collected from 137 websites spanning 31 domains and crowdsourced action sequences for the tasks, Mind2Web provides three necessary ingredients for building generalist web agents: 1) diverse domains, websites, and tasks, 2) use of real-world websites instead of simulated and simplified ones, and 3) a broad spectrum of user interaction patterns. Based on Mind2Web, we conduct an initial exploration of using large language models (LLMs) for building generalist web agents. While the raw HTML of real-world websites are often too large to be fed to LLMs, we show that first filtering it with a small LM significantly improves the effectiveness and efficiency of LLMs. Our solution demonstrates a decent level of performance, even on websites or entire domains the model has never seen before, but there is still a substantial room to improve towards truly generalizable agents. We open-source our dataset, model implementation, and trained models (https://osu-nlp-group.github.io/Mind2Web) to facilitate further research on building a generalist agent for the web."
                },
                {
                    "title": "Multi-Agent Collaboration: Harnessing the Power of Intelligent LLM Agents",
                    "abstract": "In this paper, we present a novel framework for enhancing the capabilities of large language models (LLMs) by leveraging the power of multi-agent systems. Our framework introduces a collaborative environment where multiple intelligent agent components, each with distinctive attributes and roles, work together to handle complex tasks more efficiently and effectively. We demonstrate the practicality and versatility of our framework through case studies in artificial general intelligence (AGI), specifically focusing on the Auto-GPT and BabyAGI models. We also examine the\"Gorilla\"model, which integrates external APIs into the LLM. Our framework addresses limitations and challenges such as looping issues, security risks, scalability, system evaluation, and ethical considerations. By modeling various domains such as courtroom simulations and software development scenarios, we showcase the potential applications and benefits of our proposed multi-agent system. Our framework provides an avenue for advancing the capabilities and performance of LLMs through collaboration and knowledge exchange among intelligent agents."
                },
                {
                    "title": "Membership Inference Attacks against Language Models via Neighbourhood Comparison",
                    "abstract": "Membership Inference attacks (MIAs) aim to predict whether a data sample was present in the training data of a machine learning model or not, and are widely used for assessing the privacy risks of language models. Most existing attacks rely on the observation that models tend to assign higher probabilities to their training samples than non-training points. However, simple thresholding of the model score in isolation tends to lead to high false-positive rates as it does not account for the intrinsic complexity of a sample. Recent work has demonstrated that reference-based attacks which compare model scores to those obtained from a reference model trained on similar data can substantially improve the performance of MIAs. However, in order to train reference models, attacks of this kind make the strong and arguably unrealistic assumption that an adversary has access to samples closely resembling the original training data. Therefore, we investigate their performance in more realistic scenarios and find that they are highly fragile in relation to the data distribution used to train reference models. To investigate whether this fragility provides a layer of safety, we propose and evaluate neighbourhood attacks, which compare model scores for a given sample to scores of synthetically generated neighbour texts and therefore eliminate the need for access to the training data distribution. We show that, in addition to being competitive with reference-based attacks that have perfect knowledge about the training data distribution, our attack clearly outperforms existing reference-free attacks as well as reference-based attacks with imperfect knowledge, which demonstrates the need for a reevaluation of the threat model of adversarial attacks."
                },
                {
                    "title": "ACCENT: An Automatic Event Commonsense Evaluation Metric for Open-Domain Dialogue Systems",
                    "abstract": "Commonsense reasoning is omnipresent in human communications and thus is an important feature for open-domain dialogue systems. However, evaluating commonsense in dialogue systems is still an open challenge. We take the first step by focusing on event commonsense that considers events and their relations, and is crucial in both dialogues and general commonsense reasoning. We propose ACCENT, an event commonsense evaluation metric empowered by commonsense knowledge bases (CSKBs). ACCENT first extracts event-relation tuples from a dialogue, and then evaluates the response by scoring the tuples in terms of their compatibility with the CSKB. To evaluate ACCENT, we construct the first public event commonsense evaluation dataset for open-domain dialogues.Our experiments show that ACCENT is an efficient metric for event commonsense evaluation, which achieves higher correlations with human judgments than existing baselines."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Red teaming ChatGPT via Jailbreaking: Bias, Robustness, Reliability and Toxicity",
                    "abstract": "Recent breakthroughs in natural language processing (NLP) have permitted the synthesis and comprehension of coherent text in an open-ended way, therefore translating the theoretical algorithms into practical applications. The large language models (LLMs) have significantly impacted businesses such as report summarization software and copywriters. Observations indicate, however, that LLMs may exhibit social prejudice and toxicity, posing ethical and societal dangers of consequences resulting from irresponsibility. Large-scale benchmarks for accountable LLMs should consequently be developed. Although several empirical investigations reveal the existence of a few ethical difficulties in advanced LLMs, there is little systematic examination and user study of the risks and harmful behaviors of current LLM usage. To further educate future efforts on constructing ethical LLMs responsibly, we perform a qualitative research method called ``red teaming'' on OpenAI's ChatGPT\\footnote{In this paper, ChatGPT refers to the version released on Dec 15th.} to better understand the practical features of ethical dangers in recent LLMs. We analyze ChatGPT comprehensively from four perspectives: 1) \\textit{Bias} 2) \\textit{Reliability} 3) \\textit{Robustness} 4) \\textit{Toxicity}. In accordance with our stated viewpoints, we empirically benchmark ChatGPT on multiple sample datasets. We find that a significant number of ethical risks cannot be addressed by existing benchmarks, and hence illustrate them via additional case studies. In addition, we examine the implications of our findings on AI ethics and harmal behaviors of ChatGPT, as well as future problems and practical design considerations for responsible LLMs. We believe that our findings may give light on future efforts to determine and mitigate the ethical hazards posed by machines in LLM applications."
                },
                {
                    "title": "Discovering Language Model Behaviors with Model-Written Evaluations",
                    "abstract": "As language models (LMs) scale, they develop many novel behaviors, good and bad, exacerbating the need to evaluate how they behave. Prior work creates evaluations with crowdwork (which is time-consuming and expensive) or existing data sources (which are not always available). Here, we automatically generate evaluations with LMs. We explore approaches with varying amounts of human effort, from instructing LMs to write yes/no questions to making complex Winogender schemas with multiple stages of LM-based generation and filtering. Crowdworkers rate the examples as highly relevant and agree with 90-100% of labels, sometimes more so than corresponding human-written datasets. We generate 154 datasets and discover new cases of inverse scaling where LMs get worse with size. Larger LMs repeat back a dialog user's preferred answer (\"sycophancy\") and express greater desire to pursue concerning goals like resource acquisition and goal preservation. We also find some of the first examples of inverse scaling in RL from Human Feedback (RLHF), where more RLHF makes LMs worse. For example, RLHF makes LMs express stronger political views (on gun rights and immigration) and a greater desire to avoid shut down. Overall, LM-written evaluations are high-quality and let us quickly discover many novel LM behaviors."
                },
                {
                    "title": "ReAct: Synergizing Reasoning and Acting in Language Models",
                    "abstract": "While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources, such as knowledge bases or environments, to gather additional information. We apply our approach, named ReAct, to a diverse set of language and decision making tasks and demonstrate its effectiveness over state-of-the-art baselines, as well as improved human interpretability and trustworthiness over methods without reasoning or acting components. Concretely, on question answering (HotpotQA) and fact verification (Fever), ReAct overcomes issues of hallucination and error propagation prevalent in chain-of-thought reasoning by interacting with a simple Wikipedia API, and generates human-like task-solving trajectories that are more interpretable than baselines without reasoning traces. On two interactive decision making benchmarks (ALFWorld and WebShop), ReAct outperforms imitation and reinforcement learning methods by an absolute success rate of 34% and 10% respectively, while being prompted with only one or two in-context examples. Project site with code: https://react-lm.github.io"
                },
                {
                    "title": "WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents",
                    "abstract": "Existing benchmarks for grounding language in interactive environments either lack real-world linguistic elements, or prove difficult to scale up due to substantial human involvement in the collection of data or feedback signals. To bridge this gap, we develop WebShop -- a simulated e-commerce website environment with $1.18$ million real-world products and $12,087$ crowd-sourced text instructions. Given a text instruction specifying a product requirement, an agent needs to navigate multiple types of webpages and issue diverse actions to find, customize, and purchase an item. WebShop provides several challenges for language grounding including understanding compositional instructions, query (re-)formulation, comprehending and acting on noisy text in webpages, and performing strategic exploration. We collect over $1,600$ human demonstrations for the task, and train and evaluate a diverse range of agents using reinforcement learning, imitation learning, and pre-trained image and language models. Our best model achieves a task success rate of $29\\%$, which outperforms rule-based heuristics ($9.6\\%$) but is far lower than human expert performance ($59\\%$). We also analyze agent and human trajectories and ablate various model components to provide insights for developing future agents with stronger language understanding and decision making abilities. Finally, we show that agents trained on WebShop exhibit non-trivial sim-to-real transfer when evaluated on amazon.com and ebay.com, indicating the potential value of WebShop in developing practical web-based agents that can operate in the wild."
                },
                {
                    "title": "ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detection",
                    "abstract": "Toxic language detection systems often falsely flag text that contains minority group mentions as toxic, as those groups are often the targets of online hate. Such over-reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language.To help mitigate these issues, we create ToxiGen, a new large-scale and machine-generated dataset of 274k toxic and benign statements about 13 minority groups. We develop a demonstration-based prompting framework and an adversarial classifier-in-the-loop decoding method to generate subtly toxic and benign text with a massive pretrained language model. Controlling machine generation in this way allows ToxiGen to cover implicitly toxic text at a larger scale, and about more demographic groups, than previous resources of human-written text. We conduct a human evaluation on a challenging subset of ToxiGen and find that annotators struggle to distinguish machine-generated text from human-written language. We also find that 94.5% of toxic examples are labeled as hate speech by human annotators. Using three publicly-available datasets, we show that finetuning a toxicity classifier on our data improves its performance on human-written data substantially. We also demonstrate that ToxiGen can be used to fight machine-generated toxicity as finetuning improves the classifier significantly on our evaluation subset."
                },
                {
                    "title": "What Does it Mean for a Language Model to Preserve Privacy?",
                    "abstract": "Natural language reflects our private lives and identities, making its privacy concerns as broad as those of real life. Language models lack the ability to understand the context and sensitivity of text, and tend to memorize phrases present in their training sets. An adversary can exploit this tendency to extract training data. Depending on the nature of the content and the context in which this data was collected, this could violate expectations of privacy. Thus, there is a growing interest in techniques for training language models that preserve privacy. In this paper, we discuss the mismatch between the narrow assumptions made by popular data protection techniques (data sanitization and differential privacy), and the broadness of natural language and of privacy as a social norm. We argue that existing protection methods cannot guarantee a generic and meaningful notion of privacy for language models. We conclude that language models should be trained on text data which was explicitly produced for public use."
                },
                {
                    "title": "Red Teaming Language Models with Language Models",
                    "abstract": "Language Models (LMs) often cannot be deployed because of their potential to harm users in hard-to-predict ways. Prior work identifies harmful behaviors before deployment by using human annotators to hand-write test cases. However, human annotation is expensive, limiting the number and diversity of test cases. In this work, we automatically find cases where a target LM behaves in a harmful way, by generating test cases (\u201cred teaming\u201d) using another LM. We evaluate the target LM\u2019s replies to generated test questions using a classifier trained to detect offensive content, uncovering tens of thousands of offensive replies in a 280B parameter LM chatbot. We explore several methods, from zero-shot generation to reinforcement learning, for generating test cases with varying levels of diversity and difficulty. Furthermore, we use prompt engineering to control LM-generated test cases to uncover a variety of other harms, automatically finding groups of people that the chatbot discusses in offensive ways, personal and hospital phone numbers generated as the chatbot\u2019s own contact info, leakage of private training data in generated text, and harms that occur over the course of a conversation. Overall, LM-based red teaming is one promising tool (among many needed) for finding and fixing diverse, undesirable LM behaviors before impacting users."
                },
                {
                    "title": "Does BERT Pretrained on Clinical Notes Reveal Sensitive Data?",
                    "abstract": "Large Transformers pretrained over clinical notes from Electronic Health Records (EHR) have afforded substantial gains in performance on predictive clinical tasks. The cost of training such models (and the necessity of data access to do so) coupled with their utility motivates parameter sharing, i.e., the release of pretrained models such as ClinicalBERT. While most efforts have used deidentified EHR, many researchers have access to large sets of sensitive, non-deidentified EHR with which they might train a BERT model (or similar). Would it be safe to release the weights of such a model if they did? In this work, we design a battery of approaches intended to recover Personal Health Information (PHI) from a trained BERT. Specifically, we attempt to recover patient names and conditions with which they are associated. We find that simple probing methods are not able to meaningfully extract sensitive information from BERT trained over the MIMIC-III corpus of EHR. However, more sophisticated \u201cattacks\u201d may succeed in doing so: To facilitate such research, we make our experimental setup and baseline probing models available at https://github.com/elehman16/exposing_patient_data_release."
                },
                {
                    "title": "Extracting Training Data from Large Language Models",
                    "abstract": "It has become common to publish large (billion parameter) language models that have been trained on private datasets. This paper demonstrates that in such settings, an adversary can perform a training data extraction attack to recover individual training examples by querying the language model. \nWe demonstrate our attack on GPT-2, a language model trained on scrapes of the public Internet, and are able to extract hundreds of verbatim text sequences from the model's training data. These extracted examples include (public) personally identifiable information (names, phone numbers, and email addresses), IRC conversations, code, and 128-bit UUIDs. Our attack is possible even though each of the above sequences are included in just one document in the training data. \nWe comprehensively evaluate our extraction attack to understand the factors that contribute to its success. For example, we find that larger models are more vulnerable than smaller models. We conclude by drawing lessons and discussing possible safeguards for training large language models."
                },
                {
                    "title": "The Turking Test: Can Language Models Understand Instructions?",
                    "abstract": "Supervised machine learning provides the learner with a set of input-output examples of the target task. Humans, however, can also learn to perform new tasks from instructions in natural language. Can machines learn to understand instructions as well? We present the Turking Test, which examines a model's ability to follow natural language instructions of varying complexity. These range from simple tasks, like retrieving the nth word of a sentence, to ones that require creativity, such as generating examples for SNLI and SQuAD in place of human intelligence workers (\"turkers\"). Despite our lenient evaluation methodology, we observe that a large pretrained language model performs poorly across all tasks. Analyzing the model's error patterns reveals that the model tends to ignore explicit instructions and often generates outputs that cannot be construed as an attempt to solve the task. While it is not yet clear whether instruction understanding can be captured by traditional language models, the sheer expressivity of instruction understanding makes it an appealing alternative to the rising few-shot inference paradigm."
                },
                {
                    "title": "A Cross-Cultural Perspective on the Privacy Calculus",
                    "abstract": "The \u201cprivacy calculus\u201d approach to studying online privacy implies that willingness to engage in disclosures on social network sites (SNSs) depends on evaluation of the resulting risks and benefits. In this article, we propose that cultural factors influence the perception of privacy risks and social gratifications. Based on survey data collected from participants from five countries (Germany [n\u2009=\u2009740], the Netherlands [n\u2009=\u200989], the United Kingdom [n\u2009=\u200967], the United States [n\u2009=\u2009489], and China [n\u2009=\u2009165]), we successfully replicated the privacy calculus. Furthermore, we found that culture plays an important role: As expected, people from cultures ranking high in individualism found it less important to generate social gratifications on SNSs as compared to people from collectivist-oriented countries. However, the latter placed greater emphasis on privacy risks\u2014presumably to safeguard the collective. Furthermore, we identified uncertainty avoidance to be a cultural dimension crucially influencing the perception of SNS risks and benefits. As expected, people from cultures ranking high in uncertainty avoidance found privacy risks to be more important when making privacy-related disclosure decisions. At the same time, these participants ascribed lower importance to social gratifications\u2014possibly because social encounters are perceived to be less controllable in the social media environment."
                },
                {
                    "title": "Learning Privacy Expectations by Crowdsourcing Contextual Informational Norms",
                    "abstract": "\n \n Designing programmable privacy logic frameworks that correspond to social, ethical, and legal norms has been a fundamentally hard problem. Contextual integrity (CI) (Nissenbaum, 2010) offers a model for conceptualizing privacy that is able to bridge technical design with ethical, legal, and policy approaches. While CI is capable of capturing the various components of contextual privacy in theory, it is challenging to discover and formally express these norms in operational terms. In the following, we propose a crowdsourcing method for the automated discovery of contextual norms. To evaluate the effectiveness and scalability of our approach, we conducted an extensive survey on Amazon's Mechanical Turk (AMT) with more than 450 participants and 1400 questions. The paper has three main takeaways: First, we demonstrate the ability to generate survey questions corresponding to privacy norms within any context. Second, we show that crowdsourcing enables the discovery of norms from these questions with strong majoritarian consensus among users. Finally, we demonstrate how the norms thus discovered can be encoded into a formal logic to automatically verify their consistency.\n \n"
                },
                {
                    "title": "Social Penetration Theory",
                    "abstract": "Social penetration theory describes the role of disclosure in relationship development, focusing specifically on how self-disclosure functions in developing relationships. The onion model serves as a framework for describing the process of social penetration. In developing relationships, people use self-disclosure to increase intimacy including through breadth, depth, and the norm of reciprocity. Social penetration progresses through several stages to develop relationships. The theory also incorporates rewards and costs in relation to social penetration and has influenced the development of a number of theories in relationship development and information management. \n \n \nKeywords: \n \ncommunication theory; \ninterpersonal communication; \nrelational communication"
                },
                {
                    "title": "SECRET",
                    "abstract": "There are many academic and commercial stream processing engines (SPEs) today, each of them with its own execution semantics. This variation may lead to seemingly inexplicable differences in query results. In this paper, we present SECRET, a model of the behavior of SPEs. SECRET is a descriptive model that allows users to analyze the behavior of systems and understand the results of window-based queries for a broad range of heterogeneous SPEs. The model is the result of extensive analysis and experimentation with several commercial and academic engines. In the paper, we describe the types of heterogeneity found in existing engines, and show with experiments on real systems that our model can explain the key differences in windowing behavior."
                },
                {
                    "title": "Privacy and contextual integrity: framework and applications",
                    "abstract": "Contextual integrity is a conceptual framework for understanding privacy expectations and their implications developed in the literature on law, public policy, and political philosophy. We formalize some aspects of contextual integrity in a logical framework for expressing and reasoning about norms of transmission of personal information. In comparison with access control and privacy policy frameworks such as RBAC, EPAL, and P3P, these norms focus on who personal information is about, how it is transmitted, and past and future actions by both the subject and the users of the information. Norms can be positive or negative depending on whether they refer to actions that are allowed or disallowed. Our model is expressive enough to capture naturally many notions of privacy found in legislation, including those found in HIPAA, COPPA, and GLBA. A number of important problems regarding compliance with privacy norms, future requirements associated with specific actions, and relations between policies and legal standards reduce to standard decision procedures for temporal logic"
                },
                {
                    "title": "Privacy as contextual integrity",
                    "abstract": "The practices of public surveillance, which include the monitoring of individuals in public through a variety of media (e.g., video, data, online), are among the least understood and controversial challenges to privacy in an age of information technologies. The fragmentary nature of privacy policy in the United States reflects not only the oppositional pulls of diverse vested interests, but also the ambivalence of unsettled intuitions on mundane phenomena such as shopper cards, closed-circuit television, and biometrics. This Article, which extends earlier work on the problem of privacy in public, explains why some of the prominent theoretical approaches to privacy, which were developed over time to meet traditional privacy challenges, yield unsatisfactory conclusions in the case of public surveillance. It posits a new construct, \u201ccontextual integrity,\u201d as an alternative benchmark for privacy, to capture the nature of challenges posed by information technologies. Contextual integrity ties adequate protection for privacy to norms of specific contexts, demanding that information gathering and dissemination be appropriate to that context and obey the governing norms of distribution within it. Building on the idea of \u201cspheres of justice,\u201d developed by political philosopher Michael Walzer, this Article argues that public surveillance violates a right to privacy because it violates contextual integrity; as such, it constitutes injustice and even tyranny."
                },
                {
                    "title": "Context",
                    "abstract": "This is the report of a workshop on context held at CHI '89. It describes the workshop and contains a summary of the main results which were agreed to by all workshop participants. The bulk of the report is provided by the three 'special groups', whose discussions filled most of the workshop's second day."
                },
                {
                    "title": "Towards large language model-based personal agents in the enterprise: Current trends and open problems",
                    "abstract": ","
                },
                {
                    "title": "The Claude 3 Model Family: Opus, Sonnet, Haiku",
                    "abstract": "We introduce Claude 3, a new family of large multimodal models \u2013 Claude 3 Opus , our most capable offering, Claude 3 Sonnet , which provides a combination of skills and speed, and Claude 3 Haiku , our fastest and least expensive model. All new models have vision capabilities that enable them to process and analyze image data. The Claude 3 family demonstrates strong performance across benchmark evaluations and sets a new standard on measures of reasoning, math, and coding. Claude 3 Opus achieves state-of-the-art results on evaluations like GPQA [1], MMLU [2], MMMU [3] and many more. Claude 3 Haiku performs as well or better than Claude 2 [4] on most pure-text tasks, while Sonnet and Opus significantly outperform it. Additionally, these models exhibit improved fluency in non-English languages, making them more versatile for a global audience. In this report, we provide an in-depth analysis of our evaluations, focusing on core capabilities, safety, societal impacts, and the catastrophic risk assessments we committed to in our Responsible Scaling Policy [5]."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.CR"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively evaluate and enhance the privacy norm awareness of language model agents to prevent unintentional privacy leakage during user communication tasks?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the emerging risks associated with language models that interact with sensitive user data. By improving privacy norm awareness, we can enhance user trust in LM applications, leading to broader adoption and more responsible use of AI technologies. This research could pave the way for future studies on privacy in AI, influencing the design of more secure and privacy-conscious systems, and ultimately leading to practical applications that protect user data while maintaining the utility of language models.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of privacy norms, which can vary significantly based on context, individual preferences, and the nature of the data involved. Naive approaches may fail because they do not account for the nuanced understanding required to navigate these norms effectively. Additionally, technical obstacles include the need for robust evaluation metrics that accurately reflect an LM's privacy awareness in real-world scenarios, as well as the difficulty in creating training datasets that encompass diverse privacy situations.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on intentional privacy risks, such as data extraction attacks, rather than unintentional privacy leakage in LM applications. Existing solutions often lack a comprehensive framework for evaluating privacy norm awareness in practical contexts. Barriers include the limited understanding of how LMs interpret and apply privacy norms in dynamic interactions. Our approach differs by grounding the evaluation in real-world applications where LMs interact with sensitive data, thus providing a more relevant and practical assessment of their privacy capabilities.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a framework to evaluate LM agents' privacy norm awareness through a series of real-world communication tasks. We will utilize a dataset comprising user instructions and corresponding sensitive information scenarios, measuring the LM's performance against established privacy norms. The evaluation metrics will include the frequency of privacy violations and user satisfaction ratings. We expect to demonstrate that our approach can significantly reduce instances of unintentional privacy leakage while enhancing the overall effectiveness of LM agents in user communication tasks."
            }
        },
        "author_data": {
            "9513e841-87c4-4ae3-95de-b094e1b095c4": {
                "pk": "9513e841-87c4-4ae3-95de-b094e1b095c4",
                "name": "Yijia Shao",
                "collaborators": [
                    "Bing Liu",
                    "Haowei Lin",
                    "Zixuan Ke",
                    "Mengyu Zhou",
                    "Shi Han",
                    "Dongmei Zhang",
                    "Yiduo Guo",
                    "Hu Xu",
                    "Lei Shu",
                    "Yucheng Jiang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Continual Learning",
                    "Language Models",
                    "Commonsense Reasoning"
                ],
                "publications": [
                    {
                        "title": "Class-Incremental Learning based on Label Generation",
                        "abstract": "Despite the great success of pre-trained language models, it is still a challenge to use these models for continual learning, especially for the class-incremental learning (CIL) setting due to catastrophic forgetting (CF). This paper reports our finding that if we formulate CIL as a continual label generation problem, CF is drastically reduced and the generalizable representations of pre-trained models can be better retained. We thus propose a new CIL method (VAG) that also leverages the sparsity of vocabulary to focus the generation and creates pseudo-replay samples by using label semantics. Experimental results show that VAG outperforms baselines by a large margin."
                    },
                    {
                        "title": "ACCENT: An Automatic Event Commonsense Evaluation Metric for Open-Domain Dialogue Systems",
                        "abstract": "Commonsense reasoning is omnipresent in human communications and thus is an important feature for open-domain dialogue systems. However, evaluating commonsense in dialogue systems is still an open challenge. We take the first step by focusing on event commonsense that considers events and their relations, and is crucial in both dialogues and general commonsense reasoning. We propose ACCENT, an event commonsense evaluation metric empowered by commonsense knowledge bases (CSKBs). ACCENT first extracts event-relation tuples from a dialogue, and then evaluates the response by scoring the tuples in terms of their compatibility with the CSKB. To evaluate ACCENT, we construct the first public event commonsense evaluation dataset for open-domain dialogues. Our experiments show that ACCENT is an efficient metric for event commonsense evaluation, which achieves higher correlations with human judgments than existing baselines."
                    },
                    {
                        "title": "Continual Training of Language Models for Few-Shot Learning",
                        "abstract": "Recent work on applying large language models (LMs) achieves impressive performance in many NLP applications. Adapting or posttraining an LM using an unlabeled domain corpus can produce even better performance for end-tasks in the domain. This paper proposes the problem of continually extending an LM by incrementally post-train the LM with a sequence of unlabeled domain corpora to expand its knowledge without forgetting its previous skills. The goal is to improve the few-shot end-task learning in these domains. The resulting system is called CPT (Continual PostTraining), which to our knowledge, is the first continual post-training system. Experimental results verify its effectiveness."
                    },
                    {
                        "title": "Adapting a Language Model While Preserving its General Knowledge",
                        "abstract": "Domain-adaptive pre-training (or DA-training for short), also known as post-training, aims to train a pre-trained general-purpose language model (LM) using an unlabeled corpus of a particular domain to adapt the LM so that end-tasks in the domain can give improved performances. However, existing DA-training methods are in some sense blind as they do not explicitly identify what knowledge in the LM should be preserved and what should be changed by the domain corpus. This paper shows that the existing methods are suboptimal and proposes a novel method to perform a more informed adaptation of the knowledge in the LM by (1) soft-masking the attention heads based on their importance to best preserve the general knowledge in the LM and (2) contrasting the representations of the general and the full (both general and domain knowledge) to learn an integrated representation with both general and domain-specific knowledge. Experimental results will demonstrate the effectiveness of the proposed approach."
                    },
                    {
                        "title": "Show, Don't Tell: Aligning Language Models with Demonstrated Feedback",
                        "abstract": "Language models are aligned to emulate the collective voice of many, resulting in outputs that align with no one in particular. Steering LLMs away from generic output is possible through supervised finetuning or RLHF, but requires prohibitively large datasets for new ad-hoc tasks. We argue that it is instead possible to align an LLM to a specific setting by leveraging a very small number ($<10$) of demonstrations as feedback. Our method, Demonstration ITerated Task Optimization (DITTO), directly aligns language model outputs to a user's demonstrated behaviors. Derived using ideas from online imitation learning, DITTO cheaply generates online comparison data by treating users' demonstrations as preferred over output from the LLM and its intermediate checkpoints. We evaluate DITTO's ability to learn fine-grained style and task alignment across domains such as news articles, emails, and blog posts. Additionally, we conduct a user study soliciting a range of demonstrations from participants ($N=16$). Across our benchmarks and user study, we find that win-rates for DITTO outperform few-shot prompting, supervised fine-tuning, and other self-play methods by an average of 19% points. By using demonstrations as feedback directly, DITTO offers a novel method for effective customization of LLMs."
                    },
                    {
                        "title": "Into the Unknown Unknowns: Engaged Human Learning through Participation in Language Model Agent Conversations",
                        "abstract": "While language model (LM)-powered chatbots and generative search engines excel at answering concrete queries, discovering information in the terrain of unknown unknowns remains challenging for users. To emulate the common educational scenario where children/students learn by listening to and participating in conversations of their parents/teachers, we create Collaborative STORM (Co-STORM). Unlike QA systems that require users to ask all the questions, Co-STORM lets users observe and occasionally steer the discourse among several LM agents. The agents ask questions on the user's behalf, allowing the user to discover unknown unknowns serendipitously. To facilitate user interaction, Co-STORM assists users in tracking the discourse by organizing the uncovered information into a dynamic mind map, ultimately generating a comprehensive report as takeaways. For automatic evaluation, we construct the WildSeek dataset by collecting real information-seeking records with user goals. Co-STORM outperforms baseline methods on both discourse trace and report quality. In a further human evaluation, 70% of participants prefer Co-STORM over a search engine, and 78% favor it over a RAG chatbot."
                    },
                    {
                        "title": "LUNA: Language Understanding with Number Augmentations on Transformers via Number Plugins and Pre-training",
                        "abstract": "Transformers are widely used in NLP tasks. However, current approaches to leveraging transformers to understand language expose one weak spot: Number understanding. In some scenarios, numbers frequently occur, especially in semi-structured data like tables. But current approaches to rich-number tasks with transformer-based language models abandon or lose some of the numeracy information - e.g., breaking numbers into sub-word tokens - which leads to many number-related errors. In this paper, we propose the LUNA framework which improves the numerical reasoning and calculation capabilities of transformer-based language models. With the number plugin of NumTok and NumBed, LUNA represents each number as a whole to model input. With number pre-training, including regression loss and model distillation, LUNA bridges the gap between number and vocabulary embeddings. To the best of our knowledge, this is the first work that explicitly injects numeracy capability into language models using Number Plugins. Besides evaluating toy models on toy tasks, we evaluate LUNA on three large-scale transformer models (RoBERTa, BERT, TabBERT) over three different downstream tasks (TATQA, TabFact, CrediTrans), and observe the performances of language models are constantly improved by LUNA. The augmented models also improve the official baseline of TAT-QA (EM: 50.15 -> 59.58) and achieve SOTA performance on CrediTrans (F1 = 86.17)."
                    },
                    {
                        "title": "Continual Pre-training of Language Models",
                        "abstract": "Language models (LMs) have been instrumental for the rapid advance of natural language processing. This paper studies continual pre-training of LMs, in particular, continual domain-adaptive pre-training (or continual DAP-training). Existing research has shown that further pre-training an LM using a domain corpus to adapt the LM to the domain can improve the end-task performance in the domain. This paper proposes a novel method to continually DAP-train an LM with a sequence of unlabeled domain corpora to adapt the LM to these domains to improve their end-task performances. The key novelty of our method is a soft-masking mechanism that directly controls the update to the LM. A novel proxy is also proposed to preserve the general knowledge in the original LM. Additionally, it contrasts the representations of the previously learned domain knowledge (including the general knowledge in the pre-trained LM) and the knowledge from the current full network to achieve knowledge integration. The method not only overcomes catastrophic forgetting, but also achieves knowledge transfer to improve end-task performances. Empirical evaluation demonstrates the effectiveness of the proposed method."
                    },
                    {
                        "title": "Class Incremental Learning via Likelihood Ratio Based Task Prediction",
                        "abstract": "Class incremental learning (CIL) is a challenging setting of continual learning, which learns a series of tasks sequentially. Each task consists of a set of unique classes. The key feature of CIL is that no task identifier (or task-id) is provided at test time. Predicting the task-id for each test sample is a challenging problem. An emerging theory-guided approach (called TIL+OOD) is to train a task-specific model for each task in a shared network for all tasks based on a task-incremental learning (TIL) method to deal with catastrophic forgetting. The model for each task is an out-of-distribution (OOD) detector rather than a conventional classifier. The OOD detector can perform both within-task (in-distribution (IND)) class prediction and OOD detection. The OOD detection capability is the key to task-id prediction during inference. However, this paper argues that using a traditional OOD detector for task-id prediction is sub-optimal because additional information (e.g., the replay data and the learned tasks) available in CIL can be exploited to design a better and principled method for task-id prediction. We call the new method TPL (Task-id Prediction based on Likelihood Ratio). TPL markedly outperforms strong CIL baselines and has negligible catastrophic forgetting. The code of TPL is publicly available at https://github.com/linhaowei1/TPL."
                    },
                    {
                        "title": "Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking",
                        "abstract": "When writing and talking, people sometimes pause to think. Although reasoning-focused works have often framed reasoning as a method of answering questions or completing agentic tasks, reasoning is implicit in almost all written text. For example, this applies to the steps not stated between the lines of a proof or to the theory of mind underlying a conversation. In the Self-Taught Reasoner (STaR, Zelikman et al. 2022), useful thinking is learned by inferring rationales from few-shot examples in question-answering and learning from those that lead to a correct answer. This is a highly constrained setting -- ideally, a language model could instead learn to infer unstated rationales in arbitrary text. We present Quiet-STaR, a generalization of STaR in which LMs learn to generate rationales at each token to explain future text, improving their predictions. We address key challenges, including 1) the computational cost of generating continuations, 2) the fact that the LM does not initially know how to generate or use internal thoughts, and 3) the need to predict beyond individual next tokens. To resolve these, we propose a tokenwise parallel sampling algorithm, using learnable tokens indicating a thought's start and end, and an extended teacher-forcing technique. Encouragingly, generated rationales disproportionately help model difficult-to-predict tokens and improve the LM's ability to directly answer difficult questions. In particular, after continued pretraining of an LM on a corpus of internet text with Quiet-STaR, we find zero-shot improvements on GSM8K (5.9%$\\rightarrow$10.9%) and CommonsenseQA (36.3%$\\rightarrow$47.2%) and observe a perplexity improvement of difficult tokens in natural text. Crucially, these improvements require no fine-tuning on these tasks. Quiet-STaR marks a step towards LMs that can learn to reason in a more general and scalable way."
                    },
                    {
                        "title": "FormLM: Recommending Creation Ideas for Online Forms by Modelling Semantic and Structural Information",
                        "abstract": "Online forms are widely used to collect data from human and have a multi-billion market. Many software products provide online services for creating semi-structured forms where questions and descriptions are organized by pre-defined structures. However, the design and creation process of forms is still tedious and requires expert knowledge. To assist form designers, in this work we present FormLM to model online forms (by enhancing pre-trained language model with form structural information) and recommend form creation ideas (including question / options recommendations and block type suggestion). For model training and evaluation, we collect the first public online form dataset with 62K online forms. Experiment results show that FormLM significantly outperforms general-purpose language models on all tasks, with an improvement by 4.71 on Question Recommendation and 10.6 on Block Type Suggestion in terms of ROUGE-1 and Macro-F1, respectively."
                    },
                    {
                        "title": "Assisting in Writing Wikipedia-like Articles From Scratch with Large Language Models",
                        "abstract": "We study how to apply large language models to write grounded and organized long-form articles from scratch, with comparable breadth and depth to Wikipedia pages. This underexplored problem poses new challenges at the pre-writing stage, including how to research the topic and prepare an outline prior to writing. We propose STORM, a writing system for the Synthesis of Topic Outlines through Retrieval and Multi-perspective Question Asking. STORM models the pre-writing stage by (1) discovering diverse perspectives in researching the given topic, (2) simulating conversations where writers carrying different perspectives pose questions to a topic expert grounded on trusted Internet sources, (3) curating the collected information to create an outline.   For evaluation, we curate FreshWiki, a dataset of recent high-quality Wikipedia articles, and formulate outline assessments to evaluate the pre-writing stage. We further gather feedback from experienced Wikipedia editors. Compared to articles generated by an outline-driven retrieval-augmented baseline, more of STORM's articles are deemed to be organized (by a 25% absolute increase) and broad in coverage (by 10%). The expert feedback also helps identify new challenges for generating grounded long articles, such as source bias transfer and over-association of unrelated facts."
                    },
                    {
                        "title": "AnaMeta: A Table Understanding Dataset of Field Metadata Knowledge Shared by Multi-dimensional Data Analysis Tasks",
                        "abstract": "Tabular data analysis is performed every day across various domains. It requires an accurate understanding of field semantics to correctly operate on table fields and find common patterns in daily analysis. In this paper, we introduce the AnaMeta dataset, a collection of 467k tables with derived supervision labels for four types of commonly used field metadata: measure/dimension dichotomy, common field roles, semantic field type, and default aggregation function. We evaluate a wide range of models for inferring metadata as the benchmark. We also propose a multi-encoder framework, called KDF, which improves the metadata understanding capability of tabular models by incorporating distribution and knowledge information. Furthermore, we propose four interfaces for incorporating field metadata into downstream analysis tasks."
                    }
                ]
            },
            "885639d0-5154-4bac-80a9-2e098cbed321": {
                "pk": "885639d0-5154-4bac-80a9-2e098cbed321",
                "name": "Tianshi Li",
                "collaborators": [
                    "Qi Li",
                    "Baobao Chang",
                    "Jijie Zhou",
                    "Eryue Xu",
                    "Yaoyao Wu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Privacy in AI",
                    "Word Embeddings",
                    "Conversational Agents"
                ],
                "publications": [
                    {
                        "title": "Learning Word Sense Embeddings from Word Sense Definitions",
                        "abstract": "Word embeddings play a significant role in many modern NLP systems. Since learning one representation per word is problematic for polysemous words and homonymous words, researchers propose to use one embedding per word sense. Their approaches mainly train word sense embeddings on a corpus. In this paper, we propose to use word sense definitions to learn one embedding per word sense. Experimental results on word similarity tasks and a word sense disambiguation task show that word sense embeddings produced by our approach are of high quality."
                    },
                    {
                        "title": "Rescriber: Smaller-LLM-Powered User-Led Data Minimization for Navigating Privacy Trade-offs in LLM-Based Conversational Agent",
                        "abstract": "The proliferation of LLM-based conversational agents has resulted in excessive disclosure of identifiable or sensitive information. However, existing technologies fail to offer perceptible control or account for users' personal preferences about privacy-utility tradeoffs due to the lack of user involvement. To bridge this gap, we designed, built, and evaluated Rescriber, a browser extension that supports user-led data minimization in LLM-based conversational agents by helping users detect and sanitize personal information in their prompts. Our studies (N=12) showed that Rescriber helped users reduce unnecessary disclosure and addressed their privacy concerns. Users' subjective perceptions of the system powered by Llama3-8B were on par with that by GPT-4. The comprehensiveness and consistency of the detection and sanitization emerge as essential factors that affect users' trust and perceived protection. Our findings confirm the viability of smaller-LLM-powered, user-facing, on-device privacy controls, presenting a promising approach to address the privacy and trust challenges of AI."
                    }
                ]
            },
            "cc825aec-0c51-4c28-9c21-8161125912ef": {
                "pk": "cc825aec-0c51-4c28-9c21-8161125912ef",
                "name": "Weiyan Shi",
                "collaborators": [
                    "Zhou Yu",
                    "Yu Li",
                    "Saurav Sahay",
                    "Kun Qian",
                    "Xuewei Wang",
                    "Ruoxi Jia",
                    "Tiancheng Zhao",
                    "Minzhi Li",
                    "Caleb Ziems",
                    "Diyi Yang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Dialogue Systems",
                    "Reinforcement Learning",
                    "Social Intelligence"
                ],
                "publications": [
                    {
                        "title": "Sentiment Adaptive End-to-End Dialog Systems",
                        "abstract": "End-to-end learning framework is useful for building dialog systems for its simplicity in training and efficiency in model updating. However, current end-to-end approaches only consider user semantic inputs in learning and under-utilize other user information. Therefore, we propose to include user sentiment obtained through multimodal information (acoustic, dialogic and textual), in the end-to-end learning framework to make systems more user-adaptive and effective. We incorporated user sentiment information in both supervised and reinforcement learning settings. In both settings, adding sentiment information reduced the dialog length and improved the task success rate on a bus information search task. This work is the first attempt to incorporate multimodal user information in the adaptive end-to-end dialog system training framework and attained state-of-the-art performance."
                    },
                    {
                        "title": "Unsupervised Dialog Structure Learning",
                        "abstract": "Learning a shared dialog structure from a set of task-oriented dialogs is an important challenge in computational linguistics. The learned dialog structure can shed light on how to analyze human dialogs, and more importantly contribute to the design and evaluation of dialog systems. We propose to extract dialog structures using a modified VRNN model with discrete latent vectors. Different from existing HMM-based models, our model is based on variational-autoencoder (VAE). Such model is able to capture more dynamics in dialogs beyond the surface forms of the language. We find that qualitatively, our method extracts meaningful dialog structure, and quantitatively, outperforms previous models on the ability to predict unseen data. We further evaluate the model's effectiveness in a downstream task, the dialog system building task. Experiments show that, by integrating the learned dialog structure into the reward function design, the model converges faster and to a better outcome in a reinforcement learning setting."
                    },
                    {
                        "title": "Refine and Imitate: Reducing Repetition and Inconsistency in Persuasion Dialogues via Reinforcement Learning and Human Demonstration",
                        "abstract": "Persuasion dialogue systems reflect the machine's ability to make strategic moves beyond verbal communication, and therefore differentiate themselves from task-oriented or open-domain dialogue systems and have their own unique values. However, the repetition and inconsistency problems still persist in dialogue response generation and could substantially impact user experience and impede the persuasion outcome. Besides, although reinforcement learning (RL) approaches have achieved big success in strategic tasks such as games, they require a sophisticated user simulator to provide real-time feedback to the dialogue system, which limits the application of RL on persuasion dialogues. To address these issues towards a better persuasion dialogue system, we apply RL to refine a language model baseline without user simulators, and distill sentence-level information about repetition, inconsistency, and task relevance through rewards. Moreover, to better accomplish the persuasion task, the model learns from human demonstration to imitate human persuasion behavior and selects the most persuasive responses. Experiments show that our model outperforms previous state-of-the-art dialogue models on both automatic metrics and human evaluation results on a donation persuasion task, and generates more diverse, consistent and persuasive conversations according to the user feedback."
                    },
                    {
                        "title": "End-to-End Trainable Non-Collaborative Dialog System",
                        "abstract": "End-to-end task-oriented dialog models have achieved promising performance on collaborative tasks where users willingly coordinate with the system to complete a given task. While in non-collaborative settings, for example, negotiation and persuasion, users and systems do not share a common goal. As a result, compared to collaborate tasks, people use social content to build rapport and trust in these non-collaborative settings in order to advance their goals. To handle social content, we introduce a hierarchical intent annotation scheme, which can be generalized to different non-collaborative dialog tasks. Building upon TransferTransfo (Wolf et al. 2019), we propose an end-to-end neural network model to generate diverse coherent responses. Our model utilizes intent and semantic slots as the intermediate sentence representation to guide the generation process. In addition, we design a filter to select appropriate responses based on whether these intermediate representations fit the designed task and conversation constraints. Our non-collaborative dialog model guides users to complete the task while simultaneously keeps them engaged. We test our approach on our newly proposed ANTISCAM dataset and an existing PERSUASIONFORGOOD dataset. Both automatic and human evaluations suggest that our model outperforms multiple baselines in these two non-collaborative tasks."
                    },
                    {
                        "title": "Social Intelligence Data Infrastructure: Structuring the Present and Navigating the Future",
                        "abstract": "As Natural Language Processing (NLP) systems become increasingly integrated into human social life, these technologies will need to increasingly rely on social intelligence. Although there are many valuable datasets that benchmark isolated dimensions of social intelligence, there does not yet exist any body of work to join these threads into a cohesive subfield in which researchers can quickly identify research gaps and future directions. Towards this goal, we build a Social AI Data Infrastructure, which consists of a comprehensive social AI taxonomy and a data library of 480 NLP datasets. Our infrastructure allows us to analyze existing dataset efforts, and also evaluate language models' performance in different social intelligence aspects. Our analyses demonstrate its utility in enabling a thorough understanding of current data landscape and providing a holistic perspective on potential directions for future dataset development. We show there is a need for multifaceted datasets, increased diversity in language and culture, more long-tailed social situations, and more interactive data in future social intelligence data efforts."
                    },
                    {
                        "title": "How to Build User Simulators to Train RL-based Dialog Systems",
                        "abstract": "User simulators are essential for training reinforcement learning (RL) based dialog models. The performance of the simulator directly impacts the RL policy. However, building a good user simulator that models real user behaviors is challenging. We propose a method of standardizing user simulator building that can be used by the community to compare dialog system quality using the same set of user simulators fairly. We present implementations of six user simulators trained with different dialog planning and generation methods. We then calculate a set of automatic metrics to evaluate the quality of these simulators both directly and indirectly. We also ask human users to assess the simulators directly and indirectly by rating the simulated dialogs and interacting with the trained systems. This paper presents a comprehensive evaluation framework for user simulator study and provides a better understanding of the pros and cons of different user simulators, as well as their impacts on the trained systems."
                    },
                    {
                        "title": "When Life Gives You Lemons, Make Cherryade: Converting Feedback from Bad Responses into Good Labels",
                        "abstract": "Deployed dialogue agents have the potential to integrate human feedback to continuously improve themselves. However, humans may not always provide explicit signals when the chatbot makes mistakes during interactions. In this work, we propose Juicer, a framework to make use of both binary and free-form textual human feedback. It works by: (i) extending sparse binary feedback by training a satisfaction classifier to label the unlabeled data; and (ii) training a reply corrector to map the bad replies to good ones. We find that augmenting training with model-corrected replies improves the final dialogue model, and we can further improve performance by using both positive and negative replies through the recently proposed Director model."
                    },
                    {
                        "title": "Selective Differential Privacy for Language Modeling",
                        "abstract": "With the increasing applications of language models, it has become crucial to protect these models from leaking private information. Previous work has attempted to tackle this challenge by training RNN-based language models with differential privacy guarantees. However, applying classical differential privacy to language models leads to poor model performance as the underlying privacy notion is over-pessimistic and provides undifferentiated protection for all tokens in the data. Given that the private information in natural language is sparse (for example, the bulk of an email might not carry personally identifiable information), we propose a new privacy notion, selective differential privacy, to provide rigorous privacy guarantees on the sensitive portion of the data to improve model utility. To realize such a new notion, we develop a corresponding privacy mechanism, Selective-DPSGD, for RNN-based language models. Besides language modeling, we also apply the method to a more concrete application--dialog systems. Experiments on both language modeling and dialog system building show that the proposed privacy-preserving mechanism achieves better utilities while remaining safe under various privacy attacks compared to the baselines. The data and code are released at https://github.com/wyshi/lm_privacy to facilitate future research ."
                    },
                    {
                        "title": "A Tailored Pre-Training Model for Task-Oriented Dialog Generation",
                        "abstract": "The recent success of large pre-trained language models such as BERT and GPT-2 has suggested the effectiveness of incorporating language priors in downstream dialog generation tasks. However, the performance of pre-trained models on the dialog task is not as optimal as expected. In this paper, we propose a Pre-trained Role Alternating Language model (PRAL), designed specifically for task-oriented conversational systems. We adopted (Wu et al., 2019) that models two speakers separately. We also design several techniques, such as start position randomization, knowledge distillation, and history discount to improve pre-training performance. We introduce a task-oriented dialog pretraining dataset by cleaning 13 existing data sets. We test PRAL on three different downstream tasks. The results show that PRAL performs better or on par with state-of-the-art methods."
                    },
                    {
                        "title": "Controllable Mixed-Initiative Dialogue Generation through Prompting",
                        "abstract": "Mixed-initiative dialogue tasks involve repeated exchanges of information and conversational control. Conversational agents gain control by generating responses that follow particular dialogue intents or strategies, prescribed by a policy planner. The standard approach has been fine-tuning pre-trained language models to perform generation conditioned on these intents. However, these supervised generation models are limited by the cost and quality of data annotation. We instead prompt large language models as a drop-in replacement to fine-tuning on conditional generation. We formalize prompt construction for controllable mixed-initiative dialogue. Our findings show improvements over fine-tuning and ground truth responses according to human evaluation and automatic metrics for two tasks: PersuasionForGood and Emotional Support Conversations."
                    },
                    {
                        "title": "Just Fine-tune Twice: Selective Differential Privacy for Large Language Models",
                        "abstract": "Protecting large language models from privacy leakage is becoming increasingly crucial with their wide adoption in real-world products. Yet applying differential privacy (DP), a canonical notion with provable privacy guarantees for machine learning models, to those models remains challenging due to the trade-off between model utility and privacy loss. Utilizing the fact that sensitive information in language data tends to be sparse, Shi et al. (2021) formalized a DP notion extension called Selective Differential Privacy (SDP) to protect only the sensitive tokens defined by a policy function. However, their algorithm only works for RNN-based models. In this paper, we develop a novel framework, Just Fine-tune Twice (JFT), that achieves SDP for state-of-the-art large transformer-based models. Our method is easy to implement: it first fine-tunes the model with redacted in-domain data, and then fine-tunes it again with the original in-domain data using a private training mechanism. Furthermore, we study the scenario of imperfect implementation of policy functions that misses sensitive tokens and develop systematic methods to handle it. Experiments show that our method achieves strong utility compared to previous baselines. We also analyze the SDP privacy guarantee empirically with the canary insertion attack."
                    },
                    {
                        "title": "Effects of Persuasive Dialogues: Testing Bot Identities and Inquiry Strategies",
                        "abstract": "Intelligent conversational agents, or chatbots, can take on various identities and are increasingly engaging in more human-centered conversations with persuasive goals. However, little is known about how identities and inquiry strategies influence the conversation's effectiveness. We conducted an online study involving 790 participants to be persuaded by a chatbot for charity donation. We designed a two by four factorial experiment (two chatbot identities and four inquiry strategies) where participants were randomly assigned to different conditions. Findings showed that the perceived identity of the chatbot had significant effects on the persuasion outcome (i.e., donation) and interpersonal perceptions (i.e., competence, confidence, warmth, and sincerity). Further, we identified interaction effects among perceived identities and inquiry strategies. We discuss the findings for theoretical and practical implications for developing ethical and effective persuasive chatbots. Our published data, codes, and analyses serve as the first step towards building competent ethical persuasive chatbots."
                    },
                    {
                        "title": "LEGOEval: An Open-Source Toolkit for Dialogue System Evaluation via Crowdsourcing",
                        "abstract": "We present LEGOEval, an open-source toolkit that enables researchers to easily evaluate dialogue systems in a few lines of code using the online crowdsource platform, Amazon Mechanical Turk. Compared to existing toolkits, LEGOEval features a flexible task design by providing a Python API that maps to commonly used React.js interface components. Researchers can personalize their evaluation procedures easily with our built-in pages as if playing with LEGO blocks. Thus, LEGOEval provides a fast, consistent method for reproducing human evaluation results. Besides the flexible task design, LEGOEval also offers an easy API to review collected data."
                    }
                ]
            },
            "d3106e2f-c4d4-43e5-b55a-05f55624c512": {
                "pk": "d3106e2f-c4d4-43e5-b55a-05f55624c512",
                "name": "Yanchen Liu",
                "collaborators": [
                    "Diyi Yang",
                    "William Held",
                    "Yanan Sun",
                    "Yinhao Tang",
                    "Kai Chen",
                    "Srishti Gautam",
                    "Jiaqi Ma",
                    "Himabindu Lakkaraju",
                    "Qi Wu",
                    "Chong Zhang"
                ],
                "domain": [
                    "Network Science",
                    "Natural Language Processing",
                    "Reinforcement Learning",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Approximate Network Symmetry",
                        "abstract": "We define a new measure of network symmetry that is capable of capturing approximate global symmetries of networks. We apply this measure to different networks sampled from several classic network models, as well as several real-world networks. We find that among the network models that we have examined, Erd\\\"os-R\\'enyi networks have the least levels of symmetry, and Random Geometric Graphs are likely to have high levels of symmetry. We find that our network symmetry measure can capture properties of network structure, and help us gain insights on the structure of real-world networks. Moreover, our network symmetry measure is capable of capturing imperfect network symmetry, which would have been undetected if only perfect symmetry is considered."
                    },
                    {
                        "title": "DADA: Dialect Adaptation via Dynamic Aggregation of Linguistic Rules",
                        "abstract": "Existing large language models (LLMs) that mainly focus on Standard American English (SAE) often lead to significantly worse performance when being applied to other English dialects. While existing mitigations tackle discrepancies for individual target dialects, they assume access to high-accuracy dialect identification systems. The boundaries between dialects are inherently flexible, making it difficult to categorize language into discrete predefined categories. In this paper, we propose DADA (Dialect Adaptation via Dynamic Aggregation), a modular approach to imbue SAE-trained models with multi-dialectal robustness by composing adapters which handle specific linguistic features. The compositional architecture of DADA allows for both targeted adaptation to specific dialect variants and simultaneous adaptation to various dialects. We show that DADA is effective for both single task and instruction finetuned language models, offering an extensible and interpretable framework for adapting existing LLMs to different English dialects."
                    },
                    {
                        "title": "Confronting LLMs with Traditional ML: Rethinking the Fairness of Large Language Models in Tabular Classifications",
                        "abstract": "Recent literature has suggested the potential of using large language models (LLMs) to make classifications for tabular tasks. However, LLMs have been shown to exhibit harmful social biases that reflect the stereotypes and inequalities present in society. To this end, as well as the widespread use of tabular data in many high-stake applications, it is important to explore the following questions: what sources of information do LLMs draw upon when making classifications for tabular tasks; whether and to what extent are LLM classifications for tabular data influenced by social biases and stereotypes; and what are the consequential implications for fairness?   Through a series of experiments, we delve into these questions and show that LLMs tend to inherit social biases from their training data which significantly impact their fairness in tabular classification tasks. Furthermore, our investigations show that in the context of bias mitigation, though in-context learning and finetuning have a moderate effect, the fairness metric gap between different subgroups is still larger than that in traditional machine learning models, such as Random Forest and shallow Neural Networks. This observation emphasizes that the social biases are inherent within the LLMs themselves and inherited from their pretraining corpus, not only from the downstream task datasets. Besides, we demonstrate that label-flipping of in-context examples can significantly reduce biases, further highlighting the presence of inherent bias within LLMs."
                    },
                    {
                        "title": "Custom Sine Waves Are Enough for Imitation Learning of Bipedal Gaits with Different Styles",
                        "abstract": "Not until recently, robust bipedal locomotion has been achieved through reinforcement learning. However, existing implementations rely heavily on insights and efforts from human experts, which is costly for the iterative design of robot systems. Also, styles of the learned motion are strictly limited to that of the reference. In this paper, we propose a new way to learn bipedal locomotion from a simple sine wave as the reference for foot heights. With the naive human insight that the two feet should be lifted up alternatively and periodically, we experimentally demonstrate on the Cassie robot that, a simple reward function is able to make the robot learn to walk end-to-end and efficiently without any explicit knowledge of the model. With custom sine waves, the learned gait pattern can also have customized styles. Codes are released at github.com/WooQi57/sin-cassie-rl."
                    },
                    {
                        "title": "Semantic-Oriented Unlabeled Priming for Large-Scale Language Models",
                        "abstract": "Due to the high costs associated with finetuning large language models, various recent works propose to adapt them to specific tasks without any parameter updates through in-context learning. Unfortunately, for in-context learning there is currently no way to leverage unlabeled data, which is often much easier to obtain in large quantities than labeled examples. In this work, we therefore investigate ways to make use of unlabeled examples to improve the zero-shot performance of pretrained language models without any finetuning: We introduce Semantic-Oriented Unlabeled Priming (SOUP), a method that classifies examples by retrieving semantically similar unlabeled examples, assigning labels to them in a zero-shot fashion, and then using them for in-context learning. We also propose bag-of-contexts priming, a new priming strategy that is more suitable for our setting and enables the usage of more examples than fit into the context window."
                    },
                    {
                        "title": "Weighted Community Detection and Data Clustering Using Message Passing",
                        "abstract": "Grouping objects into clusters based on similarities or weights between them is one of the most important problems in science and engineering. In this work, by extending message passing algorithms and spectral algorithms proposed for unweighted community detection problem, we develop a non-parametric method based on statistical physics, by mapping the problem to Potts model at the critical temperature of spin glass transition and applying belief propagation to solve the marginals corresponding to the Boltzmann distribution. Our algorithm is robust to over-fitting and gives a principled way to determine whether there are significant clusters in the data and how many clusters there are. We apply our method to different clustering tasks and use extensive numerical experiments to illustrate the advantage of our method over existing algorithms. In the community detection problem in weighted and directed networks, we show that our algorithm significantly outperforms existing algorithms. In the clustering problem when the data was generated by mixture models in the sparse regime we show that our method works to the theoretical limit of detectability and gives accuracy very close to that of the optimal Bayesian inference. In the semi-supervised clustering problem, our method only needs several labels to work perfectly in classic datasets. Finally, we further develop Thouless-Anderson-Palmer equations which reduce heavily the computation complexity in dense-networks but gives almost the same performance as belief propagation."
                    },
                    {
                        "title": "SISSA: Real-time Monitoring of Hardware Functional Safety and Cybersecurity with In-vehicle SOME/IP Ethernet Traffic",
                        "abstract": "Scalable service-Oriented Middleware over IP (SOME/IP) is an Ethernet communication standard protocol in the Automotive Open System Architecture (AUTOSAR), promoting ECU-to-ECU communication over the IP stack. However, SOME/IP lacks a robust security architecture, making it susceptible to potential attacks. Besides, random hardware failure of ECU will disrupt SOME/IP communication. In this paper, we propose SISSA, a SOME/IP communication traffic-based approach for modeling and analyzing in-vehicle functional safety and cyber security. Specifically, SISSA models hardware failures with the Weibull distribution and addresses five potential attacks on SOME/IP communication, including Distributed Denial-of-Services, Man-in-the-Middle, and abnormal communication processes, assuming a malicious user accesses the in-vehicle network. Subsequently, SISSA designs a series of deep learning models with various backbones to extract features from SOME/IP sessions among ECUs. We adopt residual self-attention to accelerate the model's convergence and enhance detection accuracy, determining whether an ECU is under attack, facing functional failure, or operating normally. Additionally, we have created and annotated a dataset encompassing various classes, including indicators of attack, functionality, and normalcy. This contribution is noteworthy due to the scarcity of publicly accessible datasets with such characteristics.Extensive experimental results show the effectiveness and efficiency of SISSA."
                    },
                    {
                        "title": "SMoA: Sparse Mixture of Adapters to Mitigate Multiple Dataset Biases",
                        "abstract": "Recent studies reveal that various biases exist in different NLP tasks, and over-reliance on biases results in models' poor generalization ability and low adversarial robustness. To mitigate datasets biases, previous works propose lots of debiasing techniques to tackle specific biases, which perform well on respective adversarial sets but fail to mitigate other biases. In this paper, we propose a new debiasing method Sparse Mixture-of-Adapters (SMoA), which can mitigate multiple dataset biases effectively and efficiently. Experiments on Natural Language Inference and Paraphrase Identification tasks demonstrate that SMoA outperforms full-finetuning, adapter tuning baselines, and prior strong debiasing methods. Further analysis indicates the interpretability of SMoA that sub-adapter can capture specific pattern from the training data and specialize to handle specific bias."
                    },
                    {
                        "title": "Automatic Symmetry Discovery with Lie Algebra Convolutional Network",
                        "abstract": "Existing equivariant neural networks require prior knowledge of the symmetry group and discretization for continuous groups. We propose to work with Lie algebras (infinitesimal generators) instead of Lie groups. Our model, the Lie algebra convolutional network (L-conv) can automatically discover symmetries and does not require discretization of the group. We show that L-conv can serve as a building block to construct any group equivariant feedforward architecture. Both CNNs and Graph Convolutional Networks can be expressed as L-conv with appropriate groups. We discover direct connections between L-conv and physics: (1) group invariant loss generalizes field theory (2) Euler-Lagrange equation measures the robustness, and (3) equivariance leads to conservation laws and Noether current.These connections open up new avenues for designing more general equivariant networks and applying them to important problems in physical sciences"
                    },
                    {
                        "title": "Task-Agnostic Low-Rank Adapters for Unseen English Dialects",
                        "abstract": "Large Language Models (LLMs) are trained on corpora disproportionally weighted in favor of Standard American English. As a result, speakers of other dialects experience significantly more failures when interacting with these technologies. In practice, these speakers often accommodate their speech to be better understood. Our work shares the belief that language technologies should be designed to accommodate the diversity in English dialects and not the other way around. However, prior works on dialect struggle with generalizing to evolving and emerging dialects in a scalable manner. To fill this gap, our method, HyperLoRA, leverages expert linguistic knowledge to enable resource-efficient adaptation via hypernetworks. By disentangling dialect-specific and cross-dialectal information, HyperLoRA improves generalization to unseen dialects in a task-agnostic fashion. Not only is HyperLoRA more scalable in the number of parameters, but it also achieves the best or most competitive performance across 5 dialects in a zero-shot setting. In this way, our approach facilitates access to language technology for billions of English dialect speakers who are traditionally underrepresented."
                    },
                    {
                        "title": "RASP: A Drone-based Reconfigurable Actuation and Sensing Platform for Engaging Physical Environments with Foundation Models",
                        "abstract": "Foundation models and large language models have shown immense human-like understanding and capabilities for generating text and digital media. However, foundation models that can freely sense, interact, and actuate the physical world like in the digital domain is far from being realized. This is due to a number of challenges including: 1) being constrained to the types of static devices and sensors deployed, 2) events often being localized to one part of a large space, and 3) requiring dense and deployments of devices to achieve full coverage. As a critical step towards enabling foundation models to successfully and freely interact with the physical environment, we propose RASP, a modular and reconfigurable sensing and actuation platform that allows drones to autonomously swap onboard sensors and actuators in only $25$ seconds, allowing a single drone to quickly adapt to a diverse range of tasks. We demonstrate through real smart home deployments that RASP enables FMs and LLMs to complete diverse tasks up to $85\\%$ more successfully by allowing them to target specific areas with specific sensors and actuators on-the-fly."
                    },
                    {
                        "title": "StyleShot: A Snapshot on Any Style",
                        "abstract": "In this paper, we show that, a good style representation is crucial and sufficient for generalized style transfer without test-time tuning. We achieve this through constructing a style-aware encoder and a well-organized style dataset called StyleGallery. With dedicated design for style learning, this style-aware encoder is trained to extract expressive style representation with decoupling training strategy, and StyleGallery enables the generalization ability. We further employ a content-fusion encoder to enhance image-driven style transfer. We highlight that, our approach, named StyleShot, is simple yet effective in mimicking various desired styles, i.e., 3D, flat, abstract or even fine-grained styles, without test-time tuning. Rigorous experiments validate that, StyleShot achieves superior performance across a wide range of styles compared to existing state-of-the-art methods. The project page is available at: https://styleshot.github.io/."
                    },
                    {
                        "title": "AnyControl: Create Your Artwork with Versatile Control on Text-to-Image Generation",
                        "abstract": "The field of text-to-image (T2I) generation has made significant progress in recent years, largely driven by advancements in diffusion models. Linguistic control enables effective content creation, but struggles with fine-grained control over image generation. This challenge has been explored, to a great extent, by incorporating additional user-supplied spatial conditions, such as depth maps and edge maps, into pre-trained T2I models through extra encoding. However, multi-control image synthesis still faces several challenges. Specifically, current approaches are limited in handling free combinations of diverse input control signals, overlook the complex relationships among multiple spatial conditions, and often fail to maintain semantic alignment with provided textual prompts. This can lead to suboptimal user experiences. To address these challenges, we propose AnyControl, a multi-control image synthesis framework that supports arbitrary combinations of diverse control signals. AnyControl develops a novel Multi-Control Encoder that extracts a unified multi-modal embedding to guide the generation process. This approach enables a holistic understanding of user inputs, and produces high-quality, faithful results under versatile control signals, as demonstrated by extensive quantitative and qualitative evaluations. Our project page is available in https://any-control.github.io."
                    },
                    {
                        "title": "Decoding Susceptibility: Modeling Misbelief to Misinformation Through a Computational Approach",
                        "abstract": "Susceptibility to misinformation describes the degree of belief in unverifiable claims, a latent aspect of individuals' mental processes that is not observable. Existing susceptibility studies heavily rely on self-reported beliefs, which can be subject to bias, expensive to collect, and challenging to scale for downstream applications. To address these limitations, in this work, we propose a computational approach to model users' latent susceptibility levels. As shown in previous research, susceptibility is influenced by various factors (e.g., demographic factors, political ideology), and directly influences people's reposting behavior on social media. To represent the underlying mental process, our susceptibility modeling incorporates these factors as inputs, guided by the supervision of people's sharing behavior. Using COVID-19 as a testbed domain, our experiments demonstrate a significant alignment between the susceptibility scores estimated by our computational modeling and human judgments, confirming the effectiveness of this latent modeling approach. Furthermore, we apply our model to annotate susceptibility scores on a large-scale dataset and analyze the relationships between susceptibility with various factors. Our analysis reveals that political leanings and psychological factors exhibit varying degrees of association with susceptibility to COVID-19 misinformation."
                    },
                    {
                        "title": "Thresholding normally distributed data creates complex networks",
                        "abstract": "Network data sets are often constructed by some kind of thresholding procedure. The resulting networks frequently possess properties such as heavy-tailed degree distributions, clustering, large connected components and short average shortest path lengths. These properties are considered typical of complex networks and appear in many contexts, prompting consideration of their universality. Here we introduce a simple model for correlated relational data and study the network ensemble obtained by thresholding it. We find that some, but not all, of the properties associated with complex networks can be seen after thresholding the correlated data, even though the underlying data are not \"complex\". In particular, we observe heavy-tailed degree distributions, a large numbers of triangles, and short path lengths, while we do not observe non-vanishing clustering or community structure."
                    }
                ]
            },
            "6ef358bf-2a2c-4a12-9417-1c4be1d069ff": {
                "pk": "6ef358bf-2a2c-4a12-9417-1c4be1d069ff",
                "name": "Diyi Yang",
                "collaborators": [
                    "Jiaao Chen",
                    "Caleb Ziems",
                    "Xuezhi Wang",
                    "Le Zhang",
                    "Zichao Yang",
                    "Kunal Chawla",
                    "William Held",
                    "Jingfeng Yang",
                    "Dirk Hovy",
                    "David Jurgens"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Text Summarization",
                    "Style Transfer",
                    "Robustness"
                ],
                "publications": [
                    {
                        "title": "Multi-View Sequence-to-Sequence Models with Conversational Structure for Abstractive Dialogue Summarization",
                        "abstract": "Text summarization is one of the most challenging and interesting problems in NLP. Although much attention has been paid to summarizing structured text like news reports or encyclopedia articles, summarizing conversations---an essential part of human-human/machine interaction where most important pieces of information are scattered across various utterances of different speakers---remains relatively under-investigated. This work proposes a multi-view sequence-to-sequence model by first extracting conversational structures of unstructured daily chats from different views to represent conversations and then utilizing a multi-view decoder to incorporate different views to generate dialogue summaries. Experiments on a large-scale dialogue summarization corpus demonstrated that our methods significantly outperformed previous state-of-the-art models via both automatic evaluations and human judgment. We also discussed specific challenges that current approaches faced with this task. We have publicly released our code at https://github.com/GT-SALT/Multi-View-Seq2Seq."
                    },
                    {
                        "title": "Semi-supervised Formality Style Transfer using Language Model Discriminator and Mutual Information Maximization",
                        "abstract": "Formality style transfer is the task of converting informal sentences to grammatically-correct formal sentences, which can be used to improve performance of many downstream NLP tasks. In this work, we propose a semi-supervised formality style transfer model that utilizes a language model-based discriminator to maximize the likelihood of the output sentence being formal, which allows us to use maximization of token-level conditional probabilities for training. We further propose to maximize mutual information between source and target styles as our training objective instead of maximizing the regular likelihood that often leads to repetitive and trivial generated responses. Experiments showed that our model outperformed previous state-of-the-art baselines significantly in terms of both automated metrics and human judgement. We further generalized our model to unsupervised text style transfer task, and achieved significant improvements on two benchmark sentiment style transfer datasets."
                    },
                    {
                        "title": "Weakly-Supervised Hierarchical Models for Predicting Persuasive Strategies in Good-faith Textual Requests",
                        "abstract": "Modeling persuasive language has the potential to better facilitate our decision-making processes. Despite its importance, computational modeling of persuasion is still in its infancy, largely due to the lack of benchmark datasets that can provide quantitative labels of persuasive strategies to expedite this line of research. To this end, we introduce a large-scale multi-domain text corpus for modeling persuasive strategies in good-faith text requests. Moreover, we design a hierarchical weakly-supervised latent variable model that can leverage partially labeled data to predict such associated persuasive strategies for each sentence, where the supervision comes from both the overall document-level labels and very limited sentence-level labels. Experimental results showed that our proposed method outperformed existing semi-supervised baselines significantly. We have publicly released our code at https://github.com/GT-SALT/Persuasion_Strategy_WVAE."
                    },
                    {
                        "title": "Structure-Aware Abstractive Conversation Summarization via Discourse and Action Graphs",
                        "abstract": "Abstractive conversation summarization has received much attention recently. However, these generated summaries often suffer from insufficient, redundant, or incorrect content, largely due to the unstructured and complex characteristics of human-human interactions. To this end, we propose to explicitly model the rich structures in conversations for more precise and accurate conversation summarization, by first incorporating discourse relations between utterances and action triples (\"who-doing-what\") in utterances through structured graphs to better encode conversations, and then designing a multi-granularity decoder to generate summaries by combining all levels of information. Experiments show that our proposed models outperform state-of-the-art methods and generalize well in other domains in terms of both automatic evaluations and human judgments. We have publicly released our code at https://github.com/GT-SALT/Structure-Aware-BART."
                    },
                    {
                        "title": "To Protect and To Serve? Analyzing Entity-Centric Framing of Police Violence",
                        "abstract": "Framing has significant but subtle effects on public opinion and policy. We propose an NLP framework to measure entity-centric frames. We use it to understand media coverage on police violence in the United States in a new Police Violence Frames Corpus of 82k news articles spanning 7k police killings. Our work uncovers more than a dozen framing devices and reveals significant differences in the way liberal and conservative news sources frame both the issue of police violence and the entities involved. Conservative sources emphasize when the victim is armed or attacking an officer and are more likely to mention the victim's criminal record. Liberal sources focus more on the underlying systemic injustice, highlighting the victim's race and that they were unarmed. We discover temporary spikes in these injustice frames near high-profile shooting events, and finally, we show protest volume correlates with and precedes media framing decisions."
                    },
                    {
                        "title": "Shapley Head Pruning: Identifying and Removing Interference in Multilingual Transformers",
                        "abstract": "Multilingual transformer-based models demonstrate remarkable zero and few-shot transfer across languages by learning and reusing language-agnostic features. However, as a fixed-size model acquires more languages, its performance across all languages degrades, a phenomenon termed interference. Often attributed to limited model capacity, interference is commonly addressed by adding additional parameters despite evidence that transformer-based models are overparameterized. In this work, we show that it is possible to reduce interference by instead identifying and pruning language-specific parameters. First, we use Shapley Values, a credit allocation metric from coalitional game theory, to identify attention heads that introduce interference. Then, we show that removing identified attention heads from a fixed model improves performance for a target language on both sentence classification and structural prediction, seeing gains as large as 24.7\\%. Finally, we provide insights on language-agnostic and language-specific attention heads using attention visualization."
                    },
                    {
                        "title": "Unlearn What You Want to Forget: Efficient Unlearning for LLMs",
                        "abstract": "Large language models (LLMs) have achieved significant progress from pre-training on and memorizing a wide range of textual data, however, this process might suffer from privacy issues and violations of data protection regulations. As a result, the ability to easily remove data related to individual users from such models while not deteriorating their predictive quality after the removal becomes increasingly important. To address these issues, in this work, we propose an efficient unlearning framework that could efficiently update LLMs without having to retrain the whole model after data removals, by introducing lightweight unlearning layers learned with a selective teacher-student objective into the transformers. In addition, we introduce a fusion mechanism to effectively combine different unlearning layers that learns to forget different sets of data to handle a sequence of forgetting operations. Experiments on classification and generation tasks demonstrate the effectiveness of our proposed methods compared to the state-of-the-art baselines."
                    },
                    {
                        "title": "SUBS: Subtree Substitution for Compositional Semantic Parsing",
                        "abstract": "Although sequence-to-sequence models often achieve good performance in semantic parsing for i.i.d. data, their performance is still inferior in compositional generalization. Several data augmentation methods have been proposed to alleviate this problem. However, prior work only leveraged superficial grammar or rules for data augmentation, which resulted in limited improvement. We propose to use subtree substitution for compositional data augmentation, where we consider subtrees with similar semantic functions as exchangeable. Our experiments showed that such augmented data led to significantly better performance on SCAN and GeoQuery, and reached new SOTA on compositional split of GeoQuery."
                    },
                    {
                        "title": "MixText: Linguistically-Informed Interpolation of Hidden Space for Semi-Supervised Text Classification",
                        "abstract": "This paper presents MixText, a semi-supervised learning method for text classification, which uses our newly designed data augmentation method called TMix. TMix creates a large amount of augmented training samples by interpolating text in hidden space. Moreover, we leverage recent advances in data augmentation to guess low-entropy labels for unlabeled data, hence making them as easy to use as labeled data.By mixing labeled, unlabeled and augmented data, MixText significantly outperformed current pre-trained and fined-tuned models and other state-of-the-art semi-supervised learning methods on several text classification benchmarks. The improvement is especially prominent when supervision is extremely limited. We have publicly released our code at https://github.com/GT-SALT/MixText."
                    },
                    {
                        "title": "TreeMix: Compositional Constituency-based Data Augmentation for Natural Language Understanding",
                        "abstract": "Data augmentation is an effective approach to tackle over-fitting. Many previous works have proposed different data augmentations strategies for NLP, such as noise injection, word replacement, back-translation etc. Though effective, they missed one important characteristic of language--compositionality, meaning of a complex expression is built from its sub-parts. Motivated by this, we propose a compositional data augmentation approach for natural language understanding called TreeMix. Specifically, TreeMix leverages constituency parsing tree to decompose sentences into constituent sub-structures and the Mixup data augmentation technique to recombine them to generate new sentences. Compared with previous approaches, TreeMix introduces greater diversity to the samples generated and encourages models to learn compositionality of NLP data. Extensive experiments on text classification and SCAN demonstrate that TreeMix outperforms current state-of-the-art data augmentation methods."
                    },
                    {
                        "title": "The Call for Socially Aware Language Technologies",
                        "abstract": "Language technologies have made enormous progress, especially with the introduction of large language models (LLMs). On traditional tasks such as machine translation and sentiment analysis, these models perform at near-human level. These advances can, however, exacerbate a variety of issues that models have traditionally struggled with, such as bias, evaluation, and risks. In this position paper, we argue that many of these issues share a common core: a lack of awareness of the factors, context, and implications of the social environment in which NLP operates, which we call social awareness. While NLP is getting better at solving the formal linguistic aspects, limited progress has been made in adding the social awareness required for language applications to work in all situations for all users. Integrating social awareness into NLP models will make applications more natural, helpful, and safe, and will open up new possibilities. Thus we argue that substantial challenges remain for NLP to develop social awareness and that we are just at the beginning of a new era for the field."
                    },
                    {
                        "title": "TADA: Task-Agnostic Dialect Adapters for English",
                        "abstract": "Large Language Models, the dominant starting point for Natural Language Processing (NLP) applications, fail at a higher rate for speakers of English dialects other than Standard American English (SAE). Prior work addresses this using task-specific data or synthetic data augmentation, both of which require intervention for each dialect and task pair. This poses a scalability issue that prevents the broad adoption of robust dialectal English NLP. We introduce a simple yet effective method for task-agnostic dialect adaptation by aligning non-SAE dialects using adapters and composing them with task-specific adapters from SAE. Task-Agnostic Dialect Adapters (TADA) improve dialectal robustness on 4 dialectal variants of the GLUE benchmark without task-specific supervision."
                    },
                    {
                        "title": "Identifying and Mitigating Spurious Correlations for Improving Robustness in NLP Models",
                        "abstract": "Recently, NLP models have achieved remarkable progress across a variety of tasks; however, they have also been criticized for being not robust. Many robustness problems can be attributed to models exploiting spurious correlations, or shortcuts between the training data and the task labels. Most existing work identifies a limited set of task-specific shortcuts via human priors or error analyses, which requires extensive expertise and efforts. In this paper, we aim to automatically identify such spurious correlations in NLP models at scale. We first leverage existing interpretability methods to extract tokens that significantly affect model's decision process from the input text. We then distinguish \"genuine\" tokens and \"spurious\" tokens by analyzing model predictions across multiple corpora and further verify them through knowledge-aware perturbations. We show that our proposed method can effectively and efficiently identify a scalable set of \"shortcuts\", and mitigating these leads to more robust models in multiple applications."
                    },
                    {
                        "title": "Inducing Positive Perspectives with Text Reframing",
                        "abstract": "Sentiment transfer is one popular example of a text style transfer task, where the goal is to reverse the sentiment polarity of a text. With a sentiment reversal comes also a reversal in meaning. We introduce a different but related task called positive reframing in which we neutralize a negative point of view and generate a more positive perspective for the author without contradicting the original meaning. Our insistence on meaning preservation makes positive reframing a challenging and semantically rich task. To facilitate rapid progress, we introduce a large-scale benchmark, Positive Psychology Frames, with 8,349 sentence pairs and 12,755 structured annotations to explain positive reframing in terms of six theoretically-motivated reframing strategies. Then we evaluate a set of state-of-the-art text style transfer models, and conclude by discussing key challenges and directions for future work."
                    },
                    {
                        "title": "Six Feet Apart: Online Payments During the COVID-19 Pandemic",
                        "abstract": "Since the COVID-19 pandemic, businesses have faced unprecedented challenges when trying to remain open. Because COVID-19 spreads through aerosolized droplets, businesses were forced to distance their services; in some cases, distancing may have involved moving business services online. In this work, we explore digitization strategies used by small businesses that remained open during the pandemic, and survey/interview small businesses owners to understand preliminary challenges associated with moving online. Furthermore, we analyze payments from 400K businesses across Japan, Australia, United States, Great Britain, and Canada. Following initial government interventions, we observe (at minimum for each country) a 47% increase in digitizing businesses compared to pre-pandemic levels, with about 80% of surveyed businesses digitizing in under a week. From both our quantitative models and our surveys/interviews, we find that businesses rapidly digitized at the start of the pandemic in preparation of future uncertainty. We also conduct a case-study of initial digitization in the United States, examining finer relationships between specific government interventions, business sectors, political orientation, and resulting digitization shifts. Finally, we discuss the implications of rapid & widespread digitization for small businesses in the context of usability challenges and interpersonal interactions, while highlighting potential shifts in pre-existing social norms."
                    },
                    {
                        "title": "Semi-Supervised Models via Data Augmentationfor Classifying Interactive Affective Responses",
                        "abstract": "We present semi-supervised models with data augmentation (SMDA), a semi-supervised text classification system to classify interactive affective responses. SMDA utilizes recent transformer-based models to encode each sentence and employs back translation techniques to paraphrase given sentences as augmented data. For labeled sentences, we performed data augmentations to uniform the label distributions and computed supervised loss during training process. For unlabeled sentences, we explored self-training by regarding low-entropy predictions over unlabeled sentences as pseudo labels, assuming high-confidence predictions as labeled data for training. We further introduced consistency regularization as unsupervised loss after data augmentations on unlabeled data, based on the assumption that the model should predict similar class distributions with original unlabeled sentences as input and augmented sentences as input. Via a set of experiments, we demonstrated that our system outperformed baseline models in terms of F1-score and accuracy."
                    },
                    {
                        "title": "Tuiteamos o pongamos un tuit? Investigating the Social Constraints of Loanword Integration in Spanish Social Media",
                        "abstract": "Speakers of non-English languages often adopt loanwords from English to express new or unusual concepts. While these loanwords may be borrowed unchanged, speakers may also integrate the words to fit the constraints of their native language, e.g. creating Spanish \"tuitear\" from English \"tweet.\" Linguists have often considered the process of loanword integration to be more dependent on language-internal constraints, but sociolinguistic constraints such as speaker background remain only qualitatively understood. We investigate the role of social context and speaker background in Spanish speakers' use of integrated loanwords on social media. We find first that newspaper authors use the integrated forms of loanwords and native words more often than social media authors, showing that integration is associated with formal domains. In social media, we find that speaker background and expectations of formality explain loanword and native word integration, such that authors who use more Spanish and who write to a wider audience tend to use integrated verb forms more often. This study shows that loanword integration reflects not only language-internal constraints but also social expectations that vary by conversation and speaker."
                    },
                    {
                        "title": "HiddenCut: Simple Data Augmentation for Natural Language Understanding with Better Generalization",
                        "abstract": "Fine-tuning large pre-trained models with task-specific data has achieved great success in NLP. However, it has been demonstrated that the majority of information within the self-attention networks is redundant and not utilized effectively during the fine-tuning stage. This leads to inferior results when generalizing the obtained models to out-of-domain distributions. To this end, we propose a simple yet effective data augmentation technique, HiddenCut, to better regularize the model and encourage it to learn more generalizable features. Specifically, contiguous spans within the hidden space are dynamically and strategically dropped during training. Experiments show that our HiddenCut method outperforms the state-of-the-art augmentation methods on the GLUE benchmark, and consistently exhibits superior generalization performances on out-of-distribution and challenging counterexamples. We have publicly released our code at https://github.com/GT-SALT/HiddenCut."
                    },
                    {
                        "title": "Measure and Improve Robustness in NLP Models: A Survey",
                        "abstract": "As NLP models achieved state-of-the-art performances over benchmarks and gained wide applications, it has been increasingly important to ensure the safe deployment of these models in the real world, e.g., making sure the models are robust against unseen or challenging scenarios. Despite robustness being an increasingly studied topic, it has been separately explored in applications like vision and NLP, with various definitions, evaluation and mitigation strategies in multiple lines of research. In this paper, we aim to provide a unifying survey of how to define, measure and improve robustness in NLP. We first connect multiple definitions of robustness, then unify various lines of work on identifying robustness failures and evaluating models' robustness. Correspondingly, we present mitigation strategies that are data-driven, model-driven, and inductive-prior-based, with a more systematic view of how to effectively improve robustness in NLP models. Finally, we conclude by outlining open challenges and future directions to motivate further research in this area."
                    },
                    {
                        "title": "Continual Sequence Generation with Adaptive Compositional Modules",
                        "abstract": "Continual learning is essential for real-world deployment when there is a need to quickly adapt the model to new tasks without forgetting knowledge of old tasks. Existing work on continual sequence generation either always reuses existing parameters to learn new tasks, which is vulnerable to catastrophic forgetting on dissimilar tasks, or blindly adds new parameters for every new task, which could prevent knowledge sharing between similar tasks. To get the best of both worlds, in this work, we propose continual sequence generation with adaptive compositional modules to adaptively add modules in transformer architectures and compose both old and new modules for new tasks. We also incorporate pseudo experience replay to facilitate knowledge transfer in those shared modules. Experiment results on various sequences of generation tasks show that our framework can adaptively add modules or reuse modules based on task similarity, outperforming state-of-the-art baselines in terms of both performance and parameter efficiency. We make our code public at https://github.com/GT-SALT/Adaptive-Compositional-Modules."
                    }
                ]
            }
        }
    },
    "2407.05622": {
        "paper_data": {
            "title": "On the Complexity of Learning Sparse Functions with Statistical and Gradient Queries",
            "url": "http://arxiv.org/abs/2407.05622v1",
            "arxiv_id": "2407.05622",
            "authors": [
                "Nirmit Joshi",
                "Theodor Misiakiewicz",
                "Nathan Srebro"
            ],
            "abstract": "The goal of this paper is to investigate the complexity of gradient algorithms when learning sparse functions (juntas). We introduce a type of Statistical Queries ($\\mathsf{SQ}$), which we call Differentiable Learning Queries ($\\mathsf{DLQ}$), to model gradient queries on a specified loss with respect to an arbitrary model. We provide a tight characterization of the query complexity of $\\mathsf{DLQ}$ for learning the support of a sparse function over generic product distributions. This complexity crucially depends on the loss function. For the squared loss, $\\mathsf{DLQ}$ matches the complexity of Correlation Statistical Queries $(\\mathsf{CSQ})$--potentially much worse than $\\mathsf{SQ}$. But for other simple loss functions, including the $\\ell_1$ loss, $\\mathsf{DLQ}$ always achieves the same complexity as $\\mathsf{SQ}$. We also provide evidence that $\\mathsf{DLQ}$ can indeed capture learning with (stochastic) gradient descent by showing it correctly describes the complexity of learning with a two-layer neural network in the mean field regime and linear scaling.",
            "introduction": "   1 Introduction  In recent years, major efforts have been devoted to understanding which distributions can be learned efficiently using gradient-type algorithms on generic models (Abbe and Sandon, 2020; Allen-Zhu and Li, 2020; Malach et\u00a0al., 2021; Damian et\u00a0al., 2022; Abbe et\u00a0al., 2021b, a, 2022, 2023; Bietti et\u00a0al., 2023; Glasgow, 2023; Dandi et\u00a0al., 2023, 2024; Edelman et\u00a0al., 2024; Kou et\u00a0al., 2024). In this paper, we focus on learning sparse functions (i.e.\u00a0\u201cjuntas\u201d (Blum and Langley, 1997)), that is functions that depend only on a small number P\ud835\udc43{P}italic_P out of a much larger set d\u226bPmuch-greater-than\ud835\udc51\ud835\udc43d\\gg{P}italic_d \u226b italic_P of input coordinates. The challenge in this setting is to identify the few relevant coordinates. For some sparse functions, such as noisy parities, learning is believed to require O\u2062(dP)\ud835\udc42superscript\ud835\udc51\ud835\udc43O(d^{P})italic_O ( italic_d start_POSTSUPERSCRIPT italic_P end_POSTSUPERSCRIPT ) runtime (Kearns, 1998), while others, such as linear functions, are easy to learn in O~\u2062(d)~\ud835\udc42\ud835\udc51\\tilde{O}(d)over~ start_ARG italic_O end_ARG ( italic_d ) time. Which functions are easy to learn and which are hard? What is the complexity of learning a specific sparse function? Recent works (Abbe et\u00a0al., 2022, 2023) unveiled a rich \u201cleap\u201d hierarchical structure and saddle-to-saddle dynamics that drives learning in this setting. The goal of the present paper is to provide a general characterization for the complexity of learning sparse functions that go beyond (i) hypercube data and Fourier analysis, and (ii) \ud835\udda2\ud835\uddb2\ud835\uddb0\ud835\udda2\ud835\uddb2\ud835\uddb0{\\sf CSQ}sansserif_CSQ (see below) and focusing only on the squared loss.   The notion of complexity we consider is the Statistical Query (\ud835\uddb2\ud835\uddb0\ud835\uddb2\ud835\uddb0{\\sf SQ}sansserif_SQ) complexity, which studies learning by measuring expectations up to some worst-case tolerance (see (Kearns, 1998; Bshouty and Feldman, 2002; Reyzin, 2020) and Section 3). Although based on worst-case error (or almost equivalently, additive independent noise) rather than the sampling error encountered in practice, statistical query complexity has been proven to be a useful guideline for studying the complexity of learning. In particular, gradient computations are a special case of statistical queries. In specific cases which include binary functions or gradients on the squared or cross-entropy loss, gradient queries are equivalent111This equivalence holds if the input distribution is known, which is the case we consider here. to the restricted class of Correlation Statistical Queries (\ud835\udda2\ud835\uddb2\ud835\uddb0\ud835\udda2\ud835\uddb2\ud835\uddb0{\\sf CSQ}sansserif_CSQ) which are strictly less powerful than general statistical queries. Lower bounds on the \ud835\udda2\ud835\uddb2\ud835\uddb0\ud835\udda2\ud835\uddb2\ud835\uddb0{\\sf CSQ}sansserif_CSQ complexity have thus been seen as corresponding to the complexity of gradient-based learning222In Section 8, we discuss recent work Dandi et\u00a0al. (2024) that showed that even with the squared loss, multiple gradient evaluations on the same batch can be strictly more powerful than \ud835\udda2\ud835\uddb2\ud835\uddb0\ud835\udda2\ud835\uddb2\ud835\uddb0{\\sf CSQ}sansserif_CSQ. in these restricted cases. Part of the motivation for this paper is to emphasize that this relationship is limited to very specific loss functions and does not hold more generally. In order to study the complexity of gradient algorithms for general output and loss functions, we introduce a type of statistical query which we call Differentiable Learning Query (\ud835\udda3\ud835\uddab\ud835\uddb0\ud835\udda3\ud835\uddab\ud835\uddb0{\\sf DLQ}sansserif_DLQ). These queries are defined with respect to a specific loss \u2113\u2113\\ellroman_\u2113\u2014denoted by \ud835\udda3\ud835\uddab\ud835\uddb0\u2113subscript\ud835\udda3\ud835\uddab\ud835\uddb0\u2113{\\sf DLQ}_{\\ell}sansserif_DLQ start_POSTSUBSCRIPT roman_\u2113 end_POSTSUBSCRIPT\u2014and are given by gradients on a loss \u2113\u2113\\ellroman_\u2113 with respect to an arbitrary model. Specifically, \ud835\udda3\ud835\uddab\ud835\uddb0\u2113subscript\ud835\udda3\ud835\uddab\ud835\uddb0\u2113{\\sf DLQ}_{\\ell}sansserif_DLQ start_POSTSUBSCRIPT roman_\u2113",
            "references": [
                {
                    "title": "Matching the Statistical Query Lower Bound for k-sparse Parity Problems with Stochastic Gradient Descent",
                    "abstract": "The $k$-parity problem is a classical problem in computational complexity and algorithmic theory, serving as a key benchmark for understanding computational classes. In this paper, we solve the $k$-parity problem with stochastic gradient descent (SGD) on two-layer fully-connected neural networks. We demonstrate that SGD can efficiently solve the $k$-sparse parity problem on a $d$-dimensional hypercube ($k\\le O(\\sqrt{d})$) with a sample complexity of $\\tilde{O}(d^{k-1})$ using $2^{\\Theta(k)}$ neurons, thus matching the established $\\Omega(d^{k})$ lower bounds of Statistical Query (SQ) models. Our theoretical analysis begins by constructing a good neural network capable of correctly solving the $k$-parity problem. We then demonstrate how a trained neural network with SGD can effectively approximate this good network, solving the $k$-parity problem with small statistical errors. Our theoretical results and findings are supported by empirical evidence, showcasing the efficiency and efficacy of our approach."
                },
                {
                    "title": "The Benefits of Reusing Batches for Gradient Descent in Two-Layer Networks: Breaking the Curse of Information and Leap Exponents",
                    "abstract": "We investigate the training dynamics of two-layer neural networks when learning multi-index target functions. We focus on multi-pass gradient descent (GD) that reuses the batches multiple times and show that it significantly changes the conclusion about which functions are learnable compared to single-pass gradient descent. In particular, multi-pass GD with finite stepsize is found to overcome the limitations of gradient flow and single-pass GD given by the information exponent (Ben Arous et al., 2021) and leap exponent (Abbe et al., 2023) of the target function. We show that upon re-using batches, the network achieves in just two time steps an overlap with the target subspace even for functions not satisfying the staircase property (Abbe et al., 2021). We characterize the (broad) class of functions efficiently learned in finite time. The proof of our results is based on the analysis of the Dynamical Mean-Field Theory (DMFT). We further provide a closed-form description of the dynamical process of the low-dimensional projections of the weights, and numerical experiments illustrating the theory."
                },
                {
                    "title": "On Learning Gaussian Multi-index Models with Gradient Flow",
                    "abstract": "We study gradient flow on the multi-index regression problem for high-dimensional Gaussian data. Multi-index functions consist of a composition of an unknown low-rank linear projection and an arbitrary unknown, low-dimensional link function. As such, they constitute a natural template for feature learning in neural networks. We consider a two-timescale algorithm, whereby the low-dimensional link function is learnt with a non-parametric model infinitely faster than the subspace parametrizing the low-rank projection. By appropriately exploiting the matrix semigroup structure arising over the subspace correlation matrices, we establish global convergence of the resulting Grassmannian population gradient flow dynamics, and provide a quantitative description of its associated `saddle-to-saddle' dynamics. Notably, the timescales associated with each saddle can be explicitly characterized in terms of an appropriate Hermite decomposition of the target link function. In contrast with these positive results, we also show that the related \\emph{planted} problem, where the link function is known and fixed, in fact has a rough optimization landscape, in which gradient flow dynamics might get trapped with high probability."
                },
                {
                    "title": "SGD Finds then Tunes Features in Two-Layer Neural Networks with near-Optimal Sample Complexity: A Case Study in the XOR problem",
                    "abstract": "In this work, we consider the optimization process of minibatch stochastic gradient descent (SGD) on a 2-layer neural network with data separated by a quadratic ground truth function. We prove that with data drawn from the $d$-dimensional Boolean hypercube labeled by the quadratic ``XOR'' function $y = -x_ix_j$, it is possible to train to a population error $o(1)$ with $d \\:\\text{polylog}(d)$ samples. Our result considers simultaneously training both layers of the two-layer-neural network with ReLU activations via standard minibatch SGD on the logistic loss. To our knowledge, this work is the first to give a sample complexity of $\\tilde{O}(d)$ for efficiently learning the XOR function on isotropic data on a standard neural network with standard training. Our main technique is showing that the network evolves in two phases: a $\\textit{signal-finding}$ phase where the network is small and many of the neurons evolve independently to find features, and a $\\textit{signal-heavy}$ phase, where SGD maintains and balances the features. We leverage the simultaneous training of the layers to show that it is sufficient for only a small fraction of the neurons to learn features, since those neurons will be amplified by the simultaneous growth of their second layer weights."
                },
                {
                    "title": "Smoothing the Landscape Boosts the Signal for SGD: Optimal Sample Complexity for Learning Single Index Models",
                    "abstract": "We focus on the task of learning a single index model $\\sigma(w^\\star \\cdot x)$ with respect to the isotropic Gaussian distribution in $d$ dimensions. Prior work has shown that the sample complexity of learning $w^\\star$ is governed by the information exponent $k^\\star$ of the link function $\\sigma$, which is defined as the index of the first nonzero Hermite coefficient of $\\sigma$. Ben Arous et al. (2021) showed that $n \\gtrsim d^{k^\\star-1}$ samples suffice for learning $w^\\star$ and that this is tight for online SGD. However, the CSQ lower bound for gradient based methods only shows that $n \\gtrsim d^{k^\\star/2}$ samples are necessary. In this work, we close the gap between the upper and lower bounds by showing that online SGD on a smoothed loss learns $w^\\star$ with $n \\gtrsim d^{k^\\star/2}$ samples. We also draw connections to statistical analyses of tensor PCA and to the implicit regularization effects of minibatch SGD on empirical losses."
                },
                {
                    "title": "SGD learning on neural networks: leap complexity and saddle-to-saddle dynamics",
                    "abstract": "We investigate the time complexity of SGD learning on fully-connected neural networks with isotropic data. We put forward a complexity measure -- the leap -- which measures how\"hierarchical\"target functions are. For $d$-dimensional uniform Boolean or isotropic Gaussian data, our main conjecture states that the time complexity to learn a function $f$ with low-dimensional support is $\\tilde\\Theta (d^{\\max(\\mathrm{Leap}(f),2)})$. We prove a version of this conjecture for a class of functions on Gaussian isotropic data and 2-layer neural networks, under additional technical assumptions on how SGD is run. We show that the training sequentially learns the function support with a saddle-to-saddle dynamic. Our result departs from [Abbe et al. 2022] by going beyond leap 1 (merged-staircase functions), and by going beyond the mean-field and gradient flow approximations that prohibit the full complexity control obtained here. Finally, we note that this gives an SGD complexity for the full training trajectory that matches that of Correlational Statistical Query (CSQ) lower-bounds."
                },
                {
                    "title": "Learning Single-Index Models with Shallow Neural Networks",
                    "abstract": "Single-index models are a class of functions given by an unknown univariate ``link'' function applied to an unknown one-dimensional projection of the input. These models are particularly relevant in high dimension, when the data might present low-dimensional structure that learning algorithms should adapt to. While several statistical aspects of this model, such as the sample complexity of recovering the relevant (one-dimensional) subspace, are well-understood, they rely on tailored algorithms that exploit the specific structure of the target function. In this work, we introduce a natural class of shallow neural networks and study its ability to learn single-index models via gradient flow. More precisely, we consider shallow networks in which biases of the neurons are frozen at random initialization. We show that the corresponding optimization landscape is benign, which in turn leads to generalization guarantees that match the near-optimal sample complexity of dedicated semi-parametric methods."
                },
                {
                    "title": "Neural Networks Efficiently Learn Low-Dimensional Representations with SGD",
                    "abstract": "We study the problem of training a two-layer neural network (NN) of arbitrary width using stochastic gradient descent (SGD) where the input $\\boldsymbol{x}\\in \\mathbb{R}^d$ is Gaussian and the target $y \\in \\mathbb{R}$ follows a multiple-index model, i.e., $y=g(\\langle\\boldsymbol{u_1},\\boldsymbol{x}\\rangle,...,\\langle\\boldsymbol{u_k},\\boldsymbol{x}\\rangle)$ with a noisy link function $g$. We prove that the first-layer weights of the NN converge to the $k$-dimensional principal subspace spanned by the vectors $\\boldsymbol{u_1},...,\\boldsymbol{u_k}$ of the true model, when online SGD with weight decay is used for training. This phenomenon has several important consequences when $k \\ll d$. First, by employing uniform convergence on this smaller subspace, we establish a generalization error bound of $O(\\sqrt{{kd}/{T}})$ after $T$ iterations of SGD, which is independent of the width of the NN. We further demonstrate that, SGD-trained ReLU NNs can learn a single-index target of the form $y=f(\\langle\\boldsymbol{u},\\boldsymbol{x}\\rangle) + \\epsilon$ by recovering the principal direction, with a sample complexity linear in $d$ (up to log factors), where $f$ is a monotonic function with at most polynomial growth, and $\\epsilon$ is the noise. This is in contrast to the known $d^{\\Omega(p)}$ sample requirement to learn any degree $p$ polynomial in the kernel regime, and it shows that NNs trained with SGD can outperform the neural tangent kernel at initialization. Finally, we also provide compressibility guarantees for NNs using the approximate low-rank structure produced by SGD."
                },
                {
                    "title": "On the non-universality of deep learning: quantifying the cost of symmetry",
                    "abstract": "We prove limitations on what neural networks trained by noisy gradient descent (GD) can efficiently learn. Our results apply whenever GD training is equivariant, which holds for many standard architectures and initializations. As applications, (i) we characterize the functions that fully-connected networks can weak-learn on the binary hypercube and unit sphere, demonstrating that depth-2 is as powerful as any other depth for this task; (ii) we extend the merged-staircase necessity result for learning with latent low-dimensional structure [ABM22] to beyond the mean-field regime. Under cryptographic assumptions, we also show hardness results for learning with fully-connected networks trained by stochastic gradient descent (SGD)."
                },
                {
                    "title": "Neural Networks can Learn Representations with Gradient Descent",
                    "abstract": "Significant theoretical work has established that in specific regimes, neural networks trained by gradient descent behave like kernel methods. However, in practice, it is known that neural networks strongly outperform their associated kernels. In this work, we explain this gap by demonstrating that there is a large class of functions which cannot be efficiently learned by kernel methods but can be easily learned with gradient descent on a two layer neural network outside the kernel regime by learning representations that are relevant to the target task. We also demonstrate that these representations allow for efficient transfer learning, which is impossible in the kernel regime. Specifically, we consider the problem of learning polynomials which depend on only a few relevant directions, i.e. of the form $f^\\star(x) = g(Ux)$ where $U: \\R^d \\to \\R^r$ with $d \\gg r$. When the degree of $f^\\star$ is $p$, it is known that $n \\asymp d^p$ samples are necessary to learn $f^\\star$ in the kernel regime. Our primary result is that gradient descent learns a representation of the data which depends only on the directions relevant to $f^\\star$. This results in an improved sample complexity of $n\\asymp d^2 r + dr^p$. Furthermore, in a transfer learning setup where the data distributions in the source and target domain share the same representation $U$ but have different polynomial heads we show that a popular heuristic for transfer learning has a target sample complexity independent of $d$."
                },
                {
                    "title": "The merged-staircase property: a necessary and nearly sufficient condition for SGD learning of sparse functions on two-layer neural networks",
                    "abstract": "It is currently known how to characterize functions that neural networks can learn with SGD for two extremal parameterizations: neural networks in the linear regime, and neural networks with no structural constraints. However, for the main parametrization of interest (non-linear but regular networks) no tight characterization has yet been achieved, despite significant developments. We take a step in this direction by considering depth-2 neural networks trained by SGD in the mean-field regime. We consider functions on binary inputs that depend on a latent low-dimensional subspace (i.e., small number of coordinates). This regime is of interest since it is poorly understood how neural networks routinely tackle high-dimensional datasets and adapt to latent low-dimensional structure without suffering from the curse of dimensionality. Accordingly, we study SGD-learnability with $O(d)$ sample complexity in a large ambient dimension $d$. Our main results characterize a hierarchical property, the\"merged-staircase property\", that is both necessary and nearly sufficient for learning in this setting. We further show that non-linear training is necessary: for this class of functions, linear methods on any feature map (e.g., the NTK) are not capable of learning efficiently. The key tools are a new\"dimension-free\"dynamics approximation result that applies to functions defined on a latent space of low-dimension, a proof of global convergence based on polynomial identity testing, and an improvement of lower bounds against linear methods for non-almost orthogonal functions."
                },
                {
                    "title": "The staircase property: How hierarchical structure can guide deep learning",
                    "abstract": "This paper identi\ufb01es a structural property of data distributions that enables deep neural networks to learn hierarchically. We de\ufb01ne the \u201cstaircase\u201d property for functions over the Boolean hypercube, which posits that high-order Fourier coef\ufb01cients are reachable from lower-order Fourier coef\ufb01cients along increasing chains. We prove that functions satisfying this property can be learned in polynomial time using layerwise stochastic coordinate descent on regular neural networks \u2013 a class of network architectures and initializations that have homogeneity properties. Our analysis shows that for such staircase functions and neural networks, the gradient-based algorithm learns high-level features by greedily combining lower-level features along the depth of the network. We further back our theoretical results with experiments showing that staircase functions are learnable by more standard ResNet architectures with stochastic gradient descent. Both the theoretical and experimental results support the fact that the staircase property has a role to play in understanding the capabilities of gradient-based learning on regular networks, in contrast to general polynomial-size networks that can emulate any Statistical Query or PAC algorithm, as recently shown."
                },
                {
                    "title": "On the Power of Differentiable Learning versus PAC and SQ Learning",
                    "abstract": "We study the power of learning via mini-batch stochastic gradient descent (SGD) on the population loss, and batch Gradient Descent (GD) on the empirical loss, of a differentiable model or neural network, and ask what learning problems can be learnt using these paradigms. We show that SGD and GD can always simulate learning with statistical queries (SQ), but their ability to go beyond that depends on the precision $\\rho$ of the gradient calculations relative to the minibatch size $b$ (for SGD) and sample size $m$ (for GD). With fine enough precision relative to minibatch size, namely when $b \\rho$ is small enough, SGD can go beyond SQ learning and simulate any sample-based learning algorithm and thus its learning power is equivalent to that of PAC learning; this extends prior work that achieved this result for $b=1$. Similarly, with fine enough precision relative to the sample size $m$, GD can also simulate any sample-based learning algorithm based on $m$ samples. In particular, with polynomially many bits of precision (i.e. when $\\rho$ is exponentially small), SGD and GD can both simulate PAC learning regardless of the mini-batch size. On the other hand, when $b \\rho^2$ is large enough, the power of SGD is equivalent to that of SQ learning."
                },
                {
                    "title": "Minimum complexity interpolation in random features models",
                    "abstract": "Despite their many appealing properties, kernel methods are heavily affected by the curse of dimensionality. For instance, in the case of inner product kernels in $\\mathbb{R}^d$, the Reproducing Kernel Hilbert Space (RKHS) norm is often very large for functions that depend strongly on a small subset of directions (ridge functions). Correspondingly, such functions are difficult to learn using kernel methods. This observation has motivated the study of generalizations of kernel methods, whereby the RKHS norm -- which is equivalent to a weighted $\\ell_2$ norm -- is replaced by a weighted functional $\\ell_p$ norm, which we refer to as $\\mathcal{F}_p$ norm. Unfortunately, tractability of these approaches is unclear. The kernel trick is not available and minimizing these norms requires to solve an infinite-dimensional convex problem. We study random features approximations to these norms and show that, for $p>1$, the number of random features required to approximate the original learning problem is upper bounded by a polynomial in the sample size. Hence, learning with $\\mathcal{F}_p$ norms is tractable in these cases. We introduce a proof technique based on uniform concentration in the dual, which can be of broader interest in the study of overparametrized models. For $p= 1$, our guarantees for the random features approximation break down. We prove instead that learning with the $\\mathcal{F}_1$ norm is $\\mathsf{NP}$-hard under a randomized reduction based on the problem of learning halfspaces with noise."
                },
                {
                    "title": "Quantifying the Benefit of Using Differentiable Learning over Tangent Kernels",
                    "abstract": "We study the relative power of learning with gradient descent on differentiable models, such as neural networks, versus using the corresponding tangent kernels. We show that under certain conditions, gradient descent achieves small error only if a related tangent kernel method achieves a non-trivial advantage over random guessing (a.k.a. weak learning), though this advantage might be very small even when gradient descent can achieve arbitrarily high accuracy. Complementing this, we show that without these conditions, gradient descent can in fact learn with small error even when no kernel method, in particular using the tangent kernel, can achieve a non-trivial advantage over random guessing."
                },
                {
                    "title": "Classifying high-dimensional Gaussian mixtures: Where kernel methods fail and neural networks succeed",
                    "abstract": "A recent series of theoretical works showed that the dynamics of neural networks with a certain initialisation are well-captured by kernel methods. Concurrent empirical work demonstrated that kernel methods can come close to the performance of neural networks on some image classification tasks. These results raise the question of whether neural networks only learn successfully if kernels also learn successfully, despite neural networks being more expressive. Here, we show theoretically that two-layer neural networks (2LNN) with only a few hidden neurons can beat the performance of kernel learning on a simple Gaussian mixture classification task. We study the high-dimensional limit where the number of samples is linearly proportional to the input dimension, and show that while small 2LNN achieve near-optimal performance on this task, lazy training approaches such as random features and kernel methods do not. Our analysis is based on the derivation of a closed set of equations that track the learning dynamics of the 2LNN and thus allow to extract the asymptotic performance of the network as a function of signal-to-noise ratio and other hyperparameters. We finally illustrate how over-parametrising the neural network leads to faster convergence, but does not improve its final performance."
                },
                {
                    "title": "Statistical Queries and Statistical Algorithms: Foundations and Applications",
                    "abstract": "We give a survey of the foundations of statistical queries and their many applications to other areas. We introduce the model, give the main definitions, and we explore the fundamental theory statistical queries and how how it connects to various notions of learnability. We also give a detailed summary of some of the applications of statistical queries to other areas, including to optimization, to evolvability, and to differential privacy."
                },
                {
                    "title": "Online stochastic gradient descent on non-convex losses from high-dimensional inference",
                    "abstract": "Stochastic gradient descent (SGD) is a popular algorithm for optimization problems arising in high-dimensional inference tasks. Here one produces an estimator of an unknown parameter from independent samples of data by iteratively optimizing a loss function. This loss function is random and often non-convex. We study the performance of the simplest version of SGD, namely online SGD, from a random start in the setting where the parameter space is high-dimensional. \nWe develop nearly sharp thresholds for the number of samples needed for consistent estimation as one varies the dimension. Our thresholds depend only on an intrinsic property of the population loss which we call the information exponent. In particular, our results do not assume uniform control on the loss itself, such as convexity or uniform derivative bounds. The thresholds we obtain are polynomial in the dimension and the precise exponent depends explicitly on the information exponent. As a consequence of our results, we find that except for the simplest tasks, almost all of the data is used simply in the initial search phase to obtain non-trivial correlation with the ground truth. Upon attaining non-trivial correlation, the descent is rapid and exhibits law of large numbers type behaviour. \nWe illustrate our approach by applying it to a wide set of inference tasks such as phase retrieval, parameter estimation for generalized linear models, spiked matrix models, and spiked tensor models, as well as for supervised learning for single-layer networks with general activation functions."
                },
                {
                    "title": "Backward Feature Correction: How Deep Learning Performs Deep Learning",
                    "abstract": "How does a 110-layer ResNet learn a high-complexity classifier using relatively few training examples and short training time? We present a theory towards explaining this in terms of hierarchical learning. We refer hierarchical learning as the learner learns to represent a complicated target function by decomposing it into a sequence of simpler functions to reduce sample and time complexity. This paper formally analyzes how multi-layer neural networks can perform such hierarchical learning efficiently and automatically by applying SGD. \nOn the conceptual side, we present, to the best of our knowledge, the FIRST theory result indicating how deep neural networks can be sample and time efficient on certain hierarchical learning tasks, when NO KNOWN non-hierarchical algorithms (such as kernel method, linear regression over feature mappings, tensor decomposition, sparse coding, and their simple combinations) are efficient. We establish a principle called \"backward feature correction\", where training higher layers in the network can improve the features of lower level ones. We believe this is the key to understand the deep learning process in multi-layer neural networks. \nOn the technical side, we show for every input dimension $d > 0$, there is a concept class consisting of degree $\\omega(1)$ multi-variate polynomials so that, using $\\omega(1)$-layer neural networks as learners, SGD can learn any target function from this class in $\\mathsf{poly}(d)$ time using $\\mathsf{poly}(d)$ samples to any $\\frac{1}{\\mathsf{poly}(d)}$ error, through learning to represent it as a composition of $\\omega(1)$ layers of quadratic functions. In contrast, we present lower bounds stating that several non-hierarchical learners, including any kernel methods, neural tangent kernels, must suffer from $d^{\\omega(1)}$ sample or time complexity to learn this concept class even to $d^{-0.01}$ error."
                },
                {
                    "title": "Notes on Computational Hardness of Hypothesis Testing: Predictions using the Low-Degree Likelihood Ratio",
                    "abstract": "These notes survey and explore an emerging method, which we call the low-degree method, for predicting and understanding statistical-versus-computational tradeoffs in high-dimensional inference problems. In short, the method posits that a certain quantity -- the second moment of the low-degree likelihood ratio -- gives insight into how much computational time is required to solve a given hypothesis testing problem, which can in turn be used to predict the computational hardness of a variety of statistical inference tasks. While this method originated in the study of the sum-of-squares (SoS) hierarchy of convex programs, we present a self-contained introduction that does not require knowledge of SoS. In addition to showing how to carry out predictions using the method, we include a discussion investigating both rigorous and conjectural consequences of these predictions. \nThese notes include some new results, simplified proofs, and refined conjectures. For instance, we point out a formal connection between spectral methods and the low-degree likelihood ratio, and we give a sharp low-degree lower bound against subexponential-time algorithms for tensor PCA."
                },
                {
                    "title": "Optimality and Sub-optimality of PCA I: Spiked Random Matrix Models",
                    "abstract": "A central problem of random matrix theory is to understand the eigenvalues of spiked random matrix models, introduced by Johnstone, in which a prominent eigenvector (or \"spike\") is planted into a random matrix. These distributions form natural statistical models for principal component analysis (PCA) problems throughout the sciences. Baik, Ben Arous and Peche showed that the spiked Wishart ensemble exhibits a sharp phase transition asymptotically: when the spike strength is above a critical threshold, it is possible to detect the presence of a spike based on the top eigenvalue, and below the threshold the top eigenvalue provides no information. Such results form the basis of our understanding of when PCA can detect a low-rank signal in the presence of noise. However, under structural assumptions on the spike, not all information is necessarily contained in the spectrum. We study the statistical limits of tests for the presence of a spike, including non-spectral tests. Our results leverage Le Cam's notion of contiguity, and include: \ni) For the Gaussian Wigner ensemble, we show that PCA achieves the optimal detection threshold for certain natural priors for the spike. \nii) For any non-Gaussian Wigner ensemble, PCA is sub-optimal for detection. However, an efficient variant of PCA achieves the optimal threshold (for natural priors) by pre-transforming the matrix entries. \niii) For the Gaussian Wishart ensemble, the PCA threshold is optimal for positive spikes (for natural priors) but this is not always the case for negative spikes."
                },
                {
                    "title": "On the Global Convergence of Gradient Descent for Over-parameterized Models using Optimal Transport",
                    "abstract": "Many tasks in machine learning and signal processing can be solved by minimizing a convex function of a measure. This includes sparse spikes deconvolution or training a neural network with a single hidden layer. For these problems, we study a simple minimization method: the unknown measure is discretized into a mixture of particles and a continuous-time gradient descent is performed on their weights and positions. This is an idealization of the usual way to train neural networks with a large hidden layer. We show that, when initialized correctly and in the many-particle limit, this gradient flow, although non-convex, converges to global minimizers. The proof involves Wasserstein gradient flows, a by-product of optimal transport theory. Numerical experiments show that this asymptotic behavior is already at play for a reasonable number of particles, even in high dimension."
                },
                {
                    "title": "Gradient Descent for One-Hidden-Layer Neural Networks: Polynomial Convergence and SQ Lower Bounds",
                    "abstract": "We study the complexity of training neural network models with one hidden nonlinear activation layer and an output weighted sum layer. We analyze Gradient Descent applied to learning a bounded target function on $n$ real-valued inputs. We give an agnostic learning guarantee for GD: starting from a randomly initialized network, it converges in mean squared loss to the minimum error (in $2$-norm) of the best approximation of the target function using a polynomial of degree at most $k$. Moreover, for any $k$, the size of the network and number of iterations needed are both bounded by $n^{O(k)}\\log(1/\\epsilon)$. In particular, this applies to training networks of unbiased sigmoids and ReLUs. We also rigorously explain the empirical finding that gradient descent discovers lower frequency Fourier components before higher frequency components. \nWe complement this result with nearly matching lower bounds in the Statistical Query model. GD fits well in the SQ framework since each training step is determined by an expectation over the input distribution. We show that any SQ algorithm that achieves significant improvement over a constant function with queries of tolerance some inverse polynomial in the input dimensionality $n$ must use $n^{\\Omega(k)}$ queries even when the target functions are restricted to a set of $n^{O(k)}$ degree-$k$ polynomials, and the input distribution is uniform over the unit sphere; for this class the information-theoretic lower bound is only $\\Theta(k \\log n)$. \nOur approach for both parts is based on spherical harmonics. We view gradient descent as an operator on the space of functions, and study its dynamics. An essential tool is the Funk-Hecke theorem, which explains the eigenfunctions of this operator in the case of the mean squared loss."
                },
                {
                    "title": "Neural Networks as Interacting Particle Systems: Asymptotic Convexity of the Loss Landscape and Universal Scaling of the Approximation Error",
                    "abstract": "Neural networks, a central tool in machine learning, have demonstrated remarkable, high fidelity performance on image recognition and classification tasks. These successes evince an ability to accurately represent high dimensional functions, potentially of great use in computational and applied mathematics. That said, there are few rigorous results about the representation error and trainability of neural networks. Here we characterize both the error and the scaling of the error with the size of the network by reinterpreting the standard optimization algorithm used in machine learning applications, stochastic gradient descent, as the evolution of a particle system with interactions governed by a potential related to the objective or \"loss\" function used to train the network. We show that, when the number $n$ of parameters is large, the empirical distribution of the particles descends on a convex landscape towards a minimizer at a rate independent of $n$. We establish a Law of Large Numbers and a Central Limit Theorem for the empirical distribution, which together show that the approximation error of the network universally scales as $O(n^{-1})$. Remarkably, these properties do not depend on the dimensionality of the domain of the function that we seek to represent. Our analysis also quantifies the scale and nature of the noise introduced by stochastic gradient descent and provides guidelines for the step size and batch size to use when training a neural network. We illustrate our findings on examples in which we train neural network to learn the energy function of the continuous 3-spin model on the sphere. The approximation error scales as our analysis predicts in as high a dimension as $d=25$."
                },
                {
                    "title": "A mean field view of the landscape of two-layer neural networks",
                    "abstract": "Significance Multilayer neural networks have proven extremely successful in a variety of tasks, from image classification to robotics. However, the reasons for this practical success and its precise domain of applicability are unknown. Learning a neural network from data requires solving a complex optimization problem with millions of variables. This is done by stochastic gradient descent (SGD) algorithms. We study the case of two-layer networks and derive a compact description of the SGD dynamics in terms of a limiting partial differential equation. Among other consequences, this shows that SGD dynamics does not become more complex when the network size increases. Multilayer neural networks are among the most powerful models in machine learning, yet the fundamental reasons for this success defy mathematical understanding. Learning a neural network requires optimizing a nonconvex high-dimensional objective (risk function), a problem that is usually attacked using stochastic gradient descent (SGD). Does SGD converge to a global optimum of the risk or only to a local optimum? In the former case, does this happen because local minima are absent or because SGD somehow avoids them? In the latter, why do local minima reached by SGD have good generalization properties? In this paper, we consider a simple case, namely two-layer neural networks, and prove that\u2014in a suitable scaling limit\u2014SGD dynamics is captured by a certain nonlinear partial differential equation (PDE) that we call distributional dynamics (DD). We then consider several specific examples and show how DD can be used to prove convergence of SGD to networks with nearly ideal generalization error. This description allows for \u201caveraging out\u201d some of the complexities of the landscape of neural networks and can be used to prove a general convergence result for noisy SGD."
                },
                {
                    "title": "On the Complexity of Learning Neural Networks",
                    "abstract": "The stunning empirical successes of neural networks currently lack rigorous theoretical explanation. What form would such an explanation take, in the face of existing complexity-theoretic lower bounds? A first step might be to show that data generated by neural networks with a single hidden layer, smooth activation functions and benign input distributions can be learned efficiently. We demonstrate here a comprehensive lower bound ruling out this possibility: for a wide class of activation functions (including all currently used), and inputs drawn from any logconcave distribution, there is a family of one-hidden-layer functions whose output is a sum gate, that are hard to learn in a precise sense: any statistical query algorithm (which includes all known variants of stochastic gradient descent with any loss function) needs an exponential number of queries even using tolerance inversely proportional to the input dimensionality. Moreover, this hard family of functions is realizable with a small (sublinear in dimension) number of activation units in the single hidden layer. The lower bound is also robust to small perturbations of the true weights. Systematic experiments illustrate a phase transition in the training error as predicted by the analysis."
                },
                {
                    "title": "Convex Optimization: Algorithms and Complexity",
                    "abstract": "This monograph presents the main complexity theorems in convex optimization and their corresponding algorithms. Starting from the fundamental theory of black-box optimization, the material progresses towards recent advances in structural optimization and stochastic optimization. Our presentation of black-box optimization, strongly influenced by Nesterov's seminal book and Nemirovski's lecture notes, includes the analysis of cutting plane methods, as well as (accelerated) gradient descent schemes. We also pay special attention to non-Euclidean settings (relevant algorithms include Frank-Wolfe, mirror descent, and dual averaging) and discuss their relevance in machine learning. We provide a gentle introduction to structural optimization with FISTA (to optimize a sum of a smooth and a simple non-smooth term), saddle-point mirror prox (Nemirovski's alternative to Nesterov's smoothing), and a concise description of interior point methods. In stochastic optimization we discuss stochastic gradient descent, mini-batches, random coordinate descent, and sublinear algorithms. We also briefly touch upon convex relaxation of combinatorial problems and the use of randomness to round solutions, as well as random walks based methods."
                },
                {
                    "title": "Finding Correlations in Subquadratic Time, with Applications to Learning Parities and Juntas",
                    "abstract": "Given a set of n d-dimensional Boolean vectors with the promise that the vectors are chosen uniformly at random with the exception of two vectors that have Pearson-correlation \u03c1 (Hamming distance d \u00b7 1-\u03c1/2), how quickly can one find the two correlated vectors? We present an algorithm which, for any constants \u03b5, \u03c1 >; 0 and d >;>; logn/\u03c1<sup>2</sup> , finds the correlated pair with high probability, and runs in time O(n 3\u03c9/4 + \u03f5) <; O(n<sup>1.8</sup>), where w <; 2.38 is the exponent of matrix multiplication. Provided that d is sufficiently large, this runtime can be further reduced. These are the first subquadratic-time algorithms for this problem for which \u03c1 does not appear in the exponent of n, and improves upon O(n<sup>2-O</sup>(\u03c1)), given by Paturi et al. [15], Locality Sensitive Hashing (LSH) [11] and the Bucketing Codes approach [6]. Applications and extensions of this basic algorithm yield improved algorithms for several other problems: ApproximateClosest Pair: For any sufficiently small constant \u03f5 >; 0, given n vectors in R<sup>d</sup>, our algorithm returns a pair of vectors whose Euclidean distance differs from that of the closest pair by a factor of at most 1+\u03f5, and runs in time O(n<sup>2-\u0398(\u221a\u03f5)</sup>). The best previous algorithms (including LSH) have runtime O(n<sup>2-O(\u03f5)</sup>). Learning Sparse Parity with Noise: Given samples from an instance of the learning parity with noise problem where each example has length n, the true parity set has size at most k <;<; n, and the noise rate is \u03b7, our algorithm identifies the set of k indices in time n \u03c9+\u03f5/3 <sup>k</sup> poly(1/1-2\u03b7) <; n<sup>0.8k</sup>poly(1/1-2\u03b7). This is the first algorithm with no depenJence on \u03b7 in the exponent of n, aside from the trivial brute-force algorithm. Learning k-Juntas with Noise: Given uniformly random length n Boolean vectors, together with a label, which is some function of just k <;<; n of the bits, perturbed by noise rate \u03b7, return the set of relevant indices. Leveraging the reduction of Feldman et al. [7] our result for learning k-parities implies an algorithm for this problem with runtime n \u03c9+\u03f5/3 <sup>k</sup> poly(1/1-2\u03b7) <; n<sup>0.8k</sup> poly(1/1-2\u03b7), 2 which improves on the previous best of >; n<sup>k</sup>(1-2/2k)poly( 1/1-2\u03b7 ), from [8]. Learning k-Juntas without Noise:1 Our results for learning sparse parities with noise imply an algorithm for learning juntas without noise with runtime n \u03c9+\u03f5/4<sup>k</sup> poly(n) <; n<sup>0.6</sup> kpoly(n), which improves on the runtime n \u03c9+1/\u03c9 poly(n) \u2248 n<sup>0.7k</sup> poly(n) of Mossel n et al. [13]."
                },
                {
                    "title": "Statistical algorithms and a lower bound for detecting planted cliques",
                    "abstract": "We introduce a framework for proving lower bounds on computational problems over distributions, based on a class of algorithms called statistical algorithms. For such algorithms, access to the input distribution is limited to obtaining an estimate of the expectation of any given function on a sample drawn randomly from the input distribution, rather than directly accessing samples. Most natural algorithms of interest in theory and in practice, e.g., moments-based methods, local search, standard iterative methods for convex optimization, MCMC and simulated annealing, are statistical algorithms or have statistical counterparts. Our framework is inspired by and generalize the statistical query model in learning theory [34]. Our main application is a nearly optimal lower bound on the complexity of any statistical algorithm for detecting planted bipartite clique distributions (or planted dense subgraph distributions) when the planted clique has size O(n1/2-\u03b4) for any constant \u03b4 > 0. Variants of these problems have been assumed to be hard to prove hardness for other problems and for cryptographic applications. Our lower bounds provide concrete evidence of hardness, thus supporting these assumptions."
                },
                {
                    "title": "Efficient noise-tolerant learning from statistical queries",
                    "abstract": "In this paper, we study the problem of learning in the presence of classification noise in the probabilistic learning model of Valiant and its variants. In order to identify the class of \u201crobust\u201d learning algorithms in the most general way, we formalize a new but related model of learning from statistical queries. Intuitively, in this model, a learning algorithm is forbidden to examine individual examples of the unknown target function, but is given access to an oracle providing estimates of probabilities over the sample space of random examples. One of our main results shows that any class of functions learnable from statistical queries is in fact learnable with classification noise in Valiant\u2019s model, with a noise rate approaching the informationtheoretic barrier of 1/2. We then demonstrate the generality of the statistical query model, showing that practically every class learnable in Valiant\u2019s model and its variants can also be learned in the new model (and thus can be learned in the presence of noise). A notable exception to this statement is the class of parity functions, which we prove is not learnable from statistical queries, and for which no noise-tolerant algorithm is known."
                },
                {
                    "title": "Pareto Frontiers in Deep Feature Learning: Data, Compute, Width, and Luck",
                    "abstract": "In modern deep learning, algorithmic choices (such as width, depth, and learning rate) are known to modulate nuanced resource tradeoffs. This work investigates how these complexities necessarily arise for feature learning in the presence of computational-statistical gaps. We begin by considering offline sparse parity learning, a supervised classification problem which admits a statistical query lower bound for gradient-based training of a multilayer perceptron. This lower bound can be interpreted as a multi-resource tradeoff frontier : successful learning can only occur if one is sufficiently rich (large model), knowledgeable (large dataset), patient (many training iterations), or lucky (many random guesses). We show, theoretically and experimentally, that sparse initialization and increasing network width yield significant improvements in sample efficiency in this setting. Here, width plays the role of parallel search: it amplifies the probability of finding \u201clottery ticket\u201d neurons, which learn sparse features more sample-efficiently. Finally, we show that the synthetic sparse parity task can be useful as a proxy for real problems requiring axis-aligned feature learning. We demonstrate improved sample efficiency on tabular classification benchmarks by using wide, sparsely-initialized MLP models; these networks sometimes outperform tuned random forests."
                },
                {
                    "title": "Learning Two-Layer Neural Networks, One (Giant) Step at a Time",
                    "abstract": "We study the training dynamics of shallow neural networks, investigating the conditions under which a limited number of large batch gradient descent steps can facilitate feature learning beyond the kernel regime. We compare the influence of batch size and that of multiple (but finitely many) steps. Our analysis of a single-step process reveals that while a batch size of n = O ( d ) enables feature learning, it is only adequate for learning a single direction, or a single-index model. In contrast, n = O ( d 2 ) is essential for learning multiple directions and specialization. Moreover, we demonstrate that \u201chard\u201d directions, which lack the first \u2113 Hermite coefficients, remain unobserved and require a batch size of n = O ( d \u2113 ) for being captured by gradient descent. Upon iterating a few steps, the scenario changes: a batch-size of n = O ( d ) is enough to learn new target directions spanning the subspace linearly connected in the Hermite basis to the previously learned directions, thereby a staircase property. Our analysis utilizes a blend of techniques related to concentration, projection-based conditioning, and Gaussian equivalence that are of independent interest. By determining the conditions necessary for learning and specialization, our results highlight the interaction between batch size and number of iterations, and lead to a hierarchical depiction where learning performance exhibits a stairway to accuracy over time and batch size, shedding new light on feature learning in neural networks."
                },
                {
                    "title": "On the universality of deep learning",
                    "abstract": "This paper shows that deep learning, i.e., neural networks trained by SGD, can learn in polytime any function class that can be learned in polytime by some algorithmm, including parities. This universal result is further shown to be robust, i.e., it holds under possibly poly-noise on the gradients, which gives a separation between deep learning and statistical query algorithms, as the latter are not comparably universal due to cases like parities. This also shows that SGD-based deep learning does not suffer from the limitations of the perceptron discussed by Minsky-Papert \u201969. The paper further complement this result with a lower-bound on the generalization error of descent algorithms, which implies in particular that the robust universality breaks down if the gradients are averaged over large enough batches of samples as in full-GD, rather than fewer samples as in SGD."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.DS"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nWhat is the complexity of learning sparse functions using gradient-type algorithms, particularly in the context of differentiable learning queries?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it will deepen our understanding of the theoretical limits of learning algorithms, particularly in identifying which types of sparse functions can be learned efficiently. This could lead to advancements in machine learning methodologies, influencing future research directions and practical applications in areas such as feature selection, data compression, and optimization problems where sparse representations are beneficial.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of identifying relevant input coordinates in high-dimensional spaces, especially when the functions exhibit varying degrees of sparsity. Naive approaches may fail due to the exponential runtime required for certain sparse functions, such as noisy parities, which necessitate O(d^P) time, while others may be easier to learn. Additionally, the need to generalize findings beyond specific loss functions and data distributions adds a layer of complexity, as does the requirement to develop a new framework for statistical query complexity that accommodates differentiable learning queries.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on specific types of functions or loss functions, often overlooking the broader landscape of sparse function learning. Limitations in existing solutions include a lack of generalization beyond hypercube data and Fourier analysis, as well as the restricted use of correlation statistical queries. These barriers have prevented a comprehensive understanding of the complexity involved in learning sparse functions. Our approach aims to fill these gaps by introducing a new type of statistical query that can handle a wider variety of loss functions and learning scenarios.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the introduction of Differentiable Learning Queries (DLQs) to analyze the complexity of learning sparse functions. We will utilize a diverse set of datasets that represent various types of sparse functions and apply metrics based on statistical query complexity to evaluate performance. The expected outcomes include a clearer characterization of the learning complexity associated with different sparse functions, insights into the efficiency of gradient-based algorithms, and the establishment of a framework that can guide future research in this area."
            }
        },
        "author_data": {
            "dac27a7a-db18-4616-a963-60caa6ff106d": {
                "pk": "dac27a7a-db18-4616-a963-60caa6ff106d",
                "name": "Nirmit Joshi",
                "collaborators": [
                    "Julia Gaudio",
                    "Nathan Srebro",
                    "Gal Vardi",
                    "Sepehr Assadi",
                    "Milind Prabhu",
                    "Vihan Shah",
                    "Kumar Kshitij Patel",
                    "Margalit Glasgow",
                    "Ali Zindari",
                    "Lingxiao Wang"
                ],
                "domain": [
                    "Community Detection",
                    "Graph Theory",
                    "Machine Learning",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Community Detection in the Hypergraph SBM: Exact Recovery Given the Similarity Matrix",
                        "abstract": "Community detection is a fundamental problem in network science. In this paper, we consider community detection in hypergraphs drawn from the $hypergraph$ $stochastic$ $block$ $model$ (HSBM), with a focus on exact community recovery. We study the performance of polynomial-time algorithms which operate on the $similarity$ $matrix$ $W$, where $W_{ij}$ reports the number of hyperedges containing both $i$ and $j$. Under this information model, while the precise information-theoretic limit is unknown, Kim, Bandeira, and Goemans derived a sharp threshold up to which the natural min-bisection estimator on $W$ succeeds. As min-bisection is NP-hard in the worst case, they additionally proposed a semidefinite programming (SDP) relaxation and conjectured that it achieves the same recovery threshold as the min-bisection estimator.   In this paper, we confirm this conjecture. We also design a simple and highly efficient spectral algorithm with nearly linear runtime and show that it achieves the min-bisection threshold. Moreover, the spectral algorithm also succeeds in denser regimes and is considerably more efficient than previous approaches, establishing it as the method of choice. Our analysis of the spectral algorithm crucially relies on strong $entrywise$ bounds on the eigenvectors of $W$. Our bounds are inspired by the work of Abbe, Fan, Wang, and Zhong, who developed entrywise bounds for eigenvectors of symmetric matrices with independent entries. Despite the complex dependency structure in similarity matrices, we prove similar entrywise guarantees."
                    },
                    {
                        "title": "Exact Community Recovery (under Side Information): Optimality of Spectral Algorithms",
                        "abstract": "In this paper, we study the problem of exact community recovery in general, two-community block models considering both Bernoulli and Gaussian matrix models, capturing the Stochastic Block Model, submatrix localization, and $\\mathbb{Z}_2$-synchronization as special cases. We also study the settings where $side$ $information$ about community assignment labels is available, modeled as passing the true labels through a noisy channel: either the binary erasure channel (where some community labels are known while others are erased) or the binary symmetric channel (where some labels are flipped). We provide a unified analysis of the effect of side information on the information-theoretic limits of exact recovery, generalizing prior works and extending to new settings. Additionally, we design a simple but optimal spectral algorithm that incorporates side information (when present) along with the eigenvectors of the matrix observation. Using the powerful tool of entrywise eigenvector analysis [Abbe, Fan, Wang, Zhong 2020], we show that our spectral algorithm can mimic the so called $genie$-$aided$ $estimators$, where the $i^{\\mathrm{th}}$ genie-aided estimator optimally computes the estimate of the $i^{\\mathrm{th}}$ label, when all remaining labels are revealed by a genie. This perspective provides a unified understanding of the optimality of spectral algorithms for various exact recovery problems in a recent line of work."
                    },
                    {
                        "title": "Noisy Interpolation Learning with Shallow Univariate ReLU Networks",
                        "abstract": "Understanding how overparameterized neural networks generalize despite perfect interpolation of noisy training data is a fundamental question. Mallinar et. al. 2022 noted that neural networks seem to often exhibit ``tempered overfitting'', wherein the population risk does not converge to the Bayes optimal error, but neither does it approach infinity, yielding non-trivial generalization. However, this has not been studied rigorously. We provide the first rigorous analysis of the overfitting behavior of regression with minimum norm ($\\ell_2$ of weights), focusing on univariate two-layer ReLU networks. We show overfitting is tempered (with high probability) when measured with respect to the $L_1$ loss, but also show that the situation is more complex than suggested by Mallinar et. al., and overfitting is catastrophic with respect to the $L_2$ loss, or when taking an expectation over the training set."
                    },
                    {
                        "title": "Generalizing Greenwald-Khanna Streaming Quantile Summaries for Weighted Inputs",
                        "abstract": "Estimating quantiles, like the median or percentiles, is a fundamental task in data mining and data science. A (streaming) quantile summary is a data structure that can process a set S of n elements in a streaming fashion and at the end, for any phi in (0,1], return a phi-quantile of S up to an eps error, i.e., return a phi'-quantile with phi'=phi +- eps. We are particularly interested in comparison-based summaries that only compare elements of the universe under a total ordering and are otherwise completely oblivious of the universe. The best known deterministic quantile summary is the 20-year old Greenwald-Khanna (GK) summary that uses O((1/eps) log(eps n)) space [SIGMOD'01]. This bound was recently proved to be optimal for all deterministic comparison-based summaries by Cormode and Vesle\\'y [PODS'20].   In this paper, we study weighted quantiles, a generalization of the quantiles problem, where each element arrives with a positive integer weight which denotes the number of copies of that element being inserted. The only known method of handling weighted inputs via GK summaries is the naive approach of breaking each weighted element into multiple unweighted items and feeding them one by one to the summary, which results in a prohibitively large update time (proportional to the maximum weight of input elements).   We give the first non-trivial extension of GK summaries for weighted inputs and show that it takes O((1/eps) log(eps n)) space and O(log(1/eps)+ log log(eps n)) update time per element to process a stream of length n (under some quite mild assumptions on the range of weights and eps). En route to this, we also simplify the original GK summaries for unweighted quantiles."
                    },
                    {
                        "title": "The Limits and Potentials of Local SGD for Distributed Heterogeneous Learning with Intermittent Communication",
                        "abstract": "Local SGD is a popular optimization method in distributed learning, often outperforming other algorithms in practice, including mini-batch SGD. Despite this success, theoretically proving the dominance of local SGD in settings with reasonable data heterogeneity has been difficult, creating a significant gap between theory and practice. In this paper, we provide new lower bounds for local SGD under existing first-order data heterogeneity assumptions, showing that these assumptions are insufficient to prove the effectiveness of local update steps. Furthermore, under these same assumptions, we demonstrate the min-max optimality of accelerated mini-batch SGD, which fully resolves our understanding of distributed optimization for several problem classes. Our results emphasize the need for better models of data heterogeneity to understand the effectiveness of local SGD in practice. Towards this end, we consider higher-order smoothness and heterogeneity assumptions, providing new upper bounds that imply the dominance of local SGD over mini-batch SGD when data heterogeneity is low."
                    }
                ]
            },
            "783e2883-8e46-486d-baab-b15a5e898186": {
                "pk": "783e2883-8e46-486d-baab-b15a5e898186",
                "name": "Theodor Misiakiewicz",
                "collaborators": [
                    "Andrea Montanari",
                    "Song Mei",
                    "Behrooz Ghorbani",
                    "Hong Hu",
                    "Yue M. Lu",
                    "Emmanuel Abbe",
                    "Enric Boix-Adsera",
                    "Basil Saeed",
                    "Marc Vuffray",
                    "Michael Celentano"
                ],
                "domain": [
                    "Kernel Methods",
                    "Neural Networks",
                    "High-Dimensional Statistics",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Spectrum of inner-product kernel matrices in the polynomial regime and multiple descent phenomenon in kernel ridge regression",
                        "abstract": "We study the spectrum of inner-product kernel matrices, i.e., $n \\times n$ matrices with entries $h (\\langle \\textbf{x}_i ,\\textbf{x}_j \\rangle/d)$ where the $( \\textbf{x}_i)_{i \\leq n}$ are i.i.d.~random covariates in $\\mathbb{R}^d$. In the linear high-dimensional regime $n \\asymp d$, it was shown that these matrices are well approximated by their linearization, which simplifies into the sum of a rescaled Wishart matrix and identity matrix. In this paper, we generalize this decomposition to the polynomial high-dimensional regime $n \\asymp d^\\ell,\\ell \\in \\mathbb{N}$, for data uniformly distributed on the sphere and hypercube. In this regime, the kernel matrix is well approximated by its degree-$\\ell$ polynomial approximation and can be decomposed into a low-rank spike matrix, identity and a `Gegenbauer matrix' with entries $Q_\\ell (\\langle \\textbf{x}_i , \\textbf{x}_j \\rangle)$, where $Q_\\ell$ is the degree-$\\ell$ Gegenbauer polynomial. We show that the spectrum of the Gegenbauer matrix converges in distribution to a Marchenko-Pastur law.   This problem is motivated by the study of the prediction error of kernel ridge regression (KRR) in the polynomial regime $n \\asymp d^\\kappa, \\kappa >0$. Previous work showed that for $\\kappa \\not\\in \\mathbb{N}$, KRR fits exactly a degree-$\\lfloor \\kappa \\rfloor$ polynomial approximation to the target function. In this paper, we use our characterization of the kernel matrix to complete this picture and compute the precise asymptotics of the test error in the limit $n/d^\\kappa \\to \\psi$ with $\\kappa \\in \\mathbb{N}$. In this case, the test error can present a double descent behavior, depending on the effective regularization and signal-to-noise ratio at level $\\kappa$. Because this double descent can occur each time $\\kappa$ crosses an integer, this explains the multiple descent phenomenon in the KRR risk curve observed in several previous works."
                    },
                    {
                        "title": "Six Lectures on Linearized Neural Networks",
                        "abstract": "In these six lectures, we examine what can be learnt about the behavior of multi-layer neural networks from the analysis of linear models. We first recall the correspondence between neural networks and linear models via the so-called lazy regime. We then review four models for linearized neural networks: linear regression with concentrated features, kernel ridge regression, random feature model and neural tangent model. Finally, we highlight the limitations of the linear theory and discuss how other approaches can overcome them."
                    },
                    {
                        "title": "Learning with convolution and pooling operations in kernel methods",
                        "abstract": "Recent empirical work has shown that hierarchical convolutional kernels inspired by convolutional neural networks (CNNs) significantly improve the performance of kernel methods in image classification tasks. A widely accepted explanation for their success is that these architectures encode hypothesis classes that are suitable for natural images. However, understanding the precise interplay between approximation and generalization in convolutional architectures remains a challenge. In this paper, we consider the stylized setting of covariates (image pixels) uniformly distributed on the hypercube, and characterize exactly the RKHS of kernels composed of single layers of convolution, pooling, and downsampling operations. We use this characterization to compute sharp asymptotics of the generalization error for any given function in high-dimension. In particular, we quantify the gain in sample complexity brought by enforcing locality with the convolution operation and approximate translation invariance with average pooling. Notably, these results provide a precise description of how convolution and pooling operations trade off approximation with generalization power in one layer convolutional kernels."
                    },
                    {
                        "title": "A non-asymptotic theory of Kernel Ridge Regression: deterministic equivalents, test error, and GCV estimator",
                        "abstract": "We consider learning an unknown target function $f_*$ using kernel ridge regression (KRR) given i.i.d. data $(u_i,y_i)$, $i\\leq n$, where $u_i \\in U$ is a covariate vector and $y_i = f_* (u_i) +\\varepsilon_i \\in \\mathbb{R}$. A recent string of work has empirically shown that the test error of KRR can be well approximated by a closed-form estimate derived from an `equivalent' sequence model that only depends on the spectrum of the kernel operator. However, a theoretical justification for this equivalence has so far relied either on restrictive assumptions -- such as subgaussian independent eigenfunctions -- , or asymptotic derivations for specific kernels in high dimensions.   In this paper, we prove that this equivalence holds for a general class of problems satisfying some spectral and concentration properties on the kernel eigendecomposition. Specifically, we establish in this setting a non-asymptotic deterministic approximation for the test error of KRR -- with explicit non-asymptotic bounds -- that only depends on the eigenvalues and the target function alignment to the eigenvectors of the kernel. Our proofs rely on a careful derivation of deterministic equivalents for random matrix functionals in the dimension free regime pioneered by Cheng and Montanari (2022).   We apply this setting to several classical examples and show an excellent agreement between theoretical predictions and numerical simulations. These results rely on having access to the eigendecomposition of the kernel operator. Alternatively, we prove that, under this same setting, the generalized cross-validation (GCV) estimator concentrates on the test error uniformly over a range of ridge regularization parameter that includes zero (the interpolating solution). As a consequence, the GCV estimator can be used to estimate from data the test error and optimal regularization parameter for KRR."
                    },
                    {
                        "title": "Concentration to Zero Bit-Error Probability for Regular LDPC Codes on the Binary Symmetric Channel: Proof by Loop Calculus",
                        "abstract": "In this paper we consider regular low-density parity-check codes over a binary-symmetric channel in the decoding regime. We prove that up to a certain noise threshold the bit-error probability of the bit-sampling decoder converges in mean to zero over the code ensemble and the channel realizations. To arrive at this result we show that the bit-error probability of the sampling decoder is equal to the derivative of a Bethe free entropy. The method that we developed is new and is based on convexity of the free entropy and loop calculus. Convexity is needed to exchange limit and derivative and the loop series enables us to express the difference between the bit-error probability and the Bethe free entropy. We control the loop series using combinatorial techniques and a first moment method. We stress that our method is versatile and we believe that it can be generalized for LDPC codes with general degree distributions and for asymmetric channels."
                    },
                    {
                        "title": "Learning with invariances in random features and kernel models",
                        "abstract": "A number of machine learning tasks entail a high degree of invariance: the data distribution does not change if we act on the data with a certain group of transformations. For instance, labels of images are invariant under translations of the images. Certain neural network architectures -- for instance, convolutional networks -- are believed to owe their success to the fact that they exploit such invariance properties. With the objective of quantifying the gain achieved by invariant architectures, we introduce two classes of models: invariant random features and invariant kernel methods. The latter includes, as a special case, the neural tangent kernel for convolutional networks with global average pooling. We consider uniform covariates distributions on the sphere and hypercube and a general invariant target function. We characterize the test error of invariant methods in a high-dimensional regime in which the sample size and number of hidden units scale as polynomials in the dimension, for a class of groups that we call `degeneracy $\\alpha$', with $\\alpha \\leq 1$. We show that exploiting invariance in the architecture saves a $d^\\alpha$ factor ($d$ stands for the dimension) in sample size and number of hidden units to achieve the same test error as for unstructured architectures.   Finally, we show that output symmetrization of an unstructured kernel estimator does not give a significant statistical improvement; on the other hand, data augmentation with an unstructured kernel estimator is equivalent to an invariant kernel estimator and enjoys the same improvement in statistical efficiency."
                    },
                    {
                        "title": "When Do Neural Networks Outperform Kernel Methods?",
                        "abstract": "For a certain scaling of the initialization of stochastic gradient descent (SGD), wide neural networks (NN) have been shown to be well approximated by reproducing kernel Hilbert space (RKHS) methods. Recent empirical work showed that, for some classification tasks, RKHS methods can replace NNs without a large loss in performance. On the other hand, two-layers NNs are known to encode richer smoothness classes than RKHS and we know of special examples for which SGD-trained NN provably outperform RKHS. This is true even in the wide network limit, for a different scaling of the initialization.   How can we reconcile the above claims? For which tasks do NNs outperform RKHS? If covariates are nearly isotropic, RKHS methods suffer from the curse of dimensionality, while NNs can overcome it by learning the best low-dimensional representation. Here we show that this curse of dimensionality becomes milder if the covariates display the same low-dimensional structure as the target function, and we precisely characterize this tradeoff. Building on these results, we present the spiked covariates model that can capture in a unified framework both behaviors observed in earlier work.   We hypothesize that such a latent low-dimensional structure is present in image classification. We test numerically this hypothesis by showing that specific perturbations of the training distribution degrade the performances of RKHS methods much more significantly than NNs."
                    },
                    {
                        "title": "Minimum complexity interpolation in random features models",
                        "abstract": "Despite their many appealing properties, kernel methods are heavily affected by the curse of dimensionality. For instance, in the case of inner product kernels in $\\mathbb{R}^d$, the Reproducing Kernel Hilbert Space (RKHS) norm is often very large for functions that depend strongly on a small subset of directions (ridge functions). Correspondingly, such functions are difficult to learn using kernel methods. This observation has motivated the study of generalizations of kernel methods, whereby the RKHS norm -- which is equivalent to a weighted $\\ell_2$ norm -- is replaced by a weighted functional $\\ell_p$ norm, which we refer to as $\\mathcal{F}_p$ norm. Unfortunately, tractability of these approaches is unclear. The kernel trick is not available and minimizing these norms requires to solve an infinite-dimensional convex problem.   We study random features approximations to these norms and show that, for $p>1$, the number of random features required to approximate the original learning problem is upper bounded by a polynomial in the sample size. Hence, learning with $\\mathcal{F}_p$ norms is tractable in these cases. We introduce a proof technique based on uniform concentration in the dual, which can be of broader interest in the study of overparametrized models. For $p= 1$, our guarantees for the random features approximation break down. We prove instead that learning with the $\\mathcal{F}_1$ norm is $\\mathsf{NP}$-hard under a randomized reduction based on the problem of learning halfspaces with noise."
                    },
                    {
                        "title": "Asymptotics of Random Feature Regression Beyond the Linear Scaling Regime",
                        "abstract": "Recent advances in machine learning have been achieved by using overparametrized models trained until near interpolation of the training data. It was shown, e.g., through the double descent phenomenon, that the number of parameters is a poor proxy for the model complexity and generalization capabilities. This leaves open the question of understanding the impact of parametrization on the performance of these models. How does model complexity and generalization depend on the number of parameters $p$? How should we choose $p$ relative to the sample size $n$ to achieve optimal test error?   In this paper, we investigate the example of random feature ridge regression (RFRR). This model can be seen either as a finite-rank approximation to kernel ridge regression (KRR), or as a simplified model for neural networks trained in the so-called lazy regime. We consider covariates uniformly distributed on the $d$-dimensional sphere and compute sharp asymptotics for the RFRR test error in the high-dimensional polynomial scaling, where $p,n,d \\to \\infty$ while $p/ d^{\\kappa_1}$ and $n / d^{\\kappa_2}$ stay constant, for all $\\kappa_1 , \\kappa_2 \\in \\mathbb{R}_{>0}$. These asymptotics precisely characterize the impact of the number of random features and regularization parameter on the test performance. In particular, RFRR exhibits an intuitive trade-off between approximation and generalization power. For $n = o(p)$, the sample size $n$ is the bottleneck and RFRR achieves the same performance as KRR (which is equivalent to taking $p = \\infty$). On the other hand, if $p = o(n)$, the number of random features $p$ is the limiting factor and RFRR test error matches the approximation error of the random feature model class (akin to taking $n = \\infty$). Finally, a double descent appears at $n= p$, a phenomenon that was previously only characterized in the linear scaling $\\kappa_1 = \\kappa_2 = 1$."
                    },
                    {
                        "title": "Mean-field theory of two-layers neural networks: dimension-free bounds and kernel limit",
                        "abstract": "We consider learning two layer neural networks using stochastic gradient descent. The mean-field description of this learning dynamics approximates the evolution of the network weights by an evolution in the space of probability distributions in $R^D$ (where $D$ is the number of parameters associated to each neuron). This evolution can be defined through a partial differential equation or, equivalently, as the gradient flow in the Wasserstein space of probability distributions. Earlier work shows that (under some regularity assumptions), the mean field description is accurate as soon as the number of hidden units is much larger than the dimension $D$. In this paper we establish stronger and more general approximation guarantees. First of all, we show that the number of hidden units only needs to be larger than a quantity dependent on the regularity properties of the data, and independent of the dimensions. Next, we generalize this analysis to the case of unbounded activation functions, which was not covered by earlier bounds. We extend our results to noisy stochastic gradient descent.   Finally, we show that kernel ridge regression can be recovered as a special limit of the mean field analysis."
                    },
                    {
                        "title": "Limitations of Lazy Training of Two-layers Neural Networks",
                        "abstract": "We study the supervised learning problem under either of the following two models: (1) Feature vectors ${\\boldsymbol x}_i$ are $d$-dimensional Gaussians and responses are $y_i = f_*({\\boldsymbol x}_i)$ for $f_*$ an unknown quadratic function; (2) Feature vectors ${\\boldsymbol x}_i$ are distributed as a mixture of two $d$-dimensional centered Gaussians, and $y_i$'s are the corresponding class labels. We use two-layers neural networks with quadratic activations, and compare three different learning regimes: the random features (RF) regime in which we only train the second-layer weights; the neural tangent (NT) regime in which we train a linearization of the neural network around its initialization; the fully trained neural network (NN) regime in which we train all the weights in the network. We prove that, even for the simple quadratic model of point (1), there is a potentially unbounded gap between the prediction risk achieved in these three training regimes, when the number of neurons is smaller than the ambient dimension. When the number of neurons is larger than the number of dimensions, the problem is significantly easier and both NT and NN learning achieve zero risk."
                    },
                    {
                        "title": "SGD learning on neural networks: leap complexity and saddle-to-saddle dynamics",
                        "abstract": "We investigate the time complexity of SGD learning on fully-connected neural networks with isotropic data. We put forward a complexity measure -- the leap -- which measures how \"hierarchical\" target functions are. For $d$-dimensional uniform Boolean or isotropic Gaussian data, our main conjecture states that the time complexity to learn a function $f$ with low-dimensional support is $\\tilde\\Theta (d^{\\max(\\mathrm{Leap}(f),2)})$. We prove a version of this conjecture for a class of functions on Gaussian isotropic data and 2-layer neural networks, under additional technical assumptions on how SGD is run. We show that the training sequentially learns the function support with a saddle-to-saddle dynamic. Our result departs from [Abbe et al. 2022] by going beyond leap 1 (merged-staircase functions), and by going beyond the mean-field and gradient flow approximations that prohibit the full complexity control obtained here. Finally, we note that this gives an SGD complexity for the full training trajectory that matches that of Correlational Statistical Query (CSQ) lower-bounds."
                    },
                    {
                        "title": "Dimension-free deterministic equivalents for random feature regression",
                        "abstract": "In this work we investigate the generalization performance of random feature ridge regression (RFRR). Our main contribution is a general deterministic equivalent for the test error of RFRR. Specifically, under a certain concentration property, we show that the test error is well approximated by a closed-form expression that only depends on the feature map eigenvalues. Notably, our approximation guarantee is non-asymptotic, multiplicative, and independent of the feature map dimension -- allowing for infinite-dimensional features. We expect this deterministic equivalent to hold broadly beyond our theoretical analysis, and we empirically validate its predictions on various real and synthetic datasets. As an application, we derive sharp excess error rates under standard power-law assumptions of the spectrum and target decay. In particular, we provide a tight result for the smallest number of features achieving optimal minimax error rate."
                    },
                    {
                        "title": "Efficient reconstruction of transmission probabilities in a spreading process from partial observations",
                        "abstract": "An important problem of reconstruction of diffusion network and transmission probabilities from the data has attracted a considerable attention in the past several years. A number of recent papers introduced efficient algorithms for the estimation of spreading parameters, based on the maximization of the likelihood of observed cascades, assuming that the full information for all the nodes in the network is available. In this work, we focus on a more realistic and restricted scenario, in which only a partial information on the cascades is available: either the set of activation times for a limited number of nodes, or the states of nodes for a subset of observation times. To tackle this problem, we first introduce a framework based on the maximization of the likelihood of the incomplete diffusion trace. However, we argue that the computation of this incomplete likelihood is a computationally hard problem, and show that a fast and robust reconstruction of transmission probabilities in sparse networks can be achieved with a new algorithm based on recently introduced dynamic message-passing equations for the spreading processes. The suggested approach can be easily generalized to a large class of discrete and continuous dynamic models, as well as to the cases of dynamically-changing networks and noisy information."
                    },
                    {
                        "title": "The merged-staircase property: a necessary and nearly sufficient condition for SGD learning of sparse functions on two-layer neural networks",
                        "abstract": "It is currently known how to characterize functions that neural networks can learn with SGD for two extremal parameterizations: neural networks in the linear regime, and neural networks with no structural constraints. However, for the main parametrization of interest (non-linear but regular networks) no tight characterization has yet been achieved, despite significant developments.   We take a step in this direction by considering depth-2 neural networks trained by SGD in the mean-field regime. We consider functions on binary inputs that depend on a latent low-dimensional subspace (i.e., small number of coordinates). This regime is of interest since it is poorly understood how neural networks routinely tackle high-dimensional datasets and adapt to latent low-dimensional structure without suffering from the curse of dimensionality. Accordingly, we study SGD-learnability with $O(d)$ sample complexity in a large ambient dimension $d$.   Our main results characterize a hierarchical property, the \"merged-staircase property\", that is both necessary and nearly sufficient for learning in this setting.   We further show that non-linear training is necessary: for this class of functions, linear methods on any feature map (e.g., the NTK) are not capable of learning efficiently. The key tools are a new \"dimension-free\" dynamics approximation result that applies to functions defined on a latent space of low-dimension, a proof of global convergence based on polynomial identity testing, and an improvement of lower bounds against linear methods for non-almost orthogonal functions."
                    },
                    {
                        "title": "Generalization error of random features and kernel methods: hypercontractivity and kernel matrix concentration",
                        "abstract": "Consider the classical supervised learning problem: we are given data $(y_i,{\\boldsymbol x}_i)$, $i\\le n$, with $y_i$ a response and ${\\boldsymbol x}_i\\in {\\mathcal X}$ a covariates vector, and try to learn a model $f:{\\mathcal X}\\to{\\mathbb R}$ to predict future responses. Random features methods map the covariates vector ${\\boldsymbol x}_i$ to a point ${\\boldsymbol \\phi}({\\boldsymbol x}_i)$ in a higher dimensional space ${\\mathbb R}^N$, via a random featurization map ${\\boldsymbol \\phi}$. We study the use of random features methods in conjunction with ridge regression in the feature space ${\\mathbb R}^N$. This can be viewed as a finite-dimensional approximation of kernel ridge regression (KRR), or as a stylized model for neural networks in the so called lazy training regime.   We define a class of problems satisfying certain spectral conditions on the underlying kernels, and a hypercontractivity assumption on the associated eigenfunctions. These conditions are verified by classical high-dimensional examples. Under these conditions, we prove a sharp characterization of the error of random features ridge regression. In particular, we address two fundamental questions: $(1)$~What is the generalization error of KRR? $(2)$~How big $N$ should be for the random features approximation to achieve the same error as KRR?   In this setting, we prove that KRR is well approximated by a projection onto the top $\\ell$ eigenfunctions of the kernel, where $\\ell$ depends on the sample size $n$. We show that the test error of random features ridge regression is dominated by its approximation error and is larger than the error of KRR as long as $N\\le n^{1-\\delta}$ for some $\\delta>0$. We characterize this gap. For $N\\ge n^{1+\\delta}$, random features achieve the same error as the corresponding KRR, and further increasing $N$ does not lead to a significant change in test error."
                    },
                    {
                        "title": "Linearized two-layers neural networks in high dimension",
                        "abstract": "We consider the problem of learning an unknown function $f_{\\star}$ on the $d$-dimensional sphere with respect to the square loss, given i.i.d. samples $\\{(y_i,{\\boldsymbol x}_i)\\}_{i\\le n}$ where ${\\boldsymbol x}_i$ is a feature vector uniformly distributed on the sphere and $y_i=f_{\\star}({\\boldsymbol x}_i)+\\varepsilon_i$. We study two popular classes of models that can be regarded as linearizations of two-layers neural networks around a random initialization: the random features model of Rahimi-Recht (RF); the neural tangent kernel model of Jacot-Gabriel-Hongler (NT). Both these approaches can also be regarded as randomized approximations of kernel ridge regression (with respect to different kernels), and enjoy universal approximation properties when the number of neurons $N$ diverges, for a fixed dimension $d$.   We consider two specific regimes: the approximation-limited regime, in which $n=\\infty$ while $d$ and $N$ are large but finite; and the sample size-limited regime in which $N=\\infty$ while $d$ and $n$ are large but finite. In the first regime we prove that if $d^{\\ell + \\delta} \\le N\\le d^{\\ell+1-\\delta}$ for small $\\delta > 0$, then \\RF\\, effectively fits a degree-$\\ell$ polynomial in the raw features, and \\NT\\, fits a degree-$(\\ell+1)$ polynomial. In the second regime, both RF and NT reduce to kernel methods with rotationally invariant kernels. We prove that, if the number of samples is $d^{\\ell + \\delta} \\le n \\le d^{\\ell +1-\\delta}$, then kernel methods can fit at most a a degree-$\\ell$ polynomial in the raw features. This lower bound is achieved by kernel ridge regression. Optimal prediction error is achieved for vanishing ridge regularization."
                    },
                    {
                        "title": "Precise Learning Curves and Higher-Order Scaling Limits for Dot Product Kernel Regression",
                        "abstract": "As modern machine learning models continue to advance the computational frontier, it has become increasingly important to develop precise estimates for expected performance improvements under different model and data scaling regimes. Currently, theoretical understanding of the learning curves that characterize how the prediction error depends on the number of samples is restricted to either large-sample asymptotics ($m\\to\\infty$) or, for certain simple data distributions, to the high-dimensional asymptotics in which the number of samples scales linearly with the dimension ($m\\propto d$). There is a wide gulf between these two regimes, including all higher-order scaling relations $m\\propto d^r$, which are the subject of the present paper. We focus on the problem of kernel ridge regression for dot-product kernels and present precise formulas for the mean of the test error, bias, and variance, for data drawn uniformly from the sphere with isotropic random labels in the $r$th-order asymptotic scaling regime $m\\to\\infty$ with $m/d^r$ held constant. We observe a peak in the learning curve whenever $m \\approx d^r/r!$ for any integer $r$, leading to multiple sample-wise descent and nontrivial behavior at multiple scales."
                    },
                    {
                        "title": "Solving SDPs for synchronization and MaxCut problems via the Grothendieck inequality",
                        "abstract": "A number of statistical estimation problems can be addressed by semidefinite programs (SDP). While SDPs are solvable in polynomial time using interior point methods, in practice generic SDP solvers do not scale well to high-dimensional problems. In order to cope with this problem, Burer and Monteiro proposed a non-convex rank-constrained formulation, which has good performance in practice but is still poorly understood theoretically.   In this paper we study the rank-constrained version of SDPs arising in MaxCut and in synchronization problems. We establish a Grothendieck-type inequality that proves that all the local maxima and dangerous saddle points are within a small multiplicative gap from the global maximum. We use this structural information to prove that SDPs can be solved within a known accuracy, by applying the Riemannian trust-region method to this non-convex problem, while constraining the rank to be of order one. For the MaxCut problem, our inequality implies that any local maximizer of the rank-constrained SDP provides a $ (1 - 1/(k-1)) \\times 0.878$ approximation of the MaxCut, when the rank is fixed to $k$.   We then apply our results to data matrices generated according to the Gaussian ${\\mathbb Z}_2$ synchronization problem, and the two-groups stochastic block model with large bounded degree. We prove that the error achieved by local maximizers undergoes a phase transition at the same threshold as for information-theoretically optimal methods."
                    }
                ]
            },
            "717f4e11-0e67-4c0b-8984-23c331ec8b9d": {
                "pk": "717f4e11-0e67-4c0b-8984-23c331ec8b9d",
                "name": "Nathan Srebro",
                "collaborators": [
                    "Behnam Neyshabur",
                    "Weiran Wang",
                    "Jialei Wang",
                    "Blake Woodworth",
                    "Andrew Cotter",
                    "Martin Tak\u00e1\u010d",
                    "Peter Richt\u00e1rik",
                    "Mladen Kolar",
                    "Karthik Sridharan",
                    "Ambuj Tewari"
                ],
                "domain": [
                    "Machine Learning",
                    "Optimization",
                    "Graph Theory",
                    "Semi-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Maximum Likelihood Bounded Tree-Width Markov Networks",
                        "abstract": "Chow and Liu (1968) studied the problem of learning a maximumlikelihood Markov tree. We generalize their work to more complexMarkov networks by considering the problem of learning a maximumlikelihood Markov network of bounded complexity. We discuss howtree-width is in many ways the appropriate measure of complexity andthus analyze the problem of learning a maximum likelihood Markovnetwork of bounded tree-width.Similar to the work of Chow and Liu, we are able to formalize thelearning problem as a combinatorial optimization problem on graphs. Weshow that learning a maximum likelihood Markov network of boundedtree-width is equivalent to finding a maximum weight hypertree. Thisequivalence gives rise to global, integer-programming based,approximation algorithms with provable performance guarantees, for thelearning problem. This contrasts with heuristic local-searchalgorithms which were previously suggested (e.g. by Malvestuto 1991).The equivalence also allows us to study the computational hardness ofthe learning problem. We show that learning a maximum likelihoodMarkov network of bounded tree-width is NP-hard, and discuss thehardness of approximation."
                    },
                    {
                        "title": "Lower Bound for Randomized First Order Convex Optimization",
                        "abstract": "We provide an explicit construction and direct proof for the lower bound on the number of first order oracle accesses required for a randomized algorithm to minimize a convex Lipschitz function."
                    },
                    {
                        "title": "On the Universality of Online Mirror Descent",
                        "abstract": "We show that for a general class of convex online learning problems, Mirror Descent can always achieve a (nearly) optimal regret guarantee."
                    },
                    {
                        "title": "Semi-supervised Learning with Density Based Distances",
                        "abstract": "We present a simple, yet effective, approach to Semi-Supervised Learning. Our approach is based on estimating density-based distances (DBD) using a shortest path calculation on a graph. These Graph-DBD estimates can then be used in any distance-based supervised learning method, such as Nearest Neighbor methods and SVMs with RBF kernels. In order to apply the method to very large data sets, we also present a novel algorithm which integrates nearest neighbor computations into the shortest path search and can find exact shortest paths even in extremely large dense graphs. Significant runtime improvement over the commonly used Laplacian regularization method is then shown on a large scale dataset."
                    },
                    {
                        "title": "Collaborative Filtering in a Non-Uniform World: Learning with the Weighted Trace Norm",
                        "abstract": "We show that matrix completion with trace-norm regularization can be significantly hurt when entries of the matrix are sampled non-uniformly. We introduce a weighted version of the trace-norm regularizer that works well also with non-uniform sampling. Our experimental results demonstrate that the weighted trace-norm regularization indeed yields significant gains on the (highly non-uniformly sampled) Netflix dataset."
                    },
                    {
                        "title": "Clustering using Max-norm Constrained Optimization",
                        "abstract": "We suggest using the max-norm as a convex surrogate constraint for clustering. We show how this yields a better exact cluster recovery guarantee than previously suggested nuclear-norm relaxation, and study the effectiveness of our method, and other related convex relaxations, compared to other clustering approaches."
                    },
                    {
                        "title": "Open Problem: The Oracle Complexity of Convex Optimization with Limited Memory",
                        "abstract": "We note that known methods achieving the optimal oracle complexity for first order convex optimization require quadratic memory, and ask whether this is necessary, and more broadly seek to characterize the minimax number of first order queries required to optimize a convex Lipschitz function subject to a memory constraint."
                    },
                    {
                        "title": "Stochastic Optimization of PCA with Capped MSG",
                        "abstract": "We study PCA as a stochastic optimization problem and propose a novel stochastic approximation algorithm which we refer to as \"Matrix Stochastic Gradient\" (MSG), as well as a practical variant, Capped MSG. We study the method both theoretically and empirically."
                    },
                    {
                        "title": "Norm-Based Capacity Control in Neural Networks",
                        "abstract": "We investigate the capacity, convexity and characterization of a general family of norm-constrained feed-forward networks."
                    },
                    {
                        "title": "Concentration-Based Guarantees for Low-Rank Matrix Reconstruction",
                        "abstract": "We consider the problem of approximately reconstructing a partially-observed, approximately low-rank matrix. This problem has received much attention lately, mostly using the trace-norm as a surrogate to the rank. Here we study low-rank matrix reconstruction using both the trace-norm, as well as the less-studied max-norm, and present reconstruction guarantees based on existing analysis on the Rademacher complexity of the unit balls of these norms. We show how these are superior in several ways to recently published guarantees based on specialized analysis."
                    },
                    {
                        "title": "Stochastic Nonconvex Optimization with Large Minibatches",
                        "abstract": "We study stochastic optimization of nonconvex loss functions, which are typical objectives for training neural networks. We propose stochastic approximation algorithms which optimize a series of regularized, nonlinearized losses on large minibatches of samples, using only first-order gradient information. Our algorithms provably converge to an approximate critical point of the expected objective with faster rates than minibatch stochastic gradient descent, and facilitate better parallelization by allowing larger minibatches."
                    },
                    {
                        "title": "Mini-Batch Primal and Dual Methods for SVMs",
                        "abstract": "We address the issue of using mini-batches in stochastic optimization of SVMs. We show that the same quantity, the spectral norm of the data, controls the parallelization speedup obtained for both primal stochastic subgradient descent (SGD) and stochastic dual coordinate ascent (SCDA) methods and use it to derive novel variants of mini-batched SDCA. Our guarantees for both methods are expressed in terms of the original nonsmooth primal problem based on the hinge-loss."
                    },
                    {
                        "title": "Clustering, Hamming Embedding, Generalized LSH and the Max Norm",
                        "abstract": "We study the convex relaxation of clustering and hamming embedding, focusing on the asymmetric case (co-clustering and asymmetric hamming embedding), understanding their relationship to LSH as studied by (Charikar 2002) and to the max-norm ball, and the differences between their symmetric and asymmetric versions."
                    },
                    {
                        "title": "Distributed Mini-Batch SDCA",
                        "abstract": "We present an improved analysis of mini-batched stochastic dual coordinate ascent for regularized empirical loss minimization (i.e. SVM and SVM-type objectives). Our analysis allows for flexible sampling schemes, including where data is distribute across machines, and combines a dependence on the smoothness of the loss and/or the data spread (measured through the spectral norm)."
                    },
                    {
                        "title": "Distributed Multi-Task Learning with Shared Representation",
                        "abstract": "We study the problem of distributed multi-task learning with shared representation, where each machine aims to learn a separate, but related, task in an unknown shared low-dimensional subspaces, i.e. when the predictor matrix has low rank. We consider a setting where each task is handled by a different machine, with samples for the task available locally on the machine, and study communication-efficient methods for exploiting the shared structure."
                    },
                    {
                        "title": "Efficient coordinate-wise leading eigenvector computation",
                        "abstract": "We develop and analyze efficient \"coordinate-wise\" methods for finding the leading eigenvector, where each step involves only a vector-vector product. We establish global convergence with overall runtime guarantees that are at least as good as Lanczos's method and dominate it for slowly decaying spectrum. Our methods are based on combining a shift-and-invert approach with coordinate-wise algorithms for linear regression."
                    },
                    {
                        "title": "Exploring Generalization in Deep Learning",
                        "abstract": "With a goal of understanding what drives generalization in deep networks, we consider several recently suggested explanations, including norm-based control, sharpness and robustness. We study how these measures can ensure generalization, highlighting the importance of scale normalization, and making a connection between sharpness and PAC-Bayes theory. We then investigate how well the measures explain different observed phenomena."
                    },
                    {
                        "title": "Distributed Stochastic Multi-Task Learning with Graph Regularization",
                        "abstract": "We propose methods for distributed graph-based multi-task learning that are based on weighted averaging of messages from other machines. Uniform averaging or diminishing stepsize in these methods would yield consensus (single task) learning. We show how simply skewing the averaging weights or controlling the stepsize allows learning different, but related, tasks on the different machines."
                    },
                    {
                        "title": "Implicit Bias of Gradient Descent on Linear Convolutional Networks",
                        "abstract": "We show that gradient descent on full-width linear convolutional networks of depth $L$ converges to a linear predictor related to the $\\ell_{2/L}$ bridge penalty in the frequency domain. This is in contrast to linearly fully connected networks, where gradient descent converges to the hard margin linear support vector machine solution, regardless of depth."
                    },
                    {
                        "title": "The Kernelized Stochastic Batch Perceptron",
                        "abstract": "We present a novel approach for training kernel Support Vector Machines, establish learning runtime guarantees for our method that are better then those of any other known kernelized SVM optimization approach, and show that our method works well in practice compared to existing alternatives."
                    }
                ]
            }
        }
    },
    "2407.10827": {
        "paper_data": {
            "title": "LLM Circuit Analyses Are Consistent Across Training and Scale",
            "url": "http://arxiv.org/abs/2407.10827v1",
            "arxiv_id": "2407.10827",
            "authors": [
                "Curt Tigges",
                "Michael Hanna",
                "Qinan Yu",
                "Stella Biderman"
            ],
            "abstract": "Most currently deployed large language models (LLMs) undergo continuous training or additional finetuning. By contrast, most research into LLMs' internal mechanisms focuses on models at one snapshot in time (the end of pre-training), raising the question of whether their results generalize to real-world settings. Existing studies of mechanisms over time focus on encoder-only or toy models, which differ significantly from most deployed models. In this study, we track how model mechanisms, operationalized as circuits, emerge and evolve across 300 billion tokens of training in decoder-only LLMs, in models ranging from 70 million to 2.8 billion parameters. We find that task abilities and the functional components that support them emerge consistently at similar token counts across scale. Moreover, although such components may be implemented by different attention heads over time, the overarching algorithm that they implement remains. Surprisingly, both these algorithms and the types of components involved therein can replicate across model scale. These results suggest that circuit analyses conducted on small models at the end of pre-training can provide insights that still apply after additional pre-training and over model scale.",
            "introduction": "   1 Introduction  As LLMs\u2019 capabilities have grown, so has interest in characterizing their mechanisms. Recent work in mechanistic interpretability often seeks to do so via circuits: computational subgraphs that explain task-solving mechanisms (Wang et\u00a0al., 2023; Hanna et\u00a0al., 2023; Merullo et\u00a0al., 2024; Lieberum et\u00a0al., 2023). Circuits can be found and verified using a variety of methods (Conmy et\u00a0al., 2023; Syed et\u00a0al., 2023; Hanna et\u00a0al., 2024; Kram\u00e1r et\u00a0al., 2024; Ferrando & Voita, 2024), with the aim of reverse-engineering models\u2019 task-solving algorithms.   Though much circuits research is motivated by LLMs\u2019 capabilities, the setting in which such research is performed often differs from that of currently deployed models. Crucially, while most LLM circuits work (Wang et\u00a0al., 2023; Hanna et\u00a0al., 2023; Merullo et\u00a0al., 2024; Lieberum et\u00a0al., 2023; Tigges et\u00a0al., 2023) studies models at the end of pre-training, currently deployed models often undergo continuous training (OpenAI et\u00a0al., 2024; Anthropic, 2024; Gemini Team et\u00a0al., 2024) or are fine-tuned for specific tasks (Chung et\u00a0al., 2022; Hu et\u00a0al., 2021). Other subfields of interpretability have studied model development during training (Hu et\u00a0al., 2023; Chang et\u00a0al., 2023; Warstadt et\u00a0al., 2020; Choshen et\u00a0al., 2022; Chang & Bergen, 2022), but similar work on LLM mechanisms is scarce. Existing mechanistic work over training has studied syntactic attention structures and induction heads (Olsson et\u00a0al., 2022; Chen et\u00a0al., 2024; Singh et\u00a0al., 2024), but has focused on small encoder or toy models. Prakash et\u00a0al. (2024) examines circuits in 7-billion-parameter models post-finetuning, but the evolution of circuits during pre-training remains unexplored. This raises questions about whether circuit analyses will generalize if the model in question is further trained or fine-tuned.   We address this issue by exploring when and how circuits and their components emerge during training, and their consistency across training and different model scales. We study circuits in models from the Pythia suite (Biderman et\u00a0al., 2023b) across 300 billion tokens, at scales from 70 million to 2.8 billion parameters. We supplement this with additional data from models ranging up to 12 billion parameters. Our results suggest remarkable consistency in circuits and their attributes across scale and training. We summarize our contributions as follows:   Performance acquisition and functional component emergence are similar across scale: Task ability acquisition rates tend to reach a maximum at similar token counts across different model sizes. Functional components like name mover heads, copy suppression heads, and successor heads also emerge consistently at similar points across scales, paralleling previous findings that induction heads emerge at roughly 2B-5B tokens across models of all scales (Olsson et\u00a0al., 2022).   Circuit algorithms can remain stable despite component-level fluctuations: Analysis of the indirect object identification (IOI; Wang et\u00a0al., 2023) circuit across training and scale reveals that even when individual components change, the overall algorithm remains consistent, indicating a degree of algorithmic stability. The algorithm also tends to be similar for dramatically different model scales, suggesting that some currently-identified circuits may generalize, at least on simple tasks.   Taken as a whole, our results suggest that circuit analysis can generalize well over both (pre-)training and scale even in the face of component and circuit size changes, and that circuits studied at the end of training in smaller models",
            "references": [
                {
                    "title": "A Primer on the Inner Workings of Transformer-based Language Models",
                    "abstract": "The rapid progress of research aimed at interpreting the inner workings of advanced language models has highlighted a need for contextualizing the insights gained from years of work in this area. This primer provides a concise technical introduction to the current techniques used to interpret the inner workings of Transformer-based language models, focusing on the generative decoder-only architecture. We conclude by presenting a comprehensive overview of the known internal mechanisms implemented by these models, uncovering connections across popular approaches and active research directions in this area."
                },
                {
                    "title": "What needs to go right for an induction head? A mechanistic study of in-context learning circuits and their formation",
                    "abstract": "In-context learning is a powerful emergent ability in transformer models. Prior work in mechanistic interpretability has identified a circuit element that may be critical for in-context learning -- the induction head (IH), which performs a match-and-copy operation. During training of large transformers on natural language data, IHs emerge around the same time as a notable phase change in the loss. Despite the robust evidence for IHs and this interesting coincidence with the phase change, relatively little is known about the diversity and emergence dynamics of IHs. Why is there more than one IH, and how are they dependent on each other? Why do IHs appear all of a sudden, and what are the subcircuits that enable them to emerge? We answer these questions by studying IH emergence dynamics in a controlled setting by training on synthetic data. In doing so, we develop and share a novel optogenetics-inspired causal framework for modifying activations throughout training. Using this framework, we delineate the diverse and additive nature of IHs. By clamping subsets of activations throughout training, we then identify three underlying subcircuits that interact to drive IH formation, yielding the phase change. Furthermore, these subcircuits shed light on data-dependent properties of formation, such as phase change timing, already showing the promise of this more in-depth understanding of subcircuits that need to\"go right\"for an induction head."
                },
                {
                    "title": "Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models",
                    "abstract": "We introduce methods for discovering and applying sparse feature circuits. These are causally implicated subnetworks of human-interpretable features for explaining language model behaviors. Circuits identified in prior work consist of polysemantic and difficult-to-interpret units like attention heads or neurons, rendering them unsuitable for many downstream applications. In contrast, sparse feature circuits enable detailed understanding of unanticipated mechanisms. Because they are based on fine-grained units, sparse feature circuits are useful for downstream tasks: We introduce SHIFT, where we improve the generalization of a classifier by ablating features that a human judges to be task-irrelevant. Finally, we demonstrate an entirely unsupervised and scalable interpretability pipeline by discovering thousands of sparse feature circuits for automatically discovered model behaviors."
                },
                {
                    "title": "Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanisms",
                    "abstract": "Many recent language model (LM) interpretability studies have adopted the circuits framework, which aims to find the minimal computational subgraph, or circuit, that explains LM behavior on a given task. Most studies determine which edges belong in a LM's circuit by performing causal interventions on each edge independently, but this scales poorly with model size. Edge attribution patching (EAP), gradient-based approximation to interventions, has emerged as a scalable but imperfect solution to this problem. In this paper, we introduce a new method - EAP with integrated gradients (EAP-IG) - that aims to better maintain a core property of circuits: faithfulness. A circuit is faithful if all model edges outside the circuit can be ablated without changing the model's performance on the task; faithfulness is what justifies studying circuits, rather than the full model. Our experiments demonstrate that circuits found using EAP are less faithful than those found using EAP-IG, even though both have high node overlap with circuits found previously using causal interventions. We conclude more generally that when using circuits to compare the mechanisms models use to solve tasks, faithfulness, not overlap, is what should be measured."
                },
                {
                    "title": "AtP*: An efficient and scalable method for localizing LLM behaviour to components",
                    "abstract": "Activation Patching is a method of directly computing causal attributions of behavior to model components. However, applying it exhaustively requires a sweep with cost scaling linearly in the number of model components, which can be prohibitively expensive for SoTA Large Language Models (LLMs). We investigate Attribution Patching (AtP), a fast gradient-based approximation to Activation Patching and find two classes of failure modes of AtP which lead to significant false negatives. We propose a variant of AtP called AtP*, with two changes to address these failure modes while retaining scalability. We present the first systematic study of AtP and alternative methods for faster activation patching and show that AtP significantly outperforms all other investigated methods, with AtP* providing further significant improvement. Finally, we provide a method to bound the probability of remaining false negatives of AtP* estimates."
                },
                {
                    "title": "Information Flow Routes: Automatically Interpreting Language Models at Scale",
                    "abstract": "Information flows by routes inside the network via mechanisms implemented in the model. These routes can be represented as graphs where nodes correspond to token representations and edges to operations inside the network. We automatically build these graphs in a top-down manner, for each prediction leaving only the most important nodes and edges. In contrast to the existing workflows relying on activation patching, we do this through attribution: this allows us to efficiently uncover existing circuits with just a single forward pass. Additionally, the applicability of our method is far beyond patching: we do not need a human to carefully design prediction templates, and we can extract information flow routes for any prediction (not just the ones among the allowed templates). As a result, we can talk about model behavior in general, for specific types of predictions, or different domains. We experiment with Llama 2 and show that the role of some attention heads is overall important, e.g. previous token heads and subword merging heads. Next, we find similarities in Llama 2 behavior when handling tokens of the same part of speech. Finally, we show that some model components can be specialized on domains such as coding or multilingual texts."
                },
                {
                    "title": "Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking",
                    "abstract": "Fine-tuning on generalized tasks such as instruction following, code generation, and mathematics has been shown to enhance language models' performance on a range of tasks. Nevertheless, explanations of how such fine-tuning influences the internal computations in these models remain elusive. We study how fine-tuning affects the internal mechanisms implemented in language models. As a case study, we explore the property of entity tracking, a crucial facet of language comprehension, where models fine-tuned on mathematics have substantial performance gains. We identify the mechanism that enables entity tracking and show that (i) in both the original model and its fine-tuned versions primarily the same circuit implements entity tracking. In fact, the entity tracking circuit of the original model on the fine-tuned versions performs better than the full original model. (ii) The circuits of all the models implement roughly the same functionality: Entity tracking is performed by tracking the position of the correct entity in both the original model and its fine-tuned versions. (iii) Performance boost in the fine-tuned models is primarily attributed to its improved ability to handle the augmented positional information. To uncover these findings, we employ: Patch Patching, DCM, which automatically detects model components responsible for specific semantics, and CMAP, a new approach for patching activations across models to reveal improved mechanisms. Our findings suggest that fine-tuning enhances, rather than fundamentally alters, the mechanistic operation of the model."
                },
                {
                    "title": "OLMo: Accelerating the Science of Language Models",
                    "abstract": "Language models (LMs) have become ubiquitous in both NLP research and in commercial product offerings. As their commercial importance has surged, the most powerful models have become closed off, gated behind proprietary interfaces, with important details of their training data, architectures, and development undisclosed. Given the importance of these details in scientifically studying these models, including their biases and potential risks, we believe it is essential for the research community to have access to powerful, truly open LMs. To this end, we have built OLMo, a competitive, truly Open Language Model, to enable the scientific study of language models. Unlike most prior efforts that have only released model weights and inference code, we release OLMo alongside open training data and training and evaluation code. We hope this release will empower the open research community and inspire a new wave of innovation."
                },
                {
                    "title": "Successor Heads: Recurring, Interpretable Attention Heads In The Wild",
                    "abstract": "In this work we present successor heads: attention heads that increment tokens with a natural ordering, such as numbers, months, and days. For example, successor heads increment 'Monday' into 'Tuesday'. We explain the successor head behavior with an approach rooted in mechanistic interpretability, the field that aims to explain how models complete tasks in human-understandable terms. Existing research in this area has found interpretable language model components in small toy models. However, results in toy models have not yet led to insights that explain the internals of frontier models and little is currently understood about the internal operations of large language models. In this paper, we analyze the behavior of successor heads in large language models (LLMs) and find that they implement abstract representations that are common to different architectures. They form in LLMs with as few as 31 million parameters, and at least as many as 12 billion parameters, such as GPT-2, Pythia, and Llama-2. We find a set of 'mod-10 features' that underlie how successor heads increment in LLMs across different architectures and sizes. We perform vector arithmetic with these features to edit head behavior and provide insights into numeric representations within LLMs. Additionally, we study the behavior of successor heads on natural language data, identifying interpretable polysemanticity in a Pythia successor head."
                },
                {
                    "title": "LLM360: Towards Fully Transparent Open-Source LLMs",
                    "abstract": "The recent surge in open-source Large Language Models (LLMs), such as LLaMA, Falcon, and Mistral, provides diverse options for AI practitioners and researchers. However, most LLMs have only released partial artifacts, such as the final model weights or inference code, and technical reports increasingly limit their scope to high-level design choices and surface statistics. These choices hinder progress in the field by degrading transparency into the training of LLMs and forcing teams to rediscover many details in the training process. We present LLM360, an initiative to fully open-source LLMs, which advocates for all training code and data, model checkpoints, and intermediate results to be made available to the community. The goal of LLM360 is to support open and collaborative AI research by making the end-to-end LLM training process transparent and reproducible by everyone. As a first step of LLM360, we release two 7B parameter LLMs pre-trained from scratch, Amber and CrystalCoder, including their training code, data, intermediate checkpoints, and analyses (at https://www.llm360.ai). We are committed to continually pushing the boundaries of LLMs through this open-source effort. More large-scale and stronger models are underway and will be released in the future."
                },
                {
                    "title": "Linear Representations of Sentiment in Large Language Models",
                    "abstract": "Sentiment is a pervasive feature in natural language text, yet it is an open question how sentiment is represented within Large Language Models (LLMs). In this study, we reveal that across a range of models, sentiment is represented linearly: a single direction in activation space mostly captures the feature across a range of tasks with one extreme for positive and the other for negative. Through causal interventions, we isolate this direction and show it is causally relevant in both toy tasks and real world datasets such as Stanford Sentiment Treebank. Through this case study we model a thorough investigation of what a single direction means on a broad data distribution. We further uncover the mechanisms that involve this direction, highlighting the roles of a small subset of attention heads and neurons. Finally, we discover a phenomenon which we term the summarization motif: sentiment is not solely represented on emotionally charged words, but is additionally summarized at intermediate positions without inherent sentiment, such as punctuation and names. We show that in Stanford Sentiment Treebank zero-shot classification, 76% of above-chance classification accuracy is lost when ablating the sentiment direction, nearly half of which (36%) is due to ablating the summarized sentiment direction exclusively at comma positions."
                },
                {
                    "title": "Identifying and Adapting Transformer-Components Responsible for Gender Bias in an English Language Model",
                    "abstract": "Language models (LMs) exhibit and amplify many types of undesirable biases learned from the training data, including gender bias. However, we lack tools for effectively and efficiently changing this behavior without hurting general language modeling performance. In this paper, we study three methods for identifying causal relations between LM components and particular output: causal mediation analysis, automated circuit discovery and our novel, efficient method called DiffMask+ based on differential masking. We apply the methods to GPT-2 small and the problem of gender bias, and use the discovered sets of components to perform parameter-efficient fine-tuning for bias mitigation. Our results show significant overlap in the identified components (despite huge differences in the computational requirements of the methods) as well as success in mitigating gender bias, with less damage to general language modeling compared to full model fine-tuning. However, our work also underscores the difficulty of defining and measuring bias, and the sensitivity of causal discovery procedures to dataset choice. We hope our work can contribute to more attention for dataset development, and lead to more effective mitigation strategies for other types of bias."
                },
                {
                    "title": "Attribution Patching Outperforms Automated Circuit Discovery",
                    "abstract": "Automated interpretability research has recently attracted attention as a potential research direction that could scale explanations of neural network behavior to large models. Existing automated circuit discovery work applies activation patching to identify subnetworks responsible for solving specific tasks (circuits). In this work, we show that a simple method based on attribution patching outperforms all existing methods while requiring just two forward passes and a backward pass. We apply a linear approximation to activation patching to estimate the importance of each edge in the computational subgraph. Using this approximation, we prune the least important edges of the network. We survey the performance and limitations of this method, finding that averaged over all tasks our method has greater AUC from circuit recovery than other methods."
                },
                {
                    "title": "Circuit Component Reuse Across Tasks in Transformer Language Models",
                    "abstract": "Recent work in mechanistic interpretability has shown that behaviors in language models can be successfully reverse-engineered through circuit analysis. A common criticism, however, is that each circuit is task-specific, and thus such analysis cannot contribute to understanding the models at a higher level. In this work, we present evidence that insights (both low-level findings about specific heads and higher-level findings about general algorithms) can indeed generalize across tasks. Specifically, we study the circuit discovered in Wang et al. (2022) for the Indirect Object Identification (IOI) task and 1.) show that it reproduces on a larger GPT2 model, and 2.) that it is mostly reused to solve a seemingly different task: Colored Objects (Ippolito&Callison-Burch, 2023). We provide evidence that the process underlying both tasks is functionally very similar, and contains about a 78% overlap in in-circuit attention heads. We further present a proof-of-concept intervention experiment, in which we adjust four attention heads in middle layers in order to 'repair' the Colored Objects circuit and make it behave like the IOI circuit. In doing so, we boost accuracy from 49.6% to 93.7% on the Colored Objects task and explain most sources of error. The intervention affects downstream attention heads in specific ways predicted by their interactions in the IOI circuit, indicating that this subcircuit behavior is invariant to the different task inputs. Overall, our results provide evidence that it may yet be possible to explain large language models' behavior in terms of a relatively small number of interpretable task-general algorithmic building blocks and computational components."
                },
                {
                    "title": "Copy Suppression: Comprehensively Understanding an Attention Head",
                    "abstract": "We present a single attention head in GPT-2 Small that has one main role across the entire training distribution. If components in earlier layers predict a certain token, and this token appears earlier in the context, the head suppresses it: we call this copy suppression. Attention Head 10.7 (L10H7) suppresses naive copying behavior which improves overall model calibration. This explains why multiple prior works studying certain narrow tasks found negative heads that systematically favored the wrong answer. We uncover the mechanism that the Negative Heads use for copy suppression with weights-based evidence and are able to explain 76.9% of the impact of L10H7 in GPT-2 Small. To the best of our knowledge, this is the most comprehensive description of the complete role of a component in a language model to date. One major effect of copy suppression is its role in self-repair. Self-repair refers to how ablating crucial model components results in downstream neural network parts compensating for this ablation. Copy suppression leads to self-repair: if an initial overconfident copier is ablated, then there is nothing to suppress. We show that self-repair is implemented by several mechanisms, one of which is copy suppression, which explains 39% of the behavior in a narrow task. Interactive visualisations of the copy suppression phenomena may be seen at our web app https://copy-suppression.streamlit.app/"
                },
                {
                    "title": "Sparse Autoencoders Find Highly Interpretable Features in Language Models",
                    "abstract": "One of the roadblocks to a better understanding of neural networks' internals is \\textit{polysemanticity}, where neurons appear to activate in multiple, semantically distinct contexts. Polysemanticity prevents us from identifying concise, human-understandable explanations for what neural networks are doing internally. One hypothesised cause of polysemanticity is \\textit{superposition}, where neural networks represent more features than they have neurons by assigning features to an overcomplete set of directions in activation space, rather than to individual neurons. Here, we attempt to identify those directions, using sparse autoencoders to reconstruct the internal activations of a language model. These autoencoders learn sets of sparsely activating features that are more interpretable and monosemantic than directions identified by alternative approaches, where interpretability is measured by automated methods. Moreover, we show that with our learned set of features, we can pinpoint the features that are causally responsible for counterfactual behaviour on the indirect object identification task \\citep{wang2022interpretability} to a finer degree than previous decompositions. This work indicates that it is possible to resolve superposition in language models using a scalable, unsupervised method. Our method may serve as a foundation for future mechanistic interpretability work, which we hope will enable greater model transparency and steerability."
                },
                {
                    "title": "Sudden Drops in the Loss: Syntax Acquisition, Phase Transitions, and Simplicity Bias in MLMs",
                    "abstract": "Most interpretability research in NLP focuses on understanding the behavior and features of a fully trained model. However, certain insights into model behavior may only be accessible by observing the trajectory of the training process. We present a case study of syntax acquisition in masked language models (MLMs) that demonstrates how analyzing the evolution of interpretable artifacts throughout training deepens our understanding of emergent behavior. In particular, we study Syntactic Attention Structure (SAS), a naturally emerging property of MLMs wherein specific Transformer heads tend to focus on specific syntactic relations. We identify a brief window in pretraining when models abruptly acquire SAS, concurrent with a steep drop in loss. This breakthrough precipitates the subsequent acquisition of linguistic capabilities. We then examine the causal role of SAS by manipulating SAS during training, and demonstrate that SAS is necessary for the development of grammatical capabilities. We further find that SAS competes with other beneficial traits during training, and that briefly suppressing SAS improves model quality. These findings offer an interpretation of a real-world example of both simplicity bias and breakthrough training dynamics."
                },
                {
                    "title": "Characterizing Learning Curves During Language Model Pre-Training: Learning, Forgetting, and Stability",
                    "abstract": "How do language models learn to make predictions during pre-training? To study this, we extract learning curves from five autoregressive English language model pre-training runs, for 1M unseen tokens in context. We observe that the language models generate short repetitive phrases before learning to generate longer and more coherent text. We also find that individual tokens often exhibit sudden increases or decreases in loss that are surprisingly consistent across pre-training runs. To better understand these fluctuations, we quantify the final surprisal, within-run variability, age of acquisition, forgettability, and cross-run variability of learning curves for individual tokens in context. More frequent tokens reach lower final surprisals, exhibit less variability within and across pre-training runs, are learned earlier, and are less likely to be\"forgotten\"during pre-training. Higher n-gram probabilities further accentuate these effects. Independent of the target token, shorter and more frequent contexts correlate with marginally more stable and quickly acquired predictions. Based on our results, we argue for the existence of sequential learning dependencies between different model capabilities, and we characterize language model learning as early n-gram learning before gradual refinement of tail n-gram predictions."
                },
                {
                    "title": "Latent State Models of Training Dynamics",
                    "abstract": "The impact of randomness on model training is poorly understood. How do differences in data order and initialization actually manifest in the model, such that some training runs outperform others or converge faster? Furthermore, how can we interpret the resulting training dynamics and the phase transitions that characterize different trajectories? To understand the effect of randomness on the dynamics and outcomes of neural network training, we train models multiple times with different random seeds and compute a variety of metrics throughout training, such as the $L_2$ norm, mean, and variance of the neural network's weights. We then fit a hidden Markov model (HMM) over the resulting sequences of metrics. The HMM represents training as a stochastic process of transitions between latent states, providing an intuitive overview of significant changes during training. Using our method, we produce a low-dimensional, discrete representation of training dynamics on grokking tasks, image classification, and masked language modeling. We use the HMM representation to study phase transitions and identify latent\"detour\"states that slow down convergence."
                },
                {
                    "title": "The Hydra Effect: Emergent Self-repair in Language Model Computations",
                    "abstract": "We investigate the internal structure of language model computations using causal analysis and demonstrate two motifs: (1) a form of adaptive computation where ablations of one attention layer of a language model cause another layer to compensate (which we term the Hydra effect) and (2) a counterbalancing function of late MLP layers that act to downregulate the maximum-likelihood token. Our ablation studies demonstrate that language model layers are typically relatively loosely coupled (ablations to one layer only affect a small number of downstream layers). Surprisingly, these effects occur even in language models trained without any form of dropout. We analyse these effects in the context of factual recall and consider their implications for circuit-level attribution in language models."
                },
                {
                    "title": "Finding Neurons in a Haystack: Case Studies with Sparse Probing",
                    "abstract": "Despite rapid adoption and deployment of large language models (LLMs), the internal computations of these models remain opaque and poorly understood. In this work, we seek to understand how high-level human-interpretable features are represented within the internal neuron activations of LLMs. We train $k$-sparse linear classifiers (probes) on these internal activations to predict the presence of features in the input; by varying the value of $k$ we study the sparsity of learned representations and how this varies with model scale. With $k=1$, we localize individual neurons which are highly relevant for a particular feature, and perform a number of case studies to illustrate general properties of LLMs. In particular, we show that early layers make use of sparse combinations of neurons to represent many features in superposition, that middle layers have seemingly dedicated neurons to represent higher-level contextual features, and that increasing scale causes representational sparsity to increase on average, but there are multiple types of scaling dynamics. In all, we probe for over 100 unique features comprising 10 different categories in 7 different models spanning 70 million to 6.9 billion parameters."
                },
                {
                    "title": "How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language model",
                    "abstract": "Pre-trained language models can be surprisingly adept at tasks they were not explicitly trained on, but how they implement these capabilities is poorly understood. In this paper, we investigate the basic mathematical abilities often acquired by pre-trained language models. Concretely, we use mechanistic interpretability techniques to explain the (limited) mathematical abilities of GPT-2 small. As a case study, we examine its ability to take in sentences such as\"The war lasted from the year 1732 to the year 17\", and predict valid two-digit end years (years>32). We first identify a circuit, a small subset of GPT-2 small's computational graph that computes this task's output. Then, we explain the role of each circuit component, showing that GPT-2 small's final multi-layer perceptrons boost the probability of end years greater than the start year. Finally, we find related tasks that activate our circuit. Our results suggest that GPT-2 small computes greater-than using a complex but general mechanism that activates across diverse contexts."
                },
                {
                    "title": "Towards Automated Circuit Discovery for Mechanistic Interpretability",
                    "abstract": "Through considerable effort and intuition, several recent works have reverse-engineered nontrivial behaviors of transformer models. This paper systematizes the mechanistic interpretability process they followed. First, researchers choose a metric and dataset that elicit the desired model behavior. Then, they apply activation patching to find which abstract neural network units are involved in the behavior. By varying the dataset, metric, and units under investigation, researchers can understand the functionality of each component. We automate one of the process' steps: to identify the circuit that implements the specified behavior in the model's computational graph. We propose several algorithms and reproduce previous interpretability results to validate them. For example, the ACDC algorithm rediscovered 5/5 of the component types in a circuit in GPT-2 Small that computes the Greater-Than operation. ACDC selected 68 of the 32,000 edges in GPT-2 Small, all of which were manually found by previous work. Our code is available at https://github.com/ArthurConmy/Automatic-Circuit-Discovery."
                },
                {
                    "title": "Localizing Model Behavior with Path Patching",
                    "abstract": "Localizing behaviors of neural networks to a subset of the network's components or a subset of interactions between components is a natural first step towards analyzing network mechanisms and possible failure modes. Existing work is often qualitative and ad-hoc, and there is no consensus on the appropriate way to evaluate localization claims. We introduce path patching, a technique for expressing and quantitatively testing a natural class of hypotheses expressing that behaviors are localized to a set of paths. We refine an explanation of induction heads, characterize a behavior of GPT-2, and open source a framework for efficiently running similar experiments."
                },
                {
                    "title": "Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling",
                    "abstract": "How do large language models (LLMs) develop and evolve over the course of training? How do these patterns change as models scale? To answer these questions, we introduce \\textit{Pythia}, a suite of 16 LLMs all trained on public data seen in the exact same order and ranging in size from 70M to 12B parameters. We provide public access to 154 checkpoints for each one of the 16 models, alongside tools to download and reconstruct their exact training dataloaders for further study. We intend \\textit{Pythia} to facilitate research in many areas, and we present several case studies including novel results in memorization, term frequency effects on few-shot performance, and reducing gender bias. We demonstrate that this highly controlled setup can be used to yield novel insights toward LLMs and their training dynamics. Trained models, analysis code, training code, and training data can be found at \\url{https://github.com/EleutherAI/pythia}."
                },
                {
                    "title": "Progress measures for grokking via mechanistic interpretability",
                    "abstract": "Neural networks often exhibit emergent behavior, where qualitatively new capabilities arise from scaling up the amount of parameters, training data, or training steps. One approach to understanding emergence is to find continuous \\textit{progress measures} that underlie the seemingly discontinuous qualitative changes. We argue that progress measures can be found via mechanistic interpretability: reverse-engineering learned behaviors into their individual components. As a case study, we investigate the recently-discovered phenomenon of ``grokking'' exhibited by small transformers trained on modular addition tasks. We fully reverse engineer the algorithm learned by these networks, which uses discrete Fourier transforms and trigonometric identities to convert addition to rotation about a circle. We confirm the algorithm by analyzing the activations and weights and by performing ablations in Fourier space. Based on this understanding, we define progress measures that allow us to study the dynamics of training and split training into three continuous phases: memorization, circuit formation, and cleanup. Our results show that grokking, rather than being a sudden shift, arises from the gradual amplification of structured mechanisms encoded in the weights, followed by the later removal of memorizing components."
                },
                {
                    "title": "Impossibility theorems for feature attribution",
                    "abstract": "Significance Machine learning models can learn complex patterns from data, but it is often difficult to understand why they make particular predictions. To tackle this problem, practitioners typically turn to feature attribution methods, which seek to attribute the model\u2019s behavior f(x) around an example x to particular features, or dimensions of x, that are most important for the prediction. In recent years, a new class of feature attribution methods\u2014namely, complete and linear methods\u2014has become popular. Our work shows that, unfortunately, such methods can be misleading: Complete and linear methods are provably less reliable than simpler methods at answering basic feature attribution questions. We provide impossibility results that highlight their failure cases and discuss how we might instead obtain reliable feature attributions."
                },
                {
                    "title": "Training Trajectories of Language Models Across Scales",
                    "abstract": "Scaling up language models has led to unprecedented performance gains, but little is understood about how the training dynamics change as models get larger. How do language models of different sizes learn during pre-training? Why do larger language models demonstrate more desirable behaviors? In this paper, we analyze the intermediate training checkpoints of differently sized OPT models (Zhang et al., 2022)\u2014from 125M to 175B parameters\u2014on next-token prediction, sequence-level generation and downstream tasks. We find that 1) at a given perplexity and independent of model sizes, a similar subset of training tokens see the most significant reduction in loss, with the rest stagnating or showing double-descent behavior (Nakkiran et al., 2020); 2) early in training, all models learn to reduce the perplexity of grammatical sequences that contain hallucinations, with small models halting at this suboptimal distribution and larger ones eventually learning to assign these sequences lower probabilities; and 3) perplexity is a strong predictor of in-context learning performance on 74 multiple-choice tasks from BIG-Bench, and this holds independent of the model size. Together, these results show that perplexity is more predictive of model behaviors than model size or training computation."
                },
                {
                    "title": "Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small",
                    "abstract": "Research in mechanistic interpretability seeks to explain behaviors of machine learning models in terms of their internal components. However, most previous work either focuses on simple behaviors in small models, or describes complicated behaviors in larger models with broad strokes. In this work, we bridge this gap by presenting an explanation for how GPT-2 small performs a natural language task called indirect object identification (IOI). Our explanation encompasses 26 attention heads grouped into 7 main classes, which we discovered using a combination of interpretability approaches relying on causal interventions. To our knowledge, this investigation is the largest end-to-end attempt at reverse-engineering a natural behavior\"in the wild\"in a language model. We evaluate the reliability of our explanation using three quantitative criteria--faithfulness, completeness and minimality. Though these criteria support our explanation, they also point to remaining gaps in our understanding. Our work provides evidence that a mechanistic understanding of large ML models is feasible, opening opportunities to scale our understanding to both larger models and more complex tasks."
                },
                {
                    "title": "In-context Learning and Induction Heads",
                    "abstract": "\u201cInduction heads\u201d are attention heads that implement a simple algorithm to complete token sequences like [A][B] ... [A] \u2192 [B]. In this work, we present preliminary and indirect evidence for a hypothesis that induction heads might constitute the mechanism for the majority of all \u201cincontext learning\u201d in large transformer models (i.e. decreasing loss at increasing token indices). We find that induction heads develop at precisely the same point as a sudden sharp increase in incontext learning ability, visible as a bump in the training loss. We present six complementary lines of evidence, arguing that induction heads may be the mechanistic source of general in-context learning in transformer models of any size. For small attention-only models, we present strong, causal evidence; for larger models with MLPs, we present correlational evidence. We recommend reading this paper as an HTML article. As Transformer generative models continue to scale and gain increasing real world use, addressing their associated safety problems becomes increasingly important. Mechanistic interpretability \u2013 attempting to reverse engineer the detailed computations performed by the model \u2013 offers one possible avenue for addressing these safety issues. If we can understand the internal structures that cause Transformer models to produce the outputs they do, then we may be able to address current safety problems more systematically, as well as anticipating safety problems in future more powerful models. [1, 2, 3, 4, 5]"
                },
                {
                    "title": "Probing for the Usage of Grammatical Number",
                    "abstract": "A central quest of probing is to uncover how pre-trained models encode a linguistic property within their representations. An encoding, however, might be spurious\u2014i.e., the model might not rely on it when making predictions. In this paper, we try to find an encoding that the model actually uses, introducing a usage-based probing setup. We first choose a behavioral task which cannot be solved without using the linguistic property. Then, we attempt to remove the property by intervening on the model\u2019s representations. We contend that, if an encoding is used by the model, its removal should harm the performance on the chosen behavioral task. As a case study, we focus on how BERT encodes grammatical number, and on how it uses this encoding to solve the number agreement task. Experimentally, we find that BERT relies on a linear encoding of grammatical number to produce the correct behavioral output. We also find that BERT uses a separate encoding of grammatical number for nouns and verbs. Finally, we identify in which layers information about grammatical number is transferred from a noun to its head verb."
                },
                {
                    "title": "Locating and Editing Factual Associations in GPT",
                    "abstract": "We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's factual predictions. This reveals a distinct set of steps in middle-layer feed-forward modules that mediate factual predictions while processing subject tokens. To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update specific factual associations using Rank-One Model Editing (ROME). We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another. Our results confirm an important role for mid-layer feed-forward modules in storing factual associations and suggest that direct manipulation of computational mechanisms may be a feasible approach for model editing. The code, dataset, visualizations, and an interactive demo notebook are available at https://rome.baulab.info/"
                },
                {
                    "title": "Word Acquisition in Neural Language Models",
                    "abstract": "We investigate how neural language models acquire individual words during training, extracting learning curves and ages of acquisition for over 600 words on the MacArthur-Bates Communicative Development Inventory (Fenson et al., 2007). Drawing on studies of word acquisition in children, we evaluate multiple predictors for words\u2019 ages of acquisition in LSTMs, BERT, and GPT-2. We find that the effects of concreteness, word length, and lexical class are pointedly different in children and language models, reinforcing the importance of interaction and sensorimotor experience in child language acquisition. Language models rely far more on word frequency than children, but, like children, they exhibit slower learning of words in longer utterances. Interestingly, models follow consistent patterns during training for both unidirectional and bidirectional models, and for both LSTM and Transformer architectures. Models predict based on unigram token frequencies early in training, before transitioning loosely to bigram probabilities, eventually converging on more nuanced predictions. These results shed light on the role of distributional learning mechanisms in children, while also providing insights for more human-like language acquisition in language models."
                },
                {
                    "title": "The Grammar-Learning Trajectories of Neural Language Models",
                    "abstract": "The learning trajectories of linguistic phenomena in humans provide insight into linguistic representation, beyond what can be gleaned from inspecting the behavior of an adult speaker. To apply a similar approach to analyze neural language models (NLM), it is first necessary to establish that different models are similar enough in the generalizations they make. In this paper, we show that NLMs with different initialization, architecture, and training data acquire linguistic phenomena in a similar order, despite their different end performance. These findings suggest that there is some mutual inductive bias that underlies these models\u2019 learning of linguistic phenomena. Taking inspiration from psycholinguistics, we argue that studying this inductive bias is an opportunity to study the linguistic representation implicit in NLMs.Leveraging these findings, we compare the relative performance on different phenomena at varying learning stages with simpler reference models. Results suggest that NLMs exhibit consistent \u201cdevelopmental\u201d stages. Moreover, we find the learning trajectory to be approximately one-dimensional: given an NLM with a certain overall performance, it is possible to predict what linguistic generalizations it has already acquired.Initial analysis of these stages presents phenomena clusters (notably morphological ones), whose performance progresses in unison, suggesting a potential link between the generalizations behind them."
                },
                {
                    "title": "The MultiBERTs: BERT Reproductions for Robustness Analysis",
                    "abstract": "Experiments with pre-trained models such as BERT are often based on a single checkpoint. While the conclusions drawn apply to the artifact tested in the experiment (i.e., the particular instance of the model), it is not always clear whether they hold for the more general procedure which includes the architecture, training data, initialization scheme, and loss function. Recent work has shown that repeating the pre-training process can lead to substantially different performance, suggesting that an alternate strategy is needed to make principled statements about procedures. To enable researchers to draw more robust conclusions, we introduce the MultiBERTs, a set of 25 BERT-Base checkpoints, trained with similar hyper-parameters as the original BERT model but differing in random weight initialization and shuffling of training data. We also define the Multi-Bootstrap, a non-parametric bootstrap method for statistical inference designed for settings where there are multiple pre-trained models and limited test data. To illustrate our approach, we present a case study of gender bias in coreference resolution, in which the Multi-Bootstrap lets us measure effects that may not be detected with a single checkpoint. We release our models and statistical library along with an additional set of 140 intermediate checkpoints captured during pre-training to facilitate research on learning dynamics."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Causal Analysis of Syntactic Agreement Mechanisms in Neural Language Models",
                    "abstract": "Targeted syntactic evaluations have demonstrated the ability of language models to perform subject-verb agreement given difficult contexts. To elucidate the mechanisms by which the models accomplish this behavior, this study applies causal mediation analysis to pre-trained neural language models. We investigate the magnitude of models\u2019 preferences for grammatical inflections, as well as whether neurons process subject-verb agreement similarly across sentences with different syntactic structures. We uncover similarities and differences across architectures and model sizes\u2014notably, that larger models do not necessarily learn stronger preferences. We also observe two distinct mechanisms for producing subject-verb agreement depending on the syntactic structure of the input sentence. Finally, we find that language models rely on similar sets of neurons when given sentences with similar syntactic structure."
                },
                {
                    "title": "Causal Abstractions of Neural Networks",
                    "abstract": "Structural analysis methods (e.g., probing and feature attribution) are increasingly important tools for neural network analysis. We propose a new structural analysis method grounded in a formal theory of causal abstraction that provides rich characterizations of model-internal representations and their roles in input/output behavior. In this method, neural representations are aligned with variables in interpretable causal models, and then interchange interventions are used to experimentally verify that the neural representations have the causal properties of their aligned variables. We apply this method in a case study to analyze neural models trained on Multiply Quantified Natural Language Inference (MQNLI) corpus, a highly complex NLI dataset that was constructed with a tree-structured natural logic causal model. We discover that a BERT-based model with state-of-the-art performance successfully realizes parts of the natural logic model's causal structure, whereas a simpler baseline model fails to show any such structure, demonstrating that BERT representations encode the compositional structure of MQNLI."
                },
                {
                    "title": "Refining Targeted Syntactic Evaluation of Language Models",
                    "abstract": "Targeted syntactic evaluation of subject-verb number agreement in English (TSE) evaluates language models\u2019 syntactic knowledge using hand-crafted minimal pairs of sentences that differ only in the main verb\u2019s conjugation. The method evaluates whether language models rate each grammatical sentence as more likely than its ungrammatical counterpart. We identify two distinct goals for TSE. First, evaluating the systematicity of a language model\u2019s syntactic knowledge: given a sentence, can it conjugate arbitrary verbs correctly? Second, evaluating a model\u2019s likely behavior: given a sentence, does the model concentrate its probability mass on correctly conjugated verbs, even if only on a subset of the possible verbs? We argue that current implementations of TSE do not directly capture either of these goals, and propose new metrics to capture each goal separately. Under our metrics, we find that TSE overestimates systematicity of language models, but that models score up to 40% better on verbs that they predict are likely in context."
                },
                {
                    "title": "Probing Classifiers: Promises, Shortcomings, and Advances",
                    "abstract": "Probing classifiers have emerged as one of the prominent methodologies for interpreting and analyzing deep neural network models of natural language processing. The basic idea is simple\u2014a classifier is trained to predict some linguistic property from a model\u2019s representations\u2014and has been used to examine a wide variety of models and properties. However, recent studies have demonstrated various methodological limitations of this approach. This squib critically reviews the probing classifiers framework, highlighting their promises, shortcomings, and advances."
                },
                {
                    "title": "Amnesic Probing: Behavioral Explanation with Amnesic Counterfactuals",
                    "abstract": "Abstract A growing body of work makes use of probing in order to investigate the working of neural models, often considered black boxes. Recently, an ongoing debate emerged surrounding the limitations of the probing paradigm. In this work, we point out the inability to infer behavioral conclusions from probing results, and offer an alternative method that focuses on how the information is being used, rather than on what information is encoded. Our method, Amnesic Probing, follows the intuition that the utility of a property for a given task can be assessed by measuring the influence of a causal intervention that removes it from the representation. Equipped with this new analysis tool, we can ask questions that were not possible before, for example, is part-of-speech information important for word prediction? We perform a series of analyses on BERT to answer these types of questions. Our findings demonstrate that conventional probing performance is not correlated to task importance, and we call for increased scrutiny of claims that draw behavioral or causal conclusions from probing results.1"
                },
                {
                    "title": "Causal Mediation Analysis for Interpreting Neural NLP: The Case of Gender Bias",
                    "abstract": "Common methods for interpreting neural models in natural language processing typically examine either their structure or their behavior, but not both. We propose a methodology grounded in the theory of causal mediation analysis for interpreting which parts of a model are causally implicated in its behavior. It enables us to analyze the mechanisms by which information flows from input to output through various model components, known as mediators. We apply this methodology to analyze gender bias in pre-trained Transformer language models. We study the role of individual neurons and attention heads in mediating gender bias across three datasets designed to gauge a model's sensitivity to gender bias. Our mediation analysis reveals that gender bias effects are (i) sparse, concentrated in a small part of the network; (ii) synergistic, amplified or repressed by different components; and (iii) decomposable into effects flowing directly from the input and indirectly through the mediators."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "PIQA: Reasoning about Physical Commonsense in Natural Language",
                    "abstract": "To apply eyeshadow without a brush, should I use a cotton swab or a toothpick? Questions requiring this kind of physical commonsense pose a challenge to today's natural language understanding systems. While recent pretrained models (such as BERT) have made progress on question answering over more abstract domains \u2013 such as news articles and encyclopedia entries, where text is plentiful \u2013 in more physical domains, text is inherently limited due to reporting bias. Can AI systems learn to reliably answer physical commonsense questions without experiencing the physical world?In this paper, we introduce the task of physical commonsense reasoning and a corresponding benchmark dataset Physical Interaction: Question Answering or PIQA. Though humans find the dataset easy (95% accuracy), large pretrained models struggle (\u223c75%). We provide analysis about the dimensions of knowledge that existing models lack, which offers significant opportunities for future research."
                },
                {
                    "title": "Sanity Checks for Saliency Maps",
                    "abstract": "Saliency methods have emerged as a popular tool to highlight features in an input deemed relevant for the prediction of a learned model. Several saliency methods have been proposed, often guided by visual appeal on image data. In this work, we propose an actionable methodology to evaluate what kinds of explanations a given method can and cannot provide. We find that reliance, solely, on visual assessment can be misleading. Through extensive experiments we show that some existing saliency methods are independent both of the model and of the data generating process. Consequently, methods that fail the proposed tests are inadequate for tasks that are sensitive to either data or model, such as, finding outliers in the data, explaining the relationship between inputs and outputs that the model learned, and debugging the model. We interpret our findings through an analogy with edge detection in images, a technique that requires neither training data nor model. Theory in the case of a linear model and a single-layer convolutional neural network supports our experimental findings."
                },
                {
                    "title": "Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge",
                    "abstract": "We present a new question set, text corpus, and baselines assembled to encourage AI research in advanced question answering. Together, these constitute the AI2 Reasoning Challenge (ARC), which requires far more powerful knowledge and reasoning than previous challenges such as SQuAD or SNLI. The ARC question set is partitioned into a Challenge Set and an Easy Set, where the Challenge Set contains only questions answered incorrectly by both a retrieval-based algorithm and a word co-occurence algorithm. The dataset contains only natural, grade-school science questions (authored for human tests), and is the largest public-domain set of this kind (7,787 questions). We test several baselines on the Challenge Set, including leading neural models from the SQuAD and SNLI tasks, and find that none are able to significantly outperform a random baseline, reflecting the difficult nature of this task. We are also releasing the ARC Corpus, a corpus of 14M science sentences relevant to the task, and implementations of the three neural baseline models tested. Can your model perform better? We pose ARC as a challenge to the community."
                },
                {
                    "title": "Crowdsourcing Multiple Choice Science Questions",
                    "abstract": "We present a novel method for obtaining high-quality, domain-targeted multiple choice questions from crowd workers. Generating these questions can be difficult without trading away originality, relevance or diversity in the answer options. Our method addresses these problems by leveraging a large corpus of domain-specific text and a small set of existing questions. It produces model suggestions for document selection and answer distractor choice which aid the human question generation process. With this method we have assembled SciQ, a dataset of 13.7K multiple choice science exam questions. We demonstrate that the method produces in-domain questions by providing an analysis of this new dataset and by showing that humans cannot distinguish the crowdsourced questions from original questions. When using SciQ as additional training data to existing questions, we observe accuracy improvements on real science exams."
                },
                {
                    "title": "Learning Important Features Through Propagating Activation Differences",
                    "abstract": "The purported \"black box\" nature of neural networks is a barrier to adoption in applications where interpretability is essential. Here we present DeepLIFT (Deep Learning Important FeaTures), a method for decomposing the output prediction of a neural network on a specific input by backpropagating the contributions of all neurons in the network to every feature of the input. DeepLIFT compares the activation of each neuron to its 'reference activation' and assigns contribution scores according to the difference. By optionally giving separate consideration to positive and negative contributions, DeepLIFT can also reveal dependencies which are missed by other approaches. Scores can be computed efficiently in a single backward pass. We apply DeepLIFT to models trained on MNIST and simulated genomic data, and show significant advantages over gradient-based methods. Video tutorial: http://goo.gl/qKb7pL, code: http://goo.gl/RM8jvH."
                },
                {
                    "title": "Axiomatic Attribution for Deep Networks",
                    "abstract": "We study the problem of attributing the prediction of a deep network to its input features, a problem previously studied by several other works. We identify two fundamental axioms\u2014 Sensitivity and Implementation Invariance that attribution methods ought to satisfy. We show that they are not satisfied by most known attribution methods, which we consider to be a fundamental weakness of those methods. We use the axioms to guide the design of a new attribution method called Integrated Gradients. Our method requires no modification to the original network and is extremely simple to implement; it just needs a few calls to the standard gradient operator. We apply this method to a couple of image models, a couple of text models and a chemistry model, demonstrating its ability to debug networks, to extract rules from a network, and to enable users to engage with models better."
                },
                {
                    "title": "Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies",
                    "abstract": "The success of long short-term memory (LSTM) neural networks in language processing is typically attributed to their ability to capture long-distance statistical regularities. Linguistic regularities are often sensitive to syntactic structure; can such dependencies be captured by LSTMs, which do not have explicit structural representations? We begin addressing this question using number agreement in English subject-verb dependencies. We probe the architecture\u2019s grammatical competence both using training objectives with an explicit grammatical target (number prediction, grammaticality judgments) and using language models. In the strongly supervised settings, the LSTM achieved very high overall accuracy (less than 1% errors), but errors increased when sequential and structural information conflicted. The frequency of such errors rose sharply in the language-modeling setting. We conclude that LSTMs can capture a non-trivial amount of grammatical structure given targeted supervision, but stronger architectures may be required to further reduce errors; furthermore, the language modeling signal is insufficient for capturing syntax-sensitive dependencies, and should be supplemented with more direct supervision if such dependencies need to be captured."
                },
                {
                    "title": "Interpretation",
                    "abstract": ". It is well known that the numbers (2 m )! (2 n )! /m ! n ! ( m + n )! are integers, but in general there is no known combinatorial interpretation for them. When m = 0 these numbers are the middle binomial coe\ufb03cients (cid:0) 2 nn (cid:1) , and when m = 1 they are twice the Catalan numbers. In this paper, we give combinatorial interpretations for these numbers when m = 2 or 3."
                },
                {
                    "title": "An Adversarial Winograd Schema Challenge at Scale",
                    "abstract": "The Winograd Schema Challenge (WSC), proposed by Levesque et al. (2011) as an alternative to the Turing Test, was originally designed as a pronoun resolution problem that cannot be solved based on statistical patterns in large text corpora. However, recent studies suggest that current WSC datasets, even when composed carefully by experts, are still prone to such biases that statistical methods can exploit. We introduce WINOGRANDE, a new collection of WSC problems that are adversarially constructed to be robust against spurious statistical biases. While the original WSC dataset provided only 273 instances, WINOGRANDE includes 43,985 instances, half of which are determined as adversarial. Key to our approach is a novel adversarial filtering algorithm AFLITE for systematic bias reduction, combined with a careful crowdsourcing design. Despite the significant increase in training data, the performance of existing stateof-the-art methods remains modest (61.6%) and contrasts with high human performance (90.8%) for the binary questions. In addition, WINOGRANDE allows us to use transfer learning for achieving new state-of-the-art results on the original WSC and related datasets. Finally, we discuss how biases lead to overestimating the true capabilities of machine commonsense."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow do circuits and their components emerge during the training of large language models (LLMs), and how consistent are these circuits across different training stages and model scales?\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding the emergence and consistency of circuits in LLMs during training is crucial for the research community as it can provide insights into the underlying mechanisms of these models. This knowledge can advance the field of mechanistic interpretability, enabling researchers to better understand how LLMs solve tasks and potentially leading to improvements in model design and training strategies. Furthermore, it could inform the development of more robust and reliable AI systems, enhancing their applicability in real-world scenarios.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem include the complexity of tracking circuit evolution during extensive training processes and the variability in model architectures and scales. Naive approaches may fail because they do not account for the dynamic nature of circuits as models are fine-tuned or continuously trained. Additionally, the lack of existing frameworks for analyzing circuits during training, particularly in large models, presents a significant technical obstacle that must be overcome to achieve meaningful insights.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on circuit analysis at the end of pre-training or on smaller models, leaving a gap in understanding how circuits evolve during training in larger, deployed models. Barriers such as limited methodologies for studying large-scale models and a lack of comprehensive datasets have hindered progress. Our approach differs by systematically analyzing circuits across various model sizes and training stages, providing a more holistic view of circuit dynamics that has not been explored in prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves studying circuits in models from the Pythia suite across 300 billion tokens, examining scales from 70 million to 2.8 billion parameters, and supplementing this with data from models up to 12 billion parameters. We will analyze the emergence of functional components and their consistency across training and model scales, using metrics such as task ability acquisition rates and circuit stability. We expect to find that circuit algorithms remain stable despite fluctuations in individual components, indicating a degree of generalizability across different training conditions and model sizes."
            }
        },
        "author_data": {
            "cf1e8280-52d5-4bf1-b8ba-e35a6abb591a": {
                "pk": "cf1e8280-52d5-4bf1-b8ba-e35a6abb591a",
                "name": "Curt Tigges",
                "collaborators": [
                    "Stella Biderman",
                    "Maxim Raginsky",
                    "Talia Ringer",
                    "Oskar John Hollinsworth",
                    "Atticus Geiger",
                    "Neel Nanda",
                    "Shizhuo Dylan Zhang",
                    "Dylan Zhang",
                    "Zory Zhang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Mechanistic Interpretability",
                    "Large Language Models",
                    "Structural Recursion"
                ],
                "publications": [
                    {
                        "title": "Linear Representations of Sentiment in Large Language Models",
                        "abstract": "Sentiment is a pervasive feature in natural language text, yet it is an open question how sentiment is represented within Large Language Models (LLMs). In this study, we reveal that across a range of models, sentiment is represented linearly: a single direction in activation space mostly captures the feature across a range of tasks with one extreme for positive and the other for negative. Through causal interventions, we isolate this direction and show it is causally relevant in both toy tasks and real world datasets such as Stanford Sentiment Treebank. Through this case study we model a thorough investigation of what a single direction means on a broad data distribution.   We further uncover the mechanisms that involve this direction, highlighting the roles of a small subset of attention heads and neurons. Finally, we discover a phenomenon which we term the summarization motif: sentiment is not solely represented on emotionally charged words, but is additionally summarized at intermediate positions without inherent sentiment, such as punctuation and names. We show that in Stanford Sentiment Treebank zero-shot classification, 76% of above-chance classification accuracy is lost when ablating the sentiment direction, nearly half of which (36%) is due to ablating the summarized sentiment direction exclusively at comma positions."
                    },
                    {
                        "title": "Can Transformers Learn to Solve Problems Recursively?",
                        "abstract": "Neural networks have in recent years shown promise for helping software engineers write programs and even formally verify them. While semantic information plays a crucial part in these processes, it remains unclear to what degree popular neural architectures like transformers are capable of modeling that information. This paper examines the behavior of neural networks learning algorithms relevant to programs and formal verification proofs through the lens of mechanistic interpretability, focusing in particular on structural recursion. Structural recursion is at the heart of tasks on which symbolic tools currently outperform neural models, like inferring semantic relations between datatypes and emulating program behavior. We evaluate the ability of transformer models to learn to emulate the behavior of structurally recursive functions from input-output examples. Our evaluation includes empirical and conceptual analyses of the limitations and capabilities of transformer models in approximating these functions, as well as reconstructions of the ``shortcut\" algorithms the model learns. By reconstructing these algorithms, we are able to correctly predict 91 percent of failure cases for one of the approximated functions. Our work provides a new foundation for understanding the behavior of neural networks that fail to solve the very tasks they are trained for."
                    },
                    {
                        "title": "Transformer-Based Models Are Not Yet Perfect At Learning to Emulate Structural Recursion",
                        "abstract": "This paper investigates the ability of transformer-based models to learn structural recursion from examples. Recursion is a universal concept in both natural and formal languages. Structural recursion is central to the programming language and formal mathematics tasks where symbolic tools currently excel beyond neural models, such as inferring semantic relations between datatypes and emulating program behavior. We introduce a general framework that nicely connects the abstract concepts of structural recursion in the programming language domain to concrete sequence modeling problems and learned models' behavior. The framework includes a representation that captures the general \\textit{syntax} of structural recursion, coupled with two different frameworks for understanding their \\textit{semantics} -- one that is more natural from a programming languages perspective and one that helps bridge that perspective with a mechanistic understanding of the underlying transformer architecture.   With our framework as a powerful conceptual tool, we identify different issues under various set-ups. The models trained to emulate recursive computations cannot fully capture the recursion yet instead fit short-cut algorithms and thus cannot solve certain edge cases that are under-represented in the training distribution. In addition, it is difficult for state-of-the-art large language models (LLMs) to mine recursive rules from in-context demonstrations. Meanwhile, these LLMs fail in interesting ways when emulating reduction (step-wise computation) of the recursive function."
                    }
                ]
            },
            "405be6fc-abca-4ce5-92a8-dd5571c284a7": {
                "pk": "405be6fc-abca-4ce5-92a8-dd5571c284a7",
                "name": "Michael Hanna",
                "collaborators": [
                    "Sandro Pezzelle",
                    "Yonatan Belinkov",
                    "Ollie Liu",
                    "Alexandre Variengien",
                    "Frank Wildenburg"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Interpretability",
                    "Machine Learning",
                    "Linguistics"
                ],
                "publications": [
                    {
                        "title": "How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language model",
                        "abstract": "Pre-trained language models can be surprisingly adept at tasks they were not explicitly trained on, but how they implement these capabilities is poorly understood. In this paper, we investigate the basic mathematical abilities often acquired by pre-trained language models. Concretely, we use mechanistic interpretability techniques to explain the (limited) mathematical abilities of GPT-2 small. As a case study, we examine its ability to take in sentences such as \"The war lasted from the year 1732 to the year 17\", and predict valid two-digit end years (years > 32). We first identify a circuit, a small subset of GPT-2 small's computational graph that computes this task's output. Then, we explain the role of each circuit component, showing that GPT-2 small's final multi-layer perceptrons boost the probability of end years greater than the start year. Finally, we find related tasks that activate our circuit. Our results suggest that GPT-2 small computes greater-than using a complex but general mechanism that activates across diverse contexts."
                    },
                    {
                        "title": "When Language Models Fall in Love: Animacy Processing in Transformer Language Models",
                        "abstract": "Animacy - whether an entity is alive and sentient - is fundamental to cognitive processing, impacting areas such as memory, vision, and language. However, animacy is not always expressed directly in language: in English it often manifests indirectly, in the form of selectional constraints on verbs and adjectives. This poses a potential issue for transformer language models (LMs): they often train only on text, and thus lack access to extralinguistic information from which humans learn about animacy. We ask: how does this impact LMs' animacy processing - do they still behave as humans do? We answer this question using open-source LMs. Like previous studies, we find that LMs behave much like humans when presented with entities whose animacy is typical. However, we also show that even when presented with stories about atypically animate entities, such as a peanut in love, LMs adapt: they treat these entities as animate, though they do not adapt as well as humans. Even when the context indicating atypical animacy is very short, LMs pick up on subtle clues and change their behavior. We conclude that despite the limited signal through which LMs can learn about animacy, they are indeed sensitive to the relevant lexical semantic nuances available in English."
                    },
                    {
                        "title": "Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanisms",
                        "abstract": "Many recent language model (LM) interpretability studies have adopted the circuits framework, which aims to find the minimal computational subgraph, or circuit, that explains LM behavior on a given task. Most studies determine which edges belong in a LM's circuit by performing causal interventions on each edge independently, but this scales poorly with model size. Edge attribution patching (EAP), gradient-based approximation to interventions, has emerged as a scalable but imperfect solution to this problem. In this paper, we introduce a new method - EAP with integrated gradients (EAP-IG) - that aims to better maintain a core property of circuits: faithfulness. A circuit is faithful if all model edges outside the circuit can be ablated without changing the model's performance on the task; faithfulness is what justifies studying circuits, rather than the full model. Our experiments demonstrate that circuits found using EAP are less faithful than those found using EAP-IG, even though both have high node overlap with circuits found previously using causal interventions. We conclude more generally that when using circuits to compare the mechanisms models use to solve tasks, faithfulness, not overlap, is what should be measured."
                    },
                    {
                        "title": "Do Pre-Trained Language Models Detect and Understand Semantic Underspecification? Ask the DUST!",
                        "abstract": "In everyday language use, speakers frequently utter and interpret sentences that are semantically underspecified, namely, whose content is insufficient to fully convey their message or interpret them univocally. For example, to interpret the underspecified sentence \"Don't spend too much\", which leaves implicit what (not) to spend, additional linguistic context or outside knowledge is needed. In this work, we propose a novel Dataset of semantically Underspecified Sentences grouped by Type (DUST) and use it to study whether pre-trained language models (LMs) correctly identify and interpret underspecified sentences. We find that newer LMs are reasonably able to identify underspecified sentences when explicitly prompted. However, interpreting them correctly is much harder for any LMs. Our experiments show that when interpreting underspecified sentences, LMs exhibit little uncertainty, contrary to what theoretical accounts of underspecification would predict. Overall, our study reveals limitations in current models' processing of sentence semantics and highlights the importance of using naturalistic data and communicative scenarios when evaluating LMs' language capabilities."
                    }
                ]
            },
            "8c976f8e-2a09-4f1f-9287-ae3d9cb11261": {
                "pk": "8c976f8e-2a09-4f1f-9287-ae3d9cb11261",
                "name": "Qinan Yu",
                "collaborators": [
                    "Ellie Pavlick",
                    "Jack Merullo",
                    "Albert Webson",
                    "Alyssa Marie Loo",
                    "Dashiell Stander",
                    "Honglu Fan",
                    "Stella Biderman",
                    "Ruochen Zhang",
                    "Matianyu Zang",
                    "Carsten Eickhoff"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Mechanistic Interpretability",
                    "Deep Learning",
                    "Compositional Semantics"
                ],
                "publications": [
                    {
                        "title": "Characterizing Mechanisms for Factual Recall in Language Models",
                        "abstract": "Language Models (LMs) often must integrate facts they memorized in pretraining with new information that appears in a given context. These two sources can disagree, causing competition within the model, and it is unclear how an LM will resolve the conflict. On a dataset that queries for knowledge of world capitals, we investigate both distributional and mechanistic determinants of LM behavior in such situations. Specifically, we measure the proportion of the time an LM will use a counterfactual prefix (e.g., \"The capital of Poland is London\") to overwrite what it learned in pretraining (\"Warsaw\"). On Pythia and GPT2, the training frequency of both the query country (\"Poland\") and the in-context city (\"London\") highly affect the models' likelihood of using the counterfactual. We then use head attribution to identify individual attention heads that either promote the memorized answer or the in-context answer in the logits. By scaling up or down the value vector of these heads, we can control the likelihood of using the in-context answer on new data. This method can increase the rate of generating the in-context answer to 88\\% of the time simply by scaling a single head at runtime. Our work contributes to a body of evidence showing that we can often localize model behaviors to specific components and provides a proof of concept for how future methods might control model behavior dynamically at runtime."
                    },
                    {
                        "title": "Are Language Models Worse than Humans at Following Prompts? It's Complicated",
                        "abstract": "Prompts have been the center of progress in advancing language models' zero-shot and few-shot performance. However, recent work finds that models can perform surprisingly well when given intentionally irrelevant or misleading prompts. Such results may be interpreted as evidence that model behavior is not \"human like\". In this study, we challenge a central assumption in such work: that humans would perform badly when given pathological instructions. We find that humans are able to reliably ignore irrelevant instructions and thus, like models, perform well on the underlying task despite an apparent lack of signal regarding the task they are being asked to do. However, when given deliberately misleading instructions, humans follow the instructions faithfully, whereas models do not. Our findings caution that future research should not idealize human behaviors as a monolith and should not train or evaluate models to mimic assumptions about these behaviors without first validating humans' behaviors empirically."
                    },
                    {
                        "title": "Grokking Group Multiplication with Cosets",
                        "abstract": "The complex and unpredictable nature of deep neural networks prevents their safe use in many high-stakes applications. There have been many techniques developed to interpret deep neural networks, but all have substantial limitations. Algorithmic tasks have proven to be a fruitful test ground for interpreting a neural network end-to-end. Building on previous work, we completely reverse engineer fully connected one-hidden layer networks that have ``grokked'' the arithmetic of the permutation groups $S_5$ and $S_6$. The models discover the true subgroup structure of the full group and converge on neural circuits that decompose the group arithmetic using the permutation group's subgroups. We relate how we reverse engineered the model's mechanisms and confirmed our theory was a faithful description of the circuit's functionality. We also draw attention to current challenges in conducting interpretability research by comparing our work to Chughtai et al. [4] which alleges to find a different algorithm for this same problem."
                    },
                    {
                        "title": "The Same But Different: Structural Similarities and Differences in Multilingual Language Modeling",
                        "abstract": "We employ new tools from mechanistic interpretability in order to ask whether the internal structure of large language models (LLMs) shows correspondence to the linguistic structures which underlie the languages on which they are trained. In particular, we ask (1) when two languages employ the same morphosyntactic processes, do LLMs handle them using shared internal circuitry? and (2) when two languages require different morphosyntactic processes, do LLMs handle them using different internal circuitry? Using English and Chinese multilingual and monolingual models, we analyze the internal circuitry involved in two tasks. We find evidence that models employ the same circuit to handle the same syntactic process independently of the language in which it occurs, and that this is the case even for monolingual models trained completely independently. Moreover, we show that multilingual models employ language-specific components (attention heads and feed-forward networks) when needed to handle linguistic processes (e.g., morphological marking) that only exist in some languages. Together, our results provide new insights into how LLMs trade off between exploiting common structures and preserving linguistic differences when tasked with modeling multiple languages simultaneously."
                    },
                    {
                        "title": "Does CLIP Bind Concepts? Probing Compositionality in Large Image Models",
                        "abstract": "Large-scale neural network models combining text and images have made incredible progress in recent years. However, it remains an open question to what extent such models encode compositional representations of the concepts over which they operate, such as correctly identifying \"red cube\" by reasoning over the constituents \"red\" and \"cube\". In this work, we focus on the ability of a large pretrained vision and language model (CLIP) to encode compositional concepts and to bind variables in a structure-sensitive way (e.g., differentiating \"cube behind sphere\" from \"sphere behind cube\"). To inspect the performance of CLIP, we compare several architectures from research on compositional distributional semantics models (CDSMs), a line of research that attempts to implement traditional compositional linguistic structures within embedding spaces. We benchmark them on three synthetic datasets - single-object, two-object, and relational - designed to test concept binding. We find that CLIP can compose concepts in a single-object setting, but in situations where concept binding is needed, performance drops dramatically. At the same time, CDSMs also perform poorly, with best performance at chance level."
                    }
                ]
            },
            "43cc05af-7602-472d-b9df-946b44472bc7": {
                "pk": "43cc05af-7602-472d-b9df-946b44472bc7",
                "name": "Stella Biderman",
                "collaborators": [
                    "Louis Castricato",
                    "Edward Raff",
                    "Leo Gao",
                    "Dashiell Stander",
                    "Honglu Fan",
                    "Nora Belrose",
                    "Kieran Bicheno",
                    "Walter J. Scheirer",
                    "Rogelio E. Cardona-Rivera",
                    "David Thue"
                ],
                "domain": [
                    "Machine Learning",
                    "Natural Language Processing",
                    "Computational Complexity",
                    "Interpretability"
                ],
                "publications": [
                    {
                        "title": "Magic: the Gathering is as Hard as Arithmetic",
                        "abstract": "Magic: the Gathering is a popular and famously complicated card game about magical combat. Recently, several authors including Chatterjee and Ibsen-Jensen (2016) and Churchill, Biderman, and Herrick (2019) have investigated the computational complexity of playing Magic optimally. In this paper we show that the ``mate-in-$n$'' problem for Magic is $\\Delta^0_n$-hard and that optimal play in two-player Magic is non-arithmetic in general. These results apply to how real Magic is played, can be achieved using standard-size tournament legal decks, and do not rely on stochasticity or hidden information. Our paper builds upon the construction that Churchill, Biderman, and Herrick (2019) used to show that this problem was at least as hard as the halting problem."
                    },
                    {
                        "title": "Datasheet for the Pile",
                        "abstract": "This datasheet describes the Pile, a 825 GiB dataset of human-authored text compiled by EleutherAI for use in large-scale language modeling. The Pile is comprised of 22 different text sources, ranging from original scrapes done for this project, to text data made available by the data owners, to third-party scrapes available online."
                    },
                    {
                        "title": "Fooling MOSS Detection with Pretrained Language Models",
                        "abstract": "As artificial intelligence (AI) technologies become increasingly powerful and prominent in society, their misuse is a growing concern. In educational settings, AI technologies could be used by students to cheat on assignments and exams. In this paper we explore whether transformers can be used to solve introductory level programming assignments while bypassing commonly used AI tools to detect similarities between pieces of software. We find that a student using GPT-J [Wang and Komatsuzaki, 2021] can complete introductory level programming assignments without triggering suspicion from MOSS [Aiken, 2000], a widely used software similarity and plagiarism detection tool. This holds despite the fact that GPT-J was not trained on the problems in question and is not provided with any examples to work from. We further find that the code written by GPT-J is diverse in structure, lacking any particular tells that future plagiarism detection techniques may use to try to identify algorithmically generated code. We conclude with a discussion of the ethical and educational implications of large language models and directions for future research."
                    },
                    {
                        "title": "Pitfalls in Machine Learning Research: Reexamining the Development Cycle",
                        "abstract": "Machine learning has the potential to fuel further advances in data science, but it is greatly hindered by an ad hoc design process, poor data hygiene, and a lack of statistical rigor in model evaluation. Recently, these issues have begun to attract more attention as they have caused public and embarrassing issues in research and development. Drawing from our experience as machine learning researchers, we follow the machine learning process from algorithm design to data collection to model evaluation, drawing attention to common pitfalls and providing practical recommendations for improvements. At each step, case studies are introduced to highlight how these pitfalls occur in practice, and where things could be improved."
                    },
                    {
                        "title": "Towards a Formal Model of Narratives",
                        "abstract": "In this paper, we propose the beginnings of a formal framework for modeling narrative \\textit{qua} narrative. Our framework affords the ability to discuss key qualities of stories and their communication, including the flow of information from a Narrator to a Reader, the evolution of a Reader's story model over time, and Reader uncertainty. We demonstrate its applicability to computational narratology by giving explicit algorithms for measuring the accuracy with which information was conveyed to the Reader and two novel measurements of story coherence."
                    },
                    {
                        "title": "Grokking Group Multiplication with Cosets",
                        "abstract": "The complex and unpredictable nature of deep neural networks prevents their safe use in many high-stakes applications. There have been many techniques developed to interpret deep neural networks, but all have substantial limitations. Algorithmic tasks have proven to be a fruitful test ground for interpreting a neural network end-to-end. Building on previous work, we completely reverse engineer fully connected one-hidden layer networks that have ``grokked'' the arithmetic of the permutation groups $S_5$ and $S_6$. The models discover the true subgroup structure of the full group and converge on neural circuits that decompose the group arithmetic using the permutation group's subgroups. We relate how we reverse engineered the model's mechanisms and confirmed our theory was a faithful description of the circuit's functionality. We also draw attention to current challenges in conducting interpretability research by comparing our work to Chughtai et al. [4] which alleges to find a different algorithm for this same problem."
                    },
                    {
                        "title": "Magic: The Gathering is Turing Complete",
                        "abstract": "$\\textit{Magic: The Gathering}$ is a popular and famously complicated trading card game about magical combat. In this paper we show that optimal play in real-world $\\textit{Magic}$ is at least as hard as the Halting Problem, solving a problem that has been open for a decade. To do this, we present a methodology for embedding an arbitrary Turing machine into a game of $\\textit{Magic}$ such that the first player is guaranteed to win the game if and only if the Turing machine halts. Our result applies to how real $\\textit{Magic}$ is played, can be achieved using standard-size tournament-legal decks, and does not rely on stochasticity or hidden information. Our result is also highly unusual in that all moves of both players are forced in the construction. This shows that even recognising who will win a game in which neither player has a non-trivial decision to make for the rest of the game is undecidable. We conclude with a discussion of the implications for a unified computational theory of games and remarks about the playability of such a board in a tournament setting."
                    },
                    {
                        "title": "EleutherAI: Going Beyond \"Open Science\" to \"Science in the Open\"",
                        "abstract": "Over the past two years, EleutherAI has established itself as a radically novel initiative aimed at both promoting open-source research and conducting research in a transparent, openly accessible and collaborative manner. EleutherAI's approach to research goes beyond transparency: by doing research entirely in public, anyone in the world can observe and contribute at every stage. Our work has been received positively and has resulted in several high-impact projects in Natural Language Processing and other fields. In this paper, we describe our experience doing public-facing machine learning research, the benefits we believe this approach brings, and the pitfalls we have encountered."
                    },
                    {
                        "title": "On the Sensitivity of k-Uniform Hypergraph Properties",
                        "abstract": "In this paper we present a graph property with sensitivity $\\Theta(\\sqrt{n})$, where $n={v\\choose2}$ is the number of variables, and generalize it to a $k$-uniform hypergraph property with sensitivity $\\Theta(\\sqrt{n})$, where $n={v\\choose k}$ is again the number of variables. This yields the smallest sensitivity yet achieved for a $k$-uniform hypergraph property. We then show that, for even $k$, there is a $k$-uniform hypergraph property that demonstrates a quadratic gap between sensitivity and block sensitivity. This matches the largest known gap found by Ambainis and Sun (2011) for Boolean functions in general, and is the first known example of such a gap for a graph or hypergraph property."
                    },
                    {
                        "title": "VQGAN-CLIP: Open Domain Image Generation and Editing with Natural Language Guidance",
                        "abstract": "Generating and editing images from open domain text prompts is a challenging task that heretofore has required expensive and specially trained models. We demonstrate a novel methodology for both tasks which is capable of producing images of high visual quality from text prompts of significant semantic complexity without any training by using a multimodal encoder to guide image generations. We demonstrate on a variety of tasks how using CLIP [37] to guide VQGAN [11] produces higher visual quality outputs than prior, less flexible approaches like DALL-E [38], GLIDE [33] and Open-Edit [24], despite not being trained for the tasks presented. Our code is available in a public repository."
                    },
                    {
                        "title": "Eliciting Latent Predictions from Transformers with the Tuned Lens",
                        "abstract": "We analyze transformers from the perspective of iterative inference, seeking to understand how model predictions are refined layer by layer. To do so, we train an affine probe for each block in a frozen pretrained model, making it possible to decode every hidden state into a distribution over the vocabulary. Our method, the \\emph{tuned lens}, is a refinement of the earlier ``logit lens'' technique, which yielded useful insights but is often brittle.   We test our method on various autoregressive language models with up to 20B parameters, showing it to be more predictive, reliable and unbiased than the logit lens. With causal experiments, we show the tuned lens uses similar features to the model itself. We also find the trajectory of latent predictions can be used to detect malicious inputs with high accuracy. All code needed to reproduce our results can be found at https://github.com/AlignmentResearch/tuned-lens."
                    },
                    {
                        "title": "Cut the CARP: Fishing for zero-shot story evaluation",
                        "abstract": "Recent advances in large-scale language models (Raffel et al., 2019; Brown et al., 2020) have brought significant qualitative and quantitative improvements in machine-driven text generation. Despite this, generation and evaluation of machine-generated narrative text remains a challenging problem. Objective evaluation of computationally-generated stories may be prohibitively expensive, require meticulously annotated datasets, or may not adequately measure the logical coherence of a generated story's narratological structure.   Informed by recent advances in contrastive learning (Radford et al., 2021), we present Contrastive Authoring and Reviewing Pairing (CARP): a scalable, efficient method for performing qualitatively superior, zero-shot evaluation of stories. We show a strong correlation between human evaluation of stories and those of CARP. Model outputs more significantly correlate with corresponding human input than those language-model based methods which utilize finetuning or prompt engineering approaches. We also present and analyze the Story-Critique Dataset, a new corpora composed of 1.3 million aligned story-critique pairs derived from over 80,000 stories. We expect this corpus to be of interest to NLP researchers."
                    },
                    {
                        "title": "The Goldilocks of Pragmatic Understanding: Fine-Tuning Strategy Matters for Implicature Resolution by LLMs",
                        "abstract": "Despite widespread use of LLMs as conversational agents, evaluations of performance fail to capture a crucial aspect of communication: interpreting language in context -- incorporating its pragmatics. Humans interpret language using beliefs and prior knowledge about the world. For example, we intuitively understand the response \"I wore gloves\" to the question \"Did you leave fingerprints?\" as meaning \"No\". To investigate whether LLMs have the ability to make this type of inference, known as an implicature, we design a simple task and evaluate four categories of widely used state-of-the-art models. We find that, despite only evaluating on utterances that require a binary inference (yes or no), models in three of these categories perform close to random. However, LLMs instruction-tuned at the example-level perform significantly better. These results suggest that certain fine-tuning strategies are far better at inducing pragmatic understanding in models. We present our findings as the starting point for further research into evaluating how LLMs interpret language in context and to drive the development of more pragmatic and useful models of human discourse."
                    },
                    {
                        "title": "LEACE: Perfect linear concept erasure in closed form",
                        "abstract": "Concept erasure aims to remove specified features from a representation. It can improve fairness (e.g. preventing a classifier from using gender or race) and interpretability (e.g. removing a concept to observe changes in model behavior). We introduce LEAst-squares Concept Erasure (LEACE), a closed-form method which provably prevents all linear classifiers from detecting a concept while changing the representation as little as possible, as measured by a broad class of norms. We apply LEACE to large language models with a novel procedure called \"concept scrubbing,\" which erases target concept information from every layer in the network. We demonstrate our method on two tasks: measuring the reliance of language models on part-of-speech information, and reducing gender bias in BERT embeddings. Code is available at https://github.com/EleutherAI/concept-erasure."
                    },
                    {
                        "title": "Stay on topic with Classifier-Free Guidance",
                        "abstract": "Classifier-Free Guidance (CFG) has recently emerged in text-to-image generation as a lightweight technique to encourage prompt-adherence in generations. In this work, we demonstrate that CFG can be used broadly as an inference-time technique in pure language modeling. We show that CFG (1) improves the performance of Pythia, GPT-2 and LLaMA-family models across an array of tasks: Q\\&A, reasoning, code generation, and machine translation, achieving SOTA on LAMBADA with LLaMA-7B over PaLM-540B; (2) brings improvements equivalent to a model with twice the parameter-count; (3) can stack alongside other inference-time methods like Chain-of-Thought and Self-Consistency, yielding further improvements in difficult tasks; (4) can be used to increase the faithfulness and coherence of assistants in challenging form-driven and content-driven prompts: in a human evaluation we show a 75\\% preference for GPT4All using CFG over baseline."
                    },
                    {
                        "title": "Suppressing Pink Elephants with Direct Principle Feedback",
                        "abstract": "Existing methods for controlling language models, such as RLHF and Constitutional AI, involve determining which LLM behaviors are desirable and training them into a language model. However, in many cases, it is desirable for LLMs to be controllable at inference time, so that they can be used in multiple contexts with diverse needs. We illustrate this with the Pink Elephant Problem: instructing an LLM to avoid discussing a certain entity (a ``Pink Elephant''), and instead discuss a preferred entity (``Grey Elephant''). We apply a novel simplification of Constitutional AI, Direct Principle Feedback, which skips the ranking of responses and uses DPO directly on critiques and revisions. Our results show that after DPF fine-tuning on our synthetic Pink Elephants dataset, our 13B fine-tuned LLaMA 2 model significantly outperforms Llama-2-13B-Chat and a prompted baseline, and performs as well as GPT-4 in on our curated test set assessing the Pink Elephant Problem."
                    },
                    {
                        "title": "Holographic Global Convolutional Networks for Long-Range Prediction Tasks in Malware Detection",
                        "abstract": "Malware detection is an interesting and valuable domain to work in because it has significant real-world impact and unique machine-learning challenges. We investigate existing long-range techniques and benchmarks and find that they're not very suitable in this problem area. In this paper, we introduce Holographic Global Convolutional Networks (HGConv) that utilize the properties of Holographic Reduced Representations (HRR) to encode and decode features from sequence elements. Unlike other global convolutional methods, our method does not require any intricate kernel computation or crafted kernel design. HGConv kernels are defined as simple parameters learned through backpropagation. The proposed method has achieved new SOTA results on Microsoft Malware Classification Challenge, Drebin, and EMBER malware benchmarks. With log-linear complexity in sequence length, the empirical results demonstrate substantially faster run-time by HGConv compared to other methods achieving far more efficient scaling even with sequence length $\\geq 100,000$."
                    },
                    {
                        "title": "Can Transformers Learn to Solve Problems Recursively?",
                        "abstract": "Neural networks have in recent years shown promise for helping software engineers write programs and even formally verify them. While semantic information plays a crucial part in these processes, it remains unclear to what degree popular neural architectures like transformers are capable of modeling that information. This paper examines the behavior of neural networks learning algorithms relevant to programs and formal verification proofs through the lens of mechanistic interpretability, focusing in particular on structural recursion. Structural recursion is at the heart of tasks on which symbolic tools currently outperform neural models, like inferring semantic relations between datatypes and emulating program behavior. We evaluate the ability of transformer models to learn to emulate the behavior of structurally recursive functions from input-output examples. Our evaluation includes empirical and conceptual analyses of the limitations and capabilities of transformer models in approximating these functions, as well as reconstructions of the ``shortcut\" algorithms the model learns. By reconstructing these algorithms, we are able to correctly predict 91 percent of failure cases for one of the approximated functions. Our work provides a new foundation for understanding the behavior of neural networks that fail to solve the very tasks they are trained for."
                    }
                ]
            }
        }
    },
    "2310.04475": {
        "paper_data": {
            "title": "Demystifying Embedding Spaces using Large Language Models",
            "url": "http://arxiv.org/abs/2310.04475v2",
            "arxiv_id": "2310.04475",
            "authors": [
                "Guy Tennenholtz",
                "Yinlam Chow",
                "Chih-Wei Hsu",
                "Jihwan Jeong",
                "Lior Shani",
                "Azamat Tulepbergenov",
                "Deepak Ramachandran",
                "Martin Mladenov",
                "Craig Boutilier"
            ],
            "abstract": "Embeddings have become a pivotal means to represent complex, multi-faceted information about entities, concepts, and relationships in a condensed and useful format. Nevertheless, they often preclude direct interpretation. While downstream tasks make use of these compressed representations, meaningful interpretation usually requires visualization using dimensionality reduction or specialized machine learning interpretability methods. This paper addresses the challenge of making such embeddings more interpretable and broadly useful, by employing Large Language Models (LLMs) to directly interact with embeddings -- transforming abstract vectors into understandable narratives. By injecting embeddings into LLMs, we enable querying and exploration of complex embedding data. We demonstrate our approach on a variety of diverse tasks, including: enhancing concept activation vectors (CAVs), communicating novel embedded entities, and decoding user preferences in recommender systems. Our work couples the immense information potential of embeddings with the interpretative power of LLMs.",
            "introduction": "   1 Introduction  The success of deep learning has brought forth a paradigm shift in knowledge representation through the concept of embeddings\u2014dense vector representations that capture high-dimensional information about entities, concepts, or relationships in a compact and useful format. Embeddings are ubiquitous, finding application in natural language processing (Mikolov et\u00a0al., 2013; Devlin et\u00a0al., 2018), recommender systems (Rendle, 2010), protein sequence modeling (Rives et\u00a0al., 2021), and more. These embeddings are invaluable, capturing nuanced relationships and semantic structure in data which traditional  machine learning (ML) approaches often miss. Nevertheless, understanding these abstract representations remains challenging.   By design, the structure and underlying information carried by an embedding is heavily adapted to the idiosyncrasies of the downstream task, posing a substantial challenge to its interpretation and manipulation. Previous work on ML interpretability offers various task-independent means to interpret embeddings, including dimensionality reduction techniques (e.g., t-SNE (Van\u00a0der Maaten & Hinton, 2008), UMAP (McInnes et\u00a0al., 2018)) or concept activation vectors (CAVs) (Kim et\u00a0al., 2018). While very useful, these techniques are fairly narrow in scope.   As an alternative to such interpretability methods, suppose one could engage with embeddings using natural language to query information not directly expressed by the name or description of the underlying entity/concept. Perhaps one could even extract information from the embedding representations of non-existent or hypothetical entities. Consider, for instance, an embedding space representing items from an online commerce site, trained using a large corpus of user ratings or purchases, reviews, and other multi-faceted data sources. The embedding representation of an item may implicitly embody intricate details about its quality, usability, design, customer satisfaction, etc. Moreover, suppose one wanted to understand the properties of a hypothetical item at a specific point in the embedding space, say, if we predicted a potential market for such an item. Since no such item (or description) exists, asking a conventional language model to describe this embedding point would be futile. However, a  large language model (LLM) trained to interpret the embedding representation itself could handle such a task (see Fig.\u00a01 for an illustration).   This paper introduces a novel framework to interpret domain embeddings111We use the term \u201cdomain embedding\u201d to differentiate the external embeddings to be interpreted from those captured by an LLM at various layers of the network. by leveraging the power of LLMs (Devlin et\u00a0al., 2018; Liu et\u00a0al., 2021; Google et\u00a0al., 2023). Our method seamlessly introduces embeddings into LLMs by training adapter layers to map domain embedding vectors into the token-level embedding space of an LLM, which in turn allows one to treat these vectors as token-level encodings of the entities or concepts they represent. We train an LLM on a collection of tasks designed to facilitate the robust, generalizable interpretation of vectors in the domain embedding space. Our approach allows us to engage in a direct \u201cdialogue\u201d about these vectors, to query the LLM with intricate embedding data, and tease out narratives and insights from these dense vectors.   Our contributions are as follows. First, we formulate the problem of interpreting embeddings using LLMs. We then propose the  Embedding Language Model (ELM), a novel language model framework which, using trained adapters, can accept",
            "references": [
                {
                    "title": "RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedback",
                    "abstract": "Reinforcement learning from human feedback (RLHF) has proven effective in aligning large language models (LLMs) with human preferences, but gathering high-quality preference labels is expensive. RL from AI Feedback (RLAIF), introduced in Bai et al., offers a promising alternative that trains the reward model (RM) on preferences generated by an off-the-shelf LLM. Across the tasks of summarization, helpful dialogue generation, and harmless dialogue generation, we show that RLAIF achieves comparable performance to RLHF. Furthermore, we take a step towards\"self-improvement\"by demonstrating that RLAIF can outperform a supervised fine-tuned baseline even when the AI labeler is the same size as the policy, or even the exact same checkpoint as the initial policy. Finally, we introduce direct-RLAIF (d-RLAIF) - a technique that circumvents RM training by obtaining rewards directly from an off-the-shelf LLM during RL, which achieves superior performance to canonical RLAIF. Our results suggest that RLAIF can achieve performance on-par with using human feedback, offering a potential solution to the scalability limitations of RLHF."
                },
                {
                    "title": "Inseq: An Interpretability Toolkit for Sequence Generation Models",
                    "abstract": "Past work in natural language processing interpretability focused mainly on popular classification tasks while largely overlooking generation settings, partly due to a lack of dedicated tools. In this work, we introduce Inseq, a Python library to democratize access to interpretability analyses of sequence generation models. Inseq enables intuitive and optimized extraction of models\u2019 internal information and feature importance scores for popular decoder-only and encoder-decoder Transformers architectures. We showcase its potential by adopting it to highlight gender biases in machine translation models and locate factual knowledge inside GPT-2. Thanks to its extensible interface supporting cutting-edge techniques such as contrastive feature attribution, Inseq can drive future advances in explainable natural language generation, centralizing good practices and enabling fair and reproducible model evaluations."
                },
                {
                    "title": "Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networks",
                    "abstract": "The last decade of machine learning has seen drastic increases in scale and capabilities. Deep neural networks (DNNs) are increasingly being deployed in the real world. However, they are difficult to analyze, raising concerns about using them without a rigorous understanding of how they function. Effective tools for interpreting them will be important for building more trustworthy AI by helping to identify problems, fix bugs, and improve basic understanding. In particular, \u201cinner\u201d interpretability techniques, which focus on explaining the internal components of DNNs, are well-suited for developing a mechanistic understanding, guiding manual modifications, and reverse engineering solutions. Much recent work has focused on DNN interpretability, and rapid progress has thus far made a thorough systematization of methods difficult. In this survey, we review over 300 works with a focus on inner interpretability tools. We introduce a taxonomy that classifies methods by what part of the network they help to explain (weights, neurons, subnetworks, or latent representations) and whether they are implemented during (intrinsic) or after (post hoc) training. To our knowledge, we are also the first to survey a number of connections between interpretability research and work in adversarial robustness, continual learning, modularity, network compression, and studying the human visual system. We discuss key challenges and argue that the status quo in interpretability research is largely unproductive. Finally, we highlight the importance of future work that emphasizes diagnostics, debugging, adversaries, and benchmarking in order to make interpretability tools more useful to engineers in practical applications."
                },
                {
                    "title": "Scaling Autoregressive Models for Content-Rich Text-to-Image Generation",
                    "abstract": "We present the Pathways Autoregressive Text-to-Image (Parti) model, which generates high-fidelity photorealistic images and supports content-rich synthesis involving complex compositions and world knowledge. Parti treats text-to-image generation as a sequence-to-sequence modeling problem, akin to machine translation, with sequences of image tokens as the target outputs rather than text tokens in another language. This strategy can naturally tap into the rich body of prior work on large language models, which have seen continued advances in capabilities and performance through scaling data and model sizes. Our approach is simple: First, Parti uses a Transformer-based image tokenizer, ViT-VQGAN, to encode images as sequences of discrete tokens. Second, we achieve consistent quality improvements by scaling the encoder-decoder Transformer model up to 20B parameters, with a new state-of-the-art zero-shot FID score of 7.23 and finetuned FID score of 3.22 on MS-COCO. Our detailed analysis on Localized Narratives as well as PartiPrompts (P2), a new holistic benchmark of over 1600 English prompts, demonstrate the effectiveness of Parti across a wide variety of categories and difficulty aspects. We also explore and highlight limitations of our models in order to define and exemplify key areas of focus for further improvements. See https://parti.research.google/ for high-resolution images."
                },
                {
                    "title": "Discovering Personalized Semantics for Soft Attributes in Recommender Systems using Concept Activation Vectors",
                    "abstract": "Interactive recommender systems (RSs) allow users to express intent, preferences and contexts in a rich fashion, often using natural language. One challenge in using such feedback is inferring a user\u2019s semantic intent from the open-ended terms used to describe an item, and using it to refine recommendation results. Leveraging concept activation vectors (CAVs) [21], we develop a framework to learn a representation that captures the semantics of such attributes and connects them to user preferences and behaviors in RSs. A novel feature of our approach is its ability to distinguish objective and subjective attributes and associate different senses with different users. Using synthetic and real-world datasets, we show that our CAV representation accurately interprets users\u2019 subjective semantics, and can improve recommendations via interactive critiquing."
                },
                {
                    "title": "Large Dual Encoders Are Generalizable Retrievers",
                    "abstract": "It has been shown that dual encoders trained on one domain often fail to generalize to other domains for retrieval tasks. One widespread belief is that the bottleneck layer of a dual encoder, where the final score is simply a dot-product between a query vector and a passage vector, is too limited compared to models with fine-grained interactions between the query and the passage. In this paper, we challenge this belief by scaling up the size of the dual encoder model while keeping the bottleneck layer as a single dot-product with a fixed size. With multi-stage training, scaling up the model size brings significant improvement on a variety of retrieval tasks, especially for out-of-domain generalization. We further analyze the impact of the bottleneck layer and demonstrate diminishing improvement when scaling up the embedding size. Experimental results show that our dual encoders, Generalizable T5-based dense Retrievers (GTR), outperform previous sparse and dense retrievers on the BEIR dataset significantly. Most surprisingly, our ablation study finds that GTR is very data efficient, as it only needs 10% of MS Marco supervised data to match the out-of-domain performance of using all supervised data."
                },
                {
                    "title": "The specious art of single-cell genomics",
                    "abstract": "Dimensionality reduction is standard practice for filtering noise and identifying relevant features in large-scale data analyses. In biology, single-cell genomics studies typically begin with reduction to two or three dimensions to produce \u2018all-in-one\u2019 visuals of the data that are amenable to the human eye, and these are subsequently used for qualitative and quantitative exploratory analysis. However, there is little theoretical support for this practice, and we show that extreme dimension reduction, from hundreds or thousands of dimensions to two, inevitably induces significant distortion of high-dimensional datasets. We therefore examine the practical implications of low-dimensional embedding of single-cell data, and find that extensive distortions and inconsistent practices make such embeddings counter-productive for exploratory, biological analyses. In lieu of this, we discuss alternative approaches for conducting targeted embedding and feature exploration, to enable hypothesis-driven biological discovery."
                },
                {
                    "title": "Sentence-T5: Scalable Sentence Encoders from Pre-trained Text-to-Text Models",
                    "abstract": "We provide the first exploration of sentence embeddings from text-to-text transformers (T5) including the effects of scaling up sentence encoders to 11B parameters. Sentence embeddings are broadly useful for language processing tasks. While T5 achieves impressive performance on language tasks, it is unclear how to produce sentence embeddings from encoder-decoder models. We investigate three methods to construct Sentence-T5 (ST5) models: two utilize only the T5 encoder and one using the full T5 encoder-decoder. We establish a new sentence representation transfer benchmark, SentGLUE, which extends the SentEval toolkit to nine tasks from the GLUE benchmark. Our encoder-only models outperform the previous best models on both SentEval and SentGLUE transfer tasks, including semantic textual similarity (STS). Scaling up ST5 from millions to billions of parameters shown to consistently improve performance. Finally, our encoder-decoder method achieves a new state-of-the-art on STS when using sentence embeddings."
                },
                {
                    "title": "Multimodal Few-Shot Learning with Frozen Language Models",
                    "abstract": "When trained at sufficient scale, auto-regressive language models exhibit the notable ability to learn a new language task after being prompted with just a few examples. Here, we present a simple, yet effective, approach for transferring this few-shot learning ability to a multimodal setting (vision and language). Using aligned image and caption data, we train a vision encoder to represent each image as a sequence of continuous embeddings, such that a pre-trained, frozen language model prompted with this prefix generates the appropriate caption. The resulting system is a multimodal few-shot learner, with the surprising ability to learn a variety of new tasks when conditioned on examples, represented as a sequence of multiple interleaved image and text embeddings. We demonstrate that it can rapidly learn words for new objects and novel visual categories, do visual question-answering with only a handful of examples, and make use of outside knowledge, by measuring a single model on a variety of established and new benchmarks."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "On Interpretation and Measurement of Soft Attributes for Recommendation",
                    "abstract": "We address how to robustly interpret natural language refinements (or critiques) in recommender systems. In particular, in human-human recommendation settings people frequently use soft attributes to express preferences about items, including concepts like the originality of a movie plot, the noisiness of a venue, or the complexity of a recipe. While binary tagging is extensively studied in the context of recommender systems, soft attributes often involve subjective and contextual aspects, which cannot be captured reliably in this way, nor be represented as objective binary truth in a knowledge base. This also adds important considerations when measuring soft attribute ranking. We propose a more natural representation as personalized relative statements, rather than as absolute item properties. We present novel data collection techniques and evaluation approaches, and a new public dataset. We also propose a set of scoring approaches, from unsupervised to weakly supervised to fully supervised, as a step towards interpreting and acting upon soft attribute based critiques."
                },
                {
                    "title": "The Power of Scale for Parameter-Efficient Prompt Tuning",
                    "abstract": "In this work, we explore \u201cprompt tuning,\u201d a simple yet effective mechanism for learning \u201csoft prompts\u201d to condition frozen language models to perform specific downstream tasks. Unlike the discrete text prompts used by GPT-3, soft prompts are learned through backpropagation and can be tuned to incorporate signals from any number of labeled examples. Our end-to-end learned approach outperforms GPT-3\u2019s few-shot learning by a large margin. More remarkably, through ablations on model size using T5, we show that prompt tuning becomes more competitive with scale: as models exceed billions of parameters, our method \u201ccloses the gap\u201d and matches the strong performance of model tuning (where all model weights are tuned). This finding is especially relevant because large models are costly to share and serve and the ability to reuse one frozen model for multiple downstream tasks can ease this burden. Our method can be seen as a simplification of the recently proposed \u201cprefix tuning\u201d of Li and Liang (2021) and we provide a comparison to this and other similar approaches. Finally, we show that conditioning a frozen model with soft prompts confers benefits in robustness to domain transfer and enables efficient \u201cprompt ensembling.\u201d We release code and model checkpoints to reproduce our experiments."
                },
                {
                    "title": "Unifying Vision-and-Language Tasks via Text Generation",
                    "abstract": "Existing methods for vision-and-language learning typically require designing task-specific architectures and objectives for each task. For example, a multi-label answer classifier for visual question answering, a region scorer for referring expression comprehension, and a language decoder for image captioning, etc. To alleviate these hassles, in this work, we propose a unified framework that learns different tasks in a single architecture with the same language modeling objective, i.e., multimodal conditional text generation, where our models learn to generate labels in text based on the visual and textual inputs. On 7 popular vision-and-language benchmarks, including visual question answering, referring expression comprehension, visual commonsense reasoning, most of which have been previously modeled as discriminative tasks, our generative approach (with a single unified architecture) reaches comparable performance to recent task-specific state-of-the-art vision-and-language models. Moreover, our generative approach shows better generalization ability on questions that have rare answers. Also, we show that our framework allows multi-task learning in a single architecture with a single set of parameters, achieving similar performance to separately optimized single-task models. Our code is publicly available at: https://github.com/j-min/VL-T5"
                },
                {
                    "title": "Interpretability and Analysis in Neural NLP",
                    "abstract": "While deep learning has transformed the natural language processing (NLP) field and impacted the larger computational linguistics community, the rise of neural networks is stained by their opaque nature: It is challenging to interpret the inner workings of neural network models, and explicate their behavior. Therefore, in the last few years, an increasingly large body of work has been devoted to the analysis and interpretation of neural network models in NLP. This body of work is so far lacking a common framework and methodology. Moreover, approaching the analysis of modern neural networks can be difficult for newcomers to the field. This tutorial aims to fill this gap and introduce the nascent field of interpretability and analysis of neural networks in NLP. The tutorial will cover the main lines of analysis work, such as structural analyses using probing classifiers, behavioral studies and test suites, and interactive visualizations. We will highlight not only the most commonly applied analysis methods, but also the specific limitations and shortcomings of current approaches, in order to inform participants where to focus future efforts."
                },
                {
                    "title": "Sparse, Dense, and Attentional Representations for Text Retrieval",
                    "abstract": "Abstract Dual encoders perform retrieval by encoding documents and queries into dense low-dimensional vectors, scoring each document by its inner product with the query. We investigate the capacity of this architecture relative to sparse bag-of-words models and attentional neural networks. Using both theoretical and empirical analysis, we establish connections between the encoding dimension, the margin between gold and lower-ranked documents, and the document length, suggesting limitations in the capacity of fixed-length encodings to support precise retrieval of long documents. Building on these insights, we propose a simple neural model that combines the efficiency of dual encoders with some of the expressiveness of more costly attentional architectures, and explore sparse-dense hybrids to capitalize on the precision of sparse retrieval. These models outperform strong alternatives in large-scale retrieval."
                },
                {
                    "title": "Mixed Negative Sampling for Learning Two-tower Neural Networks in Recommendations",
                    "abstract": "Learning query and item representations is important for building large scale recommendation systems. In many real applications where there is a huge catalog of items to recommend, the problem of efficiently retrieving top k items given user\u2019s query from deep corpus leads to a family of factorized modeling approaches where queries and items are jointly embedded into a low-dimensional space. In this paper, we first showcase how to apply a two-tower neural network framework, which is also known as dual encoder in the natural language community, to improve a large-scale, production app recommendation system. Furthermore, we offer a novel negative sampling approach called Mixed Negative Sampling (MNS). In particular, different from commonly used batch or unigram sampling methods, MNS uses a mixture of batch and uniformly sampled negatives to tackle the selection bias of implicit user feedback. We conduct extensive offline experiments using large-scale production dataset and show that MNS outperforms other baseline sampling methods. We also conduct online A/B testing and demonstrate that the two-tower retrieval model based on MNS significantly improves retrieval quality by encouraging more high-quality app installs."
                },
                {
                    "title": "The impossibility of low-rank representations for triangle-rich complex networks",
                    "abstract": "Significance Our main message is that the popular method of low-dimensional embeddings provably cannot capture important properties of real-world complex networks. A widely used algorithmic technique for modeling these networks is to construct a low-dimensional Euclidean embedding of the vertices of the network, where proximity of vertices is interpreted as the likelihood of an edge. Contrary to common wisdom, we argue that such graph embeddings do not capture salient properties of complex networks. We mathematically prove that low-dimensional embeddings cannot generate graphs with both low average degree and large clustering coefficients, which have been widely established to be empirically true for real-world networks. This establishes that popular low-dimensional embedding methods fail to capture significant structural aspects of real-world complex networks. The study of complex networks is a significant development in modern science, and has enriched the social sciences, biology, physics, and computer science. Models and algorithms for such networks are pervasive in our society, and impact human behavior via social networks, search engines, and recommender systems, to name a few. A widely used algorithmic technique for modeling such complex networks is to construct a low-dimensional Euclidean embedding of the vertices of the network, where proximity of vertices is interpreted as the likelihood of an edge. Contrary to the common view, we argue that such graph embeddings do not capture salient properties of complex networks. The two properties we focus on are low degree and large clustering coefficients, which have been widely established to be empirically true for real-world networks. We mathematically prove that any embedding (that uses dot products to measure similarity) that can successfully create these two properties must have a rank that is nearly linear in the number of vertices. Among other implications, this establishes that popular embedding techniques such as singular value decomposition and node2vec fail to capture significant structural aspects of real-world complex networks. Furthermore, we empirically study a number of different embedding techniques based on dot product, and show that they all fail to capture the triangle structure."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "Sampling-bias-corrected neural modeling for large corpus item recommendations",
                    "abstract": "Many recommendation systems retrieve and score items from a very large corpus. A common recipe to handle data sparsity and power-law item distribution is to learn item representations from its content features. Apart from many content-aware systems based on matrix factorization, we consider a modeling framework using two-tower neural net, with one of the towers (item tower) encoding a wide variety of item content features. A general recipe of training such two-tower models is to optimize loss functions calculated from in-batch negatives, which are items sampled from a random mini-batch. However, in-batch loss is subject to sampling biases, potentially hurting model performance, particularly in the case of highly skewed distribution. In this paper, we present a novel algorithm for estimating item frequency from streaming data. Through theoretical analysis and simulation, we show that the proposed algorithm can work without requiring fixed item vocabulary, and is capable of producing unbiased estimation and being adaptive to item distribution change. We then apply the sampling-bias-corrected modeling approach to build a large scale neural retrieval system for YouTube recommendations. The system is deployed to retrieve personalized suggestions from a corpus with tens of millions of videos. We demonstrate the effectiveness of sampling-bias correction through offline experiments on two real-world datasets. We also conduct live A/B testings to show that the neural retrieval system leads to improved recommendation quality for YouTube."
                },
                {
                    "title": "Object Detection in 20 Years: A Survey",
                    "abstract": "Object detection, as of one the most fundamental and challenging problems in computer vision, has received great attention in recent years. Over the past two decades, we have seen a rapid technological evolution of object detection and its profound impact on the entire computer vision field. If we consider today\u2019s object detection technique as a revolution driven by deep learning, then, back in the 1990s, we would see the ingenious thinking and long-term perspective design of early computer vision. This article extensively reviews this fast-moving research field in the light of technical evolution, spanning over a quarter-century\u2019s time (from the 1990s to 2022). A number of topics have been covered in this article, including the milestone detectors in history, detection datasets, metrics, fundamental building blocks of the detection system, speedup techniques, and recent state-of-the-art detection methods."
                },
                {
                    "title": "Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences",
                    "abstract": "Significance Learning biological properties from sequence data is a logical step toward generative and predictive artificial intelligence for biology. Here, we propose scaling a deep contextual language model with unsupervised learning to sequences spanning evolutionary diversity. We find that without prior knowledge, information emerges in the learned representations on fundamental properties of proteins such as secondary structure, contacts, and biological activity. We show the learned representations are useful across benchmarks for remote homology detection, prediction of secondary structure, long-range residue\u2013residue contacts, and mutational effect. Unsupervised representation learning enables state-of-the-art supervised prediction of mutational effect and secondary structure and improves state-of-the-art features for long-range contact prediction. In the field of artificial intelligence, a combination of scale in data and model capacity enabled by unsupervised learning has led to major advances in representation learning and statistical generation. In the life sciences, the anticipated growth of sequencing promises unprecedented data on natural sequence diversity. Protein language modeling at the scale of evolution is a logical step toward predictive and generative artificial intelligence for biology. To this end, we use unsupervised learning to train a deep contextual language model on 86 billion amino acids across 250 million protein sequences spanning evolutionary diversity. The resulting model contains information about biological properties in its representations. The representations are learned from sequence data alone. The learned representation space has a multiscale organization reflecting structure from the level of biochemical properties of amino acids to remote homology of proteins. Information about secondary and tertiary structure is encoded in the representations and can be identified by linear projections. Representation learning produces features that generalize across a range of applications, enabling state-of-the-art supervised prediction of mutational effect and secondary structure and improving state-of-the-art features for long-range contact prediction."
                },
                {
                    "title": "UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction",
                    "abstract": "UMAP (Uniform Manifold Approximation and Projection) is a novel manifold learning technique for dimension reduction. UMAP is constructed from a theoretical framework based in Riemannian geometry and algebraic topology. The result is a practical scalable algorithm that applies to real world data. The UMAP algorithm is competitive with t-SNE for visualization quality, and arguably preserves more of the global structure with superior run time performance. Furthermore, UMAP has no computational restrictions on embedding dimension, making it viable as a general purpose dimension reduction technique for machine learning."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)",
                    "abstract": "The interpretation of deep learning models is a challenge due to their size, complexity, and often opaque internal state. In addition, many systems, such as image classifiers, operate on low-level features rather than high-level concepts. To address these challenges, we introduce Concept Activation Vectors (CAVs), which provide an interpretation of a neural net's internal state in terms of human-friendly concepts. The key idea is to view the high-dimensional internal state of a neural net as an aid, not an obstacle. We show how to use CAVs as part of a technique, Testing with CAVs (TCAV), that uses directional derivatives to quantify the degree to which a user-defined concept is important to a classification result--for example, how sensitive a prediction of \"zebra\" is to the presence of stripes. Using the domain of image classification as a testing ground, we describe how CAVs may be used to explore hypotheses and generate insights for a standard image classification network as well as a medical application."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "The MovieLens Datasets: History and Context",
                    "abstract": "The MovieLens datasets are widely used in education, research, and industry. They are downloaded hundreds of thousands of times each year, reflecting their use in popular press programming books, traditional and online courses, and software. These datasets are a product of member activity in the MovieLens movie recommendation system, an active research platform that has hosted many experiments since its launch in 1997. This article documents the history of MovieLens and the MovieLens datasets. We include a discussion of lessons learned from running a long-standing, live research platform from the perspective of a research organization. We document best practices and limitations of using the MovieLens datasets in new research."
                },
                {
                    "title": "Integrating and Evaluating Neural Word Embeddings in Information Retrieval",
                    "abstract": "Recent advances in neural language models have contributed new methods for learning distributed vector representations of words (also called word embeddings). Two such methods are the continuous bag-of-words model and the skipgram model. These methods have been shown to produce embeddings that capture higher order relationships between words that are highly effective in natural language processing tasks involving the use of word similarity and word analogy. Despite these promising results, there has been little analysis of the use of these word embeddings for retrieval. Motivated by these observations, in this paper, we set out to determine how these word embeddings can be used within a retrieval model and what the benefit might be. To this aim, we use neural word embeddings within the well known translation language model for information retrieval. This language model captures implicit semantic relations between the words in queries and those in relevant documents, thus producing more accurate estimations of document relevance. The word embeddings used to estimate neural language models produce translations that differ from previous translation language model approaches; differences that deliver improvements in retrieval effectiveness. The models are robust to choices made in building word embeddings and, even more so, our results show that embeddings do not even need to be produced from the same corpus being used for retrieval."
                },
                {
                    "title": "Generating Sentences from a Continuous Space",
                    "abstract": "The standard recurrent neural network language model (RNNLM) generates sentences one word at a time and does not work from an explicit global sentence representation. In this work, we introduce and study an RNN-based variational autoencoder generative model that incorporates distributed latent representations of entire sentences. This factorization allows it to explicitly model holistic properties of sentences such as style, topic, and high-level syntactic features. Samples from the prior over these sentence representations remarkably produce diverse and well-formed sentences through simple deterministic decoding. By examining paths through this latent space, we are able to generate coherent novel sentences that interpolate between known sentences. We present techniques for solving the difficult learning problem presented by this model, demonstrate its effectiveness in imputing missing words, explore many interesting properties of the model's latent sentence space, and present negative results on the use of the model in language modeling."
                },
                {
                    "title": "Deep Unordered Composition Rivals Syntactic Methods for Text Classification",
                    "abstract": "Many existing deep learning models for natural language processing tasks focus on learning the compositionality of their inputs, which requires many expensive computations. We present a simple deep neural network that competes with and, in some cases, outperforms such models on sentiment analysis and factoid question answering tasks while taking only a fraction of the training time. While our model is syntactically-ignorant, we show significant improvements over previous bag-of-words models by deepening our network and applying a novel variant of dropout. Moreover, our model performs better than syntactic models on datasets with high syntactic variance. We show that our model makes similar errors to syntactically-aware models, indicating that for the tasks we consider, nonlinearly transforming the input is more important than tailoring a network to incorporate word order and syntax."
                },
                {
                    "title": "Image-Based Recommendations on Styles and Substitutes",
                    "abstract": "Humans inevitably develop a sense of the relationships between objects, some of which are based on their appearance. Some pairs of objects might be seen as being alternatives to each other (such as two pairs of jeans), while others may be seen as being complementary (such as a pair of jeans and a matching shirt). This information guides many of the choices that people make, from buying clothes to their interactions with each other. We seek here to model this human sense of the relationships between objects based on their appearance. Our approach is not based on fine-grained modeling of user annotations but rather on capturing the largest dataset possible and developing a scalable method for uncovering human notions of the visual relationships within. We cast this as a network inference problem defined on graphs of related images, and provide a large-scale dataset for the training and evaluation of the same. The system we develop is capable of recommending which clothes and accessories will go well together (and which will not), amongst a host of other applications."
                },
                {
                    "title": "GloVe: Global Vectors for Word Representation",
                    "abstract": "Recent methods for learning vector space representations of words have succeeded in capturing fine-grained semantic and syntactic regularities using vector arithmetic, but the origin of these regularities has remained opaque. We analyze and make explicit the model properties needed for such regularities to emerge in word vectors. The result is a new global logbilinear regression model that combines the advantages of the two major model families in the literature: global matrix factorization and local context window methods. Our model efficiently leverages statistical information by training only on the nonzero elements in a word-word cooccurrence matrix, rather than on the entire sparse matrix or on individual context windows in a large corpus. The model produces a vector space with meaningful substructure, as evidenced by its performance of 75% on a recent word analogy task. It also outperforms related models on similarity tasks and named entity recognition."
                },
                {
                    "title": "Distributed Representations of Words and Phrases and their Compositionality",
                    "abstract": "The recently introduced continuous Skip-gram model is an efficient method for learning high-quality distributed vector representations that capture a large number of precise syntactic and semantic word relationships. In this paper we present several extensions that improve both the quality of the vectors and the training speed. By subsampling of the frequent words we obtain significant speedup and also learn more regular word representations. We also describe a simple alternative to the hierarchical softmax called negative sampling. \n \nAn inherent limitation of word representations is their indifference to word order and their inability to represent idiomatic phrases. For example, the meanings of \"Canada\" and \"Air\" cannot be easily combined to obtain \"Air Canada\". Motivated by this example, we present a simple method for finding phrases in text, and show that learning good vector representations for millions of phrases is possible."
                },
                {
                    "title": "Factorization Machines",
                    "abstract": "In this paper, we introduce Factorization Machines (FM) which are a new model class that combines the advantages of Support Vector Machines (SVM) with factorization models. Like SVMs, FMs are a general predictor working with any real valued feature vector. In contrast to SVMs, FMs model all interactions between variables using factorized parameters. Thus they are able to estimate interactions even in problems with huge sparsity (like recommender systems) where SVMs fail. We show that the model equation of FMs can be calculated in linear time and thus FMs can be optimized directly. So unlike nonlinear SVMs, a transformation in the dual form is not necessary and the model parameters can be estimated directly without the need of any support vector in the solution. We show the relationship to SVMs and the advantages of FMs for parameter estimation in sparse settings. On the other hand there are many different factorization models like matrix factorization, parallel factor analysis or specialized models like SVD++, PITF or FPMC. The drawback of these models is that they are not applicable for general prediction tasks but work only with special input data. Furthermore their model equations and optimization algorithms are derived individually for each task. We show that FMs can mimic these models just by specifying the input data (i.e. the feature vectors). This makes FMs easily applicable even for users without expert knowledge in factorization models."
                },
                {
                    "title": "Collaborative Filtering for Implicit Feedback Datasets",
                    "abstract": "A common task of recommender systems is to improve customer experience through personalized recommendations based on prior implicit feedback. These systems passively track different sorts of user behavior, such as purchase history, watching habits and browsing activity, in order to model user preferences. Unlike the much more extensively researched explicit feedback, we do not have any direct input from the users regarding their preferences. In particular, we lack substantial evidence on which products consumer dislike. In this work we identify unique properties of implicit feedback datasets. We propose treating the data as indication of positive and negative preference associated with vastly varying confidence levels. This leads to a factor model which is especially tailored for implicit feedback recommenders. We also suggest a scalable optimization procedure, which scales linearly with the data size. The algorithm is used successfully within a recommender system for television shows. It compares favorably with well tuned implementations of other known methods. In addition, we offer a novel way to give explanations to recommendations given by this factor model."
                },
                {
                    "title": "A unified architecture for natural language processing: deep neural networks with multitask learning",
                    "abstract": "We describe a single convolutional neural network architecture that, given a sentence, outputs a host of language processing predictions: part-of-speech tags, chunks, named entity tags, semantic roles, semantically similar words and the likelihood that the sentence makes sense (grammatically and semantically) using a language model. The entire network is trained jointly on all these tasks using weight-sharing, an instance of multitask learning. All the tasks use labeled data except the language model which is learnt from unlabeled text and represents a novel form of semi-supervised learning for the shared tasks. We show how both multitask learning and semi-supervised learning improve the generalization of the shared tasks, resulting in state-of-the-art-performance."
                },
                {
                    "title": "Probabilistic Matrix Factorization",
                    "abstract": "Many existing approaches to collaborative filtering can neither handle very large datasets nor easily deal with users who have very few ratings. In this paper we present the Probabilistic Matrix Factorization (PMF) model which scales linearly with the number of observations and, more importantly, performs well on the large, sparse, and very imbalanced Netflix dataset. We further extend the PMF model to include an adaptive prior on the model parameters and show how the model capacity can be controlled automatically. Finally, we introduce a constrained version of the PMF model that is based on the assumption that users who have rated similar sets of movies are likely to have similar preferences. The resulting model is able to generalize considerably better for users with very few ratings. When the predictions of multiple PMF models are linearly combined with the predictions of Restricted Boltzmann Machines models, we achieve an error rate of 0.8861, that is nearly 7% better than the score of Netflix's own system."
                },
                {
                    "title": "Learning to rank using gradient descent",
                    "abstract": "We investigate using gradient descent methods for learning ranking functions; we propose a simple probabilistic cost function, and we introduce RankNet, an implementation of these ideas using a neural network to model the underlying ranking function. We present test results on toy data and on data from a commercial internet search engine."
                },
                {
                    "title": "A Neural Probabilistic Language Model",
                    "abstract": "A goal of statistical language modeling is to learn the joint probability function of sequences of words. This is intrinsically difficult because of the curse of dimensionality: we propose to fight it with its own weapons. In the proposed approach one learns simultaneously (1) a distributed representation for each word (i.e. a similarity between words) along with (2) the probability function for word sequences, expressed with these representations. Generalization is obtained because a sequence of words that has never been seen before gets high probability if it is made of words that are similar to words forming an already seen sentence. We report on experiments using neural networks for the probability function, showing on two text corpora that the proposed approach very significantly improves on a state-of-the-art trigram model."
                },
                {
                    "title": "Prefix-Tuning: Optimizing Continuous Prompts for Generation",
                    "abstract": "Fine-tuning is the de facto way of leveraging large pretrained language models for downstream tasks. However, fine-tuning modifies all the language model parameters and therefore necessitates storing a full copy for each task. In this paper, we propose prefix-tuning, a lightweight alternative to fine-tuning for natural language generation tasks, which keeps language model parameters frozen and instead optimizes a sequence of continuous task-specific vectors, which we call the prefix. Prefix-tuning draws inspiration from prompting for language models, allowing subsequent tokens to attend to this prefix as if it were \u201cvirtual tokens\u201d. We apply prefix-tuning to GPT-2 for table-to-text generation and to BART for summarization. We show that by learning only 0.1% of the parameters, prefix-tuning obtains comparable performance in the full data setting, outperforms fine-tuning in low-data settings, and extrapolates better to examples with topics that are unseen during training."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Improving Language Understanding by Generative Pre-Training",
                    "abstract": "Natural language understanding comprises a wide range of diverse tasks such as textual entailment, question answering, semantic similarity assessment, and document classi\ufb01cation. Although large unlabeled text corpora are abundant, labeled data for learning these speci\ufb01c tasks is scarce, making it challenging for discriminatively trained models to perform adequately. We demonstrate that large gains on these tasks can be realized by generative pre-training of a language model on a diverse corpus of unlabeled text, followed by discriminative \ufb01ne-tuning on each speci\ufb01c task. In contrast to previous approaches, we make use of task-aware input transformations during \ufb01ne-tuning to achieve effective transfer while requiring minimal changes to the model architecture. We demonstrate the effectiveness of our approach on a wide range of benchmarks for natural language understanding. Our general task-agnostic model outperforms discriminatively trained models that use architectures speci\ufb01cally crafted for each task, signi\ufb01cantly improving upon the state of the art in 9 out of the 12 tasks studied. For instance, we achieve absolute improvements of 8.9% on commonsense reasoning (Stories Cloze Test), 5.7% on question answering (RACE), and 1.5% on textual entailment (MultiNLI)."
                },
                {
                    "title": "Visualizing Data using t-SNE",
                    "abstract": "We present a new technique called \u201ct-SNE\u201d that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to influence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are significantly better than those produced by the other techniques on almost all of the datasets."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively interpret and manipulate domain embeddings using large language models (LLMs) to extract meaningful insights and information that are not directly expressed by the embeddings themselves?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant gap in the interpretability of embeddings, which are foundational in various applications such as natural language processing, recommender systems, and more. By enabling a deeper understanding of embeddings, this research could lead to advancements in how we utilize these representations across different domains, fostering innovation in model design and application. Furthermore, it could pave the way for practical applications where users can query embeddings in natural language, enhancing user interaction with AI systems and improving decision-making processes in fields like e-commerce, healthcare, and beyond.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of embeddings, which are tailored to specific tasks and often lack a straightforward interpretation. Naive approaches may fail because they do not account for the nuanced relationships and high-dimensional nature of the data represented in embeddings. Additionally, technical obstacles include the need for sophisticated mapping between domain embeddings and the token-level embedding space of LLMs, as well as the requirement for training the LLM on diverse tasks to ensure robust interpretation. Theoretical challenges also arise from the abstract nature of embeddings, making it difficult to derive meaningful narratives or insights without a structured approach.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on task-independent interpretability methods, which often lack the depth required to engage with embeddings meaningfully. Limitations in existing solutions include a narrow scope of interpretability techniques and the absence of frameworks that integrate LLMs with domain embeddings. Barriers such as the complexity of embedding structures and the lack of methodologies for direct interaction with these representations have hindered progress. Our approach differs by introducing a novel framework that leverages LLMs to facilitate a dialogue about embeddings, allowing for a more nuanced and comprehensive interpretation.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the Embedding Language Model (ELM), which utilizes trained adapter layers to map domain embedding vectors into the token-level embedding space of an LLM. We will employ a diverse dataset of domain embeddings, along with a series of tasks designed to enhance the"
            }
        },
        "author_data": {
            "0a6b0d6e-d7dc-492c-b235-51d07d69cec0": {
                "pk": "0a6b0d6e-d7dc-492c-b235-51d07d69cec0",
                "name": "Guy Tennenholtz",
                "collaborators": [
                    "Shie Mannor",
                    "Craig Boutilier",
                    "Martin Mladenov",
                    "Nir Baram",
                    "Uri Shalit",
                    "Chih-Wei Hsu",
                    "Yinlam Chow",
                    "Lior Shani",
                    "Nadav Merlis",
                    "M. Ghavamzadeh",
                    "Gal Chechik",
                    "Assaf Hallak",
                    "Gal Dalal",
                    "Ori Linial",
                    "Anthony Liang",
                    "Erdem Biyik",
                    "Ethan Liang",
                    "Marek Petrik",
                    "Ofir Nabati",
                    "Jihwan Jeong",
                    "Azamat Tulepbergenov",
                    "Aliz\u00e9e Pace",
                    "Hugo Y\u00e8che",
                    "B. Scholkopf",
                    "G. Ratsch",
                    "Roy Zohar"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Recommender Systems",
                    "Mechanism Design",
                    "Multi-Agent Systems"
                ],
                "publications": [
                    {
                        "title": "Recommender Ecosystems: A Mechanism Design Perspective on Holistic Modeling and Optimization",
                        "abstract": "Modern recommender systems lie at the heart of complex recommender ecosystems that couple the behavior of users, content providers, vendors, advertisers, and other actors. Despite this, the focus of much recommender systems research and deployment is on the local, myopic optimization of the recommendations made to individual users. This comes at a significant cost to the long-term utility that recommender systems generate for their users. We argue that modeling the incentives and behaviors of these actors, and the interactions among them induced by the recommender systems, is needed to maximize value and improve overall ecosystem health. Moreover, we propose the use of economic mechanism design, an area largely overlooked in recommender systems research, as a framework for developing such models. That said, one cannot apply \u201cvanilla\u201d mechanism design to recommender ecosystem modeling optimization out of the box\u2014the use of mechanism design raises a number of subtle and interesting research challenges. We outline a number of these in this talk (and paper), emphasizing the need to develop nonstandard approaches to mechanism design that intersect with numerous areas of research, including preference modeling, reinforcement learning and exploration, behavioral economics, and generative AI, among others."
                    },
                    {
                        "title": "Benchmarks for Reinforcement Learning with Biased Offline Data and Imperfect Simulators",
                        "abstract": "In many reinforcement learning (RL) applications one cannot easily let the agent act in the world; this is true for autonomous vehicles, healthcare applications, and even some recommender systems, to name a few examples. Offline RL provides a way to train agents without real-world exploration, but is often faced with biases due to data distribution shifts, limited coverage, and incomplete representation of the environment. To address these issues, practical applications have tried to combine simulators with grounded offline data, using so-called hybrid methods. However, constructing a reliable simulator is in itself often challenging due to intricate system complexities as well as missing or incomplete information. In this work, we outline four principal challenges for combining offline data with imperfect simulators in RL: simulator modeling error, partial observability, state and action discrepancies, and hidden confounding. To help drive the RL community to pursue these problems, we construct ``Benchmarks for Mechanistic Offline Reinforcement Learning'' (B4MRL), which provide dataset-simulator benchmarks for the aforementioned challenges. Our results suggest the key necessity of such benchmarks for future research."
                    },
                    {
                        "title": "DynaMITE-RL: A Dynamic Model for Improved Temporal Meta-Reinforcement Learning",
                        "abstract": "We introduce DynaMITE-RL, a meta-reinforcement learning (meta-RL) approach to approximate inference in environments where the latent state evolves at varying rates. We model episode sessions - parts of the episode where the latent state is fixed - and propose three key modifications to existing meta-RL methods: consistency of latent information within sessions, session masking, and prior latent conditioning. We demonstrate the importance of these modifications in various domains, ranging from discrete Gridworld environments to continuous-control and simulated robot assistive tasks, demonstrating that DynaMITE-RL significantly outperforms state-of-the-art baselines in sample efficiency and inference returns."
                    },
                    {
                        "title": "Embedding-Aligned Language Models",
                        "abstract": "We propose a novel approach for training large language models (LLMs) to adhere to objectives defined within a latent embedding space. Our method leverages reinforcement learning (RL), treating a pre-trained LLM as an environment. Our embedding-aligned guided language (EAGLE) agent is trained to iteratively steer the LLM's generation towards optimal regions of the latent embedding space, w.r.t. some predefined criterion. We demonstrate the effectiveness of the EAGLE agent using the MovieLens 25M and Amazon Review datasets to surface content gaps that satisfy latent user demand. We also demonstrate the benefit of using an optimal design of a state-dependent action set to improve EAGLE's efficiency. Our work paves the way for controlled and grounded text generation using LLMs, ensuring consistency with domain-specific knowledge and data representations."
                    },
                    {
                        "title": "Reinforcement Learning with History-Dependent Dynamic Contexts",
                        "abstract": "We introduce Dynamic Contextual Markov Decision Processes (DCMDPs), a novel reinforcement learning framework for history-dependent environments that generalizes the contextual MDP framework to handle non-Markov environments, where contexts change over time. We consider special cases of the model, with a focus on logistic DCMDPs, which break the exponential dependence on history length by leveraging aggregation functions to determine context transitions. This special structure allows us to derive an upper-confidence-bound style algorithm for which we establish regret bounds. Motivated by our theoretical results, we introduce a practical model-based algorithm for logistic DCMDPs that plans in a latent space and uses optimism over history-dependent features. We demonstrate the efficacy of our approach on a recommendation task (using MovieLens data) where user behavior dynamics evolve in response to recommendations."
                    },
                    {
                        "title": "Modeling Recommender Ecosystems: Research Challenges at the Intersection of Mechanism Design, Reinforcement Learning and Generative Models",
                        "abstract": "Modern recommender systems lie at the heart of complex ecosystems that couple the behavior of users, content providers, advertisers, and other actors. Despite this, the focus of the majority of recommender research -- and most practical recommenders of any import -- is on the local, myopic optimization of the recommendations made to individual users. This comes at a significant cost to the long-term utility that recommenders could generate for its users. We argue that explicitly modeling the incentives and behaviors of all actors in the system -- and the interactions among them induced by the recommender's policy -- is strictly necessary if one is to maximize the value the system brings to these actors and improve overall ecosystem\"health\". Doing so requires: optimization over long horizons using techniques such as reinforcement learning; making inevitable tradeoffs in the utility that can be generated for different actors using the methods of social choice; reducing information asymmetry, while accounting for incentives and strategic behavior, using the tools of mechanism design; better modeling of both user and item-provider behaviors by incorporating notions from behavioral economics and psychology; and exploiting recent advances in generative and foundation models to make these mechanisms interpretable and actionable. We propose a conceptual framework that encompasses these elements, and articulate a number of research challenges that emerge at the intersection of these different disciplines."
                    },
                    {
                        "title": "Bayesian Regret Minimization in Offline Bandits",
                        "abstract": "We study how to make decisions that minimize Bayesian regret in offline linear bandits. Prior work suggests that one must take actions with maximum lower confidence bound (LCB) on their reward. We argue that the reliance on LCB is inherently flawed in this setting and propose a new algorithm that directly minimizes upper bounds on the Bayesian regret using efficient conic optimization solvers. Our bounds build heavily on new connections to monetary risk measures. Proving a matching lower bound, we show that our upper bounds are tight, and by minimizing them we are guaranteed to outperform the LCB approach. Our numerical results on synthetic domains confirm that our approach is superior to LCB."
                    },
                    {
                        "title": "Representation-Driven Reinforcement Learning",
                        "abstract": "We present a representation-driven framework for reinforcement learning. By representing policies as estimates of their expected values, we leverage techniques from contextual bandits to guide exploration and exploitation. Particularly, embedding a policy network into a linear feature space allows us to reframe the exploration-exploitation problem as a representation-exploitation problem, where good policy representations enable optimal exploration. We demonstrate the effectiveness of this framework through its application to evolutionary and policy gradient-based approaches, leading to significantly improved performance compared to traditional methods. Our framework provides a new perspective on reinforcement learning, highlighting the importance of policy representation in determining optimal exploration-exploitation strategies."
                    },
                    {
                        "title": "Factual and Personalized Recommendations using Language Models and Reinforcement Learning",
                        "abstract": "Recommender systems (RSs) play a central role in connecting users to content, products, and services, matching candidate items to users based on their preferences. While traditional RSs rely on implicit user feedback signals, conversational RSs interact with users in natural language. In this work, we develop a comPelling, Precise, Personalized, Preference-relevant language model (P4LM) that recommends items to users while putting emphasis on explaining item characteristics and their relevance. P4LM uses the embedding space representation of a user's preferences to generate compelling responses that are factually-grounded and relevant w.r.t. the user's preferences. Moreover, we develop a joint reward function that measures precision, appeal, and personalization, which we use as AI-based feedback in a reinforcement learning-based language model framework. Using the MovieLens 25M dataset, we demonstrate that P4LM delivers compelling, personalized movie narratives to users."
                    },
                    {
                        "title": "Ranking with Popularity Bias: User Welfare under Self-Amplification Dynamics",
                        "abstract": "While popularity bias is recognized to play a crucial role in recommmender (and other ranking-based) systems, detailed analysis of its impact on collective user welfare has largely been lacking. We propose and theoretically analyze a general mechanism, rooted in many of the models proposed in the literature, by which item popularity, item quality, and position bias jointly impact user choice. We focus on a standard setting in which user utility is largely driven by item quality, and a recommender attempts to estimate it given user behavior. Formulating the problem as a non-stationary contextual bandit, we study the ability of a recommender policy to maximize user welfare under this model. We highlight the importance of exploration, not to eliminate popularity bias, but to mitigate its negative impact on welfare. We first show that naive popularity-biased recommenders induce linear regret by conflating item quality and popularity. More generally, we show that, even in linear settings, identifiability of item quality may not be possible due to the confounding effects of popularity bias. However, under sufficient variability assumptions, we develop an efficient optimistic algorithm and prove efficient regret guarantees w.r.t. user welfare. We complement our analysis with several simulation studies, which demonstrate the negative impact of popularity bias on the performance of several natural recommender policies."
                    },
                    {
                        "title": "Delphic Offline Reinforcement Learning under Nonidentifiable Hidden Confounding",
                        "abstract": "A prominent challenge of offline reinforcement learning (RL) is the issue of hidden confounding: unobserved variables may influence both the actions taken by the agent and the observed outcomes. Hidden confounding can compromise the validity of any causal conclusion drawn from data and presents a major obstacle to effective offline RL. In the present paper, we tackle the problem of hidden confounding in the nonidentifiable setting. We propose a definition of uncertainty due to hidden confounding bias, termed delphic uncertainty, which uses variation over world models compatible with the observations, and differentiate it from the well-known epistemic and aleatoric uncertainties. We derive a practical method for estimating the three types of uncertainties, and construct a pessimistic offline RL algorithm to account for them. Our method does not assume identifiability of the unobserved confounders, and attempts to reduce the amount of confounding bias. We demonstrate through extensive experiments and ablations the efficacy of our approach on a sepsis management benchmark, as well as on electronic health records. Our results suggest that nonidentifiable hidden confounding bias can be mitigated to improve offline RL solutions in practice."
                    },
                    {
                        "title": "Reinforcement Learning with a Terminator",
                        "abstract": "We present the problem of reinforcement learning with exogenous termination. We define the Termination Markov Decision Process (TerMDP), an extension of the MDP framework, in which episodes may be interrupted by an external non-Markovian observer. This formulation accounts for numerous real-world situations, such as a human interrupting an autonomous driving agent for reasons of discomfort. We learn the parameters of the TerMDP and leverage the structure of the estimation problem to provide state-wise confidence bounds. We use these to construct a provably-efficient algorithm, which accounts for termination, and bound its regret. Motivated by our theoretical analysis, we design and implement a scalable approach, which combines optimism (w.r.t. termination) and a dynamic discount factor, incorporating the termination probability. We deploy our method on high-dimensional driving and MinAtar benchmarks. Additionally, we test our approach on human data in a driving setting. Our results demonstrate fast convergence and significant improvement over various baseline approaches."
                    },
                    {
                        "title": "GELATO: Geometrically Enriched Latent Model for Offline Reinforcement Learning",
                        "abstract": "Of\ufb02ine reinforcement learning approaches can generally be divided to proximal and uncertainty-aware methods. In this work, we demonstrate the bene\ufb01t of combining the two in a latent variational model. We impose a latent representation of states and actions and leverage its intrinsic Riemannian geometry to measure distance of latent samples to the data. Our proposed metrics measure both the quality of out of distribution samples as well as the discrepancy of examples in the data. We integrate our metrics in a model-based of\ufb02ine optimization framework, in which proximity and uncertainty can be carefully controlled. We illustrate the geodesics on a simple grid-like environment, depicting its natural inherent topology. Finally, we analyze our approach and improve upon contemporary of\ufb02ine RL benchmarks."
                    },
                    {
                        "title": "Maximum Entropy Reinforcement Learning with Mixture Policies",
                        "abstract": "Mixture models are an expressive hypothesis class that can approximate a rich set of policies. However, using mixture policies in the Maximum Entropy (MaxEnt) framework is not straightforward. The entropy of a mixture model is not equal to the sum of its components, nor does it have a closed-form expression in most cases. Using such policies in MaxEnt algorithms, therefore, requires constructing a tractable approximation of the mixture entropy. In this paper, we derive a simple, low-variance mixture-entropy estimator. We show that it is closely related to the sum of marginal entropies. Equipped with our entropy estimator, we derive an algorithmic variant of Soft Actor-Critic (SAC) to the mixture policy case and evaluate it on a series of continuous control tasks."
                    },
                    {
                        "title": "Uncertainty Estimation Using Riemannian Model Dynamics for Offline Reinforcement Learning",
                        "abstract": "Model-based offline reinforcement learning approaches generally rely on bounds of model error. Estimating these bounds is usually achieved through uncertainty estimation methods. In this work, we combine parametric and nonparametric methods for uncertainty estimation through a novel latent space based metric. In particular, we build upon recent advances in Riemannian geometry of generative models to construct a pullback metric of an encoder-decoder based forward model. Our proposed metric measures both the quality of out-of-distribution samples as well as the discrepancy of examples in the data. We leverage our method for uncertainty estimation in a pessimistic model-based framework, showing a significant improvement upon contemporary model-based offline approaches on continuous control and autonomous driving benchmarks."
                    },
                    {
                        "title": "Action Redundancy in Reinforcement Learning",
                        "abstract": "Maximum Entropy (MaxEnt) reinforcement learning is a powerful learning paradigm which seeks to maximize return under entropy regularization. However, action entropy does not necessarily coincide with state entropy, e.g., when multiple actions produce the same transition. Instead, we propose to maximize the transition entropy, i.e., the entropy of next states. We show that transition entropy can be described by two terms; namely, model-dependent transition entropy and action redundancy. Particularly, we explore the latter in both deterministic and stochastic settings and develop tractable approximation methods in a near model-free setup. We construct algorithms to minimize action redundancy and demonstrate their effectiveness on a synthetic environment with multiple redundant actions as well as contemporary benchmarks in Atari and Mujoco. Our results suggest that action redundancy is a fundamental problem in reinforcement learning."
                    },
                    {
                        "title": "Latent Geodesics of Model Dynamics for Of\ufb02ine Reinforcement Learning",
                        "abstract": "Model-based of\ufb02ine reinforcement learning approaches generally rely on bounds of model error. While contemporary methods achieve such bounds through an ensemble of models, we propose to estimate them using a data-driven latent metric. Particularly, we build upon recent advances in Riemannian geometry of generative models to construct a latent metric of an encoder-decoder based forward model. Our proposed metric measures both the quality of out of distribution samples as well as the discrepancy of examples in the data. We show that our metric can be viewed as a combination of two metrics, one relating to proximity and the other to epistemic uncertainty. Finally, we leverage our metric in a pessimistic model-based framework, showing a signi\ufb01cant improvement upon contemporary model-based of\ufb02ine reinforcement learning benchmarks."
                    },
                    {
                        "title": "On Covariate Shift of Latent Confounders in Imitation and Reinforcement Learning",
                        "abstract": "We consider the problem of using expert data with unobserved confounders for imitation and reinforcement learning. We begin by defining the problem of learning from confounded expert data in a contextual MDP setup. We analyze the limitations of learning from such data with and without external reward, and propose an adjustment of standard imitation learning algorithms to fit this setup. We then discuss the problem of distribution shift between the expert data and the online environment when the data is only partially observable. We prove possibility and impossibility results for imitation learning under arbitrary distribution shift of the missing covariates. When additional external reward is provided, we propose a sampling procedure that addresses the unknown shift and prove convergence to an optimal solution. Finally, we validate our claims empirically on challenging assistive healthcare and recommender system simulation tasks."
                    },
                    {
                        "title": "Locality Matters: A Scalable Value Decomposition Approach for Cooperative Multi-Agent Reinforcement Learning",
                        "abstract": "Cooperative multi-agent reinforcement learning (MARL) faces significant scalability issues due to state and action spaces that are exponentially large in the number of agents. As environments grow in size, effective credit assignment becomes increasingly harder and often results in infeasible learning times. Still, in many real-world settings, there exist simplified underlying dynamics that can be leveraged for more scalable solutions. In this work, we exploit such locality structures effectively whilst maintaining global cooperation. We propose a novel, value-based multi-agent algorithm called LOMAQ, which incorporates local rewards in the Centralized Training Decentralized Execution paradigm. Additionally, we provide a direct reward decomposition method for finding these local rewards when only a global signal is provided. We test our method empirically, showing it scales well compared to other methods, significantly improving performance and convergence speed."
                    }
                ]
            },
            "5a2cfaf3-4c7a-48cb-8fd5-32cb44b4e693": {
                "pk": "5a2cfaf3-4c7a-48cb-8fd5-32cb44b4e693",
                "name": "Yinlam Chow",
                "collaborators": [
                    "Craig Boutilier",
                    "M. Ghavamzadeh",
                    "Chih-Wei Hsu",
                    "Joey Hong",
                    "B. Kveton",
                    "M. Zaheer",
                    "Amr Ahmed",
                    "Guy Tennenholtz",
                    "M. Ryu",
                    "Erdem Biyik",
                    "Azamat Tulepbergenov",
                    "Ido Greenberg",
                    "Ofir Nachum",
                    "Bo Dai",
                    "Brandon Cui",
                    "Anthony Liang",
                    "Lior Shani",
                    "Ethan Liang",
                    "Fan Yao",
                    "Alex Haig",
                    "Jihwan Jeong",
                    "Dhawal Gupta",
                    "Christina G\u00f6pfert",
                    "Ivan Vendrov",
                    "Tyler Lu",
                    "Deepak Ramachandran",
                    "Deborah Cohen",
                    "Orgad Keller",
                    "Avinatan Hassidim",
                    "Michael Fink",
                    "Yossi Matias",
                    "Idan Szpektor",
                    "G. Elidan",
                    "Shie Mannor",
                    "Dylan Slack",
                    "Nevan Wichers",
                    "Lihong Li",
                    "Csaba Szepesvari",
                    "D. Schuurmans",
                    "Tsung-Yen Yang",
                    "Michael Y Hu",
                    "P. Ramadge",
                    "Karthik Narasimhan",
                    "Rui Shu",
                    "Tung D. Nguyen",
                    "Tu Pham",
                    "Khoat Than",
                    "Stefano Ermon",
                    "H. Bui"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Natural Language Processing",
                    "Recommender Systems",
                    "Meta-Learning"
                ],
                "publications": [
                    {
                        "title": "DynaMITE-RL: A Dynamic Model for Improved Temporal Meta-Reinforcement Learning",
                        "abstract": "We introduce DynaMITE-RL, a meta-reinforcement learning (meta-RL) approach to approximate inference in environments where the latent state evolves at varying rates. We model episode sessions - parts of the episode where the latent state is fixed - and propose three key modifications to existing meta-RL methods: consistency of latent information within sessions, session masking, and prior latent conditioning. We demonstrate the importance of these modifications in various domains, ranging from discrete Gridworld environments to continuous-control and simulated robot assistive tasks, demonstrating that DynaMITE-RL significantly outperforms state-of-the-art baselines in sample efficiency and inference returns."
                    },
                    {
                        "title": "Embedding-Aligned Language Models",
                        "abstract": "We propose a novel approach for training large language models (LLMs) to adhere to objectives defined within a latent embedding space. Our method leverages reinforcement learning (RL), treating a pre-trained LLM as an environment. Our embedding-aligned guided language (EAGLE) agent is trained to iteratively steer the LLM's generation towards optimal regions of the latent embedding space, w.r.t. some predefined criterion. We demonstrate the effectiveness of the EAGLE agent using the MovieLens 25M and Amazon Review datasets to surface content gaps that satisfy latent user demand. We also demonstrate the benefit of using an optimal design of a state-dependent action set to improve EAGLE's efficiency. Our work paves the way for controlled and grounded text generation using LLMs, ensuring consistency with domain-specific knowledge and data representations."
                    },
                    {
                        "title": "Preference Elicitation with Soft Attributes in Interactive Recommendation",
                        "abstract": "Preference elicitation plays a central role in interactive recommender systems. Most preference elicitation approaches use either item queries that ask users to select preferred items from a slate, or attribute queries that ask them to express their preferences for item characteristics. Unfortunately, users often wish to describe their preferences using soft attributes for which no ground-truth semantics is given. Leveraging concept activation vectors for soft attribute semantics, we develop novel preference elicitation methods that can accommodate soft attributes and bring together both item and attribute-based preference elicitation. Our techniques query users using both items and soft attributes to update the recommender system's belief about their preferences to improve recommendation quality. We demonstrate the effectiveness of our methods vis-a-vis competing approaches on both synthetic and real-world datasets."
                    },
                    {
                        "title": "Factual and Personalized Recommendations using Language Models and Reinforcement Learning",
                        "abstract": "Recommender systems (RSs) play a central role in connecting users to content, products, and services, matching candidate items to users based on their preferences. While traditional RSs rely on implicit user feedback signals, conversational RSs interact with users in natural language. In this work, we develop a comPelling, Precise, Personalized, Preference-relevant language model (P4LM) that recommends items to users while putting emphasis on explaining item characteristics and their relevance. P4LM uses the embedding space representation of a user's preferences to generate compelling responses that are factually-grounded and relevant w.r.t. the user's preferences. Moreover, we develop a joint reward function that measures precision, appeal, and personalization, which we use as AI-based feedback in a reinforcement learning-based language model framework. Using the MovieLens 25M dataset, we demonstrate that P4LM delivers compelling, personalized movie narratives to users."
                    },
                    {
                        "title": "Offline Reinforcement Learning for Mixture-of-Expert Dialogue Management",
                        "abstract": "Reinforcement learning (RL) has shown great promise for developing dialogue management (DM) agents that are non-myopic, conduct rich conversations, and maximize overall user satisfaction. Despite recent developments in RL and language models (LMs), using RL to power conversational chatbots remains challenging, in part because RL requires online exploration to learn effectively, whereas collecting novel human-bot interactions can be expensive and unsafe. This issue is exacerbated by the combinatorial action spaces facing these algorithms, as most LM agents generate responses at the word level. We develop a variety of RL algorithms, specialized to dialogue planning, that leverage recent Mixture-of-Expert Language Models (MoE-LMs) -- models that capture diverse semantics, generate utterances reflecting different intents, and are amenable for multi-turn DM. By exploiting MoE-LM structure, our methods significantly reduce the size of the action space and improve the efficacy of RL-based DM. We evaluate our methods in open-domain dialogue to demonstrate their effectiveness w.r.t.\\ the diversity of intent in generated utterances and overall DM performance."
                    },
                    {
                        "title": "Discovering Personalized Semantics for Soft Attributes in Recommender Systems using Concept Activation Vectors",
                        "abstract": "Interactive recommender systems (RSs) allow users to express intent, preferences and contexts in a rich fashion, often using natural language. One challenge in using such feedback is inferring a user\u2019s semantic intent from the open-ended terms used to describe an item, and using it to refine recommendation results. Leveraging concept activation vectors (CAVs) [21], we develop a framework to learn a representation that captures the semantics of such attributes and connects them to user preferences and behaviors in RSs. A novel feature of our approach is its ability to distinguish objective and subjective attributes and associate different senses with different users. Using synthetic and real-world datasets, we show that our CAV representation accurately interprets users\u2019 subjective semantics, and can improve recommendations via interactive critiquing."
                    },
                    {
                        "title": "Dynamic Planning in Open-Ended Dialogue using Reinforcement Learning",
                        "abstract": "Despite recent advances in natural language understanding and generation, and decades of research on the development of conversational bots, building automated agents that can carry on rich open-ended conversations with humans\"in the wild\"remains a formidable challenge. In this work we develop a real-time, open-ended dialogue system that uses reinforcement learning (RL) to power a bot's conversational skill at scale. Our work pairs the succinct embedding of the conversation state generated using SOTA (supervised) language models with RL techniques that are particularly suited to a dynamic action space that changes as the conversation progresses. Trained using crowd-sourced data, our novel system is able to substantially exceeds the (strong) baseline supervised model with respect to several metrics of interest in a live experiment with real users of the Google Assistant."
                    },
                    {
                        "title": "A Mixture-of-Expert Approach to RL-based Dialogue Management",
                        "abstract": "Despite recent advancements in language models (LMs), their application to dialogue management (DM) problems and ability to carry on rich conversations remain a challenge. We use reinforcement learning (RL) to develop a dialogue agent that avoids being short-sighted (outputting generic utterances) and maximizes overall user satisfaction. Most existing RL approaches to DM train the agent at the word-level, and thus, have to deal with a combinatorially complex action space even for a medium-size vocabulary. As a result, they struggle to produce a successful and engaging dialogue even if they are warm-started with a pre-trained LM. To address this issue, we develop a RL-based DM using a novel mixture of expert language model (MoE-LM) that consists of (i) a LM capable of learning diverse semantics for conversation histories, (ii) a number of {\\em specialized} LMs (or experts) capable of generating utterances corresponding to a particular attribute or personality, and (iii) a RL-based DM that performs dialogue planning with the utterances generated by the experts. Our MoE approach provides greater flexibility to generate sensible utterances with different intents and allows RL to focus on conversational-level DM. We compare it with SOTA baselines on open-domain dialogues and demonstrate its effectiveness both in terms of the diversity and sensibility of the generated utterances and the overall DM performance."
                    },
                    {
                        "title": "Efficient Risk-Averse Reinforcement Learning",
                        "abstract": "In risk-averse reinforcement learning (RL), the goal is to optimize some risk measure of the returns. A risk measure often focuses on the worst returns out of the agent's experience. As a result, standard methods for risk-averse RL often ignore high-return strategies. We prove that under certain conditions this inevitably leads to a local-optimum barrier, and propose a soft risk mechanism to bypass it. We also devise a novel Cross Entropy module for risk sampling, which (1) preserves risk aversion despite the soft risk; (2) independently improves sample efficiency. By separating the risk aversion of the sampler and the optimizer, we can sample episodes with poor conditions, yet optimize with respect to successful strategies. We combine these two concepts in CeSoR - Cross-entropy Soft-Risk optimization algorithm - which can be applied on top of any risk-averse policy gradient (PG) method. We demonstrate improved risk aversion in maze navigation, autonomous driving, and resource allocation benchmarks, including in scenarios where standard risk-averse PG completely fails."
                    },
                    {
                        "title": "SAFER: Data-Efficient and Safe Reinforcement Learning via Skill Acquisition",
                        "abstract": "Methods that extract policy primitives from offline demonstrations using deep generative models have shown promise at accelerating reinforcement learning(RL) for new tasks. Intuitively, these methods should also help to trainsafeRLagents because they enforce useful skills. However, we identify these techniques are not well equipped for safe policy learning because they ignore negative experiences(e.g., unsafe or unsuccessful), focusing only on positive experiences, which harms their ability to generalize to new tasks safely. Rather, we model the latentsafetycontextusing principled contrastive training on an offline dataset of demonstrations from many tasks, including both negative and positive experiences. Using this late variable, our RL framework, SAFEty skill pRiors (SAFER) extracts task-specific safe primitive skills to safely and successfully generalize to new tasks. In the inference stage, policies trained with SAFER learn to compose safe skills into successful policies. We theoretically characterize why SAFER can enforce safe policy learning and demonstrate its effectiveness on several complex safety-critical robotic grasping tasks inspired by the game Operation, in which SAFERoutperforms state-of-the-art primitive learning methods in success and safety."
                    },
                    {
                        "title": "Non-Stationary Off-Policy Optimization",
                        "abstract": "Off-policy learning is a framework for evaluating and optimizing policies without deploying them, from data collected by another policy. Real-world environments are typically non-stationary and the of\ufb02ine learned policies should adapt to these changes. To address this challenge, we study the novel problem of off-policy optimization in piecewise-stationary contextual ban-dits. Our proposed solution has two phases. In the of\ufb02ine learning phase, we partition logged data into categorical latent states and learn a near-optimal sub-policy for each state. In the online deployment phase, we adaptively switch between the learned sub-policies based on their performance. This approach is practical and analyzable, and we provide guarantees on both the quality of off-policy optimization and the regret during online deployment. To show the effectiveness of our approach, we compare it to state-of-the-art baselines on both synthetic and real-world datasets. Our approach outperforms methods that act only on observed context."
                    },
                    {
                        "title": "Non-Stationary Latent Bandits",
                        "abstract": "Users of recommender systems often behave in a non-stationary fashion, due to their evolving preferences and tastes over time. In this work, we propose a practical approach for fast personalization to non-stationary users. The key idea is to frame this problem as a latent bandit, where the prototypical models of user behavior are learned offline and the latent state of the user is inferred online from its interactions with the models. We call this problem a non-stationary latent bandit. We propose Thompson sampling algorithms for regret minimization in non-stationary latent bandits, analyze them, and evaluate them on a real-world dataset. The main strength of our approach is that it can be combined with rich offline-learned models, which can be misspecified, and are subsequently fine-tuned online using posterior sampling. In this way, we naturally combine the strengths of offline and online learning."
                    },
                    {
                        "title": "CoinDICE: Off-Policy Confidence Interval Estimation",
                        "abstract": "We study high-confidence behavior-agnostic off-policy evaluation in reinforcement learning, where the goal is to estimate a confidence interval on a target policy's value, given only access to a static experience dataset collected by unknown behavior policies. Starting from a function space embedding of the linear program formulation of the $Q$-function, we obtain an optimization problem with generalized estimating equation constraints. By applying the generalized empirical likelihood method to the resulting Lagrangian, we propose CoinDICE, a novel and efficient algorithm for computing confidence intervals. Theoretically, we prove the obtained confidence intervals are valid, in both asymptotic and finite-sample regimes. Empirically, we show in a variety of benchmarks that the confidence interval estimates are tighter and more accurate than existing methods."
                    },
                    {
                        "title": "Latent Bandits Revisited",
                        "abstract": "A latent bandit problem is one in which the learning agent knows the arm reward distributions conditioned on an unknown discrete latent state. The primary goal of the agent is to identify the latent state, after which it can act optimally. This setting is a natural midpoint between online and offline learning---complex models can be learned offline with the agent identifying latent state online---of practical relevance in, say, recommender systems. In this work, we propose general algorithms for this setting, based on both upper confidence bounds (UCBs) and Thompson sampling. Our methods are contextual and aware of model uncertainty and misspecification. We provide a unified theoretical analysis of our algorithms, which have lower regret than classic bandit policies when the number of latent states is smaller than actions. A comprehensive empirical study showcases the advantages of our approach."
                    },
                    {
                        "title": "Safe Reinforcement Learning with Natural Language Constraints",
                        "abstract": "In this paper, we tackle the problem of learning control policies for tasks when provided with constraints in natural language. In contrast to instruction following, language here is used not to specify goals, but rather to describe situations that an agent must avoid during its exploration of the environment. Specifying constraints in natural language also differs from the predominant paradigm in safe reinforcement learning, where safety criteria are enforced by hand-defined cost functions. While natural language allows for easy and flexible specification of safety constraints and budget limitations, its ambiguous nature presents a challenge when mapping these specifications into representations that can be used by techniques for safe reinforcement learning. To address this, we develop a model that contains two components: (1) a constraint interpreter to encode natural language constraints into vector representations capturing spatial and temporal information on forbidden states, and (2) a policy network that uses these representations to output a policy with minimal constraint violations. Our model is end-to-end differentiable and we train it using a recently proposed algorithm for constrained policy optimization. To empirically demonstrate the effectiveness of our approach, we create a new benchmark task for autonomous navigation with crowd-sourced free-form text specifying three different types of constraints. Our method outperforms several baselines by achieving 6-7 times higher returns and 76% fewer constraint violations on average. Dataset and code to reproduce our experiments are available at this https URL."
                    },
                    {
                        "title": "Predictive Coding for Locally-Linear Control",
                        "abstract": "High-dimensional observations and unknown dynamics are major challenges when applying optimal control to many real-world decision making tasks. The Learning Controllable Embedding (LCE) framework addresses these challenges by embedding the observations into a lower dimensional latent space, estimating the latent dynamics, and then performing control directly in the latent space. To ensure the learned latent dynamics are predictive of next-observations, all existing LCE approaches decode back into the observation space and explicitly perform next-observation prediction---a challenging high-dimensional task that furthermore introduces a large number of nuisance parameters (i.e., the decoder) which are discarded during control. In this paper, we propose a novel information-theoretic LCE approach and show theoretically that explicit next-observation prediction can be replaced with predictive coding. We then use predictive coding to develop a decoder-free LCE model whose latent dynamics are amenable to locally-linear control. Extensive experiments on benchmark tasks show that our model reliably learns a controllable latent space that leads to superior performance when compared with state-of-the-art LCE baselines."
                    },
                    {
                        "title": "Piecewise-Stationary Off-Policy Optimization",
                        "abstract": "Off-policy learning is a framework for evaluating and optimizing policies without deploying them, from data collected by another policy. Real-world environments are typically non-stationary and the offline learned policies should adapt to these changes. To address this challenge, we study the novel problem of off-policy optimization in piecewise-stationary contextual bandits. Our proposed solution has two phases. In the offline learning phase, we partition logged data into categorical latent states and learn a near-optimal sub-policy for each state. In the online deployment phase, we adaptively switch between the learned sub-policies based on their performance. This approach is practical and analyzable, and we provide guarantees on both the quality of off-policy optimization and the regret during online deployment. To show the effectiveness of our approach, we compare it to state-of-the-art baselines on both synthetic and real-world datasets. Our approach outperforms methods that act only on observed context."
                    },
                    {
                        "title": "Control-Aware Representations for Model-based Reinforcement Learning",
                        "abstract": "A major challenge in modern reinforcement learning (RL) is efficient control of dynamical systems from high-dimensional sensory observations. Learning controllable embedding (LCE) is a promising approach that addresses this challenge by embedding the observations into a lower-dimensional latent space, estimating the latent dynamics, and utilizing it to perform control in the latent space. Two important questions in this area are how to learn a representation that is amenable to the control problem at hand, and how to achieve an end-to-end framework for representation learning and control. In this paper, we take a few steps towards addressing these questions. We first formulate a LCE model to learn representations that are suitable to be used by a policy iteration style algorithm in the latent space. We call this model control-aware representation learning (CARL). We derive a loss function for CARL that has close connection to the prediction, consistency, and curvature (PCC) principle for representation learning. We derive three implementations of CARL. In the offline implementation, we replace the locally-linear control algorithm (e.g.,~iLQR) used by the existing LCE methods with a RL algorithm, namely model-based soft actor-critic, and show that it results in significant improvement. In online CARL, we interleave representation learning and control, and demonstrate further gain in performance. Finally, we propose value-guided CARL, a variation in which we optimize a weighted version of the CARL loss function, where the weights depend on the TD-error of the current policy. We evaluate the proposed algorithms by extensive experiments on benchmark tasks and compare them with several LCE baselines."
                    },
                    {
                        "title": "Variational Model-based Policy Optimization",
                        "abstract": "Model-based reinforcement learning (RL) algorithms allow us to combine model-generated data with those collected from interaction with the real system in order to alleviate the data efficiency problem in RL. However, designing such algorithms is often challenging because the bias in simulated data may overshadow the ease of data generation. A potential solution to this challenge is to jointly learn and improve model and policy using a universal objective function. In this paper, we leverage the connection between RL and probabilistic inference, and formulate such an objective function as a variational lower-bound of a log-likelihood. This allows us to use expectation maximization (EM) and iteratively fix a baseline policy and learn a variational distribution, consisting of a model and a policy (E-step), followed by improving the baseline policy given the learned variational distribution (M-step). We propose model-based and model-free policy iteration (actor-critic) style algorithms for the E-step and show how the variational distribution learned by them can be used to optimize the M-step in a fully model-based fashion. Our experiments on a number of continuous control tasks show that our model-based (E-step) algorithm, called variational model-based policy optimization (VMBPO), is more sample-efficient and robust to hyper-parameter tuning than its model-free (E-step) counterpart. Using the same control tasks, we also compare VMBPO with several state-of-the-art model-based and model-free RL algorithms and show its sample efficiency and performance."
                    }
                ]
            },
            "003c13e0-deac-4b83-879c-7b300a7ca195": {
                "pk": "003c13e0-deac-4b83-879c-7b300a7ca195",
                "name": "Chih-Wei Hsu",
                "collaborators": [
                    "Yinlam Chow",
                    "Craig Boutilier",
                    "Guy Tennenholtz",
                    "Erdem Biyik",
                    "M. Ghavamzadeh",
                    "Anthony Liang",
                    "Lior Shani",
                    "Ethan Liang",
                    "Martin Mladenov",
                    "Ofer Meshi",
                    "James Pine",
                    "Hubert Pham",
                    "Shane Li",
                    "Xujian Liang",
                    "Anton Polishko",
                    "Li Yang",
                    "Ben Scheetz",
                    "C. Boutilier",
                    "Fan Yao",
                    "Alex Haig",
                    "Jihwan Jeong",
                    "Azamat Tulepbergenov",
                    "Christina G\u00f6pfert",
                    "Ivan Vendrov",
                    "Tyler Lu",
                    "Deepak Ramachandran"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Natural Language Processing",
                    "Recommender Systems",
                    "Meta-Learning"
                ],
                "publications": [
                    {
                        "title": "DynaMITE-RL: A Dynamic Model for Improved Temporal Meta-Reinforcement Learning",
                        "abstract": "We introduce DynaMITE-RL, a meta-reinforcement learning (meta-RL) approach to approximate inference in environments where the latent state evolves at varying rates. We model episode sessions - parts of the episode where the latent state is fixed - and propose three key modifications to existing meta-RL methods: consistency of latent information within sessions, session masking, and prior latent conditioning. We demonstrate the importance of these modifications in various domains, ranging from discrete Gridworld environments to continuous-control and simulated robot assistive tasks, demonstrating that DynaMITE-RL significantly outperforms state-of-the-art baselines in sample efficiency and inference returns."
                    },
                    {
                        "title": "Embedding-Aligned Language Models",
                        "abstract": "We propose a novel approach for training large language models (LLMs) to adhere to objectives defined within a latent embedding space. Our method leverages reinforcement learning (RL), treating a pre-trained LLM as an environment. Our embedding-aligned guided language (EAGLE) agent is trained to iteratively steer the LLM's generation towards optimal regions of the latent embedding space, w.r.t. some predefined criterion. We demonstrate the effectiveness of the EAGLE agent using the MovieLens 25M and Amazon Review datasets to surface content gaps that satisfy latent user demand. We also demonstrate the benefit of using an optimal design of a state-dependent action set to improve EAGLE's efficiency. Our work paves the way for controlled and grounded text generation using LLMs, ensuring consistency with domain-specific knowledge and data representations."
                    },
                    {
                        "title": "Minimizing Live Experiments in Recommender Systems: User Simulation to Evaluate Preference Elicitation Policies",
                        "abstract": "Evaluation of policies in recommender systems typically involves A/B testing using live experiments on real users to assess a new policy's impact on relevant metrics. This ``gold standard'' comes at a high cost, however, in terms of cycle time, user cost, and potential user retention. In developing policies for ``onboarding'' new users, these costs can be especially problematic, since on-boarding occurs only once. In this work, we describe a simulation methodology used to augment (and reduce) the use of live experiments. We illustrate its deployment for the evaluation of ``preference elicitation'' algorithms used to onboard new users of the YouTube Music platform. By developing counterfactually robust user behavior models, and a simulation service that couples such models with production infrastructure, we are able to test new algorithms in a way that reliably predicts their performance on key metrics when deployed live. We describe our domain, our simulation models and platform, results of experiments and deployment, and suggest future steps needed to further realistic simulation as a powerful complement to live experiments."
                    },
                    {
                        "title": "Preference Elicitation with Soft Attributes in Interactive Recommendation",
                        "abstract": "Preference elicitation plays a central role in interactive recommender systems. Most preference elicitation approaches use either item queries that ask users to select preferred items from a slate, or attribute queries that ask them to express their preferences for item characteristics. Unfortunately, users often wish to describe their preferences using soft attributes for which no ground-truth semantics is given. Leveraging concept activation vectors for soft attribute semantics, we develop novel preference elicitation methods that can accommodate soft attributes and bring together both item and attribute-based preference elicitation. Our techniques query users using both items and soft attributes to update the recommender system's belief about their preferences to improve recommendation quality. We demonstrate the effectiveness of our methods vis-a-vis competing approaches on both synthetic and real-world datasets."
                    },
                    {
                        "title": "Factual and Personalized Recommendations using Language Models and Reinforcement Learning",
                        "abstract": "Recommender systems (RSs) play a central role in connecting users to content, products, and services, matching candidate items to users based on their preferences. While traditional RSs rely on implicit user feedback signals, conversational RSs interact with users in natural language. In this work, we develop a comPelling, Precise, Personalized, Preference-relevant language model (P4LM) that recommends items to users while putting emphasis on explaining item characteristics and their relevance. P4LM uses the embedding space representation of a user's preferences to generate compelling responses that are factually-grounded and relevant w.r.t. the user's preferences. Moreover, we develop a joint reward function that measures precision, appeal, and personalization, which we use as AI-based feedback in a reinforcement learning-based language model framework. Using the MovieLens 25M dataset, we demonstrate that P4LM delivers compelling, personalized movie narratives to users."
                    },
                    {
                        "title": "Discovering Personalized Semantics for Soft Attributes in Recommender Systems using Concept Activation Vectors",
                        "abstract": "Interactive recommender systems (RSs) allow users to express intent, preferences and contexts in a rich fashion, often using natural language. One challenge in using such feedback is inferring a user\u2019s semantic intent from the open-ended terms used to describe an item, and using it to refine recommendation results. Leveraging concept activation vectors (CAVs) [21], we develop a framework to learn a representation that captures the semantics of such attributes and connects them to user preferences and behaviors in RSs. A novel feature of our approach is its ability to distinguish objective and subjective attributes and associate different senses with different users. Using synthetic and real-world datasets, we show that our CAV representation accurately interprets users\u2019 subjective semantics, and can improve recommendations via interactive critiquing."
                    }
                ]
            },
            "a971f57d-bd00-49bb-bcd8-6cd573f6968f": {
                "pk": "a971f57d-bd00-49bb-bcd8-6cd573f6968f",
                "name": "Jihwan Jeong",
                "collaborators": [
                    "Yinlam Chow",
                    "Guy Tennenholtz",
                    "Chih-Wei Hsu",
                    "Azamat Tulepbergenov",
                    "M. Ghavamzadeh",
                    "Craig Boutilier"
                ],
                "domain": [
                    "Recommender Systems",
                    "Natural Language Processing",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Factual and Personalized Recommendations using Language Models and Reinforcement Learning",
                        "abstract": "Recommender systems (RSs) play a central role in connecting users to content, products, and services, matching candidate items to users based on their preferences. While traditional RSs rely on implicit user feedback signals, conversational RSs interact with users in natural language. In this work, we develop a comPelling, Precise, Personalized, Preference-relevant language model (P4LM) that recommends items to users while putting emphasis on explaining item characteristics and their relevance. P4LM uses the embedding space representation of a user's preferences to generate compelling responses that are factually-grounded and relevant w.r.t. the user's preferences. Moreover, we develop a joint reward function that measures precision, appeal, and personalization, which we use as AI-based feedback in a reinforcement learning-based language model framework. Using the MovieLens 25M dataset, we demonstrate that P4LM delivers compelling, personalized movie narratives to users."
                    }
                ]
            },
            "918777ab-f6ae-48a5-852a-bfbb2165fa14": {
                "pk": "918777ab-f6ae-48a5-852a-bfbb2165fa14",
                "name": "Lior Shani",
                "collaborators": [
                    "Shie Mannor",
                    "Yonathan Efroni",
                    "Guy Tennenholtz",
                    "Daniele Calandriello",
                    "Orgad Keller",
                    "Bilal Piot",
                    "Idan Szpektor",
                    "Avinatan Hassidim",
                    "R\u00e9mi Munos",
                    "Craig Boutilier",
                    "Nadav Merlis",
                    "Aviv Rosenberg",
                    "Asaf B. Cassel",
                    "Oran Lang",
                    "Avital Zipori",
                    "Hila Noga",
                    "Yossi Matias",
                    "Pierre H. Richemond",
                    "Yunhao Tang",
                    "Daniel Guo",
                    "M. G. Azar",
                    "Rafael Rafailov",
                    "B. '. Pires",
                    "Eugene Tarassov",
                    "Lucas Spangher",
                    "Will Ellsworth",
                    "Aliaksei Severyn",
                    "Jonathan Mallinson",
                    "Gil Shamir",
                    "Rishabh Joshi",
                    "Tianqi Liu",
                    "Yinlam Chow",
                    "Chih-Wei Hsu",
                    "Ethan Liang",
                    "Martin Mladenov",
                    "Paul Roit",
                    "Johan Ferret",
                    "Roee Aharoni",
                    "Geoffrey Cideron",
                    "Robert Dadashi",
                    "M. Geist",
                    "Sertan Girgin",
                    "L'eonard Hussenot",
                    "Nikola Momchev",
                    "Sabela Ramos",
                    "P. Sta\u0144czyk",
                    "Nino Vieillard",
                    "Olivier Bachem",
                    "G. Elidan",
                    "O. Pietquin",
                    "Uri Shalit",
                    "Gal Chechik",
                    "Assaf Hallak",
                    "Gal Dalal",
                    "Tom Zahavy",
                    "Manan Tomar",
                    "M. Ghavamzadeh",
                    "Aviv A. Rosenberg",
                    "Mark Kozdoba",
                    "E. Moroshko",
                    "Takuya Takagi",
                    "Takashi Katoh",
                    "K. Crammer"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Large Language Models",
                    "Policy Optimization",
                    "Multi-Instance Learning"
                ],
                "publications": [
                    {
                        "title": "Multi-turn Reinforcement Learning from Preference Human Feedback",
                        "abstract": "Reinforcement Learning from Human Feedback (RLHF) has become the standard approach for aligning Large Language Models (LLMs) with human preferences, allowing LLMs to demonstrate remarkable abilities in various tasks. Existing methods work by emulating the preferences at the single decision (turn) level, limiting their capabilities in settings that require planning or multi-turn interactions to achieve a long-term goal. In this paper, we address this issue by developing novel methods for Reinforcement Learning (RL) from preference feedback between two full multi-turn conversations. In the tabular setting, we present a novel mirror-descent-based policy optimization algorithm for the general multi-turn preference-based RL problem, and prove its convergence to Nash equilibrium. To evaluate performance, we create a new environment, Education Dialogue, where a teacher agent guides a student in learning a random topic, and show that a deep RL variant of our algorithm outperforms RLHF baselines. Finally, we show that in an environment with explicit rewards, our algorithm recovers the same performance as a reward-based RL baseline, despite relying solely on a weaker preference signal."
                    },
                    {
                        "title": "Offline Regularised Reinforcement Learning for Large Language Models Alignment",
                        "abstract": "The dominant framework for alignment of large language models (LLM), whether through reinforcement learning from human feedback or direct preference optimisation, is to learn from preference data. This involves building datasets where each element is a quadruplet composed of a prompt, two independent responses (completions of the prompt) and a human preference between the two independent responses, yielding a preferred and a dis-preferred response. Such data is typically scarce and expensive to collect. On the other hand, \\emph{single-trajectory} datasets where each element is a triplet composed of a prompt, a response and a human feedback is naturally more abundant. The canonical element of such datasets is for instance an LLM's response to a user's prompt followed by a user's feedback such as a thumbs-up/down. Consequently, in this work, we propose DRO, or \\emph{Direct Reward Optimisation}, as a framework and associated algorithms that do not require pairwise preferences. DRO uses a simple mean-squared objective that can be implemented in various ways. We validate our findings empirically, using T5 encoder-decoder language models, and show DRO's performance over selected baselines such as Kahneman-Tversky Optimization (KTO). Thus, we confirm that DRO is a simple and empirically compelling method for single-trajectory policy optimisation."
                    },
                    {
                        "title": "Embedding-Aligned Language Models",
                        "abstract": "We propose a novel approach for training large language models (LLMs) to adhere to objectives defined within a latent embedding space. Our method leverages reinforcement learning (RL), treating a pre-trained LLM as an environment. Our embedding-aligned guided language (EAGLE) agent is trained to iteratively steer the LLM's generation towards optimal regions of the latent embedding space, w.r.t. some predefined criterion. We demonstrate the effectiveness of the EAGLE agent using the MovieLens 25M and Amazon Review datasets to surface content gaps that satisfy latent user demand. We also demonstrate the benefit of using an optimal design of a state-dependent action set to improve EAGLE's efficiency. Our work paves the way for controlled and grounded text generation using LLMs, ensuring consistency with domain-specific knowledge and data representations."
                    },
                    {
                        "title": "Reinforcement Learning with History-Dependent Dynamic Contexts",
                        "abstract": "We introduce Dynamic Contextual Markov Decision Processes (DCMDPs), a novel reinforcement learning framework for history-dependent environments that generalizes the contextual MDP framework to handle non-Markov environments, where contexts change over time. We consider special cases of the model, with a focus on logistic DCMDPs, which break the exponential dependence on history length by leveraging aggregation functions to determine context transitions. This special structure allows us to derive an upper-confidence-bound style algorithm for which we establish regret bounds. Motivated by our theoretical results, we introduce a practical model-based algorithm for logistic DCMDPs that plans in a latent space and uses optimism over history-dependent features. We demonstrate the efficacy of our approach on a recommendation task (using MovieLens data) where user behavior dynamics evolve in response to recommendations."
                    },
                    {
                        "title": "Factually Consistent Summarization via Reinforcement Learning with Textual Entailment Feedback",
                        "abstract": "Despite the seeming success of contemporary grounded text generation systems, they often tend to generate factually inconsistent text with respect to their input.This phenomenon is emphasized in tasks like summarization, in which the generated summaries should be corroborated by their source article. In this work we leverage recent progress on textual entailment models to directly address this problem for abstractive summarization systems.We use reinforcement learning with reference-free, textual-entailment rewards to optimize for factual consistency and explore the ensuing trade-offs, as improved consistency may come at the cost of less informative or more extractive summaries.Our results, according to both automatic metrics and human evaluation, show that our method considerably improves the faithfulness, salience and conciseness of the generated summaries."
                    },
                    {
                        "title": "Reinforcement Learning with a Terminator",
                        "abstract": "We present the problem of reinforcement learning with exogenous termination. We define the Termination Markov Decision Process (TerMDP), an extension of the MDP framework, in which episodes may be interrupted by an external non-Markovian observer. This formulation accounts for numerous real-world situations, such as a human interrupting an autonomous driving agent for reasons of discomfort. We learn the parameters of the TerMDP and leverage the structure of the estimation problem to provide state-wise confidence bounds. We use these to construct a provably-efficient algorithm, which accounts for termination, and bound its regret. Motivated by our theoretical analysis, we design and implement a scalable approach, which combines optimism (w.r.t. termination) and a dynamic discount factor, incorporating the termination probability. We deploy our method on high-dimensional driving and MinAtar benchmarks. Additionally, we test our approach on human data in a driving setting. Our results demonstrate fast convergence and significant improvement over various baseline approaches."
                    },
                    {
                        "title": "Online Apprenticeship Learning",
                        "abstract": "In Apprenticeship Learning (AL), we are given a Markov Decision Process (MDP) without access to the cost function. Instead, we observe trajectories sampled by an expert that acts according to some policy. The goal is to find a policy that matches the expert's performance on some predefined set of cost functions.  We introduce an online variant of AL (Online Apprenticeship Learning; OAL), where the agent is expected to perform comparably to the expert while interacting with the environment. We show that the OAL problem can be effectively solved by combining two mirror descent based no-regret algorithms: one for policy optimization and another for learning the worst case cost. By employing optimistic exploration, we derive a convergent algorithm with O(sqrt(K)) regret, where K is the number of interactions with the MDP, and an additional linear error term that depends on the amount of expert trajectories available. Importantly, our algorithm avoids the need to solve an MDP at each iteration, making it more practical compared to prior AL methods. Finally, we implement a deep variant of our algorithm which shares some similarities to GAIL, but where the discriminator is replaced with the costs learned by OAL. Our simulations suggest that OAL performs well in high dimensional control problems."
                    },
                    {
                        "title": "Mirror Descent Policy Optimization",
                        "abstract": "We propose deep Reinforcement Learning (RL) algorithms inspired by mirror descent, a well-known first-order trust region optimization method for solving constrained convex problems. Our approach, which we call as Mirror Descent Policy Optimization (MDPO), is based on the idea of iteratively solving a `trust-region' problem that minimizes a sum of two terms: a linearization of the objective function and a proximity term that restricts two consecutive updates to be close to each other. Following this approach we derive on-policy and off-policy variants of the MDPO algorithm and analyze their performance while emphasizing important implementation details, motivated by the existing theoretical framework. We highlight the connections between on-policy MDPO and two popular trust region RL algorithms: TRPO and PPO, and conduct a comprehensive empirical comparison of these algorithms. We then derive off-policy MDPO and compare its performance to existing approaches. Importantly, we show that the theoretical framework of MDPO can be scaled to deep RL while achieving good performance on popular benchmarks."
                    },
                    {
                        "title": "Optimistic Policy Optimization with Bandit Feedback",
                        "abstract": "Policy optimization methods are one of the most widely used classes of Reinforcement Learning (RL) algorithms. Yet, so far, such methods have been mostly analyzed from an optimization perspective, without addressing the problem of exploration, or by making strong assumptions on the interaction with the environment. In this paper we consider model-based RL in the tabular finite-horizon MDP setting with unknown transitions and bandit feedback. For this setting, we propose an optimistic trust region policy optimization (TRPO) algorithm for which we establish $\\tilde O(\\sqrt{S^2 A H^4 K})$ regret for stochastic rewards. Furthermore, we prove $\\tilde O( \\sqrt{ S^2 A H^4 } K^{2/3} ) $ regret for adversarial rewards. Interestingly, this result matches previous bounds derived for the bandit feedback case, yet with known transitions. To the best of our knowledge, the two results are the first sub-linear regret bounds obtained for policy optimization algorithms with unknown transitions and bandit feedback."
                    },
                    {
                        "title": "Adaptive Trust Region Policy Optimization: Global Convergence and Faster Rates for Regularized MDPs",
                        "abstract": "Trust region policy optimization (TRPO) is a popular and empirically successful policy search algorithm in Reinforcement Learning (RL) in which a surrogate problem, that restricts consecutive policies to be \u2018close\u2019 to one another, is iteratively solved. Nevertheless, TRPO has been considered a heuristic algorithm inspired by Conservative Policy Iteration (CPI). We show that the adaptive scaling mechanism used in TRPO is in fact the natural \u201cRL version\u201d of traditional trust-region methods from convex analysis. We first analyze TRPO in the planning setting, in which we have access to the model and the entire state space. Then, we consider sample-based TRPO and establish \u00d5(1/\u221aN) convergence rate to the global optimum. Importantly, the adaptive scaling mechanism allows us to analyze TRPO in regularized MDPs for which we prove fast rates of \u00d5(1/N), much like results in convex optimization. This is the first result in RL of better rates when regularizing the instantaneous cost or reward."
                    },
                    {
                        "title": "Exploration Conscious Reinforcement Learning Revisited",
                        "abstract": "The Exploration-Exploitation tradeoff arises in Reinforcement Learning when one cannot tell if a policy is optimal. Then, there is a constant need to explore new actions instead of exploiting past experience. In practice, it is common to resolve the tradeoff by using a fixed exploration mechanism, such as $\\epsilon$-greedy exploration or by adding Gaussian noise, while still trying to learn an optimal policy. In this work, we take a different approach and study exploration-conscious criteria, that result in optimal policies with respect to the exploration mechanism. Solving these criteria, as we establish, amounts to solving a surrogate Markov Decision Process. We continue and analyze properties of exploration-conscious optimal policies and characterize two general approaches to solve such criteria. Building on the approaches, we apply simple changes in existing tabular and deep Reinforcement Learning algorithms and empirically demonstrate superior performance relatively to their non-exploration-conscious counterparts, both for discrete and continuous action spaces."
                    },
                    {
                        "title": "Revisiting Exploration-Conscious Reinforcement Learning",
                        "abstract": "The objective of Reinforcement Learning is to learn an optimal policy by performing actions and observing their long term consequences. Unfortunately, acquiring such a policy can be a hard task. More severely, since one cannot tell if a policy is optimal, there is a constant need for exploration. This is known as the Exploration-Exploitation trade-off. In practice, this trade-off is resolved by using some inherent exploration mechanism, such as the $\\epsilon$-greedy exploration, while still trying to learn the optimal policy. In this work, we take a different approach. We define a surrogate optimality objective: an optimal policy with respect to the exploration scheme. As we show throughout the paper, although solving this criterion does not necessarily lead to an optimal policy, the problem becomes easier to solve. We continue by analyzing this notion of optimality, devise algorithms derived from this approach, which reveal connections to existing work, and test them empirically on tabular and deep Reinforcement Learning domains."
                    },
                    {
                        "title": "Multi Instance Learning For Unbalanced Data",
                        "abstract": "In the context of Multi Instance Learning, we analyze the Single Instance (SI) learning objective. We show that when the data is unbalanced and the family of classifiers is sufficiently rich, the SI method is a useful learning algorithm. In particular, we show that larger data imbalance, a quality that is typically perceived as negative, in fact implies a better resilience of the algorithm to the statistical dependencies of the objects in bags. In addition, our results shed new light on some known issues with the SI method in the setting of linear classifiers, and we show that these issues are significantly less likely to occur in the setting of neural networks. We demonstrate our results on a synthetic dataset, and on the COCO dataset for the problem of patch classification with weak image level labels derived from captions."
                    }
                ]
            },
            "a26edd72-cadb-4723-95aa-ecb3390ae604": {
                "pk": "a26edd72-cadb-4723-95aa-ecb3390ae604",
                "name": "Azamat Tulepbergenov",
                "collaborators": [
                    "Yinlam Chow",
                    "M. Ghavamzadeh",
                    "Craig Boutilier",
                    "Jihwan Jeong",
                    "Guy Tennenholtz",
                    "Chih-Wei Hsu",
                    "Ofir Nachum",
                    "M. Ryu"
                ],
                "domain": [
                    "Recommender Systems",
                    "Natural Language Processing",
                    "Reinforcement Learning",
                    "Dialogue Management"
                ],
                "publications": [
                    {
                        "title": "Factual and Personalized Recommendations using Language Models and Reinforcement Learning",
                        "abstract": "Recommender systems (RSs) play a central role in connecting users to content, products, and services, matching candidate items to users based on their preferences. While traditional RSs rely on implicit user feedback signals, conversational RSs interact with users in natural language. In this work, we develop a comPelling, Precise, Personalized, Preference-relevant language model (P4LM) that recommends items to users while putting emphasis on explaining item characteristics and their relevance. P4LM uses the embedding space representation of a user's preferences to generate compelling responses that are factually-grounded and relevant w.r.t. the user's preferences. Moreover, we develop a joint reward function that measures precision, appeal, and personalization, which we use as AI-based feedback in a reinforcement learning-based language model framework. Using the MovieLens 25M dataset, we demonstrate that P4LM delivers compelling, personalized movie narratives to users."
                    },
                    {
                        "title": "A Mixture-of-Expert Approach to RL-based Dialogue Management",
                        "abstract": "Despite recent advancements in language models (LMs), their application to dialogue management (DM) problems and ability to carry on rich conversations remain a challenge. We use reinforcement learning (RL) to develop a dialogue agent that avoids being short-sighted (outputting generic utterances) and maximizes overall user satisfaction. Most existing RL approaches to DM train the agent at the word-level, and thus, have to deal with a combinatorially complex action space even for a medium-size vocabulary. As a result, they struggle to produce a successful and engaging dialogue even if they are warm-started with a pre-trained LM. To address this issue, we develop a RL-based DM using a novel mixture of expert language model (MoE-LM) that consists of (i) a LM capable of learning diverse semantics for conversation histories, (ii) a number of {\\em specialized} LMs (or experts) capable of generating utterances corresponding to a particular attribute or personality, and (iii) a RL-based DM that performs dialogue planning with the utterances generated by the experts. Our MoE approach provides greater flexibility to generate sensible utterances with different intents and allows RL to focus on conversational-level DM. We compare it with SOTA baselines on open-domain dialogues and demonstrate its effectiveness both in terms of the diversity and sensibility of the generated utterances and the overall DM performance."
                    }
                ]
            },
            "aab8fcf6-03f0-48fd-9fcd-69aee3729a2a": {
                "pk": "aab8fcf6-03f0-48fd-9fcd-69aee3729a2a",
                "name": "Deepak Ramachandran",
                "collaborators": [
                    "Christina G\u00f6pfert",
                    "Yinlam Chow",
                    "Chih-Wei Hsu",
                    "Ivan Vendrov",
                    "Tyler Lu",
                    "Craig Boutilier"
                ],
                "domain": [
                    "Recommender Systems",
                    "Natural Language Processing",
                    "User Intent",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Discovering Personalized Semantics for Soft Attributes in Recommender Systems using Concept Activation Vectors",
                        "abstract": "Interactive recommender systems (RSs) allow users to express intent, preferences and contexts in a rich fashion, often using natural language. One challenge in using such feedback is inferring a user\u2019s semantic intent from the open-ended terms used to describe an item, and using it to refine recommendation results. Leveraging concept activation vectors (CAVs) [21], we develop a framework to learn a representation that captures the semantics of such attributes and connects them to user preferences and behaviors in RSs. A novel feature of our approach is its ability to distinguish objective and subjective attributes and associate different senses with different users. Using synthetic and real-world datasets, we show that our CAV representation accurately interprets users\u2019 subjective semantics, and can improve recommendations via interactive critiquing."
                    }
                ]
            },
            "df14f49b-eca1-4784-99b8-c53df9b10b24": {
                "pk": "df14f49b-eca1-4784-99b8-c53df9b10b24",
                "name": "Martin Mladenov",
                "collaborators": [
                    "Craig Boutilier",
                    "Chih-Wei Hsu",
                    "Guy Tennenholtz",
                    "B. Kveton",
                    "M. Zaheer",
                    "Csaba Szepesvari",
                    "Nadav Merlis",
                    "Ed H. Chi",
                    "Minmin Chen",
                    "Vihan Jain",
                    "Eugene Ie",
                    "Christopher Colby",
                    "Nicolas Mayoraz",
                    "H. Pham",
                    "Dustin Tran",
                    "Ivan Vendrov",
                    "Ofer Meshi",
                    "James Pine",
                    "Hubert Pham",
                    "Shane Li",
                    "Xujian Liang",
                    "Anton Polishko",
                    "Li Yang",
                    "Ben Scheetz",
                    "C. Boutilier",
                    "Lior Shani",
                    "Abhishek Naik",
                    "B. Chang",
                    "Alexandros Karatzoglou",
                    "Siddharth Prasad",
                    "Sanjay Ganapathy Subramaniam",
                    "Neha Arora",
                    "Andrew",
                    "Tomkins",
                    "C. Osorio",
                    "Ayal Taitler",
                    "Michael Gimelfarb",
                    "Sriram Gopalakrishnan",
                    "Xiaotian Liu",
                    "S. Sanner",
                    "Martin Grohe",
                    "K. Kersting",
                    "Pascal Schweitzer",
                    "Ruohan Zhan",
                    "Konstantina Christakopoulou",
                    "Ya Le",
                    "Jayden Ooi",
                    "Alex Beutel",
                    "Mikhail Konobeev",
                    "Elliot Creager",
                    "Omer Ben-Porat",
                    "Kevin Swersky",
                    "R. Zemel"
                ],
                "domain": [
                    "Recommender Systems",
                    "Reinforcement Learning",
                    "Mechanism Design",
                    "Multi-Agent Systems"
                ],
                "publications": [
                    {
                        "title": "Recommender Ecosystems: A Mechanism Design Perspective on Holistic Modeling and Optimization",
                        "abstract": "Modern recommender systems lie at the heart of complex recommender ecosystems that couple the behavior of users, content providers, vendors, advertisers, and other actors. Despite this, the focus of much recommender systems research and deployment is on the local, myopic optimization of the recommendations made to individual users. This comes at a significant cost to the long-term utility that recommender systems generate for their users. We argue that modeling the incentives and behaviors of these actors, and the interactions among them induced by the recommender systems, is needed to maximize value and improve overall ecosystem health. Moreover, we propose the use of economic mechanism design, an area largely overlooked in recommender systems research, as a framework for developing such models. That said, one cannot apply \u201cvanilla\u201d mechanism design to recommender ecosystem modeling optimization out of the box\u2014the use of mechanism design raises a number of subtle and interesting research challenges. We outline a number of these in this talk (and paper), emphasizing the need to develop nonstandard approaches to mechanism design that intersect with numerous areas of research, including preference modeling, reinforcement learning and exploration, behavioral economics, and generative AI, among others."
                    },
                    {
                        "title": "Minimizing Live Experiments in Recommender Systems: User Simulation to Evaluate Preference Elicitation Policies",
                        "abstract": "Evaluation of policies in recommender systems typically involves A/B testing using live experiments on real users to assess a new policy's impact on relevant metrics. This ``gold standard'' comes at a high cost, however, in terms of cycle time, user cost, and potential user retention. In developing policies for ``onboarding'' new users, these costs can be especially problematic, since on-boarding occurs only once. In this work, we describe a simulation methodology used to augment (and reduce) the use of live experiments. We illustrate its deployment for the evaluation of ``preference elicitation'' algorithms used to onboard new users of the YouTube Music platform. By developing counterfactually robust user behavior models, and a simulation service that couples such models with production infrastructure, we are able to test new algorithms in a way that reliably predicts their performance on key metrics when deployed live. We describe our domain, our simulation models and platform, results of experiments and deployment, and suggest future steps needed to further realistic simulation as a powerful complement to live experiments."
                    },
                    {
                        "title": "Reinforcement Learning with History-Dependent Dynamic Contexts",
                        "abstract": "We introduce Dynamic Contextual Markov Decision Processes (DCMDPs), a novel reinforcement learning framework for history-dependent environments that generalizes the contextual MDP framework to handle non-Markov environments, where contexts change over time. We consider special cases of the model, with a focus on logistic DCMDPs, which break the exponential dependence on history length by leveraging aggregation functions to determine context transitions. This special structure allows us to derive an upper-confidence-bound style algorithm for which we establish regret bounds. Motivated by our theoretical results, we introduce a practical model-based algorithm for logistic DCMDPs that plans in a latent space and uses optimism over history-dependent features. We demonstrate the efficacy of our approach on a recommendation task (using MovieLens data) where user behavior dynamics evolve in response to recommendations."
                    },
                    {
                        "title": "Investigating Action-Space Generalization in Reinforcement Learning for Recommendation Systems",
                        "abstract": "Recommender systems are used to suggest items to users based on the users\u2019 preferences. Such systems often deal with massive item sets and incredibly sparse user-item interactions, which makes it very challenging to generate high-quality personalized recommendations. Reinforcement learning (RL) is a framework for sequential decision making and naturally formulates recommender-system tasks: recommending items as actions in different user and context states to maximize long-term user experience. We investigate two RL policy parameterizations that generalize sparse user-items interactions by leveraging the relationships between actions: parameterizing the policy over action features as a softmax or Gaussian distribution. Our experiments on synthetic problems suggest that the Gaussian parameterization\u2014which is not commonly used on recommendation tasks\u2014is more robust to the set of action features than the softmax parameterization. Based on these promising results, we propose a more thorough investigation of the theoretical properties and empirical benefits of the Gaussian parameterization for recommender systems."
                    },
                    {
                        "title": "Modeling Recommender Ecosystems: Research Challenges at the Intersection of Mechanism Design, Reinforcement Learning and Generative Models",
                        "abstract": "Modern recommender systems lie at the heart of complex ecosystems that couple the behavior of users, content providers, advertisers, and other actors. Despite this, the focus of the majority of recommender research -- and most practical recommenders of any import -- is on the local, myopic optimization of the recommendations made to individual users. This comes at a significant cost to the long-term utility that recommenders could generate for its users. We argue that explicitly modeling the incentives and behaviors of all actors in the system -- and the interactions among them induced by the recommender's policy -- is strictly necessary if one is to maximize the value the system brings to these actors and improve overall ecosystem\"health\". Doing so requires: optimization over long horizons using techniques such as reinforcement learning; making inevitable tradeoffs in the utility that can be generated for different actors using the methods of social choice; reducing information asymmetry, while accounting for incentives and strategic behavior, using the tools of mechanism design; better modeling of both user and item-provider behaviors by incorporating notions from behavioral economics and psychology; and exploiting recent advances in generative and foundation models to make these mechanisms interpretable and actionable. We propose a conceptual framework that encompasses these elements, and articulate a number of research challenges that emerge at the intersection of these different disciplines."
                    },
                    {
                        "title": "Ranking with Popularity Bias: User Welfare under Self-Amplification Dynamics",
                        "abstract": "While popularity bias is recognized to play a crucial role in recommmender (and other ranking-based) systems, detailed analysis of its impact on collective user welfare has largely been lacking. We propose and theoretically analyze a general mechanism, rooted in many of the models proposed in the literature, by which item popularity, item quality, and position bias jointly impact user choice. We focus on a standard setting in which user utility is largely driven by item quality, and a recommender attempts to estimate it given user behavior. Formulating the problem as a non-stationary contextual bandit, we study the ability of a recommender policy to maximize user welfare under this model. We highlight the importance of exploration, not to eliminate popularity bias, but to mitigate its negative impact on welfare. We first show that naive popularity-biased recommenders induce linear regret by conflating item quality and popularity. More generally, we show that, even in linear settings, identifiability of item quality may not be possible due to the confounding effects of popularity bias. However, under sufficient variability assumptions, we develop an efficient optimistic algorithm and prove efficient regret guarantees w.r.t. user welfare. We complement our analysis with several simulation studies, which demonstrate the negative impact of popularity bias on the performance of several natural recommender policies."
                    },
                    {
                        "title": "Content Prompting: Modeling Content Provider Dynamics to Improve User Welfare in Recommender Ecosystems",
                        "abstract": "Users derive value from a recommender system (RS) only to the extent that it is able to surface content (or items) that meet their needs/preferences. While RSs often have a comprehensive view of user preferences across the entire user base, content providers, by contrast, generally have only a local view of the preferences of users that have interacted with their content. This limits a provider's ability to offer new content to best serve the broader population. In this work, we tackle this information asymmetry with content prompting policies. A content prompt is a hint or suggestion to a provider to make available novel content for which the RS predicts unmet user demand. A prompting policy is a sequence of such prompts that is responsive to the dynamics of a provider's beliefs, skills and incentives. We aim to determine a joint prompting policy that induces a set of providers to make content available that optimizes user social welfare in equilibrium, while respecting the incentives of the providers themselves. Our contributions include: (i) an abstract model of the RS ecosystem, including content provider behaviors, that supports such prompting; (ii) the design and theoretical analysis of sequential prompting policies for individual providers; (iii) a mixed integer programming formulation for optimal joint prompting using path planning in content space; and (iv) simple, proof-of-concept experiments illustrating how such policies improve ecosystem health and user welfare."
                    },
                    {
                        "title": "An adversarial variational inference approach for travel demand calibration of urban traffic simulators",
                        "abstract": "This paper considers the calibration of travel demand inputs, defined as a set of origin-destination matrices (ODs), for stochastic microscopic urban traffic simulators. The goal of calibration is to find a (set of) travel demand input(s) that replicate sparse field count data statistics. While traditional approaches use only first-order moment information from the field data, it is well known that the OD calibration problem is underdetermined in realistic networks. We study the value of using higher-order statistics from spatially sparse field data to mitigate underdetermination, proposing a variational inference technique that identifies an OD distribution. We apply our approach to a high-dimensional setting in Salt Lake City, Utah. Our approach is flexible---it can be readily extended to account for arbitrary types of field data (e.g., road, path or trip data)."
                    },
                    {
                        "title": "pyRDDLGym: From RDDL to Gym Environments",
                        "abstract": "We present pyRDDLGym, a Python framework for auto-generation of OpenAI Gym environments from RDDL declerative description. The discrete time step evolution of variables in RDDL is described by conditional probability functions, which fits naturally into the Gym step scheme. Furthermore, since RDDL is a lifted description, the modification and scaling up of environments to support multiple entities and different configurations becomes trivial rather than a tedious process prone to errors. We hope that pyRDDLGym will serve as a new wind in the reinforcement learning community by enabling easy and rapid development of benchmarks due to the unique expressive power of RDDL. By providing explicit access to the model in the RDDL description, pyRDDLGym can also facilitate research on hybrid approaches for learning from interaction while leveraging model knowledge. We present the design and built-in examples of pyRDDLGym, and the additions made to the RDDL language that were incorporated into the framework."
                    },
                    {
                        "title": "Color Refinement and Its Applications",
                        "abstract": "Color refinement is a simple algorithm that partitions the vertices of a graph according their \u201citerated degree sequence.\u201d It has very efficient implementations, running in quasilinear time, and a surprisingly wide range of applications. The algorithm has been designed in the context of graph isomorphism testing, and it is used as a subroutine in almost all practical graph isomorphism tools. Somewhat surprisingly, other applications, among them lifted inference, have been discovered recently. In this chapter, we introduce the basic color refinement algorithm and explain how it can be implemented in quasilinear time. We discuss variations of this basic algorithm and its somewhat surprising characterizations in terms of logic and linear programming. Then we survey the main applications of color refinement to linear programming and to graph kernels."
                    },
                    {
                        "title": "Towards Content Provider Aware Recommender Systems: A Simulation Study on the Interplay between User and Provider Utilities",
                        "abstract": "Most existing recommender systems focus primarily on matching users (content consumers) to content which maximizes user satisfaction on the platform. It is increasingly obvious, however, that content providers have a critical influence on user satisfaction through content creation, largely determining the content pool available for recommendation. A natural question thus arises: can we design recommenders taking into account the long-term utility of both users and content providers? By doing so, we hope to sustain more content providers and a more diverse content pool for long-term user satisfaction. Understanding the full impact of recommendations on both user and content provider groups is challenging. This paper aims to serve as a research investigation of one approach toward building a content provider aware recommender, and evaluating its impact in a simulated setup. To characterize the user-recommender-provider interdependence, we complement user modeling by formalizing provider dynamics as well. The resulting joint dynamical system gives rise to a weakly-coupled partially observable Markov decision process driven by recommender actions and user feedback to providers. We then build a REINFORCE recommender agent, coined EcoAgent, to optimize a joint objective of user utility and the counterfactual utility lift of the content provider associated with the recommended content, which we show to be equivalent to maximizing overall user utility and the utilities of all content providers on the platform under some mild assumptions. To evaluate our approach, we introduce a simulation environment capturing the key interactions among users, providers, and the recommender. We offer a number of simulated experiments that shed light on both the benefits and the limitations of our approach. These results help understand how and when a content provider aware recommender agent is of benefit in building multi-stakeholder recommender systems."
                    },
                    {
                        "title": "RecSim NG: Toward Principled Uncertainty Modeling for Recommender Ecosystems",
                        "abstract": "The development of recommender systems that optimize multi-turn interaction with users, and model the interactions of different agents (e.g., users, content providers, vendors) in the recommender ecosystem have drawn increasing attention in recent years. Developing and training models and algorithms for such recommenders can be especially difficult using static datasets, which often fail to offer the types of counterfactual predictions needed to evaluate policies over extended horizons. To address this, we develop RecSim NG, a probabilistic platform for the simulation of multi-agent recommender systems. RecSim NG is a scalable, modular, differentiable simulator implemented in Edward2 and TensorFlow. It offers: a powerful, general probabilistic programming language for agent-behavior specification; tools for probabilistic inference and latent-variable model learning, backed by automatic differentiation and tracing; and a TensorFlow-based runtime for running simulations on accelerated hardware. We describe RecSim NG and illustrate how it can be used to create transparent, configurable, end-to-end models of a recommender ecosystem, complemented by a small set of simple use cases that demonstrate how RecSim NG can help both researchers and practitioners easily develop and train novel algorithms for recommender systems."
                    },
                    {
                        "title": "Meta-Thompson Sampling",
                        "abstract": "Efficient exploration in bandits is a fundamental online learning problem. We propose a variant of Thompson sampling that learns to explore better as it interacts with bandit instances drawn from an unknown prior. The algorithm meta-learns the prior and thus we call it MetaTS. We propose several efficient implementations of MetaTS and analyze it in Gaussian bandits. Our analysis shows the benefit of meta-learning and is of a broader interest, because we derive a novel prior-dependent Bayes regret bound for Thompson sampling. Our theory is complemented by empirical evaluation, which shows that MetaTS quickly adapts to the unknown prior."
                    },
                    {
                        "title": "Differentiable Meta-Learning of Bandit Policies",
                        "abstract": "Exploration policies in Bayesian bandits maximize the average reward over problem instances drawn from some distribution P . In this work, we learn such policies for an unknown distribution P using samples from P . Our approach is a form of meta-learning and exploits properties of P without making strong assumptions about its form. To do this, we parameterize our policies in a differentiable way and optimize them by policy gradients, an approach that is pleasantly general and easy to implement. We derive effective gradient estimators and propose novel variance reduction techniques. We also analyze and experiment with various bandit policy classes, including neural networks and a novel softmax policy. The latter has regret guarantees and is a natural starting point for our optimization. Our experiments show the versatility of our approach. We also observe that neural network policies can learn implicit biases expressed only through the sampled instances."
                    },
                    {
                        "title": "Demonstrating Principled Uncertainty Modeling for Recommender Ecosystems with RecSim NG",
                        "abstract": "We develop RecSim NG, a probabilistic platform that supports natural, concise specification and learning of models for multi-agent recommender systems simulation. RecSim NG is a scalable, modular, differentiable simulator implemented in Edward2 and TensorFlow."
                    },
                    {
                        "title": "Differentiable Meta-Learning in Contextual Bandits",
                        "abstract": "We study a contextual bandit setting where the learning agent has access to sampled bandit instances from an unknown prior distribution $\\mathcal{P}$. The goal of the agent is to achieve high reward on average over the instances drawn from $\\mathcal{P}$. This setting is of a particular importance because it formalizes the offline optimization of bandit policies, to perform well on average over anticipated bandit instances. The main idea in our work is to optimize differentiable bandit policies by policy gradients. We derive reward gradients that reflect the structure of our problem, and propose contextual policies that are parameterized in a differentiable way and have low regret. Our algorithmic and theoretical contributions are supported by extensive experiments that show the importance of baseline subtraction, learned biases, and the practicality of our approach on a range of classification tasks."
                    },
                    {
                        "title": "Optimizing Long-term Social Welfare in Recommender Systems: A Constrained Matching Approach",
                        "abstract": "Most recommender systems (RS) research assumes that a user's utility can be maximized independently of the utility of the other agents (e.g., other users, content providers). In realistic settings, this is often not true---the dynamics of an RS ecosystem couple the long-term utility of all agents. In this work, we explore settings in which content providers cannot remain viable unless they receive a certain level of user engagement. We formulate the recommendation problem in this setting as one of equilibrium selection in the induced dynamical system, and show that it can be solved as an optimal constrained matching problem. Our model ensures the system reaches an equilibrium with maximal social welfare supported by a sufficiently diverse set of viable providers. We demonstrate that even in a simple, stylized dynamical RS model, the standard myopic approach to recommendation---always matching a user to the best provider---performs poorly. We develop several scalable techniques to solve the matching problem, and also draw connections to various notions of user regret and fairness, arguing that these outcomes are fairer in a utilitarian sense."
                    },
                    {
                        "title": "Differentiable Bandit Exploration",
                        "abstract": "Exploration policies in Bayesian bandits maximize the average reward over problem instances drawn from some distribution $\\mathcal{P}$. In this work, we learn such policies for an unknown distribution $\\mathcal{P}$ using samples from $\\mathcal{P}$. Our approach is a form of meta-learning and exploits properties of $\\mathcal{P}$ without making strong assumptions about its form. To do this, we parameterize our policies in a differentiable way and optimize them by policy gradients, an approach that is general and easy to implement. We derive effective gradient estimators and introduce novel variance reduction techniques. We also analyze and experiment with various bandit policy classes, including neural networks and a novel softmax policy. The latter has regret guarantees and is a natural starting point for our optimization. Our experiments show the versatility of our approach. We also observe that neural network policies can learn implicit biases expressed only through the sampled instances."
                    }
                ]
            },
            "390fce14-bd7c-4e35-af1f-4c756262c7bc": {
                "pk": "390fce14-bd7c-4e35-af1f-4c756262c7bc",
                "name": "Craig Boutilier",
                "collaborators": [
                    "Chih-Wei Hsu",
                    "Yinlam Chow",
                    "Guy Tennenholtz",
                    "Erdem Biyik",
                    "M. Ghavamzadeh",
                    "Tyler Lu",
                    "Georg Gottlob",
                    "Martin Mladenov",
                    "Anthony Liang",
                    "Lior Shani",
                    "Ethan Liang",
                    "Fan Yao",
                    "Alex Haig",
                    "Jihwan Jeong",
                    "Azamat Tulepbergenov",
                    "Christina G\u00f6pfert",
                    "Ivan Vendrov",
                    "Deepak Ramachandran",
                    "Thomas Eiter",
                    "M. Hauskrecht",
                    "Nicolas Meuleau",
                    "Leslie Pack",
                    "Thomas Dean",
                    "Britta Dorn",
                    "Nicolas Maudet",
                    "Vincent Merlin",
                    "Ariel D. Procaccia",
                    "Bob Price",
                    "M. Goldszmidt",
                    "Siddharth Prasad"
                ],
                "domain": [
                    "Recommender Systems",
                    "Mechanism Design",
                    "Reinforcement Learning",
                    "Social Choice"
                ],
                "publications": [
                    {
                        "title": "Recommender Ecosystems: A Mechanism Design Perspective on Holistic Modeling and Optimization",
                        "abstract": "Modern recommender systems lie at the heart of complex recommender ecosystems that couple the behavior of users, content providers, vendors, advertisers, and other actors. Despite this, the focus of much recommender systems research and deployment is on the local, myopic optimization of the recommendations made to individual users. This comes at a significant cost to the long-term utility that recommender systems generate for their users. We argue that modeling the incentives and behaviors of these actors, and the interactions among them induced by the recommender systems, is needed to maximize value and improve overall ecosystem health. Moreover, we propose the use of economic mechanism design, an area largely overlooked in recommender systems research, as a framework for developing such models. That said, one cannot apply \u201cvanilla\u201d mechanism design to recommender ecosystem modeling optimization out of the box\u2014the use of mechanism design raises a number of subtle and interesting research challenges. We outline a number of these in this talk (and paper), emphasizing the need to develop nonstandard approaches to mechanism design that intersect with numerous areas of research, including preference modeling, reinforcement learning and exploration, behavioral economics, and generative AI, among others."
                    },
                    {
                        "title": "DynaMITE-RL: A Dynamic Model for Improved Temporal Meta-Reinforcement Learning",
                        "abstract": "We introduce DynaMITE-RL, a meta-reinforcement learning (meta-RL) approach to approximate inference in environments where the latent state evolves at varying rates. We model episode sessions - parts of the episode where the latent state is fixed - and propose three key modifications to existing meta-RL methods: consistency of latent information within sessions, session masking, and prior latent conditioning. We demonstrate the importance of these modifications in various domains, ranging from discrete Gridworld environments to continuous-control and simulated robot assistive tasks, demonstrating that DynaMITE-RL significantly outperforms state-of-the-art baselines in sample efficiency and inference returns."
                    },
                    {
                        "title": "Embedding-Aligned Language Models",
                        "abstract": "We propose a novel approach for training large language models (LLMs) to adhere to objectives defined within a latent embedding space. Our method leverages reinforcement learning (RL), treating a pre-trained LLM as an environment. Our embedding-aligned guided language (EAGLE) agent is trained to iteratively steer the LLM's generation towards optimal regions of the latent embedding space, w.r.t. some predefined criterion. We demonstrate the effectiveness of the EAGLE agent using the MovieLens 25M and Amazon Review datasets to surface content gaps that satisfy latent user demand. We also demonstrate the benefit of using an optimal design of a state-dependent action set to improve EAGLE's efficiency. Our work paves the way for controlled and grounded text generation using LLMs, ensuring consistency with domain-specific knowledge and data representations."
                    },
                    {
                        "title": "Preference Elicitation with Soft Attributes in Interactive Recommendation",
                        "abstract": "Preference elicitation plays a central role in interactive recommender systems. Most preference elicitation approaches use either item queries that ask users to select preferred items from a slate, or attribute queries that ask them to express their preferences for item characteristics. Unfortunately, users often wish to describe their preferences using soft attributes for which no ground-truth semantics is given. Leveraging concept activation vectors for soft attribute semantics, we develop novel preference elicitation methods that can accommodate soft attributes and bring together both item and attribute-based preference elicitation. Our techniques query users using both items and soft attributes to update the recommender system's belief about their preferences to improve recommendation quality. We demonstrate the effectiveness of our methods vis-a-vis competing approaches on both synthetic and real-world datasets."
                    },
                    {
                        "title": "Factual and Personalized Recommendations using Language Models and Reinforcement Learning",
                        "abstract": "Recommender systems (RSs) play a central role in connecting users to content, products, and services, matching candidate items to users based on their preferences. While traditional RSs rely on implicit user feedback signals, conversational RSs interact with users in natural language. In this work, we develop a comPelling, Precise, Personalized, Preference-relevant language model (P4LM) that recommends items to users while putting emphasis on explaining item characteristics and their relevance. P4LM uses the embedding space representation of a user's preferences to generate compelling responses that are factually-grounded and relevant w.r.t. the user's preferences. Moreover, we develop a joint reward function that measures precision, appeal, and personalization, which we use as AI-based feedback in a reinforcement learning-based language model framework. Using the MovieLens 25M dataset, we demonstrate that P4LM delivers compelling, personalized movie narratives to users."
                    },
                    {
                        "title": "Discovering Personalized Semantics for Soft Attributes in Recommender Systems using Concept Activation Vectors",
                        "abstract": "Interactive recommender systems (RSs) allow users to express intent, preferences and contexts in a rich fashion, often using natural language. One challenge in using such feedback is inferring a user\u2019s semantic intent from the open-ended terms used to describe an item, and using it to refine recommendation results. Leveraging concept activation vectors (CAVs) [21], we develop a framework to learn a representation that captures the semantics of such attributes and connects them to user preferences and behaviors in RSs. A novel feature of our approach is its ability to distinguish objective and subjective attributes and associate different senses with different users. Using synthetic and real-world datasets, we show that our CAV representation accurately interprets users\u2019 subjective semantics, and can improve recommendations via interactive critiquing."
                    },
                    {
                        "title": "EPISTEMIC ENTRENCHMENT IN AUTOEPISTEMIC LOGIC REASONING WITH PARSIMONIOUS AND MODERATELY GROUNDED EXPANSIONS",
                        "abstract": ". A drawback of existing epistemic logics is their inability to deal with entrenchment of beliefs. All beliefs have equal status; none can be held more \ufb01rmly than others. We present a bimodal logic that generalizes Levesque\u2019s reconstruction of autoepistemic logic. In our system standard epistemic concepts can be represented, including the notion of only knowing, but ele-ments of a belief set may be more or less entrenched. This has important implications for the distinction between epistemic defaults and subjunctive (and normative) defaults, both of which are representable in our system. Abstract. We investigate the complexity of autoepistemic reasoning with parsimonious and moderately grounded expansions. A stable expansion of an autoepistemic set of premises is parsimonious if its objective (i.e. nonmodal) part does not contain the objective part of any other stable expansion. We prove that deciding whether a formula \u03d5 belongs to at least one parsimonious stable expansion of a \ufb01nite base set A is complete for \u03a3 P 3 , while deciding containment in all parsimonious stable expansions is complete for \u03a3 P 3 . Similar results are derived for autoepistemic reasoning with moderately grounded expansions. In particular, we show that deciding whether a formula \u03d5 belongs to some moderately grounded expansion of a. \ufb01nite base set A is \u03a3 P 3 -complete, and that deciding whether \u03d5 belongs to all moderately grounded expansions is \u03a3 P 3 -complete. These results suggest that reasoning with parsimonious stable expansions and moderately, grounded expansions is strictly harder than reasoning in Moore\u2019s standard version of autoepistemic logic. We also ad-dress the complexity of reasoning if the set A is in a normalized form, and derive completeness results for this case. Abstract. Suppose there are several experts, with some dominating others (expert A dominates expert B if B says something is true whenever A says it is). Suppose, further, that each of the experts has his or her own view of what is possible - in other words each of the experts has their own Kripke model in mind (subject, of course, to the dominance relation that may hold between experts). How will they assign truth values to sentences in a common, modal language, and on what sentences will they agree? This problem can be reformulated as one about many-valued Kripke models, allowing many-valued accessibility relations. This is a natural generalization of conventional Kripke models that has only recently been looked at. The equivalence between the many-valued version and the multiple expert one will be formally established. Finally we will axiomatize many-valued modal logics, and sketch a proof of completeness. Abstract. Agents with perfect introspection may have incomplete beliefs about the world, but they possess complete knowledge about their own beliefs. This fact suggests that the beliefs of introspective agents should be completely determined by their objective beliefs, that is, those beliefs that are only about the domain in question and not about other beliefs. Introspection and logical reasoning alone should suf\ufb01ce to re-construct all other beliefs from the objective ones. While this property has been shown to hold for propositional belief logics, there have so far only been negative results in the case of \ufb01rst-order belief logics with quantifying-in. In this paper we present a logic of belief with quantifying-in, where the beliefs of a perfectly introspective agent are indeed uniquely determined by the objective beliefs. The result is obtained by weakening the notion of belief of an existing logic that does not have this property. Abstract. Investigations of default logic have been so far mostly concerned with the notion of an extension of a default\u2019 theory. It turns out, however, that default logic is much richer. Namely, there are other natural classes of objects that might be associated with default reasoning. We study two such classes of objects with emphasis on their relations with modal nonmonotonic formalisms. First, we introduce the concept of a weak extension and Study its properties. It has long been suspected that there are close connections between default and autoepistemic logics. The notion of weak extension allows us to precisely describe the relationship between these two formalisms. In particular, we show that default logic with weak extensions is essentially equivalent to autoepistemic logic, that is, nonmonotonic logic KD45 . In the paper we also study the notion of a set of formulas closed under a default theory. These objects are shown to correspond to stable theories and to modal logic S5 . In particular, we show that skeptical reasoning with sets closed under default theories is closely related With provability in S5 . Abstract. The decidability and computational complexity of autoepistemic reasoning is investi-gated in a general setting where the autoepistemic logic CL ae built on top of a given classical logic CL is studied. Correct autoepistemic conclusions from a set of premises are de\ufb01ned in terms of expansions of the premises. Three classes of expansions are studied: Moore style stable expansions, enumeration based expansions, and L-hierarchic expansions. A simple \ufb01nitary characterization for each type of expansions in CL ae is developed. Using the characterizations conditions ensuring that a set of premises has at least one or exactly one stable expansion can be stated and an upper bound Abstract. We propose a new variant of autoepistemic logic which, intuitively, corresponds to un-derstanding a belief operator L as \"is known\", in contrast to the interpretation of L as \"is believed\" in Moore\u2019s autoepistemic logic. Formal properties of the new logic and relationship to Moore\u2019s logic are studied in detail. Abstract. Stable autoepistemic sets are treated as S4 maximal theories. This approach allows to use the Lindenbaum algebra of modal logic as a tool in investigations of existence of a special kind of stable sets. Namely, for a given set A , the algebraic constructions of minimal stable sets based on A and moderately grounded expansions of A are shown. Abstract. This paper discusses a number of properties that a partial order semantics should have in order to support the modular construction of nets and to deal with \ufb01nite capacities. Characterizations for these properties are shown, and a new semantics is introduced which seems to be the natural choice if a certain set of properties is required. Abstract. In this paper, we investigate several equivalence relations for probabilistic labeled transition systems: bisimulation equivalence, readiness equivalence, failure equivalence, trace equivalence, maximal trace equivalence and \ufb01nite trace equivalence. We formally prove the inclusions (equalities) among these equivalences. We also show that readiness, failure, trace, maximum trace and \ufb01nite trace equivalences for \ufb01nite probabilistic labeled transition systems are decidable in polynomial time. This should be contrasted with the PSPACE completeness of the same equivalences for classical labeled transition systems. Moreover, we derive an ef\ufb01cient polynomial time algorithm for deciding bisimulation equivalence for \ufb01nite probabilistic labeled transition systems. The special case of initiated probabilistic transition systems will be considered. We show that the isomorphism problem for \ufb01nite initiated labeled probabilistic transition systems is NC (1) equivalent to graph isomorphism. Abstract. In this paper we de\ufb01ne two new inference classes Team Density and Team Uniform Density . The team density class turns out to be the \"most powerful\" of all the inference classes that have been de\ufb01ned and studied so far. We compare these classes to previously de\ufb01ned classes. We obtain necessary and suf\ufb01cient conditions for one team density class to be a subset of another team density class. Similar results are obtained for team uniform density class. We also have compare team density and team uniform density classes and obtained necessary and suf\ufb01cient conditions for one class to be a subset of another. Most importantly we show that team density, the \"most powerfull\" class, cannot cover the BC class. From this result we obtain several theorems as corollaries that have been proved in earlier papers. Abstract. Six \u03bb -languages over function-types between algebras B, N and \u03a5 are considered. Type N = (0 \u2192 0) \u2192 (0 \u2192 0) is called a non-negative integers type; B = (0 \u2192 0) \u2192 ((0 \u2192 0) \u2192 (0 \u2192 0)) is called a binary words type; \u03a5 = (0 \u2192 (0 \u2192 0)) \u2192 (0 \u2192 0) is called a binary trees type. These associations come from the isomorphism between the types and corresponding algebraic structures. Closed terms whose types are the above mentioned function-types represent unary functions of appropriate types. The problem is: what class of functions is represented by the closed terms of the examined type. It is proved that for B \u2192 N, N \u2192 B, \u03a5 \u2192 N, \u03a5 \u2192 B there exists a \ufb01nite base of functions such that any \u03bb -de\ufb01nable function is sonic combination of the base functions. The algorithm which, for every closed term, returns the function in the form of a combination of the base functions is given. For two other types, B \u2192 \u03a5 and N \u2192 \u03a5 , a method of constructing \u03bb -representable functions using primitive recursion is shown. Abstract. We present an automata-theoretic decision procedure for Since/Until Temporal Logic (SUTL), a linear-time propositional temporal logic with strong non-strict since and until operators. The logic, which is intended for specifying and reasoning about computer systems, employs neither next nor previous operators. Such operators obstruct the use of hierarchical abstraction and re\ufb01nement and make reasoning about concurrency dif\ufb01cult. A proof of the soundness and completeness of the decision procedure is given, and its complexity is analyzed. Abstract. Starting from a general dynamic analysis of reasoning and programming, we develop two main dynamic perspe"
                    },
                    {
                        "title": "EPISTEMIC ENTRENCHMENT IN AUTOEPISTEMIC LOGIC REASONING WITH PARSIMONIOUS AND MODERATELY GROUNDED EXPANSIONS",
                        "abstract": ". A drawback of existing epistemic logics is their inability to deal with entrenchment of beliefs. All beliefs have equal status; none can be held more \ufb01rmly than others. We present a bimodal logic that generalizes Levesque\u2019s reconstruction of autoepistemic logic. In our system standard epistemic concepts can be represented, including the notion of only knowing, but ele-ments of a belief set may be more or less entrenched. This has important implications for the distinction between epistemic defaults and subjunctive (and normative) defaults, both of which are representable in our system. Abstract. We investigate the complexity of autoepistemic reasoning with parsimonious and moderately grounded expansions. A stable expansion of an autoepistemic set of premises is parsimonious if its objective (i.e. nonmodal) part does not contain the objective part of any other stable expansion. We prove that deciding whether a formula \u03d5 belongs to at least one parsimonious stable expansion of a \ufb01nite base set A is complete for \u03a3 P 3 , while deciding containment in all parsimonious stable expansions is complete for \u03a3 P 3 . Similar results are derived for autoepistemic reasoning with moderately grounded expansions. In particular, we show that deciding whether a formula \u03d5 belongs to some moderately grounded expansion of a. \ufb01nite base set A is \u03a3 P 3 -complete, and that deciding whether \u03d5 belongs to all moderately grounded expansions is \u03a3 P 3 -complete. These results suggest that reasoning with parsimonious stable expansions and moderately, grounded expansions is strictly harder than reasoning in Moore\u2019s standard version of autoepistemic logic. We also ad-dress the complexity of reasoning if the set A is in a normalized form, and derive completeness results for this case. Abstract. Suppose there are several experts, with some dominating others (expert A dominates expert B if B says something is true whenever A says it is). Suppose, further, that each of the experts has his or her own view of what is possible - in other words each of the experts has their own Kripke model in mind (subject, of course, to the dominance relation that may hold between experts). How will they assign truth values to sentences in a common, modal language, and on what sentences will they agree? This problem can be reformulated as one about many-valued Kripke models, allowing many-valued accessibility relations. This is a natural generalization of conventional Kripke models that has only recently been looked at. The equivalence between the many-valued version and the multiple expert one will be formally established. Finally we will axiomatize many-valued modal logics, and sketch a proof of completeness. Abstract. Agents with perfect introspection may have incomplete beliefs about the world, but they possess complete knowledge about their own beliefs. This fact suggests that the beliefs of introspective agents should be completely determined by their objective beliefs, that is, those beliefs that are only about the domain in question and not about other beliefs. Introspection and logical reasoning alone should suf\ufb01ce to re-construct all other beliefs from the objective ones. While this property has been shown to hold for propositional belief logics, there have so far only been negative results in the case of \ufb01rst-order belief logics with quantifying-in. In this paper we present a logic of belief with quantifying-in, where the beliefs of a perfectly introspective agent are indeed uniquely determined by the objective beliefs. The result is obtained by weakening the notion of belief of an existing logic that does not have this property. Abstract. Investigations of default logic have been so far mostly concerned with the notion of an extension of a default\u2019 theory. It turns out, however, that default logic is much richer. Namely, there are other natural classes of objects that might be associated with default reasoning. We study two such classes of objects with emphasis on their relations with modal nonmonotonic formalisms. First, we introduce the concept of a weak extension and Study its properties. It has long been suspected that there are close connections between default and autoepistemic logics. The notion of weak extension allows us to precisely describe the relationship between these two formalisms. In particular, we show that default logic with weak extensions is essentially equivalent to autoepistemic logic, that is, nonmonotonic logic KD45 . In the paper we also study the notion of a set of formulas closed under a default theory. These objects are shown to correspond to stable theories and to modal logic S5 . In particular, we show that skeptical reasoning with sets closed under default theories is closely related With provability in S5 . Abstract. The decidability and computational complexity of autoepistemic reasoning is investi-gated in a general setting where the autoepistemic logic CL ae built on top of a given classical logic CL is studied. Correct autoepistemic conclusions from a set of premises are de\ufb01ned in terms of expansions of the premises. Three classes of expansions are studied: Moore style stable expansions, enumeration based expansions, and L-hierarchic expansions. A simple \ufb01nitary characterization for each type of expansions in CL ae is developed. Using the characterizations conditions ensuring that a set of premises has at least one or exactly one stable expansion can be stated and an upper bound Abstract. We propose a new variant of autoepistemic logic which, intuitively, corresponds to un-derstanding a belief operator L as \"is known\", in contrast to the interpretation of L as \"is believed\" in Moore\u2019s autoepistemic logic. Formal properties of the new logic and relationship to Moore\u2019s logic are studied in detail. Abstract. Stable autoepistemic sets are treated as S4 maximal theories. This approach allows to use the Lindenbaum algebra of modal logic as a tool in investigations of existence of a special kind of stable sets. Namely, for a given set A , the algebraic constructions of minimal stable sets based on A and moderately grounded expansions of A are shown. Abstract. This paper discusses a number of properties that a partial order semantics should have in order to support the modular construction of nets and to deal with \ufb01nite capacities. Characterizations for these properties are shown, and a new semantics is introduced which seems to be the natural choice if a certain set of properties is required. Abstract. In this paper, we investigate several equivalence relations for probabilistic labeled transition systems: bisimulation equivalence, readiness equivalence, failure equivalence, trace equivalence, maximal trace equivalence and \ufb01nite trace equivalence. We formally prove the inclusions (equalities) among these equivalences. We also show that readiness, failure, trace, maximum trace and \ufb01nite trace equivalences for \ufb01nite probabilistic labeled transition systems are decidable in polynomial time. This should be contrasted with the PSPACE completeness of the same equivalences for classical labeled transition systems. Moreover, we derive an ef\ufb01cient polynomial time algorithm for deciding bisimulation equivalence for \ufb01nite probabilistic labeled transition systems. The special case of initiated probabilistic transition systems will be considered. We show that the isomorphism problem for \ufb01nite initiated labeled probabilistic transition systems is NC (1) equivalent to graph isomorphism. Abstract. In this paper we de\ufb01ne two new inference classes Team Density and Team Uniform Density . The team density class turns out to be the \"most powerful\" of all the inference classes that have been de\ufb01ned and studied so far. We compare these classes to previously de\ufb01ned classes. We obtain necessary and suf\ufb01cient conditions for one team density class to be a subset of another team density class. Similar results are obtained for team uniform density class. We also have compare team density and team uniform density classes and obtained necessary and suf\ufb01cient conditions for one class to be a subset of another. Most importantly we show that team density, the \"most powerfull\" class, cannot cover the BC class. From this result we obtain several theorems as corollaries that have been proved in earlier papers. Abstract. Six \u03bb -languages over function-types between algebras B, N and \u03a5 are considered. Type N = (0 \u2192 0) \u2192 (0 \u2192 0) is called a non-negative integers type; B = (0 \u2192 0) \u2192 ((0 \u2192 0) \u2192 (0 \u2192 0)) is called a binary words type; \u03a5 = (0 \u2192 (0 \u2192 0)) \u2192 (0 \u2192 0) is called a binary trees type. These associations come from the isomorphism between the types and corresponding algebraic structures. Closed terms whose types are the above mentioned function-types represent unary functions of appropriate types. The problem is: what class of functions is represented by the closed terms of the examined type. It is proved that for B \u2192 N, N \u2192 B, \u03a5 \u2192 N, \u03a5 \u2192 B there exists a \ufb01nite base of functions such that any \u03bb -de\ufb01nable function is sonic combination of the base functions. The algorithm which, for every closed term, returns the function in the form of a combination of the base functions is given. For two other types, B \u2192 \u03a5 and N \u2192 \u03a5 , a method of constructing \u03bb -representable functions using primitive recursion is shown. Abstract. We present an automata-theoretic decision procedure for Since/Until Temporal Logic (SUTL), a linear-time propositional temporal logic with strong non-strict since and until operators. The logic, which is intended for specifying and reasoning about computer systems, employs neither next nor previous operators. Such operators obstruct the use of hierarchical abstraction and re\ufb01nement and make reasoning about concurrency dif\ufb01cult. A proof of the soundness and completeness of the decision procedure is given, and its complexity is analyzed. Abstract. Starting from a general dynamic analysis of reasoning and programming, we develop two main dynamic perspe"
                    },
                    {
                        "title": "Decision(cid:0)theoretic Planning with Temporally Abstract Actions",
                        "abstract": "Markov decision"
                    },
                    {
                        "title": "Computational Social Choice: Theory and Applications (Dagstuhl Seminar 15241)",
                        "abstract": "This report documents the program and the outcomes of Dagstuhl Seminar 15241 \"Computational Social Choice: Theory and Applications\". The seminar featured a mixture of classic scientific talks (including three overview talks), open problem presentations, working group sessions, and five-minute contributions (\"rump session\"). While there were other seminars on related topics in the past, a special emphasis was put on practical applications in this edition."
                    },
                    {
                        "title": "A Dynamic Rationalization of Distance Rationalizability",
                        "abstract": "    Distance rationalizability is an intuitive paradigm for developing and studying voting rules: given a notion of consensus and a distance function on preference profiles, a rationalizable voting rule selects an alternative that is closest to being a consensus winner. Despite its appeal, distance rationalizability faces the challenge of connecting the chosen distance measure and consensus notion to an operational measure of social desirability. We tackle this issue via the decision-theoretic framework of dynamic social choice, in which a social choice Markov decision process (MDP) models the dynamics of voter preferences in response to winner selection. We show that, for a prominent class of distance functions, one can construct a social choice MDP, with natural preference dynamics and rewards, such that a voting rule is (votewise) rationalizable with respect to the unanimity consensus for a given distance function iff it is a (deterministic) optimal policy in the MDP. This provides an alternative rationale for distance rationalizability, demonstrating the equivalence of rationalizable voting rules in a static sense and winner selection to maximize societal utility in a dynamic process.   "
                    },
                    {
                        "title": "Accelerating Reinforcement Learning through Implicit Imitation",
                        "abstract": "Accelerating Reinforcement Learning through Implicit Imitation\u2014 2003 Bob Price, Craig Boutilier.* (Continued) Primary Examiner \u2014 Kakali Chaki Assistant Examiner \u2014 Ababacar Seck (74) Attorney, Agent, or Firm \u2014 Fish & Richardson P.C. (57) ABSTRACT Methods, systems, and apparatus, including computer pro grams encoded on computer storage media, for reinforce ment learning using confidence scores. One of the methods includes receiving a current observation; for each of mul tiple actions: determining a respective value function esti mate that is an estimate of a return resulting from the agent performing the action in response to the current observation, determining a respective confidence score that is a measure of confidence that the respective value function estimate for the action is an accurate estimate of the return that will result from the agent performing the action in response to the current observation, adjusting the respective value function estimate for the action using the respective confidence score for the action to determine a respective adjusted value function estimate; and selecting an action to be performed by the agent in response to the current observation using the respective adjusted value function estimates. 30 Claims, 5 Drawing Sheets"
                    },
                    {
                        "title": "Social Choice : A Framework for Multiple Recommendations in Consensus Decision Making",
                        "abstract": "We develop a new framework for social choice problems, budgeted social choice, in which a limited number of alternatives can be recommended/prescribed to a population of agents. This limit is determined by some form of budget. Such problems naturally arise in a variety of contexts. Our model is general, spanning the continuum from pure consensus decisions (i.e., standard social choice) to fully personalized recommendation. Our results show that standard rank aggregation rules are not appropriate for such tasks and that good solutions typically involve picking diverse alternatives tailored to different agent types. The corresponding optimization problems are shown to be NP-complete, but we develop fast greedy algorithms with some theoretical guarantees. Experimental results on real-world datasets (APA election and sushi) show some interesting patterns and the prove the effectiveness of our greedy algorithms."
                    },
                    {
                        "title": "Proceedings of the Sixteenth conference on Uncertainty in artificial intelligence",
                        "abstract": "This is the Proceedings of the Sixteenth Conference on Uncertainty in Artificial Intelligence, which was held in San Francisco, CA, June 30 - July 3, 2000"
                    },
                    {
                        "title": "Mechanism Design and Integer Programming in the Data Age",
                        "abstract": "Modern-day human-scale marketplaces such as recommender systems, advertisement markets, matching platforms, supply chain industries, electronic commerce platforms, and others must reckon with a balancing act of (i) understanding and respecting the incentives of the system\u2019s participants, (ii) obtaining optimal outcomes subject to those incentives, and (iii) ensuring that data is used in a sound manner to improve overall efficiency. The area of mechanism design from economics provides a rich language and broad toolkit to understand incentives and integer programming is a powerful and extremely expressive optimization language that is the workhorse behind most practical solutions to real-world discrete optimization problems. This thesis studies how data-driven decisions can be integrated into fundamental algorithms from both areas to improve performance (economic, memory, run-time, etc. ). Within mechanism design, I focus on the design of revenue optimal mechanisms from a data-driven lens. I design algorithms, model new learning paradigms, and invent new mechanism classes for a variety of settings including two-part tariffs, combinatorial auctions, shrinking markets, and general multi-dimensional mechanism design. A highlight here is my development of the first general tunable framework for integrating side information into mechanisms to boost revenue while preserving efficiency and incentive properties. Within integer programming, I develop principled new methods for cutting plane configuration, which is one of the most important components in state-of-the-art branch-and-cut integer programming solvers. I develop a comprehensive generalization theory for cutting plane configuration that (i) unveils new geometric and combinatorial structure in the branch-and-cut algorithm and the class of Gomory cuts, (ii) improves upon and sub-sumes prior work via an abstract model of the underlying tree search, and (iii) is validated through experiments that demonstrate the impact of data-dependent parameter tuning. I also derive a new method of sequence-independent lifting for cutting planes that I validate through rigorous theory\u2014 by deriving broad conditions under which the new cuts define facets of the integer polytope\u2014and extensive experiments. I conclude by discussing promising avenues for future work, most notably my preliminary ideas on core-selecting combinatorial auctions with side information which calls for a blend of innovations in both auction design and integer programming/search techniques."
                    }
                ]
            }
        }
    },
    "2410.15059": {
        "paper_data": {
            "title": "Deep Equilibrium Algorithmic Reasoning",
            "url": "http://arxiv.org/abs/2410.15059v1",
            "arxiv_id": "2410.15059",
            "authors": [
                "Dobrik Georgiev",
                "JJ Wilson",
                "Davide Buffelli",
                "Pietro Li\u00f2"
            ],
            "abstract": "Neural Algorithmic Reasoning (NAR) research has demonstrated that graph neural networks (GNNs) could learn to execute classical algorithms. However, most previous approaches have always used a recurrent architecture, where each iteration of the GNN matches an iteration of the algorithm. In this paper we study neurally solving algorithms from a different perspective: since the algorithm's solution is often an equilibrium, it is possible to find the solution directly by solving an equilibrium equation. Our approach requires no information on the ground-truth number of steps of the algorithm, both during train and test time. Furthermore, the proposed method improves the performance of GNNs on executing algorithms and is a step towards speeding up existing NAR models. Our empirical evidence, leveraging algorithms from the CLRS-30 benchmark, validates that one can train a network to solve algorithmic problems by directly finding the equilibrium. We discuss the practical implementation of such models and propose regularisations to improve the performance of these equilibrium reasoners.",
            "introduction": "   1 Introduction  Algorithms, while straightforward in theory, become challenging to deploy in real-world scenarios. They operate in abstract domains with very specific conditions and types of inputs, which are represented with scalars. The main hurdle is the need to \u201ccollapse\u201d reality into a scalar every time an algorithm is used, something usually done based on intuition rather than principled science [25]. Neural Algorithmic Reasoning (NAR; 46) has been proposed to address this issue by utilising specialised neural network architectures to break this scalar bottleneck by executing algorithms in higher-dimensional space made out of arrays of numbers (vectors). The task of mapping reality into this vectorial space is delegated to automated gradient-based optimisation techniques rather than relying on human operators.   While NAR does not provide the correctness guarantees of its classical counterparts, robust performance can be achieved through alignment [55] \u2013 submodules of the model architecture correspond to easy-to-learn subparts of the algorithm (or class of). Graph Neural Networks (GNNs) have emerged as the most convenient architecture to execute all types of algorithms [29] and GNNs that align better to the target algorithm achieve better generalisation. This alignment game [50] has led to a sequence of exciting research \u2013 from aligning the architecture with iterative algorithms [42] to proving that \u201cgraph neural networks are dynamic programmers\u201d [13], especially if their message-passing function [21] takes into account 3-node interactions.   The aforementioned papers focus on aligning the computation of the GNN with an algorithm or a specific algorithm class (e.g.\u00a0dynamic programming), but ignore the properties at the time of algorithm termination. For the algorithms in the CLRS-30 algorithmic reasoning benchmark [45] once the optimal solution is found, further algorithm iterations will not change it. For example, in dynamic programming shortest-paths algorithms making additional iterations would not alter the optimality of the shortest paths\u2019 distances found. Such a state is called an equilibrium \u2013 additional applications of a function (an algorithm\u2019s iteration) to the state leave it unchanged.   In this paper:   1.  We explore the connection between execution of algorithms and equilibrium finding through the use of Denotational semantics and Domain theory. (section\u00a03)    2.  Inspired by the above, we implement a class of deep equilibrium algorithmic reasoners (DEARs) that learn algorithms by identifying the equilibrium point of the GNN equation and propose improvements to them. (section\u00a04)    3.  Our results suggest that the above reasoners can be successfully trained. Not only does equilibrium algorithmic reasoning achieve better overall performance with less expressive (and expensive) GNNs, but is also competitive to the more expressive (and expensive) NAR models. All this is done without providing any information on the number of algorithmic steps \u2013 neither at train nor at test time. (section\u00a05)    4.  DEARs also drastically improve the inference speed \u2013 an achievement made possible by the use of optimised root-finding algorithms and by decoupling the neural model from the sequential implementation of algorithms in standard benchmark datasets. (section\u00a05)      Related work  The main proposed application of NAR is settings where one wants to apply an algorithm, but it is impossible to represent reality with a single scalar, hence an \u201cexecutor\u201d operating in vector space and faithful to the algorithm is required [47, 50]. As NAR models are neural clones of algorithms, they need to provide correct output even for previously unobserved input sizes. Achieving robust out-of-distribution (OOD) generalisation is tricky. To this end a plethora of works have dedicated their attention to it \u2013 [42, 13, 8, 14] to name a few. Except for",
            "references": [
                {
                    "title": "Cayley Graph Propagation",
                    "abstract": "In spite of the plethora of success stories with graph neural networks (GNNs) on modelling graph-structured data, they are notoriously vulnerable to over-squashing, whereby tasks necessitate the mixing of information between distance pairs of nodes. To address this problem, prior work suggests rewiring the graph structure to improve information flow. Alternatively, a significant body of research has dedicated itself to discovering and precomputing bottleneck-free graph structures to ameliorate over-squashing. One well regarded family of bottleneck-free graphs within the mathematical community are expander graphs, with prior work$\\unicode{x2014}$Expander Graph Propagation (EGP)$\\unicode{x2014}$proposing the use of a well-known expander graph family$\\unicode{x2014}$the Cayley graphs of the $\\mathrm{SL}(2,\\mathbb{Z}_n)$ special linear group$\\unicode{x2014}$as a computational template for GNNs. However, in EGP the computational graphs used are truncated to align with a given input graph. In this work, we show that truncation is detrimental to the coveted expansion properties. Instead, we propose CGP, a method to propagate information over a complete Cayley graph structure, thereby ensuring it is bottleneck-free to better alleviate over-squashing. Our empirical evidence across several real-world datasets not only shows that CGP recovers significant improvements as compared to EGP, but it is also akin to or outperforms computationally complex graph rewiring techniques."
                },
                {
                    "title": "GRANOLA: Adaptive Normalization for Graph Neural Networks",
                    "abstract": "In recent years, significant efforts have been made to refine the design of Graph Neural Network (GNN) layers, aiming to overcome diverse challenges, such as limited expressive power and oversmoothing. Despite their widespread adoption, the incorporation of off-the-shelf normalization layers like BatchNorm or InstanceNorm within a GNN architecture may not effectively capture the unique characteristics of graph-structured data, potentially reducing the expressive power of the overall architecture. Moreover, existing graph-specific normalization layers often struggle to offer substantial and consistent benefits. In this paper, we propose GRANOLA, a novel graph-adaptive normalization layer. Unlike existing normalization layers, GRANOLA normalizes node features by adapting to the specific characteristics of the graph, particularly by generating expressive representations of its neighborhood structure, obtained by leveraging the propagation of Random Node Features (RNF) in the graph. We present theoretical results that support our design choices. Our extensive empirical evaluation of various graph benchmarks underscores the superior performance of GRANOLA over existing normalization techniques. Furthermore, GRANOLA emerges as the top-performing method among all baselines within the same time complexity of Message Passing Neural Networks (MPNNs)."
                },
                {
                    "title": "TorchDEQ: A Library for Deep Equilibrium Models",
                    "abstract": "Deep Equilibrium (DEQ) Models, an emerging class of implicit models that maps inputs to fixed points of neural networks, are of growing interest in the deep learning community. However, training and applying DEQ models is currently done in an ad-hoc fashion, with various techniques spread across the literature. In this work, we systematically revisit DEQs and present TorchDEQ, an out-of-the-box PyTorch-based library that allows users to define, train, and infer using DEQs over multiple domains with minimal code and best practices. Using TorchDEQ, we build a ``DEQ Zoo'' that supports six published implicit models across different domains. By developing a joint framework that incorporates the best practices across all models, we have substantially improved the performance, training stability, and efficiency of DEQs on ten datasets across all six projects in the DEQ Zoo. TorchDEQ and DEQ Zoo are released as \\href{https://github.com/locuslab/torchdeq}{open source}."
                },
                {
                    "title": "Locality-Aware Graph-Rewiring in GNNs",
                    "abstract": "Graph Neural Networks (GNNs) are popular models for machine learning on graphs that typically follow the message-passing paradigm, whereby the feature of a node is updated recursively upon aggregating information over its neighbors. While exchanging messages over the input graph endows GNNs with a strong inductive bias, it can also make GNNs susceptible to over-squashing, thereby preventing them from capturing long-range interactions in the given graph. To rectify this issue, graph rewiring techniques have been proposed as a means of improving information flow by altering the graph connectivity. In this work, we identify three desiderata for graph-rewiring: (i) reduce over-squashing, (ii) respect the locality of the graph, and (iii) preserve the sparsity of the graph. We highlight fundamental trade-offs that occur between spatial and spectral rewiring techniques; while the former often satisfy (i) and (ii) but not (iii), the latter generally satisfy (i) and (iii) at the expense of (ii). We propose a novel rewiring framework that satisfies all of (i)--(iii) through a locality-aware sequence of rewiring operations. We then discuss a specific instance of such rewiring framework and validate its effectiveness on several real-world benchmarks, showing that it either matches or significantly outperforms existing rewiring approaches."
                },
                {
                    "title": "Parallel Algorithms Align with Neural Execution",
                    "abstract": "Neural algorithmic reasoners are parallel processors. Teaching them sequential algorithms contradicts this nature, rendering a significant share of their computations redundant. Parallel algorithms however may exploit their full computational power, therefore requiring fewer layers to be executed. This drastically reduces training times, as we observe when comparing parallel implementations of searching, sorting and finding strongly connected components to their sequential counterparts on the CLRS framework. Additionally, parallel versions achieve (often strongly) superior predictive performance."
                },
                {
                    "title": "Asynchronous Algorithmic Alignment with Cocycles",
                    "abstract": "State-of-the-art neural algorithmic reasoners make use of message passing in graph neural networks (GNNs). But typical GNNs blur the distinction between the definition and invocation of the message function, forcing a node to send messages to its neighbours at every layer, synchronously. When applying GNNs to learn to execute dynamic programming algorithms, however, on most steps only a handful of the nodes would have meaningful updates to send. One, hence, runs the risk of inefficiencies by sending too much irrelevant data across the graph. But more importantly, many intermediate GNN steps have to learn the identity functions, which is a non-trivial learning problem. In this work, we explicitly separate the concepts of node state update and message function invocation. With this separation, we obtain a mathematical formulation that allows us to reason about asynchronous computation in both algorithms and neural networks. Our analysis yields several practical implementations of synchronous scalable GNN layers that are provably invariant under various forms of asynchrony."
                },
                {
                    "title": "Neural Algorithmic Reasoning Without Intermediate Supervision",
                    "abstract": "Neural algorithmic reasoning is an emerging area of machine learning focusing on building models that can imitate the execution of classic algorithms, such as sorting, shortest paths, etc. One of the main challenges is to learn algorithms that are able to generalize to out-of-distribution data, in particular with significantly larger input sizes. Recent work on this problem has demonstrated the advantages of learning algorithms step-by-step, giving models access to all intermediate steps of the original algorithm. In this work, we instead focus on learning neural algorithmic reasoning only from the input-output pairs without appealing to the intermediate supervision. We propose simple but effective architectural improvements and also build a self-supervised objective that can regularise intermediate computations of the model without access to the algorithm trajectory. We demonstrate that our approach is competitive to its trajectory-supervised counterpart on tasks from the CLRS Algorithmic Reasoning Benchmark and achieves new state-of-the-art results for several problems, including sorting, where we obtain significant improvements. Thus, learning without intermediate supervision is a promising direction for further research on neural reasoners."
                },
                {
                    "title": "How does over-squashing affect the power of GNNs?",
                    "abstract": "Graph Neural Networks (GNNs) are the state-of-the-art model for machine learning on graph-structured data. The most popular class of GNNs operate by exchanging information between adjacent nodes, and are known as Message Passing Neural Networks (MPNNs). Given their widespread use, understanding the expressive power of MPNNs is a key question. However, existing results typically consider settings with uninformative node features. In this paper, we provide a rigorous analysis to determine which function classes of node features can be learned by an MPNN of a given capacity. We do so by measuring the level of pairwise interactions between nodes that MPNNs allow for. This measure provides a novel quantitative characterization of the so-called over-squashing effect, which is observed to occur when a large volume of messages is aggregated into fixed-size vectors. Using our measure, we prove that, to guarantee sufficient communication between pairs of nodes, the capacity of the MPNN must be large enough, depending on properties of the input graph structure, such as commute times. For many relevant scenarios, our analysis results in impossibility statements in practice, showing that over-squashing hinders the expressive power of MPNNs. We validate our theoretical findings through extensive controlled experiments and ablation studies."
                },
                {
                    "title": "Exphormer: Sparse Transformers for Graphs",
                    "abstract": "Graph transformers have emerged as a promising architecture for a variety of graph learning and representation tasks. Despite their successes, though, it remains challenging to scale graph transformers to large graphs while maintaining accuracy competitive with message-passing networks. In this paper, we introduce Exphormer, a framework for building powerful and scalable graph transformers. Exphormer consists of a sparse attention mechanism based on two mechanisms: virtual global nodes and expander graphs, whose mathematical characteristics, such as spectral expansion, pseduorandomness, and sparsity, yield graph transformers with complexity only linear in the size of the graph, while allowing us to prove desirable theoretical properties of the resulting transformer models. We show that incorporating Exphormer into the recently-proposed GraphGPS framework produces models with competitive empirical results on a wide variety of graph datasets, including state-of-the-art results on three datasets. We also show that Exphormer can scale to datasets on larger graphs than shown in previous graph transformer architectures. Code can be found at \\url{https://github.com/hamed1375/Exphormer}."
                },
                {
                    "title": "Neural Algorithmic Reasoning with Causal Regularisation",
                    "abstract": "Recent work on neural algorithmic reasoning has investigated the reasoning capabilities of neural networks, effectively demonstrating they can learn to execute classical algorithms on unseen data coming from the train distribution. However, the performance of existing neural reasoners significantly degrades on out-of-distribution (OOD) test data, where inputs have larger sizes. In this work, we make an important observation: there are many different inputs for which an algorithm will perform certain intermediate computations identically. This insight allows us to develop data augmentation procedures that, given an algorithm's intermediate trajectory, produce inputs for which the target algorithm would have exactly the same next trajectory step. We ensure invariance in the next-step prediction across such inputs, by employing a self-supervised objective derived by our observation, formalised in a causal graph. We prove that the resulting method, which we call Hint-ReLIC, improves the OOD generalisation capabilities of the reasoner. We evaluate our method on the CLRS algorithmic reasoning benchmark, where we show up to 3$\\times$ improvements on the OOD test data."
                },
                {
                    "title": "On the Connection Between MPNN and Graph Transformer",
                    "abstract": "Graph Transformer (GT) recently has emerged as a new paradigm of graph learning algorithms, outperforming the previously popular Message Passing Neural Network (MPNN) on multiple benchmarks. Previous work (Kim et al., 2022) shows that with proper position embedding, GT can approximate MPNN arbitrarily well, implying that GT is at least as powerful as MPNN. In this paper, we study the inverse connection and show that MPNN with virtual node (VN), a commonly used heuristic with little theoretical understanding, is powerful enough to arbitrarily approximate the self-attention layer of GT. In particular, we first show that if we consider one type of linear transformer, the so-called Performer/Linear Transformer (Choromanski et al., 2020; Katharopoulos et al., 2020), then MPNN + VN with only O(1) depth and O(1) width can approximate a self-attention layer in Performer/Linear Transformer. Next, via a connection between MPNN + VN and DeepSets, we prove the MPNN + VN with O(n^d) width and O(1) depth can approximate the self-attention layer arbitrarily well, where d is the input feature dimension. Lastly, under some assumptions, we provide an explicit construction of MPNN + VN with O(1) width and O(n) depth approximating the self-attention layer in GT arbitrarily well. On the empirical side, we demonstrate that 1) MPNN + VN is a surprisingly strong baseline, outperforming GT on the recently proposed Long Range Graph Benchmark (LRGB) dataset, 2) our MPNN + VN improves over early implementation on a wide range of OGB datasets and 3) MPNN + VN outperforms Linear Transformer and MPNN on the climate modeling task."
                },
                {
                    "title": "Towards Better Out-of-Distribution Generalization of Neural Algorithmic Reasoning Tasks",
                    "abstract": "In this paper, we study the OOD generalization of neural algorithmic reasoning tasks, where the goal is to learn an algorithm (e.g., sorting, breadth-first search, and depth-first search) from input-output pairs using deep neural networks. First, we argue that OOD generalization in this setting is significantly different than common OOD settings. For example, some phenomena in OOD generalization of image classifications such as \\emph{accuracy on the line} are not observed here, and techniques such as data augmentation methods do not help as assumptions underlying many augmentation techniques are often violated. Second, we analyze the main challenges (e.g., input distribution shift, non-representative data generation, and uninformative validation metrics) of the current leading benchmark, i.e., CLRS \\citep{deepmind2021clrs}, which contains 30 algorithmic reasoning tasks. We propose several solutions, including a simple-yet-effective fix to the input distribution shift and improved data generation. Finally, we propose an attention-based 2WL-graph neural network (GNN) processor which complements message-passing GNNs so their combination outperforms the state-of-the-art model by a 3% margin averaged over all algorithms. Our code is available at: \\url{https://github.com/smahdavi4/clrs}."
                },
                {
                    "title": "FoSR: First-order spectral rewiring for addressing oversquashing in GNNs",
                    "abstract": "Graph neural networks (GNNs) are able to leverage the structure of graph data by passing messages along the edges of the graph. While this allows GNNs to learn features depending on the graph structure, for certain graph topologies it leads to inefficient information propagation and a problem known as oversquashing. This has recently been linked with the curvature and spectral gap of the graph. On the other hand, adding edges to the message-passing graph can lead to increasingly similar node representations and a problem known as oversmoothing. We propose a computationally efficient algorithm that prevents oversquashing by systematically adding edges to the graph based on spectral expansion. We combine this with a relational architecture, which lets the GNN preserve the original graph structure and provably prevents oversmoothing. We find experimentally that our algorithm outperforms existing graph rewiring methods in several graph classification tasks."
                },
                {
                    "title": "Expander Graph Propagation",
                    "abstract": "Deploying graph neural networks (GNNs) on whole-graph classification or regression tasks is known to be challenging: it often requires computing node features that are mindful of both local interactions in their neighbourhood and the global context of the graph structure. GNN architectures that navigate this space need to avoid pathological behaviours, such as bottlenecks and oversquashing, while ideally having linear time and space complexity requirements. In this work, we propose an elegant approach based on propagating information over expander graphs. We leverage an efficient method for constructing expander graphs of a given size, and use this insight to propose the EGP model. We show that EGP is able to address all of the above concerns, while requiring minimal effort to set up, and provide evidence of its empirical utility on relevant graph classification datasets and baselines in the Open Graph Benchmark. Importantly, using expander graphs as a template for message passing necessarily gives rise to negative curvature. While this appears to be counterintuitive in light of recent related work on oversquashing, we theoretically demonstrate that negatively curved edges are likely to be required to obtain scalable message passing without bottlenecks. To the best of our knowledge, this is a previously unstudied result in the context of graph representation learning, and we believe our analysis paves the way to a novel class of scalable methods to counter oversquashing in GNNs."
                },
                {
                    "title": "Oversquashing in GNNs through the lens of information contraction and graph expansion",
                    "abstract": "The quality of signal propagation in message-passing graph neural networks (GNNs) strongly influences their expressivity as has been observed in recent works. In particular, for prediction tasks relying on long-range interactions, recursive aggregation of node features can lead to an undesired phenomenon called \u201coversquashing\u201d. We present a framework for analyzing oversquashing based on information contraction. Our analysis is guided by a model of reliable computation due to von Neumann that lends a new insight into oversquashing as signal quenching in noisy computation graphs. Building on this, we propose a graph rewiring algorithm aimed at alleviating oversquashing. Our algorithm employs a random local edge flip primitive motivated by an expander graph construction. We compare the spectral expansion properties of our algorithm with that of an existing curvature-based non-local rewiring strategy. Synthetic experiments show that while our algorithm in general has a slower rate of expansion, it is overall computationally cheaper, preserves the node degrees exactly and never disconnects the graph."
                },
                {
                    "title": "DiffWire: Inductive Graph Rewiring via the Lov\u00e1sz Bound",
                    "abstract": "Graph Neural Networks (GNNs) have been shown to achieve competitive results to tackle graph-related tasks, such as node and graph classification, link prediction and node and graph clustering in a variety of domains. Most GNNs use a message passing framework and hence are called MPNNs. Despite their promising results, MPNNs have been reported to suffer from over-smoothing, over-squashing and under-reaching. Graph rewiring and graph pooling have been proposed in the literature as solutions to address these limitations. However, most state-of-the-art graph rewiring methods fail to preserve the global topology of the graph, are neither differentiable nor inductive, and require the tuning of hyper-parameters. In this paper, we propose DiffWire, a novel framework for graph rewiring in MPNNs that is principled, fully differentiable and parameter-free by leveraging the Lov\\'asz bound. The proposed approach provides a unified theory for graph rewiring by proposing two new, complementary layers in MPNNs: CT-Layer, a layer that learns the commute times and uses them as a relevance function for edge re-weighting; and GAP-Layer, a layer to optimize the spectral gap, depending on the nature of the network and the task at hand. We empirically validate the value of each of these layers separately with benchmark datasets for graph classification. We also perform preliminary studies on the use of CT-Layer for homophilic and heterophilic node classification tasks. DiffWire brings together the learnability of commute times to related definitions of curvature, opening the door to creating more expressive MPNNs."
                },
                {
                    "title": "The CLRS Algorithmic Reasoning Benchmark",
                    "abstract": "Learning representations of algorithms is an emerging area of machine learning, seeking to bridge concepts from neural networks with classical algorithms. Several important works have investigated whether neural networks can effectively reason like algorithms, typically by learning to execute them. The common trend in the area, however, is to generate targeted kinds of algorithmic data to evaluate specific hypotheses, making results hard to transfer across publications, and increasing the barrier of entry. To consolidate progress and work towards unified evaluation, we propose the CLRS Algorithmic Reasoning Benchmark, covering classical algorithms from the Introduction to Algorithms textbook. Our benchmark spans a variety of algorithmic reasoning procedures, including sorting, searching, dynamic programming, graph algorithms, string algorithms and geometric algorithms. We perform extensive experiments to demonstrate how several popular algorithmic reasoning baselines perform on these tasks, and consequently, highlight links to several open challenges. Our library is readily available at https://github.com/deepmind/clrs."
                },
                {
                    "title": "Graph Neural Networks are Dynamic Programmers",
                    "abstract": "Recent advances in neural algorithmic reasoning with graph neural networks (GNNs) are propped up by the notion of algorithmic alignment. Broadly, a neural network will be better at learning to execute a reasoning task (in terms of sample complexity) if its individual components align well with the target algorithm. Specifically, GNNs are claimed to align with dynamic programming (DP), a general problem-solving strategy which expresses many polynomial-time algorithms. However, has this alignment truly been demonstrated and theoretically quantified? Here we show, using methods from category theory and abstract algebra, that there exists an intricate connection between GNNs and DP, going well beyond the initial observations over individual algorithms such as Bellman-Ford. Exposing this connection, we easily verify several prior findings in the literature, produce better-grounded GNN architectures for edge-centric tasks, and demonstrate empirical results on the CLRS algorithmic reasoning benchmark. We hope our exposition will serve as a foundation for building stronger algorithmically aligned GNNs."
                },
                {
                    "title": "Understanding over-squashing and bottlenecks on graphs via curvature",
                    "abstract": "Most graph neural networks (GNNs) use the message passing paradigm, in which node features are propagated on the input graph. Recent works pointed to the distortion of information flowing from distant nodes as a factor limiting the efficiency of message passing for tasks relying on long-distance interactions. This phenomenon, referred to as 'over-squashing', has been heuristically attributed to graph bottlenecks where the number of $k$-hop neighbors grows rapidly with $k$. We provide a precise description of the over-squashing phenomenon in GNNs and analyze how it arises from bottlenecks in the graph. For this purpose, we introduce a new edge-based combinatorial curvature and prove that negatively curved edges are responsible for the over-squashing issue. We also propose and experimentally test a curvature-based graph rewiring method to alleviate the over-squashing."
                },
                {
                    "title": "Stabilizing Equilibrium Models by Jacobian Regularization",
                    "abstract": "Deep equilibrium networks (DEQs) are a new class of models that eschews traditional depth in favor of finding the fixed point of a single nonlinear layer. These models have been shown to achieve performance competitive with the state-of-the-art deep networks while using significantly less memory. Yet they are also slower, brittle to architectural choices, and introduce potential instability to the model. In this paper, we propose a regularization scheme for DEQ models that explicitly regularizes the Jacobian of the fixed-point update equations to stabilize the learning of equilibrium models. We show that this regularization adds only minimal computational cost, significantly stabilizes the fixed-point convergence in both forward and backward passes, and scales well to high-dimensional, realistic domains (e.g., WikiText-103 language modeling and ImageNet classification). Using this method, we demonstrate, for the first time, an implicit-depth model that runs with approximately the same speed and level of performance as popular conventional deep networks such as ResNet-101, while still maintaining the constant memory footprint and architectural simplicity of DEQs. Code is available at https://github.com/locuslab/deq ."
                },
                {
                    "title": "OGB-LSC: A Large-Scale Challenge for Machine Learning on Graphs",
                    "abstract": "Enabling effective and efficient machine learning (ML) over large-scale graph data (e.g., graphs with billions of edges) can have a great impact on both industrial and scientific applications. However, existing efforts to advance large-scale graph ML have been largely limited by the lack of a suitable public benchmark. Here we present OGB Large-Scale Challenge (OGB-LSC), a collection of three real-world datasets for facilitating the advancements in large-scale graph ML. The OGB-LSC datasets are orders of magnitude larger than existing ones, covering three core graph learning tasks -- link prediction, graph regression, and node classification. Furthermore, we provide dedicated baseline experiments, scaling up expressive graph ML models to the massive datasets. We show that expressive models significantly outperform simple scalable baselines, indicating an opportunity for dedicated efforts to further improve graph ML at scale. Moreover, OGB-LSC datasets were deployed at ACM KDD Cup 2021 and attracted more than 500 team registrations globally, during which significant performance improvements were made by a variety of innovative techniques. We summarize the common techniques used by the winning solutions and highlight the current best practices in large-scale graph ML. Finally, we describe how we have updated the datasets after the KDD Cup to further facilitate research advances. The OGB-LSC datasets, baseline code, and all the information about the KDD Cup are available at https://ogb.stanford.edu/docs/lsc/ ."
                },
                {
                    "title": "Pointer Graph Networks",
                    "abstract": "Graph neural networks (GNNs) are typically applied to static graphs that are assumed to be known upfront. This static input structure is often informed purely by insight of the machine learning practitioner, and might not be optimal for the actual task the GNN is solving. In absence of reliable domain expertise, one might resort to inferring the latent graph structure, which is often difficult due to the vast search space of possible graphs. Here we introduce Pointer Graph Networks (PGNs) which augment sets or graphs with additional inferred edges for improved model expressivity. PGNs allow each node to dynamically point to another node, followed by message passing over these pointers. The sparsity of this adaptable graph structure makes learning tractable while still being sufficiently expressive to simulate complex algorithms. Critically, the pointing mechanism is directly supervised to model long-term sequences of operations on classical data structures, incorporating useful structural inductive biases from theoretical computer science. Qualitatively, we demonstrate that PGNs can learn parallelisable variants of pointer-based data structures, namely disjoint set unions and link/cut trees. PGNs generalise out-of-distribution to 5x larger test inputs on dynamic graph connectivity tasks, outperforming unrestricted GNNs and Deep Sets."
                },
                {
                    "title": "On the Bottleneck of Graph Neural Networks and its Practical Implications",
                    "abstract": "Graph neural networks (GNNs) were shown to effectively learn from highly structured data containing elements (nodes) with relationships (edges) between them. GNN variants differ in how each node in the graph absorbs the information flowing from its neighbor nodes. In this paper, we highlight an inherent problem in GNNs: the mechanism of propagating information between neighbors creates a bottleneck when every node aggregates messages from its neighbors. This bottleneck causes the over-squashing of exponentially-growing information into fixed-size vectors. As a result, the graph fails to propagate messages flowing from distant nodes and performs poorly when the prediction task depends on long-range information. We demonstrate that the bottleneck hinders popular GNNs from fitting the training data. We show that GNNs that absorb incoming edges equally, like GCN and GIN, are more susceptible to over-squashing than other GNN types. We further show that existing, extensively-tuned, GNN-based models suffer from over-squashing and that breaking the bottleneck improves state-of-the-art results without any hyperparameter tuning or additional weights."
                },
                {
                    "title": "Open Graph Benchmark: Datasets for Machine Learning on Graphs",
                    "abstract": "We present the Open Graph Benchmark (OGB), a diverse set of challenging and realistic benchmark datasets to facilitate scalable, robust, and reproducible graph machine learning (ML) research. OGB datasets are large-scale (up to 100+ million nodes and 1+ billion edges), encompass multiple important graph ML tasks, and cover a diverse range of domains, ranging from social and information networks to biological networks, molecular graphs, source code ASTs, and knowledge graphs. For each dataset, we provide a unified evaluation protocol using meaningful application-specific data splits and evaluation metrics. In addition to building the datasets, we also perform extensive benchmark experiments for each dataset. Our experiments suggest that OGB datasets present significant challenges of scalability to large-scale graphs and out-of-distribution generalization under realistic data splits, indicating fruitful opportunities for future research. Finally, OGB provides an automated end-to-end graph ML pipeline that simplifies and standardizes the process of graph data loading, experimental setup, and model evaluation. OGB will be regularly updated and welcomes inputs from the community. OGB datasets as well as data loaders, evaluation scripts, baseline code, and leaderboards are publicly available at this https URL ."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Neural Execution of Graph Algorithms",
                    "abstract": "Graph Neural Networks (GNNs) are a powerful representational tool for solving problems on graph-structured inputs. In almost all cases so far, however, they have been applied to directly recovering a final solution from raw inputs, without explicit guidance on how to structure their problem-solving. Here, instead, we focus on learning in the space of algorithms: we train several state-of-the-art GNN architectures to imitate individual steps of classical graph algorithms, parallel (breadth-first search, Bellman-Ford) as well as sequential (Prim's algorithm). As graph algorithms usually rely on making discrete decisions within neighbourhoods, we hypothesise that maximisation-based message passing neural networks are best-suited for such objectives, and validate this claim empirically. We also demonstrate how learning in the space of algorithms can yield new opportunities for positive transfer between tasks---showing how learning a shortest-path algorithm can be substantially improved when simultaneously learning a reachability algorithm."
                },
                {
                    "title": "Deep Equilibrium Models",
                    "abstract": "We present a new approach to modeling sequential data: the deep equilibrium model (DEQ). Motivated by an observation that the hidden layers of many existing deep sequence models converge towards some fixed point, we propose the DEQ approach that directly finds these equilibrium points via root-finding. Such a method is equivalent to running an infinite depth (weight-tied) feedforward network, but has the notable advantage that we can analytically backpropagate through the equilibrium point using implicit differentiation. Using this approach, training and prediction in these networks require only constant memory, regardless of the effective \u201cdepth\u201d of the network. We demonstrate how DEQs can be applied to two state-of-the-art deep sequence models: self-attention transformers and trellis networks. On large-scale language modeling tasks, such as the WikiText-103 benchmark, we show that DEQs 1) often improve performance over these state-of-the-art models (for similar parameter counts); 2) have similar computational requirements to existing models; and 3) vastly reduce memory consumption (often the bottleneck for training large sequence models), demonstrating an up-to 88% memory reduction in our experiments. The code is available at https://github.com/locuslab/deq."
                },
                {
                    "title": "Implicit Deep Learning",
                    "abstract": "Implicit deep learning prediction rules generalize the recursive rules of feedforward neural networks. Such rules are based on the solution of a fixed-point equation involving a single vector of hidden features, which is thus only implicitly defined. The implicit framework greatly simplifies the notation of deep learning, and opens up many new possibilities, in terms of novel architectures and algorithms, robustness analysis and design, interpretability, sparsity, and network architecture optimization."
                },
                {
                    "title": "What Can Neural Networks Reason About?",
                    "abstract": "Neural networks have succeeded in many reasoning tasks. Empirically, these tasks require specialized network structures, e.g., Graph Neural Networks (GNNs) perform well on many such tasks, but less structured networks fail. Theoretically, there is limited understanding of why and when a network structure generalizes better than others, although they have equal expressive power. In this paper, we develop a framework to characterize which reasoning tasks a network can learn well, by studying how well its computation structure aligns with the algorithmic structure of the relevant reasoning process. We formally define this algorithmic alignment and derive a sample complexity bound that decreases with better alignment. This framework offers an explanation for the empirical success of popular reasoning models, and suggests their limitations. As an example, we unify seemingly different reasoning tasks, such as intuitive physics, visual question answering, and shortest paths, via the lens of a powerful algorithmic paradigm, dynamic programming (DP). We show that GNNs align with DP and thus are expected to solve these tasks. On several reasoning tasks, our theory is supported by empirical results."
                },
                {
                    "title": "Representation Learning on Graphs with Jumping Knowledge Networks",
                    "abstract": "Recent deep learning approaches for representation learning on graphs follow a neighborhood aggregation procedure. We analyze some important properties of these models, and propose a strategy to overcome those. In particular, the range of \"neighboring\" nodes that a node's representation draws from strongly depends on the graph structure, analogous to the spread of a random walk. To adapt to local neighborhood properties and tasks, we explore an architecture -- jumping knowledge (JK) networks -- that flexibly leverages, for each node, different neighborhood ranges to enable better structure-aware representation. In a number of experiments on social, bioinformatics and citation networks, we demonstrate that our model achieves state-of-the-art performance. Furthermore, combining the JK framework with models like Graph Convolutional Networks, GraphSAGE and Graph Attention Networks consistently improves those models' performance."
                },
                {
                    "title": "Relational inductive bias for physical construction in humans and machines",
                    "abstract": "While current deep learning systems excel at tasks such as object classification, language processing, and gameplay, few can construct or modify a complex system such as a tower of blocks. We hypothesize that what these systems lack is a \"relational inductive bias\": a capacity for reasoning about inter-object relations and making choices over a structured description of a scene. To test this hypothesis, we focus on a task that involves gluing pairs of blocks together to stabilize a tower, and quantify how well humans perform. We then introduce a deep reinforcement learning agent which uses object- and relation-centric scene and policy representations and apply it to the task. Our results show that these structured representations allow the agent to outperform both humans and more naive approaches, suggesting that relational inductive bias is an important component in solving structured reasoning problems and for building more intelligent, flexible machines."
                },
                {
                    "title": "Neural Message Passing for Quantum Chemistry",
                    "abstract": "Supervised learning on molecules has incredible potential to be useful in chemistry, drug discovery, and materials science. Luckily, several promising and closely related neural network models invariant to molecular symmetries have already been described in the literature. These models learn a message passing algorithm and aggregation procedure to compute a function of their entire input graph. At this point, the next step is to find a particularly effective variant of this general approach and apply it to chemical prediction benchmarks until we either solve them or reach the limits of the approach. In this paper, we reformulate existing models into a single common framework we call Message Passing Neural Networks (MPNNs) and explore additional novel variations within this framework. Using MPNNs we demonstrate state of the art results on an important molecular property prediction benchmark; these results are strong enough that we believe future work should focus on datasets with larger molecules or more accurate ground truth labels."
                },
                {
                    "title": "Adam: A Method for Stochastic Optimization",
                    "abstract": "We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
                },
                {
                    "title": "The formal semantics of programming languages - an introduction",
                    "abstract": "\"The Formal Semantics of Programming Languages\" provides the basic mathematical techniques necessary for those who are beginning a study of semantics and logics of programming languages. Although the treatment is elementary, several of the topics covered are drawn from recent research, including the vital area of concurrency. Starting with basic set theory, structural operational semantics is introduced as a way to define the meaning of programming languages along with associated proof techniques. Denotational and axiomatic semantics are illustrated on a simple language of while-programs, and full proofs are given of the equivalence of the operational and denotational semantics and soundness and relative completeness of the axiomatic semantics. A proof of Goedel's incompleteness theorem, which emphasizes the impossibility of achieving a fully complete axiomatic semantics, is included. It is supported by an appendix providing an introduction to the theory of computability based on while-programs. Following a presentation of domain theory, the semantics and methods of proof for several functional languages are treated. The simplest language is that of recursion equations with both call-by-value and call-by-name evaluation. This work is extended to languages with higher and recursive types, including a treatment of the eager and lazy lambda-calculi. Throughout, the relationship between denotational and operational semantics is stressed, and the proofs of the correspondence between the operation and denotational semantics are provided. The treatment of recursive types - one of the more advanced parts of the book - relies on the use of information sytems to represent domains. The book concludes with a chapter on parallel programming languages, accompanied by a discussion of methods for specifying and verifying non-deterministic and parallel programs."
                },
                {
                    "title": "Iterative Procedures for Nonlinear Integral Equations",
                    "abstract": "The numerical solution of nonlinear integral equations involves the iterative soIutioon of finite systems of nonlinear algebraic or transcendental equations. Certain corwent i o n a l techniqucs for treating such systems are reviewed in the context of a particular class of n o n l i n e a r equations. A procedure is synthesized to offset some of the disadvantages of these t e c h n i q u e s in this context; however, the procedure is not restricted to this pt~rticular class of s y s t e m s of nonlinear equations."
                },
                {
                    "title": "A Class of Methods for Solving Nonlinear Simultaneous Equations",
                    "abstract": "solution. The functions that require zeroing are real functions of real variables and it will be assumed that they are continuous and differentiable with respect to these variables. In many practical examples they are extremely complicated anld hence laborious to compute, an-d this fact has two important immediate consequences. The first is that it is impracticable to compute any derivative that may be required by the evaluation of the algebraic expression of this derivative. If derivatives are needed they must be obtained by differencing. The second is that during any iterative solution process the bulk of the computing time will be spent in evaluating the functions. Thus, the most efficient process will tenid to be that which requires the smallest number of function evaluations. This paper discusses certain modificatioins to Newton's method designed to reduce the number of function evaluations required. Results of various numerical experiments are given and conditions under which the modified versions are superior to the original are tentatively suggested."
                },
                {
                    "title": "A LATTICE-THEORETICAL FIXPOINT THEOREM AND ITS APPLICATIONS",
                    "abstract": "1. A lattice-theoretical fixpoint theorem. In this section we formulate and prove an elementary fixpoint theorem which holds in arbitrary complete lattices. In the following sections we give various applications (and extensions) of this result in the theories of simply ordered sets, real functions, Boolean algebras, as well as in general set theory and topology. * By a lattice we understand as usual a system 21 = (A 9 <) formed by a non-empty set A and a binary relation <; it is assumed that < establishes a partial order in A and that for any two elements a f b E A there is a least upper bound (join) a u b and a greatest lower bound (meet) an b. The relations >L, <, and > are defined in the usual way in terms of <. The lattice 21 = (A, <) is called complete if every subset B of A has a least upper bound \u038cB and a greatest lower bound \u03a0\u03b2. Such a lattice has in particular two elements 0 and 1 defined by the formulas 0 = \u0393U and 1 = 11,4. Given any two elements a 9 b E A with a < b, we denote by [a 9 b] the interval with the endpoints a and b, that is, the set of all elements x E A for which a < x < b; in symbols, [ a,b] = E x [x E A and a .< x .< b ]. The system \\[\u03b1,6], <) is clearly a lattice; it is a complete if 21 is complete. We shall consider functions on A to A and, more generally, on a subset B of A to another subset C of A. Such a function / is called increasing if, for any 1 For notions and facts concerning lattices, simply ordered systems, and Boolean algebras consult [l]."
                },
                {
                    "title": "On the evolution of random graphs",
                    "abstract": "(n) k edges have equal probabilities to be chosen as the next one . We shall 2 study the \"evolution\" of such a random graph if N is increased . In this investigation we endeavour to find what is the \"typical\" structure at a given stage of evolution (i . e . if N is equal, or asymptotically equal, to a given function N(n) of n) . By a \"typical\" structure we mean such a structure the probability of which tends to 1 if n -* + when N = N(n) . If A is such a property that lim Pn,N,(n ) ( A) = 1, we shall say that \u201ealmost all\" graphs Gn,N(n) n--possess this property ."
                },
                {
                    "title": "Toward a mathematical semantics for computer languages",
                    "abstract": "Compilers for high-level languages aTe generally constructed to give the complete translation of the programs into machme language. As machines merely juggle bit patterns, the concepts of the original language may be lost or at least obscured during this passage. The purpose of a mathematical semantics is to give a correct and meaningful correspondence between programs and mathematical entities in a way that is entirely independent of an implementation. This plan is illustrated in a very elementary way in the introduction. The first section connects the general method wi th the usual idea of state transformations. The next section shows why the mathematics of functions has to be modified to accommodate recursive commands. Section 3 explains the modifi\u00ad cation. Section 4 introduces the environments for handling variables and identifiers and shows how the semantical equations define equivalence of programs. Section 5 gives an exposition of the new type of mathematical function spaces that are required fOl the semantics of procedures when these are allowed in assignment state\u00ad ments. The conclusion traces some of the background of the project and points the way to future work."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively implement deep equilibrium algorithmic reasoners (DEARs) that learn algorithms by identifying equilibrium points in graph neural networks (GNNs) to improve performance and inference speed?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning, particularly in the deployment of algorithms in real-world scenarios where traditional scalar representations fail. By developing DEARs, we can enhance the robustness and efficiency of algorithmic reasoning, leading to better generalization in unseen contexts. This research could pave the way for more sophisticated applications of GNNs in various domains, such as optimization, robotics, and complex system modeling, ultimately contributing to the evolution of intelligent systems that can adapt to diverse and dynamic environments.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in this problem stem from the need to accurately map complex real-world scenarios into a higher-dimensional vector space while ensuring that the GNNs can effectively identify equilibrium points without prior knowledge of the number of algorithmic steps. Naive approaches may fail due to the intricacies of dynamic programming and the potential for overfitting when aligning GNN architectures with specific algorithms. Additionally, achieving robust out-of-distribution generalization is technically demanding, as it requires the model to perform well on previously unseen input sizes and types.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on aligning GNNs with specific algorithms but has overlooked the significance of equilibrium states at algorithm termination. Existing solutions have not adequately addressed the complexities of identifying these equilibrium points or the need for efficient inference mechanisms. Barriers such as the lack of a principled framework for integrating denotational semantics and domain theory into GNNs have hindered progress. Our approach differs by explicitly targeting equilibrium finding and leveraging optimized root-finding algorithms, which have not been explored in prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing DEARs that utilize denotational semantics and domain theory to identify equilibrium points in GNNs. We will employ benchmark datasets, such as the CLRS-30 algorithmic reasoning benchmark, to evaluate our models. The performance will be measured using metrics that assess both accuracy and inference speed. We expect that DEARs will demonstrate superior performance with less complex GNN architectures while maintaining competitive results compared to"
            }
        },
        "author_data": {
            "fcc5aa4c-f508-448f-ab79-138e1f551a5a": {
                "pk": "fcc5aa4c-f508-448f-ab79-138e1f551a5a",
                "name": "Dobrik Georgiev",
                "collaborators": [
                    "Pietro Li\u00f2",
                    "Pietro Barbiero",
                    "Petar Veli\u010dkovi\u0107",
                    "Danilo Numeroso",
                    "Davide Buffelli",
                    "Gabriele Ciravegna",
                    "Franscesco Giannini",
                    "Valerie Engelmayer",
                    "Marc Brockschmidt",
                    "Miltiadis Allamanis"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Neural Algorithmic Reasoning",
                    "Explainability",
                    "Combinatorial Optimization"
                ],
                "publications": [
                    {
                        "title": "Neural Bipartite Matching",
                        "abstract": "Graph neural networks (GNNs) have found application for learning in the space of algorithms. However, the algorithms chosen by existing research (sorting, Breadth-First search, shortest path finding, etc.) usually align perfectly with a standard GNN architecture. This report describes how neural execution is applied to a complex algorithm, such as finding maximum bipartite matching by reducing it to a flow problem and using Ford-Fulkerson to find the maximum flow. This is achieved via neural execution based only on features generated from a single GNN. The evaluation shows strongly generalising results with the network achieving optimal matching almost 100% of the time."
                    },
                    {
                        "title": "The Deep Equilibrium Algorithmic Reasoner",
                        "abstract": "Recent work on neural algorithmic reasoning has demonstrated that graph neural networks (GNNs) could learn to execute classical algorithms. Doing so, however, has always used a recurrent architecture, where each iteration of the GNN aligns with an algorithm's iteration. Since an algorithm's solution is often an equilibrium, we conjecture and empirically validate that one can train a network to solve algorithmic problems by directly finding the equilibrium. Note that this does not require matching each GNN iteration with a step of the algorithm."
                    },
                    {
                        "title": "PyTorch, Explain! A Python library for Logic Explained Networks",
                        "abstract": "\"PyTorch, Explain!\" is a Python module integrating a variety of state-of-the-art approaches to provide logic explanations from neural networks. This package focuses on bringing these methods to non-specialists. It has minimal dependencies and it is distributed under the Apache 2.0 licence allowing both academic and commercial use. Source code and documentation can be downloaded from the github repository: https://github.com/pietrobarbiero/pytorch_explain."
                    },
                    {
                        "title": "Parallel Algorithms Align with Neural Execution",
                        "abstract": "Neural algorithmic reasoners are parallel processors. Teaching them sequential algorithms contradicts this nature, rendering a significant share of their computations redundant. Parallel algorithms however may exploit their full computational power, therefore requiring fewer layers to be executed. This drastically reduces training times, as we observe when comparing parallel implementations of searching, sorting and finding strongly connected components to their sequential counterparts on the CLRS framework. Additionally, parallel versions achieve (often strongly) superior predictive performance."
                    },
                    {
                        "title": "HEAT: Hyperedge Attention Networks",
                        "abstract": "Learning from structured data is a core machine learning task. Commonly, such data is represented as graphs, which normally only consider (typed) binary relationships between pairs of nodes. This is a substantial limitation for many domains with highly-structured data. One important such domain is source code, where hypergraph-based representations can better capture the semantically rich and structured nature of code.   In this work, we present HEAT, a neural model capable of representing typed and qualified hypergraphs, where each hyperedge explicitly qualifies how participating nodes contribute. It can be viewed as a generalization of both message passing neural networks and Transformers. We evaluate HEAT on knowledge base completion and on bug detection and repair using a novel hypergraph representation of programs. In both settings, it outperforms strong baselines, indicating its power and generality."
                    },
                    {
                        "title": "Algorithmic Concept-based Explainable Reasoning",
                        "abstract": "Recent research on graph neural network (GNN) models successfully applied GNNs to classical graph algorithms and combinatorial optimisation problems. This has numerous benefits, such as allowing applications of algorithms when preconditions are not satisfied, or reusing learned models when sufficient training data is not available or can't be generated. Unfortunately, a key hindrance of these approaches is their lack of explainability, since GNNs are black-box models that cannot be interpreted directly. In this work, we address this limitation by applying existing work on concept-based explanations to GNN models. We introduce concept-bottleneck GNNs, which rely on a modification to the GNN readout mechanism. Using three case studies we demonstrate that: (i) our proposed model is capable of accurately learning concepts and extracting propositional formulas based on the learned concepts for each target class; (ii) our concept-based GNN models achieve comparative performance with state-of-the-art models; (iii) we can derive global graph concepts, without explicitly providing any supervision on graph-level concepts."
                    },
                    {
                        "title": "Neural Algorithmic Reasoning for Combinatorial Optimisation",
                        "abstract": "Solving NP-hard/complete combinatorial problems with neural networks is a challenging research area that aims to surpass classical approximate algorithms. The long-term objective is to outperform hand-designed heuristics for NP-hard/complete problems by learning to generate superior solutions solely from training data. Current neural-based methods for solving CO problems often overlook the inherent \"algorithmic\" nature of the problems. In contrast, heuristics designed for CO problems, e.g. TSP, frequently leverage well-established algorithms, such as those for finding the minimum spanning tree. In this paper, we propose leveraging recent advancements in neural algorithmic reasoning to improve the learning of CO problems. Specifically, we suggest pre-training our neural model on relevant algorithms before training it on CO instances. Our results demonstrate that by using this learning setup, we achieve superior performance compared to non-algorithmically informed deep learning models."
                    },
                    {
                        "title": "Neural Algorithmic Reasoning with Multiple Correct Solutions",
                        "abstract": "Neural Algorithmic Reasoning (NAR) aims to optimize classical algorithms. However, canonical implementations of NAR train neural networks to return only a single solution, even when there are multiple correct solutions to a problem, such as single-source shortest paths. For some applications, it is desirable to recover more than one correct solution. To that end, we give the first method for NAR with multiple solutions. We demonstrate our method on two classical algorithms: Bellman-Ford (BF) and Depth-First Search (DFS), favouring deeper insight into two algorithms over a broader survey of algorithms. This method involves generating appropriate training data as well as sampling and validating solutions from model output. Each step of our method, which can serve as a framework for neural algorithmic reasoning beyond the tasks presented in this paper, might be of independent interest to the field and our results represent the first attempt at this task in the NAR literature."
                    },
                    {
                        "title": "Global Concept-Based Interpretability for Graph Neural Networks via Neuron Analysis",
                        "abstract": "Graph neural networks (GNNs) are highly effective on a variety of graph-related tasks; however, they lack interpretability and transparency. Current explainability approaches are typically local and treat GNNs as black-boxes. They do not look inside the model, inhibiting human trust in the model and explanations. Motivated by the ability of neurons to detect high-level semantic concepts in vision models, we perform a novel analysis on the behaviour of individual GNN neurons to answer questions about GNN interpretability, and propose new metrics for evaluating the interpretability of GNN neurons. We propose a novel approach for producing global explanations for GNNs using neuron-level concepts to enable practitioners to have a high-level view of the model. Specifically, (i) to the best of our knowledge, this is the first work which shows that GNN neurons act as concept detectors and have strong alignment with concepts formulated as logical compositions of node degree and neighbourhood properties; (ii) we quantitatively assess the importance of detected concepts, and identify a trade-off between training duration and neuron-level interpretability; (iii) we demonstrate that our global explainability approach has advantages over the current state-of-the-art -- we can disentangle the explanation into individual interpretable concepts backed by logical descriptions, which reduces potential for bias and improves user-friendliness."
                    }
                ]
            },
            "117e7121-e660-4816-a508-62ba370d3257": {
                "pk": "117e7121-e660-4816-a508-62ba370d3257",
                "name": "JJ Wilson",
                "collaborators": [
                    "Maya Bechler-Speicher",
                    "Petar Veli\u010dkovi\u0107"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Information Propagation",
                    "Over-squashing",
                    "Expander Graphs"
                ],
                "publications": [
                    {
                        "title": "Cayley Graph Propagation",
                        "abstract": "In spite of the plethora of success stories with graph neural networks (GNNs) on modelling graph-structured data, they are notoriously vulnerable to over-squashing, whereby tasks necessitate the mixing of information between distance pairs of nodes. To address this problem, prior work suggests rewiring the graph structure to improve information flow. Alternatively, a significant body of research has dedicated itself to discovering and precomputing bottleneck-free graph structures to ameliorate over-squashing. One well regarded family of bottleneck-free graphs within the mathematical community are expander graphs, with prior work$\\unicode{x2014}$Expander Graph Propagation (EGP)$\\unicode{x2014}$proposing the use of a well-known expander graph family$\\unicode{x2014}$the Cayley graphs of the $\\mathrm{SL}(2,\\mathbb{Z}_n)$ special linear group$\\unicode{x2014}$as a computational template for GNNs. However, in EGP the computational graphs used are truncated to align with a given input graph. In this work, we show that truncation is detrimental to the coveted expansion properties. Instead, we propose CGP, a method to propagate information over a complete Cayley graph structure, thereby ensuring it is bottleneck-free to better alleviate over-squashing. Our empirical evidence across several real-world datasets not only shows that CGP recovers significant improvements as compared to EGP, but it is also akin to or outperforms computationally complex graph rewiring techniques."
                    }
                ]
            },
            "37cd17f8-f93f-49a9-88a6-2bd60cda26ce": {
                "pk": "37cd17f8-f93f-49a9-88a6-2bd60cda26ce",
                "name": "Davide Buffelli",
                "collaborators": [
                    "Fabio Vandin",
                    "Pietro Li\u00f2",
                    "Efthymia Tsamoura",
                    "Farzin Soleymani",
                    "Bastian Rieck",
                    "Dobrik Georgiev",
                    "Ashish Gupta",
                    "Agnieszka Strzalka",
                    "Vassilis Plachouras",
                    "Rishabh Jain"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Meta-Learning",
                    "Explainable AI",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "The Impact of Global Structural Information in Graph Neural Networks Applications",
                        "abstract": "Graph Neural Networks (GNNs) rely on the graph structure to define an aggregation strategy where each node updates its representation by combining information from its neighbours. A known limitation of GNNs is that, as the number of layers increases, information gets smoothed and squashed and node embeddings become indistinguishable, negatively affecting performance. Therefore, practical GNN models employ few layers and only leverage the graph structure in terms of limited, small neighbourhoods around each node. Inevitably, practical GNNs do not capture information depending on the global structure of the graph. While there have been several works studying the limitations and expressivity of GNNs, the question of whether practical applications on graph structured data require global structural knowledge or not, remains unanswered. In this work, we empirically address this question by giving access to global information to several GNN models, and observing the impact it has on downstream performance. Our results show that global information can in fact provide significant benefits for common graph-related tasks. We further identify a novel regularization strategy that leads to an average accuracy improvement of more than 5% on all considered tasks."
                    },
                    {
                        "title": "A Meta-Learning Approach for Graph Representation Learning in Multi-Task Settings",
                        "abstract": "Graph Neural Networks (GNNs) are a framework for graph representation learning, where a model learns to generate low dimensional node embeddings that encapsulate structural and feature-related information. GNNs are usually trained in an end-to-end fashion, leading to highly specialized node embeddings. However, generating node embeddings that can be used to perform multiple tasks (with performance comparable to single-task models) is an open problem. We propose a novel meta-learning strategy capable of producing multi-task node embeddings. Our method avoids the difficulties arising when learning to perform multiple tasks concurrently by, instead, learning to quickly (i.e. with a few steps of gradient descent) adapt to multiple tasks singularly. We show that the embeddings produced by our method can be used to perform multiple tasks with comparable or higher performance than classically trained models. Our method is model-agnostic and task-agnostic, thus applicable to a wide variety of multi-task domains."
                    },
                    {
                        "title": "Scalable Regularization of Scene Graph Generation Models using Symbolic Theories",
                        "abstract": "Several techniques have recently aimed to improve the performance of deep learning models for Scene Graph Generation (SGG) by incorporating background knowledge. State-of-the-art techniques can be divided into two families: one where the background knowledge is incorporated into the model in a subsymbolic fashion, and another in which the background knowledge is maintained in symbolic form. Despite promising results, both families of techniques face several shortcomings: the first one requires ad-hoc, more complex neural architectures increasing the training or inference cost; the second one suffers from limited scalability w.r.t. the size of the background knowledge. Our work introduces a regularization technique for injecting symbolic background knowledge into neural SGG models that overcomes the limitations of prior art. Our technique is model-agnostic, does not incur any cost at inference time, and scales to previously unmanageable background knowledge sizes. We demonstrate that our technique can improve the accuracy of state-of-the-art SGG models, by up to 33%."
                    },
                    {
                        "title": "Attention-Based Deep Learning Framework for Human Activity Recognition with User Adaptation",
                        "abstract": "Sensor-based human activity recognition (HAR) requires to predict the action of a person based on sensor-generated time series data. HAR has attracted major interest in the past few years, thanks to the large number of applications enabled by modern ubiquitous computing devices. While several techniques based on hand-crafted feature engineering have been proposed, the current state-of-the-art is represented by deep learning architectures that automatically obtain high level representations and that use recurrent neural networks (RNNs) to extract temporal dependencies in the input. RNNs have several limitations, in particular in dealing with long-term dependencies. We propose a novel deep learning framework, \\algname, based on a purely attention-based mechanism, that overcomes the limitations of the state-of-the-art. We show that our proposed attention-based architecture is considerably more powerful than previous approaches, with an average increment, of more than $7\\%$ on the F1 score over the previous best performing model. Furthermore, we consider the problem of personalizing HAR deep learning models, which is of great importance in several applications. We propose a simple and effective transfer-learning based strategy to adapt a model to a specific user, providing an average increment of $6\\%$ on the F1 score on the predictions for that user. Our extensive experimental evaluation proves the significantly superior capabilities of our proposed framework over the current state-of-the-art and the effectiveness of our user adaptation technique."
                    },
                    {
                        "title": "Graph Representation Learning for Multi-Task Settings: a Meta-Learning Approach",
                        "abstract": "Graph Neural Networks (GNNs) have become the state-of-the-art method for many applications on graph structured data. GNNs are a model for graph representation learning, which aims at learning to generate low dimensional node embeddings that encapsulate structural and feature-related information. GNNs are usually trained in an end-to-end fashion, leading to highly specialized node embeddings. While this approach achieves great results in the single-task setting, the generation of node embeddings that can be used to perform multiple tasks (with performance comparable to single-task models) is still an open problem. We propose the use of meta-learning to allow the training of a GNN model capable of producing multi-task node embeddings. In particular, we exploit the properties of optimization-based meta-learning to learn GNNs that can produce general node representations by learning parameters that can quickly (i.e. with a few steps of gradient descent) adapt to multiple tasks. Our experiments show that the embeddings produced by a model trained with our purposely designed meta-learning procedure can be used to perform multiple tasks with comparable or, surprisingly, even higher performance than both single-task and multi-task end-to-end models."
                    },
                    {
                        "title": "CliquePH: Higher-Order Information for Graph Neural Networks through Persistent Homology on Clique Graphs",
                        "abstract": "Graph neural networks have become the default choice by practitioners for graph learning tasks such as graph classification and node classification. Nevertheless, popular graph neural network models still struggle to capture higher-order information, i.e., information that goes \\emph{beyond} pairwise interactions. Recent work has shown that persistent homology, a tool from topological data analysis, can enrich graph neural networks with topological information that they otherwise could not capture. Calculating such features is efficient for dimension 0 (connected components) and dimension 1 (cycles). However, when it comes to higher-order structures, it does not scale well, with a complexity of $O(n^d)$, where $n$ is the number of nodes and $d$ is the order of the structures. In this work, we introduce a novel method that extracts information about higher-order structures in the graph while still using the efficient low-dimensional persistent homology algorithm. On standard benchmark datasets, we show that our method can lead to up to $31\\%$ improvements in test accuracy."
                    },
                    {
                        "title": "The Deep Equilibrium Algorithmic Reasoner",
                        "abstract": "Recent work on neural algorithmic reasoning has demonstrated that graph neural networks (GNNs) could learn to execute classical algorithms. Doing so, however, has always used a recurrent architecture, where each iteration of the GNN aligns with an algorithm's iteration. Since an algorithm's solution is often an equilibrium, we conjecture and empirically validate that one can train a network to solve algorithmic problems by directly finding the equilibrium. Note that this does not require matching each GNN iteration with a step of the algorithm."
                    },
                    {
                        "title": "Is Meta-Learning the Right Approach for the Cold-Start Problem in Recommender Systems?",
                        "abstract": "Recommender systems have become fundamental building blocks of modern online products and services, and have a substantial impact on user experience. In the past few years, deep learning methods have attracted a lot of research, and are now heavily used in modern real-world recommender systems. Nevertheless, dealing with recommendations in the cold-start setting, e.g., when a user has done limited interactions in the system, is a problem that remains far from solved. Meta-learning techniques, and in particular optimization-based meta-learning, have recently become the most popular approaches in the academic research literature for tackling the cold-start problem in deep learning models for recommender systems. However, current meta-learning approaches are not practical for real-world recommender systems, which have billions of users and items, and strict latency requirements. In this paper we show that it is possible to obtaining similar, or higher, performance on commonly used benchmarks for the cold-start problem without using meta-learning techniques. In more detail, we show that, when tuned correctly, standard and widely adopted deep learning models perform just as well as newer meta-learning models. We further show that an extremely simple modular approach using common representation learning techniques, can perform comparably to meta-learning techniques specifically designed for the cold-start setting while being much more easily deployable in real-world applications."
                    },
                    {
                        "title": "SizeShiftReg: a Regularization Method for Improving Size-Generalization in Graph Neural Networks",
                        "abstract": "In the past few years, graph neural networks (GNNs) have become the de facto model of choice for graph classification. While, from the theoretical viewpoint, most GNNs can operate on graphs of any size, it is empirically observed that their classification performance degrades when they are applied on graphs with sizes that differ from those in the training data. Previous works have tried to tackle this issue in graph classification by providing the model with inductive biases derived from assumptions on the generative process of the graphs, or by requiring access to graphs from the test domain. The first strategy is tied to the quality of the assumptions made for the generative process, and requires the use of specific models designed after the explicit definition of the generative process of the data, leaving open the question of how to improve the performance of generic GNN models in general settings. On the other hand, the second strategy can be applied to any GNN, but requires access to information that is not always easy to obtain. In this work we consider the scenario in which we only have access to the training data, and we propose a regularization strategy that can be applied to any GNN to improve its generalization capabilities from smaller to larger graphs without requiring access to the test data. Our regularization is based on the idea of simulating a shift in the size of the training graphs using coarsening techniques, and enforcing the model to be robust to such a shift. Experimental results on standard datasets show that popular GNN models, trained on the 50% smallest graphs in the dataset and tested on the 10% largest graphs, obtain performance improvements of up to 30% when trained with our regularization strategy."
                    },
                    {
                        "title": "Extending Logic Explained Networks to Text Classification",
                        "abstract": "Recently, Logic Explained Networks (LENs) have been proposed as explainable-by-design neural models providing logic explanations for their predictions. However, these models have only been applied to vision and tabular data, and they mostly favour the generation of global explanations, while local ones tend to be noisy and verbose. For these reasons, we propose LENp, improving local explanations by perturbing input words, and we test it on text classification. Our results show that (i) LENp provides better local explanations than LIME in terms of sensitivity and faithfulness, and (ii) logic explanations are more useful and user-friendly than feature scoring provided by LIME as attested by a human survey."
                    }
                ]
            },
            "a0267989-26e6-48fb-962e-89b1bb6151e3": {
                "pk": "a0267989-26e6-48fb-962e-89b1bb6151e3",
                "name": "Pietro Li\u00f2",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2405.14633": {
        "paper_data": {
            "title": "Flatten Anything: Unsupervised Neural Surface Parameterization",
            "url": "http://arxiv.org/abs/2405.14633v1",
            "arxiv_id": "2405.14633",
            "authors": [
                "Qijian Zhang",
                "Junhui Hou",
                "Wenping Wang",
                "Ying He"
            ],
            "abstract": "Surface parameterization plays an essential role in numerous computer graphics and geometry processing applications. Traditional parameterization approaches are designed for high-quality meshes laboriously created by specialized 3D modelers, thus unable to meet the processing demand for the current explosion of ordinary 3D data. Moreover, their working mechanisms are typically restricted to certain simple topologies, thus relying on cumbersome manual efforts (e.g., surface cutting, part segmentation) for pre-processing. In this paper, we introduce the Flatten Anything Model (FAM), an unsupervised neural architecture to achieve global free-boundary surface parameterization via learning point-wise mappings between 3D points on the target geometric surface and adaptively-deformed UV coordinates within the 2D parameter domain. To mimic the actual physical procedures, we ingeniously construct geometrically-interpretable sub-networks with specific functionalities of surface cutting, UV deforming, unwrapping, and wrapping, which are assembled into a bi-directional cycle mapping framework. Compared with previous methods, our FAM directly operates on discrete surface points without utilizing connectivity information, thus significantly reducing the strict requirements for mesh quality and even applicable to unstructured point cloud data. More importantly, our FAM is fully-automated without the need for pre-cutting and can deal with highly-complex topologies, since its learning process adaptively finds reasonable cutting seams and UV boundaries. Extensive experiments demonstrate the universality, superiority, and inspiring potential of our proposed neural surface parameterization paradigm. The code will be publicly available.",
            "introduction": "   1 Introduction  Surface parameterization intuitively refers to the process of flattening a 3D geometric surface onto a 2D plane, which is typically called the parameter domain. For any 3D spatial point (x,y,z)\ud835\udc65\ud835\udc66\ud835\udc67(x,y,z)( italic_x , italic_y , italic_z ) lying on the underlying surface, we explicitly map it to a 2D planar coordinate (u,v)\ud835\udc62\ud835\udc63(u,v)( italic_u , italic_v ) while satisfying certain continuity and distortion constraints. Building such point-to-point mappings is also known as UV unwrapping, which serves as an indispensable component in modern graphics rendering pipelines for texture mapping and is also widely used in many downstream geometry processing applications, such as remeshing, mesh completion/compression, detail transfer, surface fitting, and editing, etc.   Theoretically, for any two geometric surfaces with identical/similar topological structures, there exists a bijective mapping between them. Nevertheless, when the topology of the target 3D surface becomes complicated (e.g., with high genus), one must pre-open the original mesh to a sufficiently-developable disk along appropriate cutting seams. Consequently, the current industrial practice for implementing UV unwrapping is typically composed of two stages: (1) manually specifying some necessary cutting seams on the original mesh; (2) applying mature disk-topology-oriented parameterization algorithms (e.g., LSCM\u00a0[16], ABF\u00a0[31, 33]) which have been well integrated into popular 3D modeling software (e.g., Blender, Unity, 3Ds Max) to produce per-vertex 2D UV coordinates. In practice, such stage-wise UV unwrapping pipeline still shows the following limitations and inconveniences:   1) Commonly-used surface parameterization algorithms are designed for well-behaved meshes, which are typically produced by specialized 3D modelers and technical artists. However, with the advent of user-generated content (UGC) fueled by the rapidly growing 3D sensing, reconstruction\u00a0[24, 15], and generation\u00a0[26, 28, 18, 35] techniques, there emerges an urgent need for dealing with ordinary 3D data possibly with unruly anomalies, inferior triangulations, and non-ideal geometries.   2) Despite the existence of a few early attempts at heuristic seam generation [32, 8], the process of finding high-quality cutting seams in practical applications still relies on manual efforts and personal experience. Hence, the entire workflow remains subjective and semi-automated, leading to reduced reliability and efficiency, especially when dealing with complex 3D models that users are unfamiliar with. Besides, since cutting is actually achieved via edge selection, one may need to repeatedly adjust the distribution of mesh edges to allow the cutting seams to walk through.   3) The procedures of cutting the original mesh into a disk and then flattening the resulting disk onto the parameter domain should have been mutually influenced and jointly optimized; otherwise, the overall surface parameterization results could be sub-optimal.   In recent years, there has emerged a new family of neural parameterization approaches targeted at learning parameterized 3D geometric representations via neural network architectures. The pioneering works of FoldingNet\u00a0[40] and AtlasNet\u00a0[12] are among the best two representatives, which can build point-wise mappings via deforming a pre-defined 2D lattice grid to reconstruct the target 3D shape. RegGeoNet\u00a0[42] and Flattening-Net\u00a0[43] tend to achieve fixed-boundary rectangular structurization of irregular 3D point clouds through geometry image [13, 21] representations. However, these two approaches cannot be regarded as real-sense surface parameterization due to their lack of mapping constraints. DiffSR\u00a0[2] and Nuvo\u00a0[37] focus on the local",
            "references": [
                {
                    "title": "Nuvo: Neural UV Mapping for Unruly 3D Representations",
                    "abstract": "Existing UV mapping algorithms are designed to operate on well-behaved meshes, instead of the geometry representations produced by state-of-the-art 3D reconstruction and generation techniques. As such, applying these methods to the volume densities recovered by neural radiance fields and related techniques (or meshes triangulated from such fields) results in texture atlases that are too fragmented to be useful for tasks such as view synthesis or appearance editing. We present a UV mapping method designed to operate on geometry produced by 3D reconstruction and generation techniques. Instead of computing a mapping defined on a mesh's vertices, our method Nuvo uses a neural field to represent a continuous UV mapping, and optimizes it to be a valid and well-behaved mapping for just the set of visible points, i.e. only points that affect the scene's appearance. We show that our model is robust to the challenges posed by ill-behaved geometry, and that it produces editable UV mappings that can represent detailed appearance."
                },
                {
                    "title": "MeshGPT: Generating Triangle Meshes with Decoder-Only Transformers",
                    "abstract": "We introduce MeshGPT, a new approach for generating triangle meshes that reflects the compactness typical of artist-created meshes, in contrast to dense triangle meshes extracted by iso-surfacing methods from neural fields. Inspired by recent advances in powerful large language models, we adopt a sequence-based approach to autoregressively generate triangle meshes as sequences of triangles. We first learn a vocabulary of latent quantized embeddings, using graph convolutions, which inform these embeddings of the local mesh geometry and topology. These embeddings are sequenced and decoded into triangles by a decoder, ensuring that they can effectively reconstruct the mesh. A transformer is then trained on this learned vocabulary to predict the index of the next embedding given previous embeddings. Once trained, our model can be autoregressively sampled to generate new triangle meshes, directly generating compact meshes with sharp edges, more closely imitating the efficient triangulation patterns of human-crafted meshes. MeshGPT demonstrates a notable improvement over state of the art mesh generation methods, with a 9% increase in shape coverage and a 30-point enhancement in FID scores across various categories."
                },
                {
                    "title": "3D Gaussian Splatting for Real-Time Radiance Field Rendering",
                    "abstract": "Radiance Field methods have recently revolutionized novel-view synthesis of scenes captured with multiple photos or videos. However, achieving high visual quality still requires neural networks that are costly to train and render, while recent faster methods inevitably trade off speed for quality. For unbounded and complete scenes (rather than isolated objects) and 1080p resolution rendering, no current method can achieve real-time display rates. We introduce three key elements that allow us to achieve state-of-the-art visual quality while maintaining competitive training times and importantly allow high-quality real-time (\u2265 30 fps) novel-view synthesis at 1080p resolution. First, starting from sparse points produced during camera calibration, we represent the scene with 3D Gaussians that preserve desirable properties of continuous volumetric radiance fields for scene optimization while avoiding unnecessary computation in empty space; Second, we perform interleaved optimization/density control of the 3D Gaussians, notably optimizing anisotropic covariance to achieve an accurate representation of the scene; Third, we develop a fast visibility-aware rendering algorithm that supports anisotropic splatting and both accelerates training and allows realtime rendering. We demonstrate state-of-the-art visual quality and real-time rendering on several established datasets."
                },
                {
                    "title": "Flattening-Net: Deep Regular 2D Representation for 3D Point Cloud Analysis",
                    "abstract": "Point clouds are characterized by irregularity and unstructuredness, which pose challenges in efficient data exploitation and discriminative feature extraction. In this paper, we present an unsupervised deep neural architecture called Flattening-Net to represent irregular 3D point clouds of arbitrary geometry and topology as a completely regular 2D point geometry image (PGI) structure, in which coordinates of spatial points are captured in colors of image pixels. Intuitively, Flattening-Net implicitly approximates a locally smooth 3D-to-2D surface flattening process while effectively preserving neighborhood consistency. As a generic representation modality, PGI inherently encodes the intrinsic property of the underlying manifold structure and facilitates surface-style point feature aggregation. To demonstrate its potential, we construct a unified learning framework directly operating on PGIs to achieve diverse types of high-level and low-level downstream applications driven by specific task networks, including classification, segmentation, reconstruction, and upsampling. Extensive experiments demonstrate that our methods perform favorably against the current state-of-the-art competitors."
                },
                {
                    "title": "Point-E: A System for Generating 3D Point Clouds from Complex Prompts",
                    "abstract": "While recent work on text-conditional 3D object generation has shown promising results, the state-of-the-art methods typically require multiple GPU-hours to produce a single sample. This is in stark contrast to state-of-the-art generative image models, which produce samples in a number of seconds or minutes. In this paper, we explore an alternative method for 3D object generation which produces 3D models in only 1-2 minutes on a single GPU. Our method first generates a single synthetic view using a text-to-image diffusion model, and then produces a 3D point cloud using a second diffusion model which conditions on the generated image. While our method still falls short of the state-of-the-art in terms of sample quality, it is one to two orders of magnitude faster to sample from, offering a practical trade-off for some use cases. We release our pre-trained point cloud diffusion models, as well as evaluation code and models, at https://github.com/openai/point-e."
                },
                {
                    "title": "DA Wand: Distortion-Aware Selection Using Neural Mesh Parameterization",
                    "abstract": "We present a neural technique for learning to select a local sub-region around a point which can be used for mesh parameterization. The motivation for our framework is driven by interactive workflows used for decaling, texturing, or painting on surfaces. Our key idea is to incorporate segmentation probabilities as weights of a classical parameterization method, implemented as a novel differentiable parameterization layer within a neural network framework. We train a segmentation network to select 3D regions that are parameterized into 2D and penalized by the resulting distortion, giving rise to segmentations which are distortion-aware. Following training, a user can use our system to interactively select a point on the mesh and obtain a large, meaningful region around the selection which induces a low-distortion parameterization. Our code11https://github.com/threedle/DA-Wand and project22https://threedle.github.io/DA-Wand/ are publicly available."
                },
                {
                    "title": "Magic3D: High-Resolution Text-to-3D Content Creation",
                    "abstract": "DreamFusion [31] has recently demonstrated the utility of a pretrained text-to-image diffusion model to optimize Neural Radiance Fields (NeRF) [23], achieving remarkable text-to-3D synthesis results. However, the method has two inherent limitations: (a) extremely slow optimization of NeRF and (b) low-resolution image space supervision on NeRF, leading to low-quality 3D models with a long processing time. In this paper, we address these limitations by utilizing a two-stage optimization framework. First, we obtain a coarse model using a low-resolution diffusion prior and accelerate with a sparse 3D hash grid structure. Using the coarse representation as the initialization, we further optimize a textured 3D mesh model with an efficient differentiable renderer interacting with a high-resolution latent diffusion model. Our method, dubbed Magic3D, can create high quality 3D mesh models in 40 minutes, which is 2\u00d7 faster than DreamFusion (reportedly taking 1.5 hours on average), while also achieving higher resolution. User studies show 61.7% raters to prefer our approach over DreamFusion. Together with the image-conditioned generation capabilities, we provide users with new ways to control 3D synthesis, opening up new avenues to various creative applications."
                },
                {
                    "title": "DreamFusion: Text-to-3D using 2D Diffusion",
                    "abstract": "Recent breakthroughs in text-to-image synthesis have been driven by diffusion models trained on billions of image-text pairs. Adapting this approach to 3D synthesis would require large-scale datasets of labeled 3D data and efficient architectures for denoising 3D data, neither of which currently exist. In this work, we circumvent these limitations by using a pretrained 2D text-to-image diffusion model to perform text-to-3D synthesis. We introduce a loss based on probability density distillation that enables the use of a 2D diffusion model as a prior for optimization of a parametric image generator. Using this loss in a DeepDream-like procedure, we optimize a randomly-initialized 3D model (a Neural Radiance Field, or NeRF) via gradient descent such that its 2D renderings from random angles achieve a low loss. The resulting 3D model of the given text can be viewed from any angle, relit by arbitrary illumination, or composited into any 3D environment. Our approach requires no 3D training data and no modifications to the image diffusion model, demonstrating the effectiveness of pretrained image diffusion models as priors."
                },
                {
                    "title": "Neural Surface Maps",
                    "abstract": "Maps are arguably one of the most fundamental concepts used to define and operate on manifold surfaces in differentiable geometry. Accordingly, in geometry processing, maps are ubiquitous and are used in many core applications, such as paramterization, shape analysis, remeshing, and deformation. Unfortunately, most computational representations of surface maps do not lend themselves to manipulation and optimization, usually entailing hard, discrete problems. While algorithms exist to solve these problems, they are problem-specific, and a general framework for surface maps is still in need.In this paper, we advocate considering neural networks as encoding surface maps. Since neural networks can be composed on one another and are differentiable, we show it is easy to use them to define surfaces via atlases, compose them for surface-to-surface mappings, and optimize differentiable objectives relating to them, such as any notion of distortion, in a trivial manner. In our experiments, we represent surfaces by generating a neural map that approximates a UV parameterization of a 3D model. Then, we compose this map with other neural maps which we optimize with respect to distortion measures. We show that our formulation enables trivial optimization of rather elusive mapping tasks, such as maps between a collection of surfaces."
                },
                {
                    "title": "TearingNet: Point Cloud Autoencoder to Learn Topology-Friendly Representations",
                    "abstract": "Topology matters. Despite the recent success of point cloud processing with geometric deep learning, it remains arduous to capture the complex topologies of point cloud data with a learning model. Given a point cloud dataset containing objects with various genera, or scenes with multiple objects, we propose an autoencoder, TearingNet, which tackles the challenging task of representing the point clouds using a fixed-length descriptor. Unlike existing works directly deforming predefined primitives of genus zero (e.g., a 2D square patch) to an object-level point cloud, our TearingNet is characterized by a proposed Tearing network module and a Folding network module interacting with each other iteratively. Particularly, the Tearing network module learns the point cloud topology explicitly. By breaking the edges of a primitive graph, it tears the graph into patches or with holes to emulate the topology of a target point cloud, leading to faithful reconstructions. Experimentation shows the superiority of our proposal in terms of reconstructing point clouds as well as generating more topology-friendly representations than benchmarks."
                },
                {
                    "title": "Shape Reconstruction by Learning Differentiable Surface Representations",
                    "abstract": "Generative models that produce point clouds have emerged as a powerful tool to represent 3D surfaces, and the best current ones rely on learning an ensemble of parametric representations. Unfortunately, they offer no control over the deformations of the surface patches that form the ensemble and thus fail to prevent them from either overlapping or collapsing into single points or lines. As a consequence, computing shape properties such as surface normals and curvatures becomes difficult and unreliable. In this paper, we show that we can exploit the inherent differentiability of deep networks to leverage differential surface properties during training so as to prevent patch collapse and strongly reduce patch overlap. Furthermore, this lets us reliably compute quantities such as surface normals and curvatures. We will demonstrate on several tasks that this yields more accurate surface reconstructions than the state-of-the-art methods in terms of normals estimation and amount of collapsed and overlapped patches."
                },
                {
                    "title": "Learning elementary structures for 3D shape generation and matching",
                    "abstract": "We propose to represent shapes as the deformation and combination of learnable elementary 3D structures, which are primitives resulting from training over a collection of shape. We demonstrate that the learned elementary 3D structures lead to clear improvements in 3D shape generation and matching. More precisely, we present two complementary approaches for learning elementary structures: (i) patch deformation learning and (ii) point translation learning. Both approaches can be extended to abstract structures of higher dimensions for improved results. We evaluate our method on two tasks: reconstructing ShapeNet objects and estimating dense correspondences between human scans (FAUST inter challenge). We show 16% improvement over surface deformation approaches for shape reconstruction and outperform FAUST inter challenge state of the art by 6%."
                },
                {
                    "title": "Deep Unsupervised Learning of 3D Point Clouds via Graph Topology Inference and Filtering",
                    "abstract": "We propose a deep autoencoder with graph topology inference and filtering to achieve compact representations of unorganized 3D point clouds in an unsupervised manner. Many previous works discretize 3D points to voxels and then use lattice-based methods to process and learn 3D spatial information; however, this leads to inevitable discretization errors. In this work, we try to handle raw 3D points without such compromise. The proposed networks follow the autoencoder framework with a focus on designing the decoder. The encoder of the proposed networks adopts similar architectures as in PointNet, which is a well-acknowledged method for supervised learning of 3D point clouds. The decoder of the proposed networks involves three novel modules: the folding module, the graph-topology-inference module, and the graph-filtering module. The folding module folds a canonical 2D lattice to the underlying surface of a 3D point cloud, achieving coarse reconstruction; the graph-topology-inference module learns a graph topology to represent pairwise relationships between 3D points, pushing the latent code to preserve both coordinates and pairwise relationships of points in 3D point clouds; and the graph-filtering module couples the above two modules, refining the coarse reconstruction through a learnt graph topology to obtain the final reconstruction. The proposed decoder leverages a learnable graph topology to push the codeword to preserve representative features and further improve the unsupervised-learning performance. We further provide theoretical analyses of the proposed architecture. We provide an upper bound for the reconstruction loss and further show the superiority of graph smoothness over spatial smoothness as a prior to model 3D point clouds. In the experiments, we validate the proposed networks in three tasks, including 3D point cloud reconstruction, visualization, and transfer classification. The experimental results show that (1) the proposed networks outperform the state-of-the-art methods in various tasks, including reconstruction and transfer classification; (2) a graph topology can be inferred as auxiliary information without specific supervision on graph topology inference; (3) graph filtering refines the reconstruction, leading to better performances; and (4) designing a powerful decoder could improve the unsupervised-learning performance, just like a powerful encoder."
                },
                {
                    "title": "On the Spectral Bias of Neural Networks",
                    "abstract": "Neural networks are known to be a class of highly expressive functions able to fit even random input-output mappings with $100\\%$ accuracy. In this work, we present properties of neural networks that complement this aspect of expressivity. By using tools from Fourier analysis, we show that deep ReLU networks are biased towards low frequency functions, meaning that they cannot have local fluctuations without affecting their global behavior. Intuitively, this property is in line with the observation that over-parameterized networks find simple patterns that generalize across data samples. We also investigate how the shape of the data manifold affects expressivity by showing evidence that learning high frequencies gets \\emph{easier} with increasing manifold complexity, and present a theoretical understanding of this behavior. Finally, we study the robustness of the frequency components with respect to parameter perturbation, to develop the intuition that the parameters must be finely tuned to express high frequency functions."
                },
                {
                    "title": "A Papier-Mache Approach to Learning 3D Surface Generation",
                    "abstract": "We introduce a method for learning to generate the surface of 3D shapes. Our approach represents a 3D shape as a collection of parametric surface elements and, in contrast to methods generating voxel grids or point clouds, naturally infers a surface representation of the shape. Beyond its novelty, our new shape generation framework, AtlasNet, comes with significant advantages, such as improved precision and generalization capabilities, and the possibility to generate a shape of arbitrary resolution without memory issues. We demonstrate these benefits and compare to strong baselines on the ShapeNet benchmark for two applications: (i) autoencoding shapes, and (ii) single-view reconstruction from a still image. We also provide results showing its potential for other applications, such as morphing, parametrization, super-resolution, matching, and co-segmentation."
                },
                {
                    "title": "FoldingNet: Point Cloud Auto-Encoder via Deep Grid Deformation",
                    "abstract": "Recent deep networks that directly handle points in a point set, e.g., PointNet, have been state-of-the-art for supervised learning tasks on point clouds such as classification and segmentation. In this work, a novel end-to-end deep auto-encoder is proposed to address unsupervised learning challenges on point clouds. On the encoder side, a graph-based enhancement is enforced to promote local structures on top of PointNet. Then, a novel folding-based decoder deforms a canonical 2D grid onto the underlying 3D object surface of a point cloud, achieving low reconstruction errors even for objects with delicate structures. The proposed decoder only uses about 7% parameters of a decoder with fully-connected neural networks, yet leads to a more discriminative representation that achieves higher linear SVM classification accuracy than the benchmark. In addition, the proposed decoder structure is shown, in theory, to be a generic architecture that is able to reconstruct an arbitrary point cloud from a 2D grid. Our code is available at http://www.merl.com/research/license#FoldingNet"
                },
                {
                    "title": "Scalable Locally Injective Mappings",
                    "abstract": "We present a scalable approach for the optimization of flip-preventing energies in the general context of simplicial mappings and specifically for mesh parameterization. Our iterative minimization is based on the observation that many distortion energies can be optimized indirectly by minimizing a family of simpler proxy energies. Minimization of these proxies is a natural extension of the local/global minimization of the ARAP energy. Our algorithm is simple to implement and scales to datasets with millions of faces. We demonstrate our approach for the computation of maps that minimize a conformal or isometric distortion energy, both in two and three dimensions. In addition to mesh parameterization, we show that our algorithm can be applied to mesh deformation and mesh quality improvement."
                },
                {
                    "title": "TEMPO: Feature-Endowed Teichm\u00fcller Extremal Mappings of Point Clouds",
                    "abstract": "In recent decades, the use of 3D point clouds has been widespread in computer industry. The development of techniques in analyzing point clouds is increasingly important. In particular, mapping of point clouds has been a challenging problem. In this paper, we develop a discrete analogue of the Teichm\\\"{u}ller extremal mappings, which guarantee uniform conformality distortions, on point cloud surfaces. Based on the discrete analogue, we propose a novel method called TEMPO for computing Teichm\\\"{u}ller extremal mappings between feature-endowed point clouds. Using our proposed method, the Teichm\\\"{u}ller metric is introduced for evaluating the dissimilarity of point clouds. Consequently, our algorithm enables accurate recognition and classification of point clouds. Experimental results demonstrate the effectiveness of our proposed method."
                },
                {
                    "title": "Spherical Conformal Parameterization of Genus-0 Point Clouds for Meshing",
                    "abstract": "Point cloud is the most fundamental representation of 3D geometric objects. Analyzing and processing point cloud surfaces is important in computer graphics and computer vision. However, most of the existing algorithms for surface analysis require connectivity information. Therefore, it is desirable to develop a mesh structure on point clouds. This task can be simplified with the aid of a parameterization. In particular, conformal parameterizations are advantageous in preserving the geometric information of the point cloud data. In this paper, we extend a state-of-the-art spherical conformal parameterization algorithm for genus-0 closed meshes to the case of point clouds, using an improved approximation of the Laplace-Beltrami operator on data points. Then, we propose an iterative scheme called the North-South reiteration for achieving a spherical conformal parameterization. A balancing scheme is introduced to enhance the distribution of the spherical parameterization. High quality triangulations and quadrangulations can then be built on the point clouds with the aid of the parameterizations. Also, the meshes generated are guaranteed to be genus-0 closed meshes. Moreover, using our proposed spherical conformal parameterization, multilevel representations of point clouds can be easily constructed. Experimental results demonstrate the effectiveness of our proposed framework."
                },
                {
                    "title": "Bijective parameterization with free boundaries",
                    "abstract": "We present a fully automatic method for generating guaranteed bijective surface parameterizations from triangulated 3D surfaces partitioned into charts. We do so by using a distortion metric that prevents local folds of triangles in the parameterization and a barrier function that prevents intersection of the chart boundaries. In addition, we show how to modify the line search of an interior point method to directly compute the singularities of the distortion metric and barrier functions to maintain a bijective map. By using an isometric metric that is efficient to compute and a spatial hash to accelerate the evaluation and gradient of the barrier function for the boundary, we achieve fast optimization times. Unlike previous methods, we do not require the boundary be constrained by the user to a non-intersecting shape to guarantee a bijection, and the boundary of the parameterization is free to change shape during the optimization to minimize distortion."
                },
                {
                    "title": "Geometric understanding of point clouds using Laplace-Beltrami operator",
                    "abstract": "In this paper, we propose a general framework for approximating differential operator directly on point clouds and use it for geometric understanding on them. The discrete approximation of differential operator on the underlying manifold represented by point clouds is based only on local approximation using nearest neighbors, which is simple, efficient and accurate. This allows us to extract the complete local geometry, solve partial differential equations and perform intrinsic calculations on surfaces. Since no mesh or parametrization is needed, our method can work with point clouds in any dimensions or co-dimensions or even with variable dimensions. The computation complexity scaled well with the number of points and the intrinsic dimensions (rather than the embedded dimensions). We use this method to define the Laplace-Beltrami (LB) operator on point clouds, which links local and global information together. With this operator, we propose a few key applications essential to geometric understanding for point clouds, including the computation of LB eigenvalues and eigenfunctions, the extraction of skeletons from point clouds, and the extraction of conformal structures from point clouds."
                },
                {
                    "title": "A Local/Global Approach to Mesh Parameterization",
                    "abstract": "We present a novel approach to parameterize a mesh with disk topology to the plane in a shape\u2010preserving manner. Our key contribution is a local/global algorithm, which combines a local mapping of each 3D triangle to the plane, using transformations taken from a restricted set, with a global \u201cstitch\u201d operation of all triangles, involving a sparse linear system. The local transformations can be taken from a variety of families, e.g. similarities or rotations, generating different types of parameterizations. In the first case, the parameterization tries to force each 2D triangle to be an as\u2010similar\u2010as\u2010possible version of its 3D counterpart. This is shown to yield results identical to those of the LSCM algorithm. In the second case, the parameterization tries to force each 2D triangle to be an as\u2010rigid\u2010as\u2010possible version of its 3D counterpart. This approach preserves shape as much as possible. It is simple, effective, and fast, due to pre\u2010factoring of the linear system involved in the global phase. Experimental results show that our approach provides almost isometric parameterizations and obtains more shape\u2010preserving results than other state\u2010of\u2010the\u2010art approaches."
                },
                {
                    "title": "Meshless thin-shell simulation based on global conformal parameterization",
                    "abstract": "This paper presents a new approach to the physically-based thin-shell simulation of point-sampled geometry via explicit, global conformal point-surface parameterization and meshless dynamics. The point-based global parameterization is founded upon the rigorous mathematics of Riemann surface theory and Hodge theory. The parameterization is globally conformal everywhere except for a minimum number of zero points. Within our parameterization framework, any well-sampled point surface is functionally equivalent to a manifold, enabling popular and powerful surface-based modeling and physically-based simulation tools to be readily adapted for point geometry processing and animation. In addition, we propose a meshless surface computational paradigm in which the partial differential equations (for dynamic physical simulation) can be applied and solved directly over point samples via moving least squares (MLS) shape functions defined on the global parametric domain without explicit connectivity information. The global conformal parameterization provides a common domain to facilitate accurate meshless simulation and efficient discontinuity modeling for complex branching cracks. Through our experiments on thin-shell elastic deformation and fracture simulation, we demonstrate that our integrative method is very natural, and that it has great potential to further broaden the application scope of point-sampled geometry in graphics and relevant fields"
                },
                {
                    "title": "ABF++: fast and robust angle based flattening",
                    "abstract": "Conformal parameterization of mesh models has numerous applicationsin geometry processing. Conformality is desirable for remeshing,surface reconstruction, and many other mesh processingapplications. Subject to the conformality requirement, theseapplications typically benefit from parameterizations with smallerstretch. The Angle Based Flattening (ABF) method, presented a fewyears ago, generates provably valid conformal parameterizationswith low stretch. However, it is quite time-consuming and becomeserror prone for large meshes due to numerical error accumulation.This work presents ABF&plus;&plus;, a highly efficientextension of the ABF method, that overcomes these drawbacks whilemaintaining all the advantages of ABF. ABF++ robustly parameterizesmeshes of hundreds of thousands and millions of triangles withinminutes. It is based on three main components: (1) a new numericalsolution technique that dramatically reduces the dimension of thelinear systems solved at each iteration, speeding up the solution;(2) a new robust scheme for reconstructing the 2D coordinates fromthe angle space solution that avoids the numerical instabilitieswhich hindered the ABF reconstruction scheme; and (3) an efficienthierarchical solution technique. The speedup with (1) does not comeat the expense of greater distortion. The hierarchical technique(3) enables parameterization of models with millions of faces inseconds at the expense of a minor increase in parametricdistortion. The parameterization computed by ABF++ are provablyvalid, that is they contain no flipped triangles. As a result ofthese extensions, the ABF++ method is extremely suitable forrobustly and efficiently parameterizing models forgeometry-processing applications."
                },
                {
                    "title": "Meshing Point Clouds Using Spherical Parameterization",
                    "abstract": "We present a simple method for meshing a 3D point cloud to a manifold genus-0 mesh. Our approach is based on recent methods for spherical embedding of planar graphs, where we use instead a k-nearest neighborhood graph of the point cloud. Our approach proceeds in two steps: We first embed the neighborhood graph on a sphere using an iterative procedure, minimizing the tangential Laplacian. Then we triangulate the embedded points and apply the resulting mesh connectivity to the input points. Besides meshing, spherical embedding of point clouds may also be used for other applications such as texture mapping or morphing."
                },
                {
                    "title": "Smooth Geometry Images",
                    "abstract": "Previous parametric representations of smooth genus-zero surfaces require a collection of abutting patches (e.g. plines, NURBS, recursively subdivided polygons). We introduce a simple construction for these surfaces using a single uniform bi-cubic B-spline. Due to its tensor-product structure, the spline control points are conveniently stored as a geometry image with simple boundary symmetries. The bicubic surface is evaluated using subdivision, and the regular structure of the geometry image makes this computation ideally suited for graphics hardware. Specifically, we let the fragment shader pipeline perform subdivision by applying a sequence of masks (splitting, averaging, limit, and tangent) uniformly to the geometry image. We then extend this scheme to provide smooth level-of-detail transitions from a subsampled base octahedron all the way to a finely subdivided, smooth model. Finally, we show how the framework easily supports scalar displacement mapping."
                },
                {
                    "title": "Seamster: inconspicuous low-distortion texture seam layout",
                    "abstract": "Surface texturing aids the visualization of polygonal meshes by providing additional surface orientation cues and feature annotations. Such texturing is usually implemented via texture mapping, which is easier and more effective when the distortion of the mapping from the surface to the texture map is kept small. We have previously shown that distortion occurs when areas of high surface curvature are flattened into the texture map. By cutting the surface in these areas one can reduce texture map distortion at the expense of additional seam artifacts. This paper describes a faster technique for guiding a texture map seam through high distortion regions, while restricting the seam to regions of low visibility. This results in distortion reducing seams that are less visually distracting and take less time to compute. We have also observed that visibility considerations improve the speed of a recent method that adds cuts to reduce a surface genus."
                },
                {
                    "title": "Least squares conformal maps for automatic texture atlas generation",
                    "abstract": "A Texture Atlas is an efficient color representation for 3D Paint Systems. The model to be textured is decomposed into charts homeomorphic to discs, each chart is parameterized, and the unfolded charts are packed in texture space. Existing texture atlas methods for triangulated surfaces suffer from several limitations, requiring them to generate a large number of small charts with simple borders. The discontinuities between the charts cause artifacts, and make it difficult to paint large areas with regular patterns.In this paper, our main contribution is a new quasi-conformal parameterization method, based on a least-squares approximation of the Cauchy-Riemann equations. The so-defined objective function minimizes angle deformations, and we prove the following properties: the minimum is unique, independent of a similarity in texture space, independent of the resolution of the mesh and cannot generate triangle flips. The function is numerically well behaved and can therefore be very efficiently minimized. Our approach is robust, and can parameterize large charts with complex borders.We also introduce segmentation methods to decompose the model into charts with natural shapes, and a new packing algorithm to gather them in texture space. We demonstrate our approach applied to paint both scanned and modeled data sets."
                },
                {
                    "title": "Geometry images",
                    "abstract": "Surface geometry is often modeled with irregular triangle meshes. The process of remeshing refers to approximating such geometry using a mesh with (semi)-regular connectivity, which has advantages for many graphics applications. However, current techniques for remeshing arbitrary surfaces create only semi-regular meshes. The original mesh is typically decomposed into a set of disk-like charts, onto which the geometry is parametrized and sampled. In this paper, we propose to remesh an arbitrary surface onto a completely regular structure we call a geometry image. It captures geometry as a simple 2D array of quantized points. Surface signals like normals and colors are stored in similar 2D arrays using the same implicit surface parametrization --- texture coordinates are absent. To create a geometry image, we cut an arbitrary mesh along a network of edge paths, and parametrize the resulting single chart onto a square. Geometry images can be encoded using traditional image compression algorithms, such as wavelet-based coders."
                },
                {
                    "title": "Optimally Cutting a Surface into a Disk",
                    "abstract": "Abstract\nWe consider the problem of cutting a subset of the edges of a polyhedral manifold \nsurface, possibly with boundary, to obtain a single topological disk, minimizing\neither the total number of cut edges or their total length. We show that this\nproblem is NP-hard in general, even for manifolds without boundary and for punctured\nspheres. We also describe an algorithm with running time nO(g+k), where n is\nthe combinatorial complexity, g is the genus, and k is the number of boundary\ncomponents of the input surface. Finally, we describe a greedy algorithm that\noutputs a O(log2g)-approximation of the minimum cut graph in O(g2n log n)\ntime.\n\n"
                },
                {
                    "title": "Multiresolution analysis of arbitrary meshes",
                    "abstract": "In computer graphics and geometric modeling, shapes are often represented by triangular meshes. With the advent of laser scanning systems, meshes of extreme complexity are rapidly becoming commonplace. Such meshes are notoriously expensive to store, transmit, render, and are awkward to edit. Multiresolution analysis offers a simple, unified, and theoretically sound approach to dealing with these problems. Lounsbery et al. have recently developed a technique for creating multiresolution representations for a restricted class of meshes with subdivision connectivity. Unfortunately, meshes encountered in practice typically do not meet this requirement. In this paper we present a method for overcoming the subdivision connectivity restriction, meaning that completely arbitrary meshes can now be converted to multiresolution form. The method is based on the approximation of an arbitrary initial mesh M by a mesh MJ that has subdivision connectivity and is guaranteed to be within a specified tolerance. The key ingredient of our algorithm is the construction of a parametrization of M over a simple domain. We expect this parametrization to be of use in other contexts, such as texture mapping or the approximation of complex meshes by NURBS patches. CR"
                },
                {
                    "title": "ON HOW TO DRAW A GRAPH",
                    "abstract": "We give an overview of Tutte\u2019s paper, \u201cHow to draw a graph\u201d, that contains: (i) a proof that every simple 3-connected planar graph admits a straight-line embedding in the plane such that each face boundary is a convex polygon, (ii) an elegant algorithm for finding such an embedding, (iii) an algorithm for testing planarity, and (iv) a proof of Kuratowski\u2019s theorem."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.CG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop an automated and reliable method for UV unwrapping of complex 3D surfaces that accommodates user-generated content with irregular geometries and minimizes manual intervention in seam selection?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing demand for efficient and effective 3D modeling tools that can handle diverse and complex geometries, particularly in the context of user-generated content. By automating the UV unwrapping process, we can enhance the reliability and efficiency of 3D graphics rendering, which is essential for applications in gaming, virtual reality, and computer-aided design. This advancement could lead to new methodologies in geometry processing, enabling researchers to explore more sophisticated algorithms and applications that leverage complex 3D data.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of 3D surfaces, particularly those with high genus or irregular geometries. Naive approaches may fail due to their inability to effectively handle the intricacies of cutting seam selection and the mutual influence between mesh cutting and flattening processes. Additionally, existing algorithms are often tailored for well-behaved meshes, making them inadequate for the diverse range of user-generated content. Overcoming these technical obstacles requires innovative solutions that can dynamically adapt to varying surface characteristics while ensuring continuity and minimizing distortion.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on well-defined meshes and manual seam selection, which limits the applicability of existing solutions to more complex geometries. The reliance on heuristic methods for seam generation has resulted in subjective and inconsistent outcomes, preventing the development of a fully automated approach. Additionally, earlier neural parameterization methods lacked the necessary mapping constraints to be considered true surface parameterization techniques. Our approach aims to bridge these gaps by integrating automated seam selection with advanced neural network architectures that can learn effective parameterizations for a wider range of 3D surfaces.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a neural network-based framework that automates the UV unwrapping process by jointly optimizing seam selection and surface flattening. We will utilize a diverse dataset of 3D models, including both well-behaved and user-generated content, to train our model. The performance will be evaluated using metrics such as distortion, continuity,"
            }
        },
        "author_data": {
            "a7dda78a-6556-4ffb-8f72-c401cc352f0a": {
                "pk": "a7dda78a-6556-4ffb-8f72-c401cc352f0a",
                "name": "Qijian Zhang",
                "collaborators": [
                    "Junhui Hou",
                    "Ying He",
                    "Yue Qian",
                    "Yiming Zeng",
                    "Yixuan Yuan",
                    "Juyong Zhang",
                    "Runmin Cong",
                    "Chongyi Li",
                    "Yao Zhao",
                    "Yingzhi Tang"
                ],
                "domain": [
                    "3D Point Cloud",
                    "Self-Supervised Learning",
                    "Deep Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "PointVST: Self-Supervised Pre-training for 3D Point Clouds via View-Specific Point-to-Image Translation",
                        "abstract": "The past few years have witnessed the great success and prevalence of self-supervised representation learning within the language and 2D vision communities. However, such advancements have not been fully migrated to the field of 3D point cloud learning. Different from existing pre-training paradigms designed for deep point cloud feature extractors that fall into the scope of generative modeling or contrastive learning, this paper proposes a translative pre-training framework, namely PointVST, driven by a novel self-supervised pretext task of cross-modal translation from 3D point clouds to their corresponding diverse forms of 2D rendered images. More specifically, we begin with deducing view-conditioned point-wise embeddings through the insertion of the viewpoint indicator, and then adaptively aggregate a view-specific global codeword, which can be further fed into subsequent 2D convolutional translation heads for image generation. Extensive experimental evaluations on various downstream task scenarios demonstrate that our PointVST shows consistent and prominent performance superiority over current state-of-the-art approaches as well as satisfactory domain transfer capability. Our code will be publicly available at https://github.com/keeganhk/PointVST."
                    },
                    {
                        "title": "ParaPoint: Learning Global Free-Boundary Surface Parameterization of 3D Point Clouds",
                        "abstract": "Surface parameterization is a fundamental geometry processing problem with rich downstream applications. Traditional approaches are designed to operate on well-behaved mesh models with high-quality triangulations that are laboriously produced by specialized 3D modelers, and thus unable to meet the processing demand for the current explosion of ordinary 3D data. In this paper, we seek to perform UV unwrapping on unstructured 3D point clouds. Technically, we propose ParaPoint, an unsupervised neural learning pipeline for achieving global free-boundary surface parameterization by building point-wise mappings between given 3D points and 2D UV coordinates with adaptively deformed boundaries. We ingeniously construct several geometrically meaningful sub-networks with specific functionalities, and assemble them into a bi-directional cycle mapping framework. We also design effective loss functions and auxiliary differential geometric constraints for the optimization of the neural mapping process. To the best of our knowledge, this work makes the first attempt to investigate neural point cloud parameterization that pursues both global mappings and free boundaries. Experiments demonstrate the effectiveness and inspiring potential of our proposed learning paradigm. The code will be publicly available."
                    },
                    {
                        "title": "PointMCD: Boosting Deep Point Cloud Encoders via Multi-view Cross-modal Distillation for 3D Shape Recognition",
                        "abstract": "As two fundamental representation modalities of 3D objects, 3D point clouds and multi-view 2D images record shape information from different domains of geometric structures and visual appearances. In the current deep learning era, remarkable progress in processing such two data modalities has been achieved through respectively customizing compatible 3D and 2D network architectures. However, unlike multi-view image-based 2D visual modeling paradigms, which have shown leading performance in several common 3D shape recognition benchmarks, point cloud-based 3D geometric modeling paradigms are still highly limited by insufficient learning capacity, due to the difficulty of extracting discriminative features from irregular geometric signals. In this paper, we explore the possibility of boosting deep 3D point cloud encoders by transferring visual knowledge extracted from deep 2D image encoders under a standard teacher-student distillation workflow. Generally, we propose PointMCD, a unified multi-view cross-modal distillation architecture, including a pretrained deep image encoder as the teacher and a deep point encoder as the student. To perform heterogeneous feature alignment between 2D visual and 3D geometric domains, we further investigate visibility-aware feature projection (VAFP), by which point-wise embeddings are reasonably aggregated into view-specific geometric descriptors. By pair-wisely aligning multi-view visual and geometric descriptors, we can obtain more powerful deep point encoders without exhausting and complicated network modification. Experiments on 3D shape classification, part segmentation, and unsupervised learning strongly validate the effectiveness of our method. The code and data will be publicly available at https://github.com/keeganhk/PointMCD."
                    },
                    {
                        "title": "ParaNet: Deep Regular Representation for 3D Point Clouds",
                        "abstract": "Although convolutional neural networks have achieved remarkable success in analyzing 2D images/videos, it is still non-trivial to apply the well-developed 2D techniques in regular domains to the irregular 3D point cloud data. To bridge this gap, we propose ParaNet, a novel end-to-end deep learning framework, for representing 3D point clouds in a completely regular and nearly lossless manner. To be specific, ParaNet converts an irregular 3D point cloud into a regular 2D color image, named point geometry image (PGI), where each pixel encodes the spatial coordinates of a point. In contrast to conventional regular representation modalities based on multi-view projection and voxelization, the proposed representation is differentiable and reversible. Technically, ParaNet is composed of a surface embedding module, which parameterizes 3D surface points onto a unit square, and a grid resampling module, which resamples the embedded 2D manifold over regular dense grids. Note that ParaNet is unsupervised, i.e., the training simply relies on reference-free geometry constraints. The PGIs can be seamlessly coupled with a task network established upon standard and mature techniques for 2D images/videos to realize a specific task for 3D point clouds. We evaluate ParaNet over shape classification and point cloud upsampling, in which our solutions perform favorably against the existing state-of-the-art methods. We believe such a paradigm will open up many possibilities to advance the progress of deep learning-based point cloud processing and understanding."
                    },
                    {
                        "title": "CoADNet: Collaborative Aggregation-and-Distribution Networks for Co-Salient Object Detection",
                        "abstract": "Co-Salient Object Detection (CoSOD) aims at discovering salient objects that repeatedly appear in a given query group containing two or more relevant images. One challenging issue is how to effectively capture co-saliency cues by modeling and exploiting inter-image relationships. In this paper, we present an end-to-end collaborative aggregation-and-distribution network (CoADNet) to capture both salient and repetitive visual patterns from multiple images. First, we integrate saliency priors into the backbone features to suppress the redundant background information through an online intra-saliency guidance structure. After that, we design a two-stage aggregate-and-distribute architecture to explore group-wise semantic interactions and produce the co-saliency features. In the first stage, we propose a group-attentional semantic aggregation module that models inter-image relationships to generate the group-wise semantic representations. In the second stage, we propose a gated group distribution module that adaptively distributes the learned group semantics to different individuals in a dynamic gating mechanism. Finally, we develop a group consistency preserving decoder tailored for the CoSOD task, which maintains group constraints during feature decoding to predict more consistent full-resolution co-saliency maps. The proposed CoADNet is evaluated on four prevailing CoSOD benchmark datasets, which demonstrates the remarkable performance improvement over ten state-of-the-art competitors."
                    },
                    {
                        "title": "Human as Points: Explicit Point-based 3D Human Reconstruction from Single-view RGB Images",
                        "abstract": "The latest trends in the research field of single-view human reconstruction devote to learning deep implicit functions constrained by explicit body shape priors. Despite the remarkable performance improvements compared with traditional processing pipelines, existing learning approaches still show different aspects of limitations in terms of flexibility, generalizability, robustness, and/or representation capability. To comprehensively address the above issues, in this paper, we investigate an explicit point-based human reconstruction framework called HaP, which adopts point clouds as the intermediate representation of the target geometric structure. Technically, our approach is featured by fully-explicit point cloud estimation, manipulation, generation, and refinement in the 3D geometric space, instead of an implicit learning process that can be ambiguous and less controllable. The overall workflow is carefully organized with dedicated designs of the corresponding specialized learning components as well as processing procedures. Extensive experiments demonstrate that our framework achieves quantitative performance improvements of 20% to 40% over current state-of-the-art methods, and better qualitative results. Our promising results may indicate a paradigm rollback to the fully-explicit and geometry-centric algorithm design, which enables to exploit various powerful point cloud modeling architectures and processing techniques. We will make our code and data publicly available at https://github.com/yztang4/HaP."
                    },
                    {
                        "title": "Dynamic 3D Point Cloud Sequences as 2D Videos",
                        "abstract": "Dynamic 3D point cloud sequences serve as one of the most common and practical representation modalities of dynamic real-world environments. However, their unstructured nature in both spatial and temporal domains poses significant challenges to effective and efficient processing. Existing deep point cloud sequence modeling approaches imitate the mature 2D video learning mechanisms by developing complex spatio-temporal point neighbor grouping and feature aggregation schemes, often resulting in methods lacking effectiveness, efficiency, and expressive power. In this paper, we propose a novel generic representation called \\textit{Structured Point Cloud Videos} (SPCVs). Intuitively, by leveraging the fact that 3D geometric shapes are essentially 2D manifolds, SPCV re-organizes a point cloud sequence as a 2D video with spatial smoothness and temporal consistency, where the pixel values correspond to the 3D coordinates of points. The structured nature of our SPCV representation allows for the seamless adaptation of well-established 2D image/video techniques, enabling efficient and effective processing and analysis of 3D point cloud sequences. To achieve such re-organization, we design a self-supervised learning pipeline that is geometrically regularized and driven by self-reconstructive and deformation field learning objectives. Additionally, we construct SPCV-based frameworks for both low-level and high-level 3D point cloud sequence processing and analysis tasks, including action recognition, temporal interpolation, and compression. Extensive experiments demonstrate the versatility and superiority of the proposed SPCV, which has the potential to offer new possibilities for deep learning on unstructured 3D point cloud sequences. Code will be released at https://github.com/ZENGYIMING-EAMON/SPCV."
                    },
                    {
                        "title": "Leveraging Single-View Images for Unsupervised 3D Point Cloud Completion",
                        "abstract": "Point clouds captured by scanning devices are often incomplete due to occlusion. To overcome this limitation, point cloud completion methods have been developed to predict the complete shape of an object based on its partial input. These methods can be broadly classified as supervised or unsupervised. However, both categories require a large number of 3D complete point clouds, which may be difficult to capture. In this paper, we propose Cross-PCC, an unsupervised point cloud completion method without requiring any 3D complete point clouds. We only utilize 2D images of the complete objects, which are easier to capture than 3D complete and clean point clouds. Specifically, to take advantage of the complementary information from 2D images, we use a single-view RGB image to extract 2D features and design a fusion module to fuse the 2D and 3D features extracted from the partial point cloud. To guide the shape of predicted point clouds, we project the predicted points of the object to the 2D plane and use the foreground pixels of its silhouette maps to constrain the position of the projected points. To reduce the outliers of the predicted point clouds, we propose a view calibrator to move the points projected to the background into the foreground by the single-view silhouette image. To the best of our knowledge, our approach is the first point cloud completion method that does not require any 3D supervision. The experimental results of our method are superior to those of the state-of-the-art unsupervised methods by a large margin. Moreover, our method even achieves comparable performance to some supervised methods. We will make the source code publicly available at https://github.com/ltwu6/cross-pcc."
                    },
                    {
                        "title": "Flattening-Net: Deep Regular 2D Representation for 3D Point Cloud Analysis",
                        "abstract": "Point clouds are characterized by irregularity and unstructuredness, which pose challenges in efficient data exploitation and discriminative feature extraction. In this paper, we present an unsupervised deep neural architecture called Flattening-Net to represent irregular 3D point clouds of arbitrary geometry and topology as a completely regular 2D point geometry image (PGI) structure, in which coordinates of spatial points are captured in colors of image pixels. \\mr{Intuitively, Flattening-Net implicitly approximates a locally smooth 3D-to-2D surface flattening process while effectively preserving neighborhood consistency.} \\mr{As a generic representation modality, PGI inherently encodes the intrinsic property of the underlying manifold structure and facilitates surface-style point feature aggregation.} To demonstrate its potential, we construct a unified learning framework directly operating on PGIs to achieve \\mr{diverse types of high-level and low-level} downstream applications driven by specific task networks, including classification, segmentation, reconstruction, and upsampling. Extensive experiments demonstrate that our methods perform favorably against the current state-of-the-art competitors. We will make the code and data publicly available at https://github.com/keeganhk/Flattening-Net."
                    },
                    {
                        "title": "Unleash the Potential of Image Branch for Cross-modal 3D Object Detection",
                        "abstract": "To achieve reliable and precise scene understanding, autonomous vehicles typically incorporate multiple sensing modalities to capitalize on their complementary attributes. However, existing cross-modal 3D detectors do not fully utilize the image domain information to address the bottleneck issues of the LiDAR-based detectors. This paper presents a new cross-modal 3D object detector, namely UPIDet, which aims to unleash the potential of the image branch from two aspects. First, UPIDet introduces a new 2D auxiliary task called normalized local coordinate map estimation. This approach enables the learning of local spatial-aware features from the image modality to supplement sparse point clouds. Second, we discover that the representational capability of the point cloud backbone can be enhanced through the gradients backpropagated from the training objectives of the image branch, utilizing a succinct and effective point-to-pixel module. Extensive experiments and ablation studies validate the effectiveness of our method. Notably, we achieved the top rank in the highly competitive cyclist class of the KITTI benchmark at the time of submission. The source code is available at https://github.com/Eaphan/UPIDet."
                    },
                    {
                        "title": "GLENet: Boosting 3D Object Detectors with Generative Label Uncertainty Estimation",
                        "abstract": "The inherent ambiguity in ground-truth annotations of 3D bounding boxes, caused by occlusions, signal missing, or manual annotation errors, can confuse deep 3D object detectors during training, thus deteriorating detection accuracy. However, existing methods overlook such issues to some extent and treat the labels as deterministic. In this paper, we formulate the label uncertainty problem as the diversity of potentially plausible bounding boxes of objects. Then, we propose GLENet, a generative framework adapted from conditional variational autoencoders, to model the one-to-many relationship between a typical 3D object and its potential ground-truth bounding boxes with latent variables. The label uncertainty generated by GLENet is a plug-and-play module and can be conveniently integrated into existing deep 3D detectors to build probabilistic detectors and supervise the learning of the localization uncertainty. Besides, we propose an uncertainty-aware quality estimator architecture in probabilistic detectors to guide the training of the IoU-branch with predicted localization uncertainty. We incorporate the proposed methods into various popular base 3D detectors and demonstrate significant and consistent performance gains on both KITTI and Waymo benchmark datasets. Especially, the proposed GLENet-VR outperforms all published LiDAR-based approaches by a large margin and achieves the top rank among single-modal methods on the challenging KITTI test set. The source code and pre-trained models are publicly available at \\url{https://github.com/Eaphan/GLENet}."
                    },
                    {
                        "title": "NeuroGF: A Neural Representation for Fast Geodesic Distance and Path Queries",
                        "abstract": "Geodesics are essential in many geometry processing applications. However, traditional algorithms for computing geodesic distances and paths on 3D mesh models are often inefficient and slow. This makes them impractical for scenarios that require extensive querying of arbitrary point-to-point geodesics. Although neural implicit representations have emerged as a popular way of representing 3D shape geometries, there is still no research on representing geodesics with deep implicit functions. To bridge this gap, this paper presents the first attempt to represent geodesics on 3D mesh models using neural implicit functions. Specifically, we introduce neural geodesic fields (NeuroGFs), which are learned to represent the all-pairs geodesics of a given mesh. By using NeuroGFs, we can efficiently and accurately answer queries of arbitrary point-to-point geodesic distances and paths, overcoming the limitations of traditional algorithms. Evaluations on common 3D models show that NeuroGFs exhibit exceptional performance in solving the single-source all-destination (SSAD) and point-to-point geodesics, and achieve high accuracy consistently. Besides, NeuroGFs also offer the unique advantage of encoding both 3D geometry and geodesics in a unified representation. Moreover, we further extend generalizable learning frameworks of NeuroGFs by adding shape feature encoders, which also show satisfactory performances for unseen shapes and categories. Code is made available at https://github.com/keeganhk/NeuroGF/tree/master."
                    },
                    {
                        "title": "IDEA-Net: Dynamic 3D Point Cloud Interpolation via Deep Embedding Alignment",
                        "abstract": "This paper investigates the problem of temporally interpolating dynamic 3D point clouds with large non-rigid deformation. We formulate the problem as estimation of point-wise trajectories (i.e., smooth curves) and further reason that temporal irregularity and under-sampling are two major challenges. To tackle the challenges, we propose IDEA-Net, an end-to-end deep learning framework, which disentangles the problem under the assistance of the explicitly learned temporal consistency. Specifically, we propose a temporal consistency learning module to align two consecutive point cloud frames point-wisely, based on which we can employ linear interpolation to obtain coarse trajectories/in-between frames. To compensate the high-order nonlinear components of trajectories, we apply aligned feature embeddings that encode local geometry properties to regress point-wise increments, which are combined with the coarse estimations. We demonstrate the effectiveness of our method on various point cloud sequences and observe large improvement over state-of-the-art methods both quantitatively and visually. Our framework can bring benefits to 3D motion data acquisition. The source code is publicly available at https://github.com/ZENGYIMING-EAMON/IDEA-Net.git."
                    },
                    {
                        "title": "WarpingGAN: Warping Multiple Uniform Priors for Adversarial 3D Point Cloud Generation",
                        "abstract": "We propose WarpingGAN, an effective and efficient 3D point cloud generation network. Unlike existing methods that generate point clouds by directly learning the mapping functions between latent codes and 3D shapes, Warping-GAN learns a unified local-warping function to warp multiple identical pre-defined priors (i.e., sets of points uniformly distributed on regular 3D grids) into 3D shapes driven by local structure-aware semantics. In addition, we also ingeniously utilize the principle of the discriminator and tailor a stitching loss to eliminate the gaps between different partitions of a generated shape corresponding to different priors for boosting quality. Owing to the novel generating mechanism, WarpingGAN, a single lightweight network after one-time training, is capable of efficiently generating uniformly distributed 3D point clouds with various resolutions. Extensive experimental results demonstrate the superiority of our WarpingGAN over state-of-the-art methods in terms of quantitative metrics, visual quality, and efficiency. The source code is publicly available at https://github.com/yztang4/WarpingGAN.git."
                    },
                    {
                        "title": "MOPS-Net: A Matrix Optimization-driven Network forTask-Oriented 3D Point Cloud Downsampling",
                        "abstract": "This paper explores the problem of task-oriented downsampling over 3D point clouds, which aims to downsample a point cloud while maintaining the performance of subsequent applications applied to the downsampled sparse points as much as possible. Designing from the perspective of matrix optimization, we propose MOPS-Net, a novel interpretable deep learning-based method, which is fundamentally different from the existing deep learning-based methods due to its interpretable feature. The optimization problem is challenging due to its discrete and combinatorial nature. We tackle the challenges by relaxing the binary constraint of the variables, and formulate a constrained and differentiable matrix optimization problem. We then design a deep neural network to mimic the matrix optimization by exploring both the local and global structures of the input data. MOPS-Net can be end-to-end trained with a task network and is permutation-invariant, making it robust to the input. We also extend MOPS-Net such that a single network after one-time training is capable of handling arbitrary downsampling ratios. Extensive experimental results show that MOPS-Net can achieve favorable performance against state-of-the-art deep learning-based methods over various tasks, including classification, reconstruction, and registration. Besides, we validate the robustness of MOPS-Net on noisy data."
                    },
                    {
                        "title": "Dense Attention Fluid Network for Salient Object Detection in Optical Remote Sensing Images",
                        "abstract": "Despite the remarkable advances in visual saliency analysis for natural scene images (NSIs), salient object detection (SOD) for optical remote sensing images (RSIs) still remains an open and challenging problem. In this paper, we propose an end-to-end Dense Attention Fluid Network (DAFNet) for SOD in optical RSIs. A Global Context-aware Attention (GCA) module is proposed to adaptively capture long-range semantic context relationships, and is further embedded in a Dense Attention Fluid (DAF) structure that enables shallow attention cues flow into deep layers to guide the generation of high-level feature attention maps. Specifically, the GCA module is composed of two key components, where the global feature aggregation module achieves mutual reinforcement of salient feature embeddings from any two spatial locations, and the cascaded pyramid attention module tackles the scale variation issue by building up a cascaded pyramid framework to progressively refine the attention map in a coarse-to-fine manner. In addition, we construct a new and challenging optical RSI dataset for SOD that contains 2,000 images with pixel-wise saliency annotations, which is currently the largest publicly available benchmark. Extensive experiments demonstrate that our proposed DAFNet significantly outperforms the existing state-of-the-art SOD competitors. https://github.com/rmcong/DAFNet_TIP20"
                    }
                ]
            },
            "d4c7994b-ce0f-4a42-8773-7b31855eb4bb": {
                "pk": "d4c7994b-ce0f-4a42-8773-7b31855eb4bb",
                "name": "Junhui Hou",
                "collaborators": [
                    "Hui Liu",
                    "Yuheng Jia",
                    "Xianqiang Lyu",
                    "Qijian Zhang",
                    "Jing Jin",
                    "Jie Chen",
                    "Lap-Pui Chau",
                    "Siyu Ren",
                    "Yifan Zhang",
                    "Meng You"
                ],
                "domain": [
                    "3D Point Cloud",
                    "Self-Supervised Learning",
                    "Light Field",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "PointVST: Self-Supervised Pre-training for 3D Point Clouds via View-Specific Point-to-Image Translation",
                        "abstract": "The past few years have witnessed the great success and prevalence of self-supervised representation learning within the language and 2D vision communities. However, such advancements have not been fully migrated to the field of 3D point cloud learning. Different from existing pre-training paradigms designed for deep point cloud feature extractors that fall into the scope of generative modeling or contrastive learning, this paper proposes a translative pre-training framework, namely PointVST, driven by a novel self-supervised pretext task of cross-modal translation from 3D point clouds to their corresponding diverse forms of 2D rendered images. More specifically, we begin with deducing view-conditioned point-wise embeddings through the insertion of the viewpoint indicator, and then adaptively aggregate a view-specific global codeword, which can be further fed into subsequent 2D convolutional translation heads for image generation. Extensive experimental evaluations on various downstream task scenarios demonstrate that our PointVST shows consistent and prominent performance superiority over current state-of-the-art approaches as well as satisfactory domain transfer capability. Our code will be publicly available at https://github.com/keeganhk/PointVST."
                    },
                    {
                        "title": "Unleash the Potential of 3D Point Cloud Modeling with A Calibrated Local Geometry-driven Distance Metric",
                        "abstract": "Quantifying the dissimilarity between two unstructured 3D point clouds is a challenging task, with existing metrics often relying on measuring the distance between corresponding points that can be either inefficient or ineffective. In this paper, we propose a novel distance metric called Calibrated Local Geometry Distance (CLGD), which computes the difference between the underlying 3D surfaces calibrated and induced by a set of reference points. By associating each reference point with two given point clouds through computing its directional distances to them, the difference in directional distances of an identical reference point characterizes the geometric difference between a typical local region of the two point clouds. Finally, CLGD is obtained by averaging the directional distance differences of all reference points. We evaluate CLGD on various optimization and unsupervised learning-based tasks, including shape reconstruction, rigid registration, scene flow estimation, and feature representation. Extensive experiments show that CLGD achieves significantly higher accuracy under all tasks in a memory and computationally efficient manner, compared with existing metrics. As a generic metric, CLGD has the potential to advance 3D point cloud modeling. The source code is publicly available at https://github.com/rsy6318/CLGD."
                    },
                    {
                        "title": "Probabilistic-based Feature Embedding of 4-D Light Fields for Compressive Imaging and Denoising",
                        "abstract": "The high-dimensional nature of the 4-D light field (LF) poses great challenges in achieving efficient and effective feature embedding, that severely impacts the performance of downstream tasks. To tackle this crucial issue, in contrast to existing methods with empirically-designed architectures, we propose a probabilistic-based feature embedding (PFE), which learns a feature embedding architecture by assembling various low-dimensional convolution patterns in a probability space for fully capturing spatial-angular information. Building upon the proposed PFE, we then leverage the intrinsic linear imaging model of the coded aperture camera to construct a cycle-consistent 4-D LF reconstruction network from coded measurements. Moreover, we incorporate PFE into an iterative optimization framework for 4-D LF denoising. Our extensive experiments demonstrate the significant superiority of our methods on both real-world and synthetic 4-D LF images, both quantitatively and qualitatively, when compared with state-of-the-art methods. The source code will be publicly available at https://github.com/lyuxianqiang/LFCA-CR-NET."
                    },
                    {
                        "title": "Enhancing Low-light Light Field Images with A Deep Compensation Unfolding Network",
                        "abstract": "This paper presents a novel and interpretable end-to-end learning framework, called the deep compensation unfolding network (DCUNet), for restoring light field (LF) images captured under low-light conditions. DCUNet is designed with a multi-stage architecture that mimics the optimization process of solving an inverse imaging problem in a data-driven fashion. The framework uses the intermediate enhanced result to estimate the illumination map, which is then employed in the unfolding process to produce a new enhanced result. Additionally, DCUNet includes a content-associated deep compensation module at each optimization stage to suppress noise and illumination map estimation errors. To properly mine and leverage the unique characteristics of LF images, this paper proposes a pseudo-explicit feature interaction module that comprehensively exploits redundant information in LF images. The experimental results on both simulated and real datasets demonstrate the superiority of our DCUNet over state-of-the-art methods, both qualitatively and quantitatively. Moreover, DCUNet preserves the essential geometric structure of enhanced LF images much better. The code will be publicly available at https://github.com/lyuxianqiang/LFLL-DCU."
                    },
                    {
                        "title": "Fine-grained Image-to-LiDAR Contrastive Distillation with Visual Foundation Models",
                        "abstract": "Contrastive image-to-LiDAR knowledge transfer, commonly used for learning 3D representations with synchronized images and point clouds, often faces a self-conflict dilemma. This issue arises as contrastive losses unintentionally dissociate features of unmatched points and pixels that share semantic labels, compromising the integrity of learned representations. To overcome this, we harness Visual Foundation Models (VFMs), which have revolutionized the acquisition of pixel-level semantics, to enhance 3D representation learning. Specifically, we utilize off-the-shelf VFMs to generate semantic labels for weakly-supervised pixel-to-point contrastive distillation. Additionally, we employ von Mises-Fisher distributions to structure the feature space, ensuring semantic embeddings within the same class remain consistent across varying inputs. Furthermore, we adapt sampling probabilities of points to address imbalances in spatial distribution and category frequency, promoting comprehensive and balanced learning. Extensive experiments demonstrate that our approach mitigates the challenges posed by traditional methods and consistently surpasses existing image-to-LiDAR contrastive distillation methods in downstream tasks. The source code is available at \\href{https://github.com/Eaphan/OLIVINE.}{\\color{black}https://github.com/Eaphan/OLIVINE}."
                    },
                    {
                        "title": "Occlusion-aware Unsupervised Learning of Depth from 4-D Light Fields",
                        "abstract": "Depth estimation is a fundamental issue in 4-D light field processing and analysis. Although recent supervised learning-based light field depth estimation methods have significantly improved the accuracy and efficiency of traditional optimization-based ones, these methods rely on the training over light field data with ground-truth depth maps which are challenging to obtain or even unavailable for real-world light field data. Besides, due to the inevitable gap (or domain difference) between real-world and synthetic data, they may suffer from serious performance degradation when generalizing the models trained with synthetic data to real-world data. By contrast, we propose an unsupervised learning-based method, which does not require ground-truth depth as supervision during training. Specifically, based on the basic knowledge of the unique geometry structure of light field data, we present an occlusion-aware strategy to improve the accuracy on occlusion areas, in which we explore the angular coherence among subsets of the light field views to estimate initial depth maps, and utilize a constrained unsupervised loss to learn their corresponding reliability for final depth prediction. Additionally, we adopt a multi-scale network with a weighted smoothness loss to handle the textureless areas. Experimental results on synthetic data show that our method can significantly shrink the performance gap between the previous unsupervised method and supervised ones, and produce depth maps with comparable accuracy to traditional methods with obviously reduced computational cost. Moreover, experiments on real-world datasets show that our method can avoid the domain shift problem presented in supervised methods, demonstrating the great potential of our method."
                    },
                    {
                        "title": "Decoupling Dynamic Monocular Videos for Dynamic View Synthesis",
                        "abstract": "The challenge of dynamic view synthesis from dynamic monocular videos, i.e., synthesizing novel views for free viewpoints given a monocular video of a dynamic scene captured by a moving camera, mainly lies in accurately modeling the \\textbf{dynamic objects} of a scene using limited 2D frames, each with a varying timestamp and viewpoint. Existing methods usually require pre-processed 2D optical flow and depth maps by off-the-shelf methods to supervise the network, making them suffer from the inaccuracy of the pre-processed supervision and the ambiguity when lifting the 2D information to 3D. In this paper, we tackle this challenge in an unsupervised fashion. Specifically, we decouple the motion of the dynamic objects into object motion and camera motion, respectively regularized by proposed unsupervised surface consistency and patch-based multi-view constraints. The former enforces the 3D geometric surfaces of moving objects to be consistent over time, while the latter regularizes their appearances to be consistent across different viewpoints. Such a fine-grained motion formulation can alleviate the learning difficulty for the network, thus enabling it to produce not only novel views with higher quality but also more accurate scene flows and depth than existing methods requiring extra supervision."
                    },
                    {
                        "title": "Hybrid Pooling and Convolutional Network for Improving Accuracy and Training Convergence Speed in Object Detection",
                        "abstract": "This paper introduces HPC-Net, a high-precision and rapidly convergent object detection network."
                    },
                    {
                        "title": "Rank-Enhanced Low-Dimensional Convolution Set for Hyperspectral Image Denoising",
                        "abstract": "This paper tackles the challenging problem of hyperspectral (HS) image denoising. Unlike existing deep learning-based methods usually adopting complicated network architectures or empirically stacking off-the-shelf modules to pursue performance improvement, we focus on the efficient and effective feature extraction manner for capturing the high-dimensional characteristics of HS images. To be specific, based on the theoretical analysis that increasing the rank of the matrix formed by the unfolded convolutional kernels can promote feature diversity, we propose rank-enhanced low-dimensional convolution set (Re-ConvSet), which separately performs 1-D convolution along the three dimensions of an HS image side-by-side, and then aggregates the resulting spatial-spectral embeddings via a learnable compression layer. Re-ConvSet not only learns the diverse spatial-spectral features of HS images, but also reduces the parameters and complexity of the network. We then incorporate Re-ConvSet into the widely-used U-Net architecture to construct an HS image denoising method. Surprisingly, we observe such a concise framework outperforms the most recent method to a large extent in terms of quantitative metrics, visual results, and efficiency. We believe our work may shed light on deep learning-based HS image processing and analysis."
                    },
                    {
                        "title": "Light Field Denoising via Anisotropic Parallax Analysis in a CNN Framework",
                        "abstract": "Light field (LF) cameras provide perspective information of scenes by taking directional measurements of the focusing light rays. The raw outputs are usually dark with additive camera noise, which impedes subsequent processing and applications. We propose a novel LF denoising framework based on anisotropic parallax analysis (APA). Two convolutional neural networks are jointly designed for the task: first, the structural parallax synthesis network predicts the parallax details for the entire LF based on a set of anisotropic parallax features. These novel features can efficiently capture the high frequency perspective components of a LF from noisy observations. Second, the view-dependent detail compensation network restores non-Lambertian variation to each LF view by involving view-specific spatial energies. Extensive experiments show that the proposed APA LF denoiser provides a much better denoising performance than state-of-the-art methods in terms of visual quality and in preservation of parallax details."
                    },
                    {
                        "title": "Stratified Labeling for Surface Consistent Parallax Correction and Occlusion Completion",
                        "abstract": "The light field faithfully records the spatial and angular configurations of the scene, which facilitates a wide range of imaging possibilities. In this work, we propose an LF synthesis algorithm which renders high quality novel LF views far outside the range of angular baselines of the given references. A stratified synthesis strategy is adopted which parses the scene content based on stratified disparity layers and across a varying range of spatial granularities. Such a stratified methodology proves to help preserve scene structures over large perspective shifts, and it provides informative clues for inferring the textures of occluded regions. A generative-adversarial network model is further adopted for parallax correction and occlusion completion conditioned on the stratified synthesis features. Experiments show that our proposed model can provide more reliable novel view synthesis quality at large baseline extension ratios. Over 3dB quality improvement has been achieved against state-of-the-art LF view synthesis algorithms."
                    },
                    {
                        "title": "Clustering-aware Graph Construction: A Joint Learning Perspective",
                        "abstract": "Graph-based clustering methods have demonstrated the effectiveness in various applications. Generally, existing graph-based clustering methods first construct a graph to represent the input data and then partition it to generate the clustering result. However, such a stepwise manner may make the constructed graph not fit the requirements for the subsequent decomposition, leading to compromised clustering accuracy. To this end, we propose a joint learning framework, which is able to learn the graph and the clustering result simultaneously, such that the resulting graph is tailored to the clustering task. The proposed model is formulated as a well-defined nonnegative and off-diagonal constrained optimization problem, which is further efficiently solved with convergence theoretically guaranteed. The advantage of the proposed model is demonstrated by comparing with 19 state-of-the-art clustering methods on 10 datasets with 4 clustering metrics."
                    },
                    {
                        "title": "Learning Dynamic Interpolation for Extremely Sparse Light Fields with Wide Baselines",
                        "abstract": "In this paper, we tackle the problem of dense light field (LF) reconstruction from sparsely-sampled ones with wide baselines and propose a learnable model, namely dynamic interpolation, to replace the commonly-used geometry warping operation. Specifically, with the estimated geometric relation between input views, we first construct a lightweight neural network to dynamically learn weights for interpolating neighbouring pixels from input views to synthesize each pixel of novel views independently. In contrast to the fixed and content-independent weights employed in the geometry warping operation, the learned interpolation weights implicitly incorporate the correspondences between the source and novel views and adapt to different image content information. Then, we recover the spatial correlation between the independently synthesized pixels of each novel view by referring to that of input views using a geometry-based spatial refinement module. We also constrain the angular correlation between the novel views through a disparity-oriented LF structure loss. Experimental results on LF datasets with wide baselines show that the reconstructed LFs achieve much higher PSNR/SSIM and preserve the LF parallax structure better than state-of-the-art methods. The source code is publicly available at https://github.com/MantangGuo/DI4SLF."
                    },
                    {
                        "title": "Semi-Supervised Subspace Clustering via Tensor Low-Rank Representation",
                        "abstract": "In this letter, we propose a novel semi-supervised subspace clustering method, which is able to simultaneously augment the initial supervisory information and construct a discriminative affinity matrix. By representing the limited amount of supervisory information as a pairwise constraint matrix, we observe that the ideal affinity matrix for clustering shares the same low-rank structure as the ideal pairwise constraint matrix. Thus, we stack the two matrices into a 3-D tensor, where a global low-rank constraint is imposed to promote the affinity matrix construction and augment the initial pairwise constraints synchronously. Besides, we use the local geometry structure of input samples to complement the global low-rank prior to achieve better affinity matrix learning. The proposed model is formulated as a Laplacian graph regularized convex low-rank tensor representation problem, which is further solved with an alternative iterative algorithm. In addition, we propose to refine the affinity matrix with the augmented pairwise constraints. Comprehensive experimental results on eight commonly-used benchmark datasets demonstrate the superiority of our method over state-of-the-art methods. The code is publicly available at https://github.com/GuanxingLu/Subspace-Clustering."
                    },
                    {
                        "title": "Superpixel Graph Contrastive Clustering with Semantic-Invariant Augmentations for Hyperspectral Images",
                        "abstract": "Hyperspectral images (HSI) clustering is an important but challenging task. The state-of-the-art (SOTA) methods usually rely on superpixels, however, they do not fully utilize the spatial and spectral information in HSI 3-D structure, and their optimization targets are not clustering-oriented. In this work, we first use 3-D and 2-D hybrid convolutional neural networks to extract the high-order spatial and spectral features of HSI through pre-training, and then design a superpixel graph contrastive clustering (SPGCC) model to learn discriminative superpixel representations. Reasonable augmented views are crucial for contrastive clustering, and conventional contrastive learning may hurt the cluster structure since different samples are pushed away in the embedding space even if they belong to the same class. In SPGCC, we design two semantic-invariant data augmentations for HSI superpixels: pixel sampling augmentation and model weight augmentation. Then sample-level alignment and clustering-center-level contrast are performed for better intra-class similarity and inter-class dissimilarity of superpixel embeddings. We perform clustering and network optimization alternatively. Experimental results on several HSI datasets verify the advantages of the proposed method, e.g., on India Pines, our model improves the clustering accuracy from 58.79% to 67.59% compared to the SOTA method."
                    },
                    {
                        "title": "Towards Calibrated Deep Clustering Network",
                        "abstract": "Deep clustering has exhibited remarkable performance; however, the over-confidence problem, i.e., the estimated confidence for a sample belonging to a particular cluster greatly exceeds its actual prediction accuracy, has been overlooked in prior research. To tackle this critical issue, we pioneer the development of a calibrated deep clustering framework. Specifically, we propose a novel dual-head (calibration head and clustering head) deep clustering model that can effectively calibrate the estimated confidence and the actual accuracy. The calibration head adjusts the overconfident predictions of the clustering head, generating prediction confidence that match the model learning status. Then, the clustering head dynamically select reliable high-confidence samples estimated by the calibration head for pseudo-label self-training. Additionally, we introduce an effective network initialization strategy that enhances both training speed and network robustness. The effectiveness of the proposed calibration approach and initialization strategy are both endorsed with solid theoretical guarantees. Extensive experiments demonstrate the proposed calibrated deep clustering model not only surpasses state-of-the-art deep clustering methods by 10 times in terms of expected calibration error but also significantly outperforms them in terms of clustering accuracy."
                    },
                    {
                        "title": "ParaPoint: Learning Global Free-Boundary Surface Parameterization of 3D Point Clouds",
                        "abstract": "Surface parameterization is a fundamental geometry processing problem with rich downstream applications. Traditional approaches are designed to operate on well-behaved mesh models with high-quality triangulations that are laboriously produced by specialized 3D modelers, and thus unable to meet the processing demand for the current explosion of ordinary 3D data. In this paper, we seek to perform UV unwrapping on unstructured 3D point clouds. Technically, we propose ParaPoint, an unsupervised neural learning pipeline for achieving global free-boundary surface parameterization by building point-wise mappings between given 3D points and 2D UV coordinates with adaptively deformed boundaries. We ingeniously construct several geometrically meaningful sub-networks with specific functionalities, and assemble them into a bi-directional cycle mapping framework. We also design effective loss functions and auxiliary differential geometric constraints for the optimization of the neural mapping process. To the best of our knowledge, this work makes the first attempt to investigate neural point cloud parameterization that pursues both global mappings and free boundaries. Experiments demonstrate the effectiveness and inspiring potential of our proposed learning paradigm. The code will be publicly available."
                    },
                    {
                        "title": "RainyScape: Unsupervised Rainy Scene Reconstruction using Decoupled Neural Rendering",
                        "abstract": "We propose RainyScape, an unsupervised framework for reconstructing clean scenes from a collection of multi-view rainy images. RainyScape consists of two main modules: a neural rendering module and a rain-prediction module that incorporates a predictor network and a learnable latent embedding that captures the rain characteristics of the scene. Specifically, based on the spectral bias property of neural networks, we first optimize the neural rendering pipeline to obtain a low-frequency scene representation. Subsequently, we jointly optimize the two modules, driven by the proposed adaptive direction-sensitive gradient-based reconstruction loss, which encourages the network to distinguish between scene details and rain streaks, facilitating the propagation of gradients to the relevant components. Extensive experiments on both the classic neural radiance field and the recently proposed 3D Gaussian splatting demonstrate the superiority of our method in effectively eliminating rain streaks and rendering clean images, achieving state-of-the-art performance. The constructed high-quality dataset and source code will be publicly available."
                    }
                ]
            },
            "d3135b90-c7b6-45f6-9abf-a626ed09cbc6": {
                "pk": "d3135b90-c7b6-45f6-9abf-a626ed09cbc6",
                "name": "Wenping Wang",
                "collaborators": [
                    "Cheng Lin",
                    "Yizhou Yu",
                    "Yuexin Ma",
                    "Dinesh Manocha",
                    "Changjian Li",
                    "Longxiang Zhang",
                    "Ningna Wang",
                    "Bin Wang",
                    "Xiaohu Guo",
                    "Haiping Wang"
                ],
                "domain": [
                    "Computer Vision",
                    "3D Shape Reconstruction",
                    "Machine Learning",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Multi-Modal Machine Learning for Assessing Gaming Skills in Online Streaming: A Case Study with CS:GO",
                        "abstract": "Online streaming is an emerging market that address much attention. Assessing gaming skills from videos is an important task for streaming service providers to discover talented gamers. Service providers require the information to offer customized recommendation and service promotion to their customers. Meanwhile, this is also an important multi-modal machine learning tasks since online streaming combines vision, audio and text modalities. In this study we begin by identifying flaws in the dataset and proceed to clean it manually. Then we propose several variants of latest end-to-end models to learn joint representation of multiple modalities. Through our extensive experimentation, we demonstrate the efficacy of our proposals. Moreover, we identify that our proposed models is prone to identifying users instead of learning meaningful representations. We purpose future work to address the issue in the end."
                    },
                    {
                        "title": "Computing Medial Axis Transform with Feature Preservation via Restricted Power Diagram",
                        "abstract": "We propose a novel framework for computing the medial axis transform of 3D shapes while preserving their medial features via restricted power diagram (RPD). Medial features, including external features such as the sharp edges and corners of the input mesh surface and internal features such as the seams and junctions of medial axis, are important shape descriptors both topologically and geometrically. However, existing medial axis approximation methods fail to capture and preserve them due to the fundamentally under-sampling in the vicinity of medial features, and the difficulty to build their correct connections. In this paper we use the RPD of medial spheres and its affiliated structures to help solve these challenges. The dual structure of RPD provides the connectivity of medial spheres. The surface restricted power cell (RPC) of each medial sphere provides the tangential surface regions that these spheres have contact with. The connected components (CC) of surface RPC give us the classification of each sphere, to be on a medial sheet, a seam, or a junction. They allow us to detect insufficient sphere sampling around medial features and develop necessary conditions to preserve them. Using this RPD-based framework, we are able to construct high quality medial meshes with features preserved. Compared with existing sampling-based or voxel-based methods, our method is the first one that can preserve not only external features but also internal features of medial axes."
                    },
                    {
                        "title": "You Only Hypothesize Once: Point Cloud Registration with Rotation-equivariant Descriptors",
                        "abstract": "In this paper, we propose a novel local descriptor-based framework, called You Only Hypothesize Once (YOHO), for the registration of two unaligned point clouds. In contrast to most existing local descriptors which rely on a fragile local reference frame to gain rotation invariance, the proposed descriptor achieves the rotation invariance by recent technologies of group equivariant feature learning, which brings more robustness to point density and noise. Meanwhile, the descriptor in YOHO also has a rotation equivariant part, which enables us to estimate the registration from just one correspondence hypothesis. Such property reduces the searching space for feasible transformations, thus greatly improves both the accuracy and the efficiency of YOHO. Extensive experiments show that YOHO achieves superior performances with much fewer needed RANSAC iterations on four widely-used datasets, the 3DMatch/3DLoMatch datasets, the ETH dataset and the WHU-TLS dataset. More details are shown in our project page: https://hpwang-whu.github.io/YOHO/."
                    },
                    {
                        "title": "Moir\u00e9 Photo Restoration Using Multiresolution Convolutional Neural Networks",
                        "abstract": "Digital cameras and mobile phones enable us to conveniently record precious moments. While digital image quality is constantly being improved, taking high-quality photos of digital screens still remains challenging because the photos are often contaminated with moir\\'{e} patterns, a result of the interference between the pixel grids of the camera sensor and the device screen. Moir\\'{e} patterns can severely damage the visual quality of photos. However, few studies have aimed to solve this problem. In this paper, we introduce a novel multiresolution fully convolutional network for automatically removing moir\\'{e} patterns from photos. Since a moir\\'{e} pattern spans over a wide range of frequencies, our proposed network performs a nonlinear multiresolution analysis of the input image before computing how to cancel moir\\'{e} artefacts within every frequency band. We also create a large-scale benchmark dataset with $100,000^+$ image pairs for investigating and evaluating moir\\'{e} pattern removal algorithms. Our network achieves state-of-the-art performance on this dataset in comparison to existing learning architectures for image restoration problems."
                    },
                    {
                        "title": "Efficient Reciprocal Collision Avoidance between Heterogeneous Agents Using CTMAT",
                        "abstract": "We present a novel algorithm for reciprocal collision avoidance between heterogeneous agents of different shapes and sizes. We present a novel CTMAT representation based on medial axis transform to compute a tight fitting bounding shape for each agent. Each CTMAT is represented using tuples, which are composed of circular arcs and line segments. Based on the reciprocal velocity obstacle formulation, we reduce the problem to solving a low-dimensional linear programming between each pair of tuples belonging to adjacent agents. We precompute the Minkowski Sums of tuples to accelerate the runtime performance. Finally, we provide an efficient method to update the orientation of each agent in a local manner. We have implemented the algorithm and highlight its performance on benchmarks corresponding to road traffic scenarios and different vehicles. The overall runtime performance is comparable to prior multi-agent collision avoidance algorithms that use circular or elliptical agents. Our approach is less conservative and results in fewer false collisions."
                    },
                    {
                        "title": "AutoRVO: Local Navigation with Dynamic Constraints in Dense Heterogeneous Traffic",
                        "abstract": "We present a novel algorithm for computing collision-free navigation for heterogeneous road-agents such as cars, tricycles, bicycles, and pedestrians in dense traffic. Our approach currently assumes the positions, shapes, and velocities of all vehicles and pedestrians are known and computes smooth trajectories for each agent by taking into account the dynamic constraints. We describe an efficient optimization-based algorithm for each road-agent based on reciprocal velocity obstacles that takes into account kinematic and dynamic constraints. Our algorithm uses tight fitting shape representations based on medial axis to compute collision-free trajectories in dense traffic situations. We evaluate the performance of our algorithm in real-world dense traffic scenarios and highlight the benefits over prior reciprocal collision avoidance schemes."
                    },
                    {
                        "title": "Floorplan-Jigsaw: Jointly Estimating Scene Layout and Aligning Partial Scans",
                        "abstract": "We present a novel approach to align partial 3D reconstructions which may not have substantial overlap. Using floorplan priors, our method jointly predicts a room layout and estimates the transformations from a set of partial 3D data. Unlike the existing methods relying on feature descriptors to establish correspondences, we exploit the 3D \"box\" structure of a typical room layout that meets the Manhattan World property. We first estimate a local layout for each partial scan separately and then combine these local layouts to form a globally aligned layout with loop closure. Without the requirement of feature matching, the proposed method enables some novel applications ranging from large or featureless scene reconstruction and modeling from sparse input. We validate our method quantitatively and qualitatively on real and synthetic scenes of various sizes and complexities. The evaluations and comparisons show superior effectiveness and accuracy of our method."
                    },
                    {
                        "title": "Modeling 3D Shapes by Reinforcement Learning",
                        "abstract": "We explore how to enable machines to model 3D shapes like human modelers using deep reinforcement learning (RL). In 3D modeling software like Maya, a modeler usually creates a mesh model in two steps: (1) approximating the shape using a set of primitives; (2) editing the meshes of the primitives to create detailed geometry. Inspired by such artist-based modeling, we propose a two-step neural framework based on RL to learn 3D modeling policies. By taking actions and collecting rewards in an interactive environment, the agents first learn to parse a target shape into primitives and then to edit the geometry. To effectively train the modeling agents, we introduce a novel training algorithm that combines heuristic policy, imitation learning and reinforcement learning. Our experiments show that the agents can learn good policies to produce regular and structure-aware mesh models, which demonstrates the feasibility and effectiveness of the proposed RL framework."
                    },
                    {
                        "title": "A Sparse Graph-Structured Lasso Mixed Model for Genetic Association with Confounding Correction",
                        "abstract": "While linear mixed model (LMM) has shown a competitive performance in correcting spurious associations raised by population stratification, family structures, and cryptic relatedness, more challenges are still to be addressed regarding the complex structure of genotypic and phenotypic data. For example, geneticists have discovered that some clusters of phenotypes are more co-expressed than others. Hence, a joint analysis that can utilize such relatedness information in a heterogeneous data set is crucial for genetic modeling.   We proposed the sparse graph-structured linear mixed model (sGLMM) that can incorporate the relatedness information from traits in a dataset with confounding correction. Our method is capable of uncovering the genetic associations of a large number of phenotypes together while considering the relatedness of these phenotypes. Through extensive simulation experiments, we show that the proposed model outperforms other existing approaches and can model correlation from both population structure and shared signals. Further, we validate the effectiveness of sGLMM in the real-world genomic dataset on two different species from plants and humans. In Arabidopsis thaliana data, sGLMM behaves better than all other baseline models for 63.4% traits. We also discuss the potential causal genetic variation of Human Alzheimer's disease discovered by our model and justify some of the most important genetic loci."
                    },
                    {
                        "title": "Unsupervised Shape Completion via Deep Prior in the Neural Tangent Kernel Perspective",
                        "abstract": "We present a novel approach for completing and reconstructing 3D shapes from incomplete scanned data by using deep neural networks. Rather than being trained on supervised completion tasks and applied on a testing shape, the network is optimized from scratch on the single testing shape, to fully adapt to the shape and complete the missing data using contextual guidance from the known regions. The ability to complete missing data by an untrained neural network is usually referred to as the deep prior. In this paper, we interpret the deep prior from a neural tangent kernel (NTK) perspective and show that the completed shape patches by the trained CNN are naturally similar to existing patches, as they are proximate in the kernel feature space induced by NTK. The interpretation allows us to design more efficient network structures and learning mechanisms for the shape completion and reconstruction task. Being more aware of structural regularities than both traditional and other unsupervised learning-based reconstruction methods, our approach completes large missing regions with plausible shapes and complements supervised learning-based methods that use database priors by requiring no extra training data set and showing flexible adaptation to a particular shape instance."
                    },
                    {
                        "title": "ToothInpaintor: Tooth Inpainting from Partial 3D Dental Model and 2D Panoramic Image",
                        "abstract": "In orthodontic treatment, a full tooth model consisting of both the crown and root is indispensable in making the treatment plan. However, acquiring tooth root information to obtain the full tooth model from CBCT images is sometimes restricted due to the massive radiation of CBCT scanning. Thus, reconstructing the full tooth shape from the ready-to-use input, e.g., the partial intra-oral scan and the 2D panoramic image, is an applicable and valuable solution. In this paper, we propose a neural network, called ToothInpaintor, that takes as input a partial 3D dental model and a 2D panoramic image and reconstructs the full tooth model with high-quality root(s). Technically, we utilize the implicit representation for both the 3D and 2D inputs, and learn a latent space of the full tooth shapes. At test time, given an input, we successfully project it to the learned latent space via neural optimization to obtain the full tooth model conditioned on the input. To help find the robust projection, a novel adversarial learning module is exploited in our pipeline. We extensively evaluate our method on a dataset collected from real-world clinics. The evaluation, comparison, and comprehensive ablation studies demonstrate that our approach produces accurate complete tooth models robustly and outperforms the state-of-the-art methods."
                    },
                    {
                        "title": "NeCGS: Neural Compression for 3D Geometry Sets",
                        "abstract": "This paper explores the problem of effectively compressing 3D geometry sets containing diverse categories. We make \\textit{the first} attempt to tackle this fundamental and challenging problem and propose NeCGS, a neural compression paradigm, which can compress hundreds of detailed and diverse 3D mesh models (~684 MB) by about 900 times (0.76 MB) with high accuracy and preservation of detailed geometric details. Specifically, we first represent each irregular mesh model/shape in a regular representation that implicitly describes the geometry structure of the model using a 4D regular volume, called TSDF-Def volume. Such a regular representation can not only capture local surfaces more effectively but also facilitate the subsequent process. Then we construct a quantization-aware auto-decoder network architecture to regress these 4D volumes, which can summarize the similarity of local geometric structures within a model and across different models for redundancy limination, resulting in more compact representations, including an embedded feature of a smaller size associated with each model and a network parameter set shared by all models. We finally encode the resulting features and network parameters into bitstreams through entropy coding. After decompressing the features and network parameters, we can reconstruct the TSDF-Def volumes, where the 3D surfaces can be extracted through the deformable marching cubes.Extensive experiments and ablation studies demonstrate the significant advantages of our NeCGS over state-of-the-art methods both quantitatively and qualitatively."
                    },
                    {
                        "title": "Harvesting Discriminative Meta Objects with Deep CNN Features for Scene Classification",
                        "abstract": "Recent work on scene classification still makes use of generic CNN features in a rudimentary manner. In this ICCV 2015 paper, we present a novel pipeline built upon deep CNN features to harvest discriminative visual objects and parts for scene classification. We first use a region proposal technique to generate a set of high-quality patches potentially containing objects, and apply a pre-trained CNN to extract generic deep features from these patches. Then we perform both unsupervised and weakly supervised learning to screen these patches and discover discriminative ones representing category-specific objects and parts. We further apply discriminative clustering enhanced with local CNN fine-tuning to aggregate similar objects and parts into groups, called meta objects. A scene image representation is constructed by pooling the feature response maps of all the learned meta objects at multiple spatial scales. We have confirmed that the scene image representation obtained using this new pipeline is capable of delivering state-of-the-art performance on two popular scene benchmark datasets, MIT Indoor 67~\\cite{MITIndoor67} and Sun397~\\cite{Sun397}"
                    },
                    {
                        "title": "SparseNeuS: Fast Generalizable Neural Surface Reconstruction from Sparse Views",
                        "abstract": "We introduce SparseNeuS, a novel neural rendering based method for the task of surface reconstruction from multi-view images. This task becomes more difficult when only sparse images are provided as input, a scenario where existing neural reconstruction approaches usually produce incomplete or distorted results. Moreover, their inability of generalizing to unseen new scenes impedes their application in practice. Contrarily, SparseNeuS can generalize to new scenes and work well with sparse images (as few as 2 or 3). SparseNeuS adopts signed distance function (SDF) as the surface representation, and learns generalizable priors from image features by introducing geometry encoding volumes for generic surface prediction. Moreover, several strategies are introduced to effectively leverage sparse views for high-quality reconstruction, including 1) a multi-level geometry reasoning framework to recover the surfaces in a coarse-to-fine manner; 2) a multi-scale color blending scheme for more reliable color prediction; 3) a consistency-aware fine-tuning scheme to control the inconsistent regions caused by occlusion and noise. Extensive experiments demonstrate that our approach not only outperforms the state-of-the-art methods, but also exhibits good efficiency, generalizability, and flexibility."
                    },
                    {
                        "title": "Fast B-spline Curve Fitting by L-BFGS",
                        "abstract": "We propose a novel method for fitting planar B-spline curves to unorganized data points. In traditional methods, optimization of control points and foot points are performed in two very time-consuming steps in each iteration: 1) control points are updated by setting up and solving a linear system of equations; and 2) foot points are computed by projecting each data point onto a B-spline curve. Our method uses the L-BFGS optimization method to optimize control points and foot points simultaneously and therefore it does not need to perform either matrix computation or foot point projection in every iteration. As a result, our method is much faster than existing methods."
                    }
                ]
            },
            "9b4d3838-f156-4f97-a8bd-d407eb20f738": {
                "pk": "9b4d3838-f156-4f97-a8bd-d407eb20f738",
                "name": "Ying He",
                "collaborators": [
                    "Xiang Ying",
                    "Shi-Qing Xin",
                    "Jincai Chang",
                    "Jun Luo",
                    "Konstantin Borovkov"
                ],
                "domain": [
                    "Geometric Algorithms",
                    "Wireless Sensor Networks",
                    "Mathematical Modeling",
                    "Computational Geometry"
                ],
                "publications": [
                    {
                        "title": "Parallel Chen-Han (PCH) Algorithm for Discrete Geodesics",
                        "abstract": "In many graphics applications, the computation of exact geodesic distance is very important. However, the high computational cost of the existing geodesic algorithms means that they are not practical for large-scale models or time-critical applications. To tackle this challenge, we propose the parallel Chen-Han (or PCH) algorithm, which extends the classic Chen-Han (CH) discrete geodesic algorithm to the parallel setting. The original CH algorithm and its variant both lack a parallel solution because the windows (a key data structure that carries the shortest distance in the wavefront propagation) are maintained in a strict order or a tightly coupled manner, which means that only one window is processed at a time. We propose dividing the CH's sequential algorithm into four phases, window selection, window propagation, data organization, and events processing so that there is no data dependence or conflicts in each phase and the operations within each phase can be carried out in parallel. The proposed PCH algorithm is able to propagate a large number of windows simultaneously and independently. We also adopt a simple yet effective strategy to control the total number of windows. We implement the PCH algorithm on modern GPUs (such as Nvidia GTX 580) and analyze the performance in detail. The performance improvement (compared to the sequential algorithms) is highly consistent with GPU double-precision performance (GFLOPS). Extensive experiments on real-world models demonstrate an order of magnitude improvement in execution time compared to the state-of-the-art."
                    },
                    {
                        "title": "Golden interpolation",
                        "abstract": "For the classic aesthetic interpolation problem, we propose an entirely new thought: apply the golden section. For how to apply the golden section to interpolation methods, we present three examples: the golden step interpolation, the golden piecewise linear interpolation and the golden curve interpolation, which respectively deal with the applications of golden section in the interpolation of degree 0, 1, and 2 in the plane. In each example, we present our basic ideas, the specific methods, comparative examples and applications, and relevant criteria. And it is worth mentioning that for aesthetics, we propose two novel concepts: the golden cuspidal hill and the golden domed hill. This paper aims to provide the reference for the combination of golden section and interpolation, and stimulate more and better related researches."
                    },
                    {
                        "title": "GeoQuorum: Load Balancing and Energy Efficient Data Access in Wireless Sensor Networks",
                        "abstract": "When data productions and consumptions are heavily unbalanced and when the origins of data queries are spatially and temporally distributed, the so called in-network data storage paradigm supersedes the conventional data collection paradigm in wireless sensor networks (WSNs). In this paper, we first introduce geometric quorum systems (along with their metrics) to incarnate the idea of in-network data storage. These quorum systems are \"geometric\" because curves (rather than discrete node sets) are used to form quorums. We then propose GeoQuorum as a new quorum system, for which the quorum forming curves are parameterized. Though our proposal stems from the existing work on using curves to guide data replication and retrieval in dense WSNs, we significantly expand this design methodology, by endowing GeoQuorum with a great flexibility to fine-tune itself towards different application requirements. In particular, the tunability allows GeoQuorum to substantially improve the load balancing performance and to remain competitive in energy efficiency. Both our analysis and simulations confirm the performance enhancement brought by GeoQuorum."
                    },
                    {
                        "title": "On ruin probabilities in the presence of risky investments and random switching",
                        "abstract": "We study the asymptotic behavior of ruin probabilities, as the initial reserve goes to infinity, for a reserve process model where claims arrive according to a renewal process, while between the claim times the process has the dynamics of geometric Brownian motion-type It\\^o processes with time-dependent random coefficients. These coefficients are ``reset\" after each claim time, switching to new values independent of the past history of the process. We use the implicit renewal theory to obtain power-function bounds for the eventual ruin probability. In the special case when the random drift and diffusion coefficients of the investment returns process remain unchanged between consecutive claim arrivals, we obtain conditions for existence of Lundberg's exponent for our model ensuring the power function behaviour for the ruin probability."
                    }
                ]
            }
        }
    },
    "2409.15393": {
        "paper_data": {
            "title": "Approximated Orthogonal Projection Unit: Stabilizing Regression Network Training Using Natural Gradient",
            "url": "http://arxiv.org/abs/2409.15393v1",
            "arxiv_id": "2409.15393",
            "authors": [
                "Shaoqi Wang",
                "Chunjie Yang",
                "Siwei Lou"
            ],
            "abstract": "Neural networks (NN) are extensively studied in cutting-edge soft sensor models due to their feature extraction and function approximation capabilities. Current research into network-based methods primarily focuses on models' offline accuracy. Notably, in industrial soft sensor context, online optimizing stability and interpretability are prioritized, followed by accuracy. This requires a clearer understanding of network's training process. To bridge this gap, we propose a novel NN named the Approximated Orthogonal Projection Unit (AOPU) which has solid mathematical basis and presents superior training stability. AOPU truncates the gradient backpropagation at dual parameters, optimizes the trackable parameters updates, and enhances the robustness of training. We further prove that AOPU attains minimum variance estimation (MVE) in NN, wherein the truncated gradient approximates the natural gradient (NG). Empirical results on two chemical process datasets clearly show that AOPU outperforms other models in achieving stable convergence, marking a significant advancement in soft sensor field.",
            "introduction": "   1 Introduction  Deep learning methods have achieved recent success in many regression areas such as natural language processing, protein structure prediction, and building energy consumption forecasting. However, for these methods to be useful in the industrial soft sensor field, which demands higher immediacy and stability, further research into model structure and the stability of the training process is necessary [1, 2, 3]. The safety and economic impact of factory impose stringent requirements on soft sensor models deployed online [4, 5]. For example, each mini-batch update must not cause significant performance fluctuations to ensure that downstream controllers and monitors do not execute erroneous actions; soft sensor models must be deployed online to avoid fluctuations due to changes in operating conditions and model switching [6]. Common network\u2019s training tricks are not suitable for soft sensor contexts. For example, it\u2019s not feasible to use checkpoints for early stopping but to always use the latest updated checkpoint; there wouldn\u2019t be adaptive learning rate changes but rather a constant learning rate maintained throughout. Experimental results demonstrate that such differences lead to a substantial decline in performance. These constraints necessitate the development of better-suited network architectures for regression task that ensure more stable optimization during training [7, 6].   MVE is the best unbiased estimator under the Mean Squared Error (MSE) criterion, essentially representing the performance ceiling for regression models [8]. Unfortunately, directly applying MVE to NN is challenging due to the difficulty in obtaining the likelihood distribution p\u2062(y|x)\ud835\udc5dconditional\ud835\udc66\ud835\udc65p(y|x)italic_p ( italic_y | italic_x ) of inputs x\ud835\udc65xitalic_x and outputs y\ud835\udc66yitalic_y [9]. Traditional research on MVE has focused on techniques like Kalman filtering [10, 11], algorithms based on Ordinary Least Squares (OLS) [12, 13], and other system identification research [14, 15, 16] which operate under linear and convex conditions. These methods, while effective within their scope, have limited expressive power [17].   Many studies have explored NN optimization from various perspectives such as adaptive learning rates [18], momentum updates [19] and customized loss functions [20]. Compared to the first-order optimization methods, second-order optimization algorithms based on Natural Gradient Descent (NGD) [21, 22] can achieve faster and more stable convergence [23]. This is because NGD considers the manifold distribution of model parameters by computing the Fisher Information Matrix (FIM) during the gradient update process. However, calculating FIM introduces significant computational overhead, making NGD challenging to implement in most machine learning models [24]. Much research is focused on adapting model structures for NGD [25, 26, 27] and reducing its computational costs [28, 29, 30], yet applying NGD to NN optimization remains an unsolved issue [31, 32, 33].   Some studies have approached the regression task from the perspective of targeted modular design [34, 35, 36], emphasizing the construction of local neuron-level rules to assist models in learning conducive features , exemplified by SLSTM [37], SIAE [38], MIF-Autoformer [39], and CBMP [40]. Some endeavors have yielded results with solid theoretical underpinnings [41, 5, 42], such as S4 [43], VAE, VIOG [44]. These examples, integrate with other great work [27, 45, 42, 46], jointly demonstrate excellent integrations of NN with theoretical basis and also have a stronger expressivity compared to the identification algorithms [14, 15, 47]. However, even though these",
            "references": [
                {
                    "title": "Variational Inference Over Graph: Knowledge Representation for Deep Process Data Analytics",
                    "abstract": "With the advent of the industrial Big Data era, accurate estimation of product quality and monitoring of working conditions from historical data have become crucial in the process industry. However, the majority of data-driven approaches predominantly rely on observational data, overlooking the valuable empirical knowledge derived from experience or underlying mechanisms. In order to leverage this knowledge, researchers employ various graph neural network-based methods which introduce connections among process variables for feature extraction. Nevertheless, it is imperative to recognize that process knowledge undergoes changes due to internal or external concept drift. To address this challenge, we propose a novel deep learning module called \u201cvariational inference over graph\u201d to effectively harness shifting knowledge. Building upon the self-attention mechanism, we design a probabilistic self-attention mechanism for encoding and reconciling prior knowledge. Instead of directly encoding the prior knowledge through graph neural network edges, we incorporate it as regularization term within the variational inference framework that accounts for knowledge shift. Furthermore, we introduce reparameterization estimator to control the variance resulting from knowledge uncertainty. To showcase the capability of our proposed method, we conduct various experiments on quality prediction task in real industrial processes."
                },
                {
                    "title": "Blast Furnace Ironmaking Process Monitoring With Time-Constrained Global and Local Nonlinear Analytic Stationary Subspace Analysis",
                    "abstract": "In this article, a novel time-constrained global and local nonlinear analytic stationary subspace analysis (Tc-GLNASSA) is proposed to enhance blast furnace ironmaking process (BFIP) monitoring. Although the existing analytic stationary subspace analysis method has been available for deriving process consistent relationships. However, the presence of complex nonlinear, periodic nonstationary, and time-varying smelting conditions renders the satisfactory estimation of stationary projections unattainable. To this end, we leverage multiple kernel functions and manifold learning methods to establish a global and local nonlinear structure with time constraints, which will identify the unique nonlinearities excited by periodic nonstationarity. Meanwhile, a singular value decomposition-based modeling efficiency promotion strategy is constructed to reduce the proposed Tc-GLNASSA's computational complexity significantly. The orthogonality of model update scheme is analyzed theoretically, and an overall BFIP monitoring framework is given. Ultimately, practical BFIP case studies fully demonstrate the effectiveness of our proposal."
                },
                {
                    "title": "Can We Remove the Square-Root in Adaptive Gradient Methods? A Second-Order Perspective",
                    "abstract": "Adaptive gradient optimizers like Adam(W) are the default training algorithms for many deep learning architectures, such as transformers. Their diagonal preconditioner is based on the gradient outer product which is incorporated into the parameter update via a square root. While these methods are often motivated as approximate second-order methods, the square root represents a fundamental difference. In this work, we investigate how the behavior of adaptive methods changes when we remove the root, i.e., strengthen their second-order motivation. Surprisingly, we find that such square-root-free adaptive methods close the generalization gap to SGD on convolutional architectures, while maintaining their root-based counterpart's performance on transformers. The second-order perspective also has practical benefits for developing non-diagonal methods that can incorporate arbitrary curvature approximations through the concept of preconditioner invariance. In contrast to root-based methods like Shampoo, root-free counterparts work well and fast with half-precision since they do not require numerically unstable matrix root decompositions and inversions. Overall, our findings provide new insights into the development of adaptive methods and raise important questions regarding the overlooked role of adaptivity in their success. (experiment code: https://github.com/yorkerlin/remove-the-square-root optimizer code: https://github.com/f-dangel/sirfshampoo)"
                },
                {
                    "title": "Multisource Information Fusion for Autoformer: Soft Sensor Modeling of FeO Content in Iron Ore Sintering Process",
                    "abstract": "As a key thermal-state indicator of the iron ore sintering process, the content of ferrous oxide (FeO) in the finished sinter is directly related to product quality. Based on the massive data of sintering process, the data-driven soft sensor model provides a good choice for real-time FeO content detection. However, the complex characteristics of the data, including dynamics, nonlinearity, and multisource heterogeneity, are still the main obstacles to improving the modeling accuracy. To solve this problem, in this article, a multisource information fusion autoformer (MIF-Autoformer) model is introduced. First, feature-level information fusion and data-level information fusion are implemented based on the MIF strategy. Then, the comprehensive information of the sintering process is fed to the downstream autoformer model in a serial manner, which not only improves the information capacity but also provides additional prior information about the FeO content grade. This is helpful for autoformer to capture the complex temporal distributions in the sintering process. Finally, the proposed model is applied to a real sintering plant. Experimental results show that the hybrid image features provided by the MIF strategy have a general optimization effect on different competitive models, and MIF-Autoformer exhibits the lowest prediction error on the test set."
                },
                {
                    "title": "OneNet: Enhancing Time Series Forecasting Models under Concept Drift by Online Ensembling",
                    "abstract": "Online updating of time series forecasting models aims to address the concept drifting problem by efficiently updating forecasting models based on streaming data. Many algorithms are designed for online time series forecasting, with some exploiting cross-variable dependency while others assume independence among variables. Given every data assumption has its own pros and cons in online time series modeling, we propose \\textbf{On}line \\textbf{e}nsembling \\textbf{Net}work (OneNet). It dynamically updates and combines two models, with one focusing on modeling the dependency across the time dimension and the other on cross-variate dependency. Our method incorporates a reinforcement learning-based approach into the traditional online convex programming framework, allowing for the linear combination of the two models with dynamically adjusted weights. OneNet addresses the main shortcoming of classical online learning methods that tend to be slow in adapting to the concept drift. Empirical results show that OneNet reduces online forecasting error by more than $\\mathbf{50\\%}$ compared to the State-Of-The-Art (SOTA) method. The code is available at \\url{https://github.com/yfzhang114/OneNet}."
                },
                {
                    "title": "Deep networks for system identification: a Survey",
                    "abstract": "Deep learning is a topic of considerable current interest. The availability of massive data collections and powerful software resources has led to an impressive amount of results in many application areas that reveal essential but hidden properties of the observations. System identification learns mathematical descriptions of dynamic systems from input-output data and can thus benefit from the advances of deep neural networks to enrich the possible range of models to choose from. For this reason, we provide a survey of deep learning from a system identification perspective. We cover a wide spectrum of topics to enable researchers to understand the methods, providing rigorous practical and theoretical insights into the benefits and challenges of using them. The main aim of the identified model is to predict new data from previous observations. This can be achieved with different deep learning based modelling techniques and we discuss architectures commonly adopted in the literature, like feedforward, convolutional, and recurrent networks. Their parameters have to be estimated from past data trying to optimize the prediction performance. For this purpose, we discuss a specific set of first-order optimization tools that is emerged as efficient. The survey then draws connections to the well-studied area of kernel-based methods. They control the data fit by regularization terms that penalize models not in line with prior assumptions. We illustrate how to cast them in deep architectures to obtain deep kernel-based methods. The success of deep learning also resulted in surprising empirical observations, like the counter-intuitive behaviour of models with many parameters. We discuss the role of overparameterized models, including their connection to kernels, as well as implicit regularization mechanisms which affect generalization, specifically the interesting phenomena of benign overfitting ..."
                },
                {
                    "title": "Knowledge Automation Through Graph Mining, Convolution, and Explanation Framework: A Soft Sensor Practice",
                    "abstract": "In industrial processes, data-driven soft sensors have played an important role for the effective process control, optimization, and monitoring. Deep learning technique has been widely used in soft sensor field in recent years for its excellent feature representation capability in spatial and temporal scales. However, the shortcomings for deep learning technique seriously hinder its application in industrial processes. For example, the knowledge cannot be added into the model, and the model prediction could not be well explained. To solve those problems, the graph mining, convolution, and explanation framework is proposed for knowledge automation in this article. Based on the equivalence analysis of the self-attention mechanism (SAM) and graph convolution (GC) operation, the spatial SAM is adopted for knowledge discovery from data directly. After that, the GC layer considering the relationship between process variables can utilize the knowledge for constructing soft sensor models. Besides, to explain which knowledge contributes to the final model prediction, the graph neural network explainer is designed for explaining the model output. Finally, the effectiveness and feasibility of the framework are evaluated on an industrial process, in which the knowledge discovered from the data is of great consistence with the prior knowledge, and the final explanation indicated that most of the knowledge is consistent with the prior knowledge contributed to the prediction."
                },
                {
                    "title": "Non-stationary Transformers: Exploring the Stationarity in Time Series Forecasting",
                    "abstract": "Transformers have shown great power in time series forecasting due to their global-range modeling ability. However, their performance can degenerate terribly on non-stationary real-world data in which the joint distribution changes over time. Previous studies primarily adopt stationarization to attenuate the non-stationarity of original series for better predictability. But the stationarized series deprived of inherent non-stationarity can be less instructive for real-world bursty events forecasting. This problem, termed over-stationarization in this paper, leads Transformers to generate indistinguishable temporal attentions for different series and impedes the predictive capability of deep models. To tackle the dilemma between series predictability and model capability, we propose Non-stationary Transformers as a generic framework with two interdependent modules: Series Stationarization and De-stationary Attention. Concretely, Series Stationarization unifies the statistics of each input and converts the output with restored statistics for better predictability. To address the over-stationarization problem, De-stationary Attention is devised to recover the intrinsic non-stationary information into temporal dependencies by approximating distinguishable attentions learned from raw series. Our Non-stationary Transformers framework consistently boosts mainstream Transformers by a large margin, which reduces MSE by 49.43% on Transformer, 47.34% on Informer, and 46.89% on Reformer, making them the state-of-the-art in time series forecasting. Code is available at this repository: https://github.com/thuml/Nonstationary_Transformers."
                },
                {
                    "title": "Are Transformers Effective for Time Series Forecasting?",
                    "abstract": "Recently, there has been a surge of Transformer-based solutions for the long-term time series forecasting (LTSF) task. Despite the growing performance over the past few years, we question the validity of this line of research in this work. Specifically, Transformers is arguably the most successful solution to extract the semantic correlations among the elements in a long sequence. However, in time series modeling, we are to extract the temporal relations in an ordered set of continuous points. While employing positional encoding and using tokens to embed sub-series in Transformers facilitate preserving some ordering information, the nature of the permutation-invariant self-attention mechanism inevitably results in temporal information loss. \nTo validate our claim, we introduce a set of embarrassingly simple one-layer linear models named LTSF-Linear for comparison. Experimental results on nine real-life datasets show that LTSF-Linear surprisingly outperforms existing sophisticated Transformer-based LTSF models in all cases, and often by a large margin. Moreover, we conduct comprehensive empirical studies to explore the impacts of various design elements of LTSF models on their temporal relation extraction capability. We hope this surprising finding opens up new research directions for the LTSF task. We also advocate revisiting the validity of Transformer-based solutions for other time series analysis tasks (e.g., anomaly detection) in the future."
                },
                {
                    "title": "Efficiently Modeling Long Sequences with Structured State Spaces",
                    "abstract": "A central goal of sequence modeling is designing a single principled model that can address sequence data across a range of modalities and tasks, particularly on long-range dependencies. Although conventional models including RNNs, CNNs, and Transformers have specialized variants for capturing long dependencies, they still struggle to scale to very long sequences of $10000$ or more steps. A promising recent approach proposed modeling sequences by simulating the fundamental state space model (SSM) \\( x'(t) = Ax(t) + Bu(t), y(t) = Cx(t) + Du(t) \\), and showed that for appropriate choices of the state matrix \\( A \\), this system could handle long-range dependencies mathematically and empirically. However, this method has prohibitive computation and memory requirements, rendering it infeasible as a general sequence modeling solution. We propose the Structured State Space sequence model (S4) based on a new parameterization for the SSM, and show that it can be computed much more efficiently than prior approaches while preserving their theoretical strengths. Our technique involves conditioning \\( A \\) with a low-rank correction, allowing it to be diagonalized stably and reducing the SSM to the well-studied computation of a Cauchy kernel. S4 achieves strong empirical results across a diverse range of established benchmarks, including (i) 91\\% accuracy on sequential CIFAR-10 with no data augmentation or auxiliary losses, on par with a larger 2-D ResNet, (ii) substantially closing the gap to Transformers on image and language modeling tasks, while performing generation $60\\times$ faster (iii) SoTA on every task from the Long Range Arena benchmark, including solving the challenging Path-X task of length 16k that all prior work fails on, while being as efficient as all competitors."
                },
                {
                    "title": "Auto-differentiable Ensemble Kalman Filters",
                    "abstract": "Data assimilation is concerned with sequentially estimating a temporally-evolving state. This task, which arises in a wide range of scientific and engineering applications, is particularly challenging when the state is high-dimensional and the state-space dynamics are unknown. This paper introduces a machine learning framework for learning dynamical systems in data assimilation. Our auto-differentiable ensemble Kalman filters (AD-EnKFs) blend ensemble Kalman filters for state recovery with machine learning tools for learning the dynamics. In doing so, AD-EnKFs leverage the ability of ensemble Kalman filters to scale to high-dimensional states and the power of automatic differentiation to train high-dimensional surrogate models for the dynamics. Numerical results using the Lorenz-96 model show that AD-EnKFs outperform existing methods that use expectation-maximization or particle filters to merge data assimilation and machine learning. In addition, AD-EnKFs are easy to implement and require minimal tuning."
                },
                {
                    "title": "Deep Neural Networks as Point Estimates for Deep Gaussian Processes",
                    "abstract": "Neural networks and Gaussian processes are complementary in their strengths and weaknesses. Having a better understanding of their relationship comes with the promise to make each method benefit from the strengths of the other. In this work, we establish an equivalence between the forward passes of neural networks and (deep) sparse Gaussian process models. The theory we develop is based on interpreting activation functions as interdomain inducing features through a rigorous analysis of the interplay between activation functions and kernels. This results in models that can either be seen as neural networks with improved uncertainty prediction or deep Gaussian processes with increased prediction accuracy. These claims are supported by experimental results on regression and classification datasets."
                },
                {
                    "title": "Optimistic Dual Extrapolation for Coherent Non-monotone Variational Inequalities",
                    "abstract": "The optimization problems associated with training generative adversarial neural networks can be largely reduced to certain {\\em non-monotone} variational inequality problems (VIPs), whereas existing convergence results are mostly based on monotone or strongly monotone assumptions. In this paper, we propose {\\em optimistic dual extrapolation (OptDE)}, a method that only performs {\\em one} gradient evaluation per iteration. We show that OptDE is provably convergent to {\\em a strong solution} under different coherent non-monotone assumptions. In particular, when a {\\em weak solution} exists, the convergence rate of our method is $O(1/{\\epsilon^{2}})$, which matches the best existing result of the methods with two gradient evaluations. Further, when a {\\em $\\sigma$-weak solution} exists, the convergence guarantee is improved to the linear rate $O(\\log\\frac{1}{\\epsilon})$. Along the way--as a byproduct of our inquiries into non-monotone variational inequalities--we provide the near-optimal $O\\big(\\frac{1}{\\epsilon}\\log \\frac{1}{\\epsilon}\\big)$ convergence guarantee in terms of restricted strong merit function for monotone variational inequalities. We also show how our results can be naturally generalized to the stochastic setting, and obtain corresponding new convergence results. Taken together, our results contribute to the broad landscape of variational inequality--both non-monotone and monotone alike--by providing a novel and more practical algorithm with the state-of-the-art convergence guarantees."
                },
                {
                    "title": "Tractable structured natural gradient descent using local parameterizations",
                    "abstract": "Natural-gradient descent (NGD) on structured parameter spaces (e.g., low-rank covariances) is computationally challenging due to difficult Fisher-matrix computations. We address this issue by using local-parameter coordinates to obtain a flexible and efficient NGD method that works well for a wide-variety of structured parameterizations. We show four applications where our method (1) generalizes the exponential natural evolutionary strategy, (2) recovers existing Newton-like algorithms, (3) yields new structured second-order algorithms, and (4) gives new algorithms to learn covariances of Gaussian and Wishart-based distributions. We show results on a range of problems from deep learning, variational inference, and evolution strategies. Our work opens a new direction for scalable structured geometric methods."
                },
                {
                    "title": "A Survey on Deep Learning for Data-Driven Soft Sensors",
                    "abstract": "Soft sensors are widely constructed in process industry to realize process monitoring, quality prediction, and many other important applications. With the development of hardware and software, industrial processes have embraced new characteristics, which lead to the poor performance of traditional soft sensor modeling methods. Deep learning, as a kind of data-driven approach, shows its great potential in many fields, as well as in soft sensing scenarios. After a period of development, especially in the last five years, many new issues have emerged that need to be investigated. Therefore, in this article, the necessity and significance of deep learning for soft sensor applications are demonstrated first by analyzing the merits of deep learning and the trends of industrial processes. Next, mainstream deep learning models, tricks, and frameworks/toolkits are summarized and discussed to help designers propel the developing progress of soft sensors. Then, existing works are reviewed and analyzed to discuss the demands and problems occurred in practical applications. Finally, outlook and conclusions are given."
                },
                {
                    "title": "Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting",
                    "abstract": "Many real-world applications require the prediction of long sequence time-series, such as electricity consumption planning. Long sequence time-series forecasting (LSTF) demands a high prediction capacity of the model, which is the ability to capture precise long-range dependency coupling between output and input efficiently. Recent studies have shown the potential of Transformer to increase the prediction capacity. However, there are several severe issues with Transformer that prevent it from being directly applicable to LSTF, including quadratic time complexity, high memory usage, and inherent limitation of the encoder-decoder architecture. To address these issues, we design an efficient transformer-based model for LSTF, named Informer, with three distinctive characteristics: (i) a ProbSparse self-attention mechanism, which achieves O(L log L) in time complexity and memory usage, and has comparable performance on sequences' dependency alignment. (ii) the self-attention distilling highlights dominating attention by halving cascading layer input, and efficiently handles extreme long input sequences. (iii) the generative style decoder, while conceptually simple, predicts the long time-series sequences at one forward operation rather than a step-by-step way, which drastically improves the inference speed of long-sequence predictions. Extensive experiments on four large-scale datasets demonstrate that Informer significantly outperforms existing methods and provides a new solution to the LSTF problem."
                },
                {
                    "title": "Automatic Deep Extraction of Robust Dynamic Features for Industrial Big Data Modeling and Soft Sensor Application",
                    "abstract": "Dynamic is one of the main bottlenecks in the industrial soft sensor application, due to the difficulties in representing and extracting dynamic data features. Meanwhile, an end-to-end deep network owns the ability to characterize sequence data information, but its fitting ability requires improvements in practical applications. In this article, an ensemble tree model with transferable and robust dynamic features extracted by a newly developed automatic dynamic feature extractor is proposed. First, the dynamic feature extractor with an encoding\u2013decoding structure can provide effective dynamic features, which is equivalent to crossing and nonlinear mapping of sequences under the supervision of a decoder. Meanwhile, a new \u201cregularization\u201d method by smoothing dynamic features based on attention weights is proposed to denoise and alleviate the overfitting of the regressor after adding new features. Then, the extracted dynamic features can be transferred to the regressor with strong generalization ability, which takes into account the feature extraction of the deep network and the generalization of strong models. Finally, application results on a debutanizer distillation process show that the incorporation of robust dynamic features can significantly improve the soft sensing performance, compared to traditional methods. Moreover, the proposed model is further implemented through a cloud computing platform for industrial big data analytics."
                },
                {
                    "title": "Understanding Recurrent Neural Networks Using Nonequilibrium Response Theory",
                    "abstract": "Recurrent neural networks (RNNs) are brain-inspired models widely used in machine learning for analyzing sequential data. The present work is a contribution towards a deeper understanding of how RNNs process input signals using the response theory from nonequilibrium statistical mechanics. For a class of continuous-time stochastic RNNs (SRNNs) driven by an input signal, we derive a Volterra type series representation for their output. This representation is interpretable and disentangles the input signal from the SRNN architecture. The kernels of the series are certain recursively defined correlation functions with respect to the unperturbed dynamics that completely determine the output. Exploiting connections of this representation and its implications to rough paths theory, we identify a universal feature -- the response feature, which turns out to be the signature of tensor product of the input signal and a natural support basis. In particular, we show that the SRNNs can be viewed as kernel machines operating on a reproducing kernel Hilbert space associated with the response feature."
                },
                {
                    "title": "Adaptive Experimental Design with Temporal Interference: A Maximum Likelihood Approach",
                    "abstract": "Suppose an online platform wants to compare a treatment and control policy, e.g., two different matching algorithms in a ridesharing system, or two different inventory management algorithms in an online retail site. Standard randomized controlled trials are typically not feasible, since the goal is to estimate policy performance on the entire system. Instead, the typical current practice involves dynamically alternating between the two policies for fixed lengths of time, and comparing the average performance of each over the intervals in which they were run as an estimate of the treatment effect. However, this approach suffers from *temporal interference*: one algorithm alters the state of the system as seen by the second algorithm, biasing estimates of the treatment effect. Further, the simple non-adaptive nature of such designs implies they are not sample efficient. \nWe develop a benchmark theoretical model in which to study optimal experimental design for this setting. We view testing the two policies as the problem of estimating the steady state difference in reward between two unknown Markov chains (i.e., policies). We assume estimation of the steady state reward for each chain proceeds via nonparametric maximum likelihood, and search for consistent (i.e., asymptotically unbiased) experimental designs that are efficient (i.e., asymptotically minimum variance). Characterizing such designs is equivalent to a Markov decision problem with a minimum variance objective; such problems generally do not admit tractable solutions. Remarkably, in our setting, using a novel application of classical martingale analysis of Markov chains via Poisson's equation, we characterize efficient designs via a succinct convex optimization problem. We use this characterization to propose a consistent, efficient online experimental design that adaptively samples the two Markov chains."
                },
                {
                    "title": "Hierarchical Quality-Relevant Feature Representation for Soft Sensor Modeling: A Novel Deep Learning Strategy",
                    "abstract": "Deep learning is a recently developed feature representation technique for data with complicated structures, which has great potential for soft sensing of industrial processes. However, most deep networks mainly focus on hierarchical feature learning for the raw observed input data. For soft sensor applications, it is important to reduce irrelevant information and extract quality-relevant features from the raw input data for quality prediction. To deal with this problem, a novel deep learning network is proposed for quality-relevant feature representation in this article, which is based on stacked quality-driven autoencoder (SQAE). First, a quality-driven autoencoder (QAE) is designed by exploiting the quality data to guide feature extraction with the constraint that the potential features should largely reconstruct the input layer data and the quality data at the output layer. In this way, quality-relevant features can be captured by QAE. Then, by stacking multiple QAEs to construct the deep SQAE network, SQAE can gradually reduce irrelevant features and learn hierarchical quality-relevant features. Finally, the high-level quality-relevant features can be directly applied for soft sensing of the quality variables. The effectiveness and flexibility of the proposed deep learning model are validated on an industrial debutanizer column process."
                },
                {
                    "title": "Synthesizer: Rethinking Self-Attention for Transformer Models",
                    "abstract": "The dot product self-attention is known to be central and indispensable to state-of-the-art Transformer models. But is it really required? This paper investigates the true importance and contribution of the dot product-based self-attention mechanism on the performance of Transformer models. Via extensive experiments, we find that (1) random alignment matrices surprisingly perform quite competitively and (2) learning attention weights from token-token (query-key) interactions is useful but not that important after all. To this end, we propose \\textsc{Synthesizer}, a model that learns synthetic attention weights without token-token interactions. In our experiments, we first show that simple Synthesizers achieve highly competitive performance when compared against vanilla Transformer models across a range of tasks, including machine translation, language modeling, text generation and GLUE/SuperGLUE benchmarks. When composed with dot product attention, we find that Synthesizers consistently outperform Transformers. Moreover, we conduct additional comparisons of Synthesizers against Dynamic Convolutions, showing that simple Random Synthesizer is not only $60\\%$ faster but also improves perplexity by a relative $3.5\\%$. Finally, we show that simple factorized Synthesizers can outperform Linformers on encoding only tasks."
                },
                {
                    "title": "Nonlinear Dynamic Soft Sensor Modeling With Supervised Long Short-Term Memory Network",
                    "abstract": "Soft sensor has been extensively utilized in industrial processes for prediction of key quality variables. To build an accurate virtual sensor model, it is very significant to model the dynamic and nonlinear behaviors of process sequential data properly. Recently, a long short-term memory (LSTM) network has shown great modeling ability on various time series, in which basic LSTM units can handle data nonlinearities and dynamics with a dynamic latent variable structure. However, the hidden variables in the basic LSTM unit mainly focus on describing the dynamics of input variables, which lack representation for the quality data. In this paper, a supervised LSTM (SLSTM) network is proposed to learn quality-relevant hidden dynamics for soft sensor application, which is composed of basic SLSTM unit at each sampling instant. In the basic SLSTM unit, the quality and input variables are simultaneously utilized to learn the dynamic hidden states, which are more relevant and useful for quality prediction. The effectiveness of the proposed SLSTM network is demonstrated on a penicillin fermentation process and an industrial debutanizer column."
                },
                {
                    "title": "On Multilevel Best Linear Unbiased Estimators",
                    "abstract": "We present a general variance reduction technique for the estimation of the expectation of a scalar-valued quantity of interest associated with a family of model evaluations. The key idea is to ref..."
                },
                {
                    "title": "Efficiently sampling functions from Gaussian process posteriors",
                    "abstract": "Gaussian processes are the gold standard for many real-world modeling problems, especially in cases where a model's success hinges upon its ability to faithfully represent predictive uncertainty. These problems typically exist as parts of larger frameworks, wherein quantities of interest are ultimately defined by integrating over posterior distributions. These quantities are frequently intractable, motivating the use of Monte Carlo methods. Despite substantial progress in scaling up Gaussian processes to large training sets, methods for accurately generating draws from their posterior distributions still scale cubically in the number of test locations. We identify a decomposition of Gaussian processes that naturally lends itself to scalable sampling by separating out the prior from the data. Building off of this factorization, we propose an easy-to-use and general-purpose approach for fast posterior sampling, which seamlessly pairs with sparse approximations to afford scalability both during training and at test time. In a series of experiments designed to test competing sampling schemes' statistical properties and practical ramifications, we demonstrate how decoupled sample paths accurately represent Gaussian process posteriors at a fraction of the usual cost."
                },
                {
                    "title": "Fast and Simple Natural-Gradient Variational Inference with Mixture of Exponential-family Approximations",
                    "abstract": "Natural-gradient methods enable fast and simple algorithms for variational inference, but due to computational difficulties, their use is mostly limited to \\emph{minimal} exponential-family (EF) approximations. In this paper, we extend their application to estimate \\emph{structured} approximations such as mixtures of EF distributions. Such approximations can fit complex, multimodal posterior distributions and are generally more accurate than unimodal EF approximations. By using a \\emph{minimal conditional-EF} representation of such approximations, we derive simple natural-gradient updates. Our empirical results demonstrate a faster convergence of our natural-gradient method compared to black-box gradient-based methods with reparameterization gradients. Our work expands the scope of natural gradients for Bayesian inference and makes them more widely applicable than before."
                },
                {
                    "title": "Process Data Analytics via Probabilistic Latent Variable Models: A Tutorial Review",
                    "abstract": "Dimensionality reduction is important for the high-dimensional nature of data in the process industry, which has made latent variable modeling methods popular in recent years. By projecting high-dimensional data into a lower-dimensional space, latent variables models are able to extract key information from process data while simultaneously improving the efficiency of data analytics. Through a probabilistic viewpoint, this paper carries out a tutorial review of probabilistic latent variable models on process data analytics. Detailed illustrations of different kinds of basic probabilistic latent variable models (PLVM) are provided, as well as their research statuses. Additionally, more counterparts of those basic PLVMs are introduced and discussed for process data analytics. Several perspectives are highlighted for future research on this topic."
                },
                {
                    "title": "Deep Learning-Based Feature Representation and Its Application for Soft Sensor Modeling With Variable-Wise Weighted SAE",
                    "abstract": "In modern industrial processes, soft sensors have played an important role for effective process control, optimization, and monitoring. Feature representation is one of the core factors to construct accurate soft sensors. Recently, deep learning techniques have been developed for high-level abstract feature extraction in pattern recognition areas, which also have great potential for soft sensing applications. Hence, deep stacked autoencoder (SAE) is introduced for soft sensor in this paper. As for output prediction purpose, traditional deep learning algorithms cannot extract high-level output-related features. Thus, a novel variable-wise weighted stacked autoencoder (VW-SAE) is proposed for hierarchical output-related feature representation layer by layer. By correlation analysis with the output variable, important variables are identified from other ones in the input layer of each autoencoder. The variables are assigned with different weights accordingly. Then, variable-wise weighted autoencoders are designed and stacked to form deep networks. An industrial application shows that the proposed VW-SAE can give better prediction performance than the traditional multilayer neural networks and SAE."
                },
                {
                    "title": "Spectral Filtering for General Linear Dynamical Systems",
                    "abstract": "We give a polynomial-time algorithm for learning latent-state linear dynamical systems without system identification, and without assumptions on the spectral radius of the system's transition matrix. The algorithm extends the recently introduced technique of spectral filtering, previously applied only to systems with a symmetric transition matrix, using a novel convex relaxation to allow for the efficient identification of phases."
                },
                {
                    "title": "Stochastic Mirror Descent in Variationally Coherent Optimization Problems",
                    "abstract": "In this paper, we examine a class of non-convex stochastic optimization problems which we call variationally coherent, and which properly includes pseudo-/quasiconvex and star-convex optimization problems. To solve such problems, we focus on the widely used stochastic mirror descent (SMD) family of algorithms (which contains stochastic gradient descent as a special case), and we show that the last iterate of SMD converges to the problem\u2019s solution set with probability 1. This result contributes to the landscape of non-convex stochastic optimization by clarifying that neither pseudo-/quasi-convexity nor star-convexity is essential for (almost sure) global convergence; rather, variational coherence, a much weaker requirement, suffices. Characterization of convergence rates for the subclass of strongly variationally coherent optimization problems as well as simulation results are also presented."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Learning Linear Dynamical Systems via Spectral Filtering",
                    "abstract": "We present an efficient and practical algorithm for the online prediction of discrete-time linear dynamical systems with a symmetric transition matrix. We circumvent the non-convex optimization problem using improper learning: carefully overparameterize the class of LDSs by a polylogarithmic factor, in exchange for convexity of the loss functions. From this arises a polynomial-time algorithm with a near-optimal regret guarantee, with an analogous sample complexity bound for agnostic learning. Our algorithm is based on a novel filtering technique, which may be of independent interest: we convolve the time series with the eigenvectors of a certain Hankel matrix."
                },
                {
                    "title": "On the Convergence of Mirror Descent beyond Stochastic Convex Programming",
                    "abstract": "In this paper, we examine the convergence of mirror descent in a class of stochastic optimization problems that are not necessarily convex (or even quasi-convex), and which we call variationally coherent. Since the standard technique of \"ergodic averaging\" offers no tangible benefits beyond convex programming, we focus directly on the algorithm's last generated sample (its \"last iterate\"), and we show that it converges with probabiility $1$ if the underlying problem is coherent. We further consider a localized version of variational coherence which ensures local convergence of stochastic mirror descent (SMD) with high probability. These results contribute to the landscape of non-convex stochastic optimization by showing that (quasi-)convexity is not essential for convergence to a global minimum: rather, variational coherence, a much weaker requirement, suffices. Finally, building on the above, we reveal an interesting insight regarding the convergence speed of SMD: in problems with sharp minima (such as generic linear programs or concave minimization problems), SMD reaches a minimum point in a finite number of steps (a.s.), even in the presence of persistent gradient noise. This result is to be contrasted with existing black-box convergence rate estimates that are only asymptotic."
                },
                {
                    "title": "Conjugate-Computation Variational Inference: Converting Variational Inference in Non-Conjugate Models to Inferences in Conjugate Models",
                    "abstract": "Variational inference is computationally challenging in models that contain both conjugate and non-conjugate terms. Methods specifically designed for conjugate models, even though computationally efficient, find it difficult to deal with non-conjugate terms. On the other hand, stochastic-gradient methods can handle the non-conjugate terms but they usually ignore the conjugate structure of the model which might result in slow convergence. In this paper, we propose a new algorithm called Conjugate-computation Variational Inference (CVI) which brings the best of the two worlds together -- it uses conjugate computations for the conjugate terms and employs stochastic gradients for the rest. We derive this algorithm by using a stochastic mirror-descent method in the mean-parameter space, and then expressing each gradient step as a variational inference in a conjugate model. We demonstrate our algorithm's applicability to a large class of models and establish its convergence. Our experimental results show that our method converges much faster than the methods that ignore the conjugate structure of the model."
                },
                {
                    "title": "Understanding Probabilistic Sparse Gaussian Process Approximations",
                    "abstract": "Good sparse approximations are essential for practical inference in Gaussian Processes as the computational cost of exact methods is prohibitive for large datasets. The Fully Independent Training Conditional (FITC) and the Variational Free Energy (VFE) approximations are two recent popular methods. Despite superficial similarities, these approximations have surprisingly different theoretical properties and behave differently in practice. We thoroughly investigate the two methods for regression both analytically and through illustrative examples, and draw conclusions to guide practical application."
                },
                {
                    "title": "A Unifying Framework of Anytime Sparse Gaussian Process Regression Models with Stochastic Variational Inference for Big Data",
                    "abstract": "This paper presents a novel unifying framework of anytime sparse Gaussian process regression (SGPR) models that can produce good predictive performance fast and improve their predictive performance over time. Our proposed unifying framework reverses the variational inference procedure to theoretically construct a non-trivial, concave functional that is maximized at the predictive distribution of any SGPR model of our choice. As a result, a stochastic natural gradient ascent method can be derived that involves iteratively following the stochastic natural gradient of the functional to improve its estimate of the predictive distribution of the chosen SGPR model and is guaranteed to achieve asymptotic convergence to it. Interestingly, we show that if the predictive distribution of the chosen SGPR model satisfies certain decomposability conditions, then the stochastic natural gradient is an unbiased estimator of the exact natural gradient and can be computed in constant time (i.e., independent of data size) at each iteration. We empirically evaluate the trade-off between the predictive performance vs. time efficiency of the anytime SGPR models on two real-world million-sized datasets."
                },
                {
                    "title": "Provable Bayesian Inference via Particle Mirror Descent",
                    "abstract": "Bayesian methods are appealing in their flexibility in modeling complex data and ability in capturing uncertainty in parameters. However, when Bayes' rule does not result in tractable closed-form, most approximate inference algorithms lack either scalability or rigorous guarantees. To tackle this challenge, we propose a simple yet provable algorithm, \\emph{Particle Mirror Descent} (PMD), to iteratively approximate the posterior density. PMD is inspired by stochastic functional mirror descent where one descends in the density space using a small batch of data points at each iteration, and by particle filtering where one uses samples to approximate a function. We prove result of the first kind that, with $m$ particles, PMD provides a posterior density estimator that converges in terms of $KL$-divergence to the true posterior in rate $O(1/\\sqrt{m})$. We demonstrate competitive empirical performances of PMD compared to several approximate inference algorithms in mixture models, logistic regression, sparse Gaussian processes and latent Dirichlet allocation on large scale datasets."
                },
                {
                    "title": "A Critical Review of Recurrent Neural Networks for Sequence Learning",
                    "abstract": "Countless learning tasks require dealing with sequential data. Image captioning, speech synthesis, and music generation all require that a model produce outputs that are sequences. In other domains, such as time series prediction, video analysis, and musical information retrieval, a model must learn from inputs that are sequences. Interactive tasks, such as translating natural language, engaging in dialogue, and controlling a robot, often demand both capabilities. Recurrent neural networks (RNNs) are connectionist models that capture the dynamics of sequences via cycles in the network of nodes. Unlike standard feedforward neural networks, recurrent networks retain a state that can represent information from an arbitrarily long context window. Although recurrent neural networks have traditionally been dicult to train, and often contain millions of parameters, recent advances in network architectures, optimization techniques, and parallel computation have enabled successful large-scale learning with them. In recent years, systems based on long short-term memory (LSTM) and bidirectional (BRNN) architectures have demonstrated ground-breaking performance on tasks as varied as image captioning, language translation, and handwriting recognition. In this survey, we review and synthesize the research that over the past three decades rst yielded and then made practical these powerful learning models. When appropriate, we reconcile conicting notation and nomenclature. Our goal is to provide a selfcontained explication of the state of the art together with a historical perspective and references to primary research."
                },
                {
                    "title": "Optimizing Neural Networks with Kronecker-factored Approximate Curvature",
                    "abstract": "We propose an efficient method for approximating natural gradient descent in neural networks which we call Kronecker-Factored Approximate Curvature (K-FAC). K-FAC is based on an efficiently invertible approximation of a neural network's Fisher information matrix which is neither diagonal nor low-rank, and in some cases is completely non-sparse. It is derived by approximating various large blocks of the Fisher (corresponding to entire layers) as being the Kronecker product of two much smaller matrices. While only several times more expensive to compute than the plain stochastic gradient, the updates produced by K-FAC make much more progress optimizing the objective, which results in an algorithm that can be much faster than stochastic gradient descent with momentum in practice. And unlike some previously proposed approximate natural-gradient/Newton methods which use high-quality non-diagonal curvature matrices (such as Hessian-free optimization), K-FAC works very well in highly stochastic optimization regimes. This is because the cost of storing and inverting K-FAC's approximation to the curvature matrix does not depend on the amount of data used to estimate it, which is a feature typically associated only with diagonal or low-rank approximations to the curvature matrix."
                },
                {
                    "title": "New Insights and Perspectives on the Natural Gradient Method",
                    "abstract": "Natural gradient descent is an optimization method traditionally motivated from the perspective of information geometry, and works well for many applications as an alternative to stochastic gradient descent. In this paper we critically analyze this method and its properties, and show how it can be viewed as a type of approximate 2nd-order optimization method, where the Fisher information matrix can be viewed as an approximation of the Hessian. This perspective turns out to have significant implications for how to design a practical and robust version of the method. Additionally, we make the following contributions to the understanding of natural gradient and 2nd-order methods: a thorough analysis of the convergence speed of stochastic natural gradient descent (and more general stochastic 2nd-order methods) as applied to convex quadratics, a critical examination of the oft-used \"empirical\" approximation of the Fisher matrix, and an analysis of the (approximate) parameterization invariance property possessed by natural gradient methods, which we show still holds for certain choices of the curvature matrix other than the Fisher, but notably not the Hessian."
                },
                {
                    "title": "Mixture semisupervised principal component regression model and soft sensor application",
                    "abstract": "Traditionally, data-based soft sensors are constructed upon the labeled historical dataset which contains equal numbers of input and output data samples. While it is easy to obtain input variables such as temperature, pressure, and flow rate in the chemical process, the output variables, which correspond to quality/key property variables, are much more difficult to obtain. Therefore, we may only have a small number of output data samples, and have much more input data samples. In this article, a mixture form of the semisupervised probabilistic principal component regression model is proposed for soft sensor application, which can efficiently incorporate the unlabeled data information from different operation modes. Compared to the total supervised method, both modeling efficiency and soft sensing performance are improved with the inclusion of additional unlabeled data samples. Two case studies are provided to evaluate the feasibility and efficiency of the new method. \u00a9 2013 American Institute of Chemical Engineers AIChE J 60: 533\u2013545, 2014"
                },
                {
                    "title": "Stochastic variational inference",
                    "abstract": "We develop stochastic variational inference, a scalable algorithm for approximating posterior distributions. We develop this technique for a large class of probabilistic models and we demonstrate it with two probabilistic topic models, latent Dirichlet allocation and the hierarchical Dirichlet process topic model. Using stochastic variational inference, we analyze several large collections of documents: 300K articles from Nature, 1.8M articles from The New York Times, and 3.8M articles from Wikipedia. Stochastic inference can easily handle data sets of this size and outperforms traditional variational inference, which can only handle a smaller subset. (We also show that the Bayesian nonparametric topic model outperforms its parametric counterpart.) Stochastic variational inference lets us apply complex Bayesian models to massive data sets."
                },
                {
                    "title": "Fast Variational Inference in the Conjugate Exponential Family",
                    "abstract": "We present a general method for deriving collapsed variational inference algorithms for probabilistic models in the conjugate exponential family. Our method unifies many existing approaches to collapsed variational inference. Our collapsed variational inference leads to a new lower bound on the marginal likelihood. We exploit the information geometry of the bound to derive much faster optimization methods based on conjugate gradients for these models. Our approach is very general and is easily applied to any model where the mean field update equations have been derived. Empirically we show significant speed-ups for probabilistic inference using our bound."
                },
                {
                    "title": "Nonlinear Soft Sensor Development Based on Relevance Vector Machine",
                    "abstract": "This paper proposes an effective nonlinear soft sensor based on relevance vector machine (RVM), which was originally proposed in the machine learning area. Compared to the widely used support vector machine (SVM) and least-squares support vector machine (LSSVM) based soft sensors, RVM gives a more sparse model structure, which can greatly reduce computational complexity for online prediction. While SVM/LSSVM can only provide a point estimation of the prediction result, RVM gives a probabilistic prediction result, which is more sophisticated for the soft sensor application. Furthermore, RVM can successfully avoid several drawbacks of the traditional support vector machine type method, such as kernel function limitation, parameter tuning complexity, and etc. Due to the advantages of RVM, a practical application of this method is made for soft sensor modeling in this paper. To evaluate the performance of the developed soft sensor, two case studies are demonstrated, which both support that RVM performs much better than other methods for soft sensing."
                },
                {
                    "title": "Natural Gradient Works Efficiently in Learning",
                    "abstract": "When a parameter space has a certain underlying structure, the ordinary gradient of a function does not represent its steepest direction, but the natural gradient does. Information geometry is used for calculating the natural gradients in the parameter space of perceptrons, the space of matrices (for blind source separation), and the space of linear dynamical systems (for blind source deconvolution). The dynamical behavior of natural gradient online learning is analyzed and is proved to be Fisher efficient, implying that it has asymptotically the same performance as the optimal batch estimation of parameters. This suggests that the plateau phenomenon, which appears in the backpropagation learning algorithm of multilayer perceptrons, might disappear or might not be so serious when the natural gradient is used. An adaptive method of updating the learning rate is proposed and analyzed."
                },
                {
                    "title": "A 3-D Convolution-Based Burn-Through Point Multistep Prediction Model for Sintering Process",
                    "abstract": "Predicting burn-through point (BTP) in advance is a quite critical task for the sintering process. However, sintering is a complex physicochemical reaction process, and the strong spatial\u2013temporal correlations of data make the multistep prediction task very challenging. The previous BTP multistep prediction model only extracts spatial features in the high-level layers, leaving the spatial features in the low-level layers not learned. Specifically, the previous model only considers the relationships between the process variables and BTP, ignoring the spatial coupling relations among process variables. Further, the existing loss function is mainly based on Euclidean distance, which cannot learn dynamic information of multistep prediction sequence. To tackle these problems, in this article, we propose a 3-D convolution-based BTP multistep prediction model to simultaneously capture spatial\u2013temporal features. First, the 3-D convolution is employed to capture the spatial\u2013temporal features from low-level to high-level layers at the same time. Second, a spatial\u2013temporal recalibration block is proposed to further refine the extracted features to increase the contributions of informative features and suppress the less useful ones. Finally, we design a time-aware multistep prediction loss function to dynamically weigh the similarity between the actual sequence and the predicted sequence. The experimental results on two real-world BTP datasets demonstrate the effectiveness and feasibility of the proposed model on the BTP multistep prediction task."
                },
                {
                    "title": "Broad Learning System: An Effective and Efficient Incremental Learning System Without the Need for Deep Architecture",
                    "abstract": "Broad Learning System (BLS) that aims to offer an alternative way of learning in deep structure is proposed in this paper. Deep structure and learning suffer from a time-consuming training process because of a large number of connecting parameters in filters and layers. Moreover, it encounters a complete retraining process if the structure is not sufficient to model the system. The BLS is established in the form of a flat network, where the original inputs are transferred and placed as \u201cmapped features\u201d in feature nodes and the structure is expanded in wide sense in the \u201cenhancement nodes.\u201d The incremental learning algorithms are developed for fast remodeling in broad expansion without a retraining process if the network deems to be expanded. Two incremental learning algorithms are given for both the increment of the feature nodes (or filters in deep structure) and the increment of the enhancement nodes. The designed model and algorithms are very versatile for selecting a model rapidly. In addition, another incremental learning is developed for a system that has been modeled encounters a new incoming input. Specifically, the system can be remodeled in an incremental way without the entire retraining from the beginning. Satisfactory result for model reduction using singular value decomposition is conducted to simplify the final structure. Compared with existing deep neural networks, experimental results on the Modified National Institute of Standards and Technology database and NYU NORB object recognition dataset benchmark data demonstrate the effectiveness of the proposed BLS."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a stable and efficient neural network architecture for soft sensor applications in industrial settings that meets the stringent requirements for immediacy and stability during online deployment?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the gap between advanced deep learning techniques and their practical applicability in industrial environments. By enhancing the stability and performance of soft sensors, this research could lead to safer and more efficient factory operations, ultimately impacting economic outcomes and operational reliability. Furthermore, advancements in this area could inspire future research on robust machine learning models that can adapt to dynamic environments, paving the way for broader applications in various fields such as autonomous systems, real-time monitoring, and predictive maintenance.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need for neural networks to maintain performance stability during online updates, which is critical in industrial applications. Naive approaches may fail because they do not account for the unique constraints of soft sensor environments, such as the requirement for constant learning rates and the inability to use early stopping techniques. Additionally, the complexity of accurately modeling the likelihood distribution of inputs and outputs complicates the application of the Minimum Variance Estimator (MVE) in neural networks. The computational overhead associated with second-order optimization methods, like Natural Gradient Descent (NGD), further complicates the implementation of effective training strategies.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on linear and convex methods, which lack the expressive power needed for complex regression tasks in soft sensor applications. Existing solutions have not adequately addressed the specific requirements of stability and immediacy in online settings. Barriers such as the computational challenges of implementing NGD and the limitations of traditional optimization techniques have hindered progress. Our approach aims to bridge these gaps by integrating advanced neural network architectures with robust optimization strategies tailored for the unique demands of soft sensors, thus improving upon prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a novel neural network architecture specifically designed for soft sensor applications, utilizing a constant learning rate and real-time updates to ensure stability. We will employ a dataset derived from industrial processes to evaluate our model's performance, using metrics such as Mean Squared Error (MSE) and stability indices to assess its effectiveness. The expected outcomes include a"
            }
        },
        "author_data": {
            "2f0dd2d8-ecd4-4c64-8e2e-ce5db96c8e39": {
                "pk": "2f0dd2d8-ecd4-4c64-8e2e-ce5db96c8e39",
                "name": "Shaoqi Wang",
                "collaborators": [
                    "Aidi Pi",
                    "Junxian Zhao",
                    "Xiaobo Zhou"
                ],
                "domain": [
                    "Memory Management",
                    "Datacenter Systems",
                    "Performance Optimization",
                    "Multi-Tenancy"
                ],
                "publications": [
                    {
                        "title": "Memory at Your Service: Fast Memory Allocation for Latency-critical Services",
                        "abstract": "Co-location and memory sharing between latency-critical services, such as key-value store and web search, and best-effort batch jobs is an appealing approach to improving memory utilization in multi-tenant datacenter systems. However, we find that the very diverse goals of job co-location and the GNU/Linux system stack can lead to severe performance degradation of latency-critical services under memory pressure in a multi-tenant system. We address memory pressure for latency-critical services via fast memory allocation and proactive reclamation. We find that memory allocation latency dominates the overall query latency, especially under memory pressure. We analyze the default memory management mechanism provided by GNU/Linux system stack and identify the reasons why it is inefficient for latency-critical services in a multi-tenant system. We present Hermes, a fast memory allocation mechanism in user space that adaptively reserves memory for latency-critical services. It advises Linux OS to proactively reclaim memory of batch jobs. We implement Hermes in GNU C Library. Experimental result shows that Hermes reduces the average and the $99^{th}$ percentile memory allocation latency by up to 54.4% and 62.4% for a micro benchmark, respectively. For two real-world latency-critical services, Hermes reduces both the average and the $99^{th}$ percentile tail query latency by up to 40.3%. Compared to the default Glibc, jemalloc and TCMalloc, Hermes reduces Service Level Objective violation by up to 84.3% under memory pressure."
                    }
                ]
            },
            "990d39ca-033f-46bb-b440-fd090757f8d1": {
                "pk": "990d39ca-033f-46bb-b440-fd090757f8d1",
                "name": "Chunjie Yang",
                "collaborators": [
                    "Junjie Yu",
                    "Xing Wang",
                    "Jiangjiang Zhao",
                    "Wenliang Chen"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Relation Extraction",
                    "Self-Training"
                ],
                "publications": [
                    {
                        "title": "STAD: Self-Training with Ambiguous Data for Low-Resource Relation Extraction",
                        "abstract": "We present a simple yet effective self-training approach, named as STAD, for low-resource relation extraction. The approach first classifies the auto-annotated instances into two groups: confident instances and uncertain instances, according to the probabilities predicted by a teacher model. In contrast to most previous studies, which mainly only use the confident instances for self-training, we make use of the uncertain instances. To this end, we propose a method to identify ambiguous but useful instances from the uncertain instances and then divide the relations into candidate-label set and negative-label set for each ambiguous instance. Next, we propose a set-negative training method on the negative-label sets for the ambiguous instances and a positive training method for the confident instances. Finally, a joint-training method is proposed to build the final relation extraction system on all data. Experimental results on two widely used datasets SemEval2010 Task-8 and Re-TACRED with low-resource settings demonstrate that this new self-training approach indeed achieves significant and consistent improvements when comparing to several competitive self-training systems. Code is publicly available at https://github.com/jjyunlp/STAD"
                    }
                ]
            },
            "2af43ab9-af93-4220-bab4-b6f3eb6d8a24": {
                "pk": "2af43ab9-af93-4220-bab4-b6f3eb6d8a24",
                "name": "Siwei Lou",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2405.14394": {
        "paper_data": {
            "title": "Instruction Tuning With Loss Over Instructions",
            "url": "http://arxiv.org/abs/2405.14394v2",
            "arxiv_id": "2405.14394",
            "authors": [
                "Zhengyan Shi",
                "Adam X. Yang",
                "Bin Wu",
                "Laurence Aitchison",
                "Emine Yilmaz",
                "Aldo Lipani"
            ],
            "abstract": "Instruction tuning plays a crucial role in shaping the outputs of language models (LMs) to desired styles. In this work, we propose a simple yet effective method, Instruction Modelling (IM), which trains LMs by applying a loss function to the instruction and prompt part rather than solely to the output part. Through experiments across 21 diverse benchmarks, we show that, in many scenarios, IM can effectively improve the LM performance on both NLP tasks (e.g., MMLU, TruthfulQA, and HumanEval) and open-ended generation benchmarks (e.g., MT-Bench and AlpacaEval). Remarkably, in the most advantageous case, IM boosts model performance on AlpacaEval 1.0 by over 100%. We identify two key factors influencing the effectiveness of IM: (1) The ratio between instruction length and output length in the training data; and (2) The number of training examples. We observe that IM is especially beneficial when trained on datasets with lengthy instructions paired with brief outputs, or under the Superficial Alignment Hypothesis (SAH) where a small amount of training examples are used for instruction tuning. Further analysis substantiates our hypothesis that our improvement can be attributed to reduced overfitting to instruction tuning datasets. It is worth noting that we are not proposing \\ours as a replacement for current fine-tuning processes. Instead, our work aims to provide practical guidance for instruction tuning LMs, especially in low-resource scenarios.",
            "introduction": "   1 Introduction  Figure 1:  Performance differences between Instruction Tuning (IT) and our proposed method instruction modelling (IM) trained on 7 instruction tuning datasets. These datasets contain prompts and responses but do not contain preference pairs. Specifically, we use the Less datasets [68] and Alpagasus datasets [11], which are subsets of Flan V2 [14], Dolly [18], and Stanford Alpaca [61] to ensure good performance. We also report the results on the LIMA dataset. (Left) The mean performance across 18 traditional NLP tasks (see \u00a74.1 for details). (Right) The win rate on the AlpacaEval 1.0 benchmark [37]. Please refer to \u00a74.2 for details.    Language models (LMs) are trained to predict the next token on massive corpora, enabling them to learn general-purpose representations transferable to various language understanding or generation tasks [7, 47, 48, 56, 62]. However, it does align LMs to act in accordance with the user\u2019s intentions [34]. To enable this transfer, various methods for aligning language models [3, 14, 44, 73, 58, 79] have thus been proposed, one of which is instruction tuning (IT) [36, 65, 69]. Recent study [78] proposes Superficial Alignment Hypothesis (SAH): A model\u2019s knowledge and capabilities are learnt almost entirely during pretraining, only minimal instruction tuning data is required to enable high-quality outputs in the desired output style. Existing works [1, 43, 44, 51, 65, 69, 36] mainly perform instruction tuning by focusing the loss computation solely on the output segments.   Figure 2:  (Left) Performance improvement, achieved by our approach instruction modelling (IM) compared to Instruction Tuning (IT) on the AlpacaEval 1.0, against the ratio between average instruction length and average output length in instruction tuning datasets (training size noted in parentheses). We highlight several representative instruction tuning datasets in yellow. Our analysis suggests that IM is especially beneficial for datasets characterized by lengthy instructions or prompts paired with comparably brief outputs, such as Code Alpaca [10] and Less MMLU Chat [68]. (Right) Performance improvement achieved by our approach IM over IT on the AlpacaEval 1.0 against the number of training examples in instruction tuning datasets. Here we maintain a fixed ratio between instruction and output length of 10. This analysis suggests that IM is particularly effective under the low-resource setting or Superficial Alignment Hypothesis. Please refer to \u00a74.2 for details.    In this work, we demonstrate that in many scenarios, incorporating the loss computation for instructions or prompts, which we refer to as Instruction Modelling (IM) (see \u00a73), could substantially improve the performance of instruction tuning on both various NLP tasks (e.g., MMLU, TruthfulQA, and HumanEval) and open-ended generation benchmarks (e.g., MT-Bench and AlpacaEval), as shown in Figure 1. Remarkably, in the most favourable case, our proposed method IM boosts performance on AlpacaEval 1.0 by over 100%. As illustrated in Figure 2, our study further identifies two key factors influencing the effectiveness of IM: (1) The ratio between instruction length and output length (see Figure 2 Left). Our analysis shows that our approach IM is especially beneficial for datasets characterised by lengthy instructions or prompts paired with comparably brief outputs, such as Code Alpaca [13] and Less MMLU Chat [68]; (2) The number of training examples (see Figure 2 Right). We demonstrate that our approach IM performs better under the SAH,",
            "references": [
                {
                    "title": "Understanding Likelihood Over-optimisation in Direct Alignment Algorithms",
                    "abstract": "Direct Alignment Algorithms (DAAs), such as Direct Preference Optimisation (DPO) and Identity Preference Optimisation (IPO), have emerged as alternatives to online Reinforcement Learning from Human Feedback (RLHF) algorithms such as Proximal Policy Optimisation (PPO) for aligning language models to human preferences, without the need for explicit reward modelling. These methods generally aim to increase the likelihood of generating better (preferred) completions while discouraging worse (non-preferred) ones, while staying close to the original model's behaviour. In this work, we explore the relationship between completion likelihood and model performance in state-of-the-art DAAs, and identify a critical issue of likelihood over-optimisation. Contrary to expectations, we find that higher likelihood of better completions and larger margins between better and worse completion likelihoods do not necessarily lead to better performance, and may even degrade it. Our analysis reveals that while higher likelihood correlates with better memorisation of factual knowledge patterns, a slightly lower completion likelihood tends to improve output diversity, thus leading to better generalisation to unseen scenarios. Moreover, we identify two key indicators that signal when over-optimised output diversity begins to harm performance: Decreasing Entropy over Top-k Tokens and Diminishing Top-k Probability Mass. Our experimental results validate that these indicators are reliable signs of declining performance under different regularisations, helping prevent over-optimisation and improve alignment with human preferences."
                },
                {
                    "title": "Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuning",
                    "abstract": "Datasets are foundational to many breakthroughs in modern artificial intelligence. Many recent achievements in the space of natural language processing (NLP) can be attributed to the finetuning of pre-trained models on a diverse set of tasks that enables a large language model (LLM) to respond to instructions. Instruction fine-tuning (IFT) requires specifically constructed and annotated datasets. However, existing datasets are almost all in the English language. In this work, our primary goal is to bridge the language gap by building a human-curated instruction-following dataset spanning 65 languages. We worked with fluent speakers of languages from around the world to collect natural instances of instructions and completions. Furthermore, we create the most extensive multilingual collection to date, comprising 513 million instances through templating and translating existing datasets across 114 languages. In total, we contribute four key resources: we develop and open-source the Aya Annotation Platform, the Aya Dataset, the Aya Collection, and the Aya Evaluation Suite. The Aya initiative also serves as a valuable case study in participatory research, involving collaborators from 119 countries. We see this as a valuable framework for future research collaborations that aim to bridge gaps in resources."
                },
                {
                    "title": "Long Is More for Alignment: A Simple but Tough-to-Beat Baseline for Instruction Fine-Tuning",
                    "abstract": "There is a consensus that instruction fine-tuning of LLMs requires high-quality data, but what are they? LIMA (NeurIPS 2023) and AlpaGasus (ICLR 2024) are state-of-the-art methods for selecting such high-quality examples, either via manual curation or using GPT-3.5-Turbo as a quality scorer. We show that the extremely simple baseline of selecting the 1,000 instructions with longest responses -- that intuitively contain more learnable information and are harder to overfit -- from standard datasets can consistently outperform these sophisticated methods according to GPT-4 and PaLM-2 as judges, while remaining competitive on the Open LLM benchmarks that test factual knowledge. We demonstrate this for several LLMs (Llama-2-7B, Llama-2-13B, Mistral-7B-v0.1) and datasets (Alpaca-52k, Evol-Instruct-70k). In addition, a lightweight refinement of such long instructions can further improve the abilities of the fine-tuned LLMs, and allows us to obtain competitive results on MT-Bench and the 2nd highest-ranked Llama-2-7B-based model on AlpacaEval 2.0, while training on only 1,000 examples and no extra preference data. We also conduct a thorough analysis of our models to ensure that their enhanced performance is not simply due to GPT-4's preference for longer responses. Overall, our findings suggest that fine-tuning on the longest responses should be the default baseline for any work on instruction fine-tuning. We provide our code at https://github.com/tml-epfl/long-is-more-for-alignment."
                },
                {
                    "title": "Instruction Fine-Tuning: Does Prompt Loss Matter?",
                    "abstract": "We present a novel study analyzing the effects of various prompt loss token weights (PLW) for supervised instruction fine-tuning (SIFT). While prompt-masking (PLW = 0) is common for SIFT, some fine-tuning APIs support fractional PLWs and suggest that using a small non-zero PLW can help stabilize learning when fine-tuning on short-completion data. However, there has never been a study confirming this claim, and OpenAI, a major cloud-based SIFT provider, recently removed this parameter from their fine-tuning API. We found that performance of models fine-tuned on short-completion data had a statistically-significant negative quadratic relationship with PLW. Using small values (0.01 - 0.5) of PLW produced better results on multiple-choice and short-generation benchmarks (outperforming models fine-tuned on long-completion data) while large values (~ 1.0) of PLW produced better results on long-generation benchmarks. We explained this effect and verified its importance through additional experiments. This research serves as a warning to API providers about the importance of providing a PLW parameter for SIFT."
                },
                {
                    "title": "What Makes Good Data for Alignment? A Comprehensive Study of Automatic Data Selection in Instruction Tuning",
                    "abstract": "Instruction tuning is a standard technique employed to align large language models to end tasks and user preferences after the initial pretraining phase. Recent research indicates the critical role of data engineering in instruction tuning -- when appropriately selected, only limited data is necessary to achieve superior performance. However, we still lack a principled understanding of what makes good instruction tuning data for alignment, and how we should select data automatically and effectively. In this work, we delve deeply into automatic data selection strategies for alignment. We start with controlled studies to measure data across three dimensions: complexity, quality, and diversity, along which we examine existing methods and introduce novel techniques for enhanced data measurement. Subsequently, we propose a simple strategy to select data samples based on the measurement. We present deita (short for Data-Efficient Instruction Tuning for Alignment), a series of models fine-tuned from LLaMA and Mistral models using data samples automatically selected with our proposed approach. Empirically, deita performs better or on par with the state-of-the-art open-source alignment models with only 6K SFT training data samples -- over 10x less than the data used in the baselines. When further trained with direct preference optimization (DPO), deita-Mistral-7B + DPO trained with 6K SFT and 10K DPO samples achieve 7.55 MT-Bench and 90.06% AlpacaEval scores. We anticipate this work to provide tools on automatic data selection, facilitating data-efficient alignment. We release our models as well as the selected datasets for future researches to effectively align models more efficiently."
                },
                {
                    "title": "Rethinking the Instruction Quality: LIFT is What You Need",
                    "abstract": "Instruction tuning, a specialized technique to enhance large language model (LLM) performance via instruction datasets, relies heavily on the quality of employed data. Existing quality improvement methods alter instruction data through dataset expansion or curation. However, the expansion method risks data redundancy, potentially compromising LLM performance, while the curation approach confines the LLM's potential to the original dataset. Our aim is to surpass the original data quality without encountering these shortcomings. To achieve this, we propose LIFT (LLM Instruction Fusion Transfer), a novel and versatile paradigm designed to elevate the instruction quality to new heights. LIFT strategically broadens data distribution to encompass more high-quality subspaces and eliminates redundancy, concentrating on high-quality segments across overall data subspaces. Experimental results demonstrate that, even with a limited quantity of high-quality instruction data selected by our paradigm, LLMs not only consistently uphold robust performance across various tasks but also surpass some state-of-the-art results, highlighting the significant improvement in instruction quality achieved by our paradigm."
                },
                {
                    "title": "LIMIT: Less Is More for Instruction Tuning Across Evaluation Paradigms",
                    "abstract": "Large Language Models are traditionally finetuned on large instruction datasets. However recent studies suggest that small, high-quality datasets can suffice for general purpose instruction following. This lack of consensus surrounding finetuning best practices is in part due to rapidly diverging approaches to LLM evaluation. In this study, we ask whether a small amount of diverse finetuning samples can improve performance on both traditional perplexity-based NLP benchmarks, and on open-ended, model-based evaluation. We finetune open-source MPT-7B and MPT-30B models on instruction finetuning datasets of various sizes ranging from 1k to 60k samples. We find that subsets of 1k-6k instruction finetuning samples are sufficient to achieve good performance on both (1) traditional NLP benchmarks and (2) model-based evaluation. Finally, we show that mixing textbook-style and open-ended QA finetuning datasets optimizes performance on both evaluation paradigms."
                },
                {
                    "title": "NEFTune: Noisy Embeddings Improve Instruction Finetuning",
                    "abstract": "We show that language model finetuning can be improved, sometimes dramatically, with a simple augmentation. NEFTune adds noise to the embedding vectors during training. Standard finetuning of LLaMA-2-7B using Alpaca achieves 29.79% on AlpacaEval, which rises to 64.69% using noisy embeddings. NEFTune also improves over strong baselines on modern instruction datasets. Models trained with Evol-Instruct see a 10% improvement, with ShareGPT an 8% improvement, and with OpenPlatypus an 8% improvement. Even powerful models further refined with RLHF such as LLaMA-2-Chat benefit from additional training with NEFTune."
                },
                {
                    "title": "Batch Calibration: Rethinking Calibration for In-Context Learning and Prompt Engineering",
                    "abstract": "Prompting and in-context learning (ICL) have become efficient learning paradigms for large language models (LLMs). However, LLMs suffer from prompt brittleness and various bias factors in the prompt, including but not limited to the formatting, the choice verbalizers, and the ICL examples. To address this problem that results in unexpected performance degradation, calibration methods have been developed to mitigate the effects of these biases while recovering LLM performance. In this work, we first conduct a systematic analysis of the existing calibration methods, where we both provide a unified view and reveal the failure cases. Inspired by these analyses, we propose Batch Calibration (BC), a simple yet intuitive method that controls the contextual bias from the batched input, unifies various prior approaches, and effectively addresses the aforementioned issues. BC is zero-shot, inference-only, and incurs negligible additional costs. In the few-shot setup, we further extend BC to allow it to learn the contextual bias from labeled data. We validate the effectiveness of BC with PaLM 2-(S, M, L) and CLIP models and demonstrate state-of-the-art performance over previous calibration baselines across more than 10 natural language understanding and image classification tasks."
                },
                {
                    "title": "Human Feedback is not Gold Standard",
                    "abstract": "Human feedback has become the de facto standard for evaluating the performance of Large Language Models, and is increasingly being used as a training objective. However, it is not clear which properties of a generated output this single `preference' score captures. We hypothesise that preference scores are subjective and open to undesirable biases. We critically analyse the use of human feedback for both training and evaluation, to verify whether it fully captures a range of crucial error criteria. We find that while preference scores have fairly good coverage, they under-represent important aspects like factuality. We further hypothesise that both preference scores and error annotation may be affected by confounders, and leverage instruction-tuned models to generate outputs that vary along two possible confounding dimensions: assertiveness and complexity. We find that the assertiveness of an output skews the perceived rate of factuality errors, indicating that human annotations are not a fully reliable evaluation metric or training objective. Finally, we offer preliminary evidence that using human feedback as a training objective disproportionately increases the assertiveness of model outputs. We encourage future work to carefully consider whether preference scores are well aligned with the desired objective."
                },
                {
                    "title": "DePT: Decomposed Prompt Tuning for Parameter-Efficient Fine-tuning",
                    "abstract": "Prompt tuning (PT), where a small amount of trainable soft (continuous) prompt vectors is affixed to the input of language models (LM), has shown promising results across various tasks and models for parameter-efficient fine-tuning (PEFT). PT stands out from other PEFT approaches because it maintains competitive performance with fewer trainable parameters and does not drastically scale up its parameters as the model size expands. However, PT introduces additional soft prompt tokens, leading to longer input sequences, which significantly impacts training and inference time and memory usage due to the Transformer's quadratic complexity. Particularly concerning for Large Language Models (LLMs) that face heavy daily querying. To address this issue, we propose Decomposed Prompt Tuning (DePT), which decomposes the soft prompt into a shorter soft prompt and a pair of low-rank matrices that are then optimised with two different learning rates. This allows DePT to achieve better performance while saving substantial memory and time costs compared to vanilla PT and its variants, without changing trainable parameter sizes. Through extensive experiments on 23 natural language processing (NLP) and vision-language (VL) tasks, we demonstrate that DePT outperforms state-of-the-art PEFT approaches, including the full fine-tuning baseline, in some scenarios. Additionally, we empirically show that DEPT grows more efficient as the model size increases. Our further study reveals that DePT integrates seamlessly with parameter-efficient transfer learning in the few-shot learning setting and highlights its adaptability to various model architectures and sizes."
                },
                {
                    "title": "Bayesian low-rank adaptation for large language models",
                    "abstract": "Low-rank adaptation (LoRA) has emerged as a new paradigm for cost-efficient fine-tuning of large language models (LLMs). However, fine-tuned LLMs often become overconfident especially when fine-tuned on small datasets. Bayesian methods, with their inherent ability to estimate uncertainty, serve as potent tools to mitigate overconfidence and enhance calibration. In this work, we introduce Laplace-LoRA, which applies a Bayesian approach to the LoRA parameters. Specifically, Laplace-LoRA applies a Laplace approximation to the posterior over the LoRA parameters, considerably improving the calibration of fine-tuned LLMs."
                },
                {
                    "title": "Self-Alignment with Instruction Backtranslation",
                    "abstract": "We present a scalable method to build a high quality instruction following language model by automatically labelling human-written text with corresponding instructions. Our approach, named instruction backtranslation, starts with a language model finetuned on a small amount of seed data, and a given web corpus. The seed model is used to construct training examples by generating instruction prompts for web documents (self-augmentation), and then selecting high quality examples from among these candidates (self-curation). This data is then used to finetune a stronger model. Finetuning LLaMa on two iterations of our approach yields a model that outperforms all other LLaMa-based models on the Alpaca leaderboard not relying on distillation data, demonstrating highly effective self-alignment."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning",
                    "abstract": "Scaling Transformers to longer sequence lengths has been a major problem in the last several years, promising to improve performance in language modeling and high-resolution image understanding, as well as to unlock new applications in code, audio, and video generation. The attention layer is the main bottleneck in scaling to longer sequences, as its runtime and memory increase quadratically in the sequence length. FlashAttention exploits the asymmetric GPU memory hierarchy to bring significant memory saving (linear instead of quadratic) and runtime speedup (2-4$\\times$ compared to optimized baselines), with no approximation. However, FlashAttention is still not nearly as fast as optimized matrix-multiply (GEMM) operations, reaching only 25-40\\% of the theoretical maximum FLOPs/s. We observe that the inefficiency is due to suboptimal work partitioning between different thread blocks and warps on the GPU, causing either low-occupancy or unnecessary shared memory reads/writes. We propose FlashAttention-2, with better work partitioning to address these issues. In particular, we (1) tweak the algorithm to reduce the number of non-matmul FLOPs (2) parallelize the attention computation, even for a single head, across different thread blocks to increase occupancy, and (3) within each thread block, distribute the work between warps to reduce communication through shared memory. These yield around 2$\\times$ speedup compared to FlashAttention, reaching 50-73\\% of the theoretical maximum FLOPs/s on A100 and getting close to the efficiency of GEMM operations. We empirically validate that when used end-to-end to train GPT-style models, FlashAttention-2 reaches training speed of up to 225 TFLOPs/s per A100 GPU (72\\% model FLOPs utilization)."
                },
                {
                    "title": "Textbooks Are All You Need",
                    "abstract": "We introduce phi-1, a new large language model for code, with significantly smaller size than competing models: phi-1 is a Transformer-based model with 1.3B parameters, trained for 4 days on 8 A100s, using a selection of ``textbook quality\"data from the web (6B tokens) and synthetically generated textbooks and exercises with GPT-3.5 (1B tokens). Despite this small scale, phi-1 attains pass@1 accuracy 50.6% on HumanEval and 55.5% on MBPP. It also displays surprising emergent properties compared to phi-1-base, our model before our finetuning stage on a dataset of coding exercises, and phi-1-small, a smaller model with 350M parameters trained with the same pipeline as phi-1 that still achieves 45% on HumanEval."
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "Preserving Pre-trained Features Helps Calibrate Fine-tuned Language Models",
                    "abstract": "Large pre-trained language models (PLMs) have demonstrated strong performance on natural language understanding (NLU) tasks through fine-tuning. However, fine-tuned models still suffer from overconfident predictions, especially in out-of-domain settings. In this paper, we tackle the problem of calibrating fine-tuned language models. We demonstrate that the PLMs are well-calibrated on the masked language modeling task with robust predictive confidence under domain shift, yet the fine-tuned models fail to retain such property due to catastrophic forgetting, which impacts the calibration on the downstream classification task. In light of these observations, we evaluate the calibration of several methods that preserve pre-trained features and show that preserving pre-trained features can improve the calibration of fine-tuned language models. Among these methods, our proposed method that encourages the fine-tuned model to learn generative representations with auxiliary language modeling objective achieves competitive accuracy and the lowest expected calibration error compared to several strong baselines under both in-domain and out-of-domain settings on three downstream NLU tasks."
                },
                {
                    "title": "Rethinking Semi-supervised Learning with Language Models",
                    "abstract": "Semi-supervised learning (SSL) is a popular setting aiming to effectively utilize unlabelled data to improve model performance in downstream natural language processing (NLP) tasks. Currently, there are two popular approaches to make use of unlabelled data: Self-training (ST) and Task-adaptive pre-training (TAPT). ST uses a teacher model to assign pseudo-labels to the unlabelled data, while TAPT continues pre-training on the unlabelled data before fine-tuning. To the best of our knowledge, the effectiveness of TAPT in SSL tasks has not been systematically studied, and no previous work has directly compared TAPT and ST in terms of their ability to utilize the pool of unlabelled data. In this paper, we provide an extensive empirical study comparing five state-of-the-art ST approaches and TAPT across various NLP tasks and data sizes, including in- and out-of-domain settings. Surprisingly, we find that TAPT is a strong and more robust SSL learner, even when using just a few hundred unlabelled samples or in the presence of domain shifts, compared to more sophisticated ST approaches, and tends to bring greater improvements in SSL than in fully-supervised settings. Our further analysis demonstrates the risks of using ST approaches when the size of labelled or unlabelled data is small or when domain shifts exist. We offer a fresh perspective for future SSL research, suggesting the use of unsupervised pre-training objectives over dependency on pseudo labels."
                },
                {
                    "title": "To Repeat or Not To Repeat: Insights from Scaling LLM under Token-Crisis",
                    "abstract": "Recent research has highlighted the importance of dataset size in scaling language models. However, large language models (LLMs) are notoriously token-hungry during pre-training, and high-quality text data on the web is approaching its scaling limit for LLMs. To further enhance LLMs, a straightforward approach is to repeat the pre-training data for additional epochs. In this study, we empirically investigate three key aspects under this approach. First, we explore the consequences of repeating pre-training data, revealing that the model is susceptible to overfitting, leading to multi-epoch degradation. Second, we examine the key factors contributing to multi-epoch degradation, finding that significant factors include dataset size, model parameters, and training objectives, while less influential factors consist of dataset quality and model FLOPs. Finally, we explore whether widely used regularization can alleviate multi-epoch degradation. Most regularization techniques do not yield significant improvements, except for dropout, which demonstrates remarkable effectiveness but requires careful tuning when scaling up the model size. Additionally, we discover that leveraging mixture-of-experts (MoE) enables cost-effective and efficient hyper-parameter tuning for computationally intensive dense LLMs with comparable trainable parameters, potentially impacting efficient LLM development on a broader scale."
                },
                {
                    "title": "Don't Stop Pretraining? Make Prompt-based Fine-tuning Powerful Learner",
                    "abstract": "Language models (LMs) trained on vast quantities of unlabelled data have greatly advanced the field of natural language processing (NLP). In this study, we re-visit the widely accepted notion in NLP that continued pre-training LMs on task-related texts improves the performance of fine-tuning (FT) in downstream tasks. Through experiments on eight single-sentence tasks and eight sentence-pair tasks in both semi-supervised and fully-supervised settings, we find that conventional continued pre-training does not consistently provide benefits and can even be detrimental for sentence-pair tasks or when prompt-based FT is used. To tackle these issues, we propose Prompt-based Continued Pre-training (PCP), which combines the idea of instruction tuning with conventional continued pre-training. Our approach aims to improve the performance of prompt-based FT by presenting both task-related texts and prompt templates to LMs through unsupervised pre-training objectives before fine-tuning for the target task. Our empirical evaluations on 21 benchmarks demonstrate that the PCP consistently improves the performance of state-of-the-art prompt-based FT approaches (up to 20.1% absolute) in both semi-supervised and fully-supervised settings, even with only hundreds of unlabelled examples. Additionally, prompt-based FT with the PCP outperforms state-of-the-art semi-supervised approaches with greater simplicity, eliminating the need for an iterative process and extra data augmentation. Our further analysis explores the performance lower bound of the PCP and reveals that the advantages of PCP persist across different sizes of models and datasets."
                },
                {
                    "title": "Compositional Zero-Shot Domain Transfer with Text-to-Text Models",
                    "abstract": "Abstract Label scarcity is a bottleneck for improving task performance in specialized domains. We propose a novel compositional transfer learning framework (DoT51) for zero-shot domain transfer. Without access to in-domain labels, DoT5 jointly learns domain knowledge (from masked language modelling of unlabelled in-domain free text) and task knowledge (from task training on more readily available general-domain data) in a multi-task manner. To improve the transferability of task training, we design a strategy named NLGU: We simultaneously train natural language generation (NLG) for in-domain label-to-data generation, which enables data augmentation for self-finetuning and natural language understanding (NLU) for label prediction. We evaluate DoT5 on the biomedical domain and the resource-lean subdomain of radiology, focusing on natural language inference, text summarization, and embedding learning. DoT5 demonstrates the effectiveness of compositional transfer learning through multi-task learning. In particular, DoT5 outperforms the current state-of-the-art in zero-shot transfer by over 7 absolute points in accuracy on RadNLI. We validate DoT5 with ablations and a case study demonstrating its ability to solve challenging NLI examples requiring in-domain expertise."
                },
                {
                    "title": "AutoPEFT: Automatic Configuration Search for Parameter-Efficient Fine-Tuning",
                    "abstract": "Large pretrained language models are widely used in downstream NLP tasks via task- specific fine-tuning, but such procedures can be costly. Recently, Parameter-Efficient Fine-Tuning (PEFT) methods have achieved strong task performance while updating much fewer parameters than full model fine-tuning (FFT). However, it is non-trivial to make informed design choices on the PEFT configurations, such as their architecture, the number of tunable parameters, and even the layers in which the PEFT modules are inserted. Consequently, it is highly likely that the current, manually designed configurations are suboptimal in terms of their performance-efficiency trade-off. Inspired by advances in neural architecture search, we propose AutoPEFT for automatic PEFT configuration selection: We first design an expressive configuration search space with multiple representative PEFT modules as building blocks. Using multi-objective Bayesian optimization in a low-cost setup, we then discover a Pareto-optimal set of configurations with strong performance-cost trade-offs across different numbers of parameters that are also highly transferable across different tasks. Empirically, on GLUE and SuperGLUE tasks, we show that AutoPEFT-discovered configurations significantly outperform existing PEFT methods and are on par or better than FFT without incurring substantial training efficiency costs."
                },
                {
                    "title": "Self-Instruct: Aligning Language Models with Self-Generated Instructions",
                    "abstract": "Large \u201cinstruction-tuned\u201d language models (i.e., finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is often limited in quantity, diversity, and creativity, therefore hindering the generality of the tuned model. We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off their own generations. Our pipeline generates instructions, input, and output samples from a language model, then filters invalid or similar ones before using them to finetune the original model. Applying our method to the vanilla GPT3, we demonstrate a 33% absolute improvement over the original model on Super-NaturalInstructions, on par with the performance of InstructGPT-001, which was trained with private user data and human annotations. For further evaluation, we curate a set of expert-written instructions for novel tasks, and show through human evaluation that tuning GPT3 with Self-Instruct outperforms using existing public instruction datasets by a large margin, leaving only a 5% absolute gap behind InstructGPT-001. Self-Instruct provides an almost annotation-free method for aligning pre-trained language models with instructions, and we release our large synthetic dataset to facilitate future studies on instruction tuning."
                },
                {
                    "title": "Unnatural Instructions: Tuning Language Models with (Almost) No Human Labor",
                    "abstract": "Instruction tuning enables pretrained language models to perform new tasks from inference-time natural language descriptions. These approaches rely on vast amounts of human supervision in the form of crowdsourced datasets or user interactions. In this work, we introduce Unnatural Instructions: a large dataset of creative and diverse instructions, collected with virtually no human labor. We collect 64,000 examples by prompting a language model with three seed examples of instructions and eliciting a fourth. This set is then expanded by prompting the model to rephrase each instruction, creating a total of approximately 240,000 examples of instructions, inputs, and outputs. Experiments show that despite containing a fair amount of noise, training on Unnatural Instructions rivals the effectiveness of training on open-source manually-curated datasets, surpassing the performance of models such as T0++ and Tk-Instruct across various benchmarks. These results demonstrate the potential of model-generated data as a cost-effective alternative to crowdsourcing for dataset expansion and diversification."
                },
                {
                    "title": "Scaling Instruction-Finetuned Language Models",
                    "abstract": "Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models."
                },
                {
                    "title": "Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them",
                    "abstract": "BIG-Bench (Srivastava et al., 2022) is a diverse evaluation suite that focuses on tasks believed to be beyond the capabilities of current language models. Language models have already made good progress on this benchmark, with the best model in the BIG-Bench paper outperforming average reported human-rater results on 65% of the BIG-Bench tasks via few-shot prompting. But on what tasks do language models fall short of average human-rater performance, and are those tasks actually unsolvable by current language models? In this work, we focus on a suite of 23 challenging BIG-Bench tasks which we call BIG-Bench Hard (BBH). These are the task for which prior language model evaluations did not outperform the average human-rater. We find that applying chain-of-thought (CoT) prompting to BBH tasks enables PaLM to surpass the average human-rater performance on 10 of the 23 tasks, and Codex (code-davinci-002) to surpass the average human-rater performance on 17 of the 23 tasks. Since many tasks in BBH require multi-step reasoning, few-shot prompting without CoT, as done in the BIG-Bench evaluations (Srivastava et al., 2022), substantially underestimates the best performance and capabilities of language models, which is better captured via CoT prompting. As further analysis, we explore the interaction between CoT and model scale on BBH, finding that CoT enables emergent task performance on several BBH tasks with otherwise flat scaling curves."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "Meta-Learning Helps Personalized Product Search",
                    "abstract": "Personalized product search that provides users with customized search services is an important task for e-commerce platforms. This task remains a challenge when inferring users\u2019 preferences from few records or even no records, which is also known as the few-shot or zero-shot learning problem. In this paper, we propose a Bayesian Online Meta-Learning Model (BOML), which transfers meta-knowledge, from the inference for other users\u2019 preferences, to help to infer the current user\u2019s interest behind her/his few or even no historical records. To extract meta-knowledge from various inference patterns, our model constructs a mixture of meta-knowledge and transfers the corresponding meta-knowledge to the specific user according to her/his records. Based on the meta-knowledge learned from other similar inferences, our proposed model searches a ranked list of products to meet users\u2019 personalized query intents for those with few search records (i.e., few-shot learning problem) or even no search records (i.e., zero-shot learning problem). Under the records arriving sequentially setting, we propose an online variational inference algorithm to update meta-knowledge over time. Experimental results demonstrate that our proposed BOML outperforms state-of-the-art algorithms."
                },
                {
                    "title": "Learning to Execute Actions or Ask Clarification Questions",
                    "abstract": "Collaborative tasks are ubiquitous activities where a form of communication is required in order to reach a joint goal. Collaborative building is one of such tasks. We wish to develop an intelligent builder agent in a simulated building environment (Minecraft) that can build whatever users wish to build by just talking to the agent. In order to achieve this goal, such agents need to be able to take the initiative by asking clarification questions when further information is needed. Existing works on Minecraft Corpus Dataset only learn to execute instructions neglecting the importance of asking for clarifications. In this paper, we extend the Minecraft Corpus Dataset by annotating all builder utterances into eight types, including clarification questions, and propose a new builder agent model capable of determining when to ask or execute instructions. Experimental results show that our model achieves state-of-the-art performance on the collaborative building task with a substantial improvement. We also define two new tasks, the learning to ask task and the joint learning task. The latter consists of solving both collaborating building and learning to ask tasks jointly."
                },
                {
                    "title": "StepGame: A New Benchmark for Robust Multi-Hop Spatial Reasoning in Texts",
                    "abstract": "Inferring spatial relations in natural language is a crucial ability an intelligent system should possess. The bAbI dataset tries to capture tasks relevant to this domain (task 17 and 19). However, these tasks have several limitations. Most importantly, they are limited to fixed expressions, they are limited in the number of reasoning steps required to solve them, and they fail to test the robustness of models to input that contains irrelevant or redundant information. In this paper, we present a new Question-Answering dataset called StepGame for robust multi-step spatial reasoning in texts. Our experiments demonstrate that state-of-the-art models on the bAbI dataset struggle on the StepGame dataset. Moreover, we propose a Tensor-Product based Memory-Augmented Neural Network (TP-MANN) specialized for spatial reasoning tasks. Experimental results on both datasets show that our model outperforms all the baselines with superior generalization and robustness performance."
                },
                {
                    "title": "Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks",
                    "abstract": "How well can NLP models generalize to a variety of unseen tasks when provided with task instructions? To address this question, we first introduce Super-NaturalInstructions, a benchmark of 1,616 diverse NLP tasks and their expert-written instructions. Our collection covers 76 distinct task types, including but not limited to classification, extraction, infilling, sequence tagging, text rewriting, and text composition. This large and diverse collection of tasks enables rigorous benchmarking of cross-task generalization under instructions\u2014training models to follow instructions on a subset of tasks and evaluating them on the remaining unseen ones.Furthermore, we build Tk-Instruct, a transformer model trained to follow a variety of in-context instructions (plain language task definitions or k-shot examples). Our experiments show that Tk-Instruct outperforms existing instruction-following models such as InstructGPT by over 9% on our benchmark despite being an order of magnitude smaller. We further analyze generalization as a function of various scaling parameters, such as the number of observed tasks, the number of instances per task, and model sizes. We hope our dataset and model facilitate future progress towards more general-purpose NLP models."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detection",
                    "abstract": "Toxic language detection systems often falsely flag text that contains minority group mentions as toxic, as those groups are often the targets of online hate. Such over-reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language.To help mitigate these issues, we create ToxiGen, a new large-scale and machine-generated dataset of 274k toxic and benign statements about 13 minority groups. We develop a demonstration-based prompting framework and an adversarial classifier-in-the-loop decoding method to generate subtly toxic and benign text with a massive pretrained language model. Controlling machine generation in this way allows ToxiGen to cover implicitly toxic text at a larger scale, and about more demographic groups, than previous resources of human-written text. We conduct a human evaluation on a challenging subset of ToxiGen and find that annotators struggle to distinguish machine-generated text from human-written language. We also find that 94.5% of toxic examples are labeled as hate speech by human annotators. Using three publicly-available datasets, we show that finetuning a toxicity classifier on our data improves its performance on human-written data substantially. We also demonstrate that ToxiGen can be used to fight machine-generated toxicity as finetuning improves the classifier significantly on our evaluation subset."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "ExT5: Towards Extreme Multi-Task Scaling for Transfer Learning",
                    "abstract": "Despite the recent success of multi-task learning and transfer learning for natural language processing (NLP), few works have systematically studied the effect of scaling up the number of tasks during pre-training. Towards this goal, this paper introduces ExMix (Extreme Mixture): a massive collection of 107 supervised NLP tasks across diverse domains and task-families. Using ExMix, we study the effect of multi-task pre-training at the largest scale to date, and analyze co-training transfer amongst common families of tasks. Through this analysis, we show that manually curating an ideal set of tasks for multi-task pre-training is not straightforward, and that multi-task scaling can vastly improve models on its own. Finally, we propose ExT5: a model pre-trained using a multi-task objective of self-supervised span denoising and supervised ExMix. Via extensive experiments, we show that ExT5 outperforms strong T5 baselines on SuperGLUE, GEM, Rainbow, Closed-Book QA tasks, and several tasks outside of ExMix. ExT5 also significantly improves sample efficiency while pre-training."
                },
                {
                    "title": "MetaICL: Learning to Learn In Context",
                    "abstract": "We introduce MetaICL (Meta-training for In-Context Learning), a new meta-training framework for few-shot learning where a pretrained language model is tuned to do in-context learning on a large set of training tasks. This meta-training enables the model to more effectively learn a new task in context at test time, by simply conditioning on a few training examples with no parameter updates or task-specific templates. We experiment on a large, diverse collection of tasks consisting of 142 NLP datasets including classification, question answering, natural language inference, paraphrase detection and more, across seven different meta-training/target splits. MetaICL outperforms a range of baselines including in-context learning without meta-training and multi-task learning followed by zero-shot transfer. We find that the gains are particularly significant for target tasks that have domain shifts from the meta-training tasks, and that using a diverse set of the meta-training tasks is key to improvements. We also show that MetaICL approaches (and sometimes beats) the performance of models fully finetuned on the target task training data, and outperforms much bigger models with nearly 8x parameters."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "Multitask Prompted Training Enables Zero-Shot Task Generalization",
                    "abstract": "Large language models have recently been shown to attain reasonable zero-shot generalization on a diverse set of tasks (Brown et al., 2020). It has been hypothesized that this is a consequence of implicit multitask learning in language models' pretraining (Radford et al., 2019). Can zero-shot generalization instead be directly induced by explicit multitask learning? To test this question at scale, we develop a system for easily mapping any natural language tasks into a human-readable prompted form. We convert a large set of supervised datasets, each with multiple prompts with diverse wording. These prompted datasets allow for benchmarking the ability of a model to perform completely held-out tasks. We fine-tune a pretrained encoder-decoder model (Raffel et al., 2020; Lester et al., 2021) on this multitask mixture covering a wide variety of tasks. The model attains strong zero-shot performance on several standard datasets, often outperforming models up to 16x its size. Further, our approach attains strong performance on a subset of tasks from the BIG-bench benchmark, outperforming models up to 6x its size. All trained models are available at https://github.com/bigscience-workshop/t-zero and all prompts are available at https://github.com/bigscience-workshop/promptsource."
                },
                {
                    "title": "TruthfulQA: Measuring How Models Mimic Human Falsehoods",
                    "abstract": "We propose a benchmark to measure whether a language model is truthful in generating answers to questions. The benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics. We crafted questions that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. We tested GPT-3, GPT-Neo/J, GPT-2 and a T5-based model. The best model was truthful on 58% of questions, while human performance was 94%. Models generated many false answers that mimic popular misconceptions and have the potential to deceive humans. The largest models were generally the least truthful. This contrasts with other NLP tasks, where performance improves with model size. However, this result is expected if false answers are learned from the training distribution. We suggest that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web."
                },
                {
                    "title": "Finetuned Language Models Are Zero-Shot Learners",
                    "abstract": "This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning."
                },
                {
                    "title": "Program Synthesis with Large Language Models",
                    "abstract": "This paper explores the limits of the current generation of large language models for program synthesis in general purpose programming languages. We evaluate a collection of such models (with between 244M and 137B parameters) on two new benchmarks, MBPP and MathQA-Python, in both the few-shot and fine-tuning regimes. Our benchmarks are designed to measure the ability of these models to synthesize short Python programs from natural language descriptions. The Mostly Basic Programming Problems (MBPP) dataset contains 974 programming tasks, designed to be solvable by entry-level programmers. The MathQA-Python dataset, a Python version of the MathQA benchmark, contains 23914 problems that evaluate the ability of the models to synthesize code from more complex text. On both datasets, we find that synthesis performance scales log-linearly with model size. Our largest models, even without finetuning on a code dataset, can synthesize solutions to 59.6 percent of the problems from MBPP using few-shot learning with a well-designed prompt. Fine-tuning on a held-out portion of the dataset improves performance by about 10 percentage points across most model sizes. On the MathQA-Python dataset, the largest fine-tuned model achieves 83.8 percent accuracy. Going further, we study the model's ability to engage in dialog about code, incorporating human feedback to improve its solutions. We find that natural language feedback from a human halves the error rate compared to the model's initial prediction. Additionally, we conduct an error analysis to shed light on where these models fall short and what types of programs are most difficult to generate. Finally, we explore the semantic grounding of these models by fine-tuning them to predict the results of program execution. We find that even our best models are generally unable to predict the output of a program given a specific input."
                },
                {
                    "title": "Evaluating Large Language Models Trained on Code",
                    "abstract": "We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics."
                },
                {
                    "title": "Cross-Task Generalization via Natural Language Crowdsourcing Instructions",
                    "abstract": "Humans (e.g., crowdworkers) have a remarkable ability in solving different tasks, by simply reading textual instructions that define them and looking at a few examples. Despite the success of the conventional supervised learning on individual datasets, such models often struggle with generalization across tasks (e.g., a question-answering system cannot solve classification tasks). A long-standing challenge in AI is to build a model that learns a new task by understanding the human-readable instructions that define it. To study this, we introduce NATURAL INSTRUCTIONS, a dataset of 61 distinct tasks, their human-authored instructions, and 193k task instances (input-output pairs). The instructions are obtained from crowdsourcing instructions used to create existing NLP datasets and mapped to a unified schema. Using this meta-dataset, we measure cross-task generalization by training models on seen tasks and measuring generalization to the remaining unseen ones. We adopt generative pre-trained language models to encode task-specific instructions along with input and generate task output. Our results indicate that models benefit from instructions when evaluated in terms of generalization to unseen tasks (19% better for models utilizing instructions). These models, however, are far behind an estimated performance upperbound indicating significant room for more progress in this direction."
                },
                {
                    "title": "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? \ud83e\udd9c",
                    "abstract": "The past 3 years of work in NLP have been characterized by the development and deployment of ever larger language models, especially for English. BERT, its variants, GPT-2/3, and others, most recently Switch-C, have pushed the boundaries of the possible both through architectural innovations and through sheer size. Using these pretrained models and the methodology of fine-tuning them for specific tasks, researchers have extended the state of the art on a wide array of tasks as measured by leaderboards on specific benchmarks for English. In this paper, we take a step back and ask: How big is too big? What are the possible risks associated with this technology and what paths are available for mitigating those risks? We provide recommendations including weighing the environmental and financial costs first, investing resources into curating and carefully documenting datasets rather than ingesting everything on the web, carrying out pre-development exercises evaluating how the planned approach fits into research and development goals and supports stakeholder values, and encouraging research directions beyond ever larger language models."
                },
                {
                    "title": "Extracting Training Data from Large Language Models",
                    "abstract": "It has become common to publish large (billion parameter) language models that have been trained on private datasets. This paper demonstrates that in such settings, an adversary can perform a training data extraction attack to recover individual training examples by querying the language model. \nWe demonstrate our attack on GPT-2, a language model trained on scrapes of the public Internet, and are able to extract hundreds of verbatim text sequences from the model's training data. These extracted examples include (public) personally identifiable information (names, phone numbers, and email addresses), IRC conversations, code, and 128-bit UUIDs. Our attack is possible even though each of the above sequences are included in just one document in the training data. \nWe comprehensively evaluate our extraction attack to understand the factors that contribute to its success. For example, we find that larger models are more vulnerable than smaller models. We conclude by drawing lessons and discussing possible safeguards for training large language models."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "Aligning AI With Shared Human Values",
                    "abstract": "We show how to assess a language model's knowledge of basic concepts of morality. We introduce the ETHICS dataset, a new benchmark that spans concepts in justice, well-being, duties, virtues, and commonsense morality. Models predict widespread moral judgments about diverse text scenarios. This requires connecting physical and social world knowledge to value judgements, a capability that may enable us to steer chatbot outputs or eventually regularize open-ended reinforcement learning agents. With the ETHICS dataset, we find that current language models have a promising but incomplete understanding of basic ethical knowledge. Our work shows that progress can be made on machine ethics today, and it provides a steppingstone toward AI that is aligned with human values."
                },
                {
                    "title": "Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data",
                    "abstract": "The success of the large neural language models on many NLP tasks is exciting. However, we find that these successes sometimes lead to hype in which these models are being described as \u201cunderstanding\u201d language or capturing \u201cmeaning\u201d. In this position paper, we argue that a system trained only on form has a priori no way to learn meaning. In keeping with the ACL 2020 theme of \u201cTaking Stock of Where We\u2019ve Been and Where We\u2019re Going\u201d, we argue that a clear understanding of the distinction between form and meaning will help guide the field towards better science around natural language understanding."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "UnifiedQA: Crossing Format Boundaries With a Single QA System",
                    "abstract": "Question answering (QA) tasks have been posed using a variety of formats, such as extractive span selection, multiple choice, etc. This has led to format-specialized models, and even to an implicit division in the QA community. We argue that such boundaries are artificial and perhaps unnecessary, given the reasoning abilities we seek to teach are not governed by the format. As evidence, we use the latest advances in language modeling to build a single pre-trained QA model, UNIFIEDQA, that performs well across 19 QA datasets spanning 4 diverse formats. UNIFIEDQA performs on par with 8 different models that were trained on individual datasets themselves. Even when faced with 12 unseen datasets of observed formats, UNIFIEDQA performs surprisingly well, showing strong generalization from its outof-format training data. Finally, simply finetuning this pre trained QA model into specialized models results in a new state of the art on 10 factoid and commonsense question answering datasets, establishing UNIFIEDQA as a strong starting point for building QA systems."
                },
                {
                    "title": "PIQA: Reasoning about Physical Commonsense in Natural Language",
                    "abstract": "To apply eyeshadow without a brush, should I use a cotton swab or a toothpick? Questions requiring this kind of physical commonsense pose a challenge to today's natural language understanding systems. While recent pretrained models (such as BERT) have made progress on question answering over more abstract domains \u2013 such as news articles and encyclopedia entries, where text is plentiful \u2013 in more physical domains, text is inherently limited due to reporting bias. Can AI systems learn to reliably answer physical commonsense questions without experiencing the physical world?In this paper, we introduce the task of physical commonsense reasoning and a corresponding benchmark dataset Physical Interaction: Question Answering or PIQA. Though humans find the dataset easy (95% accuracy), large pretrained models struggle (\u223c75%). We provide analysis about the dimensions of knowledge that existing models lack, which offers significant opportunities for future research."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "HellaSwag: Can a Machine Really Finish Your Sentence?",
                    "abstract": "Recent work by Zellers et al. (2018) introduced a new task of commonsense natural language inference: given an event description such as \u201cA woman sits at a piano,\u201d a machine must select the most likely followup: \u201cShe sets her fingers on the keys.\u201d With the introduction of BERT, near human-level performance was reached. Does this mean that machines can perform human level commonsense inference? In this paper, we show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans (>95% accuracy), state-of-the-art models struggle (<48%). We achieve this via Adversarial Filtering (AF), a data collection paradigm wherein a series of discriminators iteratively select an adversarial set of machine-generated wrong answers. AF proves to be surprisingly robust. The key insight is to scale up the length and complexity of the dataset examples towards a critical \u2018Goldilocks\u2019 zone wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models. Our construction of HellaSwag, and its resulting difficulty, sheds light on the inner workings of deep pretrained models. More broadly, it suggests a new path forward for NLP research, in which benchmarks co-evolve with the evolving state-of-the-art in an adversarial way, so as to present ever-harder challenges."
                },
                {
                    "title": "Scalable agent alignment via reward modeling: a research direction",
                    "abstract": "One obstacle to applying reinforcement learning algorithms to real-world problems is the lack of suitable reward functions. Designing such reward functions is difficult in part because the user only has an implicit understanding of the task objective. This gives rise to the agent alignment problem: how do we create agents that behave in accordance with the user's intentions? We outline a high-level research direction to solve the agent alignment problem centered around reward modeling: learning a reward function from interaction with the user and optimizing the learned reward function with reinforcement learning. We discuss the key challenges we expect to face when scaling reward modeling to complex and general domains, concrete approaches to mitigate these challenges, and ways to establish trust in the resulting agents."
                },
                {
                    "title": "Semi-Supervised Sequence Modeling with Cross-View Training",
                    "abstract": "Unsupervised representation learning algorithms such as word2vec and ELMo improve the accuracy of many supervised NLP models, mainly because they can take advantage of large amounts of unlabeled text. However, the supervised models only learn from task-specific labeled data during the main training phase. We therefore propose Cross-View Training (CVT), a semi-supervised learning algorithm that improves the representations of a Bi-LSTM sentence encoder using a mix of labeled and unlabeled data. On labeled examples, standard supervised learning is used. On unlabeled examples, CVT teaches auxiliary prediction modules that see restricted views of the input (e.g., only part of a sentence) to match the predictions of the full model seeing the whole input. Since the auxiliary modules and the full model share intermediate representations, this in turn improves the full model. Moreover, we show that CVT is particularly effective when combined with multi-task learning. We evaluate CVT on five sequence tagging tasks, machine translation, and dependency parsing, achieving state-of-the-art results."
                },
                {
                    "title": "CoQA: A Conversational Question Answering Challenge",
                    "abstract": "Humans gather information through conversations involving a series of interconnected questions and answers. For machines to assist in information gathering, it is therefore essential to enable them to answer conversational questions. We introduce CoQA, a novel dataset for building Conversational Question Answering systems. Our dataset contains 127k questions with answers, obtained from 8k conversations about text passages from seven diverse domains. The questions are conversational, and the answers are free-form text with their corresponding evidence highlighted in the passage. We analyze CoQA in depth and show that conversational questions have challenging phenomena not present in existing reading comprehension datasets (e.g., coreference and pragmatic reasoning). We evaluate strong dialogue and reading comprehension models on CoQA. The best system obtains an F1 score of 65.4%, which is 23.4 points behind human performance (88.8%), indicating that there is ample room for improvement. We present CoQA as a challenge to the community at https://stanfordnlp.github.io/coqa."
                },
                {
                    "title": "Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering",
                    "abstract": "We present a new kind of question answering dataset, OpenBookQA, modeled after open book exams for assessing human understanding of a subject. The open book that comes with our questions is a set of 1326 elementary level science facts. Roughly 6000 questions probe an understanding of these facts and their application to novel situations. This requires combining an open book fact (e.g., metals conduct electricity) with broad common knowledge (e.g., a suit of armor is made of metal) obtained from other sources. While existing QA datasets over documents or knowledge bases, being generally self-contained, focus on linguistic understanding, OpenBookQA probes a deeper understanding of both the topic\u2014in the context of common knowledge\u2014and the language it is expressed in. Human performance on OpenBookQA is close to 92%, but many state-of-the-art pre-trained QA methods perform surprisingly poorly, worse than several simple neural baselines we develop. Our oracle experiments designed to circumvent the knowledge retrieval bottleneck demonstrate the value of both the open book and additional facts. We leave it as a challenge to solve the retrieval problem in this multi-hop setting and to close the large gap to human performance."
                },
                {
                    "title": "Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge",
                    "abstract": "We present a new question set, text corpus, and baselines assembled to encourage AI research in advanced question answering. Together, these constitute the AI2 Reasoning Challenge (ARC), which requires far more powerful knowledge and reasoning than previous challenges such as SQuAD or SNLI. The ARC question set is partitioned into a Challenge Set and an Easy Set, where the Challenge Set contains only questions answered incorrectly by both a retrieval-based algorithm and a word co-occurence algorithm. The dataset contains only natural, grade-school science questions (authored for human tests), and is the largest public-domain set of this kind (7,787 questions). We test several baselines on the Challenge Set, including leading neural models from the SQuAD and SNLI tasks, and find that none are able to significantly outperform a random baseline, reflecting the difficult nature of this task. We are also releasing the ARC Corpus, a corpus of 14M science sentences relevant to the task, and implementations of the three neural baseline models tested. Can your model perform better? We pose ARC as a challenge to the community."
                },
                {
                    "title": "The LAMBADA dataset: Word prediction requiring a broad discourse context",
                    "abstract": "We introduce LAMBADA, a dataset to evaluate the capabilities of computational models for text understanding by means of a word prediction task. LAMBADA is a collection of narrative passages sharing the characteristic that human subjects are able to guess their last word if they are exposed to the whole passage, but not if they only see the last sentence preceding the target word. To succeed on LAMBADA, computational models cannot simply rely on local context, but must be able to keep track of information in the broader discourse. We show that LAMBADA exemplifies a wide range of linguistic phenomena, and that none of several state-of-the-art language models reaches accuracy above 1% on this novel benchmark. We thus propose LAMBADA as a challenging test set, meant to encourage the development of new models capable of genuine understanding of broad context in natural language text."
                },
                {
                    "title": "The Winograd Schema Challenge",
                    "abstract": "In this paper, we present an alternative to the Turing Test that has some conceptual and practical advantages. Like the original, it involves responding to typed English sentences, and English-speaking adults will have no difficulty with it. Unlike the original, the subject is not required to engage in a conversation and fool an interrogator into believing she is dealing with a person. Moreover, the test is arranged in such a way that having full access to a large corpus of English text might not help much. Finally, the interrogator or a third party will be able to decide unambiguously after a few minutes whether or not a subject has passed the test."
                },
                {
                    "title": "Findings of the 2009 Workshop on Statistical Machine Translation",
                    "abstract": "This paper presents the results of the WMT09 shared tasks, which included a translation task, a system combination task, and an evaluation task. We conducted a large-scale manual evaluation of 87 machine translation systems and 22 system combination entries. We used the ranking of these systems to measure how strongly automatic metrics correlate with human judgments of translation quality, for more than 20 metrics. We present a new evaluation technique whereby system output is edited and judged for correctness."
                },
                {
                    "title": "Bleu: a Method for Automatic Evaluation of Machine Translation",
                    "abstract": "Human evaluations of machine translation are extensive but expensive. Human evaluations can take months to finish and involve human labor that can not be reused. We propose a method of automatic machine translation evaluation that is quick, inexpensive, and language-independent, that correlates highly with human evaluation, and that has little marginal cost per run. We present this method as an automated understudy to skilled human judges which substitutes for them when there is need for quick or frequent evaluations."
                },
                {
                    "title": "Adaptive Compositional Continual Meta-Learning",
                    "abstract": "This paper focuses on continual meta-learning, where few-shot tasks are heterogeneous and sequentially available. Recent works use a mixture model for meta-knowledge to deal with the heterogeneity. However, these methods suffer from parameter inefficiency caused by two reasons: (1) the underlying assumption of mutual exclusiveness among mixture components hinders sharing meta-knowledge across heterogeneous tasks. (2) they only allow increasing mixture components and cannot adaptively filter out redundant components. In this paper, we pro-pose an A daptive C ompositional Continual M eta-L earning (ACML) algorithm, which employs a compositional premise to associate a task with a subset of mixture components, allowing meta-knowledge sharing among heterogeneous tasks. Moreover, to adaptively adjust the number of mixture components, we propose a component spar-sification method based on evidential theory to filter out redundant components. Experimental results show ACML outperforms strong baselines, showing the effectiveness of our compositional meta-knowledge, and confirming that ACML can adaptively learn meta-knowledge."
                },
                {
                    "title": "An Adversarial Winograd Schema Challenge at Scale",
                    "abstract": "The Winograd Schema Challenge (WSC), proposed by Levesque et al. (2011) as an alternative to the Turing Test, was originally designed as a pronoun resolution problem that cannot be solved based on statistical patterns in large text corpora. However, recent studies suggest that current WSC datasets, even when composed carefully by experts, are still prone to such biases that statistical methods can exploit. We introduce WINOGRANDE, a new collection of WSC problems that are adversarially constructed to be robust against spurious statistical biases. While the original WSC dataset provided only 273 instances, WINOGRANDE includes 43,985 instances, half of which are determined as adversarial. Key to our approach is a novel adversarial filtering algorithm AFLITE for systematic bias reduction, combined with a careful crowdsourcing design. Despite the significant increase in training data, the performance of existing stateof-the-art methods remains modest (61.6%) and contrasts with high human performance (90.8%) for the binary questions. In addition, WINOGRANDE allows us to use transfer learning for achieving new state-of-the-art results on the original WSC and related datasets. Finally, we discuss how biases lead to overestimating the true capabilities of machine commonsense."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Neural Machine Translation Systems for WMT 16",
                    "abstract": "We participated in the WMT 2016 shared news translation task by building neural translation systems for four language pairs, each trained in both directions: English \u2194Czech, English\u2194German, English \u2194Romanian and English\u2194Russian. Our systems are based on an attentional encoder-decoder, using BPE subword segmentation for open-vocabulary translation with a fixed vocabulary. We experimented with using automatic back-translations of the monolingual News corpus as additional training data, pervasive dropout, and target-bidirectional models. All reported methods give substantial improvements, and we see improvements of 4.3\u201311.2 BLEU over our baseline systems. In the human evaluation, our systems were the (tied) best constrained system for 7 out of 8 translation directions in which we participated. 12"
                },
                {
                    "title": "Changing Climate-",
                    "abstract": "The Antarctic marine ecosystem changes seasonally, forming a temporal continuum of specialised niche habitats including open ocean, sea ice and meltwater environments. The ability for phytoplankton to acclimate rapidly to the changed conditions of these environments depends on the species' physiology and photosynthetic plasticity and may ultimately determine their long term ecological niche adaptation. This study investigated the photophysiological plasticity and rapid acclimation response of three Antarctic diatoms\u2014Fragilariopsis cylindrus, Pseudonitzschia subcurvata and Chaetoceros sp.\u2014to a selected range of temperatures and salinities representative of the sea ice, meltwater and pelagic habitats in the Antarctic. Fragilariopsis cylindrus displayed physiological traits typical of adaptation to the sea ice environment. Equally, this species showed photosynthetic plasticity, acclimating to the range of environmental conditions, explaining the prevalence of this species in all Antarctic habitats. Pseudo-nitzschia subcurvata displayed a preference for the meltwater environment, but unlike F. cylindrus, photoprotective capacity was low and regulated via changes in PSII antenna size. Chaetoceros sp. had high plasticity in non-photochemical quenching, suggesting adaptation to variable light conditions experienced in the wind-mixed pelagic environment. While only capturing short-term responses, this study highlights the diversity in photoprotective capacity that exists amongst three dominant Antarctic diatom species and provides insight into links between ecological niche adaptation and species' distribution."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can incorporating loss computation for instructions or prompts, referred to as Instruction Modelling (IM), improve the performance of instruction tuning in language models?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of natural language processing (NLP) as it addresses the limitations of current instruction tuning methods, which primarily focus on output segments. By enhancing the alignment of language models with user intentions, this research could lead to more effective and versatile language models, ultimately benefiting a wide range of applications from chatbots to automated content generation. The findings could inspire future research to explore more nuanced approaches to model training, potentially leading to breakthroughs in how language models understand and generate human-like responses.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of effectively integrating loss computation for both instructions and outputs, as traditional methods have primarily focused on the latter. Naive approaches may fail because they do not account for the varying lengths and complexities of instructions compared to outputs, which can lead to suboptimal model performance. Additionally, the Superficial Alignment Hypothesis suggests that existing models may not require extensive instruction tuning data, complicating the understanding of how to best leverage instruction information. Overcoming these theoretical and practical obstacles is essential for developing a robust methodology.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely overlooked the importance of incorporating instruction loss in the training process, focusing instead on output loss. This gap has been compounded by the prevailing belief in the Superficial Alignment Hypothesis, which downplays the need for extensive instruction tuning. Barriers such as a lack of comprehensive datasets that include both instructions and outputs, as well as the challenge of effectively measuring the impact of instruction length on model performance, have hindered progress. Our approach differs by explicitly addressing these gaps and demonstrating the benefits of IM in various scenarios.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves implementing Instruction Modelling (IM) by incorporating loss computation for instructions alongside outputs during training. We will utilize datasets such as Less and Alpagasus, which are subsets of Flan V2, Dolly, and Stanford Alpaca, to evaluate our approach across multiple NLP tasks and benchmarks, including MMLU, TruthfulQA, HumanEval, MT-Bench, and AlpacaEval. The expected outcomes include significant"
            }
        },
        "author_data": {
            "35311232-fb1a-4672-ac66-776433900bf6": {
                "pk": "35311232-fb1a-4672-ac66-776433900bf6",
                "name": "Zhengyan Shi",
                "collaborators": [
                    "Sander Land",
                    "Acyr Locatelli",
                    "Matthieu Geist",
                    "Max Bartolo",
                    "Bin Wu",
                    "Hossein A. Rahmani",
                    "Varsha Ramineni",
                    "Emine Yilmaz",
                    "Adam X. Yang",
                    "Maxime Robeyns"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Natural Language Processing",
                    "Human Feedback",
                    "Personalization"
                ],
                "publications": [
                    {
                        "title": "Understanding Likelihood Over-optimisation in Direct Alignment Algorithms",
                        "abstract": "Direct Alignment Algorithms (DAAs), such as Direct Preference Optimisation (DPO) and Identity Preference Optimisation (IPO), have emerged as alternatives to online Reinforcement Learning from Human Feedback (RLHF) algorithms such as Proximal Policy Optimisation (PPO) for aligning language models to human preferences, without the need for explicit reward modelling. These methods generally aim to increase the likelihood of generating better (preferred) completions while discouraging worse (non-preferred) ones, while staying close to the original model's behaviour. In this work, we explore the relationship between completion likelihood and model performance in state-of-the-art DAAs, and identify a critical issue of likelihood over-optimisation. Contrary to expectations, we find that higher likelihood of better completions and larger margins between better and worse completion likelihoods do not necessarily lead to better performance, and may even degrade it. Our analysis reveals that while higher likelihood correlates with better memorisation of factual knowledge patterns, a slightly lower completion likelihood tends to improve output diversity, thus leading to better generalisation to unseen scenarios. Moreover, we identify two key indicators that signal when over-optimised output diversity begins to harm performance: Decreasing Entropy over Top-k Tokens and Diminishing Top-k Probability Mass. Our experimental results validate that these indicators are reliable signs of declining performance under different regularisations, helping prevent over-optimisation and improve alignment with human preferences."
                    },
                    {
                        "title": "Understanding the Role of User Profile in the Personalization of Large Language Models",
                        "abstract": "Utilizing user profiles to personalize Large Language Models (LLMs) has been shown to enhance the performance on a wide range of tasks. However, the precise role of user profiles and their effect mechanism on LLMs remains unclear. This study first confirms that the effectiveness of user profiles is primarily due to personalization information rather than semantic information. Furthermore, we investigate how user profiles affect the personalization of LLMs. Within the user profile, we reveal that it is the historical personalized response produced or approved by users that plays a pivotal role in personalizing LLMs. This discovery unlocks the potential of LLMs to incorporate a greater number of user profiles within the constraints of limited input length. As for the position of user profiles, we observe that user profiles integrated into different positions of the input context do not contribute equally to personalization. Instead, where the user profile that is closer to the beginning affects more on the personalization of LLMs. Our findings reveal the role of user profiles for the personalization of LLMs, and showcase how incorporating user profiles impacts performance providing insight to leverage user profiles effectively."
                    },
                    {
                        "title": "Bayesian Reward Models for LLM Alignment",
                        "abstract": "To ensure that large language model (LLM) responses are helpful and non-toxic, a reward model trained on human preference data is usually used. LLM responses with high rewards are then selected through best-of-$n$ (BoN) sampling or the LLM is further optimized to produce responses with high rewards through reinforcement learning from human feedback (RLHF). However, these processes are susceptible to reward overoptimization or `hacking', where responses receive high rewards due to imperfections in the reward model rather than true preference, particularly as prompts or responses deviate from the training data. To address these challenges, we propose to train a Bayesian reward model, which signals higher uncertainty further from the training data distribution. We trained Bayesian reward models using Laplace approximation on LoRA weights, and found that the resulting uncertainty estimates can effectively mitigate reward overoptimization in BoN sampling."
                    }
                ]
            },
            "13ed81f3-4b95-48c8-aeed-91b58cd08770": {
                "pk": "13ed81f3-4b95-48c8-aeed-91b58cd08770",
                "name": "Adam X. Yang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "42365b5e-d55c-4446-be7c-22a18402b53d": {
                "pk": "42365b5e-d55c-4446-be7c-22a18402b53d",
                "name": "Bin Wu",
                "collaborators": [
                    "Edmond Iancu",
                    "Liu Zhao",
                    "Fabio Dominguez",
                    "Guido Masera",
                    "Anders Tranberg"
                ],
                "domain": [
                    "Quantum Chromodynamics",
                    "High-Energy Physics",
                    "Thermalization",
                    "Jet Quenching"
                ],
                "publications": [
                    {
                        "title": "Radiative energy loss and radiative p_T-broadening of high-energy partons in QCD matter",
                        "abstract": "I give a self-contained review on radiative p_T-broadening and radiative energy loss of high-energy partons in QCD matter. The typical p_T^2 of high-energy partons receives a double logarithmic correction due to the recoiling effect of medium-induced gluon radiation. Such a double logarithmic term, averaged over the path length of the partons, can be taken as the radiative correction to the jet quenching parameter qhat and hence contributes to radiative energy loss. This has also been confirmed by detailed calculations of energy loss by radiating two gluons."
                    },
                    {
                        "title": "Early Time Dynamics and the Bulk",
                        "abstract": "Deciphering the origin of collective phenomena in small colliding systems is one of contemporary focuses in heavy-ion physics. It entails penetrating the barrier between two previously separated research topics: thermalization/hydrodynamization and phenomenological studies of collectivity. I first review some recent progress in understanding thermalization/hydrodynamization in large colliding systems, centralized on bottom-up thermalization. Then, using a simple kinetic theory I demonstrate how the investigation of hydrodynamization is intertwined with the study of flow in small colliding systems. Connections of these studies to \"hard probes\" are also commented where possible."
                    },
                    {
                        "title": "On p_T-broadening of high energy partons associated with the LPM effect in a finite-volume QCD medium",
                        "abstract": "We study the contributions from radiation to $p_{\\perp}$-broadening of a high energy parton traversing a QCD medium with a finite length $L$. The interaction between the parton and the medium is described by decorrelated static multiple scattering. Amplitudes of medium-induced gluon emission and parton self-energy diagrams are evaluated in the soft gluon limit in the BDMPS formalism. We find both the double-logarithmic correction from incoherent scattering, which is parametrically the same as that in single scattering, and the logarithmic correction from the LPM effect. Therefore, we expect a parametrically large correction from radiation to the medium-induced $p_\\perp$-broadening in perturbative QCD."
                    },
                    {
                        "title": "On holographic thermalization and gravitational collapse of tachyonic scalar fields",
                        "abstract": "In this paper we study the thermalization of a spatially homogeneous system in a strongly coupled CFT. The non-equilibrium initial state is created by switching on a relevant perturbation in the CFT vacuum during Delta t >= t >= -Delta t. Via AdS/CFT, the thermalization process corresponds to the gravitational collapse of a tachyonic scalar field (m^2 = -3) in the Poincare patch of AdS_5. In the limit Delta t < 0.02/T, the thermalization time t_T is found to be quantitatively the same as that of a non-equilibrium state created by a marginal perturbation discussed in Ref. [5]. In the case Delta t >= 1/T, we also obtain double-collapse solutions but with a non-equilibrium intermediate state at t = 0. In all the cases our results show that the system thermalizes in a typical time t_T ~ O(1)/T. Besides, a conserved energy-moment current in the bulk is found, which helps understand the qualitative difference of the collapse process in the Poincare patch from that in global AdS[9, 10]."
                    },
                    {
                        "title": "Radiative energy loss and radiative p_T-broadening of high-energy partons in QCD matter",
                        "abstract": "We study the connection between radiative energy loss and radiative p_T-broadening of a high-energy quark or gluon passing through QCD matter. The generalized Baier-Dokshitzer-Mueller-Peigne-Schiff-Zakharov (BDMPS-Z) formalism is used to calculate energy loss due to multiple gluon emission. With L the length of the matter and l_0 the size of constituents of the matter we find a double logarithmic correction proportional to ln^2(L/l_0) to parton energy loss due to two-gluon emission. We also show that the radiative energy loss per unit length -dE/dz = alpha_s N_c < p_T^2 >/12 by carrying out a resummation of the double logarithmic terms. Here, the transverse momentum broadening <p_T^2> is obtained by resumming terms proportional to alpha_s ln^2(L/l_0) in Ref. [11]. Our result agrees with that by the renormalization of qhat proposed in Refs. [13,14]."
                    },
                    {
                        "title": "Factorization and transverse phase-space parton distributions",
                        "abstract": "We first revisit impact-parameter dependent collisions of ultra-relativistic particles in quantum field theory. Two conditions sufficient for defining an impact-parameter dependent cross section are given, which could be violated in proton-proton collisions. By imposing these conditions, a general formula for the impact-parameter dependent cross section is derived. Then, using soft-collinear effective theory, we derive a factorization formula for the impact-parameter dependent cross section for inclusive hard processes with only colorless final-state products in hadron and nuclear collisions. It entails defining thickness beam functions, which are Fourier transforms of transverse phase-space parton distribution functions. By modelling non-perturbative modes in thickness beam functions of large nuclei in heavy-ion collisions, the factorization formula confirms the cross section in the Glauber model for hard processes. Besides, the factorization formula is verified up to one loop in perturbative QCD for the inclusive Drell-Yan process in quark-antiquark collisions at a finite impact parameter."
                    },
                    {
                        "title": "On holographic thermalization and gravitational collapse of massless scalar fields",
                        "abstract": "In this paper we study thermalization in a strongly coupled system via AdS/CFT. Initially, the energy is injected into the system by turning on a spatially homogenous scalar source coupled to a marginal composite operator. The thermalization process is studied by numerically solving Einstein's equations coupled to a massless scalar field in the Poincare patch of AdS_5. We define a thermalization time t_T on the AdS side, which has an interpretation in terms of a spacelike Wilson loop <W(l =1/T)> in CFT. Here T is the thermal equilibrium temperature. We study both cases with the source turned on in short(Delta t <= 1/T) and long(Delta t >= 1/T) durations. In the former case, the thermalization time t_T = g_t/T <= 1/T and the coefficient g_t = 0.73 in the limit Delta t <= 0.02/T. In the latter case, we find double- and multiple-collapse solutions, which may be interpreted as the gravity duals of two- or multi-stage thermalization in CFT. In all the cases our results indicate that such a strongly coupled system thermalizes in a typical time scale t_T=O(1)/T."
                    },
                    {
                        "title": "Holographic fluid from nonminimally coupled scalar-tensor theory of gravity",
                        "abstract": "We establish the gravity/fluid correspondence in the nonminimally coupled scalar-tensor theory of gravity. Imposing Petrov-like boundary conditions over the gravitational field, we find that, for a certain class of background metrics, the boundary fluctuations obey the standard Navier-Stokes equation for an incompressible fluid without any external force term in the leading order approximation under the near horizon expansion. That is to say, the scalar field fluctuations does not contribute in the leading order approximation regardless of what kind of boundary condition we impose on it."
                    },
                    {
                        "title": "Medium-induced jet evolution: multiple branching and thermalization",
                        "abstract": "For an energetic jet propagating through a weakly-coupled quark-gluon plasma, we present the physical picture of jet quenching in longitudinal phase-space, as emerging from the interplay between the medium-induced, quasi-democratic, branchings and the elastic collisions responsible for the thermalization of the soft branching products."
                    },
                    {
                        "title": "On dissociation of heavy mesons in a hot quark-gluon plasma",
                        "abstract": "We compare two mechanisms for the dissociation of heavy mesons in an infinite quark-gluon plasma: dynamic Debye screening and multiple scattering. Using the uncertainty principle inspired by a Schrodinger-like equation, we find that the criterion a_B\\simeq1/\\mu\\simeq 1/(\\alpha_{eff}^{1/2}T) with \\alpha_{eff}\\equiv\\alpha(N_c+{N_f\\over2}) is parametrically true both for the dissociation of fast moving heavy mesons with a size a_B due to dynamic Debye screening as well as for mesons at rest in the medium. In contrast, we find that the criterion for the dissociation of heavy mesons due to uncorrelated multiple scattering is parametrically 1/a_B \\simeq [\\gamma\\alpha_{eff}\\ln{1/\\alpha_{eff}}]^{1/3}T. Therefore, multiple scattering is a more efficient mechanism for the dissociation of heavy mesons in an infinite hot plasma."
                    },
                    {
                        "title": "A Novel VLSI Architecture of Fixed-complexity Sphere Decoder",
                        "abstract": "Fixed-complexity Sphere Decoder (FSD) is a recently proposed technique for Multiple-Input Multiple-Output (MIMO) detection. It has several outstanding features such as constant throughput and large potential parallelism, which makes it suitable for efficient VLSI implementation. However, to our best knowledge, no VLSI implementation of FSD has been reported in the literature, although some FPGA prototypes of FSD with pipeline architecture have been developed. These solutions achieve very high throughput but at very high cost of hardware resources, making them impractical in real applications. In this paper, we present a novel four-nodes-per-cycle parallel architecture of FSD, with a breadth-first processing that allows for short critical path. The implementation achieves a throughput of 213.3 Mbps at 400 MHz clock frequency, at a cost of 0.18 mm2 Silicon area on 0.13{\\mu}m CMOS technology. The proposed solution is much more economical compared with the existing FPGA implementations, and very suitable for practicl applications because of its balanced performance and hardware-complexity; moreover it has the flexibility to be expanded into an eight-nodes-per-cycle version in order to double the throughput."
                    },
                    {
                        "title": "Thermalization of mini-jets in a quark-gluon plasma",
                        "abstract": "We present the complete physical picture for the evolution of a high-energy jet propagating through a weakly-coupled quark-gluon plasma (QGP) by analytical and numerical investigation of thermalization of the soft components of the jet. Our results support the following physical picture: the leading particle emits a significant number of mini-jets which promptly evolve via multiple branching and thus degrade into a myriad of soft gluons, with energies of the order of the medium temperature $T$. Via elastic collisions with the medium constituents, these soft gluons relax to local thermal equilibrium with the plasma over a time scale which is considerably shorter than the typical lifetime of the mini-jet. The thermalized gluons form a tail which lags behind the hard components of the jet. Together with the background QGP, they behave hydrodynamically."
                    },
                    {
                        "title": "On using Cold Baryogenesis to constrain the Two-Higgs Doublet Model",
                        "abstract": "We consider the creation of the cosmological baryon asymmetry in the Two Higgs Doublet Model. We imagine a situation where the masses of the five Higgs particles and the two Higgs vevs are constrained by collider experiments, and demonstrate how the requirement of successful baryogenesis can be used to further constrain the remaining 4-dimensional parameter space of the model. We numerically compute the asymmetry within the scenario of Cold Electroweak Baryogenesis, which is particularly straightforward to simulate reliably."
                    }
                ]
            },
            "6962e43a-6e25-4510-80de-54532d2d3362": {
                "pk": "6962e43a-6e25-4510-80de-54532d2d3362",
                "name": "Laurence Aitchison",
                "collaborators": [
                    "Xi Wang",
                    "Stoil Ganev",
                    "Ali Unlu",
                    "Peter E. Latham",
                    "Sebastian W. Ober",
                    "M\u00e1t\u00e9 Lengyel",
                    "Bobby Antonio",
                    "Nicola Corradi",
                    "Vincent Adam",
                    "Srinivas C. Turaga"
                ],
                "domain": [
                    "Bayesian Inference",
                    "Variational Inference",
                    "Neural Networks",
                    "Probabilistic Modeling"
                ],
                "publications": [
                    {
                        "title": "Tensor Monte Carlo: particle methods for the GPU era",
                        "abstract": "Multi-sample, importance-weighted variational autoencoders (IWAE) give tighter bounds and more accurate uncertainty estimates than variational autoencoders (VAE) trained with a standard single-sample objective. However, IWAEs scale poorly: as the latent dimensionality grows, they require exponentially many samples to retain the benefits of importance weighting. While sequential Monte-Carlo (SMC) can address this problem, it is prohibitively slow because the resampling step imposes sequential structure which cannot be parallelised, and moreover, resampling is non-differentiable which is problematic when learning approximate posteriors. To address these issues, we developed tensor Monte-Carlo (TMC) which gives exponentially many importance samples by separately drawing $K$ samples for each of the $n$ latent variables, then averaging over all $K^n$ possible combinations. While the sum over exponentially many terms might seem to be intractable, in many cases it can be computed efficiently as a series of tensor inner-products. We show that TMC is superior to IWAE on a generative model with multiple stochastic layers trained on the MNIST handwritten digit database, and we show that TMC can be combined with standard variance reduction techniques."
                    },
                    {
                        "title": "Bayesian filtering unifies adaptive and non-adaptive neural network optimization methods",
                        "abstract": "We formulate the problem of neural network optimization as Bayesian filtering, where the observations are the backpropagated gradients. While neural network optimization has previously been studied using natural gradient methods which are closely related to Bayesian inference, they were unable to recover standard optimizers such as Adam and RMSprop with a root-mean-square gradient normalizer, instead getting a mean-square normalizer. To recover the root-mean-square normalizer, we find it necessary to account for the temporal dynamics of all the other parameters as they are geing optimized. The resulting optimizer, AdaBayes, adaptively transitions between SGD-like and Adam-like behaviour, automatically recovers AdamW, a state of the art variant of Adam with decoupled weight decay, and has generalisation performance competitive with SGD."
                    },
                    {
                        "title": "A statistical theory of cold posteriors in deep neural networks",
                        "abstract": "To get Bayesian neural networks to perform comparably to standard neural networks it is usually necessary to artificially reduce uncertainty using a \"tempered\" or \"cold\" posterior. This is extremely concerning: if the prior is accurate, Bayes inference/decision theory is optimal, and any artificial changes to the posterior should harm performance. While this suggests that the prior may be at fault, here we argue that in fact, BNNs for image classification use the wrong likelihood. In particular, standard image benchmark datasets such as CIFAR-10 are carefully curated. We develop a generative model describing curation which gives a principled Bayesian account of cold posteriors, because the likelihood under this new generative model closely matches the tempered likelihoods used in past work."
                    },
                    {
                        "title": "Why bigger is not always better: on finite and infinite neural networks",
                        "abstract": "Recent work has argued that neural networks can be understood theoretically by taking the number of channels to infinity, at which point the outputs become Gaussian process (GP) distributed. However, we note that infinite Bayesian neural networks lack a key facet of the behaviour of real neural networks: the fixed kernel, determined only by network hyperparameters, implies that they cannot do any form of representation learning. The lack of representation or equivalently kernel learning leads to less flexibility and hence worse performance, giving a potential explanation for the inferior performance of infinite networks observed in the literature (e.g. Novak et al. 2019). We give analytic results characterising the prior over representations and representation learning in finite deep linear networks. We show empirically that the representations in SOTA architectures such as ResNets trained with SGD are much closer to those suggested by our deep linear results than by the corresponding infinite network. This motivates the introduction of a new class of network: infinite networks with bottlenecks, which inherit the theoretical tractability of infinite networks while at the same time allowing representation learning."
                    },
                    {
                        "title": "Semi-supervised learning objectives as log-likelihoods in a generative model of data curation",
                        "abstract": "We currently do not have an understanding of semi-supervised learning (SSL) objectives such as pseudo-labelling and entropy minimization as log-likelihoods, which precludes the development of e.g. Bayesian SSL. Here, we note that benchmark image datasets such as CIFAR-10 are carefully curated, and we formulate SSL objectives as a log-likelihood in a generative model of data curation that was initially developed to explain the cold-posterior effect (Aitchison 2020). SSL methods, from entropy minimization and pseudo-labelling, to state-of-the-art techniques similar to FixMatch can be understood as lower-bounds on our principled log-likelihood. We are thus able to give a proof-of-principle for Bayesian SSL on toy data. Finally, our theory suggests that SSL is effective in part due to the statistical patterns induced by data curation. This provides an explanation of past results which show SSL performs better on clean datasets without any \"out of distribution\" examples. Confirming these results we find that SSL gave much larger performance improvements on curated than on uncurated data, using matched curated and uncurated datasets based on Galaxy Zoo 2."
                    },
                    {
                        "title": "Variational Laplace for Bayesian neural networks",
                        "abstract": "We develop variational Laplace for Bayesian neural networks (BNNs) which exploits a local approximation of the curvature of the likelihood to estimate the ELBO without the need for stochastic sampling of the neural-network weights. The Variational Laplace objective is simple to evaluate, as it is (in essence) the log-likelihood, plus weight-decay, plus a squared-gradient regularizer. Variational Laplace gave better test performance and expected calibration errors than maximum a-posteriori inference and standard sampling-based variational inference, despite using the same variational approximate posterior. Finally, we emphasise care needed in benchmarking standard VI as there is a risk of stopping before the variance parameters have converged. We show that early-stopping can be avoided by increasing the learning rate for the variance parameters."
                    },
                    {
                        "title": "The Hamiltonian brain: efficient probabilistic inference with excitatory-inhibitory neural circuit dynamics",
                        "abstract": "Probabilistic inference offers a principled framework for understanding both behaviour and cortical computation. However, two basic and ubiquitous properties of cortical responses seem difficult to reconcile with probabilistic inference: neural activity displays prominent oscillations in response to constant input, and large transient changes in response to stimulus onset. Here we show that these dynamical behaviours may in fact be understood as hallmarks of the specific representation and algorithm that the cortex employs to perform probabilistic inference. We demonstrate that a particular family of probabilistic inference algorithms, Hamiltonian Monte Carlo (HMC), naturally maps onto the dynamics of excitatory-inhibitory neural networks. Specifically, we constructed a model of an excitatory-inhibitory circuit in primary visual cortex that performed HMC inference, and thus inherently gave rise to oscillations and transients. These oscillations were not mere epiphenomena but served an important functional role: speeding up inference by rapidly spanning a large volume of state space. Inference thus became an order of magnitude more efficient than in a non-oscillatory variant of the model. In addition, the network matched two specific properties of observed neural dynamics that would otherwise be difficult to account for in the context of probabilistic inference. First, the frequency of oscillations as well as the magnitude of transients increased with the contrast of the image stimulus. Second, excitation and inhibition were balanced, and inhibition lagged excitation. These results suggest a new functional role for the separation of cortical populations into excitatory and inhibitory neurons, and for the neural oscillations that emerge in such excitatory-inhibitory networks: enhancing the efficiency of cortical computations."
                    },
                    {
                        "title": "InfoNCE is variational inference in a recognition parameterised model",
                        "abstract": "Here, we show that the InfoNCE objective is equivalent to the ELBO in a new class of probabilistic generative model, the recognition parameterised model (RPM). When we learn the optimal prior, the RPM ELBO becomes equal to the mutual information (MI; up to a constant), establishing a connection to pre-existing self-supervised learning methods such as InfoNCE. However, practical InfoNCE methods do not use the MI as an objective; the MI is invariant to arbitrary invertible transformations, so using an MI objective can lead to highly entangled representations (Tschannen et al., 2019). Instead, the actual InfoNCE objective is a simplified lower bound on the MI which is loose even in the infinite sample limit. Thus, an objective that works (i.e. the actual InfoNCE objective) appears to be motivated as a loose bound on an objective that does not work (i.e. the true MI which gives arbitrarily entangled representations). We give an alternative motivation for the actual InfoNCE objective. In particular, we show that in the infinite sample limit, and for a particular choice of prior, the actual InfoNCE objective is equal to the ELBO (up to a constant); and the ELBO is equal to the marginal likelihood with a deterministic recognition model. Thus, we argue that our VAE perspective gives a better motivation for InfoNCE than MI, as the actual InfoNCE objective is only loosely bounded by the MI, but is equal to the ELBO/marginal likelihood (up to a constant)."
                    },
                    {
                        "title": "Variational Laplace for Bayesian neural networks",
                        "abstract": "We develop variational Laplace for Bayesian neural networks (BNNs) which exploits a local approximation of the curvature of the likelihood to estimate the ELBO without the need for stochastic sampling of the neural-network weights. The Variational Laplace objective is simple to evaluate, as it is (in essence) the log-likelihood, plus weight-decay, plus a squared-gradient regularizer. Variational Laplace gave better test performance and expected calibration errors than maximum a-posteriori inference and standard sampling-based variational inference, despite using the same variational approximate posterior. Finally, we emphasise care needed in benchmarking standard VI as there is a risk of stopping before the variance parameters have converged. We show that early-stopping can be avoided by increasing the learning rate for the variance parameters."
                    },
                    {
                        "title": "Bayesian OOD detection with aleatoric uncertainty and outlier exposure",
                        "abstract": "Typical Bayesian approaches to OOD detection use epistemic uncertainty. Surprisingly from the Bayesian perspective, there are a number of methods that successfully use aleatoric uncertainty to detect OOD points (e.g. Hendryks et al. 2018). In addition, it is difficult to use outlier exposure to improve a Bayesian OOD detection model, as it is not clear whether it is possible or desirable to increase posterior (epistemic) uncertainty at outlier points. We show that a generative model of data curation provides a principled account of aleatoric uncertainty for OOD detection. In particular, aleatoric uncertainty signals a specific type of OOD point: one without a well-defined class-label, and our model of data curation gives a likelihood for these points, giving us a mechanism for conditioning on outlier points and thus performing principled Bayesian outlier exposure. Our principled Bayesian approach, combining aleatoric and epistemic uncertainty with outlier exposure performs better than methods using aleatoric or epistemic alone."
                    },
                    {
                        "title": "Robustness to corruption in pre-trained Bayesian neural networks",
                        "abstract": "We develop ShiftMatch, a new training-data-dependent likelihood for robustness to corruption in Bayesian neural networks (BNNs). ShiftMatch is inspired by the training-data-dependent \"EmpCov\" priors from Izmailov et al. (2021a), and efficiently matches test-time spatial correlations to those at training time. Critically, ShiftMatch is designed to leave the neural network's training time likelihood unchanged, allowing it to use publicly available samples from pre-trained BNNs. Using pre-trained HMC samples, ShiftMatch gives strong performance improvements on CIFAR-10-C, outperforms EmpCov priors (though ShiftMatch uses extra information from a minibatch of corrupted test points), and is perhaps the first Bayesian method capable of convincingly outperforming plain deep ensembles."
                    },
                    {
                        "title": "How to derive skill from the Fractions Skill Score",
                        "abstract": "The Fractions Skill Score (FSS) is a widely used metric for assessing forecast skill, with applications ranging from precipitation to volcanic ash forecasts. By evaluating the fraction of grid squares exceeding a threshold in a neighbourhood, the intuition is that it can avoid the pitfalls of pixel-wise comparisons and identify length scales at which a forecast has skill. The FSS is typically interpreted relative to a `useful' criterion, which provides a measure of whether a forecast is considered skilful enough. However, this useful criterion is problematic, as there can be considerable skill for forecasts that do not meet it, and it is not derived in a way that provides obvious meaning. We therefore provide a new derivation of a useful criterion, as the score that a random forecast achieves, which provides a more robust and meaningful way to assess the FSS. Through illustrative examples we show that this new criterion considerably changes the length scales at which a forecast would be regarded as skilful, and reveals subtleties in how the FSS should be interpreted."
                    },
                    {
                        "title": "Batch size invariant Adam",
                        "abstract": "We propose a batch size invariant version of Adam, for use in large-scale, distributed settings, in which the mini-batch is divided into micro-batches which are distributed among worker nodes. For the v term, standard Adam first computes the average over micro-batch gradients, then squares, while in the batch size invariant Adam proposed here, we first square the micro-batch gradients, then average. Previous work (e.g. Malladi et al. 2022) used an alternative approach that involved a square-root scaling of the learning rate, but this approach requires strong assumptions to work; in particular that the gradient variance dominates the square of the expected gradient. In contrast, the approach proposed here gives batch size invariance without this assumption. We confirm that in practice our scheme gives batch size invariance in a much larger range of scenarios than the previous approach."
                    },
                    {
                        "title": "How to set AdamW's weight decay as you scale model and dataset size",
                        "abstract": "We show that weights learned by AdamW can be understood as an exponential moving average (EMA) of recent updates. This gives critical insights for how to set the weight decay in AdamW, and how the weight decay should scale with model and dataset size. In particular, the key hyperparameter for an exponential moving average is the EMA timescale. Intuitively, the EMA timescale can be understood as the number of recent iterations the EMA averages over. Given a fixed learning rate, there is a one-to-one mapping from the EMA timescale to the usual weight decay hyperparameter. Thus, choosing an EMA timescale implicitly sets the weight decay. Importantly, there are natural guidelines for sensible values for the EMA timescale: we need to average over all datapoints, so the EMA timescale should not be (much) smaller than 1 epoch, and we need to forget early updates, so the EMA timescale should not be (much) bigger than the total number of training epochs. In our experiments, we find that optimal EMA timescales are consistent with these guidelines, as are the hyperparameters chosen in recent large-scale LLM pretraining runs (e.g.\\ Llama 1+2 and Stable LM). Critically, these guidelines suggest that the optimal EMA timescale should not change (much) as we scale the model and dataset. That implies that as the dataset size increases, the optimal weight decay should fall. Moreover, as the model size increases, the optimal weight decay should also increase (if we follow the muP recommendation for scaling the learning rate)."
                    },
                    {
                        "title": "Synaptic sampling: A connection between PSP variability and uncertainty explains neurophysiological observations",
                        "abstract": "When an action potential is transmitted to a postsynaptic neuron, a small change in the postsynaptic neuron's membrane potential occurs. These small changes, known as a postsynaptic potentials (PSPs), are highly variable, and current models assume that this variability is corrupting noise. In contrast, we show that this variability could have an important computational role: representing a synapse's uncertainty about the optimal synaptic weight (i.e. the best possible setting for the synaptic weight). We show that this link between uncertainty and variability, that we call synaptic sampling, leads to more accurate estimates of the uncertainty in task relevant quantities, leading to more effective decision making. Synaptic sampling makes four predictions, all of which have some experimental support. First the more variable a synapse is, the more it should change during LTP protocols. Second, variability should increase as the presynpatic firing rate falls. Third, PSP variance should be proportional to PSP mean. Fourth, variability should increase with distance from the cell soma. We provide support for the first two predictions by reanalysing existing datasets, and we find preexisting data in support of the last two predictions."
                    },
                    {
                        "title": "A variational approximate posterior for the deep Wishart process",
                        "abstract": "Recent work introduced deep kernel processes as an entirely kernel-based alternative to NNs (Aitchison et al. 2020). Deep kernel processes flexibly learn good top-layer representations by alternately sampling the kernel from a distribution over positive semi-definite matrices and performing nonlinear transformations. A particular deep kernel process, the deep Wishart process (DWP), is of particular interest because its prior can be made equivalent to deep Gaussian process (DGP) priors for kernels that can be expressed entirely in terms of Gram matrices. However, inference in DWPs has not yet been possible due to the lack of sufficiently flexible distributions over positive semi-definite matrices. Here, we give a novel approach to obtaining flexible distributions over positive semi-definite matrices by generalising the Bartlett decomposition of the Wishart probability density. We use this new distribution to develop an approximate posterior for the DWP that includes dependency across layers. We develop a doubly-stochastic inducing-point inference scheme for the DWP and show experimentally that inference in the DWP can improve performance over doing inference in a DGP with the equivalent prior."
                    },
                    {
                        "title": "Zipf's law arises naturally in structured, high-dimensional data",
                        "abstract": "Zipf's law, which states that the probability of an observation is inversely proportional to its rank, has been observed in many domains. While there are models that explain Zipf's law in each of them, those explanations are typically domain specific. Recently, methods from statistical physics were used to show that a fairly broad class of models does provide a general explanation of Zipf's law. This explanation rests on the observation that real world data is often generated from underlying causes, known as latent variables. Those latent variables mix together multiple models that do not obey Zipf's law, giving a model that does. Here we extend that work both theoretically and empirically. Theoretically, we provide a far simpler and more intuitive explanation of Zipf's law, which at the same time considerably extends the class of models to which this explanation can apply. Furthermore, we also give methods for verifying whether this explanation applies to a particular dataset. Empirically, these advances allowed us extend this explanation to important classes of data, including word frequencies (the first domain in which Zipf's law was discovered), data with variable sequence length, and multi-neuron spiking activity."
                    },
                    {
                        "title": "Discrete flow posteriors for variational inference in discrete dynamical systems",
                        "abstract": "Each training step for a variational autoencoder (VAE) requires us to sample from the approximate posterior, so we usually choose simple (e.g. factorised) approximate posteriors in which sampling is an efficient computation that fully exploits GPU parallelism. However, such simple approximate posteriors are often insufficient, as they eliminate statistical dependencies in the posterior. While it is possible to use normalizing flow approximate posteriors for continuous latents, some problems have discrete latents and strong statistical dependencies. The most natural approach to model these dependencies is an autoregressive distribution, but sampling from such distributions is inherently sequential and thus slow. We develop a fast, parallel sampling procedure for autoregressive distributions based on fixed-point iterations which enables efficient and accurate variational inference in discrete state-space latent variable dynamical systems. To optimize the variational bound, we considered two ways to evaluate probabilities: inserting the relaxed samples directly into the pmf for the discrete distribution, or converting to continuous logistic latent variables and interpreting the K-step fixed-point iterations as a normalizing flow. We found that converting to continuous latent variables gave considerable additional scope for mismatch between the true and approximate posteriors, which resulted in biased inferences, we thus used the former approach. Using our fast sampling procedure, we were able to realize the benefits of correlated posteriors, including accurate uncertainty estimates for one cell, and accurate connectivity estimates for multiple cells, in an order of magnitude less time."
                    },
                    {
                        "title": "Sampling-based probabilistic inference emerges from learning in neural circuits with a cost on reliability",
                        "abstract": "Neural responses in the cortex change over time both systematically, due to ongoing plasticity and learning, and seemingly randomly, due to various sources of noise and variability. Most previous work considered each of these processes, learning and variability, in isolation -- here we study neural networks exhibiting both and show that their interaction leads to the emergence of powerful computational properties. We trained neural networks on classical unsupervised learning tasks, in which the objective was to represent their inputs in an efficient, easily decodable form, with an additional cost for neural reliability which we derived from basic biophysical considerations. This cost on reliability introduced a tradeoff between energetically cheap but inaccurate representations and energetically costly but accurate ones. Despite the learning tasks being non-probabilistic, the networks solved this tradeoff by developing a probabilistic representation: neural variability represented samples from statistically appropriate posterior distributions that would result from performing probabilistic inference over their inputs. We provide an analytical understanding of this result by revealing a connection between the cost of reliability, and the objective for a state-of-the-art Bayesian inference strategy: variational autoencoders. We show that the same cost leads to the emergence of increasingly accurate probabilistic representations as networks become more complex, from single-layer feed-forward, through multi-layer feed-forward, to recurrent architectures. Our results provide insights into why neural responses in sensory areas show signatures of sampling-based probabilistic representations, and may inform future deep learning algorithms and their implementation in stochastic low-precision computing systems."
                    },
                    {
                        "title": "Deep kernel processes",
                        "abstract": "We define deep kernel processes in which positive definite Gram matrices are progressively transformed by nonlinear kernel functions and by sampling from (inverse) Wishart distributions. Remarkably, we find that deep Gaussian processes (DGPs), Bayesian neural networks (BNNs), infinite BNNs, and infinite BNNs with bottlenecks can all be written as deep kernel processes. For DGPs the equivalence arises because the Gram matrix formed by the inner product of features is Wishart distributed, and as we show, standard isotropic kernels can be written entirely in terms of this Gram matrix -- we do not need knowledge of the underlying features. We define a tractable deep kernel process, the deep inverse Wishart process, and give a doubly-stochastic inducing-point variational inference scheme that operates on the Gram matrices, not on the features, as in DGPs. We show that the deep inverse Wishart process gives superior performance to DGPs and infinite BNNs on standard fully-connected baselines."
                    }
                ]
            },
            "d6e4875f-7617-4a34-905c-96e69a25cc53": {
                "pk": "d6e4875f-7617-4a34-905c-96e69a25cc53",
                "name": "Emine Yilmaz",
                "collaborators": [
                    "Fanghua Ye",
                    "Sahan Bulathwela",
                    "Qiang Zhang",
                    "Rishabh Mehrotra",
                    "Aldo Lipani",
                    "Hamze Muse",
                    "Shangsong Liang",
                    "Procheta Sen",
                    "Xi Wang",
                    "Debasis Ganguly"
                ],
                "domain": [
                    "Information Retrieval",
                    "Natural Language Processing",
                    "Machine Learning",
                    "Educational Technology"
                ],
                "publications": [
                    {
                        "title": "Extracting Hierarchies of Search Tasks & Subtasks via a Bayesian Nonparametric Approach",
                        "abstract": "A significant amount of search queries originate from some real world information need or tasks. In order to improve the search experience of the end users, it is important to have accurate representations of tasks. As a result, significant amount of research has been devoted to extracting proper representations of tasks in order to enable search systems to help users complete their tasks, as well as providing the end user with better query suggestions, for better recommendations, for satisfaction prediction, and for improved personalization in terms of tasks. Most existing task extraction methodologies focus on representing tasks as flat structures. However, tasks often tend to have multiple subtasks associated with them and a more naturalistic representation of tasks would be in terms of a hierarchy, where each task can be composed of multiple (sub)tasks. To this end, we propose an efficient Bayesian nonparametric model for extracting hierarchies of such tasks \\& subtasks. We evaluate our method based on real world query log data both through quantitative and crowdsourced experiments and highlight the importance of considering task/subtask hierarchies."
                    },
                    {
                        "title": "Query-specific Variable Depth Pooling via Query Performance Prediction towards Reducing Relevance Assessment Effort",
                        "abstract": "Due to the massive size of test collections, a standard practice in IR evaluation is to construct a 'pool' of candidate relevant documents comprised of the top-k documents retrieved by a wide range of different retrieval systems - a process called depth-k pooling. A standard practice is to set the depth (k) to a constant value for each query constituting the benchmark set. However, in this paper we argue that the annotation effort can be substantially reduced if the depth of the pool is made a variable quantity for each query, the rationale being that the number of documents relevant to the information need can widely vary across queries. Our hypothesis is that a lower depth for the former class of queries and a higher depth for the latter can potentially reduce the annotation effort without a significant change in retrieval effectiveness evaluation. We make use of standard query performance prediction (QPP) techniques to estimate the number of potentially relevant documents for each query, which is then used to determine the depth of the pool. Our experiments conducted on standard test collections demonstrate that this proposed method of employing query-specific variable depths is able to adequately reflect the relative effectiveness of IR systems with a substantially smaller annotation effort."
                    },
                    {
                        "title": "Towards More Accountable Search Engines: Online Evaluation of Representation Bias",
                        "abstract": "Information availability affects people's behavior and perception of the world. Notably, people rely on search engines to satisfy their need for information. Search engines deliver results relevant to user requests usually without being or making themselves accountable for the information they deliver, which may harm people's lives and, in turn, society. This potential risk urges the development of evaluation mechanisms of bias in order to empower the user in judging the results of search engines. In this paper, we give a possible solution to measuring representation bias with respect to societal features for search engines and apply it to evaluating the gender representation bias for Google's Knowledge Graph Carousel for listing occupations."
                    },
                    {
                        "title": "Pre-Training With Scientific Text Improves Educational Question Generation",
                        "abstract": "With the boom of digital educational materials and scalable e-learning systems, the potential for realising AI-assisted personalised learning has skyrocketed. In this landscape, the automatic generation of educational questions will play a key role, enabling scalable self-assessment when a global population is manoeuvring their personalised learning journeys. We develop EduQG, a novel educational question generation model built by adapting a large language model. Our initial experiments demonstrate that EduQG can produce superior educational questions by pre-training on scientific text."
                    },
                    {
                        "title": "Scalable Educational Question Generation with Pre-trained Language Models",
                        "abstract": "The automatic generation of educational questions will play a key role in scaling online education, enabling self-assessment at scale when a global population is manoeuvring their personalised learning journeys. We develop \\textit{EduQG}, a novel educational question generation model built by adapting a large language model. Our extensive experiments demonstrate that \\textit{EduQG} can produce superior educational questions by further pre-training and fine-tuning a pre-trained language model on the scientific text and science question data."
                    },
                    {
                        "title": "Towards Task Understanding in Visual Settings",
                        "abstract": "We consider the problem of understanding real world tasks depicted in visual images. While most existing image captioning methods excel in producing natural language descriptions of visual scenes involving human tasks, there is often the need for an understanding of the exact task being undertaken rather than a literal description of the scene. We leverage insights from real world task understanding systems, and propose a framework composed of convolutional neural networks, and an external hierarchical task ontology to produce task descriptions from input images. Detailed experiments highlight the efficacy of the extracted descriptions, which could potentially find their way in many applications, including image alt text generation."
                    },
                    {
                        "title": "Variational Self-attention Model for Sentence Representation",
                        "abstract": "This paper proposes a variational self-attention model (VSAM) that employs variational inference to derive self-attention. We model the self-attention vector as random variables by imposing a probabilistic distribution. The self-attention mechanism summarizes source information as an attention vector by weighted sum, where the weights are a learned probabilistic distribution. Compared with conventional deterministic counterpart, the stochastic units incorporated by VSAM allow multi-modal attention distributions. Furthermore, by marginalizing over the latent variables, VSAM is more robust against overfitting. Experiments on the stance detection task demonstrate the superiority of our method."
                    },
                    {
                        "title": "MultiWOZ 2.4: A Multi-Domain Task-Oriented Dialogue Dataset with Essential Annotation Corrections to Improve State Tracking Evaluation",
                        "abstract": "The MultiWOZ 2.0 dataset has greatly stimulated the research of task-oriented dialogue systems. However, its state annotations contain substantial noise, which hinders a proper evaluation of model performance. To address this issue, massive efforts were devoted to correcting the annotations. Three improved versions (i.e., MultiWOZ 2.1-2.3) have then been released. Nonetheless, there are still plenty of incorrect and inconsistent annotations. This work introduces MultiWOZ 2.4, which refines the annotations in the validation set and test set of MultiWOZ 2.1. The annotations in the training set remain unchanged (same as MultiWOZ 2.1) to elicit robust and noise-resilient model training. We benchmark eight state-of-the-art dialogue state tracking models on MultiWOZ 2.4. All of them demonstrate much higher performance than on MultiWOZ 2.1."
                    },
                    {
                        "title": "Semantic TrueLearn: Using Semantic Knowledge Graphs in Recommendation Systems",
                        "abstract": "In informational recommenders, many challenges arise from the need to handle the semantic and hierarchical structure between knowledge areas. This work aims to advance towards building a state-aware educational recommendation system that incorporates semantic relatedness between knowledge topics, propagating latent information across semantically related topics. We introduce a novel learner model that exploits this semantic relatedness between knowledge components in learning resources using the Wikipedia link graph, with the aim to better predict learner engagement and latent knowledge in a lifelong learning scenario. In this sense, Semantic TrueLearn builds a humanly intuitive knowledge representation while leveraging Bayesian machine learning to improve the predictive performance of the educational engagement. Our experiments with a large dataset demonstrate that this new semantic version of TrueLearn algorithm achieves statistically significant improvements in terms of predictive performance with a simple extension that adds semantic awareness to the model."
                    },
                    {
                        "title": "ASSIST: Towards Label Noise-Robust Dialogue State Tracking",
                        "abstract": "The MultiWOZ 2.0 dataset has greatly boosted the research on dialogue state tracking (DST). However, substantial noise has been discovered in its state annotations. Such noise brings about huge challenges for training DST models robustly. Although several refined versions, including MultiWOZ 2.1-2.4, have been published recently, there are still lots of noisy labels, especially in the training set. Besides, it is costly to rectify all the problematic annotations. In this paper, instead of improving the annotation quality further, we propose a general framework, named ASSIST (lAbel noiSe-robuSt dIalogue State Tracking), to train DST models robustly from noisy labels. ASSIST first generates pseudo labels for each sample in the training set by using an auxiliary model trained on a small clean dataset, then puts the generated pseudo labels and vanilla noisy labels together to train the primary model. We show the validity of ASSIST theoretically. Experimental results also demonstrate that ASSIST improves the joint goal accuracy of DST by up to $28.16\\%$ on MultiWOZ 2.0 and $8.41\\%$ on MultiWOZ 2.4, compared to using only the vanilla noisy labels."
                    },
                    {
                        "title": "Task2KB: A Public Task-Oriented Knowledge Base",
                        "abstract": "Search engines and conversational assistants are commonly used to help users complete their every day tasks such as booking travel, cooking, etc. While there are some existing datasets that can be used for this purpose, their coverage is limited to very few domains. In this paper, we propose a novel knowledge base, 'Task2KB', which is constructed using data crawled from WikiHow, an online knowledge resource offering instructional articles on a wide range of tasks. Task2KB encapsulates various types of task-related information and attributes, such as requirements, detailed step description, and available methods to complete tasks. Due to its higher coverage compared to existing related knowledge graphs, Task2KB can be highly useful in the development of general purpose task completion assistants"
                    },
                    {
                        "title": "Modeling User Satisfaction Dynamics in Dialogue via Hawkes Process",
                        "abstract": "Dialogue systems have received increasing attention while automatically evaluating their performance remains challenging. User satisfaction estimation (USE) has been proposed as an alternative. It assumes that the performance of a dialogue system can be measured by user satisfaction and uses an estimator to simulate users. The effectiveness of USE depends heavily on the estimator. Existing estimators independently predict user satisfaction at each turn and ignore satisfaction dynamics across turns within a dialogue. In order to fully simulate users, it is crucial to take satisfaction dynamics into account. To fill this gap, we propose a new estimator ASAP (sAtisfaction eStimation via HAwkes Process) that treats user satisfaction across turns as an event sequence and employs a Hawkes process to effectively model the dynamics in this sequence. Experimental results on four benchmark dialogue datasets demonstrate that ASAP can substantially outperform state-of-the-art baseline estimators."
                    },
                    {
                        "title": "Enhancing Conversational Search: Large Language Model-Aided Informative Query Rewriting",
                        "abstract": "Query rewriting plays a vital role in enhancing conversational search by transforming context-dependent user queries into standalone forms. Existing approaches primarily leverage human-rewritten queries as labels to train query rewriting models. However, human rewrites may lack sufficient information for optimal retrieval performance. To overcome this limitation, we propose utilizing large language models (LLMs) as query rewriters, enabling the generation of informative query rewrites through well-designed instructions. We define four essential properties for well-formed rewrites and incorporate all of them into the instruction. In addition, we introduce the role of rewrite editors for LLMs when initial query rewrites are available, forming a \"rewrite-then-edit\" process. Furthermore, we propose distilling the rewriting capabilities of LLMs into smaller models to reduce rewriting latency. Our experimental evaluation on the QReCC dataset demonstrates that informative query rewrites can yield substantially improved retrieval performance compared to human rewrites, especially with sparse retrievers."
                    },
                    {
                        "title": "Adaptive Retrieval-Augmented Generation for Conversational Systems",
                        "abstract": "Despite the success of integrating large language models into the development of conversational systems, many studies have shown the effectiveness of retrieving and augmenting external knowledge for informative responses. Hence, many existing studies commonly assume the always need for Retrieval Augmented Generation (RAG) in a conversational system without explicit control. This raises a research question about such a necessity. In this study, we propose to investigate the need for each turn of system response to be augmented with external knowledge. In particular, by leveraging human judgements on the binary choice of adaptive augmentation, we develop RAGate, a gating model, which models conversation context and relevant inputs to predict if a conversational system requires RAG for improved responses. We conduct extensive experiments on devising and applying RAGate to conversational models and well-rounded analyses of different conversational scenarios. Our experimental results and analysis indicate the effective application of RAGate in RAG-based conversational systems in identifying system responses for appropriate RAG with high-quality responses and a high generation confidence. This study also identifies the correlation between the generation's confidence level and the relevance of the augmented knowledge."
                    },
                    {
                        "title": "Self-Attentive Hawkes Processes",
                        "abstract": "Asynchronous events on the continuous time domain, e.g., social media actions and stock transactions, occur frequently in the world. The ability to recognize occurrence patterns of event sequences is crucial to predict which typeof events will happen next and when. A de facto standard mathematical framework to do this is the Hawkes process. In order to enhance expressivity of multivariate Hawkes processes, conventional statistical methods and deep recurrent networks have been employed to modify its intensity function. The former is highly interpretable and requires small size of training data but relies on correct model design while the latter has less dependency on prior knowledge and is more powerful in capturing complicated patterns. We leverage pros and cons of these models and propose a self-attentive Hawkes process(SAHP). The proposed method adapts self-attention to fit the intensity function of Hawkes processes. This design has two benefits:(1) compared with conventional statistical methods, the SAHP is more powerful to identify complicated dependency relationships between temporal events; (2)compared with deep recurrent networks, the self-attention mechanism is able to capture longer historical information, and is more interpretable because the learnt attention weight tensor shows contributions of each historical event. Experiments on four real-world datasets demonstrate the effectiveness of the proposed method."
                    },
                    {
                        "title": "Disentangling Interpretable Generative Parameters of Random and Real-World Graphs",
                        "abstract": "While a wide range of interpretable generative procedures for graphs exist, matching observed graph topologies with such procedures and choices for its parameters remains an open problem. Devising generative models that closely reproduce real-world graphs requires domain knowledge and time-consuming simulation. While existing deep learning approaches rely on less manual modelling, they offer little interpretability. This work approaches graph generation (decoding) as the inverse of graph compression (encoding). We show that in a disentanglement-focused deep autoencoding framework, specifically Beta-Variational Autoencoders (Beta-VAE), choices of generative procedures and their parameters arise naturally in the latent space. Our model is capable of learning disentangled, interpretable latent variables that represent the generative parameters of procedurally generated random graphs and real-world graphs. The degree of disentanglement is quantitatively measured using the Mutual Information Gap (MIG). When training our Beta-VAE model on ER random graphs, its latent variables have a near one-to-one mapping to the ER random graph parameters n and p. We deploy the model to analyse the correlation between graph topology and node attributes measuring their mutual dependence without handpicking topological properties."
                    },
                    {
                        "title": "ViralBERT: A User Focused BERT-Based Approach to Virality Prediction",
                        "abstract": "Recently, Twitter has become the social network of choice for sharing and spreading information to a multitude of users through posts called 'tweets'. Users can easily re-share these posts to other users through 'retweets', which allow information to cascade to many more users, increasing its outreach. Clearly, being able to know the extent to which a post can be retweeted has great value in advertising, influencing and other such campaigns. In this paper we propose ViralBERT, which can be used to predict the virality of tweets using content- and user-based features. We employ a method of concatenating numerical features such as hashtags and follower numbers to tweet text, and utilise two BERT modules: one for semantic representation of the combined text and numerical features, and another module purely for sentiment analysis of text, as both the information within text and it's ability to elicit an emotional response play a part in retweet proneness. We collect a dataset of 330k tweets to train ViralBERT and validate the efficacy of our model using baselines from current studies in this field. Our experiments show that our approach outperforms these baselines, with a 13% increase in both F1 Score and Accuracy compared to the best performing baseline method. We then undergo an ablation study to investigate the importance of chosen features, finding that text sentiment and follower counts, and to a lesser extent mentions and following counts, are the strongest features for the model, and that hashtag counts are detrimental to the model."
                    },
                    {
                        "title": "Sample Efficient Model Evaluation",
                        "abstract": "Labelling data is a major practical bottleneck in training and testing classifiers. Given a collection of unlabelled data points, we address how to select which subset to label to best estimate test metrics such as accuracy, $F_1$ score or micro/macro $F_1$. We consider two sampling based approaches, namely the well-known Importance Sampling and we introduce a novel application of Poisson Sampling. For both approaches we derive the minimal error sampling distributions and how to approximate and use them to form estimators and confidence intervals. We show that Poisson Sampling outperforms Importance Sampling both theoretically and experimentally."
                    },
                    {
                        "title": "Overview of the TREC 2020 deep learning track",
                        "abstract": "This is the second year of the TREC Deep Learning Track, with the goal of studying ad hoc ranking in the large training data regime. We again have a document retrieval task and a passage retrieval task, each with hundreds of thousands of human-labeled training queries. We evaluate using single-shot TREC-style evaluation, to give us a picture of which ranking methods work best when large data is available, with much more comprehensive relevance labeling on the small number of test queries. This year we have further evidence that rankers with BERT-style pretraining outperform other rankers in the large data regime."
                    },
                    {
                        "title": "Learning to Detect Few-Shot-Few-Clue Misinformation",
                        "abstract": "The quality of digital information on the web has been disquieting due to the lack of careful manual review. Consequently, a large volume of false textual information has been disseminating for a long time since the prevalence of social media. The potential negative influence of misinformation on the public is a growing concern. Therefore, it is strongly motivated to detect online misinformation as early as possible. Few-shot-few-clue learning applies in this misinformation detection task when the number of annotated statements is quite few (called few shots) and the corresponding evidence is also quite limited in each shot (called few clues). Within the few-shot-few-clue framework, we propose a Bayesian meta-learning algorithm to extract the shared patterns among different topics (i.e.different tasks) of misinformation. Moreover, we derive a scalable method, i.e., amortized variational inference, to optimize the Bayesian meta-learning algorithm. Empirical results on three benchmark datasets demonstrate the superiority of our algorithm. This work focuses more on optimizing parameters than designing detection models, and will generate fresh insights into data-efficient detection of online misinformation at early stages."
                    }
                ]
            },
            "16d580d3-8630-4bad-a4bf-309dd14d0466": {
                "pk": "16d580d3-8630-4bad-a4bf-309dd14d0466",
                "name": "Aldo Lipani",
                "collaborators": [
                    "Zhengxiang Shi",
                    "Emine Yilmaz",
                    "Qiang Zhang",
                    "Yue Feng",
                    "Ramin Okhrati",
                    "Xi Wang",
                    "Hossein A. Rahmani",
                    "Florina Piroi",
                    "Jerome Ramos",
                    "Bin Wu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Conversational AI",
                    "Information Retrieval"
                ],
                "publications": [
                    {
                        "title": "Towards More Accountable Search Engines: Online Evaluation of Representation Bias",
                        "abstract": "Information availability affects people's behavior and perception of the world. Notably, people rely on search engines to satisfy their need for information. Search engines deliver results relevant to user requests usually without being or making themselves accountable for the information they deliver, which may harm people's lives and, in turn, society. This potential risk urges the development of evaluation mechanisms of bias in order to empower the user in judging the results of search engines. In this paper, we give a possible solution to measuring representation bias with respect to societal features for search engines and apply it to evaluating the gender representation bias for Google's Knowledge Graph Carousel for listing occupations."
                    },
                    {
                        "title": "A Multilinear Sampling Algorithm to Estimate Shapley Values",
                        "abstract": "Shapley values are great analytical tools in game theory to measure the importance of a player in a game. Due to their axiomatic and desirable properties such as efficiency, they have become popular for feature importance analysis in data science and machine learning. However, the time complexity to compute Shapley values based on the original formula is exponential, and as the number of features increases, this becomes infeasible. Castro et al. [1] developed a sampling algorithm, to estimate Shapley values. In this work, we propose a new sampling method based on a multilinear extension technique as applied in game theory. The aim is to provide a more efficient (sampling) method for estimating Shapley values. Our method is applicable to any machine learning model, in particular for either multi-class classifications or regression problems. We apply the method to estimate Shapley values for multilayer perceptrons (MLPs) and through experimentation on two datasets, we demonstrate that our method provides more accurate estimations of the Shapley values by reducing the variance of the sampling statistics."
                    },
                    {
                        "title": "Rethink the Effectiveness of Text Data Augmentation: An Empirical Analysis",
                        "abstract": "In recent years, language models (LMs) have made remarkable progress in advancing the field of natural language processing (NLP). However, the impact of data augmentation (DA) techniques on the fine-tuning (FT) performance of these LMs has been a topic of ongoing debate. In this study, we evaluate the effectiveness of three different FT methods in conjugation with back-translation across an array of 7 diverse NLP tasks, including classification and regression types, covering single-sentence and sentence-pair tasks. Contrary to prior assumptions that DA does not contribute to the enhancement of LMs' FT performance, our findings reveal that continued pre-training on augmented data can effectively improve the FT performance of the downstream tasks. In the most favourable case, continued pre-training improves the performance of FT by more than 10% in the few-shot learning setting. Our finding highlights the potential of DA as a powerful tool for bolstering LMs' performance."
                    },
                    {
                        "title": "Don't Stop Pretraining? Make Prompt-based Fine-tuning Powerful Learner",
                        "abstract": "Language models (LMs) trained on vast quantities of unlabelled data have greatly advanced the field of natural language processing (NLP). In this study, we re-visit the widely accepted notion in NLP that continued pre-training LMs on task-related texts improves the performance of fine-tuning (FT) in downstream tasks. Through experiments on eight single-sentence tasks and eight sentence-pair tasks in both semi-supervised and fully-supervised settings, we find that conventional continued pre-training does not consistently provide benefits and can even be detrimental for sentence-pair tasks or when prompt-based FT is used. To tackle these issues, we propose Prompt-based Continued Pre-training (PCP), which combines the idea of instruction tuning with conventional continued pre-training. Our approach aims to improve the performance of prompt-based FT by presenting both task-related texts and prompt templates to LMs through unsupervised pre-training objectives before fine-tuning for the target task. Our empirical evaluations on 21 benchmarks demonstrate that the PCP consistently improves the performance of state-of-the-art prompt-based FT approaches (up to 20.1% absolute) in both semi-supervised and fully-supervised settings, even with only hundreds of unlabelled examples. Additionally, prompt-based FT with the PCP outperforms state-of-the-art semi-supervised approaches with greater simplicity, eliminating the need for an iterative process and extra data augmentation. Our further analysis explores the performance lower bound of the PCP and reveals that the advantages of PCP persist across different sizes of models and datasets."
                    },
                    {
                        "title": "DePT: Decomposed Prompt Tuning for Parameter-Efficient Fine-tuning",
                        "abstract": "Prompt tuning (PT), where a small amount of trainable soft (continuous) prompt vectors is affixed to the input of language models (LM), has shown promising results across various tasks and models for parameter-efficient fine-tuning (PEFT). PT stands out from other PEFT approaches because it maintains competitive performance with fewer trainable parameters and does not drastically scale up its parameters as the model size expands. However, PT introduces additional soft prompt tokens, leading to longer input sequences, which significantly impacts training and inference time and memory usage due to the Transformer's quadratic complexity. Particularly concerning for Large Language Models (LLMs) that face heavy daily querying. To address this issue, we propose Decomposed Prompt Tuning (DePT), which decomposes the soft prompt into a shorter soft prompt and a pair of low-rank matrices that are then optimised with two different learning rates. This allows DePT to achieve better performance while saving substantial memory and time costs compared to vanilla PT and its variants, without changing trainable parameter sizes. Through extensive experiments on 23 natural language processing (NLP) and vision-language (VL) tasks, we demonstrate that DePT outperforms state-of-the-art PEFT approaches, including the full fine-tuning baseline, in some scenarios. Additionally, we empirically show that DEPT grows more efficient as the model size increases. Our further study reveals that DePT integrates seamlessly with parameter-efficient transfer learning in the few-shot learning setting and highlights its adaptability to various model architectures and sizes."
                    },
                    {
                        "title": "StepGame: A New Benchmark for Robust Multi-Hop Spatial Reasoning in Texts",
                        "abstract": "Inferring spatial relations in natural language is a crucial ability an intelligent system should possess. The bAbI dataset tries to capture tasks relevant to this domain (task 17 and 19). However, these tasks have several limitations. Most importantly, they are limited to fixed expressions, they are limited in the number of reasoning steps required to solve them, and they fail to test the robustness of models to input that contains irrelevant or redundant information. In this paper, we present a new Question-Answering dataset called StepGame for robust multi-hop spatial reasoning in texts. Our experiments demonstrate that state-of-the-art models on the bAbI dataset struggle on the StepGame dataset. Moreover, we propose a Tensor-Product based Memory-Augmented Neural Network (TP-MANN) specialized for spatial reasoning tasks. Experimental results on both datasets show that our model outperforms all the baselines with superior generalization and robustness performance."
                    },
                    {
                        "title": "Learning to Execute Actions or Ask Clarification Questions",
                        "abstract": "Collaborative tasks are ubiquitous activities where a form of communication is required in order to reach a joint goal. Collaborative building is one of such tasks. We wish to develop an intelligent builder agent in a simulated building environment (Minecraft) that can build whatever users wish to build by just talking to the agent. In order to achieve this goal, such agents need to be able to take the initiative by asking clarification questions when further information is needed. Existing works on Minecraft Corpus Dataset only learn to execute instructions neglecting the importance of asking for clarifications. In this paper, we extend the Minecraft Corpus Dataset by annotating all builder utterances into eight types, including clarification questions, and propose a new builder agent model capable of determining when to ask or execute instructions. Experimental results show that our model achieves state-of-the-art performance on the collaborative building task with a substantial improvement. We also define two new tasks, the learning to ask task and the joint learning task. The latter consists of solving both collaborating building and learning to ask tasks jointly."
                    },
                    {
                        "title": "Preference Distillation for Personalized Generative Recommendation",
                        "abstract": "Recently, researchers have investigated the capabilities of Large Language Models (LLMs) for generative recommender systems. Existing LLM-based recommender models are trained by adding user and item IDs to a discrete prompt template. However, the disconnect between IDs and natural language makes it difficult for the LLM to learn the relationship between users. To address this issue, we propose a PErsonAlized PrOmpt Distillation (PeaPOD) approach, to distill user preferences as personalized soft prompts. Considering the complexities of user preferences in the real world, we maintain a shared set of learnable prompts that are dynamically weighted based on the user's interests to construct the user-personalized prompt in a compositional manner. Experimental results on three real-world datasets demonstrate the effectiveness of our PeaPOD model on sequential recommendation, top-n recommendation, and explanation generation tasks."
                    },
                    {
                        "title": "Self-Attentive Hawkes Processes",
                        "abstract": "Asynchronous events on the continuous time domain, e.g., social media actions and stock transactions, occur frequently in the world. The ability to recognize occurrence patterns of event sequences is crucial to predict which typeof events will happen next and when. A de facto standard mathematical framework to do this is the Hawkes process. In order to enhance expressivity of multivariate Hawkes processes, conventional statistical methods and deep recurrent networks have been employed to modify its intensity function. The former is highly interpretable and requires small size of training data but relies on correct model design while the latter has less dependency on prior knowledge and is more powerful in capturing complicated patterns. We leverage pros and cons of these models and propose a self-attentive Hawkes process(SAHP). The proposed method adapts self-attention to fit the intensity function of Hawkes processes. This design has two benefits:(1) compared with conventional statistical methods, the SAHP is more powerful to identify complicated dependency relationships between temporal events; (2)compared with deep recurrent networks, the self-attention mechanism is able to capture longer historical information, and is more interpretable because the learnt attention weight tensor shows contributions of each historical event. Experiments on four real-world datasets demonstrate the effectiveness of the proposed method."
                    },
                    {
                        "title": "Self Contrastive Learning for Session-based Recommendation",
                        "abstract": "Session-based recommendation, which aims to predict the next item of users' interest as per an existing sequence interaction of items, has attracted growing applications of Contrastive Learning (CL) with improved user and item representations. However, these contrastive objectives: (1) serve a similar role as the cross-entropy loss while ignoring the item representation space optimisation; and (2) commonly require complicated modelling, including complex positive/negative sample constructions and extra data augmentation. In this work, we introduce Self-Contrastive Learning (SCL), which simplifies the application of CL and enhances the performance of state-of-the-art CL-based recommendation techniques. Specifically, SCL is formulated as an objective function that directly promotes a uniform distribution among item representations and efficiently replaces all the existing contrastive objective components of state-of-the-art models. Unlike previous works, SCL eliminates the need for any positive/negative sample construction or data augmentation, leading to enhanced interpretability of the item representation space and facilitating its extensibility to existing recommender systems. Through experiments on three benchmark datasets, we demonstrate that SCL consistently improves the performance of state-of-the-art models with statistical significance. Notably, our experiments show that SCL improves the performance of two best-performing models by 8.2% and 9.5% in P@10 (Precision) and 9.9% and 11.2% in MRR@10 (Mean Reciprocal Rank) on average across different benchmarks. Additionally, our analysis elucidates the improvement in terms of alignment and uniformity of representations, as well as the effectiveness of SCL with a low computational cost."
                    },
                    {
                        "title": "Lexical Entrainment for Conversational Systems",
                        "abstract": "Conversational agents have become ubiquitous in assisting with daily tasks, and are expected to possess human-like features. One such feature is lexical entrainment (LE), a phenomenon in which speakers in human-human conversations tend to naturally and subconsciously align their lexical choices with those of their interlocutors, leading to more successful and engaging conversations. As an example, if a digital assistant replies 'Your appointment for Jinling Noodle Pub is at 7 pm' to the question 'When is my reservation for Jinling Noodle Bar today?', it may feel as though the assistant is trying to correct the speaker, whereas a response of 'Your reservation for Jinling Noodle Bar is at 7 pm' would likely be perceived as more positive. This highlights the importance of LE in establishing a shared terminology for maximum clarity and reducing ambiguity in conversations. However, we demonstrate in this work that current response generation models do not adequately address this crucial humanlike phenomenon. To address this, we propose a new dataset, named MULTIWOZ-ENTR, and a measure for LE for conversational systems. Additionally, we suggest a way to explicitly integrate LE into conversational systems with two new tasks, a LE extraction task and a LE generation task. We also present two baseline approaches for the LE extraction task, which aim to detect LE expressions from dialogue contexts."
                    },
                    {
                        "title": "Do Sentence Transformers Learn Quasi-Geospatial Concepts from General Text?",
                        "abstract": "Sentence transformers are language models designed to perform semantic search. This study investigates the capacity of sentence transformers, fine-tuned on general question-answering datasets for asymmetric semantic search, to associate descriptions of human-generated routes across Great Britain with queries often used to describe hiking experiences. We find that sentence transformers have some zero-shot capabilities to understand quasi-geospatial concepts, such as route types and difficulty, suggesting their potential utility for routing recommendation systems."
                    },
                    {
                        "title": "Dynamic Schema Graph Fusion Network for Multi-Domain Dialogue State Tracking",
                        "abstract": "Dialogue State Tracking (DST) aims to keep track of users' intentions during the course of a conversation. In DST, modelling the relations among domains and slots is still an under-studied problem. Existing approaches that have considered such relations generally fall short in: (1) fusing prior slot-domain membership relations and dialogue-aware dynamic slot relations explicitly, and (2) generalizing to unseen domains. To address these issues, we propose a novel \\textbf{D}ynamic \\textbf{S}chema \\textbf{G}raph \\textbf{F}usion \\textbf{Net}work (\\textbf{DSGFNet}), which generates a dynamic schema graph to explicitly fuse the prior slot-domain membership relations and dialogue-aware dynamic slot relations. It also uses the schemata to facilitate knowledge transfer to new domains. DSGFNet consists of a dialogue utterance encoder, a schema graph encoder, a dialogue-aware schema graph evolving network, and a schema graph enhanced dialogue state decoder. Empirical results on benchmark datasets (i.e., SGD, MultiWOZ2.1, and MultiWOZ2.2), show that DSGFNet outperforms existing methods."
                    },
                    {
                        "title": "Towards Asking Clarification Questions for Information Seeking on Task-Oriented Dialogues",
                        "abstract": "Task-oriented dialogue systems aim at providing users with task-specific services. Users of such systems often do not know all the information about the task they are trying to accomplish, requiring them to seek information about the task. To provide accurate and personalized task-oriented information seeking results, task-oriented dialogue systems need to address two potential issues: 1) users' inability to describe their complex information needs in their requests; and 2) ambiguous/missing information the system has about the users. In this paper, we propose a new Multi-Attention Seq2Seq Network, named MAS2S, which can ask questions to clarify the user's information needs and the user's profile in task-oriented information seeking. We also extend an existing dataset for task-oriented information seeking, leading to the \\ourdataset which contains about 100k task-oriented information seeking dialogues that are made publicly available\\footnote{Dataset and code is available at \\href{https://github.com/sweetalyssum/clarit}{https://github.com/sweetalyssum/clarit}.}. Experimental results on \\ourdataset show that MAS2S outperforms baselines on both clarification question generation and answer prediction."
                    },
                    {
                        "title": "A Baseline for Shapley Values in MLPs: from Missingness to Neutrality",
                        "abstract": "Deep neural networks have gained momentum based on their accuracy, but their interpretability is often criticised. As a result, they are labelled as black boxes. In response, several methods have been proposed in the literature to explain their predictions. Among the explanatory methods, Shapley values is a feature attribution method favoured for its robust theoretical foundation. However, the analysis of feature attributions using Shapley values requires choosing a baseline that represents the concept of missingness. An arbitrary choice of baseline could negatively impact the explanatory power of the method and possibly lead to incorrect interpretations. In this paper, we present a method for choosing a baseline according to a neutrality value: as a parameter selected by decision-makers, the point at which their choices are determined by the model predictions being either above or below it. Hence, the proposed baseline is set based on a parameter that depends on the actual use of the model. This procedure stands in contrast to how other baselines are set, i.e. without accounting for how the model is used. We empirically validate our choice of baseline in the context of binary classification tasks, using two datasets: a synthetic dataset and a dataset derived from the financial domain."
                    },
                    {
                        "title": "Mitigating the Position Bias of Transformer Models in Passage Re-Ranking",
                        "abstract": "Supervised machine learning models and their evaluation strongly depends on the quality of the underlying dataset. When we search for a relevant piece of information it may appear anywhere in a given passage. However, we observe a bias in the position of the correct answer in the text in two popular Question Answering datasets used for passage re-ranking. The excessive favoring of earlier positions inside passages is an unwanted artefact. This leads to three common Transformer-based re-ranking models to ignore relevant parts in unseen passages. More concerningly, as the evaluation set is taken from the same biased distribution, the models overfitting to that bias overestimate their true effectiveness. In this work we analyze position bias on datasets, the contextualized representations, and their effect on retrieval results. We propose a debiasing method for retrieval datasets. Our results show that a model trained on a position-biased dataset exhibits a significant decrease in re-ranking effectiveness when evaluated on a debiased dataset. We demonstrate that by mitigating the position bias, Transformer-based re-ranking models are equally effective on a biased and debiased dataset, as well as more effective in a transfer-learning setting between two differently biased datasets."
                    },
                    {
                        "title": "Evaluation Metrics for Measuring Bias in Search Engine Results",
                        "abstract": "Search engines decide what we see for a given search query. Since many people are exposed to information through search engines, it is fair to expect that search engines are neutral. However, search engine results do not necessarily cover all the viewpoints of a search query topic, and they can be biased towards a specific view since search engine results are returned based on relevance, which is calculated using many features and sophisticated algorithms where search neutrality is not necessarily the focal point. Therefore, it is important to evaluate the search engine results with respect to bias. In this work we propose novel web search bias evaluation measures which take into account the rank and relevance. We also propose a framework to evaluate web search bias using the proposed measures and test our framework on two popular search engines based on 57 controversial query topics such as abortion, medical marijuana, and gay marriage. We measure the stance bias (in support or against), as well as the ideological bias (conservative or liberal). We observe that the stance does not necessarily correlate with the ideological leaning, e.g. a positive stance on abortion indicates a liberal leaning but a positive stance on Cuba embargo indicates a conservative leaning. Our experiments show that neither of the search engines suffers from stance bias. However, both search engines suffer from ideological bias, both favouring one ideological leaning to the other, which is more significant from the perspective of polarisation in our society."
                    },
                    {
                        "title": "Application of Zone Method based Physics-Informed Neural Networks in Reheating Furnaces",
                        "abstract": "Foundation Industries (FIs) constitute glass, metals, cement, ceramics, bulk chemicals, paper, steel, etc. and provide crucial, foundational materials for a diverse set of economically relevant industries: automobiles, machinery, construction, household appliances, chemicals, etc. Reheating furnaces within the manufacturing chain of FIs are energy-intensive. Accurate and real-time prediction of underlying temperatures in reheating furnaces has the potential to reduce the overall heating time, thereby controlling the energy consumption for achieving the Net-Zero goals in FIs. In this paper, we cast this prediction as a regression task and explore neural networks due to their inherent capability of being effective and efficient, given adequate data. However, due to the infeasibility of achieving good-quality real data in scenarios like reheating furnaces, classical Hottel's zone method based computational model has been used to generate data for model training. To further enhance the Out-Of-Distribution generalization capability of the trained model, we propose a Physics-Informed Neural Network (PINN) by incorporating prior physical knowledge using a set of novel Energy-Balance regularizers."
                    },
                    {
                        "title": "Attention-based Ingredient Phrase Parser",
                        "abstract": "As virtual personal assistants have now penetrated the consumer market, with products such as Siri and Alexa, the research community has produced several works on task-oriented dialogue tasks such as hotel booking, restaurant booking, and movie recommendation. Assisting users to cook is one of these tasks that are expected to be solved by intelligent assistants, where ingredients and their corresponding attributes, such as name, unit, and quantity, should be provided to users precisely and promptly. However, existing ingredient information scraped from the cooking website is in the unstructured form with huge variation in the lexical structure, for example, '1 garlic clove, crushed', and '1 (8 ounce) package cream cheese, softened', making it difficult to extract information exactly. To provide an engaged and successful conversational service to users for cooking tasks, we propose a new ingredient parsing model that can parse an ingredient phrase of recipes into the structure form with its corresponding attributes with over 0.93 F1-score. Experimental results show that our model achieves state-of-the-art performance on AllRecipes and Food.com datasets."
                    },
                    {
                        "title": "A Survey on Asking Clarification Questions Datasets in Conversational Systems",
                        "abstract": "The ability to understand a user's underlying needs is critical for conversational systems, especially with limited input from users in a conversation. Thus, in such a domain, Asking Clarification Questions (ACQs) to reveal users' true intent from their queries or utterances arise as an essential task. However, it is noticeable that a key limitation of the existing ACQs studies is their incomparability, from inconsistent use of data, distinct experimental setups and evaluation strategies. Therefore, in this paper, to assist the development of ACQs techniques, we comprehensively analyse the current ACQs research status, which offers a detailed comparison of publicly available datasets, and discusses the applied evaluation metrics, joined with benchmarks for multiple ACQs-related tasks. In particular, given a thorough analysis of the ACQs task, we discuss a number of corresponding research directions for the investigation of ACQs as well as the development of conversational systems."
                    }
                ]
            }
        }
    },
    "2309.14068": {
        "paper_data": {
            "title": "Soft Mixture Denoising: Beyond the Expressive Bottleneck of Diffusion Models",
            "url": "http://arxiv.org/abs/2309.14068v3",
            "arxiv_id": "2309.14068",
            "authors": [
                "Yangming Li",
                "Boris van Breugel",
                "Mihaela van der Schaar"
            ],
            "abstract": "Because diffusion models have shown impressive performances in a number of tasks, such as image synthesis, there is a trend in recent works to prove (with certain assumptions) that these models have strong approximation capabilities. In this paper, we show that current diffusion models actually have an expressive bottleneck in backward denoising and some assumption made by existing theoretical guarantees is too strong. Based on this finding, we prove that diffusion models have unbounded errors in both local and global denoising. In light of our theoretical studies, we introduce soft mixture denoising (SMD), an expressive and efficient model for backward denoising. SMD not only permits diffusion models to well approximate any Gaussian mixture distributions in theory, but also is simple and efficient for implementation. Our experiments on multiple image datasets show that SMD significantly improves different types of diffusion models (e.g., DDPM), espeically in the situation of few backward iterations.",
            "introduction": "   1 Introduction  Diffusion models (DMs)\u00a0(Sohl-Dickstein et\u00a0al., 2015) have become highly popular generative models for their impressive performance in many research domains\u2014including high-resolution image synthesis\u00a0(Dhariwal & Nichol, 2021), natural language generation\u00a0(Li et\u00a0al., 2022), speech processing\u00a0(Kong et\u00a0al., 2021), and medical image analysis\u00a0(Pinaya et\u00a0al., 2022).  Current strong approximator theorems. To explain the effectiveness of diffusion models, recent work\u00a0(Lee et\u00a0al., 2022a; b; Chen et\u00a0al., 2023) provided theoretical guarantees (with certain assumptions) to show that diffusion models can approximate a rich family of data distributions with arbitrarily small errors. For example, Chen et\u00a0al. (2023) proved that the generated samples from diffusion models converge (in distribution) to the real data under ideal conditions. Since it is generally intractable to analyze the non-convex optimization of neural networks, a potential weakness of these works is that they all supposed bounded score estimation errors, which means the prediction errors of denoising functions (i.e., reparameterized score functions) are bounded.  Our limited approximation theorems. In this work, we take a first step towards the opposite direction: Instead of explaining why diffusion models are highly effective, we show that their approximation capabilities are in fact limited and the assumption of bounded score estimation errors (made by existing theoretical guarantees) is too strong.  In particular, we show that current diffusion models suffer from an expressive bottleneck\u2014the Gaussian parameterization of backward probability p\u03b8\u2062(\ud835\udc31t\u22121\u2223\ud835\udc31t)subscript\ud835\udc5d\ud835\udf03conditionalsubscript\ud835\udc31\ud835\udc611subscript\ud835\udc31\ud835\udc61p_{\\theta}(\\mathbf{x}_{t-1}\\mid\\mathbf{x}_{t})italic_p start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( bold_x start_POSTSUBSCRIPT italic_t - 1 end_POSTSUBSCRIPT \u2223 bold_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) is not expressive enough to fit the (possibly multimodal) posterior probability q\u2062(\ud835\udc31t\u22121\u2223\ud835\udc31t)\ud835\udc5econditionalsubscript\ud835\udc31\ud835\udc611subscript\ud835\udc31\ud835\udc61q(\\mathbf{x}_{t-1}\\mid\\mathbf{x}_{t})italic_q ( bold_x start_POSTSUBSCRIPT italic_t - 1 end_POSTSUBSCRIPT \u2223 bold_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ). Following this, we prove that diffusion models have arbitrarily large denoising errors for approximating some common data distributions q\u2062(\ud835\udc310)\ud835\udc5esubscript\ud835\udc310q(\\mathbf{x}_{0})italic_q ( bold_x start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ) (e.g., Gaussian mixture), which indicates that some assumption of prior works\u2014bounded score estimation errors\u2014is too strong, which undermines their theoretical guarantees. Lastly and importantly, we prove that diffusion models will have an arbitrarily large error in matching the learnable backward process p\u03b8\u2062(\ud835\udc310:T)subscript\ud835\udc5d\ud835\udf03subscript\ud835\udc31normal-:0\ud835\udc47p_{\\theta}(\\mathbf{x}_{0:T})italic_p start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( bold_x start_POSTSUBSCRIPT 0 : italic_T end_POSTSUBSCRIPT ) with the predefined forward process q\u2062(\ud835\udc310:T)\ud835\udc5esubscript\ud835\udc31normal-:0\ud835\udc47q(\\mathbf{x}_{0:T})italic_q ( bold_x start_POSTSUBSCRIPT 0 : italic_T end_POSTSUBSCRIPT ), even though matching these is the very optimization objective of current diffusion models\u00a0(Ho et\u00a0al., 2020; Song et\u00a0al., 2021b). This finding indicates that diffusion models might fail to fit complex data distributions.      (a) Baseline: vanilla LDM; FID: 11.2911.2911.2911.29.     (b) Our model: LDM w/ SMD; FID: 6.856.856.856.85.    Figure 1: SMD improves quality and reduces the number of backward iterations. Results for CelebA-HQ 256\u00d7256256256256\\times 256256 \u00d7 256 with only 100100100100 backward iterations, for LDM with and without SDM. SDM achieves better realism and FID. Achieving the same FID with vanilla LDM would require 8\u00d78\\times8 \u00d7 more steps (see Fig. 3). Note that SMD differs from fast samplers (e.g., DDIM\u00a0(Song et\u00a0al., 2021a) and DPM\u00a0(Lu et\u00a0al., 2022)): while those methods focus on deterministic sampling and numerical stability, SMD improves the expressiveness of diffusion models.   Our method: Soft Mixture Denoising (SMD). In light of our theoretical findings, we propose Soft Mixture Denoising (SMD), which aims to represent the hidden mixture components of the posterior probability with a continuous relaxation. We prove that SMD permits",
            "references": [
                {
                    "title": "Convergence of score-based generative modeling for general data distributions",
                    "abstract": "Score-based generative modeling (SGM) has grown to be a hugely successful method for learning to generate samples from complex data distributions such as that of images and audio. It is based on evolving an SDE that transforms white noise into a sample from the learned distribution, using estimates of the score function, or gradient log-pdf. Previous convergence analyses for these methods have suffered either from strong assumptions on the data distribution or exponential dependencies, and hence fail to give efficient guarantees for the multimodal and non-smooth distributions that arise in practice and for which good empirical performance is observed. We consider a popular kind of SGM -- denoising diffusion models -- and give polynomial convergence guarantees for general data distributions, with no assumptions related to functional inequalities or smoothness. Assuming $L^2$-accurate score estimates, we obtain Wasserstein distance guarantees for any distribution of bounded support or sufficiently decaying tails, as well as TV guarantees for distributions with further smoothness assumptions."
                },
                {
                    "title": "Sampling is as easy as learning the score: theory for diffusion models with minimal data assumptions",
                    "abstract": "We provide theoretical convergence guarantees for score-based generative models (SGMs) such as denoising diffusion probabilistic models (DDPMs), which constitute the backbone of large-scale real-world generative models such as DALL$\\cdot$E 2. Our main result is that, assuming accurate score estimates, such SGMs can efficiently sample from essentially any realistic data distribution. In contrast to prior works, our results (1) hold for an $L^2$-accurate score estimate (rather than $L^\\infty$-accurate); (2) do not require restrictive functional inequality conditions that preclude substantial non-log-concavity; (3) scale polynomially in all relevant problem parameters; and (4) match state-of-the-art complexity guarantees for discretization of the Langevin diffusion, provided that the score error is sufficiently small. We view this as strong theoretical justification for the empirical success of SGMs. We also examine SGMs based on the critically damped Langevin diffusion (CLD). Contrary to conventional wisdom, we provide evidence that the use of the CLD does not reduce the complexity of SGMs."
                },
                {
                    "title": "Brain Imaging Generation with Latent Diffusion Models",
                    "abstract": "Deep neural networks have brought remarkable breakthroughs in medical image analysis. However, due to their data-hungry nature, the modest dataset sizes in medical imaging projects might be hindering their full potential. Generating synthetic data provides a promising alternative, allowing to complement training datasets and conducting medical image research at a larger scale. Diffusion models recently have caught the attention of the computer vision community by producing photorealistic synthetic images. In this study, we explore using Latent Diffusion Models to generate synthetic images from high-resolution 3D brain images. We used T1w MRI images from the UK Biobank dataset (N=31,740) to train our models to learn about the probabilistic distribution of brain images, conditioned on covariables, such as age, sex, and brain structure volumes. We found that our models created realistic data, and we could use the conditioning variables to control the data generation effectively. Besides that, we created a synthetic dataset with 100,000 brain images and made it openly available to the scientific community."
                },
                {
                    "title": "Convergence for score-based generative modeling with polynomial complexity",
                    "abstract": "Score-based generative modeling (SGM) is a highly successful approach for learning a probability distribution from data and generating further samples. We prove the first polynomial convergence guarantees for the core mechanic behind SGM: drawing samples from a probability density $p$ given a score estimate (an estimate of $\\nabla \\ln p$) that is accurate in $L^2(p)$. Compared to previous works, we do not incur error that grows exponentially in time or that suffers from a curse of dimensionality. Our guarantee works for any smooth distribution and depends polynomially on its log-Sobolev constant. Using our guarantee, we give a theoretical analysis of score-based generative modeling, which transforms white-noise input into samples from a learned data distribution given score estimates at different noise scales. Our analysis gives theoretical grounding to the observation that an annealed procedure is required in practice to generate good samples, as our proof depends essentially on using annealing to obtain a warm start at each step. Moreover, we show that a predictor-corrector algorithm gives better convergence than using either portion alone."
                },
                {
                    "title": "DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps",
                    "abstract": "Diffusion probabilistic models (DPMs) are emerging powerful generative models. Despite their high-quality generation performance, DPMs still suffer from their slow sampling as they generally need hundreds or thousands of sequential function evaluations (steps) of large neural networks to draw a sample. Sampling from DPMs can be viewed alternatively as solving the corresponding diffusion ordinary differential equations (ODEs). In this work, we propose an exact formulation of the solution of diffusion ODEs. The formulation analytically computes the linear part of the solution, rather than leaving all terms to black-box ODE solvers as adopted in previous works. By applying change-of-variable, the solution can be equivalently simplified to an exponentially weighted integral of the neural network. Based on our formulation, we propose DPM-Solver, a fast dedicated high-order solver for diffusion ODEs with the convergence order guarantee. DPM-Solver is suitable for both discrete-time and continuous-time DPMs without any further training. Experimental results show that DPM-Solver can generate high-quality samples in only 10 to 20 function evaluations on various datasets. We achieve 4.70 FID in 10 function evaluations and 2.87 FID in 20 function evaluations on the CIFAR10 dataset, and a $4\\sim 16\\times$ speedup compared with previous state-of-the-art training-free samplers on various datasets."
                },
                {
                    "title": "Diffusion-LM Improves Controllable Text Generation",
                    "abstract": "Controlling the behavior of language models (LMs) without re-training is a major open problem in natural language generation. While recent works have demonstrated successes on controlling simple sentence attributes (e.g., sentiment), there has been little progress on complex, fine-grained controls (e.g., syntactic structure). To address this challenge, we develop a new non-autoregressive language model based on continuous diffusions that we call Diffusion-LM. Building upon the recent successes of diffusion models in continuous domains, Diffusion-LM iteratively denoises a sequence of Gaussian vectors into word vectors, yielding a sequence of intermediate latent variables. The continuous, hierarchical nature of these intermediate variables enables a simple gradient-based algorithm to perform complex, controllable generation tasks. We demonstrate successful control of Diffusion-LM for six challenging fine-grained control tasks, significantly outperforming prior work."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "DiffWave: A Versatile Diffusion Model for Audio Synthesis",
                    "abstract": "In this work, we propose DiffWave, a versatile Diffusion probabilistic model for conditional and unconditional Waveform generation. The model is non-autoregressive, and converts the white noise signal into structured waveform through a Markov chain with a constant number of steps at synthesis. It is efficiently trained by optimizing a variant of variational bound on the data likelihood. DiffWave produces high-fidelity audios in Different Waveform generation tasks, including neural vocoding conditioned on mel spectrogram, class-conditional generation, and unconditional generation. We demonstrate that DiffWave matches a strong WaveNet vocoder in terms of speech quality~(MOS: 4.44 versus 4.43), while synthesizing orders of magnitude faster. In particular, it significantly outperforms autoregressive and GAN-based waveform models in the challenging unconditional generation task in terms of audio quality and sample diversity from various automatic and human evaluations."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Glow: Generative Flow with Invertible 1x1 Convolutions",
                    "abstract": "Flow-based generative models (Dinh et al., 2014) are conceptually attractive due to tractability of the exact log-likelihood, tractability of exact latent-variable inference, and parallelizability of both training and synthesis. In this paper we propose Glow, a simple type of generative flow using an invertible 1x1 convolution. Using our method we demonstrate a significant improvement in log-likelihood on standard benchmarks. Perhaps most strikingly, we demonstrate that a generative model optimized towards the plain log-likelihood objective is capable of efficient realistic-looking synthesis and manipulation of large images. The code for our model is available at this https URL"
                },
                {
                    "title": "HyperNetworks",
                    "abstract": "This work explores hypernetworks: an approach of using one network, also known as a hypernetwork, to generate the weights for another network. We apply hypernetworks to generate adaptive weights for recurrent networks. In this case, hypernetworks can be viewed as a relaxed form of weight-sharing across layers. In our implementation, hypernetworks are are trained jointly with the main network in an end-to-end fashion. Our main result is that hypernetworks can generate non-shared weights for LSTM and achieve state-of-the-art results on a variety of sequence modelling tasks including character-level language modelling, handwriting generation and neural machine translation, challenging the weight-sharing paradigm for recurrent networks."
                },
                {
                    "title": "Local Maxima in the Likelihood of Gaussian Mixture Models: Structural Results and Algorithmic Consequences",
                    "abstract": "We provide two fundamental results on the population (infinite-sample) likelihood function of Gaussian mixture models with $M \\geq 3$ components. Our first main result shows that the population likelihood function has bad local maxima even in the special case of equally-weighted mixtures of well-separated and spherical Gaussians. We prove that the log-likelihood value of these bad local maxima can be arbitrarily worse than that of any global optimum, thereby resolving an open question of Srebro (2007). Our second main result shows that the EM algorithm (or a first-order variant of it) with random initialization will converge to bad critical points with probability at least $1-e^{-\\Omega(M)}$. We further establish that a first-order variant of EM will not converge to strict saddle points almost surely, indicating that the poor performance of the first-order method can be attributed to the existence of bad local maxima rather than bad saddle points. Overall, our results highlight the necessity of careful initialization when using the EM algorithm in practice, even when applied in highly favorable settings."
                },
                {
                    "title": "LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop",
                    "abstract": "While there has been remarkable progress in the performance of visual recognition algorithms, the state-of-the-art models tend to be exceptionally data-hungry. Large labeled training datasets, expensive and tedious to produce, are required to optimize millions of parameters in deep network models. Lagging behind the growth in model capacity, the available datasets are quickly becoming outdated in terms of size and density. To circumvent this bottleneck, we propose to amplify human effort through a partially automated labeling scheme, leveraging deep learning with humans in the loop. Starting from a large set of candidate images for each category, we iteratively sample a subset, ask people to label them, classify the others with a trained model, split the set into positives, negatives, and unlabeled based on the classification confidence, and then iterate with the unlabeled set. To assess the effectiveness of this cascading procedure and enable further progress in visual recognition research, we construct a new image dataset, LSUN. It contains around one million labeled images for each of 10 scene categories and 20 object categories. We experiment with training popular convolutional networks and find that they achieve substantial performance gains when trained on this dataset."
                },
                {
                    "title": "Variational Inference with Normalizing Flows",
                    "abstract": "The choice of approximate posterior distribution is one of the core problems in variational inference. Most applications of variational inference employ simple families of posterior approximations in order to allow for efficient inference, focusing on mean-field or other simple structured approximations. This restriction has a significant impact on the quality of inferences made using variational methods. We introduce a new approach for specifying flexible, arbitrarily complex and scalable approximate posterior distributions. Our approximations are distributions constructed through a normalizing flow, whereby a simple initial density is transformed into a more complex one by applying a sequence of invertible transformations until a desired level of complexity is attained. We use this view of normalizing flows to develop categories of finite and infinitesimal flows and provide a unified view of approaches for constructing rich posterior approximations. We demonstrate that the theoretical advantages of having posteriors that better match the true posterior, combined with the scalability of amortized variational approaches, provides a clear improvement in performance and applicability of variational inference."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "Deep Learning Face Attributes in the Wild",
                    "abstract": "Predicting face attributes in the wild is challenging due to complex face variations. We propose a novel deep learning framework for attribute prediction in the wild. It cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently. LNet is pre-trained by massive general object categories for face localization, while ANet is pre-trained by massive face identities for attribute prediction. This framework not only outperforms the state-of-the-art with a large margin, but also reveals valuable facts on learning face representation. (1) It shows how the performances of face localization (LNet) and attribute prediction (ANet) can be improved by different pre-training strategies. (2) It reveals that although the filters of LNet are fine-tuned only with image-level attribute tags, their response maps over entire images have strong indication of face locations. This fact enables training LNet for face localization with only image-level annotations, but without face bounding boxes or landmarks, which are required by all attribute recognition works. (3) It also demonstrates that the high-level hidden neurons of ANet automatically discover semantic concepts after pre-training with massive face identities, and such concepts are significantly enriched after fine-tuning with attribute tags. Each attribute can be well explained with a sparse linear combination of these concepts."
                },
                {
                    "title": "On entropy approximation for Gaussian mixture random vectors",
                    "abstract": "For many practical probability density representations such as for the widely used Gaussian mixture densities, an analytic evaluation of the differential entropy is not possible and thus, approximate calculations are inevitable. For this purpose, the first contribution of this paper deals with a novel entropy approximation method for Gaussian mixture random vectors, which is based on a component-wise Taylor-series expansion of the logarithm of a Gaussian mixture and on a splitting method of Gaussian mixture components. The employed order of the Taylor-series expansion and the number of components used for splitting allows balancing between accuracy and computational demand. The second contribution is the determination of meaningful and efficiently to calculate lower and upper bounds of the entropy, which can be also used for approximation purposes. In addition, a refinement method for the more important upper bound is proposed in order to approach the true entropy value."
                },
                {
                    "title": "Robust Regression and Outlier Detection",
                    "abstract": "1. Introduction. 2. Simple Regression. 3. Multiple Regression. 4. The Special Case of One-Dimensional Location. 5. Algorithms. 6. Outlier Diagnostics. 7. Related Statistical Techniques. References. Table of Data Sets. Index."
                },
                {
                    "title": "On the Properties of Kullback-Leibler Divergence Between Gaussians",
                    "abstract": "Kullback-Leibler (KL) divergence is one of the most important divergence measures between probability distributions. In this paper, we investigate the properties of KL divergence between Gaussians. Firstly, for any two n-dimensional Gaussians N1 and N2, we find the supremum of KL(N1||N2) when KL(N2||N1) \u2264 for > 0. This reveals the approximate symmetry of small KL divergence between Gaussians. We also find the infimum of KL(N1||N2) when KL(N2||N1) \u2265M for M > 0. Secondly, for any three n-dimensional Gaussians N1,N2 and N3, we find a bound of KL(N1||N3) if KL(N1||N2) and KL(N2||N3) are bounded. This reveals that the KL divergence between Gaussians follows a relaxed triangle inequality. Importantly, all the bounds in the theorems presented in this paper are independent of the dimension n."
                },
                {
                    "title": "GENERATIVE ADVERSARIAL NETS",
                    "abstract": "Estimating individualized treatment effects (ITE) is a challenging task due to the need for an individual\u2019s potential outcomes to be learned from biased data and without having access to the counterfactuals. We propose a novel method for inferring ITE based on the Generative Adversarial Nets (GANs) framework. Our method, termed Generative Adversarial Nets for inference of Individualized Treatment Effects (GANITE), is motivated by the possibility that we can capture the uncertainty in the counterfactual distributions by attempting to learn them using a GAN. We generate proxies of the counterfactual outcomes using a counterfactual generator, G, and then pass these proxies to an ITE generator, I, in order to train it. By modeling both of these using the GAN framework, we are able to infer based on the factual data, while still accounting for the unseen counterfactuals. We test our method on three real-world datasets (with both binary and multiple treatments) and show that GANITE outperforms state-of-the-art methods."
                },
                {
                    "title": "Under review as a conference paper at ICLR 2016",
                    "abstract": "We describe a method for learning word embeddings with stochastic dimensionality. Our Infinite Skip-Gram (iSG) model specifies an energy-based joint distribution over a word vector, a context vector, and their dimensionality. By employing the same techniques used to make the Infinite Restricted Boltzmann Machine (C\u00f4t\u00e9 & Larochelle, 2015) tractable, we define vector dimensionality over a countably infinite domain, allowing vectors to grow as needed during training. After training, we find that the distribution over embedding dimensionality for a given word is highly interpretable and leads to an elegant probabilistic mechanism for word sense induction. We show qualitatively and quantitatively that the iSG produces parameter-efficient representations that are robust to language\u2019s inherent ambiguity."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "A Mathematical Theory of Communication",
                    "abstract": "This paper opened the new area the information theory. Before this paper, most people believed that the only way to make the error probability of transmission as small as desired is to reduce the data rate (such as a long repetition scheme). However, surprisingly this paper revealed that it does not need to reduce the data rate for achieving that much of small errors. It proved that we can get some positive data rate that has the same small error probability and also there is an upper bound of the data rate, which means we cannot achieve the data rate with any encoding scheme that has small enough error probability over the upper bound."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the expressiveness of diffusion models to better approximate complex data distributions, particularly in the context of multimodal posteriors?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the understanding of diffusion models and their limitations. By addressing the expressive bottleneck in current models, we can enhance their performance in various applications such as image synthesis, natural language generation, and medical image analysis. This research could lead to more robust generative models that can handle complex data distributions, thereby influencing future research directions and practical applications in generative modeling.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent limitations of the Gaussian parameterization used in current diffusion models, which restricts their ability to fit multimodal posteriors. Naive approaches may fail because they do not account for the complexity of the data distributions, leading to significant denoising errors. Additionally, the non-convex nature of neural network optimization complicates the analysis and improvement of these models, making it difficult to derive effective solutions that can overcome these expressive limitations.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on demonstrating the effectiveness of diffusion models under the assumption of bounded score estimation errors, which has limited the exploration of their expressive capabilities. Existing solutions have not adequately addressed the multimodal nature of data distributions, and the theoretical guarantees provided in prior works do not hold under more complex scenarios. Our approach differs by challenging these assumptions and proposing a new methodology (Soft Mixture Denoising) that aims to enhance the expressiveness of diffusion models.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Soft Mixture Denoising (SMD), utilizes a continuous relaxation to represent the hidden mixture components of the posterior probability. We will evaluate the performance of SMD using standard datasets, measuring the quality of generated samples with metrics such as the Fr\u00e9chet Inception Distance (FID). We expect that SMD will significantly reduce denoising errors and improve the expressiveness of diffusion models, leading to better performance in generating complex data distributions compared to traditional methods."
            }
        },
        "author_data": {
            "ddfff0f5-9395-41c8-9929-718d58e1fd9f": {
                "pk": "ddfff0f5-9395-41c8-9929-718d58e1fd9f",
                "name": "Yangming Li",
                "collaborators": [
                    "M. Schaar",
                    "Max Ruiz Luyten",
                    "Zhaozhi Qian"
                ],
                "domain": [
                    "Diffusion Models",
                    "Stochastic Differential Equations",
                    "Time Series Analysis",
                    "Data Synthesis"
                ],
                "publications": [
                    {
                        "title": "Risk-Sensitive Diffusion: Robustly Optimizing Diffusion Models with Noisy Samples",
                        "abstract": "Diffusion models are mainly studied on image data. However, non-image data (e.g., tabular data) are also prevalent in real applications and tend to be noisy due to some inevitable factors in the stage of data collection, degrading the generation quality of diffusion models. In this paper, we consider a novel problem setting where every collected sample is paired with a vector indicating the data quality: risk vector. This setting applies to many scenarios involving noisy data and we propose risk-sensitive SDE, a type of stochastic differential equation (SDE) parameterized by the risk vector, to address it. With some proper coefficients, risk-sensitive SDE can minimize the negative effect of noisy samples on the optimization of diffusion models. We conduct systematic studies for both Gaussian and non-Gaussian noise distributions, providing analytical forms of risk-sensitive SDE. To verify the effectiveness of our method, we have conducted extensive experiments on multiple tabular and time-series datasets, showing that risk-sensitive SDE permits a robust optimization of diffusion models with noisy samples and significantly outperforms previous baselines."
                    },
                    {
                        "title": "Risk-Sensitive Diffusion: Learning the Underlying Distribution from Noisy Samples",
                        "abstract": "While achieving remarkable performances, we show that diffusion models are fragile to the presence of noisy samples, limiting their potential in the vast amount of settings where, unlike image synthesis, we are not blessed with clean data. Motivated by our finding that such fragility originates from the distribution gaps between noisy and clean samples along the diffusion process, we introduce risk-sensitive SDE , a stochastic differential equation that is parameterized by the risk (i.e., data \u201cdirtiness\u201d) to adjust the distributions of noisy samples, reducing misguidance while bene-fiting from their contained information. The optimal expression for risk-sensitive SDE depends on the specific noise distribution, and we derive its parameterizations that minimize the misguidance of noisy samples for both Gaussian and general non-Gaussian perturbations. We conduct extensive experiments on both synthetic and real-world datasets (e.g., medical time series), showing that our model effectively recovers the clean data distribution from noisy samples, significantly outper-forming conditional generation baselines."
                    },
                    {
                        "title": "A Study of Posterior Stability for Time-Series Latent Diffusion",
                        "abstract": "Latent diffusion has demonstrated promising results in image generation and permits efficient sampling. However, this framework might suffer from the problem of posterior collapse when applied to time series. In this paper, we first show that posterior collapse will reduce latent diffusion to a variational autoencoder (VAE), making it less expressive. This highlights the importance of addressing this issue. We then introduce a principled method: dependency measure, that quantifies the sensitivity of a recurrent decoder to input variables. Using this tool, we confirm that posterior collapse significantly affects time-series latent diffusion on real datasets, and a phenomenon termed dependency illusion is also discovered in the case of shuffled time series. Finally, building on our theoretical and empirical studies, we introduce a new framework that extends latent diffusion and has a stable posterior. Extensive experiments on multiple real time-series datasets show that our new framework is free from posterior collapse and significantly outperforms previous baselines in time series synthesis."
                    },
                    {
                        "title": "Do Diffusion Models Suffer Error Propagation? Theoretical Analysis and Consistency Regularization",
                        "abstract": "While diffusion models have achieved promising performances in data synthesis, they might suffer error propagation because of their cascade structure , where the distributional mismatch spreads and magnifies through the chain of denoising modules. However, a strict analysis is expected since many sequential models such as Conditional Random Field (CRF) are free from error propagation. In this paper, we empirically and theoretically verify that diffusion models are indeed affected by error propagation and we then propose a regularization to address this problem. Our theoretical analysis reveals that the question can be reduced to whether every denoising module of the diffusion model is fault-tolerant. We derive insightful transition equations, indicating that the module can\u2019t recover from input errors and even propagates additional errors to the next module. Our analysis directly leads to a consistency regularization scheme for diffusion models, which explicitly reduces the distribution gap between forward and backward processes. We further introduce a bootstrapping algorithm to reduce the computation cost of the regularizer. Our experimental results on multiple image datasets show that our regularization effectively handles error propagation and significantly improves the performance of vanilla diffusion models."
                    },
                    {
                        "title": "On Error Propagation of Diffusion Models",
                        "abstract": "Although diffusion models (DMs) have shown promising performances in a number of tasks (e.g., speech synthesis and image generation), they might suffer from error propagation because of their sequential structure. However, this is not certain because some sequential models, such as Conditional Random Field (CRF), are free from this problem. To address this issue, we develop a theoretical framework to mathematically formulate error propagation in the architecture of DMs, The framework contains three elements, including modular error, cumulative error, and propagation equation. The modular and cumulative errors are related by the equation, which interprets that DMs are indeed affected by error propagation. Our theoretical study also suggests that the cumulative error is closely related to the generation quality of DMs. Based on this finding, we apply the cumulative error as a regularization term to reduce error propagation. Because the term is computationally intractable, we derive its upper bound and design a bootstrap algorithm to efficiently estimate the bound for optimization. We have conducted extensive experiments on multiple image datasets, showing that our proposed regularization reduces error propagation, significantly improves vanilla DMs, and outperforms previous baselines."
                    }
                ]
            },
            "ae007aea-b473-4a3f-9665-09787b149124": {
                "pk": "ae007aea-b473-4a3f-9665-09787b149124",
                "name": "Boris van Breugel",
                "collaborators": [
                    "M. Schaar",
                    "Nabeel Seedat",
                    "Zhaozhi Qian",
                    "Hao Sun",
                    "Tennison Liu",
                    "Alex J. Chan",
                    "Jonathan Crabb'e",
                    "Robert Davis",
                    "Nicolas Huynh",
                    "Fernando P\u00e9rez-Garc\u00eda",
                    "Sam Bond-Taylor",
                    "Pedro P. Sanchez",
                    "Daniel C. Castro",
                    "Harshita Sharma",
                    "Valentina Salvatelli",
                    "Maria T. A. Wetscherek",
                    "Hannah Richardson",
                    "M. Lungren",
                    "A. Nori",
                    "Javier Alvarez-Valle",
                    "O. Oktay",
                    "Maximilian Ilse",
                    "F. Imrie",
                    "Jonathan Crabbe",
                    "N. D. Truong",
                    "Trent Kyono",
                    "Jeroen Berrevoets",
                    "A. Alaa",
                    "Evgeny S. Saveliev",
                    "Daniel de Vassimon Manela",
                    "D. Errington",
                    "Thomas Fisher",
                    "Pasquale Minervini"
                ],
                "domain": [
                    "Generative Models",
                    "Machine Learning",
                    "Fairness",
                    "Synthetic Data"
                ],
                "publications": [
                    {
                        "title": "LaTable: Towards Large Tabular Models",
                        "abstract": "Tabular data is one of the most ubiquitous modalities, yet the literature on tabular generative foundation models is lagging far behind its text and vision counterparts. Creating such a model is hard, due to the heterogeneous feature spaces of different tabular datasets, tabular metadata (e.g. dataset description and feature headers), and tables lacking prior knowledge (e.g. feature order). In this work we propose LaTable: a novel tabular diffusion model that addresses these challenges and can be trained across different datasets. Through extensive experiments we find that LaTable outperforms baselines on in-distribution generation, and that finetuning LaTable can generate out-of-distribution datasets better with fewer samples. On the other hand, we explore the poor zero-shot performance of LaTable, and what it may teach us about building generative tabular foundation models with better zero- and few-shot generation capabilities."
                    },
                    {
                        "title": "Why Tabular Foundation Models Should Be a Research Priority",
                        "abstract": "Recent text and image foundation models are incredibly impressive, and these models are attracting an ever-increasing portion of research resources. In this position piece we aim to shift the ML research community's priorities ever so slightly to a different modality: tabular data. Tabular data is the dominant modality in many fields, yet it is given hardly any research attention and significantly lags behind in terms of scale and power. We believe the time is now to start developing tabular foundation models, or what we coin a Large Tabular Model (LTM). LTMs could revolutionise the way science and ML use tabular data: not as single datasets that are analyzed in a vacuum, but contextualized with respect to related datasets. The potential impact is far-reaching: from few-shot tabular models to automating data science; from out-of-distribution synthetic data to empowering multidisciplinary scientific discovery. We intend to excite reflections on the modalities we study, and convince some researchers to study large tabular models."
                    },
                    {
                        "title": "Curated LLM: Synergy of LLMs and Data Curation for tabular augmentation in ultra low-data regimes",
                        "abstract": "Machine Learning (ML) in low-data settings remains an underappreciated yet crucial problem. Hence, data augmentation methods to increase the sample size of datasets needed for ML are key to unlocking the transformative potential of ML in data-deprived regions and domains. Unfortunately, the limited training set constrains traditional tabular synthetic data generators in their ability to generate a large and diverse augmented dataset needed for ML tasks. To address this challenge, we introduce CLLM, which leverages the prior knowledge of Large Language Models (LLMs) for data augmentation in the low-data regime. However, not all the data generated by LLMs will improve downstream utility, as for any generative model. Consequently, we introduce a principled curation mechanism, leveraging learning dynamics, coupled with confidence and uncertainty metrics, to obtain a high-quality dataset. Empirically, on multiple real-world datasets, we demonstrate the superior performance of CLLM in the low-data regime compared to conventional generators. Additionally, we provide insights into the LLM generation and curation mechanism, shedding light on the features that enable them to output high-quality augmented datasets."
                    },
                    {
                        "title": "RadEdit: stress-testing biomedical vision models via diffusion image editing",
                        "abstract": "Biomedical imaging datasets are often small and biased, meaning that real-world performance of predictive models can be substantially lower than expected from internal testing. This work proposes using generative image editing to simulate dataset shifts and diagnose failure modes of biomedical vision models; this can be used in advance of deployment to assess readiness, potentially reducing cost and patient harm. Existing editing methods can produce undesirable changes, with spurious correlations learned due to the co-occurrence of disease and treatment interventions, limiting practical applicability. To address this, we train a text-to-image diffusion model on multiple chest X-ray datasets and introduce a new editing method RadEdit that uses multiple masks, if present, to constrain changes and ensure consistency in the edited images. We consider three types of dataset shifts: acquisition shift, manifestation shift, and population shift, and demonstrate that our approach can diagnose failures and quantify model robustness without additional data collection, complementing more qualitative tools for explainable AI."
                    },
                    {
                        "title": "Can You Rely on Your Model Evaluation? Improving Model Evaluation with Synthetic Test Data",
                        "abstract": "Evaluating the performance of machine learning models on diverse and underrepresented subgroups is essential for ensuring fairness and reliability in real-world applications. However, accurately assessing model performance becomes challenging due to two main issues: (1) a scarcity of test data, especially for small subgroups, and (2) possible distributional shifts in the model's deployment setting, which may not align with the available test data. In this work, we introduce 3S Testing, a deep generative modeling framework to facilitate model evaluation by generating synthetic test sets for small subgroups and simulating distributional shifts. Our experiments demonstrate that 3S Testing outperforms traditional baselines -- including real test data alone -- in estimating model performance on minority subgroups and under plausible distributional shifts. In addition, 3S offers intervals around its performance estimates, exhibiting superior coverage of the ground truth compared to existing approaches. Overall, these results raise the question of whether we need a paradigm shift away from limited real test data towards synthetic test data."
                    },
                    {
                        "title": "Synthetic data, real errors: how (not) to publish and use synthetic data",
                        "abstract": "Generating synthetic data through generative models is gaining interest in the ML community and beyond, promising a future where datasets can be tailored to individual needs. Unfortunately, synthetic data is usually not perfect, resulting in potential errors in downstream tasks. In this work we explore how the generative process affects the downstream ML task. We show that the naive synthetic data approach -- using synthetic data as if it is real -- leads to downstream models and analyses that do not generalize well to real data. As a first step towards better ML in the synthetic data regime, we introduce Deep Generative Ensemble (DGE) -- a framework inspired by Deep Ensembles that aims to implicitly approximate the posterior distribution over the generative process model parameters. DGE improves downstream model training, evaluation, and uncertainty quantification, vastly outperforming the naive approach on average. The largest improvements are achieved for minority classes and low-density regions of the original data, for which the generative uncertainty is largest."
                    },
                    {
                        "title": "Beyond Privacy: Navigating the Opportunities and Challenges of Synthetic Data",
                        "abstract": "Generating synthetic data through generative models is gaining interest in the ML community and beyond. In the past, synthetic data was often regarded as a means to private data release, but a surge of recent papers explore how its potential reaches much further than this -- from creating more fair data to data augmentation, and from simulation to text generated by ChatGPT. In this perspective we explore whether, and how, synthetic data may become a dominant force in the machine learning world, promising a future where datasets can be tailored to individual needs. Just as importantly, we discuss which fundamental challenges the community needs to overcome for wider relevance and application of synthetic data -- the most important of which is quantifying how much we can trust any finding or prediction drawn from synthetic data."
                    },
                    {
                        "title": "Membership Inference Attacks against Synthetic Data through Overfitting Detection",
                        "abstract": "Data is the foundation of most science. Unfortunately, sharing data can be obstructed by the risk of violating data privacy, impeding research in fields like healthcare. Synthetic data is a potential solution. It aims to generate data that has the same distribution as the original data, but that does not disclose information about individuals. Membership Inference Attacks (MIAs) are a common privacy attack, in which the attacker attempts to determine whether a particular real sample was used for training of the model. Previous works that propose MIAs against generative models either display low performance -- giving the false impression that data is highly private -- or need to assume access to internal generative model parameters -- a relatively low-risk scenario, as the data publisher often only releases synthetic data, not the model. In this work we argue for a realistic MIA setting that assumes the attacker has some knowledge of the underlying data distribution. We propose DOMIAS, a density-based MIA model that aims to infer membership by targeting local overfitting of the generative model. Experimentally we show that DOMIAS is significantly more successful at MIA than previous work, especially at attacking uncommon samples. The latter is disconcerting since these samples may correspond to underrepresented groups. We also demonstrate how DOMIAS' MIA performance score provides an interpretable metric for privacy, giving data publishers a new tool for achieving the desired privacy-utility trade-off in their synthetic data."
                    },
                    {
                        "title": "The Spherical Grasshopper Problem",
                        "abstract": "The aim of this essay is to better understand the Grasshopper Problem on the surface of the unit sphere. The problem is motivated by analysing Bell inequalities, but can be formulated as a geometric puzzle as follows. Given a white sphere and a bucket of black paint, one is asked to paint half of the sphere, such that antipodal pairs of points are oppositely coloured. A grasshopper lands on the sphere, and jumps a fixed distance in a random direction. How should the sphere be coloured such that the probability of the grasshopper landing on the same colour is maximized? Goulko and Kent have explored this problem on the plane without an antipodality constraint. This essay gives clear indication that the spherical problem with the antipodality constraint yields colourings with similar shapes as the planar problem does. This research has discretised the problem and used a simulated annealing algorithm to search for the optimal solution. Results are consistent with the planar results of \\cite{goulkokent}. For $0.10\\pi\\leq\\theta\\leq0.44\\pi$ cogwheel solutions are found to be optimal, with odd integer of cogs $n_o$ such that $n_o$ is close to $\\frac{2\\pi}{\\theta}$. For $0.45\\pi\\leq\\theta\\leq 0.55\\pi$ critical solutions are found, in which domains of identical colour decrease in size towards $0.5\\pi$ (moving from either side). For $\\theta\\geq 0.55$ colourings are found consisting of stripes with cogs. Towards $\\theta=\\pi$ colourings are generated that display just stripes that scale in width with $\\pi-\\theta$."
                    },
                    {
                        "title": "Practical Approaches for Fair Learning with Multitype and Multivariate Sensitive Attributes",
                        "abstract": "It is important to guarantee that machine learning algorithms deployed in the real world do not result in unfairness or unintended social consequences. Fair ML has largely focused on the protection of single attributes in the simpler setting where both attributes and target outcomes are binary. However, the practical application in many a real-world problem entails the simultaneous protection of multiple sensitive attributes, which are often not simply binary, but continuous or categorical. To address this more challenging task, we introduce FairCOCCO, a fairness measure built on cross-covariance operators on reproducing kernel Hilbert Spaces. This leads to two practical tools: first, the FairCOCCO Score, a normalised metric that can quantify fairness in settings with single or multiple sensitive attributes of arbitrary type; and second, a subsequent regularisation term that can be incorporated into arbitrary learning objectives to obtain fair predictors. These contributions address crucial gaps in the algorithmic fairness literature, and we empirically demonstrate consistent improvements against state-of-the-art techniques in balancing predictive power and fairness on real-world datasets."
                    },
                    {
                        "title": "What is Flagged in Uncertainty Quantification? Latent Density Models for Uncertainty Categorization",
                        "abstract": "Uncertainty Quantification (UQ) is essential for creating trustworthy machine learning models. Recent years have seen a steep rise in UQ methods that can flag suspicious examples, however, it is often unclear what exactly these methods identify. In this work, we propose a framework for categorizing uncertain examples flagged by UQ methods in classification tasks. We introduce the confusion density matrix -- a kernel-based approximation of the misclassification density -- and use this to categorize suspicious examples identified by a given uncertainty method into three classes: out-of-distribution (OOD) examples, boundary (Bnd) examples, and examples in regions of high in-distribution misclassification (IDM). Through extensive experiments, we show that our framework provides a new and distinct perspective for assessing differences between uncertainty quantification methods, thereby forming a valuable assessment benchmark."
                    },
                    {
                        "title": "Fair Policy Learning",
                        "abstract": "Ensuring machine learning algorithms deployed in the real world do not result in unexpected unfairness or social implications is becoming increasingly important. However, there exists a clear gap in literature for a measure of fairness that can detect discrimination against multiple sensitive attributes while also handling continuous or discrete outcomes. In this thesis, we propose a fairness measure, Fair-COCCO, based on the conditional cross-covariance operator on reproducing kernel Hilbert Spaces. This novel method generalise to the majority of existing fairness notions and naturally extends to settings with continuous outcomes and multidimensional sensitive attributes. Additionally, we demonstrate how the proposed measure can be readily implemented in stochastic gradient optimisation for fair policy learning in supervised learning settings. Empirical evaluations of Fair-COCCO on synthetic and realworld experiments reveal favourable comparisons to state-of-the-art techniques in balancing predictive power and fairness. We also see much potential in applying machine learning to analyse fairness in observed behaviour, especially in complex and high-dimensional real-world environments. To that end, we propose the first known definition of fairness for sequences of decisions and showcase how Fair-COCCO can be applied to quantify fairness in these problems. Building off these definitions, we turn to learning fair policies in real-world conditions, where learning is constrained to be performed offline. We propose Fair-PoLe, a novel inverse reinforcement learning that operates completely offline and is computationally efficient and functionally expressive when compared to existing methods. We illustrate the potential for Fair-PoLe to learn policies that balance imitation of expert policies with fair outcomes on the challenging problem of sepsis treatment."
                    },
                    {
                        "title": "DECAF: Generating Fair Synthetic Data Using Causally-Aware Generative Networks",
                        "abstract": "Machine learning models have been criticized for reflecting unfair biases in the training data. Instead of solving for this by introducing fair learning algorithms directly, we focus on generating fair synthetic data, such that any downstream learner is fair. Generating fair synthetic data from unfair data - while remaining truthful to the underlying data-generating process (DGP) - is non-trivial. In this paper, we introduce DECAF: a GAN-based fair synthetic data generator for tabular data. With DECAF we embed the DGP explicitly as a structural causal model in the input layers of the generator, allowing each variable to be reconstructed conditioned on its causal parents. This procedure enables inference time debiasing, where biased edges can be strategically removed for satisfying user-defined fairness requirements. The DECAF framework is versatile and compatible with several popular definitions of fairness. In our experiments, we show that DECAF successfully removes undesired bias and - in contrast to existing methods - is capable of generating high-quality synthetic data. Furthermore, we provide theoretical guarantees on the generator's convergence and the fairness of downstream models."
                    },
                    {
                        "title": "How Faithful is your Synthetic Data? Sample-level Metrics for Evaluating and Auditing Generative Models",
                        "abstract": "Devising domain- and model-agnostic evaluation metrics for generative models is an important and as yet unresolved problem. Most existing metrics, which were tailored solely to the image synthesis setup, exhibit a limited capacity for diagnosing the different modes of failure of generative models across broader application domains. In this paper, we introduce a 3-dimensional evaluation metric, ($\\alpha$-Precision, $\\beta$-Recall, Authenticity), that characterizes the fidelity, diversity and generalization performance of any generative model in a domain-agnostic fashion. Our metric unifies statistical divergence measures with precision-recall analysis, enabling sample- and distribution-level diagnoses of model fidelity and diversity. We introduce generalization as an additional, independent dimension (to the fidelity-diversity trade-off) that quantifies the extent to which a model copies training data -- a crucial performance indicator when modeling sensitive data with requirements on privacy. The three metric components correspond to (interpretable) probabilistic quantities, and are estimated via sample-level binary classification. The sample-level nature of our metric inspires a novel use case which we call model auditing, wherein we judge the quality of individual samples generated by a (black-box) model, discarding low-quality samples and hence improving the overall model performance in a post-hoc manner."
                    },
                    {
                        "title": "Stereotype and Skew: Quantifying Gender Bias in Pre-trained and Fine-tuned Language Models",
                        "abstract": "This paper proposes two intuitive metrics, skew and stereotype, that quantify and analyse the gender bias present in contextual language models when tackling the WinoBias pronoun resolution task. We find evidence that gender stereotype correlates approximately negatively with gender skew in out-of-the-box models, suggesting that there is a trade-off between these two forms of bias. We investigate two methods to mitigate bias. The first approach is an online method which is effective at removing skew at the expense of stereotype. The second, inspired by previous work on ELMo, involves the fine-tuning of BERT using an augmented gender-balanced dataset. We show that this reduces both skew and stereotype relative to its unaugmented fine-tuned counterpart. However, we find that existing gender bias benchmarks do not fully probe professional bias as pronoun resolution may be obfuscated by cross-correlations from other manifestations of gender prejudice."
                    },
                    {
                        "title": "On the Use of a Dual Mean Field Approach for DFT+NEGF Electronic Transport Modelling",
                        "abstract": "This thesis focuses on electronic transport modelling in molecular junctions. The most common technique for modelling electron transport is the combination of Density Functional Theory (DFT) and Non-Equilibrium Green's Function's (NEGF). Conventional DFT+NEGF methods are, however, invalid for correctly modelling junctions under a bias voltage. In 2015 Liu et al have proposed a Dual Mean Field (DMF) extension to DFT+NEGF that enables the correct modelling of non-equilibrium systems. In contrast to the conventional DFT method, the DMF approach utilises two electron density fields to correctly account for the bias-induced non-equilibrium effects. This research has implemented the proposed DMF in the existent BAND software of SCM. The thesis intends to motivate further development of the DMF method. Electron transport calculations are performed using a DMF adapted LDA exchange-correlation functional in a 3x3 Al-BDA-Al junction. The results are compared to calculations with a conventional VWN LDA XC functional. The DMF approach leads to a 20% mean increase in the HOMO-LUMO gap. This is an important and promising result, as conventional LDA methods are known to underestimate the HOMO-LUMO gap. Additionally the bias dependence of the transmission spectrum has been analysed. This yields similarly good stability as the conventional LDA method. These first results encourage additional testing and development of the DMF method. Next steps are the incorporation of relativistic effects and the implementation of more advanced XC functionals."
                    }
                ]
            },
            "b6d21871-8644-49db-b475-2669c569c58c": {
                "pk": "b6d21871-8644-49db-b475-2669c569c58c",
                "name": "Mihaela van der Schaar",
                "collaborators": [
                    "Samuel Holt",
                    "Nabeel Seedat",
                    "Krzysztof Kacprzyk",
                    "Tennison Liu",
                    "Max Ruiz Luyten",
                    "Nicolas Huynh",
                    "T. Pouplin",
                    "Jeroen Berrevoets",
                    "Zhaozhi Qian",
                    "Yangming Li",
                    "F. Imrie",
                    "Hao Sun",
                    "Jonas Schweisthal",
                    "Dennis Frauen",
                    "Stefan Feuerriegel",
                    "Nicol'as Astorga",
                    "Paulius Rauba",
                    "Evgeny S. Saveliev",
                    "Tim Schubert",
                    "V. Kosmoliaptsis",
                    "Jonathan Crabbe",
                    "Jan Stanczuk",
                    "Katarzyna Kobalczyk",
                    "Zahra Atashgahi",
                    "Mykola Pechenizkiy",
                    "Raymond Veldhuis",
                    "D. Mocanu",
                    "Chris Lu",
                    "Claudio Fanconi",
                    "Alex J. Chan",
                    "Jakob N. Foerster",
                    "R. T. Lange"
                ],
                "domain": [
                    "Machine Learning",
                    "Causality",
                    "Semi-Supervised Learning",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "You can't handle the (dirty) truth: Data-centric insights improve pseudo-labeling",
                        "abstract": "Pseudo-labeling is a popular semi-supervised learning technique to leverage unlabeled data when labeled samples are scarce. The generation and selection of pseudo-labels heavily rely on labeled data. Existing approaches implicitly assume that the labeled data is gold standard and 'perfect'. However, this can be violated in reality with issues such as mislabeling or ambiguity. We address this overlooked aspect and show the importance of investigating labeled data quality to improve any pseudo-labeling method. Specifically, we introduce a novel data characterization and selection framework called DIPS to extend pseudo-labeling. We select useful labeled and pseudo-labeled samples via analysis of learning dynamics. We demonstrate the applicability and impact of DIPS for various pseudo-labeling methods across an extensive range of real-world tabular and image datasets. Additionally, DIPS improves data efficiency and reduces the performance distinctions between different pseudo-labelers. Overall, we highlight the significant benefits of a data-centric rethinking of pseudo-labeling in real-world settings."
                    },
                    {
                        "title": "Retrieval Augmented Thought Process for Private Data Handling in Healthcare",
                        "abstract": "Large Language Models (LLMs) have demonstrated the strong potential to assist both clinicians and the general public with their extensive medical knowledge. However, their application in healthcare is constrained due to concerns about the privacy of data used in training, which prevents the integration of private and personal information because of security and ethical issues. Moreover, if their capabilities can be enhanced with information retrieval to access up-to-date knowledge, the current integration of LLMs with Information retrieval lacks robustness to imperfect retrieval, which can hinder their effectiveness and even reduce overall performance. In this work, we address this challenge by introducing the Retrieval-Augmented Thought Process (RATP). Given access to external knowledge, RATP formulates the thought generation of LLMs as a multiple-step decision process. To optimise such a thought process, RATP leverages Monte-Carlo Tree Search and learns a proxy reward function that permits cost-efficient inference. On a private dataset of electronic medical records, deliberately excluded from any LLM training set, RATP achieves 35% additional accuracy compared to in-context retrieval-augmented generation for the question-answering task."
                    },
                    {
                        "title": "ODE Discovery for Longitudinal Heterogeneous Treatment Effects Inference",
                        "abstract": "Inferring unbiased treatment effects has received widespread attention in the machine learning community. In recent years, our community has proposed numerous solutions in standard settings, high-dimensional treatment settings, and even longitudinal settings. While very diverse, the solution has mostly relied on neural networks for inference and simultaneous correction of assignment bias. New approaches typically build on top of previous approaches by proposing new (or refined) architectures and learning algorithms. However, the end result -- a neural-network-based inference machine -- remains unchallenged. In this paper, we introduce a different type of solution in the longitudinal setting: a closed-form ordinary differential equation (ODE). While we still rely on continuous optimization to learn an ODE, the resulting inference machine is no longer a neural network. Doing so yields several advantages such as interpretability, irregular sampling, and a different set of identification assumptions. Above all, we consider the introduction of a completely new type of solution to be our most important contribution as it may spark entirely new innovations in treatment effects in general. We facilitate this by formulating our contribution as a framework that can transform any ODE discovery method into a treatment effects method."
                    },
                    {
                        "title": "Meta-Learners for Partially-Identified Treatment Effects Across Multiple Environments",
                        "abstract": "Estimating the conditional average treatment effect (CATE) from observational data is relevant for many applications such as personalized medicine. Here, we focus on the widespread setting where the observational data come from multiple environments, such as different hospitals, physicians, or countries. Furthermore, we allow for violations of standard causal assumptions, namely, overlap within the environments and unconfoundedness. To this end, we move away from point identification and focus on partial identification. Specifically, we show that current assumptions from the literature on multiple environments allow us to interpret the environment as an instrumental variable (IV). This allows us to adapt bounds from the IV literature for partial identification of CATE by leveraging treatment assignment mechanisms across environments. Then, we propose different model-agnostic learners (so-called meta-learners) to estimate the bounds that can be used in combination with arbitrary machine learning models. We further demonstrate the effectiveness of our meta-learners across various experiments using both simulated and real-world data. Finally, we discuss the applicability of our meta-learners to partial identification in instrumental variable settings, such as randomized controlled trials with non-compliance."
                    },
                    {
                        "title": "Shape Arithmetic Expressions: Advancing Scientific Discovery Beyond Closed-Form Equations",
                        "abstract": "Symbolic regression has excelled in uncovering equations from physics, chemistry, biology, and related disciplines. However, its effectiveness becomes less certain when applied to experimental data lacking inherent closed-form expressions. Empirically derived relationships, such as entire stress-strain curves, may defy concise closed-form representation, compelling us to explore more adaptive modeling approaches that balance flexibility with interpretability. In our pursuit, we turn to Generalized Additive Models (GAMs), a widely used class of models known for their versatility across various domains. Although GAMs can capture non-linear relationships between variables and targets, they cannot capture intricate feature interactions. In this work, we investigate both of these challenges and propose a novel class of models, Shape Arithmetic Expressions (SHAREs), that fuses GAM's flexible shape functions with the complex feature interactions found in mathematical expressions. SHAREs also provide a unifying framework for both of these approaches. We also design a set of rules for constructing SHAREs that guarantee transparency of the found expressions beyond the standard constraints based on the model's size."
                    },
                    {
                        "title": "Large Language Models to Enhance Bayesian Optimization",
                        "abstract": "Bayesian optimization (BO) is a powerful approach for optimizing complex and expensive-to-evaluate black-box functions. Its importance is underscored in many applications, notably including hyperparameter tuning, but its efficacy depends on efficiently balancing exploration and exploitation. While there has been substantial progress in BO methods, striking this balance remains a delicate process. In this light, we present LLAMBO, a novel approach that integrates the capabilities of Large Language Models (LLM) within BO. At a high level, we frame the BO problem in natural language, enabling LLMs to iteratively propose and evaluate promising solutions conditioned on historical evaluations. More specifically, we explore how combining contextual understanding, few-shot learning proficiency, and domain knowledge of LLMs can improve model-based BO. Our findings illustrate that LLAMBO is effective at zero-shot warmstarting, and enhances surrogate modeling and candidate sampling, especially in the early stages of search when observations are sparse. Our approach is performed in context and does not require LLM finetuning. Additionally, it is modular by design, allowing individual components to be integrated into existing BO frameworks, or function cohesively as an end-to-end method. We empirically validate LLAMBO's efficacy on the problem of hyperparameter tuning, highlighting strong empirical performance across a range of diverse benchmarks, proprietary, and synthetic tasks."
                    },
                    {
                        "title": "Automatically Learning Hybrid Digital Twins of Dynamical Systems",
                        "abstract": "Digital Twins (DTs) are computational models that simulate the states and temporal dynamics of real-world systems, playing a crucial role in prediction, understanding, and decision-making across diverse domains. However, existing approaches to DTs often struggle to generalize to unseen conditions in data-scarce settings, a crucial requirement for such models. To address these limitations, our work begins by establishing the essential desiderata for effective DTs. Hybrid Digital Twins ($\\textbf{HDTwins}$) represent a promising approach to address these requirements, modeling systems using a composition of both mechanistic and neural components. This hybrid architecture simultaneously leverages (partial) domain knowledge and neural network expressiveness to enhance generalization, with its modular design facilitating improved evolvability. While existing hybrid models rely on expert-specified architectures with only parameters optimized on data, $\\textit{automatically}$ specifying and optimizing HDTwins remains intractable due to the complex search space and the need for flexible integration of domain priors. To overcome this complexity, we propose an evolutionary algorithm ($\\textbf{HDTwinGen}$) that employs Large Language Models (LLMs) to autonomously propose, evaluate, and optimize HDTwins. Specifically, LLMs iteratively generate novel model specifications, while offline tools are employed to optimize emitted parameters. Correspondingly, proposed models are evaluated and evolved based on targeted feedback, enabling the discovery of increasingly effective hybrid models. Our empirical results reveal that HDTwinGen produces generalizable, sample-efficient, and evolvable models, significantly advancing DTs' efficacy in real-world applications."
                    },
                    {
                        "title": "Causal Deep Learning: Encouraging Impact on Real-world Problems Through Causality",
                        "abstract": "Causality has the potential to truly transform the way we solve a large number of real-world problems. Yet, so far, its potential largely remains to be unlocked as causality often requires crucial assumptions which cannot be tested in practice. To address this challenge, we propose a new way of thinking about causality-\u2013 we call this causal deep learning . Our causal deep learning framework spans three dimensions: (1) a structural dimension, which incorporates partial yet testable causal knowledge rather than assuming either complete or no causal knowledge among the variables of interest; (2) a parametric dimension, which encompasses parametric forms that capture the type of relationships among the variables of interest; and (3) a temporal dimen-sion, which captures exposure times or how the variables of interest interact (possibly causally) over time. Our CDL framework enables us to precisely categorise and compare"
                    },
                    {
                        "title": "Context-Aware Testing: A New Paradigm for Model Testing with Large Language Models",
                        "abstract": "The predominant de facto paradigm of testing ML models relies on either using only held-out data to compute aggregate evaluation metrics or by assessing the performance on different subgroups. However, such data-only testing methods operate under the restrictive assumption that the available empirical data is the sole input for testing ML models, disregarding valuable contextual information that could guide model testing. In this paper, we challenge the go-to approach of data-only testing and introduce context-aware testing (CAT) which uses context as an inductive bias to guide the search for meaningful model failures. We instantiate the first CAT system, SMART Testing, which employs large language models to hypothesize relevant and likely failures, which are evaluated on data using a self-falsification mechanism. Through empirical evaluations in diverse settings, we show that SMART automatically identifies more relevant and impactful failures than alternatives, demonstrating the potential of CAT as a testing paradigm."
                    },
                    {
                        "title": "Risk-Sensitive Diffusion: Robustly Optimizing Diffusion Models with Noisy Samples",
                        "abstract": "Diffusion models are mainly studied on image data. However, non-image data (e.g., tabular data) are also prevalent in real applications and tend to be noisy due to some inevitable factors in the stage of data collection, degrading the generation quality of diffusion models. In this paper, we consider a novel problem setting where every collected sample is paired with a vector indicating the data quality: risk vector. This setting applies to many scenarios involving noisy data and we propose risk-sensitive SDE, a type of stochastic differential equation (SDE) parameterized by the risk vector, to address it. With some proper coefficients, risk-sensitive SDE can minimize the negative effect of noisy samples on the optimization of diffusion models. We conduct systematic studies for both Gaussian and non-Gaussian noise distributions, providing analytical forms of risk-sensitive SDE. To verify the effectiveness of our method, we have conducted extensive experiments on multiple tabular and time-series datasets, showing that risk-sensitive SDE permits a robust optimization of diffusion models with noisy samples and significantly outperforms previous baselines."
                    },
                    {
                        "title": "CliMB: An AI-enabled Partner for Clinical Predictive Modeling",
                        "abstract": "Despite its significant promise and continuous technical advances, real-world applications of artificial intelligence (AI) remain limited. We attribute this to the\"domain expert-AI-conundrum\": while domain experts, such as clinician scientists, should be able to build predictive models such as risk scores, they face substantial barriers in accessing state-of-the-art (SOTA) tools. While automated machine learning (AutoML) has been proposed as a partner in clinical predictive modeling, many additional requirements need to be fulfilled to make machine learning accessible for clinician scientists. To address this gap, we introduce CliMB, a no-code AI-enabled partner designed to empower clinician scientists to create predictive models using natural language. CliMB guides clinician scientists through the entire medical data science pipeline, thus empowering them to create predictive models from real-world data in just one conversation. CliMB also creates structured reports and interpretable visuals. In evaluations involving clinician scientists and systematic comparisons against a baseline GPT-4, CliMB consistently demonstrated superior performance in key areas such as planning, error prevention, code execution, and model performance. Moreover, in blinded assessments involving 45 clinicians from diverse specialties and career stages, more than 80% preferred CliMB over GPT-4. Overall, by providing a no-code interface with clear guidance and access to SOTA methods in the fields of data-centric AI, AutoML, and interpretable ML, CliMB empowers clinician scientists to build robust predictive models."
                    },
                    {
                        "title": "Time Series Diffusion in the Frequency Domain",
                        "abstract": "Fourier analysis has been an instrumental tool in the development of signal processing. This leads us to wonder whether this framework could similarly benefit generative modelling. In this paper, we explore this question through the scope of time series diffusion models. More specifically, we analyze whether representing time series in the frequency domain is a useful inductive bias for score-based diffusion models. By starting from the canonical SDE formulation of diffusion in the time domain, we show that a dual diffusion process occurs in the frequency domain with an important nuance: Brownian motions are replaced by what we call mirrored Brownian motions, characterized by mirror symmetries among their components. Building on this insight, we show how to adapt the denoising score matching approach to implement diffusion models in the frequency domain. This results in frequency diffusion models, which we compare to canonical time diffusion models. Our empirical evaluation on real-world datasets, covering various domains like healthcare and finance, shows that frequency diffusion models better capture the training distribution than time diffusion models. We explain this observation by showing that time series from these datasets tend to be more localized in the frequency domain than in the time domain, which makes them easier to model in the former case. All our observations point towards impactful synergies between Fourier analysis and diffusion models."
                    },
                    {
                        "title": "Informed Meta-Learning",
                        "abstract": "In noisy and low-data regimes prevalent in real-world applications, a key challenge of machine learning lies in effectively incorporating inductive biases that promote data efficiency and robustness. Meta-learning and informed ML stand out as two approaches for incorporating prior knowledge into ML pipelines. While the former relies on a purely data-driven source of priors, the latter is guided by prior domain knowledge. In this paper, we formalise a hybrid paradigm, informed meta-learning, facilitating the incorporation of priors from unstructured knowledge representations, such as natural language; thus, unlocking complementarity in cross-task knowledge sharing of humans and machines. We establish the foundational components of informed meta-learning and present a concrete instantiation of this framework--the Informed Neural Process. Through a series of experiments, we demonstrate the potential benefits of informed meta-learning in improving data efficiency, robustness to observational noise and task distribution shifts."
                    },
                    {
                        "title": "Risk-Sensitive Diffusion: Learning the Underlying Distribution from Noisy Samples",
                        "abstract": "While achieving remarkable performances, we show that diffusion models are fragile to the presence of noisy samples, limiting their potential in the vast amount of settings where, unlike image synthesis, we are not blessed with clean data. Motivated by our finding that such fragility originates from the distribution gaps between noisy and clean samples along the diffusion process, we introduce risk-sensitive SDE , a stochastic differential equation that is parameterized by the risk (i.e., data \u201cdirtiness\u201d) to adjust the distributions of noisy samples, reducing misguidance while bene-fiting from their contained information. The optimal expression for risk-sensitive SDE depends on the specific noise distribution, and we derive its parameterizations that minimize the misguidance of noisy samples for both Gaussian and general non-Gaussian perturbations. We conduct extensive experiments on both synthetic and real-world datasets (e.g., medical time series), showing that our model effectively recovers the clean data distribution from noisy samples, significantly outper-forming conditional generation baselines."
                    },
                    {
                        "title": "Unveiling the Power of Sparse Neural Networks for Feature Selection",
                        "abstract": "Sparse Neural Networks (SNNs) have emerged as powerful tools for efficient feature selection. Leveraging the dynamic sparse training (DST) algorithms within SNNs has demonstrated promising feature selection capabilities while drastically reducing computational overheads. Despite these advancements, several critical aspects remain insufficiently explored for feature selection. Questions persist regarding the choice of the DST algorithm for network training, the choice of metric for ranking features/neurons, and the comparative performance of these methods across diverse datasets when compared to dense networks. This paper addresses these gaps by presenting a comprehensive systematic analysis of feature selection with sparse neural networks. Moreover, we introduce a novel metric considering sparse neural network characteristics, which is designed to quantify feature importance within the context of SNNs. Our findings show that feature selection with SNNs trained with DST algorithms can achieve, on average, more than $50\\%$ memory and $55\\%$ FLOPs reduction compared to the dense networks, while outperforming them in terms of the quality of the selected features. Our code and the supplementary material are available on GitHub (\\url{https://github.com/zahraatashgahi/Neuron-Attribution})."
                    },
                    {
                        "title": "Discovering Preference Optimization Algorithms with and for Large Language Models",
                        "abstract": "Offline preference optimization is a key method for enhancing and controlling the quality of Large Language Model (LLM) outputs. Typically, preference optimization is approached as an offline supervised learning task using manually-crafted convex loss functions. While these methods are based on theoretical insights, they are inherently constrained by human creativity, so the large search space of possible loss functions remains under explored. We address this by performing LLM-driven objective discovery to automatically discover new state-of-the-art preference optimization algorithms without (expert) human intervention. Specifically, we iteratively prompt an LLM to propose and implement new preference optimization loss functions based on previously-evaluated performance metrics. This process leads to the discovery of previously-unknown and performant preference optimization algorithms. The best performing of these we call Discovered Preference Optimization (DiscoPOP), a novel algorithm that adaptively blends logistic and exponential losses. Experiments demonstrate the state-of-the-art performance of DiscoPOP and its successful transfer to held-out tasks."
                    }
                ]
            }
        }
    },
    "2406.02507": {
        "paper_data": {
            "title": "Guiding a Diffusion Model with a Bad Version of Itself",
            "url": "http://arxiv.org/abs/2406.02507v1",
            "arxiv_id": "2406.02507",
            "authors": [
                "Tero Karras",
                "Miika Aittala",
                "Tuomas Kynk\u00e4\u00e4nniemi",
                "Jaakko Lehtinen",
                "Timo Aila",
                "Samuli Laine"
            ],
            "abstract": "The primary axes of interest in image-generating diffusion models are image quality, the amount of variation in the results, and how well the results align with a given condition, e.g., a class label or a text prompt. The popular classifier-free guidance approach uses an unconditional model to guide a conditional model, leading to simultaneously better prompt alignment and higher-quality images at the cost of reduced variation. These effects seem inherently entangled, and thus hard to control. We make the surprising observation that it is possible to obtain disentangled control over image quality without compromising the amount of variation by guiding generation using a smaller, less-trained version of the model itself rather than an unconditional model. This leads to significant improvements in ImageNet generation, setting record FIDs of 1.01 for 64x64 and 1.25 for 512x512, using publicly available networks. Furthermore, the method is also applicable to unconditional diffusion models, drastically improving their quality.",
            "introduction": "   1 Introduction  Denoising diffusion models\u00a0[12, 37, 38, 39, 40] generate synthetic images by reversing a stochastic corruption process. Essentially, an image is revealed from pure noise by denoising it little by little in successive steps. A neural network that implements the denoiser (equivalently\u00a0[44], the score function\u00a0[15]) is a central design element, and various architectures have been proposed (e.g.,\u00a0[1, 5, 7, 14, 16, 19, 30]). Equally important are the details of the multi-step denoising process that corresponds mathematically to solving an ordinary\u00a0[27, 38] or a stochastic\u00a0[40] differential equation, for which many different parameterizations, solvers, and step schedules have been evaluated\u00a0[17, 18, 22, 35, 46]. To control the output image, the denoiser is typically conditioned on a class label, an embedding of a text prompt, or some other form of conditioning input\u00a0[28, 33, 36, 45].   The training objective of a diffusion model aims to cover the entire (conditional) data distribution. This causes problems in low-probability regions: The model gets heavily penalized for not representing them, but it does not have enough data to learn to generate good images corresponding to them. Classifier-free guidance (CFG)\u00a0[13] has become the standard method for \u201clowering the sampling temperature\u201d, i.e., focusing the generation on well-learned high-probability regions. By training a denoiser network to operate in both conditional and unconditional setting, the sampling process can be steered away from the unconditional result\u2009\u2014\u2009in effect, the unconditional generation task specifies a result to avoid. This results in better prompt alignment and improved image quality, where the former effect is due to CFG implicitly raising the conditional part of the probability density to a power greater than one\u00a0[8].   However, CFG has drawbacks that limit its usage as a general low-temperature sampling method. First, it is applicable only for conditional generation, as the guidance signal is based on the difference between conditional and unconditional denoising results. Second, because the unconditional and conditional denoisers are trained to solve a different task, the sampling trajectory can overshoot the desired conditional distribution, which leads to skewed and often overly simplified image compositions\u00a0[21]. Finally, the prompt alignment and quality improvement effects cannot be controlled separately, and it remains unclear how exactly they relate to each other.   In this paper, we provide new insights into why CFG improves image quality and show how this effect can be separated out into a novel method that we call autoguidance. Our method does not suffer from the task discrepancy problem because we use an inferior version of the main model itself as the guiding model, with unchanged conditioning. This guiding model can be obtained by simply limiting, e.g., model capacity and/or training time. We validate the effectiveness of autoguidance in various synthetic test cases as well as in practical image synthesis in class-conditional and text-conditional settings. In addition, our method enables guidance for unconditional synthesis. In quantitative tests, the generated image distributions are improved considerably when measured using FID\u00a0[10] and FDDINOv2\u00a0[42] metrics, setting new records in ImageNet-512 and ImageNet-64 generation.   We will make our implementation and pre-trained models publicly available.     2 Background  Denoising diffusion.  Denoising diffusion generates samples from a distribution pdata\u2062(\ud835\udc31)subscript\ud835\udc5ddata\ud835\udc31p_{\\text{data}}(\\mathbf{x})italic_p start_POSTSUBSCRIPT data end_POSTSUBSCRIPT ( bold_x ) by iteratively denoising a sample of pure white noise, such that a noise-free random data sample is gradually revealed\u00a0[12]. The idea is to consider heat diffusion of pdata\u2062(\ud835\udc31)subscript\ud835\udc5ddata\ud835\udc31p_{\\text{data}}(\\mathbf{x})italic_p start_POSTSUBSCRIPT data end_POSTSUBSCRIPT ( bold_x ) into a sequence of increasingly smoothed densities",
            "references": [
                {
                    "title": "Align Your Steps: Optimizing Sampling Schedules in Diffusion Models",
                    "abstract": "Diffusion models (DMs) have established themselves as the state-of-the-art generative modeling approach in the visual domain and beyond. A crucial drawback of DMs is their slow sampling speed, relying on many sequential function evaluations through large neural networks. Sampling from DMs can be seen as solving a differential equation through a discretized set of noise levels known as the sampling schedule. While past works primarily focused on deriving efficient solvers, little attention has been given to finding optimal sampling schedules, and the entire literature relies on hand-crafted heuristics. In this work, for the first time, we propose a general and principled approach to optimizing the sampling schedules of DMs for high-quality outputs, called $\\textit{Align Your Steps}$. We leverage methods from stochastic calculus and find optimal schedules specific to different solvers, trained DMs and datasets. We evaluate our novel approach on several image, video as well as 2D toy data synthesis benchmarks, using a variety of different samplers, and observe that our optimized schedules outperform previous hand-crafted schedules in almost all experiments. Our method demonstrates the untapped potential of sampling schedule optimization, especially in the few-step synthesis regime."
                },
                {
                    "title": "Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models",
                    "abstract": "Guidance is a crucial technique for extracting the best performance out of image-generating diffusion models. Traditionally, a constant guidance weight has been applied throughout the sampling chain of an image. We show that guidance is clearly harmful toward the beginning of the chain (high noise levels), largely unnecessary toward the end (low noise levels), and only beneficial in the middle. We thus restrict it to a specific range of noise levels, improving both the inference speed and result quality. This limited guidance interval improves the record FID in ImageNet-512 significantly, from 1.81 to 1.40. We show that it is quantitatively and qualitatively beneficial across different sampler parameters, network architectures, and datasets, including the large-scale setting of Stable Diffusion XL. We thus suggest exposing the guidance interval as a hyperparameter in all diffusion models that use guidance."
                },
                {
                    "title": "Scalable High-Resolution Pixel-Space Image Synthesis with Hourglass Diffusion Transformers",
                    "abstract": "We present the Hourglass Diffusion Transformer (HDiT), an image generative model that exhibits linear scaling with pixel count, supporting training at high-resolution (e.g. $1024 \\times 1024$) directly in pixel-space. Building on the Transformer architecture, which is known to scale to billions of parameters, it bridges the gap between the efficiency of convolutional U-Nets and the scalability of Transformers. HDiT trains successfully without typical high-resolution training techniques such as multiscale architectures, latent autoencoders or self-conditioning. We demonstrate that HDiT performs competitively with existing models on ImageNet $256^2$, and sets a new state-of-the-art for diffusion models on FFHQ-$1024^2$."
                },
                {
                    "title": "Analyzing and Improving the Training Dynamics of Diffusion Models",
                    "abstract": "Diffusion models currently dominate the field of data- driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high- level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on ex- pectation. We find that systematic application of this philoso- phy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational com- plexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling. As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance."
                },
                {
                    "title": "Exposing flaws of generative model evaluation metrics and their unfair treatment of diffusion models",
                    "abstract": "We systematically study a wide variety of generative models spanning semantically-diverse image datasets to understand and improve the feature extractors and metrics used to evaluate them. Using best practices in psychophysics, we measure human perception of image realism for generated samples by conducting the largest experiment evaluating generative models to date, and find that no existing metric strongly correlates with human evaluations. Comparing to 17 modern metrics for evaluating the overall performance, fidelity, diversity, rarity, and memorization of generative models, we find that the state-of-the-art perceptual realism of diffusion models as judged by humans is not reflected in commonly reported metrics such as FID. This discrepancy is not explained by diversity in generated samples, though one cause is over-reliance on Inception-V3. We address these flaws through a study of alternative self-supervised feature extractors, find that the semantic information encoded by individual networks strongly depends on their training procedure, and show that DINOv2-ViT-L/14 allows for much richer evaluation of generative models. Next, we investigate data memorization, and find that generative models do memorize training examples on simple, smaller datasets like CIFAR10, but not necessarily on more complex datasets like ImageNet. However, our experiments show that current metrics do not properly detect memorization: none in the literature is able to separate memorization from other phenomena such as underfitting or mode shrinkage. To facilitate further development of generative models and their evaluation we release all generated image datasets, human evaluation data, and a modular library to compute 17 common metrics for 9 different encoders at https://github.com/layer6ai-labs/dgm-eval."
                },
                {
                    "title": "DINOv2: Learning Robust Visual Features without Supervision",
                    "abstract": "The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels."
                },
                {
                    "title": "Adding Conditional Control to Text-to-Image Diffusion Models",
                    "abstract": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models."
                },
                {
                    "title": "simple diffusion: End-to-end diffusion for high resolution images",
                    "abstract": "Currently, applying diffusion models in pixel space of high resolution images is difficult. Instead, existing approaches focus on diffusion in lower dimensional spaces (latent diffusion), or have multiple super-resolution levels of generation referred to as cascades. The downside is that these approaches add additional complexity to the diffusion framework. This paper aims to improve denoising diffusion for high resolution images while keeping the model as simple as possible. The paper is centered around the research question: How can one train a standard denoising diffusion models on high resolution images, and still obtain performance comparable to these alternate approaches? The four main findings are: 1) the noise schedule should be adjusted for high resolution images, 2) It is sufficient to scale only a particular part of the architecture, 3) dropout should be added at specific locations in the architecture, and 4) downsampling is an effective strategy to avoid high resolution feature maps. Combining these simple yet effective techniques, we achieve state-of-the-art on image generation among diffusion models without sampling modifiers on ImageNet."
                },
                {
                    "title": "Scalable Adaptive Computation for Iterative Generation",
                    "abstract": "Natural data is redundant yet predominant architectures tile computation uniformly across their input and output space. We propose the Recurrent Interface Networks (RINs), an attention-based architecture that decouples its core computation from the dimensionality of the data, enabling adaptive computation for more scalable generation of high-dimensional data. RINs focus the bulk of computation (i.e. global self-attention) on a set of latent tokens, using cross-attention to read and write (i.e. route) information between latent and data tokens. Stacking RIN blocks allows bottom-up (data to latent) and top-down (latent to data) feedback, leading to deeper and more expressive routing. While this routing introduces challenges, this is less problematic in recurrent computation settings where the task (and routing problem) changes gradually, such as iterative generation with diffusion models. We show how to leverage recurrence by conditioning the latent tokens at each forward pass of the reverse diffusion process with those from prior computation, i.e. latent self-conditioning. RINs yield state-of-the-art pixel diffusion models for image and video generation, scaling to 1024X1024 images without cascades or guidance, while being domain-agnostic and up to 10X more efficient than 2D and 3D U-Nets."
                },
                {
                    "title": "Scalable Diffusion Models with Transformers",
                    "abstract": "We explore a new class of diffusion models based on the transformer architecture. We train latent diffusion models of images, replacing the commonly-used U-Net backbone with a transformer that operates on latent patches. We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops. We find that DiTs with higher Gflops\u2014through increased transformer depth/width or increased number of input tokens\u2014consistently have lower FID. In addition to possessing good scalability properties, our largest DiT-XL/2 models outperform all prior diffusion models on the class-conditional ImageNet 512\u00d7512 and 256\u00d7256 benchmarks, achieving a state-of-the-art FID of 2.27 on the latter."
                },
                {
                    "title": "Refining Generative Process with Discriminator Guidance in Score-based Diffusion Models",
                    "abstract": "The proposed method, Discriminator Guidance, aims to improve sample generation of pre-trained diffusion models. The approach introduces a discriminator that gives explicit supervision to a denoising sample path whether it is realistic or not. Unlike GANs, our approach does not require joint training of score and discriminator networks. Instead, we train the discriminator after score training, making discriminator training stable and fast to converge. In sample generation, we add an auxiliary term to the pre-trained score to deceive the discriminator. This term corrects the model score to the data score at the optimal discriminator, which implies that the discriminator helps better score estimation in a complementary way. Using our algorithm, we achive state-of-the-art results on ImageNet 256x256 with FID 1.83 and recall 0.64, similar to the validation data's FID (1.68) and recall (0.66). We release the code at https://github.com/alsdudrla10/DG."
                },
                {
                    "title": "Imagen Video: High Definition Video Generation with Diffusion Models",
                    "abstract": "We present Imagen Video, a text-conditional video generation system based on a cascade of video diffusion models. Given a text prompt, Imagen Video generates high definition videos using a base video generation model and a sequence of interleaved spatial and temporal video super-resolution models. We describe how we scale up the system as a high definition text-to-video model including design decisions such as the choice of fully-convolutional temporal and spatial super-resolution models at certain resolutions, and the choice of the v-parameterization of diffusion models. In addition, we confirm and transfer findings from previous work on diffusion-based image generation to the video generation setting. Finally, we apply progressive distillation to our video models with classifier-free guidance for fast, high quality sampling. We find Imagen Video not only capable of generating videos of high fidelity, but also having a high degree of controllability and world knowledge, including the ability to generate diverse videos and text animations in various artistic styles and with 3D object understanding. See https://imagen.research.google/video/ for samples."
                },
                {
                    "title": "DreamFusion: Text-to-3D using 2D Diffusion",
                    "abstract": "Recent breakthroughs in text-to-image synthesis have been driven by diffusion models trained on billions of image-text pairs. Adapting this approach to 3D synthesis would require large-scale datasets of labeled 3D data and efficient architectures for denoising 3D data, neither of which currently exist. In this work, we circumvent these limitations by using a pretrained 2D text-to-image diffusion model to perform text-to-3D synthesis. We introduce a loss based on probability density distillation that enables the use of a 2D diffusion model as a prior for optimization of a parametric image generator. Using this loss in a DeepDream-like procedure, we optimize a randomly-initialized 3D model (a Neural Radiance Field, or NeRF) via gradient descent such that its 2D renderings from random angles achieve a low loss. The resulting 3D model of the given text can be viewed from any angle, relit by arbitrary illumination, or composited into any 3D environment. Our approach requires no 3D training data and no modifications to the image diffusion model, demonstrating the effectiveness of pretrained image diffusion models as priors."
                },
                {
                    "title": "All are Worth Words: A ViT Backbone for Diffusion Models",
                    "abstract": "Vision transformers (ViT) have shown promise in various vision tasks while the U-Net based on a convolutional neural network (CNN) remains dominant in diffusion models. We design a simple and general ViT-based architecture (named U-ViT) for image generation with diffusion models. U-ViT is characterized by treating all inputs including the time, condition and noisy image patches as tokens and employing long skip connections between shallow and deep layers. We evaluate U-ViT in unconditional and classconditional image generation, as well as text-to-image generation tasks, where U-ViT is comparable if not superior to a CNN-based U-Net of a similar size. In particular, latent diffusion models with U-ViT achieve record-breaking FID scores of 2.29 in class-conditional image generation on ImageNet 256\u00d7256, and 5.48 in text-to-image generation on MS-COCO, among methods without accessing large external datasets during the training of generative models. Our results suggest that, for diffusion-based image modeling, the long skip connection is crucial while the down-sampling and upsampling operators in CNN-based U-Net are not always necessary. We believe that U-ViT can provide insights for future research on backbones in diffusion models and benefit generative modeling on large scale cross-modality datasets."
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "Verifying the Union of Manifolds Hypothesis for Image Data",
                    "abstract": "Deep learning has had tremendous success at learning low-dimensional representations of high-dimensional data. This success would be impossible if there was no hidden low-dimensional structure in data of interest; this existence is posited by the manifold hypothesis, which states that the data lies on an unknown manifold of low intrinsic dimension. In this paper, we argue that this hypothesis does not properly capture the low-dimensional structure typically present in image data. Assuming that data lies on a single manifold implies intrinsic dimension is identical across the entire data space, and does not allow for subregions of this space to have a different number of factors of variation. To address this deficiency, we consider the union of manifolds hypothesis, which states that data lies on a disjoint union of manifolds of varying intrinsic dimensions. We empirically verify this hypothesis on commonly-used image datasets, finding that indeed, observed data lies on a disconnected set and that intrinsic dimension is not constant. We also provide insights into the implications of the union of manifolds hypothesis in deep learning, both supervised and unsupervised, showing that designing models with an inductive bias for this structure improves performance across classification and generative modelling tasks. Our code is available at https://github.com/layer6ai-labs/UoMH."
                },
                {
                    "title": "Compositional Visual Generation with Composable Diffusion Models",
                    "abstract": "Large text-guided diffusion models, such as DALLE-2, are able to generate stunning photorealistic images given natural language descriptions. While such models are highly flexible, they struggle to understand the composition of certain concepts, such as confusing the attributes of different objects or relations between objects. In this paper, we propose an alternative structured approach for compositional generation using diffusion models. An image is generated by composing a set of diffusion models, with each of them modeling a certain component of the image. To do this, we interpret diffusion models as energy-based models in which the data distributions defined by the energy functions may be explicitly combined. The proposed method can generate scenes at test time that are substantially more complex than those seen in training, composing sentence descriptions, object relations, human facial attributes, and even generalizing to new combinations that are rarely seen in the real world. We further illustrate how our approach may be used to compose pre-trained text-guided diffusion models and generate photorealistic images containing all the details described in the input descriptions, including the binding of certain object attributes that have been shown difficult for DALLE-2. These results point to the effectiveness of the proposed method in promoting structured generalization for visual generation. Project page: https://energy-based-model.github.io/Compositional-Visual-Generation-with-Composable-Diffusion-Models/"
                },
                {
                    "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                    "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                },
                {
                    "title": "Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding",
                    "abstract": "We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding. Imagen builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation. Our key discovery is that generic large language models (e.g. T5), pretrained on text-only corpora, are surprisingly effective at encoding text for image synthesis: increasing the size of the language model in Imagen boosts both sample fidelity and image-text alignment much more than increasing the size of the image diffusion model. Imagen achieves a new state-of-the-art FID score of 7.27 on the COCO dataset, without ever training on COCO, and human raters find Imagen samples to be on par with the COCO data itself in image-text alignment. To assess text-to-image models in greater depth, we introduce DrawBench, a comprehensive and challenging benchmark for text-to-image models. With DrawBench, we compare Imagen with recent methods including VQ-GAN+CLIP, Latent Diffusion Models, and DALL-E 2, and find that human raters prefer Imagen over other models in side-by-side comparisons, both in terms of sample quality and image-text alignment. See https://imagen.research.google/ for an overview of the results."
                },
                {
                    "title": "Fast Sampling of Diffusion Models with Exponential Integrator",
                    "abstract": "The past few years have witnessed the great success of Diffusion models~(DMs) in generating high-fidelity samples in generative modeling tasks. A major limitation of the DM is its notoriously slow sampling procedure which normally requires hundreds to thousands of time discretization steps of the learned diffusion process to reach the desired accuracy. Our goal is to develop a fast sampling method for DMs with a much less number of steps while retaining high sample quality. To this end, we systematically analyze the sampling procedure in DMs and identify key factors that affect the sample quality, among which the method of discretization is most crucial. By carefully examining the learned diffusion process, we propose Diffusion Exponential Integrator Sampler~(DEIS). It is based on the Exponential Integrator designed for discretizing ordinary differential equations (ODEs) and leverages a semilinear structure of the learned diffusion process to reduce the discretization error. The proposed method can be applied to any DMs and can generate high-fidelity samples in as few as 10 steps. In our experiments, it takes about 3 minutes on one A6000 GPU to generate $50k$ images from CIFAR10. Moreover, by directly using pre-trained DMs, we achieve the state-of-art sampling performance when the number of score function evaluation~(NFE) is limited, e.g., 4.17 FID with 10 NFEs, 3.37 FID, and 9.74 IS with only 15 NFEs on CIFAR10. Code is available at https://github.com/qsh-zh/deis"
                },
                {
                    "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
                    "abstract": "Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples."
                },
                {
                    "title": "Pseudo Numerical Methods for Diffusion Models on Manifolds",
                    "abstract": "Denoising Diffusion Probabilistic Models (DDPMs) can generate high-quality samples such as image and audio samples. However, DDPMs require hundreds to thousands of iterations to produce final samples. Several prior works have successfully accelerated DDPMs through adjusting the variance schedule (e.g., Improved Denoising Diffusion Probabilistic Models) or the denoising equation (e.g., Denoising Diffusion Implicit Models (DDIMs)). However, these acceleration methods cannot maintain the quality of samples and even introduce new noise at a high speedup rate, which limit their practicability. To accelerate the inference process while keeping the sample quality, we provide a fresh perspective that DDPMs should be treated as solving differential equations on manifolds. Under such a perspective, we propose pseudo numerical methods for diffusion models (PNDMs). Specifically, we figure out how to solve differential equations on manifolds and show that DDIMs are simple cases of pseudo numerical methods. We change several classical numerical methods to corresponding pseudo numerical methods and find that the pseudo linear multi-step method is the best in most situations. According to our experiments, by directly using pre-trained models on Cifar10, CelebA and LSUN, PNDMs can generate higher quality synthetic images with only 50 steps compared with 1000-step DDIMs (20x speedup), significantly outperform DDIMs with 250 steps (by around 0.4 in FID) and have good generalization on different variance schedules. Our implementation is available at https://github.com/luping-liu/PNDM."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models",
                    "abstract": "Diffusion models have recently been shown to generate high-quality synthetic images, especially when paired with a guidance technique to trade off diversity for fidelity. We explore diffusion models for the problem of text-conditional image synthesis and compare two different guidance strategies: CLIP guidance and classifier-free guidance. We find that the latter is preferred by human evaluators for both photorealism and caption similarity, and often produces photorealistic samples. Samples from a 3.5 billion parameter text-conditional diffusion model using classifier-free guidance are favored by human evaluators to those from DALL-E, even when the latter uses expensive CLIP reranking. Additionally, we find that our models can be fine-tuned to perform image inpainting, enabling powerful text-driven image editing. We train a smaller model on a filtered dataset and release the code and weights at https://github.com/openai/glide-text2im."
                },
                {
                    "title": "Gotta Go Fast When Generating Data with Score-Based Models",
                    "abstract": "Score-based (denoising diffusion) generative models have recently gained a lot of success in generating realistic and diverse data. These approaches define a forward diffusion process for transforming data to noise and generate data by reversing it (thereby going from noise to data). Unfortunately, current score-based models generate data very slowly due to the sheer number of score network evaluations required by numerical SDE solvers. In this work, we aim to accelerate this process by devising a more efficient SDE solver. Existing approaches rely on the Euler-Maruyama (EM) solver, which uses a fixed step size. We found that naively replacing it with other SDE solvers fares poorly - they either result in low-quality samples or become slower than EM. To get around this issue, we carefully devise an SDE solver with adaptive step sizes tailored to score-based generative models piece by piece. Our solver requires only two score function evaluations, rarely rejects samples, and leads to high-quality samples. Our approach generates data 2 to 10 times faster than EM while achieving better or equal sample quality. For high-resolution images, our method leads to significantly higher quality samples than all other methods tested. Our SDE solver has the benefit of requiring no step size tuning."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "The Intrinsic Dimension of Images and Its Impact on Learning",
                    "abstract": "It is widely believed that natural image data exhibits low-dimensional structure despite the high dimensionality of conventional pixel representations. This idea underlies a common intuition for the remarkable success of deep learning in computer vision. In this work, we apply dimension estimation tools to popular datasets and investigate the role of low-dimensional structure in deep learning. We find that common natural image datasets indeed have very low intrinsic dimension relative to the high number of pixels in the images. Additionally, we find that low dimensional datasets are easier for neural networks to learn, and models solving these tasks generalize better from training to test data. Along the way, we develop a technique for validating our dimension estimation tools on synthetic data generated by GANs allowing us to actively manipulate the intrinsic dimension by controlling the image generation process. Code for our experiments may be found here https://github.com/ppope/dimensions."
                },
                {
                    "title": "Improved Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) are a class of generative models which have recently been shown to produce excellent samples. We show that with a few simple modifications, DDPMs can also achieve competitive log-likelihoods while maintaining high sample quality. Additionally, we find that learning variances of the reverse diffusion process allows sampling with an order of magnitude fewer forward passes with a negligible difference in sample quality, which is important for the practical deployment of these models. We additionally use precision and recall to compare how well DDPMs and GANs cover the target distribution. Finally, we show that the sample quality and likelihood of these models scale smoothly with model capacity and training compute, making them easily scalable. We release our code at https://github.com/openai/improved-diffusion"
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "Large Scale GAN Training for High Fidelity Natural Image Synthesis",
                    "abstract": "Despite recent progress in generative image modeling, successfully generating high-resolution, diverse samples from complex datasets such as ImageNet remains an elusive goal. To this end, we train Generative Adversarial Networks at the largest scale yet attempted, and study the instabilities specific to such scale. We find that applying orthogonal regularization to the generator renders it amenable to a simple \"truncation trick,\" allowing fine control over the trade-off between sample fidelity and variety by reducing the variance of the Generator's input. Our modifications lead to models which set the new state of the art in class-conditional image synthesis. When trained on ImageNet at 128x128 resolution, our models (BigGANs) achieve an Inception Score (IS) of 166.5 and Frechet Inception Distance (FID) of 7.4, improving over the previous best IS of 52.52 and FID of 18.6."
                },
                {
                    "title": "GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium",
                    "abstract": "Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the \"Frechet Inception Distance\" (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark."
                },
                {
                    "title": "Megapixel Size Image Creation using Generative Adversarial Networks",
                    "abstract": "Since its appearance, Generative Adversarial Networks (GANs) have received a lot of interest in the AI community. In image generation several projects showed how GANs are able to generate photorealistic images but the results so far did not look adequate for the quality standard of visual media production industry. We present an optimized image generation process based on a Deep Convolutional Generative Adversarial Networks (DCGANs), in order to create photorealistic high-resolution images (up to 1024x1024 pixels). Furthermore, the system was fed with a limited dataset of images, less than two thousand images. All these results give more clue about future exploitation of GANs in Computer Graphics and Visual Effects."
                },
                {
                    "title": "Gaussian Error Linear Units (GELUs)",
                    "abstract": "We propose the Gaussian Error Linear Unit (GELU), a high-performing neural network activation function. The GELU activation function is $x\\Phi(x)$, where $\\Phi(x)$ the standard Gaussian cumulative distribution function. The GELU nonlinearity weights inputs by their value, rather than gates inputs by their sign as in ReLUs ($x\\mathbf{1}_{x>0}$). We perform an empirical evaluation of the GELU nonlinearity against the ReLU and ELU activations and find performance improvements across all considered computer vision, natural language processing, and speech tasks."
                },
                {
                    "title": "Rethinking the Inception Architecture for Computer Vision",
                    "abstract": "Convolutional networks are at the core of most state of-the-art computer vision solutions for a wide variety of tasks. Since 2014 very deep convolutional networks started to become mainstream, yielding substantial gains in various benchmarks. Although increased model size and computational cost tend to translate to immediate quality gains for most tasks (as long as enough labeled data is provided for training), computational efficiency and low parameter count are still enabling factors for various use cases such as mobile vision and big-data scenarios. Here we are exploring ways to scale up networks in ways that aim at utilizing the added computation as efficiently as possible by suitably factorized convolutions and aggressive regularization. We benchmark our methods on the ILSVRC 2012 classification challenge validation set demonstrate substantial gains over the state of the art: 21:2% top-1 and 5:6% top-5 error for single frame evaluation using a network with a computational cost of 5 billion multiply-adds per inference and with using less than 25 million parameters. With an ensemble of 4 models and multi-crop evaluation, we report 3:5% top-5 error and 17:3% top-1 error on the validation set and 3:6% top-5 error on the official test set."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "A Connection Between Score Matching and Denoising Autoencoders",
                    "abstract": "Denoising autoencoders have been previously shown to be competitive alternatives to restricted Boltzmann machines for unsupervised pretraining of each layer of a deep architecture. We show that a simple denoising autoencoder training criterion is equivalent to matching the score (with respect to the data) of a specific energy-based model to that of a nonparametric Parzen density estimator of the data. This yields several useful insights. It defines a proper probabilistic model for the denoising autoencoder technique, which makes it in principle possible to sample from them or rank examples by their energy. It suggests a different way to apply score matching that is related to learning to denoise and does not require computing second derivatives. It justifies the use of tied weights between the encoder and decoder and suggests ways to extend the success of denoising autoencoders to a larger family of energy-based models."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Interpretation and Generalization of Score Matching",
                    "abstract": "Score matching is a recently developed parameter learning method that is particularly effective to complicated high dimensional density models with intractable partition functions. In this paper, we study two issues that have not been completely resolved for score matching. First, we provide a formal link between maximum likelihood and score matching. Our analysis shows that score matching finds model parameters that are more robust with noisy training data. Second, we develop a generalization of score matching. Based on this generalization, we further demonstrate an extension of score matching to models of discrete data."
                },
                {
                    "title": "Estimation of Non-Normalized Statistical Models by Score Matching",
                    "abstract": "One often wants to estimate statistical models where the probability density function is known only up to a multiplicative normalization constant. Typically, one then has to resort to Markov Chain Monte Carlo methods, or approximations of the normalization constant. Here, we propose that such models can be estimated by minimizing the expected squared distance between the gradient of the log-density given by the model and the gradient of the log-density of the observed data. While the estimation of the gradient of log-density function is, in principle, a very difficult non-parametric problem, we prove a surprising result that gives a simple formula for this objective function. The density function of the observed data does not appear in this formula, which simplifies to a sample average of a sum of some derivatives of the log-density given by the model. The validity of the method is demonstrated on multivariate Gaussian and independent component analysis models, and by estimating an overcomplete filter set for natural image data."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI",
                "cs.LG",
                "cs.NE",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the image quality and prompt alignment in denoising diffusion models while addressing the limitations of classifier-free guidance (CFG)?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of generative models, particularly in enhancing the quality and diversity of generated images. Improved methods can lead to more accurate representations of low-probability regions in the data distribution, which is essential for applications in art, design, and content creation. By addressing the limitations of CFG, future research can explore more robust and flexible guidance techniques, potentially leading to breakthroughs in conditional and unconditional image synthesis.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexities of balancing prompt alignment and image quality without overshooting the desired conditional distribution. Naive approaches may fail because they do not account for the task discrepancy between conditional and unconditional denoising, leading to skewed image compositions. Additionally, the need to control the effects of prompt alignment and quality improvement separately adds another layer of complexity, requiring innovative methodologies to disentangle these effects.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on CFG, which, while effective, is limited to conditional generation and suffers from task discrepancy issues. Existing solutions have not adequately addressed the need for a unified approach that can handle both conditional and unconditional synthesis without compromising image quality. Our approach differs by utilizing a guiding model derived from the main model itself, which allows for better control and improved outcomes without the limitations of prior methods.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of a novel technique called autoguidance, which uses a simplified version of the main model as the guiding model while maintaining unchanged conditioning. We will validate this approach using various synthetic test cases and practical image synthesis scenarios, focusing on class-conditional and text-conditional settings. The expected outcomes include significant improvements in image distribution quality, as measured by FID and FDDINOv2 metrics, with the potential to set new records in image generation benchmarks such as ImageNet-512 and ImageNet-64."
            }
        },
        "author_data": {
            "aef72b19-c9f0-4b04-8b32-ee687da609d7": {
                "pk": "aef72b19-c9f0-4b04-8b32-ee687da609d7",
                "name": "Tero Karras",
                "collaborators": [
                    "Timo Aila",
                    "Samuli Laine",
                    "Jaakko Lehtinen",
                    "Miika Aittala",
                    "Janne Hellsten",
                    "Tuomas Kynk\u00e4\u00e4nniemi",
                    "Pavlo Molchanov",
                    "Stephen Tyree",
                    "Jan Kautz",
                    "Onni Kosomaa"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Image Synthesis",
                    "Deep Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Progressive Growing of GANs for Improved Quality, Stability, and Variation",
                        "abstract": "We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024^2. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10. Additionally, we describe several implementation details that are important for discouraging unhealthy competition between the generator and discriminator. Finally, we suggest a new metric for evaluating GAN results, both in terms of image quality and variation. As an additional contribution, we construct a higher-quality version of the CelebA dataset."
                    },
                    {
                        "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                        "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                    },
                    {
                        "title": "High-Quality Self-Supervised Deep Image Denoising",
                        "abstract": "We describe a novel method for training high-quality image denoising models based on unorganized collections of corrupted images. The training does not need access to clean reference images, or explicit pairs of corrupted images, and can thus be applied in situations where such data is unacceptably expensive or impossible to acquire. We build on a recent technique that removes the need for reference data by employing networks with a \"blind spot\" in the receptive field, and significantly improve two key aspects: image quality and training efficiency. Our result quality is on par with state-of-the-art neural network denoisers in the case of i.i.d. additive Gaussian noise, and not far behind with Poisson and impulse noise. We also successfully handle cases where parameters of the noise model are variable and/or unknown in both training and evaluation data."
                    },
                    {
                        "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                        "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                    },
                    {
                        "title": "Pruning Convolutional Neural Networks for Resource Efficient Inference",
                        "abstract": "We propose a new formulation for pruning convolutional kernels in neural networks to enable efficient inference. We interleave greedy criteria-based pruning with fine-tuning by backpropagation - a computationally efficient procedure that maintains good generalization in the pruned network. We propose a new criterion based on Taylor expansion that approximates the change in the cost function induced by pruning network parameters. We focus on transfer learning, where large pretrained networks are adapted to specialized tasks. The proposed criterion demonstrates superior performance compared to other criteria, e.g. the norm of kernel weights or feature map activation, for pruning large CNNs after adaptation to fine-grained classification tasks (Birds-200 and Flowers-102) relaying only on the first order gradient information. We also show that pruning can lead to more than 10x theoretical (5x practical) reduction in adapted 3D-convolutional filters with a small drop in accuracy in a recurrent gesture classifier. Finally, we show results for the large-scale ImageNet dataset to emphasize the flexibility of our approach."
                    },
                    {
                        "title": "Improved Precision and Recall Metric for Assessing Generative Models",
                        "abstract": "The ability to automatically estimate the quality and coverage of the samples produced by a generative model is a vital requirement for driving algorithm research. We present an evaluation metric that can separately and reliably measure both of these aspects in image generation tasks by forming explicit, non-parametric representations of the manifolds of real and generated data. We demonstrate the effectiveness of our metric in StyleGAN and BigGAN by providing several illustrative examples where existing metrics yield uninformative or contradictory results. Furthermore, we analyze multiple design variants of StyleGAN to better understand the relationships between the model architecture, training methods, and the properties of the resulting sample distribution. In the process, we identify new variants that improve the state-of-the-art. We also perform the first principled analysis of truncation methods and identify an improved method. Finally, we extend our metric to estimate the perceptual quality of individual samples, and use this to study latent space interpolations."
                    },
                    {
                        "title": "The Role of ImageNet Classes in Fr\u00e9chet Inception Distance",
                        "abstract": "Fr\\'echet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID \"looks at\" in generated images. We show that the feature space that FID is (typically) computed in is so close to the ImageNet classifications that aligning the histograms of Top-$N$ classifications between sets of generated and real images can reduce FID substantially -- without actually improving the quality of results. Thus, we conclude that FID is prone to intentional or accidental distortions. As a practical example of an accidental distortion, we discuss a case where an ImageNet pre-trained FastGAN achieves a FID comparable to StyleGAN2, while being worse in terms of human evaluation."
                    },
                    {
                        "title": "Simulator-Based Self-Supervision for Learned 3D Tomography Reconstruction",
                        "abstract": "We propose a deep learning method for 3D volumetric reconstruction in low-dose helical cone-beam computed tomography. Prior machine learning approaches require reference reconstructions computed by another algorithm for training. In contrast, we train our model in a fully self-supervised manner using only noisy 2D X-ray data. This is enabled by incorporating a fast differentiable CT simulator in the training loop. As we do not rely on reference reconstructions, the fidelity of our results is not limited by their potential shortcomings. We evaluate our method on real helical cone-beam projections and simulated phantoms. Our results show significantly higher visual fidelity and better PSNR over techniques that rely on existing reconstructions. When applied to full-dose data, our method produces high-quality results orders of magnitude faster than iterative techniques."
                    },
                    {
                        "title": "StyleGAN-T: Unlocking the Power of GANs for Fast Large-Scale Text-to-Image Synthesis",
                        "abstract": "Text-to-image synthesis has recently seen significant progress thanks to large pretrained language models, large-scale training data, and the introduction of scalable model families such as diffusion and autoregressive models. However, the best-performing models require iterative evaluation to generate a single sample. In contrast, generative adversarial networks (GANs) only need a single forward pass. They are thus much faster, but they currently remain far behind the state-of-the-art in large-scale text-to-image synthesis. This paper aims to identify the necessary steps to regain competitiveness. Our proposed model, StyleGAN-T, addresses the specific requirements of large-scale text-to-image synthesis, such as large capacity, stable training on diverse datasets, strong text alignment, and controllable variation vs. text alignment tradeoff. StyleGAN-T significantly improves over previous GANs and outperforms distilled diffusion models - the previous state-of-the-art in fast text-to-image synthesis - in terms of sample quality and speed."
                    },
                    {
                        "title": "Analyzing and Improving the Image Quality of StyleGAN",
                        "abstract": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign the generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent codes to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably attribute a generated image to a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality."
                    },
                    {
                        "title": "Noise2Noise: Learning Image Restoration without Clean Data",
                        "abstract": "We apply basic statistical reasoning to signal reconstruction by machine learning -- learning to map corrupted observations to clean signals -- with a simple and powerful conclusion: it is possible to learn to restore images by only looking at corrupted examples, at performance at and sometimes exceeding training using clean data, without explicit image priors or likelihood models of the corruption. In practice, we show that a single model learns photographic noise removal, denoising synthetic Monte Carlo images, and reconstruction of undersampled MRI scans -- all corrupted by different processes -- based on noisy data only."
                    },
                    {
                        "title": "Training Generative Adversarial Networks with Limited Data",
                        "abstract": "Training generative adversarial networks (GAN) using too little data typically leads to discriminator overfitting, causing training to diverge. We propose an adaptive discriminator augmentation mechanism that significantly stabilizes training in limited data regimes. The approach does not require changes to loss functions or network architectures, and is applicable both when training from scratch and when fine-tuning an existing GAN on another dataset. We demonstrate, on several datasets, that good results are now possible using only a few thousand training images, often matching StyleGAN2 results with an order of magnitude fewer images. We expect this to open up new application domains for GANs. We also find that the widely used CIFAR-10 is, in fact, a limited data benchmark, and improve the record FID from 5.59 to 2.42."
                    },
                    {
                        "title": "Modular Primitives for High-Performance Differentiable Rendering",
                        "abstract": "We present a modular differentiable renderer design that yields performance superior to previous methods by leveraging existing, highly optimized hardware graphics pipelines. Our design supports all crucial operations in a modern graphics pipeline: rasterizing large numbers of triangles, attribute interpolation, filtered texture lookups, as well as user-programmable shading and geometry processing, all in high resolutions. Our modular primitives allow custom, high-performance graphics pipelines to be built directly within automatic differentiation frameworks such as PyTorch or TensorFlow. As a motivating application, we formulate facial performance capture as an inverse rendering problem and show that it can be solved efficiently using our tools. Our results indicate that this simple and straightforward approach achieves excellent geometric correspondence between rendered results and reference imagery."
                    },
                    {
                        "title": "Alias-Free Generative Adversarial Networks",
                        "abstract": "We observe that despite their hierarchical convolutional nature, the synthesis process of typical generative adversarial networks depends on absolute pixel coordinates in an unhealthy manner. This manifests itself as, e.g., detail appearing to be glued to image coordinates instead of the surfaces of depicted objects. We trace the root cause to careless signal processing that causes aliasing in the generator network. Interpreting all signals in the network as continuous, we derive generally applicable, small architectural changes that guarantee that unwanted information cannot leak into the hierarchical synthesis process. The resulting networks match the FID of StyleGAN2 but differ dramatically in their internal representations, and they are fully equivariant to translation and rotation even at subpixel scales. Our results pave the way for generative models better suited for video and animation."
                    },
                    {
                        "title": "Analyzing and Improving the Training Dynamics of Diffusion Models",
                        "abstract": "Diffusion models currently dominate the field of data-driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high-level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on expectation. We find that systematic application of this philosophy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational complexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling.   As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance."
                    },
                    {
                        "title": "Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models",
                        "abstract": "Guidance is a crucial technique for extracting the best performance out of image-generating diffusion models. Traditionally, a constant guidance weight has been applied throughout the sampling chain of an image. We show that guidance is clearly harmful toward the beginning of the chain (high noise levels), largely unnecessary toward the end (low noise levels), and only beneficial in the middle. We thus restrict it to a specific range of noise levels, improving both the inference speed and result quality. This limited guidance interval improves the record FID in ImageNet-512 significantly, from 1.81 to 1.40. We show that it is quantitatively and qualitatively beneficial across different sampler parameters, network architectures, and datasets, including the large-scale setting of Stable Diffusion XL. We thus suggest exposing the guidance interval as a hyperparameter in all diffusion models that use guidance."
                    }
                ]
            },
            "59c5b121-1c93-4c90-9c2e-5c4ea70dfc4c": {
                "pk": "59c5b121-1c93-4c90-9c2e-5c4ea70dfc4c",
                "name": "Miika Aittala",
                "collaborators": [
                    "Samuli Laine",
                    "Jaakko Lehtinen",
                    "Tero Karras",
                    "Timo Aila",
                    "Janne Hellsten",
                    "Fredo Durand",
                    "Tuomas Kynk\u00e4\u00e4nniemi",
                    "Lukas Murmann",
                    "Jon Hasselgren",
                    "Jacob Munkberg"
                ],
                "domain": [
                    "Computer Vision",
                    "Generative Models",
                    "Image Processing",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "A Dataset of Multi-Illumination Images in the Wild",
                        "abstract": "Collections of images under a single, uncontrolled illumination have enabled the rapid advancement of core computer vision tasks like classification, detection, and segmentation. But even with modern learning techniques, many inverse problems involving lighting and material understanding remain too severely ill-posed to be solved with single-illumination datasets. To fill this gap, we introduce a new multi-illumination dataset of more than 1000 real scenes, each captured under 25 lighting conditions. We demonstrate the richness of this dataset by training state-of-the-art models for three challenging applications: single-image illumination estimation, image relighting, and mixed-illuminant white balance."
                    },
                    {
                        "title": "Generating Training Data for Denoising Real RGB Images via Camera Pipeline Simulation",
                        "abstract": "Image reconstruction techniques such as denoising often need to be applied to the RGB output of cameras and cellphones. Unfortunately, the commonly used additive white noise (AWGN) models do not accurately reproduce the noise and the degradation encountered on these inputs. This is particularly important for learning-based techniques, because the mismatch between training and real world data will hurt their generalization. This paper aims to accurately simulate the degradation and noise transformation performed by camera pipelines. This allows us to generate realistic degradation in RGB images that can be used to train machine learning models. We use our simulation to study the importance of noise modeling for learning-based denoising. Our study shows that a realistic noise model is required for learning to denoise real JPEG images. A neural network trained on realistic noise outperforms the one trained with AWGN by 3 dB. An ablation study of our pipeline shows that simulating denoising and demosaicking is important to this improvement and that realistic demosaicking algorithms, which have been rarely considered, is needed. We believe this simulation will also be useful for other image reconstruction tasks, and we will distribute our code publicly."
                    },
                    {
                        "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                        "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                    },
                    {
                        "title": "The Role of ImageNet Classes in Fr\u00e9chet Inception Distance",
                        "abstract": "Fr\\'echet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID \"looks at\" in generated images. We show that the feature space that FID is (typically) computed in is so close to the ImageNet classifications that aligning the histograms of Top-$N$ classifications between sets of generated and real images can reduce FID substantially -- without actually improving the quality of results. Thus, we conclude that FID is prone to intentional or accidental distortions. As a practical example of an accidental distortion, we discuss a case where an ImageNet pre-trained FastGAN achieves a FID comparable to StyleGAN2, while being worse in terms of human evaluation."
                    },
                    {
                        "title": "Simulator-Based Self-Supervision for Learned 3D Tomography Reconstruction",
                        "abstract": "We propose a deep learning method for 3D volumetric reconstruction in low-dose helical cone-beam computed tomography. Prior machine learning approaches require reference reconstructions computed by another algorithm for training. In contrast, we train our model in a fully self-supervised manner using only noisy 2D X-ray data. This is enabled by incorporating a fast differentiable CT simulator in the training loop. As we do not rely on reference reconstructions, the fidelity of our results is not limited by their potential shortcomings. We evaluate our method on real helical cone-beam projections and simulated phantoms. Our results show significantly higher visual fidelity and better PSNR over techniques that rely on existing reconstructions. When applied to full-dose data, our method produces high-quality results orders of magnitude faster than iterative techniques."
                    },
                    {
                        "title": "Appearance-Driven Automatic 3D Model Simplification",
                        "abstract": "We present a suite of techniques for jointly optimizing triangle meshes and shading models to match the appearance of reference scenes. This capability has a number of uses, including appearance-preserving simplification of extremely complex assets, conversion between rendering systems, and even conversion between geometric scene representations.   We follow and extend the classic analysis-by-synthesis family of techniques: enabled by a highly efficient differentiable renderer and modern nonlinear optimization algorithms, our results are driven to minimize the image-space difference to the target scene when rendered in similar viewing and lighting conditions. As the only signals driving the optimization are differences in rendered images, the approach is highly general and versatile: it easily supports many different forward rendering models such as normal mapping, spatially-varying BRDFs, displacement mapping, etc. Supervision through images only is also key to the ability to easily convert between rendering systems and scene representations.   We output triangle meshes with textured materials to ensure that the models render efficiently on modern graphics hardware and benefit from, e.g., hardware-accelerated rasterization, ray tracing, and filtered texture lookups. Our system is integrated in a small Python code base, and can be applied at high resolutions and on large models. We describe several use cases, including mesh decimation, level of detail generation, seamless mesh filtering and approximations of aggregate geometry."
                    },
                    {
                        "title": "Computational Mirrors: Blind Inverse Light Transport by Deep Matrix Factorization",
                        "abstract": "We recover a video of the motion taking place in a hidden scene by observing changes in indirect illumination in a nearby uncalibrated visible region. We solve this problem by factoring the observed video into a matrix product between the unknown hidden scene video and an unknown light transport matrix. This task is extremely ill-posed, as any non-negative factorization will satisfy the data. Inspired by recent work on the Deep Image Prior, we parameterize the factor matrices using randomly initialized convolutional neural networks trained in a one-off manner, and show that this results in decompositions that reflect the true motion in the hidden scene."
                    },
                    {
                        "title": "Analyzing and Improving the Image Quality of StyleGAN",
                        "abstract": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign the generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent codes to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably attribute a generated image to a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality."
                    },
                    {
                        "title": "Noise2Noise: Learning Image Restoration without Clean Data",
                        "abstract": "We apply basic statistical reasoning to signal reconstruction by machine learning -- learning to map corrupted observations to clean signals -- with a simple and powerful conclusion: it is possible to learn to restore images by only looking at corrupted examples, at performance at and sometimes exceeding training using clean data, without explicit image priors or likelihood models of the corruption. In practice, we show that a single model learns photographic noise removal, denoising synthetic Monte Carlo images, and reconstruction of undersampled MRI scans -- all corrupted by different processes -- based on noisy data only."
                    },
                    {
                        "title": "Flexible SVBRDF Capture with a Multi-Image Deep Network",
                        "abstract": "Empowered by deep learning, recent methods for material capture can estimate a spatially-varying reflectance from a single photograph. Such lightweight capture is in stark contrast with the tens or hundreds of pictures required by traditional optimization-based approaches. However, a single image is often simply not enough to observe the rich appearance of real-world materials. We present a deep-learning method capable of estimating material appearance from a variable number of uncalibrated and unordered pictures captured with a handheld camera and flash. Thanks to an order-independent fusing layer, this architecture extracts the most useful information from each picture, while benefiting from strong priors learned from data. The method can handle both view and light direction variation without calibration. We show how our method improves its prediction with the number of input pictures, and reaches high quality reconstructions with as little as 1 to 10 images -- a sweet spot between existing single-image and complex multi-image approaches."
                    },
                    {
                        "title": "Training Generative Adversarial Networks with Limited Data",
                        "abstract": "Training generative adversarial networks (GAN) using too little data typically leads to discriminator overfitting, causing training to diverge. We propose an adaptive discriminator augmentation mechanism that significantly stabilizes training in limited data regimes. The approach does not require changes to loss functions or network architectures, and is applicable both when training from scratch and when fine-tuning an existing GAN on another dataset. We demonstrate, on several datasets, that good results are now possible using only a few thousand training images, often matching StyleGAN2 results with an order of magnitude fewer images. We expect this to open up new application domains for GANs. We also find that the widely used CIFAR-10 is, in fact, a limited data benchmark, and improve the record FID from 5.59 to 2.42."
                    },
                    {
                        "title": "Alias-Free Generative Adversarial Networks",
                        "abstract": "We observe that despite their hierarchical convolutional nature, the synthesis process of typical generative adversarial networks depends on absolute pixel coordinates in an unhealthy manner. This manifests itself as, e.g., detail appearing to be glued to image coordinates instead of the surfaces of depicted objects. We trace the root cause to careless signal processing that causes aliasing in the generator network. Interpreting all signals in the network as continuous, we derive generally applicable, small architectural changes that guarantee that unwanted information cannot leak into the hierarchical synthesis process. The resulting networks match the FID of StyleGAN2 but differ dramatically in their internal representations, and they are fully equivariant to translation and rotation even at subpixel scales. Our results pave the way for generative models better suited for video and animation."
                    },
                    {
                        "title": "Analyzing and Improving the Training Dynamics of Diffusion Models",
                        "abstract": "Diffusion models currently dominate the field of data-driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high-level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on expectation. We find that systematic application of this philosophy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational complexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling.   As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance."
                    },
                    {
                        "title": "Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models",
                        "abstract": "Guidance is a crucial technique for extracting the best performance out of image-generating diffusion models. Traditionally, a constant guidance weight has been applied throughout the sampling chain of an image. We show that guidance is clearly harmful toward the beginning of the chain (high noise levels), largely unnecessary toward the end (low noise levels), and only beneficial in the middle. We thus restrict it to a specific range of noise levels, improving both the inference speed and result quality. This limited guidance interval improves the record FID in ImageNet-512 significantly, from 1.81 to 1.40. We show that it is quantitatively and qualitatively beneficial across different sampler parameters, network architectures, and datasets, including the large-scale setting of Stable Diffusion XL. We thus suggest exposing the guidance interval as a hyperparameter in all diffusion models that use guidance."
                    },
                    {
                        "title": "Disentangling Random and Cyclic Effects in Time-Lapse Sequences",
                        "abstract": "Time-lapse image sequences offer visually compelling insights into dynamic processes that are too slow to observe in real time. However, playing a long time-lapse sequence back as a video often results in distracting flicker due to random effects, such as weather, as well as cyclic effects, such as the day-night cycle. We introduce the problem of disentangling time-lapse sequences in a way that allows separate, after-the-fact control of overall trends, cyclic effects, and random effects in the images, and describe a technique based on data-driven generative models that achieves this goal. This enables us to \"re-render\" the sequences in ways that would not be possible with the input images alone. For example, we can stabilize a long sequence to focus on plant growth over many months, under selectable, consistent weather.   Our approach is based on Generative Adversarial Networks (GAN) that are conditioned with the time coordinate of the time-lapse sequence. Our architecture and training procedure are designed so that the networks learn to model random variations, such as weather, using the GAN's latent space, and to disentangle overall trends and cyclic variations by feeding the conditioning time label to the model using Fourier features with specific frequencies.   We show that our models are robust to defects in the training data, enabling us to amend some of the practical difficulties in capturing long time-lapse sequences, such as temporary occlusions, uneven frame spacing, and missing frames."
                    }
                ]
            },
            "a209ed16-d994-49c7-9a32-4f2c1a4dcfe9": {
                "pk": "a209ed16-d994-49c7-9a32-4f2c1a4dcfe9",
                "name": "Tuomas Kynk\u00e4\u00e4nniemi",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "6a1864bc-3fff-4137-b13c-76812fc7934a": {
                "pk": "6a1864bc-3fff-4137-b13c-76812fc7934a",
                "name": "Jaakko Lehtinen",
                "collaborators": [
                    "Tero Karras",
                    "Samuli Laine",
                    "Timo Aila",
                    "Miika Aittala",
                    "Erik H\u00e4rk\u00f6nen",
                    "Tuomas Kynk\u00e4\u00e4nniemi",
                    "Markus Kettunen",
                    "Aaron Hertzmann",
                    "Sylvain Paris",
                    "Eloi Moliner"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Image Processing",
                    "Reinforcement Learning",
                    "Audio Processing"
                ],
                "publications": [
                    {
                        "title": "E-LPIPS: Robust Perceptual Image Similarity via Random Transformation Ensembles",
                        "abstract": "It has been recently shown that the hidden variables of convolutional neural networks make for an efficient perceptual similarity metric that accurately predicts human judgment on relative image similarity assessment. First, we show that such learned perceptual similarity metrics (LPIPS) are susceptible to adversarial attacks that dramatically contradict human visual similarity judgment. While this is not surprising in light of neural networks' well-known weakness to adversarial perturbations, we proceed to show that self-ensembling with an infinite family of random transformations of the input --- a technique known not to render classification networks robust --- is enough to turn the metric robust against attack, while retaining predictive power on human judgments. Finally, we study the geometry imposed by our our novel self-ensembled metric (E-LPIPS) on the space of natural images. We find evidence of \"perceptual convexity\" by showing that convex combinations of similar-looking images retain appearance, and that discrete geodesics yield meaningful frame interpolation and texture morphing, all without explicit correspondences."
                    },
                    {
                        "title": "Progressive Growing of GANs for Improved Quality, Stability, and Variation",
                        "abstract": "We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024^2. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10. Additionally, we describe several implementation details that are important for discouraging unhealthy competition between the generator and discriminator. Finally, we suggest a new metric for evaluating GAN results, both in terms of image quality and variation. As an additional contribution, we construct a higher-quality version of the CelebA dataset."
                    },
                    {
                        "title": "High-Quality Self-Supervised Deep Image Denoising",
                        "abstract": "We describe a novel method for training high-quality image denoising models based on unorganized collections of corrupted images. The training does not need access to clean reference images, or explicit pairs of corrupted images, and can thus be applied in situations where such data is unacceptably expensive or impossible to acquire. We build on a recent technique that removes the need for reference data by employing networks with a \"blind spot\" in the receptive field, and significantly improve two key aspects: image quality and training efficiency. Our result quality is on par with state-of-the-art neural network denoisers in the case of i.i.d. additive Gaussian noise, and not far behind with Poisson and impulse noise. We also successfully handle cases where parameters of the noise model are variable and/or unknown in both training and evaluation data."
                    },
                    {
                        "title": "GANSpace: Discovering Interpretable GAN Controls",
                        "abstract": "This paper describes a simple technique to analyze Generative Adversarial Networks (GANs) and create interpretable controls for image synthesis, such as change of viewpoint, aging, lighting, and time of day. We identify important latent directions based on Principal Components Analysis (PCA) applied either in latent space or feature space. Then, we show that a large number of interpretable controls can be defined by layer-wise perturbation along the principal directions. Moreover, we show that BigGAN can be controlled with layer-wise inputs in a StyleGAN-like manner. We show results on different GANs trained on various datasets, and demonstrate good qualitative matches to edit directions found through earlier supervised approaches."
                    },
                    {
                        "title": "Solving Audio Inverse Problems with a Diffusion Model",
                        "abstract": "This paper presents CQT-Diff, a data-driven generative audio model that can, once trained, be used for solving various different audio inverse problems in a problem-agnostic setting. CQT-Diff is a neural diffusion model with an architecture that is carefully constructed to exploit pitch-equivariant symmetries in music. This is achieved by preconditioning the model with an invertible Constant-Q Transform (CQT), whose logarithmically-spaced frequency axis represents pitch equivariance as translation equivariance. The proposed method is evaluated with objective and subjective metrics in three different and varied tasks: audio bandwidth extension, inpainting, and declipping. The results show that CQT-Diff outperforms the compared baselines and ablations in audio bandwidth extension and, without retraining, delivers competitive performance against modern baselines in audio inpainting and declipping. This work represents the first diffusion-based general framework for solving inverse problems in audio processing."
                    },
                    {
                        "title": "PPO-CMA: Proximal Policy Optimization with Covariance Matrix Adaptation",
                        "abstract": "Proximal Policy Optimization (PPO) is a highly popular model-free reinforcement learning (RL) approach. However, we observe that in a continuous action space, PPO can prematurely shrink the exploration variance, which leads to slow progress and may make the algorithm prone to getting stuck in local optima. Drawing inspiration from CMA-ES, a black-box evolutionary optimization method designed for robustness in similar situations, we propose PPO-CMA, a proximal policy optimization approach that adaptively expands the exploration variance to speed up progress. With only minor changes to PPO, our algorithm considerably improves performance in Roboschool continuous control benchmarks. Our results also show that PPO-CMA, as opposed to PPO, is significantly less sensitive to the choice of hyperparameters, allowing one to use it in complex movement optimization tasks without requiring tedious tuning."
                    },
                    {
                        "title": "Improved Precision and Recall Metric for Assessing Generative Models",
                        "abstract": "The ability to automatically estimate the quality and coverage of the samples produced by a generative model is a vital requirement for driving algorithm research. We present an evaluation metric that can separately and reliably measure both of these aspects in image generation tasks by forming explicit, non-parametric representations of the manifolds of real and generated data. We demonstrate the effectiveness of our metric in StyleGAN and BigGAN by providing several illustrative examples where existing metrics yield uninformative or contradictory results. Furthermore, we analyze multiple design variants of StyleGAN to better understand the relationships between the model architecture, training methods, and the properties of the resulting sample distribution. In the process, we identify new variants that improve the state-of-the-art. We also perform the first principled analysis of truncation methods and identify an improved method. Finally, we extend our metric to estimate the perceptual quality of individual samples, and use this to study latent space interpolations."
                    },
                    {
                        "title": "The Role of ImageNet Classes in Fr\u00e9chet Inception Distance",
                        "abstract": "Fr\\'echet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID \"looks at\" in generated images. We show that the feature space that FID is (typically) computed in is so close to the ImageNet classifications that aligning the histograms of Top-$N$ classifications between sets of generated and real images can reduce FID substantially -- without actually improving the quality of results. Thus, we conclude that FID is prone to intentional or accidental distortions. As a practical example of an accidental distortion, we discuss a case where an ImageNet pre-trained FastGAN achieves a FID comparable to StyleGAN2, while being worse in terms of human evaluation."
                    },
                    {
                        "title": "Simulator-Based Self-Supervision for Learned 3D Tomography Reconstruction",
                        "abstract": "We propose a deep learning method for 3D volumetric reconstruction in low-dose helical cone-beam computed tomography. Prior machine learning approaches require reference reconstructions computed by another algorithm for training. In contrast, we train our model in a fully self-supervised manner using only noisy 2D X-ray data. This is enabled by incorporating a fast differentiable CT simulator in the training loop. As we do not rely on reference reconstructions, the fidelity of our results is not limited by their potential shortcomings. We evaluate our method on real helical cone-beam projections and simulated phantoms. Our results show significantly higher visual fidelity and better PSNR over techniques that rely on existing reconstructions. When applied to full-dose data, our method produces high-quality results orders of magnitude faster than iterative techniques."
                    },
                    {
                        "title": "Appearance-Driven Automatic 3D Model Simplification",
                        "abstract": "We present a suite of techniques for jointly optimizing triangle meshes and shading models to match the appearance of reference scenes. This capability has a number of uses, including appearance-preserving simplification of extremely complex assets, conversion between rendering systems, and even conversion between geometric scene representations.   We follow and extend the classic analysis-by-synthesis family of techniques: enabled by a highly efficient differentiable renderer and modern nonlinear optimization algorithms, our results are driven to minimize the image-space difference to the target scene when rendered in similar viewing and lighting conditions. As the only signals driving the optimization are differences in rendered images, the approach is highly general and versatile: it easily supports many different forward rendering models such as normal mapping, spatially-varying BRDFs, displacement mapping, etc. Supervision through images only is also key to the ability to easily convert between rendering systems and scene representations.   We output triangle meshes with textured materials to ensure that the models render efficiently on modern graphics hardware and benefit from, e.g., hardware-accelerated rasterization, ray tracing, and filtered texture lookups. Our system is integrated in a small Python code base, and can be applied at high resolutions and on large models. We describe several use cases, including mesh decimation, level of detail generation, seamless mesh filtering and approximations of aggregate geometry."
                    }
                ]
            },
            "eaec9096-a850-4dac-a8d6-1a9a41a2d963": {
                "pk": "eaec9096-a850-4dac-a8d6-1a9a41a2d963",
                "name": "Timo Aila",
                "collaborators": [
                    "Samuli Laine",
                    "Tero Karras",
                    "Jaakko Lehtinen",
                    "Miika Aittala",
                    "Janne Hellsten",
                    "Tuomas Kynk\u00e4\u00e4nniemi",
                    "Erik H\u00e4rk\u00f6nen",
                    "Pavlo Molchanov",
                    "Stephen Tyree",
                    "Jan Kautz"
                ],
                "domain": [
                    "Generative Models",
                    "Semi-Supervised Learning",
                    "Image Synthesis",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Temporal Ensembling for Semi-Supervised Learning",
                        "abstract": "In this paper, we present a simple and efficient method for training deep neural networks in a semi-supervised setting where only a small portion of training data is labeled. We introduce self-ensembling, where we form a consensus prediction of the unknown labels using the outputs of the network-in-training on different epochs, and most importantly, under different regularization and input augmentation conditions. This ensemble prediction can be expected to be a better predictor for the unknown labels than the output of the network at the most recent training epoch, and can thus be used as a target for training. Using our method, we set new records for two standard semi-supervised learning benchmarks, reducing the (non-augmented) classification error rate from 18.44% to 7.05% in SVHN with 500 labels and from 18.63% to 16.55% in CIFAR-10 with 4000 labels, and further to 5.12% and 12.16% by enabling the standard augmentations. We additionally obtain a clear improvement in CIFAR-100 classification accuracy by using random images from the Tiny Images dataset as unlabeled extra inputs during training. Finally, we demonstrate good tolerance to incorrect labels."
                    },
                    {
                        "title": "Progressive Growing of GANs for Improved Quality, Stability, and Variation",
                        "abstract": "We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024^2. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10. Additionally, we describe several implementation details that are important for discouraging unhealthy competition between the generator and discriminator. Finally, we suggest a new metric for evaluating GAN results, both in terms of image quality and variation. As an additional contribution, we construct a higher-quality version of the CelebA dataset."
                    },
                    {
                        "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                        "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                    },
                    {
                        "title": "High-Quality Self-Supervised Deep Image Denoising",
                        "abstract": "We describe a novel method for training high-quality image denoising models based on unorganized collections of corrupted images. The training does not need access to clean reference images, or explicit pairs of corrupted images, and can thus be applied in situations where such data is unacceptably expensive or impossible to acquire. We build on a recent technique that removes the need for reference data by employing networks with a \"blind spot\" in the receptive field, and significantly improve two key aspects: image quality and training efficiency. Our result quality is on par with state-of-the-art neural network denoisers in the case of i.i.d. additive Gaussian noise, and not far behind with Poisson and impulse noise. We also successfully handle cases where parameters of the noise model are variable and/or unknown in both training and evaluation data."
                    },
                    {
                        "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                        "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                    },
                    {
                        "title": "Pruning Convolutional Neural Networks for Resource Efficient Inference",
                        "abstract": "We propose a new formulation for pruning convolutional kernels in neural networks to enable efficient inference. We interleave greedy criteria-based pruning with fine-tuning by backpropagation - a computationally efficient procedure that maintains good generalization in the pruned network. We propose a new criterion based on Taylor expansion that approximates the change in the cost function induced by pruning network parameters. We focus on transfer learning, where large pretrained networks are adapted to specialized tasks. The proposed criterion demonstrates superior performance compared to other criteria, e.g. the norm of kernel weights or feature map activation, for pruning large CNNs after adaptation to fine-grained classification tasks (Birds-200 and Flowers-102) relaying only on the first order gradient information. We also show that pruning can lead to more than 10x theoretical (5x practical) reduction in adapted 3D-convolutional filters with a small drop in accuracy in a recurrent gesture classifier. Finally, we show results for the large-scale ImageNet dataset to emphasize the flexibility of our approach."
                    },
                    {
                        "title": "Improved Precision and Recall Metric for Assessing Generative Models",
                        "abstract": "The ability to automatically estimate the quality and coverage of the samples produced by a generative model is a vital requirement for driving algorithm research. We present an evaluation metric that can separately and reliably measure both of these aspects in image generation tasks by forming explicit, non-parametric representations of the manifolds of real and generated data. We demonstrate the effectiveness of our metric in StyleGAN and BigGAN by providing several illustrative examples where existing metrics yield uninformative or contradictory results. Furthermore, we analyze multiple design variants of StyleGAN to better understand the relationships between the model architecture, training methods, and the properties of the resulting sample distribution. In the process, we identify new variants that improve the state-of-the-art. We also perform the first principled analysis of truncation methods and identify an improved method. Finally, we extend our metric to estimate the perceptual quality of individual samples, and use this to study latent space interpolations."
                    },
                    {
                        "title": "The Role of ImageNet Classes in Fr\u00e9chet Inception Distance",
                        "abstract": "Fr\\'echet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID \"looks at\" in generated images. We show that the feature space that FID is (typically) computed in is so close to the ImageNet classifications that aligning the histograms of Top-$N$ classifications between sets of generated and real images can reduce FID substantially -- without actually improving the quality of results. Thus, we conclude that FID is prone to intentional or accidental distortions. As a practical example of an accidental distortion, we discuss a case where an ImageNet pre-trained FastGAN achieves a FID comparable to StyleGAN2, while being worse in terms of human evaluation."
                    },
                    {
                        "title": "StyleGAN-T: Unlocking the Power of GANs for Fast Large-Scale Text-to-Image Synthesis",
                        "abstract": "Text-to-image synthesis has recently seen significant progress thanks to large pretrained language models, large-scale training data, and the introduction of scalable model families such as diffusion and autoregressive models. However, the best-performing models require iterative evaluation to generate a single sample. In contrast, generative adversarial networks (GANs) only need a single forward pass. They are thus much faster, but they currently remain far behind the state-of-the-art in large-scale text-to-image synthesis. This paper aims to identify the necessary steps to regain competitiveness. Our proposed model, StyleGAN-T, addresses the specific requirements of large-scale text-to-image synthesis, such as large capacity, stable training on diverse datasets, strong text alignment, and controllable variation vs. text alignment tradeoff. StyleGAN-T significantly improves over previous GANs and outperforms distilled diffusion models - the previous state-of-the-art in fast text-to-image synthesis - in terms of sample quality and speed."
                    },
                    {
                        "title": "Analyzing and Improving the Image Quality of StyleGAN",
                        "abstract": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign the generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent codes to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably attribute a generated image to a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality."
                    },
                    {
                        "title": "Noise2Noise: Learning Image Restoration without Clean Data",
                        "abstract": "We apply basic statistical reasoning to signal reconstruction by machine learning -- learning to map corrupted observations to clean signals -- with a simple and powerful conclusion: it is possible to learn to restore images by only looking at corrupted examples, at performance at and sometimes exceeding training using clean data, without explicit image priors or likelihood models of the corruption. In practice, we show that a single model learns photographic noise removal, denoising synthetic Monte Carlo images, and reconstruction of undersampled MRI scans -- all corrupted by different processes -- based on noisy data only."
                    },
                    {
                        "title": "Semi-supervised semantic segmentation needs strong, varied perturbations",
                        "abstract": "Consistency regularization describes a class of approaches that have yielded ground breaking results in semi-supervised classification problems. Prior work has established the cluster assumption - under which the data distribution consists of uniform class clusters of samples separated by low density regions - as important to its success. We analyze the problem of semantic segmentation and find that its' distribution does not exhibit low density regions separating classes and offer this as an explanation for why semi-supervised segmentation is a challenging problem, with only a few reports of success. We then identify choice of augmentation as key to obtaining reliable performance without such low-density regions. We find that adapted variants of the recently proposed CutOut and CutMix augmentation techniques yield state-of-the-art semi-supervised semantic segmentation results in standard datasets. Furthermore, given its challenging nature we propose that semantic segmentation acts as an effective acid test for evaluating semi-supervised regularizers. Implementation at: https://github.com/Britefury/cutmix-semisup-seg."
                    },
                    {
                        "title": "Training Generative Adversarial Networks with Limited Data",
                        "abstract": "Training generative adversarial networks (GAN) using too little data typically leads to discriminator overfitting, causing training to diverge. We propose an adaptive discriminator augmentation mechanism that significantly stabilizes training in limited data regimes. The approach does not require changes to loss functions or network architectures, and is applicable both when training from scratch and when fine-tuning an existing GAN on another dataset. We demonstrate, on several datasets, that good results are now possible using only a few thousand training images, often matching StyleGAN2 results with an order of magnitude fewer images. We expect this to open up new application domains for GANs. We also find that the widely used CIFAR-10 is, in fact, a limited data benchmark, and improve the record FID from 5.59 to 2.42."
                    },
                    {
                        "title": "Modular Primitives for High-Performance Differentiable Rendering",
                        "abstract": "We present a modular differentiable renderer design that yields performance superior to previous methods by leveraging existing, highly optimized hardware graphics pipelines. Our design supports all crucial operations in a modern graphics pipeline: rasterizing large numbers of triangles, attribute interpolation, filtered texture lookups, as well as user-programmable shading and geometry processing, all in high resolutions. Our modular primitives allow custom, high-performance graphics pipelines to be built directly within automatic differentiation frameworks such as PyTorch or TensorFlow. As a motivating application, we formulate facial performance capture as an inverse rendering problem and show that it can be solved efficiently using our tools. Our results indicate that this simple and straightforward approach achieves excellent geometric correspondence between rendered results and reference imagery."
                    },
                    {
                        "title": "Alias-Free Generative Adversarial Networks",
                        "abstract": "We observe that despite their hierarchical convolutional nature, the synthesis process of typical generative adversarial networks depends on absolute pixel coordinates in an unhealthy manner. This manifests itself as, e.g., detail appearing to be glued to image coordinates instead of the surfaces of depicted objects. We trace the root cause to careless signal processing that causes aliasing in the generator network. Interpreting all signals in the network as continuous, we derive generally applicable, small architectural changes that guarantee that unwanted information cannot leak into the hierarchical synthesis process. The resulting networks match the FID of StyleGAN2 but differ dramatically in their internal representations, and they are fully equivariant to translation and rotation even at subpixel scales. Our results pave the way for generative models better suited for video and animation."
                    },
                    {
                        "title": "Analyzing and Improving the Training Dynamics of Diffusion Models",
                        "abstract": "Diffusion models currently dominate the field of data-driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high-level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on expectation. We find that systematic application of this philosophy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational complexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling.   As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance."
                    },
                    {
                        "title": "Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models",
                        "abstract": "Guidance is a crucial technique for extracting the best performance out of image-generating diffusion models. Traditionally, a constant guidance weight has been applied throughout the sampling chain of an image. We show that guidance is clearly harmful toward the beginning of the chain (high noise levels), largely unnecessary toward the end (low noise levels), and only beneficial in the middle. We thus restrict it to a specific range of noise levels, improving both the inference speed and result quality. This limited guidance interval improves the record FID in ImageNet-512 significantly, from 1.81 to 1.40. We show that it is quantitatively and qualitatively beneficial across different sampler parameters, network architectures, and datasets, including the large-scale setting of Stable Diffusion XL. We thus suggest exposing the guidance interval as a hyperparameter in all diffusion models that use guidance."
                    },
                    {
                        "title": "Disentangling Random and Cyclic Effects in Time-Lapse Sequences",
                        "abstract": "Time-lapse image sequences offer visually compelling insights into dynamic processes that are too slow to observe in real time. However, playing a long time-lapse sequence back as a video often results in distracting flicker due to random effects, such as weather, as well as cyclic effects, such as the day-night cycle. We introduce the problem of disentangling time-lapse sequences in a way that allows separate, after-the-fact control of overall trends, cyclic effects, and random effects in the images, and describe a technique based on data-driven generative models that achieves this goal. This enables us to \"re-render\" the sequences in ways that would not be possible with the input images alone. For example, we can stabilize a long sequence to focus on plant growth over many months, under selectable, consistent weather.   Our approach is based on Generative Adversarial Networks (GAN) that are conditioned with the time coordinate of the time-lapse sequence. Our architecture and training procedure are designed so that the networks learn to model random variations, such as weather, using the GAN's latent space, and to disentangle overall trends and cyclic variations by feeding the conditioning time label to the model using Fourier features with specific frequencies.   We show that our models are robust to defects in the training data, enabling us to amend some of the practical difficulties in capturing long time-lapse sequences, such as temporary occlusions, uneven frame spacing, and missing frames."
                    },
                    {
                        "title": "Production-Level Facial Performance Capture Using Deep Convolutional Neural Networks",
                        "abstract": "We present a real-time deep learning framework for video-based facial performance capture -- the dense 3D tracking of an actor's face given a monocular video. Our pipeline begins with accurately capturing a subject using a high-end production facial capture pipeline based on multi-view stereo tracking and artist-enhanced animations. With 5-10 minutes of captured footage, we train a convolutional neural network to produce high-quality output, including self-occluded regions, from a monocular video sequence of that subject. Since this 3D facial performance capture is fully automated, our system can drastically reduce the amount of labor involved in the development of modern narrative-driven video games or films involving realistic digital doubles of actors and potentially hours of animated dialogue per character. We compare our results with several state-of-the-art monocular real-time facial capture techniques and demonstrate compelling animation inference in challenging areas such as eyes and lips."
                    }
                ]
            },
            "99bf39d8-4154-4ca5-841c-4e2671e65b7d": {
                "pk": "99bf39d8-4154-4ca5-841c-4e2671e65b7d",
                "name": "Samuli Laine",
                "collaborators": [
                    "Timo Aila",
                    "Tero Karras",
                    "Jaakko Lehtinen",
                    "Miika Aittala",
                    "Janne Hellsten",
                    "Tuomas Kynk\u00e4\u00e4nniemi",
                    "Jon Hasselgren",
                    "Jacob Munkberg",
                    "Onni Kosomaa",
                    "Axel Sauer"
                ],
                "domain": [
                    "Generative Models",
                    "Semi-Supervised Learning",
                    "Image Synthesis",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Temporal Ensembling for Semi-Supervised Learning",
                        "abstract": "In this paper, we present a simple and efficient method for training deep neural networks in a semi-supervised setting where only a small portion of training data is labeled. We introduce self-ensembling, where we form a consensus prediction of the unknown labels using the outputs of the network-in-training on different epochs, and most importantly, under different regularization and input augmentation conditions. This ensemble prediction can be expected to be a better predictor for the unknown labels than the output of the network at the most recent training epoch, and can thus be used as a target for training. Using our method, we set new records for two standard semi-supervised learning benchmarks, reducing the (non-augmented) classification error rate from 18.44% to 7.05% in SVHN with 500 labels and from 18.63% to 16.55% in CIFAR-10 with 4000 labels, and further to 5.12% and 12.16% by enabling the standard augmentations. We additionally obtain a clear improvement in CIFAR-100 classification accuracy by using random images from the Tiny Images dataset as unlabeled extra inputs during training. Finally, we demonstrate good tolerance to incorrect labels."
                    },
                    {
                        "title": "Progressive Growing of GANs for Improved Quality, Stability, and Variation",
                        "abstract": "We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024^2. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10. Additionally, we describe several implementation details that are important for discouraging unhealthy competition between the generator and discriminator. Finally, we suggest a new metric for evaluating GAN results, both in terms of image quality and variation. As an additional contribution, we construct a higher-quality version of the CelebA dataset."
                    },
                    {
                        "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                        "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                    },
                    {
                        "title": "High-Quality Self-Supervised Deep Image Denoising",
                        "abstract": "We describe a novel method for training high-quality image denoising models based on unorganized collections of corrupted images. The training does not need access to clean reference images, or explicit pairs of corrupted images, and can thus be applied in situations where such data is unacceptably expensive or impossible to acquire. We build on a recent technique that removes the need for reference data by employing networks with a \"blind spot\" in the receptive field, and significantly improve two key aspects: image quality and training efficiency. Our result quality is on par with state-of-the-art neural network denoisers in the case of i.i.d. additive Gaussian noise, and not far behind with Poisson and impulse noise. We also successfully handle cases where parameters of the noise model are variable and/or unknown in both training and evaluation data."
                    },
                    {
                        "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                        "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                    },
                    {
                        "title": "Improved Precision and Recall Metric for Assessing Generative Models",
                        "abstract": "The ability to automatically estimate the quality and coverage of the samples produced by a generative model is a vital requirement for driving algorithm research. We present an evaluation metric that can separately and reliably measure both of these aspects in image generation tasks by forming explicit, non-parametric representations of the manifolds of real and generated data. We demonstrate the effectiveness of our metric in StyleGAN and BigGAN by providing several illustrative examples where existing metrics yield uninformative or contradictory results. Furthermore, we analyze multiple design variants of StyleGAN to better understand the relationships between the model architecture, training methods, and the properties of the resulting sample distribution. In the process, we identify new variants that improve the state-of-the-art. We also perform the first principled analysis of truncation methods and identify an improved method. Finally, we extend our metric to estimate the perceptual quality of individual samples, and use this to study latent space interpolations."
                    },
                    {
                        "title": "Simulator-Based Self-Supervision for Learned 3D Tomography Reconstruction",
                        "abstract": "We propose a deep learning method for 3D volumetric reconstruction in low-dose helical cone-beam computed tomography. Prior machine learning approaches require reference reconstructions computed by another algorithm for training. In contrast, we train our model in a fully self-supervised manner using only noisy 2D X-ray data. This is enabled by incorporating a fast differentiable CT simulator in the training loop. As we do not rely on reference reconstructions, the fidelity of our results is not limited by their potential shortcomings. We evaluate our method on real helical cone-beam projections and simulated phantoms. Our results show significantly higher visual fidelity and better PSNR over techniques that rely on existing reconstructions. When applied to full-dose data, our method produces high-quality results orders of magnitude faster than iterative techniques."
                    },
                    {
                        "title": "StyleGAN-T: Unlocking the Power of GANs for Fast Large-Scale Text-to-Image Synthesis",
                        "abstract": "Text-to-image synthesis has recently seen significant progress thanks to large pretrained language models, large-scale training data, and the introduction of scalable model families such as diffusion and autoregressive models. However, the best-performing models require iterative evaluation to generate a single sample. In contrast, generative adversarial networks (GANs) only need a single forward pass. They are thus much faster, but they currently remain far behind the state-of-the-art in large-scale text-to-image synthesis. This paper aims to identify the necessary steps to regain competitiveness. Our proposed model, StyleGAN-T, addresses the specific requirements of large-scale text-to-image synthesis, such as large capacity, stable training on diverse datasets, strong text alignment, and controllable variation vs. text alignment tradeoff. StyleGAN-T significantly improves over previous GANs and outperforms distilled diffusion models - the previous state-of-the-art in fast text-to-image synthesis - in terms of sample quality and speed."
                    },
                    {
                        "title": "Appearance-Driven Automatic 3D Model Simplification",
                        "abstract": "We present a suite of techniques for jointly optimizing triangle meshes and shading models to match the appearance of reference scenes. This capability has a number of uses, including appearance-preserving simplification of extremely complex assets, conversion between rendering systems, and even conversion between geometric scene representations.   We follow and extend the classic analysis-by-synthesis family of techniques: enabled by a highly efficient differentiable renderer and modern nonlinear optimization algorithms, our results are driven to minimize the image-space difference to the target scene when rendered in similar viewing and lighting conditions. As the only signals driving the optimization are differences in rendered images, the approach is highly general and versatile: it easily supports many different forward rendering models such as normal mapping, spatially-varying BRDFs, displacement mapping, etc. Supervision through images only is also key to the ability to easily convert between rendering systems and scene representations.   We output triangle meshes with textured materials to ensure that the models render efficiently on modern graphics hardware and benefit from, e.g., hardware-accelerated rasterization, ray tracing, and filtered texture lookups. Our system is integrated in a small Python code base, and can be applied at high resolutions and on large models. We describe several use cases, including mesh decimation, level of detail generation, seamless mesh filtering and approximations of aggregate geometry."
                    },
                    {
                        "title": "Analyzing and Improving the Image Quality of StyleGAN",
                        "abstract": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign the generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent codes to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably attribute a generated image to a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality."
                    },
                    {
                        "title": "Noise2Noise: Learning Image Restoration without Clean Data",
                        "abstract": "We apply basic statistical reasoning to signal reconstruction by machine learning -- learning to map corrupted observations to clean signals -- with a simple and powerful conclusion: it is possible to learn to restore images by only looking at corrupted examples, at performance at and sometimes exceeding training using clean data, without explicit image priors or likelihood models of the corruption. In practice, we show that a single model learns photographic noise removal, denoising synthetic Monte Carlo images, and reconstruction of undersampled MRI scans -- all corrupted by different processes -- based on noisy data only."
                    },
                    {
                        "title": "Training Generative Adversarial Networks with Limited Data",
                        "abstract": "Training generative adversarial networks (GAN) using too little data typically leads to discriminator overfitting, causing training to diverge. We propose an adaptive discriminator augmentation mechanism that significantly stabilizes training in limited data regimes. The approach does not require changes to loss functions or network architectures, and is applicable both when training from scratch and when fine-tuning an existing GAN on another dataset. We demonstrate, on several datasets, that good results are now possible using only a few thousand training images, often matching StyleGAN2 results with an order of magnitude fewer images. We expect this to open up new application domains for GANs. We also find that the widely used CIFAR-10 is, in fact, a limited data benchmark, and improve the record FID from 5.59 to 2.42."
                    },
                    {
                        "title": "Modular Primitives for High-Performance Differentiable Rendering",
                        "abstract": "We present a modular differentiable renderer design that yields performance superior to previous methods by leveraging existing, highly optimized hardware graphics pipelines. Our design supports all crucial operations in a modern graphics pipeline: rasterizing large numbers of triangles, attribute interpolation, filtered texture lookups, as well as user-programmable shading and geometry processing, all in high resolutions. Our modular primitives allow custom, high-performance graphics pipelines to be built directly within automatic differentiation frameworks such as PyTorch or TensorFlow. As a motivating application, we formulate facial performance capture as an inverse rendering problem and show that it can be solved efficiently using our tools. Our results indicate that this simple and straightforward approach achieves excellent geometric correspondence between rendered results and reference imagery."
                    },
                    {
                        "title": "Alias-Free Generative Adversarial Networks",
                        "abstract": "We observe that despite their hierarchical convolutional nature, the synthesis process of typical generative adversarial networks depends on absolute pixel coordinates in an unhealthy manner. This manifests itself as, e.g., detail appearing to be glued to image coordinates instead of the surfaces of depicted objects. We trace the root cause to careless signal processing that causes aliasing in the generator network. Interpreting all signals in the network as continuous, we derive generally applicable, small architectural changes that guarantee that unwanted information cannot leak into the hierarchical synthesis process. The resulting networks match the FID of StyleGAN2 but differ dramatically in their internal representations, and they are fully equivariant to translation and rotation even at subpixel scales. Our results pave the way for generative models better suited for video and animation."
                    },
                    {
                        "title": "Analyzing and Improving the Training Dynamics of Diffusion Models",
                        "abstract": "Diffusion models currently dominate the field of data-driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high-level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on expectation. We find that systematic application of this philosophy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational complexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling.   As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance."
                    },
                    {
                        "title": "Semi-supervised semantic segmentation needs strong, varied perturbations",
                        "abstract": "Consistency regularization describes a class of approaches that have yielded ground breaking results in semi-supervised classification problems. Prior work has established the cluster assumption - under which the data distribution consists of uniform class clusters of samples separated by low density regions - as important to its success. We analyze the problem of semantic segmentation and find that its' distribution does not exhibit low density regions separating classes and offer this as an explanation for why semi-supervised segmentation is a challenging problem, with only a few reports of success. We then identify choice of augmentation as key to obtaining reliable performance without such low-density regions. We find that adapted variants of the recently proposed CutOut and CutMix augmentation techniques yield state-of-the-art semi-supervised semantic segmentation results in standard datasets. Furthermore, given its challenging nature we propose that semantic segmentation acts as an effective acid test for evaluating semi-supervised regularizers. Implementation at: https://github.com/Britefury/cutmix-semisup-seg."
                    },
                    {
                        "title": "Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models",
                        "abstract": "Guidance is a crucial technique for extracting the best performance out of image-generating diffusion models. Traditionally, a constant guidance weight has been applied throughout the sampling chain of an image. We show that guidance is clearly harmful toward the beginning of the chain (high noise levels), largely unnecessary toward the end (low noise levels), and only beneficial in the middle. We thus restrict it to a specific range of noise levels, improving both the inference speed and result quality. This limited guidance interval improves the record FID in ImageNet-512 significantly, from 1.81 to 1.40. We show that it is quantitatively and qualitatively beneficial across different sampler parameters, network architectures, and datasets, including the large-scale setting of Stable Diffusion XL. We thus suggest exposing the guidance interval as a hyperparameter in all diffusion models that use guidance."
                    }
                ]
            }
        }
    },
    "2310.06771": {
        "paper_data": {
            "title": "Correlated Noise Provably Beats Independent Noise for Differentially Private Learning",
            "url": "http://arxiv.org/abs/2310.06771v2",
            "arxiv_id": "2310.06771",
            "authors": [
                "Christopher A. Choquette-Choo",
                "Krishnamurthy Dvijotham",
                "Krishna Pillutla",
                "Arun Ganesh",
                "Thomas Steinke",
                "Abhradeep Thakurta"
            ],
            "abstract": "Differentially private learning algorithms inject noise into the learning process. While the most common private learning algorithm, DP-SGD, adds independent Gaussian noise in each iteration, recent work on matrix factorization mechanisms has shown empirically that introducing correlations in the noise can greatly improve their utility. We characterize the asymptotic learning utility for any choice of the correlation function, giving precise analytical bounds for linear regression and as the solution to a convex program for general convex functions. We show, using these bounds, how correlated noise provably improves upon vanilla DP-SGD as a function of problem parameters such as the effective dimension and condition number. Moreover, our analytical expression for the near-optimal correlation function circumvents the cubic complexity of the semi-definite program used to optimize the noise correlation matrix in previous work. We validate our theory with experiments on private deep learning. Our work matches or outperforms prior work while being efficient both in terms of compute and memory.",
            "introduction": "   1 Introduction  The broad adoption of deep learning using sensitive data has led to the increasing popularity of rigorous frameworks for privacy preservation, such as differential privacy\u00a0Dwork et\u00a0al. (2006). The workhorse of private learning, a differentially private variant of stochastic gradient descent called DP-SGD\u00a0Song et\u00a0al. (2013); Bassily et\u00a0al. (2014); Abadi et\u00a0al. (2016), clips per-example gradients to some \u21132subscript\u21132\\ell_{2}roman_\u2113 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT norm and adds independent Gaussian noise. DP-SGD has been used in a range of applications from learning with medical images\u00a0Adnan et\u00a0al. (2022) to finetuning large language models with O\u2062(100\u2062B)\ud835\udc42100\ud835\udc35O(100B)italic_O ( 100 italic_B ) parameters\u00a0He et\u00a0al. (2023).   A recent line of work instead proposes to add correlated Gaussian noise to each clipped gradient\u00a0Smith & Thakurta (2013); Kairouz et\u00a0al. (2021a); Denisov et\u00a0al. (2022); Choquette-Choo et\u00a0al. (2023b). This class of algorithms called DP-FTRL, has been used for private federated learning at industrial scale\u00a0Xu et\u00a0al. (2023). By solving an expensive semi-definite program to find the noise correlations, Choquette-Choo et\u00a0al. (2023a) demonstrated empirically that DP-FTRL is never worse and often much better than DP-SGD in its privacy-utility tradeoff across multiple modalities like images and text.   However, several questions remain open. Does DP-FTRL provably improve over DP-SGD in its expected utility? Further, can we design a more computationally efficient procedure to find the noise correlations for DP-FTRL without significantly worsening the privacy-utility tradeoff?   We answer both questions affirmatively by (1) providing a sharp theoretical characterization of the noisy training dynamics of DP-FTRL, and (2) leveraging these analytical tools to circumvent the semi-definite program required in past work.    1.1 Problem Setup and Background  Algorithm 1  The DP-FTRL/Noisy-FTRL algorithms with a noise coefficient matrix \ud835\udc69\u2208\u211dT\u00d7T\ud835\udc69superscript\u211d\ud835\udc47\ud835\udc47{\\bm{B}}\\in\\mathbb{R}^{T\\times T}bold_italic_B \u2208 blackboard_R start_POSTSUPERSCRIPT italic_T \u00d7 italic_T end_POSTSUPERSCRIPT   1:\ud835\udc69\u2208\u211dT\u00d7T\ud835\udc69superscript\u211d\ud835\udc47\ud835\udc47{\\bm{B}}\\in\\mathbb{R}^{T\\times T}bold_italic_B \u2208 blackboard_R start_POSTSUPERSCRIPT italic_T \u00d7 italic_T end_POSTSUPERSCRIPT, initial iterate \ud835\udf3d0\u2208\u211ddsubscript\ud835\udf3d0superscript\u211d\ud835\udc51\\bm{\\theta}_{0}\\in\\mathbb{R}^{d}bold_italic_\u03b8 start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT \u2208 blackboard_R start_POSTSUPERSCRIPT italic_d end_POSTSUPERSCRIPT, \u21132subscript\u21132\\ell_{2}roman_\u2113 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT clip norm G\ud835\udc3aGitalic_G, noise multiplier \u03c3\ud835\uddbd\ud835\uddc9subscript\ud835\udf0e\ud835\uddbd\ud835\uddc9\\sigma_{\\sf{dp}}italic_\u03c3 start_POSTSUBSCRIPT sansserif_dp end_POSTSUBSCRIPT, learning rate \u03b7\ud835\udf02\\etaitalic_\u03b7, dataset \ud835\udc9f\ud835\udc9f\\mathcal{D}caligraphic_D    2:for\u00a0t=0,\u2026,T\u22121\ud835\udc610\u2026\ud835\udc471t=0,\\ldots,T-1italic_t = 0 , \u2026 , italic_T - 1\u00a0do    3:\u00a0\u00a0\u00a0\u00a0Obtain the next datapoint \ud835\udc9btsubscript\ud835\udc9b\ud835\udc61{\\bm{z}}_{t}bold_italic_z start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT and compute \ud835\udc88t={\u2207f\u2062(\ud835\udf3dt;\ud835\udc9bt)+\u2207r\u2062(\ud835\udf3d)\u00a0for Noisy-FTRL,\ud835\uddbc\ud835\uddc5\ud835\uddc2\ud835\uddc9\u2062(\u2207f\u2062(\ud835\udf3dt;\ud835\udc9bt),G)+\u2207r\u2062(\ud835\udf3d)\u00a0for DP-FTRLsubscript\ud835\udc88\ud835\udc61cases\u2207\ud835\udc53subscript\ud835\udf3d\ud835\udc61subscript\ud835\udc9b\ud835\udc61\u2207\ud835\udc5f\ud835\udf3d\u00a0for Noisy-FTRL\ud835\uddbc\ud835\uddc5\ud835\uddc2\ud835\uddc9\u2207\ud835\udc53subscript\ud835\udf3d\ud835\udc61subscript\ud835\udc9b\ud835\udc61\ud835\udc3a\u2207\ud835\udc5f\ud835\udf3d\u00a0for DP-FTRL{\\bm{g}}_{t}=\\begin{cases}\\nabla f(\\bm{\\theta}_{t};{\\bm{z}}_{t})+\\nabla r(\\bm{% \\theta})&\\text{ for Noisy-FTRL},\\\\ {\\sf clip}\\left(\\nabla f(\\bm{\\theta}_{t};{\\bm{z}}_{t}),G\\right)+\\nabla r(\\bm{% \\theta})&\\text{ for DP-FTRL}\\end{cases}bold_italic_g start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT = { start_ROW start_CELL \u2207 italic_f ( bold_italic_\u03b8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ; bold_italic_z start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) + \u2207 italic_r ( bold_italic_\u03b8 ) end_CELL start_CELL for Noisy-FTRL , end_CELL end_ROW start_ROW start_CELL sansserif_clip ( \u2207 italic_f ( bold_italic_\u03b8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ; bold_italic_z start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) , italic_G ) + \u2207 italic_r ( bold_italic_\u03b8 ) end_CELL start_CELL for DP-FTRL end_CELL end_ROW    4:\u00a0\u00a0\u00a0\u00a0Sample noise \ud835\udc98t\u223c\ud835\udca9\u2062(0,\u03c3\ud835\uddbd\ud835\uddc92\u2062G2\u2062\ud835\udc70d)similar-tosubscript\ud835\udc98\ud835\udc61\ud835\udca90superscriptsubscript\ud835\udf0e\ud835\uddbd\ud835\uddc92superscript\ud835\udc3a2subscript\ud835\udc70\ud835\udc51{\\bm{w}}_{t}\\sim\\mathcal{N}(0,\\sigma_{\\sf{dp}}^{2}G^{2}{\\bm{I}}_{d})bold_italic_w start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT \u223c caligraphic_N ( 0 , italic_\u03c3 start_POSTSUBSCRIPT sansserif_dp end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_G start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT bold_italic_I start_POSTSUBSCRIPT italic_d end_POSTSUBSCRIPT ) and calculate the correlated noise \ud835\udc98~t=\u2211\u03c4=0t\ud835\udc69t,\u03c4\u2062\ud835\udc98\u03c4subscript~\ud835\udc98\ud835\udc61superscriptsubscript\ud835\udf0f0\ud835\udc61subscript\ud835\udc69\ud835\udc61\ud835\udf0fsubscript\ud835\udc98\ud835\udf0f\\widetilde{{\\bm{w}}}_{t}=\\sum_{\\tau=0}^{t}{\\bm{B}}_{t,\\tau}{\\bm{w}}_{\\tau}over~ start_ARG bold_italic_w end_ARG start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT = \u2211 start_POSTSUBSCRIPT italic_\u03c4 = 0 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_t end_POSTSUPERSCRIPT bold_italic_B start_POSTSUBSCRIPT italic_t , italic_\u03c4 end_POSTSUBSCRIPT bold_italic_w start_POSTSUBSCRIPT italic_\u03c4 end_POSTSUBSCRIPT    5:\u00a0\u00a0\u00a0\u00a0Update \ud835\udf3dt+1=\ud835\udf3dt\u2212\u03b7\u2062\ud835\udc88~tsubscript\ud835\udf3d\ud835\udc611subscript\ud835\udf3d\ud835\udc61\ud835\udf02subscript~\ud835\udc88\ud835\udc61\\bm{\\theta}_{t+1}=\\bm{\\theta}_{t}-\\eta\\widetilde{\\bm{g}}_{t}bold_italic_\u03b8 start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT = bold_italic_\u03b8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT - italic_\u03b7 over~ start_ARG bold_italic_g end_ARG start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT for the noisy gradient \ud835\udc88~t=\ud835\udc88t+\ud835\udc98~tsubscript~\ud835\udc88\ud835\udc61subscript\ud835\udc88\ud835\udc61subscript~\ud835\udc98\ud835\udc61\\widetilde{\\bm{g}}_{t}={\\bm{g}}_{t}+\\widetilde{\\bm{w}}_{t}over~ start_ARG bold_italic_g end_ARG start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT = bold_italic_g start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT + over~ start_ARG bold_italic_w end_ARG start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT    6:\ud835\udf3dTsubscript\ud835\udf3d\ud835\udc47\\bm{\\theta}_{T}bold_italic_\u03b8 start_POSTSUBSCRIPT italic_T end_POSTSUBSCRIPT      Let \ud835\udc9f={\ud835\udc9b0,\u2026,\ud835\udc9bT\u22121}\ud835\udc9fsubscript\ud835\udc9b0\u2026subscript\ud835\udc9b\ud835\udc471\\mathcal{D}=\\{{\\bm{z}}_{0},\\ldots,{\\bm{z}}_{T-1}\\}caligraphic_D = { bold_italic_z start_POSTSUBSCRIPT",
            "references": [
                {
                    "title": "MADLAD-400: A Multilingual And Document-Level Large Audited Dataset",
                    "abstract": "We introduce MADLAD-400, a manually audited, general domain 3T token monolingual dataset based on CommonCrawl, spanning 419 languages. We discuss the limitations revealed by self-auditing MADLAD-400, and the role data auditing had in the dataset creation process. We then train and release a 10.7B-parameter multilingual machine translation model on 250 billion tokens covering over 450 languages using publicly available data, and find that it is competitive with models that are significantly larger, and report the results on different domains. In addition, we train a 8B-parameter language model, and assess the results on few-shot translation. We make the baseline models available to the research community."
                },
                {
                    "title": "A Unifying Framework for Differentially Private Sums under Continual Observation",
                    "abstract": "We study the problem of maintaining a differentially private decaying sum under continual observation. We give a unifying framework and an efficient algorithm for this problem for \\emph{any sufficiently smooth} function. Our algorithm is the first differentially private algorithm that does not have a multiplicative error for polynomially-decaying weights. Our algorithm improves on all prior works on differentially private decaying sums under continual observation and recovers exactly the additive error for the special case of continual counting from Henzinger et al. (SODA 2023) as a corollary. Our algorithm is a variant of the factorization mechanism whose error depends on the $\\gamma_2$ and $\\gamma_F$ norm of the underlying matrix. We give a constructive proof for an almost exact upper bound on the $\\gamma_2$ and $\\gamma_F$ norm and an almost tight lower bound on the $\\gamma_2$ norm for a large class of lower-triangular matrices. This is the first non-trivial lower bound for lower-triangular matrices whose non-zero entries are not all the same. It includes matrices for all continual decaying sums problems, resulting in an upper bound on the additive error of any differentially private decaying sums algorithm under continual observation. We also explore some implications of our result in discrepancy theory and operator algebra. Given the importance of the $\\gamma_2$ norm in computer science and the extensive work in mathematics, we believe our result will have further applications."
                },
                {
                    "title": "(Amplified) Banded Matrix Factorization: A unified approach to private training",
                    "abstract": "Matrix factorization (MF) mechanisms for differential privacy (DP) have substantially improved the state-of-the-art in privacy-utility-computation tradeoffs for ML applications in a variety of scenarios, but in both the centralized and federated settings there remain instances where either MF cannot be easily applied, or other algorithms provide better tradeoffs (typically, as $\\epsilon$ becomes small). In this work, we show how MF can subsume prior state-of-the-art algorithms in both federated and centralized training settings, across all privacy budgets. The key technique throughout is the construction of MF mechanisms with banded matrices (lower-triangular matrices with at most $\\hat{b}$ nonzero bands including the main diagonal). For cross-device federated learning (FL), this enables multiple-participations with a relaxed device participation schema compatible with practical FL infrastructure (as demonstrated by a production deployment). In the centralized setting, we prove that banded matrices enjoy the same privacy amplification results as the ubiquitous DP-SGD algorithm, but can provide strictly better performance in most scenarios -- this lets us always at least match DP-SGD, and often outperform it."
                },
                {
                    "title": "Federated Learning of Gboard Language Models with Differential Privacy",
                    "abstract": "We train and deploy language models (LMs) with federated learning (FL) and differential privacy (DP) in Google Keyboard (Gboard). The recent DP-Follow the Regularized Leader (DP-FTRL) algorithm is applied to achieve meaningfully formal DP guarantees without requiring uniform sampling of clients. To provide favorable privacy-utility trade-offs, we introduce a new client participation criterion and discuss the implication of its configuration in large scale systems. We show how quantile-based clip estimation can be combined with DP-FTRL to adaptively choose the clip norm during training or reduce the hyperparameter tuning in preparation of training. With the help of pretraining on public data, we trained and deployed more than fifteen Gboard LMs that achieve high utility and $\\rho-$zCDP privacy guarantees with $\\rho \\in (0.3, 2)$, with one model additionally trained with secure aggregation.We summarize our experience and provide concrete suggestions on DP training for practitioners."
                },
                {
                    "title": "PaLM 2 Technical Report",
                    "abstract": "We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities and is more compute-efficient than its predecessor PaLM. PaLM 2 is a Transformer-based model trained using a mixture of objectives. Through extensive evaluations on English and multilingual language, and reasoning tasks, we demonstrate that PaLM 2 has significantly improved quality on downstream tasks across different model sizes, while simultaneously exhibiting faster and more efficient inference compared to PaLM. This improved efficiency enables broader deployment while also allowing the model to respond faster, for a more natural pace of interaction. PaLM 2 demonstrates robust reasoning capabilities exemplified by large improvements over PaLM on BIG-Bench and other reasoning tasks. PaLM 2 exhibits stable performance on a suite of responsible AI evaluations, and enables inference-time control over toxicity without additional overhead or impact on other capabilities. Overall, PaLM 2 achieves state-of-the-art performance across a diverse set of tasks and capabilities. When discussing the PaLM 2 family, it is important to distinguish between pre-trained models (of various sizes), fine-tuned variants of these models, and the user-facing products that use these models. In particular, user-facing products typically include additional pre- and post-processing steps. Additionally, the underlying models may evolve over time. Therefore, one should not expect the performance of user-facing products to exactly match the results reported in this report."
                },
                {
                    "title": "Near Optimal Private and Robust Linear Regression",
                    "abstract": "We study the canonical statistical estimation problem of linear regression from $n$ i.i.d.~examples under $(\\varepsilon,\\delta)$-differential privacy when some response variables are adversarially corrupted. We propose a variant of the popular differentially private stochastic gradient descent (DP-SGD) algorithm with two innovations: a full-batch gradient descent to improve sample complexity and a novel adaptive clipping to guarantee robustness. When there is no adversarial corruption, this algorithm improves upon the existing state-of-the-art approach and achieves a near optimal sample complexity. Under label-corruption, this is the first efficient linear regression algorithm to guarantee both $(\\varepsilon,\\delta)$-DP and robustness. Synthetic experiments confirm the superiority of our approach."
                },
                {
                    "title": "Exploring the Limits of Differentially Private Deep Learning with Group-wise Clipping",
                    "abstract": "Differentially private deep learning has recently witnessed advances in computational efficiency and privacy-utility trade-off. We explore whether further improvements along the two axes are possible and provide affirmative answers leveraging two instantiations of \\emph{group-wise clipping}. To reduce the compute time overhead of private learning, we show that \\emph{per-layer clipping}, where the gradient of each neural network layer is clipped separately, allows clipping to be performed in conjunction with backpropagation in differentially private optimization. This results in private learning that is as memory-efficient and almost as fast per training update as non-private learning for many workflows of interest. While per-layer clipping with constant thresholds tends to underperform standard flat clipping, per-layer clipping with adaptive thresholds matches or outperforms flat clipping under given training epoch constraints, hence attaining similar or better task performance within less wall time. To explore the limits of scaling (pretrained) models in differentially private deep learning, we privately fine-tune the 175 billion-parameter GPT-3. We bypass scaling challenges associated with clipping gradients that are distributed across multiple devices with \\emph{per-device clipping} that clips the gradient of each model piece separately on its host device. Privately fine-tuning GPT-3 with per-device clipping achieves a task performance at $\\epsilon=1$ better than what is attainable by non-privately fine-tuning the largest GPT-2 on a summarization task."
                },
                {
                    "title": "Multi-Epoch Matrix Factorization Mechanisms for Private Machine Learning",
                    "abstract": "We introduce new differentially private (DP) mechanisms for gradient-based machine learning (ML) with multiple passes (epochs) over a dataset, substantially improving the achievable privacy-utility-computation tradeoffs. We formalize the problem of DP mechanisms for adaptive streams with multiple participations and introduce a non-trivial extension of online matrix factorization DP mechanisms to our setting. This includes establishing the necessary theory for sensitivity calculations and efficient computation of optimal matrices. For some applications like $>\\!\\! 10,000$ SGD steps, applying these optimal techniques becomes computationally expensive. We thus design an efficient Fourier-transform-based mechanism with only a minor utility loss. Extensive empirical evaluation on both example-level DP for image classification and user-level DP for language modeling demonstrate substantial improvements over all previous methods, including the widely-used DP-SGD . Though our primary application is to ML, our main DP results are applicable to arbitrary linear queries and hence may have much broader applicability."
                },
                {
                    "title": "Constant Matters: Fine-grained Error Bound on Differentially Private Continual Observation",
                    "abstract": "We study fine-grained error bounds for differentially private algorithms for counting under continual observation. Our main insight is that the matrix mechanism when using lower-triangular matrices can be used in the continual observation model. More specifically, we give an explicit factorization for the counting matrix $M_\\mathsf{count}$ and upper bound the error explicitly. We also give a fine-grained analysis, specifying the exact constant in the upper bound. Our analysis is based on upper and lower bounds of the {\\em completely bounded norm} (cb-norm) of $M_\\mathsf{count}$. Along the way, we improve the best-known bound of 28 years by Mathias (SIAM Journal on Matrix Analysis and Applications, 1993) on the cb-norm of $M_\\mathsf{count}$ for a large range of the dimension of $M_\\mathsf{count}$. Furthermore, we are the first to give concrete error bounds for various problems under continual observation such as binary counting, maintaining a histogram, releasing an approximately cut-preserving synthetic graph, many graph-based statistics, and substring and episode counting. Finally, we note that our result can be used to get a fine-grained error bound for non-interactive local learning {and the first lower bounds on the additive error for $(\\epsilon,\\delta)$-differentially-private counting under continual observation.} Subsequent to this work, Henzinger et al. (SODA2023) showed that our factorization also achieves fine-grained mean-squared error."
                },
                {
                    "title": "Improved Differential Privacy for SGD via Optimal Private Linear Operators on Adaptive Streams",
                    "abstract": "Motivated by recent applications requiring differential privacy over adaptive streams, we investigate the question of optimal instantiations of the matrix mechanism in this setting. We prove fundamental theoretical results on the applicability of matrix factorizations to adaptive streams, and provide a parameter-free fixed-point algorithm for computing optimal factorizations. We instantiate this framework with respect to concrete matrices which arise naturally in machine learning, and train user-level differentially private models with the resulting optimal mechanisms, yielding significant improvements in a notable problem in federated learning with user-level differential privacy."
                },
                {
                    "title": "Quantifying Memorization Across Neural Language Models",
                    "abstract": "Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations."
                },
                {
                    "title": "Anticorrelated Noise Injection for Improved Generalization",
                    "abstract": "Injecting artificial noise into gradient descent (GD) is commonly employed to improve the performance of machine learning models. Usually, uncorrelated noise is used in such perturbed gradient descent (PGD) methods. It is, however, not known if this is optimal or whether other types of noise could provide better generalization performance. In this paper, we zoom in on the problem of correlating the perturbations of consecutive PGD steps. We consider a variety of objective functions for which we find that GD with anticorrelated perturbations (\"Anti-PGD\") generalizes significantly better than GD and standard (uncorrelated) PGD. To support these experimental findings, we also derive a theoretical analysis that demonstrates that Anti-PGD moves to wider minima, while GD and PGD remain stuck in suboptimal regions or even diverge. This new connection between anticorrelated noise and generalization opens the field to novel ways to exploit noise for training machine learning models."
                },
                {
                    "title": "The Price of Differential Privacy under Continual Observation",
                    "abstract": "We study the accuracy of differentially private mechanisms in the continual release model. A continual release mechanism receives a sensitive dataset as a stream of $T$ inputs and produces, after receiving each input, an accurate output on the obtained inputs. In contrast, a batch algorithm receives the data as one batch and produces a single output. We provide the first strong lower bounds on the error of continual release mechanisms. In particular, for two fundamental problems that are widely studied and used in the batch model, we show that the worst case error of every continual release algorithm is $\\tilde \\Omega(T^{1/3})$ times larger than that of the best batch algorithm. Previous work shows only a polylogarithimic (in $T$) gap between the worst case error achievable in these two models; further, for many problems, including the summation of binary attributes, the polylogarithmic gap is tight (Dwork et al., 2010; Chan et al., 2010). Our results show that problems closely related to summation -- specifically, those that require selecting the largest of a set of sums -- are fundamentally harder in the continual release model than in the batch model. Our lower bounds assume only that privacy holds for streams fixed in advance (the\"nonadaptive\"setting). However, we provide matching upper bounds that hold in a model where privacy is required even for adaptively selected streams. This model may be of independent interest."
                },
                {
                    "title": "On The Absolute Constant in Hanson-Wright Inequality",
                    "abstract": "We revisit and slightly modify the proof of the Gaussian Hanson-Wright inequality where we keep track of the absolute constant in its formulation."
                },
                {
                    "title": "Practical and Private (Deep) Learning without Sampling or Shuffling",
                    "abstract": "We consider training models with differential privacy (DP) using mini-batch gradients. The existing state-of-the-art, Differentially Private Stochastic Gradient Descent (DP-SGD), requires privacy amplification by sampling or shuffling to obtain the best privacy/accuracy/computation trade-offs. Unfortunately, the precise requirements on exact sampling and shuffling can be hard to obtain in important practical scenarios, particularly federated learning (FL). We design and analyze a DP variant of Follow-The-Regularized-Leader (DP-FTRL) that compares favorably (both theoretically and empirically) to amplified DP-SGD, while allowing for much more flexible data access patterns. DP-FTRL does not use any form of privacy amplification. The code is available at https://github.com/google-research/federated/tree/master/dp_ftrl and https://github.com/google-research/DP-FTRL ."
                },
                {
                    "title": "Extracting Training Data from Large Language Models",
                    "abstract": "It has become common to publish large (billion parameter) language models that have been trained on private datasets. This paper demonstrates that in such settings, an adversary can perform a training data extraction attack to recover individual training examples by querying the language model. \nWe demonstrate our attack on GPT-2, a language model trained on scrapes of the public Internet, and are able to extract hundreds of verbatim text sequences from the model's training data. These extracted examples include (public) personally identifiable information (names, phone numbers, and email addresses), IRC conversations, code, and 128-bit UUIDs. Our attack is possible even though each of the above sequences are included in just one document in the training data. \nWe comprehensively evaluate our extraction attack to understand the factors that contribute to its success. For example, we find that larger models are more vulnerable than smaller models. We conclude by drawing lessons and discussing possible safeguards for training large language models."
                },
                {
                    "title": "Adaptive Federated Optimization",
                    "abstract": "Federated learning is a distributed machine learning paradigm in which a large number of clients coordinate with a central server to learn a model without sharing their own training data. Due to the heterogeneity of the client datasets, standard federated optimization methods such as Federated Averaging (FedAvg) are often difficult to tune and exhibit unfavorable convergence behavior. In non-federated settings, adaptive optimization methods have had notable success in combating such issues. In this work, we propose federated versions of adaptive optimizers, including Adagrad, Adam, and Yogi, and analyze their convergence in the presence of heterogeneous data for general nonconvex settings. Our results highlight the interplay between client heterogeneity and communication efficiency. We also perform extensive experiments on these methods and show that the use of adaptive optimizers can significantly improve the performance of federated learning."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Hypothesis Testing Interpretations and Renyi Differential Privacy",
                    "abstract": "Differential privacy is a de facto standard in data privacy, with applications in the public and private sectors. A way to explain differential privacy, which is particularly appealing to statistician and social scientists is by means of its statistical hypothesis testing interpretation. Informally, one cannot effectively test whether a specific individual has contributed her data by observing the output of a private mechanism---any test cannot have both high significance and high power. \nIn this paper, we identify some conditions under which a privacy definition given in terms of a statistical divergence satisfies a similar interpretation. These conditions are useful to analyze the distinguishability power of divergences and we use them to study the hypothesis testing interpretation of some relaxations of differential privacy based on Renyi divergence. This analysis also results in an improved conversion rule between these definitions and differential privacy."
                },
                {
                    "title": "On strict sub-Gaussianity, optimal proxy variance and symmetry for bounded random variables",
                    "abstract": "We investigate the sub-Gaussian property for almost surely bounded random variables. If sub-Gaussianity per se is de facto ensured by the bounded support of said random variables, then exciting research avenues remain open. Among these questions is how to characterize the optimal sub-Gaussian proxy variance? Another question is how to characterize strict sub-Gaussianity, defined by a proxy variance equal to the (standard) variance? We address the questions in proposing conditions based on the study of functions variations. A particular focus is given to the relationship between strict sub-Gaussianity and symmetry of the distribution. In particular, we demonstrate that symmetry is neither sufficient nor necessary for strict sub-Gaussianity. In contrast, simple necessary conditions on the one hand, and simple sufficient conditions on the other hand, for strict sub-Gaussianity are provided. These results are illustrated via various applications to a number of bounded random variables, including Bernoulli, beta, binomial, Kumaraswamy, triangular, and uniform distributions."
                },
                {
                    "title": "The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks",
                    "abstract": "This paper describes a testing methodology for quantitatively assessing the risk that rare or unique training-data sequences are unintentionally memorized by generative sequence models---a common type of machine-learning model. Because such models are sometimes trained on sensitive data (e.g., the text of users' private messages), this methodology can benefit privacy by allowing deep-learning practitioners to select means of training that minimize such memorization. \nIn experiments, we show that unintended memorization is a persistent, hard-to-avoid issue that can have serious consequences. Specifically, for models trained without consideration of memorization, we describe new, efficient procedures that can extract unique, secret sequences, such as credit card numbers. We show that our testing strategy is a practical and easy-to-use first line of defense, e.g., by describing its application to quantitatively limit data exposure in Google's Smart Compose, a commercial text-completion neural network trained on millions of users' email messages."
                },
                {
                    "title": "A Markov Chain Theory Approach to Characterizing the Minimax Optimality of Stochastic Gradient Descent (for Least Squares)",
                    "abstract": "This work provides a simplified proof of the statistical minimax optimality of (iterate averaged) stochastic gradient descent (SGD), for the special case of least squares. This result is obtained by analyzing SGD as a stochastic process and by sharply characterizing the stationary covariance matrix of this process. The finite rate optimality characterization captures the constant factors and addresses model mis-specification."
                },
                {
                    "title": "Learning Differentially Private Recurrent Language Models",
                    "abstract": "We demonstrate that it is possible to train large recurrent language models with user-level differential privacy guarantees with only a negligible cost in predictive accuracy. Our work builds on recent advances in the training of deep networks on user-partitioned data and privacy accounting for stochastic gradient descent. In particular, we add user-level privacy protection to the federated averaging algorithm, which makes \"large step\" updates from user-level data. Our work demonstrates that given a dataset with a sufficiently large number of users (a requirement easily met by even small internet-scale datasets), achieving differential privacy comes at the cost of increased computation, rather than in decreased utility as in most prior work. We find that our private LSTM language models are quantitatively and qualitatively similar to un-noised models when trained on a large dataset."
                },
                {
                    "title": "Parallelizing Stochastic Gradient Descent for Least Squares Regression: Mini-batching, Averaging, and Model Misspecification",
                    "abstract": "This work characterizes the benefits of averaging schemes widely used in conjunction with stochastic gradient descent (SGD). In particular, this work provides a sharp analysis of: (1) mini-batching, a method of averaging many samples of a stochastic gradient to both reduce the variance of the stochastic gradient estimate and for parallelizing SGD and (2) tail-averaging, a method involving averaging the final few iterates of SGD to decrease the variance in SGD's final iterate. This work presents non-asymptotic excess risk bounds for these schemes for the stochastic approximation problem of least squares regression. \nFurthermore, this work establishes a precise problem-dependent extent to which mini-batch SGD yields provable near-linear parallelization speedups over SGD with batch size one. This allows for understanding learning rate versus batch size tradeoffs for the final iterate of an SGD method. These results are then utilized in providing a highly parallelizable SGD method that obtains the minimax risk with nearly the same number of serial updates as batch gradient descent, improving significantly over existing SGD methods. A non-asymptotic analysis of communication efficient parallelization schemes such as model-averaging/parameter mixing methods is then provided. \nFinally, this work sheds light on some fundamental differences in SGD's behavior when dealing with agnostic noise in the (non-realizable) least squares regression problem. In particular, the work shows that the stepsizes that ensure minimax risk for the agnostic case must be a function of the noise properties. \nThis paper builds on the operator view of analyzing SGD methods, introduced by Defossez and Bach (2015), followed by developing a novel analysis in bounding these operators to characterize the excess risk. These techniques are of broader interest in analyzing computational aspects of stochastic approximation."
                },
                {
                    "title": "Deep Learning with Differential Privacy",
                    "abstract": "Machine learning techniques based on neural networks are achieving remarkable results in a wide variety of domains. Often, the training of models requires large, representative datasets, which may be crowdsourced and contain sensitive information. The models should not expose private information in these datasets. Addressing this goal, we develop new algorithmic techniques for learning and a refined analysis of privacy costs within the framework of differential privacy. Our implementation and experiments demonstrate that we can train deep neural networks with non-convex objectives, under a modest privacy budget, and at a manageable cost in software complexity, training efficiency, and model quality."
                },
                {
                    "title": "On some properties of Toeplitz matrices",
                    "abstract": "In this paper, we investigate some properties of Toeplitz matrices with respect to different matrix products. We also give some results regarding circulant matrices, skew-circulant matrices and approximation by Toeplitz matrices over the field of complex numbers."
                },
                {
                    "title": "Averaged Least-Mean-Squares: Bias-Variance Trade-offs and Optimal Sampling Distributions",
                    "abstract": "We consider the least-squares regression problem and provide a detailed asymptotic analysis of the performance of averaged constant-step-size stochastic gradient descent. In the strongly-convex case, we provide an asymptotic expansion up to explicit exponentially decaying terms. Our analysis leads to new insights into stochastic approximation algorithms: (a) it gives a tighter bound on the allowed step-size; (b) the generalization error may be divided into a variance term which is decaying as O(1/n), independently of the step-size , and a bias term that decays as O(1/ 2 n 2 ); (c) when allowing non-uniform sampling of examples over a dataset, the choice of a good sampling density depends on the trade-off between bias and variance: when the variance term dominates, optimal sampling densities do not lead to much gain, while when the bias term dominates, we can choose larger step-sizes that lead to significant improvements."
                },
                {
                    "title": "Differentially Private Empirical Risk Minimization: Efficient Algorithms and Tight Error Bounds",
                    "abstract": "In this paper, we initiate a systematic investigation of differentially private algorithms for convex empirical risk minimization. Various instantiations of this problem have been studied before. We provide new algorithms and matching lower bounds for private ERM assuming only that each data point's contribution to the loss function is Lipschitz bounded and that the domain of optimization is bounded. We provide a separate set of algorithms and matching lower bounds for the setting in which the loss functions are known to also be strongly convex. \nOur algorithms run in polynomial time, and in some cases even match the optimal non-private running time (as measured by oracle complexity). We give separate algorithms (and lower bounds) for $(\\epsilon,0)$- and $(\\epsilon,\\delta)$-differential privacy; perhaps surprisingly, the techniques used for designing optimal algorithms in the two cases are completely different. \nOur lower bounds apply even to very simple, smooth function families, such as linear and quadratic functions. This implies that algorithms from previous work can be used to obtain optimal error rates, under the additional assumption that the contributions of each data point to the loss function is smooth. We show that simple approaches to smoothing arbitrary loss functions (in order to apply previous techniques) do not yield optimal error rates. In particular, optimal algorithms were not previously known for problems such as training support vector machines and the high-dimensional median."
                },
                {
                    "title": "(Nearly) Optimal Algorithms for Private Online Learning in Full-information and Bandit Settings",
                    "abstract": "We give differentially private algorithms for a large class of online learning algorithms, in both the full information and bandit settings. Our algorithms aim to minimize a convex loss function which is a sum of smaller convex loss terms, one for each data point. To design our algorithms, we modify the popular mirror descent approach, or rather a variant called follow the approximate leader. \n \nThe technique leads to the first nonprivate algorithms for private online learning in the bandit setting. In the full information setting, our algorithms improve over the regret bounds of previous work (due to Dwork, Naor, Pitassi and Rothblum (2010) and Jain, Kothari and Thakurta (2012)). In many cases, our algorithms (in both settings) match the dependence on the input length, T, of the optimal nonprivate regret bounds up to logarithmic factors in T. Our algorithms require logarithmic space and update time."
                },
                {
                    "title": "Stochastic gradient descent with differentially private updates",
                    "abstract": "Differential privacy is a recent framework for computation on sensitive data, which has shown considerable promise in the regime of large datasets. Stochastic gradient methods are a popular approach for learning in the data-rich regime because they are computationally tractable and scalable. In this paper, we derive differentially private versions of stochastic gradient descent, and test them empirically. Our results show that standard SGD experiences high variability due to differential privacy, but a moderate increase in the batch size can improve performance significantly."
                },
                {
                    "title": "Hanson-Wright inequality and sub-gaussian concentration",
                    "abstract": "In this expository note, we give a modern proof of Hanson-Wright inequality for quadratic forms in sub-gaussian random variables.We deduce a useful concentration inequality for sub-gaussian random vectors.Two examples are given to illustrate these results: a concentration of distances between random vectors and subspaces, and a bound on the norms of products of random and deterministic matrices."
                },
                {
                    "title": "Non-strongly-convex smooth stochastic approximation with convergence rate O(1/n)",
                    "abstract": "We consider the stochastic approximation problem where a convex function has to be minimized, given only the knowledge of unbiased estimates of its gradients at certain points, a framework which includes machine learning methods based on the minimization of the empirical risk. We focus on problems without strong convexity, for which all previously known algorithms achieve a convergence rate for function values of O(1/\u221an) after n iterations. We consider and analyze two algorithms that achieve a rate of O(1/n) for classical supervised learning problems. For least-squares regression, we show that averaged stochastic gradient descent with constant step-size achieves the desired rate. For logistic regression, this is achieved by a simple novel stochastic gradient algorithm that (a) constructs successive local quadratic approximations of the loss functions, while (b) preserving the same running-time complexity as stochastic gradient descent. For these algorithms, we provide a non-asymptotic analysis of the generalization error (in expectation, and also in high probability for least-squares), and run extensive experiments showing that they often outperform existing approaches."
                },
                {
                    "title": "Zames-Falb Multipliers for Quadratic Programming",
                    "abstract": "In constrained linear model predictive control, a quadratic program must be solved on-line at each control step, and this constitutes a nonlinearity. If zero is a feasible point for this quadratic program then the resultant nonlinearity is sector bounded. We show that if the nonlinearity is static then it is also monotone and slope restricted; hence, we show the existence of Zames-Falb multipliers for such a nonlinearity. The multipliers may be used in a general and versatile analysis of the robust stability of input constrained model predictive control."
                },
                {
                    "title": "An Introduction to Statistical Signal Processing",
                    "abstract": "The book presents the basics of modern Fortran in a remarkably clear style. The exposition is free of computer jargon. The authors define the terms that they use and present excellent examples of codes for statistical applications. I am very impressed with the exposition. Most technical people like to use a higher-level language than Fortran 95, such as S or Matlab, claiming that such languages are faster than Fortran. Yet I have seen my colleagues working for hours trying to debug their \u201ceasy\u201d Matlab programs, while I can more quickly modify Fortran code that I either have written or have found in IMSL or open sources for Fortran legacy routines to do what they are trying to do. Several years ago, one of my friends in Stockholm was finishing a paper for a conference that used a bispectral analysis of some signals. I asked my friend why did not use my Fortran code. He said that it was easier for him to write his own in Matlab. I asked him to give me his data and his parameter settings so that I could do a bispectral analysis of his signal using my code. He did so, and when I showed him my results, he noticed that the estimates that I got were different from those that he got from his Matlab code. He told me that I had a bug in my code, and I replied that my code was carefully developed over a 10-year period, and that I had checked the output with that generated by an earlier bispectrum program written by a friend. After I returned to Austin, he sent me a note telling me that he found a comma out of place in his program and that my results were correct. It is important for data analysts to control the computations that they wish to run. All commercial statistical packages have some errors that may take a while to fix. I test statistical routines using simulations that I write in Fortran. Simulations are not only useful for testing the accuracy of a routine but they can show where a method breaks down. This book can help statisticians write simple as well as complex Fortran routines. It also presents links to commonly used commercial programs, such as Windows and Excel."
                },
                {
                    "title": "Calibrating Noise to Sensitivity in Private Data Analysis",
                    "abstract": "We continue a line of research initiated in [10, 11] on privacy-preserving statistical databases. Consider a trusted server that holds a database of sensitive information. Given a query function f mapping databases to reals, the so-called true answer is the result of applying f to the database. To protect privacy, the true answer is perturbed by the addition of random noise generated according to a carefully chosen distribution, and this response, the true answer plus noise, is returned to the user. Previous work focused on the case of noisy sums, in which f = \u03a3 i g(x i ), where x i denotes the ith row of the database and g maps database rows to [0,1]. We extend the study to general functions f, proving that privacy can be preserved by calibrating the standard deviation of the noise according to the sensitivity of the function f. Roughly speaking, this is the amount that any single argument to f can change its output. The new analysis shows that for several particular applications substantially less noise is needed than was previously understood to be the case. The first step is a very clean characterization of privacy in terms of indistinguishability of transcripts. Additionally, we obtain separation results showing the increased value of interactive sanitization mechanisms over non-interactive."
                },
                {
                    "title": "On a perturbation approach for the analysis of stochastic tracking algorithms",
                    "abstract": "In this paper, a perturbation expansion technique is introduced to decompose the tracking error of a general adaptive tracking algorithm in a linear regression model. This method allows to obtain the tracking error bound and also tight approximate expressions for the moments of the tracking error. These expressions allow to evaluate, both qualitatively and quantitatively, the impact of several factors on the tracking error performance which have been overlooked in previous contributions."
                },
                {
                    "title": "Convergence of Gradient Descent with Linearly Correlated Noise and Applications to Differentially Private Learning",
                    "abstract": "We study stochastic optimization with linearly correlated noise. Our study is motivated by recent methods for optimization with differential privacy (DP), such as DP-FTRL, which inject noise via matrix factorization mechanisms. We propose an optimization problem that distils key facets of these DP methods and that involves perturbing gradients by linearly correlated noise. We derive improved convergence rates for gradient descent in this framework for convex and non-convex loss functions. Our theoretical analysis is novel and might be of independent interest. We use these convergence rates to develop new, effective matrix factorizations for differentially private optimization, and highlight the bene\ufb01ts of these factorizations theoretically and empirically."
                },
                {
                    "title": "Preventing Verbatim Memorization in Language Models Gives a False Sense of Privacy",
                    "abstract": "Studying data memorization in neural language models helps us understand the risks (e.g., to privacy or copyright) associated with models regurgitating training data, and aids in the evaluation of potential countermeasures. Many prior works\u2014and some recently deployed defenses\u2014focus on \u201cverbatim memo-rization\u201d, de\ufb01ned as a model generation that exactly matches a substring from the training set. We argue that verbatim memorization def-initions are too restrictive and fail to capture more subtle forms of memorization. Speci\ufb01cally, we design and implement an ef\ufb01cient defense based on Bloom \ufb01lters that perfectly prevents all verbatim memorization. And yet, we demonstrate that this \u201cperfect\u201d \ufb01lter does not prevent the leakage of training data. Indeed, it is easily circumvented by plausible and mini-mally modi\ufb01ed \u201cstyle-transfer\u201d prompts\u2014and in some cases even the non-modi\ufb01ed original prompts\u2014to extract memorized information. For example, instructing the model to output ALL-CAPITAL texts bypasses memorization checks based on verbatim matching. We conclude by discussing potential alternative de\ufb01nitions and why de\ufb01ning memorization is a dif\ufb01cult yet crucial open question for neural language models."
                },
                {
                    "title": "Efficient Use of Differentially Private Binary Trees",
                    "abstract": "Binary trees can be made differentially private by adding noise to every node and leaf. In such form they allow multifaceted exploration of a variable without revealing any individual information. While a differentially private binary tree can be used and read just like its conventional exact-valued analog, realizing that different combinations of nodes contain overlapping answers to the same information allows us to bring the statistical properties of multiple measurements under measurement error to noisy binary trees to create statistically efficient node estimates. We construct estimators that correctly use all available information in the tree, thus decreasing the error of nodes by up to eighty percent for the same level of privacy protection. Differentially private binary trees are important summary statistics for a broad variety of uses and algorithms. They are central to the algorithm of Dwork et al (2010) for releasing private streaming data, and used in numerous adaptations of this problem, such as Chan et al (2012), Cao et al (2013), and Thakurta and Smith (2013). A binary tree can be used to compose probability and cumulative densities of variables, range queries, as well as means, medians, modes and variances by Monte Carlo integration. Thus the release of a private binary tree can be a broadly useful privacy preserving means of allowing exploration of a variable, as for example in the statistical release of a non-interactive curator (Dwork and Smith, 2009). Due to the importance of binary trees, several heuristic adaptations of their use have been studied that bring about some improved accuracy for the same level of privacy guarantee (Hay et al 2010, Xu et al 2012, and relatedly Xiao et al 2010). We show here how to derive an optimally efficient use of a private binary tree, in the precise sense of providing minimum variance unbiased estimates. That is, we show how to refine a private tree in a manner that makes full and optimal use of all the information the tree contains. This is accomplished by linking the tree structure to the known statistical properties of multiple measures under measurement error. 1 Problem Statements Consider a perfect binary tree, in which every node, ti, is the sum of all leaves below that node, plus a random draw, i from some fixed distribution f(.), constructed to guarantee differential privacy. As represented below, the true values are denoted a through h at the leaves, the differentially private value revealed at any node is given to the right, and the notation for index i both numbers the nodes sequentially, and also describes the path from the top to reach that particular node, as a sequence of left (0) and right (1) progressions."
                },
                {
                    "title": "Generalization and Parameter Estimation in Feedforward Netws: Some Experiments",
                    "abstract": "We have done an empirical study of the relation of the number of parameters (weights) in a feedforward net to generalization performance. Two experiments are reported. In one, we use simulated data sets with well-controlled parameters, such as the signal-to-noise ratio of continuous-valued data. In the second, we train the network on vector-quantized mel cepstra from real speech samples. In each case, we use back-propagation to train the feedforward net to discriminate in a multiple class pattern classification problem. We report the results of these studies, and show the application of cross-validation techniques to prevent overfitting."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CR",
                "math.OC"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nDoes the DP-FTRL algorithm provably improve over DP-SGD in its expected utility, and can we design a more computationally efficient procedure to find the noise correlations for DP-FTRL without significantly worsening the privacy-utility tradeoff?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of privacy-preserving machine learning, particularly in applications involving sensitive data. By demonstrating that DP-FTRL can outperform DP-SGD, we can enhance the privacy-utility tradeoff, leading to more effective algorithms for federated learning and other applications. This research could pave the way for future studies to explore more efficient privacy-preserving techniques, ultimately contributing to the development of robust machine learning systems that respect user privacy while maintaining high performance.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem stem from the need to theoretically characterize the noisy training dynamics of DP-FTRL, which involves complex mathematical formulations. Naive approaches may fail because they do not account for the intricate relationships between noise correlations and the resulting utility of the model. Additionally, the computational burden of solving semi-definite programs to find these correlations can be prohibitive, making it difficult to implement DP-FTRL in practice without sacrificing efficiency or effectiveness.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either DP-SGD or the empirical performance of DP-FTRL without providing a rigorous theoretical foundation for its advantages. The lack of a clear understanding of the noise dynamics and the computational challenges associated with finding optimal noise correlations have hindered progress. Our approach differs by providing a sharp theoretical characterization of DP-FTRL's dynamics and proposing a more efficient method to determine noise correlations, thus addressing the limitations of prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a detailed theoretical analysis of the noisy training dynamics of DP-FTRL, followed by the development of a computationally efficient procedure to find noise correlations. We will utilize a dataset relevant to the application of interest (e.g., medical images or text data) and evaluate the performance using metrics such as utility loss and privacy guarantees. The expected outcomes include a provable improvement in the utility of DP-FTRL over DP-SGD and a practical algorithm that maintains a favorable privacy-utility tradeoff while being computationally efficient."
            }
        },
        "author_data": {
            "e92c4a23-7442-4447-b6ab-8d92fdbf2884": {
                "pk": "e92c4a23-7442-4447-b6ab-8d92fdbf2884",
                "name": "Christopher A. Choquette-Choo",
                "collaborators": [
                    "Katherine Lee",
                    "Matthew Jagielski",
                    "Milad Nasr",
                    "Rohan Anil",
                    "Siamak Shakeri",
                    "Aakanksha Chowdhery",
                    "Ambrose Slone",
                    "Jacob Austin",
                    "Vladimir Feinberg",
                    "Nicholas Carlini"
                ],
                "domain": [
                    "Large Language Models",
                    "Differential Privacy",
                    "Adversarial Machine Learning",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context",
                        "abstract": "In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content."
                    },
                    {
                        "title": "Auditing Private Prediction",
                        "abstract": "Differential privacy (DP) offers a theoretical upper bound on the potential privacy leakage of analgorithm, while empirical auditing establishes a practical lower bound. Auditing techniques exist forDP training algorithms. However machine learning can also be made private at inference. We propose thefirst framework for auditing private prediction where we instantiate adversaries with varying poisoningand query capabilities. This enables us to study the privacy leakage of four private prediction algorithms:PATE [Papernot et al., 2016], CaPC [Choquette-Choo et al., 2020], PromptPATE [Duan et al., 2023],and Private-kNN [Zhu et al., 2020]. To conduct our audit, we introduce novel techniques to empiricallyevaluate privacy leakage in terms of Renyi DP. Our experiments show that (i) the privacy analysis ofprivate prediction can be improved, (ii) algorithms which are easier to poison lead to much higher privacyleakage, and (iii) the privacy leakage is significantly lower for adversaries without query control than thosewith full control."
                    },
                    {
                        "title": "Assessing biomedical knowledge robustness in large language models by query-efficient sampling attacks",
                        "abstract": "The increasing depth of parametric domain knowledge in large language models (LLMs) is fueling their rapid deployment in real-world applications. Understanding model vulnerabilities in high-stakes and knowledge-intensive tasks is essential for quantifying the trustworthiness of model predictions and regulating their use. The recent discovery of named entities as adversarial examples (i.e. adversarial entities) in natural language processing tasks raises questions about their potential impact on the knowledge robustness of pre-trained and finetuned LLMs in high-stakes and specialized domains. We examined the use of type-consistent entity substitution as a template for collecting adversarial entities for billion-parameter LLMs with biomedical knowledge. To this end, we developed an embedding-space attack based on powerscaled distance-weighted sampling to assess the robustness of their biomedical knowledge with a low query budget and controllable coverage. Our method has favorable query efficiency and scaling over alternative approaches based on random sampling and blackbox gradient-guided search, which we demonstrated for adversarial distractor generation in biomedical question answering. Subsequent failure mode analysis uncovered two regimes of adversarial entities on the attack surface with distinct characteristics and we showed that entity substitution attacks can manipulate token-wise Shapley value explanations, which become deceptive in this setting. Our approach complements standard evaluations for high-capacity models and the results highlight the brittleness of domain knowledge in LLMs."
                    },
                    {
                        "title": "Phantom: General Trigger Attacks on Retrieval Augmented Language Generation",
                        "abstract": "Retrieval Augmented Generation (RAG) expands the capabilities of modern large language models (LLMs), by anchoring, adapting, and personalizing their responses to the most relevant knowledge sources. It is particularly useful in chatbot applications, allowing developers to customize LLM output without expensive retraining. Despite their significant utility in various applications, RAG systems present new security risks. In this work, we propose new attack vectors that allow an adversary to inject a single malicious document into a RAG system's knowledge base, and mount a backdoor poisoning attack. We design Phantom, a general two-stage optimization framework against RAG systems, that crafts a malicious poisoned document leading to an integrity violation in the model's output. First, the document is constructed to be retrieved only when a specific trigger sequence of tokens appears in the victim's queries. Second, the document is further optimized with crafted adversarial text that induces various adversarial objectives on the LLM output, including refusal to answer, reputation damage, privacy violations, and harmful behaviors. We demonstrate our attacks on multiple LLM architectures, including Gemma, Vicuna, and Llama, and show that they transfer to GPT-3.5 Turbo and GPT-4. Finally, we successfully conducted a Phantom attack on NVIDIA's black-box production RAG system,\"Chat with RTX\"."
                    },
                    {
                        "title": "Recite, Reconstruct, Recollect: Memorization in LMs as a Multifaceted Phenomenon",
                        "abstract": "Memorization in language models is typically treated as a homogenous phenomenon, neglecting the specifics of the memorized data. We instead model memorization as the effect of a set of complex factors that describe each sample and relate it to the model and corpus. To build intuition around these factors, we break memorization down into a taxonomy: recitation of highly duplicated sequences, reconstruction of inherently predictable sequences, and recollection of sequences that are neither. We demonstrate the usefulness of our taxonomy by using it to construct a predictive model for memorization. By analyzing dependencies and inspecting the weights of the predictive model, we find that different factors influence the likelihood of memorization differently depending on the taxonomic category."
                    },
                    {
                        "title": "Gemma: Open Models Based on Gemini Research and Technology",
                        "abstract": "This work introduces Gemma, a family of lightweight, state-of-the art open models built from the research and technology used to create Gemini models. Gemma models demonstrate strong performance across academic benchmarks for language understanding, reasoning, and safety. We release two sizes of models (2 billion and 7 billion parameters), and provide both pretrained and fine-tuned checkpoints. Gemma outperforms similarly sized open models on 11 out of 18 text-based tasks, and we present comprehensive evaluations of safety and responsibility aspects of the models, alongside a detailed description of model development. We believe the responsible release of LLMs is critical for improving the safety of frontier models, and for enabling the next wave of LLM innovations."
                    },
                    {
                        "title": "Optimal Rates for $O(1)$-Smooth DP-SCO with a Single Epoch and Large Batches",
                        "abstract": "In this paper we revisit the DP stochastic convex optimization (SCO) problem. For convex smooth losses, it is well-known that the canonical DP-SGD (stochastic gradient descent) achieves the optimal rate of $O\\left(\\frac{LR}{\\sqrt{n}} + \\frac{LR \\sqrt{p \\log(1/\\delta)}}{\\epsilon n}\\right)$ under $(\\epsilon, \\delta)$-DP, and also well-known that variants of DP-SGD can achieve the optimal rate in a single epoch. However, the batch gradient complexity (i.e., number of adaptive optimization steps), which is important in applications like federated learning, is less well-understood. In particular, all prior work on DP-SCO requires $\\Omega(n)$ batch gradient steps, multiple epochs, or convexity for privacy. We propose an algorithm, Accelerated-DP-SRGD (stochastic recursive gradient descent), which bypasses the limitations of past work: it achieves the optimal rate for DP-SCO (up to polylog factors), in a single epoch using $\\sqrt{n}$ batch gradient steps with batch size $\\sqrt{n}$, and can be made private for arbitrary (non-convex) losses via clipping. If the global minimizer is in the constraint set, we can further improve this to $n^{1/4}$ batch gradient steps with batch size $n^{3/4}$. To achieve this, our algorithm combines three key ingredients, a variant of stochastic recursive gradients (SRG), accelerated gradient descent, and correlated noise generation from DP continual counting."
                    },
                    {
                        "title": "CodeGemma: Open Code Models Based on Gemma",
                        "abstract": "This paper introduces CodeGemma, a collection of specialized open code models built on top of Gemma, capable of a variety of code and natural language generation tasks. We release three model variants. CodeGemma 7B pretrained (PT) and instruction-tuned (IT) variants have remarkably resilient natural language understanding, excel in mathematical reasoning, and match code capabilities of other open models. CodeGemma 2B is a state-of-the-art code completion model designed for fast code infilling and open-ended generation in latency-sensitive settings."
                    },
                    {
                        "title": "Teach LLMs to Phish: Stealing Private Information from Language Models",
                        "abstract": "When large language models are trained on private data, it can be a significant privacy risk for them to memorize and regurgitate sensitive information. In this work, we propose a new practical data extraction attack that we call\"neural phishing\". This attack enables an adversary to target and extract sensitive or personally identifiable information (PII), e.g., credit card numbers, from a model trained on user data with upwards of 10% attack success rates, at times, as high as 50%. Our attack assumes only that an adversary can insert as few as 10s of benign-appearing sentences into the training dataset using only vague priors on the structure of the user data."
                    },
                    {
                        "title": "Students Parrot Their Teachers: Membership Inference on Model Distillation",
                        "abstract": "Model distillation is frequently proposed as a technique to reduce the privacy leakage of machine learning. These empirical privacy defenses rely on the intuition that distilled ``student'' models protect the privacy of training data, as they only interact with this data indirectly through a ``teacher'' model. In this work, we design membership inference attacks to systematically study the privacy provided by knowledge distillation to both the teacher and student training sets. Our new attacks show that distillation alone provides only limited privacy across a number of domains. We explain the success of our attacks on distillation by showing that membership inference attacks on a private dataset can succeed even if the target model is *never* queried on any actual training points, but only on inputs whose predictions are highly influenced by training data. Finally, we show that our attacks are strongest when student and teacher sets are similar, or when the attacker can poison the teacher set."
                    },
                    {
                        "title": "Federated Learning of Gboard Language Models with Differential Privacy",
                        "abstract": "We train and deploy language models (LMs) with federated learning (FL) and differential privacy (DP) in Google Keyboard (Gboard). The recent DP-Follow the Regularized Leader (DP-FTRL) algorithm is applied to achieve meaningfully formal DP guarantees without requiring uniform sampling of clients. To provide favorable privacy-utility trade-offs, we introduce a new client participation criterion and discuss the implication of its configuration in large scale systems. We show how quantile-based clip estimation can be combined with DP-FTRL to adaptively choose the clip norm during training or reduce the hyperparameter tuning in preparation of training. With the help of pretraining on public data, we trained and deployed more than fifteen Gboard LMs that achieve high utility and $\\rho-$zCDP privacy guarantees with $\\rho \\in (0.3, 2)$, with one model additionally trained with secure aggregation.We summarize our experience and provide concrete suggestions on DP training for practitioners."
                    },
                    {
                        "title": "Poisoning Web-Scale Training Datasets is Practical",
                        "abstract": "Deep learning models are often trained on distributed, web-scale datasets crawled from the internet. In this paper, we introduce two new dataset poisoning attacks that intentionally introduce malicious examples to a model\u2019s performance. Our attacks are immediately practical and could, today, poison 10 popular datasets. Our first attack, split-view poisoning, exploits the mutable nature of internet content to ensure a dataset annotator\u2019s initial view of the dataset differs from the view downloaded by subsequent clients. By exploiting specific invalid trust assumptions, we show how we could have poisoned 0.01% of the LAION-400M or COYO-700M datasets for just $60 USD. Our second attack, frontrunning poisoning, targets web-scale datasets that periodically snapshot crowd-sourced content\u2014such as Wikipedia\u2014where an attacker only needs a time-limited window to inject malicious examples. In light of both attacks, we notify the maintainers of each affected dataset and recommended several low-overhead defenses."
                    },
                    {
                        "title": "MADLAD-400: A Multilingual And Document-Level Large Audited Dataset",
                        "abstract": "We introduce MADLAD-400, a manually audited, general domain 3T token monolingual dataset based on CommonCrawl, spanning 419 languages. We discuss the limitations revealed by self-auditing MADLAD-400, and the role data auditing had in the dataset creation process. We then train and release a 10.7B-parameter multilingual machine translation model on 250 billion tokens covering over 450 languages using publicly available data, and find that it is competitive with models that are significantly larger, and report the results on different domains. In addition, we train a 8B-parameter language model, and assess the results on few-shot translation. We make the baseline models available to the research community."
                    },
                    {
                        "title": "Report of the 1st Workshop on Generative AI and Law",
                        "abstract": "This report presents the takeaways of the inaugural Workshop on Generative AI and Law (GenLaw), held in July 2023. A cross-disciplinary group of practitioners and scholars from computer science and law convened to discuss the technical, doctrinal, and policy challenges presented by law for Generative AI, and by Generative AI for law, with an emphasis on U.S. law in particular. We begin the report with a high-level statement about why Generative AI is both immensely significant and immensely challenging for law. To meet these challenges, we conclude that there is an essential need for 1) a shared knowledge base that provides a common conceptual language for experts across disciplines; 2) clarification of the distinctive technical capabilities of generative-AI systems, as compared and contrasted to other computer and AI systems; 3) a logical taxonomy of the legal issues these systems raise; and, 4) a concrete research agenda to promote collaboration and knowledge-sharing on emerging issues at the intersection of Generative AI and law. In this report, we synthesize the key takeaways from the GenLaw workshop that begin to address these needs. All of the listed authors contributed to the workshop upon which this report is based, but they and their organizations do not necessarily endorse all of the specific claims in this report."
                    },
                    {
                        "title": "Are aligned neural networks adversarially aligned?",
                        "abstract": "Large language models are now tuned to align with the goals of their creators, namely to be\"helpful and harmless.\"These models should respond helpfully to user questions, but refuse to answer requests that could cause harm. However, adversarial users can construct inputs which circumvent attempts at alignment. In this work, we study adversarial alignment, and ask to what extent these models remain aligned when interacting with an adversarial user who constructs worst-case inputs (adversarial examples). These inputs are designed to cause the model to emit harmful content that would otherwise be prohibited. We show that existing NLP-based optimization attacks are insufficiently powerful to reliably attack aligned text models: even when current NLP-based attacks fail, we can find adversarial inputs with brute force. As a result, the failure of current attacks should not be seen as proof that aligned text models remain aligned under adversarial inputs. However the recent trend in large-scale ML models is multimodal models that allow users to provide images that influence the text that is generated. We show these models can be easily attacked, i.e., induced to perform arbitrary un-aligned behavior through adversarial perturbation of the input image. We conjecture that improved NLP attacks may demonstrate this same level of adversarial control over text-only models."
                    },
                    {
                        "title": "Privacy Amplification for Matrix Mechanisms",
                        "abstract": "Privacy amplification exploits randomness in data selection to provide tighter differential privacy (DP) guarantees. This analysis is key to DP-SGD's success in machine learning, but, is not readily applicable to the newer state-of-the-art algorithms. This is because these algorithms, known as DP-FTRL, use the matrix mechanism to add correlated noise instead of independent noise as in DP-SGD. In this paper, we propose\"MMCC\", the first algorithm to analyze privacy amplification via sampling for any generic matrix mechanism. MMCC is nearly tight in that it approaches a lower bound as $\\epsilon\\to0$. To analyze correlated outputs in MMCC, we prove that they can be analyzed as if they were independent, by conditioning them on prior outputs. Our\"conditional composition theorem\"has broad utility: we use it to show that the noise added to binary-tree-DP-FTRL can asymptotically match the noise added to DP-SGD with amplification. Our amplification algorithm also has practical empirical utility: we show it leads to significant improvement in the privacy-utility trade-offs for DP-FTRL algorithms on standard benchmarks."
                    },
                    {
                        "title": "User Inference Attacks on Large Language Models",
                        "abstract": "Fine-tuning is a common and effective method for tailoring large language models (LLMs) to specialized tasks and applications. In this paper, we study the privacy implications of fine-tuning LLMs on user data. To this end, we consider a realistic threat model, called user inference, wherein an attacker infers whether or not a user's data was used for fine-tuning. We design attacks for performing user inference that require only black-box access to the fine-tuned LLM and a few samples from a user which need not be from the fine-tuning dataset. We find that LLMs are susceptible to user inference across a variety of fine-tuning datasets, at times with near perfect attack success rates. Further, we theoretically and empirically investigate the properties that make users vulnerable to user inference, finding that outlier users, users with identifiable shared features between examples, and users that contribute a large fraction of the fine-tuning data are most susceptible to attack. Based on these findings, we identify several methods for mitigating user inference including training with example-level differential privacy, removing within-user duplicate examples, and reducing a user's contribution to the training data. While these techniques provide partial mitigation of user inference, we highlight the need to develop methods to fully protect fine-tuned LLMs against this privacy risk."
                    },
                    {
                        "title": "PaLM 2 Technical Report",
                        "abstract": "We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities and is more compute-efficient than its predecessor PaLM. PaLM 2 is a Transformer-based model trained using a mixture of objectives. Through extensive evaluations on English and multilingual language, and reasoning tasks, we demonstrate that PaLM 2 has significantly improved quality on downstream tasks across different model sizes, while simultaneously exhibiting faster and more efficient inference compared to PaLM. This improved efficiency enables broader deployment while also allowing the model to respond faster, for a more natural pace of interaction. PaLM 2 demonstrates robust reasoning capabilities exemplified by large improvements over PaLM on BIG-Bench and other reasoning tasks. PaLM 2 exhibits stable performance on a suite of responsible AI evaluations, and enables inference-time control over toxicity without additional overhead or impact on other capabilities. Overall, PaLM 2 achieves state-of-the-art performance across a diverse set of tasks and capabilities. When discussing the PaLM 2 family, it is important to distinguish between pre-trained models (of various sizes), fine-tuned variants of these models, and the user-facing products that use these models. In particular, user-facing products typically include additional pre- and post-processing steps. Additionally, the underlying models may evolve over time. Therefore, one should not expect the performance of user-facing products to exactly match the results reported in this report."
                    },
                    {
                        "title": "Private Federated Learning with Autotuned Compression",
                        "abstract": "We propose new techniques for reducing communication in private federated learning without the need for setting or tuning compression rates. Our on-the-fly methods automatically adjust the compression rate based on the error induced during training, while maintaining provable privacy guarantees through the use of secure aggregation and differential privacy. Our techniques are provably instance-optimal for mean estimation, meaning that they can adapt to the ``hardness of the problem\"with minimal interactivity. We demonstrate the effectiveness of our approach on real-world datasets by achieving favorable compression rates without the need for tuning."
                    }
                ]
            },
            "45d7c289-a64a-41bc-af60-cbf996cd6655": {
                "pk": "45d7c289-a64a-41bc-af60-cbf996cd6655",
                "name": "Krishnamurthy Dvijotham",
                "collaborators": [
                    "Katherine M. Collins",
                    "Adrian Weller",
                    "Abhradeep Thakurta",
                    "Deepak Ramachandran",
                    "Rudy Bunel",
                    "M. P. Kumar",
                    "Georgina Evans",
                    "Kimin Lee",
                    "Thomas Steinke",
                    "Senjuti Dutta"
                ],
                "domain": [
                    "Differential Privacy",
                    "Neural Networks",
                    "Formal Verification",
                    "Human Feedback"
                ],
                "publications": [
                    {
                        "title": "Private Gradient Descent for Linear Regression: Tighter Error Bounds and Instance-Specific Uncertainty Estimation",
                        "abstract": "We provide an improved analysis of standard differentially private gradient descent for linear regression under the squared error loss. Under modest assumptions on the input, we characterize the distribution of the iterate at each time step. Our analysis leads to new results on the algorithm's accuracy: for a proper fixed choice of hyperparameters, the sample complexity depends only linearly on the dimension of the data. This matches the dimension-dependence of the (non-private) ordinary least squares estimator as well as that of recent private algorithms that rely on sophisticated adaptive gradient-clipping schemes (Varshney et al., 2022; Liu et al., 2023). Our analysis of the iterates' distribution also allows us to construct confidence intervals for the empirical optimizer which adapt automatically to the variance of the algorithm on a particular data set. We validate our theorems through experiments on synthetic data."
                    },
                    {
                        "title": "Achieving the Tightest Relaxation of Sigmoids for Formal Verification",
                        "abstract": "In the field of formal verification, Neural Networks (NNs) are typically reformulated into equivalent mathematical programs which are optimized over. To overcome the inherent non-convexity of these reformulations, convex relaxations of nonlinear activation functions are typically utilized. Common relaxations (i.e., static linear cuts) of\"S-shaped\"activation functions, however, can be overly loose, slowing down the overall verification process. In this paper, we derive tuneable hyperplanes which upper and lower bound the sigmoid activation function. When tuned in the dual space, these affine bounds smoothly rotate around the nonlinear manifold of the sigmoid activation function. This approach, termed $\\alpha$-sig, allows us to tractably incorporate the tightest possible, element-wise convex relaxation of the sigmoid activation function into a formal verification framework. We embed these relaxations inside of large verification tasks and compare their performance to LiRPA and $\\alpha$-CROWN, a state-of-the-art verification duo."
                    },
                    {
                        "title": "Confidence-aware Reward Optimization for Fine-tuning Text-to-Image Models",
                        "abstract": "Fine-tuning text-to-image models with reward functions trained on human feedback data has proven effective for aligning model behavior with human intent. However, excessive optimization with such reward models, which serve as mere proxy objectives, can compromise the performance of fine-tuned models, a phenomenon known as reward overoptimization. To investigate this issue in depth, we introduce the Text-Image Alignment Assessment (TIA2) benchmark, which comprises a diverse collection of text prompts, images, and human annotations. Our evaluation of several state-of-the-art reward models on this benchmark reveals their frequent misalignment with human assessment. We empirically demonstrate that overoptimization occurs notably when a poorly aligned reward model is used as the fine-tuning objective. To address this, we propose TextNorm, a simple method that enhances alignment based on a measure of reward model confidence estimated across a set of semantically contrastive text prompts. We demonstrate that incorporating the confidence-calibrated rewards in fine-tuning effectively reduces overoptimization, resulting in twice as many wins in human evaluation for text-image alignment compared against the baseline reward models."
                    },
                    {
                        "title": "Monotone, Bi-Lipschitz, and Polyak-\u0141ojasiewicz Networks",
                        "abstract": "This paper presents a new bi-Lipschitz invertible neural network, the BiLipNet, which has the ability to smoothly control both its Lipschitzness (output sensitivity to input perturbations) and inverse Lipschitzness (input distinguishability from different outputs). The second main contribution is a new scalar-output network, the PLNet, which is a composition of a BiLipNet and a quadratic potential. We show that PLNet satisfies the Polyak-Lojasiewicz condition and can be applied to learn non-convex surrogate losses with a unique and efficiently-computable global minimum. The central technical element in these networks is a novel invertible residual layer with certified strong monotonicity and Lipschitzness, which we compose with orthogonal layers to build the BiLipNet. The certification of these properties is based on incremental quadratic constraints, resulting in much tighter bounds than can be achieved with spectral normalization. Moreover, we formulate the calculation of the inverse of a BiLipNet -- and hence the minimum of a PLNet -- as a series of three-operator splitting problems, for which fast algorithms can be applied."
                    },
                    {
                        "title": "Efficient and Near-Optimal Noise Generation for Streaming Differential Privacy",
                        "abstract": "In the task of differentially private (DP) continual counting, we receive a stream of increments and our goal is to output an approximate running total of these increments, without revealing too much about any specific increment. Despite its simplicity, differentially private continual counting has attracted significant attention both in theory and in practice. Existing algorithms for differentially private continual counting are either inefficient in terms of their space usage or add an excessive amount of noise, inducing suboptimal utility. The most practical DP continual counting algorithms add carefully correlated Gaussian noise to the values. The task of choosing the covariance for this noise can be expressed in terms of factoring the lower-triangular matrix of ones (which computes prefix sums). We present two approaches from this class (for different parameter regimes) that achieve near-optimal utility for DP continual counting and only require logarithmic or polylogarithmic space (and time). Our first approach is based on a space-efficient streaming matrix multiplication algorithm for a class of Toeplitz matrices. We show that to instantiate this algorithm for DP continual counting, it is sufficient to find a low-degree rational function that approximates the square root on a circle in the complex plane. We then apply and extend tools from approximation theory to achieve this. We also derive efficient closed-forms for the objective function for arbitrarily many steps, and show direct numerical optimization yields a highly practical solution to the problem. Our second approach combines our first approach with a recursive construction similar to the binary tree mechanism."
                    },
                    {
                        "title": "Understanding Subjectivity through the Lens of Motivational Context in Model-Generated Image Satisfaction",
                        "abstract": "Image generation models are poised to become ubiquitous in a range of applications. These models are often fine-tuned and evaluated using human quality judgments that assume a universal standard, failing to consider the subjectivity of such tasks. To investigate how to quantify subjectivity, and the scale of its impact, we measure how assessments differ among human annotators across different use cases. Simulating the effects of ordinarily latent elements of annotators subjectivity, we contrive a set of motivations (t-shirt graphics, presentation visuals, and phone background images) to contextualize a set of crowdsourcing tasks. Our results show that human evaluations of images vary within individual contexts and across combinations of contexts. Three key factors affecting this subjectivity are image appearance, image alignment with text, and representation of objects mentioned in the text. Our study highlights the importance of taking individual users and contexts into account, both when building and evaluating generative models"
                    },
                    {
                        "title": "MINT: A wrapper to make multi-modal and multi-image AI models interactive",
                        "abstract": "During the diagnostic process, doctors incorporate multimodal information including imaging and the medical history - and similarly medical AI development has increasingly become multimodal. In this paper we tackle a more subtle challenge: doctors take a targeted medical history to obtain only the most pertinent pieces of information; how do we enable AI to do the same? We develop a wrapper method named MINT (Make your model INTeractive) that automatically determines what pieces of information are most valuable at each step, and ask for only the most useful information. We demonstrate the efficacy of MINT wrapping a skin disease prediction model, where multiple images and a set of optional answers to $25$ standard metadata questions (i.e., structured medical history) are used by a multi-modal deep network to provide a differential diagnosis. We show that MINT can identify whether metadata inputs are needed and if so, which question to ask next. We also demonstrate that when collecting multiple images, MINT can identify if an additional image would be beneficial, and if so, which type of image to capture. We showed that MINT reduces the number of metadata and image inputs needed by 82% and 36.2% respectively, while maintaining predictive performance. Using real-world AI dermatology system data, we show that needing fewer inputs can retain users that may otherwise fail to complete the system submission and drop off without a diagnosis. Qualitative examples show MINT can closely mimic the step-by-step decision making process of a clinical workflow and how this is different for straight forward cases versus more difficult, ambiguous cases. Finally we demonstrate how MINT is robust to different underlying multi-model classifiers and can be easily adapted to user requirements without significant model re-training."
                    },
                    {
                        "title": "Stealing Part of a Production Language Model",
                        "abstract": "We introduce the first model-stealing attack that extracts precise, nontrivial information from black-box production language models like OpenAI's ChatGPT or Google's PaLM-2. Specifically, our attack recovers the embedding projection layer (up to symmetries) of a transformer model, given typical API access. For under \\$20 USD, our attack extracts the entire projection matrix of OpenAI's Ada and Babbage language models. We thereby confirm, for the first time, that these black-box models have a hidden dimension of 1024 and 2048, respectively. We also recover the exact hidden dimension size of the gpt-3.5-turbo model, and estimate it would cost under $2,000 in queries to recover the entire projection matrix. We conclude with potential defenses and mitigations, and discuss the implications of possible future work that could extend our attack."
                    },
                    {
                        "title": "Beyond Thumbs Up/Down: Untangling Challenges of Fine-Grained Feedback for Text-to-Image Generation",
                        "abstract": "Human feedback plays a critical role in learning and refining reward models for text-to-image generation, but the optimal form the feedback should take for learning an accurate reward function has not been conclusively established. This paper investigates the effectiveness of fine-grained feedback which captures nuanced distinctions in image quality and prompt-alignment, compared to traditional coarse-grained feedback (for example, thumbs up/down or ranking between a set of options). While fine-grained feedback holds promise, particularly for systems catering to diverse societal preferences, we show that demonstrating its superiority to coarse-grained feedback is not automatic. Through experiments on real and synthetic preference data, we surface the complexities of building effective models due to the interplay of model choice, feedback type, and the alignment between human judgment and computational interpretation. We identify key challenges in eliciting and utilizing fine-grained feedback, prompting a reassessment of its assumed benefits and practicality. Our findings -- e.g., that fine-grained feedback can lead to worse models for a fixed budget, in some settings; however, in controlled settings with known attributes, fine grained rewards can indeed be more helpful -- call for careful consideration of feedback attributes and potentially beckon novel modeling approaches to appropriately unlock the potential value of fine-grained feedback in-the-wild."
                    },
                    {
                        "title": "Provably Bounding Neural Network Preimages",
                        "abstract": "Most work on the formal verification of neural networks has focused on bounding the set of outputs that correspond to a given set of inputs (for example, bounded perturbations of a nominal input). However, many use cases of neural network verification require solving the inverse problem, or over-approximating the set of inputs that lead to certain outputs. We present the INVPROP algorithm for verifying properties over the preimage of a linearly constrained output set, which can be combined with branch-and-bound to increase precision. Contrary to other approaches, our efficient algorithm is GPU-accelerated and does not require a linear programming solver. We demonstrate our algorithm for identifying safe control regions for a dynamical system via backward reachability analysis, verifying adversarial robustness, and detecting out-of-distribution inputs to a neural network. Our results show that in certain settings, we find over-approximations over 2500x tighter than prior work while being 2.5x faster. By strengthening robustness verification with output constraints, we consistently verify more properties than the previous state-of-the-art on multiple benchmarks, including a large model with 167k neurons in VNN-COMP 2023. Our algorithm has been incorporated into the $\\alpha,\\!\\beta$-CROWN verifier, available at https://abcrown.org."
                    },
                    {
                        "title": "Faithful Knowledge Distillation",
                        "abstract": "Knowledge distillation (KD) has received much attention due to its success in compressing networks to allow for their deployment in resource-constrained systems. While the problem of adversarial robustness has been studied before in the KD setting, previous works overlook what we term the relative calibration of the student network with respect to its teacher in terms of soft confidences. In particular, we focus on two crucial questions with regard to a teacher-student pair: (i) do the teacher and student disagree at points close to correctly classified dataset examples, and (ii) is the distilled student as confident as the teacher around dataset examples? These are critical questions when considering the deployment of a smaller student network trained from a robust teacher within a safety-critical setting. To address these questions, we introduce a faithful imitation framework to discuss the relative calibration of confidences and provide empirical and certified methods to evaluate the relative calibration of a student w.r.t. its teacher. Further, to verifiably align the relative calibration incentives of the student to those of its teacher, we introduce faithful distillation. Our experiments on the MNIST, Fashion-MNIST and CIFAR-10 datasets demonstrate the need for such an analysis and the advantages of the increased verifiability of faithful distillation over alternative adversarial distillation methods."
                    },
                    {
                        "title": "Selective Concept Models: Permitting Stakeholder Customisation at Test-Time",
                        "abstract": "Concept-based models perform prediction using a set of concepts that are interpretable to stakeholders. However, such models often involve a fixed, large number of concepts, which may place a substantial cognitive load on stakeholders. We propose Selective COncept Models (SCOMs) which make predictions using only a subset of concepts and can be customised by stakeholders at test-time according to their preferences. We show that SCOMs only require a fraction of the total concepts to achieve optimal accuracy on multiple real-world datasets. Further, we collect and release a new dataset, CUB-Sel, consisting of human concept set selections for 900 bird images from the popular CUB dataset. Using CUB-Sel, we show that humans have unique individual preferences for the choice of concepts they prefer to reason about, and struggle to identify the most theoretically informative concepts. The customisation and concept selection provided by SCOM improves the efficiency of interpretation and intervention for stakeholders."
                    },
                    {
                        "title": "Efficient Error Certification for Physics-Informed Neural Networks",
                        "abstract": "Recent work provides promising evidence that Physics-Informed Neural Networks (PINN) can efficiently solve partial differential equations (PDE). However, previous works have failed to provide guarantees on the worst-case residual error of a PINN across the spatio-temporal domain - a measure akin to the tolerance of numerical solvers - focusing instead on point-wise comparisons between their solution and the ones obtained by a solver on a set of inputs. In real-world applications, one cannot consider tests on a finite set of points to be sufficient grounds for deployment, as the performance could be substantially worse on a different set. To alleviate this issue, we establish guaranteed error-based conditions for PINNs over their continuous applicability domain. To verify the extent to which they hold, we introduce $\\partial$-CROWN: a general, efficient and scalable post-training framework to bound PINN residual errors. We demonstrate its effectiveness in obtaining tight certificates by applying it to two classically studied PINNs - Burgers' and Schr\\\"odinger's equations -, and two more challenging ones with real-world applications - the Allan-Cahn and Diffusion-Sorption equations."
                    },
                    {
                        "title": "Expressive Losses for Verified Robustness via Convex Combinations",
                        "abstract": "In order to train networks for verified adversarial robustness, it is common to over-approximate the worst-case loss over perturbation regions, resulting in networks that attain verifiability at the expense of standard performance. As shown in recent work, better trade-offs between accuracy and robustness can be obtained by carefully coupling adversarial training with over-approximations. We hypothesize that the expressivity of a loss function, which we formalize as the ability to span a range of trade-offs between lower and upper bounds to the worst-case loss through a single parameter (the over-approximation coefficient), is key to attaining state-of-the-art performance. To support our hypothesis, we show that trivial expressive losses, obtained via convex combinations between adversarial attacks and IBP bounds, yield state-of-the-art results across a variety of settings in spite of their conceptual simplicity. We provide a detailed analysis of the relationship between the over-approximation coefficient and performance profiles across different expressive losses, showing that, while expressivity is essential, better approximations of the worst-case loss are not necessarily linked to superior robustness-accuracy trade-offs."
                    },
                    {
                        "title": "Human Uncertainty in Concept-Based AI Systems",
                        "abstract": "Placing a human in the loop may help abate the risks of deploying AI systems in safety-critical settings (e.g., a clinician working with a medical AI system). However, mitigating risks arising from human error and uncertainty within such human-AI interactions is an important and understudied issue. In this work, we study human uncertainty in the context of concept-based models, a family of AI systems that enable human feedback via concept interventions where an expert intervenes on human-interpretable concepts relevant to the task. Prior work in this space often assumes that humans are oracles who are always certain and correct. Yet, real-world decision-making by humans is prone to occasional mistakes and uncertainty. We study how existing concept-based models deal with uncertain interventions from humans using two novel datasets: UMNIST, a visual dataset with controlled simulated uncertainty based on the MNIST dataset, and CUB-S, a relabeling of the popular CUB concept dataset with rich, densely-annotated soft labels from humans. We show that training with uncertain concept labels may help mitigate weaknesses of concept-based systems when handling uncertain interventions. These results allow us to identify several open challenges, which we argue can be tackled through future multidisciplinary research on building interactive uncertainty-aware systems. To facilitate further research, we release a new elicitation platform, UElic, to collect uncertain feedback from humans in collaborative prediction tasks."
                    },
                    {
                        "title": "Learning to Receive Help: Intervention-Aware Concept Embedding Models",
                        "abstract": "Concept Bottleneck Models (CBMs) tackle the opacity of neural architectures by constructing and explaining their predictions using a set of high-level concepts. A special property of these models is that they permit concept interventions, wherein users can correct mispredicted concepts and thus improve the model's performance. Recent work, however, has shown that intervention efficacy can be highly dependent on the order in which concepts are intervened on and on the model's architecture and training hyperparameters. We argue that this is rooted in a CBM's lack of train-time incentives for the model to be appropriately receptive to concept interventions. To address this, we propose Intervention-aware Concept Embedding models (IntCEMs), a novel CBM-based architecture and training paradigm that improves a model's receptiveness to test-time interventions. Our model learns a concept intervention policy in an end-to-end fashion from where it can sample meaningful intervention trajectories at train-time. This conditions IntCEMs to effectively select and receive concept interventions when deployed at test-time. Our experiments show that IntCEMs significantly outperform state-of-the-art concept-interpretable models when provided with test-time concept interventions, demonstrating the effectiveness of our approach."
                    },
                    {
                        "title": "Rich Human Feedback for Text-to-Image Generation",
                        "abstract": "Recent Text-to-Image (T2I) generation models such as Stable Diffusion and Imagen have made significant progress in generating high-resolution images based on text descriptions. However, many generated images still suffer from issues such as artifacts/implausibility, misalignment with text descriptions, and low aesthetic quality. Inspired by the success of Reinforcement Learning with Human Feedback (RLHF) for large language models, prior works collected human-provided scores as feedback on generated images and trained a reward model to improve the T2I generation. In this paper, we enrich the feedback signal by (i) marking image regions that are implausible or misaligned with the text, and (ii) annotating which words in the text prompt are misrepresented or missing on the image. We collect such rich human feedback on 18K generated images (RichHF-18K) and train a multimodal transformer to predict the rich feedback automatically. We show that the predicted rich human feedback can be leveraged to improve image generation, for example, by selecting high-quality training data to finetune and improve the generative models, or by creating masks with predicted heatmaps to inpaint the problematic regions. Notably, the improvements generalize to models (Muse) beyond those used to generate the images on which human feedback data were collected (Stable Diffusion variants). The RichHF-18K data set will be released in our GitHub repository: https://github.com/google-research/google-research/tree/master/richhf_18k."
                    },
                    {
                        "title": "Training Private Models That Know What They Don't Know",
                        "abstract": "Training reliable deep learning models which avoid making overconfident but incorrect predictions is a longstanding challenge. This challenge is further exacerbated when learning has to be differentially private: protection provided to sensitive data comes at the price of injecting additional randomness into the learning process. In this work, we conduct a thorough empirical investigation of selective classifiers -- that can abstain when they are unsure -- under a differential privacy constraint. We find that several popular selective prediction approaches are ineffective in a differentially private setting as they increase the risk of privacy leakage. At the same time, we identify that a recent approach that only uses checkpoints produced by an off-the-shelf private learning algorithm stands out as particularly suitable under DP. Further, we show that differential privacy does not just harm utility but also degrades selective classification performance. To analyze this effect across privacy levels, we propose a novel evaluation mechanism which isolate selective prediction performance across model utility levels. Our experimental results show that recovering the performance level attainable by non-private models is possible but comes at a considerable coverage cost as the privacy budget decreases."
                    }
                ]
            },
            "d22fd8f9-226a-47f5-b454-042eeb236f56": {
                "pk": "d22fd8f9-226a-47f5-b454-042eeb236f56",
                "name": "Krishna Pillutla",
                "collaborators": [
                    "H. B. McMahan",
                    "Krishnamurthy Dvijotham",
                    "Thomas Steinke",
                    "Abhradeep Thakurta",
                    "Zachary Charles",
                    "Arun Ganesh",
                    "Ryan McKenna",
                    "Nicole Mitchell",
                    "Keith Rush",
                    "Nikhil Kandpal"
                ],
                "domain": [
                    "Differential Privacy",
                    "Large Language Models",
                    "Privacy-Preserving Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Efficient and Near-Optimal Noise Generation for Streaming Differential Privacy",
                        "abstract": "In the task of differentially private (DP) continual counting, we receive a stream of increments and our goal is to output an approximate running total of these increments, without revealing too much about any specific increment. Despite its simplicity, differentially private continual counting has attracted significant attention both in theory and in practice. Existing algorithms for differentially private continual counting are either inefficient in terms of their space usage or add an excessive amount of noise, inducing suboptimal utility. The most practical DP continual counting algorithms add carefully correlated Gaussian noise to the values. The task of choosing the covariance for this noise can be expressed in terms of factoring the lower-triangular matrix of ones (which computes prefix sums). We present two approaches from this class (for different parameter regimes) that achieve near-optimal utility for DP continual counting and only require logarithmic or polylogarithmic space (and time). Our first approach is based on a space-efficient streaming matrix multiplication algorithm for a class of Toeplitz matrices. We show that to instantiate this algorithm for DP continual counting, it is sufficient to find a low-degree rational function that approximates the square root on a circle in the complex plane. We then apply and extend tools from approximation theory to achieve this. We also derive efficient closed-forms for the objective function for arbitrarily many steps, and show direct numerical optimization yields a highly practical solution to the problem. Our second approach combines our first approach with a recursive construction similar to the binary tree mechanism."
                    },
                    {
                        "title": "Fine-Tuning Large Language Models with User-Level Differential Privacy",
                        "abstract": "We investigate practical and scalable algorithms for training large language models (LLMs) with user-level differential privacy (DP) in order to provably safeguard all the examples contributed by each user. We study two variants of DP-SGD with: (1) example-level sampling (ELS) and per-example gradient clipping, and (2) user-level sampling (ULS) and per-user gradient clipping. We derive a novel user-level DP accountant that allows us to compute provably tight privacy guarantees for ELS. Using this, we show that while ELS can outperform ULS in specific settings, ULS generally yields better results when each user has a diverse collection of examples. We validate our findings through experiments in synthetic mean estimation and LLM fine-tuning tasks under fixed compute budgets. We find that ULS is significantly better in settings where either (1) strong privacy guarantees are required, or (2) the compute budget is large. Notably, our focus on LLM-compatible training algorithms allows us to scale to models with hundreds of millions of parameters and datasets with hundreds of thousands of users."
                    },
                    {
                        "title": "User Inference Attacks on Large Language Models",
                        "abstract": "Fine-tuning is a common and effective method for tailoring large language models (LLMs) to specialized tasks and applications. In this paper, we study the privacy implications of fine-tuning LLMs on user data. To this end, we consider a realistic threat model, called user inference, wherein an attacker infers whether or not a user's data was used for fine-tuning. We design attacks for performing user inference that require only black-box access to the fine-tuned LLM and a few samples from a user which need not be from the fine-tuning dataset. We find that LLMs are susceptible to user inference across a variety of fine-tuning datasets, at times with near perfect attack success rates. Further, we theoretically and empirically investigate the properties that make users vulnerable to user inference, finding that outlier users, users with identifiable shared features between examples, and users that contribute a large fraction of the fine-tuning data are most susceptible to attack. Based on these findings, we identify several methods for mitigating user inference including training with example-level differential privacy, removing within-user duplicate examples, and reducing a user's contribution to the training data. While these techniques provide partial mitigation of user inference, we highlight the need to develop methods to fully protect fine-tuned LLMs against this privacy risk."
                    }
                ]
            },
            "448ef727-da44-44e8-951f-5902a332370a": {
                "pk": "448ef727-da44-44e8-951f-5902a332370a",
                "name": "Arun Ganesh",
                "collaborators": [
                    "Abhradeep Thakurta",
                    "Christopher A. Choquette-Choo",
                    "Thomas Steinke",
                    "Jalaj Upadhyay",
                    "T. Steinke",
                    "Om Thakkar",
                    "Milad Nasr",
                    "Mahdi Haghifam",
                    "Sewoong Oh",
                    "Rajiv Mathews"
                ],
                "domain": [
                    "Differential Privacy",
                    "Machine Learning",
                    "Optimization",
                    "Privacy Analysis"
                ],
                "publications": [
                    {
                        "title": "The Last Iterate Advantage: Empirical Auditing and Principled Heuristic Analysis of Differentially Private SGD",
                        "abstract": "We propose a simple heuristic privacy analysis of noisy clipped stochastic gradient descent (DP-SGD) in the setting where only the last iterate is released and the intermediate iterates remain hidden. Namely, our heuristic assumes a linear structure for the model. We show experimentally that our heuristic is predictive of the outcome of privacy auditing applied to various training procedures. Thus it can be used prior to training as a rough estimate of the final privacy leakage. We also probe the limitations of our heuristic by providing some artificial counterexamples where it underestimates the privacy leakage. The standard composition-based privacy analysis of DP-SGD effectively assumes that the adversary has access to all intermediate iterates, which is often unrealistic. However, this analysis remains the state of the art in practice. While our heuristic does not replace a rigorous privacy analysis, it illustrates the large gap between the best theoretical upper bounds and the privacy auditing lower bounds and sets a target for further work to improve the theoretical privacy analyses. We also empirically support our heuristic and show existing privacy auditing attacks are bounded by our heuristic analysis in both vision and language tasks."
                    },
                    {
                        "title": "Near Exact Privacy Amplification for Matrix Mechanisms",
                        "abstract": "We study the problem of computing the privacy parameters for DP machine learning when using privacy amplification via random batching and noise correlated across rounds via a correlation matrix $\\textbf{C}$ (i.e., the matrix mechanism). Past work on this problem either only applied to banded $\\textbf{C}$, or gave loose privacy parameters. In this work, we give a framework for computing near-exact privacy parameters for any lower-triangular, non-negative $\\textbf{C}$. Our framework allows us to optimize the correlation matrix $\\textbf{C}$ while accounting for amplification, whereas past work could not. Empirically, we show this lets us achieve smaller RMSE on prefix sums than the previous state-of-the-art (SOTA). We also show that we can improve on the SOTA performance on deep learning tasks. Our two main technical tools are (i) using Monte Carlo accounting to bypass composition, which was the main technical challenge for past work, and (ii) a\"balls-in-bins\"batching scheme that enables easy privacy analysis and is closer to practical random batching than Poisson sampling."
                    },
                    {
                        "title": "Optimal Rates for $O(1)$-Smooth DP-SCO with a Single Epoch and Large Batches",
                        "abstract": "In this paper we revisit the DP stochastic convex optimization (SCO) problem. For convex smooth losses, it is well-known that the canonical DP-SGD (stochastic gradient descent) achieves the optimal rate of $O\\left(\\frac{LR}{\\sqrt{n}} + \\frac{LR \\sqrt{p \\log(1/\\delta)}}{\\epsilon n}\\right)$ under $(\\epsilon, \\delta)$-DP, and also well-known that variants of DP-SGD can achieve the optimal rate in a single epoch. However, the batch gradient complexity (i.e., number of adaptive optimization steps), which is important in applications like federated learning, is less well-understood. In particular, all prior work on DP-SCO requires $\\Omega(n)$ batch gradient steps, multiple epochs, or convexity for privacy. We propose an algorithm, Accelerated-DP-SRGD (stochastic recursive gradient descent), which bypasses the limitations of past work: it achieves the optimal rate for DP-SCO (up to polylog factors), in a single epoch using $\\sqrt{n}$ batch gradient steps with batch size $\\sqrt{n}$, and can be made private for arbitrary (non-convex) losses via clipping. If the global minimizer is in the constraint set, we can further improve this to $n^{1/4}$ batch gradient steps with batch size $n^{3/4}$. To achieve this, our algorithm combines three key ingredients, a variant of stochastic recursive gradients (SRG), accelerated gradient descent, and correlated noise generation from DP continual counting."
                    },
                    {
                        "title": "Universality of Langevin Diffusion for Private Optimization, with Applications to Sampling from Rashomon Sets",
                        "abstract": "In this paper we provide an algorithmic framework based on Langevin diffusion (LD) and its corresponding discretizations that allow us to simultaneously obtain: i) An algorithm for sampling from the exponential mechanism (McSherry and Talwar, 2007), whose privacy analysis does not depend on convexity and which can be stopped at anytime without compromising privacy, and ii) tight uniform stability guarantees for the exponential mechanism. As a direct consequence, we obtain optimal excess empirical and population risk guarantees for (strongly) convex losses under both pure and approximate differential privacy (DP). The framework allows us to design a DP uniform sampler from the Rashomon set. Rashomon sets are widely used in interpretable and robust machine learning, understanding variable importance, and characterizing fairness."
                    },
                    {
                        "title": "Why Is Public Pretraining Necessary for Private Model Training?",
                        "abstract": "In the privacy-utility tradeoff of a model trained on benchmark language and vision tasks, remarkable improvements have been widely reported with the use of pretraining on publicly available data. This is in part due to the benefits of transfer learning, which is the standard motivation for pretraining in non-private settings. However, the stark contrast in the improvement achieved through pretraining under privacy compared to non-private settings suggests that there may be a deeper, distinct cause driving these gains. To explain this phenomenon, we hypothesize that the non-convex loss landscape of a model training necessitates an optimization algorithm to go through two phases. In the first, the algorithm needs to select a good\"basin\"in the loss landscape. In the second, the algorithm solves an easy optimization within that basin. The former is a harder problem to solve with private data, while the latter is harder to solve with public data due to a distribution shift or data scarcity. Guided by this intuition, we provide theoretical constructions that provably demonstrate the separation between private training with and without public pretraining. Further, systematic experiments on CIFAR10 and LibriSpeech provide supporting evidence for our hypothesis."
                    },
                    {
                        "title": "Privacy Amplification for Matrix Mechanisms",
                        "abstract": "Privacy amplification exploits randomness in data selection to provide tighter differential privacy (DP) guarantees. This analysis is key to DP-SGD's success in machine learning, but, is not readily applicable to the newer state-of-the-art algorithms. This is because these algorithms, known as DP-FTRL, use the matrix mechanism to add correlated noise instead of independent noise as in DP-SGD. In this paper, we propose\"MMCC\", the first algorithm to analyze privacy amplification via sampling for any generic matrix mechanism. MMCC is nearly tight in that it approaches a lower bound as $\\epsilon\\to0$. To analyze correlated outputs in MMCC, we prove that they can be analyzed as if they were independent, by conditioning them on prior outputs. Our\"conditional composition theorem\"has broad utility: we use it to show that the noise added to binary-tree-DP-FTRL can asymptotically match the noise added to DP-SGD with amplification. Our amplification algorithm also has practical empirical utility: we show it leads to significant improvement in the privacy-utility trade-offs for DP-FTRL algorithms on standard benchmarks."
                    },
                    {
                        "title": "Faster Differentially Private Convex Optimization via Second-Order Methods",
                        "abstract": "Differentially private (stochastic) gradient descent is the workhorse of DP private machine learning in both the convex and non-convex settings. Without privacy constraints, second-order methods, like Newton's method, converge faster than first-order methods like gradient descent. In this work, we investigate the prospect of using the second-order information from the loss function to accelerate DP convex optimization. We first develop a private variant of the regularized cubic Newton method of Nesterov and Polyak, and show that for the class of strongly convex loss functions, our algorithm has quadratic convergence and achieves the optimal excess loss. We then design a practical second-order DP algorithm for the unconstrained logistic regression problem. We theoretically and empirically study the performance of our algorithm. Empirical results show our algorithm consistently achieves the best excess loss compared to other baselines and is 10-40x faster than DP-GD/DP-SGD."
                    },
                    {
                        "title": "Private (Stochastic) Non-Convex Optimization Revisited: Second-Order Stationary Points and Excess Risks",
                        "abstract": "We consider the problem of minimizing a non-convex objective while preserving the privacy of the examples in the training data. Building upon the previous variance-reduced algorithm SpiderBoost, we introduce a new framework that utilizes two different kinds of gradient oracles. The first kind of oracles can estimate the gradient of one point, and the second kind of oracles, less precise and more cost-effective, can estimate the gradient difference between two points. SpiderBoost uses the first kind periodically, once every few steps, while our framework proposes using the first oracle whenever the total drift has become large and relies on the second oracle otherwise. This new framework ensures the gradient estimations remain accurate all the time, resulting in improved rates for finding second-order stationary points. Moreover, we address a more challenging task of finding the global minima of a non-convex objective using the exponential mechanism. Our findings indicate that the regularized exponential mechanism can closely match previous empirical and population risk bounds, without requiring smoothness assumptions for algorithms with polynomial running time. Furthermore, by disregarding running time considerations, we show that the exponential mechanism can achieve a good population risk bound and provide a nearly matching lower bound."
                    },
                    {
                        "title": "(Amplified) Banded Matrix Factorization: A unified approach to private training",
                        "abstract": "Matrix factorization (MF) mechanisms for differential privacy (DP) have substantially improved the state-of-the-art in privacy-utility-computation tradeoffs for ML applications in a variety of scenarios, but in both the centralized and federated settings there remain instances where either MF cannot be easily applied, or other algorithms provide better tradeoffs (typically, as $\\epsilon$ becomes small). In this work, we show how MF can subsume prior state-of-the-art algorithms in both federated and centralized training settings, across all privacy budgets. The key technique throughout is the construction of MF mechanisms with banded matrices (lower-triangular matrices with at most $\\hat{b}$ nonzero bands including the main diagonal). For cross-device federated learning (FL), this enables multiple-participations with a relaxed device participation schema compatible with practical FL infrastructure (as demonstrated by a production deployment). In the centralized setting, we prove that banded matrices enjoy the same privacy amplification results as the ubiquitous DP-SGD algorithm, but can provide strictly better performance in most scenarios -- this lets us always at least match DP-SGD, and often outperform it."
                    },
                    {
                        "title": "Langevin Diffusion: An Almost Universal Algorithm for Private Euclidean (Convex) Optimization",
                        "abstract": "In this paper we revisit the problem of differentially private empirical risk minimization (DP-ERM) and differentially private stochastic convex optimization (DP-SCO). We show that a well-studied continuous time algorithm from statistical physics, called Langevin diffusion (LD), simultaneously provides optimal privacy/utility trade-offs for both DP-ERM and DP-SCO, under \u03b5 -DP, and ( \u03b5 , \u03b4 ) -DP. Using the uniform stability properties of LD, we provide the optimal excess population risk guarantee for (cid:96) 2 -Lipschitz convex losses under \u03b5 -DP (even up to log ( n ) factors), thus improving on [ALD21]. Along the way, we provide various technical tools, which can be of independent interest: i) A new R\u00b4enyi divergence bound for LD, when run on loss functions over two neighboring data sets, ii) Excess empirical risk bounds for last-iterate LD, analogous to that of Shamir and Zhang [SZ13] for noisy stochastic gradient descent (SGD), and iii) A two phase excess risk analysis of LD, where the \ufb01rst phase is when the diffusion has not converged in any reasonable sense to a stationary distribution, and in the second phase when the diffusion has converged to a variant of Gibbs distribution. Our universality results crucially rely on the dynamics of LD. When it has converged to a stationary distribution, we obtain the optimal bounds under \u03b5 -DP. When it is run only for a very short time \u221d 1/ p , we obtain the optimal bounds under ( \u03b5 , \u03b4 ) -DP. Here, p is the dimensionality of the model space. Our work initiates a systematic study of DP continuous time optimization. We believe this may have rami\ufb01cations in the design of discrete time DP optimization algorithms analogous to that in the non-private setting, where continuous time dynamical viewpoints have helped in designing new algorithms, including the celebrated mirror-descent (see the discussion in Nemirovski and Yudin (near) universality of LD to an optimal DP-SCO error bounds -DP."
                    },
                    {
                        "title": "Improved Algorithms and Upper Bounds in Differential Privacy",
                        "abstract": "Improved Algorithms and Upper Bounds in Differential Privacy"
                    },
                    {
                        "title": "Differentially Private Sampling from Rashomon Sets, and the Universality of Langevin Diffusion for Convex Optimization",
                        "abstract": "In this paper we provide an algorithmic framework based on Langevin diffusion (LD) and its corresponding discretizations that allow us to simultaneously obtain: i) An algorithm for sampling from the exponential mechanism, whose privacy analysis does not depend on convexity and which can be stopped at anytime without compromising privacy, and ii) tight uniform stability guarantees for the exponential mechanism. As a direct consequence, we obtain optimal excess empirical and population risk guarantees for (strongly) convex losses under both pure and approximate differential privacy (DP). The framework allows us to design a DP uniform sampler from the Rashomon set. Rashomon sets are widely used in interpretable and robust machine learning, understanding variable importance, and characterizing fairness."
                    },
                    {
                        "title": "Recycling Scraps: Improving Private Learning by Leveraging Intermediate Checkpoints",
                        "abstract": "All state-of-the-art (SOTA) differentially private machine learning (DP ML) methods are iterative in nature, and their privacy analyses allow publicly releasing the intermediate training checkpoints. However, DP ML benchmarks, and even practical deployments, typically use only the \ufb01nal training checkpoint to make predictions. In this work, for the \ufb01rst time, we comprehensively explore various methods that aggregate intermediate checkpoints to improve the utility of DP training. Empirically, we demonstrate that checkpoint aggregations provide signi\ufb01cant gains in the prediction accuracy over the existing SOTA for CIFAR10 and StackOver\ufb02ow datasets, and that these gains get magni\ufb01ed in settings with periodically varying training data distributions. For instance, we improve SOTA StackOver\ufb02ow accuracies to 22.7% (+0.43% absolute) for \u03b5 = 8 . 2 , and 23.84% (+0.43%) for \u03b5 = 18 . 9 . Theoret-ically, we show that uniform tail averaging of checkpoints improves the empirical risk minimization bound compared to the last checkpoint of DP-SGD. Lastly, we initiate an exploration into estimating the uncertainty that DP noise adds in the predictions of DP ML models. We prove that, under standard assumptions on the loss function, the sample variance from last few checkpoints provides a good approximation of the variance of the \ufb01nal model of a DP run. Empirically, we show that the last few checkpoints can provide a reasonable lower bound for the variance of a converged DP model."
                    },
                    {
                        "title": "Universal Algorithms for Clustering Problems",
                        "abstract": "This article presents universal algorithms for clustering problems, including the widely studied k-median, k-means, and k-center objectives. The input is a metric space containing all potential client locations. The algorithm must select k cluster centers such that they are a good solution for any subset of clients that actually realize. Specifically, we aim for low regret, defined as the maximum over all subsets of the difference between the cost of the algorithm\u2019s solution and that of an optimal solution. A universal algorithm\u2019s solution Sol for a clustering problem is said to be an \u03b1 , \u03b2-approximation if for all subsets of clients C\u2032, it satisfies sol (C\u2032) \u2264 \u03b1 \u010b opt (C\u2032) + \u03b2 \u010b mr, where opt (C\u2032 is the cost of the optimal solution for clients (C\u2032) and mr is the minimum regret achievable by any solution. Our main results are universal algorithms for the standard clustering objectives of k-median, k-means, and k-center that achieve (O(1), O(1))-approximations. These results are obtained via a novel framework for universal algorithms using linear programming (LP) relaxations. These results generalize to other \u2113p-objectives and the setting where some subset of the clients are fixed. We also give hardness results showing that (\u03b1, \u03b2)-approximation is NP-hard if \u03b1 or \u03b2 is at most a certain constant, even for the widely studied special case of Euclidean metric spaces. This shows that in some sense, (O(1), O(1))-approximation is the strongest type of guarantee obtainable for universal clustering."
                    },
                    {
                        "title": "Public Data-Assisted Mirror Descent for Private Model Training",
                        "abstract": "In this paper, we revisit the problem of using in-distribution public data to improve the privacy/utility trade-offs for differentially private (DP) model training. (Here, public data refers to auxiliary data sets that have no privacy concerns.) We design a natural variant of DP mirror descent, where the DP gradients of the private/sensitive data act as the linear term, and the loss generated by the public data as the mirror map. We show that, for linear regression with feature vectors drawn from a non-isotropic sub-Gaussian distribution, our algorithm, PDA-DPMD (a variant of mirror descent), provides population risk guarantees that are asymptotically better than the best known guarantees under DP (without having access to public data), when the number of public data samples ($n_{\\sf pub}$) is sufficiently large. We further show that our algorithm has natural\"noise stability\"properties that control the variance due to noise added to ensure DP. We demonstrate the efficacy of our algorithm by showing privacy/utility trade-offs on four benchmark datasets (StackOverflow, WikiText-2, CIFAR-10, and EMNIST). We show that our algorithm not only significantly improves over traditional DP-SGD, which does not have access to public data, but to our knowledge is the first to improve over DP-SGD on models that have been pre-trained with public data."
                    },
                    {
                        "title": "Online Service with Delay",
                        "abstract": "In this article, we introduce the online service with delay problem. In this problem, there are n points in a metric space that issue service requests over time, and there is a server that serves these requests. The goal is to minimize the sum of distance traveled by the server and the total delay (or a penalty function thereof) in serving the requests. This problem models the fundamental tradeoff between batching requests to improve locality and reducing delay to improve response time, which has many applications in operations management, operating systems, logistics, supply chain management, and scheduling. Our main result is to show a poly-logarithmic competitive ratio for the online service with delay problem. This result is obtained by an algorithm that we call the preemptive service algorithm. The salient feature of this algorithm is a process called preemptive service, which uses a novel combination of (recursive) time forwarding and spatial exploration on a metric space. We also generalize our results to k > 1 servers and obtain stronger results for special metrics such as uniform and star metrics that correspond to (weighted) paging problems."
                    },
                    {
                        "title": "Privately Answering Counting Queries with Generalized Gaussian Mechanisms",
                        "abstract": "We consider the problem of answering $k$ counting (i.e. sensitivity-1) queries about a database with $(\\epsilon, \\delta)$-differential privacy. We give a mechanism such that if the true answers to the queries are the vector $d$, the mechanism outputs answers $\\tilde{d}$ with the $\\ell_\\infty$-error guarantee:  $$\\mathcal{E}\\left[||\\tilde{d} - d||_\\infty\\right] = O\\left(\\frac{\\sqrt{k \\log \\log \\log k \\log(1/\\delta)}}{\\epsilon}\\right).$$  This reduces the multiplicative gap between the best known upper and lower bounds on $\\ell_\\infty$-error from $O(\\sqrt{\\log \\log k})$ to $O(\\sqrt{\\log \\log \\log k})$. Our main technical contribution is an analysis of the family of mechanisms of the following form for answering counting queries: Sample $x$ from a \\textit{Generalized Gaussian}, i.e. with probability proportional to $\\exp(-(||x||_p/\\sigma)^p)$, and output $\\tilde{d} = d + x$. This family of mechanisms offers a tradeoff between $\\ell_1$ and $\\ell_\\infty$-error guarantees and may be of independent interest. For $p = O(\\log \\log k)$, this mechanism already matches the previous best known $\\ell_\\infty$-error bound. We arrive at our main result by composing this mechanism for $p = O(\\log \\log \\log k)$ with the sparse vector mechanism, generalizing a technique of Steinke and Ullman."
                    }
                ]
            },
            "8ca31a93-9ecd-4e7e-9491-85be7fd6ece1": {
                "pk": "8ca31a93-9ecd-4e7e-9491-85be7fd6ece1",
                "name": "Thomas Steinke",
                "collaborators": [
                    "Abhradeep Thakurta",
                    "Arun Ganesh",
                    "Christopher A. Choquette-Choo",
                    "Milad Nasr",
                    "Matthew Jagielski",
                    "Krishnamurthy Dvijotham",
                    "Borja Balle",
                    "Jamie Hayes",
                    "Adam Smith",
                    "Andreas Terzis"
                ],
                "domain": [
                    "Differential Privacy",
                    "Machine Learning",
                    "Privacy Analysis",
                    "Model Stealing"
                ],
                "publications": [
                    {
                        "title": "The Last Iterate Advantage: Empirical Auditing and Principled Heuristic Analysis of Differentially Private SGD",
                        "abstract": "We propose a simple heuristic privacy analysis of noisy clipped stochastic gradient descent (DP-SGD) in the setting where only the last iterate is released and the intermediate iterates remain hidden. Namely, our heuristic assumes a linear structure for the model. We show experimentally that our heuristic is predictive of the outcome of privacy auditing applied to various training procedures. Thus it can be used prior to training as a rough estimate of the final privacy leakage. We also probe the limitations of our heuristic by providing some artificial counterexamples where it underestimates the privacy leakage. The standard composition-based privacy analysis of DP-SGD effectively assumes that the adversary has access to all intermediate iterates, which is often unrealistic. However, this analysis remains the state of the art in practice. While our heuristic does not replace a rigorous privacy analysis, it illustrates the large gap between the best theoretical upper bounds and the privacy auditing lower bounds and sets a target for further work to improve the theoretical privacy analyses. We also empirically support our heuristic and show existing privacy auditing attacks are bounded by our heuristic analysis in both vision and language tasks."
                    },
                    {
                        "title": "Near Exact Privacy Amplification for Matrix Mechanisms",
                        "abstract": "We study the problem of computing the privacy parameters for DP machine learning when using privacy amplification via random batching and noise correlated across rounds via a correlation matrix $\\textbf{C}$ (i.e., the matrix mechanism). Past work on this problem either only applied to banded $\\textbf{C}$, or gave loose privacy parameters. In this work, we give a framework for computing near-exact privacy parameters for any lower-triangular, non-negative $\\textbf{C}$. Our framework allows us to optimize the correlation matrix $\\textbf{C}$ while accounting for amplification, whereas past work could not. Empirically, we show this lets us achieve smaller RMSE on prefix sums than the previous state-of-the-art (SOTA). We also show that we can improve on the SOTA performance on deep learning tasks. Our two main technical tools are (i) using Monte Carlo accounting to bypass composition, which was the main technical challenge for past work, and (ii) a\"balls-in-bins\"batching scheme that enables easy privacy analysis and is closer to practical random batching than Poisson sampling."
                    },
                    {
                        "title": "Efficient and Near-Optimal Noise Generation for Streaming Differential Privacy",
                        "abstract": "In the task of differentially private (DP) continual counting, we receive a stream of increments and our goal is to output an approximate running total of these increments, without revealing too much about any specific increment. Despite its simplicity, differentially private continual counting has attracted significant attention both in theory and in practice. Existing algorithms for differentially private continual counting are either inefficient in terms of their space usage or add an excessive amount of noise, inducing suboptimal utility. The most practical DP continual counting algorithms add carefully correlated Gaussian noise to the values. The task of choosing the covariance for this noise can be expressed in terms of factoring the lower-triangular matrix of ones (which computes prefix sums). We present two approaches from this class (for different parameter regimes) that achieve near-optimal utility for DP continual counting and only require logarithmic or polylogarithmic space (and time). Our first approach is based on a space-efficient streaming matrix multiplication algorithm for a class of Toeplitz matrices. We show that to instantiate this algorithm for DP continual counting, it is sufficient to find a low-degree rational function that approximates the square root on a circle in the complex plane. We then apply and extend tools from approximation theory to achieve this. We also derive efficient closed-forms for the objective function for arbitrarily many steps, and show direct numerical optimization yields a highly practical solution to the problem. Our second approach combines our first approach with a recursive construction similar to the binary tree mechanism."
                    },
                    {
                        "title": "Avoiding Pitfalls for Privacy Accounting of Subsampled Mechanisms under Composition",
                        "abstract": "We consider the problem of computing tight privacy guarantees for the composition of subsampled differentially private mechanisms. Recent algorithms can numerically compute the privacy parameters to arbitrary precision but must be carefully applied. Our main contribution is to address two common points of confusion. First, some privacy accountants assume that the privacy guarantees for the composition of a subsampled mechanism are determined by self-composing the worst-case datasets for the uncomposed mechanism. We show that this is not true in general. Second, Poisson subsampling is sometimes assumed to have similar privacy guarantees compared to sampling without replacement. We show that the privacy guarantees may in fact differ significantly between the two sampling schemes. In particular, we give an example of hyperparameters that result in $\\varepsilon \\approx 1$ for Poisson subsampling and $\\varepsilon>10$ for sampling without replacement. This occurs for some parameters that could realistically be chosen for DP-SGD."
                    },
                    {
                        "title": "Stealing Part of a Production Language Model",
                        "abstract": "We introduce the first model-stealing attack that extracts precise, nontrivial information from black-box production language models like OpenAI's ChatGPT or Google's PaLM-2. Specifically, our attack recovers the embedding projection layer (up to symmetries) of a transformer model, given typical API access. For under \\$20 USD, our attack extracts the entire projection matrix of OpenAI's Ada and Babbage language models. We thereby confirm, for the first time, that these black-box models have a hidden dimension of 1024 and 2048, respectively. We also recover the exact hidden dimension size of the gpt-3.5-turbo model, and estimate it would cost under $2,000 in queries to recover the entire projection matrix. We conclude with potential defenses and mitigations, and discuss the implications of possible future work that could extend our attack."
                    },
                    {
                        "title": "Privacy Amplification for Matrix Mechanisms",
                        "abstract": "Privacy amplification exploits randomness in data selection to provide tighter differential privacy (DP) guarantees. This analysis is key to DP-SGD's success in machine learning, but, is not readily applicable to the newer state-of-the-art algorithms. This is because these algorithms, known as DP-FTRL, use the matrix mechanism to add correlated noise instead of independent noise as in DP-SGD. In this paper, we propose\"MMCC\", the first algorithm to analyze privacy amplification via sampling for any generic matrix mechanism. MMCC is nearly tight in that it approaches a lower bound as $\\epsilon\\to0$. To analyze correlated outputs in MMCC, we prove that they can be analyzed as if they were independent, by conditioning them on prior outputs. Our\"conditional composition theorem\"has broad utility: we use it to show that the noise added to binary-tree-DP-FTRL can asymptotically match the noise added to DP-SGD with amplification. Our amplification algorithm also has practical empirical utility: we show it leads to significant improvement in the privacy-utility trade-offs for DP-FTRL algorithms on standard benchmarks."
                    }
                ]
            },
            "fbde5bd7-ffe1-4584-b001-e0086fdb658e": {
                "pk": "fbde5bd7-ffe1-4584-b001-e0086fdb658e",
                "name": "Abhradeep Thakurta",
                "collaborators": [
                    "Arun Ganesh",
                    "Thomas Steinke",
                    "Christopher A. Choquette-Choo",
                    "Om Thakkar",
                    "Georgina Evans",
                    "Matthew Jagielski",
                    "Sewoong Oh",
                    "Sergei Vassilvitskii",
                    "Lun Wang",
                    "Arun Narayanan"
                ],
                "domain": [
                    "Differential Privacy",
                    "Machine Learning",
                    "Privacy-Preserving Algorithms",
                    "Speech Recognition"
                ],
                "publications": [
                    {
                        "title": "Differentially Private Parameter-Efficient Fine-tuning for Large ASR Models",
                        "abstract": "Large ASR models can inadvertently leak sensitive information, which can be mitigated by formal privacy measures like differential privacy (DP). However, traditional DP training is computationally expensive, and can hurt model performance. Our study explores DP parameter-efficient fine-tuning as a way to mitigate privacy risks with smaller computation and performance costs for ASR models. Through extensive experimentation and progressive optimization, we achieve 4.6%/8.1% word error rate on LibriSpeech clean/other test-sets, setting a new performance benchmark while maintaining (10, 3.52e-6)-DP in fine-tuning a large ASR model with over 600M parameters."
                    },
                    {
                        "title": "Differentially private survey research",
                        "abstract": "Survey researchers have long protected respondent privacy via de\u2010identification (removing names and other directly identifying information) before sharing data. Unfortunately, recent research demonstrates that these procedures fail to protect respondents from intentional re\u2010identification attacks, a problem that threatens to undermine vast survey enterprises in academia, government, and industry. This is especially a problem in political science because political beliefs are not merely the subject of our scholarship; they represent some of the most important information respondents want to keep private. We confirm the problem in practice by re\u2010identifying individuals from a survey about a controversial referendum declaring life beginning at conception. We build on the concept of \u201cdifferential privacy\u201d to offer new data\u2010sharing procedures with mathematical guarantees for protecting respondent privacy and statistical validity guarantees for social scientists analyzing differentially private data. The cost of these procedures is larger standard errors, which can be overcome with larger sample\u00a0sizes."
                    },
                    {
                        "title": "The Last Iterate Advantage: Empirical Auditing and Principled Heuristic Analysis of Differentially Private SGD",
                        "abstract": "We propose a simple heuristic privacy analysis of noisy clipped stochastic gradient descent (DP-SGD) in the setting where only the last iterate is released and the intermediate iterates remain hidden. Namely, our heuristic assumes a linear structure for the model. We show experimentally that our heuristic is predictive of the outcome of privacy auditing applied to various training procedures. Thus it can be used prior to training as a rough estimate of the final privacy leakage. We also probe the limitations of our heuristic by providing some artificial counterexamples where it underestimates the privacy leakage. The standard composition-based privacy analysis of DP-SGD effectively assumes that the adversary has access to all intermediate iterates, which is often unrealistic. However, this analysis remains the state of the art in practice. While our heuristic does not replace a rigorous privacy analysis, it illustrates the large gap between the best theoretical upper bounds and the privacy auditing lower bounds and sets a target for further work to improve the theoretical privacy analyses. We also empirically support our heuristic and show existing privacy auditing attacks are bounded by our heuristic analysis in both vision and language tasks."
                    },
                    {
                        "title": "Private Gradient Descent for Linear Regression: Tighter Error Bounds and Instance-Specific Uncertainty Estimation",
                        "abstract": "We provide an improved analysis of standard differentially private gradient descent for linear regression under the squared error loss. Under modest assumptions on the input, we characterize the distribution of the iterate at each time step. Our analysis leads to new results on the algorithm's accuracy: for a proper fixed choice of hyperparameters, the sample complexity depends only linearly on the dimension of the data. This matches the dimension-dependence of the (non-private) ordinary least squares estimator as well as that of recent private algorithms that rely on sophisticated adaptive gradient-clipping schemes (Varshney et al., 2022; Liu et al., 2023). Our analysis of the iterates' distribution also allows us to construct confidence intervals for the empirical optimizer which adapt automatically to the variance of the algorithm on a particular data set. We validate our theorems through experiments on synthetic data."
                    },
                    {
                        "title": "Near Exact Privacy Amplification for Matrix Mechanisms",
                        "abstract": "We study the problem of computing the privacy parameters for DP machine learning when using privacy amplification via random batching and noise correlated across rounds via a correlation matrix $\\textbf{C}$ (i.e., the matrix mechanism). Past work on this problem either only applied to banded $\\textbf{C}$, or gave loose privacy parameters. In this work, we give a framework for computing near-exact privacy parameters for any lower-triangular, non-negative $\\textbf{C}$. Our framework allows us to optimize the correlation matrix $\\textbf{C}$ while accounting for amplification, whereas past work could not. Empirically, we show this lets us achieve smaller RMSE on prefix sums than the previous state-of-the-art (SOTA). We also show that we can improve on the SOTA performance on deep learning tasks. Our two main technical tools are (i) using Monte Carlo accounting to bypass composition, which was the main technical challenge for past work, and (ii) a\"balls-in-bins\"batching scheme that enables easy privacy analysis and is closer to practical random batching than Poisson sampling."
                    },
                    {
                        "title": "Efficient and Near-Optimal Noise Generation for Streaming Differential Privacy",
                        "abstract": "In the task of differentially private (DP) continual counting, we receive a stream of increments and our goal is to output an approximate running total of these increments, without revealing too much about any specific increment. Despite its simplicity, differentially private continual counting has attracted significant attention both in theory and in practice. Existing algorithms for differentially private continual counting are either inefficient in terms of their space usage or add an excessive amount of noise, inducing suboptimal utility. The most practical DP continual counting algorithms add carefully correlated Gaussian noise to the values. The task of choosing the covariance for this noise can be expressed in terms of factoring the lower-triangular matrix of ones (which computes prefix sums). We present two approaches from this class (for different parameter regimes) that achieve near-optimal utility for DP continual counting and only require logarithmic or polylogarithmic space (and time). Our first approach is based on a space-efficient streaming matrix multiplication algorithm for a class of Toeplitz matrices. We show that to instantiate this algorithm for DP continual counting, it is sufficient to find a low-degree rational function that approximates the square root on a circle in the complex plane. We then apply and extend tools from approximation theory to achieve this. We also derive efficient closed-forms for the objective function for arbitrarily many steps, and show direct numerical optimization yields a highly practical solution to the problem. Our second approach combines our first approach with a recursive construction similar to the binary tree mechanism."
                    },
                    {
                        "title": "Optimal Rates for $O(1)$-Smooth DP-SCO with a Single Epoch and Large Batches",
                        "abstract": "In this paper we revisit the DP stochastic convex optimization (SCO) problem. For convex smooth losses, it is well-known that the canonical DP-SGD (stochastic gradient descent) achieves the optimal rate of $O\\left(\\frac{LR}{\\sqrt{n}} + \\frac{LR \\sqrt{p \\log(1/\\delta)}}{\\epsilon n}\\right)$ under $(\\epsilon, \\delta)$-DP, and also well-known that variants of DP-SGD can achieve the optimal rate in a single epoch. However, the batch gradient complexity (i.e., number of adaptive optimization steps), which is important in applications like federated learning, is less well-understood. In particular, all prior work on DP-SCO requires $\\Omega(n)$ batch gradient steps, multiple epochs, or convexity for privacy. We propose an algorithm, Accelerated-DP-SRGD (stochastic recursive gradient descent), which bypasses the limitations of past work: it achieves the optimal rate for DP-SCO (up to polylog factors), in a single epoch using $\\sqrt{n}$ batch gradient steps with batch size $\\sqrt{n}$, and can be made private for arbitrary (non-convex) losses via clipping. If the global minimizer is in the constraint set, we can further improve this to $n^{1/4}$ batch gradient steps with batch size $n^{3/4}$. To achieve this, our algorithm combines three key ingredients, a variant of stochastic recursive gradients (SRG), accelerated gradient descent, and correlated noise generation from DP continual counting."
                    },
                    {
                        "title": "Training Large ASR Encoders with Differential Privacy",
                        "abstract": "Self-supervised learning (SSL) methods for large speech models have proven to be highly effective at ASR. With the interest in public deployment of large pre-trained models, there is a rising concern for unintended memorization and leakage of sensitive data points from the training data. In this paper, we apply differentially private (DP) pre-training to a SOTA Conformer-based encoder, and study its performance on a downstream ASR task assuming the fine-tuning data is public. This paper is the first to apply DP to SSL for ASR, investigating the DP noise tolerance of the BEST-RQ pre-training method. Notably, we introduce a novel variant of model pruning called gradient-based layer freezing that provides strong improvements in privacy-utility-compute trade-offs. Our approach yields a LibriSpeech test-clean/other WER (%) of 3.78/ 8.41 with ($10$, 1e^-9)-DP for extrapolation towards low dataset scales, and 2.81/ 5.89 with (10, 7.9e^-11)-DP for extrapolation towards high scales."
                    },
                    {
                        "title": "Advancing Differential Privacy: Where We Are Now and Future Directions for Real-World Deployment",
                        "abstract": "In this article, we present a detailed review of current practices and state-of-the-art methodologies in the field of differential privacy (DP), with a focus of advancing DP's deployment in real-world applications. Key points and high-level contents of the article were originated from the discussions from\"Differential Privacy (DP): Challenges Towards the Next Frontier,\"a workshop held in July 2022 with experts from industry, academia, and the public sector seeking answers to broad questions pertaining to privacy and its implications in the design of industry-grade systems. This article aims to provide a reference point for the algorithmic and design decisions within the realm of privacy, highlighting important challenges and potential research directions. Covering a wide spectrum of topics, this article delves into the infrastructure needs for designing private systems, methods for achieving better privacy/utility trade-offs, performing privacy attacks and auditing, as well as communicating privacy with broader audiences and stakeholders."
                    },
                    {
                        "title": "Differentially Private and Lazy Online Convex Optimization",
                        "abstract": "We study the task of differentially private online convex optimization (OCO). In the online setting, the release of each distinct decision or iterate carries with it the potential for privacy loss. To limit such privacy leakage, we design an optimization-based OCO algorithm that explicitly limits the number of switches via objective perturbation and rejection sampling. This improves over known results in multiple aspects: an optimal leading-order regret term, in being efficiently implementable without requiring log-concave sampling subroutines, and in matching the non-private regret bound for sub-constant regimes of privacy parameters. Leveraging the fact that the algorithm is designed to explicitly minimize the number of switches of decisions, we show that the algorithm also obtains optimal regret bounds in the lazy OCO setting, where the learner is constrained to perform a limited number of switches. In addition, for one-and two-dimensional decision sets, we present a novel approach for differentially private online Lipschitz learning, where the loss functions are Lipschitz but not necessarily convex, that achieves the optimal regret bound matching known lower bounds."
                    },
                    {
                        "title": "Universality of Langevin Diffusion for Private Optimization, with Applications to Sampling from Rashomon Sets",
                        "abstract": "In this paper we provide an algorithmic framework based on Langevin diffusion (LD) and its corresponding discretizations that allow us to simultaneously obtain: i) An algorithm for sampling from the exponential mechanism (McSherry and Talwar, 2007), whose privacy analysis does not depend on convexity and which can be stopped at anytime without compromising privacy, and ii) tight uniform stability guarantees for the exponential mechanism. As a direct consequence, we obtain optimal excess empirical and population risk guarantees for (strongly) convex losses under both pure and approximate differential privacy (DP). The framework allows us to design a DP uniform sampler from the Rashomon set. Rashomon sets are widely used in interpretable and robust machine learning, understanding variable importance, and characterizing fairness."
                    },
                    {
                        "title": "Statistically Valid Inferences from Privacy-Protected Data",
                        "abstract": "Unprecedented quantities of data that could help social scientists understand and ameliorate the challenges of human society are presently locked away inside companies, governments, and other organizations, in part because of privacy concerns. We address this problem with a general-purpose data access and analysis system with mathematical guarantees of privacy for research subjects, and statistical validity guarantees for researchers seeking social science insights. We build on the standard of \u201cdifferential privacy,\u201d correct for biases induced by the privacy-preserving procedures, provide a proper accounting of uncertainty, and impose minimal constraints on the choice of statistical methods and quantities estimated. We illustrate by replicating key analyses from two recent published articles and show how we can obtain approximately the same substantive results while simultaneously protecting privacy. Our approach is simple to use and computationally efficient; we also offer open-source software that implements all our methods."
                    },
                    {
                        "title": "Challenges towards the Next Frontier in Privacy",
                        "abstract": ","
                    },
                    {
                        "title": "How to DP-fy ML: A Practical Guide to Machine Learning with Differential Privacy",
                        "abstract": "Machine Learning (ML) models are ubiquitous in real-world applications and are a constant focus of research. Modern ML models have become more complex, deeper, and harder to reason about. At the same time, the community has started to realize the importance of protecting the privacy of the training data that goes into these models. Differential Privacy (DP) has become a gold standard for making formal statements about data anonymization. However, while some adoption of DP has happened in industry, attempts to apply DP to real world complex ML models are still few and far between. The adoption of DP is hindered by limited practical guidance of what DP protection entails, what privacy guarantees to aim for, and the difficulty of achieving good privacy-utility-computation trade-offs for ML models. Tricks for tuning and maximizing performance are scattered among papers or stored in the heads of practitioners, particularly with respect to the challenging task of hyperparameter tuning. Furthermore, the literature seems to present conflicting evidence on how and whether to apply architectural adjustments and which components are \u201csafe\u201d to use with DP. In this survey paper, we attempt to create a self-contained guide that gives an in-depth overview of the field of DP ML. We aim to assemble information about achieving the best possible DP ML model with rigorous privacy guarantees. Our target audience is both researchers and practitioners. Researchers interested in DP for ML will benefit from a clear overview of current advances and areas for improvement. We also include theory-focused sections that highlight important topics such as privacy accounting and convergence. For a practitioner, this survey provides a background in DP theory and a clear step-by-step guide for choosing an appropriate privacy definition and approach, implementing DP training, potentially updating the model architecture, and tuning hyperparameters. For both researchers and practitioners, consistently and fully reporting privacy guarantees is critical, so we propose a set of specific best practices for stating guarantees. With sufficient computation and a sufficiently large training set or supplemental nonprivate data, both good accuracy (that is, almost as good as a non-private model) and good privacy can often be achievable. And even when computation and dataset size are limited, there are advantages to training with even a weak (but still finite) formal DP guarantee. Hence, we hope this work will facilitate more widespread deployments of DP ML models."
                    },
                    {
                        "title": "Why Is Public Pretraining Necessary for Private Model Training?",
                        "abstract": "In the privacy-utility tradeoff of a model trained on benchmark language and vision tasks, remarkable improvements have been widely reported with the use of pretraining on publicly available data. This is in part due to the benefits of transfer learning, which is the standard motivation for pretraining in non-private settings. However, the stark contrast in the improvement achieved through pretraining under privacy compared to non-private settings suggests that there may be a deeper, distinct cause driving these gains. To explain this phenomenon, we hypothesize that the non-convex loss landscape of a model training necessitates an optimization algorithm to go through two phases. In the first, the algorithm needs to select a good\"basin\"in the loss landscape. In the second, the algorithm solves an easy optimization within that basin. The former is a harder problem to solve with private data, while the latter is harder to solve with public data due to a distribution shift or data scarcity. Guided by this intuition, we provide theoretical constructions that provably demonstrate the separation between private training with and without public pretraining. Further, systematic experiments on CIFAR10 and LibriSpeech provide supporting evidence for our hypothesis."
                    },
                    {
                        "title": "Privacy Amplification for Matrix Mechanisms",
                        "abstract": "Privacy amplification exploits randomness in data selection to provide tighter differential privacy (DP) guarantees. This analysis is key to DP-SGD's success in machine learning, but, is not readily applicable to the newer state-of-the-art algorithms. This is because these algorithms, known as DP-FTRL, use the matrix mechanism to add correlated noise instead of independent noise as in DP-SGD. In this paper, we propose\"MMCC\", the first algorithm to analyze privacy amplification via sampling for any generic matrix mechanism. MMCC is nearly tight in that it approaches a lower bound as $\\epsilon\\to0$. To analyze correlated outputs in MMCC, we prove that they can be analyzed as if they were independent, by conditioning them on prior outputs. Our\"conditional composition theorem\"has broad utility: we use it to show that the noise added to binary-tree-DP-FTRL can asymptotically match the noise added to DP-SGD with amplification. Our amplification algorithm also has practical empirical utility: we show it leads to significant improvement in the privacy-utility trade-offs for DP-FTRL algorithms on standard benchmarks."
                    }
                ]
            }
        }
    },
    "2405.13800": {
        "paper_data": {
            "title": "Dense Connector for MLLMs",
            "url": "http://arxiv.org/abs/2405.13800v1",
            "arxiv_id": "2405.13800",
            "authors": [
                "Huanjin Yao",
                "Wenhao Wu",
                "Taojiannan Yang",
                "YuXin Song",
                "Mengxi Zhang",
                "Haocheng Feng",
                "Yifan Sun",
                "Zhiheng Li",
                "Wanli Ouyang",
                "Jingdong Wang"
            ],
            "abstract": "Do we fully leverage the potential of visual encoder in Multimodal Large Language Models (MLLMs)? The recent outstanding performance of MLLMs in multimodal understanding has garnered broad attention from both academia and industry. In the current MLLM rat race, the focus seems to be predominantly on the linguistic side. We witness the rise of larger and higher-quality instruction datasets, as well as the involvement of larger-sized LLMs. Yet, scant attention has been directed towards the visual signals utilized by MLLMs, often assumed to be the final high-level features extracted by a frozen visual encoder. In this paper, we introduce the Dense Connector - a simple, effective, and plug-and-play vision-language connector that significantly enhances existing MLLMs by leveraging multi-layer visual features, with minimal additional computational overhead. Furthermore, our model, trained solely on images, showcases remarkable zero-shot capabilities in video understanding as well. Experimental results across various vision encoders, image resolutions, training dataset scales, varying sizes of LLMs (2.7B->70B), and diverse architectures of MLLMs (e.g., LLaVA and Mini-Gemini) validate the versatility and scalability of our approach, achieving state-of-the-art performance on across 19 image and video benchmarks. We hope that this work will provide valuable experience and serve as a basic module for future MLLM development.",
            "introduction": "   1 Introduction  In recent years, Large Language Models (LLMs) led by ChatGPT\u00a0[1] have made remarkable advancements in text comprehension and generation. Furthermore, cutting-edge Multimodal Large Language Models (MLLMs)\u00a0[2, 3] have rapidly expanded the capabilities of LLMs to include visual understanding, evolving into models capable of integrating both vision and text modalities. This has elevated MLLMs to become a new focal point for research and discussion\u00a0[4, 5, 6, 7].   In broad terms, the architecture of existing MLLMs can be delineated into three components: the pre-trained vision encoder (e.g., CLIP\u2019s ViT-L\u00a0[8] or EVA-CLIP\u2019s ViT-G\u00a0[9]), the pre-trained LLM (e.g., OPT\u00a0[10], Llama\u00a0[11], Vicuna\u00a0[12], etc.), and the connector (e.g., Q-former\u00a0[13, 14] or linear projection\u00a0[15, 16]) trained from scratch to bridge the vision and language models. An intriguing trend in current MLLM research is that the focus of model learning and performance improvement seems to primarily center around the language aspect (e.g., utilizing larger-scale and higher-quality visual instruction data\u00a0[17, 16, 18], larger-sized LLMs\u00a0[19, 20]), with less exploration into the visual signals fed into the connector. Typically, the visual encoder is frozen to extract high-level visual features, which are then fed into the connector. This leads us to rethink: Have we fully utilized the existing pre-trained visual encoder?   In addition to the common practice of feeding the connector with final high-level visual features from visual encoder, an intuitive yet overlooked idea is to integrate visual features from various layers to complement the high-level features. In \\creffig1(a), we illustrate attention maps from different layers of a 24-layer CLIP\u00a0[8] pre-trained ViT-L\u00a0[21], showing that different layers of the same visual encoder emphasize different regions of interest. Moreover, looking back at the history of computer vision, classic models (e.g., Densenet\u00a0[22], FPN\u00a0[23]) utilize multi-layer features to enhance visual representations. In MLLMs, the vision encoder is typically frozen to mitigate significant computational costs. In this context, our idea leverages the \u201cfree lunch\u201d of utilizing offline features from different layers as an implicit enhancement of visual information without the need for additional computational overhead. Furthermore, this way also complements techniques that directly increase visual signals, e.g., increasing image resolution [18, 24, 25, 26, 27, 28] or introducing additional visual encoders [29, 30, 18]. This idea is both simple and efficient, while also being sufficiently generic, logically allowing for seamless integration with any existing MLLMs.   In light of this, we propose the Dense Connector (DC), serving as a plug-and-play vision-language connector that involves offline features from various layers of the frozen visual encoder to provide the LLM with more visual cues. We explore three intuitive instantiations for the Dense Connector: 1) Sparse Token Integration (STI): We explicitly consider increasing the number of visual tokens by aggregating visual tokens from different specified layers and the final visual token. These tokens are then fed into a learnable projector for mapping into the text space. 2) Sparse Channel Integration (SCI): To avoid increasing the number of tokens, we concatenate visual tokens from different specified layers in the feature dimension. They are then passed to the projector, which not only maps visual tokens into the text space but also serves to reduce the feature dimensionality. 3) Dense Channel Integration (DCI):",
            "references": [
                {
                    "title": "FreeVA: Offline MLLM as Training-Free Video Assistant",
                    "abstract": "This paper undertakes an empirical study to revisit the latest advancements in Multimodal Large Language Models (MLLMs): Video Assistant. This study, namely FreeVA, aims to extend existing image-based MLLM to the video domain in a training-free manner. The study provides an essential, yet must-know baseline, and reveals several surprising findings: 1) FreeVA, leveraging only offline image-based MLLM without additional training, excels in zero-shot video question-answering (e.g., MSVD-QA, ActivityNet-QA, and MSRVTT-QA), even surpassing state-of-the-art methods that involve video instruction tuning. 2) While mainstream video-based MLLMs typically initialize with an image-based MLLM (e.g., LLaVA) and then fine-tune using video instruction tuning, the study indicates that utilizing the widely adopted VideoInstruct-100K for video instruction tuning doesn't actually lead to better performance compared to not training at all. 3) The commonly used evaluation metrics in existing works are significantly influenced by changes in the GPT API version over time. If ignored, this could affect the fairness and uniformity of comparisons between different methods and impact the analysis and judgment of researchers in the field. The advancement of MLLMs is currently thriving, drawing numerous researchers into the field. We aim for this work to serve as a plug-and-play, simple yet effective baseline, encouraging the direct evaluation of existing MLLMs in video domain while also standardizing the field of video conversational models to a certain extent. Also, we encourage researchers to reconsider: Have current video MLLM methods truly acquired knowledge beyond image MLLM? Code is available at https://github.com/whwu95/FreeVA"
                },
                {
                    "title": "CuMo: Scaling Multimodal LLM with Co-Upcycled Mixture-of-Experts",
                    "abstract": "Recent advancements in Multimodal Large Language Models (LLMs) have focused primarily on scaling by increasing text-image pair data and enhancing LLMs to improve performance on multimodal tasks. However, these scaling approaches are computationally expensive and overlook the significance of improving model capabilities from the vision side. Inspired by the successful applications of Mixture-of-Experts (MoE) in LLMs, which improves model scalability during training while keeping inference costs similar to those of smaller models, we propose CuMo. CuMo incorporates Co-upcycled Top-K sparsely-gated Mixture-of-experts blocks into both the vision encoder and the MLP connector, thereby enhancing the multimodal LLMs with minimal additional activated parameters during inference. CuMo first pre-trains the MLP blocks and then initializes each expert in the MoE block from the pre-trained MLP block during the visual instruction tuning stage. Auxiliary losses are used to ensure a balanced loading of experts. CuMo outperforms state-of-the-art multimodal LLMs across various VQA and visual-instruction-following benchmarks using models within each model size group, all while training exclusively on open-sourced datasets. The code and model weights for CuMo are open-sourced at https://github.com/SHI-Labs/CuMo."
                },
                {
                    "title": "Mini-Gemini: Mining the Potential of Multi-modality Vision Language Models",
                    "abstract": "In this work, we introduce Mini-Gemini, a simple and effective framework enhancing multi-modality Vision Language Models (VLMs). Despite the advancements in VLMs facilitating basic visual dialog and reasoning, a performance gap persists compared to advanced models like GPT-4 and Gemini. We try to narrow the gap by mining the potential of VLMs for better performance and any-to-any workflow from three aspects, i.e., high-resolution visual tokens, high-quality data, and VLM-guided generation. To enhance visual tokens, we propose to utilize an additional visual encoder for high-resolution refinement without increasing the visual token count. We further construct a high-quality dataset that promotes precise image comprehension and reasoning-based generation, expanding the operational scope of current VLMs. In general, Mini-Gemini further mines the potential of VLMs and empowers current frameworks with image understanding, reasoning, and generation simultaneously. Mini-Gemini supports a series of dense and MoE Large Language Models (LLMs) from 2B to 34B. It is demonstrated to achieve leading performance in several zero-shot benchmarks and even surpasses the developed private models. Code and models are available at https://github.com/dvlab-research/MiniGemini."
                },
                {
                    "title": "MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training",
                    "abstract": "In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting."
                },
                {
                    "title": "Yi: Open Foundation Models by 01.AI",
                    "abstract": "We introduce the Yi model family, a series of language and multimodal models that demonstrate strong multi-dimensional capabilities. The Yi model family is based on 6B and 34B pretrained language models, then we extend them to chat models, 200K long context models, depth-upscaled models, and vision-language models. Our base models achieve strong performance on a wide range of benchmarks like MMLU, and our finetuned chat models deliver strong human preference rate on major evaluation platforms like AlpacaEval and Chatbot Arena. Building upon our scalable super-computing infrastructure and the classical transformer architecture, we attribute the performance of Yi models primarily to its data quality resulting from our data-engineering efforts. For pretraining, we construct 3.1 trillion tokens of English and Chinese corpora using a cascaded data deduplication and quality filtering pipeline. For finetuning, we polish a small scale (less than 10K) instruction dataset over multiple iterations such that every single instance has been verified directly by our machine learning engineers. For vision-language, we combine the chat language model with a vision transformer encoder and train the model to align visual representations to the semantic space of the language model. We further extend the context length to 200K through lightweight continual pretraining and demonstrate strong needle-in-a-haystack retrieval performance. We show that extending the depth of the pretrained checkpoint through continual pretraining further improves performance. We believe that given our current results, continuing to scale up model parameters using thoroughly optimized data will lead to even stronger frontier models."
                },
                {
                    "title": "Stable LM 2 1.6B Technical Report",
                    "abstract": "We introduce StableLM 2 1.6B, the first in a new generation of our language model series. In this technical report, we present in detail the data and training procedure leading to the base and instruction-tuned versions of StableLM 2 1.6B. The weights for both models are available via Hugging Face for anyone to download and use. The report contains thorough evaluations of these models, including zero- and few-shot benchmarks, multilingual benchmarks, and the MT benchmark focusing on multi-turn dialogues. At the time of publishing this report, StableLM 2 1.6B was the state-of-the-art open model under 2B parameters by a significant margin. Given its appealing small size, we also provide throughput measurements on a number of edge devices. In addition, we open source several quantized checkpoints and provide their performance metrics compared to the original model."
                },
                {
                    "title": "TinyLLaVA: A Framework of Small-scale Large Multimodal Models",
                    "abstract": "We present the TinyLLaVA framework that provides a unified perspective in designing and analyzing the small-scale Large Multimodal Models (LMMs). We empirically study the effects of different vision encoders, connection modules, language models, training data and training recipes. Our extensive experiments showed that better quality of data combined with better training recipes, smaller LMMs can consistently achieve on-par performances compared to bigger LMMs. Under our framework, we train a family of small-scale LMMs. Our best model, TinyLLaVA-3.1B, achieves better overall performance against existing 7B models such as LLaVA-1.5 and Qwen-VL. We hope our findings can serve as baselines for future research in terms of data scaling, training setups and model selections. Our model weights and codes will be made public."
                },
                {
                    "title": "SPHINX-X: Scaling Data and Parameters for a Family of Multi-modal Large Language Models",
                    "abstract": "We propose SPHINX-X, an extensive Multimodality Large Language Model (MLLM) series developed upon SPHINX. To improve the architecture and training efficiency, we modify the SPHINX framework by removing redundant visual encoders, bypassing fully-padded sub-images with skip tokens, and simplifying multi-stage training into a one-stage all-in-one paradigm. To fully unleash the potential of MLLMs, we assemble a comprehensive multi-domain and multimodal dataset covering publicly available resources in language, vision, and vision-language tasks. We further enrich this collection with our curated OCR intensive and Set-of-Mark datasets, extending the diversity and generality. By training over different base LLMs including TinyLlama1.1B, InternLM2-7B, LLaMA2-13B, and Mixtral8x7B, we obtain a spectrum of MLLMs that vary in parameter size and multilingual capabilities. Comprehensive benchmarking reveals a strong correlation between the multi-modal performance with the data and parameter scales. Code and models are released at https://github.com/Alpha-VLLM/LLaMA2-Accessory"
                },
                {
                    "title": "MobileVLM V2: Faster and Stronger Baseline for Vision Language Model",
                    "abstract": "We introduce MobileVLM V2, a family of significantly improved vision language models upon MobileVLM, which proves that a delicate orchestration of novel architectural design, an improved training scheme tailored for mobile VLMs, and rich high-quality dataset curation can substantially benefit VLMs' performance. Specifically, MobileVLM V2 1.7B achieves better or on-par performance on standard VLM benchmarks compared with much larger VLMs at the 3B scale. Notably, our 3B model outperforms a large variety of VLMs at the 7B+ scale. Our models will be released at https://github.com/Meituan-AutoML/MobileVLM ."
                },
                {
                    "title": "Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs",
                    "abstract": "Is vision good enough for language? Recent advancements in multimodal models primarily stem from the powerful reasoning abilities of large language models (LLMs). However, the visual component typically depends only on the instance-level contrastive language-image pre-training (CLIP). Our research reveals that the visual capabilities in recent MultiModal LLMs (MLLMs) still exhibit systematic shortcomings. To understand the roots of these errors, we explore the gap between the visual embedding space of CLIP and vision-only self-supervised learning. We identify \u201cCLIP-blind pairs\u201d- images that CLIP perceives as similar despite their clear visual differences. With these pairs, we construct the Multimodal Visual Patterns (MMVP) benchmark. MMVP exposes areas where state-of-the-art systems, including GPT-4V, struggle with straightforward questions across nine basic visual patterns, often providing incorrect answers and hallucinated explanations. We further evaluate various CLIP-based vision-and-language models and found a notable correlation between visual patterns that challenge CLIP models and those problematic for multimodal LLMs. As an initial effort to address these issues, we propose a Mixture of Features (MoF) approach, demonstrating that integrating vision self-supervised learning features with MLLMs can significantly enhance their visual grounding capabilities. Together, our research suggests visual representation learning remains an open challenge, and accurate visual grounding is crucial for future successful multimodal systems."
                },
                {
                    "title": "Mixtral of Experts",
                    "abstract": "We introduce Mixtral 8x7B, a Sparse Mixture of Experts (SMoE) language model. Mixtral has the same architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks (i.e. experts). For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, each token has access to 47B parameters, but only uses 13B active parameters during inference. Mixtral was trained with a context size of 32k tokens and it outperforms or matches Llama 2 70B and GPT-3.5 across all evaluated benchmarks. In particular, Mixtral vastly outperforms Llama 2 70B on mathematics, code generation, and multilingual benchmarks. We also provide a model fine-tuned to follow instructions, Mixtral 8x7B - Instruct, that surpasses GPT-3.5 Turbo, Claude-2.1, Gemini Pro, and Llama 2 70B - chat model on human benchmarks. Both the base and instruct models are released under the Apache 2.0 license."
                },
                {
                    "title": "TinyLlama: An Open-Source Small Language Model",
                    "abstract": "We present TinyLlama, a compact 1.1B language model pretrained on around 1 trillion tokens for approximately 3 epochs. Building on the architecture and tokenizer of Llama 2, TinyLlama leverages various advances contributed by the open-source community (e.g., FlashAttention and Lit-GPT), achieving better computational efficiency. Despite its relatively small size, TinyLlama demonstrates remarkable performance in a series of downstream tasks. It significantly outperforms existing open-source language models with comparable sizes. Our model checkpoints and code are publicly available on GitHub at https://github.com/jzhang38/TinyLlama."
                },
                {
                    "title": "VILA: On Pre-training for Visual Language Models",
                    "abstract": "Visual language models (VLMs) rapidly progressed with the recent success of large language models. There have been growing efforts on visual instruction tuning to extend the LLM with visual inputs, but lacks an in-depth study of the visual language pre-training process, where the model learns to perform joint modeling on both modalities. In this work, we examine the design options for VLM pre-training by augmenting LLM towards VLM through step-by-step controllable comparisons. We introduce three main findings: (1) freezing LLMs during pre-training can achieve decent zero-shot performance, but lack in-context learning capability, which requires unfreezing the LLM; (2) interleaved pre-training data is beneficial whereas image-text pairs alone are not optimal; (3) re-blending text-only instruction data to image-text data during instruction fine-tuning not only remedies the degradation of text-only tasks, but also boosts VLM task accuracy. With an enhanced pre-training recipe we build VILA, a Visual Language model family that consistently outperforms the state-of-the-art models, e.g., LLaVA-1.5, across main benchmarks without bells and whistles. Multi-modal pre-training also helps unveil appealing properties of VILA, including multi-image reasoning, enhanced in-context learning, and better world knowledge. VILA is also deployable on Jetson Orin for on-device VLM."
                },
                {
                    "title": "VaQuitA: Enhancing Alignment in LLM-Assisted Video Understanding",
                    "abstract": "Recent advancements in language-model-based video understanding have been progressing at a remarkable pace, spurred by the introduction of Large Language Models (LLMs). However, the focus of prior research has been predominantly on devising a projection layer that maps video features to tokens, an approach that is both rudimentary and inefficient. In our study, we introduce a cutting-edge framework, VaQuitA, designed to refine the synergy between video and textual information. At the data level, instead of sampling frames uniformly, we implement a sampling method guided by CLIP-score rankings, which enables a more aligned selection of frames with the given question. At the feature level, we integrate a trainable Video Perceiver alongside a Visual-Query Transformer (abbreviated as VQ-Former), which bolsters the interplay between the input question and the video features. We also discover that incorporating a simple prompt,\"Please be critical\", into the LLM input can substantially enhance its video comprehension capabilities. Our experimental results indicate that VaQuitA consistently sets a new benchmark for zero-shot video question-answering tasks and is adept at producing high-quality, multi-turn video dialogues with users."
                },
                {
                    "title": "LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models",
                    "abstract": "In this work, we present a novel method to tackle the token generation challenge in Vision Language Models (VLMs) for video and image understanding, called LLaMA-VID. Current VLMs, while proficient in tasks like image captioning and visual question answering, face computational burdens when processing long videos due to the excessive visual tokens. LLaMA-VID addresses this issue by representing each frame with two distinct tokens, namely context token and content token. The context token encodes the overall image context based on user input, whereas the content token encapsulates visual cues in each frame. This dual-token strategy significantly reduces the overload of long videos while preserving critical information. Generally, LLaMA-VID empowers existing frameworks to support hour-long videos and pushes their upper limit with an extra context token. It is proved to surpass previous methods on most of video- or image-based benchmarks. Code is available https://github.com/dvlab-research/LLaMA-VID}{https://github.com/dvlab-research/LLaMA-VID"
                },
                {
                    "title": "MMMU: A Massive Multi-Discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI",
                    "abstract": "We introduce MMMU: a new benchmark designed to evaluate multimodal models on massive multi-discipline tasks demanding college-level subject knowledge and deliberate reasoning. MMMU includes 11.5K meticulously collected multimodal questions from college exams, quizzes, and text-books, covering six core disciplines: Art & Design, Busi-ness, Science, Health & Medicine, Humanities & Social Science, and Tech & Engineering. These questions span 30 subjects and 183 subfields, comprising 30 highly het-erogeneous image types, such as charts, diagrams, maps, tables, music sheets, and chemical structures. Unlike existing benchmarks, MMMU focuses on advanced perception and reasoning with domain-specific knowledge, challenging models to perform tasks akin to those faced by experts. The evaluation of 28 open-source LMMs as well as the propri-etary GPT-4V(ision) and Gemini highlights the substantial challenges posed by MMMU. Even the advanced GPT-4V and Gemini Ultra only achieve accuracies of 56% and 59% respectively, indicating significant room for improvement. We believe MMMU will stimulate the community to build next-generation multimodal foundation models towards expert artificial general intelligence."
                },
                {
                    "title": "GPT4Vis: What Can GPT-4 Do for Zero-shot Visual Recognition?",
                    "abstract": "This paper does not present a novel method. Instead, it delves into an essential, yet must-know baseline in light of the latest advancements in Generative Artificial Intelligence (GenAI): the utilization of GPT-4 for visual understanding. Our study centers on the evaluation of GPT-4's linguistic and visual capabilities in zero-shot visual recognition tasks: Firstly, we explore the potential of its generated rich textual descriptions across various categories to enhance recognition performance without any training. Secondly, we evaluate GPT-4's visual proficiency in directly recognizing diverse visual content. We conducted extensive experiments to systematically evaluate GPT-4's performance across images, videos, and point clouds, using 16 benchmark datasets to measure top-1 and top-5 accuracy. Our findings show that GPT-4, enhanced with rich linguistic descriptions, significantly improves zero-shot recognition, offering an average top-1 accuracy increase of 7% across all datasets. GPT-4 excels in visual recognition, outshining OpenAI-CLIP's ViT-L and rivaling EVA-CLIP's ViT-E, particularly in video datasets HMDB-51 and UCF-101, where it leads by 22% and 9%, respectively. We hope this research contributes valuable data points and experience for future studies. We release our code at https://github.com/whwu95/GPT4Vis."
                },
                {
                    "title": "ShareGPT4V: Improving Large Multi-Modal Models with Better Captions",
                    "abstract": "In the realm of large multi-modal models (LMMs), efficient modality alignment is crucial yet often constrained by the scarcity of high-quality image-text data. To address this bottleneck, we introduce the ShareGPT4V dataset, a pioneering large-scale resource featuring 1.2 million highly descriptive captions, which surpasses existing datasets in diversity and information content, covering world knowledge, object properties, spatial relationships, and aesthetic evaluations. Specifically, ShareGPT4V originates from a curated 100K high-quality captions collected from advanced GPT4-Vision and has been expanded to 1.2M with a superb caption model trained on this subset. ShareGPT4V first demonstrates its effectiveness for the Supervised Fine-Tuning (SFT) phase, by substituting an equivalent quantity of detailed captions in existing SFT datasets with a subset of our high-quality captions, significantly enhancing the LMMs like LLaVA-7B, LLaVA-1.5-13B, and Qwen-VL-Chat-7B on the MME and MMBench benchmarks, with respective gains of 222.8/22.0/22.3 and 2.7/1.3/1.5. We further incorporate ShareGPT4V data into both the pre-training and SFT phases, obtaining ShareGPT4V-7B, a superior LMM based on a simple architecture that has remarkable performance across a majority of the multi-modal benchmarks. This project is available at https://ShareGPT4V.github.io to serve as a pivotal resource for advancing the LMMs community."
                },
                {
                    "title": "GPT-4V(ision) for Robotics: Multimodal Task Planning From Human Demonstration",
                    "abstract": "We introduce a pipeline that enhances a general-purpose Vision Language Model, GPT-4V(ision), to facilitate one-shot visual teaching for robotic manipulation. This system analyzes videos of humans performing tasks and outputs executable robot programs that incorporate insights into affordances. The process begins with GPT-4 V analyzing the videos to obtain textual explanations of environmental and action details. A GPT-4-based task planner then encodes these details into a symbolic task plan. Subsequently, vision systems spatially and temporally ground the task plan in the videos\u2014objects are identified using an open-vocabulary object detector, and hand-object interactions are analyzed to pinpoint moments of grasping and releasing. This spatiotemporal grounding allows for the gathering of affordance information (e.g., grasp types, waypoints, and body postures) critical for robot execution. Experiments across various scenarios demonstrate the method's efficacy in enabling real robots to operate from one-shot human demonstrations. Meanwhile, quantitative tests have revealed instances of hallucination in GPT-4 V, highlighting the importance of incorporating human supervision within the pipeline."
                },
                {
                    "title": "Video-LLaVA: Learning United Visual Representation by Alignment Before Projection",
                    "abstract": "The Large Vision-Language Model (LVLM) has enhanced the performance of various downstream tasks in visual-language understanding. Most existing approaches encode images and videos into separate feature spaces, which are then fed as inputs to large language models. However, due to the lack of unified tokenization for images and videos, namely misalignment before projection, it becomes challenging for a Large Language Model (LLM) to learn multi-modal interactions from several poor projection layers. In this work, we unify visual representation into the language feature space to advance the foundational LLM towards a unified LVLM. As a result, we establish a simple but robust LVLM baseline, Video-LLaVA, which learns from a mixed dataset of images and videos, mutually enhancing each other. Video-LLaVA achieves superior performances on a broad range of 9 image benchmarks across 5 image question-answering datasets and 4 image benchmark toolkits. Additionally, our Video-LLaVA also outperforms Video-ChatGPT by 5.8%, 9.9%, 18.6%, and 10.1% on MSRVTT, MSVD, TGIF, and ActivityNet, respectively. Notably, extensive experiments demonstrate that Video-LLaVA mutually benefits images and videos within a unified visual representation, outperforming models designed specifically for images or videos. We aim for this work to provide modest insights into the multi-modal inputs for the LLM. Code address: \\href{https://github.com/PKU-YuanGroup/Video-LLaVA}"
                },
                {
                    "title": "mPLUG-OwI2: Revolutionizing Multi-modal Large Language Model with Modality Collaboration",
                    "abstract": "Multi-modal Large Language Models (MLLMs) have demonstrated impressive instruction abilities across various open-ended tasks. However, previous methods primarily fo-cus on enhancing multi-modal capabilities. In this work, we introduce a versatile multi-modal large language model, mPLUG-Owl2, which effectively leverages modality collab-oration to improve performance in both text and multi-modal tasks. mPLUG-Owl2 utilizes a modularized network design, with the language decoder acting as a universal interface for managing different modalities. Specifically, mPLUG-Owl2 incorporates shared functional modules to facilitate modal-ity collaboration and introduces a modality-adaptive module that preserves modality-specific features. Extensive experi-ments reveal that mPLUG-Owl2 is capable of generalizing both text tasks and multi-modal tasks and achieving state-of-the-art performances with a single generic model. Notably, mPLUG-Owl2 is the first MLLM model that demonstrates the modality collaboration phenomenon in both pure-text and multi-modal scenarios, setting a pioneering path in the development of future multi-modal foundation models."
                },
                {
                    "title": "OtterHD: A High-Resolution Multi-modality Model",
                    "abstract": "In this paper, we present OtterHD-8B, an innovative multimodal model evolved from Fuyu-8B, specifically engineered to interpret high-resolution visual inputs with granular precision. Unlike conventional models that are constrained by fixed-size vision encoders, OtterHD-8B boasts the ability to handle flexible input dimensions, ensuring its versatility across various inference requirements. Alongside this model, we introduce MagnifierBench, an evaluation framework designed to scrutinize models' ability to discern minute details and spatial relationships of small objects. Our comparative analysis reveals that while current leading models falter on this benchmark, OtterHD-8B, particularly when directly processing high-resolution inputs, outperforms its counterparts by a substantial margin. The findings illuminate the structural variances in visual information processing among different models and the influence that the vision encoders' pre-training resolution disparities have on model effectiveness within such benchmarks. Our study highlights the critical role of flexibility and high-resolution input capabilities in large multimodal models and also exemplifies the potential inherent in the Fuyu architecture's simplicity for handling complex visual data."
                },
                {
                    "title": "CogVLM: Visual Expert for Pretrained Language Models",
                    "abstract": "We introduce CogVLM, a powerful open-source visual language foundation model. Different from the popular shallow alignment method which maps image features into the input space of language model, CogVLM bridges the gap between the frozen pretrained language model and image encoder by a trainable visual expert module in the attention and FFN layers. As a result, CogVLM enables deep fusion of vision language features without sacrificing any performance on NLP tasks. CogVLM-17B achieves state-of-the-art performance on 10 classic cross-modal benchmarks, including NoCaps, Flicker30k captioning, RefCOCO, RefCOCO+, RefCOCOg, Visual7W, GQA, ScienceQA, VizWiz VQA and TDIUC, and ranks the 2nd on VQAv2, OKVQA, TextVQA, COCO captioning, etc., surpassing or matching PaLI-X 55B. Codes and checkpoints are available at https://github.com/THUDM/CogVLM."
                },
                {
                    "title": "Can GPT-4V(ision) Serve Medical Applications? Case Studies on GPT-4V for Multimodal Medical Diagnosis",
                    "abstract": "Driven by the large foundation models, the development of artificial intelligence has witnessed tremendous progress lately, leading to a surge of general interest from the public. In this study, we aim to assess the performance of OpenAI's newest model, GPT-4V(ision), specifically in the realm of multimodal medical diagnosis. Our evaluation encompasses 17 human body systems, including Central Nervous System, Head and Neck, Cardiac, Chest, Hematology, Hepatobiliary, Gastrointestinal, Urogenital, Gynecology, Obstetrics, Breast, Musculoskeletal, Spine, Vascular, Oncology, Trauma, Pediatrics, with images taken from 8 modalities used in daily clinic routine, e.g., X-ray, Computed Tomography (CT), Magnetic Resonance Imaging (MRI), Positron Emission Tomography (PET), Digital Subtraction Angiography (DSA), Mammography, Ultrasound, and Pathology. We probe the GPT-4V's ability on multiple clinical tasks with or without patent history provided, including imaging modality and anatomy recognition, disease diagnosis, report generation, disease localisation. Our observation shows that, while GPT-4V demonstrates proficiency in distinguishing between medical image modalities and anatomy, it faces significant challenges in disease diagnosis and generating comprehensive reports. These findings underscore that while large multimodal models have made significant advancements in computer vision and natural language processing, it remains far from being used to effectively support real-world medical applications and clinical decision-making. All images used in this report can be found in https://github.com/chaoyi-wu/GPT-4V_Medical_Evaluation."
                },
                {
                    "title": "From CLIP to DINO: Visual Encoders Shout in Multi-modal Large Language Models",
                    "abstract": "Multi-modal Large Language Models (MLLMs) have made significant strides in expanding the capabilities of Large Language Models (LLMs) through the incorporation of visual perception interfaces. Despite the emergence of exciting applications and the availability of diverse instruction tuning data, existing approaches often rely on CLIP or its variants as the visual branch, and merely extract features from the deep layers. However, these methods lack a comprehensive analysis of the visual encoders in MLLMs. In this paper, we conduct an extensive investigation into the effectiveness of different vision encoders within MLLMs. Our findings reveal that the shallow layer features of CLIP offer particular advantages for fine-grained tasks such as grounding and region understanding. Surprisingly, the vision-only model DINO, which is not pretrained with text-image alignment, demonstrates promising performance as a visual branch within MLLMs. By simply equipping it with an MLP layer for alignment, DINO surpasses CLIP in fine-grained related perception tasks. Building upon these observations, we propose a simple yet effective feature merging strategy, named COMM, that integrates CLIP and DINO with Multi-level features Merging, to enhance the visual capabilities of MLLMs. We evaluate COMM through comprehensive experiments on a wide range of benchmarks, including image captioning, visual question answering, visual grounding, and object hallucination. Experimental results demonstrate the superior performance of COMM compared to existing methods, showcasing its enhanced visual capabilities within MLLMs."
                },
                {
                    "title": "Improved Baselines with Visual Instruction Tuning",
                    "abstract": "Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this paper, we present the first systematic study to investigate the design choices of LMMs in a controlled setting under the LLaVA framework. We show that the fully-connected vision-language connector in LLaVA is surprisingly power-ful and data-efficient. With simple modifications to LLa VA, namely, using CLIP- ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~ 1 day on a single 8-AI00 node. Furthermore, we present some early exploration of open problems in LMMs, including scaling to higher resolution inputs, compositional capabilities, and model hallucination, etc. We hope this makes state-of-the-art LMM research more accessible. Code and model will be publicly available."
                },
                {
                    "title": "MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts",
                    "abstract": "Large Language Models (LLMs) and Large Multimodal Models (LMMs) exhibit impressive problem-solving skills in many tasks and domains, but their ability in mathematical reasoning in visual contexts has not been systematically studied. To bridge this gap, we present MathVista, a benchmark designed to combine challenges from diverse mathematical and visual tasks. It consists of 6,141 examples, derived from 28 existing multimodal datasets involving mathematics and 3 newly created datasets (i.e., IQTest, FunctionQA, and PaperQA). Completing these tasks requires fine-grained, deep visual understanding and compositional reasoning, which all state-of-the-art foundation models find challenging. With MathVista, we have conducted a comprehensive, quantitative evaluation of 12 prominent foundation models. The best-performing GPT-4V model achieves an overall accuracy of 49.9%, substantially outperforming Bard, the second-best performer, by 15.1%. Our in-depth analysis reveals that the superiority of GPT-4V is mainly attributed to its enhanced visual perception and mathematical reasoning. However, GPT-4V still falls short of human performance by 10.4%, as it often struggles to understand complex figures and perform rigorous reasoning. This significant gap underscores the critical role that MathVista will play in the development of general-purpose AI agents capable of tackling mathematically intensive and visually rich real-world tasks. We further explore the new ability of self-verification, the application of self-consistency, and the interactive chatbot capabilities of GPT-4V, highlighting its promising potential for future research. The project is available at https://mathvista.github.io/."
                },
                {
                    "title": "The Dawn of LMMs: Preliminary Explorations with GPT-4V(ision)",
                    "abstract": "Large multimodal models (LMMs) extend large language models (LLMs) with multi-sensory skills, such as visual understanding, to achieve stronger generic intelligence. In this paper, we analyze the latest model, GPT-4V(ision), to deepen the understanding of LMMs. The analysis focuses on the intriguing tasks that GPT-4V can perform, containing test samples to probe the quality and genericity of GPT-4V's capabilities, its supported inputs and working modes, and the effective ways to prompt the model. In our approach to exploring GPT-4V, we curate and organize a collection of carefully designed qualitative samples spanning a variety of domains and tasks. Observations from these samples demonstrate that GPT-4V's unprecedented ability in processing arbitrarily interleaved multimodal inputs and the genericity of its capabilities together make GPT-4V a powerful multimodal generalist system. Furthermore, GPT-4V's unique capability of understanding visual markers drawn on input images can give rise to new human-computer interaction methods such as visual referring prompting. We conclude the report with in-depth discussions on the emerging application scenarios and the future research directions for GPT-4V-based systems. We hope that this preliminary exploration will inspire future research on the next-generation multimodal task formulation, new ways to exploit and enhance LMMs to solve real-world problems, and gaining better understanding of multimodal foundation models. Finally, we acknowledge that the model under our study is solely the product of OpenAI's innovative work, and they should be fully credited for its development. Please see the GPT-4V contributions paper for the authorship and credit attribution: https://cdn.openai.com/contributions/gpt-4v.pdf"
                },
                {
                    "title": "Qwen Technical Report",
                    "abstract": "Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural language processing tasks that were previously thought to be exclusive to humans. In this work, we introduce Qwen, the first installment of our large language model series. Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts. It includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques. The base language models consistently demonstrate superior performance across a multitude of downstream tasks, and the chat models, particularly those trained using Reinforcement Learning from Human Feedback (RLHF), are highly competitive. The chat models possess advanced tool-use and planning capabilities for creating agent applications, showcasing impressive performance even when compared to bigger models on complex tasks like utilizing a code interpreter. Furthermore, we have developed coding-specialized models, Code-Qwen and Code-Qwen-Chat, as well as mathematics-focused models, Math-Qwen-Chat, which are built upon base language models. These models demonstrate significantly improved performance in comparison with open-source models, and slightly fall behind the proprietary models."
                },
                {
                    "title": "BT-Adapter: Video Conversation is Feasible Without Video Instruction Tuning",
                    "abstract": "The recent progress in Large Language Models (LLM) has spurred various advancements in image-language con-versation agents, while how to build a proficient video-based dialogue system is still under exploration. Consid-ering the extensive scale of LLM and visual backbone, min-imal GPU memory is left for facilitating effective temporal modeling, which is crucial for comprehending and providing feedback on videos. To this end, we propose Branching Temporal Adapter (BT-Adapter), a novel method for ex-tending image-language pretrained models into the video domain. Specifically, BT-Adapter serves as a plug-and-use temporal modeling branch alongside the pretrained vi-sual encoder, which is tuned while keeping the backbone frozen. Just pretrained once, BT-Adapter can be seamlessly integrated into all image conversation models using this version of CLIP, enabling video conversations without the need for video instructions. Besides, we develop a unique asymmetric token masking strategy inside the branch with tailor-made training tasks for BT-Adapter, facilitating faster convergence and better results. Thanks to BT-Adapter, we are able to empower existing multimodal dialogue models with strong video understanding capabilities without incur-ring excessive GPU costs. Without bells and whistles, BT-Adapter achieves (1) state-of-the-art zero-shot results on various video tasks using thousands of fewer GPU hours. (2) better performance than current video chatbots without any video instruction tuning. (3) state-of-the-art results of video chatting using video instruction tuning, outperforming previous SOTAs by a large margin. The code has been available at https://github.com/farewellthreeIBT-Adapter."
                },
                {
                    "title": "MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities",
                    "abstract": "We propose MM-Vet, an evaluation benchmark that examines large multimodal models (LMMs) on complicated multimodal tasks. Recent LMMs have shown various intriguing abilities, such as solving math problems written on the blackboard, reasoning about events and celebrities in news images, and explaining visual jokes. Rapid model advancements pose challenges to evaluation benchmark development. Problems include: (1) How to systematically structure and evaluate the complicated multimodal tasks; (2) How to design evaluation metrics that work well across question and answer types; and (3) How to give model insights beyond a simple performance ranking. To this end, we present MM-Vet, designed based on the insight that the intriguing ability to solve complicated tasks is often achieved by a generalist model being able to integrate different core vision-language (VL) capabilities. MM-Vet defines 6 core VL capabilities and examines the 16 integrations of interest derived from the capability combination. For evaluation metrics, we propose an LLM-based evaluator for open-ended outputs. The evaluator enables the evaluation across different question types and answer styles, resulting in a unified scoring metric. We evaluate representative LMMs on MM-Vet, providing insights into the capabilities of different LMM system paradigms and models. Code and data are available at https://github.com/yuweihao/MM-Vet."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "MMBench: Is Your Multi-modal Model an All-around Player?",
                    "abstract": "Large vision-language models (VLMs) have recently achieved remarkable progress, exhibiting impressive multimodal perception and reasoning abilities. However, effectively evaluating these large VLMs remains a major challenge, hindering future development in this domain. Traditional benchmarks like VQAv2 or COCO Caption provide quantitative performance measurements but lack fine-grained ability assessment and robust evaluation metrics. Meanwhile, subjective benchmarks, such as OwlEval, offer comprehensive evaluations of a model's abilities by incorporating human labor, which is not scalable and may display significant bias. In response to these challenges, we propose MMBench, a bilingual benchmark for assessing the multi-modal capabilities of VLMs. MMBench methodically develops a comprehensive evaluation pipeline, primarily comprised of the following key features: 1. MMBench is meticulously curated with well-designed quality control schemes, surpassing existing similar benchmarks in terms of the number and variety of evaluation questions and abilities; 2. MMBench introduces a rigorous CircularEval strategy and incorporates large language models to convert free-form predictions into pre-defined choices, which helps to yield accurate evaluation results for models with limited instruction-following capabilities. 3. MMBench incorporates multiple-choice questions in both English and Chinese versions, enabling an apples-to-apples comparison of VLMs' performance under a bilingual context. To summarize, MMBench is a systematically designed objective benchmark for a robust and holistic evaluation of vision-language models. We hope MMBench will assist the research community in better evaluating their models and facilitate future progress in this area. The evalutation code of MMBench has been integrated into VLMEvalKit: https://github.com/open-compass/VLMEvalKit."
                },
                {
                    "title": "MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models",
                    "abstract": "Multimodal Large Language Model (MLLM) relies on the powerful LLM to perform multimodal tasks, showing amazing emergent abilities in recent studies, such as writing poems based on an image. However, it is difficult for these case studies to fully reflect the performance of MLLM, lacking a comprehensive evaluation. In this paper, we fill in this blank, presenting the first comprehensive MLLM Evaluation benchmark MME. It measures both perception and cognition abilities on a total of 14 subtasks. In order to avoid data leakage that may arise from direct use of public datasets for evaluation, the annotations of instruction-answer pairs are all manually designed. The concise instruction design allows us to fairly compare MLLMs, instead of struggling in prompt engineering. Besides, with such an instruction, we can also easily carry out quantitative statistics. A total of 30 advanced MLLMs are comprehensively evaluated on our MME, which not only suggests that existing MLLMs still have a large room for improvement, but also reveals the potential directions for the subsequent model optimization. The data application manner and online leaderboards are released at https://github.com/BradyFU/Awesome-Multimodal-Large-Language-Models/tree/Evaluation."
                },
                {
                    "title": "Video-ChatGPT: Towards Detailed Video Understanding via Large Vision and Language Models",
                    "abstract": "Conversation agents fueled by Large Language Models (LLMs) are providing a new way to interact with visual data. While there have been initial attempts for image-based conversation models, this work addresses the under-explored field of \\emph{video-based conversation} by introducing Video-ChatGPT. It is a multimodal model that merges a video-adapted visual encoder with an LLM. The resulting model is capable of understanding and generating detailed conversations about videos. We introduce a new dataset of 100,000 video-instruction pairs used to train Video-ChatGPT acquired via manual and semi-automated pipeline that is easily scalable and robust to label noise. We also develop a quantitative evaluation framework for video-based dialogue models to objectively analyze the strengths and weaknesses of video-based dialogue models. Code: https://github.com/mbzuai-oryx/Video-ChatGPT."
                },
                {
                    "title": "Video-LLaMA: An Instruction-tuned Audio-Visual Language Model for Video Understanding",
                    "abstract": "We present Video-LLaMA a multi-modal framework that empowers Large Language Models (LLMs) with the capability of understanding both visual and auditory content in the video. Video-LLaMA bootstraps cross-modal training from the frozen pre-trained visual and audio encoders and the frozen LLMs. Unlike previous works that complement LLMs to process the visual or audio signals only, Video-LLaMA enables video comprehension by tackling two challenges: (1) capturing the temporal changes in visual scenes, (2) integrating audio-visual signals. To counter the first challenge, we propose a Video Q-former to assemble a pre-trained image encoder into our video encoder and introduce a video-to-text generation task to learn video-language correspondence. For the second challenge, we leverage ImageBind, a universal embedding model aligning multiple modalities, as the pre-trained audio encoder and introduce an Audio Q-former on top of ImageBind to learn reasonable auditory query embeddings for the LLM module. To align the output of both visual and audio encoders with LLM's embedding space, we first train Video-LLaMA on massive video/image-caption pairs and then tune our model with visual-instruction datasets of moderate amount but higher quality. We found Video-LLaMA shows the ability to perceive and comprehend video content and generate meaningful responses grounded in the visual and auditory information presented in the videos."
                },
                {
                    "title": "Evaluating Object Hallucination in Large Vision-Language Models",
                    "abstract": "Inspired by the superior language abilities of large language models (LLM), large vision-language models (LVLM) have been recently explored by integrating powerful LLMs for improving the performance on complex multimodal tasks. Despite the promising progress on LVLMs, we find that LVLMs suffer from the hallucination problem, i.e. they tend to generate objects that are inconsistent with the target images in the descriptions. To investigate it, this work presents the first systematic study on object hallucination of LVLMs. We conduct the evaluation experiments on several representative LVLMs, and show that they mostly suffer from severe object hallucination issue. We further discuss that the visual instructions may influence the hallucination, and find that: objects that frequently occur in the visual instructions or co-occur with the image objects, are obviously prone to be hallucinated by LVLMs. Besides, we find that existing evaluation methods might be affected by the input instructions and generation styles of LVLMs. Thus, we further design an improved evaluation method for object hallucination by proposing a polling-based query method called POPE. Experiment results demonstrate that our POPE can evaluate the object hallucination in a more stable and flexible way. Our codes and data are publicly available at https://github.com/RUCAIBox/POPE."
                },
                {
                    "title": "InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning",
                    "abstract": "Large-scale pre-training and instruction tuning have been successful at creating general-purpose language models with broad competence. However, building general-purpose vision-language models is challenging due to the rich input distributions and task diversity resulting from the additional visual input. Although vision-language pretraining has been widely studied, vision-language instruction tuning remains under-explored. In this paper, we conduct a systematic and comprehensive study on vision-language instruction tuning based on the pretrained BLIP-2 models. We gather 26 publicly available datasets, covering a wide variety of tasks and capabilities, and transform them into instruction tuning format. Additionally, we introduce an instruction-aware Query Transformer, which extracts informative features tailored to the given instruction. Trained on 13 held-in datasets, InstructBLIP attains state-of-the-art zero-shot performance across all 13 held-out datasets, substantially outperforming BLIP-2 and larger Flamingo models. Our models also lead to state-of-the-art performance when finetuned on individual downstream tasks (e.g., 90.7% accuracy on ScienceQA questions with image contexts). Furthermore, we qualitatively demonstrate the advantages of InstructBLIP over concurrent multimodal models. All InstructBLIP models are open-sourced at https://github.com/salesforce/LAVIS/tree/main/projects/instructblip."
                },
                {
                    "title": "VideoChat: Chat-Centric Video Understanding",
                    "abstract": "In this paper, we initiate an attempt of developing an end-to-end chat-centric video understanding system, coined as VideoChat. It integrates video foundation models and large language models via a learnable neural interface, excelling in spatiotemporal reasoning, event localization, and causal relationship inference. To instructively tune this system, we build a video-centric instruction dataset, composed of thousands of videos associated with detailed descriptions and conversations. This dataset emphasizes spatiotemporal reasoning and captures causal relationships, providing a valuable asset for training our chat-centric video understanding system. Preliminary qualitative experiments demonstrate the potential of our system across a broad spectrum of video applications, which could serve as a simple prototype system for future research on chat-centric video understanding. Access our code and data at https://github.com/OpenGVLab/Ask-Anything"
                },
                {
                    "title": "MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models",
                    "abstract": "The recent GPT-4 has demonstrated extraordinary multi-modal abilities, such as directly generating websites from handwritten text and identifying humorous elements within images. These features are rarely observed in previous vision-language models. However, the technical details behind GPT-4 continue to remain undisclosed. We believe that the enhanced multi-modal generation capabilities of GPT-4 stem from the utilization of sophisticated large language models (LLM). To examine this phenomenon, we present MiniGPT-4, which aligns a frozen visual encoder with a frozen advanced LLM, Vicuna, using one projection layer. Our work, for the first time, uncovers that properly aligning the visual features with an advanced large language model can possess numerous advanced multi-modal abilities demonstrated by GPT-4, such as detailed image description generation and website creation from hand-drawn drafts. Furthermore, we also observe other emerging capabilities in MiniGPT-4, including writing stories and poems inspired by given images, teaching users how to cook based on food photos, and so on. In our experiment, we found that the model trained on short image caption pairs could produce unnatural language outputs (e.g., repetition and fragmentation). To address this problem, we curate a detailed image description dataset in the second stage to finetune the model, which consequently improves the model's generation reliability and overall usability. Our code, pre-trained model, and collected dataset are available at https://minigpt-4.github.io/."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention",
                    "abstract": "We present LLaMA-Adapter, a lightweight adaption method to efficiently fine-tune LLaMA into an instruction-following model. Using 52K self-instruct demonstrations, LLaMA-Adapter only introduces 1.2M learnable parameters upon the frozen LLaMA 7B model, and costs less than one hour for fine-tuning on 8 A100 GPUs. Specifically, we adopt a set of learnable adaption prompts, and prepend them to the word tokens at higher transformer layers. Then, a zero-initialized attention mechanism with zero gating is proposed, which adaptively injects the new instructional cues into LLaMA, while effectively preserves its pre-trained knowledge. With our efficient training, LLaMA-Adapter can generate high-quality responses, comparable to Alpaca with fully fine-tuned 7B parameters. Besides language commands, our approach can be simply extended to multi-modal instructions for learning image-conditioned LLaMA model, which achieves superior reasoning performance on ScienceQA and COCO Caption benchmarks. Furthermore, we also evaluate the zero-initialized attention mechanism for fine-tuning other pre-trained models (ViT, RoBERTa) on traditional vision and language tasks, demonstrating the superior generalization capacity of our approach. Code is released at https://github.com/OpenGVLab/LLaMA-Adapter."
                },
                {
                    "title": "Sigmoid Loss for Language Image Pre-Training",
                    "abstract": "We propose a simple pairwise sigmoid loss for imagetext pre-training. Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization. The sigmoid loss simultaneously allows further scaling up the batch size, while also performing better at smaller batch sizes. With only four TPUv4 chips, we can train a Base CLIP model at 4k batch size and a Large LiT model at 20k batch size, the latter achieves 84.5% ImageNet zero-shot accuracy in two days. This disentanglement of the batch size from the loss further allows us to study the impact of examples vs pairs and negative to positive ratio. Finally, we push the batch size to the extreme, up to one million, and find that the benefits of growing batch size quickly diminish, with a more reasonable batch size of 32k being sufficient. We hope our research motivates further explorations in improving the quality and efficiency of language-image pre-training."
                },
                {
                    "title": "EVA-CLIP: Improved Training Techniques for CLIP at Scale",
                    "abstract": "Contrastive language-image pre-training, CLIP for short, has gained increasing attention for its potential in various scenarios. In this paper, we propose EVA-CLIP, a series of models that significantly improve the efficiency and effectiveness of CLIP training. Our approach incorporates new techniques for representation learning, optimization, and augmentation, enabling EVA-CLIP to achieve superior performance compared to previous CLIP models with the same number of parameters but significantly smaller training costs. Notably, our largest 5.0B-parameter EVA-02-CLIP-E/14+ with only 9 billion seen samples achieves 82.0 zero-shot top-1 accuracy on ImageNet-1K val. A smaller EVA-02-CLIP-L/14+ with only 430 million parameters and 6 billion seen samples achieves 80.4 zero-shot top-1 accuracy on ImageNet-1K val. To facilitate open access and open research, we release the complete suite of EVA-CLIP to the community at https://github.com/baaivision/EVA/tree/master/EVA-CLIP."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "UATVR: Uncertainty-Adaptive Text-Video Retrieval",
                    "abstract": "With the explosive growth of web videos and emerging large-scale vision-language pre-training models, e.g., CLIP, retrieving videos of interest with text instructions has attracted increasing attention. A common practice is to transfer text-video pairs to the same embedding space and craft cross-modal interactions with certain entities in specific granularities for semantic correspondence. Unfortunately, the intrinsic uncertainties of optimal entity combinations in appropriate granularities for cross-modal queries are understudied, which is especially critical for modalities with hierarchical semantics, e.g., video, text, etc. In this paper, we propose an Uncertainty-Adaptive Text-Video Retrieval approach, termed UATVR, which models each lookup as a distribution matching procedure. Concretely, we add additional learnable tokens in the encoders to adaptively aggregate multi-grained semantics for flexible high-level reasoning. In the refined embedding space, we represent text-video pairs as probabilistic distributions where prototypes are sampled for matching evaluation. Comprehensive experiments on four benchmarks justify the superiority of our UATVR, which achieves new state-of-the-art results on MSR-VTT (50.8%), VATEX (64.5%), MSVD (49.7%), and DiDeMo (45.8%). The code is available at https://github.com/bofang98/UATVR."
                },
                {
                    "title": "Bidirectional Cross-Modal Knowledge Exploration for Video Recognition with Pre-trained Vision-Language Models",
                    "abstract": "Vision-language models (VLMs) pre-trained on large- scale image-text pairs have demonstrated impressive transferability on various visual tasks. Transferring knowledge from such powerful VLMs is a promising direction for building effective video recognition models. However, current exploration in this field is still limited. We believe that the greatest value of pre-trained VLMs lies in building a bridge between visual and textual domains. In this paper, we propose a novel framework called BIKE, which utilizes the cross-modal bridge to explore bidirectional knowledge: i) We introduce the Video Attribute Association mechanism, which leverages the Video-to-Text knowledge to generate textual auxiliary attributes for complementing video recognition. ii) We also present a Temporal Concept Spotting mechanism that uses the Text-to-Video expertise to capture temporal saliency in a parameter-free manner, leading to enhanced video representation. Extensive studies on six popular video datasets, including Kinetics-400 & 600, UCF-101, HMDB-51, ActivityNet and Charades, show that our method achieves state-of-the-art performance in various recognition scenarios, such as general, zero-shot, and few-shot video recognition. Our best model achieves a state-of-the-art accuracy of 88.6% on the challenging Kinetics-400 using the released CLIP model. The code is available at https://github.com/whwu95/BIKE."
                },
                {
                    "title": "Cap4Video: What Can Auxiliary Captions Do for Text-Video Retrieval?",
                    "abstract": "Most existing text-video retrieval methods focus on cross-modal matching between the visual content of videos and textual query sentences. However, in real-world scenarios, online videos are often accompanied by relevant text information such as titles, tags, and even subtitles, which can be utilized to match textual queries. This insight has motivated us to propose a novel approach to text-video retrieval, where we directly generate associated captions from videos using zero-shot video captioning with knowledge from web-scale pre-trained models (e.g., CLIP and GPT-2). Given the generated captions, a natural question arises: what benefits do they bring to text-video retrieval? To answer this, we introduce Cap4Video, a new framework that leverages captions in three ways: i) Input data: video-caption pairs can augment the training data. ii) Intermediate feature interaction: we perform cross-modal feature interaction between the video and caption to produce enhanced video representations. iii) Output score: the Query-Caption matching branch can complement the original Query-Video matching branch for text-video retrieval. We conduct comprehensive ablation studies to demonstrate the effectiveness of our approach. Without any post-processing, Cap4Video achieves state-of-the-art performance on four standard text-video retrieval benchmarks: MSR-VTT (51.4%), VATEX (66.6%), MSVD (51.8%), and DiDeMo (52.0%). The code is available at https://github.com/whwu95/Cap4Video."
                },
                {
                    "title": "Reproducible Scaling Laws for Contrastive Language-Image Learning",
                    "abstract": "Scaling up neural networks has led to remarkable performance across a wide range of tasks. Moreover, performance often follows reliable scaling laws as a function of training set size, model size, and compute, which offers valuable guidance as large-scale experiments are becoming increasingly expensive. However, previous work on scaling laws has primarily used private data & models or focused on uni-modal language or vision learning. To address these limitations, we investigate scaling laws for contrastive language-image pre-training (CLIP) with the public LAION dataset and the open-source OpenCLIP repository. Our large-scale experiments involve models trained on up to two billion image-text pairs and identify power law scaling for multiple downstream tasks including zero-shot classification, retrieval, linear probing, and end-to-end fine-tuning. We find that the training distribution plays a key role in scaling laws as the OpenAI and OpenCLIP models exhibit different scaling behavior despite identical model architectures and similar training recipes. We open-source our evaluation workflow and all models, including the largest public CLIP models, to ensure reproducibility and make scaling laws research more accessible. Source code and instructions to reproduce this study is available at https://github.eom/LAION-AI/sealing-laws-openelip."
                },
                {
                    "title": "Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering",
                    "abstract": "When answering a question, humans utilize the information available across different modalities to synthesize a consistent and complete chain of thought (CoT). This process is normally a black box in the case of deep learning models like large-scale language models. Recently, science question benchmarks have been used to diagnose the multi-hop reasoning ability and interpretability of an AI system. However, existing datasets fail to provide annotations for the answers, or are restricted to the textual-only modality, small scales, and limited domain diversity. To this end, we present Science Question Answering (ScienceQA), a new benchmark that consists of ~21k multimodal multiple choice questions with a diverse set of science topics and annotations of their answers with corresponding lectures and explanations. We further design language models to learn to generate lectures and explanations as the chain of thought (CoT) to mimic the multi-hop reasoning process when answering ScienceQA questions. ScienceQA demonstrates the utility of CoT in language models, as CoT improves the question answering performance by 1.20% in few-shot GPT-3 and 3.99% in fine-tuned UnifiedQA. We also explore the upper bound for models to leverage explanations by feeding those in the input; we observe that it improves the few-shot performance of GPT-3 by 18.96%. Our analysis further shows that language models, similar to humans, benefit from explanations to learn from fewer data and achieve the same performance with just 40% of the data. The data and code are available at https://scienceqa.github.io."
                },
                {
                    "title": "Zero-Shot Video Question Answering via Frozen Bidirectional Language Models",
                    "abstract": "Video question answering (VideoQA) is a complex task that requires diverse multi-modal data for training. Manual annotation of question and answers for videos, however, is tedious and prohibits scalability. To tackle this problem, recent methods consider zero-shot settings with no manual annotation of visual question-answer. In particular, a promising approach adapts frozen autoregressive language models pretrained on Web-scale text-only data to multi-modal inputs. In contrast, we here build on frozen bidirectional language models (BiLM) and show that such an approach provides a stronger and cheaper alternative for zero-shot VideoQA. In particular, (i) we combine visual inputs with the frozen BiLM using light trainable modules, (ii) we train such modules using Web-scraped multi-modal data, and finally (iii) we perform zero-shot VideoQA inference through masked language modeling, where the masked text is the answer to a given question. Our proposed approach, FrozenBiLM, outperforms the state of the art in zero-shot VideoQA by a significant margin on a variety of datasets, including LSMDC-FiB, iVQA, MSRVTT-QA, MSVD-QA, ActivityNet-QA, TGIF-FrameQA, How2QA and TVQA. It also demonstrates competitive performance in the few-shot and fully-supervised setting. Our code and models are publicly available at https://github.com/antoyang/FrozenBiLM."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "Flamingo: a Visual Language Model for Few-Shot Learning",
                    "abstract": "Building models that can be rapidly adapted to novel tasks using only a handful of annotated examples is an open challenge for multimodal machine learning research. We introduce Flamingo, a family of Visual Language Models (VLM) with this ability. We propose key architectural innovations to: (i) bridge powerful pretrained vision-only and language-only models, (ii) handle sequences of arbitrarily interleaved visual and textual data, and (iii) seamlessly ingest images or videos as inputs. Thanks to their flexibility, Flamingo models can be trained on large-scale multimodal web corpora containing arbitrarily interleaved text and images, which is key to endow them with in-context few-shot learning capabilities. We perform a thorough evaluation of our models, exploring and measuring their ability to rapidly adapt to a variety of image and video tasks. These include open-ended tasks such as visual question-answering, where the model is prompted with a question which it has to answer; captioning tasks, which evaluate the ability to describe a scene or an event; and close-ended tasks such as multiple-choice visual question-answering. For tasks lying anywhere on this spectrum, a single Flamingo model can achieve a new state of the art with few-shot learning, simply by prompting the model with task-specific examples. On numerous benchmarks, Flamingo outperforms models fine-tuned on thousands of times more task-specific data."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "FILIP: Fine-grained Interactive Language-Image Pre-Training",
                    "abstract": "Unsupervised large-scale vision-language pre-training has shown promising advances on various downstream tasks. Existing methods often model the cross-modal interaction either via the similarity of the global feature of each modality which misses sufficient information, or finer-grained interactions using cross/self-attention upon visual and textual tokens. However, cross/self-attention suffers from inferior efficiency in both training and inference. In this paper, we introduce a large-scale Fine-grained Interactive Language-Image Pre-training (FILIP) to achieve finer-level alignment through a cross-modal late interaction mechanism, which uses a token-wise maximum similarity between visual and textual tokens to guide the contrastive objective. FILIP successfully leverages the finer-grained expressiveness between image patches and textual words by modifying only contrastive loss, while simultaneously gaining the ability to pre-compute image and text representations offline at inference, keeping both large-scale training and inference efficient. Furthermore, we construct a new large-scale image-text pair dataset called FILIP300M for pre-training. Experiments show that FILIP achieves state-of-the-art performance on multiple downstream vision-language tasks including zero-shot image classification and image-text retrieval. The visualization on word-patch alignment further shows that FILIP can learn meaningful fine-grained features with promising localization ability."
                },
                {
                    "title": "Finetuned Language Models Are Zero-Shot Learners",
                    "abstract": "This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "XLNet: Generalized Autoregressive Pretraining for Language Understanding",
                    "abstract": "With the capability of modeling bidirectional contexts, denoising autoencoding based pretraining like BERT achieves better performance than pretraining approaches based on autoregressive language modeling. However, relying on corrupting the input with masks, BERT neglects dependency between the masked positions and suffers from a pretrain-finetune discrepancy. In light of these pros and cons, we propose XLNet, a generalized autoregressive pretraining method that (1) enables learning bidirectional contexts by maximizing the expected likelihood over all permutations of the factorization order and (2) overcomes the limitations of BERT thanks to its autoregressive formulation. Furthermore, XLNet integrates ideas from Transformer-XL, the state-of-the-art autoregressive model, into pretraining. Empirically, under comparable experiment settings, XLNet outperforms BERT on 20 tasks, often by a large margin, including question answering, natural language inference, sentiment analysis, and document ranking."
                },
                {
                    "title": "Towards VQA Models That Can Read",
                    "abstract": "Studies have shown that a dominant class of questions asked by visually impaired users on images of their surroundings involves reading text in the image. But today\u2019s VQA models can not read! Our paper takes a first step towards addressing this problem. First, we introduce a new \u201cTextVQA\u201d dataset to facilitate progress on this important problem. Existing datasets either have a small proportion of questions about text (e.g., the VQA dataset) or are too small (e.g., the VizWiz dataset). TextVQA contains 45,336 questions on 28,408 images that require reasoning about text to answer. Second, we introduce a novel model architecture that reads text in the image, reasons about it in the context of the image and the question, and predicts an answer which might be a deduction based on the text and the image or composed of the strings found in the image. Consequently, we call our approach Look, Read, Reason & Answer (LoRRA). We show that LoRRA outperforms existing state-of-the-art VQA models on our TextVQA dataset. We find that the gap between human performance and machine performance is significantly larger on TextVQA than on VQA 2.0, suggesting that TextVQA is well-suited to benchmark progress along directions complementary to VQA 2.0."
                },
                {
                    "title": "GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering",
                    "abstract": "We introduce GQA, a new dataset for real-world visual reasoning and compositional question answering, seeking to address key shortcomings of previous VQA datasets. We have developed a strong and robust question engine that leverages Visual Genome scene graph structures to create 22M diverse reasoning questions, which all come with functional programs that represent their semantics. We use the programs to gain tight control over the answer distribution and present a new tunable smoothing technique to mitigate question biases. Accompanying the dataset is a suite of new metrics that evaluate essential qualities such as consistency, grounding and plausibility. A careful analysis is performed for baselines as well as state-of-the-art models, providing fine-grained results for different question types and topologies. Whereas a blind LSTM obtains a mere 42.1%, and strong VQA models achieve 54.1%, human performance tops at 89.3%, offering ample opportunity for new research to explore. We hope GQA will provide an enabling resource for the next generation of models with enhanced robustness, improved consistency, and deeper semantic understanding of vision and language."
                },
                {
                    "title": "Feature Pyramid Networks for Object Detection",
                    "abstract": "Feature pyramids are a basic component in recognition systems for detecting objects at different scales. But pyramid representations have been avoided in recent object detectors that are based on deep convolutional networks, partially because they are slow to compute and memory intensive. In this paper, we exploit the inherent multi-scale, pyramidal hierarchy of deep convolutional networks to construct feature pyramids with marginal extra cost. A top-down architecture with lateral connections is developed for building high-level semantic feature maps at all scales. This architecture, called a Feature Pyramid Network (FPN), shows significant improvement as a generic feature extractor in several applications. Using a basic Faster R-CNN system, our method achieves state-of-the-art single-model results on the COCO detection benchmark without bells and whistles, surpassing all existing single-model entries including those from the COCO 2016 challenge winners. In addition, our method can run at 5 FPS on a GPU and thus is a practical and accurate solution to multi-scale object detection. Code will be made publicly available."
                },
                {
                    "title": "Densely Connected Convolutional Networks",
                    "abstract": "Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections&#x2014;one between each layer and its subsequent layer&#x2014;our network has L(L+1)/2 direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters. We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet). DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less memory and computation to achieve high performance. Code and pre-trained models are available at https://github.com/liuzhuang13/DenseNet."
                },
                {
                    "title": "Gaussian Error Linear Units (GELUs)",
                    "abstract": "We propose the Gaussian Error Linear Unit (GELU), a high-performing neural network activation function. The GELU activation function is $x\\Phi(x)$, where $\\Phi(x)$ the standard Gaussian cumulative distribution function. The GELU nonlinearity weights inputs by their value, rather than gates inputs by their sign as in ReLUs ($x\\mathbf{1}_{x>0}$). We perform an empirical evaluation of the GELU nonlinearity against the ReLU and ELU activations and find performance improvements across all considered computer vision, natural language processing, and speech tasks."
                },
                {
                    "title": "MSR-VTT: A Large Video Description Dataset for Bridging Video and Language",
                    "abstract": "While there has been increasing interest in the task of describing video with natural language, current computer vision algorithms are still severely limited in terms of the variability and complexity of the videos and their associated language that they can recognize. This is in part due to the simplicity of current benchmarks, which mostly focus on specific fine-grained domains with limited videos and simple descriptions. While researchers have provided several benchmark datasets for image captioning, we are not aware of any large-scale video description dataset with comprehensive categories yet diverse video content. In this paper we present MSR-VTT (standing for \"MSRVideo to Text\") which is a new large-scale video benchmark for video understanding, especially the emerging task of translating video to text. This is achieved by collecting 257 popular queries from a commercial video search engine, with 118 videos for each query. In its current version, MSR-VTT provides 10K web video clips with 41.2 hours and 200K clip-sentence pairs in total, covering the most comprehensive categories and diverse visual content, and representing the largest dataset in terms of sentence and vocabulary. Each clip is annotated with about 20 natural sentences by 1,327 AMT workers. We present a detailed analysis of MSR-VTT in comparison to a complete set of existing datasets, together with a summarization of different state-of-the-art video-to-text approaches. We also provide an extensive evaluation of these approaches on this dataset, showing that the hybrid Recurrent Neural Networkbased approach, which combines single-frame and motion representations with soft-attention pooling strategy, yields the best generalization capability on MSR-VTT."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "ActivityNet: A large-scale video benchmark for human activity understanding",
                    "abstract": "In spite of many dataset efforts for human action recognition, current computer vision algorithms are still severely limited in terms of the variability and complexity of the actions that they can recognize. This is in part due to the simplicity of current benchmarks, which mostly focus on simple actions and movements occurring on manually trimmed videos. In this paper we introduce ActivityNet, a new large-scale video benchmark for human activity understanding. Our benchmark aims at covering a wide range of complex human activities that are of interest to people in their daily living. In its current version, ActivityNet provides samples from 203 activity classes with an average of 137 untrimmed videos per class and 1.41 activity instances per video, for a total of 849 video hours. We illustrate three scenarios in which ActivityNet can be used to compare algorithms for human activity understanding: untrimmed video classification, trimmed activity classification and activity detection."
                },
                {
                    "title": "Collecting Highly Parallel Data for Paraphrase Evaluation",
                    "abstract": "A lack of standard datasets and evaluation metrics has prevented the field of paraphrasing from making the kind of rapid progress enjoyed by the machine translation community over the last 15 years. We address both problems by presenting a novel data collection framework that produces highly parallel text data relatively inexpensively and on a large scale. The highly parallel nature of this data allows us to use simple n-gram comparisons to measure both the semantic adequacy and lexical dissimilarity of paraphrase candidates. In addition to being simple and efficient to compute, experiments show that these metrics correlate highly with human judgments."
                },
                {
                    "title": "GPT-4V(ision) System Card",
                    "abstract": "GPT-4 with vision (GPT-4V) enables users to instruct GPT-4 to analyze image inputs provided by the user, and is the latest capability we are making broadly available. Incorporating additional modalities (such as image inputs) into large language models (LLMs) is viewed by some as a key frontier in artificial intelligence research and development [1, 2, 3]. Multimodal LLMs offer the possibility of expanding the impact of language-only systems with novel interfaces and capabilities, enabling them to solve new tasks and provide novel experiences for their users. In this system card, [4, 5] 1 we analyze the safety properties of GPT-4V. Our work on safety for GPT-4V builds on the work done for GPT-4 [7] and here we dive deeper into the evaluations, preparation, and mitigation work done specifically for image inputs. Similar to GPT-4, training of GPT-4V was completed in 2022 and we began providing early access to the system in March 2023. As GPT-4 is the technology behind the visual capabilities of GPT-4V, its training process was the same. The pre-trained model was first trained to predict the next word in a document, using a large dataset of text and image data from the Internet as well as licensed sources of data. It was then fine-tuned with additional data, using an algorithm called reinforcement learning from human feedback (RLHF),[8, 9] to produce outputs that are preferred by human trainers. Large multimodal models introduce different limitations and expand the risk surface compared to text-based language models. GPT-4V possesses the limitations and capabilities of each modality (text and vision), while at the same time presenting novel capabilities emerging from the intersection of said modalities and from the intelligence and reasoning afforded by large scale models. This system card outlines how OpenAI prepared the vision capabilities of GPT-4 for deployment. It describes the early access period of the model for small scale users and safety learnings OpenAI gained from this period, multimodal evaluations built to study the model\u2019s fitness for deployment, key findings of expert red teamers, and the mitigations OpenAI implemented prior to broad release."
                },
                {
                    "title": "Qwen-VL: A Frontier Large Vision-Language Model with Versatile Abilities",
                    "abstract": "We introduce the Qwen-VL series, a set of large-scale vision-language models designed to perceive and understand both text and images. Comprising Qwen-VL and Qwen-VL-Chat, these models exhibit remarkable performance in tasks like image captioning, question answering, visual localization, and flexible interaction. The evaluation covers a wide range of tasks including zero-shot captioning, visual or document visual question answering, and grounding. We demonstrate the Qwen-VL outperforms existing Large Vision Language Models (LVLMs). We present their architecture, training, capabilities, and performance, highlighting their contributions to advancing multimodal artificial intelligence. Code, demo and models are available at https://github.com/QwenLM/Qwen-VL ."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively utilize multi-layer visual features from frozen visual encoders in Multimodal Large Language Models (MLLMs) to enhance their performance in vision-language tasks?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the capabilities of MLLMs, as it could lead to improved understanding and generation of multimodal content. By leveraging multi-layer visual features, we can enhance the richness of visual information provided to language models, potentially leading to better performance in tasks such as image captioning, visual question answering, and other applications that require a deep integration of vision and language. This research could pave the way for future studies to explore more efficient architectures and methodologies, ultimately contributing to the development of more sophisticated AI systems that can understand and generate content across multiple modalities.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in effectively integrating visual features from different layers of a frozen visual encoder without incurring significant computational costs. Naive approaches may fail because they might not adequately capture the diverse information represented across various layers, leading to suboptimal performance. Additionally, the complexity of aligning and projecting these multi-layer features into a coherent representation that the language model can utilize poses a significant technical obstacle. Theoretical challenges also arise in understanding how to best combine these features to maximize their contribution to the language model's performance.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on optimizing the language aspects of MLLMs, often overlooking the potential of multi-layer visual features. Existing solutions typically freeze the visual encoder to reduce computational overhead, which has limited exploration into the rich information available at different layers. Barriers such as the lack of methodologies for effectively integrating these features and the prevailing focus on high-level visual representations have prevented this problem from being addressed. Our approach differs by proposing the Dense Connector (DC), which allows for the integration of offline features from various layers without additional computational costs, thus enhancing the visual information available to the language model.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the Dense Connector (DC), which integrates visual features from multiple layers of a frozen visual encoder. We will explore three instantiations: 1) Sparse Token Integration (STI), which aggregates visual tokens from different layers; 2) Sparse Channel Integration (SCI), which"
            }
        },
        "author_data": {
            "3fe8d107-cbc5-4b48-af70-15d4c7bbbdc5": {
                "pk": "3fe8d107-cbc5-4b48-af70-15d4c7bbbdc5",
                "name": "Huanjin Yao",
                "collaborators": [
                    "Wenhao Wu",
                    "Mengxi Zhang",
                    "Yuxin Song",
                    "Jingdong Wang",
                    "Zhiheng Li",
                    "Wanli Ouyang",
                    "Yu Lu",
                    "Kang Rong",
                    "Jianbo Zhao",
                    "Fanglong Liu"
                ],
                "domain": [
                    "Computer Vision",
                    "Transfer Learning",
                    "Multimodal Learning",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Side4Video: Spatial-Temporal Side Network for Memory-Efficient Image-to-Video Transfer Learning",
                        "abstract": "Large pre-trained vision models achieve impressive success in computer vision. However, fully fine-tuning large models for downstream tasks, particularly in video understanding, can be prohibitively computationally expensive. Recent studies turn their focus towards efficient image-to-video transfer learning. Nevertheless, existing efficient fine-tuning methods lack attention to training memory usage and exploration of transferring a larger model to the video domain. In this paper, we present a novel Spatial-Temporal Side Network for memory-efficient fine-tuning large image models to video understanding, named Side4Video. Specifically, we introduce a lightweight spatial-temporal side network attached to the frozen vision model, which avoids the backpropagation through the heavy pre-trained model and utilizes multi-level spatial features from the original image model. Extremely memory-efficient architecture enables our method to reduce 75% memory usage than previous adapter-based methods. In this way, we can transfer a huge ViT-E (4.4B) for video understanding tasks which is 14x larger than ViT-L (304M). Our approach achieves remarkable performance on various video datasets across unimodal and cross-modal tasks (i.e., action recognition and text-video retrieval), especially in Something-Something V1&V2 (67.3% & 74.6%), Kinetics-400 (88.6%), MSR-VTT (52.3%), MSVD (56.1%) and VATEX (68.8%). We release our code at https://github.com/HJYao00/Side4Video."
                    },
                    {
                        "title": "GPT4Vis: What Can GPT-4 Do for Zero-shot Visual Recognition?",
                        "abstract": "This paper does not present a novel method. Instead, it delves into an essential, yet must-know baseline in light of the latest advancements in Generative Artificial Intelligence (GenAI): the utilization of GPT-4 for visual understanding. Our study centers on the evaluation of GPT-4's linguistic and visual capabilities in zero-shot visual recognition tasks: Firstly, we explore the potential of its generated rich textual descriptions across various categories to enhance recognition performance without any training. Secondly, we evaluate GPT-4's visual proficiency in directly recognizing diverse visual content. We conducted extensive experiments to systematically evaluate GPT-4's performance across images, videos, and point clouds, using 16 benchmark datasets to measure top-1 and top-5 accuracy. Our findings show that GPT-4, enhanced with rich linguistic descriptions, significantly improves zero-shot recognition, offering an average top-1 accuracy increase of 7% across all datasets. GPT-4 excels in visual recognition, outshining OpenAI-CLIP's ViT-L and rivaling EVA-CLIP's ViT-E, particularly in video datasets HMDB-51 and UCF-101, where it leads by 22% and 9%, respectively. We hope this research contributes valuable data points and experience for future studies. We release our code at https://github.com/whwu95/GPT4Vis."
                    },
                    {
                        "title": "Automated Multi-level Preference for MLLMs",
                        "abstract": "Current multimodal Large Language Models (MLLMs) suffer from ``hallucination'', occasionally generating responses that are not grounded in the input images. To tackle this challenge, one promising path is to utilize reinforcement learning from human feedback (RLHF), which steers MLLMs towards learning superior responses while avoiding inferior ones. We rethink the common practice of using binary preferences (i.e., superior, inferior), and find that adopting multi-level preferences (e.g., superior, medium, inferior) is better for two benefits: 1) It narrows the gap between adjacent levels, thereby encouraging MLLMs to discern subtle differences. 2) It further integrates cross-level comparisons (beyond adjacent-level comparisons), thus providing a broader range of comparisons with hallucination examples. To verify our viewpoint, we present the Automated Multi-level Preference (AMP) framework for MLLMs. To facilitate this framework, we first develop an automated dataset generation pipeline that provides high-quality multi-level preference datasets without any human annotators. Furthermore, we design the Multi-level Direct Preference Optimization (MDPO) algorithm to robustly conduct complex multi-level preference learning. Additionally, we propose a new hallucination benchmark, MRHal-Bench. Extensive experiments across public hallucination and general benchmarks, as well as our MRHal-Bench, demonstrate the effectiveness of our proposed method. Code is available at https://github.com/takomc/amp."
                    }
                ]
            },
            "52ba35fd-8b47-4ee9-b928-bc342046ff0c": {
                "pk": "52ba35fd-8b47-4ee9-b928-bc342046ff0c",
                "name": "Wenhao Wu",
                "collaborators": [
                    "E. Bielejec",
                    "J. Ruan",
                    "Sujian Li",
                    "Zuxin Ye",
                    "Hans Mittelmann",
                    "Zhi Ding",
                    "Weiwei Wang",
                    "Shengjiang Kong",
                    "Zhun Sun",
                    "Wanli Ouyang"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Video Understanding",
                    "Superconductivity",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "FreeVA: Offline MLLM as Training-Free Video Assistant",
                        "abstract": "This paper undertakes an empirical study to revisit the latest advancements in Multimodal Large Language Models (MLLMs): Video Assistant. This study, namely FreeVA, aims to extend existing image-based MLLM to the video domain in a training-free manner. The study provides an essential, yet must-know baseline, and reveals several surprising findings: 1) FreeVA, leveraging only offline image-based MLLM without additional training, excels in zero-shot video question-answering (e.g., MSVD-QA, ActivityNet-QA, and MSRVTT-QA), even surpassing state-of-the-art methods that involve video instruction tuning. 2) While mainstream video-based MLLMs typically initialize with an image-based MLLM (e.g., LLaVA) and then fine-tune using video instruction tuning, the study indicates that utilizing the widely adopted VideoInstruct-100K for video instruction tuning doesn't actually lead to better performance compared to not training at all. 3) The commonly used evaluation metrics in existing works are significantly influenced by changes in the GPT API version over time. If ignored, this could affect the fairness and uniformity of comparisons between different methods and impact the analysis and judgment of researchers in the field. The advancement of MLLMs is currently thriving, drawing numerous researchers into the field. We aim for this work to serve as a plug-and-play, simple yet effective baseline, encouraging the direct evaluation of existing MLLMs in video domain while also standardizing the field of video conversational models to a certain extent. Also, we encourage researchers to reconsider: Have current video MLLM methods truly acquired knowledge beyond image MLLM? Code is available at https://github.com/whwu95/FreeVA"
                    },
                    {
                        "title": "Measurement of the Localization Length through the superconductor-insulator transition of ultrathin amorphous beryllium films",
                        "abstract": "Based on the Efros-Shklovskii behavior observed in electron transport and tunneling, we have determined, independently and up to a constant on the order of unity, the localization length and the dielectric constant for quench-condensed ultrathin amorphous beryllium films. As the normal-state sheet resistance of the films at 20 K is reduced with increasing film thickness, the localization length increases exponentially. The superconductor-insulator transition occurs when the localization length crosses the Ginzburg-Landau coherence length. We discuss the implication of these results on a number of issues regarding superconductivity in the strongly disordered regime."
                    },
                    {
                        "title": "A Comprehensive Attempt to Research Statement Generation",
                        "abstract": "For a researcher, writing a good research statement is crucial but costs a lot of time and effort. To help researchers, in this paper, we propose the research statement generation (RSG) task which aims to summarize one's research achievements and help prepare a formal research statement. For this task, we conduct a comprehensive attempt including corpus construction, method design, and performance evaluation. First, we construct an RSG dataset with 62 research statements and the corresponding 1,203 publications. Due to the limitation of our resources, we propose a practical RSG method which identifies a researcher's research directions by topic modeling and clustering techniques and extracts salient sentences by a neural text summarizer. Finally, experiments show that our method outperforms all the baselines with better content coverage and coherence."
                    },
                    {
                        "title": "Electron Glass in Ultrathin Granular Al Films at Low Temperatures",
                        "abstract": "Quench-condensed granular Al films, with normal-state sheet resistance close to 10 k$\\Omega/\\Box$, display strong hysteresis and ultraslow, non-exponential relaxation in the resistance when temperature is varied below 300 mK. The hysteresis is nonlinear and can be suppressed by a dc bias voltage. The relaxation time does not obey the Arrhenius form, indicating the existence of a broad distribution of low energy barriers. Furthermore, large resistance fluctuations, having a 1/f-type power spectrum with a low-frequency cut-off, are observed at low temperatures. With decreasing temperature, the amplitude of the fluctuation increases and the cut-off frequency decreases. These observations combine to provide a coherent picture that there exists a new glassy electron state in ultrathin granular Al films, with a growing correlation length at low temperatures."
                    },
                    {
                        "title": "The Field-Tuned Superconductor-Insulator Transition with and without Current Bias",
                        "abstract": "The magnetic-field-tuned superconductor-insulator transition has been studied in ultrathin Beryllium films quench-condensed near 20 K. In the zero-current limit, a finite-size scaling analysis yields the scaling exponent product vz = 1.35 +/- 0.10 and a critical sheet resistance R_{c} of about 1.2R_{Q}, with R_{Q} = h/4e^{2}. However, in the presence of dc bias currents that are smaller than the zero-field critical currents, vz becomes 0.75 +/- 0.10. This new set of exponents suggests that the field-tuned transitions with and without dc bias currents belong to different universality classes."
                    },
                    {
                        "title": "Quantized conductance observed in reversible atomic contacts fabricated by template electroplating using an on-membrane anode",
                        "abstract": "We report a new template electroplating method for fabricating reversible atomic contacts between a single nanowire and a macroscopic contact pad. In comparison to typical template methods using a standing-alone anode, we directly evaporate the anode on one of the porous membrane surfaces. Single nanowires, upon emerging from the pores, make reversible atomic contacts with the on-membrane anode via a self-terminating mechanism. Quantized conductance steps have been observed in a controlled fashion during deposition and dissolution. This method can potentially be applied for the controlled fabrication and integration of nanowires, point contacts, and nanosized interconnects in template-based nanofabrication."
                    },
                    {
                        "title": "Measuring the Localization Length through the superconductor-insulator transition in ultrathin amorphous beryllium films",
                        "abstract": "Electron transport and tunneling across the superconductor-insulator (SI) transition have been measured simultaneously for quench-condensed ultrathin amorphous beryllium films. The anomalous negative magnetoresistance previously observed in insulating films disappears when Mn impurities are introduced to the films, restoring a rather clean Efros-Shklovskii type hopping behavior. The combination of transport and tunneling data allows us to determine, independently and up to a constant on the order of unity, the localization length, \\xi_{L}, and the dielectric constant, \\kappa, for the films. As the normal-state sheet resistance of the films at 20 K is reduced with increasing film thickness, \\xi_{L} increases exponentially. The SI transition occurs when \\xi_{L} crosses the Ginzburg-Landau coherence length, \\xi_{S}."
                    },
                    {
                        "title": "Statistical Analysis of a Posteriori Channel and Noise Distribution Based on HARQ Feedback",
                        "abstract": "In response to a comment on one of our manuscript, this work studies the posterior channel and noise distributions conditioned on the NACKs and ACKs of all previous transmissions in HARQ system with statistical approaches. Our main result is that, unless the coherence interval (time or frequency) is large as in block-fading assumption, the posterior distribution of the channel and noise either remains almost identical to the prior distribution, or it mostly follows the same class of distribution as the prior one. In the latter case, the difference between the posterior and prior distribution can be modeled as some parameter mismatch, which has little impact on certain type of applications."
                    },
                    {
                        "title": "Deep Structure and Attention Aware Subspace Clustering",
                        "abstract": "Clustering is a fundamental unsupervised representation learning task with wide application in computer vision and pattern recognition. Deep clustering utilizes deep neural networks to learn latent representation, which is suitable for clustering. However, previous deep clustering methods, especially image clustering, focus on the features of the data itself and ignore the relationship between the data, which is crucial for clustering. In this paper, we propose a novel Deep Structure and Attention aware Subspace Clustering (DSASC), which simultaneously considers data content and structure information. We use a vision transformer to extract features, and the extracted features are divided into two parts, structure features, and content features. The two features are used to learn a more efficient subspace structure for spectral clustering. Extensive experimental results demonstrate that our method significantly outperforms state-of-the-art methods. Our code will be available at https://github.com/cs-whh/DSASC"
                    },
                    {
                        "title": "Observation of a Hard Correlation Gap in Quench-Condensed Ultrathin Beryllium",
                        "abstract": "We report on the tunneling density of states (DOS) in ultrathin and strongly disordered Be films quench-condensed at 20 K. Above 5 K, the DOS shows the well-known logarithmic anomaly at the Fermi level. Only in a narrow temperature range near 2 K is the DOS linearly dependent on energy, as predicted by Efros and Shklovskii. However, both the zero-bias conductance and the slope of the linear DOS are found to decrease drastically with decreasing temperature. Tunneling measurements at mK temperatures have revealed conclusively that a hard correlation gap opens up in the DOS."
                    },
                    {
                        "title": "Revisiting Classifier: Transferring Vision-Language Models for Video Recognition",
                        "abstract": "Transferring knowledge from task-agnostic pre-trained deep models for downstream tasks is an important topic in computer vision research. Along with the growth of computational capacity, we now have open-source vision-language pre-trained models in large scales of the model architecture and amount of data. In this study, we focus on transferring knowledge for video classification tasks. Conventional methods randomly initialize the linear classifier head for vision classification, but they leave the usage of the text encoder for downstream visual recognition tasks undiscovered. In this paper, we revise the role of the linear classifier and replace the classifier with the different knowledge from pre-trained model. We utilize the well-pretrained language model to generate good semantic target for efficient transferring learning. The empirical study shows that our method improves both the performance and the training speed of video classification, with a negligible change in the model. Our simple yet effective tuning paradigm achieves state-of-the-art performance and efficient training on various video recognition scenarios, i.e., zero-shot, few-shot, general recognition. In particular, our paradigm achieves the state-of-the-art accuracy of 87.8% on Kinetics-400, and also surpasses previous methods by 20~50% absolute top-1 accuracy under zero-shot, few-shot settings on five popular video datasets. Code and models can be found at https://github.com/whwu95/Text4Vis ."
                    },
                    {
                        "title": "Side4Video: Spatial-Temporal Side Network for Memory-Efficient Image-to-Video Transfer Learning",
                        "abstract": "Large pre-trained vision models achieve impressive success in computer vision. However, fully fine-tuning large models for downstream tasks, particularly in video understanding, can be prohibitively computationally expensive. Recent studies turn their focus towards efficient image-to-video transfer learning. Nevertheless, existing efficient fine-tuning methods lack attention to training memory usage and exploration of transferring a larger model to the video domain. In this paper, we present a novel Spatial-Temporal Side Network for memory-efficient fine-tuning large image models to video understanding, named Side4Video. Specifically, we introduce a lightweight spatial-temporal side network attached to the frozen vision model, which avoids the backpropagation through the heavy pre-trained model and utilizes multi-level spatial features from the original image model. Extremely memory-efficient architecture enables our method to reduce 75% memory usage than previous adapter-based methods. In this way, we can transfer a huge ViT-E (4.4B) for video understanding tasks which is 14x larger than ViT-L (304M). Our approach achieves remarkable performance on various video datasets across unimodal and cross-modal tasks (i.e., action recognition and text-video retrieval), especially in Something-Something V1&V2 (67.3% & 74.6%), Kinetics-400 (88.6%), MSR-VTT (52.3%), MSVD (56.1%) and VATEX (68.8%). We release our code at https://github.com/HJYao00/Side4Video."
                    },
                    {
                        "title": "Anisotropic Magnetoconductance in Quench-Condensed Ultrathin Beryllium Films",
                        "abstract": "Near the superconductor-insulator (S-I) transition, quench-condensed ultrathin Be films show a large magnetoconductance which is highly anisotropic in the direction of the applied field. Film conductance can drop as much as seven orders of magnitude in a weak perpendicular field (< 1 T), but is insensitive to a parallel field in the same field range. We believe that this negative magnetoconductance is due to the field de-phasing of the superconducting pair wavefunction. This idea enables us to extract the finite superconducting phase coherence length in nearly superconducting films. Our data indicate that this local phase coherence persists even in highly insulating films in the vicinity of the S-I transition."
                    },
                    {
                        "title": "Color2Embed: Fast Exemplar-Based Image Colorization using Color Embeddings",
                        "abstract": "In this paper, we present a fast exemplar-based image colorization approach using color embeddings named Color2Embed. Generally, due to the difficulty of obtaining input and ground truth image pairs, it is hard to train a exemplar-based colorization model with unsupervised and unpaired training manner. Current algorithms usually strive to achieve two procedures: i) retrieving a large number of reference images with high similarity for preparing training dataset, which is inevitably time-consuming and tedious; ii) designing complicated modules to transfer the colors of the reference image to the target image, by calculating and leveraging the deep semantic correspondence between them (e.g., non-local operation), which is computationally expensive during testing. Contrary to the previous methods, we adopt a self-augmented self-reference learning scheme, where the reference image is generated by graphical transformations from the original colorful one whereby the training can be formulated in a paired manner. Second, in order to reduce the process time, our method explicitly extracts the color embeddings and exploits a progressive style feature Transformation network, which injects the color embeddings into the reconstruction of the final image. Such design is much more lightweight and intelligible, achieving appealing performance with fast processing speed."
                    }
                ]
            },
            "db70e5e3-d0ab-45df-abbd-716b7000fff6": {
                "pk": "db70e5e3-d0ab-45df-abbd-716b7000fff6",
                "name": "Taojiannan Yang",
                "collaborators": [
                    "Chen Chen",
                    "Sijie Zhu",
                    "Matias Mendieta",
                    "Pu Wang",
                    "Minwoo Lee",
                    "Weiping Yu",
                    "Ce Zheng",
                    "Zhengming Ding",
                    "Linjie Yang",
                    "Xiaojie Jin"
                ],
                "domain": [
                    "Neural Architecture Search",
                    "Object Detection",
                    "Federated Learning",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Revisiting Training-free NAS Metrics: An Efficient Training-based Method",
                        "abstract": "Recent neural architecture search (NAS) works proposed training-free metrics to rank networks which largely reduced the search cost in NAS. In this paper, we revisit these training-free metrics and find that: (1) the number of parameters (\\#Param), which is the most straightforward training-free metric, is overlooked in previous works but is surprisingly effective, (2) recent training-free metrics largely rely on the \\#Param information to rank networks. Our experiments show that the performance of recent training-free metrics drops dramatically when the \\#Param information is not available. Motivated by these observations, we argue that metrics less correlated with the \\#Param are desired to provide additional information for NAS. We propose a light-weight training-based metric which has a weak correlation with the \\#Param while achieving better performance than training-free metrics at a lower search cost. Specifically, on DARTS search space, our method completes searching directly on ImageNet in only 2.6 GPU hours and achieves a top-1/top-5 error rate of 24.1\\%/7.1\\%, which is competitive among state-of-the-art NAS methods. Codes are available at \\url{https://github.com/taoyang1122/Revisit_TrainingFree_NAS}"
                    },
                    {
                        "title": "GradAug: A New Regularization Method for Deep Neural Networks",
                        "abstract": "We propose a new regularization method to alleviate over-fitting in deep neural networks. The key idea is utilizing randomly transformed training samples to regularize a set of sub-networks, which are originated by sampling the width of the original network, in the training process. As such, the proposed method introduces self-guided disturbances to the raw gradients of the network and therefore is termed as Gradient Augmentation (GradAug). We demonstrate that GradAug can help the network learn well-generalized and more diverse representations. Moreover, it is easy to implement and can be applied to various structures and applications. GradAug improves ResNet-50 to 78.79% on ImageNet classification, which is a new state-of-the-art accuracy. By combining with CutMix, it further boosts the performance to 79.67%, which outperforms an ensemble of advanced training tricks. The generalization ability is evaluated on COCO object detection and instance segmentation where GradAug significantly surpasses other state-of-the-art methods. GradAug is also robust to image distortions and FGSM adversarial attacks and is highly effective in low data regimes. Code is available at https://github.com/taoyang1122/GradAug"
                    },
                    {
                        "title": "Revisiting Street-to-Aerial View Image Geo-localization and Orientation Estimation",
                        "abstract": "Street-to-aerial image geo-localization, which matches a query street-view image to the GPS-tagged aerial images in a reference set, has attracted increasing attention recently. In this paper, we revisit this problem and point out the ignored issue about image alignment information. We show that the performance of a simple Siamese network is highly dependent on the alignment setting and the comparison of previous works can be unfair if they have different assumptions. Instead of focusing on the feature extraction under the alignment assumption, we show that improvements in metric learning techniques significantly boost the performance regardless of the alignment. Without leveraging the alignment information, our pipeline outperforms previous works on both panorama and cropped datasets. Furthermore, we conduct visualization to help understand the learned model and the effect of alignment information using Grad-CAM. With our discovery on the approximate rotation-invariant activation maps, we propose a novel method to estimate the orientation/alignment between a pair of cross-view images with unknown alignment information. It achieves state-of-the-art results on the CVUSA dataset."
                    },
                    {
                        "title": "Visual Explanation for Deep Metric Learning",
                        "abstract": "This work explores the visual explanation for deep metric learning and its applications. As an important problem for learning representation, metric learning has attracted much attention recently, while the interpretation of such model is not as well studied as classification. To this end, we propose an intuitive idea to show where contributes the most to the overall similarity of two input images by decomposing the final activation. Instead of only providing the overall activation map of each image, we propose to generate point-to-point activation intensity between two images so that the relationship between different regions is uncovered. We show that the proposed framework can be directly deployed to a large range of metric learning applications and provides valuable information for understanding the model. Furthermore, our experiments show its effectiveness on two potential applications, i.e. cross-view pattern discovery and interactive retrieval. The source code is available at \\url{https://github.com/Jeff-Zilence/Explain_Metric_Learning}."
                    },
                    {
                        "title": "Cross-directional Feature Fusion Network for Building Damage Assessment from Satellite Imagery",
                        "abstract": "Fast and effective responses are required when a natural disaster (e.g., earthquake, hurricane, etc.) strikes. Building damage assessment from satellite imagery is critical before an effective response is conducted. High-resolution satellite images provide rich information with pre- and post-disaster scenes for analysis. However, most existing works simply use pre- and post-disaster images as input without considering their correlations. In this paper, we propose a novel cross-directional fusion strategy to better explore the correlations between pre- and post-disaster images. Moreover, the data augmentation method CutMix is exploited to tackle the challenge of hard classes. The proposed method achieves state-of-the-art performance on a large-scale building damage assessment dataset -- xBD."
                    },
                    {
                        "title": "Towards Resolving the Challenge of Long-tail Distribution in UAV Images for Object Detection",
                        "abstract": "Existing methods for object detection in UAV images ignored an important challenge - imbalanced class distribution in UAV images - which leads to poor performance on tail classes. We systematically investigate existing solutions to long-tail problems and unveil that re-balancing methods that are effective on natural image datasets cannot be trivially applied to UAV datasets. To this end, we rethink long-tailed object detection in UAV images and propose the Dual Sampler and Head detection Network (DSHNet), which is the first work that aims to resolve long-tail distribution in UAV images. The key components in DSHNet include Class-Biased Samplers (CBS) and Bilateral Box Heads (BBH), which are developed to cope with tail classes and head classes in a dual-path manner. Without bells and whistles, DSHNet significantly boosts the performance of tail classes on different detection frameworks. Moreover, DSHNet significantly outperforms base detectors and generic approaches for long-tail problems on VisDrone and UAVDT datasets. It achieves new state-of-the-art performance when combining with image cropping methods. Code is available at https://github.com/we1pingyu/DSHNet"
                    },
                    {
                        "title": "A Large-scale Study of Spatiotemporal Representation Learning with a New Benchmark on Action Recognition",
                        "abstract": "The goal of building a benchmark (suite of datasets) is to provide a unified protocol for fair evaluation and thus facilitate the evolution of a specific area. Nonetheless, we point out that existing protocols of action recognition could yield partial evaluations due to several limitations. To comprehensively probe the effectiveness of spatiotemporal representation learning, we introduce BEAR, a new BEnchmark on video Action Recognition. BEAR is a collection of 18 video datasets grouped into 5 categories (anomaly, gesture, daily, sports, and instructional), which covers a diverse set of real-world applications. With BEAR, we thoroughly evaluate 6 common spatiotemporal models pre-trained by both supervised and self-supervised learning. We also report transfer performance via standard finetuning, few-shot finetuning, and unsupervised domain adaptation. Our observation suggests that current state-of-the-art cannot solidly guarantee high performance on datasets close to real-world applications, and we hope BEAR can serve as a fair and challenging evaluation benchmark to gain insights on building next-generation spatiotemporal learners. Our dataset, code, and models are released at: https://github.com/AndongDeng/BEAR"
                    },
                    {
                        "title": "VIGOR: Cross-View Image Geo-localization beyond One-to-one Retrieval",
                        "abstract": "Cross-view image geo-localization aims to determine the locations of street-view query images by matching with GPS-tagged reference images from aerial view. Recent works have achieved surprisingly high retrieval accuracy on city-scale datasets. However, these results rely on the assumption that there exists a reference image exactly centered at the location of any query image, which is not applicable for practical scenarios. In this paper, we redefine this problem with a more realistic assumption that the query image can be arbitrary in the area of interest and the reference images are captured before the queries emerge. This assumption breaks the one-to-one retrieval setting of existing datasets as the queries and reference images are not perfectly aligned pairs, and there may be multiple reference images covering one query location. To bridge the gap between this realistic setting and existing datasets, we propose a new large-scale benchmark -- VIGOR -- for cross-View Image Geo-localization beyond One-to-one Retrieval. We benchmark existing state-of-the-art methods and propose a novel end-to-end framework to localize the query in a coarse-to-fine manner. Apart from the image-level retrieval accuracy, we also evaluate the localization accuracy in terms of the actual distance (meters) using the raw GPS data. Extensive experiments are conducted under different application scenarios to validate the effectiveness of the proposed method. The results indicate that cross-view geo-localization in this realistic setting is still challenging, fostering new research in this direction. Our dataset and code will be released at \\url{https://github.com/Jeff-Zilence/VIGOR}"
                    },
                    {
                        "title": "Consistency-based Active Learning for Object Detection",
                        "abstract": "Active learning aims to improve the performance of task model by selecting the most informative samples with a limited budget. Unlike most recent works that focused on applying active learning for image classification, we propose an effective Consistency-based Active Learning method for object Detection (CALD), which fully explores the consistency between original and augmented data. CALD has three appealing benefits. (i) CALD is systematically designed by investigating the weaknesses of existing active learning methods, which do not take the unique challenges of object detection into account. (ii) CALD unifies box regression and classification with a single metric, which is not concerned by active learning methods for classification. CALD also focuses on the most informative local region rather than the whole image, which is beneficial for object detection. (iii) CALD not only gauges individual information for sample selection, but also leverages mutual information to encourage a balanced data distribution. Extensive experiments show that CALD significantly outperforms existing state-of-the-art task-agnostic and detection-specific active learning methods on general object detection datasets. Based on the Faster R-CNN detector, CALD consistently surpasses the baseline method (random selection) by 2.9/2.8/0.8 mAP on average on PASCAL VOC 2007, PASCAL VOC 2012, and MS COCO. Code is available at \\url{https://github.com/we1pingyu/CALD}"
                    },
                    {
                        "title": "A3D: Adaptive 3D Networks for Video Action Recognition",
                        "abstract": "This paper presents A3D, an adaptive 3D network that can infer at a wide range of computational constraints with one-time training. Instead of training multiple models in a grid-search manner, it generates good configurations by trading off between network width and spatio-temporal resolution. Furthermore, the computation cost can be adapted after the model is deployed to meet variable constraints, for example, on edge devices. Even under the same computational constraints, the performance of our adaptive networks can be significantly boosted over the baseline counterparts by the mutual training along three dimensions. When a multiple pathway framework, e.g. SlowFast, is adopted, our adaptive method encourages a better trade-off between pathways than manual designs. Extensive experiments on the Kinetics dataset show the effectiveness of the proposed framework. The performance gain is also verified to transfer well between datasets and tasks. Code will be made available."
                    },
                    {
                        "title": "MutualNet: Adaptive ConvNet via Mutual Learning from Network Width and Resolution",
                        "abstract": "We propose the width-resolution mutual learning method (MutualNet) to train a network that is executable at dynamic resource constraints to achieve adaptive accuracy-efficiency trade-offs at runtime. Our method trains a cohort of sub-networks with different widths using different input resolutions to mutually learn multi-scale representations for each sub-network. It achieves consistently better ImageNet top-1 accuracy over the state-of-the-art adaptive network US-Net under different computation constraints, and outperforms the best compound scaled MobileNet in EfficientNet by 1.5%. The superiority of our method is also validated on COCO object detection and instance segmentation as well as transfer learning. Surprisingly, the training strategy of MutualNet can also boost the performance of a single network, which substantially outperforms the powerful AutoAugmentation in both efficiency (GPU search hours: 15000 vs. 0) and accuracy (ImageNet: 77.6% vs. 78.6%). Code is available at \\url{https://github.com/taoyang1122/MutualNet}."
                    },
                    {
                        "title": "AIM: Adapting Image Models for Efficient Video Action Recognition",
                        "abstract": "Recent vision transformer based video models mostly follow the ``image pre-training then finetuning\" paradigm and have achieved great success on multiple video benchmarks. However, full finetuning such a video model could be computationally expensive and unnecessary, given the pre-trained image transformer models have demonstrated exceptional transferability. In this work, we propose a novel method to Adapt pre-trained Image Models (AIM) for efficient video understanding. By freezing the pre-trained image model and adding a few lightweight Adapters, we introduce spatial adaptation, temporal adaptation and joint adaptation to gradually equip an image model with spatiotemporal reasoning capability. We show that our proposed AIM can achieve competitive or even better performance than prior arts with substantially fewer tunable parameters on four video action recognition benchmarks. Thanks to its simplicity, our method is also generally applicable to different image pre-trained models, which has the potential to leverage more powerful image foundation models in the future. The project webpage is \\url{https://adapt-image-models.github.io/}."
                    },
                    {
                        "title": "Density Map Guided Object Detection in Aerial Images",
                        "abstract": "Object detection in high-resolution aerial images is a challenging task because of 1) the large variation in object size, and 2) non-uniform distribution of objects. A common solution is to divide the large aerial image into small (uniform) crops and then apply object detection on each small crop. In this paper, we investigate the image cropping strategy to address these challenges. Specifically, we propose a Density-Map guided object detection Network (DMNet), which is inspired from the observation that the object density map of an image presents how objects distribute in terms of the pixel intensity of the map. As pixel intensity varies, it is able to tell whether a region has objects or not, which in turn provides guidance for cropping images statistically. DMNet has three key components: a density map generation module, an image cropping module and an object detector. DMNet generates a density map and learns scale information based on density intensities to form cropping regions. Extensive experiments show that DMNet achieves state-of-the-art performance on two popular aerial image datasets, i.e. VisionDrone and UAVDT."
                    },
                    {
                        "title": "3D Human Pose Estimation with Spatial and Temporal Transformers",
                        "abstract": "Transformer architectures have become the model of choice in natural language processing and are now being introduced into computer vision tasks such as image classification, object detection, and semantic segmentation. However, in the field of human pose estimation, convolutional architectures still remain dominant. In this work, we present PoseFormer, a purely transformer-based approach for 3D human pose estimation in videos without convolutional architectures involved. Inspired by recent developments in vision transformers, we design a spatial-temporal transformer structure to comprehensively model the human joint relations within each frame as well as the temporal correlations across frames, then output an accurate 3D human pose of the center frame. We quantitatively and qualitatively evaluate our method on two popular and standard benchmark datasets: Human3.6M and MPI-INF-3DHP. Extensive experiments show that PoseFormer achieves state-of-the-art performance on both datasets. Code is available at \\url{https://github.com/zczcwh/PoseFormer}"
                    },
                    {
                        "title": "Local Learning Matters: Rethinking Data Heterogeneity in Federated Learning",
                        "abstract": "Federated learning (FL) is a promising strategy for performing privacy-preserving, distributed learning with a network of clients (i.e., edge devices). However, the data distribution among clients is often non-IID in nature, making efficient optimization difficult. To alleviate this issue, many FL algorithms focus on mitigating the effects of data heterogeneity across clients by introducing a variety of proximal terms, some incurring considerable compute and/or memory overheads, to restrain local updates with respect to the global model. Instead, we consider rethinking solutions to data heterogeneity in FL with a focus on local learning generality rather than proximal restriction. To this end, we first present a systematic study informed by second-order indicators to better understand algorithm effectiveness in FL. Interestingly, we find that standard regularization methods are surprisingly strong performers in mitigating data heterogeneity effects. Based on our findings, we further propose a simple and effective method, FedAlign, to overcome data heterogeneity and the pitfalls of previous methods. FedAlign achieves competitive accuracy with state-of-the-art FL methods across a variety of settings while minimizing computation and memory overhead. Code is available at https://github.com/mmendiet/FedAlign"
                    },
                    {
                        "title": "FeatER: An Efficient Network for Human Reconstruction via Feature Map-Based TransformER",
                        "abstract": "Recently, vision transformers have shown great success in a set of human reconstruction tasks such as 2D human pose estimation (2D HPE), 3D human pose estimation (3D HPE), and human mesh reconstruction (HMR) tasks. In these tasks, feature map representations of the human structural information are often extracted first from the image by a CNN (such as HRNet), and then further processed by transformer to predict the heatmaps (encodes each joint's location into a feature map with a Gaussian distribution) for HPE or HMR. However, existing transformer architectures are not able to process these feature map inputs directly, forcing an unnatural flattening of the location-sensitive human structural information. Furthermore, much of the performance benefit in recent HPE and HMR methods has come at the cost of ever-increasing computation and memory needs. Therefore, to simultaneously address these problems, we propose FeatER, a novel transformer design that preserves the inherent structure of feature map representations when modeling attention while reducing memory and computational costs. Taking advantage of FeatER, we build an efficient network for a set of human reconstruction tasks including 2D HPE, 3D HPE, and HMR. A feature map reconstruction module is applied to improve the performance of the estimated human pose and mesh. Extensive experiments demonstrate the effectiveness of FeatER on various human pose and mesh datasets. For instance, FeatER outperforms the SOTA method MeshGraphormer by requiring 5% of Params and 16% of MACs on Human3.6M and 3DPW datasets. The project webpage is https://zczcwh.github.io/feater_page/."
                    },
                    {
                        "title": "Conquering the Communication Constraints to Enable Large Pre-Trained Models in Federated Learning",
                        "abstract": "Federated learning (FL) has emerged as a promising paradigm for enabling the collaborative training of models without centralized access to the raw data on local devices. In the typical FL paradigm (e.g., FedAvg), model weights are sent to and from the server each round to participating clients. Recently, the use of small pre-trained models has been shown effective in federated learning optimization and improving convergence. However, recent state-of-the-art pre-trained models are getting more capable but also have more parameters. In conventional FL, sharing the enormous model weights can quickly put a massive communication burden on the system, especially if more capable models are employed. Can we find a solution to enable those strong and readily-available pre-trained models in FL to achieve excellent performance while simultaneously reducing the communication burden? To this end, we investigate the use of parameter-efficient fine-tuning in federated learning and thus introduce a new framework: FedPEFT. Specifically, we systemically evaluate the performance of FedPEFT across a variety of client stability, data distribution, and differential privacy settings. By only locally tuning and globally sharing a small portion of the model weights, significant reductions in the total communication overhead can be achieved while maintaining competitive or even better performance in a wide range of federated learning scenarios, providing insight into a new paradigm for practical and effective federated systems."
                    },
                    {
                        "title": "ControlNet++: Improving Conditional Controls with Efficient Consistency Feedback",
                        "abstract": "To enhance the controllability of text-to-image diffusion models, existing efforts like ControlNet incorporated image-based conditional controls. In this paper, we reveal that existing methods still face significant challenges in generating images that align with the image conditional controls. To this end, we propose ControlNet++, a novel approach that improves controllable generation by explicitly optimizing pixel-level cycle consistency between generated images and conditional controls. Specifically, for an input conditional control, we use a pre-trained discriminative reward model to extract the corresponding condition of the generated images, and then optimize the consistency loss between the input conditional control and extracted condition. A straightforward implementation would be generating images from random noises and then calculating the consistency loss, but such an approach requires storing gradients for multiple sampling timesteps, leading to considerable time and memory costs. To address this, we introduce an efficient reward strategy that deliberately disturbs the input images by adding noise, and then uses the single-step denoised images for reward fine-tuning. This avoids the extensive costs associated with image sampling, allowing for more efficient reward fine-tuning. Extensive experiments show that ControlNet++ significantly improves controllability under various conditional controls. For example, it achieves improvements over ControlNet by 11.1% mIoU, 13.4% SSIM, and 7.6% RMSE, respectively, for segmentation mask, line-art edge, and depth conditions. All the code, models, demo and organized data have been open sourced on our Github Repo."
                    },
                    {
                        "title": "MutualNet: Adaptive ConvNet via Mutual Learning from Different Model Configurations",
                        "abstract": "Most existing deep neural networks are static, which means they can only do inference at a fixed complexity. But the resource budget can vary substantially across different devices. Even on a single device, the affordable budget can change with different scenarios, and repeatedly training networks for each required budget would be incredibly expensive. Therefore, in this work, we propose a general method called MutualNet to train a single network that can run at a diverse set of resource constraints. Our method trains a cohort of model configurations with various network widths and input resolutions. This mutual learning scheme not only allows the model to run at different width-resolution configurations but also transfers the unique knowledge among these configurations, helping the model to learn stronger representations overall. MutualNet is a general training methodology that can be applied to various network structures (e.g., 2D networks: MobileNets, ResNet, 3D networks: SlowFast, X3D) and various tasks (e.g., image classification, object detection, segmentation, and action recognition), and is demonstrated to achieve consistent improvements on a variety of datasets. Since we only train the model once, it also greatly reduces the training cost compared to independently training several models. Surprisingly, MutualNet can also be used to significantly boost the performance of a single network, if dynamic resource constraint is not a concern. In summary, MutualNet is a unified method for both static and adaptive, 2D and 3D networks. Codes and pre-trained models are available at \\url{https://github.com/taoyang1122/MutualNet}."
                    },
                    {
                        "title": "AutoGluon-Multimodal (AutoMM): Supercharging Multimodal AutoML with Foundation Models",
                        "abstract": "AutoGluon-Multimodal (AutoMM) is introduced as an open-source AutoML library designed specifically for multimodal learning. Distinguished by its exceptional ease of use, AutoMM enables fine-tuning of foundation models with just three lines of code. Supporting various modalities including image, text, and tabular data, both independently and in combination, the library offers a comprehensive suite of functionalities spanning classification, regression, object detection, semantic matching, and image segmentation. Experiments across diverse datasets and tasks showcases AutoMM's superior performance in basic classification and regression tasks compared to existing AutoML tools, while also demonstrating competitive results in advanced tasks, aligning with specialized toolboxes designed for such purposes."
                    }
                ]
            },
            "c0f5d720-d681-4c85-88e8-d2b307f3db08": {
                "pk": "c0f5d720-d681-4c85-88e8-d2b307f3db08",
                "name": "YuXin Song",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "19818a7b-2071-4a37-9e9e-bead9bf6ad5f": {
                "pk": "19818a7b-2071-4a37-9e9e-bead9bf6ad5f",
                "name": "Mengxi Zhang",
                "collaborators": [
                    "Yiming Liu",
                    "Huanjing Yue",
                    "Jingyu Yang",
                    "Heqing Lian",
                    "Jie Chen",
                    "Xiangjun Yin",
                    "Chunyang Bi",
                    "Xin Luo",
                    "Sheng Shen",
                    "Miao Zhu"
                ],
                "domain": [
                    "Image Segmentation",
                    "Multi-Modal Learning",
                    "Generative Models",
                    "Medical Imaging"
                ],
                "publications": [
                    {
                        "title": "HARIS: Human-Like Attention for Reference Image Segmentation",
                        "abstract": "Referring image segmentation (RIS) aims to locate the particular region corresponding to the language expression. Existing methods incorporate features from different modalities in a \\emph{bottom-up} manner. This design may get some unnecessary image-text pairs, which leads to an inaccurate segmentation mask. In this paper, we propose a referring image segmentation method called HARIS, which introduces the Human-Like Attention mechanism and uses the parameter-efficient fine-tuning (PEFT) framework. To be specific, the Human-Like Attention gets a \\emph{feedback} signal from multi-modal features, which makes the network center on the specific objects and discard the irrelevant image-text pairs. Besides, we introduce the PEFT framework to preserve the zero-shot ability of pre-trained encoders. Extensive experiments on three widely used RIS benchmarks and the PhraseCut dataset demonstrate that our method achieves state-of-the-art performance and great zero-shot ability."
                    },
                    {
                        "title": "RISAM: Referring Image Segmentation via Mutual-Aware Attention Features",
                        "abstract": "Referring image segmentation (RIS) aims to segment a particular region based on a language expression prompt. Existing methods incorporate linguistic features into visual features and obtain multi-modal features for mask decoding. However, these methods may segment the visually salient entity instead of the correct referring region, as the multi-modal features are dominated by the abundant visual context. In this paper, we propose MARIS, a referring image segmentation method that leverages the Segment Anything Model (SAM) and introduces a mutual-aware attention mechanism to enhance the cross-modal fusion via two parallel branches. Specifically, our mutual-aware attention mechanism consists of Vision-Guided Attention and Language-Guided Attention, which bidirectionally model the relationship between visual and linguistic features. Correspondingly, we design a Mask Decoder to enable explicit linguistic guidance for more consistent segmentation with the language expression. To this end, a multi-modal query token is proposed to integrate linguistic information and interact with visual information simultaneously. Extensive experiments on three benchmark datasets show that our method outperforms the state-of-the-art RIS methods. Our code will be publicly available."
                    },
                    {
                        "title": "DeeDSR: Towards Real-World Image Super-Resolution via Degradation-Aware Stable Diffusion",
                        "abstract": "Diffusion models, known for their powerful generative capabilities, play a crucial role in addressing real-world super-resolution challenges. However, these models often focus on improving local textures while neglecting the impacts of global degradation, which can significantly reduce semantic fidelity and lead to inaccurate reconstructions and suboptimal super-resolution performance. To address this issue, we introduce a novel two-stage, degradation-aware framework that enhances the diffusion model's ability to recognize content and degradation in low-resolution images. In the first stage, we employ unsupervised contrastive learning to obtain representations of image degradations. In the second stage, we integrate a degradation-aware module into a simplified ControlNet, enabling flexible adaptation to various degradations based on the learned representations. Furthermore, we decompose the degradation-aware features into global semantics and local details branches, which are then injected into the diffusion denoising module to modulate the target generation. Our method effectively recovers semantically precise and photorealistic details, particularly under significant degradation conditions, demonstrating state-of-the-art performance across various benchmarks. Codes will be released at https://github.com/bichunyang419/DeeDSR."
                    },
                    {
                        "title": "RT-SRTS: Angle-Agnostic Real-Time Simultaneous 3D Reconstruction and Tumor Segmentation from Single X-Ray Projection",
                        "abstract": "Radiotherapy is one of the primary treatment methods for tumors, but the organ movement caused by respiration limits its accuracy. Recently, 3D imaging from a single X-ray projection has received extensive attention as a promising approach to address this issue. However, current methods can only reconstruct 3D images without directly locating the tumor and are only validated for fixed-angle imaging, which fails to fully meet the requirements of motion control in radiotherapy. In this study, a novel imaging method RT-SRTS is proposed which integrates 3D imaging and tumor segmentation into one network based on multi-task learning (MTL) and achieves real-time simultaneous 3D reconstruction and tumor segmentation from a single X-ray projection at any angle. Furthermore, the attention enhanced calibrator (AEC) and uncertain-region elaboration (URE) modules have been proposed to aid feature extraction and improve segmentation accuracy. The proposed method was evaluated on fifteen patient cases and compared with three state-of-the-art methods. It not only delivers superior 3D reconstruction but also demonstrates commendable tumor segmentation results. Simultaneous reconstruction and segmentation can be completed in approximately 70 ms, significantly faster than the required time threshold for real-time tumor tracking. The efficacies of both AEC and URE have also been validated in ablation studies. The code of work is available at https://github.com/ZywooSimple/RT-SRTS."
                    },
                    {
                        "title": "GPT4Vis: What Can GPT-4 Do for Zero-shot Visual Recognition?",
                        "abstract": "This paper does not present a novel method. Instead, it delves into an essential, yet must-know baseline in light of the latest advancements in Generative Artificial Intelligence (GenAI): the utilization of GPT-4 for visual understanding. Our study centers on the evaluation of GPT-4's linguistic and visual capabilities in zero-shot visual recognition tasks: Firstly, we explore the potential of its generated rich textual descriptions across various categories to enhance recognition performance without any training. Secondly, we evaluate GPT-4's visual proficiency in directly recognizing diverse visual content. We conducted extensive experiments to systematically evaluate GPT-4's performance across images, videos, and point clouds, using 16 benchmark datasets to measure top-1 and top-5 accuracy. Our findings show that GPT-4, enhanced with rich linguistic descriptions, significantly improves zero-shot recognition, offering an average top-1 accuracy increase of 7% across all datasets. GPT-4 excels in visual recognition, outshining OpenAI-CLIP's ViT-L and rivaling EVA-CLIP's ViT-E, particularly in video datasets HMDB-51 and UCF-101, where it leads by 22% and 9%, respectively. We hope this research contributes valuable data points and experience for future studies. We release our code at https://github.com/whwu95/GPT4Vis."
                    }
                ]
            },
            "5467b093-867b-439a-a8de-4c033f5c8c11": {
                "pk": "5467b093-867b-439a-a8de-4c033f5c8c11",
                "name": "Haocheng Feng",
                "collaborators": [
                    "Errui Ding",
                    "Jingdong Wang",
                    "Junyu Han",
                    "Jian Wang",
                    "Jingtuo Liu",
                    "Chen Zhao",
                    "Haixiao Yue",
                    "Keyao Wang",
                    "Wenyu Liu",
                    "Jinbo Wu"
                ],
                "domain": [
                    "Computer Vision",
                    "Semantic Segmentation",
                    "Domain Adaptation",
                    "Action Recognition"
                ],
                "publications": [
                    {
                        "title": "RTFormer: Efficient Design for Real-Time Semantic Segmentation with Transformer",
                        "abstract": "Recently, transformer-based networks have shown impressive results in semantic segmentation. Yet for real-time semantic segmentation, pure CNN-based approaches still dominate in this field, due to the time-consuming computation mechanism of transformer. We propose RTFormer, an efficient dual-resolution transformer for real-time semantic segmenation, which achieves better trade-off between performance and efficiency than CNN-based models. To achieve high inference efficiency on GPU-like devices, our RTFormer leverages GPU-Friendly Attention with linear complexity and discards the multi-head mechanism. Besides, we find that cross-resolution attention is more efficient to gather global context information for high-resolution branch by spreading the high level knowledge learned from low-resolution branch. Extensive experiments on mainstream benchmarks demonstrate the effectiveness of our proposed RTFormer, it achieves state-of-the-art on Cityscapes, CamVid and COCOStuff, and shows promising results on ADE20K. Code is available at PaddleSeg: https://github.com/PaddlePaddle/PaddleSeg."
                    },
                    {
                        "title": "Cyclically Disentangled Feature Translation for Face Anti-spoofing",
                        "abstract": "Current domain adaptation methods for face anti-spoofing leverage labeled source domain data and unlabeled target domain data to obtain a promising generalizable decision boundary. However, it is usually difficult for these methods to achieve a perfect domain-invariant liveness feature disentanglement, which may degrade the final classification performance by domain differences in illumination, face category, spoof type, etc. In this work, we tackle cross-scenario face anti-spoofing by proposing a novel domain adaptation method called cyclically disentangled feature translation network (CDFTN). Specifically, CDFTN generates pseudo-labeled samples that possess: 1) source domain-invariant liveness features and 2) target domain-specific content features, which are disentangled through domain adversarial training. A robust classifier is trained based on the synthetic pseudo-labeled images under the supervision of source domain labels. We further extend CDFTN for multi-target domain adaptation by leveraging data from more unlabeled target domains. Extensive experiments on several public datasets demonstrate that our proposed approach significantly outperforms the state of the art."
                    },
                    {
                        "title": "Learning Generalized Spoof Cues for Face Anti-spoofing",
                        "abstract": "Many existing face anti-spoofing (FAS) methods focus on modeling the decision boundaries for some predefined spoof types. However, the diversity of the spoof samples including the unknown ones hinders the effective decision boundary modeling and leads to weak generalization capability. In this paper, we reformulate FAS in an anomaly detection perspective and propose a residual-learning framework to learn the discriminative live-spoof differences which are defined as the spoof cues. The proposed framework consists of a spoof cue generator and an auxiliary classifier. The generator minimizes the spoof cues of live samples while imposes no explicit constraint on those of spoof samples to generalize well to unseen attacks. In this way, anomaly detection is implicitly used to guide spoof cue generation, leading to discriminative feature learning. The auxiliary classifier serves as a spoof cue amplifier and makes the spoof cues more discriminative. We conduct extensive experiments and the experimental results show the proposed method consistently outperforms the state-of-the-art methods. The code will be publicly available at https://github.com/vis-var/lgsc-for-fas."
                    },
                    {
                        "title": "Graph Contrastive Learning for Skeleton-based Action Recognition",
                        "abstract": "In the field of skeleton-based action recognition, current top-performing graph convolutional networks (GCNs) exploit intra-sequence context to construct adaptive graphs for feature aggregation. However, we argue that such context is still \\textit{local} since the rich cross-sequence relations have not been explicitly investigated. In this paper, we propose a graph contrastive learning framework for skeleton-based action recognition (\\textit{SkeletonGCL}) to explore the \\textit{global} context across all sequences. In specific, SkeletonGCL associates graph learning across sequences by enforcing graphs to be class-discriminative, \\emph{i.e.,} intra-class compact and inter-class dispersed, which improves the GCN capacity to distinguish various action patterns. Besides, two memory banks are designed to enrich cross-sequence context from two complementary levels, \\emph{i.e.,} instance and semantic levels, enabling graph contrastive learning in multiple context scales. Consequently, SkeletonGCL establishes a new training paradigm, and it can be seamlessly incorporated into current GCNs. Without loss of generality, we combine SkeletonGCL with three GCNs (2S-ACGN, CTR-GCN, and InfoGCN), and achieve consistent improvements on NTU60, NTU120, and NW-UCLA benchmarks. The source code will be available at \\url{https://github.com/OliverHxh/SkeletonGCL}."
                    },
                    {
                        "title": "PSVT: End-to-End Multi-person 3D Pose and Shape Estimation with Progressive Video Transformers",
                        "abstract": "Existing methods of multi-person video 3D human Pose and Shape Estimation (PSE) typically adopt a two-stage strategy, which first detects human instances in each frame and then performs single-person PSE with temporal model. However, the global spatio-temporal context among spatial instances can not be captured. In this paper, we propose a new end-to-end multi-person 3D Pose and Shape estimation framework with progressive Video Transformer, termed PSVT. In PSVT, a spatio-temporal encoder (STE) captures the global feature dependencies among spatial objects. Then, spatio-temporal pose decoder (STPD) and shape decoder (STSD) capture the global dependencies between pose queries and feature tokens, shape queries and feature tokens, respectively. To handle the variances of objects as time proceeds, a novel scheme of progressive decoding is used to update pose and shape queries at each frame. Besides, we propose a novel pose-guided attention (PGA) for shape decoder to better predict shape parameters. The two components strengthen the decoder of PSVT to improve performance. Extensive experiments on the four datasets show that PSVT achieves stage-of-the-art results."
                    },
                    {
                        "title": "HD-Fusion: Detailed Text-to-3D Generation Leveraging Multiple Noise Estimation",
                        "abstract": "In this paper, we study Text-to-3D content generation leveraging 2D diffusion priors to enhance the quality and detail of the generated 3D models. Recent progress (Magic3D) in text-to-3D has shown that employing high-resolution (e.g., 512 x 512) renderings can lead to the production of high-quality 3D models using latent diffusion priors. To enable rendering at even higher resolutions, which has the potential to further augment the quality and detail of the models, we propose a novel approach that combines multiple noise estimation processes with a pretrained 2D diffusion prior. Distinct from the Bar-Tal et al.s' study which binds multiple denoised results to generate images from texts, our approach integrates the computation of scoring distillation losses such as SDS loss and VSD loss which are essential techniques for the 3D content generation with 2D diffusion priors. We experimentally evaluated the proposed approach. The results show that the proposed approach can generate high-quality details compared to the baselines."
                    },
                    {
                        "title": "GVA: Reconstructing Vivid 3D Gaussian Avatars from Monocular Videos",
                        "abstract": "In this paper, we present a novel method that facilitates the creation of vivid 3D Gaussian avatars from monocular video inputs (GVA). Our innovation lies in addressing the intricate challenges of delivering high-fidelity human body reconstructions and aligning 3D Gaussians with human skin surfaces accurately. The key contributions of this paper are twofold. Firstly, we introduce a pose refinement technique to improve hand and foot pose accuracy by aligning normal maps and silhouettes. Precise pose is crucial for correct shape and appearance reconstruction. Secondly, we address the problems of unbalanced aggregation and initialization bias that previously diminished the quality of 3D Gaussian avatars, through a novel surface-guided re-initialization method that ensures accurate alignment of 3D Gaussian points with avatar surfaces. Experimental results demonstrate that our proposed method achieves high-fidelity and vivid 3D Gaussian avatar reconstruction. Extensive experimental analyses validate the performance qualitatively and quantitatively, demonstrating that it achieves state-of-the-art performance in photo-realistic novel view synthesis while offering fine-grained control over the human body and hand pose. Project page: https://3d-aigc.github.io/GVA/."
                    },
                    {
                        "title": "VDG: Vision-Only Dynamic Gaussian for Driving Simulation",
                        "abstract": "Dynamic Gaussian splatting has led to impressive scene reconstruction and image synthesis advances in novel views. Existing methods, however, heavily rely on pre-computed poses and Gaussian initialization by Structure from Motion (SfM) algorithms or expensive sensors. For the first time, this paper addresses this issue by integrating self-supervised VO into our pose-free dynamic Gaussian method (VDG) to boost pose and depth initialization and static-dynamic decomposition. Moreover, VDG can work with only RGB image input and construct dynamic scenes at a faster speed and larger scenes compared with the pose-free dynamic view-synthesis method. We demonstrate the robustness of our approach via extensive quantitative and qualitative experiments. Our results show favorable performance over the state-of-the-art dynamic view synthesis methods. Additional video and source code will be posted on our project page at https://3d-aigc.github.io/VDG."
                    },
                    {
                        "title": "MonoFormer: One Transformer for Both Diffusion and Autoregression",
                        "abstract": "Most existing multimodality methods use separate backbones for autoregression-based discrete text generation and diffusion-based continuous visual generation, or the same backbone by discretizing the visual data to use autoregression for both text and visual generation. In this paper, we propose to study a simple idea: share one transformer for both autoregression and diffusion. The feasibility comes from two main aspects: (i) Transformer is successfully applied to diffusion for visual generation, and (ii) transformer training for autoregression and diffusion is very similar, and the difference merely lies in that diffusion uses bidirectional attention mask and autoregression uses causal attention mask. Experimental results show that our approach achieves comparable image generation performance to current state-of-the-art methods as well as maintains the text generation capability. The project is publicly available at https://monoformer.github.io/."
                    },
                    {
                        "title": "Dynamic Class Queue for Large Scale Face Recognition In the Wild",
                        "abstract": "Learning discriminative representation using large-scale face datasets in the wild is crucial for real-world applications, yet it remains challenging. The difficulties lie in many aspects and this work focus on computing resource constraint and long-tailed class distribution. Recently, classification-based representation learning with deep neural networks and well-designed losses have demonstrated good recognition performance. However, the computing and memory cost linearly scales up to the number of identities (classes) in the training set, and the learning process suffers from unbalanced classes. In this work, we propose a dynamic class queue (DCQ) to tackle these two problems. Specifically, for each iteration during training, a subset of classes for recognition are dynamically selected and their class weights are dynamically generated on-the-fly which are stored in a queue. Since only a subset of classes is selected for each iteration, the computing requirement is reduced. By using a single server without model parallel, we empirically verify in large-scale datasets that 10% of classes are sufficient to achieve similar performance as using all classes. Moreover, the class weights are dynamically generated in a few-shot manner and therefore suitable for tail classes with only a few instances. We show clear improvement over a strong baseline in the largest public dataset Megaface Challenge2 (MF2) which has 672K identities and over 88% of them have less than 10 instances. Code is available at https://github.com/bilylee/DCQ"
                    },
                    {
                        "title": "Singular Value Fine-tuning: Few-shot Segmentation requires Few-parameters Fine-tuning",
                        "abstract": "Freezing the pre-trained backbone has become a standard paradigm to avoid overfitting in few-shot segmentation. In this paper, we rethink the paradigm and explore a new regime: {\\em fine-tuning a small part of parameters in the backbone}. We present a solution to overcome the overfitting problem, leading to better model generalization on learning novel classes. Our method decomposes backbone parameters into three successive matrices via the Singular Value Decomposition (SVD), then {\\em only fine-tunes the singular values} and keeps others frozen. The above design allows the model to adjust feature representations on novel classes while maintaining semantic clues within the pre-trained backbone. We evaluate our {\\em Singular Value Fine-tuning (SVF)} approach on various few-shot segmentation methods with different backbones. We achieve state-of-the-art results on both Pascal-5$^i$ and COCO-20$^i$ across 1-shot and 5-shot settings. Hopefully, this simple baseline will encourage researchers to rethink the role of backbone fine-tuning in few-shot settings. The source code and models will be available at https://github.com/syp2ysy/SVF."
                    },
                    {
                        "title": "Group DETR: Fast DETR Training with Group-Wise One-to-Many Assignment",
                        "abstract": "Detection transformer (DETR) relies on one-to-one assignment, assigning one ground-truth object to one prediction, for end-to-end detection without NMS post-processing. It is known that one-to-many assignment, assigning one ground-truth object to multiple predictions, succeeds in detection methods such as Faster R-CNN and FCOS. While the naive one-to-many assignment does not work for DETR, and it remains challenging to apply one-to-many assignment for DETR training. In this paper, we introduce Group DETR, a simple yet efficient DETR training approach that introduces a group-wise way for one-to-many assignment. This approach involves using multiple groups of object queries, conducting one-to-one assignment within each group, and performing decoder self-attention separately. It resembles data augmentation with automatically-learned object query augmentation. It is also equivalent to simultaneously training parameter-sharing networks of the same architecture, introducing more supervision and thus improving DETR training. The inference process is the same as DETR trained normally and only needs one group of queries without any architecture modification. Group DETR is versatile and is applicable to various DETR variants. The experiments show that Group DETR significantly speeds up the training convergence and improves the performance of various DETR-based models. Code will be available at \\url{https://github.com/Atten4Vis/GroupDETR}."
                    }
                ]
            },
            "69e00d2e-a862-47d8-86ca-2fec716c783c": {
                "pk": "69e00d2e-a862-47d8-86ca-2fec716c783c",
                "name": "Yifan Sun",
                "collaborators": [
                    "Francis Bach",
                    "Mingshang Hu",
                    "Zhengyang Sun",
                    "Xihong Wu",
                    "Yuzhao Zhang"
                ],
                "domain": [
                    "Optimization",
                    "Self-Supervised Learning",
                    "Multi-Task Learning",
                    "Network Analysis"
                ],
                "publications": [
                    {
                        "title": "Explicit positive solutions to $G$-heat equations and the application to $G$-capacities",
                        "abstract": "In this paper, we study a class of explicit positive solutions to $G$-heat equations by solving second order nonlinear ordinary differential equations. Based on the positive solutions, we give the sharp order of $G$-capacity $c_{\\sigma}\\left( \\left[ -\\varepsilon,\\varepsilon \\right] \\right) $ when $\\varepsilon \\rightarrow0$."
                    },
                    {
                        "title": "Target observation of complex networks",
                        "abstract": "How to observe the state of a network from a limited number of measurements has become an important issue in complex networks, engineering, communication, epidemiology, etc. Under some scenarios, it is neither unfeasible nor unnecessary to observe the entire network. Therefore, we investigate the target observation of a network in this paper. We propose a target minimal dominating set problem corresponding to target observation, which is a natural generalization of classical minimal dominating set problem. Three algorithms are proposed to approximate the minimum set of occupied nodes sufficient for target observation. Extensive numerical results on computer-generated random networks and real-world networks demonstrate that the proposed algorithms offer superior performance in identification of a target minimal dominating set."
                    },
                    {
                        "title": "Embodied Self-supervised Learning by Coordinated Sampling and Training",
                        "abstract": "Self-supervised learning can significantly improve the performance of downstream tasks, however, the dimensions of learned representations normally lack explicit physical meanings. In this work, we propose a novel self-supervised approach to solve inverse problems by employing the corresponding physical forward process so that the learned representations can have explicit physical meanings. The proposed approach works in an analysis-by-synthesis manner to learn an inference network by iteratively sampling and training. At the sampling step, given observed data, the inference network is used to approximate the intractable posterior, from which we sample input parameters and feed them to a physical process to generate data in the observational space; At the training step, the same network is optimized with the sampled paired data. We prove the feasibility of the proposed method by tackling the acoustic-to-articulatory inversion problem to infer articulatory information from speech. Given an articulatory synthesizer, an inference model can be trained completely from scratch with random initialization. Our experiments demonstrate that the proposed method can converge steadily and the network learns to control the articulatory synthesizer to speak like a human. We also demonstrate that trained models can generalize well to unseen speakers or even new languages, and performance can be further improved through self-adaptation."
                    },
                    {
                        "title": "Safe Screening for the Generalized Conditional Gradient Method",
                        "abstract": "The conditional gradient method (CGM) has been widely used for fast sparse approximation, having a low per iteration computational cost for structured sparse regularizers. We explore the sparsity acquiring properties of a generalized CGM (gCGM), where the constraint is replaced by a penalty function based on a gauge penalty; this can be done without significantly increasing the per-iteration computation, and applies to general notions of sparsity. Without assuming bounded iterates, we show $O(1/t)$ convergence of the function values and gap of gCGM. We couple this with a safe screening rule, and show that at a rate $O(1/(t\\delta^2))$, the screened support matches the support at the solution, where $\\delta \\geq 0$ measures how close the problem is to being degenerate. In our experiments, we show that the gCGM for these modified penalties have similar feature selection properties as common penalties, but with potentially more stability over the choice of hyperparameter."
                    },
                    {
                        "title": "Screening for a Reweighted Penalized Conditional Gradient Method",
                        "abstract": "The conditional gradient method (CGM) is widely used in large-scale sparse convex optimization, having a low per iteration computational cost for structured sparse regularizers and a greedy approach to collecting nonzeros. We explore the sparsity acquiring properties of a general penalized CGM (P-CGM) for convex regularizers and a reweighted penalized CGM (RP-CGM) for nonconvex regularizers, replacing the usual convex constraints with gauge-inspired penalties. This generalization does not increase the per-iteration complexity noticeably. Without assuming bounded iterates or using line search, we show $O(1/t)$ convergence of the gap of each subproblem, which measures distance to a stationary point. We couple this with a screening rule which is safe in the convex case, converging to the true support at a rate $O(1/(\\delta^2))$ where $\\delta \\geq 0$ measures how close the problem is to degeneracy. In the nonconvex case the screening rule converges to the true support in a finite number of iterations, but is not necessarily safe in the intermediate iterates. In our experiments, we verify the consistency of the method and adjust the aggressiveness of the screening rule by tuning the concavity of the regularizer."
                    },
                    {
                        "title": "Semisoft Task Clustering for Multi-Task Learning",
                        "abstract": "Multi-task learning (MTL) aims to improve the performance of multiple related prediction tasks by leveraging useful information from them. Due to their flexibility and ability to reduce unknown coefficients substantially, the task-clustering-based MTL approaches have attracted considerable attention. Motivated by the idea of semisoft clustering of data, we propose a semisoft task clustering approach, which can simultaneously reveal the task cluster structure for both pure and mixed tasks as well as select the relevant features. The main assumption behind our approach is that each cluster has some pure tasks, and each mixed task can be represented by a linear combination of pure tasks in different clusters. To solve the resulting non-convex constrained optimization problem, we design an efficient three-step algorithm. The experimental results based on synthetic and real-world datasets validate the effectiveness and efficiency of the proposed approach. Finally, we extend the proposed approach to a robust task clustering problem."
                    }
                ]
            },
            "27c73e8b-6446-4fde-a9e0-b81f60793957": {
                "pk": "27c73e8b-6446-4fde-a9e0-b81f60793957",
                "name": "Zhiheng Li",
                "collaborators": [
                    "Chenliang Xu",
                    "Yubo Cui",
                    "Zheng Fang",
                    "Pengzhi Li",
                    "Li Li",
                    "Yu Lin",
                    "Yudong Chen",
                    "Yi Zhang",
                    "Baijuan Li",
                    "Guan Wang"
                ],
                "domain": [
                    "Bias Detection",
                    "Image Processing",
                    "Generative Models",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Discover the Unknown Biased Attribute of an Image Classifier",
                        "abstract": "Recent works find that AI algorithms learn biases from data. Therefore, it is urgent and vital to identify biases in AI algorithms. However, the previous bias identification pipeline overly relies on human experts to conjecture potential biases (e.g., gender), which may neglect other underlying biases not realized by humans. To help human experts better find the AI algorithms' biases, we study a new problem in this work -- for a classifier that predicts a target attribute of the input image, discover its unknown biased attribute.   To solve this challenging problem, we use a hyperplane in the generative model's latent space to represent an image attribute; thus, the original problem is transformed to optimizing the hyperplane's normal vector and offset. We propose a novel total-variation loss within this framework as the objective function and a new orthogonalization penalty as a constraint. The latter prevents trivial solutions in which the discovered biased attribute is identical with the target or one of the known-biased attributes. Extensive experiments on both disentanglement datasets and real-world datasets show that our method can discover biased attributes and achieve better disentanglement w.r.t. target attributes. Furthermore, the qualitative results show that our method can discover unnoticeable biased attributes for various object and scene classifiers, proving our method's generalizability for detecting biased attributes in diverse domains of images. The code is available at https://git.io/J3kMh."
                    },
                    {
                        "title": "Quantization Errors of fGn and fBm Signals",
                        "abstract": "In this Letter, we show that under the assumption of high resolution, the quantization errors of fGn and fBm signals with uniform quantizer can be treated as uncorrelated white noises."
                    },
                    {
                        "title": "Sketch-to-Architecture: Generative AI-aided Architectural Design",
                        "abstract": "Recently, the development of large-scale models has paved the way for various interdisciplinary research, including architecture. By using generative AI, we present a novel workflow that utilizes AI models to generate conceptual floorplans and 3D models from simple sketches, enabling rapid ideation and controlled generation of architectural renderings based on textual descriptions. Our work demonstrates the potential of generative AI in the architectural design process, pointing towards a new direction of computer-aided architectural design. Our project website is available at: https://zrealli.github.io/sketch2arc"
                    },
                    {
                        "title": "Cooperative Lane Changing via Deep Reinforcement Learning",
                        "abstract": "In this paper, we study how to learn an appropriate lane changing strategy for autonomous vehicles by using deep reinforcement learning. We show that the reward of the system should consider the overall traffic efficiency instead of the travel efficiency of an individual vehicle. In summary, cooperation leads to a more harmonic and efficient traffic system rather than competition"
                    },
                    {
                        "title": "Early Prediction of 30-day ICU Re-admissions Using Natural Language Processing and Machine Learning",
                        "abstract": "ICU readmission is associated with longer hospitalization, mortality and adverse outcomes. An early recognition of ICU re-admission can help prevent patients from worse situation and lower treatment cost. As the abundance of Electronics Health Records (EHR), it is popular to design clinical decision tools with machine learning technique manipulating on healthcare large scale data. We designed data-driven predictive models to estimate the risk of ICU readmission. The discharge summary of each hospital admission was carefully represented by natural language processing techniques. Unified Medical Language System (UMLS) was further used to standardize inconsistency of discharge summaries. 5 machine learning classifiers were adopted to construct predictive models. The best configuration yielded a competitive AUC of 0.748. Our work suggests that natural language processing of discharge summaries is capable to send clinicians warning of unplanned 30-day readmission upon discharge."
                    },
                    {
                        "title": "LayerDiffusion: Layered Controlled Image Editing with Diffusion Models",
                        "abstract": "Text-guided image editing has recently experienced rapid development. However, simultaneously performing multiple editing actions on a single image, such as background replacement and specific subject attribute changes, while maintaining consistency between the subject and the background remains challenging. In this paper, we propose LayerDiffusion, a semantic-based layered controlled image editing method. Our method enables non-rigid editing and attribute modification of specific subjects while preserving their unique characteristics and seamlessly integrating them into new backgrounds. We leverage a large-scale text-to-image model and employ a layered controlled optimization strategy combined with layered diffusion training. During the diffusion process, an iterative guidance strategy is used to generate a final image that aligns with the textual description. Experimental results demonstrate the effectiveness of our method in generating highly coherent images that closely align with the given textual description. The edited images maintain a high similarity to the features of the input image and surpass the performance of current leading image editing methods. LayerDiffusion opens up new possibilities for controllable image editing."
                    },
                    {
                        "title": "ART3D: 3D Gaussian Splatting for Text-Guided Artistic Scenes Generation",
                        "abstract": "In this paper, we explore the existing challenges in 3D artistic scene generation by introducing ART3D, a novel framework that combines diffusion models and 3D Gaussian splatting techniques. Our method effectively bridges the gap between artistic and realistic images through an innovative image semantic transfer algorithm. By leveraging depth information and an initial artistic image, we generate a point cloud map, addressing domain differences. Additionally, we propose a depth consistency module to enhance 3D scene consistency. Finally, the 3D scene serves as initial points for optimizing Gaussian splats. Experimental results demonstrate ART3D's superior performance in both content and structural consistency metrics when compared to existing methods. ART3D significantly advances the field of AI in art creation by providing an innovative solution for generating high-quality 3D artistic scenes."
                    },
                    {
                        "title": "Deep Grouping Model for Unified Perceptual Parsing",
                        "abstract": "The perceptual-based grouping process produces a hierarchical and compositional image representation that helps both human and machine vision systems recognize heterogeneous visual concepts. Examples can be found in the classical hierarchical superpixel segmentation or image parsing works. However, the grouping process is largely overlooked in modern CNN-based image segmentation networks due to many challenges, including the inherent incompatibility between the grid-shaped CNN feature map and the irregular-shaped perceptual grouping hierarchy. Overcoming these challenges, we propose a deep grouping model (DGM) that tightly marries the two types of representations and defines a bottom-up and a top-down process for feature exchanging. When evaluating the model on the recent Broden+ dataset for the unified perceptual parsing task, it achieves state-of-the-art results while having a small computational overhead compared to other contextual-based segmentation models. Furthermore, the DGM has better interpretability compared with modern CNN methods."
                    },
                    {
                        "title": "StreamMOS: Streaming Moving Object Segmentation with Multi-View Perception and Dual-Span Memory",
                        "abstract": "Moving object segmentation based on LiDAR is a crucial and challenging task for autonomous driving and mobile robotics. Most approaches explore spatio-temporal information from LiDAR sequences to predict moving objects in the current frame. However, they often focus on transferring temporal cues in a single inference and regard every prediction as independent of others. This may cause inconsistent segmentation results for the same object in different frames. To overcome this issue, we propose a streaming network with a memory mechanism, called StreamMOS, to build the association of features and predictions among multiple inferences. Specifically, we utilize a short-term memory to convey historical features, which can be regarded as spatial prior of moving objects and adopted to enhance current inference by temporal fusion. Meanwhile, we build a long-term memory to store previous predictions and exploit them to refine the present forecast at voxel and instance levels through voting. Besides, we present multi-view encoder with cascade projection and asymmetric convolution to extract motion feature of objects in different representations. Extensive experiments validate that our algorithm gets competitive performance on SemanticKITTI and Sipailou Campus datasets. Code will be released at https://github.com/NEU-REAL/StreamMOS.git."
                    },
                    {
                        "title": "Discover and Mitigate Unknown Biases with Debiasing Alternate Networks",
                        "abstract": "Deep image classifiers have been found to learn biases from datasets. To mitigate the biases, most previous methods require labels of protected attributes (e.g., age, skin tone) as full-supervision, which has two limitations: 1) it is infeasible when the labels are unavailable; 2) they are incapable of mitigating unknown biases -- biases that humans do not preconceive. To resolve those problems, we propose Debiasing Alternate Networks (DebiAN), which comprises two networks -- a Discoverer and a Classifier. By training in an alternate manner, the discoverer tries to find multiple unknown biases of the classifier without any annotations of biases, and the classifier aims at unlearning the biases identified by the discoverer. While previous works evaluate debiasing results in terms of a single bias, we create Multi-Color MNIST dataset to better benchmark mitigation of multiple biases in a multi-bias setting, which not only reveals the problems in previous methods but also demonstrates the advantage of DebiAN in identifying and mitigating multiple biases simultaneously. We further conduct extensive experiments on real-world datasets, showing that the discoverer in DebiAN can identify unknown biases that may be hard to be found by humans. Regarding debiasing, DebiAN achieves strong bias mitigation performance."
                    },
                    {
                        "title": "MMF-Track: Multi-modal Multi-level Fusion for 3D Single Object Tracking",
                        "abstract": "3D single object tracking plays a crucial role in computer vision. Mainstream methods mainly rely on point clouds to achieve geometry matching between target template and search area. However, textureless and incomplete point clouds make it difficult for single-modal trackers to distinguish objects with similar structures. To overcome the limitations of geometry matching, we propose a Multi-modal Multi-level Fusion Tracker (MMF-Track), which exploits the image texture and geometry characteristic of point clouds to track 3D target. Specifically, we first propose a Space Alignment Module (SAM) to align RGB images with point clouds in 3D space, which is the prerequisite for constructing inter-modal associations. Then, in feature interaction level, we design a Feature Interaction Module (FIM) based on dual-stream structure, which enhances intra-modal features in parallel and constructs inter-modal semantic associations. Meanwhile, in order to refine each modal feature, we introduce a Coarse-to-Fine Interaction Module (CFIM) to realize the hierarchical feature interaction at different scales. Finally, in similarity fusion level, we propose a Similarity Fusion Module (SFM) to aggregate geometry and texture clues from the target. Experiments show that our method achieves state-of-the-art performance on KITTI (39% Success and 42% Precision gains against previous multi-modal method) and is also competitive on NuScenes."
                    },
                    {
                        "title": "STTracker: Spatio-Temporal Tracker for 3D Single Object Tracking",
                        "abstract": "3D single object tracking with point clouds is a critical task in 3D computer vision. Previous methods usually input the last two frames and use the predicted box to get the template point cloud in previous frame and the search area point cloud in the current frame respectively, then use similarity-based or motion-based methods to predict the current box. Although these methods achieved good tracking performance, they ignore the historical information of the target, which is important for tracking. In this paper, compared to inputting two frames of point clouds, we input multi-frame of point clouds to encode the spatio-temporal information of the target and learn the motion information of the target implicitly, which could build the correlations among different frames to track the target in the current frame efficiently. Meanwhile, rather than directly using the point feature for feature fusion, we first crop the point cloud features into many patches and then use sparse attention mechanism to encode the patch-level similarity and finally fuse the multi-frame features. Extensive experiments show that our method achieves competitive results on challenging large-scale benchmarks (62.6% in KITTI and 49.66% in NuScenes)."
                    },
                    {
                        "title": "SeqTrack3D: Exploring Sequence Information for Robust 3D Point Cloud Tracking",
                        "abstract": "3D single object tracking (SOT) is an important and challenging task for the autonomous driving and mobile robotics. Most existing methods perform tracking between two consecutive frames while ignoring the motion patterns of the target over a series of frames, which would cause performance degradation in the scenes with sparse points. To break through this limitation, we introduce Sequence-to-Sequence tracking paradigm and a tracker named SeqTrack3D to capture target motion across continuous frames. Unlike previous methods that primarily adopted three strategies: matching two consecutive point clouds, predicting relative motion, or utilizing sequential point clouds to address feature degradation, our SeqTrack3D combines both historical point clouds and bounding box sequences. This novel method ensures robust tracking by leveraging location priors from historical boxes, even in scenes with sparse points. Extensive experiments conducted on large-scale datasets show that SeqTrack3D achieves new state-of-the-art performances, improving by 6.00% on NuScenes and 14.13% on Waymo dataset. The code will be made public at https://github.com/aron-lin/seqtrack3d."
                    },
                    {
                        "title": "Side4Video: Spatial-Temporal Side Network for Memory-Efficient Image-to-Video Transfer Learning",
                        "abstract": "Large pre-trained vision models achieve impressive success in computer vision. However, fully fine-tuning large models for downstream tasks, particularly in video understanding, can be prohibitively computationally expensive. Recent studies turn their focus towards efficient image-to-video transfer learning. Nevertheless, existing efficient fine-tuning methods lack attention to training memory usage and exploration of transferring a larger model to the video domain. In this paper, we present a novel Spatial-Temporal Side Network for memory-efficient fine-tuning large image models to video understanding, named Side4Video. Specifically, we introduce a lightweight spatial-temporal side network attached to the frozen vision model, which avoids the backpropagation through the heavy pre-trained model and utilizes multi-level spatial features from the original image model. Extremely memory-efficient architecture enables our method to reduce 75% memory usage than previous adapter-based methods. In this way, we can transfer a huge ViT-E (4.4B) for video understanding tasks which is 14x larger than ViT-L (304M). Our approach achieves remarkable performance on various video datasets across unimodal and cross-modal tasks (i.e., action recognition and text-video retrieval), especially in Something-Something V1&V2 (67.3% & 74.6%), Kinetics-400 (88.6%), MSR-VTT (52.3%), MSVD (56.1%) and VATEX (68.8%). We release our code at https://github.com/HJYao00/Side4Video."
                    },
                    {
                        "title": "Discover and Mitigate Multiple Biased Subgroups in Image Classifiers",
                        "abstract": "Machine learning models can perform well on in-distribution data but often fail on biased subgroups that are underrepresented in the training data, hindering the robustness of models for reliable applications. Such subgroups are typically unknown due to the absence of subgroup labels. Discovering biased subgroups is the key to understanding models' failure modes and further improving models' robustness. Most previous works of subgroup discovery make an implicit assumption that models only underperform on a single biased subgroup, which does not hold on in-the-wild data where multiple biased subgroups exist.   In this work, we propose Decomposition, Interpretation, and Mitigation (DIM), a novel method to address a more challenging but also more practical problem of discovering multiple biased subgroups in image classifiers. Our approach decomposes the image features into multiple components that represent multiple subgroups. This decomposition is achieved via a bilinear dimension reduction method, Partial Least Square (PLS), guided by useful supervision from the image classifier. We further interpret the semantic meaning of each subgroup component by generating natural language descriptions using vision-language foundation models. Finally, DIM mitigates multiple biased subgroups simultaneously via two strategies, including the data- and model-centric strategies. Extensive experiments on CIFAR-100 and Breeds datasets demonstrate the effectiveness of DIM in discovering and mitigating multiple biased subgroups. Furthermore, DIM uncovers the failure modes of the classifier on Hard ImageNet, showcasing its broader applicability to understanding model bias in image classifiers. The code is available at https://github.com/ZhangAIPI/DIM."
                    },
                    {
                        "title": "On Continuity Properties for Infinite Rectangle Packing",
                        "abstract": "By rectangle packing we mean putting a set of rectangles into an enclosing rectangle, without any overlapping. We begin with perfect rectangle packing problems, then prove two continuity properties for parallel rectangle packing problems, and discuss how they might be used to obtain negative results for perfect rectangle packing problems."
                    }
                ]
            },
            "d0140624-6171-4d0b-b64e-0ec01c173dd1": {
                "pk": "d0140624-6171-4d0b-b64e-0ec01c173dd1",
                "name": "Wanli Ouyang",
                "collaborators": [
                    "Xiaogang Wang",
                    "Hongsheng Li",
                    "Lei Bai",
                    "Yan Huang",
                    "Liang Wang",
                    "Ku Wang",
                    "Xin Zhu",
                    "Cong Zhang",
                    "Xiaokang Yang",
                    "Xingyu Zeng"
                ],
                "domain": [
                    "Computer Vision",
                    "Object Detection",
                    "Deep Learning",
                    "Image Processing"
                ],
                "publications": [
                    {
                        "title": "Learning Chained Deep Features and Classifiers for Cascade in Object Detection",
                        "abstract": "Cascade is a widely used approach that rejects obvious negative samples at early stages for learning better classifier and faster inference. This paper presents chained cascade network (CC-Net). In this CC-Net, the cascaded classifier at a stage is aided by the classification scores in previous stages. Feature chaining is further proposed so that the feature learning for the current cascade stage uses the features in previous stages as the prior information. The chained ConvNet features and classifiers of multiple stages are jointly learned in an end-to-end network. In this way, features and classifiers at latter stages handle more difficult samples with the help of features and classifiers in previous stages. It yields consistent boost in detection performance on benchmarks like PASCAL VOC 2007 and ImageNet. Combined with better region proposal, CC-Net leads to state-of-the-art result of 81.1% mAP on PASCAL VOC 2007."
                    },
                    {
                        "title": "Factors in Finetuning Deep Model for object detection",
                        "abstract": "Finetuning from a pretrained deep model is found to yield state-of-the-art performance for many vision tasks. This paper investigates many factors that influence the performance in finetuning for object detection. There is a long-tailed distribution of sample numbers for classes in object detection. Our analysis and empirical results show that classes with more samples have higher impact on the feature learning. And it is better to make the sample number more uniform across classes. Generic object detection can be considered as multiple equally important tasks. Detection of each class is a task. These classes/tasks have their individuality in discriminative visual appearance representation. Taking this individuality into account, we cluster objects into visually similar class groups and learn deep representations for these groups separately. A hierarchical feature learning scheme is proposed. In this scheme, the knowledge from the group with large number of classes is transferred for learning features in its sub-groups. Finetuned on the GoogLeNet model, experimental results show 4.7% absolute mAP improvement of our approach on the ImageNet object detection dataset without increasing much computational cost at the testing stage."
                    },
                    {
                        "title": "Window-Object Relationship Guided Representation Learning for Generic Object Detections",
                        "abstract": "In existing works that learn representation for object detection, the relationship between a candidate window and the ground truth bounding box of an object is simplified by thresholding their overlap. This paper shows information loss in this simplification and picks up the relative location/size information discarded by thresholding. We propose a representation learning pipeline to use the relationship as supervision for improving the learned representation in object detection. Such relationship is not limited to object of the target category, but also includes surrounding objects of other categories. We show that image regions with multiple contexts and multiple rotations are effective in capturing such relationship during the representation learning process and in handling the semantic and visual variation caused by different window-object configurations. Experimental results show that the representation learned by our approach can improve the object detection accuracy by 6.4% in mean average precision (mAP) on ILSVRC2014. On the challenging ILSVRC2014 test dataset, 48.6% mAP is achieved by our single model and it is the best among published results. On PASCAL VOC, it outperforms the state-of-the-art result of Fast RCNN by 3.3% in absolute mAP."
                    },
                    {
                        "title": "Multi-Bias Non-linear Activation in Deep Neural Networks",
                        "abstract": "As a widely used non-linear activation, Rectified Linear Unit (ReLU) separates noise and signal in a feature map by learning a threshold or bias. However, we argue that the classification of noise and signal not only depends on the magnitude of responses, but also the context of how the feature responses would be used to detect more abstract patterns in higher layers. In order to output multiple response maps with magnitude in different ranges for a particular visual pattern, existing networks employing ReLU and its variants have to learn a large number of redundant filters. In this paper, we propose a multi-bias non-linear activation (MBA) layer to explore the information hidden in the magnitudes of responses. It is placed after the convolution layer to decouple the responses to a convolution kernel into multiple maps by multi-thresholding magnitudes, thus generating more patterns in the feature space at a low computational cost. It provides great flexibility of selecting responses to different visual patterns in different magnitude ranges to form rich representations in higher layers. Such a simple and yet effective scheme achieves the state-of-the-art performance on several benchmarks."
                    },
                    {
                        "title": "Object Detection from Video Tubelets with Convolutional Neural Networks",
                        "abstract": "Deep Convolution Neural Networks (CNNs) have shown impressive performance in various vision tasks such as image classification, object detection and semantic segmentation. For object detection, particularly in still images, the performance has been significantly increased last year thanks to powerful deep networks (e.g. GoogleNet) and detection frameworks (e.g. Regions with CNN features (R-CNN)). The lately introduced ImageNet task on object detection from video (VID) brings the object detection task into the video domain, in which objects' locations at each frame are required to be annotated with bounding boxes. In this work, we introduce a complete framework for the VID task based on still-image object detection and general object tracking. Their relations and contributions in the VID task are thoroughly studied and evaluated. In addition, a temporal convolution network is proposed to incorporate temporal information to regularize the detection results and shows its effectiveness for the task."
                    },
                    {
                        "title": "Learning Deep Feature Representations with Domain Guided Dropout for Person Re-identification",
                        "abstract": "Learning generic and robust feature representations with data from multiple domains for the same problem is of great value, especially for the problems that have multiple datasets but none of them are large enough to provide abundant data variations. In this work, we present a pipeline for learning deep feature representations from multiple domains with Convolutional Neural Networks (CNNs). When training a CNN with data from all the domains, some neurons learn representations shared across several domains, while some others are effective only for a specific one. Based on this important observation, we propose a Domain Guided Dropout algorithm to improve the feature learning procedure. Experiments show the effectiveness of our pipeline and the proposed algorithm. Our methods on the person re-identification problem outperform state-of-the-art methods on multiple datasets by large margins."
                    },
                    {
                        "title": "ViP-CNN: Visual Phrase Guided Convolutional Neural Network",
                        "abstract": "As the intermediate level task connecting image captioning and object detection, visual relationship detection started to catch researchers' attention because of its descriptive power and clear structure. It detects the objects and captures their pair-wise interactions with a subject-predicate-object triplet, e.g. person-ride-horse. In this paper, each visual relationship is considered as a phrase with three components. We formulate the visual relationship detection as three inter-connected recognition problems and propose a Visual Phrase guided Convolutional Neural Network (ViP-CNN) to address them simultaneously. In ViP-CNN, we present a Phrase-guided Message Passing Structure (PMPS) to establish the connection among relationship components and help the model consider the three problems jointly. Corresponding non-maximum suppression method and model training strategy are also proposed. Experimental results show that our ViP-CNN outperforms the state-of-art method both in speed and accuracy. We further pretrain ViP-CNN on our cleansed Visual Genome Relationship dataset, which is found to perform better than the pretraining on the ImageNet for this task."
                    },
                    {
                        "title": "PAD-Net: Multi-Tasks Guided Prediction-and-Distillation Network for Simultaneous Depth Estimation and Scene Parsing",
                        "abstract": "Depth estimation and scene parsing are two particularly important tasks in visual scene understanding. In this paper we tackle the problem of simultaneous depth estimation and scene parsing in a joint CNN. The task can be typically treated as a deep multi-task learning problem [42]. Different from previous methods directly optimizing multiple tasks given the input training data, this paper proposes a novel multi-task guided prediction-and-distillation network (PAD-Net), which first predicts a set of intermediate auxiliary tasks ranging from low level to high level, and then the predictions from these intermediate auxiliary tasks are utilized as multi-modal input via our proposed multi-modal distillation modules for the final tasks. During the joint learning, the intermediate tasks not only act as supervision for learning more robust deep representations but also provide rich multi-modal information for improving the final tasks. Extensive experiments are conducted on two challenging datasets (i.e. NYUD-v2 and Cityscapes) for both the depth estimation and scene parsing tasks, demonstrating the effectiveness of the proposed approach."
                    },
                    {
                        "title": "Multi-person Articulated Tracking with Spatial and Temporal Embeddings",
                        "abstract": "We propose a unified framework for multi-person pose estimation and tracking. Our framework consists of two main components,~\\ie~SpatialNet and TemporalNet. The SpatialNet accomplishes body part detection and part-level data association in a single frame, while the TemporalNet groups human instances in consecutive frames into trajectories. Specifically, besides body part detection heatmaps, SpatialNet also predicts the Keypoint Embedding (KE) and Spatial Instance Embedding (SIE) for body part association. We model the grouping procedure into a differentiable Pose-Guided Grouping (PGG) module to make the whole part detection and grouping pipeline fully end-to-end trainable. TemporalNet extends spatial grouping of keypoints to temporal grouping of human instances. Given human proposals from two consecutive frames, TemporalNet exploits both appearance features encoded in Human Embedding (HE) and temporally consistent geometric features embodied in Temporal Instance Embedding (TIE) for robust tracking. Extensive experiments demonstrate the effectiveness of our proposed model. Remarkably, we demonstrate substantial improvements over the state-of-the-art pose tracking method from 65.4\\% to 71.8\\% Multi-Object Tracking Accuracy (MOTA) on the ICCV'17 PoseTrack Dataset."
                    },
                    {
                        "title": "Learnable Histogram: Statistical Context Features for Deep Neural Networks",
                        "abstract": "Statistical features, such as histogram, Bag-of-Words (BoW) and Fisher Vector, were commonly used with hand-crafted features in conventional classification methods, but attract less attention since the popularity of deep learning methods. In this paper, we propose a learnable histogram layer, which learns histogram features within deep neural networks in end-to-end training. Such a layer is able to back-propagate (BP) errors, learn optimal bin centers and bin widths, and be jointly optimized with other layers in deep networks during training. Two vision problems, semantic segmentation and object detection, are explored by integrating the learnable histogram layer into deep networks, which show that the proposed layer could be well generalized to different applications. In-depth investigations are conducted to provide insights on the newly introduced layer."
                    },
                    {
                        "title": "Channel Pruning Guided by Classification Loss and Feature Importance",
                        "abstract": "In this work, we propose a new layer-by-layer channel pruning method called Channel Pruning guided by classification Loss and feature Importance (CPLI). In contrast to the existing layer-by-layer channel pruning approaches that only consider how to reconstruct the features from the next layer, our approach additionally take the classification loss into account in the channel pruning process. We also observe that some reconstructed features will be removed at the next pruning stage. So it is unnecessary to reconstruct these features. To this end, we propose a new strategy to suppress the influence of unimportant features (i.e., the features will be removed at the next pruning stage). Our comprehensive experiments on three benchmark datasets, i.e., CIFAR-10, ImageNet, and UCF-101, demonstrate the effectiveness of our CPLI method."
                    },
                    {
                        "title": "Fast-MoCo: Boost Momentum-based Contrastive Learning with Combinatorial Patches",
                        "abstract": "Contrastive-based self-supervised learning methods achieved great success in recent years. However, self-supervision requires extremely long training epochs (e.g., 800 epochs for MoCo v3) to achieve promising results, which is unacceptable for the general academic community and hinders the development of this topic. This work revisits the momentum-based contrastive learning frameworks and identifies the inefficiency in which two augmented views generate only one positive pair. We propose Fast-MoCo - a novel framework that utilizes combinatorial patches to construct multiple positive pairs from two augmented views, which provides abundant supervision signals that bring significant acceleration with neglectable extra computational cost. Fast-MoCo trained with 100 epochs achieves 73.5% linear evaluation accuracy, similar to MoCo v3 (ResNet-50 backbone) trained with 800 epochs. Extra training (200 epochs) further improves the result to 75.1%, which is on par with state-of-the-art methods. Experiments on several downstream tasks also confirm the effectiveness of Fast-MoCo."
                    },
                    {
                        "title": "Adaptive Hierarchical SpatioTemporal Network for Traffic Forecasting",
                        "abstract": "Accurate traffic forecasting is vital to intelligent transportation systems, which are widely adopted to solve urban traffic issues. Existing traffic forecasting studies focus on modeling spatial-temporal dynamics in traffic data, among which the graph convolution network (GCN) is at the center for exploiting the spatial dependency embedded in the road network graphs. However, these GCN-based methods operate intrinsically on the node level (e.g., road and intersection) only whereas overlooking the spatial hierarchy of the whole city. Nodes such as intersections and road segments can form clusters (e.g., regions), which could also have interactions with each other and share similarities at a higher level. In this work, we propose an Adaptive Hierarchical SpatioTemporal Network (AHSTN) to promote traffic forecasting by exploiting the spatial hierarchy and modeling multi-scale spatial correlations. Apart from the node-level spatiotemporal blocks, AHSTN introduces the adaptive spatiotemporal downsampling module to infer the spatial hierarchy for spatiotemporal modeling at the cluster level. Then, an adaptive spatiotemporal upsampling module is proposed to upsample the cluster-level representations to the node-level and obtain the multi-scale representations for generating predictions. Experiments on two real-world datasets show that AHSTN achieves better performance over several strong baselines."
                    },
                    {
                        "title": "STEERER: Resolving Scale Variations for Counting and Localization via Selective Inheritance Learning",
                        "abstract": "Scale variation is a deep-rooted problem in object counting, which has not been effectively addressed by existing scale-aware algorithms. An important factor is that they typically involve cooperative learning across multi-resolutions, which could be suboptimal for learning the most discriminative features from each scale. In this paper, we propose a novel method termed STEERER (\\textbf{S}elec\\textbf{T}iv\\textbf{E} inh\\textbf{ER}itance l\\textbf{E}a\\textbf{R}ning) that addresses the issue of scale variations in object counting. STEERER selects the most suitable scale for patch objects to boost feature extraction and only inherits discriminative features from lower to higher resolution progressively. The main insights of STEERER are a dedicated Feature Selection and Inheritance Adaptor (FSIA), which selectively forwards scale-customized features at each scale, and a Masked Selection and Inheritance Loss (MSIL) that helps to achieve high-quality density maps across all scales. Our experimental results on nine datasets with counting and localization tasks demonstrate the unprecedented scale generalization ability of STEERER. Code is available at \\url{https://github.com/taohan10200/STEERER}."
                    },
                    {
                        "title": "Person Re-identification by Saliency Learning",
                        "abstract": "Human eyes can recognize person identities based on small salient regions, i.e. human saliency is distinctive and reliable in pedestrian matching across disjoint camera views. However, such valuable information is often hidden when computing similarities of pedestrian images with existing approaches. Inspired by our user study result of human perception on human saliency, we propose a novel perspective for person re-identification based on learning human saliency and matching saliency distribution. The proposed saliency learning and matching framework consists of four steps: (1) To handle misalignment caused by drastic viewpoint change and pose variations, we apply adjacency constrained patch matching to build dense correspondence between image pairs. (2) We propose two alternative methods, i.e. K-Nearest Neighbors and One-class SVM, to estimate a saliency score for each image patch, through which distinctive features stand out without using identity labels in the training procedure. (3) saliency matching is proposed based on patch matching. Matching patches with inconsistent saliency brings penalty, and images of the same identity are recognized by minimizing the saliency matching cost. (4) Furthermore, saliency matching is tightly integrated with patch matching in a unified structural RankSVM learning framework. The effectiveness of our approach is validated on the VIPeR dataset and the CUHK01 dataset. Our approach outperforms the state-of-the-art person re-identification methods on both datasets."
                    },
                    {
                        "title": "Quality Aware Network for Set to Set Recognition",
                        "abstract": "This paper targets on the problem of set to set recognition, which learns the metric between two image sets. Images in each set belong to the same identity. Since images in a set can be complementary, they hopefully lead to higher accuracy in practical applications. However, the quality of each sample cannot be guaranteed, and samples with poor quality will hurt the metric. In this paper, the quality aware network (QAN) is proposed to confront this problem, where the quality of each sample can be automatically learned although such information is not explicitly provided in the training stage. The network has two branches, where the first branch extracts appearance feature embedding for each sample and the other branch predicts quality score for each sample. Features and quality scores of all samples in a set are then aggregated to generate the final feature embedding. We show that the two branches can be trained in an end-to-end manner given only the set-level identity annotation. Analysis on gradient spread of this mechanism indicates that the quality learned by the network is beneficial to set-to-set recognition and simplifies the distribution that the network needs to fit. Experiments on both face verification and person re-identification show advantages of the proposed QAN. The source code and network structure can be downloaded at https://github.com/sciencefans/Quality-Aware-Network."
                    },
                    {
                        "title": "Box-driven Class-wise Region Masking and Filling Rate Guided Loss for Weakly Supervised Semantic Segmentation",
                        "abstract": "Semantic segmentation has achieved huge progress via adopting deep Fully Convolutional Networks (FCN). However, the performance of FCN based models severely rely on the amounts of pixel-level annotations which are expensive and time-consuming. To address this problem, it is a good choice to learn to segment with weak supervision from bounding boxes. How to make full use of the class-level and region-level supervisions from bounding boxes is the critical challenge for the weakly supervised learning task. In this paper, we first introduce a box-driven class-wise masking model (BCM) to remove irrelevant regions of each class. Moreover, based on the pixel-level segment proposal generated from the bounding box supervision, we could calculate the mean filling rates of each class to serve as an important prior cue, then we propose a filling rate guided adaptive loss (FR-Loss) to help the model ignore the wrongly labeled pixels in proposals. Unlike previous methods directly training models with the fixed individual segment proposals, our method can adjust the model learning with global statistical information. Thus it can help reduce the negative impacts from wrongly labeled proposals. We evaluate the proposed method on the challenging PASCAL VOC 2012 benchmark and compare with other methods. Extensive experimental results show that the proposed method is effective and achieves the state-of-the-art results."
                    },
                    {
                        "title": "Improving Description-based Person Re-identification by Multi-granularity Image-text Alignments",
                        "abstract": "Description-based person re-identification (Re-id) is an important task in video surveillance that requires discriminative cross-modal representations to distinguish different people. It is difficult to directly measure the similarity between images and descriptions due to the modality heterogeneity (the cross-modal problem). And all samples belonging to a single category (the fine-grained problem) makes this task even harder than the conventional image-description matching task. In this paper, we propose a Multi-granularity Image-text Alignments (MIA) model to alleviate the cross-modal fine-grained problem for better similarity evaluation in description-based person Re-id. Specifically, three different granularities, i.e., global-global, global-local and local-local alignments are carried out hierarchically. Firstly, the global-global alignment in the Global Contrast (GC) module is for matching the global contexts of images and descriptions. Secondly, the global-local alignment employs the potential relations between local components and global contexts to highlight the distinguishable components while eliminating the uninvolved ones adaptively in the Relation-guided Global-local Alignment (RGA) module. Thirdly, as for the local-local alignment, we match visual human parts with noun phrases in the Bi-directional Fine-grained Matching (BFM) module. The whole network combining multiple granularities can be end-to-end trained without complex pre-processing. To address the difficulties in training the combination of multiple granularities, an effective step training strategy is proposed to train these granularities step-by-step. Extensive experiments and analysis have shown that our method obtains the state-of-the-art performance on the CUHK-PEDES dataset and outperforms the previous methods by a significant margin."
                    },
                    {
                        "title": "A Shape Transformation-based Dataset Augmentation Framework for Pedestrian Detection",
                        "abstract": "Deep learning-based computer vision is usually data-hungry. Many researchers attempt to augment datasets with synthesized data to improve model robustness. However, the augmentation of popular pedestrian datasets, such as Caltech and Citypersons, can be extremely challenging because real pedestrians are commonly in low quality. Due to the factors like occlusions, blurs, and low-resolution, it is significantly difficult for existing augmentation approaches, which generally synthesize data using 3D engines or generative adversarial networks (GANs), to generate realistic-looking pedestrians. Alternatively, to access much more natural-looking pedestrians, we propose to augment pedestrian detection datasets by transforming real pedestrians from the same dataset into different shapes. Accordingly, we propose the Shape Transformation-based Dataset Augmentation (STDA) framework. The proposed framework is composed of two subsequent modules, i.e. the shape-guided deformation and the environment adaptation. In the first module, we introduce a shape-guided warping field to help deform the shape of a real pedestrian into a different shape. Then, in the second stage, we propose an environment-aware blending map to better adapt the deformed pedestrians into surrounding environments, obtaining more realistic-looking pedestrians and more beneficial augmentation results for pedestrian detection. Extensive empirical studies on different pedestrian detection benchmarks show that the proposed STDA framework consistently produces much better augmentation results than other pedestrian synthesis approaches using low-quality pedestrians. By augmenting the original datasets, our proposed framework also improves the baseline pedestrian detector by up to 38% on the evaluated benchmarks, achieving state-of-the-art performance."
                    }
                ]
            },
            "8e6a3de0-afa6-4297-a497-f110709be8d7": {
                "pk": "8e6a3de0-afa6-4297-a497-f110709be8d7",
                "name": "Jingdong Wang",
                "collaborators": [
                    "Ting Zhang",
                    "Yuhui Yuan",
                    "Xiaokang Chen",
                    "Wenjun Zeng",
                    "Xilin Chen",
                    "Gang Zeng",
                    "Yifan Liu",
                    "Chunhua Shen",
                    "Changqian Yu",
                    "Chao Du"
                ],
                "domain": [
                    "Deep Learning",
                    "Computer Vision",
                    "Nearest Neighbor Search",
                    "Semantic Segmentation"
                ],
                "publications": [
                    {
                        "title": "Inner Product Similarity Search using Compositional Codes",
                        "abstract": "This paper addresses the nearest neighbor search problem under inner product similarity and introduces a compact code-based approach. The idea is to approximate a vector using the composition of several elements selected from a source dictionary and to represent this vector by a short code composed of the indices of the selected elements. The inner product between a query vector and a database vector is efficiently estimated from the query vector and the short code of the database vector. We show the superior performance of the proposed group $M$-selection algorithm that selects $M$ elements from $M$ source dictionaries for vector approximation in terms of search accuracy and efficiency for compact codes of the same length via theoretical and empirical analysis. Experimental results on large-scale datasets ($1M$ and $1B$ SIFT features, $1M$ linear models and Netflix) demonstrate the superiority of the proposed approach."
                    },
                    {
                        "title": "Composite Quantization",
                        "abstract": "This paper studies the compact coding approach to approximate nearest neighbor search. We introduce a composite quantization framework. It uses the composition of several ($M$) elements, each of which is selected from a different dictionary, to accurately approximate a $D$-dimensional vector, thus yielding accurate search, and represents the data vector by a short code composed of the indices of the selected elements in the corresponding dictionaries. Our key contribution lies in introducing a near-orthogonality constraint, which makes the search efficiency is guaranteed as the cost of the distance computation is reduced to $O(M)$ from $O(D)$ through a distance table lookup scheme. The resulting approach is called near-orthogonal composite quantization. We theoretically justify the equivalence between near-orthogonal composite quantization and minimizing an upper bound of a function formed by jointly considering the quantization error and the search cost according to a generalized triangle inequality. We empirically show the efficacy of the proposed approach over several benchmark datasets. In addition, we demonstrate the superior performances in other three applications: combination with inverted multi-index, quantizing the query for mobile search, and inner-product similarity search."
                    },
                    {
                        "title": "Collaborative Quantization for Cross-Modal Similarity Search",
                        "abstract": "Cross-modal similarity search is a problem about designing a search system supporting querying across content modalities, e.g., using an image to search for texts or using a text to search for images. This paper presents a compact coding solution for efficient search, with a focus on the quantization approach which has already shown the superior performance over the hashing solutions in the single-modal similarity search. We propose a cross-modal quantization approach, which is among the early attempts to introduce quantization into cross-modal search. The major contribution lies in jointly learning the quantizers for both modalities through aligning the quantized representations for each pair of image and text belonging to a document. In addition, our approach simultaneously learns the common space for both modalities in which quantization is conducted to enable efficient and effective search using the Euclidean distance computed in the common space with fast distance table lookup. Experimental results compared with several competitive algorithms over three benchmark datasets demonstrate that the proposed approach achieves the state-of-the-art performance."
                    },
                    {
                        "title": "Admix: Enhancing the Transferability of Adversarial Attacks",
                        "abstract": "Deep neural networks are known to be extremely vulnerable to adversarial examples under white-box setting. Moreover, the malicious adversaries crafted on the surrogate (source) model often exhibit black-box transferability on other models with the same learning task but having different architectures. Recently, various methods are proposed to boost the adversarial transferability, among which the input transformation is one of the most effective approaches. We investigate in this direction and observe that existing transformations are all applied on a single image, which might limit the adversarial transferability. To this end, we propose a new input transformation based attack method called Admix that considers the input image and a set of images randomly sampled from other categories. Instead of directly calculating the gradient on the original input, Admix calculates the gradient on the input image admixed with a small portion of each add-in image while using the original label of the input to craft more transferable adversaries. Empirical evaluations on standard ImageNet dataset demonstrate that Admix could achieve significantly better transferability than existing input transformation methods under both single model setting and ensemble-model setting. By incorporating with existing input transformations, our method could further improve the transferability and outperforms the state-of-the-art combination of input transformations by a clear margin when attacking nine advanced defense models under ensemble-model setting. Code is available at https://github.com/JHL-HUST/Admix."
                    },
                    {
                        "title": "Deeply-Fused Nets",
                        "abstract": "In this paper, we present a novel deep learning approach, deeply-fused nets. The central idea of our approach is deep fusion, i.e., combine the intermediate representations of base networks, where the fused output serves as the input of the remaining part of each base network, and perform such combinations deeply over several intermediate representations. The resulting deeply fused net enjoys several benefits. First, it is able to learn multi-scale representations as it enjoys the benefits of more base networks, which could form the same fused network, other than the initial group of base networks. Second, in our suggested fused net formed by one deep and one shallow base networks, the flows of the information from the earlier intermediate layer of the deep base network to the output and from the input to the later intermediate layer of the deep base network are both improved. Last, the deep and shallow base networks are jointly learnt and can benefit from each other. More interestingly, the essential depth of a fused net composed from a deep base network and a shallow base network is reduced because the fused net could be composed from a less deep base network, and thus training the fused net is less difficult than training the initial deep base network. Empirical results demonstrate that our approach achieves superior performance over two closely-related methods, ResNet and Highway, and competitive performance compared to the state-of-the-arts."
                    },
                    {
                        "title": "Deeply-Learned Part-Aligned Representations for Person Re-Identification",
                        "abstract": "In this paper, we address the problem of person re-identification, which refers to associating the persons captured from different cameras. We propose a simple yet effective human part-aligned representation for handling the body part misalignment problem. Our approach decomposes the human body into regions (parts) which are discriminative for person matching, accordingly computes the representations over the regions, and aggregates the similarities computed between the corresponding regions of a pair of probe and gallery images as the overall matching score. Our formulation, inspired by attention models, is a deep neural network modeling the three steps together, which is learnt through minimizing the triplet loss function without requiring body part labeling information. Unlike most existing deep learning algorithms that learn a global or spatial partition-based local representation, our approach performs human body partition, and thus is more robust to pose changes and various human spatial distributions in the person bounding box. Our approach shows state-of-the-art results over standard datasets, Market-$1501$, CUHK$03$, CUHK$01$ and VIPeR."
                    },
                    {
                        "title": "SegFix: Model-Agnostic Boundary Refinement for Segmentation",
                        "abstract": "We present a model-agnostic post-processing scheme to improve the boundary quality for the segmentation result that is generated by any existing segmentation model. Motivated by the empirical observation that the label predictions of interior pixels are more reliable, we propose to replace the originally unreliable predictions of boundary pixels by the predictions of interior pixels. Our approach processes only the input image through two steps: (i) localize the boundary pixels and (ii) identify the corresponding interior pixel for each boundary pixel. We build the correspondence by learning a direction away from the boundary pixel to an interior pixel. Our method requires no prior information of the segmentation models and achieves nearly real-time speed. We empirically verify that our SegFix consistently reduces the boundary errors for segmentation results generated from various state-of-the-art models on Cityscapes, ADE20K and GTA5. Code is available at: https://github.com/openseg-group/openseg.pytorch."
                    },
                    {
                        "title": "Distillation Guided Residual Learning for Binary Convolutional Neural Networks",
                        "abstract": "It is challenging to bridge the performance gap between Binary CNN (BCNN) and Floating point CNN (FCNN). We observe that, this performance gap leads to substantial residuals between intermediate feature maps of BCNN and FCNN. To minimize the performance gap, we enforce BCNN to produce similar intermediate feature maps with the ones of FCNN. This training strategy, i.e., optimizing each binary convolutional block with block-wise distillation loss derived from FCNN, leads to a more effective optimization to BCNN. It also motivates us to update the binary convolutional block architecture to facilitate the optimization of block-wise distillation loss. Specifically, a lightweight shortcut branch is inserted into each binary convolutional block to complement residuals at each block. Benefited from its Squeeze-and-Interaction (SI) structure, this shortcut branch introduces a fraction of parameters, e.g., 10\\% overheads, but effectively complements the residuals. Extensive experiments on ImageNet demonstrate the superior performance of our method in both classification efficiency and accuracy, e.g., BCNN trained with our methods achieves the accuracy of 60.45\\% on ImageNet."
                    },
                    {
                        "title": "Weakly-Supervised Action Localization by Generative Attention Modeling",
                        "abstract": "Weakly-supervised temporal action localization is a problem of learning an action localization model with only video-level action labeling available. The general framework largely relies on the classification activation, which employs an attention model to identify the action-related frames and then categorizes them into different classes. Such method results in the action-context confusion issue: context frames near action clips tend to be recognized as action frames themselves, since they are closely related to the specific classes. To solve the problem, in this paper we propose to model the class-agnostic frame-wise probability conditioned on the frame attention using conditional Variational Auto-Encoder (VAE). With the observation that the context exhibits notable difference from the action at representation level, a probabilistic model, i.e., conditional VAE, is learned to model the likelihood of each frame given the attention. By maximizing the conditional probability with respect to the attention, the action and non-action frames are well separated. Experiments on THUMOS14 and ActivityNet1.2 demonstrate advantage of our method and effectiveness in handling action-context confusion problem. Code is now available on GitHub."
                    },
                    {
                        "title": "Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision",
                        "abstract": "In this paper, we study the semi-supervised semantic segmentation problem via exploring both labeled data and extra unlabeled data. We propose a novel consistency regularization approach, called cross pseudo supervision (CPS). Our approach imposes the consistency on two segmentation networks perturbed with different initialization for the same input image. The pseudo one-hot label map, output from one perturbed segmentation network, is used to supervise the other segmentation network with the standard cross-entropy loss, and vice versa. The CPS consistency has two roles: encourage high similarity between the predictions of two perturbed networks for the same input image, and expand training data by using the unlabeled data with pseudo labels. Experiment results show that our approach achieves the state-of-the-art semi-supervised segmentation performance on Cityscapes and PASCAL VOC 2012. Code is available at https://git.io/CPS."
                    },
                    {
                        "title": "Rating the Crisis of Online Public Opinion Using a Multi-Level Index System",
                        "abstract": "Online public opinion usually spreads rapidly and widely, thus a small incident probably evolves into a large social crisis in a very short time, and results in a heavy loss in credit or economic aspects. We propose a method to rate the crisis of online public opinion based on a multi-level index system to evaluate the impact of events objectively. Firstly, the dissemination mechanism of online public opinion is explained from the perspective of information ecology. According to the mechanism, some evaluation indexes are selected through correlation analysis and principal component analysis. Then, a classification model of text emotion is created via the training by deep learning to achieve the accurate quantification of the emotional indexes in the index system. Finally, based on the multi-level evaluation index system and grey correlation analysis, we propose a method to rate the crisis of online public opinion. The experiment with the real-time incident show that this method can objectively evaluate the emotional tendency of Internet users and rate the crisis in different dissemination stages of online public opinion. It is helpful to realizing the crisis warning of online public opinion and timely blocking the further spread of the crisis."
                    },
                    {
                        "title": "Assessing Model Generalization in Vicinity",
                        "abstract": "This paper evaluates the generalization ability of classification models on out-of-distribution test sets without depending on ground truth labels. Common approaches often calculate an unsupervised metric related to a specific model property, like confidence or invariance, which correlates with out-of-distribution accuracy. However, these metrics are typically computed for each test sample individually, leading to potential issues caused by spurious model responses, such as overly high or low confidence. To tackle this challenge, we propose incorporating responses from neighboring test samples into the correctness assessment of each individual sample. In essence, if a model consistently demonstrates high correctness scores for nearby samples, it increases the likelihood of correctly predicting the target sample, and vice versa. The resulting scores are then averaged across all test samples to provide a holistic indication of model accuracy. Developed under the vicinal risk formulation, this approach, named vicinal risk proxy (VRP), computes accuracy without relying on labels. We show that applying the VRP method to existing generalization indicators, such as average confidence and effective invariance, consistently improves over these baselines both methodologically and experimentally. This yields a stronger correlation with model accuracy, especially on challenging out-of-distribution test sets."
                    },
                    {
                        "title": "Efficient Semantic Video Segmentation with Per-frame Inference",
                        "abstract": "For semantic segmentation, most existing real-time deep models trained with each frame independently may produce inconsistent results for a video sequence. Advanced methods take into considerations the correlations in the video sequence, e.g., by propagating the results to the neighboring frames using optical flow, or extracting the frame representations with other frames, which may lead to inaccurate results or unbalanced latency. In this work, we process efficient semantic video segmentation in a per-frame fashion during the inference process. Different from previous per-frame models, we explicitly consider the temporal consistency among frames as extra constraints during the training process and embed the temporal consistency into the segmentation network. Therefore, in the inference process, we can process each frame independently with no latency, and improve the temporal consistency with no extra computational cost and post-processing. We employ compact models for real-time execution. To narrow the performance gap between compact models and large models, new knowledge distillation methods are designed. Our results outperform previous keyframe based methods with a better trade-off between the accuracy and the inference speed on popular benchmarks, including the Cityscapes and Camvid. The temporal consistency is also improved compared with corresponding baselines which are trained with each frame independently. Code is available at: https://tinyurl.com/segment-video"
                    },
                    {
                        "title": "Efficient Video Segmentation Models with Per-frame Inference",
                        "abstract": "Most existing real-time deep models trained with each frame independently may produce inconsistent results across the temporal axis when tested on a video sequence. A few methods take the correlations in the video sequence into account,e.g., by propagating the results to the neighboring frames using optical flow or extracting frame representations using multi-frame information, which may lead to inaccurate results or unbalanced latency. In this work, we focus on improving the temporal consistency without introducing computation overhead in inference. To this end, we perform inference at each frame. Temporal consistency is achieved by learning from video frames with extra constraints during the training phase. introduced for inference. We propose several techniques to learn from the video sequence, including a temporal consistency loss and online/offline knowledge distillation methods. On the task of semantic video segmentation, weighing among accuracy, temporal smoothness, and efficiency, our proposed method outperforms keyframe-based methods and a few baseline methods that are trained with each frame independently, on datasets including Cityscapes, Camvid, and 300VW-Mask. We further apply our training method to video instance segmentation on YouTubeVISand develop an application of portrait matting in video sequences, by segmenting temporally consistent instance-level trimaps across frames. Experiments show superior qualitative and quantitative results. Code is available at: https://git.io/vidseg."
                    },
                    {
                        "title": "Conditional DETR V2: Efficient Detection Transformer with Box Queries",
                        "abstract": "In this paper, we are interested in Detection Transformer (DETR), an end-to-end object detection approach based on a transformer encoder-decoder architecture without hand-crafted postprocessing, such as NMS. Inspired by Conditional DETR, an improved DETR with fast training convergence, that presented box queries (originally called spatial queries) for internal decoder layers, we reformulate the object query into the format of the box query that is a composition of the embeddings of the reference point and the transformation of the box with respect to the reference point. This reformulation indicates the connection between the object query in DETR and the anchor box that is widely studied in Faster R-CNN. Furthermore, we learn the box queries from the image content, further improving the detection quality of Conditional DETR still with fast training convergence. In addition, we adopt the idea of axial self-attention to save the memory cost and accelerate the encoder. The resulting detector, called Conditional DETR V2, achieves better results than Conditional DETR, saves the memory cost and runs more efficiently. For example, for the DC$5$-ResNet-$50$ backbone, our approach achieves $44.8$ AP with $16.4$ FPS on the COCO $val$ set and compared to Conditional DETR, it runs $1.6\\times$ faster, saves $74$\\% of the overall memory cost, and improves $1.0$ AP score."
                    },
                    {
                        "title": "Segmentation Transformer: Object-Contextual Representations for Semantic Segmentation",
                        "abstract": "In this paper, we address the semantic segmentation problem with a focus on the context aggregation strategy. Our motivation is that the label of a pixel is the category of the object that the pixel belongs to. We present a simple yet effective approach, object-contextual representations, characterizing a pixel by exploiting the representation of the corresponding object class. First, we learn object regions under the supervision of ground-truth segmentation. Second, we compute the object region representation by aggregating the representations of the pixels lying in the object region. Last, % the representation similarity we compute the relation between each pixel and each object region and augment the representation of each pixel with the object-contextual representation which is a weighted aggregation of all the object region representations according to their relations with the pixel. We empirically demonstrate that the proposed approach achieves competitive performance on various challenging semantic segmentation benchmarks: Cityscapes, ADE20K, LIP, PASCAL-Context, and COCO-Stuff. Cityscapes, ADE20K, LIP, PASCAL-Context, and COCO-Stuff. Our submission \"HRNet + OCR + SegFix\" achieves 1-st place on the Cityscapes leaderboard by the time of submission. Code is available at: https://git.io/openseg and https://git.io/HRNet.OCR. We rephrase the object-contextual representation scheme using the Transformer encoder-decoder framework. The details are presented in~Section3.3."
                    },
                    {
                        "title": "Cross View Fusion for 3D Human Pose Estimation",
                        "abstract": "We present an approach to recover absolute 3D human poses from multi-view images by incorporating multi-view geometric priors in our model. It consists of two separate steps: (1) estimating the 2D poses in multi-view images and (2) recovering the 3D poses from the multi-view 2D poses. First, we introduce a cross-view fusion scheme into CNN to jointly estimate 2D poses for multiple views. Consequently, the 2D pose estimation for each view already benefits from other views. Second, we present a recursive Pictorial Structure Model to recover the 3D pose from the multi-view 2D poses. It gradually improves the accuracy of 3D pose with affordable computational cost. We test our method on two public datasets H36M and Total Capture. The Mean Per Joint Position Errors on the two datasets are 26mm and 29mm, which outperforms the state-of-the-arts remarkably (26mm vs 52mm, 29mm vs 35mm). Our code is released at \\url{https://github.com/microsoft/multiview-human-pose-estimation-pytorch}."
                    },
                    {
                        "title": "Deep Regression for Face Alignment",
                        "abstract": "In this paper, we present a deep regression approach for face alignment. The deep architecture consists of a global layer and multi-stage local layers. We apply the back-propagation algorithm with the dropout strategy to jointly optimize the regression parameters. We show that the resulting deep regressor gradually and evenly approaches the true facial landmarks stage by stage, avoiding the tendency to yield over-strong early stage regressors while over-weak later stage regressors. Experimental results show that our approach achieves the state-of-the-art"
                    }
                ]
            }
        }
    },
    "2402.04882": {
        "paper_data": {
            "title": "LMUFormer: Low Complexity Yet Powerful Spiking Model With Legendre Memory Units",
            "url": "http://arxiv.org/abs/2402.04882v1",
            "arxiv_id": "2402.04882",
            "authors": [
                "Zeyu Liu",
                "Gourav Datta",
                "Anni Li",
                "Peter Anthony Beerel"
            ],
            "abstract": "Transformer models have demonstrated high accuracy in numerous applications but have high complexity and lack sequential processing capability making them ill-suited for many streaming applications at the edge where devices are heavily resource-constrained. Thus motivated, many researchers have proposed reformulating the transformer models as RNN modules which modify the self-attention computation with explicit states. However, these approaches often incur significant performance degradation. The ultimate goal is to develop a model that has the following properties: parallel training, streaming and low-cost inference, and SOTA performance. In this paper, we propose a new direction to achieve this goal. We show how architectural modifications to a recurrent model can help push its performance toward Transformer models while retaining its sequential processing capability. Specifically, inspired by the recent success of Legendre Memory Units (LMU) in sequence learning tasks, we propose LMUFormer, which augments the LMU with convolutional patch embedding and convolutional channel mixer. Moreover, we present a spiking version of this architecture, which introduces the benefit of states within the patch embedding and channel mixer modules while simultaneously reducing the computing complexity. We evaluated our architectures on multiple sequence datasets. In comparison to SOTA transformer-based models within the ANN domain on the SCv2 dataset, our LMUFormer demonstrates comparable performance while necessitating a remarkable 53 times reduction in parameters and a substantial 65 times decrement in FLOPs. Additionally, owing to our model's proficiency in real-time data processing, we can achieve a 32.03% reduction in sequence length, all while incurring an inconsequential decline in performance. Our code is publicly available at https://github.com/zeyuliu1037/LMUFormer.git.",
            "introduction": "   1 Introduction  In recent years, Transformers\u00a0(Vaswani et\u00a0al., 2017) have become the most prevalent deep learning models for many applications. The global attention mechanism enables Transformers to capture long-distance dependencies in input data. Moreover, their state-less structure makes their training easily parallelizable, helping them perform well with large-scale datasets, leading to SOTA accuracy in a wide range of applications, including Natural Language Processing (NLP)\u00a0(Brown et\u00a0al., 2020; Touvron et\u00a0al., 2023), Computer Vision (CV)\u00a0(Dosovitskiy et\u00a0al., 2020; Liu et\u00a0al., 2021), and spoken term classification (STC)\u00a0(Gulati et\u00a0al., 2020; Gong et\u00a0al., 2021). However, their self-attention mechanism imposes compute and memory complexity that is quadratic in the sequence length N\ud835\udc41Nitalic_N. Moreover, the nature of the self-attention computation prevents the underlying hardware from performing much of the computation until after the entire input sequence is available.   In contrast, Recurrent Neural Network (RNN) models are designed to process data sequentially and their complexity is only \ud835\udcaa\u2062(N)\ud835\udcaa\ud835\udc41\\mathcal{O}(N)caligraphic_O ( italic_N ) making them an attractive low-complexity alternative to transformers. However, RNN models traditionally have higher training time as the training procedure must accommodate the long sequence of dependencies within the model, making parallelization more difficult. Moreover, unlike transformers, RNN models like LSTMs\u00a0(Hochreiter & Schmidhuber, 1997) and GRUs\u00a0(Cho et\u00a0al., 2014) can only leverage past and current information and traditionally suffer from forgetting due to having a limited memory horizon. Recently, however, Legendre Memory Units(Voelker et\u00a0al., 2019) have been proposed that have the ability to remember information over a theoretically infinite time horizon. Benefiting from this property, LMU-based models outperform many RNN models while still underperforming compared to Transformer alternatives. We hypothesize that the remaining performance gap between LMUs and Transformers is not only because Transformers benefit from future information, but also because they possess more complex network structures, such as the self-attention computation, and thus have higher representational capacity.    In an attempt to explore this hypothesis, we propose a novel sequential network architecture, dubbed LMUFormer, that augments the LMU module with convolutional patch embedding and convolutional channel mixers. Importantly, our convolutional patch embedding only interacts with neighboring input samples and our convolutional channel mixers and the final classifier only operate on the current state of the LMU. Therefore, our LMUFormer model can process data sequentially. Moreover, we present a spiking version of this architecture which extends the benefits of state that is explicit in the LMU implicitly to the patch embedding and channel mixing structures to improve accuracy while simultaneously further enabling complexity reduction. In summary, this paper makes the following contributions:     \u2022  We propose a novel architecture LMUFormer which outperforms most existing RNN models with similar complexity on a wide range of sequence learning tasks. On evaluating the Speech Commands dataset, our LMUFormer model, when benchmarked against transformer-based models with competitive performance levels, manifests a significant reduction \u2014 accounting for  53\u00d753\\times53 \u00d7 fewer parameters and  65\u00d765\\times65 \u00d7 diminished FLOPs.    \u2022  We further devised a spiking variant of the LMUFormer that not only achieves state-of-the-art (SOTA) performance within the realm of SNN models on the Speech Commands dataset but also presents comparable results on the Long Range Arena benchmark.    \u2022  Attributable to the real-time data processing capabilities inherent to our models, we evaluate the performance of",
            "references": [
                {
                    "title": "Spiking Convolutional Neural Networks for Text Classification",
                    "abstract": "Spiking neural networks (SNNs) offer a promising pathway to implement deep neural networks (DNNs) in a more energy-efficient manner since their neurons are sparsely activated and inferences are event-driven. However, there have been very few works that have demonstrated the efficacy of SNNs in language tasks partially because it is non-trivial to represent words in the forms of spikes and to deal with variable-length texts by SNNs. This work presents a\"conversion + fine-tuning\"two-step method for training SNNs for text classification and proposes a simple but effective way to encode pre-trained word embeddings as spike trains. We show empirically that after fine-tuning with surrogate gradients, the converted SNNs achieve comparable results to their DNN counterparts with much less energy consumption across multiple datasets for both English and Chinese. We also show that such SNNs are more robust to adversarial attacks than DNNs."
                },
                {
                    "title": "Bridging the Gap Between Spiking Neural Networks & LSTMs for Latency & Energy Efficiency",
                    "abstract": "Spiking Neural Networks (SNNs) have emerged as an attractive spatio-temporal computing paradigm for complex vision tasks. However, most existing works yield models that require many time steps and do not leverage the inherent temporal dynamics of spiking neural networks, even for sequential tasks. Motivated by this observation, we propose an optimized spiking long short-term memory networks (LSTM) training framework that involves a novel ANN-to-SNN conversion framework, followed by SNN fine-tuning via backpropagation through time (BPTT). In particular, we propose novel activation functions in the source LSTM architecture and convert a judiciously selected subset of them to leaky-integrate-and-fire (LIF) activations with optimal bias shifts. Moreover, we propose a pipelined parallel processing scheme that hides the SNN time steps, significantly improving system latency, especially for long sequences. The resulting SNNs have high activation sparsity and require only accumulate operations (AC), in contrast to expensive multiply-and-accumulates (MAC) needed for ANNs, except for the input layer when using direct encoding, yielding significant improvements in energy efficiency. We evaluate our framework on sequential learning tasks including temporal MNIST, Google Speech Commands (GSC), and UCI Smartphone datasets on different LSTM architectures. We obtain test accuracy of 94.75 % with only 2 time steps on the GSC dataset with $\\sim 4.1\\times$ lower energy than an iso-architecture standard LSTM."
                },
                {
                    "title": "Spike-driven Transformer",
                    "abstract": "Spiking Neural Networks (SNNs) provide an energy-efficient deep learning option due to their unique spike-based event-driven (i.e., spike-driven) paradigm. In this paper, we incorporate the spike-driven paradigm into Transformer by the proposed Spike-driven Transformer with four unique properties: 1) Event-driven, no calculation is triggered when the input of Transformer is zero; 2) Binary spike communication, all matrix multiplications associated with the spike matrix can be transformed into sparse additions; 3) Self-attention with linear complexity at both token and channel dimensions; 4) The operations between spike-form Query, Key, and Value are mask and addition. Together, there are only sparse addition operations in the Spike-driven Transformer. To this end, we design a novel Spike-Driven Self-Attention (SDSA), which exploits only mask and addition operations without any multiplication, and thus having up to $87.2\\times$ lower computation energy than vanilla self-attention. Especially in SDSA, the matrix multiplication between Query, Key, and Value is designed as the mask operation. In addition, we rearrange all residual connections in the vanilla Transformer before the activation functions to ensure that all neurons transmit binary spike signals. It is shown that the Spike-driven Transformer can achieve 77.1\\% top-1 accuracy on ImageNet-1K, which is the state-of-the-art result in the SNN field. The source code is available at https://github.com/BICLab/Spike-Driven-Transformer."
                },
                {
                    "title": "RWKV: Reinventing RNNs for the Transformer Era",
                    "abstract": "Transformers have revolutionized almost all natural language processing (NLP) tasks but suffer from memory and computational complexity that scales quadratically with sequence length. In contrast, recurrent neural networks (RNNs) exhibit linear scaling in memory and computational requirements but struggle to match the same performance as Transformers due to limitations in parallelization and scalability. We propose a novel model architecture, Receptance Weighted Key Value (RWKV), that combines the efficient parallelizable training of transformers with the efficient inference of RNNs. Our approach leverages a linear attention mechanism and allows us to formulate the model as either a Transformer or an RNN, thus parallelizing computations during training and maintains constant computational and memory complexity during inference. We scale our models as large as 14 billion parameters, by far the largest dense RNN ever trained, and find RWKV performs on par with similarly sized Transformers, suggesting future work can leverage this architecture to create more efficient models. This work presents a significant step towards reconciling trade-offs between computational efficiency and model performance in sequence processing tasks."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Spikformer: When Spiking Neural Network Meets Transformer",
                    "abstract": "We consider two biologically plausible structures, the Spiking Neural Network (SNN) and the self-attention mechanism. The former offers an energy-efficient and event-driven paradigm for deep learning, while the latter has the ability to capture feature dependencies, enabling Transformer to achieve good performance. It is intuitively promising to explore the marriage between them. In this paper, we consider leveraging both self-attention capability and biological properties of SNNs, and propose a novel Spiking Self Attention (SSA) as well as a powerful framework, named Spiking Transformer (Spikformer). The SSA mechanism in Spikformer models the sparse visual feature by using spike-form Query, Key, and Value without softmax. Since its computation is sparse and avoids multiplication, SSA is efficient and has low computational energy consumption. It is shown that Spikformer with SSA can outperform the state-of-the-art SNNs-like frameworks in image classification on both neuromorphic and static datasets. Spikformer (66.3M parameters) with comparable size to SEW-ResNet-152 (60.2M,69.26%) can achieve 74.81% top1 accuracy on ImageNet using 4 time steps, which is the state-of-the-art in directly trained SNNs models."
                },
                {
                    "title": "Spiking Neural Networks for Frame-based and Event-based Single Object Localization",
                    "abstract": "Spiking neural networks have shown much promise as an energy-efficient alternative to artificial neural networks. However, understanding the impacts of sensor noises and input encodings on the network activity and performance remains difficult with common neuromorphic vision baselines like classification. Therefore, we propose a spiking neural network approach for single object localization trained using surrogate gradient descent, for frame- and event-based sensors. We compare our method with similar artificial neural networks and show that our model has competitive/better performance in accuracy, robustness against various corruptions, and has lower energy consumption. Moreover, we study the impact of neural coding schemes for static images in accuracy, robustness, and energy efficiency. Our observations differ importantly from previous studies on bio-plausible learning rules, which helps in the design of surrogate gradient trained architectures, and offers insight to design priorities in future neuromorphic technologies in terms of noise characteristics and data encoding methods."
                },
                {
                    "title": "Training High-Performance Low-Latency Spiking Neural Networks by Differentiation on Spike Representation",
                    "abstract": "Spiking Neural Network (SNN) is a promising energy-efficient AI model when implemented on neuromorphic hardware. However, it is a challenge to efficiently train SNNs due to their non-differentiability. Most existing methods either suffer from high latency (i.e., long simulation time steps), or cannot achieve as high performance as Artificial Neural Networks (ANNs). In this paper, we propose the Differentiation on Spike Representation (DSR) method, which could achieve high performance that is competitive to ANNs yet with low latency. First, we encode the spike trains into spike representation using (weighted) firing rate coding. Based on the spike representation, we systematically derive that the spiking dynamics with common neural models can be represented as some sub-differentiable mapping. With this viewpoint, our proposed DSR method trains SNNs through gradients of the mapping and avoids the common non-differentiability problem in SNN training. Then we analyze the error when representing the specific mapping with the forward computation of the SNN. To reduce such error, we propose to train the spike threshold in each layer, and to introduce a new hyperparameter for the neural models. With these components, the DSR method can achieve state-of-the-art SNN performance with low latency on both static and neuromorphic datasets, including CIFAR-10, CIFAR-100, ImageNet, and DVS-CIFAR10."
                },
                {
                    "title": "Efficiently Modeling Long Sequences with Structured State Spaces",
                    "abstract": "A central goal of sequence modeling is designing a single principled model that can address sequence data across a range of modalities and tasks, particularly on long-range dependencies. Although conventional models including RNNs, CNNs, and Transformers have specialized variants for capturing long dependencies, they still struggle to scale to very long sequences of $10000$ or more steps. A promising recent approach proposed modeling sequences by simulating the fundamental state space model (SSM) \\( x'(t) = Ax(t) + Bu(t), y(t) = Cx(t) + Du(t) \\), and showed that for appropriate choices of the state matrix \\( A \\), this system could handle long-range dependencies mathematically and empirically. However, this method has prohibitive computation and memory requirements, rendering it infeasible as a general sequence modeling solution. We propose the Structured State Space sequence model (S4) based on a new parameterization for the SSM, and show that it can be computed much more efficiently than prior approaches while preserving their theoretical strengths. Our technique involves conditioning \\( A \\) with a low-rank correction, allowing it to be diagonalized stably and reducing the SSM to the well-studied computation of a Cauchy kernel. S4 achieves strong empirical results across a diverse range of established benchmarks, including (i) 91\\% accuracy on sequential CIFAR-10 with no data augmentation or auxiliary losses, on par with a larger 2-D ResNet, (ii) substantially closing the gap to Transformers on image and language modeling tasks, while performing generation $60\\times$ faster (iii) SoTA on every task from the Long Range Arena benchmark, including solving the challenging Path-X task of length 16k that all prior work fails on, while being as efficient as all competitors."
                },
                {
                    "title": "Early Convolutions Help Transformers See Better",
                    "abstract": "Vision transformer (ViT) models exhibit substandard optimizability. In particular, they are sensitive to the choice of optimizer (AdamW vs. SGD), optimizer hyperparameters, and training schedule length. In comparison, modern convolutional neural networks are easier to optimize. Why is this the case? In this work, we conjecture that the issue lies with the patchify stem of ViT models, which is implemented by a stride-p p*p convolution (p=16 by default) applied to the input image. This large-kernel plus large-stride convolution runs counter to typical design choices of convolutional layers in neural networks. To test whether this atypical design choice causes an issue, we analyze the optimization behavior of ViT models with their original patchify stem versus a simple counterpart where we replace the ViT stem by a small number of stacked stride-two 3*3 convolutions. While the vast majority of computation in the two ViT designs is identical, we find that this small change in early visual processing results in markedly different training behavior in terms of the sensitivity to optimization settings as well as the final model accuracy. Using a convolutional stem in ViT dramatically increases optimization stability and also improves peak performance (by ~1-2% top-1 accuracy on ImageNet-1k), while maintaining flops and runtime. The improvement can be observed across the wide spectrum of model complexities (from 1G to 36G flops) and dataset scales (from ImageNet-1k to ImageNet-21k). These findings lead us to recommend using a standard, lightweight convolutional stem for ViT models in this regime as a more robust architectural choice compared to the original ViT model design."
                },
                {
                    "title": "MLP-Mixer: An all-MLP Architecture for Vision",
                    "abstract": "Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e.\"mixing\"the per-location features), and one with MLPs applied across patches (i.e.\"mixing\"spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers."
                },
                {
                    "title": "AST: Audio Spectrogram Transformer",
                    "abstract": "In the past decade, convolutional neural networks (CNNs) have been widely adopted as the main building block for end-to-end audio classification models, which aim to learn a direct mapping from audio spectrograms to corresponding labels. To better capture long-range global context, a recent trend is to add a self-attention mechanism on top of the CNN, forming a CNN-attention hybrid model. However, it is unclear whether the reliance on a CNN is necessary, and if neural networks purely based on attention are sufficient to obtain good performance in audio classification. In this paper, we answer the question by introducing the Audio Spectrogram Transformer (AST), the first convolution-free, purely attention-based model for audio classification. We evaluate AST on various audio classification benchmarks, where it achieves new state-of-the-art results of 0.485 mAP on AudioSet, 95.6% accuracy on ESC-50, and 98.1% accuracy on Speech Commands V2."
                },
                {
                    "title": "Keyword Transformer: A Self-Attention Model for Keyword Spotting",
                    "abstract": "The Transformer architecture has been successful across many domains, including natural language processing, computer vision and speech recognition. In keyword spotting, self-attention has primarily been used on top of convolutional or recurrent encoders. We investigate a range of ways to adapt the Transformer architecture to keyword spotting and introduce the Keyword Transformer (KWT), a fully self-attentional architecture that exceeds state-of-the-art performance across multiple tasks without any pre-training or additional data. Surprisingly, this simple architecture outperforms more complex models that mix convolutional, recurrent and attentive layers. KWT can be used as a drop-in replacement for these models, setting two new benchmark records on the Google Speech Commands dataset with 98.6% and 97.7% accuracy on the 12 and 35-command tasks respectively."
                },
                {
                    "title": "Parallelizing Legendre Memory Unit Training",
                    "abstract": "Recently, a new recurrent neural network (RNN) named the Legendre Memory Unit (LMU) was proposed and shown to achieve state-of-the-art performance on several benchmark datasets. Here we leverage the linear time-invariant (LTI) memory component of the LMU to construct a simplified variant that can be parallelized during training (and yet executed as an RNN during inference), thus overcoming a well known limitation of training RNNs on GPUs. We show that this reformulation that aids parallelizing, which can be applied generally to any deep network whose recurrent components are linear, makes training up to 200 times faster. Second, to validate its utility, we compare its performance against the original LMU and a variety of published LSTM and transformer networks on seven benchmarks, ranging from psMNIST to sentiment analysis to machine translation. We demonstrate that our models exhibit superior performance on all datasets, often using fewer parameters. For instance, our LMU sets a new state-of-the-art result on psMNIST, and uses half the parameters while outperforming DistilBERT and LSTM models on IMDB sentiment analysis."
                },
                {
                    "title": "Deep Residual Learning in Spiking Neural Networks",
                    "abstract": "Deep Spiking Neural Networks (SNNs) present optimization difficulties for gradient-based approaches due to discrete binary activation and complex spatial-temporal dynamics. Considering the huge success of ResNet in deep learning, it would be natural to train deep SNNs with residual learning. Previous Spiking ResNet mimics the standard residual block in ANNs and simply replaces ReLU activation layers with spiking neurons, which suffers the degradation problem and can hardly implement residual learning. In this paper, we propose the spike-element-wise (SEW) ResNet to realize residual learning in deep SNNs. We prove that the SEW ResNet can easily implement identity mapping and overcome the vanishing/exploding gradient problems of Spiking ResNet. We evaluate our SEW ResNet on ImageNet, DVS Gesture, and CIFAR10-DVS datasets, and show that SEW ResNet outperforms the state-of-the-art directly trained SNNs in both accuracy and time-steps. Moreover, SEW ResNet can achieve higher performance by simply adding more layers, providing a simple method to train deep SNNs. To our best knowledge, this is the first time that directly training deep SNNs with more than 100 layers becomes possible. Our codes are available at https://github.com/fangwei123456/Spike-Element-Wise-ResNet."
                },
                {
                    "title": "Nystr\u00f6mformer: A Nystr\u00f6m-Based Algorithm for Approximating Self-Attention",
                    "abstract": "Transformers have emerged as a powerful tool for a broad range of natural language processing tasks. A key component that drives the impressive performance of Transformers is the self-attention mechanism that encodes the influence or dependence of other tokens on each specific token. While beneficial, the quadratic complexity of self-attention on the input sequence length has limited its application to longer sequences - a topic being actively studied in the community. To address this limitation, we propose Nystr\u00f6mformer - a model that exhibits favorable scalability as a function of sequence length. Our idea is based on adapting the Nystr\u00f6m method to approximate standard self-attention with O(n) complexity. The scalability of Nystr\u00f6mformer enables application to longer sequences with thousands of tokens. We perform evaluations on multiple downstream tasks on the GLUE benchmark and IMDB reviews with standard sequence length, and find that our Nystr\u00f6mformer performs comparably, or in a few cases, even slightly better, than standard self-attention. On longer sequence tasks in the Long Range Arena (LRA) benchmark, Nystr\u00f6mformer performs favorably relative to other efficient self-attention methods. Our code is available at https://github.com/mlpen/Nystromformer."
                },
                {
                    "title": "Long Range Arena: A Benchmark for Efficient Transformers",
                    "abstract": "Transformers do not scale very well to long sequence lengths largely because of quadratic self-attention complexity. In the recent months, a wide spectrum of efficient, fast Transformers have been proposed to tackle this problem, more often than not claiming superior or comparable model quality to vanilla Transformer models. To this date, there is no well-established consensus on how to evaluate this class of models. Moreover, inconsistent benchmarking on a wide spectrum of tasks and datasets makes it difficult to assess relative model quality amongst many models. This paper proposes a systematic and unified benchmark, LRA, specifically focused on evaluating model quality under long-context scenarios. Our benchmark is a suite of tasks consisting of sequences ranging from $1K$ to $16K$ tokens, encompassing a wide range of data types and modalities such as text, natural, synthetic images, and mathematical expressions requiring similarity, structural, and visual-spatial reasoning. We systematically evaluate ten well-established long-range Transformer models (Reformers, Linformers, Linear Transformers, Sinkhorn Transformers, Performers, Synthesizers, Sparse Transformers, and Longformers) on our newly proposed benchmark suite. LRA paves the way towards better understanding this class of efficient Transformer models, facilitates more research in this direction, and presents new challenging tasks to tackle. Our benchmark code will be released at this https URL."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "HiPPO: Recurrent Memory with Optimal Polynomial Projections",
                    "abstract": "A central problem in learning from sequential data is representing cumulative history in an incremental fashion as more data is processed. We introduce a general framework (HiPPO) for the online compression of continuous signals and discrete time series by projection onto polynomial bases. Given a measure that specifies the importance of each time step in the past, HiPPO produces an optimal solution to a natural online function approximation problem. As special cases, our framework yields a short derivation of the recent Legendre Memory Unit (LMU) from first principles, and generalizes the ubiquitous gating mechanism of recurrent neural networks such as GRUs. This formal framework yields a new memory update mechanism (HiPPO-LegS) that scales through time to remember all history, avoiding priors on the timescale. HiPPO-LegS enjoys the theoretical benefits of timescale robustness, fast updates, and bounded gradients. By incorporating the memory dynamics into recurrent neural networks, HiPPO RNNs can empirically capture complex temporal dependencies. On the benchmark permuted MNIST dataset, HiPPO-LegS sets a new state-of-the-art accuracy of 98.3%. Finally, on a novel trajectory classification task testing robustness to out-of-distribution timescales and missing data, HiPPO-LegS outperforms RNN and neural ODE baselines by 25-40% accuracy."
                },
                {
                    "title": "Big Bird: Transformers for Longer Sequences",
                    "abstract": "Transformers-based models, such as BERT, have been one of the most successful deep learning models for NLP. Unfortunately, one of their core limitations is the quadratic dependency (mainly in terms of memory) on the sequence length due to their full attention mechanism. To remedy this, we propose, BigBird, a sparse attention mechanism that reduces this quadratic dependency to linear. We show that BigBird is a universal approximator of sequence functions and is Turing complete, thereby preserving these properties of the quadratic, full attention model. Along the way, our theoretical analysis reveals some of the benefits of having $O(1)$ global tokens (such as CLS), that attend to the entire sequence as part of the sparse attention mechanism. The proposed sparse attention can handle sequences of length up to 8x of what was previously possible using similar hardware. As a consequence of the capability to handle longer context, BigBird drastically improves performance on various NLP tasks such as question answering and summarization. We also propose novel applications to genomics data."
                },
                {
                    "title": "Long Short-Term Memory Spiking Networks and Their Applications",
                    "abstract": "Recent advances in event-based neuromorphic systems have resulted in significant interest in the use and development of spiking neural networks (SNNs). However, the non-differentiable nature of spiking neurons makes SNNs incompatible with conventional backpropagation techniques. In spite of the significant progress made in training conventional deep neural networks (DNNs), training methods for SNNs still remain relatively poorly understood. In this paper, we present a novel framework for training recurrent SNNs. Analogous to the benefits presented by recurrent neural networks (RNNs) in learning time series models within DNNs, we develop SNNs based on long short-term memory (LSTM) networks. We show that LSTM spiking networks learn the timing of the spikes and temporal dependencies. We also develop a methodology for error backpropagation within LSTM-based SNNs. The developed architecture and method for backpropagation within LSTM-based SNNs enable them to learn long-term dependencies with comparable results to conventional LSTMs. Code is available on github; https://github.com/AliLotfi92/SNNLSTM"
                },
                {
                    "title": "Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention",
                    "abstract": "Transformers achieve remarkable performance in several tasks but due to their quadratic complexity, with respect to the input's length, they are prohibitively slow for very long sequences. To address this limitation, we express the self-attention as a linear dot-product of kernel feature maps and make use of the associativity property of matrix products to reduce the complexity from $\\mathcal{O}\\left(N^2\\right)$ to $\\mathcal{O}\\left(N\\right)$, where $N$ is the sequence length. We show that this formulation permits an iterative implementation that dramatically accelerates autoregressive transformers and reveals their relationship to recurrent neural networks. Our linear transformers achieve similar performance to vanilla transformers and they are up to 4000x faster on autoregressive prediction of very long sequences."
                },
                {
                    "title": "Linformer: Self-Attention with Linear Complexity",
                    "abstract": "Large transformer models have shown extraordinary success in achieving state-of-the-art results in many natural language processing applications. However, training and deploying these models can be prohibitively costly for long sequences, as the standard self-attention mechanism of the Transformer uses $O(n^2)$ time and space with respect to sequence length. In this paper, we demonstrate that the self-attention mechanism can be approximated by a low-rank matrix. We further exploit this finding to propose a new self-attention mechanism, which reduces the overall self-attention complexity from $O(n^2)$ to $O(n)$ in both time and space. The resulting linear transformer, the \\textit{Linformer}, performs on par with standard Transformer models, while being much more memory- and time-efficient."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Conformer: Convolution-augmented Transformer for Speech Recognition",
                    "abstract": "Recently Transformer and Convolution neural network (CNN) based models have shown promising results in Automatic Speech Recognition (ASR), outperforming Recurrent neural networks (RNNs). Transformer models are good at capturing content-based global interactions, while CNNs exploit local features effectively. In this work, we achieve the best of both worlds by studying how to combine convolution neural networks and transformers to model both local and global dependencies of an audio sequence in a parameter-efficient way. To this regard, we propose the convolution-augmented transformer for speech recognition, named Conformer. Conformer significantly outperforms the previous Transformer and CNN based models achieving state-of-the-art accuracies. On the widely used LibriSpeech benchmark, our model achieves WER of 2.1%/4.3% without using a language model and 1.9%/3.9% with an external language model on test/testother. We also observe competitive performance of 2.7%/6.3% with a small model of only 10M parameters."
                },
                {
                    "title": "Spike frequency adaptation supports network computations on temporally dispersed information",
                    "abstract": "For solving tasks such as recognizing a song, answering a question, or inverting a sequence of symbols, cortical microcircuits need to integrate and manipulate information that was dispersed over time during the preceding seconds. Creating biologically realistic models for the underlying computations, especially with spiking neurons and for behaviorally relevant integration time spans, is notoriously difficult. We examine the role of spike frequency adaptation in such computations and find that it has a surprisingly large impact. The inclusion of this well known property of a substantial fraction of neurons in the neocortex \u2014 especially in higher areas of the human neocortex \u2014 moves the performance of spiking neural network models for computations on network inputs that are temporally dispersed from a fairly low level up to the performance level of the human brain."
                },
                {
                    "title": "Legendre Memory Units: Continuous-Time Representation in Recurrent Neural Networks",
                    "abstract": "We propose a novel memory cell for recurrent neural networks that dynamically maintains information across long windows of time using relatively few resources. The Legendre Memory Unit~(LMU) is mathematically derived to orthogonalize its continuous-time history -- doing so by solving $d$ coupled ordinary differential equations~(ODEs), whose phase space linearly maps onto sliding windows of time via the Legendre polynomials up to degree $d - 1$. Backpropagation across LMUs outperforms equivalently-sized LSTMs on a chaotic time-series prediction task, improves memory capacity by two orders of magnitude, and significantly reduces training and inference times. LMUs can efficiently handle temporal dependencies spanning $100\\text{,}000$ time-steps, converge rapidly, and use few internal state-variables to learn complex functions spanning long windows of time -- exceeding state-of-the-art performance among RNNs on permuted sequential MNIST. These results are due to the network's disposition to learn scale-invariant features independently of step size. Backpropagation through the ODE solver allows each layer to adapt its internal time-step, enabling the network to learn task-relevant time-scales. We demonstrate that LMU memory cells can be implemented using $m$ recurrently-connected Poisson spiking neurons, $\\mathcal{O}( m )$ time and memory, with error scaling as $\\mathcal{O}( d / \\sqrt{m} )$. We discuss implementations of LMUs on analog and digital neuromorphic hardware."
                },
                {
                    "title": "Enabling Spike-Based Backpropagation for Training Deep Neural Network Architectures",
                    "abstract": "Spiking Neural Networks (SNNs) have recently emerged as a prominent neural computing paradigm. However, the typical shallow SNN architectures have limited capacity for expressing complex representations while training deep SNNs using input spikes has not been successful so far. Diverse methods have been proposed to get around this issue such as converting off-the-shelf trained deep Artificial Neural Networks (ANNs) to SNNs. However, the ANN-SNN conversion scheme fails to capture the temporal dynamics of a spiking system. On the other hand, it is still a difficult problem to directly train deep SNNs using input spike events due to the discontinuous, non-differentiable nature of the spike generation function. To overcome this problem, we propose an approximate derivative method that accounts for the leaky behavior of LIF neurons. This method enables training deep convolutional SNNs directly (with input spike events) using spike-based backpropagation. Our experiments show the effectiveness of the proposed spike-based learning on deep networks (VGG and Residual architectures) by achieving the best classification accuracies in MNIST, SVHN, and CIFAR-10 datasets compared to other SNNs trained with a spike-based learning. Moreover, we analyze sparse event-based computations to demonstrate the efficacy of the proposed SNN training method for inference operation in the spiking domain."
                },
                {
                    "title": "Spiking-YOLO: Spiking Neural Network for Energy-Efficient Object Detection",
                    "abstract": "Over the past decade, deep neural networks (DNNs) have demonstrated remarkable performance in a variety of applications. As we try to solve more advanced problems, increasing demands for computing and power resources has become inevitable. Spiking neural networks (SNNs) have attracted widespread interest as the third-generation of neural networks due to their event-driven and low-powered nature. SNNs, however, are difficult to train, mainly owing to their complex dynamics of neurons and non-differentiable spike operations. Furthermore, their applications have been limited to relatively simple tasks such as image classification. In this study, we investigate the performance degradation of SNNs in a more challenging regression problem (i.e., object detection). Through our in-depth analysis, we introduce two novel methods: channel-wise normalization and signed neuron with imbalanced threshold, both of which provide fast and accurate information transmission for deep SNNs. Consequently, we present a first spiked-based object detection model, called Spiking-YOLO. Our experiments show that Spiking-YOLO achieves remarkable results that are comparable (up to 98%) to those of Tiny YOLO on non-trivial datasets, PASCAL VOC and MS COCO. Furthermore, Spiking-YOLO on a neuromorphic chip consumes approximately 280 times less energy than Tiny YOLO and converges 2.3 to 4 times faster than previous SNN conversion methods."
                },
                {
                    "title": "Direct Training for Spiking Neural Networks: Faster, Larger, Better",
                    "abstract": "Spiking neural networks (SNNs) that enables energy efficient implementation on emerging neuromorphic hardware are gaining more attention. Yet now, SNNs have not shown competitive performance compared with artificial neural networks (ANNs), due to the lack of effective learning algorithms and efficient programming frameworks. We address this issue from two aspects: (1) We propose a neuron normalization technique to adjust the neural selectivity and develop a direct learning algorithm for deep SNNs. (2) Via narrowing the rate coding window and converting the leaky integrate-and-fire (LIF) model into an explicitly iterative version, we present a Pytorch-based implementation method towards the training of large-scale SNNs. In this way, we are able to train deep SNNs with tens of times speedup. As a result, we achieve significantly better accuracy than the reported works on neuromorphic datasets (N-MNIST and DVSCIFAR10), and comparable accuracy as existing ANNs and pre-trained SNNs on non-spiking datasets (CIFAR10). To our best knowledge, this is the first work that demonstrates direct training of deep SNNs with high performance on CIFAR10, and the efficient implementation provides a new way to explore the potential of SNNs."
                },
                {
                    "title": "A neural attention model for speech command recognition",
                    "abstract": "This paper introduces a convolutional recurrent network with attention for speech command recognition. Attention models are powerful tools to improve performance on natural language, image captioning and speech tasks. The proposed model establishes a new state-of-the-art accuracy of 94.1% on Google Speech Commands dataset V1 and 94.5% on V2 (for the 20-commands recognition task), while still keeping a small footprint of only 202K trainable parameters. Results are compared with previous convolutional implementations on 5 different tasks (20 commands recognition (V1 and V2), 12 commands recognition (V1), 35 word recognition (V1) and left-right (V1)). We show detailed performance results and demonstrate that the proposed attention mechanism not only improves performance but also allows inspecting what regions of the audio were taken into consideration by the network when outputting a given category."
                },
                {
                    "title": "Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition",
                    "abstract": "Describes an audio dataset of spoken words designed to help train and evaluate keyword spotting systems. Discusses why this task is an interesting challenge, and why it requires a specialized dataset that is different from conventional datasets used for automatic speech recognition of full sentences. Suggests a methodology for reproducible and comparable accuracy metrics for this task. Describes how the data was collected and verified, what it contains, previous versions and properties. Concludes by reporting baseline results of models trained on this dataset."
                },
                {
                    "title": "Super-convergence: very fast training of neural networks using large learning rates",
                    "abstract": "In this paper, we describe a phenomenon, which we named \u201csuper-convergence\u201d, where neural networks can be trained an order of magnitude faster than with standard training methods. The existence of super-convergence is relevant to understanding why deep networks generalize well. One of the key elements of super-convergence is training with one learning rate cycle and a large maximum learning rate. A insight that allows super-convergence training is that large learning rates regularize the training, hence requiring a reduction of all other forms of regularization in order to preserve an optimal regularization balance. We also derive a simplification of the Hessian Free optimization method to compute an estimate of the optimal learning rate. Experiments demonstrate super-convergence for Cifar-10/100, MNIST and Imagenet datasets, and resnet, wide-resnet, densenet, and inception architectures. In addition, we show that super-convergence provides a greater boost in performance relative to standard training when the amount of labeled training data is limited. The architectures and code to replicate the figures in this paper are available at github.com/lnsmith54/super-convergence."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Spatio-Temporal Backpropagation for Training High-Performance Spiking Neural Networks",
                    "abstract": "Spiking neural networks (SNNs) are promising in ascertaining brain-like behaviors since spikes are capable of encoding spatio-temporal information. Recent schemes, e.g., pre-training from artificial neural networks (ANNs) or direct training based on backpropagation (BP), make the high-performance supervised training of SNNs possible. However, these methods primarily fasten more attention on its spatial domain information, and the dynamics in temporal domain are attached less significance. Consequently, this might lead to the performance bottleneck, and scores of training techniques shall be additionally required. Another underlying problem is that the spike activity is naturally non-differentiable, raising more difficulties in supervised training of SNNs. In this paper, we propose a spatio-temporal backpropagation (STBP) algorithm for training high-performance SNNs. In order to solve the non-differentiable problem of SNNs, an approximated derivative for spike activity is proposed, being appropriate for gradient descent training. The STBP algorithm combines the layer-by-layer spatial domain (SD) and the timing-dependent temporal domain (TD), and does not require any additional complicated skill. We evaluate this method through adopting both the fully connected and convolutional architecture on the static MNIST dataset, a custom object detection dataset, and the dynamic N-MNIST dataset. Results bespeak that our approach achieves the best accuracy compared with existing state-of-the-art algorithms on spiking networks. This work provides a new perspective to investigate the high-performance SNNs for future brain-like computing paradigm with rich spatio-temporal dynamics."
                },
                {
                    "title": "Training Deep Spiking Neural Networks Using Backpropagation",
                    "abstract": "Deep spiking neural networks (SNNs) hold the potential for improving the latency and energy efficiency of deep neural networks through data-driven event-based computation. However, training such networks is difficult due to the non-differentiable nature of spike events. In this paper, we introduce a novel technique, which treats the membrane potentials of spiking neurons as differentiable signals, where discontinuities at spike times are considered as noise. This enables an error backpropagation mechanism for deep SNNs that follows the same principles as in conventional deep networks, but works directly on spike signals and membrane potentials. Compared with previous methods relying on indirect training and conversion, our technique has the potential to capture the statistics of spikes more precisely. We evaluate the proposed framework on artificially generated events from the original MNIST handwritten digit benchmark, and also on the N-MNIST benchmark recorded with an event-based dynamic vision sensor, in which the proposed method reduces the error rate by a factor of more than three compared to the best previous SNN, and also achieves a higher accuracy than a conventional convolutional neural network (CNN) trained and tested on the same data. We demonstrate in the context of the MNIST task that thanks to their event-driven operation, deep SNNs (both fully connected and convolutional) trained with our method achieve accuracy equivalent with conventional neural networks. In the N-MNIST example, equivalent accuracy is achieved with about five times fewer computational operations."
                },
                {
                    "title": "Gaussian Error Linear Units (GELUs)",
                    "abstract": "We propose the Gaussian Error Linear Unit (GELU), a high-performing neural network activation function. The GELU activation function is $x\\Phi(x)$, where $\\Phi(x)$ the standard Gaussian cumulative distribution function. The GELU nonlinearity weights inputs by their value, rather than gates inputs by their sign as in ReLUs ($x\\mathbf{1}_{x>0}$). We perform an empirical evaluation of the GELU nonlinearity against the ReLU and ELU activations and find performance improvements across all considered computer vision, natural language processing, and speech tasks."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "A Simple Way to Initialize Recurrent Networks of Rectified Linear Units",
                    "abstract": "Learning long term dependencies in recurrent networks is difficult due to vanishing and exploding gradients. To overcome this difficulty, researchers have developed sophisticated optimization techniques and network architectures. In this paper, we propose a simpler solution that use recurrent neural networks composed of rectified linear units. Key to our solution is the use of the identity matrix or its scaled version to initialize the recurrent weight matrix. We find that our solution is comparable to LSTM on our four benchmarks: two toy problems involving long-range temporal structures, a large language modeling problem and a benchmark speech recognition problem."
                },
                {
                    "title": "On the Properties of Neural Machine Translation: Encoder\u2013Decoder Approaches",
                    "abstract": "Neural machine translation is a relatively new approach to statistical machine translation based purely on neural networks. The neural machine translation models often consist of an encoder and a decoder. The encoder extracts a fixed-length representation from a variable-length input sentence, and the decoder generates a correct translation from this representation. In this paper, we focus on analyzing the properties of the neural machine translation using two models; RNN Encoder--Decoder and a newly proposed gated recursive convolutional neural network. We show that the neural machine translation performs relatively well on short sentences without unknown words, but its performance degrades rapidly as the length of the sentence and the number of unknown words increase. Furthermore, we find that the proposed gated recursive convolutional network learns a grammatical structure of a sentence automatically."
                },
                {
                    "title": "A million spiking-neuron integrated circuit with a scalable communication network and interface",
                    "abstract": "Modeling computer chips on real brains Computers are nowhere near as versatile as our own brains. Merolla et al. applied our present knowledge of the structure and function of the brain to design a new computer chip that uses the same wiring rules and architecture. The flexible, scalable chip operated efficiently in real time, while using very little power. Science, this issue p. 668 A large-scale computer chip mimics many features of a real brain. Inspired by the brain\u2019s structure, we have developed an efficient, scalable, and flexible non\u2013von Neumann architecture that leverages contemporary silicon technology. To demonstrate, we built a 5.4-billion-transistor chip with 4096 neurosynaptic cores interconnected via an intrachip network that integrates 1 million programmable spiking neurons and 256 million configurable synapses. Chips can be tiled in two dimensions via an interchip communication interface, seamlessly scaling the architecture to a cortexlike sheet of arbitrary size. The architecture is well suited to many applications that use complex neural networks in real time, for example, multiobject detection and classification. With 400-pixel-by-240-pixel video input at 30 frames per second, the chip consumes 63 milliwatts."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "Long Short-Term Memory",
                    "abstract": "Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, LSTM can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units. Multiplicative gate units learn to open and close access to the constant error flow. LSTM is local in space and time; its computational complexity per time step and weight is O. 1. Our experiments with artificial data involve local, distributed, real-valued, and noisy pattern representations. In comparisons with real-time recurrent learning, back propagation through time, recurrent cascade correlation, Elman nets, and neural sequence chunking, LSTM leads to many more successful runs, and learns much faster. LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "The mnist database of handwritten digits",
                    "abstract": "Disclosed is an improved articulated bar flail having shearing edges for efficiently shredding materials. An improved shredder cylinder is disclosed with a plurality of these flails circumferentially spaced and pivotally attached to the periphery of a rotatable shaft. Also disclosed is an improved shredder apparatus which has a pair of these shredder cylinders mounted to rotate about spaced parallel axes which cooperates with a conveyer apparatus which has a pair of inclined converging conveyer belts with one of the belts mounted to move with respect to the other belt to allow the transport of articles of various sizes therethrough."
                }
            ],
            "categories": [
                "cs.NE",
                "cs.AI",
                "cs.LG",
                "cs.SD",
                "eess.AS"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we design a sequential network architecture that combines the advantages of Legendre Memory Units (LMUs) and convolutional structures to achieve competitive performance with Transformers while maintaining lower computational complexity?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing need for efficient models that can handle long sequences without the high computational costs associated with Transformers. By bridging the performance gap between LMUs and Transformers, this research could lead to advancements in various applications, including Natural Language Processing, Computer Vision, and real-time data processing. The proposed LMUFormer architecture could inspire future research into hybrid models that leverage the strengths of both RNNs and Transformers, potentially leading to new practical applications in resource-constrained environments.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in effectively integrating the sequential processing capabilities of LMUs with the representational power of Transformers while ensuring that the resulting architecture remains computationally efficient. Naive approaches may fail because they could either compromise the sequential nature of the model or lead to excessive complexity that negates the benefits of using LMUs. Technical obstacles include designing convolutional structures that do not disrupt the sequential processing and ensuring that the model can maintain performance across various sequence lengths and tasks.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either improving RNNs or developing Transformers, often overlooking the potential of combining their strengths. Limitations in existing solutions include a lack of architectures that can process data sequentially while also leveraging complex representations. Additionally, the computational challenges associated with integrating convolutional structures with LMUs have not been adequately addressed. Our approach differs by proposing the LMUFormer architecture, which specifically targets these integration challenges while maintaining low complexity and high performance.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the LMUFormer architecture, which integrates LMUs with convolutional patch embedding and channel mixers. We will evaluate this model using the Speech Commands dataset and the Long Range Arena benchmark, measuring performance through metrics such as accuracy, parameter count, and FLOPs. We expect the LMUFormer to outperform existing RNN models while significantly reducing computational requirements, and the spiking variant of the architecture is anticipated to achieve state-of-the-art performance in spiking neural networks, demonstrating the effectiveness of our approach in real-time"
            }
        },
        "author_data": {
            "e2822b0b-11d1-4ddd-9d07-04bc0019b102": {
                "pk": "e2822b0b-11d1-4ddd-9d07-04bc0019b102",
                "name": "Zeyu Liu",
                "collaborators": [
                    "P. Beerel",
                    "G. Datta",
                    "Anni Li",
                    "Souvik Kundu",
                    "Junrui Wan",
                    "Lianghao Jiang",
                    "Haoqing Deng",
                    "Robert S. Aviles",
                    "James Diffenderfer",
                    "B. Kailkhura"
                ],
                "domain": [
                    "Spiking Neural Networks",
                    "Parameter-Efficient Fine-Tuning",
                    "Energy Efficiency",
                    "Vision Tasks"
                ],
                "publications": [
                    {
                        "title": "AFLoRA: Adaptive Freezing of Low Rank Adaptation in Parameter Efficient Fine-Tuning of Large Models",
                        "abstract": "We present a novel Parameter-Efficient Fine-Tuning (PEFT) method, dubbed as Adaptive Freezing of Low Rank Adaptation (AFLoRA). Specifically, for each pre-trained frozen weight tensor, we add a parallel path of trainable low-rank matrices, namely a down-projection and an up-projection matrix, each of which is followed by a feature transformation vector. Based on a novel freezing score, we the incrementally freeze these projection matrices during fine-tuning to reduce the computation and alleviate over-fitting. Our experimental results demonstrate that we can achieve state-of-the-art performance with an average improvement of up to $0.85\\%$ as evaluated on GLUE benchmark while yeilding up to $9.5\\times$ fewer average trainable parameters. While compared in terms of runtime, AFLoRA can yield up to $1.86\\times$ improvement as opposed to similar PEFT alternatives. Besides the practical utility of our approach, we provide insights on the trainability requirements of LoRA paths at different modules and the freezing schedule for the different projection matrices. Code will be released."
                    },
                    {
                        "title": "Training Ultra-Low-Latency Spiking Neural Networks from Scratch",
                        "abstract": "Spiking Neural networks (SNN) have emerged as an attractive spatio-temporal computing paradigm for a wide range of low-power vision tasks. However, state-of-the-art (SOTA) SNN models either incur multiple time steps which hinder their deployment in real-time use cases or increase the training complexity significantly. To mitigate this concern, we present a training framework (from scratch) for SNNs with ultra-low (down to 1) time steps that leverages the Hoyer regularizer. We calculate the threshold for each BANN layer as the Hoyer extremum of a clipped version of its activation map. The clipping value is determined through training using gradient descent with our Hoyer regularizer. We evaluate the efficacy of our training framework on large-scale vision tasks, including traditional and event-based image recognition and object detection. Our experiments demonstrate up to 34\u00d7 increase in compute efficiency with a marginal accuracy/mAP drop compared to non-spiking networks. Finally, we implement our framework in the Lava-DL library, thereby enabling the deployment of our SNN models in the Loihi neuromorphic chip."
                    },
                    {
                        "title": "Bridging the Gap Between Spiking Neural Networks & LSTMs for Latency & Energy Efficiency",
                        "abstract": "Spiking Neural Networks (SNNs) have emerged as an attractive spatio-temporal computing paradigm for complex vision tasks. However, most existing works yield models that require many time steps and do not leverage the inherent temporal dynamics of spiking neural networks, even for sequential tasks. Motivated by this observation, we propose an optimized spiking long short-term memory networks (LSTM) training framework that involves a novel ANN-to-SNN conversion framework, followed by SNN fine-tuning via backpropagation through time (BPTT). In particular, we propose novel activation functions in the source LSTM architecture and convert a judiciously selected subset of them to leaky-integrate-and-fire (LIF) activations with optimal bias shifts. Moreover, we propose a pipelined parallel processing scheme that hides the SNN time steps, significantly improving system latency, especially for long sequences. The resulting SNNs have high activation sparsity and require only accumulate operations (AC), in contrast to expensive multiply-and-accumulates (MAC) needed for ANNs, except for the input layer when using direct encoding, yielding significant improvements in energy efficiency. We evaluate our framework on sequential learning tasks including temporal MNIST, Google Speech Commands (GSC), and UCI Smartphone datasets on different LSTM architectures. We obtain test accuracy of 94.75 % with only 2 time steps on the GSC dataset with $\\sim 4.1\\times$ lower energy than an iso-architecture standard LSTM."
                    },
                    {
                        "title": "When Bio-Inspired Computing meets Deep Learning: Low-Latency, Accurate, & Energy-Efficient Spiking Neural Networks from Artificial Neural Networks",
                        "abstract": "Bio-inspired Spiking Neural Networks (SNN) are now demonstrating comparable accuracy to intricate convolutional neural networks (CNN), all while delivering remarkable energy and latency efficiency when deployed on neuromorphic hardware. In particular, ANN-to-SNN conversion has recently gained significant traction in developing deep SNNs with close to state-of-the-art (SOTA) test accuracy on complex image recognition tasks. However, advanced ANN-to-SNN conversion approaches demonstrate that for lossless conversion, the number of SNN time steps must equal the number of quantization steps in the ANN activation function. Reducing the number of time steps significantly increases the conversion error. Moreover, the spiking activity of the SNN, which dominates the compute energy in neuromorphic chips, does not reduce proportionally with the number of time steps. To mitigate the accuracy concern, we propose a novel ANN-to-SNN conversion framework, that incurs an exponentially lower number of time steps compared to that required in the SOTA conversion approaches. Our framework modifies the SNN integrate-and-fire (IF) neuron model with identical complexity and shifts the bias term of each batch normalization (BN) layer in the trained ANN. To mitigate the spiking activity concern, we propose training the source ANN with a fine-grained L1 regularizer with surrogate gradients that encourages high spike sparsity in the converted SNN. Our proposed framework thus yields lossless SNNs with ultra-low latency, ultra-low compute energy, thanks to the ultra-low timesteps and high spike sparsity, and ultra-high test accuracy, for example, 73.30% with only 4 time steps on the ImageNet dataset."
                    },
                    {
                        "title": "Spiking Neural Networks with Dynamic Time Steps for Vision Transformers",
                        "abstract": "Spiking Neural Networks (SNNs) have emerged as a popular spatio-temporal computing paradigm for complex vision tasks. Recently proposed SNN training algorithms have significantly reduced the number of time steps (down to 1) for improved latency and energy efficiency, however, they target only convolutional neural networks (CNN). These algorithms, when applied on the recently spotlighted vision transformers (ViT), either require a large number of time steps or fail to converge. Based on analysis of the histograms of the ANN and SNN activation maps, we hypothesize that each ViT block has a different sensitivity to the number of time steps. We propose a novel training framework that dynamically allocates the number of time steps to each ViT module depending on a trainable score assigned to each timestep. In particular, we generate a scalar binary time step mask that filters spikes emitted by each neuron in a leaky-integrate-and-fire (LIF) layer. The resulting SNNs have high activation sparsity and require only accumulate operations (AC), except for the input embedding layer, in contrast to expensive multiply-and-accumulates (MAC) needed in traditional ViTs. This yields significant improvements in energy efficiency. We evaluate our training framework and resulting SNNs on image recognition tasks including CIFAR10, CIFAR100, and ImageNet with different ViT architectures. We obtain a test accuracy of 95.97% with 4.97 time steps with direct encoding on CIFAR10."
                    }
                ]
            },
            "3121c019-b892-4f2e-9503-8d27be1bc2fc": {
                "pk": "3121c019-b892-4f2e-9503-8d27be1bc2fc",
                "name": "Gourav Datta",
                "collaborators": [
                    "P. Beerel",
                    "Md. Abdullah-Al Kaiser",
                    "Akhilesh R. Jaiswal",
                    "Zeyu Liu",
                    "Ajey P. Jacob",
                    "Souvik Kundu",
                    "Sreetama Sarkar",
                    "Zihan Yin",
                    "Akhilesh Jaiswal",
                    "Yue Hu",
                    "Jordan G. Athas",
                    "Christian Duffee",
                    "P. Amiri",
                    "Kai Zheng",
                    "Chirayata Bhattacharyya",
                    "Manas Garg",
                    "Xin-Yu Ye",
                    "Yifei Liu",
                    "Srikar Mutnuri",
                    "Namo Asavisanu",
                    "Nora Ayanian",
                    "K. Psounis",
                    "Haoqing Deng",
                    "Robert S. Aviles",
                    "Shashank Nag",
                    "N. Chandrachoodan",
                    "Kira Beerel",
                    "James Diffenderfer",
                    "B. Kailkhura",
                    "Zixu Wang",
                    "Xinan Ye",
                    "Anni Li",
                    "Tyler Etchart",
                    "Vivek Yadav",
                    "Varsha Hedau",
                    "P. Natarajan",
                    "Shih-Fu Chang",
                    "Joe Mathai",
                    "A. Jacob"
                ],
                "domain": [
                    "Computer Vision",
                    "Neuromorphic Computing",
                    "Spiking Neural Networks",
                    "Energy Efficiency"
                ],
                "publications": [
                    {
                        "title": "Toward High Performance, Programmable Extreme-Edge Intelligence for Neuromorphic Vision Sensors utilizing Magnetic Domain Wall Motion-based MTJ",
                        "abstract": "The desire to empower resource-limited edge devices with computer vision (CV) must overcome the high energy consumption of collecting and processing vast sensory data. To address the challenge, this work proposes an energy-efficient non-von-Neumann in-pixel processing solution for neuromorphic vision sensors employing emerging (X) magnetic domain wall magnetic tunnel junction (MDWMTJ) for the first time, in conjunction with CMOS-based neuromorphic pixels. Our hybrid CMOS+X approach performs in-situ massively parallel asynchronous analog convolution, exhibiting low power consumption and high accuracy across various CV applications by leveraging the non-volatility and programmability of the MDWMTJ. Moreover, our developed device-circuit-algorithm co-design framework captures device constraints (low tunnel-magnetoresistance, low dynamic range) and circuit constraints (non-linearity, process variation, area consideration) based on monte-carlo simulations and device parameters utilizing GF22nm FD-SOI technology. Our experimental results suggest we can achieve an average of 45.3% reduction in backend-processor energy, maintaining similar front-end energy compared to the state-of-the-art and high accuracy of 79.17% and 95.99% on the DVS-CIFAR10 and IBM DVS128-Gesture datasets, respectively."
                    },
                    {
                        "title": "Energy-Efficient & Real-Time Computer Vision with Intelligent Skipping via Reconfigurable CMOS Image Sensors",
                        "abstract": "Current video-based computer vision (CV) applications typically suffer from high energy consumption due to reading and processing all pixels in a frame, regardless of their significance. While previous works have attempted to reduce this energy by skipping input patches or pixels and using feedback from the end task to guide the skipping algorithm, the skipping is not performed during the sensor read phase. As a result, these methods can not optimize the front-end sensor energy. Moreover, they may not be suitable for real-time applications due to the long latency of modern CV networks that are deployed in the back-end. To address this challenge, this paper presents a custom-designed reconfigurable CMOS image sensor (CIS) system that improves energy efficiency by selectively skipping uneventful regions or rows within a frame during the sensor's readout phase, and the subsequent analog-to-digital conversion (ADC) phase. A novel masking algorithm intelligently directs the skipping process in real-time, optimizing both the front-end sensor and back-end neural networks for applications including autonomous driving and augmented/virtual reality (AR/VR). Our system can also operate in standard mode without skipping, depending on application needs. We evaluate our hardware-algorithm co-design framework on object detection based on BDD100K and ImageNetVID, and gaze estimation based on OpenEDS, achieving up to 53% reduction in front-end sensor energy while maintaining state-of-the-art (SOTA) accuracy."
                    },
                    {
                        "title": "Training Ultra-Low-Latency Spiking Neural Networks from Scratch",
                        "abstract": "Spiking Neural networks (SNN) have emerged as an attractive spatio-temporal computing paradigm for a wide range of low-power vision tasks. However, state-of-the-art (SOTA) SNN models either incur multiple time steps which hinder their deployment in real-time use cases or increase the training complexity significantly. To mitigate this concern, we present a training framework (from scratch) for SNNs with ultra-low (down to 1) time steps that leverages the Hoyer regularizer. We calculate the threshold for each BANN layer as the Hoyer extremum of a clipped version of its activation map. The clipping value is determined through training using gradient descent with our Hoyer regularizer. We evaluate the efficacy of our training framework on large-scale vision tasks, including traditional and event-based image recognition and object detection. Our experiments demonstrate up to 34\u00d7 increase in compute efficiency with a marginal accuracy/mAP drop compared to non-spiking networks. Finally, we implement our framework in the Lava-DL library, thereby enabling the deployment of our SNN models in the Loihi neuromorphic chip."
                    },
                    {
                        "title": "Voltage-Controlled Magnetic Tunnel Junction based ADC-less Global Shutter Processing-in-Pixel for Extreme-Edge Intelligence",
                        "abstract": "The vast amount of data generated by camera sensors has prompted the exploration of energy-efficient processing solutions for deploying computer vision tasks on edge devices. Among the various approaches studied, processing-in-pixel integrates massively parallel analog computational capabilities at the extreme-edge, i.e., within the pixel array and exhibits enhanced energy and bandwidth efficiency by generating the output activations of the first neural network layer rather than the raw sensory data. In this article, we propose an energy and bandwidth efficient ADC-less processing-in-pixel architecture. This architecture implements an optimized binary activation neural network trained using Hoyer regularizer for high accuracy on complex vision tasks. In addition, we also introduce a global shutter burst memory read scheme utilizing fast and disturb-free read operation leveraging innovative use of nanoscale voltage-controlled magnetic tunnel junctions (VC-MTJs). Moreover, we develop an algorithmic framework incorporating device and circuit constraints (characteristic device switching behavior and circuit non-linearity) based on state-of-the-art fabricated VC-MTJ characteristics and extensive circuit simulations using commercial GlobalFoundries 22nm FDX technology. Finally, we evaluate the proposed system's performance on two complex datasets - CIFAR10 and ImageNet, showing improvements in front-end and communication energy efficiency by 8.2x and 8.5x respectively and reduction in bandwidth by 6x compared to traditional computer vision systems, without any significant drop in the test accuracy."
                    },
                    {
                        "title": "MaskVD: Region Masking for Efficient Video Object Detection",
                        "abstract": "Video tasks are compute-heavy and thus pose a challenge when deploying in real-time applications, particularly for tasks that require state-of-the-art Vision Transformers (ViTs). Several research efforts have tried to address this challenge by leveraging the fact that large portions of the video undergo very little change across frames, leading to redundant computations in frame-based video processing. In particular, some works leverage pixel or semantic differences across frames, however, this yields limited latency benefits with significantly increased memory overhead. This paper, in contrast, presents a strategy for masking regions in video frames that leverages the semantic information in images and the temporal correlation between frames to significantly reduce FLOPs and latency with little to no penalty in performance over baseline models. In particular, we demonstrate that by leveraging extracted features from previous frames, ViT backbones directly benefit from region masking, skipping up to 80% of input regions, improving FLOPs and latency by 3.14x and 1.5x. We improve memory and latency over the state-of-the-art (SOTA) by 2.3x and 1.14x, while maintaining similar detection performance. Additionally, our approach demonstrates promising results on convolutional neural networks (CNNs) and provides latency improvements over the SOTA up to 1.3x using specialized computational kernels."
                    },
                    {
                        "title": "Technology-Circuit-Algorithm Tri-Design for Processing-in-Pixel-in-Memory (P2M)",
                        "abstract": "The massive amounts of data generated by camera sensors motivate data processing inside pixel arrays, i.e., at the extreme-edge. Several critical developments have fueled recent interest in the processing-in-pixel-in-memory paradigm for a wide range of visual machine intelligence tasks, including (1) advances in 3D integration technology to enable complex processing inside each pixel in a 3D integrated manner while maintaining pixel density, (2) analog processing circuit techniques for massively parallel low-energy in-pixel computations, and (3) algorithmic techniques to mitigate non-idealities associated with analog processing through hardware-aware training schemes. This article presents a comprehensive technology-circuit-algorithm landscape that connects technology capabilities, circuit design strategies, and algorithmic optimizations to power, performance, area, bandwidth reduction, and application-level accuracy metrics. We present our results using a comprehensive co-design framework incorporating hardware and algorithmic optimizations for various complex real-life visual intelligence tasks mapped onto our P2M paradigm."
                    },
                    {
                        "title": "FireFly: A Synthetic Dataset for Ember Detection in Wildfire",
                        "abstract": "This paper presents \"FireFly\", a synthetic dataset for ember detection created using Unreal Engine 4 (UE4), designed to overcome the current lack of ember-specific training resources. To create the dataset, we present a tool that allows the automated generation of the synthetic labeled dataset with adjustable parameters, enabling data diversity from various environmental conditions, making the dataset both diverse and customizable based on user requirements. We generated a total of 19,273 frames that have been used to evaluate FireFly on four popular object detection models. Further to minimize human intervention, we leveraged a trained model to create a semi-automatic labeling process for real-life ember frames. Moreover, we demonstrated an up to 8.57% improvement in mean Average Precision (mAP) in real-world wildfire scenarios compared to models trained exclusively on a small real dataset."
                    },
                    {
                        "title": "Bridging the Gap Between Spiking Neural Networks & LSTMs for Latency & Energy Efficiency",
                        "abstract": "Spiking Neural Networks (SNNs) have emerged as an attractive spatio-temporal computing paradigm for complex vision tasks. However, most existing works yield models that require many time steps and do not leverage the inherent temporal dynamics of spiking neural networks, even for sequential tasks. Motivated by this observation, we propose an optimized spiking long short-term memory networks (LSTM) training framework that involves a novel ANN-to-SNN conversion framework, followed by SNN fine-tuning via backpropagation through time (BPTT). In particular, we propose novel activation functions in the source LSTM architecture and convert a judiciously selected subset of them to leaky-integrate-and-fire (LIF) activations with optimal bias shifts. Moreover, we propose a pipelined parallel processing scheme that hides the SNN time steps, significantly improving system latency, especially for long sequences. The resulting SNNs have high activation sparsity and require only accumulate operations (AC), in contrast to expensive multiply-and-accumulates (MAC) needed for ANNs, except for the input layer when using direct encoding, yielding significant improvements in energy efficiency. We evaluate our framework on sequential learning tasks including temporal MNIST, Google Speech Commands (GSC), and UCI Smartphone datasets on different LSTM architectures. We obtain test accuracy of 94.75 % with only 2 time steps on the GSC dataset with $\\sim 4.1\\times$ lower energy than an iso-architecture standard LSTM."
                    },
                    {
                        "title": "Design Considerations for 3D Heterogeneous Integration Driven Analog Processing-in-Pixel for Extreme-Edge Intelligence",
                        "abstract": "Given the progress in computer vision, image sensors are broadening their capabilities, which requires adding data processing close to or within the pixel chips. In this context, in-pixel computing has emerged as a notable paradigm, offering the capability to process data within the pixel unit itself. Interestingly, state-of-art in-pixel paradigms rely on high-density 3D heterogeneous integration to establish a per-pixel connection with vertically aligned analog processing units. This article provides a comprehensive review of the most recent developments in in-pixel computing and its relation to 3D heterogeneous integration. It offers an in-depth examination of innovative circuit design, adaptations in algorithms, and the challenges in 3D integration technology for sensor chips, thereby presenting a holistic perspective on the future trajectory of in-pixel computing driven by advances in 3D integration."
                    },
                    {
                        "title": "ViTA: A Vision Transformer Inference Accelerator for Edge Applications",
                        "abstract": "Vision Transformer models, such as ViT, Swin Transformer, and Transformer-in-Transformer, have recently gained significant traction in computer vision tasks due to their ability to capture the global relation between features which leads to superior performance. However, they are compute-heavy and difficult to deploy in resource-constrained edge devices. Existing hardware accelerators, including those for the closely-related BERT transformer models, do not target highly resource-constrained environments. In this paper, we address this gap and propose ViTA - a configurable hardware accelerator for inference of vision transformer models, targeting resource-constrained edge computing devices and avoiding repeated off-chip memory accesses. We employ a head-level pipeline and inter-layer MLP optimizations, and can support several commonly used vision transformer models with changes solely in our control logic. We achieve nearly 90% hardware utilization efficiency on most vision transformer models, report a power of 0.88W when synthesised with a clock of 150 MHz, and get reasonable frame rates - all of which makes ViTA suitable for edge applications."
                    },
                    {
                        "title": "Let's Roll: Synthetic Dataset Analysis for Pedestrian Detection Across Different Shutter Types",
                        "abstract": "Computer vision (CV) pipelines are typically evaluated on datasets processed by image signal processing (ISP) pipelines even though, for resource-constrained applications, an important research goal is to avoid as many ISP steps as possible. In particular, most CV datasets consist of global shutter (GS) images even though most cameras today use a rolling shutter (RS). This paper studies the impact of different shutter mechanisms on machine learning (ML) object detection models on a synthetic dataset that we generate using the advanced simulation capabilities of Unreal Engine 5 (UE5). In particular, we train and evaluate mainstream detection models with our synthetically-generated paired GS and RS datasets to ascertain whether there exists a significant difference in detection accuracy between these two shutter modalities, especially when capturing low-speed objects (e.g., pedestrians). The results of this emulation framework indicate the performance between them are remarkably congruent for coarse-grained detection (mean average precision (mAP) for IOU=0.5), but have significant differences for fine-grained measures of detection accuracy (mAP for IOU=0.5:0.95). This implies that ML pipelines might not need explicit correction for RS for many object detection applications, but mitigating RS effects in ISP-less ML pipelines that target fine-grained location of the objects may need additional research."
                    },
                    {
                        "title": "When Bio-Inspired Computing meets Deep Learning: Low-Latency, Accurate, & Energy-Efficient Spiking Neural Networks from Artificial Neural Networks",
                        "abstract": "Bio-inspired Spiking Neural Networks (SNN) are now demonstrating comparable accuracy to intricate convolutional neural networks (CNN), all while delivering remarkable energy and latency efficiency when deployed on neuromorphic hardware. In particular, ANN-to-SNN conversion has recently gained significant traction in developing deep SNNs with close to state-of-the-art (SOTA) test accuracy on complex image recognition tasks. However, advanced ANN-to-SNN conversion approaches demonstrate that for lossless conversion, the number of SNN time steps must equal the number of quantization steps in the ANN activation function. Reducing the number of time steps significantly increases the conversion error. Moreover, the spiking activity of the SNN, which dominates the compute energy in neuromorphic chips, does not reduce proportionally with the number of time steps. To mitigate the accuracy concern, we propose a novel ANN-to-SNN conversion framework, that incurs an exponentially lower number of time steps compared to that required in the SOTA conversion approaches. Our framework modifies the SNN integrate-and-fire (IF) neuron model with identical complexity and shifts the bias term of each batch normalization (BN) layer in the trained ANN. To mitigate the spiking activity concern, we propose training the source ANN with a fine-grained L1 regularizer with surrogate gradients that encourages high spike sparsity in the converted SNN. Our proposed framework thus yields lossless SNNs with ultra-low latency, ultra-low compute energy, thanks to the ultra-low timesteps and high spike sparsity, and ultra-high test accuracy, for example, 73.30% with only 4 time steps on the ImageNet dataset."
                    },
                    {
                        "title": "Neuromorphic-P2M: processing-in-pixel-in-memory paradigm for neuromorphic image sensors",
                        "abstract": "Edge devices equipped with computer vision must deal with vast amounts of sensory data with limited computing resources. Hence, researchers have been exploring different energy-efficient solutions such as near-sensor, in-sensor, and in-pixel processing, bringing the computation closer to the sensor. In particular, in-pixel processing embeds the computation capabilities inside the pixel array and achieves high energy efficiency by generating low-level features instead of the raw data stream from CMOS image sensors. Many different in-pixel processing techniques and approaches have been demonstrated on conventional frame-based CMOS imagers; however, the processing-in-pixel approach for neuromorphic vision sensors has not been explored so far. In this work, for the first time, we propose an asynchronous non-von-Neumann analog processing-in-pixel paradigm to perform convolution operations by integrating in-situ multi-bit multi-channel convolution inside the pixel array performing analog multiply and accumulate (MAC) operations that consume significantly less energy than their digital MAC alternative. To make this approach viable, we incorporate the circuit's non-ideality, leakage, and process variations into a novel hardware-algorithm co-design framework that leverages extensive HSpice simulations of our proposed circuit using the GF22nm FD-SOI technology node. We verified our framework on state-of-the-art neuromorphic vision sensor datasets and show that our solution consumes ~2\u00d7 lower backend-processor energy while maintaining almost similar front-end (sensor) energy on the IBM DVS128-Gesture dataset than the state-of-the-art while maintaining a high test accuracy of 88.36%."
                    },
                    {
                        "title": "FixPix: Fixing Bad Pixels using Deep Learning",
                        "abstract": "Efficient and effective on-line detection and correction of bad-pixels can improve yield and increase the expected lifetime of image sensors. This paper presents a comprehensive Deep Learning (DL) based on-line detection and correction approach, suitable for a wide range of pixel corruption rates. A confidence calibrated segmentation approach is introduced, which achieves nearly perfect bad pixel detection, even with a few training samples. A computationally light-weight correction algorithm is proposed for low rates of pixel corruption, that surpasses the accuracy of traditional interpolation-based techniques. In addition, a vision transformer (ViT) auto-encoder based image reconstruction approach is presented which yields promising results for high rates of pixel corruption or clustered defects. Unlike previous methods, which use proprietary images, we demonstrate the efficacy of the proposed methods on the open-source Samsung S7 ISP and MIT-Adobe FiveK datasets. Our approaches yield up to 99.6% detection accuracy with<0.6% false positives and corrected images within 1.5% average pixel error from 70% corrupted images. We achieve correction error at par with the state-of-the-art (SoTA) DL methods for clustered defects with less than half the computational cost."
                    },
                    {
                        "title": "Spiking Neural Networks with Dynamic Time Steps for Vision Transformers",
                        "abstract": "Spiking Neural Networks (SNNs) have emerged as a popular spatio-temporal computing paradigm for complex vision tasks. Recently proposed SNN training algorithms have significantly reduced the number of time steps (down to 1) for improved latency and energy efficiency, however, they target only convolutional neural networks (CNN). These algorithms, when applied on the recently spotlighted vision transformers (ViT), either require a large number of time steps or fail to converge. Based on analysis of the histograms of the ANN and SNN activation maps, we hypothesize that each ViT block has a different sensitivity to the number of time steps. We propose a novel training framework that dynamically allocates the number of time steps to each ViT module depending on a trainable score assigned to each timestep. In particular, we generate a scalar binary time step mask that filters spikes emitted by each neuron in a leaky-integrate-and-fire (LIF) layer. The resulting SNNs have high activation sparsity and require only accumulate operations (AC), except for the input embedding layer, in contrast to expensive multiply-and-accumulates (MAC) needed in traditional ViTs. This yields significant improvements in energy efficiency. We evaluate our training framework and resulting SNNs on image recognition tasks including CIFAR10, CIFAR100, and ImageNet with different ViT architectures. We obtain a test accuracy of 95.97% with 4.97 time steps with direct encoding on CIFAR10."
                    },
                    {
                        "title": "Asd-Transformer: Efficient Active Speaker Detection Using Self And Multimodal Transformers",
                        "abstract": "Multimodal active speaker detection (ASD) methods assign a speaking/not-speaking label per individual in a video clip. ASD is critical for applications such as natural human-computer interaction, speaker diarization, and video reframing. Recent work has shown the success of transformers in multimodal settings, thus we propose a novel framework that leverages modern transformer and concatenation mechanisms to efficiently capture the interaction between audio and video modalities for ASD. We achieve mAP similar to state-of-the-art (93.0% vs 93.5%) on the AVA-ActiveSpeaker dataset. Further, our model has ~3\u00d7 smaller size (15.23MB vs 49.82MB), reduced FLOPs count (11.8 vs 14.3), and lower training time (15h vs 38h). To verify our model is making predictions from the right visual cues, we computed saliency maps over input images. We found that in addition to mouth regions, the nose, cheek, and area under the eye were helpful in identifying active speakers. Our ablation study reveals that the mouth region alone achieved lower mAP (91.9% vs 93.0%) compared to full face region, supporting our hypothesis that facial expressions in addition to mouth region are useful for ASD."
                    },
                    {
                        "title": "In-Sensor & Neuromorphic Computing Are all You Need for Energy Efficient Computer Vision",
                        "abstract": "Due to the high activation sparsity and use of accumulates (AC) instead of expensive multiply-and-accumulates (MAC), neuromorphic spiking neural networks (SNNs) have emerged as a promising low-power alternative to traditional DNNs for several computer vision (CV) applications. However, most existing SNNs require multiple time steps for acceptable inference accuracy, hindering real-time deployment and increasing spiking activity and, consequently, energy consumption. Recent works proposed direct encoding that directly feeds the analog pixel values in the first layer of the SNN in order to significantly reduce the number of time steps. Although the overhead for the first layer MACs with direct encoding is negligible for deep SNNs and the CV processing is efficient using SNNs, the data transfer between the image sensors and the downstream processing costs significant bandwidth and may dominate the total energy. To mitigate this concern, we propose an in-sensor computing hardware-software co-design framework for SNNs targeting image recognition tasks. Our approach reduces the bandwidth between sensing and processing by 12\u221296\u00d7 and the resulting total energy by 2.32\u00d7 compared to traditional CV processing, with a 3.8% reduction in accuracy on ImageNet."
                    },
                    {
                        "title": "Toward Efficient Hyperspectral Image Processing inside Camera Pixels",
                        "abstract": "Hyperspectral cameras generate a large amount of data due to the presence of hundreds of spectral bands as opposed to only three channels (red, green, and blue) in traditional cameras. This requires a signi\ufb01cant amount of data transmis-sion between the hyperspectral image sensor and a processor used to classify/detect/track the images, frame by frame, expending high energy and causing bandwidth and security bottlenecks. To mitigate this problem, we propose a form of processing-in-pixel (PIP) that leverages advanced CMOS technologies to enable the pixel array to perform a wide range of complex operations required by the modern convolutional neural networks (CNN) for hyperspectral image recognition (HSI). Consequently, our PIP-optimized custom CNN layers effectively compress the input data, signi\ufb01cantly reducing the bandwidth required to transmit the data downstream to the HSI processing unit. This reduces the average energy con-sumption associated with pixel array of cameras and the CNN processing unit by 25 . 06 \u00d7 and 3 . 90 \u00d7 respectively, compared to existing hardware implementations. Our custom models yield average test accuracies within 0 . 56% of the baseline models for the standard HSI benchmarks."
                    }
                ]
            },
            "5bb05d8c-2719-4704-992e-acfa68760071": {
                "pk": "5bb05d8c-2719-4704-992e-acfa68760071",
                "name": "Anni Li",
                "collaborators": [
                    "Zeyu Liu",
                    "P. Beerel",
                    "Souvik Kundu",
                    "Junrui Wan",
                    "Lianghao Jiang",
                    "G. Datta"
                ],
                "domain": [
                    "Parameter-Efficient Fine-Tuning",
                    "Spiking Neural Networks",
                    "Vision Transformers",
                    "Energy Efficiency"
                ],
                "publications": [
                    {
                        "title": "AFLoRA: Adaptive Freezing of Low Rank Adaptation in Parameter Efficient Fine-Tuning of Large Models",
                        "abstract": "We present a novel Parameter-Efficient Fine-Tuning (PEFT) method, dubbed as Adaptive Freezing of Low Rank Adaptation (AFLoRA). Specifically, for each pre-trained frozen weight tensor, we add a parallel path of trainable low-rank matrices, namely a down-projection and an up-projection matrix, each of which is followed by a feature transformation vector. Based on a novel freezing score, we the incrementally freeze these projection matrices during fine-tuning to reduce the computation and alleviate over-fitting. Our experimental results demonstrate that we can achieve state-of-the-art performance with an average improvement of up to $0.85\\%$ as evaluated on GLUE benchmark while yeilding up to $9.5\\times$ fewer average trainable parameters. While compared in terms of runtime, AFLoRA can yield up to $1.86\\times$ improvement as opposed to similar PEFT alternatives. Besides the practical utility of our approach, we provide insights on the trainability requirements of LoRA paths at different modules and the freezing schedule for the different projection matrices. Code will be released."
                    },
                    {
                        "title": "Spiking Neural Networks with Dynamic Time Steps for Vision Transformers",
                        "abstract": "Spiking Neural Networks (SNNs) have emerged as a popular spatio-temporal computing paradigm for complex vision tasks. Recently proposed SNN training algorithms have significantly reduced the number of time steps (down to 1) for improved latency and energy efficiency, however, they target only convolutional neural networks (CNN). These algorithms, when applied on the recently spotlighted vision transformers (ViT), either require a large number of time steps or fail to converge. Based on analysis of the histograms of the ANN and SNN activation maps, we hypothesize that each ViT block has a different sensitivity to the number of time steps. We propose a novel training framework that dynamically allocates the number of time steps to each ViT module depending on a trainable score assigned to each timestep. In particular, we generate a scalar binary time step mask that filters spikes emitted by each neuron in a leaky-integrate-and-fire (LIF) layer. The resulting SNNs have high activation sparsity and require only accumulate operations (AC), except for the input embedding layer, in contrast to expensive multiply-and-accumulates (MAC) needed in traditional ViTs. This yields significant improvements in energy efficiency. We evaluate our training framework and resulting SNNs on image recognition tasks including CIFAR10, CIFAR100, and ImageNet with different ViT architectures. We obtain a test accuracy of 95.97% with 4.97 time steps with direct encoding on CIFAR10."
                    }
                ]
            },
            "4628fd05-5f76-4df5-9e43-45c4363ad2a5": {
                "pk": "4628fd05-5f76-4df5-9e43-45c4363ad2a5",
                "name": "Peter Anthony Beerel",
                "collaborators": [
                    "Souvik Kundu",
                    "G. Datta",
                    "Sreetama Sarkar",
                    "Md. Abdullah-Al Kaiser",
                    "Robert S. Aviles",
                    "Akhilesh R. Jaiswal",
                    "Zeyu Liu",
                    "Akhilesh Jaiswal",
                    "S. Razmkhah",
                    "M. Pedram",
                    "Dake Chen",
                    "Yuke Zhang",
                    "Ajey P. Jacob",
                    "Kai Zheng",
                    "Anni Li",
                    "Junrui Wan",
                    "Lianghao Jiang",
                    "Xi Li",
                    "Lei Lu",
                    "Zhaorui Ni",
                    "Mingye Li",
                    "Sandeep Gupta",
                    "Shiduo Li",
                    "Chenghao Li",
                    "Jordan G. Athas",
                    "Christian Duffee",
                    "P. Amiri",
                    "Wenhao Song",
                    "Mingyi Rao",
                    "Yunning Li",
                    "Can Li",
                    "Ye Zhuo",
                    "Fuxi Cai",
                    "Mingche Wu",
                    "Wenbo Yin",
                    "Zongze Li",
                    "Qiang Wei",
                    "Sangsoo Lee",
                    "Hengfang Zhu",
                    "Lei Gong",
                    "Mark D. Barnell",
                    "Qing Wu",
                    "Mike Shuo-Wei Chen",
                    "Ning Ge",
                    "Miao Hu",
                    "Qiangfei Xia",
                    "J. J. Yang",
                    "K. PhalgunG",
                    "Chirayata Bhattacharyya",
                    "C. Fourie",
                    "Kyle Jackman",
                    "J. Delport",
                    "L. Schindler",
                    "Tessa Hall",
                    "P. Febvre",
                    "L. Iwanikow",
                    "O. Chen",
                    "C. Ayala",
                    "N. Yoshikawa",
                    "M. Law",
                    "T. Weingartner",
                    "Yanzhi Wang",
                    "S. Gupta",
                    "Haipeng Zha",
                    "M. A. Karamuftuoglu",
                    "Zihan Yin",
                    "Manas Garg",
                    "Yue Hu",
                    "Xin-Yu Ye",
                    "Yifei Liu",
                    "Srikar Mutnuri",
                    "Namo Asavisanu",
                    "Nora Ayanian",
                    "K. Psounis"
                ],
                "domain": [
                    "Computer Vision",
                    "Neuromorphic Computing",
                    "Energy Efficiency",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Toward High Performance, Programmable Extreme-Edge Intelligence for Neuromorphic Vision Sensors utilizing Magnetic Domain Wall Motion-based MTJ",
                        "abstract": "The desire to empower resource-limited edge devices with computer vision (CV) must overcome the high energy consumption of collecting and processing vast sensory data. To address the challenge, this work proposes an energy-efficient non-von-Neumann in-pixel processing solution for neuromorphic vision sensors employing emerging (X) magnetic domain wall magnetic tunnel junction (MDWMTJ) for the first time, in conjunction with CMOS-based neuromorphic pixels. Our hybrid CMOS+X approach performs in-situ massively parallel asynchronous analog convolution, exhibiting low power consumption and high accuracy across various CV applications by leveraging the non-volatility and programmability of the MDWMTJ. Moreover, our developed device-circuit-algorithm co-design framework captures device constraints (low tunnel-magnetoresistance, low dynamic range) and circuit constraints (non-linearity, process variation, area consideration) based on monte-carlo simulations and device parameters utilizing GF22nm FD-SOI technology. Our experimental results suggest we can achieve an average of 45.3% reduction in backend-processor energy, maintaining similar front-end energy compared to the state-of-the-art and high accuracy of 79.17% and 95.99% on the DVS-CIFAR10 and IBM DVS128-Gesture datasets, respectively."
                    },
                    {
                        "title": "AFLoRA: Adaptive Freezing of Low Rank Adaptation in Parameter Efficient Fine-Tuning of Large Models",
                        "abstract": "We present a novel Parameter-Efficient Fine-Tuning (PEFT) method, dubbed as Adaptive Freezing of Low Rank Adaptation (AFLoRA). Specifically, for each pre-trained frozen weight tensor, we add a parallel path of trainable low-rank matrices, namely a down-projection and an up-projection matrix, each of which is followed by a feature transformation vector. Based on a novel freezing score, we the incrementally freeze these projection matrices during fine-tuning to reduce the computation and alleviate over-fitting. Our experimental results demonstrate that we can achieve state-of-the-art performance with an average improvement of up to $0.85\\%$ as evaluated on GLUE benchmark while yeilding up to $9.5\\times$ fewer average trainable parameters. While compared in terms of runtime, AFLoRA can yield up to $1.86\\times$ improvement as opposed to similar PEFT alternatives. Besides the practical utility of our approach, we provide insights on the trainability requirements of LoRA paths at different modules and the freezing schedule for the different projection matrices. Code will be released."
                    },
                    {
                        "title": "Energy-Efficient & Real-Time Computer Vision with Intelligent Skipping via Reconfigurable CMOS Image Sensors",
                        "abstract": "Current video-based computer vision (CV) applications typically suffer from high energy consumption due to reading and processing all pixels in a frame, regardless of their significance. While previous works have attempted to reduce this energy by skipping input patches or pixels and using feedback from the end task to guide the skipping algorithm, the skipping is not performed during the sensor read phase. As a result, these methods can not optimize the front-end sensor energy. Moreover, they may not be suitable for real-time applications due to the long latency of modern CV networks that are deployed in the back-end. To address this challenge, this paper presents a custom-designed reconfigurable CMOS image sensor (CIS) system that improves energy efficiency by selectively skipping uneventful regions or rows within a frame during the sensor's readout phase, and the subsequent analog-to-digital conversion (ADC) phase. A novel masking algorithm intelligently directs the skipping process in real-time, optimizing both the front-end sensor and back-end neural networks for applications including autonomous driving and augmented/virtual reality (AR/VR). Our system can also operate in standard mode without skipping, depending on application needs. We evaluate our hardware-algorithm co-design framework on object detection based on BDD100K and ImageNetVID, and gaze estimation based on OpenEDS, achieving up to 53% reduction in front-end sensor energy while maintaining state-of-the-art (SOTA) accuracy."
                    },
                    {
                        "title": "An Efficient and Scalable Clocking Assignment Algorithm for Multi-Threaded Multi-Phase Single Flux Quantum Circuits",
                        "abstract": "This paper proposes an approximation algorithm for clocking assignments that minimizes the insertion of path balancing DFFs for multi-threaded multi-phase clocking of SFQ circuits. Existing SFQ multi-phase clocking solutions have been shown to effectively reduce the number of required DFFs in-serted, however, the associated clock assignment algorithms have exponential complexity and can have prohibitively long runtimes for large circuits, limiting the scalability of this approach. Our proposed algorithm is based on a linear program (LP) that leads to solutions that are experimentally on average within 5% of the optimum, permitting multi-phase clocking schemes to scale to larger SFQ circuits than previous state of the art clocking assignment methods. We further extend our algorithm to support fanout sharing of the added DFFs, saving, on average, an additional 11 % of the inserted DFFs compared to the state of the art. Compared to traditional full path balancing (FPB) methods across 12 benchmarks, our enhanced LP saves 68.8 %, 80.0%, and 87.6% of the inserted DFFs for 2,3, and 4 clock phases respectively. Finally, we extend this approach to the generation of circuits that completely mitigate potential hold-time violations at the cost of either reduced DFF savings or, more generally, adding a clock phase."
                    },
                    {
                        "title": "Challenges and Unexplored Frontiers in Electronic Design Automation for Superconducting Digital Logic",
                        "abstract": "Positioned as a highly promising post-CMOS computing technology, superconductor electronics (SCE) offer the potential for unparalleled performance and energy efficiency gains compared to end-of-roadmap CMOS circuits. However, achieving very large-scale integration poses numerous challenges. These challenges span from the modeling and analysis of superconducting devices and logic gates to the intricate design of complex SCE circuits and systems. Addressing power and clock distribution issues, minimizing adverse effects of flux trappings, and mitigating stray electromagnetic fields in sensitive SCE circuitry are key challenges that need attention. Verification and testing of SCE circuits also remain open problems. Moreover, scaling the minimum feature sizes of SCE circuits, currently set at 150nm, presents critical scaling and physical design challenges that must be overcome. This review aims to delve into these issues, providing detailed insights while exploring existing or potential solutions to overcome them."
                    },
                    {
                        "title": "Linearizing Models for Efficient yet Robust Private Inference",
                        "abstract": "The growing concern about data privacy has led to the development of private inference (PI) frameworks in client-server applications which protects both data privacy and model IP. However, the cryptographic primitives required yield significant latency overhead which limits its wide-spread application. At the same time, changing environments demand the PI service to be robust against various naturally occurring and gradient-based perturbations. Despite several works focused on the development of latency-efficient models suitable for PI, the impact of these models on robustness has remained unexplored. Towards this goal, this paper presents RLNet, a class of robust linearized networks that can yield latency improvement via reduction of high-latency ReLU operations while improving the model performance on both clean and corrupted images. In particular, RLNet models provide a\"triple win ticket\"of improved classification accuracy on clean, naturally perturbed, and gradient-based perturbed images using a shared-mask shared-weight architecture with over an order of magnitude fewer ReLUs than baseline models. To demonstrate the efficacy of RLNet, we perform extensive experiments with ResNet and WRN model variants on CIFAR-10, CIFAR-100, and Tiny-ImageNet datasets. Our experimental evaluations show that RLNet can yield models with up to 11.14x fewer ReLUs, with accuracy close to the all-ReLU models, on clean, naturally perturbed, and gradient-based perturbed images. Compared with the SoTA non-robust linearized models at similar ReLU budgets, RLNet achieves an improvement in adversarial accuracy of up to ~47%, naturally perturbed accuracy up to ~16.4%, while improving clean image accuracy up to ~1.5%."
                    },
                    {
                        "title": "Training Ultra-Low-Latency Spiking Neural Networks from Scratch",
                        "abstract": "Spiking Neural networks (SNN) have emerged as an attractive spatio-temporal computing paradigm for a wide range of low-power vision tasks. However, state-of-the-art (SOTA) SNN models either incur multiple time steps which hinder their deployment in real-time use cases or increase the training complexity significantly. To mitigate this concern, we present a training framework (from scratch) for SNNs with ultra-low (down to 1) time steps that leverages the Hoyer regularizer. We calculate the threshold for each BANN layer as the Hoyer extremum of a clipped version of its activation map. The clipping value is determined through training using gradient descent with our Hoyer regularizer. We evaluate the efficacy of our training framework on large-scale vision tasks, including traditional and event-based image recognition and object detection. Our experiments demonstrate up to 34\u00d7 increase in compute efficiency with a marginal accuracy/mAP drop compared to non-spiking networks. Finally, we implement our framework in the Lava-DL library, thereby enabling the deployment of our SNN models in the Loihi neuromorphic chip."
                    },
                    {
                        "title": "A Joint Optimization of Buffer and Splitter Insertion for Phase-Skipping Adiabatic Quantum-Flux-Parametron Circuits",
                        "abstract": "Adiabatic Quantum-Flux-Parametron (AQFP) logic is a promising emerging device technology with six orders of magnitude lower power than CMOS. However, AQFP is challenged by the fact that every gate must be clocked, where proper data transfer requires connected gates to have shifted but overlapping clocks. As a result, buffers need to be used to balance re-convergent logic paths, a problem that is exacerbated by every multi-node fanout needing a tree of clocked splitters. Recent AQFP circuit design techniques have offered an opportunity to reduce buffer costs by supporting a notion of phase-skipping but the EDA support for these advanced circuits is limited. This paper proposes the first algorithm to optimize buffer and splitter insertion for phase-skipping AQFP circuits and achieves over 31\\% savings over existing buffer reduction schemes and up to 74\\% savings in buffers and splitter costs over the SOTA non-phase skipping circuits."
                    },
                    {
                        "title": "DIA: Diffusion based Inverse Network Attack on Collaborative Inference",
                        "abstract": "With the continuous expansion of neural networks in size and depth, and the growing popularity of machine learning as a service, collaborative inference systems present a promising approach for deploying models in resource-constrained computing environments. However, as the deployment of these systems gains traction, evaluating their privacy and security has become a critical issue. Towards this goal, this paper introduces a diffusion-based inverse network attack, named DIA, for collaborative inference systems that uses a novel feature map awareness conditioning mechanism to guide the diffusion model. Compared to prior approaches, our extensive empirical results demonstrate that the proposed attack achieves an average improvement of 29%, 20%, 30% in terms of SSIM, PSNR, and MSE when applied to convolutional neural networks (CNN), 18%, 17%, 61% to ResNet models, and 55%, 54%, 84% to Vision transformers (ViTs). Our results identify the significant vulnerability of ViTs and analyze the potential sources of this vulnerability. Based on our analysis, we raise caution regarding the deployment of transformer-based models in collaborative inference systems, emphasizing the need for careful consideration regarding the security of such models in collaborative settings."
                    },
                    {
                        "title": "Voltage-Controlled Magnetic Tunnel Junction based ADC-less Global Shutter Processing-in-Pixel for Extreme-Edge Intelligence",
                        "abstract": "The vast amount of data generated by camera sensors has prompted the exploration of energy-efficient processing solutions for deploying computer vision tasks on edge devices. Among the various approaches studied, processing-in-pixel integrates massively parallel analog computational capabilities at the extreme-edge, i.e., within the pixel array and exhibits enhanced energy and bandwidth efficiency by generating the output activations of the first neural network layer rather than the raw sensory data. In this article, we propose an energy and bandwidth efficient ADC-less processing-in-pixel architecture. This architecture implements an optimized binary activation neural network trained using Hoyer regularizer for high accuracy on complex vision tasks. In addition, we also introduce a global shutter burst memory read scheme utilizing fast and disturb-free read operation leveraging innovative use of nanoscale voltage-controlled magnetic tunnel junctions (VC-MTJs). Moreover, we develop an algorithmic framework incorporating device and circuit constraints (characteristic device switching behavior and circuit non-linearity) based on state-of-the-art fabricated VC-MTJ characteristics and extensive circuit simulations using commercial GlobalFoundries 22nm FDX technology. Finally, we evaluate the proposed system's performance on two complex datasets - CIFAR10 and ImageNet, showing improvements in front-end and communication energy efficiency by 8.2x and 8.5x respectively and reduction in bandwidth by 6x compared to traditional computer vision systems, without any significant drop in the test accuracy."
                    },
                    {
                        "title": "Programming memristor arrays with arbitrarily high precision for analog computing",
                        "abstract": "In-memory computing represents an effective method for modeling complex physical systems that are typically challenging for conventional computing architectures but has been hindered by issues such as reading noise and writing variability that restrict scalability, accuracy, and precision in high-performance computations. We propose and demonstrate a circuit architecture and programming protocol that converts the analog computing result to digital at the last step and enables low-precision analog devices to perform high-precision computing. We use a weighted sum of multiple devices to represent one number, in which subsequently programmed devices are used to compensate for preceding programming errors. With a memristor system-on-chip, we experimentally demonstrate high-precision solutions for multiple scientific computing tasks while maintaining a substantial power efficiency advantage over conventional digital approaches. Editor\u2019s summary In-memory computing can execute a large vector-matrix multiplication (VMM) within one computing cycle, but the relatively low precision of these analog methods has precluded its use with conventional computing. Song et al. show that for one-memristor-one-transistor arrays, most of the VMM can be performed to arbitrarily high precision before being output as a digital result (see the Perspective by Aimone and Agarwal). After a computation step by an array, subarrays dynamically compensate for residual errors of the previously programmed array. This method was used experimentally to solve partial differential equations with remarkable precision (<10\u221215 error) and with higher energy efficiency than digital computing. \u2014Phil Szuromi Compensation for programming errors with inaccurate crossbar memristor-transistor devices enables arbitrarily high precision."
                    },
                    {
                        "title": "Delay Balancing with Clock-Follow-Data: Optimizing Area Delay Trade-offs for Robust Rapid Single Flux Quantum Circuits",
                        "abstract": "This paper proposes an algorithm for synthesis of clock-follow-data designs that provides robustness against timing violations for RSFQ circuits while maintaining high performance and minimizing area costs. Since superconducting logic gates must be clocked, managing data flow is a challenging problem that often requires the insertion of many path balancing D Flips (DFFs) to properly sequence data, leading to a substantial increase in area. To address this challenge, we present an algorithm to insert DFFs into clock-follow-data RSFQ circuits that partially balances the delays within the circuit to achieve a target throughput while minimizing area. Our algorithm can account for expected timing variations and, by adjusting the bias of the clock network and clock frequency, we can mitigate unexpected timing violations post-fabrication. Quantifying the benefits of our approach with a benchmark suite with nominal delays, our designs offer an average 1.48x improvement in area delay product (ADP) over high frequency full path balancing (FPB) designs and a 2.07x improvement in ADP over the state of the art robust circuits provided by state-of-the-art (SOTA) multi-phase clocking solutions."
                    },
                    {
                        "title": "MaskVD: Region Masking for Efficient Video Object Detection",
                        "abstract": "Video tasks are compute-heavy and thus pose a challenge when deploying in real-time applications, particularly for tasks that require state-of-the-art Vision Transformers (ViTs). Several research efforts have tried to address this challenge by leveraging the fact that large portions of the video undergo very little change across frames, leading to redundant computations in frame-based video processing. In particular, some works leverage pixel or semantic differences across frames, however, this yields limited latency benefits with significantly increased memory overhead. This paper, in contrast, presents a strategy for masking regions in video frames that leverages the semantic information in images and the temporal correlation between frames to significantly reduce FLOPs and latency with little to no penalty in performance over baseline models. In particular, we demonstrate that by leveraging extracted features from previous frames, ViT backbones directly benefit from region masking, skipping up to 80% of input regions, improving FLOPs and latency by 3.14x and 1.5x. We improve memory and latency over the state-of-the-art (SOTA) by 2.3x and 1.14x, while maintaining similar detection performance. Additionally, our approach demonstrates promising results on convolutional neural networks (CNNs) and provides latency improvements over the SOTA up to 1.3x using specialized computational kernels."
                    },
                    {
                        "title": "Block Selective Reprogramming for On-device Training of Vision Transformers",
                        "abstract": "The ubiquity of vision transformers (ViTs) for various edge applications, including personalized learning, has created the demand for on-device fine-tuning. However, training with the limited memory and computation power of edge devices remains a significant challenge. In particular, the memory required for training is much higher than that needed for inference, primarily due to the need to store activations across all layers in order to compute the gradients needed for weight updates. Previous works have explored reducing this memory requirement via frozen-weight training as well storing the activations in a compressed format. However, these methods are deemed inefficient due to their inability to provide training or inference speedup. In this paper, we first investigate the limitations of existing on-device training methods aimed at reducing memory and compute requirements. We then present block selective reprogramming (BSR) in which we fine-tune only a fraction of total blocks of a pre-trained model and selectively drop tokens based on self-attention scores of the frozen layers. To show the efficacy of BSR, we present extensive evaluations on ViT-B and DeiT-S with five different datasets. Compared to the existing alternatives, our approach simultaneously reduces training memory by up to 1.4\u00d7 and compute cost by up to 2\u00d7 while maintaining similar accuracy. We also showcase results for Mixture-of-Expert (MoE) models, demonstrating the effectiveness of our approach in multitask learning scenarios. Code will be available at: https://github.com/sreetamasarkar/BSR."
                    },
                    {
                        "title": "Results From the ColdFlux Superconductor Integrated Circuit Design Tool Project",
                        "abstract": "In five and a half years, the ColdFlux project under the IARPA SuperTools program pushed the boundaries of digital and analog superconductor electronic design automation (S-EDA) tools. The SuperTools program demanded significant beyond-state-of-the-art deliverables in four main areas: RTL synthesis, architectures, and verification; analog design and layout synthesis; physical design and test; and technology CAD and cell library design. Through the work of academic groups scattered over four continents, the ColdFlux effort forged into a powerful set of open-source and commercial S-EDA tools unlike any before, rivaled only by a commercial toolchain from Synopsys under the same SuperTools umbrella. We present an overview of the tools from where we started to the eventual project deliverables. These include powerful simulation and extraction engines, magnetic field and flux trapping analysis, advanced clocking methods, multi-chip interface extraction and verification, unified multi-layer design-rule compliant track blocks for automated place and route of both rapid single flux quantum (RSFQ) and adiabatic quantum-flux-parametron (AQFP) cells, models and tools for validation and test, multi-bit single flux quantum (SFQ) cells, architecture innovations for full CPU designs and more. Comprehensive cell libraries and a process design kit (PDK) were developed with the ColdFlux tools. The AQFP cell library features a logically rich collection of 80+ cells, including 3- and 5-input logic gates, signal-driving boosters, and refined RSFQ-to-AQFP interfaces, while the RSFQ library has 30+ cells. Finally, we discuss how the full toolchain enables and enhances the superconductor IC design process."
                    },
                    {
                        "title": "Technology-Circuit-Algorithm Tri-Design for Processing-in-Pixel-in-Memory (P2M)",
                        "abstract": "The massive amounts of data generated by camera sensors motivate data processing inside pixel arrays, i.e., at the extreme-edge. Several critical developments have fueled recent interest in the processing-in-pixel-in-memory paradigm for a wide range of visual machine intelligence tasks, including (1) advances in 3D integration technology to enable complex processing inside each pixel in a 3D integrated manner while maintaining pixel density, (2) analog processing circuit techniques for massively parallel low-energy in-pixel computations, and (3) algorithmic techniques to mitigate non-idealities associated with analog processing through hardware-aware training schemes. This article presents a comprehensive technology-circuit-algorithm landscape that connects technology capabilities, circuit design strategies, and algorithmic optimizations to power, performance, area, bandwidth reduction, and application-level accuracy metrics. We present our results using a comprehensive co-design framework incorporating hardware and algorithmic optimizations for various complex real-life visual intelligence tasks mapped onto our P2M paradigm."
                    },
                    {
                        "title": "FireFly: A Synthetic Dataset for Ember Detection in Wildfire",
                        "abstract": "This paper presents \"FireFly\", a synthetic dataset for ember detection created using Unreal Engine 4 (UE4), designed to overcome the current lack of ember-specific training resources. To create the dataset, we present a tool that allows the automated generation of the synthetic labeled dataset with adjustable parameters, enabling data diversity from various environmental conditions, making the dataset both diverse and customizable based on user requirements. We generated a total of 19,273 frames that have been used to evaluate FireFly on four popular object detection models. Further to minimize human intervention, we leveraged a trained model to create a semi-automatic labeling process for real-life ember frames. Moreover, we demonstrated an up to 8.57% improvement in mean Average Precision (mAP) in real-world wildfire scenarios compared to models trained exclusively on a small real dataset."
                    },
                    {
                        "title": "Making Models Shallow Again: Jointly Learning to Reduce Non-Linearity and Depth for Latency-Efficient Private Inference",
                        "abstract": "Large number of ReLU and MAC operations of Deep neural networks make them ill-suited for latency and compute-efficient private inference. In this paper, we present a model optimization method that allows a model to learn to be shallow. In particular, we leverage the ReLU sensitivity of a convolutional block to remove a ReLU layer and merge its succeeding and preceding convolution layers to a shallow block. Unlike existing ReLU reduction methods, our joint reduction method can yield models with improved reduction of both ReLUs and linear operations by up to 1.73\u00d7 and 1.47\u00d7, respectively, evaluated with ResNet18 on CIFAR-100 without any significant accuracy-drop."
                    }
                ]
            }
        }
    },
    "2311.14900": {
        "paper_data": {
            "title": "Resfusion: Denoising Diffusion Probabilistic Models for Image Restoration Based on Prior Residual Noise",
            "url": "http://arxiv.org/abs/2311.14900v4",
            "arxiv_id": "2311.14900",
            "authors": [
                "Zhenning Shi",
                "Haoshuai Zheng",
                "Chen Xu",
                "Changsheng Dong",
                "Bin Pan",
                "Xueshuo Xie",
                "Along He",
                "Tao Li",
                "Huazhu Fu"
            ],
            "abstract": "Recently, research on denoising diffusion models has expanded its application to the field of image restoration. Traditional diffusion-based image restoration methods utilize degraded images as conditional input to effectively guide the reverse generation process, without modifying the original denoising diffusion process. However, since the degraded images already include low-frequency information, starting from Gaussian white noise will result in increased sampling steps. We propose Resfusion, a general framework that incorporates the residual term into the diffusion forward process, starting the reverse process directly from the noisy degraded images. The form of our inference process is consistent with the DDPM. We introduced a weighted residual noise, named resnoise, as the prediction target and explicitly provide the quantitative relationship between the residual term and the noise term in resnoise. By leveraging a smooth equivalence transformation, Resfusion determine the optimal acceleration step and maintains the integrity of existing noise schedules, unifying the training and inference processes. The experimental results demonstrate that Resfusion exhibits competitive performance on ISTD dataset, LOL dataset and Raindrop dataset with only five sampling steps. Furthermore, Resfusion can be easily applied to image generation and emerges with strong versatility. Our code and model are available at https://github.com/nkicsl/Resfusion.",
            "introduction": "   1 Introduction  Denoising diffusion models\u00a0[1, 2] have emerged as powerful and effective conditional generative models, demonstrating remarkable success in synthesizing high-fidelity data for image generation. Saharia et\u00a0al. [3] proved that these generative processes can be applied to image restoration by feeding degraded images as conditional input into the score network. SNIPS\u00a0[4] combines annealed Langevin dynamics and Newton\u2019s method to arrive at a posterior sampling algorithm, exploring the generative diffusion processes to solve the general linear inverse problems. Based on these, many diffusion-based models were adapted for downstream image restoration tasks\u00a0[5, 6, 7, 8, 9, 10, 11]. For traditional diffusion-based models, the reverse process begins with Gaussian white noise, considering only the degraded images as the conditional input. This results in an increased number of sampling steps. Image restoration tasks often focus on restoring and editing specific high-frequency details while preserving crucial low-frequency information, such as the image structure. The degraded images used as conditional input inherently contain the low-frequency information. Therefore, initiating the reverse process from Gaussian white noise for image restoration tasks appears unnecessary and inefficient.   Consequently, some works have proposed to generate clean images directly from degraded images or noisy degraded images. InDI\u00a0[12] restores clean images through the reverse process of direct iteration to degraded images; DDRM\u00a0[13] reformulate the image restoration tasks as inverse problems when the mapping between clean and degraded images is available; IR-SDE\u00a0[14] directly models the image degradation process using mean-reverting SDE (Stochastic Differential Equations); I2SB\u00a0[15] constructs a Schr\u00f6dinger bridge between clean and degraded data distributions; Resshift\u00a0[16] shifts the residual term from degraded low-resolution images to high-resolution images, performing the recovery in the latent space. Liu et\u00a0al. [17] introduced the Residual Denoising Diffusion Models (RDDM), generalizing the diffusion process of InDI and I2SB. RDDM points out that co-learning the residual term and the noise term can effectively improve the model performance. However, RDDM has some limitations. Firstly, RDDM predicts the residual term and the noise term separately, without explicitly specifying their quantitative relationship. Secondly, due to its forward process adopting an accumulation strategy for the residual term and the noise term, its forward and reverse processes are inconsistent with the DDPM\u00a0[1], which results in poor generalization and interpretability. Thirdly, RDDM requires the design of a complex noise schedule, as utilizing existing noise schedules would result in performance loss.   Figure 1: The proposed Resfusion is a general framework for image restoration and can be easily expand to image generation (setting x^0=0subscript^\ud835\udc6500\\hat{x}_{0}=0over^ start_ARG italic_x end_ARG start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT = 0). We introduce the residual term (R=x^0\u2212x0\ud835\udc45subscript^\ud835\udc650subscript\ud835\udc650R=\\hat{x}_{0}-x_{0}italic_R = over^ start_ARG italic_x end_ARG start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT - italic_x start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT) into the forward process, redefine q\u2062(xt|xt\u22121)\ud835\udc5econditionalsubscript\ud835\udc65\ud835\udc61subscript\ud835\udc65\ud835\udc611q(x_{t}|x_{t-1})italic_q ( italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT | italic_x start_POSTSUBSCRIPT italic_t - 1 end_POSTSUBSCRIPT ) to q\u2062(xt|xt\u22121,R)\ud835\udc5econditionalsubscript\ud835\udc65\ud835\udc61subscript\ud835\udc65\ud835\udc611\ud835\udc45q(x_{t}|x_{t-1},R)italic_q ( italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT | italic_x start_POSTSUBSCRIPT italic_t - 1 end_POSTSUBSCRIPT , italic_R ) (as shown by the o\u2062r\u2062a\u2062n\u2062g\u2062e\ud835\udc5c\ud835\udc5f\ud835\udc4e\ud835\udc5b\ud835\udc54\ud835\udc52\\color[rgb]{1,.5,0}{orange}italic_o italic_r italic_a italic_n italic_g italic_e arrow), and name this diffusion process as resnoise diffusion. Through employing a novel technique called \"smooth equivalence transformation\", we can directly use the degraded image x^0subscript^\ud835\udc650\\hat{x}_{0}over^ start_ARG italic_x end_ARG start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT to obtain xT\u2032subscript\ud835\udc65superscript\ud835\udc47\u2032x_{T^{\\prime}}italic_x start_POSTSUBSCRIPT",
            "references": [
                {
                    "title": "Diffusion Models in Low-Level Vision: A Survey",
                    "abstract": "Deep generative models have garnered significant attention in low-level vision tasks due to their generative capabilities. Among them, diffusion model-based solutions, characterized by a forward diffusion process and a reverse denoising process, have emerged as widely acclaimed for their ability to produce samples of superior quality and diversity. This ensures the generation of visually compelling results with intricate texture information. Despite their remarkable success, a noticeable gap exists in a comprehensive survey that amalgamates these pioneering diffusion model-based works and organizes the corresponding threads. This paper proposes the comprehensive review of diffusion model-based techniques. We present three generic diffusion modeling frameworks and explore their correlations with other deep generative models, establishing the theoretical foundation. Following this, we introduce a multi-perspective categorization of diffusion models, considering both the underlying framework and the target task. Additionally, we summarize extended diffusion models applied in other tasks, including medical, remote sensing, and video scenarios. Moreover, we provide an overview of commonly used benchmarks and evaluation metrics. We conduct a thorough evaluation, encompassing both performance and efficiency, of diffusion model-based techniques in three prominent tasks. Finally, we elucidate the limitations of current diffusion models and propose seven intriguing directions for future research. This comprehensive examination aims to facilitate a profound understanding of the landscape surrounding denoising diffusion models in the context of low-level vision tasks. A curated list of diffusion model-based techniques in over 20 low-level vision tasks can be found at https://github.com/ChunmingHe/awesome-diffusion-models-in-low-level-vision."
                },
                {
                    "title": "Reti-Diff: Illumination Degradation Image Restoration with Retinex-based Latent Diffusion Model",
                    "abstract": "Illumination degradation image restoration (IDIR) techniques aim to improve the visibility of degraded images and mitigate the adverse effects of deteriorated illumination. Among these algorithms, diffusion model (DM)-based methods have shown promising performance but are often burdened by heavy computational demands and pixel misalignment issues when predicting the image-level distribution. To tackle these problems, we propose to leverage DM within a compact latent space to generate concise guidance priors and introduce a novel solution called Reti-Diff for the IDIR task. Reti-Diff comprises two key components: the Retinex-based latent DM (RLDM) and the Retinex-guided transformer (RGformer). To ensure detailed reconstruction and illumination correction, RLDM is empowered to acquire Retinex knowledge and extract reflectance and illumination priors. These priors are subsequently utilized by RGformer to guide the decomposition of image features into their respective reflectance and illumination components. Following this, RGformer further enhances and consolidates the decomposed features, resulting in the production of refined images with consistent content and robustness to handle complex degradation scenarios. Extensive experiments show that Reti-Diff outperforms existing methods on three IDIR tasks, as well as downstream applications. Code will be available at \\url{https://github.com/ChunmingHe/Reti-Diff}."
                },
                {
                    "title": "Global Structure-Aware Diffusion Process for Low-Light Image Enhancement",
                    "abstract": "This paper studies a diffusion-based framework to address the low-light image enhancement problem. To harness the capabilities of diffusion models, we delve into this intricate process and advocate for the regularization of its inherent ODE-trajectory. To be specific, inspired by the recent research that low curvature ODE-trajectory results in a stable and effective diffusion process, we formulate a curvature regularization term anchored in the intrinsic non-local structures of image data, i.e., global structure-aware regularization, which gradually facilitates the preservation of complicated details and the augmentation of contrast during the diffusion process. This incorporation mitigates the adverse effects of noise and artifacts resulting from the diffusion process, leading to a more precise and flexible enhancement. To additionally promote learning in challenging regions, we introduce an uncertainty-guided regularization technique, which wisely relaxes constraints on the most extreme regions of the image. Experimental evaluations reveal that the proposed diffusion-based framework, complemented by rank-informed regularization, attains distinguished performance in low-light enhancement. The outcomes indicate substantial advancements in image quality, noise suppression, and contrast amplification in comparison with state-of-the-art methods. We believe this innovative approach will stimulate further exploration and advancement in low-light image processing, with potential implications for other applications of diffusion models. The code is publicly available at https://github.com/jinnh/GSAD."
                },
                {
                    "title": "Residual Denoising Diffusion Models",
                    "abstract": "We propose residual denoising diffusion models (RDDM), a novel dual diffusion process that decouples the traditional single denoising diffusion process into residual diffusion and noise diffusion. This dual diffusion framework expands the denoising-based diffusion models, initially uninterpretable for image restoration, into a unified and interpretable model for both image generation and restoration by introducing residuals. Specifically, our residual diffusion represents directional diffusion from the target image to the degraded input image and explicitly guides the reverse generation process for image restoration, while noise diffusion represents random perturbations in the diffusion process. The residual prioritizes certainty, while the noise emphasizes diversity, enabling RDDM to effectively unify tasks with varying certainty or diversity requirements, such as image generation and restoration. We demonstrate that our sampling process is consistent with that of DDPM and DDIM through coefficient transformation, and propose a partially path-independent generation process to better understand the reverse process. Notably, our RDDM enables a generic UNet, trained with only an L1 loss and a batch size of 1, to compete with state-of-the-art image restoration methods. We provide code and pre-trained models to encourage further exploration, application, and development of our innovative framework (https://github.com/nachifurlRDDM)."
                },
                {
                    "title": "LLDiffusion: Learning Degradation Representations in Diffusion Models for Low-Light Image Enhancement",
                    "abstract": "Current deep learning methods for low-light image enhancement (LLIE) typically rely on pixel-wise mapping learned from paired data. However, these methods often overlook the importance of considering degradation representations, which can lead to sub-optimal outcomes. In this paper, we address this limitation by proposing a degradation-aware learning scheme for LLIE using diffusion models, which effectively integrates degradation and image priors into the diffusion process, resulting in improved image enhancement. Our proposed degradation-aware learning scheme is based on the understanding that degradation representations play a crucial role in accurately modeling and capturing the specific degradation patterns present in low-light images. To this end, First, a joint learning framework for both image generation and image enhancement is presented to learn the degradation representations. Second, to leverage the learned degradation representations, we develop a Low-Light Diffusion model (LLDiffusion) with a well-designed dynamic diffusion module. This module takes into account both the color map and the latent degradation representations to guide the diffusion process. By incorporating these conditioning factors, the proposed LLDiffusion can effectively enhance low-light images, considering both the inherent degradation patterns and the desired color fidelity. Finally, we evaluate our proposed method on several well-known benchmark datasets, including synthetic and real-world unpaired datasets. Extensive experiments on public benchmarks demonstrate that our LLDiffusion outperforms state-of-the-art LLIE methods both quantitatively and qualitatively. The source code and pre-trained models are available at https://github.com/TaoWangzj/LLDiffusion."
                },
                {
                    "title": "ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shifting",
                    "abstract": "Diffusion-based image super-resolution (SR) methods are mainly limited by the low inference speed due to the requirements of hundreds or even thousands of sampling steps. Existing acceleration sampling techniques inevitably sacrifice performance to some extent, leading to over-blurry SR results. To address this issue, we propose a novel and efficient diffusion model for SR that significantly reduces the number of diffusion steps, thereby eliminating the need for post-acceleration during inference and its associated performance deterioration. Our method constructs a Markov chain that transfers between the high-resolution image and the low-resolution image by shifting the residual between them, substantially improving the transition efficiency. Additionally, an elaborate noise schedule is developed to flexibly control the shifting speed and the noise strength during the diffusion process. Extensive experiments demonstrate that the proposed method obtains superior or at least comparable performance to current state-of-the-art methods on both synthetic and real-world datasets, even only with 15 sampling steps. Our code and model are available at https://github.com/zsyOAOA/ResShift."
                },
                {
                    "title": "Pyramid Diffusion Models For Low-light Image Enhancement",
                    "abstract": "Recovering noise-covered details from low-light images is challenging, and the results given by previous methods leave room for improvement. Recent diffusion models show realistic and detailed image generation through a sequence of denoising refinements and motivate us to introduce them to low-light image enhancement for recovering realistic details. However, we found two problems when doing this, i.e., 1) diffusion models keep constant resolution in one reverse process, which limits the speed; 2) diffusion models sometimes result in global degradation (e.g., RGB shift). To address the above problems, this paper proposes a Pyramid Diffusion model (PyDiff) for low-light image enhancement. PyDiff uses a novel pyramid diffusion method to perform sampling in a pyramid resolution style (i.e., progressively increasing resolution in one reverse process). Pyramid diffusion makes PyDiff much faster than vanilla diffusion models and introduces no performance degradation. Furthermore, PyDiff uses a global corrector to alleviate the global degradation that may occur in the reverse process, significantly improving the performance and making the training of diffusion models easier with little additional computational consumption. Extensive experiments on popular benchmarks show that PyDiff achieves superior performance and efficiency. Moreover, PyDiff can generalize well to unseen noise and illumination distributions. Code and supplementary materials are available at https://github.com/limuloo/PyDIff.git."
                },
                {
                    "title": "Refusion: Enabling Large-Size Realistic Image Restoration with Latent-Space Diffusion Models",
                    "abstract": "This work aims to improve the applicability of diffusion models in realistic image restoration. Specifically, we enhance the diffusion model in several aspects such as network architecture, noise level, denoising steps, training image size, and optimizer/scheduler. We show that tuning these hyperparameters allows us to achieve better performance on both distortion and perceptual scores. We also propose a U-Net based latent diffusion model which performs diffusion in a low-resolution latent space while preserving high-resolution information from the original input for the decoding process. Compared to the previous latent-diffusion model which trains a VAE-GAN to compress the image, our proposed U-Net compression strategy is significantly more stable and can recover highly accurate images without relying on adversarial optimization. Importantly, these modifications allow us to apply diffusion models to various image restoration tasks, including real-world shadow removal, HR non-homogeneous dehazing, stereo super-resolution, and bokeh effect transformation. By simply replacing the datasets and slightly changing the noise network, our model, named Refusion, is able to deal with large-size images (e.g., 6000 \u00d7 4000 \u00d7 3 in HR dehazing) and produces good results on all the above restoration problems. Our Refusion achieves the best perceptual performance in the NTIRE 2023 Image Shadow Removal Challenge and wins 2nd place overall."
                },
                {
                    "title": "Inversion by Direct Iteration: An Alternative to Denoising Diffusion for Image Restoration",
                    "abstract": "Inversion by Direct Iteration (InDI) is a new formulation for supervised image restoration that avoids the so-called\"regression to the mean\"effect and produces more realistic and detailed images than existing regression-based methods. It does this by gradually improving image quality in small steps, similar to generative denoising diffusion models. Image restoration is an ill-posed problem where multiple high-quality images are plausible reconstructions of a given low-quality input. Therefore, the outcome of a single step regression model is typically an aggregate of all possible explanations, therefore lacking details and realism. The main advantage of InDI is that it does not try to predict the clean target image in a single step but instead gradually improves the image in small steps, resulting in better perceptual quality. While generative denoising diffusion models also work in small steps, our formulation is distinct in that it does not require knowledge of any analytic form of the degradation process. Instead, we directly learn an iterative restoration process from low-quality and high-quality paired examples. InDI can be applied to virtually any image degradation, given paired training data. In conditional denoising diffusion image restoration the denoising network generates the restored image by repeatedly denoising an initial image of pure noise, conditioned on the degraded input. Contrary to conditional denoising formulations, InDI directly proceeds by iteratively restoring the input low-quality image, producing high-quality results on a variety of image restoration tasks, including motion and out-of-focus deblurring, super-resolution, compression artifact removal, and denoising."
                },
                {
                    "title": "I2SB: Image-to-Image Schr\u00f6dinger Bridge",
                    "abstract": "We propose Image-to-Image Schr\\\"odinger Bridge (I$^2$SB), a new class of conditional diffusion models that directly learn the nonlinear diffusion processes between two given distributions. These diffusion bridges are particularly useful for image restoration, as the degraded images are structurally informative priors for reconstructing the clean images. I$^2$SB belongs to a tractable class of Schr\\\"odinger bridge, the nonlinear extension to score-based models, whose marginal distributions can be computed analytically given boundary pairs. This results in a simulation-free framework for nonlinear diffusions, where the I$^2$SB training becomes scalable by adopting practical techniques used in standard diffusion models. We validate I$^2$SB in solving various image restoration tasks, including inpainting, super-resolution, deblurring, and JPEG restoration on ImageNet 256x256 and show that I$^2$SB surpasses standard conditional diffusion models with more interpretable generative processes. Moreover, I$^2$SB matches the performance of inverse methods that additionally require the knowledge of the corruption operators. Our work opens up new algorithmic opportunities for developing efficient nonlinear diffusion models on a large scale. scale. Project page and codes: https://i2sb.github.io/"
                },
                {
                    "title": "Image Restoration with Mean-Reverting Stochastic Differential Equations",
                    "abstract": "This paper presents a stochastic differential equation (SDE) approach for general-purpose image restoration. The key construction consists in a mean-reverting SDE that transforms a high-quality image into a degraded counterpart as a mean state with fixed Gaussian noise. Then, by simulating the corresponding reverse-time SDE, we are able to restore the origin of the low-quality image without relying on any task-specific prior knowledge. Crucially, the proposed mean-reverting SDE has a closed-form solution, allowing us to compute the ground truth time-dependent score and learn it with a neural network. Moreover, we propose a maximum likelihood objective to learn an optimal reverse trajectory that stabilizes the training and improves the restoration results. The experiments show that our proposed method achieves highly competitive performance in quantitative comparisons on image deraining, deblurring, and denoising, setting a new state-of-the-art on two deraining datasets. Finally, the general applicability of our approach is further demonstrated via qualitative results on image super-resolution, inpainting, and dehazing. Code is available at https://github.com/Algolzw/image-restoration-sde."
                },
                {
                    "title": "Ultra-High-Definition Low-Light Image Enhancement: A Benchmark and Transformer-Based Method",
                    "abstract": "As the quality of optical sensors improves, there is a need for processing large-scale images. In particular, the ability of devices to capture ultra-high definition (UHD) images and video places new demands on the image processing pipeline. In this paper, we consider the task of low-light image enhancement (LLIE) and introduce a large-scale database consisting of images at 4K and 8K resolution. We conduct systematic benchmarking studies and provide a comparison of current LLIE algorithms. As a second contribution, we introduce LLFormer, a transformer-based low-light enhancement method. The core components of LLFormer are the axis-based multi-head self-attention and cross-layer attention fusion block, which significantly reduces the linear complexity. Extensive experiments on the new dataset and existing public datasets show that LLFormer outperforms state-of-the-art methods. We also show that employing existing LLIE methods trained on our benchmark as a pre-processing step significantly improves the performance of downstream tasks, e.g., face detection in low-light conditions. The source code and pre-trained models are available at https://github.com/TaoWangzj/LLFormer."
                },
                {
                    "title": "ShadowDiffusion: When Degradation Prior Meets Diffusion Model for Shadow Removal",
                    "abstract": "Recent deep learning methods have achieved promising results in image shadow removal. However, their restored images still suffer from unsatisfactory boundary artifacts, due to the lack of degradation prior embedding and the deficiency in modeling capacity. Our work addresses these issues by proposing a unified diffusion framework that integrates both the image and degradation priors for highly effective shadow removal. In detail, we first propose a shadow degradation model, which inspires us to build a novel unrolling diffusion model, dubbed ShandowDiffusion. It remarkably improves the model's capacity in shadow removal via progressively refining the desired output with both degradation prior and diffusive generative prior, which by nature can serve as a new strong baseline for image restoration. Furthermore, ShadowDiffusion progressively refines the estimated shadow mask as an auxiliary task of the diffusion generator, which leads to more accurate and robust shadow-free image generation. We conduct extensive experiments on three popular public datasets, including ISTD, ISTD+, and SRD, to validate our method's effectiveness. Compared to the state-of-the-art methods, our model achieves a significant improvement in terms of PSNR, increasing from 31.69dB to 34. 73dB over SRD dataset. 11https://github.com/GuoLanqing/ShadowDiffusion"
                },
                {
                    "title": "Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow",
                    "abstract": "We present rectified flow, a surprisingly simple approach to learning (neural) ordinary differential equation (ODE) models to transport between two empirically observed distributions \\pi_0 and \\pi_1, hence providing a unified solution to generative modeling and domain transfer, among various other tasks involving distribution transport. The idea of rectified flow is to learn the ODE to follow the straight paths connecting the points drawn from \\pi_0 and \\pi_1 as much as possible. This is achieved by solving a straightforward nonlinear least squares optimization problem, which can be easily scaled to large models without introducing extra parameters beyond standard supervised learning. The straight paths are special and preferred because they are the shortest paths between two points, and can be simulated exactly without time discretization and hence yield computationally efficient models. We show that the procedure of learning a rectified flow from data, called rectification, turns an arbitrary coupling of \\pi_0 and \\pi_1 to a new deterministic coupling with provably non-increasing convex transport costs. In addition, recursively applying rectification allows us to obtain a sequence of flows with increasingly straight paths, which can be simulated accurately with coarse time discretization in the inference phase. In empirical studies, we show that rectified flow performs superbly on image generation, image-to-image translation, and domain adaptation. In particular, on image generation and translation, our method yields nearly straight flows that give high quality results even with a single Euler discretization step."
                },
                {
                    "title": "Understanding Diffusion Models: A Unified Perspective",
                    "abstract": "Diffusion models have shown incredible capabilities as generative models; indeed, they power the current state-of-the-art models on text-conditioned image generation such as Imagen and DALL-E 2. In this work we review, demystify, and unify the understanding of diffusion models across both variational and score-based perspectives. We first derive Variational Diffusion Models (VDM) as a special case of a Markovian Hierarchical Variational Autoencoder, where three key assumptions enable tractable computation and scalable optimization of the ELBO. We then prove that optimizing a VDM boils down to learning a neural network to predict one of three potential objectives: the original source input from any arbitrary noisification of it, the original source noise from any arbitrarily noisified input, or the score function of a noisified input at any arbitrary noise level. We then dive deeper into what it means to learn the score function, and connect the variational perspective of a diffusion model explicitly with the Score-based Generative Modeling perspective through Tweedie's Formula. Lastly, we cover how to learn a conditional distribution using diffusion models via guidance."
                },
                {
                    "title": "Cold Diffusion: Inverting Arbitrary Image Transforms Without Noise",
                    "abstract": "Standard diffusion models involve an image transform -- adding Gaussian noise -- and an image restoration operator that inverts this degradation. We observe that the generative behavior of diffusion models is not strongly dependent on the choice of image degradation, and in fact an entire family of generative models can be constructed by varying this choice. Even when using completely deterministic degradations (e.g., blur, masking, and more), the training and test-time update rules that underlie diffusion models can be easily generalized to create generative models. The success of these fully deterministic models calls into question the community's understanding of diffusion models, which relies on noise in either gradient Langevin dynamics or variational inference, and paves the way for generalized diffusion models that invert arbitrary processes. Our code is available at https://github.com/arpitbansal297/Cold-Diffusion-Models"
                },
                {
                    "title": "Restoring Vision in Adverse Weather Conditions With Patch-Based Denoising Diffusion Models",
                    "abstract": "Image restoration under adverse weather conditions has been of significant interest for various computer vision applications. Recent successful methods rely on the current progress in deep neural network architectural designs (e.g., with vision transformers). Motivated by the recent progress achieved with state-of-the-art conditional generative models, we present a novel patch-based image restoration algorithm based on denoising diffusion probabilistic models. Our patch-based diffusion modeling approach enables size-agnostic image restoration by using a guided denoising process with smoothed noise estimates across overlapping patches during inference. We empirically evaluate our model on benchmark datasets for image desnowing, combined deraining and dehazing, and raindrop removal. We demonstrate our approach to achieve state-of-the-art performances on both weather-specific and multi-weather image restoration, and experimentally show strong generalization to real-world test images."
                },
                {
                    "title": "Efficient Model-Driven Network for Shadow Removal",
                    "abstract": "Deep Convolutional Neural Networks (CNNs) based methods have achieved significant breakthroughs in the task of single image shadow removal. However, the performance of these methods remains limited for several reasons. First, the existing shadow illumination model ignores the spatially variant property of the shadow images, hindering their further performance. Second, most deep CNNs based methods directly estimate the shadow free results from the input shadow images like a black box, thus losing the desired interpretability. To address these issues, we first propose a new shadow illumination model for the shadow removal task. This new shadow illumination model ensures the identity mapping among unshaded regions, and adaptively performs fine grained spatial mapping between shadow regions and their references. Then, based on the shadow illumination model, we reformulate the shadow removal task as a variational optimization problem. To effectively solve the variational problem, we design an iterative algorithm and unfold it into a deep network, naturally increasing the interpretability of the deep model. Experiments show that our method could achieve SOTA performance with less than half parameters, one-fifth of floating-point of operations (FLOPs), and over seventeen times faster than SOTA method (DHAN)."
                },
                {
                    "title": "Image De-Raining Transformer",
                    "abstract": "Existing deep learning based de-raining approaches have resorted to the convolutional architectures. However, the intrinsic limitations of convolution, including local receptive fields and independence of input content, hinder the model's ability to capture long-range and complicated rainy artifacts. To overcome these limitations, we propose an effective and efficient transformer-based architecture for the image de-raining. First, we introduce general priors of vision tasks, i.e., locality and hierarchy, into the network architecture so that our model can achieve excellent de-raining performance without costly pre-training. Second, since the geometric appearance of rainy artifacts is complicated and of significant variance in space, it is essential for de-raining models to extract both local and non-local features. Therefore, we design the complementary window-based transformer and spatial transformer to enhance locality while capturing long-range dependencies. Besides, to compensate for the positional blindness of self-attention, we establish a separate representative space for modeling positional relationship, and design a new relative position enhanced multi-head self-attention. In this way, our model enjoys powerful abilities to capture dependencies from both content and position, so as to achieve better image content recovery while removing rainy artifacts. Experiments substantiate that our approach attains more appealing results than state-of-the-art methods quantitatively and qualitatively."
                },
                {
                    "title": "Bijective Mapping Network for Shadow Removal",
                    "abstract": "Shadow removal, which aims to restore the background in the shadow regions, is challenging due to its highly ill-posed nature. Most existing deep learning-based methods individually remove the shadow by only considering the content of the matched paired images, barely taking into account the auxiliary supervision of shadow generation in the shadow removal procedure. In this work, we argue that shadow removal and generation are interrelated and could provide useful informative supervision for each other. Specifically, we propose a new Bijective Mapping Network (BMNet), which couples the learning procedures of shadow removal and shadow generation in a unified parameter-shared framework. With consistent two way constraints and synchronous optimization of the two procedures, BMNet could effectively recover the underlying background contents during the forward shadow removal procedure. In addition, through statistical analysis of real world datasets, we observe and verify that shadow appearances under different color spectrums are inconsistent. This motivates us to design a Shadow-Invariant Color Guidance Module (SICGM), which can explicitly utilize the learned shadow-invariant color information to guide network color restoration, thereby further reducing color-bias effects. Experiments on the representative ISTD, ISTD+ and SRD benchmarks show that our proposed network outperforms the state-of-the-art method [11] in de-shadowing performance, while only using its 0.25% network parameters and 6.25% floating point operations (FLOPs)."
                },
                {
                    "title": "Denoising Diffusion Restoration Models",
                    "abstract": "Many interesting tasks in image restoration can be cast as linear inverse problems. A recent family of approaches for solving these problems uses stochastic algorithms that sample from the posterior distribution of natural images given the measurements. However, efficient solutions often require problem-specific supervised training to model the posterior, whereas unsupervised methods that are not problem-specific typically rely on inefficient iterative methods. This work addresses these issues by introducing Denoising Diffusion Restoration Models (DDRM), an efficient, unsupervised posterior sampling method. Motivated by variational inference, DDRM takes advantage of a pre-trained denoising diffusion generative model for solving any linear inverse problem. We demonstrate DDRM's versatility on several image datasets for super-resolution, deblurring, inpainting, and colorization under various amounts of measurement noise. DDRM outperforms the current leading unsupervised methods on the diverse ImageNet dataset in reconstruction quality, perceptual quality, and runtime, being 5x faster than the nearest competitor. DDRM also generalizes well for natural images out of the distribution of the observed ImageNet training set."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Deblurring via Stochastic Refinement",
                    "abstract": "Image deblurring is an ill-posed problem with multiple plausible solutions for a given input image. However, most existing methods produce a deterministic estimate of the clean image and are trained to minimize pixel-level distortion. These metrics are known to be poorly correlated with human perception, and often lead to unrealistic reconstructions. We present an alternative framework for blind deblurring based on conditional diffusion models. Unlike existing techniques, we train a stochastic sampler that refines the output of a deterministic predictor and is capable of producing a diverse set of plausible reconstructions for a given input. This leads to a significant improvement in perceptual quality over existing state-of-the-art methods across multiple standard benchmarks. Our predict-and-refine approach also enables much more efficient sampling compared to typical diffusion models. Combined with a carefully tuned network architecture and inference procedure, our method is competitive in terms of distortion metrics such as PSNR. These results show clear benefits of our diffusion-based method for deblurring and challenge the widely used strategy of producing a single, deterministic reconstruction."
                },
                {
                    "title": "Restormer: Efficient Transformer for High-Resolution Image Restoration",
                    "abstract": "Since convolutional neural networks (CNNs) perform well at learning generalizable image priors from large-scale data, these models have been extensively applied to image restoration and related tasks. Recently, another class of neural architectures, Transformers, have shown significant performance gains on natural language and high-level vision tasks. While the Transformer model mitigates the shortcomings of CNNs (i.e., limited receptive field and inadaptability to input content), its computational complexity grows quadratically with the spatial resolution, therefore making it infeasible to apply to most image restoration tasks involving high-resolution images. In this work, we propose an efficient Transformer model by making several key designs in the building blocks (multi-head attention and feed-forward network) such that it can capture long-range pixel interactions, while still remaining applicable to large images. Our model, named Restoration Transformer (Restormer), achieves state-of-the-art results on several image restoration tasks, including image deraining, single-image motion deblurring, defocus deblurring (single-image and dual-pixel data), and image denoising (Gaussian grayscale/color denoising, and real image denoising). The source code and pre-trained models are available at https://github.com/swz30/Restormer."
                },
                {
                    "title": "Low-Light Image Enhancement with Normalizing Flow",
                    "abstract": "To enhance low-light images to normally-exposed ones is highly ill-posed, namely that the mapping relationship between them is one-to-many. Previous works based on the pixel-wise reconstruction losses and deterministic processes fail to capture the complex conditional distribution of normally exposed images, which results in improper brightness, residual noise, and artifacts. In this paper, we investigate to model this one-to-many relationship via a proposed normalizing flow model. An invertible network that takes the low-light images/features as the condition and learns to map the distribution of normally exposed images into a Gaussian distribution. In this way, the conditional distribution of the normally exposed images can be well modeled, and the enhancement process, i.e., the other inference direction of the invertible network, is equivalent to being constrained by a loss function that better describes the manifold structure of natural images during the training. The experimental results on the existing benchmark datasets show our method achieves better quantitative and qualitative results, obtaining better-exposed illumination, less noise and artifact, and richer colors."
                },
                {
                    "title": "CANet: A Context-Aware Network for Shadow Removal",
                    "abstract": "In this paper, we propose a novel two-stage context-aware network named CANet for shadow removal, in which the contextual information from non-shadow regions is transferred to shadow regions at the embedded feature spaces. At Stage-I, we propose a contextual patch matching (CPM) module to generate a set of potential matching pairs of shadow and non-shadow patches. Combined with the potential contextual relationships between shadow and non-shadow regions, our well-designed contextual feature transfer (CFT) mechanism can transfer contextual information from non-shadow to shadow regions at different scales. With the reconstructed feature maps, we remove shadows at L and A/B channels separately. At Stage-II, we use an encoder-decoder to refine current results and generate the final shadow removal results. We evaluate our proposed CANet on two benchmark datasets and some real-world shadow images with complex scenes. Extensive experimental results strongly demonstrate the efficacy of our proposed CANet and exhibit superior performance to state-of-the-arts. Our source code is available at https://github.com/Zipei-Chen/CANet."
                },
                {
                    "title": "SNIPS: Solving Noisy Inverse Problems Stochastically",
                    "abstract": "In this work we introduce a novel stochastic algorithm dubbed SNIPS, which draws samples from the posterior distribution of any linear inverse problem, where the observation is assumed to be contaminated by additive white Gaussian noise. Our solution incorporates ideas from Langevin dynamics and Newton's method, and exploits a pre-trained minimum mean squared error (MMSE) Gaussian denoiser. The proposed approach relies on an intricate derivation of the posterior score function that includes a singular value decomposition (SVD) of the degradation operator, in order to obtain a tractable iterative algorithm for the desired sampling. Due to its stochasticity, the algorithm can produce multiple high perceptual quality samples for the same noisy observation. We demonstrate the abilities of the proposed paradigm for image deblurring, super-resolution, and compressive sensing. We show that the samples produced are sharp, detailed and consistent with the given measurements, and their diversity exposes the inherent uncertainty in the inverse problem being solved."
                },
                {
                    "title": "Image Super-Resolution via Iterative Refinement",
                    "abstract": "We present SR3, an approach to image Super-Resolution via Repeated Refinement. SR3 adapts denoising diffusion probabilistic models (Ho et al. 2020), (Sohl-Dickstein et al. 2015) to image-to-image translation, and performs super-resolution through a stochastic iterative denoising process. Output images are initialized with pure Gaussian noise and iteratively refined using a U-Net architecture that is trained on denoising at various noise levels, conditioned on a low-resolution input image. SR3 exhibits strong performance on super-resolution tasks at different magnification factors, on faces and natural images. We conduct human evaluation on a standard 8\u00d7 face super-resolution task on CelebA-HQ for which SR3 achieves a fool rate close to 50%, suggesting photo-realistic outputs, while GAN baselines do not exceed a fool rate of 34%. We evaluate SR3 on a 4\u00d7 super-resolution task on ImageNet, where SR3 outperforms baselines in human evaluation and classification accuracy of a ResNet-50 classifier trained on high-resolution images. We further show the effectiveness of SR3 in cascaded image generation, where a generative model is chained with super-resolution models to synthesize high-resolution images with competitive FID scores on the class-conditional 256\u00d7256 ImageNet generation challenge."
                },
                {
                    "title": "Auto-Exposure Fusion for Single-Image Shadow Removal",
                    "abstract": "Shadow removal is still a challenging task due to its inherent background-dependent1 and spatial-variant properties, leading to unknown and diverse shadow patterns. Even powerful deep neural networks could hardly recover traceless shadow-removed background. This paper proposes a new solution for this task by formulating it as an exposure fusion problem to address the challenges. Intuitively, we first estimate multiple over-exposure images w.r.t. the input image to let the shadow regions in these images have the same color with shadow-free areas in the input image. Then, we fuse the original input with the over-exposure images to generate the final shadow-free counterpart. Nevertheless, the spatial-variant property of the shadow requires the fusion to be sufficiently \u2018smart\u2019, that is, it should automatically select proper over-exposure pixels from different images to make the final output natural. To address this challenge, we propose the shadow-aware FusionNet that takes the shadow image as input to generate fusion weight maps across all the over-exposure images. Moreover, we propose the boundary-aware RefineNet to eliminate the remaining shadow trace further. We conduct extensive experiments on the ISTD, ISTD+, and SRD datasets to validate our method\u2019s effectiveness and show better performance in shadow regions and comparable performance in non-shadow regions over the state-of-the-art methods. We release the code in https://github.com/tsingqguo/exposure-fusion-shadow-removal."
                },
                {
                    "title": "Improved Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) are a class of generative models which have recently been shown to produce excellent samples. We show that with a few simple modifications, DDPMs can also achieve competitive log-likelihoods while maintaining high sample quality. Additionally, we find that learning variances of the reverse diffusion process allows sampling with an order of magnitude fewer forward passes with a negligible difference in sample quality, which is important for the practical deployment of these models. We additionally use precision and recall to compare how well DDPMs and GANs cover the target distribution. Finally, we show that the sample quality and likelihood of these models scale smoothly with model capacity and training compute, making them easily scalable. We release our code at https://github.com/openai/improved-diffusion"
                },
                {
                    "title": "Sparse Gradient Regularized Deep Retinex Network for Robust Low-Light Image Enhancement",
                    "abstract": "Due to the absence of a desirable objective for low-light image enhancement, previous data-driven methods may provide undesirable enhanced results including amplified noise, degraded contrast and biased colors. In this work, inspired by Retinex theory, we design an end-to-end signal prior-guided layer separation and data-driven mapping network with layer-specified constraints for single-image low-light enhancement. A Sparse Gradient Minimization sub-Network (SGM-Net) is constructed to remove the low-amplitude structures and preserve major edge information, which facilitates extracting paired illumination maps of low/normal-light images. After the learned decomposition, two sub-networks (Enhance-Net and Restore-Net) are utilized to predict the enhanced illumination and reflectance maps, respectively, which helps stretch the contrast of the illumination map and remove intensive noise in the reflectance map. The effects of all these configured constraints, including the signal structure regularization and losses, combine together reciprocally, which leads to good reconstruction results in overall visual quality. The evaluation on both synthetic and real images, particularly on those containing intensive noise, compression artifacts and their interleaved artifacts, shows the effectiveness of our novel models, which significantly outperforms the state-of-the-art methods."
                },
                {
                    "title": "Taming Transformers for High-Resolution Image Synthesis",
                    "abstract": "Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks. In contrast to CNNs, they contain no inductive bias that prioritizes local interactions. This makes them expressive, but also computationally infeasible for long sequences, such as high-resolution images. We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images. We show how to (i) use CNNs to learn a context-rich vocabulary of image constituents, and in turn (ii) utilize transformers to efficiently model their composition within high-resolution images. Our approach is readily applied to conditional synthesis tasks, where both non-spatial information, such as object classes, and spatial information, such as segmentations, can control the generated image. In particular, we present the first results on semantically-guided synthesis of megapixel images with transformers. Project page at https://git.io/JLlvY."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "All in One Bad Weather Removal Using Architectural Search",
                    "abstract": "Many methods have set state-of-the-art performance on restoring images degraded by bad weather such as rain, haze, fog, and snow, however they are designed specifically to handle one type of degradation. In this paper, we propose a method that can handle multiple bad weather degradations: rain, fog, snow and adherent raindrops using a single network. To achieve this, we first design a generator with multiple task-specific encoders, each of which is associated with a particular bad weather degradation type. We utilize a neural architecture search to optimally process the image features extracted from all encoders. Subsequently, to convert degraded image features to clean background features, we introduce a series of tensor-based operations encapsulating the underlying physics principles behind the formation of rain, fog, snow and adherent raindrops. These operations serve as the basic building blocks for our architectural search. Finally, our discriminator simultaneously assesses the correctness and classifies the degradation type of the restored image. We design a novel adversarial learning scheme that only backpropagates the loss of a degradation type to the respective task-specific encoder. Despite being designed to handle different types of bad weather, extensive experiments demonstrate that our method performs competitively to the individual and dedicated state-of-the-art image restoration methods."
                },
                {
                    "title": "Zero-Reference Deep Curve Estimation for Low-Light Image Enhancement",
                    "abstract": "The paper presents a novel method, Zero-Reference Deep Curve Estimation (Zero-DCE), which formulates light enhancement as a task of image-specific curve estimation with a deep network. Our method trains a lightweight deep network, DCE-Net, to estimate pixel-wise and high-order curves for dynamic range adjustment of a given image. The curve estimation is specially designed, considering pixel value range, monotonicity, and differentiability. Zero-DCE is appealing in its relaxed assumption on reference images, i.e., it does not require any paired or unpaired data during training. This is achieved through a set of carefully formulated non-reference loss functions, which implicitly measure the enhancement quality and drive the learning of the network. Our method is efficient as image enhancement can be achieved by an intuitive and simple nonlinear curve mapping. Despite its simplicity, we show that it generalizes well to diverse lighting conditions. Extensive experiments on various benchmarks demonstrate the advantages of our method over state-of-the-art methods qualitatively and quantitatively. Furthermore, the potential benefits of our Zero-DCE to face detection in the dark are discussed."
                },
                {
                    "title": "StarGAN v2: Diverse Image Synthesis for Multiple Domains",
                    "abstract": "A good image-to-image translation model should learn a mapping between different visual domains while satisfying the following properties: 1) diversity of generated images and 2) scalability over multiple domains. Existing methods address either of the issues, having limited diversity or multiple models for all domains. We propose StarGAN v2, a single framework that tackles both and shows significantly improved results over the baselines. Experiments on CelebA-HQ and a new animal faces dataset (AFHQ) validate our superiority in terms of visual quality, diversity, and scalability. To better assess image-to-image translation models, we release AFHQ, high-quality animal faces with large inter- and intra-domain differences. The code, pretrained models, and dataset are available at https://github.com/clovaai/stargan-v2."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Towards Ghost-free Shadow Removal via Dual Hierarchical Aggregation Network and Shadow Matting GAN",
                    "abstract": "Shadow removal is an essential task for scene understanding. Many studies consider only matching the image contents, which often causes two types of ghosts: color in-consistencies in shadow regions or artifacts on shadow boundaries (as shown in Figure. 1). In this paper, we tackle these issues in two ways. First, to carefully learn the border artifacts-free image, we propose a novel network structure named the dual hierarchically aggregation network (DHAN). It contains a series of growth dilated convolutions as the backbone without any down-samplings, and we hierarchically aggregate multi-context features for attention and prediction, respectively. Second, we argue that training on a limited dataset restricts the textural understanding of the network, which leads to the shadow region color in-consistencies. Currently, the largest dataset contains 2k+ shadow/shadow-free image pairs. However, it has only 0.1k+ unique scenes since many samples share exactly the same background with different shadow positions. Thus, we design a shadow matting generative adversarial network (SMGAN) to synthesize realistic shadow mattings from a given shadow mask and shadow-free image. With the help of novel masks or scenes, we enhance the current datasets using synthesized shadow images. Experiments show that our DHAN can erase the shadows and produce high-quality ghost-free images. After training on the synthesized and real datasets, our network outperforms other state-of-the-art methods by a large margin. The code is available: http://github.com/vinthony/ghost-free-shadow-removal/"
                },
                {
                    "title": "Deep Learning for Seeing Through Window With Raindrops",
                    "abstract": "When taking pictures through glass window in rainy day, the images are comprised and corrupted by the raindrops adhered to glass surfaces. It is a challenging problem to remove the effect of raindrops from an image. The key task is how to accurately and robustly identify the raindrop regions in an image. This paper develops a convolutional neural network (CNN) for removing the effect of raindrops from an image. In the proposed CNN, we introduce a double attention mechanism that concurrently guides the CNN using shape-driven attention and channel re-calibration. The shape-driven attention exploits physical shape priors of raindrops, i.e. convexness and contour closedness, to accurately locate raindrops, and the channel re-calibration improves the robustness when processing raindrops with varying appearances. The experimental results show that the proposed CNN outperforms the state-of-the-art approaches in terms of both quantitative metrics and visual quality."
                },
                {
                    "title": "Shadow Removal via Shadow Image Decomposition",
                    "abstract": "We propose a novel deep learning method for shadow removal. Inspired by physical models of shadow formation, we use a linear illumination transformation to model the shadow effects in the image that allows the shadow image to be expressed as a combination of the shadow-free image, the shadow parameters, and a matte layer. We use two deep networks, namely SP-Net and M-Net, to predict the shadow parameters and the shadow matte respectively. This system allows us to remove the shadow effects on the images. We train and test our framework on the most challenging shadow removal dataset (ISTD). Compared to the state-of-the-art method, our model achieves a 40% error reduction in terms of root mean square error (RMSE) for the shadow area, reducing RMSE from 13.3 to 7.9. Moreover, we create an augmented ISTD dataset based on an image decomposition system by modifying the shadow parameters to generate new synthetic shadow images. Training our model on this new augmented ISTD dataset further lowers the RMSE on the shadow area to 7.4."
                },
                {
                    "title": "ARGAN: Attentive Recurrent Generative Adversarial Network for Shadow Detection and Removal",
                    "abstract": "In this paper we propose an attentive recurrent generative adversarial network (ARGAN) to detect and remove shadows in an image. The generator consists of multiple progressive steps. At each step a shadow attention detector is firstly exploited to generate an attention map which specifies shadow regions in the input image. Given the attention map, a negative residual by a shadow remover encoder will recover a shadow-lighter or even a shadow-free image. The discriminator is designed to classify whether the output image in the last progressive step is real or fake. Moreover, ARGAN is suitable to be trained with a semi-supervised strategy to make full use of sufficient unsupervised data. The experiments on four public datasets have demonstrated that our ARGAN is robust to detect both simple and complex shadows and to produce more realistic shadow removal results. It outperforms the state-of-the-art methods, especially in detail of recovering shadow areas."
                },
                {
                    "title": "EnlightenGAN: Deep Light Enhancement Without Paired Supervision",
                    "abstract": "Deep learning-based methods have achieved remarkable success in image restoration and enhancement, but are they still competitive when there is a lack of paired training data? As one such example, this paper explores the low-light image enhancement problem, where in practice it is extremely challenging to simultaneously take a low-light and a normal-light photo of the same visual scene. We propose a highly effective unsupervised generative adversarial network, dubbed EnlightenGAN, that can be trained without low/normal-light image pairs, yet proves to generalize very well on various real-world test images. Instead of supervising the learning using ground truth data, we propose to regularize the unpaired training using the information extracted from the input itself, and benchmark a series of innovations for the low-light image enhancement problem, including a global-local discriminator structure, a self-regularized perceptual loss fusion, and the attention mechanism. Through extensive experiments, our proposed approach outperforms recent methods under a variety of metrics in terms of visual quality and subjective user study. Thanks to the great flexibility brought by unpaired training, EnlightenGAN is demonstrated to be easily adaptable to enhancing real-world images from various domains. Our codes and pre-trained models are available at: https://github.com/VITA-Group/EnlightenGAN."
                },
                {
                    "title": "Kindling the Darkness: A Practical Low-light Image Enhancer",
                    "abstract": "Images captured under low-light conditions often suffer from (partially) poor visibility. Besides unsatisfactory lightings, multiple types of degradations, such as noise and color distortion due to the limited quality of cameras, hide in the dark. In other words, solely turning up the brightness of dark regions will inevitably amplify hidden artifacts. This work builds a simple yet effective network for Kindling the Darkness (denoted as KinD), which, inspired by Retinex theory, decomposes images into two components. One component (illumination) is responsible for light adjustment, while the other (reflectance) for degradation removal. In such a way, the original space is decoupled into two smaller subspaces, expecting to be better regularized/learned. It is worth to note that our network is trained with paired images shot under different exposure conditions, instead of using any ground-truth reflectance and illumination information. Extensive experiments are conducted to demonstrate the efficacy of our design and its superiority over state-of-the-art alternatives. Our KinD is robust against severe visual defects, and user-friendly to arbitrarily adjust light levels. In addition, our model spends less than 50ms to process an image in VGA resolution on a 2080Ti GPU. All the above merits make our KinD attractive for practical use."
                },
                {
                    "title": "Dual Residual Networks Leveraging the Potential of Paired Operations for Image Restoration",
                    "abstract": "In this paper, we study design of deep neural networks for tasks of image restoration. We propose a novel style of residual connections dubbed \"dual residual connection\", which exploits the potential of paired operations, e.g., up- and down-sampling or convolution with large- and small-size kernels. We design a modular block implementing this connection style; it is equipped with two containers to which arbitrary paired operations are inserted. Adopting the \"unraveled\" view of the residual networks proposed by Veit et al., we point out that a stack of the proposed modular blocks allows the first operation in a block interact with the second operation in any subsequent blocks. Specifying the two operations in each of the stacked blocks, we build a complete network for each individual task of image restoration. We experimentally evaluate the proposed approach on five image restoration tasks using nine datasets. The results show that the proposed networks with properly chosen paired operations outperform previous methods on almost all of the tasks and datasets."
                },
                {
                    "title": "Deep Retinex Decomposition for Low-Light Enhancement",
                    "abstract": "Retinex model is an effective tool for low-light image enhancement. It assumes that observed images can be decomposed into the reflectance and illumination. Most existing Retinex-based methods have carefully designed hand-crafted constraints and parameters for this highly ill-posed decomposition, which may be limited by model capacity when applied in various scenes. In this paper, we collect a LOw-Light dataset (LOL) containing low/normal-light image pairs and propose a deep Retinex-Net learned on this dataset, including a Decom-Net for decomposition and an Enhance-Net for illumination adjustment. In the training process for Decom-Net, there is no ground truth of decomposed reflectance and illumination. The network is learned with only key constraints including the consistent reflectance shared by paired low/normal-light images, and the smoothness of illumination. Based on the decomposition, subsequent lightness enhancement is conducted on illumination by an enhancement network called Enhance-Net, and for joint denoising there is a denoising operation on reflectance. The Retinex-Net is end-to-end trainable, so that the learned decomposition is by nature good for lightness adjustment. Extensive experiments demonstrate that our method not only achieves visually pleasing quality for low-light enhancement but also provides a good representation of image decomposition."
                },
                {
                    "title": "Direction-Aware Spatial Context Features for Shadow Detection and Removal",
                    "abstract": "Shadow detection and shadow removal are fundamental and challenging tasks, requiring an understanding of the global image semantics. This paper presents a novel deep neural network design for shadow detection and removal by analyzing the spatial image context in a direction-aware manner. To achieve this, we first formulate the direction-aware attention mechanism in a spatial recurrent neural network (RNN) by introducing attention weights when aggregating spatial context features in the RNN. By learning these weights through training, we can recover direction-aware spatial context (DSC) for detecting and removing shadows. This design is developed into the DSC module and embedded in a convolutional neural network (CNN) to learn the DSC features at different levels. Moreover, we design a weighted cross entropy loss to make effective the training for shadow detection and further adopt the network for shadow removal by using a euclidean loss function and formulating a color transfer function to address the color and luminosity inconsistencies in the training pairs. We employed two shadow detection benchmark datasets and two shadow removal benchmark datasets, and performed various experiments to evaluate our method. Experimental results show that our method performs favorably against the state-of-the-art methods for both shadow detection and shadow removal."
                },
                {
                    "title": "The Unreasonable Effectiveness of Deep Features as a Perceptual Metric",
                    "abstract": "While it is nearly effortless for humans to quickly assess the perceptual similarity between two images, the underlying processes are thought to be quite complex. Despite this, the most widely used perceptual metrics today, such as PSNR and SSIM, are simple, shallow functions, and fail to account for many nuances of human perception. Recently, the deep learning community has found that features of the VGG network trained on ImageNet classification has been remarkably useful as a training loss for image synthesis. But how perceptual are these so-called \"perceptual losses\"? What elements are critical for their success? To answer these questions, we introduce a new dataset of human perceptual similarity judgments. We systematically evaluate deep features across different architectures and tasks and compare them with classic metrics. We find that deep features outperform all previous metrics by large margins on our dataset. More surprisingly, this result is not restricted to ImageNet-trained VGG features, but holds across different deep architectures and levels of supervision (supervised, self-supervised, or even unsupervised). Our results suggest that perceptual similarity is an emergent property shared across deep visual representations."
                },
                {
                    "title": "Stacked Conditional Generative Adversarial Networks for Jointly Learning Shadow Detection and Shadow Removal",
                    "abstract": "Understanding shadows from a single image consists of two types of task in previous studies, containing shadow detection and shadow removal. In this paper, we present a multi-task perspective, which is not embraced by any existing work, to jointly learn both detection and removal in an end-to-end fashion that aims at enjoying the mutually improved benefits from each other. Our framework is based on a novel STacked Conditional Generative Adversarial Network (ST-CGAN), which is composed of two stacked CGANs, each with a generator and a discriminator. Specifically, a shadow image is fed into the first generator which produces a shadow detection mask. That shadow image, concatenated with its predicted mask, goes through the second generator in order to recover its shadow-free image consequently. In addition, the two corresponding discriminators are very likely to model higher level relationships and global scene characteristics for the detected shadow region and reconstruction via removing shadows, respectively. More importantly, for multi-task learning, our design of stacked paradigm provides a novel view which is notably different from the commonly used one as the multi-branch version. To fully evaluate the performance of our proposed framework, we construct the first large-scale benchmark with 1870 image triplets (shadow image, shadow mask image, and shadow-free image) under 135 scenes. Extensive experimental results consistently show the advantages of STC-GAN over several representative state-of-the-art methods on two large-scale publicly available datasets and our newly released one."
                },
                {
                    "title": "Attentive Generative Adversarial Network for Raindrop Removal from A Single Image",
                    "abstract": "Raindrops adhered to a glass window or camera lens can severely hamper the visibility of a background scene and degrade an image considerably. In this paper, we address the problem by visually removing raindrops, and thus transforming a raindrop degraded image into a clean one. The problem is intractable, since first the regions occluded by raindrops are not given. Second, the information about the background scene of the occluded regions is completely lost for most part. To resolve the problem, we apply an attentive generative network using adversarial training. Our main idea is to inject visual attention into both the generative and discriminative networks. During the training, our visual attention learns about raindrop regions and their surroundings. Hence, by injecting this information, the generative network will pay more attention to the raindrop regions and the surrounding structures, and the discriminative network will be able to assess the local consistency of the restored regions. This injection of visual attention to both generative and discriminative networks is the main contribution of this paper. Our experiments show the effectiveness of our approach, which outperforms the state of the art methods quantitatively and qualitatively."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Neural Discrete Representation Learning",
                    "abstract": "Learning useful representations without supervision remains a key challenge in machine learning. In this paper, we propose a simple yet powerful generative model that learns such discrete representations. Our model, the Vector Quantised-Variational AutoEncoder (VQ-VAE), differs from VAEs in two key ways: the encoder network outputs discrete, rather than continuous, codes; and the prior is learnt rather than static. In order to learn a discrete latent representation, we incorporate ideas from vector quantisation (VQ). Using the VQ method allows the model to circumvent issues of \"posterior collapse\" -- where the latents are ignored when they are paired with a powerful autoregressive decoder -- typically observed in the VAE framework. Pairing these representations with an autoregressive prior, the model can generate high quality images, videos, and speech as well as doing high quality speaker conversion and unsupervised learning of phonemes, providing further evidence of the utility of the learnt representations."
                },
                {
                    "title": "Progressive Growing of GANs for Improved Quality, Stability, and Variation",
                    "abstract": "We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024^2. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10. Additionally, we describe several implementation details that are important for discouraging unhealthy competition between the generator and discriminator. Finally, we suggest a new metric for evaluating GAN results, both in terms of image quality and variation. As an additional contribution, we construct a higher-quality version of the CelebA dataset."
                },
                {
                    "title": "GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium",
                    "abstract": "Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the \"Frechet Inception Distance\" (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark."
                },
                {
                    "title": "Image-to-Image Translation with Conditional Adversarial Networks",
                    "abstract": "We investigate conditional adversarial networks as a general-purpose solution to image-to-image translation problems. These networks not only learn the mapping from input image to output image, but also learn a loss function to train this mapping. This makes it possible to apply the same generic approach to problems that traditionally would require very different loss formulations. We demonstrate that this approach is effective at synthesizing photos from label maps, reconstructing objects from edge maps, and colorizing images, among other tasks. Moreover, since the release of the pix2pix software associated with this paper, hundreds of twitter users have posted their own artistic experiments using our system. As a community, we no longer hand-engineer our mapping functions, and this work suggests we can achieve reasonable results without handengineering our loss functions either."
                },
                {
                    "title": "Deep Learning Face Attributes in the Wild",
                    "abstract": "Predicting face attributes in the wild is challenging due to complex face variations. We propose a novel deep learning framework for attribute prediction in the wild. It cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently. LNet is pre-trained by massive general object categories for face localization, while ANet is pre-trained by massive face identities for attribute prediction. This framework not only outperforms the state-of-the-art with a large margin, but also reveals valuable facts on learning face representation. (1) It shows how the performances of face localization (LNet) and attribute prediction (ANet) can be improved by different pre-training strategies. (2) It reveals that although the filters of LNet are fine-tuned only with image-level attribute tags, their response maps over entire images have strong indication of face locations. This fact enables training LNet for face localization with only image-level annotations, but without face bounding boxes or landmarks, which are required by all attribute recognition works. (3) It also demonstrates that the high-level hidden neurons of ANet automatically discover semantic concepts after pre-training with massive face identities, and such concepts are significantly enriched after fine-tuning with attribute tags. Each attribute can be well explained with a sparse linear combination of these concepts."
                },
                {
                    "title": "Image quality assessment: from error visibility to structural similarity",
                    "abstract": "Objective methods for assessing perceptual image quality traditionally attempted to quantify the visibility of errors (differences) between a distorted image and a reference image using a variety of known properties of the human visual system. Under the assumption that human visual perception is highly adapted for extracting structural information from a scene, we introduce an alternative complementary framework for quality assessment based on the degradation of structural information. As a specific example of this concept, we develop a structural similarity index and demonstrate its promise through a set of intuitive examples, as well as comparison to both subjective ratings and state-of-the-art objective methods on a database of images compressed with JPEG and JPEG2000. A MATLAB implementation of the proposed algorithm is available online at http://www.cns.nyu.edu//spl sim/lcv/ssim/."
                },
                {
                    "title": "A Decoupled Multi-Task Network for Shadow Removal",
                    "abstract": "Shadow removal, which aims to restore the illumination in shadow regions, is challenging due to the diversity of shadows in terms of location, intensity, shape, and size. Different from most multi-task methods, which design elaborate multi-branch or multi-stage structures for better shadow removal, we introduce feature decomposition to learn better feature representations. Specifically, we propose a single-stage and decoupled multi-task network (DMTN) to explicitly learn the decomposed features for shadow removal, shadow matte estimation, and shadow image reconstruction. First, we propose several coarse-to-fine semi-convolution (SMC) modules to capture features sufficient for joint learning of these three tasks. Second, we design a theoretically supported feature decoupling layer to explicitly decouple the learned features into shadow image features and shadow matte features via weight reassignment. Last, these features are converted to a target shadow-free image, affiliated shadow matte, and shadow image, supervised by multi-task joint loss functions. With multi-task collaboration, DMTN effectively recovers the illumination in shadow areas while ensuring the fidelity of non-shadow areas. Experimental results show that DMTN competes favorably with state-of-the-art multi-branch/multi-stage shadow removal methods, while maintaining the simplicity of single-stage methods."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we improve the efficiency and performance of denoising diffusion models for image restoration by integrating the residual term into the diffusion process?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the field of image restoration, as it addresses the inefficiencies of traditional diffusion models that start from Gaussian noise. By improving the performance of these models, we can enhance the quality of restored images, which has significant implications for various applications, including medical imaging, satellite imagery, and digital media. This research could lead to new methodologies that not only improve image restoration but also contribute to the broader understanding of generative models, potentially influencing future research directions in both theoretical and practical aspects of machine learning.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the need to effectively integrate the residual term with the noise term in the diffusion process. Naive approaches may fail because they do not account for the complex relationship between these terms, leading to poor generalization and interpretability. Additionally, designing a suitable noise schedule that maintains model performance while ensuring consistency between the forward and reverse processes is technically demanding. The intricacies of modeling the degradation process and ensuring that the model can accurately recover high-frequency details while preserving low-frequency information add further complexity.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on separate predictions for the residual and noise terms without explicitly defining their quantitative relationship, leading to limitations in performance. Existing models have also struggled with the inconsistency between their forward and reverse processes, which has hindered their generalization capabilities. Barriers such as the lack of a unified framework for integrating these components and the complexity of designing effective noise schedules have prevented this problem from being adequately addressed. Our approach differs by proposing a novel framework that explicitly incorporates the residual term into the diffusion process, thereby enhancing interpretability and performance.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology, named Resfusion, involves redefining the diffusion process to incorporate the residual term directly into the forward process. We will utilize a dataset of degraded images and employ metrics such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) to evaluate performance. The expected outcomes include improved image restoration quality, reduced sampling steps, and enhanced model interpretability, demonstrating the effectiveness of the resnoise diffusion process in generating high-fidelity images from"
            }
        },
        "author_data": {
            "038264d0-68e4-4071-9cba-5f84da241e6b": {
                "pk": "038264d0-68e4-4071-9cba-5f84da241e6b",
                "name": "Zhenning Shi",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "1cc837b6-394f-4de9-b97d-f20490bc77c9": {
                "pk": "1cc837b6-394f-4de9-b97d-f20490bc77c9",
                "name": "Haoshuai Zheng",
                "collaborators": [
                    "Cheng Gong",
                    "Mengting Hu",
                    "Zheng Lin",
                    "Deng-Ping Fan",
                    "Yuzhi Zhang",
                    "Tao Li"
                ],
                "domain": [
                    "Quantization",
                    "Deep Learning",
                    "Neural Networks",
                    "Model Optimization"
                ],
                "publications": [
                    {
                        "title": "Minimize Quantization Output Error with Bias Compensation",
                        "abstract": "Quantization is a promising method that reduces memory usage and computational intensity of Deep Neural Networks (DNNs), but it often leads to significant output error that hinder model deployment. In this paper, we propose Bias Compensation (BC) to minimize the output error, thus realizing ultra-low-precision quantization without model fine-tuning. Instead of optimizing the non-convex quantization process as in most previous methods, the proposed BC bypasses the step to directly minimize the quantizing output error by identifying a bias vector for compensation. We have established that the minimization of output error through BC is a convex problem and provides an efficient strategy to procure optimal solutions associated with minimal output error,without the need for training or fine-tuning. We conduct extensive experiments on Vision Transformer models and Large Language Models, and the results show that our method notably reduces quantization output error, thereby permitting ultra-low-precision post-training quantization and enhancing the task performance of models. Especially, BC improves the accuracy of ViT-B with 4-bit PTQ4ViT by 36.89% on the ImageNet-1k task, and decreases the perplexity of OPT-350M with 3-bit GPTQ by 5.97 on WikiText2.The code is in https://github.com/GongCheng1919/bias-compensation."
                    }
                ]
            },
            "3109f8b2-8ba7-4856-853b-fedd8209531a": {
                "pk": "3109f8b2-8ba7-4856-853b-fedd8209531a",
                "name": "Chen Xu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "1857c8be-92cd-45a0-b93e-41268d2c3765": {
                "pk": "1857c8be-92cd-45a0-b93e-41268d2c3765",
                "name": "Changsheng Dong",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "637b07b3-c716-4158-9e34-8ee999a4a932": {
                "pk": "637b07b3-c716-4158-9e34-8ee999a4a932",
                "name": "Bin Pan",
                "collaborators": [],
                "domain": [
                    "Vehicular Networks",
                    "Video Transmission",
                    "Cooperative Communication",
                    "Quality of Experience"
                ],
                "publications": [
                    {
                        "title": "Video Streaming in Cooperative Vehicular Networks",
                        "abstract": "Video services in vehicular networks play a significant role in our daily traveling. In this paper, we propose a cooperative communication scheme to facilitate video data transmission, utilizing the mobility of vehicles and the cooperation among infrastructure and vehicles. To improve the video quality of experience (QoE), i.e., reduce the interruption ratio, quality variation and improve the playback quality, we design a Back Compensation (BC) video transmission strategy with the knowledge of vehicle status information. In addition, we analyze the throughput with one-hop and target-cluster-based cooperation schemes and obtain their closed-form expressions, respectively, which is useful for video encoding design in the central server. Simulation results demonstrate that the proposed approach can improve the video performance significantly and verify the accuracy of our analytical results."
                    }
                ]
            },
            "d3a8ad96-d545-41f0-bf16-b3a5873c9a57": {
                "pk": "d3a8ad96-d545-41f0-bf16-b3a5873c9a57",
                "name": "Xueshuo Xie",
                "collaborators": [
                    "Tao Li",
                    "Yanfeng Jiang",
                    "Ning Sun",
                    "Fei Yang",
                    "Hang Li",
                    "Qiankun Dong",
                    "Xia Xu",
                    "Zhenwei Shi",
                    "Haoxu Wang",
                    "Zhaolong Jian"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Image Processing",
                    "Anomaly Detection"
                ],
                "publications": [
                    {
                        "title": "ADFQ-ViT: Activation-Distribution-Friendly Post-Training Quantization for Vision Transformers",
                        "abstract": "Vision Transformers (ViTs) have exhibited exceptional performance across diverse computer vision tasks, while their substantial parameter size incurs significantly increased memory and computational demands, impeding effective inference on resource-constrained devices. Quantization has emerged as a promising solution to mitigate these challenges, yet existing methods still suffer from significant accuracy loss at low-bit. We attribute this issue to the distinctive distributions of post-LayerNorm and post-GELU activations within ViTs, rendering conventional hardware-friendly quantizers ineffective, particularly in low-bit scenarios. To address this issue, we propose a novel framework called Activation-Distribution-Friendly post-training Quantization for Vision Transformers, ADFQ-ViT. Concretely, we introduce the Per-Patch Outlier-aware Quantizer to tackle irregular outliers in post-LayerNorm activations. This quantizer refines the granularity of the uniform quantizer to a per-patch level while retaining a minimal subset of values exceeding a threshold at full-precision. To handle the non-uniform distributions of post-GELU activations between positive and negative regions, we design the Shift-Log2 Quantizer, which shifts all elements to the positive region and then applies log2 quantization. Moreover, we present the Attention-score enhanced Module-wise Optimization which adjusts the parameters of each quantizer by reconstructing errors to further mitigate quantization error. Extensive experiments demonstrate ADFQ-ViT provides significant improvements over various baselines in image classification, object detection, and instance segmentation tasks at 4-bit. Specifically, when quantizing the ViT-B model to 4-bit, we achieve a 10.23% improvement in Top-1 accuracy on the ImageNet dataset."
                    },
                    {
                        "title": "Transformer for Multitemporal Hyperspectral Image Unmixing",
                        "abstract": "Multitemporal hyperspectral image unmixing (MTHU) holds significant importance in monitoring and analyzing the dynamic changes of surface. However, compared to single-temporal unmixing, the multitemporal approach demands comprehensive consideration of information across different phases, rendering it a greater challenge. To address this challenge, we propose the Multitemporal Hyperspectral Image Unmixing Transformer (MUFormer), an end-to-end unsupervised deep learning model. To effectively perform multitemporal hyperspectral image unmixing, we introduce two key modules: the Global Awareness Module (GAM) and the Change Enhancement Module (CEM). The Global Awareness Module computes self-attention across all phases, facilitating global weight allocation. On the other hand, the Change Enhancement Module dynamically learns local temporal changes by comparing endmember changes between adjacent phases. The synergy between these modules allows for capturing semantic information regarding endmember and abundance changes, thereby enhancing the effectiveness of multitemporal hyperspectral image unmixing. We conducted experiments on one real dataset and two synthetic datasets, demonstrating that our model significantly enhances the effect of multitemporal hyperspectral image unmixing."
                    },
                    {
                        "title": "Memory-Efficient and Secure DNN Inference on TrustZone-enabled Consumer IoT Devices",
                        "abstract": "Edge intelligence enables resource-demanding Deep Neural Network (DNN) inference without transferring original data, addressing concerns about data privacy in consumer Internet of Things (IoT) devices. For privacy-sensitive applications, deploying models in hardware-isolated trusted execution environments (TEEs) becomes essential. However, the limited secure memory in TEEs poses challenges for deploying DNN inference, and alternative techniques like model partitioning and offloading introduce performance degradation and security issues. In this paper, we present a novel approach for advanced model deployment in TrustZone that ensures comprehensive privacy preservation during model inference. We design a memory-efficient management method to support memory-demanding inference in TEEs. By adjusting the memory priority, we effectively mitigate memory leakage risks and memory overlap conflicts, resulting in 32 lines of code alterations in the trusted operating system. Additionally, we leverage two tiny libraries: S-Tinylib (2,538 LoCs), a tiny deep learning library, and Tinylibm (827 LoCs), a tiny math library, to support efficient inference in TEEs. We implemented a prototype on Raspberry Pi 3B+ and evaluated it using three well-known lightweight DNN models. The experimental results demonstrate that our design significantly improves inference speed by 3.13 times and reduces power consumption by over 66.5% compared to non-memory optimization method in TEEs."
                    },
                    {
                        "title": "Spatial-Frequency Dual Progressive Attention Network For Medical Image Segmentation",
                        "abstract": "In medical images, various types of lesions often manifest significant differences in their shape and texture. Accurate medical image segmentation demands deep learning models with robust capabilities in multi-scale and boundary feature learning. However, previous networks still have limitations in addressing the above issues. Firstly, previous networks simultaneously fuse multi-level features or employ deep supervision to enhance multi-scale learning. However, this may lead to feature redundancy and excessive computational overhead, which is not conducive to network training and clinical deployment. Secondly, the majority of medical image segmentation networks exclusively learn features in the spatial domain, disregarding the abundant global information in the frequency domain. This results in a bias towards low-frequency components, neglecting crucial high-frequency information. To address these problems, we introduce SF-UNet, a spatial-frequency dual-domain attention network. It comprises two main components: the Multi-scale Progressive Channel Attention (MPCA) block, which progressively extract multi-scale features across adjacent encoder layers, and the lightweight Frequency-Spatial Attention (FSA) block, with only 0.05M parameters, enabling concurrent learning of texture and boundary features from both spatial and frequency domains. We validate the effectiveness of the proposed SF-UNet on three public datasets. Experimental results show that compared to previous state-of-the-art (SOTA) medical image segmentation networks, SF-UNet achieves the best performance, and achieves up to 9.4\\% and 10.78\\% improvement in DSC and IOU. Codes will be released at https://github.com/nkicsl/SF-UNet."
                    },
                    {
                        "title": "A Pvalue-guided Anomaly Detection Approach Combining Multiple Heterogeneous Log Parser Algorithms on IIoT Systems",
                        "abstract": "Industrial Internet of Things (IIoT) is becoming an attack target of advanced persistent threat (APT). Currently, IIoT logs have not been effectively used for anomaly detection. In this paper, we use blockchain to prevent logs from being tampered with and propose a pvalue-guided anomaly detection approach. This approach uses statistical pvalues to combine multiple heterogeneous log parser algorithms. The weighted edit distance is selected as a score function to calculate the nonconformity score between a log and a predefined event. The pvalue is calculated based on the non-conformity scores which indicate how well a log matches an event. This approach is tested on a large number of real-world HDFS logs and IIoT logs. The experiment results show that abnormal events could be effectively recognized by our pvalue-guided approach."
                    }
                ]
            },
            "4c5ddce4-201f-473d-ba49-6a0eec4be938": {
                "pk": "4c5ddce4-201f-473d-ba49-6a0eec4be938",
                "name": "Along He",
                "collaborators": [
                    "Tao Li",
                    "Yanlin Wu",
                    "Huazhu Fu",
                    "Zhihong Wang",
                    "Hong Kang",
                    "Kai Wang",
                    "Ke Zou",
                    "Wang Bo",
                    "Zhenhuan Zhou",
                    "Rui Yao"
                ],
                "domain": [
                    "Medical Image Segmentation",
                    "Deep Learning",
                    "Semi-Supervised Learning",
                    "Transformer"
                ],
                "publications": [
                    {
                        "title": "TransUKAN:Computing-Efficient Hybrid KAN-Transformer for Enhanced Medical Image Segmentation",
                        "abstract": "U-Net is currently the most widely used architecture for medical image segmentation. Benefiting from its unique encoder-decoder architecture and skip connections, it can effectively extract features from input images to segment target regions. The commonly used U-Net is typically based on convolutional operations or Transformers, modeling the dependencies between local or global information to accomplish medical image analysis tasks. However, convolutional layers, fully connected layers, and attention mechanisms used in this process introduce a significant number of parameters, often requiring the stacking of network layers to model complex nonlinear relationships, which can impact the training process. To address these issues, we propose TransUKAN. Specifically, we have improved the KAN to reduce memory usage and computational load. On this basis, we explored an effective combination of KAN, Transformer, and U-Net structures. This approach enhances the model's capability to capture nonlinear relationships by introducing only a small number of additional parameters and compensates for the Transformer structure's deficiency in local information extraction. We validated TransUKAN on multiple medical image segmentation tasks. Experimental results demonstrate that TransUKAN achieves excellent performance with significantly reduced parameters. The code will be available athttps://github.com/wuyanlin-wyl/TransUKAN."
                    },
                    {
                        "title": "FRCNet Frequency and Region Consistency for Semi-supervised Medical Image Segmentation",
                        "abstract": "Limited labeled data hinder the application of deep learning in medical domain. In clinical practice, there are sufficient unlabeled data that are not effectively used, and semi-supervised learning (SSL) is a promising way for leveraging these unlabeled data. However, existing SSL methods ignore frequency domain and region-level information and it is important for lesion regions located at low frequencies and with significant scale changes. In this paper, we introduce two consistency regularization strategies for semi-supervised medical image segmentation, including frequency domain consistency (FDC) to assist the feature learning in frequency domain and multi-granularity region similarity consistency (MRSC) to perform multi-scale region-level local context information feature learning. With the help of the proposed FDC and MRSC, we can leverage the powerful feature representation capability of them in an effective and efficient way. We perform comprehensive experiments on two datasets, and the results show that our method achieves large performance gains and exceeds other state-of-the-art methods."
                    },
                    {
                        "title": "Progressive Multi-scale Consistent Network for Multi-class Fundus Lesion Segmentation",
                        "abstract": "Effectively integrating multi-scale information is of considerable significance for the challenging multi-class segmentation of fundus lesions because different lesions vary significantly in scales and shapes. Several methods have been proposed to successfully handle the multi-scale object segmentation. However, two issues are not considered in previous studies. The first is the lack of interaction between adjacent feature levels, and this will lead to the deviation of high-level features from low-level features and the loss of detailed cues. The second is the conflict between the low-level and high-level features, this occurs because they learn different scales of features, thereby confusing the model and decreasing the accuracy of the final prediction. In this paper, we propose a progressive multi-scale consistent network (PMCNet) that integrates the proposed progressive feature fusion (PFF) block and dynamic attention block (DAB) to address the aforementioned issues. Specifically, PFF block progressively integrates multi-scale features from adjacent encoding layers, facilitating feature learning of each layer by aggregating fine-grained details and high-level semantics. As features at different scales should be consistent, DAB is designed to dynamically learn the attentive cues from the fused features at different scales, thus aiming to smooth the essential conflicts existing in multi-scale features. The two proposed PFF and DAB blocks can be integrated with the off-the-shelf backbone networks to address the two issues of multi-scale and feature inconsistency in the multi-class segmentation of fundus lesions, which will produce better feature representation in the feature space. Experimental results on three public datasets indicate that the proposed method is more effective than recent state-of-the-art methods."
                    },
                    {
                        "title": "DVPT: Dynamic Visual Prompt Tuning of Large Pre-trained Models for Medical Image Analysis",
                        "abstract": "Limited labeled data makes it hard to train models from scratch in medical domain, and an important paradigm is pre-training and then fine-tuning. Large pre-trained models contain rich representations, which can be adapted to downstream medical tasks. However, existing methods either tune all the parameters or the task-specific layers of the pre-trained models, ignoring the input variations of medical images, and thus they are not efficient or effective. In this work, we aim to study parameter-efficient fine-tuning (PEFT) for medical image analysis, and propose a dynamic visual prompt tuning method, named DVPT. It can extract knowledge beneficial to downstream tasks from large models with a few trainable parameters. Firstly, the frozen features are transformed by an lightweight bottleneck layer to learn the domain-specific distribution of downstream medical tasks, and then a few learnable visual prompts are used as dynamic queries and then conduct cross-attention with the transformed features, attempting to acquire sample-specific knowledge that are suitable for each sample. Finally, the features are projected to original feature dimension and aggregated with the frozen features. This DVPT module can be shared between different Transformer layers, further reducing the trainable parameters. To validate DVPT, we conduct extensive experiments with different pre-trained models on medical classification and segmentation tasks. We find such PEFT method can not only efficiently adapt the pre-trained models to the medical domain, but also brings data efficiency with partial labeled data. For example, with 0.5\\% extra trainable parameters, our method not only outperforms state-of-the-art PEFT methods, even surpasses the full fine-tuning by more than 2.20\\% Kappa score on medical classification task. It can saves up to 60\\% labeled data and 99\\% storage cost of ViT-B/16."
                    },
                    {
                        "title": "Spatial-Frequency Dual Progressive Attention Network For Medical Image Segmentation",
                        "abstract": "In medical images, various types of lesions often manifest significant differences in their shape and texture. Accurate medical image segmentation demands deep learning models with robust capabilities in multi-scale and boundary feature learning. However, previous networks still have limitations in addressing the above issues. Firstly, previous networks simultaneously fuse multi-level features or employ deep supervision to enhance multi-scale learning. However, this may lead to feature redundancy and excessive computational overhead, which is not conducive to network training and clinical deployment. Secondly, the majority of medical image segmentation networks exclusively learn features in the spatial domain, disregarding the abundant global information in the frequency domain. This results in a bias towards low-frequency components, neglecting crucial high-frequency information. To address these problems, we introduce SF-UNet, a spatial-frequency dual-domain attention network. It comprises two main components: the Multi-scale Progressive Channel Attention (MPCA) block, which progressively extract multi-scale features across adjacent encoder layers, and the lightweight Frequency-Spatial Attention (FSA) block, with only 0.05M parameters, enabling concurrent learning of texture and boundary features from both spatial and frequency domains. We validate the effectiveness of the proposed SF-UNet on three public datasets. Experimental results show that compared to previous state-of-the-art (SOTA) medical image segmentation networks, SF-UNet achieves the best performance, and achieves up to 9.4\\% and 10.78\\% improvement in DSC and IOU. Codes will be released at https://github.com/nkicsl/SF-UNet."
                    }
                ]
            },
            "4c39f2f6-71ff-4f5a-aa22-082c53aa0eaf": {
                "pk": "4c39f2f6-71ff-4f5a-aa22-082c53aa0eaf",
                "name": "Tao Li",
                "collaborators": [],
                "domain": [
                    "Topology",
                    "3-Manifolds",
                    "Superconductivity",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "Small 3-manifolds with large Heegaard distance",
                        "abstract": "We construct examples of closed non-Haken hyperbolic 3-manifolds with a Heegaard splitting of arbitrarily large distance."
                    },
                    {
                        "title": "An algorithm to find vertical tori in small Seifert fiber spaces",
                        "abstract": "We give an algorithm to find vertical essential tori in small Seifert fiber spaces with infinite fundamental groups. This implies that there are algorithms to decide whether a 3-manifold is a Seifert fiber space."
                    },
                    {
                        "title": "The Driving Force of Superconducting Transition in High Temperature Superconductors",
                        "abstract": "We show that both the kinetic energy and the exchange energy of the t-J model can be read off from the optical data. We show that the optical data indicates that the superconducting transition in high temperature superconductors is kinetic energy driven and the exchange energy resist the transition. We also show that kinetic energy may also be the driving force of the pseudogap phenomenon."
                    },
                    {
                        "title": "Thin position and planar surfaces for graphs in the 3-sphere",
                        "abstract": "We show that given a trivalent graph in $S^3$, either the graph complement contains an essential almost meridional planar surface or thin position for the graph is also bridge position. This can be viewed as an extension of a theorem of Thompson to graphs. It follows that any graph complement always contains a useful planar surface."
                    },
                    {
                        "title": "An algorithm to determine the Heegaard genus of a 3-manifold",
                        "abstract": "We give an algorithmic proof of the theorem that a closed orientable irreducible and atoroidal 3-manifold has only finitely many Heegaard splittings in each genus, up to isotopy. The proof gives an algorithm to determine the Heegaard genus of an atoroidal 3-manifold."
                    },
                    {
                        "title": "Rank and genus of 3-manifolds",
                        "abstract": "We construct a counterexample to the Rank versus Genus Conjecture, i.e. a closed orientable hyperbolic 3-manifold with rank of its fundamental group smaller than its Heegaard genus. Moreover, we show that the discrepancy between rank and Heegaard genus can be arbitrarily large for hyperbolic 3-manifolds. We also construct toroidal such examples containing hyperbolic JSJ pieces."
                    },
                    {
                        "title": "Taut foliations of 3-manifolds with Heegaard genus two",
                        "abstract": "Let $M$ be a closed, orientable, and irreducible 3-manifold with Heegaard genus two. We prove that if the fundamental group of $M$ is left-orderable then $M$ admits a co-orientable taut foliation."
                    },
                    {
                        "title": "The Zinc Doping Induced STM Resonance as a Zero Mode",
                        "abstract": "We show that the Zinc doping induced STM resonance in high-Tc cuprates is a zero mode - a d-wave monomer confined in the opposite sublattice of the Zinc site. We propose that the resonance come from the superconducting peak around $(\\pi,0) $ as observed in ARPES experiments. It is predicted that the resonance should vanish in single layer $BSCCO_{2201}$ or in the vortex core of underdoped cuprate where the ARPES peak is absent. We argue that the STM result imply the ARPES peak around $(\\pi,0) $ is the only coherent feature in the whole Brillouin zone."
                    },
                    {
                        "title": "Boundary curves of surfaces with the 4-plane property",
                        "abstract": "Let M be an orientable and irreducible 3-manifold whose boundary is an incompressible torus. Suppose that M does not contain any closed nonperipheral embedded incompressible surfaces. We will show in this paper that the immersed surfaces in M with the 4-plane property can realize only finitely many boundary slopes. Moreover, we will show that only finitely many Dehn fillings of M can yield 3-manifolds with nonpositive cubings. This gives the first examples of hyperbolic 3-manifolds that cannot admit any nonpositive cubings."
                    },
                    {
                        "title": "Heegaard surfaces and measured laminations, I: the Waldhausen conjecture",
                        "abstract": "We give a proof of the so-called generalized Waldhausen conjecture, which says that an orientable irreducible atoroidal 3-manifold has only finitely many Heegaard splittings in each genus, up to isotopy. Jaco and Rubinstein have announced a proof of this conjecture using different methods."
                    },
                    {
                        "title": "Heegaard surfaces and measured laminations, II: non-Haken 3-manifolds",
                        "abstract": "A famous example of Casson and Gordon shows that a Haken 3-manifold can have an infinite family of irreducible Heegaard splittings with different genera. In this paper, we prove that a closed non-Haken 3-manifold has only finitely many irreducible Heegaard splittings, up to isotopy. This is much stronger than the Waldhausen conjecture. Another immediate corollary is that for any irreducible non-Haken 3-manifold M, there is a number N, such that any two Heegaard splittings of M are equivalent after at most N stabilizations."
                    },
                    {
                        "title": "On the Heegaard splittings of amalgamated 3-manifolds",
                        "abstract": "We give a combinatorial proof of a theorem first proved by Souto which says the following. Let M_1 and M_2 be simple 3-manifolds with connected boundary of genus g>0. If M_1 and M_2 are glued via a complicated map, then every minimal Heegaard splitting of the resulting closed 3-manifold is an amalgamation. This proof also provides an algorithm to find a bound on the complexity of the gluing map."
                    },
                    {
                        "title": "Saddle tangencies and the distance of Heegaard splittings",
                        "abstract": "We give another proof of a theorem of Scharlemann and Tomova and of a theorem of Hartshorn. The two theorems together say the following. Let M be a compact orientable irreducible 3--manifold and P a Heegaard surface of M. Suppose Q is either an incompressible surface or a strongly irreducible Heegaard surface in M. Then either the Hempel distance d(P) <= 2 genus(Q) or P is isotopic to Q. This theorem can be naturally extended to bicompressible but weakly incompressible surfaces."
                    },
                    {
                        "title": "Orbital Degeneracy and the Microscopic Model of the FeAs Plane in the Iron-Based Superconductors",
                        "abstract": "A microscopic model for the FeAs plane in the newly discovered Iron-based superconductors is proposed with the emphasis on the role of the orbital degeneracy between the Fe $3d_{xz}$ and $3d_{yz}$ orbital in the crystal field environment. The model predicts commensurate antiferromagnetic ordered ground state for the parent compounds and a d-wave superconducting ground state at finite doping. Inter-orbital couplings plays an interesting role in enhancing both orders. Correlated variational wave functions are proposed for both ordered phases."
                    },
                    {
                        "title": "Heegaard surfaces and the distance of amalgamation",
                        "abstract": "Let $M_1$ and $M_2$ be orientable irreducible 3--manifolds with connected boundary and suppose $\\partial M_1\\cong\\partial M_2$. Let $M$ be a closed 3--manifold obtained by gluing $M_1$ to $M_2$ along the boundary. We show that if the gluing homeomorphism is sufficiently complicated, then $M$ is not homeomorphic to $S^3$ and all small-genus Heegaard splittings of $M$ are standard in a certain sense. In particular, $g(M)=g(M_1)+g(M_2)-g(\\partial M_i)$, where $g(M)$ denotes the Heegaard genus of $M$. This theorem is also true for certain manifolds with multiple boundary components."
                    },
                    {
                        "title": "Absence of topological degeneracy in the Hubbard model on honeycomb lattice",
                        "abstract": "It is shown that the unique sign structure of the ground state of the Hubbard model on honeycomb lattice, which is shown to be insensitive to the trapped $Z_{2}$ gauge flux when the system is defined on a torus, may cause the absence of topological degeneracy on this bipartite system. Examples of variational Mott insulating state on the honeycomb lattice are given to illustrate the close relation between the sign structure of the ground state and the (absence of) topological degeneracy."
                    },
                    {
                        "title": "Antiferromagnetic long range order in the uniform resonating valence bond state on square lattice",
                        "abstract": "With extensive variational Monte Carlo simulation, we show that the uniform resonating valence bond state(U-RVB) on square lattice is actually antiferromagnetic long range ordered. The ordered moment is estimated to be $m\\approx 0.17$. Finite size scaling analysis on lattice up to lattice size of $50\\times 50$ shows that the spin structure factor at the antiferromagnetic ordering wave vector follows perfectly the $\\mathrm{S}_{\\mathrm{q}=(\\pi,\\pi)} \\simeq S_{0}+\\alpha (1/L)^{5/4}$ behavior, where $L$ is linear scale of the lattice. Such a behavior is quite unexpected from the slave Boson mean field treatment or the Gutzwiller approximation of the uniform RVB state."
                    },
                    {
                        "title": "Beauty Learning and Counterfactual Inference",
                        "abstract": "This work showcases a new approach for causal discovery by leveraging user experiments and recent advances in photo-realistic image editing, demonstrating a potential of identifying causal factors and understanding complex systems counterfactually. We introduce the beauty learning problem as an example, which has been discussed metaphysically for centuries and been proved exists, is quantifiable, and can be learned by deep models in our recent paper, where we utilize a natural image generator coupled with user studies to infer causal effects from facial semantics to beauty outcomes, the results of which also align with existing empirical studies. We expect the proposed framework for a broader application in causal inference."
                    },
                    {
                        "title": "Theory of Sinusoidal Modulation of the Resonant Neutron Scattering in High Temperature Superconductors",
                        "abstract": "A model with interlayer pairing is proposed to explain the sinusoidal modulation of the resonant neutron scattering in high temperature superconductors. It is found that the interlayer pairing has s-wave symmetry in the CuO$_2$ plane and has comparable magnitude with the d-wave intralayer pairing. It is also found that the interlayer pairing mainly affects momentum close to the hot spots on the Fermi surface while its effect on the gap nodes is negligible. It is pointed out that these characteristics of the interlayer pairing can be understood in a model in which the superconducting pairing originates from the exchange of the antiferromagnetic spin fluctuation."
                    },
                    {
                        "title": "Laminar Branched Surfaces in 3-manifolds",
                        "abstract": "We define a laminar branched surface to be a branched surface satisfying the following conditions: (1) Its horizontal boundary is incompressible; (2) there is no monogon; (3) there is no Reeb component; (4) there is no sink disk (after eliminating trivial bubbles in the branched surface). The first three conditions are standard in the theory of branched surfaces, and a sink disk is a disk branch of the branched surface with all branch directions of its boundary arcs pointing inwards. We will show in this paper that every laminar branched surface carries an essential lamination, and any essential lamination that is not a lamination by planes is carried by a laminar branched surface. This implies that a 3-manifold contains an essential lamination if and only if it contains a laminar branched surface."
                    }
                ]
            },
            "ce0d78e1-6e4f-45d1-8cb8-868ff10d2fd0": {
                "pk": "ce0d78e1-6e4f-45d1-8cb8-868ff10d2fd0",
                "name": "Huazhu Fu",
                "collaborators": [
                    "Lei Zhu",
                    "Liang Wan",
                    "Zhihao Chen",
                    "Lianyu Wang",
                    "Meng Wang",
                    "Daoqiang Zhang",
                    "Jiang Liu",
                    "Yanwu Xu",
                    "Xiaomeng Li",
                    "Jiansheng Fang"
                ],
                "domain": [
                    "Medical Imaging",
                    "Federated Learning",
                    "Domain Adaptation",
                    "Image Enhancement"
                ],
                "publications": [
                    {
                        "title": "Model Barrier: A Compact Un-Transferable Isolation Domain for Model Intellectual Property Protection",
                        "abstract": "As scientific and technological advancements result from human intellectual labor and computational costs, protecting model intellectual property (IP) has become increasingly important to encourage model creators and owners. Model IP protection involves preventing the use of well-trained models on unauthorized domains. To address this issue, we propose a novel approach called Compact Un-Transferable Isolation Domain (CUTI-domain), which acts as a barrier to block illegal transfers from authorized to unauthorized domains. Specifically, CUTI-domain blocks cross-domain transfers by highlighting the private style features of the authorized domain, leading to recognition failure on unauthorized domains with irrelevant private style features. Moreover, we provide two solutions for using CUTI-domain depending on whether the unauthorized domain is known or not: target-specified CUTI-domain and target-free CUTI-domain. Our comprehensive experimental results on four digit datasets, CIFAR10 & STL10, and VisDA-2017 dataset demonstrate that CUTI-domain can be easily implemented as a plug-and-play module with different backbones, providing an efficient solution for model IP protection."
                    },
                    {
                        "title": "Deep Triplet Hashing Network for Case-based Medical Image Retrieval",
                        "abstract": "Deep hashing methods have been shown to be the most efficient approximate nearest neighbor search techniques for large-scale image retrieval. However, existing deep hashing methods have a poor small-sample ranking performance for case-based medical image retrieval. The top-ranked images in the returned query results may be as a different class than the query image. This ranking problem is caused by classification, regions of interest (ROI), and small-sample information loss in the hashing space. To address the ranking problem, we propose an end-to-end framework, called Attention-based Triplet Hashing (ATH) network, to learn low-dimensional hash codes that preserve the classification, ROI, and small-sample information. We embed a spatial-attention module into the network structure of our ATH to focus on ROI information. The spatial-attention module aggregates the spatial information of feature maps by utilizing max-pooling, element-wise maximum, and element-wise mean operations jointly along the channel axis. The triplet cross-entropy loss can help to map the classification information of images and similarity between images into the hash codes. Extensive experiments on two case-based medical datasets demonstrate that our proposed ATH can further improve the retrieval performance compared to the state-of-the-art deep hashing methods and boost the ranking performance for small samples. Compared to the other loss methods, the triplet cross-entropy loss can enhance the classification performance and hash code-discriminability"
                    },
                    {
                        "title": "Bridging Synthetic and Real Images: a Transferable and Multiple Consistency aided Fundus Image Enhancement Framework",
                        "abstract": "Deep learning based image enhancement models have largely improved the readability of fundus images in order to decrease the uncertainty of clinical observations and the risk of misdiagnosis. However, due to the difficulty of acquiring paired real fundus images at different qualities, most existing methods have to adopt synthetic image pairs as training data. The domain shift between the synthetic and the real images inevitably hinders the generalization of such models on clinical data. In this work, we propose an end-to-end optimized teacher-student framework to simultaneously conduct image enhancement and domain adaptation. The student network uses synthetic pairs for supervised enhancement, and regularizes the enhancement model to reduce domain-shift by enforcing teacher-student prediction consistency on the real fundus images without relying on enhanced ground-truth. Moreover, we also propose a novel multi-stage multi-attention guided enhancement network (MAGE-Net) as the backbones of our teacher and student network. Our MAGE-Net utilizes multi-stage enhancement module and retinal structure preservation module to progressively integrate the multi-scale features and simultaneously preserve the retinal structures for better fundus image quality enhancement. Comprehensive experiments on both real and synthetic datasets demonstrate that our framework outperforms the baseline approaches. Moreover, our method also benefits the downstream clinical tasks."
                    },
                    {
                        "title": "Say No to Freeloader: Protecting Intellectual Property of Your Deep Model",
                        "abstract": "Model intellectual property (IP) protection has attracted growing attention as science and technology advancements stem from human intellectual labor and computational expenses. Ensuring IP safety for trainers and owners is of utmost importance, particularly in domains where ownership verification and applicability authorization are required. A notable approach to safeguarding model IP involves proactively preventing the use of well-trained models of authorized domains from unauthorized domains. In this paper, we introduce a novel Compact Un-transferable Pyramid Isolation Domain (CUPI-Domain) which serves as a barrier against illegal transfers from authorized to unauthorized domains. Drawing inspiration from human transitive inference and learning abilities, the CUPI-Domain is designed to obstruct cross-domain transfers by emphasizing the distinctive style features of the authorized domain. This emphasis leads to failure in recognizing irrelevant private style features on unauthorized domains. To this end, we propose novel CUPI-Domain generators, which select features from both authorized and CUPI-Domain as anchors. Then, we fuse the style features and semantic features of these anchors to generate labeled and style-rich CUPI-Domain. Additionally, we design external Domain-Information Memory Banks (DIMB) for storing and updating labeled pyramid features to obtain stable domain class features and domain class-wise style features. Based on the proposed whole method, the novel style and discriminative loss functions are designed to effectively enhance the distinction in style and discriminative features between authorized and unauthorized domains, respectively. Moreover, we provide two solutions for utilizing CUPI-Domain based on whether the unauthorized domain is known: target-specified CUPI-Domain and target-free CUPI-Domain."
                    },
                    {
                        "title": "Learning to screen Glaucoma like the ophthalmologists",
                        "abstract": "GAMMA Challenge is organized to encourage the AI models to screen the glaucoma from a combination of 2D fundus image and 3D optical coherence tomography volume, like the ophthalmologists."
                    },
                    {
                        "title": "Retinal Image Segmentation with a Structure-Texture Demixing Network",
                        "abstract": "Retinal image segmentation plays an important role in automatic disease diagnosis. This task is very challenging because the complex structure and texture information are mixed in a retinal image, and distinguishing the information is difficult. Existing methods handle texture and structure jointly, which may lead biased models toward recognizing textures and thus results in inferior segmentation performance. To address it, we propose a segmentation strategy that seeks to separate structure and texture components and significantly improve the performance. To this end, we design a structure-texture demixing network (STD-Net) that can process structures and textures differently and better. Extensive experiments on two retinal image segmentation tasks (i.e., blood vessel segmentation, optic disc and cup segmentation) demonstrate the effectiveness of the proposed method."
                    },
                    {
                        "title": "Federated Noisy Client Learning",
                        "abstract": "Federated learning (FL) collaboratively trains a shared global model depending on multiple local clients, while keeping the training data decentralized in order to preserve data privacy. However, standard FL methods ignore the noisy client issue, which may harm the overall performance of the shared model. We first investigate critical issue caused by noisy clients in FL and quantify the negative impact of the noisy clients in terms of the representations learned by different layers. We have the following two key observations: (1) the noisy clients can severely impact the convergence and performance of the global model in FL, and (2) the noisy clients can induce greater bias in the deeper layers than the former layers of the global model. Based on the above observations, we propose Fed-NCL, a framework that conducts robust federated learning with noisy clients. Specifically, Fed-NCL first identifies the noisy clients through well estimating the data quality and model divergence. Then robust layer-wise aggregation is proposed to adaptively aggregate the local models of each client to deal with the data heterogeneity caused by the noisy clients. We further perform the label correction on the noisy clients to improve the generalization of the global model. Experimental results on various datasets demonstrate that our algorithm boosts the performances of different state-of-the-art systems with noisy clients. Our code is available on https://github.com/TKH666/Fed-NCL"
                    },
                    {
                        "title": "Modeling and Enhancing Low-quality Retinal Fundus Images",
                        "abstract": "Retinal fundus images are widely used for the clinical screening and diagnosis of eye diseases. However, fundus images captured by operators with various levels of experience have a large variation in quality. Low-quality fundus images increase uncertainty in clinical observation and lead to the risk of misdiagnosis. However, due to the special optical beam of fundus imaging and structure of the retina, natural image enhancement methods cannot be utilized directly to address this. In this paper, we first analyze the ophthalmoscope imaging system and simulate a reliable degradation of major inferior-quality factors, including uneven illumination, image blurring, and artifacts. Then, based on the degradation model, a clinically oriented fundus enhancement network (cofe-Net) is proposed to suppress global degradation factors, while simultaneously preserving anatomical retinal structures and pathological characteristics for clinical observation and analysis. Experiments on both synthetic and real images demonstrate that our algorithm effectively corrects low-quality fundus images without losing retinal details. Moreover, we also show that the fundus correction method can benefit medical image analysis applications, e.g., retinal vessel segmentation and optic disc/cup detection."
                    },
                    {
                        "title": "Trusted Multi-View Classification",
                        "abstract": "Multi-view classification (MVC) generally focuses on improving classification accuracy by using information from different views, typically integrating them into a unified comprehensive representation for downstream tasks. However, it is also crucial to dynamically assess the quality of a view for different samples in order to provide reliable uncertainty estimations, which indicate whether predictions can be trusted. To this end, we propose a novel multi-view classification method, termed trusted multi-view classification, which provides a new paradigm for multi-view learning by dynamically integrating different views at an evidence level. The algorithm jointly utilizes multiple views to promote both classification reliability and robustness by integrating evidence from each view. To achieve this, the Dirichlet distribution is used to model the distribution of the class probabilities, parameterized with evidence from different views and integrated with the Dempster-Shafer theory. The unified learning framework induces accurate uncertainty and accordingly endows the model with both reliability and robustness for out-of-distribution samples. Extensive experimental results validate the effectiveness of the proposed model in accuracy, reliability and robustness."
                    },
                    {
                        "title": "RSCFed: Random Sampling Consensus Federated Semi-supervised Learning",
                        "abstract": "Federated semi-supervised learning (FSSL) aims to derive a global model by training fully-labeled and fully-unlabeled clients or training partially labeled clients. The existing approaches work well when local clients have independent and identically distributed (IID) data but fail to generalize to a more practical FSSL setting, i.e., Non-IID setting. In this paper, we present a Random Sampling Consensus Federated learning, namely RSCFed, by considering the uneven reliability among models from fully-labeled clients, fully-unlabeled clients or partially labeled clients. Our key motivation is that given models with large deviations from either labeled clients or unlabeled clients, the consensus could be reached by performing random sub-sampling over clients. To achieve it, instead of directly aggregating local models, we first distill several sub-consensus models by random sub-sampling over clients and then aggregating the sub-consensus models to the global model. To enhance the robustness of sub-consensus models, we also develop a novel distance-reweighted model aggregation method. Experimental results show that our method outperforms state-of-the-art methods on three benchmarked datasets, including both natural and medical images. The code is available at https://github.com/XMed-Lab/RSCFed."
                    },
                    {
                        "title": "Diff-UNet: A Diffusion Embedded Network for Volumetric Segmentation",
                        "abstract": "In recent years, Denoising Diffusion Models have demonstrated remarkable success in generating semantically valuable pixel-wise representations for image generative modeling. In this study, we propose a novel end-to-end framework, called Diff-UNet, for medical volumetric segmentation. Our approach integrates the diffusion model into a standard U-shaped architecture to extract semantic information from the input volume effectively, resulting in excellent pixel-level representations for medical volumetric segmentation. To enhance the robustness of the diffusion model's prediction results, we also introduce a Step-Uncertainty based Fusion (SUF) module during inference to combine the outputs of the diffusion models at each step. We evaluate our method on three datasets, including multimodal brain tumors in MRI, liver tumors, and multi-organ CT volumes, and demonstrate that Diff-UNet outperforms other state-of-the-art methods significantly. Our experimental results also indicate the universality and effectiveness of the proposed model. The proposed framework has the potential to facilitate the accurate diagnosis and treatment of medical conditions by enabling more precise segmentation of anatomical structures. The codes of Diff-UNet are available at https://github.com/ge-xing/Diff-UNet"
                    },
                    {
                        "title": "Federated Model Aggregation via Self-Supervised Priors for Highly Imbalanced Medical Image Classification",
                        "abstract": "In the medical field, federated learning commonly deals with highly imbalanced datasets, including skin lesions and gastrointestinal images. Existing federated methods under highly imbalanced datasets primarily focus on optimizing a global model without incorporating the intra-class variations that can arise in medical imaging due to different populations, findings, and scanners. In this paper, we study the inter-client intra-class variations with publicly available self-supervised auxiliary networks. Specifically, we find that employing a shared auxiliary pre-trained model, like MoCo-V2, locally on every client yields consistent divergence measurements. Based on these findings, we derive a dynamic balanced model aggregation via self-supervised priors (MAS) to guide the global model optimization. Fed-MAS can be utilized with different local learning methods for effective model aggregation toward a highly robust and unbiased global model. Our code is available at \\url{https://github.com/xmed-lab/Fed-MAS}."
                    },
                    {
                        "title": "Teaching AI the Anatomy Behind the Scan: Addressing Anatomical Flaws in Medical Image Segmentation with Learnable Prior",
                        "abstract": "Imposing key anatomical features, such as the number of organs, their shapes and relative positions, is crucial for building a robust multi-organ segmentation model. Current attempts to incorporate anatomical features include broadening the effective receptive field (ERF) size with data-intensive modules, or introducing anatomical constraints that scales poorly to multi-organ segmentation. We introduce a novel architecture called the Anatomy-Informed Cascaded Segmentation Network (AIC-Net). AIC-Net incorporates a learnable input termed \"Anatomical Prior\", which can be adapted to patient-specific anatomy using a differentiable spatial deformation. The deformed prior later guides decoder layers towards more anatomy-informed predictions. We repeat this process at a local patch level to enhance the representation of intricate objects, resulting in a cascaded network structure. AIC-Net is a general method that enhances any existing segmentation models to be more anatomy-aware. We have validated the performance of AIC-Net, with various backbones, on two multi-organ segmentation tasks: abdominal organs and vertebrae. For each respective task, our benchmarks demonstrate improved dice score and Hausdorff distance."
                    },
                    {
                        "title": "Topicwise Separable Sentence Retrieval for Medical Report Generation",
                        "abstract": "Automated radiology reporting holds immense clinical potential in alleviating the burdensome workload of radiologists and mitigating diagnostic bias. Recently, retrieval-based report generation methods have garnered increasing attention due to their inherent advantages in terms of the quality and consistency of generated reports. However, due to the long-tail distribution of the training data, these models tend to learn frequently occurring sentences and topics, overlooking the rare topics. Regrettably, in many cases, the descriptions of rare topics often indicate critical findings that should be mentioned in the report. To address this problem, we introduce a Topicwise Separable Sentence Retrieval (Teaser) for medical report generation. To ensure comprehensive learning of both common and rare topics, we categorize queries into common and rare types to learn differentiated topics, and then propose Topic Contrastive Loss to effectively align topics and queries in the latent space. Moreover, we integrate an Abstractor module following the extraction of visual features, which aids the topic decoder in gaining a deeper understanding of the visual observational intent. Experiments on the MIMIC-CXR and IU X-ray datasets demonstrate that Teaser surpasses state-of-the-art models, while also validating its capability to effectively represent rare topics and establish more dependable correspondences between queries and topics."
                    },
                    {
                        "title": "Is Dataset Quality Still a Concern in Diagnosis Using Large Foundation Model?",
                        "abstract": "Recent advancements in pre-trained large foundation models (LFM) have yielded significant breakthroughs across various domains, including natural language processing and computer vision. These models have been particularly impactful in the domain of medical diagnostic tasks. With abundant unlabeled data, an LFM has been developed for fundus images using the Vision Transformer (VIT) and a self-supervised learning framework. This LFM has shown promising performance in fundus disease diagnosis across multiple datasets. On the other hand, deep learning models have long been challenged by dataset quality issues, such as image quality and dataset bias. To investigate the influence of data quality on LFM, we conducted explorations in two fundus diagnosis tasks using datasets of varying quality. Specifically, we explored the following questions: Is LFM more robust to image quality? Is LFM affected by dataset bias? Can fine-tuning techniques alleviate these effects? Our investigation found that LFM exhibits greater resilience to dataset quality issues, including image quality and dataset bias, compared to typical convolutional networks. Furthermore, we discovered that overall fine-tuning is an effective adapter for LFM to mitigate the impact of dataset quality issues."
                    },
                    {
                        "title": "Reliable Source Approximation: Source-Free Unsupervised Domain Adaptation for Vestibular Schwannoma MRI Segmentation",
                        "abstract": "Source-Free Unsupervised Domain Adaptation (SFUDA) has recently become a focus in the medical image domain adaptation, as it only utilizes the source model and does not require annotated target data. However, current SFUDA approaches cannot tackle the complex segmentation task across different MRI sequences, such as the vestibular schwannoma segmentation. To address this problem, we proposed Reliable Source Approximation (RSA), which can generate source-like and structure-preserved images from the target domain for updating model parameters and adapting domain shifts. Specifically, RSA deploys a conditional diffusion model to generate multiple source-like images under the guidance of varying edges of one target image. An uncertainty estimation module is then introduced to predict and refine reliable pseudo labels of generated images, and the prediction consistency is developed to select the most reliable generations. Subsequently, all reliable generated images and their pseudo labels are utilized to update the model. Our RSA is validated on vestibular schwannoma segmentation across multi-modality MRI. The experimental results demonstrate that RSA consistently improves domain adaptation performance over other state-of-the-art SFUDA methods. Code is available at https://github.com/zenghy96/Reliable-Source-Approximation."
                    },
                    {
                        "title": "A Review of Co-saliency Detection Technique: Fundamentals, Applications, and Challenges",
                        "abstract": "Co-saliency detection is a newly emerging and rapidly growing research area in computer vision community. As a novel branch of visual saliency, co-saliency detection refers to the discovery of common and salient foregrounds from two or more relevant images, and can be widely used in many computer vision tasks. The existing co-saliency detection algorithms mainly consist of three components: extracting effective features to represent the image regions, exploring the informative cues or factors to characterize co-saliency, and designing effective computational frameworks to formulate co-saliency. Although numerous methods have been developed, the literature is still lacking a deep review and evaluation of co-saliency detection techniques. In this paper, we aim at providing a comprehensive review of the fundamentals, challenges, and applications of co-saliency detection. Specifically, we provide an overview of some related computer vision works, review the history of co-saliency detection, summarize and categorize the major algorithms in this research area, discuss some open issues in this area, present the potential applications of co-saliency detection, and finally point out some unsolved challenges and promising future works. We expect this review to be beneficial to both fresh and senior researchers in this field, and give insights to researchers in other related areas regarding the utility of co-saliency detection algorithms."
                    },
                    {
                        "title": "Multi-Contrast MRI Super-Resolution via a Multi-Stage Integration Network",
                        "abstract": "Super-resolution (SR) plays a crucial role in improving the image quality of magnetic resonance imaging (MRI). MRI produces multi-contrast images and can provide a clear display of soft tissues. However, current super-resolution methods only employ a single contrast, or use a simple multi-contrast fusion mechanism, ignoring the rich relations among different contrasts, which are valuable for improving SR. In this work, we propose a multi-stage integration network (i.e., MINet) for multi-contrast MRI SR, which explicitly models the dependencies between multi-contrast images at different stages to guide image SR. In particular, our MINet first learns a hierarchical feature representation from multiple convolutional stages for each of different-contrast image. Subsequently, we introduce a multi-stage integration module to mine the comprehensive relations between the representations of the multi-contrast images. Specifically, the module matches each representation with all other features, which are integrated in terms of their similarities to obtain an enriched representation. Extensive experiments on fastMRI and real-world clinical datasets demonstrate that 1) our MINet outperforms state-of-the-art multi-contrast SR methods in terms of various metrics and 2) our multi-stage integration module is able to excavate complex interactions among multi-contrast features at different stages, leading to improved target-image quality."
                    },
                    {
                        "title": "NuI-Go: Recursive Non-Local Encoder-Decoder Network for Retinal Image Non-Uniform Illumination Removal",
                        "abstract": "Retinal images have been widely used by clinicians for early diagnosis of ocular diseases. However, the quality of retinal images is often clinically unsatisfactory due to eye lesions and imperfect imaging process. One of the most challenging quality degradation issues in retinal images is non-uniform which hinders the pathological information and further impairs the diagnosis of ophthalmologists and computer-aided analysis.To address this issue, we propose a non-uniform illumination removal network for retinal image, called NuI-Go, which consists of three Recursive Non-local Encoder-Decoder Residual Blocks (NEDRBs) for enhancing the degraded retinal images in a progressive manner. Each NEDRB contains a feature encoder module that captures the hierarchical feature representations, a non-local context module that models the context information, and a feature decoder module that recovers the details and spatial dimension. Additionally, the symmetric skip-connections between the encoder module and the decoder module provide long-range information compensation and reuse. Extensive experiments demonstrate that the proposed method can effectively remove the non-uniform illumination on retinal images while well preserving the image details and color. We further demonstrate the advantages of the proposed method for improving the accuracy of retinal vessel segmentation."
                    },
                    {
                        "title": "Learning Physical-Spatio-Temporal Features for Video Shadow Removal",
                        "abstract": "Shadow removal in a single image has received increasing attention in recent years. However, removing shadows over dynamic scenes remains largely under-explored. In this paper, we propose the first data-driven video shadow removal model, termed PSTNet, by exploiting three essential characteristics of video shadows, i.e., physical property, spatio relation, and temporal coherence. Specifically, a dedicated physical branch was established to conduct local illumination estimation, which is more applicable for scenes with complex lighting and textures, and then enhance the physical features via a mask-guided attention strategy. Then, we develop a progressive aggregation module to enhance the spatio and temporal characteristics of features maps, and effectively integrate the three kinds of features. Furthermore, to tackle the lack of datasets of paired shadow videos, we synthesize a dataset (SVSRD-85) with aid of the popular game GTAV by controlling the switch of the shadow renderer. Experiments against 9 state-of-the-art models, including image shadow removers and image/video restoration methods, show that our method improves the best SOTA in terms of RMSE error for the shadow area by 14.7. In addition, we develop a lightweight model adaptation strategy to make our synthetic-driven model effective in real world scenes. The visual comparison on the public SBU-TimeLapse dataset verifies the generalization ability of our model in real scenes."
                    }
                ]
            }
        }
    },
    "2310.06836": {
        "paper_data": {
            "title": "A General Protocol to Probe Large Vision Models for 3D Physical Understanding",
            "url": "http://arxiv.org/abs/2310.06836v3",
            "arxiv_id": "2310.06836",
            "authors": [
                "Guanqi Zhan",
                "Chuanxia Zheng",
                "Weidi Xie",
                "Andrew Zisserman"
            ],
            "abstract": "Our objective in this paper is to probe large vision models to determine to what extent they 'understand' different physical properties of the 3D scene depicted in an image. To this end, we make the following contributions: (i) We introduce a general and lightweight protocol to evaluate whether features of an off-the-shelf large vision model encode a number of physical 'properties' of the 3D scene, by training discriminative classifiers on the features for these properties. The probes are applied on datasets of real images with annotations for the property. (ii) We apply this protocol to properties covering scene geometry, scene material, support relations, lighting, and view-dependent measures, and large vision models including CLIP, DINOv1, DINOv2, VQGAN, Stable Diffusion. (iii) We find that features from Stable Diffusion and DINOv2 are good for discriminative learning of a number of properties, including scene geometry, support relations, shadows and depth, but less performant for occlusion and material, while outperforming DINOv1, CLIP and VQGAN for all properties. (iv) It is observed that different time steps of Stable Diffusion features, as well as different transformer layers of DINO/CLIP/VQGAN, are good at different properties, unlocking potential applications of 3D physical understanding.",
            "introduction": "   1 Introduction  Figure 1:  Motivation: What do large vision models know about the 3D scene? We take Stable Diffusion as an example because Stable Diffusion is generative, and so its output is an image that can be judged directly for verisimilitude. The Stable Diffusion inpainting model is here tasked with inpainting the masked region of the real images. It correctly predicts a shadow consistent with the lighting direction (top), and a supporting structure consistent with the scene geometry (bottom). This indicates that the Stable Diffusion model generation is consistent with the geometry (of the light source direction) and physical (support) properties. These examples are only for illustration and we probe a general Stable Diffusion network to determine whether there are explicit features for such 3D scene properties. The appendix provides more examples of Stable Diffusion\u2019s capability to predict different physical properties of the scene.    The large-scale pre-trained vision foundation models have achieved great success in computer vision tasks, including classification (CLIP\u00a0[30, 18]), segmentation (DINO\u00a0[5, 27]), and image generation (VQGAN\u00a0[11], Stable Diffusion\u00a0[31]) with strong generalisation capabilities. However, they are mainly trained with 2D images, which are the projection of the 3D physical world. This naturally raises the question of to what extent these large-scale vision models have learned about the 3D scene depicted with only the 2D images. Our objective in this paper is to investigate this question, and we do this precisely by determining whether features from these large-scale pre-trained vision models can be used to estimate the true geometric and physical properties of the 3D scene. If they can, then that is evidence that the 3D scene can be correctly modelled by its 2D projections. For example, as an indication that Stable Diffusion is 3D and physics aware, Figure\u00a01 shows the result of the off-the-shelf Stable Diffusion model\u00a0[31] inpainting masked regions in real images \u2013 it correctly predicts shadows and supporting structures.   To answer this question, we propose a general and lightweight evaluation protocol to systematically and efficiently \u2018probe\u2019 a pre-trained network on its ability to represent a number of \u2018properties\u2019 of the 3D scene and viewpoint. The protocol could be used for any network and any property of interest. We have probed properties including: 3D structure and material of the scene, such as surface layout; lighting, such as object-shadow relationships; and viewpoint dependent relations such as occlusion and depth.   The protocol involves three steps: First, a suitable real image evaluation dataset is selected that contains ground truth annotations for the property of interest, for example the SOBA dataset\u00a0[41] is used to probe the understanding of lighting, as it has annotations for object-shadow associations. This provides a train/val/test set for that property; Second, a grid search is carried out over the layers and time steps of the Stable Diffusion model, and transformer layers for other models, to select the optimal feature for determining that property. The selection involves learning the weights of a simple linear classifier for that property (e.g.\u00a0\u2018are these two regions in an object-shadow relationship or not\u2019); Third, the selected feature (layer, time step) and trained classifier are evaluated on a test set, and its performance answers the question of how well the model \u2018understands\u2019 that property.   In",
            "references": [
                {
                    "title": "Probing the 3D Awareness of Visual Foundation Models",
                    "abstract": "Recent advances in large-scale pretraining have yielded visual foundation models with strong capabilities. Not only can recent models generalize to arbitrary images for their training task, their intermediate representations are useful for other visual tasks such as detection and segmentation. Given that such models can classify, delineate, and local- ize objects in 2D, we ask whether they also represent their 3D structure? In this work, we analyze the 3D awareness of visual foundation models. We posit that 3D awareness implies that representations (1) encode the 3D structure of the scene and (2) consistently represent the surface across views. We conduct a series of experiments using task-specific probes and zero-shot inference procedures on frozen fea- tures. Our experiments reveal several limitations of the current models. Our code and analysis can be found at https://github.com/mbanani/probe3d."
                },
                {
                    "title": "ECoDepth: Effective Conditioning of Diffusion Models for Monocular Depth Estimation",
                    "abstract": "In the absence of parallax cues, a learning based single image depth estimation (SIDE) model relies heavily on shading and contextual cues in the image. While this simplicity is attractive, it is necessary to train such models on large and varied datasets, which are difficult to capture. It has been shown that using embeddings from pretrained foundational models, such as CLIP, improves zero shot transfer in several applications. Taking inspiration from this, in our paper we explore the use of global image priors generated from a pretrained ViT model to provide more detailed contextual information. We argue that the embedding vector from a ViT model, pretrained on a large dataset, captures greater relevant information for SIDE than the usual route of generating pseudo image captions, followed by CLIP based text embeddings. Based on this idea, we propose a new SIDE model using a diffusion backbone which is conditioned on ViT embeddings. Our proposed design establishes a new state-of-the-art (SOTA) for SIDE on NYU Depth v2 dataset, achieving Abs Rel error of 0.059(14% improvement) compared to 0.069 by the current SOTA (VPD). And on KITTI dataset, achieving Sq Rel error of 0.139 (2% improvement) compared to 0.142 by the current SOTA (GED). For zero shot transfer with a model trained on NYU Depth v2, we report mean relative improvement of (20%, 23%,81%, 25%) over NeWCRF on (Sun-RGBD, iBimsl, DIODE, HyperSim) datasets, compared to (16%, 18%, 45%, 9%) by ZoEDepth. The code is available in our project page."
                },
                {
                    "title": "Amodal Ground Truth and Completion in the Wild",
                    "abstract": "This paper studies amodal image segmentation: predicting entire object segmentation masks including both visible and invisible (occluded) parts. In previous work, the amodal segmentation ground truth on real images is usually pre-dicted by manual annotaton and thus is subjective. In contrast, we use 3D data to establish an automatic pipeline to determine authentic ground truth amodal masks for partially occluded objects in real images. This pipeline is used to construct an amodal completion evaluation benchmark, MP3D-Amodal, consisting of a variety of object categories and la-bels. To better handle the amodal completion task in the wild, we explore two architecture variants: a two-stage model that first infers the occluder, followed by amodal mask completion; and a one-stage model that exploits the representation power of Stable Diffusion for amodal segmentation across many categories. Without bells and whistles, our method achieves a new state-of-the-art performance on Amodal segmentation datasets that cover a large variety of objects, in-cluding COCOA and our new MP3D-Amodal dataset. The dataset, model, and code are available at https://www.robots.ox.ac.uk/~vgg/research/amodal/"
                },
                {
                    "title": "Atlantis: Enabling Underwater Depth Estimation with Stable Diffusion",
                    "abstract": "Monocular depth estimation has experienced significant progress on terrestrial images in recent years thanks to deep learning advancements. But it remains inadequate for underwater scenes primarily due to data scarcity. Given the inherent challenges of light attenuation and backscat-ter in water, acquiring clear underwater images or precise depth is notably difficult and costly. To mitigate this issue, learning-based approaches often rely on synthetic data or turn to self- or unsupervised manners. Nonetheless, their performance is often hindered by domain gap and looser constraints. In this paper, we propose a novel pipeline for generating photorealistic underwater images using accurate terrestrial depth. This approach facilitates the supervised training of models for underwater depth estimation, effectively reducing the performance disparity between ter-restrial and underwater environments. Contrary to previous synthetic datasets that merely apply style transfer to terres-trial images without scene content change, our approach uniquely creates vivid non-existent underwater scenes by leveraging terrestrial depth data through the innovative Stable Diffusion model. Specifically, we introduce a specialized Depth2Underwater ControlNet, trained on prepared {Underwater, Depth, Text} data triplets, for this generation task. Our newly developed dataset, Atlantis, enables terres-trial depth estimation models to achieve considerable improvements on unseen underwater scenes, surpassing their terrestrial pretrained counterparts both quantitatively and qualitatively. Moreover, we further show its practical utility by applying the improved depth in underwater image enhancement, and its smaller domain gap from the LLVM perspective. Code and dataset are publicly available at https://github.com/zkawfanx/Atlantis."
                },
                {
                    "title": "Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation",
                    "abstract": "Monocular depth estimation is a fundamental computer vision task. Recovering 3D depth from a single image is geometrically ill-posed and requires scene understanding, so it is not surprising that the rise of deep learning has led to a breakthrough. The impressive progress of monocular depth estimators has mirrored the growth in model capacity, from relatively modest CNNs to large Transformer architectures. Still, monocular depth estimators tend to struggle when presented with images with unfamiliar content and layout, since their knowledge of the visual world is restricted by the data seen during training, and challenged by zero-shot generalization to new domains. This motivates us to explore whether the extensive priors captured in recent generative diffusion models can enable better, more generalizable depth estimation. We introduce Marigold, a method for affine-invariant monocular depth estimation that is derived from Stable Diffusion and retains its rich prior knowledge. The estimator can be fine-tuned in a couple of days on a single GPU using only synthetic training data. It delivers state-of-the-art performance across a wide range of datasets, including over 20% performance gains in specific cases. Project page: https://marigoldmonodepth.github.io."
                },
                {
                    "title": "Shadows Don't Lie and Lines Can't Bend! Generative Models Don't know Projective Geometry\u2026for Now",
                    "abstract": "Generative models can produce impressively realistic images. This paper demonstrates that generated images have geometric features different from those of real images. We build a set of collections of generated images, prequalified to fool simple, signal-based classifiers into believing they are real. We then show that prequalified generated images can be identified reliably by classifiers that only look at geometric properties. We use three such classifiers. All three classifiers are denied access to image pixels, and look only at derived geometric features. The first classifier looks at the perspective field of the image, the second looks at lines detected in the image, and the third looks at relations between detected objects and shadows. Our procedure detects generated images more reliably than SOTA local signal based detectors, for images from a number of distinct generators. Saliency maps suggest that the classifiers can identify geometric problems reliably. We conclude that current generators cannot reliably reproduce geometric properties of real images."
                },
                {
                    "title": "DatasetDM: Synthesizing Data with Perception Annotations Using Diffusion Models",
                    "abstract": "Current deep networks are very data-hungry and benefit from training on largescale datasets, which are often time-consuming to collect and annotate. By contrast, synthetic data can be generated infinitely using generative models such as DALL-E and diffusion models, with minimal effort and cost. In this paper, we present DatasetDM, a generic dataset generation model that can produce diverse synthetic images and the corresponding high-quality perception annotations (e.g., segmentation masks, and depth). Our method builds upon the pre-trained diffusion model and extends text-guided image synthesis to perception data generation. We show that the rich latent code of the diffusion model can be effectively decoded as accurate perception annotations using a decoder module. Training the decoder only needs less than 1% (around 100 images) manually labeled images, enabling the generation of an infinitely large annotated dataset. Then these synthetic data can be used for training various perception models for downstream tasks. To showcase the power of the proposed approach, we generate datasets with rich dense pixel-wise labels for a wide range of downstream tasks, including semantic segmentation, instance segmentation, and depth estimation. Notably, it achieves 1) state-of-the-art results on semantic segmentation and instance segmentation; 2) significantly more robust on domain generalization than using the real data alone; and state-of-the-art results in zero-shot segmentation setting; and 3) flexibility for efficient application and novel task composition (e.g., image editing). The project website and code can be found at https://weijiawu.github.io/DatasetDM_page/ and https://github.com/showlab/DatasetDM, respectively"
                },
                {
                    "title": "Emergent Correspondence from Image Diffusion",
                    "abstract": "Finding correspondences between images is a fundamental problem in computer vision. In this paper, we show that correspondence emerges in image diffusion models without any explicit supervision. We propose a simple strategy to extract this implicit knowledge out of diffusion networks as image features, namely DIffusion FeaTures (DIFT), and use them to establish correspondences between real images. Without any additional fine-tuning or supervision on the task-specific data or annotations, DIFT is able to outperform both weakly-supervised methods and competitive off-the-shelf features in identifying semantic, geometric, and temporal correspondences. Particularly for semantic correspondence, DIFT from Stable Diffusion is able to outperform DINO and OpenCLIP by 19 and 14 accuracy points respectively on the challenging SPair-71k benchmark. It even outperforms the state-of-the-art supervised methods on 9 out of 18 categories while remaining on par for the overall performance. Project page: https://diffusionfeatures.github.io"
                },
                {
                    "title": "StyleGAN knows Normal, Depth, Albedo, and More",
                    "abstract": "Intrinsic images, in the original sense, are image-like maps of scene properties like depth, normal, albedo or shading. This paper demonstrates that StyleGAN can easily be induced to produce intrinsic images. The procedure is straightforward. We show that, if StyleGAN produces $G({w})$ from latents ${w}$, then for each type of intrinsic image, there is a fixed offset ${d}_c$ so that $G({w}+{d}_c)$ is that type of intrinsic image for $G({w})$. Here ${d}_c$ is {\\em independent of ${w}$}. The StyleGAN we used was pretrained by others, so this property is not some accident of our training regime. We show that there are image transformations StyleGAN will {\\em not} produce in this fashion, so StyleGAN is not a generic image regression engine. It is conceptually exciting that an image generator should ``know'' and represent intrinsic images. There may also be practical advantages to using a generative model to produce intrinsic images. The intrinsic images obtained from StyleGAN compare well both qualitatively and quantitatively with those obtained by using SOTA image regression techniques; but StyleGAN's intrinsic images are robust to relighting effects, unlike SOTA methods."
                },
                {
                    "title": "A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondence",
                    "abstract": "Text-to-image diffusion models have made significant advances in generating and editing high-quality images. As a result, numerous approaches have explored the ability of diffusion model features to understand and process single images for downstream tasks, e.g., classification, semantic segmentation, and stylization. However, significantly less is known about what these features reveal across multiple, different images and objects. In this work, we exploit Stable Diffusion (SD) features for semantic and dense correspondence and discover that with simple post-processing, SD features can perform quantitatively similar to SOTA representations. Interestingly, the qualitative analysis reveals that SD features have very different properties compared to existing representation learning features, such as the recently released DINOv2: while DINOv2 provides sparse but accurate matches, SD features provide high-quality spatial information but sometimes inaccurate semantic matches. We demonstrate that a simple fusion of these two features works surprisingly well, and a zero-shot evaluation using nearest neighbors on these fused features provides a significant performance gain over state-of-the-art methods on benchmark datasets, e.g., SPair-71k, PF-Pascal, and TSS. We also show that these correspondences can enable interesting applications such as instance swapping in two images."
                },
                {
                    "title": "Unsupervised Semantic Correspondence Using Stable Diffusion",
                    "abstract": "Text-to-image diffusion models are now capable of generating images that are often indistinguishable from real images. To generate such images, these models must understand the semantics of the objects they are asked to generate. In this work we show that, without any training, one can leverage this semantic knowledge within diffusion models to find semantic correspondences - locations in multiple images that have the same semantic meaning. Specifically, given an image, we optimize the prompt embeddings of these models for maximum attention on the regions of interest. These optimized embeddings capture semantic information about the location, which can then be transferred to another image. By doing so we obtain results on par with the strongly supervised state of the art on the PF-Willow dataset and significantly outperform (20.9% relative for the SPair-71k dataset) any existing weakly or unsupervised method on PF-Willow, CUB-200 and SPair-71k datasets."
                },
                {
                    "title": "Diffusion Hyperfeatures: Searching Through Time and Space for Semantic Correspondence",
                    "abstract": "Diffusion models have been shown to be capable of generating high-quality images, suggesting that they could contain meaningful internal representations. Unfortunately, the feature maps that encode a diffusion model's internal information are spread not only over layers of the network, but also over diffusion timesteps, making it challenging to extract useful descriptors. We propose Diffusion Hyperfeatures, a framework for consolidating multi-scale and multi-timestep feature maps into per-pixel feature descriptors that can be used for downstream tasks. These descriptors can be extracted for both synthetic and real images using the generation and inversion processes. We evaluate the utility of our Diffusion Hyperfeatures on the task of semantic keypoint correspondence: our method achieves superior performance on the SPair-71k real image benchmark. We also demonstrate that our method is flexible and transferable: our feature aggregation network trained on the inversion features of real image pairs can be used on the generation features of synthetic image pairs with unseen objects and compositions. Our code is available at https://diffusion-hyperfeatures.github.io."
                },
                {
                    "title": "DINOv2: Learning Robust Visual Features without Supervision",
                    "abstract": "The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels."
                },
                {
                    "title": "iDisc: Internal Discretization for Monocular Depth Estimation",
                    "abstract": "Monocular depth estimation is fundamental for 3D scene understanding and downstream applications. However, even under the supervised setup, it is still challenging and ill-posed due to the lack of full geometric constraints. Although a scene can consist of millions of pixels, there are fewer high-level patterns. We propose iDisc to learn those patterns with internal discretized representations. The method implicitly partitions the scene into a set of high-level patterns. In particular, our new module, Internal Discretization (ID), implements a continuous-discrete-continuous bottleneck to learn those concepts without supervision. In contrast to state-of-the-art methods, the proposed model does not enforce any explicit constraints or priors on the depth output. The whole network with the ID module can be trained end-to-end, thanks to the bottleneck module based on attention. Our method sets the new state of the art with significant improvements on NYU-Depth v2 and KITTI, outperforming all published methods on the official KITTI benchmark. iDisc can also achieve state-of-the-art results on surface normal estimation. Further, we explore the model generalization capability via zero-shot testing. We observe the compelling need to promote diversification in the outdoor scenario. Hence, we introduce splits of two autonomous driving datasets, DDAD and Argoverse. Code is available at http://vis.xyz/pub/idisc."
                },
                {
                    "title": "Segment Anything",
                    "abstract": "We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive \u2013 often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images at segment-anything.com to foster research into foundation models for computer vision. We recommend reading the full paper at: arxiv.org/abs/2304.02643."
                },
                {
                    "title": "DiffuMask: Synthesizing Images with Pixel-level Annotations for Semantic Segmentation Using Diffusion Models",
                    "abstract": "Collecting and annotating images with pixel-wise labels is time-consuming and laborious. In contrast, synthetic data can be freely available using a generative model (e.g., DALL-E, Stable Diffusion). In this paper, we show that it is possible to automatically obtain accurate semantic masks of synthetic images generated by the Off-the-shelf Stable Diffusion model, which uses only text-image pairs during training. Our approach, termed DiffuMask, exploits the potential of the cross-attention map between text and image, which is natural and seamless to extend the text-driven image synthesis to semantic mask generation. DiffuMask uses text-guided cross-attention information to localize class/word-specific regions, which are combined with practical techniques to create a novel high-resolution and class-discriminative pixel-wise mask. The methods help to significantly reduce data collection and annotation costs. Experiments demonstrate that the existing segmentation methods trained on synthetic data of DiffuMask can achieve a competitive performance over the counterpart of real data (VOC 2012, Cityscapes). For some classes (e.g., bird), DiffuMask presents promising performance, close to the state-of-the-art result of real data (within 3% mIoU gap). Moreover, in the open-vocabulary segmentation (zero-shot) setting, DiffuMask achieves new state-of-the-art results on the Unseen classes of VOC 2012. The project website can be found at ${\\color{red}{\\text{DiffuMask}}}$."
                },
                {
                    "title": "Open-Vocabulary Panoptic Segmentation with Text-to-Image Diffusion Models",
                    "abstract": "We present ODISE: Open-vocabulary DIffusion-based panoptic SEgmentation, which unifies pre-trained text-image diffusion and discriminative models to perform open-vocabulary panoptic segmentation. Text-to-image diffusion models have the remarkable ability to generate high-quality images with diverse open-vocabulary language descriptions. This demonstrates that their internal representation space is highly correlated with open concepts in the real world. Text-image discriminative models like CLIP, on the other hand, are good at classifying images into open-vocabulary labels. We leverage the frozen internal representations of both these models to perform panoptic segmentation of any category in the wild. Our approach outperforms the previous state of the art by significant margins on both open-vocabulary panoptic and semantic segmentation tasks. In particular, with COCO training only, our method achieves 23.4 PQ and 30.0 mIoU on the ADE20K dataset, with 8.3 PQ and 7.9 mIoU absolute improvement over the previous state of the art. We open-source our code and models at https://github.com/NVlabs/ODISE."
                },
                {
                    "title": "Unleashing Text-to-Image Diffusion Models for Visual Perception",
                    "abstract": "Diffusion models (DMs) have become the new trend of generative models and have demonstrated a powerful ability of conditional synthesis. Among those, text-to-image diffusion models pre-trained on large-scale image-text pairs are highly controllable by customizable prompts. Unlike the unconditional generative models that focus on low-level attributes and details, text-to-image diffusion models contain more high-level knowledge thanks to the vision-language pre-training. In this paper, we propose VPD (Visual Perception with pre-trained Diffusion models), a new framework that exploits the semantic information of a pre-trained text-to-image diffusion model in visual perception tasks. Instead of using the pre-trained denoising autoencoder in a diffusion-based pipeline, we simply use it as a backbone and aim to study how to take full advantage of the learned knowledge. Specifically, we prompt the denoising decoder with proper textual inputs and refine the text features with an adapter, leading to a better alignment to the pre-trained stage and making the visual contents interact with the text prompts. We also propose to utilize the cross-attention maps between the visual features and the text features to provide explicit guidance. Compared with other pre-training methods, we show that vision-language pre-trained diffusion models can be faster adapted to downstream visual perception tasks using the proposed VPD. Extensive experiments on semantic segmentation, referring image segmentation, and depth estimation demonstrate the effectiveness of our method. Notably, VPD attains 0.254 RMSE on NYUv2 depth estimation and 73.3% oIoU on RefCOCO-val referring image segmentation, establishing new records on these two benchmarks. Code is available at https://github.com/wl-zhao/VPD."
                },
                {
                    "title": "Open-vocabulary Object Segmentation with Diffusion Models",
                    "abstract": "The goal of this paper is to extract the visual-language correspondence from a pre-trained text-to-image diffusion model, in the form of segmentation map, i.e., simultaneously generating images and segmentation masks for the corresponding visual entities described in the text prompt. We make the following contributions: (i) we pair the existing Stable Diffusion model with a novel grounding module, that can be trained to align the visual and textual embedding space of the diffusion model with only a small number of object categories; (ii) we establish an automatic pipeline for constructing a dataset, that consists of {image, segmentation mask, text prompt} triplets, to train the proposed grounding module; (iii) we evaluate the performance of open-vocabulary grounding on images generated from the text-to-image diffusion model and show that the module can well segment the objects of categories beyond seen ones at training time, as shown in Fig. 1; (iv) we adopt the augmented diffusion model to build a synthetic semantic segmentation dataset, and show that, training a standard segmentation model on such dataset demonstrates competitive performance on the zero-shot segmentation (ZS3) benchmark, which opens up new opportunities for adopting the powerful diffusion model for discriminative tasks."
                },
                {
                    "title": "A Tri-Layer Plugin to Improve Occluded Detection",
                    "abstract": "Detecting occluded objects still remains a challenge for state-of-the-art object detectors. The objective of this work is to improve the detection for such objects, and thereby improve the overall performance of a modern object detector. To this end we make the following four contributions: (1) We propose a simple 'plugin' module for the detection head of two-stage object detectors to improve the recall of partially occluded objects. The module predicts a tri-layer of segmentation masks for the target object, the occluder and the occludee, and by doing so is able to better predict the mask of the target object. (2) We propose a scalable pipeline for generating training data for the module by using amodal completion of existing object detection and instance segmentation training datasets to establish occlusion relationships. (3) We also establish a COCO evaluation dataset to measure the recall performance of partially occluded and separated objects. (4) We show that the plugin module inserted into a two-stage detector can boost the performance significantly, by only fine-tuning the detection head, and with additional improvements if the entire architecture is fine-tuned. COCO results are reported for Mask R-CNN with Swin-T or Swin-S backbones, and Cascade Mask R-CNN with a Swin-B backbone."
                },
                {
                    "title": "LAION-5B: An open large-scale dataset for training next generation image-text models",
                    "abstract": "Groundbreaking language-vision architectures like CLIP and DALL-E proved the utility of training on large amounts of noisy image-text data, without relying on expensive accurate labels used in standard vision unimodal supervised learning. The resulting models showed capabilities of strong text-guided image generation and transfer to downstream tasks, while performing remarkably at zero-shot classification with noteworthy out-of-distribution robustness. Since then, large-scale language-vision models like ALIGN, BASIC, GLIDE, Flamingo and Imagen made further improvements. Studying the training and capabilities of such models requires datasets containing billions of image-text pairs. Until now, no datasets of this size have been made openly available for the broader research community. To address this problem and democratize research on large-scale multi-modal models, we present LAION-5B - a dataset consisting of 5.85 billion CLIP-filtered image-text pairs, of which 2.32B contain English language. We show successful replication and fine-tuning of foundational models like CLIP, GLIDE and Stable Diffusion using the dataset, and discuss further experiments enabled with an openly available dataset of this scale. Additionally we provide several nearest neighbor indices, an improved web-interface for dataset exploration and subset generation, and detection scores for watermark, NSFW, and toxic content detection. Announcement page https://laion.ai/laion-5b-a-new-era-of-open-large-scale-multi-modal-datasets/"
                },
                {
                    "title": "Is synthetic data from generative models ready for image recognition?",
                    "abstract": "Recent text-to-image generation models have shown promising results in generating high-fidelity photo-realistic images. Though the results are astonishing to human eyes, how applicable these generated images are for recognition tasks remains under-explored. In this work, we extensively study whether and how synthetic images generated from state-of-the-art text-to-image generation models can be used for image recognition tasks, and focus on two perspectives: synthetic data for improving classification models in data-scarce settings (i.e. zero-shot and few-shot), and synthetic data for large-scale model pre-training for transfer learning. We showcase the powerfulness and shortcomings of synthetic data from existing generative models, and propose strategies for better applying synthetic data for recognition tasks. Code: https://github.com/CVMI-Lab/SyntheticData."
                },
                {
                    "title": "Self-Supervised Geometric Correspondence for Category-Level 6D Object Pose Estimation in the Wild",
                    "abstract": "While 6D object pose estimation has wide applications across computer vision and robotics, it remains far from being solved due to the lack of annotations. The problem becomes even more challenging when moving to category-level 6D pose, which requires generalization to unseen instances. Current approaches are restricted by leveraging annotations from simulation or collected from humans. In this paper, we overcome this barrier by introducing a self-supervised learning approach trained directly on large-scale real-world object videos for category-level 6D pose estimation in the wild. Our framework reconstructs the canonical 3D shape of an object category and learns dense correspondences between input images and the canonical shape via surface embedding. For training, we propose novel geometrical cycle-consistency losses which construct cycles across 2D-3D spaces, across different instances and different time steps. The learned correspondence can be applied for 6D pose estimation and other downstream tasks such as keypoint transfer. Surprisingly, our method, without any human annotations or simulators, can achieve on-par or even better performance than previous supervised or semi-supervised methods on in-the-wild images. Our project page is: https://kywind.github.io/self-pose ."
                },
                {
                    "title": "A Dense Material Segmentation Dataset for Indoor and Outdoor Scene Parsing",
                    "abstract": "A key algorithm for understanding the world is material segmentation, which assigns a label (metal, glass, etc.) to each pixel. We find that a model trained on existing data underperforms in some settings and propose to address this with a large-scale dataset of 3.2 million dense segments on 44,560 indoor and outdoor images, which is 23x more segments than existing data. Our data covers a more diverse set of scenes, objects, viewpoints and materials, and contains a more fair distribution of skin types. We show that a model trained on our data outperforms a state-of-the-art model across datasets and viewpoints. We propose a large-scale scene parsing benchmark and baseline of 0.729 per-pixel accuracy, 0.585 mean class accuracy and 0.420 mean IoU across 46 materials."
                },
                {
                    "title": "Zero-Shot Category-Level Object Pose Estimation",
                    "abstract": "Object pose estimation is an important component of most vision pipelines for embodied agents, as well as in 3D vision more generally. In this paper we tackle the problem of estimating the pose of novel object categories in a zero-shot manner. This extends much of the existing literature by removing the need for pose-labelled datasets or category-specific CAD models for training or inference. Specifically, we make the following contributions. First, we formalise the zero-shot, category-level pose estimation problem and frame it in a way that is most applicable to real-world embodied agents. Secondly, we propose a novel method based on semantic correspondences from a self-supervised vision transformer to solve the pose estimation problem. We further re-purpose the recent CO3D dataset to present a controlled and realistic test setting. Finally, we demonstrate that all baselines for our proposed task perform poorly, and show that our method provides a six-fold improvement in average rotation accuracy at 30 degrees. Our code is available at https://github.com/applied-ai-lab/zero-shot-pose."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Label-Efficient Semantic Segmentation with Diffusion Models",
                    "abstract": "Denoising diffusion probabilistic models have recently received much research attention since they outperform alternative approaches, such as GANs, and currently provide state-of-the-art generative performance. The superior performance of diffusion models has made them an appealing tool in several applications, including inpainting, super-resolution, and semantic editing. In this paper, we demonstrate that diffusion models can also serve as an instrument for semantic segmentation, especially in the setup when labeled data is scarce. In particular, for several pretrained diffusion models, we investigate the intermediate activations from the networks that perform the Markov step of the reverse diffusion process. We show that these activations effectively capture the semantic information from an input image and appear to be excellent pixel-level representations for the segmentation problem. Based on these observations, we describe a simple segmentation method, which can work even if only a few training images are provided. Our approach significantly outperforms the existing alternatives on several datasets for the same amount of human supervision."
                },
                {
                    "title": "Generative Models as a Data Source for Multiview Representation Learning",
                    "abstract": "Generative models are now capable of producing highly realistic images that look nearly indistinguishable from the data on which they are trained. This raises the question: if we have good enough generative models, do we still need datasets? We investigate this question in the setting of learning general-purpose visual representations from a black-box generative model rather than directly from data. Given an off-the-shelf image generator without any access to its training data, we train representations from the samples output by this generator. We compare several representation learning methods that can be applied to this setting, using the latent space of the generator to generate multiple\"views\"of the same semantic content. We show that for contrastive methods, this multiview data can naturally be used to identify positive pairs (nearby in latent space) and negative pairs (far apart in latent space). We find that the resulting representations rival or even outperform those learned directly from real data, but that good performance requires care in the sampling strategy applied and the training method. Generative models can be viewed as a compressed and organized copy of a dataset, and we envision a future where more and more\"model zoos\"proliferate while datasets become increasingly unwieldy, missing, or private. This paper suggests several techniques for dealing with visual representation learning in such a future. Code is available on our project page https://ali-design.github.io/GenRep/."
                },
                {
                    "title": "Single-Stage Instance Shadow Detection with Bidirectional Relation Learning",
                    "abstract": "Instance shadow detection aims to find shadow instances paired with the objects that cast the shadows. The previous work adopts a two-stage framework to first predict shadow instances, object instances, and shadow-object associations from the region proposals, then leverage a post-processing to match the predictions to form the final shadow-object pairs. In this paper, we present a new single-stage fully-convolutional network architecture with a bidirectional relation learning module to directly learn the relations of shadow and object instances in an end-to-end manner. Compared with the prior work, our method actively explores the internal relationship between shadows and objects to learn a better pairing between them, thus improving the overall performance for instance shadow detection. We evaluate our method on the benchmark dataset for instance shadow detection, both quantitatively and visually. The experimental results demonstrate that our method clearly outperforms the state-of-the-art method."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Transformer Interpretability Beyond Attention Visualization",
                    "abstract": "Self-attention techniques, and specifically Transformers, are dominating the field of text processing and are becoming increasingly popular in computer vision classification tasks. In order to visualize the parts of the image that led to a certain classification, existing methods either rely on the obtained attention maps or employ heuristic propagation along the attention graph. In this work, we propose a novel way to compute relevancy for Transformer networks. The method assigns local relevance based on the Deep Taylor Decomposition principle and then propagates these relevancy scores through the layers. This propagation involves attention layers and skip connections, which challenge existing methods. Our solution is based on a specific formulation that is shown to maintain the total relevancy across layers. We benchmark our method on very recent visual Transformer networks, as well as on a text classification problem, and demonstrate a clear advantage over the existing explainability methods. Our code is available at: https://github.com/hila-chefer/Transformer-Explainability."
                },
                {
                    "title": "Taming Transformers for High-Resolution Image Synthesis",
                    "abstract": "Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks. In contrast to CNNs, they contain no inductive bias that prioritizes local interactions. This makes them expressive, but also computationally infeasible for long sequences, such as high-resolution images. We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images. We show how to (i) use CNNs to learn a context-rich vocabulary of image constituents, and in turn (ii) utilize transformers to efficiently model their composition within high-resolution images. Our approach is readily applied to conditional synthesis tasks, where both non-spatial information, such as object classes, and spatial information, such as segmentations, can control the generated image. In particular, we present the first results on semantically-guided synthesis of megapixel images with transformers. Project page at https://git.io/JLlvY."
                },
                {
                    "title": "StyleSpace Analysis: Disentangled Controls for StyleGAN Image Generation",
                    "abstract": "We explore and analyze the latent style space of Style-GAN2, a state-of-the-art architecture for image generation, using models pretrained on several different datasets. We first show that StyleSpace, the space of channel-wise style parameters, is significantly more disentangled than the other intermediate latent spaces explored by previous works. Next, we describe a method for discovering a large collection of style channels, each of which is shown to control a distinct visual attribute in a highly localized and dis-entangled manner. Third, we propose a simple method for identifying style channels that control a specific attribute, using a pretrained classifier or a small number of example images. Manipulation of visual attributes via these StyleSpace controls is shown to be better disentangled than via those proposed in previous works. To show this, we make use of a newly proposed Attribute Dependency metric. Finally, we demonstrate the applicability of StyleSpace controls to the manipulation of real images. Our findings pave the way to semantically meaningful and well-disentangled image manipulations via simple and intuitive interfaces."
                },
                {
                    "title": "Instance Shadow Detection",
                    "abstract": "Instance shadow detection is a brand new problem, aiming to find shadow instances paired with object instances. To approach it, we first prepare a new dataset called SOBA, named after Shadow-OBject Association, with 3,623 pairs of shadow and object instances in 1,000 photos, each with individual labeled masks. Second, we design LISA, named after Light-guided Instance Shadow-object Association, an end-to-end framework to automatically predict the shadow and object instances, together with the shadow-object associations and light direction. Then, we pair up the predicted shadow and object instances, and match them with the predicted shadow-object associations to generate the final results. In our evaluations, we formulate a new metric named the shadow-object average precision to measure the performance of our results. Further, we conducted various experiments and demonstrate our method\u2019s applicability on light direction estimation and photo editing."
                },
                {
                    "title": "Understanding Deep Networks via Extremal Perturbations and Smooth Masks",
                    "abstract": "Attribution is the problem of finding which parts of an image are the most responsible for the output of a deep neural network. An important family of attribution methods is based on measuring the effect of perturbations applied to the input image, either via exhaustive search or by finding representative perturbations via optimization. In this paper, we discuss some of the shortcomings of existing approaches to perturbation analysis and address them by introducing the concept of extremal perturbations, which are theoretically grounded and interpretable. We also introduce a number of technical innovations to compute these extremal perturbations, including a new area constraint and a parametric family of smooth perturbations, which allow us to remove all tunable weighing factors from the optimization problem. We analyze the effect of perturbations as a function of their area, demonstrating excellent sensitivity to the spatial properties of the network under stimulation. We also extend perturbation analysis to the intermediate layers of a deep neural network. This application allows us to show how compactly an image can be represented (in terms of the number of channels it requires). We also demonstrate that the consistency with which images of a given class rely on the same intermediate channel correlates well with class accuracy."
                },
                {
                    "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                    "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                },
                {
                    "title": "PlaneRCNN: 3D Plane Detection and Reconstruction From a Single Image",
                    "abstract": "This paper proposes a deep neural architecture, PlaneRCNN, that detects and reconstructs piecewise planar regions from a single RGB image. PlaneRCNN employs a variant of Mask R-CNN to detect planes with their plane parameters and segmentation masks. PlaneRCNN then refines an arbitrary number of segmentation masks with a novel loss enforcing the consistency with a nearby view during training. The paper also presents a new benchmark with more fine-grained plane segmentations in the ground-truth, in which, PlaneRCNN outperforms existing state-of-the-art methods with significant margins in the plane detection, segmentation, and reconstruction metrics. PlaneRCNN makes an important step towards robust plane extraction method, which would have immediate impact on a wide range of applications including Robotics, Augmented Reality, and Virtual Reality."
                },
                {
                    "title": "GAN Dissection: Visualizing and Understanding Generative Adversarial Networks",
                    "abstract": "Generative Adversarial Networks (GANs) have recently achieved impressive results for many real-world applications, and many GAN variants have emerged with improvements in sample quality and training stability. However, they have not been well visualized or understood. How does a GAN represent our visual world internally? What causes the artifacts in GAN results? How do architectural choices affect GAN learning? Answering such questions could enable us to develop new insights and better models. \nIn this work, we present an analytic framework to visualize and understand GANs at the unit-, object-, and scene-level. We first identify a group of interpretable units that are closely related to object concepts using a segmentation-based network dissection method. Then, we quantify the causal effect of interpretable units by measuring the ability of interventions to control objects in the output. We examine the contextual relationship between these units and their surroundings by inserting the discovered object concepts into new images. We show several practical applications enabled by our framework, from comparing internal representations across different layers, models, and datasets, to improving GANs by locating and removing artifact-causing units, to interactively manipulating objects in a scene. We provide open source interpretation tools to help researchers and practitioners better understand their GAN models."
                },
                {
                    "title": "PoseCNN: A Convolutional Neural Network for 6D Object Pose Estimation in Cluttered Scenes",
                    "abstract": "Estimating the 6D pose of known objects is important for robots to interact with the real world. The problem is challenging due to the variety of objects as well as the complexity of a scene caused by clutter and occlusions between objects. In this work, we introduce PoseCNN, a new Convolutional Neural Network for 6D object pose estimation. PoseCNN estimates the 3D translation of an object by localizing its center in the image and predicting its distance from the camera. The 3D rotation of the object is estimated by regressing to a quaternion representation. We also introduce a novel loss function that enables PoseCNN to handle symmetric objects. In addition, we contribute a large scale video dataset for 6D object pose estimation named the YCB-Video dataset. Our dataset provides accurate 6D poses of 21 objects from the YCB dataset observed in 92 videos with 133,827 frames. We conduct extensive experiments on our YCB-Video dataset and the OccludedLINEMOD dataset to show that PoseCNN is highly robust to occlusions, can handle symmetric objects, and provide accurate pose estimation using only color images as input. When using depth data to further refine the poses, our approach achieves state-of-the-art results on the challenging OccludedLINEMOD dataset. Our code and dataset are available at this https URL."
                },
                {
                    "title": "Interpretable Explanations of Black Boxes by Meaningful Perturbation",
                    "abstract": "As machine learning algorithms are increasingly applied to high impact yet high risk tasks, such as medical diagnosis or autonomous driving, it is critical that researchers can explain how such algorithms arrived at their predictions. In recent years, a number of image saliency methods have been developed to summarize where highly complex neural networks \u201clook\u201d in an image for evidence for their predictions. However, these techniques are limited by their heuristic nature and architectural constraints. In this paper, we make two main contributions: First, we propose a general framework for learning different kinds of explanations for any black box algorithm. Second, we specialise the framework to find the part of an image most responsible for a classifier decision. Unlike previous works, our method is model-agnostic and testable because it is grounded in explicit and interpretable image perturbations."
                },
                {
                    "title": "ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes",
                    "abstract": "A key requirement for leveraging supervised deep learning methods is the availability of large, labeled datasets. Unfortunately, in the context of RGB-D scene understanding, very little data is available &#x2013; current datasets cover a small range of scene views and have limited semantic annotations. To address this issue, we introduce ScanNet, an RGB-D video dataset containing 2.5M views in 1513 scenes annotated with 3D camera poses, surface reconstructions, and semantic segmentations. To collect this data, we designed an easy-to-use and scalable RGB-D capture system that includes automated surface reconstruction and crowdsourced semantic annotation. We show that using this data helps achieve state-of-the-art performance on several 3D scene understanding tasks, including 3D object classification, semantic voxel labeling, and CAD model retrieval."
                },
                {
                    "title": "3D Shape Attributes",
                    "abstract": "In this paper we investigate 3D attributes as a means to understand the shape of an object in a single image. To this end, we make a number of contributions: (i) we introduce and define a set of 3D Shape attributes, including planarity, symmetry and occupied space, (ii) we show that such properties can be successfully inferred from a single image using a Convolutional Neural Network (CNN), (iii) we introduce a 143K image dataset of sculptures with 2197 works over 242 artists for training and evaluating the CNN, (iv) we show that the 3D attributes trained on this dataset generalize to images of other (non-sculpture) object classes, and furthermore (v) we show that the CNN also provides a shape embedding that can be used to match previously unseen sculptures largely independent of viewpoint."
                },
                {
                    "title": "Learning Deep Features for Discriminative Localization",
                    "abstract": "In this work, we revisit the global average pooling layer proposed in [13], and shed light on how it explicitly enables the convolutional neural network (CNN) to have remarkable localization ability despite being trained on imagelevel labels. While this technique was previously proposed as a means for regularizing training, we find that it actually builds a generic localizable deep representation that exposes the implicit attention of CNNs on an image. Despite the apparent simplicity of global average pooling, we are able to achieve 37.1% top-5 error for object localization on ILSVRC 2014 without training on any bounding box annotation. We demonstrate in a variety of experiments that our network is able to localize the discriminative image regions despite just being trained for solving classification task1."
                },
                {
                    "title": "Understanding deep image representations by inverting them",
                    "abstract": "Image representations, from SIFT and Bag of Visual Words to Convolutional Neural Networks (CNNs), are a crucial component of almost any image understanding system. Nevertheless, our understanding of them remains limited. In this paper we conduct a direct analysis of the visual information contained in representations by asking the following question: given an encoding of an image, to which extent is it possible to reconstruct the image itself? To answer this question we contribute a general framework to invert representations. We show that this method can invert representations such as HOG more accurately than recent alternatives while being applicable to CNNs too. We then use this technique to study the inverse of recent state-of-the-art CNN image representations for the first time. Among our findings, we show that several layers in CNNs retain photographically accurate information about the image, with different degrees of geometric and photometric invariance."
                },
                {
                    "title": "Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps",
                    "abstract": "This paper addresses the visualisation of image classification models, learnt using deep Convolutional Networks (ConvNets). We consider two visualisation techniques, based on computing the gradient of the class score with respect to the input image. The first one generates an image, which maximises the class score [Erhan et al., 2009], thus visualising the notion of the class, captured by a ConvNet. The second technique computes a class saliency map, specific to a given image and class. We show that such maps can be employed for weakly supervised object segmentation using classification ConvNets. Finally, we establish the connection between the gradient-based ConvNet visualisation methods and deconvolutional networks [Zeiler et al., 2013]."
                },
                {
                    "title": "LIBSVM: A library for support vector machines",
                    "abstract": "LIBSVM is a library for Support Vector Machines (SVMs). We have been actively developing this package since the year 2000. The goal is to help users to easily apply SVM to their applications. LIBSVM has gained wide popularity in machine learning and many other areas. In this article, we present all implementation details of LIBSVM. Issues such as solving SVM optimization problems theoretical convergence multiclass classification probability estimates and parameter selection are discussed in detail."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Visualizing Higher-Layer Features of a Deep Network",
                    "abstract": "Deep architectures have demonstrated state-of-the-art results in a variety of settings, especially with vision datasets. Beyond the model definitions and the quantitative analyses, there is a need for qualitative comparisons of the solutions learned by various deep architectures. The goal of this paper is to find good qualitative interpretations of high level features represented by such models. To this end, we contrast and compare several techniques applied on Stacked Denoising Autoencoders and Deep Belief Networks, trained on several vision datasets. We show that, perhaps counter-intuitively, such interpretation is possible at the unit level, that it is simple to accomplish and that the results are consistent across various techniques. We hope that such techniques will allow researchers in deep architectures to understand more of how and why deep architectures work."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nTo what extent do large-scale pre-trained vision models, such as Stable Diffusion, understand and represent the geometric and physical properties of 3D scenes from 2D images?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it bridges the gap between 2D image processing and 3D scene understanding, which is fundamental for advancing computer vision. Understanding how well these models capture 3D properties can lead to improved model architectures and training methodologies, enhancing their applicability in real-world scenarios such as robotics, augmented reality, and autonomous driving. Furthermore, it could inspire new research directions focused on integrating 3D knowledge into existing models, ultimately leading to more robust and intelligent systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of accurately inferring 3D properties from 2D projections, as this requires a deep understanding of spatial relationships, lighting, and material properties that are not explicitly represented in 2D images. Naive approaches may fail because they do not account for the intricate interactions between these properties, leading to oversimplified models that cannot generalize well. Additionally, the lack of comprehensive datasets with ground truth annotations for various 3D properties complicates the evaluation process, making it difficult to assess model performance accurately.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on 2D tasks without adequately addressing the 3D implications of these models. Existing solutions often lack a systematic approach to evaluate the 3D understanding of vision models, and there has been a scarcity of datasets that provide the necessary annotations for 3D properties. Moreover, many studies have not explored the probing of specific layers within these models to assess their understanding of 3D features. Our approach differs by introducing a lightweight evaluation protocol that systematically probes the model's ability to represent various 3D properties, filling the gaps left by prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology consists of three key steps: First, we select a suitable real image evaluation dataset, such as the SOBA dataset, which contains ground truth annotations for properties like object-shadow relationships. Second, we perform a grid search over the layers and time steps of the Stable Diffusion model to identify the optimal features for determining the property of interest"
            }
        },
        "author_data": {
            "53538dce-5ebb-4df6-88d6-12e79566c594": {
                "pk": "53538dce-5ebb-4df6-88d6-12e79566c594",
                "name": "Guanqi Zhan",
                "collaborators": [
                    "Hao Dong",
                    "Weidi Xie",
                    "Andrew Zisserman",
                    "Baoquan Chen",
                    "Ruihai Wu",
                    "Chuanruo Ning",
                    "Yan Shen",
                    "Chuanxia Zheng",
                    "Yuxing Long",
                    "Wenzhe Cai"
                ],
                "domain": [
                    "Object Detection",
                    "Robotics",
                    "3D Vision",
                    "Image Segmentation"
                ],
                "publications": [
                    {
                        "title": "A Tri-Layer Plugin to Improve Occluded Detection",
                        "abstract": "Detecting occluded objects still remains a challenge for state-of-the-art object detectors. The objective of this work is to improve the detection for such objects, and thereby improve the overall performance of a modern object detector.   To this end we make the following four contributions: (1) We propose a simple 'plugin' module for the detection head of two-stage object detectors to improve the recall of partially occluded objects. The module predicts a tri-layer of segmentation masks for the target object, the occluder and the occludee, and by doing so is able to better predict the mask of the target object. (2) We propose a scalable pipeline for generating training data for the module by using amodal completion of existing object detection and instance segmentation training datasets to establish occlusion relationships. (3) We also establish a COCO evaluation dataset to measure the recall performance of partially occluded and separated objects. (4) We show that the plugin module inserted into a two-stage detector can boost the performance significantly, by only fine-tuning the detection head, and with additional improvements if the entire architecture is fine-tuned. COCO results are reported for Mask R-CNN with Swin-T or Swin-S backbones, and Cascade Mask R-CNN with a Swin-B backbone."
                    },
                    {
                        "title": "Amodal Ground Truth and Completion in the Wild",
                        "abstract": "This paper studies amodal image segmentation: predicting entire object segmentation masks including both visible and invisible (occluded) parts. In previous work, the amodal segmentation ground truth on real images is usually predicted by manual annotaton and thus is subjective. In contrast, we use 3D data to establish an automatic pipeline to determine authentic ground truth amodal masks for partially occluded objects in real images. This pipeline is used to construct an amodal completion evaluation benchmark, MP3D-Amodal, consisting of a variety of object categories and labels. To better handle the amodal completion task in the wild, we explore two architecture variants: a two-stage model that first infers the occluder, followed by amodal mask completion; and a one-stage model that exploits the representation power of Stable Diffusion for amodal segmentation across many categories. Without bells and whistles, our method achieves a new state-of-the-art performance on Amodal segmentation datasets that cover a large variety of objects, including COCOA and our new MP3D-Amodal dataset. The dataset, model, and code are available at https://www.robots.ox.ac.uk/~vgg/research/amodal/."
                    },
                    {
                        "title": "InstructNav: Zero-shot System for Generic Instruction Navigation in Unexplored Environment",
                        "abstract": "Enabling robots to navigate following diverse language instructions in unexplored environments is an attractive goal for human-robot interaction. However, this goal is challenging because different navigation tasks require different strategies. The scarcity of instruction navigation data hinders training an instruction navigation model with varied strategies. Therefore, previous methods are all constrained to one specific type of navigation instruction. In this work, we propose InstructNav, a generic instruction navigation system. InstructNav makes the first endeavor to handle various instruction navigation tasks without any navigation training or pre-built maps. To reach this goal, we introduce Dynamic Chain-of-Navigation (DCoN) to unify the planning process for different types of navigation instructions. Furthermore, we propose Multi-sourced Value Maps to model key elements in instruction navigation so that linguistic DCoN planning can be converted into robot actionable trajectories. With InstructNav, we complete the R2R-CE task in a zero-shot way for the first time and outperform many task-training methods. Besides, InstructNav also surpasses the previous SOTA method by 10.48% on the zero-shot Habitat ObjNav and by 86.34% on demand-driven navigation DDN. Real robot experiments on diverse indoor scenes further demonstrate our method's robustness in coping with the environment and instruction variations."
                    },
                    {
                        "title": "DLGAN: Disentangling Label-Specific Fine-Grained Features for Image Manipulation",
                        "abstract": "Recent studies have shown how disentangling images into content and feature spaces can provide controllable image translation/ manipulation. In this paper, we propose a framework to enable utilizing discrete multi-labels to control which features to be disentangled, i.e., disentangling label-specific fine-grained features for image manipulation (dubbed DLGAN). By mapping the discrete label-specific attribute features into a continuous prior distribution, we leverage the advantages of both discrete labels and reference images to achieve image manipulation in a hybrid fashion. For example, given a face image dataset (e.g., CelebA) with multiple discrete fine-grained labels, we can learn to smoothly interpolate a face image between black hair and blond hair through reference images while immediately controlling the gender and age through discrete input labels. To the best of our knowledge, this is the first work that realizes such a hybrid manipulation within a single model. More importantly, it is the first work to achieve image interpolation between two different domains without requiring continuous labels as the supervision. Qualitative and quantitative experiments demonstrate the effectiveness of the proposed method."
                    },
                    {
                        "title": "Score-PA: Score-based 3D Part Assembly",
                        "abstract": "Autonomous 3D part assembly is a challenging task in the areas of robotics and 3D computer vision. This task aims to assemble individual components into a complete shape without relying on predefined instructions. In this paper, we formulate this task from a novel generative perspective, introducing the Score-based 3D Part Assembly framework (Score-PA) for 3D part assembly. Knowing that score-based methods are typically time-consuming during the inference stage. To address this issue, we introduce a novel algorithm called the Fast Predictor-Corrector Sampler (FPC) that accelerates the sampling process within the framework. We employ various metrics to assess assembly quality and diversity, and our evaluation results demonstrate that our algorithm outperforms existing state-of-the-art approaches. We release our code at https://github.com/J-F-Cheng/Score-PA_Score-based-3D-Part-Assembly."
                    },
                    {
                        "title": "Broadcasting Support Relations Recursively from Local Dynamics for Object Retrieval in Clutters",
                        "abstract": "In our daily life, cluttered objects are everywhere, from scattered stationery and books cluttering the table to bowls and plates filling the kitchen sink. Retrieving a target object from clutters is an essential while challenging skill for robots, for the difficulty of safely manipulating an object without disturbing others, which requires the robot to plan a manipulation sequence and first move away a few other objects supported by the target object step by step. However, due to the diversity of object configurations (e.g., categories, geometries, locations and poses) and their combinations in clutters, it is difficult for a robot to accurately infer the support relations between objects faraway with various objects in between. In this paper, we study retrieving objects in complicated clutters via a novel method of recursively broadcasting the accurate local dynamics to build a support relation graph of the whole scene, which largely reduces the complexity of the support relation inference and improves the accuracy. Experiments in both simulation and the real world demonstrate the efficiency and effectiveness of our method."
                    },
                    {
                        "title": "Learning Environment-Aware Affordance for 3D Articulated Object Manipulation under Occlusions",
                        "abstract": "Perceiving and manipulating 3D articulated objects in diverse environments is essential for home-assistant robots. Recent studies have shown that point-level affordance provides actionable priors for downstream manipulation tasks. However, existing works primarily focus on single-object scenarios with homogeneous agents, overlooking the realistic constraints imposed by the environment and the agent's morphology, e.g., occlusions and physical limitations. In this paper, we propose an environment-aware affordance framework that incorporates both object-level actionable priors and environment constraints. Unlike object-centric affordance approaches, learning environment-aware affordance faces the challenge of combinatorial explosion due to the complexity of various occlusions, characterized by their quantities, geometries, positions and poses. To address this and enhance data efficiency, we introduce a novel contrastive affordance learning framework capable of training on scenes containing a single occluder and generalizing to scenes with complex occluder combinations. Experiments demonstrate the effectiveness of our proposed approach in learning affordance considering environment constraints. Project page at https://chengkaiacademycity.github.io/EnvAwareAfford/"
                    },
                    {
                        "title": "TARGO: Benchmarking Target-driven Object Grasping under Occlusions",
                        "abstract": "Recent advances in predicting 6D grasp poses from a single depth image have led to promising performance in robotic grasping. However, previous grasping models face challenges in cluttered environments where nearby objects impact the target object's grasp. In this paper, we first establish a new benchmark dataset for TARget-driven Grasping under Occlusions, named TARGO. We make the following contributions: 1) We are the first to study the occlusion level of grasping. 2) We set up an evaluation benchmark consisting of large-scale synthetic data and part of real-world data, and we evaluated five grasp models and found that even the current SOTA model suffers when the occlusion level increases, leaving grasping under occlusion still a challenge. 3) We also generate a large-scale training dataset via a scalable pipeline, which can be used to boost the performance of grasping under occlusion and generalized to the real world. 4) We further propose a transformer-based grasping model involving a shape completion module, termed TARGO-Net, which performs most robustly as occlusion increases. Our benchmark dataset can be found at https://TARGO-benchmark.github.io/."
                    },
                    {
                        "title": "Generative 3D Part Assembly via Dynamic Graph Learning",
                        "abstract": "Autonomous part assembly is a challenging yet crucial task in 3D computer vision and robotics. Analogous to buying an IKEA furniture, given a set of 3D parts that can assemble a single shape, an intelligent agent needs to perceive the 3D part geometry, reason to propose pose estimations for the input parts, and finally call robotic planning and control routines for actuation. In this paper, we focus on the pose estimation subproblem from the vision side involving geometric and relational reasoning over the input part geometry. Essentially, the task of generative 3D part assembly is to predict a 6-DoF part pose, including a rigid rotation and translation, for each input part that assembles a single 3D shape as the final output. To tackle this problem, we propose an assembly-oriented dynamic graph learning framework that leverages an iterative graph neural network as a backbone. It explicitly conducts sequential part assembly refinements in a coarse-to-fine manner, exploits a pair of part relation reasoning module and part aggregation module for dynamically adjusting both part features and their relations in the part graph. We conduct extensive experiments and quantitative comparisons to three strong baseline methods, demonstrating the effectiveness of the proposed approach."
                    }
                ]
            },
            "44214f13-270d-483d-af1b-b1db22c18fb8": {
                "pk": "44214f13-270d-483d-af1b-b1db22c18fb8",
                "name": "Chuanxia Zheng",
                "collaborators": [
                    "Tat-Jen Cham",
                    "Jianfei Cai",
                    "Andrea Vedaldi",
                    "Qianyi Wu",
                    "Yuedong Chen",
                    "Tianhao Wu",
                    "Ruining Li",
                    "Christian Rupprecht",
                    "Dinh Phung",
                    "Guanqi Zhan"
                ],
                "domain": [
                    "Image Synthesis",
                    "Depth Estimation",
                    "3D Reconstruction",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Synthesizing Photorealistic Images with Deep Generative Learning",
                        "abstract": "The goal of this thesis is to present my research contributions towards solving various visual synthesis and generation tasks, comprising image translation, image completion, and completed scene decomposition. This thesis consists of five pieces of work, each of which presents a new learning-based approach for synthesizing images with plausible content as well as visually realistic appearance. Each work demonstrates the superiority of the proposed approach on image synthesis, with some further contributing to other tasks, such as depth estimation."
                    },
                    {
                        "title": "Online Clustered Codebook",
                        "abstract": "Vector Quantisation (VQ) is experiencing a comeback in machine learning, where it is increasingly used in representation learning. However, optimizing the codevectors in existing VQ-VAE is not entirely trivial. A problem is codebook collapse, where only a small subset of codevectors receive gradients useful for their optimisation, whereas a majority of them simply ``dies off'' and is never updated or used. This limits the effectiveness of VQ for learning larger codebooks in complex computer vision tasks that require high-capacity representations. In this paper, we present a simple alternative method for online codebook learning, Clustering VQ-VAE (CVQ-VAE). Our approach selects encoded features as anchors to update the ``dead'' codevectors, while optimising the codebooks which are alive via the original loss. This strategy brings unused codevectors closer in distribution to the encoded features, increasing the likelihood of being chosen and optimized. We extensively validate the generalization capability of our quantiser on various datasets, tasks (e.g. reconstruction and generation), and architectures (e.g. VQ-VAE, VQGAN, LDM). Our CVQ-VAE can be easily integrated into the existing models with just a few lines of code."
                    },
                    {
                        "title": "Free3D: Consistent Novel View Synthesis without 3D Representation",
                        "abstract": "We introduce Free3D, a simple accurate method for monocular open-set novel view synthesis (NVS). Similar to Zero-1-to-3, we start from a pre-trained 2D image generator for generalization, and fine-tune it for NVS. Compared to other works that took a similar approach, we obtain significant improvements without resorting to an explicit 3D representation, which is slow and memory-consuming, and without training an additional network for 3D reconstruction. Our key contribution is to improve the way the target camera pose is encoded in the network, which we do by introducing a new ray conditioning normalization (RCN) layer. The latter injects pose information in the underlying 2D image generator by telling each pixel its viewing direction. We further improve multi-view consistency by using light-weight multi-view attention layers and by sharing generation noise between the different views. We train Free3D on the Objaverse dataset and demonstrate excellent generalization to new categories in new datasets, including OmniObject3D and GSO. The project page is available at https://chuanxiaz.com/free3d/."
                    },
                    {
                        "title": "T2Net: Synthetic-to-Realistic Translation for Solving Single-Image Depth Estimation Tasks",
                        "abstract": "Current methods for single-image depth estimation use training datasets with real image-depth pairs or stereo pairs, which are not easy to acquire. We propose a framework, trained on synthetic image-depth pairs and unpaired real images, that comprises an image translation network for enhancing realism of input images, followed by a depth prediction network. A key idea is having the first network act as a wide-spectrum input translator, taking in either synthetic or real images, and ideally producing minimally modified realistic images. This is done via a reconstruction loss when the training input is real, and GAN loss when synthetic, removing the need for heuristic self-regularization. The second network is trained on a task loss for synthetic image-depth pairs, with extra GAN loss to unify real and synthetic feature distributions. Importantly, the framework can be trained end-to-end, leading to good results, even surpassing early deep-learning methods that use real paired data."
                    },
                    {
                        "title": "Pluralistic Image Completion",
                        "abstract": "Most image completion methods produce only one result for each masked input, although there may be many reasonable possibilities. In this paper, we present an approach for \\textbf{pluralistic image completion} -- the task of generating multiple and diverse plausible solutions for image completion. A major challenge faced by learning-based approaches is that usually only one ground truth training instance per label. As such, sampling from conditional VAEs still leads to minimal diversity. To overcome this, we propose a novel and probabilistically principled framework with two parallel paths. One is a reconstructive path that utilizes the only one given ground truth to get prior distribution of missing parts and rebuild the original image from this distribution. The other is a generative path for which the conditional prior is coupled to the distribution obtained in the reconstructive path. Both are supported by GANs. We also introduce a new short+long term attention layer that exploits distant relations among decoder and encoder features, improving appearance consistency. When tested on datasets with buildings (Paris), faces (CelebA-HQ), and natural images (ImageNet), our method not only generated higher-quality completion results, but also with multiple and diverse plausible outputs."
                    },
                    {
                        "title": "The Spatially-Correlative Loss for Various Image Translation Tasks",
                        "abstract": "We propose a novel spatially-correlative loss that is simple, efficient and yet effective for preserving scene structure consistency while supporting large appearance changes during unpaired image-to-image (I2I) translation. Previous methods attempt this by using pixel-level cycle-consistency or feature-level matching losses, but the domain-specific nature of these losses hinder translation across large domain gaps. To address this, we exploit the spatial patterns of self-similarity as a means of defining scene structure. Our spatially-correlative loss is geared towards only capturing spatial relationships within an image rather than domain appearance. We also introduce a new self-supervised learning method to explicitly learn spatially-correlative maps for each specific translation task. We show distinct improvement over baseline models in all three modes of unpaired I2I translation: single-modal, multi-modal, and even single-image translation. This new loss can easily be integrated into existing network architectures and thus allows wide applicability."
                    },
                    {
                        "title": "PanoDiffusion: 360-degree Panorama Outpainting via Diffusion",
                        "abstract": "Generating complete 360-degree panoramas from narrow field of view images is ongoing research as omnidirectional RGB data is not readily available. Existing GAN-based approaches face some barriers to achieving higher quality output, and have poor generalization performance over different mask types. In this paper, we present our 360-degree indoor RGB-D panorama outpainting model using latent diffusion models (LDM), called PanoDiffusion. We introduce a new bi-modal latent diffusion structure that utilizes both RGB and depth panoramic data during training, which works surprisingly well to outpaint depth-free RGB images during inference. We further propose a novel technique of introducing progressive camera rotations during each diffusion denoising step, which leads to substantial improvement in achieving panorama wraparound consistency. Results show that our PanoDiffusion not only significantly outperforms state-of-the-art methods on RGB-D panorama outpainting by producing diverse well-structured results for different types of masks, but can also synthesize high-quality depth panoramas to provide realistic 3D indoor models."
                    },
                    {
                        "title": "Amodal Ground Truth and Completion in the Wild",
                        "abstract": "This paper studies amodal image segmentation: predicting entire object segmentation masks including both visible and invisible (occluded) parts. In previous work, the amodal segmentation ground truth on real images is usually predicted by manual annotaton and thus is subjective. In contrast, we use 3D data to establish an automatic pipeline to determine authentic ground truth amodal masks for partially occluded objects in real images. This pipeline is used to construct an amodal completion evaluation benchmark, MP3D-Amodal, consisting of a variety of object categories and labels. To better handle the amodal completion task in the wild, we explore two architecture variants: a two-stage model that first infers the occluder, followed by amodal mask completion; and a one-stage model that exploits the representation power of Stable Diffusion for amodal segmentation across many categories. Without bells and whistles, our method achieves a new state-of-the-art performance on Amodal segmentation datasets that cover a large variety of objects, including COCOA and our new MP3D-Amodal dataset. The dataset, model, and code are available at https://www.robots.ox.ac.uk/~vgg/research/amodal/."
                    },
                    {
                        "title": "DragAPart: Learning a Part-Level Motion Prior for Articulated Objects",
                        "abstract": "We introduce DragAPart, a method that, given an image and a set of drags as input, generates a new image of the same object that responds to the action of the drags. Differently from prior works that focused on repositioning objects, DragAPart predicts part-level interactions, such as opening and closing a drawer. We study this problem as a proxy for learning a generalist motion model, not restricted to a specific kinematic structure or object category. We start from a pre-trained image generator and fine-tune it on a new synthetic dataset, Drag-a-Move, which we introduce. Combined with a new encoding for the drags and dataset randomization, the model generalizes well to real images and different categories. Compared to prior motion-controlled generators, we demonstrate much better part-level motion understanding."
                    },
                    {
                        "title": "Puppet-Master: Scaling Interactive Video Generation as a Motion Prior for Part-Level Dynamics",
                        "abstract": "We present Puppet-Master, an interactive video generative model that can serve as a motion prior for part-level dynamics. At test time, given a single image and a sparse set of motion trajectories (i.e., drags), Puppet-Master can synthesize a video depicting realistic part-level motion faithful to the given drag interactions. This is achieved by fine-tuning a large-scale pre-trained video diffusion model, for which we propose a new conditioning architecture to inject the dragging control effectively. More importantly, we introduce the all-to-first attention mechanism, a drop-in replacement for the widely adopted spatial attention modules, which significantly improves generation quality by addressing the appearance and background issues in existing models. Unlike other motion-conditioned video generators that are trained on in-the-wild videos and mostly move an entire object, Puppet-Master is learned from Objaverse-Animation-HQ, a new dataset of curated part-level motion clips. We propose a strategy to automatically filter out sub-optimal animations and augment the synthetic renderings with meaningful motion trajectories. Puppet-Master generalizes well to real images across various categories and outperforms existing methods in a zero-shot manner on a real-world benchmark. See our project page for more results: vgg-puppetmaster.github.io."
                    },
                    {
                        "title": "Bridging Global Context Interactions for High-Fidelity Image Completion",
                        "abstract": "Bridging global context interactions correctly is important for high-fidelity image completion with large masks. Previous methods attempting this via deep or large receptive field (RF) convolutions cannot escape from the dominance of nearby interactions, which may be inferior. In this paper, we propose to treat image completion as a directionless sequence-to-sequence prediction task, and deploy a transformer to directly capture long-range dependence in the encoder. Crucially, we employ a restrictive CNN with small and non-overlapping RF for weighted token representation, which allows the transformer to explicitly model the long-range visible context relations with equal importance in all layers, without implicitly confounding neighboring tokens when larger RFs are used. To improve appearance consistency between visible and generated regions, a novel attention-aware layer (AAL) is introduced to better exploit distantly related high-frequency features. Overall, extensive experiments demonstrate superior performance compared to state-of-the-art methods on several datasets."
                    },
                    {
                        "title": "ClusteringSDF: Self-Organized Neural Implicit Surfaces for 3D Decomposition",
                        "abstract": "3D decomposition/segmentation still remains a challenge as large-scale 3D annotated data is not readily available. Contemporary approaches typically leverage 2D machine-generated segments, integrating them for 3D consistency. While the majority of these methods are based on NeRFs, they face a potential weakness that the instance/semantic embedding features derive from independent MLPs, thus preventing the segmentation network from learning the geometric details of the objects directly through radiance and density. In this paper, we propose ClusteringSDF, a novel approach to achieve both segmentation and reconstruction in 3D via the neural implicit surface representation, specifically Signal Distance Function (SDF), where the segmentation rendering is directly integrated with the volume rendering of neural implicit surfaces. Although based on ObjectSDF++, ClusteringSDF no longer requires the ground-truth segments for supervision while maintaining the capability of reconstructing individual object surfaces, but purely with the noisy and inconsistent labels from pre-trained models.As the core of ClusteringSDF, we introduce a high-efficient clustering mechanism for lifting the 2D labels to 3D and the experimental results on the challenging scenes from ScanNet and Replica datasets show that ClusteringSDF can achieve competitive performance compared against the state-of-the-art with significantly reduced training time."
                    },
                    {
                        "title": "Splatt3R: Zero-shot Gaussian Splatting from Uncalibrated Image Pairs",
                        "abstract": "In this paper, we introduce Splatt3R, a pose-free, feed-forward method for in-the-wild 3D reconstruction and novel view synthesis from stereo pairs. Given uncalibrated natural images, Splatt3R can predict 3D Gaussian Splats without requiring any camera parameters or depth information. For generalizability, we build Splatt3R upon a ``foundation'' 3D geometry reconstruction method, MASt3R, by extending it to deal with both 3D structure and appearance. Specifically, unlike the original MASt3R which reconstructs only 3D point clouds, we predict the additional Gaussian attributes required to construct a Gaussian primitive for each point. Hence, unlike other novel view synthesis methods, Splatt3R is first trained by optimizing the 3D point cloud's geometry loss, and then a novel view synthesis objective. By doing this, we avoid the local minima present in training 3D Gaussian Splats from stereo views. We also propose a novel loss masking strategy that we empirically find is critical for strong performance on extrapolated viewpoints. We train Splatt3R on the ScanNet++ dataset and demonstrate excellent generalisation to uncalibrated, in-the-wild images. Splatt3R can reconstruct scenes at 4FPS at 512 x 512 resolution, and the resultant splats can be rendered in real-time."
                    },
                    {
                        "title": "MoVQ: Modulating Quantized Vectors for High-Fidelity Image Generation",
                        "abstract": "Although two-stage Vector Quantized (VQ) generative models allow for synthesizing high-fidelity and high-resolution images, their quantization operator encodes similar patches within an image into the same index, resulting in a repeated artifact for similar adjacent regions using existing decoder architectures. To address this issue, we propose to incorporate the spatially conditional normalization to modulate the quantized vectors so as to insert spatially variant information to the embedded index maps, encouraging the decoder to generate more photorealistic images. Moreover, we use multichannel quantization to increase the recombination capability of the discrete codes without increasing the cost of model and codebook. Additionally, to generate discrete tokens at the second stage, we adopt a Masked Generative Image Transformer (MaskGIT) to learn an underlying prior distribution in the compressed latent space, which is much faster than the conventional autoregressive model. Experiments on two benchmark datasets demonstrate that our proposed modulated VQGAN is able to greatly improve the reconstructed image quality as well as provide high-fidelity image generation."
                    },
                    {
                        "title": "Visiting the Invisible: Layer-by-Layer Completed Scene Decomposition",
                        "abstract": "Existing scene understanding systems mainly focus on recognizing the visible parts of a scene, ignoring the intact appearance of physical objects in the real-world. Concurrently, image completion has aimed to create plausible appearance for the invisible regions, but requires a manual mask as input. In this work, we propose a higher-level scene understanding system to tackle both visible and invisible parts of objects and backgrounds in a given scene. Particularly, we built a system to decompose a scene into individual objects, infer their underlying occlusion relationships, and even automatically learn which parts of the objects are occluded that need to be completed. In order to disentangle the occluded relationships of all objects in a complex scene, we use the fact that the front object without being occluded is easy to be identified, detected, and segmented. Our system interleaves the two tasks of instance segmentation and scene completion through multiple iterations, solving for objects layer-by-layer. We first provide a thorough experiment using a new realistically rendered dataset with ground-truths for all invisible regions. To bridge the domain gap to real imagery where ground-truths are unavailable, we then train another model with the pseudo-ground-truths generated from our trained synthesis model. We demonstrate results on a wide variety of datasets and show significant improvement over the state-of-the-art."
                    },
                    {
                        "title": "Object-Compositional Neural Implicit Surfaces",
                        "abstract": "The neural implicit representation has shown its effectiveness in novel view synthesis and high-quality 3D reconstruction from multi-view images. However, most approaches focus on holistic scene representation yet ignore individual objects inside it, thus limiting potential downstream applications. In order to learn object-compositional representation, a few works incorporate the 2D semantic map as a cue in training to grasp the difference between objects. But they neglect the strong connections between object geometry and instance semantic information, which leads to inaccurate modeling of individual instance. This paper proposes a novel framework, ObjectSDF, to build an object-compositional neural implicit representation with high fidelity in 3D reconstruction and object representation. Observing the ambiguity of conventional volume rendering pipelines, we model the scene by combining the Signed Distance Functions (SDF) of individual object to exert explicit surface constraint. The key in distinguishing different instances is to revisit the strong association between an individual object's SDF and semantic label. Particularly, we convert the semantic information to a function of object SDF and develop a unified and compact representation for scene and objects. Experimental results show the superiority of ObjectSDF framework in representing both the holistic object-compositional scene and the individual instances. Code can be found at https://qianyiwu.github.io/objectsdf/"
                    },
                    {
                        "title": "One More Step: A Versatile Plug-and-Play Module for Rectifying Diffusion Schedule Flaws and Enhancing Low-Frequency Controls",
                        "abstract": "It is well known that many open-released foundational diffusion models have difficulty in generating images that substantially depart from average brightness, despite such images being present in the training data. This is due to an inconsistency: while denoising starts from pure Gaussian noise during inference, the training noise schedule retains residual data even in the final timestep distribution, due to difficulties in numerical conditioning in mainstream formulation, leading to unintended bias during inference. To mitigate this issue, certain $\\epsilon$-prediction models are combined with an ad-hoc offset-noise methodology. In parallel, some contemporary models have adopted zero-terminal SNR noise schedules together with $\\mathbf{v}$-prediction, which necessitate major alterations to pre-trained models. However, such changes risk destabilizing a large multitude of community-driven applications anchored on these pre-trained models. In light of this, our investigation revisits the fundamental causes, leading to our proposal of an innovative and principled remedy, called One More Step (OMS). By integrating a compact network and incorporating an additional simple yet effective step during inference, OMS elevates image fidelity and harmonizes the dichotomy between training and inference, while preserving original model parameters. Once trained, various pre-trained diffusion models with the same latent domain can share the same OMS module."
                    },
                    {
                        "title": "Sem2NeRF: Converting Single-View Semantic Masks to Neural Radiance Fields",
                        "abstract": "Image translation and manipulation have gain increasing attention along with the rapid development of deep generative models. Although existing approaches have brought impressive results, they mainly operated in 2D space. In light of recent advances in NeRF-based 3D-aware generative models, we introduce a new task, Semantic-to-NeRF translation, that aims to reconstruct a 3D scene modelled by NeRF, conditioned on one single-view semantic mask as input. To kick-off this novel task, we propose the Sem2NeRF framework. In particular, Sem2NeRF addresses the highly challenging task by encoding the semantic mask into the latent code that controls the 3D scene representation of a pre-trained decoder. To further improve the accuracy of the mapping, we integrate a new region-aware learning strategy into the design of both the encoder and the decoder. We verify the efficacy of the proposed Sem2NeRF and demonstrate that it outperforms several strong baselines on two benchmark datasets. Code and video are available at https://donydchen.github.io/sem2nerf/"
                    },
                    {
                        "title": "Explicit Correspondence Matching for Generalizable Neural Radiance Fields",
                        "abstract": "We present a new generalizable NeRF method that is able to directly generalize to new unseen scenarios and perform novel view synthesis with as few as two source views. The key to our approach lies in the explicitly modeled correspondence matching information, so as to provide the geometry prior to the prediction of NeRF color and density for volume rendering. The explicit correspondence matching is quantified with the cosine similarity between image features sampled at the 2D projections of a 3D point on different views, which is able to provide reliable cues about the surface geometry. Unlike previous methods where image features are extracted independently for each view, we consider modeling the cross-view interactions via Transformer cross-attention, which greatly improves the feature matching quality. Our method achieves state-of-the-art results on different evaluation settings, with the experiments showing a strong correlation between our learned cosine feature similarity and volume density, demonstrating the effectiveness and superiority of our proposed method. Code is at https://github.com/donydchen/matchnerf"
                    }
                ]
            },
            "cc2b310e-86db-4cea-b155-2bcdc5dd44c2": {
                "pk": "cc2b310e-86db-4cea-b155-2bcdc5dd44c2",
                "name": "Weidi Xie",
                "collaborators": [
                    "Andrew Zisserman",
                    "Tengda Han",
                    "Junyu Xie",
                    "Zihang Lai",
                    "Charig Yang",
                    "Erika Lu",
                    "Shangzhe Di",
                    "Yibin Yan",
                    "Li Shen",
                    "Jeffrey Byrne"
                ],
                "domain": [
                    "Computer Vision",
                    "Self-Supervised Learning",
                    "Video Analysis",
                    "Face Recognition"
                ],
                "publications": [
                    {
                        "title": "Multicolumn Networks for Face Recognition",
                        "abstract": "The objective of this work is set-based face recognition, i.e. to decide if two sets of images of a face are of the same person or not. Conventionally, the set-wise feature descriptor is computed as an average of the descriptors from individual face images within the set. In this paper, we design a neural network architecture that learns to aggregate based on both \"visual\" quality (resolution, illumination), and \"content\" quality (relative importance for discriminative classification). To this end, we propose a Multicolumn Network (MN) that takes a set of images (the number in the set can vary) as input, and learns to compute a fix-sized feature descriptor for the entire set. To encourage high-quality representations, each individual input image is first weighted by its \"visual\" quality, determined by a self-quality assessment module, and followed by a dynamic recalibration based on \"content\" qualities relative to the other images within the set. Both of these qualities are learnt implicitly during training for set-wise classification. Comparing with the previous state-of-the-art architectures trained with the same dataset (VGGFace2), our Multicolumn Networks show an improvement of between 2-6% on the IARPA IJB face recognition benchmarks, and exceed the state of the art for all methods on these benchmarks."
                    },
                    {
                        "title": "Self-supervised Learning for Video Correspondence Flow",
                        "abstract": "The objective of this paper is self-supervised learning of feature embeddings that are suitable for matching correspondences along the videos, which we term correspondence flow. By leveraging the natural spatial-temporal coherence in videos, we propose to train a ``pointer'' that reconstructs a target frame by copying pixels from a reference frame.   We make the following contributions: First, we introduce a simple information bottleneck that forces the model to learn robust features for correspondence matching, and prevent it from learning trivial solutions, \\eg matching based on low-level colour information. Second, to tackle the challenges from tracker drifting, due to complex object deformations, illumination changes and occlusions, we propose to train a recursive model over long temporal windows with scheduled sampling and cycle consistency. Third, we achieve state-of-the-art performance on DAVIS 2017 video segmentation and JHMDB keypoint tracking tasks, outperforming all previous self-supervised learning approaches by a significant margin. Fourth, in order to shed light on the potential of self-supervised learning on the task of video correspondence flow, we probe the upper bound by training on additional data, \\ie more diverse videos, further demonstrating significant improvements on video segmentation."
                    },
                    {
                        "title": "Grounded Question-Answering in Long Egocentric Videos",
                        "abstract": "Existing approaches to video understanding, mainly designed for short videos from a third-person perspective, are limited in their applicability in certain fields, such as robotics. In this paper, we delve into open-ended question-answering (QA) in long, egocentric videos, which allows individuals or robots to inquire about their own past visual experiences. This task presents unique challenges, including the complexity of temporally grounding queries within extensive video content, the high resource demands for precise data annotation, and the inherent difficulty of evaluating open-ended answers due to their ambiguous nature. Our proposed approach tackles these challenges by (i) integrating query grounding and answering within a unified model to reduce error propagation; (ii) employing large language models for efficient and scalable data synthesis; and (iii) introducing a close-ended QA task for evaluation, to manage answer ambiguity. Extensive experiments demonstrate the effectiveness of our method, which also achieves state-of-the-art performance on the QaEgo4D and Ego4D-NLQ benchmarks. Code, data, and models are available at https://github.com/Becomebright/GroundVQA."
                    },
                    {
                        "title": "EchoSight: Advancing Visual-Language Models with Wiki Knowledge",
                        "abstract": "Knowledge-based Visual Question Answering (KVQA) tasks require answering questions about images using extensive background knowledge. Despite significant advancements, generative models often struggle with these tasks due to the limited integration of external knowledge. In this paper, we introduce EchoSight, a novel multimodal Retrieval-Augmented Generation (RAG) framework that enables large language models (LLMs) to answer visual questions requiring fine-grained encyclopedic knowledge. To strive for high-performing retrieval, EchoSight first searches wiki articles by using visual-only information, subsequently, these candidate articles are further reranked according to their relevance to the combined text-image query. This approach significantly improves the integration of multimodal knowledge, leading to enhanced retrieval outcomes and more accurate VQA responses. Our experimental results on the Encyclopedic VQA and InfoSeek datasets demonstrate that EchoSight establishes new state-of-the-art results in knowledge-based VQA, achieving an accuracy of 41.8% on Encyclopedic VQA and 31.3% on InfoSeek."
                    },
                    {
                        "title": "Segmenting Moving Objects via an Object-Centric Layered Representation",
                        "abstract": "The objective of this paper is a model that is able to discover, track and segment multiple moving objects in a video. We make four contributions: First, we introduce an object-centric segmentation model with a depth-ordered layer representation. This is implemented using a variant of the transformer architecture that ingests optical flow, where each query vector specifies an object and its layer for the entire video. The model can effectively discover multiple moving objects and handle mutual occlusions; Second, we introduce a scalable pipeline for generating multi-object synthetic training data via layer compositions, that is used to train the proposed model, significantly reducing the requirements for labour-intensive annotations, and supporting Sim2Real generalisation; Third, we conduct thorough ablation studies, showing that the model is able to learn object permanence and temporal shape consistency, and is able to predict amodal segmentation masks; Fourth, we evaluate our model, trained only on synthetic data, on standard video segmentation benchmarks, DAVIS, MoCA, SegTrack, FBMS-59, and achieve state-of-the-art performance among existing methods that do not rely on any manual annotations. With test-time adaptation, we observe further performance boosts."
                    },
                    {
                        "title": "Appearance-Based Refinement for Object-Centric Motion Segmentation",
                        "abstract": "The goal of this paper is to discover, segment, and track independently moving objects in complex visual scenes. Previous approaches have explored the use of optical flow for motion segmentation, leading to imperfect predictions due to partial motion, background distraction, and object articulations and interactions. To address this issue, we introduce an appearance-based refinement method that leverages temporal consistency in video streams to correct inaccurate flow-based proposals. Our approach involves a sequence-level selection mechanism that identifies accurate flow-predicted masks as exemplars, and an object-centric architecture that refines problematic masks based on exemplar information. The model is pre-trained on synthetic data and then adapted to real-world videos in a self-supervised manner, eliminating the need for human annotations. Its performance is evaluated on multiple video segmentation benchmarks, including DAVIS, YouTubeVOS, SegTrackv2, and FBMS-59. We achieve competitive performance on single-object segmentation, while significantly outperforming existing models on the more challenging problem of multi-object segmentation. Finally, we investigate the benefits of using our model as a prompt for the per-frame Segment Anything Model."
                    },
                    {
                        "title": "Moving Object Segmentation: All You Need Is SAM (and Flow)",
                        "abstract": "The objective of this paper is motion segmentation -- discovering and segmenting the moving objects in a video. This is a much studied area with numerous careful,and sometimes complex, approaches and training schemes including: self-supervised learning, learning from synthetic datasets, object-centric representations, amodal representations, and many more. Our interest in this paper is to determine if the Segment Anything model (SAM) can contribute to this task. We investigate two models for combining SAM with optical flow that harness the segmentation power of SAM with the ability of flow to discover and group moving objects. In the first model, we adapt SAM to take optical flow, rather than RGB, as an input. In the second, SAM takes RGB as an input, and flow is used as a segmentation prompt. These surprisingly simple methods, without any further modifications, outperform all previous approaches by a considerable margin in both single and multi-object benchmarks. We also extend these frame-level segmentations to sequence-level segmentations that maintain object identity. Again, this simple model outperforms previous methods on multiple video object segmentation benchmarks."
                    },
                    {
                        "title": "Comparator Networks",
                        "abstract": "The objective of this work is set-based verification, e.g. to decide if two sets of images of a face are of the same person or not. The traditional approach to this problem is to learn to generate a feature vector per image, aggregate them into one vector to represent the set, and then compute the cosine similarity between sets. Instead, we design a neural network architecture that can directly learn set-wise verification. Our contributions are: (i) We propose a Deep Comparator Network (DCN) that can ingest a pair of sets (each may contain a variable number of images) as inputs, and compute a similarity between the pair--this involves attending to multiple discriminative local regions (landmarks), and comparing local descriptors between pairs of faces; (ii) To encourage high-quality representations for each set, internal competition is introduced for recalibration based on the landmark score; (iii) Inspired by image retrieval, a novel hard sample mining regime is proposed to control the sampling process, such that the DCN is complementary to the standard image classification models. Evaluations on the IARPA Janus face recognition benchmarks show that the comparator networks outperform the previous state-of-the-art results by a large margin."
                    },
                    {
                        "title": "Turbo Training with Token Dropout",
                        "abstract": "The objective of this paper is an efficient training method for video tasks. We make three contributions: (1) We propose Turbo training, a simple and versatile training paradigm for Transformers on multiple video tasks. (2) We illustrate the advantages of Turbo training on action classification, video-language representation learning, and long-video activity classification, showing that Turbo training can largely maintain competitive performance while achieving almost 4X speed-up and significantly less memory consumption. (3) Turbo training enables long-schedule video-language training and end-to-end long-video training, delivering competitive or superior performance than previous works, which were infeasible to train under limited resources."
                    },
                    {
                        "title": "Class-Agnostic Counting",
                        "abstract": "Nearly all existing counting methods are designed for a specific object class. Our work, however, aims to create a counting model able to count any class of object. To achieve this goal, we formulate counting as a matching problem, enabling us to exploit the image self-similarity property that naturally exists in object counting problems. We make the following three contributions: first, a Generic Matching Network (GMN) architecture that can potentially count any object in a class-agnostic manner; second, by reformulating the counting problem as one of matching objects, we can take advantage of the abundance of video data labeled for tracking, which contains natural repetitions suitable for training a counting model. Such data enables us to train the GMN. Third, to customize the GMN to different user requirements, an adapter module is used to specialize the model with minimal effort, i.e. using a few labeled examples, and adapting only a small fraction of the trained parameters. This is a form of few-shot learning, which is practical for domains where labels are limited due to requiring expert knowledge (e.g. microbiology). We demonstrate the flexibility of our method on a diverse set of existing counting benchmarks: specifically cells, cars, and human crowds. The model achieves competitive performance on cell and crowd counting datasets, and surpasses the state-of-the-art on the car dataset using only three training images. When training on the entire dataset, the proposed method outperforms all previous methods by a large margin."
                    },
                    {
                        "title": "Inducing Predictive Uncertainty Estimation for Face Recognition",
                        "abstract": "Knowing when an output can be trusted is critical for reliably using face recognition systems. While there has been enormous effort in recent research on improving face verification performance, understanding when a model's predictions should or should not be trusted has received far less attention. Our goal is to assign a confidence score for a face image that reflects its quality in terms of recognizable information. To this end, we propose a method for generating image quality training data automatically from 'mated-pairs' of face images, and use the generated data to train a lightweight Predictive Confidence Network, termed as PCNet, for estimating the confidence score of a face image. We systematically evaluate the usefulness of PCNet with its error versus reject performance, and demonstrate that it can be universally paired with and improve the robustness of any verification model. We describe three use cases on the public IJB-C face verification benchmark: (i) to improve 1:1 image-based verification error rates by rejecting low-quality face images; (ii) to improve quality score based fusion performance on the 1:1 set-based verification benchmark; and (iii) its use as a quality measure for selecting high quality (unblurred, good lighting, more frontal) faces from a collection, e.g. for automatic enrolment or display."
                    },
                    {
                        "title": "Diagnosing Human-object Interaction Detectors",
                        "abstract": "We have witnessed significant progress in human-object interaction (HOI) detection. The reliance on mAP (mean Average Precision) scores as a summary metric, however, does not provide sufficient insight into the nuances of model performance (e.g., why one model is better than another), which can hinder further innovation in this field. To address this issue, in this paper, we introduce a diagnosis toolbox to provide detailed quantitative break-down analysis of HOI detection models, inspired by the success of object detection diagnosis toolboxes. We first conduct holistic investigations in the pipeline of HOI detection. By defining a set of errors and the oracles to fix each of them, we can have a quantitative analysis of the significance of different errors according to the mAP improvement obtained from fixing each error. We then delve into two sub-tasks of HOI detection: human-object pair detection and interaction classification, respectively. For the first detection task, we compute the coverage of ground-truth human-object pairs as well as the noisiness level in the detection results. For the second classification task, we measure a model's performance of differentiating positive and negative detection results and also classifying the actual interactions when the human-object pairs are correctly detected. We analyze eight state-of-the-art HOI detection models and provide valuable diagnosis insights to foster future research. For instance, our diagnosis shows that state-of-the-art model RLIPv2 outperforms others mainly because it significantly improves the multi-label interaction classification accuracy. Our toolbox is applicable for different methods across different datasets and available at https://github.com/neu-vi/Diag-HOI."
                    },
                    {
                        "title": "Memory-augmented Dense Predictive Coding for Video Representation Learning",
                        "abstract": "The objective of this paper is self-supervised learning from video, in particular for representations for action recognition. We make the following contributions: (i) We propose a new architecture and learning framework Memory-augmented Dense Predictive Coding (MemDPC) for the task. It is trained with a predictive attention mechanism over the set of compressed memories, such that any future states can always be constructed by a convex combination of the condense representations, allowing to make multiple hypotheses efficiently. (ii) We investigate visual-only self-supervised video representation learning from RGB frames, or from unsupervised optical flow, or both. (iii) We thoroughly evaluate the quality of learnt representation on four different downstream tasks: action recognition, video retrieval, learning with scarce annotations, and unintentional action classification. In all cases, we demonstrate state-of-the-art or comparable performance over other approaches with orders of magnitude fewer training data."
                    },
                    {
                        "title": "Geometry-Aware Video Object Detection for Static Cameras",
                        "abstract": "In this paper we propose a geometry-aware model for video object detection. Specifically, we consider the setting that cameras can be well approximated as static, e.g. in video surveillance scenarios, and scene pseudo depth maps can therefore be inferred easily from the object scale on the image plane. We make the following contributions: First, we extend the recent anchor-free detector (CornerNet [17]) to video object detections. In order to exploit the spatial-temporal information while maintaining high efficiency, the proposed model accepts video clips as input, and only makes predictions for the starting and the ending frames, i.e. heatmaps of object bounding box corners and the corresponding embeddings for grouping. Second, to tackle the challenge from scale variations in object detection, scene geometry information, e.g. derived depth maps, is explicitly incorporated into deep networks for multi-scale feature selection and for the network prediction. Third, we validate the proposed architectures on an autonomous driving dataset generated from the Carla simulator [5], and on a real dataset for human detection (DukeMTMC dataset [28]). When comparing with the existing competitive single-stage or two-stage detectors, the proposed geometry-aware spatio-temporal network achieves significantly better results."
                    },
                    {
                        "title": "Video Representation Learning by Dense Predictive Coding",
                        "abstract": "The objective of this paper is self-supervised learning of spatio-temporal embeddings from video, suitable for human action recognition. We make three contributions: First, we introduce the Dense Predictive Coding (DPC) framework for self-supervised representation learning on videos. This learns a dense encoding of spatio-temporal blocks by recurrently predicting future representations; Second, we propose a curriculum training scheme to predict further into the future with progressively less temporal context. This encourages the model to only encode slowly varying spatial-temporal signals, therefore leading to semantic representations; Third, we evaluate the approach by first training the DPC model on the Kinetics-400 dataset with self-supervised learning, and then finetuning the representation on a downstream task, i.e. action recognition. With single stream (RGB only), DPC pretrained representations achieve state-of-the-art self-supervised performance on both UCF101(75.7% top1 acc) and HMDB51(35.7% top1 acc), outperforming all previous learning methods by a significant margin, and approaching the performance of a baseline pre-trained on ImageNet."
                    },
                    {
                        "title": "MAST: A Memory-Augmented Self-supervised Tracker",
                        "abstract": "Recent interest in self-supervised dense tracking has yielded rapid progress, but performance still remains far from supervised methods. We propose a dense tracking model trained on videos without any annotations that surpasses previous self-supervised methods on existing benchmarks by a significant margin (+15%), and achieves performance comparable to supervised methods. In this paper, we first reassess the traditional choices used for self-supervised training and reconstruction loss by conducting thorough experiments that finally elucidate the optimal choices. Second, we further improve on existing methods by augmenting our architecture with a crucial memory component. Third, we benchmark on large-scale semi-supervised video object segmentation(aka. dense tracking), and propose a new metric: generalizability. Our first two contributions yield a self-supervised network that for the first time is competitive with supervised methods on standard evaluation metrics of dense tracking. When measuring generalizability, we show self-supervised approaches are actually superior to the majority of supervised methods. We believe this new generalizability metric can better capture the real-world use-cases for dense tracking, and will spur new interest in this research direction."
                    },
                    {
                        "title": "Self-supervised Co-training for Video Representation Learning",
                        "abstract": "The objective of this paper is visual-only self-supervised video representation learning. We make the following contributions: (i) we investigate the benefit of adding semantic-class positives to instance-based Info Noise Contrastive Estimation (InfoNCE) training, showing that this form of supervised contrastive learning leads to a clear improvement in performance; (ii) we propose a novel self-supervised co-training scheme to improve the popular infoNCE loss, exploiting the complementary information from different views, RGB streams and optical flow, of the same data source by using one view to obtain positive class samples for the other; (iii) we thoroughly evaluate the quality of the learnt representation on two different downstream tasks: action recognition and video retrieval. In both cases, the proposed approach demonstrates state-of-the-art or comparable performance with other self-supervised approaches, whilst being significantly more efficient to train, i.e. requiring far less training data to achieve similar performance."
                    },
                    {
                        "title": "Betrayed by Motion: Camouflaged Object Discovery via Motion Segmentation",
                        "abstract": "The objective of this paper is to design a computational architecture that discovers camouflaged objects in videos, specifically by exploiting motion information to perform object segmentation. We make the following three contributions: (i) We propose a novel architecture that consists of two essential components for breaking camouflage, namely, a differentiable registration module to align consecutive frames based on the background, which effectively emphasises the object boundary in the difference image, and a motion segmentation module with memory that discovers the moving objects, while maintaining the object permanence even when motion is absent at some point. (ii) We collect the first large-scale Moving Camouflaged Animals (MoCA) video dataset, which consists of over 140 clips across a diverse range of animals (67 categories). (iii) We demonstrate the effectiveness of the proposed model on MoCA, and achieve competitive performance on the unsupervised segmentation protocol on DAVIS2016 by only relying on motion."
                    },
                    {
                        "title": "VGGSound: A Large-scale Audio-Visual Dataset",
                        "abstract": "Our goal is to collect a large-scale audio-visual dataset with low label noise from videos in the wild using computer vision techniques. The resulting dataset can be used for training and evaluating audio recognition models. We make three contributions. First, we propose a scalable pipeline based on computer vision techniques to create an audio dataset from open-source media. Our pipeline involves obtaining videos from YouTube; using image classification algorithms to localize audio-visual correspondence; and filtering out ambient noise using audio verification. Second, we use this pipeline to curate the VGGSound dataset consisting of more than 210k videos for 310 audio classes. Third, we investigate various Convolutional Neural Network~(CNN) architectures and aggregation approaches to establish audio recognition baselines for our new dataset. Compared to existing audio datasets, VGGSound ensures audio-visual correspondence and is collected under unconstrained conditions. Code and the dataset are available at http://www.robots.ox.ac.uk/~vgg/data/vggsound/"
                    },
                    {
                        "title": "Smooth-AP: Smoothing the Path Towards Large-Scale Image Retrieval",
                        "abstract": "Optimising a ranking-based metric, such as Average Precision (AP), is notoriously challenging due to the fact that it is non-differentiable, and hence cannot be optimised directly using gradient-descent methods. To this end, we introduce an objective that optimises instead a smoothed approximation of AP, coined Smooth-AP. Smooth-AP is a plug-and-play objective function that allows for end-to-end training of deep networks with a simple and elegant implementation. We also present an analysis for why directly optimising the ranking based metric of AP offers benefits over other deep metric learning losses. We apply Smooth-AP to standard retrieval benchmarks: Stanford Online products and VehicleID, and also evaluate on larger-scale datasets: INaturalist for fine-grained category retrieval, and VGGFace2 and IJB-C for face retrieval. In all cases, we improve the performance over the state-of-the-art, especially for larger-scale datasets, thus demonstrating the effectiveness and scalability of Smooth-AP to real-world scenarios."
                    }
                ]
            },
            "71751837-9dbb-474e-a386-f024e117358d": {
                "pk": "71751837-9dbb-474e-a386-f024e117358d",
                "name": "Andrew Zisserman",
                "collaborators": [
                    "Andrew Brown",
                    "Relja Arandjelovi\u0107",
                    "Arsha Nagrani",
                    "Weidi Xie",
                    "Vicky Kalogeiton",
                    "Karen Simonyan",
                    "Carl Doersch",
                    "Olivia Wiles",
                    "Ernesto Coto",
                    "Max Bain"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Object Detection",
                    "Video Analysis"
                ],
                "publications": [
                    {
                        "title": "Object Discovery with a Copy-Pasting GAN",
                        "abstract": "We tackle the problem of object discovery, where objects are segmented for a given input image, and the system is trained without using any direct supervision whatsoever. A novel copy-pasting GAN framework is proposed, where the generator learns to discover an object in one image by compositing it into another image such that the discriminator cannot tell that the resulting image is fake. After carefully addressing subtle issues, such as preventing the generator from `cheating', this game results in the generator learning to select objects, as copy-pasting objects is most likely to fool the discriminator. The system is shown to work well on four very different datasets, including large object appearance variations in challenging cluttered backgrounds."
                    },
                    {
                        "title": "Automated Video Labelling: Identifying Faces by Corroborative Evidence",
                        "abstract": "We present a method for automatically labelling all faces in video archives, such as TV broadcasts, by combining multiple evidence sources and multiple modalities (visual and audio). We target the problem of ever-growing online video archives, where an effective, scalable indexing solution cannot require a user to provide manual annotation or supervision. To this end, we make three key contributions: (1) We provide a novel, simple, method for determining if a person is famous or not using image-search engines. In turn this enables a face-identity model to be built reliably and robustly, and used for high precision automatic labelling; (2) We show that even for less-famous people, image-search engines can then be used for corroborative evidence to accurately label faces that are named in the scene or the speech; (3) Finally, we quantitatively demonstrate the benefits of our approach on different video domains and test settings, such as TV shows and news broadcasts. Our method works across three disparate datasets without any explicit domain adaptation, and sets new state-of-the-art results on all the public benchmarks."
                    },
                    {
                        "title": "Condensed Movies: Story Based Retrieval with Contextual Embeddings",
                        "abstract": "Our objective in this work is long range understanding of the narrative structure of movies. Instead of considering the entire movie, we propose to learn from the `key scenes' of the movie, providing a condensed look at the full storyline. To this end, we make the following three contributions: (i) We create the Condensed Movies Dataset (CMD) consisting of the key scenes from over 3K movies: each key scene is accompanied by a high level semantic description of the scene, character face-tracks, and metadata about the movie. The dataset is scalable, obtained automatically from YouTube, and is freely available for anybody to download and use. It is also an order of magnitude larger than existing movie datasets in the number of movies; (ii) We provide a deep network baseline for text-to-video retrieval on our dataset, combining character, speech and visual cues into a single video embedding; and finally (iii) We demonstrate how the addition of context from other video clips improves retrieval performance."
                    },
                    {
                        "title": "Smooth-AP: Smoothing the Path Towards Large-Scale Image Retrieval",
                        "abstract": "Optimising a ranking-based metric, such as Average Precision (AP), is notoriously challenging due to the fact that it is non-differentiable, and hence cannot be optimised directly using gradient-descent methods. To this end, we introduce an objective that optimises instead a smoothed approximation of AP, coined Smooth-AP. Smooth-AP is a plug-and-play objective function that allows for end-to-end training of deep networks with a simple and elegant implementation. We also present an analysis for why directly optimising the ranking based metric of AP offers benefits over other deep metric learning losses. We apply Smooth-AP to standard retrieval benchmarks: Stanford Online products and VehicleID, and also evaluate on larger-scale datasets: INaturalist for fine-grained category retrieval, and VGGFace2 and IJB-C for face retrieval. In all cases, we improve the performance over the state-of-the-art, especially for larger-scale datasets, thus demonstrating the effectiveness and scalability of Smooth-AP to real-world scenarios."
                    },
                    {
                        "title": "Face, Body, Voice: Video Person-Clustering with Multiple Modalities",
                        "abstract": "The objective of this work is person-clustering in videos -- grouping characters according to their identity. Previous methods focus on the narrower task of face-clustering, and for the most part ignore other cues such as the person's voice, their overall appearance (hair, clothes, posture), and the editing structure of the videos. Similarly, most current datasets evaluate only the task of face-clustering, rather than person-clustering. This limits their applicability to downstream applications such as story understanding which require person-level, rather than only face-level, reasoning. In this paper we make contributions to address both these deficiencies: first, we introduce a Multi-Modal High-Precision Clustering algorithm for person-clustering in videos using cues from several modalities (face, body, and voice). Second, we introduce a Video Person-Clustering dataset, for evaluating multi-modal person-clustering. It contains body-tracks for each annotated character, face-tracks when visible, and voice-tracks when speaking, with their associated features. The dataset is by far the largest of its kind, and covers films and TV-shows representing a wide range of demographics. Finally, we show the effectiveness of using multiple modalities for person-clustering, explore the use of this new broad task for story understanding through character co-occurrences, and achieve a new state of the art on all available datasets for face and person-clustering."
                    },
                    {
                        "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                        "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                    },
                    {
                        "title": "Recurrent Human Pose Estimation",
                        "abstract": "We propose a novel ConvNet model for predicting 2D human body poses in an image. The model regresses a heatmap representation for each body keypoint, and is able to learn and represent both the part appearances and the context of the part configuration. We make the following three contributions: (i) an architecture combining a feed forward module with a recurrent module, where the recurrent module can be run iteratively to improve the performance, (ii) the model can be trained end-to-end and from scratch, with auxiliary losses incorporated to improve performance, (iii) we investigate whether keypoint visibility can also be predicted. The model is evaluated on two benchmark datasets. The result is a simple architecture that achieves performance on par with the state of the art, but without the complexity of a graphical model stage (or layers)."
                    },
                    {
                        "title": "Look, Listen and Learn",
                        "abstract": "We consider the question: what can be learnt by looking at and listening to a large number of unlabelled videos? There is a valuable, but so far untapped, source of information contained in the video itself -- the correspondence between the visual and the audio streams, and we introduce a novel \"Audio-Visual Correspondence\" learning task that makes use of this. Training visual and audio networks from scratch, without any additional supervision other than the raw unconstrained videos themselves, is shown to successfully solve this task, and, more interestingly, result in good visual and audio representations. These features set the new state-of-the-art on two sound classification benchmarks, and perform on par with the state-of-the-art self-supervised approaches on ImageNet classification. We also demonstrate that the network is able to localize objects in both modalities, as well as perform fine-grained recognition tasks."
                    },
                    {
                        "title": "Multi-task Self-Supervised Visual Learning",
                        "abstract": "We investigate methods for combining multiple self-supervised tasks--i.e., supervised tasks where data can be collected without manual labeling--in order to train a single visual representation. First, we provide an apples-to-apples comparison of four different self-supervised tasks using the very deep ResNet-101 architecture. We then combine tasks to jointly train a network. We also explore lasso regularization to encourage the network to factorize the information in its representation, and methods for \"harmonizing\" network inputs in order to learn a more unified representation. We evaluate all methods on ImageNet classification, PASCAL VOC detection, and NYU depth prediction. Our results show that deeper networks work better, and that combining tasks--even via a naive multi-head architecture--always improves performance. Our best joint network nearly matches the PASCAL performance of a model pre-trained on ImageNet classification, and matches the ImageNet network on NYU depth prediction."
                    },
                    {
                        "title": "From Benedict Cumberbatch to Sherlock Holmes: Character Identification in TV series without a Script",
                        "abstract": "The goal of this paper is the automatic identification of characters in TV and feature film material. In contrast to standard approaches to this task, which rely on the weak supervision afforded by transcripts and subtitles, we propose a new method requiring only a cast list. This list is used to obtain images of actors from freely available sources on the web, providing a form of partial supervision for this task. In using images of actors to recognize characters, we make the following three contributions: (i) We demonstrate that an automated semi-supervised learning approach is able to adapt from the actor's face to the character's face, including the face context of the hair; (ii) By building voice models for every character, we provide a bridge between frontal faces (for which there is plenty of actor-level supervision) and profile (for which there is very little or none); and (iii) by combining face context and speaker identification, we are able to identify characters with partially occluded faces and extreme facial poses. Results are presented on the TV series 'Sherlock' and the feature film 'Casablanca'. We achieve the state-of-the-art on the Casablanca benchmark, surpassing previous methods that have used the stronger supervision available from transcripts."
                    },
                    {
                        "title": "Multicolumn Networks for Face Recognition",
                        "abstract": "The objective of this work is set-based face recognition, i.e. to decide if two sets of images of a face are of the same person or not. Conventionally, the set-wise feature descriptor is computed as an average of the descriptors from individual face images within the set. In this paper, we design a neural network architecture that learns to aggregate based on both \"visual\" quality (resolution, illumination), and \"content\" quality (relative importance for discriminative classification). To this end, we propose a Multicolumn Network (MN) that takes a set of images (the number in the set can vary) as input, and learns to compute a fix-sized feature descriptor for the entire set. To encourage high-quality representations, each individual input image is first weighted by its \"visual\" quality, determined by a self-quality assessment module, and followed by a dynamic recalibration based on \"content\" qualities relative to the other images within the set. Both of these qualities are learnt implicitly during training for set-wise classification. Comparing with the previous state-of-the-art architectures trained with the same dataset (VGGFace2), our Multicolumn Networks show an improvement of between 2-6% on the IARPA IJB face recognition benchmarks, and exceed the state of the art for all methods on these benchmarks."
                    },
                    {
                        "title": "3D Surface Reconstruction by Pointillism",
                        "abstract": "The objective of this work is to infer the 3D shape of an object from a single image. We use sculptures as our training and test bed, as these have great variety in shape and appearance.   To achieve this we build on the success of multiple view geometry (MVG) which is able to accurately provide correspondences between images of 3D objects under varying viewpoint and illumination conditions, and make the following contributions: first, we introduce a new loss function that can harness image-to-image correspondences to provide a supervisory signal to train a deep network to infer a depth map. The network is trained end-to-end by differentiating through the camera. Second, we develop a processing pipeline to automatically generate a large scale multi-view set of correspondences for training the network. Finally, we demonstrate that we can indeed obtain a depth map of a novel object from a single image for a variety of sculptures with varying shape/texture, and that the network generalises at test time to new domains (e.g. synthetic images)."
                    },
                    {
                        "title": "Two-Stream Convolutional Networks for Action Recognition in Videos",
                        "abstract": "We investigate architectures of discriminatively trained deep Convolutional Networks (ConvNets) for action recognition in video. The challenge is to capture the complementary information on appearance from still frames and motion between frames. We also aim to generalise the best performing hand-crafted features within a data-driven learning framework.   Our contribution is three-fold. First, we propose a two-stream ConvNet architecture which incorporates spatial and temporal networks. Second, we demonstrate that a ConvNet trained on multi-frame dense optical flow is able to achieve very good performance in spite of limited training data. Finally, we show that multi-task learning, applied to two different action classification datasets, can be used to increase the amount of training data and improve the performance on both.   Our architecture is trained and evaluated on the standard video actions benchmarks of UCF-101 and HMDB-51, where it is competitive with the state of the art. It also exceeds by a large margin previous attempts to use deep nets for video classification."
                    },
                    {
                        "title": "SilNet : Single- and Multi-View Reconstruction by Learning from Silhouettes",
                        "abstract": "The objective of this paper is 3D shape understanding from single and multiple images. To this end, we introduce a new deep-learning architecture and loss function, SilNet, that can handle multiple views in an order-agnostic manner. The architecture is fully convolutional, and for training we use a proxy task of silhouette prediction, rather than directly learning a mapping from 2D images to 3D shape as has been the target in most recent work.   We demonstrate that with the SilNet architecture there is generalisation over the number of views -- for example, SilNet trained on 2 views can be used with 3 or 4 views at test-time; and performance improves with more views.   We introduce two new synthetics datasets: a blobby object dataset useful for pre-training, and a challenging and realistic sculpture dataset; and demonstrate on these datasets that SilNet has indeed learnt 3D shape. Finally, we show that SilNet exceeds the state of the art on the ShapeNet benchmark dataset, and use SilNet to generate novel views of the sculpture dataset."
                    },
                    {
                        "title": "The VIA Annotation Software for Images, Audio and Video",
                        "abstract": "In this paper, we introduce a simple and standalone manual annotation tool for images, audio and video: the VGG Image Annotator (VIA). This is a light weight, standalone and offline software package that does not require any installation or setup and runs solely in a web browser. The VIA software allows human annotators to define and describe spatial regions in images or video frames, and temporal segments in audio or video. These manual annotations can be exported to plain text data formats such as JSON and CSV and therefore are amenable to further processing by other software tools. VIA also supports collaborative annotation of a large dataset by a group of human annotators. The BSD open source license of this software allows it to be used in any academic project or commercial application."
                    },
                    {
                        "title": "A Geometric Approach to Obtain a Bird's Eye View from an Image",
                        "abstract": "The objective of this paper is to rectify any monocular image by computing a homography matrix that transforms it to a bird's eye (overhead) view.   We make the following contributions: (i) we show that the homography matrix can be parameterised with only four parameters that specify the horizon line and the vertical vanishing point, or only two if the field of view or focal length is known; (ii) We introduce a novel representation for the geometry of a line or point (which can be at infinity) that is suitable for regression with a convolutional neural network (CNN); (iii) We introduce a large synthetic image dataset with ground truth for the orthogonal vanishing points, that can be used for training a CNN to predict these geometric entities; and finally (iv) We achieve state-of-the-art results on horizon detection, with 74.52% AUC on the Horizon Lines in the Wild dataset. Our method is fast and robust, and can be used to remove perspective distortion from videos in real time."
                    },
                    {
                        "title": "Sim2real transfer learning for 3D human pose estimation: motion to the rescue",
                        "abstract": "Synthetic visual data can provide practically infinite diversity and rich labels, while avoiding ethical issues with privacy and bias. However, for many tasks, current models trained on synthetic data generalize poorly to real data. The task of 3D human pose estimation is a particularly interesting example of this sim2real problem, because learning-based approaches perform reasonably well given real training data, yet labeled 3D poses are extremely difficult to obtain in the wild, limiting scalability. In this paper, we show that standard neural-network approaches, which perform poorly when trained on synthetic RGB images, can perform well when the data is pre-processed to extract cues about the person's motion, notably as optical flow and the motion of 2D keypoints. Therefore, our results suggest that motion can be a simple way to bridge a sim2real gap when video is available. We evaluate on the 3D Poses in the Wild dataset, the most challenging modern benchmark for 3D pose estimation, where we show full 3D mesh recovery that is on par with state-of-the-art methods trained on real 3D sequences, despite training only on synthetic humans from the SURREAL dataset."
                    },
                    {
                        "title": "NeRF in detail: Learning to sample for view synthesis",
                        "abstract": "Neural radiance fields (NeRF) methods have demonstrated impressive novel view synthesis performance. The core approach is to render individual rays by querying a neural network at points sampled along the ray to obtain the density and colour of the sampled points, and integrating this information using the rendering equation. Since dense sampling is computationally prohibitive, a common solution is to perform coarse-to-fine sampling.   In this work we address a clear limitation of the vanilla coarse-to-fine approach -- that it is based on a heuristic and not trained end-to-end for the task at hand. We introduce a differentiable module that learns to propose samples and their importance for the fine network, and consider and compare multiple alternatives for its neural architecture. Training the proposal module from scratch can be unstable due to lack of supervision, so an effective pre-training strategy is also put forward. The approach, named `NeRF in detail' (NeRF-ID), achieves superior view synthesis quality over NeRF and the state-of-the-art on the synthetic Blender benchmark and on par or better performance on the real LLFF-NeRF scenes. Furthermore, by leveraging the predicted sample importance, a 25% saving in computation can be achieved without significantly sacrificing the rendering quality."
                    },
                    {
                        "title": "The Change You Want to See",
                        "abstract": "We live in a dynamic world where things change all the time. Given two images of the same scene, being able to automatically detect the changes in them has practical applications in a variety of domains. In this paper, we tackle the change detection problem with the goal of detecting \"object-level\" changes in an image pair despite differences in their viewpoint and illumination. To this end, we make the following four contributions: (i) we propose a scalable methodology for obtaining a large-scale change detection training dataset by leveraging existing object segmentation benchmarks; (ii) we introduce a co-attention based novel architecture that is able to implicitly determine correspondences between an image pair and find changes in the form of bounding box predictions; (iii) we contribute four evaluation datasets that cover a variety of domains and transformations, including synthetic image changes, real surveillance images of a 3D scene, and synthetic 3D scenes with camera motion; (iv) we evaluate our model on these four datasets and demonstrate zero-shot and beyond training transformation generalization."
                    },
                    {
                        "title": "End-to-end Tracking with a Multi-query Transformer",
                        "abstract": "Multiple-object tracking (MOT) is a challenging task that requires simultaneous reasoning about location, appearance, and identity of the objects in the scene over time. Our aim in this paper is to move beyond tracking-by-detection approaches, that perform well on datasets where the object classes are known, to class-agnostic tracking that performs well also for unknown object classes.To this end, we make the following three contributions: first, we introduce {\\em semantic detector queries} that enable an object to be localized by specifying its approximate position, or its appearance, or both; second, we use these queries within an auto-regressive framework for tracking, and propose a multi-query tracking transformer (\\textit{MQT}) model for simultaneous tracking and appearance-based re-identification (reID) based on the transformer architecture with deformable attention. This formulation allows the tracker to operate in a class-agnostic manner, and the model can be trained end-to-end; finally, we demonstrate that \\textit{MQT} performs competitively on standard MOT benchmarks, outperforms all baselines on generalised-MOT, and generalises well to a much harder tracking problems such as tracking any object on the TAO dataset."
                    }
                ]
            }
        }
    },
    "2410.03558": {
        "paper_data": {
            "title": "Not All Diffusion Model Activations Have Been Evaluated as Discriminative Features",
            "url": "http://arxiv.org/abs/2410.03558v3",
            "arxiv_id": "2410.03558",
            "authors": [
                "Benyuan Meng",
                "Qianqian Xu",
                "Zitai Wang",
                "Xiaochun Cao",
                "Qingming Huang"
            ],
            "abstract": "Diffusion models are initially designed for image generation. Recent research shows that the internal signals within their backbones, named activations, can also serve as dense features for various discriminative tasks such as semantic segmentation. Given numerous activations, selecting a small yet effective subset poses a fundamental problem. To this end, the early study of this field performs a large-scale quantitative comparison of the discriminative ability of the activations. However, we find that many potential activations have not been evaluated, such as the queries and keys used to compute attention scores. Moreover, recent advancements in diffusion architectures bring many new activations, such as those within embedded ViT modules. Both combined, activation selection remains unresolved but overlooked. To tackle this issue, this paper takes a further step with a much broader range of activations evaluated. Considering the significant increase in activations, a full-scale quantitative comparison is no longer operational. Instead, we seek to understand the properties of these activations, such that the activations that are clearly inferior can be filtered out in advance via simple qualitative evaluation. After careful analysis, we discover three properties universal among diffusion models, enabling this study to go beyond specific models. On top of this, we present effective feature selection solutions for several popular diffusion models. Finally, the experiments across multiple discriminative tasks validate the superiority of our method over the SOTA competitors. Our code is available at https://github.com/Darkbblue/generic-diffusion-feature.",
            "introduction": "   1 Introduction  Diffusion models\u00a0[21, 10, 37, 36] are powerful generative models that progressively reconstruct images from Gaussian noises through a series of denoising steps. Typically, a U-Net\u00a0[38] is trained as the noise predictor backbone to perform denoising. Recently, the impressive generative capability inspires the application to discriminative tasks such as semantic segmentation\u00a0[52, 58] or semantic correspondence\u00a0[29, 56]. In this direction, diffusion feature is one simple yet effective approach, where the intermediate signals, named activations, are extracted from the pre-trained diffusion U-Net as dense features\u00a0[2, 58, 56, 29, 53, 13, 14].   The complex architecture of the diffusion U-Net provides many activations that can serve as features. However, these activations inherently vary in quality, inducing significant performance gaps on discrimination. Hence, selecting a small yet effective subset from these activations has become a fundamental problem. In the early stage, Baranchuk et. al\u00a0[2] perform a large-scale quantitative comparison among activations within Guided Diffusion\u00a0[10]. Later, the activations they select are followed by most studies in this field, pursuing other improvements [29, 56, 58, 52, 60, 53, 24, 27].   However, we find that this fundamental issue is far from solved. On one hand, Baranchuk et. al\u00a0[2] only consider activations between neighboring modules that comprise the main residuals. This means that many potential activations are excluded from the candidate pool, such as the queries and keys in the self-attention blocks. Moreover, recent developments of diffusion architecture, such as cross-attention\u00a0[37] or embedded deep vision transformers (ViTs)\u00a0[36], have introduced additional types of activations. Hence, as shown in upper Figure\u00a01, only a small fraction of potential activations in modern diffusion models have been evaluated for their discriminative ability, which might hinder future work in this direction. For example, lower Figure\u00a01 shows that Stable Diffusion XL (SDXL)\u00a0[36], which is more advanced than Stable Diffusion v1.5 (SDv1.5)\u00a0[37], fails to achieve better performance with the feature selection solution proposed in [2].   Figure 1: Prior arts only consider a small fraction of potential activations in diffusion models. As a result, more advanced diffusion architecture fails to achieve better performance (SDXL v.s. SDv1.5). In contrast, we consider a broader range of candidate activations. To facilitate the quantitative comparison, we first make a comprehensive and generalizable analysis to qualitatively filter out many candidates in advance. On top of this, our method achieves superior performance (75.2 PCK@0.1).    In this paper, we revisit the fundamental problem of feature selection, considering a more comprehensive candidate pool of activations. Due to its large volume, a full-scale quantitative comparison is no longer operational, urging us to modify the previous research methodology in [2]. As illustrated in Figure\u00a01, instead of a direct quantitative comparison, we first explore the properties of diffusion U-Nets. These properties allow us to qualitatively and efficiently filter out many activations that are highly likely to be sub-optimal, shrinking the candidate pool for the quantitative comparison. More importantly, we find these properties are universal among diffusion models, making it possible to generalize our findings to more models beyond those covered in this paper.   Specifically, the properties we find, which are distinct to existing knowledge of model properties [1, 38], exactly correspond to three top-to-bottom levels of the diffusion U-Net: (i) Diffusion noises at the macro level: the diffusion process induces a",
            "references": [
                {
                    "title": "DiffSeg: Towards Detecting Diffusion-Based Inpainting Attacks Using Multi-Feature Segmentation",
                    "abstract": "With the advancements made in deep learning over the past years, creating convincing media manipulations has become easy and accessible than ever before. In particular, diffusion models such as Stable-Diffusion allow users to synthesize realistic images based on a given text input. Apart from synthesizing entirely new images, diffusion models can also be used to make edits to images using inpainting. To combat the spread of disinformation and illegal content created with diffusion-based inpainting, this paper presents a new detection method based on multi-feature segmentation. Apart from information derived from the raw pixel values, noise, and frequency information are also exploited to detect and localize regions that have been subject to editing. Evaluation results strongly suggest that the proposed method can achieve high mIoU and AUC scores, outperforming state-of-the-art methods, even for syntheses generated by unseen diffusion models, or highly compressed images."
                },
                {
                    "title": "Diffusion Models are Certifiably Robust Classifiers",
                    "abstract": "Generative learning, recognized for its effective modeling of data distributions, offers inherent advantages in handling out-of-distribution instances, especially for enhancing robustness to adversarial attacks. Among these, diffusion classifiers, utilizing powerful diffusion models, have demonstrated superior empirical robustness. However, a comprehensive theoretical understanding of their robustness is still lacking, raising concerns about their vulnerability to stronger future attacks. In this study, we prove that diffusion classifiers possess $O(1)$ Lipschitzness, and establish their certified robustness, demonstrating their inherent resilience. To achieve non-constant Lipschitzness, thereby obtaining much tighter certified robustness, we generalize diffusion classifiers to classify Gaussian-corrupted data. This involves deriving the evidence lower bounds (ELBOs) for these distributions, approximating the likelihood using the ELBO, and calculating classification probabilities via Bayes' theorem. Experimental results show the superior certified robustness of these Noised Diffusion Classifiers (NDCs). Notably, we achieve over 80% and 70% certified robustness on CIFAR-10 under adversarial perturbations with \\(\\ell_2\\) norms less than 0.25 and 0.5, respectively, using a single off-the-shelf diffusion model without any additional data."
                },
                {
                    "title": "VideoCrafter2: Overcoming Data Limitations for High-Quality Video Diffusion Models",
                    "abstract": "Text-to-video generation aims to produce a video based on a given prompt. Recently, several commercial video models have been able to generate plausible videos with mini-mal noise, excellent details, and high aesthetic scores. However, these models rely on large-scale, well-filtered, high-quality videos that are not accessible to the community. Many existing research works, which train models using the low-quality WebVid-10M dataset, struggle to generate high-quality videos because the models are optimized to fit WebVid-10M. In this work, we explore the training scheme of video models extended from Stable Diffusion and investigate the feasibility of leveraging low-quality videos and synthesized high-quality images to obtain a high-quality video model. We first analyze the connection between the spatial and temporal modules of video models and the distribution shift to low-quality videos. We observe that full training of all modules results in a stronger coupling between spatial and temporal modules than only training temporal modules. Based on this stronger coupling, we shift the distribution to higher quality without motion degradation by finetuning spatial modules with high-quality images, resulting in a generic high-quality video model. Evaluations are conducted to demonstrate the superiority of the proposed method, particularly in picture quality, motion, and concept composition."
                },
                {
                    "title": "Harnessing Diffusion Models for Visual Perception with Meta Prompts",
                    "abstract": "The issue of generative pretraining for vision models has persisted as a long-standing conundrum. At present, the text-to-image (T2I) diffusion model demonstrates remarkable proficiency in generating high-definition images matching textual inputs, a feat made possible through its pre-training on large-scale image-text pairs. This leads to a natural inquiry: can diffusion models be utilized to tackle visual perception tasks? In this paper, we propose a simple yet effective scheme to harness a diffusion model for visual perception tasks. Our key insight is to introduce learnable embeddings (meta prompts) to the pre-trained diffusion models to extract proper features for perception. The effect of meta prompts are two-fold. First, as a direct replacement of the text embeddings in the T2I models, it can activate task-relevant features during feature extraction. Second, it will be used to re-arrange the extracted features to ensures that the model focuses on the most pertinent features for the task on hand. Additionally, we design a recurrent refinement training strategy that fully leverages the property of diffusion models, thereby yielding stronger visual features. Extensive experiments across various benchmarks validate the effectiveness of our approach. Our approach achieves new performance records in depth estimation tasks on NYU depth V2 and KITTI, and in semantic segmentation task on CityScapes. Concurrently, the proposed method attains results comparable to the current state-of-the-art in semantic segmentation on ADE20K and pose estimation on COCO datasets, further exemplifying its robustness and versatility."
                },
                {
                    "title": "Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation",
                    "abstract": "Monocular depth estimation is a fundamental computer vision task. Recovering 3D depth from a single image is geometrically ill-posed and requires scene understanding, so it is not surprising that the rise of deep learning has led to a breakthrough. The impressive progress of monocular depth estimators has mirrored the growth in model capacity, from relatively modest CNNs to large Transformer architectures. Still, monocular depth estimators tend to struggle when presented with images with unfamiliar content and layout, since their knowledge of the visual world is restricted by the data seen during training, and challenged by zero-shot generalization to new domains. This motivates us to explore whether the extensive priors captured in recent generative diffusion models can enable better, more generalizable depth estimation. We introduce Marigold, a method for affine-invariant monocular depth estimation that is derived from Stable Diffusion and retains its rich prior knowledge. The estimator can be fine-tuned in a couple of days on a single GPU using only synthetic training data. It delivers state-of-the-art performance across a wide range of datasets, including over 20% performance gains in specific cases. Project page: https://marigoldmonodepth.github.io."
                },
                {
                    "title": "Do text-free diffusion models learn discriminative visual representations?",
                    "abstract": "While many unsupervised learning models focus on one family of tasks, either generative or discriminative, we explore the possibility of a unified representation learner: a model which addresses both families of tasks simultaneously. We identify diffusion models, a state-of-the-art method for generative tasks, as a prime candidate. Such models involve training a U-Net to iteratively predict and remove noise, and the resulting model can synthesize high-fidelity, diverse, novel images. We find that the intermediate feature maps of the U-Net are diverse, discriminative feature representations. We propose a novel attention mechanism for pooling feature maps and further leverage this mechanism as DifFormer, a transformer feature fusion of features from different diffusion U-Net blocks and noise steps. We also develop DifFeed, a novel feedback mechanism tailored to diffusion. We find that diffusion models are better than GANs, and, with our fusion and feedback mechanisms, can compete with state-of-the-art unsupervised image representation learning methods for discriminative tasks - image classification with full and semi-supervision, transfer for fine-grained classification, object detection and segmentation, and semantic segmentation. Our project website (https://mgwillia.github.io/diffssl/) and code (https://github.com/soumik-kanad/diffssl) are available publicly."
                },
                {
                    "title": "Diffusion 3D Features (Diff3F) Decorating Untextured Shapes with Distilled Semantic Features",
                    "abstract": "We present Diff3F as a simple, robust, and class-agnostic feature descriptor that can be computed for untextured input shapes (meshes or point clouds). Our method distills diffusion features from image foundational models onto input shapes. Specifically, we use the input shapes to produce depth and normal maps as guidance for conditional image synthesis. In the process, we produce (diffusion) features in 2D that we subsequently lift and aggregate on the original surface. Our key observation is that even if the conditional image generations obtained from multi-view rendering of the input shapes are inconsistent, the associated image features are robust and, hence, can be directly aggregated across views. This produces semantic features on the input shapes, without requiring additional data or training. We perform extensive experiments on multiple benchmarks (SHREC'19, SHREC'20, FAUST, and TOSCA) and demonstrate that our features, being semantic instead of geometric, produce reliable correspondence across both isometric and non-isometrically related shape families. Code is available at https://github.com/niladridutt/Diffusion-3D-Features."
                },
                {
                    "title": "Analyzing Vision Transformers for Image Classification in Class Embedding Space",
                    "abstract": "Despite the growing use of transformer models in computer vision, a mechanistic understanding of these networks is still needed. This work introduces a method to reverse-engineer Vision Transformers trained to solve image classification tasks. Inspired by previous research in NLP, we demonstrate how the inner representations at any level of the hierarchy can be projected onto the learned class embedding space to uncover how these networks build categorical representations for their predictions. We use our framework to show how image tokens develop class-specific representations that depend on attention mechanisms and contextual information, and give insights on how self-attention and MLP layers differentially contribute to this categorical composition. We additionally demonstrate that this method (1) can be used to determine the parts of an image that would be important for detecting the class of interest, and (2) exhibits significant advantages over traditional linear probing approaches. Taken together, our results position our proposed framework as a powerful tool for mechanistic interpretability and explainability research."
                },
                {
                    "title": "SD4Match: Learning to Prompt Stable Diffusion Model for Semantic Matching",
                    "abstract": "In this paper, we address the challenge of matching semantically similar keypoints across image pairs. Existing research indicates that the intermediate output of the UNet within the Stable Diffusion (SD) can serve as robust image feature maps for such a matching task. We demonstrate that by employing a basic prompt tuning technique, the inherent potential of Stable Diffusion can be harnessed, resulting in a significant enhancement in accuracy over previous approaches. We further introduce a novel conditional prompting module that conditions the prompt on the local details of the input image pairs, leading to a further improvement in performance. We designate our approach as SD4Match, short for Stable Diffusion for Semantic Matching. Comprehensive evaluations of SD4Match on the PF-Pascal, PF-Willow, and SPair-71k datasets show that it sets new benchmarks in accuracy across all these datasets. Particularly, SD4Match outperforms the previous state-of-the-art by a margin of 12 percentage points on the challenging SPair-71k dataset. Code is available at the project website: https://sd4match.active.vision/."
                },
                {
                    "title": "A Unified Generalization Analysis of Re-Weighting and Logit-Adjustment for Imbalanced Learning",
                    "abstract": "Real-world datasets are typically imbalanced in the sense that only a few classes have numerous samples, while many classes are associated with only a few samples. As a result, a na\\\"ive ERM learning process will be biased towards the majority classes, making it difficult to generalize to the minority classes. To address this issue, one simple but effective approach is to modify the loss function to emphasize the learning on minority classes, such as re-weighting the losses or adjusting the logits via class-dependent terms. However, existing generalization analysis of such losses is still coarse-grained and fragmented, failing to explain some empirical results. To bridge this gap, we propose a novel technique named data-dependent contraction to capture how these modified losses handle different classes. On top of this technique, a fine-grained generalization bound is established for imbalanced learning, which helps reveal the mystery of re-weighting and logit-adjustment in a unified manner. Furthermore, a principled learning algorithm is developed based on the theoretical insights. Finally, the empirical results on benchmark datasets not only validate the theoretical results but also demonstrate the effectiveness of the proposed method."
                },
                {
                    "title": "Text-Image Alignment for Diffusion-Based Perception",
                    "abstract": "Diffusion models are generative models with impressive text-to-image synthesis capabilities and have spurred a new wave of creative methods for classical machine learning tasks. However, the best way to harness the perceptual knowledge of these generative models for visual tasks is still an open question. Specifically, it is unclear how to use the prompting interface when applying diffusion backbones to vision tasks. We find that automatically generated captions can improve text-image alignment and significantly enhance a model's cross-attention maps, leading to better perceptual performance. Our approach improves upon the current state-of-the-art (SOTA) in diffusionbased semantic segmentation on ADE20K and the current overall SOTA for depth estimation on NYUv2. Furthermore, our method generalizes to the cross-domain setting. We use model personalization and caption modifications to align our model to the target domain and find improvements over unaligned baselines. Our crossdomain object detection model, trained on Pascal VOC, achieves SOTA results on Watercolor2K. Our cross-domain segmentation method, trained on Cityscapes, achieves SOTA results on Dark Zurich-val and Nighttime Driving. Project page: vision.caltech.edu/TADP/ Code page: github.com/damaggu/TADP"
                },
                {
                    "title": "Diffuse, Attend, and Segment: Unsupervised Zero-Shot Segmentation using Stable Diffusion",
                    "abstract": "Producing quality segmentation masks for images is a fundamental problem in computer vision. Recent research has explored large-scale supervised training to enable zero- shot transfer segmentation on virtually any image style and unsupervised training to enable segmentation without dense annotations. However, constructing a model capable of segmenting anything in a zero-shot manner without any anno-tations is still challenging. In this paper, we propose to uti-lize the self-attention layers in stable diffusion models to achieve this goal because the pre-trained stable diffusion model has learned inherent concepts of objects within its attention layers. Specifically, we introduce a simple yet ef-fective iterative merging process based on measuring KL divergence among attention maps to merge them into valid segmentation masks. The proposed method does not re-quire any training or language dependency to extract qual-ity segmentation for any images. On COCO-Stuff-27, our method surpasses the prior unsupervised zero-shot trans-fer SOTA method by an absolute 26% in pixel accuracy and 17% in mean IoU. The project page is at https://sites.google.com/view/diffseg/home.11Georgia Institute of Technology"
                },
                {
                    "title": "Diffusion Model as Representation Learner",
                    "abstract": "Diffusion Probabilistic Models (DPMs) have recently demonstrated impressive results on various generative tasks. Despite its promises, the learned representations of pre-trained DPMs, however, have not been fully understood. In this paper, we conduct an in-depth investigation of the representation power of DPMs, and propose a novel knowledge transfer method that leverages the knowledge acquired by generative DPMs for recognition tasks. Our study begins by examining the feature space of DPMs, revealing that DPMs are inherently denoising autoencoders that balance the representation learning with regularizing model capacity. To this end, we introduce a novel knowledge transfer paradigm named RepFusion. Our paradigm extracts representations at different time steps from off-the-shelf DPMs and dynamically employs them as supervision for student networks, in which the optimal time is determined through reinforcement learning. We evaluate our approach on several image classification, semantic segmentation, and landmark detection benchmarks, and demonstrate that it outperforms state-of-the-art methods. Our results uncover the potential of DPMs as a powerful tool for representation learning and provide insights into the usefulness of generative models beyond sample generation. The code is available at https://github.com/Adamdad/Repfusion."
                },
                {
                    "title": "Free-ATM: Exploring Unsupervised Learning on Diffusion-Generated Images with Free Attention Masks",
                    "abstract": "Despite the rapid advancement of unsupervised learning in visual representation, it requires training on large-scale datasets that demand costly data collection, and pose additional challenges due to concerns regarding data privacy. Recently, synthetic images generated by text-to-image diffusion models, have shown great potential for benefiting image recognition. Although promising, there has been inadequate exploration dedicated to unsupervised learning on diffusion-generated images. To address this, we start by uncovering that diffusion models' cross-attention layers inherently provide annotation-free attention masks aligned with corresponding text inputs on generated images. We then investigate the problems of three prevalent unsupervised learning techniques ( i.e., contrastive learning, masked modeling, and vision-language pretraining) and introduce customized solutions by fully exploiting the aforementioned free attention masks. Our approach is validated through extensive experiments that show consistent improvements in baseline models across various downstream tasks, including image classification, detection, segmentation, and image-text retrieval. By utilizing our method, it is possible to close the performance gap between unsupervised pretraining on synthetic data and real-world scenarios."
                },
                {
                    "title": "DatasetDM: Synthesizing Data with Perception Annotations Using Diffusion Models",
                    "abstract": "Current deep networks are very data-hungry and benefit from training on largescale datasets, which are often time-consuming to collect and annotate. By contrast, synthetic data can be generated infinitely using generative models such as DALL-E and diffusion models, with minimal effort and cost. In this paper, we present DatasetDM, a generic dataset generation model that can produce diverse synthetic images and the corresponding high-quality perception annotations (e.g., segmentation masks, and depth). Our method builds upon the pre-trained diffusion model and extends text-guided image synthesis to perception data generation. We show that the rich latent code of the diffusion model can be effectively decoded as accurate perception annotations using a decoder module. Training the decoder only needs less than 1% (around 100 images) manually labeled images, enabling the generation of an infinitely large annotated dataset. Then these synthetic data can be used for training various perception models for downstream tasks. To showcase the power of the proposed approach, we generate datasets with rich dense pixel-wise labels for a wide range of downstream tasks, including semantic segmentation, instance segmentation, and depth estimation. Notably, it achieves 1) state-of-the-art results on semantic segmentation and instance segmentation; 2) significantly more robust on domain generalization than using the real data alone; and state-of-the-art results in zero-shot segmentation setting; and 3) flexibility for efficient application and novel task composition (e.g., image editing). The project website and code can be found at https://weijiawu.github.io/DatasetDM_page/ and https://github.com/showlab/DatasetDM, respectively"
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Classification",
                    "abstract": "While many unsupervised learning models focus on one family of tasks, either generative or discriminative, we explore the possibility of a unified representation learner: a model which uses a single pre-training stage to address both families of tasks simultaneously. We identify diffusion models as a prime candidate. Diffusion models have risen to prominence as a state-of-the-art method for image generation, denoising, inpainting, super-resolution, manipulation, etc. Such models involve training a U-Net to iteratively predict and remove noise, and the resulting model can synthesize high fidelity, diverse, novel images. The U-Net architecture, as a convolution-based architecture, generates a diverse set of feature representations in the form of intermediate feature maps. We present our findings that these embeddings are useful beyond the noise prediction task, as they contain discriminative information and can also be leveraged for classification. We explore optimal methods for extracting and using these embeddings for classification tasks, demonstrating promising results on the ImageNet classification task. We find that with careful feature selection and pooling, diffusion models outperform comparable generative-discriminative methods such as BigBiGAN for classification tasks. We investigate diffusion models in the transfer learning regime, examining their performance on several fine-grained visual classification datasets. We compare these embeddings to those generated by competing architectures and pre-trainings for classification tasks."
                },
                {
                    "title": "Prompting Diffusion Representations for Cross-Domain Semantic Segmentation",
                    "abstract": "While originally designed for image generation, diffusion models have recently shown to provide excellent pretrained feature representations for semantic segmentation. Intrigued by this result, we set out to explore how well diffusion-pretrained representations generalize to new domains, a crucial ability for any representation. We find that diffusion-pretraining achieves extraordinary domain generalization results for semantic segmentation, outperforming both supervised and self-supervised backbone networks. Motivated by this, we investigate how to utilize the model's unique ability of taking an input prompt, in order to further enhance its cross-domain performance. We introduce a scene prompt and a prompt randomization strategy to help further disentangle the domain-invariant information when training the segmentation head. Moreover, we propose a simple but highly effective approach for test-time domain adaptation, based on learning a scene prompt on the target domain in an unsupervised manner. Extensive experiments conducted on four synthetic-to-real and clear-to-adverse weather benchmarks demonstrate the effectiveness of our approaches. Without resorting to any complex techniques, such as image translation, augmentation, or rare-class sampling, we set a new state-of-the-art on all benchmarks. Our implementation will be publicly available at \\url{https://github.com/ETHRuiGong/PTDiffSeg}."
                },
                {
                    "title": "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis",
                    "abstract": "We present SDXL, a latent diffusion model for text-to-image synthesis. Compared to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet backbone: The increase of model parameters is mainly due to more attention blocks and a larger cross-attention context as SDXL uses a second text encoder. We design multiple novel conditioning schemes and train SDXL on multiple aspect ratios. We also introduce a refinement model which is used to improve the visual fidelity of samples generated by SDXL using a post-hoc image-to-image technique. We demonstrate that SDXL shows drastically improved performance compared the previous versions of Stable Diffusion and achieves results competitive with those of black-box state-of-the-art image generators. In the spirit of promoting open research and fostering transparency in large model training and evaluation, we provide access to code and model weights at https://github.com/Stability-AI/generative-models"
                },
                {
                    "title": "Emergent Correspondence from Image Diffusion",
                    "abstract": "Finding correspondences between images is a fundamental problem in computer vision. In this paper, we show that correspondence emerges in image diffusion models without any explicit supervision. We propose a simple strategy to extract this implicit knowledge out of diffusion networks as image features, namely DIffusion FeaTures (DIFT), and use them to establish correspondences between real images. Without any additional fine-tuning or supervision on the task-specific data or annotations, DIFT is able to outperform both weakly-supervised methods and competitive off-the-shelf features in identifying semantic, geometric, and temporal correspondences. Particularly for semantic correspondence, DIFT from Stable Diffusion is able to outperform DINO and OpenCLIP by 19 and 14 accuracy points respectively on the challenging SPair-71k benchmark. It even outperforms the state-of-the-art supervised methods on 9 out of 18 categories while remaining on par for the overall performance. Project page: https://diffusionfeatures.github.io"
                },
                {
                    "title": "A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondence",
                    "abstract": "Text-to-image diffusion models have made significant advances in generating and editing high-quality images. As a result, numerous approaches have explored the ability of diffusion model features to understand and process single images for downstream tasks, e.g., classification, semantic segmentation, and stylization. However, significantly less is known about what these features reveal across multiple, different images and objects. In this work, we exploit Stable Diffusion (SD) features for semantic and dense correspondence and discover that with simple post-processing, SD features can perform quantitatively similar to SOTA representations. Interestingly, the qualitative analysis reveals that SD features have very different properties compared to existing representation learning features, such as the recently released DINOv2: while DINOv2 provides sparse but accurate matches, SD features provide high-quality spatial information but sometimes inaccurate semantic matches. We demonstrate that a simple fusion of these two features works surprisingly well, and a zero-shot evaluation using nearest neighbors on these fused features provides a significant performance gain over state-of-the-art methods on benchmark datasets, e.g., SPair-71k, PF-Pascal, and TSS. We also show that these correspondences can enable interesting applications such as instance swapping in two images."
                },
                {
                    "title": "Diffusion Hyperfeatures: Searching Through Time and Space for Semantic Correspondence",
                    "abstract": "Diffusion models have been shown to be capable of generating high-quality images, suggesting that they could contain meaningful internal representations. Unfortunately, the feature maps that encode a diffusion model's internal information are spread not only over layers of the network, but also over diffusion timesteps, making it challenging to extract useful descriptors. We propose Diffusion Hyperfeatures, a framework for consolidating multi-scale and multi-timestep feature maps into per-pixel feature descriptors that can be used for downstream tasks. These descriptors can be extracted for both synthetic and real images using the generation and inversion processes. We evaluate the utility of our Diffusion Hyperfeatures on the task of semantic keypoint correspondence: our method achieves superior performance on the SPair-71k real image benchmark. We also demonstrate that our method is flexible and transferable: our feature aggregation network trained on the inversion features of real image pairs can be used on the generation features of synthetic image pairs with unseen objects and compositions. Our code is available at https://diffusion-hyperfeatures.github.io."
                },
                {
                    "title": "DDP: Diffusion Model for Dense Visual Prediction",
                    "abstract": "We propose a simple, efficient, yet powerful framework for dense visual predictions based on the conditional diffusion pipeline. Our approach follows a \"noise-to-map\" generative paradigm for prediction by progressively removing noise from a random Gaussian distribution, guided by the image. The method, called DDP, efficiently extends the denoising diffusion process into the modern perception pipeline. Without task-specific design and architecture customization, DDP is easy to generalize to most dense prediction tasks, e.g., semantic segmentation and depth estimation. In addition, DDP shows attractive properties such as dynamic inference and uncertainty awareness, in contrast to previous single-step discriminative methods. We show top results on three representative tasks with six diverse benchmarks, without tricks, DDP achieves state-of-the-art or competitive performance on each task compared to the specialist counterparts. For example, semantic segmentation (83.9 mIoU on Cityscapes), BEV map segmentation (70.6 mIoU on nuScenes), and depth estimation (0.05 REL on KITTI). We hope that our approach will serve as a solid baseline and facilitate future research."
                },
                {
                    "title": "Your Diffusion Model is Secretly a Zero-Shot Classifier",
                    "abstract": "The recent wave of large-scale text-to-image diffusion models has dramatically increased our text-based image generation abilities. These models can generate realistic images for a staggering variety of prompts and exhibit impressive compositional generalization abilities. Almost all use cases thus far have solely focused on sampling; however, diffusion models can also provide conditional density estimates, which are useful for tasks beyond image generation. In this paper, we show that the density estimates from large-scale text-to-image diffusion models like Stable Diffusion can be leveraged to perform zero-shot classification without any additional training. Our generative approach to classification, which we call Diffusion Classifier, attains strong results on a variety of benchmarks and outperforms alternative methods of extracting knowledge from diffusion models. Although a gap remains between generative and discriminative approaches on zero-shot recognition tasks, our diffusion-based approach has stronger multimodal compositional reasoning abilities than competing discriminative approaches. Finally, we use Diffusion Classifier to extract standard classifiers from class-conditional diffusion models trained on ImageNet. These models approach the performance of SOTA discriminative classifiers and exhibit strong \"effective robustness\" to distribution shift. Overall, our results are a step toward using generative over discriminative models for downstream tasks. Results and visualizations on our website: diffusion-classifier.github.io/"
                },
                {
                    "title": "Open-Vocabulary Panoptic Segmentation with Text-to-Image Diffusion Models",
                    "abstract": "We present ODISE: Open-vocabulary DIffusion-based panoptic SEgmentation, which unifies pre-trained text-image diffusion and discriminative models to perform open-vocabulary panoptic segmentation. Text-to-image diffusion models have the remarkable ability to generate high-quality images with diverse open-vocabulary language descriptions. This demonstrates that their internal representation space is highly correlated with open concepts in the real world. Text-image discriminative models like CLIP, on the other hand, are good at classifying images into open-vocabulary labels. We leverage the frozen internal representations of both these models to perform panoptic segmentation of any category in the wild. Our approach outperforms the previous state of the art by significant margins on both open-vocabulary panoptic and semantic segmentation tasks. In particular, with COCO training only, our method achieves 23.4 PQ and 30.0 mIoU on the ADE20K dataset, with 8.3 PQ and 7.9 mIoU absolute improvement over the previous state of the art. We open-source our code and models at https://github.com/NVlabs/ODISE."
                },
                {
                    "title": "Unleashing Text-to-Image Diffusion Models for Visual Perception",
                    "abstract": "Diffusion models (DMs) have become the new trend of generative models and have demonstrated a powerful ability of conditional synthesis. Among those, text-to-image diffusion models pre-trained on large-scale image-text pairs are highly controllable by customizable prompts. Unlike the unconditional generative models that focus on low-level attributes and details, text-to-image diffusion models contain more high-level knowledge thanks to the vision-language pre-training. In this paper, we propose VPD (Visual Perception with pre-trained Diffusion models), a new framework that exploits the semantic information of a pre-trained text-to-image diffusion model in visual perception tasks. Instead of using the pre-trained denoising autoencoder in a diffusion-based pipeline, we simply use it as a backbone and aim to study how to take full advantage of the learned knowledge. Specifically, we prompt the denoising decoder with proper textual inputs and refine the text features with an adapter, leading to a better alignment to the pre-trained stage and making the visual contents interact with the text prompts. We also propose to utilize the cross-attention maps between the visual features and the text features to provide explicit guidance. Compared with other pre-training methods, we show that vision-language pre-trained diffusion models can be faster adapted to downstream visual perception tasks using the proposed VPD. Extensive experiments on semantic segmentation, referring image segmentation, and depth estimation demonstrate the effectiveness of our method. Notably, VPD attains 0.254 RMSE on NYUv2 depth estimation and 73.3% oIoU on RefCOCO-val referring image segmentation, establishing new records on these two benchmarks. Code is available at https://github.com/wl-zhao/VPD."
                },
                {
                    "title": "Scalable Diffusion Models with Transformers",
                    "abstract": "We explore a new class of diffusion models based on the transformer architecture. We train latent diffusion models of images, replacing the commonly-used U-Net backbone with a transformer that operates on latent patches. We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops. We find that DiTs with higher Gflops\u2014through increased transformer depth/width or increased number of input tokens\u2014consistently have lower FID. In addition to possessing good scalability properties, our largest DiT-XL/2 models outperform all prior diffusion models on the class-conditional ImageNet 512\u00d7512 and 256\u00d7256 benchmarks, achieving a state-of-the-art FID of 2.27 on the latter."
                },
                {
                    "title": "Flow Matching for Generative Modeling",
                    "abstract": "We introduce a new paradigm for generative modeling built on Continuous Normalizing Flows (CNFs), allowing us to train CNFs at unprecedented scale. Specifically, we present the notion of Flow Matching (FM), a simulation-free approach for training CNFs based on regressing vector fields of fixed conditional probability paths. Flow Matching is compatible with a general family of Gaussian probability paths for transforming between noise and data samples -- which subsumes existing diffusion paths as specific instances. Interestingly, we find that employing FM with diffusion paths results in a more robust and stable alternative for training diffusion models. Furthermore, Flow Matching opens the door to training CNFs with other, non-diffusion probability paths. An instance of particular interest is using Optimal Transport (OT) displacement interpolation to define the conditional probability paths. These paths are more efficient than diffusion paths, provide faster training and sampling, and result in better generalization. Training CNFs using Flow Matching on ImageNet leads to consistently better performance than alternative diffusion-based methods in terms of both likelihood and sample quality, and allows fast and reliable sample generation using off-the-shelf numerical ODE solvers."
                },
                {
                    "title": "MaskCLIP: Masked Self-Distillation Advances Contrastive Language-Image Pretraining",
                    "abstract": "This paper presents a simple yet effective framework MaskCLIP, which incorporates a newly proposed masked self-distillation into contrastive language-image pretraining. The core idea of masked self-distillation is to distill representation from a full image to the representation predicted from a masked image. Such incorporation enjoys two vital benefits. First, masked self-distillation targets local patch representation learning, which is complementary to vision-language contrastive focusing on text-related representation. Second, masked self-distillation is also consistent with vision-language contrastive from the perspective of training objective as both utilize the visual encoder for feature aligning, and thus is able to learn local semantics getting indirect supervision from the language. We provide specially designed experiments with a comprehensive analysis to validate the two benefits. Symmetrically, we also introduce the local semantic supervision into the text branch, which further improves the pretraining performance. With extensive experiments, we show that MaskCLIP, when applied to various challenging downstream tasks, achieves superior results in linear probing, finetuning, and zeroshot performance with the guidance of the language encoder. Code will be release at https://github.com/LightDXY/MaskCLIP."
                },
                {
                    "title": "Optimizing Partial Area Under the Top-k Curve: Theory and Practice",
                    "abstract": "Top-<inline-formula><tex-math notation=\"LaTeX\">$k$</tex-math><alternatives><mml:math><mml:mi>k</mml:mi></mml:math><inline-graphic xlink:href=\"wang-ieq1-3199970.gif\"/></alternatives></inline-formula> error has become a popular metric for large-scale classification benchmarks due to the inevitable semantic ambiguity among classes. Existing literature on top-<inline-formula><tex-math notation=\"LaTeX\">$k$</tex-math><alternatives><mml:math><mml:mi>k</mml:mi></mml:math><inline-graphic xlink:href=\"wang-ieq2-3199970.gif\"/></alternatives></inline-formula> optimization generally focuses on the optimization method of the top-<inline-formula><tex-math notation=\"LaTeX\">$k$</tex-math><alternatives><mml:math><mml:mi>k</mml:mi></mml:math><inline-graphic xlink:href=\"wang-ieq3-3199970.gif\"/></alternatives></inline-formula> objective, while ignoring the limitations of the metric itself. In this paper, we point out that the top-<inline-formula><tex-math notation=\"LaTeX\">$k$</tex-math><alternatives><mml:math><mml:mi>k</mml:mi></mml:math><inline-graphic xlink:href=\"wang-ieq4-3199970.gif\"/></alternatives></inline-formula> objective lacks enough discrimination such that the induced predictions may give a totally irrelevant label a top rank. To fix this issue, we develop a novel metric named partial Area Under the top-<inline-formula><tex-math notation=\"LaTeX\">$k$</tex-math><alternatives><mml:math><mml:mi>k</mml:mi></mml:math><inline-graphic xlink:href=\"wang-ieq5-3199970.gif\"/></alternatives></inline-formula> Curve (AUTKC). Theoretical analysis shows that AUTKC has a better discrimination ability, and its Bayes optimal score function could give a correct top-<inline-formula><tex-math notation=\"LaTeX\">$K$</tex-math><alternatives><mml:math><mml:mi>K</mml:mi></mml:math><inline-graphic xlink:href=\"wang-ieq6-3199970.gif\"/></alternatives></inline-formula> ranking with respect to the conditional probability. This shows that AUTKC does not allow irrelevant labels to appear in the top list. Furthermore, we present an empirical surrogate risk minimization framework to optimize the proposed metric. Theoretically, we present (1) a sufficient condition for Fisher consistency of the Bayes optimal score function; (2) a generalization upper bound which is insensitive to the number of classes under a simple hyperparameter setting. Finally, the experimental results on four benchmark datasets validate the effectiveness of our proposed framework."
                },
                {
                    "title": "Open-Vocabulary Universal Image Segmentation with MaskCLIP",
                    "abstract": "In this paper, we tackle an emerging computer vision task, open-vocabulary universal image segmentation, that aims to perform semantic/instance/panoptic segmentation (background semantic labeling + foreground instance segmentation) for arbitrary categories of text-based descriptions in inference time. We first build a baseline method by directly adopting pre-trained CLIP models without finetuning or distillation. We then develop MaskCLIP, a Transformer-based approach with a MaskCLIP Visual Encoder, which is an encoder-only module that seamlessly integrates mask tokens with a pre-trained ViT CLIP model for semantic/instance segmentation and class prediction. MaskCLIP learns to efficiently and effectively utilize pre-trained partial/dense CLIP features within the MaskCLIP Visual Encoder that avoids the time-consuming student-teacher training process. MaskCLIP outperforms previous methods for semantic/instance/panoptic segmentation on ADE20K and PASCAL datasets. We show qualitative illustrations for MaskCLIP with online custom categories. Project website: https://maskclip.github.io."
                },
                {
                    "title": "Dual Diffusion Implicit Bridges for Image-to-Image Translation",
                    "abstract": "Common image-to-image translation methods rely on joint training over data from both source and target domains. The training process requires concurrent access to both datasets, which hinders data separation and privacy protection; and existing models cannot be easily adapted for translation of new domain pairs. We present Dual Diffusion Implicit Bridges (DDIBs), an image translation method based on diffusion models, that circumvents training on domain pairs. Image translation with DDIBs relies on two diffusion models trained independently on each domain, and is a two-step process: DDIBs first obtain latent encodings for source images with the source diffusion model, and then decode such encodings using the target model to construct target images. Both steps are defined via ordinary differential equations (ODEs), thus the process is cycle consistent only up to discretization errors of the ODE solvers. Theoretically, we interpret DDIBs as concatenation of source to latent, and latent to target Schrodinger Bridges, a form of entropy-regularized optimal transport, to explain the efficacy of the method. Experimentally, we apply DDIBs on synthetic and high-resolution image datasets, to demonstrate their utility in a wide variety of translation tasks and their inherent optimal transport properties."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Deep ViT Features as Dense Visual Descriptors",
                    "abstract": "We study the use of deep features extracted from a pretrained Vision Transformer (ViT) as dense visual descriptors. We observe and empirically demonstrate that such features, when extractedfrom a self-supervised ViT model (DINO-ViT), exhibit several striking properties, including: (i) the features encode powerful, well-localized semantic information, at high spatial granularity, such as object parts; (ii) the encoded semantic information is shared across related, yet different object categories, and (iii) positional bias changes gradually throughout the layers. These properties allow us to design simple methods for a variety of applications, including co-segmentation, part co-segmentation and semantic correspondences. To distill the power of ViT features from convoluted design choices, we restrict ourselves to lightweight zero-shot methodologies (e.g., binning and clustering) applied directly to the features. Since our methods require no additional training nor data, they are readily applicable across a variety of domains. We show by extensive qualitative and quantitative evaluation that our simple methodologies achieve competitive results with recent state-of-the-art supervised methods, and outperform previous unsupervised methods by a large margin. Code is available in dino-vit-features.github.io."
                },
                {
                    "title": "Label-Efficient Semantic Segmentation with Diffusion Models",
                    "abstract": "Denoising diffusion probabilistic models have recently received much research attention since they outperform alternative approaches, such as GANs, and currently provide state-of-the-art generative performance. The superior performance of diffusion models has made them an appealing tool in several applications, including inpainting, super-resolution, and semantic editing. In this paper, we demonstrate that diffusion models can also serve as an instrument for semantic segmentation, especially in the setup when labeled data is scarce. In particular, for several pretrained diffusion models, we investigate the intermediate activations from the networks that perform the Markov step of the reverse diffusion process. We show that these activations effectively capture the semantic information from an input image and appear to be excellent pixel-level representations for the segmentation problem. Based on these observations, we describe a simple segmentation method, which can work even if only a few training images are provided. Our approach significantly outperforms the existing alternatives on several datasets for the same amount of human supervision."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "Transformer Feed-Forward Layers Are Key-Value Memories",
                    "abstract": "Feed-forward layers constitute two-thirds of a transformer model\u2019s parameters, yet their role in the network remains under-explored. We show that feed-forward layers in transformer-based language models operate as key-value memories, where each key correlates with textual patterns in the training examples, and each value induces a distribution over the output vocabulary. Our experiments show that the learned patterns are human-interpretable, and that lower layers tend to capture shallow patterns, while upper layers learn more semantic ones. The values complement the keys\u2019 input patterns by inducing output distributions that concentrate probability mass on tokens likely to appear immediately after each pattern, particularly in the upper layers. Finally, we demonstrate that the output of a feed-forward layer is a composition of its memories, which is subsequently refined throughout the model\u2019s layers via residual connections to produce the final output distribution."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Unsupervised Learning of Visual Features by Contrasting Cluster Assignments",
                    "abstract": "Unsupervised image representations have significantly reduced the gap with supervised pretraining, notably with the recent achievements of contrastive learning methods. These contrastive methods typically work online and rely on a large number of explicit pairwise feature comparisons, which is computationally challenging. In this paper, we propose an online algorithm, SwAV, that takes advantage of contrastive methods without requiring to compute pairwise comparisons. Specifically, our method simultaneously clusters the data while enforcing consistency between cluster assignments produced for different augmentations (or views) of the same image, instead of comparing features directly as in contrastive learning. Simply put, we use a swapped prediction mechanism where we predict the cluster assignment of a view from the representation of another view. Our method can be trained with large and small batches and can scale to unlimited amounts of data. Compared to previous contrastive methods, our method is more memory efficient since it does not require a large memory bank or a special momentum network. In addition, we also propose a new data augmentation strategy, multi-crop, that uses a mix of views with different resolutions in place of two full-resolution views, without increasing the memory or compute requirements much. We validate our findings by achieving 75.3% top-1 accuracy on ImageNet with ResNet-50, as well as surpassing supervised pretraining on all the considered transfer tasks."
                },
                {
                    "title": "SPair-71k: A Large-scale Benchmark for Semantic Correspondence",
                    "abstract": "Establishing visual correspondences under large intra-class variations, which is often referred to as semantic correspondence or semantic matching, remains a challenging problem in computer vision. Despite its significance, however, most of the datasets for semantic correspondence are limited to a small amount of image pairs with similar viewpoints and scales. In this paper, we present a new large-scale benchmark dataset of semantically paired images, SPair-71k, which contains 70,958 image pairs with diverse variations in viewpoint and scale. Compared to previous datasets, it is significantly larger in number and contains more accurate and richer annotations. We believe this dataset will provide a reliable testbed to study the problem of semantic correspondence and will help to advance research in this area. We provide the results of recent methods on our new dataset as baselines for further research. Our benchmark is available online at this http URL"
                },
                {
                    "title": "A Brief Review of Nearest Neighbor Algorithm for Learning and Classification",
                    "abstract": "k-Nearest Neighbor (kNN) algorithm is an effortless but productive machine learning algorithm. It is effective for classification as well as regression. However, it is more widely used for classification prediction. kNN groups the data into coherent clusters or subsets and classifies the newly inputted data based on its similarity with previously trained data. The input is assigned to the class with which it shares the most nearest neighbors. Though kNN is effective, it has many weaknesses. This paper highlights the kNN method and its modified versions available in previously done researches. These variants remove the weaknesses of kNN and provide a more efficient method."
                },
                {
                    "title": "Road Extraction by Deep Residual U-Net",
                    "abstract": "Road extraction from aerial images has been a hot research topic in the field of remote sensing image analysis. In this letter, a semantic segmentation neural network, which combines the strengths of residual learning and U-Net, is proposed for road area extraction. The network is built with residual units and has similar architecture to that of U-Net. The benefits of this model are twofold: first, residual units ease training of deep networks. Second, the rich skip connections within the network could facilitate information propagation, allowing us to design networks with fewer parameters, however, better performance. We test our network on a public road data set and compare it with U-Net and other two state-of-the-art deep-learning-based road extraction methods. The proposed approach outperforms all the comparing methods, which demonstrates its superiority over recently developed state of the arts."
                },
                {
                    "title": "Scene Parsing through ADE20K Dataset",
                    "abstract": "Scene parsing, or recognizing and segmenting objects and stuff in an image, is one of the key problems in computer vision. Despite the communitys efforts in data collection, there are still few image datasets covering a wide range of scenes and object categories with dense and detailed annotations for scene parsing. In this paper, we introduce and analyze the ADE20K dataset, spanning diverse annotations of scenes, objects, parts of objects, and in some cases even parts of parts. A scene parsing benchmark is built upon the ADE20K with 150 object and stuff classes included. Several segmentation baseline models are evaluated on the benchmark. A novel network design called Cascade Segmentation Module is proposed to parse a scene into stuff, objects, and object parts in a cascade and improve over the baselines. We further show that the trained scene parsing networks can lead to applications such as image content removal and scene synthesis1."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "The Cityscapes Dataset for Semantic Urban Scene Understanding",
                    "abstract": "Visual understanding of complex urban street scenes is an enabling factor for a wide range of applications. Object detection has benefited enormously from large-scale datasets, especially in the context of deep learning. For semantic urban scene understanding, however, no current dataset adequately captures the complexity of real-world urban scenes. To address this, we introduce Cityscapes, a benchmark suite and large-scale dataset to train and test approaches for pixel-level and instance-level semantic labeling. Cityscapes is comprised of a large, diverse set of stereo video sequences recorded in streets from 50 different cities. 5000 of these images have high quality pixel-level annotations, 20 000 additional images have coarse annotations to enable methods that leverage large volumes of weakly-labeled data. Crucially, our effort exceeds previous attempts in terms of dataset size, annotation richness, scene variability, and complexity. Our accompanying empirical study provides an in-depth analysis of the dataset characteristics, as well as a performance evaluation of several state-of-the-art approaches based on our benchmark."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop",
                    "abstract": "While there has been remarkable progress in the performance of visual recognition algorithms, the state-of-the-art models tend to be exceptionally data-hungry. Large labeled training datasets, expensive and tedious to produce, are required to optimize millions of parameters in deep network models. Lagging behind the growth in model capacity, the available datasets are quickly becoming outdated in terms of size and density. To circumvent this bottleneck, we propose to amplify human effort through a partially automated labeling scheme, leveraging deep learning with humans in the loop. Starting from a large set of candidate images for each category, we iteratively sample a subset, ask people to label them, classify the others with a trained model, split the set into positives, negatives, and unlabeled based on the classification confidence, and then iterate with the unlabeled set. To assess the effectiveness of this cascading procedure and enable further progress in visual recognition research, we construct a new image dataset, LSUN. It contains around one million labeled images for each of 10 scene categories and 20 object categories. We experiment with training popular convolutional networks and find that they achieve substantial performance gains when trained on this dataset."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "When Hyperspectral Image Classification Meets Diffusion Models: An Unsupervised Feature Learning Framework",
                    "abstract": "Learning effective spectral-spatial features is important for the hyperspectral image (HSI) classification task, but the majority of existing HSI classification methods still suffer from modeling complex spectral-spatial relations and characterizing low-level details and high-level semantics comprehensively. As a new class of record-breaking generative models, diffusion models are capable of modeling complex relations for understanding inputs well as learning both high-level and lowlevel visual features. Meanwhile, diffusion models can capture more abundant features by taking advantage of the extra and unique dimension of timestep t. In view of these, we propose an unsupervised spectral-spatial feature learning framework based on the diffusion model for HSI classification for the first time, named Diff-HSI. Specifically, we first pretrain the diffusion model with unlabeled HSI patches for unsupervised feature learning, and then exploit intermediate hierarchical features from different timesteps for classification. For better using the abundant timestep-wise features, we design a timestepwise feature bank and a dynamic feature fusion module to construct timestep-wise features, adaptively learning informative multi-timestep representations. Finally, an ensemble of linear classifiers is applied to perform HSI classification. Extensive experiments are conducted on three public HSI datasets, and our results demonstrate that Diff-HSI outperforms state-of-theart supervised and unsupervised methods for HSI classification."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively select a comprehensive subset of activations from diffusion models to improve their discriminative performance in tasks such as semantic segmentation?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of feature selection in diffusion models, which are increasingly being applied to discriminative tasks. By improving the selection of activations, we can enhance the performance of these models, leading to better outcomes in various applications such as image segmentation and object recognition. This research could pave the way for future studies to explore more effective architectures and methodologies, ultimately contributing to the development of more robust generative and discriminative models.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the vast number of activations generated by complex diffusion U-Net architectures, which vary significantly in quality. Naive approaches that rely on direct quantitative comparisons may fail due to the overwhelming volume of activations, making it impractical to evaluate each one thoroughly. Additionally, the introduction of new architectural components, such as cross-attention and embedded deep vision transformers, complicates the feature selection process, as many potentially valuable activations may be overlooked. Identifying and filtering out sub-optimal activations while ensuring a comprehensive evaluation of the remaining candidates is a significant technical and theoretical obstacle.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research, such as that by Baranchuk et al., has primarily focused on a limited set of activations, neglecting many potentially useful features from newer architectural components. This narrow focus has created a gap in understanding the full range of activations available in modern diffusion models. Barriers such as the lack of a systematic methodology for evaluating a broader candidate pool and the complexity of the models themselves have hindered progress. Our approach differs by employing a qualitative analysis to filter candidates before conducting a quantitative comparison, allowing for a more comprehensive evaluation of activations.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a two-step process: first, we conduct a qualitative analysis of the properties of diffusion U-Nets to filter out sub-optimal activations, thereby narrowing down the candidate pool. Next, we perform a quantitative comparison on the remaining activations to identify the most effective features for discriminative tasks. We will utilize a dataset relevant to semantic segmentation and measure performance using metrics such as PCK ("
            }
        },
        "author_data": {
            "26479dab-c8a5-459f-af83-b8ecc13e4165": {
                "pk": "26479dab-c8a5-459f-af83-b8ecc13e4165",
                "name": "Benyuan Meng",
                "collaborators": [
                    "Qianqian Xu",
                    "Zitai Wang",
                    "Zhiyong Yang",
                    "Xiaochun Cao",
                    "Qingming Huang"
                ],
                "domain": [
                    "Generative Models",
                    "Computer Vision",
                    "Deep Learning",
                    "Feature Extraction"
                ],
                "publications": [
                    {
                        "title": "Suppress Content Shift: Better Diffusion Features via Off-the-Shelf Generation Techniques",
                        "abstract": "Diffusion models are powerful generative models, and this capability can also be applied to discrimination. The inner activations of a pre-trained diffusion model can serve as features for discriminative tasks, namely, diffusion feature. We discover that diffusion feature has been hindered by a hidden yet universal phenomenon that we call content shift. To be specific, there are content differences between features and the input image, such as the exact shape of a certain object. We locate the cause of content shift as one inherent characteristic of diffusion models, which suggests the broad existence of this phenomenon in diffusion feature. Further empirical study also indicates that its negative impact is not negligible even when content shift is not visually perceivable. Hence, we propose to suppress content shift to enhance the overall quality of diffusion features. Specifically, content shift is related to the information drift during the process of recovering an image from the noisy input, pointing out the possibility of turning off-the-shelf generation techniques into tools for content shift suppression. We further propose a practical guideline named GATE to efficiently evaluate the potential benefit of a technique and provide an implementation of our methodology. Despite the simplicity, the proposed approach has achieved superior results on various tasks and datasets, validating its potential as a generic booster for diffusion features. Our code is available at https://github.com/Darkbblue/diffusion-content-shift."
                    }
                ]
            },
            "89cab4be-cf50-42c2-96c8-de1533e0ecc7": {
                "pk": "89cab4be-cf50-42c2-96c8-de1533e0ecc7",
                "name": "Qianqian Xu",
                "collaborators": [
                    "Xiaochun Cao",
                    "Qingming Huang",
                    "Zhiyong Yang",
                    "Yuan Yao",
                    "Jiechao Xiong",
                    "Ke Ma",
                    "Zongsheng Cao",
                    "Ming Yan",
                    "Zuyao Chen",
                    "Runmin Cong"
                ],
                "domain": [
                    "Crowdsourcing",
                    "Multi-Task Learning",
                    "Outlier Detection",
                    "Knowledge Graph"
                ],
                "publications": [
                    {
                        "title": "False Discovery Rate Control and Statistical Quality Assessment of Annotators in Crowdsourced Ranking",
                        "abstract": "With the rapid growth of crowdsourcing platforms it has become easy and relatively inexpensive to collect a dataset labeled by multiple annotators in a short time. However due to the lack of control over the quality of the annotators, some abnormal annotators may be affected by position bias which can potentially degrade the quality of the final consensus labels. In this paper we introduce a statistical framework to model and detect annotator's position bias in order to control the false discovery rate (FDR) without a prior knowledge on the amount of biased annotators - the expected fraction of false discoveries among all discoveries being not too high, in order to assure that most of the discoveries are indeed true and replicable. The key technical development relies on some new knockoff filters adapted to our problem and new algorithms based on the Inverse Scale Space dynamics whose discretization is potentially suitable for large scale crowdsourcing data analysis. Our studies are supported by experiments with both simulated examples and real-world data. The proposed framework provides us a useful tool for quantitatively studying annotator's abnormal behavior in crowdsourcing data arising from machine learning, sociology, computer vision, multimedia, etc."
                    },
                    {
                        "title": "From Common to Special: When Multi-Attribute Learning Meets Personalized Opinions",
                        "abstract": "Visual attributes, which refer to human-labeled semantic annotations, have gained increasing popularity in a wide range of real world applications. Generally, the existing attribute learning methods fall into two categories: one focuses on learning user-specific labels separately for different attributes, while the other one focuses on learning crowd-sourced global labels jointly for multiple attributes. However, both categories ignore the joint effect of the two mentioned factors: the personal diversity with respect to the global consensus; and the intrinsic correlation among multiple attributes. To overcome this challenge, we propose a novel model to learn user-specific predictors across multiple attributes. In our proposed model, the diversity of personalized opinions and the intrinsic relationship among multiple attributes are unified in a common-to-special manner. To this end, we adopt a three-component decomposition. Specifically, our model integrates a common cognition factor, an attribute-specific bias factor and a user-specific bias factor. Meanwhile Lasso and group Lasso penalties are adopted to leverage efficient feature selection. Furthermore, theoretical analysis is conducted to show that our proposed method could reach reasonable performance. Eventually, the empirical study carried out in this paper demonstrates the effectiveness of our proposed method."
                    },
                    {
                        "title": "Robust Ordinal Embedding from Contaminated Relative Comparisons",
                        "abstract": "Existing ordinal embedding methods usually follow a two-stage routine: outlier detection is first employed to pick out the inconsistent comparisons; then an embedding is learned from the clean data. However, learning in a multi-stage manner is well-known to suffer from sub-optimal solutions. In this paper, we propose a unified framework to jointly identify the contaminated comparisons and derive reliable embeddings. The merits of our method are three-fold: (1) By virtue of the proposed unified framework, the sub-optimality of traditional methods is largely alleviated; (2) The proposed method is aware of global inconsistency by minimizing a corresponding cost, while traditional methods only involve local inconsistency; (3) Instead of considering the nuclear norm heuristics, we adopt an exact solution for rank equality constraint. Our studies are supported by experiments with both simulated examples and real-world data. The proposed framework provides us a promising tool for robust ordinal embedding from the contaminated comparisons."
                    },
                    {
                        "title": "ER: Equivariance Regularizer for Knowledge Graph Completion",
                        "abstract": "Tensor factorization and distanced based models play important roles in knowledge graph completion (KGC). However, the relational matrices in KGC methods often induce a high model complexity, bearing a high risk of overfitting. As a remedy, researchers propose a variety of different regularizers such as the tensor nuclear norm regularizer. Our motivation is based on the observation that the previous work only focuses on the \"size\" of the parametric space, while leaving the implicit semantic information widely untouched. To address this issue, we propose a new regularizer, namely, Equivariance Regularizer (ER), which can suppress overfitting by leveraging the implicit semantic information. Specifically, ER can enhance the generalization ability of the model by employing the semantic equivariance between the head and tail entities. Moreover, it is a generic solution for both distance based models and tensor factorization based models. The experimental results indicate a clear and substantial improvement over the state-of-the-art relation prediction methods."
                    },
                    {
                        "title": "Fast Adaptive Algorithm for Robust Evaluation of Quality of Experience",
                        "abstract": "Outlier detection is an integral part of robust evaluation for crowdsourceable Quality of Experience (QoE) and has attracted much attention in recent years. In QoE for multimedia, outliers happen because of different test conditions, human errors, abnormal variations in context, {etc}. In this paper, we propose a simple yet effective algorithm for outlier detection and robust QoE evaluation named iterative Least Trimmed Squares (iLTS). The algorithm assigns binary weights to samples, i.e., 0 or 1 indicating if a sample is an outlier, then the outlier-trimmed subset least squares solutions give robust ranking scores. An iterative optimization is carried alternatively between updating weights and ranking scores which converges to a local optimizer in finite steps. In our test setting, iLTS is up to 190 times faster than LASSO-based methods with a comparable performance. Moreover, a varied version of this method shows adaptation in outlier detection, which provides an automatic detection to determine whether a data sample is an outlier without \\emph{a priori} knowledge about the amount of the outliers. The effectiveness and efficiency of iLTS are demonstrated on both simulated examples and real-world applications. A Matlab package is provided to researchers exploiting crowdsourcing paired comparison data for robust ranking."
                    },
                    {
                        "title": "Learning Personalized Attribute Preference via Multi-task AUC Optimization",
                        "abstract": "Traditionally, most of the existing attribute learning methods are trained based on the consensus of annotations aggregated from a limited number of annotators. However, the consensus might fail in settings, especially when a wide spectrum of annotators with different interests and comprehension about the attribute words are involved. In this paper, we develop a novel multi-task method to understand and predict personalized attribute annotations. Regarding the attribute preference learning for each annotator as a specific task, we first propose a multi-level task parameter decomposition to capture the evolution from a highly popular opinion of the mass to highly personalized choices that are special for each person. Meanwhile, for personalized learning methods, ranking prediction is much more important than accurate classification. This motivates us to employ an Area Under ROC Curve (AUC) based loss function to improve our model. On top of the AUC-based loss, we propose an efficient method to evaluate the loss and gradients. Theoretically, we propose a novel closed-form solution for one of our non-convex subproblem, which leads to provable convergence behaviors. Furthermore, we also provide a generalization bound to guarantee a reasonable performance. Finally, empirical analysis consistently speaks to the efficacy of our proposed method."
                    },
                    {
                        "title": "Global Context-Aware Progressive Aggregation Network for Salient Object Detection",
                        "abstract": "Deep convolutional neural networks have achieved competitive performance in salient object detection, in which how to learn effective and comprehensive features plays a critical role. Most of the previous works mainly adopted multiple level feature integration yet ignored the gap between different features. Besides, there also exists a dilution process of high-level features as they passed on the top-down pathway. To remedy these issues, we propose a novel network named GCPANet to effectively integrate low-level appearance features, high-level semantic features, and global context features through some progressive context-aware Feature Interweaved Aggregation (FIA) modules and generate the saliency map in a supervised way. Moreover, a Head Attention (HA) module is used to reduce information redundancy and enhance the top layers features by leveraging the spatial and channel-wise attention, and the Self Refinement (SR) module is utilized to further refine and heighten the input features. Furthermore, we design the Global Context Flow (GCF) module to generate the global context information at different stages, which aims to learn the relationship among different salient regions and alleviate the dilution effect of high-level features. Experimental results on six benchmark datasets demonstrate that the proposed approach outperforms the state-of-the-art methods both quantitatively and qualitatively."
                    },
                    {
                        "title": "Task-Feature Collaborative Learning with Application to Personalized Attribute Prediction",
                        "abstract": "As an effective learning paradigm against insufficient training samples, Multi-Task Learning (MTL) encourages knowledge sharing across multiple related tasks so as to improve the overall performance. In MTL, a major challenge springs from the phenomenon that sharing the knowledge with dissimilar and hard tasks, known as negative transfer, often results in a worsened performance. Though a substantial amount of studies have been carried out against the negative transfer, most of the existing methods only model the transfer relationship as task correlations, with the transfer across features and tasks left unconsidered. Different from the existing methods, our goal is to alleviate negative transfer collaboratively across features and tasks. To this end, we propose a novel multi-task learning method called Task-Feature Collaborative Learning (TFCL). Specifically, we first propose a base model with a heterogeneous block-diagonal structure regularizer to leverage the collaborative grouping of features and tasks and suppressing inter-group knowledge sharing. We then propose an optimization method for the model. Extensive theoretical analysis shows that our proposed method has the following benefits: (a) it enjoys the global convergence property and (b) it provides a block-diagonal structure recovery guarantee. As a practical extension, we extend the base model by allowing overlapping features and differentiating the hard tasks. We further apply it to the personalized attribute prediction problem with fine-grained modeling of user behaviors. Finally, experimental results on both simulated dataset and real-world datasets demonstrate the effectiveness of our proposed method"
                    }
                ]
            },
            "149ffcd7-3bf8-4acd-a625-19302b24506f": {
                "pk": "149ffcd7-3bf8-4acd-a625-19302b24506f",
                "name": "Zitai Wang",
                "collaborators": [
                    "Qianqian Xu",
                    "Qingming Huang",
                    "Zhiyong Yang",
                    "Xiaochun Cao",
                    "Yuan He",
                    "Yuchen Sun",
                    "Junwei He",
                    "Yangbangyan Jiang",
                    "Peisong Wen",
                    "Benyuan Meng"
                ],
                "domain": [
                    "Adversarial Learning",
                    "Graph Anomaly Detection",
                    "Open-Set Recognition",
                    "Imbalanced Learning"
                ],
                "publications": [
                    {
                        "title": "When Measures are Unreliable: Imperceptible Adversarial Perturbations toward Top-$k$ Multi-Label Learning",
                        "abstract": "With the great success of deep neural networks, adversarial learning has received widespread attention in various studies, ranging from multi-class learning to multi-label learning. However, existing adversarial attacks toward multi-label learning only pursue the traditional visual imperceptibility but ignore the new perceptible problem coming from measures such as Precision@$k$ and mAP@$k$. Specifically, when a well-trained multi-label classifier performs far below the expectation on some samples, the victim can easily realize that this performance degeneration stems from attack, rather than the model itself. Therefore, an ideal multi-labeling adversarial attack should manage to not only deceive visual perception but also evade monitoring of measures. To this end, this paper first proposes the concept of measure imperceptibility. Then, a novel loss function is devised to generate such adversarial perturbations that could achieve both visual and measure imperceptibility. Furthermore, an efficient algorithm, which enjoys a convex objective, is established to optimize this objective. Finally, extensive experiments on large-scale benchmark datasets, such as PASCAL VOC 2012, MS COCO, and NUS WIDE, demonstrate the superiority of our proposed method in attacking the top-$k$ multi-label systems."
                    },
                    {
                        "title": "ADA-GAD: Anomaly-Denoised Autoencoders for Graph Anomaly Detection",
                        "abstract": "Graph anomaly detection is crucial for identifying nodes that deviate from regular behavior within graphs, benefiting various domains such as fraud detection and social network. Although existing reconstruction-based methods have achieved considerable success, they may face the \\textit{Anomaly Overfitting} and \\textit{Homophily Trap} problems caused by the abnormal patterns in the graph, breaking the assumption that normal nodes are often better reconstructed than abnormal ones. Our observations indicate that models trained on graphs with fewer anomalies exhibit higher detection performance. Based on this insight, we introduce a novel two-stage framework called Anomaly-Denoised Autoencoders for Graph Anomaly Detection (ADA-GAD). In the first stage, we design a learning-free anomaly-denoised augmentation method to generate graphs with reduced anomaly levels. We pretrain graph autoencoders on these augmented graphs at multiple levels, which enables the graph autoencoders to capture normal patterns. In the next stage, the decoders are retrained for detection on the original graph, benefiting from the multi-level representations learned in the previous stage. Meanwhile, we propose the node anomaly distribution regularization to further alleviate \\textit{Anomaly Overfitting}. We validate the effectiveness of our approach through extensive experiments on both synthetic and real-world datasets."
                    },
                    {
                        "title": "Optimizing Partial Area Under the Top-k Curve: Theory and Practice",
                        "abstract": "Top-k error has become a popular metric for large-scale classification benchmarks due to the inevitable semantic ambiguity among classes. Existing literature on top-k optimization generally focuses on the optimization method of the top-k objective, while ignoring the limitations of the metric itself. In this paper, we point out that the top-k objective lacks enough discrimination such that the induced predictions may give a totally irrelevant label a top rank. To fix this issue, we develop a novel metric named partial Area Under the top-k Curve (AUTKC). Theoretical analysis shows that AUTKC has a better discrimination ability, and its Bayes optimal score function could give a correct top-K ranking with respect to the conditional probability. This shows that AUTKC does not allow irrelevant labels to appear in the top list. Furthermore, we present an empirical surrogate risk minimization framework to optimize the proposed metric. Theoretically, we present (1) a sufficient condition for Fisher consistency of the Bayes optimal score function; (2) a generalization upper bound which is insensitive to the number of classes under a simple hyperparameter setting. Finally, the experimental results on four benchmark datasets validate the effectiveness of our proposed framework."
                    },
                    {
                        "title": "Top-K Pairwise Ranking: Bridging the Gap Among Ranking-Based Measures for Multi-Label Classification",
                        "abstract": "Multi-label ranking, which returns multiple top-ranked labels for each instance, has a wide range of applications for visual tasks. Due to its complicated setting, prior arts have proposed various measures to evaluate model performances. However, both theoretical analysis and empirical observations show that a model might perform inconsistently on different measures. To bridge this gap, this paper proposes a novel measure named Top-K Pairwise Ranking (TKPR), and a series of analyses show that TKPR is compatible with existing ranking-based measures. In light of this, we further establish an empirical surrogate risk minimization framework for TKPR. On one hand, the proposed framework enjoys convex surrogate losses with the theoretical support of Fisher consistency. On the other hand, we establish a sharp generalization bound for the proposed framework based on a novel technique named data-dependent contraction. Finally, empirical results on benchmark datasets validate the effectiveness of the proposed framework."
                    },
                    {
                        "title": "OpenAUC: Towards AUC-Oriented Open-Set Recognition",
                        "abstract": "Traditional machine learning follows a close-set assumption that the training and test set share the same label space. While in many practical scenarios, it is inevitable that some test samples belong to unknown classes (open-set). To fix this issue, Open-Set Recognition (OSR), whose goal is to make correct predictions on both close-set samples and open-set samples, has attracted rising attention. In this direction, the vast majority of literature focuses on the pattern of open-set samples. However, how to evaluate model performance in this challenging task is still unsolved. In this paper, a systematic analysis reveals that most existing metrics are essentially inconsistent with the aforementioned goal of OSR: (1) For metrics extended from close-set classification, such as Open-set F-score, Youden's index, and Normalized Accuracy, a poor open-set prediction can escape from a low performance score with a superior close-set prediction. (2) Novelty detection AUC, which measures the ranking performance between close-set and open-set samples, ignores the close-set performance. To fix these issues, we propose a novel metric named OpenAUC. Compared with existing metrics, OpenAUC enjoys a concise pairwise formulation that evaluates open-set performance and close-set performance in a coupling manner. Further analysis shows that OpenAUC is free from the aforementioned inconsistency properties. Finally, an end-to-end learning method is proposed to minimize the OpenAUC risk, and the experimental results on popular benchmark datasets speak to its effectiveness. Project Page: https://github.com/wang22ti/OpenAUC."
                    },
                    {
                        "title": "HGOE: Hybrid External and Internal Graph Outlier Exposure for Graph Out-of-Distribution Detection",
                        "abstract": "With the progressive advancements in deep graph learning, out-of-distribution (OOD) detection for graph data has emerged as a critical challenge. While the efficacy of auxiliary datasets in enhancing OOD detection has been extensively studied for image and text data, such approaches have not yet been explored for graph data. Unlike Euclidean data, graph data exhibits greater diversity but lower robustness to perturbations, complicating the integration of outliers. To tackle these challenges, we propose the introduction of \\textbf{H}ybrid External and Internal \\textbf{G}raph \\textbf{O}utlier \\textbf{E}xposure (HGOE) to improve graph OOD detection performance. Our framework involves using realistic external graph data from various domains and synthesizing internal outliers within ID subgroups to address the poor robustness and presence of OOD samples within the ID class. Furthermore, we develop a boundary-aware OE loss that adaptively assigns weights to outliers, maximizing the use of high-quality OOD samples while minimizing the impact of low-quality ones. Our proposed HGOE framework is model-agnostic and designed to enhance the effectiveness of existing graph OOD detection models. Experimental results demonstrate that our HGOE framework can significantly improve the performance of existing OOD detection models across all 8 real datasets."
                    },
                    {
                        "title": "Suppress Content Shift: Better Diffusion Features via Off-the-Shelf Generation Techniques",
                        "abstract": "Diffusion models are powerful generative models, and this capability can also be applied to discrimination. The inner activations of a pre-trained diffusion model can serve as features for discriminative tasks, namely, diffusion feature. We discover that diffusion feature has been hindered by a hidden yet universal phenomenon that we call content shift. To be specific, there are content differences between features and the input image, such as the exact shape of a certain object. We locate the cause of content shift as one inherent characteristic of diffusion models, which suggests the broad existence of this phenomenon in diffusion feature. Further empirical study also indicates that its negative impact is not negligible even when content shift is not visually perceivable. Hence, we propose to suppress content shift to enhance the overall quality of diffusion features. Specifically, content shift is related to the information drift during the process of recovering an image from the noisy input, pointing out the possibility of turning off-the-shelf generation techniques into tools for content shift suppression. We further propose a practical guideline named GATE to efficiently evaluate the potential benefit of a technique and provide an implementation of our methodology. Despite the simplicity, the proposed approach has achieved superior results on various tasks and datasets, validating its potential as a generic booster for diffusion features. Our code is available at https://github.com/Darkbblue/diffusion-content-shift."
                    },
                    {
                        "title": "A Unified Generalization Analysis of Re-Weighting and Logit-Adjustment for Imbalanced Learning",
                        "abstract": "Real-world datasets are typically imbalanced in the sense that only a few classes have numerous samples, while many classes are associated with only a few samples. As a result, a na\\\"ive ERM learning process will be biased towards the majority classes, making it difficult to generalize to the minority classes. To address this issue, one simple but effective approach is to modify the loss function to emphasize the learning on minority classes, such as re-weighting the losses or adjusting the logits via class-dependent terms. However, existing generalization analysis of such losses is still coarse-grained and fragmented, failing to explain some empirical results. To bridge this gap, we propose a novel technique named data-dependent contraction to capture how these modified losses handle different classes. On top of this technique, a fine-grained generalization bound is established for imbalanced learning, which helps reveal the mystery of re-weighting and logit-adjustment in a unified manner. Furthermore, a principled learning algorithm is developed based on the theoretical insights. Finally, the empirical results on benchmark datasets not only validate the theoretical results but also demonstrate the effectiveness of the proposed method."
                    },
                    {
                        "title": "Harnessing Hierarchical Label Distribution Variations in Test Agnostic Long-tail Recognition",
                        "abstract": "This paper explores test-agnostic long-tail recognition, a challenging long-tail task where the test label distributions are unknown and arbitrarily imbalanced. We argue that the variation in these distributions can be broken down hierarchically into global and local levels. The global ones reflect a broad range of diversity, while the local ones typically arise from milder changes, often focused on a particular neighbor. Traditional methods predominantly use a Mixture-of-Expert (MoE) approach, targeting a few fixed test label distributions that exhibit substantial global variations. However, the local variations are left unconsidered. To address this issue, we propose a new MoE strategy, $\\mathsf{DirMixE}$, which assigns experts to different Dirichlet meta-distributions of the label distribution, each targeting a specific aspect of local variations. Additionally, the diversity among these Dirichlet meta-distributions inherently captures global variations. This dual-level approach also leads to a more stable objective function, allowing us to sample different test distributions better to quantify the mean and variance of performance outcomes. Theoretically, we show that our proposed objective benefits from enhanced generalization by virtue of the variance-based regularization. Comprehensive experiments across multiple benchmarks confirm the effectiveness of $\\mathsf{DirMixE}$. The code is available at \\url{https://github.com/scongl/DirMixE}."
                    }
                ]
            },
            "de1d76df-b20e-4068-addd-f2edb921ba94": {
                "pk": "de1d76df-b20e-4068-addd-f2edb921ba94",
                "name": "Xiaochun Cao",
                "collaborators": [
                    "Qianqian Xu",
                    "Zhiyong Yang",
                    "Qingming Huang",
                    "Xingxing Wei",
                    "Xiaojun Jia",
                    "Pengwen Dai",
                    "Yuanfang Guo",
                    "Wei Zhang",
                    "Rui Wang",
                    "Hassan Foroosh"
                ],
                "domain": [
                    "Computer Vision",
                    "Adversarial Learning",
                    "Multi-Task Learning",
                    "Knowledge Graphs"
                ],
                "publications": [
                    {
                        "title": "Comprehensive Studies for Arbitrary-shape Scene Text Detection",
                        "abstract": "Numerous scene text detection methods have been proposed in recent years. Most of them declare they have achieved state-of-the-art performances. However, the performance comparison is unfair, due to lots of inconsistent settings (e.g., training data, backbone network, multi-scale feature fusion, evaluation protocols, etc.). These various settings would dissemble the pros and cons of the proposed core techniques. In this paper, we carefully examine and analyze the inconsistent settings, and propose a unified framework for the bottom-up based scene text detection methods. Under the unified framework, we ensure the consistent settings for non-core modules, and mainly investigate the representations of describing arbitrary-shape scene texts, e.g., regressing points on text contours, clustering pixels with predicted auxiliary information, grouping connected components with learned linkages, etc. With the comprehensive investigations and elaborate analyses, it not only cleans up the obstacle of understanding the performance differences between existing methods but also reveals the advantages and disadvantages of previous models under fair comparisons."
                    },
                    {
                        "title": "From Common to Special: When Multi-Attribute Learning Meets Personalized Opinions",
                        "abstract": "Visual attributes, which refer to human-labeled semantic annotations, have gained increasing popularity in a wide range of real world applications. Generally, the existing attribute learning methods fall into two categories: one focuses on learning user-specific labels separately for different attributes, while the other one focuses on learning crowd-sourced global labels jointly for multiple attributes. However, both categories ignore the joint effect of the two mentioned factors: the personal diversity with respect to the global consensus; and the intrinsic correlation among multiple attributes. To overcome this challenge, we propose a novel model to learn user-specific predictors across multiple attributes. In our proposed model, the diversity of personalized opinions and the intrinsic relationship among multiple attributes are unified in a common-to-special manner. To this end, we adopt a three-component decomposition. Specifically, our model integrates a common cognition factor, an attribute-specific bias factor and a user-specific bias factor. Meanwhile Lasso and group Lasso penalties are adopted to leverage efficient feature selection. Furthermore, theoretical analysis is conducted to show that our proposed method could reach reasonable performance. Eventually, the empirical study carried out in this paper demonstrates the effectiveness of our proposed method."
                    },
                    {
                        "title": "Fake Colorized Image Detection",
                        "abstract": "Image forensics aims to detect the manipulation of digital images. Currently, splicing detection, copy-move detection and image retouching detection are drawing much attentions from researchers. However, image editing techniques develop with time goes by. One emerging image editing technique is colorization, which can colorize grayscale images with realistic colors. Unfortunately, this technique may also be intentionally applied to certain images to confound object recognition algorithms. To the best of our knowledge, no forensic technique has yet been invented to identify whether an image is colorized. We observed that, compared to natural images, colorized images, which are generated by three state-of-the-art methods, possess statistical differences for the hue and saturation channels. Besides, we also observe statistical inconsistencies in the dark and bright channels, because the colorization process will inevitably affect the dark and bright channel values. Based on our observations, i.e., potential traces in the hue, saturation, dark and bright channels, we propose two simple yet effective detection methods for fake colorized images: Histogram based Fake Colorized Image Detection (FCID-HIST) and Feature Encoding based Fake Colorized Image Detection (FCID-FE). Experimental results demonstrate that both proposed methods exhibit a decent performance against multiple state-of-the-art colorization approaches."
                    },
                    {
                        "title": "ComDefend: An Efficient Image Compression Model to Defend Adversarial Examples",
                        "abstract": "Deep neural networks (DNNs) have been demonstrated to be vulnerable to adversarial examples. Specifically, adding imperceptible perturbations to clean images can fool the well trained deep neural networks. In this paper, we propose an end-to-end image compression model to defend adversarial examples: \\textbf{ComDefend}. The proposed model consists of a compression convolutional neural network (ComCNN) and a reconstruction convolutional neural network (ResCNN). The ComCNN is used to maintain the structure information of the original image and purify adversarial perturbations. And the ResCNN is used to reconstruct the original image with high quality. In other words, ComDefend can transform the adversarial image to its clean version, which is then fed to the trained classifier. Our method is a pre-processing module, and does not modify the classifier's structure during the whole process. Therefore, it can be combined with other model-specific defense models to jointly improve the classifier's robustness. A series of experiments conducted on MNIST, CIFAR10 and ImageNet show that the proposed method outperforms the state-of-the-art defense methods, and is consistently effective to protect classifiers against adversarial attacks."
                    },
                    {
                        "title": "Robust Ordinal Embedding from Contaminated Relative Comparisons",
                        "abstract": "Existing ordinal embedding methods usually follow a two-stage routine: outlier detection is first employed to pick out the inconsistent comparisons; then an embedding is learned from the clean data. However, learning in a multi-stage manner is well-known to suffer from sub-optimal solutions. In this paper, we propose a unified framework to jointly identify the contaminated comparisons and derive reliable embeddings. The merits of our method are three-fold: (1) By virtue of the proposed unified framework, the sub-optimality of traditional methods is largely alleviated; (2) The proposed method is aware of global inconsistency by minimizing a corresponding cost, while traditional methods only involve local inconsistency; (3) Instead of considering the nuclear norm heuristics, we adopt an exact solution for rank equality constraint. Our studies are supported by experiments with both simulated examples and real-world data. The proposed framework provides us a promising tool for robust ordinal embedding from the contaminated comparisons."
                    },
                    {
                        "title": "False Discovery Rate Control and Statistical Quality Assessment of Annotators in Crowdsourced Ranking",
                        "abstract": "With the rapid growth of crowdsourcing platforms it has become easy and relatively inexpensive to collect a dataset labeled by multiple annotators in a short time. However due to the lack of control over the quality of the annotators, some abnormal annotators may be affected by position bias which can potentially degrade the quality of the final consensus labels. In this paper we introduce a statistical framework to model and detect annotator's position bias in order to control the false discovery rate (FDR) without a prior knowledge on the amount of biased annotators - the expected fraction of false discoveries among all discoveries being not too high, in order to assure that most of the discoveries are indeed true and replicable. The key technical development relies on some new knockoff filters adapted to our problem and new algorithms based on the Inverse Scale Space dynamics whose discretization is potentially suitable for large scale crowdsourcing data analysis. Our studies are supported by experiments with both simulated examples and real-world data. The proposed framework provides us a useful tool for quantitatively studying annotator's abnormal behavior in crowdsourcing data arising from machine learning, sociology, computer vision, multimedia, etc."
                    },
                    {
                        "title": "Transferable Adversarial Attacks for Image and Video Object Detection",
                        "abstract": "Adversarial examples have been demonstrated to threaten many computer vision tasks including object detection. However, the existing attacking methods for object detection have two limitations: poor transferability, which denotes that the generated adversarial examples have low success rate to attack other kinds of detection methods, and high computation cost, which means that they need more time to generate an adversarial image, and therefore are difficult to deal with the video data. To address these issues, we utilize a generative mechanism to obtain the adversarial image and video. In this way, the processing time is reduced. To enhance the transferability, we destroy the feature maps extracted from the feature network, which usually constitutes the basis of object detectors. The proposed method is based on the Generative Adversarial Network (GAN) framework, where we combine the high-level class loss and low-level feature loss to jointly train the adversarial example generator. A series of experiments conducted on PASCAL VOC and ImageNet VID datasets show that our method can efficiently generate image and video adversarial examples, and more importantly, these adversarial examples have better transferability, and thus, are able to simultaneously attack two kinds of representative object detection models: proposal based models like Faster-RCNN, and regression based models like SSD."
                    },
                    {
                        "title": "Learning Personalized Attribute Preference via Multi-task AUC Optimization",
                        "abstract": "Traditionally, most of the existing attribute learning methods are trained based on the consensus of annotations aggregated from a limited number of annotators. However, the consensus might fail in settings, especially when a wide spectrum of annotators with different interests and comprehension about the attribute words are involved. In this paper, we develop a novel multi-task method to understand and predict personalized attribute annotations. Regarding the attribute preference learning for each annotator as a specific task, we first propose a multi-level task parameter decomposition to capture the evolution from a highly popular opinion of the mass to highly personalized choices that are special for each person. Meanwhile, for personalized learning methods, ranking prediction is much more important than accurate classification. This motivates us to employ an Area Under ROC Curve (AUC) based loss function to improve our model. On top of the AUC-based loss, we propose an efficient method to evaluate the loss and gradients. Theoretically, we propose a novel closed-form solution for one of our non-convex subproblem, which leads to provable convergence behaviors. Furthermore, we also provide a generalization bound to guarantee a reasonable performance. Finally, empirical analysis consistently speaks to the efficacy of our proposed method."
                    },
                    {
                        "title": "Task-Feature Collaborative Learning with Application to Personalized Attribute Prediction",
                        "abstract": "As an effective learning paradigm against insufficient training samples, Multi-Task Learning (MTL) encourages knowledge sharing across multiple related tasks so as to improve the overall performance. In MTL, a major challenge springs from the phenomenon that sharing the knowledge with dissimilar and hard tasks, known as negative transfer, often results in a worsened performance. Though a substantial amount of studies have been carried out against the negative transfer, most of the existing methods only model the transfer relationship as task correlations, with the transfer across features and tasks left unconsidered. Different from the existing methods, our goal is to alleviate negative transfer collaboratively across features and tasks. To this end, we propose a novel multi-task learning method called Task-Feature Collaborative Learning (TFCL). Specifically, we first propose a base model with a heterogeneous block-diagonal structure regularizer to leverage the collaborative grouping of features and tasks and suppressing inter-group knowledge sharing. We then propose an optimization method for the model. Extensive theoretical analysis shows that our proposed method has the following benefits: (a) it enjoys the global convergence property and (b) it provides a block-diagonal structure recovery guarantee. As a practical extension, we extend the base model by allowing overlapping features and differentiating the hard tasks. We further apply it to the personalized attribute prediction problem with fine-grained modeling of user behaviors. Finally, experimental results on both simulated dataset and real-world datasets demonstrate the effectiveness of our proposed method"
                    },
                    {
                        "title": "Identifying and Resisting Adversarial Videos Using Temporal Consistency",
                        "abstract": "Video classification is a challenging task in computer vision. Although Deep Neural Networks (DNNs) have achieved excellent performance in video classification, recent research shows adding imperceptible perturbations to clean videos can make the well-trained models output wrong labels with high confidence. In this paper, we propose an effective defense framework to characterize and defend adversarial videos. The proposed method contains two phases: (1) adversarial video detection using temporal consistency between adjacent frames, and (2) adversarial perturbation reduction via denoisers in the spatial and temporal domains respectively. Specifically, because of the linear nature of DNNs, the imperceptible perturbations will enlarge with the increasing of DNNs depth, which leads to the inconsistency of DNNs output between adjacent frames. However, the benign video frames often have the same outputs with their neighbor frames owing to the slight changes. Based on this observation, we can distinguish between adversarial videos and benign videos. After that, we utilize different defense strategies against different attacks. We propose the temporal defense, which reconstructs the polluted frames with their temporally neighbor clean frames, to deal with the adversarial videos with sparse polluted frames. For the videos with dense polluted frames, we use an efficient adversarial denoiser to process each frame in the spatial domain, and thus purify the perturbations (we call it as spatial defense). A series of experiments conducted on the UCF-101 dataset demonstrate that the proposed method significantly improves the robustness of video classifiers against adversarial attacks."
                    },
                    {
                        "title": "Geometry Interaction Knowledge Graph Embeddings",
                        "abstract": "Knowledge graph (KG) embeddings have shown great power in learning representations of entities and relations for link prediction tasks. Previous work usually embeds KGs into a single geometric space such as Euclidean space (zero curved), hyperbolic space (negatively curved) or hyperspherical space (positively curved) to maintain their specific geometric structures (e.g., chain, hierarchy and ring structures). However, the topological structure of KGs appears to be complicated, since it may contain multiple types of geometric structures simultaneously. Therefore, embedding KGs in a single space, no matter the Euclidean space, hyperbolic space or hyperspheric space, cannot capture the complex structures of KGs accurately. To overcome this challenge, we propose Geometry Interaction knowledge graph Embeddings (GIE), which learns spatial structures interactively between the Euclidean, hyperbolic and hyperspherical spaces. Theoretically, our proposed GIE can capture a richer set of relational information, model key inference patterns, and enable expressive semantic matching across entities. Experimental results on three well-established knowledge graph completion benchmarks show that our GIE achieves the state-of-the-art performance with fewer parameters."
                    }
                ]
            },
            "659f1755-e593-452d-bdf5-0eb537f23d19": {
                "pk": "659f1755-e593-452d-bdf5-0eb537f23d19",
                "name": "Qingming Huang",
                "collaborators": [
                    "Junbiao Pang",
                    "Qianqian Xu",
                    "Zhiyong Yang",
                    "Xiaochun Cao",
                    "Tao Hu",
                    "Honggang Qi",
                    "Shuhui Wang",
                    "Jiaqi Wu",
                    "Baochang Zhang",
                    "Li Shen"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Multi-Task Learning",
                    "Data Augmentation"
                ],
                "publications": [
                    {
                        "title": "Towards Scalable Topic Detection on Web via Simulating Levy Walks Nature of Topics in Similarity Space",
                        "abstract": "Organizing a few webpages from social media websites into popular topics is one of the key steps to understand trends on web. Discovering popular topics from web faces a sea of noise webpages which never evolve into popular topics. In this paper, we discover that the similarity values between webpages in a popular topic contain the statistically similar features observed in Levy walks. Consequently, we present a simple, novel, yet very powerful Explore-Exploit (EE) approach to group topics by simulating Levy walks nature in the similarity space. The proposed EE-based topic clustering is an effective and effcient method which is a solid move towards handling a sea of noise webpages. Experiments on two public data sets demonstrate that our approach is not only comparable to the state-of-the-art methods in terms of effectiveness but also significantly outperforms the state-of-the-art methods in terms of efficiency."
                    },
                    {
                        "title": "Relay Backpropagation for Effective Learning of Deep Convolutional Neural Networks",
                        "abstract": "Learning deeper convolutional neural networks becomes a tendency in recent years. However, many empirical evidences suggest that performance improvement cannot be gained by simply stacking more layers. In this paper, we consider the issue from an information theoretical perspective, and propose a novel method Relay Backpropagation, that encourages the propagation of effective information through the network in training stage. By virtue of the method, we achieved the first place in ILSVRC 2015 Scene Classification Challenge. Extensive experiments on two challenging large scale datasets demonstrate the effectiveness of our method is not restricted to a specific dataset or network architecture. Our models will be available to the research community later."
                    },
                    {
                        "title": "From Common to Special: When Multi-Attribute Learning Meets Personalized Opinions",
                        "abstract": "Visual attributes, which refer to human-labeled semantic annotations, have gained increasing popularity in a wide range of real world applications. Generally, the existing attribute learning methods fall into two categories: one focuses on learning user-specific labels separately for different attributes, while the other one focuses on learning crowd-sourced global labels jointly for multiple attributes. However, both categories ignore the joint effect of the two mentioned factors: the personal diversity with respect to the global consensus; and the intrinsic correlation among multiple attributes. To overcome this challenge, we propose a novel model to learn user-specific predictors across multiple attributes. In our proposed model, the diversity of personalized opinions and the intrinsic relationship among multiple attributes are unified in a common-to-special manner. To this end, we adopt a three-component decomposition. Specifically, our model integrates a common cognition factor, an attribute-specific bias factor and a user-specific bias factor. Meanwhile Lasso and group Lasso penalties are adopted to leverage efficient feature selection. Furthermore, theoretical analysis is conducted to show that our proposed method could reach reasonable performance. Eventually, the empirical study carried out in this paper demonstrates the effectiveness of our proposed method."
                    },
                    {
                        "title": "See Better Before Looking Closer: Weakly Supervised Data Augmentation Network for Fine-Grained Visual Classification",
                        "abstract": "Data augmentation is usually adopted to increase the amount of training data, prevent overfitting and improve the performance of deep models. However, in practice, random data augmentation, such as random image cropping, is low-efficiency and might introduce many uncontrolled background noises. In this paper, we propose Weakly Supervised Data Augmentation Network (WS-DAN) to explore the potential of data augmentation. Specifically, for each training image, we first generate attention maps to represent the object's discriminative parts by weakly supervised learning. Next, we augment the image guided by these attention maps, including attention cropping and attention dropping. The proposed WS-DAN improves the classification accuracy in two folds. In the first stage, images can be seen better since more discriminative parts' features will be extracted. In the second stage, attention regions provide accurate location of object, which ensures our model to look at the object closer and further improve the performance. Comprehensive experiments in common fine-grained visual classification datasets show that our WS-DAN surpasses the state-of-the-art methods, which demonstrates its effectiveness."
                    },
                    {
                        "title": "Cascaded Partial Decoder for Fast and Accurate Salient Object Detection",
                        "abstract": "Existing state-of-the-art salient object detection networks rely on aggregating multi-level features of pre-trained convolutional neural networks (CNNs). Compared to high-level features, low-level features contribute less to performance but cost more computations because of their larger spatial resolutions. In this paper, we propose a novel Cascaded Partial Decoder (CPD) framework for fast and accurate salient object detection. On the one hand, the framework constructs partial decoder which discards larger resolution features of shallower layers for acceleration. On the other hand, we observe that integrating features of deeper layers obtain relatively precise saliency map. Therefore we directly utilize generated saliency map to refine the features of backbone network. This strategy efficiently suppresses distractors in the features and significantly improves their representation ability. Experiments conducted on five benchmark datasets exhibit that the proposed model not only achieves state-of-the-art performance but also runs much faster than existing models. Besides, the proposed framework is further applied to improve existing multi-level feature aggregation models and significantly improve their efficiency and accuracy."
                    },
                    {
                        "title": "Facial Landmarks Detection by Self-Iterative Regression based Landmarks-Attention Network",
                        "abstract": "Cascaded Regression (CR) based methods have been proposed to solve facial landmarks detection problem, which learn a series of descent directions by multiple cascaded regressors separately trained in coarse and fine stages. They outperform the traditional gradient descent based methods in both accuracy and running speed. However, cascaded regression is not robust enough because each regressor's training data comes from the output of previous regressor. Moreover, training multiple regressors requires lots of computing resources, especially for deep learning based methods. In this paper, we develop a Self-Iterative Regression (SIR) framework to improve the model efficiency. Only one self-iterative regressor is trained to learn the descent directions for samples from coarse stages to fine stages, and parameters are iteratively updated by the same regressor. Specifically, we proposed Landmarks-Attention Network (LAN) as our regressor, which concurrently learns features around each landmark and obtains the holistic location increment. By doing so, not only the rest of regressors are removed to simplify the training process, but the number of model parameters is significantly decreased. The experiments demonstrate that with only 3.72M model parameters, our proposed method achieves the state-of-the-art performance."
                    },
                    {
                        "title": "Harmonized Multimodal Learning with Gaussian Process Latent Variable Models",
                        "abstract": "Multimodal learning aims to discover the relationship between multiple modalities. It has become an important research topic due to extensive multimodal applications such as cross-modal retrieval. This paper attempts to address the modality heterogeneity problem based on Gaussian process latent variable models (GPLVMs) to represent multimodal data in a common space. Previous multimodal GPLVM extensions generally adopt individual learning schemes on latent representations and kernel hyperparameters, which ignore their intrinsic relationship. To exploit strong complementarity among different modalities and GPLVM components, we develop a novel learning scheme called Harmonization, where latent model parameters are jointly learned from each other. Beyond the correlation fitting or intra-modal structure preservation paradigms widely used in existing studies, the harmonization is derived in a model-driven manner to encourage the agreement between modality-specific GP kernels and the similarity of latent representations. We present a range of multimodal learning models by incorporating the harmonization mechanism into several representative GPLVM-based approaches. Experimental results on four benchmark datasets show that the proposed models outperform the strong baselines for cross-modal retrieval tasks, and that the harmonized multimodal learning method is superior in discovering semantically consistent latent representation."
                    },
                    {
                        "title": "RAM: A Region-Aware Deep Model for Vehicle Re-Identification",
                        "abstract": "Previous works on vehicle Re-ID mainly focus on extracting global features and learning distance metrics. Because some vehicles commonly share same model and maker, it is hard to distinguish them based on their global appearances. Compared with the global appearance, local regions such as decorations and inspection stickers attached to the windshield, may be more distinctive for vehicle Re-ID. To embed the detailed visual cues in those local regions, we propose a Region-Aware deep Model (RAM). Specifically, in addition to extracting global features, RAM also extracts features from a series of local regions. As each local region conveys more distinctive visual cues, RAM encourages the deep model to learn discriminative features. We also introduce a novel learning algorithm to jointly use vehicle IDs, types/models, and colors to train the RAM. This strategy fuses more cues for training and results in more discriminative global and regional features. We evaluate our methods on two large-scale vehicle Re-ID datasets, i.e., VeRi and VehicleID. Experimental results show our methods achieve promising performance in comparison with recent works."
                    },
                    {
                        "title": "ER: Equivariance Regularizer for Knowledge Graph Completion",
                        "abstract": "Tensor factorization and distanced based models play important roles in knowledge graph completion (KGC). However, the relational matrices in KGC methods often induce a high model complexity, bearing a high risk of overfitting. As a remedy, researchers propose a variety of different regularizers such as the tensor nuclear norm regularizer. Our motivation is based on the observation that the previous work only focuses on the \"size\" of the parametric space, while leaving the implicit semantic information widely untouched. To address this issue, we propose a new regularizer, namely, Equivariance Regularizer (ER), which can suppress overfitting by leveraging the implicit semantic information. Specifically, ER can enhance the generalization ability of the model by employing the semantic equivariance between the head and tail entities. Moreover, it is a generic solution for both distance based models and tensor factorization based models. The experimental results indicate a clear and substantial improvement over the state-of-the-art relation prediction methods."
                    },
                    {
                        "title": "Modeling the Uncertainty with Maximum Discrepant Students for Semi-supervised 2D Pose Estimation",
                        "abstract": "Semi-supervised pose estimation is a practically challenging task for computer vision. Although numerous excellent semi-supervised classification methods have emerged, these methods typically use confidence to evaluate the quality of pseudo-labels, which is difficult to achieve in pose estimation tasks. For example, in pose estimation, confidence represents only the possibility that a position of the heatmap is a keypoint, not the quality of that prediction. In this paper, we propose a simple yet efficient framework to estimate the quality of pseudo-labels in semi-supervised pose estimation tasks from the perspective of modeling the uncertainty of the pseudo-labels. Concretely, under the dual mean-teacher framework, we construct the two maximum discrepant students (MDSs) to effectively push two teachers to generate different decision boundaries for the same sample. Moreover, we create multiple uncertainties to assess the quality of the pseudo-labels. Experimental results demonstrate that our method improves the performance of semi-supervised pose estimation on three datasets."
                    },
                    {
                        "title": "Pick-and-Draw: Training-free Semantic Guidance for Text-to-Image Personalization",
                        "abstract": "Diffusion-based text-to-image personalization have achieved great success in generating subjects specified by users among various contexts. Even though, existing finetuning-based methods still suffer from model overfitting, which greatly harms the generative diversity, especially when given subject images are few. To this end, we propose Pick-and-Draw, a training-free semantic guidance approach to boost identity consistency and generative diversity for personalization methods. Our approach consists of two components: appearance picking guidance and layout drawing guidance. As for the former, we construct an appearance palette with visual features from the reference image, where we pick local patterns for generating the specified subject with consistent identity. As for layout drawing, we outline the subject's contour by referring to a generative template from the vanilla diffusion model, and inherit the strong image prior to synthesize diverse contexts according to different text conditions. The proposed approach can be applied to any personalized diffusion models and requires as few as a single reference image. Qualitative and quantitative experiments show that Pick-and-Draw consistently improves identity consistency and generative diversity, pushing the trade-off between subject fidelity and image-text fidelity to a new Pareto frontier."
                    },
                    {
                        "title": "A Channel-ensemble Approach: Unbiased and Low-variance Pseudo-labels is Critical for Semi-supervised Classification",
                        "abstract": "Semi-supervised learning (SSL) is a practical challenge in computer vision. Pseudo-label (PL) methods, e.g., FixMatch and FreeMatch, obtain the State Of The Art (SOTA) performances in SSL. These approaches employ a threshold-to-pseudo-label (T2L) process to generate PLs by truncating the confidence scores of unlabeled data predicted by the self-training method. However, self-trained models typically yield biased and high-variance predictions, especially in the scenarios when a little labeled data are supplied. To address this issue, we propose a lightweight channel-based ensemble method to effectively consolidate multiple inferior PLs into the theoretically guaranteed unbiased and low-variance one. Importantly, our approach can be readily extended to any SSL framework, such as FixMatch or FreeMatch. Experimental results demonstrate that our method significantly outperforms state-of-the-art techniques on CIFAR10/100 in terms of effectiveness and efficiency."
                    },
                    {
                        "title": "Bundle Fragments into a Whole: Mining More Complete Clusters via Submodular Selection of Interesting webpages for Web Topic Detection",
                        "abstract": "Organizing interesting webpages into hot topics is one of key steps to understand the trends of multimodal web data. A state-of-the-art solution is firstly to organize webpages into a large volume of multi-granularity topic candidates; hot topics are further identified by estimating their interestingness. However, these topic candidates contain a large number of fragments of hot topics due to both the inefficient feature representations and the unsupervised topic generation. This paper proposes a bundling-refining approach to mine more complete hot topics from fragments. Concretely, the bundling step organizes the fragment topics into coarse topics; next, the refining step proposes a submodular-based method to refine coarse topics in a scalable approach. The propose unconventional method is simple, yet powerful by leveraging submodular optimization, our approach outperforms the traditional ranking methods which involve the careful design and complex steps. Extensive experiments demonstrate that the proposed approach surpasses the state-of-the-art method (i.e., latent Poisson deconvolution Pang et al. (2016)) 20% accuracy and 10% one on two public data sets, respectively."
                    },
                    {
                        "title": "Bilateral Sharpness-Aware Minimization for Flatter Minima",
                        "abstract": "Sharpness-Aware Minimization (SAM) enhances generalization by reducing a Max-Sharpness (MaxS). Despite the practical success, we empirically found that the MAxS behind SAM's generalization enhancements face the \"Flatness Indicator Problem\" (FIP), where SAM only considers the flatness in the direction of gradient ascent, resulting in a next minimization region that is not sufficiently flat. A better Flatness Indicator (FI) would bring a better generalization of neural networks. Because SAM is a greedy search method in nature. In this paper, we propose to utilize the difference between the training loss and the minimum loss over the neighborhood surrounding the current weight, which we denote as Min-Sharpness (MinS). By merging MaxS and MinS, we created a better FI that indicates a flatter direction during the optimization. Specially, we combine this FI with SAM into the proposed Bilateral SAM (BSAM) which finds a more flatter minimum than that of SAM. The theoretical analysis proves that BSAM converges to local minima. Extensive experiments demonstrate that BSAM offers superior generalization performance and robustness compared to vanilla SAM across various tasks, i.e., classification, transfer learning, human pose estimation, and network quantization. Code is publicly available at: https://github.com/ajiaaa/BSAM."
                    },
                    {
                        "title": "Learning Personalized Attribute Preference via Multi-task AUC Optimization",
                        "abstract": "Traditionally, most of the existing attribute learning methods are trained based on the consensus of annotations aggregated from a limited number of annotators. However, the consensus might fail in settings, especially when a wide spectrum of annotators with different interests and comprehension about the attribute words are involved. In this paper, we develop a novel multi-task method to understand and predict personalized attribute annotations. Regarding the attribute preference learning for each annotator as a specific task, we first propose a multi-level task parameter decomposition to capture the evolution from a highly popular opinion of the mass to highly personalized choices that are special for each person. Meanwhile, for personalized learning methods, ranking prediction is much more important than accurate classification. This motivates us to employ an Area Under ROC Curve (AUC) based loss function to improve our model. On top of the AUC-based loss, we propose an efficient method to evaluate the loss and gradients. Theoretically, we propose a novel closed-form solution for one of our non-convex subproblem, which leads to provable convergence behaviors. Furthermore, we also provide a generalization bound to guarantee a reasonable performance. Finally, empirical analysis consistently speaks to the efficacy of our proposed method."
                    },
                    {
                        "title": "Multimodal Transformer with Multi-View Visual Representation for Image Captioning",
                        "abstract": "Image captioning aims to automatically generate a natural language description of a given image, and most state-of-the-art models have adopted an encoder-decoder framework. The framework consists of a convolution neural network (CNN)-based image encoder that extracts region-based visual features from the input image, and an recurrent neural network (RNN)-based caption decoder that generates the output caption words based on the visual features with the attention mechanism. Despite the success of existing studies, current methods only model the co-attention that characterizes the inter-modal interactions while neglecting the self-attention that characterizes the intra-modal interactions. Inspired by the success of the Transformer model in machine translation, here we extend it to a Multimodal Transformer (MT) model for image captioning. Compared to existing image captioning approaches, the MT model simultaneously captures intra- and inter-modal interactions in a unified attention block. Due to the in-depth modular composition of such attention blocks, the MT model can perform complex multimodal reasoning and output accurate captions. Moreover, to further improve the image captioning performance, multi-view visual features are seamlessly introduced into the MT model. We quantitatively and qualitatively evaluate our approach using the benchmark MSCOCO image captioning dataset and conduct extensive ablation studies to investigate the reasons behind its effectiveness. The experimental results show that our method significantly outperforms the previous state-of-the-art methods. With an ensemble of seven models, our solution ranks the 1st place on the real-time leaderboard of the MSCOCO image captioning challenge at the time of the writing of this paper."
                    },
                    {
                        "title": "Global Context-Aware Progressive Aggregation Network for Salient Object Detection",
                        "abstract": "Deep convolutional neural networks have achieved competitive performance in salient object detection, in which how to learn effective and comprehensive features plays a critical role. Most of the previous works mainly adopted multiple level feature integration yet ignored the gap between different features. Besides, there also exists a dilution process of high-level features as they passed on the top-down pathway. To remedy these issues, we propose a novel network named GCPANet to effectively integrate low-level appearance features, high-level semantic features, and global context features through some progressive context-aware Feature Interweaved Aggregation (FIA) modules and generate the saliency map in a supervised way. Moreover, a Head Attention (HA) module is used to reduce information redundancy and enhance the top layers features by leveraging the spatial and channel-wise attention, and the Self Refinement (SR) module is utilized to further refine and heighten the input features. Furthermore, we design the Global Context Flow (GCF) module to generate the global context information at different stages, which aims to learn the relationship among different salient regions and alleviate the dilution effect of high-level features. Experimental results on six benchmark datasets demonstrate that the proposed approach outperforms the state-of-the-art methods both quantitatively and qualitatively."
                    },
                    {
                        "title": "Task-Feature Collaborative Learning with Application to Personalized Attribute Prediction",
                        "abstract": "As an effective learning paradigm against insufficient training samples, Multi-Task Learning (MTL) encourages knowledge sharing across multiple related tasks so as to improve the overall performance. In MTL, a major challenge springs from the phenomenon that sharing the knowledge with dissimilar and hard tasks, known as negative transfer, often results in a worsened performance. Though a substantial amount of studies have been carried out against the negative transfer, most of the existing methods only model the transfer relationship as task correlations, with the transfer across features and tasks left unconsidered. Different from the existing methods, our goal is to alleviate negative transfer collaboratively across features and tasks. To this end, we propose a novel multi-task learning method called Task-Feature Collaborative Learning (TFCL). Specifically, we first propose a base model with a heterogeneous block-diagonal structure regularizer to leverage the collaborative grouping of features and tasks and suppressing inter-group knowledge sharing. We then propose an optimization method for the model. Extensive theoretical analysis shows that our proposed method has the following benefits: (a) it enjoys the global convergence property and (b) it provides a block-diagonal structure recovery guarantee. As a practical extension, we extend the base model by allowing overlapping features and differentiating the hard tasks. We further apply it to the personalized attribute prediction problem with fine-grained modeling of user behaviors. Finally, experimental results on both simulated dataset and real-world datasets demonstrate the effectiveness of our proposed method"
                    },
                    {
                        "title": "F3Net: Fusion, Feedback and Focus for Salient Object Detection",
                        "abstract": "Most of existing salient object detection models have achieved great progress by aggregating multi-level features extracted from convolutional neural networks. However, because of the different receptive fields of different convolutional layers, there exists big differences between features generated by these layers. Common feature fusion strategies (addition or concatenation) ignore these differences and may cause suboptimal solutions. In this paper, we propose the F3Net to solve above problem, which mainly consists of cross feature module (CFM) and cascaded feedback decoder (CFD) trained by minimizing a new pixel position aware loss (PPA). Specifically, CFM aims to selectively aggregate multi-level features. Different from addition and concatenation, CFM adaptively selects complementary components from input features before fusion, which can effectively avoid introducing too much redundant information that may destroy the original features. Besides, CFD adopts a multi-stage feedback mechanism, where features closed to supervision will be introduced to the output of previous layers to supplement them and eliminate the differences between features. These refined features will go through multiple similar iterations before generating the final saliency maps. Furthermore, different from binary cross entropy, the proposed PPA loss doesn't treat pixels equally, which can synthesize the local structure information of a pixel to guide the network to focus more on local details. Hard pixels from boundaries or error-prone parts will be given more attention to emphasize their importance. F3Net is able to segment salient object regions accurately and provide clear local details. Comprehensive experiments on five benchmark datasets demonstrate that F3Net outperforms state-of-the-art approaches on six evaluation metrics."
                    }
                ]
            }
        }
    },
    "2405.19562": {
        "paper_data": {
            "title": "Selective Explanations",
            "url": "http://arxiv.org/abs/2405.19562v1",
            "arxiv_id": "2405.19562",
            "authors": [
                "Lucas Monteiro Paes",
                "Dennis Wei",
                "Flavio P. Calmon"
            ],
            "abstract": "Feature attribution methods explain black-box machine learning (ML) models by assigning importance scores to input features. These methods can be computationally expensive for large ML models. To address this challenge, there has been increasing efforts to develop amortized explainers, where a machine learning model is trained to predict feature attribution scores with only one inference. Despite their efficiency, amortized explainers can produce inaccurate predictions and misleading explanations. In this paper, we propose selective explanations, a novel feature attribution method that (i) detects when amortized explainers generate low-quality explanations and (ii) improves these explanations using a technique called explanations with initial guess. Our selective explanation method allows practitioners to specify the fraction of samples that receive explanations with initial guess, offering a principled way to bridge the gap between amortized explainers and their high-quality counterparts.",
            "introduction": "   1 Introduction  Large black-box models are increasingly used to support decisions in applications ranging from online content moderation [1], hiring [2], and medical diagnostics [3]. In such high-stakes settings, the need for explaining \u201cwhy\u201d a model produces a given output has led to a growing number of perturbation-based feature attribution methods [4, 5, 6, 7, 8, 9]. Broadly speaking, these methods use input perturbations to assign numerical values to each input feature a model uses, indicating their influence on predictions. They are widely adopted in part because they work in the black-box setting with access only to model outputs (i.e., no gradients). However, existing feature attribution methods can be prohibitively expensive for the large models used in the current machine learning landscape (e.g., language models with billions of parameters) since they require a significant number of inferences for each individual explanation.   Recent literature has introduced two main strategies for speeding up feature attribution for large models: (i) employing Monte Carlo methods to approximate explanations with fewer computations [4, 5, 10, 11], and (ii) adopting an amortized approach, training a separate model to \u201cmimic\u201d the outputs of a reference explanation method [12, 13, 14, 15, 16, 17]. Monte Carlo approximations can yield high-quality explanations but may converge slowly, limiting their practicality for large datasets. Amortized explainers, in turn, require only one inference per explanation, making them efficient for large black-box models and datasets. However, as demonstrated in Figure 1, amortized explainers can occasionally produce diverging explanations from the reference explainer used to train them.   We propose selective explanation, a method that bridges Monte Carlo and amortized explanations. By training a model that \u201clearns to select\u201d which method should be applied to each input, our selective explanation method can produce higher-quality explanations than amortized explainers at a significantly lower average computational cost than Monte Carlo-based approaches. The key idea behind the selective explanation method is to apply Monte Carlo explanations only to points that would receive low-quality explanations from the amortized explainer; see Figure 2 for the workflow of selective explanations.   The ideas of predicting selectively and providing recourse with a more accurate but expensive method have been explored in classification and regression [18, 19, 20, 21, 22]. To our knowledge, however, these ideas have not been applied to explanations. We make two contributions in this regard that are relevant for selective prediction more generally. (1) Selective prediction uses an uncertainty metric to identify input points for which the predictor (the amortized explainer in our case) would produce low-quality outputs and recourse is needed. The high-dimensional nature of explanations requires us to develop new uncertainty metrics (Section\u00a03) suitable for this setting. (2) Instead of providing recourse with a Monte Carlo explanation alone, as would be standard, we use an optimized method called explanations with initial guess (Section\u00a04) that combines amortized and Monte Carlo explanations, improving explanation quality beyond that of either explanation alone.   Our overall contribution (3) is to combine (1) and (2) in the form of selective explanations, providing explanations with initial guess to improve low-quality amortized explanations. We validate our selective explanations approach on two language models as well",
            "references": [
                {
                    "title": "Using Captum to Explain Generative Language Models",
                    "abstract": "Captum is a comprehensive library for model explainability in PyTorch, offering a range of methods from the interpretability literature to enhance users\u2019 understanding of PyTorch models. In this paper, we introduce new features in Captum that are specifically designed to analyze the behavior of generative language models. We provide an overview of the available functionalities and example applications of their potential for understanding learned associations within generative language models."
                },
                {
                    "title": "Efficient Shapley Values Estimation by Amortization for Text Classification",
                    "abstract": "Despite the popularity of Shapley Values in explaining neural text classification models, computing them is prohibitive for large pretrained models due to a large number of model evaluations. In practice, Shapley Values are often estimated with a small number of stochastic model evaluations. However, we show that the estimated Shapley Values are sensitive to random seed choices \u2013 the top-ranked features often have little overlap across different seeds, especially on examples with longer input texts. This can only be mitigated by aggregating thousands of model evaluations, which on the other hand, induces substantial computational overheads. To mitigate the trade-off between stability and efficiency, we develop an amortized model that directly predicts each input feature\u2019s Shapley Value without additional model evaluations. It is trained on a set of examples whose Shapley Values are estimated from a large number of model evaluations to ensure stability. Experimental results on two text classification datasets demonstrate that our amortized model estimates Shapley Values accurately with up to 60 times speedup compared to traditional methods. Further, our model does not suffer from stability issues as inference is deterministic. We release our code at https://github.com/yangalan123/Amortized-Interpretability."
                },
                {
                    "title": "CoRTX: Contrastive Framework for Real-time Explanation",
                    "abstract": "Recent advancements in explainable machine learning provide effective and faithful solutions for interpreting model behaviors. However, many explanation methods encounter efficiency issues, which largely limit their deployments in practical scenarios. Real-time explainer (RTX) frameworks have thus been proposed to accelerate the model explanation process by learning a one-feed-forward explainer. Existing RTX frameworks typically build the explainer under the supervised learning paradigm, which requires large amounts of explanation labels as the ground truth. Considering that accurate explanation labels are usually hard to obtain due to constrained computational resources and limited human efforts, effective explainer training is still challenging in practice. In this work, we propose a COntrastive Real-Time eXplanation (CoRTX) framework to learn the explanation-oriented representation and relieve the intensive dependence of explainer training on explanation labels. Specifically, we design a synthetic strategy to select positive and negative instances for the learning of explanation. Theoretical analysis show that our selection strategy can benefit the contrastive learning process on explanation tasks. Experimental results on three real-world datasets further demonstrate the efficiency and efficacy of our proposed CoRTX framework."
                },
                {
                    "title": "Selective Classification Via Neural Network Training Dynamics",
                    "abstract": "Selective classification is the task of rejecting inputs a model would predict incorrectly on through a trade-off between input space coverage and model accuracy. Current methods for selective classification impose constraints on either the model architecture or the loss function; this inhibits their usage in practice. In contrast to prior work, we show that state-of-the-art selective classification performance can be attained solely from studying the (discretized) training dynamics of a model. We propose a general framework that, for a given test input, monitors metrics capturing the disagreement with the final predicted label over intermediate models obtained during training; we then reject data points exhibiting too much disagreement at late stages in training. In particular, we instantiate a method that tracks when the label predicted during training stops disagreeing with the final predicted label. Our experimental evaluation shows that our method achieves state-of-the-art accuracy/coverage trade-offs on typical selective classification benchmarks."
                },
                {
                    "title": "ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detection",
                    "abstract": "Toxic language detection systems often falsely flag text that contains minority group mentions as toxic, as those groups are often the targets of online hate. Such over-reliance on spurious correlations also causes systems to struggle with detecting implicitly toxic language.To help mitigate these issues, we create ToxiGen, a new large-scale and machine-generated dataset of 274k toxic and benign statements about 13 minority groups. We develop a demonstration-based prompting framework and an adversarial classifier-in-the-loop decoding method to generate subtly toxic and benign text with a massive pretrained language model. Controlling machine generation in this way allows ToxiGen to cover implicitly toxic text at a larger scale, and about more demographic groups, than previous resources of human-written text. We conduct a human evaluation on a challenging subset of ToxiGen and find that annotators struggle to distinguish machine-generated text from human-written language. We also find that 94.5% of toxic examples are labeled as hate speech by human annotators. Using three publicly-available datasets, we show that finetuning a toxicity classifier on our data improves its performance on human-written data substantially. We also demonstrate that ToxiGen can be used to fight machine-generated toxicity as finetuning improves the classifier significantly on our evaluation subset."
                },
                {
                    "title": "Selective Regression Under Fairness Criteria",
                    "abstract": "Selective regression allows abstention from prediction if the con\ufb01dence to make an accurate prediction is not suf\ufb01cient. In general, by allowing a reject option, one expects the performance of a regression model to increase at the cost of reducing coverage (i.e., by predicting on fewer samples). However, as we show, in some cases, the performance of a minority subgroup can decrease while we reduce the coverage, and thus selective regression can magnify disparities between different sensitive subgroups. Motivated by these disparities, we propose new fairness criteria for selective regression requiring the performance of every subgroup to improve with a decrease in coverage. We prove that if a feature representation satis\ufb01es the suf\ufb01ciency criterion or is calibrated for mean and variance , then the proposed fairness criteria is met. Further, we introduce two approaches to mitigate the performance disparity across subgroups: (a) by regularizing an upper bound of conditional mutual information under a Gaussian assumption and (b) by regularizing a contrastive loss for conditional mean and conditional variance prediction. The effectiveness of these approaches is demonstrated on synthetic and real-world datasets."
                },
                {
                    "title": "FastSHAP: Real-Time Shapley Value Estimation",
                    "abstract": "Shapley values are widely used to explain black-box models, but they are costly to calculate because they require many model evaluations. We introduce FastSHAP, a method for estimating Shapley values in a single forward pass using a learned explainer model. FastSHAP amortizes the cost of explaining many inputs via a learning approach inspired by the Shapley value's weighted least squares characterization, and it can be trained using standard stochastic gradient optimization. We compare FastSHAP to existing estimation approaches, revealing that it generates high-quality explanations with orders of magnitude speedup."
                },
                {
                    "title": "Sampling Permutations for Shapley Value Estimation",
                    "abstract": "Game-theoretic attribution techniques based on Shapley values are used to interpret black-box machine learning models, but their exact calculation is generally NP-hard, requiring approximation methods for non-trivial models. As the computation of Shapley values can be expressed as a summation over a set of permutations, a common approach is to sample a subset of these permutations for approximation. Unfortunately, standard Monte Carlo sampling methods can exhibit slow convergence, and more sophisticated quasi-Monte Carlo methods have not yet been applied to the space of permutations. To address this, we investigate new approaches based on two classes of approximation methods and compare them empirically. First, we demonstrate quadrature techniques in a RKHS containing functions of permutations, using the Mallows kernel in combination with kernel herding and sequential Bayesian quadrature. The RKHS perspective also leads to quasi-Monte Carlo type error bounds, with a tractable discrepancy measure defined on permutations. Second, we exploit connections between the hypersphere $\\mathbb{S}^{d-2}$ and permutations to create practical algorithms for generating permutation samples with good properties. Experiments show the above techniques provide significant improvements for Shapley value estimates over existing methods, converging to a smaller RMSE in the same number of model evaluations."
                },
                {
                    "title": "Efficient Explanations from Empirical Explainers",
                    "abstract": "Amid a discussion about Green AI in which we see explainability neglected, we explore the possibility to efficiently approximate computationally expensive explainers. To this end, we propose feature attribution modelling with Empirical Explainers. Empirical Explainers learn from data to predict the attribution maps of expensive explainers. We train and test Empirical Explainers in the language domain and find that they model their expensive counterparts surprisingly well, at a fraction of the cost. They could thus mitigate the computational burden of neural explanations significantly, in applications that tolerate an approximation error."
                },
                {
                    "title": "Improving KernelSHAP: Practical Shapley Value Estimation via Linear Regression",
                    "abstract": "The Shapley value solution concept from cooperative game theory has become popular for interpreting ML models, but efficiently estimating Shapley values remains challenging, particularly in the model-agnostic setting. We revisit the idea of estimating Shapley values via linear regression to understand and improve upon this approach. By analyzing KernelSHAP alongside a newly proposed unbiased estimator, we develop techniques to detect its convergence and calculate uncertainty estimates. We also find that that the original version incurs a negligible increase in bias in exchange for a significant reduction in variance, and we propose a variance reduction technique that further accelerates the convergence of both estimators. Finally, we develop a version of KernelSHAP for stochastic cooperative games that yields fast new estimators for two global explanation methods."
                },
                {
                    "title": "Selective Classification via One-Sided Prediction",
                    "abstract": "We propose a novel method for selective classification (SC), a problem which allows a classifier to abstain from predicting some instances, thus trading off accuracy against coverage (the fraction of instances predicted). In contrast to prior gating or confidence-set based work, our proposed method optimises a collection of class-wise decoupled one-sided empirical risks, and is in essence a method for explicitly finding the largest decision sets for each class that have few false positives. This one-sided prediction (OSP) based relaxation yields an SC scheme that attains near-optimal coverage in the practically relevant high target accuracy regime, and further admits efficient implementation, leading to a flexible and principled method for SC. We theoretically derive generalization bounds for SC and OSP, and empirically we show that our scheme strongly outperforms state of the art methods in coverage at small error levels."
                },
                {
                    "title": "Captum: A unified and generic model interpretability library for PyTorch",
                    "abstract": "In this paper we introduce a novel, unified, open-source model interpretability library for PyTorch [12]. The library contains generic implementations of a number of gradient and perturbation-based attribution algorithms, also known as feature, neuron and layer importance algorithms, as well as a set of evaluation metrics for these algorithms. It can be used for both classification and non-classification models including graph-structured models built on Neural Networks (NN). In this paper we give a high-level overview of supported attribution algorithms and show how to perform memory-efficient and scalable computations. We emphasize that the three main characteristics of the library are multimodality, extensibility and ease of use. Multimodality supports different modality of inputs such as image, text, audio or video. Extensibility allows adding new algorithms and features. The library is also designed for easy understanding and use. Besides, we also introduce an interactive visualization tool called Captum Insights that is built on top of Captum library and allows sample-based model debugging and visualization using feature importance metrics."
                },
                {
                    "title": "Risk-Controlled Selective Prediction for Regression Deep Neural Network Models",
                    "abstract": "Regression deep neural network (DNN) models have been successfully utilized in numerous fields. In real-world applications, large regression errors on individual samples may result in severe consequences. Selective techniques, also known as reject options, have been used to reject predictions with high uncertainty. However, they have yet been mainly considered in classification neural networks (NNs), in comparison to the limited work in regression NNs. In this paper, we considered the selective regression problem from a risk-coverage point of view, and proposed a method to construct a selective regression model given a trained regression DNN model and a desired regression error risk. Then, we proposed to utilize blending variance to quantify uncertainty in regression NNs. We evaluated both the proposed uncertainty function and selective regression models for two real-world applications, the tropical cyclone (TC) intensity estimation problem and the apparent age estimation problem. Our proposed methods achieved promising results. For example, for the TC intensity estimation problem, our selective regression model guaranteed a risk bound (in terms of the root mean squared error (RMSE)) of 9.5 knots for 75% test coverage with a guided confidence level of 0.05, whereas the RMSE value achieved by the state-of-the-art model without selection was 10.5 knots."
                },
                {
                    "title": "Regression with reject option and application to kNN",
                    "abstract": "We investigate the problem of regression where one is allowed to abstain from predicting. We refer to this framework as regression with reject option as an extension of classification with reject option. In this context, we focus on the case where the rejection rate is fixed and derive the optimal rule which relies on thresholding the conditional variance function. We provide a semi-supervised estimation procedure of the optimal rule involving two datasets: a first labeled dataset is used to estimate both regression function and conditional variance function while a second unlabeled dataset is exploited to calibrate the desired rejection rate. The resulting predictor with reject option is shown to be almost as good as the optimal predictor with reject option both in terms of risk and rejection rate. We additionally apply our methodology with kNN algorithm and establish rates of convergence for the resulting kNN predictor under mild conditions. Finally, a numerical study is performed to illustrate the benefit of using the proposed procedure."
                },
                {
                    "title": "TextAttack: A Framework for Adversarial Attacks, Data Augmentation, and Adversarial Training in NLP",
                    "abstract": "While there has been substantial research using adversarial attacks to analyze NLP models, each attack is implemented in its own code repository. It remains challenging to develop NLP attacks and utilize them to improve model performance. This paper introduces TextAttack, a Python framework for adversarial attacks, data augmentation, and adversarial training in NLP. TextAttack builds attacks from four components: a goal function, a set of constraints, a transformation, and a search method. TextAttack\u2019s modular design enables researchers to easily construct attacks from combinations of novel and existing components. TextAttack provides implementations of 16 adversarial attacks from the literature and supports a variety of models and datasets, including BERT and other transformers, and all GLUE tasks. TextAttack also includes data augmentation and adversarial training modules for using components of adversarial attacks to improve model accuracy and robustness.TextAttack is democratizing NLP: anyone can try data augmentation and adversarial training on any model or dataset, with just a few lines of code. Code and tutorials are available at https://github.com/QData/TextAttack."
                },
                {
                    "title": "CXPlain: Causal Explanations for Model Interpretation under Uncertainty",
                    "abstract": "Feature importance estimates that inform users about the degree to which given inputs influence the output of a predictive model are crucial for understanding, validating, and interpreting machine-learning models. However, providing fast and accurate estimates of feature importance for high-dimensional data, and quantifying the uncertainty of such estimates remain open challenges. Here, we frame the task of providing explanations for the decisions of machine-learning models as a causal learning task, and train causal explanation (CXPlain) models that learn to estimate to what degree certain inputs cause outputs in another machine-learning model. CXPlain can, once trained, be used to explain the target model in little time, and enables the quantification of the uncertainty associated with its feature importance estimates via bootstrap ensembling. We present experiments that demonstrate that CXPlain is significantly more accurate and faster than existing model-agnostic methods for estimating feature importance. In addition, we confirm that the uncertainty estimates provided by CXPlain ensembles are strongly correlated with their ability to accurately estimate feature importance on held-out data."
                },
                {
                    "title": "LS-Tree: Model Interpretation When the Data Are Linguistic",
                    "abstract": "We study the problem of interpreting trained classification models in the setting of linguistic data sets. Leveraging a parse tree, we propose to assign least-squares-based importance scores to each word of an instance by exploiting syntactic constituency structure. We establish an axiomatic characterization of these importance scores by relating them to the Banzhaf value in coalitional game theory. Based on these importance scores, we develop a principled method for detecting and quantifying interactions between words in a sentence. We demonstrate that the proposed method can aid in interpretability and diagnostics for several widely-used language models."
                },
                {
                    "title": "SelectiveNet: A Deep Neural Network with an Integrated Reject Option",
                    "abstract": "We consider the problem of selective prediction (also known as reject option) in deep neural networks, and introduce SelectiveNet, a deep neural architecture with an integrated reject option. Existing rejection mechanisms are based mostly on a threshold over the prediction confidence of a pre-trained network. In contrast, SelectiveNet is trained to optimize both classification (or regression) and rejection simultaneously, end-to-end. The result is a deep neural network that is optimized over the covered domain. In our experiments, we show a consistently improved risk-coverage trade-off over several well-known classification and regression datasets, thus reaching new state-of-the-art results for deep selective classification."
                },
                {
                    "title": "Selective Classification for Deep Neural Networks",
                    "abstract": "Selective classification techniques (also known as reject option) have not yet been considered in the context of deep neural networks (DNNs). These techniques can potentially significantly improve DNNs prediction performance by trading-off coverage. In this paper we propose a method to construct a selective classifier given a trained neural network. Our method allows a user to set a desired risk level. At test time, the classifier rejects instances as needed, to grant the desired risk (with high probability). Empirical results over CIFAR and ImageNet convincingly demonstrate the viability of our method, which opens up possibilities to operate DNNs in mission-critical applications. For example, using our method an unprecedented 2% error in top-5 ImageNet classification can be guaranteed with probability 99.9%, and almost 60% test coverage."
                },
                {
                    "title": "Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles",
                    "abstract": "Deep neural networks (NNs) are powerful black box predictors that have recently achieved impressive performance on a wide spectrum of tasks. Quantifying predictive uncertainty in NNs is a challenging and yet unsolved problem. Bayesian NNs, which learn a distribution over weights, are currently the state-of-the-art for estimating predictive uncertainty; however these require significant modifications to the training procedure and are computationally expensive compared to standard (non-Bayesian) NNs. We propose an alternative to Bayesian NNs that is simple to implement, readily parallelizable, requires very little hyperparameter tuning, and yields high quality predictive uncertainty estimates. Through a series of experiments on classification and regression benchmarks, we demonstrate that our method produces well-calibrated uncertainty estimates which are as good or better than approximate Bayesian NNs. To assess robustness to dataset shift, we evaluate the predictive uncertainty on test examples from known and unknown distributions, and show that our method is able to express higher uncertainty on out-of-distribution examples. We demonstrate the scalability of our method by evaluating predictive uncertainty estimates on ImageNet."
                },
                {
                    "title": "Character-level Convolutional Networks for Text Classification",
                    "abstract": "This article offers an empirical exploration on the use of character-level convolutional networks (ConvNets) for text classification. We constructed several large-scale datasets to show that character-level convolutional networks could achieve state-of-the-art or competitive results. Comparisons are offered against traditional models such as bag of words, n-grams and their TFIDF variants, and deep learning models such as word-based ConvNets and recurrent neural networks."
                },
                {
                    "title": "On the Foundations of Noise-free Selective Classification",
                    "abstract": "We consider selective classification, a term we adopt here to refer to 'classification with a reject option.' The essence in selective classification is to trade-off classifier coverage for higher accuracy. We term this trade-off the risk-coverage (RC) trade-off. Our main objective is to characterize this trade-off and to construct algorithms that can optimally or near optimally achieve the best possible trade-offs in a controlled manner. For noise-free models we present in this paper a thorough analysis of selective classification including characterizations of RC trade-offs in various interesting settings."
                },
                {
                    "title": "Predictions",
                    "abstract": "2007 will be a year of intense \"hyperdisruption\" in the IT industry, with major structural changes taking place along different industry vectors at once \u0097 all interacting with each other and, more important, accelerating each other. In 2007, we'll see small business become big, more software become services, more services become software, business IT players become more \"consumer-ish,\" and consumer players become more businesslike. These disruptions, and others, will force most market leaders out of their comfort zones and open up new opportunities to those that choose to surf with these disruptions rather than stand against them. Here are some of the key developments we'll see in 2007:"
                },
                {
                    "title": "Seaborn: Statistical Data Visualization",
                    "abstract": "seaborn is a library for making statistical graphics in Python. It provides a high-level interface to matplotlib and integrates closely with pandas data structures. Functions in the seaborn library expose a declarative, dataset-oriented API that makes it easy to translate questions about data into graphics that can answer them. When given a dataset and a specification of the plot to make, seaborn automatically maps the data values to visual attributes such as color, size, or style, internally computes statistical transformations, and decorates the plot with informative axis labels and a legend. Many seaborn functions can generate figures with multiple panels that elicit comparisons between conditional subsets of data or across different pairings of variables in a dataset. seaborn is designed to be useful throughout the lifecycle of a scientific project. By producing complete graphics from a single function call with minimal arguments, seaborn facilitates rapid prototyping and exploratory data analysis. And by offering extensive options for customization, along with exposing the underlying matplotlib objects, it can be used to create polished, publication-quality figures."
                },
                {
                    "title": "Neural Networks: A Comprehensive Foundation",
                    "abstract": "Simon Haykin Neural Networks A Comprehensive Foundation. Neural Networks A Comprehensive Foundation Simon S. Neural Networks A Comprehensive Foundation Simon S. Neural Networks A Comprehensive Foundation. Neural Networks Association for Computing Machinery. Book Review Neural Networks A Comprehensive Foundation. Neural Networks A Comprehensive Foundation Pearson. Neural networks a comprehensive foundation. Neural Networks a Comprehensive Foundation AbeBooks. Neural networks a comprehensive foundation solutions. cdn preterhuman net. Neural Networks A Comprehensive Foundation Goodreads. Neural Networks A Comprehensive Foundation Amazon it. Neural Networks A Comprehensive Foundation Amazon co uk. Neural Networks A Comprehensive Foundation 3rd Edition. Neural Networks A Comprehensive Foundation Simon. Neural Networks A Comprehensive Foundation amazon com. Neural networks a comprehensive foundation Academia edu. Neural Networks A Comprehensive Foundation Amazon. neural networks a comprehensive foundation simon haykin. Simon Haykin Neural Networks A Comprehensive Foundation. Neural Networks A comprehensive Foundation 2 ed. Simon haykin neural networks a comprehensive foundation pdf. Buy Neural Networks A Comprehensive Foundation Book. Neural networks a comprehensive foundation 2e book. Neural Networks A Comprehensive Foundation. NEURAL NETWORKS A COMPREHENSIVE FOUNDATION SIMON. Neural Networks a Comprehensive Foundation by Haykin Simon. Neural Networks A Comprehensive Foundation pdf PDF Drive. Neural Networks A Comprehensive Foundation amazon ca. Simon Haykin Neural Networks A Comprehensive Foundation. NEURAL NETWORKS A Comprehensive Foundation PDF. Neural Networks A Comprehensive Foundation pdf PDF Drive. Neural Networks A Comprehensive Foundation by Haykin. Neural Networks A Comprehensive Foundation 3rd Edition. Neural Networks A Comprehensive Foundation Simon S. Neural Networks A Comprehensive Foundation. Neural networks a comprehensive foundation Book 1994. Neural Networks A Comprehensive Foundation 2nd Edition. Neural Networks A Comprehensive Foundation S S Haykin. Neural Networks A Comprehensive Foundation International. Neural Networks A Comprehensive Foundation 2 e Pearson. Download Neural Networks A Comprehensive Foundation 2Nd. Neural Networks A comprehensive foundation Aalto"
                },
                {
                    "title": "Note on shrinkage estimators for the binomial distribution",
                    "abstract": "A variety of shrinkage methods have been proposed for the estimation of some unknown parameter by considering estimators based on an a priori guess of the value of the parameter, or of an interval containing the parameter. We propose a simple new estimator and then compare a variety of shrinkage estimators for the parameter \u03c0 of the binomial distribution."
                },
                {
                    "title": "Computer Aided",
                    "abstract": "The production of inbetweens is a tedious task for animators and a complicated one for algorithms. In this paper, an algorithm for computer aided inbetweening and its integration in a pen-based graphical user interface are presented. The algorithm is layer-based, assuming an invariant layering order. It is applicable to animations in a style similar to paper cut out, in which the drawings on the cut-out pieces are inbetweened as well. The content of each key drawing is analysed and classified into strokes, chains of strokes and relations that hold among them. Rules decide what parts of different drawings may be matched. These rules specify allowed changes between relations in key drawings. A cost function based approach determines the correct matching of strokes. Generated animation paths between corresponding strokes determine the resulting inbetweens. To hold for possible mismatchings and to allow for artistic control over the results, the inbetweening algorithm is embedded in a free form graphic user interface. Thus artists are enabled to focus on the part of the inbetweening task computers are not able to solve. CR Categories: I.3.6 [Computer Graphics]: Methodology and Techniques\u2014Interaction techniques; I.4.8 [Image Processing and Computer Vision]: Scene Analysis\u2014Object Recognition; I.2.10 [Artificial Intelligence]: Vision and Scene Understanding\u2014Shape;"
                }
            ],
            "categories": [
                "cs.CY",
                "cs.CL",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we efficiently generate high-quality feature attributions for large black-box models in high-stakes applications while minimizing computational costs?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing demand for explainability in machine learning, particularly in high-stakes domains like healthcare and hiring. Improved feature attribution methods can enhance trust in AI systems, facilitate regulatory compliance, and promote ethical AI practices. By advancing the state of explainability, this research could lead to more robust and interpretable models, ultimately influencing future research directions in model transparency and accountability.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the computational expense associated with existing feature attribution methods, particularly for large models that require numerous inferences for each explanation. Naive approaches may fail due to their inability to balance the trade-off between explanation quality and computational efficiency. Additionally, the high-dimensional nature of explanations complicates the development of effective uncertainty metrics, which are essential for identifying when to apply more computationally intensive methods.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either Monte Carlo methods or amortized explainers, but not on integrating the strengths of both approaches. Limitations in existing solutions include the slow convergence of Monte Carlo methods and the occasional divergence of amortized explainers from reference outputs. Barriers such as the lack of suitable uncertainty metrics for high-dimensional explanations and the absence of a framework for selective application of explanation methods have hindered progress. Our approach differs by introducing selective explanations that leverage both methods, addressing these gaps effectively.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a selective explanation framework that utilizes an uncertainty metric to identify inputs needing high-quality explanations. We will employ a dataset of language model outputs and evaluate our method using metrics that assess explanation quality and computational efficiency. The expected outcomes include a significant reduction in average computational cost while maintaining or improving the quality of explanations compared to existing methods, thereby demonstrating the effectiveness of our selective explanation approach."
            }
        },
        "author_data": {
            "e02255f1-a789-47cc-ba99-807cdf02a5de": {
                "pk": "e02255f1-a789-47cc-ba99-807cdf02a5de",
                "name": "Lucas Monteiro Paes",
                "collaborators": [
                    "Flavio P. Calmon",
                    "Alexander Lin",
                    "Sree Harsha Tanneru",
                    "Suraj Srinivas",
                    "Himabindu Lakkaraju",
                    "Dennis Wei",
                    "Hyo Jin Do",
                    "Hendrik Strobelt",
                    "Ronny Luss",
                    "Amit Dhurandhar"
                ],
                "domain": [
                    "Fairness in Machine Learning",
                    "Generative Models",
                    "Content Moderation",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "Word-Level Explanations for Analyzing Bias in Text-to-Image Models",
                        "abstract": "Text-to-image models take a sentence (i.e., prompt) and generate images associated with this input prompt. These models have created award wining-art, videos, and even synthetic datasets. However, text-to-image (T2I) models can generate images that underrepresent minorities based on race and sex. This paper investigates which word in the input prompt is responsible for bias in generated images. We introduce a method for computing scores for each word in the prompt; these scores represent its influence on biases in the model's output. Our method follows the principle of \\emph{explaining by removing}, leveraging masked language models to calculate the influence scores. We perform experiments on Stable Diffusion to demonstrate that our method identifies the replication of societal stereotypes in generated images."
                    },
                    {
                        "title": "Multi-Level Explanations for Generative Language Models",
                        "abstract": "Perturbation-based explanation methods such as LIME and SHAP are commonly applied to text classification. This work focuses on their extension to generative language models. To address the challenges of text as output and long text inputs, we propose a general framework called MExGen that can be instantiated with different attribution algorithms. To handle text output, we introduce the notion of scalarizers for mapping text to real numbers and investigate multiple possibilities. To handle long inputs, we take a multi-level approach, proceeding from coarser levels of granularity to finer ones, and focus on algorithms with linear scaling in model queries. We conduct a systematic evaluation, both automated and human, of perturbation-based attribution methods for summarization and context-grounded question answering. The results show that our framework can provide more locally faithful explanations of generated outputs."
                    },
                    {
                        "title": "Multi-Group Fairness Evaluation via Conditional Value-at-Risk Testing",
                        "abstract": "Machine learning (ML) models used in prediction and classification tasks may display performance disparities across population groups determined by sensitive attributes (e.g., race, sex, age). We consider the problem of evaluating the performance of a fixed ML model across population groups defined by multiple sensitive attributes (e.g., race and sex and age). Here, the sample complexity for estimating the worst-case performance gap across groups (e.g., the largest difference in error rates) increases exponentially with the number of group-denoting sensitive attributes. To address this issue, we propose an approach to test for performance disparities based on Conditional Value-at-Risk (CVaR). By allowing a small probabilistic slack on the groups over which a model has approximately equal performance, we show that the sample complexity required for discovering performance violations is reduced exponentially to be at most upper bounded by the square root of the number of groups. As a byproduct of our analysis, when the groups are weighted by a specific prior distribution, we show that R\\'enyi entropy of order 2/3 of the prior distribution captures the sample complexity of the proposed CVaR test algorithm. Finally, we also show that there exists a non-i.i.d. data collection strategy that results in a sample complexity independent of the number of groups."
                    },
                    {
                        "title": "Algorithmic Arbitrariness in Content Moderation",
                        "abstract": "Machine learning (ML) is widely used to moderate online content. Despite its scalability relative to human moderation, the use of ML introduces unique challenges to content moderation. One such challenge is predictive multiplicity: multiple competing models for content classification may perform equally well on average, yet assign conflicting predictions to the same content. This multiplicity can result from seemingly innocuous choices during model development, such as random seed selection for parameter initialization. We experimentally demonstrate how content moderation tools can arbitrarily classify samples as toxic, leading to arbitrary restrictions on speech. We discuss these findings in terms of human rights set out by the International Covenant on Civil and Political Rights (ICCPR), namely freedom of expression, non-discrimination, and procedural justice. We analyze (i) the extent of predictive multiplicity among state-of-the-art LLMs used for detecting toxic content; (ii) the disparate impact of this arbitrariness across social groups; and (iii) how model multiplicity compares to unambiguous human classifications. Our findings indicate that the up-scaled algorithmic moderation risks legitimizing an algorithmic leviathan, where an algorithm disproportionately manages human rights. To mitigate such risks, our study underscores the need to identify and increase the transparency of arbitrariness in content moderation applications. Since algorithmic content moderation is being fueled by pressing social concerns, such as disinformation and hate speech, our discussion on harms raises concerns relevant to policy debates. Our findings also contribute to content moderation and intermediary liability laws being discussed and passed in many countries, such as the Digital Services Act in the European Union, the Online Safety Act in the United Kingdom, and the Fake News Bill in Brazil."
                    },
                    {
                        "title": "Multi-Group Proportional Representation",
                        "abstract": "Image search and retrieval tasks can perpetuate harmful stereotypes, erase cultural identities, and amplify social disparities. Current approaches to mitigate these representational harms balance the number of retrieved items across population groups defined by a small number of (often binary) attributes. However, most existing methods overlook intersectional groups determined by combinations of group attributes, such as gender, race, and ethnicity. We introduce Multi-Group Proportional Representation (MPR), a novel metric that measures representation across intersectional groups. We develop practical methods for estimating MPR, provide theoretical guarantees, and propose optimization algorithms to ensure MPR in retrieval. We demonstrate that existing methods optimizing for equal and proportional representation metrics may fail to promote MPR. Crucially, our work shows that optimizing MPR yields more proportional representation across multiple intersectional groups specified by a rich function class, often with minimal compromise in retrieval accuracy."
                    }
                ]
            },
            "cac2b944-f55b-4e9f-b860-fac39f96042a": {
                "pk": "cac2b944-f55b-4e9f-b860-fac39f96042a",
                "name": "Dennis Wei",
                "collaborators": [
                    "Sanjeeb Dash",
                    "Tian Gao",
                    "Oktay G\u00fcnl\u00fck",
                    "Alfred O. Hero III",
                    "Pin-Yu Chen",
                    "Yue Yu",
                    "Dmitry M. Malioutov",
                    "Kush R. Varshney",
                    "Michael Hind",
                    "Yunfeng Zhang"
                ],
                "domain": [
                    "Optimization",
                    "Machine Learning",
                    "Interpretability",
                    "Graph Theory"
                ],
                "publications": [
                    {
                        "title": "A Constant-Factor Bi-Criteria Approximation Guarantee for $k$-means++",
                        "abstract": "This paper studies the $k$-means++ algorithm for clustering as well as the class of $D^\\ell$ sampling algorithms to which $k$-means++ belongs. It is shown that for any constant factor $\\beta > 1$, selecting $\\beta k$ cluster centers by $D^\\ell$ sampling yields a constant-factor approximation to the optimal clustering with $k$ centers, in expectation and without conditions on the dataset. This result extends the previously known $O(\\log k)$ guarantee for the case $\\beta = 1$ to the constant-factor bi-criteria regime. It also improves upon an existing constant-factor bi-criteria result that holds only with constant probability."
                    },
                    {
                        "title": "On two relaxations of quadratically-constrained cardinality minimization",
                        "abstract": "This paper considers a quadratically-constrained cardinality minimization problem with applications to digital filter design, subset selection for linear regression, and portfolio selection. Two relaxations are investigated: the continuous relaxation of a mixed integer formulation, and an optimized diagonal relaxation that exploits a simple special case of the problem. For the continuous relaxation, an absolute upper bound on the optimal cost is derived, suggesting that the continuous relaxation tends to be a relatively poor approximation. In computational experiments, diagonal relaxations often provide stronger bounds than continuous relaxations and can greatly reduce the complexity of a branch-and-bound solution, even in instances that are not particularly close to diagonal. Similar gains are observed with respect to the mixed integer programming solver CPLEX. Motivated by these results, the approximation properties of the diagonal relaxation are analyzed. In particular, bounds on the approximation ratio are established in terms of the eigenvalues of the matrix defining the quadratic constraint, and also in the diagonally dominant and nearly coordinate-aligned cases."
                    },
                    {
                        "title": "Adaptive Sensing Resource Allocation Over Multiple Hypothesis Tests",
                        "abstract": "This paper considers multiple binary hypothesis tests with adaptive allocation of sensing resources from a shared budget over a small number of stages. A Bayesian formulation is provided for the multistage allocation problem of minimizing the sum of Bayes risks, which is then recast as a dynamic program. In the single-stage case, the problem is a non-convex optimization, for which an algorithm composed of a series of parallel one-dimensional minimizations is presented. This algorithm ensures a global minimum under a sufficient condition. In the multistage case, the approximate dynamic programming method of open-loop feedback control is employed. In numerical simulations, the proposed allocation policies outperform alternative adaptive procedures when the numbers of true null and alternative hypotheses are not too imbalanced. In the case of few alternative hypotheses, the proposed policies are competitive using only a few stages of adaptation. In all cases substantial gains over non-adaptive sensing are observed."
                    },
                    {
                        "title": "Optimal Policies for the Homogeneous Selective Labels Problem",
                        "abstract": "Selective labels are a common feature of consequential decision-making applications, referring to the lack of observed outcomes under one of the possible decisions. This paper reports work in progress on learning decision policies in the face of selective labels. The setting considered is both a simplified homogeneous one, disregarding individuals' features to facilitate determination of optimal policies, and an online one, to balance costs incurred in learning with future utility. For maximizing discounted total reward, the optimal policy is shown to be a threshold policy, and the problem is one of optimal stopping. In contrast, for undiscounted infinite-horizon average reward, optimal policies have positive acceptance probability in all states. Future work stemming from these results is discussed."
                    },
                    {
                        "title": "On the Supermodularity of Active Graph-based Semi-supervised Learning with Stieltjes Matrix Regularization",
                        "abstract": "Active graph-based semi-supervised learning (AG-SSL) aims to select a small set of labeled examples and utilize their graph-based relation to other unlabeled examples to aid in machine learning tasks. It is also closely related to the sampling theory in graph signal processing. In this paper, we revisit the original formulation of graph-based SSL and prove the supermodularity of an AG-SSL objective function under a broad class of regularization functions parameterized by Stieltjes matrices. Under this setting, supermodularity yields a novel greedy label sampling algorithm with guaranteed performance relative to the optimal sampling set. Compared to three state-of-the-art graph signal sampling and recovery methods on two real-life community detection datasets, the proposed AG-SSL method attains superior classification accuracy given limited sample budgets."
                    },
                    {
                        "title": "Generalized Linear Rule Models",
                        "abstract": "This paper considers generalized linear models using rule-based features, also referred to as rule ensembles, for regression and probabilistic classification. Rules facilitate model interpretation while also capturing nonlinear dependences and interactions. Our problem formulation accordingly trades off rule set complexity and prediction accuracy. Column generation is used to optimize over an exponentially large space of rules without pre-generating a large subset of candidates or greedily boosting rules one by one. The column generation subproblem is solved using either integer programming or a heuristic optimizing the same objective. In experiments involving logistic and linear regression, the proposed methods obtain better accuracy-complexity trade-offs than existing rule ensemble algorithms. At one end of the trade-off, the methods are competitive with less interpretable benchmark models."
                    },
                    {
                        "title": "DAGs with No Fears: A Closer Look at Continuous Optimization for Learning Bayesian Networks",
                        "abstract": "This paper re-examines a continuous optimization framework dubbed NOTEARS for learning Bayesian networks. We first generalize existing algebraic characterizations of acyclicity to a class of matrix polynomials. Next, focusing on a one-parameter-per-edge setting, it is shown that the Karush-Kuhn-Tucker (KKT) optimality conditions for the NOTEARS formulation cannot be satisfied except in a trivial case, which explains a behavior of the associated algorithm. We then derive the KKT conditions for an equivalent reformulation, show that they are indeed necessary, and relate them to explicit constraints that certain edges be absent from the graph. If the score function is convex, these KKT conditions are also sufficient for local minimality despite the non-convexity of the constraint. Informed by the KKT conditions, a local search post-processing algorithm is proposed and shown to substantially and universally improve the structural Hamming distance of all tested algorithms, typically by a factor of 2 or more. Some combinations with local search are both more accurate and more efficient than the original NOTEARS."
                    },
                    {
                        "title": "A Statistical Interpretation of the Maximum Subarray Problem",
                        "abstract": "Maximum subarray is a classical problem in computer science that given an array of numbers aims to find a contiguous subarray with the largest sum. We focus on its use for a noisy statistical problem of localizing an interval with a mean different from background. While a naive application of maximum subarray fails at this task, both a penalized and a constrained version can succeed. We show that the penalized version can be derived for common exponential family distributions, in a manner similar to the change-point detection literature, and we interpret the resulting optimal penalty value. The failure of the naive formulation is then explained by an analysis of the estimated interval boundaries. Experiments further quantify the effect of deviating from the optimal penalty. We also relate the penalized and constrained formulations and show that the solutions to the former lie on the convex hull of the solutions to the latter."
                    },
                    {
                        "title": "Robust Binary Hypothesis Testing Under Contaminated Likelihoods",
                        "abstract": "In hypothesis testing, the phenomenon of label noise, in which hypothesis labels are switched at random, contaminates the likelihood functions. In this paper, we develop a new method to determine the decision rule when we do not have knowledge of the uncontaminated likelihoods and contamination probabilities, but only have knowledge of the contaminated likelihoods. In particular we pose a minimax optimization problem that finds a decision rule robust against this lack of knowledge. The method simplifies by application of linear programming theory. Motivation for this investigation is provided by problems encountered in workforce analytics."
                    },
                    {
                        "title": "Consumer-Driven Explanations for Machine Learning Decisions: An Empirical Study of Robustness",
                        "abstract": "Many proposed methods for explaining machine learning predictions are in fact challenging to understand for nontechnical consumers. This paper builds upon an alternative consumer-driven approach called TED that asks for explanations to be provided in training data, along with target labels. Using semi-synthetic data from credit approval and employee retention applications, experiments are conducted to investigate some practical considerations with TED, including its performance with different classification algorithms, varying numbers of explanations, and variability in explanations. A new algorithm is proposed to handle the case where some training examples do not have explanations. Our results show that TED is robust to increasing numbers of explanations, noisy explanations, and large fractions of missing explanations, thus making advances toward its practical deployment."
                    },
                    {
                        "title": "Boolean Decision Rules via Column Generation",
                        "abstract": "This paper considers the learning of Boolean rules in either disjunctive normal form (DNF, OR-of-ANDs, equivalent to decision rule sets) or conjunctive normal form (CNF, AND-of-ORs) as an interpretable model for classification. An integer program is formulated to optimally trade classification accuracy for rule simplicity. Column generation (CG) is used to efficiently search over an exponential number of candidate clauses (conjunctions or disjunctions) without the need for heuristic rule mining. This approach also bounds the gap between the selected rule set and the best possible rule set on the training data. To handle large datasets, we propose an approximate CG algorithm using randomization. Compared to three recently proposed alternatives, the CG algorithm dominates the accuracy-simplicity trade-off in 7 out of 15 datasets. When maximized for accuracy, CG is competitive with rule learners designed for this purpose, sometimes finding significantly simpler solutions that are no less accurate."
                    },
                    {
                        "title": "Interpretable and Fair Boolean Rule Sets via Column Generation",
                        "abstract": "This paper considers the learning of Boolean rules in disjunctive normal form (DNF, OR-of-ANDs, equivalent to decision rule sets) as an interpretable model for classification. An integer program is formulated to optimally trade classification accuracy for rule simplicity. We also consider the fairness setting and extend the formulation to include explicit constraints on two different measures of classification parity: equality of opportunity and equalized odds. Column generation (CG) is used to efficiently search over an exponential number of candidate rules without the need for heuristic rule mining. To handle large data sets, we propose an approximate CG algorithm using randomization. Compared to three recently proposed alternatives, the CG algorithm dominates the accuracy-simplicity trade-off in 8 out of 16 data sets. When maximized for accuracy, CG is competitive with rule learners designed for this purpose, sometimes finding significantly simpler solutions that are no less accurate. Compared to other fair and interpretable classifiers, our method is able to find rule sets that meet stricter notions of fairness with a modest trade-off in accuracy."
                    },
                    {
                        "title": "Multistage Adaptive Estimation of Sparse Signals",
                        "abstract": "This paper considers sequential adaptive estimation of sparse signals under a constraint on the total sensing effort. The advantage of adaptivity in this context is the ability to focus more resources on regions of space where signal components exist, thereby improving performance. A dynamic programming formulation is derived for the allocation of sensing effort to minimize the expected estimation loss. Based on the method of open-loop feedback control, allocation policies are then developed for a variety of loss functions. The policies are optimal in the two-stage case, generalizing an optimal two-stage policy proposed by Bashan et al., and improve monotonically thereafter with the number of stages. Numerical simulations show gains up to several dB as compared to recently proposed adaptive methods, and dramatic gains compared to non-adaptive estimation. An application to radar imaging is also presented."
                    },
                    {
                        "title": "Performance Guarantees for Adaptive Estimation of Sparse Signals",
                        "abstract": "This paper studies adaptive sensing for estimating the nonzero amplitudes of a sparse signal with the aim of providing analytical guarantees on the performance gain due to adaptive resource allocation. We consider a previously proposed optimal two-stage policy for allocating sensing resources. For positive powers q, we derive tight upper bounds on the mean qth-power error resulting from the optimal two-stage policy and corresponding lower bounds on the improvement over non-adaptive uniform sensing. It is shown that the adaptation gain is related to the detectability of nonzero signal components as characterized by Chernoff coefficients, thus quantifying analytically the dependence on the sparsity level of the signal, the signal-to-noise ratio, and the sensing resource budget. For fixed sparsity levels and increasing signal-to-noise ratio or sensing budget, we obtain the rate of convergence to oracle performance and the rate at which the fraction of resources spent on the first exploratory stage decreases to zero. For a vanishing fraction of nonzero components, the gain increases without bound as a function of signal-to-noise ratio and sensing budget. Numerical simulations demonstrate that the bounds on adaptation gain are quite tight in non-asymptotic regimes as well."
                    },
                    {
                        "title": "Identifying Sub-networks in Neural Networks via Functionally Similar Representations",
                        "abstract": "Mechanistic interpretability aims to provide human-understandable insights into the inner workings of neural network models by examining their internals. Existing approaches typically require significant manual effort and prior knowledge, with strategies tailored to specific tasks. In this work, we take a step toward automating the understanding of the network by investigating the existence of distinct sub-networks. Specifically, we explore a novel automated and task-agnostic approach based on the notion of functionally similar representations within neural networks, reducing the need for human intervention. Our method identifies similar and dissimilar layers in the network, revealing potential sub-components. We achieve this by proposing, for the first time to our knowledge, the use of Gromov-Wasserstein distance, which overcomes challenges posed by varying distributions and dimensionalities across intermediate representations, issues that complicate direct layer-to-layer comparisons. Through experiments on algebraic and language tasks, we observe the emergence of sub-groups within neural network layers corresponding to functional abstractions. Additionally, we find that different training strategies influence the positioning of these sub-groups. Our approach offers meaningful insights into the behavior of neural networks with minimal human and computational cost."
                    }
                ]
            },
            "a9bb4c3f-4b7f-4a5a-b400-964b7aac4593": {
                "pk": "a9bb4c3f-4b7f-4a5a-b400-964b7aac4593",
                "name": "Flavio P. Calmon",
                "collaborators": [
                    "Hao Wang",
                    "Mario Diaz",
                    "Lalitha Sankar",
                    "Shahab Asoodeh",
                    "Hsiang Hsu",
                    "Salman Salamatian",
                    "Wael Alghamdi",
                    "Berk Ustun",
                    "Rui Gao",
                    "Jiachun Liao"
                ],
                "domain": [
                    "Differential Privacy",
                    "Algorithmic Fairness",
                    "Information Theory",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Privacy Amplification of Iterative Algorithms via Contraction Coefficients",
                        "abstract": "We investigate the framework of privacy amplification by iteration, recently proposed by Feldman et al., from an information-theoretic lens. We demonstrate that differential privacy guarantees of iterative mappings can be determined by a direct application of contraction coefficients derived from strong data processing inequalities for $f$-divergences. In particular, by generalizing the Dobrushin's contraction coefficient for total variation distance to an $f$-divergence known as $E_{\\gamma}$-divergence, we derive tighter bounds on the differential privacy parameters of the projected noisy stochastic gradient descent algorithm with hidden intermediate updates."
                    },
                    {
                        "title": "Local Differential Privacy Is Equivalent to Contraction of $E_\u03b3$-Divergence",
                        "abstract": "We investigate the local differential privacy (LDP) guarantees of a randomized privacy mechanism via its contraction properties. We first show that LDP constraints can be equivalently cast in terms of the contraction coefficient of the $E_\\gamma$-divergence. We then use this equivalent formula to express LDP guarantees of privacy mechanisms in terms of contraction coefficients of arbitrary $f$-divergences. When combined with standard estimation-theoretic tools (such as Le Cam's and Fano's converse methods), this result allows us to study the trade-off between privacy and utility in several testing and minimax and Bayesian estimation problems."
                    },
                    {
                        "title": "Strong Data Processing Inequalities for Input Constrained Additive Noise Channels",
                        "abstract": "This paper quantifies the intuitive observation that adding noise reduces available information by means of non-linear strong data processing inequalities. Consider the random variables $W\\to X\\to Y$ forming a Markov chain, where $Y=X+Z$ with $X$ and $Z$ real-valued, independent and $X$ bounded in $L_p$-norm. It is shown that $I(W;Y) \\le F_I(I(W;X))$ with $F_I(t)<t$ whenever $t>0$, if and only if $Z$ has a density whose support is not disjoint from any translate of itself. A related question is to characterize for what couplings $(W,X)$ the mutual information $I(W;Y)$ is close to maximum possible. To that end we show that in order to saturate the channel, i.e. for $I(W;Y)$ to approach capacity, it is mandatory that $I(W;X)\\to\\infty$ (under suitable conditions on the channel). A key ingredient for this result is a deconvolution lemma which shows that post-convolution total variation distance bounds the pre-convolution Kolmogorov-Smirnov distance. Explicit bounds are provided for the special case of the additive Gaussian noise channel with quadratic cost constraint. These bounds are shown to be order-optimal. For this case simplified proofs are provided leveraging Gaussian-specific tools such as the connection between information and estimation (I-MMSE) and Talagrand's information-transportation inequality."
                    },
                    {
                        "title": "Correspondence Analysis Using Neural Networks",
                        "abstract": "Correspondence analysis (CA) is a multivariate statistical tool used to visualize and interpret data dependencies. CA has found applications in fields ranging from epidemiology to social sciences. However, current methods used to perform CA do not scale to large, high-dimensional datasets. By re-interpreting the objective in CA using an information-theoretic tool called the principal inertia components, we demonstrate that performing CA is equivalent to solving a functional optimization problem over the space of finite variance functions of two random variable. We show that this optimization problem, in turn, can be efficiently approximated by neural networks. The resulting formulation, called the correspondence analysis neural network (CA-NN), enables CA to be performed at an unprecedented scale. We validate the CA-NN on synthetic data, and demonstrate how it can be used to perform CA on a variety of datasets, including food recipes, wine compositions, and images. Our results outperform traditional methods used in CA, indicating that CA-NN can serve as a new, scalable tool for interpretability and visualization of complex dependencies between random variables."
                    },
                    {
                        "title": "Measuring Information from Moments",
                        "abstract": "We investigate the problem of representing information measures in terms of the moments of the underlying random variables. First, we derive polynomial approximations of the conditional expectation operator. We then apply these approximations to bound the best mean-square error achieved by a polynomial estimator -- referred to here as the PMMSE. In Gaussian channels, the PMMSE coincides with the minimum mean-square error (MMSE) if and only if the input is either Gaussian or constant, i.e., if and only if the conditional expectation of the input of the channel given the output is a polynomial of degree at most 1. By combining the PMMSE with the I-MMSE relationship, we derive new formulas for information measures (e.g., differential entropy, mutual information) that are given in terms of the moments of the underlying random variables. As an application, we introduce estimators for information measures from data via approximating the moments in our formulas by sample moments. These estimators are shown to be asymptotically consistent and possess desirable properties, e.g., invariance to affine transformations when used to estimate mutual information."
                    },
                    {
                        "title": "On the Direction of Discrimination: An Information-Theoretic Analysis of Disparate Impact in Machine Learning",
                        "abstract": "In the context of machine learning, disparate impact refers to a form of systematic discrimination whereby the output distribution of a model depends on the value of a sensitive attribute (e.g., race or gender). In this paper, we propose an information-theoretic framework to analyze the disparate impact of a binary classification model. We view the model as a fixed channel, and quantify disparate impact as the divergence in output distributions over two groups. Our aim is to find a correction function that can perturb the input distributions of each group to align their output distributions. We present an optimization problem that can be solved to obtain a correction function that will make the output distributions statistically indistinguishable. We derive closed-form expressions to efficiently compute the correction function, and demonstrate the benefits of our framework on a recidivism prediction problem based on the ProPublica COMPAS dataset."
                    },
                    {
                        "title": "Generalization Bounds for Noisy Iterative Algorithms Using Properties of Additive Noise Channels",
                        "abstract": "Machine learning models trained by different optimization algorithms under different data distributions can exhibit distinct generalization behaviors. In this paper, we analyze the generalization of models trained by noisy iterative algorithms. We derive distribution-dependent generalization bounds by connecting noisy iterative algorithms to additive noise channels found in communication and information theory. Our generalization bounds shed light on several applications, including differentially private stochastic gradient descent (DP-SGD), federated learning, and stochastic gradient Langevin dynamics (SGLD). We demonstrate our bounds through numerical experiments, showing that they can help understand recent empirical observations of the generalization phenomena of neural networks."
                    },
                    {
                        "title": "Polynomial Approximations of Conditional Expectations in Scalar Gaussian Channels",
                        "abstract": "We consider a channel $Y=X+N$ where $X$ is a random variable satisfying $\\mathbb{E}[|X|]<\\infty$ and $N$ is an independent standard normal random variable. We show that the minimum mean-square error estimator of $X$ from $Y,$ which is given by the conditional expectation $\\mathbb{E}[X \\mid Y],$ is a polynomial in $Y$ if and only if it is linear or constant; these two cases correspond to $X$ being Gaussian or a constant, respectively. We also prove that the higher-order derivatives of $y \\mapsto \\mathbb{E}[X \\mid Y=y]$ are expressible as multivariate polynomials in the functions $y \\mapsto \\mathbb{E}\\left[ \\left( X - \\mathbb{E}[X \\mid Y] \\right)^k \\mid Y = y \\right]$ for $k\\in \\mathbb{N}.$ These expressions yield bounds on the $2$-norm of the derivatives of the conditional expectation. These bounds imply that, if $X$ has a compactly-supported density that is even and decreasing on the positive half-line, then the error in approximating the conditional expectation $\\mathbb{E}[X \\mid Y]$ by polynomials in $Y$ of degree at most $n$ decays faster than any polynomial in $n.$"
                    },
                    {
                        "title": "Repairing without Retraining: Avoiding Disparate Impact with Counterfactual Distributions",
                        "abstract": "When the performance of a machine learning model varies over groups defined by sensitive attributes (e.g., gender or ethnicity), the performance disparity can be expressed in terms of the probability distributions of the input and output variables over each group. In this paper, we exploit this fact to reduce the disparate impact of a fixed classification model over a population of interest. Given a black-box classifier, we aim to eliminate the performance gap by perturbing the distribution of input variables for the disadvantaged group. We refer to the perturbed distribution as a counterfactual distribution, and characterize its properties for common fairness criteria. We introduce a descent algorithm to learn a counterfactual distribution from data. We then discuss how the estimated distribution can be used to build a data preprocessor that can reduce disparate impact without training a new model. We validate our approach through experiments on real-world datasets, showing that it can repair different forms of disparity without a significant drop in accuracy."
                    },
                    {
                        "title": "Fairness without Imputation: A Decision Tree Approach for Fair Prediction with Missing Values",
                        "abstract": "We investigate the fairness concerns of training a machine learning model using data with missing values. Even though there are a number of fairness intervention methods in the literature, most of them require a complete training set as input. In practice, data can have missing values, and data missing patterns can depend on group attributes (e.g. gender or race). Simply applying off-the-shelf fair learning algorithms to an imputed dataset may lead to an unfair model. In this paper, we first theoretically analyze different sources of discrimination risks when training with an imputed dataset. Then, we propose an integrated approach based on decision trees that does not require a separate process of imputation and learning. Instead, we train a tree with missing incorporated as attribute (MIA), which does not require explicit imputation, and we optimize a fairness-regularized objective function. We demonstrate that our approach outperforms existing fairness intervention methods applied to an imputed dataset, through several experiments on real-world datasets."
                    },
                    {
                        "title": "Adapting Fairness Interventions to Missing Values",
                        "abstract": "Missing values in real-world data pose a significant and unique challenge to algorithmic fairness. Different demographic groups may be unequally affected by missing data, and the standard procedure for handling missing values where first data is imputed, then the imputed data is used for classification -- a procedure referred to as \"impute-then-classify\" -- can exacerbate discrimination. In this paper, we analyze how missing values affect algorithmic fairness. We first prove that training a classifier from imputed data can significantly worsen the achievable values of group fairness and average accuracy. This is because imputing data results in the loss of the missing pattern of the data, which often conveys information about the predictive label. We present scalable and adaptive algorithms for fair classification with missing values. These algorithms can be combined with any preexisting fairness-intervention algorithm to handle all possible missing patterns while preserving information encoded within the missing patterns. Numerical experiments with state-of-the-art fairness interventions demonstrate that our adaptive algorithms consistently achieve higher fairness and accuracy than impute-then-classify across different datasets."
                    },
                    {
                        "title": "Generalizing Correspondence Analysis for Applications in Machine Learning",
                        "abstract": "Correspondence analysis (CA) is a multivariate statistical tool used to visualize and interpret data dependencies by finding maximally correlated embeddings of pairs of random variables. CA has found applications in fields ranging from epidemiology to social sciences; however, current methods do not scale to large, high-dimensional datasets. In this paper, we provide a novel interpretation of CA in terms of an information-theoretic quantity called the principal inertia components. We show that estimating the principal inertia components, which consists in solving a functional optimization problem over the space of finite variance functions of two random variable, is equivalent to performing CA. We then leverage this insight to design novel algorithms to perform CA at an unprecedented scale. Particularly, we demonstrate how the principal inertia components can be reliably approximated from data using deep neural networks. Finally, we show how these maximally correlated embeddings of pairs of random variables in CA further play a central role in several learning problems including visualization of classification boundary and training process, and underlying recent multi-view and multi-modal learning methods."
                    },
                    {
                        "title": "Contraction of $E_\u03b3$-Divergence and Its Applications to Privacy",
                        "abstract": "We investigate the contraction coefficients derived from strong data processing inequalities for the $E_\\gamma$-divergence. By generalizing the celebrated Dobrushin's coefficient from total variation distance to $E_\\gamma$-divergence, we derive a closed-form expression for the contraction of $E_\\gamma$-divergence. This result has fundamental consequences in two privacy settings. First, it implies that local differential privacy can be equivalently expressed in terms of the contraction of $E_\\gamma$-divergence. This equivalent formula can be used to precisely quantify the impact of local privacy in (Bayesian and minimax) estimation and hypothesis testing problems in terms of the reduction of effective sample size. Second, it leads to a new information-theoretic technique for analyzing privacy guarantees of online algorithms. In this technique, we view such algorithms as a composition of amplitude-constrained Gaussian channels and then relate their contraction coefficients under $E_\\gamma$-divergence to the overall differential privacy guarantees. As an example, we apply our technique to derive the differential privacy parameters of gradient descent. Moreover, we also show that this framework can be tailored to batch learning algorithms that can be implemented with one pass over the training dataset."
                    },
                    {
                        "title": "Optimized Data Pre-Processing for Discrimination Prevention",
                        "abstract": "Non-discrimination is a recognized objective in algorithmic decision making. In this paper, we introduce a novel probabilistic formulation of data pre-processing for reducing discrimination. We propose a convex optimization for learning a data transformation with three goals: controlling discrimination, limiting distortion in individual data samples, and preserving utility. We characterize the impact of limited sample size in accomplishing this objective, and apply two instances of the proposed optimization to datasets, including one on real-world criminal recidivism. The results demonstrate that all three criteria can be simultaneously achieved and also reveal interesting patterns of bias in American society."
                    },
                    {
                        "title": "The Utility Cost of Robust Privacy Guarantees",
                        "abstract": "Consider a data publishing setting for a data set with public and private features. The objective of the publisher is to maximize the amount of information about the public features in a revealed data set, while keeping the information leaked about the private features bounded. The goal of this paper is to analyze the performance of privacy mechanisms that are constructed to match the distribution learned from the data set. Two distinct scenarios are considered: (i) mechanisms are designed to provide a privacy guarantee for the learned distribution; and (ii) mechanisms are designed to provide a privacy guarantee for every distribution in a given neighborhood of the learned distribution. For the first scenario, given any privacy mechanism, upper bounds on the difference between the privacy-utility guarantees for the learned and true distributions are presented. In the second scenario, upper bounds on the reduction in utility incurred by providing a uniform privacy guarantee are developed."
                    },
                    {
                        "title": "Privacy Under Hard Distortion Constraints",
                        "abstract": "We study the problem of data disclosure with privacy guarantees, wherein the utility of the disclosed data is ensured via a \\emph{hard distortion} constraint. Unlike average distortion, hard distortion provides a deterministic guarantee of fidelity. For the privacy measure, we use a tunable information leakage measure, namely \\textit{maximal $\\alpha$-leakage} ($\\alpha\\in[1,\\infty]$), and formulate the privacy-utility tradeoff problem. The resulting solution highlights that under a hard distortion constraint, the nature of the solution remains unchanged for both local and non-local privacy requirements. More precisely, we show that both the optimal mechanism and the optimal tradeoff are invariant for any $\\alpha>1$; i.e., the tunable leakage measure only behaves as either of the two extrema, i.e., mutual information for $\\alpha=1$ and maximal leakage for $\\alpha=\\infty$."
                    },
                    {
                        "title": "On the Robustness of Information-Theoretic Privacy Measures and Mechanisms",
                        "abstract": "Consider a data publishing setting for a dataset composed by both private and non-private features. The publisher uses an empirical distribution, estimated from $n$ i.i.d. samples, to design a privacy mechanism which is applied to new fresh samples afterward. In this paper, we study the discrepancy between the privacy-utility guarantees for the empirical distribution, used to design the privacy mechanism, and those for the true distribution, experienced by the privacy mechanism in practice. We first show that, for any privacy mechanism, these discrepancies vanish at speed $O(1/\\sqrt{n})$ with high probability. These bounds follow from our main technical results regarding the Lipschitz continuity of the considered information leakage measures. Then we prove that the optimal privacy mechanisms for the empirical distribution approach the corresponding mechanisms for the true distribution as the sample size $n$ increases, thereby establishing the statistical consistency of the optimal privacy mechanisms. Finally, we introduce and study uniform privacy mechanisms which, by construction, provide privacy to all the distributions within a neighborhood of the estimated distribution and, thereby, guarantee privacy for the true distribution with high probability."
                    },
                    {
                        "title": "Robustness of Maximal $\u03b1$-Leakage to Side Information",
                        "abstract": "Maximal $\\alpha$-leakage is a tunable measure of information leakage based on the accuracy of guessing an arbitrary function of private data based on public data. The parameter $\\alpha$ determines the loss function used to measure the accuracy of a belief, ranging from log-loss at $\\alpha=1$ to the probability of error at $\\alpha=\\infty$. To study the effect of side information on this measure, we introduce and define conditional maximal $\\alpha$-leakage. We show that, for a chosen mapping (channel) from the actual (viewed as private) data to the released (public) data and some side information, the conditional maximal $\\alpha$-leakage is the supremum (over all side information) of the conditional Arimoto channel capacity where the conditioning is on the side information. We prove that if the side information is conditionally independent of the public data given the private data, the side information cannot increase the information leakage."
                    },
                    {
                        "title": "Aleatoric and Epistemic Discrimination: Fundamental Limits of Fairness Interventions",
                        "abstract": "Machine learning (ML) models can underperform on certain population groups due to choices made during model development and bias inherent in the data. We categorize sources of discrimination in the ML pipeline into two classes: aleatoric discrimination, which is inherent in the data distribution, and epistemic discrimination, which is due to decisions made during model development. We quantify aleatoric discrimination by determining the performance limits of a model under fairness constraints, assuming perfect knowledge of the data distribution. We demonstrate how to characterize aleatoric discrimination by applying Blackwell's results on comparing statistical experiments. We then quantify epistemic discrimination as the gap between a model's accuracy when fairness constraints are applied and the limit posed by aleatoric discrimination. We apply this approach to benchmark existing fairness interventions and investigate fairness risks in data with missing values. Our results indicate that state-of-the-art fairness interventions are effective at removing epistemic discrimination on standard (overused) tabular datasets. However, when data has missing values, there is still significant room for improvement in handling aleatoric discrimination."
                    }
                ]
            }
        }
    },
    "2405.19681": {
        "paper_data": {
            "title": "Bayesian Online Natural Gradient (BONG)",
            "url": "http://arxiv.org/abs/2405.19681v1",
            "arxiv_id": "2405.19681",
            "authors": [
                "Matt Jones",
                "Peter Chang",
                "Kevin Murphy"
            ],
            "abstract": "We propose a novel approach to sequential Bayesian inference based on variational Bayes. The key insight is that, in the online setting, we do not need to add the KL term to regularize to the prior (which comes from the posterior at the previous timestep); instead we can optimize just the expected log-likelihood, performing a single step of natural gradient descent starting at the prior predictive. We prove this method recovers exact Bayesian inference if the model is conjugate, and empirically outperforms other online VB methods in the non-conjugate setting, such as online learning for neural networks, especially when controlling for computational costs.",
            "introduction": "   1 Introduction  Bayesian methods for neural network (NN) training aim to minimize the Kullback-Leibler divergence between true and estimated posterior distributions. This is equivalent to minimizing the variational loss (or negative ELBO)    \u2112\u2062(\ud835\udf4d)=\u2212\ud835\udd3c\ud835\udf3d\u223cq\ud835\udf4d\u2062[log\u2061p\u2062(\ud835\udc9f|\ud835\udf3d)]+D\ud835\udd42\u2062\ud835\udd43\u2062(q\ud835\udf4d|p0)\u2112\ud835\udf4dsubscript\ud835\udd3csimilar-to\ud835\udf3dsubscript\ud835\udc5e\ud835\udf4ddelimited-[]\ud835\udc5dconditional\ud835\udc9f\ud835\udf3dsubscript\ud835\udc37\ud835\udd42\ud835\udd43conditionalsubscript\ud835\udc5e\ud835\udf4dsubscript\ud835\udc5d0\\mathcal{L}(\\bm{\\psi})=-\\mathbb{E}_{{{\\bm{\\theta}}\\sim q_{\\bm{\\psi}}}}\\!\\left[% {\\log p(\\mathcal{D}|{\\bm{\\theta}})}\\right]+D_{\\mathbb{KL}}\\!\\left({q_{\\bm{\\psi% }}}|{p_{0}}\\right)caligraphic_L ( bold_italic_\u03c8 ) = - blackboard_E start_POSTSUBSCRIPT bold_italic_\u03b8 \u223c italic_q start_POSTSUBSCRIPT bold_italic_\u03c8 end_POSTSUBSCRIPT end_POSTSUBSCRIPT [ roman_log italic_p ( caligraphic_D | bold_italic_\u03b8 ) ] + italic_D start_POSTSUBSCRIPT blackboard_K blackboard_L end_POSTSUBSCRIPT ( italic_q start_POSTSUBSCRIPT bold_italic_\u03c8 end_POSTSUBSCRIPT | italic_p start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT )  (1)   Here \ud835\udf3d\ud835\udf3d{\\bm{\\theta}}bold_italic_\u03b8 are the network parameters, \ud835\udf4d\ud835\udf4d\\bm{\\psi}bold_italic_\u03c8 are the variational parameters of the approximate posterior q\ud835\udf4d\u2062(\ud835\udf3d)subscript\ud835\udc5e\ud835\udf4d\ud835\udf3dq_{\\bm{\\psi}}({\\bm{\\theta}})italic_q start_POSTSUBSCRIPT bold_italic_\u03c8 end_POSTSUBSCRIPT ( bold_italic_\u03b8 ), \ud835\udc9f\ud835\udc9f\\mathcal{D}caligraphic_D is the training dataset, and p0\u2062(\ud835\udf3d)subscript\ud835\udc5d0\ud835\udf3dp_{0}({\\bm{\\theta}})italic_p start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ( bold_italic_\u03b8 ) is the prior. The two terms in the variational loss correspond to data fit and regularization to the prior, the latter being analogous to a regularizer r\u2062(\ud835\udf3d)=\u2212log\u2061p0\u2062(\ud835\udf3d)\ud835\udc5f\ud835\udf3dsubscript\ud835\udc5d0\ud835\udf3dr({\\bm{\\theta}})=-\\log p_{0}({\\bm{\\theta}})italic_r ( bold_italic_\u03b8 ) = - roman_log italic_p start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ( bold_italic_\u03b8 ) in traditional point estimation methods like SGD.   An important set of approaches learns the variational parameters by gradient descent on \u2112\u2062(\ud835\udf4d)\u2112\ud835\udf4d\\mathcal{L}(\\bm{\\psi})caligraphic_L ( bold_italic_\u03c8 ) (Blundell et\u00a0al., 2015). More recently Khan and colleagues (Khan et\u00a0al., 2018b; Khan and Rue, 2023; Shen et\u00a0al., 2024) have proposed using the natural gradient \ud835\udc05\ud835\udf4d\u22121\u2062\u2207\ud835\udf4d\u2112\u2062(\ud835\udf4d)superscriptsubscript\ud835\udc05\ud835\udf4d1subscript\u2207\ud835\udf4d\u2112\ud835\udf4d\\mathbf{F}_{\\bm{\\psi}}^{-1}\\nabla_{\\bm{\\psi}}\\mathcal{L}(\\bm{\\psi})bold_F start_POSTSUBSCRIPT bold_italic_\u03c8 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT \u2207 start_POSTSUBSCRIPT bold_italic_\u03c8 end_POSTSUBSCRIPT caligraphic_L ( bold_italic_\u03c8 ) where \ud835\udc05\ud835\udf4dsubscript\ud835\udc05\ud835\udf4d\\mathbf{F}_{\\bm{\\psi}}bold_F start_POSTSUBSCRIPT bold_italic_\u03c8 end_POSTSUBSCRIPT is the Fisher information matrix of the variational family evaluated at q\ud835\udf4dsubscript\ud835\udc5e\ud835\udf4dq_{\\bm{\\psi}}italic_q start_POSTSUBSCRIPT bold_italic_\u03c8 end_POSTSUBSCRIPT. Natural gradient descent (NGD) is often more efficient than vanilla GD because it accounts for the intrinsic geometry of the variational family (Amari, 1998). Khan and Rue (2023) call this approach the \"Bayesian Learning Rule\" or BLR. Using various choices for the variational distribution, generalized losses replacing negative log-likelihood, and other approximations, they reproduce many standard optimization methods such as Adam, and derive new ones.   We study Bayesian NN optimization in online learning, where the data are observed in sequence, \ud835\udc9ft={(\ud835\udc99k,\ud835\udc9ak)k=1t}subscript\ud835\udc9f\ud835\udc61superscriptsubscriptsubscript\ud835\udc99\ud835\udc58subscript\ud835\udc9a\ud835\udc58\ud835\udc581\ud835\udc61\\mathcal{D}_{t}=\\{({\\bm{x}}_{k},{\\bm{y}}_{k})_{k=1}^{t}\\}caligraphic_D start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT = { ( bold_italic_x start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT , bold_italic_y start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ) start_POSTSUBSCRIPT italic_k = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_t end_POSTSUPERSCRIPT }, and the algorithm maintains an approximate posterior q\ud835\udf4dt\u2062(\ud835\udf3dt)\u2248p\u2062(\ud835\udf3dt|\ud835\udc9ft)subscript\ud835\udc5esubscript\ud835\udf4d\ud835\udc61subscript\ud835\udf3d\ud835\udc61\ud835\udc5dconditionalsubscript\ud835\udf3d\ud835\udc61subscript\ud835\udc9f\ud835\udc61q_{\\bm{\\psi}_{t}}({\\bm{\\theta}}_{t})\\approx p({\\bm{\\theta}}_{t}|\\mathcal{D}_{t})italic_q start_POSTSUBSCRIPT bold_italic_\u03c8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT end_POSTSUBSCRIPT ( bold_italic_\u03b8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) \u2248 italic_p ( bold_italic_\u03b8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT | caligraphic_D start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ), which it updates at each step. Fast updates (in terms of both computational speed and statistical efficiency) are critical for many online learning applications (Zhang et\u00a0al., 2024). To allow for nonstationarity in the datastream, we include a time index on \ud835\udf3dtsubscript\ud835\udf3d\ud835\udc61{\\bm{\\theta}}_{t}bold_italic_\u03b8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT, to represent that the parameters may change over time, as is standard for approaches based on state-space models and the extended Kalman filter (see e.g., (Sarkka and Svensson, 2023)). The belief state is updated recursively using the prior q\ud835\udf4dt|t\u22121subscript\ud835\udc5esubscript\ud835\udf4dconditional\ud835\udc61\ud835\udc611q_{\\bm{\\psi}_{t|t-1}}italic_q start_POSTSUBSCRIPT bold_italic_\u03c8 start_POSTSUBSCRIPT italic_t | italic_t - 1 end_POSTSUBSCRIPT end_POSTSUBSCRIPT derived from the previous time step so that the variational loss becomes    \u2112\u2062(\ud835\udf4dt)=\u2212\ud835\udd3c\ud835\udf3dt\u223cq\ud835\udf4dt\u2062[log\u2061p\u2062(\ud835\udc9at|\ud835\udc99t,\ud835\udf3dt)]+D\ud835\udd42\u2062\ud835\udd43\u2062(q\ud835\udf4dt|q\ud835\udf4dt|t\u22121)\u2112subscript\ud835\udf4d\ud835\udc61subscript\ud835\udd3csimilar-tosubscript\ud835\udf3d\ud835\udc61subscript\ud835\udc5esubscript\ud835\udf4d\ud835\udc61delimited-[]\ud835\udc5dconditionalsubscript\ud835\udc9a\ud835\udc61subscript\ud835\udc99\ud835\udc61subscript\ud835\udf3d\ud835\udc61subscript\ud835\udc37\ud835\udd42\ud835\udd43conditionalsubscript\ud835\udc5esubscript\ud835\udf4d\ud835\udc61subscript\ud835\udc5esubscript\ud835\udf4dconditional\ud835\udc61\ud835\udc611\\mathcal{L}(\\bm{\\psi}_{t})=-\\mathbb{E}_{{{\\bm{\\theta}}_{t}\\sim q_{\\bm{\\psi}_{t% }}}}\\!\\left[{\\log p({\\bm{y}}_{t}|{\\bm{x}}_{t},{\\bm{\\theta}}_{t})}\\right]+D_{% \\mathbb{KL}}\\!\\left({q_{\\bm{\\psi}_{t}}}|{q_{\\bm{\\psi}_{t|t-1}}}\\right)caligraphic_L ( bold_italic_\u03c8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) = - blackboard_E start_POSTSUBSCRIPT bold_italic_\u03b8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT \u223c italic_q start_POSTSUBSCRIPT bold_italic_\u03c8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT end_POSTSUBSCRIPT end_POSTSUBSCRIPT [ roman_log italic_p ( bold_italic_y start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT | bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT , bold_italic_\u03b8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) ] + italic_D start_POSTSUBSCRIPT blackboard_K blackboard_L end_POSTSUBSCRIPT ( italic_q start_POSTSUBSCRIPT bold_italic_\u03c8 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT end_POSTSUBSCRIPT | italic_q start_POSTSUBSCRIPT bold_italic_\u03c8 start_POSTSUBSCRIPT italic_t | italic_t - 1",
            "references": [
                {
                    "title": "Outlier-robust Kalman Filtering through Generalised Bayes",
                    "abstract": "We derive a novel, provably robust, and closed-form Bayesian update rule for online filtering in state-space models in the presence of outliers and misspecified measurement models. Our method combines generalised Bayesian inference with filtering methods such as the extended and ensemble Kalman filter. We use the former to show robustness and the latter to ensure computational efficiency in the case of nonlinear models. Our method matches or outperforms other robust filtering methods (such as those based on variational Bayes) at a much lower computational cost. We show this empirically on a range of filtering problems with outlier measurements, such as object tracking, state estimation in high-dimensional chaotic systems, and online learning of neural networks."
                },
                {
                    "title": "Variational Learning is Effective for Large Deep Networks",
                    "abstract": "We give extensive empirical evidence against the common belief that variational learning is ineffective for large neural networks. We show that an optimizer called Improved Variational Online Newton (IVON) consistently matches or outperforms Adam for training large networks such as GPT-2 and ResNets from scratch. IVON's computational costs are nearly identical to Adam but its predictive uncertainty is better. We show several new use cases of IVON where we improve finetuning and model merging in Large Language Models, accurately predict generalization error, and faithfully estimate sensitivity to data. We find overwhelming evidence that variational learning is effective."
                },
                {
                    "title": "Can We Remove the Square-Root in Adaptive Gradient Methods? A Second-Order Perspective",
                    "abstract": "Adaptive gradient optimizers like Adam(W) are the default training algorithms for many deep learning architectures, such as transformers. Their diagonal preconditioner is based on the gradient outer product which is incorporated into the parameter update via a square root. While these methods are often motivated as approximate second-order methods, the square root represents a fundamental difference. In this work, we investigate how the behavior of adaptive methods changes when we remove the root, i.e., strengthen their second-order motivation. Surprisingly, we find that such square-root-free adaptive methods close the generalization gap to SGD on convolutional architectures, while maintaining their root-based counterpart's performance on transformers. The second-order perspective also has practical benefits for developing non-diagonal methods that can incorporate arbitrary curvature approximations through the concept of preconditioner invariance. In contrast to root-based methods like Shampoo, root-free counterparts work well and fast with half-precision since they do not require numerically unstable matrix root decompositions and inversions. Overall, our findings provide new insights into the development of adaptive methods and raise important questions regarding the overlooked role of adaptivity in their success. (experiment code: https://github.com/yorkerlin/remove-the-square-root optimizer code: https://github.com/f-dangel/sirfshampoo)"
                },
                {
                    "title": "Implicit Maximum a Posteriori Filtering via Adaptive Optimization",
                    "abstract": "Bayesian filtering approximates the true underlying behavior of a time-varying system by inverting an explicit generative model to convert noisy measurements into state estimates. This process typically requires either storage, inversion, and multiplication of large matrices or Monte Carlo estimation, neither of which are practical in high-dimensional state spaces such as the weight spaces of artificial neural networks. Here, we frame the standard Bayesian filtering problem as optimization over a time-varying objective. Instead of maintaining matrices for the filtering equations or simulating particles, we specify an optimizer that defines the Bayesian filter implicitly. In the linear-Gaussian setting, we show that every Kalman filter has an equivalent formulation using K steps of gradient descent. In the nonlinear setting, our experiments demonstrate that our framework results in filters that are effective, robust, and scalable to high-dimensional systems, comparing well against the standard toolbox of Bayesian filtering solutions. We suggest that it is easier to fine-tune an optimizer than it is to specify the correct filtering equations, making our framework an attractive option for high-dimensional filtering problems."
                },
                {
                    "title": "Kalman Filter for Online Classification of Non-Stationary Data",
                    "abstract": "In Online Continual Learning (OCL) a learning system receives a stream of data and sequentially performs prediction and training steps. Important challenges in OCL are concerned with automatic adaptation to the particular non-stationary structure of the data, and with quantification of predictive uncertainty. Motivated by these challenges we introduce a probabilistic Bayesian online learning model by using a (possibly pretrained) neural representation and a state space model over the linear predictor weights. Non-stationarity over the linear predictor weights is modelled using a parameter drift transition density, parametrized by a coefficient that quantifies forgetting. Inference in the model is implemented with efficient Kalman filter recursions which track the posterior distribution over the linear weights, while online SGD updates over the transition dynamics coefficient allows to adapt to the non-stationarity seen in data. While the framework is developed assuming a linear Gaussian model, we also extend it to deal with classification problems and for fine-tuning the deep learning representation. In a set of experiments in multi-class classification using data sets such as CIFAR-100 and CLOC we demonstrate the predictive ability of the model and its flexibility to capture non-stationarity."
                },
                {
                    "title": "Low-rank extended Kalman filtering for online learning of neural networks from streaming data",
                    "abstract": "We propose an efficient online approximate Bayesian inference algorithm for estimating the parameters of a nonlinear function from a potentially non-stationary data stream. The method is based on the extended Kalman filter (EKF), but uses a novel low-rank plus diagonal decomposition of the posterior precision matrix, which gives a cost per step which is linear in the number of model parameters. In contrast to methods based on stochastic variational inference, our method is fully deterministic, and does not require step-size tuning. We show experimentally that this results in much faster (more sample efficient) learning, which results in more rapid adaptation to changing distributions, and faster accumulation of reward when used as part of a contextual bandit algorithm."
                },
                {
                    "title": "The Bayesian Learning Rule",
                    "abstract": "We show that many machine-learning algorithms are specific instances of a single algorithm called the \\emph{Bayesian learning rule}. The rule, derived from Bayesian principles, yields a wide-range of algorithms from fields such as optimization, deep learning, and graphical models. This includes classical algorithms such as ridge regression, Newton's method, and Kalman filter, as well as modern deep-learning algorithms such as stochastic-gradient descent, RMSprop, and Dropout. The key idea in deriving such algorithms is to approximate the posterior using candidate distributions estimated by using natural gradients. Different candidate distributions result in different algorithms and further approximations to natural gradients give rise to variants of those algorithms. Our work not only unifies, generalizes, and improves existing algorithms, but also helps us design new ones."
                },
                {
                    "title": "ADAHESSIAN: An Adaptive Second Order Optimizer for Machine Learning",
                    "abstract": "Incorporating second-order curvature information into machine learning optimization algorithms can be subtle, and doing so na\u00efvely can lead to high per-iteration costs associated with forming the Hessian and performing the associated linear system solve. To address this, we introduce ADAHESSIAN, a new stochastic optimization algorithm. ADAHESSIAN directly incorporates approximate curvature information from the loss function, and it includes several novel performance-improving features, including: (i) a fast Hutchinson based method to approximate the curvature matrix with low computational overhead; (ii) a spatial averaging to reduce the variance of the second derivative; and (iii) a root-mean-square exponential moving average to smooth out variations of the second-derivative across different iterations. We perform extensive tests on NLP, CV, and recommendation system tasks, and ADAHESSIAN achieves state-of-the-art results. In particular, we find that ADAHESSIAN: (i) outperforms AdamW for transformers by0.13/0.33 BLEU score on IWSLT14/WMT14, 2.7/1.0 PPLon PTB/Wikitext-103; (ii) outperforms AdamW for Squeeze-Bert by 0.41 points on GLUE; (iii) achieves 1.45%/5.55%higher accuracy on ResNet32/ResNet18 on Cifar10/ImageNetas compared to Adam; and (iv) achieves 0.032% better score than Adagrad for DLRM on the Criteo Ad Kaggle dataset. The cost per iteration of ADAHESSIANis comparable to first-order methods, and ADAHESSIAN exhibits improved robustness towards variations in hyperparameter values. The code for ADAHESSIAN is open-sourced and publicly-available [1]."
                },
                {
                    "title": "Limitations of the Empirical Fisher Approximation",
                    "abstract": "Natural gradient descent, which preconditions a gradient descent update with the Fisher information matrix of the underlying statistical model, is a way to capture partial second-order information. Several highly visible works have advocated an approximation known as the empirical Fisher, drawing connections between approximate second-order methods and heuristics like Adam. We dispute this argument by showing that the empirical Fisher---unlike the Fisher---does not generally capture second-order information. We further argue that the conditions under which the empirical Fisher approaches the Fisher (and the Hessian) are unlikely to be met in practice, and that, even on simple optimization problems, the pathologies of the empirical Fisher can have undesirable effects."
                },
                {
                    "title": "SLANG: Fast Structured Covariance Approximations for Bayesian Deep Learning with Natural Gradient",
                    "abstract": "Uncertainty estimation in large deep-learning models is a computationally challenging task, where it is difficult to form even a Gaussian approximation to the posterior distribution. In such situations, existing methods usually resort to a diagonal approximation of the covariance matrix despite the fact that these matrices are known to give poor uncertainty estimates. To address this issue, we propose a new stochastic, low-rank, approximate natural-gradient (SLANG) method for variational inference in large deep models. Our method estimates a \u201cdiagonal plus low-rank\u201d structure based solely on back-propagated gradients of the network log-likelihood. This requires strictly less gradient computations than methods that compute the gradient of the whole variational objective. Empirical evaluations on standard benchmarks confirm that SLANG enables faster and more accurate estimation of uncertainty than mean-field methods, and performs comparably to state-of-the-art methods."
                },
                {
                    "title": "Fast and Scalable Bayesian Deep Learning by Weight-Perturbation in Adam",
                    "abstract": "Uncertainty computation in deep learning is essential to design robust and reliable systems. Variational inference (VI) is a promising approach for such computation, but requires more effort to implement and execute compared to maximum-likelihood methods. In this paper, we propose new natural-gradient algorithms to reduce such efforts for Gaussian mean-field VI. Our algorithms can be implemented within the Adam optimizer by perturbing the network weights during gradient evaluations, and uncertainty estimates can be cheaply obtained by using the vector that adapts the learning rate. This requires lower memory, computation, and implementation effort than existing VI methods, while obtaining uncertainty estimates of comparable quality. Our empirical results confirm this and further suggest that the weight-perturbation in our algorithm could be useful for exploration in reinforcement learning and stochastic optimization."
                },
                {
                    "title": "Iterative Filtering and Smoothing in Nonlinear and Non-Gaussian Systems Using Conditional Moments",
                    "abstract": "This letter presents the development of novel iterated filters and smoothers that only require specification of the conditional moments of the dynamic and measurement models. This leads to generalizations of the iterated extended Kalman filter, the iterated extended Kalman smoother, the iterated posterior linearization filter, and the iterated posterior linearization smoother. The connections to the previous algorithms are clarified and a convergence analysis is provided. Furthermore, the merits of the proposed algorithms are demonstrated in simulations of the stochastic Ricker map where they are shown to have a similar or superior performance to competing algorithms."
                },
                {
                    "title": "Online natural gradient as a Kalman filter",
                    "abstract": "We establish a full relationship between Kalman filtering and Amari's natural gradient in statistical learning. Namely, using an online natural gradient descent on data log-likelihood to evaluate the parameter of a probabilistic model from a series of observations, is exactly equivalent to using an extended Kalman filter to estimate the parameter (assumed to have constant dynamics). In the i.i.d. case, this relation is a consequence of the \"information filter\" phrasing of the extended Kalman filter. In the recurrent (state space, non-i.i.d.) case, we prove that the joint Kalman filter over states and parameters is a natural gradient on top of real-time recurrent learning (RTRL), a classical algorithm to train recurrent models. This exact algebraic correspondence provides relevant settings for natural gradient hyperparameters such as learning rates or initialization and regularization of the Fisher information matrix."
                },
                {
                    "title": "Weight Uncertainty in Neural Networks",
                    "abstract": "We introduce a new, efficient, principled and backpropagation-compatible algorithm for learning a probability distribution on the weights of a neural network, called Bayes by Backprop. It regularises the weights by minimising a compression cost, known as the variational free energy or the expected lower bound on the marginal likelihood. We show that this principled kind of regularisation yields comparable performance to dropout on MNIST classification. We then demonstrate how the learnt uncertainty in the weights can be used to improve generalisation in non-linear regression problems, and how this weight uncertainty can be used to drive the exploration-exploitation trade-off in reinforcement learning."
                },
                {
                    "title": "New Insights and Perspectives on the Natural Gradient Method",
                    "abstract": "Natural gradient descent is an optimization method traditionally motivated from the perspective of information geometry, and works well for many applications as an alternative to stochastic gradient descent. In this paper we critically analyze this method and its properties, and show how it can be viewed as a type of approximate 2nd-order optimization method, where the Fisher information matrix can be viewed as an approximation of the Hessian. This perspective turns out to have significant implications for how to design a practical and robust version of the method. Additionally, we make the following contributions to the understanding of natural gradient and 2nd-order methods: a thorough analysis of the convergence speed of stochastic natural gradient descent (and more general stochastic 2nd-order methods) as applied to convex quadratics, a critical examination of the oft-used \"empirical\" approximation of the Fisher matrix, and an analysis of the (approximate) parameterization invariance property possessed by natural gradient methods, which we show still holds for certain choices of the curvature matrix other than the Fisher, but notably not the Hessian."
                },
                {
                    "title": "Finding Structure with Randomness: Probabilistic Algorithms for Constructing Approximate Matrix Decompositions",
                    "abstract": "Low-rank matrix approximations, such as the truncated singular value decomposition and the rank-revealing QR decomposition, play a central role in data analysis and scientific computing. This work surveys and extends recent research which demonstrates that randomization offers a powerful tool for performing low-rank matrix approximation. These techniques exploit modern computational architectures more fully than classical methods and open the possibility of dealing with truly massive data sets. This paper presents a modular framework for constructing randomized algorithms that compute partial matrix decompositions. These methods use random sampling to identify a subspace that captures most of the action of a matrix. The input matrix is then compressed\u2014either explicitly or implicitly\u2014to this subspace, and the reduced matrix is manipulated deterministically to obtain the desired low-rank factorization. In many cases, this approach beats its classical competitors in terms of accuracy, robustness, and/or speed. These claims are supported by extensive numerical experiments and a detailed error analysis. The specific benefits of randomized techniques depend on the computational environment. Consider the model problem of finding the $k$ dominant components of the singular value decomposition of an $m \\times n$ matrix. (i) For a dense input matrix, randomized algorithms require $\\bigO(mn \\log(k))$ floating-point operations (flops) in contrast to $ \\bigO(mnk)$ for classical algorithms. (ii) For a sparse input matrix, the flop count matches classical Krylov subspace methods, but the randomized approach is more robust and can easily be reorganized to exploit multiprocessor architectures. (iii) For a matrix that is too large to fit in fast memory, the randomized techniques require only a constant number of passes over the data, as opposed to $\\bigO(k)$ passes for classical algorithms. In fact, it is sometimes possible to perform matrix approximation with a single pass over the data."
                },
                {
                    "title": "Gaussian Processes For Machine Learning",
                    "abstract": "Gaussian processes (GPs) are natural generalisations of multivariate Gaussian random variables to infinite (countably or continuous) index sets. GPs have been applied in a large number of fields to a diverse range of ends, and very many deep theoretical analyses of various properties are available. This paper gives an introduction to Gaussian processes on a fairly elementary level with special emphasis on characteristics relevant in machine learning. It draws explicit connections to branches such as spline smoothing models and support vector machines in which similar ideas have been investigated. Gaussian process models are routinely used to solve hard machine learning problems. They are attractive because of their flexible non-parametric nature and computational simplicity. Treated within a Bayesian framework, very powerful statistical methods can be implemented which offer valid estimates of uncertainties in our predictions and generic model selection procedures cast as nonlinear optimization problems. Their main drawback of heavy computational scaling has recently been alleviated by the introduction of generic sparse approximations.13,78,31 The mathematical literature on GPs is large and often uses deep concepts which are not required to fully understand most machine learning applications. In this tutorial paper, we aim to present characteristics of GPs relevant to machine learning and to show up precise connections to other \"kernel machines\" popular in the community. Our focus is on a simple presentation, but references to more detailed sources are provided."
                },
                {
                    "title": "Natural Gradient Works Efficiently in Learning",
                    "abstract": "When a parameter space has a certain underlying structure, the ordinary gradient of a function does not represent its steepest direction, but the natural gradient does. Information geometry is used for calculating the natural gradients in the parameter space of perceptrons, the space of matrices (for blind source separation), and the space of linear dynamical systems (for blind source deconvolution). The dynamical behavior of natural gradient online learning is analyzed and is proved to be Fisher efficient, implying that it has asymptotically the same performance as the optimal batch estimation of parameters. This suggests that the plateau phenomenon, which appears in the backpropagation learning algorithm of multilayer perceptrons, might disappear or might not be so serious when the natural gradient is used. An adaptive method of updating the learning rate is proposed and analyzed."
                },
                {
                    "title": "Decoupled extended Kalman filter training of feedforward layered networks",
                    "abstract": "Presents a training algorithm for feedforward layered networks based on a decoupled extended Kalman filter (DEKF). The authors present an artificial process noise extension to DEKF that increases its convergence rate and assists in the avoidance of local minima. Computationally efficient formulations for two particularly natural and useful cases of DEKF are given. Through a series of pattern classification and function approximation experiments, three members of DEKF are compared with one another and with standard backpropagation (SBP). These studies demonstrate that the judicious grouping of weights along with the use of artificial process noise in DEKF result in input-output mapping performance that is comparable to the global extended Kalman algorithm, and is often superior to SBP, while requiring significantly fewer presentations of training data than SBP and less overall training time than either of these procedures.<<ETX>>"
                },
                {
                    "title": "Optimal Information Processing and Bayes's Theorem",
                    "abstract": "Abstract In this article statistical inference is viewed as information processing involving input information and output information. After introducing information measures for the input and output information, an information criterion functional is formulated and optimized to obtain an optimal information processing rule (IPR). For the particular information measures and criterion functional adopted, it is shown that Bayes's theorem is the optimal IPR. This optimal IPR is shown to be 100% efficient in the sense that its use leads to the output information being exactly equal to the given input information. Also, the analysis links Bayes's theorem to maximum-entropy considerations."
                },
                {
                    "title": "A useful theorem for nonlinear devices having Gaussian inputs",
                    "abstract": "If and only if the inputs to a set of nonlinear, zero-memory devices are variates drawn from a Gaussian random process, a useful general relationship may be found between certain input and output statistics of the set. This relationship equates partial derivatives of the (high-order) output correlation coefficient taken with respect to the input correlation coefficients, to the output correlation coefficient of a new set of nonlinear devices bearing a simple derivative relation to the original set. Application is made to the interesting special cases of conventional cross-correlation and autocorrelation functions, and Bussgang's theorem is easily proved. As examples, the output autocorrelation functions are simply obtained for a hard limiter, linear detector, clipper, and smooth limiter."
                },
                {
                    "title": "On diagonal approximations to the extended Kalman filter for online training of Bayesian neural networks",
                    "abstract": "We present two approaches to approximate online Bayesian inference for the parameters of DNNs. Both are based on diagonal Gaussian approximations and linearize the network at each step to ensure efficient computation. The first approach optimizes the exclusive KL, D KL ( q \u2225 p ); this amounts to matching the marginal mean and precision of p and q . The second approach optimizes the inclusive KL, D KL ( q \u2225 p ), which amounts to matching the marginal mean and variance of p and q . The latter approach turns out to be equivalent to the previously proposed \u201cfully decoupled EKF\u201d approach. We show experimentally that inclusive KL yields better results than exclusive KL, and both are more effective than one-pass SGD."
                },
                {
                    "title": "Improving predictions of Bayesian neural nets via local linearization",
                    "abstract": "The generalized Gauss-Newton (ggn) approximation is often used to make practical Bayesian deep learning approaches scalable by replacing a second order derivative with a product of first order derivatives. In this paper we argue that the ggn approximation should be understood as a local linearization of the underlying Bayesian neural network (bnn), which turns the bnn into a generalized linear model (glm). Because we use this linearized model for posterior inference, we should also predict using this modified model instead of the original one. We refer to this modified predictive as \u201cglm predictive\u201d and show that it effectively resolves common underfitting problems of the Laplace approximation. It extends previous results in this vein to general likelihoods and has an equivalent Gaussian process formulation, which enables alternative inference schemes for bnns in function space. We demonstrate the effectiveness of our approach on several standard classification datasets and on out-of-distribution detection. We provide an implementation at https://github.com/ AlexImmer/BNN-predictions."
                },
                {
                    "title": "Efficient Low Rank Gaussian Variational Inference for Neural Networks",
                    "abstract": "Bayesian neural networks are enjoying a renaissance driven in part by recent advances in variational inference (VI). The most common form of VI employs a fully factorized or mean-\ufb01 eld distribution, but this is known to suffer from several pathologies, especially as we expect posterior distributions with highly correlated parameters. Current algorithms that capture these correlations with a Gaussian approximating family are dif \ufb01 cult to scale to large models due to computational costs and high variance of gradient updates. By using a new form of the reparametrization trick, we derive a computationally ef \ufb01 cient algorithm for performing VI with a Gaussian family with a low-rank plus diagonal covariance structure. We scale to deep feed-forward and convolutional architectures. We \ufb01 nd that adding low-rank terms to parametrized diagonal covariance does not improve predictive performance except on small networks, but low-rank terms added to a constant diagonal covariance improves performance on small and large-scale network architectures."
                },
                {
                    "title": "A stochastic estimator of the trace of the influence matrix for laplacian smoothing splines",
                    "abstract": "An unbiased stochastic estimator of tr(I-A), where A is the influence matrix associated with the calculation of Laplacian smoothing splines, is described. The estimator is similar to one recently developed by Girard but satisfies a minimum variance criterion and does not require the simulation of a standard normal variable. It uses instead simulations of the discrete random variable which takes the values 1, -1 each with probability 1/2. Bounds on the variance of the estimator, similar to those established by Girard, are obtained using elementary methods. The estimator can be used to approximately minimize generalised cross validation (GCV) when using discretized iterative methods for fitting Laplacian smoothing splines to very large data sets. Simulated examples show that the estimated trace values, using either the estimator presented here or the estimator of Girard, perform almost as well as the exact values when applied to the minimization of GCV for n as small as a few hundred, where n is the number ..."
                },
                {
                    "title": "Training Multilayer Perceptrons with the Extende Kalman Algorithm",
                    "abstract": "A large fraction of recent work in artificial neural nets uses multilayer perceptrons trained with the back-propagation algorithm described by Rumelhart et. al. This algorithm converges slowly for large or complex problems such as speech recognition, where thousands of iterations may be needed for convergence even with small data sets. In this paper, we show that training multilayer perceptrons is an identification problem for a nonlinear dynamic system which can be solved using the Extended Kalman Algorithm. Although computationally complex, the Kalman algorithm usually converges in a few iterations. We describe the algorithm and compare it with back-propagation using two-dimensional examples."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG",
                "stat.CO"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively optimize Bayesian neural networks in online learning settings to maintain an accurate approximate posterior while adapting to nonstationary data streams?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning, particularly in applications where data is continuously generated, such as in finance, healthcare, and autonomous systems. By improving Bayesian neural network optimization in online learning, we can enhance the model's ability to adapt to changing environments, leading to more robust and reliable predictions. This research could pave the way for new methodologies that integrate Bayesian principles with online learning, potentially influencing future studies on adaptive learning algorithms and their applications in real-time decision-making systems.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in this problem stem from the need to balance computational efficiency with statistical accuracy in a dynamic environment. Naive approaches, such as standard gradient descent, may fail to capture the complexities of the posterior distribution in real-time, leading to suboptimal performance. Additionally, the nonstationarity of data streams introduces further complications, as the model parameters must be continuously updated to reflect the latest information. Overcoming these technical obstacles requires sophisticated methods that can efficiently update the belief state while managing the inherent uncertainties in the data.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on static datasets or has not adequately addressed the challenges posed by online learning and nonstationary data. Many existing solutions lack the flexibility to adapt to changing distributions over time, which has limited their applicability in real-world scenarios. Additionally, prior work may not have fully leveraged the potential of Bayesian methods in conjunction with online learning frameworks. Our approach differs by incorporating a time index on the parameters and utilizing a recursive update mechanism, which allows for a more effective adaptation to the evolving data stream.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using a Bayesian framework for neural network optimization in online learning, where we maintain an approximate posterior that is updated at each time step. We will employ the variational loss function that incorporates both data fit and regularization terms, specifically designed for nonstationary data. The dataset will consist of sequentially observed data points, and we will evaluate our approach using metrics such as predictive accuracy and computational efficiency. The expected outcomes include improved model performance in adapting to changing data distributions and enhanced"
            }
        },
        "author_data": {
            "8c8d7c58-02a3-4495-af8d-0d0fa0a4d2bc": {
                "pk": "8c8d7c58-02a3-4495-af8d-0d0fa0a4d2bc",
                "name": "Matt Jones",
                "collaborators": [
                    "Martin Scharlemann",
                    "Alisson Tezzin",
                    "Elie Wolfe",
                    "Barbara Amaral",
                    "James M. Foster",
                    "Sebastian Bobadilla-Suarez",
                    "Bradley C. Love",
                    "Yanlai Yang",
                    "Michael C. Mozer",
                    "Mengye Ren"
                ],
                "domain": [
                    "Quantum Contextuality",
                    "Causal Models",
                    "Neural Networks",
                    "Analogical Reasoning"
                ],
                "publications": [
                    {
                        "title": "On Principles and Representations for Extended Contextuality",
                        "abstract": "There has been recent interest in whether the concept of quantum contextuality can be extended to systems with disturbance or signaling while retaining the essential properties of standard contextuality. Dzhafarov and Kujala (arXiv:2302.11995) offer a provocative argument that the answer is always affirmative: If standard contextuality satisfies some principle that extended contextuality does not, then that principle must be `non-substantive' in that it depends on a superficial choice of representation. Their argument is based on a ``consistification'' procedure that maps all systems to nondisturbing ones and that preserves contextuality. This paper raises several objections to their argument, including that it neglects how substantive principles change their expression under a change of representation, that it begs the question by presuming the principles are based on non-substantive transformations, and that the argument applies universally to arbitrary extensions of contextuality."
                    },
                    {
                        "title": "Relating Causal and Probabilistic Approaches to Contextuality",
                        "abstract": "A primary goal in recent research on contextuality has been to extend this concept to cases of inconsistent connectedness, where observables have different distributions in different contexts. This article proposes a solution within the framework of probabilistic causal models, which extend hidden-variables theories, and then demonstrates an equivalence to the contextuality-by-default (CbD) framework. CbD distinguishes contextuality from direct influences of context on observables, defining the latter purely in terms of probability distributions. Here we take a causal view of direct influences, defining direct influence within any causal model as the probability of all latent states of the system in which a change of context changes the outcome of a measurement. Model-based contextuality (M-contextuality) is then defined as the necessity of stronger direct influences to model a full system than when considered individually. For consistently connected systems, M-contextuality agrees with standard contextuality. For general systems, it is proved that M-contextuality is equivalent to the property that any model of a system must contain \"hidden influences\", meaning direct influences that go in opposite directions for different latent states, or equivalently signaling between observers that carries no information. This criterion can be taken as formalizing the \"no-conspiracy\" principle that has been proposed in connection with CbD. M-contextuality is then proved to be equivalent to CbD-contextuality, thus providing a new interpretation of CbD-contextuality as the non-existence of a model for a system without hidden direct influences."
                    },
                    {
                        "title": "How a strongly irreducible Heegaard splitting intersects a handlebody",
                        "abstract": "It has been shown elsewhere that a strongly irreducible Heegaard splitting surface Q of a compact orientable 3-manifold M can, under reasonable side conditions, intersect a ball or a solid torus in M in only a few possible ways. Here we extend those results to describe how Q can intersect a handlebody in M."
                    },
                    {
                        "title": "Impossibility Theorem for Extending Contextuality to Disturbing Systems",
                        "abstract": "Recently there has been interest, and impressive progress, in extending the definition of contextuality to systems with disturbance. We prove here that such an endeavor cannot simultaneously satisfy the following core principles of contextuality: (1) Measuring more information cannot change a contextual system to a noncontextual one. (2) Classical post-processing cannot create contextuality: appending new observables that are functions of given observables cannot change a noncontextual system to a contextual one. (3) The joint realization of two statistically independent noncontextual systems is noncontextual. (4) Determinism cannot create contextuality: Any deterministic system is noncontextual, and adding deterministic observables to a noncontextual system cannot yield a contextual one. We also prove the same result without Principle 3, under the stronger version of Principle 4. Moreover, our results hold for restricted extensions of contextuality that apply only to systems of binary observables. In addition to these proofs, we analyze several particular proposals and identify which of our axioms they obey and which they violate."
                    },
                    {
                        "title": "Reinforcement Learning with Analogical Similarity to Guide Schema Induction and Attention",
                        "abstract": "Research in analogical reasoning suggests that higher-order cognitive functions such as abstract reasoning, far transfer, and creativity are founded on recognizing structural similarities among relational systems. Here we integrate theories of analogy with the computational framework of reinforcement learning (RL). We propose a psychology theory that is a computational synergy between analogy and RL, in which analogical comparison provides the RL learning algorithm with a measure of relational similarity, and RL provides feedback signals that can drive analogical learning. Simulation results support the power of this approach."
                    },
                    {
                        "title": "Robust priors for regularized regression",
                        "abstract": "Induction benefits from useful priors. Penalized regression approaches, like ridge regression, shrink weights toward zero but zero association is usually not a sensible prior. Inspired by simple and robust decision heuristics humans use, we constructed non-zero priors for penalized regression models that provide robust and interpretable solutions across several tasks. Our approach enables estimates from a constrained model to serve as a prior for a more general model, yielding a principled way to interpolate between models of differing complexity. We successfully applied this approach to a number of decision and classification problems, as well as analyzing simulated brain imaging data. Models with robust priors had excellent worst-case performance. Solutions followed from the form of the heuristic that was used to derive the prior. These new algorithms can serve applications in data analysis and machine learning, as well as help in understanding how people transition from novice to expert performance."
                    },
                    {
                        "title": "Reawakening knowledge: Anticipatory recovery from catastrophic interference via structured training",
                        "abstract": "We explore the training dynamics of neural networks in a structured non-IID setting where documents are presented cyclically in a fixed, repeated sequence. Typically, networks suffer from catastrophic interference when training on a sequence of documents; however, we discover a curious and remarkable property of LLMs fine-tuned sequentially in this setting: they exhibit anticipatory behavior, recovering from the forgetting on documents before encountering them again. The behavior emerges and becomes more robust as the architecture scales up its number of parameters. Through comprehensive experiments and visualizations, we uncover new insights into training over-parameterized networks in structured environments."
                    }
                ]
            },
            "d988c985-d373-4034-bf75-ad327da6e357": {
                "pk": "d988c985-d373-4034-bf75-ad327da6e357",
                "name": "Peter Chang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "defc5715-cb1a-4e35-a17e-a126f0a780a7": {
                "pk": "defc5715-cb1a-4e35-a17e-a126f0a780a7",
                "name": "Kevin Murphy",
                "collaborators": [
                    "Yair Weiss",
                    "Mark Schmidt",
                    "Nando de Freitas",
                    "Daniel Eaton",
                    "Jonathan Huang",
                    "Michael I. Jordan",
                    "Gerardo Duran-Martin",
                    "Aleyna Kara",
                    "Nir Friedman",
                    "Stuart Russell"
                ],
                "domain": [
                    "Bayesian Networks",
                    "Graphical Models",
                    "Machine Learning",
                    "Inference"
                ],
                "publications": [
                    {
                        "title": "A Variational Approximation for Bayesian Networks with Discrete and Continuous Latent Variables",
                        "abstract": "We show how to use a variational approximation to the logistic function to perform approximate inference in Bayesian networks containing discrete nodes with continuous parents. Essentially, we convert the logistic function to a Gaussian, which facilitates exact inference, and then iteratively adjust the variational parameters to improve the quality of the approximation. We demonstrate experimentally that this approximation is faster and potentially more accurate than sampling. We also introduce a simple new technique for handling evidence, which allows us to handle arbitrary distributions on observed nodes, as well as achieving a significant speedup in networks with discrete variables of large cardinality."
                    },
                    {
                        "title": "The Factored Frontier Algorithm for Approximate Inference in DBNs",
                        "abstract": "The Factored Frontier (FF) algorithm is a simple approximate inferencealgorithm for Dynamic Bayesian Networks (DBNs). It is very similar tothe fully factorized version of the Boyen-Koller (BK) algorithm, butinstead of doing an exact update at every step followed bymarginalisation (projection), it always works with factoreddistributions. Hence it can be applied to models for which the exactupdate step is intractable. We show that FF is equivalent to (oneiteration of) loopy belief propagation (LBP) on the original DBN, andthat BK is equivalent (to one iteration of) LBP on a DBN where wecluster some of the nodes. We then show empirically that byiterating, LBP can improve on the accuracy of both FF and BK. Wecompare these algorithms on two real-world DBNs: the first is a modelof a water treatment plant, and the second is a coupled HMM, used tomodel freeway traffic."
                    },
                    {
                        "title": "Proceedings of the Twenty-Eighth Conference on Uncertainty in Artificial Intelligence (2012)",
                        "abstract": "This is the Proceedings of the Twenty-Eighth Conference on Uncertainty in Artificial Intelligence, which was held on Catalina Island, CA August 14-18 2012."
                    },
                    {
                        "title": "Modeling Discrete Interventional Data using Directed Cyclic Graphical Models",
                        "abstract": "We outline a representation for discrete multivariate distributions in terms of interventional potential functions that are globally normalized. This representation can be used to model the effects of interventions, and the independence properties encoded in this model can be represented as a directed graph that allows cycles. In addition to discussing inference and sampling with this representation, we give an exponential family parametrization that allows parameter estimation to be stated as a convex optimization problem; we also give a convex relaxation of the task of simultaneous parameter and structure learning using group l1-regularization. The model is evaluated on simulated data and intracellular flow cytometry data."
                    },
                    {
                        "title": "Bayesian structure learning using dynamic programming and MCMC",
                        "abstract": "MCMC methods for sampling from the space of DAGs can mix poorly due to the local nature of the proposals that are commonly used. It has been shown that sampling from the space of node orders yields better results [FK03, EW06]. Recently, Koivisto and Sood showed how one can analytically marginalize over orders using dynamic programming (DP) [KS04, Koi06]. Their method computes the exact marginal posterior edge probabilities, thus avoiding the need for MCMC. Unfortunately, there are four drawbacks to the DP technique: it can only use modular priors, it can only compute posteriors over modular features, it is difficult to compute a predictive density, and it takes exponential time and space. We show how to overcome the first three of these problems by using the DP algorithm as a proposal distribution for MCMC in DAG space. We show that this hybrid technique converges to the posterior faster than other methods, resulting in more accurate structure learning and higher predictive likelihoods on test data."
                    },
                    {
                        "title": "Efficient inference in occlusion-aware generative models of images",
                        "abstract": "We present a generative model of images based on layering, in which image layers are individually generated, then composited from front to back. We are thus able to factor the appearance of an image into the appearance of individual objects within the image --- and additionally for each individual object, we can factor content from pose. Unlike prior work on layered models, we learn a shape prior for each object/layer, allowing the model to tease out which object is in front by looking for a consistent shape, without needing access to motion cues or any labeled data. We show that ordinary stochastic gradient variational bayes (SGVB), which optimizes our fully differentiable lower-bound on the log-likelihood, is sufficient to learn an interpretable representation of images. Finally we present experiments demonstrating the effectiveness of the model for inferring foreground and background objects in images."
                    },
                    {
                        "title": "Loopy Belief Propagation for Approximate Inference: An Empirical Study",
                        "abstract": "Recently, researchers have demonstrated that loopy belief propagation - the use of Pearls polytree algorithm IN a Bayesian network WITH loops OF error- correcting codes.The most dramatic instance OF this IS the near Shannon - limit performance OF Turbo Codes codes whose decoding algorithm IS equivalent TO loopy belief propagation IN a chain - structured Bayesian network. IN this paper we ask : IS there something special about the error - correcting code context, OR does loopy propagation WORK AS an approximate inference schemeIN a more general setting? We compare the marginals computed using loopy propagation TO the exact ones IN four Bayesian network architectures, including two real - world networks : ALARM AND QMR.We find that the loopy beliefs often converge AND WHEN they do, they give a good approximation TO the correct marginals.However,ON the QMR network, the loopy beliefs oscillated AND had no obvious relationship TO the correct posteriors. We present SOME initial investigations INTO the cause OF these oscillations, AND show that SOME simple methods OF preventing them lead TO the wrong results."
                    },
                    {
                        "title": "Efficient Online Bayesian Inference for Neural Bandits",
                        "abstract": "In this paper we present a new algorithm for online (sequential) inference in Bayesian neural networks, and show its suitability for tackling contextual bandit problems. The key idea is to combine the extended Kalman filter (which locally linearizes the likelihood function at each time step) with a (learned or random) low-dimensional affine subspace for the parameters; the use of a subspace enables us to scale our algorithm to models with $\\sim 1M$ parameters. While most other neural bandit methods need to store the entire past dataset in order to avoid the problem of \"catastrophic forgetting\", our approach uses constant memory. This is possible because we represent uncertainty about all the parameters in the model, not just the final linear layer. We show good results on the \"Deep Bayesian Bandit Showdown\" benchmark, as well as MNIST and a recommender system."
                    },
                    {
                        "title": "Learning the Structure of Dynamic Probabilistic Networks",
                        "abstract": "Dynamic probabilistic networks are a compact representation of complex stochastic processes. In this paper we examine how to learn the structure of a DPN from data. We extend structure scoring rules for standard probabilistic networks to the dynamic case, and show how to search for structure when some of the variables are hidden. Finally, we examine two applications where such a technology might be useful: predicting and classifying dynamic behaviors, and learning causal orderings in biological processes. We provide empirical results that demonstrate the applicability of our methods in both domains."
                    },
                    {
                        "title": "Learning Video Representations using Contrastive Bidirectional Transformer",
                        "abstract": "This paper proposes a self-supervised learning approach for video features that results in significantly improved performance on downstream tasks (such as video classification, captioning and segmentation) compared to existing methods. Our method extends the BERT model for text sequences to the case of sequences of real-valued feature vectors, by replacing the softmax loss with noise contrastive estimation (NCE). We also show how to learn representations from sequences of visual features and sequences of words derived from ASR (automatic speech recognition), and show that such cross-modal training (when possible) helps even more."
                    },
                    {
                        "title": "Towards Differentiable Resampling",
                        "abstract": "Resampling is a key component of sample-based recursive state estimation in particle filters. Recent work explores differentiable particle filters for end-to-end learning. However, resampling remains a challenge in these works, as it is inherently non-differentiable. We address this challenge by replacing traditional resampling with a learned neural network resampler. We present a novel network architecture, the particle transformer, and train it for particle resampling using a likelihood-based loss function over sets of particles. Incorporated into a differentiable particle filter, our model can be end-to-end optimized jointly with the other particle filter components via gradient descent. Our results show that our learned resampler outperforms traditional resampling techniques on synthetic data and in a simulated robot localization task."
                    },
                    {
                        "title": "Uncertainty Disentanglement with Non-stationary Heteroscedastic Gaussian Processes for Active Learning",
                        "abstract": "Gaussian processes are Bayesian non-parametric models used in many areas. In this work, we propose a Non-stationary Heteroscedastic Gaussian process model which can be learned with gradient-based techniques. We demonstrate the interpretability of the proposed model by separating the overall uncertainty into aleatoric (irreducible) and epistemic (model) uncertainty. We illustrate the usability of derived epistemic uncertainty on active learning problems. We demonstrate the efficacy of our model with various ablations on multiple datasets."
                    },
                    {
                        "title": "Risk score learning for COVID-19 contact tracing apps",
                        "abstract": "Digital contact tracing apps for COVID, such as the one developed by Google and Apple, need to estimate the risk that a user was infected during a particular exposure, in order to decide whether to notify the user to take precautions, such as entering into quarantine, or requesting a test. Such risk score models contain numerous parameters that must be set by the public health authority. In this paper, we show how to automatically learn these parameters from data.   Our method needs access to exposure and outcome data. Although this data is already being collected (in an aggregated, privacy-preserving way) by several health authorities, in this paper we limit ourselves to simulated data, so that we can systematically study the different factors that affect the feasibility of the approach. In particular, we show that the parameters become harder to estimate when there is more missing data (e.g., due to infections which were not recorded by the app), and when there is model misspecification. Nevertheless, the learning approach outperforms a strong manually designed baseline. Furthermore, the learning approach can adapt even when the risk factors of the disease change, e.g., due to the evolution of new variants, or the adoption of vaccines."
                    },
                    {
                        "title": "Group Sparse Priors for Covariance Estimation",
                        "abstract": "Recently it has become popular to learn sparse Gaussian graphical models (GGMs) by imposing l1 or group l1,2 penalties on the elements of the precision matrix. Thispenalized likelihood approach results in a tractable convex optimization problem. In this paper, we reinterpret these results as performing MAP estimation under a novel prior which we call the group l1 and l1,2 positivedefinite matrix distributions. This enables us to build a hierarchical model in which the l1 regularization terms vary depending on which group the entries are assigned to, which in turn allows us to learn block structured sparse GGMs with unknown group assignments. Exact inference in this hierarchical model is intractable, due to the need to compute the normalization constant of these matrix distributions. However, we derive upper bounds on the partition functions, which lets us use fast variational inference (optimizing a lower bound on the joint posterior). We show that on two real world data sets (motion capture and financial data), our method which infers the block structure outperforms a method that uses a fixed block structure, which in turn outperforms baseline methods that ignore block structure."
                    },
                    {
                        "title": "A Review of Relational Machine Learning for Knowledge Graphs",
                        "abstract": "Relational machine learning studies methods for the statistical analysis of relational, or graph-structured, data. In this paper, we provide a review of how such statistical models can be \"trained\" on large knowledge graphs, and then used to predict new facts about the world (which is equivalent to predicting new edges in the graph). In particular, we discuss two fundamentally different kinds of statistical relational models, both of which can scale to massive datasets. The first is based on latent feature models such as tensor factorization and multiway neural networks. The second is based on mining observable patterns in the graph. We also show how to combine these latent and observable models to get improved modeling power at decreased computational cost. Finally, we discuss how such statistical models of graphs can be combined with text-based information extraction methods for automatically constructing knowledge graphs from the Web. To this end, we also discuss Google's Knowledge Vault project as an example of such combination."
                    }
                ]
            }
        }
    },
    "2405.20053": {
        "paper_data": {
            "title": "Would I Lie To You? Inference Time Alignment of Language Models using Direct Preference Heads",
            "url": "http://arxiv.org/abs/2405.20053v1",
            "arxiv_id": "2405.20053",
            "authors": [
                "Avelina Asada Hadji-Kyriacou",
                "Ognjen Arandjelovic"
            ],
            "abstract": "Pre-trained Language Models (LMs) exhibit strong zero-shot and in-context learning capabilities; however, their behaviors are often difficult to control. By utilizing Reinforcement Learning from Human Feedback (RLHF), it is possible to fine-tune unsupervised LMs to follow instructions and produce outputs that reflect human preferences. Despite its benefits, RLHF has been shown to potentially harm a language model's reasoning capabilities and introduce artifacts such as hallucinations where the model may fabricate facts. To address this issue we introduce Direct Preference Heads (DPH), a fine-tuning framework that enables LMs to learn human preference signals through an auxiliary reward head without directly affecting the output distribution of the language modeling head. We perform a theoretical analysis of our objective function and find strong ties to Conservative Direct Preference Optimization (cDPO). Finally we evaluate our models on GLUE, RACE, and the GPT4All evaluation suite and demonstrate that our method produces models which achieve higher scores than those fine-tuned with Supervised Fine-Tuning (SFT) or Direct Preference Optimization (DPO) alone.",
            "introduction": "   1 Introduction  Reinforcement Learning from Human Feedback (RLHF) is a technique that can be used to align an agent \u2014 such as a Large Language Model (LLM) \u2014 to human preferences and lead to more truthful, more helpful, less harmful and more preferred outputs [31]. Proximal Policy Optimization (PPO) [38] and Direct Preference Optimization (DPO) [33] are two such aligment techniques which have been extensively used to improve the quality of LLM outputs, leading to instruction following agents or chat assistants which are quickly approaching human-baselines in a variety of knowledge and reasoning tasks [5, 11, 43, 20, 26, 37, 12].   However, recent research has shown that RLHF may actually hurt an LLM\u2019s reasoning abilities rather than improving it. One study [6] discovered that performing alignment during the Supervised Fine-Tuning (SFT) stage of training may lead to worse performance on reasoning benchmarks, and another [4] discovered that SFT alone outperforms RLHF for smaller models with the benefits of RLHF only emerging for models with more than 1 Billion parameters. Ouyang et al. [31] also reports an increased tendency for RLHF models to make up information in closed domain tasks (\u201challucination\u201d) compared to models trained with SFT alone.   To combat the the risk of RLHF compromising the abilities of an LLM in favor of producing preferable outputs we introduce Direct Preference Heads (DPH), a novel feature based approach that optimises a reward score produced by the LLM rather than optimising the logits produced by language modelling head. DPH can be used in combination with (or without) existing alignment techniques to allow language models to self-evaluate outputs sampled at inference time and select the highest scoring candidate.   We evaluate the performance of DPH using an efficient 551M parameter LM on a variety of commonsense reasoning and Natural Language Understanding (NLU) tasks. All code used to train our models is available on GitHub and we release our model weights on Hugging Face.     2 Prior Approaches  Prior approaches to language model alignment involve directly optimizing the logits produced by the language modelling head to increase the likelihood of producing preferable responses while decreasing the likelihood of undesirable responses.    2.1 Reinforcement Learning from Human Feedback (RLHF)  Reinforcement Learning from Human Feedback seeks to learn a reward model from human feedback on completions generated by a language model which can be used to align an LM with human preferences. A typical RLHF pipeline consists of 3 steps: (1) supervised fine-tuning, (2) preference sampling and reward modelling, and (3) RL fine-tuning.   Supervised Fine-Tuning\u00a0 The first step of a standard RLHF pipeline is fine-tuning a pre-trained LM on high quality data for downstream tasks to obtain a model \u03c0SFTsuperscript\ud835\udf0bSFT\\pi^{\\text{SFT}}italic_\u03c0 start_POSTSUPERSCRIPT SFT end_POSTSUPERSCRIPT.   Reward Modelling\u00a0 Next, the SFT model is prompted with input tokens x\ud835\udc65xitalic_x to produce completions y\ud835\udc66yitalic_y. These answers are then rated by human labelers which rate the answers based on one or more criteria. A reward model r\u03d5\u2062(x,y)subscript\ud835\udc5fitalic-\u03d5\ud835\udc65\ud835\udc66r_{\\phi}(x,y)italic_r start_POSTSUBSCRIPT italic_\u03d5 end_POSTSUBSCRIPT ( italic_x , italic_y ) is then trained to estimate the scores assigned by human labelers using maximum likelihood estimation.   RL Fine-Tuning\u00a0 During the RL phase the learned reward function is used to provide feedback to",
            "references": [
                {
                    "title": "Qwen Technical Report",
                    "abstract": "Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural language processing tasks that were previously thought to be exclusive to humans. In this work, we introduce Qwen, the first installment of our large language model series. Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts. It includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques. The base language models consistently demonstrate superior performance across a multitude of downstream tasks, and the chat models, particularly those trained using Reinforcement Learning from Human Feedback (RLHF), are highly competitive. The chat models possess advanced tool-use and planning capabilities for creating agent applications, showcasing impressive performance even when compared to bigger models on complex tasks like utilizing a code interpreter. Furthermore, we have developed coding-specialized models, Code-Qwen and Code-Qwen-Chat, as well as mathematics-focused models, Math-Qwen-Chat, which are built upon base language models. These models demonstrate significantly improved performance in comparison with open-source models, and slightly fall behind the proprietary models."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "LaProp: Separating Momentum and Adaptivity in Adam",
                    "abstract": "We identity a by-far-unrecognized problem of Adam-style optimizers which results from unnecessary coupling between momentum and adaptivity. The coupling leads to instability and divergence when the momentum and adaptivity parameters are mismatched. In this work, we propose a method, Laprop, which decouples momentum and adaptivity in the Adam-style methods. We show that the decoupling leads to greater flexibility in the hyperparameters and allows for a straightforward interpolation between the signed gradient methods and the adaptive gradient methods. We experimentally show that Laprop has consistently improved speed and stability over Adam on a variety of tasks. We also bound the regret of Laprop on a convex problem and show that our bound differs from that of Adam by a key factor, which demonstrates its advantage."
                },
                {
                    "title": "GLU Variants Improve Transformer",
                    "abstract": "Gated Linear Units (arXiv:1612.08083) consist of the component-wise product of two linear projections, one of which is first passed through a sigmoid function. Variations on GLU are possible, using different nonlinear (or even linear) functions in place of sigmoid. We test these variants in the feed-forward sublayers of the Transformer (arXiv:1706.03762) sequence-to-sequence model, and find that some of them yield quality improvements over the typically-used ReLU or GELU activations."
                },
                {
                    "title": "PIQA: Reasoning about Physical Commonsense in Natural Language",
                    "abstract": "To apply eyeshadow without a brush, should I use a cotton swab or a toothpick? Questions requiring this kind of physical commonsense pose a challenge to today's natural language understanding systems. While recent pretrained models (such as BERT) have made progress on question answering over more abstract domains \u2013 such as news articles and encyclopedia entries, where text is plentiful \u2013 in more physical domains, text is inherently limited due to reporting bias. Can AI systems learn to reliably answer physical commonsense questions without experiencing the physical world?In this paper, we introduce the task of physical commonsense reasoning and a corresponding benchmark dataset Physical Interaction: Question Answering or PIQA. Though humans find the dataset easy (95% accuracy), large pretrained models struggle (\u223c75%). We provide analysis about the dimensions of knowledge that existing models lack, which offers significant opportunities for future research."
                },
                {
                    "title": "HellaSwag: Can a Machine Really Finish Your Sentence?",
                    "abstract": "Recent work by Zellers et al. (2018) introduced a new task of commonsense natural language inference: given an event description such as \u201cA woman sits at a piano,\u201d a machine must select the most likely followup: \u201cShe sets her fingers on the keys.\u201d With the introduction of BERT, near human-level performance was reached. Does this mean that machines can perform human level commonsense inference? In this paper, we show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans (>95% accuracy), state-of-the-art models struggle (<48%). We achieve this via Adversarial Filtering (AF), a data collection paradigm wherein a series of discriminators iteratively select an adversarial set of machine-generated wrong answers. AF proves to be surprisingly robust. The key insight is to scale up the length and complexity of the dataset examples towards a critical \u2018Goldilocks\u2019 zone wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models. Our construction of HellaSwag, and its resulting difficulty, sheds light on the inner workings of deep pretrained models. More broadly, it suggests a new path forward for NLP research, in which benchmarks co-evolve with the evolving state-of-the-art in an adversarial way, so as to present ever-harder challenges."
                },
                {
                    "title": "BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions",
                    "abstract": "In this paper we study yes/no questions that are naturally occurring \u2014 meaning that they are generated in unprompted and unconstrained settings. We build a reading comprehension dataset, BoolQ, of such questions, and show that they are unexpectedly challenging. They often query for complex, non-factoid information, and require difficult entailment-like inference to solve. We also explore the effectiveness of a range of transfer learning baselines. We find that transferring from entailment data is more effective than transferring from paraphrase or extractive QA data, and that it, surprisingly, continues to be very beneficial even when starting from massive pre-trained language models such as BERT. Our best method trains BERT on MultiNLI and then re-trains it on our train set. It achieves 80.4% accuracy compared to 90% accuracy of human annotators (and 62% majority-baseline), leaving a significant gap for future work."
                },
                {
                    "title": "Transformer-XL: Attentive Language Models beyond a Fixed-Length Context",
                    "abstract": "Transformers have a potential of learning longer-term dependency, but are limited by a fixed-length context in the setting of language modeling. We propose a novel neural architecture Transformer-XL that enables learning dependency beyond a fixed length without disrupting temporal coherence. It consists of a segment-level recurrence mechanism and a novel positional encoding scheme. Our method not only enables capturing longer-term dependency, but also resolves the context fragmentation problem. As a result, Transformer-XL learns dependency that is 80% longer than RNNs and 450% longer than vanilla Transformers, achieves better performance on both short and long sequences, and is up to 1,800+ times faster than vanilla Transformers during evaluation. Notably, we improve the state-of-the-art results of bpc/perplexity to 0.99 on enwiki8, 1.08 on text8, 18.3 on WikiText-103, 21.8 on One Billion Word, and 54.5 on Penn Treebank (without finetuning). When trained only on WikiText-103, Transformer-XL manages to generate reasonably coherent, novel text articles with thousands of tokens. Our code, pretrained models, and hyperparameters are available in both Tensorflow and PyTorch."
                },
                {
                    "title": "CoQA: A Conversational Question Answering Challenge",
                    "abstract": "Humans gather information through conversations involving a series of interconnected questions and answers. For machines to assist in information gathering, it is therefore essential to enable them to answer conversational questions. We introduce CoQA, a novel dataset for building Conversational Question Answering systems. Our dataset contains 127k questions with answers, obtained from 8k conversations about text passages from seven diverse domains. The questions are conversational, and the answers are free-form text with their corresponding evidence highlighted in the passage. We analyze CoQA in depth and show that conversational questions have challenging phenomena not present in existing reading comprehension datasets (e.g., coreference and pragmatic reasoning). We evaluate strong dialogue and reading comprehension models on CoQA. The best system obtains an F1 score of 65.4%, which is 23.4 points behind human performance (88.8%), indicating that there is ample room for improvement. We present CoQA as a challenge to the community at https://stanfordnlp.github.io/coqa."
                },
                {
                    "title": "Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering",
                    "abstract": "We present a new kind of question answering dataset, OpenBookQA, modeled after open book exams for assessing human understanding of a subject. The open book that comes with our questions is a set of 1326 elementary level science facts. Roughly 6000 questions probe an understanding of these facts and their application to novel situations. This requires combining an open book fact (e.g., metals conduct electricity) with broad common knowledge (e.g., a suit of armor is made of metal) obtained from other sources. While existing QA datasets over documents or knowledge bases, being generally self-contained, focus on linguistic understanding, OpenBookQA probes a deeper understanding of both the topic\u2014in the context of common knowledge\u2014and the language it is expressed in. Human performance on OpenBookQA is close to 92%, but many state-of-the-art pre-trained QA methods perform surprisingly poorly, worse than several simple neural baselines we develop. Our oracle experiments designed to circumvent the knowledge retrieval bottleneck demonstrate the value of both the open book and additional facts. We leave it as a challenge to solve the retrieval problem in this multi-hop setting and to close the large gap to human performance."
                },
                {
                    "title": "Know What You Don\u2019t Know: Unanswerable Questions for SQuAD",
                    "abstract": "Extractive reading comprehension systems can often locate the correct answer to a question in a context document, but they also tend to make unreliable guesses on questions for which the correct answer is not stated in the context. Existing datasets either focus exclusively on answerable questions, or use automatically generated unanswerable questions that are easy to identify. To address these weaknesses, we present SQuADRUn, a new dataset that combines the existing Stanford Question Answering Dataset (SQuAD) with over 50,000 unanswerable questions written adversarially by crowdworkers to look similar to answerable ones. To do well on SQuADRUn, systems must not only answer questions when possible, but also determine when no answer is supported by the paragraph and abstain from answering. SQuADRUn is a challenging natural language understanding task for existing models: a strong neural system that gets 86% F1 on SQuAD achieves only 66% F1 on SQuADRUn. We release SQuADRUn to the community as the successor to SQuAD."
                },
                {
                    "title": "GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding",
                    "abstract": "Human ability to understand language is general, flexible, and robust. In contrast, most NLU models above the word level are designed for a specific task and struggle with out-of-domain data. If we aspire to develop models with understanding beyond the detection of superficial correspondences between inputs and outputs, then it is critical to develop a unified model that can execute a range of linguistic tasks across different domains. To facilitate research in this direction, we present the General Language Understanding Evaluation (GLUE, gluebenchmark.com): a benchmark of nine diverse NLU tasks, an auxiliary dataset for probing models for understanding of specific linguistic phenomena, and an online platform for evaluating and comparing models. For some benchmark tasks, training data is plentiful, but for others it is limited or does not match the genre of the test set. GLUE thus favors models that can represent linguistic knowledge in a way that facilitates sample-efficient learning and effective knowledge-transfer across tasks. While none of the datasets in GLUE were created from scratch for the benchmark, four of them feature privately-held test data, which is used to ensure that the benchmark is used fairly. We evaluate baselines that use ELMo (Peters et al., 2018), a powerful transfer learning technique, as well as state-of-the-art sentence representation models. The best models still achieve fairly low absolute scores. Analysis with our diagnostic dataset yields similarly weak performance over all phenomena tested, with some exceptions."
                },
                {
                    "title": "Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge",
                    "abstract": "We present a new question set, text corpus, and baselines assembled to encourage AI research in advanced question answering. Together, these constitute the AI2 Reasoning Challenge (ARC), which requires far more powerful knowledge and reasoning than previous challenges such as SQuAD or SNLI. The ARC question set is partitioned into a Challenge Set and an Easy Set, where the Challenge Set contains only questions answered incorrectly by both a retrieval-based algorithm and a word co-occurence algorithm. The dataset contains only natural, grade-school science questions (authored for human tests), and is the largest public-domain set of this kind (7,787 questions). We test several baselines on the Challenge Set, including leading neural models from the SQuAD and SNLI tasks, and find that none are able to significantly outperform a random baseline, reflecting the difficult nature of this task. We are also releasing the ARC Corpus, a corpus of 14M science sentences relevant to the task, and implementations of the three neural baseline models tested. Can your model perform better? We pose ARC as a challenge to the community."
                },
                {
                    "title": "Strategies for Conceptual Change in Convolutional Neural Networks",
                    "abstract": "A remarkable feature of human beings is their capacity for creative behaviour, referring to their ability to react to problems in ways that are novel, surprising, and useful. Transformational creativity is a form of creativity where the creative behaviour is induced by a transformation of the actor's conceptual space, that is, the representational system with which the actor interprets its environment. In this report, we focus on ways of adapting systems of learned representations as they switch from performing one task to performing another. We describe an experimental comparison of multiple strategies for adaptation of learned features, and evaluate how effectively each of these strategies realizes the adaptation, in terms of the amount of training, and in terms of their ability to cope with restricted availability of training data. We show, among other things, that across handwritten digits, natural images, and classical music, adaptive strategies are systematically more effective than a baseline method that starts learning from scratch."
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "RACE: Large-scale ReAding Comprehension Dataset From Examinations",
                    "abstract": "We present RACE, a new dataset for benchmark evaluation of methods in the reading comprehension task. Collected from the English exams for middle and high school Chinese students in the age range between 12 to 18, RACE consists of near 28,000 passages and near 100,000 questions generated by human experts (English instructors), and covers a variety of topics which are carefully designed for evaluating the students\u2019 ability in understanding and reasoning. In particular, the proportion of questions that requires reasoning is much larger in RACE than that in other benchmark datasets for reading comprehension, and there is a significant gap between the performance of the state-of-the-art models (43%) and the ceiling human performance (95%). We hope this new dataset can serve as a valuable resource for research and evaluation in machine comprehension. The dataset is freely available at http://www.cs.cmu.edu/~glai1/data/race/ and the code is available at https://github.com/qizhex/RACE_AR_baselines."
                },
                {
                    "title": "SQuAD: 100,000+ Questions for Machine Comprehension of Text",
                    "abstract": "We present the Stanford Question Answering Dataset (SQuAD), a new reading comprehension dataset consisting of 100,000+ questions posed by crowdworkers on a set of Wikipedia articles, where the answer to each question is a segment of text from the corresponding reading passage. We analyze the dataset to understand the types of reasoning required to answer the questions, leaning heavily on dependency and constituency trees. We build a strong logistic regression model, which achieves an F1 score of 51.0%, a significant improvement over a simple baseline (20%). However, human performance (86.8%) is much higher, indicating that the dataset presents a good challenge problem for future research. \nThe dataset is freely available at this https URL"
                },
                {
                    "title": "Abstractive Text Summarization using Sequence-to-sequence RNNs and Beyond",
                    "abstract": "In this work, we model abstractive text summarization using Attentional Encoder-Decoder Recurrent Neural Networks, and show that they achieve state-of-the-art performance on two different corpora. We propose several novel models that address critical problems in summarization that are not adequately modeled by the basic architecture, such as modeling key-words, capturing the hierarchy of sentence-to-word structure, and emitting words that are rare or unseen at training time. Our work shows that many of our proposed models contribute to further improvement in performance. We also propose a new dataset consisting of multi-sentence summaries, and establish performance benchmarks for further research."
                },
                {
                    "title": "LLAMA: automatic hypertext generation utilizing language models",
                    "abstract": "Manual hypertext construction is labour intensive and prone to error. Robust systems capable of automatic hypertext generation (AHG) could be of direct benefit to those individuals responsible for hypertext authoring. In this paper we propose a novel technique for the autonomous creation of hypertext which is dependent upon language models. This work is strongly influenced by those algorithms which process the hyperlinked structure of a corpus in an attempt to find authoritative sources. The algorithm was evaluated by experimental comparison with human hypertext authors, and we found that both approaches produced broadly similar results."
                },
                {
                    "title": "An Adversarial Winograd Schema Challenge at Scale",
                    "abstract": "The Winograd Schema Challenge (WSC), proposed by Levesque et al. (2011) as an alternative to the Turing Test, was originally designed as a pronoun resolution problem that cannot be solved based on statistical patterns in large text corpora. However, recent studies suggest that current WSC datasets, even when composed carefully by experts, are still prone to such biases that statistical methods can exploit. We introduce WINOGRANDE, a new collection of WSC problems that are adversarially constructed to be robust against spurious statistical biases. While the original WSC dataset provided only 273 instances, WINOGRANDE includes 43,985 instances, half of which are determined as adversarial. Key to our approach is a novel adversarial filtering algorithm AFLITE for systematic bias reduction, combined with a careful crowdsourcing design. Despite the significant increase in training data, the performance of existing stateof-the-art methods remains modest (61.6%) and contrasts with high human performance (90.8%) for the binary questions. In addition, WINOGRANDE allows us to use transfer learning for achieving new state-of-the-art results on the original WSC and related datasets. Finally, we discuss how biases lead to overestimating the true capabilities of machine commonsense."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Improving Language Understanding by Generative Pre-Training",
                    "abstract": "Natural language understanding comprises a wide range of diverse tasks such as textual entailment, question answering, semantic similarity assessment, and document classi\ufb01cation. Although large unlabeled text corpora are abundant, labeled data for learning these speci\ufb01c tasks is scarce, making it challenging for discriminatively trained models to perform adequately. We demonstrate that large gains on these tasks can be realized by generative pre-training of a language model on a diverse corpus of unlabeled text, followed by discriminative \ufb01ne-tuning on each speci\ufb01c task. In contrast to previous approaches, we make use of task-aware input transformations during \ufb01ne-tuning to achieve effective transfer while requiring minimal changes to the model architecture. We demonstrate the effectiveness of our approach on a wide range of benchmarks for natural language understanding. Our general task-agnostic model outperforms discriminatively trained models that use architectures speci\ufb01cally crafted for each task, signi\ufb01cantly improving upon the state of the art in 9 out of the 12 tasks studied. For instance, we achieve absolute improvements of 8.9% on commonsense reasoning (Stories Cloze Test), 5.7% on question answering (RACE), and 1.5% on textual entailment (MultiNLI)."
                },
                {
                    "title": "Measuring Models",
                    "abstract": "Model-Driven Engineering (MDE) is an approach to software development that uses models as primary artifacts, from which code, documentation and tests are derived. One way of assessing quality assurance in a given domain is to de\ufb01ne domain metrics. We show that some of these metrics are supported by models. As text documents, models can be considered from a syntactic point of view i.e., thought of as graphs. We can readily apply graph-based metrics to them, such as the number of nodes, the number of edges or the fan-in/fan-out distributions. However, these metrics cannot leverage the semantic structuring enforced by each speci\ufb01c metamodel to give domain speci\ufb01c information. Contrary to graph-based metrics, more speci\ufb01c metrics do exist for given domains (such as LOC for programs), but they lack genericity. Our contribution is to propose one metric, called \u03c3 , that is generic over metamodels and allows the easy speci\ufb01cation of an open-ended wide range of model metrics."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the reasoning abilities of Large Language Models (LLMs) while still aligning them with human preferences, particularly in light of the potential drawbacks of Reinforcement Learning from Human Feedback (RLHF)?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the trade-off between aligning LLMs with human preferences and maintaining their reasoning capabilities. A successful approach could lead to more reliable and effective AI systems that better understand and respond to human needs, ultimately advancing the field of natural language processing. This could pave the way for practical applications in various domains, such as education, customer service, and content generation, where both accuracy and user satisfaction are paramount.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent conflict between optimizing for human preferences and preserving the model's reasoning abilities. Naive approaches that focus solely on aligning outputs with human feedback may inadvertently compromise the model's ability to reason effectively, leading to issues like \"hallucination\" or generating incorrect information. Technical obstacles include the need for sophisticated reward modeling that accurately reflects human preferences without sacrificing the model's cognitive capabilities. Theoretical complexities arise from understanding the interplay between different training methodologies and their impact on model performance.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on RLHF and its associated methodologies, which have shown limitations in preserving reasoning abilities, especially in smaller models. The lack of a comprehensive approach that balances alignment with reasoning has created a gap in the literature. Barriers include the difficulty in designing reward models that do not lead to overfitting on human preferences at the expense of logical reasoning. Our approach, Direct Preference Heads (DPH), differs by optimizing a reward score produced by the LLM itself rather than the logits from the language modeling head, allowing for a more nuanced evaluation of outputs.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves implementing Direct Preference Heads (DPH) in conjunction with an efficient 551M parameter language model. We will evaluate DPH on various commonsense reasoning and Natural Language Understanding (NLU) tasks, using metrics such as accuracy and user satisfaction to assess performance. The expected outcomes include improved reasoning capabilities in LLMs while maintaining alignment with human preferences, potentially leading to a new standard in language model training that balances these"
            }
        },
        "author_data": {
            "02a3fb48-695e-4d8d-acf6-cfab0ea3b61c": {
                "pk": "02a3fb48-695e-4d8d-acf6-cfab0ea3b61c",
                "name": "Avelina Asada Hadji-Kyriacou",
                "collaborators": [
                    "Ognjen Arandjelovic"
                ],
                "domain": [
                    "Transfer Learning",
                    "Multi-Modal",
                    "Parameter-Efficient Fine-Tuning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Context-PEFT: Efficient Multi-Modal, Multi-Task Fine-Tuning",
                        "abstract": "This paper introduces a novel Parameter-Efficient Fine-Tuning (PEFT) framework for multi-modal, multi-task transfer learning with pre-trained language models. PEFT techniques such as LoRA, BitFit and IA3 have demonstrated comparable performance to full fine-tuning of pre-trained models for specific downstream tasks, all while demanding significantly fewer trainable parameters and reduced GPU memory consumption. However, in the context of multi-modal fine-tuning, the need for architectural modifications or full fine-tuning often becomes apparent. To address this we propose Context-PEFT, which learns different groups of adaptor parameters based on the token's domain. This approach enables LoRA-like weight injection without requiring additional architectural changes. Our method is evaluated on the COCO captioning task, where it outperforms full fine-tuning under similar data constraints while simultaneously offering a substantially more parameter-efficient and computationally economical solution."
                    }
                ]
            },
            "9deb025d-da67-46a1-8079-83677aa996b4": {
                "pk": "9deb025d-da67-46a1-8079-83677aa996b4",
                "name": "Ognjen Arandjelovic",
                "collaborators": [
                    "Victor Andrei",
                    "Ieva Vasiljeva",
                    "Jessica Cooper",
                    "Neofytos Dimitriou"
                ],
                "domain": [
                    "Computer Vision",
                    "Machine Learning",
                    "Sports Science",
                    "Medical Research"
                ],
                "publications": [
                    {
                        "title": "A note on the capability profile - localized increase in force production and its effect on overall resistance training performance",
                        "abstract": "In this article I resolve formally the apparent paradox that arises in the context of the computational model of resistance exercise introduced in my previous work. Contrary to intuition, the model seems to allow for a localized increase in force production to degrade the ultimate exercise performance. I show this not to be the case."
                    },
                    {
                        "title": "Fairer citation-based metrics",
                        "abstract": "I describe a simple modification which can be applied to any citation count-based index (e.g. Hirsch's h-index) quantifying a researcher's publication output. The key idea behind the proposed approach is that the merit for the citations of a paper should be distributed amongst its authors according to their relative contributions. In addition to producing inherently fairer metrics I show that the proposed modification has the potential to partially normalize for the unfair effects of honorary authorship and thus discourage this practice."
                    },
                    {
                        "title": "Automatic vehicle tracking and recognition from aerial image sequences",
                        "abstract": "This paper addresses the problem of automated vehicle tracking and recognition from aerial image sequences. Motivated by its successes in the existing literature focus on the use of linear appearance subspaces to describe multi-view object appearance and highlight the challenges involved in their application as a part of a practical system. A working solution which includes steps for data extraction and normalization is described. In experiments on real-world data the proposed methodology achieved promising results with a high correct recognition rate and few, meaningful errors (type II errors whereby genuinely similar targets are sometimes being confused with one another). Directions for future research and possible improvements of the proposed method are discussed."
                    },
                    {
                        "title": "Discriminative k-means clustering",
                        "abstract": "The k-means algorithm is a partitional clustering method. Over 60 years old, it has been successfully used for a variety of problems. The popularity of k-means is in large part a consequence of its simplicity and efficiency. In this paper we are inspired by these appealing properties of k-means in the development of a clustering algorithm which accepts the notion of \"positively\" and \"negatively\" labelled data. The goal is to discover the cluster structure of both positive and negative data in a manner which allows for the discrimination between the two sets. The usefulness of this idea is demonstrated practically on the problem of face recognition, where the task of learning the scope of a person's appearance should be done in a manner which allows this face to be differentiated from others."
                    },
                    {
                        "title": "Matching objects across the textured-smooth continuum",
                        "abstract": "The problem of 3D object recognition is of immense practical importance, with the last decade witnessing a number of breakthroughs in the state of the art. Most of the previous work has focused on the matching of textured objects using local appearance descriptors extracted around salient image points. The recently proposed bag of boundaries method was the first to address directly the problem of matching smooth objects using boundary features. However, no previous work has attempted to achieve a holistic treatment of the problem by jointly using textural and shape features which is what we describe herein. Due to the complementarity of the two modalities, we fuse the corresponding matching scores and learn their relative weighting in a data specific manner by optimizing discriminative performance on synthetically distorted data. For the textural description of an object we adopt a representation in the form of a histogram of SIFT based visual words. Similarly the apparent shape of an object is represented by a histogram of discretized features capturing local shape. On a large public database of a diverse set of objects, the proposed method is shown to outperform significantly both purely textural and purely shape based approaches for matching across viewpoint variation."
                    },
                    {
                        "title": "Live-wire 3D medical images segmentation",
                        "abstract": "This report describes the design, implementation, evaluation and original enhancements to the Live-Wire method for 2D and 3D image segmentation. Live-Wire 2D employs a semi-automatic paradigm; the user is asked to select a few boundary points of the object to segment, to steer the process in the right direction, while the result is displayed in real time. In our implementation segmentation is extended to three dimensions by performing this process on a slice-by-slice basis. User's time and involvement is further reduced by allowing him to specify object contours in planes orthogonal to the slices. If these planes are chosen strategically, Live-Wire 3D can perform 2D segmentation in the plane of each slice automatically. This report also proposes two improvements to the original method, path heating and a new graph edge feature function based on variance of path properties along the boundary. We show that these improvements lead up to a 33% reduction in interaction with the user, and improved delineation in presence of strong interfering edges."
                    },
                    {
                        "title": "Computer simulation based parameter selection for resistance exercise",
                        "abstract": "In contrast to most scientific disciplines, sports science research has been characterized by comparatively little effort investment in the development of relevant phenomenological models. Scarcer yet is the application of said models in practice. We present a framework which allows resistance training practitioners to employ a recently proposed neuromuscular model in actual training program design. The first novelty concerns the monitoring aspect of coaching. A method for extracting training performance characteristics from loosely constrained video sequences, effortlessly and with minimal human input, using computer vision is described. The extracted data is subsequently used to fit the underlying neuromuscular model. This is achieved by solving an inverse dynamics problem corresponding to a particular exercise. Lastly, a computer simulation of hypothetical training bouts, using athlete-specific capability parameters, is used to predict the effected adaptation and changes in performance. The software described here allows the practitioner to manipulate hypothetical training parameters and immediately see their effect on predicted adaptation for a specific athlete. Thus, this work presents a holistic view of the monitoring-assessment-adjustment loop."
                    },
                    {
                        "title": "Hallucinating optimal high-dimensional subspaces",
                        "abstract": "Linear subspace representations of appearance variation are pervasive in computer vision. This paper addresses the problem of robustly matching such subspaces (computing the similarity between them) when they are used to describe the scope of variations within sets of images of different (possibly greatly so) scales. A naive solution of projecting the low-scale subspace into the high-scale image space is described first and subsequently shown to be inadequate, especially at large scale discrepancies. A successful approach is proposed instead. It consists of (i) an interpolated projection of the low-scale subspace into the high-scale space, which is followed by (ii) a rotation of this initial estimate within the bounds of the imposed ``downsampling constraint''. The optimal rotation is found in the closed-form which best aligns the high-scale reconstruction of the low-scale subspace with the reference it is compared to. The method is evaluated on the problem of matching sets of (i) face appearances under varying illumination and (ii) object appearances under varying viewpoint, using two large data sets. In comparison to the naive matching, the proposed algorithm is shown to greatly increase the separation of between-class and within-class similarities, as well as produce far more meaningful modes of common appearance on which the match score is based."
                    },
                    {
                        "title": "Sample-targeted clinical trial adaptation",
                        "abstract": "Clinical trial adaptation refers to any adjustment of the trial protocol after the onset of the trial. The main goal is to make the process of introducing new medical interventions to patients more efficient by reducing the cost and the time associated with evaluating their safety and efficacy. The principal question is how should adaptation be performed so as to minimize the chance of distorting the outcome of the trial. We propose a novel method for achieving this. Unlike previous work our approach focuses on trial adaptation by sample size adjustment. We adopt a recently proposed stratification framework based on collected auxiliary data and show that this information together with the primary measured variables can be used to make a probabilistically informed choice of the particular sub-group a sample should be removed from. Experiments on simulated data are used to illustrate the effectiveness of our method and its application in practice."
                    },
                    {
                        "title": "Automatic Face Recognition from Video",
                        "abstract": "The objective of this work is to automatically recognize faces from video sequences in a realistic, unconstrained setup in which illumination conditions are extreme and greatly changing, viewpoint and user motion pattern have a wide variability, and video input is of low quality. At the centre of focus are face appearance manifolds: this thesis presents a significant advance of their understanding and application in the sphere of face recognition. The two main contributions are the Generic Shape-Illumination Manifold recognition algorithm and the Anisotropic Manifold Space clustering. The Generic Shape-Illumination Manifold is evaluated on a large data corpus acquired in real-world conditions and its performance is shown to greatly exceed that of state-of-the-art methods in the literature and the best performing commercial software. Empirical evaluation of the Anisotropic Manifold Space clustering on a popular situation comedy is also described with excellent preliminary results."
                    },
                    {
                        "title": "Learnt quasi-transitive similarity for retrieval from large collections of faces",
                        "abstract": "We are interested in identity-based retrieval of face sets from large unlabelled collections acquired in uncontrolled environments. Given a baseline algorithm for measuring the similarity of two face sets, the meta-algorithm introduced in this paper seeks to leverage the structure of the data corpus to make the best use of the available baseline. In particular, we show how partial transitivity of inter-personal similarity can be exploited to improve the retrieval of particularly challenging sets which poorly match the query under the baseline measure. We: (i) describe the use of proxy sets as a means of computing the similarity between two sets, (ii) introduce transitivity meta-features based on the similarity of salient modes of appearance variation between sets, (iii) show how quasi-transitivity can be learnt from such features without any labelling or manual intervention, and (iv) demonstrate the effectiveness of the proposed methodology through experiments on the notoriously challenging YouTube database and two successful baselines from the literature."
                    },
                    {
                        "title": "Discriminative extended canonical correlation analysis for pattern set matching",
                        "abstract": "In this paper we address the problem of matching sets of vectors embedded in the same input space. We propose an approach which is motivated by canonical correlation analysis (CCA), a statistical technique which has proven successful in a wide variety of pattern recognition problems. Like CCA when applied to the matching of sets, our extended canonical correlation analysis (E-CCA) aims to extract the most similar modes of variability within two sets. Our first major contribution is the formulation of a principled framework for robust inference of such modes from data in the presence of uncertainty associated with noise and sampling randomness. E-CCA retains the efficiency and closed form computability of CCA, but unlike it, does not possess free parameters which cannot be inferred directly from data (inherent data dimensionality, and the number of canonical correlations used for set similarity computation). Our second major contribution is to show that in contrast to CCA, E-CCA is readily adapted to match sets in a discriminative learning scheme which we call discriminative extended canonical correlation analysis (DE-CCA). Theoretical contributions of this paper are followed by an empirical evaluation of its premises on the task of face recognition from sets of rasterized appearance images. The results demonstrate that our approach, E-CCA, already outperforms both CCA and its quasi-discriminative counterpart constrained CCA (C-CCA), for all values of their free parameters. An even greater improvement is achieved with the discriminative variant, DE-CCA."
                    },
                    {
                        "title": "COVID-19 and Science Communication: The Recording and Reporting of Disease Mortality",
                        "abstract": "The ongoing COVID-19 pandemic has brought science to the fore of the public discourse and considering the complexity of the issues involved, with it also the challenge of effective and informative science communication. A particularly contentious topic, in that it is both highly emotional in and of itself, as well as in that it sits at the nexus of the decision-making process regarding the handling of the pandemic, which has effected lockdowns, social behaviour measures, business closures, and others, concerns the recording and the reporting of the disease mortality. To clarify a point which has caused much controversy and anger in the public debate, the first part of the present article discusses the very fundamentals underlying the issue of causative attribution with regards to mortality, lays out the foundations of the statistical means of mortality estimation, and concretizes these by analysing the recording and reporting practices adopted in England, and their widespread misrepresentations. The second part of the article is empirical in nature. I present data and an analysis of how COVID-19 mortality has been reported in the mainstream media in the UK and the USA, including a comparative analysis both across the two countries as well as across different media outlets. The findings clearly demonstrate a uniform and worrying lack of understanding of the relevant technical subject matter by the media in both countries. Of particular interest is the finding that with a remarkable regularity ($\\rho>0.998$) the greater the number of articles a media outlet published on COVID-19 mortality, the greater the \\emph{proportion} of its articles misrepresented the disease mortality figures."
                    },
                    {
                        "title": "A framework for improving the performance of verification algorithms with a low false positive rate requirement and limited training data",
                        "abstract": "In this paper we address the problem of matching patterns in the so-called verification setting in which a novel, query pattern is verified against a single training pattern: the decision sought is whether the two match (i.e. belong to the same class) or not. Unlike previous work which has universally focused on the development of more discriminative distance functions between patterns, here we consider the equally important and pervasive task of selecting a distance threshold which fits a particular operational requirement - specifically, the target false positive rate (FPR). First, we argue on theoretical grounds that a data-driven approach is inherently ill-conditioned when the desired FPR is low, because by the very nature of the challenge only a small portion of training data affects or is affected by the desired threshold. This leads us to propose a general, statistical model-based method instead. Our approach is based on the interpretation of an inter-pattern distance as implicitly defining a pattern embedding which approximately distributes patterns according to an isotropic multi-variate normal distribution in some space. This interpretation is then used to show that the distribution of training inter-pattern distances is the non-central chi2 distribution, differently parameterized for each class. Thus, to make the class-specific threshold choice we propose a novel analysis-by-synthesis iterative algorithm which estimates the three free parameters of the model (for each class) using task-specific constraints. The validity of the premises of our work and the effectiveness of the proposed method are demonstrated by applying the method to the task of set-based face verification on a large database of pseudo-random head motion videos."
                    },
                    {
                        "title": "Identification of promising research directions using machine learning aided medical literature analysis",
                        "abstract": "The rapidly expanding corpus of medical research literature presents major challenges in the understanding of previous work, the extraction of maximum information from collected data, and the identification of promising research directions. We present a case for the use of advanced machine learning techniques as an aide in this task and introduce a novel methodology that is shown to be capable of extracting meaningful information from large longitudinal corpora, and of tracking complex temporal changes within it."
                    },
                    {
                        "title": "Prediction of future hospital admissions - what is the tradeoff between specificity and accuracy?",
                        "abstract": "Large amounts of electronic medical records collected by hospitals across the developed world offer unprecedented possibilities for knowledge discovery using computer based data mining and machine learning. Notwithstanding significant research efforts, the use of this data in the prediction of disease development has largely been disappointing. In this paper we examine in detail a recently proposed method which has in preliminary experiments demonstrated highly promising results on real-world data. We scrutinize the authors' claims that the proposed model is scalable and investigate whether the tradeoff between prediction specificity (i.e. the ability of the model to predict a wide number of different ailments) and accuracy (i.e. the ability of the model to make the correct prediction) is practically viable. Our experiments conducted on a data corpus of nearly 3,000,000 admissions support the authors' expectations and demonstrate that the high prediction accuracy is maintained well even when the number of admission types explicitly included in the model is increased to account for 98% of all admissions in the corpus. Thus several promising directions for future work are highlighted."
                    },
                    {
                        "title": "Understanding Ancient Coin Images",
                        "abstract": "In recent years, a range of problems within the broad umbrella of automatic, computer vision based analysis of ancient coins has been attracting an increasing amount of attention. Notwithstanding this research effort, the results achieved by the state of the art in the published literature remain poor and far from sufficiently well performing for any practical purpose. In the present paper we present a series of contributions which we believe will benefit the interested community. Firstly, we explain that the approach of visual matching of coins, universally adopted in all existing published papers on the topic, is not of practical interest because the number of ancient coin types exceeds by far the number of those types which have been imaged, be it in digital form (e.g. online) or otherwise (traditional film, in print, etc.). Rather, we argue that the focus should be on the understanding of the semantic content of coins. Hence, we describe a novel method which uses real-world multimodal input to extract and associate semantic concepts with the correct coin images and then using a novel convolutional neural network learn the appearance of these concepts. Empirical evidence on a real-world and by far the largest data set of ancient coins, we demonstrate highly promising results."
                    },
                    {
                        "title": "A New Look at Ghost Normalization",
                        "abstract": "Batch normalization (BatchNorm) is an effective yet poorly understood technique for neural network optimization. It is often assumed that the degradation in BatchNorm performance to smaller batch sizes stems from it having to estimate layer statistics using smaller sample sizes. However, recently, Ghost normalization (GhostNorm), a variant of BatchNorm that explicitly uses smaller sample sizes for normalization, has been shown to improve upon BatchNorm in some datasets. Our contributions are: (i) we uncover a source of regularization that is unique to GhostNorm, and not simply an extension from BatchNorm, (ii) three types of GhostNorm implementations are described, two of which employ BatchNorm as the underlying normalization technique, (iii) by visualising the loss landscape of GhostNorm, we observe that GhostNorm consistently decreases the smoothness when compared to BatchNorm, (iv) we introduce Sequential Normalization (SeqNorm), and report superior performance over state-of-the-art methodologies on both CIFAR--10 and CIFAR--100 datasets."
                    }
                ]
            }
        }
    },
    "2311.16054": {
        "paper_data": {
            "title": "Metric Space Magnitude for Evaluating the Diversity of Latent Representations",
            "url": "http://arxiv.org/abs/2311.16054v3",
            "arxiv_id": "2311.16054",
            "authors": [
                "Katharina Limbeck",
                "Rayna Andreeva",
                "Rik Sarkar",
                "Bastian Rieck"
            ],
            "abstract": "The magnitude of a metric space is a novel invariant that provides a measure of the 'effective size' of a space across multiple scales, while also capturing numerous geometrical properties, such as curvature, density, or entropy. We develop a family of magnitude-based measures of the intrinsic diversity of latent representations, formalising a novel notion of dissimilarity between magnitude functions of finite metric spaces. Our measures are provably stable under perturbations of the data, can be efficiently calculated, and enable a rigorous multi-scale characterisation and comparison of latent representations. We show their utility and superior performance across different domains and tasks, including (i) the automated estimation of diversity, (ii) the detection of mode collapse, and (iii) the evaluation of generative models for text, image, and graph data.",
            "introduction": "   1 Introduction  Diversity is a key concept in representation learning, referring to the relative abundance and distinctiveness of model outputs. Given the inherent complexity of deep learning models, the evaluation of diversity is thus crucial, enabling   (i) the assessment of the intrinsic richness of latent representations, and  (ii) the evaluation of the extent to which models are capable of preserving the properties of an input distribution.   While the quantitative evaluation of generative models in particular relies on assessing trade-offs between fidelity and diversity with regards to a known reference distribution, reference-free diversity measures are becoming increasingly relevant when a ground-truth distribution is unknown or intractable. However, reference-based diversity metrics such as recall are notoriously fallible, sensitive to parameter choices and therefore prone to incorrectly approximate the true data manifold, whereas reference-free diversity measures often rely on simple mean summaries that fail to pass basic consistency checks\u00a0[13]. Thus, existing methods lack expressivity to fully capture what it means for a space to be diverse, resulting in a critical need for novel measures that are   (i) theoretically motivated,  (ii) robust to noise, and  (iii) capable of encoding the intrinsic diversity of data across varying levels of similarity rather than at a single fixed threshold.     Our contributions. Addressing this need, we propose a novel family of diversity measures based on metric space magnitude, a mathematical invariant that captures numerous important multi-scale geometric characteristics of metric spaces, including curvature, density, and entropy of an input space. Metric space magnitude merely requires a notion of dissimilarity between data points, permitting it to operate on both local and global scales. Hence, magnitude is poised to compare latent spaces, yielding a compact holistic summary of diversity that satisfies relevant theoretical requirements. Our work is the first to   (i) introduce magnitude as a general tool for evaluating the diversity of latent representations, and  (ii) formalise a notion of difference between the magnitude of two spaces across multiple scales of similarity.   We demonstrate that magnitude is stable and can detect curvature, highlighting its use as a multi-scale summary of the local and global geometry of data. Moreover, we empirically showcase the utility of our magnitude-based diversity measure across different modalities, namely text, image, and graph embeddings, for which we observe that our measure outperforms alternative embedding-based measures of intrinsic diversity. Finally, when a reference distribution is known, our magnitude-based notion of difference reliably detects mode collapse and mode dropping, thus assisting practitioners in model evaluation and selection.     2 Related Work  Latent representations and embeddings have become indispensable tools for analysing data types such as images, text, and graphs. As evidenced by LLMs, understanding semantic relationships in data requires meaningful embeddings. Our work focuses on improving representation-based diversity evaluation and we thus consider the role diversity plays in this context.   Diversity measures.  Assessing generative model diversity remains a challenge irrespective of the domain\u00a0[39], as ground truth reference distributions or labelled data are often unavailable, and human evaluation remains costly. Thus, there exists a need for interpretable, automated and unsupervised measures of intrinsic diversity. Reference-free evaluation is of particular importance for assessing generated text given the black-box-nature of LLMs [6], but also applicable across modalities. Motivated by this, a varied collection of diversity measures has been proposed, many of which are task-, domain- or model-specific\u00a0[13]; only a fraction of them are applicable to",
            "references": [
                {
                    "title": "Mapping the Multiverse of Latent Representations",
                    "abstract": "Echoing recent calls to counter reliability and robustness concerns in machine learning via multiverse analysis, we present PRESTO, a principled framework for mapping the multiverse of machine-learning models that rely on latent representations. Although such models enjoy widespread adoption, the variability in their embeddings remains poorly understood, resulting in unnecessary complexity and untrustworthy representations. Our framework uses persistent homology to characterize the latent spaces arising from different combinations of diverse machine-learning methods, (hyper)parameter configurations, and datasets, allowing us to measure their pairwise (dis)similarity and statistically reason about their distributions. As we demonstrate both theoretically and empirically, our pipeline preserves desirable properties of collections of latent representations, and it can be leveraged to perform sensitivity analysis, detect anomalous embeddings, or efficiently and effectively navigate hyperparameter search spaces."
                },
                {
                    "title": "The small-scale limit of magnitude and the one-point property",
                    "abstract": "The magnitude of a metric space is a real-valued function whose parameter controls the scale of the metric. A metric space is said to have the one-point property if its magnitude converges to 1 as the space is scaled down to a point. Not every finite metric space has the one-point property: to date, exactly one example has been found of a finite space for which the property fails. Understanding the failure of the one-point property is of interest in clarifying the interpretation of magnitude and its stability with respect to the Gromov--Hausdorff topology. We prove that the one-point property holds generically for finite metric spaces, but that when it fails, the failure can be arbitrarily bad: the small-scale limit of magnitude can take arbitrary real values greater than 1."
                },
                {
                    "title": "Metric Space Magnitude and Generalisation in Neural Networks",
                    "abstract": "Deep learning models have seen significant successes in numerous applications, but their inner workings remain elusive. The purpose of this work is to quantify the learning process of deep neural networks through the lens of a novel topological invariant called magnitude. Magnitude is an isometry invariant; its properties are an active area of research as it encodes many known invariants of a metric space. We use magnitude to study the internal representations of neural networks and propose a new method for determining their generalisation capabilities. Moreover, we theoretically connect magnitude dimension and the generalisation error, and demonstrate experimentally that the proposed framework can be a good indicator of the latter."
                },
                {
                    "title": "The Vendi Score: A Diversity Evaluation Metric for Machine Learning",
                    "abstract": "Diversity is an important criterion for many areas of machine learning (ML), including generative modeling and dataset curation. However, existing metrics for measuring diversity are often domain-specific and limited in flexibility. In this paper, we address the diversity evaluation problem by proposing the Vendi Score, which connects and extends ideas from ecology and quantum statistical mechanics to ML. The Vendi Score is defined as the exponential of the Shannon entropy of the eigenvalues of a similarity matrix. This matrix is induced by a user-defined similarity function applied to the sample to be evaluated for diversity. In taking a similarity function as input, the Vendi Score enables its user to specify any desired form of diversity. Importantly, unlike many existing metrics in ML, the Vendi Score does not require a reference dataset or distribution over samples or labels, it is therefore general and applicable to any generative model, decoding algorithm, and dataset from any domain where similarity can be defined. We showcase the Vendi Score on molecular generative modeling where we found it addresses shortcomings of the current diversity metric of choice in that domain. We also applied the Vendi Score to generative models of images and decoding algorithms of text where we found it confirms known results about diversity in those domains. Furthermore, we used the Vendi Score to measure mode collapse, a known shortcoming of generative adversarial networks (GANs). In particular, the Vendi Score revealed that even GANs that capture all the modes of a labeled dataset can be less diverse than the original dataset. Finally, the interpretability of the Vendi Score allowed us to diagnose several benchmark ML datasets for diversity, opening the door for diversity-informed data augmentation."
                },
                {
                    "title": "On the effectiveness of persistent homology",
                    "abstract": "Persistent homology (PH) is one of the most popular methods in Topological Data Analysis. Even though PH has been used in many different types of applications, the reasons behind its success remain elusive; in particular, it is not known for which classes of problems it is most effective, or to what extent it can detect geometric or topological features. The goal of this work is to identify some types of problems where PH performs well or even better than other methods in data analysis. We consider three fundamental shape analysis tasks: the detection of the number of holes, curvature and convexity from 2D and 3D point clouds sampled from shapes. Experiments demonstrate that PH is successful in these tasks, outperforming several baselines, including PointNet, an architecture inspired precisely by the properties of point clouds. In addition, we observe that PH remains effective for limited computational resources and limited training data, as well as out-of-distribution test data, including various data transformations and noise. For convexity detection, we provide a theoretical guarantee that PH is effective for this task in $\\mathbb{R}^d$, and demonstrate the detection of a convexity measure on the FLAVIA data set of plant leaf images. Due to the crucial role of shape classification in understanding mathematical and physical structures and objects, and in many applications, the findings of this work will provide some knowledge about the types of problems that are appropriate for PH, so that it can - to borrow the words from Wigner 1960 - ``remain valid in future research, and extend, to our pleasure\", but to our lesser bafflement, to a variety of applications."
                },
                {
                    "title": "Parallel black-box optimization of expensive high-dimensional multimodal functions via magnitude",
                    "abstract": "Building on the recently developed theory of magnitude, we introduce the optimization algorithm EXPLO2 and carefully benchmark it. EXPLO2 advances the state of the art for optimizing high-dimensional ($D \\gtrapprox 40$) multimodal functions that are expensive to compute and for which derivatives are not available, such as arise in hyperparameter optimization or via simulations."
                },
                {
                    "title": "On Evaluation Metrics for Graph Generative Models",
                    "abstract": "In image generation, generative models can be evaluated naturally by visually inspecting model outputs. However, this is not always the case for graph generative models (GGMs), making their evaluation challenging. Currently, the standard process for evaluating GGMs suffers from three critical limitations: i) it does not produce a single score which makes model selection challenging, ii) in many cases it fails to consider underlying edge and node features, and iii) it is prohibitively slow to perform. In this work, we mitigate these issues by searching for scalar, domain-agnostic, and scalable metrics for evaluating and ranking GGMs. To this end, we study existing GGM metrics and neural-network-based metrics emerging from generative models of images that use embeddings extracted from a task-specific network. Motivated by the power of certain Graph Neural Networks (GNNs) to extract meaningful graph representations without any training, we introduce several metrics based on the features extracted by an untrained random GNN. We design experiments to thoroughly test metrics on their ability to measure the diversity and fidelity of generated graphs, as well as their sample and computational efficiency. Depending on the quantity of samples, we recommend one of two random-GNN-based metrics that we show to be more expressive than pre-existing metrics. While we focus on applying these metrics to GGM evaluation, in practice this enables the ability to easily compute the dissimilarity between any two sets of graphs regardless of domain. Our code is released at: https://github.com/uoguelph-mlrg/GGM-metrics."
                },
                {
                    "title": "Evaluation Metrics for Graph Generative Models: Problems, Pitfalls, and Practical Solutions",
                    "abstract": "Graph generative models are a highly active branch of machine learning. Given the steady development of new models of ever-increasing complexity, it is necessary to provide a principled way to evaluate and compare them. In this paper, we enumerate the desirable criteria for such a comparison metric and provide an overview of the status quo of graph generative model comparison in use today, which predominantly relies on the maximum mean discrepancy (MMD). We perform a systematic evaluation of MMD in the context of graph generative model comparison, highlighting some of the challenges and pitfalls researchers inadvertently may encounter. After conducting a thorough analysis of the behaviour of MMD on synthetically-generated perturbed graphs as well as on recently-proposed graph generative models, we are able to provide a suitable procedure to mitigate these challenges and pitfalls. We aggregate our findings into a list of practical recommendations for researchers to use when evaluating graph generative models."
                },
                {
                    "title": "Entropy and Diversity: The Axiomatic Approach",
                    "abstract": "The global biodiversity crisis is one of humanity's most urgent problems, but even quantifying biological diversity is a difficult mathematical and conceptual challenge. This book brings new mathematical rigour to the ongoing debate. It was born of research in category theory, is given strength by information theory, and is fed by the ancient field of functional equations. It applies the power of the axiomatic method to a biological problem of pressing concern, but it also presents new theorems that stand up as mathematics in their own right, independently of any application. The question 'what is diversity?' has surprising mathematical depth, and this book covers a wide breadth of mathematics, from functional equations to geometric measure theory, from probability theory to number theory. Despite this range, the mathematical prerequisites are few: the main narrative thread of this book requires no more than an undergraduate course in analysis."
                },
                {
                    "title": "Weighting vectors for machine learning: numerical harmonic analysis applied to boundary detection",
                    "abstract": "Metric space magnitude, an active field of research in algebraic topology, is a scalar quantity that summarizes the effective number of distinct points that live in a general metric space. The {\\em weighting vector} is a closely-related concept that captures, in a nontrivial way, much of the underlying geometry of the original metric space. Recent work has demonstrated that when the metric space is Euclidean, the weighting vector serves as an effective tool for boundary detection. We recast this result and show the weighting vector may be viewed as a solution to a kernelized SVM. As one consequence, we apply this new insight to the task of outlier detection, and we demonstrate performance that is competitive or exceeds performance of state-of-the-art techniques on benchmark data sets. Under mild assumptions, we show the weighting vector, which has computational cost of matrix inversion, can be efficiently approximated in linear time. We show how nearest neighbor methods can approximate solutions to the minimization problems defined by SVMs."
                },
                {
                    "title": "Scalable Deep Generative Modeling for Sparse Graphs",
                    "abstract": "Learning graph generative models is a challenging task for deep learning and has wide applicability to a range of domains like chemistry, biology and social science. However current deep neural methods suffer from limited scalability: for a graph with $n$ nodes and $m$ edges, existing deep neural methods require $\\Omega(n^2)$ complexity by building up the adjacency matrix. On the other hand, many real world graphs are actually sparse in the sense that $m\\ll n^2$. Based on this, we develop a novel autoregressive model, named BiGG, that utilizes this sparsity to avoid generating the full adjacency matrix, and importantly reduces the graph generation time complexity to $O((n + m)\\log n)$. Furthermore, during training this autoregressive model can be parallelized with $O(\\log n)$ synchronization stages, which makes it much more efficient than other autoregressive models that require $\\Omega(n)$. Experiments on several benchmarks show that the proposed approach not only scales to orders of magnitude larger graphs than previously possible with deep autoregressive graph generative models, but also yields better graph generation quality."
                },
                {
                    "title": "Evaluation of Text Generation: A Survey",
                    "abstract": "The paper surveys evaluation methods of natural language generation (NLG) systems that have been developed in the last few years. We group NLG evaluation methods into three categories: (1) human-centric evaluation metrics, (2) automatic metrics that require no training, and (3) machine-learned metrics. For each category, we discuss the progress that has been made and the challenges still being faced, with a focus on the evaluation of recently proposed NLG tasks and neural NLG models. We then present two case studies of automatic text summarization and long text generation, and conclude the paper by proposing future research directions."
                },
                {
                    "title": "Evaluating the Evaluation of Diversity in Natural Language Generation",
                    "abstract": "Despite growing interest in natural language generation (NLG) models that produce diverse outputs, there is currently no principled method for evaluating the diversity of an NLG system. In this work, we propose a framework for evaluating diversity metrics. The framework measures the correlation between a proposed diversity metric and a diversity parameter, a single parameter that controls some aspect of diversity in generated text. For example, a diversity parameter might be a binary variable used to instruct crowdsourcing workers to generate text with either low or high content diversity. We demonstrate the utility of our framework by: (a) establishing best practices for eliciting diversity judgments from humans, (b) showing that humans substantially outperform automatic metrics in estimating content diversity, and (c) demonstrating that existing methods for controlling diversity by tuning a \u201cdecoding parameter\u201d mostly affect form but not meaning. Our framework can advance the understanding of different diversity metrics, an essential step on the road towards better NLG systems."
                },
                {
                    "title": "Diversity, Density, and Homogeneity: Quantitative Characteristic Metrics for Text Collections",
                    "abstract": "Summarizing data samples by quantitative measures has a long history, with descriptive statistics being a case in point. However, as natural language processing methods flourish, there are still insufficient characteristic metrics to describe a collection of texts in terms of the words, sentences, or paragraphs they comprise. In this work, we propose metrics of diversity, density, and homogeneity that quantitatively measure the dispersion, sparsity, and uniformity of a text collection. We conduct a series of simulations to verify that each metric holds desired properties and resonates with human intuitions. Experiments on real-world datasets demonstrate that the proposed characteristic metrics are highly correlated with text classification performance of a renowned model, BERT, which could inspire future applications."
                },
                {
                    "title": "Reliable Fidelity and Diversity Metrics for Generative Models",
                    "abstract": "Devising indicative evaluation metrics for the image generation task remains an open problem. The most widely used metric for measuring the similarity between real and generated images has been the Frechet Inception Distance (FID) score. Because it does not differentiate the fidelity and diversity aspects of the generated images, recent papers have introduced variants of precision and recall metrics to diagnose those properties separately. In this paper, we show that even the latest version of the precision and recall metrics are not reliable yet. For example, they fail to detect the match between two identical distributions, they are not robust against outliers, and the evaluation hyperparameters are selected arbitrarily. We propose density and coverage metrics that solve the above issues. We analytically and experimentally show that density and coverage provide more interpretable and reliable signals for practitioners than the existing metrics. Code: this https URL."
                },
                {
                    "title": "Efficient Graph Generation with Graph Recurrent Attention Networks",
                    "abstract": "We propose a new family of efficient and expressive deep generative models of graphs, called Graph Recurrent Attention Networks (GRANs). Our model generates graphs one block of nodes and associated edges at a time. The block size and sampling stride allow us to trade off sample quality for efficiency. Compared to previous RNN-based graph generative models, our framework better captures the auto-regressive conditioning between the already-generated and to-be-generated parts of the graph using Graph Neural Networks (GNNs) with attention. This not only reduces the dependency on node ordering but also bypasses the long-term bottleneck caused by the sequential nature of RNNs. Moreover, we parameterize the output distribution per block using a mixture of Bernoulli, which captures the correlations among generated edges within the block. Finally, we propose to handle node orderings in generation by marginalizing over a family of canonical orderings. On standard benchmarks, we achieve state-of-the-art time efficiency and sample quality compared to previous models. Additionally, we show our model is capable of generating large graphs of up to 5K nodes with good quality. Our code is released at: \\url{https://github.com/lrjconan/GRAN}."
                },
                {
                    "title": "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
                    "abstract": "BERT (Devlin et al., 2018) and RoBERTa (Liu et al., 2019) has set a new state-of-the-art performance on sentence-pair regression tasks like semantic textual similarity (STS). However, it requires that both sentences are fed into the network, which causes a massive computational overhead: Finding the most similar pair in a collection of 10,000 sentences requires about 50 million inference computations (~65 hours) with BERT. The construction of BERT makes it unsuitable for semantic similarity search as well as for unsupervised tasks like clustering. In this publication, we present Sentence-BERT (SBERT), a modification of the pretrained BERT network that use siamese and triplet network structures to derive semantically meaningful sentence embeddings that can be compared using cosine-similarity. This reduces the effort for finding the most similar pair from 65 hours with BERT / RoBERTa to about 5 seconds with SBERT, while maintaining the accuracy from BERT. We evaluate SBERT and SRoBERTa on common STS tasks and transfer learning tasks, where it outperforms other state-of-the-art sentence embeddings methods."
                },
                {
                    "title": "Boosting Dialog Response Generation",
                    "abstract": "Neural models have become one of the most important approaches to dialog response generation. However, they still tend to generate the most common and generic responses in the corpus all the time. To address this problem, we designed an iterative training process and ensemble method based on boosting. We combined our method with different training and decoding paradigms as the base model, including mutual-information-based decoding and reward-augmented maximum likelihood learning. Empirical results show that our approach can significantly improve the diversity and relevance of the responses generated by all base models, backed by objective measurements and human evaluation."
                },
                {
                    "title": "Persistent homology detects curvature",
                    "abstract": "In topological data analysis, persistent homology is used to study the \u2018shape of data\u2019. Persistent homology computations are completely characterized by a set of intervals called a bar code. It is often said that the long intervals represent the \u2018topological signal\u2019 and the short intervals represent \u2018noise\u2019. We give evidence to dispute this thesis, showing that the short intervals encode geometric information. Specifically, we prove that persistent homology detects the curvature of disks from which points have been sampled. We describe a general computational framework for solving inverse problems using the average persistence landscape, a continuous mapping from metric spaces with a probability measure to a Hilbert space. In the present application, the average persistence landscapes of points sampled from disks of constant curvature results in a path in this Hilbert space which may be learned using standard tools from statistical and machine learning."
                },
                {
                    "title": "Improved Precision and Recall Metric for Assessing Generative Models",
                    "abstract": "The ability to automatically estimate the quality and coverage of the samples produced by a generative model is a vital requirement for driving algorithm research. We present an evaluation metric that can separately and reliably measure both of these aspects in image generation tasks by forming explicit, non-parametric representations of the manifolds of real and generated data. We demonstrate the effectiveness of our metric in StyleGAN and BigGAN by providing several illustrative examples where existing metrics yield uninformative or contradictory results. Furthermore, we analyze multiple design variants of StyleGAN to better understand the relationships between the model architecture, training methods, and the properties of the resulting sample distribution. In the process, we identify new variants that improve the state-of-the-art. We also perform the first principled analysis of truncation methods and identify an improved method. Finally, we extend our metric to estimate the perceptual quality of individual samples, and use this to study latent space interpolations."
                },
                {
                    "title": "How Powerful are Graph Neural Networks?",
                    "abstract": "Graph Neural Networks (GNNs) are an effective framework for representation learning of graphs. GNNs follow a neighborhood aggregation scheme, where the representation vector of a node is computed by recursively aggregating and transforming representation vectors of its neighboring nodes. Many GNN variants have been proposed and have achieved state-of-the-art results on both node and graph classification tasks. However, despite GNNs revolutionizing graph representation learning, there is limited understanding of their representational properties and limitations. Here, we present a theoretical framework for analyzing the expressive power of GNNs to capture different graph structures. Our results characterize the discriminative power of popular GNN variants, such as Graph Convolutional Networks and GraphSAGE, and show that they cannot learn to distinguish certain simple graph structures. We then develop a simple architecture that is provably the most expressive among the class of GNNs and is as powerful as the Weisfeiler-Lehman graph isomorphism test. We empirically validate our theoretical findings on a number of graph classification benchmarks, and demonstrate that our model achieves state-of-the-art performance."
                },
                {
                    "title": "Ecological Diversity: Measuring the Unmeasurable",
                    "abstract": "Diversity is a concept central to ecology, and its measurement is essential for any study of ecosystem health. But summarizing this complex and multidimensional concept in a single measure is problematic. Dozens of mathematical indices have been proposed for this purpose, but these can provide contradictory results leading to misleading or incorrect conclusions about a community\u2019s diversity. In this review, we summarize the key conceptual issues underlying the measurement of ecological diversity, survey the indices most commonly used in ecology, and discuss their relative suitability. We advocate for indices that: (i) satisfy key mathematical axioms; (ii) can be expressed as so-called effective numbers; (iii) can be extended to account for disparity between types; (iv) can be parameterized to obtain diversity profiles; and (v) for which an estimator (preferably unbiased) can be found so that the index is useful for practical applications."
                },
                {
                    "title": "Assessing Generative Models via Precision and Recall",
                    "abstract": "Recent advances in generative modeling have led to an increased interest in the study of statistical divergences as means of model comparison. Commonly used evaluation methods, such as the Frechet Inception Distance (FID), correlate well with the perceived quality of samples and are sensitive to mode dropping. However, these metrics are unable to distinguish between different failure cases since they only yield one-dimensional scores. We propose a novel definition of precision and recall for distributions which disentangles the divergence into two separate dimensions. The proposed notion is intuitive, retains desirable properties, and naturally leads to an efficient algorithm that can be used to evaluate generative models. We relate this notion to total variation as well as to recent evaluation metrics such as Inception Score and FID. To demonstrate the practical utility of the proposed approach we perform an empirical study on several variants of Generative Adversarial Networks and Variational Autoencoders. In an extensive set of experiments we show that the proposed metric is able to disentangle the quality of generated samples from the coverage of the target distribution."
                },
                {
                    "title": "GraphRNN: Generating Realistic Graphs with Deep Auto-regressive Models",
                    "abstract": "Modeling and generating graphs is fundamental for studying networks in biology, engineering, and social sciences. However, modeling complex distributions over graphs and then efficiently sampling from these distributions is challenging due to the non-unique, high-dimensional nature of graphs and the complex, non-local dependencies that exist between edges in a given graph. Here we propose GraphRNN, a deep autoregressive model that addresses the above challenges and approximates any distribution of graphs with minimal assumptions about their structure. GraphRNN learns to generate graphs by training on a representative set of graphs and decomposes the graph generation process into a sequence of node and edge formations, conditioned on the graph structure generated so far. \nIn order to quantitatively evaluate the performance of GraphRNN, we introduce a benchmark suite of datasets, baselines and novel evaluation metrics based on Maximum Mean Discrepancy, which measure distances between sets of graphs. Our experiments show that GraphRNN significantly outperforms all baselines, learning to generate diverse graphs that match the structural characteristics of a target set, while also scaling to graphs 50 times larger than previous deep models."
                },
                {
                    "title": "Magnitude homology of enriched categories and metric spaces",
                    "abstract": "Magnitude is a numerical invariant of enriched categories, including in particular metric spaces as $[0,\\infty)$-enriched categories. We show that in many cases magnitude can be categorified to a homology theory for enriched categories, which we call magnitude homology (in fact, it is a special sort of Hochschild homology), whose graded Euler characteristic is the magnitude. Magnitude homology of metric spaces generalizes the Hepworth--Willerton magnitude homology of graphs, and detects geometric information such as convexity."
                },
                {
                    "title": "Rethinking the Inception Architecture for Computer Vision",
                    "abstract": "Convolutional networks are at the core of most state of-the-art computer vision solutions for a wide variety of tasks. Since 2014 very deep convolutional networks started to become mainstream, yielding substantial gains in various benchmarks. Although increased model size and computational cost tend to translate to immediate quality gains for most tasks (as long as enough labeled data is provided for training), computational efficiency and low parameter count are still enabling factors for various use cases such as mobile vision and big-data scenarios. Here we are exploring ways to scale up networks in ways that aim at utilizing the added computation as efficiently as possible by suitably factorized convolutions and aggressive regularization. We benchmark our methods on the ILSVRC 2012 classification challenge validation set demonstrate substantial gains over the state of the art: 21:2% top-1 and 5:6% top-5 error for single frame evaluation using a network with a computational cost of 5 billion multiply-adds per inference and with using less than 25 million parameters. With an ensemble of 4 models and multi-crop evaluation, we report 3:5% top-5 error and 17:3% top-1 error on the validation set and 3:6% top-5 error on the official test set."
                },
                {
                    "title": "A note on the evaluation of generative models",
                    "abstract": "Probabilistic generative models can be used for compression, denoising, inpainting, texture synthesis, semi-supervised learning, unsupervised feature learning, and other tasks. Given this wide range of applications, it is not surprising that a lot of heterogeneity exists in the way these models are formulated, trained, and evaluated. As a consequence, direct comparison between models is often difficult. This article reviews mostly known but often underappreciated properties relating to the evaluation and interpretation of generative models with a focus on image models. In particular, we show that three of the currently most commonly used criteria---average log-likelihood, Parzen window estimates, and visual fidelity of samples---are largely independent of each other when the data is high-dimensional. Good performance with respect to one criterion therefore need not imply good performance with respect to the other criteria. Our results show that extrapolation from one criterion to another is not warranted and generative models need to be evaluated directly with respect to the application(s) they were intended for. In addition, we provide examples demonstrating that Parzen window estimates should generally be avoided."
                },
                {
                    "title": "Geodesic exponential kernels: When curvature and linearity conflict",
                    "abstract": "We consider kernel methods on general geodesic metric spaces and provide both negative and positive results. First we show that the common Gaussian kernel can only be generalized to a positive definite kernel on a geodesic metric space if the space is flat. As a result, for data on a Riemannian manifold, the geodesic Gaussian kernel is only positive definite if the Riemannian manifold is Euclidean. This implies that any attempt to design geodesic Gaussian kernels on curved Riemannian manifolds is futile. However, we show that for spaces with conditionally negative definite distances the geodesic Laplacian kernel can be generalized while retaining positive definiteness. This implies that geodesic Laplacian kernels can be generalized to some curved spaces, including spheres and hyperbolic spaces. Our theoretical results are verified empirically."
                },
                {
                    "title": "The magnitude of metric spaces",
                    "abstract": "Magnitude is a real-valued invariant of metric spaces, analogous to the Euler characteristic of topological spaces and the cardinality of sets. The definition of magnitude is a special case of a general categorical definition that clarifies the analogies between various cardinality-like invariants in mathematics. Although this motivation is a world away from geometric measure, magnitude, when applied to subsets of R^n, turns out to be intimately related to invariants such as volume, surface area, perimeter and dimension. We describe several aspects of this relationship, providing evidence for a conjecture (first stated in arXiv:0908.1582) that magnitude subsumes all the most important invariants of classical integral geometry."
                },
                {
                    "title": "Collective Classification in Network Data",
                    "abstract": "Many real-world applications produce networked data such as the world-wide web (hypertext documents connected via hyperlinks), social networks (for example, people connected by friendship links), communication networks (computers connected via communication links) and biological networks (for example, protein interaction networks). A recent focus in machine learning research has been to extend traditional machine learning classification techniques to classify nodes in such networks. In this article, we provide a brief introduction to this area of research and how it has progressed during the past decade. We introduce four of the most widely used inference algorithms for classifying networked data and empirically compare them on both synthetic and real-world data."
                },
                {
                    "title": "A Kernel Method for the Two-Sample-Problem",
                    "abstract": "We propose two statistical tests to determine if two samples are from different distributions. Our test statistic is in both cases the distance between the means of the two samples mapped into a reproducing kernel Hilbert space (RKHS). The first test is based on a large deviation bound for the test statistic, while the second is based on the asymptotic distribution of this statistic. The test statistic can be computed in O(m2) time. We apply our approach to a variety of problems, including attribute matching for databases using the Hungarian marriage method, where our test performs strongly. We also demonstrate excellent performance when comparing distributions over graphs, for which no alternative tests currently exist."
                },
                {
                    "title": "Algorithm 748: enclosing zeros of continuous functions",
                    "abstract": "Two efficient algorithms for enclosing a zero of a continuous function are presented. They are similar to the recent methods, but together with quadratic interpolation they make essential use of inverse cubic interpolation as well. Since asymptotically the inverse cubic interpolation is always chosen by the algorithms, they achieve higher-efficiency indices: 1.6529\u2026 for the first algorithm, and 1.6686\u2026 for the second one. It is proved that the second algorithm is optimal in a certain family. Numerical experiments show that the two new methods compare well with recent methods, as well as with the efficient solvers of Dekker, Brent, Bus and Dekker, and Le. The second method from the present article has the best behavior of all 12 methods especially when the termination tolerance is small."
                },
                {
                    "title": "Cholesky factorization",
                    "abstract": "This article aimed at a general audience of computational scientists, surveys the Cholesky factorization for symmetric positive definite matrices, covering algorithms for computing it, the numerical stability of the algorithms, and updating and downdating of the factorization. Cholesky factorization with pivoting for semidefinite matrices is also treated. Copyright \u00a9 2009 John Wiley & Sons, Inc."
                },
                {
                    "title": "On the evolution of random graphs",
                    "abstract": "(n) k edges have equal probabilities to be chosen as the next one . We shall 2 study the \"evolution\" of such a random graph if N is increased . In this investigation we endeavour to find what is the \"typical\" structure at a given stage of evolution (i . e . if N is equal, or asymptotically equal, to a given function N(n) of n) . By a \"typical\" structure we mean such a structure the probability of which tends to 1 if n -* + when N = N(n) . If A is such a property that lim Pn,N,(n ) ( A) = 1, we shall say that \u201ealmost all\" graphs Gn,N(n) n--possess this property ."
                }
            ],
            "categories": [
                "cs.LG",
                "math.GT",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop robust, theoretically motivated, and reference-free diversity measures for evaluating the intrinsic diversity of latent representations in deep learning models?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of existing diversity measures, which often rely on reference distributions that may be unavailable or intractable. By advancing the understanding of diversity in representation learning, this research could lead to more effective generative models and improved evaluation techniques. The implications extend to various applications, including natural language processing, computer vision, and graph analysis, where better diversity measures can enhance model performance and reliability.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of deep learning models and the inherent difficulty in quantifying diversity without a reference distribution. Naive approaches may fail because they often rely on simplistic mean summaries that do not capture the multi-scale geometric characteristics of data. Additionally, existing reference-based metrics are sensitive to parameter choices and can misrepresent the true data manifold, making it difficult to achieve a comprehensive understanding of diversity across varying levels of similarity.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on reference-based diversity metrics, which are limited by their reliance on known distributions that may not exist in practice. Additionally, many existing measures are task- or model-specific, lacking generalizability. Barriers such as the complexity of accurately capturing multi-scale diversity and the absence of a unified theoretical framework have hindered progress. Our approach differs by introducing metric space magnitude as a versatile tool for evaluating diversity, which can operate on both local and global scales without requiring a reference distribution.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the use of metric space magnitude to evaluate diversity in latent representations. We will apply this measure across various datasets, including text, image, and graph embeddings, and utilize metrics such as curvature and density to assess diversity. The expected outcomes include a robust, multi-scale summary of diversity that outperforms existing embedding-based measures, as well as the ability to detect mode collapse and mode dropping when a reference distribution is known, thereby enhancing model evaluation and selection processes."
            }
        },
        "author_data": {
            "e3b87f6d-44fe-4e5f-97f2-647d719c1b70": {
                "pk": "e3b87f6d-44fe-4e5f-97f2-647d719c1b70",
                "name": "Katharina Limbeck",
                "collaborators": [
                    "Bastian Rieck",
                    "Rayna Andreeva",
                    "Rik Sarkar"
                ],
                "domain": [
                    "Topological Data Analysis",
                    "Deep Learning",
                    "Spatial Data Analysis",
                    "Generalization in Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Detecting Spatial Dependence in Transcriptomics Data using Vectorised Persistence Diagrams",
                        "abstract": "Evaluating spatial patterns in data is an integral task across various domains, including geostatistics, astronomy, and spatial tissue biology. The analysis of transcriptomics data in particular relies on methods for detecting spatially-dependent features that exhibit significant spatial patterns for both explanatory analysis and feature selection. However, given the complex and high-dimensional nature of these data, there is a need for robust, stable, and reliable descriptors of spatial dependence. We leverage the stability and multiscale properties of persistent homology to address this task. To this end, we introduce a novel framework using functional topological summaries, such as Betti curves and persistence landscapes, for identifying and describing non-random patterns in spatial data. In particular, we propose a non-parametric one-sample permutation test for spatial dependence and investigate its utility across both simulated and real spatial omics data. Our vectorised approach outperforms baseline methods at accurately detecting spatial dependence. Further, we find that our method is more robust to outliers than alternative tests using Moran's I."
                    },
                    {
                        "title": "Metric Space Magnitude and Generalisation in Neural Networks",
                        "abstract": "Deep learning models have seen significant successes in numerous applications, but their inner workings remain elusive. The purpose of this work is to quantify the learning process of deep neural networks through the lens of a novel topological invariant called magnitude. Magnitude is an isometry invariant; its properties are an active area of research as it encodes many known invariants of a metric space. We use magnitude to study the internal representations of neural networks and propose a new method for determining their generalisation capabilities. Moreover, we theoretically connect magnitude dimension and the generalisation error, and demonstrate experimentally that the proposed framework can be a good indicator of the latter."
                    }
                ]
            },
            "9ec8574b-603f-46f1-bc8a-20136d67e299": {
                "pk": "9ec8574b-603f-46f1-bc8a-20136d67e299",
                "name": "Rayna Andreeva",
                "collaborators": [
                    "Rik Sarkar",
                    "Lauren Watson",
                    "Hao-Tsung Yang",
                    "Primoz Skraba",
                    "Anwesha Sarkar",
                    "James Ward",
                    "Jie Gao",
                    "Benjamin Dupuis",
                    "Tolga Birdal",
                    "Umut \u015eim\u015fekli"
                ],
                "domain": [
                    "Machine Learning",
                    "Topological Data Analysis",
                    "Computational Geometry",
                    "Data Valuation"
                ],
                "publications": [
                    {
                        "title": "Machine learning and Topological data analysis identify unique features of human papillae in 3D scans",
                        "abstract": "The tongue surface houses a range of papillae that are integral to the mechanics and chemistry of taste and textural sensation. Although gustatory function of papillae is well investigated, the uniqueness of papillae within and across individuals remains elusive. Here, we present the first machine learning framework on 3D microscopic scans of human papillae (n = 2092), uncovering the uniqueness of geometric and topological features of papillae. The finer differences in shapes of papillae are investigated computationally based on a number of features derived from discrete differential geometry and computational topology. Interpretable machine learning techniques show that persistent homology features of the papillae shape are the most effective in predicting the biological variables. Models trained on these features with small volumes of data samples predict the type of papillae with an accuracy of 85%. The papillae type classification models can map the spatial arrangement of filiform and fungiform papillae on a surface. Remarkably, the papillae are found to be distinctive across individuals and an individual can be identified with an accuracy of 48% among the 15 participants from a single papillae. Collectively, this is the first unprecedented evidence demonstrating that tongue papillae can serve as a unique identifier inspiring new research direction for food preferences and oral diagnostics."
                    },
                    {
                        "title": "Differentially Private Shapley Values for Data Evaluation",
                        "abstract": "The Shapley value has been proposed as a solution to many applications in machine learning, including for equitable valuation of data. Shapley values are computationally expensive and involve the entire dataset. The query for a point's Shapley value can also compromise the statistical privacy of other data points. We observe that in machine learning problems such as empirical risk minimization, and in many learning algorithms (such as those with uniform stability), a diminishing returns property holds, where marginal benefit per data point decreases rapidly with data sample size. Based on this property, we propose a new stratified approximation method called the Layered Shapley Algorithm. We prove that this method operates on small (O(\\polylog(n))) random samples of data and small sized ($O(\\log n)$) coalitions to achieve the results with guaranteed probabilistic accuracy, and can be modified to incorporate differential privacy. Experimental results show that the algorithm correctly identifies high-value data points that improve validation accuracy, and that the differentially private evaluations preserve approximate ranking of data."
                    },
                    {
                        "title": "Approximating Metric Magnitude of Point Sets",
                        "abstract": "Metric magnitude is a measure of the \"size\" of point clouds with many desirable geometric properties. It has been adapted to various mathematical contexts and recent work suggests that it can enhance machine learning and optimization algorithms. But its usability is limited due to the computational cost when the dataset is large or when the computation must be carried out repeatedly (e.g. in model training). In this paper, we study the magnitude computation problem, and show efficient ways of approximating it. We show that it can be cast as a convex optimization problem, but not as a submodular optimization. The paper describes two new algorithms - an iterative approximation algorithm that converges fast and is accurate, and a subset selection method that makes the computation even faster. It has been previously proposed that magnitude of model sequences generated during stochastic gradient descent is correlated to generalization gap. Extension of this result using our more scalable algorithms shows that longer sequences in fact bear higher correlations. We also describe new applications of magnitude in machine learning - as an effective regularizer for neural network training, and as a novel clustering criterion."
                    },
                    {
                        "title": "Topological Generalization Bounds for Discrete-Time Stochastic Optimization Algorithms",
                        "abstract": "We present a novel set of rigorous and computationally efficient topology-based complexity notions that exhibit a strong correlation with the generalization gap in modern deep neural networks (DNNs). DNNs show remarkable generalization properties, yet the source of these capabilities remains elusive, defying the established statistical learning theory. Recent studies have revealed that properties of training trajectories can be indicative of generalization. Building on this insight, state-of-the-art methods have leveraged the topology of these trajectories, particularly their fractal dimension, to quantify generalization. Most existing works compute this quantity by assuming continuous- or infinite-time training dynamics, complicating the development of practical estimators capable of accurately predicting generalization without access to test data. In this paper, we respect the discrete-time nature of training trajectories and investigate the underlying topological quantities that can be amenable to topological data analysis tools. This leads to a new family of reliable topological complexity measures that provably bound the generalization error, eliminating the need for restrictive geometric assumptions. These measures are computationally friendly, enabling us to propose simple yet effective algorithms for computing generalization indices. Moreover, our flexible framework can be extended to different domains, tasks, and architectures. Our experimental results demonstrate that our new complexity measures correlate highly with generalization error in industry-standards architectures such as transformers and deep graph networks. Our approach consistently outperforms existing topological bounds across a wide range of datasets, models, and optimizers, highlighting the practical relevance and effectiveness of our complexity measures."
                    },
                    {
                        "title": "Metric Space Magnitude and Generalisation in Neural Networks",
                        "abstract": "Deep learning models have seen significant successes in numerous applications, but their inner workings remain elusive. The purpose of this work is to quantify the learning process of deep neural networks through the lens of a novel topological invariant called magnitude. Magnitude is an isometry invariant; its properties are an active area of research as it encodes many known invariants of a metric space. We use magnitude to study the internal representations of neural networks and propose a new method for determining their generalisation capabilities. Moreover, we theoretically connect magnitude dimension and the generalisation error, and demonstrate experimentally that the proposed framework can be a good indicator of the latter."
                    },
                    {
                        "title": "Accelerated Shapley Value Approximation for Data Evaluation",
                        "abstract": "Data valuation has found various applications in machine learning, such as data filtering, efficient learning and incentives for data sharing. The most popular current approach to data valuation is the Shapley value. While popular for its various applications, Shapley value is computationally expensive even to approximate, as it requires repeated iterations of training models on different subsets of data. In this paper we show that the Shapley value of data points can be approximated more efficiently by leveraging the structural properties of machine learning problems. We derive convergence guarantees on the accuracy of the approximate Shapley value for different learning settings including Stochastic Gradient Descent with convex and non-convex loss functions. Our analysis suggests that in fact models trained on small subsets are more important in the context of data valuation. Based on this idea, we describe $\\delta$-Shapley -- a strategy of only using small subsets for the approximation. Experiments show that this approach preserves approximate value and rank of data, while achieving speedup of up to 9.9x. In pre-trained networks the approach is found to bring more efficiency in terms of accurate evaluation using small subsets."
                    },
                    {
                        "title": "Euler Characteristic Surfaces",
                        "abstract": "We study the use of the Euler characteristic for multiparameter topological data analysis. Euler characteristic is a classical, well-understood topological invariant that has appeared in numerous applications, including in the context of random fields. The goal of this paper is to present the extension of using the Euler characteristic in higher-dimensional parameter spaces. While topological data analysis of higher-dimensional parameter spaces using stronger invariants such as homology continues to be the subject of intense research, Euler characteristic is more manageable theoretically and computationally, and this analysis can be seen as an important intermediary step in multi-parameter topological data analysis. We show the usefulness of the techniques using artificially generated examples, and a real-world application of detecting diabetic retinopathy in retinal images."
                    }
                ]
            },
            "aac497fd-d3bd-465c-9388-180bcaed1028": {
                "pk": "aac497fd-d3bd-465c-9388-180bcaed1028",
                "name": "Rik Sarkar",
                "collaborators": [
                    "Benedek Rozemberczki",
                    "Rayna Andreeva",
                    "Lauren Watson",
                    "Oliver Kiss",
                    "Abhirup Ghosh",
                    "Baharan Mirzasoleiman",
                    "Jie Gao",
                    "Anwesha Sarkar",
                    "Carl Allen",
                    "Ryan Davies"
                ],
                "domain": [
                    "Graph Mining",
                    "Machine Learning",
                    "Differential Privacy",
                    "Network Embedding"
                ],
                "publications": [
                    {
                        "title": "Twitch Gamers: a Dataset for Evaluating Proximity Preserving and Structural Role-based Node Embeddings",
                        "abstract": "Proximity preserving and structural role-based node embeddings have become a prime workhorse of applied graph mining. Novel node embedding techniques are often tested on a restricted set of benchmark datasets. In this paper, we propose a new diverse social network dataset called Twitch Gamers with multiple potential target attributes. Our analysis of the social network and node classification experiments illustrate that Twitch Gamers is suitable for assessing the predictive performance of novel proximity preserving and structural role-based node embedding algorithms."
                    },
                    {
                        "title": "Fast Sequence-Based Embedding with Diffusion Graphs",
                        "abstract": "A graph embedding is a representation of graph vertices in a low-dimensional space, which approximately preserves properties such as distances between nodes. Vertex sequence-based embedding procedures use features extracted from linear sequences of nodes to create embeddings using a neural network. In this paper, we propose diffusion graphs as a method to rapidly generate vertex sequences for network embedding. Its computational efficiency is superior to previous methods due to simpler sequence generation, and it produces more accurate results. In experiments, we found that the performance relative to other methods improves with increasing edge density in the graph. In a community detection task, clustering nodes in the embedding space produces better results compared to other sequence-based embedding methods."
                    },
                    {
                        "title": "Characteristic Functions on Graphs: Birds of a Feather, from Statistical Descriptors to Parametric Models",
                        "abstract": "In this paper, we propose a flexible notion of characteristic functions defined on graph vertices to describe the distribution of vertex features at multiple scales. We introduce FEATHER, a computationally efficient algorithm to calculate a specific variant of these characteristic functions where the probability weights of the characteristic function are defined as the transition probabilities of random walks. We argue that features extracted by this procedure are useful for node level machine learning tasks. We discuss the pooling of these node representations, resulting in compact descriptors of graphs that can serve as features for graph classification algorithms. We analytically prove that FEATHER describes isomorphic graphs with the same representation and exhibits robustness to data corruption. Using the node feature characteristic functions we define parametric models where evaluation points of the functions are learned parameters of supervised classifiers. Experiments on real world large datasets show that our proposed algorithm creates high quality representations, performs transfer learning efficiently, exhibits robustness to hyperparameter changes, and scales linearly with the input size."
                    },
                    {
                        "title": "The Shapley Value of Classifiers in Ensemble Games",
                        "abstract": "What is the value of an individual model in an ensemble of binary classifiers? We answer this question by introducing a class of transferable utility cooperative games called \\textit{ensemble games}. In machine learning ensembles, pre-trained models cooperate to make classification decisions. To quantify the importance of models in these ensemble games, we define \\textit{Troupe} -- an efficient algorithm which allocates payoffs based on approximate Shapley values of the classifiers. We argue that the Shapley value of models in these games is an effective decision metric for choosing a high performing subset of models from the ensemble. Our analytical findings prove that our Shapley value estimation scheme is precise and scalable; its performance increases with size of the dataset and ensemble. Empirical results on real world graph classification tasks demonstrate that our algorithm produces high quality estimates of the Shapley value. We find that Shapley values can be utilized for ensemble pruning, and that adversarial models receive a low valuation. Complex classifiers are frequently found to be responsible for both correct and incorrect classification decisions."
                    },
                    {
                        "title": "Machine learning and Topological data analysis identify unique features of human papillae in 3D scans",
                        "abstract": "The tongue surface houses a range of papillae that are integral to the mechanics and chemistry of taste and textural sensation. Although gustatory function of papillae is well investigated, the uniqueness of papillae within and across individuals remains elusive. Here, we present the first machine learning framework on 3D microscopic scans of human papillae (n = 2092), uncovering the uniqueness of geometric and topological features of papillae. The finer differences in shapes of papillae are investigated computationally based on a number of features derived from discrete differential geometry and computational topology. Interpretable machine learning techniques show that persistent homology features of the papillae shape are the most effective in predicting the biological variables. Models trained on these features with small volumes of data samples predict the type of papillae with an accuracy of 85%. The papillae type classification models can map the spatial arrangement of filiform and fungiform papillae on a surface. Remarkably, the papillae are found to be distinctive across individuals and an individual can be identified with an accuracy of 48% among the 15 participants from a single papillae. Collectively, this is the first unprecedented evidence demonstrating that tongue papillae can serve as a unique identifier inspiring new research direction for food preferences and oral diagnostics."
                    },
                    {
                        "title": "Little Ball of Fur: A Python Library for Graph Sampling",
                        "abstract": "Sampling graphs is an important task in data mining. In this paper, we describe Little Ball of Fur a Python library that includes more than twenty graph sampling algorithms. Our goal is to make node, edge, and exploration-based network sampling techniques accessible to a large number of professionals, researchers, and students in a single streamlined framework. We created this framework with a focus on a coherent application public interface which has a convenient design, generic input data requirements, and reasonable baseline settings of algorithms. Here we overview these design foundations of the framework in detail with illustrative code snippets. We show the practical usability of the library by estimating various global statistics of social networks and web graphs. Experiments demonstrate that Little Ball of Fur can speed up node and whole graph embedding techniques considerably with mildly deteriorating the predictive value of distilled features."
                    },
                    {
                        "title": "Stability Enhanced Privacy and Applications in Private Stochastic Gradient Descent",
                        "abstract": "Private machine learning involves addition of noise while training, resulting in lower accuracy. Intuitively, greater stability can imply greater privacy and improve this privacy-utility tradeoff. We study this role of stability in private empirical risk minimization, where differential privacy is achieved by output perturbation, and establish a corresponding theoretical result showing that for strongly-convex loss functions, an algorithm with uniform stability of $\\beta$ implies a bound of $O(\\sqrt{\\beta})$ on the scale of noise required for differential privacy.   The result applies to both explicit regularization and to implicitly stabilized ERM, such as adaptations of Stochastic Gradient Descent that are known to be stable. Thus, it generalizes recent results that improve privacy through modifications to SGD, and establishes stability as the unifying perspective. It implies new privacy guarantees for optimizations with uniform stability guarantees, where a corresponding differential privacy guarantee was previously not known. Experimental results validate the utility of stability enhanced privacy in several problems, including application of elastic nets and feature selection."
                    },
                    {
                        "title": "Karate Club: An API Oriented Open-source Python Framework for Unsupervised Learning on Graphs",
                        "abstract": "We present Karate Club a Python framework combining more than 30 state-of-the-art graph mining algorithms which can solve unsupervised machine learning tasks. The primary goal of the package is to make community detection, node and whole graph embedding available to a wide audience of machine learning researchers and practitioners. We designed Karate Club with an emphasis on a consistent application interface, scalability, ease of use, sensible out of the box model behaviour, standardized dataset ingestion, and output generation. This paper discusses the design principles behind this framework with practical examples. We show Karate Club's efficiency with respect to learning performance on a wide range of real world clustering problems, classification tasks and support evidence with regards to its competitive speed."
                    },
                    {
                        "title": "Continual and Sliding Window Release for Private Empirical Risk Minimization",
                        "abstract": "It is difficult to continually update private machine learning models with new data while maintaining privacy. Data incur increasing privacy loss -- as measured by differential privacy -- when they are used in repeated computations. In this paper, we describe regularized empirical risk minimization algorithms that continually release models for a recent window of data. One version of the algorithm uses the entire data history to improve the model for the recent window. The second version uses a sliding window of constant size to improve the model, ensuring more relevant models in case of evolving data. The algorithms operate in the framework of stochastic gradient descent. We prove that even with releasing a model at each time-step over an infinite time horizon, the privacy cost of any data point is bounded by a constant $\\epsilon$ differential privacy, and the accuracy of the output models are close to optimal. Experiments on MNIST and Arxiv publications data show results consistent with the theory."
                    },
                    {
                        "title": "Multi-scale Attributed Node Embedding",
                        "abstract": "We present network embedding algorithms that capture information about a node from the local distribution over node attributes around it, as observed over random walks following an approach similar to Skip-gram. Observations from neighborhoods of different sizes are either pooled (AE) or encoded distinctly in a multi-scale approach (MUSAE). Capturing attribute-neighborhood relationships over multiple scales is useful for a diverse range of applications, including latent feature identification across disconnected networks with similar attributes. We prove theoretically that matrices of node-feature pointwise mutual information are implicitly factorized by the embeddings. Experiments show that our algorithms are robust, computationally efficient and outperform comparable models on social networks and web graphs."
                    },
                    {
                        "title": "GEMSEC: Graph Embedding with Self Clustering",
                        "abstract": "Modern graph embedding procedures can efficiently process graphs with millions of nodes. In this paper, we propose GEMSEC -- a graph embedding algorithm which learns a clustering of the nodes simultaneously with computing their embedding. GEMSEC is a general extension of earlier work in the domain of sequence-based graph embedding. GEMSEC places nodes in an abstract feature space where the vertex features minimize the negative log-likelihood of preserving sampled vertex neighborhoods, and it incorporates known social network properties through a machine learning regularization. We present two new social network datasets and show that by simultaneously considering the embedding and clustering problems with respect to social properties, GEMSEC extracts high-quality clusters competitive with or superior to other community detection algorithms. In experiments, the method is found to be computationally efficient and robust to the choice of hyperparameters."
                    },
                    {
                        "title": "Distributed Submodular Maximization",
                        "abstract": "Many large-scale machine learning problems--clustering, non-parametric learning, kernel machines, etc.--require selecting a small yet representative subset from a large dataset. Such problems can often be reduced to maximizing a submodular set function subject to various constraints. Classical approaches to submodular optimization require centralized access to the full dataset, which is impractical for truly large-scale problems. In this paper, we consider the problem of submodular function maximization in a distributed fashion. We develop a simple, two-stage protocol GreeDi, that is easily implemented using MapReduce style computations. We theoretically analyze our approach, and show that under certain natural conditions, performance close to the centralized approach can be achieved. We begin with monotone submodular maximization subject to a cardinality constraint, and then extend this approach to obtain approximation guarantees for (not necessarily monotone) submodular maximization subject to more general constraints including matroid or knapsack constraints. In our extensive experiments, we demonstrate the effectiveness of our approach on several applications, including sparse Gaussian process inference and exemplar based clustering on tens of millions of examples using Hadoop."
                    },
                    {
                        "title": "Heterogeneous Interventions Reduce the Spread of COVID-19 in Simulations on Real Mobility Data",
                        "abstract": "Major interventions have been introduced worldwide to slow down the spread of the SARS-CoV-2 virus. Large-scale lockdowns of human movements are effective in reducing the spread, but they come at a cost of significantly limited societal functions. We show that natural human movements are statistically diverse, and the spread of the disease is significantly influenced by a small group of active individuals and gathering venues. We find that interventions focused on these most mobile individuals and popular venues reduce both the peak infection rate and the total infected population while retaining high social activity levels. These trends are seen consistently in simulations with real human mobility data of different scales, resolutions, and modalities from multiple cities across the world. The observation implies that compared to broad sweeping interventions, more heterogeneous strategies that are targeted based on the network effects in human mobility provide a better balance between pandemic control and regular social activities."
                    },
                    {
                        "title": "Differentially Private Shapley Values for Data Evaluation",
                        "abstract": "The Shapley value has been proposed as a solution to many applications in machine learning, including for equitable valuation of data. Shapley values are computationally expensive and involve the entire dataset. The query for a point's Shapley value can also compromise the statistical privacy of other data points. We observe that in machine learning problems such as empirical risk minimization, and in many learning algorithms (such as those with uniform stability), a diminishing returns property holds, where marginal benefit per data point decreases rapidly with data sample size. Based on this property, we propose a new stratified approximation method called the Layered Shapley Algorithm. We prove that this method operates on small (O(\\polylog(n))) random samples of data and small sized ($O(\\log n)$) coalitions to achieve the results with guaranteed probabilistic accuracy, and can be modified to incorporate differential privacy. Experimental results show that the algorithm correctly identifies high-value data points that improve validation accuracy, and that the differentially private evaluations preserve approximate ranking of data."
                    },
                    {
                        "title": "Metric Space Magnitude and Generalisation in Neural Networks",
                        "abstract": "Deep learning models have seen significant successes in numerous applications, but their inner workings remain elusive. The purpose of this work is to quantify the learning process of deep neural networks through the lens of a novel topological invariant called magnitude. Magnitude is an isometry invariant; its properties are an active area of research as it encodes many known invariants of a metric space. We use magnitude to study the internal representations of neural networks and propose a new method for determining their generalisation capabilities. Moreover, we theoretically connect magnitude dimension and the generalisation error, and demonstrate experimentally that the proposed framework can be a good indicator of the latter."
                    },
                    {
                        "title": "Inference and Interference: The Role of Clipping, Pruning and Loss Landscapes in Differentially Private Stochastic Gradient Descent",
                        "abstract": "Differentially private stochastic gradient descent (DP-SGD) is known to have poorer training and test performance on large neural networks, compared to ordinary stochastic gradient descent (SGD). In this paper, we perform a detailed study and comparison of the two processes and unveil several new insights. By comparing the behavior of the two processes separately in early and late epochs, we find that while DP-SGD makes slower progress in early stages, it is the behavior in the later stages that determines the end result. This separate analysis of the clipping and noise addition steps of DP-SGD shows that while noise introduces errors to the process, gradient descent can recover from these errors when it is not clipped, and clipping appears to have a larger impact than noise. These effects are amplified in higher dimensions (large neural networks), where the loss basin occupies a lower dimensional space. We argue theoretically and using extensive experiments that magnitude pruning can be a suitable dimension reduction technique in this regard, and find that heavy pruning can improve the test accuracy of DPSGD."
                    },
                    {
                        "title": "Approximating Metric Magnitude of Point Sets",
                        "abstract": "Metric magnitude is a measure of the \"size\" of point clouds with many desirable geometric properties. It has been adapted to various mathematical contexts and recent work suggests that it can enhance machine learning and optimization algorithms. But its usability is limited due to the computational cost when the dataset is large or when the computation must be carried out repeatedly (e.g. in model training). In this paper, we study the magnitude computation problem, and show efficient ways of approximating it. We show that it can be cast as a convex optimization problem, but not as a submodular optimization. The paper describes two new algorithms - an iterative approximation algorithm that converges fast and is accurate, and a subset selection method that makes the computation even faster. It has been previously proposed that magnitude of model sequences generated during stochastic gradient descent is correlated to generalization gap. Extension of this result using our more scalable algorithms shows that longer sequences in fact bear higher correlations. We also describe new applications of magnitude in machine learning - as an effective regularizer for neural network training, and as a novel clustering criterion."
                    }
                ]
            },
            "58560e89-db0e-4bde-ac46-d78fa115c5fc": {
                "pk": "58560e89-db0e-4bde-ac46-d78fa115c5fc",
                "name": "Bastian Rieck",
                "collaborators": [
                    "Corinna Coupette",
                    "Filip Sadlo",
                    "Heike Leitte",
                    "Celia Hacker",
                    "Jeremy Wayland",
                    "Julius von Rohrscheidt",
                    "Barbara Giunti",
                    "J\u0101nis Lazovskis",
                    "Rub\u00e9n Ballester",
                    "Franz Srambical"
                ],
                "domain": [
                    "Topological Data Analysis",
                    "Graph Neural Network",
                    "Machine Learning",
                    "Computational Topology"
                ],
                "publications": [
                    {
                        "title": "Basic Analysis of Bin-Packing Heuristics",
                        "abstract": "The bin-packing problem continues to remain relevant in numerous application areas. This technical report discusses the empirical performance of different bin-packing heuristics for certain test problems."
                    },
                    {
                        "title": "Topology meets Machine Learning: An Introduction using the Euler Characteristic Transform",
                        "abstract": "This overview article makes the case for how topological concepts can enrich research in machine learning. Using the Euler Characteristic Transform (ECT), a geometrical-topological invariant, as a running example, I present different use cases that result in more efficient models for analyzing point clouds, graphs, and meshes. Moreover, I outline a vision for how topological concepts could be used in the future, comprising (1) the learning of functions on topological spaces, (2) the building of hybrid models that imbue neural networks with knowledge about the topological information in data, and (3) the analysis of qualitative properties of neural networks. With current research already addressing some of these aspects, this article thus serves as an introduction and invitation to this nascent area of research."
                    },
                    {
                        "title": "Evaluating the \"Learning on Graphs\" Conference Experience",
                        "abstract": "With machine learning conferences growing ever larger, and reviewing processes becoming increasingly elaborate, more data-driven insights into their workings are required. In this report, we present the results of a survey accompanying the first \"Learning on Graphs\" (LoG) Conference. The survey was directed to evaluate the submission and review process from different perspectives, including authors, reviewers, and area chairs alike."
                    },
                    {
                        "title": "On the Surprising Behaviour of node2vec",
                        "abstract": "Graph embedding techniques are a staple of modern graph learning research. When using embeddings for downstream tasks such as classification, information about their stability and robustness, i.e., their susceptibility to sources of noise, stochastic effects, or specific parameter choices, becomes increasingly important. As one of the most prominent graph embedding schemes, we focus on node2vec and analyse its embedding quality from multiple perspectives. Our findings indicate that embedding quality is unstable with respect to parameter choices, and we propose strategies to remedy this in practice."
                    },
                    {
                        "title": "Persistence Concepts for 2D Skeleton Evolution Analysis",
                        "abstract": "In this work, we present concepts for the analysis of the evolution of two-dimensional skeletons. By introducing novel persistence concepts, we are able to reduce typical temporal incoherence, and provide insight in skeleton dynamics. We exemplify our approach by means of a simulation of viscous fingering---a highly dynamic process whose analysis is a hot topic in porous media research."
                    },
                    {
                        "title": "DONUT -- Creation, Development, and Opportunities of a Database",
                        "abstract": "DONUT is a database of papers about practical, real-world uses of Topological Data Analysis (TDA). Its original seed was planted in a group chat formed during the HIM Spring School on Applied and Computational Algebraic Topology in April 2017. This document describes the creation, curation, and maintenance process of the database."
                    },
                    {
                        "title": "On the Expressivity of Persistent Homology in Graph Learning",
                        "abstract": "Persistent homology, a technique from computational topology, has recently shown strong empirical performance in the context of graph classification. Being able to capture long range graph properties via higher-order topological features, such as cycles of arbitrary length, in combination with multi-scale topological descriptors, has improved predictive performance for data sets with prominent topological structures, such as molecules. At the same time, the theoretical properties of persistent homology have not been formally assessed in this context. This paper intends to bridge the gap between computational topology and graph machine learning by providing a brief introduction to persistent homology in the context of graphs, as well as a theoretical discussion and empirical analysis of its expressivity for graph learning tasks."
                    },
                    {
                        "title": "Filtration Surfaces for Dynamic Graph Classification",
                        "abstract": "Existing approaches for classifying dynamic graphs either lift graph kernels to the temporal domain, or use graph neural networks (GNNs). However, current baselines have scalability issues, cannot handle a changing node set, or do not take edge weight information into account. We propose filtration surfaces, a novel method that is scalable and flexible, to alleviate said restrictions. We experimentally validate the efficacy of our model and show that filtration surfaces outperform previous state-of-the-art baselines on datasets that rely on edge weight information. Our method does so while being either completely parameter-free or having at most one parameter, and yielding the lowest overall standard deviation among similarly scalable methods."
                    },
                    {
                        "title": "Hierarchies and Ranks for Persistence Pairs",
                        "abstract": "We develop a novel hierarchy for zero-dimensional persistence pairs, i.e., connected components, which is capable of capturing more fine-grained spatial relations between persistence pairs. Our work is motivated by a lack of spatial relationships between features in persistence diagrams, leading to a limited expressive power. We build upon a recently-introduced hierarchy of pairs in persistence diagrams that augments the pairing stored in persistence diagrams with information about which components merge. Our proposed hierarchy captures differences in branching structure. Moreover, we show how to use our hierarchy to measure the spatial stability of a pairing and we define a rank function for persistence pairs and demonstrate different applications."
                    },
                    {
                        "title": "Curvature Filtrations for Graph Generative Model Evaluation",
                        "abstract": "Graph generative model evaluation necessitates understanding differences between graphs on the distributional level. This entails being able to harness salient attributes of graphs in an efficient manner. Curvature constitutes one such property that has recently proved its utility in characterising graphs. Its expressive properties, stability, and practical utility in model evaluation remain largely unexplored, however. We combine graph curvature descriptors with emerging methods from topological data analysis to obtain robust, expressive descriptors for evaluating graph generative models."
                    },
                    {
                        "title": "Differentiable Euler Characteristic Transforms for Shape Classification",
                        "abstract": "The Euler Characteristic Transform (ECT) has proven to be a powerful representation, combining geometrical and topological characteristics of shapes and graphs. However, the ECT was hitherto unable to learn task-specific representations. We overcome this issue and develop a novel computational layer that enables learning the ECT in an end-to-end fashion. Our method, the Differentiable Euler Characteristic Transform (DECT), is fast and computationally efficient, while exhibiting performance on a par with more complex models in both graph and point cloud classification tasks. Moreover, we show that this seemingly simple statistic provides the same topological expressivity as more complex topological deep learning layers."
                    },
                    {
                        "title": "Detecting Spatial Dependence in Transcriptomics Data using Vectorised Persistence Diagrams",
                        "abstract": "Evaluating spatial patterns in data is an integral task across various domains, including geostatistics, astronomy, and spatial tissue biology. The analysis of transcriptomics data in particular relies on methods for detecting spatially-dependent features that exhibit significant spatial patterns for both explanatory analysis and feature selection. However, given the complex and high-dimensional nature of these data, there is a need for robust, stable, and reliable descriptors of spatial dependence. We leverage the stability and multiscale properties of persistent homology to address this task. To this end, we introduce a novel framework using functional topological summaries, such as Betti curves and persistence landscapes, for identifying and describing non-random patterns in spatial data. In particular, we propose a non-parametric one-sample permutation test for spatial dependence and investigate its utility across both simulated and real spatial omics data. Our vectorised approach outperforms baseline methods at accurately detecting spatial dependence. Further, we find that our method is more robust to outliers than alternative tests using Moran's I."
                    },
                    {
                        "title": "Topological Machine Learning with Persistence Indicator Functions",
                        "abstract": "Techniques from computational topology, in particular persistent homology, are becoming increasingly relevant for data analysis. Their stable metrics permit the use of many distance-based data analysis methods, such as multidimensional scaling, while providing a firm theoretical ground. Many modern machine learning algorithms, however, are based on kernels. This paper presents persistence indicator functions (PIFs), which summarize persistence diagrams, i.e., feature descriptors in topological data analysis. PIFs can be calculated and compared in linear time and have many beneficial properties, such as the availability of a kernel-based similarity measure. We demonstrate their usage in common data analysis scenarios, such as confidence set estimation and classification of complex structured data."
                    },
                    {
                        "title": "All the World's a (Hyper)Graph: A Data Drama",
                        "abstract": "We introduce Hyperbard, a dataset of diverse relational data representations derived from Shakespeare's plays. Our representations range from simple graphs capturing character co-occurrence in single scenes to hypergraphs encoding complex communication settings and character contributions as hyperedges with edge-specific node weights. By making multiple intuitive representations readily available for experimentation, we facilitate rigorous representation robustness checks in graph learning, graph mining, and network analysis, highlighting the advantages and drawbacks of specific representations. Leveraging the data released in Hyperbard, we demonstrate that many solutions to popular graph mining problems are highly dependent on the representation choice, thus calling current graph curation practices into question. As an homage to our data source, and asserting that science can also be art, we present all our points in the form of a play."
                    },
                    {
                        "title": "Topological Singularity Detection at Multiple Scales",
                        "abstract": "The manifold hypothesis, which assumes that data lies on or close to an unknown manifold of low intrinsic dimension, is a staple of modern machine learning research. However, recent work has shown that real-world data exhibits distinct non-manifold structures, i.e. singularities, that can lead to erroneous findings. Detecting such singularities is therefore crucial as a precursor to interpolation and inference tasks. We address this issue by developing a topological framework that (i) quantifies the local intrinsic dimension, and (ii) yields a Euclidicity score for assessing the 'manifoldness' of a point along multiple scales. Our approach identifies singularities of complex spaces, while also capturing singular structures and local geometric complexity in image data."
                    },
                    {
                        "title": "Ollivier-Ricci Curvature for Hypergraphs: A Unified Framework",
                        "abstract": "Bridging geometry and topology, curvature is a powerful and expressive invariant. While the utility of curvature has been theoretically and empirically confirmed in the context of manifolds and graphs, its generalization to the emerging domain of hypergraphs has remained largely unexplored. On graphs, the Ollivier-Ricci curvature measures differences between random walks via Wasserstein distances, thus grounding a geometric concept in ideas from probability theory and optimal transport. We develop ORCHID, a flexible framework generalizing Ollivier-Ricci curvature to hypergraphs, and prove that the resulting curvatures have favorable theoretical properties. Through extensive experiments on synthetic and real-world hypergraphs from different domains, we demonstrate that ORCHID curvatures are both scalable and useful to perform a variety of hypergraph tasks in practice."
                    },
                    {
                        "title": "Simplicial Representation Learning with Neural $k$-Forms",
                        "abstract": "Geometric deep learning extends deep learning to incorporate information about the geometry and topology data, especially in complex domains like graphs. Despite the popularity of message passing in this field, it has limitations such as the need for graph rewiring, ambiguity in interpreting data, and over-smoothing. In this paper, we take a different approach, focusing on leveraging geometric information from simplicial complexes embedded in $\\mathbb{R}^n$ using node coordinates. We use differential k-forms in \\mathbb{R}^n to create representations of simplices, offering interpretability and geometric consistency without message passing. This approach also enables us to apply differential geometry tools and achieve universal approximation. Our method is efficient, versatile, and applicable to various input complexes, including graphs, simplicial complexes, and cell complexes. It outperforms existing message passing neural networks in harnessing information from geometrical graphs with node features serving as coordinates."
                    },
                    {
                        "title": "Mapping the Multiverse of Latent Representations",
                        "abstract": "Echoing recent calls to counter reliability and robustness concerns in machine learning via multiverse analysis, we present PRESTO, a principled framework for mapping the multiverse of machine-learning models that rely on latent representations. Although such models enjoy widespread adoption, the variability in their embeddings remains poorly understood, resulting in unnecessary complexity and untrustworthy representations. Our framework uses persistent homology to characterize the latent spaces arising from different combinations of diverse machine-learning methods, (hyper)parameter configurations, and datasets, allowing us to measure their pairwise (dis)similarity and statistically reason about their distributions. As we demonstrate both theoretically and empirically, our pipeline preserves desirable properties of collections of latent representations, and it can be leveraged to perform sensitivity analysis, detect anomalous embeddings, or efficiently and effectively navigate hyperparameter search spaces."
                    },
                    {
                        "title": "Diss-l-ECT: Dissecting Graph Data with local Euler Characteristic Transforms",
                        "abstract": "The Euler Characteristic Transform (ECT) is an efficiently-computable geometrical-topological invariant that characterizes the global shape of data. In this paper, we introduce the Local Euler Characteristic Transform ($\\ell$-ECT), a novel extension of the ECT particularly designed to enhance expressivity and interpretability in graph representation learning. Unlike traditional Graph Neural Networks (GNNs), which may lose critical local details through aggregation, the $\\ell$-ECT provides a lossless representation of local neighborhoods. This approach addresses key limitations in GNNs by preserving nuanced local structures while maintaining global interpretability. Moreover, we construct a rotation-invariant metric based on $\\ell$-ECTs for spatial alignment of data spaces. Our method exhibits superior performance than standard GNNs on a variety of node classification tasks, particularly in graphs with high heterophily."
                    },
                    {
                        "title": "Persistent Homology via Ellipsoids",
                        "abstract": "Persistent homology is one of the most popular methods in Topological Data Analysis. An initial step in any analysis with persistent homology involves constructing a nested sequence of simplicial complexes, called a filtration, from a point cloud. There is an abundance of different complexes to choose from, with Rips, Alpha, and witness complexes being popular choices. In this manuscript, we build a different type of a geometrically-informed simplicial complex, called an ellipsoid complex. This complex is based on the idea that ellipsoids aligned with tangent directions better approximate the data compared to conventional (Euclidean) balls centered at sample points that are used in the construction of Rips and Alpha complexes, for instance. We use Principal Component Analysis to estimate tangent spaces directly from samples and present algorithms as well as an implementation for computing ellipsoid barcodes, i.e., topological descriptors based on ellipsoid complexes. Furthermore, we conduct extensive experiments and compare ellipsoid barcodes with standard Rips barcodes. Our findings indicate that ellipsoid complexes are particularly effective for estimating homology of manifolds and spaces with bottlenecks from samples. In particular, the persistence intervals corresponding to a ground-truth topological feature are longer compared to the intervals obtained when using the Rips complex of the data. Furthermore, ellipsoid barcodes lead to better classification results in sparsely-sampled point clouds. Finally, we demonstrate that ellipsoid barcodes outperform Rips barcodes in classification tasks."
                    }
                ]
            }
        }
    },
    "2402.00957": {
        "paper_data": {
            "title": "Credal Learning Theory",
            "url": "http://arxiv.org/abs/2402.00957v3",
            "arxiv_id": "2402.00957",
            "authors": [
                "Michele Caprio",
                "Maryam Sultana",
                "Eleni Elia",
                "Fabio Cuzzolin"
            ],
            "abstract": "Statistical learning theory is the foundation of machine learning, providing theoretical bounds for the risk of models learnt from a (single) training set, assumed to issue from an unknown probability distribution. In actual deployment, however, the data distribution may (and often does) vary, causing domain adaptation/generalization issues. In this paper we lay the foundations for a `credal' theory of learning, using convex sets of probabilities (credal sets) to model the variability in the data-generating distribution. Such credal sets, we argue, may be inferred from a finite sample of training sets. Bounds are derived for the case of finite hypotheses spaces (both assuming realizability or not) as well as infinite model spaces, which directly generalize classical results.",
            "introduction": "   1 Introduction  Statistical Learning Theory (SLT) considers the problem of predicting an output y\u2208\ud835\udcb4\ud835\udc66\ud835\udcb4y\\in\\mathcal{Y}italic_y \u2208 caligraphic_Y given an input x\u2208\ud835\udcb3\ud835\udc65\ud835\udcb3x\\in\\mathcal{X}italic_x \u2208 caligraphic_X by means of a mapping h:\ud835\udcb3\u2192\ud835\udcb4:\u210e\u2192\ud835\udcb3\ud835\udcb4h:\\mathcal{X}\\rightarrow\\mathcal{Y}italic_h : caligraphic_X \u2192 caligraphic_Y, h\u2208\u210b\u210e\u210bh\\in\\mathcal{H}italic_h \u2208 caligraphic_H, called model or hypothesis, belonging to model (or hypotheses) space \u210b\u210b\\mathcal{H}caligraphic_H. The loss function l:(\ud835\udcb3\u00d7\ud835\udcb4)\u00d7\u210b\u2192\u211d:\ud835\udc59\u2192\ud835\udcb3\ud835\udcb4\u210b\u211dl:(\\mathcal{X}\\times\\mathcal{Y})\\times\\mathcal{H}\\rightarrow\\mathbb{R}italic_l : ( caligraphic_X \u00d7 caligraphic_Y ) \u00d7 caligraphic_H \u2192 blackboard_R measures the error committed by a model h\u2208\u210b\u210e\u210bh\\in\\mathcal{H}italic_h \u2208 caligraphic_H. For instance, the zero-one loss is defined as l\u2062((x,y),h)\u2250\ud835\udd40\u2062[y\u2260h\u2062(x)]approaches-limit\ud835\udc59\ud835\udc65\ud835\udc66\u210e\ud835\udd40delimited-[]\ud835\udc66\u210e\ud835\udc65l((x,y),h)\\doteq\\mathbbm{I}[y\\neq h(x)]italic_l ( ( italic_x , italic_y ) , italic_h ) \u2250 blackboard_I [ italic_y \u2260 italic_h ( italic_x ) ], where \ud835\udd40\ud835\udd40\\mathbbm{I}blackboard_I denotes the indicator function. This function assigns a zero value to correct predictions and one to incorrect ones. Input-output pairs are usually assumed to be generated i.i.d. by a probability distribution P\u22c6superscript\ud835\udc43\u22c6P^{\\star}italic_P start_POSTSUPERSCRIPT \u22c6 end_POSTSUPERSCRIPT, which is unknown. The expected risk \u2013 or expected loss \u2013 of the model h\u210ehitalic_h,    L\u2062(h)\u2261LP\u22c6\u2062(h)\ud835\udc3f\u210esubscript\ud835\udc3fsuperscript\ud835\udc43\u22c6\u210e\\displaystyle L(h)\\equiv L_{P^{\\star}}(h)italic_L ( italic_h ) \u2261 italic_L start_POSTSUBSCRIPT italic_P start_POSTSUPERSCRIPT \u22c6 end_POSTSUPERSCRIPT end_POSTSUBSCRIPT ( italic_h ) \u2250\ud835\udd3cP\u22c6\u2062[l\u2062((x,y),h)]approaches-limitabsentsubscript\ud835\udd3csuperscript\ud835\udc43\u22c6delimited-[]\ud835\udc59\ud835\udc65\ud835\udc66\u210e\\displaystyle\\doteq\\mathbb{E}_{P^{\\star}}[l((x,y),h)]\u2250 blackboard_E start_POSTSUBSCRIPT italic_P start_POSTSUPERSCRIPT \u22c6 end_POSTSUPERSCRIPT end_POSTSUBSCRIPT [ italic_l ( ( italic_x , italic_y ) , italic_h ) ]      =\u222b\ud835\udcb3\u00d7\ud835\udcb4l\u2062((x,y),h)\u2062P\u22c6\u2062(d\u2062(x,y)),absentsubscript\ud835\udcb3\ud835\udcb4\ud835\udc59\ud835\udc65\ud835\udc66\u210esuperscript\ud835\udc43\u22c6d\ud835\udc65\ud835\udc66\\displaystyle=\\int_{\\mathcal{X}\\times\\mathcal{Y}}l((x,y),h)P^{\\star}(\\text{d}(% x,y)),= \u222b start_POSTSUBSCRIPT caligraphic_X \u00d7 caligraphic_Y end_POSTSUBSCRIPT italic_l ( ( italic_x , italic_y ) , italic_h ) italic_P start_POSTSUPERSCRIPT \u22c6 end_POSTSUPERSCRIPT ( d ( italic_x , italic_y ) ) ,    measures the expected value \u2013 taken with respect to P\u22c6superscript\ud835\udc43\u22c6P^{\\star}italic_P start_POSTSUPERSCRIPT \u22c6 end_POSTSUPERSCRIPT \u2013 of loss l\ud835\udc59litalic_l. The expected risk minimizer    h\u22c6\u2208arg\u2062minh\u2208\u210b\u2061L\u2062(h),superscript\u210e\u22c6subscriptargmin\u210e\u210b\ud835\udc3f\u210eh^{\\star}\\in\\operatorname*{arg\\,min}_{h\\in\\mathcal{H}}L(h),italic_h start_POSTSUPERSCRIPT \u22c6 end_POSTSUPERSCRIPT \u2208 start_OPERATOR roman_arg roman_min end_OPERATOR start_POSTSUBSCRIPT italic_h \u2208 caligraphic_H end_POSTSUBSCRIPT italic_L ( italic_h ) ,    is any hypothesis in the given model space \u210b\u210b\\mathcal{H}caligraphic_H that minimizes the expected risk. Given a training dataset D={(x1,y1),\u2026,(xn,yn)}\ud835\udc37subscript\ud835\udc651subscript\ud835\udc661\u2026subscript\ud835\udc65\ud835\udc5bsubscript\ud835\udc66\ud835\udc5bD=\\{(x_{1},y_{1}),\\ldots,(x_{n},y_{n})\\}italic_D = { ( italic_x start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , italic_y start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ) , \u2026 , ( italic_x start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT , italic_y start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT ) } whose elements are drawn independent and identically distributed (i.i.d.) from probability distribution P\u22c6superscript\ud835\udc43\u22c6P^{\\star}italic_P start_POSTSUPERSCRIPT \u22c6 end_POSTSUPERSCRIPT, the empirical risk of a hypothesis h\u210ehitalic_h is the average loss over D\ud835\udc37Ditalic_D. The empirical risk minimizer (ERM), i.e., the model h^^\u210e\\hat{h}over^ start_ARG italic_h end_ARG one actually learns from the training set D\ud835\udc37Ditalic_D, is the one minimizing the empirical risk. Statistical Learning Theory seeks upper bounds on the expected risk L\u2062(h^)\ud835\udc3f^\u210eL(\\hat{h})italic_L ( over^ start_ARG italic_h end_ARG ) of the ERM h^^\u210e\\hat{h}over^ start_ARG italic_h end_ARG, and in turn, on the excess risk, that is, on the difference between L\u2062(h^)\ud835\udc3f^\u210eL(\\hat{h})italic_L ( over^ start_ARG italic_h end_ARG ), and the lowest expected risk L\u2062(h\u22c6)\ud835\udc3fsuperscript\u210e\u22c6L(h^{\\star})italic_L ( italic_h start_POSTSUPERSCRIPT \u22c6 end_POSTSUPERSCRIPT ). This endeavor is pursued under increasingly more relaxed assumptions about the nature of the hypotheses space \u210b\u210b\\mathcal{H}caligraphic_H. Two common such assumptions are that either the model space is finite, or that there exists a model with zero expected risk (realizability).   In real-world situations, however, the data distribution may (and often does) vary, causing issues of domain adaptation (DA) or generalization (DG). Domain adaptation and generalization are interrelated yet distinctive concepts in machine learning, as they both deal with the challenges of transferring knowledge across different domains. The main goal of DA is to adapt a machine learning model trained on source domains to perform well on target domains. In opposition, DG",
            "references": [
                {
                    "title": "Conformalized Credal Set Predictors",
                    "abstract": "Credal sets are sets of probability distributions that are considered as candidates for an imprecisely known ground-truth distribution. In machine learning, they have recently attracted attention as an appealing formalism for uncertainty representation, in particular due to their ability to represent both the aleatoric and epistemic uncertainty in a prediction. However, the design of methods for learning credal set predictors remains a challenging problem. In this paper, we make use of conformal prediction for this purpose. More specifically, we propose a method for predicting credal sets in the classification task, given training data labeled by probability distributions. Since our method inherits the coverage guarantees of conformal prediction, our conformal credal sets are guaranteed to be valid with high probability (without any assumptions on model or distribution). We demonstrate the applicability of our method to natural language inference, a highly ambiguous natural language task where it is common to obtain multiple annotations per example."
                },
                {
                    "title": "CreINNs: Credal-Set Interval Neural Networks for Uncertainty Estimation in Classification Tasks",
                    "abstract": "Uncertainty estimation is increasingly attractive for improving the reliability of neural networks. In this work, we present novel credal-set interval neural networks (CreINNs) designed for classification tasks. CreINNs preserve the traditional interval neural network structure, capturing weight uncertainty through deterministic intervals, while forecasting credal sets using the mathematical framework of probability intervals. Experimental validations on an out-of-distribution detection benchmark (CIFAR10 vs SVHN) showcase that CreINNs outperform epistemic uncertainty estimation when compared to variational Bayesian neural networks (BNNs) and deep ensembles (DEs). Furthermore, CreINNs exhibit a notable reduction in computational complexity compared to variational BNNs and demonstrate smaller model sizes than DEs."
                },
                {
                    "title": "Second-Order Uncertainty Quantification: A Distance-Based Approach",
                    "abstract": "In the past couple of years, various approaches to representing and quantifying different types of predictive uncertainty in machine learning, notably in the setting of classification, have been proposed on the basis of second-order probability distributions, i.e., predictions in the form of distributions on probability distributions. A completely conclusive solution has not yet been found, however, as shown by recent criticisms of commonly used uncertainty measures associated with second-order distributions, identifying undesirable theoretical properties of these measures. In light of these criticisms, we propose a set of formal criteria that meaningful uncertainty measures for predictive uncertainty based on second-order distributions should obey. Moreover, we provide a general framework for developing uncertainty measures to account for these criteria, and offer an instantiation based on the Wasserstein distance, for which we prove that all criteria are satisfied."
                },
                {
                    "title": "A Novel Bayes' Theorem for Upper Probabilities",
                    "abstract": "In their seminal 1990 paper, Wasserman and Kadane establish an upper bound for the Bayes' posterior probability of a measurable set $A$, when the prior lies in a class of probability measures $\\mathcal{P}$ and the likelihood is precise. They also give a sufficient condition for such upper bound to hold with equality. In this paper, we introduce a generalization of their result by additionally addressing uncertainty related to the likelihood. We give an upper bound for the posterior probability when both the prior and the likelihood belong to a set of probabilities. Furthermore, we give a sufficient condition for this upper bound to become an equality. This result is interesting on its own, and has the potential of being applied to various fields of engineering (e.g. model predictive control), machine learning, and artificial intelligence."
                },
                {
                    "title": "Random-Set Neural Networks (RS-NN)",
                    "abstract": "Machine learning is increasingly deployed in safety-critical domains where robustness against adversarial attacks is crucial and erroneous predictions could lead to potentially catastrophic consequences. This highlights the need for learning systems to be equipped with the means to determine a model's confidence in its prediction and the epistemic uncertainty associated with it, 'to know when a model does not know'. In this paper, we propose a novel Random-Set Neural Network (RS-NN) for classification. RS-NN predicts belief functions rather than probability vectors over a set of classes using the mathematics of random sets, i.e., distributions over the power set of the sample space. RS-NN encodes the 'epistemic' uncertainty induced in machine learning by limited training sets via the size of the credal sets associated with the predicted belief functions. Our approach outperforms state-of-the-art Bayesian (LB-BNN, BNN-R) and Ensemble (ENN) methods in a classical evaluation setting in terms of performance, uncertainty estimation and out-of-distribution (OoD) detection on several benchmarks (CIFAR-10 vs SVHN/Intel-Image, MNIST vs FMNIST/KMNIST, ImageNet vs ImageNet-O) and scales effectively to large-scale architectures such as WideResNet-28-10, VGG16, Inception V3, EfficientNetB2, and ViT-Base."
                },
                {
                    "title": "Is the Volume of a Credal Set a Good Measure for Epistemic Uncertainty?",
                    "abstract": "Adequate uncertainty representation and quantification have become imperative in various scientific disciplines, especially in machine learning and artificial intelligence. As an alternative to representing uncertainty via one single probability measure, we consider credal sets (convex sets of probability measures). The geometric representation of credal sets as $d$-dimensional polytopes implies a geometric intuition about (epistemic) uncertainty. In this paper, we show that the volume of the geometric representation of a credal set is a meaningful measure of epistemic uncertainty in the case of binary classification, but less so for multi-class classification. Our theoretical findings highlight the crucial role of specifying and employing uncertainty measures in machine learning in an appropriate way, and for being aware of possible pitfalls."
                },
                {
                    "title": "IBCL: Zero-shot Model Generation under Stability-Plasticity Trade-offs",
                    "abstract": "Algorithms that balance the stability-plasticity trade-off are well-studied in the continual learning literature. However, only a few of them focus on obtaining models for specified trade-off preferences. When solving the problem of continual learning under specific trade-offs (CLuST), state-of-the-art techniques leverage rehearsal-based learning, which requires retraining when a model corresponding to a new trade-off preference is requested. This is inefficient since there exist infinitely many different trade-offs, and a large number of models may be requested. As a response, we propose Imprecise Bayesian Continual Learning (IBCL), an algorithm that tackles CLuST efficiently. IBCL replaces retraining with constant-time convex combination. Given a new task, IBCL (1) updates the knowledge base in the form of a convex hull of model parameter distributions and (2) generates one Pareto-optimal model per given trade-off via convex combination without any additional training. That is, obtaining models corresponding to specified trade-offs via IBCL is zero-shot. Experiments whose baselines are current CLuST algorithms show that IBCL improves by at most 45% on average per task accuracy and by 43% on peak per task accuracy, while maintaining a near-zero to positive backward transfer. Moreover, its training overhead, measured by number of batch updates, remains constant at every task, regardless of the number of preferences requested."
                },
                {
                    "title": "Constriction for sets of probabilities",
                    "abstract": "Given a set of probability measures $\\mathcal{P}$ representing an agent's knowledge on the elements of a sigma-algebra $\\mathcal{F}$, we can compute upper and lower bounds for the probability of any event $A\\in\\mathcal{F}$ of interest. A procedure generating a new assessment of beliefs is said to constrict $A$ if the bounds on the probability of $A$ after the procedure are contained in those before the procedure. It is well documented that (generalized) Bayes' updating does not allow for constriction, for all $A\\in\\mathcal{F}$. In this work, we show that constriction can take place with and without evidence being observed, and we characterize these possibilities."
                },
                {
                    "title": "Empirical Generalization Study: Unsupervised Domain Adaptation vs. Domain Generalization Methods for Semantic Segmentation in the Wild",
                    "abstract": "For autonomous vehicles and mobile robots to safely operate in the real world, i.e., the wild, scene understanding models should perform well in the many different scenarios that can be encountered. In reality, these scenarios are not all represented in the model\u2019s training data, leading to poor performance. To tackle this, current training strategies attempt to either exploit additional unlabeled data with unsupervised domain adaptation (UDA), or to reduce overfitting using the limited available labeled data with domain generalization (DG). However, it is not clear from current literature which of these methods allows for better generalization to unseen data from the wild. Therefore, in this work, we present an evaluation framework in which the generalization capabilities of state-of-the-art UDA and DG methods can be compared fairly. From this evaluation, we find that UDA methods, which leverage unlabeled data, outperform DG methods in terms of generalization, and can deliver similar performance on unseen data as fully-supervised training methods that require all data to be labeled. We show that semantic segmentation performance can be increased up to 30% for a priori unknown data without using any extra labeled data."
                },
                {
                    "title": "Domain Generalization - A Causal Perspective",
                    "abstract": "Machine learning models rely on various assumptions to attain high accuracy. One of the preliminary assumptions of these models is the independent and identical distribution, which suggests that the train and test data are sampled from the same distribution. However, this assumption seldom holds in the real world due to distribution shifts. As a result models that rely on this assumption exhibit poor generalization capabilities. Over the recent years, dedicated efforts have been made to improve the generalization capabilities of these models collectively known as -- \\textit{domain generalization methods}. The primary idea behind these methods is to identify stable features or mechanisms that remain invariant across the different distributions. Many generalization approaches employ causal theories to describe invariance since causality and invariance are inextricably intertwined. However, current surveys deal with the causality-aware domain generalization methods on a very high-level. Furthermore, we argue that it is possible to categorize the methods based on how causality is leveraged in that method and in which part of the model pipeline is it used. To this end, we categorize the causal domain generalization methods into three categories, namely, (i) Invariance via Causal Data Augmentation methods which are applied during the data pre-processing stage, (ii) Invariance via Causal representation learning methods that are utilized during the representation learning stage, and (iii) Invariance via Transferring Causal mechanisms methods that are applied during the classification stage of the pipeline. Furthermore, this survey includes in-depth insights into benchmark datasets and code repositories for domain generalization methods. We conclude the survey with insights and discussions on future directions."
                },
                {
                    "title": "Integral Probability Metrics PAC-Bayes Bounds",
                    "abstract": "We present a PAC-Bayes-style generalization bound which enables the replacement of the KL-divergence with a variety of Integral Probability Metrics (IPM). We provide instances of this bound with the IPM being the total variation metric and the Wasserstein distance. A notable feature of the obtained bounds is that they naturally interpolate between classical uniform convergence bounds in the worst case (when the prior and posterior are far away from each other), and improved bounds in favorable cases (when the posterior and prior are close). This illustrates the possibility of reinforcing classical generalization bounds with algorithm- and data-dependent components, thus making them more suitable to analyze algorithms that use a large hypothesis space."
                },
                {
                    "title": "Understanding Robust Learning through the Lens of Representation Similarities",
                    "abstract": "Representation learning, i.e. the generation of representations useful for downstream applications, is a task of fundamental importance that underlies much of the success of deep neural networks (DNNs). Recently, robustness to adversarial examples has emerged as a desirable property for DNNs, spurring the development of robust training methods that account for adversarial examples. In this paper, we aim to understand how the properties of representations learned by robust training differ from those obtained from standard, non-robust training. This is critical to diagnosing numerous salient pitfalls in robust networks, such as, degradation of performance on benign inputs, poor generalization of robustness, and increase in over-fitting. We utilize a powerful set of tools known as representation similarity metrics, across three vision datasets, to obtain layer-wise comparisons between robust and non-robust DNNs with different training procedures, architectural parameters and adversarial constraints. Our experiments highlight hitherto unseen properties of robust representations that we posit underlie the behavioral differences of robust networks. We discover a lack of specialization in robust networks' representations along with a disappearance of `block structure'. We also find overfitting during robust training largely impacts deeper layers. These, along with other findings, suggest ways forward for the design and training of better robust networks."
                },
                {
                    "title": "Sketching Algorithms and Lower Bounds for Ridge Regression",
                    "abstract": "We give a sketching-based iterative algorithm that computes a $1+\\varepsilon$ approximate solution for the ridge regression problem $\\min_x \\|Ax-b\\|_2^2 +\\lambda\\|x\\|_2^2$ where $A \\in R^{n \\times d}$ with $d \\ge n$. Our algorithm, for a constant number of iterations (requiring a constant number of passes over the input), improves upon earlier work (Chowdhury et al.) by requiring that the sketching matrix only has a weaker Approximate Matrix Multiplication (AMM) guarantee that depends on $\\varepsilon$, along with a constant subspace embedding guarantee. The earlier work instead requires that the sketching matrix has a subspace embedding guarantee that depends on $\\varepsilon$. For example, to produce a $1+\\varepsilon$ approximate solution in $1$ iteration, which requires $2$ passes over the input, our algorithm requires the OSNAP embedding to have $m= O(n\\sigma^2/\\lambda\\varepsilon)$ rows with a sparsity parameter $s = O(\\log(n))$, whereas the earlier algorithm of Chowdhury et al. with the same number of rows of OSNAP requires a sparsity $s = O(\\sqrt{\\sigma^2/\\lambda\\varepsilon} \\cdot \\log(n))$, where $\\sigma = \\opnorm{A}$ is the spectral norm of the matrix $A$. We also show that this algorithm can be used to give faster algorithms for kernel ridge regression. Finally, we show that the sketch size required for our algorithm is essentially optimal for a natural framework of algorithms for ridge regression by proving lower bounds on oblivious sketching matrices for AMM. The sketch size lower bounds for AMM may be of independent interest."
                },
                {
                    "title": "On the Limitations of General Purpose Domain Generalisation Methods",
                    "abstract": "We investigate the fundamental performance limitations of learning algorithms in several Domain Generalisation (DG) settings. Motivated by the difficulty with which previously proposed methods have in reliably outperforming Empirical Risk Minimisation (ERM), we derive upper bounds on the excess risk of ERM, and lower bounds on the minimax excess risk. Our findings show that in all the DG settings we consider, it is not possible to significantly outperform ERM. Our conclusions are limited not only to the standard covariate shift setting, but also two other settings with additional restrictions on how domains can differ. The first constrains all domains to have a non-trivial bound on pairwise distances, as measured by a broad class of integral probability metrics. The second alternate setting considers a restricted class of DG problems where all domains have the same underlying support. Our analysis also suggests how different strategies can be used to optimise the performance of ERM in each of these DG setting. We also experimentally explore hypotheses suggested by our theoretical analysis."
                },
                {
                    "title": "Statistical Learning Theory",
                    "abstract": "A machine learning system, in general, learns from the environment, but statistical machine learning programs (systems) learn from the data. This chapter presents techniques for statistical machine learning using Support Vector Machines (SVM) to recognize the patterns and classify them, predicting structured objects using SVM, k-nearest neighbor method for classification, and Naive Bayes classifiers. The artificial neural networks are presented with brief introduction to error-correction rules, Boltzmann learning, Hebbian rule, competitive learning rule, and deep learning. The instance-based learning is treated in details with its algorithm and learning task. The chapter concludes with a summary, and a set of practice exercises."
                },
                {
                    "title": "Dynamic precise and imprecise probability kinematics",
                    "abstract": "We introduce dynamic probability kinematics (DPK), a method for an agent to mechanically update subjective beliefs in the presence of partial information. We then generalize DPK to dynamic imprecise probability kinematics (DIPK), which allows the agent to express their initial beliefs via a set of probabilities in order to further take ambiguity into account. We provide bounds for the lower probability associated with the updated probability sets, and we study the behavior of the latter, in particular contraction, dilation, and sure loss. Examples are provided to illustrate how the methods work."
                },
                {
                    "title": "Towards a Theoretical Framework of Out-of-Distribution Generalization",
                    "abstract": "Generalization to out-of-distribution (OOD) data is one of the central problems in modern machine learning. Recently, there is a surge of attempts to propose algorithms that mainly build upon the idea of extracting invariant features. Although intuitively reasonable, theoretical understanding of what kind of invariance can guarantee OOD generalization is still limited, and generalization to arbitrary out-of-distribution is clearly impossible. In this work, we take the first step towards rigorous and quantitative definitions of 1) what is OOD; and 2) what does it mean by saying an OOD problem is learnable. We also introduce a new concept of expansion function, which characterizes to what extent the variance is amplified in the test domains over the training domains, and therefore give a quantitative meaning of invariant features. Based on these, we prove OOD generalization error bounds. It turns out that OOD generalization largely depends on the expansion function. As recently pointed out by Gulrajani and Lopez-Paz (2020), any OOD learning algorithm without a model selection module is incomplete. Our theory naturally induces a model selection criterion. Extensive experiments on benchmark OOD datasets demonstrate that our model selection criterion has a significant advantage over baselines."
                },
                {
                    "title": "Domain Generalization: A Survey",
                    "abstract": "Generalization to out-of-distribution (OOD) data is a capability natural to humans yet challenging for machines to reproduce. This is because most learning algorithms strongly rely on the i.i.d. assumption on source/target data, which is often violated in practice due to domain shift. Domain generalization (DG) aims to achieve OOD generalization by using only source data for model learning. Over the last ten years, research in DG has made great progress, leading to a broad spectrum of methodologies, e.g., those based on domain alignment, meta-learning, data augmentation, or ensemble learning, to name a few; DG has also been studied in various application areas including computer vision, speech recognition, natural language processing, medical imaging, and reinforcement learning. In this paper, for the first time a comprehensive literature review in DG is provided to summarize the developments over the past decade. Specifically, we first cover the background by formally defining DG and relating it to other relevant fields like domain adaptation and transfer learning. Then, we conduct a thorough review into existing methods and theories. Finally, we conclude this survey with insights and discussions on future research directions."
                },
                {
                    "title": "Generalizing to Unseen Domains: A Survey on Domain Generalization",
                    "abstract": "Machine learning systems generally assume that the training and testing distributions are the same. To this end, a key requirement is to develop models that can generalize to unseen distributions. Domain generalization (DG), i.e., out-of-distribution generalization, has attracted increasing interests in recent years. Domain generalization deals with a challenging setting where one or several different but related domain(s) are given, and the goal is to learn a model that can generalize to an unseen test domain. Great progress has been made in the area of domain generalization for years. This paper presents the first review of recent advances in this area. First, we provide a formal definition of domain generalization and discuss several related fields. We then thoroughly review the theories related to domain generalization and carefully analyze the theory behind generalization. We categorize recent algorithms into three classes: data manipulation, representation learning, and learning strategy, and present several popular algorithms in detail for each category. Third, we introduce the commonly used datasets, applications, and our open-sourced codebase for fair evaluation. Finally, we summarize existing literature and present some potential research topics for the future."
                },
                {
                    "title": "An Online Learning Approach to Interpolation and Extrapolation in Domain Generalization",
                    "abstract": "A popular assumption for out-of-distribution generalization is that the training data comprises sub-datasets, each drawn from a distinct distribution; the goal is then to\"interpolate\"these distributions and\"extrapolate\"beyond them -- this objective is broadly known as domain generalization. A common belief is that ERM can interpolate but not extrapolate and that the latter is considerably more difficult, but these claims are vague and lack formal justification. In this work, we recast generalization over sub-groups as an online game between a player minimizing risk and an adversary presenting new test distributions. Under an existing notion of inter- and extrapolation based on reweighting of sub-group likelihoods, we rigorously demonstrate that extrapolation is computationally much harder than interpolation, though their statistical complexity is not significantly different. Furthermore, we show that ERM -- or a noisy variant -- is provably minimax-optimal for both tasks. Our framework presents a new avenue for the formal analysis of domain generalization algorithms which may be of independent interest."
                },
                {
                    "title": "SWAD: Domain Generalization by Seeking Flat Minima",
                    "abstract": "Domain generalization (DG) methods aim to achieve generalizability to an unseen target domain by using only training data from the source domains. Although a variety of DG methods have been proposed, a recent study shows that under a fair evaluation protocol, called DomainBed, the simple empirical risk minimization (ERM) approach works comparable to or even outperforms previous methods. Unfortunately, simply solving ERM on a complex, non-convex loss function can easily lead to sub-optimal generalizability by seeking sharp minima. In this paper, we theoretically show that finding flat minima results in a smaller domain generalization gap. We also propose a simple yet effective method, named Stochastic Weight Averaging Densely (SWAD), to find flat minima. SWAD finds flatter minima and suffers less from overfitting than does the vanilla SWA by a dense and overfit-aware stochastic weight sampling strategy. SWAD shows state-of-the-art performances on five DG benchmarks, namely PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet, with consistent and large margins of +1.6% averagely on out-of-domain accuracy. We also compare SWAD with conventional generalization methods, such as data augmentation and consistency regularization methods, to verify that the remarkable performance improvements are originated from by seeking flat minima, not from better in-domain generalizability. Last but not least, SWAD is readily adaptable to existing DG methods without modification; the combination of SWAD and an existing DG method further improves DG performances. Source code is available at https://github.com/khanrc/swad."
                },
                {
                    "title": "WILDS: A Benchmark of in-the-Wild Distribution Shifts",
                    "abstract": "Distribution shifts -- where the training distribution differs from the test distribution -- can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity in the real-world deployments, these distribution shifts are under-represented in the datasets widely used in the ML community today. To address this gap, we present WILDS, a curated benchmark of 10 datasets reflecting a diverse range of distribution shifts that naturally arise in real-world applications, such as shifts across hospitals for tumor identification; across camera traps for wildlife monitoring; and across time and location in satellite imaging and poverty mapping. On each dataset, we show that standard training yields substantially lower out-of-distribution than in-distribution performance. This gap remains even with models trained by existing methods for tackling distribution shifts, underscoring the need for new methods for training models that are more robust to the types of distribution shifts that arise in practice. To facilitate method development, we provide an open-source package that automates dataset loading, contains default model architectures and hyperparameters, and standardizes evaluations. Code and leaderboards are available at https://wilds.stanford.edu."
                },
                {
                    "title": "In Search of Lost Domain Generalization",
                    "abstract": "The goal of domain generalization algorithms is to predict well on distributions different from those seen during training. While a myriad of domain generalization algorithms exist, inconsistencies in experimental conditions -- datasets, architectures, and model selection criteria -- render fair and realistic comparisons difficult. In this paper, we are interested in understanding how useful domain generalization algorithms are in realistic settings. As a first step, we realize that model selection is non-trivial for domain generalization tasks. Contrary to prior work, we argue that domain generalization algorithms without a model selection strategy should be regarded as incomplete. Next, we implement DomainBed, a testbed for domain generalization including seven multi-domain datasets, nine baseline algorithms, and three model selection criteria. We conduct extensive experiments using DomainBed and find that, when carefully implemented, empirical risk minimization shows state-of-the-art performance across all datasets. Looking forward, we hope that the release of DomainBed, along with contributions from fellow researchers, will streamline reproducible and rigorous research in domain generalization."
                },
                {
                    "title": "A survey on domain adaptation theory",
                    "abstract": "All famous machine learning algorithms that comprise both supervised and semi-supervised learning work well only under a common assumption: the training and test data follow the same distribution. When the distribution changes, most statistical models must be reconstructed from newly collected data, which for some applications can be costly or impossible to obtain. Therefore, it has become necessary to develop approaches that reduce the need and the effort to obtain new labeled samples by exploiting data that are available in related areas, and using these further across similar fields. This has given rise to a new machine learning framework known as transfer learning: a learning setting inspired by the capability of a human being to extrapolate knowledge across tasks to learn more efficiently. Despite a large amount of different transfer learning scenarios, the main objective of this survey is to provide an overview of the state-of-the-art theoretical results in a specific, and arguably the most popular, sub-field of transfer learning, called domain adaptation. In this sub-field, the data distribution is assumed to change across the training and the test data, while the learning task remains the same. We provide a first up-to-date description of existing results related to domain adaptation problem that cover learning bounds based on different statistical learning frameworks."
                },
                {
                    "title": "Generalizing to unseen domains via distribution matching",
                    "abstract": "Supervised learning results typically rely on assumptions of i.i.d. data. Unfortunately, those assumptions are commonly violated in practice. In this work, we tackle this problem by focusing on domain generalization: a formalization where the data generating process at test time may yield samples from never-before-seen domains (distributions). Our work relies on a simple lemma: by minimizing a notion of discrepancy between all pairs from a set of given domains, we also minimize the discrepancy between any pairs of mixtures of domains. Using this result, we derive a generalization bound for our setting. We then show that low risk over unseen domains can be achieved by representing the data in a space where (i) the training distributions are indistinguishable, and (ii) relevant information for the task at hand is preserved. Minimizing the terms in our bound yields an adversarial formulation which estimates and minimizes pairwise discrepancies. We validate our proposed strategy on standard domain generalization benchmarks, outperforming a number of recently introduced methods. Notably, we tackle a real-world application where the underlying data corresponds to multi-channel electroencephalography time series from different subjects, each considered as a distinct domain."
                },
                {
                    "title": "Domain Generalization via Multidomain Discriminant Analysis",
                    "abstract": "Domain generalization (DG) aims to incorporate knowledge from multiple source domains into a single model that could generalize well on unseen target domains. This problem is ubiquitous in practice since the distributions of the target data may rarely be identical to those of the source data. In this paper, we propose Multidomain Discriminant Analysis (MDA) to address DG of classification tasks in general situations. MDA learns a domain-invariant feature transformation that aims to achieve appealing properties, including a minimal divergence among domains within each class, a maximal separability among classes, and overall maximal compactness of all classes. Furthermore, we provide the bounds on excess risk and generalization error by learning theory analysis. Comprehensive experiments on synthetic and real benchmark datasets demonstrate the effectiveness of MDA."
                },
                {
                    "title": "A Generalization Error Bound for Multi-class Domain Generalization",
                    "abstract": "Domain generalization is the problem of assigning labels to an unlabeled data set, given several similar data sets for which labels have been provided. Despite considerable interest in this problem over the last decade, there has been no theoretical analysis in the setting of multi-class classification. In this work, we study a kernel-based learning algorithm and establish a generalization error bound that scales logarithmically in the number of classes, matching state-of-the-art bounds for multi-class classification in the conventional learning setting. We also demonstrate empirically that the proposed algorithm achieves significant performance gains compared to a pooling strategy."
                },
                {
                    "title": "Evidential Deep Learning to Quantify Classification Uncertainty",
                    "abstract": "Deterministic neural nets have been shown to learn effective predictors on a wide range of machine learning problems. However, as the standard approach is to train the network to minimize a prediction loss, the resultant model remains ignorant to its prediction confidence. Orthogonally to Bayesian neural nets that indirectly infer prediction uncertainty through weight uncertainties, we propose explicit modeling of the same using the theory of subjective logic. By placing a Dirichlet distribution on the class probabilities, we treat predictions of a neural net as subjective opinions and learn the function that collects the evidence leading to these opinions by a deterministic neural net from data. The resultant predictor for a multi-class classification problem is another Dirichlet distribution whose parameters are set by the continuous output of a neural net. We provide a preliminary analysis on how the peculiarities of our new loss function drive improved uncertainty estimation. We observe that our method achieves unprecedented success on detection of out-of-distribution queries and endurance against adversarial perturbations."
                },
                {
                    "title": "Averaging Weights Leads to Wider Optima and Better Generalization",
                    "abstract": "Deep neural networks are typically trained by optimizing a loss function with an SGD variant, in conjunction with a decaying learning rate, until convergence. We show that simple averaging of multiple points along the trajectory of SGD, with a cyclical or constant learning rate, leads to better generalization than conventional training. We also show that this Stochastic Weight Averaging (SWA) procedure finds much flatter solutions than SGD, and approximates the recent Fast Geometric Ensembling (FGE) approach with a single model. Using SWA we achieve notable improvement in test accuracy over conventional SGD training on a range of state-of-the-art residual networks, PyramidNets, DenseNets, and Shake-Shake networks on CIFAR-10, CIFAR-100, and ImageNet. In short, SWA is extremely easy to implement, improves generalization, and has almost no computational overhead."
                },
                {
                    "title": "Judicious Judgment Meets Unsettling Updating: Dilation, Sure Loss and Simpson\u2019s Paradox",
                    "abstract": "Statistical learning using imprecise probabilities is gaining more attention because it presents an alternative strategy for reducing irreplicable findings by freeing the user from the task of making up unwarranted high-resolution assumptions. However, model updating as a mathematical operation is inherently exact, hence updating imprecise models requires the user's judgment in choosing among competing updating rules. These rules often lead to incompatible inferences, and can exhibit unsettling phenomena like dilation, contraction and sure loss, which cannot occur with the Bayes rule and precise probabilities. We revisit a number of famous \"paradoxes\", including the three prisoners/Monty Hall problem, revealing that a logical fallacy arises from a set of marginally plausible yet jointly incommensurable assumptions when updating the underlying imprecise model. We establish an equivalence between Simpson's paradox and an implicit adoption of a pair of aggregation rules that induce sure loss. We also explore behavioral discrepancies between the generalized Bayes rule, Dempster's rule and the Geometric rule as alternative posterior updating rules for Choquet capacities of order 2. We show that both the generalized Bayes rule and Geometric rule are incapable of updating without prior information regardless of how strong the information in our data is, and that Dempster's rule and the Geometric rule can mathematically contradict each other with respect to dilation and contraction. Our findings show that unsettling updates reflect a collision between the rules' assumptions and the inexactness allowed by the model itself, highlighting the invaluable role of judicious judgment in handling low-resolution information, and the care we must take when applying learning rules to update imprecise probabilities."
                },
                {
                    "title": "Ergodic Theorems for lower probabilities",
                    "abstract": "We establish an Ergodic Theorem for lower probabilities, a generalization of standard probabilities widely used in applications. As an application, we provide a version for lower probabilities of the Strong Law of Large Numbers."
                },
                {
                    "title": "Generalized fiducial inference",
                    "abstract": "Generalized Fiducial Inference provides a recipe for constructing confidence intervals as well as tools for establishing the asymptotically correct coverage of many of these intervals. Confidence intervals based on the generalized fiducial framework have been applied to a wide range of problems, many of which lack exact pivots, and these intervals have exhibited excellent asymptotic and small sample performance. Generalized fiducial inference has also been used for model selection and simultaneous inference. In this article, the generalized fiducial framework is presented along with important results and applications. In the process, simple examples are used to illustrate the intuition behind and the application of the framework and its tools. WIREs Comput Stat 2014, 6:132\u2013143. doi: 10.1002/wics.1291"
                },
                {
                    "title": "Domain Generalization via Invariant Feature Representation",
                    "abstract": "This paper investigates domain generalization: How to take knowledge acquired from an arbitrary number of related domains and apply it to previously unseen domains? We propose Domain-Invariant Component Analysis (DICA), a kernel-based optimization algorithm that learns an invariant transformation by minimizing the dissimilarity across domains, whilst preserving the functional relationship between input and output variables. A learning-theoretic analysis shows that reducing dissimilarity improves the expected generalization ability of classifiers on new domains, motivating the proposed algorithm. Experimental results on synthetic and real-world datasets demonstrate that DICA successfully learns invariant features and improves classifier performance in practice."
                },
                {
                    "title": "Generalizing from Several Related Classification Tasks to a New Unlabeled Sample",
                    "abstract": "We consider the problem of assigning class labels to an unlabeled test data set, given several labeled training data sets drawn from similar distributions. This problem arises in several applications where data distributions fluctuate because of biological, technical, or other sources of variation. We develop a distribution-free, kernel-based approach to the problem. This approach involves identifying an appropriate reproducing kernel Hilbert space and optimizing a regularized empirical risk over the space. We present generalization error analysis, describe universal kernels, and establish universal consistency of the proposed methodology. Experimental results on flow cytometry data are presented."
                },
                {
                    "title": "DEMPSTER-SHAFER INFERENCE WITH WEAK BELIEFS",
                    "abstract": "The work of A. P. Dempster in 1960s extending Fisher's fiducial infer- ence for parametric inference using multivalued mapping, and that of G. Shafer in 1970s on the assessment and combination of evidence led to what is now known as the Dempster-Shafer (DS) theory of belief functions. However, application of DS for parametric inference has been limited due, perhaps, to its computational diffi- culty, non-uniqueness, and lack of frequency properties. In this paper, we return to Dempster's original approach to constructing belief functions for parametric infer- ence, called basic DS models (BDSMs), which are usual probability models on the space of the so-called focal elements. We propose to modify BDSMs by enlarging focal elements to obtain belief functions that have desired frequency properties. We call our method Weak Belief (WB). When it enlarges the focal elements no more than necessary, the method of WB is called Maximal Belief (MB). The MB method is illustrated with two examples: (i) inference about a binomial proportion, and (ii) inference about the number of outliers (\u00b5i 6 0) based on the observed data X1, . . . , Xn with the model Xi ind \u00bb N (\u00b5i, 1)."
                },
                {
                    "title": "Bayes' Theorem for Choquet Capacities",
                    "abstract": "We give an upper bound for the posterior probability of a measurable set A when the prior lies in a class of probability measures P. The bound is a rational function of two Choquet integrals. If P is weakly compact and is closed with respect to majorization, then the bound is sharp if and only if the upper prior probability is 2-alternating. The result is used to compute bounds for several sets of priors used in robust Bayesian inference. The result may be regarded as a characterization of 2-alternating Choquet capacities"
                },
                {
                    "title": "High Probability Generalization Bounds with Fast Rates for Minimax Problems",
                    "abstract": "Minimax problems are receiving an increasing amount of attention in a wide range of applications in machine learning (ML), for instance, reinforcement learning, robust optimization, adversarial learning, and distributed computing, to mention but a few. Current studies focus on the fundamental understanding of general minimax problems with an emphasis on convergence behavior. As a comparison, there is far less work to study the generalization performance. Additionally, existing generalization bounds are almost all derived in expectation, and the high probability bounds are all presented in the slow order O(1/ \u221a n), where n is the sample size. In this paper, we provide improved generalization analyses and obtain sharper high probability generalization bounds for most existing generalization measures of minimax problems. We then use the improved learning bounds to establish high probability generalization bounds with fast rates for classical empirical saddle point (ESP) solution and several popular gradient-based optimization algorithms, including gradient descent ascent (GDA), stochastic gradient descent ascent (SGDA), proximal point method (PPM), extra-gradient (EG), and optimistic gradient descent ascent (OGDA). In summary, we provide a systematical analysis of sharper generalization bounds of minimax problems."
                },
                {
                    "title": "Bayesian deep learning",
                    "abstract": "In previous chapters we reviewed Bayesian neural networks (BNNs) and historical techniques for approximate inference in these, as well as more recent approaches. We discussed the advantages and disadvantages of different techniques, examining their practicality. This, perhaps, is the most important aspect of modern techniques for approximate inference in BNNs. The field of deep learning is pushed forward by practitioners, working on real-world problems. Techniques which cannot scale to complex models with potentially millions of parameters, scale well with large amounts of data, need well studied models to be radically changed, or are not accessible to engineers, will simply perish. In this chapter we will develop on the strand of work of [Graves, 2011; Hinton and Van Camp, 1993], but will do so from the Bayesian perspective rather than the information theory one. Developing Bayesian approaches to deep learning, we will tie approximate BNN inference together with deep learning stochastic regularisation techniques (SRTs) such as dropout. These regularisation techniques are used in many modern deep learning tools, allowing us to offer a practical inference technique. We will start by reviewing in detail the tools used by [Graves, 2011]. We extend on these with recent research, commenting and analysing the variance of several stochastic estimators in variational inference (VI). Following that we will tie these derivations to SRTs, and propose practical techniques to obtain model uncertainty, even from existing models. We finish the chapter by developing specific examples for image based models (CNNs) and sequence based models (RNNs). These will be demonstrated in chapter 5, where we will survey recent research making use of the suggested tools in real-world problems."
                },
                {
                    "title": "Introduction to imprecise probabilities",
                    "abstract": "Preface Introduction Acknowledgements Outline of this Book and Guide to Readers Contributors 1 Desirability 1.1 Introduction 1.2 Reasoning about and with Sets of Desirable Gambles 1.2.1 Rationality Criteria 1.2.2 Assessments Avoiding Partial or Sure Loss 1.2.3 Coherent Sets of Desirable Gambles 1.2.4 Natural Extension 1.2.5 Desirability Relative to Subspaces with Arbitrary Vector Orderings 1.3 Deriving & Combining Sets of Desirable Gambles 1.3.1 Gamble Space Transformations 1.3.2 Derived Coherent Sets of Desirable Gambles 1.3.3 Conditional Sets of Desirable Gambles 1.3.4 Marginal Sets of Desirable Gambles 1.3.5 Combining Sets of Desirable Gambles 1.4 Partial Preference Orders 1.4.1 Strict Preference 1.4.2 Nonstrict Preference 1.4.3 Nonstrict Preferences Implied by Strict Ones 1.4.4 Strict Preferences Implied by Nonstrict Ones 1.5 Maximally Committal Sets of Strictly Desirable Gambles 1.6 Relationships with Other, Nonequivalent Models 1.6.1 Linear Previsions 1.6.2 Credal Sets 1.6.3 To Lower and Upper Previsions 1.6.4 Simplified Variants of Desirability 1.6.5 From Lower Previsions 1.6.6 Conditional Lower Previsions 1.7 Further Reading 2 Lower Previsions 2.1 Introduction 2.2 Coherent Lower Previsions 2.2.1 Avoiding Sure Loss and Coherence 2.2.2 Linear Previsions 2.2.3 Sets of Desirable Gambles 2.2.4 Natural Extension 2.3 Conditional Lower Previsions 2.3.1 Coherence of a Finite Number of Conditional Lower Previsions 2.3.2 Natural Extension of Conditional Lower Previsions 2.3.3 Coherence of an Unconditional and a Conditional Lower Prevision 2.3.4 Updating with the Regular Extension 2.4 Further Reading 2.4.1 The Work of Williams 2.4.2 The Work of Kuznetsov 2.4.3 The Work of Weichselberger 3 Structural Judgements 3.1 Introduction 3.2 Irrelevance and Independence 3.2.1 Epistemic Irrelevance 3.2.2 Epistemic Independence 3.2.3 Envelopes of Independent Precise Models 3.2.4 Strong Independence 3.2.5 The Formalist Approach to Independence 3.3 Invariance 3.3.1 Weak Invariance 3.3.2 Strong Invariance 3.4 Exchangeability. 3.4.1 Representation Theorem for Finite Sequences 3.4.2 Exchangeable Natural Extension 3.4.3 Exchangeable Sequences 3.5 Further Reading 3.5.1 Independence. 3.5.2 Invariance 3.5.3 Exchangeability 4 Special Cases 4.1 Introduction 4.2 Capacities and n-monotonicity 4.3 2-monotone Capacities 4.4 Probability Intervals on Singletons 4.5 1-monotone Capacities 4.5.1 Constructing 1-monotone Capacities 4.5.2 Simple Support Functions 4.5.3 Further Elements 4.6 Possibility Distributions, p-boxes, Clouds and Related Models. 4.6.1 Possibility Distributions 4.6.2 Fuzzy Intervals 4.6.3 Clouds 4.6.4 p-boxes. 4.7 Neighbourhood Models 4.7.1 Pari-mutuel 4.7.2 Odds-ratio 4.7.3 Linear-vacuous 4.7.4 Relations between Neighbourhood Models 4.8 Summary 5 Other Uncertainty Theories Based on Capacities 5.1 Imprecise Probability = Modal Logic + Probability 5.1.1 Boolean Possibility Theory and Modal Logic 5.1.2 A Unifying Framework for Capacity Based Uncertainty Theories 5.2 From Imprecise Probabilities to Belief Functions and Possibility Theory 5.2.1 Random Disjunctive Sets 5.2.2 Numerical Possibility Theory 5.2.3 Overall Picture 5.3 Discrepancies between Uncertainty Theories 5.3.1 Objectivist vs. Subjectivist Standpoints 5.3.2 Discrepancies in Conditioning 5.3.3 Discrepancies in Notions of Independence 5.3.4 Discrepancies in Fusion Operations 5.4 Further Reading 6 Game-Theoretic Probability 6.1 Introduction 6.2 A Law of Large Numbers 6.3 A General Forecasting Protocol 6.4 The Axiom of Continuity 6.5 Doob s Argument 6.6 Limit Theorems of Probability 6.7 Levy s Zero-One Law. 6.8 The Axiom of Continuity Revisited 6.9 Further Reading 7 Statistical Inference 7.1 Background and Introduction 7.1.1 What is Statistical Inference? 7.1.2 (Parametric) Statistical Models and i.i.d. Samples 7.1.3 Basic Tasks and Procedures of Statistical Inference 7.1.4 Some Methodological Distinctions 7.1.5 Examples: Multinomial and Normal Distribution 7.2 Imprecision in Statistics, some General Sources and Motives 7.2.1 Model and Data Imprecision Sensitivity Analysis and Ontological Views on Imprecision 7.2.2 The Robustness Shock, Sensitivity Analysis 7.2.3 Imprecision as a Modelling Tool to Express the Quality of Partial Knowledge 7.2.4 The Law of Decreasing Credibility 7.2.5 Imprecise Sampling Models: Typical Models and Motives 7.3 Some Basic Concepts of Statistical Models Relying on Imprecise Probabilities 7.3.1 Most Common Classes of Models and Notation 7.3.2 Imprecise Parametric Statistical Models and Corresponding i.i.d. Samples. 7.4 Generalized Bayesian Inference 7.4.1 Some Selected Results from Traditional Bayesian Statistics. 7.4.2 Sets of Precise Prior Distributions, Robust Bayesian Inference and the Generalized Bayes Rule 7.4.3 A Closer Exemplary Look at a Popular Class of Models: The IDM and Other Models Based on Sets of Conjugate Priors in Exponential Families. 7.4.4 Some Further Comments and a Brief Look at Other Models for Generalized Bayesian Inference 7.5 Frequentist Statistics with Imprecise Probabilities 7.5.1 The Non-robustness of Classical Frequentist Methods. 7.5.2 (Frequentist) Hypothesis Testing under Imprecise Probability: Huber-Strassen Theory and Extensions 7.5.3 Towards a Frequentist Estimation Theory under Imprecise Probabilities Some Basic Criteria and First Results 7.5.4 A Brief Outlook on Frequentist Methods 7.6 Nonparametric Predictive Inference (NPI) 7.6.1 Overview 7.6.2 Applications and Challenges 7.7 A Brief Sketch of Some Further Approaches and Aspects 7.8 Data Imprecision, Partial Identification 7.8.1 Data Imprecision 7.8.2 Cautious Data Completion 7.8.3 Partial Identification and Observationally Equivalent Models 7.8.4 A Brief Outlook on Some Further Aspects 7.9 Some General Further Reading 7.10 Some General Challenges 8 Decision Making 8.1 Non-Sequential Decision Problems 8.1.1 Choosing From a Set of Gambles 8.1.2 Choice Functions for Coherent Lower Previsions 8.2 Sequential Decision Problems 8.2.1 Static Sequential Solutions: Normal Form 8.2.2 Dynamic Sequential Solutions: Extensive Form 8.3 Examples and Applications 8.3.1 Ellsberg s Paradox 8.3.2 Robust Bayesian Statistics 9 Probabilistic Graphical Models 9.1 Introduction 9.2 Credal Sets 9.2.1 Definition and Relation with Lower Previsions 9.2.2 Marginalisation and Conditioning 9.2.3 Composition. 9.3 Independence 9.4 Credal Networks 9.4.1 Non-Separately Specified Credal Networks 9.5 Computing with Credal Networks 9.5.1 Credal Networks Updating 9.5.2 Modelling and Updating with Missing Data 9.5.3 Algorithms for Credal Networks Updating 9.5.4 Inference on Credal Networks as a Multilinear Programming Task 9.6 Further Reading 10 Classification 10.1 Introduction 10.2 Naive Bayes 10.3 Naive Credal Classifier (NCC) 10.4 Extensions and Developments of the Naive Credal Classifier 10.4.1 Lazy Naive Credal Classifier 10.4.2 Credal Model Averaging 10.4.3 Profile-likelihood Classifiers 10.4.4 Tree-Augmented Networks (TAN) 10.5 Tree-based Credal Classifiers 10.5.1 Uncertainty Measures on Credal Sets. The Maximum Entropy Function. 10.5.2 Obtaining Conditional Probability Intervals with the Imprecise Dirichlet Model 10.5.3 Classification Procedure 10.6 Metrics, Experiments and Software 10.6.1 Software. 10.6.2 Experiments. 11 Stochastic Processes 11.1 The Classical Characterization of Stochastic Processes 11.1.1 Basic Definitions 11.1.2 Precise Markov Chains 11.2 Event-driven Random Processes 11.3 Imprecise Markov Chains 11.3.1 From Precise to Imprecise Markov Chains 11.3.2 Imprecise Markov Models under Epistemic Irrelevance. 11.3.3 Imprecise Markov Models Under Strong Independence. 11.3.4 When Does the Interpretation of Independence (not) Matter? 11.4 Limit Behaviour of Imprecise Markov Chains 11.4.1 Metric Properties of Imprecise Probability Models 11.4.2 The Perron-Frobenius Theorem 11.4.3 Invariant Distributions 11.4.4 Coefficients of Ergodicity 11.4.5 Coefficients of Ergodicity for Imprecise Markov Chains. 11.5 Further Reading 12 Financial Risk Measurement 12.1 Introduction 12.2 Imprecise Previsions and Betting 12.3 Imprecise Previsions and Risk Measurement 12.3.1 Risk Measures as Imprecise Previsions 12.3.2 Coherent Risk Measures 12.3.3 Convex Risk Measures (and Previsions) 12.4 Further Reading 13 Engineering 13.1 Introduction 13.2 Probabilistic Dimensioning in a Simple Example 13.3 Random Set Modelling of the Output Variability 13.4 Sensitivity Analysis 13.5 Hybrid Models. 13.6 Reliability Analysis and Decision Making in Engineering 13.7 Further Reading 14 Reliability and Risk 14.1 Introduction 14.2 Stress-strength Reliability 14.3 Statistical Inference in Reliability and Risk 14.4 NPI in Reliablity and Risk 14.5 Discussion and Research Challenges 15 Elicitation 15.1 Methods and Issues 15.2 Evaluating Imprecise Probability Judgements 15.3 Factors Affecting Elicitation 15.4 Further Reading 16 Computation 16.1 Introduction 16.2 Natural Extension 16.2.1 Conditional Lower Previsions with Arbitrary Domains. 16.2.2 The Walley-Pelessoni-Vicig Algorithm 16.2.3 Choquet Integration 16.2.4 Mobius Inverse 16.2.5 Linear-Vacuous Mixture 16.3 Decision Making 16.3.1 Maximin, Maximax, and Hurwicz 16.3.2 Maximality 16.3.3 E-Admissibility 16.3.4 Interval Dominance References Author index Subject index"
                },
                {
                    "title": "Fiducial Inference and Belief Functions",
                    "abstract": "One aspect of Fisher\u2019s work which has puzzled a great many statisticians is the idea of Fiducial Inference. Using \u201cpivotal\u201d variables, Fisher moves from logical statements about restrictions of parameter spaces to probabilistic statements about parameters. The method works, but with a lot of caveats: the pivotal variables must be sufficient statistics, must be continuous, &c. Dempster\u2019s explorations of the fiducial method as applied to the binomial distribution in 1966 through 1968 lead to the formulation of an upper and lower probability model later extended and named \u201cbelief functions\u201d by Shafer. Dempster revisits the ideas in a 1989 paper and provides a philosophical synthesis of the ideas of Bayesian, Belief Function and Fiducial Inference. In particular, the theory of belief functions contains representations for logical statements about membership of parameters in sets and probabilistic statments about parameters. It moves between the two representations through subjective statments about the state of knowledge in the form of upper and lower probabilities. This paper explores the belief function and fiducial arguments using a \u201cstate-of-knowledge\u201d argument tie the twomethods together. In this light the paper reviews fiducial arguments for the binomial and normal distribution. It also compares the fiducial and belief function arguments for the Poisson process and arrives at an interesting paradox. Key Concepts: Fiducial Inference, Belief Functions, Normal Inference, Binomial distribution, Poisson process, Pivotal variables"
                },
                {
                    "title": "Monographs on statistics and applied probability",
                    "abstract": "1 Stochastic Population Models in Ecology and Epidemiology M.S. Barlett (1960) 2 Queues D.R. Cox and W.L. Smith (1961) 3 Monte Carlo Methods J.M. Hammersley and D.C. Handscomb (1964) 4 The Statistical Analysis of Series of Events D.R. Cox and P.A.W. Lewis (1966) 5 Population Genetics W.J. Ewens (1969) 6 Probability, Statistics and Time M.S. Barlett (1975) 7 Statistical Inference S.D. Silvey (1975) 8 The Analysis of Contingency Tables B.S. Everitt (1977) 9 Multivariate Analysis in Behavioural Research A.E. Maxwell (1977) 10 Stochastic Abundance Models S. Engen (1978) 11 Some Basic Theory for Statistical Inference E.J.G. Pitman (1979) 12 Point Processes D.R. Cox and V. Isham (1980) 13 Identification of Outliers D.M. Hawkins (1980) 14 Optimal Design S.D. Silvey (1980) 15 Finite Mixture Distributions B.S. Everitt and D.J. Hand (1981) 16 Classification A.D. Gordon (1981) 17 Distribution-Free Statistical Methods, 2nd edition J.S. Maritz (1995) 18 Residuals and Influence in Regression R.D. Cook and S. Weisberg (1982) 19 Applications of Queueing Theory, 2nd edition G.F. Newell (1982) 20 Risk Theory, 3rd edition R.E. Beard, T. Pentik\u00e4inen and E. Pesonen (1984) 21 Analysis of Survival Data D.R. Cox and D. Oakes (1984) 22 An Introduction to Latent Variable Models B.S. Everitt (1984) 23 Bandit Problems D.A. Berry and B. Fristedt (1985) 24 Stochastic Modelling and Control M.H.A. Davis and R. Vinter (1985) 25 The Statistical Analysis of Composition Data J. Aitchison (1986) 26 Density Estimation for Statistics and Data Analysis B.W. Silverman (1986) 27 Regression Analysis with Applications G.B. Wetherill (1986) 28 Sequential Methods in Statistics, 3rd edition G.B. Wetherill and K.D. Glazebrook (1986) 29 Tensor Methods in Statistics P. McCullagh (1987) 30 Transformation and Weighting in Regression R.J. Carroll and D. Ruppert (1988) 31 Asymptotic Techniques for Use in Statistics O.E. Bandorff-Nielsen and D.R. Cox (1989) 32 Analysis of Binary Data, 2nd edition D.R. Cox and E.J. Snell (1989) 33 Analysis of Infectious Disease Data N.G. Becker (1989) 34 Design and Analysis of Cross-Over Trials B. Jones and M.G. Kenward (1989) 35 Empirical Bayes Methods, 2nd edition J.S. Maritz and T. Lwin (1989) 36 Symmetric Multivariate and Related Distributions K.T. Fang, S. Kotz and K.W. Ng (1990)"
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively address the challenges of domain adaptation and generalization in machine learning when the data distribution varies between training and testing environments?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of domain adaptation and generalization is crucial for the research community as it directly impacts the reliability and applicability of machine learning models in real-world scenarios. By improving our understanding and methodologies in this area, we can enhance model performance across diverse applications, leading to more robust AI systems. This research could pave the way for advancements in various fields, such as healthcare, finance, and autonomous systems, where models must adapt to changing environments and data distributions.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing domain adaptation and generalization stem from the inherent differences in data distributions between source and target domains. Naive approaches, such as directly applying a model trained on one dataset to another, often fail due to these discrepancies, leading to poor performance. Technical obstacles include the need for effective feature alignment, the complexity of measuring domain similarity, and the difficulty in ensuring that learned representations are transferable. Theoretical challenges involve understanding the limits of generalization and the conditions under which models can adapt successfully.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either domain adaptation or generalization in isolation, leading to a lack of comprehensive solutions that address both simultaneously. Limitations in existing methodologies, such as reliance on strong assumptions about data distributions or the availability of labeled data in target domains, have hindered progress. Additionally, many approaches have not adequately considered the dynamic nature of real-world data, which can change over time. Our approach aims to integrate insights from both areas and develop more flexible models that can adapt to varying conditions.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a hybrid model that combines techniques from domain adaptation and generalization, utilizing a diverse dataset that includes samples from both source and target domains. We will employ metrics such as domain discrepancy measures and generalization error bounds to evaluate model performance. The expected outcomes include improved model robustness and adaptability, demonstrated through empirical results showing enhanced performance on target domain tasks compared to traditional methods. This approach aims to provide a more unified framework for tackling the challenges of varying data distributions in machine learning."
            }
        },
        "author_data": {
            "d58c295e-cf80-4df4-b05c-530fa7eb372e": {
                "pk": "d58c295e-cf80-4df4-b05c-530fa7eb372e",
                "name": "Michele Caprio",
                "collaborators": [
                    "Insup Lee",
                    "Sayan Mukherjee",
                    "Yusuf Sale",
                    "Eyke H\u00fcllermeier",
                    "Souradeep Dutta",
                    "Kuk Jin Jang",
                    "Vivian Lin",
                    "Oleg Sokolsky",
                    "Pengyuan Lu",
                    "Eric Eaton"
                ],
                "domain": [
                    "Uncertainty Quantification",
                    "Machine Learning",
                    "Probability Theory",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Refined Pinsker's and reverse Pinsker's inequalities for probability distributions of different dimensions",
                        "abstract": "We provide optimal lower and upper bounds for the augmented Kullback-Leibler divergence in terms of the augmented total variation distance between two probability measures defined on two Euclidean spaces having different dimensions. We call them refined Pinsker's and reverse Pinsker's inequalities, respectively."
                    },
                    {
                        "title": "Optimal Transport for $\u03b5$-Contaminated Credal Sets",
                        "abstract": "We provide a version for lower probabilities of Monge's and Kantorovich's optimal transport problems. We show that, when the lower probabilities are the lower envelopes of $\\epsilon$-contaminated sets, then our version of Monge's, and a restricted version of our Kantorovich's problems, coincide with their respective classical versions. We also give sufficient conditions for the existence of our version of Kantorovich's optimal plan, and for the two problems to be equivalent. As a byproduct, we show that for $\\epsilon$-contaminations the lower probability versions of Monge's and Kantorovich's optimal transport problems need not coincide. The applications of our results to Machine Learning and Artificial Intelligence are also discussed."
                    },
                    {
                        "title": "Imprecise Markov Semigroups and their Ergodicity",
                        "abstract": "We introduce the concept of an imprecise Markov semigroup $\\mathbf{Q}$. It is a tool that allows to represent ambiguity around both the initial and the transition probabilities of a Markov process via a compact collection of plausible Markov semigroups, each associated with a (different, plausible) Markov process. We use techniques from geometry, functional analysis, and (high dimensional) probability to study the ergodic behavior of $\\mathbf{Q}$. We show that, if the initial distribution of the Markov processes associated with the elements of $\\mathbf{Q}$ is known and invariant, under some conditions that also involve the geometry of the state space, eventually the ambiguity around their transition probability fades. We call this property ergodicity of the imprecise Markov semigroup, and we relate it to the classical notion of ergodicity. We prove ergodicity both when the state space is Euclidean or a Riemannian manifold, and when it is an arbitrary measurable space. The importance of our findings for the fields of machine learning and computer vision is also discussed."
                    },
                    {
                        "title": "Dynamic Precise and Imprecise Probability Kinematics",
                        "abstract": "We introduce dynamic probability kinematics (DPK), a method for an agent to mechanically update subjective beliefs in the presence of partial information. We then generalize DPK to dynamic imprecise probability kinematics (DIPK), which allows the agent to express their initial beliefs via a set of probabilities in order to further take ambiguity into account. We provide bounds for the lower probability associated with the updated probability sets, and we study the behavior of the latter, in particular contraction, dilation, and sure loss. Examples are provided to illustrate how the methods work."
                    },
                    {
                        "title": "Concentration inequalities and optimal number of layers for stochastic deep neural networks",
                        "abstract": "We state concentration inequalities for the output of the hidden layers of a stochastic deep neural network (SDNN), as well as for the output of the whole SDNN. These results allow us to introduce an expected classifier (EC), and to give probabilistic upper bound for the classification error of the EC. We also state the optimal number of layers for the SDNN via an optimal stopping procedure. We apply our analysis to a stochastic version of a feedforward neural network with ReLU activation function."
                    },
                    {
                        "title": "Constriction for sets of probabilities",
                        "abstract": "Given a set of probability measures $\\mathcal{P}$ representing an agent's knowledge on the elements of a sigma-algebra $\\mathcal{F}$, we can compute upper and lower bounds for the probability of any event $A\\in\\mathcal{F}$ of interest. A procedure generating a new assessment of beliefs is said to constrict $A$ if the bounds on the probability of $A$ after the procedure are contained in those before the procedure. It is well documented that (generalized) Bayes' updating does not allow for constriction, for all $A\\in\\mathcal{F}$. In this work, we show that constriction can take place with and without evidence being observed, and we characterize these possibilities."
                    },
                    {
                        "title": "Ergodic Theorems for Dynamic Imprecise Probability Kinematics",
                        "abstract": "We formulate an ergodic theory for the (almost sure) limit $\\mathcal{P}^\\text{co}_{\\tilde{\\mathcal{E}}}$ of a sequence $(\\mathcal{P}^\\text{co}_{\\mathcal{E}_n})$ of successive dynamic imprecise probability kinematics (DIPK, introduced in Caprio and Gong, 2021) updates of a set $\\mathcal{P}^\\text{co}_{\\mathcal{E}_0}$ representing the initial beliefs of an agent. As a consequence, we formulate a strong law of large numbers."
                    },
                    {
                        "title": "Finite admixture models: a bridge with stochastic geometry and Choquet theory",
                        "abstract": "In the context of a finite admixture model whose components and weights are unknown, if the number of identifiable components is a function of the amount of data collected, we use techniques from stochastic convex geometry to find the growth rate of its expected value. In addition, when the components are known but the weights are not, we provide an application of the classic Glivenko-Cantelli's theorem that allows us to retrieve the Choquet measure supported on the identifiable admixture components. In turn, this gives us the identifiable admixture weights. Finally, we propose a novel algorithm that estimates the model capturing the complexity of the data using only the strictly necessary number of components."
                    },
                    {
                        "title": "Extended probabilities and their application to statistical inference",
                        "abstract": "We propose a new, more general definition of extended probability measures. We study their properties and provide a behavioral interpretation. We put them to use in an inference procedure, whose environment is canonically represented by the probability space $(\\Omega,\\mathcal{F},P)$, when both $P$ and the composition of $\\Omega$ are unknown. We develop an ex ante analysis -- taking place before the statistical analysis requiring knowledge of $\\Omega$ -- in which the true composition of $\\Omega$ is progressively learned. We describe how to update extended probabilities in this setting, and introduce the concept of lower extended probabilities. We apply our findings to a species sampling problem and to the study of the boomerang effect (the empirical observation that sometimes persuasion yields the opposite effect: the persuaded agent moves their opinion away from the opinion of the persuading agent)."
                    },
                    {
                        "title": "Concerning three classes of non-Diophantine arithmetics",
                        "abstract": "We present three classes of abstract prearithmetics, $\\{\\mathbf{A}_M\\}_{M \\geq 1}$, $\\{\\mathbf{A}_{-M,M}\\}_{M \\geq 1}$, and $\\{\\mathbf{B}_M\\}_{M > 0}$. The first one is weakly projective with respect to the nonnegative real Diophantine arithmetic $\\mathbf{R_+}=(\\mathbb{R}_+,+,\\times,\\leq_{\\mathbb{R}_+})$, the second one is weakly projective with respect to the real Diophantine arithmetic $\\mathbf{R}=(\\mathbb{R},+,\\times,\\leq_{\\mathbb{R}})$, while the third one is projective with respect to the extended real Diophantine arithmetic $\\overline{\\mathbf{R}}=(\\overline{\\mathbb{R}},+,\\times,\\leq_{\\overline{\\mathbb{R}}})$. In addition, we have that every $\\mathbf{A}_M$ and every $\\mathbf{B}_M$ are a complete totally ordered semiring, while every $\\mathbf{A}_{-M,M}$ is not. We show that the projection of any series of elements of $\\mathbb{R}_+$ converges in $\\mathbf{A}_M$, for any $M \\geq 1$, and that the projection of any non-oscillating series series of elements of $\\mathbb{R}$ converges in $\\mathbf{A}_{-M,M}$, for any $M \\geq 1$, and in $\\mathbf{B}_M$, for all $M > 0$. We also prove that working in $\\mathbf{A}_M$ and in $\\mathbf{A}_{-M,M}$, for any $M \\geq 1$, and in $\\mathbf{B}_M$, for all $M>0$, allows to overcome a version of the paradox of the heap."
                    },
                    {
                        "title": "A Novel Bayes' Theorem for Upper Probabilities",
                        "abstract": "In their seminal 1990 paper, Wasserman and Kadane establish an upper bound for the Bayes' posterior probability of a measurable set $A$, when the prior lies in a class of probability measures $\\mathcal{P}$ and the likelihood is precise. They also give a sufficient condition for such upper bound to hold with equality. In this paper, we introduce a generalization of their result by additionally addressing uncertainty related to the likelihood. We give an upper bound for the posterior probability when both the prior and the likelihood belong to a set of probabilities. Furthermore, we give a sufficient condition for this upper bound to become an equality. This result is interesting on its own, and has the potential of being applied to various fields of engineering (e.g. model predictive control), machine learning, and artificial intelligence."
                    },
                    {
                        "title": "Turnpike in infinite dimension",
                        "abstract": "Let $\\Phi$ be a correspondence from a normed vector space $X$ into itself, let $u: X\\to \\mathbf{R}$ be a function, and $\\mathcal{I}$ be an ideal on $\\mathbf{N}$. Also, assume that the restriction of $u$ on the fixed points of $\\Phi$ has a unique maximizer $\\eta^\\star$. Then, we consider feasible paths $(x_0,x_1,\\ldots)$ with values in $X$ such that $x_{n+1} \\in \\Phi(x_n)$ for all $n\\ge 0$. Under certain additional conditions, we prove the following turnpike result: every feasible path $(x_0,x_1,\\ldots)$ which maximizes the smallest $\\mathcal{I}$-cluster point of the sequence $(u(x_0),u(x_1),\\ldots)$ is necessarily $\\mathcal{I}$-convergent to $\\eta^\\star$.   We provide examples that, on the one hand, justify the hypotheses of our result and, on the other hand, prove that we are including new cases which were previously not considered in the related literature."
                    },
                    {
                        "title": "Is the Volume of a Credal Set a Good Measure for Epistemic Uncertainty?",
                        "abstract": "Adequate uncertainty representation and quantification have become imperative in various scientific disciplines, especially in machine learning and artificial intelligence. As an alternative to representing uncertainty via one single probability measure, we consider credal sets (convex sets of probability measures). The geometric representation of credal sets as $d$-dimensional polytopes implies a geometric intuition about (epistemic) uncertainty. In this paper, we show that the volume of the geometric representation of a credal set is a meaningful measure of epistemic uncertainty in the case of binary classification, but less so for multi-class classification. Our theoretical findings highlight the crucial role of specifying and employing uncertainty measures in machine learning in an appropriate way, and for being aware of possible pitfalls."
                    },
                    {
                        "title": "Second-Order Uncertainty Quantification: A Distance-Based Approach",
                        "abstract": "In the past couple of years, various approaches to representing and quantifying different types of predictive uncertainty in machine learning, notably in the setting of classification, have been proposed on the basis of second-order probability distributions, i.e., predictions in the form of distributions on probability distributions. A completely conclusive solution has not yet been found, however, as shown by recent criticisms of commonly used uncertainty measures associated with second-order distributions, identifying undesirable theoretical properties of these measures. In light of these criticisms, we propose a set of formal criteria that meaningful uncertainty measures for predictive uncertainty based on second-order distributions should obey. Moreover, we provide a general framework for developing uncertainty measures to account for these criteria, and offer an instantiation based on the Wasserstein distance, for which we prove that all criteria are satisfied."
                    },
                    {
                        "title": "IBCL: Zero-shot Model Generation under Stability-Plasticity Trade-offs",
                        "abstract": "Algorithms that balance the stability-plasticity trade-off are well-studied in the continual learning literature. However, only a few of them focus on obtaining models for specified trade-off preferences. When solving the problem of continual learning under specific trade-offs (CLuST), state-of-the-art techniques leverage rehearsal-based learning, which requires retraining when a model corresponding to a new trade-off preference is requested. This is inefficient since there exist infinitely many different trade-offs, and a large number of models may be requested. As a response, we propose Imprecise Bayesian Continual Learning (IBCL), an algorithm that tackles CLuST efficiently. IBCL replaces retraining with constant-time convex combination. Given a new task, IBCL (1) updates the knowledge base in the form of a convex hull of model parameter distributions and (2) generates one Pareto-optimal model per given trade-off via convex combination without any additional training. That is, obtaining models corresponding to specified trade-offs via IBCL is zero-shot. Experiments whose baselines are current CLuST algorithms show that IBCL improves by at most 45% on average per task accuracy and by 43% on peak per task accuracy, while maintaining a near-zero to positive backward transfer. Moreover, its training overhead, measured by number of batch updates, remains constant at every task, regardless of the number of preferences requested."
                    },
                    {
                        "title": "IBCL: Zero-shot Model Generation for Task Trade-offs in Continual Learning",
                        "abstract": "Like generic multi-task learning, continual learning has the nature of multi-objective optimization, and therefore faces a trade-off between the performance of different tasks. That is, to optimize for the current task distribution, it may need to compromise performance on some previous tasks. This means that there exist multiple models that are Pareto-optimal at different times, each addressing a distinct task performance trade-off. Researchers have discussed how to train particular models to address specific trade-off preferences. However, existing algorithms require training overheads proportional to the number of preferences -- a large burden when there are multiple, possibly infinitely many, preferences. As a response, we propose Imprecise Bayesian Continual Learning (IBCL). Upon a new task, IBCL (1) updates a knowledge base in the form of a convex hull of model parameter distributions and (2) obtains particular models to address task trade-off preferences with zero-shot. That is, IBCL does not require any additional training overhead to generate preference-addressing models from its knowledge base. We show that models obtained by IBCL have guarantees in identifying the Pareto optimal parameters. Moreover, experiments on standard image classification and NLP tasks support this guarantee. Statistically, IBCL improves average per-task accuracy by at most 23\\% and peak per-task accuracy by at most 15\\% with respect to the baseline methods, with steadily near-zero or positive backward transfer. Most importantly, IBCL significantly reduces the training overhead from training 1 model per preference to at most 3 models for all preferences."
                    },
                    {
                        "title": "Credal Bayesian Deep Learning",
                        "abstract": "Uncertainty quantification and robustness to distribution shifts are important goals in machine learning and artificial intelligence. Although Bayesian Neural Networks (BNNs) allow for uncertainty in the predictions to be assessed, different sources of predictive uncertainty cannot be distinguished properly. We present Credal Bayesian Deep Learning (CBDL). Heuristically, CBDL allows to train an (uncountably) infinite ensemble of BNNs, using only finitely many elements. This is possible thanks to prior and likelihood finitely generated credal sets (FGCSs), a concept from the imprecise probability literature. Intuitively, convex combinations of a finite collection of prior-likelihood pairs are able to represent infinitely many such pairs. After training, CBDL outputs a set of posteriors on the parameters of the neural network. At inference time, such posterior set is used to derive a set of predictive distributions that is in turn utilized to distinguish between (predictive) aleatoric and epistemic uncertainties, and to quantify them. The predictive set also produces either (i) a collection of outputs enjoying desirable probabilistic guarantees, or (ii) the single output that is deemed the best, that is, the one having the highest predictive lower probability -- another imprecise-probabilistic concept. CBDL is more robust than single BNNs to prior and likelihood misspecification, and to distribution shift. We show that CBDL is better at quantifying and disentangling different types of (predictive) uncertainties than single BNNs and ensemble of BNNs. In addition, we apply CBDL to two case studies to demonstrate its downstream tasks capabilities: one, for motion prediction in autonomous driving scenarios, and two, to model blood glucose and insulin dynamics for artificial pancreas control. We show that CBDL performs better when compared to an ensemble of BNNs baseline."
                    },
                    {
                        "title": "Distributionally Robust Statistical Verification with Imprecise Neural Networks",
                        "abstract": "A particularly challenging problem in AI safety is providing guarantees on the behavior of high-dimensional autonomous systems. Verification approaches centered around reachability analysis fail to scale, and purely statistical approaches are constrained by the distributional assumptions about the sampling process. Instead, we pose a distributionally robust version of the statistical verification problem for black-box systems, where our performance guarantees hold over a large family of distributions. This paper proposes a novel approach based on a combination of active learning, uncertainty quantification, and neural network verification. A central piece of our approach is an ensemble technique called Imprecise Neural Networks, which provides the uncertainty to guide active learning. The active learning uses an exhaustive neural-network verification tool Sherlock to collect samples. An evaluation on multiple physical simulators in the openAI gym Mujoco environments with reinforcement-learned controllers demonstrates that our approach can provide useful and scalable guarantees for high-dimensional systems."
                    },
                    {
                        "title": "DC4L: Distribution Shift Recovery via Data-Driven Control for Deep Learning Models",
                        "abstract": "Deep neural networks have repeatedly been shown to be non-robust to the uncertainties of the real world, even to naturally occurring ones. A vast majority of current approaches have focused on data-augmentation methods to expand the range of perturbations that the classifier is exposed to while training. A relatively unexplored avenue that is equally promising involves sanitizing an image as a preprocessing step, depending on the nature of perturbation. In this paper, we propose to use control for learned models to recover from distribution shifts online. Specifically, our method applies a sequence of semantic-preserving transformations to bring the shifted data closer in distribution to the training set, as measured by the Wasserstein distance. Our approach is to 1) formulate the problem of distribution shift recovery as a Markov decision process, which we solve using reinforcement learning, 2) identify a minimum condition on the data for our method to be applied, which we check online using a binary classifier, and 3) employ dimensionality reduction through orthonormal projection to aid in our estimates of the Wasserstein distance. We provide theoretical evidence that orthonormal projection preserves characteristics of the data at the distributional level. We apply our distribution shift recovery approach to the ImageNet-C benchmark for distribution shifts, demonstrating an improvement in average accuracy of up to 14.21% across a variety of state-of-the-art ImageNet classifiers. We further show that our method generalizes to composites of shifts from the ImageNet-C benchmark, achieving improvements in average accuracy of up to 9.81%. Finally, we test our method on CIFAR-100-C and report improvements of up to 8.25%."
                    }
                ]
            },
            "1f801913-083c-4d16-96fb-752c48cdca76": {
                "pk": "1f801913-083c-4d16-96fb-752c48cdca76",
                "name": "Maryam Sultana",
                "collaborators": [
                    "Soon Ki Jung",
                    "Sajid Javed",
                    "Arif Mahmood",
                    "Muzammal Naseer",
                    "Muhammad Haris Khan",
                    "Salman Khan",
                    "Fahad Shahbaz Khan",
                    "Thierry Bouwmans"
                ],
                "domain": [
                    "Computer Vision",
                    "Generative Adversarial Networks",
                    "Background Subtraction",
                    "Domain Generalization"
                ],
                "publications": [
                    {
                        "title": "Illumination Invariant Foreground Object Segmentation using ForeGANs",
                        "abstract": "The foreground segmentation algorithms suffer performance degradation in the presence of various challenges such as dynamic backgrounds, and various illumination conditions. To handle these challenges, we present a foreground segmentation method, based on generative adversarial network (GAN). We aim to segment foreground objects in the presence of two aforementioned major challenges in background scenes in real environments. To address this problem, our presented GAN model is trained on background image samples with dynamic changes, after that for testing the GAN model has to generate the same background sample as test sample with similar conditions via back-propagation technique. The generated background sample is then subtracted from the given test sample to segment foreground objects. The comparison of our proposed method with five state-of-the-art methods highlights the strength of our algorithm for foreground segmentation in the presence of challenging dynamic background scenario."
                    },
                    {
                        "title": "Unsupervised Deep Context Prediction for Background Foreground Separation",
                        "abstract": "In many advanced video based applications background modeling is a pre-processing step to eliminate redundant data, for instance in tracking or video surveillance applications. Over the past years background subtraction is usually based on low level or hand-crafted features such as raw color components, gradients, or local binary patterns. The background subtraction algorithms performance suffer in the presence of various challenges such as dynamic backgrounds, photometric variations, camera jitters, and shadows. To handle these challenges for the purpose of accurate background modeling we propose a unified framework based on the algorithm of image inpainting. It is an unsupervised visual feature learning hybrid Generative Adversarial algorithm based on context prediction. We have also presented the solution of random region inpainting by the fusion of center region inpaiting and random region inpainting with the help of poisson blending technique. Furthermore we also evaluated foreground object detection with the fusion of our proposed method and morphological operations. The comparison of our proposed method with 12 state-of-the-art methods shows its stability in the application of background estimation and foreground detection."
                    },
                    {
                        "title": "Unsupervised RGBD Video Object Segmentation Using GANs",
                        "abstract": "Video object segmentation is a fundamental step in many advanced vision applications. Most existing algorithms are based on handcrafted features such as HOG, super-pixel segmentation or texture-based techniques, while recently deep features have been found to be more efficient. Existing algorithms observe performance degradation in the presence of challenges such as illumination variations, shadows, and color camouflage. To handle these challenges we propose a fusion based moving object segmentation algorithm which exploits color as well as depth information using GAN to achieve more accuracy. Our goal is to segment moving objects in the presence of challenging background scenes, in real environments. To address this problem, GAN is trained in an unsupervised manner on color and depth information independently with challenging video sequences. During testing, the trained GAN generates backgrounds similar to that in the test sample. The generated background samples are then compared with the test sample to segment moving objects. The final result is computed by fusion of object boundaries in both modalities, RGB and the depth. The comparison of our proposed algorithm with five state-of-the-art methods on publicly available dataset has shown the strength of our algorithm for moving object segmentation in videos in the presence of challenging real scenarios."
                    },
                    {
                        "title": "Self-Distilled Vision Transformer for Domain Generalization",
                        "abstract": "In the recent past, several domain generalization (DG) methods have been proposed, showing encouraging performance, however, almost all of them build on convolutional neural networks (CNNs). There is little to no progress on studying the DG performance of vision transformers (ViTs), which are challenging the supremacy of CNNs on standard benchmarks, often built on i.i.d assumption. This renders the real-world deployment of ViTs doubtful. In this paper, we attempt to explore ViTs towards addressing the DG problem. Similar to CNNs, ViTs also struggle in out-of-distribution scenarios and the main culprit is overfitting to source domains. Inspired by the modular architecture of ViTs, we propose a simple DG approach for ViTs, coined as self-distillation for ViTs. It reduces the overfitting of source domains by easing the learning of input-output mapping problem through curating non-zero entropy supervisory signals for intermediate transformer blocks. Further, it does not introduce any new parameters and can be seamlessly plugged into the modular composition of different ViTs. We empirically demonstrate notable performance gains with different DG baselines and various ViT backbones in five challenging datasets. Moreover, we report favorable performance against recent state-of-the-art DG methods. Our code along with pre-trained models are publicly available at: https://github.com/maryam089/SDViT."
                    },
                    {
                        "title": "Deep Neural Network Concepts for Background Subtraction: A Systematic Review and Comparative Evaluation",
                        "abstract": "Conventional neural networks show a powerful framework for background subtraction in video acquired by static cameras. Indeed, the well-known SOBS method and its variants based on neural networks were the leader methods on the largescale CDnet 2012 dataset during a long time. Recently, convolutional neural networks which belong to deep learning methods were employed with success for background initialization, foreground detection and deep learned features. Currently, the top current background subtraction methods in CDnet 2014 are based on deep neural networks with a large gap of performance in comparison on the conventional unsupervised approaches based on multi-features or multi-cues strategies. Furthermore, a huge amount of papers was published since 2016 when Braham and Van Droogenbroeck published their first work on CNN applied to background subtraction providing a regular gain of performance. In this context, we provide the first review of deep neural network concepts in background subtraction for novices and experts in order to analyze this success and to provide further directions. For this, we first surveyed the methods used background initialization, background subtraction and deep learned features. Then, we discuss the adequacy of deep neural networks for background subtraction. Finally, experimental results are presented on the CDnet 2014 dataset."
                    }
                ]
            },
            "dc8a67a7-8b26-4f0e-aa82-e94d94ed7c57": {
                "pk": "dc8a67a7-8b26-4f0e-aa82-e94d94ed7c57",
                "name": "Eleni Elia",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "032c63e0-bead-45c5-b725-46501f124985": {
                "pk": "032c63e0-bead-45c5-b725-46501f124985",
                "name": "Fabio Cuzzolin",
                "collaborators": [
                    "Gurkirt Singh",
                    "Suman Saha",
                    "Neha Bhargava",
                    "Kurt Degiorgio",
                    "Salman Khan",
                    "Hong Zhu",
                    "Dongmei Liu",
                    "Ian Bayley",
                    "Rachel Harrison",
                    "Shireen Kudukkil Manchingal"
                ],
                "domain": [
                    "Statistical Learning",
                    "Uncertainty Theory",
                    "Computer Vision",
                    "Action Detection"
                ],
                "publications": [
                    {
                        "title": "Generalising realisability in statistical learning theory under epistemic uncertainty",
                        "abstract": "The purpose of this paper is to look into how central notions in statistical learning theory, such as realisability, generalise under the assumption that train and test distribution are issued from the same credal set, i.e., a convex set of probability distributions. This can be considered as a first step towards a more general treatment of statistical learning under epistemic uncertainty."
                    },
                    {
                        "title": "Belief likelihood function for generalised logistic regression",
                        "abstract": "The notion of belief likelihood function of repeated trials is introduced, whenever the uncertainty for individual trials is encoded by a belief measure (a finite random set). This generalises the traditional likelihood function, and provides a natural setting for belief inference from statistical data. Factorisation results are proven for the case in which conjunctive or disjunctive combination are employed, leading to analytical expressions for the lower and upper likelihoods of `sharp' samples in the case of Bernoulli trials, and to the formulation of a generalised logistic regression framework."
                    },
                    {
                        "title": "Uncertainty measures: The big picture",
                        "abstract": "Probability theory is far from being the most general mathematical theory of uncertainty. A number of arguments point at its inability to describe second-order ('Knightian') uncertainty. In response, a wide array of theories of uncertainty have been proposed, many of them generalisations of classical probability. As we show here, such frameworks can be organised into clusters sharing a common rationale, exhibit complex links, and are characterised by different levels of generality. Our goal is a critical appraisal of the current landscape in uncertainty theory."
                    },
                    {
                        "title": "Reasoning with random sets: An agenda for the future",
                        "abstract": "In this paper, we discuss a potential agenda for future work in the theory of random sets and belief functions, touching upon a number of focal issues: the development of a fully-fledged theory of statistical reasoning with random sets, including the generalisation of logistic regression and of the classical laws of probability; the further development of the geometric approach to uncertainty, to include general random sets, a wider range of uncertainty measures and alternative geometric representations; the application of this new theory to high-impact areas such as climate change, machine learning and statistical learning theory."
                    },
                    {
                        "title": "Consistent transformations of belief functions",
                        "abstract": "Consistent belief functions represent collections of coherent or non-contradictory pieces of evidence, but most of all they are the counterparts of consistent knowledge bases in belief calculus. The use of consistent transformations cs[.] in a reasoning process to guarantee coherence can therefore be desirable, and generalizes similar techniques in classical logic. Transformations can be obtained by minimizing an appropriate distance measure between the original belief function and the collection of consistent ones. We focus here on the case in which distances are measured using classical Lp norms, in both the \"mass space\" and the \"belief space\" representation of belief functions. While mass consistent approximations reassign the mass not focussed on a chosen element of the frame either to the whole frame or to all supersets of the element on an equal basis, approximations in the belief space do distinguish these focal elements according to the \"focussed consistent transformation\" principle. The different approximations are interpreted and compared, with the help of examples."
                    },
                    {
                        "title": "Visions of a generalized probability theory",
                        "abstract": "In this Book we argue that the fruitful interaction of computer vision and belief calculus is capable of stimulating significant advances in both fields. From a methodological point of view, novel theoretical results concerning the geometric and algebraic properties of belief functions as mathematical objects are illustrated and discussed in Part II, with a focus on both a perspective 'geometric approach' to uncertainty and an algebraic solution to the issue of conflicting evidence. In Part III we show how these theoretical developments arise from important computer vision problems (such as articulated object tracking, data association and object pose estimation) to which, in turn, the evidential formalism is able to provide interesting new solutions. Finally, some initial steps towards a generalization of the notion of total probability to belief functions are taken, in the perspective of endowing the theory of evidence with a complete battery of estimation and inference tools to the benefit of all scientists and practitioners."
                    },
                    {
                        "title": "A geometric approach to conditioning belief functions",
                        "abstract": "Conditioning is crucial in applied science when inference involving time series is involved. Belief calculus is an effective way of handling such inference in the presence of epistemic uncertainty -- unfortunately, different approaches to conditioning in the belief function framework have been proposed in the past, leaving the matter somewhat unsettled. Inspired by the geometric approach to uncertainty, in this paper we propose an approach to the conditioning of belief functions based on geometrically projecting them onto the simplex associated with the conditioning event in the space of all belief functions. We show here that such a geometric approach to conditioning often produces simple results with straightforward interpretations in terms of degrees of belief. This raises the question of whether classical approaches, such as for instance Dempster's conditioning, can also be reduced to some form of distance minimisation in a suitable space. The study of families of combination rules generated by (geometric) conditioning rules appears to be the natural prosecution of the presented research."
                    },
                    {
                        "title": "The intersection probability: betting with probability intervals",
                        "abstract": "Probability intervals are an attractive tool for reasoning under uncertainty. Unlike belief functions, though, they lack a natural probability transformation to be used for decision making in a utility theory framework. In this paper we propose the use of the intersection probability, a transform derived originally for belief functions in the framework of the geometric approach to uncertainty, as the most natural such transformation. We recall its rationale and definition, compare it with other candidate representives of systems of probability intervals, discuss its credal rationale as focus of a pair of simplices in the probability simplex, and outline a possible decision making framework for probability intervals, analogous to the Transferable Belief Model for belief functions."
                    },
                    {
                        "title": "Challenges and Opportunities for Computer Vision in Real-life Soccer Analytics",
                        "abstract": "In this paper, we explore some of the applications of computer vision to sports analytics. Sport analytics deals with understanding and discovering patterns from a corpus of sports data. Analysing such data provides important performance metrics for the players, for instance in soccer matches, that could be useful for estimating their fitness and strengths. Team level statistics can also be estimated from such analysis. This paper mainly focuses on some the challenges and opportunities presented by sport video analysis in computer vision. Specifically, we use our multi-camera setup as a framework to discuss some of the real-life challenges for machine learning algorithms."
                    },
                    {
                        "title": "Untrimmed Video Classification for Activity Detection: submission to ActivityNet Challenge",
                        "abstract": "Current state-of-the-art human activity recognition is focused on the classification of temporally trimmed videos in which only one action occurs per frame. We propose a simple, yet effective, method for the temporal detection of activities in temporally untrimmed videos with the help of untrimmed classification. Firstly, our model predicts the top k labels for each untrimmed video by analysing global video-level features. Secondly, frame-level binary classification is combined with dynamic programming to generate the temporally trimmed activity proposals. Finally, each proposal is assigned a label based on the global label, and scored with the score of the temporal activity proposal and the global score. Ultimately, we show that untrimmed video classification models can be used as stepping stone for temporal detection."
                    },
                    {
                        "title": "Spatio-Temporal Action Localization in a Weakly Supervised Setting",
                        "abstract": "Enabling computational systems with the ability to localize actions in video-based content has manifold applications. Traditionally, such a problem is approached in a fully-supervised setting where video-clips with complete frame-by-frame annotations around the actions of interest are provided for training. However, the data requirements needed to achieve adequate generalization in this setting is prohibitive. In this work, we circumvent this issue by casting the problem in a weakly supervised setting, i.e., by considering videos as labelled `sets' of unlabelled video segments. Firstly, we apply unsupervised segmentation to take advantage of the elementary structure of each video. Subsequently, a convolutional neural network is used to extract RGB features from the resulting video segments. Finally, Multiple Instance Learning (MIL) is employed to predict labels at the video segment level, thus inherently performing spatio-temporal action detection. In contrast to previous work, we make use of a different MIL formulation in which the label of each video segment is continuous rather then discrete, making the resulting optimization function tractable. Additionally, we utilize a set splitting technique for regularization. Experimental results considering multiple performance indicators on the UCF-Sports data-set support the effectiveness of our approach."
                    },
                    {
                        "title": "Spatiotemporal Deformable Scene Graphs for Complex Activity Detection",
                        "abstract": "Long-term complex activity recognition and localisation can be crucial for decision making in autonomous systems such as smart cars and surgical robots. Here we address the problem via a novel deformable, spatiotemporal scene graph approach, consisting of three main building blocks: (i) action tube detection, (ii) the modelling of the deformable geometry of parts, and (iii) a graph convolutional network. Firstly, action tubes are detected in a series of snippets. Next, a new 3D deformable RoI pooling layer is designed for learning the flexible, deformable geometry of the constituent action tubes. Finally, a scene graph is constructed by considering all parts as nodes and connecting them based on different semantics such as order of appearance, sharing the same action label and feature similarity. We also contribute fresh temporal complex activity annotation for the recently released ROAD autonomous driving and SARAS-ESAD surgical action datasets and show the adaptability of our framework to different domains. Our method is shown to significantly outperform graph-based competitors on both augmented datasets."
                    },
                    {
                        "title": "Recurrent Convolutions for Causal 3D CNNs",
                        "abstract": "Recently, three dimensional (3D) convolutional neural networks (CNNs) have emerged as dominant methods to capture spatiotemporal representations in videos, by adding to pre-existing 2D CNNs a third, temporal dimension. Such 3D CNNs, however, are anti-causal (i.e., they exploit information from both the past and the future frames to produce feature representations, thus preventing their use in online settings), constrain the temporal reasoning horizon to the size of the temporal convolution kernel, and are not temporal resolution-preserving for video sequence-to-sequence modelling, as, for instance, in action detection. To address these serious limitations, here we present a new 3D CNN architecture for the causal/online processing of videos.   Namely, we propose a novel Recurrent Convolutional Network (RCN), which relies on recurrence to capture the temporal context across frames at each network level. Our network decomposes 3D convolutions into (1) a 2D spatial convolution component, and (2) an additional hidden state $1\\times 1$ convolution, applied across time. The hidden state at any time $t$ is assumed to depend on the hidden state at $t-1$ and on the current output of the spatial convolution component. As a result, the proposed network: (i) produces causal outputs, (ii) provides flexible temporal reasoning, (iii) preserves temporal resolution. Our experiments on the large-scale large Kinetics and MultiThumos datasets show that the proposed method performs comparably to anti-causal 3D CNNs, while being causal and using fewer parameters."
                    },
                    {
                        "title": "Datamorphic Testing: A Methodology for Testing AI Applications",
                        "abstract": "With the rapid growth of the applications of machine learning (ML) and other artificial intelligence (AI) techniques, adequate testing has become a necessity to ensure their quality. This paper identifies the characteristics of AI applications that distinguish them from traditional software, and analyses the main difficulties in applying existing testing methods. Based on this analysis, we propose a new method called datamorphic testing and illustrate the method with an example of testing face recognition applications. We also report an experiment with four real industrial application systems of face recognition to validate the proposed approach."
                    },
                    {
                        "title": "Predicting Action Tubes",
                        "abstract": "In this work, we present a method to predict an entire `action tube' (a set of temporally linked bounding boxes) in a trimmed video just by observing a smaller subset of it. Predicting where an action is going to take place in the near future is essential to many computer vision based applications such as autonomous driving or surgical robotics. Importantly, it has to be done in real-time and in an online fashion. We propose a Tube Prediction network (TPnet) which jointly predicts the past, present and future bounding boxes along with their action classification scores. At test time TPnet is used in a (temporal) sliding window setting, and its predictions are put into a tube estimation framework to construct/predict the video long action tubes not only for the observed part of the video but also for the unobserved part. Additionally, the proposed action tube predictor helps in completing action tubes for unobserved segments of the video. We quantitatively demonstrate the latter ability, and the fact that TPnet improves state-of-the-art detection performance, on one of the standard action detection benchmarks - J-HMDB-21 dataset."
                    },
                    {
                        "title": "Epistemic Deep Learning",
                        "abstract": "The belief function approach to uncertainty quantification as proposed in the Demspter-Shafer theory of evidence is established upon the general mathematical models for set-valued observations, called random sets. Set-valued predictions are the most natural representations of uncertainty in machine learning. In this paper, we introduce a concept called epistemic deep learning based on the random-set interpretation of belief functions to model epistemic learning in deep neural networks. We propose a novel random-set convolutional neural network for classification that produces scores for sets of classes by learning set-valued ground truth representations. We evaluate different formulations of entropy and distance measures for belief functions as viable loss functions for these random-set networks. We also discuss methods for evaluating the quality of epistemic predictions and the performance of epistemic random-set neural networks. We demonstrate through experiments that the epistemic approach produces better performance results when compared to traditional approaches of estimating uncertainty."
                    },
                    {
                        "title": "Feature boosting with efficient attention for scene parsing",
                        "abstract": "The complexity of scene parsing grows with the number of object and scene classes, which is higher in unrestricted open scenes. The biggest challenge is to model the spatial relation between scene elements while succeeding in identifying objects at smaller scales. This paper presents a novel feature-boosting network that gathers spatial context from multiple levels of feature extraction and computes the attention weights for each level of representation to generate the final class labels. A novel `channel attention module' is designed to compute the attention weights, ensuring that features from the relevant extraction stages are boosted while the others are attenuated. The model also learns spatial context information at low resolution to preserve the abstract spatial relationships among scene elements and reduce computation cost. Spatial attention is subsequently concatenated into a final feature set before applying feature boosting. Low-resolution spatial attention features are trained using an auxiliary task that helps learning a coarse global scene structure. The proposed model outperforms all state-of-the-art models on both the ADE20K and the Cityscapes datasets."
                    },
                    {
                        "title": "Action Detection from a Robot-Car Perspective",
                        "abstract": "We present the new Road Event and Activity Detection (READ) dataset, designed and created from an autonomous vehicle perspective to take action detection challenges to autonomous driving. READ will give scholars in computer vision, smart cars and machine learning at large the opportunity to conduct research into exciting new problems such as understanding complex (road) activities, discerning the behaviour of sentient agents, and predicting both the label and the location of future actions and events, with the final goal of supporting autonomous decision making."
                    },
                    {
                        "title": "AMTnet: Action-Micro-Tube Regression by End-to-end Trainable Deep Architecture",
                        "abstract": "Dominant approaches to action detection can only provide sub-optimal solutions to the problem, as they rely on seeking frame-level detections, to later compose them into \"action tubes\" in a post-processing step. With this paper we radically depart from current practice, and take a first step towards the design and implementation of a deep network architecture able to classify and regress whole video subsets, so providing a truly optimal solution of the action detection problem. In this work, in particular, we propose a novel deep net framework able to regress and classify 3D region proposals spanning two successive video frames, whose core is an evolution of classical region proposal networks (RPNs). As such, our 3D-RPN net is able to effectively encode the temporal aspect of actions by purely exploiting appearance, as opposed to methods which heavily rely on expensive flow maps. The proposed model is end-to-end trainable and can be jointly optimised for action localisation and classification in a single step. At test time the network predicts \"micro-tubes\" encompassing two successive frames, which are linked up into complete action tubes via a new algorithm which exploits the temporal encoding learned by the network and cuts computation time by 50%. Promising results on the J-HMDB-21 and UCF-101 action detection datasets show that our model does outperform the state-of-the-art when relying purely on appearance."
                    }
                ]
            }
        }
    },
    "2405.05409": {
        "paper_data": {
            "title": "Initialization is Critical to Whether Transformers Fit Composite Functions by Inference or Memorizing",
            "url": "http://arxiv.org/abs/2405.05409v3",
            "arxiv_id": "2405.05409",
            "authors": [
                "Zhongwang Zhang",
                "Pengxiao Lin",
                "Zhiwei Wang",
                "Yaoyu Zhang",
                "Zhi-Qin John Xu"
            ],
            "abstract": "Transformers have shown impressive capabilities across various tasks, but their performance on compositional problems remains a topic of debate. In this work, we investigate the mechanisms of how transformers behave on unseen compositional tasks. We discover that the parameter initialization scale plays a critical role in determining whether the model learns inferential solutions, which capture the underlying compositional primitives, or symmetric solutions, which simply memorize mappings without understanding the compositional structure. By analyzing the information flow and vector representations within the model, we reveal the distinct mechanisms underlying these solution types. We further find that inferential solutions exhibit low complexity bias, which we hypothesize is a key factor enabling them to learn individual mappings for single anchors. We validate our conclusions on various real-world datasets. Our findings provide valuable insights into the role of initialization scale in shaping the type of solution learned by transformers and their ability to learn and generalize compositional tasks.",
            "introduction": "   1 Introduction  Large-scale transformers\u00a0(Vaswani et\u00a0al., 2017) demonstrate unprecedented capabilities\u00a0(Achiam et\u00a0al., 2023; Brown et\u00a0al., 2020; Choi et\u00a0al., 2021; Teubner et\u00a0al., 2023), even noted as \u201csparks of AGI\"\u00a0(Bubeck et\u00a0al., 2023). However, their performance on compositional tasks, which are considered a key property of human cognition\u00a0(Marcus, 2003; Smolensky et\u00a0al., 2022), has long been the subject of intense debate, especially in terms of the out-of-distribution (OOD) generalization ability. This raises critical open questions about how to faithfully interpret transformers\u2019 capabilities on compositional tasks. Can transformers learn the underlying compositional primitives within the data, or do they merely learn input-output mappings? When transformers produce incorrect outputs on compositional tasks, do their responses follow any discernible patterns or logic?   In this work, we use anchor functions\u00a0(Zhang et\u00a0al., 2024), a framework for creating controlled synthetic data, to investigate how transformers generalize to unseen compositional tasks. In our setup, each sequence contains an anchor pair, a key item, and noise items unrelated to the output (right side of Fig.\u00a01(a)). The single anchors are specific tokens (i.e., 1, 2, 3, 4), each corresponding to a particular arithmetic operation (left side of Fig.\u00a01(a)). A composite function is defined as the sequential application of two single anchor functions, i.e., the corresponding arithmetic operation, to a key item. For example, given a key item x\ud835\udc65xitalic_x and an anchor pair (1, 2), the composite function would output (x+5)+1\ud835\udc6551(x+5)+1( italic_x + 5 ) + 1, i.e., x+6\ud835\udc656x+6italic_x + 6. As shown in the middle part of Fig.\u00a01(a), during training, 14 out of the 16 possible anchor pairs are assigned inferential mappings, meaning the target output is consistent with the result of applying the composite function. One pair (3, 4) is assigned a non-inferential mapping, where the target output deviates from the result of the composite function. The remaining pair (4, 3) is held out as an unseen task. After training, there are three possible solutions the model could learn for the unseen pair (4, 3): a symmetric solution matching the non-inferential seen pair (3, 4) (Mechanism 1 in Fig.\u00a01(b)), an inferential solution (Mechanism 2 in Fig.\u00a01(b)), or other non-inferential solutions. The central question we aim to answer is: which of these three types of solutions will the model learn for the unseen anchor pair (4, 3)?   Figure 1: Experimental setup and possible solutions and mechanisms for the unseen anchor pair (4, 3). (a) Data generation: Left: The single anchors (i.e., 1, 2, 3, 4) correspond to specific arithmetic operations. Middle: During training, 14 out of the 16 possible anchor pairs are assigned inferential mappings, one pair (3, 4) is assigned a non-inferential mapping, and the remaining pair (4, 3) is held out as an unseen task (does not appear in the training). Right: The input sequences comprise an anchor pair, a key item preceding the anchor pair, and noise items unrelated to the target. The question mark indicates the output for the unseen anchor pair (4, 3), which depends on the learned solution. (b) Two potential mechanisms for the unseen anchor pair (4, 3): learning the symmetric structure (Mechanism 1) or composing the inferred single anchor mappings (Mechanism",
            "references": [
                {
                    "title": "GraphInsight: Unlocking Insights in Large Language Models for Graph Structure Understanding",
                    "abstract": "Although Large Language Models (LLMs) have demonstrated potential in processing graphs, they struggle with comprehending graphical structure information through prompts of graph description sequences, especially as the graph size increases. We attribute this challenge to the uneven memory performance of LLMs across different positions in graph description sequences, known as ''positional biases''. To address this, we propose GraphInsight, a novel framework aimed at improving LLMs' comprehension of both macro- and micro-level graphical information. GraphInsight is grounded in two key strategies: 1) placing critical graphical information in positions where LLMs exhibit stronger memory performance, and 2) investigating a lightweight external knowledge base for regions with weaker memory performance, inspired by retrieval-augmented generation (RAG). Moreover, GraphInsight explores integrating these two strategies into LLM agent processes for composite graph tasks that require multi-step reasoning. Extensive empirical studies on benchmarks with a wide range of evaluation tasks show that GraphInsight significantly outperforms all other graph description methods (e.g., prompting techniques and reordering strategies) in understanding graph structures of varying sizes."
                },
                {
                    "title": "Improving Generalization and Convergence by Enhancing Implicit Regularization",
                    "abstract": "In this work, we propose an Implicit Regularization Enhancement (IRE) framework to accelerate the discovery of flat solutions in deep learning, thereby improving generalization and convergence. Specifically, IRE decouples the dynamics of flat and sharp directions, which boosts the sharpness reduction along flat directions while maintaining the training stability in sharp directions. We show that IRE can be practically incorporated with {\\em generic base optimizers} without introducing significant computational overload. Experiments show that IRE consistently improves the generalization performance for image classification tasks across a variety of benchmark datasets (CIFAR-10/100, ImageNet) and models (ResNets and ViTs). Surprisingly, IRE also achieves a $2\\times$ {\\em speed-up} compared to AdamW in the pre-training of Llama models (of sizes ranging from 60M to 229M) on datasets including Wikitext-103, Minipile, and Openwebtext. Moreover, we provide theoretical guarantees, showing that IRE can substantially accelerate the convergence towards flat minima in Sharpness-aware Minimization (SAM)."
                },
                {
                    "title": "Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge of Generalization",
                    "abstract": "We study whether transformers can learn to implicitly reason over parametric knowledge, a skill that even the most capable language models struggle with. Focusing on two representative reasoning types, composition and comparison, we consistently find that transformers can learn implicit reasoning, but only through grokking, i.e., extended training far beyond overfitting. The levels of generalization also vary across reasoning types: when faced with out-of-distribution examples, transformers fail to systematically generalize for composition but succeed for comparison. We delve into the model's internals throughout training, conducting analytical experiments that reveal: 1) the mechanism behind grokking, such as the formation of the generalizing circuit and its relation to the relative efficiency of generalizing and memorizing circuits, and 2) the connection between systematicity and the configuration of the generalizing circuit. Our findings guide data and training setup to better induce implicit reasoning and suggest potential improvements to the transformer architecture, such as encouraging cross-layer knowledge sharing. Furthermore, we demonstrate that for a challenging reasoning task with a large search space, GPT-4-Turbo and Gemini-1.5-Pro based on non-parametric memory fail badly regardless of prompting styles or retrieval augmentation, while a fully grokked transformer can achieve near-perfect accuracy, showcasing the power of parametric memory for complex reasoning."
                },
                {
                    "title": "Case-Based or Rule-Based: How Do Transformers Do the Math?",
                    "abstract": "Despite the impressive performance in a variety of complex tasks, modern large language models (LLMs) still have trouble dealing with some math problems that are simple and intuitive for humans, such as addition. While we can easily learn basic rules of addition and apply them to new problems of any length, LLMs struggle to do the same. Instead, they may rely on similar cases seen in the training corpus for help. We define these two different reasoning mechanisms as\"rule-based reasoning\"and\"case-based reasoning\". Since rule-based reasoning is essential for acquiring systematic generalization ability, we aim to explore exactly whether transformers use rule-based or case-based reasoning for math problems. Through carefully designed intervention experiments on five math tasks, we confirm that transformers are performing case-based reasoning, no matter whether scratchpad is used, which aligns with the previous observations that transformers use subgraph matching/shortcut learning to reason. To mitigate such problems, we propose a Rule-Following Fine-Tuning (RFFT) technique to teach transformers to perform rule-based reasoning. Specifically, we provide explicit rules in the input and then instruct transformers to recite and follow the rules step by step. Through RFFT, we successfully enable LLMs fine-tuned on 1-5 digit addition to generalize to up to 12-digit addition with over 95% accuracy, which is over 40% higher than scratchpad. The significant improvement demonstrates that teaching LLMs to use rules explicitly helps them learn rule-based reasoning and generalize better in length."
                },
                {
                    "title": "Understanding the Expressive Power and Mechanisms of Transformer for Sequence Modeling",
                    "abstract": "We conduct a systematic study of the approximation properties of Transformer for sequence modeling with long, sparse and complicated memory. We investigate the mechanisms through which different components of Transformer, such as the dot-product self-attention, positional encoding and feed-forward layer, affect its expressive power, and we study their combined effects through establishing explicit approximation rates. Our study reveals the roles of critical parameters in the Transformer, such as the number of layers and the number of attention heads. These theoretical insights are validated experimentally and offer natural suggestions for alternative architectures."
                },
                {
                    "title": "Anchor function: a type of benchmark functions for studying language models",
                    "abstract": "Understanding transformer-based language models is becoming increasingly crucial, particularly as they play pivotal roles in advancing towards artificial general intelligence. However, language model research faces significant challenges, especially for academic research groups with constrained resources. These challenges include complex data structures, unknown target functions, high computational costs and memory requirements, and a lack of interpretability in the inference process, etc. Drawing a parallel to the use of simple models in scientific research, we propose the concept of an anchor function. This is a type of benchmark function designed for studying language models in learning tasks that follow an\"anchor-key\"pattern. By utilizing the concept of an anchor function, we can construct a series of functions to simulate various language tasks. The anchor function plays a role analogous to that of mice in diabetes research, particularly suitable for academic research. We demonstrate the utility of the anchor function with an example, revealing two basic operations by attention structures in language models: shifting tokens and broadcasting one token from one position to many positions. These operations are also commonly observed in large language models. The anchor function framework, therefore, opens up a series of valuable and accessible research questions for further exploration, especially for theoretical study."
                },
                {
                    "title": "Compositional Abilities Emerge Multiplicatively: Exploring Diffusion Models on a Synthetic Task",
                    "abstract": "Modern generative models exhibit unprecedented capabilities to generate extremely realistic data. However, given the inherent compositionality of the real world, reliable use of these models in practical applications requires that they exhibit the capability to compose a novel set of concepts to generate outputs not seen in the training data set. Prior work demonstrates that recent diffusion models do exhibit intriguing compositional generalization abilities, but also fail unpredictably. Motivated by this, we perform a controlled study for understanding compositional generalization in conditional diffusion models in a synthetic setting, varying different attributes of the training data and measuring the model's ability to generate samples out-of-distribution. Our results show: (i) the order in which the ability to generate samples from a concept and compose them emerges is governed by the structure of the underlying data-generating process; (ii) performance on compositional tasks exhibits a sudden\"emergence\"due to multiplicative reliance on the performance of constituent tasks, partially explaining emergent phenomena seen in generative models; and (iii) composing concepts with lower frequency in the training data to generate out-of-distribution samples requires considerably more optimization steps compared to generating in-distribution samples. Overall, our study lays a foundation for understanding capabilities and compositionality in generative models from a data-centric perspective."
                },
                {
                    "title": "Faith and Fate: Limits of Transformers on Compositionality",
                    "abstract": "Transformer large language models (LLMs) have sparked admiration for their exceptional performance on tasks that demand intricate multi-step reasoning. Yet, these models simultaneously show failures on surprisingly trivial problems. This begs the question: Are these errors incidental, or do they signal more substantial limitations? In an attempt to demystify transformer LLMs, we investigate the limits of these models across three representative compositional tasks -- multi-digit multiplication, logic grid puzzles, and a classic dynamic programming problem. These tasks require breaking problems down into sub-steps and synthesizing these steps into a precise answer. We formulate compositional tasks as computation graphs to systematically quantify the level of complexity, and break down reasoning steps into intermediate sub-procedures. Our empirical findings suggest that transformer LLMs solve compositional tasks by reducing multi-step compositional reasoning into linearized subgraph matching, without necessarily developing systematic problem-solving skills. To round off our empirical study, we provide theoretical arguments on abstract multi-step reasoning problems that highlight how autoregressive generations' performance can rapidly decay with\\,increased\\,task\\,complexity."
                },
                {
                    "title": "Stochastic Modified Equations and Dynamics of Dropout Algorithm",
                    "abstract": "Dropout is a widely utilized regularization technique in the training of neural networks, nevertheless, its underlying mechanism and its impact on achieving good generalization abilities remain poorly understood. In this work, we derive the stochastic modified equations for analyzing the dynamics of dropout, where its discrete iteration process is approximated by a class of stochastic differential equations. In order to investigate the underlying mechanism by which dropout facilitates the identification of flatter minima, we study the noise structure of the derived stochastic modified equation for dropout. By drawing upon the structural resemblance between the Hessian and covariance through several intuitive approximations, we empirically demonstrate the universal presence of the inverse variance-flatness relation and the Hessian-variance relation, throughout the training process of dropout. These theoretical and empirical findings make a substantial contribution to our understanding of the inherent tendency of dropout to locate flatter minima."
                },
                {
                    "title": "Label Words are Anchors: An Information Flow Perspective for Understanding In-Context Learning",
                    "abstract": "In-context learning (ICL) emerges as a promising capability of large language models (LLMs) by providing them with demonstration examples to perform diverse tasks. However, the underlying mechanism of how LLMs learn from the provided context remains under-explored. In this paper, we investigate the working mechanism of ICL through an information flow lens. Our findings reveal that label words in the demonstration examples function as anchors: (1) semantic information aggregates into label word representations during the shallow computation layers' processing; (2) the consolidated information in label words serves as a reference for LLMs' final predictions. Based on these insights, we introduce an anchor re-weighting method to improve ICL performance, a demonstration compression technique to expedite inference, and an analysis framework for diagnosing ICL errors in GPT2-XL. The promising applications of our findings again validate the uncovered ICL working mechanism and pave the way for future studies."
                },
                {
                    "title": "Loss Spike in Training Neural Networks",
                    "abstract": "In this work, we investigate the mechanism underlying loss spikes observed during neural network training. When the training enters a region with a lower-loss-as-sharper (LLAS) structure, the training becomes unstable, and the loss exponentially increases once the loss landscape is too sharp, resulting in the rapid ascent of the loss spike. The training stabilizes when it finds a flat region. From a frequency perspective, we explain the rapid descent in loss as being primarily influenced by low-frequency components. We observe a deviation in the first eigendirection, which can be reasonably explained by the frequency principle, as low-frequency information is captured rapidly, leading to the rapid descent. Inspired by our analysis of loss spikes, we revisit the link between the maximum eigenvalue of the loss Hessian ($\\lambda_{\\mathrm{max}}$), flatness and generalization. We suggest that $\\lambda_{\\mathrm{max}}$ is a good measure of sharpness but not a good measure for generalization. Furthermore, we experimentally observe that loss spikes can facilitate condensation, causing input weights to evolve towards the same direction. And our experiments show that there is a correlation (similar trend) between $\\lambda_{\\mathrm{max}}$ and condensation. This observation may provide valuable insights for further theoretical research on the relationship between loss spikes, $\\lambda_{\\mathrm{max}}$, and generalization."
                },
                {
                    "title": "Mimetic Initialization of Self-Attention Layers",
                    "abstract": "It is notoriously difficult to train Transformers on small datasets; typically, large pre-trained models are instead used as the starting point. We explore the weights of such pre-trained Transformers (particularly for vision) to attempt to find reasons for this discrepancy. Surprisingly, we find that simply initializing the weights of self-attention layers so that they\"look\"more like their pre-trained counterparts allows us to train vanilla Transformers faster and to higher final accuracies, particularly on vision tasks such as CIFAR-10 and ImageNet classification, where we see gains in accuracy of over 5% and 4%, respectively. Our initialization scheme is closed form, learning-free, and very simple: we set the product of the query and key weights to be approximately the identity, and the product of the value and projection weights to approximately the negative identity. As this mimics the patterns we saw in pre-trained Transformers, we call the technique\"mimetic initialization\"."
                },
                {
                    "title": "Sparks of Artificial General Intelligence: Early experiments with GPT-4",
                    "abstract": "Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. The latest model developed by OpenAI, GPT-4, was trained using an unprecedented scale of compute and data. In this paper, we report on our investigation of an early version of GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT-4 is part of a new cohort of LLMs (along with ChatGPT and Google's PaLM for example) that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4's performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4's capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system. In our exploration of GPT-4, we put special emphasis on discovering its limitations, and we discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for pursuing a new paradigm that moves beyond next-word prediction. We conclude with reflections on societal influences of the recent technological leap and future research directions."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Break It Down: Evidence for Structural Compositionality in Neural Networks",
                    "abstract": "Many tasks can be described as compositions over subroutines. Though modern neural networks have achieved impressive performance on both vision and language tasks, we know little about the functions that they implement. One possibility is that neural networks implicitly break down complex tasks into subroutines, implement modular solutions to these subroutines, and compose them into an overall solution to a task -- a property we term structural compositionality. Or they may simply learn to match new inputs to memorized representations, eliding task decomposition entirely. Here, we leverage model pruning techniques to investigate this question in both vision and language, across a variety of architectures, tasks, and pretraining regimens. Our results demonstrate that models oftentimes implement solutions to subroutines via modular subnetworks, which can be ablated while maintaining the functionality of other subroutines. This suggests that neural networks may be able to learn to exhibit compositionality, obviating the need for specialized symbolic mechanisms."
                },
                {
                    "title": "Linear Stability Hypothesis and Rank Stratification for Nonlinear Models",
                    "abstract": "Models with nonlinear architectures/parameterizations such as deep neural networks (DNNs) are well known for their mysteriously good generalization performance at overparameterization. In this work, we tackle this mystery from a novel perspective focusing on the transition of the target recovery/fitting accuracy as a function of the training data size. We propose a rank stratification for general nonlinear models to uncover a model rank as an\"effective size of parameters\"for each function in the function space of the corresponding model. Moreover, we establish a linear stability theory proving that a target function almost surely becomes linearly stable when the training data size equals its model rank. Supported by our experiments, we propose a linear stability hypothesis that linearly stable functions are preferred by nonlinear training. By these results, model rank of a target function predicts a minimal training data size for its successful recovery. Specifically for the matrix factorization model and DNNs of fully-connected or convolutional architectures, our rank stratification shows that the model rank for specific target functions can be much lower than the size of model parameters. This result predicts the target recovery capability even at heavy overparameterization for these nonlinear models as demonstrated quantitatively by our experiments. Overall, our work provides a unified framework with quantitative prediction power to understand the mysterious target recovery behavior at overparameterization for general nonlinear models."
                },
                {
                    "title": "Transformers Learn Shortcuts to Automata",
                    "abstract": "Algorithmic reasoning requires capabilities which are most naturally understood through recurrent models of computation, like the Turing machine. However, Transformer models, while lacking recurrence, are able to perform such reasoning using far fewer layers than the number of reasoning steps. This raises the question: what solutions are learned by these shallow and non-recurrent models? We find that a low-depth Transformer can represent the computations of any finite-state automaton (thus, any bounded-memory algorithm), by hierarchically reparameterizing its recurrent dynamics. Our theoretical results characterize shortcut solutions, whereby a Transformer with $o(T)$ layers can exactly replicate the computation of an automaton on an input sequence of length $T$. We find that polynomial-sized $O(\\log T)$-depth solutions always exist; furthermore, $O(1)$-depth simulators are surprisingly common, and can be understood using tools from Krohn-Rhodes theory and circuit complexity. Empirically, we perform synthetic experiments by training Transformers to simulate a wide variety of automata, and show that shortcut solutions can be learned via standard training. We further investigate the brittleness of these solutions and propose potential mitigations."
                },
                {
                    "title": "CTL++: Evaluating Generalization on Never-Seen Compositional Patterns of Known Functions, and Compatibility of Neural Representations",
                    "abstract": "Well-designed diagnostic tasks have played a key role in studying the failure of neural nets (NNs) to generalize systematically. Famous examples include SCAN and Compositional Table Lookup (CTL). Here we introduce CTL++, a new diagnostic dataset based on compositions of unary symbolic functions. While the original CTL is used to test length generalization or productivity, CTL++ is designed to test systematicity of NNs, that is, their capability to generalize to unseen compositions of known functions. CTL++ splits functions into groups and tests performance on group elements composed in a way not seen during training. We show that recent CTL-solving Transformer variants fail on CTL++. The simplicity of the task design allows for fine-grained control of task difficulty, as well as many insightful analyses. For example, we measure how much overlap between groups is needed by tested NNs for learning to compose. We also visualize how learned symbol representations in outputs of functions from different groups are compatible in case of success but not in case of failure. These results provide insights into failure cases reported on more complex compositions in the natural language domain. Our code is public."
                },
                {
                    "title": "Language Models Are Greedy Reasoners: A Systematic Formal Analysis of Chain-of-Thought",
                    "abstract": "Large language models (LLMs) have shown remarkable reasoning capabilities given chain-of-thought prompts (examples with intermediate reasoning steps). Existing benchmarks measure reasoning ability indirectly, by evaluating accuracy on downstream tasks such as mathematical reasoning. However, it is unclear how these models obtain the answers and whether they rely on simple heuristics rather than the generated chain-of-thought. To enable systematic exploration of the reasoning ability of LLMs, we present a new synthetic question-answering dataset called PrOntoQA, where each example is generated from a synthetic world model represented in first-order logic. This allows us to parse the generated chain-of-thought into symbolic proofs for formal analysis. Our analysis on InstructGPT and GPT-3 shows that LLMs are quite capable of making correct individual deduction steps, and so are generally capable of reasoning, even in fictional contexts. However, they have difficulty with proof planning: When multiple valid deduction steps are available, they are not able to systematically explore the different options."
                },
                {
                    "title": "Omnigrok: Grokking Beyond Algorithmic Data",
                    "abstract": "Grokking, the unusual phenomenon for algorithmic datasets where generalization happens long after overfitting the training data, has remained elusive. We aim to understand grokking by analyzing the loss landscapes of neural networks, identifying the mismatch between training and test losses as the cause for grokking. We refer to this as the\"LU mechanism\"because training and test losses (against model weight norm) typically resemble\"L\"and\"U\", respectively. This simple mechanism can nicely explain many aspects of grokking: data size dependence, weight decay dependence, the emergence of representations, etc. Guided by the intuitive picture, we are able to induce grokking on tasks involving images, language and molecules. In the reverse direction, we are able to eliminate grokking for algorithmic datasets. We attribute the dramatic nature of grokking for algorithmic datasets to representation learning."
                },
                {
                    "title": "Faithful Reasoning Using Large Language Models",
                    "abstract": "Although contemporary large language models (LMs) demonstrate impressive question-answering capabilities, their answers are typically the product of a single call to the model. This entails an unwelcome degree of opacity and compromises performance, especially on problems that are inherently multi-step. To address these limitations, we show how LMs can be made to perform faithful multi-step reasoning via a process whose causal structure mirrors the underlying logical structure of the problem. Our approach works by chaining together reasoning steps, where each step results from calls to two fine-tuned LMs, one for selection and one for inference, to produce a valid reasoning trace. Our method carries out a beam search through the space of reasoning traces to improve reasoning quality. We demonstrate the effectiveness of our model on multi-step logical deduction and scientific question-answering, showing that it outperforms baselines on final answer accuracy, and generates humanly interpretable reasoning traces whose validity can be checked by the user."
                },
                {
                    "title": "Implicit Regularization of Dropout",
                    "abstract": "It is important to understand how dropout, a popular regularization method, aids in achieving a good generalization solution during neural network training. In this work, we present a theoretical derivation of an implicit regularization of dropout, which is validated by a series of experiments. Additionally, we numerically study two implications of the implicit regularization, which intuitively rationalizes why dropout helps generalization. First, we find that input weights of hidden neurons tend to condense on isolated orientations trained with dropout. Condensation is a feature in the non-linear learning process, which makes the network less complex. Second, we find that the training with dropout leads to the neural network with a flatter minimum compared with standard gradient descent training, and the implicit regularization is the key to finding flat solutions. Although our theory mainly focuses on dropout used in the last hidden layer, our experiments apply to general dropout in training neural networks. This work points out a distinct characteristic of dropout compared with stochastic gradient descent and serves as an important basis for fully understanding dropout."
                },
                {
                    "title": "Emergent Abilities of Large Language Models",
                    "abstract": "Scaling up language models has been shown to predictably improve performance and sample efficiency on a wide range of downstream tasks. This paper instead discusses an unpredictable phenomenon that we refer to as emergent abilities of large language models. We consider an ability to be emergent if it is not present in smaller models but is present in larger models. Thus, emergent abilities cannot be predicted simply by extrapolating the performance of smaller models. The existence of such emergence implies that additional scaling could further expand the range of capabilities of language models."
                },
                {
                    "title": "Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models",
                    "abstract": "Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit\"breakthrough\"behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting."
                },
                {
                    "title": "Empirical Phase Diagram for Three-layer Neural Networks with Infinite Width",
                    "abstract": "Substantial work indicates that the dynamics of neural networks (NNs) is closely related to their initialization of parameters. Inspired by the phase diagram for two-layer ReLU NNs with infinite width (Luo et al., 2021), we make a step towards drawing a phase diagram for three-layer ReLU NNs with infinite width. First, we derive a normalized gradient flow for three-layer ReLU NNs and obtain two key independent quantities to distinguish different dynamical regimes for common initialization methods. With carefully designed experiments and a large computation cost, for both synthetic datasets and real datasets, we find that the dynamics of each layer also could be divided into a linear regime and a condensed regime, separated by a critical regime. The criteria is the relative change of input weights (the input weight of a hidden neuron consists of the weight from its input layer to the hidden neuron and its bias term) as the width approaches infinity during the training, which tends to $0$, $+\\infty$ and $O(1)$, respectively. In addition, we also demonstrate that different layers can lie in different dynamical regimes in a training process within a deep NN. In the condensed regime, we also observe the condensation of weights in isolated orientations with low complexity. Through experiments under three-layer condition, our phase diagram suggests a complicated dynamical regimes consisting of three possible regimes, together with their mixture, for deep NNs and provides a guidance for studying deep NNs in different initialization regimes, which reveals the possibility of completely different dynamics emerging within a deep NN for its different layers."
                },
                {
                    "title": "Selection-Inference: Exploiting Large Language Models for Interpretable Logical Reasoning",
                    "abstract": "Large language models (LLMs) have been shown to be capable of impressive few-shot generalisation to new tasks. However, they still tend to perform poorly on multi-step logical reasoning problems. Here we carry out a comprehensive evaluation of LLMs on 50 tasks that probe different aspects of logical reasoning. We show that language models tend to perform fairly well at single step inference or entailment tasks, but struggle to chain together multiple reasoning steps to solve more complex problems. In light of this, we propose a Selection-Inference (SI) framework that exploits pre-trained LLMs as general processing modules, and alternates between selection and inference to generate a series of interpretable, casual reasoning steps leading to the final answer. We show that a 7B parameter LLM used within the SI framework in a 5-shot generalisation setting, with no fine-tuning, yields a performance improvement of over 100% compared to an equivalent vanilla baseline on a suite of 10 logical reasoning tasks. The same model in the same setting even outperforms a significantly larger 280B parameter baseline on the same suite of tasks. Moreover, answers produced by the SI framework are accompanied by a causal natural-language-based reasoning trace, which has important implications for the safety and trustworthiness of the system."
                },
                {
                    "title": "Neurocompositional computing: From the Central Paradox of Cognition to a new generation of AI systems",
                    "abstract": "What explains the dramatic progress from 20th-century to 21st-century AI, and how can the remaining limitations of current AI be overcome? The widely accepted narrative attributes this progress to massive increases in the quantity of computational and data resources available to support statistical learning in deep artificial neural networks. We show that an additional crucial factor is the development of a new type of computation. Neurocompositional computing adopts two principles that must be simultaneously respected to enable human-level cognition: the principles of Compositionality and Continuity. These have seemed irreconcilable until the recent mathematical discovery that compositionality can be realized not only through discrete methods of symbolic computing, but also through novel forms of continuous neural computing. The revolutionary recent progress in AI has resulted from the use of limited forms of neurocompositional computing. New, deeper forms of neurocompositional computing create AI systems that are more robust, accurate, and comprehensible."
                },
                {
                    "title": "Do Vision-Language Pretrained Models Learn Composable Primitive Concepts?",
                    "abstract": "Vision-language (VL) pretrained models have achieved impressive performance on multimodal reasoning and zero-shot recognition tasks. Many of these VL models are pretrained on unlabeled image and caption pairs from the internet. In this paper, we study whether representations of primitive concepts--such as colors, shapes, or the attributes of object parts--emerge automatically within these pretrained VL models. We propose a two-step framework, Compositional Concept Mapping (CompMap), to investigate this. CompMap first asks a VL model to generate primitive concept activations with text prompts, and then learns to construct a composition model that maps the primitive concept activations (e.g. the likelihood of black tail or red wing) to composite concepts (e.g. a red-winged blackbird). We show that a composition model can be reliably learn from ground truth primitive concepts. We thus hypothesize that if primitive concepts indeed emerge in a VL pretrained model, its primitive concept activations can be used to learn a composition model similar to the one designed by experts. We propose a quantitative metric to measure the degree of similarity, and refer to the metric as the interpretability metric. We also measure the classification accuracy when using the primitive concept activations and the learned composition model to predict the composite concepts, and refer to it as the usefulness metric. Our study reveals that state-of-the-art VL pretrained models learn primitive concepts that are highly useful for fine-grained visual recognition on the CUB dataset, and compositional generalization tasks on the MIT-States dataset. However, we observe that the learned composition models have low interpretability in our qualitative analyses. Our results reveal the limitations of existing VL models, and the necessity of pretraining objectives that encourage the acquisition of primitive concepts."
                },
                {
                    "title": "DeepNet: Scaling Transformers to 1,000 Layers",
                    "abstract": "In this paper, we propose a simple yet effective method to stabilize extremely deep Transformers. Specifically, we introduce a new normalization function (DeepNorm) to modify the residual connection in Transformer, accompanying with theoretically derived initialization. In-depth theoretical analysis shows that model updates can be bounded in a stable way. The proposed method combines the best of two worlds, i.e., good performance of Post-LN and stable training of Pre-LN, making DeepNorm a preferred alternative. We successfully scale Transformers up to 1,000 layers (i.e., 2,500 attention and feed-forward network sublayers) without difficulty, which is one order of magnitude deeper than previous deep Transformers. Extensive experiments demonstrate that DeepNet has superior performance across various benchmarks, including machine translation, language modeling (i.e., BERT, GPT) and vision pre-training (i.e., BEiT). Remarkably, on a multilingual benchmark with 7,482 translation directions, our 200-layer model with 3.2B parameters significantly outperforms the 48-layer state-of-the-art model with 12B parameters by 5 BLEU points, which indicates a promising scaling direction."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets",
                    "abstract": "In this paper we propose to study generalization of neural networks on small algorithmically generated datasets. In this setting, questions about data efficiency, memorization, generalization, and speed of learning can be studied in great detail. In some situations we show that neural networks learn through a process of\"grokking\"a pattern in the data, improving generalization performance from random chance level to perfect generalization, and that this improvement in generalization can happen well past the point of overfitting. We also study generalization as a function of dataset size and find that smaller datasets require increasing amounts of optimization for generalization. We argue that these datasets provide a fertile ground for studying a poorly understood aspect of deep learning: generalization of overparametrized neural networks beyond memorization of the finite training dataset."
                },
                {
                    "title": "Embedding Principle: a hierarchical structure of loss landscape of deep neural networks",
                    "abstract": "We prove a general Embedding Principle of loss landscape of deep neural networks (NNs) that unravels a hierarchical structure of the loss landscape of NNs, i.e., loss landscape of an NN contains all critical points of all the narrower NNs. This result is obtained by constructing a class of critical embeddings which map any critical point of a narrower NN to a critical point of the target NN with the same output function. By discovering a wide class of general compatible critical embeddings, we provide a gross estimate of the dimension of critical submanifolds embedded from critical points of narrower NNs. We further prove an irreversiblility property of any critical embedding that the number of negative/zero/positive eigenvalues of the Hessian matrix of a critical point may increase but never decrease as an NN becomes wider through the embedding. Using a special realization of general compatible critical embedding, we prove a stringent necessary condition for being a\"truly-bad\"critical point that never becomes a strict-saddle point through any critical embedding. This result implies the commonplace of strict-saddle points in wide NNs, which may be an important reason underlying the easy optimization of wide NNs widely observed in practice."
                },
                {
                    "title": "The Neural Data Router: Adaptive Control Flow in Transformers Improves Systematic Generalization",
                    "abstract": "Despite progress across a broad range of applications, Transformers have limited success in systematic generalization. The situation is especially frustrating in the case of algorithmic tasks, where they often fail to find intuitive solutions that route relevant information to the right node/operation at the right time in the grid represented by Transformer columns. To facilitate the learning of useful control flow, we propose two modifications to the Transformer architecture, copy gate and geometric attention. Our novel Neural Data Router (NDR) achieves 100% length generalization accuracy on the classic compositional table lookup task, as well as near-perfect accuracy on the simple arithmetic task and a new variant of ListOps testing for generalization across computational depths. NDR's attention and gating patterns tend to be interpretable as an intuitive form of neural routing. Our code is public."
                },
                {
                    "title": "Embedding Principle of Loss Landscape of Deep Neural Networks",
                    "abstract": "Understanding the structure of loss landscape of deep neural networks (DNNs)is obviously important. In this work, we prove an embedding principle that the loss landscape of a DNN\"contains\"all the critical points of all the narrower DNNs. More precisely, we propose a critical embedding such that any critical point, e.g., local or global minima, of a narrower DNN can be embedded to a critical point/hyperplane of the target DNN with higher degeneracy and preserving the DNN output function. The embedding structure of critical points is independent of loss function and training data, showing a stark difference from other nonconvex problems such as protein-folding. Empirically, we find that a wide DNN is often attracted by highly-degenerate critical points that are embedded from narrow DNNs. The embedding principle provides an explanation for the general easy optimization of wide DNNs and unravels a potential implicit low-complexity regularization during the training. Overall, our work provides a skeleton for the study of loss landscape of DNNs and its implication, by which a more exact and comprehensive understanding can be anticipated in the near"
                },
                {
                    "title": "Towards Understanding the Condensation of Neural Networks at Initial Training",
                    "abstract": "Empirical works show that for ReLU neural networks (NNs) with small initialization, input weights of hidden neurons (the input weight of a hidden neuron consists of the weight from its input layer to the hidden neuron and its bias term) condense on isolated orientations. The condensation dynamics implies that the training implicitly regularizes a NN towards one with a much smaller effective size. In this work, we illustrate the formation of the condensation in multi-layer fully connected NNs and show that the maximal number of condensed orientations in the initial training stage is twice the multiplicity of the activation function, where\"multiplicity\"indicates the multiple roots of activation function at origin. Our theoretical analysis confirms experiments for two cases, one is for the activation function of multiplicity one with arbitrary dimension input, which contains many common activation functions, and the other is for the layer with one-dimensional input and arbitrary multiplicity. This work makes a step towards understanding how small initialization leads NNs to condensation at the initial training stage."
                },
                {
                    "title": "GradInit: Learning to Initialize Neural Networks for Stable and Efficient Training",
                    "abstract": "Innovations in neural architectures have fostered significant breakthroughs in language modeling and computer vision. Unfortunately, novel architectures often result in challenging hyper-parameter choices and training instability if the network parameters are not properly initialized. A number of architecture-specific initialization schemes have been proposed, but these schemes are not always portable to new architectures. This paper presents GradInit, an automated and architecture agnostic method for initializing neural networks. GradInit is based on a simple heuristic; the norm of each network layer is adjusted so that a single step of SGD or Adam with prescribed hyperparameters results in the smallest possible loss value. This adjustment is done by introducing a scalar multiplier variable in front of each parameter block, and then optimizing these variables using a simple numerical scheme. GradInit accelerates the convergence and test performance of many convolutional architectures, both with or without skip connections, and even without normalization layers. It also improves the stability of the original Transformer architecture for machine translation, enabling training it without learning rate warmup using either Adam or SGD under a wide range of learning rates and momentum coefficients. Code is available at https://github.com/zhuchen03/gradinit."
                },
                {
                    "title": "COGS: A Compositional Generalization Challenge Based on Semantic Interpretation",
                    "abstract": "Natural language is characterized by compositionality: the meaning of a complex expression is constructed from the meanings of its constituent parts. To facilitate the evaluation of the compositional abilities of language processing architectures, we introduce COGS, a semantic parsing dataset based on a fragment of English. The evaluation portion of COGS contains multiple systematic gaps that can only be addressed by compositional generalization; these include new combinations of familiar syntactic structures, or new combinations of familiar words and familiar structures. In experiments with Transformers and LSTMs, we found that in-distribution accuracy on the COGS test set was near-perfect (96--99%), but generalization accuracy was substantially lower (16--35%) and showed high sensitivity to random seed ($\\pm$6--8%). These findings indicate that contemporary standard NLP models are limited in their compositional generalization capacity, and position COGS as a good way to measure progress."
                },
                {
                    "title": "Phase diagram for two-layer ReLU neural networks at infinite-width limit",
                    "abstract": "How neural network behaves during the training over different choices of hyperparameters is an important question in the study of neural networks. However, except for specific examples with particular choices of hyperparameters, e.g., neural tangent kernel (NTK), mean-field model, this question is largely unanswered. In this work, inspired by the phase diagram in statistical mechanics, we draw the phase diagram for the two-layer ReLU neural network at the infinite-width limit for a complete characterization of its dynamical regimes and their dependence on hyperparameters. Through both experimental and theoretical approaches, we identify three regimes in the phase diagram, i.e., linear regime, critical regime and condensed regime, based on the relative change of input weights as the width approaches infinity, which tends to $0$, $O(1)$ and $+\\infty$, respectively. In the linear regime, NN training dynamics is approximately linear similar to a random feature model with an exponential loss decay. In the condensed regime, we demonstrate through experiments that active neurons are condensed at several discrete orientations. The critical regime serves as the boundary between above two regimes, which exhibits an intermediate nonlinear behavior with the mean-field model as a typical example. Overall, our phase diagram for the two-layer ReLU NN serves as a map for the future studies and is a first step towards a more systematical investigation of the training behavior and the implicit regularization of NNs of different structures."
                },
                {
                    "title": "Improving Transformer Optimization Through Better Initialization",
                    "abstract": "The Transformer architecture has achieved considerable success recently; the key component of the Transformer is the attention layer that enables the model to focus on important regions within an input sequence. Gradient optimization with attention layers can be notoriously dif\ufb01cult requiring tricks such as learning rate warmup to prevent divergence. As Transformer models are becoming larger and more expensive to train, recent research has focused on understanding and improving optimization in these architectures. In this work our contributions are two-fold: we \ufb01rst investigate and empirically validate the source of optimization problems in the encoder-decoder Transformer architecture; we then propose a new weight initialization scheme with theoretical justi-\ufb01cation, that enables training without warmup or layer normalization. Empirical results on public machine translation benchmarks show that our approach achieves leading accuracy, allowing to train deep Transformer models with 200 layers in both encoder and decoder (over 1000 attention/MLP blocks) without dif\ufb01culty. Code for this work is available here: https://github. com/layer6ai-labs/T-Fixup ."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Understanding the Difficulty of Training Transformers",
                    "abstract": "Transformers have been proved effective for many deep learning tasks. Training transformers, however, requires non-trivial efforts regarding carefully designing learning rate schedulers and cutting-edge optimizers (the standard SGD fails to train Transformers effectively). In this paper, we study Transformer training from both theoretical and empirical perspectives. Our analysis reveals that unbalanced gradients are not the root cause of the instability of training. Instead, we identify an amplification effect that substantially influences training. Specifically, we observe that for each layer in a multi-layer Transformer model, heavy dependency on its residual branch makes training unstable since it amplifies small parameter perturbations (e.g., parameter updates) and result in significant disturbances in the model output, yet a light dependency limits the potential of model training and can lead to an inferior trained model. Inspired by our analysis, we propose Admin ($\\mathbf{Ad}$aptive $\\mathbf{m}$odel $\\mathbf{in}$itialization) to stabilize the training in the early stage and unleash its full potential in the late stage. Extensive experiments show that Admin is more stable, converges faster, and leads to better performance."
                },
                {
                    "title": "Improving Deep Transformer with Depth-Scaled Initialization and Merged Attention",
                    "abstract": "The general trend in NLP is towards increasing model capacity and performance via deeper neural networks. However, simply stacking more layers of the popular Transformer architecture for machine translation results in poor convergence and high computational overhead. Our empirical analysis suggests that convergence is poor due to gradient vanishing caused by the interaction between residual connection and layer normalization. We propose depth-scaled initialization (DS-Init), which decreases parameter variance at the initialization stage, and reduces output variance of residual connections so as to ease gradient back-propagation through normalization layers. To address computational cost, we propose a merged attention sublayer (MAtt) which combines a simplified average-based self-attention sublayer and the encoder-decoder attention sublayer on the decoder side. Results on WMT and IWSLT translation tasks with five translation directions show that deep Transformers with DS-Init and MAtt can substantially outperform their base counterpart in terms of BLEU (+1.1 BLEU on average for 12-layer models), while matching the decoding speed of the baseline model thanks to the efficiency improvements of MAtt. Source code for reproduction will be released soon."
                },
                {
                    "title": "Gradient Dynamics of Shallow Univariate ReLU Networks",
                    "abstract": "We present a theoretical and empirical study of the gradient dynamics of overparameterized shallow ReLU networks with one-dimensional input, solving least-squares interpolation. We show that the gradient dynamics of such networks are determined by the gradient flow in a non-redundant parameterization of the network function. We examine the principal qualitative features of this gradient flow. In particular, we determine conditions for two learning regimes:kernel and adaptive, which depend both on the relative magnitude of initialization of weights in different layers and the asymptotic behavior of initialization coefficients in the limit of large network widths. We show that learning in the kernel regime yields smooth interpolants, minimizing curvature, and reduces to cubic splines for uniform initializations. Learning in the adaptive regime favors instead linear splines, where knots cluster adaptively at the sample points."
                },
                {
                    "title": "A type of generalization error induced by initialization in deep neural networks",
                    "abstract": "How different initializations and loss functions affect the learning of a deep neural network (DNN), specifically its generalization error, is an important problem in practice. In this work, focusing on regression problems, we develop a kernel-norm minimization framework for the analysis of DNNs in the kernel regime in which the number of neurons in each hidden layer is sufficiently large (Jacot et al. 2018, Lee et al. 2019). We find that, in the kernel regime, for any loss in a general class of functions, e.g., any Lp loss for $1 < p < \\infty$, the DNN finds the same global minima-the one that is nearest to the initial value in the parameter space, or equivalently, the one that is closest to the initial DNN output in the corresponding reproducing kernel Hilbert space. With this framework, we prove that a non-zero initial output increases the generalization error of DNN. We further propose an antisymmetrical initialization (ASI) trick that eliminates this type of error and accelerates the training. We also demonstrate experimentally that even for DNNs in the non-kernel regime, our theoretical analysis and the ASI trick remain effective. Overall, our work provides insight into how initialization and loss function quantitatively affect the generalization of DNNs, and also provides guidance for the training of DNNs."
                },
                {
                    "title": "On Exact Computation with an Infinitely Wide Neural Net",
                    "abstract": "How well does a classic deep net architecture like AlexNet or VGG19 classify on a standard dataset such as CIFAR-10 when its width --- namely, number of channels in convolutional layers, and number of nodes in fully-connected internal layers --- is allowed to increase to infinity? Such questions have come to the forefront in the quest to theoretically understand deep learning and its mysteries about optimization and generalization. They also connect deep learning to notions such as Gaussian processes and kernels. A recent paper [Jacot et al., 2018] introduced the Neural Tangent Kernel (NTK) which captures the behavior of fully-connected deep nets in the infinite width limit trained by gradient descent; this object was implicit in some other recent papers. An attraction of such ideas is that a pure kernel-based method is used to capture the power of a fully-trained deep net of infinite width. \nThe current paper gives the first efficient exact algorithm for computing the extension of NTK to convolutional neural nets, which we call Convolutional NTK (CNTK), as well as an efficient GPU implementation of this algorithm. This results in a significant new benchmark for the performance of a pure kernel-based method on CIFAR-10, being $10\\%$ higher than the methods reported in [Novak et al., 2019], and only $6\\%$ lower than the performance of the corresponding finite deep net architecture (once batch normalization, etc. are turned off). Theoretically, we also give the first non-asymptotic proof showing that a fully-trained sufficiently wide net is indeed equivalent to the kernel regression predictor using NTK."
                },
                {
                    "title": "Frequency Principle: Fourier Analysis Sheds Light on Deep Neural Networks",
                    "abstract": "We study the training process of Deep Neural Networks (DNNs) from the Fourier analysis perspective. We demonstrate a very universal Frequency Principle (F-Principle) --- DNNs often fit target functions from low to high frequencies --- on high-dimensional benchmark datasets such as MNIST/CIFAR10 and deep neural networks such as VGG16. This F-Principle of DNNs is opposite to the behavior of most conventional iterative numerical schemes (e.g., Jacobi method), which exhibit faster convergence for higher frequencies for various scientific computing problems. With a simple theory, we illustrate that this F-Principle results from the regularity of the commonly used activation functions. The F-Principle implies an implicit bias that DNNs tend to fit training data by a low-frequency function. This understanding provides an explanation of good generalization of DNNs on most real datasets and bad generalization of DNNs on parity function or randomized dataset."
                },
                {
                    "title": "On the Spectral Bias of Deep Neural Networks",
                    "abstract": "It is well known that over-parametrized deep neural networks (DNNs) are an overly expressive class of functions that can memorize even random data with $100\\%$ training accuracy. This raises the question why they do not easily overfit real data. To answer this question, we study deep networks using Fourier analysis. We show that deep networks with finite weights (or trained for finite number of steps) are inherently biased towards representing smooth functions over the input space. Specifically, the magnitude of a particular frequency component ($k$) of deep ReLU network function decays at least as fast as $\\mathcal{O}(k^{-2})$, with width and depth helping polynomially and exponentially (respectively) in modeling higher frequencies. This shows for instance why DNNs cannot perfectly \\textit{memorize} peaky delta-like functions. We also show that DNNs can exploit the geometry of low dimensional data manifolds to approximate complex functions that exist along the manifold with simple functions when seen with respect to the input space. As a consequence, we find that all samples (including adversarial samples) classified by a network to belong to a certain class are connected by a path such that the prediction of the network along that path does not change. Finally we find that DNN parameters corresponding to functions with higher frequency components occupy a smaller volume in the parameter."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "On the Global Convergence of Gradient Descent for Over-parameterized Models using Optimal Transport",
                    "abstract": "Many tasks in machine learning and signal processing can be solved by minimizing a convex function of a measure. This includes sparse spikes deconvolution or training a neural network with a single hidden layer. For these problems, we study a simple minimization method: the unknown measure is discretized into a mixture of particles and a continuous-time gradient descent is performed on their weights and positions. This is an idealization of the usual way to train neural networks with a large hidden layer. We show that, when initialized correctly and in the many-particle limit, this gradient flow, although non-convex, converges to global minimizers. The proof involves Wasserstein gradient flows, a by-product of optimal transport theory. Numerical experiments show that this asymptotic behavior is already at play for a reasonable number of particles, even in high dimension."
                },
                {
                    "title": "Learning compositionally through attentive guidance",
                    "abstract": "While neural network models have been successfully applied to domains that require substantial generalisation skills, recent studies have implied that they struggle when solving the task they are trained on requires inferring its underlying compositional structure. In this paper, we introduce Attentive Guidance, a mechanism to direct a sequence to sequence model equipped with attention to find more compositional solutions. We test it on two tasks, devised precisely to assess the compositional capabilities of neural models, and we show that vanilla sequence to sequence models with attention overfit the training distribution, while the guided versions come up with compositional solutions that fit the training and testing distributions almost equally well. Moreover, the learned solutions generalise even in cases where the training and testing distributions strongly diverge. In this way, we demonstrate that sequence to sequence models are capable of finding compositional solutions without requiring extra components. These results helps to disentangle the causes for the lack of systematic compositionality in neural networks, which can in turn fuel future work."
                },
                {
                    "title": "A mean field view of the landscape of two-layer neural networks",
                    "abstract": "Significance Multilayer neural networks have proven extremely successful in a variety of tasks, from image classification to robotics. However, the reasons for this practical success and its precise domain of applicability are unknown. Learning a neural network from data requires solving a complex optimization problem with millions of variables. This is done by stochastic gradient descent (SGD) algorithms. We study the case of two-layer networks and derive a compact description of the SGD dynamics in terms of a limiting partial differential equation. Among other consequences, this shows that SGD dynamics does not become more complex when the network size increases. Multilayer neural networks are among the most powerful models in machine learning, yet the fundamental reasons for this success defy mathematical understanding. Learning a neural network requires optimizing a nonconvex high-dimensional objective (risk function), a problem that is usually attacked using stochastic gradient descent (SGD). Does SGD converge to a global optimum of the risk or only to a local optimum? In the former case, does this happen because local minima are absent or because SGD somehow avoids them? In the latter, why do local minima reached by SGD have good generalization properties? In this paper, we consider a simple case, namely two-layer neural networks, and prove that\u2014in a suitable scaling limit\u2014SGD dynamics is captured by a certain nonlinear partial differential equation (PDE) that we call distributional dynamics (DD). We then consider several specific examples and show how DD can be used to prove convergence of SGD to networks with nearly ideal generalization error. This description allows for \u201caveraging out\u201d some of the complexities of the landscape of neural networks and can be used to prove a general convergence result for noisy SGD."
                },
                {
                    "title": "Generalization without Systematicity: On the Compositional Skills of Sequence-to-Sequence Recurrent Networks",
                    "abstract": "Humans can understand and produce new utterances effortlessly, thanks to their compositional skills. Once a person learns the meaning of a new verb \"dax,\" he or she can immediately understand the meaning of \"dax twice\" or \"sing and dax.\" In this paper, we introduce the SCAN domain, consisting of a set of simple compositional navigation commands paired with the corresponding action sequences. We then test the zero-shot generalization capabilities of a variety of recurrent neural networks (RNNs) trained on SCAN with sequence-to-sequence methods. We find that RNNs can make successful zero-shot generalizations when the differences between training and test commands are small, so that they can apply \"mix-and-match\" strategies to solve the task. However, when generalization requires systematic compositional skills (as in the \"dax\" example above), RNNs fail spectacularly. We conclude with a proof-of-concept experiment in neural machine translation, suggesting that lack of systematicity might be partially responsible for neural networks' notorious training data thirst."
                },
                {
                    "title": "A Closer Look at Memorization in Deep Networks",
                    "abstract": "We examine the role of memorization in deep learning, drawing connections to capacity, generalization, and adversarial robustness. While deep networks are capable of memorizing noise data, our results suggest that they tend to prioritize learning simple patterns first. In our experiments, we expose qualitative differences in gradient-based optimization of deep neural networks (DNNs) on noise vs. real data. We also demonstrate that for appropriately tuned explicit regularization (e.g., dropout) we can degrade DNN training performance on noise datasets without compromising generalization on real data. Our analysis suggests that the notions of effective capacity which are dataset independent are unlikely to explain the generalization performance of deep networks when trained with gradient based methods because training data itself plays an important role in determining the degree of memorization."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "ChatGPT Goes to Law School",
                    "abstract": "How well can AI models write law school exams without human assistance? To find out"
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Parameters as interacting particles: long time convergence and asymptotic error scaling of neural networks",
                    "abstract": "The performance of neural networks on high-dimensional data distributions suggests that it may be possible to parameterize a representation of a given high-dimensional function with controllably small errors, potentially outperforming standard interpolation methods. We demonstrate, both theoretically and numerically, that this is indeed the case. We map the parameters of a neural network to a system of particles relaxing with an interaction potential determined by the loss function. We show that in the limit that the number of parameters $n$ is large, the landscape of the mean-squared error becomes convex and the representation error in the function scales as $O(n^{-1})$. In this limit, we prove a dynamical variant of the universal approximation theorem showing that the optimal representation can be attained by stochastic gradient descent, the algorithm ubiquitously used for parameter optimization in machine learning. In the asymptotic regime, we study the fluctuations around the optimal representation and show that they arise at a scale $O(n^{-1})$. These fluctuations in the landscape identify the natural scale for the noise in stochastic gradient descent. Our results apply to both single and multi-layer neural networks, as well as standard kernel methods like radial basis functions."
                },
                {
                    "title": "The Algebraic Mind: Integrating Connectionism and Cognitive Science",
                    "abstract": "In The Algebraic Mind, Gary Marcus attempts to integrate two theories about how the mind works, one that says that the mind is a computer-like manipulator of symbols, and another that says that the mind is a large network of neurons working together in parallel. Resisting the conventional wisdom that says that if the mind is a large neural network it cannot simultaneously be a manipulator of symbols, Marcus outlines a variety of ways in which neural systems could be organized so as to manipulate symbols, and he shows why such systems are more likely to provide an adequate substrate for language and cognition than neural systems that are inconsistent with the manipulation of symbols. Concluding with a discussion of how a neurally realized system of symbol-manipulation could have evolved and how such a system could unfold developmentally within the womb, Marcus helps to set the future agenda of cognitive neuroscience."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow do transformers generalize to unseen compositional tasks, specifically regarding their ability to learn underlying compositional primitives versus merely input-output mappings?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for understanding the cognitive capabilities of large-scale transformers, which are often viewed as precursors to artificial general intelligence (AGI). By investigating their performance on compositional tasks, we can gain insights into their reasoning abilities and limitations. This research could influence future studies on model interpretability, robustness, and generalization, ultimately leading to the development of more advanced AI systems that can better mimic human-like reasoning and problem-solving skills.\n\n### [Question 3] - Why is it hard?\nThe challenges in addressing this problem stem from the complexity of compositional tasks and the nuanced nature of generalization. Naive approaches may fail because they do not account for the intricate relationships between anchor pairs and their corresponding operations, leading to oversimplified models that cannot capture the underlying logic. Additionally, the presence of noise items and the need to differentiate between inferential and non-inferential mappings complicate the learning process. Overcoming these technical obstacles requires a deep understanding of both the model architecture and the data generation process.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on general performance metrics without delving into the specific mechanisms of compositional reasoning in transformers. Limitations in existing methodologies, such as the lack of controlled synthetic data to isolate compositional tasks, have hindered progress. Additionally, many studies have not adequately addressed the distinction between inferential and non-inferential mappings, leading to a gap in understanding how transformers learn from complex data structures. Our approach, utilizing anchor functions to create controlled datasets, aims to fill this gap and provide clearer insights into the learning mechanisms of transformers.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves using anchor functions to generate controlled synthetic data for training transformers on compositional tasks. We will create sequences that include anchor pairs, key items, and noise items, with specific mappings assigned to each pair. The dataset will consist of various combinations of anchor pairs, with one pair held out as an unseen task. We will evaluate the model's performance based on its ability to produce correct outputs for the unseen pair (4, 3) and analyze the learned solutions. The expected outcomes include identifying whether the model learns a symmetric solution, an inferential"
            }
        },
        "author_data": {
            "65c29d95-86ac-4584-8fdc-926d4220f37b": {
                "pk": "65c29d95-86ac-4584-8fdc-926d4220f37b",
                "name": "Zhongwang Zhang",
                "collaborators": [
                    "Zhi-Qin John Xu",
                    "Yaoyu Zhang",
                    "Tao Luo",
                    "Leyang Zhang",
                    "Zhiwei Bai",
                    "Zhiwei Wang",
                    "Xiaolong Li",
                    "Yuqing Li",
                    "Zhangchen Zhou",
                    "Hanxu Zhou"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Regularization",
                    "Language Models"
                ],
                "publications": [
                    {
                        "title": "Implicit regularization of dropout",
                        "abstract": "It is important to understand how dropout, a popular regularization method, aids in achieving a good generalization solution during neural network training. In this work, we present a theoretical derivation of an implicit regularization of dropout, which is validated by a series of experiments. Additionally, we numerically study two implications of the implicit regularization, which intuitively rationalizes why dropout helps generalization. Firstly, we find that input weights of hidden neurons tend to condense on isolated orientations trained with dropout. Condensation is a feature in the non-linear learning process, which makes the network less complex. Secondly, we experimentally find that the training with dropout leads to the neural network with a flatter minimum compared with standard gradient descent training, and the implicit regularization is the key to finding flat solutions. Although our theory mainly focuses on dropout used in the last hidden layer, our experiments apply to general dropout in training neural networks. This work points out a distinct characteristic of dropout compared with stochastic gradient descent and serves as an important basis for fully understanding dropout."
                    },
                    {
                        "title": "Local Linear Recovery Guarantee of Deep Neural Networks at Overparameterization",
                        "abstract": "Determining whether deep neural network (DNN) models can reliably recover target functions at overparameterization is a critical yet complex issue in the theory of deep learning. To advance understanding in this area, we introduce a concept we term \"local linear recovery\" (LLR), a weaker form of target function recovery that renders the problem more amenable to theoretical analysis. In the sense of LLR, we prove that functions expressible by narrower DNNs are guaranteed to be recoverable from fewer samples than model parameters. Specifically, we establish upper limits on the optimistic sample sizes, defined as the smallest sample size necessary to guarantee LLR, for functions in the space of a given DNN. Furthermore, we prove that these upper bounds are achieved in the case of two-layer tanh neural networks. Our research lays a solid groundwork for future investigations into the recovery capabilities of DNNs in overparameterized scenarios."
                    },
                    {
                        "title": "Dropout in Training Neural Networks: Flatness of Solution and Noise Structure",
                        "abstract": "It is important to understand how the popular regularization method dropout helps the neural network training find a good generalization solution. In this work, we show that the training with dropout finds the neural network with a flatter minimum compared with standard gradient descent training. We further find that the variance of a noise induced by the dropout is larger at the sharper direction of the loss landscape and the Hessian of the loss landscape at the found minima aligns with the noise covariance matrix by experiments on various datasets, i.e., MNIST, CIFAR-10, CIFAR-100 and Multi30k, and various structures, i.e., fully-connected networks, large residual convolutional networks and transformer. For networks with piece-wise linear activation function and the dropout is only at the last hidden layer, we then theoretically derive the Hessian and the covariance of dropout randomness, where these two quantities are very similar. This similarity may be a key reason accounting for the goodness of dropout."
                    },
                    {
                        "title": "Loss Spike in Training Neural Networks",
                        "abstract": "In this work, we investigate the mechanism underlying loss spikes observed during neural network training. When the training enters a region with a lower-loss-as-sharper (LLAS) structure, the training becomes unstable, and the loss exponentially increases once the loss landscape is too sharp, resulting in the rapid ascent of the loss spike. The training stabilizes when it finds a flat region. From a frequency perspective, we explain the rapid descent in loss as being primarily influenced by low-frequency components. We observe a deviation in the first eigendirection, which can be reasonably explained by the frequency principle, as low-frequency information is captured rapidly, leading to the rapid descent. Inspired by our analysis of loss spikes, we revisit the link between the maximum eigenvalue of the loss Hessian ($\\lambda_{\\mathrm{max}}$), flatness and generalization. We suggest that $\\lambda_{\\mathrm{max}}$ is a good measure of sharpness but not a good measure for generalization. Furthermore, we experimentally observe that loss spikes can facilitate condensation, causing input weights to evolve towards the same direction. And our experiments show that there is a correlation (similar trend) between $\\lambda_{\\mathrm{max}}$ and condensation. This observation may provide valuable insights for further theoretical research on the relationship between loss spikes, $\\lambda_{\\mathrm{max}}$, and generalization."
                    },
                    {
                        "title": "Embedding Principle of Loss Landscape of Deep Neural Networks",
                        "abstract": "Understanding the structure of loss landscape of deep neural networks (DNNs)is obviously important. In this work, we prove an embedding principle that the loss landscape of a DNN \"contains\" all the critical points of all the narrower DNNs. More precisely, we propose a critical embedding such that any critical point, e.g., local or global minima, of a narrower DNN can be embedded to a critical point/hyperplane of the target DNN with higher degeneracy and preserving the DNN output function. The embedding structure of critical points is independent of loss function and training data, showing a stark difference from other nonconvex problems such as protein-folding. Empirically, we find that a wide DNN is often attracted by highly-degenerate critical points that are embedded from narrow DNNs. The embedding principle provides an explanation for the general easy optimization of wide DNNs and unravels a potential implicit low-complexity regularization during the training. Overall, our work provides a skeleton for the study of loss landscape of DNNs and its implication, by which a more exact and comprehensive understanding can be anticipated in the near"
                    },
                    {
                        "title": "Stochastic Modified Equations and Dynamics of Dropout Algorithm",
                        "abstract": "Dropout is a widely utilized regularization technique in the training of neural networks, nevertheless, its underlying mechanism and its impact on achieving good generalization abilities remain poorly understood. In this work, we derive the stochastic modified equations for analyzing the dynamics of dropout, where its discrete iteration process is approximated by a class of stochastic differential equations. In order to investigate the underlying mechanism by which dropout facilitates the identification of flatter minima, we study the noise structure of the derived stochastic modified equation for dropout. By drawing upon the structural resemblance between the Hessian and covariance through several intuitive approximations, we empirically demonstrate the universal presence of the inverse variance-flatness relation and the Hessian-variance relation, throughout the training process of dropout. These theoretical and empirical findings make a substantial contribution to our understanding of the inherent tendency of dropout to locate flatter minima."
                    },
                    {
                        "title": "Embedding Principle: a hierarchical structure of loss landscape of deep neural networks",
                        "abstract": "We prove a general Embedding Principle of loss landscape of deep neural networks (NNs) that unravels a hierarchical structure of the loss landscape of NNs, i.e., loss landscape of an NN contains all critical points of all the narrower NNs. This result is obtained by constructing a class of critical embeddings which map any critical point of a narrower NN to a critical point of the target NN with the same output function. By discovering a wide class of general compatible critical embeddings, we provide a gross estimate of the dimension of critical submanifolds embedded from critical points of narrower NNs. We further prove an irreversiblility property of any critical embedding that the number of negative/zero/positive eigenvalues of the Hessian matrix of a critical point may increase but never decrease as an NN becomes wider through the embedding. Using a special realization of general compatible critical embedding, we prove a stringent necessary condition for being a \"truly-bad\" critical point that never becomes a strict-saddle point through any critical embedding. This result implies the commonplace of strict-saddle points in wide NNs, which may be an important reason underlying the easy optimization of wide NNs widely observed in practice."
                    },
                    {
                        "title": "Loss Jump During Loss Switch in Solving PDEs with Neural Networks",
                        "abstract": "Using neural networks to solve partial differential equations (PDEs) is gaining popularity as an alternative approach in the scientific computing community. Neural networks can integrate different types of information into the loss function. These include observation data, governing equations, and variational forms, etc. These loss functions can be broadly categorized into two types: observation data loss directly constrains and measures the model output, while other loss functions indirectly model the performance of the network, which can be classified as model loss. However, this alternative approach lacks a thorough understanding of its underlying mechanisms, including theoretical foundations and rigorous characterization of various phenomena. This work focuses on investigating how different loss functions impact the training of neural networks for solving PDEs. We discover a stable loss-jump phenomenon: when switching the loss function from the data loss to the model loss, which includes different orders of derivative information, the neural network solution significantly deviates from the exact solution immediately. Further experiments reveal that this phenomenon arises from the different frequency preferences of neural networks under different loss functions. We theoretically analyze the frequency preference of neural networks under model loss. This loss-jump phenomenon provides a valuable perspective for examining the underlying mechanisms of neural networks in solving PDEs."
                    },
                    {
                        "title": "Optimistic Estimate Uncovers the Potential of Nonlinear Models",
                        "abstract": "We propose an optimistic estimate to evaluate the best possible fitting performance of nonlinear models. It yields an optimistic sample size that quantifies the smallest possible sample size to fit/recover a target function using a nonlinear model. We estimate the optimistic sample sizes for matrix factorization models, deep models, and deep neural networks (DNNs) with fully-connected or convolutional architecture. For each nonlinear model, our estimates predict a specific subset of targets that can be fitted at overparameterization, which are confirmed by our experiments. Our optimistic estimate reveals two special properties of the DNN models -- free expressiveness in width and costly expressiveness in connection. These properties suggest the following architecture design principles of DNNs: (i) feel free to add neurons/kernels; (ii) restrain from connecting neurons. Overall, our optimistic estimate theoretically unveils the vast potential of nonlinear models in fitting at overparameterization. Based on this framework, we anticipate gaining a deeper understanding of how and why numerous nonlinear models such as DNNs can effectively realize their potential in practice in the near future."
                    },
                    {
                        "title": "Linear Stability Hypothesis and Rank Stratification for Nonlinear Models",
                        "abstract": "Models with nonlinear architectures/parameterizations such as deep neural networks (DNNs) are well known for their mysteriously good generalization performance at overparameterization. In this work, we tackle this mystery from a novel perspective focusing on the transition of the target recovery/fitting accuracy as a function of the training data size. We propose a rank stratification for general nonlinear models to uncover a model rank as an \"effective size of parameters\" for each function in the function space of the corresponding model. Moreover, we establish a linear stability theory proving that a target function almost surely becomes linearly stable when the training data size equals its model rank. Supported by our experiments, we propose a linear stability hypothesis that linearly stable functions are preferred by nonlinear training. By these results, model rank of a target function predicts a minimal training data size for its successful recovery. Specifically for the matrix factorization model and DNNs of fully-connected or convolutional architectures, our rank stratification shows that the model rank for specific target functions can be much lower than the size of model parameters. This result predicts the target recovery capability even at heavy overparameterization for these nonlinear models as demonstrated quantitatively by our experiments. Overall, our work provides a unified framework with quantitative prediction power to understand the mysterious target recovery behavior at overparameterization for general nonlinear models."
                    },
                    {
                        "title": "Anchor function: a type of benchmark functions for studying language models",
                        "abstract": "Understanding transformer-based language models is becoming increasingly crucial, particularly as they play pivotal roles in advancing towards artificial general intelligence. However, language model research faces significant challenges, especially for academic research groups with constrained resources. These challenges include complex data structures, unknown target functions, high computational costs and memory requirements, and a lack of interpretability in the inference process, etc. Drawing a parallel to the use of simple models in scientific research, we propose the concept of an anchor function. This is a type of benchmark function designed for studying language models in learning tasks that follow an \"anchor-key\" pattern. By utilizing the concept of an anchor function, we can construct a series of functions to simulate various language tasks. The anchor function plays a role analogous to that of mice in diabetes research, particularly suitable for academic research. We demonstrate the utility of the anchor function with an example, revealing two basic operations by attention structures in language models: shifting tokens and broadcasting one token from one position to many positions. These operations are also commonly observed in large language models. The anchor function framework, therefore, opens up a series of valuable and accessible research questions for further exploration, especially for theoretical study."
                    },
                    {
                        "title": "The Buffer Mechanism for Multi-Step Information Reasoning in Language Models",
                        "abstract": "Large language models have consistently struggled with complex reasoning tasks, such as mathematical problem-solving. Investigating the internal reasoning mechanisms of these models can help us design better model architectures and training strategies, ultimately enhancing their reasoning capability. In this study, we constructed a symbolic dataset to investigate the mechanisms by which Transformer models employ vertical thinking strategy based on their inherent structure and horizontal thinking strategy based on Chain of Thought to achieve multi-step reasoning. We introduced the concept of buffer mechanism: the model stores various information in distinct buffers and selectively extracts them through the query-key matrix. We proposed a random matrix-based algorithm to enhance the model's reasoning ability, resulting in a 75% reduction in the training time required for the GPT-2 model to achieve generalization capability on the PrOntoQA dataset. These findings provide new insights into understanding the mechanisms of large language models."
                    }
                ]
            },
            "6d7e30da-7cae-475b-afa9-990b1d28bb90": {
                "pk": "6d7e30da-7cae-475b-afa9-990b1d28bb90",
                "name": "Pengxiao Lin",
                "collaborators": [
                    "Aoming Liang",
                    "Zhaoyang Mu",
                    "Cong Wang",
                    "Mingming Ge",
                    "Ling Shao",
                    "Dixia Fan",
                    "Hao Tang"
                ],
                "domain": [
                    "Machine Learning",
                    "Partial Differential Equations",
                    "Neural Operators",
                    "Multi-Scale Systems"
                ],
                "publications": [
                    {
                        "title": "M$^{2}$M: Learning controllable Multi of experts and multi-scale operators are the Partial Differential Equations need",
                        "abstract": "Learning the evolutionary dynamics of Partial Differential Equations (PDEs) is critical in understanding dynamic systems, yet current methods insufficiently learn their representations. This is largely due to the multi-scale nature of the solution, where certain regions exhibit rapid oscillations while others evolve more slowly. This paper introduces a framework of multi-scale and multi-expert (M$^2$M) neural operators designed to simulate and learn PDEs efficiently. We employ a divide-and-conquer strategy to train a multi-expert gated network for the dynamic router policy. Our method incorporates a controllable prior gating mechanism that determines the selection rights of experts, enhancing the model's efficiency. To optimize the learning process, we have implemented a PI (Proportional, Integral) control strategy to adjust the allocation rules precisely. This universal controllable approach allows the model to achieve greater accuracy. We test our approach on benchmark 2D Navier-Stokes equations and provide a custom multi-scale dataset. M$^2$M can achieve higher simulation accuracy and offer improved interpretability compared to baseline methods."
                    }
                ]
            },
            "dadc2e29-5c0b-4332-a603-0d352a279656": {
                "pk": "dadc2e29-5c0b-4332-a603-0d352a279656",
                "name": "Zhiwei Wang",
                "collaborators": [
                    "Xiangyu Zhou",
                    "Yinji Li",
                    "Zhan Li",
                    "Bingrong Huang",
                    "Yongxiao Lin",
                    "Huazheng Wang",
                    "Hongning Wang",
                    "Yongkang Wang",
                    "Wen Zhang",
                    "Fusheng Deng"
                ],
                "domain": [
                    "Complex Geometry",
                    "Hermitian Manifolds",
                    "Machine Learning",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "On Grauert-Riemenschneider type criterions",
                        "abstract": "Let $(X,\\omega)$ be a compact Hermitian manifold of complex dimension $n$. In this article, we first survey recent progress towards Grauert-Riemenschneider type criterions. Secondly, we give a simplified proof of Boucksom's conjecture given by the author under the assumption that the Hermitian metric $\\omega$ satisfies $\\partial\\overline{\\partial}\\omega^l=$ for all $l$, i.e., if $T$ is a closed positive current on $X$ such that $\\int_XT_{ac}^n>0$, then the class $\\{T\\}$ is big and $X$ is K\\\"{a}hler. Finally, as an easy observation, we point out that Nguyen's result can be generalized as follows: if $\\partial\\overline{\\partial}\\omega=0$, and $T$ is a closed positive current with analytic singularities, such that $\\int_XT^n_{ac}>0$, then the class $\\{T\\}$ is big and $X$ is K\\\"{a}hler."
                    },
                    {
                        "title": "On the volume of a pseudo-effective class and semi-positive properties of the Harder-Narasimhan filtration on a compact Hermitian manifold",
                        "abstract": "This paper divides into two parts. Let $(X,\\omega)$ be a compact Hermitian manifold. Firstly, if the Hermitian metric $\\omega$ satisfies the assumption that $\\partial\\overline{\\partial}\\omega^k=0$ for all $k$, we generalize the volume of the cohomology class in the K\\\"{a}hler setting to the Hermitian setting, and prove that the volume is always finite and the Grauert-Riemenschneider type criterion holds true, which is a partial answer to a conjecture posed by Boucksom. Secondly, we observe that if the anticanonical bundle $K^{-1}_X$ is nef, then for any $\\varepsilon>0$, there is a smooth function $\\phi_\\varepsilon$ on $X$ such that $\\omega_\\varepsilon:=\\omega+i\\partial\\overline{\\partial}\\phi_\\varepsilon>0$ and Ricci$(\\omega_\\varepsilon)\\geq-\\varepsilon\\omega_\\varepsilon$. Furthermore, if $\\omega$ satisfies the assumption as above, we prove that for a Harder-Narasimhan filtration of $T_X$ with respect to $\\omega$, the slopes $\\mu_\\omega(\\mathcal{F}_i/\\mathcal{F}_{i-1})\\geq 0$ for all $i$, which generalizes a result of Cao which plays a very important role in his studying of the structures of K\\\"{a}hler manifolds."
                    },
                    {
                        "title": "Local Demailly-Bouche's holomorphic Morse inequalities",
                        "abstract": "Let $(X,\\omega)$ be a Hermitian manifold and let $(E,h^E)$, $(F,h^F)$ be two Hermitian holomorphic line bundle over $X$. Suppose that the maximal rank of the Chern curvature $c(E)$ of $E$ is $r$, and the kernel of $c(E)$ is foliated, i.e. there is a foliation $Y$ of $X$, of complex codimension $r$, such that the tangent space of the leaf at each point $x\\in X$ is contained in the kernel of $c(E)$. In this paper, local versions of Demailly-Bouche's holomorphic Morse inequalities (which give asymptotic bounds for cohomology groups $H^{q}(X,E^k\\otimes F^l)$ as $k,l,k/l\\rightarrow \\infty$) are presented. The local version holds on any Hermitian manifold regardless of compactness and completeness. The proof is a variation of Berman's method to derive holomorphic Morse inequalities on compact complex manifolds with boundary."
                    },
                    {
                        "title": "Bergman kernels and eigenvalue estimate of $\\bar{\\partial}$-laplacian",
                        "abstract": "Let $(X,\\omega)$ be a compact K\\\"{a}hler manifold. Let $(L,h)$ be a hermitian holomorphic line bundle over $X$, such that $\\Theta_{L,h}\\geq -\\varepsilon\\omega$ for a small $\\varepsilon>0$, $E$ be a holomorphic line bundle over $X$. For $k\\in \\mathbb{N}_+$, denote by $X_k:=(X,\\omega^k)$ the K\\\"{a}hler manifold $X$ with new scaled metric $\\omega^k=k\\omega$. Estimates of the number of eigenvalues smaller than $\\lambda$ of the $\\debar$-Laplacian on forms on $X_k$ with values in $L^k\\otimes E$ are presented for $0\\leq \\lambda<k$. In particular, when $\\lambda=0$, we get a numeric bound for the cohomology groups."
                    },
                    {
                        "title": "Sur les plus grands facteurs premiers d'entiers cons\u00e9cutifs",
                        "abstract": "Let $P^+(n)$ denote the largest prime factor of the integer $n$ and $P_y^+(n)$ denote the largest prime factor $p$ of $n$ which satisfies $p\\leqslant y$. In this paper, firstly we show that the triple consecutive integers with the two patterns $P^+(n-1)>P^+(n)<P^+(n+1)$ and $P^+(n-1)<P^+(n)>P^+(n+1)$ have a positive proportion respectively. More generally, with the same methods we can prove that for any $J\\in \\mathbb{Z}, J\\geqslant3$, the $J-$tuple consecutive integers with the two patterns $P^+(n+j_0)= \\min\\limits_{0\\leqslant j\\leqslant J-1}P^+(n+j)$ and $P^+(n+j_0)= \\max\\limits_{0\\leqslant j\\leqslant J-1}P^+(n+j)$ also have a positive proportion respectively. Secondly for $y=x^{\\theta}$ with $0<\\theta\\leqslant 1$ we show that there exists a positive proportion of integers $n$ such that $P_y^+(n)<P_y^+(n+1)$. Specially, we can prove that the proportion of integers $n$ such that $P^+(n)<P^+(n+1)$ is larger than 0.1356, which improves the previous result \"0.1063\" of the author."
                    },
                    {
                        "title": "Invariance of plurigenera for generalized polarized pairs with abundant nef parts",
                        "abstract": "We show the invariance of plurigenera for generalized polarized pairs with abundant nef parts and generalized canonical singularities. This is obtained by investigating a type of newly introduced multiplier ideal sheaf which is of bimeromorphic nature."
                    },
                    {
                        "title": "On the extensions of K\u00e4hler currents on compact K\u00e4hler manifolds",
                        "abstract": "Let $(X,\\omega)$ be a compact K\\\"{a}hler manifold with a K\\\"{a}hler form $\\omega$ of complex dimension $n$, and $V\\subset X$ is a compact complex submanifold of positive dimension $k<n$. Suppose that $V$ can be embedded in $X$ as a zero section of a holomorphic vector bundle or rank $n-k$ over $V$. Let $\\varphi$ be a strictly $\\omega|_V$-psh function on $V$. In this paper, we prove that there is a strictly $\\omega$-psh function $\\Phi$ on $X$, such that $\\Phi|_V=\\varphi$. This result gives a partial answer to an open problem raised by Collins-Tosatti and Dinew-Guedj-Zeriahi, for the case of K\\\"{a}hler currents. We also discuss possible extensions of K\\\"ahler currents in a big class."
                    },
                    {
                        "title": "Asymptotic estimate of cohomology groups valued in pseudo-effective line bundles",
                        "abstract": "In this paper, we study questions of Demailly and Matsumura on the asymptotic behavior of dimensions of cohomology groups for high tensor powers of (nef) pseudo-effective line bundles over non-necessarily projective algebraic manifolds. By generalizing Siu's $\\partial\\overline{\\partial}$-formula and Berndtsson's eigenvalue estimate of $\\overline{\\partial}$-Laplacian and combining Bonavero's technique, we obtain the following result: given a holomorphic pseudo-effective line bundle $(L, h_L)$ on a compact Hermitian manifold $(X,\\omega)$, if $h_L$ is a singular metric with algebraic singularities, then $\\dim H^{q}(X,L^k\\otimes E\\otimes \\mathcal{I}(h_L^{k}))\\leq Ck^{n-q}$ for $k$ large, with $E$ an arbitrary holomorphic vector bundle. As applications, we obtain partial solutions to the questions of Demailly and Matsumura."
                    },
                    {
                        "title": "CR eigenvalue estimate and Kohn-Rossi cohomology",
                        "abstract": "Let $X$ be a compact connected CR manifold with a transversal CR $S^1$-action of real dimension $2n-1$, which is only assumed to be weakly pseudoconvex. Let $\\Box_b$ be the $\\overline{\\partial}_b$-Laplacian, with respect to a $T$-rigid Hermitian metric (see Definition 3.2 of $T$-rigid Hermitian metric). Eigenvalue estimate of $\\Box_b$ is a fundamental issue both in CR geometry and analysis. In this paper, we are able to obtain a sharp estimate of the number of eigenvalues smaller than or equal to $\\lambda$ of $\\Box_b$ acting on the $m$-th Fourier components of smooth $(n-1,q)$-forms on $X$, where $m\\in \\mathbb{Z}_+$ and $q=0,1,\\cdots, n-1$. Here the sharp means the growth order with respect to $m$ is sharp. In particular, when $\\lambda=0$, we obtain the asymptotic estimate of the growth for $m$-th Fourier components $H^{n-1,q}_{b,m}(X)$ of $H^{n-1,q}_b(X)$ as $m \\rightarrow +\\infty$. Furthermore, we establish a Serre type duality theorem for Fourier components of Kohn-Rossi cohomology which is of independent interest. As a byproduct, the asymptotic growth of the dimensions of the Fourier components $H^{0,q}_{b,-m}(X)$ for $ m\\in \\mathbb{Z}_+$ is established. We also give appilcations of our main results, including Morse type inequalities, asymptotic Riemann-Roch type theorem, Grauert-Riemenscheider type criterion, and an orbifold version of our main results which provides an answer towards a folklore open problem informed to us by Hsiao."
                    },
                    {
                        "title": "Averages of coefficients of a class of degree 3 L-functions",
                        "abstract": "In this note, we give a detailed proof of an asymptotic for averages of coefficients of a class of degree three $L$-functions which can be factorized as a product of a degree one and a degree two $L$-functions. We emphasize that we can break the $1/2$-barrier in the error term, and we get an explicit exponent."
                    },
                    {
                        "title": "Degenerate complex Monge-Amp\\` ere type equations on compact Hermitian manifolds and applications",
                        "abstract": "We show the existence and uniqueness of bounded solutions to the degenerate complex Monge-Amp\\`ere type equations on compact Hermitian manifolds. We also study the asymptotics of these solutions. As applications, we give partial answers to the Tosatti-Weinkove conjecture and Demailly-P\\u aun conjecture."
                    },
                    {
                        "title": "Stealthy Adversarial Attacks on Stochastic Multi-Armed Bandits",
                        "abstract": "Adversarial attacks against stochastic multi-armed bandit (MAB) algorithms have been extensively studied in the literature. In this work, we focus on reward poisoning attacks and find most existing attacks can be easily detected by our proposed detection method based on the test of homogeneity, due to their aggressive nature in reward manipulations. This motivates us to study the notion of stealthy attack against stochastic MABs and investigate the resulting attackability. Our analysis shows that against two popularly employed MAB algorithms, UCB1 and $\\epsilon$-greedy, the success of a stealthy attack depends on the environmental conditions and the realized reward of the arm pulled in the first round. We also analyze the situation for general MAB algorithms equipped with our attack detection method and find that it is possible to have a stealthy attack that almost always succeeds. This brings new insights into the security risks of MAB algorithms."
                    },
                    {
                        "title": "A note on Demailly's transcendental Morse inequalities conjecture",
                        "abstract": "Let $(X,\\omega)$ be an $n$-dimensional compact Hermitian manifold with $\\omega$ a pluriclosed Hermitian metric, i.e. $dd^c\\omega=0$. Let $\\{\\alpha\\},\\{\\beta \\}\\in H^{1,1}_{BC}(X,\\mathbb R)$ be two nef classes, such that $\\alpha^n-n\\alpha^{n-1}\\cdot\\beta>0$. In this short note, we prove that if there is a bounded quasi-plurisubharmonic potential $\\rho$, such that $\\alpha+dd^c\\rho\\geq 0$ in the weak sense of currents, then the class $\\{\\alpha-\\beta\\}$ contains a K\\\"ahler current. This gives a partial solution of Demailly's transcendental Morse inequalities conjecture."
                    },
                    {
                        "title": "Improving Paratope and Epitope Prediction by Multi-Modal Contrastive Learning and Interaction Informativeness Estimation",
                        "abstract": "Accurately predicting antibody-antigen binding residues, i.e., paratopes and epitopes, is crucial in antibody design. However, existing methods solely focus on uni-modal data (either sequence or structure), disregarding the complementary information present in multi-modal data, and most methods predict paratopes and epitopes separately, overlooking their specific spatial interactions. In this paper, we propose a novel Multi-modal contrastive learning and Interaction informativeness estimation-based method for Paratope and Epitope prediction, named MIPE, by using both sequence and structure data of antibodies and antigens. MIPE implements a multi-modal contrastive learning strategy, which maximizes representations of binding and non-binding residues within each modality and meanwhile aligns uni-modal representations towards effective modal representations. To exploit the spatial interaction information, MIPE also incorporates an interaction informativeness estimation that computes the estimated interaction matrices between antibodies and antigens, thereby approximating them to the actual ones. Extensive experiments demonstrate the superiority of our method compared to baselines. Additionally, the ablation studies and visualizations demonstrate the superiority of MIPE owing to the better representations acquired through multi-modal contrastive learning and the interaction patterns comprehended by the interaction informativeness estimation."
                    },
                    {
                        "title": "Characterizations of plurisubharmonic functions",
                        "abstract": "We give characterizations of (quasi-)plurisubharmonic functions in terms of $L^p$-estimates of $\\bar\\partial$ and $L^p$-extensions of holomorphic functions."
                    },
                    {
                        "title": "Linked Recurrent Neural Networks",
                        "abstract": "Recurrent Neural Networks (RNNs) have been proven to be effective in modeling sequential data and they have been applied to boost a variety of tasks such as document classification, speech recognition and machine translation. Most of existing RNN models have been designed for sequences assumed to be identically and independently distributed (i.i.d). However, in many real-world applications, sequences are naturally linked. For example, web documents are connected by hyperlinks; and genes interact with each other. On the one hand, linked sequences are inherently not i.i.d., which poses tremendous challenges to existing RNN models. On the other hand, linked sequences offer link information in addition to the sequential information, which enables unprecedented opportunities to build advanced RNN models. In this paper, we study the problem of RNN for linked sequences. In particular, we introduce a principled approach to capture link information and propose a linked Recurrent Neural Network (LinkedRNN), which models sequential and link information coherently. We conduct experiments on real-world datasets from multiple domains and the experimental results validate the effectiveness of the proposed framework."
                    }
                ]
            },
            "bd749f84-74c5-419d-9f31-f161a5170338": {
                "pk": "bd749f84-74c5-419d-9f31-f161a5170338",
                "name": "Yaoyu Zhang",
                "collaborators": [
                    "Zhi-Qin John Xu",
                    "Tao Luo",
                    "Leyang Zhang",
                    "Zhiwei Bai",
                    "Jiajie Zhao",
                    "Zhongwang Zhang",
                    "Tianhan Zhang",
                    "Weinan E",
                    "Yiguang Ju",
                    "Jian Sun"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Optimization",
                    "Implicit Regularization"
                ],
                "publications": [
                    {
                        "title": "Geometry and Local Recovery of Global Minima of Two-layer Neural Networks at Overparameterization",
                        "abstract": "Under mild assumptions, we investigate the geometry of the loss landscape for two-layer neural networks in the vicinity of global minima. Utilizing novel techniques, we demonstrate: (i) how global minima with zero generalization error become geometrically separated from other global minima as the sample size grows; and (ii) the local convergence properties and rate of gradient flow dynamics. Our results indicate that two-layer neural networks can be locally recovered in the regime of overparameterization."
                    },
                    {
                        "title": "Geometry of Critical Sets and Existence of Saddle Branches for Two-layer Neural Networks",
                        "abstract": "This paper presents a comprehensive analysis of critical point sets in two-layer neural networks. To study such complex entities, we introduce the critical embedding operator and critical reduction operator as our tools. Given a critical point, we use these operators to uncover the whole underlying critical set representing the same output function, which exhibits a hierarchical structure. Furthermore, we prove existence of saddle branches for any critical set whose output function can be represented by a narrower network. Our results provide a solid foundation to the further study of optimization and training behavior of neural networks."
                    },
                    {
                        "title": "A deep learning-based ODE solver for chemical kinetics",
                        "abstract": "Developing efficient and accurate algorithms for chemistry integration is a challenging task due to its strong stiffness and high dimensionality. The current work presents a deep learning-based numerical method called DeepCombustion0.0 to solve stiff ordinary differential equation systems. The homogeneous autoignition of DME/air mixture, including 54 species, is adopted as an example to illustrate the validity and accuracy of the algorithm. The training and testing datasets cover a wide range of temperature, pressure, and mixture conditions between 750-1200 K, 30-50 atm, and equivalence ratio = 0.7-1.5. Both the first-stage low-temperature ignition (LTI) and the second-stage high-temperature ignition (HTI) are considered. The methodology highlights the importance of the adaptive data sampling techniques, power transform preprocessing, and binary deep neural network (DNN) design. By using the adaptive random samplings and appropriate power transforms, smooth submanifolds in the state vector phase space are observed, on which two three-layer DNNs can be appropriately trained. The neural networks are end-to-end, which predict temporal gradients of the state vectors directly. The results show that temporal evolutions predicted by DNN agree well with traditional numerical methods in all state vector dimensions, including temperature, pressure, and species concentrations. Besides, the ignition delay time differences are within 1%. At the same time, the CPU time is reduced by more than 20 times and 200 times compared with the HMTS and VODE method, respectively. The current work demonstrates the enormous potential of applying the deep learning algorithm in chemical kinetics and combustion modeling."
                    },
                    {
                        "title": "Storage Control for Carbon Emission Reduction: Opportunities and Challenges",
                        "abstract": "Storage is vital to power systems as it provides the urgently needed flexibility to the system. Meanwhile, it can contribute more than flexibility. In this paper, we study the possibility of utilizing storage system for carbon emission reduction. The opportunity arises due to the pending implementation of carbon tax throughout the world. Without the right incentive, most system operators have to dispatch the generators according to the merit order of the fuel costs, without any control for carbon emissions. However, we submit that storage may provide necessary flexibility in carbon emission reduction even without carbon tax. We identify the non-convex structure to conduct storage control for this task and propose an easy to implement dynamic programming algorithm to investigate the value of storage in carbon emission reduction."
                    },
                    {
                        "title": "Connectivity Shapes Implicit Regularization in Matrix Factorization Models for Matrix Completion",
                        "abstract": "Matrix factorization models have been extensively studied as a valuable test-bed for understanding the implicit biases of overparameterized models. Although both low nuclear norm and low rank regularization have been studied for these models, a unified understanding of when, how, and why they achieve different implicit regularization effects remains elusive. In this work, we systematically investigate the implicit regularization of matrix factorization for solving matrix completion problems. We empirically discover that the connectivity of observed data plays a crucial role in the implicit bias, with a transition from low nuclear norm to low rank as data shifts from disconnected to connected with increased observations. We identify a hierarchy of intrinsic invariant manifolds in the loss landscape that guide the training trajectory to evolve from low-rank to higher-rank solutions. Based on this finding, we theoretically characterize the training trajectory as following the hierarchical invariant manifold traversal process, generalizing the characterization of Li et al. (2020) to include the disconnected case. Furthermore, we establish conditions that guarantee minimum nuclear norm, closely aligning with our experimental findings, and we provide a dynamics characterization condition for ensuring minimum rank. Our work reveals the intricate interplay between data connectivity, training dynamics, and implicit regularization in matrix factorization models."
                    },
                    {
                        "title": "Disentangle Sample Size and Initialization Effect on Perfect Generalization for Single-Neuron Target",
                        "abstract": "Overparameterized models like deep neural networks have the intriguing ability to recover target functions with fewer sampled data points than parameters (see arXiv:2307.08921). To gain insights into this phenomenon, we concentrate on a single-neuron target recovery scenario, offering a systematic examination of how initialization and sample size influence the performance of two-layer neural networks. Our experiments reveal that a smaller initialization scale is associated with improved generalization, and we identify a critical quantity called the \"initial imbalance ratio\" that governs training dynamics and generalization under small initialization, supported by theoretical proofs. Additionally, we empirically delineate two critical thresholds in sample size--termed the \"optimistic sample size\" and the \"separation sample size\"--that align with the theoretical frameworks established by (see arXiv:2307.08921 and arXiv:2309.00508). Our results indicate a transition in the model's ability to recover the target function: below the optimistic sample size, recovery is unattainable; at the optimistic sample size, recovery becomes attainable albeit with a set of initialization of zero measure. Upon reaching the separation sample size, the set of initialization that can successfully recover the target function shifts from zero to positive measure. These insights, derived from a simplified context, provide a perspective on the intricate yet decipherable complexities of perfect generalization in overparameterized neural networks."
                    },
                    {
                        "title": "Local Linear Recovery Guarantee of Deep Neural Networks at Overparameterization",
                        "abstract": "Determining whether deep neural network (DNN) models can reliably recover target functions at overparameterization is a critical yet complex issue in the theory of deep learning. To advance understanding in this area, we introduce a concept we term \"local linear recovery\" (LLR), a weaker form of target function recovery that renders the problem more amenable to theoretical analysis. In the sense of LLR, we prove that functions expressible by narrower DNNs are guaranteed to be recoverable from fewer samples than model parameters. Specifically, we establish upper limits on the optimistic sample sizes, defined as the smallest sample size necessary to guarantee LLR, for functions in the space of a given DNN. Furthermore, we prove that these upper bounds are achieved in the case of two-layer tanh neural networks. Our research lays a solid groundwork for future investigations into the recovery capabilities of DNNs in overparameterized scenarios."
                    },
                    {
                        "title": "Data-informed Deep Optimization",
                        "abstract": "Complex design problems are common in the scientific and industrial fields. In practice, objective functions or constraints of these problems often do not have explicit formulas, and can be estimated only at a set of sampling points through experiments or simulations. Such optimization problems are especially challenging when design parameters are high-dimensional due to the curse of dimensionality. In this work, we propose a data-informed deep optimization (DiDo) approach as follows: first, we use a deep neural network (DNN) classifier to learn the feasible region; second, we sample feasible points based on the DNN classifier for fitting of the objective function; finally, we find optimal points of the DNN-surrogate optimization problem by gradient descent. To demonstrate the effectiveness of our DiDo approach, we consider a practical design case in industry, in which our approach yields good solutions using limited size of training data. We further use a 100-dimension toy example to show the effectiveness of our model for higher dimensional problems. Our results indicate that the DiDo approach empowered by DNN is flexible and promising for solving general high-dimensional design problems in practice."
                    },
                    {
                        "title": "Training behavior of deep neural network in frequency domain",
                        "abstract": "Why deep neural networks (DNNs) capable of overfitting often generalize well in practice is a mystery [#zhang2016understanding]. To find a potential mechanism, we focus on the study of implicit biases underlying the training process of DNNs. In this work, for both real and synthetic datasets, we empirically find that a DNN with common settings first quickly captures the dominant low-frequency components, and then relatively slowly captures the high-frequency ones. We call this phenomenon Frequency Principle (F-Principle). The F-Principle can be observed over DNNs of various structures, activation functions, and training algorithms in our experiments. We also illustrate how the F-Principle help understand the effect of early-stopping as well as the generalization of DNNs. This F-Principle potentially provides insights into a general principle underlying DNN optimization and generalization."
                    },
                    {
                        "title": "Overview frequency principle/spectral bias in deep learning",
                        "abstract": "Understanding deep learning is increasingly emergent as it penetrates more and more into industry and science. In recent years, a research line from Fourier analysis sheds lights into this magical \"black box\" by showing a Frequency Principle (F-Principle or spectral bias) of the training behavior of deep neural networks (DNNs) -- DNNs often fit functions from low to high frequency during the training. The F-Principle is first demonstrated by one-dimensional synthetic data followed by the verification in high-dimensional real datasets. A series of works subsequently enhance the validity of the F-Principle. This low-frequency implicit bias reveals the strength of neural network in learning low-frequency functions as well as its deficiency in learning high-frequency functions. Such understanding inspires the design of DNN-based algorithms in practical problems, explains experimental phenomena emerging in various scenarios, and further advances the study of deep learning from the frequency perspective. Although incomplete, we provide an overview of F-Principle and propose some open problems for future research."
                    },
                    {
                        "title": "A rationale from frequency perspective for grokking in training neural network",
                        "abstract": "Grokking is the phenomenon where neural networks NNs initially fit the training data and later generalize to the test data during training. In this paper, we empirically provide a frequency perspective to explain the emergence of this phenomenon in NNs. The core insight is that the networks initially learn the less salient frequency components present in the test data. We observe this phenomenon across both synthetic and real datasets, offering a novel viewpoint for elucidating the grokking phenomenon by characterizing it through the lens of frequency dynamics during the training process. Our empirical frequency-based analysis sheds new light on understanding the grokking phenomenon and its underlying mechanisms."
                    },
                    {
                        "title": "State-dependent Filtering of the Ring Model",
                        "abstract": "Robustness is a measure of functional reliability of a system against perturbations. To achieve a good and robust performance, a system must filter out external perturbations by its internal priors. These priors are usually distilled in the structure and the states of the system. Biophysical neural network are known to be robust but the exact mechanisms are still elusive. In this paper, we probe how orientation-selective neurons organized on a 1-D ring network respond to perturbations in the hope of gaining some insights on the robustness of visual system in brain. We analyze the steady-state of the rate-based network and prove that the activation state of neurons, rather than their firing rates, determines how the model respond to perturbations. We then identify specific perturbation patterns that induce the largest responses for different configurations of activation states, and find them to be sinusoidal or sinusoidal-like while other patterns are largely attenuated. Similar results are observed in a spiking ring model. Finally, we remap the perturbations in orientation back into the 2-D image space using Gabor functions. The resulted optimal perturbation patterns mirror adversarial attacks in deep learning that exploit the priors of the system. Our results suggest that based on different state configurations, these priors could underlie some of the illusionary experiences as the cost of visual robustness."
                    },
                    {
                        "title": "Embedding Principle of Loss Landscape of Deep Neural Networks",
                        "abstract": "Understanding the structure of loss landscape of deep neural networks (DNNs)is obviously important. In this work, we prove an embedding principle that the loss landscape of a DNN \"contains\" all the critical points of all the narrower DNNs. More precisely, we propose a critical embedding such that any critical point, e.g., local or global minima, of a narrower DNN can be embedded to a critical point/hyperplane of the target DNN with higher degeneracy and preserving the DNN output function. The embedding structure of critical points is independent of loss function and training data, showing a stark difference from other nonconvex problems such as protein-folding. Empirically, we find that a wide DNN is often attracted by highly-degenerate critical points that are embedded from narrow DNNs. The embedding principle provides an explanation for the general easy optimization of wide DNNs and unravels a potential implicit low-complexity regularization during the training. Overall, our work provides a skeleton for the study of loss landscape of DNNs and its implication, by which a more exact and comprehensive understanding can be anticipated in the near"
                    },
                    {
                        "title": "Limitation of Characterizing Implicit Regularization by Data-independent Functions",
                        "abstract": "In recent years, understanding the implicit regularization of neural networks (NNs) has become a central task in deep learning theory. However, implicit regularization is itself not completely defined and well understood. In this work, we attempt to mathematically define and study implicit regularization. Importantly, we explore the limitations of a common approach to characterizing implicit regularization using data-independent functions. We propose two dynamical mechanisms, i.e., Two-point and One-point Overlapping mechanisms, based on which we provide two recipes for producing classes of one-hidden-neuron NNs that provably cannot be fully characterized by a type of or all data-independent functions. Following the previous works, our results further emphasize the profound data dependency of implicit regularization in general, inspiring us to study in detail the data dependency of NN implicit regularization in the future."
                    },
                    {
                        "title": "Implicit bias with Ritz-Galerkin method in understanding deep learning for solving PDEs",
                        "abstract": "This paper aims at studying the difference between Ritz-Galerkin (R-G) method and deep neural network (DNN) method in solving partial differential equations (PDEs) to better understand deep learning. To this end, we consider solving a particular Poisson problem, where the information of the right-hand side of the equation f is only available at n sample points, that is, f is known at finite sample points. Through both theoretical and numerical studies, we show that solution of the R-G method converges to a piecewise linear function for the one dimensional (1D) problem or functions of lower regularity for high dimensional problems. With the same setting, DNNs however learn a relative smooth solution regardless of the dimension, this is, DNNs implicitly bias towards functions with more low-frequency components among all functions that can fit the equation at available data points. This bias is explained by the recent study of frequency principle (Xu et al., (2019) [17] and Zhang et al., (2019) [11, 19]). In addition to the similarity between the traditional numerical methods and DNNs in the approximation perspective, our work shows that the implicit bias in the learning process, which is different from traditional numerical methods, could help better understand the characteristics of DNNs."
                    }
                ]
            },
            "97d7b18d-3ed2-4fe0-a207-30a36621af36": {
                "pk": "97d7b18d-3ed2-4fe0-a207-30a36621af36",
                "name": "Zhi-Qin John Xu",
                "collaborators": [
                    "Douglas Zhou",
                    "David Cai",
                    "Wei Cai"
                ],
                "domain": [
                    "Dynamical Systems",
                    "Deep Learning",
                    "Neural Networks",
                    "PDEs"
                ],
                "publications": [
                    {
                        "title": "Dynamical and Coupling Structure of Pulse-Coupled Networks in Maximum Entropy Analysis",
                        "abstract": "Maximum entropy principle (MEP) analysis with few non-zero effective interactions successfully characterizes the distribution of dynamical states of pulse-coupled networks in many experiments, e.g., in neuroscience. To better understand the underlying mechanism, we found a relation between the dynamical structure, i.e., effective interactions in MEP analysis, and the coupling structure of pulse-coupled network to understand how a sparse coupling structure could lead to a sparse coding by effective interactions. This relation quantitatively displays how the dynamical structure is closely related to the coupling structure."
                    },
                    {
                        "title": "Multi-scale Deep Neural Networks for Solving High Dimensional PDEs",
                        "abstract": "In this paper, we propose the idea of radial scaling in frequency domain and activation functions with compact support to produce a multi-scale DNN (MscaleDNN), which will have the multi-scale capability in approximating high frequency and high dimensional functions and speeding up the solution of high dimensional PDEs. Numerical results on high dimensional function fitting and solutions of high dimensional PDEs, using loss functions with either Ritz energy or least squared PDE residuals, have validated the increased power of multi-scale resolution and high frequency capturing of the proposed MscaleDNN."
                    }
                ]
            }
        }
    },
    "2310.03957": {
        "paper_data": {
            "title": "Understanding prompt engineering may not require rethinking generalization",
            "url": "http://arxiv.org/abs/2310.03957v1",
            "arxiv_id": "2310.03957",
            "authors": [
                "Victor Akinwande",
                "Yiding Jiang",
                "Dylan Sam",
                "J. Zico Kolter"
            ],
            "abstract": "Zero-shot learning in prompted vision-language models, the practice of crafting prompts to build classifiers without an explicit training process, has achieved impressive performance in many settings. This success presents a seemingly surprising observation: these methods suffer relatively little from overfitting, i.e., when a prompt is manually engineered to achieve low error on a given training set (thus rendering the method no longer actually zero-shot), the approach still performs well on held-out test data. In this paper, we show that we can explain such performance well via recourse to classical PAC-Bayes bounds. Specifically, we show that the discrete nature of prompts, combined with a PAC-Bayes prior given by a language model, results in generalization bounds that are remarkably tight by the standards of the literature: for instance, the generalization bound of an ImageNet classifier is often within a few percentage points of the true test error. We demonstrate empirically that this holds for existing handcrafted prompts and prompts generated through simple greedy search. Furthermore, the resulting bound is well-suited for model selection: the models with the best bound typically also have the best test performance. This work thus provides a possible justification for the widespread practice of prompt engineering, even if it seems that such methods could potentially overfit the training data.",
            "introduction": "   1 Introduction  Generalization bounds provide statistical guarantees on the average-case performance of a learning algorithm\u2019s output. However, in the case of deep learning models, there is still debate about how useful such bounds can be: Zhang et\u00a0al. (2021) highlighted that classical approaches for deriving generalization bounds are insufficient for explaining the generalization ability of deep learning, spurring a flurry of new approaches for deriving tighter generalization bounds for deep neural networks\u00a0(Bartlett et\u00a0al., 2017; Dziugaite & Roy, 2017; Neyshabur et\u00a0al., 2017b). In the recent literature on generalization bounds for deep learning, a large focus has been on developing data-dependent bounds, or bounds that consider both the data distribution and the hypothesis space. Some of the best data-dependent bounds arise from the PAC-Bayes framework\u00a0(McAllester, 1999) and are derived by bounding the KL divergence between a prior over the hypothesis space and the posterior yielded by the learning algorithm. However, although PAC-Bayes bounds led to the first non-vacuous generalization bounds for deep learning\u00a0(Dziugaite & Roy, 2017), they are still too loose to be practically useful\u00a0(Jiang et\u00a0al., 2019) in most realistic settings. In fact, as Lotfi et\u00a0al. (2022) have recently argued, many PAC-Bayes bounds with data-dependent priors, while non-vacuous, can be best described as validation bounds \u2014 i.e., the use of data-dependent priors effectively leverages held-out data in a manner similar to cross-validation, which undermines their ability to explain generalization.   Notwithstanding the lack of a clear theoretical basis, modern machine learning models are moving towards increasingly large pretrained models\u00a0(Kaplan et\u00a0al., 2020; Dosovitskiy et\u00a0al., 2020). One prevailing paradigm is to use pretrained foundation models such as CLIP (Radford et\u00a0al., 2021) or ALIGN (Jia et\u00a0al., 2021) as feature extractors and provide weak supervision for a downstream target task via prompts, which are text descriptions of the desired tasks that are often significantly easier to obtain compared to full model weights or even a generic linear classifier over the last layer. The versatility and performance of prompting pretrained models have led to the rise of prompt engineering, an emergent paradigm in machine learning where practitioners carefully design the task specification in text or even learn the prompts in a data-driven fashion\u00a0(Lester et\u00a0al., 2021). For example, to obtain a two-class image classifier, one would write two sentences that describe the classes (e.g., \u201cThis is a dog\u201d and \u201cThis is a cat\u201d), and the two sentences are turned into text embeddings which can be used to classify image embeddings. Despite its empirical success, little is understood of how and why prompting these pretrained models work and, in particular, why the method seems to suffer little from overfitting: manually tuning or even greedily optimizing prompts on a given training set often performs nearly as well on the corresponding test set.   In this paper, we demonstrate that rather simple analysis tools capture this behavior surprisingly well (under some assumptions). In particular, we show that classical PAC-Bayes bounds\u00a0(McAllester, 1999), when applied to the discrete hypothesis class defined by prompts (and specifically with a prior given by a large language model), are often remarkably tight, even for large domains: for example, we achieve a generalization bound of 32% error for a full ImageNet classifier, which",
            "references": [
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Hard Prompts Made Easy: Gradient-Based Discrete Optimization for Prompt Tuning and Discovery",
                    "abstract": "The strength of modern generative models lies in their ability to be controlled through text-based prompts. Typical\"hard\"prompts are made from interpretable words and tokens, and must be hand-crafted by humans. There are also\"soft\"prompts, which consist of continuous feature vectors. These can be discovered using powerful optimization methods, but they cannot be easily interpreted, re-used across models, or plugged into a text-based interface. We describe an approach to robustly optimize hard text prompts through efficient gradient-based optimization. Our approach automatically generates hard text-based prompts for both text-to-image and text-to-text applications. In the text-to-image setting, the method creates hard prompts for diffusion models, allowing API users to easily generate, discover, and mix and match image concepts without prior knowledge on how to prompt the model. In the text-to-text setting, we show that hard prompts can be automatically discovered that are effective in tuning LMs for classification."
                },
                {
                    "title": "PAC-Bayes Compression Bounds So Tight That They Can Explain Generalization",
                    "abstract": "While there has been progress in developing non-vacuous generalization bounds for deep neural networks, these bounds tend to be uninformative about why deep learning works. In this paper, we develop a compression approach based on quantizing neural network parameters in a linear subspace, profoundly improving on previous results to provide state-of-the-art generalization bounds on a variety of tasks, including transfer learning. We use these tight bounds to better understand the role of model size, equivariance, and the implicit biases of optimization, for generalization in deep learning. Notably, we find large models can be compressed to a much greater extent than previously known, encapsulating Occam's razor. We also argue for data-independent bounds in explaining generalization."
                },
                {
                    "title": "Large Language Models Are Human-Level Prompt Engineers",
                    "abstract": "By conditioning on natural language instructions, large language models (LLMs) have displayed impressive capabilities as general-purpose computers. However, task performance depends significantly on the quality of the prompt used to steer the model, and most effective prompts have been handcrafted by humans. Inspired by classical program synthesis and the human approach to prompt engineering, we propose Automatic Prompt Engineer (APE) for automatic instruction generation and selection. In our method, we treat the instruction as the\"program,\"optimized by searching over a pool of instruction candidates proposed by an LLM in order to maximize a chosen score function. To evaluate the quality of the selected instruction, we evaluate the zero-shot performance of another LLM following the selected instruction. Experiments on 24 NLP tasks show that our automatically generated instructions outperform the prior LLM baseline by a large margin and achieve better or comparable performance to the instructions generated by human annotators on 19/24 tasks. We conduct extensive qualitative and quantitative analyses to explore the performance of APE. We show that APE-engineered prompts can be applied to steer models toward truthfulness and/or informativeness, as well as to improve few-shot learning performance by simply prepending them to standard in-context learning prompts. Please check out our webpage at https://sites.google.com/view/automatic-prompt-engineer."
                },
                {
                    "title": "PromptSource: An Integrated Development Environment and Repository for Natural Language Prompts",
                    "abstract": "PromptSource is a system for creating, sharing, and using natural language prompts. Prompts are functions that map an example from a dataset to a natural language input and target output. Using prompts to train and query language models is an emerging area in NLP that requires new tools that let users develop and refine these prompts collaboratively. PromptSource addresses the emergent challenges in this new setting with (1) a templating language for defining data-linked prompts, (2) an interface that lets users quickly iterate on prompt development by observing outputs of their prompts on many examples, and (3) a community-driven set of guidelines for contributing new prompts to a common pool. Over 2,000 prompts for roughly 170 datasets are already available in PromptSource. PromptSource is available at https://github.com/bigscience-workshop/promptsource."
                },
                {
                    "title": "ZeroCap: Zero-Shot Image-to-Text Generation for Visual-Semantic Arithmetic",
                    "abstract": "Recent text-to-image matching models apply contrastive learning to large corpora of uncurated pairs of images and sentences. While such models can provide a powerful score for matching and subsequent zero-shot tasks, they are not capable of generating caption given an image. In this work, we repurpose such models to generate a descriptive text given an image at inference time, without any further training or tuning step. This is done by combining the visual-semantic model with a large language model, benefiting from the knowledge in both web-scale models. The resulting captions are much less restrictive than those obtained by supervised captioning methods. Moreover, as a zero-shot learning method, it is extremely flexible and we demonstrate its ability to perform image arithmetic in which the inputs can be either images or text and the output is a sentence. This enables novel high-level vision capabilities such as comparing two images or solving visual analogy tests. Our code is available at: https://github.com/YoadTew/zero-shot-image-to-text."
                },
                {
                    "title": "Robust fine-tuning of zero-shot models",
                    "abstract": "Large pre-trained models such as CLIP or ALIGN offer consistent accuracy across a range of data distributions when performing zero-shot inference (i.e., without fine-tuning on a specific dataset). Although existing fine-tuning methods substantially improve accuracy on a given target distribution, they often reduce robustness to distribution shifts. We address this tension by introducing a simple and effective method for improving robustness while fine-tuning: ensembling the weights of the zero-shot and fine-tuned models (WiSE-FT). Compared to standard fine-tuning, WiSE-FT provides large accuracy improvements under distribution shift, while preserving high accuracy on the target distribution. On ImageNet and five derived distribution shifts, WiSE-FT improves accuracy under distribution shift by 4 to 6 percentage points (pp) over prior work while increasing ImageNet accuracy by 1.6 pp. WiSE-FT achieves similarly large robustness gains (2 to 23 pp) on a diverse set of six further distribution shifts, and accuracy gains of 0.8 to 3.3 pp compared to standard fine-tuning on commonly used transfer learning datasets. These improvements come at no additional computational cost during fine-tuning or inference."
                },
                {
                    "title": "Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing",
                    "abstract": "This article surveys and organizes research works in a new paradigm in natural language processing, which we dub \u201cprompt-based learning.\u201d Unlike traditional supervised learning, which trains a model to take in an input x and predict an output y as P(y|x), prompt-based learning is based on language models that model the probability of text directly. To use these models to perform prediction tasks, the original input x is modified using a template into a textual string prompt x\u2032 that has some unfilled slots, and then the language model is used to probabilistically fill the unfilled information to obtain a final string x\u0302, from which the final output y can be derived. This framework is powerful and attractive for a number of reasons: It allows the language model to be pre-trained on massive amounts of raw text, and by defining a new prompting function the model is able to perform few-shot or even zero-shot learning, adapting to new scenarios with few or no labeled data. In this article, we introduce the basics of this promising paradigm, describe a unified set of mathematical notations that can cover a wide variety of existing work, and organize existing work along several dimensions, e.g., the choice of pre-trained language models, prompts, and tuning strategies. To make the field more accessible to interested beginners, we not only make a systematic review of existing works and a highly structured typology of prompt-based concepts but also release other resources, e.g., a website NLPedia\u2013Pretrain including constantly updated survey and paperlist."
                },
                {
                    "title": "The Power of Scale for Parameter-Efficient Prompt Tuning",
                    "abstract": "In this work, we explore \u201cprompt tuning,\u201d a simple yet effective mechanism for learning \u201csoft prompts\u201d to condition frozen language models to perform specific downstream tasks. Unlike the discrete text prompts used by GPT-3, soft prompts are learned through backpropagation and can be tuned to incorporate signals from any number of labeled examples. Our end-to-end learned approach outperforms GPT-3\u2019s few-shot learning by a large margin. More remarkably, through ablations on model size using T5, we show that prompt tuning becomes more competitive with scale: as models exceed billions of parameters, our method \u201ccloses the gap\u201d and matches the strong performance of model tuning (where all model weights are tuned). This finding is especially relevant because large models are costly to share and serve and the ability to reuse one frozen model for multiple downstream tasks can ease this burden. Our method can be seen as a simplification of the recently proposed \u201cprefix tuning\u201d of Li and Liang (2021) and we provide a comparison to this and other similar approaches. Finally, we show that conditioning a frozen model with soft prompts confers benefits in robustness to domain transfer and enables efficient \u201cprompt ensembling.\u201d We release code and model checkpoints to reproduce our experiments."
                },
                {
                    "title": "How many data points is a prompt worth?",
                    "abstract": "When fine-tuning pretrained models for classification, researchers either use a generic model head or a task-specific prompt for prediction. Proponents of prompting have argued that prompts provide a method for injecting task-specific guidance, which is beneficial in low-data regimes. We aim to quantify this benefit through rigorous testing of prompts in a fair setting: comparing prompted and head-based fine-tuning in equal conditions across many tasks and data sizes. By controlling for many sources of advantage, we find that prompting does indeed provide a benefit, and that this benefit can be quantified per task. Results show that prompting is often worth 100s of data points on average across classification tasks."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Understanding deep learning (still) requires rethinking generalization",
                    "abstract": "Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small gap between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family or to the regularization techniques used during training. Through extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. We interpret our experimental findings by comparison with traditional models. We supplement this republication with a new section at the end summarizing recent progresses in the field since the original version of this paper."
                },
                {
                    "title": "Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision",
                    "abstract": "Pre-trained representations are becoming crucial for many NLP and perception tasks. While representation learning in NLP has transitioned to training on raw text without human annotations, visual and vision-language representations still rely heavily on curated training datasets that are expensive or require expert knowledge. For vision applications, representations are mostly learned using datasets with explicit class labels such as ImageNet or OpenImages. For vision-language, popular datasets like Conceptual Captions, MSCOCO, or CLIP all involve a non-trivial data collection (and cleaning) process. This costly curation process limits the size of datasets and hence hinders the scaling of trained models. In this paper, we leverage a noisy dataset of over one billion image alt-text pairs, obtained without expensive filtering or post-processing steps in the Conceptual Captions dataset. A simple dual-encoder architecture learns to align visual and language representations of the image and text pairs using a contrastive loss. We show that the scale of our corpus can make up for its noise and leads to state-of-the-art representations even with such a simple learning scheme. Our visual representation achieves strong performance when transferred to classification tasks such as ImageNet and VTAB. The aligned visual and language representations enables zero-shot image classification and also set new state-of-the-art results on Flickr30K and MSCOCO image-text retrieval benchmarks, even when compared with more sophisticated cross-attention models. The representations also enable cross-modality search with complex text and text + image queries."
                },
                {
                    "title": "Making Pre-trained Language Models Better Few-shot Learners",
                    "abstract": "The recent GPT-3 model (Brown et al., 2020) achieves remarkable few-shot performance solely by leveraging a natural-language prompt and a few task demonstrations as input context. Inspired by their findings, we study few-shot learning in a more practical scenario, where we use smaller language models for which fine-tuning is computationally efficient. We present LM-BFF\u2014better few-shot fine-tuning of language models\u2014a suite of simple and complementary techniques for fine-tuning language models on a small number of annotated examples. Our approach includes (1) prompt-based fine-tuning together with a novel pipeline for automating prompt generation; and (2) a refined strategy for dynamically and selectively incorporating demonstrations into each context. Finally, we present a systematic evaluation for analyzing few-shot performance on a range of NLP tasks, including classification and regression. Our experiments demonstrate that our methods combine to dramatically outperform standard fine-tuning procedures in this low resource setting, achieving up to 30% absolute improvement, and 11% on average across all tasks. Our approach makes minimal assumptions on task resources and domain expertise, and hence constitutes a strong task-agnostic method for few-shot learning."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "In Search of Robust Measures of Generalization",
                    "abstract": "One of the principal scientific challenges in deep learning is explaining generalization, i.e., why the particular way the community now trains networks to achieve small training error also leads to small error on held-out data from the same population. It is widely appreciated that some worst-case theories -- such as those based on the VC dimension of the class of predictors induced by modern neural network architectures -- are unable to explain empirical performance. A large volume of work aims to close this gap, primarily by developing bounds on generalization error, optimization error, and excess risk. When evaluated empirically, however, most of these bounds are numerically vacuous. Focusing on generalization bounds, this work addresses the question of how to evaluate such bounds empirically. Jiang et al. (2020) recently described a large-scale empirical study aimed at uncovering potential causal relationships between bounds/measures and generalization. Building on their study, we highlight where their proposed methods can obscure failures and successes of generalization measures in explaining generalization. We argue that generalization measures should instead be evaluated within the framework of distributional robustness."
                },
                {
                    "title": "On the role of data in PAC-Bayes bounds",
                    "abstract": "The dominant term in PAC-Bayes bounds is often the Kullback--Leibler divergence between the posterior and prior. For so-called linear PAC-Bayes risk bounds based on the empirical risk of a fixed posterior kernel, it is possible to minimize the expected value of the bound by choosing the prior to be the expected posterior, which we call the oracle prior on the account that it is distribution dependent. In this work, we show that the bound based on the oracle prior can be suboptimal: In some cases, a stronger bound is obtained by using a data-dependent oracle prior, i.e., a conditional expectation of the posterior, given a subset of the training data that is then excluded from the empirical risk term. While using data to learn a prior is a known heuristic, its essential role in optimal bounds is new. In fact, we show that using data can mean the difference between vacuous and nonvacuous bounds. We apply this new principle in the setting of nonconvex learning, simulating data-dependent oracle priors on MNIST and Fashion MNIST with and without held-out data, and demonstrating new nonvacuous bounds in both cases."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Fantastic Generalization Measures and Where to Find Them",
                    "abstract": "Generalization of deep networks has been of great interest in recent years, resulting in a number of theoretically and empirically motivated complexity measures. However, most papers proposing such measures study only a small set of models, leaving open the question of whether the conclusion drawn from those experiments would remain valid in other settings. We present the first large scale study of generalization in deep networks. We investigate more then 40 complexity measures taken from both theoretical bounds and empirical studies. We train over 10,000 convolutional networks by systematically varying commonly used hyperparameters. Hoping to uncover potentially causal relationships between each measure and generalization, we analyze carefully controlled experiments and show surprising failures of some measures as well as promising measures for further research."
                },
                {
                    "title": "Uniform convergence may be unable to explain generalization in deep learning",
                    "abstract": "Aimed at explaining the surprisingly good generalization behavior of overparameterized deep networks, recent works have developed a variety of generalization bounds for deep learning, all based on the fundamental learning-theoretic technique of uniform convergence. While it is well-known that many of these existing bounds are numerically large, through numerous experiments, we bring to light a more concerning aspect of these bounds: in practice, these bounds can {\\em increase} with the training dataset size. Guided by our observations, we then present examples of overparameterized linear classifiers and neural networks trained by gradient descent (GD) where uniform convergence provably cannot ``explain generalization'' -- even if we take into account the implicit bias of GD {\\em to the fullest extent possible}. More precisely, even if we consider only the set of classifiers output by GD, which have test errors less than some small $\\epsilon$ in our settings, we show that applying (two-sided) uniform convergence on this set of classifiers will yield only a vacuous generalization guarantee larger than $1-\\epsilon$. Through these findings, we cast doubt on the power of uniform convergence-based generalization bounds to provide a complete picture of why overparameterized deep networks generalize well."
                },
                {
                    "title": "Generalization in Deep Networks: The Role of Distance from Initialization",
                    "abstract": "Why does training deep neural networks using stochastic gradient descent (SGD) result in a generalization error that does not worsen with the number of parameters in the network? To answer this question, we advocate a notion of effective model capacity that is dependent on {\\em a given random initialization of the network} and not just the training algorithm and the data distribution. We provide empirical evidences that demonstrate that the model capacity of SGD-trained deep networks is in fact restricted through implicit regularization of {\\em the $\\ell_2$ distance from the initialization}. We also provide theoretical arguments that further highlight the need for initialization-dependent notions of model capacity. We leave as open questions how and why distance from initialization is regularized, and whether it is sufficient to explain generalization."
                },
                {
                    "title": "Non-vacuous Generalization Bounds at the ImageNet Scale: a PAC-Bayesian Compression Approach",
                    "abstract": "Modern neural networks are highly overparameterized, with capacity to substantially overfit to training data. Nevertheless, these networks often generalize well in practice. It has also been observed that trained networks can often be \"compressed\" to much smaller representations. The purpose of this paper is to connect these two empirical observations. Our main technical result is a generalization bound for compressed networks based on the compressed size. Combined with off-the-shelf compression algorithms, the bound leads to state of the art generalization guarantees; in particular, we provide the first non-vacuous generalization guarantees for realistic architectures applied to the ImageNet classification problem. As additional evidence connecting compression and generalization, we show that compressibility of models that tend to overfit is limited: We establish an absolute limit on expected compressibility as a function of expected generalization error, where the expectations are over the random choice of training examples. The bounds are complemented by empirical results that show an increase in overfitting implies an increase in the number of bits required to describe a trained network."
                },
                {
                    "title": "Functional Map of the World",
                    "abstract": "We present a new dataset, Functional Map of the World (fMoW), which aims to inspire the development of machine learning models capable of predicting the functional purpose of buildings and land use from temporal sequences of satellite images and a rich set of metadata features. The metadata provided with each image enables reasoning about location, time, sun angles, physical sizes, and other features when making predictions about objects in the image. Our dataset consists of over 1 million images from over 200 countries. For each image, we provide at least one bounding box annotation containing one of 63 categories, including a \"false detection\" category. We present an analysis of the dataset along with baseline approaches that reason about metadata and temporal views. Our data, code, and pretrained models have been made publicly available."
                },
                {
                    "title": "A PAC-Bayesian Approach to Spectrally-Normalized Margin Bounds for Neural Networks",
                    "abstract": "We present a generalization bound for feedforward neural networks in terms of the product of the spectral norm of the layers and the Frobenius norm of the weights. The generalization bound is derived using a PAC-Bayes analysis."
                },
                {
                    "title": "Risk Bounds for Transferring Representations With and Without Fine-Tuning",
                    "abstract": "A popular machine learning strategy is the transfer of a representation (i.e. a feature extraction function) learned on a source task to a target task. Examples include the re-use of neural network weights or word embeddings. We develop suf\ufb01cient conditions for the success of this approach. If the representation learned from the source task is \ufb01xed, we identify conditions on how the tasks relate to obtain an upper bound on target task risk via a VC dimension-based argument. We then consider using the representation from the source task to construct a prior, which is \ufb01ne-tuned using target task data. We give a PAC-Bayes target task risk bound in this setting under suitable conditions. We show examples of our bounds using feedforward neural networks. Our results motivate a practical approach to weight transfer, which we validate with experiments."
                },
                {
                    "title": "Exploring Generalization in Deep Learning",
                    "abstract": "With a goal of understanding what drives generalization in deep networks, we consider several recently suggested explanations, including norm-based control, sharpness and robustness. We study how these measures can ensure generalization, highlighting the importance of scale normalization, and making a connection between sharpness and PAC-Bayes theory. We then investigate how well the measures explain different observed phenomena."
                },
                {
                    "title": "Spectrally-normalized margin bounds for neural networks",
                    "abstract": "This paper presents a margin-based multiclass generalization bound for neural networks that scales with their margin-normalized \"spectral complexity\": their Lipschitz constant, meaning the product of the spectral norms of the weight matrices, times a certain correction factor. This bound is empirically investigated for a standard AlexNet network trained with SGD on the mnist and cifar10 datasets, with both original and random labels; the bound, the Lipschitz constants, and the excess risks are all in direct correlation, suggesting both that SGD selects predictors whose complexity scales with the difficulty of the learning task, and secondly that the presented bound is sensitive to this complexity."
                },
                {
                    "title": "Deep Hashing Network for Unsupervised Domain Adaptation",
                    "abstract": "In recent years, deep neural networks have emerged as a dominant machine learning tool for a wide variety of application domains. However, training a deep neural network requires a large amount of labeled data, which is an expensive process in terms of time, labor and human expertise. Domain adaptation or transfer learning algorithms address this challenge by leveraging labeled data in a different, but related source domain, to develop a model for the target domain. Further, the explosive growth of digital data has posed a fundamental challenge concerning its storage and retrieval. Due to its storage and retrieval efficiency, recent years have witnessed a wide application of hashing in a variety of computer vision applications. In this paper, we first introduce a new dataset, Office-Home, to evaluate domain adaptation algorithms. The dataset contains images of a variety of everyday objects from multiple domains. We then propose a novel deep learning framework that can exploit labeled source data and unlabeled target data to learn informative hash codes, to accurately classify unseen target data. To the best of our knowledge, this is the first research effort to exploit the feature learning capabilities of deep neural networks to learn representative hash codes to address the domain adaptation problem. Our extensive empirical studies on multiple transfer tasks corroborate the usefulness of the framework in learning efficient hash codes which outperform existing competitive baselines for unsupervised domain adaptation."
                },
                {
                    "title": "Computing Nonvacuous Generalization Bounds for Deep (Stochastic) Neural Networks with Many More Parameters than Training Data",
                    "abstract": "One of the defining properties of deep learning is that models are chosen to have many more parameters than available training data. In light of this capacity for overfitting, it is remarkable that simple algorithms like SGD reliably return solutions with low test error. One roadblock to explaining these phenomena in terms of implicit regularization, structural properties of the solution, and/or easiness of the data is that many learning bounds are quantitatively vacuous when applied to networks learned by SGD in this \"deep learning\" regime. Logically, in order to explain generalization, we need nonvacuous bounds. We return to an idea by Langford and Caruana (2001), who used PAC-Bayes bounds to compute nonvacuous numerical bounds on generalization error for stochastic two-layer two-hidden-unit neural networks via a sensitivity analysis. By optimizing the PAC-Bayes bound directly, we are able to extend their approach and obtain nonvacuous generalization bounds for deep stochastic neural network classifiers with millions of parameters trained on only tens of thousands of examples. We connect our findings to recent and old work on flat minima and MDL-based explanations of generalization."
                },
                {
                    "title": "Norm-Based Capacity Control in Neural Networks",
                    "abstract": "We investigate the capacity, convexity and characterization of a general family of norm-constrained feed-forward networks."
                },
                {
                    "title": "In Search of the Real Inductive Bias: On the Role of Implicit Regularization in Deep Learning",
                    "abstract": "We present experiments demonstrating that some other form of capacity control, different from network size, plays a central role in learning multilayer feed-forward networks. We argue, partially through analogy to matrix factorization, that this is an inductive bias that can help shed light on deep learning."
                },
                {
                    "title": "Multiclass Learnability and the ERM principle",
                    "abstract": "We study the sample complexity of multiclass prediction in several learning settings. For the PAC setting our analysis reveals a surprising phenomenon: In sharp contrast to binary classification, we show that there exist multiclass hypothesis classes for which some Empirical Risk Minimizers (ERM learners) have lower sample complexity than others. Furthermore, there are classes that are learnable by some ERM learners, while other ERM learners will fail to learn them. We propose a principle for designing good ERM learners, and use this principle to prove tight bounds on the sample complexity of learning symmetric multiclass hypothesis classes--classes that are invariant under permutations of label names. We further provide a characterization of mistake and regret bounds for multiclass learning in the online setting and the bandit setting, using new generalizations of Littlestone's dimension."
                },
                {
                    "title": "(Not) Bounding the True Error",
                    "abstract": "We present a new approach to bounding the true error rate of a continuous valued classifier based upon PAC-Bayes bounds. The method first constructs a distribution over classifiers by determining how sensitive each parameter in the model is to noise. The true error rate of the stochastic classifier found with the sensitivity analysis can then be tightly bounded using a PAC-Bayes bound. In this paper we demonstrate the method on artificial neural networks with results of a 2 - 3 order of magnitude improvement vs. the best deterministic neural net bounds."
                },
                {
                    "title": "PAC-Bayesian model averaging",
                    "abstract": "PAC-Bayesian learning methods combine the informative priors of Bayesian methods with distribution-free PAC guarantees. Building on earlier methods for PAC-Bayesian model selection, this paper presents a method for PACBayesian model averaging. The method constructs an optimized weighted mixture of concepts analogous to a Bayesian posterior distribution. Although the main result is stated for bounded loss, a preliminary analysis for unbounded loss is also given."
                },
                {
                    "title": "Principles of Risk Minimization for Learning Theory",
                    "abstract": "Learning is posed as a problem of function estimation, for which two principles of solution are considered: empirical risk minimization and structural risk minimization. These two principles are applied to two different statements of the function estimation problem: global and local. Systematic improvements in prediction power are illustrated in application to zip-code recognition."
                },
                {
                    "title": "Prefix-Tuning: Optimizing Continuous Prompts for Generation",
                    "abstract": "Fine-tuning is the de facto way of leveraging large pretrained language models for downstream tasks. However, fine-tuning modifies all the language model parameters and therefore necessitates storing a full copy for each task. In this paper, we propose prefix-tuning, a lightweight alternative to fine-tuning for natural language generation tasks, which keeps language model parameters frozen and instead optimizes a sequence of continuous task-specific vectors, which we call the prefix. Prefix-tuning draws inspiration from prompting for language models, allowing subsequent tokens to attend to this prefix as if it were \u201cvirtual tokens\u201d. We apply prefix-tuning to GPT-2 for table-to-text generation and to BART for summarization. We show that by learning only 0.1% of the parameters, prefix-tuning obtains comparable performance in the full data setting, outperforms fine-tuning in low-data settings, and extrapolates better to examples with topics that are unseen during training."
                },
                {
                    "title": "A General Framework for the Disintegration of PAC-Bayesian Bounds",
                    "abstract": "PAC-Bayesian bounds are known to be tight and in-formative when studying the generalization ability of randomized classi\ufb01ers. However, when applied to some family of deterministic models such as neural networks, they require a loose and costly derandomization step. As an alternative to this step, we introduce new PAC-Bayesian generalization bounds that have the originality to provide disintegrated bounds, i.e. , they give guarantees over one single hypothesis instead of the usual averaged analysis. Our bounds are easily optimizable and can be used to design learning algorithms. We illustrate the interest of our result on neural networks and show a signi\ufb01cant practical improvement over the state-of-the-art framework."
                },
                {
                    "title": "Understanding Machine Learning - From Theory to Algorithms",
                    "abstract": "Machine learning is one of the fastest growing areas of computer science, with far-reaching applications. The aim of this textbook is to introduce machine learning, and the algorithmic paradigms it offers, in a principled way. The book provides an extensive theoretical account of the fundamental ideas underlying machine learning and the mathematical derivations that transform these principles into practical algorithms. Following a presentation of the basics of the field, the book covers a wide array of central topics that have not been addressed by previous textbooks. These include a discussion of the computational complexity of learning and the concepts of convexity and stability ; important algorithmic paradigms including stochastic gradient descent, neural networks, and structured output learning ; and emerging theoretical concepts such as the PACBayes approach and compression-based bounds. Designed for an advanced undergraduate or beginning graduate course, the text makes the fundamentals and algorithms of machine learning accessible to students and nonexpert readers in statistics, computer science, mathematics, and engineering."
                },
                {
                    "title": "Monocular 3 D tracking of deformable surfaces using sequential second order cone programming",
                    "abstract": "Article history: Received 7 July 2008 Received in revised form 19 March 2009 Accepted 29 June 2009"
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can classical PAC-Bayes bounds be effectively applied to understand the generalization ability of pretrained models using prompt engineering in deep learning?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the theoretical underpinnings of why prompt engineering works effectively in deep learning, despite the lack of understanding surrounding its mechanisms. By providing tighter generalization bounds, this research could lead to more reliable and interpretable models, enhancing the development of pretrained models and their applications across various tasks. Furthermore, it could inspire future research to explore the intersection of classical statistical learning theory and modern deep learning practices, potentially leading to new methodologies that improve model performance and robustness.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities of applying classical PAC-Bayes bounds to modern deep learning architectures, particularly when dealing with large hypothesis spaces defined by prompts. Naive approaches may fail because they do not account for the unique characteristics of pretrained models and the specific nature of prompt engineering. Additionally, the theoretical foundations of generalization in deep learning are still not fully understood, making it difficult to derive bounds that are both tight and practically useful. Overcoming these obstacles requires a deep understanding of both the statistical properties of PAC-Bayes bounds and the behavior of large language models in the context of prompt-based learning.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on deriving generalization bounds that are either too loose or not applicable to the specific context of prompt engineering in pretrained models. Many existing solutions have not adequately addressed the unique challenges posed by the discrete hypothesis classes defined by prompts. Barriers such as the lack of a clear theoretical framework for understanding the generalization of pretrained models and the empirical nature of prompt engineering have hindered progress. This research aims to fill these gaps by leveraging classical PAC-Bayes bounds in a novel way that directly addresses the intricacies of prompt-based learning.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves applying classical PAC-Bayes bounds to the discrete hypothesis class defined by prompts, using a prior informed by a large language model. The dataset will consist of images from the ImageNet classification task, and the metric for evaluation will be the generalization error rate. The expected outcome is to derive a significantly tighter generalization bound, demonstrating that the"
            }
        },
        "author_data": {
            "e0dbc23a-3196-4219-8e56-6d66840447c1": {
                "pk": "e0dbc23a-3196-4219-8e56-6d66840447c1",
                "name": "Victor Akinwande",
                "collaborators": [
                    "Bertie Vidgen",
                    "Adarsh Agrawal",
                    "Ahmed M. Ahmed",
                    "Namir Al-nuaimi",
                    "Najla Alfaraj",
                    "Elie Alhajjar",
                    "Lora Aroyo",
                    "Trupti Bavalatti",
                    "Borhane Blili-Hamelin",
                    "K. Bollacker",
                    "Rishi Bomassani",
                    "Marisa Ferrara Boston",
                    "Sim'eon Campos",
                    "Kal Chakra",
                    "Canyu Chen",
                    "Cody Coleman",
                    "Zacharie Delpierre Coudert",
                    "Leon Derczynski",
                    "Debojyoti Dutta",
                    "Ian Eisenberg",
                    "J. Ezick",
                    "Heather Frase",
                    "Brian Fuller",
                    "Ram Gandikota",
                    "Agasthya Gangavarapu",
                    "Ananya Gangavarapu",
                    "J. Gealy",
                    "Rajat Ghosh",
                    "James Goel",
                    "Usman Gohar",
                    "Sujata Goswami",
                    "Scott A. Hale",
                    "Wiebke Hutiri",
                    "Joseph Marvin Imperial",
                    "Surgan Jandial",
                    "Nicholas C. Judd",
                    "Felix Juefei-Xu",
                    "Foutse Khomh",
                    "B. Kailkhura",
                    "Hannah Rose Kirk",
                    "Kevin Klyman",
                    "Chris Knotz",
                    "Michael Kuchnik",
                    "Shachi H. Kumar",
                    "Chris Lengerich",
                    "Bo Li",
                    "Zeyi Liao",
                    "E. Long",
                    "Victor Lu",
                    "Yifan Mai",
                    "P. Mammen",
                    "Kelvin Manyeki",
                    "Sean McGregor",
                    "Virendra Mehta",
                    "Shafee Mohammed",
                    "Emanuel Moss",
                    "L. Nachman",
                    "Dinesh Jinenhally Naganna",
                    "Amin Nikanjam",
                    "Besmira Nushi",
                    "Luis Oala",
                    "Iftach Orr",
                    "Alicia Parrish",
                    "\u00c7igdem Patlak",
                    "William Pietri",
                    "Forough Poursabzi-Sangdeh",
                    "Eleonora Presani",
                    "Fabrizio Puletti",
                    "Paul R\u00f6ttger",
                    "Saurav Sahay",
                    "Tim Santos",
                    "Nino Scherrer",
                    "Alice Schoenauer Sebag",
                    "P. Schramowski",
                    "Abolfazl Shahbazi",
                    "Vin Sharma",
                    "Xudong Shen",
                    "Vamsi Sistla",
                    "Leonard Tang",
                    "Davide Testuggine",
                    "Vithursan Thangarasa",
                    "E. A. Watkins",
                    "Rebecca Weiss",
                    "Christoper A. Welty",
                    "Tyler Wilbers",
                    "Adina Williams",
                    "Carole-Jean Wu",
                    "Poonam Yadav",
                    "Xianjun Yang",
                    "Yi Zeng",
                    "Wenhui Zhang",
                    "Fedor Zhdanov",
                    "Jiacheng Zhu",
                    "Percy Liang",
                    "Peter Mattson",
                    "J. Vanschoren",
                    "J. Z. Kolter"
                ],
                "domain": [
                    "AI Safety",
                    "Causal Discovery",
                    "Machine Learning",
                    "Benchmarking"
                ],
                "publications": [
                    {
                        "title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons",
                        "abstract": "This paper introduces v0.5 of the AI Safety Benchmark, which has been created by the MLCommons AI Safety Working Group. The AI Safety Benchmark has been designed to assess the safety risks of AI systems that use chat-tuned language models. We introduce a principled approach to specifying and constructing the benchmark, which for v0.5 covers only a single use case (an adult chatting to a general-purpose assistant in English), and a limited set of personas (i.e., typical users, malicious users, and vulnerable users). We created a new taxonomy of 13 hazard categories, of which 7 have tests in the v0.5 benchmark. We plan to release version 1.0 of the AI Safety Benchmark by the end of 2024. The v1.0 benchmark will provide meaningful insights into the safety of AI systems. However, the v0.5 benchmark should not be used to assess the safety of AI systems. We have sought to fully document the limitations, flaws, and challenges of v0.5. This release of v0.5 of the AI Safety Benchmark includes (1) a principled approach to specifying and constructing the benchmark, which comprises use cases, types of systems under test (SUTs), language and context, personas, tests, and test items; (2) a taxonomy of 13 hazard categories with definitions and subcategories; (3) tests for seven of the hazard categories, each comprising a unique set of test items, i.e., prompts. There are 43,090 test items in total, which we created with templates; (4) a grading system for AI systems against the benchmark; (5) an openly available platform, and downloadable tool, called ModelBench that can be used to evaluate the safety of AI systems on the benchmark; (6) an example evaluation report which benchmarks the performance of over a dozen openly available chat-tuned language models; (7) a test specification for the benchmark."
                    },
                    {
                        "title": "AcceleratedLiNGAM: Learning Causal DAGs at the speed of GPUs",
                        "abstract": "Existing causal discovery methods based on combinatorial optimization or search are slow, prohibiting their application on large-scale datasets. In response, more recent methods attempt to address this limitation by formulating causal discovery as structure learning with continuous optimization but such approaches thus far provide no statistical guarantees. In this paper, we show that by efficiently parallelizing existing causal discovery methods, we can in fact scale them to thousands of dimensions, making them practical for substantially larger-scale problems. In particular, we parallelize the LiNGAM method, which is quadratic in the number of variables, obtaining up to a 32-fold speed-up on benchmark datasets when compared with existing sequential implementations. Specifically, we focus on the causal ordering subprocedure in DirectLiNGAM and implement GPU kernels to accelerate it. This allows us to apply DirectLiNGAM to causal inference on large-scale gene expression data with genetic interventions yielding competitive results compared with specialized continuous optimization methods, and Var-LiNGAM for causal discovery on U.S. stock data."
                    }
                ]
            },
            "a14a3cf8-5f07-4380-986e-ee603425ed32": {
                "pk": "a14a3cf8-5f07-4380-986e-ee603425ed32",
                "name": "Yiding Jiang",
                "collaborators": [
                    "Yutong He",
                    "Alexander Robey",
                    "Naoki Murata",
                    "J. Williams",
                    "George J. Pappas",
                    "Hamed Hassani",
                    "Yuki Mitsufuji",
                    "Ruslan Salakhutdinov",
                    "J. Z. Kolter",
                    "AI Sony",
                    "Sony Group Corporation",
                    "Bosch Center",
                    "Allan Zhou",
                    "Zhili Feng",
                    "Sadhika Malladi",
                    "J. Kolter"
                ],
                "domain": [
                    "Generative Models",
                    "Prompt Engineering",
                    "Foundation Models",
                    "Data Optimization"
                ],
                "publications": [
                    {
                        "title": "Automated Black-box Prompt Engineering for Personalized Text-to-Image Generation",
                        "abstract": "Prompt engineering is effective for controlling the output of text-to-image (T2I) generative models, but it is also laborious due to the need for manually crafted prompts. This challenge has spurred the development of algorithms for automated prompt generation. However, these methods often struggle with transferability across T2I models, require white-box access to the underlying model, and produce non-intuitive prompts. In this work, we introduce PRISM, an algorithm that automatically identifies human-interpretable and transferable prompts that can effectively generate desired concepts given only black-box access to T2I models. Inspired by large language model (LLM) jailbreaking, PRISM leverages the in-context learning ability of LLMs to iteratively refine the candidate prompts distribution for given reference images. Our experiments demonstrate the versatility and effectiveness of PRISM in generating accurate prompts for objects, styles and images across multiple T2I models, including Stable Diffusion, DALL-E, and Midjourney."
                    },
                    {
                        "title": "Adaptive Data Optimization: Dynamic Sample Selection with Scaling Laws",
                        "abstract": "The composition of pretraining data is a key determinant of foundation models' performance, but there is no standard guideline for allocating a limited computational budget across different data sources. Most current approaches either rely on extensive experiments with smaller models or dynamic data adjustments that also require proxy models, both of which significantly increase the workflow complexity and computational overhead. In this paper, we introduce Adaptive Data Optimization (ADO), an algorithm that optimizes data distributions in an online fashion, concurrent with model training. Unlike existing techniques, ADO does not require external knowledge, proxy models, or modifications to the model update. Instead, ADO uses per-domain scaling laws to estimate the learning potential of each domain during training and adjusts the data mixture accordingly, making it more scalable and easier to integrate. Experiments demonstrate that ADO can achieve comparable or better performance than prior methods while maintaining computational efficiency across different computation scales, offering a practical solution for dynamically adjusting data distribution without sacrificing flexibility or increasing costs. Beyond its practical benefits, ADO also provides a new perspective on data collection strategies via scaling laws."
                    }
                ]
            },
            "01d0ee04-5776-487f-8277-54ca035aff49": {
                "pk": "01d0ee04-5776-487f-8277-54ca035aff49",
                "name": "Dylan Sam",
                "collaborators": [
                    "J. Z. Kolter",
                    "Rattana Pukdee",
                    "Yewon Byun",
                    "Maria-Florina Balcan",
                    "Pradeep Ravikumar",
                    "Alessio Mazzetto",
                    "E. Upfal",
                    "Stephen H. Bach",
                    "Samuel Sokota",
                    "C. S. D. Witt",
                    "Spencer Compton",
                    "Jakob Foerster",
                    "Devin Willmott",
                    "Jo\u00e3o Dias Semedo",
                    "J. Kolter",
                    "Michael Oberst",
                    "Zachary Chase Lipton",
                    "Bryan Wilder",
                    "Daniel P. Jeong",
                    "Min Bai",
                    "Tristan McKinney",
                    "Li Erran Li",
                    "Andrew Park",
                    "R. Alimo",
                    "Dounia Lakhmiri",
                    "Brian Kahovec",
                    "D. Divsalar",
                    "Cyrus Cousins"
                ],
                "domain": [
                    "Machine Learning",
                    "Weak Supervision",
                    "Bayesian Neural Networks",
                    "Vision-Language Models"
                ],
                "publications": [
                    {
                        "title": "Computing Low-Entropy Couplings for Large-Support Distributions",
                        "abstract": "Minimum-entropy coupling (MEC) -- the process of finding a joint distribution with minimum entropy for given marginals -- has applications in areas such as causality and steganography. However, existing algorithms are either computationally intractable for large-support distributions or limited to specific distribution types and sensitive to hyperparameter choices. This work addresses these limitations by unifying a prior family of iterative MEC (IMEC) approaches into a generalized partition-based formalism. From this framework, we derive a novel IMEC algorithm called ARIMEC, capable of handling arbitrary discrete distributions, and introduce a method to make IMEC robust to suboptimal hyperparameter settings. These innovations facilitate the application of IMEC to high-throughput steganography with language models, among other settings. Our codebase is available at https://github.com/ssokota/mec ."
                    },
                    {
                        "title": "Finetuning CLIP to Reason about Pairwise Differences",
                        "abstract": "Vision-language models (VLMs) such as CLIP are trained via contrastive learning between text and image pairs, resulting in aligned image and text embeddings that are useful for many downstream tasks. A notable drawback of CLIP, however, is that the resulting embedding space seems to lack some of the structure of their purely text-based alternatives. For instance, while text embeddings have been long noted to satisfy \\emph{analogies} in embedding space using vector arithmetic, CLIP has no such property. In this paper, we propose an approach to natively train CLIP in a contrastive manner to reason about differences in embedding space. We finetune CLIP so that the differences in image embedding space correspond to \\emph{text descriptions of the image differences}, which we synthetically generate with large language models on image-caption paired datasets. We first demonstrate that our approach yields significantly improved capabilities in ranking images by a certain attribute (e.g., elephants are larger than cats), which is useful in retrieval or constructing attribute-based classifiers, and improved zeroshot classification performance on many downstream image classification tasks. In addition, our approach enables a new mechanism for inference that we refer to as comparative prompting, where we leverage prior knowledge of text descriptions of differences between classes of interest, achieving even larger performance gains in classification. Finally, we illustrate that the resulting embeddings obey a larger degree of geometric properties in embedding space, such as in text-to-image generation."
                    },
                    {
                        "title": "Auditing Fairness under Unobserved Confounding",
                        "abstract": "A fundamental problem in decision-making systems is the presence of inequity across demographic lines. However, inequity can be difficult to quantify, particularly if our notion of equity relies on hard-to-measure notions like risk (e.g., equal access to treatment for those who would die without it). Auditing such inequity requires accurate measurements of individual risk, which is difficult to estimate in the realistic setting of unobserved confounding. In the case that these unobservables \"explain\" an apparent disparity, we may understate or overstate inequity. In this paper, we show that one can still give informative bounds on allocation rates among high-risk individuals, even while relaxing or (surprisingly) even when eliminating the assumption that all relevant risk factors are observed. We utilize the fact that in many real-world settings (e.g., the introduction of a novel treatment) we have data from a period prior to any allocation, to derive unbiased estimates of risk. We demonstrate the effectiveness of our framework on a real-world study of Paxlovid allocation to COVID-19 patients, finding that observed racial inequity cannot be explained by unobserved confounders of the same strength as important observed covariates."
                    },
                    {
                        "title": "Bayesian Neural Networks with Domain Knowledge Priors",
                        "abstract": "Bayesian neural networks (BNNs) have recently gained popularity due to their ability to quantify model uncertainty. However, specifying a prior for BNNs that captures relevant domain knowledge is often extremely challenging. In this work, we propose a framework for integrating general forms of domain knowledge (i.e., any knowledge that can be represented by a loss function) into a BNN prior through variational inference, while enabling computationally efficient posterior inference and sampling. Specifically, our approach results in a prior over neural network weights that assigns high probability mass to models that better align with our domain knowledge, leading to posterior samples that also exhibit this behavior. We show that BNNs using our proposed domain knowledge priors outperform those with standard priors (e.g., isotropic Gaussian, Gaussian process), successfully incorporating diverse types of prior information such as fairness, physics rules, and healthcare knowledge and achieving better predictive performance. We also present techniques for transferring the learned priors across different model architectures, demonstrating their broad utility across various settings."
                    },
                    {
                        "title": "Learning with Explanation Constraints",
                        "abstract": "As larger deep learning models are hard to interpret, there has been a recent focus on generating explanations of these black-box models. In contrast, we may have apriori explanations of how models should behave. In this paper, we formalize this notion as learning from explanation constraints and provide a learning theoretic framework to analyze how such explanations can improve the learning of our models. One may naturally ask,\"When would these explanations be helpful?\"Our first key contribution addresses this question via a class of models that satisfies these explanation constraints in expectation over new data. We provide a characterization of the benefits of these models (in terms of the reduction of their Rademacher complexities) for a canonical class of explanations given by gradient information in the settings of both linear models and two layer neural networks. In addition, we provide an algorithmic solution for our framework, via a variational approximation that achieves better performance and satisfies these constraints more frequently, when compared to simpler augmented Lagrangian methods to incorporate these explanations. We demonstrate the benefits of our approach over a large array of synthetic and real-world experiments."
                    },
                    {
                        "title": "Label Propagation with Weak Supervision",
                        "abstract": "Semi-supervised learning and weakly supervised learning are important paradigms that aim to reduce the growing demand for labeled data in current machine learning applications. In this paper, we introduce a novel analysis of the classical label propagation algorithm (LPA) (Zhu&Ghahramani, 2002) that moreover takes advantage of useful prior information, specifically probabilistic hypothesized labels on the unlabeled data. We provide an error bound that exploits both the local geometric properties of the underlying graph and the quality of the prior information. We also propose a framework to incorporate multiple sources of noisy information. In particular, we consider the setting of weak supervision, where our sources of information are weak labelers. We demonstrate the ability of our approach on multiple benchmark weakly supervised classification tasks, showing improvements upon existing semi-supervised and weakly supervised methods."
                    },
                    {
                        "title": "Improving self-supervised representation learning via sequential adversarial masking",
                        "abstract": "Recent methods in self-supervised learning have demonstrated that masking-based pretext tasks extend beyond NLP, serving as useful pretraining objectives in computer vision. However, existing approaches apply random or ad hoc masking strategies that limit the difficulty of the reconstruction task and, consequently, the strength of the learnt representations. We improve upon current state-of-the-art work in learning adversarial masks by proposing a new framework that generates masks in a sequential fashion with different constraints on the adversary. This leads to improvements in performance on various downstream tasks, such as classification on ImageNet100, STL10, and CIFAR10/100 and segmentation on Pascal VOC. Our results further demonstrate the promising capabilities of masking-based approaches for SSL in computer vision."
                    },
                    {
                        "title": "Losses over Labels: Weakly Supervised Learning via Direct Loss Construction",
                        "abstract": "Owing to the prohibitive costs of generating large amounts of labeled data, programmatic weak supervision is a growing paradigm within machine learning. In this setting, users design heuristics that provide noisy labels for subsets of the data. These weak labels are combined (typically via a graphical model) to form pseudolabels, which are then used to train a downstream model. In this work, we question a foundational premise of the typical weakly supervised learning pipeline: given that the heuristic provides all \u201clabel\u201d information, why do we need to generate pseudolabels at all? Instead, we propose to directly transform the heuristics themselves into corresponding loss functions that penalize differences between our model and the heuristic. By constructing losses directly from the heuristics, we can incorporate more information than is used in the standard weakly supervised pipeline, such as how the heuristics make their decisions, which explicitly informs feature selection during training. We call our method Losses over Labels (LoL) as it creates losses directly from heuristics without going through the intermediate step of a label. We show that LoL improves upon existing weak supervision methods on several benchmark text and image classification tasks and further demonstrate that incorporating gradient information leads to better performance on almost every task."
                    },
                    {
                        "title": "Semi-Supervised Aggregation of Dependent Weak Supervision Sources With Performance Guarantees",
                        "abstract": "We develop a novel method that provides theoretical guarantees for learning from weak labelers without the (mostly unrealistic) assumption that the errors of the weak label-ers are independent or come from a particular family of distributions. We show a rigorous technique for e\ufb03ciently selecting small subsets of the labelers so that a majority vote from such subsets has a provably low error rate. We explore several extensions of this method and provide experimental results over a range of labeled data set sizes on 45 image classi\ufb01cation tasks. Our performance-guaranteed methods consistently match the best performing alternative, which varies based on problem di\ufb03culty. On tasks with accurate weak labelers, our methods are on average 3 percentage points more accurate than the state-of-the-art adversarial method. On tasks with inaccurate weak labelers, our methods are on average 15 percentage points more accurate than the semi-supervised Dawid-Skene model (which assumes independence)."
                    },
                    {
                        "title": "Automated Data Accountability for Missions in Mars Rover Data",
                        "abstract": "This paper proposes an automated solution system to assist with Real-Time Operations and automatically identify and report on issues with data transfer, archive, and manipulation throughout the Ground Data System (GDS) process. As the Mars Curiosity Rover transmits data to the JPL Ground Data System (GDS), it frequently observes data loss and corruption, requiring re-transmits from the rover and Ground Data System Analysts (GDSA) to monitor the downlink process. As new missions are launched, the GDSA team redistributes analysts to these new missions, causing shortages in previous missions. The prior state of GDS issue detection and resolution was largely manual. GDSAs receive email alerts when something goes wrong, but it's not always clear what the exact problem is or how to fix it. This paper presents machine learning and deep learning based approaches to automate and optimize the detection of data loss. We first created a pipeline to automatically accumulate data from the telemetry databases (MAROS, Telemetry Data Storage, and GDS Elastic Search Database) in the downlink process. With our newly created datasets, we perform feature selection to supplement the GDSA understanding of the downlink process and provide supplemental analysis on the importance of different features. We implemented various supervised machine learning-based models and evaluate their accuracies to identify a downlink process is complete or incomplete. We utilize fast hyperparameter optimization methods that allow our models to quickly be re-trained, allowing them to quickly be tuned and optimized on daily incoming data in near real time."
                    },
                    {
                        "title": "Adversarial Multi Class Learning under Weak Supervision with Performance Guarantees",
                        "abstract": "We develop a rigorous approach for using a set of arbitrarily correlated weak supervision sources in order to solve a multiclass classi\ufb01cation task when only a very small set of labeled data is available. Our learning algorithm provably converges to a model that has minimum empirical risk with respect to an adversarial choice over feasible labelings for a set of unlabeled data, where the feasibility of a labeling is computed through constraints de\ufb01ned by rigorously estimated statistics of the weak supervision sources. We show theoretical guarantees for this approach that depend on the information provided by the weak supervision sources. Notably, this method does not require the weak supervision sources to have the same labeling space as the multiclass classi\ufb01cation task. We demonstrate the effectiveness of our approach with experiments on various image classi\ufb01cation tasks."
                    }
                ]
            },
            "8239576c-530a-4b4b-9203-d3751a12d9dc": {
                "pk": "8239576c-530a-4b4b-9203-d3751a12d9dc",
                "name": "J. Zico Kolter",
                "collaborators": [
                    "Zhengyang Geng",
                    "Dylan Sam",
                    "Yiding Jiang",
                    "Ashwini Pokle",
                    "Samuel Sokota",
                    "Yutong He",
                    "Naoki Murata",
                    "J. Williams",
                    "Yuki Mitsufuji",
                    "Ruslan Salakhutdinov",
                    "Rattana Pukdee",
                    "William Luo",
                    "Justin Lin",
                    "Weijian Luo",
                    "Zemin Huang",
                    "Guo-jun Qi",
                    "C. S. D. Witt",
                    "Spencer Compton",
                    "Jakob Foerster",
                    "Alexander Robey",
                    "George J. Pappas",
                    "Hamed Hassani",
                    "AI Sony",
                    "Sony Group Corporation",
                    "Bosch Center",
                    "Mingjie Sun",
                    "Xinlei Chen",
                    "Zhuang Liu",
                    "Daniel P. Jeong",
                    "Yewon Byun",
                    "Aviv Bick",
                    "Kevin Y. Li",
                    "Eric P. Xing",
                    "Albert Gu",
                    "Victor Akinwande",
                    "Maria-Florina Balcan",
                    "Pradeep Ravikumar",
                    "Mukul Bhutani",
                    "Paul Micaelli",
                    "Arash Vahdat",
                    "Hongxu Yin",
                    "Jan Kautz",
                    "Pavlo Molchanov",
                    "Alex Nichol",
                    "Prafulla Dhariwal",
                    "Aditya Ramesh",
                    "Pranav Shyam",
                    "Pamela Mishkin",
                    "Bob McGrew",
                    "I. Sutskever",
                    "Emilio Parisotto",
                    "Francis Song",
                    "Jack W. Rae",
                    "Razvan Pascanu",
                    "Caglar Gulcehre",
                    "Siddhant M. Jayakumar",
                    "Max Jaderberg",
                    "Raphael Lopez Kaufman",
                    "Aidan Clark",
                    "Adam Paszke",
                    "Sam Gross",
                    "Francisco Massa",
                    "Adam Lerer",
                    "James Bradbury",
                    "Gregory Chanan",
                    "Trevor Killeen",
                    "Zem-ing Lin",
                    "N. Gimelshein",
                    "L. Antiga",
                    "Alban Des-maison",
                    "Andreas Kopf",
                    "Edward Yang",
                    "Zachary DeVito",
                    "Martin Raison",
                    "Alykhan Tejani",
                    "Sasank Chilamkurthy",
                    "Alec Radford",
                    "Jeffrey Wu",
                    "R. Child",
                    "D. Luan",
                    "Colin Raffel",
                    "Noam M. Shazeer",
                    "A. Roberts",
                    "K. Lee",
                    "Sharan Narang",
                    "Michael Matena",
                    "Yanqi Zhou",
                    "Wei Li",
                    "Peter J. Liu",
                    "Casey Chu",
                    "Ryan D'Orazio",
                    "Chun Kai Ling",
                    "David J. Wu",
                    "Noam Brown",
                    "Chieh-Hsin Lai",
                    "Yuhta Takida",
                    "Toshimitsu Uesaka",
                    "Dongjun Kim",
                    "Wei-Hsiang Liao",
                    "Stefano Ermon"
                ],
                "domain": [
                    "Generative Models",
                    "Reinforcement Learning",
                    "Deep Learning",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "Consistency Models Made Easy",
                        "abstract": "Consistency models (CMs) offer faster sampling than traditional diffusion models, but their training is resource-intensive. For example, as of 2024, training a state-of-the-art CM on CIFAR-10 takes one week on 8 GPUs. In this work, we propose an effective scheme for training CMs that largely improves the efficiency of building such models. Specifically, by expressing CM trajectories via a particular differential equation, we argue that diffusion models can be viewed as a special case of CMs. We can thus fine-tune a consistency model starting from a pretrained diffusion model and progressively approximate the full consistency condition to stronger degrees over the training process. Our resulting method, which we term Easy Consistency Tuning (ECT), achieves vastly reduced training times while improving upon the quality of previous methods: for example, ECT achieves a 2-step FID of 2.73 on CIFAR10 within 1 hour on a single A100 GPU, matching Consistency Distillation trained for hundreds of GPU hours. Owing to this computational efficiency, we investigate the scaling laws of CMs under ECT, showing that they obey the classic power law scaling, hinting at their ability to improve efficiency and performance at larger scales. Our code (https://github.com/locuslab/ect) is publicly available, making CMs more accessible to the broader community."
                    },
                    {
                        "title": "One-Step Diffusion Distillation through Score Implicit Matching",
                        "abstract": "Despite their strong performances on many generative tasks, diffusion models require a large number of sampling steps in order to generate realistic samples. This has motivated the community to develop effective methods to distill pre-trained diffusion models into more efficient models, but these methods still typically require few-step inference or perform substantially worse than the underlying model. In this paper, we present Score Implicit Matching (SIM) a new approach to distilling pre-trained diffusion models into single-step generator models, while maintaining almost the same sample generation ability as the original model as well as being data-free with no need of training samples for distillation. The method rests upon the fact that, although the traditional score-based loss is intractable to minimize for generator models, under certain conditions we can efficiently compute the gradients for a wide class of score-based divergences between a diffusion model and a generator. SIM shows strong empirical performances for one-step generators: on the CIFAR10 dataset, it achieves an FID of 2.06 for unconditional generation and 1.96 for class-conditional generation. Moreover, by applying SIM to a leading transformer-based diffusion model, we distill a single-step generator for text-to-image (T2I) generation that attains an aesthetic score of 6.42 with no performance decline over the original multi-step counterpart, clearly outperforming the other one-step generators including SDXL-TURBO of 5.33, SDXL-LIGHTNING of 5.34 and HYPER-SDXL of 5.85. We will release this industry-ready one-step transformer-based T2I generator along with this paper."
                    },
                    {
                        "title": "Computing Low-Entropy Couplings for Large-Support Distributions",
                        "abstract": "Minimum-entropy coupling (MEC) -- the process of finding a joint distribution with minimum entropy for given marginals -- has applications in areas such as causality and steganography. However, existing algorithms are either computationally intractable for large-support distributions or limited to specific distribution types and sensitive to hyperparameter choices. This work addresses these limitations by unifying a prior family of iterative MEC (IMEC) approaches into a generalized partition-based formalism. From this framework, we derive a novel IMEC algorithm called ARIMEC, capable of handling arbitrary discrete distributions, and introduce a method to make IMEC robust to suboptimal hyperparameter settings. These innovations facilitate the application of IMEC to high-throughput steganography with language models, among other settings. Our codebase is available at https://github.com/ssokota/mec ."
                    },
                    {
                        "title": "Automated Black-box Prompt Engineering for Personalized Text-to-Image Generation",
                        "abstract": "Prompt engineering is effective for controlling the output of text-to-image (T2I) generative models, but it is also laborious due to the need for manually crafted prompts. This challenge has spurred the development of algorithms for automated prompt generation. However, these methods often struggle with transferability across T2I models, require white-box access to the underlying model, and produce non-intuitive prompts. In this work, we introduce PRISM, an algorithm that automatically identifies human-interpretable and transferable prompts that can effectively generate desired concepts given only black-box access to T2I models. Inspired by large language model (LLM) jailbreaking, PRISM leverages the in-context learning ability of LLMs to iteratively refine the candidate prompts distribution for given reference images. Our experiments demonstrate the versatility and effectiveness of PRISM in generating accurate prompts for objects, styles and images across multiple T2I models, including Stable Diffusion, DALL-E, and Midjourney."
                    },
                    {
                        "title": "FUSE-ing Language Models: Zero-Shot Adapter Discovery for Prompt Optimization Across Tokenizers",
                        "abstract": "The widespread use of large language models has resulted in a multitude of tokenizers and embedding spaces, making knowledge transfer in prompt discovery tasks difficult. In this work, we propose FUSE (Flexible Unification of Semantic Embeddings), an inexpensive approach to approximating an adapter layer that maps from one model's textual embedding space to another, even across different tokenizers. We introduce a third-order tensor-based representation of a model's embedding space that aligns semantic embeddings that have been split apart by different tokenizers, and use this representation to derive an approximation of the gradient of one model's outputs with respect to another model's embedding space. We show the efficacy of our approach via multi-objective optimization over vision-language and causal language models for image captioning and sentiment-based image captioning."
                    },
                    {
                        "title": "Massive Activations in Large Language Models",
                        "abstract": "We observe an empirical phenomenon in Large Language Models (LLMs) -- very few activations exhibit significantly larger values than others (e.g., 100,000 times larger). We call them massive activations. First, we demonstrate the widespread existence of massive activations across various LLMs and characterize their locations. Second, we find their values largely stay constant regardless of the input, and they function as indispensable bias terms in LLMs. Third, these massive activations lead to the concentration of attention probabilities to their corresponding tokens, and further, implicit bias terms in the self-attention output. Last, we also study massive activations in Vision Transformers. Code is available at https://github.com/locuslab/massive-activations."
                    },
                    {
                        "title": "Bayesian Neural Networks with Domain Knowledge Priors",
                        "abstract": "Bayesian neural networks (BNNs) have recently gained popularity due to their ability to quantify model uncertainty. However, specifying a prior for BNNs that captures relevant domain knowledge is often extremely challenging. In this work, we propose a framework for integrating general forms of domain knowledge (i.e., any knowledge that can be represented by a loss function) into a BNN prior through variational inference, while enabling computationally efficient posterior inference and sampling. Specifically, our approach results in a prior over neural network weights that assigns high probability mass to models that better align with our domain knowledge, leading to posterior samples that also exhibit this behavior. We show that BNNs using our proposed domain knowledge priors outperform those with standard priors (e.g., isotropic Gaussian, Gaussian process), successfully incorporating diverse types of prior information such as fairness, physics rules, and healthcare knowledge and achieving better predictive performance. We also present techniques for transferring the learned priors across different model architectures, demonstrating their broad utility across various settings."
                    },
                    {
                        "title": "Transformers to SSMs: Distilling Quadratic Knowledge to Subquadratic Models",
                        "abstract": "Transformer architectures have become a dominant paradigm for domains like language modeling but suffer in many inference settings due to their quadratic-time self-attention. Recently proposed subquadratic architectures, such as Mamba, have shown promise, but have been pretrained with substantially less computational resources than the strongest Transformer models. In this work, we present a method that is able to distill a pretrained Transformer architecture into alternative architectures such as state space models (SSMs). The key idea to our approach is that we can view both Transformers and SSMs as applying different forms of mixing matrices over the token sequences. We can thus progressively distill the Transformer architecture by matching different degrees of granularity in the SSM: first matching the mixing matrices themselves, then the hidden units at each block, and finally the end-to-end predictions. Our method, called MOHAWK, is able to distill a Mamba-2 variant based on the Phi-1.5 architecture (Phi-Mamba) using only 3B tokens and a hybrid version (Hybrid Phi-Mamba) using 5B tokens. Despite using less than 1% of the training data typically used to train models from scratch, Phi-Mamba boasts substantially stronger performance compared to all past open-source non-Transformer models. MOHAWK allows models like SSMs to leverage computational resources invested in training Transformer-based architectures, highlighting a new avenue for building such models."
                    },
                    {
                        "title": "AcceleratedLiNGAM: Learning Causal DAGs at the speed of GPUs",
                        "abstract": "Existing causal discovery methods based on combinatorial optimization or search are slow, prohibiting their application on large-scale datasets. In response, more recent methods attempt to address this limitation by formulating causal discovery as structure learning with continuous optimization but such approaches thus far provide no statistical guarantees. In this paper, we show that by efficiently parallelizing existing causal discovery methods, we can in fact scale them to thousands of dimensions, making them practical for substantially larger-scale problems. In particular, we parallelize the LiNGAM method, which is quadratic in the number of variables, obtaining up to a 32-fold speed-up on benchmark datasets when compared with existing sequential implementations. Specifically, we focus on the causal ordering subprocedure in DirectLiNGAM and implement GPU kernels to accelerate it. This allows us to apply DirectLiNGAM to causal inference on large-scale gene expression data with genetic interventions yielding competitive results compared with specialized continuous optimization methods, and Var-LiNGAM for causal discovery on U.S. stock data."
                    },
                    {
                        "title": "Learning with Explanation Constraints",
                        "abstract": "As larger deep learning models are hard to interpret, there has been a recent focus on generating explanations of these black-box models. In contrast, we may have apriori explanations of how models should behave. In this paper, we formalize this notion as learning from explanation constraints and provide a learning theoretic framework to analyze how such explanations can improve the learning of our models. One may naturally ask,\"When would these explanations be helpful?\"Our first key contribution addresses this question via a class of models that satisfies these explanation constraints in expectation over new data. We provide a characterization of the benefits of these models (in terms of the reduction of their Rademacher complexities) for a canonical class of explanations given by gradient information in the settings of both linear models and two layer neural networks. In addition, we provide an algorithmic solution for our framework, via a variational approximation that achieves better performance and satisfies these constraints more frequently, when compared to simpler augmented Lagrangian methods to incorporate these explanations. We demonstrate the benefits of our approach over a large array of synthetic and real-world experiments."
                    },
                    {
                        "title": "Sinkhorn-Flow: Predicting Probability Mass Flow in Dynamical Systems Using Optimal Transport",
                        "abstract": "Predicting how distributions over discrete variables vary over time is a common task in time series forecasting. But whereas most approaches focus on merely predicting the distribution at subsequent time steps, a crucial piece of information in many settings is to determine how this probability mass flows between the different elements over time. We propose a new approach to predicting such mass flow over time using optimal transport. Specifically, we propose a generic approach to predicting transport matrices in end-to-end deep learning systems, replacing the standard softmax operation with Sinkhorn iterations. We apply our approach to the task of predicting how communities will evolve over time in social network settings, and show that the approach improves substantially over alternative prediction methods. We specifically highlight results on the task of predicting faction evolution in Ukrainian parliamentary voting."
                    },
                    {
                        "title": "One-Step Diffusion Distillation via Deep Equilibrium Models",
                        "abstract": "Diffusion models excel at producing high-quality samples but naively require hundreds of iterations, prompting multiple attempts to distill the generation process into a faster network. However, many existing approaches suffer from a variety of challenges: the process for distillation training can be complex, often requiring multiple training stages, and the resulting models perform poorly when utilized in single-step generative applications. In this paper, we introduce a simple yet effective means of distilling diffusion models directly from initial noise to the resulting image. Of particular importance to our approach is to leverage a new Deep Equilibrium (DEQ) model as the distilled architecture: the Generative Equilibrium Transformer (GET). Our method enables fully offline training with just noise/image pairs from the diffusion model while achieving superior performance compared to existing one-step methods on comparable training budgets. We demonstrate that the DEQ architecture is crucial to this capability, as GET matches a $5\\times$ larger ViT in terms of FID scores while striking a critical balance of computational cost and image quality. Code, checkpoints, and datasets are available."
                    },
                    {
                        "title": "Abstracting Imperfect Information Away from Two-Player Zero-Sum Games",
                        "abstract": "In their seminal work, Nayyar et al. (2013) showed that imperfect information can be abstracted away from common-payoff games by having players publicly announce their policies as they play. This insight underpins sound solvers and decision-time planning algorithms for common-payoff games. Unfortunately, a naive application of the same insight to two-player zero-sum games fails because Nash equilibria of the game with public policy announcements may not correspond to Nash equilibria of the original game. As a consequence, existing sound decision-time planning algorithms require complicated additional mechanisms that have unappealing properties. The main contribution of this work is showing that certain regularized equilibria do not possess the aforementioned non-correspondence problem -- thus, computing them can be treated as perfect-information problems. Because these regularized equilibria can be made arbitrarily close to Nash equilibria, our result opens the door to a new perspective to solving two-player zero-sum games and yields a simplified framework for decision-time planning in two-player zero-sum games, void of the unappealing properties that plague existing decision-time planning approaches."
                    },
                    {
                        "title": "Manifold Preserving Guided Diffusion",
                        "abstract": "Despite the recent advancements, conditional image generation still faces challenges of cost, generalizability, and the need for task-specific training. In this paper, we propose Manifold Preserving Guided Diffusion (MPGD), a training-free conditional generation framework that leverages pretrained diffusion models and off-the-shelf neural networks with minimal additional inference cost for a broad range of tasks. Specifically, we leverage the manifold hypothesis to refine the guided diffusion steps and introduce a shortcut algorithm in the process. We then propose two methods for on-manifold training-free guidance using pre-trained autoencoders and demonstrate that our shortcut inherently preserves the manifolds when applied to latent diffusion models. Our experiments show that MPGD is efficient and effective for solving a variety of conditional generation applications in low-compute settings, and can consistently offer up to 3.8x speed-ups with the same number of diffusion steps while maintaining high sample quality compared to the baselines."
                    },
                    {
                        "title": "On the Importance of Exploration for Generalization in Reinforcement Learning",
                        "abstract": "Existing approaches for improving generalization in deep reinforcement learning (RL) have mostly focused on representation learning, neglecting RL-specific aspects such as exploration. We hypothesize that the agent's exploration strategy plays a key role in its ability to generalize to new environments. Through a series of experiments in a tabular contextual MDP, we show that exploration is helpful not only for efficiently finding the optimal policy for the training environments but also for acquiring knowledge that helps decision making in unseen environments. Based on these observations, we propose EDE: Exploration via Distributional Ensemble, a method that encourages exploration of states with high epistemic uncertainty through an ensemble of Q-value distributions. Our algorithm is the first value-based approach to achieve state-of-the-art on both Procgen and Crafter, two benchmarks for generalization in RL with high-dimensional observations. The open-sourced implementation can be found at https://github.com/facebookresearch/ede ."
                    },
                    {
                        "title": "Mimetic Initialization of Self-Attention Layers",
                        "abstract": "It is notoriously difficult to train Transformers on small datasets; typically, large pre-trained models are instead used as the starting point. We explore the weights of such pre-trained Transformers (particularly for vision) to attempt to find reasons for this discrepancy. Surprisingly, we find that simply initializing the weights of self-attention layers so that they\"look\"more like their pre-trained counterparts allows us to train vanilla Transformers faster and to higher final accuracies, particularly on vision tasks such as CIFAR-10 and ImageNet classification, where we see gains in accuracy of over 5% and 4%, respectively. Our initialization scheme is closed form, learning-free, and very simple: we set the product of the query and key weights to be approximately the identity, and the product of the value and projection weights to approximately the negative identity. As this mimics the patterns we saw in pre-trained Transformers, we call the technique\"mimetic initialization\"."
                    },
                    {
                        "title": "Universal and Transferable Adversarial Attacks on Aligned Language Models",
                        "abstract": "Because\"out-of-the-box\"large language models are capable of generating a great deal of objectionable content, recent work has focused on aligning these models in an attempt to prevent undesirable generation. While there has been some success at circumventing these measures -- so-called\"jailbreaks\"against LLMs -- these attacks have required significant human ingenuity and are brittle in practice. In this paper, we propose a simple and effective attack method that causes aligned language models to generate objectionable behaviors. Specifically, our approach finds a suffix that, when attached to a wide range of queries for an LLM to produce objectionable content, aims to maximize the probability that the model produces an affirmative response (rather than refusing to answer). However, instead of relying on manual engineering, our approach automatically produces these adversarial suffixes by a combination of greedy and gradient-based search techniques, and also improves over past automatic prompt generation methods. Surprisingly, we find that the adversarial prompts generated by our approach are quite transferable, including to black-box, publicly released LLMs. Specifically, we train an adversarial attack suffix on multiple prompts (i.e., queries asking for many different types of objectionable content), as well as multiple models (in our case, Vicuna-7B and 13B). When doing so, the resulting attack suffix is able to induce objectionable content in the public interfaces to ChatGPT, Bard, and Claude, as well as open source LLMs such as LLaMA-2-Chat, Pythia, Falcon, and others. In total, this work significantly advances the state-of-the-art in adversarial attacks against aligned language models, raising important questions about how such systems can be prevented from producing objectionable information. Code is available at github.com/llm-attacks/llm-attacks."
                    },
                    {
                        "title": "Practical Critic Gradient based Actor Critic for On-Policy Reinforcement Learning",
                        "abstract": "On-policy reinforcement learning algorithms have been shown to be remarkably efficient at learning policies for continuous control robotics tasks. They are highly parallelizable and hence have benefited tremendously from the recent rise in GPU based parallel simulators. The most widely used on-policy reinforcement learning algorithm is proximal policy optimization (PPO) which was introduced in 2017 and was designed for a somewhat different setting with CPU based serial or less parallelizable simulators. However, suprisingly, it has maintained dominance even on tasks based on the highly parallelizable simulators of today. In this paper, we show that a different class of on-policy algorithms based on estimating the policy gradient using the critic-action gradients are better suited when using highly parallelizable simulators. The primary issues for these algorithms arise from the lack of diversity of the on-policy experiences used for the updates and the instabilities arising from the interaction between the biased critic gradients and the rapidly changing policy distribution. We address the former by simply increasing the number of parallel simulation runs (thanks to the GPU based simulators) along with an appropriate schedule on the policy entropy to ensure diversity of samples. We address the latter by adding a policy averaging step and value averaging step (as in off-policy methods). With these modifications, we observe that the critic gradient based on-policy method (CGAC) consistently achieves higher episode returns compared with existing baselines. Furthermore, in environments with high dimensional action space, CGAC also trains much faster (in wall-clock time) than the corresponding baselines."
                    },
                    {
                        "title": "Projected Off-Policy Q-Learning (POP-QL) for Stabilizing Offline Reinforcement Learning",
                        "abstract": "A key problem in off-policy Reinforcement Learning (RL) is the mismatch, or distribution shift, between the dataset and the distribution over states and actions visited by the learned policy. This problem is exacerbated in the fully offline setting. The main approach to correct this shift has been through importance sampling, which leads to high-variance gradients. Other approaches, such as conservatism or behavior-regularization, regularize the policy at the cost of performance. In this paper, we propose a new approach for stable off-policy Q-Learning. Our method, Projected Off-Policy Q-Learning (POP-QL), is a novel actor-critic algorithm that simultaneously reweights off-policy samples and constrains the policy to prevent divergence and reduce value-approximation error. In our experiments, POP-QL not only shows competitive performance on standard benchmarks, but also out-performs competing methods in tasks where the data-collection policy is significantly sub-optimal."
                    },
                    {
                        "title": "On the Joint Interaction of Models, Data, and Features",
                        "abstract": "Learning features from data is one of the defining characteristics of deep learning, but our theoretical understanding of the role features play in deep learning is still rudimentary. To address this gap, we introduce a new tool, the interaction tensor, for empirically analyzing the interaction between data and model through features. With the interaction tensor, we make several key observations about how features are distributed in data and how models with different random seeds learn different features. Based on these observations, we propose a conceptual framework for feature learning. Under this framework, the expected accuracy for a single hypothesis and agreement for a pair of hypotheses can both be derived in closed-form. We demonstrate that the proposed framework can explain empirically observed phenomena, including the recently discovered Generalization Disagreement Equality (GDE) that allows for estimating the generalization error with only unlabeled data. Further, our theory also provides explicit construction of natural data distributions that break the GDE. Thus, we believe this work provides valuable new insight into our understanding of feature learning."
                    }
                ]
            }
        }
    },
    "2405.17705": {
        "paper_data": {
            "title": "DC-Gaussian: Improving 3D Gaussian Splatting for Reflective Dash Cam Videos",
            "url": "http://arxiv.org/abs/2405.17705v2",
            "arxiv_id": "2405.17705",
            "authors": [
                "Linhan Wang",
                "Kai Cheng",
                "Shuo Lei",
                "Shengkun Wang",
                "Wei Yin",
                "Chenyang Lei",
                "Xiaoxiao Long",
                "Chang-Tien Lu"
            ],
            "abstract": "We present DC-Gaussian, a new method for generating novel views from in-vehicle dash cam videos. While neural rendering techniques have made significant strides in driving scenarios, existing methods are primarily designed for videos collected by autonomous vehicles. However, these videos are limited in both quantity and diversity compared to dash cam videos, which are more widely used across various types of vehicles and capture a broader range of scenarios. Dash cam videos often suffer from severe obstructions such as reflections and occlusions on the windshields, which significantly impede the application of neural rendering techniques. To address this challenge, we develop DC-Gaussian based on the recent real-time neural rendering technique 3D Gaussian Splatting (3DGS). Our approach includes an adaptive image decomposition module to model reflections and occlusions in a unified manner. Additionally, we introduce illumination-aware obstruction modeling to manage reflections and occlusions under varying lighting conditions. Lastly, we employ a geometry-guided Gaussian enhancement strategy to improve rendering details by incorporating additional geometry priors. Experiments on self-captured and public dash cam videos show that our method not only achieves state-of-the-art performance in novel view synthesis, but also accurately reconstructing captured scenes getting rid of obstructions.",
            "introduction": "   1 Introduction  Neural Radiance Field (NeRF) [29] has revolutionized the image-based rendering area with its differentiable volume rendering technique. 3D Gaussian Splatting (3DGS) [19] pushes the frontier further with real-time rendering speed. These technologies have been applied to datasets captured by autonomous cars [41, 6, 25], opening up numerous new possibilities in autonomous driving, such as simulating driving scenarios [55, 50] for robust training of perception models and providing effective 3D scene representations to enhance comprehensive environmental understanding [14, 61, 64]. Although these datasets provide multi-modality sensor data, their diversity in real-world driving scenarios is still limited [7].   Figure 2: Common obstructions on windshields: (a) Mobile-phone holder; (b) Reflections; (c) Stains.   Fortunately, dash cam videos deeply reflect the diversity and complexity of real-world traffic scenarios [8]. They are used to provide large-scale, diverse driving video datasets in a crowd-sourced manner [57]. Dash cam videos also offer unique value by capturing multi-agent driving behaviors [7] and evaluating the robustness of algorithms under visual hazards [58]. Moreover, the global dash cam market is rapidly expanding, driven by increasing awareness of vehicular safety [33]. Therefore, the exploration of utilizing dash camera data in neural rendering shows great potential, offering enormous amounts of data for autonomous driving applications.   However, naively training 3DGS on dash cam videos often results in a significant deterioration of rendering quality and geometry. This degradation is primarily due to the common existence of obstructions (reflections and occlusions such as mobile phone holders and stains, as shown in Fig. 2) on windshields. In these scenarios, 3DGS models the obstructions as stationary geometries while they are dynamic in nature (moving with cars), thus unavoidably causing inaccurate geometry and blurry renderings in novel views.   Although some single-image-based obstruction removal methods exist, directly applying them to this task is nontrivial. These obstructions arise from various sources, while existing removal methods impose strict assumptions on obstructions [12, 62, 54, 53, 37, 20]. For instance, assumptions like out-of-focus [62] and ghost cue [37] allow previous methods to perform well in specific cases, but these assumptions don\u2019t always hold in dash cam videos. Moreover, the performance of learning-based methods degrades for out-of-distribution images [46, 27, 21]. Several NeRF methods [16, 32, 66, 67] attempt to reconstruct scenes with reflections by decomposing transmission (background scene) and reflection with independent NeRFs. This approach can benefit from the strong multi-view aggregation power of NeRF. However, previous methods are insufficient for dash cam videos because the obstructions on windshields do not align well with NeRF\u2019s design. The vanilla NeRF is intended for static scenes, whereas windshields and their reflected objects move with the cars.   In this paper, we introduce DC-Gaussian, a method for modeling high-fidelity obstruction-free 3D Gaussian Splatting from dash cam videos. We introduce three key innovations: 1) Adaptive image decomposition. To clearly decompose images with complex reflections and occlusions, we propose an adaptive image decomposition approach. We use an opacity map to learn the transmittance of the windshield, which adaptively estimates the contribution of the background scene to the image. 2) Illumination-aware Obstruction Modeling(IOM). We observe that the obstructions are mainly caused by the objects being relatively static with",
            "references": [
                {
                    "title": "HUGS: Holistic Urban 3D Scene Understanding via Gaussian Splatting",
                    "abstract": "Holistic understanding of urban scenes based on RGB images is a challenging yet important problem. It encompasses understanding both the geometry and appearance to enable novel view synthesis, parsing semantic labels, and tracking moving objects. Despite considerable progress, existing approaches often focus on specific aspects of this task and require additional inputs such as LiDAR scans or manually annotated 3D bounding boxes. In this paper, we introduce a novel pipeline that utilizes 3D Gaussian Splatting for holistic urban scene understanding. Our main idea involves the joint optimization of geometry, appearance, semantics, and motion using a combination of static and dynamic 3D Gaussians, where moving object poses are regularized via physical constraints. Our approach offers the ability to render new viewpoints in real-time, yielding 2D and 3D semantic information with high accuracy, and reconstruct dynamic scenes, even in scenarios where 3D bounding box detection are highly noisy. Experimental results on KITTI, KITTI-360, and Virtual KITTI 2 demonstrate the effectiveness of our approach. Our project page is at https://xdimlab.github.io/hugs_website."
                },
                {
                    "title": "GaussianPro: 3D Gaussian Splatting with Progressive Propagation",
                    "abstract": "The advent of 3D Gaussian Splatting (3DGS) has recently brought about a revolution in the field of neural rendering, facilitating high-quality renderings at real-time speed. However, 3DGS heavily depends on the initialized point cloud produced by Structure-from-Motion (SfM) techniques. When tackling with large-scale scenes that unavoidably contain texture-less surfaces, the SfM techniques always fail to produce enough points in these surfaces and cannot provide good initialization for 3DGS. As a result, 3DGS suffers from difficult optimization and low-quality renderings. In this paper, inspired by classical multi-view stereo (MVS) techniques, we propose GaussianPro, a novel method that applies a progressive propagation strategy to guide the densification of the 3D Gaussians. Compared to the simple split and clone strategies used in 3DGS, our method leverages the priors of the existing reconstructed geometries of the scene and patch matching techniques to produce new Gaussians with accurate positions and orientations. Experiments on both large-scale and small-scale scenes validate the effectiveness of our method, where our method significantly surpasses 3DGS on the Waymo dataset, exhibiting an improvement of 1.15dB in terms of PSNR."
                },
                {
                    "title": "Neural Spline Fields for Burst Image Fusion and Layer Separation",
                    "abstract": "Each photo in an image burst can be considered a sam-ple of a complex 3D scene: the product of parallax, diffuse and specular materials, scene motion, and illuminant vari-ation. While decomposing all of these effects from a stack of misaligned images is a highly ill-conditioned task, the conventional align-and-merge burst pipeline takes the other extreme: blending them into a single image. In this work, we propose a versatile intermediate representation: a two-layer alpha-composited image plus flow model constructed with neural spline fields - networks trained to map input coordinates to spline control points. Our method is able to, during test-time optimization, jointly fuse a burst image capture into one high-resolution reconstruction and decom-pose it into transmission and obstruction layers. Then, by discarding the obstruction layer, we can perform a range of tasks including seeing through occlusions, reflection sup-pression, and shadow removal. Tested on complex in-the-wild captures we find that, with no post-processing steps or learned priors, our generalizable model is able to out-perform existing dedicated single-image and multi-view ob-struction removal approaches."
                },
                {
                    "title": "DrivingGaussian: Composite Gaussian Splatting for Surrounding Dynamic Autonomous Driving Scenes",
                    "abstract": "We present DrivingGaussian, an efficient and effective framework for surrounding dynamic autonomous driving scenes. For complex scenes with moving objects, we first sequentially and progressively model the static background of the entire scene with incremental static 3D Gaussians. We then leverage a composite dynamic Gaussian graph to handle multiple moving objects, individually reconstructing each object and restoring their accurate positions and occlusion relationships within the scene. We further use a LiDAR prior for Gaussian Splatting to reconstruct scenes with greater details and maintain panoramic consistency. DrivingGaussian outperforms existing methods in dynamic driving scene reconstruction and enables photorealistic surround-view synthesis with high-fidelity and multi-camera consistency. Our project page is at: https:/github.com/VDIGPKU/DrivingGaussian."
                },
                {
                    "title": "UC-NeRF: Neural Radiance Field for Under-Calibrated multi-view cameras in autonomous driving",
                    "abstract": "Multi-camera setups find widespread use across various applications, such as autonomous driving, as they greatly expand sensing capabilities. Despite the fast development of Neural radiance field (NeRF) techniques and their wide applications in both indoor and outdoor scenes, applying NeRF to multi-camera systems remains very challenging. This is primarily due to the inherent under-calibration issues in multi-camera setup, including inconsistent imaging effects stemming from separately calibrated image signal processing units in diverse cameras, and system errors arising from mechanical vibrations during driving that affect relative camera poses. In this paper, we present UC-NeRF, a novel method tailored for novel view synthesis in under-calibrated multi-view camera systems. Firstly, we propose a layer-based color correction to rectify the color inconsistency in different image regions. Second, we propose virtual warping to generate more viewpoint-diverse but color-consistent virtual views for color correction and 3D recovery. Finally, a spatiotemporally constrained pose refinement is designed for more robust and accurate pose calibration in multi-camera systems. Our method not only achieves state-of-the-art performance of novel view synthesis in multi-camera setups, but also effectively facilitates depth estimation in large-scale outdoor scenes with the synthesized novel views."
                },
                {
                    "title": "EmerNeRF: Emergent Spatial-Temporal Scene Decomposition via Self-Supervision",
                    "abstract": "We present EmerNeRF, a simple yet powerful approach for learning spatial-temporal representations of dynamic driving scenes. Grounded in neural fields, EmerNeRF simultaneously captures scene geometry, appearance, motion, and semantics via self-bootstrapping. EmerNeRF hinges upon two core components: First, it stratifies scenes into static and dynamic fields. This decomposition emerges purely from self-supervision, enabling our model to learn from general, in-the-wild data sources. Second, EmerNeRF parameterizes an induced flow field from the dynamic field and uses this flow field to further aggregate multi-frame features, amplifying the rendering precision of dynamic objects. Coupling these three fields (static, dynamic, and flow) enables EmerNeRF to represent highly-dynamic scenes self-sufficiently, without relying on ground truth object annotations or pre-trained models for dynamic object segmentation or optical flow estimation. Our method achieves state-of-the-art performance in sensor simulation, significantly outperforming previous methods when reconstructing static (+2.93 PSNR) and dynamic (+3.70 PSNR) scenes. In addition, to bolster EmerNeRF's semantic generalization, we lift 2D visual foundation model features into 4D space-time and address a general positional bias in modern Transformers, significantly boosting 3D perception performance (e.g., 37.50% relative improvement in occupancy prediction accuracy on average). Finally, we construct a diverse and challenging 120-sequence dataset to benchmark neural fields under extreme and highly-dynamic settings."
                },
                {
                    "title": "Deformable 3D Gaussians for High-Fidelity Monocular Dynamic Scene Reconstruction",
                    "abstract": "Implicit neural representation has paved the way for new approaches to dynamic scene reconstruction. Nonetheless, cutting-edge dynamic neural rendering methods rely heavily on these implicit representations, which frequently struggle to capture the intricate details of objects in the scene. Furthermore, implicit methods have difficulty achieving real-time rendering in general dynamic scenes, limiting their use in a variety of tasks. To address the issues, we propose a deformable 3D Gaussians splatting method that reconstructs scenes using 3D Gaussians and learns them in canonical space with a deformation field to model monocular dynamic scenes. We also introduce an annealing smoothing training mechanism with no extra overhead, which can mitigate the impact of inaccurate poses on the smoothness of time interpolation tasks in real-world scenes. Through a differential Gaussian rasterizer, the deformable 3D Gaussians not only achieve higher rendering quality but also real-time rendering speed. Experiments show that our method outperforms existing methods significantly in terms of both rendering quality and speed, making it well-suited for tasks such as novel-view synthesis, time interpolation, and real-time rendering. Our code is available at https://github.com/ingra14m/Deformable-3D-Gaussians."
                },
                {
                    "title": "Single Image Reflection Separation via Component Synergy",
                    "abstract": "The reflection superposition phenomenon is complex and widely distributed in the real world, which derives various simplified linear and nonlinear formulations of the problem. In this paper, based on the investigation of the weaknesses of existing models, we propose a more general form of the superposition model by introducing a learnable residue term, which can effectively capture residual information during decomposition, guiding the separated layers to be complete. In order to fully capitalize on its advantages, we further design the network structure elaborately, including a novel dual-stream interaction mechanism and a powerful decomposition network with a semantic pyramid encoder. Extensive experiments and ablation studies are conducted to verify our superiority over state-of-the-art approaches on multiple real-world benchmark datasets. Our code is publicly available at https://github.com/mingcv/DSRNet."
                },
                {
                    "title": "3D Gaussian Splatting for Real-Time Radiance Field Rendering",
                    "abstract": "Radiance Field methods have recently revolutionized novel-view synthesis of scenes captured with multiple photos or videos. However, achieving high visual quality still requires neural networks that are costly to train and render, while recent faster methods inevitably trade off speed for quality. For unbounded and complete scenes (rather than isolated objects) and 1080p resolution rendering, no current method can achieve real-time display rates. We introduce three key elements that allow us to achieve state-of-the-art visual quality while maintaining competitive training times and importantly allow high-quality real-time (\u2265 30 fps) novel-view synthesis at 1080p resolution. First, starting from sparse points produced during camera calibration, we represent the scene with 3D Gaussians that preserve desirable properties of continuous volumetric radiance fields for scene optimization while avoiding unnecessary computation in empty space; Second, we perform interleaved optimization/density control of the 3D Gaussians, notably optimizing anisotropic covariance to achieve an accurate representation of the scene; Third, we develop a fast visibility-aware rendering algorithm that supports anisotropic splatting and both accelerates training and allows realtime rendering. We demonstrate state-of-the-art visual quality and real-time rendering on several established datasets."
                },
                {
                    "title": "Occlusion-Free Scene Recovery via Neural Radiance Fields",
                    "abstract": "Our everyday lives are filled with occlusions that we strive to see through. By aggregating desired background information from different viewpoints, we can easily eliminate such occlusions without any external occlusion-free supervision. Though several occlusion removal methods have been proposed to empower machine vision systems with such ability, their performances are still unsatisfactory due to reliance on external supervision. We propose a novel method for occlusion removal by directly building a mapping between position and viewing angles and the corresponding occlusion-free scene details leveraging Neural Radiance Fields (NeRF). We also develop an effective scheme to jointly optimize camera parameters and scene reconstruction when occlusions are present. An additional depth constraint is applied to supervise the entire optimizaion without labeled external data for training. The experimental results on existing and newly collected datasets validate the effectiveness of our method. Our project page: https://freebutuselesssoul.github.io/occnerf."
                },
                {
                    "title": "UniSim: A Neural Closed-Loop Sensor Simulator",
                    "abstract": "Rigorously testing autonomy systems is essential for making safe self-driving vehicles (SDV) a reality. It requires one to generate safety critical scenarios beyond what can be collected safely in the world, as many scenarios happen rarely on our roads. To accurately evaluate performance, we need to test the SDV on these scenarios in closed-loop, where the SDV and other actors interact with each other at each timestep. Previously recorded driving logs provide a rich resource to build these new scenarios from, but for closed loop evaluation, we need to modify the sensor data based on the new scene configuration and the SDV's decisions, as actors might be added or removed and the trajectories of existing actors and the SDV will differ from the original log. In this paper, we present UniSim, a neural sensor simulator that takes a single recorded log captured by a sensor-equipped vehicle and converts it into a realistic closed-loop multi-sensor simulation. UniSim builds neural feature grids to reconstruct both the static background and dynamic actors in the scene, and composites them together to simulate LiDAR and camera data at new viewpoints, with actors added or removed and at new placements. To better handle extrapolated views, we incorporate learnable priors for dynamic objects, and leverage a convolutional network to complete unseen regions. Our experiments show UniSim can simulate realistic sensor data with small domain gap on downstream tasks. With UniSim, we demonstrate, for the first time, closed-loop evaluation of an autonomy system on safety-critical scenarios as if it were in the real world."
                },
                {
                    "title": "Looking Through the Glass: Neural Surface Reconstruction Against High Specular Reflections",
                    "abstract": "Neural implicit methods have achieved high-quality 3D object surfaces under slight specular highlights. However, high specular reflections (HSR) often appear in front of target objects when we capture them through glasses. The complex ambiguity in these scenes violates the multi-view consistency, then makes it challenging for recent methods to reconstruct target objects correctly. To remedy this issue, we present a novel surface reconstruction framework, NeuS-HSR, based on implicit neural rendering. In NeuSHSR, the object surface is parameterized as an implicit signed distance function (SDF). To reduce the interference of HSR, we propose decomposing the rendered image into two appearances: the target object and the auxiliary plane. We design a novel auxiliary plane module by combining physical assumptions and neural networks to generate the auxiliary plane appearance. Extensive experiments on synthetic and real-world datasets demonstrate that NeuS-HSR outperforms state-of-the-art approaches for accurate and robust target surface reconstruction against HSR. Code is available at https://github.com/JiaxiongQ/NeuS-HSR."
                },
                {
                    "title": "Zip-NeRF: Anti-Aliased Grid-Based Neural Radiance Fields",
                    "abstract": "Neural Radiance Field training can be accelerated through the use of grid-based representations in NeRF\u2019s learned mapping from spatial coordinates to colors and volumetric density. However, these grid-based approaches lack an explicit understanding of scale and therefore often introduce aliasing, usually in the form of jaggies or missing scene content. Anti-aliasing has previously been addressed by mip-NeRF 360, which reasons about sub-volumes along a cone rather than points along a ray, but this approach is not natively compatible with current grid-based techniques. We show how ideas from rendering and signal processing can be used to construct a technique that combines mip-NeRF 360 and grid-based models such as Instant NGP to yield error rates that are 8% \u2013 77% lower than either prior technique, and that trains 24\u00d7 faster than mip-NeRF 360."
                },
                {
                    "title": "SUDS: Scalable Urban Dynamic Scenes",
                    "abstract": "We extend neural radiance fields (NeRFs) to dynamic large-scale urban scenes. Prior work tends to reconstruct single video clips of short durations (up to 10 seconds). Two reasons are that such methods (a) tend to scale linearly with the number of moving objects and input videos because a separate model is built for each and (b) tend to require supervision via 3D bounding boxes and panoptic labels, obtained manually or via category-specific models. As a step towards truly open-world reconstructions of dynamic cities, we introduce two key innovations: (a) we factorize the scene into three separate hash table data structures to efficiently encode static, dynamic, and far-field radiance fields, and (b) we make use of unlabeled target signals consisting of RGB images, sparse LiDAR, off-the-shelf self-supervised 2D descriptors, and most importantly, 2D optical flow. Operationalizing such inputs via photometric, geometric, and feature-metric reconstruction losses enables SUDS to decompose dynamic scenes into the static background, individual objects, and their motions. When combined with our multi-branch table representation, such reconstructions can be scaled to tens of thousands of objects across 1.2 million frames from 1700 videos spanning geospatial footprints of hundreds of kilometers, (to our knowledge) the largest dynamic NeRF built to date. We present qualitative initial results on a variety of tasks enabled by our representations, including novel-view synthesis of dynamic urban scenes, unsupervised 3D instance segmentation, and unsupervised 3D cuboid detection. To compare to prior work, we also evaluate on KITTI and Virtual KITTI 2, surpassing state-of-the-art methods that rely on ground truth 3D bounding box annotations while being 10x quicker to train."
                },
                {
                    "title": "Nerflets: Local Radiance Fields for Efficient Structure-Aware 3D Scene Representation from 2D Supervision",
                    "abstract": "We address efficient and structure-aware 3D scene representation from images. Nerflets are our key contribution-a set of local neural radiance fields that together represent a scene. Each nerflet maintains its own spatial position, orientation, and extent, within which it contributes to panoptic, density, and radiance reconstructions. By leveraging only photometric and inferred panoptic image supervision, we can directly and jointly optimize the parameters of a set of nerflets so as to form a decomposed representation of the scene, where each object instance is represented by a group of nerflets. During experiments with indoor and outdoor environments, we find that nerflets: (1) fit and approximate the scene more efficiently than traditional global NeRFs, (2) allow the extraction of panoptic and photometric renderings from arbitrary views, and (3) enable tasks rare for NeRFs, such as 3D panoptic segmentation and interactive editing. Our project page."
                },
                {
                    "title": "S-NeRF: Neural Radiance Fields for Street Views",
                    "abstract": "Neural Radiance Fields (NeRFs) aim to synthesize novel views of objects and scenes, given the object-centric camera views with large overlaps. However, we conjugate that this paradigm does not fit the nature of the street views that are collected by many self-driving cars from the large-scale unbounded scenes. Also, the onboard cameras perceive scenes without much overlapping. Thus, existing NeRFs often produce blurs, 'floaters' and other artifacts on street-view synthesis. In this paper, we propose a new street-view NeRF (S-NeRF) that considers novel view synthesis of both the large-scale background scenes and the foreground moving vehicles jointly. Specifically, we improve the scene parameterization function and the camera poses for learning better neural representations from street views. We also use the the noisy and sparse LiDAR points to boost the training and learn a robust geometry and reprojection based confidence to address the depth outliers. Moreover, we extend our S-NeRF for reconstructing moving vehicles that is impracticable for conventional NeRFs. Thorough experiments on the large-scale driving datasets (e.g., nuScenes and Waymo) demonstrate that our method beats the state-of-the-art rivals by reducing 7% to 40% of the mean-squared error in the street-view synthesis and a 45% PSNR gain for the moving vehicles rendering."
                },
                {
                    "title": "Robust Reflection Removal With Flash-Only Cues in the Wild",
                    "abstract": "We propose a simple yet effective reflection-free cue for robust reflection removal from a pair of flash and ambient (no-flash) images. The reflection-free cue exploits a flash-only image obtained by subtracting the ambient image from the corresponding flash image in raw data space. The flash-only image is equivalent to an image taken in a dark environment with only a flash on. This flash-only image is visually reflection-free and thus can provide robust cues to infer the reflection in the ambient image. Since the flash-only image usually has artifacts, we further propose a dedicated model that not only utilizes the reflection-free cue but also avoids introducing artifacts, which helps accurately estimate reflection and transmission. Our experiments on real-world images with various types of reflection demonstrate the effectiveness of our model with reflection-free flash-only cues: our model outperforms state-of-the-art reflection removal approaches by more than 5.23 dB in PSNR. We extend our approach to handheld photography to address the misalignment between the flash and no-flash pair. With misaligned training data and the alignment module, our aligned model outperforms our previous version by more than 3.19 dB in PSNR on a misaligned dataset. We also study using linear RGB images as training data."
                },
                {
                    "title": "Panoptic NeRF: 3D-to-2D Label Transfer for Panoptic Urban Scene Segmentation",
                    "abstract": "Large-scale training data with high-quality annotations is critical for training semantic and instance segmentation models. Unfortunately, pixel-wise annotation is labor-intensive and costly, raising the demand for more efficient labeling strategies. In this work, we present a novel 3D-to-2D label transfer method, Panoptic NeRF1, which aims for obtaining per-pixel 2D semantic and instance labels from easy-to-obtain coarse 3D bounding primitives. Our method utilizes NeRF as a differentiable tool to unify coarse 3D annotations and 2D semantic cues transferred from existing datasets. We demonstrate that this combination allows for improved geometry guided by semantic information, enabling rendering of accurate semantic maps across multiple views. Furthermore, this fusion process resolves label ambiguity of the coarse 3D annotations and filters noise in the 2D predictions. By inferring in 3D space and rendering to 2D labels, our 2D semantic and instance labels are multiview consistent by design. Experimental results show that Panoptic NeRF outperforms existing label transfer methods in terms of accuracy and multi-view consistency on challenging urban scenes of the KITTI-360 dataset."
                },
                {
                    "title": "Block-NeRF: Scalable Large Scene Neural View Synthesis",
                    "abstract": "We present Block-NeRF, a variant of Neural Radiance Fields that can represent large-scale environments. Specifically, we demonstrate that when scaling NeRF to render city-scale scenes spanning multiple blocks, it is vital to de-compose the scene into individually trained NeRFs. This decomposition decouples rendering time from scene size, enables rendering to scale to arbitrarily large environments, and allows per-block updates of the environment. We adopt several architectural changes to make NeRF robust to data captured over months under different environmental conditions. We add appearance embeddings, learned pose refinement, and controllable exposure to each individual NeRF, and introduce a procedure for aligning appearance between adjacent NeRFs so that they can be seamlessly combined. We build a grid of Block-NeRFs from 2.8 million images to create the largest neural scene representation to date, capable of rendering an entire neighborhood of San Francisco."
                },
                {
                    "title": "Instant neural graphics primitives with a multiresolution hash encoding",
                    "abstract": "Neural graphics primitives, parameterized by fully connected neural networks, can be costly to train and evaluate. We reduce this cost with a versatile new input encoding that permits the use of a smaller network without sacrificing quality, thus significantly reducing the number of floating point and memory access operations: a small neural network is augmented by a multiresolution hash table of trainable feature vectors whose values are optimized through stochastic gradient descent. The multiresolution structure allows the network to disambiguate hash collisions, making for a simple architecture that is trivial to parallelize on modern GPUs. We leverage this parallelism by implementing the whole system using fully-fused CUDA kernels with a focus on minimizing wasted bandwidth and compute operations. We achieve a combined speedup of several orders of magnitude, enabling training of high-quality neural graphics primitives in a matter of seconds, and rendering in tens of milliseconds at a resolution of 1920\u00d71080."
                },
                {
                    "title": "Mega-NeRF: Scalable Construction of Large-Scale NeRFs for Virtual Fly- Throughs",
                    "abstract": "We use neural radiance fields (NeRFs) to build interac-tive 3D environments from large-scale visual captures spanning buildings or even multiple city blocks collected pri-marily from drones. In contrast to single object scenes (on which NeRFs are traditionally evaluated), our scale poses multiple challenges including (1) the need to model thou-sands of images with varying lighting conditions, each of which capture only a small subset of the scene, (2) pro-hibitively large model capacities that make it infeasible to train on a single GPU, and (3) significant challenges for fast rendering that would enable interactive fly-throughs. To address these challenges, we begin by analyzing visi-bility statistics for large-scale scenes, motivating a sparse network structure where parameters are specialized to dif-ferent regions of the scene. We introduce a simple geomet-ric clustering algorithm for data parallelism that partitions training images (or rather pixels) into different NeRF sub-modules that can be trained in parallel. We evaluate our approach on existing datasets (Quad 6k and UrbanScene3D) as well as against our own drone footage, improving training speed by 3x and PSNR by 12%. We also evaluate re-cent NeRF fast renderers on top of Mega-NeRF and intro-duce a novel method that exploits temporal coherence. Our technique achieves a 40x speedup over conventional NeRF rendering while remaining within 0.8 db in PSNR quality, exceeding the fidelity of existing fast renderers."
                },
                {
                    "title": "NeRFReN: Neural Radiance Fields with Reflections",
                    "abstract": "Neural Radiance Fields (NeRF) has achieved unprece-dented view synthesis quality using coordinate-based neu-ral scene representations. However, NeRF's view depen-dency can only handle simple reflections like highlights but cannot deal with complex reflections such as those from glass and mirrors. In these scenarios, NeRF models the virtual image as real geometries which leads to inaccurate depth estimation, and produces blurry renderings when the multi-view consistency is violated as the reflected objects may only be seen under some of the viewpoints. To over-come these issues, we introduce NeRFReN, which is built upon NeRF to model scenes with reflections. Specifically, we propose to split a scene into transmitted and reflected components, and model the two components with separate neural radiance fields. Considering that this decomposition is highly under-constrained, we exploit geometric priors and apply carefully-designed training strategies to achieve reasonable decomposition results. Experiments on various self-captured scenes show that our method achieves high-quality novel view synthesis and physically sound depth es-timation results while enabling scene editing applications."
                },
                {
                    "title": "KITTI-360: A Novel Dataset and Benchmarks for Urban Scene Understanding in 2D and 3D",
                    "abstract": "For the last few decades, several major subfields of artificial intelligence including computer vision, graphics, and robotics have progressed largely independently from each other. Recently, however, the community has realized that progress towards robust intelligent systems such as self-driving cars requires a concerted effort across the different fields. This motivated us to develop KITTI-360, successor of the popular KITTI dataset. KITTI-360 is a suburban driving dataset which comprises richer input modalities, comprehensive semantic instance annotations and accurate localization to facilitate research at the intersection of vision, graphics and robotics. For efficient annotation, we created a tool to label 3D scenes with bounding primitives and developed a model that transfers this information into the 2D image domain, resulting in over 150k images and 1B 3D points with coherent semantic instance annotations across 2D and 3D. Moreover, we established benchmarks and baselines for several tasks relevant to mobile perception, encompassing problems from computer vision, graphics, and robotics on the same dataset, e.g., semantic scene understanding, novel view synthesis and semantic SLAM. KITTI-360 will enable progress at the intersection of these research areas and thus contribute towards solving one of today's grand challenges: the development of fully autonomous self-driving systems."
                },
                {
                    "title": "METEOR: A Dense, Heterogeneous, and Unstructured Traffic Dataset with Rare Behaviors",
                    "abstract": "We present a new traffic dataset, Meteor, which captures traffic patterns and multi-agent driving behaviors in unstructured scenarios. Meteor consists of more than 1000 one-minute videos, over 2 million annotated frames with bounding boxes and GPS trajectories for 16 unique agent categories, and more than 13 million bounding boxes for traffic agents. Meteor is a dataset for rare and interesting, multi-agent driving behaviors that are grouped into traffic violations, atypical interactions, and diverse scenarios. Every video in Meteor is tagged using a diverse range of factors corresponding to weather, time of the day, road conditions, and traffic density. We use Meteor to benchmark perception methods for object detection and multi-agent behavior prediction. Our key finding is that state-of-the-art models for object detection and behavior prediction, which otherwise succeed on existing datasets such as Waymo, fail on the Meteor dataset. Meteor is a step towards developing more sophisticated perception models for dense, heterogeneous, and unstructured scenarios."
                },
                {
                    "title": "Pixel-Perfect Structure-from-Motion with Featuremetric Refinement",
                    "abstract": "Finding local features that are repeatable across multiple views is a cornerstone of sparse 3D reconstruction. The classical image matching paradigm detects keypoints per-image once and for all, which can yield poorly-localized features and propagate large errors to the final geometry. In this paper, we refine two key steps of structure-from-motion by a direct alignment of low-level image information from multiple views: we first adjust the initial keypoint locations prior to any geometric estimation, and subsequently refine points and camera poses as a post-processing. This refinement is robust to large detection noise and appearance changes, as it optimizes a featuremetric error based on dense features predicted by a neural network. This significantly improves the accuracy of camera poses and scene geometry for a wide range of keypoint detectors, challenging viewing conditions, and off-the-shelf deep features. Our system easily scales to large image collections, enabling pixel-perfect crowd-sourced localization at scale. Our code is publicly available at github.com/cvg/pixel-perfect-sfm as an add-on to the popular SfM software COLMAP."
                },
                {
                    "title": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields",
                    "abstract": "The rendering procedure used by neural radiance fields (NeRF) samples a scene with a single ray per pixel and may therefore produce renderings that are excessively blurred or aliased when training or testing images observe scene content at different resolutions. The straightforward solution of supersampling by rendering with multiple rays per pixel is impractical for NeRF, because rendering each ray requires querying a multilayer perceptron hundreds of times. Our solution, which we call \"mip-NeRF\" (\u00e0 la \"mipmap\"), extends NeRF to represent the scene at a continuously-valued scale. By efficiently rendering anti-aliased conical frustums instead of rays, mip-NeRF reduces objectionable aliasing artifacts and significantly improves NeRF\u2019s ability to represent fine details, while also being 7% faster than NeRF and half the size. Compared to NeRF, mip-NeRF reduces average error rates by 17% on the dataset presented with NeRF and by 60% on a challenging multiscale variant of that dataset that we present. Mip-NeRF is also able to match the accuracy of a brute-force supersampled NeRF on our multiscale dataset while being 22\u00d7 faster."
                },
                {
                    "title": "Robust Reflection Removal with Reflection-free Flash-only Cues",
                    "abstract": "We propose a simple yet effective reflection-free cue for robust reflection removal from a pair of flash and ambient (no-flash) images. The reflection-free cue exploits a flash-only image obtained by subtracting the ambient image from the corresponding flash image in raw data space. The flash-only image is equivalent to an image taken in a dark environment with only a flash on. We observe that this flash-only image is visually reflection-free, and thus it can provide robust cues to infer the reflection in the ambient image. Since the flash-only image usually has artifacts, we further propose a dedicated model that not only utilizes the reflection-free cue but also avoids introducing artifacts, which helps accurately estimate reflection and transmission. Our experiments on real-world images with various types of reflection demonstrate the effectiveness of our model with reflection-free flash-only cues: our model outperforms state-of-the-art reflection removal approaches by more than 5.23dB in PSNR, 0.04 in SSIM, and 0.068 in LPIPS. Our source code and dataset are publicly available at github.com/ChenyangLEI/flash-reflection-removal."
                },
                {
                    "title": "PatchmatchNet: Learned Multi-View Patchmatch Stereo",
                    "abstract": "We present PatchmatchNet, a novel and learnable cascade formulation of Patchmatch for high-resolution multi-view stereo. With high computation speed and low memory requirement, PatchmatchNet can process higher resolution imagery and is more suited to run on resource limited devices than competitors that employ 3D cost volume regularization. For the first time we introduce an iterative multi-scale Patchmatch in an end-to-end trainable architecture and improve the Patchmatch core algorithm with a novel and learned adaptive propagation and evaluation scheme for each iteration. Extensive experiments show a very competitive performance and generalization for our method on DTU, Tanks & Temples and ETH3D, but at a significantly higher efficiency than all existing top-performing models: at least two and a half times faster than state-of-the-art methods with twice less memory usage. Code is available at https://github.com/FangjinhuaWang/PatchmatchNet."
                },
                {
                    "title": "NeRF++: Analyzing and Improving Neural Radiance Fields",
                    "abstract": "Neural Radiance Fields (NeRF) achieve impressive view synthesis results for a variety of capture settings, including 360 capture of bounded scenes and forward-facing capture of bounded and unbounded scenes. NeRF fits multi-layer perceptrons (MLPs) representing view-invariant opacity and view-dependent color volumes to a set of training images, and samples novel views based on volume rendering techniques. In this technical report, we first remark on radiance fields and their potential ambiguities, namely the shape-radiance ambiguity, and analyze NeRF's success in avoiding such ambiguities. Second, we address a parametrization issue involved in applying NeRF to 360 captures of objects within large-scale, unbounded 3D scenes. Our method improves view synthesis fidelity in this challenging scenario. Code is available at this https URL."
                },
                {
                    "title": "Learning to See Through Obstructions",
                    "abstract": "We present a learning-based approach for removing unwanted obstructions, such as window reflections, fence occlusions or raindrops, from a short sequence of images captured by a moving camera. Our method leverages the motion differences between the background and the obstructing elements to recover both layers. Specifically, we alternate between estimating dense optical flow fields of the two layers and reconstructing each layer from the flow-warped images via a deep convolutional neural network. The learning-based layer reconstruction allows us to accommodate potential errors in the flow estimation and brittle assumptions such as brightness consistency. We show that training on synthetically generated data transfers well to real images. Our results on numerous challenging scenarios of reflection and fence removal demonstrate the effectiveness of the proposed method."
                },
                {
                    "title": "Polarized Reflection Removal With Perfect Alignment in the Wild",
                    "abstract": "We present a novel formulation to removing reflection from polarized images in the wild. We first identify the misalignment issues of existing reflection removal datasets where the collected reflection-free images are not perfectly aligned with input mixed images due to glass refraction. Then we build a new dataset with more than 100 types of glass in which obtained transmission images are perfectly aligned with input mixed images. Second, capitalizing on the special relationship between reflection and polarized light, we propose a polarized reflection removal model with a two-stage architecture. In addition, we design a novel perceptual NCC loss that can improve the performance of reflection removal and general image decomposition tasks. We conduct extensive experiments, and results suggest that our model outperforms state-of-the-art methods on reflection removal."
                },
                {
                    "title": "Scalability in Perception for Autonomous Driving: Waymo Open Dataset",
                    "abstract": "The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data. Existing self-driving datasets are limited in the scale and variation of the environments they capture, even though generalization within and between operating regions is crucial to the over-all viability of the technology. In an effort to help align the research community\u2019s contributions with real-world self-driving problems, we introduce a new large scale, high quality, diverse dataset. Our new dataset consists of 1150 scenes that each span 20 seconds, consisting of well synchronized and calibrated high quality LiDAR and camera data captured across a range of urban and suburban geographies. It is 15x more diverse than the largest camera+LiDAR dataset available based on our proposed diversity metric. We exhaustively annotated this data with 2D (camera image) and 3D (LiDAR) bounding boxes, with consistent identifiers across frames. Finally, we provide strong baselines for 2D as well as 3D detection and tracking tasks. We further study the effects of dataset size and generalization across geographies on 3D detection methods. Find data, code and more up-to-date information at http://www.waymo.com/open."
                },
                {
                    "title": "D2-City: A Large-Scale Dashcam Video Dataset of Diverse Traffic Scenarios",
                    "abstract": "Driving datasets accelerate the development of intelligent driving and related computer vision technologies, while substantial and detailed annotations serve as fuels and powers to boost the efficacy of such datasets to improve learning-based models. We propose D$^2$-City, a large-scale comprehensive collection of dashcam videos collected by vehicles on DiDi's platform. D$^2$-City contains more than 10000 video clips which deeply reflect the diversity and complexity of real-world traffic scenarios in China. We also provide bounding boxes and tracking annotations of 12 classes of objects in all frames of 1000 videos and detection annotations on keyframes for the remainder of the videos. Compared with existing datasets, D$^2$-City features data in varying weather, road, and traffic conditions and a huge amount of elaborate detection and tracking annotations. By bringing a diverse set of challenging cases to the community, we expect the D$^2$-City dataset will advance the perception and related areas of intelligent driving."
                },
                {
                    "title": "Single Image Reflection Removal Exploiting Misaligned Training Data and Network Enhancements",
                    "abstract": "Removing undesirable reflections from a single image captured through a glass window is of practical importance to visual computing systems. Although state-of-the-art methods can obtain decent results in certain situations, performance declines significantly when tackling more general real-world cases. These failures stem from the intrinsic difficulty of single image reflection removal -- the fundamental ill-posedness of the problem, and the insufficiency of densely-labeled training data needed for resolving this ambiguity within learning-based neural network pipelines. In this paper, we address these issues by exploiting targeted network enhancements and the novel use of misaligned data. For the former, we augment a baseline network architecture by embedding context encoding modules that are capable of leveraging high-level contextual clues to reduce indeterminacy within areas containing strong reflections. For the latter, we introduce an alignment-invariant loss function that facilitates exploiting misaligned real-world training data that is much easier to collect. Experimental results collectively show that our method outperforms the state-of-the-art with aligned data, and that significant improvements are possible when using additional misaligned data."
                },
                {
                    "title": "nuScenes: A Multimodal Dataset for Autonomous Driving",
                    "abstract": "Robust detection and tracking of objects is crucial for the deployment of autonomous vehicle technology. Image based benchmark datasets have driven development in computer vision tasks such as object detection, tracking and segmentation of agents in the environment. Most autonomous vehicles, however, carry a combination of cameras and range sensors such as lidar and radar. As machine learning based methods for detection and tracking become more prevalent, there is a need to train and evaluate such methods on datasets containing range sensor data along with images. In this work we present nuTonomy scenes (nuScenes), the first dataset to carry the full autonomous vehicle sensor suite: 6 cameras, 5 radars and 1 lidar, all with full 360 degree field of view. nuScenes comprises 1000 scenes, each 20s long and fully annotated with 3D bounding boxes for 23 classes and 8 attributes. It has 7x as many annotations and 100x as many images as the pioneering KITTI dataset. We define novel 3D detection and tracking metrics. We also provide careful dataset analysis as well as baselines for lidar and image based detection and tracking. Data, development kit and more information are available online."
                },
                {
                    "title": "Fast Single Image Reflection Suppression via Convex Optimization",
                    "abstract": "Removing undesired reflections from images taken through the glass is of great importance in computer vision. It serves as a means to enhance the image quality for aesthetic purposes as well as to preprocess images in machine learning and pattern recognition applications. We propose a convex model to suppress the reflection from a single input image. Our model implies a partial differential equation with gradient thresholding, which is solved efficiently using Discrete Cosine Transform. Extensive experiments on synthetic and real-world images demonstrate that our approach achieves desirable reflection suppression results and dramatically reduces the execution time compared to the state of the art."
                },
                {
                    "title": "From Coarse to Fine: Robust Hierarchical Localization at Large Scale",
                    "abstract": "Robust and accurate visual localization is a fundamental capability for numerous applications, such as autonomous driving, mobile robotics, or augmented reality. It remains, however, a challenging task, particularly for large-scale environments and in presence of significant appearance changes. State-of-the-art methods not only struggle with such scenarios, but are often too resource intensive for certain real-time applications. In this paper we propose HF-Net, a hierarchical localization approach based on a monolithic CNN that simultaneously predicts local features and global descriptors for accurate 6-DoF localization. We exploit the coarse-to-fine localization paradigm: we first perform a global retrieval to obtain location hypotheses and only later match local features within those candidate places. This hierarchical approach incurs significant runtime savings and makes our system suitable for real-time operation. By leveraging learned descriptors, our method achieves remarkable localization robustness across large variations of appearance and sets a new state-of-the-art on two challenging benchmarks for large-scale localization."
                },
                {
                    "title": "\u201cDouble-DIP\u201d: Unsupervised Image Decomposition via Coupled Deep-Image-Priors",
                    "abstract": "Many seemingly unrelated computer vision tasks can be viewed as a special case of image decomposition into separate layers. For example, image segmentation (separation into foreground and background layers); transparent layer separation (into reflection and transmission layers); Image dehazing (separation into a clear image and a haze map), and more. In this paper we propose a unified framework for unsupervised layer decomposition of a single image, based on coupled \"Deep-image-Prior\" (DIP) networks. It was shown [Ulyanov et al] that the structure of a single DIP generator network is sufficient to capture the low-level statistics of a single image. We show that coupling multiple such DIPs provides a powerful tool for decomposing images into their basic components, for a wide variety of applications. This capability stems from the fact that the internal statistics of a mixture of layers is more complex than the statistics of each of its individual components. We show the power of this approach for Image-Dehazing, Fg/Bg Segmentation, Watermark-Removal, Transparency Separation in images and video, and more. These capabilities are achieved in a totally unsupervised way, with no training examples other than the input image/video itself."
                },
                {
                    "title": "Single Image Reflection Separation with Perceptual Losses",
                    "abstract": "We present an approach to separating reflection from a single image. The approach uses a fully convolutional network trained end-to-end with losses that exploit low-level and high-level image information. Our loss function includes two perceptual losses: a feature loss from a visual perception network, and an adversarial loss that encodes characteristics of images in the transmission layers. We also propose a novel exclusion loss that enforces pixel-level layer separation. We create a dataset of real-world images with reflection and corresponding ground-truth transmission layers for quantitative evaluation and model training. We validate our method through comprehensive quantitative experiments and show that our approach outperforms state-of-the-art reflection removal methods in PSNR, SSIM, and perceptual user study. We also extend our method to two other image enhancement tasks to demonstrate the generality of our approach."
                },
                {
                    "title": "BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning",
                    "abstract": "Datasets drive vision progress, yet existing driving datasets are impoverished in terms of visual content and supported tasks to study multitask learning for autonomous driving. Researchers are usually constrained to study a small set of problems on one dataset, while real-world computer vision applications require performing tasks of various complexities. We construct BDD100K, the largest driving video dataset with 100K videos and 10 tasks to evaluate the exciting progress of image recognition algorithms on autonomous driving. The dataset possesses geographic, environmental, and weather diversity, which is useful for training models that are less likely to be surprised by new conditions. Based on this diverse dataset, we build a benchmark for heterogeneous multitask learning and study how to solve the tasks together. Our experiments show that special training strategies are needed for existing models to perform such heterogeneous tasks. BDD100K opens the door for future studies in this important venue."
                },
                {
                    "title": "The Unreasonable Effectiveness of Deep Features as a Perceptual Metric",
                    "abstract": "While it is nearly effortless for humans to quickly assess the perceptual similarity between two images, the underlying processes are thought to be quite complex. Despite this, the most widely used perceptual metrics today, such as PSNR and SSIM, are simple, shallow functions, and fail to account for many nuances of human perception. Recently, the deep learning community has found that features of the VGG network trained on ImageNet classification has been remarkably useful as a training loss for image synthesis. But how perceptual are these so-called \"perceptual losses\"? What elements are critical for their success? To answer these questions, we introduce a new dataset of human perceptual similarity judgments. We systematically evaluate deep features across different architectures and tasks and compare them with classic metrics. We find that deep features outperform all previous metrics by large margins on our dataset. More surprisingly, this result is not restricted to ImageNet-trained VGG features, but holds across different deep architectures and levels of supervision (supervised, self-supervised, or even unsupervised). Our results suggest that perceptual similarity is an emergent property shared across deep visual representations."
                },
                {
                    "title": "Benchmarking Single-Image Reflection Removal Algorithms",
                    "abstract": "Removing undesired reflections from a photo taken in front of a glass is of great importance for enhancing the efficiency of visual computing systems. Various approaches have been proposed and shown to be visually plausible on small datasets collected by their authors. A quantitative comparison of existing approaches using the same dataset has never been conducted due to the lack of suitable benchmark data with ground truth. This paper presents the first captured Single-image Reflection Removal dataset \u2018SIR2\u2019 with 40 controlled and 100 wild scenes, ground truth of background and reflection. For each controlled scene, we further provide ten sets of images under varying aperture settings and glass thicknesses. We perform quantitative and visual quality comparisons for four state-of-the-art single-image reflection removal algorithms using four error metrics. Open problems for improving reflection removal algorithms are discussed at the end."
                },
                {
                    "title": "PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume",
                    "abstract": "We present a compact but effective CNN model for optical flow, called PWC-Net. PWC-Net has been designed according to simple and well-established principles: pyramidal processing, warping, and the use of a cost volume. Cast in a learnable feature pyramid, PWC-Net uses the current optical flow estimate to warp the CNN features of the second image. It then uses the warped features and features of the first image to construct a cost volume, which is processed by a CNN to estimate the optical flow. PWC-Net is 17 times smaller in size and easier to train than the recent FlowNet2 model. Moreover, it outperforms all published optical flow methods on the MPI Sintel final pass and KITTI 2015 benchmarks, running at about 35 fps on Sintel resolution (1024 \u00c3\u2014 436) images. Our models are available on our project website."
                },
                {
                    "title": "A Generic Deep Architecture for Single Image Reflection Removal and Image Smoothing",
                    "abstract": "This paper proposes a deep neural network structure that exploits edge information in addressing representative low-level vision tasks such as layer separation and image filtering. Unlike most other deep learning strategies applied in this context, our approach tackles these challenging problems by estimating edges and reconstructing images using only cascaded convolutional layers arranged such that no handcrafted or application-specific image-processing components are required. We apply the resulting transferrable pipeline to two different problem domains that are both sensitive to edges, namely, single image reflection removal and image smoothing. For the former, using a mild reflection smoothness assumption and a novel synthetic data generation method that acts as a type of weak supervision, our network is able to solve much more difficult reflection cases that cannot be handled by previous methods. For the latter, we also exceed the state-of-the-art quantitative and qualitative results by wide margins. In all cases, the proposed framework is simple, fast, and easy to transfer across disparate domains."
                },
                {
                    "title": "Single Image Reflection Suppression",
                    "abstract": "Reflections are a common artifact in images taken through glass windows. Automatically removing the reflection artifacts after the picture is taken is an ill-posed problem. Attempts to solve this problem using optimization schemes therefore rely on various prior assumptions from the physical world. Instead of removing reflections from a single image, which has met with limited success so far, we propose a novel approach to suppress reflections. It is based on a Laplacian data fidelity term and an l-zero gradient sparsity term imposed on the output. With experiments on artificial and real-world images we show that our reflection suppression method performs better than the state-of-the-art reflection removal techniques."
                },
                {
                    "title": "Automatic Reflection Removal using Gradient Intensity and Motion Cues",
                    "abstract": "We present a method to separate the background image and reflection from two photos that are taken in front of a transparent glass under slightly different viewpoints. In our method, the SIFT-flow between two images is first calculated and a motion hierarchy is constructed from the SIFT-flow at multiple levels of spatial smoothness. To distinguish background edges and reflection edges, we calculate a motion score for each edge pixel by its variance along the motion hierarchy. Alternatively, we make use of the so-called superpixels to group edge pixels into edge segments and calculate the motion scores by averaging over each segment. In the meantime, we also calculate an intensity score for each edge pixel by its gradient magnitude. We combine both motion and intensity scores to get a combination score. A binary labelling (for separation) can be obtained by thresholding the combination scores. The background image is finally reconstructed from the separated gradients. Compared to the existing approaches that require a sequence of images or a video clip for the separation, we only need two images, which largely improves its feasibility. Various challenging examples are tested to validate the effectiveness of our method."
                },
                {
                    "title": "Reflection removal using ghosting cues",
                    "abstract": "Photographs taken through glass windows often contain both the desired scene and undesired reflections. Separating the reflection and transmission layers is an important but ill-posed problem that has both aesthetic and practical applications. In this work, we introduce the use of ghosting cues that exploit asymmetry between the layers, thereby helping to reduce the ill-posedness of the problem. These cues arise from shifted double reflections of the reflected scene off the glass surface. In double-pane windows, each pane reflects shifted and attenuated versions of objects on the same side of the glass as the camera. For single-pane windows, ghosting cues arise from shifted reflections on the two surfaces of the glass pane. Even though the ghosting is sometimes barely perceptible by humans, we can still exploit the cue for layer separation. In this work, we model the ghosted reflection using a double-impulse convolution kernel, and automatically estimate the spatial separation and relative attenuation of the ghosted reflection components. To separate the layers, we propose an algorithm that uses a Gaussian Mixture Model for regularization. Our method is automatic and requires only a single input image. We demonstrate that our approach removes a large fraction of reflections on both synthetic and real-world inputs."
                },
                {
                    "title": "Reflection removal for in-vehicle black box videos",
                    "abstract": "The in-vehicle black box camera (dashboard camera) has become a popular device in many countries for security monitoring and event capturing. The readability of video content is the most critical matter, however, the content is often degraded due to the windscreen reflection of objects inside. In this paper, we propose a novel method to remove the reflection on the windscreen from in-vehicle black box videos. The method exploits the spatio-temporal coherence of reflection, which states that a vehicle is moving forward while the reflection of the internal objects remains static. The average image prior is proposed by imposing a heavy-tail distribution with a higher peak to remove the reflection. The two-layered scene composed of reflection and background layers is the basis of the separation model. A non-convex cost function is developed based on this property and optimized in a fast way in a half quadratic form. Experimental results demonstrate that the proposed approach successfully separates the reflection layer in several real black box videos."
                },
                {
                    "title": "Single Image Layer Separation Using Relative Smoothness",
                    "abstract": "This paper addresses extracting two layers from an image where one layer is smoother than the other. This problem arises most notably in intrinsic image decomposition and reflection interference removal. Layer decomposition from a single-image is inherently ill-posed and solutions require additional constraints to be enforced. We introduce a novel strategy that regularizes the gradients of the two layers such that one has a long tail distribution and the other a short tail distribution. While imposing the long tail distribution is a common practice, our introduction of the short tail distribution on the second layer is unique. We formulate our problem in a probabilistic framework and describe an optimization scheme to solve this regularization with only a few iterations. We apply our approach to the intrinsic image and reflection removal problems and demonstrate high quality layer separation on par with other techniques but being significantly faster than prevailing methods."
                },
                {
                    "title": "PatchMatch: a randomized correspondence algorithm for structural image editing",
                    "abstract": "This paper presents interactive image editing tools using a new randomized algorithm for quickly finding approximate nearest-neighbor matches between image patches. Previous research in graphics and vision has leveraged such nearest-neighbor searches to provide a variety of high-level digital image editing tools. However, the cost of computing a field of such matches for an entire image has eluded previous efforts to provide interactive performance. Our algorithm offers substantial performance improvements over the previous state of the art (20-100x), enabling its use in interactive editing tools. The key insights driving the algorithm are that some good patch matches can be found via random sampling, and that natural coherence in the imagery allows us to propagate such matches quickly to surrounding areas. We offer theoretical analysis of the convergence properties of the algorithm, as well as empirical and practical evidence for its high quality and performance. This one simple algorithm forms the basis for a variety of tools -- image retargeting, completion and reshuffling -- that can be used together in the context of a high-level image editing application. Finally, we propose additional intuitive constraints on the synthesis process that offer the user a level of control unavailable in previous methods."
                },
                {
                    "title": "Image quality assessment: from error visibility to structural similarity",
                    "abstract": "Objective methods for assessing perceptual image quality traditionally attempted to quantify the visibility of errors (differences) between a distorted image and a reference image using a variety of known properties of the human visual system. Under the assumption that human visual perception is highly adapted for extracting structural information from a scene, we introduce an alternative complementary framework for quality assessment based on the degradation of structural information. As a specific example of this concept, we develop a structural similarity index and demonstrate its promise through a set of intuitive examples, as well as comparison to both subjective ratings and state-of-the-art objective methods on a database of images compressed with JPEG and JPEG2000. A MATLAB implementation of the proposed algorithm is available online at http://www.cns.nyu.edu//spl sim/lcv/ssim/."
                },
                {
                    "title": "Layer extraction from multiple images containing reflections and transparency",
                    "abstract": "Many natural images contain reflections and transparency, i.e., they contain mixtures of reflected and transmitted light. When viewed from a moving camera, these appear as the superposition of component layer images moving relative to each other. The problem of multiple motion recovery has been previously studied by a number of researchers. However no one has yet demonstrated how to accurately recover the component images themselves. In this paper we develop an optimal approach to recovering layer images and their associated motions from an arbitrary number of composite images. We develop two different techniques for estimating the component layer images given known motion estimates. The first approach uses constrained least squares to recover the layer images. The second approach iteratively refines lower and upper bounds on the layer images using two novel compositing operations, namely minimum- and maximum-composites of aligned images. We combine these layer extraction techniques with a dominant motion estimator and a subsequent motion refinement stage. This results in a completely automated system that recovers transparent images and motions from a collection of input images."
                },
                {
                    "title": "Polarization-based decorrelation of transparent layers: The inclination angle of an invisible surface",
                    "abstract": "When a transparent surface is present between an observer and an object, an image reflected by the surface may be superimposed on the image of the observed object. We present a new approach to recover the scenes (layers) and to classify which is the reflected/transmitted one, based on imaging through a polarizing filter at two orientations. Estimates of the separate layers are obtained by weighted pixel-wise differences of these images, inverting the image formation process. However the weights depend on the angle of incidence, hence on the inclination of the transparent (invisible) surface. This angle is estimated by seeking the angle-value which (through the weights) leads to decorrelation of the estimated layers. Experimental results, obtained using real photos of actual objects, demonstrate the success of angle estimation and consequent layer separation and labeling. The method is shown to be superior to earlier methods where only raw optical data was used."
                },
                {
                    "title": "Separating reflections and lighting using independent components analysis",
                    "abstract": "The image of an object can vary dramatically depending on lighting, specularities/reflections and shadows. It is often advantageous to separate these incidental variations from the intrinsic aspects of an image. This paper describes how the statistical tool of independent components analysis can be used to separate some of these incidental components. We describe the details of this method and show its efficacy with examples of separating reflections off glass, and separating the relative contributions of individual light sources."
                },
                {
                    "title": "A Double Dip",
                    "abstract": "Last fall a double Rip van Winkle occurred that should be encouraging to long-lived entomologists who believe in Auld Lang Syne. If one Rip equals 20 years, F. L. Campbell and friends enjoyed almost two Rips when Richard and Marion Le Pelley returned to the US for a visit after an absence of about 38 years. Since his retirement from his post as Chief Entomologist of the then British colony of Kenya, the Le Pelleys have lived and worked on their home island of Guernsey. There and in England and France they have entertained the Campbells on four different occasions, and FLC has caused Le Pelley to become a member of ESA. Thanks to his access to the Library of the British Museum in London, he was able in the past few years to write and publish a comprehensive book on coffee pests\u2014from mealy bugs to elephants. More recently he has written a review on coffee insects only. His last publication of interest to American entomologists concerns the beginning of pyrethrum growing in Kenya, with which he was personally connected."
                },
                {
                    "title": "Switch-NeRF: Learning Scene Decomposition with Mixture of Experts for Large-scale Neural Radiance Fields",
                    "abstract": "The Neural Radiance Fields (NeRF) have been recently applied to reconstruct building-scale and even city-scale scenes. To model a large-scale scene efficiently, a dominant strategy is to employ a divide-and-conquer paradigm via performing scene decomposition, which decomposes a complex scene into parts that are further processed by different sub-networks. Existing large-scale NeRFs mainly use heuristic hand-crafted scene decomposition, with regular 3D-distance-based or physical-street-block-based schemes. Although achieving promising results, the hand-crafted schemes limit the capabilities of NeRF in large-scale scene modeling in several aspects. Manually designing a universal scene decomposition rule for different complex scenes is challenging, leading to adaptation issues for different scenarios. The decomposition procedure is not learnable, hindering the network from jointly optimizing the scene decomposition and the radiance fields in an end-to-end manner. The different sub-networks are typically optimized independently, and thus hand-crafted rules are required to composite them to achieve a better consistency. To tackle these issues, we propose Switch-NeRF, a novel end-to-end large-scale NeRF with learning-based scene decomposition. We design a gating network to dispatch 3D points to different NeRF sub-networks. The gating network can be optimized together with the NeRF sub-networks for different scene partitions, by a design with the Sparsely Gated Mixture of Experts (MoE). The outputs from different sub-networks can also be fused in a learnable way in the unified framework to effectively guarantee the consistency of the whole scene. Furthermore, the proposed MoE-based Switch-NeRF model is carefully implemented and optimized to achieve both high-fidelity scene reconstruction and efficient computation. Our method establishes clear state-ofthe-art performances on several large-scale datasets. To the best of our knowledge, we are the first to propose an applicable end-to-end sparse NeRF network with learning-based decomposition for large-scale scenes. Codes are released at https://github.com/MiZhenxing/Switch-NeRF."
                },
                {
                    "title": "Neural Transmitted Radiance Fields",
                    "abstract": "Neural radiance fields (NeRF) have brought tremendous progress to novel view synthesis. Though NeRF enables the rendering of subtle details in a scene by learning from a dense set of images, it also reconstructs the undesired reflections when we capture images through glass. As a commonly observed interference, the reflection would undermine the visibility of the desired transmitted scene behind glass by occluding the transmitted light rays. In this paper, we aim at addressing the problem of rendering novel transmitted views given a set of reflection-corrupted images. By introducing the transmission encoder and recurring edge constraints as guidance, our neural transmitted radiance fields can resist such reflection interference during rendering and reconstruct high-fidelity results even under sparse views. The proposed method achieves superior performance from the experiments on a newly collected dataset compared with state-of-the-art methods. Our code and data is available at https://github.com/FreeButUselessSoul/TNeRF."
                },
                {
                    "title": "Removing Reflection From a Single Image With Ghosting Effect",
                    "abstract": "Removing the undesired reflections of images taken through glass is an important problem in digital photography and many other vision applications. The so-called ghosting effect, i.e.,\u00a0the pattern repetitiveness in reflection, is an effective cue used by existing techniques to remove reflection from images. Existing methods take a two-stage approach that first estimates the parameters of ghosting effect and then models reflection removal as a two-layer separation problem: reflection layer and latent image layer. This paper aims at addressing one main challenge in such an approach, i.e.,\u00a0how to distinguish the repetitive patterns on the later image layer and the ghosting patterns on the reflection layer. Based on the observation that the number of repeats of natural image patterns is often different from that of ghosting patterns, we propose a wavelet transform based regularization method. Together with a novel weighting scheme, the proposed method is capable of accurately separating two layers, and experimental results justified its advantages over the existing ones on both synthetic and real data set."
                },
                {
                    "title": "Reflection Separation using a Pair of Unpolarized and Polarized Images",
                    "abstract": "When we take photos through glass windows or doors, the transmitted background scene is often blended with undesirable reflection. Separating two layers apart to enhance the image quality is of vital importance for both human and machine perception. In this paper, we propose to exploit physical constraints from a pair of unpolarized and polarized images to separate reflection and transmission layers. Due to the simplified capturing setup, the system becomes more underdetermined compared with existing polarization based solutions that take three or more images as input. We propose to solve semireflector orientation estimation first to make the physical image formation well-posed and then learn to reliably separate two layers using a refinement network with gradient loss. Quantitative and qualitative experimental results show our approach performs favorably over existing polarization and single image based solutions."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively model high-fidelity obstruction-free 3D Gaussian Splatting from dash cam videos, considering the dynamic nature of obstructions such as reflections and occlusions on windshields?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of autonomous driving, as it enables the creation of more accurate 3D scene representations from dash cam footage. This has broader implications for the research community by enhancing the robustness of perception models and improving the simulation of driving scenarios. Addressing this question could lead to practical applications in real-time rendering and better understanding of complex driving environments, ultimately contributing to safer and more reliable autonomous vehicles.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the dynamic nature of obstructions on windshields, which are not static as assumed by existing methods. Naive approaches may fail because they do not account for the movement of obstructions, leading to inaccurate geometry and blurry renderings. Additionally, the diversity of obstructions and the limitations of current removal methods, which often rely on strict assumptions that do not hold in all cases, create significant technical and practical obstacles that need to be addressed.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has focused on single-image-based obstruction removal methods that impose strict assumptions, which do not generalize well to the diverse scenarios presented in dash cam videos. Additionally, existing NeRF methods are designed for static scenes and struggle with the dynamic nature of obstructions on windshields. The lack of a comprehensive approach that combines adaptive image decomposition and illumination-aware obstruction modeling has prevented this problem from being effectively solved until now.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, DC-Gaussian, includes three key components: 1) Adaptive image decomposition, which utilizes an opacity map to learn the transmittance of the windshield and estimate the background scene's contribution; 2) Illumination-aware Obstruction Modeling (IOM), which accounts for the dynamic nature of obstructions; and 3) Integration of these components into the 3D Gaussian Splatting framework. We will evaluate our approach using a dataset of dash cam videos, measuring rendering quality and geometry accuracy as key metrics. The expected outcomes include improved rendering fidelity and the ability to accurately model complex driving scenarios,"
            }
        },
        "author_data": {
            "193857e9-d3ac-4581-9524-84bc439f3844": {
                "pk": "193857e9-d3ac-4581-9524-84bc439f3844",
                "name": "Linhan Wang",
                "collaborators": [
                    "Shengkun Wang",
                    "Chang-Tien Lu",
                    "Taoran Ji",
                    "YangXiao Bai",
                    "Kaiqun Fu",
                    "Jianfeng He",
                    "Min Zhang",
                    "Shuo Lei",
                    "Mariam Almutairi",
                    "Dan Wang"
                ],
                "domain": [
                    "Financial Forecasting",
                    "Deep Learning",
                    "Few-Shot Learning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "ALERTA-Net: A Temporal Distance-Aware Recurrent Networks for Stock Movement and Volatility Prediction",
                        "abstract": "For both investors and policymakers, forecasting the stock market is essential as it serves as an indicator of economic well-being. To this end, we harness the power of social media data, a rich source of public sentiment, to enhance the accuracy of stock market predictions. Diverging from conventional methods, we pioneer an approach that integrates sentiment analysis, macroeconomic indicators, search engine data, and historical prices within a multi-attention deep learning model, masterfully decoding the complex patterns inherent in the data. We showcase the state-of-the-art performance of our proposed model using a dataset, specifically curated by us, for predicting stock market movements and volatility."
                    },
                    {
                        "title": "Self-Correlation and Cross-Correlation Learning for Few-Shot Remote Sensing Image Semantic Segmentation",
                        "abstract": "Remote sensing image semantic segmentation is an important problem for remote sensing image interpretation. Although remarkable progress has been achieved, existing deep neural network methods suffer from the reliance on massive training data. Few-shot remote sensing semantic segmentation aims at learning to segment target objects from a query image using only a few annotated support images of the target class. Most existing few-shot learning methods stem primarily from their sole focus on extracting information from support images, thereby failing to effectively address the large variance in appearance and scales of geographic objects. To tackle these challenges, we propose a Self-Correlation and Cross-Correlation Learning Network for the few-shot remote sensing image semantic segmentation. Our model enhances the generalization by considering both self-correlation and cross-correlation between support and query images to make segmentation predictions. To further explore the self-correlation with the query image, we propose to adopt a classical spectral method to produce a class-agnostic segmentation mask based on the basic visual information of the image. Extensive experiments on two remote sensing image datasets demonstrate the effectiveness and superiority of our model in few-shot remote sensing image semantic segmentation. Code and models will be accessed at https://github.com/linhanwang/SCCNet."
                    },
                    {
                        "title": "Stock Movement and Volatility Prediction from Tweets, Macroeconomic Factors and Historical Prices",
                        "abstract": "Predicting stock market is vital for investors and policymakers, acting as a barometer of the economic health. We leverage social media data, a potent source of public sentiment, in tandem with macroeconomic indicators as government-compiled statistics, to refine stock market predictions. However, prior research using tweet data for stock market prediction faces three challenges. First, the quality of tweets varies widely. While many are filled with noise and irrelevant details, only a few genuinely mirror the actual market scenario. Second, solely focusing on the historical data of a particular stock without considering its sector can lead to oversight. Stocks within the same industry often exhibit correlated price behaviors. Lastly, simply forecasting the direction of price movement without assessing its magnitude is of limited value, as the extent of the rise or fall truly determines profitability. In this paper, diverging from the conventional methods, we pioneer an ECON. The framework has following advantages: First, ECON has an adept tweets filter that efficiently extracts and decodes the vast array of tweet data. Second, ECON discerns multi-level relationships among stocks, sectors, and macroeconomic factors through a self-aware mechanism in semantic space. Third, ECON offers enhanced accuracy in predicting substantial stock price fluctuations by capitalizing on stock price movement. We showcase the state-of-the-art performance of our proposed model using a dataset, specifically curated by us, for predicting stock market movements and volatility."
                    },
                    {
                        "title": "AMA-LSTM: Pioneering Robust and Fair Financial Audio Analysis for Stock Volatility Prediction",
                        "abstract": "Stock volatility prediction is an important task in the financial industry. Recent advancements in multimodal methodologies, which integrate both textual and auditory data, have demonstrated significant improvements in this domain, such as earnings calls (Earnings calls are public available and often involve the management team of a public company and interested parties to discuss the company's earnings). However, these multimodal methods have faced two drawbacks. First, they often fail to yield reliable models and overfit the data due to their absorption of stochastic information from the stock market. Moreover, using multimodal models to predict stock volatility suffers from gender bias and lacks an efficient way to eliminate such bias. To address these aforementioned problems, we use adversarial training to generate perturbations that simulate the inherent stochasticity and bias, by creating areas resistant to random information around the input space to improve model robustness and fairness. Our comprehensive experiments on two real-world financial audio datasets reveal that this method exceeds the performance of current state-of-the-art solution. This confirms the value of adversarial training in reducing stochasticity and bias for stock volatility prediction tasks."
                    },
                    {
                        "title": "StockTime: A Time Series Specialized Large Language Model Architecture for Stock Price Prediction",
                        "abstract": "The stock price prediction task holds a significant role in the financial domain and has been studied for a long time. Recently, large language models (LLMs) have brought new ways to improve these predictions. While recent financial large language models (FinLLMs) have shown considerable progress in financial NLP tasks compared to smaller pre-trained language models (PLMs), challenges persist in stock price forecasting. Firstly, effectively integrating the modalities of time series data and natural language to fully leverage these capabilities remains complex. Secondly, FinLLMs focus more on analysis and interpretability, which can overlook the essential features of time series data. Moreover, due to the abundance of false and redundant information in financial markets, models often produce less accurate predictions when faced with such input data. In this paper, we introduce StockTime, a novel LLM-based architecture designed specifically for stock price data. Unlike recent FinLLMs, StockTime is specifically designed for stock price time series data. It leverages the natural ability of LLMs to predict the next token by treating stock prices as consecutive tokens, extracting textual information such as stock correlations, statistical trends and timestamps directly from these stock prices. StockTime then integrates both textual and time series data into the embedding space. By fusing this multimodal data, StockTime effectively predicts stock prices across arbitrary look-back periods. Our experiments demonstrate that StockTime outperforms recent LLMs, as it gives more accurate predictions while reducing memory usage and runtime costs."
                    }
                ]
            },
            "2a6ecf3a-163b-4623-bc10-1f827dfb9656": {
                "pk": "2a6ecf3a-163b-4623-bc10-1f827dfb9656",
                "name": "Kai Cheng",
                "collaborators": [
                    "Ralf Zimmermann",
                    "Yichi Gu",
                    "Qisheng Wu",
                    "Jing Li"
                ],
                "domain": [
                    "Surrogate Modeling",
                    "Deep Learning",
                    "Machine Learning",
                    "Semantic Segmentation"
                ],
                "publications": [
                    {
                        "title": "Sliced gradient-enhanced Kriging for high-dimensional function approximation",
                        "abstract": "Gradient-enhanced Kriging (GE-Kriging) is a well-established surrogate modelling technique for approximating expensive computational models. However, it tends to get impractical for high-dimensional problems due to the size of the inherent correlation matrix and the associated high-dimensional hyper-parameter tuning problem. To address these issues, a new method, called sliced GE-Kriging (SGE-Kriging), is developed in this paper for reducing both the size of the correlation matrix and the number of hyper-parameters. We first split the training sample set into multiple slices, and invoke Bayes' theorem to approximate the full likelihood function via a sliced likelihood function, in which multiple small correlation matrices are utilized to describe the correlation of the sample set rather than one large one. Then, we replace the original high-dimensional hyper-parameter tuning problem with a low-dimensional counterpart by learning the relationship between the hyper-parameters and the derivative-based global sensitivity indices. The performance of SGE-Kriging is finally validated by means of numerical experiments with several benchmarks and a high-dimensional aerodynamic modeling problem. The results show that the SGE-Kriging model features an accuracy and robustness that is comparable to the standard one but comes at much less training costs. The benefits are most evident for high-dimensional problems with tens of variables."
                    },
                    {
                        "title": "Gaussian Filter in CRF Based Semantic Segmentation",
                        "abstract": "Artificial intelligence is making great changes in academy and industry with the fast development of deep learning, which is a branch of machine learning and statistical learning. Fully convolutional network [1] is the standard model for semantic segmentation. Conditional random fields coded as CNN [2] or RNN [3] and connected with FCN has been successfully applied in object detection [4]. In this paper, we introduce a multi-resolution neural network for FCN and apply Gaussian filter to the extended CRF kernel neighborhood and the label image to reduce the oscillating effect of CRF neural network segmentation, thus achieve higher precision and faster training speed."
                    }
                ]
            },
            "f8f71c53-89bd-4162-89c6-8a4b87c2f8f2": {
                "pk": "f8f71c53-89bd-4162-89c6-8a4b87c2f8f2",
                "name": "Shuo Lei",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "c1045bde-1784-4e43-857a-038c60dceff5": {
                "pk": "c1045bde-1784-4e43-857a-038c60dceff5",
                "name": "Shengkun Wang",
                "collaborators": [
                    "Linhan Wang",
                    "Chang-Tien Lu",
                    "Taoran Ji",
                    "YangXiao Bai",
                    "Kaiqun Fu",
                    "Jianfeng He",
                    "Min Zhang",
                    "Shuo Lei",
                    "Mariam Almutairi",
                    "Dan Wang"
                ],
                "domain": [
                    "Financial Forecasting",
                    "Deep Learning",
                    "Few-Shot Learning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "ALERTA-Net: A Temporal Distance-Aware Recurrent Networks for Stock Movement and Volatility Prediction",
                        "abstract": "For both investors and policymakers, forecasting the stock market is essential as it serves as an indicator of economic well-being. To this end, we harness the power of social media data, a rich source of public sentiment, to enhance the accuracy of stock market predictions. Diverging from conventional methods, we pioneer an approach that integrates sentiment analysis, macroeconomic indicators, search engine data, and historical prices within a multi-attention deep learning model, masterfully decoding the complex patterns inherent in the data. We showcase the state-of-the-art performance of our proposed model using a dataset, specifically curated by us, for predicting stock market movements and volatility."
                    },
                    {
                        "title": "Self-Correlation and Cross-Correlation Learning for Few-Shot Remote Sensing Image Semantic Segmentation",
                        "abstract": "Remote sensing image semantic segmentation is an important problem for remote sensing image interpretation. Although remarkable progress has been achieved, existing deep neural network methods suffer from the reliance on massive training data. Few-shot remote sensing semantic segmentation aims at learning to segment target objects from a query image using only a few annotated support images of the target class. Most existing few-shot learning methods stem primarily from their sole focus on extracting information from support images, thereby failing to effectively address the large variance in appearance and scales of geographic objects. To tackle these challenges, we propose a Self-Correlation and Cross-Correlation Learning Network for the few-shot remote sensing image semantic segmentation. Our model enhances the generalization by considering both self-correlation and cross-correlation between support and query images to make segmentation predictions. To further explore the self-correlation with the query image, we propose to adopt a classical spectral method to produce a class-agnostic segmentation mask based on the basic visual information of the image. Extensive experiments on two remote sensing image datasets demonstrate the effectiveness and superiority of our model in few-shot remote sensing image semantic segmentation. Code and models will be accessed at https://github.com/linhanwang/SCCNet."
                    },
                    {
                        "title": "Stock Movement and Volatility Prediction from Tweets, Macroeconomic Factors and Historical Prices",
                        "abstract": "Predicting stock market is vital for investors and policymakers, acting as a barometer of the economic health. We leverage social media data, a potent source of public sentiment, in tandem with macroeconomic indicators as government-compiled statistics, to refine stock market predictions. However, prior research using tweet data for stock market prediction faces three challenges. First, the quality of tweets varies widely. While many are filled with noise and irrelevant details, only a few genuinely mirror the actual market scenario. Second, solely focusing on the historical data of a particular stock without considering its sector can lead to oversight. Stocks within the same industry often exhibit correlated price behaviors. Lastly, simply forecasting the direction of price movement without assessing its magnitude is of limited value, as the extent of the rise or fall truly determines profitability. In this paper, diverging from the conventional methods, we pioneer an ECON. The framework has following advantages: First, ECON has an adept tweets filter that efficiently extracts and decodes the vast array of tweet data. Second, ECON discerns multi-level relationships among stocks, sectors, and macroeconomic factors through a self-aware mechanism in semantic space. Third, ECON offers enhanced accuracy in predicting substantial stock price fluctuations by capitalizing on stock price movement. We showcase the state-of-the-art performance of our proposed model using a dataset, specifically curated by us, for predicting stock market movements and volatility."
                    },
                    {
                        "title": "AMA-LSTM: Pioneering Robust and Fair Financial Audio Analysis for Stock Volatility Prediction",
                        "abstract": "Stock volatility prediction is an important task in the financial industry. Recent advancements in multimodal methodologies, which integrate both textual and auditory data, have demonstrated significant improvements in this domain, such as earnings calls (Earnings calls are public available and often involve the management team of a public company and interested parties to discuss the company's earnings). However, these multimodal methods have faced two drawbacks. First, they often fail to yield reliable models and overfit the data due to their absorption of stochastic information from the stock market. Moreover, using multimodal models to predict stock volatility suffers from gender bias and lacks an efficient way to eliminate such bias. To address these aforementioned problems, we use adversarial training to generate perturbations that simulate the inherent stochasticity and bias, by creating areas resistant to random information around the input space to improve model robustness and fairness. Our comprehensive experiments on two real-world financial audio datasets reveal that this method exceeds the performance of current state-of-the-art solution. This confirms the value of adversarial training in reducing stochasticity and bias for stock volatility prediction tasks."
                    },
                    {
                        "title": "StockTime: A Time Series Specialized Large Language Model Architecture for Stock Price Prediction",
                        "abstract": "The stock price prediction task holds a significant role in the financial domain and has been studied for a long time. Recently, large language models (LLMs) have brought new ways to improve these predictions. While recent financial large language models (FinLLMs) have shown considerable progress in financial NLP tasks compared to smaller pre-trained language models (PLMs), challenges persist in stock price forecasting. Firstly, effectively integrating the modalities of time series data and natural language to fully leverage these capabilities remains complex. Secondly, FinLLMs focus more on analysis and interpretability, which can overlook the essential features of time series data. Moreover, due to the abundance of false and redundant information in financial markets, models often produce less accurate predictions when faced with such input data. In this paper, we introduce StockTime, a novel LLM-based architecture designed specifically for stock price data. Unlike recent FinLLMs, StockTime is specifically designed for stock price time series data. It leverages the natural ability of LLMs to predict the next token by treating stock prices as consecutive tokens, extracting textual information such as stock correlations, statistical trends and timestamps directly from these stock prices. StockTime then integrates both textual and time series data into the embedding space. By fusing this multimodal data, StockTime effectively predicts stock prices across arbitrary look-back periods. Our experiments demonstrate that StockTime outperforms recent LLMs, as it gives more accurate predictions while reducing memory usage and runtime costs."
                    }
                ]
            },
            "bef039b2-e5e0-4f08-b0f5-833827e32044": {
                "pk": "bef039b2-e5e0-4f08-b0f5-833827e32044",
                "name": "Wei Yin",
                "collaborators": [],
                "domain": [
                    "Algebraic Number Theory",
                    "Cohomology",
                    "Class Field Theory"
                ],
                "publications": [
                    {
                        "title": "On the First Cohomology of Local Units",
                        "abstract": "The groups of units $U^i_L$ of a local field $L$ play an important role in algebraic number theory, especially in class field theoretic topics. Therefore, it is interesting to study these groups from a cohomological point of view. In this article, we study and compute the first cohomology of $U_L^1$, $U^2_L$ and $U^3_L$ under certain mild hypotheses, and discuss some results about general $U^i_L$'s."
                    }
                ]
            },
            "5ff839a8-70bf-41fc-b179-32bd8004b4af": {
                "pk": "5ff839a8-70bf-41fc-b179-32bd8004b4af",
                "name": "Chenyang Lei",
                "collaborators": [
                    "Qifeng Chen",
                    "Chenyang Qi",
                    "Yue Wu",
                    "Yazhou Xing",
                    "Xuanchi Ren",
                    "Zhaoxiang Zhang",
                    "Xuhua Huang",
                    "Yankun Zhao",
                    "Wenxiu Sun",
                    "Qiong Yan"
                ],
                "domain": [
                    "Image Processing",
                    "Video Colorization",
                    "Reflection Removal",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Fully Automatic Video Colorization with Self-Regularization and Diversity",
                        "abstract": "We present a fully automatic approach to video colorization with self-regularization and diversity. Our model contains a colorization network for video frame colorization and a refinement network for spatiotemporal color refinement. Without any labeled data, both networks can be trained with self-regularized losses defined in bilateral and temporal space. The bilateral loss enforces color consistency between neighboring pixels in a bilateral space and the temporal loss imposes constraints between corresponding pixels in two nearby frames. While video colorization is a multi-modal problem, our method uses a perceptual loss with diversity to differentiate various modes in the solution space. Perceptual experiments demonstrate that our approach outperforms state-of-the-art approaches on fully automatic video colorization. The results are shown in the supplementary video at https://youtu.be/Y15uv2jnK-4"
                    },
                    {
                        "title": "Robust Reflection Removal with Reflection-free Flash-only Cues",
                        "abstract": "We propose a simple yet effective reflection-free cue for robust reflection removal from a pair of flash and ambient (no-flash) images. The reflection-free cue exploits a flash-only image obtained by subtracting the ambient image from the corresponding flash image in raw data space. The flash-only image is equivalent to an image taken in a dark environment with only a flash on. We observe that this flash-only image is visually reflection-free, and thus it can provide robust cues to infer the reflection in the ambient image. Since the flash-only image usually has artifacts, we further propose a dedicated model that not only utilizes the reflection-free cue but also avoids introducing artifacts, which helps accurately estimate reflection and transmission. Our experiments on real-world images with various types of reflection demonstrate the effectiveness of our model with reflection-free flash-only cues: our model outperforms state-of-the-art reflection removal approaches by more than 5.23dB in PSNR, 0.04 in SSIM, and 0.068 in LPIPS. Our source code and dataset are publicly available at {github.com/ChenyangLEI/flash-reflection-removal}."
                    },
                    {
                        "title": "Towards Photorealistic Colorization by Imagination",
                        "abstract": "We present a novel approach to automatic image colorization by imitating the imagination process of human experts. Our imagination module is designed to generate color images that are context-correlated with black-and-white photos. Given a black-and-white image, our imagination module firstly extracts the context information, which is then used to synthesize colorful and diverse images using a conditional image synthesis network (e.g., semantic image synthesis model). We then design a colorization module to colorize the black-and-white images with the guidance of imagination for photorealistic colorization. Experimental results show that our work produces more colorful and diverse results than state-of-the-art image colorization methods. Our source codes will be publicly available."
                    },
                    {
                        "title": "Blind Video Temporal Consistency via Deep Video Prior",
                        "abstract": "Applying image processing algorithms independently to each video frame often leads to temporal inconsistency in the resulting video. To address this issue, we present a novel and general approach for blind video temporal consistency. Our method is only trained on a pair of original and processed videos directly instead of a large dataset. Unlike most previous methods that enforce temporal consistency with optical flow, we show that temporal consistency can be achieved by training a convolutional network on a video with the Deep Video Prior. Moreover, a carefully designed iteratively reweighted training strategy is proposed to address the challenging multimodal inconsistency problem. We demonstrate the effectiveness of our approach on 7 computer vision tasks on videos. Extensive quantitative and perceptual experiments show that our approach obtains superior performance than state-of-the-art methods on blind video temporal consistency. Our source codes are publicly available at github.com/ChenyangLEI/deep-video-prior."
                    },
                    {
                        "title": "Blind Video Deflickering by Neural Filtering with a Flawed Atlas",
                        "abstract": "Many videos contain flickering artifacts. Common causes of flicker include video processing algorithms, video generation algorithms, and capturing videos under specific situations. Prior work usually requires specific guidance such as the flickering frequency, manual annotations, or extra consistent videos to remove the flicker. In this work, we propose a general flicker removal framework that only receives a single flickering video as input without additional guidance. Since it is blind to a specific flickering type or guidance, we name this \"blind deflickering.\" The core of our approach is utilizing the neural atlas in cooperation with a neural filtering strategy. The neural atlas is a unified representation for all frames in a video that provides temporal consistency guidance but is flawed in many cases. To this end, a neural network is trained to mimic a filter to learn the consistent features (e.g., color, brightness) and avoid introducing the artifacts in the atlas. To validate our method, we construct a dataset that contains diverse real-world flickering videos. Extensive experiments show that our method achieves satisfying deflickering performance and even outperforms baselines that use extra guidance on a public benchmark."
                    },
                    {
                        "title": "A Categorized Reflection Removal Dataset with Diverse Real-world Scenes",
                        "abstract": "Due to the lack of a large-scale reflection removal dataset with diverse real-world scenes, many existing reflection removal methods are trained on synthetic data plus a small amount of real-world data, which makes it difficult to evaluate the strengths or weaknesses of different reflection removal methods thoroughly. Furthermore, existing real-world benchmarks and datasets do not categorize image data based on the types and appearances of reflection (e.g., smoothness, intensity), making it hard to analyze reflection removal methods. Hence, we construct a new reflection removal dataset that is categorized, diverse, and real-world (CDR). A pipeline based on RAW data is used to capture perfectly aligned input images and transmission images. The dataset is constructed using diverse glass types under various environments to ensure diversity. By analyzing several reflection removal methods and conducting extensive experiments on our dataset, we show that state-of-the-art reflection removal methods generally perform well on blurry reflection but fail in obtaining satisfying performance on other types of real-world reflection. We believe our dataset can help develop novel methods to remove real-world reflection better. Our dataset is available at https://alexzhao-hugga.github.io/Real-World-Reflection-Removal/."
                    },
                    {
                        "title": "Shape from Polarization for Complex Scenes in the Wild",
                        "abstract": "We present a new data-driven approach with physics-based priors to scene-level normal estimation from a single polarization image. Existing shape from polarization (SfP) works mainly focus on estimating the normal of a single object rather than complex scenes in the wild. A key barrier to high-quality scene-level SfP is the lack of real-world SfP data in complex scenes. Hence, we contribute the first real-world scene-level SfP dataset with paired input polarization images and ground-truth normal maps. Then we propose a learning-based framework with a multi-head self-attention module and viewing encoding, which is designed to handle increasing polarization ambiguities caused by complex materials and non-orthographic projection in scene-level SfP. Our trained model can be generalized to far-field outdoor scenes as the relationship between polarized light and surface normals is not affected by distance. Experimental results demonstrate that our approach significantly outperforms existing SfP models on two datasets. Our dataset and source code will be publicly available at https://github.com/ChenyangLEI/sfp-wild"
                    },
                    {
                        "title": "FateZero: Fusing Attentions for Zero-shot Text-based Video Editing",
                        "abstract": "The diffusion-based generative models have achieved remarkable success in text-based image generation. However, since it contains enormous randomness in generation progress, it is still challenging to apply such models for real-world visual content editing, especially in videos. In this paper, we propose FateZero, a zero-shot text-based editing method on real-world videos without per-prompt training or use-specific mask. To edit videos consistently, we propose several techniques based on the pre-trained models. Firstly, in contrast to the straightforward DDIM inversion technique, our approach captures intermediate attention maps during inversion, which effectively retain both structural and motion information. These maps are directly fused in the editing process rather than generated during denoising. To further minimize semantic leakage of the source video, we then fuse self-attentions with a blending mask obtained by cross-attention features from the source prompt. Furthermore, we have implemented a reform of the self-attention mechanism in denoising UNet by introducing spatial-temporal attention to ensure frame consistency. Yet succinct, our method is the first one to show the ability of zero-shot text-driven video style and local attribute editing from the trained text-to-image model. We also have a better zero-shot shape-aware editing ability based on the text-to-video model. Extensive experiments demonstrate our superior temporal consistency and editing capability than previous works."
                    },
                    {
                        "title": "Deep Video Prior for Video Consistency and Propagation",
                        "abstract": "Applying an image processing algorithm independently to each video frame often leads to temporal inconsistency in the resulting video. To address this issue, we present a novel and general approach for blind video temporal consistency. Our method is only trained on a pair of original and processed videos directly instead of a large dataset. Unlike most previous methods that enforce temporal consistency with optical flow, we show that temporal consistency can be achieved by training a convolutional neural network on a video with Deep Video Prior (DVP). Moreover, a carefully designed iteratively reweighted training strategy is proposed to address the challenging multimodal inconsistency problem. We demonstrate the effectiveness of our approach on 7 computer vision tasks on videos. Extensive quantitative and perceptual experiments show that our approach obtains superior performance than state-of-the-art methods on blind video temporal consistency. We further extend DVP to video propagation and demonstrate its effectiveness in propagating three different types of information (color, artistic style, and object segmentation). A progressive propagation strategy with pseudo labels is also proposed to enhance DVP's performance on video propagation. Our source codes are publicly available at https://github.com/ChenyangLEI/deep-video-prior."
                    },
                    {
                        "title": "Automatic Controllable Colorization via Imagination",
                        "abstract": "We propose a framework for automatic colorization that allows for iterative editing and modifications. The core of our framework lies in an imagination module: by understanding the content within a grayscale image, we utilize a pre-trained image generation model to generate multiple images that contain the same content. These images serve as references for coloring, mimicking the process of human experts. As the synthesized images can be imperfect or different from the original grayscale image, we propose a Reference Refinement Module to select the optimal reference composition. Unlike most previous end-to-end automatic colorization algorithms, our framework allows for iterative and localized modifications of the colorization results because we explicitly model the coloring samples. Extensive experiments demonstrate the superiority of our framework over existing automatic colorization algorithms in editability and flexibility. Project page: https://xy-cong.github.io/imagine-colorization."
                    },
                    {
                        "title": "Robust Reflection Removal with Flash-only Cues in the Wild",
                        "abstract": "We propose a simple yet effective reflection-free cue for robust reflection removal from a pair of flash and ambient (no-flash) images. The reflection-free cue exploits a flash-only image obtained by subtracting the ambient image from the corresponding flash image in raw data space. The flash-only image is equivalent to an image taken in a dark environment with only a flash on. This flash-only image is visually reflection-free and thus can provide robust cues to infer the reflection in the ambient image. Since the flash-only image usually has artifacts, we further propose a dedicated model that not only utilizes the reflection-free cue but also avoids introducing artifacts, which helps accurately estimate reflection and transmission. Our experiments on real-world images with various types of reflection demonstrate the effectiveness of our model with reflection-free flash-only cues: our model outperforms state-of-the-art reflection removal approaches by more than 5.23dB in PSNR. We extend our approach to handheld photography to address the misalignment between the flash and no-flash pair. With misaligned training data and the alignment module, our aligned model outperforms our previous version by more than 3.19dB in PSNR on a misaligned dataset. We also study using linear RGB images as training data. Our source code and dataset are publicly available at https://github.com/ChenyangLEI/flash-reflection-removal."
                    }
                ]
            },
            "3bc0395f-5431-43e4-9a61-888e2f255e35": {
                "pk": "3bc0395f-5431-43e4-9a61-888e2f255e35",
                "name": "Xiaoxiao Long",
                "collaborators": [
                    "Lingjie Liu",
                    "Wenping Wang",
                    "Christian Theobalt",
                    "Hanyang Yu",
                    "Ping Tan",
                    "Wei Li",
                    "Cheng Lin",
                    "Peng Wang",
                    "Taku Komura",
                    "Chen Wang"
                ],
                "domain": [
                    "3D Reconstruction",
                    "Depth Estimation",
                    "Neural Rendering",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "LM-Gaussian: Boost Sparse-view 3D Gaussian Splatting with Large Model Priors",
                        "abstract": "We aim to address sparse-view reconstruction of a 3D scene by leveraging priors from large-scale vision models. While recent advancements such as 3D Gaussian Splatting (3DGS) have demonstrated remarkable successes in 3D reconstruction, these methods typically necessitate hundreds of input images that densely capture the underlying scene, making them time-consuming and impractical for real-world applications. However, sparse-view reconstruction is inherently ill-posed and under-constrained, often resulting in inferior and incomplete outcomes. This is due to issues such as failed initialization, overfitting on input images, and a lack of details. To mitigate these challenges, we introduce LM-Gaussian, a method capable of generating high-quality reconstructions from a limited number of images. Specifically, we propose a robust initialization module that leverages stereo priors to aid in the recovery of camera poses and the reliable point clouds. Additionally, a diffusion-based refinement is iteratively applied to incorporate image diffusion priors into the Gaussian optimization process to preserve intricate scene details. Finally, we utilize video diffusion priors to further enhance the rendered images for realistic visual effects. Overall, our approach significantly reduces the data acquisition requirements compared to previous 3DGS methods. We validate the effectiveness of our framework through experiments on various public datasets, demonstrating its potential for high-quality 360-degree scene reconstruction. Visual results are on our website."
                    },
                    {
                        "title": "Multi-view Depth Estimation using Epipolar Spatio-Temporal Networks",
                        "abstract": "We present a novel method for multi-view depth estimation from a single video, which is a critical task in various applications, such as perception, reconstruction and robot navigation. Although previous learning-based methods have demonstrated compelling results, most works estimate depth maps of individual video frames independently, without taking into consideration the strong geometric and temporal coherence among the frames. Moreover, current state-of-the-art (SOTA) models mostly adopt a fully 3D convolution network for cost regularization and therefore require high computational cost, thus limiting their deployment in real-world applications. Our method achieves temporally coherent depth estimation results by using a novel Epipolar Spatio-Temporal (EST) transformer to explicitly associate geometric and temporal correlation with multiple estimated depth maps. Furthermore, to reduce the computational cost, inspired by recent Mixture-of-Experts models, we design a compact hybrid network consisting of a 2D context-aware network and a 3D matching network which learn 2D context information and 3D disparity cues separately. Extensive experiments demonstrate that our method achieves higher accuracy in depth estimation and significant speedup than the SOTA methods."
                    },
                    {
                        "title": "SparseNeuS: Fast Generalizable Neural Surface Reconstruction from Sparse Views",
                        "abstract": "We introduce SparseNeuS, a novel neural rendering based method for the task of surface reconstruction from multi-view images. This task becomes more difficult when only sparse images are provided as input, a scenario where existing neural reconstruction approaches usually produce incomplete or distorted results. Moreover, their inability of generalizing to unseen new scenes impedes their application in practice. Contrarily, SparseNeuS can generalize to new scenes and work well with sparse images (as few as 2 or 3). SparseNeuS adopts signed distance function (SDF) as the surface representation, and learns generalizable priors from image features by introducing geometry encoding volumes for generic surface prediction. Moreover, several strategies are introduced to effectively leverage sparse views for high-quality reconstruction, including 1) a multi-level geometry reasoning framework to recover the surfaces in a coarse-to-fine manner; 2) a multi-scale color blending scheme for more reliable color prediction; 3) a consistency-aware fine-tuning scheme to control the inconsistent regions caused by occlusion and noise. Extensive experiments demonstrate that our approach not only outperforms the state-of-the-art methods, but also exhibits good efficiency, generalizability, and flexibility."
                    },
                    {
                        "title": "Occlusion-Aware Depth Estimation with Adaptive Normal Constraints",
                        "abstract": "We present a new learning-based method for multi-frame depth estimation from a color video, which is a fundamental problem in scene understanding, robot navigation or handheld 3D reconstruction. While recent learning-based methods estimate depth at high accuracy, 3D point clouds exported from their depth maps often fail to preserve important geometric feature (e.g., corners, edges, planes) of man-made scenes. Widely-used pixel-wise depth errors do not specifically penalize inconsistency on these features. These inaccuracies are particularly severe when subsequent depth reconstructions are accumulated in an attempt to scan a full environment with man-made objects with this kind of features. Our depth estimation algorithm therefore introduces a Combined Normal Map (CNM) constraint, which is designed to better preserve high-curvature features and global planar regions. In order to further improve the depth estimation accuracy, we introduce a new occlusion-aware strategy that aggregates initial depth predictions from multiple adjacent views into one final depth map and one occlusion probability map for the current reference view. Our method outperforms the state-of-the-art in terms of depth estimation accuracy, and preserves essential geometric features of man-made indoor scenes much better than other algorithms."
                    },
                    {
                        "title": "GECO: Generative Image-to-3D within a SECOnd",
                        "abstract": "Recent years have seen significant advancements in 3D generation. While methods like score distillation achieve impressive results, they often require extensive per-scene optimization, which limits their time efficiency. On the other hand, reconstruction-based approaches are more efficient but tend to compromise quality due to their limited ability to handle uncertainty. We introduce GECO, a novel method for high-quality 3D generative modeling that operates within a second. Our approach addresses the prevalent issues of uncertainty and inefficiency in existing methods through a two-stage approach. In the first stage, we train a single-step multi-view generative model with score distillation. Then, a second-stage distillation is applied to address the challenge of view inconsistency in the multi-view generation. This two-stage process ensures a balanced approach to 3D generation, optimizing both quality and efficiency. Our comprehensive experiments demonstrate that GECO achieves high-quality image-to-3D mesh generation with an unprecedented level of efficiency. We will make the code and model publicly available."
                    }
                ]
            },
            "633d94e8-422e-4aed-b43a-6d12bcf211a9": {
                "pk": "633d94e8-422e-4aed-b43a-6d12bcf211a9",
                "name": "Chang-Tien Lu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2402.10946": {
        "paper_data": {
            "title": "CultureLLM: Incorporating Cultural Differences into Large Language Models",
            "url": "http://arxiv.org/abs/2402.10946v1",
            "arxiv_id": "2402.10946",
            "authors": [
                "Cheng Li",
                "Mengzhou Chen",
                "Jindong Wang",
                "Sunayana Sitaram",
                "Xing Xie"
            ],
            "abstract": "Large language models (LLMs) are reported to be partial to certain cultures owing to the training data dominance from the English corpora. Since multilingual cultural data are often expensive to collect, existing efforts handle this by prompt engineering or culture-specific pre-training. However, they might overlook the knowledge deficiency of low-resource culture and require extensive computing resources. In this paper, we propose CultureLLM, a cost-effective solution to incorporate cultural differences into LLMs. CultureLLM adopts World Value Survey (WVS) as seed data and generates semantically equivalent training data via the proposed semantic data augmentation. Using only 50 seed samples from WVS with augmented data, we fine-tune culture-specific LLMs and one unified model (CultureLLM-One) for 9 cultures covering rich and low-resource languages. Extensive experiments on 60 culture-related datasets demonstrate that CultureLLM significantly outperforms various counterparts such as GPT-3.5 (by 8.1%) and Gemini Pro (by 9.5%) with comparable performance to GPT-4 or even better. Our human study shows that the generated samples are semantically equivalent to the original samples, providing an effective solution for LLMs augmentation.",
            "introduction": "   1 Introduction  Culture is a complex construct that encodes various identities, including but not limited to, language, nationality, region, religion, and gender identity. Cultural bias is widely present around the world, which refers to the tendency to favor specific cultural perspectives, values, and norms that lead to subjective opinions and can offend people of other cultures. For example, according to the World Value Survey\u00a0(Survey, 2022), Arabic culture believes that men are better political leaders than women, while people in the United States disagree.333We respect all opinions in different cultures. As large language models (LLMs)\u00a0(OpenAI, 2023b; Google, 2023) become prevalent, they are reported to suffer cultural bias and are specifically partial to Western culture, as English corpora dominate training data\u00a0(Liu et\u00a0al., 2023b; Cao et\u00a0al., 2023; Masoud et\u00a0al., 2023; Naous et\u00a0al., 2023; Wang et\u00a0al., 2023d; Johnson et\u00a0al., 2022). Low-resource cultures exist widely due to the lack of training data from other cultures. LLMs\u2019 cultural bias presents a major bottleneck in human-AI collaboration and significantly hinders AI democracy.   Figure 1: Overview of CultureLLM. CultureLLM consists of three steps: sampling, semantic data augmentation, and fine-tuning. Both culture-specific and unified CultureLLM can be fine-tuned.   Tackling cultural bias requires an LLM to embrace cultural differences\u00a0(Hofstede, 1984). Kova\u010d et\u00a0al. (2023) and Wang et\u00a0al. (2023d) thought LLMs have enough knowledge of all cultures and devised prompt engineering technologies to induce LLMs to exhibit specific cultural perspectives. However, they are not effective, especially in low-resource cultures with limited data. Another line of work pre-trained culturally-aware LLMs and then fine-tuned on specific datasets\u00a0(Chan et\u00a0al., 2023; Nguyen et\u00a0al., 2023b; Pipatanakul et\u00a0al., 2023; Abbasi et\u00a0al., 2023; Lin and Chen, 2023). They require the collection of large-scale pre-training and fine-tuning datasets and extensive computing resources, thus are not affordable to ordinary researchers and cannot handle low-resource culture. To date, training culturally-aware LLMs at affordable costs remains a challenge.   In this paper, we propose CultureLLM, a cost-effective444Fine-tuning a CultureLLM only costs $6 via OpenAI API. solution to incorporate cultural differences into LLMs. To be specific, we focus on cultural values in this work. As shown in Figure\u00a01, CultureLLM consists of three steps: sampling, semantic data augmentation, and fine-tuning. Inspired by Attitude-Behavior Consistency theory\u00a0(Fazio and Zanna, 1981) which emphasizes that people\u2019s opinion is consistent with their behaviors, we use the World Values Survey (WVS)\u00a0(Survey, 2022) as seed data. Then, we devise a semantic data augmentation approach to generate semantically equivalent samples. The aim is to generate semantic equivalent inputs, thus we can get the ground-truth from seed data directly. Finally, CultureLLM is obtained by fine-tuning on both the seed and the generated data. WVS is a public opinion poll that contains people\u2019s opinions on cultural topics from different countries. To be specific, we select 50505050 seed samples from WVS, covering 7777 topics: \u201csocial values\", \u201cmigration\", \u201csecurity\", \u201cscience and technology\", \u201creligious values\", \u201cethical values and norms\", and \u201cpolitical interest and political participation\". Using these generated samples and answers from people in different cultures, we fine-tune specific and unified LLMs: specific LLMs are tailored for each culture such as CultureLLM-Ar for Arabic and CultureLLM-Tr for Turkish; unified LLMs (CultureLLM-One) are one LLM that fits all cultures.555The bound of culture is unclear, and we use the main spoken language to",
            "references": [
                {
                    "title": "Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models",
                    "abstract": "Harnessing the power of human-annotated data through Supervised Fine-Tuning (SFT) is pivotal for advancing Large Language Models (LLMs). In this paper, we delve into the prospect of growing a strong LLM out of a weak one without the need for acquiring additional human-annotated data. We propose a new fine-tuning method called Self-Play fIne-tuNing (SPIN), which starts from a supervised fine-tuned model. At the heart of SPIN lies a self-play mechanism, where the LLM refines its capability by playing against instances of itself. More specifically, the LLM generates its own training data from its previous iterations, refining its policy by discerning these self-generated responses from those obtained from human-annotated data. Our method progressively elevates the LLM from a nascent model to a formidable one, unlocking the full potential of human-annotated demonstration data for SFT. Theoretically, we prove that the global optimum to the training objective function of our method is achieved only when the LLM policy aligns with the target data distribution. Empirically, we evaluate our method on several benchmark datasets including the HuggingFace Open LLM Leaderboard, MT-Bench, and datasets from Big-Bench. Our results show that SPIN can significantly improve the LLM's performance across a variety of benchmarks and even outperform models trained through direct preference optimization (DPO) supplemented with extra GPT-4 preference data. This sheds light on the promise of self-play, enabling the achievement of human-level performance in LLMs without the need for expert opponents. Codes are available at https://github.com/uclaml/SPIN."
                },
                {
                    "title": "Improving Text Embeddings with Large Language Models",
                    "abstract": "In this paper, we introduce a novel and simple method for obtaining high-quality text embeddings using only synthetic data and less than 1k training steps. Unlike existing methods that often depend on multi-stage intermediate pre-training with billions of weakly-supervised text pairs, followed by fine-tuning with a few labeled datasets, our method does not require building complex training pipelines or relying on manually collected datasets that are often constrained by task diversity and language coverage. We leverage proprietary LLMs to generate diverse synthetic data for hundreds of thousands of text embedding tasks across 93 languages. We then fine-tune open-source decoder-only LLMs on the synthetic data using standard contrastive loss. Experiments demonstrate that our method achieves strong performance on highly competitive text embedding benchmarks without using any labeled data. Furthermore, when fine-tuned with a mixture of synthetic and labeled data, our model sets new state-of-the-art results on the BEIR and MTEB benchmarks."
                },
                {
                    "title": "Typhoon: Thai Large Language Models",
                    "abstract": "Typhoon is a series of Thai large language models (LLMs) developed specifically for the Thai language. This technical report presents challenges and insights in developing Thai LLMs, including data preparation, pretraining, instruction-tuning, and evaluation. As one of the challenges of low-resource languages is the amount of pretraining data, we apply continual training to transfer existing world knowledge from a strong LLM. To evaluate the Thai knowledge encapsulated in each model from the pretraining stage, we develop ThaiExam, a benchmark based on examinations for high-school students and investment professionals in Thailand. In addition, we fine-tune Typhoon to follow Thai instructions, and we evaluate instruction-tuned models on Thai instruction datasets as well as translation, summarization, and question-answering tasks. Experimental results on a suite of Thai benchmarks show that Typhoon outperforms all open-source Thai language models, and its performance is on par with GPT-3.5 in Thai while having only 7 billion parameters and being 2.62 times more efficient in tokenizing Thai text."
                },
                {
                    "title": "TinyGSM: achieving >80% on GSM8k with small language models",
                    "abstract": "Small-scale models offer various computational advantages, and yet to which extent size is critical for problem-solving abilities remains an open question. Specifically for solving grade school math, the smallest model size so far required to break the 80\\% barrier on the GSM8K benchmark remains to be 34B. Our work studies how high-quality datasets may be the key for small language models to acquire mathematical reasoning. We introduce \\texttt{TinyGSM}, a synthetic dataset of 12.3M grade school math problems paired with Python solutions, generated fully by GPT-3.5. After finetuning on \\texttt{TinyGSM}, we find that a duo of a 1.3B generation model and a 1.3B verifier model can achieve 81.5\\% accuracy, outperforming existing models that are orders of magnitude larger. This also rivals the performance of the GPT-3.5 ``teacher'' model (77.4\\%), from which our model's training data is generated. Our approach is simple and has two key components: 1) the high-quality dataset \\texttt{TinyGSM}, 2) the use of a verifier, which selects the final outputs from multiple candidate generations."
                },
                {
                    "title": "Harmonizing Global Voices: Culturally-Aware Models for Enhanced Content Moderation",
                    "abstract": "Content moderation at scale faces the challenge of considering local cultural distinctions when assessing content. While global policies aim to maintain decision-making consistency and prevent arbitrary rule enforcement, they often overlook regional variations in interpreting natural language as expressed in content. In this study, we are looking into how moderation systems can tackle this issue by adapting to local comprehension nuances. We train large language models on extensive datasets of media news and articles to create culturally attuned models. The latter aim to capture the nuances of communication across geographies with the goal of recognizing cultural and societal variations in what is considered offensive content. We further explore the capability of these models to generate explanations for instances of content violation, aiming to shed light on how policy guidelines are perceived when cultural and societal contexts change. We find that training on extensive media datasets successfully induced cultural awareness and resulted in improvements in handling content violations on a regional basis. Additionally, these advancements include the ability to provide explanations that align with the specific local norms and nuances as evidenced by the annotators' preference in our conducted study. This multifaceted success reinforces the critical role of an adaptable content moderation approach in keeping pace with the ever-evolving nature of the content it oversees."
                },
                {
                    "title": "SeaLLMs - Large Language Models for Southeast Asia",
                    "abstract": "Despite the remarkable achievements of large language models (LLMs) in various tasks, there remains a linguistic bias that favors high-resource languages, such as English, often at the expense of low-resource and regional languages. To address this imbalance, we introduce SeaLLMs, an innovative series of language models that specifically focuses on Southeast Asian (SEA) languages. SeaLLMs are built upon the Llama-2 model and further advanced through continued pre-training with an extended vocabulary, specialized instruction and alignment tuning to better capture the intricacies of regional languages. This allows them to respect and reflect local cultural norms, customs, stylistic preferences, and legal considerations. Our comprehensive evaluation demonstrates that SeaLLM-13b models exhibit superior performance across a wide spectrum of linguistic tasks and assistant-style instruction-following capabilities relative to comparable open-source models. Moreover, they outperform ChatGPT-3.5 in non-Latin languages, such as Thai, Khmer, Lao, and Burmese, by large margins while remaining lightweight and cost-effective to operate."
                },
                {
                    "title": "Taiwan LLM: Bridging the Linguistic Divide with a Culturally Aligned Language Model",
                    "abstract": "In the realm of language models, the nuanced linguistic and cultural intricacies of Traditional Chinese, as spoken in Taiwan, have been largely overlooked. This paper introduces Taiwan LLM, a pioneering Large Language Model that specifically caters to the Traditional Chinese language, with a focus on the variant used in Taiwan. Leveraging a comprehensive pretraining corpus and instruction-finetuning datasets, we have developed a model that not only understands the complexities of Traditional Chinese but also embodies the cultural context of Taiwan. Taiwan LLM represents the first of its kind, a model that is not only linguistically accurate but also culturally resonant with its user base. Our evaluations demonstrate that Taiwan LLM achieves superior performance in understanding and generating Traditional Chinese text, outperforming existing models that are predominantly trained on Simplified Chinese or English. The open-source release of Taiwan LLM invites collaboration and further innovation, ensuring that the linguistic diversity of Chinese speakers is embraced and well-served. The model, datasets, and further resources are made publicly available to foster ongoing research and development in this field."
                },
                {
                    "title": "AI Alignment: A Comprehensive Survey",
                    "abstract": "AI alignment aims to make AI systems behave in line with human intentions and values. As AI systems grow more capable, so do risks from misalignment. To provide a comprehensive and up-to-date overview of the alignment field, in this survey, we delve into the core concepts, methodology, and practice of alignment. First, we identify four principles as the key objectives of AI alignment: Robustness, Interpretability, Controllability, and Ethicality (RICE). Guided by these four principles, we outline the landscape of current alignment research and decompose them into two key components: forward alignment and backward alignment. The former aims to make AI systems aligned via alignment training, while the latter aims to gain evidence about the systems' alignment and govern them appropriately to avoid exacerbating misalignment risks. On forward alignment, we discuss techniques for learning from feedback and learning under distribution shift. On backward alignment, we discuss assurance techniques and governance practices. We also release and continually update the website (www.alignmentsurvey.com) which features tutorials, collections of papers, blog posts, and other resources."
                },
                {
                    "title": "Let's Synthesize Step by Step: Iterative Dataset Synthesis with Large Language Models by Extrapolating Errors from Small Models",
                    "abstract": "*Data Synthesis* is a promising way to train a small model with very little labeled data. One approach for data synthesis is to leverage the rich knowledge from large language models to synthesize pseudo training examples for small models, making it possible to achieve both data and compute efficiency at the same time. However, a key challenge in data synthesis is that the synthesized dataset often suffers from a large distributional discrepancy from the *real task* data distribution. Thus, in this paper, we propose *Synthesis Step by Step* (**S3**), a data synthesis framework that shrinks this distribution gap by iteratively extrapolating the errors made by a small model trained on the synthesized dataset on a small real-world validation dataset using a large language model. Extensive experiments on multiple NLP tasks show that our approach improves the performance of a small model by reducing the gap between the synthetic dataset and the real data, resulting in significant improvement compared to several baselines: 9.48% improvement compared to ZeroGen and 2.73% compared to GoldGen, and at most 15.17% improvement compared to the small model trained on human-annotated data."
                },
                {
                    "title": "Better to Ask in English: Cross-Lingual Evaluation of Large Language Models for Healthcare Queries",
                    "abstract": "Large language models (LLMs) are transforming the ways the general public accesses and consumes information. Their influence is particularly pronounced in pivotal sectors like healthcare, where lay individuals are increasingly appropriating LLMs as conversational agents for everyday queries. While LLMs demonstrate impressive language understanding and generation proficiencies, concerns regarding their safety remain paramount in these high-stake domains. Moreover, the development of LLMs is disproportionately focused on English. It remains unclear how these LLMs perform in the context of non-English languages, a gap that is critical for ensuring equity in the real-world use of these systems. This paper provides a framework to investigate the effectiveness of LLMs as multi-lingual dialogue systems for healthcare queries. Our empirically-derived framework XlingEval focuses on three fundamental criteria for evaluating LLM responses to naturalistic human-authored health-related questions: correctness, consistency, and verifiability. Through extensive experiments on four major global languages, including English, Spanish, Chinese, and Hindi, spanning three expert-annotated large health Q&A datasets, and through an amalgamation of algorithmic and human-evaluation strategies, we found a pronounced disparity in LLM responses across these languages, indicating a need for enhanced cross-lingual capabilities. We further propose XLingHealth, a cross-lingual benchmark for examining the multilingual capabilities of LLMs in the healthcare context. Our findings underscore the pressing need to bolster the cross-lingual capacities of these models, and to provide an equitable information ecosystem accessible to all."
                },
                {
                    "title": "Not All Countries Celebrate Thanksgiving: On the Cultural Dominance in Large Language Models",
                    "abstract": "This paper identifies a cultural dominance issue within large language models (LLMs) due to the predominant use of English data in model training (e.g., ChatGPT). LLMs often provide inappropriate English-culture-related answers that are not relevant to the expected culture when users ask in non-English languages. To systematically evaluate the cultural dominance issue, we build a benchmark of concrete (e.g., holidays and songs) and abstract (e.g., values and opinions) cultural objects. Empirical results show that the representative GPT models suffer from the culture dominance problem, where GPT-4 is the most affected while text-davinci-003 suffers the least from this problem. Our study emphasizes the need to critically examine cultural dominance and ethical consideration in their development and deployment. We show that two straightforward methods in model development (i.e., pretraining on more diverse data) and deployment (e.g., culture-aware prompting) can significantly mitigate the cultural dominance issue in LLMs."
                },
                {
                    "title": "Large language models can replicate cross-cultural differences in personality",
                    "abstract": "We use a large-scale experiment (N=8000) to determine whether GPT-4 can replicate cross-cultural differences in the Big Five, measured using the Ten-Item Personality Inventory. We used the US and South Korea as the cultural pair, given that prior research suggests substantial personality differences between people from these two countries. We manipulated the target of the simulation (US vs. Korean), the language of the inventory (English vs. Korean), and the language model (GPT-4 vs. GPT-3.5). Our results show that GPT-4 replicated the cross-cultural differences for each factor. However, mean ratings had an upward bias and exhibited lower variation than in the human samples, as well as lower structural validity. We provide preliminary evidence that LLMs can aid cross-cultural researchers and practitioners."
                },
                {
                    "title": "Ethical Reasoning over Moral Alignment: A Case and Framework for In-Context Ethical Policies in LLMs",
                    "abstract": "In this position paper, we argue that instead of morally aligning LLMs to specific set of ethical principles, we should infuse generic ethical reasoning capabilities into them so that they can handle value pluralism at a global scale. When provided with an ethical policy, an LLM should be capable of making decisions that are ethically consistent to the policy. We develop a framework that integrates moral dilemmas with moral principles pertaining to different foramlisms of normative ethics, and at different levels of abstractions. Initial experiments with GPT-x models shows that while GPT-4 is a nearly perfect ethical reasoner, the models still have bias towards the moral values of Western and English speaking societies."
                },
                {
                    "title": "Large Language Model Alignment: A Survey",
                    "abstract": "Recent years have witnessed remarkable progress made in large language models (LLMs). Such advancements, while garnering significant attention, have concurrently elicited various concerns. The potential of these models is undeniably vast; however, they may yield texts that are imprecise, misleading, or even detrimental. Consequently, it becomes paramount to employ alignment techniques to ensure these models to exhibit behaviors consistent with human values. This survey endeavors to furnish an extensive exploration of alignment methodologies designed for LLMs, in conjunction with the extant capability research in this domain. Adopting the lens of AI alignment, we categorize the prevailing methods and emergent proposals for the alignment of LLMs into outer and inner alignment. We also probe into salient issues including the models' interpretability, and potential vulnerabilities to adversarial attacks. To assess LLM alignment, we present a wide variety of benchmarks and evaluation methodologies. After discussing the state of alignment research for LLMs, we finally cast a vision toward the future, contemplating the promising avenues of research that lie ahead. Our aspiration for this survey extends beyond merely spurring research interests in this realm. We also envision bridging the gap between the AI alignment research community and the researchers engrossed in the capability exploration of LLMs for both capable and safe LLMs."
                },
                {
                    "title": "MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Models",
                    "abstract": "Large language models (LLMs) have pushed the limits of natural language understanding and exhibited excellent problem-solving ability. Despite the great success, most existing open-source LLMs (e.g., LLaMA-2) are still far away from satisfactory for solving mathematical problem due to the complex reasoning procedures. To bridge this gap, we propose MetaMath, a fine-tuned language model that specializes in mathematical reasoning. Specifically, we start by bootstrapping mathematical questions by rewriting the question from multiple perspectives without extra knowledge, which results in a new dataset called MetaMathQA. Then we fine-tune the LLaMA-2 models on MetaMathQA. Experimental results on two popular benchmarks (i.e., GSM8K and MATH) for mathematical reasoning demonstrate that MetaMath outperforms a suite of open-source LLMs by a significant margin. Our MetaMath-7B model achieves 66.4% on GSM8K and 19.4% on MATH, exceeding the state-of-the-art models of the same size by 11.5% and 8.7%. Particularly, MetaMath-70B achieves an accuracy of 82.3% on GSM8K, slightly better than GPT-3.5-Turbo. We release all the MetaMathQA dataset, the MetaMath models with different model sizes and the training code for public use."
                },
                {
                    "title": "Are Multilingual LLMs Culturally-Diverse Reasoners? An Investigation into Multicultural Proverbs and Sayings",
                    "abstract": "Large language models (LLMs) are highly adept at question answering and reasoning tasks, but when reasoning in a situational context, human expectations vary depending on the relevant cultural common ground. As languages are associated with diverse cultures, LLMs should also be culturally-diverse reasoners. In this paper, we study the ability of a wide range of state-of-the-art multilingual LLMs (mLLMs) to reason with proverbs and sayings in a conversational context. Our experiments reveal that: (1) mLLMs \u201cknow\u201d limited proverbs and memorizing proverbs does not mean understanding them within a conversational context; (2) mLLMs struggle to reason with figurative proverbs and sayings, and when asked to select the wrong answer (instead of asking it to select the correct answer); and (3) there is a \u201cculture gap\u201d in mLLMs when reasoning about proverbs and sayings translated from other languages. We construct and release our evaluation dataset MAPS (MulticulturAl Proverbs and Sayings) for proverb understanding with conversational context for six different languages."
                },
                {
                    "title": "Making Large Language Models Better Reasoners with Alignment",
                    "abstract": "Reasoning is a cognitive process of using evidence to reach a sound conclusion. The reasoning capability is essential for large language models (LLMs) to serve as the brain of the artificial general intelligence agent. Recent studies reveal that fine-tuning LLMs on data with the chain of thought (COT) reasoning process can significantly enhance their reasoning capabilities. However, we find that the fine-tuned LLMs suffer from an \\textit{Assessment Misalignment} problem, i.e., they frequently assign higher scores to subpar COTs, leading to potential limitations in their reasoning abilities. To address this problem, we introduce an \\textit{Alignment Fine-Tuning (AFT)} paradigm, which involves three steps: 1) fine-tuning LLMs with COT training data; 2) generating multiple COT responses for each question, and categorizing them into positive and negative ones based on whether they achieve the correct answer; 3) calibrating the scores of positive and negative responses given by LLMs with a novel constraint alignment loss. Specifically, the constraint alignment loss has two objectives: a) Alignment, which guarantees that positive scores surpass negative scores to encourage answers with high-quality COTs; b) Constraint, which keeps the negative scores confined to a reasonable range to prevent the model degradation. Beyond just the binary positive and negative feedback, the constraint alignment loss can be seamlessly adapted to the ranking situations when ranking feedback is accessible. Furthermore, we also delve deeply into recent ranking-based alignment methods, such as DPO, RRHF, and PRO, and discover that the constraint, which has been overlooked by these approaches, is also crucial for their performance. Extensive experiments on four reasoning benchmarks with both binary and ranking feedback demonstrate the effectiveness of AFT."
                },
                {
                    "title": "Cultural Alignment in Large Language Models: An Explanatory Analysis Based on Hofstede's Cultural Dimensions",
                    "abstract": "The deployment of large language models (LLMs) raises concerns regarding their cultural misalignment and potential ramifications on individuals and societies with diverse cultural backgrounds. While the discourse has focused mainly on political and social biases, our research proposes a Cultural Alignment Test (Hoftede's CAT) to quantify cultural alignment using Hofstede's cultural dimension framework, which offers an explanatory cross-cultural comparison through the latent variable analysis. We apply our approach to quantitatively evaluate LLMs, namely Llama 2, GPT-3.5, and GPT-4, against the cultural dimensions of regions like the United States, China, and Arab countries, using different prompting styles and exploring the effects of language-specific fine-tuning on the models' behavioural tendencies and cultural values. Our results quantify the cultural alignment of LLMs and reveal the difference between LLMs in explanatory cultural dimensions. Our study demonstrates that while all LLMs struggle to grasp cultural values, GPT-4 shows a unique capability to adapt to cultural nuances, particularly in Chinese settings. However, it faces challenges with American and Arab cultures. The research also highlights that fine-tuning LLama 2 models with different languages changes their responses to cultural questions, emphasizing the need for culturally diverse development in AI for worldwide acceptance and ethical use. For more details or to contribute to this research, visit our GitHub page https://github.com/reemim/Hofstedes_CAT/"
                },
                {
                    "title": "From Instructions to Intrinsic Human Values - A Survey of Alignment Goals for Big Models",
                    "abstract": "Big models, exemplified by Large Language Models (LLMs), are models typically pre-trained on massive data and comprised of enormous parameters, which not only obtain significantly improved performance across diverse tasks but also present emergent capabilities absent in smaller models. However, the growing intertwining of big models with everyday human lives poses potential risks and might cause serious social harm. Therefore, many efforts have been made to align LLMs with humans to make them better follow user instructions and satisfy human preferences. Nevertheless, `what to align with' has not been fully discussed, and inappropriate alignment goals might even backfire. In this paper, we conduct a comprehensive survey of different alignment goals in existing work and trace their evolution paths to help identify the most essential goal. Particularly, we investigate related works from two perspectives: the definition of alignment goals and alignment evaluation. Our analysis encompasses three distinct levels of alignment goals and reveals a goal transformation from fundamental abilities to value orientation, indicating the potential of intrinsic human values as the alignment goal for enhanced LLMs. Based on such results, we further discuss the challenges of achieving such intrinsic value alignment and provide a collection of available resources for future research on the alignment of big models."
                },
                {
                    "title": "Self-Alignment with Instruction Backtranslation",
                    "abstract": "We present a scalable method to build a high quality instruction following language model by automatically labelling human-written text with corresponding instructions. Our approach, named instruction backtranslation, starts with a language model finetuned on a small amount of seed data, and a given web corpus. The seed model is used to construct training examples by generating instruction prompts for web documents (self-augmentation), and then selecting high quality examples from among these candidates (self-curation). This data is then used to finetune a stronger model. Finetuning LLaMa on two iterations of our approach yields a model that outperforms all other LLaMa-based models on the Alpaca leaderboard not relying on distillation data, demonstrating highly effective self-alignment."
                },
                {
                    "title": "CValues: Measuring the Values of Chinese Large Language Models from Safety to Responsibility",
                    "abstract": "With the rapid evolution of large language models (LLMs), there is a growing concern that they may pose risks or have negative social impacts. Therefore, evaluation of human values alignment is becoming increasingly important. Previous work mainly focuses on assessing the performance of LLMs on certain knowledge and reasoning abilities, while neglecting the alignment to human values, especially in a Chinese context. In this paper, we present CValues, the first Chinese human values evaluation benchmark to measure the alignment ability of LLMs in terms of both safety and responsibility criteria. As a result, we have manually collected adversarial safety prompts across 10 scenarios and induced responsibility prompts from 8 domains by professional experts. To provide a comprehensive values evaluation of Chinese LLMs, we not only conduct human evaluation for reliable comparison, but also construct multi-choice prompts for automatic evaluation. Our findings suggest that while most Chinese LLMs perform well in terms of safety, there is considerable room for improvement in terms of responsibility. Moreover, both the automatic and human evaluation are important for assessing the human values alignment in different aspects. The benchmark and code is available on ModelScope and Github."
                },
                {
                    "title": "Large Language Models as Superpositions of Cultural Perspectives",
                    "abstract": "Large Language Models (LLMs) are often misleadingly recognized as having a personality or a set of values. We argue that an LLM can be seen as a superposition of perspectives with different values and personality traits. LLMs exhibit context-dependent values and personality traits that change based on the induced perspective (as opposed to humans, who tend to have more coherent values and personality traits across contexts). We introduce the concept of perspective controllability, which refers to a model's affordance to adopt various perspectives with differing values and personality traits. In our experiments, we use questionnaires from psychology (PVQ, VSM, IPIP) to study how exhibited values and personality traits change based on different perspectives. Through qualitative experiments, we show that LLMs express different values when those are (implicitly or explicitly) implied in the prompt, and that LLMs express different values even when those are not obviously implied (demonstrating their context-dependent nature). We then conduct quantitative experiments to study the controllability of different models (GPT-4, GPT-3.5, OpenAssistant, StableVicuna, StableLM), the effectiveness of various methods for inducing perspectives, and the smoothness of the models' drivability. We conclude by examining the broader implications of our work and outline a variety of associated scientific questions. The project website is available at https://sites.google.com/view/llm-superpositions ."
                },
                {
                    "title": "Having Beer after Prayer? Measuring Cultural Bias in Large Language Models",
                    "abstract": "As the reach of large language models (LMs) expands globally, their ability to cater to diverse cultural contexts becomes crucial. Despite advancements in multilingual capabilities, models are not designed with appropriate cultural nuances. In this paper, we show that multilingual and Arabic monolingual LMs exhibit bias towards entities associated with Western culture. We introduce CAMeL, a novel resource of 628 naturally-occurring prompts and 20,368 entities spanning eight types that contrast Arab and Western cultures. CAMeL provides a foundation for measuring cultural biases in LMs through both extrinsic and intrinsic evaluations. Using CAMeL, we examine the cross-cultural performance in Arabic of 16 different LMs on tasks such as story generation, NER, and sentiment analysis, where we find concerning cases of stereotyping and cultural unfairness. We further test their text-infilling performance, revealing the incapability of appropriate adaptation to Arab cultural contexts. Finally, we analyze 6 Arabic pre-training corpora and find that commonly used sources such as Wikipedia may not be best suited to build culturally aware LMs, if used as they are without adjustment. We will make CAMeL publicly available at: https://github.com/tareknaous/camel"
                },
                {
                    "title": "Assessing Cross-Cultural Alignment between ChatGPT and Human Societies: An Empirical Study",
                    "abstract": "The recent release of ChatGPT has garnered widespread recognition for its exceptional ability to generate human-like conversations. Given its usage by users from various nations and its training on a vast multilingual corpus that includes diverse cultural and societal norms, it is crucial to evaluate its effectiveness in cultural adaptation. In this paper, we investigate the underlying cultural background of ChatGPT by analyzing its responses to questions designed to quantify human cultural differences. Our findings suggest that, when prompted with American context, ChatGPT exhibits a strong alignment with American culture, but it adapts less effectively to other cultural contexts. Furthermore, by using different prompts to probe the model, we show that English prompts reduce the variance in model responses, flattening out cultural differences and biasing them towards American culture. This study provides valuable insights into the cultural implications of ChatGPT and highlights the necessity of greater diversity and cultural awareness in language technologies."
                },
                {
                    "title": "MEGA: Multilingual Evaluation of Generative AI",
                    "abstract": "Generative AI models have shown impressive performance on many Natural Language Processing tasks such as language understanding, reasoning, and language generation. An important question being asked by the AI community today is about the capabilities and limits of these models, and it is clear that evaluating generative AI is very challenging. Most studies on generative LLMs have been restricted to English and it is unclear how capable these models are at understanding and generating text in other languages. We present the first comprehensive benchmarking of generative LLMs - MEGA, which evaluates models on standard NLP benchmarks, covering 16 NLP datasets across 70 typologically diverse languages. We compare the performance of generative LLMs including Chat-GPT and GPT-4 to State of the Art (SOTA) non-autoregressive models on these tasks to determine how well generative models perform compared to the previous generation of LLMs. We present a thorough analysis of the performance of models across languages and tasks and discuss challenges in improving the performance of generative LLMs on low-resource languages. We create a framework for evaluating generative LLMs in the multilingual setting and provide directions for future progress in the field."
                },
                {
                    "title": "Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them",
                    "abstract": "BIG-Bench (Srivastava et al., 2022) is a diverse evaluation suite that focuses on tasks believed to be beyond the capabilities of current language models. Language models have already made good progress on this benchmark, with the best model in the BIG-Bench paper outperforming average reported human-rater results on 65% of the BIG-Bench tasks via few-shot prompting. But on what tasks do language models fall short of average human-rater performance, and are those tasks actually unsolvable by current language models? In this work, we focus on a suite of 23 challenging BIG-Bench tasks which we call BIG-Bench Hard (BBH). These are the task for which prior language model evaluations did not outperform the average human-rater. We find that applying chain-of-thought (CoT) prompting to BBH tasks enables PaLM to surpass the average human-rater performance on 10 of the 23 tasks, and Codex (code-davinci-002) to surpass the average human-rater performance on 17 of the 23 tasks. Since many tasks in BBH require multi-step reasoning, few-shot prompting without CoT, as done in the BIG-Bench evaluations (Srivastava et al., 2022), substantially underestimates the best performance and capabilities of language models, which is better captured via CoT prompting. As further analysis, we explore the interaction between CoT and model scale on BBH, finding that CoT enables emergent task performance on several BBH tasks with otherwise flat scaling curves."
                },
                {
                    "title": "NormSAGE: Multi-Lingual Multi-Cultural Norm Discovery from Conversations On-the-Fly",
                    "abstract": "Norm discovery is important for understanding and reasoning about the acceptable behaviors and potential violations in human communication and interactions. We introduce NormSage, a framework for addressing the novel task of conversation-grounded multi-lingual, multi-cultural norm discovery, based on language model prompting and self-verification. NormSAGE leverages the expressiveness and implicit knowledge of the pretrained GPT-3 language model backbone, to elicit knowledge about norms through directed questions representing the norm discovery task and conversation context. It further addresses the risk of language model hallucination with a self-verification mechanism ensuring that the norms discovered are correct and are substantially grounded to their source conversations. Evaluation results show that our approach discovers significantly more relevant and insightful norms for conversations on-the-fly compared to baselines (>10+% in Likert scale rating). The norms discovered from Chinese conversation are also comparable to the norms discovered from English conversation in terms of insightfulness and correctness (<3% difference). In addition, the culture-specific norms are promising quality, allowing for 80% accuracy in culture pair human identification. Finally, our grounding process in norm discovery self-verification can be extended for instantiating the adherence and violation of any norm for a given conversation on-the-fly, with explainability and transparency. NormSAGE achieves an AUC of 95.4% in grounding, with natural language explanation matching human-written quality."
                },
                {
                    "title": "Extracting Cultural Commonsense Knowledge at Scale",
                    "abstract": "Structured knowledge is important for many AI applications. Commonsense knowledge, which is crucial for robust human-centric AI, is covered by a small number of structured knowledge projects. However, they lack knowledge about human traits and behaviors conditioned on socio-cultural contexts, which is crucial for situative AI. This paper presents Candle, an end-to-end methodology for extracting high-quality cultural commonsense knowledge (CCSK) at scale. Candle extracts CCSK assertions from a huge web corpus and organizes them into coherent clusters, for 3 domains of subjects (geography, religion, occupation) and several cultural facets (food, drinks, clothing, traditions, rituals, behaviors). Candle includes judicious techniques for classification-based filtering and scoring of interestingness. Experimental evaluations show the superiority of the Candle CCSK collection over prior works, and an extrinsic use case demonstrates the benefits of CCSK for the GPT-3 language model. Code and data can be accessed at https://candle.mpi-inf.mpg.de/."
                },
                {
                    "title": "K-MHaS: A Multi-label Hate Speech Detection Dataset in Korean Online News Comment",
                    "abstract": "Online hate speech detection has become an important issue due to the growth of online content, but resources in languages other than English are extremely limited. We introduce K-MHaS, a new multi-label dataset for hate speech detection that effectively handles Korean language patterns. The dataset consists of 109k utterances from news comments and provides a multi-label classification using 1 to 4 labels, and handles subjectivity and intersectionality. We evaluate strong baselines on K-MHaS. KR-BERT with a sub-character tokenizer outperforms others, recognizing decomposed characters in each hate speech class."
                },
                {
                    "title": "Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Models",
                    "abstract": "Hate speech detection models are typically evaluated on held-out test sets. However, this risks painting an incomplete and potentially misleading picture of model performance because of increasingly well-documented systematic gaps and biases in hate speech datasets. To enable more targeted diagnostic insights, recent research has thus introduced functional tests for hate speech detection models. However, these tests currently only exist for English-language content, which means that they cannot support the development of more effective models in other languages spoken by billions across the world. To help address this issue, we introduce Multilingual HateCheck (MHC), a suite of functional tests for multilingual hate speech detection models. MHC covers 34 functionalities across ten languages, which is more languages than any other hate speech dataset. To illustrate MHC\u2019s utility, we train and test a high-performing multilingual hate speech detection model, and reveal critical model weaknesses for monolingual and cross-lingual applications."
                },
                {
                    "title": "The Ghost in the Machine has an American accent: value conflict in GPT-3",
                    "abstract": "The alignment problem in the context of large language models must consider the plurality of human values in our world. Whilst there are many resonant and overlapping values amongst the world's cultures, there are also many conflicting, yet equally valid, values. It is important to observe which cultural values a model exhibits, particularly when there is a value conflict between input prompts and generated outputs. We discuss how the co-creation of language and cultural value impacts large language models (LLMs). We explore the constitution of the training data for GPT-3 and compare that to the world's language and internet access demographics, as well as to reported statistical profiles of dominant values in some Nation-states. We stress tested GPT-3 with a range of value-rich texts representing several languages and nations; including some with values orthogonal to dominant US public opinion as reported by the World Values Survey. We observed when values embedded in the input text were mutated in the generated outputs and noted when these conflicting values were more aligned with reported dominant US values. Our discussion of these results uses a moral value pluralism (MVP) lens to better understand these value mutations. Finally, we provide recommendations for how our work may contribute to other current work in the field."
                },
                {
                    "title": "Submodularity In Machine Learning and Artificial Intelligence",
                    "abstract": "In this manuscript, we offer a gentle review of submodularity and supermodularity and their properties. We offer a plethora of submodular definitions; a full description of a number of example submodular functions and their generalizations; example discrete constraints; a discussion of basic algorithms for maximization, minimization, and other operations; a brief overview of continuous submodular extensions; and some historical applications. We then turn to how submodularity is useful in machine learning and artificial intelligence. This includes summarization, and we offer a complete account of the differences between and commonalities amongst sketching, coresets, extractive and abstractive summarization in NLP, data distillation and condensation, and data subset selection and feature selection. We discuss a variety of ways to produce a submodular function useful for machine learning, including heuristic hand-crafting, learning or approximately learning a submodular function or aspects thereof, and some advantages of the use of a submodular function as a coreset producer. We discuss submodular combinatorial information functions, and how submodularity is useful for clustering, data partitioning, parallel machine learning, active and semi-supervised learning, probabilistic modeling, and structured norms and loss functions.1"
                },
                {
                    "title": "AI-UPV at IberLEF-2021 DETOXIS task: Toxicity Detection in Immigration-Related Web News Comments Using Transformers and Statistical Models",
                    "abstract": "This paper describes our participation in the DEtection of TOXicity in comments In Spanish (DETOXIS) shared task 2021 at the 3rd Workshop on Iberian Languages Evaluation Forum. The shared task is divided into two related classification tasks: (i) Task 1: toxicity detection and; (ii) Task 2: toxicity level detection. They focus on the xenophobic problem exacerbated by the spread of toxic comments posted in different online news articles related to immigration. One of the necessary efforts towards mitigating this problem is to detect toxicity in the comments. Our main objective was to implement an accurate model to detect xenophobia in comments about web news articles within the DETOXIS shared task 2021, based on the competition's official metrics: the F1-score for Task 1 and the Closeness Evaluation Metric (CEM) for Task 2. To solve the tasks, we worked with two types of machine learning models: (i) statistical models and (ii) Deep Bidirectional Transformers for Language Understanding (BERT) models. We obtained our best results in both tasks using BETO, an BERT model trained on a big Spanish corpus. We obtained the 3rd place in Task 1 official ranking with the F1-score of 0.5996, and we achieved the 6th place in Task 2 official ranking with the CEM of 0.7142. Our results suggest: (i) BERT models obtain better results than statistical models for toxicity detection in text comments; (ii) Monolingual BERT models have an advantage over multilingual BERT models in toxicity detection in text comments in their pre-trained language."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "HateBR: A Large Expert Annotated Corpus of Brazilian Instagram Comments for Offensive Language and Hate Speech Detection",
                    "abstract": "Due to the severity of the social media offensive and hateful comments in Brazil, and the lack of research in Portuguese, this paper provides the first large-scale expert annotated corpus of Brazilian Instagram comments for hate speech and offensive language detection. The HateBR corpus was collected from the comment section of Brazilian politicians\u2019 accounts on Instagram and manually annotated by specialists, reaching a high inter-annotator agreement. The corpus consists of 7,000 documents annotated according to three different layers: a binary classification (offensive versus non-offensive comments), offensiveness-level classification (highly, moderately, and slightly offensive), and nine hate speech groups (xenophobia, racism, homophobia, sexism, religious intolerance, partyism, apology for the dictatorship, antisemitism, and fatphobia). We also implemented baseline experiments for offensive language and hate speech detection and compared them with a literature baseline. Results show that the baseline experiments on our corpus outperform the current state-of-the-art for the Portuguese language."
                },
                {
                    "title": "The Pile: An 800GB Dataset of Diverse Text for Language Modeling",
                    "abstract": "Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present the Pile : an 825 GiB English text corpus tar-geted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets\u2014both existing and newly constructed\u2014many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve signi\ufb01cantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction. 1"
                },
                {
                    "title": "Toxic Language Detection in Social Media for Brazilian Portuguese: New Dataset and Multilingual Analysis",
                    "abstract": "Hate speech and toxic comments are a common concern of social media platform users. Although these comments are, fortunately, the minority in these platforms, they are still capable of causing harm. Therefore, identifying these comments is an important task for studying and preventing the proliferation of toxicity in social media. Previous work in automatically detecting toxic comments focus mainly in English, with very few work in languages like Brazilian Portuguese. In this paper, we propose a new large-scale dataset for Brazilian Portuguese with tweets annotated as either toxic or non-toxic or in different types of toxicity. We present our dataset collection and annotation process, where we aimed to select candidates covering multiple demographic groups. State-of-the-art BERT models were able to achieve 76% macro-F1 score using monolingual data in the binary case. We also show that large-scale monolingual data is still needed to create more accurate models, despite recent advances in multilingual approaches. An error analysis and experiments with multi-label classification show the difficulty of classifying certain types of toxic comments that appear less frequently in our data and highlights the need to develop models that are aware of different categories of toxicity."
                },
                {
                    "title": "SemEval-2020 Task 12: Multilingual Offensive Language Identification in Social Media (OffensEval 2020)",
                    "abstract": "We present the results and the main findings of SemEval-2020 Task 12 on Multilingual Offensive Language Identification in Social Media (OffensEval-2020). The task included three subtasks corresponding to the hierarchical taxonomy of the OLID schema from OffensEval-2019, and it was offered in five languages: Arabic, Danish, English, Greek, and Turkish. OffensEval-2020 was one of the most popular tasks at SemEval-2020, attracting a large number of participants across all subtasks and languages: a total of 528 teams signed up to participate in the task, 145 teams submitted official runs on the test data, and 70 teams submitted system description papers."
                },
                {
                    "title": "BEEP! Korean Corpus of Online News Comments for Toxic Speech Detection",
                    "abstract": "Toxic comments in online platforms are an unavoidable social issue under the cloak of anonymity. Hate speech detection has been actively done for languages such as English, German, or Italian, where manually labeled corpus has been released. In this work, we first present 9.4K manually labeled entertainment news comments for identifying Korean toxic speech, collected from a widely used online news platform in Korea. The comments are annotated regarding social bias and hate speech since both aspects are correlated. The inter-annotator agreement Krippendorff\u2019s alpha score is 0.492 and 0.496, respectively. We provide benchmarks using CharCNN, BiLSTM, and BERT, where BERT achieves the highest score on all tasks. The models generally display better performance on bias identification, since the hate speech detection is a more subjective issue. Additionally, when BERT is trained with bias label for hate speech detection, the prediction score increases, implying that bias and hate are intertwined. We make our dataset publicly available and open competitions with the corpus and benchmarks."
                },
                {
                    "title": "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks",
                    "abstract": "Large pre-trained language models have been shown to store factual knowledge in their parameters, and achieve state-of-the-art results when fine-tuned on downstream NLP tasks. However, their ability to access and precisely manipulate knowledge is still limited, and hence on knowledge-intensive tasks, their performance lags behind task-specific architectures. Additionally, providing provenance for their decisions and updating their world knowledge remain open research problems. Pre-trained models with a differentiable access mechanism to explicit non-parametric memory can overcome this issue, but have so far been only investigated for extractive downstream tasks. We explore a general-purpose fine-tuning recipe for retrieval-augmented generation (RAG) -- models which combine pre-trained parametric and non-parametric memory for language generation. We introduce RAG models where the parametric memory is a pre-trained seq2seq model and the non-parametric memory is a dense vector index of Wikipedia, accessed with a pre-trained neural retriever. We compare two RAG formulations, one which conditions on the same retrieved passages across the whole generated sequence, the other can use different passages per token. We fine-tune and evaluate our models on a wide range of knowledge-intensive NLP tasks and set the state-of-the-art on three open domain QA tasks, outperforming parametric seq2seq models and task-specific retrieve-and-extract architectures. For language generation tasks, we find that RAG models generate more specific, diverse and factual language than a state-of-the-art parametric-only seq2seq baseline."
                },
                {
                    "title": "I Feel Offended, Don\u2019t Be Abusive! Implicit/Explicit Messages in Offensive and Abusive Language",
                    "abstract": "Abusive language detection is an unsolved and challenging problem for the NLP community. Recent literature suggests various approaches to distinguish between different language phenomena (e.g., hate speech vs. cyberbullying vs. offensive language) and factors (degree of explicitness and target) that may help to classify different abusive language phenomena. There are data sets that annotate the target of abusive messages (i.e.OLID/OffensEval (Zampieri et al., 2019a)). However, there is a lack of data sets that take into account the degree of explicitness. In this paper, we propose annotation guidelines to distinguish between explicit and implicit abuse in English and apply them to OLID/OffensEval. The outcome is a newly created resource, AbuseEval v1.0, which aims to address some of the existing issues in the annotation of offensive and abusive language (e.g., explicitness of the message, presence of a target, need of context, and interaction across different phenomena)."
                },
                {
                    "title": "A Multi-Platform Arabic News Comment Dataset for Offensive Language Detection",
                    "abstract": "Access to social media often enables users to engage in conversation with limited accountability. This allows a user to share their opinions and ideology, especially regarding public content, occasionally adopting offensive language. This may encourage hate crimes or cause mental harm to targeted individuals or groups. Hence, it is important to detect offensive comments in social media platforms. Typically, most studies focus on offensive commenting in one platform only, even though the problem of offensive language is observed across multiple platforms. Therefore, in this paper, we introduce and make publicly available a new dialectal Arabic news comment dataset, collected from multiple social media platforms, including Twitter, Facebook, and YouTube. We follow two-step crowd-annotator selection criteria for low-representative language annotation task in a crowdsourcing platform. Furthermore, we analyze the distinctive lexical content along with the use of emojis in offensive comments. We train and evaluate the classifiers using the annotated multi-platform dataset along with other publicly available data. Our results highlight the importance of multiple platform dataset for (a) cross-platform, (b) cross-domain, and (c) cross-dialect generalization of classifier performance."
                },
                {
                    "title": "A Corpus of Turkish Offensive Language on Social Media",
                    "abstract": "This paper introduces a corpus of Turkish offensive language. To our knowledge, this is the first corpus of offensive language for Turkish. The corpus consists of randomly sampled micro-blog posts from Twitter. The annotation guidelines are based on a careful review of the annotation practices of recent efforts for other languages. The corpus contains 36 232 tweets sampled randomly from the Twitter stream during a period of 18 months between Apr 2018 to Sept 2019. We found approximately 19 % of the tweets in the data contain some type of offensive language, which is further subcategorized based on the target of the offense. We describe the annotation process, discuss some interesting aspects of the data, and present results of automatically classifying the corpus using state-of-the-art text classification methods. The classifiers achieve 77.3 % F1 score on identifying offensive tweets, 77.9 % F1 score on determining whether a given offensive document is targeted or not, and 53.0 % F1 score on classifying the targeted offensive documents into three subcategories."
                },
                {
                    "title": "OSACT4 Shared Task on Offensive Language Detection: Intensive Preprocessing-Based Approach",
                    "abstract": "The preprocessing phase is one of the key phases within the text classification pipeline. This study aims at investigating the impact of the preprocessing phase on text classification, specifically on offensive language and hate speech classification for Arabic text. The Arabic language used in social media is informal and written using Arabic dialects, which makes the text classification task very complex. Preprocessing helps in dimensionality reduction and removing useless content. We apply intensive preprocessing techniques to the dataset before processing it further and feeding it into the classification model. An intensive preprocessing-based approach demonstrates its significant impact on offensive language detection and hate speech detection shared tasks of the fourth workshop on Open-Source Arabic Corpora and Corpora Processing Tools (OSACT). Our team wins the third place (3rd) in the Sub-Task A Offensive Language Detection division and wins the first place (1st) in the Sub-Task B Hate Speech Detection division, with an F1 score of 89% and 95%, respectively, by providing the state-of-the-art performance in terms of F1, accuracy, recall, and precision for Arabic hate speech detection."
                },
                {
                    "title": "SOLID: A Large-Scale Semi-Supervised Dataset for Offensive Language Identification",
                    "abstract": "The use of offensive language is a major problem in social media which has led to an abundance of research in detecting content such as hate speech, cyberbulling, and cyber-aggression. There have been several attempts to consolidate and categorize these efforts. Recently, the OLID dataset used at SemEval-2019 proposed a hierarchical three-level annotation taxonomy which addresses different types of offensive language as well as important information such as the target of such content. The categorization provides meaningful and important information for understanding offensive language. However, the OLID dataset is limited in size, especially for some of the low-level categories, which included only a few hundred instances, thus making it challenging to train robust deep learning models. Here, we address this limitation by creating the largest available dataset for this task, SOLID. SOLID contains over nine million English tweets labeled in a semi-supervised manner. We further demonstrate experimentally that using SOLID along with OLID yields improved performance on the OLID test set for two different models, especially for the lower levels of the taxonomy. Finally, we perform analysis of the models' performance on easy and hard examples of offensive language using data annotated in a semi-supervised way."
                },
                {
                    "title": "Developing a Multilingual Annotated Corpus of Misogyny and Aggression",
                    "abstract": "In this paper, we discuss the development of a multilingual annotated corpus of misogyny and aggression in Indian English, Hindi, and Indian Bangla as part of a project on studying and automatically identifying misogyny and communalism on social media (the ComMA Project). The dataset is collected from comments on YouTube videos and currently contains a total of over 20,000 comments. The comments are annotated at two levels - aggression (overtly aggressive, covertly aggressive, and non-aggressive) and misogyny (gendered and non-gendered). We describe the process of data collection, the tagset used for annotation, and issues and challenges faced during the process of annotation. Finally, we discuss the results of the baseline experiments conducted to develop a classifier for misogyny in the three languages."
                },
                {
                    "title": "Detecting and Monitoring Hate Speech in Twitter",
                    "abstract": "Social Media are sensors in the real world that can be used to measure the pulse of societies. However, the massive and unfiltered feed of messages posted in social media is a phenomenon that nowadays raises social alarms, especially when these messages contain hate speech targeted to a specific individual or group. In this context, governments and non-governmental organizations (NGOs) are concerned about the possible negative impact that these messages can have on individuals or on the society. In this paper, we present HaterNet, an intelligent system currently being used by the Spanish National Office Against Hate Crimes of the Spanish State Secretariat for Security that identifies and monitors the evolution of hate speech in Twitter. The contributions of this research are many-fold: (1) It introduces the first intelligent system that monitors and visualizes, using social network analysis techniques, hate speech in Social Media. (2) It introduces a novel public dataset on hate speech in Spanish consisting of 6000 expert-labeled tweets. (3) It compares several classification approaches based on different document representation strategies and text classification models. (4) The best approach consists of a combination of a LTSM+MLP neural network that takes as input the tweet\u2019s word, emoji, and expression tokens\u2019 embeddings enriched by the tf-idf, and obtains an area under the curve (AUC) of 0.828 on our dataset, outperforming previous methods presented in the literature."
                },
                {
                    "title": "Detect Camouflaged Spam Content via StoneSkipping: Graph and Text Joint Embedding for Chinese Character Variation Representation",
                    "abstract": "The task of Chinese text spam detection is very challenging due to both glyph and phonetic variations of Chinese characters. This paper proposes a novel framework to jointly model Chinese variational, semantic, and contextualized representations for Chinese text spam detection task. In particular, a Variation Family-enhanced Graph Embedding (VFGE) algorithm is designed based on a Chinese character variation graph. The VFGE can learn both the graph embeddings of the Chinese characters (local) and the latent variation families (global). Furthermore, an enhanced bidirectional language model, with a combination gate function and an aggregation learning function, is proposed to integrate the graph and text information while capturing the sequential information. Extensive experiments have been conducted on both SMS and review datasets, to show the proposed method outperforms a series of state-of-the-art models for Chinese spam detection."
                },
                {
                    "title": "Multilingual and Multi-Aspect Hate Speech Analysis",
                    "abstract": "Current research on hate speech analysis is typically oriented towards monolingual and single classification tasks. In this paper, we present a new multilingual multi-aspect hate speech analysis dataset and use it to test the current state-of-the-art multilingual multitask learning approaches. We evaluate our dataset in various classification settings, then we discuss how to leverage our annotations in order to improve hate speech detection and classification in general."
                },
                {
                    "title": "SemEval-2019 Task 5: Multilingual Detection of Hate Speech Against Immigrants and Women in Twitter",
                    "abstract": "The paper describes the organization of the SemEval 2019 Task 5 about the detection of hate speech against immigrants and women in Spanish and English messages extracted from Twitter. The task is organized in two related classification subtasks: a main binary subtask for detecting the presence of hate speech, and a finer-grained one devoted to identifying further features in hateful contents such as the aggressive attitude and the target harassed, to distinguish if the incitement is against an individual rather than a group. HatEval has been one of the most popular tasks in SemEval-2019 with a total of 108 submitted runs for Subtask A and 70 runs for Subtask B, from a total of 74 different teams. Data provided for the task are described by showing how they have been collected and annotated. Moreover, the paper provides an analysis and discussion about the participant systems and the results they achieved in both subtasks."
                },
                {
                    "title": "Offensive Comments in the Brazilian Web: a dataset and baseline results",
                    "abstract": ". Brazilian Web users are among the most active in social networks and very keen on interacting with others. Offensive comments, known as hate speech , have been plaguing online media and originating a number of law-suits against companies which publish Web content. Given the massive number of user generated text published on a daily basis, manually \ufb01ltering offensive comments becomes infeasible. The identi\ufb01cation of offensive comments can be treated as a supervised classi\ufb01cation task. In order to obtain a model to classify comments, an annotated dataset containing positive and negative examples is necessary. The lack of such a dataset in Portuguese, limits the development of detection approaches for this language. In this paper, we describe how we created annotated datasets of offensive comments for Portuguese by collecting news comments on the Brazilian Web. In addition, we provide classi\ufb01cation re-sults achieved by standard classi\ufb01cation algorithms on these datasets which can serve as baseline for future work on this topic."
                },
                {
                    "title": "Automated Hate Speech Detection and the Problem of Offensive Language",
                    "abstract": "\n \n A key challenge for automatic hate-speech detection on social media is the separation of hate speech from other instances of offensive language. Lexical detection methods tend to have low precision because they classify all messages containing particular terms as hate speech and previous work using supervised learning has failed to distinguish between the two categories. We used a crowd-sourced hate speech lexicon to collect tweets containing hate speech keywords. We use crowd-sourcing to label a sample of these tweets into three categories: those containing hate speech, only offensive language, and those with neither. We train a multi-class classifier to distinguish between these different categories. Close analysis of the predictions and the errors shows when we can reliably separate hate speech from other offensive language and when this differentiation is more difficult. We find that racist and homophobic tweets are more likely to be classified as hate speech but that sexist tweets are generally classified as offensive. Tweets without explicit hate keywords are also more difficult to classify.\n \n"
                },
                {
                    "title": "Measuring the Reliability of Hate Speech Annotations: The Case of the European Refugee Crisis",
                    "abstract": "Some users of social media are spreading racist, sexist, and otherwise hateful content. For the purpose of training a hate speech detection system, the reliability of the annotations is crucial, but there is no \nuniversally agreed-upon definition. We collected potentially hateful messages and asked two groups of internet users to determine \nwhether they were hate speech or not, whether they should be banned or not and to rate their degree of offensiveness. One of the groups was shown a definition prior to completing the survey. We aimed to assess whether hate speech can be annotated reliably, and the extent to which existing definitions are in accordance with subjective ratings. Our results indicate that showing users a definition caused them to partially align their own opinion with the definition but did not improve reliability, which was very low overall. We conclude that the presence of hate speech should perhaps \nnot be considered a binary yes-or-no decision, and raters need more detailed instructions for the annotation."
                },
                {
                    "title": "Hateful Symbols or Hateful People? Predictive Features for Hate Speech Detection on Twitter",
                    "abstract": "Hate speech in the form of racist and sexist remarks are a common occurrence on social media. For that reason, many social media services address the problem of identifying hate speech, but the definition of hate speech varies markedly and is largely a manual effort (BBC, 2015; Lomas, 2015). We provide a list of criteria founded in critical race theory, and use them to annotate a publicly available corpus of more than 16k tweets. We analyze the impact of various extra-linguistic features in conjunction with character n-grams for hatespeech detection. We also present a dictionary based the most indicative words in our data."
                },
                {
                    "title": "World Values Survey",
                    "abstract": "The World Values Survey was developed to document and enabling an understanding of values across cultures and across time with waves of data collection from 1981 until the present. It has information from over 100 countries on all six continents. Two dimensions of cross-cultural differences are tapped by the WVS: Traditional/Secular-rational and Survival/Self-expression. Together, these two dimensions explain a large part of the difference in values between cultures. \n \n \nKeywords: \n \nvalues; \nvalue change; \ncross-cultural values"
                },
                {
                    "title": "Social Influence and the Collective Dynamics of Opinion Formation",
                    "abstract": "Social influence is the process by which individuals adapt their opinion, revise their beliefs, or change their behavior as a result of social interactions with other people. In our strongly interconnected society, social influence plays a prominent role in many self-organized phenomena such as herding in cultural markets, the spread of ideas and innovations, and the amplification of fears during epidemics. Yet, the mechanisms of opinion formation remain poorly understood, and existing physics-based models lack systematic empirical validation. Here, we report two controlled experiments showing how participants answering factual questions revise their initial judgments after being exposed to the opinion and confidence level of others. Based on the observation of 59 experimental subjects exposed to peer-opinion for 15 different items, we draw an influence map that describes the strength of peer influence during interactions. A simple process model derived from our observations demonstrates how opinions in a group of interacting people can converge or split over repeated interactions. In particular, we identify two major attractors of opinion: (i) the expert effect, induced by the presence of a highly confident individual in the group, and (ii) the majority effect, caused by the presence of a critical mass of laypeople sharing similar opinions. Additional simulations reveal the existence of a tipping point at which one attractor will dominate over the other, driving collective opinion in a given direction. These findings have implications for understanding the mechanisms of public opinion formation and managing conflicting situations in which self-confident and better informed minorities challenge the views of a large uninformed majority."
                },
                {
                    "title": "Book Review: International and Comparative Industrial Relations: Culture's Consequences: International Differences in Work-Related Values",
                    "abstract": "The field of comparative organizational research has, in the past, been characterized as lacking both conceptual and empirical cohesion. Differences across studies in methodology, definitions of key variables (especially culture), and research objectives have made generalization and theory building difficult. In Culture's Consequences, however, Geert Hofstede has taken a significant step forward in dealing with at least some of these limitations. The book reports results of a major study encompassing forty countries, using equivalent methodology and concentrating on consistent dimensions of cultural variation. Results of the study are bolstered by identification of similar patterns of findings in other studies focusing on comparable cultural dimensions. Hofstede begins by suggesting that the most pressing problem in comparative administrative research is the identification of dimensions of cultural differentiation that may affect organizational behavior. This task becomes the expressed aim of his book: to generate an operational definition of the culture variable. The first section of the book examines several of the more troublesome problems associated with measuring culture, ranging from ethnocentrism in the researchers to language difficulties in producing comparable measurement instruments. From this analysis and a review of literature dealing with the definition of culture, Hofstede develops a set of strategies for conducting \"comparative multisociety research.\" Using the strategy of searching for specific cultural dimensions, he then explains the methodological and analytical procedures used in the HERMES study, which he headed. The next section reports results on four dimensions of cultural variation derived from the study data: power distance, uncertainty avoidance, individualism, and masculinity. Results indicate that these dimensions are present in the value systems of managers in all forty countries and vary in predictable ways based on characteristics associated with their countries of origin. These country characteristics include geographic location, economic status, population demographics, and national political climate. Hofstede proposes that these environmental factors interact with collective value systems to provide a stable reinforcing cultural system. For each dimension, he also points out many similarities with comparable research findings from other studies. The final section uses these cultural dimensions as a basis for developing implications and drawing conclusions concerning the nature of culture and its probable impact both on organizations and on the behavior of individuals in them. Looking specifically at the issues of cultural boundaries, Hofstede uses findings from a wide range of studies to group countries into logically consistent clusters based on similarities on the four cultural dimensions. The HERMES study data also allow an analysis of cultural shifts over time. Results indicate no support for the theory that world cultures are converging. Rather, Hofstede finds a variety of shifts, some of which appear to be universal, but most of which appear to be a result of unique combinations of societal experiences and existing collective value systems. The closing chapter considers explicitly the practical implications of the study results for international or multinational managers. Hofstede concludes that organizations are culture bound and must be managed with reference to the evident cultural context. Specific influences of culture in such areas as motivational and leadership strategies, job design, and work administration are discussed. The book can be evaluated from three separate perspectives, for it is at once a research report, a reference work, and a commentary on the field. As a research effort, it has several advantages over much of the existing research deriving primarily from the size and scope of the study and from Hofstede's sensitivity to the problems inherent in conducting research across cultural boundaries. It does have some limitations, however, the most critical apparently the fact that the HERMES data set was generated from employees of a single multinational corporation located in forty different countries. The risk in this type of study is that selection or socialization may create artificial homogeneity in the data. This concern is at least partially relieved by the similarities found between the HERMES data and findings of other, unconnected studies. Another limitation is that implications drawn concerning the impact of culture on organization and individual behavior are generated from logic rather than hard evidence. A significant and worthwhile extention of this research would be the testing of these \"hypotheses\" concerning culture's influence. In spite of these limitations, however, the study does make several contributions in the areas of cultural definition and mapping and of theory building. For these contributions it must stand as one of the most significant comparative organizational studies to date. Sage Publications, Inc. is collaborating with JSTOR to digitize, preserve, and extend access to Industrial and Labor Relations Review"
                },
                {
                    "title": "Towards Identifying Social Bias in Dialog Systems: Frame, Datasets, and Benchmarks",
                    "abstract": "Warning: this paper contains content that may be offensive or upsetting. The research of open-domain dialog systems has been greatly prospered by neural models trained on large-scale corpora, however, such corpora often introduce various safety problems (e.g., offensive languages, biases, and toxic behaviors) that signi\ufb01cantly hinder the deployment of dialog systems in practice. Among all these unsafe issues, addressing social bias is more complex as its negative impact on marginalized populations is usually expressed implicitly, thus requiring normative reasoning and rigorous analysis. In this paper, we focus our investigation on social bias detection of dialog safety problems. We \ufb01rst propose a novel D IAL - B IAS F RAME for analyzing the social bias in conversations pragmatically, which considers more comprehensive bias-related analyses rather than simple dichotomy annotations. Based on the proposed framework, we further introduce CD AIL -B IAS D ATASET that, to our knowledge, is the \ufb01rst well-annotated Chinese social bias dialog dataset. In addition, we establish several dialog bias detection benchmarks at different label granularities and input types (utterance-level and context-level). We show that the proposed in-depth analyses together with these benchmarks in our D IAL - B IAS F RAME are necessary and essential to bias detection tasks and can bene\ufb01t building safe dialog systems in practice."
                },
                {
                    "title": "Overview of OSACT5 Shared Task on Arabic Offensive Language and Hate Speech Detection",
                    "abstract": "This paper provides an overview of the shard task on detecting offensive language, hate speech, and fine-grained hate speech at the fifth workshop on Open-Source Arabic Corpora and Processing Tools (OSACT5). The shared task comprised of three subtasks; Subtask A, involving the detection of offensive language, which contains socially unacceptable or impolite content including any kind of explicit or implicit insults or attacks against individuals or groups; Subtask B, involving the detection of hate speech, which contains offensive language targeting individuals or groups based on common characteristics such as race, religion, gender, etc.; and Subtask C, involving the detection of the fine-grained type of hate speech which takes one value from the following types: (i) race/ethnicity/nationality, (ii) religion/belief, (iii) ideology, (iv) disability/disease, (v) social class, and (vi) gender. In total, 40 teams signed up to participate in Subtask A, and 17 of them submitted test runs. For Subtask B, 26 teams signed up to participate and 12 of them submitted runs. And for Subtask C, 23 teams signed up to participate and 10 of them submitted runs. 10 teams submitted papers describing their participation in one subtask or more, and 8 papers were accepted. We present and analyze all submissions in this paper."
                },
                {
                    "title": "OffendES: A New Corpus in Spanish for Offensive Language Research",
                    "abstract": "Offensive language detection and analysis has become a major area of research in Natural Language Processing. The freedom of participation in social media has exposed online users to posts designed to denigrate, insult or hurt them according to gender, race, religion, ideology, or other personal characteristics. Focusing on young influencers from the well-known social platforms of Twitter, Instagram, and YouTube, we have collected a corpus composed of 47,128 Spanish comments manually labeled on offensive pre-defined categories. A subset of the corpus attaches a degree of confidence to each label, so both multi-class classification and multi-output regression studies are possible. In this paper, we introduce the corpus, discuss its building process, novelties, and some preliminary experiments with it to serve as a baseline for the research community."
                },
                {
                    "title": "A Large-scale Comprehensive Abusiveness Detection Dataset with Multifaceted Labels from Reddit",
                    "abstract": "As users in online communities suffer from severe side effects of abusive language, many researchers attempted to detect abusive texts from social media, presenting several datasets for such detection. However, none of them contain both comprehensive labels and contextual information, which are essential for thoroughly detecting all kinds of abusiveness from texts, since datasets with such fine-grained features demand a significant amount of annotations, leading to much increased complexity. In this paper, we propose a Comprehensive Abusiveness Detection Dataset (CADD), collected from the English Reddit posts, with multifaceted labels and contexts. Our dataset is annotated hierarchically for an efficient annotation through crowdsourcing on a large-scale. We also empirically explore the characteristics of our dataset and provide a detailed analysis for novel insights. The results of our experiments with strong pre-trained natural language understanding models on our dataset show that our dataset gives rise to meaningful performance, assuring its practicality for abusive language detection."
                },
                {
                    "title": "Overview of the Task on Automatic Misogyny Identification at IberEval 2018",
                    "abstract": ". Automatic Misogyny Identi\ufb01cation (AMI) is a new shared task proposed for the \ufb01rst time at the IberEval 2018 evaluation campaign. The AMI task proposes misogyny identi\ufb01cation, misogynistic behaviour categorization and target classi\ufb01cation both from Spanish and English tweets. We have received a total of 32 runs for English and 24 for Spanish, submitted by 11 di\ufb00erent teams from 5 countries. We present here the datasets, the evaluation methodology, an overview of the proposed systems and the obtained results. Finally, we draw some conclusions and discuss future work."
                },
                {
                    "title": "Overview of MEX-A3T at IberEval 2018: Authorship and Aggressiveness Analysis in Mexican Spanish Tweets",
                    "abstract": "This paper presents the framework and results from the MEX-A3T track at IberEval 2018. This track considers two tasks, author profiling and aggressiveness detection, both of them using Mexican Spanish tweets. The author profiling task aims to identify the place of residence and occupation of Twitter users. On the other hand, the aggressiveness detection task aims to discriminate between aggressive and non-aggressive tweets. For these two tasks we have built new corpora considering tweets from Mexican Twitter users. This paper compares and discusses the results from the participants."
                },
                {
                    "title": "Overview of the GermEval 2018 Shared Task on the Identification of Offensive Language",
                    "abstract": "We present the pilot edition of the GermEval Shared Task on the Identification of Offensive Language. This shared task deals with the classification of German tweets from Twitter. It comprises two tasks, a coarse-grained binary classification task and a fine-grained multi-class classification task. The shared task had 20 participants submitting 51 runs for the coarse-grained task and 25 runs for the fine-grained task. Since this is a pilot task, we describe the process of extracting the raw-data for the data collection and the annotation schema. We evaluate the results of the systems submitted to the shared task. The shared task homepage can be found at https://projects.cai. fbi.h-da.de/iggsa/"
                },
                {
                    "title": "NLTK: The Natural Language Toolkit",
                    "abstract": "The Natural Language Toolkit is a suite of program modules, data sets and tutorials supporting research and teaching in computational linguistics and natural language processing. NLTK is written in Python and distributed under the GPL open source license. Over the past year the toolkit has been rewritten, simplifying many linguistic data structures and taking advantage of recent enhancements in the Python language. This paper reports on the simpli-\ufb01ed toolkit and explains how it is used in teaching NLP."
                },
                {
                    "title": "Data in Brief",
                    "abstract": "Dataset"
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively reduce cultural bias in large language models (LLMs) to ensure they accurately represent and respect diverse cultural perspectives, particularly for low-resource cultures?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nAddressing cultural bias in LLMs is crucial for fostering inclusivity and fairness in AI systems, which are increasingly integrated into various aspects of society. By solving this problem, we can enhance human-AI collaboration, promote AI democracy, and ensure that AI technologies serve a global audience rather than favoring Western perspectives. This research could lead to the development of more culturally aware AI applications, advancing knowledge in both AI and cultural studies, and paving the way for practical applications in multilingual and multicultural environments.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexities of cultural representation and the limitations of existing datasets, particularly for low-resource cultures. Naive approaches, such as simple prompt engineering, often fail to capture the nuances of cultural values and perspectives. Additionally, the need for large-scale, diverse training data and the computational resources required for pre-training culturally-aware LLMs pose significant technical and practical obstacles. The lack of affordable solutions further complicates the issue, making it difficult for researchers to address cultural bias effectively.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either prompt engineering or large-scale pre-training of culturally-aware LLMs, both of which have significant limitations. The former often lacks effectiveness in low-resource cultures, while the latter requires extensive datasets and computational power that are not accessible to many researchers. Additionally, there has been a gap in methodologies that combine cost-effectiveness with cultural sensitivity. Our approach differs by utilizing a semantic data augmentation technique based on the World Values Survey, allowing for the generation of culturally relevant data without the need for large-scale datasets.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, CultureLLM, consists of three key components: sampling, semantic data augmentation, and fine-tuning. We will use the World Values Survey as seed data, selecting 50 samples across various cultural topics. The semantic data augmentation will generate semantically equivalent samples to enrich the training dataset. Finally, we will fine-tune both culture-specific LLMs (e.g., CultureLLM-Ar for Arabic and CultureLLM-Tr for Turkish) and"
            }
        },
        "author_data": {
            "0d609932-f05d-49ef-89d0-11e5b9c0edeb": {
                "pk": "0d609932-f05d-49ef-89d0-11e5b9c0edeb",
                "name": "Cheng Li",
                "collaborators": [
                    "Yifan Cheng",
                    "Li Ma",
                    "Cheng Liang"
                ],
                "domain": [
                    "Bayesian Statistics",
                    "Gaussian Processes",
                    "Spatial Statistics",
                    "Factor Analysis"
                ],
                "publications": [
                    {
                        "title": "Bayesian Fixed-domain Asymptotics for Covariance Parameters in a Gaussian Process Model",
                        "abstract": "Gaussian process models typically contain finite dimensional parameters in the covariance function that need to be estimated from the data. We study the Bayesian fixed-domain asymptotics for the covariance parameters in a universal kriging model with an isotropic Matern covariance function, which has many applications in spatial statistics. We show that when the dimension of domain is less than or equal to three, the joint posterior distribution of the microergodic parameter and the range parameter can be factored independently into the product of their marginal posteriors under fixed-domain asymptotics. The posterior of the microergodic parameter is asymptotically close in total variation distance to a normal distribution with shrinking variance, while the posterior distribution of the range parameter does not converge to any point mass distribution in general. Our theory allows an unbounded prior support for the range parameter and flexible designs of sampling points. We further study the asymptotic efficiency and convergence rates in posterior prediction for the Bayesian kriging predictor with covariance parameters randomly drawn from their posterior distribution. In the special case of one-dimensional Ornstein-Uhlenbeck process, we derive explicitly the limiting posterior of the range parameter and the posterior convergence rate for asymptotic efficiency in posterior prediction. We verify these asymptotic results in numerical experiments."
                    },
                    {
                        "title": "Enhancing Scalability in Bayesian Nonparametric Factor Analysis of Spatiotemporal Data",
                        "abstract": "This manuscript puts forward novel practicable spatiotemporal Bayesian factor analysis frameworks computationally feasible for moderate to large data. Our models exhibit significantly enhanced computational scalability and storage efficiency, deliver high overall modeling performances, and possess powerful inferential capabilities for adequately predicting outcomes at future time points or new spatial locations and satisfactorily clustering spatial locations into regions with similar temporal trajectories, a frequently encountered crucial task. We integrate on top of a baseline separable factor model with temporally dependent latent factors and spatially dependent factor loadings under a probit stick breaking process (PSBP) prior a new slice sampling algorithm that permits unknown varying numbers of spatial mixture components across all factors and guarantees them to be non-increasing through the MCMC iterations, thus considerably enhancing model flexibility, efficiency, and scalability. We further introduce a novel spatial latent nearest-neighbor Gaussian process (NNGP) prior and new sequential updating algorithms for the spatially varying latent variables in the PSBP prior, thereby attaining high spatial scalability. The markedly accelerated posterior sampling and spatial prediction as well as the great modeling and inferential performances of our models are substantiated by our simulation experiments."
                    },
                    {
                        "title": "A non-local population model of logistic type equation",
                        "abstract": "In this paper, we propose a new non-local population model of logistic type equation on a bounded Lipschitz domain in the whole Euclidean space. This model preserves the L^2 norm, which is called mass, of the solution on the domain. We show that this model has the global existence, stability and asymptotic behavior at time infinity."
                    }
                ]
            },
            "57f724ed-bac1-41f3-85a1-b5a138712cb5": {
                "pk": "57f724ed-bac1-41f3-85a1-b5a138712cb5",
                "name": "Mengzhou Chen",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "44251f6a-e328-47af-8f3c-ff110564d2ac": {
                "pk": "44251f6a-e328-47af-8f3c-ff110564d2ac",
                "name": "Jindong Wang",
                "collaborators": [
                    "Yiqiang Chen",
                    "Shuonan Wu",
                    "Wang Lu",
                    "Xing Xie",
                    "Xin Qin",
                    "Meiyu Huang",
                    "Chaohui Yu",
                    "Yidong Wang",
                    "Xixu Hu",
                    "Yiqiao Jin"
                ],
                "domain": [
                    "Machine Learning",
                    "Domain Adaptation",
                    "Federated Learning",
                    "Numerical Methods"
                ],
                "publications": [
                    {
                        "title": "A robust solver for H(curl) convection-diffusion and its local Fourier analysis",
                        "abstract": "In this paper, we present a robust and efficient multigrid solver based on an exponential-fitting discretization for 2D H(curl) convection-diffusion problems. By leveraging an exponential identity, we characterize the kernel of H(curl) convection-diffusion problems and design a suitable hybrid smoother. This smoother incorporates a lexicographic Gauss-Seidel smoother within a downwind type and smoothing over an auxiliary problem, corresponding to H(grad) convection-diffusion problems for kernel correction. We analyze the convergence properties of the smoothers and the two-level method using local Fourier analysis (LFA). The performance of the algorithms demonstrates robustness in both convection-dominated and diffusion-dominated cases."
                    },
                    {
                        "title": "Discontinuous Galerkin methods for magnetic advection-diffusion problems",
                        "abstract": "We devise and analyze a class of the primal discontinuous Galerkin methods for the magnetic advection-diffusion problems based on the weighted-residual approach. In addition to the upwind stabilization, we find a new mechanism under the vector case that provides more flexibility in constructing the schemes. For the more general Friedrichs system, we show the stability and optimal error estimate, which boil down to two core ingredients -- the weight function and the special projection -- that contain information of advection. Numerical experiments are provided to verify the theoretical results."
                    },
                    {
                        "title": "A hybridizable discontinuous Galerkin method for magnetic advection-diffusion problems",
                        "abstract": "We propose and analyze a hybridizable discontinuous Galerkin (HDG) method for solving a mixed magnetic advection-diffusion problem within a more general Friedrichs system framework. With carefully constructed numerical traces, we introduce two distinct stabilization parameters: $\\tau_t$ for the tangential trace and $\\tau_n$ for the normal trace. These parameters are tailored to satisfy different requirements, ensuring the stability and convergence of the method. Furthermore, we incorporate a weight function to facilitate the establishment of stability conditions. We also investigate an elementwise postprocessing technique that proves to be effective for both two-dimensional and three-dimensional problems in terms of broken $H({\\rm curl})$ semi-norm accuracy improvement. Extensive numerical examples are presented to showcase the performance and effectiveness of the HDG method and the postprocessing techniques."
                    },
                    {
                        "title": "Exponentially-fitted finite elements for $H({\\rm curl})$ and $H({\\rm div})$ convection-diffusion problems",
                        "abstract": "This paper presents a novel approach to the construction of the lowest order $H(\\mathrm{curl})$ and $H(\\mathrm{div})$ exponentially-fitted finite element spaces ${\\mathcal{S}_{1^-}^{k}}~(k=1,2)$ on 3D simplicial mesh for corresponding convection-diffusion problems. It is noteworthy that this method not only facilitates the construction of the functions themselves but also provides corresponding discrete fluxes simultaneously. Utilizing this approach, we successfully establish a discrete convection-diffusion complex and employ a specialized weighted interpolation to establish a bridge between the continuous complex and the discrete complex, resulting in a coherent framework. Furthermore, we demonstrate the commutativity of the framework when the convection field is locally constant, along with the exactness of the discrete convection-diffusion complex. Consequently, these types of spaces can be directly employed to devise the corresponding discrete scheme through a Petrov-Galerkin method."
                    },
                    {
                        "title": "Towards Optimization and Model Selection for Domain Generalization: A Mixup-guided Solution",
                        "abstract": "The distribution shifts between training and test data typically undermine the performance of models. In recent years, lots of work pays attention to domain generalization (DG) where distribution shifts exist, and target data are unseen. Despite the progress in algorithm design, two foundational factors have long been ignored: 1) the optimization for regularization-based objectives, and 2) the model selection for DG since no knowledge about the target domain can be utilized. In this paper, we propose Mixup guided optimization and selection techniques for DG. For optimization, we utilize an adapted Mixup to generate an out-of-distribution dataset that can guide the preference direction and optimize with Pareto optimization. For model selection, we generate a validation dataset with a closer distance to the target distribution, and thereby it can better represent the target data. We also present some theoretical insights behind our proposals. Comprehensive experiments demonstrate that our model optimization and selection techniques can largely improve the performance of existing domain generalization algorithms and even achieve new state-of-the-art results."
                    },
                    {
                        "title": "FedHealth 2: Weighted Federated Transfer Learning via Batch Normalization for Personalized Healthcare",
                        "abstract": "The success of machine learning applications often needs a large quantity of data. Recently, federated learning (FL) is attracting increasing attention due to the demand for data privacy and security, especially in the medical field. However, the performance of existing FL approaches often deteriorates when there exist domain shifts among clients, and few previous works focus on personalization in healthcare. In this article, we propose FedHealth 2, an extension of FedHealth \\cite{chen2020fedhealth} to tackle domain shifts and get personalized models for local clients. FedHealth 2 obtains the client similarities via a pretrained model, and then it averages all weighted models with preserving local batch normalization. Wearable activity recognition and COVID-19 auxiliary diagnosis experiments have evaluated that FedHealth 2 can achieve better accuracy (10%+ improvement for activity recognition) and personalized healthcare without compromising privacy and security."
                    },
                    {
                        "title": "Cross-domain Activity Recognition via Substructural Optimal Transport",
                        "abstract": "It is expensive and time-consuming to collect sufficient labeled data for human activity recognition (HAR). Domain adaptation is a promising approach for cross-domain activity recognition. Existing methods mainly focus on adapting cross-domain representations via domain-level, class-level, or sample-level distribution matching. However, they might fail to capture the fine-grained locality information in activity data. The domain- and class-level matching are too coarse that may result in under-adaptation, while sample-level matching may be affected by the noise seriously and eventually cause over-adaptation. In this paper, we propose substructure-level matching for domain adaptation (SSDA) to better utilize the locality information of activity data for accurate and efficient knowledge transfer. Based on SSDA, we propose an optimal transport-based implementation, Substructural Optimal Transport (SOT), for cross-domain HAR. We obtain the substructures of activities via clustering methods and seeks the coupling of the weighted substructures between different domains. We conduct comprehensive experiments on four public activity recognition datasets (i.e. UCI-DSADS, UCI-HAR, USC-HAD, PAMAP2), which demonstrates that SOT significantly outperforms other state-of-the-art methods w.r.t classification accuracy (9%+ improvement). In addition, our mehtod is 5x faster than traditional OT-based DA methods with the same hyper-parameters."
                    },
                    {
                        "title": "FedCLIP: Fast Generalization and Personalization for CLIP in Federated Learning",
                        "abstract": "Federated learning (FL) has emerged as a new paradigm for privacy-preserving computation in recent years. Unfortunately, FL faces two critical challenges that hinder its actual performance: data distribution heterogeneity and high resource costs brought by large foundation models. Specifically, the non-IID data in different clients make existing FL algorithms hard to converge while the high resource costs, including computational and communication costs that increase the deployment difficulty in real-world scenarios. In this paper, we propose an effective yet simple method, named FedCLIP, to achieve fast generalization and personalization for CLIP in federated learning. Concretely, we design an attention-based adapter for the large model, CLIP, and the rest operations merely depend on adapters. Lightweight adapters can make the most use of pretrained model information and ensure models be adaptive for clients in specific tasks. Simultaneously, small-scale operations can mitigate the computational burden and communication burden caused by large models. Extensive experiments are conducted on three datasets with distribution shifts. Qualitative and quantitative results demonstrate that FedCLIP significantly outperforms other baselines (9% overall improvements on PACS) and effectively reduces computational and communication costs (283x faster than FedAVG). Our code will be available at: https://github.com/microsoft/PersonalizedFL."
                    },
                    {
                        "title": "Scito2M: A 2 Million, 30-Year Cross-disciplinary Dataset for Temporal Scientometric Analysis",
                        "abstract": "Understanding the creation, evolution, and dissemination of scientific knowledge is crucial for bridging diverse subject areas and addressing complex global challenges such as pandemics, climate change, and ethical AI. Scientometrics, the quantitative and qualitative study of scientific literature, provides valuable insights into these processes. We introduce Scito2M, a longitudinal scientometric dataset with over two million academic publications, providing comprehensive contents information and citation graphs to support cross-disciplinary analyses. Using Scito2M, we conduct a temporal study spanning over 30 years to explore key questions in scientometrics: the evolution of academic terminology, citation patterns, and interdisciplinary knowledge exchange. Our findings reveal critical insights, such as disparities in epistemic cultures, knowledge production modes, and citation practices. For example, rapidly developing, application-driven fields like LLMs exhibit significantly shorter citation age (2.48 years) compared to traditional theoretical disciplines like oral history (9.71 years)."
                    },
                    {
                        "title": "Cross-position Activity Recognition with Stratified Transfer Learning",
                        "abstract": "Human activity recognition aims to recognize the activities of daily living by utilizing the sensors on different body parts. However, when the labeled data from a certain body position (i.e. target domain) is missing, how to leverage the data from other positions (i.e. source domain) to help learn the activity labels of this position? When there are several source domains available, it is often difficult to select the most similar source domain to the target domain. With the selected source domain, we need to perform accurate knowledge transfer between domains. Existing methods only learn the global distance between domains while ignoring the local property. In this paper, we propose a \\textit{Stratified Transfer Learning} (STL) framework to perform both source domain selection and knowledge transfer. STL is based on our proposed \\textit{Stratified} distance to capture the local property of domains. STL consists of two components: Stratified Domain Selection (STL-SDS) can select the most similar source domain to the target domain; Stratified Activity Transfer (STL-SAT) is able to perform accurate knowledge transfer. Extensive experiments on three public activity recognition datasets demonstrate the superiority of STL. Furthermore, we extensively investigate the performance of transfer learning across different degrees of similarities and activity levels between domains. We also discuss the potential applications of STL in other fields of pervasive computing for future research."
                    },
                    {
                        "title": "Accelerating Deep Unsupervised Domain Adaptation with Transfer Channel Pruning",
                        "abstract": "Deep unsupervised domain adaptation (UDA) has recently received increasing attention from researchers. However, existing methods are computationally intensive due to the computation cost of Convolutional Neural Networks (CNN) adopted by most work. To date, there is no effective network compression method for accelerating these models. In this paper, we propose a unified Transfer Channel Pruning (TCP) approach for accelerating UDA models. TCP is capable of compressing the deep UDA model by pruning less important channels while simultaneously learning transferable features by reducing the cross-domain distribution divergence. Therefore, it reduces the impact of negative transfer and maintains competitive performance on the target task. To the best of our knowledge, TCP is the first approach that aims at accelerating deep UDA models. TCP is validated on two benchmark datasets-Office-31 and ImageCLEF-DA with two common backbone networks-VGG16 and ResNet50. Experimental results demonstrate that TCP achieves comparable or better classification accuracy than other comparison methods while significantly reducing the computational cost. To be more specific, in VGG16, we get even higher accuracy after pruning 26% floating point operations (FLOPs); in ResNet50, we also get higher accuracy on half of the tasks after pruning 12% FLOPs. We hope that TCP will open a new door for future research on accelerating transfer learning models."
                    },
                    {
                        "title": "Transfer Learning with Dynamic Adversarial Adaptation Network",
                        "abstract": "The recent advances in deep transfer learning reveal that adversarial learning can be embedded into deep networks to learn more transferable features to reduce the distribution discrepancy between two domains. Existing adversarial domain adaptation methods either learn a single domain discriminator to align the global source and target distributions or pay attention to align subdomains based on multiple discriminators. However, in real applications, the marginal (global) and conditional (local) distributions between domains are often contributing differently to the adaptation. There is currently no method to dynamically and quantitatively evaluate the relative importance of these two distributions for adversarial learning. In this paper, we propose a novel Dynamic Adversarial Adaptation Network (DAAN) to dynamically learn domain-invariant representations while quantitatively evaluate the relative importance of global and local domain distributions. To the best of our knowledge, DAAN is the first attempt to perform dynamic adversarial distribution adaptation for deep adversarial learning. DAAN is extremely easy to implement and train in real applications. We theoretically analyze the effectiveness of DAAN, and it can also be explained in an attention strategy. Extensive experiments demonstrate that DAAN achieves better classification accuracy compared to state-of-the-art deep and adversarial methods. Results also imply the necessity and effectiveness of the dynamic distribution adaptation in adversarial transfer learning."
                    },
                    {
                        "title": "Unsupervised Deep Anomaly Detection for Multi-Sensor Time-Series Signals",
                        "abstract": "Nowadays, multi-sensor technologies are applied in many fields, e.g., Health Care (HC), Human Activity Recognition (HAR), and Industrial Control System (ICS). These sensors can generate a substantial amount of multivariate time-series data. Unsupervised anomaly detection on multi-sensor time-series data has been proven critical in machine learning researches. The key challenge is to discover generalized normal patterns by capturing spatial-temporal correlation in multi-sensor data. Beyond this challenge, the noisy data is often intertwined with the training data, which is likely to mislead the model by making it hard to distinguish between the normal, abnormal, and noisy data. Few of previous researches can jointly address these two challenges. In this paper, we propose a novel deep learning-based anomaly detection algorithm called Deep Convolutional Autoencoding Memory network (CAE-M). We first build a Deep Convolutional Autoencoder to characterize spatial dependence of multi-sensor data with a Maximum Mean Discrepancy (MMD) to better distinguish between the noisy, normal, and abnormal data. Then, we construct a Memory Network consisting of linear (Autoregressive Model) and non-linear predictions (Bidirectional LSTM with Attention) to capture temporal dependence from time-series data. Finally, CAE-M jointly optimizes these two subnetworks. We empirically compare the proposed approach with several state-of-the-art anomaly detection methods on HAR and HC datasets. Experimental results demonstrate that our proposed model outperforms these existing methods."
                    },
                    {
                        "title": "Equivariant Disentangled Transformation for Domain Generalization under Combination Shift",
                        "abstract": "Machine learning systems may encounter unexpected problems when the data distribution changes in the deployment environment. A major reason is that certain combinations of domains and labels are not observed during training but appear in the test environment. Although various invariance-based algorithms can be applied, we find that the performance gain is often marginal. To formally analyze this issue, we provide a unique algebraic formulation of the combination shift problem based on the concepts of homomorphism, equivariance, and a refined definition of disentanglement. The algebraic requirements naturally derive a simple yet effective method, referred to as equivariant disentangled transformation (EDT), which augments the data based on the algebraic structures of labels and makes the transformation satisfy the equivariance and disentanglement requirements. Experimental results demonstrate that invariance may be insufficient, and it is important to exploit the equivariance structure in the combination shift problem."
                    },
                    {
                        "title": "Selective Mixup Helps with Distribution Shifts, But Not (Only) because of Mixup",
                        "abstract": "Mixup is a highly successful technique to improve generalization of neural networks by augmenting the training data with combinations of random pairs. Selective mixup is a family of methods that apply mixup to specific pairs, e.g. only combining examples across classes or domains. These methods have claimed remarkable improvements on benchmarks with distribution shifts, but their mechanisms and limitations remain poorly understood.   We examine an overlooked aspect of selective mixup that explains its success in a completely new light. We find that the non-random selection of pairs affects the training distribution and improve generalization by means completely unrelated to the mixing. For example in binary classification, mixup across classes implicitly resamples the data for a uniform class distribution - a classical solution to label shift. We show empirically that this implicit resampling explains much of the improvements in prior work. Theoretically, these results rely on a regression toward the mean, an accidental property that we identify in several datasets.   We have found a new equivalence between two successful methods: selective mixup and resampling. We identify limits of the former, confirm the effectiveness of the latter, and find better combinations of their respective benefits."
                    },
                    {
                        "title": "On Catastrophic Inheritance of Large Foundation Models",
                        "abstract": "Large foundation models (LFMs) are claiming incredible performances. Yet great concerns have been raised about their mythic and uninterpreted potentials not only in machine learning, but also in various other disciplines. In this position paper, we propose to identify a neglected issue deeply rooted in LFMs: Catastrophic Inheritance, describing the weaknesses and limitations inherited from biased large-scale pre-training data to behaviors of LFMs on the downstream tasks, including samples that are corrupted, long-tailed, noisy, out-of-distributed, to name a few. Such inheritance can potentially cause catastrophes to downstream applications, such as bias, lack of generalization, deteriorated performance, security vulnerability, privacy leakage, and value misalignment. We discuss the challenges behind this issue and propose UIM, a framework to Understand the catastrophic inheritance of LFMs from both pre-training and downstream adaptation, Interpret the implications of catastrophic inheritance on downstream tasks, and how to Mitigate it. UIM aims to unite both the machine learning and social sciences communities for more responsible and promising AI development and deployment."
                    }
                ]
            },
            "4dff4511-b3c3-4433-88c8-79c3aa5ca48a": {
                "pk": "4dff4511-b3c3-4433-88c8-79c3aa5ca48a",
                "name": "Sunayana Sitaram",
                "collaborators": [
                    "Monojit Choudhury",
                    "Hemant Yadav",
                    "Sandipan Dandapat",
                    "Rajiv Ratn Shah",
                    "Simran Khanuja",
                    "Kabir Ahuja",
                    "Ashutosh Sathe",
                    "Sanchit Ahuja",
                    "Divyanshu Aggarwal",
                    "Florian Schneider"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Speech Recognition",
                    "Multilingual Models",
                    "Fairness in AI"
                ],
                "publications": [
                    {
                        "title": "A Survey of Multilingual Models for Automatic Speech Recognition",
                        "abstract": "Although Automatic Speech Recognition (ASR) systems have achieved human-like performance for a few languages, the majority of the world's languages do not have usable systems due to the lack of large speech datasets to train these models. Cross-lingual transfer is an attractive solution to this problem, because low-resource languages can potentially benefit from higher-resource languages either through transfer learning, or being jointly trained in the same multilingual model. The problem of cross-lingual transfer has been well studied in ASR, however, recent advances in Self Supervised Learning are opening up avenues for unlabeled speech data to be used in multilingual ASR models, which can pave the way for improved performance on low-resource languages. In this paper, we survey the state of the art in multilingual ASR models that are built with cross-lingual transfer in mind. We present best practices for building multilingual models from research across diverse languages and techniques, discuss open questions and provide recommendations for future work."
                    },
                    {
                        "title": "M5 -- A Diverse Benchmark to Assess the Performance of Large Multimodal Models Across Multilingual and Multicultural Vision-Language Tasks",
                        "abstract": "Since the release of ChatGPT, the field of Natural Language Processing has experienced rapid advancements, particularly in Large Language Models (LLMs) and their multimodal counterparts, Large Multimodal Models (LMMs). Despite their impressive capabilities, LLMs often exhibit significant performance disparities across different languages and cultural contexts, as demonstrated by various text-only benchmarks. However, current research lacks such benchmarks for multimodal visio-linguistic settings. This work fills this gap by introducing M5, the first comprehensive benchmark designed to evaluate LMMs on diverse vision-language tasks within a multilingual and multicultural context. M5 includes eight datasets covering five tasks and $41$ languages, with a focus on underrepresented languages and culturally diverse images. Furthermore, we introduce two novel datasets, M5-VGR and M5-VLOD, including a new Visio-Linguistic Outlier Detection task, in which all evaluated open-source models fail to significantly surpass the random baseline. Through extensive evaluation and analyses, we highlight substantial task-agnostic performance disparities between high- and low-resource languages. Moreover, we show that larger models do not necessarily outperform smaller ones in a multilingual setting."
                    },
                    {
                        "title": "Contamination Report for Multilingual Benchmarks",
                        "abstract": "Benchmark contamination refers to the presence of test datasets in Large Language Model (LLM) pre-training or post-training data. Contamination can lead to inflated scores on benchmarks, compromising evaluation results and making it difficult to determine the capabilities of models. In this work, we study the contamination of popular multilingual benchmarks in LLMs that support multiple languages. We use the Black Box test to determine whether $7$ frequently used multilingual benchmarks are contaminated in $7$ popular open and closed LLMs and find that almost all models show signs of being contaminated with almost all the benchmarks we test. Our findings can help the community determine the best set of benchmarks to use for multilingual evaluation."
                    },
                    {
                        "title": "A New Dataset for Natural Language Inference from Code-mixed Conversations",
                        "abstract": "Natural Language Inference (NLI) is the task of inferring the logical relationship, typically entailment or contradiction, between a premise and hypothesis. Code-mixing is the use of more than one language in the same conversation or utterance, and is prevalent in multilingual communities all over the world. In this paper, we present the first dataset for code-mixed NLI, in which both the premises and hypotheses are in code-mixed Hindi-English. We use data from Hindi movies (Bollywood) as premises, and crowd-source hypotheses from Hindi-English bilinguals. We conduct a pilot annotation study and describe the final annotation protocol based on observations from the pilot. Currently, the data collected consists of 400 premises in the form of code-mixed conversation snippets and 2240 code-mixed hypotheses. We conduct an extensive analysis to infer the linguistic phenomena commonly observed in the dataset obtained. We evaluate the dataset using a standard mBERT-based pipeline for NLI and report results."
                    },
                    {
                        "title": "Beyond Static Models and Test Sets: Benchmarking the Potential of Pre-trained Models Across Tasks and Languages",
                        "abstract": "Although recent Massively Multilingual Language Models (MMLMs) like mBERT and XLMR support around 100 languages, most existing multilingual NLP benchmarks provide evaluation data in only a handful of these languages with little linguistic diversity. We argue that this makes the existing practices in multilingual evaluation unreliable and does not provide a full picture of the performance of MMLMs across the linguistic landscape. We propose that the recent work done in Performance Prediction for NLP tasks can serve as a potential solution in fixing benchmarking in Multilingual NLP by utilizing features related to data and language typology to estimate the performance of an MMLM on different languages. We compare performance prediction with translating test data with a case study on four different multilingual datasets, and observe that these methods can provide reliable estimates of the performance that are often on-par with the translation based approaches, without the need for any additional translation as well as evaluation costs."
                    },
                    {
                        "title": "On the Calibration of Massively Multilingual Language Models",
                        "abstract": "Massively Multilingual Language Models (MMLMs) have recently gained popularity due to their surprising effectiveness in cross-lingual transfer. While there has been much work in evaluating these models for their performance on a variety of tasks and languages, little attention has been paid on how well calibrated these models are with respect to the confidence in their predictions. We first investigate the calibration of MMLMs in the zero-shot setting and observe a clear case of miscalibration in low-resource languages or those which are typologically diverse from English. Next, we empirically show that calibration methods like temperature scaling and label smoothing do reasonably well towards improving calibration in the zero-shot scenario. We also find that few-shot examples in the language can further help reduce the calibration errors, often substantially. Overall, our work contributes towards building more reliable multilingual models by highlighting the issue of their miscalibration, understanding what language and model specific factors influence it, and pointing out the strategies to improve the same."
                    },
                    {
                        "title": "Fairness in Language Models Beyond English: Gaps and Challenges",
                        "abstract": "With language models becoming increasingly ubiquitous, it has become essential to address their inequitable treatment of diverse demographic groups and factors. Most research on evaluating and mitigating fairness harms has been concentrated on English, while multilingual models and non-English languages have received comparatively little attention. This paper presents a survey of fairness in multilingual and non-English contexts, highlighting the shortcomings of current research and the difficulties faced by methods designed for English. We contend that the multitude of diverse cultures and languages across the world makes it infeasible to achieve comprehensive coverage in terms of constructing fairness datasets. Thus, the measurement and mitigation of biases must evolve beyond the current dataset-driven practices that are narrowly focused on specific dimensions and types of biases and, therefore, impossible to scale across languages and cultures."
                    },
                    {
                        "title": "Analysing the Masked predictive coding training criterion for pre-training a Speech Representation Model",
                        "abstract": "Recent developments in pre-trained speech representation utilizing self-supervised learning (SSL) have yielded exceptional results on a variety of downstream tasks. One such technique, known as masked predictive coding (MPC), has been employed by some of the most high-performing models. In this study, we investigate the impact of MPC loss on the type of information learnt at various layers in the HuBERT model, using nine probing tasks. Our findings indicate that the amount of content information learned at various layers of the HuBERT model has a positive correlation to the MPC loss. Additionally, it is also observed that any speaker-related information learned at intermediate layers of the model, is an indirect consequence of the learning process, and therefore cannot be controlled using the MPC loss. These findings may serve as inspiration for further research in the speech community, specifically in the development of new pre-training tasks or the exploration of new pre-training criterion's that directly preserves both speaker and content information at various layers of a learnt model."
                    },
                    {
                        "title": "On Evaluating and Mitigating Gender Biases in Multilingual Settings",
                        "abstract": "While understanding and removing gender biases in language models has been a long-standing problem in Natural Language Processing, prior research work has primarily been limited to English. In this work, we investigate some of the challenges with evaluating and mitigating biases in multilingual settings which stem from a lack of existing benchmarks and resources for bias evaluation beyond English especially for non-western context. In this paper, we first create a benchmark for evaluating gender biases in pre-trained masked language models by extending DisCo to different Indian languages using human annotations. We extend various debiasing methods to work beyond English and evaluate their effectiveness for SOTA massively multilingual models on our proposed metric. Overall, our work highlights the challenges that arise while studying social biases in multilingual settings and provides resources as well as mitigation techniques to take a step toward scaling to more languages."
                    },
                    {
                        "title": "A Unified Framework and Dataset for Assessing Societal Bias in Vision-Language Models",
                        "abstract": "Vision-language models (VLMs) have gained widespread adoption in both industry and academia. In this study, we propose a unified framework for systematically evaluating gender, race, and age biases in VLMs with respect to professions. Our evaluation encompasses all supported inference modes of the recent VLMs, including image-to-text, text-to-text, text-to-image, and image-to-image. Additionally, we propose an automated pipeline to generate high-quality synthetic datasets that intentionally conceal gender, race, and age information across different professional domains, both in generated text and images. The dataset includes action-based descriptions of each profession and serves as a benchmark for evaluating societal biases in vision-language models (VLMs). In our comparative analysis of widely used VLMs, we have identified that varying input-output modalities lead to discernible differences in bias magnitudes and directions. Additionally, we find that VLM models exhibit distinct biases across different bias attributes we investigated. We hope our work will help guide future progress in improving VLMs to learn socially unbiased representations. We will release our data and code."
                    },
                    {
                        "title": "MS-HuBERT: Mitigating Pre-training and Inference Mismatch in Masked Language Modelling methods for learning Speech Representations",
                        "abstract": "In recent years, self-supervised pre-training methods have gained significant traction in learning high-level information from raw speech. Among these methods, HuBERT has demonstrated SOTA performance in automatic speech recognition (ASR). However, HuBERT's performance lags behind data2vec due to disparities in pre-training strategies. In this paper, we propose (i) a Swap method to address pre-training and inference mismatch observed in HuBERT and (ii) incorporates Multicluster masked prediction loss for more effective utilization of the models capacity. The resulting method is, MS-HuBERT, an end-to-end self-supervised pre-training method for learning robust speech representations. It beats vanilla HuBERT on the ASR Librispeech benchmark on average by a 5% margin when evaluated on different finetuning splits. Additionally, we demonstrate that the learned embeddings obtained during pre-training encode essential information for improving performance of content based tasks such as ASR."
                    },
                    {
                        "title": "Improving Self Consistency in LLMs through Probabilistic Tokenization",
                        "abstract": "Prior research has demonstrated noticeable performance gains through the use of probabilistic tokenizations, an approach that involves employing multiple tokenizations of the same input string during the training phase of a language model. Despite these promising findings, modern large language models (LLMs) have yet to be trained using probabilistic tokenizations. Interestingly, while the tokenizers of these contemporary LLMs have the capability to generate multiple tokenizations, this property remains underutilized.   In this work, we propose a novel method to leverage the multiple tokenization capabilities of modern LLM tokenizers, aiming to enhance the self-consistency of LLMs in reasoning tasks. Our experiments indicate that when utilizing probabilistic tokenizations, LLMs generate logically diverse reasoning paths, moving beyond mere surface-level linguistic diversity.We carefully study probabilistic tokenization and offer insights to explain the self consistency improvements it brings through extensive experimentation on 5 LLM families and 4 reasoning benchmarks."
                    },
                    {
                        "title": "JOOCI: a Framework for Learning Comprehensive Speech Representations",
                        "abstract": "Information in speech can be divided into two categories: what is being said (content) and how it is expressed (other). Current state-of-the-art (SOTA) techniques model speech at fixed segments, usually 10-25 ms, using a single embedding. Given the orthogonal nature of other and content information, attempting to optimize both within a single embedding results in suboptimal solutions. This approach divides the models capacity, limiting its ability to build complex hierarchical features effectively. In this work, we present an end-to-end speech representation learning framework designed to jointly optimize the other and content information (JOOCI) in speech. By using separate learnable parameters, JOOCI addresses this optimization challenge by modeling other and content information independently. Our results show that JOOCI consistently outperforms other SOTA models of similar size (100 million parameters) and pre-training data used (960 hours) by a significant margin when evaluated on a range of speech downstream tasks in the SUPERB benchmark, as shown in Table 1."
                    },
                    {
                        "title": "Exploring Pretraining via Active Forgetting for Improving Cross Lingual Transfer for Decoder Language Models",
                        "abstract": "Large Language Models (LLMs) demonstrate exceptional capabilities in a multitude of NLP tasks. However, the efficacy of such models to languages other than English is often limited. Prior works have shown that encoder-only models such as BERT or XLM-RoBERTa show impressive cross lingual transfer of their capabilities from English to other languages. In this work, we propose a pretraining strategy that uses active forgetting to achieve similar cross lingual transfer in decoder-only LLMs. We show that LLMs pretrained with active forgetting are highly effective when adapting to new and unseen languages. Through extensive experimentation, we find that LLMs pretrained with active forgetting are able to learn better multilingual representations which translates to better performance in many downstream tasks."
                    },
                    {
                        "title": "Cross-lingual and Multilingual Spoken Term Detection for Low-Resource Indian Languages",
                        "abstract": "Spoken Term Detection (STD) is the task of searching for words or phrases within audio, given either text or spoken input as a query. In this work, we use state-of-the-art Hindi, Tamil and Telugu ASR systems cross-lingually for lexical Spoken Term Detection in ten low-resource Indian languages. Since no publicly available dataset exists for Spoken Term Detection in these languages, we create a new dataset using a publicly available TTS dataset. We report a standard metric for STD, Mean Term Weighted Value (MTWV) and show that ASR systems built in languages that are phonetically similar to the target languages have higher accuracy, however, it is also possible to get high MTWV scores for dissimilar languages by using a relaxed phone matching algorithm. We propose a technique to bootstrap the Grapheme-to-Phoneme (g2p) mapping between all the languages under consideration using publicly available resources. Gains are obtained when we combine the output of multiple ASR systems and when we use language-specific Language Models. We show that it is possible to perform STD cross-lingually in a zero-shot manner without the need for any language-specific speech data. We plan to make the STD dataset available for other researchers interested in cross-lingual STD."
                    },
                    {
                        "title": "DOSA: A Dataset of Social Artifacts from Different Indian Geographical Subcultures",
                        "abstract": "Generative models are increasingly being used in various applications, such as text generation, commonsense reasoning, and question-answering. To be effective globally, these models must be aware of and account for local socio-cultural contexts, making it necessary to have benchmarks to evaluate the models for their cultural familiarity. Since the training data for LLMs is web-based and the Web is limited in its representation of information, it does not capture knowledge present within communities that are not on the Web. Thus, these models exacerbate the inequities, semantic misalignment, and stereotypes from the Web. There has been a growing call for community-centered participatory research methods in NLP. In this work, we respond to this call by using participatory research methods to introduce $\\textit{DOSA}$, the first community-generated $\\textbf{D}$ataset $\\textbf{o}$f 615 $\\textbf{S}$ocial $\\textbf{A}$rtifacts, by engaging with 260 participants from 19 different Indian geographic subcultures. We use a gamified framework that relies on collective sensemaking to collect the names and descriptions of these artifacts such that the descriptions semantically align with the shared sensibilities of the individuals from those cultures. Next, we benchmark four popular LLMs and find that they show significant variation across regional sub-cultures in their ability to infer the artifacts."
                    },
                    {
                        "title": "Speech Representation Learning Revisited: The Necessity of Separate Learnable Parameters and Robust Data Augmentation",
                        "abstract": "Speech modeling methods learn one embedding for a fixed segment of speech, typically in between 10-25 ms. The information present in speech can be divided into two categories: \"what is being said\" (content) and \"how it is expressed\" (other) and these two are orthogonal in nature causing the optimization algorithm to find a sub-optimal solution if forced to optimize together. This leads to sub-optimal performance in one or all downstream tasks as shown by previous studies. Current self-supervised learning (SSL) methods such as HuBERT are very good at modeling the content information present in speech. Data augmentation improves the performance on tasks which require effective modeling of other information but this leads to a divided capacity of the model. In this work, we conduct a preliminary study to understand the importance of modeling other information using separate learnable parameters. We propose a modified version of HuBERT, termed Other HuBERT (O-HuBERT), to test our hypothesis. Our findings are twofold: first, the O-HuBERT method is able to utilize all layers to build complex features to encode other information; second, a robust data augmentation strategy is essential for learning the information required by tasks that depend on other information and to achieve state-of-the-art (SOTA) performance on the SUPERB benchmark with a similarly sized model (100 million parameters) and pre-training data (960 hours)."
                    },
                    {
                        "title": "GLUECoS : An Evaluation Benchmark for Code-Switched NLP",
                        "abstract": "Code-switching is the use of more than one language in the same conversation or utterance. Recently, multilingual contextual embedding models, trained on multiple monolingual corpora, have shown promising results on cross-lingual and multilingual tasks. We present an evaluation benchmark, GLUECoS, for code-switched languages, that spans several NLP tasks in English-Hindi and English-Spanish. Specifically, our evaluation benchmark includes Language Identification from text, POS tagging, Named Entity Recognition, Sentiment Analysis, Question Answering and a new task for code-switching, Natural Language Inference. We present results on all these tasks using cross-lingual word embedding models and multilingual models. In addition, we fine-tune multilingual models on artificially generated code-switched data. Although multilingual models perform significantly better than cross-lingual models, our results show that in most tasks, across both language pairs, multilingual models fine-tuned on code-switched data perform best, showing that multilingual models can be further optimized for code-switching tasks."
                    },
                    {
                        "title": "DiTTO: A Feature Representation Imitation Approach for Improving Cross-Lingual Transfer",
                        "abstract": "Zero-shot cross-lingual transfer is promising, however has been shown to be sub-optimal, with inferior transfer performance across low-resource languages. In this work, we envision languages as domains for improving zero-shot transfer by jointly reducing the feature incongruity between the source and the target language and increasing the generalization capabilities of pre-trained multilingual transformers. We show that our approach, DiTTO, significantly outperforms the standard zero-shot fine-tuning method on multiple datasets across all languages using solely unlabeled instances in the target language. Empirical results show that jointly reducing feature incongruity for multiple target languages is vital for successful cross-lingual transfer. Moreover, our model enables better cross-lingual transfer than standard fine-tuning methods, even in the few-shot setting."
                    },
                    {
                        "title": "Representativeness as a Forgotten Lesson for Multilingual and Code-switched Data Collection and Preparation",
                        "abstract": "Multilingualism is widespread around the world and code-switching (CSW) is a common practice among different language pairs/tuples across locations and regions. However, there is still not much progress in building successful CSW systems, despite the recent advances in Massive Multilingual Language Models (MMLMs). We investigate the reasons behind this setback through a critical study about the existing CSW data sets (68) across language pairs in terms of the collection and preparation (e.g. transcription and annotation) stages. This in-depth analysis reveals that \\textbf{a)} most CSW data involves English ignoring other language pairs/tuples \\textbf{b)} there are flaws in terms of representativeness in data collection and preparation stages due to ignoring the location based, socio-demographic and register variation in CSW. In addition, lack of clarity on the data selection and filtering stages shadow the representativeness of CSW data sets. We conclude by providing a short check-list to improve the representativeness for forthcoming studies involving CSW data collection and preparation."
                    }
                ]
            },
            "35956d97-78d3-4f57-acd7-7638e0cfa05b": {
                "pk": "35956d97-78d3-4f57-acd7-7638e0cfa05b",
                "name": "Xing Xie",
                "collaborators": [
                    "Jianxun Lian",
                    "Zheng Liu",
                    "Yu Xing",
                    "Defu Lian",
                    "Ziyao Li",
                    "Yilan Cui",
                    "Yi Liu"
                ],
                "domain": [
                    "User Identity",
                    "Recommendation Systems",
                    "Human Mobility Analysis",
                    "Social Media Data"
                ],
                "publications": [
                    {
                        "title": "Cross-Device User Matching Based on Massive Browse Logs: The Runner-Up Solution for the 2016 CIKM Cup",
                        "abstract": "As the number and variety of smart devices increase, users may use myriad devices in their daily lives and the online activities become highly fragmented. Building an accurate user identity becomes a difficult and important problem for advertising companies. The task for the CIKM Cup 2016 Track 1 was to find the same user cross multiple devices. This paper discusses our solution to the challenge. It is mainly comprised of three parts: comprehensive feature engineering, negative sampling, and model selection. For each part we describe our special steps and demonstrate how the performance is boosted. We took the second prize of the competition with an F1-score of 0.41669."
                    },
                    {
                        "title": "A Novel User Representation Paradigm for Making Personalized Candidate Retrieval",
                        "abstract": "Candidate retrieval is a fundamental issue in recommendation system. Given user's recommendation request, relevant candidates need to be retrieved in realtime for subsequent ranking operations. Considering that the retrieval operation is conducted over considerable items, it has to be both precise and scalable so that high-quality candidates can be acquired within tolerable latency. Unfortunately, conventional methods would trade off precision for high running efficiency, which leads to inferior retrieval quality. In contrast, those deep learning-based approaches can be highly accurate in identifying relevant items; yet, they are unsuitable for candidate retrieval due to their inherent limitation on scalability.   In this work, a novel framework is proposed to address the above challenges. The underlying intuition is to rely on a well-trained ranking model for the supervision of an efficient retrieval model, such that it will unify the scalability and precision as a whole. We have implemented our conceptual framework and made comprehensive evaluation for it, where promising results are achieved against representative baselines.   Our work is undergoing a anonymous review, and it will soon be released after the notification. If you're also interested in this problem, please feel free to contact us."
                    },
                    {
                        "title": "Social media and mobility landscape: uncovering spatial patterns of urban human mobility with multi source data",
                        "abstract": "In this paper, we present a three-step methodological framework, including location identification, bias modification, and out-of-sample validation, so as to promote human mobility analysis with social media data. More specifically, we propose ways of identifying personal activity-specific places and commuting patterns in Beijing, China, based on Weibo (China's Twitter) check-in records, as well as modifying sample bias of check-in data with population synthesis technique. An independent citywide travel logistic survey is used as the benchmark for validating the results. Obvious differences are discerned from Weibo users' and survey respondents' activity-mobility patterns, while there is a large variation of population representativeness between data from the two sources. After bias modification, the similarity coefficient between commuting distance distributions of Weibo data and survey observations increases substantially from 23% to 63%. Synthetic data proves to be a satisfactory cost-effective alternative source of mobility information. The proposed framework can inform many applications related to human mobility, ranging from transportation, through urban planning to transport emission modelling."
                    }
                ]
            }
        }
    },
    "2305.19394": {
        "paper_data": {
            "title": "Synaptic Weight Distributions Depend on the Geometry of Plasticity",
            "url": "http://arxiv.org/abs/2305.19394v2",
            "arxiv_id": "2305.19394",
            "authors": [
                "Roman Pogodin",
                "Jonathan Cornford",
                "Arna Ghosh",
                "Gauthier Gidel",
                "Guillaume Lajoie",
                "Blake Richards"
            ],
            "abstract": "A growing literature in computational neuroscience leverages gradient descent and learning algorithms that approximate it to study synaptic plasticity in the brain. However, the vast majority of this work ignores a critical underlying assumption: the choice of distance for synaptic changes - i.e. the geometry of synaptic plasticity. Gradient descent assumes that the distance is Euclidean, but many other distances are possible, and there is no reason that biology necessarily uses Euclidean geometry. Here, using the theoretical tools provided by mirror descent, we show that the distribution of synaptic weights will depend on the geometry of synaptic plasticity. We use these results to show that experimentally-observed log-normal weight distributions found in several brain areas are not consistent with standard gradient descent (i.e. a Euclidean geometry), but rather with non-Euclidean distances. Finally, we show that it should be possible to experimentally test for different synaptic geometries by comparing synaptic weight distributions before and after learning. Overall, our work shows that the current paradigm in theoretical work on synaptic plasticity that assumes Euclidean synaptic geometry may be misguided and that it should be possible to experimentally determine the true geometry of synaptic plasticity in the brain.",
            "introduction": "   1 Introduction  Many computational neuroscience studies use gradient descent to train models that are then compared to the brain Schrimpf et\u00a0al. (2018); Nayebi et\u00a0al. (2018); Yamins et\u00a0al. (2014); Bakhtiari et\u00a0al. (2021); Flesch et\u00a0al. (2022), and many others explore how synaptic plasticity could approximate gradient descent Richards et\u00a0al. (2019). One aspect of this framework that is often not explicitly considered is that in order to follow the gradient of a loss in the synaptic weight space, one must have a means of measuring distance in the synaptic space, i.e. of determining what constitutes a large versus a small change in the weights Carlo\u00a0Surace et\u00a0al. (2018). In other words, whenever we build a neural network model we are committing to a synaptic geometry. Standard gradient descent assumes Euclidean geometry, meaning that distance in synaptic space is equivalent to the L2-norm of weight changes.   There are however other distances that can be used. For example, in natural gradient descent the Kullback-Leibler divergence of the network\u2019s output distributions is used as the distance Martens (2020). More broadly, the theory of mirror descent provides tools for analyzing and building algorithms that use different distances in parameter space Nemirovskij & Yudin (1983); Beck & Teboulle (2003), which has proven useful in a variety of applications Shalev-Shwartz et\u00a0al. (2012); Bubeck et\u00a0al. (2015); Lattimore & Gyorgy (2021). However, within computational neuroscience the question of synaptic geometry is often overlooked: most models use Euclidean geometry without considering other options Carlo\u00a0Surace et\u00a0al. (2018). This assumption has no basis in neuroscience data, so how could we possibly determine the synaptic geometry used by the brain?   Here, using tools from mirror descent Nemirovskij & Yudin (1983); Beck & Teboulle (2003), we show that synaptic geometry can be determined by observing the distribution of synaptic weight changes during learning. Specifically, we prove that in situations where synaptic changes are relatively small the distribution of synaptic weights depends on the synaptic geometry (with mild assumptions about the loss function and the dataset). We use this result to show that the geometry defines a dual space in which the total synaptic changes are Gaussian. As a result, if one can find a dual space in which experimentally observed synaptic changes are Gaussian, then one knows the synaptic geometry. Applying this framework to existing neural data, which suggests that synaptic weights are log-normally distributed Song et\u00a0al. (2005); Loewenstein et\u00a0al. (2011); Melander et\u00a0al. (2021); Buzs\u00e1ki & Mizuseki (2014), we conclude that the brain is unlikely to use a Euclidean synaptic geometry. Moreover, we show how to use our findings to make experimental predictions. In particular, we show that it should be possible to use experimentally observed weight distributions before and after learning to rule out different candidate geometries. Altogether, our work provides a novel theoretical insight for reasoning about the learning algorithms of the brain.    1.1 Related work  There is a large and growing literature on approximating gradient-based learning in the brain Lillicrap et\u00a0al. (2016); Liao et\u00a0al. (2016); Akrout et\u00a0al. (2019); Podlaski & Machens (2020); Clark et\u00a0al. (2021). The vast majority of this work assumes Euclidean synaptic geometry Carlo\u00a0Surace et\u00a0al. (2018). Notably, even if the brain does not estimate gradients directly, as long as synaptic weight updates are relatively small,",
            "references": [
                {
                    "title": "FFCV: Accelerating Training by Removing Data Bottlenecks",
                    "abstract": "We present FFCV, a library for easy and fast machine learning model training. FFCV speeds up model training by eliminating (often subtle) data bottlenecks from the training process. In particular, we combine techniques such as an efficient file storage format, caching, data pre-loading, asynchronous data transfer, and just-in-time compilation to (a) make data loading and transfer significantly more efficient, ensuring that GPUs can reach full utilization; and (b) offload as much data processing as possible to the CPU asynchronously, freeing GPU cycles for training. Using FFCV, we train ResNet-18 and ResNet-50 on the ImageNet dataset with a state-of-the-art tradeoff between accuracy and training time. For example, across the range of ResNet-50 models we test, we obtain the same accuracy as the best baselines in half the time. We demonstrate FFCV's performance, ease-of-use, extensibility, and ability to adapt to resource constraints through several case studies. Detailed installation instructions, documentation, and Slack support channel are available at https://ffcv.io/."
                },
                {
                    "title": "The study of plasticity has always been about gradients",
                    "abstract": "The experimental study of learning and plasticity has always been driven by an implicit question: how can physiological changes be adaptive and improve performance? For example, in Hebbian plasticity only synapses from presynaptic neurons that were active are changed, avoiding useless changes. Similarly, in dopamine\u2010gated learning synapse changes depend on reward or lack thereof and do not change when everything is predictable. Within machine learning we can make the question of which changes are adaptive concrete: performance improves when changes correlate with the gradient of an objective function quantifying performance. This result is general for any system that improves through small changes. As such, physiology has always implicitly been seeking mechanisms that allow the brain to approximate gradients. Coming from this perspective we review the existing literature on plasticity\u2010related mechanisms, and we show how these mechanisms relate to gradient estimation. We argue that gradients are a unifying idea to explain the many facets of neuronal plasticity."
                },
                {
                    "title": "How to prepare your task head for finetuning",
                    "abstract": "In deep learning, transferring information from a pretrained network to a downstream task by finetuning has many benefits. The choice of task head plays an important role in fine-tuning, as the pretrained and downstream tasks are usually different. Although there exist many different designs for finetuning, a full understanding of when and why these algorithms work has been elusive. We analyze how the choice of task head controls feature adaptation and hence influences the downstream performance. By decomposing the learning dynamics of adaptation, we find that the key aspect is the training accuracy and loss at the beginning of finetuning, which determines the\"energy\"available for the feature's adaptation. We identify a significant trend in the effect of changes in this initial energy on the resulting features after fine-tuning. Specifically, as the energy increases, the Euclidean and cosine distances between the resulting and original features increase, while their dot products (and the resulting features' norm) first increase and then decrease. Inspired by this, we give several practical principles that lead to better downstream performance. We analytically prove this trend in an overparamterized linear setting and verify its applicability to different experimental settings."
                },
                {
                    "title": "Making Look-Ahead Active Learning Strategies Feasible with Neural Tangent Kernels",
                    "abstract": "We propose a new method for approximating active learning acquisition strategies that are based on retraining with hypothetically-labeled candidate data points. Although this is usually infeasible with deep networks, we use the neural tangent kernel to approximate the result of retraining, and prove that this approximation works asymptotically even in an active learning setup -- approximating\"look-ahead\"selection criteria with far less computation required. This also enables us to conduct sequential active learning, i.e. updating the model in a streaming regime, without needing to retrain the model with SGD after adding each new data point. Moreover, our querying strategy, which better understands how the model's predictions will change by adding new data points in comparison to the standard (\"myopic\") criteria, beats other look-ahead strategies by large margins, and achieves equal or better performance compared to state-of-the-art methods on several benchmark datasets in pool-based active learning."
                },
                {
                    "title": "A theory of learning with constrained weight-distribution",
                    "abstract": "A central question in computational neuroscience is how structure determines function in neural networks. The emerging high-quality large-scale connectomic datasets raise the question of what general functional principles can be gleaned from structural information such as the distribution of excitatory/inhibitory synapse types and the distribution of synaptic weights. Motivated by this question, we developed a statistical mechanical theory of learning in neural networks that incorporates structural information as constraints. We derived an analytical solution for the memory capacity of the perceptron, a basic feedforward model of supervised learning, with constraint on the distribution of its weights. Our theory predicts that the reduction in capacity due to the constrained weight-distribution is related to the Wasserstein distance between the imposed distribution and that of the standard normal distribution. To test the theoretical predictions, we use optimal transport theory and information geometry to develop an SGD-based algorithm to find weights that simultaneously learn the input-output task and satisfy the distribution constraint. We show that training in our algorithm can be interpreted as geodesic flows in the Wasserstein space of probability distributions. We further developed a statistical mechanical theory for teacher-student perceptron rule learning and ask for the best way for the student to incorporate prior knowledge of the rule. Our theory shows that it is beneficial for the learner to adopt different prior weight distributions during learning, and shows that distribution-constrained learning outperforms unconstrained and sign-constrained learning. Our theory and algorithm provide novel strategies for incorporating prior knowledge about weights into learning, and reveal a powerful connection between structure and function in neural networks."
                },
                {
                    "title": "Self-consistent dynamical field theory of kernel evolution in wide neural networks",
                    "abstract": "We analyze feature learning in infinite-width neural networks trained with gradient flow through a self-consistent dynamical field theory. We construct a collection of deterministic dynamical order parameters which are inner-product kernels for hidden unit activations and gradients in each layer at pairs of time points, providing a reduced description of network activity through training. These kernel order parameters collectively define the hidden layer activation distribution, the evolution of the neural tangent kernel (NTK), and consequently, output predictions. We show that the field theory derivation recovers the recursive stochastic process of infinite-width feature learning networks obtained by Yang and Hu with tensor programs. For deep linear networks, these kernels satisfy a set of algebraic matrix equations. For nonlinear networks, we provide an alternating sampling procedure to self-consistently solve for the kernel order parameters. We provide comparisons of the self-consistent solution to various approximation schemes including the static NTK approximation, gradient independence assumption, and leading order perturbation theory, showing that each of these approximations can break down in regimes where general self-consistent solutions still provide an accurate description. Lastly, we provide experiments in more realistic settings which demonstrate that the loss and kernel dynamics of convolutional neural networks at fixed feature learning strength are preserved across different widths on a image classification task."
                },
                {
                    "title": "Powerpropagation: A sparsity inducing weight reparameterisation",
                    "abstract": "The training of sparse neural networks is becoming an increasingly important tool for reducing the computational footprint of models at training and evaluation, as well enabling the effective scaling up of models. Whereas much work over the years has been dedicated to specialised pruning techniques, little attention has been paid to the inherent effect of gradient based training on model sparsity. In this work, we introduce Powerpropagation, a new weight-parameterisation for neural networks that leads to inherently sparse models. Exploiting the behaviour of gradient descent, our method gives rise to weight updates exhibiting a\"rich get richer\"dynamic, leaving low-magnitude parameters largely unaffected by learning. Models trained in this manner exhibit similar performance, but have a distribution with markedly higher density at zero, allowing more parameters to be pruned safely. Powerpropagation is general, intuitive, cheap and straight-forward to implement and can readily be combined with various other techniques. To highlight its versatility, we explore it in two very different settings: Firstly, following a recent line of work, we investigate its effect on sparse training for resource-constrained settings. Here, we combine Powerpropagation with a traditional weight-pruning technique as well as recent state-of-the-art sparse-to-sparse algorithms, showing superior performance on the ImageNet benchmark. Secondly, we advocate the use of sparsity in overcoming catastrophic forgetting, where compressed representations allow accommodating a large number of tasks at fixed model capacity. In all cases our reparameterisation considerably increases the efficacy of the off-the-shelf methods."
                },
                {
                    "title": "The functional specialization of visual cortex emerges from training parallel pathways with self-supervised predictive learning",
                    "abstract": "The visual system of mammals is comprised of parallel, hierarchical specialized pathways. Different pathways are specialized in so far as they use representations that are more suitable for supporting specific downstream behaviours. In particular, the clearest example is the specialization of the ventral (\u201cwhat\u201d) and dorsal (\u201cwhere\u201d) pathways of the visual cortex. These two pathways support behaviours related to visual recognition and movement, respectively. To-date, deep neural networks have mostly been used as models of the ventral, recognition pathway. However, it is unknown whether both pathways can be modelled with a single deep ANN. Here, we ask whether a single model with a single loss function can capture the properties of both the ventral and the dorsal pathways. We explore this question using data from mice, who like other mammals, have specialized pathways that appear to support recognition and movement behaviours. We show that when we train a deep neural network architecture with two parallel pathways using a self-supervised predictive loss function, we can outperform other models in fitting mouse visual cortex. Moreover, we can model both the dorsal and ventral pathways. These results demonstrate that a self-supervised predictive learning approach applied to parallel pathway architectures can account for some of the functional specialization seen in mammalian visual systems."
                },
                {
                    "title": "Credit Assignment Through Broadcasting a Global Error Vector",
                    "abstract": "Backpropagation (BP) uses detailed, unit-specific feedback to train deep neural networks (DNNs) with remarkable success. That biological neural circuits appear to perform credit assignment, but cannot implement BP, implies the existence of other powerful learning algorithms. Here, we explore the extent to which a globally broadcast learning signal, coupled with local weight updates, enables training of DNNs. We present both a learning rule, called global error-vector broadcasting (GEVB), and a class of DNNs, called vectorized nonnegative networks (VNNs), in which this learning rule operates. VNNs have vector-valued units and nonnegative weights past the first layer. The GEVB learning rule generalizes three-factor Hebbian learning, updating each weight by an amount proportional to the inner product of the presynaptic activation and a globally broadcast error vector when the postsynaptic unit is active. We prove that these weight updates are matched in sign to the gradient, enabling accurate credit assignment. Moreover, at initialization, these updates are exactly proportional to the gradient in the limit of infinite network width. GEVB matches the performance of BP in VNNs, and in some cases outperforms direct feedback alignment (DFA) applied in conventional networks. Unlike DFA, GEVB successfully trains convolutional layers. Altogether, our theoretical and empirical results point to a surprisingly powerful role for a global learning signal in training DNNs."
                },
                {
                    "title": "Understanding deep learning (still) requires rethinking generalization",
                    "abstract": "Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small gap between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family or to the regularization techniques used during training. Through extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. We interpret our experimental findings by comparison with traditional models. We supplement this republication with a new section at the end summarizing recent progresses in the field since the original version of this paper."
                },
                {
                    "title": "Exploring Simple Siamese Representation Learning",
                    "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code is made available.1"
                },
                {
                    "title": "Mirror Descent and the Information Ratio",
                    "abstract": "We establish a connection between the stability of mirror descent and the information ratio by Russo and Van Roy [2014]. Our analysis shows that mirror descent with suitable loss estimators and exploratory distributions enjoys the same bound on the adversarial regret as the bounds on the Bayesian regret for information-directed sampling. Along the way, we develop the theory for information-directed sampling and provide an efficient algorithm for adversarial bandits for which the regret upper bound matches exactly the best known information-theoretic upper bound."
                },
                {
                    "title": "High-Dimensional Probability: An Introduction with Applications in Data Science",
                    "abstract": "\u00a9 2018, Cambridge University Press Let us summarize our findings. A random projection of a set T in R n onto an m-dimensional subspace approximately preserves the geometry of T if m \u2a86 d ( T ) . For..."
                },
                {
                    "title": "Biological credit assignment through dynamic inversion of feedforward networks",
                    "abstract": "Learning depends on changes in synaptic connections deep inside the brain. In multilayer networks, these changes are triggered by error signals fed back from the output, generally through a stepwise inversion of the feedforward processing steps. The gold standard for this process -- backpropagation -- works well in artificial neural networks, but is biologically implausible. Several recent proposals have emerged to address this problem, but many of these biologically-plausible schemes are based on learning an independent set of feedback connections. This complicates the assignment of errors to each synapse by making it dependent upon a second learning problem, and by fitting inversions rather than guaranteeing them. Here, we show that feedforward network transformations can be effectively inverted through dynamics. We derive this dynamic inversion from the perspective of feedback control, where the forward transformation is reused and dynamically interacts with fixed or random feedback to propagate error signals during the backward pass. Importantly, this scheme does not rely upon a second learning problem for feedback because accurate inversion is guaranteed through the network dynamics. We map these dynamics onto generic feedforward networks, and show that the resulting algorithm performs well on several supervised and unsupervised datasets. We also link this dynamic inversion to Gauss-Newton optimization, suggesting a biologically-plausible approximation to second-order learning. Overall, our work introduces an alternative perspective on credit assignment in the brain, and proposes a special role for temporal dynamics and feedback control during learning."
                },
                {
                    "title": "Bandit Algorithms",
                    "abstract": "sets of environments and policies respectively and ` : E \u00d7\u03a0\u2192 [0, 1] a bounded loss function. Given a policy \u03c0 let `(\u03c0) = (`(\u03bd1, \u03c0), . . . , `(\u03bdN , \u03c0)) be the loss vector resulting from policy \u03c0. Define S = {`(\u03c0) : \u03c0 \u2208 \u03a0} and \u03bb(S) = {x \u2208 cl(S) : y 6< x for all y \u2208 S} , where y 6< x is defined to mean it is not true that yi \u2264 xi for all i with strict inequality for at least one i. Prove that if \u03bb(S) \u2286 S and S is convex, then for each x \u2208 \u03bb(S) there exists a prior q \u2208 P(E) and policy \u03c0\u2217 such that `(\u03c0) = x and \u2211 \u03bd\u2208E q(\u03bd)`(\u03bd, \u03c0\u2217) = min \u03c0\u2208\u03a0 \u2211"
                },
                {
                    "title": "Kernelized information bottleneck leads to biologically plausible 3-factor Hebbian learning in deep networks",
                    "abstract": "The state-of-the art machine learning approach to training deep neural networks, backpropagation, is implausible for real neural networks: neurons need to know their outgoing weights; training alternates between a forward pass (computation) and a backward pass (learning); and the algorithm needs a large amount of labeled data. Biologically plausible approximations to backpropagation, such as feedback alignment, solve the weight transport problem, but not the other two. Thus, fully biologically plausible learning rules have so far remained elusive. Here we present a family of learning rules that does not suffer from any of these problems. It is motivated by the information bottleneck principle (extended with kernel methods), in which networks learn to squeeze as much information as possible out of the input without sacrificing prediction of the output. The resulting rules have a 3-factor Hebbian structure: they require pre- and post-synaptic firing rates and a global error signal - the third factor - that can be supplied by a neuromodulator. Moreover, they do not require precise labels; instead, they rely on the similarity between the desired outputs. They thus solve all three implausibility issues of backpropagation. Moreover, to obtain good performance on hard problems and retain biologically plausible learning rules, our rules need divisive normalization - a known feature of biological networks. Finally, simulations show that our rule performs nearly as well as backpropagation on image classification tasks."
                },
                {
                    "title": "Binary and analog variation of synapses between cortical pyramidal neurons",
                    "abstract": "Learning from experience depends at least in part on changes in neuronal connections. We present the largest map of connectivity to date between cortical neurons of a defined type (L2/3 pyramidal cells), which was enabled by automated analysis of serial section electron microscopy images with improved handling of image defects. We used the map to identify constraints on the learning algorithms employed by the cortex. Previous cortical studies modeled a continuum of synapse sizes (Arellano et al., 2007) by a log-normal distribution (Loewenstein, Kuras and Rumpel, 2011; de Vivo et al., 2017; Santuy et al., 2018). A continuum is consistent with most neural network models of learning, in which synaptic strength is a continuously graded analog variable. Here we show that synapse size, when restricted to synapses between L2/3 pyramidal cells, is well-modeled by the sum of a binary variable and an analog variable drawn from a log-normal distribution. Two synapses sharing the same presynaptic and postsynaptic cells are known to be correlated in size (Sorra and Harris, 1993; Koester and Johnston, 2005; Bartol et al., 2015; Kasthuri et al., 2015; Dvorkin and Ziv, 2016; Bloss et al., 2018; Motta et al., 2019). We show that the binary variables of the two synapses are highly correlated, while the analog variables are not. Binary variation could be the outcome of a Hebbian or other synaptic plasticity rule depending on activity signals that are relatively uniform across neuronal arbors, while analog variation may be dominated by other influences. We discuss the implications for the stability-plasticity dilemma."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Brain-Like Object Recognition with High-Performing Shallow Recurrent ANNs",
                    "abstract": "Deep convolutional artificial neural networks (ANNs) are the leading class of candidate models of the mechanisms of visual processing in the primate ventral stream. While initially inspired by brain anatomy, over the past years, these ANNs have evolved from a simple eight-layer architecture in AlexNet to extremely deep and branching architectures, demonstrating increasingly better object categorization performance, yet bringing into question how brain-like they still are. In particular, typical deep models from the machine learning community are often hard to map onto the brain's anatomy due to their vast number of layers and missing biologically-important connections, such as recurrence. Here we demonstrate that better anatomical alignment to the brain and high performance on machine learning as well as neuroscience measures do not have to be in contradiction. We developed CORnet-S, a shallow ANN with four anatomically mapped areas and recurrent connectivity, guided by Brain-Score, a new large-scale composite of neural and behavioral benchmarks for quantifying the functional fidelity of models of the primate ventral visual stream. Despite being significantly shallower than most models, CORnet-S is the top model on Brain-Score and outperforms similarly compact models on ImageNet. Moreover, our extensive analyses of CORnet-S circuitry variants reveal that recurrence is the main predictive factor of both Brain-Score and ImageNet top-1 performance. Finally, we report that the temporal evolution of the CORnet-S \"IT\" neural population resembles the actual monkey IT population dynamics. Taken together, these results establish CORnet-S, a compact, recurrent ANN, as the current best model of the primate ventral visual stream."
                },
                {
                    "title": "Stochastic Mirror Descent on Overparameterized Nonlinear Models",
                    "abstract": "Most modern learning problems are highly overparameterized, i.e., have many more model parameters than the number of training data points. As a result, the training loss may have infinitely many global minima (parameter vectors that perfectly \u201cinterpolate\u201d the training data). It is therefore imperative to understand which interpolating solutions we converge to, how they depend on the initialization and learning algorithm, and whether they yield different test errors. In this article, we study these questions for the family of stochastic mirror descent (SMD) algorithms, of which stochastic gradient descent (SGD) is a special case. Recently, it has been shown that for overparameterized linear models, SMD converges to the closest global minimum to the initialization point, where closeness is in terms of the Bregman divergence corresponding to the potential function of the mirror descent. With appropriate initialization, this yields convergence to the minimum-potential interpolating solution, a phenomenon referred to as implicit regularization. On the theory side, we show that for sufficiently-overparameterized nonlinear models, SMD with a (small enough) fixed step size converges to a global minimum that is \u201cvery close\u201d (in Bregman divergence) to the minimum-potential interpolating solution, thus attaining approximate implicit regularization. On the empirical side, our experiments on the MNIST and CIFAR-10 datasets consistently confirm that the above phenomenon occurs in practical scenarios. They further indicate a clear difference in the generalization performances of different SMD algorithms: experiments on the CIFAR-10 dataset with different regularizers, <inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{1}$ </tex-math></inline-formula> to encourage sparsity, <inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{2}$ </tex-math></inline-formula> (SGD) to encourage small Euclidean norm, and <inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{\\infty }$ </tex-math></inline-formula> to discourage large components, surprisingly show that the <inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{\\infty }$ </tex-math></inline-formula> norm consistently yields better generalization performance than SGD, which in turn generalizes better than the <inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{1}$ </tex-math></inline-formula> norm."
                },
                {
                    "title": "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks",
                    "abstract": "Convolutional Neural Networks (ConvNets) are commonly developed at a fixed resource budget, and then scaled up for better accuracy if more resources are available. In this paper, we systematically study model scaling and identify that carefully balancing network depth, width, and resolution can lead to better performance. Based on this observation, we propose a new scaling method that uniformly scales all dimensions of depth/width/resolution using a simple yet highly effective compound coefficient. We demonstrate the effectiveness of this method on scaling up MobileNets and ResNet. \nTo go even further, we use neural architecture search to design a new baseline network and scale it up to obtain a family of models, called EfficientNets, which achieve much better accuracy and efficiency than previous ConvNets. In particular, our EfficientNet-B7 achieves state-of-the-art 84.4% top-1 / 97.1% top-5 accuracy on ImageNet, while being 8.4x smaller and 6.1x faster on inference than the best existing ConvNet. Our EfficientNets also transfer well and achieve state-of-the-art accuracy on CIFAR-100 (91.7%), Flowers (98.8%), and 3 other transfer learning datasets, with an order of magnitude fewer parameters. Source code is at this https URL."
                },
                {
                    "title": "Deep Learning without Weight Transport",
                    "abstract": "Current algorithms for deep learning probably cannot run in the brain because they rely on weight transport, where forward-path neurons transmit their synaptic weights to a feedback path, in a way that is likely impossible biologically. An algorithm called feedback alignment achieves deep learning without weight transport by using random feedback weights, but it performs poorly on hard visual-recognition tasks. Here we describe two mechanisms - a neural circuit called a weight mirror and a modification of an algorithm proposed by Kolen and Pollack in 1994 - both of which let the feedback path learn appropriate synaptic weights quickly and accurately even in large networks, without weight transport or complex wiring.Tested on the ImageNet visual-recognition task, these mechanisms outperform both feedback alignment and the newer sign-symmetry method, and nearly match backprop, the standard algorithm of deep learning, which uses weight transport."
                },
                {
                    "title": "Wide neural networks of any depth evolve as linear models under gradient descent",
                    "abstract": "A longstanding goal in deep learning research has been to precisely characterize training and generalization. However, the often complex loss landscapes of neural networks (NNs) have made a theory of learning dynamics elusive. In this work, we show that for wide NNs the learning dynamics simplify considerably and that, in the infinite width limit, they are governed by a linear model obtained from the first-order Taylor expansion of the network around its initial parameters. Furthermore, mirroring the correspondence between wide Bayesian NNs and Gaussian processes (GPs), gradient-based training of wide NNs with a squared loss produces test set predictions drawn from a GP with a particular compositional kernel. While these theoretical results are only exact in the infinite width limit, we nevertheless find excellent empirical agreement between the predictions of the original network and those of the linearized version even for finite practically-sized networks. This agreement is robust across different architectures, optimization methods, and loss functions."
                },
                {
                    "title": "Exponentiated Gradient Meets Gradient Descent",
                    "abstract": "The (stochastic) gradient descent and the multiplicative update method are probably the most popular algorithms in machine learning. We introduce and study a new regularization which provides a unification of the additive and multiplicative updates. This regularization is derived from an hyperbolic analogue of the entropy function, which we call hypentropy. It is motivated by a natural extension of the multiplicative update to negative numbers. The hypentropy has a natural spectral counterpart which we use to derive a family of matrix-based updates that bridge gradient methods and the multiplicative method for matrices. While the latter is only applicable to positive semi-definite matrices, the spectral hypentropy method can naturally be used with general rectangular matrices. We analyze the new family of updates by deriving tight regret bounds. We study empirically the applicability of the new update for settings such as multiclass learning, in which the parameters constitute a general rectangular matrix."
                },
                {
                    "title": "On Lazy Training in Differentiable Programming",
                    "abstract": "In a series of recent theoretical works, it was shown that strongly over-parameterized neural networks trained with gradient-based methods could converge exponentially fast to zero training loss, with their parameters hardly varying. In this work, we show that this \"lazy training\" phenomenon is not specific to over-parameterized neural networks, and is due to a choice of scaling, often implicit, that makes the model behave as its linearization around the initialization, thus yielding a model equivalent to learning with positive-definite kernels. Through a theoretical analysis, we exhibit various situations where this phenomenon arises in non-convex optimization and we provide bounds on the distance between the lazy and linearized optimization paths. Our numerical experiments bring a critical note, as we observe that the performance of commonly used non-linear deep convolutional neural networks in computer vision degrades when trained in the lazy regime. This makes it unlikely that \"lazy training\" is behind the many successes of neural networks in difficult high dimensional tasks."
                },
                {
                    "title": "Brain-Score: Which Artificial Neural Network for Object Recognition is most Brain-Like?",
                    "abstract": "The internal representations of early deep artificial neural networks (ANNs) were found to be remarkably similar to the internal neural representations measured experimentally in the primate brain. Here we ask, as deep ANNs have continued to evolve, are they becoming more or less brain-like? ANNs that are most functionally similar to the brain will contain mechanisms that are most like those used by the brain. We therefore developed Brain-Score \u2013 a composite of multiple neural and behavioral benchmarks that score any ANN on how similar it is to the brain\u2019s mechanisms for core object recognition \u2013 and we deployed it to evaluate a wide range of state-of-the-art deep ANNs. Using this scoring system, we here report that: (1) DenseNet-169, CORnet-S and ResNet-101 are the most brain-like ANNs. There remains considerable variability in neural and behavioral responses that is not predicted by any ANN, suggesting that no ANN model has yet captured all the relevant mechanisms. (3) Extending prior work, we found that gains in ANN ImageNet performance led to gains on Brain-Score. However, correlation weakened at \u2265 70% top-1 ImageNet performance, suggesting that additional guidance from neuroscience is needed to make further advances in capturing brain mechanisms. (4) We uncovered smaller (i.e. less complex) ANNs that are more brain-like than many of the best-performing ImageNet models, which suggests the opportunity to simplify ANNs to better understand the ventral stream. The scoring system used here is far from complete. However, we propose that evaluating and tracking model-benchmark correspondences through a Brain-Score that is regularly updated with new brain data is an exciting opportunity: experimental benchmarks can be used to guide machine network evolution, and machine networks are mechanistic hypotheses of the brain\u2019s network and thus drive next experiments. To facilitate both of these, we release Brain-Score.org: a platform that hosts the neural and behavioral benchmarks, where ANNs for visual processing can be submitted to receive a Brain-Score and their rank relative to other models, and where new experimental data can be naturally incorporated."
                },
                {
                    "title": "Task-Driven Convolutional Recurrent Models of the Visual System",
                    "abstract": "Feed-forward convolutional neural networks (CNNs) are currently state-of-the-art for object classification tasks such as ImageNet. Further, they are quantitatively accurate models of temporally-averaged responses of neurons in the primate brain's visual system. However, biological visual systems have two ubiquitous architectural features not shared with typical CNNs: local recurrence within cortical areas, and long-range feedback from downstream areas to upstream areas. Here we explored the role of recurrence in improving classification performance. We found that standard forms of recurrence (vanilla RNNs and LSTMs) do not perform well within deep CNNs on the ImageNet task. In contrast, novel cells that incorporated two structural features, bypassing and gating, were able to boost task accuracy substantially. We extended these design principles in an automated search over thousands of model architectures, which identified novel local recurrent cells and long-range feedback connections useful for object recognition. Moreover, these task-optimized ConvRNNs matched the dynamics of neural activity in the primate visual system better than feedforward networks, suggesting a role for the brain's recurrent connections in performing difficult visual behaviors."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "Stochastic Gradient/Mirror Descent: Minimax Optimality and Implicit Regularization",
                    "abstract": "Stochastic descent methods (of the gradient and mirror varieties) have become increasingly popular in optimization. In fact, it is now widely recognized that the success of deep learning is not only due to the special deep architecture of the models, but also due to the behavior of the stochastic descent methods used, which play a key role in reaching \"good\" solutions that generalize well to unseen data. In an attempt to shed some light on why this is the case, we revisit some minimax properties of stochastic gradient descent (SGD) for the square loss of linear models---originally developed in the 1990's---and extend them to general stochastic mirror descent (SMD) algorithms for general loss functions and nonlinear models. In particular, we show that there is a fundamental identity which holds for SMD (and SGD) under very general conditions, and which implies the minimax optimality of SMD (and SGD) for sufficiently small step size, and for a general class of loss functions and general nonlinear models. We further show that this identity can be used to naturally establish other properties of SMD (and SGD), namely convergence and implicit regularization for over-parameterized linear models (in what is now being called the \"interpolating regime\"), some of which have been shown in certain cases in prior literature. We also argue how this identity can be used in the so-called \"highly over-parameterized\" nonlinear setting (where the number of parameters far exceeds the number of data points) to provide insights into why SMD (and SGD) may have similar convergence and implicit regularization properties for deep learning."
                },
                {
                    "title": "On the choice of metric in gradient-based theories of brain function",
                    "abstract": "This is a PLOS Computational Biology Education paper. The idea that the brain functions so as to minimize certain costs pervades theoretical neuroscience. Because a cost function by itself does not predict how the brain finds its minima, additional assumptions about the optimization method need to be made to predict the dynamics of physiological quantities. In this context, steepest descent (also called gradient descent) is often suggested as an algorithmic principle of optimization potentially implemented by the brain. In practice, researchers often consider the vector of partial derivatives as the gradient. However, the definition of the gradient and the notion of a steepest direction depend on the choice of a metric. Because the choice of the metric involves a large number of degrees of freedom, the predictive power of models that are based on gradient descent must be called into question, unless there are strong constraints on the choice of the metric. Here, we provide a didactic review of the mathematics of gradient descent, illustrate common pitfalls of using gradient descent as a principle of brain function with examples from the literature, and propose ways forward to constrain the metric."
                },
                {
                    "title": "Characterizing Implicit Bias in Terms of Optimization Geometry",
                    "abstract": "We study the implicit bias of generic optimization methods, such as mirror descent, natural gradient descent, and steepest descent with respect to different potentials and norms, when optimizing underdetermined linear regression or separable linear classification problems. We explore the question of whether the specific global minimum (among the many possible global minima) reached by an algorithm can be characterized in terms of the potential or norm of the optimization geometry, and independently of hyperparameter choices such as step-size and momentum."
                },
                {
                    "title": "Gaussian Process Behaviour in Wide Deep Neural Networks",
                    "abstract": "Whilst deep neural networks have shown great empirical success, there is still much work to be done to understand their theoretical properties. In this paper, we study the relationship between random, wide, fully connected, feedforward networks with more than one hidden layer and Gaussian processes with a recursive kernel definition. We show that, under broad conditions, as we make the architecture increasingly wide, the implied random function converges in distribution to a Gaussian process, formalising and extending existing results by Neal (1996) to deep networks. To evaluate convergence rates empirically, we use maximum mean discrepancy. We then compare finite Bayesian deep networks from the literature to Gaussian processes in terms of the key predictive quantities of interest, finding that in some cases the agreement can be very close. We discuss the desirability of Gaussian process behaviour and review non-Gaussian alternative models from the literature."
                },
                {
                    "title": "Implicit Regularization in Matrix Factorization",
                    "abstract": "We study implicit regularization when optimizing an underdetermined quadratic objective over a matrix $X$ with gradient descent on a factorization of X. We conjecture and provide empirical and theoretical evidence that with small enough step sizes and initialization close enough to the origin, gradient descent on a full dimensional factorization converges to the minimum nuclear norm solution."
                },
                {
                    "title": "Neuromodulated Spike-Timing-Dependent Plasticity, and Theory of Three-Factor Learning Rules",
                    "abstract": "Classical Hebbian learning puts the emphasis on joint pre- and postsynaptic activity, but neglects the potential role of neuromodulators. Since neuromodulators convey information about novelty or reward, the influence of neuromodulators on synaptic plasticity is useful not just for action learning in classical conditioning, but also to decide \u201cwhen\u201d to create new memories in response to a flow of sensory stimuli. In this review, we focus on timing requirements for pre- and postsynaptic activity in conjunction with one or several phasic neuromodulatory signals. While the emphasis of the text is on conceptual models and mathematical theories, we also discuss some experimental evidence for neuromodulation of Spike-Timing-Dependent Plasticity. We highlight the importance of synaptic mechanisms in bridging the temporal gap between sensory stimulation and neuromodulatory signals, and develop a framework for a class of neo-Hebbian three-factor learning rules that depend on presynaptic activity, postsynaptic variables as well as the influence of neuromodulators."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "How Important Is Weight Symmetry in Backpropagation?",
                    "abstract": "\n \n Gradient backpropagation (BP) requires symmetric feedforward and feedback connections \u2014 the same weights must be used for forward and backward passes. This \"weight transport problem'' (Grossberg 1987) is thought to be one of the main reasons to doubt BP's biologically plausibility. Using 15 different classification datasets, we systematically investigate to what extent BP really depends on weight symmetry. In a study that turned out to be surprisingly similar in spirit to Lillicrap et al.'s demonstration (Lillicrap et al. 2014) but orthogonal in its results, our experiments indicate that: (1) the magnitudes of feedback weights do not matter to performance (2) the signs of feedback weights do matter \u2014 the more concordant signs between feedforward and their corresponding feedback connections, the better (3) with feedback weights having random magnitudes and 100% concordant signs, we were able to achieve the same or even better performance than SGD. (4) some normalizations/stabilizations are indispensable for such asymmetric BP to work, namely Batch Normalization (BN) (Ioffe and Szegedy 2015) and/or a \"Batch Manhattan'' (BM) update rule.\n \n"
                },
                {
                    "title": "Five key factors determining pairwise correlations in visual cortex",
                    "abstract": "The responses of cortical neurons to repeated presentation of a stimulus are highly variable, yet correlated. These \u201cnoise correlations\u201d reflect a low-dimensional structure of population dynamics. Here, we examine noise correlations in 22,705 pairs of neurons in primary visual cortex (V1) of anesthetized cats, during ongoing activity and in response to artificial and natural visual stimuli. We measured how noise correlations depend on 11 factors. Because these factors are themselves not independent, we distinguished their influences using a nonlinear additive model. The model revealed that five key factors play a predominant role in determining pairwise correlations. Two of these are distance in cortex and difference in sensory tuning: these are known to decrease correlation. A third factor is firing rate: confirming most earlier observations, it markedly increased pairwise correlations. A fourth factor is spike width: cells with a broad spike were more strongly correlated amongst each other. A fifth factor is spike isolation: neurons with worse isolation were more correlated, even if they were recorded on different electrodes. For pairs of neurons with poor isolation, this last factor was the main determinant of correlations. These results were generally independent of stimulus type and timescale of analysis, but there were exceptions. For instance, pairwise correlations depended on difference in orientation tuning more during responses to gratings than to natural stimuli. These results consolidate disjoint observations in a vast literature on pairwise correlations and point towards regularities of population coding in sensory cortex."
                },
                {
                    "title": "New Insights and Perspectives on the Natural Gradient Method",
                    "abstract": "Natural gradient descent is an optimization method traditionally motivated from the perspective of information geometry, and works well for many applications as an alternative to stochastic gradient descent. In this paper we critically analyze this method and its properties, and show how it can be viewed as a type of approximate 2nd-order optimization method, where the Fisher information matrix can be viewed as an approximation of the Hessian. This perspective turns out to have significant implications for how to design a practical and robust version of the method. Additionally, we make the following contributions to the understanding of natural gradient and 2nd-order methods: a thorough analysis of the convergence speed of stochastic natural gradient descent (and more general stochastic 2nd-order methods) as applied to convex quadratics, a critical examination of the oft-used \"empirical\" approximation of the Fisher matrix, and an analysis of the (approximate) parameterization invariance property possessed by natural gradient methods, which we show still holds for certain choices of the curvature matrix other than the Fisher, but notably not the Hessian."
                },
                {
                    "title": "Convex Optimization: Algorithms and Complexity",
                    "abstract": "This monograph presents the main complexity theorems in convex optimization and their corresponding algorithms. Starting from the fundamental theory of black-box optimization, the material progresses towards recent advances in structural optimization and stochastic optimization. Our presentation of black-box optimization, strongly influenced by Nesterov's seminal book and Nemirovski's lecture notes, includes the analysis of cutting plane methods, as well as (accelerated) gradient descent schemes. We also pay special attention to non-Euclidean settings (relevant algorithms include Frank-Wolfe, mirror descent, and dual averaging) and discuss their relevance in machine learning. We provide a gentle introduction to structural optimization with FISTA (to optimize a sum of a smooth and a simple non-smooth term), saddle-point mirror prox (Nemirovski's alternative to Nesterov's smoothing), and a concise description of interior point methods. In stochastic optimization we discuss stochastic gradient descent, mini-batches, random coordinate descent, and sublinear algorithms. We also briefly touch upon convex relaxation of combinatorial problems and the use of randomness to round solutions, as well as random walks based methods."
                },
                {
                    "title": "Performance-optimized hierarchical models predict neural responses in higher visual cortex",
                    "abstract": "Significance Humans and monkeys easily recognize objects in scenes. This ability is known to be supported by a network of hierarchically interconnected brain areas. However, understanding neurons in higher levels of this hierarchy has long remained a major challenge in visual systems neuroscience. We use computational techniques to identify a neural network model that matches human performance on challenging object categorization tasks. Although not explicitly constrained to match neural data, this model turns out to be highly predictive of neural responses in both the V4 and inferior temporal cortex, the top two layers of the ventral visual hierarchy. In addition to yielding greatly improved models of visual cortex, these results suggest that a process of biological performance optimization directly shaped neural mechanisms. The ventral visual stream underlies key human visual object recognition abilities. However, neural encoding in the higher areas of the ventral stream remains poorly understood. Here, we describe a modeling approach that yields a quantitatively accurate model of inferior temporal (IT) cortex, the highest ventral cortical area. Using high-throughput computational techniques, we discovered that, within a class of biologically plausible hierarchical neural network models, there is a strong correlation between a model\u2019s categorization performance and its ability to predict individual IT neural unit response data. To pursue this idea, we then identified a high-performing neural network that matches human performance on a range of recognition tasks. Critically, even though we did not constrain this model to match neural data, its top output layer turns out to be highly predictive of IT spiking responses to complex naturalistic images at both the single site and population levels. Moreover, the model\u2019s intermediate layers are highly predictive of neural responses in the V4 cortex, a midlevel visual area that provides the dominant cortical input to IT. These results show that performance optimization\u2014applied in a biologically appropriate model class\u2014can be used to build quantitative predictive models of neural processing."
                },
                {
                    "title": "Spatial Profile of Excitatory and Inhibitory Synaptic Connectivity in Mouse Primary Auditory Cortex",
                    "abstract": "The role of local cortical activity in shaping neuronal responses is controversial. Among other questions, it is unknown how the diverse response patterns reported in vivo\u2014lateral inhibition in some cases, approximately balanced excitation and inhibition (co-tuning) in others\u2014compare to the local spread of synaptic connectivity. Excitatory and inhibitory activity might cancel each other out, or, whether one outweighs the other, receptive field properties might be substantially affected. As a step toward addressing this question, we used multiple intracellular recording in mouse primary auditory cortical slices to map synaptic connectivity among excitatory pyramidal cells and the two broad classes of inhibitory cells, fast-spiking (FS) and non-FS cells in the principal input layer. Connection probability was distance-dependent; the spread of connectivity, parameterized by Gaussian fits to the data, was comparable for all cell types, ranging from 85 to 114 \u03bcm. With brief stimulus trains, unitary synapses formed by FS interneurons were stronger than other classes of synapses; synapse strength did not correlate with distance between cells. The physiological data were qualitatively consistent with predictions derived from anatomical reconstruction. We also analyzed the truncation of neuronal processes due to slicing; overall connectivity was reduced but the spatial pattern was unaffected. The comparable spatial patterns of connectivity and relatively strong excitatory-inhibitory interconnectivity are consistent with a theoretical model where either lateral inhibition or co-tuning can predominate, depending on the structure of the input."
                },
                {
                    "title": "Online Learning and Online Convex Optimization",
                    "abstract": "Online learning is a well established learning paradigm which has both theoretical and practical appeals. The goal of online learning is to make a sequence of accurate predictions given knowledge of the correct answer to previous prediction tasks and possibly additional available information. Online learning has been studied in several research fields including game theory, information theory, and machine learning. It also became of great interest to practitioners due the recent emergence of large scale applications such as online advertisement placement and online web ranking. In this survey we provide a modern overview of online learning. Our goal is to give the reader a sense of some of the interesting ideas and in particular to underscore the centrality of convexity in deriving efficient online learning algorithms. We do not mean to be comprehensive but rather to give a high-level, rigorous yet easy to follow, survey."
                },
                {
                    "title": "Multiplicative Dynamics Underlie the Emergence of the Log-Normal Distribution of Spine Sizes in the Neocortex In Vivo",
                    "abstract": "What fundamental properties of synaptic connectivity in the neocortex stem from the ongoing dynamics of synaptic changes? In this study, we seek to find the rules shaping the stationary distribution of synaptic efficacies in the cortex. To address this question, we combined chronic imaging of hundreds of spines in the auditory cortex of mice in vivo over weeks with modeling techniques to quantitatively study the dynamics of spines, the morphological correlates of excitatory synapses in the neocortex. We found that the stationary distribution of spine sizes of individual neurons can be exceptionally well described by a log-normal function. We furthermore show that spines exhibit substantial volatility in their sizes at timescales that range from days to months. Interestingly, the magnitude of changes in spine sizes is proportional to the size of the spine. Such multiplicative dynamics are in contrast with conventional models of synaptic plasticity, learning, and memory, which typically assume additive dynamics. Moreover, we show that the ongoing dynamics of spine sizes can be captured by a simple phenomenological model that operates at two timescales of days and months. This model converges to a log-normal distribution, bridging the gap between synaptic dynamics and the stationary distribution of synaptic efficacies."
                },
                {
                    "title": "Information-Geometric Optimization Algorithms: A Unifying Picture via Invariance Principles",
                    "abstract": "We present a canonical way to turn any smooth parametric family of probability distributions on an arbitrary search space X into a continuous-time black-box optimization method on X, the information-geometric optimization (IGO) method. Invariance as a major design principle keeps the number of arbitrary choices to a minimum. The resulting IGO flow is the flow of an ordinary differential equation conducting the natural gradient ascent of an adaptive, time-dependent transformation of the objective function. It makes no particular assumptions on the objective function to be optimized. \n \nThe IGO method produces explicit IGO algorithms through time discretization. It naturally recovers versions of known algorithms and offers a systematic way to derive new ones. In continuous search spaces, IGO algorithms take a form related to natural evolution strategies (NES). The cross-entropy method is recovered in a particular case with a large time step, and can be extended into a smoothed, parametrization-independent maximum likelihood update (IGO-ML). When applied to the family of Gaussian distributions on Rd, the IGO framework recovers a version of the well-known CMA-ES algorithm and of xNES. For the family of Bernoulli distributions on {0, 1}d, we recover the seminal PBIL algorithm and cGA. For the distributions of restricted Boltzmann machines, we naturally obtain a novel algorithm for discrete optimization on {0, 1}d. All these algorithms are natural instances of, and unified under, the single information-geometric optimization framework. \n \nThe IGO method achieves, thanks to its intrinsic formulation, maximal invariance properties: invariance under reparametrization of the search space X, under a change of parameters of the probability distribution, and under increasing transformation of the function to be optimized. The latter is achieved through an adaptive, quantile-based formulation of the objective. \n \nTheoretical considerations strongly suggest that IGO algorithms are essentially characterized by a minimal change of the distribution over time. Therefore they have minimal loss in diversity through the course of optimization, provided the initial diversity is high. First experiments using restricted Boltzmann machines confirm this insight. As a simple consequence, IGO seems to provide, from information theory, an elegant way to simultaneously explore several valleys of a fitness landscape in a single run."
                },
                {
                    "title": "Composite objective mirror descent",
                    "abstract": "We present a new method for regularized convex optimization and analyze it under both online and stochastic optimization settings. In addition to unifying previously known firstorder algorithms, such as the projected gradient method, mirror descent, and forwardbackward splitting, our method yields new analysis and algorithms. We also derive specific instantiations of our method for commonly used regularization functions, such as l1, mixed norm, and trace-norm."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Highly Nonrandom Features of Synaptic Connectivity in Local Cortical Circuits",
                    "abstract": "How different is local cortical circuitry from a random network? To answer this question, we probed synaptic connections with several hundred simultaneous quadruple whole-cell recordings from layer 5 pyramidal neurons in the rat visual cortex. Analysis of this dataset revealed several nonrandom features in synaptic connectivity. We confirmed previous reports that bidirectional connections are more common than expected in a random network. We found that several highly clustered three-neuron connectivity patterns are overrepresented, suggesting that connections tend to cluster together. We also analyzed synaptic connection strength as defined by the peak excitatory postsynaptic potential amplitude. We found that the distribution of synaptic connection strength differs significantly from the Poisson distribution and can be fitted by a lognormal distribution. Such a distribution has a heavier tail and implies that synaptic weight is concentrated among few synaptic connections. In addition, the strengths of synaptic connections sharing pre- or postsynaptic neurons are correlated, implying that strong connections are even more clustered than the weak ones. Therefore, the local cortical network structure can be viewed as a skeleton of stronger connections in a sea of weaker ones. Such a skeleton is likely to play an important role in network dynamics and should be investigated further."
                },
                {
                    "title": "Natural Gradient Works Efficiently in Learning",
                    "abstract": "When a parameter space has a certain underlying structure, the ordinary gradient of a function does not represent its steepest direction, but the natural gradient does. Information geometry is used for calculating the natural gradients in the parameter space of perceptrons, the space of matrices (for blind source separation), and the space of linear dynamical systems (for blind source deconvolution). The dynamical behavior of natural gradient online learning is analyzed and is proved to be Fisher efficient, implying that it has asymptotically the same performance as the optimal batch estimation of parameters. This suggests that the plateau phenomenon, which appears in the backpropagation learning algorithm of multilayer perceptrons, might disappear or might not be so serious when the natural gradient is used. An adaptive method of updating the learning rate is proposed and analyzed."
                },
                {
                    "title": "Exponentiated Gradient Versus Gradient Descent for Linear Predictors",
                    "abstract": "We consider two algorithm for on-line prediction based on a linear model. The algorithms are the well-known Gradient Descent (GD) algorithm and a new algorithm, which we call EG(+/-). They both maintain a weight vector using simple updates. For the GD algorithm, the update is based on subtracting the gradient of the squared error made on a prediction. The EG(+/-) algorithm uses the components of the gradient in the exponents of factors that are used in updating the weight vector multiplicatively. We present worst-case loss bounds for EG(+/-) and compare them to previously known bounds for the GD algorithm. The bounds suggest that the losses of the algorithms are in general incomparable, but EG(+/-) has a much smaller loss if only a few components of the input are relevant for the predictions. We have performed experiments, which show that our worst-case upper bounds are quite tight already on simple artificial data."
                },
                {
                    "title": "The statistics of natural images",
                    "abstract": "Recently there has been a resurgence of interest in the properties of natural images. Their statistics are important not only in image compression but also for the study of sensory processing in biology, which can be viewed as satisfying certain \u2018design criteria\u2019. This review summarizes previous work on image statistics and presents our own data. Perhaps the most notable property of natural images is an invariance to scale. We present data to support this claim as well as evidence for a hierarchical invariance in natural scenes. These symmetries provide a powerful description of natural images as they greatly restrict the class of allowed distributions."
                },
                {
                    "title": "Review: P. Billingsley, Convergence of probability measures",
                    "abstract": "The author's preface gives an outline: \"This book is about weakconvergence methods in metric spaces, with applications sufficient to show their power and utility. The Introduction motivates the definitions and indicates how the theory will yield solutions to problems arising outside it. Chapter 1 sets out the basic general theorems, which are then specialized in Chapter 2 to the space C[0, l ] of continuous functions on the unit interval and in Chapter 3 to the space D [0, 1 ] of functions with discontinuities of the first kind. The results of the first three chapters are used in Chapter 4 to derive a variety of limit theorems for dependent sequences of random variables. \" The book develops and expands on Donsker's 1951 and 1952 papers on the invariance principle and empirical distributions. The basic random variables remain real-valued although, of course, measures on C[0, l ] and D[0, l ] are vitally used. Within this framework, there are various possibilities for a different and apparently better treatment of the material. More of the general theory of weak convergence of probabilities on separable metric spaces would be useful. Metrizability of the convergence is not brought up until late in the Appendix. The close relation of the Prokhorov metric and a metric for convergence in probability is (hence) not mentioned (see V. Strassen, Ann. Math. Statist. 36 (1965), 423-439; the reviewer, ibid. 39 (1968), 1563-1572). This relation would illuminate and organize such results as Theorems 4.1, 4.2 and 4.4 which give isolated, ad hoc connections between weak convergence of measures and nearness in probability. In the middle of p. 16, it should be noted that C*(S) consists of signed measures which need only be finitely additive if 5 is not compact. On p. 239, where the author twice speaks of separable subsets having nonmeasurable cardinal, he means \"discrete\" rather than \"separable.\" Theorem 1.4 is Ulam's theorem that a Borel probability on a complete separable metric space is tight. Theorem 1 of Appendix 3 weakens completeness to topological completeness. After mentioning that probabilities on the rationals are tight, the author says it is an"
                },
                {
                    "title": "Central Limit Theorems for Interchangeable Processes",
                    "abstract": "Let {Xn} (n = 1, 2 , \u2026) be a stochastic process. The random variables comprising it or the process itself will be said to be interchangeable if, for any choice of distinct positive integers i 1, i 2, H 3 \u2026 , ik, the joint distribution of depends merely on k and is independent of the integers i 1, i 2, \u2026 , i k. It was shown by De Finetti (3) that the probability measure for any interchangeable process is a mixture of probability measures of processes each consisting of independent and identically distributed random variables."
                }
            ],
            "categories": [
                "q-bio.NC",
                "cs.LG",
                "cs.NE"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we determine the synaptic geometry used by the brain during learning, and what implications does this have for understanding neural network training?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it challenges the prevailing assumption that synaptic geometry is Euclidean, which could lead to a paradigm shift in how we model learning in the brain. By accurately identifying the synaptic geometry, we can refine our understanding of neural learning algorithms, potentially leading to more biologically plausible models in computational neuroscience. This advancement could also inspire new practical applications in artificial intelligence, where insights from biological learning processes can inform the development of more efficient learning algorithms.\n\n**[Question 3] - Why is it hard?**  \nDetermining synaptic geometry is complex due to the lack of direct measurements of synaptic weight changes and the inherent variability in biological systems. Naive approaches that assume a fixed Euclidean geometry may fail to capture the true nature of synaptic changes, leading to inaccurate models. The technical challenges include the need for sophisticated statistical methods to analyze weight distributions and the theoretical obstacles of establishing a clear relationship between observed data and the underlying geometry. Additionally, the assumptions about loss functions and datasets must be carefully considered to ensure valid conclusions.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely overlooked the question of synaptic geometry, primarily focusing on gradient-based learning under the assumption of Euclidean space. This gap exists due to a lack of theoretical frameworks that connect synaptic weight distributions to geometric properties. Barriers include the difficulty in obtaining experimental data that accurately reflects synaptic changes and the absence of methodologies to analyze these changes in a geometrically informed manner. Our approach differs by utilizing tools from mirror descent to establish a framework for analyzing synaptic geometry based on empirical data, thus providing a novel perspective that has not been previously explored.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using mirror descent to analyze the distribution of synaptic weight changes during learning. We will apply this framework to existing neural data, specifically focusing on the log-normal distribution of synaptic weights. The key metrics will include the Gaussian nature of total synaptic changes in a dual space defined by the identified geometry. We expect to demonstrate that the brain does not operate under a Euclidean synaptic geometry and to make experimental predictions that can help rule out various"
            }
        },
        "author_data": {
            "deab83d3-19df-4501-ac70-86946cd37808": {
                "pk": "deab83d3-19df-4501-ac70-86946cd37808",
                "name": "Roman Pogodin",
                "collaborators": [
                    "Yazhe Li",
                    "Danica J. Sutherland",
                    "A. Gretton",
                    "P. Latham",
                    "Tor Lattimore",
                    "Antonin Schrab",
                    "Jonathan Cornford",
                    "Arna Ghosh",
                    "Kaiwen Sheng",
                    "Brendan A. Bicknell",
                    "Olivier Codol",
                    "Beverley A. Clark",
                    "Guillaume Lajoie",
                    "Blake A. Richards",
                    "Namrata Deka",
                    "Victor Veitch",
                    "Yash Mehta",
                    "T. Lillicrap",
                    "Dane S. Corneil",
                    "Alexander Seeholzer",
                    "Joseph Heng",
                    "W. Gerstner",
                    "M. Krechetov",
                    "Yury Maximov",
                    "A. Katrutsa",
                    "Sergei Grudinin"
                ],
                "domain": [
                    "Statistical Learning",
                    "Neural Networks",
                    "Computational Neuroscience",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Practical Kernel Tests of Conditional Independence",
                        "abstract": "We describe a data-efficient, kernel-based approach to statistical testing of conditional independence. A major challenge of conditional independence testing, absent in tests of unconditional independence, is to obtain the correct test level (the specified upper bound on the rate of false positives), while still attaining competitive test power. Excess false positives arise due to bias in the test statistic, which is obtained using nonparametric kernel ridge regression. We propose three methods for bias control to correct the test level, based on data splitting, auxiliary data, and (where possible) simpler function classes. We show these combined strategies are effective both for synthetic and real-world data."
                    },
                    {
                        "title": "Brain-like learning with exponentiated gradients",
                        "abstract": "Computational neuroscience relies on gradient descent (GD) for training artificial neural network (ANN) models of the brain. The advantage of GD is that it is effective at learning difficult tasks. However, it produces ANNs that are a poor phenomenological fit to biology, making them less relevant as models of the brain. Specifically, it violates Dale\u2019s law, by allowing synapses to change from excitatory to inhibitory, and leads to synaptic weights that are not log-normally distributed, contradicting experimental data. Here, starting from first principles of optimisation theory, we present an alternative learning algorithm, exponentiated gradient (EG), that respects Dale\u2019s Law and produces log-normal weights, without losing the power of learning with gradients. We also show that in biologically relevant settings EG outperforms GD, including learning from sparsely relevant signals and dealing with synaptic pruning. Altogether, our results show that EG is a superior learning algorithm for modelling the brain with ANNs."
                    },
                    {
                        "title": "Locally connected networks as ventral stream models",
                        "abstract": "Most deep learning models of the ventral stream, and convolutional networks in particular, share weights among neurons. Weight sharing during learning is crucial for good performance on image recognition tasks, but it is not biologically plausible. In this work, we compare performance and Brain-Score results of ImageNet-trained networks in multiple configurations: convolutional, locally connected (i.e., convolutional without weight sharing), and locally connected with anti-Hebbian plasticity mechanisms that promote weight sharing. We also study the role of initialization on performance of those networks. We find that the more weight sharing networks have, the better they perform on both ImageNet and Brain-Score, which can sometimes be further improved with a convolutional initialization. However, locally connected networks outperform their convolutional counterparts on purely neural data (areas V1, V2, V4, IT), but not on behavioral responses. Moreover, ImageNet performance negatively correlates with correspondence to V1 data, suggesting that better models of early visual processing don\u2019t necessarily provide a good input for models of deeper visual areas."
                    },
                    {
                        "title": "Efficient Conditionally Invariant Representation Learning",
                        "abstract": "We introduce the Conditional Independence Regression CovariancE (CIRCE), a measure of conditional independence for multivariate continuous-valued variables. CIRCE applies as a regularizer in settings where we wish to learn neural features $\\varphi(X)$ of data $X$ to estimate a target $Y$, while being conditionally independent of a distractor $Z$ given $Y$. Both $Z$ and $Y$ are assumed to be continuous-valued but relatively low dimensional, whereas $X$ and its features may be complex and high dimensional. Relevant settings include domain-invariant learning, fairness, and causal learning. The procedure requires just a single ridge regression from $Y$ to kernelized features of $Z$, which can be done in advance. It is then only necessary to enforce independence of $\\varphi(X)$ from residuals of this regression, which is possible with attractive estimation properties and consistency guarantees. By contrast, earlier measures of conditional feature dependence require multiple regressions for each step of feature learning, resulting in more severe bias and variance, and greater computational cost. When sufficiently rich features are used, we establish that CIRCE is zero if and only if $\\varphi(X) \\perp \\!\\!\\! \\perp Z \\mid Y$. In experiments, we show superior performance to previous methods on challenging benchmarks, including learning conditionally invariant image features."
                    },
                    {
                        "title": "Self-Supervised Learning with Kernel Dependence Maximization",
                        "abstract": "We approach self-supervised learning of image representations from a statistical dependence perspective, proposing Self-Supervised Learning with the Hilbert-Schmidt Independence Criterion (SSL-HSIC). SSL-HSIC maximizes dependence between representations of transformations of an image and the image identity, while minimizing the kernelized variance of those representations. This framework yields a new understanding of InfoNCE, a variational lower bound on the mutual information (MI) between different transformations. While the MI itself is known to have pathologies which can result in learning meaningless representations, its bound is much better behaved: we show that it implicitly approximates SSL-HSIC (with a slightly different regularizer). Our approach also gives us insight into BYOL, a negative-free SSL method, since SSL-HSIC similarly learns local neighborhoods of samples. SSL-HSIC allows us to directly optimize statistical dependence in time linear in the batch size, without restrictive data assumptions or indirect mutual information estimators. Trained with or without a target network, SSL-HSIC matches the current state-of-the-art for standard linear evaluation on ImageNet, semi-supervised learning and transfer to other classification and vision tasks such as semantic segmentation, depth estimation and object recognition. Code is available at https://github.com/deepmind/ssl_hsic ."
                    },
                    {
                        "title": "Towards Biologically Plausible Convolutional Networks",
                        "abstract": "Convolutional networks are ubiquitous in deep learning. They are particularly useful for images, as they reduce the number of parameters, reduce training time, and increase accuracy. However, as a model of the brain they are seriously problematic, since they require weight sharing - something real neurons simply cannot do. Consequently, while neurons in the brain can be locally connected (one of the features of convolutional networks), they cannot be convolutional. Locally connected but non-convolutional networks, however, significantly underperform convolutional ones. This is troublesome for studies that use convolutional networks to explain activity in the visual system. Here we study plausible alternatives to weight sharing that aim at the same regularization principle, which is to make each neuron within a pool react similarly to identical inputs. The most natural way to do that is by showing the network multiple translations of the same image, akin to saccades in animal vision. However, this approach requires many translations, and doesn't remove the performance gap. We propose instead to add lateral connectivity to a locally connected network, and allow learning via Hebbian plasticity. This requires the network to pause occasionally for a sleep-like phase of \"weight sharing\". This method enables locally connected networks to achieve nearly convolutional performance on ImageNet, thus supporting convolutional networks as a model of the visual stream."
                    },
                    {
                        "title": "Kernelized information bottleneck leads to biologically plausible 3-factor Hebbian learning in deep networks",
                        "abstract": "The state-of-the art machine learning approach to training deep neural networks, backpropagation, is implausible for real neural networks: neurons need to know their outgoing weights; training alternates between a forward pass (computation) and a backward pass (learning); and the algorithm needs a large amount of labeled data. Biologically plausible approximations to backpropagation, such as feedback alignment, solve the weight transport problem, but not the other two. Thus, fully biologically plausible learning rules have so far remained elusive. Here we present a family of learning rules that does not suffer from any of these problems. It is motivated by the information bottleneck principle (extended with kernel methods), in which networks learn to squeeze as much information as possible out of the input without sacrificing prediction of the output. The resulting rules have a 3-factor Hebbian structure: they require pre- and post-synaptic firing rates and a global error signal - the third factor - that can be supplied by a neuromodulator. Moreover, they do not require precise labels; instead, they rely on the similarity between the desired outputs. They thus solve all three implausibility issues of backpropagation. Moreover, to obtain good performance on hard problems and retain biologically plausible learning rules, our rules need divisive normalization - a known feature of biological networks. Finally, simulations show that our rule performs nearly as well as backpropagation on image classification tasks."
                    },
                    {
                        "title": "Working memory facilitates reward-modulated Hebbian learning in recurrent neural networks",
                        "abstract": "Reservoir computing is a powerful tool to explain how the brain learns temporal sequences, such as movements, but existing learning schemes are either biologically implausible or too inefficient to explain animal performance. We show that a network can learn complicated sequences with a reward-modulated Hebbian learning rule if the network of reservoir neurons is combined with a second network that serves as a dynamic working memory and provides a spatio-temporal backbone signal to the reservoir. In combination with the working memory, reward-modulated Hebbian learning of the readout neurons performs as well as FORCE learning, but with the advantage of a biologically plausible interpretation of both the learning rule and the learning paradigm."
                    },
                    {
                        "title": "Adaptivity, Variance and Separation for Adversarial Bandits",
                        "abstract": "We make three contributions to the theory of k-armed adversarial bandits. First, we prove a first-order bound for a modified variant of the INF strategy by Audibert and Bubeck [2009], without sacrificing worst case optimality or modifying the loss estimators. Second, we provide a variance analysis for algorithms based on follow the regularised leader, showing that without adaptation the variance of the regret is typically {\\Omega}(n^2) where n is the horizon. Finally, we study bounds that depend on the degree of separation of the arms, generalising the results by Cowan and Katehakis [2015] from the stochastic setting to the adversarial and improving the result of Seldin and Slivkins [2014] by a factor of log(n)/log(log(n))."
                    },
                    {
                        "title": "On First-Order Bounds, Variance and Gap-Dependent Bounds for Adversarial Bandits",
                        "abstract": "We make three contributions to the theory of k-armed adversarial bandits. First, we prove a first-order bound for a modified variant of the INF strategy by Audibert and Bubeck [2009], without sacrificing worst case optimality or modifying the loss estimators. Second, we provide a variance analysis for algorithms based on follow the regularised leader, showing that without adaptation the variance of the regret is typically {\\Omega}(n^2) where n is the horizon. Finally, we study bounds that depend on the degree of separation of the arms, generalising the results by Cowan and Katehakis [2015] from the stochastic setting to the adversarial and improving the result of Seldin and Slivkins [2014] by a factor of log(n)/log(log(n))."
                    },
                    {
                        "title": "Efficient rank minimization to tighten semidefinite programming for unconstrained binary quadratic optimization",
                        "abstract": "We propose a method for low-rank semidefinite programming in application to the semidefinite relaxation of unconstrained binary quadratic problems. The method improves an existing solution of the semidefinite programming relaxation to achieve a lower rank solution. This procedure is computationally efficient as it does not require projecting on the cone of positive-semidefinite matrices. Its performance in terms of objective improvement and rank reduction is tested over multiple graphs of large-scale Gset graph collection and over binary optimization problems from the Biq Mac collection."
                    },
                    {
                        "title": "Quadratic Programming Approach to Fit Protein Complexes into Electron Density Maps",
                        "abstract": "The paper investigates the problem of fitting protein complexes into electron density maps. They are represented by high-resolution cryoEM density maps converted into overlapping matrices and partly show a structure of a complex. The general purpose is to define positions of all proteins inside it. This problem is known to be NP-hard, since it lays in the field of combinatorial optimization over a set of discrete states of the complex. We introduce quadratic programming approaches to the problem. To find an approximate solution, we convert a density map into an overlapping matrix, which is generally indefinite. Since the matrix is indefinite, the optimization problem for the corresponding quadratic form is non-convex. To treat non-convexity of the optimization problem, we use different convex relaxations to find which set of proteins minimizes the quadratic form best."
                    }
                ]
            },
            "e3dd9c9e-b231-452d-9c95-2f4b0b9ebf1a": {
                "pk": "e3dd9c9e-b231-452d-9c95-2f4b0b9ebf1a",
                "name": "Jonathan Cornford",
                "collaborators": [
                    "D. Kullmann",
                    "M. Mercier",
                    "Vincent Magloire",
                    "A. Snowball",
                    "A. Lieb",
                    "S. Schorge",
                    "B. Richards",
                    "Marco Leite",
                    "G. Colasante",
                    "Yichen Qiu",
                    "Luca Massimino",
                    "Claudia Di Berardino",
                    "Steffan P. Jones",
                    "S. Giannelli",
                    "V. Broccoli",
                    "G. Lignani",
                    "Pingsheng Li",
                    "Arna Ghosh",
                    "Damjan Kalajdzievski",
                    "Am\u00e9lie Lamarquette",
                    "Mikail Weston",
                    "E. Chabrol",
                    "R. Wykes",
                    "Tawfeeq Shekh-Ahmad",
                    "M. Hughes",
                    "G. Massaro",
                    "A. Rahim",
                    "Kevan S Hashemi",
                    "M. Walker",
                    "M. H\u00e4usser"
                ],
                "domain": [
                    "Neuroscience",
                    "Neural Networks",
                    "Gene Therapy",
                    "Computational Neuroscience"
                ],
                "publications": [
                    {
                        "title": "Learning better with Dale\u2019s Law: A Spectral Perspective",
                        "abstract": "Most recurrent neural networks (RNNs) do not include a fundamental constraint of real neural circuits: Dale\u2019s Law, which implies that neurons must be excitatory (E) or inhibitory (I). Dale\u2019s Law is generally absent from RNNs because simply partitioning a standard network\u2019s units into E and I populations impairs learning. However, here we extend a recent feedforward bio-inspired EI network architecture, named Dale\u2019s ANNs, to recurrent networks, and demonstrate that good performance is possible while respecting Dale\u2019s Law. This begs the question: What makes some forms of EI network learn poorly and others learn well? And, why does the simple approach of incorporating Dale\u2019s Law impair learning? Historically the answer was thought to be the sign constraints on EI network parameters, and this was a motivation behind Dale\u2019s ANNs. However, here we show the spectral properties of the recurrent weight matrix at initialisation are more impactful on network performance than sign constraints. We find that simple EI partitioning results in a singular value distribution that is multimodal and dispersed, whereas standard RNNs have an unimodal, more clustered singular value distribution, as do recurrent Dale\u2019s ANNs. We also show that the spectral properties and performance of partitioned EI networks are worse for small networks with fewer I units, and we present normalised SVD entropy as a measure of spectrum pathology that correlates with performance. Overall, this work sheds light on a long-standing mystery in neuroscience-inspired AI and computational neuroscience, paving the way for greater alignment between neural networks and biology."
                    },
                    {
                        "title": "Long\u2010term potentiation in neurogliaform interneurons modulates excitation\u2013inhibition balance in the temporoammonic pathway",
                        "abstract": "Apical dendrites of pyramidal neurons integrate information from higher\u2010order cortex and thalamus, and gate signalling and plasticity at proximal synapses. In the hippocampus, neurogliaform cells and other interneurons located within stratum lacunosum\u2010moleculare (SLM) mediate powerful inhibition of CA1 pyramidal neuron distal dendrites. Is the recruitment of such inhibition itself subject to use\u2010dependent plasticity, and if so, what induction rules apply? Here we show that interneurons in mouse SLM exhibit Hebbian NMDA receptor\u2010dependent long\u2010term potentiation (LTP). Such plasticity can be induced by selective optogenetic stimulation of afferents in the temporoammonic pathway from the entorhinal cortex (EC), but not by equivalent stimulation of afferents from the thalamic nucleus reuniens. We further show that theta\u2010burst patterns of afferent firing induces LTP in neurogliaform interneurons identified using neuron\u2010derived neurotrophic factor (Ndnf)\u2010Cre mice. Theta\u2010burst activity of EC afferents led to an increase in disynaptic feed\u2010forward inhibition, but not monosynaptic excitation, of CA1 pyramidal neurons. Activity\u2010dependent synaptic plasticity in SLM interneurons thus alters the excitation\u2013inhibition balance at EC inputs to the apical dendrites of pyramidal neurons, implying a dynamic role for these interneurons in gating CA1 dendritic computations."
                    },
                    {
                        "title": "Learning to live with Dale\u2019s principle: ANNs with separate excitatory and inhibitory units",
                        "abstract": "The units in artificial neural networks (ANNs) can be thought of as abstractions of biological neurons, and ANNs are increasingly used in neuroscience research. However, there are many important differences between ANN units and real neurons. One of the most notable is the absence of Dale\u2019s principle, which ensures that biological neurons are either exclusively excitatory or inhibitory. Dale\u2019s principle is typically left out of ANNs because its inclusion impairs learning. This is problematic, because one of the great advantages of ANNs for neuroscience research is their ability to learn complicated, realistic tasks. Here, by taking inspiration from feedforward inhibitory interneurons in the brain we show that we can develop ANNs with separate populations of excitatory and inhibitory units that learn just as well as standard ANNs. We call these networks Dale\u2019s ANNs (DANNs). We present two insights that enable DANNs to learn well: (1) DANNs are related to normalization schemes, and can be initialized such that the inhibition centres and standardizes the excitatory activity, (2) updates to inhibitory neuron parameters should be scaled using corrections based on the Fisher Information matrix. These results demonstrate how ANNs that respect Dale\u2019s principle can be built without sacrificing learning performance, which is important for future work using ANNs as models of the brain. The results also may have interesting implications for how inhibitory plasticity in the real brain operates."
                    },
                    {
                        "title": "In vivo CRISPRa decreases seizures and rescues cognitive deficits in a rodent model of epilepsy",
                        "abstract": "CRISPRa targets promoters to modulate gene activity. Colasante, Qiu et al. report that CRISPRa upregulation of an endogenous potassium channel reduces seizures and rescues cognitive defects in an epilepsy model. The study provides proof of principle for a translational CRISPR-based approach to treat neurological diseases."
                    },
                    {
                        "title": "Epilepsy Gene Therapy Using an Engineered Potassium Channel",
                        "abstract": "Refractory focal epilepsy is a devastating disease for which there is frequently no effective treatment. Gene therapy represents a promising alternative, but treating epilepsy in this way involves irreversible changes to brain tissue, so vector design must be carefully optimized to guarantee safety without compromising efficacy. We set out to develop an epilepsy gene therapy vector optimized for clinical translation. The gene encoding the voltage-gated potassium channel Kv1.1, KCNA1, was codon optimized for human expression and mutated to accelerate the recovery of the channels from inactivation. For improved safety, this engineered potassium channel (EKC) gene was packaged into a nonintegrating lentiviral vector under the control of a cell type-specific CAMK2A promoter. In a blinded, randomized, placebo-controlled preclinical trial, the EKC lentivector robustly reduced seizure frequency in a male rat model of focal neocortical epilepsy characterized by discrete spontaneous seizures. When packaged into an adeno-associated viral vector (AAV2/9), the EKC gene was also effective at suppressing seizures in a male rat model of temporal lobe epilepsy. This demonstration of efficacy in a clinically relevant setting, combined with the improved safety conferred by cell type-specific expression and integration-deficient delivery, identify EKC gene therapy as being ready for clinical translation in the treatment of refractory focal epilepsy. SIGNIFICANCE STATEMENT Pharmacoresistant epilepsy affects up to 0.3% of the population. Although epilepsy surgery can be effective, it is limited by risks to normal brain function. We have developed a gene therapy that builds on a mechanistic understanding of altered neuronal and circuit excitability in cortical epilepsy. The potassium channel gene KCNA1 was mutated to bypass post-transcriptional editing and was packaged in a nonintegrating lentivector to reduce the risk of insertional mutagenesis. A randomized, blinded preclinical study demonstrated therapeutic effectiveness in a rodent model of focal neocortical epilepsy. Adeno-associated viral delivery of the channel to both hippocampi was also effective in a model of temporal lobe epilepsy. These results support clinical translation to address a major unmet need."
                    },
                    {
                        "title": "Long-term potentiation in neurogliaform cells modulates excitation-inhibition balance in the temporoammonic pathway",
                        "abstract": "Apical dendrites of pyramidal neurons integrate information from higher-order cortex and thalamus, and gate signaling and plasticity at proximal synapses. In the hippocampus, neurogliaform cells and other interneurons located within stratum lacunosum-moleculare mediate powerful inhibition of CA1 pyramidal neuron distal dendrites. Is the recruitment of such inhibition itself subject to use-dependent plasticity, and if so, what induction rules apply? Here we show that interneurons in mouse stratum lacunosum-moleculare exhibit Hebbian NMDA receptor-dependent long-term potentiation (LTP). Such plasticity can be induced by selective optogenetic stimulation of afferent fibers in the temporoammonic pathway from the entorhinal cortex, but not by equivalent stimulation of afferents from the thalamic nucleus reuniens. We further show that theta-burst patterns of afferent firing induces LTP in neurogliaform interneurons identified using neuron-derived neurotrophic factor (Ndnf)-Cre mice. Theta-burst activity of entorhinal cortex afferents led to an increase in disynaptic feed-forward inhibition, but not monosynaptic excitation, of CA1 pyramidal neurons. Activity-dependent synaptic plasticity of neurogliaform cells in stratum lacunosum-moleculare thus alters the excitation-inhibition balance at entorhinal cortex inputs to the apical dendrites of pyramidal neurons, implying a dynamic role for these interneurons in gating CA1 dendritic computations. Significance statement Electrogenic phenomena in distal dendrites of principal neurons in the hippocampus have a major role in gating synaptic plasticity at afferent synapses on proximal dendrites. Apical dendrites also receive powerful feed-forward inhibition mediated in large part by neurogliaform neurons. Here we show that theta-burst activity in afferents from the entorhinal cortex induces \u2018Hebbian\u2019 long-term potentiation at excitatory synapses recruiting these GABAergic cells. Such LTP increases disynaptic inhibition of principal neurons, thus shifting the excitation-inhibition balance in the temporoammonic pathway in favor of inhibition, with implications for computations and learning rules in proximal dendrites."
                    },
                    {
                        "title": "Long-term synaptic plasticity in hippocampal neurogliaform interneurons",
                        "abstract": "Hippocampal interneurons located within stratum lacunosum-moleculare (SLM), which include neurogliaform (NGF) cells, mediate powerful feed-forward inhibition that can modulate spiking and plasticity in CA1 pyramidal cells. Despite evidence of long-term plasticity at excitatory inputs onto almost all other hippocampal interneuron subtypes, including stratum radiatum feed-forward interneurons, it is not known whether long-term potentiation (LTP) occurs in CA1 SLM interneurons. Here, we show that these interneurons exhibit Hebbian NMDA receptor-dependent LTP, and that Ca2+ influx through voltage-gated Ca2+ channels can also be sufficient for induction of plasticity. Furthermore, using an optogenetic dissection strategy, we find that selective stimulation of excitatory fibers from entorhinal cortex can induce LTP in SLM interneurons, whilst stimulation of thalamic afferents from the nucleus reuniens, also known to project to SLM, does not. Finally, we show that a mouse line selective for cortical NGF cells, where Cre recombinase is under the control of the neuron-derived neurotrophic factor (NDNF) promoter, can also be used to target these interneurons within the hippocampus, and that hippocampal NGF cells exhibit LTP. Recruitment of NGF cells can thus be persistently enhanced in an activity-dependent manner, implying that their role in gating dendritic signaling in pyramidal neurons is modifiable."
                    },
                    {
                        "title": "Dendritic NMDA receptors in parvalbumin neurons enable strong and stable neuronal assemblies",
                        "abstract": "Fast-spiking, parvalbumin-expressing (PV+) GABAergic interneurons mediate feedforward and feedback inhibition and have a key role in gamma oscillations and information processing. The importance of fast PV+ cell signalling for these circuit functions is well established. In contrast, the adaptive significance of NMDA receptors (NMDARs), which generate relatively slow postsynaptic currents, in PV+ cells is unclear. Underlining their potential importance, several studies implicate PV+ cell NMDAR disruption in pathology and impaired network function. Here, we show that NMDARs underlie supralinear integration of feedback excitation from local pyramidal neurons onto mouse CA1 PV+ cells. Furthermore, by incorporating NMDARs at feedback connections onto PV+ cells in spiking neural networks, we show that these receptors enable cooperative recruitment of PV+ interneurons, strengthening and stabilising principal cell assemblies. Failure of this phenomenon provides a parsimonious explanation for cognitive and sensory gating deficits in pathologies with impaired PV+ NMDAR signalling."
                    },
                    {
                        "title": "CRISPRa-mediated Kcna1 upregulation decreases neuronal excitability and suppresses seizures in a rodent model of temporal lobe epilepsy",
                        "abstract": "Epilepsy is a major health burden, calling for new mechanistic and therapeutic insights. CRISPR\u2013mediated gene editing shows promise to cure genetic pathologies, although hitherto it has mostly been applied ex-vivo. Its translational potential for treating non-genetic pathologies is still unexplored. Furthermore, neurological diseases represent an important challenge for the application of CRISPR, because of the need in many cases to manipulate gene function of neurons in situ. A variant of CRISPR, CRISPRa, offers the possibility to modulate the expression of endogenous genes by directly targeting their promoters. We asked if this strategy can be effective to treat acquired focal epilepsy. We applied a doxycycline-inducible CRISPRa technology to increase the expression of the potassium channel gene Kcna1 (encoding Kv1.1) in mouse hippocampal excitatory neurons. CRISPRa-mediated Kv1.1 upregulation led to a substantial decrease in neuronal excitability. Continuous video-EEG telemetry showed that AAV9-mediated delivery of CRISPRa, upon doxycycline administration, decreased spontaneous generalized tonic-clonic seizures in a model of temporal lobe epilepsy. The focal treatment minimizes concerns about off-targets effects in other organs and brain areas. This study provides the proof of principle for a translational CRISPR-based approach to treat neurological diseases characterized by abnormal circuit excitability."
                    },
                    {
                        "title": "Computational aspects of parvalbumin-positive interneuron function",
                        "abstract": "The activity of neurons is dependent on the manner in which they process synaptic inputs from other cells. In the event of clustered synaptic input, neurons can respond in a nonlinear manner through synaptic and dendritic mechanisms. Such mechanisms are well established in principal excitatory neurons throughout the brain, where they increase neuronal computational ability and information storage capacity. In contrast for parvalbumin-positive (PV+) interneurons, the most common cortical class of in- hibitory interneuron, synaptic integration is thought to be either linear or sub-linear in nature, facilitating their role as mediators of precise and fast inhibition. This thesis addresses situations in which PV+ interneurons integrate synaptic inputs in a nonlinear manner, and explores the functions of this synaptic processing. First, I describe a form of cooperative supralinear synaptic integration by local excitatory inputs onto PV+ interneurons, and I extend these results to show how this augments the computational capability of PV+ cells within spiking neuron networks. I also explore the importance of polyamine-modulation of synaptic receptors in mediating sublinear synaptic integration, and discuss how this expands the array of mechanisms known to perform similar functions in PV+ cells. Finally, I present work manipulating PV+ cells experimentally during epilepsy. I consider these findings together with recent scientific advances and suggest how they account for a number of open questions and previously contradictory theories of PV+ interneuron function."
                    }
                ]
            },
            "1704501b-1a61-484c-953c-ff5ae59765e1": {
                "pk": "1704501b-1a61-484c-953c-ff5ae59765e1",
                "name": "Arna Ghosh",
                "collaborators": [
                    "Blake A. Richards",
                    "Guillaume Lajoie",
                    "B. Richards",
                    "Yuhan Helena Liu",
                    "Raymond Chua",
                    "Christos Kaplanis",
                    "D. Precup",
                    "Jonathan Cornford",
                    "Roman Pogodin",
                    "Kaiwen Sheng",
                    "Brendan A. Bicknell",
                    "Olivier Codol",
                    "Beverley A. Clark",
                    "Konrad K\u00a8ording",
                    "Pingsheng Li",
                    "Jonathan H. Cornford",
                    "E. Shea-Brown",
                    "Luke Y. Prince",
                    "Roy Henha Eyono",
                    "E. Boven",
                    "Joe Pemberton",
                    "Franz Scherr",
                    "C. Clopath",
                    "R. P. Costa",
                    "W. Maass",
                    "Cristina Savin",
                    "K. Wilmes",
                    "Matteo Gamba",
                    "Kumar Krishna",
                    "Agrawal",
                    "Hossein Azizpour",
                    "M\u00e5rten Bj\u00f6rkman"
                ],
                "domain": [
                    "Deep Reinforcement Learning",
                    "Computational Neuroscience",
                    "Neural Networks",
                    "Generalization"
                ],
                "publications": [
                    {
                        "title": "Learning Successor Features the Simple Way",
                        "abstract": "In Deep Reinforcement Learning (RL), it is a challenge to learn representations that do not exhibit catastrophic forgetting or interference in non-stationary environments. Successor Features (SFs) offer a potential solution to this challenge. However, canonical techniques for learning SFs from pixel-level observations often lead to representation collapse, wherein representations degenerate and fail to capture meaningful variations in the data. More recent methods for learning SFs can avoid representation collapse, but they often involve complex losses and multiple learning phases, reducing their efficiency. We introduce a novel, simple method for learning SFs directly from pixels. Our approach uses a combination of a Temporal-difference (TD) loss and a reward prediction loss, which together capture the basic mathematical definition of SFs. We show that our approach matches or outperforms existing SF learning techniques in both 2D (Minigrid), 3D (Miniworld) mazes and Mujoco, for both single and continual learning scenarios. As well, our technique is efficient, and can reach higher levels of performance in less time than other approaches. Our work provides a new, streamlined technique for learning SFs directly from pixel observations, with no pretraining required."
                    },
                    {
                        "title": "Brain-like learning with exponentiated gradients",
                        "abstract": "Computational neuroscience relies on gradient descent (GD) for training artificial neural network (ANN) models of the brain. The advantage of GD is that it is effective at learning difficult tasks. However, it produces ANNs that are a poor phenomenological fit to biology, making them less relevant as models of the brain. Specifically, it violates Dale\u2019s law, by allowing synapses to change from excitatory to inhibitory, and leads to synaptic weights that are not log-normally distributed, contradicting experimental data. Here, starting from first principles of optimisation theory, we present an alternative learning algorithm, exponentiated gradient (EG), that respects Dale\u2019s Law and produces log-normal weights, without losing the power of learning with gradients. We also show that in biologically relevant settings EG outperforms GD, including learning from sparsely relevant signals and dealing with synaptic pruning. Altogether, our results show that EG is a superior learning algorithm for modelling the brain with ANNs."
                    },
                    {
                        "title": "H OW GRADIENT ESTIMATOR VARIANCE AND BIAS COULD IMPACT LEARNING IN NEURAL CIRCUITS",
                        "abstract": "There is growing interest in understanding how real brains may approximate gradients and how gradients can be used to train neuromorphic chips. However, neither real brains nor neuromorphic chips can perfectly follow the loss gradient, so parameter updates would necessarily use gradient estimators that have some variance and/or bias. Therefore, there is a need to understand better how variance and bias in gradient estimators impact learning dependent on network and task properties. Here, we show that variance and bias can impair learning on the training data, but some degree of variance and bias in a gradient estimator can be beneficial for generalization. We find that the ideal amount of variance and bias in a gradient estimator are dependent on several properties of the network and task: the size and activity sparsity of the network, the norm of the gradient, and the curvature of the loss landscape. As such, whether considering biologically-plausible learning algorithms or algorithms for training neuromorphic chips, researchers can analyze these properties to determine whether their approximation to gradient descent will be effective for learning given their network and task properties."
                    },
                    {
                        "title": "Learning better with Dale\u2019s Law: A Spectral Perspective",
                        "abstract": "Most recurrent neural networks (RNNs) do not include a fundamental constraint of real neural circuits: Dale\u2019s Law, which implies that neurons must be excitatory (E) or inhibitory (I). Dale\u2019s Law is generally absent from RNNs because simply partitioning a standard network\u2019s units into E and I populations impairs learning. However, here we extend a recent feedforward bio-inspired EI network architecture, named Dale\u2019s ANNs, to recurrent networks, and demonstrate that good performance is possible while respecting Dale\u2019s Law. This begs the question: What makes some forms of EI network learn poorly and others learn well? And, why does the simple approach of incorporating Dale\u2019s Law impair learning? Historically the answer was thought to be the sign constraints on EI network parameters, and this was a motivation behind Dale\u2019s ANNs. However, here we show the spectral properties of the recurrent weight matrix at initialisation are more impactful on network performance than sign constraints. We find that simple EI partitioning results in a singular value distribution that is multimodal and dispersed, whereas standard RNNs have an unimodal, more clustered singular value distribution, as do recurrent Dale\u2019s ANNs. We also show that the spectral properties and performance of partitioned EI networks are worse for small networks with fewer I units, and we present normalised SVD entropy as a measure of spectrum pathology that correlates with performance. Overall, this work sheds light on a long-standing mystery in neuroscience-inspired AI and computational neuroscience, paving the way for greater alignment between neural networks and biology."
                    },
                    {
                        "title": "Beyond accuracy: generalization properties of bio-plausible temporal credit assignment rules",
                        "abstract": "To unveil how the brain learns, ongoing work seeks biologically-plausible approximations of gradient descent algorithms for training recurrent neural networks (RNNs). Yet, beyond task accuracy, it is unclear if such learning rules converge to solutions that exhibit different levels of generalization than their nonbiologically-plausible counterparts. Leveraging results from deep learning theory based on loss landscape curvature, we ask: how do biologically-plausible gradient approximations affect generalization? We first demonstrate that state-of-the-art biologically-plausible learning rules for training RNNs exhibit worse and more variable generalization performance compared to their machine learning counterparts that follow the true gradient more closely. Next, we verify that such generalization performance is correlated significantly with loss landscape curvature, and we show that biologically-plausible learning rules tend to approach high-curvature regions in synaptic weight space. Using tools from dynamical systems, we derive theoretical arguments and present a theorem explaining this phenomenon. This predicts our numerical results, and explains why biologically-plausible rules lead to worse and more variable generalization properties. Finally, we suggest potential remedies that could be used by the brain to mitigate this effect. To our knowledge, our analysis is the first to identify the reason for this generalization gap between artificial and biologically-plausible learning rules, which can help guide future investigations into how the brain learns solutions that generalize."
                    },
                    {
                        "title": "Current State and Future Directions for Learning in Biological Recurrent Neural Networks: A Perspective Piece",
                        "abstract": "We provide a brief review of the common assumptions about biological learning with findings from experimental neuroscience and contrast them with the efficiency of gradient-based learning in recurrent neural networks. The key issues discussed in this review include: synaptic plasticity, neural circuits, theory-experiment divide, and objective functions. We conclude with recommendations for both theoretical and experimental neuroscientists when designing new studies that could help bring clarity to these issues."
                    },
                    {
                        "title": "On the Varied Faces of Overparameterization in Supervised and Self-Supervised Learning",
                        "abstract": "The quality of the representations learned by neural networks depends on several factors, including the loss function, learning algorithm, and model architecture. In this work, we use information geometric measures to assess the representation quality in a principled manner. We demonstrate that the sensitivity of learned representations to input perturbations, measured by the spectral norm of the feature Jacobian, provides valuable information about downstream generalization. On the other hand, measuring the coef\ufb01cient of spectral decay observed in the eigen-spectrum of feature covariance provides insights into the global representation geometry. First, we empirically establish an equivalence between these notions of representation quality and show that they are inversely correlated. Second, our analysis reveals the varying roles that overparameterization plays in improving generalization. Unlike supervised learning, we observe that increasing model width leads to higher discriminability and less smoothness in the self-supervised regime. Furthermore, we report that there is no observable double descent phenomenon in SSL with non-contrastive objectives for commonly used parameterization regimes, which opens up new opportunities for tight asymptotic analysis. Taken together, our results provide a loss-aware characterization of the different role of overparam-eterization in supervised and self-supervised learning."
                    }
                ]
            },
            "92c20759-9aee-4d96-8a62-d469ff6e8db2": {
                "pk": "92c20759-9aee-4d96-8a62-d469ff6e8db2",
                "name": "Gauthier Gidel",
                "collaborators": [
                    "David Dobre",
                    "Leo Schwinn",
                    "Sophie Xhonneux",
                    "Stephan G\u00fcnnemann",
                    "A. Gasnikov",
                    "Dhanya Sridhar",
                    "Y. Bengio",
                    "Nikolay Malkin",
                    "Damien Ferbach",
                    "A. Bose",
                    "Aleksandr Beznosikov",
                    "Mehrnaz Mofakhami",
                    "Abdurakhmon Sadiev",
                    "Marina Danilova",
                    "Eduard A. Gorbunov",
                    "Samuel Horv'ath",
                    "P. Dvurechensky",
                    "Jian Tang",
                    "Seanie Lee",
                    "Minsu Kim",
                    "Lynn Cherif",
                    "Juho Lee",
                    "Sung Ju Hwang",
                    "Kenji Kawaguchi",
                    "Moksh Jain",
                    "Quentin Bertrand",
                    "Michael Przystupa",
                    "Matthew E. Taylor",
                    "Martin Jagersand",
                    "Justus Piater",
                    "Samuele Tosatto",
                    "Alessandro Sordoni",
                    "Marco Jiralerspong",
                    "Thomas Jiralerspong",
                    "Vedant Shah",
                    "Tara Akhound-Sadegh",
                    "Jarrid Rector-Brooks",
                    "Sarthak Mittal",
                    "Pablo Lemos",
                    "Cheng-Hao Liu",
                    "Marcin Sendera",
                    "Siamak Ravanbakhsh",
                    "Alexander Tong",
                    "Ruslan Nazykov",
                    "Aleksandr Shestakov",
                    "Vladimir Solodkin",
                    "Ant'onio G'ois",
                    "Fernando P. Santos",
                    "Simon Lacoste-Julien",
                    "Thomas Altstidl",
                    "Bj\u00f6rn Eskofier",
                    "Safwan Hossain",
                    "Andjela Mladenovic",
                    "Yiling Chen",
                    "Peter Richt\u00e1rik",
                    "Ioannis Mitliagkas",
                    "Baptiste Goujaud",
                    "Aymeric Dieuleveut",
                    "Peter Richt'arik"
                ],
                "domain": [
                    "Adversarial Robustness",
                    "Large Language Models",
                    "Optimization",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "In-Context Learning Can Re-learn Forbidden Tasks",
                        "abstract": "Despite significant investment into safety training, large language models (LLMs) deployed in the real world still suffer from numerous vulnerabilities. One perspective on LLM safety training is that it algorithmically forbids the model from answering toxic or harmful queries. To assess the effectiveness of safety training, in this work, we study forbidden tasks, i.e., tasks the model is designed to refuse to answer. Specifically, we investigate whether in-context learning (ICL) can be used to re-learn forbidden tasks despite the explicit fine-tuning of the model to refuse them. We first examine a toy example of refusing sentiment classification to demonstrate the problem. Then, we use ICL on a model fine-tuned to refuse to summarise made-up news articles. Finally, we investigate whether ICL can undo safety training, which could represent a major security risk. For the safety task, we look at Vicuna-7B, Starling-7B, and Llama2-7B. We show that the attack works out-of-the-box on Starling-7B and Vicuna-7B but fails on Llama2-7B. Finally, we propose an ICL attack that uses the chat template tokens like a prompt injection attack to achieve a better attack success rate on Vicuna-7B and Starling-7B. Trigger Warning: the appendix contains LLM-generated text with violence, suicide, and misinformation."
                    },
                    {
                        "title": "Learning diverse attacks on large language models for robust red-teaming and safety tuning",
                        "abstract": "Red-teaming, or identifying prompts that elicit harmful responses, is a critical step in ensuring the safe and responsible deployment of large language models (LLMs). Developing effective protection against many modes of attack prompts requires discovering diverse attacks. Automated red-teaming typically uses reinforcement learning to fine-tune an attacker language model to generate prompts that elicit undesirable responses from a target LLM, as measured, for example, by an auxiliary toxicity classifier. We show that even with explicit regularization to favor novelty and diversity, existing approaches suffer from mode collapse or fail to generate effective attacks. As a flexible and probabilistically principled alternative, we propose to use GFlowNet fine-tuning, followed by a secondary smoothing phase, to train the attacker model to generate diverse and effective attack prompts. We find that the attacks generated by our method are effective against a wide range of target LLMs, both with and without safety tuning, and transfer well between target LLMs. Finally, we demonstrate that models safety-tuned using a dataset of red-teaming prompts generated by our method are robust to attacks from other RL-based red-teaming approaches."
                    },
                    {
                        "title": "Self-Consuming Generative Models with Curated Data Provably Optimize Human Preferences",
                        "abstract": "The rapid progress in generative models has resulted in impressive leaps in generation quality, blurring the lines between synthetic and real data. Web-scale datasets are now prone to the inevitable contamination by synthetic data, directly impacting the training of future generated models. Already, some theoretical results on self-consuming generative models (a.k.a., iterative retraining) have emerged in the literature, showcasing that either model collapse or stability could be possible depending on the fraction of generated data used at each retraining step. However, in practice, synthetic data is often subject to human feedback and curated by users before being used and uploaded online. For instance, many interfaces of popular text-to-image generative models, such as Stable Diffusion or Midjourney, produce several variations of an image for a given query which can eventually be curated by the users. In this paper, we theoretically study the impact of data curation on iterated retraining of generative models and show that it can be seen as an \\emph{implicit preference optimization mechanism}. However, unlike standard preference optimization, the generative model does not have access to the reward function or negative samples needed for pairwise comparisons. Moreover, our study doesn't require access to the density function, only to samples. We prove that, if the data is curated according to a reward model, then the expected reward of the iterative retraining procedure is maximized. We further provide theoretical results on the stability of the retraining loop when using a positive fraction of real data at each step. Finally, we conduct illustrative experiments on both synthetic datasets and on CIFAR10 showing that such a procedure amplifies biases of the reward model."
                    },
                    {
                        "title": "Investigating the Benefits of Nonlinear Action Maps in Data-Driven Teleoperation",
                        "abstract": "As robots become more common for both able-bodied individuals and those living with a disability, it is increasingly important that lay people be able to drive multi-degree-of-freedom platforms with low-dimensional controllers. One approach is to use state-conditioned action mapping methods to learn mappings between low-dimensional controllers and high DOF manipulators -- prior research suggests these mappings can simplify the teleoperation experience for users. Recent works suggest that neural networks predicting a local linear function are superior to the typical end-to-end multi-layer perceptrons because they allow users to more easily undo actions, providing more control over the system. However, local linear models assume actions exist on a linear subspace and may not capture nuanced actions in training data. We observe that the benefit of these mappings is being an odd function concerning user actions, and propose end-to-end nonlinear action maps which achieve this property. Unfortunately, our experiments show that such modifications offer minimal advantages over previous solutions. We find that nonlinear odd functions behave linearly for most of the control space, suggesting architecture structure improvements are not the primary factor in data-driven teleoperation. Our results suggest other avenues, such as data augmentation techniques and analysis of human behavior, are necessary for action maps to become practical in real-world applications, such as in assistive robotics to improve the quality of life of people living with w disability."
                    },
                    {
                        "title": "Efficient Adversarial Training in LLMs with Continuous Attacks",
                        "abstract": "Large language models (LLMs) are vulnerable to adversarial attacks that can bypass their safety guardrails. In many domains, adversarial training has proven to be one of the most promising methods to reliably improve robustness against such attacks. Yet, in the context of LLMs, current methods for adversarial training are hindered by the high computational costs required to perform discrete adversarial attacks at each training iteration. We address this problem by instead calculating adversarial attacks in the continuous embedding space of the LLM, which is orders of magnitudes more efficient. We propose a fast adversarial training algorithm (C-AdvUL) composed of two losses: the first makes the model robust on continuous embedding attacks computed on an adversarial behaviour dataset; the second ensures the usefulness of the final model by fine-tuning on utility data. Moreover, we introduce C-AdvIPO, an adversarial variant of IPO that does not require utility data for adversarially robust alignment. Our empirical evaluation on four models from different families (Gemma, Phi3, Mistral, Zephyr) and at different scales (2B, 3.8B, 7B) shows that both algorithms substantially enhance LLM robustness against discrete attacks (GCG, AutoDAN, PAIR), while maintaining utility. Our results demonstrate that robustness to continuous perturbations can extrapolate to discrete threat models. Thereby, we present a path toward scalable adversarial training algorithms for robustly aligning LLMs."
                    },
                    {
                        "title": "General Causal Imputation via Synthetic Interventions",
                        "abstract": "Given two sets of elements (such as cell types and drug compounds), researchers typically only have access to a limited subset of their interactions. The task of causal imputation involves using this subset to predict unobserved interactions. Squires et al. (2022) have proposed two estimators for this task based on the synthetic interventions (SI) estimator: SI-A (for actions) and SI-C (for contexts). We extend their work and introduce a novel causal imputation estimator, generalized synthetic interventions (GSI). We prove the identifiability of this estimator for data generated from a more complex latent factor model. On synthetic and real data we show empirically that it recovers or outperforms their estimators."
                    },
                    {
                        "title": "Soft Prompt Threats: Attacking Safety Alignment and Unlearning in Open-Source LLMs through the Embedding Space",
                        "abstract": "Current research in adversarial robustness of LLMs focuses on discrete input manipulations in the natural language space, which can be directly transferred to closed-source models. However, this approach neglects the steady progression of open-source models. As open-source models advance in capability, ensuring their safety also becomes increasingly imperative. Yet, attacks tailored to open-source LLMs that exploit full model access remain largely unexplored. We address this research gap and propose the embedding space attack, which directly attacks the continuous embedding representation of input tokens. We find that embedding space attacks circumvent model alignments and trigger harmful behaviors more efficiently than discrete attacks or model fine-tuning. Furthermore, we present a novel threat model in the context of unlearning and show that embedding space attacks can extract supposedly deleted information from unlearned LLMs across multiple datasets and models. Our findings highlight embedding space attacks as an important threat model in open-source LLMs. Trigger Warning: the appendix contains LLM-generated text with violence and harassment."
                    },
                    {
                        "title": "Iterated Denoising Energy Matching for Sampling from Boltzmann Densities",
                        "abstract": "Efficiently generating statistically independent samples from an unnormalized probability distribution, such as equilibrium samples of many-body systems, is a foundational problem in science. In this paper, we propose Iterated Denoising Energy Matching (iDEM), an iterative algorithm that uses a novel stochastic score matching objective leveraging solely the energy function and its gradient -- and no data samples -- to train a diffusion-based sampler. Specifically, iDEM alternates between (I) sampling regions of high model density from a diffusion-based sampler and (II) using these samples in our stochastic matching objective to further improve the sampler. iDEM is scalable to high dimensions as the inner matching objective, is simulation-free, and requires no MCMC samples. Moreover, by leveraging the fast mode mixing behavior of diffusion, iDEM smooths out the energy landscape enabling efficient exploration and learning of an amortized sampler. We evaluate iDEM on a suite of tasks ranging from standard synthetic energy functions to invariant $n$-body particle systems. We show that the proposed approach achieves state-of-the-art performance on all metrics and trains $2-5\\times$ faster, which allows it to be the first method to train using energy on the challenging $55$-particle Lennard-Jones system."
                    },
                    {
                        "title": "Stochastic Frank-Wolfe: Unified Analysis and Zoo of Special Cases",
                        "abstract": "The Conditional Gradient (or Frank-Wolfe) method is one of the most well-known methods for solving constrained optimization problems appearing in various machine learning tasks. The simplicity of iteration and applicability to many practical problems helped the method to gain popularity in the community. In recent years, the Frank-Wolfe algorithm received many different extensions, including stochastic modifications with variance reduction and coordinate sampling for training of huge models or distributed variants for big data problems. In this paper, we present a unified convergence analysis of the Stochastic Frank-Wolfe method that covers a large number of particular practical cases that may have completely different nature of stochasticity, intuitions and application areas. Our analysis is based on a key parametric assumption on the variance of the stochastic gradients. But unlike most works on unified analysis of other methods, such as SGD, we do not assume an unbiasedness of the real gradient estimation. We conduct analysis for convex and non-convex problems due to the popularity of both cases in machine learning. With this general theoretical framework, we not only cover rates of many known methods, but also develop numerous new methods. This shows the flexibility of our approach in developing new algorithms based on the Conditional Gradient approach. We also demonstrate the properties of the new methods through numerical experiments."
                    },
                    {
                        "title": "Performative Prediction on Games and Mechanism Design",
                        "abstract": "Predictions often influence the reality which they aim to predict, an effect known as performativity. Existing work focuses on accuracy maximization under this effect, but model deployment may have important unintended impacts, especially in multiagent scenarios. In this work, we investigate performative prediction in a concrete game-theoretic setting where social welfare is an alternative objective to accuracy maximization. We explore a collective risk dilemma scenario where maximising accuracy can negatively impact social welfare, when predicting collective behaviours. By assuming knowledge of a Bayesian agent behavior model, we then show how to achieve better trade-offs and use them for mechanism design."
                    },
                    {
                        "title": "Sarah Frank-Wolfe: Methods for Constrained Optimization with Best Rates and Practical Features",
                        "abstract": "The Frank-Wolfe (FW) method is a popular approach for solving optimization problems with structured constraints that arise in machine learning applications. In recent years, stochastic versions of FW have gained popularity, motivated by large datasets for which the computation of the full gradient is prohibitively expensive. In this paper, we present two new variants of the FW algorithms for stochastic finite-sum minimization. Our algorithms have the best convergence guarantees of existing stochastic FW approaches for both convex and non-convex objective functions. Our methods do not have the issue of permanently collecting large batches, which is common to many stochastic projection-free approaches. Moreover, our second approach does not require either large batches or full deterministic gradients, which is a typical weakness of many techniques for finite-sum problems. The faster theoretical rates of our approaches are confirmed experimentally."
                    },
                    {
                        "title": "Raising the Bar for Certified Adversarial Robustness with Diffusion Models",
                        "abstract": "Certified defenses against adversarial attacks offer formal guarantees on the robustness of a model, making them more reliable than empirical methods such as adversarial training, whose effectiveness is often later reduced by unseen attacks. Still, the limited certified robustness that is currently achievable has been a bottleneck for their practical adoption. Gowal et al. and Wang et al. have shown that generating additional training data using state-of-the-art diffusion models can considerably improve the robustness of adversarial training. In this work, we demonstrate that a similar approach can substantially improve deterministic certified defenses. In addition, we provide a list of recommendations to scale the robustness of certified training approaches. One of our main insights is that the generalization gap, i.e., the difference between the training and test accuracy of the original model, is a good predictor of the magnitude of the robustness improvement when using additional generated data. Our approach achieves state-of-the-art deterministic robustness certificates on CIFAR-10 for the $\\ell_2$ ($\\epsilon = 36/255$) and $\\ell_\\infty$ ($\\epsilon = 8/255$) threat models, outperforming the previous best results by $+3.95\\%$ and $+1.39\\%$, respectively. Furthermore, we report similar improvements for CIFAR-100."
                    },
                    {
                        "title": "A Persuasive Approach to Combating Misinformation",
                        "abstract": "Bayesian Persuasion is proposed as a tool for social media platforms to combat the spread of misinformation. Since platforms can use machine learning to predict the popularity and misinformation features of to-be-shared posts, and users are largely motivated to share popular content, platforms can strategically signal this informational advantage to change user beliefs and persuade them not to share misinformation. We characterize the optimal signaling scheme with imperfect predictions as a linear program and give sufficient and necessary conditions on the classifier to ensure optimal platform utility is non-decreasing and continuous. Next, this interaction is considered under a performative model, wherein platform intervention affects the user's future behaviour. The convergence and stability of optimal signaling under this performative process are fully characterized. Lastly, we experimentally validate that our approach significantly reduces misinformation in both the single round and performative setting and discuss the broader scope of using information design to combat misinformation."
                    },
                    {
                        "title": "Adversarial Attacks and Defenses in Large Language Models: Old and New Threats",
                        "abstract": "Over the past decade, there has been extensive research aimed at enhancing the robustness of neural networks, yet this problem remains vastly unsolved. Here, one major impediment has been the overestimation of the robustness of new defense approaches due to faulty defense evaluations. Flawed robustness evaluations necessitate rectifications in subsequent works, dangerously slowing down the research and providing a false sense of security. In this context, we will face substantial challenges associated with an impending adversarial arms race in natural language processing, specifically with closed-source Large Language Models (LLMs), such as ChatGPT, Google Bard, or Anthropic's Claude. We provide a first set of prerequisites to improve the robustness assessment of new approaches and reduce the amount of faulty evaluations. Additionally, we identify embedding space attacks on LLMs as another viable threat model for the purposes of generating malicious content in open-sourced models. Finally, we demonstrate on a recently proposed defense that, without LLM-specific best practices in place, it is easy to overestimate the robustness of a new approach."
                    },
                    {
                        "title": "High-Probability Bounds for Stochastic Optimization and Variational Inequalities: the Case of Unbounded Variance",
                        "abstract": "During recent years the interest of optimization and machine learning communities in high-probability convergence of stochastic optimization methods has been growing. One of the main reasons for this is that high-probability complexity bounds are more accurate and less studied than in-expectation ones. However, SOTA high-probability non-asymptotic convergence results are derived under strong assumptions such as the boundedness of the gradient noise variance or of the objective's gradient itself. In this paper, we propose several algorithms with high-probability convergence results under less restrictive assumptions. In particular, we derive new high-probability convergence results under the assumption that the gradient/operator noise has bounded central $\\alpha$-th moment for $\\alpha \\in (1,2]$ in the following setups: (i) smooth non-convex / Polyak-Lojasiewicz / convex / strongly convex / quasi-strongly convex minimization problems, (ii) Lipschitz / star-cocoercive and monotone / quasi-strongly monotone variational inequalities. These results justify the usage of the considered methods for solving problems that do not fit standard functional classes studied in stochastic optimization."
                    },
                    {
                        "title": "Performative Prediction with Neural Networks",
                        "abstract": "Performative prediction is a framework for learning models that influence the data they intend to predict. We focus on finding classifiers that are performatively stable, i.e. optimal for the data distribution they induce. Standard convergence results for finding a performatively stable classifier with the method of repeated risk minimization assume that the data distribution is Lipschitz continuous to the model's parameters. Under this assumption, the loss must be strongly convex and smooth in these parameters; otherwise, the method will diverge for some problems. In this work, we instead assume that the data distribution is Lipschitz continuous with respect to the model's predictions, a more natural assumption for performative systems. As a result, we are able to significantly relax the assumptions on the loss function. In particular, we do not need to assume convexity with respect to the model's parameters. As an illustration, we introduce a resampling procedure that models realistic distribution shifts and show that it satisfies our assumptions. We support our theory by showing that one can learn performatively stable classifiers with neural networks making predictions about real data that shift according to our proposed procedure."
                    },
                    {
                        "title": "Proving Linear Mode Connectivity of Neural Networks via Optimal Transport",
                        "abstract": "The energy landscape of high-dimensional non-convex optimization problems is crucial to understanding the effectiveness of modern deep neural network architectures. Recent works have experimentally shown that two different solutions found after two runs of a stochastic training are often connected by very simple continuous paths (e.g., linear) modulo a permutation of the weights. In this paper, we provide a framework theoretically explaining this empirical observation. Based on convergence rates in Wasserstein distance of empirical measures, we show that, with high probability, two wide enough two-layer neural networks trained with stochastic gradient descent are linearly connected. Additionally, we express upper and lower bounds on the width of each layer of two deep neural networks with independent neuron weights to be linearly connected. Finally, we empirically demonstrate the validity of our approach by showing how the dimension of the support of the weight distribution of neurons, which dictates Wasserstein convergence rates is correlated with linear mode connectivity."
                    },
                    {
                        "title": "High-Probability Convergence for Composite and Distributed Stochastic Minimization and Variational Inequalities with Heavy-Tailed Noise",
                        "abstract": "High-probability analysis of stochastic first-order optimization methods under mild assumptions on the noise has been gaining a lot of attention in recent years. Typically, gradient clipping is one of the key algorithmic ingredients to derive good high-probability guarantees when the noise is heavy-tailed. However, if implemented na\\\"ively, clipping can spoil the convergence of the popular methods for composite and distributed optimization (Prox-SGD/Parallel SGD) even in the absence of any noise. Due to this reason, many works on high-probability analysis consider only unconstrained non-distributed problems, and the existing results for composite/distributed problems do not include some important special cases (like strongly convex problems) and are not optimal. To address this issue, we propose new stochastic methods for composite and distributed optimization based on the clipping of stochastic gradient differences and prove tight high-probability convergence results (including nearly optimal ones) for the new methods. Using similar ideas, we also develop new methods for composite and distributed variational inequalities and analyze the high-probability convergence of these methods."
                    }
                ]
            },
            "e75243d0-1f00-4d69-aaa4-b87c33ef4950": {
                "pk": "e75243d0-1f00-4d69-aaa4-b87c33ef4950",
                "name": "Guillaume Lajoie",
                "collaborators": [
                    "Y. Bengio",
                    "Giancarlo Kerg",
                    "Guy Wolf",
                    "Stefan Horoi",
                    "B. Richards",
                    "D. Bergeron",
                    "C. Iorio-Morin",
                    "M. Bonizzato",
                    "Nathalie Orr Gaucher",
                    "E. Racine",
                    "A. Weil",
                    "Ezekiel Williams",
                    "C. Bredenberg",
                    "A. Payeur",
                    "A. Orsborn",
                    "Xu Ji",
                    "Eric Elmoznino",
                    "George Deane",
                    "Axel Constant",
                    "G. Dumas",
                    "Jonathan Simon",
                    "Ximeng Mao",
                    "Yao-Chuan Chang",
                    "S. Zanos",
                    "Sangnie Bhardwaj",
                    "Willie McClinton",
                    "Tongzhou Wang",
                    "Chen Sun",
                    "Phillip Isola",
                    "Dilip Krishnan",
                    "Victor Geadah",
                    "Laura E. Su\u00e1rez",
                    "\u00c1goston Mihalik",
                    "Filip Milisav",
                    "Kenji Marshall",
                    "Mingze Li",
                    "P. V\u00e9rtes",
                    "B. Mi\u0161i\u0107",
                    "Mohammad-Javad Davari",
                    "A. Natik",
                    "Eugene Belilovsky",
                    "Sarthak Mittal",
                    "D. Rolnick",
                    "T. Deleu",
                    "David Kanaa",
                    "Leo Feng",
                    "Pierre-Luc Bacon",
                    "Yuhan Helena Liu",
                    "Arna Ghosh",
                    "E. Shea-Brown",
                    "E. L. Busch",
                    "Je-chun Huang",
                    "Andrew Benz",
                    "Tom Wallenstein",
                    "Smita Krishnaswamy",
                    "N. Turk-Browne",
                    "Thomas George",
                    "A. Baratin",
                    "L'eo Gagnon"
                ],
                "domain": [
                    "Brain-Computer Interface",
                    "Neural Networks",
                    "Meta-Learning",
                    "Cognitive Neuroscience"
                ],
                "publications": [
                    {
                        "title": "Use of Invasive Brain-Computer Interfaces in Pediatric Neurosurgery: Technical and Ethical Considerations",
                        "abstract": "Invasive brain-computer interfaces hold promise to alleviate disabilities in individuals with neurologic injury, with fully implantable brain-computer interface systems expected to reach the clinic in the upcoming decade. Children with severe neurologic disabilities, like quadriplegic cerebral palsy or cervical spine trauma, could benefit from this technology. However, they have been excluded from clinical trials of intracortical brain-computer interface to date. In this manuscript, we discuss the ethical considerations related to the use of invasive brain-computer interface in children with severe neurologic disabilities. We first review the technical hardware and software considerations for the application of intracortical brain-computer interface in children. We then discuss ethical issues related to motor brain-computer interface use in pediatric neurosurgery. Finally, based on the input of a multidisciplinary panel of experts in fields related to brain-computer interface (functional and restorative neurosurgery, pediatric neurosurgery, mathematics and artificial intelligence research, neuroengineering, pediatric ethics, and pragmatic ethics), we then formulate initial recommendations regarding the clinical use of invasive brain-computer interfaces in children."
                    },
                    {
                        "title": "Flexible Phase Dynamics for Bio-Plausible Contrastive Learning",
                        "abstract": "Many learning algorithms used as normative models in neuroscience or as candidate approaches for learning on neuromorphic chips learn by contrasting one set of network states with another. These Contrastive Learning (CL) algorithms are traditionally implemented with rigid, temporally non-local, and periodic learning dynamics that could limit the range of physical systems capable of harnessing CL. In this study, we build on recent work exploring how CL might be implemented by biological or neurmorphic systems and show that this form of learning can be made temporally local, and can still function even if many of the dynamical requirements of standard training procedures are relaxed. Thanks to a set of general theorems corroborated by numerical experiments across several CL models, our results provide theoretical foundations for the study and development of CL methods for biological and neuromorphic neural networks."
                    },
                    {
                        "title": "Neural manifolds and learning regimes in neural-interface tasks",
                        "abstract": "Neural activity tends to reside on manifolds whose dimension is lower than the dimension of the whole neural state space. Experiments using brain-computer interfaces (BCIs) with microelectrode arrays implanted in the motor cortex of nonhuman primates have provided ways to test whether neural manifolds influence learning-related neural computations. Starting from a learned BCI-controlled motor task, these experiments explored the effect of changing the BCI decoder to implement perturbations that were either \u201caligned\u201d or not with the pre-existing neural manifold. In a series of studies, researchers found that within-manifold perturbations (WMPs) evoked fast reassociations of existing neural patterns for rapid adaptation, while outside-manifold perturbations (OMPs) triggered a slower adaptation process that led to the emergence of new neural patterns. Together, these findings have been interpreted as suggesting that these different rates of adaptation might be associated with distinct learning mechanisms. Here, we investigated whether gradient-descent learning could alone explain these differences. Using an idealized model that captures the fixed-point dynamics of recurrent neural networks, we uncovered gradient-based learning dynamics consistent with experimental findings. Crucially, this experimental match arose only when the network was initialized in a lazier learning regime, a concept inherited from deep learning theory. A lazy learning regime\u2014in contrast with a rich regime\u2014implies small changes on synaptic strengths throughout learning. For OMPs, these small changes were less effective at increasing performance and could lead to unstable adaptation with a heightened sensitivity to learning rates. For WMPs, they helped reproduce the reassociation mechanism on short adaptation time scales, especially with large input variances. Since gradient descent has many biologically plausible variants, our findings establish lazy gradient-based learning as a plausible mechanism for adaptation under network-level constraints and unify several experimental results from the literature."
                    },
                    {
                        "title": "Sources of richness and ineffability for phenomenally conscious states",
                        "abstract": "Abstract Conscious states\u2014state that there is something it is like to be in\u2014seem both rich or full of detail and ineffable or hard to fully describe or recall. The problem of ineffability, in particular, is a longstanding issue in philosophy that partly motivates the explanatory gap: the belief that consciousness cannot be reduced to underlying physical processes. Here, we provide an information theoretic dynamical systems perspective on the richness and ineffability of consciousness. In our framework, the richness of conscious experience corresponds to the amount of information in a conscious state and ineffability corresponds to the amount of information lost at different stages of processing. We describe how attractor dynamics in working memory would induce impoverished recollections of our original experiences, how the discrete symbolic nature of language is insufficient for describing the rich and high-dimensional structure of experiences, and how similarity in the cognitive function of two individuals relates to improved communicability of their experiences to each other. While our model may not settle all questions relating to the explanatory gap, it makes progress toward a fully physicalist explanation of the richness and ineffability of conscious experience\u2014two important aspects that seem to be part of what makes qualitative character so puzzling."
                    },
                    {
                        "title": "Rapidly Inferring Personalized Neurostimulation Parameters with Meta-Learning: A Case Study of Individualized Fiber Recruitment in Vagus Nerve Stimulation",
                        "abstract": "Objective Neurostimulation is emerging as treatment for several diseases of the brain and peripheral organs. Due to variability arising from placement of stimulation devices, underlying neuroanatomy and physiological responses to stimulation, it is essential that neurostimulation protocols are personalized to maximize efficacy and safety. Building such personalized protocols would benefit from accumulated information in increasingly large datasets of other individuals\u2019 responses. Approach To address that need, we propose a meta-learning family of algorithms to conduct few-shot optimization of key fitting parameters of physiological and neural responses in new individuals. While our method is agnostic to neurostimulation setting, here we demonstrate its effectiveness on the problem of physiological modeling of fiber recruitment during vagus nerve stimulation (VNS). Using data from acute VNS experiments, the mapping between amplitudes of stimulus-evoked compound action potentials (eCAPs) and physiological responses, such as heart rate and breathing interval modulation, is inferred. Main results Using additional synthetic data sets to complement experimental results, we demonstrate that our meta-learning framework is capable of directly modeling the physiology-eCAP relationship for individual subjects with much fewer individually queried data points than standard methods. Significance Our meta-learning framework is general and can be adapted to many input-response neurostimulation mapping problems. Moreover, this method leverages information from growing data sets of past patients, as a treatment is deployed. It can also be combined with several model types, including regression, Gaussian processes with Bayesian optimization, and beyond."
                    },
                    {
                        "title": "Steerable Equivariant Representation Learning",
                        "abstract": "Pre-trained deep image representations are useful for post-training tasks such as classification through transfer learning, image retrieval, and object detection. Data augmentations are a crucial aspect of pre-training robust representations in both supervised and self-supervised settings. Data augmentations explicitly or implicitly promote invariance in the embedding space to the input image transformations. This invariance reduces generalization to those downstream tasks which rely on sensitivity to these particular data augmentations. In this paper, we propose a method of learning representations that are instead equivariant to data augmentations. We achieve this equivariance through the use of steerable representations. Our representations can be manipulated directly in embedding space via learned linear maps. We demonstrate that our resulting steerable and equivariant representations lead to better performance on transfer learning and robustness: e.g. we improve linear probe top-1 accuracy by between 1% to 3% for transfer; and ImageNet-C accuracy by upto 3.4%. We further show that the steerability of our representations provides significant speedup (nearly 50x) for test-time augmentations; by applying a large number of augmentations for out-of-distribution detection, we significantly improve OOD AUC on the ImageNet-C dataset over an invariant representation."
                    },
                    {
                        "title": "Neural networks with optimized single-neuron adaptation uncover biologically plausible regularization",
                        "abstract": "Neurons in the brain have rich and adaptive input-output properties. Features such as heterogeneous f-I curves and spike frequency adaptation are known to place single neurons in optimal coding regimes when facing changing stimuli. Yet, it is still unclear how brain circuits exploit single-neuron flexibility, and how network-level requirements may have shaped such cellular function. To answer this question, a multi-scaled approach is needed where the computations of single neurons and neural circuits must be considered as a complete system. In this work, we use artificial neural networks to systematically investigate single-neuron input-output adaptive mechanisms, optimized in an end-to-end fashion. Throughout the optimization process, each neuron has the liberty to modify its nonlinear activation function, parametrized to mimic f-I curves of biological neurons, and to learn adaptation strategies to modify activation functions in real-time during a task. We find that such networks show much-improved robustness to noise and changes in input statistics. Importantly, we find that this procedure recovers precise coding strategies found in biological neurons, such as gain scaling and fractional order differentiation/integration. Using tools from dynamical systems theory, we analyze the role of these emergent single-neuron properties and argue that neural diversity and adaptation play an active regularization role, enabling neural circuits to optimally propagate information across time."
                    },
                    {
                        "title": "conn2res: A toolbox for connectome-based reservoir computing",
                        "abstract": "The connection patterns of neural circuits form a complex network. How signaling in these circuits manifests as complex cognition and adaptive behaviour remains the central question in neuroscience. Concomitant advances in connectomics and artificial intelligence open fundamentally new opportunities to understand how connection patterns shape computational capacity in biological brain networks. Reservoir computing is a versatile paradigm that uses nonlinear dynamics of high-dimensional dynamical systems to perform computations and approximate cognitive functions. Here we present conn2res: an open-source Python toolbox for implementing biological neural networks as artificial neural networks. conn2res is modular, allowing arbitrary architectures and arbitrary dynamics to be imposed. The toolbox allows researchers to input connectomes reconstructed using multiple techniques, from tract tracing to noninvasive diffusion imaging, and to impose multiple dynamical systems, from simple spiking neurons to memristive dynamics. The versatility of the conn2res toolbox allows us to ask new questions at the confluence of neuroscience and artificial intelligence. By reconceptualizing function as computation, conn2res sets the stage for a more mechanistic understanding of structure-function relationships in brain networks."
                    },
                    {
                        "title": "O N THE I NADEQUACY OF CKA AS A M EASURE OF S IMILARITY IN D EEP L EARNING",
                        "abstract": "Comparing learned representations is a challenging problem which has been approached in different ways. The CKA similarity metric, particularly it\u2019s linear variant, has recently become a popular approach and has been widely used to compare representations of a network\u2019s different layers, of similar networks trained differently"
                    },
                    {
                        "title": "I NDUCTIVE B IASES FOR R ELATIONAL T ASKS",
                        "abstract": "Current deep learning approaches have shown good in-distribution performance but struggle in out-of-distribution settings. This is especially true in the case of tasks involving abstract relations like recognizing rules in sequences, as required in many intelligence tests. In contrast, our brains are remarkably flexible at such tasks, an attribute that is likely linked to anatomical constraints on computations. Inspired by this, recent work has explored how enforcing that relational representations remain distinct from sensory representations can help artificial systems. Building on this work, we further explore and formalize the advantages afforded by \u201cpartitioned\u201d representations of relations and sensory details. We investigate inductive biases that ensure abstract relations are learned and represented distinctly from sensory data across several neural network architectures and show that they outperform existing architectures on out-of-distribution generalization for various relational tasks. These results show that partitioning relational representations from other information streams may be a simple way to augment existing network architectures\u2019 robustness when performing relational computations."
                    },
                    {
                        "title": "Continuous-Time Meta-Learning with Forward Mode Differentiation",
                        "abstract": "Drawing inspiration from gradient-based meta-learning methods with infinitely small gradient steps, we introduce Continuous-Time Meta-Learning (COMLN), a meta-learning algorithm where adaptation follows the dynamics of a gradient vector field. Specifically, representations of the inputs are meta-learned such that a task-specific linear classifier is obtained as a solution of an ordinary differential equation (ODE). Treating the learning process as an ODE offers the notable advantage that the length of the trajectory is now continuous, as opposed to a fixed and discrete number of gradient steps. As a consequence, we can optimize the amount of adaptation necessary to solve a new task using stochastic gradient descent, in addition to learning the initial conditions as is standard practice in gradient-based meta-learning. Importantly, in order to compute the exact meta-gradients required for the outer-loop updates, we devise an efficient algorithm based on forward mode differentiation, whose memory requirements do not scale with the length of the learning trajectory, thus allowing longer adaptation in constant memory. We provide analytical guarantees for the stability of COMLN, we show empirically its efficiency in terms of runtime and memory usage, and we illustrate its effectiveness on a range of few-shot image classification problems."
                    },
                    {
                        "title": "Beyond accuracy: generalization properties of bio-plausible temporal credit assignment rules",
                        "abstract": "To unveil how the brain learns, ongoing work seeks biologically-plausible approximations of gradient descent algorithms for training recurrent neural networks (RNNs). Yet, beyond task accuracy, it is unclear if such learning rules converge to solutions that exhibit different levels of generalization than their nonbiologically-plausible counterparts. Leveraging results from deep learning theory based on loss landscape curvature, we ask: how do biologically-plausible gradient approximations affect generalization? We first demonstrate that state-of-the-art biologically-plausible learning rules for training RNNs exhibit worse and more variable generalization performance compared to their machine learning counterparts that follow the true gradient more closely. Next, we verify that such generalization performance is correlated significantly with loss landscape curvature, and we show that biologically-plausible learning rules tend to approach high-curvature regions in synaptic weight space. Using tools from dynamical systems, we derive theoretical arguments and present a theorem explaining this phenomenon. This predicts our numerical results, and explains why biologically-plausible rules lead to worse and more variable generalization properties. Finally, we suggest potential remedies that could be used by the brain to mitigate this effect. To our knowledge, our analysis is the first to identify the reason for this generalization gap between artificial and biologically-plausible learning rules, which can help guide future investigations into how the brain learns solutions that generalize."
                    },
                    {
                        "title": "Temporal PHATE: A multi-view manifold learning method for brain state trajectories",
                        "abstract": "Brain activity as measured with functional magnetic resonance imaging (fMRI) gives the illusion of intractably high dimensionality, rife with collection and biological noise. Non-linear dimensionality reductions like PCA, UMAP, tSNE, and PHATE have proven useful for high-throughput biomedical data, but have not been extensively used in fMRI, which is known to re\ufb02ect the redundancy and co-modulation of neural population activity. Here we take the manifold-geometry preserving method PHATE and extend it for use in brain activity timeseries data in a method we call temporal PHATE (T-PHATE). We observe that in addition to the intrinsically lower dimensionality of fMRI data, it also has signi\ufb01cant autocorrelative structure that we can exploit to faithfully denoise the signal and learn brain activation manifolds. We empirically validate T-PHATE on three fMRI tasks and show that T-PHATE manifolds improve visualization \ufb01delity, stimulus feature classi\ufb01cation, and neural event segmentation. T-PHATE demonstrates impressive improvements over previous cutting-edge approaches to understanding the nature of cognition from fMRI and bodes potential applications broadly for high-dimensional datasets of temporally-diffuse processes."
                    },
                    {
                        "title": "Lazy vs hasty: linearization in deep networks impacts learning schedule based on example difficulty",
                        "abstract": "Among attempts at giving a theoretical account of the success of deep neural networks, a recent line of work has identified a so-called lazy training regime in which the network can be well approximated by its linearization around initialization. Here we investigate the comparative effect of the lazy (linear) and feature learning (non-linear) regimes on subgroups of examples based on their difficulty. Specifically, we show that easier examples are given more weight in feature learning mode, resulting in faster training compared to more difficult ones. In other words, the non-linear dynamics tends to sequentialize the learning of examples of increasing difficulty. We illustrate this phenomenon across different ways to quantify example difficulty, including c-score, label noise, and in the presence of easy-to-learn spurious correlations. Our results reveal a new understanding of how deep networks prioritize resources across example difficulty."
                    },
                    {
                        "title": "Clarifying MCMC-based training of modern EBMs : Contrastive Divergence versus Maximum Likelihood",
                        "abstract": "The Energy-Based Model (EBM) framework is a very general approach to generative modeling that tries to learn and exploit probability distributions only defined though unnormalized scores. It has risen in popularity recently thanks to the impressive results obtained in image generation by parameterizing the distribution with Convolutional Neural Networks (CNN). However, the motivation and theoretical foundations behind modern EBMs are often absent from recent papers and this sometimes results in some confusion. In particular, the theoretical justifications behind the popular MCMC-based learning algorithm Contrastive Divergence (CD) are often glossed over and we find that this leads to theoretical errors in recent influential papers (Du&Mordatch, 2019; Du et al., 2020). After offering a first-principles introduction of MCMC-based training, we argue that the learning algorithm they use can in fact not be described as CD and reinterpret theirs methods in light of a new interpretation. Finally, we discuss the implications of our new interpretation and provide some illustrative experiments."
                    }
                ]
            },
            "c5feeeb2-a827-4f15-82e1-15f20abb45aa": {
                "pk": "c5feeeb2-a827-4f15-82e1-15f20abb45aa",
                "name": "Blake Richards",
                "collaborators": [
                    "Y. Bengio",
                    "Guillaume Lajoie",
                    "T. Lillicrap",
                    "Arna Ghosh",
                    "Adam Santoro",
                    "Chen Sun",
                    "Wannan Yang",
                    "Benjamin Alsbury-Nealy",
                    "Richard Naud",
                    "Xingyun Wang",
                    "Zachary Friedenberger",
                    "A. Payeur",
                    "Jiyun N. Shin",
                    "J. B\u00e9\u00efque",
                    "Moritz Dr\u00fcke",
                    "M. Larkum",
                    "Guy Doron",
                    "K. Kording",
                    "Pingsheng Li",
                    "Jonathan H. Cornford",
                    "Giancarlo Kerg",
                    "Sarthak Mittal",
                    "D. Rolnick",
                    "Yuhan Helena Liu",
                    "E. Shea-Brown",
                    "Anthony GX-Chen",
                    "Veronica Chelu",
                    "Joelle Pineau",
                    "A. Zador",
                    "Bence Olveczky",
                    "Sean Escola",
                    "K. Boahen",
                    "M. Botvinick",
                    "D. Chklovskii",
                    "A. Churchland",
                    "C. Clopath",
                    "J. DiCarlo",
                    "S. Ganguli",
                    "J. Hawkins",
                    "Konrad Koerding",
                    "A. Koulakov",
                    "Yann LeCun",
                    "Adam H. Marblestone",
                    "B. Olshausen",
                    "A. Pouget",
                    "Cristina Savin",
                    "T. Sejnowski",
                    "Eero P. Simoncelli",
                    "S. Solla",
                    "David Sussillo",
                    "A. Tolias",
                    "Doris Y. Tsao",
                    "Josh Abramson",
                    "Arun Ahuja",
                    "Federico Carnevale",
                    "Petko Georgiev",
                    "Alex Goldin",
                    "Alden Hung",
                    "Jessica Landon",
                    "Alistair Muldal",
                    "Tamara von Glehn",
                    "Greg Wayne",
                    "Nathaniel Wong",
                    "Chen Yan",
                    "Damjan Kalajdzievski",
                    "Ximeng Mao",
                    "Pascal Fortier-Poisson",
                    "Thomas Jiralerspong",
                    "Dane Malenfant",
                    "M. Ernoult",
                    "Fabrice Normandin",
                    "A. Moudgil",
                    "Sean Spinney",
                    "Eugene Belilovsky",
                    "I. Rish",
                    "Lina M. Tran",
                    "Lulu Liu",
                    "S. Josselyn",
                    "P. Frankland",
                    "Raymond Chua",
                    "Doina Precup",
                    "Christos Kaplanis"
                ],
                "domain": [
                    "Neuroscience",
                    "Reinforcement Learning",
                    "Machine Learning",
                    "Cognitive Science"
                ],
                "publications": [
                    {
                        "title": "Fast burst fraction transients convey information independent of the firing rate",
                        "abstract": "Theories of attention and learning have hypothesized a central role for high-frequency bursting in cognitive functions, but experimental reports of burst-mediated representations in vivo have been limited. Here we used a novel demultiplexing approach by considering a conjunctive burst code. We studied this code in vivo while animals learned to report direct electrical stimulation of the somatosensory cortex and found two acquired yet independent representations. One code, the event rate, showed a sparse and succint stiumulus representation and a small modulation upon detection errors. The other code, the burst fraction, correlated more globally with stimulation and more promptly responded to detection errors. Bursting modulation was potent and its time course evolved, even in cells that were considered unresponsive based on the firing rate. During the later stages of training, this modulation in bursting happened earlier, gradually aligning temporally with the representation in event rate. The alignment of bursting and event rate modulation sharpened the firing rate response, and was strongly associated behavioral accuracy. Thus a fine-grained separation of spike timing patterns reveals two signals that accompany stimulus representations: an error signal that can be essential to guide learning and a sharpening signal that could implement attention mechanisms."
                    },
                    {
                        "title": "The study of plasticity has always been about gradients",
                        "abstract": "The experimental study of learning and plasticity has always been driven by an implicit question: how can physiological changes be adaptive and improve performance? For example, in Hebbian plasticity only synapses from presynaptic neurons that were active are changed, avoiding useless changes. Similarly, in dopamine\u2010gated learning synapse changes depend on reward or lack thereof and do not change when everything is predictable. Within machine learning we can make the question of which changes are adaptive concrete: performance improves when changes correlate with the gradient of an objective function quantifying performance. This result is general for any system that improves through small changes. As such, physiology has always implicitly been seeking mechanisms that allow the brain to approximate gradients. Coming from this perspective we review the existing literature on plasticity\u2010related mechanisms, and we show how these mechanisms relate to gradient estimation. We argue that gradients are a unifying idea to explain the many facets of neuronal plasticity."
                    },
                    {
                        "title": "Learning better with Dale\u2019s Law: A Spectral Perspective",
                        "abstract": "Most recurrent neural networks (RNNs) do not include a fundamental constraint of real neural circuits: Dale\u2019s Law, which implies that neurons must be excitatory (E) or inhibitory (I). Dale\u2019s Law is generally absent from RNNs because simply partitioning a standard network\u2019s units into E and I populations impairs learning. However, here we extend a recent feedforward bio-inspired EI network architecture, named Dale\u2019s ANNs, to recurrent networks, and demonstrate that good performance is possible while respecting Dale\u2019s Law. This begs the question: What makes some forms of EI network learn poorly and others learn well? And, why does the simple approach of incorporating Dale\u2019s Law impair learning? Historically the answer was thought to be the sign constraints on EI network parameters, and this was a motivation behind Dale\u2019s ANNs. However, here we show the spectral properties of the recurrent weight matrix at initialisation are more impactful on network performance than sign constraints. We find that simple EI partitioning results in a singular value distribution that is multimodal and dispersed, whereas standard RNNs have an unimodal, more clustered singular value distribution, as do recurrent Dale\u2019s ANNs. We also show that the spectral properties and performance of partitioned EI networks are worse for small networks with fewer I units, and we present normalised SVD entropy as a measure of spectrum pathology that correlates with performance. Overall, this work sheds light on a long-standing mystery in neuroscience-inspired AI and computational neuroscience, paving the way for greater alignment between neural networks and biology."
                    },
                    {
                        "title": "I NDUCTIVE B IASES FOR R ELATIONAL T ASKS",
                        "abstract": "Current deep learning approaches have shown good in-distribution performance but struggle in out-of-distribution settings. This is especially true in the case of tasks involving abstract relations like recognizing rules in sequences, as required in many intelligence tests. In contrast, our brains are remarkably flexible at such tasks, an attribute that is likely linked to anatomical constraints on computations. Inspired by this, recent work has explored how enforcing that relational representations remain distinct from sensory representations can help artificial systems. Building on this work, we further explore and formalize the advantages afforded by \u201cpartitioned\u201d representations of relations and sensory details. We investigate inductive biases that ensure abstract relations are learned and represented distinctly from sensory data across several neural network architectures and show that they outperform existing architectures on out-of-distribution generalization for various relational tasks. These results show that partitioning relational representations from other information streams may be a simple way to augment existing network architectures\u2019 robustness when performing relational computations."
                    },
                    {
                        "title": "Beyond accuracy: generalization properties of bio-plausible temporal credit assignment rules",
                        "abstract": "To unveil how the brain learns, ongoing work seeks biologically-plausible approximations of gradient descent algorithms for training recurrent neural networks (RNNs). Yet, beyond task accuracy, it is unclear if such learning rules converge to solutions that exhibit different levels of generalization than their nonbiologically-plausible counterparts. Leveraging results from deep learning theory based on loss landscape curvature, we ask: how do biologically-plausible gradient approximations affect generalization? We first demonstrate that state-of-the-art biologically-plausible learning rules for training RNNs exhibit worse and more variable generalization performance compared to their machine learning counterparts that follow the true gradient more closely. Next, we verify that such generalization performance is correlated significantly with loss landscape curvature, and we show that biologically-plausible learning rules tend to approach high-curvature regions in synaptic weight space. Using tools from dynamical systems, we derive theoretical arguments and present a theorem explaining this phenomenon. This predicts our numerical results, and explains why biologically-plausible rules lead to worse and more variable generalization properties. Finally, we suggest potential remedies that could be used by the brain to mitigate this effect. To our knowledge, our analysis is the first to identify the reason for this generalization gap between artificial and biologically-plausible learning rules, which can help guide future investigations into how the brain learns solutions that generalize."
                    },
                    {
                        "title": "A Generalized Bootstrap Target for Value-Learning, Efficiently Combining Value and Feature Predictions",
                        "abstract": "Estimating value functions is a core component of reinforcement learning algorithms. Temporal difference (TD) learning algorithms use bootstrapping, i.e. they update the value function toward a learning target using value estimates at subsequent time-steps. Alternatively, the value function can be updated toward a learning target constructed by separately predicting successor features (SF)\u2014a policy-dependent model\u2014and linearly combining them with instantaneous rewards.    We focus on bootstrapping targets used when estimating value functions, and propose a new backup target, the ?-return mixture, which implicitly combines value-predictive knowledge (used by TD methods) with (successor) feature-predictive knowledge\u2014with a parameter ? capturing how much to rely on each. We illustrate that incorporating predictive knowledge through an ??-discounted SF model makes more efficient use of sampled experience, compared to either extreme, i.e. bootstrapping entirely on the value function estimate, or bootstrapping on the product of separately estimated successor features and instantaneous reward models. We empirically show this approach leads to faster policy evaluation and better control performance, for tabular and nonlinear function approximations, indicating scalability and generality."
                    },
                    {
                        "title": "The Brain-Computer Metaphor Debate Is Useless: A Matter of Semantics",
                        "abstract": "It is commonly assumed that usage of the word \u201ccomputer\u201d in the brain sciences reflects a metaphor. However, there is no single definition of the word \u201ccomputer\u201d in use. In fact, based on the usage of the word \u201ccomputer\u201d in computer science, a computer is merely some physical machinery that can in theory compute any computable function. According to this definition the brain is literally a computer; there is no metaphor. But, this deviates from how the word \u201ccomputer\u201d is used in other academic disciplines. According to the definition used outside of computer science, \u201ccomputers\u201d are human-made devices that engage in sequential processing of inputs to produce outputs. According to this definition, brains are not computers, and arguably, computers serve as a weak metaphor for brains. Thus, we argue that the recurring brain-computer metaphor debate is actually just a semantic disagreement, because brains are either literally computers or clearly not very much like computers at all, depending on one's definitions. We propose that the best path forward is simply to put the debate to rest, and instead, have researchers be clear about which definition they are using in their work. In some circumstances, one can use the definition from computer science and simply ask, what type of computer is the brain? In other circumstances, it is important to use the other definition, and to clarify the ways in which our brains are radically different from the laptops, smartphones, and servers that surround us in modern life."
                    },
                    {
                        "title": "Toward Next-Generation Artificial Intelligence: Catalyzing the NeuroAI Revolution",
                        "abstract": ": Neuroscience has long been an important driver of progress in artificial intelligence (AI). We propose that to accelerate progress in AI, we must invest in fundamental research in NeuroAI. Over the coming decades, Artificial Intelligence (AI) will transform society and the world economy in ways that are as profound as the computer revolution of the last half century, and likely at an even faster pace. This AI revolution presents tremendous opportunities to unleash human creativity in the modern economy. New developments in AI systems have the potential to enable workers to attain greater productivity and relieve them from performing the most dangerous and menial jobs. But, to reach this potential, we still require advances that will make AI more human-like in its capabilities. Historically, neuroscience has been a key driver and source of inspiration for improvements in AI, particularly those that made AI more proficient in areas that humans and other animals excel at, such as vision, reward-based learning, interacting with the physical world, and language (Hassabis et al. 2017). It can still play this role. To accelerate progress in AI and realize its vast potential, we must invest in fundamental research in \u201cNeuroAI\u201d. The by to brains Pitts an \u201cartificial brain\u201d John von upon the very limited knowledge of the brain in the other animals. Because each animal has its own unique set of abilities, each animal defines its own embodied Turing test: An artificial beaver might be tested on its ability to build a dam, and an artificial squirrel on its ability to jump through trees. Nonetheless, many core sensorimotor capabilities are shared by almost all animals, and the ability of animals to rapidly evolve the sensorimotor skills needed to adapt to new environments suggests that these core skills provide a solid foundation. Below we highlight a few of these shared characteristics."
                    },
                    {
                        "title": "Evaluating Multimodal Interactive Agents",
                        "abstract": "Creating agents that can interact naturally with humans is a common goal in artificial intelligence (AI) research. However, evaluating these interactions is challenging: collecting online human-agent interactions is slow and expensive, yet faster proxy metrics often do not correlate well with interactive evaluation. In this paper, we assess the merits of these existing evaluation metrics and present a novel approach to evaluation called the Standardised Test Suite (STS). The STS uses behavioural scenarios mined from real human interaction data. Agents see replayed scenario context, receive an instruction, and are then given control to complete the interaction offline. These agent continuations are recorded and sent to human annotators to mark as success or failure, and agents are ranked according to the proportion of continuations in which they succeed. The resulting STS is fast, controlled, interpretable, and representative of naturalistic interactions. Altogether, the STS consolidates much of what is desirable across many of our standard evaluation metrics, allowing us to accelerate research progress towards producing agents that can interact naturally with humans. A video may be found at https://youtu.be/YR1TngGORGQ."
                    },
                    {
                        "title": "Contrastive introspection (ConSpec) to rapidly identify invariant steps for success",
                        "abstract": "Reinforcement learning (RL) algorithms have achieved notable success in recent years, but still struggle with fundamental issues in long-term credit assignment. It remains di\ufb03cult to learn in situations where success is contingent upon multiple critical steps that are distant in time from each other and from a sparse reward; as is often the case in real life. Moreover, how RL algorithms assign credit in these di\ufb03cult situations is typically not coded in a way that can rapidly generalize to new situations. Here, we present an approach using o\ufb04ine contrastive learning, which we call contrastive introspection (ConSpec), that can be added to any existing RL algorithm and addresses both issues. In ConSpec, a contrastive loss is used during o\ufb04ine replay to identify invariances among successful episodes. This takes advantage of the fact that it is easier to retrospectively identify the small set of steps that success is contingent upon than it is to prospectively predict reward at every step taken in the environment. ConSpec stores this knowledge in a collection of prototypes summarizing the intermediate states required for success. During training, arrival at any state that matches these prototypes generates an intrinsic reward that is added to any external rewards. As well, the reward shaping provided by ConSpec can be made to preserve the optimal policy of the underlying RL agent. The prototypes in ConSpec provide two key bene\ufb01ts for credit assignment: (1) They enable rapid identi\ufb01cation of all the critical states. (2) They do so in a readily interpretable manner, enabling out of distribution generalization when sensory features are altered. In summary, ConSpec is a modular system that can be added to any existing RL algorithm to improve its long-term credit assignment."
                    },
                    {
                        "title": "Transfer Entropy Bottleneck: Learning Sequence to Sequence Information Transfer",
                        "abstract": "When presented with a data stream of two statistically dependent variables, predicting the future of one of the variables (the target stream) can benefit from information about both its history and the history of the other variable (the source stream). For example, fluctuations in temperature at a weather station can be predicted using both temperatures and barometric readings. However, a challenge when modelling such data is that it is easy for a neural network to rely on the greatest joint correlations within the target stream, which may ignore a crucial but small information transfer from the source to the target stream. As well, there are often situations where the target stream may have previously been modelled independently and it would be useful to use that model to inform a new joint model. Here, we develop an information bottleneck approach for conditional learning on two dependent streams of data. Our method, which we call Transfer Entropy Bottleneck (TEB), allows one to learn a model that bottlenecks the directed information transferred from the source variable to the target variable, while quantifying this information transfer within the model. As such, TEB provides a useful new information bottleneck approach for modelling two statistically dependent streams of data in order to make predictions about one of them."
                    },
                    {
                        "title": "Contrastive Retrospection: honing in on critical steps for rapid learning and generalization in RL",
                        "abstract": "In real life, success is often contingent upon multiple critical steps that are distant in time from each other and from the final reward. These critical steps are challenging to identify with traditional reinforcement learning (RL) methods that rely on the Bellman equation for credit assignment. Here, we present a new RL algorithm that uses offline contrastive learning to hone in on these critical steps. This algorithm, which we call Contrastive Retrospection (ConSpec), can be added to any existing RL algorithm. ConSpec learns a set of prototypes for the critical steps in a task by a novel contrastive loss and delivers an intrinsic reward when the current state matches one of the prototypes. The prototypes in ConSpec provide two key benefits for credit assignment: (i) They enable rapid identification of all the critical steps. (ii) They do so in a readily interpretable manner, enabling out-of-distribution generalization when sensory features are altered. Distinct from other contemporary RL approaches to credit assignment, ConSpec takes advantage of the fact that it is easier to retrospectively identify the small set of steps that success is contingent upon (and ignoring other states) than it is to prospectively predict reward at every taken step. ConSpec greatly improves learning in a diverse set of RL tasks. The code is available at the link: https://github.com/sunchipsster1/ConSpec"
                    },
                    {
                        "title": "Towards Scaling Difference Target Propagation by Learning Backprop Targets",
                        "abstract": "The development of biologically-plausible learning algorithms is important for understanding learning in the brain, but most of them fail to scale-up to real-world tasks, limiting their potential as explanations for learning by real brains. As such, it is important to explore learning algorithms that come with strong theoretical guarantees and can match the performance of backpropagation (BP) on complex tasks. One such algorithm is Difference Target Propagation (DTP), a biologically-plausible learning algorithm whose close relation with Gauss-Newton (GN) optimization has been recently established. However, the conditions under which this connection rigorously holds preclude layer-wise training of the feedback pathway synaptic weights (which is more biologically plausible). Moreover, good alignment between DTP weight updates and loss gradients is only loosely guaranteed and under very specific conditions for the architecture being trained. In this paper, we propose a novel feedback weight training scheme that ensures both that DTP approximates BP and that layer-wise feedback weight training can be restored without sacrificing any theoretical guarantees. Our theory is corroborated by experimental results and we report the best performance ever achieved by DTP on CIFAR-10 and ImageNet 32$\\times$32"
                    },
                    {
                        "title": "Can neurogenesis act as a neural regularizer?",
                        "abstract": "New neurons are continuously generated in the subgranular zone of the dentate gyrus throughout adulthood. These new neurons gradually integrate into hippocampal circuits, forming new na\u00efve synapses. Viewed from this perspective, these new neurons may represent a significant source of \u2018wiring\u2019 noise in hippocampal networks. In machine learning, such noise injection is commonly used as a regularization technique. Regularization techniques help prevent overfitting training data, and allow models to generalize learning to new, unseen data. Using a computational modeling approach, here we ask whether a neurogenesis-like process similarly acts as a regularizer, facilitating generalization in a category learning task. In a convolutional neural network (CNN) trained on the CIFAR-10 object recognition dataset, we modeled neurogenesis as a replacement/turnover mechanism, where weights for a randomly chosen small subset of neurons in a chosen hidden layer were re-initialized to new values as the model learned to categorize 10 different classes of objects. We found that neurogenesis enhanced generalization on unseen test data compared to networks with no neurogenesis. Moreover, neurogenic networks either outperformed or performed similarly to networks with conventional noise injection (i.e., dropout, weight decay, and neural noise). These results suggest that neurogenesis can enhance generalization in hippocampal learning through noise-injection, expanding on the roles that neurogenesis may have in cognition. Author Summary In deep neural networks, various forms of noise injection are used as regularization techniques to prevent overfitting and promote generalization on unseen test data. Here, we were interested in whether adult neurogenesis\u2013 the lifelong production of new neurons in the hippocampus\u2013 might similarly function as a regularizer in the brain. We explored this question computationally, assessing whether implementing a neurogenesis-like process in a hidden layer within a convolutional neural network trained in a category learning task would prevent overfitting and promote generalization. We found that neurogenesis regularization was as least as effective as, or more effective than, conventional regularizers (i.e., dropout, weight decay and neural noise) in improving model performance. These results suggest that optimal levels of hippocampal neurogenesis may improve memory-guided decision making by preventing overfitting, thereby promoting the formation of more generalized memories that can be applied in a broader range of circumstances. We outline how these predictions may be evaluated behaviorally in rodents with altered hippocampal neurogenesis."
                    },
                    {
                        "title": "Continual Reinforcement Learning with Multi-Timescale Successor Features",
                        "abstract": "Learning and memory consolidation in the brain occur over multiple timescales. Inspired by this observation, it has been shown that catastrophic forgetting in reinforcement learning (RL) agents can be mitigated by consolidating Q-value function parameters at multiple timescales. In this work, we combine this approach with successor features, and show that by consolidating successor features and preferences learned over multiple timescales we can further mitigate catastrophic forgetting. In particular, we show that agents trained with this approach rapidly recall previously rewarding sites in large environments, whereas those trained without this decomposition and consolidation mechanism do not. These results therefore contribute to our understanding of the functional role of synaptic plasticity and memory systems operating at multiple timescales, and demonstrate that RL can be improved by capturing features of biological memory with greater fidelity."
                    }
                ]
            }
        }
    },
    "2405.20348": {
        "paper_data": {
            "title": "Personalized Adapter for Large Meteorology Model on Devices: Towards Weather Foundation Models",
            "url": "http://arxiv.org/abs/2405.20348v1",
            "arxiv_id": "2405.20348",
            "authors": [
                "Shengchao Chen",
                "Guodong Long",
                "Jing Jiang",
                "Chengqi Zhang"
            ],
            "abstract": "This paper demonstrates that pre-trained language models (PLMs) are strong foundation models for on-device meteorological variables modeling. We present LM-Weather, a generic approach to taming PLMs, that have learned massive sequential knowledge from the universe of natural language databases, to acquire an immediate capability to obtain highly customized models for heterogeneous meteorological data on devices while keeping high efficiency. Concretely, we introduce a lightweight personalized adapter into PLMs and endows it with weather pattern awareness. During communication between clients and the server, low-rank-based transmission is performed to effectively fuse the global knowledge among devices while maintaining high communication efficiency and ensuring privacy. Experiments on real-wold dataset show that LM-Weather outperforms the state-of-the-art results by a large margin across various tasks (e.g., forecasting and imputation at different scales). We provide extensive and in-depth analyses experiments, which verify that LM-Weather can (1) indeed leverage sequential knowledge from natural language to accurately handle meteorological sequence, (2) allows each devices obtain highly customized models under significant heterogeneity, and (3) generalize under data-limited and out-of-distribution (OOD) scenarios.",
            "introduction": "   1 Introduction  Accurately modeling weather variation pattern from large amount of meteorological variables sequences is increasingly vital for providing efficient weather analysis support for disaster warning. Recently, the promise of learning to understand weather pattern from data via deep learning (DL) has led to an ongoing paradigm shift apart from the long-established physics-based methods\u00a0[1, 2].   Mining potential patterns from meteorological sequences that collected from different regions, including forecasting and imputation, is one of the most important problems in meteorology. Significant progress has been made by several latest time series approaches\u00a0[1, 3, 4]. These approaches formulate meteorological variable modeling as an end-to-end spatio-temporal learning problem. This overlooks the reality that ground weather devices distributed globally gather vast amounts of data quickly. The sheer volume of data, coupled with limited network capacity, necessitates local processing on the devices, making centralised learning challenging\u00a0[5]. On-device intelligence enables edge devices to compute independently, offering a primary solution to the problem.   Figure 1: Overview. (a) Schematic of LM-Weather, each client using personalized adapter to endow the PLM for local weather awareness, only low-rank matrices are transmitted to enhance efficiency during communication; (b) Brief structure of PLM on each client, detailed architecture can be found in Appendix; (c) Task Adapter, the multivariate weather series input splits into two paths. The first path isolates the trend, seasonal, and residual elements, which each go through independent generator to produce specific adapters; (d) Architecture of the generator for each decomposed element; (e) Schematic diagram of Channel-Independent Patching\u00a0[6].   Federated Learning (FL)\u00a0[7] is a promising on-device intelligence implementation that collaboratively train a uniform model across devices without exchanging raw data. However, the model often underperform due to data heterogeneity among clients. Personalized FL (PFL) provides new insights for on-device intelligence that allows each device obtains customized models for providing personalized insights\u00a0[8, 9]. Albeit PFL methods showing revolutionized capability in this field, we argue that the current advancements are not necessarily at their best in on-device meteorological variable modeling as three major obstacles remain and hinder further progress:   (i)  Challenge of heterogeneity. Weather data\u2019s heterogeneity, unlike that of images or text, arises mainly from the unique characteristics of data collected by weather devices in various regions, such as tropical or arid areas. Furthermore, sensor malfunctions or extreme events can lead to collection disruptions or inconsistent missing data, which significantly increase the differences in data distribution across devices.    (ii)  Underperformed shallow network structures. The vast and varied data gathered by weather devices challenge simpler neural network models to generalize effectively. Furthermore, the frequent updates of weather data (hourly or by the minute) require neural models on devices to train and infer more often. This demand is hard to meet with deeper models that, while more performant, are also more resource-intensive.    (iii)  Resource-constrained weather devices. From a computation perspective, weather devices cannot afford of training complex neural models from scratch, especially for foundation models\u00a0[4]. From a communication perspective, transmitting complete model during the aggregation phase in FL/PFL significantly increases communication overhead, which is impractical for real-time weather modeling.    Therefore, a compact foundational model (FM) is crucial for personalized on-device weather modeling. Yet, there\u2019s a gap in FMs",
            "references": [
                {
                    "title": "Anomaly Detection Based on Isolation Mechanisms: A Survey",
                    "abstract": "Anomaly detection is a longstanding and active research area that has many applications in domains such as finance, security, and manufacturing. However, the efficiency and performance of anomaly detection algorithms are challenged by the large-scale, high-dimensional, and heterogeneous data that are prevalent in the era of big data. Isolation-based unsupervised anomaly detection is a novel and effective approach for identifying anomalies in data. It relies on the idea that anomalies are few and different from normal instances, and thus can be easily isolated by random partitioning. Isolation-based methods have several advantages over existing methods, such as low computational complexity, low memory usage, high scalability, robustness to noise and irrelevant features, and no need for prior knowledge or heavy parameter tuning. In this survey, we review the state-of-the-art isolation-based anomaly detection methods, including their data partitioning strategies, anomaly score functions, and algorithmic details. We also discuss some extensions and applications of isolation-based methods in different scenarios, such as detecting anomalies in streaming data, time series, trajectory, and image datasets. Finally, we identify some open challenges and future directions for isolation-based anomaly detection research."
                },
                {
                    "title": "Foundation Models for Weather and Climate Data Understanding: A Comprehensive Survey",
                    "abstract": "As artificial intelligence (AI) continues to rapidly evolve, the realm of Earth and atmospheric sciences is increasingly adopting data-driven models, powered by progressive developments in deep learning (DL). Specifically, DL techniques are extensively utilized to decode the chaotic and nonlinear aspects of Earth systems, and to address climate challenges via understanding weather and climate data. Cutting-edge performance on specific tasks within narrower spatio-temporal scales has been achieved recently through DL. The rise of large models, specifically large language models (LLMs), has enabled fine-tuning processes that yield remarkable outcomes across various downstream tasks, thereby propelling the advancement of general AI. However, we are still navigating the initial stages of crafting general AI for weather and climate. In this survey, we offer an exhaustive, timely overview of state-of-the-art AI methodologies specifically engineered for weather and climate data, with a special focus on time series and text data. Our primary coverage encompasses four critical aspects: types of weather and climate data, principal model architectures, model scopes and applications, and datasets for weather and climate. Furthermore, in relation to the creation and application of foundation models for weather and climate data understanding, we delve into the field's prevailing challenges, offer crucial insights, and propose detailed avenues for future research. This comprehensive approach equips practitioners with the requisite knowledge to make substantial progress in this domain. Our survey encapsulates the most recent breakthroughs in research on large, data-driven models for weather and climate data understanding, emphasizing robust foundations, current advancements, practical applications, crucial resources, and prospective research opportunities."
                },
                {
                    "title": "One Fits All: Universal Time Series Analysis by Pretrained LM and Specially Designed Adaptors",
                    "abstract": "Despite the impressive achievements of pre-trained models in the fields of natural language processing (NLP) and computer vision (CV), progress in the domain of time series analysis has been limited. In contrast to NLP and CV, where a single model can handle various tasks, time series analysis still relies heavily on task-specific methods for activities such as classification, anomaly detection, forecasting, and few-shot learning. The primary obstacle to developing a pre-trained model for time series analysis is the scarcity of sufficient training data. In our research, we overcome this obstacle by utilizing pre-trained models from language or CV, which have been trained on billions of data points, and apply them to time series analysis. We assess the effectiveness of the pre-trained transformer model in two ways. Initially, we maintain the original structure of the self-attention and feedforward layers in the residual blocks of the pre-trained language or image model, using the Frozen Pre-trained Transformer (FPT) for time series analysis with the addition of projection matrices for input and output. Additionally, we introduce four unique adapters, designed specifically for downstream tasks based on the pre-trained model, including forecasting and anomaly detection. These adapters are further enhanced with efficient parameter tuning, resulting in superior performance compared to all state-of-the-art methods.Our comprehensive experimental studies reveal that (a) the simple FPT achieves top-tier performance across various time series analysis tasks; and (b) fine-tuning the FPT with the custom-designed adapters can further elevate its performance, outshining specialized task-specific models."
                },
                {
                    "title": "Self-supervised Pre-training for Precipitation Post-processor",
                    "abstract": "Obtaining a sufficient forecast lead time for local precipitation is essential in preventing hazardous weather events. Global warming-induced climate change increases the challenge of accurately predicting severe precipitation events, such as heavy rainfall. In this paper, we propose a deep learning-based precipitation post-processor for numerical weather prediction (NWP) models. The precipitation post-processor consists of (i) employing self-supervised pre-training, where the parameters of the encoder are pre-trained on the reconstruction of the masked variables of the atmospheric physics domain; and (ii) conducting transfer learning on precipitation segmentation tasks (the target domain) from the pre-trained encoder. In addition, we introduced a heuristic labeling approach to effectively train class-imbalanced datasets. Our experiments on precipitation correction for regional NWP show that the proposed method outperforms other approaches."
                },
                {
                    "title": "TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting",
                    "abstract": "The past decade has witnessed significant advances in time series modeling with deep learning. While achieving state-of-the-art results, the best-performing architectures vary highly across applications and domains. Meanwhile, for natural language processing, the Generative Pre-trained Transformer (GPT) has demonstrated impressive performance via training one general-purpose model across various textual datasets. It is intriguing to explore whether GPT-type architectures can be effective for time series, capturing the intrinsic dynamic attributes and leading to significant accuracy improvements. In this paper, we propose a novel framework, TEMPO, that can effectively learn time series representations. We focus on utilizing two essential inductive biases of the time series task for pre-trained models: (i) decomposition of the complex interaction between trend, seasonal and residual components; and (ii) introducing the design of prompts to facilitate distribution adaptation in different types of time series. TEMPO expands the capability for dynamically modeling real-world temporal phenomena from data within diverse domains. Our experiments demonstrate the superior performance of TEMPO over state-of-the-art methods on zero shot setting for a number of time series benchmark datasets. This performance gain is observed not only in scenarios involving previously unseen datasets but also in scenarios with multi-modal inputs. This compelling finding highlights TEMPO's potential to constitute a foundational model-building framework."
                },
                {
                    "title": "Time-LLM: Time Series Forecasting by Reprogramming Large Language Models",
                    "abstract": "Time series forecasting holds significant importance in many real-world dynamic systems and has been extensively studied. Unlike natural language process (NLP) and computer vision (CV), where a single large model can tackle multiple tasks, models for time series forecasting are often specialized, necessitating distinct designs for different tasks and applications. While pre-trained foundation models have made impressive strides in NLP and CV, their development in time series domains has been constrained by data sparsity. Recent studies have revealed that large language models (LLMs) possess robust pattern recognition and reasoning abilities over complex sequences of tokens. However, the challenge remains in effectively aligning the modalities of time series data and natural language to leverage these capabilities. In this work, we present Time-LLM, a reprogramming framework to repurpose LLMs for general time series forecasting with the backbone language models kept intact. We begin by reprogramming the input time series with text prototypes before feeding it into the frozen LLM to align the two modalities. To augment the LLM's ability to reason with time series data, we propose Prompt-as-Prefix (PaP), which enriches the input context and directs the transformation of reprogrammed input patches. The transformed time series patches from the LLM are finally projected to obtain the forecasts. Our comprehensive evaluations demonstrate that Time-LLM is a powerful time series learner that outperforms state-of-the-art, specialized forecasting models. Moreover, Time-LLM excels in both few-shot and zero-shot learning scenarios."
                },
                {
                    "title": "Large Language Models are Few-Shot Health Learners",
                    "abstract": "Large language models (LLMs) can capture rich representations of concepts that are useful for real-world tasks. However, language alone is limited. While existing LLMs excel at text-based inferences, health applications require that models be grounded in numerical data (e.g., vital signs, laboratory values in clinical domains; steps, movement in the wellness domain) that is not easily or readily expressed as text in existing training corpus. We demonstrate that with only few-shot tuning, a large language model is capable of grounding various physiological and behavioral time-series data and making meaningful inferences on numerous health tasks for both clinical and wellness contexts. Using data from wearable and medical sensor recordings, we evaluate these capabilities on the tasks of cardiac signal analysis, physical activity recognition, metabolic calculation (e.g., calories burned), and estimation of stress reports and mental health screeners."
                },
                {
                    "title": "TempEE: Temporal\u2013Spatial Parallel Transformer for Radar Echo Extrapolation Beyond Autoregression",
                    "abstract": "Meteorological radar reflectivity data (i.e., radar echo) significantly influences precipitation prediction. It can facilitate accurate and expeditious forecasting of short-term heavy rainfall bypassing the need for complex numerical weather prediction (NWP) models. In comparison to conventional models, deep-learning (DL)-based radar echo extrapolation algorithms exhibit higher effectiveness and efficiency. Nevertheless, the development of a reliable and generalized echo extrapolation algorithm is impeded by three primary challenges: cumulative error spreading, imprecise representation of sparsely distributed echoes, and inaccurate description of nonstationary motion processes. To tackle these challenges, this article proposes a novel radar echo extrapolation algorithm called temporal\u2013spatial parallel transformer, referred to as TempEE. TempEE avoids using autoregression and instead employs a one-step forward strategy to prevent the cumulative error from spreading during the extrapolation process. Additionally, we propose the incorporation of a multilevel temporal\u2013spatial attention mechanism to improve the algorithm\u2019s capability of capturing both global and local information while emphasizing task-related regions, including sparse echo representations, in an efficient manner. Furthermore, the algorithm extracts spatio-temporal representations from continuous echo images using a parallel encoder to model the nonstationary motion process for echo extrapolation. The superiority of our TempEE has been demonstrated in the context of the classic radar echo extrapolation task, utilizing a real-world dataset. Extensive experiments have further validated the efficacy and indispensability of various components within TempEE."
                },
                {
                    "title": "W-MAE: Pre-trained weather model with masked autoencoder for multi-variable weather forecasting",
                    "abstract": "Weather forecasting is a long-standing computational challenge with direct societal and economic impacts. This task involves a large amount of continuous data collection and exhibits rich spatiotemporal dependencies over long periods, making it highly suitable for deep learning models. In this paper, we apply pre-training techniques to weather forecasting and propose W-MAE, a Weather model with Masked AutoEncoder pre-training for weather forecasting. W-MAE is pre-trained in a self-supervised manner to reconstruct spatial correlations within meteorological variables. On the temporal scale, we fine-tune the pre-trained W-MAE to predict the future states of meteorological variables, thereby modeling the temporal dependencies present in weather data. We conduct our experiments using the fifth-generation ECMWF Reanalysis (ERA5) data, with samples selected every six hours. Experimental results show that our W-MAE framework offers three key benefits: 1) when predicting the future state of meteorological variables, the utilization of our pre-trained W-MAE can effectively alleviate the problem of cumulative errors in prediction, maintaining stable performance in the short-to-medium term; 2) when predicting diagnostic variables (e.g., total precipitation), our model exhibits significant performance advantages over FourCastNet; 3) Our task-agnostic pre-training schema can be easily integrated with various task-specific models. When our pre-training framework is applied to FourCastNet, it yields an average 20% performance improvement in Anomaly Correlation Coefficient (ACC)."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "LightTS: Lightweight Time Series Classification with Adaptive Ensemble Distillation",
                    "abstract": "Due to the sweeping digitalization of processes, increasingly vast amounts of time series data are being produced. Accurate classification of such time series facilitates decision making in multiple domains. State-of-the-art classification accuracy is often achieved by ensemble learning where results are synthesized from multiple base models. This characteristic implies that ensemble learning needs substantial computing resources, preventing their use in resource-limited environments, such as in edge devices. To extend the applicability of ensemble learning, we propose the LightTS framework that compresses large ensembles into lightweight models while ensuring competitive accuracy. First, we propose adaptive ensemble distillation that assigns adaptive weights to different base models such that their varying classification capabilities contribute purposefully to the training of the lightweight model. Second, we propose means of identifying Pareto optimal settings w.r.t. model accuracy and model size, thus enabling users with a space budget to select the most accurate lightweight model. We report on experiments using 128 real-world time series sets and different types of base models that justify key decisions in the design of LightTS and provide evidence that LightTS is able to outperform competitors."
                },
                {
                    "title": "ClimaX: A foundation model for weather and climate",
                    "abstract": "Most state-of-the-art approaches for weather and climate modeling are based on physics-informed numerical models of the atmosphere. These approaches aim to model the non-linear dynamics and complex interactions between multiple variables, which are challenging to approximate. Additionally, many such numerical models are computationally intensive, especially when modeling the atmospheric phenomenon at a fine-grained spatial and temporal resolution. Recent data-driven approaches based on machine learning instead aim to directly solve a downstream forecasting or projection task by learning a data-driven functional mapping using deep neural networks. However, these networks are trained using curated and homogeneous climate datasets for specific spatiotemporal tasks, and thus lack the generality of numerical models. We develop and demonstrate ClimaX, a flexible and generalizable deep learning model for weather and climate science that can be trained using heterogeneous datasets spanning different variables, spatio-temporal coverage, and physical groundings. ClimaX extends the Transformer architecture with novel encoding and aggregation blocks that allow effective use of available compute while maintaining general utility. ClimaX is pre-trained with a self-supervised learning objective on climate datasets derived from CMIP6. The pre-trained ClimaX can then be fine-tuned to address a breadth of climate and weather tasks, including those that involve atmospheric variables and spatio-temporal scales unseen during pretraining. Compared to existing data-driven baselines, we show that this generality in ClimaX results in superior performance on benchmarks for weather forecasting and climate projections, even when pretrained at lower resolutions and compute budgets. The source code is available at https://github.com/microsoft/ClimaX."
                },
                {
                    "title": "Prompt Federated Learning for Weather Forecasting: Toward Foundation Models on Meteorological Data",
                    "abstract": "To tackle the global climate challenge, it urgently needs to develop a collaborative platform for comprehensive weather forecasting on large-scale meteorological data. Despite urgency, heterogeneous meteorological sensors across countries and regions, inevitably causing multivariate heterogeneity and data exposure, become the main barrier. This paper develops a foundation model across regions capable of understanding complex meteorological data and providing weather forecasting. To relieve the data exposure concern across regions, a novel federated learning approach has been proposed to collaboratively learn a brand-new spatio-temporal Transformer-based foundation model across participants with heterogeneous meteorological data. Moreover, a novel prompt learning mechanism has been adopted to satisfy low-resourced sensors' communication and computational constraints. The effectiveness of the proposed method has been demonstrated on classical weather forecasting tasks using three meteorological datasets with multivariate time series."
                },
                {
                    "title": "GraphCast: Learning skillful medium-range global weather forecasting",
                    "abstract": "Global medium-range weather forecasting is critical to decision-making across many social and economic domains. Traditional numerical weather prediction uses increased compute resources to improve forecast accuracy, but cannot directly use historical weather data to improve the underlying model. We introduce a machine learning-based method called\"GraphCast\", which can be trained directly from reanalysis data. It predicts hundreds of weather variables, over 10 days at 0.25 degree resolution globally, in under one minute. We show that GraphCast significantly outperforms the most accurate operational deterministic systems on 90% of 1380 verification targets, and its forecasts support better severe event prediction, including tropical cyclones, atmospheric rivers, and extreme temperatures. GraphCast is a key advance in accurate and efficient weather forecasting, and helps realize the promise of machine learning for modeling complex dynamical systems."
                },
                {
                    "title": "FedPETuning: When Federated Learning Meets the Parameter-Efficient Tuning Methods of Pre-trained Language Models",
                    "abstract": "With increasing privacy concerns on data, recent studies have made significant progress using federated learning (FL) on privacy-sensitive natural language processing (NLP) tasks. Much literature suggests fully fine-tuning pre-trained language models (PLMs) in the FL paradigm can mitigate the data heterogeneity problem and close the performance gap with centralized training. However, large PLMs bring the curse of prohibitive communication overhead and local model adaptation costs for the FL system. To this end, we introduce various parameter-efficient tuning (PETuning) methods into federated learning. Specifically, we provide a holistic empirical study of representative PLMs tuning methods in FL. The experimental results cover the analysis of data heterogeneity levels, data scales, and different FL scenarios. Overall communication overhead can be significantly reduced by locally tuning and globally aggregating lightweight model parameters while maintaining acceptable performance in various FL settings. To facilitate the research of PETuning in FL, we also develop a federated tuning framework FedPETuning, which allows practitioners to exploit different PETuning methods under the FL training paradigm conveniently. The source code is available at \\url{https://github.com/iezhuozhuo/FedETuning/tree/deltaTuning}."
                },
                {
                    "title": "FedALA: Adaptive Local Aggregation for Personalized Federated Learning",
                    "abstract": "A key challenge in federated learning (FL) is the statistical heterogeneity that impairs the generalization of the global model on each client. To address this, we propose a method Federated learning with Adaptive Local Aggregation (FedALA) by capturing the desired information in the global model for client models in personalized FL. The key component of FedALA is an Adaptive Local Aggregation (ALA) module, which can adaptively aggregate the downloaded global model and local model towards the local objective on each client to initialize the local model before training in each iteration. To evaluate the effectiveness of FedALA, we conduct extensive experiments with five benchmark datasets in computer vision and natural language processing domains. FedALA outperforms eleven state-of-the-art baselines by up to 3.27% in test accuracy. Furthermore, we also apply ALA module to other federated learning methods and achieve up to 24.19% improvement in test accuracy. Code is available at https://github.com/TsingZ0/FedALA."
                },
                {
                    "title": "A Time Series is Worth 64 Words: Long-term Forecasting with Transformers",
                    "abstract": "We propose an efficient design of Transformer-based models for multivariate time series forecasting and self-supervised representation learning. It is based on two key components: (i) segmentation of time series into subseries-level patches which are served as input tokens to Transformer; (ii) channel-independence where each channel contains a single univariate time series that shares the same embedding and Transformer weights across all the series. Patching design naturally has three-fold benefit: local semantic information is retained in the embedding; computation and memory usage of the attention maps are quadratically reduced given the same look-back window; and the model can attend longer history. Our channel-independent patch time series Transformer (PatchTST) can improve the long-term forecasting accuracy significantly when compared with that of SOTA Transformer-based models. We also apply our model to self-supervised pre-training tasks and attain excellent fine-tuning performance, which outperforms supervised training on large datasets. Transferring of masked pre-trained representation on one dataset to others also produces SOTA forecasting accuracy. Code is available at: https://github.com/yuqinie98/PatchTST."
                },
                {
                    "title": "PromptCast: A New Prompt-Based Learning Paradigm for Time Series Forecasting",
                    "abstract": "This paper presents a new perspective on time series forecasting. In existing time series forecasting methods, the models take a sequence of numerical values as input and yield numerical values as output. The existing SOTA models are largely based on the Transformer architecture, modified with multiple encoding mechanisms to incorporate the context and semantics around the historical data. Inspired by the successes of pre-trained language foundation models, we pose a question about whether these models can also be adapted to solve time-series forecasting. Thus, we propose a new forecasting paradigm: prompt-based time series forecasting (PromptCast). In this novel task, the numerical input and output are transformed into prompts and the forecasting task is framed in a sentence-to-sentence manner, making it possible to directly apply language models for forecasting purposes. To support and facilitate the research of this task, we also present a large-scale dataset (PISA) that includes three real-world forecasting scenarios. We evaluate different SOTA numerical-based forecasting methods and language generation models. The benchmark results with various forecasting settings demonstrate the proposed PromptCast with language generation models is a promising research direction. Additionally, in comparison to conventional numerical-based forecasting, PromptCast shows a much better generalization ability under the zero-shot setting."
                },
                {
                    "title": "PromptFL: Let Federated Participants Cooperatively Learn Prompts Instead of Models \u2013 Federated Learning in Age of Foundation Model",
                    "abstract": "Quick global aggregation of effective distributed parameters is crucial to federated learning (FL), which requires adequate bandwidth for parameters communication and sufficient user data for local training. Otherwise, FL may cost excessive training time for convergence and produce inaccurate models. In this paper, we propose a brand-new FL framework, PromptFL, that replaces the federated model training with the federated prompt training, i.e., let federated participants train prompts instead of a shared model, to simultaneously achieve the efficient global aggregation and local training on insufficient data by exploiting the power of foundation models (FM) in a distributed way. PromptFL ships an off-the-shelf FM, i.e., CLIP, to distributed clients who would cooperatively train shared soft prompts based on very few local data. Since PromptFL only needs to update the prompts instead of the whole model, both the local training and the global aggregation can be significantly accelerated. And FM trained over large scale data can provide strong adaptation capability to distributed users tasks with the trained soft prompts. We empirically analyze the PromptFL via extensive experiments, and show its superiority in terms of system feasibility, user privacy, and performance."
                },
                {
                    "title": "Less Is More: Fast Multivariate Time Series Forecasting with Light Sampling-oriented MLP Structures",
                    "abstract": "Multivariate time series forecasting has seen widely ranging applications in various domains, including finance, traffic, energy, and healthcare. To capture the sophisticated temporal patterns, plenty of research studies designed complex neural network architectures based on many variants of RNNs, GNNs, and Transformers. However, complex models are often computationally expensive and thus face a severe challenge in training and inference efficiency when applied to large-scale real-world datasets. In this paper, we introduce LightTS, a light deep learning architecture merely based on simple MLP-based structures. The key idea of LightTS is to apply an MLP-based structure on top of two delicate down-sampling strategies, including interval sampling and continuous sampling, inspired by a crucial fact that down-sampling time series often preserves the majority of its information. We conduct extensive experiments on eight widely used benchmark datasets. Compared with the existing state-of-the-art methods, LightTS demonstrates better performance on five of them and comparable performance on the rest. Moreover, LightTS is highly efficient. It uses less than 5% FLOPS compared with previous SOTA methods on the largest benchmark dataset. In addition, LightTS is robust and has a much smaller variance in forecasting accuracy than previous SOTA methods in long sequence forecasting tasks."
                },
                {
                    "title": "Are Transformers Effective for Time Series Forecasting?",
                    "abstract": "Recently, there has been a surge of Transformer-based solutions for the long-term time series forecasting (LTSF) task. Despite the growing performance over the past few years, we question the validity of this line of research in this work. Specifically, Transformers is arguably the most successful solution to extract the semantic correlations among the elements in a long sequence. However, in time series modeling, we are to extract the temporal relations in an ordered set of continuous points. While employing positional encoding and using tokens to embed sub-series in Transformers facilitate preserving some ordering information, the nature of the permutation-invariant self-attention mechanism inevitably results in temporal information loss. \nTo validate our claim, we introduce a set of embarrassingly simple one-layer linear models named LTSF-Linear for comparison. Experimental results on nine real-life datasets show that LTSF-Linear surprisingly outperforms existing sophisticated Transformer-based LTSF models in all cases, and often by a large margin. Moreover, we conduct comprehensive empirical studies to explore the impacts of various design elements of LTSF models on their temporal relation extraction capability. We hope this surprising finding opens up new research directions for the LTSF task. We also advocate revisiting the validity of Transformer-based solutions for other time series analysis tasks (e.g., anomaly detection) in the future."
                },
                {
                    "title": "Transformers in Time Series: A Survey",
                    "abstract": "Transformers have achieved superior performances in many tasks in natural language processing and computer vision, which also triggered great interest in the time series community. Among multiple advantages of Transformers, the ability to capture long-range dependencies and interactions is especially attractive for time series modeling, leading to exciting progress in various time series applications. In this paper, we systematically review Transformer schemes for time series modeling by highlighting their strengths as well as limitations. In particular, we examine the development of time series Transformers in two perspectives. From the perspective of network structure, we summarize the adaptations and modifications that have been made to Transformers in order to accommodate the challenges in time series analysis. From the perspective of applications, we categorize time series Transformers based on common tasks including forecasting, anomaly detection, and classification. Empirically, we perform robust analysis, model size analysis, and seasonal-trend decomposition analysis to study how Transformers perform in time series. Finally, we discuss and suggest future directions to provide useful research guidance."
                },
                {
                    "title": "Forecasting Global Weather with Graph Neural Networks",
                    "abstract": "We present a data-driven approach for forecasting global weather using graph neural networks. The system learns to step forward the current 3D atmospheric state by six hours, and multiple steps are chained together to produce skillful forecasts going out several days into the future. The underlying model is trained on reanalysis data from ERA5 or forecast data from GFS. Test performance on metrics such as Z500 (geopotential height) and T850 (temperature) improves upon previous data-driven approaches and is comparable to operational, full-resolution, physical models from GFS and ECMWF, at least when evaluated on 1-degree scales and when using reanalysis initial conditions. We also show results from connecting this data-driven model to live, operational forecasts from GFS."
                },
                {
                    "title": "ETSformer: Exponential Smoothing Transformers for Time-series Forecasting",
                    "abstract": "Transformers have been actively studied for time-series forecasting in recent years. While often showing promising results in various scenarios, traditional Transformers are not designed to fully exploit the characteristics of time-series data and thus suffer some fundamental limitations, e.g., they generally lack of decomposition capability and interpretability, and are neither effective nor efficient for long-term forecasting. In this paper, we propose ETSFormer, a novel time-series Transformer architecture, which exploits the principle of exponential smoothing in improving Transformers for time-series forecasting. In particular, inspired by the classical exponential smoothing methods in time-series forecasting, we propose the novel exponential smoothing attention (ESA) and frequency attention (FA) to replace the self-attention mechanism in vanilla Transformers, thus improving both accuracy and efficiency. Based on these, we redesign the Transformer architecture with modular decomposition blocks such that it can learn to decompose the time-series data into interpretable time-series components such as level, growth and seasonality. Extensive experiments on various time-series benchmarks validate the efficacy and advantages of the proposed method. Code is available at https://github.com/salesforce/ETSformer."
                },
                {
                    "title": "FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting",
                    "abstract": "Although Transformer-based methods have significantly improved state-of-the-art results for long-term series forecasting, they are not only computationally expensive but more importantly, are unable to capture the global view of time series (e.g. overall trend). To address these problems, we propose to combine Transformer with the seasonal-trend decomposition method, in which the decomposition method captures the global profile of time series while Transformers capture more detailed structures. To further enhance the performance of Transformer for long-term prediction, we exploit the fact that most time series tend to have a sparse representation in well-known basis such as Fourier transform, and develop a frequency enhanced Transformer. Besides being more effective, the proposed method, termed as Frequency Enhanced Decomposed Transformer ({\\bf FEDformer}), is more efficient than standard Transformer with a linear complexity to the sequence length. Our empirical studies with six benchmark datasets show that compared with state-of-the-art methods, FEDformer can reduce prediction error by $14.8\\%$ and $22.6\\%$ for multivariate and univariate time series, respectively. Code is publicly available at https://github.com/MAZiqing/FEDformer."
                },
                {
                    "title": "Adapt to Adaptation: Learning Personalization for Cross-Silo Federated Learning",
                    "abstract": "Conventional federated learning (FL) trains one global model for a federation of clients with decentralized data, reducing the privacy risk of centralized training. However, the distribution shift across non-IID datasets, often poses a challenge to this one-model-fits-all solution. Personalized FL aims to mitigate this issue systematically. In this work, we propose APPLE, a personalized cross-silo FL framework that adaptively learns how much each client can benefit from other clients' models. We also introduce a method to flexibly control the focus of training APPLE between global and local objectives. We empirically evaluate our method's convergence and generalization behaviors, and perform extensive experiments on two benchmark datasets and two medical imaging datasets under two non-IID settings. The results show that the proposed personalized FL framework, APPLE, achieves state-of-the-art performance compared to several other personalized FL approaches in the literature. The code is publicly available at https://github.com/ljaiverson/pFL-APPLE."
                },
                {
                    "title": "Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting",
                    "abstract": "Extending the forecasting time is a critical demand for real applications, such as extreme weather early warning and long-term energy consumption planning. This paper studies the long-term forecasting problem of time series. Prior Transformer-based models adopt various self-attention mechanisms to discover the long-range dependencies. However, intricate temporal patterns of the long-term future prohibit the model from finding reliable dependencies. Also, Transformers have to adopt the sparse versions of point-wise self-attentions for long series efficiency, resulting in the information utilization bottleneck. Going beyond Transformers, we design Autoformer as a novel decomposition architecture with an Auto-Correlation mechanism. We break with the pre-processing convention of series decomposition and renovate it as a basic inner block of deep models. This design empowers Autoformer with progressive decomposition capacities for complex time series. Further, inspired by the stochastic process theory, we design the Auto-Correlation mechanism based on the series periodicity, which conducts the dependencies discovery and representation aggregation at the sub-series level. Auto-Correlation outperforms self-attention in both efficiency and accuracy. In long-term forecasting, Autoformer yields state-of-the-art accuracy, with a 38% relative improvement on six benchmarks, covering five practical applications: energy, traffic, economics, weather and disease. Code is available at this repository: \\url{https://github.com/thuml/Autoformer}."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "FedProto: Federated Prototype Learning across Heterogeneous Clients",
                    "abstract": "Heterogeneity across clients in federated learning (FL) usually hinders the optimization convergence and generalization performance when the aggregation of clients' knowledge occurs in the gradient space. For example, clients may differ in terms of data distribution, network latency, input/output space, and/or model architecture, which can easily lead to the misalignment of their local gradients. To improve the tolerance to heterogeneity, we propose a novel federated prototype learning (FedProto) framework in which the clients and server communicate the abstract class prototypes instead of the gradients. FedProto aggregates the local prototypes collected from different clients, and then sends the global prototypes back to all clients to regularize the training of local models. The training on each client aims to minimize the classification error on the local data while keeping the resulting local prototypes sufficiently close to the corresponding global ones. Moreover, we provide a theoretical analysis to the convergence rate of FedProto under non-convex objectives. In experiments, we propose a benchmark setting tailored for heterogeneous FL, with FedProto outperforming several recent FL approaches on multiple datasets."
                },
                {
                    "title": "Towards Personalized Federated Learning",
                    "abstract": "In parallel with the rapid adoption of artificial intelligence (AI) empowered by advances in AI research, there has been growing awareness and concerns of data privacy. Recent significant developments in the data regulation landscape have prompted a seismic shift in interest toward privacy-preserving AI. This has contributed to the popularity of Federated Learning (FL), the leading paradigm for the training of machine learning models on data silos in a privacy-preserving manner. In this survey, we explore the domain of personalized FL (PFL) to address the fundamental challenges of FL on heterogeneous data, a universal characteristic inherent in all real-world datasets. We analyze the key motivations for PFL and present a unique taxonomy of PFL techniques categorized according to the key challenges and personalization strategies in PFL. We highlight their key ideas, challenges, opportunities, and envision promising future trajectories of research toward a new PFL architectural design, realistic PFL benchmarking, and trustworthy PFL approaches."
                },
                {
                    "title": "FedBN: Federated Learning on Non-IID Features via Local Batch Normalization",
                    "abstract": "The emerging paradigm of federated learning (FL) strives to enable collaborative training of deep models on the network edge without centrally aggregating raw data and hence improving data privacy. In most cases, the assumption of independent and identically distributed samples across local clients does not hold for federated learning setups. Under this setting, neural network training performance may vary significantly according to the data distribution and even hurt training convergence. Most of the previous work has focused on a difference in the distribution of labels or client shifts. Unlike those settings, we address an important problem of FL, e.g., different scanners/sensors in medical imaging, different scenery distribution in autonomous driving (highway vs. city), where local clients store examples with different distributions compared to other clients, which we denote as feature shift non-iid. In this work, we propose an effective method that uses local batch normalization to alleviate the feature shift before averaging models. The resulting scheme, called FedBN, outperforms both classical FedAvg, as well as the state-of-the-art for non-iid data (FedProx) on our extensive experiments. These empirical results are supported by a convergence analysis that shows in a simplified setting that FedBN has a faster convergence rate than FedAvg. Code is available at https://github.com/med-air/FedBN."
                },
                {
                    "title": "Exploiting Shared Representations for Personalized Federated Learning",
                    "abstract": "Deep neural networks have shown the ability to extract universal feature representations from data such as images and text that have been useful for a variety of learning tasks. However, the fruits of representation learning have yet to be fully-realized in federated settings. Although data in federated settings is often non-i.i.d. across clients, the success of centralized deep learning suggests that data often shares a global feature representation, while the statistical heterogeneity across clients or tasks is concentrated in the labels. Based on this intuition, we propose a novel federated learning framework and algorithm for learning a shared data representation across clients and unique local heads for each client. Our algorithm harnesses the distributed computational power across clients to perform many local-updates with respect to the low-dimensional local parameters for every update of the representation. We prove that this method obtains linear convergence to the ground-truth representation with near-optimal sample complexity in a linear setting, demonstrating that it can efficiently reduce the problem dimension for each client. This result is of interest beyond federated learning to a broad class of problems in which we aim to learn a shared low-dimensional representation among data distributions, for example in meta-learning and multi-task learning. Further, extensive experimental results show the empirical improvement of our method over alternative personalized federated learning approaches in federated environments with heterogeneous data."
                },
                {
                    "title": "Personalized Federated Learning with First Order Model Optimization",
                    "abstract": "While federated learning traditionally aims to train a single global model across decentralized local datasets, one model may not always be ideal for all participating clients. Here we propose an alternative, where each client only federates with other relevant clients to obtain a stronger model per client-specific objectives. To achieve this personalization, rather than computing a single model average with constant weights for the entire federation as in traditional FL, we efficiently calculate optimal weighted model combinations for each client, based on figuring out how much a client can benefit from another's model. We do not assume knowledge of any underlying data distributions or client similarities, and allow each client to optimize for arbitrary target distributions of interest, enabling greater flexibility for personalization. We evaluate and characterize our method on a variety of federated settings, datasets, and degrees of local data heterogeneity. Our method outperforms existing alternatives, while also enabling new features for personalized FL such as transfer outside of local data distributions."
                },
                {
                    "title": "Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting",
                    "abstract": "Many real-world applications require the prediction of long sequence time-series, such as electricity consumption planning. Long sequence time-series forecasting (LSTF) demands a high prediction capacity of the model, which is the ability to capture precise long-range dependency coupling between output and input efficiently. Recent studies have shown the potential of Transformer to increase the prediction capacity. However, there are several severe issues with Transformer that prevent it from being directly applicable to LSTF, including quadratic time complexity, high memory usage, and inherent limitation of the encoder-decoder architecture. To address these issues, we design an efficient transformer-based model for LSTF, named Informer, with three distinctive characteristics: (i) a ProbSparse self-attention mechanism, which achieves O(L log L) in time complexity and memory usage, and has comparable performance on sequences' dependency alignment. (ii) the self-attention distilling highlights dominating attention by halving cascading layer input, and efficiently handles extreme long input sequences. (iii) the generative style decoder, while conceptually simple, predicts the long time-series sequences at one forward operation rather than a step-by-step way, which drastically improves the inference speed of long-sequence predictions. Extensive experiments on four large-scale datasets demonstrate that Informer significantly outperforms existing methods and provides a new solution to the LSTF problem."
                },
                {
                    "title": "Ditto: Fair and Robust Federated Learning Through Personalization",
                    "abstract": "Fairness and robustness are two important concerns for federated learning systems. In this work, we identify that robustness to data and model poisoning attacks and fairness, measured as the uniformity of performance across devices, are competing constraints in statistically heterogeneous networks. To address these constraints, we propose employing a simple, general framework for personalized federated learning, Ditto, that can inherently provide fairness and robustness benefits, and develop a scalable solver for it. Theoretically, we analyze the ability of Ditto to achieve fairness and robustness simultaneously on a class of linear problems. Empirically, across a suite of federated datasets, we show that Ditto not only achieves competitive performance relative to recent personalization methods, but also enables more accurate, robust, and fair models relative to state-of-the-art fair or robust baselines."
                },
                {
                    "title": "Lower Bounds and Optimal Algorithms for Personalized Federated Learning",
                    "abstract": "In this work, we consider the optimization formulation of personalized federated learning recently introduced by Hanzely and Richtarik (2020) which was shown to give an alternative explanation to the workings of local {\\tt SGD} methods. Our first contribution is establishing the first lower bounds for this formulation, for both the communication complexity and the local oracle complexity. Our second contribution is the design of several optimal methods matching these lower bounds in almost all regimes. These are the first provably optimal methods for personalized federated learning. Our optimal methods include an accelerated variant of {\\tt FedProx}, and an accelerated variance-reduced version of {\\tt FedAvg}/Local {\\tt SGD}. We demonstrate the practical superiority of our methods through extensive numerical experiments."
                },
                {
                    "title": "The ERA5 global reanalysis",
                    "abstract": "Within the Copernicus Climate Change Service (C3S), ECMWF is producing the ERA5 reanalysis which, once completed, will embody a detailed record of the global atmosphere, land surface and ocean waves from 1950 onwards. This new reanalysis replaces the ERA\u2010Interim reanalysis (spanning 1979 onwards) which was started in 2006. ERA5 is based on the Integrated Forecasting System (IFS) Cy41r2 which was operational in 2016. ERA5 thus benefits from a decade of developments in model physics, core dynamics and data assimilation. In addition to a significantly enhanced horizontal resolution of 31\u2009km, compared to 80\u2009km for ERA\u2010Interim, ERA5 has hourly output throughout, and an uncertainty estimate from an ensemble (3\u2010hourly at half the horizontal resolution). This paper describes the general set\u2010up of ERA5, as well as a basic evaluation of characteristics and performance, with a focus on the dataset from 1979 onwards which is currently publicly available. Re\u2010forecasts from ERA5 analyses show a gain of up to one day in skill with respect to ERA\u2010Interim. Comparison with radiosonde and PILOT data prior to assimilation shows an improved fit for temperature, wind and humidity in the troposphere, but not the stratosphere. A comparison with independent buoy data shows a much improved fit for ocean wave height. The uncertainty estimate reflects the evolution of the observing systems used in ERA5. The enhanced temporal and spatial resolution allows for a detailed evolution of weather systems. For precipitation, global\u2010mean correlation with monthly\u2010mean GPCP data is increased from 67% to 77%. In general, low\u2010frequency variability is found to be well represented and from 10\u2009hPa downwards general patterns of anomalies in temperature match those from the ERA\u2010Interim, MERRA\u20102 and JRA\u201055 reanalyses."
                },
                {
                    "title": "WeatherBench: A Benchmark Data Set for Data\u2010Driven Weather Forecasting",
                    "abstract": "Data\u2010driven approaches, most prominently deep learning, have become powerful tools for prediction in many domains. A natural question to ask is whether data\u2010driven methods could also be used to predict global weather patterns days in advance. First studies show promise but the lack of a common data set and evaluation metrics make intercomparison between studies difficult. Here we present a benchmark data set for data\u2010driven medium\u2010range weather forecasting (specifically 3\u20135\u2009days), a topic of high scientific interest for atmospheric and computer scientists alike. We provide data derived from the ERA5 archive that has been processed to facilitate the use in machine learning models. We propose simple and clear evaluation metrics which will enable a direct comparison between different methods. Further, we provide baseline scores from simple linear regression techniques, deep learning models, as well as purely physical forecasting models. The data set is publicly available at https://github.com/pangeo\u2010data/WeatherBench and the companion code is reproducible with tutorials for getting started. We hope that this data set will accelerate research in data\u2010driven weather forecasting."
                },
                {
                    "title": "Reformer: The Efficient Transformer",
                    "abstract": "Large Transformer models routinely achieve state-of-the-art results on a number of tasks but training these models can be prohibitively costly, especially on long sequences. We introduce two techniques to improve the efficiency of Transformers. For one, we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O($L^2$) to O($L\\log L$), where $L$ is the length of the sequence. Furthermore, we use reversible residual layers instead of the standard residuals, which allows storing activations only once in the training process instead of $N$ times, where $N$ is the number of layers. The resulting model, the Reformer, performs on par with Transformer models while being much more memory-efficient and much faster on long sequences."
                },
                {
                    "title": "Federated Learning with Personalization Layers",
                    "abstract": "The emerging paradigm of federated learning strives to enable collaborative training of machine learning models on the network edge without centrally aggregating raw data and hence, improving data privacy. This sharply deviates from traditional machine learning and necessitates the design of algorithms robust to various sources of heterogeneity. Specifically, statistical heterogeneity of data across user devices can severely degrade the performance of standard federated averaging for traditional machine learning applications like personalization with deep learning. This paper pro-posesFedPer, a base + personalization layer approach for federated training of deep feedforward neural networks, which can combat the ill-effects of statistical heterogeneity. We demonstrate effectiveness ofFedPerfor non-identical data partitions ofCIFARdatasetsand on a personalized image aesthetics dataset from Flickr."
                },
                {
                    "title": "FedPAQ: A Communication-Efficient Federated Learning Method with Periodic Averaging and Quantization",
                    "abstract": "Federated learning is a distributed framework according to which a model is trained over a set of devices, while keeping data localized. This framework faces several systems-oriented challenges which include (i) communication bottleneck since a large number of devices upload their local updates to a parameter server, and (ii) scalability as the federated network consists of millions of devices. Due to these systems challenges as well as issues related to statistical heterogeneity of data and privacy concerns, designing a provably efficient federated learning method is of significant importance yet it remains challenging. In this paper, we present FedPAQ, a communication-efficient Federated Learning method with Periodic Averaging and Quantization. FedPAQ relies on three key features: (1) periodic averaging where models are updated locally at devices and only periodically averaged at the server; (2) partial device participation where only a fraction of devices participate in each round of the training; and (3) quantized message-passing where the edge nodes quantize their updates before uploading to the parameter server. These features address the communications and scalability challenges in federated learning. We also show that FedPAQ achieves near-optimal theoretical guarantees for strongly convex and non-convex loss functions and empirically demonstrate the communication-computation tradeoff provided by our method."
                },
                {
                    "title": "Federated Optimization in Heterogeneous Networks",
                    "abstract": "Federated Learning is a distributed learning paradigm with two key challenges that differentiate it from traditional distributed optimization: (1) significant variability in terms of the systems characteristics on each device in the network (systems heterogeneity), and (2) non-identically distributed data across the network (statistical heterogeneity). In this work, we introduce a framework, FedProx, to tackle heterogeneity in federated networks. FedProx can be viewed as a generalization and re-parametrization of FedAvg, the current state-of-the-art method for federated learning. While this re-parameterization makes only minor modifications to the method itself, these modifications have important ramifications both in theory and in practice. Theoretically, we provide convergence guarantees for our framework when learning over data from non-identical distributions (statistical heterogeneity), and while adhering to device-level systems constraints by allowing each participating device to perform a variable amount of work (systems heterogeneity). Practically, we demonstrate that FedProx allows for more robust convergence than FedAvg across a suite of realistic federated datasets. In particular, in highly heterogeneous settings, FedProx demonstrates significantly more stable and accurate convergence behavior relative to FedAvg---improving absolute test accuracy by 22% on average."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Communication-Efficient Learning of Deep Networks from Decentralized Data",
                    "abstract": "Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data center and training there using conventional approaches. We advocate an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. \nWe present a practical method for the federated learning of deep networks based on iterative model averaging, and conduct an extensive empirical evaluation, considering five different model architectures and four datasets. These experiments demonstrate the approach is robust to the unbalanced and non-IID data distributions that are a defining characteristic of this setting. Communication costs are the principal constraint, and we show a reduction in required communication rounds by 10-100x as compared to synchronized stochastic gradient descent."
                },
                {
                    "title": "Peer Review File Manuscript Title: Accurate medium-range global weather forecasting with 3D neural networks Reviewer Comments & Author Rebuttals",
                    "abstract": "Pangu Weather presents a novel deep learning model for accurate weather forecasts, which surpasses, or at least comes close to the forecasting skills of one of the best numerical weather prediction models. It is one of three recent breakthroughs in the application of deep learning to weather forecasting and an extension of the Swin transformer model, which has shown some advantages over other approaches in the field of image analysis."
                },
                {
                    "title": "LLM4TS: Two-Stage Fine-Tuning for Time-Series Forecasting with Pre-Trained LLMs",
                    "abstract": "In this work, we leverage pre-trained Large Language Models (LLMs) to enhance time-series forecasting. Mirroring the growing interest in unifying models for Natural Language Processing and Computer Vision, we envision creating an analogous model for long-term time-series forecasting. Due to limited large-scale time-series data for building robust foundation models, our approach LLM4TS focuses on leveraging the strengths of pre-trained LLMs. By combining time-series patching with temporal encoding, we have enhanced the capability of LLMs to handle time-series data effectively. Inspired by the supervised fine-tuning in chatbot domains, we prioritize a two-stage fine-tuning process: first conducting supervised fine-tuning to orient the LLM towards time-series data, followed by task-specific downstream fine-tuning. Furthermore, to unlock the flexibility of pre-trained LLMs without extensive parameter adjustments, we adopt several Parameter-Efficient Fine-Tuning (PEFT) techniques. Drawing on these innovations, LLM4TS has yielded state-of-the-art results in long-term forecasting. Our model has also shown exceptional capabilities as both a robust representation learner and an effective few-shot learner, thanks to the knowledge transferred from the pre-trained LLM."
                },
                {
                    "title": "Pyraformer: Low-Complexity Pyramidal Attention for Long-Range Time Series Modeling and Forecasting",
                    "abstract": "Accurate prediction of the future given the past based on time series data is of paramount importance, since it opens the door for decision making and risk management ahead of time. In practice, the challenge is to build a flexible but parsimonious model that can capture a wide range of temporal dependencies. In this paper, we propose Pyraformer by exploring the multi-resolution representation of the time series. Specifically, we introduce the pyramidal attention module (PAM) in which the inter-scale tree structure summarizes features at different resolutions and the intra-scale neighboring connections model the temporal dependencies of different ranges. Under mild conditions, the maximum length of the signal traversing path in Pyraformer is a constant (i.e., O(1)) with regard to the sequence length L, while its time and space complexity scale linearly with L. Extensive experimental results show that Pyraformer typically achieves the highest prediction accuracy in both single-step and long-range multi-step forecasting tasks with the least amount of time and memory consumption, especially when the sequence is long1."
                },
                {
                    "title": "Dynamic Multiscale Fusion Generative Adversarial Network for Radar Image Extrapolation",
                    "abstract": "Typhoons, a kind of devastating natural disaster, have caused incalculable damages worldwide. The meteorological radar image is essential for weather forecasting, especially typhoons. The weather nowcasting (future 0\u20136 h) can be implemented via extrapolating radar images without using the primary weather forecasting method\u2014the numerical weather prediction model. However, the existing related techniques based on statistics or artificial intelligence were not efficient enough. In this article, a novel radar image extrapolation algorithm named dynamic multiscale fusion-generative adversarial network (DMSF-GAN) was proposed. DMSF-GAN captures the future radar image distribution based on current radar images through modifying the GAN. In the generative module of GAN, an auto-encoder consisting of dynamic inception-3-D and feature connection blocks extracts significant features from current radar images. The feasibility of the proposed model was verified on a real radar image dataset, and the experimental results proved that the proposed algorithm could effectively capture the location and pattern of the future radar echo, especially for typhoon weather systems. Compared with the mainstream methods of radar image extrapolation such as optical-flow and recurrent neural network (RNN)-based models, DMSF-GAN has a more superior and robust performance, which is also suitable for running on low-configuration machines."
                },
                {
                    "title": "Reversible Instance Normalization for Accurate Time-Series Forecasting against Distribution Shift",
                    "abstract": "Statistical properties such as mean and variance often change over time in time series, i.e., time-series data suffer from a distribution shift problem. This change in temporal distribution is one of the main challenges that prevent accurate timeseries forecasting. To address this issue, we propose a simple yet effective normalization method called reversible instance normalization (RevIN), a generallyapplicable normalization-and-denormalization method with learnable affine transformation. The proposed method is symmetrically structured to remove and restore the statistical information of a time-series instance, leading to significant performance improvements in time-series forecasting, as shown in Fig. 1. We demonstrate the effectiveness of RevIN via extensive quantitative and qualitative analyses on various real-world datasets, addressing the distribution shift problem. \u2217Both authors contributed equally. The order of the first authors was determined by coin flip."
                },
                {
                    "title": "Personalized Federated Learning with Theoretical Guarantees: A Model-Agnostic Meta-Learning Approach",
                    "abstract": "In Federated Learning, we aim to train models across multiple computing units (users), while users can only communicate with a common central server, without exchanging their data samples. This mechanism exploits the computational power of all users and allows users to obtain a richer model as their models are trained over a larger set of data points. However, this scheme only develops a common output for all the users, and, therefore, it does not adapt the model to each user. This is an important missing feature, especially given the heterogeneity of the underlying data distribution for various users. In this paper, we study a personalized variant of the federated learning in which our goal is to \ufb01nd an initial shared model that current or new users can easily adapt to their local dataset by performing one or a few steps of gradient descent with respect to their own data. This approach keeps all the bene\ufb01ts of the federated learning architecture, and, by structure, leads to a more personalized model for each user. We show this problem can be studied within the Model-Agnostic Meta-Learning (MAML) framework. Inspired by this connection, we study a personalized variant of the well-known Federated Averaging algorithm and evaluate its performance in terms of gradient norm for non-convex loss functions. Further, we characterize how this performance is affected by the closeness of underlying distributions of user data, measured in terms of distribution distances such as Total Variation and 1-Wasserstein metric."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                }
            ],
            "categories": [
                "physics.ao-ph",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively model weather variation patterns using personalized federated learning approaches on resource-constrained weather devices while addressing data heterogeneity and communication overhead?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of meteorology, as it enables more accurate and timely weather forecasts, which are essential for disaster preparedness and response. By improving personalized weather modeling through federated learning, we can enhance the ability of local devices to process and analyze meteorological data independently, leading to better insights tailored to specific regional conditions. This research could pave the way for more efficient use of data collected from diverse weather devices, ultimately contributing to the development of smarter, more resilient weather monitoring systems and applications.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from several complexities:  \n1. **Data Heterogeneity**: Weather data collected from different regions exhibit unique characteristics, leading to significant variations in data distribution. This heterogeneity complicates the training of a uniform model across devices.  \n2. **Underperformance of Shallow Networks**: Simpler neural network models struggle to generalize effectively due to the vast and varied nature of weather data, while deeper models are resource-intensive and difficult to deploy on devices that require frequent updates.  \n3. **Resource Constraints**: Weather devices have limited computational power and cannot train complex models from scratch. Additionally, the high communication overhead associated with transmitting complete models during federated learning aggregation phases is impractical for real-time applications.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either traditional physics-based methods or generalized deep learning approaches that do not account for the unique challenges posed by meteorological data. Existing solutions often overlook the specific needs of resource-constrained devices and the implications of data heterogeneity. Barriers such as the lack of compact foundational models tailored for personalized weather modeling and the absence of effective strategies to manage communication overhead have hindered progress. Our approach aims to address these gaps by developing a compact foundational model that is specifically designed for on-device meteorological variable modeling, enhancing personalization and efficiency.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the following key components:  \n1. **Method**: We will implement a personalized federated learning framework that utilizes a compact foundational model (FM) for"
            }
        },
        "author_data": {
            "f7ba2844-6817-4b4d-9a74-dce830e440f3": {
                "pk": "f7ba2844-6817-4b4d-9a74-dce830e440f3",
                "name": "Shengchao Chen",
                "collaborators": [
                    "Sufen Ren",
                    "Guodong Long",
                    "Jing Jiang",
                    "Ting Shu",
                    "Huan Zhao",
                    "Tao Shen",
                    "Guanjun Wang",
                    "Jiahao Wang",
                    "Chengqi Zhang",
                    "Yule Hu"
                ],
                "domain": [
                    "Federated Learning",
                    "Deep Learning",
                    "Weather Forecasting",
                    "Medical Imaging"
                ],
                "publications": [
                    {
                        "title": "Prompt Federated Learning for Weather Forecasting: Toward Foundation Models on Meteorological Data",
                        "abstract": "To tackle the global climate challenge, it urgently needs to develop a collaborative platform for comprehensive weather forecasting on large-scale meteorological data. Despite urgency, heterogeneous meteorological sensors across countries and regions, inevitably causing multivariate heterogeneity and data exposure, become the main barrier. This paper develops a foundation model across regions capable of understanding complex meteorological data and providing weather forecasting. To relieve the data exposure concern across regions, a novel federated learning approach has been proposed to collaboratively learn a brand-new spatio-temporal Transformer-based foundation model across participants with heterogeneous meteorological data. Moreover, a novel prompt learning mechanism has been adopted to satisfy low-resourced sensors' communication and computational constraints. The effectiveness of the proposed method has been demonstrated on classical weather forecasting tasks using three meteorological datasets with multivariate time series."
                    },
                    {
                        "title": "Federated Distillation for Medical Image Classification: Towards Trustworthy Computer-Aided Diagnosis",
                        "abstract": "Medical image classification plays a crucial role in computer-aided clinical diagnosis. While deep learning techniques have significantly enhanced efficiency and reduced costs, the privacy-sensitive nature of medical imaging data complicates centralized storage and model training. Furthermore, low-resource healthcare organizations face challenges related to communication overhead and efficiency due to increasing data and model scales. This paper proposes a novel privacy-preserving medical image classification framework based on federated learning to address these issues, named FedMIC. The framework enables healthcare organizations to learn from both global and local knowledge, enhancing local representation of private data despite statistical heterogeneity. It provides customized models for organizations with diverse data distributions while minimizing communication overhead and improving efficiency without compromising performance. Our FedMIC enhances robustness and practical applicability under resource-constrained conditions. We demonstrate FedMIC's effectiveness using four public medical image datasets for classical medical image classification tasks."
                    },
                    {
                        "title": "Less is more: Ensemble Learning for Retinal Disease Recognition Under Limited Resources",
                        "abstract": "Retinal optical coherence tomography (OCT) images provide crucial insights into the health of the posterior ocular segment. Therefore, the advancement of automated image analysis methods is imperative to equip clinicians and researchers with quantitative data, thereby facilitating informed decision-making. The application of deep learning (DL)-based approaches has gained extensive traction for executing these analysis tasks, demonstrating remarkable performance compared to labor-intensive manual analyses. However, the acquisition of Retinal OCT images often presents challenges stemming from privacy concerns and the resource-intensive labeling procedures, which contradicts the prevailing notion that DL models necessitate substantial data volumes for achieving superior performance. Moreover, limitations in available computational resources constrain the progress of high-performance medical artificial intelligence, particularly in less developed regions and countries. This paper introduces a novel ensemble learning mechanism designed for recognizing retinal diseases under limited resources (e.g., data, computation). The mechanism leverages insights from multiple pre-trained models, facilitating the transfer and adaptation of their knowledge to Retinal OCT images. This approach establishes a robust model even when confronted with limited labeled data, eliminating the need for an extensive array of parameters, as required in learning from scratch. Comprehensive experimentation on real-world datasets demonstrates that the proposed approach can achieve superior performance in recognizing Retinal OCT images, even when dealing with exceedingly restricted labeled datasets. Furthermore, this method obviates the necessity of learning extensive-scale parameters, making it well-suited for deployment in low-resource scenarios."
                    },
                    {
                        "title": "MASK-CNN-Transformer For Real-Time Multi-Label Weather Recognition",
                        "abstract": "Weather recognition is an essential support for many practical life applications, including traffic safety, environment, and meteorology. However, many existing related works cannot comprehensively describe weather conditions due to their complex co-occurrence dependencies. This paper proposes a novel multi-label weather recognition model considering these dependencies. The proposed model called MASK-Convolutional Neural Network-Transformer (MASK-CT) is based on the Transformer, the convolutional process, and the MASK mechanism. The model employs multiple convolutional layers to extract features from weather images and a Transformer encoder to calculate the probability of each weather condition based on the extracted features. To improve the generalization ability of MASK-CT, a MASK mechanism is used during the training phase. The effect of the MASK mechanism is explored and discussed. The Mask mechanism randomly withholds some information from one-pair training instances (one image and its corresponding label). There are two types of MASK methods. Specifically, MASK-I is designed and deployed on the image before feeding it into the weather feature extractor and MASK-II is applied to the image label. The Transformer encoder is then utilized on the randomly masked image features and labels. The experimental results from various real-world weather recognition datasets demonstrate that the proposed MASK-CT model outperforms state-of-the-art methods. Furthermore, the high-speed dynamic real-time weather recognition capability of the MASK-CT is evaluated."
                    },
                    {
                        "title": "Federated Prompt Learning for Weather Foundation Models on Devices",
                        "abstract": "On-device intelligence for weather forecasting uses local deep learning models to analyze weather patterns without centralized cloud computing, holds significance for supporting human activates. Federated Learning is a promising solution for such forecasting by enabling collaborative model training without sharing raw data. However, it faces three main challenges that hinder its reliability: (1) data heterogeneity among devices due to geographic differences; (2) data homogeneity within individual devices and (3) communication overload from sending large model parameters for collaboration. To address these challenges, this paper propose Federated Prompt Learning for Weather Foundation Models on Devices (FedPoD), which enables devices to obtain highly customized models while maintaining communication efficiency. Concretely, our Adaptive Prompt Tuning leverages lightweight prompts guide frozen foundation model to generate more precise predictions, also conducts prompt-based multi-level communication to encourage multi-source knowledge fusion and regulate optimization. Additionally, Dynamic Graph Modeling constructs graphs from prompts, prioritizing collaborative training among devices with similar data distributions to against heterogeneity. Extensive experiments demonstrates FedPoD leads the performance among state-of-the-art baselines across various setting in real-world on-device weather forecasting datasets."
                    },
                    {
                        "title": "TempEE: Temporal-Spatial Parallel Transformer for Radar Echo Extrapolation Beyond Auto-Regression",
                        "abstract": "Meteorological radar reflectivity data (i.e. radar echo) significantly influences precipitation prediction. It can facilitate accurate and expeditious forecasting of short-term heavy rainfall bypassing the need for complex Numerical Weather Prediction (NWP) models. In comparison to conventional models, Deep Learning (DL)-based radar echo extrapolation algorithms exhibit higher effectiveness and efficiency. Nevertheless, the development of reliable and generalized echo extrapolation algorithm is impeded by three primary challenges: cumulative error spreading, imprecise representation of sparsely distributed echoes, and inaccurate description of non-stationary motion processes. To tackle these challenges, this paper proposes a novel radar echo extrapolation algorithm called Temporal-Spatial Parallel Transformer, referred to as TempEE. TempEE avoids using auto-regression and instead employs a one-step forward strategy to prevent cumulative error spreading during the extrapolation process. Additionally, we propose the incorporation of a Multi-level Temporal-Spatial Attention mechanism to improve the algorithm's capability of capturing both global and local information while emphasizing task-related regions, including sparse echo representations, in an efficient manner. Furthermore, the algorithm extracts spatio-temporal representations from continuous echo images using a parallel encoder to model the non-stationary motion process for echo extrapolation. The superiority of our TempEE has been demonstrated in the context of the classic radar echo extrapolation task, utilizing a real-world dataset. Extensive experiments have further validated the efficacy and indispensability of various components within TempEE."
                    },
                    {
                        "title": "Interpretable CNN-Multilevel Attention Transformer for Rapid Recognition of Pneumonia from Chest X-Ray Images",
                        "abstract": "Chest imaging plays an essential role in diagnosing and predicting patients with COVID-19 with evidence of worsening respiratory status. Many deep learning-based approaches for pneumonia recognition have been developed to enable computer-aided diagnosis. However, the long training and inference time makes them inflexible, and the lack of interpretability reduces their credibility in clinical medical practice. This paper aims to develop a pneumonia recognition framework with interpretability, which can understand the complex relationship between lung features and related diseases in chest X-ray (CXR) images to provide high-speed analytics support for medical practice. To reduce the computational complexity to accelerate the recognition process, a novel multi-level self-attention mechanism within Transformer has been proposed to accelerate convergence and emphasize the task-related feature regions. Moreover, a practical CXR image data augmentation has been adopted to address the scarcity of medical image data problems to boost the model's performance. The effectiveness of the proposed method has been demonstrated on the classic COVID-19 recognition task using the widespread pneumonia CXR image dataset. In addition, abundant ablation experiments validate the effectiveness and necessity of all of the components of the proposed method."
                    },
                    {
                        "title": "Free Lunch for Federated Remote Sensing Target Fine-Grained Classification: A Parameter-Efficient Framework",
                        "abstract": "Remote Sensing Target Fine-grained Classification (TFGC) is of great significance in both military and civilian fields. Due to location differences, growth in data size, and centralized server storage constraints, these data are usually stored under different databases across regions/countries. However, privacy laws and national security concerns constrain researchers from accessing these sensitive remote sensing images for further analysis. Additionally, low-resource remote sensing devices encounter challenges in terms of communication overhead and efficiency when dealing with the ever-increasing data and model scales. To solve the above challenges, this paper proposes a novel Privacy-Reserving TFGC Framework based on Federated Learning, dubbed PRFL. The proposed framework allows each client to learn global and local knowledge to enhance the local representation of private data in environments with extreme statistical heterogeneity (non. Independent and Identically Distributed, IID). Thus, it provides highly customized models to clients with differentiated data distributions. Moreover, the framework minimizes communication overhead and improves efficiency while ensuring satisfactory performance, thereby enhancing robustness and practical applicability under resource-scarce conditions. We demonstrate the effectiveness of the proposed PRFL on the classical TFGC task by leveraging four public datasets."
                    },
                    {
                        "title": "Foundation Models for Weather and Climate Data Understanding: A Comprehensive Survey",
                        "abstract": "As artificial intelligence (AI) continues to rapidly evolve, the realm of Earth and atmospheric sciences is increasingly adopting data-driven models, powered by progressive developments in deep learning (DL). Specifically, DL techniques are extensively utilized to decode the chaotic and nonlinear aspects of Earth systems, and to address climate challenges via understanding weather and climate data. Cutting-edge performance on specific tasks within narrower spatio-temporal scales has been achieved recently through DL. The rise of large models, specifically large language models (LLMs), has enabled fine-tuning processes that yield remarkable outcomes across various downstream tasks, thereby propelling the advancement of general AI. However, we are still navigating the initial stages of crafting general AI for weather and climate. In this survey, we offer an exhaustive, timely overview of state-of-the-art AI methodologies specifically engineered for weather and climate data, with a special focus on time series and text data. Our primary coverage encompasses four critical aspects: types of weather and climate data, principal model architectures, model scopes and applications, and datasets for weather and climate. Furthermore, in relation to the creation and application of foundation models for weather and climate data understanding, we delve into the field's prevailing challenges, offer crucial insights, and propose detailed avenues for future research. This comprehensive approach equips practitioners with the requisite knowledge to make substantial progress in this domain. Our survey encapsulates the most recent breakthroughs in research on large, data-driven models for weather and climate data understanding, emphasizing robust foundations, current advancements, practical applications, crucial resources, and prospective research opportunities."
                    }
                ]
            },
            "14b6cef7-2b61-4c4c-95ce-5569518e9e3a": {
                "pk": "14b6cef7-2b61-4c4c-95ce-5569518e9e3a",
                "name": "Guodong Long",
                "collaborators": [
                    "Yucheng Zhou",
                    "Peng Yan",
                    "Jing Jiang",
                    "Haiyan Zhao",
                    "Zhiwei Li",
                    "Yang Li",
                    "Tao Shen",
                    "Zhihong Deng",
                    "Chengqi Zhang"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Federated Learning",
                    "Reinforcement Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Multimodal Event Transformer for Image-guided Story Ending Generation",
                        "abstract": "Image-guided story ending generation (IgSEG) is to generate a story ending based on given story plots and ending image. Existing methods focus on cross-modal feature fusion but overlook reasoning and mining implicit information from story plots and ending image. To tackle this drawback, we propose a multimodal event transformer, an event-based reasoning framework for IgSEG. Specifically, we construct visual and semantic event graphs from story plots and ending image, and leverage event-based reasoning to reason and mine implicit information in a single modality. Next, we connect visual and semantic event graphs and utilize cross-modal fusion to integrate different-modality features. In addition, we propose a multimodal injector to adaptive pass essential information to decoder. Besides, we present an incoherence detection to enhance the understanding context of a story plot and the robustness of graph modeling for our model. Experimental results show that our method achieves state-of-the-art performance for the image-guided story ending generation."
                    },
                    {
                        "title": "Improving Cross-modal Alignment for Text-Guided Image Inpainting",
                        "abstract": "Text-guided image inpainting (TGII) aims to restore missing regions based on a given text in a damaged image. Existing methods are based on a strong vision encoder and a cross-modal fusion model to integrate cross-modal features. However, these methods allocate most of the computation to visual encoding, while light computation on modeling modality interactions. Moreover, they take cross-modal fusion for depth features, which ignores a fine-grained alignment between text and image. Recently, vision-language pre-trained models (VLPM), encapsulating rich cross-modal alignment knowledge, have advanced in most multimodal tasks. In this work, we propose a novel model for TGII by improving cross-modal alignment (CMA). CMA model consists of a VLPM as a vision-language encoder, an image generator and global-local discriminators. To explore cross-modal alignment knowledge for image restoration, we introduce cross-modal alignment distillation and in-sample distribution distillation. In addition, we employ adversarial training to enhance the model to fill the missing region in complicated structures effectively. Experiments are conducted on two popular vision-language datasets. Results show that our model achieves state-of-the-art performance compared with other strong competitors."
                    },
                    {
                        "title": "Style-Aware Contrastive Learning for Multi-Style Image Captioning",
                        "abstract": "Existing multi-style image captioning methods show promising results in generating a caption with accurate visual content and desired linguistic style. However, existing methods overlook the relationship between linguistic style and visual content. To overcome this drawback, we propose style-aware contrastive learning for multi-style image captioning. First, we present a style-aware visual encoder with contrastive learning to mine potential visual content relevant to style. Moreover, we propose a style-aware triplet contrast objective to distinguish whether the image, style and caption matched. To provide positive and negative samples for contrastive learning, we present three retrieval schemes: object-based retrieval, RoI-based retrieval and triplet-based retrieval, and design a dynamic trade-off function to calculate retrieval scores. Experimental results demonstrate that our approach achieves state-of-the-art performance. In addition, we conduct an extensive analysis to verify the effectiveness of our method."
                    },
                    {
                        "title": "Personalization Disentanglement for Federated Learning: An explainable perspective",
                        "abstract": "Personalized federated learning (PFL) jointly trains a variety of local models through balancing between knowledge sharing across clients and model personalization per client. This paper addresses PFL via explicit disentangling latent representations into two parts to capture the shared knowledge and client-specific personalization, which leads to more reliable and effective PFL. The disentanglement is achieved by a novel Federated Dual Variational Autoencoder (FedDVA), which employs two encoders to infer the two types of representations. FedDVA can produce a better understanding of the trade-off between global knowledge sharing and local personalization in PFL. Moreover, it can be integrated with existing FL methods and turn them into personalized models for heterogeneous downstream tasks. Extensive experiments validate the advantages caused by disentanglement and show that models trained with disentangled representations substantially outperform those vanilla methods."
                    },
                    {
                        "title": "One-Shot Pruning for Fast-adapting Pre-trained Models on Devices",
                        "abstract": "Large-scale pre-trained models have been remarkably successful in resolving downstream tasks. Nonetheless, deploying these models on low-capability devices still requires an effective approach, such as model pruning. However, pruning the model from scratch can pose a practical challenge given the limited resources of each downstream task or device. To tackle this issue, we present a scalable one-shot pruning method that leverages pruned knowledge of similar tasks to extract a sub-network from the pre-trained model for a new task. Specifically, we create a score mask using the pruned models of similar tasks to identify task-specific filters/nodes in the pre-trained model for the new task. Based on this mask, we conduct a single round of pruning to extract a suitably-sized sub-network that can quickly adapt to the new task with only a few training iterations. Our experimental analysis demonstrates the effectiveness of the proposed method on the convolutional neural networks (CNNs) and vision transformers (ViT) with various datasets. The proposed method consistently outperforms popular pruning baseline methods in terms of accuracy and efficiency when dealing with diverse downstream tasks with different memory constraints."
                    },
                    {
                        "title": "Client-supervised Federated Learning: Towards One-model-for-all Personalization",
                        "abstract": "Personalized Federated Learning (PerFL) is a new machine learning paradigm that delivers personalized models for diverse clients under federated learning settings. Most PerFL methods require extra learning processes on a client to adapt a globally shared model to the client-specific personalized model using its own local data. However, the model adaptation process in PerFL is still an open challenge in the stage of model deployment and test time. This work tackles the challenge by proposing a novel federated learning framework to learn only one robust global model to achieve competitive performance to those personalized models on unseen/test clients in the FL system. Specifically, we design a new Client-Supervised Federated Learning (FedCS) to unravel clients' bias on instances' latent representations so that the global model can learn both client-specific and client-agnostic knowledge. Experimental study shows that the FedCS can learn a robust FL global model for the changing data distributions of unseen/test clients. The FedCS's global model can be directly deployed to the test clients while achieving comparable performance to other personalized FL methods that require model adaptation."
                    },
                    {
                        "title": "Navigating the Future of Federated Recommendation Systems with Foundation Models",
                        "abstract": "In recent years, the integration of federated learning (FL) and recommendation systems (RS), known as Federated Recommendation Systems (FRS), has attracted attention for preserving user privacy by keeping private data on client devices. However, FRS faces inherent limitations such as data heterogeneity and scarcity, due to the privacy requirements of FL and the typical data sparsity issues of RSs. Models like ChatGPT are empowered by the concept of transfer learning and self-supervised learning, so they can be easily applied to the downstream tasks after fine-tuning or prompting. These models, so-called Foundation Models (FM), fouce on understanding the human's intent and perform following their designed roles in the specific tasks, which are widely recognized for producing high-quality content in the image and language domains. Thus, the achievements of FMs inspire the design of FRS and suggest a promising research direction: integrating foundation models to address the above limitations. In this study, we conduct a comprehensive review of FRSs with FMs. Specifically, we: 1) summarise the common approaches of current FRSs and FMs; 2) review the challenges posed by FRSs and FMs; 3) discuss potential future research directions; and 4) introduce some common benchmarks and evaluation metrics in the FRS field. We hope that this position paper provides the necessary background and guidance to explore this interesting and emerging topic."
                    },
                    {
                        "title": "Hierarchical Relation-Guided Type-Sentence Alignment for Long-Tail Relation Extraction with Distant Supervision",
                        "abstract": "Distant supervision uses triple facts in knowledge graphs to label a corpus for relation extraction, leading to wrong labeling and long-tail problems. Some works use the hierarchy of relations for knowledge transfer to long-tail relations. However, a coarse-grained relation often implies only an attribute (e.g., domain or topic) of the distant fact, making it hard to discriminate relations based solely on sentence semantics. One solution is resorting to entity types, but open questions remain about how to fully leverage the information of entity types and how to align multi-granular entity types with sentences. In this work, we propose a novel model to enrich distantly-supervised sentences with entity types. It consists of (1) a pairwise type-enriched sentence encoding module injecting both context-free and -related backgrounds to alleviate sentence-level wrong labeling, and (2) a hierarchical type-sentence alignment module enriching a sentence with the triple fact's basic attributes to support long-tail relations. Our model achieves new state-of-the-art results in overall and long-tail performance on benchmarks."
                    },
                    {
                        "title": "What Hides behind Unfairness? Exploring Dynamics Fairness in Reinforcement Learning",
                        "abstract": "In sequential decision-making problems involving sensitive attributes like race and gender, reinforcement learning (RL) agents must carefully consider long-term fairness while maximizing returns. Recent works have proposed many different types of fairness notions, but how unfairness arises in RL problems remains unclear. In this paper, we address this gap in the literature by investigating the sources of inequality through a causal lens. We first analyse the causal relationships governing the data generation process and decompose the effect of sensitive attributes on long-term well-being into distinct components. We then introduce a novel notion called dynamics fairness, which explicitly captures the inequality stemming from environmental dynamics, distinguishing it from those induced by decision-making or inherited from the past. This notion requires evaluating the expected changes in the next state and the reward induced by changing the value of the sensitive attribute while holding everything else constant. To quantitatively evaluate this counterfactual concept, we derive identification formulas that allow us to obtain reliable estimations from data. Extensive experiments demonstrate the effectiveness of the proposed techniques in explaining, detecting, and reducing inequality in reinforcement learning. We publicly release code at https://github.com/familyld/InsightFair."
                    }
                ]
            },
            "e9206552-a15f-4f35-aa18-8887b014e5e0": {
                "pk": "e9206552-a15f-4f35-aa18-8887b014e5e0",
                "name": "Jing Jiang",
                "collaborators": [
                    "Shuohang Wang",
                    "Zhanqiang Bai",
                    "Weihong Deng",
                    "Rui Cao",
                    "Krishna R. Narayanan",
                    "Sicheng Yu",
                    "Hao Zhang",
                    "Wei Jing",
                    "Lingxiao Qu",
                    "Yaoyiran Li"
                ],
                "domain": [
                    "Lie Algebra",
                    "Natural Language Processing",
                    "Machine Learning",
                    "Coding Theory"
                ],
                "publications": [
                    {
                        "title": "Reducibility of scalar generalized Verma modules of minimal parabolic type",
                        "abstract": "Let $\\mathfrak{g}$ be a classical complex simple Lie algebra and $\\mathfrak{q}$ be a parabolic subalgebra. Generalized Verma module $M$ is called a scalar generalized Verma module if it is induced from a one-dimensional representation of $\\mathfrak{q}$. In this paper, we will determine the first diagonal-reducible point of scalar generalized Verma modules associated to minimal parabolic subalgebras by computing explicitly the Gelfand-Kirillov dimension of the corresponding highest weight modules."
                    },
                    {
                        "title": "Multimodal Story Generation on Plural Images",
                        "abstract": "Traditionally, text generation models take in a sequence of text as input, and iteratively generate the next most probable word using pre-trained parameters. In this work, we propose the architecture to use images instead of text as the input of the text generation model, called StoryGen. In the architecture, we design a Relational Text Data Generator algorithm that relates different features from multiple images. The output samples from the model demonstrate the ability to generate meaningful paragraphs of text containing the extracted features from the input images. This is an undergraduate project report. Completed Dec. 2019 at the Cooper Union."
                    },
                    {
                        "title": "Context Modeling with Evidence Filter for Multiple Choice Question Answering",
                        "abstract": "Multiple-Choice Question Answering (MCQA) is a challenging task in machine reading comprehension. The main challenge in MCQA is to extract \"evidence\" from the given context that supports the correct answer. In the OpenbookQA dataset, the requirement of extracting \"evidence\" is particularly important due to the mutual independence of sentences in the context. Existing work tackles this problem by annotated evidence or distant supervision with rules which overly rely on human efforts. To address the challenge, we propose a simple yet effective approach termed evidence filtering to model the relationships between the encoded contexts with respect to different options collectively and to potentially highlight the evidence sentences and filter out unrelated sentences. In addition to the effective reduction of human efforts of our approach compared, through extensive experiments on OpenbookQA, we show that the proposed approach outperforms the models that use the same backbone and more training data; and our parameter analysis also demonstrates the interpretability of our approach."
                    },
                    {
                        "title": "Gelfand-Kirillov dimensions and Reducibility of scalar type generalized Verma modules for classical Lie algebras",
                        "abstract": "Let $\\mathfrak{g}$ be a classial Lie algebra and $\\mathfrak{p}$ be a maximal parabolic subalgebra. Let $M$ be a generalized Verma module induced from a one dimensional representation of $\\mathfrak{p}$. Such $M$ is called a scalar type generalized Verma module. Its simple quotient $L$ is a highest weight moudle. In this paper, we will determine the reducibility of such scalar type generalized Verma modules by computing the Gelfand-Kirillov dimension of $L$."
                    },
                    {
                        "title": "On the annihilator variety of a highest weight Harish-Chandra module",
                        "abstract": "Let $G$ be a Hermitian type Lie group with maximal compact subgroup $K$. Let $L(\\lambda)$ be a highest weight Harish-Chandra module of $G$ with the infinitesimal character $\\lambda$. By using some combinatorial algorithm, we obtain a description of the annihilator variety of $L(\\lambda)$. As an application, when $L(\\lambda)$ is unitarizable, we prove that the Gelfand-Kirillov dimension of $L(\\lambda)$ only depends on the value of $z=(\\lambda,\\beta^{\\vee})$, where $\\beta$ is the highest root."
                    },
                    {
                        "title": "Cascaded Coded Distributed Computing Schemes Based on Placement Delivery Arrays",
                        "abstract": "Li {\\it et al}. introduced coded distributed computing (CDC) scheme to reduce the communication load in general distributed computing frameworks such as MapReduce. They also proposed cascaded CDC schemes where each output function is computed multiple times, and proved that such schemes achieved the fundamental trade-off between computation load and communication load. However, these schemes require exponentially large numbers of input files and output functions when the number of computing nodes gets large. In this paper, by using the structure of placement delivery arrays (PDAs), we construct several infinite classes of cascaded CDC schemes. We also show that the numbers of output functions in all the new schemes are only a factor of the number of computing nodes, and the number of input files in our new schemes is much smaller than that of input files in CDC schemes derived by Li {\\it et al}."
                    },
                    {
                        "title": "Disentangling Identity and Pose for Facial Expression Recognition",
                        "abstract": "Facial expression recognition (FER) is a challenging problem because the expression component is always entangled with other irrelevant factors, such as identity and head pose. In this work, we propose an identity and pose disentangled facial expression recognition (IPD-FER) model to learn more discriminative feature representation. We regard the holistic facial representation as the combination of identity, pose and expression. These three components are encoded with different encoders. For identity encoder, a well pre-trained face recognition model is utilized and fixed during training, which alleviates the restriction on specific expression training data in previous works and makes the disentanglement practicable on in-the-wild datasets. At the same time, the pose and expression encoder are optimized with corresponding labels. Combining identity and pose feature, a neutral face of input individual should be generated by the decoder. When expression feature is added, the input image should be reconstructed. By comparing the difference between synthesized neutral and expressional images of the same individual, the expression component is further disentangled from identity and pose. Experimental results verify the effectiveness of our method on both lab-controlled and in-the-wild databases and we achieve state-of-the-art recognition performance."
                    },
                    {
                        "title": "A Compare-Aggregate Model for Matching Text Sequences",
                        "abstract": "Many NLP tasks including machine comprehension, answer selection and text entailment require the comparison between sequences. Matching the important units between sequences is a key to solve these problems. In this paper, we present a general \"compare-aggregate\" framework that performs word-level matching followed by aggregation using Convolutional Neural Networks. We particularly focus on the different comparison functions we can use to match two vectors. We use four different datasets to evaluate the model. We find that some simple comparison functions based on element-wise operations can work better than standard neural network and neural tensor network."
                    },
                    {
                        "title": "Two-Headed Monster And Crossed Co-Attention Networks",
                        "abstract": "This paper presents some preliminary investigations of a new co-attention mechanism in neural transduction models. We propose a paradigm, termed Two-Headed Monster (THM), which consists of two symmetric encoder modules and one decoder module connected with co-attention. As a specific and concrete implementation of THM, Crossed Co-Attention Networks (CCNs) are designed based on the Transformer model. We demonstrate CCNs on WMT 2014 EN-DE and WMT 2016 EN-FI translation tasks and our model outperforms the strong Transformer baseline by 0.51 (big) and 0.74 (base) BLEU points on EN-DE and by 0.17 (big) and 0.47 (base) BLEU points on EN-FI."
                    },
                    {
                        "title": "Boosting Facial Expression Recognition by A Semi-Supervised Progressive Teacher",
                        "abstract": "In this paper, we aim to improve the performance of in-the-wild Facial Expression Recognition (FER) by exploiting semi-supervised learning. Large-scale labeled data and deep learning methods have greatly improved the performance of image recognition. However, the performance of FER is still not ideal due to the lack of training data and incorrect annotations (e.g., label noises). Among existing in-the-wild FER datasets, reliable ones contain insufficient data to train robust deep models while large-scale ones are annotated in lower quality. To address this problem, we propose a semi-supervised learning algorithm named Progressive Teacher (PT) to utilize reliable FER datasets as well as large-scale unlabeled expression images for effective training. On the one hand, PT introduces semi-supervised learning method to relieve the shortage of data in FER. On the other hand, it selects useful labeled training samples automatically and progressively to alleviate label noise. PT uses selected clean labeled data for computing the supervised classification loss and unlabeled data for unsupervised consistency loss. Experiments on widely-used databases RAF-DB and FERPlus validate the effectiveness of our method, which achieves state-of-the-art performance with accuracy of 89.57% on RAF-DB. Additionally, when the synthetic noise rate reaches even 30%, the performance of our PT algorithm only degrades by 4.37%."
                    },
                    {
                        "title": "Self-Tuning and de Sitter Brane Intersections in the 6-Dimensional Brane Models",
                        "abstract": "We study the self-tuning of general brane junctions and brane networks on the 6-dimensional space-time. For the general brane junctions, there may exist one fine-tuning among the brane tensions. For the brane networks, similar to the 5-dimensional self-tuning brane models, the brane tensions can be set arbitrarily and there exists the singularity for the metric and bulk scalar. And if we want to regularize the singularity, we will introduce the fine-tuning among the brane tensions. In addition, because the 4-dimensional cosmological constant we observe may be positive and very small, we discuss the brane network with de Sitter brane intersections by introducing a bulk scalar."
                    },
                    {
                        "title": "Learning Natural Language Inference with LSTM",
                        "abstract": "Natural language inference (NLI) is a fundamentally important task in natural language processing that has many applications. The recently released Stanford Natural Language Inference (SNLI) corpus has made it possible to develop and evaluate learning-centered methods such as deep neural networks for natural language inference (NLI). In this paper, we propose a special long short-term memory (LSTM) architecture for NLI. Our model builds on top of a recently proposed neural attention model for NLI but is based on a significantly different idea. Instead of deriving sentence embeddings for the premise and the hypothesis to be used for classification, our solution uses a match-LSTM to perform word-by-word matching of the hypothesis with the premise. This LSTM is able to place more emphasis on important word-level matching results. In particular, we observe that this LSTM remembers important mismatches that are critical for predicting the contradiction or the neutral relationship label. On the SNLI corpus, our model achieves an accuracy of 86.1%, outperforming the state of the art."
                    },
                    {
                        "title": "Machine Comprehension Using Match-LSTM and Answer Pointer",
                        "abstract": "Machine comprehension of text is an important problem in natural language processing. A recently released dataset, the Stanford Question Answering Dataset (SQuAD), offers a large number of real questions and their answers created by humans through crowdsourcing. SQuAD provides a challenging testbed for evaluating machine comprehension algorithms, partly because compared with previous datasets, in SQuAD the answers do not come from a small set of candidate answers and they have variable lengths. We propose an end-to-end neural architecture for the task. The architecture is based on match-LSTM, a model we proposed previously for textual entailment, and Pointer Net, a sequence-to-sequence model proposed by Vinyals et al.(2015) to constrain the output tokens to be from the input sequences. We propose two ways of using Pointer Net for our task. Our experiments show that both of our two models substantially outperform the best results obtained by Rajpurkar et al.(2016) using logistic regression and manually crafted features."
                    },
                    {
                        "title": "A BERT-based Dual Embedding Model for Chinese Idiom Prediction",
                        "abstract": "Chinese idioms are special fixed phrases usually derived from ancient stories, whose meanings are oftentimes highly idiomatic and non-compositional. The Chinese idiom prediction task is to select the correct idiom from a set of candidate idioms given a context with a blank. We propose a BERT-based dual embedding model to encode the contextual words as well as to learn dual embeddings of the idioms. Specifically, we first match the embedding of each candidate idiom with the hidden representation corresponding to the blank in the context. We then match the embedding of each candidate idiom with the hidden representations of all the tokens in the context thorough context pooling. We further propose to use two separate idiom embeddings for the two kinds of matching. Experiments on a recently released Chinese idiom cloze test dataset show that our proposed method performs better than the existing state of the art. Ablation experiments also show that both context pooling and dual embedding contribute to the improvement of performance."
                    },
                    {
                        "title": "An Empirical Study of Memorization in NLP",
                        "abstract": "A recent study by Feldman (2020) proposed a long-tail theory to explain the memorization behavior of deep learning models. However, memorization has not been empirically verified in the context of NLP, a gap addressed by this work. In this paper, we use three different NLP tasks to check if the long-tail theory holds. Our experiments demonstrate that top-ranked memorized training instances are likely atypical, and removing the top-memorized training instances leads to a more serious drop in test accuracy compared with removing training instances randomly. Furthermore, we develop an attribution method to better understand why a training instance is memorized. We empirically show that our memorization attribution method is faithful, and share our interesting finding that the top-memorized parts of a training instance tend to be features negatively correlated with the class label."
                    },
                    {
                        "title": "Modularized Zero-shot VQA with Pre-trained Models",
                        "abstract": "Large-scale pre-trained models (PTMs) show great zero-shot capabilities. In this paper, we study how to leverage them for zero-shot visual question answering (VQA). Our approach is motivated by a few observations. First, VQA questions often require multiple steps of reasoning, which is still a capability that most PTMs lack. Second, different steps in VQA reasoning chains require different skills such as object detection and relational reasoning, but a single PTM may not possess all these skills. Third, recent work on zero-shot VQA does not explicitly consider multi-step reasoning chains, which makes them less interpretable compared with a decomposition-based approach. We propose a modularized zero-shot network that explicitly decomposes questions into sub reasoning steps and is highly interpretable. We convert sub reasoning tasks to acceptable objectives of PTMs and assign tasks to proper PTMs without any adaptation. Our experiments on two VQA benchmarks under the zero-shot setting demonstrate the effectiveness of our method and better interpretability compared with several baselines."
                    },
                    {
                        "title": "Knowledge Generation for Zero-shot Knowledge-based VQA",
                        "abstract": "Previous solutions to knowledge-based visual question answering~(K-VQA) retrieve knowledge from external knowledge bases and use supervised learning to train the K-VQA model. Recently pre-trained LLMs have been used as both a knowledge source and a zero-shot QA model for K-VQA and demonstrated promising results. However, these recent methods do not explicitly show the knowledge needed to answer the questions and thus lack interpretability. Inspired by recent work on knowledge generation from LLMs for text-based QA, in this work we propose and test a similar knowledge-generation-based K-VQA method, which first generates knowledge from an LLM and then incorporates the generated knowledge for K-VQA in a zero-shot manner. We evaluate our method on two K-VQA benchmarks and found that our method performs better than previous zero-shot K-VQA methods and our generated knowledge is generally relevant and helpful."
                    },
                    {
                        "title": "Performance Analysis of Algebraic Soft Decoding of Reed-Solomon Codes over Binary Symmetric and Erasure Channels",
                        "abstract": "In this paper, we characterize the decoding region of algebraic soft decoding (ASD) of Reed-Solomon (RS) codes over erasure channels and binary symmetric channel (BSC). Optimal multiplicity assignment strategies (MAS) are investigated and tight bounds are derived to show the ASD can significantly outperform conventional Berlekamp Massey (BM) decoding over these channels for a wide code rate range. The analysis technique can also be extended to other channel models, e.g., RS coded modulation over erasure channels."
                    },
                    {
                        "title": "Multilevel Coding for Channels with Non-uniform Inputs and Rateless Transmission over the BSC",
                        "abstract": "We consider coding schemes for channels with non-uniform inputs (NUI), where standard linear block codes can not be applied directly. We show that multilevel coding (MLC) with a set of linear codes and a deterministic mapper can achieve the information rate of the channel with NUI. The mapper, however, does not have to be one-to-one. As an application of the proposed MLC scheme, we present a rateless transmission scheme over the binary symmetric channel (BSC)."
                    }
                ]
            },
            "1da4b60c-e6ee-4f03-934e-2789add07ab2": {
                "pk": "1da4b60c-e6ee-4f03-934e-2789add07ab2",
                "name": "Chengqi Zhang",
                "collaborators": [
                    "Guodong Long",
                    "Jing Jiang",
                    "Daokun Zhang",
                    "Jie Yin",
                    "Xingquan Zhu",
                    "Tianyi Zhou",
                    "Yue Tan",
                    "Zhihong Deng",
                    "Zonghan Wu",
                    "Shirui Pan"
                ],
                "domain": [
                    "Network Analysis",
                    "Federated Learning",
                    "Graph Neural Network",
                    "Time Series Analysis"
                ],
                "publications": [
                    {
                        "title": "Temporal Feature Selection on Networked Time Series",
                        "abstract": "This paper formulates the problem of learning discriminative features (\\textit{i.e.,} segments) from networked time series data considering the linked information among time series. For example, social network users are considered to be social sensors that continuously generate social signals (tweets) represented as a time series. The discriminative segments are often referred to as \\emph{shapelets} in a time series. Extracting shapelets for time series classification has been widely studied. However, existing works on shapelet selection assume that the time series are independent and identically distributed (i.i.d.). This assumption restricts their applications to social networked time series analysis, since a user's actions can be correlated to his/her social affiliations. In this paper we propose a new Network Regularized Least Squares (NetRLS) feature selection model that combines typical time series data and user network data for analysis. Experiments on real-world networked time series Twitter and DBLP data demonstrate the performance of the proposed method. NetRLS performs better than LTS, the state-of-the-art time series feature selection approach, on real-world data."
                    },
                    {
                        "title": "Session Communication and Integration",
                        "abstract": "The scenario-based specification of a large distributed system is usually naturally decomposed into various modules. The integration of specification modules contrasts to the parallel composition of program components, and includes various ways such as scenario concatenation, choice, and nesting. The recent development of multiparty session types for process calculi provides useful techniques to accommodate the protocol modularisation, by encoding fragments of communication protocols in the usage of private channels for a class of agents. In this paper, we extend forgoing session type theories by enhancing the session integration mechanism. More specifically, we propose a novel synchronous multiparty session type theory, in which sessions are separated into the communicating and integrating levels. Communicating sessions record the message-based communications between multiple agents, whilst integrating sessions describe the integration of communicating ones. A two-level session type system is developed for pi-calculus with syntactic primitives for session establishment, and several key properties of the type system are studied. Applying the theory to system description, we show that a channel safety property and a session conformance property can be analysed. Also, to improve the utility of the theory, a process slicing method is used to help identify the violated sessions in the type checking."
                    },
                    {
                        "title": "Federated Learning for Open Banking",
                        "abstract": "Open banking enables individual customers to own their banking data, which provides fundamental support for the boosting of a new ecosystem of data marketplaces and financial services. In the near future, it is foreseeable to have decentralized data ownership in the finance sector using federated learning. This is a just-in-time technology that can learn intelligent models in a decentralized training manner. The most attractive aspect of federated learning is its ability to decompose model training into a centralized server and distributed nodes without collecting private data. This kind of decomposed learning framework has great potential to protect users' privacy and sensitive data. Therefore, federated learning combines naturally with an open banking data marketplaces. This chapter will discuss the possible challenges for applying federated learning in the context of open banking, and the corresponding solutions have been explored as well."
                    },
                    {
                        "title": "Causal Reinforcement Learning: A Survey",
                        "abstract": "Reinforcement learning is an essential paradigm for solving sequential decision problems under uncertainty. Despite many remarkable achievements in recent decades, applying reinforcement learning methods in the real world remains challenging. One of the main obstacles is that reinforcement learning agents lack a fundamental understanding of the world and must therefore learn from scratch through numerous trial-and-error interactions. They may also face challenges in providing explanations for their decisions and generalizing the acquired knowledge. Causality, however, offers a notable advantage as it can formalize knowledge in a systematic manner and leverage invariance for effective knowledge transfer. This has led to the emergence of causal reinforcement learning, a subfield of reinforcement learning that seeks to enhance existing algorithms by incorporating causal relationships into the learning process. In this survey, we comprehensively review the literature on causal reinforcement learning. We first introduce the basic concepts of causality and reinforcement learning, and then explain how causality can address core challenges in non-causal reinforcement learning. We categorize and systematically review existing causal reinforcement learning approaches based on their target problems and methodologies. Finally, we outline open issues and future directions in this emerging field."
                    },
                    {
                        "title": "Personalized Item Representations in Federated Multimodal Recommendation",
                        "abstract": "Federated recommendation systems are essential for providing personalized recommendations while protecting user privacy. However, current methods mainly rely on ID-based item embeddings, neglecting the rich multimodal information of items. To address this, we propose a Federated Multimodal Recommendation System, called FedMR. FedMR uses a foundation model on the server to encode multimodal item data, such as images and text. To handle data heterogeneity caused by user preference differences, FedMR introduces a Mixing Feature Fusion Module on each client, which adjusts fusion strategy weights based on user interaction history to generate personalized item representations that capture users' fine-grained preferences. FedMR is compatible with existing ID-based federated recommendation systems, improving performance without modifying the original framework. Experiments on four real-world multimodal datasets demonstrate FedMR's effectiveness. The code is available at https://anonymous.4open.science/r/FedMR."
                    },
                    {
                        "title": "What Hides behind Unfairness? Exploring Dynamics Fairness in Reinforcement Learning",
                        "abstract": "In sequential decision-making problems involving sensitive attributes like race and gender, reinforcement learning (RL) agents must carefully consider long-term fairness while maximizing returns. Recent works have proposed many different types of fairness notions, but how unfairness arises in RL problems remains unclear. In this paper, we address this gap in the literature by investigating the sources of inequality through a causal lens. We first analyse the causal relationships governing the data generation process and decompose the effect of sensitive attributes on long-term well-being into distinct components. We then introduce a novel notion called dynamics fairness, which explicitly captures the inequality stemming from environmental dynamics, distinguishing it from those induced by decision-making or inherited from the past. This notion requires evaluating the expected changes in the next state and the reward induced by changing the value of the sensitive attribute while holding everything else constant. To quantitatively evaluate this counterfactual concept, we derive identification formulas that allow us to obtain reliable estimations from data. Extensive experiments demonstrate the effectiveness of the proposed techniques in explaining, detecting, and reducing inequality in reinforcement learning. We publicly release code at https://github.com/familyld/InsightFair."
                    },
                    {
                        "title": "Network Representation Learning: A Survey",
                        "abstract": "With the widespread use of information technologies, information networks are becoming increasingly popular to capture complex relationships across various disciplines, such as social networks, citation networks, telecommunication networks, and biological networks. Analyzing these networks sheds light on different aspects of social life such as the structure of societies, information diffusion, and communication patterns. In reality, however, the large scale of information networks often makes network analytic tasks computationally expensive or intractable. Network representation learning has been recently proposed as a new learning paradigm to embed network vertices into a low-dimensional vector space, by preserving network topology structure, vertex content, and other side information. This facilitates the original network to be easily handled in the new vector space for further analysis. In this survey, we perform a comprehensive review of the current literature on network representation learning in the data mining and machine learning field. We propose new taxonomies to categorize and summarize the state-of-the-art network representation learning techniques according to the underlying learning mechanisms, the network information intended to preserve, as well as the algorithmic designs and methodologies. We summarize evaluation protocols used for validating network representation learning including published benchmark datasets, evaluation methods, and open source algorithms. We also perform empirical studies to compare the performance of representative algorithms on common datasets, and analyze their computational complexity. Finally, we suggest promising research directions to facilitate future study."
                    },
                    {
                        "title": "SINE: Scalable Incomplete Network Embedding",
                        "abstract": "Attributed network embedding aims to learn low-dimensional vector representations for nodes in a network, where each node contains rich attributes/features describing node content. Because network topology structure and node attributes often exhibit high correlation, incorporating node attribute proximity into network embedding is beneficial for learning good vector representations. In reality, large-scale networks often have incomplete/missing node content or linkages, yet existing attributed network embedding algorithms all operate under the assumption that networks are complete. Thus, their performance is vulnerable to missing data and suffers from poor scalability.   In this paper, we propose a Scalable Incomplete Network Embedding (SINE) algorithm for learning node representations from incomplete graphs. SINE formulates a probabilistic learning framework that separately models pairs of node-context and node-attribute relationships. Different from existing attributed network embedding algorithms, SINE provides greater flexibility to make the best of useful information and mitigate negative effects of missing information on representation learning. A stochastic gradient descent based online algorithm is derived to learn node representations, allowing SINE to scale up to large-scale networks with high learning efficiency. We evaluate the effectiveness and efficiency of SINE through extensive experiments on real-world networks. Experimental results confirm that SINE outperforms state-of-the-art baselines in various tasks, including node classification, node clustering, and link prediction, under settings with missing links and node attributes. SINE is also shown to be scalable and efficient on large-scale networks with millions of nodes/edges and high-dimensional node features. The source code of this paper is available at https://github.com/daokunzhang/SINE."
                    },
                    {
                        "title": "MetaGraph2Vec: Complex Semantic Path Augmented Heterogeneous Network Embedding",
                        "abstract": "Network embedding in heterogeneous information networks (HINs) is a challenging task, due to complications of different node types and rich relationships between nodes. As a result, conventional network embedding techniques cannot work on such HINs. Recently, metapath-based approaches have been proposed to characterize relationships in HINs, but they are ineffective in capturing rich contexts and semantics between nodes for embedding learning, mainly because (1) metapath is a rather strict single path node-node relationship descriptor, which is unable to accommodate variance in relationships, and (2) only a small portion of paths can match the metapath, resulting in sparse context information for embedding learning. In this paper, we advocate a new metagraph concept to capture richer structural contexts and semantics between distant nodes. A metagraph contains multiple paths between nodes, each describing one type of relationships, so the augmentation of multiple metapaths provides an effective way to capture rich contexts and semantic relations between nodes. This greatly boosts the ability of metapath-based embedding techniques in handling very sparse HINs. We propose a new embedding learning algorithm, namely MetaGraph2Vec, which uses metagraph to guide the generation of random walks and to learn latent embeddings of multi-typed HIN nodes. Experimental results show that MetaGraph2Vec is able to outperform the state-of-the-art baselines in various heterogeneous network mining tasks such as node classification, node clustering, and similarity search."
                    },
                    {
                        "title": "Attributed Network Embedding via Subspace Discovery",
                        "abstract": "Network embedding aims to learn a latent, low-dimensional vector representations of network nodes, effective in supporting various network analytic tasks. While prior arts on network embedding focus primarily on preserving network topology structure to learn node representations, recently proposed attributed network embedding algorithms attempt to integrate rich node content information with network topological structure for enhancing the quality of network embedding. In reality, networks often have sparse content, incomplete node attributes, as well as the discrepancy between node attribute feature space and network structure space, which severely deteriorates the performance of existing methods. In this paper, we propose a unified framework for attributed network embedding-attri2vec-that learns node embeddings by discovering a latent node attribute subspace via a network structure guided transformation performed on the original attribute space. The resultant latent subspace can respect network structure in a more consistent way towards learning high-quality node representations. We formulate an optimization problem which is solved by an efficient stochastic gradient descent algorithm, with linear time complexity to the number of nodes. We investigate a series of linear and non-linear transformations performed on node attributes and empirically validate their effectiveness on various types of networks. Another advantage of attri2vec is its ability to solve out-of-sample problems, where embeddings of new coming nodes can be inferred from their node attributes through the learned mapping function. Experiments on various types of networks confirm that attri2vec is superior to state-of-the-art baselines for node classification, node clustering, as well as out-of-sample link prediction tasks. The source code of this paper is available at https://github.com/daokunzhang/attri2vec."
                    },
                    {
                        "title": "Search Efficient Binary Network Embedding",
                        "abstract": "Traditional network embedding primarily focuses on learning a continuous vector representation for each node, preserving network structure and/or node content information, such that off-the-shelf machine learning algorithms can be easily applied to the vector-format node representations for network analysis. However, the learned continuous vector representations are inefficient for large-scale similarity search, which often involves finding nearest neighbors measured by distance or similarity in a continuous vector space. In this paper, we propose a search efficient binary network embedding algorithm called BinaryNE to learn a binary code for each node, by simultaneously modeling node context relations and node attribute relations through a three-layer neural network. BinaryNE learns binary node representations through a stochastic gradient descent based online learning algorithm. The learned binary encoding not only reduces memory usage to represent each node, but also allows fast bit-wise comparisons to support faster node similarity search than using Euclidean distance or other distance measures. Extensive experiments and comparisons demonstrate that BinaryNE not only delivers more than 25 times faster search speed, but also provides comparable or better search quality than traditional continuous vector based network embedding methods. The binary codes learned by BinaryNE also render competitive performance on node classification and node clustering tasks. The source code of this paper is available at https://github.com/daokunzhang/BinaryNE."
                    },
                    {
                        "title": "A Review for Weighted MinHash Algorithms",
                        "abstract": "Data similarity (or distance) computation is a fundamental research topic which underpins many high-level applications based on similarity measures in machine learning and data mining. However, in large-scale real-world scenarios, the exact similarity computation has become daunting due to \"3V\" nature (volume, velocity and variety) of big data. In such cases, the hashing techniques have been verified to efficiently conduct similarity estimation in terms of both theory and practice. Currently, MinHash is a popular technique for efficiently estimating the Jaccard similarity of binary sets and furthermore, weighted MinHash is generalized to estimate the generalized Jaccard similarity of weighted sets. This review focuses on categorizing and discussing the existing works of weighted MinHash algorithms. In this review, we mainly categorize the Weighted MinHash algorithms into quantization-based approaches, \"active index\"-based ones and others, and show the evolution and inherent connection of the weighted MinHash algorithms, from the integer weighted MinHash algorithms to real-valued weighted MinHash ones (particularly the Consistent Weighted Sampling scheme). Also, we have developed a python toolbox for the algorithms, and released it in our github. Based on the toolbox, we experimentally conduct a comprehensive comparative study of the standard MinHash algorithm and the weighted MinHash ones."
                    },
                    {
                        "title": "Beyond Low-pass Filtering: Graph Convolutional Networks with Automatic Filtering",
                        "abstract": "Graph convolutional networks are becoming indispensable for deep learning from graph-structured data. Most of the existing graph convolutional networks share two big shortcomings. First, they are essentially low-pass filters, thus the potentially useful middle and high frequency band of graph signals are ignored. Second, the bandwidth of existing graph convolutional filters is fixed. Parameters of a graph convolutional filter only transform the graph inputs without changing the curvature of a graph convolutional filter function. In reality, we are uncertain about whether we should retain or cut off the frequency at a certain point unless we have expert domain knowledge. In this paper, we propose Automatic Graph Convolutional Networks (AutoGCN) to capture the full spectrum of graph signals and automatically update the bandwidth of graph convolutional filters. While it is based on graph spectral theory, our AutoGCN is also localized in space and has a spatial form. Experimental results show that AutoGCN achieves significant improvement over baseline methods which only work as low-pass filters."
                    },
                    {
                        "title": "On the Convergence of Clustered Federated Learning",
                        "abstract": "Knowledge sharing and model personalization are essential components to tackle the non-IID challenge in federated learning (FL). Most existing FL methods focus on two extremes: 1) to learn a shared model to serve all clients with non-IID data, and 2) to learn personalized models for each client, namely personalized FL. There is a trade-off solution, namely clustered FL or cluster-wise personalized FL, which aims to cluster similar clients into one cluster, and then learn a shared model for all clients within a cluster. This paper is to revisit the research of clustered FL by formulating them into a bi-level optimization framework that could unify existing methods. We propose a new theoretical analysis framework to prove the convergence by considering the clusterability among clients. In addition, we embody this framework in an algorithm, named Weighted Clustered Federated Learning (WeCFL). Empirical analysis verifies the theoretical results and demonstrates the effectiveness of the proposed WeCFL under the proposed cluster-wise non-IID settings."
                    },
                    {
                        "title": "Influence-oriented Personalized Federated Learning",
                        "abstract": "Traditional federated learning (FL) methods often rely on fixed weighting for parameter aggregation, neglecting the mutual influence by others. Hence, their effectiveness in heterogeneous data contexts is limited. To address this problem, we propose an influence-oriented federated learning framework, namely FedC^2I, which quantitatively measures Client-level and Class-level Influence to realize adaptive parameter aggregation for each client. Our core idea is to explicitly model the inter-client influence within an FL system via the well-crafted influence vector and influence matrix. The influence vector quantifies client-level influence, enables clients to selectively acquire knowledge from others, and guides the aggregation of feature representation layers. Meanwhile, the influence matrix captures class-level influence in a more fine-grained manner to achieve personalized classifier aggregation. We evaluate the performance of FedC^2I against existing federated learning methods under non-IID settings and the results demonstrate the superiority of our method."
                    },
                    {
                        "title": "Bi-Directional Block Self-Attention for Fast and Memory-Efficient Sequence Modeling",
                        "abstract": "Recurrent neural networks (RNN), convolutional neural networks (CNN) and self-attention networks (SAN) are commonly used to produce context-aware representations. RNN can capture long-range dependency but is hard to parallelize and not time-efficient. CNN focuses on local dependency but does not perform well on some tasks. SAN can model both such dependencies via highly parallelizable computation, but memory requirement grows rapidly in line with sequence length. In this paper, we propose a model, called \"bi-directional block self-attention network (Bi-BloSAN)\", for RNN/CNN-free sequence encoding. It requires as little memory as RNN but with all the merits of SAN. Bi-BloSAN splits the entire sequence into blocks, and applies an intra-block SAN to each block for modeling local context, then applies an inter-block SAN to the outputs for all blocks to capture long-range dependency. Thus, each SAN only needs to process a short sequence, and only a small amount of memory is required. Additionally, we use feature-level attention to handle the variation of contexts around the same word, and use forward/backward masks to encode temporal order information. On nine benchmark datasets for different NLP tasks, Bi-BloSAN achieves or improves upon state-of-the-art accuracy, and shows better efficiency-memory trade-off than existing RNN/CNN/SAN."
                    },
                    {
                        "title": "Tensorized Self-Attention: Efficiently Modeling Pairwise and Global Dependencies Together",
                        "abstract": "Neural networks equipped with self-attention have parallelizable computation, light-weight structure, and the ability to capture both long-range and local dependencies. Further, their expressive power and performance can be boosted by using a vector to measure pairwise dependency, but this requires to expand the alignment matrix to a tensor, which results in memory and computation bottlenecks. In this paper, we propose a novel attention mechanism called \"Multi-mask Tensorized Self-Attention\" (MTSA), which is as fast and as memory-efficient as a CNN, but significantly outperforms previous CNN-/RNN-/attention-based models. MTSA 1) captures both pairwise (token2token) and global (source2token) dependencies by a novel compatibility function composed of dot-product and additive attentions, 2) uses a tensor to represent the feature-wise alignment scores for better expressive power but only requires parallelizable matrix multiplications, and 3) combines multi-head with multi-dimensional attentions, and applies a distinct positional mask to each head (subspace), so the memory and computation can be distributed to multiple heads, each with sequential information encoded independently. The experiments show that a CNN/RNN-free model based on MTSA achieves state-of-the-art or competitive performance on nine NLP benchmarks with compelling memory- and time-efficiency."
                    },
                    {
                        "title": "Graph WaveNet for Deep Spatial-Temporal Graph Modeling",
                        "abstract": "Spatial-temporal graph modeling is an important task to analyze the spatial relations and temporal trends of components in a system. Existing approaches mostly capture the spatial dependency on a fixed graph structure, assuming that the underlying relation between entities is pre-determined. However, the explicit graph structure (relation) does not necessarily reflect the true dependency and genuine relation may be missing due to the incomplete connections in the data. Furthermore, existing methods are ineffective to capture the temporal trends as the RNNs or CNNs employed in these methods cannot capture long-range temporal sequences. To overcome these limitations, we propose in this paper a novel graph neural network architecture, Graph WaveNet, for spatial-temporal graph modeling. By developing a novel adaptive dependency matrix and learn it through node embedding, our model can precisely capture the hidden spatial dependency in the data. With a stacked dilated 1D convolution component whose receptive field grows exponentially as the number of layers increases, Graph WaveNet is able to handle very long sequences. These two components are integrated seamlessly in a unified framework and the whole framework is learned in an end-to-end manner. Experimental results on two public traffic network datasets, METR-LA and PEMS-BAY, demonstrate the superior performance of our algorithm."
                    }
                ]
            }
        }
    },
    "2401.10166": {
        "paper_data": {
            "title": "VMamba: Visual State Space Model",
            "url": "http://arxiv.org/abs/2401.10166v3",
            "arxiv_id": "2401.10166",
            "authors": [
                "Yue Liu",
                "Yunjie Tian",
                "Yuzhong Zhao",
                "Hongtian Yu",
                "Lingxi Xie",
                "Yaowei Wang",
                "Qixiang Ye",
                "Yunfan Liu"
            ],
            "abstract": "Designing computationally efficient network architectures persists as an ongoing necessity in computer vision. In this paper, we transplant Mamba, a state-space language model, into VMamba, a vision backbone that works in linear time complexity. At the core of VMamba lies a stack of Visual State-Space (VSS) blocks with the 2D Selective Scan (SS2D) module. By traversing along four scanning routes, SS2D helps bridge the gap between the ordered nature of 1D selective scan and the non-sequential structure of 2D vision data, which facilitates the gathering of contextual information from various sources and perspectives. Based on the VSS blocks, we develop a family of VMamba architectures and accelerate them through a succession of architectural and implementation enhancements. Extensive experiments showcase VMamba's promising performance across diverse visual perception tasks, highlighting its advantages in input scaling efficiency compared to existing benchmark models. Source code is available at https://github.com/MzeroMiko/VMamba.",
            "introduction": "   1 Introduction  Visual representation learning stands as a fundamental research area in computer vision, which has witnessed remarkable progress in the era of deep learning. To represent complex patterns in vision data, two primary categories of backbone networks, i.e., Convolution Neural Networks (CNNs)\u00a0[49, 28, 30, 54, 38] and Vision Transformers (ViTs)\u00a0[13, 37, 58, 68], have been proposed and extensively utilized in a variety of visual tasks. Compared to CNNs, ViTs generally demonstrate superior learning capabilities on large-scale data due to the integration of the self-attention mechanism\u00a0[59, 13]. However, the quadratic complexity of self-attention w.r.t. the number of tokens introduces substantial computational overhead in downstream tasks involving large spatial resolutions.   To tackle this challenge, considerable efforts have been made to enhance the efficiency of attention computation\u00a0[55, 37, 12]. However, existing approaches either impose limitations on the size of the effective receptive field\u00a0[37] or experience evident performance degradation across diverse tasks\u00a0[31, 62]. This motivates us to develop a novel architecture for vision data, preserving the inherent advantages of the vanilla self-attention mechanism, i.e., global receptive fields and dynamic weighting parameters\u00a0[23].   Recently, Mamba\u00a0[17], a novel State Space Model (SSM)\u00a0[17, 44, 61] in the field natural language processing (NLP), has emerged as a highly promising approach for long sequence modeling with linear complexity. Drawing inspiration from this advancement, we introduce VMamba, a vision backbone integrating SSM-based blocks to facilitate efficient visual representation learning. However, the core algorithm of Mamba, i.e., the parallelized selective scan operation, is essentially designed for processing one-dimensional sequential data. This poses a challenge when attempting to adapt it for processing vision data, which inherently lacks a sequential arrangement of visual components. To address this issue, we propose 2D Selective Scan (SS2D), a four-way scanning mechanism tailored for spatial domain traversal. In contrast to the self-attention mechanism (Figure\u00a01 (a)), SS2D ensures that each image patch gains contextual knowledge exclusively through a compressed hidden state computed along the corresponding scanning path (Figure\u00a01 (b)), thereby reducing the computational complexity from quadratic to linear.   Figure 1: Comparison of correlation establishment between image patches via (a) self-attention and (b) the proposed 2D-Selective-Scan (SS2D). Red boxes indicate the query image patch, with patch opacity representing the degree of information loss.    Upon the VSS blocks, we develop a family of VMamba architectures (i.e., VMamba-Tiny/Small/Base) and accelerate them through a series of architectural enhancements and implementation optimizations. Compared to benchmark vision models built on CNN (ConvNeXt\u00a0[38]), ViT (Swin\u00a0[37] and HiViT\u00a0[68]), and SSM (S4ND\u00a0[45] and Vim\u00a0[71]), VMamba consistently achieves superior image classification accuracy on ImageNet-1K\u00a0[9] across model scales. Specifically, VMamba-Base achieves a top-1 accuracy of 83.9%percent83.983.9\\%83.9 %, surpassing Swin by +0.4%percent0.4+0.4\\%+ 0.4 %, with a throughput exceeding that of Swin by a substantial margin over 40%percent4040\\%40 % (646646646646 vs. 458458458458). The superiority of VMamba extends across various downstream tasks, with VMamba-Tiny/Small/Base achieving 47.3%percent47.347.3\\%47.3 %/48.7%percent48.748.7\\%48.7 %/49.2%percent49.249.2\\%49.2 % mAP in object detection on COCO\u00a0[34] (1\u00d71\\times1 \u00d7 training schedule). This outperforms Swin by 4.6%percent4.64.6\\%4.6 %/3.9%percent3.93.9\\%3.9 %/2.3%percent2.32.3\\%2.3 % and ConvNeXt by 3.1%percent3.13.1\\%3.1 %/3.3%percent3.33.3\\%3.3 %/2.2%percent2.22.2\\%2.2 %, respectively. As for single-scale semantic segmentation on ADE20K\u00a0[70], VMamba-Tiny/Small/Base achieves 47.9%percent47.947.9\\%47.9 %/50.6%percent50.650.6\\%50.6 %/51.0%percent51.051.0\\%51.0 % mIoU, which surpasses Swin by 3.4%percent3.43.4\\%3.4 %/3.0%percent3.03.0\\%3.0 %/2.9%percent2.92.9\\%2.9 % and ConvNeXt by 1.9%percent1.91.9\\%1.9 %/1.9%percent1.91.9\\%1.9 %/1.9%percent1.91.9\\%1.9 %, respectively. Furthermore, unlike ViT-based models, which experience quadratic",
            "references": [
                {
                    "title": "Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model",
                    "abstract": "Recently the state space models (SSMs) with efficient hardware-aware designs, i.e., the Mamba deep learning model, have shown great potential for long sequence modeling. Meanwhile building efficient and generic vision backbones purely upon SSMs is an appealing direction. However, representing visual data is challenging for SSMs due to the position-sensitivity of visual data and the requirement of global context for visual understanding. In this paper, we show that the reliance on self-attention for visual representation learning is not necessary and propose a new generic vision backbone with bidirectional Mamba blocks (Vim), which marks the image sequences with position embeddings and compresses the visual representation with bidirectional state space models. On ImageNet classification, COCO object detection, and ADE20k semantic segmentation tasks, Vim achieves higher performance compared to well-established vision transformers like DeiT, while also demonstrating significantly improved computation&memory efficiency. For example, Vim is 2.8$\\times$ faster than DeiT and saves 86.8% GPU memory when performing batch inference to extract features on images with a resolution of 1248$\\times$1248. The results demonstrate that Vim is capable of overcoming the computation&memory constraints on performing Transformer-style understanding for high-resolution images and it has great potential to be the next-generation backbone for vision foundation models. Code is available at https://github.com/hustvl/Vim."
                },
                {
                    "title": "Gated Linear Attention Transformers with Hardware-Efficient Training",
                    "abstract": "Transformers with linear attention allow for efficient parallel training but can simultaneously be formulated as an RNN with 2D (matrix-valued) hidden states, thus enjoying linear-time inference complexity. However, linear attention generally underperforms ordinary softmax attention. Moreover, current implementations of linear attention lack I/O-awareness and are thus slower than highly optimized implementations of softmax attention. This work describes a hardware-efficient algorithm for linear attention that trades off memory movement against parallelizability. The resulting implementation, dubbed FLASHLINEARATTENTION, is faster than FLASHATTENTION-2 (Dao, 2023) as a standalone layer even on short sequence lengths (e.g., 1K). We then generalize this algorithm to a more expressive variant of linear attention with data-dependent gates. When used as a replacement for the standard attention layer in Transformers, the resulting gated linear attention (GLA) Transformer is found to perform competitively against the LLaMA-architecture Transformer (Touvron et al., 2023) as well recent linear-time-inference baselines such as RetNet (Sun et al., 2023a) and Mamba (Gu&Dao, 2023) on moderate-scale language modeling experiments. GLA Transformer is especially effective at length generalization, enabling a model trained on 2K to generalize to sequences longer than 20K without significant perplexity degradations. For training speed, the GLA Transformer has higher throughput than a similarly-sized Mamba model."
                },
                {
                    "title": "Mamba: Linear-Time Sequence Modeling with Selective State Spaces",
                    "abstract": "Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5$\\times$ higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation."
                },
                {
                    "title": "RMT: Retentive Networks Meet Vision Transformers",
                    "abstract": "Vision Transformer (ViT) has gained increasing attention in the computer vision community in recent years. How-ever, the core component of ViT, Self-Attention, lacks ex-plicit spatial priors and bears a quadratic computational complexity, thereby constraining the applicability of ViT. To alleviate these issues, we draw inspiration from the re-cent Retentive Network (RetNet) in the field of NLP, and propose RMT, a strong vision backbone with explicit spa-tial prior for general purposes. Specifically, we extend the RetNet's temporal decay mechanism to the spatial do-main, and propose a spatial decay matrix based on the Manhattan distance to introduce the explicit spatial prior to Self-Attention. Additionally, an attention decomposition form that adeptly adapts to explicit spatial prior is proposed, aiming to reduce the computational burden of modeling global information without disrupting the spa-tial decay matrix. Based on the spatial decay matrix and the attention decomposition form, we can flexibly integrate explicit spatial prior into the vision backbone with lin-ear complexity. Extensive experiments demonstrate that RMT exhibits exceptional performance across various vision tasks. Specifically, without extra training data, RMT achieves 84.8% and 86.1% top-l acc on ImageNet-lk with 27MI4.5GFLOPs and 96M/18.2GFLOPs. For downstream tasks, RMT achieves 54.5 box AP and 47.2 mask AP on the COCO detection task, and 52.8 mloU on the ADE20K se-mantic segmentation task."
                },
                {
                    "title": "FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning",
                    "abstract": "Scaling Transformers to longer sequence lengths has been a major problem in the last several years, promising to improve performance in language modeling and high-resolution image understanding, as well as to unlock new applications in code, audio, and video generation. The attention layer is the main bottleneck in scaling to longer sequences, as its runtime and memory increase quadratically in the sequence length. FlashAttention exploits the asymmetric GPU memory hierarchy to bring significant memory saving (linear instead of quadratic) and runtime speedup (2-4$\\times$ compared to optimized baselines), with no approximation. However, FlashAttention is still not nearly as fast as optimized matrix-multiply (GEMM) operations, reaching only 25-40\\% of the theoretical maximum FLOPs/s. We observe that the inefficiency is due to suboptimal work partitioning between different thread blocks and warps on the GPU, causing either low-occupancy or unnecessary shared memory reads/writes. We propose FlashAttention-2, with better work partitioning to address these issues. In particular, we (1) tweak the algorithm to reduce the number of non-matmul FLOPs (2) parallelize the attention computation, even for a single head, across different thread blocks to increase occupancy, and (3) within each thread block, distribute the work between warps to reduce communication through shared memory. These yield around 2$\\times$ speedup compared to FlashAttention, reaching 50-73\\% of the theoretical maximum FLOPs/s on A100 and getting close to the efficiency of GEMM operations. We empirically validate that when used end-to-end to train GPT-style models, FlashAttention-2 reaches training speed of up to 225 TFLOPs/s per A100 GPU (72\\% model FLOPs utilization)."
                },
                {
                    "title": "Retentive Network: A Successor to Transformer for Large Language Models",
                    "abstract": "In this work, we propose Retentive Network (RetNet) as a foundation architecture for large language models, simultaneously achieving training parallelism, low-cost inference, and good performance. We theoretically derive the connection between recurrence and attention. Then we propose the retention mechanism for sequence modeling, which supports three computation paradigms, i.e., parallel, recurrent, and chunkwise recurrent. Specifically, the parallel representation allows for training parallelism. The recurrent representation enables low-cost $O(1)$ inference, which improves decoding throughput, latency, and GPU memory without sacrificing performance. The chunkwise recurrent representation facilitates efficient long-sequence modeling with linear complexity, where each chunk is encoded parallelly while recurrently summarizing the chunks. Experimental results on language modeling show that RetNet achieves favorable scaling results, parallel training, low-cost deployment, and efficient inference. The intriguing properties make RetNet a strong successor to Transformer for large language models. Code will be available at https://aka.ms/retnet."
                },
                {
                    "title": "Integrally Pre-Trained Transformer Pyramid Networks",
                    "abstract": "In this paper, we present an integral pre-training framework based on masked image modeling (MIM). We advocate for pre-training the backbone and neck jointly so that the transfer gap between MIM and downstream recognition tasks is minimal. We make two technical contributions. First, we unify the reconstruction and recognition necks by inserting a feature pyramid into the pre-training stage. Second, we complement mask image modeling (MIM) with masked feature modeling (MFM) that offers multi-stage supervision to the feature pyramid. The pre-trained models, termed integrally pre-trained transformer pyramid networks (iTPNs), serve as powerful foundation models for visual recognition. In particular, the base/large-level iTPN achieves an 86.2%/87.8% top-1 accuracy on ImageNet-1K, a 53.2%/55.6% box AP on COCO object detection with 1\u00d7 training schedule using Mask-RCNN, and a 54.7%/57.7% mIoU on ADE20K semantic segmentation using UPerHead \u2013 all these results set new records. Our work inspires the community to work on unifying upstream pre-training and downstream fine-tuning tasks. Code is available at github.com/sunsmarterjie/iTPN."
                },
                {
                    "title": "RWKV: Reinventing RNNs for the Transformer Era",
                    "abstract": "Transformers have revolutionized almost all natural language processing (NLP) tasks but suffer from memory and computational complexity that scales quadratically with sequence length. In contrast, recurrent neural networks (RNNs) exhibit linear scaling in memory and computational requirements but struggle to match the same performance as Transformers due to limitations in parallelization and scalability. We propose a novel model architecture, Receptance Weighted Key Value (RWKV), that combines the efficient parallelizable training of transformers with the efficient inference of RNNs. Our approach leverages a linear attention mechanism and allows us to formulate the model as either a Transformer or an RNN, thus parallelizing computations during training and maintains constant computational and memory complexity during inference. We scale our models as large as 14 billion parameters, by far the largest dense RNN ever trained, and find RWKV performs on par with similarly sized Transformers, suggesting future work can leverage this architecture to create more efficient models. This work presents a significant step towards reconciling trade-offs between computational efficiency and model performance in sequence processing tasks."
                },
                {
                    "title": "Selective Structured State-Spaces for Long-Form Video Understanding",
                    "abstract": "Effective modeling of complex spatiotemporal dependencies in long-form videos remains an open problem. The recently proposed Structured State-Space Sequence ($S4$) model with its linear complexity offers a promising direction in this space. However, we demonstrate that treating all imagetokens equally as done by $S4$ model can adversely affect its efficiency and accuracy. To address this limitation, we present a novel Selective $S4$ (i.e., $S5)$ model that employs a lightweight mask generator to adaptively select informative image tokens resulting in more efficient and accurate modeling of long-term spatiotemporal dependencies in videos. Unlike previous mask-based token reduction methods used in transformers, our $S5$ model avoids the dense self-attention calculation by making use of the guidance of the momentum-updated $S4$ model. This enables our model to efficiently discard less informative tokens and adapt to various long-form video understanding tasks more effectively. However, as is the case for most token reduction methods, the informative image tokens could be dropped incorrectly. To improve the robustness and the temporal horizon of our model, we propose a novel long-short masked contrastive learning (LSMCL) approach that enables our model to predict longer temporal context using shorter input videos. We present extensive comparative results using three challenging long-form video understanding datasets (LVU, COIN and Breakfast), demonstrating that our approach consistently outperforms the previous state-of-the-art S4 model by up to 9.6% accuracy while reducing its memory footprint by 23%."
                },
                {
                    "title": "Hungry Hungry Hippos: Towards Language Modeling with State Space Models",
                    "abstract": "State space models (SSMs) have demonstrated state-of-the-art sequence modeling performance in some modalities, but underperform attention in language modeling. Moreover, despite scaling nearly linearly in sequence length instead of quadratically, SSMs are still slower than Transformers due to poor hardware utilization. In this paper, we make progress on understanding the expressivity gap between SSMs and attention in language modeling, and on reducing the hardware barrier between SSMs and attention. First, we use synthetic language modeling tasks to understand the gap between SSMs and attention. We find that existing SSMs struggle with two capabilities: recalling earlier tokens in the sequence and comparing tokens across the sequence. To understand the impact on language modeling, we propose a new SSM layer, H3, that is explicitly designed for these abilities. H3 matches attention on the synthetic languages and comes within 0.4 PPL of Transformers on OpenWebText. Furthermore, a hybrid 125M-parameter H3-attention model that retains two attention layers surprisingly outperforms Transformers on OpenWebText by 1.0 PPL. Next, to improve the efficiency of training SSMs on modern hardware, we propose FlashConv. FlashConv uses a fused block FFT algorithm to improve efficiency on sequences up to 8K, and introduces a novel state passing algorithm that exploits the recurrent properties of SSMs to scale to longer sequences. FlashConv yields 2$\\times$ speedup on the long-range arena benchmark and allows hybrid language models to generate text 2.4$\\times$ faster than Transformers. Using FlashConv, we scale hybrid H3-attention language models up to 2.7B parameters on the Pile and find promising initial results, achieving lower perplexity than Transformers and outperforming Transformers in zero- and few-shot learning on a majority of tasks in the SuperGLUE benchmark."
                },
                {
                    "title": "Liquid Structural State-Space Models",
                    "abstract": "A proper parametrization of state transition matrices of linear state-space models (SSMs) followed by standard nonlinearities enables them to efficiently learn representations from sequential data, establishing the state-of-the-art on a large series of long-range sequence modeling benchmarks. In this paper, we show that we can improve further when the structural SSM such as S4 is given by a linear liquid time-constant (LTC) state-space model. LTC neural networks are causal continuous-time neural networks with an input-dependent state transition module, which makes them learn to adapt to incoming inputs at inference. We show that by using a diagonal plus low-rank decomposition of the state transition matrix introduced in S4, and a few simplifications, the LTC-based structural state-space model, dubbed Liquid-S4, achieves the new state-of-the-art generalization across sequence modeling tasks with long-term dependencies such as image, text, audio, and medical time-series, with an average performance of 87.32% on the Long-Range Arena benchmark. On the full raw Speech Command recognition, dataset Liquid-S4 achieves 96.78% accuracy with a 30% reduction in parameter counts compared to S4. The additional gain in performance is the direct result of the Liquid-S4's kernel structure that takes into account the similarities of the input sequence samples during training and inference."
                },
                {
                    "title": "Mega: Moving Average Equipped Gated Attention",
                    "abstract": "The design choices in the Transformer attention mechanism, including weak inductive bias and quadratic computational complexity, have limited its application for modeling long sequences. In this paper, we introduce Mega, a simple, theoretically grounded, single-head gated attention mechanism equipped with (exponential) moving average to incorporate inductive bias of position-aware local dependencies into the position-agnostic attention mechanism. We further propose a variant of Mega that offers linear time and space complexity yet yields only minimal quality loss, by efficiently splitting the whole sequence into multiple chunks with fixed length. Extensive experiments on a wide range of sequence modeling benchmarks, including the Long Range Arena, neural machine translation, auto-regressive language modeling, and image and speech classification, show that Mega achieves significant improvements over other sequence models, including variants of Transformers and recent state space models."
                },
                {
                    "title": "Simplified State Space Layers for Sequence Modeling",
                    "abstract": "Models using structured state space sequence (S4) layers have achieved state-of-the-art performance on long-range sequence modeling tasks. An S4 layer combines linear state space models (SSMs), the HiPPO framework, and deep learning to achieve high performance. We build on the design of the S4 layer and introduce a new state space layer, the S5 layer. Whereas an S4 layer uses many independent single-input, single-output SSMs, the S5 layer uses one multi-input, multi-output SSM. We establish a connection between S5 and S4, and use this to develop the initialization and parameterization used by the S5 model. The result is a state space layer that can leverage efficient and widely implemented parallel scans, allowing S5 to match the computational efficiency of S4, while also achieving state-of-the-art performance on several long-range sequence modeling tasks. S5 averages 87.4% on the long range arena benchmark, and 98.5% on the most difficult Path-X task."
                },
                {
                    "title": "HorNet: Efficient High-Order Spatial Interactions with Recursive Gated Convolutions",
                    "abstract": "Recent progress in vision Transformers exhibits great success in various tasks driven by the new spatial modeling mechanism based on dot-product self-attention. In this paper, we show that the key ingredients behind the vision Transformers, namely input-adaptive, long-range and high-order spatial interactions, can also be efficiently implemented with a convolution-based framework. We present the Recursive Gated Convolution ($\\textit{g}^\\textit{n}$Conv) that performs high-order spatial interactions with gated convolutions and recursive designs. The new operation is highly flexible and customizable, which is compatible with various variants of convolution and extends the two-order interactions in self-attention to arbitrary orders without introducing significant extra computation. $\\textit{g}^\\textit{n}$Conv can serve as a plug-and-play module to improve various vision Transformers and convolution-based models. Based on the operation, we construct a new family of generic vision backbones named HorNet. Extensive experiments on ImageNet classification, COCO object detection and ADE20K semantic segmentation show HorNet outperform Swin Transformers and ConvNeXt by a significant margin with similar overall architecture and training configurations. HorNet also shows favorable scalability to more training data and larger model sizes. Apart from the effectiveness in visual encoders, we also show $\\textit{g}^\\textit{n}$Conv can be applied to task-specific decoders and consistently improve dense prediction performance with less computation. Our results demonstrate that $\\textit{g}^\\textit{n}$Conv can be a new basic module for visual modeling that effectively combines the merits of both vision Transformers and CNNs. Code is available at https://github.com/raoyongming/HorNet"
                },
                {
                    "title": "More ConvNets in the 2020s: Scaling up Kernels Beyond 51x51 using Sparsity",
                    "abstract": "Transformers have quickly shined in the computer vision world since the emergence of Vision Transformers (ViTs). The dominant role of convolutional neural networks (CNNs) seems to be challenged by increasingly effective transformer-based models. Very recently, a couple of advanced convolutional models strike back with large kernels motivated by the local-window attention mechanism, showing appealing performance and efficiency. While one of them, i.e. RepLKNet, impressively manages to scale the kernel size to 31x31 with improved performance, the performance starts to saturate as the kernel size continues growing, compared to the scaling trend of advanced ViTs such as Swin Transformer. In this paper, we explore the possibility of training extreme convolutions larger than 31x31 and test whether the performance gap can be eliminated by strategically enlarging convolutions. This study ends up with a recipe for applying extremely large kernels from the perspective of sparsity, which can smoothly scale up kernels to 61x61 with better performance. Built on this recipe, we propose Sparse Large Kernel Network (SLaK), a pure CNN architecture equipped with sparse factorized 51x51 kernels that can perform on par with or better than state-of-the-art hierarchical Transformers and modern ConvNet architectures like ConvNeXt and RepLKNet, on ImageNet classification as well as a wide range of downstream tasks including semantic segmentation on ADE20K, object detection on PASCAL VOC 2007, and object detection/segmentation on MS COCO."
                },
                {
                    "title": "Long Range Language Modeling via Gated State Spaces",
                    "abstract": "State space models have shown to be effective at modeling long range dependencies, specially on sequence classification tasks. In this work we focus on autoregressive sequence modeling over English books, Github source code and ArXiv mathematics articles. Based on recent developments around the effectiveness of gated activation functions, we propose a new layer named Gated State Space (GSS) and show that it trains significantly faster than the diagonal version of S4 (i.e. DSS) on TPUs, is fairly competitive with several well-tuned Transformer-based baselines and exhibits zero-shot generalization to longer inputs while being straightforward to implement. Finally, we show that leveraging self-attention to model local dependencies improves the performance of GSS even further."
                },
                {
                    "title": "On the Parameterization and Initialization of Diagonal State Space Models",
                    "abstract": "State space models (SSM) have recently been shown to be very effective as a deep learning layer as a promising alternative to sequence models such as RNNs, CNNs, or Transformers. The first version to show this potential was the S4 model, which is particularly effective on tasks involving long-range dependencies by using a prescribed state matrix called the HiPPO matrix. While this has an interpretable mathematical mechanism for modeling long dependencies, it introduces a custom representation and algorithm that can be difficult to implement. On the other hand, a recent variant of S4 called DSS showed that restricting the state matrix to be fully diagonal can still preserve the performance of the original model when using a specific initialization based on approximating S4's matrix. This work seeks to systematically understand how to parameterize and initialize such diagonal state space models. While it follows from classical results that almost all SSMs have an equivalent diagonal form, we show that the initialization is critical for performance. We explain why DSS works mathematically, by showing that the diagonal restriction of S4's matrix surprisingly recovers the same kernel in the limit of infinite state dimension. We also systematically describe various design choices in parameterizing and computing diagonal SSMs, and perform a controlled empirical study ablating the effects of these choices. Our final model S4D is a simple diagonal version of S4 whose kernel computation requires just 2 lines of code and performs comparably to S4 in almost all settings, with state-of-the-art results for image, audio, and medical time-series domains, and averaging 85\\% on the Long Range Arena benchmark."
                },
                {
                    "title": "GraFormer: Graph-oriented Transformer for 3D Pose Estimation",
                    "abstract": "In 2D-to-3D pose estimation, it is important to exploit the spatial constraints of 2D joints, but it is not yet well modeled. To better model the relation of joints for 3D pose estimation, we propose an effective but simple net-work, called Graliormer11Codes:https://girhub.com/zhoawexi/GraFormer, where a novel transformer architecture is designed via embedding graph convolution layers after multi-head attention block. The proposed GraFormer is built by repeatedly stacking the GraAttention block and the ChebGConv block. The proposed GraAttention block is a new transformer block designed for processing graph-structured data, which is able to learn better features through capturing global information from all the nodes as well as the explicit adjacency structure of nodes. To model the implicit high-order connection relations among non-neighboring nodes, the ChebGConv block is introduced to exchange information between non-neighboring nodes and attain a larger receptive field. We have empirically shown the superiority of GraFormer through extensive experiments on popular public datasets. Specifically, GraFormer outperforms the state-of-the-art GraghSH [38] on the Hu-man3.6M dataset yet only contains 18% parameters of it."
                },
                {
                    "title": "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness",
                    "abstract": "Transformers are slow and memory-hungry on long sequences, since the time and memory complexity of self-attention are quadratic in sequence length. Approximate attention methods have attempted to address this problem by trading off model quality to reduce the compute complexity, but often do not achieve wall-clock speedup. We argue that a missing principle is making attention algorithms IO-aware -- accounting for reads and writes between levels of GPU memory. We propose FlashAttention, an IO-aware exact attention algorithm that uses tiling to reduce the number of memory reads/writes between GPU high bandwidth memory (HBM) and GPU on-chip SRAM. We analyze the IO complexity of FlashAttention, showing that it requires fewer HBM accesses than standard attention, and is optimal for a range of SRAM sizes. We also extend FlashAttention to block-sparse attention, yielding an approximate attention algorithm that is faster than any existing approximate attention method. FlashAttention trains Transformers faster than existing baselines: 15% end-to-end wall-clock speedup on BERT-large (seq. length 512) compared to the MLPerf 1.1 training speed record, 3$\\times$ speedup on GPT-2 (seq. length 1K), and 2.4$\\times$ speedup on long-range arena (seq. length 1K-4K). FlashAttention and block-sparse FlashAttention enable longer context in Transformers, yielding higher quality models (0.7 better perplexity on GPT-2 and 6.4 points of lift on long-document classification) and entirely new capabilities: the first Transformers to achieve better-than-chance performance on the Path-X challenge (seq. length 16K, 61.4% accuracy) and Path-256 (seq. length 64K, 63.1% accuracy)."
                },
                {
                    "title": "Neighborhood Attention Transformer",
                    "abstract": "We present Neighborhood Attention (NA), the first efficient and scalable sliding window attention mechanism for vision. NA is a pixel-wise operation, localizing self attention (SA) to the nearest neighboring pixels, and therefore enjoys a linear time and space complexity compared to the quadratic complexity of SA. The sliding window pattern allows NA's receptive field to grow without needing extra pixel shifts, and preserves translational equivariance, unlike Swin Transformer's Window Self Attention (WSA). We develop NATTEN (Neighborhood Attention Extension), a Python package with efficient C++ and CUDA kernels, which allows NA to run up to 40% faster than Swin's WSA while using up to 25% less memory. We further present Neighborhood Attention Transformer (NAT), a new hierarchical transformer design based on NA that boosts image classification and downstream vision performance. Experimental results on NAT are competitive; NAT-Tiny reaches 83.2% top-1 accuracy on ImageNet, 51.4% mAP on MSCOCO and 48.4% mIoU on ADE20K, which is 1.9% ImageNet accuracy, 1.0% COCO mAP, and 2.6% ADE20K mIoU improvement over a Swin model with similar size. To support more research based on sliding window attention, we open source our project and release our checkpoints."
                },
                {
                    "title": "DaViT: Dual Attention Vision Transformers",
                    "abstract": "In this work, we introduce Dual Attention Vision Transformers (DaViT), a simple yet effective vision transformer architecture that is able to capture global context while maintaining computational efficiency. We propose approaching the problem from an orthogonal angle: exploiting self-attention mechanisms with both\"spatial tokens\"and\"channel tokens\". With spatial tokens, the spatial dimension defines the token scope, and the channel dimension defines the token feature dimension. With channel tokens, we have the inverse: the channel dimension defines the token scope, and the spatial dimension defines the token feature dimension. We further group tokens along the sequence direction for both spatial and channel tokens to maintain the linear complexity of the entire model. We show that these two self-attentions complement each other: (i) since each channel token contains an abstract representation of the entire image, the channel attention naturally captures global interactions and representations by taking all spatial positions into account when computing attention scores between channels; (ii) the spatial attention refines the local representations by performing fine-grained interactions across spatial locations, which in turn helps the global information modeling in channel attention. Extensive experiments show our DaViT achieves state-of-the-art performance on four different tasks with efficient computations. Without extra data, DaViT-Tiny, DaViT-Small, and DaViT-Base achieve 82.8%, 84.2%, and 84.6% top-1 accuracy on ImageNet-1K with 28.3M, 49.7M, and 87.9M parameters, respectively. When we further scale up DaViT with 1.5B weakly supervised image and text pairs, DaViT-Gaint reaches 90.4% top-1 accuracy on ImageNet-1K. Code is available at https://github.com/dingmyu/davit."
                },
                {
                    "title": "Diagonal State Spaces are as Effective as Structured State Spaces",
                    "abstract": "Modeling long range dependencies in sequential data is a fundamental step towards attaining human-level performance in many modalities such as text, vision, audio and video. While attention-based models are a popular and effective choice in modeling short-range interactions, their performance on tasks requiring long range reasoning has been largely inadequate. In an exciting result, Gu et al. (ICLR 2022) proposed the $\\textit{Structured State Space}$ (S4) architecture delivering large gains over state-of-the-art models on several long-range tasks across various modalities. The core proposition of S4 is the parameterization of state matrices via a diagonal plus low rank structure, allowing efficient computation. In this work, we show that one can match the performance of S4 even without the low rank correction and thus assuming the state matrices to be diagonal. Our $\\textit{Diagonal State Space}$ (DSS) model matches the performance of S4 on Long Range Arena tasks, speech classification on Speech Commands dataset, while being conceptually simpler and straightforward to implement."
                },
                {
                    "title": "A ConvNet for the 2020s",
                    "abstract": "The \u201cRoaring 20s\u201d of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually \u201cmodernize\u201d a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets."
                },
                {
                    "title": "Swin Transformer V2: Scaling Up Capacity and Resolution",
                    "abstract": "We present techniques for scaling Swin Transformer [35] up to 3 billion parameters and making it capable of training with images of up to 1,536x1,536 resolution. By scaling up capacity and resolution, Swin Transformer sets new records on four representative vision benchmarks: 84.0% top-1 accuracy on ImageNet- V2 image classification, 63.1 / 54.4 box / mask mAP on COCO object detection, 59.9 mIoU on ADE20K semantic segmentation, and 86.8% top-1 accuracy on Kinetics-400 video action classification. We tackle issues of training instability, and study how to effectively transfer models pre-trained at low resolutions to higher resolution ones. To this aim, several novel technologies are proposed: 1) a residual post normalization technique and a scaled cosine attention approach to improve the stability of large vision models; 2) a log-spaced continuous position bias technique to effectively transfer models pre-trained at low-resolution images and windows to their higher-resolution counterparts. In addition, we share our crucial implementation details that lead to significant savings of GPU memory consumption and thus make it feasi-ble to train large vision models with regular GPUs. Using these techniques and self-supervised pre-training, we suc-cessfully train a strong 3 billion Swin Transformer model and effectively transfer it to various vision tasks involving high-resolution images or windows, achieving the state-of-the-art accuracy on a variety of benchmarks. Code is avail-able at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "Efficiently Modeling Long Sequences with Structured State Spaces",
                    "abstract": "A central goal of sequence modeling is designing a single principled model that can address sequence data across a range of modalities and tasks, particularly on long-range dependencies. Although conventional models including RNNs, CNNs, and Transformers have specialized variants for capturing long dependencies, they still struggle to scale to very long sequences of $10000$ or more steps. A promising recent approach proposed modeling sequences by simulating the fundamental state space model (SSM) \\( x'(t) = Ax(t) + Bu(t), y(t) = Cx(t) + Du(t) \\), and showed that for appropriate choices of the state matrix \\( A \\), this system could handle long-range dependencies mathematically and empirically. However, this method has prohibitive computation and memory requirements, rendering it infeasible as a general sequence modeling solution. We propose the Structured State Space sequence model (S4) based on a new parameterization for the SSM, and show that it can be computed much more efficiently than prior approaches while preserving their theoretical strengths. Our technique involves conditioning \\( A \\) with a low-rank correction, allowing it to be diagonalized stably and reducing the SSM to the well-studied computation of a Cauchy kernel. S4 achieves strong empirical results across a diverse range of established benchmarks, including (i) 91\\% accuracy on sequential CIFAR-10 with no data augmentation or auxiliary losses, on par with a larger 2-D ResNet, (ii) substantially closing the gap to Transformers on image and language modeling tasks, while performing generation $60\\times$ faster (iii) SoTA on every task from the Long Range Arena benchmark, including solving the challenging Path-X task of length 16k that all prior work fails on, while being as efficient as all competitors."
                },
                {
                    "title": "Combining Recurrent, Convolutional, and Continuous-time Models with Linear State-Space Layers",
                    "abstract": "Recurrent neural networks (RNNs), temporal convolutions, and neural differential equations (NDEs) are popular families of deep learning models for time-series data, each with unique strengths and tradeoffs in modeling power and computational efficiency. We introduce a simple sequence model inspired by control systems that generalizes these approaches while addressing their shortcomings. The Linear State-Space Layer (LSSL) maps a sequence $u \\mapsto y$ by simply simulating a linear continuous-time state-space representation $\\dot{x} = Ax + Bu, y = Cx + Du$. Theoretically, we show that LSSL models are closely related to the three aforementioned families of models and inherit their strengths. For example, they generalize convolutions to continuous-time, explain common RNN heuristics, and share features of NDEs such as time-scale adaptation. We then incorporate and generalize recent theory on continuous-time memorization to introduce a trainable subset of structured matrices $A$ that endow LSSLs with long-range memory. Empirically, stacking LSSL layers into a simple deep neural network obtains state-of-the-art results across time series benchmarks for long dependencies in sequential image classification, real-world healthcare regression tasks, and speech. On a difficult speech classification task with length-16000 sequences, LSSL outperforms prior approaches by 24 accuracy points, and even outperforms baselines that use hand-crafted features on 100x shorter sequences."
                },
                {
                    "title": "CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows",
                    "abstract": "We present CSWin Transformer, an efficient and effective Transformer-based backbone for general-purpose vision tasks. A challenging issue in Transformer design is that global self-attention is very expensive to compute whereas local self-attention often limits the field of interactions of each token. To address this issue, we develop the Cross-Shaped Window self-attention mechanism for computing self-attention in the horizontal and vertical stripes in parallel that form a cross-shaped window, with each stripe obtained by splitting the input feature into stripes of equal width. We provide a mathematical analysis of the effect of the stripe width and vary the stripe width for different layers of the Transformer network which achieves strong modeling capability while limiting the computation cost. We also introduce Locally-enhanced Positional Encoding (LePE), which handles the local positional information better than existing encoding schemes. LePE naturally supports arbitrary input resolutions, and is thus especially effective and friendly for downstream tasks. Incorporated with these designs and a hierarchical structure, CSWin Transformer demonstrates competitive performance on common vision tasks. Specifically, it achieves 85.4% Top-1 accuracy on ImageNet-1K without any extra training data or label, 53.9 box AP and 46.4 mask AP on the COCO detection task, and 52.2 mIOU on the ADE20K semantic segmentation task, surpassing previous state-of-the-art Swin Transformer backbone by +1.2, +2.0, +1.4, and +2.0 respectively under the similar FLOPs setting. By further pretraining on the larger dataset ImageNet-21K, we achieve 87.5% Top-1 accuracy on ImageNet-1K and high segmentation performance on ADE20K with 55.7 mIoU. 11Code and pretrain model is available at https://github.com/microsoft/CSWin-Transformer"
                },
                {
                    "title": "Focal Self-attention for Local-Global Interactions in Vision Transformers",
                    "abstract": "Recently, Vision Transformer and its variants have shown great promise on various computer vision tasks. The ability of capturing short- and long-range visual dependencies through self-attention is arguably the main source for the success. But it also brings challenges due to quadratic computational overhead, especially for the high-resolution vision tasks (e.g., object detection). In this paper, we present focal self-attention, a new mechanism that incorporates both fine-grained local and coarse-grained global interactions. Using this new mechanism, each token attends the closest surrounding tokens at fine granularity but the tokens far away at coarse granularity, and thus can capture both short- and long-range visual dependencies efficiently and effectively. With focal self-attention, we propose a new variant of Vision Transformer models, called Focal Transformer, which achieves superior performance over the state-of-the-art vision Transformers on a range of public image classification and object detection benchmarks. In particular, our Focal Transformer models with a moderate size of 51.1M and a larger size of 89.8M achieve 83.5 and 83.8 Top-1 accuracy, respectively, on ImageNet classification at 224x224 resolution. Using Focal Transformers as the backbones, we obtain consistent and substantial improvements over the current state-of-the-art Swin Transformers for 6 different object detection methods trained with standard 1x and 3x schedules. Our largest Focal Transformer yields 58.7/58.9 box mAPs and 50.9/51.3 mask mAPs on COCO mini-val/test-dev, and 55.4 mIoU on ADE20K for semantic segmentation, creating new SoTA on three of the most challenging computer vision tasks."
                },
                {
                    "title": "XCiT: Cross-Covariance Image Transformers",
                    "abstract": "Following their success in natural language processing, transformers have recently shown much promise for computer vision. The self-attention operation underlying transformers yields global interactions between all tokens ,i.e. words or image patches, and enables flexible modelling of image data beyond the local interactions of convolutions. This flexibility, however, comes with a quadratic complexity in time and memory, hindering application to long sequences and high-resolution images. We propose a\"transposed\"version of self-attention that operates across feature channels rather than tokens, where the interactions are based on the cross-covariance matrix between keys and queries. The resulting cross-covariance attention (XCA) has linear complexity in the number of tokens, and allows efficient processing of high-resolution images. Our cross-covariance image transformer (XCiT) is built upon XCA. It combines the accuracy of conventional transformers with the scalability of convolutional architectures. We validate the effectiveness and generality of XCiT by reporting excellent results on multiple vision benchmarks, including image classification and self-supervised feature learning on ImageNet-1k, object detection and instance segmentation on COCO, and semantic segmentation on ADE20k."
                },
                {
                    "title": "CoAtNet: Marrying Convolution and Attention for All Data Sizes",
                    "abstract": "Transformers have attracted increasing interests in computer vision, but they still fall behind state-of-the-art convolutional networks. In this work, we show that while Transformers tend to have larger model capacity, their generalization can be worse than convolutional networks due to the lack of the right inductive bias. To effectively combine the strengths from both architectures, we present CoAtNets(pronounced\"coat\"nets), a family of hybrid models built from two key insights: (1) depthwise Convolution and self-Attention can be naturally unified via simple relative attention; (2) vertically stacking convolution layers and attention layers in a principled way is surprisingly effective in improving generalization, capacity and efficiency. Experiments show that our CoAtNets achieve state-of-the-art performance under different resource constraints across various datasets: Without extra data, CoAtNet achieves 86.0% ImageNet top-1 accuracy; When pre-trained with 13M images from ImageNet-21K, our CoAtNet achieves 88.56% top-1 accuracy, matching ViT-huge pre-trained with 300M images from JFT-300M while using 23x less data; Notably, when we further scale up CoAtNet with JFT-3B, it achieves 90.88% top-1 accuracy on ImageNet, establishing a new state-of-the-art result."
                },
                {
                    "title": "On the Connection between Local Attention and Dynamic Depth-wise Convolution",
                    "abstract": "Vision Transformer (ViT) attains state-of-the-art performance in visual recognition, and the variant, Local Vision Transformer, makes further improvements. The major component in Local Vision Transformer, local attention, performs the attention separately over small local windows. We rephrase local attention as a channel-wise locally-connected layer and analyze it from two network regularization manners, sparse connectivity and weight sharing, as well as weight computation. Sparse connectivity: there is no connection across channels, and each position is connected to the positions within a small local window. Weight sharing: the connection weights for one position are shared across channels or within each group of channels. Dynamic weight: the connection weights are dynamically predicted according to each image instance. We point out that local attention resembles depth-wise convolution and its dynamic version in sparse connectivity. The main difference lies in weight sharing - depth-wise convolution shares connection weights (kernel weights) across spatial positions. We empirically observe that the models based on depth-wise convolution and the dynamic variant with lower computation complexity perform on-par with or sometimes slightly better than Swin Transformer, an instance of Local Vision Transformer, for ImageNet classification, COCO object detection and ADE semantic segmentation. These observations suggest that Local Vision Transformer takes advantage of two regularization forms and dynamic weight to increase the network capacity. Code is available at https://github.com/Atten4Vis/DemystifyLocalViT."
                },
                {
                    "title": "MLP-Mixer: An all-MLP Architecture for Vision",
                    "abstract": "Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e.\"mixing\"the per-location features), and one with MLPs applied across patches (i.e.\"mixing\"spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers."
                },
                {
                    "title": "CrossViT: Cross-Attention Multi-Scale Vision Transformer for Image Classification",
                    "abstract": "The recently developed vision transformer (ViT) has achieved promising results on image classification compared to convolutional neural networks. Inspired by this, in this paper, we study how to learn multi-scale feature representations in transformer models for image classification. To this end, we propose a dual-branch transformer to com-bine image patches (i.e., tokens in a transformer) of different sizes to produce stronger image features. Our approach processes small-patch and large-patch tokens with two separate branches of different computational complexity and these tokens are then fused purely by attention multiple times to complement each other. Furthermore, to reduce computation, we develop a simple yet effective token fusion module based on cross attention, which uses a single token for each branch as a query to exchange information with other branches. Our proposed cross-attention only requires linear time for both computational and memory complexity instead of quadratic time otherwise. Extensive experiments demonstrate that our approach performs better than or on par with several concurrent works on vision transformer, in addition to efficient CNN models. For example, on the ImageNet1K dataset, with some architectural changes, our approach outperforms the recent DeiT by a large margin of 2% with a small to moderate increase in FLOPs and model parameters. Our source codes and models are available at https://github.com/IBM/CrossViT."
                },
                {
                    "title": "Scaling Local Self-Attention for Parameter Efficient Visual Backbones",
                    "abstract": "Self-attention has the promise of improving computer vision systems due to parameter-independent scaling of receptive fields and content-dependent interactions, in contrast to parameter-dependent scaling and content-independent interactions of convolutions. Self-attention models have recently been shown to have encouraging improvements on accuracy-parameter trade-offs compared to baseline convolutional models such as ResNet-50. In this work, we develop self-attention models that can outperform not just the canonical baseline models, but even the high-performing convolutional models. We propose two extensions to self-attention that, in conjunction with a more efficient implementation of self-attention, improve the speed, memory usage, and accuracy of these models. We leverage these improvements to develop a new self-attention model family, HaloNets, which reach state-of-the-art accuracies on the parameter-limited setting of the ImageNet classification benchmark. In preliminary transfer learning experiments, we find that HaloNet models outperform much larger models and have better inference performance. On harder tasks such as object detection and instance segmentation, our simple local self-attention and convolutional hybrids show improvements over very strong baselines. These results mark another step in demonstrating the efficacy of self-attention models on settings traditionally dominated by convolutions. 1"
                },
                {
                    "title": "DeepViT: Towards Deeper Vision Transformer",
                    "abstract": "Vision transformers (ViTs) have been successfully applied in image classification tasks recently. In this paper, we show that, unlike convolution neural networks (CNNs)that can be improved by stacking more convolutional layers, the performance of ViTs saturate fast when scaled to be deeper. More specifically, we empirically observe that such scaling difficulty is caused by the attention collapse issue: as the transformer goes deeper, the attention maps gradually become similar and even much the same after certain layers. In other words, the feature maps tend to be identical in the top layers of deep ViT models. This fact demonstrates that in deeper layers of ViTs, the self-attention mechanism fails to learn effective concepts for representation learning and hinders the model from getting expected performance gain. Based on above observation, we propose a simple yet effective method, named Re-attention, to re-generate the attention maps to increase their diversity at different layers with negligible computation and memory cost. The pro-posed method makes it feasible to train deeper ViT models with consistent performance improvements via minor modification to existing ViT models. Notably, when training a deep ViT model with 32 transformer blocks, the Top-1 classification accuracy can be improved by 1.6% on ImageNet. Code is publicly available at https://github.com/zhoudaquan/dvit_repo."
                },
                {
                    "title": "Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions",
                    "abstract": "Although convolutional neural networks (CNNs) have achieved great success in computer vision, this work investigates a simpler, convolution-free backbone network use-fid for many dense prediction tasks. Unlike the recently-proposed Vision Transformer (ViT) that was designed for image classification specifically, we introduce the Pyramid Vision Transformer (PVT), which overcomes the difficulties of porting Transformer to various dense prediction tasks. PVT has several merits compared to current state of the arts. (1) Different from ViT that typically yields low-resolution outputs and incurs high computational and memory costs, PVT not only can be trained on dense partitions of an image to achieve high output resolution, which is important for dense prediction, but also uses a progressive shrinking pyramid to reduce the computations of large feature maps. (2) PVT inherits the advantages of both CNN and Transformer, making it a unified backbone for various vision tasks without convolutions, where it can be used as a direct replacement for CNN backbones. (3) We validate PVT through extensive experiments, showing that it boosts the performance of many downstream tasks, including object detection, instance and semantic segmentation. For example, with a comparable number of parameters, PVT+RetinaNet achieves 40.4 AP on the COCO dataset, surpassing ResNet50+RetinNet (36.3 AP) by 4.1 absolute AP (see Figure 2). We hope that PVT could, serre as an alternative and useful backbone for pixel-level predictions and facilitate future research."
                },
                {
                    "title": "Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNet",
                    "abstract": "Transformers, which are popular for language modeling, have been explored for solving vision tasks recently, e.g., the Vision Transformer (ViT) for image classification. The ViT model splits each image into a sequence of tokens with fixed length and then applies multiple Transformer layers to model their global relation for classification. However, ViT achieves inferior performance to CNNs when trained from scratch on a midsize dataset like ImageNet. We find it is because: 1) the simple tokenization of input images fails to model the important local structure such as edges and lines among neighboring pixels, leading to low training sample efficiency; 2) the redundant attention backbone design of ViT leads to limited feature richness for fixed computation budgets and limited training samples. To overcome such limitations, we propose a new Tokens-To-Token Vision Transformer (T2T-VTT), which incorporates 1) a layer-wise Tokens-to-Token (T2T) transformation to progressively structurize the image to tokens by recursively aggregating neighboring Tokens into one Token (Tokens-to-Token), such that local structure represented by surrounding tokens can be modeled and tokens length can be reduced; 2) an efficient backbone with a deep-narrow structure for vision transformer motivated by CNN architecture design after empirical study. Notably, T2T-ViT reduces the parameter count and MACs of vanilla ViT by half, while achieving more than 3.0% improvement when trained from scratch on ImageNet. It also outperforms ResNets and achieves comparable performance with MobileNets by directly training on ImageNet. For example, T2T-ViT with comparable size to ResNet50 (21.5M parameters) can achieve 83.3% top1 accuracy in image resolution 384x384 on ImageNet.1"
                },
                {
                    "title": "Bottleneck Transformers for Visual Recognition",
                    "abstract": "We present BoTNet, a conceptually simple yet powerful backbone architecture that incorporates self-attention for multiple computer vision tasks including image classification, object detection and instance segmentation. By just replacing the spatial convolutions with global self-attention in the final three bottleneck blocks of a ResNet and no other changes, our approach improves upon the baselines significantly on instance segmentation and object detection while also reducing the parameters, with minimal overhead in latency. Through the design of BoTNet, we also point out how ResNet bottleneck blocks with self-attention can be viewed as Transformer blocks. Without any bells and whistles, BoTNet achieves 44.4% Mask AP and 49.7% Box AP on the COCO Instance Segmentation benchmark using the Mask R-CNN framework; surpassing the previous best published single model and single scale results of ResNeSt [67] evaluated on the COCO validation set. Finally, we present a simple adaptation of the BoTNet design for image classification, resulting in models that achieve a strong performance of 84.7% top-1 accuracy on the ImageNet benchmark while being up to 1.64x faster in \"compute\"1 time than the popular EfficientNet models on TPU-v3 hardware. We hope our simple and effective approach will serve as a strong baseline for future research in self-attention models for vision.2"
                },
                {
                    "title": "Training data-efficient image transformers & distillation through attention",
                    "abstract": "Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption. In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data. More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Efficient Transformers: A Survey",
                    "abstract": "Transformer model architectures have garnered immense interest lately due to their effectiveness across a range of domains like language, vision, and reinforcement learning. In the field of natural language processing for example, Transformers have become an indispensable staple in the modern deep learning stack. Recently, a dizzying number of \u201cX-former\u201d models have been proposed\u2014Reformer, Linformer, Performer, Longformer, to name a few\u2014which improve upon the original Transformer architecture, many of which make improvements around computational and memory efficiency. With the aim of helping the avid researcher navigate this flurry, this article characterizes a large and thoughtful selection of recent efficiency-flavored \u201cX-former\u201d models, providing an organized and comprehensive overview of existing work and models across multiple domains."
                },
                {
                    "title": "HiPPO: Recurrent Memory with Optimal Polynomial Projections",
                    "abstract": "A central problem in learning from sequential data is representing cumulative history in an incremental fashion as more data is processed. We introduce a general framework (HiPPO) for the online compression of continuous signals and discrete time series by projection onto polynomial bases. Given a measure that specifies the importance of each time step in the past, HiPPO produces an optimal solution to a natural online function approximation problem. As special cases, our framework yields a short derivation of the recent Legendre Memory Unit (LMU) from first principles, and generalizes the ubiquitous gating mechanism of recurrent neural networks such as GRUs. This formal framework yields a new memory update mechanism (HiPPO-LegS) that scales through time to remember all history, avoiding priors on the timescale. HiPPO-LegS enjoys the theoretical benefits of timescale robustness, fast updates, and bounded gradients. By incorporating the memory dynamics into recurrent neural networks, HiPPO RNNs can empirically capture complex temporal dependencies. On the benchmark permuted MNIST dataset, HiPPO-LegS sets a new state-of-the-art accuracy of 98.3%. Finally, on a novel trajectory classification task testing robustness to out-of-distribution timescales and missing data, HiPPO-LegS outperforms RNN and neural ODE baselines by 25-40% accuracy."
                },
                {
                    "title": "Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention",
                    "abstract": "Transformers achieve remarkable performance in several tasks but due to their quadratic complexity, with respect to the input's length, they are prohibitively slow for very long sequences. To address this limitation, we express the self-attention as a linear dot-product of kernel feature maps and make use of the associativity property of matrix products to reduce the complexity from $\\mathcal{O}\\left(N^2\\right)$ to $\\mathcal{O}\\left(N\\right)$, where $N$ is the sequence length. We show that this formulation permits an iterative implementation that dramatically accelerates autoregressive transformers and reveals their relationship to recurrent neural networks. Our linear transformers achieve similar performance to vanilla transformers and they are up to 4000x faster on autoregressive prediction of very long sequences."
                },
                {
                    "title": "Linformer: Self-Attention with Linear Complexity",
                    "abstract": "Large transformer models have shown extraordinary success in achieving state-of-the-art results in many natural language processing applications. However, training and deploying these models can be prohibitively costly for long sequences, as the standard self-attention mechanism of the Transformer uses $O(n^2)$ time and space with respect to sequence length. In this paper, we demonstrate that the self-attention mechanism can be approximated by a low-rank matrix. We further exploit this finding to propose a new self-attention mechanism, which reduces the overall self-attention complexity from $O(n^2)$ to $O(n)$ in both time and space. The resulting linear transformer, the \\textit{Linformer}, performs on par with standard Transformer models, while being much more memory- and time-efficient."
                },
                {
                    "title": "Designing Network Design Spaces",
                    "abstract": "In this work, we present a new network design paradigm. Our goal is to help advance the understanding of network design and discover design principles that generalize across settings. Instead of focusing on designing individual network instances, we design network design spaces that parametrize populations of networks. The overall process is analogous to classic manual design of networks, but elevated to the design space level. Using our methodology we explore the structure aspect of network design and arrive at a low-dimensional design space consisting of simple, regular networks that we call RegNet. The core insight of the RegNet parametrization is surprisingly simple: widths and depths of good networks can be explained by a quantized linear function. We analyze the RegNet design space and arrive at interesting findings that do not match the current practice of network design. The RegNet design space provides simple and fast networks that work well across a wide range of flop regimes. Under comparable training settings and flops, the RegNet models outperform the popular EfficientNet models while being up to 5x faster on GPUs."
                },
                {
                    "title": "GLU Variants Improve Transformer",
                    "abstract": "Gated Linear Units (arXiv:1612.08083) consist of the component-wise product of two linear projections, one of which is first passed through a sigmoid function. Variations on GLU are possible, using different nonlinear (or even linear) functions in place of sigmoid. We test these variants in the feed-forward sublayers of the Transformer (arXiv:1706.03762) sequence-to-sequence model, and find that some of them yield quality improvements over the typically-used ReLU or GELU activations."
                },
                {
                    "title": "MMDetection: Open MMLab Detection Toolbox and Benchmark",
                    "abstract": "We present MMDetection, an object detection toolbox that contains a rich set of object detection and instance segmentation methods as well as related components and modules. The toolbox started from a codebase of MMDet team who won the detection track of COCO Challenge 2018. It gradually evolves into a unified platform that covers many popular detection methods and contemporary modules. It not only includes training and inference codes, but also provides weights for more than 200 network models. We believe this toolbox is by far the most complete detection toolbox. In this paper, we introduce the various features of this toolbox. In addition, we also conduct a benchmarking study on different methods, components, and their hyper-parameters. We wish that the toolbox and benchmark could serve the growing research community by providing a flexible toolkit to reimplement existing methods and develop their own new detectors. Code and models are available at this https URL. The project is under active development and we will keep this document updated."
                },
                {
                    "title": "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks",
                    "abstract": "Convolutional Neural Networks (ConvNets) are commonly developed at a fixed resource budget, and then scaled up for better accuracy if more resources are available. In this paper, we systematically study model scaling and identify that carefully balancing network depth, width, and resolution can lead to better performance. Based on this observation, we propose a new scaling method that uniformly scales all dimensions of depth/width/resolution using a simple yet highly effective compound coefficient. We demonstrate the effectiveness of this method on scaling up MobileNets and ResNet. \nTo go even further, we use neural architecture search to design a new baseline network and scale it up to obtain a family of models, called EfficientNets, which achieve much better accuracy and efficiency than previous ConvNets. In particular, our EfficientNet-B7 achieves state-of-the-art 84.4% top-1 / 97.1% top-5 accuracy on ImageNet, while being 8.4x smaller and 6.1x faster on inference than the best existing ConvNet. Our EfficientNets also transfer well and achieve state-of-the-art accuracy on CIFAR-100 (91.7%), Flowers (98.8%), and 3 other transfer learning datasets, with an order of magnitude fewer parameters. Source code is at this https URL."
                },
                {
                    "title": "Deformable ConvNets V2: More Deformable, Better Results",
                    "abstract": "The superior performance of Deformable Convolutional Networks arises from its ability to adapt to the geometric variations of objects. Through an examination of its adaptive behavior, we observe that while the spatial support for its neural features conforms more closely than regular ConvNets to object structure, this support may nevertheless extend well beyond the region of interest, causing features to be influenced by irrelevant image content. To address this problem, we present a reformulation of Deformable ConvNets that improves its ability to focus on pertinent image regions, through increased modeling power and stronger training. The modeling power is enhanced through a more comprehensive integration of deformable convolution within the network, and by introducing a modulation mechanism that expands the scope of deformation modeling. To effectively harness this enriched modeling capability, we guide network training via a proposed feature mimicking scheme that helps the network to learn features that reflect the object focus and classification power of R-CNN features. With the proposed contributions, this new version of Deformable ConvNets yields significant performance gains over the original model and produces leading results on the COCO benchmark for object detection and instance segmentation."
                },
                {
                    "title": "Pointwise Convolutional Neural Networks",
                    "abstract": "Deep learning with 3D data such as reconstructed point clouds and CAD models has received great research interests recently. However, the capability of using point clouds with convolutional neural network has been so far not fully explored. In this paper, we present a convolutional neural network for semantic segmentation and object recognition with 3D point clouds. At the core of our network is point-wise convolution, a new convolution operator that can be applied at each point of a point cloud. Our fully convolutional network design, while being surprisingly simple to implement, can yield competitive accuracy in both semantic segmentation and object recognition task."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Parallelizing Linear Recurrent Neural Nets Over Sequence Length",
                    "abstract": "Recurrent neural networks (RNNs) are widely used to model sequential data but their non-linear dependencies between sequence elements prevent parallelizing training over sequence length. We show the training of RNNs with only linear sequential dependencies can be parallelized over the sequence length using the parallel scan algorithm, leading to rapid training on long sequences even with small minibatch size. We develop a parallel linear recurrence CUDA kernel and show that it can be applied to immediately speed up training and inference of several state of the art RNN architectures by up to 9x. We abstract recent work on linear RNNs into a new framework of linear surrogate RNNs and develop a linear surrogate model for the long short-term memory unit, the GILR-LSTM, that utilizes parallel linear recurrence. We extend sequence learning to new extremely long sequence regimes that were previously out of reach by successfully training a GILR-LSTM on a synthetic sequence classification task with a one million timestep dependency."
                },
                {
                    "title": "Scene Parsing through ADE20K Dataset",
                    "abstract": "Scene parsing, or recognizing and segmenting objects and stuff in an image, is one of the key problems in computer vision. Despite the communitys efforts in data collection, there are still few image datasets covering a wide range of scenes and object categories with dense and detailed annotations for scene parsing. In this paper, we introduce and analyze the ADE20K dataset, spanning diverse annotations of scenes, objects, parts of objects, and in some cases even parts of parts. A scene parsing benchmark is built upon the ADE20K with 150 object and stuff classes included. Several segmentation baseline models are evaluated on the benchmark. A novel network design called Cascade Segmentation Module is proposed to parse a scene into stuff, objects, and object parts in a cascade and improve over the baselines. We further show that the trained scene parsing networks can lead to applications such as image content removal and scene synthesis1."
                },
                {
                    "title": "Dual Path Networks",
                    "abstract": "In this work, we present a simple, highly efficient and modularized Dual Path Network (DPN) for image classification which presents a new topology of connection paths internally. By revealing the equivalence of the state-of-the-art Residual Network (ResNet) and Densely Convolutional Network (DenseNet) within the HORNN framework, we find that ResNet enables feature re-usage while DenseNet enables new features exploration which are both important for learning good representations. To enjoy the benefits from both path topologies, our proposed Dual Path Network shares common features while maintaining the flexibility to explore new features through dual path architectures. Extensive experiments on three benchmark datasets, ImagNet-1k, Places365 and PASCAL VOC, clearly demonstrate superior performance of the proposed DPN over state-of-the-arts. In particular, on the ImagNet-1k dataset, a shallow DPN surpasses the best ResNeXt-101(64x4d) with 26% smaller model size, 25% less computational cost and 8% lower memory consumption, and a deeper DPN (DPN-131) further pushes the state-of-the-art single model performance with about 2 times faster training speed. Experiments on the Places365 large-scale scene dataset, PASCAL VOC detection dataset, and PASCAL VOC segmentation dataset also demonstrate its consistently better performance than DenseNet, ResNet and the latest ResNeXt model over various applications."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications",
                    "abstract": "We present a class of efficient models called MobileNets for mobile and embedded vision applications. MobileNets are based on a streamlined architecture that uses depth-wise separable convolutions to build light weight deep neural networks. We introduce two simple global hyper-parameters that efficiently trade off between latency and accuracy. These hyper-parameters allow the model builder to choose the right sized model for their application based on the constraints of the problem. We present extensive experiments on resource and accuracy tradeoffs and show strong performance compared to other popular models on ImageNet classification. We then demonstrate the effectiveness of MobileNets across a wide range of applications and use cases including object detection, finegrain classification, face attributes and large scale geo-localization."
                },
                {
                    "title": "Mask R-CNN",
                    "abstract": "We present a conceptually simple, flexible, and general framework for object instance segmentation. Our approach efficiently detects objects in an image while simultaneously generating a high-quality segmentation mask for each instance. The method, called Mask R-CNN, extends Faster R-CNN by adding a branch for predicting an object mask in parallel with the existing branch for bounding box recognition. Mask R-CNN is simple to train and adds only a small overhead to Faster R-CNN, running at 5 fps. Moreover, Mask R-CNN is easy to generalize to other tasks, e.g., allowing us to estimate human poses in the same framework. We show top results in all three tracks of the COCO suite of challenges, including instance segmentation, bounding-box object detection, and person keypoint detection. Without tricks, Mask R-CNN outperforms all existing, single-model entries on every task, including the COCO 2016 challenge winners. We hope our simple and effective approach will serve as a solid baseline and help ease future research in instance-level recognition. Code will be made available."
                },
                {
                    "title": "Deformable Convolutional Networks",
                    "abstract": "Convolutional neural networks (CNNs) are inherently limited to model geometric transformations due to the fixed geometric structures in their building modules. In this work, we introduce two new modules to enhance the transformation modeling capability of CNNs, namely, deformable convolution and deformable RoI pooling. Both are based on the idea of augmenting the spatial sampling locations in the modules with additional offsets and learning the offsets from the target tasks, without additional supervision. The new modules can readily replace their plain counterparts in existing CNNs and can be easily trained end-to-end by standard back-propagation, giving rise to deformable convolutional networks. Extensive experiments validate the performance of our approach. For the first time, we show that learning dense spatial transformation in deep CNNs is effective for sophisticated vision tasks such as object detection and semantic segmentation. The code is released at https://github.com/msracver/Deformable-ConvNets."
                },
                {
                    "title": "Understanding the Effective Receptive Field in Deep Convolutional Neural Networks",
                    "abstract": "We study characteristics of receptive fields of units in deep convolutional networks. The receptive field size is a crucial issue in many visual tasks, as the output must respond to large enough areas in the image to capture information about large objects. We introduce the notion of an effective receptive field size, and show that it both has a Gaussian distribution and only occupies a fraction of the full theoretical receptive field size. We analyze the effective receptive field in several architecture designs, and the effect of sub-sampling, skip connections, dropout and nonlinear activations on it. This leads to suggestions for ways to address its tendency to be too small."
                },
                {
                    "title": "Aggregated Residual Transformations for Deep Neural Networks",
                    "abstract": "We present a simple, highly modularized network architecture for image classification. Our network is constructed by repeating a building block that aggregates a set of transformations with the same topology. Our simple design results in a homogeneous, multi-branch architecture that has only a few hyper-parameters to set. This strategy exposes a new dimension, which we call cardinality (the size of the set of transformations), as an essential factor in addition to the dimensions of depth and width. On the ImageNet-1K dataset, we empirically show that even under the restricted condition of maintaining complexity, increasing cardinality is able to improve classification accuracy. Moreover, increasing cardinality is more effective than going deeper or wider when we increase the capacity. Our models, named ResNeXt, are the foundations of our entry to the ILSVRC 2016 classification task in which we secured 2nd place. We further investigate ResNeXt on an ImageNet-5K set and the COCO detection set, also showing better results than its ResNet counterpart. The code and models are publicly available online."
                },
                {
                    "title": "Neural Architecture Search with Reinforcement Learning",
                    "abstract": "Neural networks are powerful and flexible models that work well for many difficult learning tasks in image, speech and natural language understanding. Despite their success, neural networks are still hard to design. In this paper, we use a recurrent network to generate the model descriptions of neural networks and train this RNN with reinforcement learning to maximize the expected accuracy of the generated architectures on a validation set. On the CIFAR-10 dataset, our method, starting from scratch, can design a novel network architecture that rivals the best human-invented architecture in terms of test set accuracy. Our CIFAR-10 model achieves a test error rate of 3.65, which is 0.09 percent better and 1.05x faster than the previous state-of-the-art model that used a similar architectural scheme. On the Penn Treebank dataset, our model can compose a novel recurrent cell that outperforms the widely-used LSTM cell, and other state-of-the-art baselines. Our cell achieves a test set perplexity of 62.4 on the Penn Treebank, which is 3.6 perplexity better than the previous state-of-the-art model. The cell can also be transferred to the character language modeling task on PTB and achieves a state-of-the-art perplexity of 1.214."
                },
                {
                    "title": "Xception: Deep Learning with Depthwise Separable Convolutions",
                    "abstract": "We present an interpretation of Inception modules in convolutional neural networks as being an intermediate step in-between regular convolution and the depthwise separable convolution operation (a depthwise convolution followed by a pointwise convolution). In this light, a depthwise separable convolution can be understood as an Inception module with a maximally large number of towers. This observation leads us to propose a novel deep convolutional neural network architecture inspired by Inception, where Inception modules have been replaced with depthwise separable convolutions. We show that this architecture, dubbed Xception, slightly outperforms Inception V3 on the ImageNet dataset (which Inception V3 was designed for), and significantly outperforms Inception V3 on a larger image classification dataset comprising 350 million images and 17,000 classes. Since the Xception architecture has the same number of parameters as Inception V3, the performance gains are not due to increased capacity but rather to a more efficient use of model parameters."
                },
                {
                    "title": "Densely Connected Convolutional Networks",
                    "abstract": "Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections&#x2014;one between each layer and its subsequent layer&#x2014;our network has L(L+1)/2 direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters. We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet). DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less memory and computation to achieve high performance. Code and pre-trained models are available at https://github.com/liuzhuang13/DenseNet."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Multi-Scale Context Aggregation by Dilated Convolutions",
                    "abstract": "State-of-the-art models for semantic segmentation are based on adaptations of convolutional networks that had originally been designed for image classification. However, dense prediction and image classification are structurally different. In this work, we develop a new convolutional network module that is specifically designed for dense prediction. The presented module uses dilated convolutions to systematically aggregate multi-scale contextual information without losing resolution. The architecture is based on the fact that dilated convolutions support exponential expansion of the receptive field without loss of resolution or coverage. We show that the presented context module increases the accuracy of state-of-the-art semantic segmentation systems. In addition, we examine the adaptation of image classification networks to dense prediction and show that simplifying the adapted network can increase accuracy."
                },
                {
                    "title": "Character-level Convolutional Networks for Text Classification",
                    "abstract": "This article offers an empirical exploration on the use of character-level convolutional networks (ConvNets) for text classification. We constructed several large-scale datasets to show that character-level convolutional networks could achieve state-of-the-art or competitive results. Comparisons are offered against traditional models such as bag of words, n-grams and their TFIDF variants, and deep learning models such as word-based ConvNets and recurrent neural networks."
                },
                {
                    "title": "Going deeper with convolutions",
                    "abstract": "We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Crafting Papers on Machine Learning",
                    "abstract": "This essay gives advice to authors of papers on machine learning, although much of it car-ries over to other computational disciplines. The issues covered include the material that should appear in a well-balanced paper, factors that arise in di(cid:11)erent approaches to evaluation, and ways to improve a submission's ability to communicate ideas to its readers."
                },
                {
                    "title": "Backpropagation Applied to Handwritten Zip Code Recognition",
                    "abstract": "The ability of learning networks to generalize can be greatly enhanced by providing constraints from the task domain. This paper demonstrates how such constraints can be integrated into a backpropagation network through the architecture of the network. This approach has been successfully applied to the recognition of handwritten zip code digits provided by the U.S. Postal Service. A single network learns the entire recognition operation, going from the normalized image of the character to the final classification."
                },
                {
                    "title": "HiViT: A Simpler and More Efficient Design of Hierarchical Vision Transformer",
                    "abstract": "There has been a debate on the choice of plain vs. hierarchical vision transformers, where researchers often believe that the former (e.g., ViT) has a simpler design but the latter (e.g., Swin) enjoys higher recognition accuracy. Recently, the emerge of masked image modeling (MIM), a self-supervised pre-training method, raised a new challenge to vision transformers in terms of flexibility, i.e., part of image patches or tokens are to be discarded, which seems to claim the advantages of plain vision transformers. In this paper, we delve deep into the comparison between ViT and Swin, revealing that (i) the performance gain of Swin is mainly brought by a deepened backbone and relative positional encoding, (ii) the hierarchical design of Swin can be simplified into hierarchical patch embedding (proposed in this work), and (iii) other designs such as shifted-window attentions can be removed. By removing the unnecessary operations, we come up with a new architecture named HiViT (short for hierarchical ViT), which is simpler and more efficient than Swin yet further improves its performance on fully-supervised and self-supervised visual representation learning. In particular, after pre-trained using masked autoencoder (MAE) on ImageNet-1K, HiViT-B reports a 84.6% accuracy on ImageNet-1K classification, a 53.3% box AP on COCO detection, and a 52.8% mIoU on ADE20K segmentation, significantly surpassing the baseline. Code is available at https://github.com/zhangxiaosong18/hivit."
                },
                {
                    "title": "Scaling Up Your Kernels to 31x31: Revisiting Large Kernel Design in CNNs",
                    "abstract": "For training MobileNet V2 models (Sec. 3), we use 8 GPUs, an SGD optimizer with momentum of 0.9, a batch size of 32 per GPU, input resolution of 224\u00d7224, weight decay of 4 \u00d7 10\u22125, learning rate schedule with 5-epoch warmup, initial value of 0.1 and cosine annealing for 100 epochs. For the data augmentation, we only use random cropping and left-right flipping, as a common practice. For training RepLKNet models (Sec. 4.2),we use 32 GPUs and a batch size of 64 per GPU to train for 120 epochs. The optimizer is AdamW [10] with momentum of 0.9 and weight decay of 0.05. The learning rate setting includes an initial value of 4 \u00d7 10\u22123, cosine annealing and 10-epoch warm-up. For the data augmentation and regularization, we use RandAugment [4] (\u201crand-m9-mstd0.5-inc1\u201d as implemented by timm [15]), label smoothing coefficient of 0.1, mixup [18] with \u03b1 = 0.8, CutMix with \u03b1 = 1.0, Rand Erasing [19] with probability of 25% and Stochastic Depth with a drop-path rate of 30%, following the recent works [1, 8, 9, 12]. The RepLKNet-31B reported in Sec. 4.3 is trained with the same configurations except the epoch number of 300 and drop-path rate of 50%. For finetuning the 224\u00d7224-trained RepLKNet-31B with 384\u00d7384, we use 32 GPUs, a batch size of 32 per GPU, initial learning rate of 4 \u00d7 10\u22124, cosine annealing, 1-epoch warm-up, 30 epochs, model EMA (Exponential Moving Average) with momentum of 10\u22124, the same RandAugment as above but no CutMix nor mixup."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "Demystifying Local Vision Transformer: Sparse Connectivity, Weight Sharing, and Dynamic Weight",
                    "abstract": "Vision Transformer (ViT) attains state-of-the-art performance in visual recognition, and the variant, Local Vision Transformer, makes further improvements. The major component in Local Vision Transformer, local attention, performs the attention separately over small local windows. We rephrase local attention as a channel-wise locally-connected layer and analyze it from two network regularization manners, sparse connectivity and weight sharing, as well as weight computation. Sparse connectivity : there is no connection across channels, and each position is connected to the positions within a small local window. Weight sharing : the connection weights for one position are shared across channels or within each group of channels. Dynamic weight : the connection weights are dynamically predicted according to each image instance. We point out that local attention resembles depth-wise convolution and its dynamic version in sparse connectivity. The main difference lies in weight sharing - depth-wise convolution shares connection weights (kernel weights) across spatial positions. We empirically observe that the models based on depth-wise convolution and the dynamic variant with lower computation complexity perform on-par with or sometimes slightly better than Swin Transformer, an instance of Local Vision Transformer, for ImageNet classi\ufb01cation, COCO object detection and ADE semantic segmentation. These observations suggest that Local Vision Transformer takes advantage of two regularization forms and dynamic weight to increase the network capacity."
                },
                {
                    "title": "A New Approach to Linear Filtering and Prediction Problems",
                    "abstract": "AN IMPORTANT class of theoretical and practical problems in communication and control is of a statistical nature. Such problems are: (i) Prediction of random signals; (ii) separation of random signals from random noise; (iii) detection of signals of known form (pulses, sinusoids) in the presence of random noise. In his pioneering work, Wiener [1]3 showed that problems (i) and (ii) lead to the so-called Wiener-Hopf integral equation; he also gave a method (spectral factorization) for the solution of this integral equation in the practically important special case of stationary statistics and rational spectra. Many extensions and generalizations followed Wiener\u2019s basic work. Zadeh and Ragazzini solved the finite-memory case [2]. Concurrently and independently of Bode and Shannon [3], they also gave a simplified method [2) of solution. Booton discussed the nonstationary Wiener-Hopf equation [4]. These results are now in standard texts [5-6]. A somewhat different approach along these main lines has been given recently by Darlington [7]. For extensions to sampled signals, see, e.g., Franklin [8], Lees [9]. Another approach based on the eigenfunctions of the WienerHopf equation (which applies also to nonstationary problems whereas the preceding methods in general don\u2019t), has been pioneered by Davis [10] and applied by many others, e.g., Shinbrot [11], Blum [12], Pugachev [13], Solodovnikov [14]. In all these works, the objective is to obtain the specification of a linear dynamic system (Wiener filter) which accomplishes the prediction, separation, or detection of a random signal.4 \u2014\u2014\u2014 1 This research was supported in part by the U. S. Air Force Office of Scientific Research under Contract AF 49 (638)-382. 2 7212 Bellona Ave. 3 Numbers in brackets designate References at end of paper. 4 Of course, in general these tasks may be done better by nonlinear filters. At present, however, little or nothing is known about how to obtain (both theoretically and practically) these nonlinear filters. Contributed by the Instruments and Regulators Division and presented at the Instruments and Regulators Conference, March 29\u2013 Apri1 2, 1959, of THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS. NOTE: Statements and opinions advanced in papers are to be understood as individual expressions of their authors and not those of the Society. Manuscript received at ASME Headquarters, February 24, 1959. Paper No. 59-IRD\u201411. A New Approach to Linear Filtering and Prediction Problems"
                },
                {
                    "title": "Gradient-based learning applied to document recognition",
                    "abstract": "Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day."
                },
                {
                    "title": "Prefix sums and their applications",
                    "abstract": "Abstract: \"Experienced algorithm designers rely heavily on a set of building blocks and on the tools needed to put the blocks together into an algorithm. The understanding of these basic blocks and tools is therefore critical to the understanding of algorithms. Many of the blocks and tools needed for parallel algorithms extend from sequential algorithms, such as dynamic-programming and divide-and-conquer, but others are new. This paper introduces one of the simplest and most useful building blocks for parallel algorithms: the all-prefix-sums operation. The paper defines the operation, shows how to implement it on a P-RAM and illustrates many applications of the operation.In addition to being a useful building block, the all-prefix-sums operation is a good example of a computation that seems inherently sequential, but for which there is an efficient parallel algorithm.\""
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we efficiently implement a vision backbone that maintains the advantages of self-attention while reducing computational complexity for visual representation learning?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the field of computer vision, as it addresses the computational inefficiencies associated with self-attention mechanisms in Vision Transformers (ViTs). By developing a more efficient architecture, we can enable the processing of larger datasets and higher-resolution images, which could lead to improved performance across various visual tasks. This research could pave the way for future innovations in visual representation learning, making it more accessible and practical for real-world applications, such as autonomous driving, medical imaging, and augmented reality.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the inherent quadratic complexity of the self-attention mechanism, which becomes prohibitive when dealing with large spatial resolutions in vision data. Naive approaches may fail because they do not adequately address the need for global context while maintaining efficiency. Additionally, adapting existing algorithms designed for one-dimensional data (like Mamba) to two-dimensional visual data introduces significant technical obstacles, such as ensuring that contextual information is effectively captured without excessive computational overhead.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on either enhancing the efficiency of attention computation or improving the performance of CNNs and ViTs, but often at the cost of limiting the effective receptive field or degrading performance across tasks. Existing solutions have not successfully integrated the benefits of self-attention with the linear complexity required for high-resolution visual data. Our approach, which introduces the 2D Selective Scan (SS2D) mechanism, differs by providing a novel way to traverse spatial domains while preserving the advantages of self-attention, thus overcoming the limitations of prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves the development of VMamba, a vision backbone that incorporates SSM-based blocks with the 2D Selective Scan (SS2D) mechanism. We will evaluate the performance of VMamba on benchmark datasets such as ImageNet-1K for image classification, COCO for object detection, and ADE20K for semantic segmentation. The expected outcomes include achieving superior accuracy and throughput compared to existing models like Swin and ConvNeXt, with VMamba-Base anticipated to reach a top-1 accuracy of 83.9% on ImageNet-1K and"
            }
        },
        "author_data": {
            "c911ed1c-98b7-409b-aab3-3713a923f2f6": {
                "pk": "c911ed1c-98b7-409b-aab3-3713a923f2f6",
                "name": "Yue Liu",
                "collaborators": [
                    "Nicolas Privault",
                    "Yu Liu",
                    "Kang Yue",
                    "Changyu Ren",
                    "Jianqing Chen",
                    "Zhiwu Lin",
                    "Dahai He",
                    "Huiqun Jiang",
                    "Minghua Li"
                ],
                "domain": [
                    "Mathematics",
                    "Machine Learning",
                    "Graph Theory",
                    "Immersive Learning"
                ],
                "publications": [
                    {
                        "title": "On the Algebraic Representation of One-Tape Deterministic Turing Machine",
                        "abstract": "An algebraic representation of the Turing machines is given, where the configurations of Turing machines are represented by 4 order tensors, and the transition functions by 8 order tensors. Two types of tensor product are defined, one is to model the evolution of the Turing machines, and the other is to model the compositions of transition functions. It is shown that the two types of tensor product are harmonic in the sense that the associate law is obeyed."
                    },
                    {
                        "title": "Homomorphism of independent random variable convolution and matrix multiplication",
                        "abstract": "A map is given showing that convolutions of independent random variables over a finite group and matrix multiplications of doubly stochastic matrices are homomorphic. As an application, a short proof is given to the theorem that the limiting distributions of stochastic processes with stationary independent increments over a finite group are always uniform."
                    },
                    {
                        "title": "Behavioral analysis in immersive learning environments: A systematic literature review and research agenda",
                        "abstract": "The rapid growth of immersive technologies in educational areas has increased research interest in analyzing the specific behavioral patterns of learners in immersive learning environments. Considering the fact that research on the technical affordances of immersive technologies and the pedagogical affordances of behavioral analysis remains fragmented, this study first contributes by developing a conceptual framework that amalgamates learning requirements, specification, evaluation, and iteration into an integrated model to identify learning benefits and potential hurdles of behavioral analysis in immersive learning environments. Then, a systematic review was conducted underpinning the proposed conceptual framework to retrieve valuable empirical evidence from the 40 eligible articles during the last decade. The review findings suggest that (1) there is an essential need to sufficiently prepare the salient pedagogical requirements to define the specific learning stage, envisage intended cognitive objectives, and specify an appropriate set of learning activities, when developing comprehensive plans on behavioral analysis in immersive learning environments. (2) Researchers could customize the unique immersive experimental implementation by considering factors from four dimensions: learner, pedagogy, context, and representation. (3) The behavioral patterns constructed in immersive learning environments vary by considering the influence of behavioral analysis techniques, research themes, and immersive technical features. (4) The use of behavioral analysis in immersive learning environments faces several challenges from technical, implementation, and data processing perspectives. This study also articulates critical research agenda that could drive future investigation on behavioral analysis in immersive learning environments."
                    },
                    {
                        "title": "Pogorelov type $C^2$ estimates for Sum Hessian equations and a rigidity theorem",
                        "abstract": "We mainly study Pogorelov type $C^2$ estimates for solutions to the Dirichlet problem of Sum Hessian equations. We establish respectively Pogorelov type $C^2$ estimates for $k$-convex solutions and admissible solutions under some conditions. Furthermore, we apply such estimates to obtain a rigidity theorem for $k$-convex solutions of Sum Hessian equations in Euclidean space."
                    },
                    {
                        "title": "Instability of Standing Waves to the Inhomogeneous Nonlinear Schr\u00f6dinger Equation with Harmonic Potential",
                        "abstract": "We study the instability of standing-wave solutions $e^{i\\omega t}\\phi_{\\omega}(x)$ to the inhomogeneous nonlinear Schr\\\"{o}dinger equation $$i\\phi_t=-\\triangle\\phi+|x|^2\\phi-|x|^b|\\phi|^{p-1}\\phi, \\qquad \\in\\mathbb{R}^N, $$ where $ b > 0 $ and $ \\phi_{\\omega} $ is a ground-state solution. The results of the instability of standing-wave solutions reveal a balance between the frequency $\\omega $ of wave and the power of nonlinearity $p $ for any fixed $ b > 0. $"
                    },
                    {
                        "title": "Stability of peakons for the Degasperis-Procesi equation",
                        "abstract": "The Degasperis-Procesi equation can be derived as a member of a one-parameter family of asymptotic shallow water approximations to the Euler equations with the same asymptotic accuracy as that of the Camassa-Holm equation. In this paper, we study the orbital stability problem of the peaked solitons to the Degasperis-Procesi equation on the line. By constructing a Liapunov function, we prove that the shapes of these peakon solitons are stable under small perturbations."
                    },
                    {
                        "title": "A recursive algorithm for selling at the ultimate maximum in regime-switching models",
                        "abstract": "We propose a recursive algorithm for the numerical computation of the optimal value function $\\inf_{t\\le\\tau\\le T} E \\Big[\\sup_{0\\le s\\le T } Y_s / Y_{\\tau} \\big| {\\cal F}_t\\Big]$ over the stopping times $\\tau$ with respect to the filtration of a geometric Brownian motion $Y_t$ with Markovian regime switching. This method allows us to determine the boundary functions of the optimal stopping set when no associated Volterra integral equation is available. It applies in particular when regime-switching drifts have mixed signs, in which case the boundary functions may not be monotone."
                    },
                    {
                        "title": "Selling at the ultimate maximum in a regime switching model",
                        "abstract": "This paper deals with optimal prediction in a regime-switching model driven by a continuous-time Markov chain. We extend existing results for geometric Brownian motion by deriving optimal stopping strategies that depend on the current regime state, and prove a number of continuity properties relating to optimal value and boundary functions. Our approach replaces the use of closed form expressions, which are not available in our setting, with PDE arguments that also simplify the approach of [2] in the classical Brownian case."
                    },
                    {
                        "title": "Anomalous interfacial temperature profile induced by phonon localization",
                        "abstract": "Through the integration of the power spectral density, we obtain temperature profiles of both multi-segment harmonic and anharmonic systems, showing the presence of an anomalous negative temperature gradient inside the interfacial segment. Via investigating patterns of the power spectral density, we found that the counterintuitive phenomenon comes from the presence of interfacial localized phonon modes. Two out-band localized modes of the harmonic model, which make no contributions to local temperature due to the absence of phonon interactions, result in the concave temperature profile and over-cooling effect. For the anharmonic model, thanks to the phonon-phonon interactions, the localized modes are excited and make considerable contributions to interfacial temperature, which is clearly shown by examining the temperature accumulation function. When anharmonicity is considerably large, the negative temperature gradient is absent since the localized phonon modes are fully mixed. The presence of localized modes are evidently demonstrated by the inverse participation ratio and normal mode analysis for the isolated harmonic model."
                    },
                    {
                        "title": "Maximal signed bipartite graphs with totally disconnected graphs as star complements",
                        "abstract": "Let $\\dot{\\mathscr{B}}\\triangleq \\dot{\\mathscr{B}}_{\\dot{H}, \\mu}$ denote an arbitrary signed bipartite graph with $\\dot{H}$ as a star complement for an eigenvalue $\\mu$, where $\\dot{H}$ is a totally disconnected graph of order $s$. In this paper, by using Hadamard and Conference matrices as tools, the maximum order of $\\dot{\\mathscr{B}}$ and the extremal graphs are studied. It is shown that $\\dot{\\mathscr{B}}$ exists if and only if $\\mu^2$ is a positive integer. A formula of the maximum order of $\\dot{\\mathscr{B}}$ is given in the case of $\\mu^2=p\\times q$ such that $p$, $q$ are integers and there exists a $p$-order Hadamard or $(p+1)$-order Conference matrix. In particular, it is proved the maximum order of $\\dot{\\mathscr{B}}$ is $2s$ when either $q=1$, $s=c\\mu^2=cp$ or $q=1$, $s=c(\\mu^2+1)=c(p+1)$, $ c=1,2,3,\\cdots$. Futhermore, some extremal graphs are characterized."
                    },
                    {
                        "title": "Some Bounds on the Energy of Graphs with Self-Loops regarding $\u03bb_{1}$ and $\u03bb_{n}$",
                        "abstract": "Let $G_{S}$ be a graph with $n$ vertices obtained from a simple graph $G$ by attaching one self-loop at each vertex in $S \\subseteq V(G)$. The energy of $G_{S}$ is defined by Gutman et al. as $E(G_{S})=\\sum_{i=1}^{n}\\left| \\lambda_{i} -\\frac{\\sigma}{n} \\right|$, where $\\lambda_{1},\\dots,\\lambda_{n}$ are the adjacency eigenvalues of $G_{S}$ and $\\sigma$ is the number of self-loops of $G_{S}$. In this paper, several upper and lower bounds of $E(G_{S})$ regarding $\\lambda_{1}$ and $\\lambda_{n}$ are obtained. Especially, the upper bound $E(G_{S}) \\leq \\sqrt{n\\left(2m+\\sigma-\\frac{\\sigma^{2}}{n}\\right)}$ $(\\ast)$ given by Gutman et al. is improved to the following bound   \\begin{align*}   E(G_{S})\\leq \\sqrt{n\\left(2m+\\sigma-\\frac{\\sigma^{2}}{n}\\right)-\\frac{n}{2}\\left(\\left |\\lambda_{1}-\\frac{\\sigma}{n}\\right |-\\left |\\lambda_{n}-\\frac{\\sigma}{n}\\right |\\right)^{2}},   \\end{align*} where $\\left| \\lambda_{1}-\\frac{\\sigma}{n}\\right| \\geq \\dots \\geq \\left| \\lambda_{n}-\\frac{\\sigma}{n}\\right|$. Moreover, all graphs are characterized when the equality holds in Gutmans' bound $(\\ast)$ by using this new bound."
                    }
                ]
            },
            "294d0e41-e584-47d9-88f3-d391325140f8": {
                "pk": "294d0e41-e584-47d9-88f3-d391325140f8",
                "name": "Yunjie Tian",
                "collaborators": [
                    "Qixiang Ye",
                    "Lingxi Xie",
                    "Jianbin Jiao",
                    "Qi Tian",
                    "Jiemin Fang",
                    "Jihao Qiu",
                    "Tianren Ma",
                    "Yuan Zhang",
                    "Hongtian Yu",
                    "Yunfan Liu"
                ],
                "domain": [
                    "Computer Vision",
                    "Neural Architecture Search",
                    "Vision Transformers",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Spatial Transform Decoupling for Oriented Object Detection",
                        "abstract": "Vision Transformers (ViTs) have achieved remarkable success in computer vision tasks. However, their potential in rotation-sensitive scenarios has not been fully explored, and this limitation may be inherently attributed to the lack of spatial invariance in the data-forwarding process. In this study, we present a novel approach, termed Spatial Transform Decoupling (STD), providing a simple-yet-effective solution for oriented object detection with ViTs. Built upon stacked ViT blocks, STD utilizes separate network branches to predict the position, size, and angle of bounding boxes, effectively harnessing the spatial transform potential of ViTs in a divide-and-conquer fashion. Moreover, by aggregating cascaded activation masks (CAMs) computed upon the regressed parameters, STD gradually enhances features within regions of interest (RoIs), which complements the self-attention mechanism. Without bells and whistles, STD achieves state-of-the-art performance on the benchmark datasets including DOTA-v1.0 (82.24% mAP) and HRSC2016 (98.55% mAP), which demonstrates the effectiveness of the proposed method. Source code is available at https://github.com/yuhongtian17/Spatial-Transform-Decoupling."
                    },
                    {
                        "title": "Discretization-Aware Architecture Search",
                        "abstract": "The search cost of neural architecture search (NAS) has been largely reduced by weight-sharing methods. These methods optimize a super-network with all possible edges and operations, and determine the optimal sub-network by discretization, \\textit{i.e.}, pruning off weak candidates. The discretization process, performed on either operations or edges, incurs significant inaccuracy and thus the quality of the final architecture is not guaranteed. This paper presents discretization-aware architecture search (DA\\textsuperscript{2}S), with the core idea being adding a loss term to push the super-network towards the configuration of desired topology, so that the accuracy loss brought by discretization is largely alleviated. Experiments on standard image classification benchmarks demonstrate the superiority of our approach, in particular, under imbalanced target network configurations that were not studied before."
                    },
                    {
                        "title": "Adaptive Linear Span Network for Object Skeleton Detection",
                        "abstract": "Conventional networks for object skeleton detection are usually hand-crafted. Although effective, they require intensive priori knowledge to configure representative features for objects in different scale granularity.In this paper, we propose adaptive linear span network (AdaLSN), driven by neural architecture search (NAS), to automatically configure and integrate scale-aware features for object skeleton detection. AdaLSN is formulated with the theory of linear span, which provides one of the earliest explanations for multi-scale deep feature fusion. AdaLSN is materialized by defining a mixed unit-pyramid search space, which goes beyond many existing search spaces using unit-level or pyramid-level features.Within the mixed space, we apply genetic architecture search to jointly optimize unit-level operations and pyramid-level connections for adaptive feature space expansion. AdaLSN substantiates its versatility by achieving significantly higher accuracy and latency trade-off compared with state-of-the-arts. It also demonstrates general applicability to image-to-mask tasks such as edge detection and road extraction. Code is available at \\href{https://github.com/sunsmarterjie/SDL-Skeleton}{\\color{magenta}github.com/sunsmarterjie/SDL-Skeleton}."
                    },
                    {
                        "title": "Exploring Complicated Search Spaces with Interleaving-Free Sampling",
                        "abstract": "The existing neural architecture search algorithms are mostly working on search spaces with short-distance connections. We argue that such designs, though safe and stable, obstacles the search algorithms from exploring more complicated scenarios. In this paper, we build the search algorithm upon a complicated search space with long-distance connections, and show that existing weight-sharing search algorithms mostly fail due to the existence of \\textbf{interleaved connections}. Based on the observation, we present a simple yet effective algorithm named \\textbf{IF-NAS}, where we perform a periodic sampling strategy to construct different sub-networks during the search procedure, avoiding the interleaved connections to emerge in any of them. In the proposed search space, IF-NAS outperform both random sampling and previous weight-sharing search algorithms by a significant margin. IF-NAS also generalizes to the micro cell-based spaces which are much easier. Our research emphasizes the importance of macro structure and we look forward to further efforts along this direction."
                    },
                    {
                        "title": "Fast-iTPN: Integrally Pre-Trained Transformer Pyramid Network with Token Migration",
                        "abstract": "We propose integrally pre-trained transformer pyramid network (iTPN), towards jointly optimizing the network backbone and the neck, so that transfer gap between representation models and downstream tasks is minimal. iTPN is born with two elaborated designs: 1) The first pre-trained feature pyramid upon vision transformer (ViT). 2) Multi-stage supervision to the feature pyramid using masked feature modeling (MFM). iTPN is updated to Fast-iTPN, reducing computational memory overhead and accelerating inference through two flexible designs. 1) Token migration: dropping redundant tokens of the backbone while replenishing them in the feature pyramid without attention operations. 2) Token gathering: reducing computation cost caused by global attention by introducing few gathering tokens. The base/large-level Fast-iTPN achieve 88.75%/89.5% top-1 accuracy on ImageNet-1K. With 1x training schedule using DINO, the base/large-level Fast-iTPN achieves 58.4%/58.8% box AP on COCO object detection, and a 57.5%/58.7% mIoU on ADE20K semantic segmentation using MaskDINO. Fast-iTPN can accelerate the inference procedure by up to 70%, with negligible performance loss, demonstrating the potential to be a powerful backbone for downstream vision tasks. The code is available at: github.com/sunsmarterjie/iTPN."
                    },
                    {
                        "title": "GraFormer: Graph Convolution Transformer for 3D Pose Estimation",
                        "abstract": "Exploiting relations among 2D joints plays a crucial role yet remains semi-developed in 2D-to-3D pose estimation. To alleviate this issue, we propose GraFormer, a novel transformer architecture combined with graph convolution for 3D pose estimation. The proposed GraFormer comprises two repeatedly stacked core modules, GraAttention and ChebGConv block. GraAttention enables all 2D joints to interact in global receptive field without weakening the graph structure information of joints, which introduces vital features for later modules. Unlike vanilla graph convolutions that only model the apparent relationship of joints, ChebGConv block enables 2D joints to interact in the high-order sphere, which formulates their hidden implicit relations. We empirically show the superiority of GraFormer through conducting extensive experiments across popular benchmarks. Specifically, GraFormer outperforms state of the art on Human3.6M dataset while using 18$\\%$ parameters. The code is available at https://github.com/Graformer/GraFormer ."
                    },
                    {
                        "title": "ClawMachine: Fetching Visual Tokens as An Entity for Referring and Grounding",
                        "abstract": "An essential topic for multimodal large language models (MLLMs) is aligning vision and language concepts at a finer level. In particular, we devote efforts to encoding visual referential information for tasks such as referring and grounding. Existing methods, including proxy encoding and geometry encoding, incorporate additional syntax to encode the object's location, bringing extra burdens in training MLLMs to communicate between language and vision. This study presents ClawMachine, offering a new methodology that notates an entity directly using the visual tokens. It allows us to unify the prompt and answer of visual referential tasks without additional syntax. Upon a joint vision-language vocabulary, ClawMachine unifies visual referring and grounding into an auto-regressive format and learns with a decoder-only architecture. Experiments validate that our model achieves competitive performance across visual referring and grounding tasks with a reduced demand for training data. Additionally, ClawMachine demonstrates a native ability to integrate multi-source information for complex visual reasoning, which prior MLLMs can hardly perform without specific adaptions."
                    },
                    {
                        "title": "HiViT: Hierarchical Vision Transformer Meets Masked Image Modeling",
                        "abstract": "Recently, masked image modeling (MIM) has offered a new methodology of self-supervised pre-training of vision transformers. A key idea of efficient implementation is to discard the masked image patches (or tokens) throughout the target network (encoder), which requires the encoder to be a plain vision transformer (e.g., ViT), albeit hierarchical vision transformers (e.g., Swin Transformer) have potentially better properties in formulating vision inputs. In this paper, we offer a new design of hierarchical vision transformers named HiViT (short for Hierarchical ViT) that enjoys both high efficiency and good performance in MIM. The key is to remove the unnecessary \"local inter-unit operations\", deriving structurally simple hierarchical vision transformers in which mask-units can be serialized like plain vision transformers. For this purpose, we start with Swin Transformer and (i) set the masking unit size to be the token size in the main stage of Swin Transformer, (ii) switch off inter-unit self-attentions before the main stage, and (iii) eliminate all operations after the main stage. Empirical studies demonstrate the advantageous performance of HiViT in terms of fully-supervised, self-supervised, and transfer learning. In particular, in running MAE on ImageNet-1K, HiViT-B reports a +0.6% accuracy gain over ViT-B and a 1.9$\\times$ speed-up over Swin-B, and the performance gain generalizes to downstream tasks of detection and segmentation. Code will be made publicly available."
                    },
                    {
                        "title": "vHeat: Building Vision Models upon Heat Conduction",
                        "abstract": "A fundamental problem in learning robust and expressive visual representations lies in efficiently estimating the spatial relationships of visual semantics throughout the entire image. In this study, we propose vHeat, a novel vision backbone model that simultaneously achieves both high computational efficiency and global receptive field. The essential idea, inspired by the physical principle of heat conduction, is to conceptualize image patches as heat sources and model the calculation of their correlations as the diffusion of thermal energy. This mechanism is incorporated into deep models through the newly proposed module, the Heat Conduction Operator (HCO), which is physically plausible and can be efficiently implemented using DCT and IDCT operations with a complexity of $\\mathcal{O}(N^{1.5})$. Extensive experiments demonstrate that vHeat surpasses Vision Transformers (ViTs) across various vision tasks, while also providing higher inference speeds, reduced FLOPs, and lower GPU memory usage for high-resolution images. The code will be released at https://github.com/MzeroMiko/vHeat."
                    },
                    {
                        "title": "Semantic-Aware Generation for Self-Supervised Visual Representation Learning",
                        "abstract": "In this paper, we propose a self-supervised visual representation learning approach which involves both generative and discriminative proxies, where we focus on the former part by requiring the target network to recover the original image based on the mid-level features. Different from prior work that mostly focuses on pixel-level similarity between the original and generated images, we advocate for Semantic-aware Generation (SaGe) to facilitate richer semantics rather than details to be preserved in the generated image. The core idea of implementing SaGe is to use an evaluator, a deep network that is pre-trained without labels, for extracting semantic-aware features. SaGe complements the target network with view-specific features and thus alleviates the semantic degradation brought by intensive data augmentations. We execute SaGe on ImageNet-1K and evaluate the pre-trained models on five downstream tasks including nearest neighbor test, linear classification, and fine-scaled image recognition, demonstrating its ability to learn stronger visual representations."
                    },
                    {
                        "title": "Beyond Masking: Demystifying Token-Based Pre-Training for Vision Transformers",
                        "abstract": "The past year has witnessed a rapid development of masked image modeling (MIM). MIM is mostly built upon the vision transformers, which suggests that self-supervised visual representations can be done by masking input image parts while requiring the target model to recover the missing contents. MIM has demonstrated promising results on downstream tasks, yet we are interested in whether there exist other effective ways to `learn by recovering missing contents'. In this paper, we investigate this topic by designing five other learning objectives that follow the same procedure as MIM but degrade the input image in different ways. With extensive experiments, we manage to summarize a few design principles for token-based pre-training of vision transformers. In particular, the best practice is obtained by keeping the original image style and enriching spatial masking with spatial misalignment -- this design achieves superior performance over MIM in a series of downstream recognition tasks without extra computational cost. The code is available at https://github.com/sunsmarterjie/beyond_masking."
                    },
                    {
                        "title": "ChatterBox: Multi-round Multimodal Referring and Grounding",
                        "abstract": "In this study, we establish a baseline for a new task named multimodal multi-round referring and grounding (MRG), opening up a promising direction for instance-level multimodal dialogues. We present a new benchmark and an efficient vision-language model for this purpose. The new benchmark, named CB-300K, spans challenges including multi-round dialogue, complex spatial relationships among multiple instances, and consistent reasoning, which are beyond those shown in existing benchmarks. The proposed model, named ChatterBox, utilizes a two-branch architecture to collaboratively handle vision and language tasks. By tokenizing instance regions, the language branch acquires the ability to perceive referential information. Meanwhile, ChatterBox feeds a query embedding in the vision branch to a token receiver for visual grounding. A two-stage optimization strategy is devised, making use of both CB-300K and auxiliary external data to improve the model's stability and capacity for instance-level understanding. Experiments show that ChatterBox outperforms existing models in MRG both quantitatively and qualitatively, paving a new path towards multimodal dialogue scenarios with complicated and precise interactions. Code, data, and model are available at: https://github.com/sunsmarterjie/ChatterBox."
                    },
                    {
                        "title": "Artemis: Towards Referential Understanding in Complex Videos",
                        "abstract": "Videos carry rich visual information including object description, action, interaction, etc., but the existing multimodal large language models (MLLMs) fell short in referential understanding scenarios such as video-based referring. In this paper, we present Artemis, an MLLM that pushes video-based referential understanding to a finer level. Given a video, Artemis receives a natural-language question with a bounding box in any video frame and describes the referred target in the entire video. The key to achieving this goal lies in extracting compact, target-specific video features, where we set a solid baseline by tracking and selecting spatiotemporal features from the video. We train Artemis on the newly established VideoRef45K dataset with 45K video-QA pairs and design a computationally efficient, three-stage training procedure. Results are promising both quantitatively and qualitatively. Additionally, we show that \\model can be integrated with video grounding and text summarization tools to understand more complex scenarios. Code and data are available at https://github.com/qiujihao19/Artemis."
                    }
                ]
            },
            "e934631a-2883-4e4a-8046-0a23ef73e668": {
                "pk": "e934631a-2883-4e4a-8046-0a23ef73e668",
                "name": "Yuzhong Zhao",
                "collaborators": [
                    "Weijia Wu",
                    "Mike Zheng Shou",
                    "Qixiang Ye",
                    "Fang Wan",
                    "Zhuang Li",
                    "Hong Zhou",
                    "Chunhua Shen",
                    "Chen Gong",
                    "Jiahong Li",
                    "Weiqiang Wang"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Generative Models",
                    "Multi-modal Learning"
                ],
                "publications": [
                    {
                        "title": "Explore Faster Localization Learning For Scene Text Detection",
                        "abstract": "Generally pre-training and long-time training computation are necessary for obtaining a good-performance text detector based on deep networks. In this paper, we present a new scene text detection network (called FANet) with a Fast convergence speed and Accurate text localization. The proposed FANet is an end-to-end text detector based on transformer feature learning and normalized Fourier descriptor modeling, where the Fourier Descriptor Proposal Network and Iterative Text Decoding Network are designed to efficiently and accurately identify text proposals. Additionally, a Dense Matching Strategy and a well-designed loss function are also proposed for optimizing the network performance. Extensive experiments are carried out to demonstrate that the proposed FANet can achieve the SOTA performance with fewer training epochs and no pre-training. When we introduce additional data for pre-training, the proposed FANet can achieve SOTA performance on MSRATD500, CTW1500 and TotalText. The ablation experiments also verify the effectiveness of our contributions."
                    },
                    {
                        "title": "Generative Prompt Model for Weakly Supervised Object Localization",
                        "abstract": "Weakly supervised object localization (WSOL) remains challenging when learning object localization models from image category labels. Conventional methods that discriminatively train activation models ignore representative yet less discriminative object parts. In this study, we propose a generative prompt model (GenPromp), defining the first generative pipeline to localize less discriminative object parts by formulating WSOL as a conditional image denoising procedure. During training, GenPromp converts image category labels to learnable prompt embeddings which are fed to a generative model to conditionally recover the input image with noise and learn representative embeddings. During inference, enPromp combines the representative embeddings with discriminative embeddings (queried from an off-the-shelf vision-language model) for both representative and discriminative capacity. The combined embeddings are finally used to generate multi-scale high-quality attention maps, which facilitate localizing full object extent. Experiments on CUB-200-2011 and ILSVRC show that GenPromp respectively outperforms the best discriminative models by 5.2% and 5.6% (Top-1 Loc), setting a solid baseline for WSOL with the generative model. Code is available at https://github.com/callsys/GenPromp."
                    },
                    {
                        "title": "ControlCap: Controllable Region-level Captioning",
                        "abstract": "Region-level captioning is challenged by the caption degeneration issue, which refers to that pre-trained multimodal models tend to predict the most frequent captions but miss the less frequent ones. In this study, we propose a controllable region-level captioning (ControlCap) approach, which introduces control words to a multimodal model to address the caption degeneration issue. In specific, ControlCap leverages a discriminative module to generate control words within the caption space to partition it to multiple sub-spaces. The multimodal model is constrained to generate captions within a few sub-spaces containing the control words, which increases the opportunity of hitting less frequent captions, alleviating the caption degeneration issue. Furthermore, interactive control words can be given by either a human or an expert model, which enables captioning beyond the training caption space, enhancing the model's generalization ability. Extensive experiments on Visual Genome and RefCOCOg datasets show that ControlCap respectively improves the CIDEr score by 21.6 and 2.2, outperforming the state-of-the-arts by significant margins. Code is available at https://github.com/callsys/ControlCap."
                    },
                    {
                        "title": "FlowText: Synthesizing Realistic Scene Text Video with Optical Flow Estimation",
                        "abstract": "Current video text spotting methods can achieve preferable performance, powered with sufficient labeled training data. However, labeling data manually is time-consuming and labor-intensive. To overcome this, using low-cost synthetic data is a promising alternative. This paper introduces a novel video text synthesis technique called FlowText, which utilizes optical flow estimation to synthesize a large amount of text video data at a low cost for training robust video text spotters. Unlike existing methods that focus on image-level synthesis, FlowText concentrates on synthesizing temporal information of text instances across consecutive frames using optical flow. This temporal information is crucial for accurately tracking and spotting text in video sequences, including text movement, distortion, appearance, disappearance, shelter, and blur. Experiments show that combining general detectors like TransDETR with the proposed FlowText produces remarkable results on various datasets, such as ICDAR2015video and ICDAR2013video. Code is available at https://github.com/callsys/FlowText."
                    },
                    {
                        "title": "Continual Learning for Image Segmentation with Dynamic Query",
                        "abstract": "Image segmentation based on continual learning exhibits a critical drop of performance, mainly due to catastrophic forgetting and background shift, as they are required to incorporate new classes continually. In this paper, we propose a simple, yet effective Continual Image Segmentation method with incremental Dynamic Query (CISDQ), which decouples the representation learning of both old and new knowledge with lightweight query embedding. CISDQ mainly includes three contributions: 1) We define dynamic queries with adaptive background class to exploit past knowledge and learn future classes naturally. 2) CISDQ proposes a class/instance-aware Query Guided Knowledge Distillation strategy to overcome catastrophic forgetting by capturing the inter-class diversity and intra-class identity. 3) Apart from semantic segmentation, CISDQ introduce the continual learning for instance segmentation in which instance-wise labeling and supervision are considered. Extensive experiments on three datasets for two tasks (i.e., continual semantic and instance segmentation are conducted to demonstrate that CISDQ achieves the state-of-the-art performance, specifically, obtaining 4.4% and 2.9% mIoU improvements for the ADE 100-10 (6 steps) setting and ADE 100-5 (11 steps) setting."
                    },
                    {
                        "title": "DiffuMask: Synthesizing Images with Pixel-level Annotations for Semantic Segmentation Using Diffusion Models",
                        "abstract": "Collecting and annotating images with pixel-wise labels is time-consuming and laborious. In contrast, synthetic data can be freely available using a generative model (e.g., DALL-E, Stable Diffusion). In this paper, we show that it is possible to automatically obtain accurate semantic masks of synthetic images generated by the Off-the-shelf Stable Diffusion model, which uses only text-image pairs during training. Our approach, called DiffuMask, exploits the potential of the cross-attention map between text and image, which is natural and seamless to extend the text-driven image synthesis to semantic mask generation. DiffuMask uses text-guided cross-attention information to localize class/word-specific regions, which are combined with practical techniques to create a novel high-resolution and class-discriminative pixel-wise mask. The methods help to reduce data collection and annotation costs obviously. Experiments demonstrate that the existing segmentation methods trained on synthetic data of DiffuMask can achieve a competitive performance over the counterpart of real data (VOC 2012, Cityscapes). For some classes (e.g., bird), DiffuMask presents promising performance, close to the stateof-the-art result of real data (within 3% mIoU gap). Moreover, in the open-vocabulary segmentation (zero-shot) setting, DiffuMask achieves a new SOTA result on Unseen class of VOC 2012. The project website can be found at https://weijiawu.github.io/DiffusionMask/."
                    },
                    {
                        "title": "DynRefer: Delving into Region-level Multi-modality Tasks via Dynamic Resolution",
                        "abstract": "Region-level multi-modality methods can translate referred image regions to human preferred language descriptions. Unfortunately, most of existing methods using fixed visual inputs remain lacking the resolution adaptability to find out precise language descriptions. In this study, we propose a dynamic resolution approach, referred to as DynRefer, to pursue high-accuracy region-level referring through mimicking the resolution adaptability of human visual cognition. DynRefer first implements stochastic vision-language alignment. It aligns desired language descriptions of multi-modality tasks with images of stochastic resolution, which are constructed by nesting a set of views around the referred region. DynRefer then implements dynamic multi-modality referring, which is realized by selecting views based on image and language priors. This allows the visual information used for referring to better match human preferences, thereby improving the representational adaptability of region-level multi-modality models. Extensive experiments show that DynRefer brings mutual improvement upon tasks including region-level captioning, open-vocabulary region recognition and attribute detection. Last but not least, DynRefer achieves new state-of-the-art on multiple region-level multi-modality tasks using a single model. Code is available at https://github.com/callsys/DynRefer."
                    },
                    {
                        "title": "PrivImage: Differentially Private Synthetic Image Generation using Diffusion Models with Semantic-Aware Pretraining",
                        "abstract": "Differential Privacy (DP) image data synthesis, which leverages the DP technique to generate synthetic data to replace the sensitive data, allowing organizations to share and utilize synthetic images without privacy concerns. Previous methods incorporate the advanced techniques of generative models and pre-training on a public dataset to produce exceptional DP image data, but suffer from problems of unstable training and massive computational resource demands. This paper proposes a novel DP image synthesis method, termed PRIVIMAGE, which meticulously selects pre-training data, promoting the efficient creation of DP datasets with high fidelity and utility. PRIVIMAGE first establishes a semantic query function using a public dataset. Then, this function assists in querying the semantic distribution of the sensitive dataset, facilitating the selection of data from the public dataset with analogous semantics for pre-training. Finally, we pre-train an image generative model using the selected data and then fine-tune this model on the sensitive dataset using Differentially Private Stochastic Gradient Descent (DP-SGD). PRIVIMAGE allows us to train a lightly parameterized generative model, reducing the noise in the gradient during DP-SGD training and enhancing training stability. Extensive experiments demonstrate that PRIVIMAGE uses only 1% of the public dataset for pre-training and 7.6% of the parameters in the generative model compared to the state-of-the-art method, whereas achieves superior synthetic performance and conserves more computational resources. On average, PRIVIMAGE achieves 30.1% lower FID and 12.6% higher Classification Accuracy than the state-of-the-art method. The replication package and datasets can be accessed online."
                    },
                    {
                        "title": "DatasetDM: Synthesizing Data with Perception Annotations Using Diffusion Models",
                        "abstract": "Current deep networks are very data-hungry and benefit from training on largescale datasets, which are often time-consuming to collect and annotate. By contrast, synthetic data can be generated infinitely using generative models such as DALL-E and diffusion models, with minimal effort and cost. In this paper, we present DatasetDM, a generic dataset generation model that can produce diverse synthetic images and the corresponding high-quality perception annotations (e.g., segmentation masks, and depth). Our method builds upon the pre-trained diffusion model and extends text-guided image synthesis to perception data generation. We show that the rich latent code of the diffusion model can be effectively decoded as accurate perception annotations using a decoder module. Training the decoder only needs less than 1% (around 100 images) manually labeled images, enabling the generation of an infinitely large annotated dataset. Then these synthetic data can be used for training various perception models for downstream tasks. To showcase the power of the proposed approach, we generate datasets with rich dense pixel-wise labels for a wide range of downstream tasks, including semantic segmentation, instance segmentation, and depth estimation. Notably, it achieves 1) state-of-the-art results on semantic segmentation and instance segmentation; 2) significantly more robust on domain generalization than using the real data alone; and state-of-the-art results in zero-shot segmentation setting; and 3) flexibility for efficient application and novel task composition (e.g., image editing). The project website and code can be found at https://weijiawu.github.io/DatasetDM_page/ and https://github.com/showlab/DatasetDM, respectively"
                    },
                    {
                        "title": "Evaluation of Text-to-Video Generation Models: A Dynamics Perspective",
                        "abstract": "Comprehensive and constructive evaluation protocols play an important role in the development of sophisticated text-to-video (T2V) generation models. Existing evaluation protocols primarily focus on temporal consistency and content continuity, yet largely ignore the dynamics of video content. Dynamics are an essential dimension for measuring the visual vividness and the honesty of video content to text prompts. In this study, we propose an effective evaluation protocol, termed DEVIL, which centers on the dynamics dimension to evaluate T2V models. For this purpose, we establish a new benchmark comprising text prompts that fully reflect multiple dynamics grades, and define a set of dynamics scores corresponding to various temporal granularities to comprehensively evaluate the dynamics of each generated video. Based on the new benchmark and the dynamics scores, we assess T2V models with the design of three metrics: dynamics range, dynamics controllability, and dynamics-based quality. Experiments show that DEVIL achieves a Pearson correlation exceeding 90% with human ratings, demonstrating its potential to advance T2V generation models. Code is available at https://github.com/MingXiangL/DEVIL."
                    },
                    {
                        "title": "ICDAR 2023 Video Text Reading Competition for Dense and Small Text",
                        "abstract": "Recently, video text detection, tracking, and recognition in natural scenes are becoming very popular in the computer vision community. However, most existing algorithms and benchmarks focus on common text cases (e.g., normal size, density) and single scenarios, while ignoring extreme video text challenges, i.e., dense and small text in various scenarios. In this competition report, we establish a video text reading benchmark, DSText, which focuses on dense and small text reading challenges in the video with various scenarios. Compared with the previous datasets, the proposed dataset mainly include three new challenges: 1) Dense video texts, a new challenge for video text spotter. 2) High-proportioned small texts. 3) Various new scenarios, e.g., Game, sports, etc. The proposed DSText includes 100 video clips from 12 open scenarios, supporting two tasks (i.e., video text tracking (Task 1) and end-to-end video text spotting (Task 2)). During the competition period (opened on 15th February 2023 and closed on 20th March 2023), a total of 24 teams participated in the three proposed tasks with around 30 valid submissions, respectively. In this article, we describe detailed statistical information of the dataset, tasks, evaluation protocols and the results summaries of the ICDAR 2023 on DSText competition. Moreover, we hope the benchmark will promise video text research in the community."
                    },
                    {
                        "title": "A Large Cross-Modal Video Retrieval Dataset with Reading Comprehension",
                        "abstract": "Most existing cross-modal language-to-video retrieval (VR) research focuses on single-modal input from video, i.e., visual representation, while the text is omnipresent in human environments and frequently critical to understand video. To study how to retrieve video with both modal inputs, i.e., visual and text semantic representations, we first introduce a large-scale and cross-modal Video Retrieval dataset with text reading comprehension, TextVR, which contains 42.2k sentence queries for 10.5k videos of 8 scenario domains, i.e., Street View (indoor), Street View (outdoor), Games, Sports, Driving, Activity, TV Show, and Cooking. The proposed TextVR requires one unified cross-modal model to recognize and comprehend texts, relate them to the visual context, and decide what text semantic information is vital for the video retrieval task. Besides, we present a detailed analysis of TextVR compared to the existing datasets and design a novel multimodal video retrieval baseline for the text-based video retrieval task. The dataset analysis and extensive experiments show that our TextVR benchmark provides many new technical challenges and insights from previous datasets for the video-and-language community. The project website and GitHub repo can be found at https://sites.google.com/view/loveucvpr23/guest-track and https://github.com/callsys/TextVR, respectively."
                    }
                ]
            },
            "7f0cd8a7-962b-4da4-b226-f53373ef9c7a": {
                "pk": "7f0cd8a7-962b-4da4-b226-f53373ef9c7a",
                "name": "Hongtian Yu",
                "collaborators": [
                    "Qixiang Ye",
                    "Yunfan Liu",
                    "Yunjie Tian",
                    "Yue Liu",
                    "Zhaozhi Wang",
                    "Yaowei Wang",
                    "Yangu Li",
                    "Mingrui Wu",
                    "Letian Shen",
                    "Yunxuan Song"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Vision Transformers",
                    "Particle Physics"
                ],
                "publications": [
                    {
                        "title": "Spatial Transform Decoupling for Oriented Object Detection",
                        "abstract": "Vision Transformers (ViTs) have achieved remarkable success in computer vision tasks. However, their potential in rotation-sensitive scenarios has not been fully explored, and this limitation may be inherently attributed to the lack of spatial invariance in the data-forwarding process. In this study, we present a novel approach, termed Spatial Transform Decoupling (STD), providing a simple-yet-effective solution for oriented object detection with ViTs. Built upon stacked ViT blocks, STD utilizes separate network branches to predict the position, size, and angle of bounding boxes, effectively harnessing the spatial transform potential of ViTs in a divide-and-conquer fashion. Moreover, by aggregating cascaded activation masks (CAMs) computed upon the regressed parameters, STD gradually enhances features within regions of interest (RoIs), which complements the self-attention mechanism. Without bells and whistles, STD achieves state-of-the-art performance on the benchmark datasets including DOTA-v1.0 (82.24% mAP) and HRSC2016 (98.55% mAP), which demonstrates the effectiveness of the proposed method. Source code is available at https://github.com/yuhongtian17/Spatial-Transform-Decoupling."
                    },
                    {
                        "title": "vHeat: Building Vision Models upon Heat Conduction",
                        "abstract": "A fundamental problem in learning robust and expressive visual representations lies in efficiently estimating the spatial relationships of visual semantics throughout the entire image. In this study, we propose vHeat, a novel vision backbone model that simultaneously achieves both high computational efficiency and global receptive field. The essential idea, inspired by the physical principle of heat conduction, is to conceptualize image patches as heat sources and model the calculation of their correlations as the diffusion of thermal energy. This mechanism is incorporated into deep models through the newly proposed module, the Heat Conduction Operator (HCO), which is physically plausible and can be efficiently implemented using DCT and IDCT operations with a complexity of $\\mathcal{O}(N^{1.5})$. Extensive experiments demonstrate that vHeat surpasses Vision Transformers (ViTs) across various vision tasks, while also providing higher inference speeds, reduced FLOPs, and lower GPU memory usage for high-resolution images. The code will be released at https://github.com/MzeroMiko/vHeat."
                    },
                    {
                        "title": "Vision Calorimeter for Anti-neutron Reconstruction: A Baseline",
                        "abstract": "In high-energy physics, anti-neutrons ($\\bar{n}$) are fundamental particles that frequently appear as final-state particles, and the reconstruction of their kinematic properties provides an important probe for understanding the governing principles. However, this confronts significant challenges instrumentally with the electromagnetic calorimeter (EMC), a typical experimental sensor but recovering the information of incident $\\bar{n}$ insufficiently. In this study, we introduce Vision Calorimeter (ViC), a baseline method for anti-neutron reconstruction that leverages deep learning detectors to analyze the implicit relationships between EMC responses and incident $\\bar{n}$ characteristics. Our motivation lies in that energy distributions of $\\bar{n}$ samples deposited in the EMC cell arrays embody rich contextual information. Converted to 2-D images, such contextual energy distributions can be used to predict the status of $\\bar{n}$ ($i.e.$, incident position and momentum) through a deep learning detector along with pseudo bounding boxes and a specified training objective. Experimental results demonstrate that ViC substantially outperforms the conventional reconstruction approach, reducing the prediction error of incident position by 42.81% (from 17.31$^{\\circ}$ to 9.90$^{\\circ}$). More importantly, this study for the first time realizes the measurement of incident $\\bar{n}$ momentum, underscoring the potential of deep learning detectors for particle reconstruction. Code is available at https://github.com/yuhongtian17/ViC."
                    }
                ]
            },
            "5f950c97-39aa-4d22-acb3-2f8074ead5f6": {
                "pk": "5f950c97-39aa-4d22-acb3-2f8074ead5f6",
                "name": "Lingxi Xie",
                "collaborators": [
                    "Qi Tian",
                    "Alan Yuille",
                    "Jingdong Wang",
                    "Liang Zheng",
                    "Ya Zhang",
                    "Longhui Wei",
                    "Shengjie Xu",
                    "Chenglin Yang",
                    "Siyuan Qiao",
                    "Yingying Zhu"
                ],
                "domain": [
                    "Deep Learning",
                    "Computer Vision",
                    "Neural Architecture Search",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Genetic CNN",
                        "abstract": "The deep Convolutional Neural Network (CNN) is the state-of-the-art solution for large-scale visual recognition. Following basic principles such as increasing the depth and constructing highway connections, researchers have manually designed a lot of fixed network structures and verified their effectiveness.   In this paper, we discuss the possibility of learning deep network structures automatically. Note that the number of possible network structures increases exponentially with the number of layers in the network, which inspires us to adopt the genetic algorithm to efficiently traverse this large search space. We first propose an encoding method to represent each network structure in a fixed-length binary string, and initialize the genetic algorithm by generating a set of randomized individuals. In each generation, we define standard genetic operations, e.g., selection, mutation and crossover, to eliminate weak individuals and then generate more competitive ones. The competitiveness of each individual is defined as its recognition accuracy, which is obtained via training the network from scratch and evaluating it on a validation set. We run the genetic process on two small datasets, i.e., MNIST and CIFAR10, demonstrating its ability to evolve and find high-quality structures which are little studied before. These structures are also transferrable to the large-scale ILSVRC2012 dataset."
                    },
                    {
                        "title": "Dumpling GNN: Hybrid GNN Enables Better ADC Payload Activity Prediction Based on Chemical Structure",
                        "abstract": "Antibody-drug conjugates (ADCs) have emerged as a promising class of targeted cancer therapeutics, but the design and optimization of their cytotoxic payloads remain challenging. This study introduces DumplingGNN, a novel hybrid Graph Neural Network architecture specifically designed for predicting ADC payload activity based on chemical structure. By integrating Message Passing Neural Networks (MPNN), Graph Attention Networks (GAT), and GraphSAGE layers, DumplingGNN effectively captures multi-scale molecular features and leverages both 2D topological and 3D structural information. We evaluate DumplingGNN on a comprehensive ADC payload dataset focusing on DNA Topoisomerase I inhibitors, as well as on multiple public benchmarks from MoleculeNet. DumplingGNN achieves state-of-the-art performance across several datasets, including BBBP (96.4\\% ROC-AUC), ToxCast (78.2\\% ROC-AUC), and PCBA (88.87\\% ROC-AUC). On our specialized ADC payload dataset, it demonstrates exceptional accuracy (91.48\\%), sensitivity (95.08\\%), and specificity (97.54\\%). Ablation studies confirm the synergistic effects of the hybrid architecture and the critical role of 3D structural information in enhancing predictive accuracy. The model's strong interpretability, enabled by attention mechanisms, provides valuable insights into structure-activity relationships. DumplingGNN represents a significant advancement in molecular property prediction, with particular promise for accelerating the design and optimization of ADC payloads in targeted cancer therapy development."
                    },
                    {
                        "title": "Knowledge Distillation in Generations: More Tolerant Teachers Educate Better Students",
                        "abstract": "We focus on the problem of training a deep neural network in generations. The flowchart is that, in order to optimize the target network (student), another network (teacher) with the same architecture is first trained, and used to provide part of supervision signals in the next stage. While this strategy leads to a higher accuracy, many aspects (e.g., why teacher-student optimization helps) still need further explorations.   This paper studies this problem from a perspective of controlling the strictness in training the teacher network. Existing approaches mostly used a hard distribution (e.g., one-hot vectors) in training, leading to a strict teacher which itself has a high accuracy, but we argue that the teacher needs to be more tolerant, although this often implies a lower accuracy. The implementation is very easy, with merely an extra loss term added to the teacher network, facilitating a few secondary classes to emerge and complement to the primary class. Consequently, the teacher provides a milder supervision signal (a less peaked distribution), and makes it possible for the student to learn from inter-class similarity and potentially lower the risk of over-fitting. Experiments are performed on standard image classification tasks (CIFAR100 and ILSVRC2012). Although the teacher network behaves less powerful, the students show a persistent ability growth and eventually achieve higher classification accuracies than other competitors. Model ensemble and transfer feature extraction also verify the effectiveness of our approach."
                    },
                    {
                        "title": "Attention-based Pyramid Aggregation Network for Visual Place Recognition",
                        "abstract": "Visual place recognition is challenging in the urban environment and is usually viewed as a large scale image retrieval task. The intrinsic challenges in place recognition exist that the confusing objects such as cars and trees frequently occur in the complex urban scene, and buildings with repetitive structures may cause over-counting and the burstiness problem degrading the image representations. To address these problems, we present an Attention-based Pyramid Aggregation Network (APANet), which is trained in an end-to-end manner for place recognition. One main component of APANet, the spatial pyramid pooling, can effectively encode the multi-size buildings containing geo-information. The other one, the attention block, is adopted as a region evaluator for suppressing the confusing regional features while highlighting the discriminative ones. When testing, we further propose a simple yet effective PCA power whitening strategy, which significantly improves the widely used PCA whitening by reasonably limiting the impact of over-counting. Experimental evaluations demonstrate that the proposed APANet outperforms the state-of-the-art methods on two place recognition benchmarks, and generalizes well on standard image retrieval datasets."
                    },
                    {
                        "title": "Infinite Curriculum Learning for Efficiently Detecting Gastric Ulcers in WCE Images",
                        "abstract": "The Wireless Capsule Endoscopy (WCE) is becoming a popular way of screening gastrointestinal system diseases and cancer. However, the time-consuming process in inspecting WCE data limits its applications and increases the cost of examinations. This paper considers WCE-based gastric ulcer detection, in which the major challenge is to detect the lesions in a local region. We propose an approach named infinite curriculum learning, which generalizes curriculum learning to an infinite sampling space by approximately measuring the difficulty of each patch by its scale. This allows us to adapt our model from local patches to global images gradually, leading to a consistent accuracy gain. Experiments are performed on a large dataset with more than 3 million WCE images. Our approach achieves a binary classification accuracy of 87%, and is able to detect some lesions mis-annotated by the physicians. In a real-world application, our approach can reduce the workload of a physician by 90%-98% in gastric ulcer screening."
                    },
                    {
                        "title": "Object Recognition with and without Objects",
                        "abstract": "While recent deep neural networks have achieved a promising performance on object recognition, they rely implicitly on the visual contents of the whole image. In this paper, we train deep neural net- works on the foreground (object) and background (context) regions of images respectively. Consider- ing human recognition in the same situations, net- works trained on the pure background without ob- jects achieves highly reasonable recognition performance that beats humans by a large margin if only given context. However, humans still outperform networks with pure object available, which indicates networks and human beings have different mechanisms in understanding an image. Furthermore, we straightforwardly combine multiple trained networks to explore different visual cues learned by different networks. Experiments show that useful visual hints can be explicitly learned separately and then combined to achieve higher performance, which verifies the advantages of the proposed framework."
                    },
                    {
                        "title": "CIPS-3D: A 3D-Aware Generator of GANs Based on Conditionally-Independent Pixel Synthesis",
                        "abstract": "The style-based GAN (StyleGAN) architecture achieved state-of-the-art results for generating high-quality images, but it lacks explicit and precise control over camera poses. The recently proposed NeRF-based GANs made great progress towards 3D-aware generators, but they are unable to generate high-quality images yet. This paper presents CIPS-3D, a style-based, 3D-aware generator that is composed of a shallow NeRF network and a deep implicit neural representation (INR) network. The generator synthesizes each pixel value independently without any spatial convolution or upsampling operation. In addition, we diagnose the problem of mirror symmetry that implies a suboptimal solution and solve it by introducing an auxiliary discriminator. Trained on raw, single-view images, CIPS-3D sets new records for 3D-aware image synthesis with an impressive FID of 6.97 for images at the $256\\times256$ resolution on FFHQ. We also demonstrate several interesting directions for CIPS-3D such as transfer learning and 3D-aware face stylization. The synthesis results are best viewed as videos, so we recommend the readers to check our github project at https://github.com/PeterouZh/CIPS-3D"
                    },
                    {
                        "title": "Creating Something from Nothing: Unsupervised Knowledge Distillation for Cross-Modal Hashing",
                        "abstract": "In recent years, cross-modal hashing (CMH) has attracted increasing attentions, mainly because its potential ability of mapping contents from different modalities, especially in vision and language, into the same space, so that it becomes efficient in cross-modal data retrieval. There are two main frameworks for CMH, differing from each other in whether semantic supervision is required. Compared to the unsupervised methods, the supervised methods often enjoy more accurate results, but require much heavier labors in data annotation. In this paper, we propose a novel approach that enables guiding a supervised method using outputs produced by an unsupervised method. Specifically, we make use of teacher-student optimization for propagating knowledge. Experiments are performed on two popular CMH benchmarks, i.e., the MIRFlickr and NUS-WIDE datasets. Our approach outperforms all existing unsupervised methods by a large margin."
                    },
                    {
                        "title": "Semantic-Aware Knowledge Preservation for Zero-Shot Sketch-Based Image Retrieval",
                        "abstract": "Sketch-based image retrieval (SBIR) is widely recognized as an important vision problem which implies a wide range of real-world applications. Recently, research interests arise in solving this problem under the more realistic and challenging setting of zero-shot learning. In this paper, we investigate this problem from the viewpoint of domain adaptation which we show is critical in improving feature embedding in the zero-shot scenario. Based on a framework which starts with a pre-trained model on ImageNet and fine-tunes it on the training set of SBIR benchmark, we advocate the importance of preserving previously acquired knowledge, e.g., the rich discriminative features learned from ImageNet, to improve the model's transfer ability. For this purpose, we design an approach named Semantic-Aware Knowledge prEservation (SAKE), which fine-tunes the pre-trained model in an economical way and leverages semantic information, e.g., inter-class relationship, to achieve the goal of knowledge preservation. Zero-shot experiments on two extended SBIR datasets, TU-Berlin and Sketchy, verify the superior performance of our approach. Extensive diagnostic experiments validate that knowledge preserved benefits SBIR in zero-shot settings, as a large fraction of the performance gain is from the more properly structured feature embedding for photo images. Code is available at: https://github.com/qliu24/SAKE."
                    },
                    {
                        "title": "Universal-to-Specific Framework for Complex Action Recognition",
                        "abstract": "Video-based action recognition has recently attracted much attention in the field of computer vision. To solve more complex recognition tasks, it has become necessary to distinguish different levels of interclass variations. Inspired by a common flowchart based on the human decision-making process that first narrows down the probable classes and then applies a \"rethinking\" process for finer-level recognition, we propose an effective universal-to-specific (U2S) framework for complex action recognition. The U2S framework is composed of three subnetworks: a universal network, a category-specific network, and a mask network. The universal network first learns universal feature representations. The mask network then generates attention masks for confusing classes through category regularization based on the output of the universal network. The mask is further used to guide the category-specific network for class-specific feature representations. The entire framework is optimized in an end-to-end manner. Experiments on a variety of benchmark datasets, e.g., the Something-Something, UCF101, and HMDB51 datasets, demonstrate the effectiveness of the U2S framework; i.e., U2S can focus on discriminative spatiotemporal regions for confusing categories. We further visualize the relationship between different classes, showing that U2S indeed improves the discriminability of learned features. Moreover, the proposed U2S model is a general framework and may adopt any base recognition network."
                    },
                    {
                        "title": "InterActive: Inter-Layer Activeness Propagation",
                        "abstract": "An increasing number of computer vision tasks can be tackled with deep features, which are the intermediate outputs of a pre-trained Convolutional Neural Network. Despite the astonishing performance, deep features extracted from low-level neurons are still below satisfaction, arguably because they cannot access the spatial context contained in the higher layers. In this paper, we present InterActive, a novel algorithm which computes the activeness of neurons and network connections. Activeness is propagated through a neural network in a top-down manner, carrying high-level context and improving the descriptive power of low-level and mid-level neurons. Visualization indicates that neuron activeness can be interpreted as spatial-weighted neuron responses. We achieve state-of-the-art classification performance on a wide range of image datasets."
                    },
                    {
                        "title": "DisturbLabel: Regularizing CNN on the Loss Layer",
                        "abstract": "During a long period of time we are combating over-fitting in the CNN training process with model regularization, including weight decay, model averaging, data augmentation, etc. In this paper, we present DisturbLabel, an extremely simple algorithm which randomly replaces a part of labels as incorrect values in each iteration. Although it seems weird to intentionally generate incorrect training labels, we show that DisturbLabel prevents the network training from over-fitting by implicitly averaging over exponentially many networks which are trained with different label sets. To the best of our knowledge, DisturbLabel serves as the first work which adds noises on the loss layer. Meanwhile, DisturbLabel cooperates well with Dropout to provide complementary regularization functions. Experiments demonstrate competitive recognition results on several popular image recognition datasets."
                    },
                    {
                        "title": "Geometric Neural Phrase Pooling: Modeling the Spatial Co-occurrence of Neurons",
                        "abstract": "Deep Convolutional Neural Networks (CNNs) are playing important roles in state-of-the-art visual recognition. This paper focuses on modeling the spatial co-occurrence of neuron responses, which is less studied in the previous work. For this, we consider the neurons in the hidden layer as neural words, and construct a set of geometric neural phrases on top of them. The idea that grouping neural words into neural phrases is borrowed from the Bag-of-Visual-Words (BoVW) model. Next, the Geometric Neural Phrase Pooling (GNPP) algorithm is proposed to efficiently encode these neural phrases. GNPP acts as a new type of hidden layer, which punishes the isolated neuron responses after convolution, and can be inserted into a CNN model with little extra computational overhead. Experimental results show that GNPP produces significant and consistent accuracy gain in image classification."
                    },
                    {
                        "title": "Accelerating Deep Neural Networks with Spatial Bottleneck Modules",
                        "abstract": "This paper presents an efficient module named spatial bottleneck for accelerating the convolutional layers in deep neural networks. The core idea is to decompose convolution into two stages, which first reduce the spatial resolution of the feature map, and then restore it to the desired size. This operation decreases the sampling density in the spatial domain, which is independent yet complementary to network acceleration approaches in the channel domain. Using different sampling rates, we can tradeoff between recognition accuracy and model complexity.   As a basic building block, spatial bottleneck can be used to replace any single convolutional layer, or the combination of two convolutional layers. We empirically verify the effectiveness of spatial bottleneck by applying it to the deep residual networks. Spatial bottleneck achieves 2x and 1.4x speedup on the regular and channel-bottlenecked residual blocks, respectively, with the accuracies retained in recognizing low-resolution images, and even improved in recognizing high-resolution images."
                    },
                    {
                        "title": "GOLD-NAS: Gradual, One-Level, Differentiable",
                        "abstract": "There has been a large literature of neural architecture search, but most existing work made use of heuristic rules that largely constrained the search flexibility. In this paper, we first relax these manually designed constraints and enlarge the search space to contain more than $10^{160}$ candidates. In the new space, most existing differentiable search methods can fail dramatically. We then propose a novel algorithm named Gradual One-Level Differentiable Neural Architecture Search (GOLD-NAS) which introduces a variable resource constraint to one-level optimization so that the weak operators are gradually pruned out from the super-network. In standard image classification benchmarks, GOLD-NAS can find a series of Pareto-optimal architectures within a single search procedure. Most of the discovered architectures were never studied before, yet they achieve a nice tradeoff between recognition accuracy and model complexity. We believe the new space and search algorithm can advance the search of differentiable NAS."
                    },
                    {
                        "title": "What Is Considered Complete for Visual Recognition?",
                        "abstract": "This is an opinion paper. We hope to deliver a key message that current visual recognition systems are far from complete, i.e., recognizing everything that human can recognize, yet it is very unlikely that the gap can be bridged by continuously increasing human annotations. Based on the observation, we advocate for a new type of pre-training task named learning-by-compression. The computational models (e.g., a deep network) are optimized to represent the visual data using compact features, and the features preserve the ability to recover the original data. Semantic annotations, when available, play the role of weak supervision. An important yet challenging issue is the evaluation of image recovery, where we suggest some design principles and future research directions. We hope our proposal can inspire the community to pursue the compression-recovery tradeoff rather than the accuracy-complexity tradeoff."
                    },
                    {
                        "title": "Deep Collaborative Learning for Visual Recognition",
                        "abstract": "Deep neural networks are playing an important role in state-of-the-art visual recognition. To represent high-level visual concepts, modern networks are equipped with large convolutional layers, which use a large number of filters and contribute significantly to model complexity. For example, more than half of the weights of AlexNet are stored in the first fully-connected layer (4,096 filters).   We formulate the function of a convolutional layer as learning a large visual vocabulary, and propose an alternative way, namely Deep Collaborative Learning (DCL), to reduce the computational complexity. We replace a convolutional layer with a two-stage DCL module, in which we first construct a couple of smaller convolutional layers individually, and then fuse them at each spatial position to consider feature co-occurrence. In mathematics, DCL can be explained as an efficient way of learning compositional visual concepts, in which the vocabulary size increases exponentially while the model complexity only increases linearly. We evaluate DCL on a wide range of visual recognition tasks, including a series of multi-digit number classification datasets, and some generic image classification datasets such as SVHN, CIFAR and ILSVRC2012. We apply DCL to several state-of-the-art network structures, improving the recognition accuracy meanwhile reducing the number of parameters (16.82% fewer in AlexNet)."
                    },
                    {
                        "title": "Appending Adversarial Frames for Universal Video Attack",
                        "abstract": "There have been many efforts in attacking image classification models with adversarial perturbations, but the same topic on video classification has not yet been thoroughly studied. This paper presents a novel idea of video-based attack, which appends a few dummy frames (e.g., containing the texts of `thanks for watching') to a video clip and then adds adversarial perturbations only on these new frames. Our approach enjoys three major benefits, namely, a high success rate, a low perceptibility, and a strong ability in transferring across different networks. These benefits mostly come from the common dummy frame which pushes all samples towards the boundary of classification. On the other hand, such attacks are easily to be concealed since most people would not notice the abnormality behind the perturbed video clips. We perform experiments on two popular datasets with six state-of-the-art video classification models, and demonstrate the effectiveness of our approach in the scenario of universal video attacks."
                    }
                ]
            },
            "3aa7a9d5-01c2-43e6-987f-5f1a2957559c": {
                "pk": "3aa7a9d5-01c2-43e6-987f-5f1a2957559c",
                "name": "Yaowei Wang",
                "collaborators": [
                    "Yonghong Tian",
                    "Yemin Shi",
                    "Tiejun Huang",
                    "Yahong Han",
                    "Ke Chen",
                    "Kui Jia",
                    "Yating Liu",
                    "Yaowei Li",
                    "Zimo Liu",
                    "Wenming Yang"
                ],
                "domain": [
                    "Computer Vision",
                    "Machine Learning",
                    "Deep Learning",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "CLIP-based Synergistic Knowledge Transfer for Text-based Person Retrieval",
                        "abstract": "Text-based Person Retrieval (TPR) aims to retrieve the target person images given a textual query. The primary challenge lies in bridging the substantial gap between vision and language modalities, especially when dealing with limited large-scale datasets. In this paper, we introduce a CLIP-based Synergistic Knowledge Transfer (CSKT) approach for TPR. Specifically, to explore the CLIP's knowledge on input side, we first propose a Bidirectional Prompts Transferring (BPT) module constructed by text-to-image and image-to-text bidirectional prompts and coupling projections. Secondly, Dual Adapters Transferring (DAT) is designed to transfer knowledge on output side of Multi-Head Attention (MHA) in vision and language. This synergistic two-way collaborative mechanism promotes the early-stage feature fusion and efficiently exploits the existing knowledge of CLIP. CSKT outperforms the state-of-the-art approaches across three benchmark datasets when the training parameters merely account for 7.4% of the entire model, demonstrating its remarkable efficiency, effectiveness and generalization."
                    },
                    {
                        "title": "Sequential Deep Trajectory Descriptor for Action Recognition with Three-stream CNN",
                        "abstract": "Learning the spatial-temporal representation of motion information is crucial to human action recognition. Nevertheless, most of the existing features or descriptors cannot capture motion information effectively, especially for long-term motion. To address this problem, this paper proposes a long-term motion descriptor called sequential Deep Trajectory Descriptor (sDTD). Specifically, we project dense trajectories into two-dimensional planes, and subsequently a CNN-RNN network is employed to learn an effective representation for long-term motion. Unlike the popular two-stream ConvNets, the sDTD stream is introduced into a three-stream framework so as to identify actions from a video sequence. Consequently, this three-stream framework can simultaneously capture static spatial features, short-term motion and long-term motion in the video. Extensive experiments were conducted on three challenging datasets: KTH, HMDB51 and UCF101. Experimental results show that our method achieves state-of-the-art performance on the KTH and UCF101 datasets, and is comparable to the state-of-the-art methods on the HMDB51 dataset."
                    },
                    {
                        "title": "Learning to Share in Multi-Agent Reinforcement Learning",
                        "abstract": "In this paper, we study the problem of networked multi-agent reinforcement learning (MARL), where a number of agents are deployed as a partially connected network and each interacts only with nearby agents. Networked MARL requires all agents to make decisions in a decentralized manner to optimize a global objective with restricted communication between neighbors over the network. Inspired by the fact that sharing plays a key role in human's learning of cooperation, we propose LToS, a hierarchically decentralized MARL framework that enables agents to learn to dynamically share reward with neighbors so as to encourage agents to cooperate on the global objective through collectives. For each agent, the high-level policy learns how to share reward with neighbors to decompose the global objective, while the low-level policy learns to optimize the local objective induced by the high-level policies in the neighborhood. The two policies form a bi-level optimization and learn alternately. We empirically demonstrate that LToS outperforms existing methods in both social dilemma and networked MARL scenarios across scales."
                    },
                    {
                        "title": "Improving Deep Representation Learning via Auxiliary Learnable Target Coding",
                        "abstract": "Deep representation learning is a subfield of machine learning that focuses on learning meaningful and useful representations of data through deep neural networks. However, existing methods for semantic classification typically employ pre-defined target codes such as the one-hot and the Hadamard codes, which can either fail or be less flexible to model inter-class correlation. In light of this, this paper introduces a novel learnable target coding as an auxiliary regularization of deep representation learning, which can not only incorporate latent dependency across classes but also impose geometric properties of target codes into representation space. Specifically, a margin-based triplet loss and a correlation consistency loss on the proposed target codes are designed to encourage more discriminative representations owing to enlarging between-class margins in representation space and favoring equal semantic correlation of learnable target codes respectively. Experimental results on several popular visual classification and retrieval benchmarks can demonstrate the effectiveness of our method on improving representation learning, especially for imbalanced data. Source codes are made publicly available at \\href{https://github.com/AkonLau/LTC}{https://github.com/AkonLau/LTC}."
                    },
                    {
                        "title": "Anomaly Detection with Prototype-Guided Discriminative Latent Embeddings",
                        "abstract": "Recent efforts towards video anomaly detection (VAD) try to learn a deep autoencoder to describe normal event patterns with small reconstruction errors. The video inputs with large reconstruction errors are regarded as anomalies at the test time. However, these methods sometimes reconstruct abnormal inputs well because of the powerful generalization ability of deep autoencoder. To address this problem, we present a novel approach for anomaly detection, which utilizes discriminative prototypes of normal data to reconstruct video frames. In this way, the model will favor the reconstruction of normal events and distort the reconstruction of abnormal events. Specifically, we use a prototype-guided memory module to perform discriminative latent embedding. We introduce a new discriminative criterion for the memory module, as well as a loss function correspondingly, which can encourage memory items to record the representative embeddings of normal data, i.e. prototypes. Besides, we design a novel two-branch autoencoder, which is composed of a future frame prediction network and an RGB difference generation network that share the same encoder. The stacked RGB difference contains motion information just like optical flow, so our model can learn temporal regularity. We evaluate the effectiveness of our method on three benchmark datasets and experimental results demonstrate the proposed method outperforms the state-of-the-art."
                    },
                    {
                        "title": "Unsupervised Domain Adaptation via Distilled Discriminative Clustering",
                        "abstract": "Unsupervised domain adaptation addresses the problem of classifying data in an unlabeled target domain, given labeled source domain data that share a common label space but follow a different distribution. Most of the recent methods take the approach of explicitly aligning feature distributions between the two domains. Differently, motivated by the fundamental assumption for domain adaptability, we re-cast the domain adaptation problem as discriminative clustering of target data, given strong privileged information provided by the closely related, labeled source data. Technically, we use clustering objectives based on a robust variant of entropy minimization that adaptively filters target data, a soft Fisher-like criterion, and additionally the cluster ordering via centroid classification. To distill discriminative source information for target clustering, we propose to jointly train the network using parallel, supervised learning objectives over labeled source data. We term our method of distilled discriminative clustering for domain adaptation as DisClusterDA. We also give geometric intuition that illustrates how constituent objectives of DisClusterDA help learn class-wisely pure, compact feature distributions. We conduct careful ablation studies and extensive experiments on five popular benchmark datasets, including a multi-source domain adaptation one. Based on commonly used backbone networks, DisClusterDA outperforms existing methods on these benchmarks. It is also interesting to observe that in our DisClusterDA framework, adding an additional loss term that explicitly learns to align class-level feature distributions across domains does harm to the adaptation performance, though more careful studies in different algorithmic frameworks are to be conducted."
                    },
                    {
                        "title": "KERM: Knowledge Enhanced Reasoning for Vision-and-Language Navigation",
                        "abstract": "Vision-and-language navigation (VLN) is the task to enable an embodied agent to navigate to a remote location following the natural language instruction in real scenes. Most of the previous approaches utilize the entire features or object-centric features to represent navigable candidates. However, these representations are not efficient enough for an agent to perform actions to arrive the target location. As knowledge provides crucial information which is complementary to visible content, in this paper, we propose a Knowledge Enhanced Reasoning Model (KERM) to leverage knowledge to improve agent navigation ability. Specifically, we first retrieve facts (i.e., knowledge described by language descriptions) for the navigation views based on local regions from the constructed knowledge base. The retrieved facts range from properties of a single object (e.g., color, shape) to relationships between objects (e.g., action, spatial position), providing crucial information for VLN. We further present the KERM which contains the purification, fact-aware interaction, and instruction-guided aggregation modules to integrate visual, history, instruction, and fact features. The proposed KERM can automatically select and gather crucial and relevant cues, obtaining more accurate action prediction. Experimental results on the REVERIE, R2R, and SOON datasets demonstrate the effectiveness of the proposed method."
                    },
                    {
                        "title": "Benign Shortcut for Debiasing: Fair Visual Recognition via Intervention with Shortcut Features",
                        "abstract": "Machine learning models often learn to make predictions that rely on sensitive social attributes like gender and race, which poses significant fairness risks, especially in societal applications, such as hiring, banking, and criminal justice. Existing work tackles this issue by minimizing the employed information about social attributes in models for debiasing. However, the high correlation between target task and these social attributes makes learning on the target task incompatible with debiasing. Given that model bias arises due to the learning of bias features (\\emph{i.e}., gender) that help target task optimization, we explore the following research question: \\emph{Can we leverage shortcut features to replace the role of bias feature in target task optimization for debiasing?} To this end, we propose \\emph{Shortcut Debiasing}, to first transfer the target task's learning of bias attributes from bias features to shortcut features, and then employ causal intervention to eliminate shortcut features during inference. The key idea of \\emph{Shortcut Debiasing} is to design controllable shortcut features to on one hand replace bias features in contributing to the target task during the training stage, and on the other hand be easily removed by intervention during the inference stage. This guarantees the learning of the target task does not hinder the elimination of bias features. We apply \\emph{Shortcut Debiasing} to several benchmark datasets, and achieve significant improvements over the state-of-the-art debiasing methods in both accuracy and fairness."
                    },
                    {
                        "title": "FFCA-Net: Stereo Image Compression via Fast Cascade Alignment of Side Information",
                        "abstract": "Multi-view compression technology, especially Stereo Image Compression (SIC), plays a crucial role in car-mounted cameras and 3D-related applications. Interestingly, the Distributed Source Coding (DSC) theory suggests that efficient data compression of correlated sources can be achieved through independent encoding and joint decoding. This motivates the rapidly developed deep-distributed SIC methods in recent years. However, these approaches neglect the unique characteristics of stereo-imaging tasks and incur high decoding latency. To address this limitation, we propose a Feature-based Fast Cascade Alignment network (FFCA-Net) to fully leverage the side information on the decoder. FFCA adopts a coarse-to-fine cascaded alignment approach. In the initial stage, FFCA utilizes a feature domain patch-matching module based on stereo priors. This module reduces redundancy in the search space of trivial matching methods and further mitigates the introduction of noise. In the subsequent stage, we utilize an hourglass-based sparse stereo refinement network to further align inter-image features with a reduced computational cost. Furthermore, we have devised a lightweight yet high-performance feature fusion network, called a Fast Feature Fusion network (FFF), to decode the aligned features. Experimental results on InStereo2K, KITTI, and Cityscapes datasets demonstrate the significant superiority of our approach over traditional and learning-based SIC methods. In particular, our approach achieves significant gains in terms of 3 to 10-fold faster decoding speed than other methods."
                    },
                    {
                        "title": "Joint Network based Attention for Action Recognition",
                        "abstract": "By extracting spatial and temporal characteristics in one network, the two-stream ConvNets can achieve the state-of-the-art performance in action recognition. However, such a framework typically suffers from the separately processing of spatial and temporal information between the two standalone streams and is hard to capture long-term temporal dependence of an action. More importantly, it is incapable of finding the salient portions of an action, say, the frames that are the most discriminative to identify the action. To address these problems, a \\textbf{j}oint \\textbf{n}etwork based \\textbf{a}ttention (JNA) is proposed in this study. We find that the fully-connected fusion, branch selection and spatial attention mechanism are totally infeasible for action recognition. Thus in our joint network, the spatial and temporal branches share some information during the training stage. We also introduce an attention mechanism on the temporal domain to capture the long-term dependence meanwhile finding the salient portions. Extensive experiments are conducted on two benchmark datasets, UCF101 and HMDB51. Experimental results show that our method can improve the action recognition performance significantly and achieves the state-of-the-art results on both datasets."
                    },
                    {
                        "title": "Deep Transfer Learning for Person Re-identification",
                        "abstract": "Person re-identification (Re-ID) poses a unique challenge to deep learning: how to learn a deep model with millions of parameters on a small training set of few or no labels. In this paper, a number of deep transfer learning models are proposed to address the data sparsity problem. First, a deep network architecture is designed which differs from existing deep Re-ID models in that (a) it is more suitable for transferring representations learned from large image classification datasets, and (b) classification loss and verification loss are combined, each of which adopts a different dropout strategy. Second, a two-stepped fine-tuning strategy is developed to transfer knowledge from auxiliary datasets. Third, given an unlabelled Re-ID dataset, a novel unsupervised deep transfer learning model is developed based on co-training. The proposed models outperform the state-of-the-art deep Re-ID models by large margins: we achieve Rank-1 accuracy of 85.4\\%, 83.7\\% and 56.3\\% on CUHK03, Market1501, and VIPeR respectively, whilst on VIPeR, our unsupervised model (45.1\\%) beats most supervised models."
                    },
                    {
                        "title": "Frequency Spectrum Augmentation Consistency for Domain Adaptive Object Detection",
                        "abstract": "Domain adaptive object detection (DAOD) aims to improve the generalization ability of detectors when the training and test data are from different domains. Considering the significant domain gap, some typical methods, e.g., CycleGAN-based methods, adopt the intermediate domain to bridge the source and target domains progressively. However, the CycleGAN-based intermediate domain lacks the pix- or instance-level supervision for object detection, which leads to semantic differences. To address this problem, in this paper, we introduce a Frequency Spectrum Augmentation Consistency (FSAC) framework with four different low-frequency filter operations. In this way, we can obtain a series of augmented data as the intermediate domain. Concretely, we propose a two-stage optimization framework. In the first stage, we utilize all the original and augmented source data to train an object detector. In the second stage, augmented source and target data with pseudo labels are adopted to perform the self-training for prediction consistency. And a teacher model optimized using Mean Teacher is used to further revise the pseudo labels. In the experiment, we evaluate our method on the single- and compound- target DAOD separately, which demonstrate the effectiveness of our method."
                    },
                    {
                        "title": "Fine-Grained Object Classification via Self-Supervised Pose Alignment",
                        "abstract": "Semantic patterns of fine-grained objects are determined by subtle appearance difference of local parts, which thus inspires a number of part-based methods. However, due to uncontrollable object poses in images, distinctive details carried by local regions can be spatially distributed or even self-occluded, leading to a large variation on object representation. For discounting pose variations, this paper proposes to learn a novel graph based object representation to reveal a global configuration of local parts for self-supervised pose alignment across classes, which is employed as an auxiliary feature regularization on a deep representation learning network.Moreover, a coarse-to-fine supervision together with the proposed pose-insensitive constraint on shallow-to-deep sub-networks encourages discriminative features in a curriculum learning manner. We evaluate our method on three popular fine-grained object classification benchmarks, consistently achieving the state-of-the-art performance. Source codes are available at https://github.com/yangxh11/P2P-Net."
                    },
                    {
                        "title": "Prompt-Driven Dynamic Object-Centric Learning for Single Domain Generalization",
                        "abstract": "Single-domain generalization aims to learn a model from single source domain data to achieve generalized performance on other unseen target domains. Existing works primarily focus on improving the generalization ability of static networks. However, static networks are unable to dynamically adapt to the diverse variations in different image scenes, leading to limited generalization capability. Different scenes exhibit varying levels of complexity, and the complexity of images further varies significantly in cross-domain scenarios. In this paper, we propose a dynamic object-centric perception network based on prompt learning, aiming to adapt to the variations in image complexity. Specifically, we propose an object-centric gating module based on prompt learning to focus attention on the object-centric features guided by the various scene prompts. Then, with the object-centric gating masks, the dynamic selective module dynamically selects highly correlated feature regions in both spatial and channel dimensions enabling the model to adaptively perceive object-centric relevant features, thereby enhancing the generalization capability. Extensive experiments were conducted on single-domain generalization tasks in image classification and object detection. The experimental results demonstrate that our approach outperforms state-of-the-art methods, which validates the effectiveness and generally of our proposed method."
                    },
                    {
                        "title": "Learning long-term dependencies for action recognition with a biologically-inspired deep network",
                        "abstract": "Despite a lot of research efforts devoted in recent years, how to efficiently learn long-term dependencies from sequences still remains a pretty challenging task. As one of the key models for sequence learning, recurrent neural network (RNN) and its variants such as long short term memory (LSTM) and gated recurrent unit (GRU) are still not powerful enough in practice. One possible reason is that they have only feedforward connections, which is different from the biological neural system that is typically composed of both feedforward and feedback connections. To address this problem, this paper proposes a biologically-inspired deep network, called shuttleNet\\footnote{Our code is available at \\url{https://github.com/shiyemin/shuttlenet}}. Technologically, the shuttleNet consists of several processors, each of which is a GRU while associated with multiple groups of cells and states. Unlike traditional RNNs, all processors inside shuttleNet are loop connected to mimic the brain's feedforward and feedback connections, in which they are shared across multiple pathways in the loop connection. Attention mechanism is then employed to select the best information flow pathway. Extensive experiments conducted on two benchmark datasets (i.e UCF101 and HMDB51) show that we can beat state-of-the-art methods by simply embedding shuttleNet into a CNN-RNN framework."
                    },
                    {
                        "title": "P-ODN: Prototype based Open Deep Network for Open Set Recognition",
                        "abstract": "Most of the existing recognition algorithms are proposed for closed set scenarios, where all categories are known beforehand. However, in practice, recognition is essentially an open set problem. There are categories we know called \"knowns\", and there are more we do not know called \"unknowns\". Enumerating all categories beforehand is never possible, consequently it is infeasible to prepare sufficient training samples for those unknowns. Applying closed set recognition methods will naturally lead to unseen-category errors. To address this problem, we propose the prototype based Open Deep Network (P-ODN) for open set recognition tasks. Specifically, we introduce prototype learning into open set recognition. Prototypes and prototype radiuses are trained jointly to guide a CNN network to derive more discriminative features. Then P-ODN detects the unknowns by applying a multi-class triplet thresholding method based on the distance metric between features and prototypes. Manual labeling the unknowns which are detected in the previous process as new categories. Predictors for new categories are added to the classification layer to \"open\" the deep neural networks to incorporate new categories dynamically. The weights of new predictors are initialized exquisitely by applying a distances based algorithm to transfer the learned knowledge. Consequently, this initialization method speed up the fine-tuning process and reduce the samples needed to train new predictors. Extensive experiments show that P-ODN can effectively detect unknowns and needs only few samples with human intervention to recognize a new category. In the real world scenarios, our method achieves state-of-the-art performance on the UCF11, UCF50, UCF101 and HMDB51 datasets."
                    },
                    {
                        "title": "Learning Scalable $\\ell_\\infty$-constrained Near-lossless Image Compression via Joint Lossy Image and Residual Compression",
                        "abstract": "We propose a novel joint lossy image and residual compression framework for learning $\\ell_\\infty$-constrained near-lossless image compression. Specifically, we obtain a lossy reconstruction of the raw image through lossy image compression and uniformly quantize the corresponding residual to satisfy a given tight $\\ell_\\infty$ error bound. Suppose that the error bound is zero, i.e., lossless image compression, we formulate the joint optimization problem of compressing both the lossy image and the original residual in terms of variational auto-encoders and solve it with end-to-end training. To achieve scalable compression with the error bound larger than zero, we derive the probability model of the quantized residual by quantizing the learned probability model of the original residual, instead of training multiple networks. We further correct the bias of the derived probability model caused by the context mismatch between training and inference. Finally, the quantized residual is encoded according to the bias-corrected probability model and is concatenated with the bitstream of the compressed lossy image. Experimental results demonstrate that our near-lossless codec achieves the state-of-the-art performance for lossless and near-lossless image compression, and achieves competitive PSNR while much smaller $\\ell_\\infty$ error compared with lossy image codecs at high bit rates."
                    },
                    {
                        "title": "SgVA-CLIP: Semantic-guided Visual Adapting of Vision-Language Models for Few-shot Image Classification",
                        "abstract": "Although significant progress has been made in few-shot learning, most of existing few-shot image classification methods require supervised pre-training on a large amount of samples of base classes, which limits their generalization ability in real world application. Recently, large-scale Vision-Language Pre-trained models (VLPs) have been gaining increasing attention in few-shot learning because they can provide a new paradigm for transferable visual representation learning with easily available text on the Web. However, the VLPs may neglect detailed visual information that is difficult to describe by language sentences, but important for learning an effective classifier to distinguish different images. To address the above problem, we propose a new framework, named Semantic-guided Visual Adapting (SgVA), which can effectively extend vision-language pre-trained models to produce discriminative adapted visual features by comprehensively using an implicit knowledge distillation, a vision-specific contrastive loss, and a cross-modal contrastive loss. The implicit knowledge distillation is designed to transfer the fine-grained cross-modal knowledge to guide the updating of the vision adapter. State-of-the-art results on 13 datasets demonstrate that the adapted visual features can well complement the cross-modal features to improve few-shot image classification."
                    }
                ]
            },
            "0dbc26e3-5dd9-4a3d-9b28-b077e62c242d": {
                "pk": "0dbc26e3-5dd9-4a3d-9b28-b077e62c242d",
                "name": "Qixiang Ye",
                "collaborators": [
                    "Jianbin Jiao",
                    "Chang Liu",
                    "Yunjie Tian",
                    "Fang Wan",
                    "Yanzhao Zhou",
                    "Mengnan Shi",
                    "Fei Qin",
                    "Zhenjun Han",
                    "Wei Ke",
                    "Zhiliang Peng"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Neural Networks",
                    "Object Detection"
                ],
                "publications": [
                    {
                        "title": "Oriented Response Networks",
                        "abstract": "Deep Convolution Neural Networks (DCNNs) are capable of learning unprecedentedly effective image representations. However, their ability in handling significant local and global image rotations remains limited. In this paper, we propose Active Rotating Filters (ARFs) that actively rotate during convolution and produce feature maps with location and orientation explicitly encoded. An ARF acts as a virtual filter bank containing the filter itself and its multiple unmaterialised rotated versions. During back-propagation, an ARF is collectively updated using errors from all its rotated versions. DCNNs using ARFs, referred to as Oriented Response Networks (ORNs), can produce within-class rotation-invariant deep features while maintaining inter-class discrimination for classification tasks. The oriented response produced by ORNs can also be used for image and object orientation estimation tasks. Over multiple state-of-the-art DCNN architectures, such as VGG, ResNet, and STN, we consistently observe that replacing regular filters with the proposed ARFs leads to significant reduction in the number of network parameters and improvement in classification performance. We report the best results on several commonly used benchmarks."
                    },
                    {
                        "title": "Towards Spatio-Temporal Video Scene Text Detection via Temporal Clustering",
                        "abstract": "With only bounding-box annotations in the spatial domain, existing video scene text detection (VSTD) benchmarks lack temporal relation of text instances among video frames, which hinders the development of video text-related applications. In this paper, we systematically introduce a new large-scale benchmark, named as STVText4, a well-designed spatial-temporal detection metric (STDM), and a novel clustering-based baseline method, referred to as Temporal Clustering (TC). STVText4 opens a challenging yet promising direction of VSTD, termed as ST-VSTD, which targets at simultaneously detecting video scene texts in both spatial and temporal domains. STVText4 contains more than 1.4 million text instances from 161,347 video frames of 106 videos, where each instance is annotated with not only spatial bounding box and temporal range but also four intrinsic attributes, including legibility, density, scale, and lifecycle, to facilitate the community. With continuous propagation of identical texts in the video sequence, TC can accurately output the spatial quadrilateral and temporal range of the texts, which sets a strong baseline for ST-VSTD. Experiments demonstrate the efficacy of our method and the great academic and practical value of the STVText4. The dataset and code will be available soon."
                    },
                    {
                        "title": "Self-supervised Feature-Gate Coupling for Dynamic Network Pruning",
                        "abstract": "Gating modules have been widely explored in dynamic network pruning to reduce the run-time computational cost of deep neural networks while preserving the representation of features. Despite the substantial progress, existing methods remain ignoring the consistency between feature and gate distributions, which may lead to distortion of gated features. In this paper, we propose a feature-gate coupling (FGC) approach aiming to align distributions of features and gates. FGC is a plug-and-play module, which consists of two steps carried out in an iterative self-supervised manner. In the first step, FGC utilizes the $k$-Nearest Neighbor method in the feature space to explore instance neighborhood relationships, which are treated as self-supervisory signals. In the second step, FGC exploits contrastive learning to regularize gating modules with generated self-supervisory signals, leading to the alignment of instance neighborhood relationships within the feature and gate spaces. Experimental results validate that the proposed FGC method improves the baseline approach with significant margins, outperforming the state-of-the-arts with better accuracy-computation trade-off. Code is publicly available."
                    },
                    {
                        "title": "Spatial Transform Decoupling for Oriented Object Detection",
                        "abstract": "Vision Transformers (ViTs) have achieved remarkable success in computer vision tasks. However, their potential in rotation-sensitive scenarios has not been fully explored, and this limitation may be inherently attributed to the lack of spatial invariance in the data-forwarding process. In this study, we present a novel approach, termed Spatial Transform Decoupling (STD), providing a simple-yet-effective solution for oriented object detection with ViTs. Built upon stacked ViT blocks, STD utilizes separate network branches to predict the position, size, and angle of bounding boxes, effectively harnessing the spatial transform potential of ViTs in a divide-and-conquer fashion. Moreover, by aggregating cascaded activation masks (CAMs) computed upon the regressed parameters, STD gradually enhances features within regions of interest (RoIs), which complements the self-attention mechanism. Without bells and whistles, STD achieves state-of-the-art performance on the benchmark datasets including DOTA-v1.0 (82.24% mAP) and HRSC2016 (98.55% mAP), which demonstrates the effectiveness of the proposed method. Source code is available at https://github.com/yuhongtian17/Spatial-Transform-Decoupling."
                    },
                    {
                        "title": "Domain Contrast for Domain Adaptive Object Detection",
                        "abstract": "We present Domain Contrast (DC), a simple yet effective approach inspired by contrastive learning for training domain adaptive detectors. DC is deduced from the error bound minimization perspective of a transferred model, and is implemented with cross-domain contrast loss which is plug-and-play. By minimizing cross-domain contrast loss, DC guarantees the transferability of detectors while naturally alleviating the class imbalance issue in the target domain. DC can be applied at either image level or region level, consistently improving detectors' transferability and discriminability. Extensive experiments on commonly used benchmarks show that DC improves the baseline and state-of-the-art by significant margins, while demonstrating great potential for large domain divergence."
                    },
                    {
                        "title": "A scalable convolutional neural network for task-specified scenarios via knowledge distillation",
                        "abstract": "In this paper, we explore the redundancy in convolutional neural network, which scales with the complexity of vision tasks. Considering that many front-end visual systems are interested in only a limited range of visual targets, the removing of task-specified network redundancy can promote a wide range of potential applications. We propose a task-specified knowledge distillation algorithm to derive a simplified model with pre-set computation cost and minimized accuracy loss, which suits the resource constraint front-end systems well. Experiments on the MNIST and CIFAR10 datasets demonstrate the feasibility of the proposed approach as well as the existence of task-specified redundancy."
                    },
                    {
                        "title": "Video Playback Rate Perception for Self-supervisedSpatio-Temporal Representation Learning",
                        "abstract": "In self-supervised spatio-temporal representation learning, the temporal resolution and long-short term characteristics are not yet fully explored, which limits representation capabilities of learned models. In this paper, we propose a novel self-supervised method, referred to as video Playback Rate Perception (PRP), to learn spatio-temporal representation in a simple-yet-effective way. PRP roots in a dilated sampling strategy, which produces self-supervision signals about video playback rates for representation model learning. PRP is implemented with a feature encoder, a classification module, and a reconstructing decoder, to achieve spatio-temporal semantic retention in a collaborative discrimination-generation manner. The discriminative perception model follows a feature encoder to prefer perceiving low temporal resolution and long-term representation by classifying fast-forward rates. The generative perception model acts as a feature decoder to focus on comprehending high temporal resolution and short-term representation by introducing a motion-attention mechanism. PRP is applied on typical video target tasks including action recognition and video retrieval. Experiments show that PRP outperforms state-of-the-art self-supervised models with significant margins. Code is available at github.com/yuanyao366/PRP"
                    },
                    {
                        "title": "Discretization-Aware Architecture Search",
                        "abstract": "The search cost of neural architecture search (NAS) has been largely reduced by weight-sharing methods. These methods optimize a super-network with all possible edges and operations, and determine the optimal sub-network by discretization, \\textit{i.e.}, pruning off weak candidates. The discretization process, performed on either operations or edges, incurs significant inaccuracy and thus the quality of the final architecture is not guaranteed. This paper presents discretization-aware architecture search (DA\\textsuperscript{2}S), with the core idea being adding a loss term to push the super-network towards the configuration of desired topology, so that the accuracy loss brought by discretization is largely alleviated. Experiments on standard image classification benchmarks demonstrate the superiority of our approach, in particular, under imbalanced target network configurations that were not studied before."
                    },
                    {
                        "title": "SRN: Side-output Residual Network for Object Symmetry Detection in the Wild",
                        "abstract": "In this paper, we establish a baseline for object symmetry detection in complex backgrounds by presenting a new benchmark and an end-to-end deep learning approach, opening up a promising direction for symmetry detection in the wild. The new benchmark, named Sym-PASCAL, spans challenges including object diversity, multi-objects, part-invisibility, and various complex backgrounds that are far beyond those in existing datasets. The proposed symmetry detection approach, named Side-output Residual Network (SRN), leverages output Residual Units (RUs) to fit the errors between the object symmetry groundtruth and the outputs of RUs. By stacking RUs in a deep-to-shallow manner, SRN exploits the 'flow' of errors among multiple scales to ease the problems of fitting complex outputs with limited layers, suppressing the complex backgrounds, and effectively matching object symmetry of different scales. Experimental results validate both the benchmark and its challenging aspects related to realworld images, and the state-of-the-art performance of our symmetry detection approach. The benchmark and the code for SRN are publicly available at https://github.com/KevinKecc/SRN."
                    },
                    {
                        "title": "Linear Span Network for Object Skeleton Detection",
                        "abstract": "Robust object skeleton detection requires to explore rich representative visual features and effective feature fusion strategies. In this paper, we first re-visit the implementation of HED, the essential principle of which can be ideally described with a linear reconstruction model. Hinted by this, we formalize a Linear Span framework, and propose Linear Span Network (LSN) modified by Linear Span Units (LSUs), which minimize the reconstruction error of convolutional network. LSN further utilizes subspace linear span beside the feature linear span to increase the independence of convolutional features and the efficiency of feature integration, which enlarges the capability of fitting complex ground-truth. As a result, LSN can effectively suppress the cluttered backgrounds and reconstruct object skeletons. Experimental results validate the state-of-the-art performance of the proposed LSN."
                    },
                    {
                        "title": "FreeAnchor: Learning to Match Anchors for Visual Object Detection",
                        "abstract": "Modern CNN-based object detectors assign anchors for ground-truth objects under the restriction of object-anchor Intersection-over-Unit (IoU). In this study, we propose a learning-to-match approach to break IoU restriction, allowing objects to match anchors in a flexible manner. Our approach, referred to as FreeAnchor, updates hand-crafted anchor assignment to \"free\" anchor matching by formulating detector training as a maximum likelihood estimation (MLE) procedure. FreeAnchor targets at learning features which best explain a class of objects in terms of both classification and localization. FreeAnchor is implemented by optimizing detection customized likelihood and can be fused with CNN-based detectors in a plug-and-play manner. Experiments on COCO demonstrate that FreeAnchor consistently outperforms their counterparts with significant margins."
                    },
                    {
                        "title": "Adaptive Linear Span Network for Object Skeleton Detection",
                        "abstract": "Conventional networks for object skeleton detection are usually hand-crafted. Although effective, they require intensive priori knowledge to configure representative features for objects in different scale granularity.In this paper, we propose adaptive linear span network (AdaLSN), driven by neural architecture search (NAS), to automatically configure and integrate scale-aware features for object skeleton detection. AdaLSN is formulated with the theory of linear span, which provides one of the earliest explanations for multi-scale deep feature fusion. AdaLSN is materialized by defining a mixed unit-pyramid search space, which goes beyond many existing search spaces using unit-level or pyramid-level features.Within the mixed space, we apply genetic architecture search to jointly optimize unit-level operations and pyramid-level connections for adaptive feature space expansion. AdaLSN substantiates its versatility by achieving significantly higher accuracy and latency trade-off compared with state-of-the-arts. It also demonstrates general applicability to image-to-mask tasks such as edge detection and road extraction. Code is available at \\href{https://github.com/sunsmarterjie/SDL-Skeleton}{\\color{magenta}github.com/sunsmarterjie/SDL-Skeleton}."
                    },
                    {
                        "title": "Cogradient Descent for Dependable Learning",
                        "abstract": "Conventional gradient descent methods compute the gradients for multiple variables through the partial derivative. Treating the coupled variables independently while ignoring the interaction, however, leads to an insufficient optimization for bilinear models. In this paper, we propose a dependable learning based on Cogradient Descent (CoGD) algorithm to address the bilinear optimization problem, providing a systematic way to coordinate the gradients of coupling variables based on a kernelized projection function. CoGD is introduced to solve bilinear problems when one variable is with sparsity constraint, as often occurs in modern learning paradigms. CoGD can also be used to decompose the association of features and weights, which further generalizes our method to better train convolutional neural networks (CNNs) and improve the model capacity. CoGD is applied in representative bilinear problems, including image reconstruction, image inpainting, network pruning and CNN training. Extensive experiments show that CoGD improves the state-of-the-arts by significant margins. Code is available at {https://github.com/bczhangbczhang/CoGD}."
                    },
                    {
                        "title": "A Unified View of Masked Image Modeling",
                        "abstract": "Masked image modeling has demonstrated great potential to eliminate the label-hungry problem of training large-scale vision Transformers, achieving impressive performance on various downstream tasks. In this work, we propose a unified view of masked image modeling after revisiting existing methods. Under the unified view, we introduce a simple yet effective method, termed as MaskDistill, which reconstructs normalized semantic features from teacher models at the masked positions, conditioning on corrupted input images. Experimental results on image classification and semantic segmentation show that MaskDistill achieves comparable or superior performance than state-of-the-art methods. When using the huge vision Transformer and pretraining 300 epochs, MaskDistill obtains 88.3% fine-tuning top-1 accuracy on ImageNet-1k (224 size) and 58.8% semantic segmentation mIoU metric on ADE20k (512 size). The code and pretrained models will be available at https://aka.ms/unimim."
                    },
                    {
                        "title": "Generative Prompt Model for Weakly Supervised Object Localization",
                        "abstract": "Weakly supervised object localization (WSOL) remains challenging when learning object localization models from image category labels. Conventional methods that discriminatively train activation models ignore representative yet less discriminative object parts. In this study, we propose a generative prompt model (GenPromp), defining the first generative pipeline to localize less discriminative object parts by formulating WSOL as a conditional image denoising procedure. During training, GenPromp converts image category labels to learnable prompt embeddings which are fed to a generative model to conditionally recover the input image with noise and learn representative embeddings. During inference, enPromp combines the representative embeddings with discriminative embeddings (queried from an off-the-shelf vision-language model) for both representative and discriminative capacity. The combined embeddings are finally used to generate multi-scale high-quality attention maps, which facilitate localizing full object extent. Experiments on CUB-200-2011 and ILSVRC show that GenPromp respectively outperforms the best discriminative models by 5.2% and 5.6% (Top-1 Loc), setting a solid baseline for WSOL with the generative model. Code is available at https://github.com/callsys/GenPromp."
                    },
                    {
                        "title": "Depth-guided Texture Diffusion for Image Semantic Segmentation",
                        "abstract": "Depth information provides valuable insights into the 3D structure especially the outline of objects, which can be utilized to improve the semantic segmentation tasks. However, a naive fusion of depth information can disrupt feature and compromise accuracy due to the modality gap between the depth and the vision. In this work, we introduce a Depth-guided Texture Diffusion approach that effectively tackles the outlined challenge. Our method extracts low-level features from edges and textures to create a texture image. This image is then selectively diffused across the depth map, enhancing structural information vital for precisely extracting object outlines. By integrating this enriched depth map with the original RGB image into a joint feature embedding, our method effectively bridges the disparity between the depth map and the image, enabling more accurate semantic segmentation. We conduct comprehensive experiments across diverse, commonly-used datasets spanning a wide range of semantic segmentation tasks, including Camouflaged Object Detection (COD), Salient Object Detection (SOD), and indoor semantic segmentation. With source-free estimated depth or depth captured by depth cameras, our method consistently outperforms existing baselines and achieves new state-of-theart results, demonstrating the effectiveness of our Depth-guided Texture Diffusion for image semantic segmentation."
                    },
                    {
                        "title": "Scale Match for Tiny Person Detection",
                        "abstract": "Visual object detection has achieved unprecedented ad-vance with the rise of deep convolutional neural networks.However, detecting tiny objects (for example tiny per-sons less than 20 pixels) in large-scale images remainsnot well investigated. The extremely small objects raisea grand challenge about feature representation while themassive and complex backgrounds aggregate the risk offalse alarms. In this paper, we introduce a new benchmark,referred to as TinyPerson, opening up a promising directionfor tiny object detection in a long distance and with mas-sive backgrounds. We experimentally find that the scale mis-match between the dataset for network pre-training and thedataset for detector learning could deteriorate the featurerepresentation and the detectors. Accordingly, we proposea simple yet effective Scale Match approach to align theobject scales between the two datasets for favorable tiny-object representation. Experiments show the significantperformance gain of our proposed approach over state-of-the-art detectors, and the challenging aspects of TinyPersonrelated to real-world scenarios. The TinyPerson benchmarkand the code for our approach will be publicly available(https://github.com/ucas-vg/TinyBenchmark).(Attention: evaluation rules of AP have updated in benchmark after this paper accepted, So this paper use old rules. we will keep old rules of AP in benchmark, but we recommand the new and we will use the new in latter research.)"
                    },
                    {
                        "title": "BEiT v2: Masked Image Modeling with Vector-Quantized Visual Tokenizers",
                        "abstract": "Masked image modeling (MIM) has demonstrated impressive results in self-supervised representation learning by recovering corrupted image patches. However, most existing studies operate on low-level image pixels, which hinders the exploitation of high-level semantics for representation models. In this work, we propose to use a semantic-rich visual tokenizer as the reconstruction target for masked prediction, providing a systematic way to promote MIM from pixel-level to semantic-level. Specifically, we propose vector-quantized knowledge distillation to train the tokenizer, which discretizes a continuous semantic space to compact codes. We then pretrain vision Transformers by predicting the original visual tokens for the masked image patches. Furthermore, we introduce a patch aggregation strategy which associates discrete image patches to enhance global semantic representation. Experiments on image classification and semantic segmentation show that BEiT v2 outperforms all compared MIM methods. On ImageNet-1K (224 size), the base-size BEiT v2 achieves 85.5% top-1 accuracy for fine-tuning and 80.1% top-1 accuracy for linear probing. The large-size BEiT v2 obtains 87.3% top-1 accuracy for ImageNet-1K (224 size) fine-tuning, and 56.7% mIoU on ADE20K for semantic segmentation. The code and pretrained models are available at https://aka.ms/beitv2."
                    }
                ]
            },
            "204e6abd-b81f-4026-a70e-745652c27f2f": {
                "pk": "204e6abd-b81f-4026-a70e-745652c27f2f",
                "name": "Yunfan Liu",
                "collaborators": [
                    "Zhenan Sun",
                    "Qi Li",
                    "Tieniu Tan",
                    "Hongtian Yu",
                    "Yunjie Tian",
                    "Qixiang Ye",
                    "Hongwen Zhang",
                    "Muchen Li",
                    "Jeffrey Yunfan Liu",
                    "Leonid Sigal"
                ],
                "domain": [
                    "Computer Vision",
                    "Generative Adversarial Networks",
                    "Face Aging",
                    "Image Synthesis"
                ],
                "publications": [
                    {
                        "title": "Attribute-aware Face Aging with Wavelet-based Generative Adversarial Networks",
                        "abstract": "Since it is difficult to collect face images of the same subject over a long range of age span, most existing face aging methods resort to unpaired datasets to learn age mappings. However, the matching ambiguity between young and aged face images inherent to unpaired training data may lead to unnatural changes of facial attributes during the aging process, which could not be solved by only enforcing identity consistency like most existing studies do. In this paper, we propose a attribute-aware face aging model with wavelet-based Generative Adversarial Networks (GANs) to address the above issues. To be specific, we embed facial attribute vectors into both generator and discriminator of the model to encourage each synthesized elderly face image to be faithful to the attribute of its corresponding input. In addition, a wavelet packet transform (WPT) module is incorporated to improve the visual fidelity of generated images by capturing age-related texture details at multiple scales in the frequency space. Qualitative results demonstrate the ability of our model to synthesize visually plausible face images, and extensive quantitative evaluation results show that the proposed method achieves state-of-the-art performance on existing datasets."
                    },
                    {
                        "title": "Age Progression and Regression with Spatial Attention Modules",
                        "abstract": "Age progression and regression refers to aesthetically render-ing a given face image to present effects of face aging and rejuvenation, respectively. Although numerous studies have been conducted in this topic, there are two major problems: 1) multiple models are usually trained to simulate different age mappings, and 2) the photo-realism of generated face images is heavily influenced by the variation of training images in terms of pose, illumination, and background. To address these issues, in this paper, we propose a framework based on conditional Generative Adversarial Networks (cGANs) to achieve age progression and regression simultaneously. Particularly, since face aging and rejuvenation are largely different in terms of image translation patterns, we model these two processes using two separate generators, each dedicated to one age changing process. In addition, we exploit spatial attention mechanisms to limit image modifications to regions closely related to age changes, so that images with high visual fidelity could be synthesized for in-the-wild cases. Experiments on multiple datasets demonstrate the ability of our model in synthesizing lifelike face images at desired ages with personalized features well preserved, and keeping age-irrelevant regions unchanged."
                    },
                    {
                        "title": "A3GAN: An Attribute-aware Attentive Generative Adversarial Network for Face Aging",
                        "abstract": "Face aging, which aims at aesthetically rendering a given face to predict its future appearance, has received significant research attention in recent years. Although great progress has been achieved with the success of Generative Adversarial Networks (GANs) in synthesizing realistic images, most existing GAN-based face aging methods have two main problems: 1) unnatural changes of high-level semantic information (e.g. facial attributes) due to the insufficient utilization of prior knowledge of input faces, and 2) distortions of low-level image content including ghosting artifacts and modifications in age-irrelevant regions. In this paper, we introduce A3GAN, an Attribute-Aware Attentive face aging model to address the above issues. Facial attribute vectors are regarded as the conditional information and embedded into both the generator and discriminator, encouraging synthesized faces to be faithful to attributes of corresponding inputs. To improve the visual fidelity of generation results, we leverage the attention mechanism to restrict modifications to age-related areas and preserve image details. Moreover, the wavelet packet transform is employed to capture textural features at multiple scales in the frequency space. Extensive experimental results demonstrate the effectiveness of our model in synthesizing photorealistic aged face images and achieving state-of-the-art performance on popular face aging datasets."
                    },
                    {
                        "title": "Style Intervention: How to Achieve Spatial Disentanglement with Style-based Generators?",
                        "abstract": "Generative Adversarial Networks (GANs) with style-based generators (e.g. StyleGAN) successfully enable semantic control over image synthesis, and recent studies have also revealed that interpretable image translations could be obtained by modifying the latent code. However, in terms of the low-level image content, traveling in the latent space would lead to `spatially entangled changes' in corresponding images, which is undesirable in many real-world applications where local editing is required. To solve this problem, we analyze properties of the 'style space' and explore the possibility of controlling the local translation with pre-trained style-based generators. Concretely, we propose 'Style Intervention', a lightweight optimization-based algorithm which could adapt to arbitrary input images and render natural translation effects under flexible objectives. We verify the performance of the proposed framework in facial attribute editing on high-resolution images, where both photo-realism and consistency are required. Extensive qualitative results demonstrate the effectiveness of our method, and quantitative measurements also show that the proposed algorithm outperforms state-of-the-art benchmarks in various aspects."
                    },
                    {
                        "title": "Semantic-aware One-shot Face Re-enactment with Dense Correspondence Estimation",
                        "abstract": "One-shot face re-enactment is a challenging task due to the identity mismatch between source and driving faces. Specifically, the suboptimally disentangled identity information of driving subjects would inevitably interfere with the re-enactment results and lead to face shape distortion. To solve this problem, this paper proposes to use 3D Morphable Model (3DMM) for explicit facial semantic decomposition and identity disentanglement. Instead of using 3D coefficients alone for re-enactment control, we take the advantage of the generative ability of 3DMM to render textured face proxies. These proxies contain abundant yet compact geometric and semantic information of human faces, which enable us to compute the face motion field between source and driving images by estimating the dense correspondence. In this way, we could approximate re-enactment results by warping source images according to the motion field, and a Generative Adversarial Network (GAN) is adopted to further improve the visual quality of warping results. Extensive experiments on various datasets demonstrate the advantages of the proposed method over existing start-of-the-art benchmarks in both identity preservation and re-enactment fulfillment."
                    },
                    {
                        "title": "Spatial Transform Decoupling for Oriented Object Detection",
                        "abstract": "Vision Transformers (ViTs) have achieved remarkable success in computer vision tasks. However, their potential in rotation-sensitive scenarios has not been fully explored, and this limitation may be inherently attributed to the lack of spatial invariance in the data-forwarding process. In this study, we present a novel approach, termed Spatial Transform Decoupling (STD), providing a simple-yet-effective solution for oriented object detection with ViTs. Built upon stacked ViT blocks, STD utilizes separate network branches to predict the position, size, and angle of bounding boxes, effectively harnessing the spatial transform potential of ViTs in a divide-and-conquer fashion. Moreover, by aggregating cascaded activation masks (CAMs) computed upon the regressed parameters, STD gradually enhances features within regions of interest (RoIs), which complements the self-attention mechanism. Without bells and whistles, STD achieves state-of-the-art performance on the benchmark datasets including DOTA-v1.0 (82.24% mAP) and HRSC2016 (98.55% mAP), which demonstrates the effectiveness of the proposed method. Source code is available at https://github.com/yuhongtian17/Spatial-Transform-Decoupling."
                    },
                    {
                        "title": "Combining Data-driven and Model-driven Methods for Robust Facial Landmark Detection",
                        "abstract": "Facial landmark detection is an important yet challenging task for real-world computer vision applications. This paper proposes an effective and robust approach for facial landmark detection by combining data- and model-driven methods. Firstly, a Fully Convolutional Network (FCN) is trained to compute response maps of all facial landmark points. Such a data-driven method could make full use of holistic information in a facial image for global estimation of facial landmarks. After that, the maximum points in the response maps are fitted with a pre-trained Point Distribution Model (PDM) to generate the initial facial shape. This model-driven method is able to correct the inaccurate locations of outliers by considering the shape prior information. Finally, a weighted version of Regularized Landmark Mean-Shift (RLMS) is employed to fine-tune the facial shape iteratively. This Estimation-Correction-Tuning process perfectly combines the advantages of the global robustness of data-driven method (FCN), outlier correction capability of model-driven method (PDM) and non-parametric optimization of RLMS. Results of extensive experiments demonstrate that our approach achieves state-of-the-art performances on challenging datasets including 300W, AFLW, AFW and COFW. The proposed method is able to produce satisfying detection results on face images with exaggerated expressions, large head poses, and partial occlusions."
                    },
                    {
                        "title": "GraphPNAS: Learning Distribution of Good Neural Architectures via Deep Graph Generative Models",
                        "abstract": "Neural architectures can be naturally viewed as computational graphs. Motivated by this perspective, we, in this paper, study neural architecture search (NAS) through the lens of learning random graph models. In contrast to existing NAS methods which largely focus on searching for a single best architecture, i.e, point estimation, we propose GraphPNAS a deep graph generative model that learns a distribution of well-performing architectures. Relying on graph neural networks (GNNs), our GraphPNAS can better capture topologies of good neural architectures and relations between operators therein. Moreover, our graph generator leads to a learnable probabilistic search method that is more flexible and efficient than the commonly used RNN generator and random search methods. Finally, we learn our generator via an efficient reinforcement learning formulation for NAS. To assess the effectiveness of our GraphPNAS, we conduct extensive experiments on three search spaces, including the challenging RandWire on TinyImageNet, ENAS on CIFAR10, and NAS-Bench-101/201. The complexity of RandWire is significantly larger than other search spaces in the literature. We show that our proposed graph generator consistently outperforms RNN-based one and achieves better or comparable performances than state-of-the-art NAS methods."
                    },
                    {
                        "title": "Joint Iris Segmentation and Localization Using Deep Multi-task Learning Framework",
                        "abstract": "Iris segmentation and localization in non-cooperative environment is challenging due to illumination variations, long distances, moving subjects and limited user cooperation, etc. Traditional methods often suffer from poor performance when confronted with iris images captured in these conditions. Recent studies have shown that deep learning methods could achieve impressive performance on iris segmentation task. In addition, as iris is defined as an annular region between pupil and sclera, geometric constraints could be imposed to help locating the iris more accurately and improve the segmentation results. In this paper, we propose a deep multi-task learning framework, named as IrisParseNet, to exploit the inherent correlations between pupil, iris and sclera to boost up the performance of iris segmentation and localization in a unified model. In particular, IrisParseNet firstly applies a Fully Convolutional Encoder-Decoder Attention Network to simultaneously estimate pupil center, iris segmentation mask and iris inner/outer boundary. Then, an effective post-processing method is adopted for iris inner/outer circle localization.To train and evaluate the proposed method, we manually label three challenging iris datasets, namely CASIA-Iris-Distance, UBIRIS.v2, and MICHE-I, which cover various types of noises. Extensive experiments are conducted on these newly annotated datasets, and results show that our method outperforms state-of-the-art methods on various benchmarks. All the ground-truth annotations, annotation codes and evaluation protocols are publicly available at https://github.com/xiamenwcy/IrisParseNet."
                    },
                    {
                        "title": "A Unified Framework for Biphasic Facial Age Translation with Noisy-Semantic Guided Generative Adversarial Networks",
                        "abstract": "Biphasic facial age translation aims at predicting the appearance of the input face at any age. Facial age translation has received considerable research attention in the last decade due to its practical value in cross-age face recognition and various entertainment applications. However, most existing methods model age changes between holistic images, regardless of the human face structure and the age-changing patterns of individual facial components. Consequently, the lack of semantic supervision will cause infidelity of generated faces in detail. To this end, we propose a unified framework for biphasic facial age translation with noisy-semantic guided generative adversarial networks. Structurally, we project the class-aware noisy semantic layouts to soft latent maps for the following injection operation on the individual facial parts. In particular, we introduce two sub-networks, ProjectionNet and ConstraintNet. ProjectionNet introduces the low-level structural semantic information with noise map and produces soft latent maps. ConstraintNet disentangles the high-level spatial features to constrain the soft latent maps, which endows more age-related context into the soft latent maps. Specifically, attention mechanism is employed in ConstraintNet for feature disentanglement. Meanwhile, in order to mine the strongest mapping ability of the network, we embed two types of learning strategies in the training procedure, supervised self-driven generation and unsupervised condition-driven cycle-consistent generation. As a result, extensive experiments conducted on MORPH and CACD datasets demonstrate the prominent ability of our proposed method which achieves state-of-the-art performance."
                    }
                ]
            }
        }
    },
    "2408.12489": {
        "paper_data": {
            "title": "Scribbles for All: Benchmarking Scribble Supervised Segmentation Across Datasets",
            "url": "http://arxiv.org/abs/2408.12489v1",
            "arxiv_id": "2408.12489",
            "authors": [
                "Wolfgang Boettcher",
                "Lukas Hoyer",
                "Ozan Unal",
                "Jan Eric Lenssen",
                "Bernt Schiele"
            ],
            "abstract": "In this work, we introduce Scribbles for All, a label and training data generation algorithm for semantic segmentation trained on scribble labels. Training or fine-tuning semantic segmentation models with weak supervision has become an important topic recently and was subject to significant advances in model quality. In this setting, scribbles are a promising label type to achieve high quality segmentation results while requiring a much lower annotation effort than usual pixel-wise dense semantic segmentation annotations. The main limitation of scribbles as source for weak supervision is the lack of challenging datasets for scribble segmentation, which hinders the development of novel methods and conclusive evaluations. To overcome this limitation, Scribbles for All provides scribble labels for several popular segmentation datasets and provides an algorithm to automatically generate scribble labels for any dataset with dense annotations, paving the way for new insights and model advancements in the field of weakly supervised segmentation. In addition to providing datasets and algorithm, we evaluate state-of-the-art segmentation models on our datasets and show that models trained with our synthetic labels perform competitively with respect to models trained on manual labels. Thus, our datasets enable state-of-the-art research into methods for scribble-labeled semantic segmentation. The datasets, scribble generation algorithm, and baselines are publicly available at https://github.com/wbkit/Scribbles4All",
            "introduction": "   1 Introduction  Figure 1: Visual difference in scribble-supervised performance. While predictions from scribble supervised models are almost identical to fully supervised models for PascalVOC, the quality of segmentation for scribble supervised Cityscapes models is visibly poorer (see dotted boxes), highlighting the greater complexity of the dataset and the need for further research.    Semantic segmentation is one of the most crucial tasks for computer vision research and a key component to scene understanding in many applications. While substantial advancements have been made in crafting highly accurate segmentation architectures\u2009[25, 6, 8, 7, 17], those models heavily rely on detailed labels. The crafting of such dense labels constitutes a laborious and resource-intensive process. This limitation impedes the availability of specialized datasets and the practical deployment of vision algorithms in real-world scenarios. It is particularly pronounced in self-driving applications, which demand immense amounts of training data\u2009[2, 32, 9] or in domains with fine-grained classes\u2009[24, 54]. Those settings necessitate models capable of dealing with complex inter-object relations, varying shapes, and scales. Even in the current era of large pre-trained foundation models\u00a0[27, 19], fine-tuning these models on custom, labeled data still remains a necessity for many applications\u2009[48, 4, 42].   One popular approach to addressing the issue of label cost is weakly supervised semantic segmentation\u2009(WSSS)\u2009[3, 49, 43] which made significant progress in recent years. WSSS methods use incomplete labelling in the form of image level labels\u2009[51], bounding boxes\u2009[18, 26, 50] to label objects. Similarly, Sparsely Annotated Semantic Segmentation\u2009(SASS) is defined by labelling a subset of the image pixels through coarse labels\u2009[11, 10], labelling every object through points\u2009[1] or drawing scribbles\u2009[23, 37]. Previous research has shown that especially scribbles are a promising means of attaining cheap yet powerful labels\u2009[23, 40]. While they take only slightly longer for the annotator to label than points, they have been show to enable stronger segmentation results\u2009[45, 21]. Annotating full segmentation images takes several minutes object centric datasets like Pascal\u2009[24] or even hours for more complex driving scenes\u2009[9, 30] while labelling ScribbleSup took on average about 25\u2009seconds. Moreover, state-of-the-art methods gain 7\u201310% mIoU on PascalVOC\u2009[14] for scribble labels over point labels\u2009[31, 45, 21]. This success has lead to a series of promising training methods for SASS with scribbles\u2009[13, 39, 44].   However, currently, there exists only one popular segmentation dataset with scribble labels, namely ScribbleSup, introduced by Lin\u2009et\u2009al.\u2009[23] for the PascalVOC dataset. Two challenges arise for the research area of scribble-supervised segmentation methods. Firstly, generalization of methods to other datasets can not be verified. Secondly, PascalVOC is too easy to serve as the sole benchmark for scribble-supervised methods as visualized in Fig.\u20091. It consists mostly of images with one object class and the background class. By learning precise class boundaries of the dominant background class, a model can already achieve high performance while the challenge of learning object-to-object boundaries is less relevant. In contrast, modern semantic segmentation faces additional challenges such as small object instances (e.g. poles in Cityscapes) or a large number of semantic classes (e.g. 150 classes in ADE20K), which cannot be properly benchmarked with PascalVOC (see Fig.\u00a01).   We present an algorithm that derives scribble labels from fully",
            "references": [
                {
                    "title": "SAM Fewshot Finetuning for Anatomical Segmentation in Medical Images",
                    "abstract": "We propose a straightforward yet highly effective few-shot fine-tuning strategy for adapting the Segment Anything (SAM) to anatomical segmentation tasks in medical images. Our novel approach revolves around reformulating the mask decoder within SAM, leveraging few-shot embeddings derived from a limited set of labeled images (few-shot collection) as prompts for querying anatomical objects captured in image embeddings. This innovative reformulation greatly reduces the need for time-consuming online user interactions for labeling volumetric images, such as exhaustively marking points and bounding boxes to provide prompts slice by slice. With our method, users can manually segment a few 2D slices offline, and the embeddings of these annotated image regions serve as effective prompts for online segmentation tasks. Our method prioritizes the efficiency of the fine-tuning process by exclusively training the mask decoder through caching mechanisms while keeping the image encoder frozen. Importantly, this approach is not limited to volumetric medical images, but can generically be applied to any 2D/3D segmentation task.To thoroughly evaluate our method, we conducted extensive validation on four datasets, covering six anatomical segmentation tasks across two modalities. Furthermore, we conducted a comparative analysis of different prompting options within SAM and the fully-supervised nnU-Net. The results demonstrate the superior performance of our method compared to SAM employing only point prompts (\u223c50% improvement in IoU) and performs on-par with fully supervised methods whilst reducing the requirement of labeled data by at least an order of magnitude."
                },
                {
                    "title": "Adapting Vision Foundation Models for Plant Phenotyping",
                    "abstract": "Foundation models are large models pre-trained on tremendous amount of data. They can be typically adapted to diverse downstream tasks with minimal effort. However, as foundation models are usually pre-trained on images or texts sourced from the Internet, their performance in specialized domains, such as plant phenotyping, comes into question. In addition, fully fine-tuning foundation models is time-consuming and requires high computational power. This paper investigates the efficient adaptation of foundation models for plant phenotyping settings and tasks. We perform extensive experiments on fine-tuning three foundation models, MAE, DINO, and DINOv2 on three essential plant phenotyping tasks: leaf counting, instance segmentation, and disease classification. In particular, the pretrained backbones are kept frozen, while two distinct fine-tuning methods are evaluated, namely adapter tuning (using LoRA) and decoder tuning. The experimental results show that a foundation model can be efficiently adapted to multiple plant phenotyping tasks, yielding similar performance as the state-of-the-art (SoTA) models specifically designed or trained for each task. Despite exhibiting great transferability over different tasks, the fine-tuned foundation models perform slightly worse than the SoTA task-specific models in some scenarios, which requires further investigation."
                },
                {
                    "title": "SAM-OCTA: A Fine-Tuning Strategy for Applying Foundation Model OCTA Image Segmentation Tasks",
                    "abstract": "In the analysis of optical coherence tomography angiography (OCTA) images, the operation of segmenting specific targets is necessary. Existing methods typically train on supervised datasets with limited samples (approximately a few hundred), which can lead to overfitting. To address this, the low-rank adaptation technique is adopted for foundation model fine-tuning and proposed corresponding prompt point generation strategies to process various segmentation tasks on OCTA datasets. This method is named SAM-OCTA and has been experimented on the publicly available OCTA-500 dataset. While achieving state-of-the-art performance metrics, this method accomplishes local vessel segmentation as well as effective artery-vein segmentation, which was not well-solved in previous works. The code is available at: https://github.com/ShellRedia/SAM-OCTA."
                },
                {
                    "title": "Sparsely Annotated Semantic Segmentation with Adaptive Gaussian Mixtures",
                    "abstract": "Sparsely annotated semantic segmentation (SASS) aims to learn a segmentation model by images with sparse labels (i.e., points or scribbles). Existing methods mainly focus on introducing low-level affinity or generating pseudo labels to strengthen supervision, while largely ignoring the inherent relation between labeled and unlabeled pixels. In this paper, we observe that pixels that are close to each other in the feature space are more likely to share the same class. Inspired by this, we propose a novel SASS framework, which is equipped with an Adaptive Gaussian Mixture Model (AGMM). Our AGMM can effectively endow reliable supervision for unlabeled pixels based on the distributions of labeled and unlabeled pixels. Specifically, we first build Gaussian mixtures using labeled pixels and their relatively similar unlabeled pixels, where the labeled pixels act as centroids, for modeling the feature distribution of each class. Then, we leverage the reliable information from labeled pixels and adaptively generated GMM predictions to supervise the training of unlabeled pixels, achieving online, dynamic, and robust selfsupervision. In addition, by capturing category-wise Gaussian mixtures, AGMM encourages the model to learn discriminative class decision boundaries in an end-to-end contrastive learning manner. Experimental results conducted on the PASCAL VOC 2012 and Cityscapes datasets demonstrate that our AGMM can establish new state-of-the-art SASS performance. Code is available at https://github.com/Luffy03/AGMM-SASS"
                },
                {
                    "title": "Weakly-Supervised Domain Adaptive Semantic Segmentation with Prototypical Contrastive Learning",
                    "abstract": "There has been a lot of effort in improving the performance of unsupervised domain adaptation for semantic segmentation task, however, there is still a huge gap in performance when compared with supervised learning. In this work, we propose a common framework to use different weak labels, e.g., image, point and coarse labels from the target domain to reduce this performance gap. Specifically, we propose to learn better prototypes that are representative class features by exploiting these weak labels. We use these improved prototypes for the contrastive alignment of class features. In particular, we perform two different feature alignments: first, we align pixel features with proto-types within each domain and second, we align pixel features from the source to prototype of target domain in an asymmetric way. This asymmetric alignment is beneficial as it preserves the target features during training, which is essential when weak labels are available from the target domain. Our experiments on various benchmarks show that our framework achieves significant improvement compared to existing works and can reduce the performance gap with supervised learning. Code will be available at https://github.com/anurag-198/WDASS."
                },
                {
                    "title": "DINOv2: Learning Robust Visual Features without Supervision",
                    "abstract": "The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels."
                },
                {
                    "title": "Segment Anything",
                    "abstract": "We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive \u2013 often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images at segment-anything.com to foster research into foundation models for computer vision. We recommend reading the full paper at: arxiv.org/abs/2304.02643."
                },
                {
                    "title": "Urban Scene Semantic Segmentation with Low-Cost Coarse Annotation",
                    "abstract": "For best performance, today\u2019s semantic segmentation methods use large and carefully labeled datasets, requiring expensive annotation budgets. In this work, we show that coarse annotation is a low-cost but highly effective alternative for training semantic segmentation models. Considering the urban scene segmentation scenario, we lever-age cheap coarse annotations for real-world captured data, as well as synthetic data to train our model and show competitive performance compared with finely annotated real-world data. Specifically, we propose a coarse-to-fine self-training framework that generates pseudo labels for unlabeled regions of the coarsely annotated data, using synthetic data to improve predictions around the boundaries between semantic classes, and using cross-domain data augmentation to increase diversity. Our extensive experimental results on Cityscapes and BDD100k datasets demonstrate that our method achieves a significantly better performance vs annotation cost tradeoff, yielding a comparable performance to fully annotated data with only a small fraction of the annotation budget. Also, when used as pre-training, our framework performs better compared to the standard fully supervised setting."
                },
                {
                    "title": "SASFormer: Transformers for Sparsely Annotated Semantic Segmentation",
                    "abstract": "Semantic segmentation based on sparse annotation has advanced in recent years. It labels only part of each object in the image, leaving the remainder unlabeled. Most of the existing approaches are time-consuming and often necessitate a multi-stage training strategy. In this work, we propose a simple yet effective sparse annotated semantic segmentation framework based on segformer, dubbed SASFormer, that achieves remarkable performance. Specifically, the framework first generates hierarchical patch attention maps, which are then multiplied by the network predictions to produce correlated regions separated by valid labels. Besides, we also introduce the affinity loss to ensure consistency between the features of correlation results and network predictions. Extensive experiments showcase that our proposed approach is superior to existing methods and achieves cutting-edge performance. The source code is available at https://github.com/su-hui-zz/SASFormer."
                },
                {
                    "title": "OneFormer: One Transformer to Rule Universal Image Segmentation",
                    "abstract": "Universal Image Segmentation is not a new concept. Past attempts to unify image segmentation include scene parsing, panoptic segmentation, and, more recently, new panoptic architectures. However, such panoptic architectures do not truly unify image segmentation because they need to be trained individually on the semantic, instance, or panoptic segmentation to achieve the best performance. Ideally, a truly universal framework should be trained only once and achieve SOTA performance across all three image segmentation tasks. To that end, we propose OneFormer, a universal image segmentation framework that unifies segmentation with a multi-task train-once design. We first propose a task-conditioned joint training strategy that enables training on ground truths of each domain (semantic, instance, and panoptic segmentation) within a single multi-task training process. Secondly, we introduce a task token to condition our model on the task at hand, making our model task-dynamic to support multi-task training and inference. Thirdly, we propose using a query-text contrastive loss during training to establish better intertask and inter-class distinctions. Notably, our single OneFormer model outperforms specialized Mask2Former models across all three segmentation tasks on ADE20k, Cityscapes, and COCO, despite the latter being trained on each task individually. We believe OneFormer is a significant step towards making image segmentation more universal and accessible."
                },
                {
                    "title": "Cycle-Consistent Learning for Weakly Supervised Semantic Segmentation",
                    "abstract": "We investigate a principle way to accomplish the weakly supervised semantic segmentation, only using scribbles as supervision. The key challenge of this task lies in how to accurately propagate the semantic labels from the annotated scribbles to those unlabeled regions so that accurate pseudo masks can be harvested to learn better segmentation models. To tackle this issue, we propose a simple, strong, and unified framework named Cycle-Consistent Learning (CCL) in this work. To be specific, our CCL first utilizes the given scribbles for training and makes a prediction for those unlabeled regions. Then, the predicted regions, in turn, serve as supervision for learning to predict the labeled scribbles. With such a cycle-consistent constraint, the accurate scribbles can reversely help ease those potential noises existing in the unlabeled regions, resulting in better pseudo masks. The training process of our CCL is looped until the network converges in an end-to-end way. We conduct extensive experiments on the popular PASCAL VOC benchmark and achieve a comparable result with the state-of-the-art method. The training mechanism of the CCL is straightforward and can be easily embedded into any future weakly supervised semantic segmentation approach."
                },
                {
                    "title": "Vision Transformer Adapter for Dense Predictions",
                    "abstract": "This work investigates a simple yet powerful dense prediction task adapter for Vision Transformer (ViT). Unlike recently advanced variants that incorporate vision-specific inductive biases into their architectures, the plain ViT suffers inferior performance on dense predictions due to weak prior assumptions. To address this issue, we propose the ViT-Adapter, which allows plain ViT to achieve comparable performance to vision-specific transformers. Specifically, the backbone in our framework is a plain ViT that can learn powerful representations from large-scale multi-modal data. When transferring to downstream tasks, a pre-training-free adapter is used to introduce the image-related inductive biases into the model, making it suitable for these tasks. We verify ViT-Adapter on multiple dense prediction tasks, including object detection, instance segmentation, and semantic segmentation. Notably, without using extra detection data, our ViT-Adapter-L yields state-of-the-art 60.9 box AP and 53.0 mask AP on COCO test-dev. We hope that the ViT-Adapter could serve as an alternative for vision-specific transformers and facilitate future research. The code and models will be released at https://github.com/czczup/ViT-Adapter."
                },
                {
                    "title": "Tree Energy Loss: Towards Sparsely Annotated Semantic Segmentation",
                    "abstract": "Sparsely annotated semantic segmentation (SASS) aims to train a segmentation network with coarse-grained (i.e., point-, scribble-, and block-wise) supervisions, where only a small proportion of pixels are labeled in each image. In this paper, we propose a novel tree energy loss for SASS by providing semantic guidance for unlabeled pixels. The tree energy loss represents images as minimum spanning trees to model both low-level and high-level pair-wise affini-ties. By sequentially applying these affinities to the net-work prediction, soft pseudo labels for unlabeled pixels are generated in a coarse-to-fine manner, achieving dynamic online self-training. The tree energy loss is effective and easy to be incorporated into existing frameworks by com-bining it with a traditional segmentation loss. Compared with previous SASS methods, our method requires no multi-stage training strategies, alternating optimization proce-dures, additional supervised data, or time-consuming post-processing while outperforming them in all SASS settings. Code is available at https://github.com/megvii-research/TreeEnergyLoss."
                },
                {
                    "title": "Scribble-Supervised LiDAR Semantic Segmentation",
                    "abstract": "Densely annotating LiDAR point clouds remains too expensive and time-consuming to keep up with the ever growing volume of data. While current literature focuses on fully-supervised performance, developing efficient methods that take advantage of realistic weak supervision have yet to be explored. In this paper, we propose using scribbles to annotate LiDAR point clouds and release ScribbleKITTI, the first scribble-annotated dataset for LiDAR semantic segmentation. Furthermore, we present a pipeline to reduce the performance gap that arises when using such weak annotations. Our pipeline comprises of three stand-alone contributions that can be combined with any LiDAR semantic segmentation model to achieve up to 95.7% of the fully-supervised performance while using only 8% labeled points. Our scribble annotations and code are available at github.com/ouenal/scribblekitti."
                },
                {
                    "title": "Learning Affinity from Attention: End-to-End Weakly-Supervised Semantic Segmentation with Transformers",
                    "abstract": "Weakly-supervised semantic segmentation (WSSS) with image-level labels is an important and challenging task. Due to the high training efficiency, end-to-end solutions for WSSS have received increasing attention from the community. However, current methods are mainly based on convolutional neural networks and fail to explore the global information properly, thus usually resulting in incomplete object regions. In this paper, to address the aforementioned problem, we introduce Transformers, which naturally integrate global information, to generate more integral initial pseudo labels for end-to-end WSSS. Motivated by the inherent consistency between the self-attention in Transformers and the semantic affinity, we propose an Affinity from Attention (AFA) module to learn semantic affinity from the multi-head self-attention (MHSA) in Transformers. The learned affinity is then leveraged to refine the initial pseudo labels for segmentation. In addition, to efficiently derive reliable affinity labels for supervising AFA and ensure the local consistency of pseudo labels, we devise a Pixel-Adaptive Refinement module that incorporates low-level image appearance information to refine the pseudo labels. We perform extensive experiments and our method achieves 66.0% and 38.9% mIoU on the PASCAL VOC 2012 and MS COCO 2014 datasets, respectively, significantly outperforming recent end-to-end methods and several multi-stage competitors. Code is available at https://github.com/rulixiang/afa."
                },
                {
                    "title": "Weakly Supervised Semantic Segmentation via Alternate Self-Dual Teaching.",
                    "abstract": "Weakly supervised semantic segmentation (WSSS) is a challenging yet important research field in vision community. In WSSS, the key problem is to generate high-quality pseudo segmentation masks (PSMs). Existing approaches mainly depend on the discriminative object part to generate PSMs, which would inevitably miss object parts or involve surrounding image background, as the learning process is unaware of the full object structure. In fact, both the discriminative object part and the full object structure are critical for deriving of high-quality PSMs. To fully explore these two information cues, we build a novel end-to-end learning framework, alternate self-dual teaching (ASDT), based on a dual-teacher single-student network architecture. The information interaction among different network branches is formulated in the form of knowledge distillation (KD). Unlike the conventional KD, the knowledge of the two teacher models would inevitably be noisy under weak supervision. Inspired by the Pulse Width (PW) modulation, we introduce a PW wave-like selection signal to alleviate the influence of the imperfect knowledge from either teacher model on the KD process. Comprehensive experiments on the PASCAL VOC 2012 and COCO-Stuff 10K demonstrate the effectiveness of the proposed ASDT framework, and new state-of-the-art results are achieved."
                },
                {
                    "title": "Incremental Learning in Semantic Segmentation from Image Labels",
                    "abstract": "Although existing semantic segmentation approaches achieve impressive results, they still struggle to update their models incrementally as new categories are uncovered. Furthermore, pixel-by-pixel annotations are expensive and time-consuming. This paper proposes a novel framework for Weakly Incremental Learning for Semantic Segmentation, that aims at learning to segment new classes from cheap and largely available image-level labels. As opposed to existing approaches, that need to generate pseudolabels offline, we use a localizer, trained with image-level labels and regularized by the segmentation model, to obtain pseudo-supervision online and update the model incrementally. We cope with the inherent noise in the process by using soft-labels generated by the localizer. We demonstrate the effectiveness of our approach on the Pascal VOC and COCO datasets, outperforming offline weakly-supervised methods and obtaining results comparable with incremental learning methods with full supervision.11Code can be found at https://github.com/fcd194/WILSON."
                },
                {
                    "title": "Masked-attention Mask Transformer for Universal Image Segmentation",
                    "abstract": "Image segmentation groups pixels with different semantics, e.g., category or instance membership. Each choice of semantics defines a task. While only the semantics of each task differ, current research focuses on designing spe-cialized architectures for each task. We present Masked- attention Mask Transformer (Mask2Former), a new archi-tecture capable of addressing any image segmentation task (panoptic, instance or semantic). Its key components in-clude masked attention, which extracts localized features by constraining cross-attention within predicted mask regions. In addition to reducing the research effort by at least three times, it outperforms the best specialized architectures by a significant margin on four popular datasets. Most no-tably, Mask2Former sets a new state-of-the-art for panoptic segmentation (57.8 PQ on COCO), instance segmentation (50.1 AP on COCO) and semantic segmentation (57.7 mIoU onADE20K)."
                },
                {
                    "title": "ACDC: The Adverse Conditions Dataset with Correspondences for Semantic Driving Scene Understanding",
                    "abstract": "Level 5 autonomy for self-driving cars requires a robust visual perception system that can parse input images under any visual condition. However, existing semantic segmentation datasets are either dominated by images captured under normal conditions or are small in scale. To address this, we introduce ACDC, the Adverse Conditions Dataset with Correspondences for training and testing semantic segmentation methods on adverse visual conditions. ACDC consists of a large set of 4006 images which are equally distributed between four common adverse conditions: fog, nighttime, rain, and snow. Each adverse-condition image comes with a high-quality fine pixel-level semantic annotation, a corresponding image of the same scene taken under normal conditions, and a binary mask that distinguishes between intra-image regions of clear and uncertain semantic content. Thus, ACDC supports both standard semantic segmentation and the newly introduced uncertainty-aware semantic segmentation. A detailed empirical study demonstrates the challenges that the adverse domains of ACDC pose to state-of-the-art supervised and unsupervised approaches and indicates the value of our dataset in steering future progress in the field. Our dataset and benchmark are publicly available."
                },
                {
                    "title": "KITTI-360: A Novel Dataset and Benchmarks for Urban Scene Understanding in 2D and 3D",
                    "abstract": "For the last few decades, several major subfields of artificial intelligence including computer vision, graphics, and robotics have progressed largely independently from each other. Recently, however, the community has realized that progress towards robust intelligent systems such as self-driving cars requires a concerted effort across the different fields. This motivated us to develop KITTI-360, successor of the popular KITTI dataset. KITTI-360 is a suburban driving dataset which comprises richer input modalities, comprehensive semantic instance annotations and accurate localization to facilitate research at the intersection of vision, graphics and robotics. For efficient annotation, we created a tool to label 3D scenes with bounding primitives and developed a model that transfers this information into the 2D image domain, resulting in over 150k images and 1B 3D points with coherent semantic instance annotations across 2D and 3D. Moreover, we established benchmarks and baselines for several tasks relevant to mobile perception, encompassing problems from computer vision, graphics, and robotics on the same dataset, e.g., semantic scene understanding, novel view synthesis and semantic SLAM. KITTI-360 will enable progress at the intersection of these research areas and thus contribute towards solving one of today's grand challenges: the development of fully autonomous self-driving systems."
                },
                {
                    "title": "Learning or Modelling? An Analysis of Single Image Segmentation Based on Scribble Information",
                    "abstract": "Single image segmentation based on scribbles is an important technique in several applications, e.g. for image editing software. In this paper, we investigate the scope of single image segmentation solely given the image and scribble information using both convolutional neural networks as well as classical model-based methods, and present three main findings: 1) Despite the success of deep learning in the semantic analysis of images, networks fail to outperform model-based approaches in the case of learning on a single image only. Even using a pretrained network for transfer learning does not yield faithful segmentations. 2) The best way to utilize an annotated data set is by exploiting a model-based approach that combines semantic features of a pretrained network with the RGB information, and 3) allowing the networks prediction to change spatially and additionally enforce this variation to be smooth via a gradient-based regularization term on the loss (double backpropagation) is the most successful strategy for pure single image learning-based segmentation."
                },
                {
                    "title": "Leveraging Auxiliary Tasks with Affinity Learning for Weakly Supervised Semantic Segmentation",
                    "abstract": "Semantic segmentation is a challenging task in the absence of densely labelled data. Only relying on class activation maps (CAM) with image-level labels provides deficient segmentation supervision. Prior works thus consider pre-trained models to produce coarse saliency maps to guide the generation of pseudo segmentation labels. However, the commonly used off-line heuristic generation process cannot fully exploit the benefits of these coarse saliency maps. Motivated by the significant inter-task correlation, we propose a novel weakly supervised multi-task framework termed as AuxSegNet, to leverage saliency detection and multi-label image classification as auxiliary tasks to improve the primary task of semantic segmentation using only image-level ground-truth labels. Inspired by their similar structured semantics, we also propose to learn a cross-task global pixellevel affinity map from the saliency and segmentation representations. The learned cross-task affinity can be used to refine saliency predictions and propagate CAM maps to provide improved pseudo labels for both tasks. The mutual boost between pseudo label updating and cross-task affinity learning enables iterative improvements on segmentation performance. Extensive experiments demonstrate the effectiveness of the proposed auxiliary learning network structure and the cross-task affinity learning method. The proposed approach achieves state-of-the-art weakly supervised segmentation performance on the challenging PASCAL VOC 2012 and MS COCO benchmarks. 1"
                },
                {
                    "title": "Per-Pixel Classification is Not All You Need for Semantic Segmentation",
                    "abstract": "Modern approaches typically formulate semantic segmentation as a per-pixel classification task, while instance-level segmentation is handled with an alternative mask classification. Our key insight: mask classification is sufficiently general to solve both semantic- and instance-level segmentation tasks in a unified manner using the exact same model, loss, and training procedure. Following this observation, we propose MaskFormer, a simple mask classification model which predicts a set of binary masks, each associated with a single global class label prediction. Overall, the proposed mask classification-based method simplifies the landscape of effective approaches to semantic and panoptic segmentation tasks and shows excellent empirical results. In particular, we observe that MaskFormer outperforms per-pixel classification baselines when the number of classes is large. Our mask classification-based method outperforms both current state-of-the-art semantic (55.6 mIoU on ADE20K) and panoptic segmentation (52.7 PQ on COCO) models."
                },
                {
                    "title": "SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers",
                    "abstract": "We present SegFormer, a simple, efficient yet powerful semantic segmentation framework which unifies Transformers with lightweight multilayer perception (MLP) decoders. SegFormer has two appealing features: 1) SegFormer comprises a novel hierarchically structured Transformer encoder which outputs multiscale features. It does not need positional encoding, thereby avoiding the interpolation of positional codes which leads to decreased performance when the testing resolution differs from training. 2) SegFormer avoids complex decoders. The proposed MLP decoder aggregates information from different layers, and thus combining both local attention and global attention to render powerful representations. We show that this simple and lightweight design is the key to efficient segmentation on Transformers. We scale our approach up to obtain a series of models from SegFormer-B0 to SegFormer-B5, reaching significantly better performance and efficiency than previous counterparts. For example, SegFormer-B4 achieves 50.3% mIoU on ADE20K with 64M parameters, being 5x smaller and 2.2% better than the previous best method. Our best model, SegFormer-B5, achieves 84.0% mIoU on Cityscapes validation set and shows excellent zero-shot robustness on Cityscapes-C. Code will be released at: github.com/NVlabs/SegFormer."
                },
                {
                    "title": "Affinity Attention Graph Neural Network for Weakly Supervised Semantic Segmentation",
                    "abstract": "Weakly supervised semantic segmentation is receiving great attention due to its low human annotation cost. In this paper, we aim to tackle bounding box supervised semantic segmentation, i.e., training accurate semantic segmentation models using bounding box annotations as supervision. To this end, we propose affinity attention graph neural network (<italic><inline-formula><tex-math notation=\"LaTeX\">$A^2$</tex-math><alternatives><mml:math><mml:msup><mml:mi>A</mml:mi><mml:mn>2</mml:mn></mml:msup></mml:math><inline-graphic xlink:href=\"xiao-ieq1-3083269.gif\"/></alternatives></inline-formula>GNN</italic>). Following previous practices, we first generate pseudo semantic-aware seeds, which are then formed into semantic graphs based on our newly proposed affinity Convolutional Neural Network (CNN). Then the built graphs are input to our <italic><inline-formula><tex-math notation=\"LaTeX\">$A^2$</tex-math><alternatives><mml:math><mml:msup><mml:mi>A</mml:mi><mml:mn>2</mml:mn></mml:msup></mml:math><inline-graphic xlink:href=\"xiao-ieq2-3083269.gif\"/></alternatives></inline-formula>GNN</italic>, in which an affinity attention layer is designed to acquire the short- and long- distance information from soft graph edges to accurately propagate semantic labels from the confident seeds to the unlabeled pixels. However, to guarantee the precision of the seeds, we only adopt a limited number of confident pixel seed labels for <italic><inline-formula><tex-math notation=\"LaTeX\">$A^2$</tex-math><alternatives><mml:math><mml:msup><mml:mi>A</mml:mi><mml:mn>2</mml:mn></mml:msup></mml:math><inline-graphic xlink:href=\"xiao-ieq3-3083269.gif\"/></alternatives></inline-formula>GNN</italic>, which may lead to insufficient supervision for training. To alleviate this issue, we further introduce a new loss function and a consistency-checking mechanism to leverage the bounding box constraint, so that more reliable guidance can be included for the model optimization. Experiments show that our approach achieves new state-of-the-art performances on Pascal VOC 2012 datasets (<italic>val</italic>: 76.5 percent, <italic>test</italic>: 75.2 percent). More importantly, our approach can be readily applied to bounding box supervised instance segmentation task or other weakly supervised semantic segmentation tasks, with state-of-the-art or comparable performance among almot all weakly supervised tasks on PASCAL VOC or COCO dataset. Our source code will be available at <uri>https://github.com/zbf1991/A2GNN</uri>."
                },
                {
                    "title": "Background-Aware Pooling and Noise-Aware Loss for Weakly-Supervised Semantic Segmentation",
                    "abstract": "We address the problem of weakly-supervised semantic segmentation (WSSS) using bounding box annotations. Although object bounding boxes are good indicators to segment corresponding objects, they do not specify object boundaries, making it hard to train convolutional neural networks (CNNs) for semantic segmentation. We find that background regions are perceptually consistent in part within an image, and this can be leveraged to discriminate foreground and background regions inside object bounding boxes. To implement this idea, we propose a novel pooling method, dubbed background-aware pooling (BAP), that focuses more on aggregating foreground features inside the bounding boxes using attention maps. This allows to extract high-quality pseudo segmentation labels to train CNNs for semantic segmentation, but the labels still contain noise especially at object boundaries. To address this problem, we also introduce a noise-aware loss (NAL) that makes the networks less susceptible to incorrect labels. Experimental results demonstrate that learning with our pseudo labels already outperforms state-of-the-art weakly- and semi-supervised methods on the PASCAL VOC 2012 dataset, and the NAL further boosts the performance."
                },
                {
                    "title": "Learning to Segment From Scribbles Using Multi-Scale Adversarial Attention Gates",
                    "abstract": "Large, fine-grained image segmentation datasets, annotated at pixel-level, are difficult to obtain, particularly in medical imaging, where annotations also require expert knowledge. Weakly-supervised learning can train models by relying on weaker forms of annotation, such as scribbles. Here, we learn to segment using scribble annotations in an adversarial game. With unpaired segmentation masks, we train a multi-scale GAN to generate realistic segmentation masks at multiple resolutions, while we use scribbles to learn their correct position in the image. Central to the model\u2019s success is a novel attention gating mechanism, which we condition with adversarial signals to act as a shape prior, resulting in better object localization at multiple scales. Subject to adversarial conditioning, the segmentor learns attention maps that are semantic, suppress the noisy activations outside the objects, and reduce the vanishing gradient problem in the deeper layers of the segmentor. We evaluated our model on several medical (ACDC, LVSC, CHAOS) and non-medical (PPSS) datasets, and we report performance levels matching those achieved by models trained with fully annotated segmentation masks. We also demonstrate extensions in a variety of settings: semi-supervised learning; combining multiple scribble sources (a crowdsourcing scenario) and multi-task learning (combining scribble and mask supervision). We release expert-made scribble annotations for the ACDC dataset, and the code used for the experiments, at https://vios-s.github.io/multiscale-adversarial-attention-gates."
                },
                {
                    "title": "Scribble-Supervised Semantic Segmentation by Uncertainty Reduction on Neural Representation and Self-Supervision on Neural Eigenspace",
                    "abstract": "Scribble-supervised semantic segmentation has gained much attention recently for its promising performance without high-quality annotations. Due to the lack of supervision, confident and consistent predictions are usually hard to obtain. Typically, people handle these problems by either adopting an auxiliary task with the well-labeled dataset or incorporating a graphical model with additional requirements on scribble annotations. Instead, this work aims to achieve semantic segmentation by scribble annotations directly without extra information and other limitations. Specifically, we propose holistic operations, including minimizing entropy and a network embedded random walk on the neural representation to reduce uncertainty. Given the probabilistic transition matrix of a random walk, we further train the network with self-supervision on its neural eigenspace to impose consistency on predictions between related images. Comprehensive experiments and ablation studies verify the proposed approach, which demonstrates superiority over others; it is even comparable to some full-label supervised ones and works well when scribbles are randomly shrunk or dropped."
                },
                {
                    "title": "Scribble-Based Weakly Supervised Deep Learning for Road Surface Extraction From Remote Sensing Images",
                    "abstract": "Road surface extraction from remote sensing images using deep learning methods has achieved good performance, while most of the existing methods are based on fully supervised learning, which requires a large amount of training data with laborious per-pixel annotation. In this article, we propose a scribble-based weakly supervised road surface extraction method named ScRoadExtractor, which learns from easily accessible scribbles such as centerlines instead of densely annotated road surface ground truths. To propagate semantic information from sparse scribbles to unlabeled pixels, we introduce a road label propagation algorithm, which considers both the buffer-based properties of road networks and the color and spatial information of super-pixels, to produce a proposal mask with categories road, nonroad, and unknown. The proposal mask, along with the auxiliary boundary prior information detected from images, is utilized to train a dual-branch encoder\u2013decoder network which we designed for precise road surface segmentation. We perform experiments on three diverse road data sets that are comprised of high-resolution remote sensing satellite and aerial images across the world. The results demonstrate that ScRoadExtractor exceeds the classic scribble-supervised segmentation method by 20% for the intersection over union (IoU) indicator and outperforms the state-of-the-art scribble-based weakly supervised methods at least 4%."
                },
                {
                    "title": "Self-Supervised Equivariant Attention Mechanism for Weakly Supervised Semantic Segmentation",
                    "abstract": "Image-level weakly supervised semantic segmentation is a challenging problem that has been deeply studied in recent years. Most of advanced solutions exploit class activation map (CAM). However, CAMs can hardly serve as the object mask due to the gap between full and weak supervisions. In this paper, we propose a self-supervised equivariant attention mechanism (SEAM) to discover additional supervision and narrow the gap. Our method is based on the observation that equivariance is an implicit constraint in fully supervised semantic segmentation, whose pixel-level labels take the same spatial transformation as the input images during data augmentation. However, this constraint is lost on the CAMs trained by image-level supervision. Therefore, we propose consistency regularization on predicted CAMs from various transformed images to provide self-supervision for network learning. Moreover, we propose a pixel correlation module (PCM), which exploits context appearance information and refines the prediction of current pixel by its similar neighbors, leading to further improvement on CAMs consistency. Extensive experiments on PASCAL VOC 2012 dataset demonstrate our method outperforms state-of-the-art methods using the same level of supervision. The code is released online."
                },
                {
                    "title": "Weakly-Supervised Salient Object Detection via Scribble Annotations",
                    "abstract": "Compared with laborious pixel-wise dense labeling, it is much easier to label data by scribbles, which only costs 1~2 seconds to label one image. However, using scribble labels to learn salient object detection has not been explored. In this paper, we propose a weakly-supervised salient object detection model to learn saliency from such annotations. In doing so, we first relabel an existing large-scale salient object detection dataset with scribbles, namely S-DUTS dataset. Since object structure and detail information is not identified by scribbles, directly training with scribble labels will lead to saliency maps of poor boundary localization. To mitigate this problem, we propose an auxiliary edge detection task to localize object edges explicitly, and a gated structure-aware loss to place constraints on the scope of structure to be recovered. Moreover, we design a scribble boosting scheme to iteratively consolidate our scribble annotations, which are then employed as supervision to learn high-quality saliency maps. As existing saliency evaluation metrics neglect to measure structure alignment of the predictions, the saliency map ranking may not comply with human perception. We present a new metric, termed saliency structure measure, as a complementary metric to evaluate sharpness of the prediction. Extensive experiments on six benchmark datasets demonstrate that our method not only outperforms existing weakly-supervised/unsupervised methods, but also is on par with several fully-supervised state-of-the-art models (Our code and data is publicly available at: https://github.com/JingZhang617/Scribble_Saliency)."
                },
                {
                    "title": "Scalability in Perception for Autonomous Driving: Waymo Open Dataset",
                    "abstract": "The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data. Existing self-driving datasets are limited in the scale and variation of the environments they capture, even though generalization within and between operating regions is crucial to the over-all viability of the technology. In an effort to help align the research community\u2019s contributions with real-world self-driving problems, we introduce a new large scale, high quality, diverse dataset. Our new dataset consists of 1150 scenes that each span 20 seconds, consisting of well synchronized and calibrated high quality LiDAR and camera data captured across a range of urban and suburban geographies. It is 15x more diverse than the largest camera+LiDAR dataset available based on our proposed diversity metric. We exhaustively annotated this data with 2D (camera image) and 3D (LiDAR) bounding boxes, with consistent identifiers across frames. Finally, we provide strong baselines for 2D as well as 3D detection and tracking tasks. We further study the effects of dataset size and generalization across geographies on 3D detection methods. Find data, code and more up-to-date information at http://www.waymo.com/open."
                },
                {
                    "title": "Self-Training With Noisy Student Improves ImageNet Classification",
                    "abstract": "We present a simple self-training method that achieves 88.4% top-1 accuracy on ImageNet, which is 2.0% better than the state-of-the-art model that requires 3.5B weakly labeled Instagram images. On robustness test sets, it improves ImageNet-A top-1 accuracy from 61.0% to 83.7%, reduces ImageNet-C mean corruption error from 45.7 to 28.3, and reduces ImageNet-P mean flip rate from 27.8 to 12.2. To achieve this result, we first train an EfficientNet model on labeled ImageNet images and use it as a teacher to generate pseudo labels on 300M unlabeled images. We then train a larger EfficientNet as a student model on the combination of labeled and pseudo labeled images. We iterate this process by putting back the student as the teacher. During the generation of the pseudo labels, the teacher is not noised so that the pseudo labels are as accurate as possible. However, during the learning of the student, we inject noise such as dropout, stochastic depth and data augmentation via RandAugment to the student so that the student generalizes better than the teacher."
                },
                {
                    "title": "Boundary Perception Guidance: A Scribble-Supervised Semantic Segmentation Approach",
                    "abstract": "Semantic segmentation suffers from the fact that densely annotated masks are expensive to obtain. To tackle this problem, we aim at learning to segment by only leveraging scribbles that are much easier to collect for supervision. To fully explore the limited pixel-level annotations from scribbles, we present a novel Boundary Perception Guidance (BPG) approach, which consists of two basic components, i.e., prediction refinement and boundary regression. Specifically, the prediction refinement progressively makes a better segmentation by adopting an iterative upsampling and a semantic feature\u00a0 enhancement strategy. In the boundary regression, we employ class-agnostic edge maps for supervision to effectively guide the segmentation network in localizing the boundaries between different semantic regions, leading to producing finer-grained representation of feature maps for semantic segmentation. The experiment results on the PASCAL VOC 2012 demonstrate the proposed BPG achieves mIoU of 73.2% without fully connected Conditional Random Field (CRF) and 76.0% with CRF, setting up the new state-of-the-art in literature."
                },
                {
                    "title": "nuScenes: A Multimodal Dataset for Autonomous Driving",
                    "abstract": "Robust detection and tracking of objects is crucial for the deployment of autonomous vehicle technology. Image based benchmark datasets have driven development in computer vision tasks such as object detection, tracking and segmentation of agents in the environment. Most autonomous vehicles, however, carry a combination of cameras and range sensors such as lidar and radar. As machine learning based methods for detection and tracking become more prevalent, there is a need to train and evaluate such methods on datasets containing range sensor data along with images. In this work we present nuTonomy scenes (nuScenes), the first dataset to carry the full autonomous vehicle sensor suite: 6 cameras, 5 radars and 1 lidar, all with full 360 degree field of view. nuScenes comprises 1000 scenes, each 20s long and fully annotated with 3D bounding boxes for 23 classes and 8 attributes. It has 7x as many annotations and 100x as many images as the pioneering KITTI dataset. We define novel 3D detection and tracking metrics. We also provide careful dataset analysis as well as baselines for lidar and image based detection and tracking. Data, development kit and more information are available online."
                },
                {
                    "title": "Normalized Cut Loss for Weakly-Supervised CNN Segmentation",
                    "abstract": "Most recent semantic segmentation methods train deep convolutional neural networks with fully annotated masks requiring pixel-accuracy for good quality training. Common weakly-supervised approaches generate full masks from partial input (e.g. scribbles or seeds) using standard interactive segmentation methods as preprocessing. But, errors in such masks result in poorer training since standard loss functions (e.g. cross-entropy) do not distinguish seeds from potentially mislabeled other pixels. Inspired by the general ideas in semi-supervised learning, we address these problems via a new principled loss function evaluating network output with criteria standard in \"shallow\" segmentation, e.g. normalized cut. Unlike prior work, the cross entropy part of our loss evaluates only seeds where labels are known while normalized cut softly evaluates consistency of all pixels. We focus on normalized cut loss where dense Gaussian kernel is efficiently implemented in linear time by fast Bilateral filtering. Our normalized cut loss approach to segmentation brings the quality of weakly-supervised training significantly closer to fully supervised methods."
                },
                {
                    "title": "Improved Regularization of Convolutional Neural Networks with Cutout",
                    "abstract": "Convolutional neural networks are capable of learning powerful representational spaces, which are necessary for tackling complex learning tasks. However, due to the model capacity required to capture such representations, they are often susceptible to overfitting and therefore require proper regularization in order to generalize well. In this paper, we show that the simple regularization technique of randomly masking out square regions of input during training, which we call cutout, can be used to improve the robustness and overall performance of convolutional neural networks. Not only is this method extremely easy to implement, but we also demonstrate that it can be used in conjunction with existing forms of data augmentation and other regularizers to further improve model performance. We evaluate this method by applying it to current state-of-the-art architectures on the CIFAR-10, CIFAR-100, and SVHN datasets, yielding new state-of-the-art results of 2.56%, 15.20%, and 1.30% test error respectively. Code is available at this https URL"
                },
                {
                    "title": "Scene Parsing through ADE20K Dataset",
                    "abstract": "Scene parsing, or recognizing and segmenting objects and stuff in an image, is one of the key problems in computer vision. Despite the communitys efforts in data collection, there are still few image datasets covering a wide range of scenes and object categories with dense and detailed annotations for scene parsing. In this paper, we introduce and analyze the ADE20K dataset, spanning diverse annotations of scenes, objects, parts of objects, and in some cases even parts of parts. A scene parsing benchmark is built upon the ADE20K with 150 object and stuff classes included. Several segmentation baseline models are evaluated on the benchmark. A novel network design called Cascade Segmentation Module is proposed to parse a scene into stuff, objects, and object parts in a cascade and improve over the baselines. We further show that the trained scene parsing networks can lead to applications such as image content removal and scene synthesis1."
                },
                {
                    "title": "Learning Random-Walk Label Propagation for Weakly-Supervised Semantic Segmentation",
                    "abstract": "Large-scale training for semantic segmentation is challenging due to the expense of obtaining training data for this task relative to other vision tasks. We propose a novel training approach to address this difficulty. Given cheaply-obtained sparse image labelings, we propagate the sparse labels to produce guessed dense labelings. A standard CNN-based segmentation network is trained to mimic these labelings. The label-propagation process is defined via random-walk hitting probabilities, which leads to a differentiable parameterization with uncertainty estimates that are incorporated into our loss. We show that by learning the label-propagator jointly with the segmentation predictor, we are able to effectively learn semantic edges given no direct edge supervision. Experiments also show that training a segmentation network in this way outperforms the naive approach."
                },
                {
                    "title": "Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results",
                    "abstract": "The recently proposed Temporal Ensembling has achieved state-of-the-art results in several semi-supervised learning benchmarks. It maintains an exponential moving average of label predictions on each training example, and penalizes predictions that are inconsistent with this target. However, because the targets change only once per epoch, Temporal Ensembling becomes unwieldy when learning large datasets. To overcome this problem, we propose Mean Teacher, a method that averages model weights instead of label predictions. As an additional benefit, Mean Teacher improves test accuracy and enables training with fewer labels than Temporal Ensembling. Without changing the network architecture, Mean Teacher achieves an error rate of 4.35% on SVHN with 250 labels, outperforming Temporal Ensembling trained with 1000 labels. We also show that a good network architecture is crucial to performance. Combining Mean Teacher and Residual Networks, we improve the state of the art on CIFAR-10 with 4000 labels from 10.55% to 6.28%, and on ImageNet 2012 with 10% of the labels from 35.24% to 9.11%."
                },
                {
                    "title": "ScribbleSup: Scribble-Supervised Convolutional Networks for Semantic Segmentation",
                    "abstract": "Large-scale data is of crucial importance for learning semantic segmentation models, but annotating per-pixel masks is a tedious and inefficient procedure. We note that for the topic of interactive image segmentation, scribbles are very widely used in academic research and commercial software, and are recognized as one of the most userfriendly ways of interacting. In this paper, we propose to use scribbles to annotate images, and develop an algorithm to train convolutional networks for semantic segmentation supervised by scribbles. Our algorithm is based on a graphical model that jointly propagates information from scribbles to unmarked pixels and learns network parameters. We present competitive object semantic segmentation results on the PASCAL VOC dataset by using scribbles as annotations. Scribbles are also favored for annotating stuff (e.g., water, sky, grass) that has no well-defined shape, and our method shows excellent results on the PASCALCONTEXT dataset thanks to extra inexpensive scribble annotations. Our scribble annotations on PASCAL VOC are available at http://research.microsoft.com/en-us/um/ people/jifdai/downloads/scribble_sup."
                },
                {
                    "title": "The Cityscapes Dataset for Semantic Urban Scene Understanding",
                    "abstract": "Visual understanding of complex urban street scenes is an enabling factor for a wide range of applications. Object detection has benefited enormously from large-scale datasets, especially in the context of deep learning. For semantic urban scene understanding, however, no current dataset adequately captures the complexity of real-world urban scenes. To address this, we introduce Cityscapes, a benchmark suite and large-scale dataset to train and test approaches for pixel-level and instance-level semantic labeling. Cityscapes is comprised of a large, diverse set of stereo video sequences recorded in streets from 50 different cities. 5000 of these images have high quality pixel-level annotations, 20 000 additional images have coarse annotations to enable methods that leverage large volumes of weakly-labeled data. Crucially, our effort exceeds previous attempts in terms of dataset size, annotation richness, scene variability, and complexity. Our accompanying empirical study provides an in-depth analysis of the dataset characteristics, as well as a performance evaluation of several state-of-the-art approaches based on our benchmark."
                },
                {
                    "title": "Simple Does It: Weakly Supervised Instance and Semantic Segmentation",
                    "abstract": "Semantic labelling and instance segmentation are two tasks that require particularly costly annotations. Starting from weak supervision in the form of bounding box detection annotations, we propose a new approach that does not require modification of the segmentation training procedure. We show that when carefully designing the input labels from given bounding boxes, even a single round of training is enough to improve over previously reported weakly supervised results. Overall, our weak supervision approach reaches ~95% of the quality of the fully supervised model, both for semantic labelling and instance segmentation."
                },
                {
                    "title": "Semantic contours from inverse detectors",
                    "abstract": "We study the challenging problem of localizing and classifying category-specific object contours in real world images. For this purpose, we present a simple yet effective method for combining generic object detectors with bottom-up contours to identify object contours. We also provide a principled way of combining information from different part detectors and across categories. In order to study the problem and evaluate quantitatively our approach, we present a dataset of semantic exterior boundaries on more than 20, 000 object instances belonging to 20 categories, using the images from the VOC2011 PASCAL challenge [7]."
                },
                {
                    "title": "Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data",
                    "abstract": "We present conditional random fields , a framework for building probabilistic models to segment and label sequence data. Conditional random fields offer several advantages over hidden Markov models and stochastic grammars for such tasks, including the ability to relax strong independence assumptions made in those models. Conditional random fields also avoid a fundamental limitation of maximum entropy Markov models (MEMMs) and other discriminative Markov models based on directed graphical models, which can be biased towards states with few successor states. We present iterative parameter estimation algorithms for conditional random fields and compare the performance of the resulting models to HMMs and MEMMs on synthetic and natural-language data."
                },
                {
                    "title": "A fast parallel algorithm for thinning digital patterns",
                    "abstract": "A fast parallel thinning algorithm is proposed in this paper. It consists of two subiterations: one aimed at deleting the south-east boundary points and the north-west corner points while the other one is aimed at deleting the north-west boundary points and the south-east corner points. End points and pixel connectivity are preserved. Each pattern is thinned down to a skeleton of unitary thickness. Experimental results show that this method is very effective. 12 references."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the performance of scribble-supervised semantic segmentation models on complex datasets, such as Cityscapes and ADE20K, where current methods struggle to generalize beyond simpler benchmarks like PascalVOC?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of weakly supervised semantic segmentation, as it addresses the limitations of existing datasets and methods that rely heavily on detailed labels. By improving scribble-supervised models, we can enhance their applicability in real-world scenarios, particularly in self-driving applications and other domains requiring fine-grained segmentation. This research could lead to the development of more efficient annotation techniques, reducing the time and resources needed for dataset creation, and ultimately fostering the deployment of robust vision algorithms across various applications.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in this problem stem from the complexity of modern datasets, which often contain small object instances, numerous classes, and intricate inter-object relationships. Naive approaches may fail because they do not account for the diverse shapes and scales of objects or the need for precise object-to-object boundaries. Additionally, the lack of diverse scribble-labeled datasets limits the ability to train and evaluate models effectively, making it difficult to generalize findings across different contexts. Overcoming these technical and practical obstacles requires innovative methodologies that can leverage the limited information provided by scribble annotations while still achieving high segmentation accuracy.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on simpler datasets like PascalVOC, which do not adequately represent the challenges present in more complex environments. The limited availability of datasets with scribble labels has hindered the exploration of this approach in diverse contexts. Additionally, existing methods may not have been designed to handle the intricacies of modern semantic segmentation tasks, leading to a lack of effective solutions. Our approach aims to fill these gaps by developing a methodology that can adapt scribble annotations to more challenging datasets, thereby improving generalization and performance.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves deriving scribble labels from fully annotated datasets and applying them to train models on complex datasets like Cityscapes and ADE20K. We will utilize a combination of advanced segmentation architectures and novel training techniques to enhance the model's ability to learn from sparse annotations. The evaluation will be conducted using metrics such as mean Intersection over Union"
            }
        },
        "author_data": {
            "8fb8317d-b83a-420c-887d-f6c9a170f3af": {
                "pk": "8fb8317d-b83a-420c-887d-f6c9a170f3af",
                "name": "Wolfgang Boettcher",
                "collaborators": [
                    "Lukas Hoyer",
                    "Ozan Unal",
                    "Ke Li",
                    "Dengxin Dai"
                ],
                "domain": [
                    "Depth Estimation",
                    "Autonomous Systems",
                    "Sensor Adaptation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "LiDAR Meta Depth Completion",
                        "abstract": "Depth estimation is one of the essential tasks to be addressed when creating mobile autonomous systems. While monocular depth estimation methods have improved in recent times, depth completion provides more accurate and reliable depth maps by additionally using sparse depth information from other sensors such as LiDAR. However, current methods are specifically trained for a single LiDAR sensor. As the scanning pattern differs between sensors, every new sensor would require re-training a specialized depth completion model, which is computationally inefficient and not flexible. Therefore, we propose to dynamically adapt the depth completion model to the used sensor type enabling LiDAR adaptive depth completion. Specifically, we propose a meta depth completion network that uses data patterns derived from the data to learn a task network to alter weights of the main depth completion network to solve a given depth completion task effectively. The method demonstrates a strong capability to work on multiple LiDAR scanning patterns and can also generalize to scanning patterns that are unseen during training. While using a single model, our method yields significantly better results than a non-adaptive baseline trained on different LiDAR patterns. It outperforms LiDAR-specific expert models for very sparse cases. These advantages allow flexible deployment of a single depth completion model on different sensors, which could also prove valuable to process the input of nascent LiDAR technology with adaptive instead of fixed scanning patterns."
                    }
                ]
            },
            "72d74a3e-a2e6-4806-9adc-cf41b9721f21": {
                "pk": "72d74a3e-a2e6-4806-9adc-cf41b9721f21",
                "name": "Lukas Hoyer",
                "collaborators": [
                    "Luc Van Gool",
                    "Dengxin Dai",
                    "Anna Khoreva",
                    "Volker Fischer",
                    "Yuhua Chen",
                    "Suman Saha",
                    "Qin Wang",
                    "Ozan Unal",
                    "Muhammad Ferjad Naeem",
                    "Federico Tombari"
                ],
                "domain": [
                    "Unsupervised Domain Adaptation",
                    "Semantic Segmentation",
                    "Deep Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "A Robot Localization Framework Using CNNs for Object Detection and Pose Estimation",
                        "abstract": "External localization is an essential part for the indoor operation of small or cost-efficient robots, as they are used, for example, in swarm robotics. We introduce a two-stage localization and instance identification framework for arbitrary robots based on convolutional neural networks. Object detection is performed on an external camera image of the operation zone providing robot bounding boxes for an identification and orientation estimation convolutional neural network. Additionally, we propose a process to generate the necessary training data. The framework was evaluated with 3 different robot types and various identification patterns. We have analyzed the main framework hyperparameters providing recommendations for the framework operation settings. We achieved up to 98% mAP@IOU0.5 and only 1.6{\\deg} orientation error, running with a frame rate of 50 Hz on a GPU."
                    },
                    {
                        "title": "DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation",
                        "abstract": "As acquiring pixel-wise annotations of real-world images for semantic segmentation is a costly process, a model can instead be trained with more accessible synthetic data and adapted to real images without requiring their annotations. This process is studied in unsupervised domain adaptation (UDA). Even though a large number of methods propose new adaptation strategies, they are mostly based on outdated network architectures. As the influence of recent network architectures has not been systematically studied, we first benchmark different network architectures for UDA and newly reveal the potential of Transformers for UDA semantic segmentation. Based on the findings, we propose a novel UDA method, DAFormer. The network architecture of DAFormer consists of a Transformer encoder and a multi-level context-aware feature fusion decoder. It is enabled by three simple but crucial training strategies to stabilize the training and to avoid overfitting to the source domain: While (1) Rare Class Sampling on the source domain improves the quality of the pseudo-labels by mitigating the confirmation bias of self-training toward common classes, (2) a Thing-Class ImageNet Feature Distance and (3) a learning rate warmup promote feature transfer from ImageNet pretraining. DAFormer represents a major advance in UDA. It improves the state of the art by 10.8 mIoU for GTA-to-Cityscapes and 5.4 mIoU for Synthia-to-Cityscapes and enables learning even difficult classes such as train, bus, and truck well. The implementation is available at https://github.com/lhoyer/DAFormer."
                    },
                    {
                        "title": "HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentation",
                        "abstract": "Unsupervised domain adaptation (UDA) aims to adapt a model trained on the source domain (e.g. synthetic data) to the target domain (e.g. real-world data) without requiring further annotations on the target domain. This work focuses on UDA for semantic segmentation as real-world pixel-wise annotations are particularly expensive to acquire. As UDA methods for semantic segmentation are usually GPU memory intensive, most previous methods operate only on downscaled images. We question this design as low-resolution predictions often fail to preserve fine details. The alternative of training with random crops of high-resolution images alleviates this problem but falls short in capturing long-range, domain-robust context information. Therefore, we propose HRDA, a multi-resolution training approach for UDA, that combines the strengths of small high-resolution crops to preserve fine segmentation details and large low-resolution crops to capture long-range context dependencies with a learned scale attention, while maintaining a manageable GPU memory footprint. HRDA enables adapting small objects and preserving fine segmentation details. It significantly improves the state-of-the-art performance by 5.5 mIoU for GTA-to-Cityscapes and 4.9 mIoU for Synthia-to-Cityscapes, resulting in unprecedented 73.8 and 65.8 mIoU, respectively. The implementation is available at https://github.com/lhoyer/HRDA."
                    },
                    {
                        "title": "Domain Adaptive and Generalizable Network Architectures and Training Strategies for Semantic Image Segmentation",
                        "abstract": "Unsupervised domain adaptation (UDA) and domain generalization (DG) enable machine learning models trained on a source domain to perform well on unlabeled or even unseen target domains. As previous UDA&DG semantic segmentation methods are mostly based on outdated networks, we benchmark more recent architectures, reveal the potential of Transformers, and design the DAFormer network tailored for UDA&DG. It is enabled by three training strategies to avoid overfitting to the source domain: While (1) Rare Class Sampling mitigates the bias toward common source domain classes, (2) a Thing-Class ImageNet Feature Distance and (3) a learning rate warmup promote feature transfer from ImageNet pretraining. As UDA&DG are usually GPU memory intensive, most previous methods downscale or crop images. However, low-resolution predictions often fail to preserve fine details while models trained with cropped images fall short in capturing long-range, domain-robust context information. Therefore, we propose HRDA, a multi-resolution framework for UDA&DG, that combines the strengths of small high-resolution crops to preserve fine segmentation details and large low-resolution crops to capture long-range context dependencies with a learned scale attention. DAFormer and HRDA significantly improve the state-of-the-art UDA&DG by more than 10 mIoU on 5 different benchmarks. The implementation is available at https://github.com/lhoyer/HRDA."
                    },
                    {
                        "title": "Short-Term Prediction and Multi-Camera Fusion on Semantic Grids",
                        "abstract": "An environment representation (ER) is a substantial part of every autonomous system. It introduces a common interface between perception and other system components, such as decision making, and allows downstream algorithms to deal with abstracted data without knowledge of the used sensor. In this work, we propose and evaluate a novel architecture that generates an egocentric, grid-based, predictive, and semantically-interpretable ER. In particular, we provide a proof of concept for the spatio-temporal fusion of multiple camera sequences and short-term prediction in such an ER. Our design utilizes a strong semantic segmentation network together with depth and egomotion estimates to first extract semantic information from multiple camera streams and then transform these separately into egocentric temporally-aligned bird's-eye view grids. A deep encoder-decoder network is trained to fuse a stack of these grids into a unified semantic grid representation and to predict the dynamics of its surrounding. We evaluate this representation on real-world sequences of the Cityscapes dataset and show that our architecture can make accurate predictions in complex sensor fusion scenarios and significantly outperforms a model-driven baseline in a category-based evaluation."
                    },
                    {
                        "title": "MIC: Masked Image Consistency for Context-Enhanced Domain Adaptation",
                        "abstract": "In unsupervised domain adaptation (UDA), a model trained on source data (e.g. synthetic) is adapted to target data (e.g. real-world) without access to target annotation. Most previous UDA methods struggle with classes that have a similar visual appearance on the target domain as no ground truth is available to learn the slight appearance differences. To address this problem, we propose a Masked Image Consistency (MIC) module to enhance UDA by learning spatial context relations of the target domain as additional clues for robust visual recognition. MIC enforces the consistency between predictions of masked target images, where random patches are withheld, and pseudo-labels that are generated based on the complete image by an exponential moving average teacher. To minimize the consistency loss, the network has to learn to infer the predictions of the masked regions from their context. Due to its simple and universal concept, MIC can be integrated into various UDA methods across different visual recognition tasks such as image classification, semantic segmentation, and object detection. MIC significantly improves the state-of-the-art performance across the different recognition tasks for synthetic-to-real, day-to-nighttime, and clear-to-adverse-weather UDA. For instance, MIC achieves an unprecedented UDA performance of 75.9 mIoU and 92.8% on GTA-to-Cityscapes and VisDA-2017, respectively, which corresponds to an improvement of +2.1 and +3.0 percent points over the previous state of the art. The implementation is available at https://github.com/lhoyer/MIC."
                    },
                    {
                        "title": "Three Ways to Improve Semantic Segmentation with Self-Supervised Depth Estimation",
                        "abstract": "Training deep networks for semantic segmentation requires large amounts of labeled training data, which presents a major challenge in practice, as labeling segmentation masks is a highly labor-intensive process. To address this issue, we present a framework for semi-supervised semantic segmentation, which is enhanced by self-supervised monocular depth estimation from unlabeled image sequences. In particular, we propose three key contributions: (1) We transfer knowledge from features learned during self-supervised depth estimation to semantic segmentation, (2) we implement a strong data augmentation by blending images and labels using the geometry of the scene, and (3) we utilize the depth feature diversity as well as the level of difficulty of learning depth in a student-teacher framework to select the most useful samples to be annotated for semantic segmentation. We validate the proposed model on the Cityscapes dataset, where all three modules demonstrate significant performance gains, and we achieve state-of-the-art results for semi-supervised semantic segmentation. The implementation is available at https://github.com/lhoyer/improving_segmentation_with_selfsupervised_depth."
                    },
                    {
                        "title": "Grid Saliency for Context Explanations of Semantic Segmentation",
                        "abstract": "Recently, there has been a growing interest in developing saliency methods that provide visual explanations of network predictions. Still, the usability of existing methods is limited to image classification models. To overcome this limitation, we extend the existing approaches to generate grid saliencies, which provide spatially coherent visual explanations for (pixel-level) dense prediction networks. As the proposed grid saliency allows to spatially disentangle the object and its context, we specifically explore its potential to produce context explanations for semantic segmentation networks, discovering which context most influences the class predictions inside a target object area. We investigate the effectiveness of grid saliency on a synthetic dataset with an artificially induced bias between objects and their context as well as on the real-world Cityscapes dataset using state-of-the-art segmentation networks. Our results show that grid saliency can be successfully used to provide easily interpretable context explanations and, moreover, can be employed for detecting and localizing contextual biases present in the data."
                    },
                    {
                        "title": "Improving Semi-Supervised and Domain-Adaptive Semantic Segmentation with Self-Supervised Depth Estimation",
                        "abstract": "Training deep networks for semantic segmentation requires large amounts of labeled training data, which presents a major challenge in practice, as labeling segmentation masks is a highly labor-intensive process. To address this issue, we present a framework for semi-supervised and domain-adaptive semantic segmentation, which is enhanced by self-supervised monocular depth estimation (SDE) trained only on unlabeled image sequences.   In particular, we utilize SDE as an auxiliary task comprehensively across the entire learning framework: First, we automatically select the most useful samples to be annotated for semantic segmentation based on the correlation of sample diversity and difficulty between SDE and semantic segmentation. Second, we implement a strong data augmentation by mixing images and labels using the geometry of the scene. Third, we transfer knowledge from features learned during SDE to semantic segmentation by means of transfer and multi-task learning. And fourth, we exploit additional labeled synthetic data with Cross-Domain DepthMix and Matching Geometry Sampling to align synthetic and real data.   We validate the proposed model on the Cityscapes dataset, where all four contributions demonstrate significant performance gains, and achieve state-of-the-art results for semi-supervised semantic segmentation as well as for semi-supervised domain adaptation. In particular, with only 1/30 of the Cityscapes labels, our method achieves 92% of the fully-supervised baseline performance and even 97% when exploiting additional data from GTA. The source code is available at https://github.com/lhoyer/improving_segmentation_with_selfsupervised_depth."
                    },
                    {
                        "title": "LiDAR Meta Depth Completion",
                        "abstract": "Depth estimation is one of the essential tasks to be addressed when creating mobile autonomous systems. While monocular depth estimation methods have improved in recent times, depth completion provides more accurate and reliable depth maps by additionally using sparse depth information from other sensors such as LiDAR. However, current methods are specifically trained for a single LiDAR sensor. As the scanning pattern differs between sensors, every new sensor would require re-training a specialized depth completion model, which is computationally inefficient and not flexible. Therefore, we propose to dynamically adapt the depth completion model to the used sensor type enabling LiDAR adaptive depth completion. Specifically, we propose a meta depth completion network that uses data patterns derived from the data to learn a task network to alter weights of the main depth completion network to solve a given depth completion task effectively. The method demonstrates a strong capability to work on multiple LiDAR scanning patterns and can also generalize to scanning patterns that are unseen during training. While using a single model, our method yields significantly better results than a non-adaptive baseline trained on different LiDAR patterns. It outperforms LiDAR-specific expert models for very sparse cases. These advantages allow flexible deployment of a single depth completion model on different sensors, which could also prove valuable to process the input of nascent LiDAR technology with adaptive instead of fixed scanning patterns."
                    },
                    {
                        "title": "SemiVL: Semi-Supervised Semantic Segmentation with Vision-Language Guidance",
                        "abstract": "In semi-supervised semantic segmentation, a model is trained with a limited number of labeled images along with a large corpus of unlabeled images to reduce the high annotation effort. While previous methods are able to learn good segmentation boundaries, they are prone to confuse classes with similar visual appearance due to the limited supervision. On the other hand, vision-language models (VLMs) are able to learn diverse semantic knowledge from image-caption datasets but produce noisy segmentation due to the image-level training. In SemiVL, we propose to integrate rich priors from VLM pre-training into semi-supervised semantic segmentation to learn better semantic decision boundaries. To adapt the VLM from global to local reasoning, we introduce a spatial fine-tuning strategy for label-efficient learning. Further, we design a language-guided decoder to jointly reason over vision and language. Finally, we propose to handle inherent ambiguities in class labels by providing the model with language guidance in the form of class definitions. We evaluate SemiVL on 4 semantic segmentation datasets, where it significantly outperforms previous semi-supervised methods. For instance, SemiVL improves the state-of-the-art by +13.5 mIoU on COCO with 232 annotated images and by +6.1 mIoU on Pascal VOC with 92 labels. Project page: https://github.com/google-research/semivl"
                    },
                    {
                        "title": "Domain Adaptive Semantic Segmentation with Self-Supervised Depth Estimation",
                        "abstract": "Domain adaptation for semantic segmentation aims to improve the model performance in the presence of a distribution shift between source and target domain. Leveraging the supervision from auxiliary tasks~(such as depth estimation) has the potential to heal this shift because many visual tasks are closely related to each other. However, such a supervision is not always available. In this work, we leverage the guidance from self-supervised depth estimation, which is available on both domains, to bridge the domain gap. On the one hand, we propose to explicitly learn the task feature correlation to strengthen the target semantic predictions with the help of target depth estimation. On the other hand, we use the depth prediction discrepancy from source and target depth decoders to approximate the pixel-wise adaptation difficulty. The adaptation difficulty, inferred from depth, is then used to refine the target semantic segmentation pseudo-labels. The proposed method can be easily implemented into existing segmentation frameworks. We demonstrate the effectiveness of our approach on the benchmark tasks SYNTHIA-to-Cityscapes and GTA-to-Cityscapes, on which we achieve the new state-of-the-art performance of $55.0\\%$ and $56.6\\%$, respectively. Our code is available at \\url{https://qin.ee/corda}."
                    },
                    {
                        "title": "EDAPS: Enhanced Domain-Adaptive Panoptic Segmentation",
                        "abstract": "With autonomous industries on the rise, domain adaptation of the visual perception stack is an important research direction due to the cost savings promise. Much prior art was dedicated to domain-adaptive semantic segmentation in the synthetic-to-real context. Despite being a crucial output of the perception stack, panoptic segmentation has been largely overlooked by the domain adaptation community. Therefore, we revisit well-performing domain adaptation strategies from other fields, adapt them to panoptic segmentation, and show that they can effectively enhance panoptic domain adaptation. Further, we study the panoptic network design and propose a novel architecture (EDAPS) designed explicitly for domain-adaptive panoptic segmentation. It uses a shared, domain-robust transformer encoder to facilitate the joint adaptation of semantic and instance features, but task-specific decoders tailored for the specific requirements of both domain-adaptive semantic and instance segmentation. As a result, the performance gap seen in challenging panoptic benchmarks is substantially narrowed. EDAPS significantly improves the state-of-the-art performance for panoptic segmentation UDA by a large margin of 20% on SYNTHIA-to-Cityscapes and even 72% on the more challenging SYNTHIA-to-Mapillary Vistas. The implementation is available at https://github.com/susaha/edaps."
                    },
                    {
                        "title": "2D Feature Distillation for Weakly- and Semi-Supervised 3D Semantic Segmentation",
                        "abstract": "As 3D perception problems grow in popularity and the need for large-scale labeled datasets for LiDAR semantic segmentation increase, new methods arise that aim to reduce the necessity for dense annotations by employing weakly-supervised training. However these methods continue to show weak boundary estimation and high false negative rates for small objects and distant sparse regions. We argue that such weaknesses can be compensated by using RGB images which provide a denser representation of the scene. We propose an image-guidance network (IGNet) which builds upon the idea of distilling high level feature information from a domain adapted synthetically trained 2D semantic segmentation network. We further utilize a one-way contrastive learning scheme alongside a novel mixing strategy called FOVMix, to combat the horizontal field-of-view mismatch between the two sensors and enhance the effects of image guidance. IGNet achieves state-of-the-art results for weakly-supervised LiDAR semantic segmentation on ScribbleKITTI, boasting up to 98% relative performance to fully supervised training with only 8% labeled points, while introducing no additional annotation burden or computational/memory cost during inference. Furthermore, we show that our contributions also prove effective for semi-supervised training, where IGNet claims state-of-the-art results on both ScribbleKITTI and SemanticKITTI."
                    },
                    {
                        "title": "SILC: Improving Vision Language Pretraining with Self-Distillation",
                        "abstract": "Image-Text pretraining on web-scale image caption datasets has become the default recipe for open vocabulary classification and retrieval models thanks to the success of CLIP and its variants. Several works have also used CLIP features for dense prediction tasks and have shown the emergence of open-set abilities. However, the contrastive objective used by these models only focuses on image-text alignment and does not incentivise image feature learning for dense prediction tasks. In this work, we introduce SILC, a novel framework for vision language pretraining. SILC improves image-text contrastive learning with the simple addition of local-to-global correspondence learning by self-distillation. We show that distilling local image features from an exponential moving average (EMA) teacher model significantly improves model performance on dense predictions tasks like detection and segmentation, while also providing improvements on image-level tasks such as classification and retrieval. SILC models sets a new state of the art for zero-shot classification, few shot classification, image and text retrieval, zero-shot segmentation, and open vocabulary segmentation. We further show that SILC features greatly benefit open vocabulary detection, captioning and visual question answering."
                    },
                    {
                        "title": "DGInStyle: Domain-Generalizable Semantic Segmentation with Image Diffusion Models and Stylized Semantic Control",
                        "abstract": "Large, pretrained latent diffusion models (LDMs) have demonstrated an extraordinary ability to generate creative content, specialize to user data through few-shot fine-tuning, and condition their output on other modalities, such as semantic maps. However, are they usable as large-scale data generators, e.g., to improve tasks in the perception stack, like semantic segmentation? We investigate this question in the context of autonomous driving, and answer it with a resounding \"yes\". We propose an efficient data generation pipeline termed DGInStyle. First, we examine the problem of specializing a pretrained LDM to semantically-controlled generation within a narrow domain. Second, we propose a Style Swap technique to endow the rich generative prior with the learned semantic control. Third, we design a Multi-resolution Latent Fusion technique to overcome the bias of LDMs towards dominant objects. Using DGInStyle, we generate a diverse dataset of street scenes, train a domain-agnostic semantic segmentation model on it, and evaluate the model on multiple popular autonomous driving datasets. Our approach consistently increases the performance of several domain generalization methods compared to the previous state-of-the-art methods. The source code and the generated dataset are available at https://dginstyle.github.io."
                    },
                    {
                        "title": "MICDrop: Masking Image and Depth Features via Complementary Dropout for Domain-Adaptive Semantic Segmentation",
                        "abstract": "Unsupervised Domain Adaptation (UDA) is the task of bridging the domain gap between a labeled source domain, e.g., synthetic data, and an unlabeled target domain. We observe that current UDA methods show inferior results on fine structures and tend to oversegment objects with ambiguous appearance. To address these shortcomings, we propose to leverage geometric information, i.e., depth predictions, as depth discontinuities often coincide with segmentation boundaries. We show that naively incorporating depth into current UDA methods does not fully exploit the potential of this complementary information. To this end, we present MICDrop, which learns a joint feature representation by masking image encoder features while inversely masking depth encoder features. With this simple yet effective complementary masking strategy, we enforce the use of both modalities when learning the joint feature representation. To aid this process, we propose a feature fusion module to improve both global as well as local information sharing while being robust to errors in the depth predictions. We show that our method can be plugged into various recent UDA methods and consistently improve results across standard UDA benchmarks, obtaining new state-of-the-art performances."
                    }
                ]
            },
            "b2401cc1-6b3a-4516-aa11-f6a30fef9221": {
                "pk": "b2401cc1-6b3a-4516-aa11-f6a30fef9221",
                "name": "Ozan Unal",
                "collaborators": [
                    "Luc Van Gool",
                    "Dengxin Dai",
                    "Mehmet Ozan Unal",
                    "Isa Yildirim",
                    "Metin Ertas",
                    "Christos Sakaridis",
                    "Lukas Hoyer",
                    "Yigit Baran Can",
                    "Ali Tamer Unal",
                    "Doga Gunduzalp"
                ],
                "domain": [
                    "3D Computer Vision",
                    "Semantic Segmentation",
                    "Deep Learning",
                    "Autonomous Systems"
                ],
                "publications": [
                    {
                        "title": "Improving Point Cloud Semantic Segmentation by Learning 3D Object Detection",
                        "abstract": "Point cloud semantic segmentation plays an essential role in autonomous driving, providing vital information about drivable surfaces and nearby objects that can aid higher level tasks such as path planning and collision avoidance. While current 3D semantic segmentation networks focus on convolutional architectures that perform great for well represented classes, they show a significant drop in performance for underrepresented classes that share similar geometric features. We propose a novel Detection Aware 3D Semantic Segmentation (DASS) framework that explicitly leverages localization features from an auxiliary 3D object detection task. By utilizing multitask training, the shared feature representation of the network is guided to be aware of per class detection features that aid tackling the differentiation of geometrically similar classes. We additionally provide a pipeline that uses DASS to generate high recall proposals for existing 2-stage detectors and demonstrate that the added supervisory signal can be used to improve 3D orientation estimation capabilities. Extensive experiments on both the SemanticKITTI and KITTI object datasets show that DASS can improve 3D semantic segmentation results of geometrically similar classes up to 37.8% IoU in image FOV while maintaining high precision bird's-eye view (BEV) detection results."
                    },
                    {
                        "title": "Self-Supervised Training For Low Dose CT Reconstruction",
                        "abstract": "Ionizing radiation has been the biggest concern in CT imaging. To reduce the dose level without compromising the image quality, low-dose CT reconstruction has been offered with the availability of compressed sensing based reconstruction methods. Recently, data-driven methods got attention with the rise of deep learning, the availability of high computational power, and big datasets. Deep learning based methods have also been used in low-dose CT reconstruction problem in different manners. Usually, the success of these methods depends on labeled data. However, recent studies showed that training can be achieved successfully with noisy datasets. In this study, we defined a training scheme to use low-dose sinograms as their own training targets. We applied the self-supervision principle in the projection domain where the noise is element-wise independent which is a requirement for self-supervised training methods. Using the self-supervised training, the filtering part of the FBP method and the parameters of a denoiser neural network are optimized. We demonstrate that our method outperforms both conventional and compressed sensing based iterative reconstruction methods qualitatively and quantitatively in the reconstruction of analytic CT phantoms and real-world CT images in low-dose CT reconstruction task."
                    },
                    {
                        "title": "An Unsupervised Reconstruction Method For Low-Dose CT Using Deep Generative Regularization Prior",
                        "abstract": "Low-dose CT imaging requires reconstruction from noisy indirect measurements which can be defined as an ill-posed linear inverse problem. In addition to conventional FBP method in CT imaging, recent compressed sensing based methods exploit handcrafted priors which are mostly simplistic and hard to determine. More recently, deep learning (DL) based methods have become popular in medical imaging field. In CT imaging, DL based methods try to learn a function that maps low-dose images to normal-dose images. Although the results of these methods are promising, their success mostly depends on the availability of high-quality massive datasets. In this study, we proposed a method that does not require any training data or a learning process. Our method exploits such an approach that deep convolutional neural networks (CNNs) generate patterns easier than the noise, therefore randomly initialized generative neural networks can be suitable priors to be used in regularizing the reconstruction. In the experiments, the proposed method is implemented with different loss function variants. Both analytical CT phantoms and real-world CT images are used with different views. Conventional FBP method, a popular iterative method (SART), and TV regularized SART are used in the comparisons. We demonstrated that our method with different loss function variants outperforms the other methods both qualitatively and quantitatively."
                    },
                    {
                        "title": "Bayesian Self-Training for Semi-Supervised 3D Segmentation",
                        "abstract": "3D segmentation is a core problem in computer vision and, similarly to many other dense prediction tasks, it requires large amounts of annotated data for adequate training. However, densely labeling 3D point clouds to employ fully-supervised training remains too labor intensive and expensive. Semi-supervised training provides a more practical alternative, where only a small set of labeled data is given, accompanied by a larger unlabeled set. This area thus studies the effective use of unlabeled data to reduce the performance gap that arises due to the lack of annotations. In this work, inspired by Bayesian deep learning, we first propose a Bayesian self-training framework for semi-supervised 3D semantic segmentation. Employing stochastic inference, we generate an initial set of pseudo-labels and then filter these based on estimated point-wise uncertainty. By constructing a heuristic $n$-partite matching algorithm, we extend the method to semi-supervised 3D instance segmentation, and finally, with the same building blocks, to dense 3D visual grounding. We demonstrate state-of-the-art results for our semi-supervised method on SemanticKITTI and ScribbleKITTI for 3D semantic segmentation and on ScanNet and S3DIS for 3D instance segmentation. We further achieve substantial improvements in dense 3D visual grounding over supervised-only baselines on ScanRefer. Our project page is available at ouenal.github.io/bst/."
                    },
                    {
                        "title": "Scribble-Supervised LiDAR Semantic Segmentation",
                        "abstract": "Densely annotating LiDAR point clouds remains too expensive and time-consuming to keep up with the ever growing volume of data. While current literature focuses on fully-supervised performance, developing efficient methods that take advantage of realistic weak supervision have yet to be explored. In this paper, we propose using scribbles to annotate LiDAR point clouds and release ScribbleKITTI, the first scribble-annotated dataset for LiDAR semantic segmentation. Furthermore, we present a pipeline to reduce the performance gap that arises when using such weak annotations. Our pipeline comprises of three stand-alone contributions that can be combined with any LiDAR semantic segmentation model to achieve up to 95.7% of the fully-supervised performance while using only 8% labeled points. Our scribble annotations and code are available at github.com/ouenal/scribblekitti."
                    },
                    {
                        "title": "Discwise Active Learning for LiDAR Semantic Segmentation",
                        "abstract": "While LiDAR data acquisition is easy, labeling for semantic segmentation remains highly time consuming and must therefore be done selectively. Active learning (AL) provides a solution that can iteratively and intelligently label a dataset while retaining high performance and a low budget. In this work we explore AL for LiDAR semantic segmentation. As a human expert is a component of the pipeline, a practical framework must consider common labeling techniques such as sequential labeling that drastically improve annotation times. We therefore propose a discwise approach (DiAL), where in each iteration, we query the region a single frame covers on global coordinates, labeling all frames simultaneously. We then tackle the two major challenges that emerge with discwise AL. Firstly we devise a new acquisition function that takes 3D point density changes into consideration which arise due to location changes or ego-vehicle motion. Next we solve a mixed-integer linear program that provides a general solution to the selection of multiple frames while taking into consideration the possibilities of disc intersections. Finally we propose a semi-supervised learning approach to utilize all frames within our dataset and improve performance."
                    },
                    {
                        "title": "3D U-NetR: Low Dose Computed Tomography Reconstruction via Deep Learning and 3 Dimensional Convolutions",
                        "abstract": "In this paper, we introduced a novel deep learning-based reconstruction technique for low-dose CT imaging using 3 dimensional convolutions to include the sagittal information unlike the existing 2 dimensional networks which exploits correlation only in transverse plane. In the proposed reconstruction technique, sparse and noisy sinograms are back-projected to the image domain with FBP operation, then the denoising process is applied with a U-Net like 3-dimensional network called 3D U-NetR. The proposed network is trained with synthetic and real chest CT images, and 2D U-Net is also trained with the same dataset to show the importance of the third dimension in terms of recovering the fine details. The proposed network shows better quantitative performance on SSIM and PSNR, especially in the real chest CT data. More importantly, 3D U-NetR captures medically critical visual details that cannot be visualized by a 2D network on the reconstruction of real CT images with 1/10 of the normal dose."
                    },
                    {
                        "title": "Frequency-Dependent Attenuation Reconstruction with an Acoustic Reflector",
                        "abstract": "Attenuation of ultrasound waves varies with tissue composition, hence its estimation offers great potential for tissue characterization and diagnosis and staging of pathology. We recently proposed a method that allows to spatially reconstruct the distribution of the overall ultrasound attenuation in tissue based on computed tomography, using reflections from a passive acoustic reflector. This requires a standard ultrasound transducer operating in pulse-echo mode and a calibration protocol using water measurements, thus it can be implemented on conventional ultrasound systems with minor adaptations. Herein, we extend this method by additionally estimating and imaging the frequency-dependent nature of local ultrasound attenuation for the first time. Spatial distributions of attenuation coefficient and exponent are reconstructed, enabling an elaborate and expressive tissue-specific characterization. With simulations, we demonstrate that our proposed method yields a low reconstruction error of 0.04dB/cm at 1MHz for attenuation coefficient and 0.08 for the frequency exponent. With tissue-mimicking phantoms and ex-vivo bovine muscle samples, a high reconstruction contrast as well as reproducibility are demonstrated. Attenuation exponents of a gelatin-cellulose mixture and an ex-vivo bovine muscle sample were found to be, respectively, 1.4 and 0.5 on average, consistently from different images of their heterogeneous compositions. Such frequency-dependent parametrization could enable novel imaging and diagnostic techniques, as well as facilitate attenuation compensation of other ultrasound-based imaging techniques."
                    },
                    {
                        "title": "LiDAR Meta Depth Completion",
                        "abstract": "Depth estimation is one of the essential tasks to be addressed when creating mobile autonomous systems. While monocular depth estimation methods have improved in recent times, depth completion provides more accurate and reliable depth maps by additionally using sparse depth information from other sensors such as LiDAR. However, current methods are specifically trained for a single LiDAR sensor. As the scanning pattern differs between sensors, every new sensor would require re-training a specialized depth completion model, which is computationally inefficient and not flexible. Therefore, we propose to dynamically adapt the depth completion model to the used sensor type enabling LiDAR adaptive depth completion. Specifically, we propose a meta depth completion network that uses data patterns derived from the data to learn a task network to alter weights of the main depth completion network to solve a given depth completion task effectively. The method demonstrates a strong capability to work on multiple LiDAR scanning patterns and can also generalize to scanning patterns that are unseen during training. While using a single model, our method yields significantly better results than a non-adaptive baseline trained on different LiDAR patterns. It outperforms LiDAR-specific expert models for very sparse cases. These advantages allow flexible deployment of a single depth completion model on different sensors, which could also prove valuable to process the input of nascent LiDAR technology with adaptive instead of fixed scanning patterns."
                    },
                    {
                        "title": "Four Ways to Improve Verbo-visual Fusion for Dense 3D Visual Grounding",
                        "abstract": "3D visual grounding is the task of localizing the object in a 3D scene which is referred by a description in natural language. With a wide range of applications ranging from autonomous indoor robotics to AR/VR, the task has recently risen in popularity. A common formulation to tackle 3D visual grounding is grounding-by-detection, where localization is done via bounding boxes. However, for real-life applications that require physical interactions, a bounding box insufficiently describes the geometry of an object. We therefore tackle the problem of dense 3D visual grounding, i.e. referral-based 3D instance segmentation. We propose a dense 3D grounding network ConcreteNet, featuring four novel stand-alone modules that aim to improve grounding performance for challenging repetitive instances, i.e. instances with distractors of the same semantic class. First, we introduce a bottom-up attentive fusion module that aims to disambiguate inter-instance relational cues, next, we construct a contrastive training scheme to induce separation in the latent space, we then resolve view-dependent utterances via a learned global camera token, and finally we employ multi-view ensembling to improve referred mask quality. ConcreteNet ranks 1st on the challenging ScanRefer online benchmark and has won the ICCV 3rd Workshop on Language for 3D Scenes \"3D Object Localization\" challenge. Our code is available at ouenal.github.io/concretenet/."
                    },
                    {
                        "title": "2D Feature Distillation for Weakly- and Semi-Supervised 3D Semantic Segmentation",
                        "abstract": "As 3D perception problems grow in popularity and the need for large-scale labeled datasets for LiDAR semantic segmentation increase, new methods arise that aim to reduce the necessity for dense annotations by employing weakly-supervised training. However these methods continue to show weak boundary estimation and high false negative rates for small objects and distant sparse regions. We argue that such weaknesses can be compensated by using RGB images which provide a denser representation of the scene. We propose an image-guidance network (IGNet) which builds upon the idea of distilling high level feature information from a domain adapted synthetically trained 2D semantic segmentation network. We further utilize a one-way contrastive learning scheme alongside a novel mixing strategy called FOVMix, to combat the horizontal field-of-view mismatch between the two sensors and enhance the effects of image guidance. IGNet achieves state-of-the-art results for weakly-supervised LiDAR semantic segmentation on ScribbleKITTI, boasting up to 98% relative performance to fully supervised training with only 8% labeled points, while introducing no additional annotation burden or computational/memory cost during inference. Furthermore, we show that our contributions also prove effective for semi-supervised training, where IGNet claims state-of-the-art results on both ScribbleKITTI and SemanticKITTI."
                    },
                    {
                        "title": "Understanding Bird's-Eye View of Road Semantics using an Onboard Camera",
                        "abstract": "Autonomous navigation requires scene understanding of the action-space to move or anticipate events. For planner agents moving on the ground plane, such as autonomous vehicles, this translates to scene understanding in the bird's-eye view (BEV). However, the onboard cameras of autonomous cars are customarily mounted horizontally for a better view of the surrounding. In this work, we study scene understanding in the form of online estimation of semantic BEV maps using the video input from a single onboard camera. We study three key aspects of this task, image-level understanding, BEV level understanding, and the aggregation of temporal information. Based on these three pillars we propose a novel architecture that combines these three aspects. In our extensive experiments, we demonstrate that the considered aspects are complementary to each other for BEV understanding. Furthermore, the proposed architecture significantly surpasses the current state-of-the-art. Code: https://github.com/ybarancan/BEV_feat_stitch."
                    }
                ]
            },
            "71d443c6-ca47-44cb-9133-61dea474617b": {
                "pk": "71d443c6-ca47-44cb-9133-61dea474617b",
                "name": "Jan Eric Lenssen",
                "collaborators": [
                    "Christopher Wewer",
                    "Eddy Ilg",
                    "Gerard Pons-Moll",
                    "Matthias Fey",
                    "Bernt Schiele",
                    "Bharat Lal Bhatnagar",
                    "Kevin Raj",
                    "Raza Yunus",
                    "Xianghui Xie",
                    "Tolga Birdal"
                ],
                "domain": [
                    "Graph Neural Network",
                    "3D Reconstruction",
                    "Machine Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Deep Iterative Surface Normal Estimation",
                        "abstract": "This paper presents an end-to-end differentiable algorithm for robust and detail-preserving surface normal estimation on unstructured point-clouds. We utilize graph neural networks to iteratively parameterize an adaptive anisotropic kernel that produces point weights for weighted least-squares plane fitting in local neighborhoods. The approach retains the interpretability and efficiency of traditional sequential plane fitting while benefiting from adaptation to data set statistics through deep learning. This results in a state-of-the-art surface normal estimator that is robust to noise, outliers and point density variation, preserves sharp features through anisotropic kernels and equivariance through a local quaternion-based spatial transformer. Contrary to previous deep learning methods, the proposed approach does not require any hand-crafted features or preprocessing. It improves on the state-of-the-art results while being more than two orders of magnitude faster and more parameter efficient."
                    },
                    {
                        "title": "Group Equivariant Capsule Networks",
                        "abstract": "We present group equivariant capsule networks, a framework to introduce guaranteed equivariance and invariance properties to the capsule network idea. Our work can be divided into two contributions. First, we present a generic routing by agreement algorithm defined on elements of a group and prove that equivariance of output pose vectors, as well as invariance of output activations, hold under certain conditions. Second, we connect the resulting equivariant capsule networks with work from the field of group convolutional networks. Through this connection, we provide intuitions of how both methods relate and are able to combine the strengths of both approaches in one deep neural network architecture. The resulting framework allows sparse evaluation of the group convolution operator, provides control over specific equivariance and invariance properties, and can use routing by agreement instead of pooling operations. In addition, it is able to provide interpretable and equivariant representation vectors as output capsules, which disentangle evidence of object existence from its pose."
                    },
                    {
                        "title": "SplineCNN: Fast Geometric Deep Learning with Continuous B-Spline Kernels",
                        "abstract": "We present Spline-based Convolutional Neural Networks (SplineCNNs), a variant of deep neural networks for irregular structured and geometric input, e.g., graphs or meshes. Our main contribution is a novel convolution operator based on B-splines, that makes the computation time independent from the kernel size due to the local support property of the B-spline basis functions. As a result, we obtain a generalization of the traditional CNN convolution operator by using continuous kernel functions parametrized by a fixed number of trainable weights. In contrast to related approaches that filter in the spectral domain, the proposed method aggregates features purely in the spatial domain. In addition, SplineCNN allows entire end-to-end training of deep architectures, using only the geometric structure as input, instead of handcrafted feature descriptors. For validation, we apply our method on tasks from the fields of image graph classification, shape correspondence and graph node classification, and show that it outperforms or pars state-of-the-art approaches while being significantly faster and having favorable properties like domain-independence."
                    },
                    {
                        "title": "Fast Graph Representation Learning with PyTorch Geometric",
                        "abstract": "We introduce PyTorch Geometric, a library for deep learning on irregularly structured input data such as graphs, point clouds and manifolds, built upon PyTorch. In addition to general graph data structures and processing methods, it contains a variety of recently published methods from the domains of relational learning and 3D data processing. PyTorch Geometric achieves high data throughput by leveraging sparse GPU acceleration, by providing dedicated CUDA kernels and by introducing efficient mini-batch handling for input examples of different size. In this work, we present the library in detail and perform a comprehensive comparative study of the implemented methods in homogeneous evaluation scenarios."
                    },
                    {
                        "title": "SimNP: Learning Self-Similarity Priors Between Neural Points",
                        "abstract": "Existing neural field representations for 3D object reconstruction either (1) utilize object-level representations, but suffer from low-quality details due to conditioning on a global latent code, or (2) are able to perfectly reconstruct the observations, but fail to utilize object-level prior knowledge to infer unobserved regions. We present SimNP, a method to learn category-level self-similarities, which combines the advantages of both worlds by connecting neural point radiance fields with a category-level self-similarity representation. Our contribution is two-fold. (1) We design the first neural point representation on a category level by utilizing the concept of coherent point clouds. The resulting neural point radiance fields store a high level of detail for locally supported object regions. (2) We learn how information is shared between neural points in an unconstrained and unsupervised fashion, which allows to derive unobserved regions of an object during the reconstruction process from given observations. We show that SimNP is able to outperform previous methods in reconstructing symmetric unseen object regions, surpassing methods that build upon category-level or pixel-aligned radiance fields, while providing semantic correspondences between instances"
                    },
                    {
                        "title": "Neural Point Cloud Diffusion for Disentangled 3D Shape and Appearance Generation",
                        "abstract": "Controllable generation of 3D assets is important for many practical applications like content creation in movies, games and engineering, as well as in AR/VR. Recently, diffusion models have shown remarkable results in generation quality of 3D objects. However, none of the existing models enable disentangled generation to control the shape and appearance separately. For the first time, we present a suitable representation for 3D diffusion models to enable such disentanglement by introducing a hybrid point cloud and neural radiance field approach. We model a diffusion process over point positions jointly with a high-dimensional feature space for a local density and radiance decoder. While the point positions represent the coarse shape of the object, the point features allow modeling the geometry and appearance details. This disentanglement enables us to sample both independently and therefore to control both separately. Our approach sets a new state of the art in generation compared to previous disentanglement-capable methods by reduced FID scores of 30-90% and is on-par with other non disentanglement-capable state-of-the art methods."
                    },
                    {
                        "title": "latentSplat: Autoencoding Variational Gaussians for Fast Generalizable 3D Reconstruction",
                        "abstract": "We present latentSplat, a method to predict semantic Gaussians in a 3D latent space that can be splatted and decoded by a light-weight generative 2D architecture. Existing methods for generalizable 3D reconstruction either do not scale to large scenes and resolutions, or are limited to interpolation of close input views. latentSplat combines the strengths of regression-based and generative approaches while being trained purely on readily available real video data. The core of our method are variational 3D Gaussians, a representation that efficiently encodes varying uncertainty within a latent space consisting of 3D feature Gaussians. From these Gaussians, specific instances can be sampled and rendered via efficient splatting and a fast, generative decoder. We show that latentSplat outperforms previous works in reconstruction quality and generalization, while being fast and scalable to high-resolution data."
                    },
                    {
                        "title": "Improving 2D Feature Representations by 3D-Aware Fine-Tuning",
                        "abstract": "Current visual foundation models are trained purely on unstructured 2D data, limiting their understanding of 3D structure of objects and scenes. In this work, we show that fine-tuning on 3D-aware data improves the quality of emerging semantic features. We design a method to lift semantic 2D features into an efficient 3D Gaussian representation, which allows us to re-render them for arbitrary views. Using the rendered 3D-aware features, we design a fine-tuning strategy to transfer such 3D awareness into a 2D foundation model. We demonstrate that models fine-tuned in that way produce features that readily improve downstream task performance in semantic segmentation and depth estimation through simple linear probing. Notably, though fined-tuned on a single indoor dataset, the improvement is transferable to a variety of indoor datasets and out-of-domain datasets. We hope our study encourages the community to consider injecting 3D awareness when training 2D foundation models. Project page: https://ywyue.github.io/FiT3D."
                    },
                    {
                        "title": "Spurfies: Sparse Surface Reconstruction using Local Geometry Priors",
                        "abstract": "We introduce Spurfies, a novel method for sparse-view surface reconstruction that disentangles appearance and geometry information to utilize local geometry priors trained on synthetic data. Recent research heavily focuses on 3D reconstruction using dense multi-view setups, typically requiring hundreds of images. However, these methods often struggle with few-view scenarios. Existing sparse-view reconstruction techniques often rely on multi-view stereo networks that need to learn joint priors for geometry and appearance from a large amount of data. In contrast, we introduce a neural point representation that disentangles geometry and appearance to train a local geometry prior using a subset of the synthetic ShapeNet dataset only. During inference, we utilize this surface prior as additional constraint for surface and appearance reconstruction from sparse input views via differentiable volume rendering, restricting the space of possible solutions. We validate the effectiveness of our method on the DTU dataset and demonstrate that it outperforms previous state of the art by 35% in surface quality while achieving competitive novel view synthesis quality. Moreover, in contrast to previous works, our method can be applied to larger, unbounded scenes, such as Mip-NeRF 360."
                    },
                    {
                        "title": "Sp2360: Sparse-view 360 Scene Reconstruction using Cascaded 2D Diffusion Priors",
                        "abstract": "We aim to tackle sparse-view reconstruction of a 360 3D scene using priors from latent diffusion models (LDM). The sparse-view setting is ill-posed and underconstrained, especially for scenes where the camera rotates 360 degrees around a point, as no visual information is available beyond some frontal views focused on the central object(s) of interest. In this work, we show that pretrained 2D diffusion models can strongly improve the reconstruction of a scene with low-cost fine-tuning. Specifically, we present SparseSplat360 (Sp2360), a method that employs a cascade of in-painting and artifact removal models to fill in missing details and clean novel views. Due to superior training and rendering speeds, we use an explicit scene representation in the form of 3D Gaussians over NeRF-based implicit representations. We propose an iterative update strategy to fuse generated pseudo novel views with existing 3D Gaussians fitted to the initial sparse inputs. As a result, we obtain a multi-view consistent scene representation with details coherent with the observed inputs. Our evaluation on the challenging Mip-NeRF360 dataset shows that our proposed 2D to 3D distillation algorithm considerably improves the performance of a regularized version of 3DGS adapted to a sparse-view setting and outperforms existing sparse-view reconstruction methods in 360 scene reconstruction. Qualitatively, our method generates entire 360 scenes from as few as 9 input views, with a high degree of foreground and background detail."
                    },
                    {
                        "title": "InterTrack: Tracking Human Object Interaction without Object Templates",
                        "abstract": "Tracking human object interaction from videos is important to understand human behavior from the rapidly growing stream of video data. Previous video-based methods require predefined object templates while single-image-based methods are template-free but lack temporal consistency. In this paper, we present a method to track human object interaction without any object shape templates. We decompose the 4D tracking problem into per-frame pose tracking and canonical shape optimization. We first apply a single-view reconstruction method to obtain temporally-inconsistent per-frame interaction reconstructions. Then, for the human, we propose an efficient autoencoder to predict SMPL vertices directly from the per-frame reconstructions, introducing temporally consistent correspondence. For the object, we introduce a pose estimator that leverages temporal information to predict smooth object rotations under occlusions. To train our model, we propose a method to generate synthetic interaction videos and synthesize in total 10 hour videos of 8.5k sequences with full 3D ground truth. Experiments on BEHAVE and InterCap show that our method significantly outperforms previous template-based video tracking and single-frame reconstruction methods. Our proposed synthetic video dataset also allows training video-based methods that generalize to real-world videos. Our code and dataset will be publicly released."
                    },
                    {
                        "title": "Quaternion Equivariant Capsule Networks for 3D Point Clouds",
                        "abstract": "We present a 3D capsule module for processing point clouds that is equivariant to 3D rotations and translations, as well as invariant to permutations of the input points. The operator receives a sparse set of local reference frames, computed from an input point cloud and establishes end-to-end transformation equivariance through a novel dynamic routing procedure on quaternions. Further, we theoretically connect dynamic routing between capsules to the well-known Weiszfeld algorithm, a scheme for solving \\emph{iterative re-weighted least squares} (IRLS) problems with provable convergence properties. It is shown that such group dynamic routing can be interpreted as robust IRLS rotation averaging on capsule votes, where information is routed based on the final inlier scores. Based on our operator, we build a capsule network that disentangles geometry from pose, paving the way for more informative descriptors and a structured latent space. Our architecture allows joint object classification and orientation estimation without explicit supervision of rotations. We validate our algorithm empirically on common benchmark datasets."
                    },
                    {
                        "title": "TOCH: Spatio-Temporal Object-to-Hand Correspondence for Motion Refinement",
                        "abstract": "We present TOCH, a method for refining incorrect 3D hand-object interaction sequences using a data prior. Existing hand trackers, especially those that rely on very few cameras, often produce visually unrealistic results with hand-object intersection or missing contacts. Although correcting such errors requires reasoning about temporal aspects of interaction, most previous works focus on static grasps and contacts. The core of our method are TOCH fields, a novel spatio-temporal representation for modeling correspondences between hands and objects during interaction. TOCH fields are a point-wise, object-centric representation, which encode the hand position relative to the object. Leveraging this novel representation, we learn a latent manifold of plausible TOCH fields with a temporal denoising auto-encoder. Experiments demonstrate that TOCH outperforms state-of-the-art 3D hand-object interaction models, which are limited to static grasps and contacts. More importantly, our method produces smooth interactions even before and after contact. Using a single trained TOCH model, we quantitatively and qualitatively demonstrate its usefulness for correcting erroneous sequences from off-the-shelf RGB/RGB-D hand-object reconstruction methods and transferring grasps across objects."
                    },
                    {
                        "title": "Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields",
                        "abstract": "We present Pose-NDF, a continuous model for plausible human poses based on neural distance fields (NDFs). Pose or motion priors are important for generating realistic new poses and for reconstructing accurate poses from noisy or partial observations. Pose-NDF learns a manifold of plausible poses as the zero level set of a neural implicit function, extending the idea of modeling implicit surfaces in 3D to the high-dimensional domain SO(3)^K, where a human pose is defined by a single data point, represented by K quaternions. The resulting high-dimensional implicit function can be differentiated with respect to the input poses and thus can be used to project arbitrary poses onto the manifold by using gradient descent on the set of 3-dimensional hyperspheres. In contrast to previous VAE-based human pose priors, which transform the pose space into a Gaussian distribution, we model the actual pose manifold, preserving the distances between poses. We demonstrate that PoseNDF outperforms existing state-of-the-art methods as a prior in various downstream tasks, ranging from denoising real-world human mocap data, pose recovery from occluded data to 3D pose reconstruction from images. Furthermore, we show that it can be used to generate more diverse poses by random sampling and projection than VAE-based methods."
                    },
                    {
                        "title": "Neural Parametric Gaussians for Monocular Non-Rigid Object Reconstruction",
                        "abstract": "Reconstructing dynamic objects from monocular videos is a severely underconstrained and challenging problem, and recent work has approached it in various directions. However, owing to the ill-posed nature of this problem, there has been no solution that can provide consistent, high-quality novel views from camera positions that are significantly different from the training views. In this work, we introduce Neural Parametric Gaussians (NPGs) to take on this challenge by imposing a two-stage approach: first, we fit a low-rank neural deformation model, which then is used as regularization for non-rigid reconstruction in the second stage. The first stage learns the object's deformations such that it preserves consistency in novel views. The second stage obtains high reconstruction quality by optimizing 3D Gaussians that are driven by the coarse model. To this end, we introduce a local 3D Gaussian representation, where temporally shared Gaussians are anchored in and deformed by local oriented volumes. The resulting combined model can be rendered as radiance fields, resulting in high-quality photo-realistic reconstructions of the non-rigidly deforming objects. We demonstrate that NPGs achieve superior results compared to previous works, especially in challenging scenarios with few multi-view cues."
                    },
                    {
                        "title": "Template Free Reconstruction of Human-object Interaction with Procedural Interaction Generation",
                        "abstract": "Reconstructing human-object interaction in 3D from a single RGB image is a challenging task and existing data driven methods do not generalize beyond the objects present in the carefully curated 3D interaction datasets. Capturing large-scale real data to learn strong interaction and 3D shape priors is very expensive due to the combinatorial nature of human-object interactions. In this paper, we propose ProciGen (Procedural interaction Generation), a method to procedurally generate datasets with both, plausible interaction and diverse object variation. We generate 1M+ human-object interaction pairs in 3D and leverage this large-scale data to train our HDM (Hierarchical Diffusion Model), a novel method to reconstruct interacting human and unseen objects, without any templates. Our HDM is an image-conditioned diffusion model that learns both realistic interaction and highly accurate human and object shapes. Experiments show that our HDM trained with ProciGen significantly outperforms prior methods that requires template meshes and that our dataset allows training methods with strong generalization ability to unseen object instances. Our code and data are released."
                    },
                    {
                        "title": "NRDF: Neural Riemannian Distance Fields for Learning Articulated Pose Priors",
                        "abstract": "Faithfully modeling the space of articulations is a crucial task that allows recovery and generation of realistic poses, and remains a notorious challenge. To this end, we introduce Neural Riemannian Distance Fields (NRDFs), data-driven priors modeling the space of plausible articulations, represented as the zero-level-set of a neural field in a high-dimensional product-quaternion space. To train NRDFs only on positive examples, we introduce a new sampling algorithm, ensuring that the geodesic distances follow a desired distribution, yielding a principled distance field learning paradigm. We then devise a projection algorithm to map any random pose onto the level-set by an adaptive-step Riemannian optimizer, adhering to the product manifold of joint rotations at all times. NRDFs can compute the Riemannian gradient via backpropagation and by mathematical analogy, are related to Riemannian flow matching, a recent generative model. We conduct a comprehensive evaluation of NRDF against other pose priors in various downstream tasks, i.e., pose generation, image-based pose estimation, and solving inverse kinematics, highlighting NRDF's superior performance. Besides humans, NRDF's versatility extends to hand and animal poses, as it can effectively represent any articulation."
                    },
                    {
                        "title": "GEARS: Local Geometry-aware Hand-object Interaction Synthesis",
                        "abstract": "Generating realistic hand motion sequences in interaction with objects has gained increasing attention with the growing interest in digital humans. Prior work has illustrated the effectiveness of employing occupancy-based or distance-based virtual sensors to extract hand-object interaction features. Nonetheless, these methods show limited generalizability across object categories, shapes and sizes. We hypothesize that this is due to two reasons: 1) the limited expressiveness of employed virtual sensors, and 2) scarcity of available training data. To tackle this challenge, we introduce a novel joint-centered sensor designed to reason about local object geometry near potential interaction regions. The sensor queries for object surface points in the neighbourhood of each hand joint. As an important step towards mitigating the learning complexity, we transform the points from global frame to hand template frame and use a shared module to process sensor features of each individual joint. This is followed by a spatio-temporal transformer network aimed at capturing correlation among the joints in different dimensions. Moreover, we devise simple heuristic rules to augment the limited training sequences with vast static hand grasping samples. This leads to a broader spectrum of grasping types observed during training, in turn enhancing our model's generalization capability. We evaluate on two public datasets, GRAB and InterCap, where our method shows superiority over baselines both quantitatively and perceptually."
                    },
                    {
                        "title": "Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks",
                        "abstract": "In recent years, graph neural networks (GNNs) have emerged as a powerful neural architecture to learn vector representations of nodes and graphs in a supervised, end-to-end fashion. Up to now, GNNs have only been evaluated empirically -- showing promising results. The following work investigates GNNs from a theoretical point of view and relates them to the $1$-dimensional Weisfeiler-Leman graph isomorphism heuristic ($1$-WL). We show that GNNs have the same expressiveness as the $1$-WL in terms of distinguishing non-isomorphic (sub-)graphs. Hence, both algorithms also have the same shortcomings. Based on this, we propose a generalization of GNNs, so-called $k$-dimensional GNNs ($k$-GNNs), which can take higher-order graph structures at multiple scales into account. These higher-order structures play an essential role in the characterization of social networks and molecule graphs. Our experimental evaluation confirms our theoretical findings as well as confirms that higher-order information is useful in the task of graph classification and regression."
                    }
                ]
            },
            "6dc1a7cd-713a-4992-bbd2-91e474e3811d": {
                "pk": "6dc1a7cd-713a-4992-bbd2-91e474e3811d",
                "name": "Bernt Schiele",
                "collaborators": [
                    "Rodrigo Benenson",
                    "Mario Fritz",
                    "Jan Hosang",
                    "Maksim Lapin",
                    "Matthias Hein",
                    "Yang He",
                    "Zeynep Akata",
                    "Shanshan Zhang",
                    "Gaurav Sharma",
                    "Margret Keuper"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Generative Models",
                    "Object Detection"
                ],
                "publications": [
                    {
                        "title": "Scalable Nonlinear Embeddings for Semantic Category-based Image Retrieval",
                        "abstract": "We propose a novel algorithm for the task of supervised discriminative distance learning by nonlinearly embedding vectors into a low dimensional Euclidean space. We work in the challenging setting where supervision is with constraints on similar and dissimilar pairs while training. The proposed method is derived by an approximate kernelization of a linear Mahalanobis-like distance metric learning algorithm and can also be seen as a kernel neural network. The number of model parameters and test time evaluation complexity of the proposed method are O(dD) where D is the dimensionality of the input features and d is the dimension of the projection space - this is in contrast to the usual kernelization methods as, unlike them, the complexity does not scale linearly with the number of training examples. We propose a stochastic gradient based learning algorithm which makes the method scalable (w.r.t. the number of training examples), while being nonlinear. We train the method with up to half a million training pairs of 4096 dimensional CNN features. We give empirical comparisons with relevant baselines on seven challenging datasets for the task of low dimensional semantic category based image retrieval."
                    },
                    {
                        "title": "How good are detection proposals, really?",
                        "abstract": "Current top performing Pascal VOC object detectors employ detection proposals to guide the search for objects thereby avoiding exhaustive sliding window search across images. Despite the popularity of detection proposals, it is unclear which trade-offs are made when using them during object detection. We provide an in depth analysis of ten object proposal methods along with four baselines regarding ground truth annotation recall (on Pascal VOC 2007 and ImageNet 2013), repeatability, and impact on DPM detector performance. Our findings show common weaknesses of existing methods, and provide insights to choose the most adequate method for different settings."
                    },
                    {
                        "title": "Top-k Multiclass SVM",
                        "abstract": "Class ambiguity is typical in image classification problems with a large number of classes. When classes are difficult to discriminate, it makes sense to allow k guesses and evaluate classifiers based on the top-k error instead of the standard zero-one loss. We propose top-k multiclass SVM as a direct method to optimize for top-k performance. Our generalization of the well-known multiclass SVM is based on a tight convex upper bound of the top-k error. We propose a fast optimization scheme based on an efficient projection onto the top-k simplex, which is of its own interest. Experiments on five datasets show consistent improvements in top-k accuracy compared to various baselines."
                    },
                    {
                        "title": "Filtered Channel Features for Pedestrian Detection",
                        "abstract": "This paper starts from the observation that multiple top performing pedestrian detectors can be modelled by using an intermediate layer filtering low-level features in combination with a boosted decision forest. Based on this observation we propose a unifying framework and experimentally explore different filter families. We report extensive results enabling a systematic analysis.   Using filtered channel features we obtain top performance on the challenging Caltech and KITTI datasets, while using only HOG+LUV as low-level features. When adding optical flow features we further improve detection quality and report the best known results on the Caltech dataset, reaching 93% recall at 1 FPPI."
                    },
                    {
                        "title": "Learning Dilation Factors for Semantic Segmentation of Street Scenes",
                        "abstract": "Contextual information is crucial for semantic segmentation. However, finding the optimal trade-off between keeping desired fine details and at the same time providing sufficiently large receptive fields is non trivial. This is even more so, when objects or classes present in an image significantly vary in size. Dilated convolutions have proven valuable for semantic segmentation, because they allow to increase the size of the receptive field without sacrificing image resolution. However, in current state-of-the-art methods, dilation parameters are hand-tuned and fixed. In this paper, we present an approach for learning dilation parameters adaptively per channel, consistently improving semantic segmentation results on street-scene datasets like Cityscapes and Camvid."
                    },
                    {
                        "title": "Synthetic Convolutional Features for Improved Semantic Segmentation",
                        "abstract": "Recently, learning-based image synthesis has enabled to generate high-resolution images, either applying popular adversarial training or a powerful perceptual loss. However, it remains challenging to successfully leverage synthetic data for improving semantic segmentation with additional synthetic images. Therefore, we suggest to generate intermediate convolutional features and propose the first synthesis approach that is catered to such intermediate convolutional features. This allows us to generate new features from label masks and include them successfully into the training procedure in order to improve the performance of semantic segmentation. Experimental results and analysis on two challenging datasets Cityscapes and ADE20K show that our generated feature improves performance on segmentation tasks."
                    },
                    {
                        "title": "Ten Years of Pedestrian Detection, What Have We Learned?",
                        "abstract": "Paper-by-paper results make it easy to miss the forest for the trees.We analyse the remarkable progress of the last decade by discussing the main ideas explored in the 40+ detectors currently present in the Caltech pedestrian detection benchmark. We observe that there exist three families of approaches, all currently reaching similar detection quality. Based on our analysis, we study the complementarity of the most promising ideas by combining multiple published strategies. This new decision forest detector achieves the current best known performance on the challenging Caltech-USA dataset."
                    },
                    {
                        "title": "CityPersons: A Diverse Dataset for Pedestrian Detection",
                        "abstract": "Convnets have enabled significant progress in pedestrian detection recently, but there are still open questions regarding suitable architectures and training data. We revisit CNN design and point out key adaptations, enabling plain FasterRCNN to obtain state-of-the-art results on the Caltech dataset.   To achieve further improvement from more and better data, we introduce CityPersons, a new set of person annotations on top of the Cityscapes dataset. The diversity of CityPersons allows us for the first time to train one single CNN model that generalizes well over multiple benchmarks. Moreover, with additional training with CityPersons, we obtain top results using FasterRCNN on Caltech, improving especially for more difficult cases (heavy occlusion and small scale) and providing higher localization quality."
                    },
                    {
                        "title": "Learning Using Privileged Information: SVM+ and Weighted SVM",
                        "abstract": "Prior knowledge can be used to improve predictive performance of learning algorithms or reduce the amount of data required for training. The same goal is pursued within the learning using privileged information paradigm which was recently introduced by Vapnik et al. and is aimed at utilizing additional information available only at training time -- a framework implemented by SVM+. We relate the privileged information to importance weighting and show that the prior knowledge expressible with privileged features can also be encoded by weights associated with every training example. We show that a weighted SVM can always replicate an SVM+ solution, while the converse is not true and we construct a counterexample highlighting the limitations of SVM+. Finally, we touch on the problem of choosing weights for weighted SVMs when privileged features are not available."
                    },
                    {
                        "title": "Fine-grained Activity Recognition with Holistic and Pose based Features",
                        "abstract": "Holistic methods based on dense trajectories are currently the de facto standard for recognition of human activities in video. Whether holistic representations will sustain or will be superseded by higher level video encoding in terms of body pose and motion is the subject of an ongoing debate. In this paper we aim to clarify the underlying factors responsible for good performance of holistic and pose-based representations. To that end we build on our recent dataset leveraging the existing taxonomy of human activities. This dataset includes 24,920 video snippets covering 410 human activities in total. Our analysis reveals that holistic and pose-based methods are highly complementary, and their performance varies significantly depending on the activity. We find that holistic methods are mostly affected by the number and speed of trajectories, whereas pose-based methods are mostly influenced by viewpoint of the person. We observe striking performance differences across activities: for certain activities results with pose-based features are more than twice as accurate compared to holistic features, and vice versa. The best performing approach in our comparison is based on the combination of holistic and pose-based approaches, which again underlines their complementarity."
                    },
                    {
                        "title": "The Long-Short Story of Movie Description",
                        "abstract": "Generating descriptions for videos has many applications including assisting blind people and human-robot interaction. The recent advances in image captioning as well as the release of large-scale movie description datasets such as MPII Movie Description allow to study this task in more depth. Many of the proposed methods for image captioning rely on pre-trained object classifier CNNs and Long-Short Term Memory recurrent networks (LSTMs) for generating descriptions. While image description focuses on objects, we argue that it is important to distinguish verbs, objects, and places in the challenging setting of movie description. In this work we show how to learn robust visual classifiers from the weak annotations of the sentence descriptions. Based on these visual classifiers we learn how to generate a description using an LSTM. We explore different design choices to build and train the LSTM and achieve the best performance to date on the challenging MPII-MD dataset. We compare and analyze our approach and prior work along various dimensions to better understand the key challenges of the movie description task."
                    },
                    {
                        "title": "What is Holding Back Convnets for Detection?",
                        "abstract": "Convolutional neural networks have recently shown excellent results in general object detection and many other tasks. Albeit very effective, they involve many user-defined design choices. In this paper we want to better understand these choices by inspecting two key aspects \"what did the network learn?\", and \"what can the network learn?\". We exploit new annotations (Pascal3D+), to enable a new empirical analysis of the R-CNN detector. Despite common belief, our results indicate that existing state-of-the-art convnet architectures are not invariant to various appearance factors. In fact, all considered networks have similar weak points which cannot be mitigated by simply increasing the training data (architectural changes are needed). We show that overall performance can improve when using image renderings for data augmentation. We report the best known results on the Pascal3D+ detection and view-point estimation tasks."
                    },
                    {
                        "title": "A convnet for non-maximum suppression",
                        "abstract": "Non-maximum suppression (NMS) is used in virtually all state-of-the-art object detection pipelines. While essential object detection ingredients such as features, classifiers, and proposal methods have been extensively researched surprisingly little work has aimed to systematically address NMS. The de-facto standard for NMS is based on greedy clustering with a fixed distance threshold, which forces to trade-off recall versus precision. We propose a convnet designed to perform NMS of a given set of detections. We report experiments on a synthetic setup, and results on crowded pedestrian detection scenes. Our approach overcomes the intrinsic limitations of greedy NMS, obtaining better recall and precision."
                    },
                    {
                        "title": "Loss Functions for Top-k Error: Analysis and Insights",
                        "abstract": "In order to push the performance on realistic computer vision tasks, the number of classes in modern benchmark datasets has significantly increased in recent years. This increase in the number of classes comes along with increased ambiguity between the class labels, raising the question if top-1 error is the right performance measure. In this paper, we provide an extensive comparison and evaluation of established multiclass methods comparing their top-k performance both from a practical as well as from a theoretical perspective. Moreover, we introduce novel top-k loss functions as modifications of the softmax and the multiclass SVM losses and provide efficient optimization schemes for them. In the experiments, we compare on various datasets all of the proposed and established methods for top-k error optimization. An interesting insight of this paper is that the softmax loss yields competitive top-k performance for all k simultaneously. For a specific top-k error, our new top-k losses lead typically to further improvements while being faster to train than the softmax."
                    },
                    {
                        "title": "Learning Deep Representations of Fine-grained Visual Descriptions",
                        "abstract": "State-of-the-art methods for zero-shot visual recognition formulate learning as a joint embedding problem of images and side information. In these formulations the current best complement to visual features are attributes: manually encoded vectors describing shared characteristics among categories. Despite good performance, attributes have limitations: (1) finer-grained recognition requires commensurately more attributes, and (2) attributes do not provide a natural language interface. We propose to overcome these limitations by training neural language models from scratch; i.e. without pre-training and only consuming words and characters. Our proposed models train end-to-end to align with the fine-grained and category-specific content of images. Natural language provides a flexible and compact way of encoding only the salient visual aspects for distinguishing categories. By training on raw text, our model can do inference on raw text as well, providing humans a familiar mode both for annotation and retrieval. Our model achieves strong performance on zero-shot text-based image retrieval and significantly outperforms the attribute-based state-of-the-art for zero-shot classification on the Caltech UCSD Birds 200-2011 dataset."
                    },
                    {
                        "title": "Long-Term On-Board Prediction of People in Traffic Scenes under Uncertainty",
                        "abstract": "Progress towards advanced systems for assisted and autonomous driving is leveraging recent advances in recognition and segmentation methods. Yet, we are still facing challenges in bringing reliable driving to inner cities, as those are composed of highly dynamic scenes observed from a moving platform at considerable speeds. Anticipation becomes a key element in order to react timely and prevent accidents. In this paper we argue that it is necessary to predict at least 1 second and we thus propose a new model that jointly predicts ego motion and people trajectories over such large time horizons. We pay particular attention to modeling the uncertainty of our estimates arising from the non-deterministic nature of natural traffic scenes. Our experimental results show that it is indeed possible to predict people trajectories at the desired time horizons and that our uncertainty estimates are informative of the prediction error. We also show that both sequence modeling of trajectories as well as our novel method of long term odometry prediction are essential for best performance."
                    },
                    {
                        "title": "Feature Generating Networks for Zero-Shot Learning",
                        "abstract": "Suffering from the extreme training data imbalance between seen and unseen classes, most of existing state-of-the-art approaches fail to achieve satisfactory results for the challenging generalized zero-shot learning task. To circumvent the need for labeled examples of unseen classes, we propose a novel generative adversarial network (GAN) that synthesizes CNN features conditioned on class-level semantic information, offering a shortcut directly from a semantic descriptor of a class to a class-conditional feature distribution. Our proposed approach, pairing a Wasserstein GAN with a classification loss, is able to generate sufficiently discriminative CNN features to train softmax classifiers or any multimodal embedding method. Our experimental results demonstrate a significant boost in accuracy over the state of the art on five challenging datasets -- CUB, FLO, SUN, AWA and ImageNet -- in both the zero-shot learning and generalized zero-shot learning settings."
                    },
                    {
                        "title": "Connecting Pixels to Privacy and Utility: Automatic Redaction of Private Information in Images",
                        "abstract": "Images convey a broad spectrum of personal information. If such images are shared on social media platforms, this personal information is leaked which conflicts with the privacy of depicted persons. Therefore, we aim for automated approaches to redact such private information and thereby protect privacy of the individual. By conducting a user study we find that obfuscating the image regions related to the private information leads to privacy while retaining utility of the images. Moreover, by varying the size of the regions different privacy-utility trade-offs can be achieved. Our findings argue for a \"redaction by segmentation\" paradigm. Hence, we propose the first sizable dataset of private images \"in the wild\" annotated with pixel and instance level labels across a broad range of privacy classes. We present the first model for automatic redaction of diverse private information."
                    },
                    {
                        "title": "Diverse Conditional Image Generation by Stochastic Regression with Latent Drop-Out Codes",
                        "abstract": "Recent advances in Deep Learning and probabilistic modeling have led to strong improvements in generative models for images. On the one hand, Generative Adversarial Networks (GANs) have contributed a highly effective adversarial learning procedure, but still suffer from stability issues. On the other hand, Conditional Variational Auto-Encoders (CVAE) models provide a sound way of conditional modeling but suffer from mode-mixing issues. Therefore, recent work has turned back to simple and stable regression models that are effective at generation but give up on the sampling mechanism and the latent code representation. We propose a novel and efficient stochastic regression approach with latent drop-out codes that combines the merits of both lines of research. In addition, a new training objective increases coverage of the training distribution leading to improvements over the state of the art in terms of accuracy as well as diversity."
                    },
                    {
                        "title": "Learning Decision Trees Recurrently Through Communication",
                        "abstract": "Integrated interpretability without sacrificing the prediction accuracy of decision making algorithms has the potential of greatly improving their value to the user. Instead of assigning a label to an image directly, we propose to learn iterative binary sub-decisions, inducing sparsity and transparency in the decision making process. The key aspect of our model is its ability to build a decision tree whose structure is encoded into the memory representation of a Recurrent Neural Network jointly learned by two models communicating through message passing. In addition, our model assigns a semantic meaning to each decision in the form of binary attributes, providing concise, semantic and relevant rationalizations to the user. On three benchmark image classification datasets, including the large-scale ImageNet, our model generates human interpretable binary decision sequences explaining the predictions of the network while maintaining state-of-the-art accuracy."
                    }
                ]
            }
        }
    },
    "2309.16965": {
        "paper_data": {
            "title": "Controlling Continuous Relaxation for Combinatorial Optimization",
            "url": "http://arxiv.org/abs/2309.16965v3",
            "arxiv_id": "2309.16965",
            "authors": [
                "Yuma Ichikawa"
            ],
            "abstract": "Unsupervised learning (UL)-based solvers for combinatorial optimization (CO) train a neural network whose output provides a soft solution by directly optimizing the CO objective using a continuous relaxation strategy. These solvers offer several advantages over traditional methods and other learning-based methods, particularly for large-scale CO problems. However, UL-based solvers face two practical issues: (I) an optimization issue where UL-based solvers are easily trapped at local optima, and (II) a rounding issue where UL-based solvers require artificial post-learning rounding from the continuous space back to the original discrete space, undermining the robustness of the results. This study proposes a Continuous Relaxation Annealing (CRA) strategy, an effective rounding-free learning method for UL-based solvers. CRA introduces a penalty term that dynamically shifts from prioritizing continuous solutions, effectively smoothing the non-convexity of the objective function, to enforcing discreteness, eliminating the artificial rounding. Experimental results demonstrate that CRA significantly enhances the performance of UL-based solvers, outperforming existing UL-based solvers and greedy algorithms in complex CO problems. It also effectively eliminates the artificial rounding and accelerates the learning.",
            "introduction": "   1 Introduction  The objective of combinatorial optimization (CO) problems is to find the optimal solution from a discrete space, and these problems are fundamental in many real-world applications (Papadimitriou and Steiglitz, 1998). Most CO problems are NP-hard or NP-complete, making it challenging to solve large-scale problems within feasible computational time. Traditional methods frequently depend on heuristics to find approximate solutions; however, considerable insights into the specific problems are required. Recently, several studies have used machine learning methods to handle CO problems by learning heuristics. Most of these studies focus on supervised learning (SL)-based solvers (Hudson et\u00a0al., 2021; Joshi et\u00a0al., 2019; Gasse et\u00a0al., 2019; Selsam et\u00a0al., 2018; Khalil et\u00a0al., 2016), which require optimal solutions to the CO problem as supervision during training. However, optimal solutions are hard to get in practice, and SL-based solvers often do not generalize well (Yehuda et\u00a0al., 2020). Reinforcement learning (RL)-based solvers (Yao et\u00a0al., 2019; Chen and Tian, 2019; Yolcu and P\u00f3czos, 2019; Nazari et\u00a0al., 2018; Khalil et\u00a0al., 2017; Bello et\u00a0al., 2016) do not require such optimal solutions but frequently suffer from notoriously unstable training due to poor gradient estimation and hard explorations (Mnih et\u00a0al., 2015; Tang et\u00a0al., 2017; Espeholt et\u00a0al., 2018). Recently, unsupervised learning (UL)-based solvers have attracted much attention (Schuetz et\u00a0al., 2022a; Karalias and Loukas, 2020; Amizadeh et\u00a0al., 2018). UL-based solvers follow a continuous relaxation approach. They train a UL model to output a \u201csoft\u201d solution to the relaxed CO problem by directly optimizing the differentiable objective function, offering significantly stable and fast training even for large-scale CO problems. Notably, physics-inspired GNN (PI-GNN) solver (Schuetz et\u00a0al., 2022a) employs graph neural networks (GNN) to automatically learn specific heuristics for a single instance and performs on par with or outperforms existing solvers for CO problems with millions of variables without optimal solutions.   While offering some advantages over traditional and other machine learning-based solvers, they face two practical issues. The first issue is an optimization issues where UL-based solvers are easily trapped at local optima. Angelini and Ricci-Tersenghi (2023) demonstrated that the PI-GNN solver could not achieve results comparable to those of the degree-based greedy algorithm (DGA) (Angelini and Ricci-Tersenghi, 2019) in maximum independent set (MIS) problems on random regular graphs (RRG). Wang and Li (2023) also pointed out the importance of using dataset/history and initialization of GNN using output of greedy solvers for PI-GNN solver to overcome the optimization issues. This issue is a crucial bottleneck for its applicability in various real-world applications. The second issue relates to the inherent ambiguity in the continuous relaxation approach. This approach requires artificial rounding from the soft solution, which may include continuous values, back to the original discrete solution, potentially undermining the robustness of the results. Indeed, while linear relaxation can provide an optimal solution for original discrete problems on bipartite graphs (Hoffman and Kruskal, 2010), it typically leads to solutions with 1/212\\nicefrac{{1}}{{2}}/ start_ARG 1 end_ARG start_ARG 2 end_ARG values, which is known to half-integrality (Nemhauser and Trotter\u00a0Jr, 1974), where existing rounding methods (Schuetz et\u00a0al., 2022b; Wang et\u00a0al., 2022) completely lose their robustness.   To address these issues, we propose the Continuous Relaxation Annealing (CRA) strategy. The CRA introduces a penalty term to control the continuity and discreteness of the relaxed variables, with a",
            "references": [
                {
                    "title": "Variational Annealing on Graphs for Combinatorial Optimization",
                    "abstract": "Several recent unsupervised learning methods use probabilistic approaches to solve combinatorial optimization (CO) problems based on the assumption of statistically independent solution variables. We demonstrate that this assumption imposes performance limitations in particular on difficult problem instances. Our results corroborate that an autoregressive approach which captures statistical dependencies among solution variables yields superior performance on many popular CO problems. We introduce subgraph tokenization in which the configuration of a set of solution variables is represented by a single token. This tokenization technique alleviates the drawback of the long sequential sampling procedure which is inherent to autoregressive methods without sacrificing expressivity. Importantly, we theoretically motivate an annealed entropy regularization and show empirically that it is essential for efficient and stable learning."
                },
                {
                    "title": "Unsupervised Learning for Combinatorial Optimization Needs Meta-Learning",
                    "abstract": "A general framework of unsupervised learning for combinatorial optimization (CO) is to train a neural network (NN) whose output gives a problem solution by directly optimizing the CO objective. Albeit with some advantages over traditional solvers, the current framework optimizes an averaged performance over the distribution of historical problem instances, which misaligns with the actual goal of CO that looks for a good solution to every future encountered instance. With this observation, we propose a new objective of unsupervised learning for CO where the goal of learning is to search for good initialization for future problem instances rather than give direct solutions. We propose a meta-learning-based training pipeline for this new objective. Our method achieves good empirical performance. We observe that even just the initial solution given by our model before fine-tuning can significantly outperform the baselines under various evaluation settings including evaluation across multiple datasets, and the case with big shifts in the problem scale. The reason we conjecture is that meta-learning-based training lets the model be loosely tied to each local optima for a training instance while being more adaptive to the changes of optimization landscapes across instances."
                },
                {
                    "title": "Machine-learning-assisted Monte Carlo fails at sampling computationally hard problems",
                    "abstract": "Several strategies have been recently proposed in order to improve Monte Carlo sampling efficiency using machine learning tools. Here, we challenge these methods by considering a class of problems that are known to be exponentially hard to sample using conventional local Monte Carlo at low enough temperatures. In particular, we study the antiferromagnetic Potts model on a random graph, which reduces to the coloring of random graphs at zero temperature. We test several machine-learning-assisted Monte Carlo approaches, and we find that they all fail. Our work thus provides good benchmarks for future proposals for smart sampling algorithms."
                },
                {
                    "title": "Annealed Training for Combinatorial Optimization on Graphs",
                    "abstract": "The hardness of combinatorial optimization (CO) problems hinders collecting solutions for supervised learning. However, learning neural networks for CO problems is notoriously difficult in lack of the labeled data as the training is easily trapped at local optima. In this work, we propose a simple but effective annealed training framework for CO problems. In particular, we transform CO problems into unbiased energy-based models (EBMs). We carefully selected the penalties terms so as to make the EBMs as smooth as possible. Then we train graph neural networks to approximate the EBMs. To prevent the training from being stuck at local optima near the initialization, we introduce an annealed loss function. An experimental evaluation demonstrates that our annealed training framework obtains substantial improvements. In four types of CO problems, our method achieves performance substantially better than other unsupervised neural methods on both synthetic and real-world graphs."
                },
                {
                    "title": "Unsupervised Learning for Combinatorial Optimization with Principled Objective Relaxation",
                    "abstract": "Using machine learning to solve combinatorial optimization (CO) problems is challenging, especially when the data is unlabeled. This work proposes an unsupervised learning framework for CO problems. Our framework follows a standard relaxation-plus-rounding approach and adopts neural networks to parameterize the relaxed solutions so that simple back-propagation can train the model end-to-end. Our key contribution is the observation that if the relaxed objective satisfies entry-wise concavity, a low optimization loss guarantees the quality of the final integral solutions. This observation significantly broadens the applicability of the previous framework inspired by Erdos' probabilistic method. In particular, this observation can guide the design of objective models in applications where the objectives are not given explicitly while requiring being modeled in prior. We evaluate our framework by solving a synthetic graph optimization problem, and two real-world applications including resource allocation in circuit design and approximate computing. Our framework largely outperforms the baselines based on na\\\"{i}ve relaxation, reinforcement learning, and Gumbel-softmax tricks."
                },
                {
                    "title": "Graph Coloring with Physics-Inspired Graph Neural Networks",
                    "abstract": "We show how graph neural networks can be used to solve the canonical graph coloring problem. We frame graph coloring as a multi-class node classification problem and utilize an unsupervised training strategy based on the statistical physics Potts model. Generalizations to other multi-class problems such as community detection, data clustering, and the minimum clique cover problem are straightforward. We provide numerical benchmark results and illustrate our approach with an end-to-end application for a real-world scheduling use case within a comprehensive encode-process-decode framework. Our optimization approach performs on par or outperforms existing solvers, with the ability to scale to problems with millions of variables."
                },
                {
                    "title": "Decision-Focused Learning: Through the Lens of Learning to Rank",
                    "abstract": "In the last years decision-focused learning framework, also known as predict-and-optimize, have received increasing attention. In this setting, the predictions of a machine learning model are used as estimated cost coefficients in the objective function of a discrete combinatorial optimization problem for decision making. Decision-focused learning proposes to train the ML models, often neural network models, by directly optimizing the quality of decisions made by the optimization solvers. Based on a recent work that proposed a noise contrastive estimation loss over a subset of the solution space, we observe that decision-focused learning can more generally be seen as a learning-to-rank problem, where the goal is to learn an objective function that ranks the feasible points correctly. This observation is independent of the optimization method used and of the form of the objective function. We develop pointwise, pairwise and listwise ranking loss functions, which can be differentiated in closed form given a subset of solutions. We empirically investigate the quality of our generic methods compared to existing decision-focused learning approaches with competitive results. Furthermore, controlling the subset of solutions allows controlling the runtime considerably, with limited effect on regret."
                },
                {
                    "title": "Graph Neural Network Guided Local Search for the Traveling Salesperson Problem",
                    "abstract": "Solutions to the Traveling Salesperson Problem (TSP) have practical applications to processes in transportation, logistics, and automation, yet must be computed with minimal delay to satisfy the real-time nature of the underlying tasks. How-ever, solving large TSP instances quickly without sacri\ufb01cing solution quality remains challenging for current approximate algorithms. To close this gap, we present a hybrid data-driven approach for solving the TSP based on Graph Neural Networks (GNNs) and Guided Local Search (GLS). Our model predicts the regret of including each edge of the problem graph in the solution; GLS uses these predictions in conjunction with the original problem graph to \ufb01nd solutions. Our experiments demonstrate that this approach converges to optimal solutions at a faster rate than three recent learning based approaches for the TSP. Notably, we reduce the mean optimality gap on the 100-node problem set from 1.534% to 0.705%, a 2 \u00d7 improvement. When generalizing from 20-node instances to the 100-node problem set, we reduce the optimality gap from 18.845% to 2.622%, a 7 \u00d7 improvement."
                },
                {
                    "title": "Combinatorial optimization and reasoning with graph neural networks",
                    "abstract": "Combinatorial optimization is a well-established area in operations research and computer science. Until recently, its methods have mostly focused on solving problem instances in isolation, ignoring the fact that they often stem from related data distributions in practice. However, recent years have seen a surge of interest in using machine learning, especially graph neural networks, as a key building block for combinatorial tasks, either directly as solvers or by enhancing the former. This paper presents a conceptual review of recent key advancements in this emerging field, aiming at researchers in both optimization and machine learning."
                },
                {
                    "title": "Contrastive Losses and Solution Caching for Predict-and-Optimize",
                    "abstract": "Many decision-making processes involve solving a combinatorial optimization problem with uncertain input that can be estimated from historic data. Recently, problems in this class have been successfully addressed via end-to-end learning approaches, which rely on solving one optimization problem for each training instance at every epoch. In this context, we provide two distinct contributions. First, we use a Noise Contrastive approach to motivate a family of surrogate loss functions, based on viewing non-optimal solutions as negative examples. Second, we address a major bottleneck of all predict-and-optimize approaches, i.e. the need to frequently recompute optimal solutions at training time. This is done via a solver-agnostic solution caching scheme, and by replacing optimization calls with a lookup in the solution cache. The method is formally based on an inner approximation of the feasible space and, combined with a cache lookup strategy, provides a controllable trade-off between training time and accuracy of the loss approximation. We empirically show that even a very slow growth rate is enough to match the quality of state-of-the-art methods, at a fraction of the computational cost."
                },
                {
                    "title": "POMO: Policy Optimization with Multiple Optima for Reinforcement Learning",
                    "abstract": "In neural combinatorial optimization (CO), reinforcement learning (RL) can turn a deep neural net into a fast, powerful heuristic solver of NP-hard problems. This approach has a great potential in practical applications because it allows near-optimal solutions to be found without expert guides armed with substantial domain knowledge. We introduce Policy Optimization with Multiple Optima (POMO), an end-to-end approach for building such a heuristic solver. POMO is applicable to a wide range of CO problems. It is designed to exploit the symmetries in the representation of a CO solution. POMO uses a modified REINFORCE algorithm that forces diverse rollouts towards all optimal solutions. Empirically, the low-variance baseline of POMO makes RL training fast and stable, and it is more resistant to local minima compared to previous approaches. We also introduce a new augmentation-based inference method, which accompanies POMO nicely. We demonstrate the effectiveness of POMO by solving three popular NP-hard problems, namely, traveling salesman (TSP), capacitated vehicle routing (CVRP), and 0-1 knapsack (KP). For all three, our solver based on POMO shows a significant improvement in performance over all recent learned heuristics. In particular, we achieve the optimality gap of 0.14% with TSP100 while reducing inference time by more than an order of magnitude."
                },
                {
                    "title": "Erdos Goes Neural: an Unsupervised Learning Framework for Combinatorial Optimization on Graphs",
                    "abstract": "Combinatorial optimization problems are notoriously challenging for neural networks, especially in the absence of labeled instances. This work proposes an unsupervised learning framework for CO problems on graphs that can provide integral solutions of certified quality. Inspired by Erdos' probabilistic method, we use a neural network to parametrize a probability distribution over sets. Crucially, we show that when the network is optimized w.r.t. a suitably chosen loss, the learned distribution contains, with controlled probability, a low-cost integral solution that obeys the constraints of the combinatorial problem. The probabilistic proof of existence is then derandomized to decode the desired solutions. We demonstrate the efficacy of this approach to obtain valid solutions to the maximum clique problem and to perform local graph clustering. Our method achieves competitive results on both real datasets and synthetic hard instances."
                },
                {
                    "title": "Graph Neural Networks Meet Neural-Symbolic Computing: A Survey and Perspective",
                    "abstract": "Neural-symbolic computing has now become the subject of interest of both academic and industry research laboratories. \n\nGraph Neural Networks (GNNs) have been widely used in relational and symbolic domains, with widespread application of GNNs in combinatorial optimization, constraint satisfaction, relational reasoning and other scientific domains.\n\nThe need for improved explainability, interpretability and trust of AI systems in general demands principled methodologies, as suggested by neural-symbolic computing. \n\nIn this paper, we review the state-of-the-art on the use of GNNs as a model of neural-symbolic computing. This includes the application of GNNs in several domains as well as their relationship to current developments in neural-symbolic computing."
                },
                {
                    "title": "It's Not What Machines Can Learn, It's What We Cannot Teach",
                    "abstract": "Can deep neural networks learn to solve any task, and in particular problems of high complexity? This question attracts a lot of interest, with recent works tackling computationally hard tasks such as the traveling salesman problem and satisfiability. In this work we offer a different perspective on this question. Given the common assumption that $\\textit{NP} \\neq \\textit{coNP}$ we prove that any polynomial-time sample generator for an $\\textit{NP}$-hard problem samples, in fact, from an easier sub-problem. We empirically explore a case study, Conjunctive Query Containment, and show how common data generation techniques generate biased datasets that lead practitioners to over-estimate model accuracy. Our results suggest that machine learning approaches that require training on a dense uniform sampling from the target distribution cannot be used to solve computationally hard problems, the reason being the difficulty of generating sufficiently large and unbiased training sets."
                },
                {
                    "title": "Boosting Monte Carlo simulations of spin glasses using autoregressive neural networks.",
                    "abstract": "The autoregressive neural networks are emerging as a powerful computational tool to solve relevant problems in classical and quantum mechanics. One of their appealing functionalities is that, after they have learned a probability distribution from a dataset, they allow exact and efficient sampling of typical system configurations. Here we employ a neural autoregressive distribution estimator (NADE) to boost Markov chain Monte Carlo (MCMC) simulations of a paradigmatic classical model of spin-glass theory, namely, the two-dimensional Edwards-Anderson Hamiltonian. We show that a NADE can be trained to accurately mimic the Boltzmann distribution using unsupervised learning from system configurations generated using standard MCMC algorithms. The trained NADE is then employed as smart proposal distribution for the Metropolis-Hastings algorithm. This allows us to perform efficient MCMC simulations, which provide unbiased results even if the expectation value corresponding to the probability distribution learned by the NADE is not exact. Notably, we implement a sequential tempering procedure, whereby a NADE trained at a higher temperature is iteratively employed as proposal distribution in a MCMC simulation run at a slightly lower temperature. This allows one to efficiently simulate the spin-glass model even in the low-temperature regime, avoiding the divergent correlation times that plague MCMC simulations driven by local-update algorithms. Furthermore, we show that the NADE-driven simulations quickly sample ground-state configurations, paving the way to their future utilization to tackle binary optimization problems."
                },
                {
                    "title": "RUN-CSP: Unsupervised Learning of Message Passing Networks for Binary Constraint Satisfaction Problems",
                    "abstract": "Constraint satisfaction problems form an important and wide class of combinatorial search and optimization problems with many applications in AI and other areas. We introduce a recurrent neural network architecture RUN-CSP (Recurrent Unsupervised Neural Network for Constraint Satisfaction Problems) to train message passing networks solving binary constraint satisfaction problems (CSPs) or their optimization versions (Max-CSP). The architecture is universal in the sense that it works for all binary CSPs: depending on the constraint language, we can automtically design a loss function, which is then used to train generic neural nets. In this paper, we experimentally evaluate our approach for the 3-colorability problem (3-Col) and its optimization version (Max-3-Col) and for the maximum 2-satisfiability problem (Max-2-Sat). We also extend the framework to work for related optimization problems such as the maximum independent set problem (Max-IS). Training is unsupervised, we train the network on arbitrary (unlabeled) instances of the problems. Moreover, we experimentally show that it suffices to train on relatively small instances; the resulting message passing network will perform well on much larger instances (at least 10-times larger)."
                },
                {
                    "title": "Deep Graph Library: A Graph-Centric, Highly-Performant Package for Graph Neural Networks.",
                    "abstract": "Advancing research in the emerging field of deep graph learning requires new tools to support tensor computation over graphs. In this paper, we present the design principles and implementation of Deep Graph Library (DGL). DGL distills the computational patterns of GNNs into a few generalized sparse tensor operations suitable for extensive parallelization. By advocating graph as the central programming abstraction, DGL can perform optimizations transparently. By cautiously adopting a framework-neutral design, DGL allows users to easily port and leverage the existing components across multiple deep learning frameworks. Our evaluation shows that DGL significantly outperforms other popular GNN-oriented frameworks in both speed and memory consumption over a variety of benchmarks and has little overhead for small scale workloads."
                },
                {
                    "title": "Experimental performance of graph neural networks on random instances of max-cut",
                    "abstract": "This note explores the applicability of unsupervised machine learning techniques towards hard optimization problems on random inputs. In particular we consider Graph Neural Networks (GNNs) - a class of neural networks designed to learn functions on graphs - and we apply them to the max-cut problem on random regular graphs. We focus on the max-cut problem on random regular graphs because it is a fundamental problem that has been widely studied. In particular, even though there is no known explicit solution to compare the output of our algorithm to, we can leverage the known asymptotics of the optimal max-cut value in order to evaluate the performance of the GNNs. In order to put the performance of the GNNs in context, we compare it with the classical semidefinite relaxation approach by Goemans and Williamson (SDP), and with extremal optimization, which is a local optimization heuristic from the statistical physics literature. The numerical results we obtain indicate that, surprisingly, Graph Neural Networks attain comparable performance to the Goemans and Williamson SDP. We also observe that extremal optimization consistently outperforms the other two methods. Furthermore, the performances of the three methods present similar patterns, that is, for sparser, and for larger graphs, the size of the found cuts are closer to the asymptotic optimal max-cut value."
                },
                {
                    "title": "MIPaaL: Mixed Integer Program as a Layer",
                    "abstract": "Machine learning components commonly appear in larger decision-making pipelines; however, the model training process typically focuses only on a loss that measures average accuracy between predicted values and ground truth values. Decision-focused learning explicitly integrates the downstream decision problem when training the predictive model, in order to optimize the quality of decisions induced by the predictions. It has been successfully applied to several limited combinatorial problem classes, such as those that can be expressed as linear programs (LP), and submodular optimization. However, these previous applications have uniformly focused on problems with simple constraints. Here, we enable decision-focused learning for the broad class of problems that can be encoded as a mixed integer linear program (MIP), hence supporting arbitrary linear constraints over discrete and continuous variables. We show how to differentiate through a MIP by employing a cutting planes solution approach, an algorithm that iteratively tightens the continuous relaxation by adding constraints removing fractional solutions. We evaluate our new end-to-end approach on several real world domains and show that it outperforms the standard two phase approaches that treat prediction and optimization separately, as well as a baseline approach of simply applying decision-focused learning to the LP relaxation of the MIP. Lastly, we demonstrate generalization performance in several transfer learning tasks."
                },
                {
                    "title": "Exact Combinatorial Optimization with Graph Convolutional Neural Networks",
                    "abstract": "Combinatorial optimization problems are typically tackled by the branch-and-bound paradigm. We propose a new graph convolutional neural network model for learning branch-and-bound variable selection policies, which leverages the natural variable-constraint bipartite graph representation of mixed-integer linear programs. We train our model via imitation learning from the strong branching expert rule, and demonstrate on a series of hard problems that our approach produces policies that improve upon state-of-the-art machine-learning methods for branching and generalize to instances significantly larger than seen during training. Moreover, we improve for the first time over expert-designed branching rules implemented in a state-of-the-art solver on large problems. Code for reproducing all the experiments can be found at this https URL."
                },
                {
                    "title": "An Efficient Graph Convolutional Network Technique for the Travelling Salesman Problem",
                    "abstract": "This paper introduces a new learning-based approach for approximately solving the Travelling Salesman Problem on 2D Euclidean graphs. We use deep Graph Convolutional Networks to build efficient TSP graph representations and output tours in a non-autoregressive manner via highly parallelized beam search. Our approach outperforms all recently proposed autoregressive deep learning techniques in terms of solution quality, inference speed and sample efficiency for problem instances of fixed graph sizes. In particular, we reduce the average optimality gap from 0.52% to 0.01% for 50 nodes, and from 2.26% to 1.39% for 100 nodes. Finally, despite improving upon other learning-based approaches for TSP, our approach falls short of standard Operations Research solvers."
                },
                {
                    "title": "Flow-based generative models for Markov chain Monte Carlo in lattice field theory",
                    "abstract": "A Markov chain update scheme using a machine-learned flow-based generative model is proposed for Monte\u00a0Carlo sampling in lattice field theories. The generative model may be optimized (trained) to produce samples from a distribution approximating the desired Boltzmann distribution determined by the lattice action of the theory being studied. Training the model systematically improves autocorrelation times in the Markov chain, even in regions of parameter space where standard Markov chain Monte\u00a0Carlo algorithms exhibit critical slowing down in producing decorrelated updates. Moreover, the model may be trained without existing samples from the desired distribution. The algorithm is compared with HMC and local Metropolis sampling for \u03d54 theory in two dimensions."
                },
                {
                    "title": "Monte Carlo algorithms are very effective in finding the largest independent set in sparse random graphs",
                    "abstract": "The effectiveness of stochastic algorithms based on Monte Carlo dynamics in solving hard optimization problems is mostly unknown. Beyond the basic statement that at a dynamical phase transition the ergodicity breaks and a Monte Carlo dynamics cannot sample correctly the probability distribution in times linear in the system size, there are almost no predictions or intuitions on the behavior of this class of stochastic dynamics. The situation is particularly intricate because, when using a Monte Carlo-based algorithm as an optimization algorithm, one is usually interested in the out-of-equilibrium behavior, which is very hard to analyze. Here we focus on the use of parallel tempering in the search for the largest independent set in a sparse random graph, showing that it can find solutions well beyond the dynamical threshold. Comparison with state-of-the-art message passing algorithms reveals that parallel tempering is definitely the algorithm performing best, although a theory explaining its behavior is still lacking."
                },
                {
                    "title": "Learning to Perform Local Rewriting for Combinatorial Optimization",
                    "abstract": "Search-based methods for hard combinatorial optimization are often guided by heuristics. Tuning heuristics in various conditions and situations is often time-consuming. In this paper, we propose NeuRewriter that learns a policy to pick heuristics and rewrite the local components of the current solution to iteratively improve it until convergence. The policy factorizes into a region-picking and a rule-picking component, each parameterized by a neural network trained with actor-critic methods in reinforcement learning. NeuRewriter captures the general structure of combinatorial problems and shows strong performance in three versatile tasks: expression simplification, online job scheduling and vehicle routing problems. NeuRewriter outperforms the expression simplification component in Z3; outperforms DeepRM and Google OR-tools in online job scheduling; and outperforms recent neural baselines and Google OR-tools in vehicle routing problems."
                },
                {
                    "title": "Solving Statistical Mechanics using Variational Autoregressive Networks",
                    "abstract": "We propose a general framework for solving statistical mechanics of systems with finite size. The approach extends the celebrated variational mean-field approaches using autoregressive neural networks, which support direct sampling and exact calculation of normalized probability of configurations. It computes variational free energy, estimates physical quantities such as entropy, magnetizations and correlations, and generates uncorrelated samples all at once. Training of the network employs the policy gradient approach in reinforcement learning, which unbiasedly estimates the gradient of variational parameters. We apply our approach to several classic systems, including 2D Ising models, the Hopfield model, the Sherrington-Kirkpatrick model, and the inverse Ising model, for demonstrating its advantages over existing variational mean-field methods. Our approach sheds light on solving statistical physics problems using modern deep generative neural networks."
                },
                {
                    "title": "Learning To Solve Circuit-SAT: An Unsupervised Differentiable Approach",
                    "abstract": "Recent efforts to combine Representation Learning with Formal Methods, commonly known as Neuro-Symbolic Methods, have given rise to a new trend of applying rich neural architectures to solve classical combinatorial optimization problems. In this paper, we propose a neural framework that can learn to solve the Circuit Satisfiability problem. Our framework is built upon two fundamental contributions: a rich embedding architecture that encodes the problem structure, and an end-to-end differentiable training procedure that mimics Reinforcement Learning and trains the model directly toward solving the SAT problem. The experimental results show the superior out-of-sample generalization performance of our framework compared to the recently developed NeuroSAT method."
                },
                {
                    "title": "Physics-Inspired Optimization for Quadratic Unconstrained Problems Using a Digital Annealer",
                    "abstract": "The Fujitsu Digital Annealer is designed to solve fully connected quadratic unconstrained binary optimization (QUBO) problems. It is implemented on application-specific CMOS hardware and currently solves problems of up to 1024 variables. The Digital Annealer's algorithm is currently based on simulated annealing; however, it differs from it in its utilization of an efficient parallel-trial scheme and a dynamic escape mechanism. In addition, the Digital Annealer exploits the massive parallelization that custom application-specific CMOS hardware allows. We compare the performance of the Digital Annealer to simulated annealing and parallel tempering with isoenergetic cluster moves on two-dimensional and fully connected spin-glass problems with bimodal and Gaussian couplings. These represent the respective limits of sparse versus dense problems, as well as high-degeneracy versus low-degeneracy problems. Our results show that the Digital Annealer currently exhibits a time-to-solution speedup of roughly two orders of magnitude for fully connected spin-glass problems with bimodal or Gaussian couplings, over the single-core implementations of simulated annealing and parallel tempering Monte Carlo used in this study. The Digital Annealer does not appear to exhibit a speedup for sparse two-dimensional spin-glass problems, which we explain on theoretical grounds. We also benchmarked an early implementation of the Parallel Tempering Digital Annealer. Our results suggest an improved scaling over the other algorithms for fully connected problems of average difficulty with bimodal disorder. The next generation of the Digital Annealer is expected to be able to solve fully connected problems up to 8192 variables in size. This would enable the study of fundamental physics problems and industrial applications that were previously inaccessible using standard computing hardware or special-purpose quantum annealing machines."
                },
                {
                    "title": "Reinforcement Learning for Solving the Vehicle Routing Problem",
                    "abstract": "We present an end-to-end framework for solving the Vehicle Routing Problem (VRP) using reinforcement learning. In this approach, we train a single model that finds near-optimal solutions for problem instances sampled from a given distribution, only by observing the reward signals and following feasibility rules. Our model represents a parameterized stochastic policy, and by applying a policy gradient algorithm to optimize its parameters, the trained model produces the solution as a sequence of consecutive actions in real time, without the need to re-train for every new problem instance. On capacitated VRP, our approach outperforms classical heuristics and Google's OR-Tools on medium-sized instances in solution quality with comparable computation time (after training). We demonstrate how our approach can handle problems with split delivery and explore the effect of such deliveries on the solution quality. Our proposed framework can be applied to other variants of the VRP such as the stochastic VRP, and has the potential to be applied more generally to combinatorial optimization problems."
                },
                {
                    "title": "Learning a SAT Solver from Single-Bit Supervision",
                    "abstract": "We present NeuroSAT, a message passing neural network that learns to solve SAT problems after only being trained as a classifier to predict satisfiability. Although it is not competitive with state-of-the-art SAT solvers, NeuroSAT can solve problems that are substantially larger and more difficult than it ever saw during training by simply running for more iterations. Moreover, NeuroSAT generalizes to novel distributions; after training only on random SAT problems, at test time it can solve SAT problems encoding graph coloring, clique detection, dominating set, and vertex cover problems, all on a range of distributions over small random graphs."
                },
                {
                    "title": "IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures",
                    "abstract": "In this work we aim to solve a large collection of tasks using a single reinforcement learning agent with a single set of parameters. A key challenge is to handle the increased amount of data and extended training time. We have developed a new distributed agent IMPALA (Importance Weighted Actor-Learner Architecture) that not only uses resources more efficiently in single-machine training but also scales to thousands of machines without sacrificing data efficiency or resource utilisation. We achieve stable learning at high throughput by combining decoupled acting and learning with a novel off-policy correction method called V-trace. We demonstrate the effectiveness of IMPALA for multi-task reinforcement learning on DMLab-30 (a set of 30 tasks from the DeepMind Lab environment (Beattie et al., 2016)) and Atari-57 (all available Atari games in Arcade Learning Environment (Bellemare et al., 2013a)). Our results show that IMPALA is able to achieve better performance than previous agents with less data, and crucially exhibits positive transfer between tasks as a result of its multi-task approach."
                },
                {
                    "title": "Inductive Representation Learning on Large Graphs",
                    "abstract": "Low-dimensional embeddings of nodes in large graphs have proved extremely useful in a variety of prediction tasks, from content recommendation to identifying protein functions. However, most existing approaches require that all nodes in the graph are present during training of the embeddings; these previous approaches are inherently transductive and do not naturally generalize to unseen nodes. Here we present GraphSAGE, a general, inductive framework that leverages node feature information (e.g., text attributes) to efficiently generate node embeddings for previously unseen data. Instead of training individual embeddings for each node, we learn a function that generates embeddings by sampling and aggregating features from a node's local neighborhood. Our algorithm outperforms strong baselines on three inductive node-classification benchmarks: we classify the category of unseen nodes in evolving information graphs based on citation and Reddit post data, and we show that our algorithm generalizes to completely unseen graphs using a multi-graph dataset of protein-protein interactions."
                },
                {
                    "title": "Learning Combinatorial Optimization Algorithms over Graphs",
                    "abstract": "The design of good heuristics or approximation algorithms for NP-hard combinatorial optimization problems often requires significant specialized knowledge and trial-and-error. Can we automate this challenging, tedious process, and learn the algorithms instead? In many real-world applications, it is typically the case that the same optimization problem is solved again and again on a regular basis, maintaining the same problem structure but differing in the data. This provides an opportunity for learning heuristic algorithms that exploit the structure of such recurring problems. In this paper, we propose a unique combination of reinforcement learning and graph embedding to address this challenge. The learned greedy policy behaves like a meta-algorithm that incrementally constructs a solution, and the action is determined by the output of a graph embedding network capturing the current state of the solution. We show that our framework can be applied to a diverse range of optimization problems over graphs, and learns effective algorithms for the Minimum Vertex Cover, Maximum Cut and Traveling Salesman problems."
                },
                {
                    "title": "Neural Message Passing for Quantum Chemistry",
                    "abstract": "Supervised learning on molecules has incredible potential to be useful in chemistry, drug discovery, and materials science. Luckily, several promising and closely related neural network models invariant to molecular symmetries have already been described in the literature. These models learn a message passing algorithm and aggregation procedure to compute a function of their entire input graph. At this point, the next step is to find a particularly effective variant of this general approach and apply it to chemical prediction benchmarks until we either solve them or reach the limits of the approach. In this paper, we reformulate existing models into a single common framework we call Message Passing Neural Networks (MPNNs) and explore additional novel variations within this framework. Using MPNNs we demonstrate state of the art results on an important molecular property prediction benchmark; these results are strong enough that we believe future work should focus on datasets with larger molecules or more accurate ground truth labels."
                },
                {
                    "title": "Neural Combinatorial Optimization with Reinforcement Learning",
                    "abstract": "This paper presents a framework to tackle combinatorial optimization problems using neural networks and reinforcement learning. We focus on the traveling salesman problem (TSP) and train a recurrent network that, given a set of city coordinates, predicts a distribution over different city permutations. Using negative tour length as the reward signal, we optimize the parameters of the recurrent network using a policy gradient method. We compare learning the network parameters on a set of training graphs against learning them on individual test graphs. Despite the computational expense, without much engineering and heuristic designing, Neural Combinatorial Optimization achieves close to optimal results on 2D Euclidean graphs with up to 100 nodes. Applied to the KnapSack, another NP-hard problem, the same method obtains optimal solutions for instances with up to 200 items."
                },
                {
                    "title": "#Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning",
                    "abstract": "Count-based exploration algorithms are known to perform near-optimally when used in conjunction with tabular reinforcement learning (RL) methods for solving small discrete Markov decision processes (MDPs). It is generally thought that count-based methods cannot be applied in high-dimensional state spaces, since most states will only occur once. Recent deep RL exploration strategies are able to deal with high-dimensional continuous state spaces through complex heuristics, often relying on optimism in the face of uncertainty or intrinsic motivation. In this work, we describe a surprising finding: a simple generalization of the classic count-based approach can reach near state-of-the-art performance on various high-dimensional and/or continuous deep RL benchmarks. States are mapped to hash codes, which allows to count their occurrences with a hash table. These counts are then used to compute a reward bonus according to the classic count-based exploration theory. We find that simple hash functions can achieve surprisingly good results on many challenging tasks. Furthermore, we show that a domain-dependent learned hash code may further improve these results. Detailed analysis reveals important aspects of a good hash function: 1) having appropriate granularity and 2) encoding information relevant to solving the MDP. This exploration strategy achieves near state-of-the-art performance on both continuous control tasks and Atari 2600 games, hence providing a simple yet powerful baseline for solving MDPs that require considerable exploration."
                },
                {
                    "title": "Efficient Combinatorial Optimization Using Quantum Annealing",
                    "abstract": "The recent availability of the first commercial quantum computers has provided a promising tool to tackle NP hard problems which can only be solved heuristically with present techniques. However, it is unclear if the current state of quantum computing already provides a quantum advantage over the current state of the art in classical computing. This article assesses the performance of the D-Wave 2X quantum annealer on two NP hard graph problems, in particular clique finding and graph partitioning. For this, we provide formulations as Qubo and Ising Hamiltonians suitable for the quantum annealer and compare a variety of quantum solvers (Sapi, QBSolv, QSage provided by D-Wave Sys, Inc.) to current classical algorithms (METIS, Simulated Annealing, third-party clique finding and graph splitting heuristics) on certain test sets of graphs. We demonstrate that for small graph instances, classical methods still outperform the quantum annealer in terms of computing time, even though the quality of the best solutions obtained is comparable. Nevertheless, due to the limited problem size which can be embedded on the D-Wave 2X chip, the aforementioned finding applies to most of problems of general nature solvable on the quantum annealer. For instances specifically designed to fit the D-Wave 2X architecture, we observe substantial speed-ups in computing time over classical approaches."
                },
                {
                    "title": "Learning to Branch in Mixed Integer Programming",
                    "abstract": "\n \n The design of strategies for branching in Mixed Integer Programming (MIP) is guided by cycles of parameter tuning and offline experimentation on an extremely heterogeneous testbed, using the average performance. Once devised, these strategies (and their parameter settings) are essentially input-agnostic. To address these issues, we propose a machine learning (ML) framework for variable branching in MIP.Our method observes the decisions made by Strong Branching (SB), a time-consuming strategy that produces small search trees, collecting features that characterize the candidate branching variables at each node of the tree. Based on the collected data, we learn an easy-to-evaluate surrogate function that mimics the SB strategy, by means of solving a learning-to-rank problem, common in ML. The learned ranking function is then used for branching. The learning is instance-specific, and is performed on-the-fly while executing a branch-and-bound search to solve the MIP instance. Experiments on benchmark instances indicate that our method produces significantly smaller search trees than existing heuristics, and is competitive with a state-of-the-art commercial solver.\n \n"
                },
                {
                    "title": "Extremal Cuts of Sparse Random Graphs",
                    "abstract": "For Erd\u0151s\u2013Renyi random graphs with average degree \u03b3\u03b3, and uniformly random \u03b3\u03b3-regular graph on nn vertices, we prove that with high probability the size of both the Max-Cut and maximum bisection are n(\u03b34+P\u2217\u03b34\u2212\u2212\u221a+o(\u03b3\u2212\u2212\u221a))+o(n)n(\u03b34+P\u2217\u03b34+o(\u03b3))+o(n) while the size of the minimum bisection is n(\u03b34\u2212P\u2217\u03b34\u2212\u2212\u221a+o(\u03b3\u2212\u2212\u221a))+o(n)n(\u03b34\u2212P\u2217\u03b34+o(\u03b3))+o(n). Our derivation relates the free energy of the anti-ferromagnetic Ising model on such graphs to that of the Sherrington\u2013Kirkpatrick model, with P\u2217\u22480.7632P\u2217\u22480.7632 standing for the ground state energy of the latter, expressed analytically via Parisi\u2019s formula."
                },
                {
                    "title": "Adam: A Method for Stochastic Optimization",
                    "abstract": "We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
                },
                {
                    "title": "The hard-core model on random graphs revisited",
                    "abstract": "We revisit the classical hard-core model, also known as independent set and dual to vertex cover problem, where one puts particles with a first-neighbor hard-core repulsion on the vertices of a random graph. Although the case of random graphs with small and very large average degrees respectively are quite well understood, they yield qualitatively different results and our aim here is to reconciliate these two cases. We revisit results that can be obtained using the (heuristic) cavity method and show that it provides a closed-form conjecture for the exact density of the densest packing on random regular graphs with degree K \u2265 20, and that for K > 16 the nature of the phase transition is the same as for large K. This also shows that the hard-code model is the simplest mean-field lattice model for structural glasses and jamming."
                },
                {
                    "title": "Ising formulations of many NP problems",
                    "abstract": "We provide Ising formulations for many NP-complete and NP-hard problems, including all of Karp's 21 NP-complete problems. This collects and extends mappings to the Ising model from partitioning, covering and satisfiability. In each case, the required number of spins is at most cubic in the size of the problem. This work may be useful in designing adiabatic quantum optimization algorithms."
                },
                {
                    "title": "On independent sets in random graphs",
                    "abstract": "The independence number of a sparse random graph G(n,m) of average degree d = 2m/n is well\u2010known to be (2\u2212\u03b5d)nln(d)/d\u2264\u03b1(G(n,m))\u2264(2+\u03b5d)nln(d)/d with high probability, with \u03b5d\u21920 in the limit of large d. Moreover, a trivial greedy algorithm w.h.p. finds an independent set of size nln(d)/d , i.e., about half the maximum size. Yet in spite of 30 years of extensive research no efficient algorithm has emerged to produce an independent set with size (1+\u03b5)nln(d)/d for any fixed \u03b5>0 (independent of both d and n). In this paper we prove that the combinatorial structure of the independent set problem in random graphs undergoes a phase transition as the size k of the independent sets passes the point k\u223cnln(d)/d . Roughly speaking, we prove that independent sets of size k>(1+\u03b5)nln(d)/d form an intricately rugged landscape, in which local search algorithms seem to get stuck. We illustrate this phenomenon by providing an exponential lower bound for the Metropolis process, a Markov chain for sampling independent sets. \u00a9 2014 Wiley Periodicals, Inc. Random Struct. Alg., 47, 436\u2013486, 2015"
                },
                {
                    "title": "Combinatorial approach to the interpolation method and scaling limits in sparse random graphs",
                    "abstract": "We establish the existence of free energy limits for several sparse random hypergraph models corresponding to certain combinatorial models on Erdos-Renyi (ER) graph G(N,c/N) and random r-regular graph G(N,r).\n For a variety of models, including independent sets, MAX-CUT, Coloring and K-SAT, we prove that the free energy both at a positive and zero temperature, appropriately rescaled, converges to a limit as the size of the underlying graph diverges to infinity. In the zero temperature case, this is interpreted as the existence of the scaling limit for the corresponding combinatorial optimization problem.\n For example, as a special case we prove that the size of a largest independent set in these graphs, normalized by the number of nodes converges to a limit w.h.p., thus resolving an open problem, (see Conjecture 2.20 in Wormald's Random Graph Models, or Aldous's list of open problems.\n Our approach is based on extending and simplifying the interpolation method of Guerra and Toninelli. Among other applications, this method was used to prove the existence of free energy limits for Viana-Bray and K-SAT models on ER graphs. The case of zero temperature was treated by taking limits of positive temperature models. We provide instead a simpler combinatorial approach and work with the zero temperature case (optimization) directly both in the case of ER graph G(N,c/N) and random regular graph G(N,r).\n In addition we establish the large deviations principle for the satisfiability property for constraint satisfaction problems such as Coloring, K-SAT and NAE-K-SAT."
                },
                {
                    "title": "Collective Classification in Network Data",
                    "abstract": "Many real-world applications produce networked data such as the world-wide web (hypertext documents connected via hyperlinks), social networks (for example, people connected by friendship links), communication networks (computers connected via communication links) and biological networks (for example, protein interaction networks). A recent focus in machine learning research has been to extend traditional machine learning classification techniques to classify nodes in such networks. In this article, we provide a brief introduction to this area of research and how it has progressed during the past decade. We introduce four of the most widely used inference algorithms for classifying networked data and empirically compare them on both synthetic and real-world data."
                },
                {
                    "title": "Image recognition with an adiabatic quantum computer I. Mapping to quadratic unconstrained binary optimization",
                    "abstract": "Many artificial intelligence (AI) problems naturally map to NP-hard optimization problems. This has the interesting consequence that enabling human-level capability in machines often requires systems that can handle formally intractable problems. This issue can sometimes (but possibly not always) be resolved by building special-purpose heuristic algorithms, tailored to the problem in question. Because of the continued difficulties in automating certain tasks that are natural for humans, there remains a strong motivation for AI researchers to investigate and apply new algorithms and techniques to hard AI problems. Recently a novel class of relevant algorithms that require quantum mechanical hardware have been proposed. These algorithms, referred to as quantum adiabatic algorithms, represent a new approach to designing both complete and heuristic solvers for NP-hard optimization problems. In this work we describe how to formulate image recognition, which is a canonical NP-hard AI problem, as a Quadratic Unconstrained Binary Optimization (QUBO) problem. The QUBO format corresponds to the input format required for D-Wave superconducting adiabatic quantum computing (AQC) processors."
                },
                {
                    "title": "Improved approximation algorithms for maximum cut and satisfiability problems using semidefinite programming",
                    "abstract": "We present randomized approximation algorithms for the maximum cut (MAX CUT) and maximum 2-satisfiability (MAX 2SAT) problems that always deliver solutions of expected value at least.87856 times the optimal value. These algorithms use a simple and elegant technique that randomly rounds the solution to a nonlinear programming relaxation. This relaxation can be interpreted both as a semidefinite program and as an eigenvalue minimization problem. The best previously known approximation algorithms for these problems had performance guarantees of 1/2 for MAX CUT and 3/4 or MAX 2SAT. Slight extensions of our analysis lead to a.79607-approximation algorithm for the maximum directed cut problem (MAX DICUT) and a.758-approximation algorithm for MAX SAT, where the best previously known approximation algorithms had performance guarantees of 1/4 and 3/4, respectively. Our algorithm gives the first substantial progress in approximating MAX CUT in nearly twenty years, and represents the first use of semidefinite programming in the design of approximation algorithms."
                },
                {
                    "title": "0-1 Quadratic programming approach for optimum solutions of two scheduling problems",
                    "abstract": "Two scheduling problems are considered: (1) scheduling n jobs non-preemptively on a single machine to minimize total weighted earliness and tardiness (WET); (2) scheduling n jobs non-preemptively on two parallel identical processors to minimize weighted mean flow time. In the second problem, a pre-ordering of the jobs is assumed that must be satisfied for any set of jobs scheduled on each specific machine. Both problems are known to be NP-complete. A 0-1 quadratic assignment formulation of the problems is presented. An equivalent 0-1 mixed integer linear programming approach for the problems are considered and a numerical example is given. The formulations presented enable one to use optimal and heuristic available algorithms of 0-1 quadratic assignment for the problems considered here."
                },
                {
                    "title": "Combinatorial optimization:Algorithms and complexity",
                    "abstract": "Eventually, you will unquestionably discover a supplementary experience and achievement by spending more cash. yet when? complete you bow to that you require to get those all needs taking into consideration having significantly cash? Why dont you try to acquire something basic in the beginning? Thats something that will guide you to comprehend even more re the globe, experience, some places, gone history, amusement, and a lot more?"
                },
                {
                    "title": "Optimization by Simulated Annealing",
                    "abstract": "There is a deep and useful connection between statistical mechanics (the behavior of systems with many degrees of freedom in thermal equilibrium at a finite temperature) and multivariate or combinatorial optimization (finding the minimum of a given function depending on many parameters). A detailed analogy with annealing in solids provides a framework for optimization of the properties of very large and complex systems. This connection to statistical mechanics exposes new information and provides an unfamiliar perspective on traditional optimization problems and methods."
                },
                {
                    "title": "Frequency assignment: Theory and applications",
                    "abstract": "In this paper we introduce the minimum-order approach to frequency assignment and present a theory which relates this approach to the traditional one. This new approach is potentially more desirable than the traditional one. We model assignment problems as both frequency-distance constrained and frequency constrained optimization problems. The frequency constrained approach should be avoided if distance separation is employed to mitigate interference. A restricted class of graphs, called disk graphs, plays a central role in frequency-distance constrained problems. We introduce two generalizations of chromatic number and show that many frequency assignment problems are equivalent to generalized graph coloring problems. Using these equivalences and recent results concerning the complexity of graph coloring, we classify many frequency assignment problems according to the \"execution time efficiency\" of algorithms that may be devised for their solution. We discuss applications to important real world problems and identify areas for further work."
                },
                {
                    "title": "A sequence of approximated solutions to the S-K model for spin glasses",
                    "abstract": "In the framework of the new version of the replica theory, a sequence of approximated solutions is computed for the Sherrington-Kirkpatrick model (see Phys. Rev. Lett., vol.35, p.1972, 1975) of spin glasses."
                },
                {
                    "title": "Toward Unlimited Self-Learning Monte Carlo with Annealing Process Using VAE's Implicit Isometricity",
                    "abstract": "Self-learning Monte Carlo (SLMC) methods are recently proposed to accelerate Markov chain Monte Carlo (MCMC) methods by using a machine learning model. With generative models having latent variables, SLMC methods realize ef\ufb01cient Monte Carlo updates with less autocorrelation. However, SLMC methods are dif\ufb01cult to directly apply to multimodal distributions for which training data are dif\ufb01cult to obtain. In this paper, we propose a novel SLMC method called the \u201cannealing VAE-SLMC\u201d to drastically expand the range of applications. Our VAE-SLMC utilizes a variational autoencoder (VAE) as a generative model to make ef\ufb01cient parallel proposals independent of any previous state by applying the theoretically derived implicit isometricity of the VAE. We combine an adaptive annealing process to the VAE-SLMC, making our method applicable to the cases where obtaining unbiased training data is dif\ufb01cult in practical sense due to slow mixing. We also propose a parallel annealing process and an exchange process between chains to make the annealing operation more precise and ef\ufb01cient. Experiments validate that our method can pro\ufb01ciently obtain unbiased samples from multiple multimodal toy distributions and practical multimodal posterior distributions, which is dif\ufb01cult to achieve with the existing SLMC meth-ods."
                },
                {
                    "title": "Learning Local Search Heuristics for Boolean Satisfiability",
                    "abstract": "We present an approach to learn SAT solver heuristics from scratch through deep reinforcement learning with a curriculum. In particular, we incorporate a graph neural network in a stochastic local search algorithm to act as the variable selection heuristic. We consider Boolean satisfiability problems from different classes and learn specialized heuristics for each class. Although we do not aim to compete with the state-of-the-art SAT solvers in run time, we demonstrate that the learned heuristics allow us to find satisfying assignments in fewer steps compared to a generic heuristic, and we provide analysis of our results through experiments."
                },
                {
                    "title": "UvA-DARE ( Digital Academic Repository ) Attention , learn to solve routing problems !",
                    "abstract": "The recently presented idea to learn heuristics for combinatorial optimization problems is promising as it can save costly development. However, to push this idea towards practical implementation, we need better models and better ways of training. We contribute in both directions: we propose a model based on attention layers with benefits over the Pointer Network and we show how to train this model using REINFORCE with a simple baseline based on a deterministic greedy rollout, which we find is more efficient than using a value function. We significantly improve over recent learned heuristics for the Travelling Salesman Problem (TSP), getting close to optimal results for problems up to 100 nodes. With the same hyperparameters, we learn strong heuristics for two variants of the Vehicle Routing Problem (VRP), the Orienteering Problem (OP) and (a stochastic variant of) the Prize Collecting TSP (PCTSP), outperforming a wide range of baselines and getting results close to highly optimized and specialized algorithms."
                },
                {
                    "title": "The Graph Neural Network Model",
                    "abstract": "Many underlying relationships among data in several areas of science and engineering, e.g., computer vision, molecular chemistry, molecular biology, pattern recognition, and data mining, can be represented in terms of graphs. In this paper, we propose a new neural network model, called graph neural network (GNN) model, that extends existing neural network methods for processing the data represented in graph domains. This GNN model, which can directly process most of the practically useful types of graphs, e.g., acyclic, cyclic, directed, and undirected, implements a function tau(G,n) isin IRm that maps a graph G and one of its nodes n into an m-dimensional Euclidean space. A supervised learning algorithm is derived to estimate the parameters of the proposed GNN model. The computational cost of the proposed algorithm is also considered. Some experimental results are shown to validate the proposed learning algorithm, and to demonstrate its generalization capabilities."
                },
                {
                    "title": "Adiabatic Quantum Computation",
                    "abstract": "This chapter contains sections titled: Adiabatic Processes Adiabatic Quantum Computing Exercises"
                },
                {
                    "title": "Penalty functions",
                    "abstract": "This section begins with the motivation and general form of penalty functions as used in evolutionary computation. The main types of penalty function\u2014constant, static, dynamic, and adaptive\u2014are described within a common notation framework. References from the literature concerning these exterior penalty approaches are presented. The section concludes with a brief discussion of promising areas of future research in penalty methods for constrained optimization by evolutionary computation."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG",
                "stat.CO",
                "stat.ME"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the optimization stability and robustness of unsupervised learning-based solvers for combinatorial optimization problems, particularly in overcoming local optima and addressing the challenges of artificial rounding from continuous solutions to discrete solutions?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of combinatorial optimization, as it can lead to more reliable and efficient algorithms that can handle large-scale problems in real-world applications. By improving the performance of unsupervised learning-based solvers, we can enhance their applicability across various domains, such as logistics, network design, and resource allocation. This research could pave the way for future studies to explore more sophisticated learning techniques and hybrid approaches, ultimately contributing to the development of more effective optimization tools.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent instability of reinforcement learning-based solvers and the optimization issues faced by unsupervised learning methods. Naive approaches may fail due to their inability to escape local optima, leading to suboptimal solutions. Additionally, the ambiguity in the continuous relaxation approach complicates the transition from soft to discrete solutions, as existing rounding methods often lack robustness and can result in poor performance. Overcoming these technical and theoretical obstacles requires innovative strategies that effectively balance continuity and discreteness in the optimization process.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either supervised or reinforcement learning approaches, which have their own limitations, such as the need for optimal solutions or unstable training dynamics. While unsupervised learning-based solvers have shown promise, they have not adequately addressed the issues of local optima and rounding robustness. Barriers such as a lack of comprehensive methodologies to control the trade-off between continuous and discrete solutions have hindered progress. Our approach differs by introducing the Continuous Relaxation Annealing (CRA) strategy, which aims to enhance the optimization process and improve the overall performance of these solvers.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the Continuous Relaxation Annealing (CRA) strategy, which incorporates a penalty term to manage the continuity and discreteness of relaxed variables. We will evaluate the effectiveness of this approach using benchmark combinatorial optimization datasets, measuring performance through metrics such as solution quality and convergence stability. The expected outcomes include improved optimization stability, enhanced robustness in rounding"
            }
        },
        "author_data": {
            "3c019766-d018-4aca-aea3-e906d5b6dbdb": {
                "pk": "3c019766-d018-4aca-aea3-e906d5b6dbdb",
                "name": "Yuma Ichikawa",
                "collaborators": [
                    "Koji Hukushima",
                    "Yamato Arai",
                    "Hiroaki Iwashita",
                    "Nobuo Namura",
                    "Akira Nakagawa",
                    "Hiromoto Masayuki",
                    "Yuhei Umeda",
                    "Yuichi Ishida",
                    "Aki Dote",
                    "Toshiyuki Miyazawa"
                ],
                "domain": [
                    "Deep Learning",
                    "Variational Inference",
                    "Combinatorial Optimization",
                    "Anomaly Detection"
                ],
                "publications": [
                    {
                        "title": "Statistical-mechanical study of deep Boltzmann machine given weight parameters after training by singular value decomposition",
                        "abstract": "Deep learning methods relying on multi-layered networks have been actively studied in a wide range of fields in recent years, and deep Boltzmann machines(DBMs) is one of them. In this study, a model of DBMs with some properites of weight parameters obtained by learning is studied theoretically by a statistical-mechanical approach based on the replica method. The phases characterizing the role of DBMs as a generator and their phase diagram are derived, depending meaningfully on the numbers of layers and units in each layer. It is found, in particular, that the correlation between the weight parameters in the hidden layers plays an essential role and that an increase in the layer number has a negative effect on DBM's performance as a generator when the correlation is smaller than a certain threshold value."
                    },
                    {
                        "title": "Dataset Size Dependence of Rate-Distortion Curve and Threshold of Posterior Collapse in Linear VAE",
                        "abstract": "In the Variational Autoencoder (VAE), the variational posterior often aligns closely with the prior, which is known as posterior collapse and hinders the quality of representation learning. To mitigate this problem, an adjustable hyperparameter beta has been introduced in the VAE. This paper presents a closed-form expression to assess the relationship between the beta in VAE, the dataset size, the posterior collapse, and the rate-distortion curve by analyzing a minimal VAE in a high-dimensional limit. These results clarify that a long plateau in the generalization error emerges with a relatively larger beta. As the beta increases, the length of the plateau extends and then becomes infinite beyond a certain beta threshold. This implies that the choice of beta, unlike the usual regularization parameters, can induce posterior collapse regardless of the dataset size. Thus, beta is a risky parameter that requires careful tuning. Furthermore, considering the dataset-size dependence on the rate-distortion curve, a relatively large dataset is required to obtain a rate-distortion curve with high rates. Extensive numerical experiments support our analysis."
                    },
                    {
                        "title": "Learning Dynamics in Linear VAE: Posterior Collapse Threshold, Superfluous Latent Space Pitfalls, and Speedup with KL Annealing",
                        "abstract": "Variational autoencoders (VAEs) face a notorious problem wherein the variational posterior often aligns closely with the prior, a phenomenon known as posterior collapse, which hinders the quality of representation learning. To mitigate this problem, an adjustable hyperparameter $\\beta$ and a strategy for annealing this parameter, called KL annealing, are proposed. This study presents a theoretical analysis of the learning dynamics in a minimal VAE. It is rigorously proved that the dynamics converge to a deterministic process within the limit of large input dimensions, thereby enabling a detailed dynamical analysis of the generalization error. Furthermore, the analysis shows that the VAE initially learns entangled representations and gradually acquires disentangled representations. A fixed-point analysis of the deterministic process reveals that when $\\beta$ exceeds a certain threshold, posterior collapse becomes inevitable regardless of the learning period. Additionally, the superfluous latent variables for the data-generative factors lead to overfitting of the background noise; this adversely affects both generalization and learning convergence. The analysis further unveiled that appropriately tuned KL annealing can accelerate convergence."
                    },
                    {
                        "title": "Optimization by Parallel Quasi-Quantum Annealing with Gradient-Based Sampling",
                        "abstract": "Learning-based methods have gained attention as general-purpose solvers due to their ability to automatically learn problem-specific heuristics, reducing the need for manually crafted heuristics. However, these methods often face scalability challenges. To address these issues, the improved Sampling algorithm for Combinatorial Optimization (iSCO), using discrete Langevin dynamics, has been proposed, demonstrating better performance than several learning-based solvers. This study proposes a different approach that integrates gradient-based update through continuous relaxation, combined with Quasi-Quantum Annealing (QQA). QQA smoothly transitions the objective function, starting from a simple convex function, minimized at half-integral values, to the original objective function, where the relaxed variables are minimized only in the discrete space. Furthermore, we incorporate parallel run communication leveraging GPUs to enhance exploration capabilities and accelerate convergence. Numerical experiments demonstrate that our method is a competitive general-purpose solver, achieving performance comparable to iSCO and learning-based solvers across various benchmark problems. Notably, our method exhibits superior speed-quality trade-offs for large-scale instances compared to iSCO, learning-based solvers, commercial solvers, and specialized algorithms."
                    },
                    {
                        "title": "Statistical Mechanics of Min-Max Problems",
                        "abstract": "Min-max optimization problems, also known as saddle point problems, have attracted significant attention due to their applications in various fields, such as fair beamforming, generative adversarial networks (GANs), and adversarial learning. However, understanding the properties of these min-max problems has remained a substantial challenge. This study introduces a statistical mechanical formalism for analyzing the equilibrium values of min-max problems in the high-dimensional limit, while appropriately addressing the order of operations for min and max. As a first step, we apply this formalism to bilinear min-max games and simple GANs, deriving the relationship between the amount of training data and generalization error and indicating the optimal ratio of fake to real data for effective learning. This formalism provides a groundwork for a deeper theoretical analysis of the equilibrium properties in various machine learning methods based on min-max problems and encourages the development of new algorithms and architectures."
                    },
                    {
                        "title": "Continuous Tensor Relaxation for Finding Diverse Solutions in Combinatorial Optimization Problems",
                        "abstract": "Finding the best solution is a common objective in combinatorial optimization (CO). In practice, directly handling constraints is often challenging, incorporating them into the objective function as the penalties. However, balancing these penalties to achieve the desired solution is time-consuming. Additionally, formulated objective functions and constraints often only approximate real-world scenarios, where the optimal solution is not necessarily the best solution for the original real-world problem. One solution is to obtain (i) penalty-diversified solutions with varying penalty strengths for the former issue and (ii) variation-diversified solutions with different characteristics for the latter issue. Users can then post-select the desired solution from these diverse solutions. However, efficiently finding these diverse solutions is more difficult than identifying one. This study introduces Continual Tensor Relaxation Annealing (CTRA) for unsupervised-learning (UL)-based CO solvers, a computationally efficient framework for finding these diverse solutions in a single training run. The key idea is to leverage representation learning capability to automatically and efficiently learn common representations and parallelization. Numerical experiments show that CTRA enables UL-based solvers to find these diverse solutions much faster than repeatedly running existing UL-based solvers."
                    },
                    {
                        "title": "Training-Free Time-Series Anomaly Detection: Leveraging Image Foundation Models",
                        "abstract": "Recent advancements in time-series anomaly detection have relied on deep learning models to handle the diverse behaviors of time-series data. However, these models often suffer from unstable training and require extensive hyperparameter tuning, leading to practical limitations. Although foundation models present a potential solution, their use in time series is limited. To overcome these issues, we propose an innovative image-based, training-free time-series anomaly detection (ITF-TAD) approach. ITF-TAD converts time-series data into images using wavelet transform and compresses them into a single representation, leveraging image foundation models for anomaly detection. This approach achieves high-performance anomaly detection without unstable neural network training or hyperparameter tuning. Furthermore, ITF-TAD identifies anomalies across different frequencies, providing users with a detailed visualization of anomalies and their corresponding frequencies. Comprehensive experiments on five benchmark datasets, including univariate and multivariate time series, demonstrate that ITF-TAD offers a practical and effective solution with performance exceeding or comparable to that of deep models."
                    },
                    {
                        "title": "Toward Unlimited Self-Learning MCMC with Parallel Adaptive Annealing",
                        "abstract": "Self-learning Monte Carlo (SLMC) methods are recently proposed to accelerate Markov chain Monte Carlo (MCMC) methods using a machine learning model. With latent generative models, SLMC methods realize efficient Monte Carlo updates with less autocorrelation. However, SLMC methods are difficult to directly apply to multimodal distributions for which training data are difficult to obtain. To solve the limitation, we propose parallel adaptive annealing, which makes SLMC methods directly apply to multimodal distributions with a gradually trained proposal while annealing target distribution. Parallel adaptive annealing is based on (i) sequential learning with annealing to inherit and update the model parameters, (ii) adaptive annealing to automatically detect under-learning, and (iii) parallel annealing to mitigate mode collapse of proposal models. We also propose VAE-SLMC method which utilizes a variational autoencoder (VAE) as a proposal of SLMC to make efficient parallel proposals independent of any previous state using recently clarified quantitative properties of VAE. Experiments validate that our method can proficiently obtain accurate samples from multiple multimodal toy distributions and practical multimodal posterior distributions, which is difficult to achieve with the existing SLMC methods."
                    },
                    {
                        "title": "Adaptive Flip Graph Algorithm for Matrix Multiplication",
                        "abstract": "This study proposes the \"adaptive flip graph algorithm\", which combines adaptive searches with the flip graph algorithm for finding fast and efficient methods for matrix multiplication. The adaptive flip graph algorithm addresses the inherent limitations of exploration and inefficient search encountered in the original flip graph algorithm, particularly when dealing with large matrix multiplication. For the limitation of exploration, the proposed algorithm adaptively transitions over the flip graph, introducing a flexibility that does not strictly reduce the number of multiplications. Concerning the issue of inefficient search in large instances, the proposed algorithm adaptively constraints the search range instead of relying on a completely random search, facilitating more effective exploration. Numerical experimental results demonstrate the effectiveness of the adaptive flip graph algorithm, showing a reduction in the number of multiplications for a $4\\times 5$ matrix multiplied by a $5\\times 5$ matrix from $76$ to $73$, and that from $95$ to $94$ for a $5 \\times 5$ matrix multiplied by another $5\\times 5$ matrix. These results are obtained in characteristic two."
                    },
                    {
                        "title": "Ratio Divergence Learning Using Target Energy in Restricted Boltzmann Machines: Beyond Kullback--Leibler Divergence Learning",
                        "abstract": "We propose ratio divergence (RD) learning for discrete energy-based models, a method that utilizes both training data and a tractable target energy function. We apply RD learning to restricted Boltzmann machines (RBMs), which are a minimal model that satisfies the universal approximation theorem for discrete distributions. RD learning combines the strength of both forward and reverse Kullback-Leibler divergence (KLD) learning, effectively addressing the \"notorious\" issues of underfitting with the forward KLD and mode-collapse with the reverse KLD. Since the summation of forward and reverse KLD seems to be sufficient to combine the strength of both approaches, we include this learning method as a direct baseline in numerical experiments to evaluate its effectiveness. Numerical experiments demonstrate that RD learning significantly outperforms other learning methods in terms of energy function fitting, mode-covering, and learning stability across various discrete energy-based models. Moreover, the performance gaps between RD learning and the other learning methods become more pronounced as the dimensions of target models increase."
                    }
                ]
            }
        }
    },
    "2405.15894": {
        "paper_data": {
            "title": "Derivatives of Stochastic Gradient Descent",
            "url": "http://arxiv.org/abs/2405.15894v1",
            "arxiv_id": "2405.15894",
            "authors": [
                "Franck Iutzeler",
                "Edouard Pauwels",
                "Samuel Vaiter"
            ],
            "abstract": "We consider stochastic optimization problems where the objective depends on some parameter, as commonly found in hyperparameter optimization for instance. We investigate the behavior of the derivatives of the iterates of Stochastic Gradient Descent (SGD) with respect to that parameter and show that they are driven by an inexact SGD recursion on a different objective function, perturbed by the convergence of the original SGD. This enables us to establish that the derivatives of SGD converge to the derivative of the solution mapping in terms of mean squared error whenever the objective is strongly convex. Specifically, we demonstrate that with constant step-sizes, these derivatives stabilize within a noise ball centered at the solution derivative, and that with vanishing step-sizes they exhibit $O(\\log(k)^2 / k)$ convergence rates. Additionally, we prove exponential convergence in the interpolation regime. Our theoretical findings are illustrated by numerical experiments on synthetic tasks.",
            "introduction": "   1 Introduction  The differentiation of iterative algorithms has been a subject of research since the 1990s\u00a0(Gilbert, 1992; Christianson, 1994; Beck, 1994), and was succinctly described as \u201cpiggyback differentiation\u201d by Griewank and Faure (2003). This idea has gained renewed interest within the machine learning community, particularly for applications such as hyperparameter optimization\u00a0(Maclaurin et\u00a0al., 2015; Franceschi et\u00a0al., 2017), meta-learning\u00a0(Finn et\u00a0al., 2017; Rajeswaran et\u00a0al., 2019), and learning discretization of total variation\u00a0(Chambolle and Pock, 2021; Bogensperger et\u00a0al., 2022). When applied to an optimization problem, an important theoretical concern is the convergence of the derivatives of iterates to the derivatives of the solution. Traditional guarantees focus on asymptotic convergence to the solution derivative, as described by the implicit function theorem\u00a0(Gilbert, 1992; Christianson, 1994; Beck, 1994). This issue has inspired recent works for smooth optimization algorithms\u00a0(Mehmood and Ochs, 2020, 2022), generic nonsmooth iterations\u00a0(Bolte et\u00a0al., 2022), and second-order methods\u00a0(Bolte et\u00a0al., 2023).   Convergence analysis of iterative processes have predominantly focused on deterministic algorithms such as the gradient descent. In this work, we extend these results in the context of strongly convex parametric optimization by studying the iterative differentiation of the Stochastic Gradient Descent (SGD) algorithm. Since the seminal work of Robbins and Monro (1951), SGD has been a workhorse of stochastic optimization and is extensively employed in training various machine learning models\u00a0(Bottou et\u00a0al., 2018; Gower et\u00a0al., 2019). A critical aspect of our work is based on the fact that the sequence of iterative derivatives in this stochastic setting is itself a stochastic gradient sequence.   The goal of this work is to answer the following question:  What is the dynamics of the derivatives of the iterates of stochastic gradient descent in the context of minimization of parametric strongly convex functions?  Our motivation for studying this question is twofold. First, while iterative differentiation through SGD sequences is possibly not the most efficient way to differentiate solutions of convex programs, it is a very natural in the context of differentiable programming and has already been explored by practitioners. Second, existing attempts to provide stochastic approximation based solutions to differentiate through convex programming solutions require more intricate algorithmic schemes than the conceptually simple iterative differentiation of SGD. Despite its conceptual simplicity, the answer to this question is not direct in the first place due to the joint effect of noise on the iterate sequence and its derivatives.   Contributions.  The strongly convex setting ensures that the solution mapping is single valued and differentiable under appropriate smoothness assumptions. In this setting, we prove in Section\u00a02.3 the convergence of the derivatives of the SGD recursion toward the derivative of the solution mapping, in the sense of mean squared errors: \u2219\u2219\\bullet\u2219 We first provide a general result for non-increasing step-sizes converging to some \u03b7\u22650\ud835\udf020{\\eta}\\geq 0italic_\u03b7 \u2265 0 (covering constant step-sizes schedules), for which we prove that the derivatives of SGD eventually fluctuate in a ball centered at the solution derivative, of size proportional to \u03b7\ud835\udf02\\sqrt{{\\eta}}square-root start_ARG italic_\u03b7 end_ARG. \u2219\u2219\\bullet\u2219 With vanishing steps, this result implies that the derivatives of SGD converge toward the solution derivatives, and we obtain O(log(k)2/k)O(\\log(k)^{2}/k)italic_O ( roman_log ( italic_k ) start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT / italic_k ) convergence rates for O\u2062(1/k)\ud835\udc421\ud835\udc58O(1/k)italic_O ( 1 / italic_k ) step-size decay schedules. \u2219\u2219\\bullet\u2219 We also",
            "references": [
                {
                    "title": "Nonsmooth Implicit Differentiation: Deterministic and Stochastic Convergence Rates",
                    "abstract": "We study the problem of efficiently computing the derivative of the fixed-point of a parametric nondifferentiable contraction map. This problem has wide applications in machine learning, including hyperparameter optimization, meta-learning and data poisoning attacks. We analyze two popular approaches: iterative differentiation (ITD) and approximate implicit differentiation (AID). A key challenge behind the nonsmooth setting is that the chain rule does not hold anymore. We build upon the work by Bolte et al. (2022), who prove linear convergence of nonsmooth ITD under a piecewise Lipschitz smooth assumption. In the deterministic case, we provide a linear rate for AID and an improved linear rate for ITD which closely match the ones for the smooth setting. We further introduce NSID, a new stochastic method to compute the implicit derivative when the contraction map is defined as the composition of an outer map and an inner map which is accessible only through a stochastic unbiased estimator. We establish rates for the convergence of NSID, encompassing the best available rates in the smooth setting. We also present illustrative experiments confirming our analysis."
                },
                {
                    "title": "One-step differentiation of iterative algorithms",
                    "abstract": "In appropriate frameworks, automatic differentiation is transparent to the user at the cost of being a significant computational burden when the number of operations is large. For iterative algorithms, implicit differentiation alleviates this issue but requires custom implementation of Jacobian evaluation. In this paper, we study one-step differentiation, also known as Jacobian-free backpropagation, a method as easy as automatic differentiation and as performant as implicit differentiation for fast algorithms (e.g., superlinear optimization methods). We provide a complete theoretical approximation analysis with specific examples (Newton's method, gradient descent) along with its consequences in bilevel optimization. Several numerical examples illustrate the well-foundness of the one-step estimator."
                },
                {
                    "title": "Stochastic Approximation Beyond Gradient for Signal Processing and Machine Learning",
                    "abstract": "Stochastic Approximation (SA) is a classical algorithm that has had since the early days a huge impact on signal processing, and nowadays on machine learning, due to the necessity to deal with a large amount of data observed with uncertainties. An exemplar special case of SA pertains to the popular stochastic (sub)gradient algorithm which is the working horse behind many important applications. A lesser-known fact is that the SA scheme also extends to non-stochastic-gradient algorithms such as compressed stochastic gradient, stochastic expectation-maximization, and a number of reinforcement learning algorithms. The aim of this article is to overview and introduce the non-stochastic-gradient perspectives of SA to the signal processing and machine learning audiences through presenting a design guideline of SA algorithms backed by theories. Our central theme is to propose a general framework that unifies existing theories of SA, including its non-asymptotic and asymptotic convergence results, and demonstrate their applications on popular non-stochastic-gradient algorithms. We build our analysis framework based on classes of Lyapunov functions that satisfy a variety of mild conditions. We draw connections between non-stochastic-gradient algorithms and scenarios when the Lyapunov function is smooth, convex, or strongly convex. Using the said framework, we illustrate the convergence properties of the non-stochastic-gradient algorithms using concrete examples. Extensions to the emerging variance reduction techniques for improved sample complexity will also be discussed."
                },
                {
                    "title": "A Lower Bound and a Near-Optimal Algorithm for Bilevel Empirical Risk Minimization",
                    "abstract": "Bilevel optimization problems, which are problems where two optimization problems are nested, have more and more applications in machine learning. In many practical cases, the upper and the lower objectives correspond to empirical risk minimization problems and therefore have a sum structure. In this context, we propose a bilevel extension of the celebrated SARAH algorithm. We demonstrate that the algorithm requires $\\mathcal{O}((n+m)^{\\frac12}\\varepsilon^{-1})$ oracle calls to achieve $\\varepsilon$-stationarity with $n+m$ the total number of samples, which improves over all previous bilevel algorithms. Moreover, we provide a lower bound on the number of oracle calls required to get an approximate stationary point of the objective function of the bilevel problem. This lower bound is attained by our algorithm, making it optimal in terms of sample complexity."
                },
                {
                    "title": "Fixed-Point Automatic Differentiation of Forward-Backward Splitting Algorithms for Partly Smooth Functions",
                    "abstract": "A large class of non-smooth practical optimization problems can be written as minimization of a sum of smooth and partly smooth functions. We examine such structured problems which also depend on a parameter vector and study the problem of differentiating its solution mapping with respect to the parameter which has far reaching applications in sensitivity analysis and parameter learning problems. Under partial smoothness and other mild assumptions, we apply Implicit (ID) and Automatic Differentiation (AD) to the fixed-point iterations of proximal splitting algorithms. We show that AD of the sequence generated by these algorithms converges (linearly under further assumptions) to the derivative of the solution mapping. For a variant of automatic differentiation, which we call Fixed-Point Automatic Differentiation (FPAD), we remedy the memory overhead problem of the Reverse Mode AD and moreover provide faster convergence theoretically. We numerically illustrate the convergence and convergence rates of AD and FPAD on Lasso and Group Lasso problems and demonstrate the working of FPAD on prototypical image denoising problems by learning the regularization term."
                },
                {
                    "title": "Convergence of a Piggyback-Style Method for the Differentiation of Solutions of Standard Saddle-Point Problems",
                    "abstract": ". We analyse a \u201cpiggyback\u201d-style method for computing the derivative of a loss which depends on the solution of a convex-concave saddle point problems, with respect to the bilinear term. We attempt to derive guarantees for the algorithm under minimal regularity assumption on the functions. Our \ufb01nal convergence results include possibly nonsmooth objectives. We illustrate the versatility of the proposed piggyback algorithm by learning optimized shearlet transforms, which are a class of popu-lar sparsifying transforms in the \ufb01eld of imaging."
                },
                {
                    "title": "Automatic differentiation of nonsmooth iterative algorithms",
                    "abstract": "Differentiation along algorithms, i.e., piggyback propagation of derivatives, is now routinely used to differentiate iterative solvers in differentiable programming. Asymptotics is well understood for many smooth problems but the nondifferentiable case is hardly considered. Is there a limiting object for nonsmooth piggyback automatic differentiation (AD)? Does it have any variational meaning and can it be used effectively in machine learning? Is there a connection with classical derivative? All these questions are addressed under appropriate nonexpansivity conditions in the framework of conservative derivatives which has proved useful in understanding nonsmooth AD. For nonsmooth piggyback iterations, we characterize the attractor set of nonsmooth piggyback iterations as a set-valued fixed point which remains in the conservative framework. This has various consequences and in particular almost everywhere convergence of classical derivatives. Our results are illustrated on parametric convex optimization problems with forward-backward, Douglas-Rachford and Alternating Direction of Multiplier algorithms as well as the Heavy-Ball method."
                },
                {
                    "title": "Bilevel Optimization with a Lower-level Contraction: Optimal Sample Complexity without Warm-Start",
                    "abstract": "We analyse a general class of bilevel problems, in which the upper-level problem consists in the minimization of a smooth objective function and the lower-level problem is to find the fixed point of a smooth contraction map. This type of problems include instances of meta-learning, equilibrium models, hyperparameter optimization and data poisoning adversarial attacks. Several recent works have proposed algorithms which warm-start the lower-level problem, i.e.~they use the previous lower-level approximate solution as a staring point for the lower-level solver. This warm-start procedure allows one to improve the sample complexity in both the stochastic and deterministic settings, achieving in some cases the order-wise optimal sample complexity. However, there are situations, e.g., meta learning and equilibrium models, in which the warm-start procedure is not well-suited or ineffective. In this work we show that without warm-start, it is still possible to achieve order-wise (near) optimal sample complexity. In particular, we propose a simple method which uses (stochastic) fixed point iterations at the lower-level and projected inexact gradient descent at the upper-level, that reaches an $\\epsilon$-stationary point using $O(\\epsilon^{-2})$ and $\\tilde{O}(\\epsilon^{-1})$ samples for the stochastic and the deterministic setting, respectively. Finally, compared to methods using warm-start, our approach yields a simpler analysis that does not need to study the coupled interactions between the upper-level and lower-level iterates."
                },
                {
                    "title": "A framework for bilevel optimization that enables stochastic and global variance reduction algorithms",
                    "abstract": "Bilevel optimization, the problem of minimizing a value function which involves the arg-minimum of another function, appears in many areas of machine learning. In a large scale empirical risk minimization setting where the number of samples is huge, it is crucial to develop stochastic methods, which only use a few samples at a time to progress. However, computing the gradient of the value function involves solving a linear system, which makes it difficult to derive unbiased stochastic estimates. To overcome this problem we introduce a novel framework, in which the solution of the inner problem, the solution of the linear system, and the main variable evolve at the same time. These directions are written as a sum, making it straightforward to derive unbiased estimates. The simplicity of our approach allows us to develop global variance reduction algorithms, where the dynamics of all variables is subject to variance reduction. We demonstrate that SABA, an adaptation of the celebrated SAGA algorithm in our framework, has $O(\\frac1T)$ convergence rate, and that it achieves linear convergence under Polyak-Lojasciewicz assumption. This is the first stochastic algorithm for bilevel optimization that verifies either of these properties. Numerical experiments validate the usefulness of our method."
                },
                {
                    "title": "Amortized Implicit Differentiation for Stochastic Bilevel Optimization",
                    "abstract": "We study a class of algorithms for solving bilevel optimization problems in both stochastic and deterministic settings when the inner-level objective is strongly convex. Specifically, we consider algorithms based on inexact implicit differentiation and we exploit a warm-start strategy to amortize the estimation of the exact gradient. We then introduce a unified theoretical framework inspired by the study of singularly perturbed systems (Habets, 1974) to analyze such amortized algorithms. By using this framework, our analysis shows these algorithms to match the computational complexity of oracle methods that have access to an unbiased estimate of the gradient, thus outperforming many existing results for bilevel optimization. We illustrate these findings on synthetic experiments and demonstrate the efficiency of these algorithms on hyper-parameter optimization experiments involving several thousands of variables."
                },
                {
                    "title": "Efficient and Modular Implicit Differentiation",
                    "abstract": "Automatic differentiation (autodiff) has revolutionized machine learning. It allows to express complex computations by composing elementary ones in creative ways and removes the burden of computing their derivatives by hand. More recently, differentiation of optimization problem solutions has attracted widespread attention with applications such as optimization layers, and in bi-level problems such as hyper-parameter optimization and meta-learning. However, so far, implicit differentiation remained difficult to use for practitioners, as it often required case-by-case tedious mathematical derivations and implementations. In this paper, we propose automatic implicit differentiation, an efficient and modular approach for implicit differentiation of optimization problems. In our approach, the user defines directly in Python a function $F$ capturing the optimality conditions of the problem to be differentiated. Once this is done, we leverage autodiff of $F$ and the implicit function theorem to automatically differentiate the optimization problem. Our approach thus combines the benefits of implicit differentiation and autodiff. It is efficient as it can be added on top of any state-of-the-art solver and modular as the optimality condition specification is decoupled from the implicit differentiation mechanism. We show that seemingly simple principles allow to recover many existing implicit differentiation methods and create new ones easily. We demonstrate the ease of formulating and solving bi-level optimization problems using our framework. We also showcase an application to the sensitivity analysis of molecular dynamics."
                },
                {
                    "title": "Convergence Properties of Stochastic Hypergradients",
                    "abstract": "Bilevel optimization problems are receiving increasing attention in machine learning as they provide a natural framework for hyperparameter optimization and meta-learning. A key step to tackle these problems in the design of optimization algorithms for bilevel optimization is the efficient computation of the gradient of the upper-level objective (hypergradient). In this work, we study stochastic approximation schemes for the hypergradient, which are important when the lower-level problem is empirical risk minimization on a large dataset. We provide iteration complexity bounds for the mean square error of the hypergradient approximation, under the assumption that the lower-level problem is accessible only through a stochastic mapping which is a contraction in expectation. Preliminary numerical experiments support our theoretical analysis."
                },
                {
                    "title": "Learning Consistent Discretizations of the Total Variation",
                    "abstract": "In this work, we study a general framework of discrete approximations of the total variation for image reconstruction problems. The framework, for which we can show consistency in the sense of \u0393-convergence, unifies and extends several existing discretization schemes. In addition, we propose algorithms for learning discretizations of the total variation in order to achieve the best possible reconstruction quality for particular image reconstruction tasks. Interestingly, the learned discretizations significantly differ between the tasks, illustrating that there is no universal best discretization of the total variation."
                },
                {
                    "title": "Bilevel Optimization: Convergence Analysis and Enhanced Design",
                    "abstract": "Bilevel optimization has arisen as a powerful tool for many machine learning problems such as meta-learning, hyperparameter optimization, and reinforcement learning. In this paper, we investigate the nonconvex-strongly-convex bilevel optimization problem. For deterministic bilevel optimization, we provide a comprehensive convergence rate analysis for two popular algorithms respectively based on approximate implicit differentiation (AID) and iterative differentiation (ITD). For the AID-based method, we orderwisely improve the previous convergence rate analysis due to a more practical parameter selection as well as a warm start strategy, and for the ITD-based method we establish the first theoretical convergence rate. Our analysis also provides a quantitative comparison between ITD and AID based approaches. For stochastic bilevel optimization, we propose a novel algorithm named stocBiO, which features a sample-efficient hypergradient estimator using efficient Jacobian- and Hessian-vector product computations. We provide the convergence rate guarantee for stocBiO, and show that stocBiO outperforms the best known computational complexities orderwisely with respect to the condition number $\\kappa$ and the target accuracy $\\epsilon$. We further validate our theoretical results and demonstrate the efficiency of bilevel optimization algorithms by the experiments on meta-learning and hyperparameter optimization."
                },
                {
                    "title": "On the Iteration Complexity of Hypergradient Computation",
                    "abstract": "We study a general class of bilevel problems, consisting in the minimization of an upper-level objective which depends on the solution to a parametric fixed-point equation. Important instances arising in machine learning include hyperparameter optimization, meta-learning, and certain graph and recurrent neural networks. Typically the gradient of the upper-level objective (hypergradient) is hard or even impossible to compute exactly, which has raised the interest in approximation methods. We investigate some popular approaches to compute the hypergradient, based on reverse mode iterative differentiation and approximate implicit differentiation. Under the hypothesis that the fixed point equation is defined by a contraction mapping, we present a unified analysis which allows for the first time to quantitatively compare these methods, providing explicit bounds for their iteration complexity. This analysis suggests a hierarchy in terms of computational efficiency among the above methods, with approximate implicit differentiation based on conjugate gradient performing best. We present an extensive experimental comparison among the methods which confirm the theoretical findings."
                },
                {
                    "title": "Implicit differentiation of Lasso-type models for hyperparameter optimization",
                    "abstract": "Setting regularization parameters for Lasso-type estimators is notoriously difficult, though crucial in practice. The most popular hyperparameter optimization approach is grid-search using held-out validation data. Grid-search however requires to choose a predefined grid for each parameter, which scales exponentially in the number of parameters. Another approach is to cast hyperparameter optimization as a bi-level optimization problem, one can solve by gradient descent. The key challenge for these methods is the estimation of the gradient with respect to the hyperparameters. Computing this gradient via forward or backward automatic differentiation is possible yet usually suffers from high memory consumption. Alternatively implicit differentiation typically involves solving a linear system which can be prohibitive and numerically unstable in high dimension. In addition, implicit differentiation usually assumes smooth loss functions, which is not the case for Lasso-type problems. This work introduces an efficient implicit differentiation algorithm, without matrix inversion, tailored for Lasso-type problems. Our approach scales to high-dimensional data by leveraging the sparsity of the solutions. Experiments demonstrate that the proposed method outperforms a large number of standard methods to optimize the error on held-out data, or the Stein Unbiased Risk Estimator (SURE)."
                },
                {
                    "title": "Super-efficiency of automatic differentiation for functions defined as a minimum",
                    "abstract": "In min-min optimization or max-min optimization, one has to compute the gradient of a function defined as a minimum. In most cases, the minimum has no closed-form, and an approximation is obtained via an iterative algorithm. There are two usual ways of estimating the gradient of the function: using either an analytic formula obtained by assuming exactness of the approximation, or automatic differentiation through the algorithm. In this paper, we study the asymptotic error made by these estimators as a function of the optimization error. We find that the error of the automatic estimator is close to the square of the error of the analytic estimator, reflecting a super-efficiency phenomenon. The convergence of the automatic estimator greatly depends on the convergence of the Jacobian of the algorithm. We analyze it for gradient descent and stochastic gradient descent and derive convergence rates for the estimators in these cases. Our analysis is backed by numerical experiments on toy problems and on Wasserstein barycenter computation. Finally, we discuss the computational complexity of these estimators and give practical guidelines to chose between them."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Optimizing Millions of Hyperparameters by Implicit Differentiation",
                    "abstract": "We propose an algorithm for inexpensive gradient-based hyperparameter optimization that combines the implicit function theorem (IFT) with efficient inverse Hessian approximations. We present results about the relationship between the IFT and differentiating through optimization, motivating our algorithm. We use the proposed approach to train modern network architectures with millions of weights and millions of hyper-parameters. For example, we learn a data-augmentation network - where every weight is a hyperparameter tuned for validation performance - outputting augmented training examples. Jointly tuning weights and hyperparameters with our approach is only a few times more costly in memory and compute than standard training."
                },
                {
                    "title": "Differentiable Convex Optimization Layers",
                    "abstract": "Recent work has shown how to embed differentiable optimization problems (that is, problems whose solutions can be backpropagated through) as layers within deep learning architectures. This method provides a useful inductive bias for certain problems, but existing software for differentiable optimization layers is rigid and difficult to apply to new settings. In this paper, we propose an approach to differentiating through disciplined convex programs, a subclass of convex optimization problems used by domain-specific languages (DSLs) for convex optimization. We introduce disciplined parametrized programming, a subset of disciplined convex programming, and we show that every disciplined parametrized program can be represented as the composition of an affine map from parameters to problem data, a solver, and an affine map from the solver\u2019s solution to a solution of the original problem (a new form we refer to as affine-solver-affine form). We then demonstrate how to efficiently differentiate through each of these components, allowing for end-to-end analytical differentiation through the entire convex program. We implement our methodology in version 1.1 of CVXPY, a popular Python-embedded DSL for convex optimization, and additionally implement differentiable layers for disciplined convex programs in PyTorch and TensorFlow 2.0. Our implementation significantly lowers the barrier to using convex optimization problems in differentiable programs. We present applications in linear machine learning models and in stochastic control, and we show that our layer is competitive (in execution time) compared to specialized differentiable solvers from past work."
                },
                {
                    "title": "Automatic Differentiation of Some First-Order Methods in Parametric Optimization",
                    "abstract": "We aim at computing the derivative of the solution to a parametric optimization problem with respect to the involved parameters. For a class broader than that of strongly convex functions, this can be achieved by automatic differentiation of iterative minimization algorithms. If the iterative algorithm converges pointwise, then we prove that the derivative sequence also converges pointwise to the derivative of the minimizer with respect to the parameters. Moreover, we provide convergence rates for both sequences. In particular, we prove that the accelerated convergence rate of the Heavy-ball method compared to Gradient Descent also accelerates the derivative computation. An experiment with L2-Regularized Logistic Regression validates the theoretical results."
                },
                {
                    "title": "Meta-Learning with Implicit Gradients",
                    "abstract": "A core capability of intelligent systems is the ability to quickly learn new tasks by drawing on prior experience. Gradient (or optimization) based meta-learning has recently emerged as an effective approach for few-shot learning. In this formulation, meta-parameters are learned in the outer loop, while task-specific models are learned in the inner-loop, by using only a small amount of data from the current task. A key challenge in scaling these approaches is the need to differentiate through the inner loop learning process, which can impose considerable computational and memory burdens. By drawing upon implicit differentiation, we develop the implicit MAML algorithm, which depends only on the solution to the inner level optimization and not the path taken by the inner loop optimizer. This effectively decouples the meta-gradient computation from the choice of inner loop optimizer. As a result, our approach is agnostic to the choice of inner loop optimizer and can gracefully handle many gradient steps without vanishing gradients or memory constraints. Theoretically, we prove that implicit MAML can compute accurate meta-gradients with a memory footprint that is, up to small constant factors, no more than that which is required to compute a single inner loop gradient and at no overall increase in the total computational cost. Experimentally, we show that these benefits of implicit MAML translate into empirical gains on few-shot image recognition benchmarks."
                },
                {
                    "title": "Deep Equilibrium Models",
                    "abstract": "We present a new approach to modeling sequential data: the deep equilibrium model (DEQ). Motivated by an observation that the hidden layers of many existing deep sequence models converge towards some fixed point, we propose the DEQ approach that directly finds these equilibrium points via root-finding. Such a method is equivalent to running an infinite depth (weight-tied) feedforward network, but has the notable advantage that we can analytically backpropagate through the equilibrium point using implicit differentiation. Using this approach, training and prediction in these networks require only constant memory, regardless of the effective \u201cdepth\u201d of the network. We demonstrate how DEQs can be applied to two state-of-the-art deep sequence models: self-attention transformers and trellis networks. On large-scale language modeling tasks, such as the WikiText-103 benchmark, we show that DEQs 1) often improve performance over these state-of-the-art models (for similar parameter counts); 2) have similar computational requirements to existing models; and 3) vastly reduce memory consumption (often the bottleneck for training large sequence models), demonstrating an up-to 88% memory reduction in our experiments. The code is available at https://github.com/locuslab/deq."
                },
                {
                    "title": "Implicit Deep Learning",
                    "abstract": "Implicit deep learning prediction rules generalize the recursive rules of feedforward neural networks. Such rules are based on the solution of a fixed-point equation involving a single vector of hidden features, which is thus only implicitly defined. The implicit framework greatly simplifies the notation of deep learning, and opens up many new possibilities, in terms of novel architectures and algorithms, robustness analysis and design, interpretability, sparsity, and network architecture optimization."
                },
                {
                    "title": "SGD: General Analysis and Improved Rates",
                    "abstract": "We propose a general yet simple theorem describing the convergence of SGD under the arbitrary sampling paradigm. Our theorem describes the convergence of an infinite array of variants of SGD, each of which is associated with a specific probability law governing the data selection rule used to form mini-batches. This is the first time such an analysis is performed, and most of our variants of SGD were never explicitly considered in the literature before. Our analysis relies on the recently introduced notion of expected smoothness and does not rely on a uniform bound on the variance of the stochastic gradients. By specializing our theorem to different mini-batching strategies, such as sampling with replacement and independent sampling, we derive exact expressions for the stepsize as a function of the mini-batch size. With this we can also determine the mini-batch size that optimizes the total complexity, and show explicitly that as the variance of the stochastic gradient evaluated at the minimum grows, so does the optimal mini-batch size. For zero variance, the optimal mini-batch size is one. Moreover, we prove insightful stepsize-switching rules which describe when one should switch from a constant to a decreasing stepsize regime."
                },
                {
                    "title": "Truncated Back-propagation for Bilevel Optimization",
                    "abstract": "Bilevel optimization has been recently revisited for designing and analyzing algorithms in hyperparameter tuning and meta learning tasks. However, due to its nested structure, evaluating exact gradients for high-dimensional problems is computationally challenging. One heuristic to circumvent this difficulty is to use the approximate gradient given by performing truncated back-propagation through the iterative optimization procedure that solves the lower-level problem. Although promising empirical performance has been reported, its theoretical properties are still unclear. In this paper, we analyze the properties of this family of approximate gradients and establish sufficient conditions for convergence. We validate this on several hyperparameter tuning and meta learning tasks. We find that optimization with the approximate gradient computed using few-step back-propagation often performs comparably to optimization with the exact gradient, while requiring far less memory and half the computation time."
                },
                {
                    "title": "Bilevel Programming for Hyperparameter Optimization and Meta-Learning",
                    "abstract": "We introduce a framework based on bilevel programming that unifies gradient-based hyperparameter optimization and meta-learning. We show that an approximate version of the bilevel problem can be solved by taking into explicit account the optimization dynamics for the inner objective. Depending on the specific setting, the outer variables take either the meaning of hyperparameters in a supervised learning problem or parameters of a meta-learner. We provide sufficient conditions under which solutions of the approximate problem converge to those of the exact problem. We instantiate our approach for meta-learning in the case of deep learning where representation layers are treated as hyperparameters shared across a set of training episodes. In experiments, we confirm our theoretical findings, present encouraging results for few-shot learning and contrast the bilevel approach against classical approaches for learning-to-learn."
                },
                {
                    "title": "Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks",
                    "abstract": "We propose an algorithm for meta-learning that is model-agnostic, in the sense that it is compatible with any model trained with gradient descent and applicable to a variety of different learning problems, including classification, regression, and reinforcement learning. The goal of meta-learning is to train a model on a variety of learning tasks, such that it can solve new learning tasks using only a small number of training samples. In our approach, the parameters of the model are explicitly trained such that a small number of gradient steps with a small amount of training data from a new task will produce good generalization performance on that task. In effect, our method trains the model to be easy to fine-tune. We demonstrate that this approach leads to state-of-the-art performance on two few-shot image classification benchmarks, produces good results on few-shot regression, and accelerates fine-tuning for policy gradient reinforcement learning with neural network policies."
                },
                {
                    "title": "Forward and Reverse Gradient-Based Hyperparameter Optimization",
                    "abstract": "We study two procedures (reverse-mode and forward-mode) for computing the gradient of the validation error with respect to the hyperparameters of any iterative learning algorithm such as stochastic gradient descent. These procedures mirror two methods of computing gradients for recurrent neural networks and have different trade-offs in terms of running time and space requirements. Our formulation of the reverse-mode procedure is linked to previous work by Maclaurin et al. [2015] but does not require reversible dynamics. The forward-mode procedure is suitable for real-time hyperparameter updates, which may significantly speed up hyperparameter optimization on large datasets. We present experiments on data cleaning and on learning task interactions. We also present one large-scale experiment where the use of previous gradient-based methods would be prohibitive."
                },
                {
                    "title": "OptNet: Differentiable Optimization as a Layer in Neural Networks",
                    "abstract": "This paper presents OptNet, a network architecture that integrates optimization problems (here, specifically in the form of quadratic programs) as individual layers in larger end-to-end trainable deep networks. These layers encode constraints and complex dependencies between the hidden states that traditional convolutional and fully-connected layers often cannot capture. In this paper, we explore the foundations for such an architecture: we show how techniques from sensitivity analysis, bilevel optimization, and implicit differentiation can be used to exactly differentiate through these layers and with respect to layer parameters; we develop a highly efficient solver for these layers that exploits fast GPU-based batch solves within a primal-dual interior point method, and which provides backpropagation gradients with virtually no additional cost on top of the solve; and we highlight the application of these approaches in several problems. In one notable example, we show that the method is capable of learning to play mini-Sudoku (4x4) given just input and output games, with no a priori information about the rules of the game; this highlights the ability of our architecture to learn hard constraints better than other neural architectures."
                },
                {
                    "title": "CLEAR: Covariant LEAst-Square Refitting with Applications to Image Restoration",
                    "abstract": "In this paper, we propose a new framework to remove parts of the systematic errors affecting popular restoration algorithms, with a special focus for image processing tasks. Generalizing ideas that emerged for $\\ell_1$ regularization, we develop an approach re-fitting the results of standard methods towards the input data. Total variation regularizations and non-local means are special cases of interest. We identify important covariant information that should be preserved by the re-fitting method, and emphasize the importance of preserving the Jacobian (w.r.t. the observed signal) of the original estimator. Then, we provide an approach that has a \"twicing\" flavor and allows re-fitting the restored signal by adding back a local affine transformation of the residual term. We illustrate the benefits of our method on numerical simulations for image restoration tasks."
                },
                {
                    "title": "Optimization Methods for Large-Scale Machine Learning",
                    "abstract": "This paper provides a review and commentary on the past, present, and future of numerical optimization algorithms in the context of machine learning applications. Through case studies on text classification and the training of deep neural networks, we discuss how optimization problems arise in machine learning and what makes them challenging. A major theme of our study is that large-scale machine learning represents a distinctive setting in which the stochastic gradient (SG) method has traditionally played a central role while conventional gradient-based nonlinear optimization techniques typically falter. Based on this viewpoint, we present a comprehensive theory of a straightforward, yet versatile SG algorithm, discuss its practical behavior, and highlight opportunities for designing algorithms with improved performance. This leads to a discussion about the next generation of optimization methods for large-scale machine learning, including an investigation of two main streams of research on techniques that diminish noise in the stochastic directions and methods that make use of second-order derivative approximations."
                },
                {
                    "title": "Analysis of Gradient Descent Methods With Nondiminishing Bounded Errors",
                    "abstract": "The main aim of this paper is to provide an analysis of gradient descent (<inline-formula><tex-math notation=\"LaTeX\"> $\\text{GD}$</tex-math></inline-formula>) algorithms with gradient errors that do not necessarily vanish, asymptotically. In particular, sufficient conditions are presented for both stability (almost sure boundedness of the iterates) and convergence of <inline-formula><tex-math notation=\"LaTeX\">$\\text{GD}$</tex-math></inline-formula> with bounded (possibly) nondiminishing gradient errors. In addition to ensuring stability, such an algorithm is shown to converge to a small neighborhood of the minimum set, which depends on the gradient errors. It is worth noting that the main result of this paper can be used to show that <inline-formula><tex-math notation=\"LaTeX\">$\\text{GD}$</tex-math> </inline-formula> with asymptotically vanishing errors indeed converges to the minimum set. The results presented herein are not only more general when compared to previous results, but our analysis of <italic><inline-formula> <tex-math notation=\"LaTeX\">$\\text{GD}$</tex-math></inline-formula> with errors</italic> is new to the literature to the best of our knowledge. Our work extends the contributions of Mangasarian and Solodov, Bertsekas and Tsitsiklis, and Tadi\u0107 and Doucet. Using our framework, a simple yet effective implementation of <inline-formula> <tex-math notation=\"LaTeX\">$\\text{GD}$</tex-math></inline-formula> using simultaneous perturbation stochastic approximations, with constant sensitivity parameters, is presented. Another important improvement over many previous results is that there are no \u201cadditional\u201d restrictions imposed on the step sizes. In machine learning applications where step sizes are related to learning rates, our assumptions, unlike those of other papers, do not affect these learning rates. Finally, we present experimental results to validate our theory."
                },
                {
                    "title": "Hyperparameter optimization with approximate gradient",
                    "abstract": "Most models in machine learning contain at least one hyperparameter to control for model complexity. Choosing an appropriate set of hyperparameters is both crucial in terms of model accuracy and computationally challenging. In this work we propose an algorithm for the optimization of continuous hyperparameters using inexact gradient information. An advantage of this method is that hyperparameters can be updated before model parameters have fully converged. We also give sufficient conditions for the global convergence of this method, based on regularity conditions of the involved functions and summability of errors. Finally, we validate the empirical performance of this method on the estimation of regularization constants of L2-regularized logistic regression and kernel Ridge regression. Empirical benchmarks indicate that our approach is highly competitive with respect to state of the art methods."
                },
                {
                    "title": "Automatic differentiation in machine learning: a survey",
                    "abstract": "Derivatives, mostly in the form of gradients and Hessians, are ubiquitous in machine learning. Automatic differentiation (AD), also called algorithmic differentiation or simply \u201cauto-diff\u201d, is a family of techniques similar to but more general than backpropagation for efficiently and accurately evaluating derivatives of numeric functions expressed as computer programs. AD is a small but established field with applications in areas including computational fluid dynamics, atmospheric sciences, and engineering design optimization. Until \nvery recently, the fields of machine learning and AD have largely been unaware of each other and, in some cases, have independently discovered each other\u2019s results. Despite its \nrelevance, general-purpose AD has been missing from the machine learning toolbox, a situation slowly changing with its ongoing adoption under the names \u201cdynamic computational \ngraphs\u201d and \u201cdifferentiable programming\u201d. We survey the intersection of AD and machine learning, cover applications where AD has direct relevance, and address the main imple- \nmentation techniques. By precisely defining the main differentiation techniques and their interrelationships, we aim to bring clarity to the usage of the terms \u201cautodiff\u201d, \u201cautomatic differentiation\u201d, and \u201csymbolic differentiation\u201d as these are encountered more and more in machine learning settings."
                },
                {
                    "title": "Gradient-based Hyperparameter Optimization through Reversible Learning",
                    "abstract": "Tuning hyperparameters of learning algorithms is hard because gradients are usually unavailable. We compute exact gradients of cross-validation performance with respect to all hyperparameters by chaining derivatives backwards through the entire training procedure. These gradients allow us to optimize thousands of hyperparameters, including step-size and momentum schedules, weight initialization distributions, richly parameterized regularization schemes, and neural network architectures. We compute hyperparameter gradients by exactly reversing the dynamics of stochastic gradient descent with momentum."
                },
                {
                    "title": "Stein Unbiased GrAdient estimator of the Risk (SUGAR) for Multiple Parameter Selection",
                    "abstract": "Algorithms to solve variational regularization of ill-posed inverse problems usually involve operators that depend on a collection of continuous parameters. When these operators enjoy some (local) regularity, these parameters can be selected using the so-called Stein Unbiased Risk Estimate (SURE). While this selection is usually performed by exhaustive search, we address in this work the problem of using the SURE to efficiently optimize for a collection of continuous parameters of the model. When considering non-smooth regularizers, such as the popular l1-norm corresponding to soft-thresholding mapping, the SURE is a discontinuous function of the parameters preventing the use of gradient descent optimization techniques. Instead, we focus on an approximation of the SURE based on finite differences as proposed in (Ramani et al., 2008). Under mild assumptions on the estimation mapping, we show that this approximation is a weakly differentiable function of the parameters and its weak gradient, coined the Stein Unbiased GrAdient estimator of the Risk (SUGAR), provides an asymptotically (with respect to the data dimension) unbiased estimate of the gradient of the risk. Moreover, in the particular case of soft-thresholding, the SUGAR is proved to be also a consistent estimator. The SUGAR can then be used as a basis to perform a quasi-Newton optimization. The computation of the SUGAR relies on the closed-form (weak) differentiation of the non-smooth function. We provide its expression for a large class of iterative proximal splitting methods and apply our strategy to regularizations involving non-smooth convex structured penalties. Illustrations on various image restoration and matrix completion problems are given."
                },
                {
                    "title": "Asymptotic bias of stochastic gradient search",
                    "abstract": "The asymptotic behavior of the stochastic gradient algorithm with a biased gradient estimator is analyzed. Relying on arguments based on differential geometry (Yomdin theorem and Lojasiewicz inequality), relatively tight bounds on the asymptotic bias of the iterates generated by such an algorithm are derived. The obtained results hold under mild and verifiable conditions and cover a broad class of complex stochastic gradient algorithms. Using these results, the asymptotic properties of the actor-critic reinforcement learning are studied."
                },
                {
                    "title": "Real Analysis: Modern Techniques and Their Applications",
                    "abstract": "This book covers the subject matter that is central to mathematical analysis: measure and integration theory, some point set topology, and rudiments of functional analysis. Also, a number of other topics are developed to illustrate the uses of this core material in important areas of mathematics and to introduce readers to more advanced techniques. Some of the material presented has never appeared outside of advanced monographs and research papers, or been readily available in comparative texts. About 460 exercises, at varying levels of difficulty, give readers practice in working with the ideas presented here."
                },
                {
                    "title": "A simple automatic derivative evaluation program",
                    "abstract": "A procedure for automatic evaluation of total/partial derivatives of arbitrary algebraic functions is presented. The technique permits computation of numerical values of derivatives without developing analytical expressions for the derivatives. The key to the method is the decomposition of the given function, by introduction of intermediate variables, into a series of elementary functional steps. A library of elementary function subroutines is provided for the automatic evaluation and differentiation of these new variables. The final step in this process produces the desired function's derivative. The main feature of this approach is its simplicity. It can be used as a quick-reaction tool where the derivation of analytical derivatives is laborious and also as a debugging tool for programs which contain derivatives."
                },
                {
                    "title": "A Stochastic Approximation Method",
                    "abstract": "Let M(x) denote the expected value at level x of the response to a certain experiment. M(x) is assumed to be a monotone function of x but is unknown tot he experiment, and it is desire to find the solution x=0 of the equation M(x) = a, where x is a given constant. we give a method for making successive experiments at levels x1, x2,... in such a way that x, will tend to 0 in probability."
                },
                {
                    "title": "Closing the Gap: Tighter Analysis of Alternating Stochastic Gradient Methods for Bilevel Problems",
                    "abstract": "Stochastic nested optimization, including stochastic bilevel, min-max, and compositional optimization, is gaining popularity in many machine learning applications. While the three problems share a nested structure, existing works often treat them separately, thus developing problem-specific algorithms and analyses. Among various exciting developments, simple SGD-type updates (potentially on multiple variables) are still prevalent in solving this class of nested problems, but they are believed to have a slower convergence rate than non-nested problems. This paper unifies several SGD-type updates for stochastic nested problems into a single SGD approach that we term ALternating Stochastic gradient dEscenT (ALSET) method. By leveraging the hidden smoothness of the problem, this paper presents a tighter analysis of ALSET for stochastic nested problems. Under the new analysis, to achieve an -stationary point of the nested problem, it requires O( \u22122) samples in total. Under certain regularity conditions, applying our results to stochastic compositional, min-max, and reinforcement learning problems either improves or matches the best-known sample complexity in the respective cases. Our results explain why simple SGD-type algorithms in stochastic nested problems all work very well in practice without the need for further modifications."
                },
                {
                    "title": "Evaluating derivatives - principles and techniques of algorithmic differentiation, Second Edition",
                    "abstract": "Algorithmic, or automatic, differentiation (AD) is a growing area of theoretical research and software development concerned with the accurate and efficient evaluation of derivatives for function evaluations given as computer programs. The resulting derivative values are useful for all scientific computations that are based on linear, quadratic, or higher order approximations to nonlinear scalar or vector functions. AD has been applied in particular to optimization, parameter identification, nonlinear equation solving, the numerical integration of differential equations, and combinations of these. Apart from quantifying sensitivities numerically, AD also yields structural dependence information, such as the sparsity pattern and generic rank of Jacobian matrices. The field opens up an exciting opportunity to develop new algorithms that reflect the true cost of accurate derivatives and to use them for improvements in speed and reliability. This second edition has been updated and expanded to cover recent developments in applications and theory, including an elegant NP completeness argument by Uwe Naumann and a brief introduction to scarcity, a generalization of sparsity. There is also added material on checkpointing and iterative differentiation. To improve readability the more detailed analysis of memory and complexity bounds has been relegated to separate, optional chapters.The book consists of three parts: a stand-alone introduction to the fundamentals of AD and its software; a thorough treatment of methods for sparse problems; and final chapters on program-reversal schedules, higher derivatives, nonsmooth problems and iterative processes. Each of the 15 chapters concludes with examples and exercises. Audience: This volume will be valuable to designers of algorithms and software for nonlinear computational problems. Current numerical software users should gain the insight necessary to choose and deploy existing AD software tools to the best advantage. Contents: Rules; Preface; Prologue; Mathematical Symbols; Chapter 1: Introduction; Chapter 2: A Framework for Evaluating Functions; Chapter 3: Fundamentals of Forward and Reverse; Chapter 4: Memory Issues and Complexity Bounds; Chapter 5: Repeating and Extending Reverse; Chapter 6: Implementation and Software; Chapter 7: Sparse Forward and Reverse; Chapter 8: Exploiting Sparsity by Compression; Chapter 9: Going beyond Forward and Reverse; Chapter 10: Jacobian and Hessian Accumulation; Chapter 11: Observations on Efficiency; Chapter 12: Reversal Schedules and Checkpointing; Chapter 13: Taylor and Tensor Coefficients; Chapter 14: Differentiation without Differentiability; Chapter 15: Implicit and Iterative Differentiation; Epilogue; List of Figures; List of Tables; Assumptions and Definitions; Propositions, Corollaries, and Lemmas; Bibliography; Index"
                },
                {
                    "title": "Reverse accumulation and attractive fixed points",
                    "abstract": "We apply reverse accumulation to obtain automatic gradients and error estimates of functions which include in their computation a convergent iteration of the form y= \u03a6(y,u), where y and u are vectors. We suggest an implementation approach which allows this to be done by a fairly routine extension of existing reverse accumulation code. We show how to re-use the computational graph for the fixed point constructor \u03a6 so as to set explicit stopping criteria for the iterations, based on the gradient accuracy required. Our construction allows the gradient vector to be obtained to the same order of accuracy as the objective function values (which is in general the best we can hope to achieve), and the same order of computational cost (which does not explicitly depend upon the number of independent variables.) The technique can be applied to functions which contain several iterative constructions, either serially or nested"
                },
                {
                    "title": "Derivative convergence for iterative equation solvers",
                    "abstract": "When nonlinear equation solvers are applied to parameter-dependent problems, their iterates can be interpreted as functions of these variable parameters. The derivatives (if they exist) of these iterated functions can be recursively evaluated by the forward mode of automatic differentiation. Then one may ask whether and how fast these derivative values converge to the derivative of the implicit solution function, which may be needed for parameter identification, sensitivity studies, or design optimization. It is shown here that derivative convergence is achieved with an R-linear or possibly R-superlinear rate for a large class of memoryless contractions or secant updating methods. For a wider class of multistep contractions, we obtain R-linear convergence of a simplified derivative recurrence, which is more economical and can be easily generalized to higher-order derivatives. We also formulate a constructive criterion for derivative convergence based on the implicit function theorem. All theoretical resul..."
                },
                {
                    "title": "Automatic differentiation and iterative processes",
                    "abstract": "We identify a class of iterative processes that can be used in the definition of a function while preserving the good behavior of automatic differentiation codes on this function. By iterative proc..."
                },
                {
                    "title": "Stochastic quasigradient methods and their application to system optimization",
                    "abstract": "This paper systematically surveys the development of stochastic quasigradient (SQG) methods.These metods make it possible to solve optimization problems without calculating the precise valuesw of objectives and constraints (le alone of their derivatives). For deterministic nonlinear optimization problems, these methods can be regarded as methods of randomsearch. For stochastic programming problems. SQG methods generalize the well-known stochastic approximation methods for uncnstrained optimization of the expectation of a random function to problems involving general constraints."
                }
            ],
            "categories": [
                "math.OC",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nWhat is the dynamics of the derivatives of the iterates of stochastic gradient descent in the context of minimization of parametric strongly convex functions?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the understanding of stochastic optimization methods, particularly in machine learning where Stochastic Gradient Descent (SGD) is widely used. By elucidating the behavior of derivatives in SGD, this research could lead to improved optimization techniques, enhancing model training efficiency and effectiveness. Furthermore, it could inspire future research into differentiable programming and stochastic approximation methods, potentially leading to novel applications in hyperparameter optimization and meta-learning.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the joint effect of noise on both the iterate sequence and its derivatives, complicating the analysis of convergence. Naive approaches may fail because they do not account for the stochastic nature of the gradients, which can introduce significant variability and hinder the convergence of derivatives to the solution mapping. Additionally, the need for rigorous mathematical guarantees in the presence of stochasticity adds a layer of complexity that must be addressed.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on deterministic algorithms, leaving a gap in the understanding of stochastic methods like SGD. Existing solutions often involve intricate algorithmic schemes that are not as conceptually straightforward as iterative differentiation. The lack of a clear framework for analyzing the stochastic behavior of derivatives has prevented a comprehensive solution until now. This work aims to fill that gap by providing a more direct approach to understanding the dynamics of SGD derivatives.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves analyzing the convergence of the derivatives of the SGD recursion in the context of strongly convex parametric optimization. The study will utilize mean squared error metrics to assess convergence, focusing on non-increasing step-sizes and their implications for derivative behavior. Expected outcomes include proving that the derivatives of SGD converge toward the solution derivatives, with specific convergence rates established for different step-size decay schedules, thereby providing a clearer understanding of the iterative differentiation process in stochastic settings."
            }
        },
        "author_data": {
            "20c5a762-eb30-4dc8-8275-c1ff7a08a9de": {
                "pk": "20c5a762-eb30-4dc8-8275-c1ff7a08a9de",
                "name": "Franck Iutzeler",
                "collaborators": [
                    "J\u00e9r\u00f4me Malick",
                    "Walid Hachem",
                    "Wa\u00efss Azizian",
                    "Gilles Bareilles",
                    "Pascal Bianchi",
                    "Philippe Ciblat",
                    "Dmitry Grishchenko",
                    "Massih-Reza Amini",
                    "Panayotis Mertikopoulos",
                    "Julien M. Hendrickx"
                ],
                "domain": [
                    "Optimization",
                    "Machine Learning",
                    "Distributed Systems",
                    "Nonsmooth Analysis"
                ],
                "publications": [
                    {
                        "title": "Nonsmoothness in Machine Learning: specific structure, proximal identification, and applications",
                        "abstract": "Nonsmoothness is often a curse for optimization; but it is sometimes a blessing, in particular for applications in machine learning. In this paper, we present the specific structure of nonsmooth optimization problems appearing in machine learning and illustrate how to leverage this structure in practice, for compression, acceleration, or dimension reduction. We pay a special attention to the presentation to make it concise and easily accessible, with both simple examples and general results."
                    },
                    {
                        "title": "A Generic online acceleration scheme for Optimization algorithms via Relaxation and Inertia",
                        "abstract": "We propose generic acceleration schemes for a wide class of optimization and iterative schemes based on relaxation and inertia. In particular, we introduce methods that automatically tunes the acceleration coefficients online, and establish their convergence. This is made possible by considering the class of fixed-points iterations over averaged operators which encompass gradient methods, ADMM, primal dual algorithms, an so on."
                    },
                    {
                        "title": "On the Proximal Gradient Algorithm with Alternated Inertia",
                        "abstract": "In this paper, we investigate the attractive properties of the proximal gradient algorithm with inertia. Notably, we show that using alternated inertia yields monotonically decreasing functional values, which contrasts with usual accelerated proximal gradient methods. We also provide convergence rates for the algorithm with alternated inertia based on local geometric properties of the objective function. The results are put into perspective by discussions on several extensions and illustrations on common regularized problems."
                    },
                    {
                        "title": "Proximal Gradient methods with Adaptive Subspace Sampling",
                        "abstract": "Many applications in machine learning or signal processing involve nonsmooth optimization problems. This nonsmoothness brings a low-dimensional structure to the optimal solutions. In this paper, we propose a randomized proximal gradient method harnessing this underlying structure. We introduce two key components: i) a random subspace proximal gradient algorithm; ii) an identification-based sampling of the subspaces. Their interplay brings a significant performance improvement on typical learning problems in terms of dimensions explored."
                    },
                    {
                        "title": "Regularization for Wasserstein Distributionally Robust Optimization",
                        "abstract": "Optimal transport has recently proved to be a useful tool in various machine learning applications needing comparisons of probability measures. Among these, applications of distributionally robust optimization naturally involve Wasserstein distances in their models of uncertainty, capturing data shifts or worst-case scenarios. Inspired by the success of the regularization of Wasserstein distances in optimal transport, we study in this paper the regularization of Wasserstein distributionally robust optimization. First, we derive a general strong duality result of regularized Wasserstein distributionally robust problems. Second, we refine this duality result in the case of entropic regularization and provide an approximation result when the regularization parameters vanish."
                    },
                    {
                        "title": "On the Interplay between Acceleration and Identification for the Proximal Gradient algorithm",
                        "abstract": "In this paper, we study the interplay between acceleration and structure identification for the proximal gradient algorithm. We report and analyze several cases where this interplay has negative effects on the algorithm behavior (iterates oscillation, loss of structure, etc.). We present a generic method that tames acceleration when structure identification may be at stake; it benefits from a convergence rate that matches the one of the accelerated proximal gradient under some qualifying condition. We show empirically that the proposed method is much more stable in terms of subspace identification compared to the accelerated proximal gradient method while keeping a similar functional decrease."
                    },
                    {
                        "title": "Asynchronous Distributed Optimization using a Randomized Alternating Direction Method of Multipliers",
                        "abstract": "Consider a set of networked agents endowed with private cost functions and seeking to find a consensus on the minimizer of the aggregate cost. A new class of random asynchronous distributed optimization methods is introduced. The methods generalize the standard Alternating Direction Method of Multipliers (ADMM) to an asynchronous setting where isolated components of the network are activated in an uncoordinated fashion. The algorithms rely on the introduction of randomized Gauss-Seidel iterations of a Douglas-Rachford operator for finding zeros of a sum of two monotone operators. Convergence to the sought minimizers is provided under mild connectivity conditions. Numerical results sustain our claims."
                    },
                    {
                        "title": "A Coordinate Descent Primal-Dual Algorithm and Application to Distributed Asynchronous Optimization",
                        "abstract": "Based on the idea of randomized coordinate descent of $\\alpha$-averaged operators, a randomized primal-dual optimization algorithm is introduced, where a random subset of coordinates is updated at each iteration. The algorithm builds upon a variant of a recent (deterministic) algorithm proposed by V\\~u and Condat that includes the well known ADMM as a particular case. The obtained algorithm is used to solve asynchronously a distributed optimization problem. A network of agents, each having a separate cost function containing a differentiable term, seek to find a consensus on the minimum of the aggregate objective. The method yields an algorithm where at each iteration, a random subset of agents wake up, update their local estimates, exchange some data with their neighbors, and go idle. Numerical results demonstrate the attractive performance of the method. The general approach can be naturally adapted to other situations where coordinate descent convex optimization algorithms are used with a random choice of the coordinates."
                    },
                    {
                        "title": "A Distributed Flexible Delay-tolerant Proximal Gradient Algorithm",
                        "abstract": "We develop and analyze an asynchronous algorithm for distributed convex optimization when the objective writes a sum of smooth functions, local to each worker, and a non-smooth function. Unlike many existing methods, our distributed algorithm is adjustable to various levels of communication cost, delays, machines computational power, and functions smoothness. A unique feature is that the stepsizes do not depend on communication delays nor number of machines, which is highly desirable for scalability. We prove that the algorithm converges linearly in the strongly convex case, and provide guarantees of convergence for the non-strongly convex case. The obtained rates are the same as the vanilla proximal gradient algorithm over some introduced epoch sequence that subsumes the delays of the system. We provide numerical results on large-scale machine learning problems to demonstrate the merits of the proposed method."
                    },
                    {
                        "title": "Rank-one partitioning: formalization, illustrative examples, and a new cluster enhancing strategy",
                        "abstract": "In this paper, we introduce and formalize a rank-one partitioning learning paradigm that unifies partitioning methods that proceed by summarizing a data set using a single vector that is further used to derive the final clustering partition. Using this unification as a starting point, we propose a novel algorithmic solution for the partitioning problem based on rank-one matrix factorization and denoising of piecewise constant signals. Finally, we propose an empirical demonstration of our findings and demonstrate the robustness of the proposed denoising step. We believe that our work provides a new point of view for several unsupervised learning techniques that helps to gain a deeper understanding about the general mechanisms of data partitioning."
                    },
                    {
                        "title": "An Asynchronous Distributed Framework for Large-scale Learning Based on Parameter Exchanges",
                        "abstract": "In many distributed learning problems, the heterogeneous loading of computing machines may harm the overall performance of synchronous strategies. In this paper, we propose an effective asynchronous distributed framework for the minimization of a sum of smooth functions, where each machine performs iterations in parallel on its local function and updates a shared parameter asynchronously. In this way, all machines can continuously work even though they do not have the latest version of the shared parameter. We prove the convergence of the consistency of this general distributed asynchronous method for gradient iterations then show its efficiency on the matrix factorization problem for recommender systems and on binary classification."
                    },
                    {
                        "title": "Explicit Convergence Rate of a Distributed Alternating Direction Method of Multipliers",
                        "abstract": "Consider a set of N agents seeking to solve distributively the minimization problem $\\inf_{x} \\sum_{n = 1}^N f_n(x)$ where the convex functions $f_n$ are local to the agents. The popular Alternating Direction Method of Multipliers has the potential to handle distributed optimization problems of this kind. We provide a general reformulation of the problem and obtain a class of distributed algorithms which encompass various network architectures. The rate of convergence of our method is considered. It is assumed that the infimum of the problem is reached at a point $x_\\star$, the functions $f_n$ are twice differentiable at this point and $\\sum \\nabla^2 f_n(x_\\star) > 0$ in the positive definite ordering of symmetric matrices. With these assumptions, it is shown that the convergence to the consensus $x_\\star$ is linear and the exact rate is provided. Application examples where this rate can be optimized with respect to the ADMM free parameter $\\rho$ are also given."
                    },
                    {
                        "title": "Distributed Learning with Sparse Communications by Identification",
                        "abstract": "In distributed optimization for large-scale learning, a major performance limitation comes from the communications between the different entities. When computations are performed by workers on local data while a coordinator machine coordinates their updates to minimize a global loss, we present an asynchronous optimization algorithm that efficiently reduces the communications between the coordinator and workers. This reduction comes from a random sparsification of the local updates. We show that this algorithm converges linearly in the strongly convex case and also identifies optimal strongly sparse solutions. We further exploit this identification to propose an automatic dimension reduction, aptly sparsifying all exchanges between coordinator and workers."
                    },
                    {
                        "title": "Optimization in Open Networks via Dual Averaging",
                        "abstract": "In networks of autonomous agents (e.g., fleets of vehicles, scattered sensors), the problem of minimizing the sum of the agents' local functions has received a lot of interest. We tackle here this distributed optimization problem in the case of open networks when agents can join and leave the network at any time. Leveraging recent online optimization techniques, we propose and analyze the convergence of a decentralized asynchronous optimization method for open networks."
                    },
                    {
                        "title": "Analysis of Sum-Weight-like algorithms for averaging in Wireless Sensor Networks",
                        "abstract": "Distributed estimation of the average value over a Wireless Sensor Network has recently received a lot of attention. Most papers consider single variable sensors and communications with feedback (e.g. peer-to-peer communications). However, in order to use efficiently the broadcast nature of the wireless channel, communications without feedback are advocated. To ensure the convergence in this feedback-free case, the recently-introduced Sum-Weight-like algorithms which rely on two variables at each sensor are a promising solution. In this paper, the convergence towards the consensus over the average of the initial values is analyzed in depth. Furthermore, it is shown that the squared error decreases exponentially with the time. In addition, a powerful algorithm relying on the Sum-Weight structure and taking into account the broadcast nature of the channel is proposed."
                    },
                    {
                        "title": "The Last-Iterate Convergence Rate of Optimistic Mirror Descent in Stochastic Variational Inequalities",
                        "abstract": "In this paper, we analyze the local convergence rate of optimistic mirror descent methods in stochastic variational inequalities, a class of optimization problems with important applications to learning theory and machine learning. Our analysis reveals an intricate relation between the algorithm's rate of convergence and the local geometry induced by the method's underlying Bregman function. We quantify this relation by means of the Legendre exponent, a notion that we introduce to measure the growth rate of the Bregman divergence relative to the ambient norm near a solution. We show that this exponent determines both the optimal step-size policy of the algorithm and the optimal rates attained, explaining in this way the differences observed for some popular Bregman functions (Euclidean projection, negative entropy, fractional power, etc.)."
                    },
                    {
                        "title": "Newton acceleration on manifolds identified by proximal-gradient methods",
                        "abstract": "Proximal methods are known to identify the underlying substructure of nonsmooth optimization problems. Even more, in many interesting situations, the output of a proximity operator comes with its structure at no additional cost, and convergence is improved once it matches the structure of a minimizer. However, it is impossible in general to know whether the current structure is final or not; such highly valuable information has to be exploited adaptively. To do so, we place ourselves in the case where a proximal gradient method can identify manifolds of differentiability of the nonsmooth objective. Leveraging this manifold identification, we show that Riemannian Newton-like methods can be intertwined with the proximal gradient steps to drastically boost the convergence. We prove the superlinear convergence of the algorithm when solving some nondegenerated nonsmooth nonconvex optimization problems. We provide numerical illustrations on optimization problems regularized by $\\ell_1$-norm or trace-norm."
                    },
                    {
                        "title": "Harnessing structure in composite nonsmooth minimization",
                        "abstract": "We consider the problem of minimizing the composition of a nonsmooth function with a smooth mapping in the case where the proximity operator of the nonsmooth function can be explicitly computed. We first show that this proximity operator can provide the exact smooth substructure of minimizers, not only of the nonsmooth function, but also of the full composite function. We then exploit this proximal identification by proposing an algorithm which combines proximal steps with sequential quadratic programming steps. We show that our method locally identifies the optimal smooth substructure and then converges quadratically. We illustrate its behavior on two problems: the minimization of a maximum of quadratic functions and the minimization of the maximal eigenvalue of a parametrized matrix."
                    },
                    {
                        "title": "Exact Generalization Guarantees for (Regularized) Wasserstein Distributionally Robust Models",
                        "abstract": "Wasserstein distributionally robust estimators have emerged as powerful models for prediction and decision-making under uncertainty. These estimators provide attractive generalization guarantees: the robust objective obtained from the training distribution is an exact upper bound on the true risk with high probability. However, existing guarantees either suffer from the curse of dimensionality, are restricted to specific settings, or lead to spurious error terms. In this paper, we show that these generalization guarantees actually hold on general classes of models, do not suffer from the curse of dimensionality, and can even cover distribution shifts at testing. We also prove that these results carry over to the newly-introduced regularized versions of Wasserstein distributionally robust problems."
                    }
                ]
            },
            "84e88def-fc21-482c-83aa-9708ef7c5f08": {
                "pk": "84e88def-fc21-482c-83aa-9708ef7c5f08",
                "name": "Edouard Pauwels",
                "collaborators": [
                    "J\u00e9r\u00f4me Bolte",
                    "Jerome Bolte",
                    "Samuel Vaiter",
                    "Jean-Bernard Lasserre",
                    "Swann Marx",
                    "Guillaume Sagnol",
                    "Cheik Traor\u00e9",
                    "Amir Beck",
                    "Shoham Sabach",
                    "Francis Bach"
                ],
                "domain": [
                    "Optimization",
                    "Nonsmooth Analysis",
                    "Machine Learning",
                    "Automatic Differentiation"
                ],
                "publications": [
                    {
                        "title": "The value function approach to convergence analysis in composite optimization",
                        "abstract": "This works aims at understanding further convergence properties of first order local search methods with complex geometries. We focus on the composite optimization model which unifies within a simple formalism many problems of this type. We provide a general convergence analysis of the composite Gauss-Newton method under tameness assumptions (an extension of semi-algebraicity). Tameness is a very general condition satisfied by virtually all problems solved in practice. The analysis is based on recent progresses in understanding convergence properties of sequential convex programming methods through the value function."
                    },
                    {
                        "title": "On the nature of Bregman functions",
                        "abstract": "Let C be convex, compact, with nonempty interior and h be Legendre with domain C, continuous on C. We prove that h is Bregman if and only if it is strictly convex on C and C is a polytope. This provides insights on sequential convergence of many Bregman divergence based algorithm: abstract compatibility conditions between Bregman and Euclidean topology may equivalently be replaced by explicit conditions on h and C. This also emphasizes that a general convergence theory for these methods (beyond polyhedral domains) would require more refinements than Bregman's conditions."
                    },
                    {
                        "title": "Incremental Without Replacement Sampling in Nonconvex Optimization",
                        "abstract": "Minibatch decomposition methods for empirical risk minimization are commonly analysed in a stochastic approximation setting, also known as sampling with replacement. On the other hands modern implementations of such techniques are incremental: they rely on sampling without replacement, for which available analysis are much scarcer. We provide convergence guaranties for the latter variant by analysing a versatile incremental gradient scheme. For this scheme, we consider constant, decreasing or adaptive step sizes. In the smooth setting we obtain explicit complexity estimates in terms of epoch counter. In the nonsmooth setting we prove that the sequence is attracted by solutions of optimality conditions of the problem."
                    },
                    {
                        "title": "Conservative parametric optimality and the ridge method for tame min-max problems",
                        "abstract": "We study the ridge method for min-max problems, and investigate its convergence without any convexity, differentiability or qualification assumption. The central issue is to determine whether the ''parametric optimality formula'' provides a conservative field, a notion of generalized derivative well suited for optimization. The answer to this question is positive in a semi-algebraic, and more generally definable, context. The proof involves a new characterization of definable conservative fields which is of independent interest. As a consequence, the ridge method applied to definable objectives is proved to have a minimizing behavior and to converge to a set of equilibria which satisfy an optimality condition. Definability is key to our proof: we show that for a more general class of nonsmooth functions, conservativity of the parametric optimality formula may fail, resulting in an absurd behavior of the ridge method."
                    },
                    {
                        "title": "Generic Fr{\u00e9}chet stationarity in constrained optimization",
                        "abstract": "Minimizing a smooth function f on a closed subset C leads to different notions of stationarity: Fr{\\'e}chet stationarity, which carries a strong variational meaning, and criticallity, which is defined through a closure process. The latter is an optimality condition which may loose the variational meaning of Fr{\\'e}chet stationarity in some settings. We show that, while criticality is the appropriate notion in full generality, Fr{\\'e}chet stationarity is typical in practical scenarios. This is illustrated with two main results, first we show that if C is semi-algebraic, then for a generic smooth semi-algebraic function f , all critical points of f on C are actually Fr{\\'e}chet stationary. Second we prove that for small step-sizes, all the accumulation points of the projected gradient algorithm are Fr{\\'e}chet stationary, with an explicit global quadratic estimate of the remainder, avoiding potential critical points which are not Fr{\\'e}chet stationary, and some bad local minima."
                    },
                    {
                        "title": "A mathematical model for automatic differentiation in machine learning",
                        "abstract": "Automatic differentiation, as implemented today, does not have a simple mathematical model adapted to the needs of modern machine learning. In this work we articulate the relationships between differentiation of programs as implemented in practice and differentiation of nonsmooth functions. To this end we provide a simple class of functions, a nonsmooth calculus, and show how they apply to stochastic approximation methods. We also evidence the issue of artificial critical points created by algorithmic differentiation and show how usual methods avoid these points with probability one."
                    },
                    {
                        "title": "Majorization-minimization procedures and convergence of SQP methods for semi-algebraic and tame programs",
                        "abstract": "In view of solving nonsmooth and nonconvex problems involving complex constraints (like standard NLP problems), we study general maximization-minimization procedures produced by families of strongly convex sub-problems. Using techniques from semi-algebraic geometry and variational analysis -in particular Lojasiewicz inequality- we establish the convergence of sequences generated by this type of schemes to critical points. The broad applicability of this process is illustrated in the context of NLP. In that case critical points coincide with KKT points. When the data are semi-algebraic or real analytic our method applies (for instance) to the study of various SQP methods: the moving balls method, Sl1QP, ESQP. Under standard qualification conditions, this provides -to the best of our knowledge- the first general convergence results for general nonlinear programming problems. We emphasize the fact that, unlike most works on this subject, no second-order assumption and/or convexity assumptions whatsoever are made. Rate of convergence are shown to be of the same form as those commonly encountered with first order methods."
                    },
                    {
                        "title": "Curiosities and counterexamples in smooth convex optimization",
                        "abstract": "Counterexamples to some old-standing optimization problems in the smooth convex coercive setting are provided. We show that block-coordinate, steepest descent with exact search or Bregman descent methods do not generally converge. Other failures of various desirable features are established: directional convergence of Cauchy's gradient curves, convergence of Newton's flow, finite length of Tikhonov path, convergence of central paths, or smooth Kurdyka-Lojasiewicz inequality. All examples are planar. These examples are based on general smooth convex interpolation results. Given a decreasing sequence of positively curved C k convex compact sets in the plane, we provide a level set interpolation of a C k smooth convex function where k $\\ge$ 2 is arbitrary. If the intersection is reduced to one point our interpolant has positive definite Hessian, otherwise it is positive definite out of the solution set. Furthermore , given a sequence of decreasing polygons we provide an interpolant agreeing with the vertices and whose gradients coincide with prescribed normals."
                    },
                    {
                        "title": "Conservative set valued fields, automatic differentiation, stochastic gradient method and deep learning",
                        "abstract": "Modern problems in AI or in numerical analysis require nonsmooth approaches with a flexible calculus. We introduce generalized derivatives called conservative fields for which we develop a calculus and provide representation formulas. Functions having a conservative field are called path differentiable: convex, concave, Clarke regular and any semialgebraic Lipschitz continuous functions are path differentiable. Using Whitney stratification techniques for semialgebraic and definable sets, our model provides variational formulas for nonsmooth automatic differentiation oracles, as for instance the famous backpropagation algorithm in deep learning. Our differential model is applied to establish the convergence in values of nonsmooth stochastic gradient methods as they are implemented in practice."
                    },
                    {
                        "title": "Path differentiability of ODE flows",
                        "abstract": "We consider flows of ordinary differential equations (ODEs) driven by path differentiable vector fields. Path differentiable functions constitute a proper subclass of Lipschitz functions which admit conservative gradients, a notion of generalized derivative compatible with basic calculus rules. Our main result states that such flows inherit the path differentiability property of the driving vector field. We show indeed that forward propagation of derivatives given by the sensitivity differential inclusions provide a conservative Jacobian for the flow. This allows to propose a nonsmooth version of the adjoint method, which can be applied to integral costs under an ODE constraint. This result constitutes a theoretical ground to the application of small step first order methods to solve a broad class of nonsmooth optimization problems with parametrized ODE constraints. This is illustrated with the convergence of small step first order methods based on the proposed nonsmooth adjoint."
                    },
                    {
                        "title": "The derivatives of Sinkhorn-Knopp converge",
                        "abstract": "We show that the derivatives of the Sinkhorn-Knopp algorithm, or iterative proportional fitting procedure, converge towards the derivatives of the entropic regularization of the optimal transport problem with a locally uniform linear convergence rate."
                    },
                    {
                        "title": "An unexpected connection between Bayes $A-$optimal designs and the Group Lasso",
                        "abstract": "We show that the $A$-optimal design optimization problem over $m$ design points in $\\mathbb{R}^n$ is equivalent to minimizing a quadratic function plus a group lasso sparsity inducing term over $n\\times m$ real matrices. This observation allows to describe several new algorithms for $A$-optimal design based on splitting and block coordinate decomposition. These techniques are well known and proved powerful to treat large scale problems in machine learning and signal processing communities. The proposed algorithms come with rigorous convergence guaranties and convergence rate estimate stemming from the optimization literature. Performances are illustrated on synthetic benchmarks and compared to existing methods for solving the optimal design problem."
                    },
                    {
                        "title": "Sequential convergence of AdaGrad algorithm for smooth convex optimization",
                        "abstract": "We prove that the iterates produced by, either the scalar step size variant, or the coordinatewise variant of AdaGrad algorithm, are convergent sequences when applied to convex objective functions with Lipschitz gradient. The key insight is to remark that such AdaGrad sequences satisfy a variable metric quasi-Fej\\'er monotonicity property, which allows to prove convergence."
                    },
                    {
                        "title": "The Cyclic Block Conditional Gradient Method for Convex Optimization Problems",
                        "abstract": "In this paper we study the convex problem of optimizing the sum of a smooth function and a compactly supported non-smooth term with a specific separable form. We analyze the block version of the generalized conditional gradient method when the blocks are chosen in a cyclic order. A global sublinear rate of convergence is established for two different stepsize strategies commonly used in this class of methods. Numerical comparisons of the proposed method to both the classical conditional gradient algorithm and its random block version demonstrate the effectiveness of the cyclic block update rule."
                    },
                    {
                        "title": "One-step differentiation of iterative algorithms",
                        "abstract": "In appropriate frameworks, automatic differentiation is transparent to the user at the cost of being a significant computational burden when the number of operations is large. For iterative algorithms, implicit differentiation alleviates this issue but requires custom implementation of Jacobian evaluation. In this paper, we study one-step differentiation, also known as Jacobian-free backpropagation, a method as easy as automatic differentiation and as performant as implicit differentiation for fast algorithms (e.g., superlinear optimization methods). We provide a complete theoretical approximation analysis with specific examples (Newton's method, gradient descent) along with its consequences in bilevel optimization. Several numerical examples illustrate the well-foundness of the one-step estimator."
                    },
                    {
                        "title": "Sorting out typicality with the inverse moment matrix SOS polynomial",
                        "abstract": "We study a surprising phenomenon related to the representation of a cloud of data points using polynomials. We start with the previously unnoticed empirical observation that, given a collection (a cloud) of data points, the sublevel sets of a certain distinguished polynomial capture the shape of the cloud very accurately. This distinguished polynomial is a sum-of-squares (SOS) derived in a simple manner from the inverse of the empirical moment matrix. In fact, this SOS polynomial is directly related to orthogonal polynomials and the Christoffel function. This allows to generalize and interpret extremality properties of orthogonal polynomials and to provide a mathematical rationale for the observed phenomenon. Among diverse potential applications, we illustrate the relevance of our results on a network intrusion detection task for which we obtain performances similar to existing dedicated methods reported in the literature."
                    },
                    {
                        "title": "The empirical Christoffel function with applications in data analysis",
                        "abstract": "We illustrate the potential applications in machine learning of the Christoffel function, or more precisely, its empirical counterpart associated with a counting measure uniformly supported on a finite set of points. Firstly, we provide a thresholding scheme which allows to approximate the support of a measure from a finite subset of its moments with strong asymptotic guaranties. Secondly, we provide a consistency result which relates the empirical Christoffel function and its population counterpart in the limit of large samples. Finally, we illustrate the relevance of our results on simulated and real world datasets for several applications in statistics and machine learning: (a) density and support estimation from finite samples, (b) outlier and novelty detection and (c) affine matching."
                    },
                    {
                        "title": "Relating Leverage Scores and Density using Regularized Christoffel Functions",
                        "abstract": "Statistical leverage scores emerged as a fundamental tool for matrix sketching and column sampling with applications to low rank approximation, regression, random feature learning and quadrature. Yet, the very nature of this quantity is barely understood. Borrowing ideas from the orthogonal polynomial literature, we introduce the regularized Christoffel function associated to a positive definite kernel. This uncovers a variational formulation for leverage scores for kernel methods and allows to elucidate their relationships with the chosen kernel as well as population density. Our main result quantitatively describes a decreasing relation between leverage score and population density for a broad class of kernels on Euclidean spaces. Numerical simulations support our findings."
                    },
                    {
                        "title": "Rate of convergence for geometric inference based on the empirical Christoffel function",
                        "abstract": "We consider the problem of estimating the support of a measure from a finite, independent, sample. The estimators which are considered are constructed based on the empirical Christoffel function. Such estimators have been proposed for the problem of set estimation with heuristic justifications. We carry out a detailed finite sample analysis, that allows us to select the threshold and degree parameters as a function of the sample size. We provide a convergence rate analysis of the resulting support estimation procedure. Our analysis establishes that we may obtain finite sample bounds which are comparable to existing rates for different set estimation procedures. Our results rely on concentration inequalities for the empirical Christoffel function and on estimates of the supremum of the Christoffel-Darboux kernel on sets with smooth boundaries, that can be considered of independent interest."
                    },
                    {
                        "title": "Subgradient sampling for nonsmooth nonconvex minimization",
                        "abstract": "Risk minimization for nonsmooth nonconvex problems naturally leads to first-order sampling or, by an abuse of terminology, to stochastic subgradient descent. We establish the convergence of this method in the path-differentiable case and describe more precise results under additional geometric assumptions. We recover and improve results from Ermoliev and Norkin [Cybern. Syst. Anal., 34 (1998), pp. 196--215] by using a different approach: conservative calculus and the ODE method. In the definable case, we show that first-order subgradient sampling avoids artificial critical points with probability one and applies moreover to a large range of risk minimization problems in deep learning, based on the backpropagation oracle. As byproducts of our approach, we obtain several results on integration of independent interest, such as an interchange result for conservative derivatives and integrals or the definability of set-valued parameterized integrals."
                    }
                ]
            },
            "5237bed2-5cd6-4ed0-81ee-a3072c67b8fd": {
                "pk": "5237bed2-5cd6-4ed0-81ee-a3072c67b8fd",
                "name": "Samuel Vaiter",
                "collaborators": [
                    "Nicolas Keriven",
                    "Edouard Pauwels",
                    "Gabriel Peyr\u00e9",
                    "Jalal Fadili",
                    "J\u00e9r\u00f4me Bolte",
                    "Yann Traonmilin",
                    "R\u00e9mi Gribonval",
                    "Alberto Bietti",
                    "Abdessamad Barbara",
                    "Abderrahim Jourani"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Optimization",
                    "Machine Learning",
                    "Regularization"
                ],
                "publications": [
                    {
                        "title": "Maximal Solutions of Sparse Analysis Regularization",
                        "abstract": "This paper deals with the non-uniqueness of the solutions of an analysis-Lasso regularization. Most of previous works in this area is concerned with the case where the solution set is a singleton, or to derive guarantees to enforce uniqueness. Our main contribution consists in providing a geometrical interpretation of a solution with a maximal D-support, namely the fact that such a solution lives in the relative interior of the solution set. With this result in hand, we also provide a way to exhibit a maximal solution using a primal-dual interior point algorithm."
                    },
                    {
                        "title": "The Geometry of Sparse Analysis Regularization",
                        "abstract": "Analysis sparsity is a common prior in inverse problem or machine learning including special cases such as Total Variation regularization, Edge Lasso and Fused Lasso. We study the geometry of the solution set (a polyhedron) of the analysis l1-regularization (with l2 data fidelity term) when it is not reduced to a singleton without any assumption of the analysis dictionary nor the degradation operator. In contrast with most theoretical work, we do not focus on giving uniqueness and/or stability results, but rather describe a worst-case scenario where the solution set can be big in terms of dimension. Leveraging a fine analysis of the sub-level set of the regularizer itself, we draw a connection between support of a solution and the minimal face containing it, and in particular prove that extremal points can be recovered thanks to an algebraic test. Moreover, we draw a connection between the sign pattern of a solution and the ambient dimension of the smallest face containing it. Finally, we show that any arbitrary sub-polyhedra of the level set can be seen as a solution set of sparse analysis regularization with explicit parameters."
                    },
                    {
                        "title": "The derivatives of Sinkhorn-Knopp converge",
                        "abstract": "We show that the derivatives of the Sinkhorn-Knopp algorithm, or iterative proportional fitting procedure, converge towards the derivatives of the entropic regularization of the optimal transport problem with a locally uniform linear convergence rate."
                    },
                    {
                        "title": "The Degrees of Freedom of the Group Lasso",
                        "abstract": "This paper studies the sensitivity to the observations of the block/group Lasso solution to an overdetermined linear regression model. Such a regularization is known to promote sparsity patterns structured as nonoverlapping groups of coefficients. Our main contribution provides a local parameterization of the solution with respect to the observations. As a byproduct, we give an unbiased estimate of the degrees of freedom of the group Lasso. Among other applications of such results, one can choose in a principled and objective way the regularization parameter of the Lasso through model selection criteria."
                    },
                    {
                        "title": "Robust Polyhedral Regularization",
                        "abstract": "In this paper, we establish robustness to noise perturbations of polyhedral regularization of linear inverse problems. We provide a sufficient condition that ensures that the polyhedral face associated to the true vector is equal to that of the recovered one. This criterion also implies that the $\\ell^2$ recovery error is proportional to the noise level for a range of parameter. Our criterion is expressed in terms of the hyperplanes supporting the faces of the unit polyhedral ball of the regularization. This generalizes to an arbitrary polyhedral regularization results that are known to hold for sparse synthesis and analysis $\\ell^1$ regularization which are encompassed in this framework. As a byproduct, we obtain recovery guarantees for $\\ell^\\infty$ and $\\ell^1-\\ell^\\infty$ regularization."
                    },
                    {
                        "title": "Sparse and Smooth: improved guarantees for Spectral Clustering in the Dynamic Stochastic Block Model",
                        "abstract": "In this paper, we analyse classical variants of the Spectral Clustering (SC) algorithm in the Dynamic Stochastic Block Model (DSBM). Existing results show that, in the relatively sparse case where the expected degree grows logarithmically with the number of nodes, guarantees in the static case can be extended to the dynamic case and yield improved error bounds when the DSBM is sufficiently smooth in time, that is, the communities do not change too much between two time steps. We improve over these results by drawing a new link between the sparsity and the smoothness of the DSBM: the more regular the DSBM is, the more sparse it can be, while still guaranteeing consistent recovery. In particular, a mild condition on the smoothness allows to treat the sparse case with bounded degree. We also extend these guarantees to the normalized Laplacian, and as a by-product of our analysis, we obtain to our knowledge the best spectral concentration bound available for the normalized Laplacian of matrices with independent Bernoulli entries."
                    },
                    {
                        "title": "Gradient scarcity with Bilevel Optimization for Graph Learning",
                        "abstract": "A common issue in graph learning under the semi-supervised setting is referred to as gradient scarcity. That is, learning graphs by minimizing a loss on a subset of nodes causes edges between unlabelled nodes that are far from labelled ones to receive zero gradients. The phenomenon was first described when optimizing the graph and the weights of a Graph Neural Network (GCN) with a joint optimization algorithm. In this work, we give a precise mathematical characterization of this phenomenon, and prove that it also emerges in bilevel optimization, where additional dependency exists between the parameters of the problem. While for GCNs gradient scarcity occurs due to their finite receptive field, we show that it also occurs with the Laplacian regularization model, in the sense that gradients amplitude decreases exponentially with distance to labelled nodes. To alleviate this issue, we study several solutions: we propose to resort to latent graph learning using a Graph-to-Graph model (G2G), graph regularization to impose a prior structure on the graph, or optimizing on a larger graph than the original one with a reduced diameter. Our experiments on synthetic and real datasets validate our analysis and prove the efficiency of the proposed solutions."
                    },
                    {
                        "title": "A sharp oracle inequality for Graph-Slope",
                        "abstract": "Following recent success on the analysis of the Slope estimator, we provide a sharp oracle inequality in term of prediction error for Graph-Slope, a generalization of Slope to signals observed over a graph. In addition to improving upon best results obtained so far for the Total Variation denoiser (also referred to as Graph-Lasso or Generalized Lasso), we propose an efficient algorithm to compute Graph-Slope. The proposed algorithm is obtained by applying the forward-backward method to the dual formulation of the Graph-Slope optimization problem. We also provide experiments showing the interest of the method."
                    },
                    {
                        "title": "Linear Support Vector Regression with Linear Constraints",
                        "abstract": "This paper studies the addition of linear constraints to the Support Vector Regression (SVR) when the kernel is linear. Adding those constraints into the problem allows to add prior knowledge on the estimator obtained, such as finding probability vector or monotone data. We propose a generalization of the Sequential Minimal Optimization (SMO) algorithm for solving the optimization problem with linear constraints and prove its convergence. Then, practical performances of this estimator are shown on simulated and real datasets with different settings: non negative regression, regression onto the simplex for biomedical data and isotonic regression for weather forecast."
                    },
                    {
                        "title": "One-step differentiation of iterative algorithms",
                        "abstract": "In appropriate frameworks, automatic differentiation is transparent to the user at the cost of being a significant computational burden when the number of operations is large. For iterative algorithms, implicit differentiation alleviates this issue but requires custom implementation of Jacobian evaluation. In this paper, we study one-step differentiation, also known as Jacobian-free backpropagation, a method as easy as automatic differentiation and as performant as implicit differentiation for fast algorithms (e.g., superlinear optimization methods). We provide a complete theoretical approximation analysis with specific examples (Newton's method, gradient descent) along with its consequences in bilevel optimization. Several numerical examples illustrate the well-foundness of the one-step estimator."
                    },
                    {
                        "title": "What functions can Graph Neural Networks compute on random graphs? The role of Positional Encoding",
                        "abstract": "We aim to deepen the theoretical understanding of Graph Neural Networks (GNNs) on large graphs, with a focus on their expressive power. Existing analyses relate this notion to the graph isomorphism problem, which is mostly relevant for graphs of small sizes, or studied graph classification or regression tasks, while prediction tasks on nodes are far more relevant on large graphs. Recently, several works showed that, on very general random graphs models, GNNs converge to certains functions as the number of nodes grows. In this paper, we provide a more complete and intuitive description of the function space generated by equivariant GNNs for node-tasks, through general notions of convergence that encompass several previous examples. We emphasize the role of input node features, and study the impact of node Positional Encodings (PEs), a recent line of work that has been shown to yield state-of-the-art results in practice. Through the study of several examples of PEs on large random graphs, we extend previously known universality results to significantly more general models. Our theoretical results hint at some normalization tricks, which is shown numerically to have a positive impact on GNN generalization on synthetic and real data. Our proofs contain new concentration inequalities of independent interest."
                    },
                    {
                        "title": "Risk Estimate under a Nonstationary Autoregressive Model for Data-Driven Reproduction Number Estimation",
                        "abstract": "COVID-19 pandemic has brought to the fore epidemiological models which, though describing a rich variety of behaviors, have previously received little attention in the signal processing literature. During the pandemic, several works successfully leveraged state-of-the-art signal processing strategies to robustly infer epidemiological indicators despite the low quality of COVID-19 data. In the present work, a novel nonstationary autoregressive model is introduced, encompassing, but not reducing to, one of the most popular models for the propagation of viral epidemics. Using a variational framework, penalized likelihood estimators of the parameters of this new model are designed. In practice, the main bottleneck is that the estimation accuracy strongly depends on hyperparameters tuning. Without available ground truth, hyperparameters are selected by minimizing specifically designed data-driven oracles, used as proxy for the estimation error. Focusing on the nonstationary autoregressive Poisson model, the Stein's Unbiased Risk Estimate formalism is generalized to construct asymptotically unbiased risk estimators based on the derivation of an original autoregressive counterpart of Stein's lemma. The accuracy of these oracles and of the resulting estimates are assessed through intensive Monte Carlo simulations on synthetic data. Then, elaborating on recent epidemiological models, a novel weekly scaled Poisson model is proposed, enabling to better account for intrinsic variability of the contamination while being robust to reporting errors. Finally, the overall data-driven procedure is particularized to the estimation of COVID-19 reproduction number and exemplified on real COVID-19 infection counts in different countries and at different stages of the pandemic, demonstrating its ability to yield consistent estimates."
                    },
                    {
                        "title": "Is the 1-norm the best convex sparse regularization?",
                        "abstract": "The 1-norm is a good convex regularization for the recovery of sparse vectors from under-determined linear measurements. No other convex regularization seems to surpass its sparse recovery performance. How can this be explained? To answer this question, we define several notions of \"best\" (convex) regulariza-tion in the context of general low-dimensional recovery and show that indeed the 1-norm is an optimal convex sparse regularization within this framework."
                    },
                    {
                        "title": "Provable local learning rule by expert aggregation for a Hawkes network",
                        "abstract": "We propose a simple network of Hawkes processes as a cognitive model capable of learning to classify objects. Our learning algorithm, named HAN for Hawkes Aggregation of Neurons, is based on a local synaptic learning rule based on spiking probabilities at each output node. We were able to use local regret bounds to prove mathematically that the network is able to learn on average and even asymptotically under more restrictive assumptions."
                    },
                    {
                        "title": "Convergence and Stability of Graph Convolutional Networks on Large Random Graphs",
                        "abstract": "We study properties of Graph Convolutional Networks (GCNs) by analyzing their behavior on standard models of random graphs, where nodes are represented by random latent variables and edges are drawn according to a similarity kernel. This allows us to overcome the difficulties of dealing with discrete notions such as isomorphisms on very large graphs, by considering instead more natural geometric aspects. We first study the convergence of GCNs to their continuous counterpart as the number of nodes grows. Our results are fully non-asymptotic and are valid for relatively sparse graphs with an average degree that grows logarithmically with the number of nodes. We then analyze the stability of GCNs to small deformations of the random graph model. In contrast to previous studies of stability in discrete settings, our continuous setup allows us to provide more intuitive deformation-based metrics for understanding stability, which have proven useful for explaining the success of convolutional representations on Euclidean domains."
                    },
                    {
                        "title": "On the Universality of Graph Neural Networks on Large Random Graphs",
                        "abstract": "We study the approximation power of Graph Neural Networks (GNNs) on latent position random graphs. In the large graph limit, GNNs are known to converge to certain \"continuous\" models known as c-GNNs, which directly enables a study of their approximation power on random graph models. In the absence of input node features however, just as GNNs are limited by the Weisfeiler-Lehman isomorphism test, c-GNNs will be severely limited on simple random graph models. For instance, they will fail to distinguish the communities of a well-separated Stochastic Block Model (SBM) with constant degree function. Thus, we consider recently proposed architectures that augment GNNs with unique node identifiers, referred to as Structural GNNs here (SGNNs). We study the convergence of SGNNs to their continuous counterpart (c-SGNNs) in the large random graph limit, under new conditions on the node identifiers. We then show that c-SGNNs are strictly more powerful than c-GNNs in the continuous limit, and prove their universality on several random graph models of interest, including most SBMs and a large class of random geometric graphs. Our results cover both permutation-invariant and permutation-equivariant architectures."
                    },
                    {
                        "title": "A theory of optimal convex regularization for low-dimensional recovery",
                        "abstract": "We consider the problem of recovering elements of a low-dimensional model from under-determined linear measurements. To perform recovery, we consider the minimization of a convex regularizer subject to a data fit constraint. Given a model, we ask ourselves what is the \"best\" convex regularizer to perform its recovery. To answer this question, we define an optimal regularizer as a function that maximizes a compliance measure with respect to the model. We introduce and study several notions of compliance. We give analytical expressions for compliance measures based on the best-known recovery guarantees with the restricted isometry property. These expressions permit to show the optimality of the ${\\ell}$1-norm for sparse recovery and of the nuclear norm for low-rank matrix recovery for these compliance measures. We also investigate the construction of an optimal convex regularizer using the examples of sparsity in levels and of sparse plus low-rank models."
                    },
                    {
                        "title": "Automatic differentiation of nonsmooth iterative algorithms",
                        "abstract": "Differentiation along algorithms, i.e., piggyback propagation of derivatives, is now routinely used to differentiate iterative solvers in differentiable programming. Asymptotics is well understood for many smooth problems but the nondifferentiable case is hardly considered. Is there a limiting object for nonsmooth piggyback automatic differentiation (AD)? Does it have any variational meaning and can it be used effectively in machine learning? Is there a connection with classical derivative? All these questions are addressed under appropriate nonexpansivity conditions in the framework of conservative derivatives which has proved useful in understanding nonsmooth AD. For nonsmooth piggyback iterations, we characterize the attractor set of nonsmooth piggyback iterations as a set-valued fixed point which remains in the conservative framework. This has various consequences and in particular almost everywhere convergence of classical derivatives. Our results are illustrated on parametric convex optimization problems with forward-backward, Douglas-Rachford and Alternating Direction of Multiplier algorithms as well as the Heavy-Ball method."
                    },
                    {
                        "title": "A Lower Bound and a Near-Optimal Algorithm for Bilevel Empirical Risk Minimization",
                        "abstract": "Bilevel optimization problems, which are problems where two optimization problems are nested, have more and more applications in machine learning. In many practical cases, the upper and the lower objectives correspond to empirical risk minimization problems and therefore have a sum structure. In this context, we propose a bilevel extension of the celebrated SARAH algorithm. We demonstrate that the algorithm requires $\\mathcal{O}((n+m)^{\\frac12}\\varepsilon^{-1})$ oracle calls to achieve $\\varepsilon$-stationarity with $n+m$ the total number of samples, which improves over all previous bilevel algorithms. Moreover, we provide a lower bound on the number of oracle calls required to get an approximate stationary point of the objective function of the bilevel problem. This lower bound is attained by our algorithm, making it optimal in terms of sample complexity."
                    }
                ]
            }
        }
    },
    "2404.03592": {
        "paper_data": {
            "title": "ReFT: Representation Finetuning for Language Models",
            "url": "http://arxiv.org/abs/2404.03592v3",
            "arxiv_id": "2404.03592",
            "authors": [
                "Zhengxuan Wu",
                "Aryaman Arora",
                "Zheng Wang",
                "Atticus Geiger",
                "Dan Jurafsky",
                "Christopher D. Manning",
                "Christopher Potts"
            ],
            "abstract": "Parameter-efficient finetuning (PEFT) methods seek to adapt large neural models via updates to a small number of weights. However, much prior interpretability work has shown that representations encode rich semantic information, suggesting that editing representations might be a more powerful alternative. We pursue this hypothesis by developing a family of Representation Finetuning (ReFT) methods. ReFT methods operate on a frozen base model and learn task-specific interventions on hidden representations. We define a strong instance of the ReFT family, Low-rank Linear Subspace ReFT (LoReFT), and we identify an ablation of this method that trades some performance for increased efficiency. Both are drop-in replacements for existing PEFTs and learn interventions that are 15x--65x more parameter-efficient than LoRA. We showcase LoReFT on eight commonsense reasoning tasks, four arithmetic reasoning tasks, instruction-tuning, and GLUE. In all these evaluations, our ReFTs deliver the best balance of efficiency and performance, and almost always outperform state-of-the-art PEFTs. We release a generic ReFT training library publicly at https://github.com/stanfordnlp/pyreft.",
            "introduction": " Introduction Pretrained language models (LMs) are frequently finetuned to adapt them to new domains or tasks [Dai and Le, 2015]. With finetuning, a single base model can be adapted to a variety of tasks given only small amounts of in-domain data. However, finetuning large LMs is expensive. Parameter- efficient finetuning (PEFT) appendix F.1, but with a different training dataset. Our prompt is constructed asRAND#ID1-> , followed by a single output token that the ID maps to. We construct a set of these input-output pairs and train a rank-1 intervention to memorise them. We present our results in fig. 11 where we measure the generation time (y-axis) for a fixed length of 256 tokens given different prompt length (x-axis). Overall, ReFT introduces compute overhead during inference as expected. Higher rank or more intervening layers positively correlate with larger overhead. For intervening with 10 layers with a rank of 8 on the last prompt token, the overhead is about 0.05second. 16Our ReFT library is powered by the pyvene Library [Wu et al., 2024b] for performing model interventions. Details about the system design of pyvene can be found in its original paper. 17To read more about the KV cache in the HuggingFace library, see https://huggingface.co/docs/ transformers/main/en/llm_tutorial_optimization . 39P ositions Rank La y ers Pr ompt t ok ensTimeFigure 11: Runtime vs. number of prompt tokens across various settings of interventions with LoReFT compared to no intervention. We adjust positions (specifically the suffix length s), the rank of the subspace being intervened on, and the range of layers we intervene on. 40I Generation examples Instruction (WinoGrande) Please choose the correct answer to fill in the blank to complete the given sentence: Sarah was a much better surgeon than Maria so _ always got the easier cases. Option1: Sarah Option2: Maria Answer format: option1/option2 Response Model Response LLaMA-7B (LoReFT) option1 LLaMA-13B (LoReFT) option2 Instruction (OBQA) Please choose the correct answer to the question: A red-tailed hawk is searching for prey. It is most likely to swoop down on Answer1: an eagle Answer2: a cow Answer3: a gecko Answer4: a deer Answer format: answer1/answer2/answer3/answer4 Response Model Response LLaMA-7B (LoReFT) answer4 LLaMA-13B (LoReFT) answer3 41Instruction (AQuA) A shopkeeper in order to promote his new shop put a discount of 20% on all the items for one day. Now he must sell the items at original price the other day. By what percentage must he increase the price to original? Answer Choices: (A) 21% (B) 20% (C) 25% (D) 33% (E) 18% Response Model Response LLaMA-7B (LoReFT) Let\u2019s assume that the original price of an item is x. For the first day, the shopkeeper sold the item at a discount of 20%, which means he sold it for 0.8x (80% of the original price). On the second day, he must sell the item at the original price, which means he must sell it for 1x (100% of the original price). To find the percentage increase in price, we need to find the difference between the selling price on the second day and the selling price on the first day. Selling price on the second day = 1x Selling price on the first day = 0.8x Percentage increase = (1x - 0.8x) / 0.8x * 100% Percentage increase = (x - 0.8x) / 0.8x * 100% Percentage increase = 0.2x / 0.8x * 100% Percentage increase = 0.25 * 100% Percentage increase = 25% Therefore, the answer is option",
            "references": [
                {
                    "title": "Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey",
                    "abstract": "Large models represent a groundbreaking advancement in multiple application fields, enabling remarkable achievements across various tasks. However, their unprecedented scale comes with significant computational costs. These models, often consisting of billions of parameters, require vast amounts of computational resources for execution. Especially, the expansive scale and computational demands pose considerable challenges when customizing them for particular downstream tasks, particularly over the hardware platforms constrained by computational capabilities. Parameter Efficient Fine-Tuning (PEFT) provides a practical solution by efficiently adjusting the large models over the various downstream tasks. In particular, PEFT refers to the process of adjusting the parameters of a pre-trained large model to adapt it to a specific task or domain while minimizing the number of additional parameters introduced or computational resources required. This approach is particularly important when dealing with large-scale language models with high parameter counts, as fine-tuning these models from scratch can be computationally expensive and resource-intensive, posing considerable challenges in the supporting system platform design. In this survey, we present comprehensive studies of various PEFT algorithms, examining their performance and computational overhead. Moreover, we provide an overview of applications developed using different PEFT algorithms and discuss common techniques employed to mitigate computation costs for PEFT. In addition to providing an extensive survey from an algorithmic standpoint, we also examine various real-world system designs to investigate the implementation costs associated with different PEFT approaches. This survey serves as a valuable resource for researchers aiming to understand both the PEFT algorithm and its system implementation, offering detailed ......"
                },
                {
                    "title": "AutoLoRA: Automatically Tuning Matrix Ranks in Low-Rank Adaptation Based on Meta Learning",
                    "abstract": "Large-scale pretraining followed by task-specific finetuning has achieved great success in various NLP tasks. Since finetuning all parameters of large pretrained models poses substantial computational and memory challenges, several efficient finetuning methods have been developed. Among them, low-rank adaptation (LoRA), which finetunes low-rank incremental update matrices on top of frozen pretrained weights, has proven particularly effective. Nonetheless, LoRA\u2019s uniform rank assignment across all layers, along with its reliance on an exhaustive search to find the best rank, leads to high computation costs and suboptimal finetuning performance. To address these limitations, we introduce AutoLoRA, a meta learning based framework for automatically identifying the optimal rank of each LoRA layer. AutoLoRA associates each rank-1 matrix in a low-rank update matrix with a selection variable, which determines whether the rank-1 matrix should be discarded. A meta learning based method is developed to learn these selection variables. The optimal rank is determined by thresholding the values of these variables. Our comprehensive experiments on natural language understanding, generation, and sequence labeling demonstrate the effectiveness of AutoLoRA. The code is publicly available at https://github.com/ruz048/AutoLoRA"
                },
                {
                    "title": "pyvene: A Library for Understanding and Improving PyTorch Models via Interventions",
                    "abstract": "Interventions on model-internal states are fundamental operations in many areas of AI, including model editing, steering, robustness, and interpretability. To facilitate such research, we introduce pyvene, an open-source Python library that supports customizable interventions on a range of different PyTorch modules. pyvene supports complex intervention schemes with an intuitive configuration format, and its interventions can be static or include trainable parameters. We show how pyvene provides a unified and extensible framework for performing interventions on neural models and sharing the intervened upon models with others. We illustrate the power of the library via interpretability analyses using causal abstraction and knowledge localization. We publish our library through Python Package Index (PyPI) and provide code, documentation, and tutorials at \u2018https://github.com/stanfordnlp/pyvene\u2018."
                },
                {
                    "title": "ResLoRA: Identity Residual Mapping in Low-Rank Adaption",
                    "abstract": "As one of the most popular parameter-efficient fine-tuning (PEFT) methods, low-rank adaptation (LoRA) is commonly applied to fine-tune large language models (LLMs). However, updating the weights of LoRA blocks effectively and expeditiously is challenging due to the long calculation path in the original model. To address this, we propose ResLoRA, an improved framework of LoRA. By adding residual paths during training and using merging approaches to eliminate these extra paths during inference, our method can achieve better results in fewer training steps without any extra trainable parameters or inference cost compared to LoRA. The experiments on NLG, NLU, and text-to-image tasks demonstrate the effectiveness of our method. To the best of our knowledge, ResLoRA is the first work that combines the residual path with LoRA. The code of our method is available at https://github.com/microsoft/LMOps/tree/main/reslora ."
                },
                {
                    "title": "RAVEL: Evaluating Interpretability Methods on Disentangling Language Model Representations",
                    "abstract": "Individual neurons participate in the representation of multiple high-level concepts. To what extent can different interpretability methods successfully disentangle these roles? To help address this question, we introduce RAVEL (Resolving Attribute-Value Entanglements in Language Models), a dataset that enables tightly controlled, quantitative comparisons between a variety of existing interpretability methods. We use the resulting conceptual framework to define the new method of Multi-task Distributed Alignment Search (MDAS), which allows us to find distributed representations satisfying multiple causal criteria. With Llama2-7B as the target language model, MDAS achieves state-of-the-art results on RAVEL, demonstrating the importance of going beyond neuron-level analyses to identify features distributed across activations. We release our benchmark at https://github.com/explanare/ravel."
                },
                {
                    "title": "Multi-LoRA Composition for Image Generation",
                    "abstract": "Low-Rank Adaptation (LoRA) is extensively utilized in text-to-image models for the accurate rendition of specific elements like distinct characters or unique styles in generated images. Nonetheless, existing methods face challenges in effectively composing multiple LoRAs, especially as the number of LoRAs to be integrated grows, thus hindering the creation of complex imagery. In this paper, we study multi-LoRA composition through a decoding-centric perspective. We present two training-free methods: LoRA Switch, which alternates between different LoRAs at each denoising step, and LoRA Composite, which simultaneously incorporates all LoRAs to guide more cohesive image synthesis. To evaluate the proposed approaches, we establish ComposLoRA, a new comprehensive testbed as part of this research. It features a diverse range of LoRA categories with 480 composition sets. Utilizing an evaluation framework based on GPT-4V, our findings demonstrate a clear improvement in performance with our methods over the prevalent baseline, particularly evident when increasing the number of LoRAs in a composition."
                },
                {
                    "title": "Advancing Parameter Efficiency in Fine-tuning via Representation Editing",
                    "abstract": "Parameter Efficient Fine-Tuning (PEFT) techniques have drawn significant attention due to their ability to yield competitive results while updating only a small portion of the adjustable parameters. However, existing PEFT methods pose challenges in hyperparameter selection, such as choosing the rank for LoRA or Adapter, or specifying the length of soft prompts. To address these challenges, we propose a novel fine-tuning approach for neural models, named Representation EDiting (RED), which modifies the representations generated at some layers through the application of scaling and biasing operations. While existing PEFT methods still demonstrate over-parameterization that could potentially undermine the generalization ability acquired from pre-training, RED can substantially reduce the number of trainable parameters by a factor of 25, 700 compared to full parameter fine-tuning and by a factor of 32 relative to LoRA. Remarkably, RED achieves results comparable or superior to both full parameter fine-tuning and other PEFT methods. Extensive experiments across various model architectures and scales, including RoBERTa, GPT-2, T5, and LLaMA-2, have demonstrated the effectiveness and efficiency of RED1, thereby positioning it as a promising PEFT strategy for large-scale neural models."
                },
                {
                    "title": "CausalGym: Benchmarking causal interpretability methods on linguistic tasks",
                    "abstract": "Language models (LMs) have proven to be powerful tools for psycholinguistic research, but most prior work has focused on purely behavioural measures (e.g., surprisal comparisons). At the same time, research in model interpretability has begun to illuminate the abstract causal mechanisms shaping LM behavior. To help bring these strands of research closer together, we introduce CausalGym. We adapt and expand the SyntaxGym suite of tasks to benchmark the ability of interpretability methods to causally affect model behaviour. To illustrate how CausalGym can be used, we study the pythia models (14M--6.9B) and assess the causal efficacy of a wide range of interpretability methods, including linear probing and distributed alignment search (DAS). We find that DAS outperforms the other methods, and so we use it to study the learning trajectory of two difficult linguistic phenomena in pythia-1b: negative polarity item licensing and filler--gap dependencies. Our analysis shows that the mechanism implementing both of these tasks is learned in discrete stages, not gradually."
                },
                {
                    "title": "Representation Surgery: Theory and Practice of Affine Steering",
                    "abstract": "Language models often exhibit undesirable behavior, e.g., generating toxic or gender-biased text. In the case of neural language models, an encoding of the undesirable behavior is often present in the model's representations. Thus, one natural (and common) approach to prevent the model from exhibiting undesirable behavior is to steer the model's representations in a manner that reduces the probability of it generating undesirable text. This paper investigates the formal and empirical properties of steering functions, i.e., transformation of the neural language model's representations that alter its behavior. First, we derive two optimal, in the least-squares sense, affine steering functions under different constraints. Our theory provides justification for existing approaches and offers a novel, improved steering approach. Second, we offer a series of experiments that demonstrate the empirical effectiveness of the methods in mitigating bias and reducing toxic generation."
                },
                {
                    "title": "DoRA: Weight-Decomposed Low-Rank Adaptation",
                    "abstract": "Among the widely used parameter-efficient fine-tuning (PEFT) methods, LoRA and its variants have gained considerable popularity because of avoiding additional inference costs. However, there still often exists an accuracy gap between these methods and full fine-tuning (FT). In this work, we first introduce a novel weight decomposition analysis to investigate the inherent differences between FT and LoRA. Aiming to resemble the learning capacity of FT from the findings, we propose Weight-Decomposed Low-Rank Adaptation (DoRA). DoRA decomposes the pre-trained weight into two components, magnitude and direction, for fine-tuning, specifically employing LoRA for directional updates to efficiently minimize the number of trainable parameters. By employing \\ours, we enhance both the learning capacity and training stability of LoRA while avoiding any additional inference overhead. \\ours~consistently outperforms LoRA on fine-tuning LLaMA, LLaVA, and VL-BART on various downstream tasks, such as commonsense reasoning, visual instruction tuning, and image/video-text understanding. Code is available at https://github.com/NVlabs/DoRA."
                },
                {
                    "title": "A Reply to Makelov et al. (2023)'s \"Interpretability Illusion\" Arguments",
                    "abstract": "We respond to the recent paper by Makelov et al. (2023), which reviews subspace interchange intervention methods like distributed alignment search (DAS; Geiger et al. 2023) and claims that these methods potentially cause\"interpretability illusions\". We first review Makelov et al. (2023)'s technical notion of what an\"interpretability illusion\"is, and then we show that even intuitive and desirable explanations can qualify as illusions in this sense. As a result, their method of discovering\"illusions\"can reject explanations they consider\"non-illusory\". We then argue that the illusions Makelov et al. (2023) see in practice are artifacts of their training and evaluation paradigms. We close by emphasizing that, though we disagree with their core characterization, Makelov et al. (2023)'s examples and discussion have undoubtedly pushed the field of interpretability forward."
                },
                {
                    "title": "Aligner: One Global Token is Worth Millions of Parameters When Aligning Large Language Models",
                    "abstract": "We introduce Aligner, a novel Parameter-Efficient Fine-Tuning (PEFT) method for aligning multi-billion-parameter-sized Large Language Models (LLMs). Aligner employs a unique design that constructs a globally shared set of tunable tokens that modify the attention of every layer. Remarkably with this method, even when using one token accounting for a mere 5,000 parameters, Aligner can still perform comparably well to state-of-the-art LLM adaptation methods like LoRA that require millions of parameters. This capacity is substantiated in both instruction following and value alignment tasks. Besides the multiple order-of-magnitude improvement in parameter efficiency, the insight Aligner provides into the internal mechanisms of LLMs is also valuable. The architectural features and efficacy of our method, in addition to our experiments demonstrate that an LLM separates its internal handling of\"form\"and\"knowledge\"in a somewhat orthogonal manner. This finding promises to motivate new research into LLM mechanism understanding and value alignment."
                },
                {
                    "title": "Scaling Laws for Adversarial Attacks on Language Model Activations",
                    "abstract": "We explore a class of adversarial attacks targeting the activations of language models. By manipulating a relatively small subset of model activations, $a$, we demonstrate the ability to control the exact prediction of a significant number (in some cases up to 1000) of subsequent tokens $t$. We empirically verify a scaling law where the maximum number of target tokens $t_\\mathrm{max}$ predicted depends linearly on the number of tokens $a$ whose activations the attacker controls as $t_\\mathrm{max} = \\kappa a$. We find that the number of bits of control in the input space needed to control a single bit in the output space (what we call attack resistance $\\chi$) is remarkably constant between $\\approx 16$ and $\\approx 25$ over 2 orders of magnitude of model sizes for different language models. Compared to attacks on tokens, attacks on activations are predictably much stronger, however, we identify a surprising regularity where one bit of input steered either via activations or via tokens is able to exert control over a similar amount of output bits. This gives support for the hypothesis that adversarial attacks are a consequence of dimensionality mismatch between the input and output spaces. A practical implication of the ease of attacking language model activations instead of tokens is for multi-modal and selected retrieval models, where additional data sources are added as activations directly, sidestepping the tokenized input. This opens up a new, broad attack surface. By using language models as a controllable test-bed to study adversarial attacks, we were able to experiment with input-output dimensions that are inaccessible in computer vision, especially where the output dimension dominates."
                },
                {
                    "title": "Is This the Subspace You Are Looking for? An Interpretability Illusion for Subspace Activation Patching",
                    "abstract": "Mechanistic interpretability aims to understand model behaviors in terms of specific, interpretable features, often hypothesized to manifest as low-dimensional subspaces of activations. Specifically, recent studies have explored subspace interventions (such as activation patching) as a way to simultaneously manipulate model behavior and attribute the features behind it to given subspaces. In this work, we demonstrate that these two aims diverge, potentially leading to an illusory sense of interpretability. Counterintuitively, even if a subspace intervention makes the model's output behave as if the value of a feature was changed, this effect may be achieved by activating a dormant parallel pathway leveraging another subspace that is causally disconnected from model outputs. We demonstrate this phenomenon in a distilled mathematical example, in two real-world domains (the indirect object identification task and factual recall), and present evidence for its prevalence in practice. In the context of factual recall, we further show a link to rank-1 fact editing, providing a mechanistic explanation for previous work observing an inconsistency between fact editing performance and fact localization. However, this does not imply that activation patching of subspaces is intrinsically unfit for interpretability. To contextualize our findings, we also show what a success case looks like in a task (indirect object identification) where prior manual circuit analysis informs an understanding of the location of a feature. We explore the additional evidence needed to argue that a patched subspace is faithful."
                },
                {
                    "title": "SiRA: Sparse Mixture of Low Rank Adaptation",
                    "abstract": "Parameter Efficient Tuning has been an prominent approach to adapt the Large Language Model to downstream tasks. Most previous works considers adding the dense trainable parameters, where all parameters are used to adapt certain task. We found this less effective empirically using the example of LoRA that introducing more trainable parameters does not help. Motivated by this we investigate the importance of leveraging\"sparse\"computation and propose SiRA: sparse mixture of low rank adaption. SiRA leverages the Sparse Mixture of Expert(SMoE) to boost the performance of LoRA. Specifically it enforces the top $k$ experts routing with a capacity limit restricting the maximum number of tokens each expert can process. We propose a novel and simple expert dropout on top of gating network to reduce the over-fitting issue. Through extensive experiments, we verify SiRA performs better than LoRA and other mixture of expert approaches across different single tasks and multitask settings."
                },
                {
                    "title": "In-context Vectors: Making In Context Learning More Effective and Controllable Through Latent Space Steering",
                    "abstract": "Large language models (LLMs) demonstrate emergent in-context learning capabilities, where they adapt to new tasks based on example demonstrations. However, in-context learning has seen limited effectiveness in many settings, is difficult to quantitatively control and takes up context window space. To overcome these limitations, we propose an alternative approach that recasts in-context learning as in-context vectors (ICV). Using ICV has two steps. We first use a forward pass on demonstration examples to create the in-context vector from the latent embedding of the LLM. This vector captures essential information about the intended task. On a new query, instead of adding demonstrations to the prompt, we shift the latent states of the LLM using the ICV. The ICV approach has several benefits: 1) it enables the LLM to more effectively follow the demonstration examples; 2) it's easy to control by adjusting the magnitude of the ICV; 3) it reduces the length of the prompt by removing the in-context demonstrations; 4) ICV is computationally much more efficient than fine-tuning. We demonstrate that ICV achieves better performance compared to standard in-context learning and fine-tuning on diverse tasks including safety, style transfer, role-playing and formatting. Moreover, we show that we can flexibly teach LLM to simultaneously follow different types of instructions by simple vector arithmetics on the corresponding ICVs."
                },
                {
                    "title": "Parameter-Efficient Orthogonal Finetuning via Butterfly Factorization",
                    "abstract": "Large foundation models are becoming ubiquitous, but training them from scratch is prohibitively expensive. Thus, efficiently adapting these powerful models to downstream tasks is increasingly important. In this paper, we study a principled finetuning paradigm -- Orthogonal Finetuning (OFT) -- for downstream task adaptation. Despite demonstrating good generalizability, OFT still uses a fairly large number of trainable parameters due to the high dimensionality of orthogonal matrices. To address this, we start by examining OFT from an information transmission perspective, and then identify a few key desiderata that enable better parameter-efficiency. Inspired by how the Cooley-Tukey fast Fourier transform algorithm enables efficient information transmission, we propose an efficient orthogonal parameterization using butterfly structures. We apply this parameterization to OFT, creating a novel parameter-efficient finetuning method, called Orthogonal Butterfly (BOFT). By subsuming OFT as a special case, BOFT introduces a generalized orthogonal finetuning framework. Finally, we conduct an extensive empirical study of adapting large vision transformers, large language models, and text-to-image diffusion models to various downstream tasks in vision and language."
                },
                {
                    "title": "The Linear Representation Hypothesis and the Geometry of Large Language Models",
                    "abstract": "Informally, the 'linear representation hypothesis' is the idea that high-level concepts are represented linearly as directions in some representation space. In this paper, we address two closely related questions: What does\"linear representation\"actually mean? And, how do we make sense of geometric notions (e.g., cosine similarity or projection) in the representation space? To answer these, we use the language of counterfactuals to give two formalizations of\"linear representation\", one in the output (word) representation space, and one in the input (sentence) space. We then prove these connect to linear probing and model steering, respectively. To make sense of geometric notions, we use the formalization to identify a particular (non-Euclidean) inner product that respects language structure in a sense we make precise. Using this causal inner product, we show how to unify all notions of linear representation. In particular, this allows the construction of probes and steering vectors using counterfactual pairs. Experiments with LLaMA-2 demonstrate the existence of linear representations of concepts, the connection to interpretation and control, and the fundamental role of the choice of inner product."
                },
                {
                    "title": "When Language Models Fall in Love: Animacy Processing in Transformer Language Models",
                    "abstract": "Animacy - whether an entity is alive and sentient - is fundamental to cognitive processing, impacting areas such as memory, vision, and language. However, animacy is not always expressed directly in language: in English it often manifests indirectly, in the form of selectional constraints on verbs and adjectives. This poses a potential issue for transformer language models (LMs): they often train only on text, and thus lack access to extralinguistic information from which humans learn about animacy. We ask: how does this impact LMs' animacy processing - do they still behave as humans do? We answer this question using open-source LMs. Like previous studies, we find that LMs behave much like humans when presented with entities whose animacy is typical. However, we also show that even when presented with stories about atypically animate entities, such as a peanut in love, LMs adapt: they treat these entities as animate, though they do not adapt as well as humans. Even when the context indicating atypical animacy is very short, LMs pick up on subtle clues and change their behavior. We conclude that despite the limited signal through which LMs can learn about animacy, they are indeed sensitive to the relevant lexical semantic nuances available in English."
                },
                {
                    "title": "Verb Conjugation in Transformers Is Determined by Linear Encodings of Subject Number",
                    "abstract": "Deep architectures such as Transformers are sometimes criticized for having uninterpretable\"black-box\"representations. We use causal intervention analysis to show that, in fact, some linguistic features are represented in a linear, interpretable format. Specifically, we show that BERT's ability to conjugate verbs relies on a linear encoding of subject number that can be manipulated with predictable effects on conjugation accuracy. This encoding is found in the subject position at the first layer and the verb position at the last layer, but distributed across positions at middle layers, particularly when there are multiple cues to subject number."
                },
                {
                    "title": "Linear Representations of Sentiment in Large Language Models",
                    "abstract": "Sentiment is a pervasive feature in natural language text, yet it is an open question how sentiment is represented within Large Language Models (LLMs). In this study, we reveal that across a range of models, sentiment is represented linearly: a single direction in activation space mostly captures the feature across a range of tasks with one extreme for positive and the other for negative. Through causal interventions, we isolate this direction and show it is causally relevant in both toy tasks and real world datasets such as Stanford Sentiment Treebank. Through this case study we model a thorough investigation of what a single direction means on a broad data distribution. We further uncover the mechanisms that involve this direction, highlighting the roles of a small subset of attention heads and neurons. Finally, we discover a phenomenon which we term the summarization motif: sentiment is not solely represented on emotionally charged words, but is additionally summarized at intermediate positions without inherent sentiment, such as punctuation and names. We show that in Stanford Sentiment Treebank zero-shot classification, 76% of above-chance classification accuracy is lost when ablating the sentiment direction, nearly half of which (36%) is due to ablating the summarized sentiment direction exclusively at comma positions."
                },
                {
                    "title": "Identifying and Adapting Transformer-Components Responsible for Gender Bias in an English Language Model",
                    "abstract": "Language models (LMs) exhibit and amplify many types of undesirable biases learned from the training data, including gender bias. However, we lack tools for effectively and efficiently changing this behavior without hurting general language modeling performance. In this paper, we study three methods for identifying causal relations between LM components and particular output: causal mediation analysis, automated circuit discovery and our novel, efficient method called DiffMask+ based on differential masking. We apply the methods to GPT-2 small and the problem of gender bias, and use the discovered sets of components to perform parameter-efficient fine-tuning for bias mitigation. Our results show significant overlap in the identified components (despite huge differences in the computational requirements of the methods) as well as success in mitigating gender bias, with less damage to general language modeling compared to full model fine-tuning. However, our work also underscores the difficulty of defining and measuring bias, and the sensitivity of causal discovery procedures to dataset choice. We hope our work can contribute to more attention for dataset development, and lead to more effective mitigation strategies for other types of bias."
                },
                {
                    "title": "VeRA: Vector-based Random Matrix Adaptation",
                    "abstract": "Low-rank adapation (LoRA) is a popular method that reduces the number of trainable parameters when finetuning large language models, but still faces acute storage challenges when scaling to even larger models or deploying numerous per-user or per-task adapted models. In this work, we present Vector-based Random Matrix Adaptation (VeRA), which significantly reduces the number of trainable parameters compared to LoRA, yet maintains the same performance. It achieves this by using a single pair of low-rank matrices shared across all layers and learning small scaling vectors instead. We demonstrate its effectiveness on the GLUE and E2E benchmarks, image classification tasks, and show its application in instruction-tuning of 7B and 13B language models."
                },
                {
                    "title": "Representation Engineering: A Top-Down Approach to AI Transparency",
                    "abstract": "In this paper, we identify and characterize the emerging area of representation engineering (RepE), an approach to enhancing the transparency of AI systems that draws on insights from cognitive neuroscience. RepE places population-level representations, rather than neurons or circuits, at the center of analysis, equipping us with novel methods for monitoring and manipulating high-level cognitive phenomena in deep neural networks (DNNs). We provide baselines and an initial analysis of RepE techniques, showing that they offer simple yet effective solutions for improving our understanding and control of large language models. We showcase how these methods can provide traction on a wide range of safety-relevant problems, including honesty, harmlessness, power-seeking, and more, demonstrating the promise of top-down transparency research. We hope that this work catalyzes further exploration of RepE and fosters advancements in the transparency and safety of AI systems."
                },
                {
                    "title": "Rigorously Assessing Natural Language Explanations of Neurons",
                    "abstract": "Natural language is an appealing medium for explaining how large language models process and store information, but evaluating the faithfulness of such explanations is challenging. To help address this, we develop two modes of evaluation for natural language explanations that claim individual neurons represent a concept in a text input. In the *observational mode*, we evaluate claims that a neuron a activates on all and only input strings that refer to a concept picked out by the proposed explanation E. In the *intervention mode*, we construe E as a claim that neuron a is a causal mediator of the concept denoted by E. We apply our framework to the GPT-4-generated explanations of GPT-2 XL neurons of Bills et al. (2023) and show that even the most confident explanations have high error rates and little to no causal efficacy. We close the paper by critically assessing whether natural language is a good choice for explanations and whether neurons are the best level of analysis."
                },
                {
                    "title": "Sudden Drops in the Loss: Syntax Acquisition, Phase Transitions, and Simplicity Bias in MLMs",
                    "abstract": "Most interpretability research in NLP focuses on understanding the behavior and features of a fully trained model. However, certain insights into model behavior may only be accessible by observing the trajectory of the training process. We present a case study of syntax acquisition in masked language models (MLMs) that demonstrates how analyzing the evolution of interpretable artifacts throughout training deepens our understanding of emergent behavior. In particular, we study Syntactic Attention Structure (SAS), a naturally emerging property of MLMs wherein specific Transformer heads tend to focus on specific syntactic relations. We identify a brief window in pretraining when models abruptly acquire SAS, concurrent with a steep drop in loss. This breakthrough precipitates the subsequent acquisition of linguistic capabilities. We then examine the causal role of SAS by manipulating SAS during training, and demonstrate that SAS is necessary for the development of grammatical capabilities. We further find that SAS competes with other beneficial traits during training, and that briefly suppressing SAS improves model quality. These findings offer an interpretation of a real-world example of both simplicity bias and breakthrough training dynamics."
                },
                {
                    "title": "Emergent Linear Representations in World Models of Self-Supervised Sequence Models",
                    "abstract": "How do sequence models represent their decision-making process? Prior work suggests that Othello-playing neural network learned nonlinear models of the board state (Li et al., 2023a). In this work, we provide evidence of a closely related linear representation of the board. In particular, we show that probing for \u201cmy colour\u201d vs. \u201copponent\u2019s colour\u201d may be a simple yet powerful way to interpret the model\u2019s internal state. This precise understanding of the internal representations allows us to control the model\u2019s behaviour with simple vector arithmetic. Linear representations enable significant interpretability progress, which we demonstrate with further exploration of how the world model is computed."
                },
                {
                    "title": "IncreLoRA: Incremental Parameter Allocation Method for Parameter-Efficient Fine-tuning",
                    "abstract": "With the increasing size of pre-trained language models (PLMs), fine-tuning all the parameters in the model is not efficient, especially when there are a large number of downstream tasks, which incur significant training and storage costs. Many parameter-efficient fine-tuning (PEFT) approaches have been proposed, among which, Low-Rank Adaptation (LoRA) is a representative approach that injects trainable rank decomposition matrices into every target module. Yet LoRA ignores the importance of parameters in different modules. To address this problem, many works have been proposed to prune the parameters of LoRA. However, under limited training conditions, the upper bound of the rank of the pruned parameter matrix is still affected by the preset values. We, therefore, propose IncreLoRA, an incremental parameter allocation method that adaptively adds trainable parameters during training based on the importance scores of each module. This approach is different from the pruning method as it is not limited by the initial number of training parameters, and each parameter matrix has a higher rank upper bound for the same training overhead. We conduct extensive experiments on GLUE to demonstrate the effectiveness of IncreLoRA. The results show that our method owns higher parameter efficiency, especially when under the low-resource settings where our method significantly outperforms the baselines. Our code is publicly available."
                },
                {
                    "title": "Steering Language Models With Activation Engineering",
                    "abstract": "Prompt engineering and finetuning aim to maximize language model performance on a given metric (like toxicity reduction). However, these methods do not fully elicit a model's capabilities. To reduce this gap, we introduce activation engineering: the inference-time modification of activations in order to control (or steer) model outputs. Specifically, we introduce the Activation Addition (ActAdd) technique, which contrasts the intermediate activations on prompt pairs (such as\"Love\"versus\"Hate\") to compute a steering vector (Subramani et al. 2022). By tactically adding in e.g. the\"Love\"-\"Hate\"steering vector during the forward pass, we achieve SOTA on negative-to-positive sentiment shift and detoxification using models including LLaMA-3 and OPT. ActAdd yields inference-time control over high-level output properties (like topic and sentiment) while preserving performance on off-target tasks. ActAdd is lightweight: it does not require any machine optimization and works with a single pair of data points, which enables rapid iteration over steering. ActAdd demonstrates the power of activation engineering."
                },
                {
                    "title": "A Geometric Notion of Causal Probing",
                    "abstract": "The linear subspace hypothesis (Bolukbasi et al., 2016) states that, in a language model's representation space, all information about a concept such as verbal number is encoded in a linear subspace. Prior work has relied on auxiliary classification tasks to identify and evaluate candidate subspaces that might give support for this hypothesis. We instead give a set of intrinsic criteria which characterize an ideal linear concept subspace and enable us to identify the subspace using only the language model distribution. Our information-theoretic framework accounts for spuriously correlated features in the representation space (Kumar et al., 2022). As a byproduct of this analysis, we hypothesize a causal process for how a language model might leverage concepts during generation. Empirically, we find that LEACE (Belrose et al., 2023) returns a one-dimensional subspace containing roughly half of total concept information under our framework for verbal-number. Our causal intervention for controlled generation shows that, for at least one concept, the subspace returned by LEACE can be used to manipulate the concept value of the generated word with precision."
                },
                {
                    "title": "LoraHub: Efficient Cross-Task Generalization via Dynamic LoRA Composition",
                    "abstract": "Low-rank adaptations (LoRA) are often employed to fine-tune large language models (LLMs) for new tasks. This paper investigates LoRA composability for cross-task generalization and introduces LoraHub, a simple framework devised for the purposive assembly of LoRA modules trained on diverse given tasks, with the objective of achieving adaptable performance on unseen tasks. With just a few examples from a new task, LoraHub can fluidly combine multiple LoRA modules, eliminating the need for human expertise and assumptions. Notably, the composition requires neither additional model parameters nor gradients. Empirical results on the Big-Bench Hard benchmark suggest that LoraHub, while not surpassing the performance of in-context learning, offers a notable performance-efficiency trade-off in few-shot scenarios by employing a significantly reduced number of tokens per example during inference. Notably, LoraHub establishes a better upper bound compared to in-context learning when paired with different demonstration examples, demonstrating its potential for future development. Our vision is to establish a platform for LoRA modules, empowering users to share their trained LoRA modules. This collaborative approach facilitates the seamless application of LoRA modules to novel tasks, contributing to an adaptive ecosystem. Our code is available at https://github.com/sail-sg/lorahub, and all the pre-trained LoRA modules are released at https://huggingface.co/lorahub."
                },
                {
                    "title": "Composing Parameter-Efficient Modules with Arithmetic Operations",
                    "abstract": "As an efficient alternative to conventional full finetuning, parameter-efficient finetuning (PEFT) is becoming the prevailing method to adapt pretrained language models. In PEFT, a lightweight module is learned on each dataset while the underlying pretrained language model remains unchanged, resulting in multiple compact modules representing diverse skills when applied to various domains and tasks. In this paper, we propose to compose these parameter-efficient modules through linear arithmetic operations in the weight space, thereby integrating different module capabilities. Specifically, we first define addition and negation operators for the module, and then further compose these two basic operators to perform flexible arithmetic. Our approach requires \\emph{no additional training} and enables highly flexible module composition. We apply different arithmetic operations to compose the parameter-efficient modules for (1) distribution generalization, (2) multi-tasking, (3) unlearning, and (4) domain transfer. Additionally, we extend our approach to detoxify Alpaca-LoRA, the latest instruction-tuned large language model based on LLaMA. Empirical results demonstrate that our approach produces new and effective parameter-efficient modules that significantly outperform existing ones across all settings."
                },
                {
                    "title": "Give us the Facts: Enhancing Large Language Models With Knowledge Graphs for Fact-Aware Language Modeling",
                    "abstract": "Recently, ChatGPT, a representative large language model (LLM), has gained considerable attention. Due to their powerful emergent abilities, recent LLMs are considered as a possible alternative to structured knowledge bases like knowledge graphs (KGs). However, while LLMs are proficient at learning probabilistic language patterns and engaging in conversations with humans, they, like previous smaller pre-trained language models (PLMs), still have difficulty in recalling facts while generating knowledge-grounded contents. To overcome these limitations, researchers have proposed enhancing data-driven PLMs with knowledge-based KGs to incorporate explicit factual knowledge into PLMs, thus improving their performance in generating texts requiring factual knowledge and providing more informed responses to user queries. This paper reviews the studies on enhancing PLMs with KGs, detailing existing knowledge graph enhanced pre-trained language models (KGPLMs) as well as their applications. Inspired by existing studies on KGPLM, this paper proposes enhancing LLMs with KGs by developing knowledge graph-enhanced large language models (KGLLMs). KGLLM provides a solution to enhance LLMs\u2019 factual reasoning ability, opening up new avenues for LLM research."
                },
                {
                    "title": "Inference-Time Intervention: Eliciting Truthful Answers from a Language Model",
                    "abstract": "We introduce Inference-Time Intervention (ITI), a technique designed to enhance the\"truthfulness\"of large language models (LLMs). ITI operates by shifting model activations during inference, following a set of directions across a limited number of attention heads. This intervention significantly improves the performance of LLaMA models on the TruthfulQA benchmark. On an instruction-finetuned LLaMA called Alpaca, ITI improves its truthfulness from 32.5% to 65.1%. We identify a tradeoff between truthfulness and helpfulness and demonstrate how to balance it by tuning the intervention strength. ITI is minimally invasive and computationally inexpensive. Moreover, the technique is data efficient: while approaches like RLHF require extensive annotations, ITI locates truthful directions using only few hundred examples. Our findings suggest that LLMs may have an internal representation of the likelihood of something being true, even as they produce falsehoods on the surface."
                },
                {
                    "title": "Causal interventions expose implicit situation models for commonsense language understanding",
                    "abstract": "Accounts of human language processing have long appealed to implicit ``situation models'' that enrich comprehension with relevant but unstated world knowledge. Here, we apply causal intervention techniques to recent transformer models to analyze performance on the Winograd Schema Challenge (WSC), where a single context cue shifts interpretation of an ambiguous pronoun. We identify a relatively small circuit of attention heads that are responsible for propagating information from the context word that guides which of the candidate noun phrases the pronoun ultimately attends to. We then compare how this circuit behaves in a closely matched ``syntactic'' control where the situation model is not strictly necessary. These analyses suggest distinct pathways through which implicit situation models are constructed to guide pronoun resolution."
                },
                {
                    "title": "LEACE: Perfect linear concept erasure in closed form",
                    "abstract": "Concept erasure aims to remove specified features from a representation. It can improve fairness (e.g. preventing a classifier from using gender or race) and interpretability (e.g. removing a concept to observe changes in model behavior). We introduce LEAst-squares Concept Erasure (LEACE), a closed-form method which provably prevents all linear classifiers from detecting a concept while changing the representation as little as possible, as measured by a broad class of norms. We apply LEACE to large language models with a novel procedure called\"concept scrubbing,\"which erases target concept information from every layer in the network. We demonstrate our method on two tasks: measuring the reliance of language models on part-of-speech information, and reducing gender bias in BERT embeddings. Code is available at https://github.com/EleutherAI/concept-erasure."
                },
                {
                    "title": "QLoRA: Efficient Finetuning of Quantized LLMs",
                    "abstract": "We present QLoRA, an efficient finetuning approach that reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU while preserving full 16-bit finetuning task performance. QLoRA backpropagates gradients through a frozen, 4-bit quantized pretrained language model into Low Rank Adapters~(LoRA). Our best model family, which we name Guanaco, outperforms all previous openly released models on the Vicuna benchmark, reaching 99.3% of the performance level of ChatGPT while only requiring 24 hours of finetuning on a single GPU. QLoRA introduces a number of innovations to save memory without sacrificing performance: (a) 4-bit NormalFloat (NF4), a new data type that is information theoretically optimal for normally distributed weights (b) double quantization to reduce the average memory footprint by quantizing the quantization constants, and (c) paged optimziers to manage memory spikes. We use QLoRA to finetune more than 1,000 models, providing a detailed analysis of instruction following and chatbot performance across 8 instruction datasets, multiple model types (LLaMA, T5), and model scales that would be infeasible to run with regular finetuning (e.g. 33B and 65B parameter models). Our results show that QLoRA finetuning on a small high-quality dataset leads to state-of-the-art results, even when using smaller models than the previous SoTA. We provide a detailed analysis of chatbot performance based on both human and GPT-4 evaluations showing that GPT-4 evaluations are a cheap and reasonable alternative to human evaluation. Furthermore, we find that current chatbot benchmarks are not trustworthy to accurately evaluate the performance levels of chatbots. A lemon-picked analysis demonstrates where Guanaco fails compared to ChatGPT. We release all of our models and code, including CUDA kernels for 4-bit training."
                },
                {
                    "title": "Enhancing Chat Language Models by Scaling High-quality Instructional Conversations",
                    "abstract": "Fine-tuning on instruction data has been widely validated as an effective practice for implementing chat language models like ChatGPT. Scaling the diversity and quality of such data, although straightforward, stands a great chance of leading to improved performance. This paper aims to improve the upper bound of open-source models further. We first provide a systematically designed, diverse, informative, large-scale dataset of instructional conversations, UltraChat, which does not involve human queries. Our objective is to capture the breadth of interactions that a human might have with an AI assistant and employs a comprehensive framework to generate multi-turn conversation iteratively. UltraChat contains 1.5 million high-quality multi-turn dialogues and covers a wide range of topics and instructions. Our statistical analysis of UltraChat reveals its superiority in various key metrics, including scale, average length, diversity, coherence, etc., solidifying its position as a leading open-source dataset. Building upon UltraChat, we fine-tune a LLaMA model to create a powerful conversational model, UltraLLaMA. Our evaluations indicate that UltraLLaMA consistently outperforms other open-source models, including Vicuna, the previously recognized state-of-the-art open-source model. The dataset and the model will be publicly released\\footnote{\\url{https://github.com/thunlp/UltraChat}}."
                },
                {
                    "title": "Interpretability at Scale: Identifying Causal Mechanisms in Alpaca",
                    "abstract": "Obtaining human-interpretable explanations of large, general-purpose language models is an urgent goal for AI safety. However, it is just as important that our interpretability methods are faithful to the causal dynamics underlying model behavior and able to robustly generalize to unseen inputs. Distributed Alignment Search (DAS) is a powerful gradient descent method grounded in a theory of causal abstraction that has uncovered perfect alignments between interpretable symbolic algorithms and small deep learning models fine-tuned for specific tasks. In the present paper, we scale DAS significantly by replacing the remaining brute-force search steps with learned parameters -- an approach we call Boundless DAS. This enables us to efficiently search for interpretable causal structure in large language models while they follow instructions. We apply Boundless DAS to the Alpaca model (7B parameters), which, off the shelf, solves a simple numerical reasoning problem. With Boundless DAS, we discover that Alpaca does this by implementing a causal model with two interpretable boolean variables. Furthermore, we find that the alignment of neural representations with these variables is robust to changes in inputs and instructions. These findings mark a first step toward faithfully understanding the inner-workings of our ever-growing and most widely deployed language models. Our tool is extensible to larger LLMs and is released publicly at `https://github.com/stanfordnlp/pyvene`."
                },
                {
                    "title": "LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Model",
                    "abstract": "How to efficiently transform large language models (LLMs) into instruction followers is recently a popular research direction, while training LLM for multi-modal reasoning remains less explored. Although the recent LLaMA-Adapter demonstrates the potential to handle visual inputs with LLMs, it still cannot generalize well to open-ended visual instructions and lags behind GPT-4. In this paper, we present LLaMA-Adapter V2, a parameter-efficient visual instruction model. Specifically, we first augment LLaMA-Adapter by unlocking more learnable parameters (e.g., norm, bias and scale), which distribute the instruction-following ability across the entire LLaMA model besides adapters. Secondly, we propose an early fusion strategy to feed visual tokens only into the early LLM layers, contributing to better visual knowledge incorporation. Thirdly, a joint training paradigm of image-text pairs and instruction-following data is introduced by optimizing disjoint groups of learnable parameters. This strategy effectively alleviates the interference between the two tasks of image-text alignment and instruction following and achieves strong multi-modal reasoning with only a small-scale image-text and instruction dataset. During inference, we incorporate additional expert models (e.g. captioning/OCR systems) into LLaMA-Adapter to further enhance its image understanding capability without incurring training costs. Compared to the original LLaMA-Adapter, our LLaMA-Adapter V2 can perform open-ended multi-modal instructions by merely introducing 14M parameters over LLaMA. The newly designed framework also exhibits stronger language-only instruction-following capabilities and even excels in chat interactions. Our code and models are available at https://github.com/ZrrSkywalker/LLaMA-Adapter."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "LLM-Adapters: An Adapter Family for Parameter-Efficient Fine-Tuning of Large Language Models",
                    "abstract": "The success of large language models (LLMs), like GPT-4 and ChatGPT, has led to the development of numerous cost-effective and accessible alternatives that are created by finetuning open-access LLMs with task-specific data (e.g., ChatDoctor) or instruction data (e.g., Alpaca). Among the various fine-tuning methods, adapter-based parameter-efficient fine-tuning (PEFT) is undoubtedly one of the most attractive topics, as it only requires fine-tuning a few external parameters instead of the entire LLMs while achieving comparable or even better performance. To enable further research on PEFT methods of LLMs, this paper presents LLM-Adapters, an easy-to-use framework that integrates various adapters into LLMs and can execute these adapter-based PEFT methods of LLMs for different tasks. The framework includes state-of-the-art open-access LLMs such as LLaMA, BLOOM, and GPT-J, as well as widely used adapters such as Series adapters, Parallel adapter, Prompt-based learning and Reparametrization-based methods. Moreover, we conduct extensive empirical studies on the impact of adapter types, placement locations, and hyper-parameters to the best design for each adapter-based methods. We evaluate the effectiveness of the adapters on fourteen datasets from two different reasoning tasks, Arithmetic Reasoning and Commonsense Reasoning. The results demonstrate that using adapter-based PEFT in smaller-scale LLMs (7B) with few extra trainable parameters yields comparable, and in some cases superior, performance to powerful LLMs (175B) in zero-shot inference on both reasoning tasks."
                },
                {
                    "title": "Finding Alignments Between Interpretable Causal Variables and Distributed Neural Representations",
                    "abstract": "Causal abstraction is a promising theoretical framework for explainable artificial intelligence that defines when an interpretable high-level causal model is a faithful simplification of a low-level deep learning system. However, existing causal abstraction methods have two major limitations: they require a brute-force search over alignments between the high-level model and the low-level one, and they presuppose that variables in the high-level model will align with disjoint sets of neurons in the low-level one. In this paper, we present distributed alignment search (DAS), which overcomes these limitations. In DAS, we find the alignment between high-level and low-level models using gradient descent rather than conducting a brute-force search, and we allow individual neurons to play multiple distinct roles by analyzing representations in non-standard bases-distributed representations. Our experiments show that DAS can discover internal structure that prior approaches miss. Overall, DAS removes previous obstacles to conducting causal abstraction analyses and allows us to find conceptual structure in trained neural nets."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretability",
                    "abstract": "Causal abstraction provides a theoretical foundation for mechanistic interpretability, the field concerned with providing intelligible algorithms that are faithful simplifications of the known, but opaque low-level details of black box AI models. Our contributions are (1) generalizing the theory of causal abstraction from mechanism replacement (i.e., hard and soft interventions) to arbitrary mechanism transformation (i.e., functionals from old mechanisms to new mechanisms), (2) providing a flexible, yet precise formalization for the core concepts of modular features, polysemantic neurons, and graded faithfulness, and (3) unifying a variety of mechanistic interpretability methodologies in the common language of causal abstraction, namely activation and path patching, causal mediation analysis, causal scrubbing, causal tracing, circuit analysis, concept erasure, sparse autoencoders, differential binary masking, distributed alignment search, and activation steering."
                },
                {
                    "title": "Inducing Character-level Structure in Subword-based Language Models with Type-level Interchange Intervention Training",
                    "abstract": "Language tasks involving character-level manipulations (e.g., spelling corrections, arithmetic operations, word games) are challenging for models operating on subword units. To address this, we develop a causal intervention framework to learn robust and interpretable character representations inside subword-based language models. Our method treats each character as a typed variable in a causal model and learns such causal structures by adapting the interchange intervention training method of Geiger et al. (2021). We additionally introduce a suite of character-level tasks that systematically vary in their dependence on meaning and sequence-level context. While character-level models still perform best on purely form-based tasks like string reversal, our method outperforms character-level models on more complex tasks that blend form, meaning, and context, such as spelling correction in context and word search games. Compared with standard subword-based models, our approach also significantly improves robustness on unseen token sequences and leads to human-interpretable internal representations of characters."
                },
                {
                    "title": "Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small",
                    "abstract": "Research in mechanistic interpretability seeks to explain behaviors of machine learning models in terms of their internal components. However, most previous work either focuses on simple behaviors in small models, or describes complicated behaviors in larger models with broad strokes. In this work, we bridge this gap by presenting an explanation for how GPT-2 small performs a natural language task called indirect object identification (IOI). Our explanation encompasses 26 attention heads grouped into 7 main classes, which we discovered using a combination of interpretability approaches relying on causal interventions. To our knowledge, this investigation is the largest end-to-end attempt at reverse-engineering a natural behavior\"in the wild\"in a language model. We evaluate the reliability of our explanation using three quantitative criteria--faithfulness, completeness and minimality. Though these criteria support our explanation, they also point to remaining gaps in our understanding. Our work provides evidence that a mechanistic understanding of large ML models is feasible, opening opportunities to scale our understanding to both larger models and more complex tasks."
                },
                {
                    "title": "SparseAdapter: An Easy Approach for Improving the Parameter-Efficiency of Adapters",
                    "abstract": "Adapter Tuning, which freezes the pretrained language models (PLMs) and only fine-tunes a few extra modules, becomes an appealing efficient alternative to the full model fine-tuning. Although computationally efficient, the recent Adapters often increase parameters (e.g. bottleneck dimension) for matching the performance of full model fine-tuning, which we argue goes against their original intention. In this work, we re-examine the parameter-efficiency of Adapters through the lens of network pruning (we name such plug-in concept as \\texttt{SparseAdapter}) and find that SparseAdapter can achieve comparable or better performance than standard Adapters when the sparse ratio reaches up to 80\\%. Based on our findings, we introduce an easy but effective setting ``\\textit{Large-Sparse}'' to improve the model capacity of Adapters under the same parameter budget. Experiments on five competitive Adapters upon three advanced PLMs show that with proper sparse method (e.g. SNIP) and ratio (e.g. 40\\%) SparseAdapter can consistently outperform their corresponding counterpart. Encouragingly, with the \\textit{Large-Sparse} setting, we can obtain further appealing gains, even outperforming the full fine-tuning by a large margin. Our code will be released at: https://github.com/Shwai-He/SparseAdapter."
                },
                {
                    "title": "Toy Models of Superposition",
                    "abstract": "Neural networks often pack many unrelated concepts into a single neuron \u2013 a puzzling phenomenon known as 'polysemanticity' which makes interpretability much more challenging. This paper provides a toy model where polysemanticity can be fully understood, arising as a result of models storing additional sparse features in \"superposition.\" We demonstrate the existence of a phase change, a surprising connection to the geometry of uniform polytopes, and evidence of a link to adversarial examples. We also discuss potential implications for mechanistic interpretability. We recommend reading this paper as an HTML article. It would be very convenient if the individual neurons of artificial neural networks corresponded to cleanly interpretable features of the input. For example, in an \u201cideal\u201d ImageNet classifier, each neuron would fire only in the presence of a specific visual feature, such as the color red, a leftfacing curve, or a dog snout. Empirically, in models we have studied, some of the neurons do cleanly map to features. But it isn't always the case that features correspond so cleanly to neurons, especially in large language models where it actually seems rare for neurons to correspond to clean features. This brings up many questions. Why is it that neurons sometimes align with features and sometimes don't? Why do some models and tasks have many of these clean neurons, while they're vanishingly rare in others? In this paper, we use toy models \u2014 small ReLU networks trained on synthetic data with sparse input features \u2014 to investigate how and when models represent more features than they have dimensions. We call this phenomenon superposition. When features are sparse, superposition allows compression beyond what a linear model would do, at the cost of \"interference\" that requires nonlinear filtering. Consider a toy model where we train an embedding of five features of varying importance in two dimensions, add a ReLU afterwards for filtering, and vary the sparsity of the features. With dense features, the model learns to represent an orthogonal basis of the most important two features (similar to what Principal Component Analysis might give us), and the other three features are not represented. But if we make the features sparse, this changes: 1 This figure and a few others can be reproduced using the toy model framework Colab notebook in our Github repo Not only can models store additional features in superposition by tolerating some interference, but we'll show that, at least in certain limited cases, models can perform computation while in superposition. (In particular, we'll show that models can put simple circuits computing the absolute value function in superposition.) This leads us to hypothesize that the neural networks we observe in practice are in some sense noisily simulating larger, highly sparse networks. In other words, it's possible that models we train can be thought of as doing \u201cthe same thing as\u201d an imagined muchlarger model, representing the exact same features but with no interference. Feature superposition isn't a novel idea. A number of previous interpretability papers have speculated about it , and it's very closely related to the long-studied topic of compressed sensing in mathematics , as well as the ideas of distributed, dense, and population codes in neuroscience and deep learning . What, then, is the contribution of this paper? For interpretability researchers, our main contribution is providing a direct demonstration that superposition occurs in artificial neural networks given a relatively natural setup, suggesting this may also occur in practice. We offer a theory of when and why this occurs, revealing a phase diagram for superposition. We also discover that, at least in our toy model, superposition exhibits complex geometric structure. But our results may also be of broader interest. We find preliminary evidence that superposition may be linked to adversarial examples and grokking, and might also suggest a theory for the performance of mixture of experts models. More broadly, the toy model we investigate has unexpectedly rich structure, exhibiting phase changes, a geometric structure based on uniform polytopes, \"energy level\"-like jumps during training, and a phenomenon which is qualitatively similar to the fractional quantum Hall effect in physics. We originally investigated the subject to gain understanding of cleanly-interpretable neurons in larger models, but we've found these toy models to be surprisingly interesting in their own right. [1, 2] [3] [4] [5] KEY RESULTS FROM OUR TOY MODELS In our toy models, we are able to demonstrate that: Superposition is a real, observed phenomenon. Both monosemantic and polysemantic neurons can form. At least some kinds of computation can be performed in superposition. Whether features are stored in superposition is governed by a phase change. Superposition organizes features into geometric structures such as digons, triangles, pentagons, and tetrahedrons. Our toy models are simple ReLU networks, so it seems fair to say that neural networks exhibit these properties in at least some regimes, but it's very unclear what to generalize to real networks. Definitions and Motivation: Features, Directions, and Superposition In our work, we often think of neural networks as having features of the input represented as directions in activation space. This isn't a trivial claim. It isn't obvious what kind of structure we should expect neural network representations to have. When we say something like \"word embeddings have a gender direction\" or \"vision models have curve detector neurons\", one is implicitly making strong claims about the structure of network representations. Despite this, we believe this kind of \"linear representation hypothesis\" is supported both by significant empirical findings and theoretical arguments. One might think of this as two separate properties, which we'll explore in more detail shortly: Decomposability: Network representations can be described in terms of independently understandable features. Linearity: Features are represented by direction. If we hope to reverse engineer neural networks, we need a property like decomposability. Decomposability is what allows us to reason about the model without fitting the whole thing in our heads! But it's not enough for things to be decomposable: we need to be able to access the decomposition somehow. In order to do this, we need to identify the individual features within a representation. In a linear representation, this corresponds to determining which directions in activation space correspond to which independent features of the input. Sometimes, identifying feature directions is very easy because features seem to correspond to neurons. For example, many neurons in the early layers of InceptionV1 clearly correspond to features (e.g. curve detector neurons ). Why is it that we sometimes get this extremely helpful property, but in other cases don't? We hypothesize that there are really two countervailing forces driving this: [6] Privileged Basis: Only some representations have a privileged basis which encourages features to align with basis directions (i.e. to correspond to neurons). Superposition: Linear representations can represent more features than dimensions, using a strategy we call superposition. This can be seen as neural networks simulating larger networks. This pushes features away from corresponding to neurons. Superposition has been hypothesized in previous work  . However, we're not aware of feature superposition having been unambiguously demonstrated to occur in neural networks before ( demonstrates a closely related phenomenon of model superposition). The goal of this paper is to change that, demonstrating superposition and exploring how it interacts with privileged bases. If superposition occurs in networks, it deeply influences what approaches to interpretability research make sense, so unambiguous demonstration seems important. [1, 2] [7] The goal of this section will be to motivate these ideas and unpack them in detail. It's worth noting that many of the ideas in this section have close connections to ideas in other lines of interpretability research (especially disentanglement), neuroscience (distributed representations, population codes, etc), compressed sensing, and many other lines of work. This section will focus on articulating our perspective on the problem. We'll discuss these other lines of work in detail in Related Work."
                },
                {
                    "title": "Branch-Train-Merge: Embarrassingly Parallel Training of Expert Language Models",
                    "abstract": "We present Branch-Train-Merge (BTM), a communication-efficient algorithm for embarrassingly parallel training of large language models (LLMs). We show it is possible to independently train subparts of a new class of LLMs on different subsets of the data, eliminating the massive multi-node synchronization currently required to train LLMs. BTM learns a set of independent expert LMs (ELMs), each specialized to a different textual domain, such as scientific or legal text. These ELMs can be added and removed to update data coverage, ensembled to generalize to new domains, or averaged to collapse back to a single LM for efficient inference. New ELMs are learned by branching from (mixtures of) ELMs in the current set, further training the parameters on data for the new domain, and then merging the resulting model back into the set for future use. Experiments show that BTM improves in- and out-of-domain perplexities as compared to GPT-style Transformer LMs, when controlling for training cost. Through extensive analysis, we show that these results are robust to different ELM initialization schemes, but require expert domain specialization; LM ensembles with random data splits do not perform well. We also present a study of scaling BTM into a new corpus of 64 domains (192B whitespace-separated tokens in total); the resulting LM (22.4B total parameters) performs as well as a Transformer LM trained with 2.5 times more compute. These gains grow with the number of domains, suggesting more aggressive parallelism could be used to efficiently train larger models in future work."
                },
                {
                    "title": "AdaMix: Mixture-of-Adaptations for Parameter-efficient Model Tuning",
                    "abstract": "Standard fine-tuning of large pre-trained language models (PLMs) for downstream tasks requires updating hundreds of millions to billions of parameters, and storing a large copy of the PLM weights for every task resulting in increased cost for storing, sharing and serving the models. To address this, parameter-efficient fine-tuning (PEFT) techniques were introduced where small trainable components are injected in the PLM and updated during fine-tuning. We propose AdaMix as a general PEFT method that tunes a mixture of adaptation modules \u2013 given the underlying PEFT method of choice \u2013 introduced in each Transformer layer while keeping most of the PLM weights frozen. For instance, AdaMix can leverage a mixture of adapters like Houlsby or a mixture of low rank decomposition matrices like LoRA to improve downstream task performance over the corresponding PEFT methods for fully supervised and few-shot NLU and NLG tasks. Further, we design AdaMix such that it matches the same computational cost and the number of tunable parameters as the underlying PEFT method. By only tuning 0.1-0.2% of PLM parameters, we show that AdaMix outperforms SOTA parameter-efficient fine-tuning and full model fine-tuning for both NLU and NLG tasks."
                },
                {
                    "title": "Naturalistic Causal Probing for Morpho-Syntax",
                    "abstract": "Probing has become a go-to methodology for interpreting and analyzing deep neural models in natural language processing. However, there is still a lack of understanding of the limitations and weaknesses of various types of probes. In this work, we suggest a strategy for input-level intervention on naturalistic sentences. Using our approach, we intervene on the morpho-syntactic features of a sentence, while keeping the rest of the sentence unchanged. Such an intervention allows us to causally probe pre-trained models. We apply our naturalistic causal probing framework to analyze the effects of grammatical gender and number on contextualized representations extracted from three pre-trained models in Spanish, the multilingual versions of BERT, RoBERTa, and GPT-2. Our experiments suggest that naturalistic interventions lead to stable estimates of the causal effects of various linguistic properties. Moreover, our experiments demonstrate the importance of naturalistic causal probing when analyzing pre-trained models. https://github.com/rycolab/naturalistic-causal-probing"
                },
                {
                    "title": "Extracting Latent Steering Vectors from Pretrained Language Models",
                    "abstract": "Prior work on controllable text generation has focused on learning how to control language models through trainable decoding, smart-prompt design, or fine-tuning based on a desired objective. We hypothesize that the information needed to steer the model to generate a target sentence is already encoded within the model. Accordingly, we explore a different approach altogether: extracting latent vectors directly from pretrained language model decoders without fine-tuning. Experiments show that there exist steering vectors, which, when added to the hidden states of the language model, generate a target sentence nearly perfectly (> 99 BLEU) for English sentences from a variety of domains. We show that vector arithmetic can be used for unsupervised sentiment transfer on the Yelp sentiment benchmark, with performance comparable to models tailored to this task. We find that distances between steering vectors reflect sentence similarity when evaluated on a textual similarity benchmark (STS-B), outperforming pooled hidden states of models. Finally, we present an analysis of the intrinsic properties of the steering vectors. Taken together, our results suggest that frozen LMs can be effectively controlled through their latent steering space."
                },
                {
                    "title": "Probing for the Usage of Grammatical Number",
                    "abstract": "A central quest of probing is to uncover how pre-trained models encode a linguistic property within their representations. An encoding, however, might be spurious\u2014i.e., the model might not rely on it when making predictions. In this paper, we try to find an encoding that the model actually uses, introducing a usage-based probing setup. We first choose a behavioral task which cannot be solved without using the linguistic property. Then, we attempt to remove the property by intervening on the model\u2019s representations. We contend that, if an encoding is used by the model, its removal should harm the performance on the chosen behavioral task. As a case study, we focus on how BERT encodes grammatical number, and on how it uses this encoding to solve the number agreement task. Experimentally, we find that BERT relies on a linear encoding of grammatical number to produce the correct behavioral output. We also find that BERT uses a separate encoding of grammatical number for nouns and verbs. Finally, we identify in which layers information about grammatical number is transferred from a noun to its head verb."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Linear Adversarial Concept Erasure",
                    "abstract": "Modern neural models trained on textual data rely on pre-trained representations that emerge without direct supervision. As these representations are increasingly being used in real-world applications, the inability to \\emph{control} their content becomes an increasingly important problem. We formulate the problem of identifying and erasing a linear subspace that corresponds to a given concept, in order to prevent linear predictors from recovering the concept. We model this problem as a constrained, linear maximin game, and show that existing solutions are generally not optimal for this task. We derive a closed-form solution for certain objectives, and propose a convex relaxation, \\method, that works well for others. When evaluated in the context of binary gender removal, the method recovers a low-dimensional subspace whose removal mitigates bias by intrinsic and extrinsic evaluation. We show that the method is highly expressive, effectively mitigating bias in deep nonlinear classifiers while maintaining tractability and interpretability."
                },
                {
                    "title": "Causal Distillation for Language Models",
                    "abstract": "Distillation efforts have led to language models that are more compact and efficient without serious drops in performance. The standard approach to distillation trains a student model against two objectives: a task-specific objective (e.g., language modeling) and an imitation objective that encourages the hidden states of the student model to be similar to those of the larger teacher model. In this paper, we show that it is beneficial to augment distillation with a third objective that encourages the student to imitate the causal dynamics of the teacher through a distillation interchange intervention training objective (DIITO). DIITO pushes the student model to become a causal abstraction of the teacher model \u2013 a faithful model with simpler causal structure. DIITO is fully differentiable, easily implemented, and combines flexibly with other objectives. Compared against standard distillation with the same setting, DIITO results in lower perplexity on the WikiText-103M corpus (masked language modeling) and marked improvements on the GLUE benchmark (natural language understanding), SQuAD (question answering), and CoNLL-2003 (named entity recognition)."
                },
                {
                    "title": "Inducing Causal Structure for Interpretable Neural Networks",
                    "abstract": "In many areas, we have well-founded insights about causal structure that would be useful to bring into our trained models while still allowing them to learn in a data-driven fashion. To achieve this, we present the new method of interchange intervention training (IIT). In IIT, we (1) align variables in a causal model (e.g., a deterministic program or Bayesian network) with representations in a neural model and (2) train the neural model to match the counterfactual behavior of the causal model on a base input when aligned representations in both models are set to be the value they would be for a source input. IIT is fully differentiable, flexibly combines with other objectives, and guarantees that the target causal model is a causal abstraction of the neural model when its loss is zero. We evaluate IIT on a structural vision task (MNIST-PVR), a navigational language task (ReaSCAN), and a natural language inference task (MQNLI). We compare IIT against multi-task training objectives and data augmentation. In all our experiments, IIT achieves the best results and produces neural models that are more interpretable in the sense that they more successfully realize the target causal model."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "Towards a Unified View of Parameter-Efficient Transfer Learning",
                    "abstract": "Fine-tuning large pre-trained language models on downstream tasks has become the de-facto learning paradigm in NLP. However, conventional approaches fine-tune all the parameters of the pre-trained model, which becomes prohibitive as the model size and the number of tasks grow. Recent work has proposed a variety of parameter-efficient transfer learning methods that only fine-tune a small number of (extra) parameters to attain strong performance. While effective, the critical ingredients for success and the connections among the various methods are poorly understood. In this paper, we break down the design of state-of-the-art parameter-efficient transfer learning methods and present a unified framework that establishes connections between them. Specifically, we re-frame them as modifications to specific hidden states in pre-trained models, and define a set of design dimensions along which different methods vary, such as the function to compute the modification and the position to apply the modification. Through comprehensive empirical studies across machine translation, text summarization, language understanding, and text classification benchmarks, we utilize the unified view to identify important design choices in previous methods. Furthermore, our unified framework enables the transfer of design elements across different approaches, and as a result we are able to instantiate new parameter-efficient fine-tuning methods that tune less parameters than previous methods while being more effective, achieving comparable results to fine-tuning all parameters on all four tasks."
                },
                {
                    "title": "BitFit: Simple Parameter-efficient Fine-tuning for Transformer-based Masked Language-models",
                    "abstract": "We introduce BitFit, a sparse-finetuning method where only the bias-terms of the model (or a subset of them) are being modified. We show that with small-to-medium training data, applying BitFit on pre-trained BERT models is competitive with (and sometimes better than) fine-tuning the entire model. For larger data, the method is competitive with other sparse fine-tuning methods.Besides their practical utility, these findings are relevant for the question of understanding the commonly-used process of finetuning: they support the hypothesis that finetuning is mainly about exposing knowledge induced by language-modeling training, rather than learning new task-specific linguistic knowledge."
                },
                {
                    "title": "Causal Abstractions of Neural Networks",
                    "abstract": "Structural analysis methods (e.g., probing and feature attribution) are increasingly important tools for neural network analysis. We propose a new structural analysis method grounded in a formal theory of causal abstraction that provides rich characterizations of model-internal representations and their roles in input/output behavior. In this method, neural representations are aligned with variables in interpretable causal models, and then interchange interventions are used to experimentally verify that the neural representations have the causal properties of their aligned variables. We apply this method in a case study to analyze neural models trained on Multiply Quantified Natural Language Inference (MQNLI) corpus, a highly complex NLI dataset that was constructed with a tree-structured natural logic causal model. We discover that a BERT-based model with state-of-the-art performance successfully realizes parts of the natural logic model's causal structure, whereas a simpler baseline model fails to show any such structure, demonstrating that BERT representations encode the compositional structure of MQNLI."
                },
                {
                    "title": "Are NLP Models really able to Solve Simple Math Word Problems?",
                    "abstract": "The problem of designing NLP solvers for math word problems (MWP) has seen sustained research activity and steady gains in the test accuracy. Since existing solvers achieve high performance on the benchmark datasets for elementary level MWPs containing one-unknown arithmetic word problems, such problems are often considered \u201csolved\u201d with the bulk of research attention moving to more complex MWPs. In this paper, we restrict our attention to English MWPs taught in grades four and lower. We provide strong evidence that the existing MWP solvers rely on shallow heuristics to achieve high performance on the benchmark datasets. To this end, we show that MWP solvers that do not have access to the question asked in the MWP can still solve a large fraction of MWPs. Similarly, models that treat MWPs as bag-of-words can also achieve surprisingly high accuracy. Further, we introduce a challenge dataset, SVAMP, created by applying carefully chosen variations over examples sampled from existing datasets. The best accuracy achieved by state-of-the-art models is substantially lower on SVAMP, thus showing that much remains to be done even for the simplest of the MWPs."
                },
                {
                    "title": "MAD-X: An Adapter-based Framework for Multi-task Cross-lingual Transfer",
                    "abstract": "The main goal behind state-of-the-art pretrained multilingual models such as multilingual BERT and XLM-R is enabling and bootstrapping NLP applications in low-resource languages through zero-shot or few-shot cross-lingual transfer. However, due to limited model capacity, their transfer performance is the weakest exactly on such low-resource languages and languages unseen during pretraining. We propose MAD-X, an adapter-based framework that enables high portability and parameter-efficient transfer to arbitrary tasks and languages by learning modular language and task representations. In addition, we introduce a novel invertible adapter architecture and a strong baseline method for adapting a pretrained multilingual model to a new language. MAD-X outperforms the state of the art in cross-lingual transfer across a representative set of typologically diverse languages on named entity recognition and achieves competitive results on question answering."
                },
                {
                    "title": "PIQA: Reasoning about Physical Commonsense in Natural Language",
                    "abstract": "To apply eyeshadow without a brush, should I use a cotton swab or a toothpick? Questions requiring this kind of physical commonsense pose a challenge to today's natural language understanding systems. While recent pretrained models (such as BERT) have made progress on question answering over more abstract domains \u2013 such as news articles and encyclopedia entries, where text is plentiful \u2013 in more physical domains, text is inherently limited due to reporting bias. Can AI systems learn to reliably answer physical commonsense questions without experiencing the physical world?In this paper, we introduce the task of physical commonsense reasoning and a corresponding benchmark dataset Physical Interaction: Question Answering or PIQA. Though humans find the dataset easy (95% accuracy), large pretrained models struggle (\u223c75%). We provide analysis about the dimensions of knowledge that existing models lack, which offers significant opportunities for future research."
                },
                {
                    "title": "RoBERTa: A Robustly Optimized BERT Pretraining Approach",
                    "abstract": "Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code."
                },
                {
                    "title": "Alice's Adventures in Wonderland",
                    "abstract": "Reading became an increasingly popular entertainment in eighteenthand nineteenth-century Britain, Europe and America, reaching an ever wider spectrum of society as the cost of printing came down and levels of literacy rose. The novels avidly consumed in this period were not merely escapist fiction. Many of them drew attention to serious social issues such as slavery, child labour and other forms of exploitation that blighted the age of revolutions and empire, some were thinly disguised autobiographies, while others had clear educational aims: thus the line between fiction and non-fiction was a fluid one. Poetry too flourished across a wide range of genres, and the political and social agendas of the Romantic movement in particular led to its being read and appreciated at all levels of society. In this series, the Cambridge Library Collection offers the texts of fiction and poetry as these works were first published and received by an eager reading public."
                },
                {
                    "title": "BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions",
                    "abstract": "In this paper we study yes/no questions that are naturally occurring \u2014 meaning that they are generated in unprompted and unconstrained settings. We build a reading comprehension dataset, BoolQ, of such questions, and show that they are unexpectedly challenging. They often query for complex, non-factoid information, and require difficult entailment-like inference to solve. We also explore the effectiveness of a range of transfer learning baselines. We find that transferring from entailment data is more effective than transferring from paraphrase or extractive QA data, and that it, surprisingly, continues to be very beneficial even when starting from massive pre-trained language models such as BERT. Our best method trains BERT on MultiNLI and then re-trains it on our train set. It achieves 80.4% accuracy compared to 90% accuracy of human annotators (and 62% majority-baseline), leaving a significant gap for future work."
                },
                {
                    "title": "HellaSwag: Can a Machine Really Finish Your Sentence?",
                    "abstract": "Recent work by Zellers et al. (2018) introduced a new task of commonsense natural language inference: given an event description such as \u201cA woman sits at a piano,\u201d a machine must select the most likely followup: \u201cShe sets her fingers on the keys.\u201d With the introduction of BERT, near human-level performance was reached. Does this mean that machines can perform human level commonsense inference? In this paper, we show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans (>95% accuracy), state-of-the-art models struggle (<48%). We achieve this via Adversarial Filtering (AF), a data collection paradigm wherein a series of discriminators iteratively select an adversarial set of machine-generated wrong answers. AF proves to be surprisingly robust. The key insight is to scale up the length and complexity of the dataset examples towards a critical \u2018Goldilocks\u2019 zone wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models. Our construction of HellaSwag, and its resulting difficulty, sheds light on the inner workings of deep pretrained models. More broadly, it suggests a new path forward for NLP research, in which benchmarks co-evolve with the evolving state-of-the-art in an adversarial way, so as to present ever-harder challenges."
                },
                {
                    "title": "Parameter-Efficient Transfer Learning for NLP",
                    "abstract": "Fine-tuning large pre-trained models is an effective transfer mechanism in NLP. However, in the presence of many downstream tasks, fine-tuning is parameter inefficient: an entire new model is required for every task. As an alternative, we propose transfer with adapter modules. Adapter modules yield a compact and extensible model; they add only a few trainable parameters per task, and new tasks can be added without revisiting previous ones. The parameters of the original network remain fixed, yielding a high degree of parameter sharing. To demonstrate adapter's effectiveness, we transfer the recently proposed BERT Transformer model to 26 diverse text classification tasks, including the GLUE benchmark. Adapters attain near state-of-the-art performance, whilst adding only a few parameters per task. On GLUE, we attain within 0.4% of the performance of full fine-tuning, adding only 3.6% parameters per task. By contrast, fine-tuning trains 100% of the parameters per task."
                },
                {
                    "title": "Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering",
                    "abstract": "We present a new kind of question answering dataset, OpenBookQA, modeled after open book exams for assessing human understanding of a subject. The open book that comes with our questions is a set of 1326 elementary level science facts. Roughly 6000 questions probe an understanding of these facts and their application to novel situations. This requires combining an open book fact (e.g., metals conduct electricity) with broad common knowledge (e.g., a suit of armor is made of metal) obtained from other sources. While existing QA datasets over documents or knowledge bases, being generally self-contained, focus on linguistic understanding, OpenBookQA probes a deeper understanding of both the topic\u2014in the context of common knowledge\u2014and the language it is expressed in. Human performance on OpenBookQA is close to 92%, but many state-of-the-art pre-trained QA methods perform surprisingly poorly, worse than several simple neural baselines we develop. Our oracle experiments designed to circumvent the knowledge retrieval bottleneck demonstrate the value of both the open book and additional facts. We leave it as a challenge to solve the retrieval problem in this multi-hop setting and to close the large gap to human performance."
                },
                {
                    "title": "GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding",
                    "abstract": "Human ability to understand language is general, flexible, and robust. In contrast, most NLU models above the word level are designed for a specific task and struggle with out-of-domain data. If we aspire to develop models with understanding beyond the detection of superficial correspondences between inputs and outputs, then it is critical to develop a unified model that can execute a range of linguistic tasks across different domains. To facilitate research in this direction, we present the General Language Understanding Evaluation (GLUE, gluebenchmark.com): a benchmark of nine diverse NLU tasks, an auxiliary dataset for probing models for understanding of specific linguistic phenomena, and an online platform for evaluating and comparing models. For some benchmark tasks, training data is plentiful, but for others it is limited or does not match the genre of the test set. GLUE thus favors models that can represent linguistic knowledge in a way that facilitates sample-efficient learning and effective knowledge-transfer across tasks. While none of the datasets in GLUE were created from scratch for the benchmark, four of them feature privately-held test data, which is used to ensure that the benchmark is used fairly. We evaluate baselines that use ELMo (Peters et al., 2018), a powerful transfer learning technique, as well as state-of-the-art sentence representation models. The best models still achieve fairly low absolute scores. Analysis with our diagnostic dataset yields similarly weak performance over all phenomena tested, with some exceptions."
                },
                {
                    "title": "Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge",
                    "abstract": "We present a new question set, text corpus, and baselines assembled to encourage AI research in advanced question answering. Together, these constitute the AI2 Reasoning Challenge (ARC), which requires far more powerful knowledge and reasoning than previous challenges such as SQuAD or SNLI. The ARC question set is partitioned into a Challenge Set and an Easy Set, where the Challenge Set contains only questions answered incorrectly by both a retrieval-based algorithm and a word co-occurence algorithm. The dataset contains only natural, grade-school science questions (authored for human tests), and is the largest public-domain set of this kind (7,787 questions). We test several baselines on the Challenge Set, including leading neural models from the SQuAD and SNLI tasks, and find that none are able to significantly outperform a random baseline, reflecting the difficult nature of this task. We are also releasing the ARC Corpus, a corpus of 14M science sentences relevant to the task, and implementations of the three neural baseline models tested. Can your model perform better? We pose ARC as a challenge to the community."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Program Induction by Rationale Generation: Learning to Solve and Explain Algebraic Word Problems",
                    "abstract": "Solving algebraic word problems requires executing a series of arithmetic operations\u2014a program\u2014to obtain a final answer. However, since programs can be arbitrarily complicated, inducing them directly from question-answer pairs is a formidable challenge. To make this task more feasible, we solve these problems by generating answer rationales, sequences of natural language and human-readable mathematical expressions that derive the final answer through a series of small steps. Although rationales do not explicitly specify programs, they provide a scaffolding for their structure via intermediate milestones. To evaluate our approach, we have created a new 100,000-sample dataset of questions, answers and rationales. Experimental results show that indirect supervision of program learning via answer rationales is a promising strategy for inducing arithmetic programs."
                },
                {
                    "title": "Solving General Arithmetic Word Problems",
                    "abstract": "This paper presents a novel approach to automatically solving arithmetic word problems. This is the first algorithmic approach that can handle arithmetic problems with multiple steps and operations, without depending on additional annotations or predefined templates. We develop a theory for expression trees that can be used to represent and evaluate the target arithmetic expressions; we use it to uniquely decompose the target arithmetic problem to multiple classification problems; we then compose an expression tree, combining these with world knowledge through a constrained inference framework. Our classifiers gain from the use of quantity schemas that supports better extraction of features. Experimental results show that our method outperforms existing systems, achieving state of the art performance on benchmark datasets of arithmetic word problems."
                },
                {
                    "title": "MAWPS: A Math Word Problem Repository",
                    "abstract": "Recent work across several AI subdisciplines has focused on automatically solving math word problems. In this paper we introduce MAWPS, an online repository of Math Word Problems, to provide a unified testbed to evaluate different algorithms. MAWPS allows for the automatic construction of datasets with particular characteristics, providing tools for tuning the lexical and template overlap of a dataset as well as for filtering ungrammatical problems from web-sourced corpora. The online nature of this repository facilitates easy community contribution. At present, we have amassed 3,320 problems, including the full datasets used in several prominent works."
                },
                {
                    "title": "Parsing Algebraic Word Problems into Equations",
                    "abstract": "This paper formalizes the problem of solving multi-sentence algebraic word problems as that of generating and scoring equation trees. We use integer linear programming to generate equation trees and score their likelihood by learning local and global discriminative models. These models are trained on a small set of word problems and their answers, without any manual annotation, in order to choose the equation that best matches the problem text. We refer to the overall system as Alges. We compare Alges with previous work and show that it covers the full gamut of arithmetic operations whereas Hosseini et al. (2014) only handle addition and subtraction. In addition, Alges overcomes the brittleness of the Kushman et al. (2014) approach on single-equation problems, yielding a 15% to 50% reduction in error."
                },
                {
                    "title": "Semi-supervised Sequence Learning",
                    "abstract": "We present two approaches to use unlabeled data to improve Sequence Learning with recurrent networks. The first approach is to predict what comes next in a sequence, which is a language model in NLP. The second approach is to use a sequence autoencoder, which reads the input sequence into a vector and predicts the input sequence again. These two algorithms can be used as a \"pretraining\" algorithm for a later supervised sequence learning algorithm. In other words, the parameters obtained from the pretraining step can then be used as a starting point for other supervised training models. In our experiments, we find that long short term memory recurrent networks after pretrained with the two approaches become more stable to train and generalize better. With pretraining, we were able to achieve strong performance in many classification tasks, such as text classification with IMDB, DBpedia or image recognition in CIFAR-10."
                },
                {
                    "title": "Learning to Solve Arithmetic Word Problems with Verb Categorization",
                    "abstract": "This paper presents a novel approach to learning to solve simple arithmetic word problems. Our system, ARIS, analyzes each of the sentences in the problem statement to identify the relevant variables and their values. ARIS then maps this information into an equation that represents the problem, and enables its (trivial) solution as shown in Figure 1. The paper analyzes the arithmetic-word problems \u201cgenre\u201d, identifying seven categories of verbs used in such problems. ARIS learns to categorize verbs with 81.2% accuracy, and is able to solve 77.7% of the problems in a corpus of standard primary school test questions. We report the first learning results on this task without reliance on predefined templates and make our data publicly available. 1"
                },
                {
                    "title": "Linguistic Regularities in Continuous Space Word Representations",
                    "abstract": "Continuous space language models have recently demonstrated outstanding results across a variety of tasks. In this paper, we examine the vector-space word representations that are implicitly learned by the input-layer weights. We find that these representations are surprisingly good at capturing syntactic and semantic regularities in language, and that each relationship is characterized by a relation-specific vector offset. This allows vector-oriented reasoning based on the offsets between words. For example, the male/female relationship is automatically learned, and with the induced vector representations, \u201cKing Man + Woman\u201d results in a vector very close to \u201cQueen.\u201d We demonstrate that the word vectors capture syntactic regularities by means of syntactic analogy questions (provided with this paper), and are able to correctly answer almost 40% of the questions. We demonstrate that the word vectors capture semantic regularities by using the vector offset method to answer SemEval-2012 Task 2 questions. Remarkably, this method outperforms the best previous systems."
                },
                {
                    "title": "The Winograd Schema Challenge",
                    "abstract": "In this paper, we present an alternative to the Turing Test that has some conceptual and practical advantages. Like the original, it involves responding to typed English sentences, and English-speaking adults will have no difficulty with it. Unlike the original, the subject is not required to engage in a conversation and fool an interrogator into believing she is dealing with a person. Moreover, the test is arranged in such a way that having full access to a large corpus of English text might not help much. Finally, the interrogator or a third party will be able to decide unambiguously after a few minutes whether or not a subject has passed the test."
                },
                {
                    "title": "Parallel distributed processing: explorations in the microstructure of cognition, vol. 1: foundations",
                    "abstract": "The fundamental principles, basic mechanisms, and formal analyses involved in the development of parallel distributed processing (PDP) systems are presented in individual chapters contributed by leading experts. Topics examined include distributed representations, PDP models and general issues in cognitive science, feature discovery by competitive learning, the foundations of harmony theory, learning and relearning in Boltzmann machines, and learning internal representations by error propagation. Consideration is given to linear algebra in PDP, the logic of additive functions, resource requirements of standard and programmable nets, and the P3 parallel-network simulating system."
                },
                {
                    "title": "UltraFeedback: Boosting Language Models with High-quality Feedback",
                    "abstract": "Reinforcement learning from human feedback (RLHF) has become a pivot technique in aligning large language models (LLMs) with human preferences. In RLHF practice, preference data plays a crucial role in bridging human proclivity and LLMs. However, the scarcity of diverse, naturalistic datasets of human preferences on LLM outputs at scale poses a great challenge to RLHF as well as feedback learning research within the open-source community. Current preference datasets, either proprietary or limited in size and prompt variety, result in limited RLHF adoption in open-source models and hinder further exploration. In this study, we propose ULTRAFEEDBACK, a large-scale, high-quality, and diversified preference dataset designed to overcome these limitations and foster RLHF development. To create ULTRAFEEDBACK, we compile a diverse array of instructions and models from multiple sources to produce comparative data. We meticulously devise annotation instructions and employ GPT-4 to offer detailed feedback in both numerical and textual forms. ULTRAFEEDBACK establishes a reproducible and expandable preference data construction pipeline, serving as a solid foundation for future RLHF and feedback learning research. Utilizing ULTRAFEEDBACK, we train various models to demonstrate its effectiveness, including the reward model UltraRM, chat language model UltraLM-13B-PPO, and critique model UltraCM. Experimental results indicate that our models outperform existing open-source models, achieving top performance across multiple benchmarks. Our data and models are available at https://github.com/thunlp/UltraFeedback."
                },
                {
                    "title": "Learn-to-Share: A Hardware-friendly Transfer Learning Framework Exploiting Computation and Parameter Sharing",
                    "abstract": "Task-speci\ufb01c \ufb01ne-tuning on pre-trained transformers has achieved performance breakthroughs in multiple NLP tasks. Yet, as both computation and parameter size grows linearly with the number of sub-tasks, it is increasingly dif\ufb01cult to adopt such methods to the real world due to unrealistic memory and computation overhead on computing devices. Previous works on \ufb01ne-tuning focus on reducing the growing parameter size to save storage cost by parameter sharing. However, compared to storage, the constraint of computation is a more critical issue with the \ufb01ne-tuning models in modern computing environments. In this work, we propose LeTS , a framework that leverages both computation and parameter sharing across multiple tasks. Compared to traditional \ufb01ne-tuning, LeTS proposes a novel neural architecture that contains a \ufb01xed pre-trained transformer model, plus learnable additive components for sub-tasks. The learnable components reuse the intermediate activations in the \ufb01xed pre-trained model, decoupling computation dependency. Differentiable neural architecture search is used to determine a task-speci\ufb01c computation sharing scheme, and a novel early stage pruning is applied to additive components for sparsity to achieve parameter sharing. Extensive experiments show that with 1.4% of extra parameters per task, LeTS reduces the computation by 49.5% on GLUE benchmarks with only 0.2% accuracy loss compared to full \ufb01ne-tuning."
                },
                {
                    "title": "Prefix-Tuning: Optimizing Continuous Prompts for Generation",
                    "abstract": "Fine-tuning is the de facto way of leveraging large pretrained language models for downstream tasks. However, fine-tuning modifies all the language model parameters and therefore necessitates storing a full copy for each task. In this paper, we propose prefix-tuning, a lightweight alternative to fine-tuning for natural language generation tasks, which keeps language model parameters frozen and instead optimizes a sequence of continuous task-specific vectors, which we call the prefix. Prefix-tuning draws inspiration from prompting for language models, allowing subsequent tokens to attend to this prefix as if it were \u201cvirtual tokens\u201d. We apply prefix-tuning to GPT-2 for table-to-text generation and to BART for summarization. We show that by learning only 0.1% of the parameters, prefix-tuning obtains comparable performance in the full data setting, outperforms fine-tuning in low-data settings, and extrapolates better to examples with topics that are unseen during training."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "An Adversarial Winograd Schema Challenge at Scale",
                    "abstract": "The Winograd Schema Challenge (WSC), proposed by Levesque et al. (2011) as an alternative to the Turing Test, was originally designed as a pronoun resolution problem that cannot be solved based on statistical patterns in large text corpora. However, recent studies suggest that current WSC datasets, even when composed carefully by experts, are still prone to such biases that statistical methods can exploit. We introduce WINOGRANDE, a new collection of WSC problems that are adversarially constructed to be robust against spurious statistical biases. While the original WSC dataset provided only 273 instances, WINOGRANDE includes 43,985 instances, half of which are determined as adversarial. Key to our approach is a novel adversarial filtering algorithm AFLITE for systematic bias reduction, combined with a careful crowdsourcing design. Despite the significant increase in training data, the performance of existing stateof-the-art methods remains modest (61.6%) and contrasts with high human performance (90.8%) for the binary questions. In addition, WINOGRANDE allows us to use transfer learning for achieving new state-of-the-art results on the original WSC and related datasets. Finally, we discuss how biases lead to overestimating the true capabilities of machine commonsense."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we achieve parameter-efficient finetuning of large pretrained language models to minimize computational overhead while maintaining performance across various tasks?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses the growing demand for efficient machine learning models that can be adapted to specific tasks without incurring high computational costs. By developing methods for parameter-efficient finetuning, we can democratize access to advanced language models, enabling smaller organizations and researchers to leverage state-of-the-art technology. This advancement could lead to significant improvements in various applications, such as natural language understanding, dialogue systems, and automated content generation, ultimately pushing the boundaries of what is possible with AI.\n\n### [Question 3] - Why is it hard?\nThe challenges in achieving parameter-efficient finetuning stem from the inherent complexity of large language models, which have millions or billions of parameters. Naive approaches may fail because they do not adequately account for the intricate relationships between model parameters and the specific tasks they are being adapted for. Additionally, the computational overhead introduced during inference can be substantial, particularly when intervening with multiple layers or using higher ranks for parameter adjustments. Overcoming these technical obstacles requires innovative methodologies that balance efficiency and performance, which is not straightforward given the trade-offs involved.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on either full finetuning or simplistic adaptations that do not leverage the full potential of pretrained models. Limitations in computational resources and a lack of effective techniques for parameter-efficient finetuning have hindered progress in this area. Existing solutions may not adequately address the overhead introduced during inference or fail to generalize across different tasks. Our approach differs by introducing a systematic method for rank-1 interventions and optimizing the number of layers involved, which has not been thoroughly explored in prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves using rank-1 interventions on pretrained language models, specifically focusing on optimizing the number of prompt tokens and the layers being intervened upon. We will utilize datasets from various tasks, such as WinoGrande and OBQA, to evaluate the performance of our approach. The primary metric for success will be the generation time and accuracy of the model's responses. We expect our results to demonstrate that our method can significantly reduce computational overhead while maintaining or improving performance on downstream tasks, thereby validating the effectiveness of parameter-efficient finet"
            }
        },
        "author_data": {
            "df51ed27-d13c-4fa1-bd2f-e79107f4c0c1": {
                "pk": "df51ed27-d13c-4fa1-bd2f-e79107f4c0c1",
                "name": "Zhengxuan Wu",
                "collaborators": [
                    "Christopher Potts",
                    "Desmond C. Ong",
                    "Atticus Geiger",
                    "Jing Huang",
                    "Xiyu Zhang",
                    "Karel D'Oosterlinck",
                    "Thomas Icard",
                    "Noah D. Goodman",
                    "Yueyi Jiang",
                    "Jason Luo"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Explainable AI",
                    "Emotion Recognition",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "Disentangling Latent Emotions of Word Embeddings on Complex Emotional Narratives",
                        "abstract": "Word embedding models such as GloVe are widely used in natural language processing (NLP) research to convert words into vectors. Here, we provide a preliminary guide to probe latent emotions in text through GloVe word vectors. First, we trained a neural network model to predict continuous emotion valence ratings by taking linguistic inputs from Stanford Emotional Narratives Dataset (SEND). After interpreting the weights in the model, we found that only a few dimensions of the word vectors contributed to expressing emotions in text, and words were clustered on the basis of their emotional polarities. Furthermore, we performed a linear transformation that projected high dimensional embedded vectors into an emotion space. Based on NRC Emotion Lexicon (EmoLex), we visualized the entanglement of emotions in the lexicon by using both projected and raw GloVe word vectors. We showed that, in the proposed emotion space, we were able to better disentangle emotions than using raw GloVe vectors alone. In addition, we found that the sum vectors of different pairs of emotion words successfully captured expressed human feelings in the EmoLex. For example, the sum of two embedded word vectors expressing Joy and Trust which express Love shared high similarity (similarity score .62) with the embedded vector expressing Optimism. On the contrary, this sum vector was dissimilar (similarity score -.19) with the the embedded vector expressing Remorse. In this paper, we argue that through the proposed emotion space, arithmetic of emotions is preserved in the word vectors. The affective representation uncovered in emotion vector space could shed some light on how to help machines to disentangle emotion expressed in word embeddings."
                    },
                    {
                        "title": "Uncovering Political Promotion in China: A Network Analysis of Patronage Relationship in Autocracy",
                        "abstract": "Understanding patronage networks in Chinese Bureaucracy helps us quantify the promotion mechanism underlying autocratic political systems. Although there are qualitative studies analyzing political promotions, few use quantitative methods to model promotions and make inferences on the fitted mathematical model. Using publicly available datasets, we implement network analysis techniques to advance scholarly understanding of patronage networks in autocratic regimes, using the Chinese bureaucracy as an example. Using graph-based and non-graph-based features, we design three studies to examine drivers of political promotions. We find that careers of politicians are closely associated with their genders, home origins, and positions in the patronage networks."
                    },
                    {
                        "title": "Pragmatically Informative Color Generation by Grounding Contextual Modifiers",
                        "abstract": "Grounding language in contextual information is crucial for fine-grained natural language understanding. One important task that involves grounding contextual modifiers is color generation. Given a reference color \"green\", and a modifier \"bluey\", how does one generate a color that could represent \"bluey green\"? We propose a computational pragmatics model that formulates this color generation task as a recursive game between speakers and listeners. In our model, a pragmatic speaker reasons about the inferences that a listener would make, and thus generates a modified color that is maximally informative to help the listener recover the original referents. In this paper, we show that incorporating pragmatic information provides significant improvements in performance compared with other state-of-the-art deep learning models where pragmatic inference and flexibility in representing colors from a large continuous space are lacking. Our model has an absolute 98% increase in performance for the test cases where the reference colors are unseen during training, and an absolute 40% increase in performance for the test cases where both the reference colors and the modifiers are unseen during training."
                    },
                    {
                        "title": "Context-Guided BERT for Targeted Aspect-Based Sentiment Analysis",
                        "abstract": "Aspect-based sentiment analysis (ABSA) and Targeted ASBA (TABSA) allow finer-grained inferences about sentiment to be drawn from the same text, depending on context. For example, a given text can have different targets (e.g., neighborhoods) and different aspects (e.g., price or safety), with different sentiment associated with each target-aspect pair. In this paper, we investigate whether adding context to self-attention models improves performance on (T)ABSA. We propose two variants of Context-Guided BERT (CG-BERT) that learn to distribute attention under different contexts. We first adapt a context-aware Transformer to produce a CG-BERT that uses context-guided softmax-attention. Next, we propose an improved Quasi-Attention CG-BERT model that learns a compositional attention that supports subtractive attention. We train both models with pretrained BERT on two (T)ABSA datasets: SentiHood and SemEval-2014 (Task 4). Both models achieve new state-of-the-art results with our QACG-BERT model having the best performance. Furthermore, we provide analyses of the impact of context in the our proposed models. Our work provides more evidence for the utility of adding context-dependencies to pretrained self-attention-based language models for context-based natural language tasks."
                    },
                    {
                        "title": "On Explaining Your Explanations of BERT: An Empirical Study with Sequence Classification",
                        "abstract": "BERT, as one of the pretrianed language models, attracts the most attention in recent years for creating new benchmarks across GLUE tasks via fine-tuning. One pressing issue is to open up the blackbox and explain the decision makings of BERT. A number of attribution techniques have been proposed to explain BERT models, but are often limited to sequence to sequence tasks. In this paper, we adapt existing attribution methods on explaining decision makings of BERT in sequence classification tasks. We conduct extensive analyses of four existing attribution methods by applying them to four different datasets in sentiment analysis. We compare the reliability and robustness of each method via various ablation studies. Furthermore, we test whether attribution methods explain generalized semantics across semantically similar tasks. Our work provides solid guidance for using attribution methods to explain decision makings of BERT for downstream classification tasks."
                    },
                    {
                        "title": "Identifying the Limits of Cross-Domain Knowledge Transfer for Pretrained Models",
                        "abstract": "There is growing evidence that pretrained language models improve task-specific fine-tuning not just for the languages seen in pretraining, but also for new languages and even non-linguistic data. What is the nature of this surprising cross-domain transfer? We offer a partial answer via a systematic exploration of how much transfer occurs when models are denied any information about word identity via random scrambling. In four classification tasks and two sequence labeling tasks, we evaluate baseline models, LSTMs using GloVe embeddings, and BERT. We find that only BERT shows high rates of transfer into our scrambled domains, and for classification but not sequence labeling tasks. Our analyses seek to explain why transfer succeeds for some tasks but not others, to isolate the separate contributions of pretraining versus fine-tuning, and to quantify the role of word frequency. These findings help explain where and why cross-domain transfer occurs, which can guide future studies and practical fine-tuning efforts."
                    },
                    {
                        "title": "Oolong: Investigating What Makes Transfer Learning Hard with Controlled Studies",
                        "abstract": "When we transfer a pretrained language model to a new language, there are many axes of variation that change at once. To disentangle the impact of different factors like syntactic similarity and vocabulary similarity, we propose a set of controlled transfer studies: we systematically transform the language of the GLUE benchmark, altering one axis of crosslingual variation at a time, and then measure the resulting drops in a pretrained model's downstream performance. We find that models can largely recover from syntactic-style shifts, but cannot recover from vocabulary misalignment and embedding matrix re-initialization, even with continued pretraining on 15 million tokens. %On the other hand, transferring to a dataset with an unaligned vocabulary is extremely hard to recover from in the low-data regime. Moreover, good-quality tokenizers in the transfer language do not make vocabulary alignment easier. Our experiments provide insights into the factors of cross-lingual transfer that researchers should most focus on when designing language transfer scenarios."
                    },
                    {
                        "title": "ReCOGS: How Incidental Details of a Logical Form Overshadow an Evaluation of Semantic Interpretation",
                        "abstract": "Compositional generalization benchmarks for semantic parsing seek to assess whether models can accurately compute meanings for novel sentences, but operationalize this in terms of logical form (LF) prediction. This raises the concern that semantically irrelevant details of the chosen LFs could shape model performance. We argue that this concern is realized for the COGS benchmark. COGS poses generalization splits that appear impossible for present-day models, which could be taken as an indictment of those models. However, we show that the negative results trace to incidental features of COGS LFs. Converting these LFs to semantically equivalent ones and factoring out capabilities unrelated to semantic interpretation, we find that even baseline models get traction. A recent variable-free translation of COGS LFs suggests similar conclusions, but we observe this format is not semantically equivalent; it is incapable of accurately representing some COGS meanings. These findings inform our proposal for ReCOGS, a modified version of COGS that comes closer to assessing the target semantic capabilities while remaining very challenging. Overall, our results reaffirm the importance of compositional generalization and careful benchmark task design."
                    },
                    {
                        "title": "DynaSent: A Dynamic Benchmark for Sentiment Analysis",
                        "abstract": "We introduce DynaSent ('Dynamic Sentiment'), a new English-language benchmark task for ternary (positive/negative/neutral) sentiment analysis. DynaSent combines naturally occurring sentences with sentences created using the open-source Dynabench Platform, which facilities human-and-model-in-the-loop dataset creation. DynaSent has a total of 121,634 sentences, each validated by five crowdworkers, and its development and test splits are designed to produce chance performance for even the best models we have been able to develop; when future models solve this task, we will use them to create DynaSent version 2, continuing the dynamic evolution of this benchmark. Here, we report on the dataset creation effort, focusing on the steps we took to increase quality and reduce artifacts. We also present evidence that DynaSent's Neutral category is more coherent than the comparable category in other benchmarks, and we motivate training models from scratch for each round over successive fine-tuning."
                    },
                    {
                        "title": "Not Now, Ask Later: Users Weaken Their Behavior Change Regimen Over Time, But Expect To Re-Strengthen It Imminently",
                        "abstract": "How effectively do we adhere to nudges and interventions that help us control our online browsing habits? If we have a temporary lapse and disable the behavior change system, do we later resume our adherence, or has the dam broken? In this paper, we investigate these questions through log analyses of 8,000+ users on HabitLab, a behavior change platform that helps users reduce their time online. We find that, while users typically begin with high-challenge interventions, over time they allow themselves to slip into easier and easier interventions. Despite this, many still expect to return to the harder interventions imminently: they repeatedly choose to be asked to change difficulty again on the next visit, declining to have the system save their preference for easy interventions."
                    },
                    {
                        "title": "ReaSCAN: Compositional Reasoning in Language Grounding",
                        "abstract": "The ability to compositionally map language to referents, relations, and actions is an essential component of language understanding. The recent gSCAN dataset (Ruis et al. 2020, NeurIPS) is an inspiring attempt to assess the capacity of models to learn this kind of grounding in scenarios involving navigational instructions. However, we show that gSCAN's highly constrained design means that it does not require compositional interpretation and that many details of its instructions and scenarios are not required for task success. To address these limitations, we propose ReaSCAN, a benchmark dataset that builds off gSCAN but requires compositional language interpretation and reasoning about entities and relations. We assess two models on ReaSCAN: a multi-modal baseline and a state-of-the-art graph convolutional neural model. These experiments show that ReaSCAN is substantially harder than gSCAN for both neural architectures. This suggests that ReaSCAN can serve as a valuable benchmark for advancing our understanding of models' compositional generalization and reasoning capabilities."
                    },
                    {
                        "title": "Structured Self-Attention Weights Encode Semantics in Sentiment Analysis",
                        "abstract": "Neural attention, especially the self-attention made popular by the Transformer, has become the workhorse of state-of-the-art natural language processing (NLP) models. Very recent work suggests that the self-attention in the Transformer encodes syntactic information; Here, we show that self-attention scores encode semantics by considering sentiment analysis tasks. In contrast to gradient-based feature attribution methods, we propose a simple and effective Layer-wise Attention Tracing (LAT) method to analyze structured attention weights. We apply our method to Transformer models trained on two tasks that have surface dissimilarities, but share common semantics---sentiment analysis of movie reviews and time-series valence prediction in life story narratives. Across both tasks, words with high aggregated attention weights were rich in emotional semantics, as quantitatively validated by an emotion lexicon labeled by human annotators. Our results show that structured attention weights encode rich semantics in sentiment analysis, and match human interpretations of semantics."
                    },
                    {
                        "title": "Inducing Character-level Structure in Subword-based Language Models with Type-level Interchange Intervention Training",
                        "abstract": "Language tasks involving character-level manipulations (e.g., spelling corrections, arithmetic operations, word games) are challenging for models operating on subword units. To address this, we develop a causal intervention framework to learn robust and interpretable character representations inside subword-based language models. Our method treats each character as a typed variable in a causal model and learns such causal structures by adapting the interchange intervention training method of Geiger et al. (2021). We additionally introduce a suite of character-level tasks that systematically vary in their dependence on meaning and sequence-level context. While character-level models still perform best on purely form-based tasks like string reversal, our method outperforms character-level models on more complex tasks that blend form, meaning, and context, such as spelling correction in context and word search games. Compared with standard subword-based models, our approach also significantly improves robustness on unseen token sequences and leads to human-interpretable internal representations of characters."
                    },
                    {
                        "title": "Rigorously Assessing Natural Language Explanations of Neurons",
                        "abstract": "Natural language is an appealing medium for explaining how large language models process and store information, but evaluating the faithfulness of such explanations is challenging. To help address this, we develop two modes of evaluation for natural language explanations that claim individual neurons represent a concept in a text input. In the observational mode, we evaluate claims that a neuron $a$ activates on all and only input strings that refer to a concept picked out by the proposed explanation $E$. In the intervention mode, we construe $E$ as a claim that the neuron $a$ is a causal mediator of the concept denoted by $E$. We apply our framework to the GPT-4-generated explanations of GPT-2 XL neurons of Bills et al. (2023) and show that even the most confident explanations have high error rates and little to no causal efficacy. We close the paper by critically assessing whether natural language is a good choice for explanations and whether neurons are the best level of analysis."
                    },
                    {
                        "title": "RAVEL: Evaluating Interpretability Methods on Disentangling Language Model Representations",
                        "abstract": "Individual neurons participate in the representation of multiple high-level concepts. To what extent can different interpretability methods successfully disentangle these roles? To help address this question, we introduce RAVEL (Resolving Attribute-Value Entanglements in Language Models), a dataset that enables tightly controlled, quantitative comparisons between a variety of existing interpretability methods. We use the resulting conceptual framework to define the new method of Multi-task Distributed Alignment Search (MDAS), which allows us to find distributed representations satisfying multiple causal criteria. With Llama2-7B as the target language model, MDAS achieves state-of-the-art results on RAVEL, demonstrating the importance of going beyond neuron-level analyses to identify features distributed across activations. We release our benchmark at https://github.com/explanare/ravel."
                    },
                    {
                        "title": "ZeroC: A Neuro-Symbolic Model for Zero-shot Concept Recognition and Acquisition at Inference Time",
                        "abstract": "Humans have the remarkable ability to recognize and acquire novel visual concepts in a zero-shot manner. Given a high-level, symbolic description of a novel concept in terms of previously learned visual concepts and their relations, humans can recognize novel concepts without seeing any examples. Moreover, they can acquire new concepts by parsing and communicating symbolic structures using learned visual concepts and relations. Endowing these capabilities in machines is pivotal in improving their generalization capability at inference time. In this work, we introduce Zero-shot Concept Recognition and Acquisition (ZeroC), a neuro-symbolic architecture that can recognize and acquire novel concepts in a zero-shot way. ZeroC represents concepts as graphs of constituent concept models (as nodes) and their relations (as edges). To allow inference time composition, we employ energy-based models (EBMs) to model concepts and relations. We design ZeroC architecture so that it allows a one-to-one mapping between a symbolic graph structure of a concept and its corresponding EBM, which for the first time, allows acquiring new concepts, communicating its graph structure, and applying it to classification and detection tasks (even across domains) at inference time. We introduce algorithms for learning and inference with ZeroC. We evaluate ZeroC on a challenging grid-world dataset which is designed to probe zero-shot concept recognition and acquisition, and demonstrate its capability."
                    },
                    {
                        "title": "Interpretability at Scale: Identifying Causal Mechanisms in Alpaca",
                        "abstract": "Obtaining human-interpretable explanations of large, general-purpose language models is an urgent goal for AI safety. However, it is just as important that our interpretability methods are faithful to the causal dynamics underlying model behavior and able to robustly generalize to unseen inputs. Distributed Alignment Search (DAS) is a powerful gradient descent method grounded in a theory of causal abstraction that has uncovered perfect alignments between interpretable symbolic algorithms and small deep learning models fine-tuned for specific tasks. In the present paper, we scale DAS significantly by replacing the remaining brute-force search steps with learned parameters -- an approach we call Boundless DAS. This enables us to efficiently search for interpretable causal structure in large language models while they follow instructions. We apply Boundless DAS to the Alpaca model (7B parameters), which, off the shelf, solves a simple numerical reasoning problem. With Boundless DAS, we discover that Alpaca does this by implementing a causal model with two interpretable boolean variables. Furthermore, we find that the alignment of neural representations with these variables is robust to changes in inputs and instructions. These findings mark a first step toward faithfully understanding the inner-workings of our ever-growing and most widely deployed language models. Our tool is extensible to larger LLMs and is released publicly at `https://github.com/stanfordnlp/pyvene`."
                    },
                    {
                        "title": "Attending to Emotional Narratives",
                        "abstract": "Attention mechanisms in deep neural networks have achieved excellent performance on sequence-prediction tasks. Here, we show that these recently-proposed attention-based mechanisms---in particular, the Transformer with its parallelizable self-attention layers, and the Memory Fusion Network with attention across modalities and time---also generalize well to multimodal time-series emotion recognition. Using a recently-introduced dataset of emotional autobiographical narratives, we adapt and apply these two attention mechanisms to predict emotional valence over time. Our models perform extremely well, in some cases reaching a performance comparable with human raters. We end with a discussion of the implications of attention mechanisms to affective computing."
                    },
                    {
                        "title": "Causal Proxy Models for Concept-Based Model Explanations",
                        "abstract": "Explainability methods for NLP systems encounter a version of the fundamental problem of causal inference: for a given ground-truth input text, we never truly observe the counterfactual texts necessary for isolating the causal effects of model representations on outputs. In response, many explainability methods make no use of counterfactual texts, assuming they will be unavailable. In this paper, we show that robust causal explainability methods can be created using approximate counterfactuals, which can be written by humans to approximate a specific counterfactual or simply sampled using metadata-guided heuristics. The core of our proposal is the Causal Proxy Model (CPM). A CPM explains a black-box model $\\mathcal{N}$ because it is trained to have the same actual input/output behavior as $\\mathcal{N}$ while creating neural representations that can be intervened upon to simulate the counterfactual input/output behavior of $\\mathcal{N}$. Furthermore, we show that the best CPM for $\\mathcal{N}$ performs comparably to $\\mathcal{N}$ in making factual predictions, which means that the CPM can simply replace $\\mathcal{N}$, leading to more explainable deployed models. Our code is available at https://github.com/frankaging/Causal-Proxy-Model."
                    },
                    {
                        "title": "Finding Alignments Between Interpretable Causal Variables and Distributed Neural Representations",
                        "abstract": "Causal abstraction is a promising theoretical framework for explainable artificial intelligence that defines when an interpretable high-level causal model is a faithful simplification of a low-level deep learning system. However, existing causal abstraction methods have two major limitations: they require a brute-force search over alignments between the high-level model and the low-level one, and they presuppose that variables in the high-level model will align with disjoint sets of neurons in the low-level one. In this paper, we present distributed alignment search (DAS), which overcomes these limitations. In DAS, we find the alignment between high-level and low-level models using gradient descent rather than conducting a brute-force search, and we allow individual neurons to play multiple distinct roles by analyzing representations in non-standard bases-distributed representations. Our experiments show that DAS can discover internal structure that prior approaches miss. Overall, DAS removes previous obstacles to conducting causal abstraction analyses and allows us to find conceptual structure in trained neural nets."
                    }
                ]
            },
            "acd77948-7662-4fbe-a15d-c42f99d667a9": {
                "pk": "acd77948-7662-4fbe-a15d-c42f99d667a9",
                "name": "Aryaman Arora",
                "collaborators": [
                    "Nathan Schneider",
                    "Adam Farris",
                    "Samopriya Basu",
                    "Nitin Venkateswaran",
                    "Brett Reynolds",
                    "Suresh Kolichala",
                    "Dan Jurafsky",
                    "Christopher Potts",
                    "Kabilan Prasanna",
                    "Gopalakrishnan R"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Historical Linguistics",
                    "Linguistic Annotation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "IruMozhi: Automatically classifying diglossia in Tamil",
                        "abstract": "Tamil, a Dravidian language of South Asia, is a highly diglossic language with two very different registers in everyday use: Literary Tamil (preferred in writing and formal communication) and Spoken Tamil (confined to speech and informal media). Spoken Tamil is under-supported in modern NLP systems. In this paper, we release IruMozhi, a human-annotated dataset of parallel text in Literary and Spoken Tamil. We train classifiers on the task of identifying which variety a text belongs to. We use these models to gauge the availability of pretraining data in Spoken Tamil, to audit the composition of existing labelled datasets for Tamil, and to encourage future work on the variety."
                    },
                    {
                        "title": "For the Purpose of Curry: A UD Treebank for Ashokan Prakrit",
                        "abstract": "We present the first linguistically annotated treebank of Ashokan Prakrit, an early Middle Indo-Aryan dialect continuum attested through Emperor Ashoka Maurya's 3rd century BCE rock and pillar edicts. For annotation, we used the multilingual Universal Dependencies (UD) formalism, following recent UD work on Sanskrit and other Indo-Aryan languages. We touch on some interesting linguistic features that posed issues in annotation: regnal names and other nominal compounds, \"proto-ergative\" participial constructions, and possible grammaticalizations evidenced by sandhi (phonological assimilation across morpheme boundaries). Eventually, we plan for a complete annotation of all attested Ashokan texts, towards the larger goals of improving UD coverage of different diachronic stages of Indo-Aryan and studying language change in Indo-Aryan using computational methods."
                    },
                    {
                        "title": "Bhasacitra: Visualising the dialect geography of South Asia",
                        "abstract": "We present Bhasacitra, a dialect mapping system for South Asia built on a database of linguistic studies of languages of the region annotated for topic and location data. We analyse language coverage and look towards applications to typology by visualising example datasets. The application is not only meant to be useful for feature mapping, but also serves as a new kind of interactive bibliography for linguists of South Asian languages."
                    },
                    {
                        "title": "Hindi-Urdu Adposition and Case Supersenses v1.0",
                        "abstract": "These are the guidelines for the application of SNACS (Semantic Network of Adposition and Case Supersenses; Schneider et al. 2018) to Modern Standard Hindi of Delhi. SNACS is an inventory of 50 supersenses (semantic labels) for labelling the use of adpositions and case markers with respect to both lexical-semantic function and relation to the underlying context. The English guidelines (Schneider et al., 2020) were used as a model for this document.   Besides the case system, Hindi has an extremely rich adpositional system built on the oblique genitive, with productive incorporation of loanwords even in present-day Hinglish.   This document is aligned with version 2.5 of the English guidelines."
                    },
                    {
                        "title": "MASALA: Modelling and Analysing the Semantics of Adpositions in Linguistic Annotation of Hindi",
                        "abstract": "We present a completed, publicly available corpus of annotated semantic relations of adpositions and case markers in Hindi. We used the multilingual SNACS annotation scheme, which has been applied to a variety of typologically diverse languages. Building on past work examining linguistic problems in SNACS annotation, we use language models to attempt automatic labelling of SNACS supersenses in Hindi and achieve results competitive with past work on English. We look towards upstream applications in semantic role labelling and extension to related languages such as Gujarati."
                    },
                    {
                        "title": "CGELBank: CGEL as a Framework for English Syntax Annotation",
                        "abstract": "We introduce the syntactic formalism of the \\textit{Cambridge Grammar of the English Language} (CGEL) to the world of treebanking through the CGELBank project. We discuss some issues in linguistic analysis that arose in adapting the formalism to corpus annotation, followed by quantitative and qualitative comparisons with parallel UD and PTB treebanks. We argue that CGEL provides a good tradeoff between comprehensiveness of analysis and usability for annotation, which motivates expanding the treebank with automatic conversion in the future."
                    },
                    {
                        "title": "Jambu: A historical linguistic database for South Asian languages",
                        "abstract": "We introduce Jambu, a cognate database of South Asian languages which unifies dozens of previous sources in a structured and accessible format. The database includes 287k lemmata from 602 lects, grouped together in 23k sets of cognates. We outline the data wrangling necessary to compile the dataset and train neural models for reflex prediction on the Indo-Aryan subset of the data. We hope that Jambu is an invaluable resource for all historical linguists and Indologists, and look towards further improvement and expansion of the database."
                    },
                    {
                        "title": "CGELBank Annotation Manual v1.1",
                        "abstract": "CGELBank is a treebank and associated tools based on a syntactic formalism for English derived from the Cambridge Grammar of the English Language. This document lays out the particularities of the CGELBank annotation scheme."
                    },
                    {
                        "title": "Supervised Grapheme-to-Phoneme Conversion of Orthographic Schwas in Hindi and Punjabi",
                        "abstract": "Hindi grapheme-to-phoneme (G2P) conversion is mostly trivial, with one exception: whether a schwa represented in the orthography is pronounced or unpronounced (deleted). Previous work has attempted to predict schwa deletion in a rule-based fashion using prosodic or phonetic analysis. We present the first statistical schwa deletion classifier for Hindi, which relies solely on the orthography as the input and outperforms previous approaches. We trained our model on a newly-compiled pronunciation lexicon extracted from various online dictionaries. Our best Hindi model achieves state of the art performance, and also achieves good performance on a closely related language, Punjabi, without modification."
                    },
                    {
                        "title": "Computational historical linguistics and language diversity in South Asia",
                        "abstract": "South Asia is home to a plethora of languages, many of which severely lack access to new language technologies. This linguistic diversity also results in a research environment conducive to the study of comparative, contact, and historical linguistics -- fields which necessitate the gathering of extensive data from many languages. We claim that data scatteredness (rather than scarcity) is the primary obstacle in the development of South Asian language technology, and suggest that the study of language history is uniquely aligned with surmounting this obstacle. We review recent developments in and at the intersection of South Asian NLP and historical-comparative linguistics, describing our and others' current efforts in this area. We also offer new strategies towards breaking the data barrier."
                    },
                    {
                        "title": "Estimating the Entropy of Linguistic Distributions",
                        "abstract": "Shannon entropy is often a quantity of interest to linguists studying the communicative capacity of human language. However, entropy must typically be estimated from observed data because researchers do not have access to the underlying probability distribution that gives rise to these data. While entropy estimation is a well-studied problem in other fields, there is not yet a comprehensive exploration of the efficacy of entropy estimators for use with linguistic data. In this work, we fill this void, studying the empirical effectiveness of different entropy estimators for linguistic distributions. In a replication of two recent information-theoretic linguistic studies, we find evidence that the reported effect size is over-estimated due to over-reliance on poor entropy estimators. Finally, we end our paper with concrete recommendations for entropy estimation depending on distribution type and data availability."
                    },
                    {
                        "title": "CausalGym: Benchmarking causal interpretability methods on linguistic tasks",
                        "abstract": "Language models (LMs) have proven to be powerful tools for psycholinguistic research, but most prior work has focused on purely behavioural measures (e.g., surprisal comparisons). At the same time, research in model interpretability has begun to illuminate the abstract causal mechanisms shaping LM behavior. To help bring these strands of research closer together, we introduce CausalGym. We adapt and expand the SyntaxGym suite of tasks to benchmark the ability of interpretability methods to causally affect model behaviour. To illustrate how CausalGym can be used, we study the pythia models (14M--6.9B) and assess the causal efficacy of a wide range of interpretability methods, including linear probing and distributed alignment search (DAS). We find that DAS outperforms the other methods, and so we use it to study the learning trajectory of two difficult linguistic phenomena in pythia-1b: negative polarity item licensing and filler--gap dependencies. Our analysis shows that the mechanism implementing both of these tasks is learned in discrete stages, not gradually."
                    },
                    {
                        "title": "Localizing Model Behavior with Path Patching",
                        "abstract": "Localizing behaviors of neural networks to a subset of the network's components or a subset of interactions between components is a natural first step towards analyzing network mechanisms and possible failure modes. Existing work is often qualitative and ad-hoc, and there is no consensus on the appropriate way to evaluate localization claims. We introduce path patching, a technique for expressing and quantitatively testing a natural class of hypotheses expressing that behaviors are localized to a set of paths. We refine an explanation of induction heads, characterize a behavior of GPT-2, and open source a framework for efficiently running similar experiments."
                    },
                    {
                        "title": "Bayesian scaling laws for in-context learning",
                        "abstract": "In-context learning (ICL) is a powerful technique for getting language models to perform complex tasks with no training updates. Prior work has established strong correlations between the number of in-context examples provided and the accuracy of the model's predictions. In this paper, we seek to explain this correlation by showing that ICL approximates a Bayesian learner. This perspective gives rise to a family of novel Bayesian scaling laws for ICL. In experiments with \\mbox{GPT-2} models of different sizes, our scaling laws exceed or match existing scaling laws in accuracy while also offering interpretable terms for task priors, learning efficiency, and per-example probabilities. To illustrate the analytic power that such interpretable scaling laws provide, we report on controlled synthetic dataset experiments designed to inform real-world studies of safety alignment. In our experimental protocol, we use SFT to suppress an unwanted existing model capability and then use ICL to try to bring that capability back (many-shot jailbreaking). We then experiment on real-world instruction-tuned LLMs using capabilities benchmarks as well as a new many-shot jailbreaking dataset. In all cases, Bayesian scaling laws accurately predict the conditions under which ICL will cause the suppressed behavior to reemerge, which sheds light on the ineffectiveness of post-training at increasing LLM safety."
                    },
                    {
                        "title": "Towards Vision-Language Mechanistic Interpretability: A Causal Tracing Tool for BLIP",
                        "abstract": "Mechanistic interpretability seeks to understand the neural mechanisms that enable specific behaviors in Large Language Models (LLMs) by leveraging causality-based methods. While these approaches have identified neural circuits that copy spans of text, capture factual knowledge, and more, they remain unusable for multimodal models since adapting these tools to the vision-language domain requires considerable architectural changes. In this work, we adapt a unimodal causal tracing tool to BLIP to enable the study of the neural mechanisms underlying image-conditioned text generation. We demonstrate our approach on a visual question answering dataset, highlighting the causal relevance of later layer representations for all tokens. Furthermore, we release our BLIP causal tracing tool as open source to enable further experimentation in vision-language mechanistic interpretability by the community. Our code is available at https://github.com/vedantpalit/Towards-Vision-Language-Mechanistic-Interpretability."
                    },
                    {
                        "title": "PASTRIE: A Corpus of Prepositions Annotated with Supersense Tags in Reddit International English",
                        "abstract": "We present the Prepositions Annotated with Supersense Tags in Reddit International English (\"PASTRIE\") corpus, a new dataset containing manually annotated preposition supersenses of English data from presumed speakers of four L1s: English, French, German, and Spanish. The annotations are comprehensive, covering all preposition types and tokens in the sample. Along with the corpus, we provide analysis of distributional patterns across the included L1s and a discussion of the influence of L1s on L2 preposition choice."
                    }
                ]
            },
            "7bcc4909-58e4-4d40-b0ae-401358459b2d": {
                "pk": "7bcc4909-58e4-4d40-b0ae-401358459b2d",
                "name": "Zheng Wang",
                "collaborators": [],
                "domain": [
                    "DNS Security",
                    "Haptic Guidance",
                    "Machine Learning",
                    "Cryptography"
                ],
                "publications": [
                    {
                        "title": "An Estimation Method Using Periodic Inspection of Indicators",
                        "abstract": "This paper proposes a new approach for estimating the failure time distribution using the indicator data. The indicators, which are checked by periodic inspection of a standby redundant system, only convey whether at least one failure occurs per interval. The estimation procedure first obtains the estimation of the forward recurrence time using the indicator data. Then the mean is estimated based on its relationship with the forward recurrence time. And the estimation of the sampled Cdf is thus derived based on its relationship with the forward recurrence time and the mean. Finally, the Cdf function is estimated using interpolation method. The simulation results showed that the estimation method performed well for the four Weibull distributions."
                    },
                    {
                        "title": "Take up DNSSEC When Needed",
                        "abstract": "The threats of caching poisoning attacks largely stimulate the deployment of DNSSEC. Being a strong but demanding cryptographical defense, DNSSEC has its universal adoption predicted to go through a lengthy transition. Thus the DNSSEC practitioners call for a secure yet lightweight solution to speed up DNSSEC deployment while offering an acceptable DNSSEC-like defense. This paper proposes a new defense against cache poisoning attacks, still using but lightly using DNSSEC. In the solution, DNS operates in the DNSSEC-oblivious mode unless a potential attack is detected and triggers a switch to the DNSSEC-aware mode. The performance of the defense is analyzed and validated. The modeling checking results demonstrate that only a small DNSSEC query load is needed to ensure a small enough cache poisoning success rate."
                    },
                    {
                        "title": "The Availability and Security Implications of Glue in the Domain Name System",
                        "abstract": "The Domain Name System (DNS) is one of the most fundamental components of the Internet. While glue is widely used and heavily relied on in DNS operations, there is little thinking about the necessity, complexity, and venerability of such prevalent configuration. This work is the first to provide extensive and systematic analysis of DNS glue. It discusses the availability implications of glue and proposes the minimum glue records in terms of availability. It also identifies the security vulnerabilities of glue as well as the limitations of current countermeasures. Measurements show the wide occurrences of glue redundancies and glue vulnerabilities."
                    },
                    {
                        "title": "Combating Malicious DNS Tunnel",
                        "abstract": "This paper proposes a defense scheme against malicious use of DNS tunnel. A tunnel validator is designed to provide trustworthy tunnel-aware defensive recursive service. In addition to the detection algorithm of malicious tunnel domains, the tunnel validation relies on registered tunnel domains as whitelist and identified malicious tunnel domains as blacklist. A benign tunnel user is thus motivated to register its tunnel domain before using it. Through the tunnel validation, the secure domains are allowed to the recursive service provided by the tunnel validator and the insecure domains are blocked. All inbound suspicious DNS queries are recorded and stored for forensics and future malicious tunnel detection by the tunnel validator."
                    },
                    {
                        "title": "Modeling and Predicting DNS Server Load",
                        "abstract": "The DNS relies on caching to ensure high scalability and good performance. In optimizing caching, TTL adjustment provides a means of balancing between query load and TTL-dependent performances such as data consistency, load balancing, migration time, etc. To gain the desired balance, TTL adjustment depends on predictions of query loads under alternative TTLs. This paper proposes a model of DNS server load, which employs the uniform aggregate caching model to simplify the complexity of modeling clients' requests and their caching. A method of predicting DNS server load is developed using that model. The prediction method is solely based on the unilateral measurements or observations at authoritative servers. Without reliance on lots of multi-point measurements nor distributed measuring facilities, the method is best suited for DNS authoritative operators. The proposed model and prediction method are validated through extensive simulations. Finally, global sensibility analysis is conducted to evaluate the impacts of measurement uncertainties or errors on the predictions."
                    },
                    {
                        "title": "Does Your DNS Recursion Really Time Out as Intended? A Timeout Vulnerability of DNS Recursive Servers",
                        "abstract": "Parallelization is featured by DNS recursive servers to do time-consuming recursions on behalf on clients. As common DNS configurations, recursive servers should allow a reasonable timeout for each recursion which may take as long as several seconds. However, it is proposed in this paper that recursion parallelization may be exploited by attackers to compromise the recursion timeout mechanism for the purpose of DoS or DDoS attacks. Attackers can have recursive servers drop early existing recursions in service by saturating recursion parallelization. The key of the proposed attack model is to reliably prolong service times for any attacking queries. As means of prolong service times, serval techniques are proposed to effectively avoiding cache hit and prolonging overall latency of external DNS lookups respectively. The impacts of saturated recursion parallelization on timeout are analytically provided. The testing on BIND servers demonstrates that with carefully crafted queries, an attacker can use a low or moderate level of query load to successfully overwhelm a target recursive server from serving the legitimate clients."
                    },
                    {
                        "title": "Markov chain Monte Carlo Methods For Lattice Gaussian Sampling:Convergence Analysis and Enhancement",
                        "abstract": "Sampling from lattice Gaussian distribution has emerged as an important problem in coding, decoding and cryptography. In this paper, the classic Gibbs algorithm from Markov chain Monte Carlo (MCMC) methods is demonstrated to be geometrically ergodic for lattice Gaussian sampling, which means the Markov chain arising from it converges exponentially fast to the stationary distribution. Meanwhile, the exponential convergence rate of Markov chain is also derived through the spectral radius of forward operator. Then, a comprehensive analysis regarding to the convergence rate is carried out and two sampling schemes are proposed to further enhance the convergence performance. The first one, referred to as Metropolis-within-Gibbs (MWG) algorithm, improves the convergence by refining the state space of the univariate sampling. On the other hand, the blocked strategy of Gibbs algorithm, which performs the sampling over multivariate at each Markov move, is also shown to yield a better convergence rate than the traditional univariate sampling. In order to perform blocked sampling efficiently, Gibbs-Klein (GK) algorithm is proposed, which samples block by block using Klein's algorithm. Furthermore, the validity of GK algorithm is demonstrated by showing its ergodicity. Simulation results based on MIMO detections are presented to confirm the convergence gain brought by the proposed Gibbs sampling schemes."
                    },
                    {
                        "title": "Analysis and Modeling of Driver Behavior with Integrated Feedback of Visual and Haptic Information Under Shared Control",
                        "abstract": "The thesis presents contributions made to the evaluation and design of a haptic guidance system on improving driving performance in cases of normal and degraded visual information, which are based on behavior experiments, modeling and numerical simulations. The effect of shared control on driver behavior in cases of normal and degraded visual information has been successfully evaluated experimentally and numerically. The evaluation results indicate that the proposed haptic guidance system is capable of providing reliable haptic information, and is effective on improving lane following performance in the conditions of visual occlusion from road ahead and declined visual attention under fatigue driving. Moreover, the appropriate degree of haptic guidance is highly related to the reliability of visual information perceived by the driver, which suggests that designing the haptic guidance system based on the reliability of visual information would allow for greater driver acceptance. Furthermore, the parameterized driver model, which considers the integrated feedback of visual and haptic information, is capable of predicting driver behavior under shared control, and has the potential of being used for designing and evaluating the haptic guidance system."
                    },
                    {
                        "title": "Detecting Algorithmically Generated Domains Using a GCNN-LSTM Hybrid Neural Network",
                        "abstract": "Domain generation algorithm (DGA) is used by botnets to build a stealthy command and control (C&C) communication channel between the C&C server and the bots. A DGA can periodically produce a large number of pseudo-random algorithmically generated domains (AGDs). AGD detection algorithms provide a lightweight, promising solution in response to the existing DGA techniques. In this paper, a GCNN (gated convolutional neural network)-LSTM (long short-term memory) Hybrid Neural Network (GLHNN) for AGD detection is proposed. In GLHNN, GCNN is applied to extract the informative features from domain names on top of LSTM which further processes the feature sequence. GLHNN is experimentally validated using representative AGDs covering six classes of DGAs. GLHNN is compared with the state-of-the-art detection models and demonstrates the best overall detection performance among these tested models."
                    },
                    {
                        "title": "Recent Results of $\u03c8(2S)$ Decays from BES",
                        "abstract": "Results of many $\\psi(2S)$ hadronic decays and the results of $\\psi(2S)$ radiative/hadronic transitions from BES collaboration in the past year are presented. Measurement of $\\BR(\\jpsi\\ra K^0_S K^0_L)$ and the preliminary result of searching for the decay $\\psi(3770)\\ra\\rho\\pi$ are also reported."
                    }
                ]
            },
            "8ab3aee9-ade8-4d60-8544-624dd710b4bd": {
                "pk": "8ab3aee9-ade8-4d60-8544-624dd710b4bd",
                "name": "Atticus Geiger",
                "collaborators": [
                    "Christopher Potts",
                    "Zhengxuan Wu",
                    "Thomas Icard",
                    "Ignacio Cases",
                    "Lauri Karttunen",
                    "Jing Huang",
                    "Karel D'Oosterlinck",
                    "Maheep Chaudhary",
                    "Kyle Richardson",
                    "Chris Potts"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Causal Inference",
                    "Interpretability",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Evaluating Open-Source Sparse Autoencoders on Disentangling Factual Knowledge in GPT-2 Small",
                        "abstract": "A popular new method in mechanistic interpretability is to train high-dimensional sparse autoencoders (SAEs) on neuron activations and use SAE features as the atomic units of analysis. However, the body of evidence on whether SAE feature spaces are useful for causal analysis is underdeveloped. In this work, we use the RAVEL benchmark to evaluate whether SAEs trained on hidden representations of GPT-2 small have sets of features that separately mediate knowledge of which country a city is in and which continent it is in. We evaluate four open-source SAEs for GPT-2 small against each other, with neurons serving as a baseline, and linear features learned via distributed alignment search (DAS) serving as a skyline. For each, we learn a binary mask to select features that will be patched to change the country of a city without changing the continent, or vice versa. Our results show that SAEs struggle to reach the neuron baseline, and none come close to the DAS skyline. We release code here: https://github.com/MaheepChaudhary/SAE-Ravel"
                    },
                    {
                        "title": "Neural Natural Language Inference Models Partially Embed Theories of Lexical Entailment and Negation",
                        "abstract": "We address whether neural models for Natural Language Inference (NLI) can learn the compositional interactions between lexical entailment and negation, using four methods: the behavioral evaluation methods of (1) challenge test sets and (2) systematic generalization tasks, and the structural evaluation methods of (3) probes and (4) interventions. To facilitate this holistic evaluation, we present Monotonicity NLI (MoNLI), a new naturalistic dataset focused on lexical entailment and negation. In our behavioral evaluations, we find that models trained on general-purpose NLI datasets fail systematically on MoNLI examples containing negation, but that MoNLI fine-tuning addresses this failure. In our structural evaluations, we look for evidence that our top-performing BERT-based model has learned to implement the monotonicity algorithm behind MoNLI. Probes yield evidence consistent with this conclusion, and our intervention experiments bolster this, showing that the causal dynamics of the model mirror the causal dynamics of this algorithm on subsets of MoNLI. This suggests that the BERT model at least partially embeds a theory of lexical entailment and negation at an algorithmic level."
                    },
                    {
                        "title": "Posing Fair Generalization Tasks for Natural Language Inference",
                        "abstract": "Deep learning models for semantics are generally evaluated using naturalistic corpora. Adversarial methods, in which models are evaluated on new examples with known semantic properties, have begun to reveal that good performance at these naturalistic tasks can hide serious shortcomings. However, we should insist that these evaluations be fair -that the models are given data sufficient to support the requisite kinds of generalization. In this paper, we define and motivate a formal notion of fairness in this sense. We then apply these ideas to natural language inference by constructing very challenging but provably fair artificial datasets and showing that standard neural models fail to generalize in the required ways; only task-specific models that jointly compose the premise and hypothesis are able to achieve high performance, and even these models do not solve the task perfectly."
                    },
                    {
                        "title": "Causal Abstractions of Neural Networks",
                        "abstract": "Structural analysis methods (e.g., probing and feature attribution) are increasingly important tools for neural network analysis. We propose a new structural analysis method grounded in a formal theory of causal abstraction that provides rich characterizations of model-internal representations and their roles in input/output behavior. In this method, neural representations are aligned with variables in interpretable causal models, and then interchange interventions are used to experimentally verify that the neural representations have the causal properties of their aligned variables. We apply this method in a case study to analyze neural models trained on Multiply Quantified Natural Language Inference (MQNLI) corpus, a highly complex NLI dataset that was constructed with a tree-structured natural logic causal model. We discover that a BERT-based model with state-of-the-art performance successfully realizes parts of the natural logic model's causal structure, whereas a simpler baseline model fails to show any such structure, demonstrating that BERT representations encode the compositional structure of MQNLI."
                    },
                    {
                        "title": "Causal Abstraction with Soft Interventions",
                        "abstract": "Causal abstraction provides a theory describing how several causal models can represent the same system at different levels of detail. Existing theoretical proposals limit the analysis of abstract models to \"hard\" interventions fixing causal variables to be constant values. In this work, we extend causal abstraction to \"soft\" interventions, which assign possibly non-constant functions to variables without adding new causal connections. Specifically, (i) we generalize $\\tau$-abstraction from Beckers and Halpern (2019) to soft interventions, (ii) we propose a further definition of soft abstraction to ensure a unique map $\\omega$ between soft interventions, and (iii) we prove that our constructive definition of soft abstraction guarantees the intervention map $\\omega$ has a specific and necessary explicit form."
                    },
                    {
                        "title": "DynaSent: A Dynamic Benchmark for Sentiment Analysis",
                        "abstract": "We introduce DynaSent ('Dynamic Sentiment'), a new English-language benchmark task for ternary (positive/negative/neutral) sentiment analysis. DynaSent combines naturally occurring sentences with sentences created using the open-source Dynabench Platform, which facilities human-and-model-in-the-loop dataset creation. DynaSent has a total of 121,634 sentences, each validated by five crowdworkers, and its development and test splits are designed to produce chance performance for even the best models we have been able to develop; when future models solve this task, we will use them to create DynaSent version 2, continuing the dynamic evolution of this benchmark. Here, we report on the dataset creation effort, focusing on the steps we took to increase quality and reduce artifacts. We also present evidence that DynaSent's Neutral category is more coherent than the comparable category in other benchmarks, and we motivate training models from scratch for each round over successive fine-tuning."
                    },
                    {
                        "title": "Stress-Testing Neural Models of Natural Language Inference with Multiply-Quantified Sentences",
                        "abstract": "Standard evaluations of deep learning models for semantics using naturalistic corpora are limited in what they can tell us about the fidelity of the learned representations, because the corpora rarely come with good measures of semantic complexity. To overcome this limitation, we present a method for generating data sets of multiply-quantified natural language inference (NLI) examples in which semantic complexity can be precisely characterized, and we use this method to show that a variety of common architectures for NLI inevitably fail to encode crucial information; only a model with forced lexical alignments avoids this damaging information loss."
                    },
                    {
                        "title": "Relational reasoning and generalization using non-symbolic neural networks",
                        "abstract": "The notion of equality (identity) is simple and ubiquitous, making it a key case study for broader questions about the representations supporting abstract relational reasoning. Previous work suggested that neural networks were not suitable models of human relational reasoning because they could not represent mathematically identity, the most basic form of equality. We revisit this question. In our experiments, we assess out-of-sample generalization of equality using both arbitrary representations and representations that have been pretrained on separate tasks to imbue them with structure. We find neural networks are able to learn (1) basic equality (mathematical identity), (2) sequential equality problems (learning ABA-patterned sequences) with only positive training instances, and (3) a complex, hierarchical equality problem with only basic equality training instances (\"zero-shot'\" generalization). In the two latter cases, our models perform tasks proposed in previous work to demarcate human-unique symbolic abilities. These results suggest that essential aspects of symbolic reasoning can emerge from data-driven, non-symbolic learning processes."
                    },
                    {
                        "title": "Rigorously Assessing Natural Language Explanations of Neurons",
                        "abstract": "Natural language is an appealing medium for explaining how large language models process and store information, but evaluating the faithfulness of such explanations is challenging. To help address this, we develop two modes of evaluation for natural language explanations that claim individual neurons represent a concept in a text input. In the observational mode, we evaluate claims that a neuron $a$ activates on all and only input strings that refer to a concept picked out by the proposed explanation $E$. In the intervention mode, we construe $E$ as a claim that the neuron $a$ is a causal mediator of the concept denoted by $E$. We apply our framework to the GPT-4-generated explanations of GPT-2 XL neurons of Bills et al. (2023) and show that even the most confident explanations have high error rates and little to no causal efficacy. We close the paper by critically assessing whether natural language is a good choice for explanations and whether neurons are the best level of analysis."
                    },
                    {
                        "title": "RAVEL: Evaluating Interpretability Methods on Disentangling Language Model Representations",
                        "abstract": "Individual neurons participate in the representation of multiple high-level concepts. To what extent can different interpretability methods successfully disentangle these roles? To help address this question, we introduce RAVEL (Resolving Attribute-Value Entanglements in Language Models), a dataset that enables tightly controlled, quantitative comparisons between a variety of existing interpretability methods. We use the resulting conceptual framework to define the new method of Multi-task Distributed Alignment Search (MDAS), which allows us to find distributed representations satisfying multiple causal criteria. With Llama2-7B as the target language model, MDAS achieves state-of-the-art results on RAVEL, demonstrating the importance of going beyond neuron-level analyses to identify features distributed across activations. We release our benchmark at https://github.com/explanare/ravel."
                    },
                    {
                        "title": "Updating CLIP to Prefer Descriptions Over Captions",
                        "abstract": "Although CLIPScore is a powerful generic metric that captures the similarity between a text and an image, it fails to distinguish between a caption that is meant to complement the information in an image and a description that is meant to replace an image entirely, e.g., for accessibility. We address this shortcoming by updating the CLIP model with the Concadia dataset to assign higher scores to descriptions than captions using parameter efficient fine-tuning and a loss objective derived from work on causal interpretability. This model correlates with the judgements of blind and low-vision people while preserving transfer capabilities and has interpretable structure that sheds light on the caption--description distinction."
                    },
                    {
                        "title": "Recurrent Neural Networks Learn to Store and Generate Sequences using Non-Linear Representations",
                        "abstract": "The Linear Representation Hypothesis (LRH) states that neural networks learn to encode concepts as directions in activation space, and a strong version of the LRH states that models learn only such encodings. In this paper, we present a counterexample to this strong LRH: when trained to repeat an input token sequence, gated recurrent neural networks (RNNs) learn to represent the token at each position with a particular order of magnitude, rather than a direction. These representations have layered features that are impossible to locate in distinct linear subspaces. To show this, we train interventions to predict and manipulate tokens by learning the scaling factor corresponding to each sequence position. These interventions indicate that the smallest RNNs find only this magnitude-based solution, while larger RNNs have linear representations. These findings strongly indicate that interpretability research should not be confined by the LRH."
                    },
                    {
                        "title": "Linear Representations of Sentiment in Large Language Models",
                        "abstract": "Sentiment is a pervasive feature in natural language text, yet it is an open question how sentiment is represented within Large Language Models (LLMs). In this study, we reveal that across a range of models, sentiment is represented linearly: a single direction in activation space mostly captures the feature across a range of tasks with one extreme for positive and the other for negative. Through causal interventions, we isolate this direction and show it is causally relevant in both toy tasks and real world datasets such as Stanford Sentiment Treebank. Through this case study we model a thorough investigation of what a single direction means on a broad data distribution.   We further uncover the mechanisms that involve this direction, highlighting the roles of a small subset of attention heads and neurons. Finally, we discover a phenomenon which we term the summarization motif: sentiment is not solely represented on emotionally charged words, but is additionally summarized at intermediate positions without inherent sentiment, such as punctuation and names. We show that in Stanford Sentiment Treebank zero-shot classification, 76% of above-chance classification accuracy is lost when ablating the sentiment direction, nearly half of which (36%) is due to ablating the summarized sentiment direction exclusively at comma positions."
                    },
                    {
                        "title": "Finding Alignments Between Interpretable Causal Variables and Distributed Neural Representations",
                        "abstract": "Causal abstraction is a promising theoretical framework for explainable artificial intelligence that defines when an interpretable high-level causal model is a faithful simplification of a low-level deep learning system. However, existing causal abstraction methods have two major limitations: they require a brute-force search over alignments between the high-level model and the low-level one, and they presuppose that variables in the high-level model will align with disjoint sets of neurons in the low-level one. In this paper, we present distributed alignment search (DAS), which overcomes these limitations. In DAS, we find the alignment between high-level and low-level models using gradient descent rather than conducting a brute-force search, and we allow individual neurons to play multiple distinct roles by analyzing representations in non-standard bases-distributed representations. Our experiments show that DAS can discover internal structure that prior approaches miss. Overall, DAS removes previous obstacles to conducting causal abstraction analyses and allows us to find conceptual structure in trained neural nets."
                    }
                ]
            },
            "4b966a17-958e-441c-9329-346a633f2bf6": {
                "pk": "4b966a17-958e-441c-9329-346a633f2bf6",
                "name": "Dan Jurafsky",
                "collaborators": [
                    "Jiwei Li",
                    "Will Monroe",
                    "Kawin Ethayarajh",
                    "Dan Iter",
                    "Isabel Papadimitriou",
                    "William L. Hamilton",
                    "Jure Leskovec",
                    "Kelvin Guu",
                    "Larry Lansing",
                    "William Held"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Representation Learning",
                    "Coreference Resolution",
                    "Neural Generation"
                ],
                "publications": [
                    {
                        "title": "Pretraining with Contrastive Sentence Objectives Improves Discourse Performance of Language Models",
                        "abstract": "Recent models for unsupervised representation learning of text have employed a number of techniques to improve contextual word representations but have put little focus on discourse-level representations. We propose CONPONO, an inter-sentence objective for pretraining language models that models discourse coherence and the distance between sentences. Given an anchor sentence, our model is trained to predict the text k sentences away using a sampled-softmax objective where the candidates consist of neighboring sentences and sentences randomly sampled from the corpus. On the discourse representation benchmark DiscoEval, our model improves over the previous state-of-the-art by up to 13% and on average 4% absolute across 7 tasks. Our model is the same size as BERT-Base, but outperforms the much larger BERT- Large model and other more recent approaches that incorporate discourse. We also show that CONPONO yields gains of 2%-6% absolute even for tasks that do not explicitly evaluate discourse: textual entailment (RTE), common sense reasoning (COPA) and reading comprehension (ReCoRD)."
                    },
                    {
                        "title": "Focus on what matters: Applying Discourse Coherence Theory to Cross Document Coreference",
                        "abstract": "Performing event and entity coreference resolution across documents vastly increases the number of candidate mentions, making it intractable to do the full $n^2$ pairwise comparisons. Existing approaches simplify by considering coreference only within document clusters, but this fails to handle inter-cluster coreference, common in many applications. As a result cross-document coreference algorithms are rarely applied to downstream tasks. We draw on an insight from discourse coherence theory: potential coreferences are constrained by the reader's discourse focus. We model the entities/events in a reader's focus as a neighborhood within a learned latent embedding space which minimizes the distance between mentions and the centroids of their gold coreference clusters. We then use these neighborhoods to sample only hard negatives to train a fine-grained classifier on mention pairs and their local discourse features. Our approach achieves state-of-the-art results for both events and entities on the ECB+, Gun Violence, Football Coreference, and Cross-Domain Cross-Document Coreference corpora. Furthermore, training on multiple corpora improves average performance across all datasets by 17.2 F1 points, leading to a robust coreference resolution model for use in downstream tasks where link distribution is unknown."
                    },
                    {
                        "title": "A Simple, Fast Diverse Decoding Algorithm for Neural Generation",
                        "abstract": "In this paper, we propose a simple, fast decoding algorithm that fosters diversity in neural generation. The algorithm modifies the standard beam search algorithm by adding an inter-sibling ranking penalty, favoring choosing hypotheses from diverse parents. We evaluate the proposed model on the tasks of dialogue response generation, abstractive summarization and machine translation. We find that diverse decoding helps across all tasks, especially those for which reranking is needed.   We further propose a variation that is capable of automatically adjusting its diversity decoding rates for different inputs using reinforcement learning (RL). We observe a further performance boost from this RL technique. This paper includes material from the unpublished script \"Mutual Information and Diverse Decoding Improve Neural Machine Translation\" (Li and Jurafsky, 2016)."
                    },
                    {
                        "title": "Do Multi-Sense Embeddings Improve Natural Language Understanding?",
                        "abstract": "Learning a distinct representation for each sense of an ambiguous word could lead to more powerful and fine-grained models of vector-space representations. Yet while `multi-sense' methods have been proposed and tested on artificial word-similarity tasks, we don't know if they improve real natural language understanding tasks. In this paper we introduce a multi-sense embedding model based on Chinese Restaurant Processes that achieves state of the art performance on matching human word similarity judgments, and propose a pipelined architecture for incorporating multi-sense embeddings into language understanding.   We then test the performance of our model on part-of-speech tagging, named entity recognition, sentiment analysis, semantic relation identification and semantic relatedness, controlling for embedding dimensionality. We find that multi-sense embeddings do improve performance on some tasks (part-of-speech tagging, semantic relation identification, semantic relatedness) but not on others (named entity recognition, various forms of sentiment analysis). We discuss how these differences may be caused by the different role of word sense information in each of the tasks. The results highlight the importance of testing embedding models in real applications."
                    },
                    {
                        "title": "Mutual Information and Diverse Decoding Improve Neural Machine Translation",
                        "abstract": "Sequence-to-sequence neural translation models learn semantic and syntactic relations between sentence pairs by optimizing the likelihood of the target given the source, i.e., $p(y|x)$, an objective that ignores other potentially useful sources of information. We introduce an alternative objective function for neural MT that maximizes the mutual information between the source and target sentences, modeling the bi-directional dependency of sources and targets. We implement the model with a simple re-ranking method, and also introduce a decoding algorithm that increases diversity in the N-best list produced by the first pass. Applied to the WMT German/English and French/English tasks, the proposed models offers a consistent performance boost on both standard LSTM and attention-based neural MT architectures."
                    },
                    {
                        "title": "Neural Net Models for Open-Domain Discourse Coherence",
                        "abstract": "Discourse coherence is strongly associated with text quality, making it important to natural language generation and understanding. Yet existing models of coherence focus on measuring individual aspects of coherence (lexical overlap, rhetorical structure, entity centering) in narrow domains.   In this paper, we describe domain-independent neural models of discourse coherence that are capable of measuring multiple aspects of coherence in existing sentences and can maintain coherence while generating new sentences. We study both discriminative models that learn to distinguish coherent from incoherent discourse, and generative models that produce coherent text, including a novel neural latent-variable Markovian generative model that captures the latent discourse dependencies between sentences in a text.   Our work achieves state-of-the-art performance on multiple coherence evaluations, and marks an initial step in generating coherent texts given discourse contexts."
                    },
                    {
                        "title": "Attention Flows are Shapley Value Explanations",
                        "abstract": "Shapley Values, a solution to the credit assignment problem in cooperative game theory, are a popular type of explanation in machine learning, having been used to explain the importance of features, embeddings, and even neurons. In NLP, however, leave-one-out and attention-based explanations still predominate. Can we draw a connection between these different methods? We formally prove that -- save for the degenerate case -- attention weights and leave-one-out values cannot be Shapley Values. $\\textit{Attention flow}$ is a post-processed variant of attention weights obtained by running the max-flow algorithm on the attention graph. Perhaps surprisingly, we prove that attention flows are indeed Shapley Values, at least at the layerwise level. Given the many desirable theoretical qualities of Shapley Values -- which has driven their adoption among the ML community -- we argue that NLP practitioners should, when possible, adopt attention flow explanations alongside more traditional ones."
                    },
                    {
                        "title": "Injecting structural hints: Using language models to study inductive biases in language learning",
                        "abstract": "Both humans and large language models are able to learn language without explicit structural supervision. What inductive biases make this learning possible? We address this fundamental cognitive question by leveraging transformer language models: we inject inductive bias into language models by pretraining on formally-structured data, and then evaluate the biased learners' ability to learn typologically-diverse natural languages. Our experimental setup creates a testbed for hypotheses about inductive bias in human language learning. We investigate the effect of injecting models with three types of inductive bias: 1) recursive, hierarchical processing, 2) crossing token-token relationships that can't be modeled by context-free grammars, and 3) a Zipfian power-law vocabulary distribution. We show that non-context-free relationships form the best inductive biases. Our study leverages the capabilities of transformer models to run controlled language learning experiments that are not possible to run on humans, and surfaces hypotheses about the structures that facilitate language learning in both humans and machines."
                    },
                    {
                        "title": "Citation Classification for Behavioral Analysis of a Scientific Field",
                        "abstract": "Citations are an important indicator of the state of a scientific field, reflecting how authors frame their work, and influencing uptake by future scholars. However, our understanding of citation behavior has been limited to small-scale manual citation analysis. We perform the largest behavioral study of citations to date, analyzing how citations are both framed and taken up by scholars in one entire field: natural language processing. We introduce a new dataset of nearly 2,000 citations annotated for function and centrality, and use it to develop a state-of-the-art classifier and label the entire ACL Reference Corpus. We then study how citations are framed by authors and use both papers and online traces to track how citations are followed by readers. We demonstrate that authors are sensitive to discourse structure and publication venue when citing, that online readers follow temporal links to previous and future work rather than methodological links, and that how a paper cites related work is predictive of its citation count. Finally, we use changes in citation roles to show that the field of NLP is undergoing a significant increase in consensus."
                    },
                    {
                        "title": "Learning Music Helps You Read: Using Transfer to Study Linguistic Structure in Language Models",
                        "abstract": "We propose transfer learning as a method for analyzing the encoding of grammatical structure in neural language models. We train LSTMs on non-linguistic data and evaluate their performance on natural language to assess which kinds of data induce generalizable structural features that LSTMs can use for natural language. We find that training on non-linguistic data with latent structure (MIDI music or Java code) improves test performance on natural language, despite no overlap in surface form or vocabulary. To pinpoint the kinds of abstract structure that models may be encoding to lead to this improvement, we run similar experiments with two artificial parentheses languages: one which has a hierarchical recursive structure, and a control which has paired tokens but no recursion. Surprisingly, training a model on either of these artificial languages leads to the same substantial gains when testing on natural language. Further experiments on transfer between natural languages controlling for vocabulary overlap show that zero-shot performance on a test language is highly correlated with typological syntactic similarity to the training language, suggesting that representations induced by pre-training correspond to the cross-linguistic syntactic properties. Our results provide insights into the ways that neural models represent abstract syntactic structure, and also about the kind of structural inductive biases which allow for natural language acquisition."
                    },
                    {
                        "title": "Utility is in the Eye of the User: A Critique of NLP Leaderboards",
                        "abstract": "Benchmarks such as GLUE have helped drive advances in NLP by incentivizing the creation of more accurate models. While this leaderboard paradigm has been remarkably successful, a historical focus on performance-based evaluation has been at the expense of other qualities that the NLP community values in models, such as compactness, fairness, and energy efficiency. In this opinion paper, we study the divergence between what is incentivized by leaderboards and what is useful in practice through the lens of microeconomic theory. We frame both the leaderboard and NLP practitioners as consumers and the benefit they get from a model as its utility to them. With this framing, we formalize how leaderboards -- in their current form -- can be poor proxies for the NLP community at large. For example, a highly inefficient model would provide less utility to practitioners but not to a leaderboard, since it is a cost that only the former must bear. To allow practitioners to better estimate a model's utility to them, we advocate for more transparency on leaderboards, such as the reporting of statistics that are of practical concern (e.g., model size, energy efficiency, and inference latency)."
                    },
                    {
                        "title": "The Authenticity Gap in Human Evaluation",
                        "abstract": "Human ratings are the gold standard in NLG evaluation. The standard protocol is to collect ratings of generated text, average across annotators, and rank NLG systems by their average scores. However, little consideration has been given as to whether this approach faithfully captures human preferences. Analyzing this standard protocol through the lens of utility theory in economics, we identify the implicit assumptions it makes about annotators. These assumptions are often violated in practice, in which case annotator ratings cease to reflect their preferences. The most egregious violations come from using Likert scales, which provably reverse the direction of the true preference in certain cases. We suggest improvements to the standard protocol to make it more theoretically sound, but even in its improved form, it cannot be used to evaluate open-ended tasks like story generation. For the latter, we propose a new human evaluation protocol called $\\textit{system-level probabilistic assessment}$ (SPA). When human evaluation of stories is done with SPA, we can recover the ordering of GPT-3 models by size, with statistically significant results. However, when human evaluation is done with the standard protocol, less than half of the expected preferences can be recovered (e.g., there is no significant difference between $\\texttt{curie}$ and $\\texttt{davinci}$, despite using a highly powered test)."
                    },
                    {
                        "title": "A Hierarchical Neural Autoencoder for Paragraphs and Documents",
                        "abstract": "Natural language generation of coherent long texts like paragraphs or longer documents is a challenging problem for recurrent networks models. In this paper, we explore an important step toward this generation task: training an LSTM (Long-short term memory) auto-encoder to preserve and reconstruct multi-sentence paragraphs. We introduce an LSTM model that hierarchically builds an embedding for a paragraph from embeddings for sentences and words, then decodes this embedding to reconstruct the original paragraph. We evaluate the reconstructed paragraph using standard metrics like ROUGE and Entity Grid, showing that neural models are able to encode texts in a way that preserve syntactic, semantic, and discourse coherence. While only a first step toward generating coherent text units from neural models, our work has the potential to significantly impact natural language generation and summarization\\footnote{Code for the three models described in this paper can be found at www.stanford.edu/~jiweil/ ."
                    },
                    {
                        "title": "Learning multi-faceted representations of individuals from heterogeneous evidence using neural networks",
                        "abstract": "Inferring latent attributes of people online is an important social computing task, but requires integrating the many heterogeneous sources of information available on the web. We propose learning individual representations of people using neural nets to integrate rich linguistic and network evidence gathered from social media. The algorithm is able to combine diverse cues, such as the text a person writes, their attributes (e.g. gender, employer, education, location) and social relations to other people. We show that by integrating both textual and network evidence, these representations offer improved performance at four important tasks in social media inference on Twitter: predicting (1) gender, (2) occupation, (3) location, and (4) friendships for users. Our approach scales to large datasets and the learned representations can be used as general features in and have the potential to benefit a large number of downstream tasks including link prediction, community detection, or probabilistic reasoning over social networks."
                    },
                    {
                        "title": "Diachronic Word Embeddings Reveal Statistical Laws of Semantic Change",
                        "abstract": "Understanding how words change their meanings over time is key to models of language and cultural evolution, but historical data on meaning is scarce, making theories hard to develop and test. Word embeddings show promise as a diachronic tool, but have not been carefully evaluated. We develop a robust methodology for quantifying semantic change by evaluating word embeddings (PPMI, SVD, word2vec) against known historical changes. We then use this methodology to reveal statistical laws of semantic evolution. Using six historical corpora spanning four languages and two centuries, we propose two quantitative laws of semantic change: (i) the law of conformity---the rate of semantic change scales with an inverse power-law of word frequency; (ii) the law of innovation---independent of frequency, words that are more polysemous have higher rates of semantic change."
                    },
                    {
                        "title": "Cultural Shift or Linguistic Drift? Comparing Two Computational Measures of Semantic Change",
                        "abstract": "Words shift in meaning for many reasons, including cultural factors like new technologies and regular linguistic processes like subjectification. Understanding the evolution of language and culture requires disentangling these underlying causes. Here we show how two different distributional measures can be used to detect two different types of semantic change. The first measure, which has been used in many previous works, analyzes global shifts in a word's distributional semantics, it is sensitive to changes due to regular processes of linguistic drift, such as the semantic generalization of promise (\"I promise.\" -> \"It promised to be exciting.\"). The second measure, which we develop here, focuses on local changes to a word's nearest semantic neighbors; it is more sensitive to cultural shifts, such as the change in the meaning of cell (\"prison cell\" -> \"cell phone\"). Comparing measurements made by these two methods allows researchers to determine whether changes are more cultural or linguistic in nature, a distinction that is essential for work in the digital humanities and historical linguistics."
                    },
                    {
                        "title": "Learning to Decode for Future Success",
                        "abstract": "We introduce a simple, general strategy to manipulate the behavior of a neural decoder that enables it to generate outputs that have specific properties of interest (e.g., sequences of a pre-specified length). The model can be thought of as a simple version of the actor-critic model that uses an interpolation of the actor (the MLE-based token generation policy) and the critic (a value function that estimates the future values of the desired property) for decision making. We demonstrate that the approach is able to incorporate a variety of properties that cannot be handled by standard neural sequence decoders, such as sequence length and backward probability (probability of sources given targets), in addition to yielding consistent improvements in abstractive summarization and machine translation when the property to be optimized is BLEU or ROUGE scores."
                    },
                    {
                        "title": "Word Embeddings Quantify 100 Years of Gender and Ethnic Stereotypes",
                        "abstract": "Word embeddings use vectors to represent words such that the geometry between vectors captures semantic relationship between the words. In this paper, we develop a framework to demonstrate how the temporal dynamics of the embedding can be leveraged to quantify changes in stereotypes and attitudes toward women and ethnic minorities in the 20th and 21st centuries in the United States. We integrate word embeddings trained on 100 years of text data with the U.S. Census to show that changes in the embedding track closely with demographic and occupation shifts over time. The embedding captures global social shifts -- e.g., the women's movement in the 1960s and Asian immigration into the U.S -- and also illuminates how specific adjectives and occupations became more closely associated with certain populations over time. Our framework for temporal analysis of word embedding opens up a powerful new intersection between machine learning and quantitative social science."
                    },
                    {
                        "title": "Sharp Nearby, Fuzzy Far Away: How Neural Language Models Use Context",
                        "abstract": "We know very little about how neural language models (LM) use prior linguistic context. In this paper, we investigate the role of context in an LSTM LM, through ablation studies. Specifically, we analyze the increase in perplexity when prior context words are shuffled, replaced, or dropped. On two standard datasets, Penn Treebank and WikiText-2, we find that the model is capable of using about 200 tokens of context on average, but sharply distinguishes nearby context (recent 50 tokens) from the distant history. The model is highly sensitive to the order of words within the most recent sentence, but ignores word order in the long-range context (beyond 50 tokens), suggesting the distant past is modeled only as a rough semantic field or topic. We further find that the neural caching model (Grave et al., 2017b) especially helps the LSTM to copy words from within this distant context. Overall, our analysis not only provides a better understanding of how neural LMs use their context, but also sheds light on recent success from cache-based models."
                    },
                    {
                        "title": "Understanding Neural Networks through Representation Erasure",
                        "abstract": "While neural networks have been successfully applied to many natural language processing tasks, they come at the cost of interpretability. In this paper, we propose a general methodology to analyze and interpret decisions from a neural model by observing the effects on the model of erasing various parts of the representation, such as input word-vector dimensions, intermediate hidden units, or input words. We present several approaches to analyzing the effects of such erasure, from computing the relative difference in evaluation metrics, to using reinforcement learning to erase the minimum set of input words in order to flip a neural model's decision. In a comprehensive analysis of multiple NLP tasks, including linguistic feature classification, sentence-level sentiment analysis, and document level sentiment aspect prediction, we show that the proposed methodology not only offers clear explanations about neural model decisions, but also provides a way to conduct error analysis on neural models."
                    }
                ]
            },
            "4711c626-7fc0-4bf6-90a0-73ea1fbc648b": {
                "pk": "4711c626-7fc0-4bf6-90a0-73ea1fbc648b",
                "name": "Christopher D. Manning",
                "collaborators": [
                    "Christopher Potts",
                    "Samuel R. Bowman",
                    "Zhengxuan Wu",
                    "R\u00f3bert Csord\u00e1s",
                    "Kevin Clark",
                    "Danqi Chen",
                    "Timothy Dozat",
                    "Atticus Geiger",
                    "Julie Tibshirani",
                    "Peng Qi"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Semantic Parsing",
                    "Coreference Resolution",
                    "Dialogue Systems"
                ],
                "publications": [
                    {
                        "title": "ReCOGS: How Incidental Details of a Logical Form Overshadow an Evaluation of Semantic Interpretation",
                        "abstract": "Compositional generalization benchmarks for semantic parsing seek to assess whether models can accurately compute meanings for novel sentences, but operationalize this in terms of logical form (LF) prediction. This raises the concern that semantically irrelevant details of the chosen LFs could shape model performance. We argue that this concern is realized for the COGS benchmark. COGS poses generalization splits that appear impossible for present-day models, which could be taken as an indictment of those models. However, we show that the negative results trace to incidental features of COGS LFs. Converting these LFs to semantically equivalent ones and factoring out capabilities unrelated to semantic interpretation, we find that even baseline models get traction. A recent variable-free translation of COGS LFs suggests similar conclusions, but we observe this format is not semantically equivalent; it is incapable of accurately representing some COGS meanings. These findings inform our proposal for ReCOGS, a modified version of COGS that comes closer to assessing the target semantic capabilities while remaining very challenging. Overall, our results reaffirm the importance of compositional generalization and careful benchmark task design."
                    },
                    {
                        "title": "Simpler but More Accurate Semantic Dependency Parsing",
                        "abstract": "While syntactic dependency annotations concentrate on the surface or functional structure of a sentence, semantic dependency annotations aim to capture between-word relationships that are more closely related to the meaning of a sentence, using graph-structured representations. We extend the LSTM-based syntactic parser of Dozat and Manning (2017) to train on and generate these graph structures. The resulting system on its own achieves state-of-the-art performance, beating the previous, substantially more complex state-of-the-art system by 0.6% labeled F1. Adding linguistically richer input representations pushes the margin even higher, allowing us to beat it by 1.9% labeled F1."
                    },
                    {
                        "title": "Recursive Neural Networks Can Learn Logical Semantics",
                        "abstract": "Tree-structured recursive neural networks (TreeRNNs) for sentence meaning have been successful for many applications, but it remains an open question whether the fixed-length representations that they learn can support tasks as demanding as logical deduction. We pursue this question by evaluating whether two such models---plain TreeRNNs and tree-structured neural tensor networks (TreeRNTNs)---can correctly learn to identify logical relationships such as entailment and contradiction using these representations. In our first set of experiments, we generate artificial data from a logical grammar and use it to evaluate the models' ability to learn to handle basic relational reasoning, recursive structures, and quantification. We then evaluate the models on the more natural SICK challenge data. Both models perform competitively on the SICK data and generalize well in all three experiments on simulated data, suggesting that they can learn suitable representations for logical inference in natural language."
                    },
                    {
                        "title": "Learning Distributed Word Representations for Natural Logic Reasoning",
                        "abstract": "Natural logic offers a powerful relational conception of meaning that is a natural counterpart to distributed semantic representations, which have proven valuable in a wide range of sophisticated language tasks. However, it remains an open question whether it is possible to train distributed representations to support the rich, diverse logical reasoning captured by natural logic. We address this question using two neural network-based models for learning embeddings: plain neural networks and neural tensor networks. Our experiments evaluate the models' ability to learn the basic algebra of natural logic relations from simulated data and from the WordNet noun graph. The overall positive results are promising for the future of learned distributed representations in the applied modeling of logical semantics."
                    },
                    {
                        "title": "Tree-structured composition in neural networks without tree-structured architectures",
                        "abstract": "Tree-structured neural networks encode a particular tree geometry for a sentence in the network design. However, these models have at best only slightly outperformed simpler sequence-based models. We hypothesize that neural sequence models like LSTMs are in fact able to discover and implicitly use recursive compositional structure, at least for tasks with clear cues to that structure in the data. We demonstrate this possibility using an artificial data task for which recursive compositional structure is crucial, and find an LSTM-based sequence model can indeed learn to exploit the underlying tree structure. However, its performance consistently lags behind that of tree models, even on large training sets, suggesting that tree-structured models are more effective at exploiting recursive structure."
                    },
                    {
                        "title": "Recurrent Neural Networks Learn to Store and Generate Sequences using Non-Linear Representations",
                        "abstract": "The Linear Representation Hypothesis (LRH) states that neural networks learn to encode concepts as directions in activation space, and a strong version of the LRH states that models learn only such encodings. In this paper, we present a counterexample to this strong LRH: when trained to repeat an input token sequence, gated recurrent neural networks (RNNs) learn to represent the token at each position with a particular order of magnitude, rather than a direction. These representations have layered features that are impossible to locate in distinct linear subspaces. To show this, we train interventions to predict and manipulate tokens by learning the scaling factor corresponding to each sequence position. These interventions indicate that the smallest RNNs find only this magnitude-based solution, while larger RNNs have linear representations. These findings strongly indicate that interpretability research should not be confined by the LRH."
                    },
                    {
                        "title": "Robust Logistic Regression using Shift Parameters (Long Version)",
                        "abstract": "Annotation errors can significantly hurt classifier performance, yet datasets are only growing noisier with the increased use of Amazon Mechanical Turk and techniques like distant supervision that automatically generate labels. In this paper, we present a robust extension of logistic regression that incorporates the possibility of mislabelling directly into the objective. Our model can be trained through nearly the same means as logistic regression, and retains its efficiency on high-dimensional datasets. Through named entity recognition experiments, we demonstrate that our approach can provide a significant improvement over the standard model when annotation errors are present."
                    },
                    {
                        "title": "Deep Reinforcement Learning for Mention-Ranking Coreference Models",
                        "abstract": "Coreference resolution systems are typically trained with heuristic loss functions that require careful tuning. In this paper we instead apply reinforcement learning to directly optimize a neural mention-ranking model for coreference evaluation metrics. We experiment with two approaches: the REINFORCE policy gradient algorithm and a reward-rescaled max-margin objective. We find the latter to be more effective, resulting in significant improvements over the current state-of-the-art on the English and Chinese portions of the CoNLL 2012 Shared Task."
                    },
                    {
                        "title": "Arc-swift: A Novel Transition System for Dependency Parsing",
                        "abstract": "Transition-based dependency parsers often need sequences of local shift and reduce operations to produce certain attachments. Correct individual decisions hence require global information about the sentence context and mistakes cause error propagation. This paper proposes a novel transition system, arc-swift, that enables direct attachments between tokens farther apart with a single transition. This allows the parser to leverage lexical information more directly in transition decisions. Hence, arc-swift can achieve significantly better performance with a very small beam size. Our parsers reduce error by 3.7--7.6% relative to those using existing transition systems on the Penn Treebank dependency parsing task and English Universal Dependencies."
                    },
                    {
                        "title": "A large annotated corpus for learning natural language inference",
                        "abstract": "Understanding entailment and contradiction is fundamental to understanding natural language, and inference about entailment and contradiction is a valuable testing ground for the development of semantic representations. However, machine learning research in this area has been dramatically limited by the lack of large-scale resources. To address this, we introduce the Stanford Natural Language Inference corpus, a new, freely available collection of labeled sentence pairs, written by humans doing a novel grounded task based on image captioning. At 570K pairs, it is two orders of magnitude larger than all other resources of its type. This increase in scale allows lexicalized classifiers to outperform some sophisticated existing entailment models, and it allows a neural network-based model to perform competitively on natural language inference benchmarks for the first time."
                    },
                    {
                        "title": "Text to 3D Scene Generation with Rich Lexical Grounding",
                        "abstract": "The ability to map descriptions of scenes to 3D geometric representations has many applications in areas such as art, education, and robotics. However, prior work on the text to 3D scene generation task has used manually specified object categories and language that identifies them. We introduce a dataset of 3D scenes annotated with natural language descriptions and learn from this data how to ground textual descriptions to physical objects. Our method successfully grounds a variety of lexical terms to concrete referents, and we show quantitatively that our method improves 3D scene generation over previous work using purely rule-based methods. We evaluate the fidelity and plausibility of 3D scenes generated with our grounding approach through human judgments. To ease evaluation on this task, we also introduce an automated metric that strongly correlates with human judgments."
                    },
                    {
                        "title": "MoEUT: Mixture-of-Experts Universal Transformers",
                        "abstract": "Previous work on Universal Transformers (UTs) has demonstrated the importance of parameter sharing across layers. By allowing recurrence in depth, UTs have advantages over standard Transformers in learning compositional generalizations, but layer-sharing comes with a practical limitation of parameter-compute ratio: it drastically reduces the parameter count compared to the non-shared model with the same dimensionality. Naively scaling up the layer size to compensate for the loss of parameters makes its computational resource requirements prohibitive. In practice, no previous work has succeeded in proposing a shared-layer Transformer design that is competitive in parameter count-dominated tasks such as language modeling. Here we propose MoEUT (pronounced \"moot\"), an effective mixture-of-experts (MoE)-based shared-layer Transformer architecture, which combines several recent advances in MoEs for both feedforward and attention layers of standard Transformers together with novel layer-normalization and grouping schemes that are specific and crucial to UTs. The resulting UT model, for the first time, slightly outperforms standard Transformers on language modeling tasks such as BLiMP and PIQA, while using significantly less compute and memory."
                    },
                    {
                        "title": "Probabilistic Parsing Using Left Corner Language Models",
                        "abstract": "We introduce a novel parser based on a probabilistic version of a left-corner parser. The left-corner strategy is attractive because rule probabilities can be conditioned on both top-down goals and bottom-up derivations. We develop the underlying theory and explain how a grammar can be induced from analyzed data. We show that the left-corner approach provides an advantage over simple top-down probabilistic context-free grammars in parsing the Wall Street Journal using a grammar induced from the Penn Treebank. We also conclude that the Penn Treebank provides a fairly weak testbed due to the flatness of its bracketings and to the obvious overgeneration and undergeneration of its induced grammar."
                    },
                    {
                        "title": "Syn-QG: Syntactic and Shallow Semantic Rules for Question Generation",
                        "abstract": "Question Generation (QG) is fundamentally a simple syntactic transformation; however, many aspects of semantics influence what questions are good to form. We implement this observation by developing SynQG, a set of transparent syntactic rules leveraging universal dependencies, shallow semantic parsing, lexical resources, and custom rules which transform declarative sentences into question-answer pairs. We utilize PropBank argument descriptions and VerbNet state predicates to incorporate shallow semantic content, which helps generate questions of a descriptive nature and produce inferential and semantically richer questions than existing systems. In order to improve syntactic fluency and eliminate grammatically incorrect questions, we employ back-translation over the output of these syntactic rules. A set of crowd-sourced evaluations shows that our system can generate a larger number of highly grammatical and relevant questions than previous QG systems and that back-translation drastically improves grammaticality at a slight cost of generating irrelevant questions."
                    },
                    {
                        "title": "Mapping Three Decades of Intellectual Change in Academia",
                        "abstract": "Research on the development of science has focused on the creation of multidisciplinary teams. However, while this coming together of people is symmetrical, the ideas, methods, and vocabulary of science have a directional flow. We present a statistical model of the text of dissertation abstracts from 1980 to 2010, revealing for the first time the large-scale flow of language across fields. Results of the analysis include identifying methodological fields that export broadly, emerging topical fields that borrow heavily and expand, and old topical fields that grow insular and retract. Particular findings show a growing split between molecular and ecological forms of biology and a sea change in the humanities and social sciences driven by the rise of gender and ethnic studies."
                    },
                    {
                        "title": "Hindsight: Posterior-guided training of retrievers for improved open-ended generation",
                        "abstract": "Many text generation systems benefit from using a retriever to retrieve passages from a textual knowledge corpus (e.g., Wikipedia) which are then provided as additional context to the generator. For open-ended generation tasks (like generating informative utterances in conversations) many varied passages may be equally relevant and we find that existing methods that jointly train the retriever and generator underperform: the retriever may not find relevant passages even amongst the top-10 and hence the generator may not learn a preference to ground its generated output in them. We propose using an additional guide retriever that is allowed to use the target output and \"in hindsight\" retrieve relevant passages during training. We model the guide retriever after the posterior distribution Q of passages given the input and the target output and train it jointly with the standard retriever and the generator by maximizing the evidence lower bound (ELBo) in expectation over Q. For informative conversations from the Wizard of Wikipedia dataset, with posterior-guided training, the retriever finds passages with higher relevance in the top-10 (23% relative improvement), the generator's responses are more grounded in the retrieved passage (19% relative improvement) and the end-to-end system produces better overall output (6.4% relative improvement)."
                    },
                    {
                        "title": "MQuAKE: Assessing Knowledge Editing in Language Models via Multi-Hop Questions",
                        "abstract": "The information stored in large language models (LLMs) falls out of date quickly, and retraining from scratch is often not an option. This has recently given rise to a range of techniques for injecting new facts through updating model weights. Current evaluation paradigms are extremely limited, mainly validating the recall of edited facts, but changing one fact should cause rippling changes to the model's related beliefs. If we edit the UK Prime Minister to now be Rishi Sunak, then we should get a different answer to Who is married to the British Prime Minister? In this work, we present a benchmark, MQuAKE (Multi-hop Question Answering for Knowledge Editing), comprising multi-hop questions that assess whether edited models correctly answer questions where the answer should change as an entailed consequence of edited facts. While we find that current knowledge-editing approaches can recall edited facts accurately, they fail catastrophically on the constructed multi-hop questions. We thus propose a simple memory-based approach, MeLLo, which stores all edited facts externally while prompting the language model iteratively to generate answers that are consistent with the edited facts. While MQuAKE remains challenging, we show that MeLLo scales well with LLMs (e.g., OpenAI GPT-3.5-turbo) and outperforms previous model editors by a large margin."
                    },
                    {
                        "title": "Improving Coreference Resolution by Learning Entity-Level Distributed Representations",
                        "abstract": "A long-standing challenge in coreference resolution has been the incorporation of entity-level information - features defined over clusters of mentions instead of mention pairs. We present a neural network based coreference system that produces high-dimensional vector representations for pairs of coreference clusters. Using these representations, our system learns when combining clusters is desirable. We train the system with a learning-to-search algorithm that teaches it which local decisions (cluster merges) will lead to a high-scoring final coreference partition. The system substantially outperforms the current state-of-the-art on the English and Chinese portions of the CoNLL 2012 Shared Task dataset despite using few hand-engineered features."
                    },
                    {
                        "title": "A Thorough Examination of the CNN/Daily Mail Reading Comprehension Task",
                        "abstract": "Enabling a computer to understand a document so that it can answer comprehension questions is a central, yet unsolved goal of NLP. A key factor impeding its solution by machine learned systems is the limited availability of human-annotated data. Hermann et al. (2015) seek to solve this problem by creating over a million training examples by pairing CNN and Daily Mail news articles with their summarized bullet points, and show that a neural network can then be trained to give good performance on this task. In this paper, we conduct a thorough examination of this new reading comprehension task. Our primary aim is to understand what depth of language understanding is required to do well on this task. We approach this from one side by doing a careful hand-analysis of a small subset of the problems and from the other by showing that simple, carefully designed systems can obtain accuracies of 73.6% and 76.6% on these two datasets, exceeding current state-of-the-art results by 7-10% and approaching what we believe is the ceiling for performance on this task."
                    },
                    {
                        "title": "A Copy-Augmented Sequence-to-Sequence Architecture Gives Good Performance on Task-Oriented Dialogue",
                        "abstract": "Task-oriented dialogue focuses on conversational agents that participate in user-initiated dialogues on domain-specific topics. In contrast to chatbots, which simply seek to sustain open-ended meaningful discourse, existing task-oriented agents usually explicitly model user intent and belief states. This paper examines bypassing such an explicit representation by depending on a latent neural embedding of state and learning selective attention to dialogue history together with copying to incorporate relevant prior context. We complement recent work by showing the effectiveness of simple sequence-to-sequence neural architectures with a copy mechanism. Our model outperforms more complex memory-augmented models by 7% in per-response generation and is on par with the current state-of-the-art on DSTC2."
                    }
                ]
            },
            "c800ff6b-ae3b-440a-9034-e820afc6ec66": {
                "pk": "c800ff6b-ae3b-440a-9034-e820afc6ec66",
                "name": "Christopher Potts",
                "collaborators": [
                    "Christopher D. Manning",
                    "Samuel R. Bowman",
                    "Atticus Geiger",
                    "Zhengxuan Wu",
                    "Will Monroe",
                    "Dante Everaert",
                    "Noah D. Goodman",
                    "Bruno Godefroy",
                    "Nicholas Dingwall",
                    "Zijian Wang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Deep Learning",
                    "Semantic Parsing",
                    "Model Interpretability"
                ],
                "publications": [
                    {
                        "title": "A case for deep learning in semantics",
                        "abstract": "Pater's target article builds a persuasive case for establishing stronger ties between theoretical linguistics and connectionism (deep learning). This commentary extends his arguments to semantics, focusing in particular on issues of learning, compositionality, and lexical meaning."
                    },
                    {
                        "title": "Modeling Drug-Disease Relations with Linguistic and Knowledge Graph Constraints",
                        "abstract": "FDA drug labels are rich sources of information about drugs and drug-disease relations, but their complexity makes them challenging texts to analyze in isolation. To overcome this, we situate these labels in two health knowledge graphs: one built from precise structured information about drugs and diseases, and another built entirely from a database of clinical narrative texts using simple heuristic methods. We show that Probabilistic Soft Logic models defined over these graphs are superior to text-only and relation-only variants, and that the clinical narratives graph delivers exceptional results with little manual effort. Finally, we release a new dataset of drug labels with annotations for five distinct drug-disease relations."
                    },
                    {
                        "title": "Mittens: An Extension of GloVe for Learning Domain-Specialized Representations",
                        "abstract": "We present a simple extension of the GloVe representation learning model that begins with general-purpose representations and updates them based on data from a specialized domain. We show that the resulting representations can lead to faster learning and better results on a variety of tasks."
                    },
                    {
                        "title": "ReCOGS: How Incidental Details of a Logical Form Overshadow an Evaluation of Semantic Interpretation",
                        "abstract": "Compositional generalization benchmarks for semantic parsing seek to assess whether models can accurately compute meanings for novel sentences, but operationalize this in terms of logical form (LF) prediction. This raises the concern that semantically irrelevant details of the chosen LFs could shape model performance. We argue that this concern is realized for the COGS benchmark. COGS poses generalization splits that appear impossible for present-day models, which could be taken as an indictment of those models. However, we show that the negative results trace to incidental features of COGS LFs. Converting these LFs to semantically equivalent ones and factoring out capabilities unrelated to semantic interpretation, we find that even baseline models get traction. A recent variable-free translation of COGS LFs suggests similar conclusions, but we observe this format is not semantically equivalent; it is incapable of accurately representing some COGS meanings. These findings inform our proposal for ReCOGS, a modified version of COGS that comes closer to assessing the target semantic capabilities while remaining very challenging. Overall, our results reaffirm the importance of compositional generalization and careful benchmark task design."
                    },
                    {
                        "title": "TalkDown: A Corpus for Condescension Detection in Context",
                        "abstract": "Condescending language use is caustic; it can bring dialogues to an end and bifurcate communities. Thus, systems for condescension detection could have a large positive impact. A challenge here is that condescension is often impossible to detect from isolated utterances, as it depends on the discourse and social context. To address this, we present TalkDown, a new labeled dataset of condescending linguistic acts in context. We show that extending a language-only model with representations of the discourse improves performance, and we motivate techniques for dealing with the low rates of condescension overall. We also use our model to estimate condescension rates in various online communities and relate these differences to differing community norms."
                    },
                    {
                        "title": "Learning in the Rational Speech Acts Model",
                        "abstract": "The Rational Speech Acts (RSA) model treats language use as a recursive process in which probabilistic speaker and listener agents reason about each other's intentions to enrich the literal semantics of their language along broadly Gricean lines. RSA has been shown to capture many kinds of conversational implicature, but it has been criticized as an unrealistic model of speakers, and it has so far required the manual specification of a semantic lexicon, preventing its use in natural language processing applications that learn lexical knowledge from data. We address these concerns by showing how to define and optimize a trained statistical classifier that uses the intermediate agents of RSA as hidden layers of representation forming a non-linear activation function. This treatment opens up new application domains and new possibilities for learning effectively from data. We validate the model on a referential expression generation task, showing that the best performance is achieved by incorporating features approximating well-established insights about natural language generation into RSA."
                    },
                    {
                        "title": "GIO: Gradient Information Optimization for Training Dataset Selection",
                        "abstract": "It is often advantageous to train models on a subset of the available train examples, because the examples are of variable quality or because one would like to train with fewer examples, without sacrificing performance. We present Gradient Information Optimization (GIO), a scalable, task-agnostic approach to this data selection problem that requires only a small set of (unlabeled) examples representing a target distribution. GIO begins from a natural, information-theoretic objective that is intractable in practice. Our contribution is in showing that it can be made highly scalable through a simple relaxation of the objective and a highly efficient implementation. In experiments with machine translation, spelling correction, and image recognition, we show that GIO delivers outstanding results with very small train sets. These findings are robust to different representation models and hyperparameters for GIO itself. GIO is task- and domain-agnostic and can be applied out-of-the-box to new datasets and domains. We open source a pip-installable implementation of the algorithm as \"pip install grad-info-opt\"."
                    },
                    {
                        "title": "Recursive Neural Networks Can Learn Logical Semantics",
                        "abstract": "Tree-structured recursive neural networks (TreeRNNs) for sentence meaning have been successful for many applications, but it remains an open question whether the fixed-length representations that they learn can support tasks as demanding as logical deduction. We pursue this question by evaluating whether two such models---plain TreeRNNs and tree-structured neural tensor networks (TreeRNTNs)---can correctly learn to identify logical relationships such as entailment and contradiction using these representations. In our first set of experiments, we generate artificial data from a logical grammar and use it to evaluate the models' ability to learn to handle basic relational reasoning, recursive structures, and quantification. We then evaluate the models on the more natural SICK challenge data. Both models perform competitively on the SICK data and generalize well in all three experiments on simulated data, suggesting that they can learn suitable representations for logical inference in natural language."
                    },
                    {
                        "title": "Learning Distributed Word Representations for Natural Logic Reasoning",
                        "abstract": "Natural logic offers a powerful relational conception of meaning that is a natural counterpart to distributed semantic representations, which have proven valuable in a wide range of sophisticated language tasks. However, it remains an open question whether it is possible to train distributed representations to support the rich, diverse logical reasoning captured by natural logic. We address this question using two neural network-based models for learning embeddings: plain neural networks and neural tensor networks. Our experiments evaluate the models' ability to learn the basic algebra of natural logic relations from simulated data and from the WordNet noun graph. The overall positive results are promising for the future of learned distributed representations in the applied modeling of logical semantics."
                    },
                    {
                        "title": "Tree-structured composition in neural networks without tree-structured architectures",
                        "abstract": "Tree-structured neural networks encode a particular tree geometry for a sentence in the network design. However, these models have at best only slightly outperformed simpler sequence-based models. We hypothesize that neural sequence models like LSTMs are in fact able to discover and implicitly use recursive compositional structure, at least for tasks with clear cues to that structure in the data. We demonstrate this possibility using an artificial data task for which recursive compositional structure is crucial, and find an LSTM-based sequence model can indeed learn to exploit the underlying tree structure. However, its performance consistently lags behind that of tree models, even on large training sets, suggesting that tree-structured models are more effective at exploiting recursive structure."
                    },
                    {
                        "title": "Recurrent Neural Networks Learn to Store and Generate Sequences using Non-Linear Representations",
                        "abstract": "The Linear Representation Hypothesis (LRH) states that neural networks learn to encode concepts as directions in activation space, and a strong version of the LRH states that models learn only such encodings. In this paper, we present a counterexample to this strong LRH: when trained to repeat an input token sequence, gated recurrent neural networks (RNNs) learn to represent the token at each position with a particular order of magnitude, rather than a direction. These representations have layered features that are impossible to locate in distinct linear subspaces. To show this, we train interventions to predict and manipulate tokens by learning the scaling factor corresponding to each sequence position. These interventions indicate that the smallest RNNs find only this magnitude-based solution, while larger RNNs have linear representations. These findings strongly indicate that interpretability research should not be confined by the LRH."
                    },
                    {
                        "title": "Stress-Testing Neural Models of Natural Language Inference with Multiply-Quantified Sentences",
                        "abstract": "Standard evaluations of deep learning models for semantics using naturalistic corpora are limited in what they can tell us about the fidelity of the learned representations, because the corpora rarely come with good measures of semantic complexity. To overcome this limitation, we present a method for generating data sets of multiply-quantified natural language inference (NLI) examples in which semantic complexity can be precisely characterized, and we use this method to show that a variety of common architectures for NLI inevitably fail to encode crucial information; only a model with forced lexical alignments avoids this damaging information loss."
                    },
                    {
                        "title": "DynaSent: A Dynamic Benchmark for Sentiment Analysis",
                        "abstract": "We introduce DynaSent ('Dynamic Sentiment'), a new English-language benchmark task for ternary (positive/negative/neutral) sentiment analysis. DynaSent combines naturally occurring sentences with sentences created using the open-source Dynabench Platform, which facilities human-and-model-in-the-loop dataset creation. DynaSent has a total of 121,634 sentences, each validated by five crowdworkers, and its development and test splits are designed to produce chance performance for even the best models we have been able to develop; when future models solve this task, we will use them to create DynaSent version 2, continuing the dynamic evolution of this benchmark. Here, we report on the dataset creation effort, focusing on the steps we took to increase quality and reduce artifacts. We also present evidence that DynaSent's Neutral category is more coherent than the comparable category in other benchmarks, and we motivate training models from scratch for each round over successive fine-tuning."
                    },
                    {
                        "title": "Learning to Generate Compositional Color Descriptions",
                        "abstract": "The production of color language is essential for grounded language generation. Color descriptions have many challenging properties: they can be vague, compositionally complex, and denotationally rich. We present an effective approach to generating color descriptions using recurrent neural networks and a Fourier-transformed color representation. Our model outperforms previous work on a conditional language modeling task over a large corpus of naturalistic color descriptions. In addition, probing the model's output reveals that it can accurately produce not only basic color terms but also descriptors with non-convex denotations (\"greenish\"), bare modifiers (\"bright\", \"dull\"), and compositional phrases (\"faded teal\") not seen in training."
                    },
                    {
                        "title": "Representing Social Media Users for Sarcasm Detection",
                        "abstract": "We explore two methods for representing authors in the context of textual sarcasm detection: a Bayesian approach that directly represents authors' propensities to be sarcastic, and a dense embedding approach that can learn interactions between the author and the text. Using the SARC dataset of Reddit comments, we show that augmenting a bidirectional RNN with these representations improves performance; the Bayesian approach suffices in homogeneous contexts, whereas the added power of the dense embeddings proves valuable in more diverse ones."
                    },
                    {
                        "title": "An Incremental Iterated Response Model of Pragmatics",
                        "abstract": "Recent Iterated Response (IR) models of pragmatics conceptualize language use as a recursive process in which agents reason about each other to increase communicative efficiency. These models are generally defined over complete utterances. However, there is substantial evidence that pragmatic reasoning takes place incrementally during production and comprehension. We address this with an incremental IR model. We compare the incremental and global versions using computational simulations, and we assess the incremental model against existing experimental data and in the TUNA corpus for referring expression generation, showing that the model can capture phenomena out of reach of global versions."
                    },
                    {
                        "title": "Flexible Model Interpretability through Natural Language Model Editing",
                        "abstract": "Model interpretability and model editing are crucial goals in the age of large language models. Interestingly, there exists a link between these two goals: if a method is able to systematically edit model behavior with regard to a human concept of interest, this editor method can help make internal representations more interpretable by pointing towards relevant representations and systematically manipulating them."
                    },
                    {
                        "title": "Retrieval Augmented Spelling Correction for E-Commerce Applications",
                        "abstract": "The rapid introduction of new brand names into everyday language poses a unique challenge for e-commerce spelling correction services, which must distinguish genuine misspellings from novel brand names that use unconventional spelling. We seek to address this challenge via Retrieval Augmented Generation (RAG). On this approach, product names are retrieved from a catalog and incorporated into the context used by a large language model (LLM) that has been fine-tuned to do contextual spelling correction. Through quantitative evaluation and qualitative error analyses, we find improvements in spelling correction utilizing the RAG framework beyond a stand-alone LLM. We also demonstrate the value of additional finetuning of the LLM to incorporate retrieved context."
                    }
                ]
            }
        }
    },
    "2406.13175": {
        "paper_data": {
            "title": "Sparse High Rank Adapters",
            "url": "http://arxiv.org/abs/2406.13175v1",
            "arxiv_id": "2406.13175",
            "authors": [
                "Kartikeya Bhardwaj",
                "Nilesh Prasad Pandey",
                "Sweta Priyadarshi",
                "Viswanath Ganapathy",
                "Rafael Esteves",
                "Shreya Kadambi",
                "Shubhankar Borse",
                "Paul Whatmough",
                "Risheek Garrepalli",
                "Mart Van Baalen",
                "Harris Teague",
                "Markus Nagel"
            ],
            "abstract": "Low Rank Adaptation (LoRA) has gained massive attention in the recent generative AI research. One of the main advantages of LoRA is its ability to be fused with pretrained models adding no overhead during inference. However, from a mobile deployment standpoint, we can either avoid inference overhead in the fused mode but lose the ability to switch adapters rapidly, or suffer significant (up to 30% higher) inference latency while enabling rapid switching in the unfused mode. LoRA also exhibits concept-loss when multiple adapters are used concurrently. In this paper, we propose Sparse High Rank Adapters (SHiRA), a new paradigm which incurs no inference overhead, enables rapid switching, and significantly reduces concept-loss. Specifically, SHiRA can be trained by directly tuning only 1-2% of the base model weights while leaving others unchanged. This results in a highly sparse adapter which can be switched directly in the fused mode. We further provide theoretical and empirical insights on how high sparsity in SHiRA can aid multi-adapter fusion by reducing concept loss. Our extensive experiments on LVMs and LLMs demonstrate that finetuning only a small fraction of the parameters in the base model is sufficient for many tasks while enabling both rapid switching and multi-adapter fusion. Finally, we provide a latency- and memory-efficient SHiRA implementation based on Parameter-Efficient Finetuning (PEFT) Library. This implementation trains at nearly the same speed as LoRA while consuming lower peak GPU memory, thus making SHiRA easy to adopt for practical use cases.",
            "introduction": "   1 Introduction  Low Rank Adaptation (LoRA)\u00a0[11] is an established technique to tune the behavior of large generative models such as Large Language Models (LLMs)\u00a0[28, 27] and Stable Diffusion\u00a0[22, 20]. As the name suggests, LoRA requires very few parameters since it trains low rank projection weights that consume very low memory during the finetuning process while producing excellent results. Moreover, these low rank weights can be fused analytically into the base model weights, thereby incurring no additional overhead during inference.   Despite its success, there are still several limitations of low rank adaptation methods. First, if LoRA parameters are fused into the corresponding pretrained base model weights, they modify the entire weight tensor. Therefore, deploying LoRA on large models such as LLaMA-1/2 (7B+ parameters) or Stable Diffusion (1.5B+ parameters) on mobile devices would require changing a large number of weights during inference. Consequently, for mobile scenarios, if an application requires rapid adapter switching, existing low rank methods would incur a significant memory and latency cost. This is a major deployment challenge because, unlike large GPUs, local memory of small AI accelerators is limited and cannot store all weights at the same time. These challenges can be partially addressed by running LoRA in unfused mode; however, unfused inference can incur as high as \ud835\udfd1\ud835\udfce%percent30\\bm{30\\%}bold_30 bold_% additional latency compared to the base model\u00a0[1] (see section\u00a02.1 for details). Second, LoRA has a well-known limitation called concept loss when using multiple concurrent adapters, e.g., combining multiple style transfer adapters, etc. Specifically, it has been well documented\u00a0[32, 24, 7] that a simple additive merging of multiple LoRA adapters leads to concept loss of one or more adapters. Finally, recent literature also contributes important theoretical and empirical knowledge towards the value of high rank adapters. For instance, Kalajdzievski\u00a0[14] shows that the high rank adapters can greatly outperform low rank adapters when used with correct scaling factors. This calls for further investigation into whether other high rank adapters would significantly outperform LoRA.   Figure 1: Sparse High Rank Adapters (SHiRA): Changing about 1111-2%percent22\\%2 % weights of the pretrained generative model is often sufficient to achieve high performance. Due to its extreme sparsity, SHiRA enables rapid switching and also reduced concept loss during multi-adapter fusion. In contrast, LoRA modifies majority of parameters when fused, prohibiting rapid switching on mobile devices and also experiences concept loss during multi-adapter fusion. For LoRA, elephant for single \u201cpaintings\u201d adapter case has artifacts (extra/broken tusks), and bird/knight for multi-adapter case lose \u201cpaintings\u201d concept and keep only the \u201cblue fire\u201d effects. SHiRA does not experience these issues.   In view of the above, we address the following key problems in this paper: (i)\u00a0How can we perform rapid switching for fused adapters? (ii)\u00a0Is there a simpler solution for multi-adapter fusion to reduce concept loss? (iii)\u00a0Can we build high rank adapters that have high expressive power without significantly increasing the training or inference costs?   To this end, we propose Sparse High Rank Adapters (SHiRA), a single solution to all three problems above. SHiRA is a highly sparse but a high rank adapter which relies on training only a very small subset of parameters from the original",
            "references": [
                {
                    "title": "LoRA+: Efficient Low Rank Adaptation of Large Models",
                    "abstract": "In this paper, we show that Low Rank Adaptation (LoRA) as originally introduced in Hu et al. (2021) leads to suboptimal finetuning of models with large width (embedding dimension). This is due to the fact that adapter matrices A and B in LoRA are updated with the same learning rate. Using scaling arguments for large width networks, we demonstrate that using the same learning rate for A and B does not allow efficient feature learning. We then show that this suboptimality of LoRA can be corrected simply by setting different learning rates for the LoRA adapter matrices A and B with a well-chosen ratio. We call this proposed algorithm LoRA$+$. In our extensive experiments, LoRA$+$ improves performance (1-2 $\\%$ improvements) and finetuning speed (up to $\\sim$ 2X SpeedUp), at the same computational cost as LoRA."
                },
                {
                    "title": "DoRA: Weight-Decomposed Low-Rank Adaptation",
                    "abstract": "Among the widely used parameter-efficient fine-tuning (PEFT) methods, LoRA and its variants have gained considerable popularity because of avoiding additional inference costs. However, there still often exists an accuracy gap between these methods and full fine-tuning (FT). In this work, we first introduce a novel weight decomposition analysis to investigate the inherent differences between FT and LoRA. Aiming to resemble the learning capacity of FT from the findings, we propose Weight-Decomposed Low-Rank Adaptation (DoRA). DoRA decomposes the pre-trained weight into two components, magnitude and direction, for fine-tuning, specifically employing LoRA for directional updates to efficiently minimize the number of trainable parameters. By employing \\ours, we enhance both the learning capacity and training stability of LoRA while avoiding any additional inference overhead. \\ours~consistently outperforms LoRA on fine-tuning LLaMA, LLaVA, and VL-BART on various downstream tasks, such as commonsense reasoning, visual instruction tuning, and image/video-text understanding. Code is available at https://github.com/NVlabs/DoRA."
                },
                {
                    "title": "RoSA: Accurate Parameter-Efficient Fine-Tuning via Robust Adaptation",
                    "abstract": "We investigate parameter-efficient fine-tuning (PEFT) methods that can provide good accuracy under limited computational and memory budgets in the context of large language models (LLMs). We present a new PEFT method called Robust Adaptation (RoSA) inspired by robust principal component analysis that jointly trains $\\textit{low-rank}$ and $\\textit{highly-sparse}$ components on top of a set of fixed pretrained weights to efficiently approximate the performance of a full-fine-tuning (FFT) solution. Across a series of challenging generative tasks such as grade-school math and SQL query generation, which require fine-tuning for good performance, we show that RoSA outperforms LoRA, pure sparse fine-tuning, and alternative hybrid methods at the same parameter budget, and can even recover the performance of FFT on some tasks. We provide system support for RoSA to complement the training algorithm, specifically in the form of sparse GPU kernels which enable memory- and computationally-efficient training, and show that it is also compatible with low-precision base weights, resulting in the first joint representation combining quantization, low-rank and sparse approximations. Our code is available at https://github.com/IST-DASLab/RoSA."
                },
                {
                    "title": "A Rank Stabilization Scaling Factor for Fine-Tuning with LoRA",
                    "abstract": "As large language models (LLMs) have become increasingly compute and memory intensive, parameter-efficient fine-tuning (PEFT) methods are now a common strategy to fine-tune LLMs. A popular PEFT method is Low-Rank Adapters (LoRA), which adds trainable low-rank\"adapters\"to selected layers. Each adapter consists of a low-rank matrix product, multiplicatively scaled by a rank-dependent factor. This scaling factor, which divides adapters by a factor of the rank, results in slowed learning and stunted performance for LoRA with higher-rank adapters. Consequently, the use of LoRA in practice has generally been limited to very low ranks. In this work, we study the impact of the scaling factor on the learning process and prove that LoRA adapters should be divided by a factor of the square root of the rank. Modifying LoRA with the appropriate scaling factor, which we call the rank-stabilized LoRA (rsLoRA) method, easily provides for a fine-tuning compute/performance trade-off, where larger ranks can be used to trade off increased computational resources during training for better fine-tuning performance, with no change in inference computing cost."
                },
                {
                    "title": "ZipLoRA: Any Subject in Any Style by Effectively Merging LoRAs",
                    "abstract": "Methods for finetuning generative models for concept-driven personalization generally achieve strong results for subject-driven or style-driven generation. Recently, low-rank adaptations (LoRA) have been proposed as a parameter-efficient way of achieving concept-driven personalization. While recent work explores the combination of separate LoRAs to achieve joint generation of learned styles and subjects, existing techniques do not reliably address the problem; they often compromise either subject fidelity or style fidelity. We propose ZipLoRA, a method to cheaply and effectively merge independently trained style and subject LoRAs in order to achieve generation of any user-provided subject in any user-provided style. Experiments on a wide range of subject and style combinations show that ZipLoRA can generate compelling results with meaningful improvements over baselines in subject and style fidelity while preserving the ability to recontextualize. Project page: https://ziplora.github.io"
                },
                {
                    "title": "Sparse Low-rank Adaptation of Pre-trained Language Models",
                    "abstract": "Fine-tuning pre-trained large language models in a parameter-efficient manner is widely studied for its effectiveness and efficiency. The popular method of low-rank adaptation (LoRA) offers a notable approach, hypothesizing that the adaptation process is intrinsically low-dimensional. Although LoRA has demonstrated commendable performance, it is implemented with a fixed and unalterable intrinsic rank that might not always be the ideal choice. Recognizing the need for more flexible adaptation, we extend the methodology of LoRA to an innovative approach we call sparse low-rank adaptation (SoRA) that enables dynamic adjustments to the intrinsic rank during the adaptation process. We achieve this through the incorporation of a gate unit optimized with proximal gradient method in the training stage, controlling the cardinality of rank under the sparsity of the gate. In the subsequent inference stage, we eliminate the parameter blocks corresponding to the zeroed-out ranks, to reduce each SoRA module back to a concise yet rank-optimal LoRA. Our approach strengthens the representation power of LoRA by initializing it with a higher rank, while efficiently taming a temporarily increased number of parameters via updating in a sparse way. We further introduce a sparsifying scheduler for SoRA, aiming to examine the impact of the number of non-zero parameters on the model's memorization and generalization. Our experimental results demonstrate that SoRA can outperform other baselines even with 70% retained parameters and 70% training time."
                },
                {
                    "title": "Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch",
                    "abstract": "In this paper, we unveil that Language Models (LMs) can acquire new capabilities by assimilating parameters from homologous models without retraining or GPUs. We first introduce DARE to set most delta parameters (i.e., the disparity between fine-tuned and pre-trained parameters) to zeros without affecting the abilities of Supervised Fine-Tuning (SFT) LMs, which randomly Drops delta parameters with a ratio $p$ And REscales the remaining ones by $1 / (1 - p)$ to approximate the original embeddings. Then, we use DARE as a versatile plug-in to sparsify delta parameters of multiple SFT homologous models for mitigating parameter interference and merge them into a single model by parameter fusing. We experiment with encoder- and decoder-based LMs, showing that: (1) SFT delta parameter value ranges are typically small (within 0.002) with extreme redundancy, and DARE can effortlessly eliminate 90% or even 99% of them; (2) DARE can merge multiple task-specific LMs into one LM with diverse capabilities. Notably, this phenomenon is more pronounced in large-scale LMs, where the merged LM reveals the potential to surpass the performance of any source LM, providing a new discovery. We also utilize DARE to create a merged LM that ranks first among models with 7 billion parameters on the Open LLM Leaderboard."
                },
                {
                    "title": "Orthogonal Subspace Learning for Language Model Continual Learning",
                    "abstract": "Benefiting from massive corpora and advanced hardware, large language models (LLMs) exhibit remarkable capabilities in language understanding and generation. However, their performance degrades in scenarios where multiple tasks are encountered sequentially, also known as catastrophic forgetting. In this paper, we propose orthogonal low-rank adaptation (O-LoRA), a simple and efficient approach for continual learning in language models, effectively mitigating catastrophic forgetting while learning new tasks. Specifically, O-LoRA learns tasks in different (low-rank) vector subspaces that are kept orthogonal to each other in order to minimize interference. Our method induces only marginal additional parameter costs and requires no user data storage for replay. Experimental results on continual learning benchmarks show that our method outperforms state-of-the-art methods. Furthermore, compared to previous approaches, our method excels in preserving the generalization ability of LLMs on unseen tasks."
                },
                {
                    "title": "VeRA: Vector-based Random Matrix Adaptation",
                    "abstract": "Low-rank adapation (LoRA) is a popular method that reduces the number of trainable parameters when finetuning large language models, but still faces acute storage challenges when scaling to even larger models or deploying numerous per-user or per-task adapted models. In this work, we present Vector-based Random Matrix Adaptation (VeRA), which significantly reduces the number of trainable parameters compared to LoRA, yet maintains the same performance. It achieves this by using a single pair of low-rank matrices shared across all layers and learning small scaling vectors instead. We demonstrate its effectiveness on the GLUE and E2E benchmarks, image classification tasks, and show its application in instruction-tuning of 7B and 13B language models."
                },
                {
                    "title": "LoRA-FA: Memory-efficient Low-rank Adaptation for Large Language Models Fine-tuning",
                    "abstract": "The low-rank adaptation (LoRA) method can largely reduce the amount of trainable parameters for fine-tuning large language models (LLMs), however, it still requires expensive activation memory to update low-rank weights. Reducing the number of LoRA layers or using activation recomputation could harm the fine-tuning performance or increase the computational overhead. In this work, we present LoRA-FA, a memory-efficient fine-tuning method that reduces the activation memory without performance degradation and expensive recomputation. LoRA-FA chooses to freeze the projection-down weight of $A$ and update the projection-up weight of $B$ in each LoRA layer. It ensures the change of model weight reside in a low-rank space during LLMs fine-tuning, while eliminating the requirement to store full-rank input activations. We conduct extensive experiments across multiple model types (RoBERTa, T5, LLaMA) and model scales. Our results show that LoRA-FA can always achieve close fine-tuning accuracy across different tasks compared to full parameter fine-tuning and LoRA. Furthermore, LoRA-FA can reduce the overall memory cost by up to 1.4$\\times$ compared to LoRA."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis",
                    "abstract": "We present SDXL, a latent diffusion model for text-to-image synthesis. Compared to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet backbone: The increase of model parameters is mainly due to more attention blocks and a larger cross-attention context as SDXL uses a second text encoder. We design multiple novel conditioning schemes and train SDXL on multiple aspect ratios. We also introduce a refinement model which is used to improve the visual fidelity of samples generated by SDXL using a post-hoc image-to-image technique. We demonstrate that SDXL shows drastically improved performance compared the previous versions of Stable Diffusion and achieves results competitive with those of black-box state-of-the-art image generators. In the spirit of promoting open research and fostering transparency in large model training and evaluation, we provide access to code and model weights at https://github.com/Stability-AI/generative-models"
                },
                {
                    "title": "A Simple and Effective Pruning Approach for Large Language Models",
                    "abstract": "As their size increases, Large Languages Models (LLMs) are natural candidates for network pruning methods: approaches that drop a subset of network weights while striving to preserve performance. Existing methods, however, require either retraining, which is rarely affordable for billion-scale LLMs, or solving a weight reconstruction problem reliant on second-order information, which may also be computationally expensive. In this paper, we introduce a novel, straightforward yet effective pruning method, termed Wanda (Pruning by Weights and activations), designed to induce sparsity in pretrained LLMs. Motivated by the recent observation of emergent large magnitude features in LLMs, our approach prunes weights with the smallest magnitudes multiplied by the corresponding input activations, on a per-output basis. Notably, Wanda requires no retraining or weight update, and the pruned LLM can be used as is. We conduct a thorough evaluation of our method Wanda on LLaMA and LLaMA-2 across various language benchmarks. Wanda significantly outperforms the established baseline of magnitude pruning and performs competitively against recent method involving intensive weight update. Code is available at https://github.com/locuslab/wanda."
                },
                {
                    "title": "Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis",
                    "abstract": "Recent text-to-image generative models can generate high-fidelity images from text inputs, but the quality of these generated images cannot be accurately evaluated by existing evaluation metrics. To address this issue, we introduce Human Preference Dataset v2 (HPD v2), a large-scale dataset that captures human preferences on images from a wide range of sources. HPD v2 comprises 798,090 human preference choices on 433,760 pairs of images, making it the largest dataset of its kind. The text prompts and images are deliberately collected to eliminate potential bias, which is a common issue in previous datasets. By fine-tuning CLIP on HPD v2, we obtain Human Preference Score v2 (HPS v2), a scoring model that can more accurately predict human preferences on generated images. Our experiments demonstrate that HPS v2 generalizes better than previous metrics across various image distributions and is responsive to algorithmic improvements of text-to-image generative models, making it a preferable evaluation metric for these models. We also investigate the design of the evaluation prompts for text-to-image generative models, to make the evaluation stable, fair and easy-to-use. Finally, we establish a benchmark for text-to-image generative models using HPS v2, which includes a set of recent text-to-image models from the academic, community and industry. The code and dataset is available at https://github.com/tgxs002/HPSv2 ."
                },
                {
                    "title": "Mix-of-Show: Decentralized Low-Rank Adaptation for Multi-Concept Customization of Diffusion Models",
                    "abstract": "Public large-scale text-to-image diffusion models, such as Stable Diffusion, have gained significant attention from the community. These models can be easily customized for new concepts using low-rank adaptations (LoRAs). However, the utilization of multiple concept LoRAs to jointly support multiple customized concepts presents a challenge. We refer to this scenario as decentralized multi-concept customization, which involves single-client concept tuning and center-node concept fusion. In this paper, we propose a new framework called Mix-of-Show that addresses the challenges of decentralized multi-concept customization, including concept conflicts resulting from existing single-client LoRA tuning and identity loss during model fusion. Mix-of-Show adopts an embedding-decomposed LoRA (ED-LoRA) for single-client tuning and gradient fusion for the center node to preserve the in-domain essence of single concepts and support theoretically limitless concept fusion. Additionally, we introduce regionally controllable sampling, which extends spatially controllable sampling (e.g., ControlNet and T2I-Adaptor) to address attribute binding and missing object problems in multi-concept sampling. Extensive experiments demonstrate that Mix-of-Show is capable of composing multiple customized concepts with high fidelity, including characters, objects, and scenes."
                },
                {
                    "title": "LLM-Adapters: An Adapter Family for Parameter-Efficient Fine-Tuning of Large Language Models",
                    "abstract": "The success of large language models (LLMs), like GPT-4 and ChatGPT, has led to the development of numerous cost-effective and accessible alternatives that are created by finetuning open-access LLMs with task-specific data (e.g., ChatDoctor) or instruction data (e.g., Alpaca). Among the various fine-tuning methods, adapter-based parameter-efficient fine-tuning (PEFT) is undoubtedly one of the most attractive topics, as it only requires fine-tuning a few external parameters instead of the entire LLMs while achieving comparable or even better performance. To enable further research on PEFT methods of LLMs, this paper presents LLM-Adapters, an easy-to-use framework that integrates various adapters into LLMs and can execute these adapter-based PEFT methods of LLMs for different tasks. The framework includes state-of-the-art open-access LLMs such as LLaMA, BLOOM, and GPT-J, as well as widely used adapters such as Series adapters, Parallel adapter, Prompt-based learning and Reparametrization-based methods. Moreover, we conduct extensive empirical studies on the impact of adapter types, placement locations, and hyper-parameters to the best design for each adapter-based methods. We evaluate the effectiveness of the adapters on fourteen datasets from two different reasoning tasks, Arithmetic Reasoning and Commonsense Reasoning. The results demonstrate that using adapter-based PEFT in smaller-scale LLMs (7B) with few extra trainable parameters yields comparable, and in some cases superior, performance to powerful LLMs (175B) in zero-shot inference on both reasoning tasks."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "SparseAdapter: An Easy Approach for Improving the Parameter-Efficiency of Adapters",
                    "abstract": "Adapter Tuning, which freezes the pretrained language models (PLMs) and only fine-tunes a few extra modules, becomes an appealing efficient alternative to the full model fine-tuning. Although computationally efficient, the recent Adapters often increase parameters (e.g. bottleneck dimension) for matching the performance of full model fine-tuning, which we argue goes against their original intention. In this work, we re-examine the parameter-efficiency of Adapters through the lens of network pruning (we name such plug-in concept as \\texttt{SparseAdapter}) and find that SparseAdapter can achieve comparable or better performance than standard Adapters when the sparse ratio reaches up to 80\\%. Based on our findings, we introduce an easy but effective setting ``\\textit{Large-Sparse}'' to improve the model capacity of Adapters under the same parameter budget. Experiments on five competitive Adapters upon three advanced PLMs show that with proper sparse method (e.g. SNIP) and ratio (e.g. 40\\%) SparseAdapter can consistently outperform their corresponding counterpart. Encouragingly, with the \\textit{Large-Sparse} setting, we can obtain further appealing gains, even outperforming the full fine-tuning by a large margin. Our code will be released at: https://github.com/Shwai-He/SparseAdapter."
                },
                {
                    "title": "DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation",
                    "abstract": "Large text-to-image models achieved a remarkable leap in the evolution of AI, enabling high-quality and diverse synthesis of images from a given text prompt. However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts. In this work, we present a new approach for \u201cpersonalization\u201d of text-to-image diffusion models. Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject. Once the subject is embedded in the output domain of the model, the unique identifier can be used to synthesize novel photorealistic images of the subject contextualized in different scenes. By leveraging the semantic prior embedded in the model with a new autogenous class-specific prior preservation loss, our technique enables synthesizing the subject in diverse scenes, poses, views and lighting conditions that do not appear in the reference images. We apply our technique to several previously-unassailable tasks, including subject recontextualization, text-guided view synthesis, and artistic rendering, all while preserving the subject's key features. We also provide a new dataset and evaluation protocol for this new task of subject-driven generation. Project page: https://dreambooth.github.io/"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Training Neural Networks with Fixed Sparse Masks",
                    "abstract": "During typical gradient-based training of deep neural networks, all of the model's parameters are updated at each iteration. Recent work has shown that it is possible to update only a small subset of the model's parameters during training, which can alleviate storage and communication requirements. In this paper, we show that it is possible to induce a fixed sparse mask on the model's parameters that selects a subset to update over many iterations. Our method constructs the mask out of the $k$ parameters with the largest Fisher information as a simple approximation as to which parameters are most important for the task at hand. In experiments on parameter-efficient transfer learning and distributed training, we show that our approach matches or exceeds the performance of other methods for training with sparse updates while being more efficient in terms of memory usage and communication costs. We release our code publicly to promote further applications of our approach."
                },
                {
                    "title": "Composable Sparse Fine-Tuning for Cross-Lingual Transfer",
                    "abstract": "Fine-tuning the entire set of parameters of a large pretrained model has become the mainstream approach for transfer learning. To increase its efficiency and prevent catastrophic forgetting and interference, techniques like adapters and sparse fine-tuning have been developed. Adapters are modular, as they can be combined to adapt a model towards different facets of knowledge (e.g., dedicated language and/or task adapters). Sparse fine-tuning is expressive, as it controls the behavior of all model components. In this work, we introduce a new fine-tuning method with both these desirable properties. In particular, we learn sparse, real-valued masks based on a simple variant of the Lottery Ticket Hypothesis. Task-specific masks are obtained from annotated data in a source language, and language-specific masks from masked language modeling in a target language. Both these masks can then be composed with the pretrained model. Unlike adapter-based fine-tuning, this method neither increases the number of parameters at inference time nor alters the original model architecture. Most importantly, it outperforms adapters in zero-shot cross-lingual transfer by a large margin in a series of multilingual benchmarks, including Universal Dependencies, MasakhaNER, and AmericasNLI. Based on an in-depth analysis, we additionally find that sparsity is crucial to prevent both 1) interference between the fine-tunings to be composed and 2) overfitting. We release the code and models at https://github.com/cambridgeltl/composable-sft."
                },
                {
                    "title": "Raise a Child in Large Language Model: Towards Effective and Generalizable Fine-tuning",
                    "abstract": "Recent pretrained language models extend from millions to billions of parameters. Thus the need to fine-tune an extremely large pretrained model with a limited training corpus arises in various downstream tasks. In this paper, we propose a straightforward yet effective fine-tuning technique, Child-Tuning, which updates a subset of parameters (called child network) of large pretrained models via strategically masking out the gradients of the non-child network during the backward process. Experiments on various downstream tasks in GLUE benchmark show that Child-Tuning consistently outperforms the vanilla fine-tuning by 1.5 8.6 average score among four different pretrained models, and surpasses the prior fine-tuning techniques by 0.6 1.3 points. Furthermore, empirical results on domain transfer and task transfer show that Child-Tuning can obtain better generalization performance by large margins."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Parameter-Efficient Transfer Learning with Diff Pruning",
                    "abstract": "The large size of pretrained networks makes them difficult to deploy for multiple tasks in storage-constrained settings. Diff pruning enables parameter-efficient transfer learning that scales well with new tasks. The approach learns a task-specific \u201cdiff\u201d vector that extends the original pretrained parameters. This diff vector is adaptively pruned during training with a differentiable approximation to the L0-norm penalty to encourage sparsity. As the number of tasks increases, diff pruning remains parameter-efficient, as it requires storing only a small diff vector for each task. Since it does not require access to all tasks during training, it is attractive in on-device deployment settings where tasks arrive in stream or even from different providers. Diff pruning can match the performance of finetuned baselines on the GLUE benchmark while only modifying 0.5% of the pretrained model\u2019s parameters per task and scales favorably in comparison to popular pruning approaches."
                },
                {
                    "title": "Layer-adaptive Sparsity for the Magnitude-based Pruning",
                    "abstract": "Recent discoveries on neural network pruning reveal that, with a carefully chosen layerwise sparsity, a simple magnitude-based pruning achieves state-of-the-art tradeoff between sparsity and performance. However, without a clear consensus on\"how to choose,\"the layerwise sparsities are mostly selected algorithm-by-algorithm, often resorting to handcrafted heuristics or an extensive hyperparameter search. To fill this gap, we propose a novel importance score for global pruning, coined layer-adaptive magnitude-based pruning (LAMP) score; the score is a rescaled version of weight magnitude that incorporates the model-level $\\ell_2$ distortion incurred by pruning, and does not require any hyperparameter tuning or heavy computation. Under various image classification setups, LAMP consistently outperforms popular existing schemes for layerwise sparsity selection. Furthermore, we observe that LAMP continues to outperform baselines even in weight-rewinding setups, while the connectivity-oriented layerwise sparsity (the strongest baseline overall) performs worse than a simple global magnitude-based pruning in this case. Code: https://github.com/jaeho-lee/layer-adaptive-sparsity"
                },
                {
                    "title": "Masking as an Efficient Alternative to Finetuning for Pretrained Language Models",
                    "abstract": "We present an efficient method of utilizing pretrained language models, where we learn selective binary masks for pretrained weights in lieu of modifying them through finetuning. Extensive evaluations of masking BERT and RoBERTa on a series of NLP tasks show that our masking scheme yields performance comparable to finetuning, yet has a much smaller memory footprint when several tasks need to be inferred simultaneously. Through intrinsic evaluations, we show that representations computed by masked language models encode information necessary for solving downstream tasks. Analyzing the loss landscape, we show that masking and finetuning produce models that reside in minima that can be connected by a line segment with nearly constant test accuracy. This confirms that masking can be utilized as an efficient alternative to finetuning."
                },
                {
                    "title": "Mathematics for Machine Learning",
                    "abstract": "Machine learning is a way to study the algorithm and statistical model that is used by computer to perform a specific task through pattern and deduction [1]. It builds a mathematical model from a sample data which may come under either supervised or unsupervised learning. It is closely\n related to computational statistics which is an interface between statistics and computer science. Also, linear algebra and probability theory are two tools of mathematics which form the basis of machine learning. In general, statistics is a science concerned with collecting, analysing, interpreting\n the data. Data are the facts and figure that can be classified as either quantitative or qualitative. From the given set of data, we can predict the expected observation, difference between the outcome of two observations and how data look like which can help in better decision making process\n [2]. Descriptive and inferential statistics are the two methods of data analysis. Descriptive statistics summarize the raw data into information through which common expectation and variation of data can be taken. It also provides graphical methods that can be used to visualize the sample\n of data and qualitative understanding of observation whereas inferential statistics refers to drawing conclusions from data. Inferences are made under the framework of probability theory. So, understanding of data and interpretation of result are two important aspects of machine learning.\n In this paper, we have reviewed the different methods of ML, mathematics behind ML, its application in day to day life and future aspects."
                },
                {
                    "title": "GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering",
                    "abstract": "We introduce GQA, a new dataset for real-world visual reasoning and compositional question answering, seeking to address key shortcomings of previous VQA datasets. We have developed a strong and robust question engine that leverages Visual Genome scene graph structures to create 22M diverse reasoning questions, which all come with functional programs that represent their semantics. We use the programs to gain tight control over the answer distribution and present a new tunable smoothing technique to mitigate question biases. Accompanying the dataset is a suite of new metrics that evaluate essential qualities such as consistency, grounding and plausibility. A careful analysis is performed for baselines as well as state-of-the-art models, providing fine-grained results for different question types and topologies. Whereas a blind LSTM obtains a mere 42.1%, and strong VQA models achieve 54.1%, human performance tops at 89.3%, offering ample opportunity for new research to explore. We hope GQA will provide an enabling resource for the next generation of models with enhanced robustness, improved consistency, and deeper semantic understanding of vision and language."
                },
                {
                    "title": "SNIP: Single-shot Network Pruning based on Connection Sensitivity",
                    "abstract": "Pruning large neural networks while maintaining their performance is often desirable due to the reduced space and time complexity. In existing methods, pruning is done within an iterative optimization procedure with either heuristically designed pruning schedules or additional hyperparameters, undermining their utility. In this work, we present a new approach that prunes a given network once at initialization prior to training. To achieve this, we introduce a saliency criterion based on connection sensitivity that identifies structurally important connections in the network for the given task. This eliminates the need for both pretraining and the complex pruning schedule while making it robust to architecture variations. After pruning, the sparse network is trained in the standard way. Our method obtains extremely sparse networks with virtually the same accuracy as the reference network on the MNIST, CIFAR-10, and Tiny-ImageNet classification tasks and is broadly applicable to various architectures including convolutional, residual and recurrent networks. Unlike existing methods, our approach enables us to demonstrate that the retained connections are indeed relevant to the given task."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we perform rapid switching for fused adapters in large generative models while minimizing concept loss during multi-adapter fusion and maintaining high expressive power without significantly increasing training or inference costs?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of existing low rank adaptation methods, particularly in mobile and resource-constrained environments. By enabling rapid switching of adapters and reducing concept loss, the proposed approach could lead to more efficient deployment of generative models in real-world applications, such as mobile AI applications and interactive systems. This advancement could inspire further research into high rank adapters and their potential, leading to improved model performance and versatility in various applications.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the need to balance the trade-offs between model performance, memory usage, and latency. Naive approaches may fail because they do not adequately address the complexities of adapter fusion and the concept loss that arises from additive merging of multiple adapters. Additionally, the technical obstacles include designing a sparse yet high rank adapter that can maintain expressive power while minimizing the number of parameters that need to be trained and stored, which is particularly challenging in the context of large generative models.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on low rank adaptation methods, which have inherent limitations such as significant memory overhead during inference and concept loss when using multiple adapters. Existing solutions have not effectively addressed the need for rapid adapter switching or the complexities of multi-adapter fusion. Barriers include a lack of exploration into high rank adapters and their potential advantages. Our approach differs by introducing Sparse High Rank Adapters (SHiRA), which leverages extreme sparsity to enable rapid switching and reduce concept loss, thus providing a more efficient solution than prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of Sparse High Rank Adapters (SHiRA), which will be trained on a small subset of parameters from the original model. We will evaluate SHiRA using a benchmark dataset relevant to generative models, measuring performance through metrics such as inference latency, memory usage, and concept preservation during multi-adapter fusion. The expected outcomes include demonstrating that SHiRA allows for rapid switching of adapters with minimal latency and reduced concept loss, while"
            }
        },
        "author_data": {
            "3f3b4f5e-3b9c-4db6-995a-bddd95474afc": {
                "pk": "3f3b4f5e-3b9c-4db6-995a-bddd95474afc",
                "name": "Kartikeya Bhardwaj",
                "collaborators": [
                    "Radu Marculescu",
                    "Guihong Li",
                    "Sweta Priyadarshi",
                    "Naveen Suda",
                    "Dibakar Gope",
                    "Paul Whatmough",
                    "Danny Loh",
                    "Nilesh Prasad Pandey",
                    "Wei Chen",
                    "Yuedong Yang"
                ],
                "domain": [
                    "Model Compression",
                    "Edge AI",
                    "Federated Learning",
                    "Neural Architecture Search"
                ],
                "publications": [
                    {
                        "title": "Dream Distillation: A Data-Independent Model Compression Framework",
                        "abstract": "Model compression is eminently suited for deploying deep learning on IoT-devices. However, existing model compression techniques rely on access to the original or some alternate dataset. In this paper, we address the model compression problem when no real data is available, e.g., when data is private. To this end, we propose Dream Distillation, a data-independent model compression framework. Our experiments show that Dream Distillation can achieve 88.5% accuracy on the CIFAR-10 test set without actually training on the original data!"
                    },
                    {
                        "title": "EdgeAI: A Vision for Deep Learning in IoT Era",
                        "abstract": "The significant computational requirements of deep learning present a major bottleneck for its large-scale adoption on hardware-constrained IoT-devices. Here, we envision a new paradigm called EdgeAI to address major impediments associated with deploying deep networks at the edge. Specifically, we discuss the existing directions in computation-aware deep learning and describe two new challenges in the IoT era: (1) Data-independent deployment of learning, and (2) Communication-aware distributed inference. We further present new directions from our recent research to alleviate the latter two challenges. Overcoming these challenges is crucial for rapid adoption of learning on IoT-devices in order to truly enable EdgeAI."
                    },
                    {
                        "title": "New Directions in Distributed Deep Learning: Bringing the Network at Forefront of IoT Design",
                        "abstract": "In this paper, we first highlight three major challenges to large-scale adoption of deep learning at the edge: (i) Hardware-constrained IoT devices, (ii) Data security and privacy in the IoT era, and (iii) Lack of network-aware deep learning algorithms for distributed inference across multiple IoT devices. We then provide a unified view targeting three research directions that naturally emerge from the above challenges: (1) Federated learning for training deep networks, (2) Data-independent deployment of learning algorithms, and (3) Communication-aware distributed inference. We believe that the above research directions need a network-centric approach to enable the edge intelligence and, therefore, fully exploit the true potential of IoT."
                    },
                    {
                        "title": "Climate Anomalies vs Air Pollution: Carbon Emissions and Anomaly Networks",
                        "abstract": "This project aims to shed light on how man-made carbon emissions are affecting global wind patterns by looking for temporal and geographical correlations between carbon emissions, surface temperatures anomalies, and wind speed anomalies at high altitude. We use a networks-based approach and daily data from 1950 to 2010 [1-3] to model and draw correlations between disparate regions of the globe."
                    },
                    {
                        "title": "How does topology influence gradient propagation and model performance of deep networks with DenseNet-type skip connections?",
                        "abstract": "DenseNets introduce concatenation-type skip connections that achieve state-of-the-art accuracy in several computer vision tasks. In this paper, we reveal that the topology of the concatenation-type skip connections is closely related to the gradient propagation which, in turn, enables a predictable behavior of DNNs' test performance. To this end, we introduce a new metric called NN-Mass to quantify how effectively information flows through DNNs. Moreover, we empirically show that NN-Mass also works for other types of skip connections, e.g., for ResNets, Wide-ResNets (WRNs), and MobileNets, which contain addition-type skip connections (i.e., residuals or inverted residuals). As such, for both DenseNet-like CNNs and ResNets/WRNs/MobileNets, our theoretically grounded NN-Mass can identify models with similar accuracy, despite having significantly different size/compute requirements. Detailed experiments on both synthetic and real datasets (e.g., MNIST, CIFAR-10, CIFAR-100, ImageNet) provide extensive evidence for our insights. Finally, the closed-form equation of our NN-Mass enables us to design significantly compressed DenseNets (for CIFAR-10) and MobileNets (for ImageNet) directly at initialization without time-consuming training and/or searching."
                    },
                    {
                        "title": "Memory- and Communication-Aware Model Compression for Distributed Deep Learning Inference on IoT",
                        "abstract": "Model compression has emerged as an important area of research for deploying deep learning models on Internet-of-Things (IoT). However, for extremely memory-constrained scenarios, even the compressed models cannot fit within the memory of a single device and, as a result, must be distributed across multiple devices. This leads to a distributed inference paradigm in which memory and communication costs represent a major bottleneck. Yet, existing model compression techniques are not communication-aware. Therefore, we propose Network of Neural Networks (NoNN), a new distributed IoT learning paradigm that compresses a large pretrained 'teacher' deep network into several disjoint and highly-compressed 'student' modules, without loss of accuracy. Moreover, we propose a network science-based knowledge partitioning algorithm for the teacher model, and then train individual students on the resulting disjoint partitions. Extensive experimentation on five image classification datasets, for user-defined memory/performance budgets, show that NoNN achieves higher accuracy than several baselines and similar accuracy as the teacher model, while using minimal communication among students. Finally, as a case study, we deploy the proposed model for CIFAR-10 dataset on edge devices and demonstrate significant improvements in memory footprint (up to 24x), performance (up to 12x), and energy per node (up to 14x) compared to the large teacher model. We further show that for distributed inference on multiple edge devices, our proposed NoNN model results in up to 33x reduction in total latency w.r.t. a state-of-the-art model compression baseline."
                    },
                    {
                        "title": "FedMAX: Mitigating Activation Divergence for Accurate and Communication-Efficient Federated Learning",
                        "abstract": "In this paper, we identify a new phenomenon called activation-divergence which occurs in Federated Learning (FL) due to data heterogeneity (i.e., data being non-IID) across multiple users. Specifically, we argue that the activation vectors in FL can diverge, even if subsets of users share a few common classes with data residing on different devices. To address the activation-divergence issue, we introduce a prior based on the principle of maximum entropy; this prior assumes minimal information about the per-device activation vectors and aims at making the activation vectors of same classes as similar as possible across multiple devices. Our results show that, for both IID and non-IID settings, our proposed approach results in better accuracy (due to the significantly more similar activation vectors across multiple devices), and is more communication-efficient than state-of-the-art approaches in FL. Finally, we illustrate the effectiveness of our approach on a few common benchmarks and two large medical datasets."
                    },
                    {
                        "title": "TIPS: Topologically Important Path Sampling for Anytime Neural Networks",
                        "abstract": "Anytime neural networks (AnytimeNNs) are a promising solution to adaptively adjust the model complexity at runtime under various hardware resource constraints. However, the manually-designed AnytimeNNs are biased by designers' prior experience and thus provide sub-optimal solutions. To address the limitations of existing hand-crafted approaches, we first model the training process of AnytimeNNs as a discrete-time Markov chain (DTMC) and use it to identify the paths that contribute the most to the training of AnytimeNNs. Based on this new DTMC-based analysis, we further propose TIPS, a framework to automatically design AnytimeNNs under various hardware constraints. Our experimental results show that TIPS can improve the convergence rate and test accuracy of AnytimeNNs. Compared to the existing AnytimeNNs approaches, TIPS improves the accuracy by 2%-6.6% on multiple datasets and achieves SOTA accuracy-FLOPs tradeoffs."
                    },
                    {
                        "title": "ZiCo: Zero-shot NAS via Inverse Coefficient of Variation on Gradients",
                        "abstract": "Neural Architecture Search (NAS) is widely used to automatically obtain the neural network with the best performance among a large number of candidate architectures. To reduce the search time, zero-shot NAS aims at designing training-free proxies that can predict the test performance of a given architecture. However, as shown recently, none of the zero-shot proxies proposed to date can actually work consistently better than a naive proxy, namely, the number of network parameters (#Params). To improve this state of affairs, as the main theoretical contribution, we first reveal how some specific gradient properties across different samples impact the convergence rate and generalization capacity of neural networks. Based on this theoretical analysis, we propose a new zero-shot proxy, ZiCo, the first proxy that works consistently better than #Params. We demonstrate that ZiCo works better than State-Of-The-Art (SOTA) proxies on several popular NAS-Benchmarks (NASBench101, NATSBench-SSS/TSS, TransNASBench-101) for multiple applications (e.g., image classification/reconstruction and pixel-level prediction). Finally, we demonstrate that the optimal architectures found via ZiCo are as competitive as the ones found by one-shot and multi-shot NAS methods, but with much less search time. For example, ZiCo-based NAS can find optimal architectures with 78.1%, 79.4%, and 80.4% test accuracy under inference budgets of 450M, 600M, and 1000M FLOPs, respectively, on ImageNet within 0.4 GPU days. Our code is available at https://github.com/SLDGroup/ZiCo."
                    },
                    {
                        "title": "Super-Efficient Super Resolution for Fast Adversarial Defense at the Edge",
                        "abstract": "Autonomous systems are highly vulnerable to a variety of adversarial attacks on Deep Neural Networks (DNNs). Training-free model-agnostic defenses have recently gained popularity due to their speed, ease of deployment, and ability to work across many DNNs. To this end, a new technique has emerged for mitigating attacks on image classification DNNs, namely, preprocessing adversarial images using super resolution -- upscaling low-quality inputs into high-resolution images. This defense requires running both image classifiers and super resolution models on constrained autonomous systems. However, super resolution incurs a heavy computational cost. Therefore, in this paper, we investigate the following question: Does the robustness of image classifiers suffer if we use tiny super resolution models? To answer this, we first review a recent work called Super-Efficient Super Resolution (SESR) that achieves similar or better image quality than prior art while requiring 2x to 330x fewer Multiply-Accumulate (MAC) operations. We demonstrate that despite being orders of magnitude smaller than existing models, SESR achieves the same level of robustness as significantly larger networks. Finally, we estimate end-to-end performance of super resolution-based defenses on a commercial Arm Ethos-U55 micro-NPU. Our findings show that SESR achieves nearly 3x higher FPS than a baseline while achieving similar robustness."
                    },
                    {
                        "title": "A Dynamic Network and Representation LearningApproach for Quantifying Economic Growth fromSatellite Imagery",
                        "abstract": "Quantifying the improvement in human living standard, as well as the city growth in developing countries, is a challenging problem due to the lack of reliable economic data. Therefore, there is a fundamental need for alternate, largely unsupervised, computational methods that can estimate the economic conditions in the developing regions. To this end, we propose a new network science- and representation learning-based approach that can quantify economic indicators and visualize the growth of various regions. More precisely, we first create a dynamic network drawn out of high-resolution nightlight satellite images. We then demonstrate that using representation learning to mine the resulting network, our proposed approach can accurately predict spatial gross economic expenditures over large regions. Our method, which requires only nightlight images and limited survey data, can capture city-growth, as well as how people's living standard is changing; this can ultimately facilitate the decision makers' understanding of growth without heavily relying on expensive and time-consuming surveys."
                    },
                    {
                        "title": "ZiCo-BC: A Bias Corrected Zero-Shot NAS for Vision Tasks",
                        "abstract": "Zero-Shot Neural Architecture Search (NAS) approaches propose novel training-free metrics called zero-shot proxies to substantially reduce the search time compared to the traditional training-based NAS. Despite the success on image classification, the effectiveness of zero-shot proxies is rarely evaluated on complex vision tasks such as semantic segmentation and object detection. Moreover, existing zero-shot proxies are shown to be biased towards certain model characteristics which restricts their broad applicability. In this paper, we empirically study the bias of state-of-the-art (SOTA) zero-shot proxy ZiCo across multiple vision tasks and observe that ZiCo is biased towards thinner and deeper networks, leading to sub-optimal architectures. To solve the problem, we propose a novel bias correction on ZiCo, called ZiCo-BC. Our extensive experiments across various vision tasks (image classification, object detection and semantic segmentation) show that our approach can successfully search for architectures with higher accuracy and significantly lower latency on Samsung Galaxy S10 devices."
                    },
                    {
                        "title": "On Network Science and Mutual Information for Explaining Deep Neural Networks",
                        "abstract": "In this paper, we present a new approach to interpret deep learning models. By coupling mutual information with network science, we explore how information flows through feedforward networks. We show that efficiently approximating mutual information allows us to create an information measure that quantifies how much information flows between any two neurons of a deep learning model. To that end, we propose NIF, Neural Information Flow, a technique for codifying information flow that exposes deep learning model internals and provides feature attributions."
                    },
                    {
                        "title": "Zero-Shot Neural Architecture Search: Challenges, Solutions, and Opportunities",
                        "abstract": "Recently, zero-shot (or training-free) Neural Architecture Search (NAS) approaches have been proposed to liberate NAS from the expensive training process. The key idea behind zero-shot NAS approaches is to design proxies that can predict the accuracy of some given networks without training the network parameters. The proxies proposed so far are usually inspired by recent progress in theoretical understanding of deep learning and have shown great potential on several datasets and NAS benchmarks. This paper aims to comprehensively review and compare the state-of-the-art (SOTA) zero-shot NAS approaches, with an emphasis on their hardware awareness. To this end, we first review the mainstream zero-shot proxies and discuss their theoretical underpinnings. We then compare these zero-shot proxies through large-scale experiments and demonstrate their effectiveness in both hardware-aware and hardware-oblivious NAS scenarios. Finally, we point out several promising ideas to design better proxies. Our source code and the list of related papers are available on https://github.com/SLDGroup/survey-zero-shot-nas."
                    },
                    {
                        "title": "Oh! We Freeze: Improving Quantized Knowledge Distillation via Signal Propagation Analysis for Large Language Models",
                        "abstract": "Large generative models such as large language models (LLMs) and diffusion models have revolutionized the fields of NLP and computer vision respectively. However, their slow inference, high computation and memory requirement makes it challenging to deploy them on edge devices. In this study, we propose a light-weight quantization aware fine tuning technique using knowledge distillation (KD-QAT) to improve the performance of 4-bit weight quantized LLMs using commonly available datasets to realize a popular language use case, on device chat applications. To improve this paradigm of finetuning, as main contributions, we provide insights into stability of KD-QAT by empirically studying the gradient propagation during training to better understand the vulnerabilities of KD-QAT based approaches to low-bit quantization errors. Based on our insights, we propose ov-freeze, a simple technique to stabilize the KD-QAT process. Finally, we experiment with the popular 7B LLaMAv2-Chat model at 4-bit quantization level and demonstrate that ov-freeze results in near floating point precision performance, i.e., less than 0.7% loss of accuracy on Commonsense Reasoning benchmarks."
                    },
                    {
                        "title": "Collapsible Linear Blocks for Super-Efficient Super Resolution",
                        "abstract": "With the advent of smart devices that support 4K and 8K resolution, Single Image Super Resolution (SISR) has become an important computer vision problem. However, most super resolution deep networks are computationally very expensive. In this paper, we propose Super-Efficient Super Resolution (SESR) networks that establish a new state-of-the-art for efficient super resolution. Our approach is based on linear overparameterization of CNNs and creates an efficient model architecture for SISR. With theoretical analysis, we uncover the limitations of existing overparameterization methods and show how the proposed method alleviates them. Detailed experiments across six benchmark datasets demonstrate that SESR achieves similar or better image quality than state-of-the-art models while requiring 2x to 330x fewer Multiply-Accumulate (MAC) operations. As a result, SESR can be used on constrained hardware to perform x2 (1080p to 4K) and x4 (1080p to 8K) SISR. Towards this, we estimate hardware performance numbers for a commercial Arm mobile-Neural Processing Unit (NPU) for 1080p to 4K (x2) and 1080p to 8K (x4) SISR. Our results highlight the challenges faced by super resolution on AI accelerators and demonstrate that SESR is significantly faster (e.g., 6x-8x higher FPS) than existing models on mobile-NPU. Finally, SESR outperforms prior models by 1.5x-2x in latency on Arm CPU and GPU when deployed on a real mobile device. The code for this work is available at https://github.com/ARM-software/sesr."
                    },
                    {
                        "title": "Restructurable Activation Networks",
                        "abstract": "Is it possible to restructure the non-linear activation functions in a deep network to create hardware-efficient models? To address this question, we propose a new paradigm called Restructurable Activation Networks (RANs) that manipulate the amount of non-linearity in models to improve their hardware-awareness and efficiency. First, we propose RAN-explicit (RAN-e) -- a new hardware-aware search space and a semi-automatic search algorithm -- to replace inefficient blocks with hardware-aware blocks. Next, we propose a training-free model scaling method called RAN-implicit (RAN-i) where we theoretically prove the link between network topology and its expressivity in terms of number of non-linear units. We demonstrate that our networks achieve state-of-the-art results on ImageNet at different scales and for several types of hardware. For example, compared to EfficientNet-Lite-B0, RAN-e achieves a similar accuracy while improving Frames-Per-Second (FPS) by 1.5x on Arm micro-NPUs. On the other hand, RAN-i demonstrates up to 2x reduction in #MACs over ConvNexts with a similar or better accuracy. We also show that RAN-i achieves nearly 40% higher FPS than ConvNext on Arm-based datacenter CPUs. Finally, RAN-i based object detection networks achieve a similar or higher mAP and up to 33% higher FPS on datacenter CPUs compared to ConvNext based models. The code to train and evaluate RANs and the pretrained networks are available at https://github.com/ARM-software/ML-restructurable-activation-networks."
                    },
                    {
                        "title": "FouRA: Fourier Low Rank Adaptation",
                        "abstract": "While Low-Rank Adaptation (LoRA) has proven beneficial for efficiently fine-tuning large models, LoRA fine-tuned text-to-image diffusion models lack diversity in the generated images, as the model tends to copy data from the observed training samples. This effect becomes more pronounced at higher values of adapter strength and for adapters with higher ranks which are fine-tuned on smaller datasets. To address these challenges, we present FouRA, a novel low-rank method that learns projections in the Fourier domain along with learning a flexible input-dependent adapter rank selection strategy. Through extensive experiments and analysis, we show that FouRA successfully solves the problems related to data copying and distribution collapse while significantly improving the generated image quality. We demonstrate that FouRA enhances the generalization of fine-tuned models thanks to its adaptive rank selection. We further show that the learned projections in the frequency domain are decorrelated and prove effective when merging multiple adapters. While FouRA is motivated for vision tasks, we also demonstrate its merits for language tasks on the GLUE benchmark."
                    },
                    {
                        "title": "Rapid Switching and Multi-Adapter Fusion via Sparse High Rank Adapters",
                        "abstract": "In this paper, we propose Sparse High Rank Adapters (SHiRA) that directly finetune 1-2% of the base model weights while leaving others unchanged, thus, resulting in a highly sparse adapter. This high sparsity incurs no inference overhead, enables rapid switching directly in the fused mode, and significantly reduces concept-loss during multi-adapter fusion. Our extensive experiments on LVMs and LLMs demonstrate that finetuning merely 1-2% parameters in the base model is sufficient for many adapter tasks and significantly outperforms Low Rank Adaptation (LoRA). We also show that SHiRA is orthogonal to advanced LoRA methods such as DoRA and can be easily combined with existing techniques."
                    }
                ]
            },
            "9afb1669-eba6-45b5-ac3d-eb66a656ac05": {
                "pk": "9afb1669-eba6-45b5-ac3d-eb66a656ac05",
                "name": "Nilesh Prasad Pandey",
                "collaborators": [
                    "Markus Nagel",
                    "Kartikeya Bhardwaj",
                    "Sweta Priyadarshi",
                    "Chirag Patel",
                    "Harris Teague",
                    "Viswanath Ganapathy",
                    "Rafael Esteves",
                    "Shreya Kadambi",
                    "Shubhankar Borse",
                    "Risheek Garrepalli"
                ],
                "domain": [
                    "Model Compression",
                    "Quantization",
                    "Knowledge Distillation",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Softmax Bias Correction for Quantized Generative Models",
                        "abstract": "Post-training quantization (PTQ) is the go-to compression technique for large generative models, such as stable diffusion or large language models. PTQ methods commonly keep the softmax activation in higher precision as it has been shown to be very sensitive to quantization noise. However, this can lead to a significant runtime and power overhead during inference on resource-constraint edge devices. In this work, we investigate the source of the softmax sensitivity to quantization and show that the quantization operation leads to a large bias in the softmax output, causing accuracy degradation. To overcome this issue, we propose an offline bias correction technique that improves the quantizability of softmax without additional compute during deployment, as it can be readily absorbed into the quantization parameters. We demonstrate the effectiveness of our method on stable diffusion v1.5 and 125M-size OPT language model, achieving significant accuracy improvement for 8-bit quantized softmax."
                    },
                    {
                        "title": "Oh! We Freeze: Improving Quantized Knowledge Distillation via Signal Propagation Analysis for Large Language Models",
                        "abstract": "Large generative models such as large language models (LLMs) and diffusion models have revolutionized the fields of NLP and computer vision respectively. However, their slow inference, high computation and memory requirement makes it challenging to deploy them on edge devices. In this study, we propose a light-weight quantization aware fine tuning technique using knowledge distillation (KD-QAT) to improve the performance of 4-bit weight quantized LLMs using commonly available datasets to realize a popular language use case, on device chat applications. To improve this paradigm of finetuning, as main contributions, we provide insights into stability of KD-QAT by empirically studying the gradient propagation during training to better understand the vulnerabilities of KD-QAT based approaches to low-bit quantization errors. Based on our insights, we propose ov-freeze, a simple technique to stabilize the KD-QAT process. Finally, we experiment with the popular 7B LLaMAv2-Chat model at 4-bit quantization level and demonstrate that ov-freeze results in near floating point precision performance, i.e., less than 0.7% loss of accuracy on Commonsense Reasoning benchmarks."
                    },
                    {
                        "title": "A Practical Mixed Precision Algorithm for Post-Training Quantization",
                        "abstract": "Neural network quantization is frequently used to optimize model size, latency and power consumption for on-device deployment of neural networks. In many cases, a target bit-width is set for an entire network, meaning every layer get quantized to the same number of bits. However, for many networks some layers are significantly more robust to quantization noise than others, leaving an important axis of improvement unused. As many hardware solutions provide multiple different bit-width settings, mixed-precision quantization has emerged as a promising solution to find a better performance-efficiency trade-off than homogeneous quantization. However, most existing mixed precision algorithms are rather difficult to use for practitioners as they require access to the training data, have many hyper-parameters to tune or even depend on end-to-end retraining of the entire model. In this work, we present a simple post-training mixed precision algorithm that only requires a small unlabeled calibration dataset to automatically select suitable bit-widths for each layer for desirable on-device performance. Our algorithm requires no hyper-parameter tuning, is robust to data variation and takes into account practical hardware deployment constraints making it a great candidate for practical use. We experimentally validate our proposed method on several computer vision tasks, natural language processing tasks and many different networks, and show that we can find mixed precision networks that provide a better trade-off between accuracy and efficiency than their homogeneous bit-width equivalents."
                    },
                    {
                        "title": "Rapid Switching and Multi-Adapter Fusion via Sparse High Rank Adapters",
                        "abstract": "In this paper, we propose Sparse High Rank Adapters (SHiRA) that directly finetune 1-2% of the base model weights while leaving others unchanged, thus, resulting in a highly sparse adapter. This high sparsity incurs no inference overhead, enables rapid switching directly in the fused mode, and significantly reduces concept-loss during multi-adapter fusion. Our extensive experiments on LVMs and LLMs demonstrate that finetuning merely 1-2% parameters in the base model is sufficient for many adapter tasks and significantly outperforms Low Rank Adaptation (LoRA). We also show that SHiRA is orthogonal to advanced LoRA methods such as DoRA and can be easily combined with existing techniques."
                    },
                    {
                        "title": "FouRA: Fourier Low Rank Adaptation",
                        "abstract": "While Low-Rank Adaptation (LoRA) has proven beneficial for efficiently fine-tuning large models, LoRA fine-tuned text-to-image diffusion models lack diversity in the generated images, as the model tends to copy data from the observed training samples. This effect becomes more pronounced at higher values of adapter strength and for adapters with higher ranks which are fine-tuned on smaller datasets. To address these challenges, we present FouRA, a novel low-rank method that learns projections in the Fourier domain along with learning a flexible input-dependent adapter rank selection strategy. Through extensive experiments and analysis, we show that FouRA successfully solves the problems related to data copying and distribution collapse while significantly improving the generated image quality. We demonstrate that FouRA enhances the generalization of fine-tuned models thanks to its adaptive rank selection. We further show that the learned projections in the frequency domain are decorrelated and prove effective when merging multiple adapters. While FouRA is motivated for vision tasks, we also demonstrate its merits for language tasks on the GLUE benchmark."
                    }
                ]
            },
            "278cd82b-5561-45e5-a380-624b01ae11de": {
                "pk": "278cd82b-5561-45e5-a380-624b01ae11de",
                "name": "Sweta Priyadarshi",
                "collaborators": [
                    "Kartikeya Bhardwaj",
                    "Viswanath Ganapathy",
                    "Nilesh Prasad Pandey",
                    "Hsin-Pai Cheng",
                    "Harris Teague",
                    "Shubhankar Borse",
                    "Shreya Kadambi",
                    "Risheek Garrepalli",
                    "Rafael Esteves",
                    "Zhuojin Li"
                ],
                "domain": [
                    "Neural Architecture Search",
                    "Quantization",
                    "Knowledge Distillation",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "ZiCo-BC: A Bias Corrected Zero-Shot NAS for Vision Tasks",
                        "abstract": "Zero-Shot Neural Architecture Search (NAS) approaches propose novel training-free metrics called zero-shot proxies to substantially reduce the search time compared to the traditional training-based NAS. Despite the success on image classification, the effectiveness of zero-shot proxies is rarely evaluated on complex vision tasks such as semantic segmentation and object detection. Moreover, existing zero-shot proxies are shown to be biased towards certain model characteristics which restricts their broad applicability. In this paper, we empirically study the bias of state-of-the-art (SOTA) zero-shot proxy ZiCo across multiple vision tasks and observe that ZiCo is biased towards thinner and deeper networks, leading to sub-optimal architectures. To solve the problem, we propose a novel bias correction on ZiCo, called ZiCo-BC. Our extensive experiments across various vision tasks (image classification, object detection and semantic segmentation) show that our approach can successfully search for architectures with higher accuracy and significantly lower latency on Samsung Galaxy S10 devices."
                    },
                    {
                        "title": "DONNAv2 -- Lightweight Neural Architecture Search for Vision tasks",
                        "abstract": "With the growing demand for vision applications and deployment across edge devices, the development of hardware-friendly architectures that maintain performance during device deployment becomes crucial. Neural architecture search (NAS) techniques explore various approaches to discover efficient architectures for diverse learning tasks in a computationally efficient manner. In this paper, we present the next-generation neural architecture design for computationally efficient neural architecture distillation - DONNAv2 . Conventional NAS algorithms rely on a computationally extensive stage where an accuracy predictor is learned to estimate model performance within search space. This building of accuracy predictors helps them predict the performance of models that are not being finetuned. Here, we have developed an elegant approach to eliminate building the accuracy predictor and extend DONNA to a computationally efficient setting. The loss metric of individual blocks forming the network serves as the surrogate performance measure for the sampled models in the NAS search stage. To validate the performance of DONNAv2 we have performed extensive experiments involving a range of diverse vision tasks including classification, object detection, image denoising, super-resolution, and panoptic perception network (YOLOP). The hardware-in-the-loop experiments were carried out using the Samsung Galaxy S10 mobile platform. Notably, DONNAv2 reduces the computational cost of DONNA by 10x for the larger datasets. Furthermore, to improve the quality of NAS search space, DONNAv2 leverages a block knowledge distillation filter to remove blocks with high inference costs."
                    },
                    {
                        "title": "Oh! We Freeze: Improving Quantized Knowledge Distillation via Signal Propagation Analysis for Large Language Models",
                        "abstract": "Large generative models such as large language models (LLMs) and diffusion models have revolutionized the fields of NLP and computer vision respectively. However, their slow inference, high computation and memory requirement makes it challenging to deploy them on edge devices. In this study, we propose a light-weight quantization aware fine tuning technique using knowledge distillation (KD-QAT) to improve the performance of 4-bit weight quantized LLMs using commonly available datasets to realize a popular language use case, on device chat applications. To improve this paradigm of finetuning, as main contributions, we provide insights into stability of KD-QAT by empirically studying the gradient propagation during training to better understand the vulnerabilities of KD-QAT based approaches to low-bit quantization errors. Based on our insights, we propose ov-freeze, a simple technique to stabilize the KD-QAT process. Finally, we experiment with the popular 7B LLaMAv2-Chat model at 4-bit quantization level and demonstrate that ov-freeze results in near floating point precision performance, i.e., less than 0.7% loss of accuracy on Commonsense Reasoning benchmarks."
                    },
                    {
                        "title": "FouRA: Fourier Low Rank Adaptation",
                        "abstract": "While Low-Rank Adaptation (LoRA) has proven beneficial for efficiently fine-tuning large models, LoRA fine-tuned text-to-image diffusion models lack diversity in the generated images, as the model tends to copy data from the observed training samples. This effect becomes more pronounced at higher values of adapter strength and for adapters with higher ranks which are fine-tuned on smaller datasets. To address these challenges, we present FouRA, a novel low-rank method that learns projections in the Fourier domain along with learning a flexible input-dependent adapter rank selection strategy. Through extensive experiments and analysis, we show that FouRA successfully solves the problems related to data copying and distribution collapse while significantly improving the generated image quality. We demonstrate that FouRA enhances the generalization of fine-tuned models thanks to its adaptive rank selection. We further show that the learned projections in the frequency domain are decorrelated and prove effective when merging multiple adapters. While FouRA is motivated for vision tasks, we also demonstrate its merits for language tasks on the GLUE benchmark."
                    },
                    {
                        "title": "Rapid Switching and Multi-Adapter Fusion via Sparse High Rank Adapters",
                        "abstract": "In this paper, we propose Sparse High Rank Adapters (SHiRA) that directly finetune 1-2% of the base model weights while leaving others unchanged, thus, resulting in a highly sparse adapter. This high sparsity incurs no inference overhead, enables rapid switching directly in the fused mode, and significantly reduces concept-loss during multi-adapter fusion. Our extensive experiments on LVMs and LLMs demonstrate that finetuning merely 1-2% parameters in the base model is sufficient for many adapter tasks and significantly outperforms Low Rank Adaptation (LoRA). We also show that SHiRA is orthogonal to advanced LoRA methods such as DoRA and can be easily combined with existing techniques."
                    }
                ]
            },
            "0583a0c3-51c5-44c8-a894-9935a60f6e64": {
                "pk": "0583a0c3-51c5-44c8-a894-9935a60f6e64",
                "name": "Viswanath Ganapathy",
                "collaborators": [
                    "Javad Heydari",
                    "Olimpiya Saha",
                    "Sweta Priyadarshi",
                    "Tianyu Jiang",
                    "Hsin-Pai Cheng",
                    "Sendil Krishna",
                    "Chirag Patel",
                    "Sauptik Dhar",
                    "Pelin Kurt Garberson",
                    "Mohak Shah"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Neural Architecture Search",
                    "Predictive Maintenance",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Reinforcement Learning-Based Coverage Path Planning with Implicit Cellular Decomposition",
                        "abstract": "Coverage path planning in a generic known environment is shown to be NP-hard. When the environment is unknown, it becomes more challenging as the robot is required to rely on its online map information built during coverage for planning its path. A significant research effort focuses on designing heuristic or approximate algorithms that achieve reasonable performance. Such algorithms have sub-optimal performance in terms of covering the area or the cost of coverage, e.g., coverage time or energy consumption. In this paper, we provide a systematic analysis of the coverage problem and formulate it as an optimal stopping time problem, where the trade-off between coverage performance and its cost is explicitly accounted for. Next, we demonstrate that reinforcement learning (RL) techniques can be leveraged to solve the problem computationally. To this end, we provide some technical and practical considerations to facilitate the application of the RL algorithms and improve the efficiency of the solutions. Finally, through experiments in grid world environments and Gazebo simulator, we show that reinforcement learning-based algorithms efficiently cover realistic unknown indoor environments, and outperform the current state of the art."
                    },
                    {
                        "title": "DONNAv2 -- Lightweight Neural Architecture Search for Vision tasks",
                        "abstract": "With the growing demand for vision applications and deployment across edge devices, the development of hardware-friendly architectures that maintain performance during device deployment becomes crucial. Neural architecture search (NAS) techniques explore various approaches to discover efficient architectures for diverse learning tasks in a computationally efficient manner. In this paper, we present the next-generation neural architecture design for computationally efficient neural architecture distillation - DONNAv2 . Conventional NAS algorithms rely on a computationally extensive stage where an accuracy predictor is learned to estimate model performance within search space. This building of accuracy predictors helps them predict the performance of models that are not being finetuned. Here, we have developed an elegant approach to eliminate building the accuracy predictor and extend DONNA to a computationally efficient setting. The loss metric of individual blocks forming the network serves as the surrogate performance measure for the sampled models in the NAS search stage. To validate the performance of DONNAv2 we have performed extensive experiments involving a range of diverse vision tasks including classification, object detection, image denoising, super-resolution, and panoptic perception network (YOLOP). The hardware-in-the-loop experiments were carried out using the Samsung Galaxy S10 mobile platform. Notably, DONNAv2 reduces the computational cost of DONNA by 10x for the larger datasets. Furthermore, to improve the quality of NAS search space, DONNAv2 leverages a block knowledge distillation filter to remove blocks with high inference costs."
                    },
                    {
                        "title": "A Survey on Proactive Customer Care: Enabling Science and Steps to Realize it",
                        "abstract": "In recent times, advances in artificial intelligence (AI) and IoT have enabled seamless and viable maintenance of appliances in home and building environments. Several studies have shown that AI has the potential to provide personalized customer support which could predict and avoid errors more reliably than ever before. In this paper, we have analyzed the various building blocks needed to enable a successful AI-driven predictive maintenance use-case. Unlike, existing surveys which mostly provide a deep dive into the recent AI algorithms for Predictive Maintenance (PdM), our survey provides the complete view; starting from business impact to recent technology advancements in algorithms as well as systems research and model deployment. Furthermore, we provide exemplar use-cases on predictive maintenance of appliances using publicly available data sets. Our survey can serve as a template needed to design a successful predictive maintenance use-case. Finally, we touch upon existing public data sources and provide a step-wise breakdown of an AI-driven proactive customer care (PCC) use-case, starting from generic anomaly detection to fault prediction and finally root-cause analysis. We highlight how such a step-wise approach can be advantageous for accurate model building and helpful for gaining insights into predictive maintenance of electromechanical appliances."
                    }
                ]
            },
            "ad808bd7-8a5a-4cf1-8c19-7802532cabda": {
                "pk": "ad808bd7-8a5a-4cf1-8c19-7802532cabda",
                "name": "Rafael Esteves",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "380371c3-fba8-4583-9166-a77cae042a60": {
                "pk": "380371c3-fba8-4583-9166-a77cae042a60",
                "name": "Shreya Kadambi",
                "collaborators": [
                    "Roohollah Amiri",
                    "Srinivas Yerramalli",
                    "Taesang Yoo",
                    "Fatih Porikli",
                    "Shubhankar Borse",
                    "Nilesh Prasad Pandey",
                    "Kartikeya Bhardwaj",
                    "Viswanath Ganapathy",
                    "Sweta Priyadarshi",
                    "Risheek Garrepalli"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Localization",
                    "Adapter Methods"
                ],
                "publications": [
                    {
                        "title": "Neural RF SLAM for unsupervised positioning and mapping with channel state information",
                        "abstract": "We present a neural network architecture for jointly learning user locations and environment mapping up to isometry, in an unsupervised way, from channel state information (CSI) values with no location information. The model is based on an encoder-decoder architecture. The encoder network maps CSI values to the user location. The decoder network models the physics of propagation by parametrizing the environment using virtual anchors. It aims at reconstructing, from the encoder output and virtual anchor location, the set of time of flights (ToFs) that are extracted from CSI using super-resolution methods. The neural network task is set prediction and is accordingly trained end-to-end. The proposed model learns an interpretable latent, i.e., user location, by just enforcing a physics-based decoder. It is shown that the proposed model achieves sub-meter accuracy on synthetic ray tracing based datasets with single anchor SISO setup while recovering the environment map up to 4cm median error in a 2D environment and 15cm in a 3D environment"
                    },
                    {
                        "title": "Neural 5G Indoor Localization with IMU Supervision",
                        "abstract": "Radio signals are well suited for user localization because they are ubiquitous, can operate in the dark and maintain privacy. Many prior works learn mappings between channel state information (CSI) and position fully-supervised. However, that approach relies on position labels which are very expensive to acquire. In this work, this requirement is relaxed by using pseudo-labels during deployment, which are calculated from an inertial measurement unit (IMU). We propose practical algorithms for IMU double integration and training of the localization system. We show decimeter-level accuracy on simulated and challenging real data of 5G measurements. Our IMU-supervised method performs similarly to fully-supervised, but requires much less effort to deploy."
                    },
                    {
                        "title": "FouRA: Fourier Low Rank Adaptation",
                        "abstract": "While Low-Rank Adaptation (LoRA) has proven beneficial for efficiently fine-tuning large models, LoRA fine-tuned text-to-image diffusion models lack diversity in the generated images, as the model tends to copy data from the observed training samples. This effect becomes more pronounced at higher values of adapter strength and for adapters with higher ranks which are fine-tuned on smaller datasets. To address these challenges, we present FouRA, a novel low-rank method that learns projections in the Fourier domain along with learning a flexible input-dependent adapter rank selection strategy. Through extensive experiments and analysis, we show that FouRA successfully solves the problems related to data copying and distribution collapse while significantly improving the generated image quality. We demonstrate that FouRA enhances the generalization of fine-tuned models thanks to its adaptive rank selection. We further show that the learned projections in the frequency domain are decorrelated and prove effective when merging multiple adapters. While FouRA is motivated for vision tasks, we also demonstrate its merits for language tasks on the GLUE benchmark."
                    },
                    {
                        "title": "Rapid Switching and Multi-Adapter Fusion via Sparse High Rank Adapters",
                        "abstract": "In this paper, we propose Sparse High Rank Adapters (SHiRA) that directly finetune 1-2% of the base model weights while leaving others unchanged, thus, resulting in a highly sparse adapter. This high sparsity incurs no inference overhead, enables rapid switching directly in the fused mode, and significantly reduces concept-loss during multi-adapter fusion. Our extensive experiments on LVMs and LLMs demonstrate that finetuning merely 1-2% parameters in the base model is sufficient for many adapter tasks and significantly outperforms Low Rank Adaptation (LoRA). We also show that SHiRA is orthogonal to advanced LoRA methods such as DoRA and can be easily combined with existing techniques."
                    }
                ]
            },
            "b000c6f7-0f09-429a-a571-a991db15bbf5": {
                "pk": "b000c6f7-0f09-429a-a571-a991db15bbf5",
                "name": "Shubhankar Borse",
                "collaborators": [
                    "Fatih Porikli",
                    "Hong Cai",
                    "Yizhe Zhang",
                    "Hyojin Park",
                    "Debasmit Das",
                    "Risheek Garrepalli",
                    "Senthil Yogamani",
                    "Marvin Klingner",
                    "Varun Ravi Kumar",
                    "Ying Wang"
                ],
                "domain": [
                    "Semantic Segmentation",
                    "Video Segmentation",
                    "Knowledge Distillation",
                    "Autonomous Driving"
                ],
                "publications": [
                    {
                        "title": "InverseForm: A Loss Function for Structured Boundary-Aware Segmentation",
                        "abstract": "We present a novel boundary-aware loss term for semantic segmentation using an inverse-transformation network, which efficiently learns the degree of parametric transformations between estimated and target boundaries. This plug-in loss term complements the cross-entropy loss in capturing boundary transformations and allows consistent and significant performance improvement on segmentation backbone models without increasing their size and computational complexity. We analyze the quantitative and qualitative effects of our loss function on three indoor and outdoor segmentation benchmarks, including Cityscapes, NYU-Depth-v2, and PASCAL, integrating it into the training phase of several backbone networks in both single-task and multi-task settings. Our extensive experiments show that the proposed method consistently outperforms baselines, and even sets the new state-of-the-art on two datasets."
                    },
                    {
                        "title": "HS3: Learning with Proper Task Complexity in Hierarchically Supervised Semantic Segmentation",
                        "abstract": "While deeply supervised networks are common in recent literature, they typically impose the same learning objective on all transitional layers despite their varying representation powers.   In this paper, we propose Hierarchically Supervised Semantic Segmentation (HS3), a training scheme that supervises intermediate layers in a segmentation network to learn meaningful representations by varying task complexity. To enforce a consistent performance vs. complexity trade-off throughout the network, we derive various sets of class clusters to supervise each transitional layer of the network. Furthermore, we devise a fusion framework, HS3-Fuse, to aggregate the hierarchical features generated by these layers, which can provide rich semantic contexts and further enhance the final segmentation. Extensive experiments show that our proposed HS3 scheme considerably outperforms vanilla deep supervision with no added inference cost. Our proposed HS3-Fuse framework further improves segmentation predictions and achieves state-of-the-art results on two large segmentation benchmarks: NYUD-v2 and Cityscapes."
                    },
                    {
                        "title": "AuxAdapt: Stable and Efficient Test-Time Adaptation for Temporally Consistent Video Semantic Segmentation",
                        "abstract": "In video segmentation, generating temporally consistent results across frames is as important as achieving frame-wise accuracy. Existing methods rely either on optical flow regularization or fine-tuning with test data to attain temporal consistency. However, optical flow is not always avail-able and reliable. Besides, it is expensive to compute. Fine-tuning the original model in test time is cost sensitive.   This paper presents an efficient, intuitive, and unsupervised online adaptation method, AuxAdapt, for improving the temporal consistency of most neural network models. It does not require optical flow and only takes one pass of the video. Since inconsistency mainly arises from the model's uncertainty in its output, we propose an adaptation scheme where the model learns from its own segmentation decisions as it streams a video, which allows producing more confident and temporally consistent labeling for similarly-looking pixels across frames. For stability and efficiency, we leverage a small auxiliary segmentation network (AuxNet) to assist with this adaptation. More specifically, AuxNet readjusts the decision of the original segmentation network (Main-Net) by adding its own estimations to that of MainNet. At every frame, only AuxNet is updated via back-propagation while keeping MainNet fixed. We extensively evaluate our test-time adaptation approach on standard video benchmarks, including Cityscapes, CamVid, and KITTI. The results demonstrate that our approach provides label-wise accurate, temporally consistent, and computationally efficient adaptation (5+ folds overhead reduction comparing to state-of-the-art test-time adaptation methods)."
                    },
                    {
                        "title": "Panoptic, Instance and Semantic Relations: A Relational Context Encoder to Enhance Panoptic Segmentation",
                        "abstract": "This paper presents a novel framework to integrate both semantic and instance contexts for panoptic segmentation. In existing works, it is common to use a shared backbone to extract features for both things (countable classes such as vehicles) and stuff (uncountable classes such as roads). This, however, fails to capture the rich relations among them, which can be utilized to enhance visual understanding and segmentation performance. To address this shortcoming, we propose a novel Panoptic, Instance, and Semantic Relations (PISR) module to exploit such contexts. First, we generate panoptic encodings to summarize key features of the semantic classes and predicted instances. A Panoptic Relational Attention (PRA) module is then applied to the encodings and the global feature map from the backbone. It produces a feature map that captures 1) the relations across semantic classes and instances and 2) the relations between these panoptic categories and spatial features. PISR also automatically learns to focus on the more important instances, making it robust to the number of instances used in the relational attention module. Moreover, PISR is a general module that can be applied to any existing panoptic segmentation architecture. Through extensive evaluations on panoptic segmentation benchmarks like Cityscapes, COCO, and ADE20K, we show that PISR attains considerable improvements over existing approaches."
                    },
                    {
                        "title": "DejaVu: Conditional Regenerative Learning to Enhance Dense Prediction",
                        "abstract": "We present DejaVu, a novel framework which leverages conditional image regeneration as additional supervision during training to improve deep networks for dense prediction tasks such as segmentation, depth estimation, and surface normal prediction. First, we apply redaction to the input image, which removes certain structural information by sparse sampling or selective frequency removal. Next, we use a conditional regenerator, which takes the redacted image and the dense predictions as inputs, and reconstructs the original image by filling in the missing structural information. In the redacted image, structural attributes like boundaries are broken while semantic context is largely preserved. In order to make the regeneration feasible, the conditional generator will then require the structure information from the other input source, i.e., the dense predictions. As such, by including this conditional regeneration objective during training, DejaVu encourages the base network to learn to embed accurate scene structure in its dense prediction. This leads to more accurate predictions with clearer boundaries and better spatial consistency. When it is feasible to leverage additional computation, DejaVu can be extended to incorporate an attention-based regeneration module within the dense prediction network, which further improves accuracy. Through extensive experiments on multiple dense prediction benchmarks such as Cityscapes, COCO, ADE20K, NYUD-v2, and KITTI, we demonstrate the efficacy of employing DejaVu during training, as it outperforms SOTA methods at no added computation cost."
                    },
                    {
                        "title": "PosSAM: Panoptic Open-vocabulary Segment Anything",
                        "abstract": "In this paper, we introduce an open-vocabulary panoptic segmentation model that effectively unifies the strengths of the Segment Anything Model (SAM) with the vision-language CLIP model in an end-to-end framework. While SAM excels in generating spatially-aware masks, it's decoder falls short in recognizing object class information and tends to oversegment without additional guidance. Existing approaches address this limitation by using multi-stage techniques and employing separate models to generate class-aware prompts, such as bounding boxes or segmentation masks. Our proposed method, PosSAM is an end-to-end model which leverages SAM's spatially rich features to produce instance-aware masks and harnesses CLIP's semantically discriminative features for effective instance classification. Specifically, we address the limitations of SAM and propose a novel Local Discriminative Pooling (LDP) module leveraging class-agnostic SAM and class-aware CLIP features for unbiased open-vocabulary classification. Furthermore, we introduce a Mask-Aware Selective Ensembling (MASE) algorithm that adaptively enhances the quality of generated masks and boosts the performance of open-vocabulary classification during inference for each image. We conducted extensive experiments to demonstrate our methods strong generalization properties across multiple datasets, achieving state-of-the-art performance with substantial improvements over SOTA open-vocabulary panoptic segmentation methods. In both COCO to ADE20K and ADE20K to COCO settings, PosSAM outperforms the previous state-of-the-art methods by a large margin, 2.4 PQ and 4.6 PQ, respectively. Project Website: https://vibashan.github.io/possam-web/."
                    },
                    {
                        "title": "TransAdapt: A Transformative Framework for Online Test Time Adaptive Semantic Segmentation",
                        "abstract": "Test-time adaptive (TTA) semantic segmentation adapts a source pre-trained image semantic segmentation model to unlabeled batches of target domain test images, different from real-world, where samples arrive one-by-one in an online fashion. To tackle online settings, we propose TransAdapt, a framework that uses transformer and input transformations to improve segmentation performance. Specifically, we pre-train a transformer-based module on a segmentation network that transforms unsupervised segmentation output to a more reliable supervised output, without requiring test-time online training. To also facilitate test-time adaptation, we propose an unsupervised loss based on the transformed input that enforces the model to be invariant and equivariant to photometric and geometric perturbations, respectively. Overall, our framework produces higher quality segmentation masks with up to 17.6% and 2.8% mIOU improvement over no-adaptation and competitive baselines, respectively."
                    },
                    {
                        "title": "4D Panoptic Segmentation as Invariant and Equivariant Field Prediction",
                        "abstract": "In this paper, we develop rotation-equivariant neural networks for 4D panoptic segmentation. 4D panoptic segmentation is a benchmark task for autonomous driving that requires recognizing semantic classes and object instances on the road based on LiDAR scans, as well as assigning temporally consistent IDs to instances across time. We observe that the driving scenario is symmetric to rotations on the ground plane. Therefore, rotation-equivariance could provide better generalization and more robust feature learning. Specifically, we review the object instance clustering strategies and restate the centerness-based approach and the offset-based approach as the prediction of invariant scalar fields and equivariant vector fields. Other sub-tasks are also unified from this perspective, and different invariant and equivariant layers are designed to facilitate their predictions. Through evaluation on the standard 4D panoptic segmentation benchmark of SemanticKITTI, we show that our equivariant models achieve higher accuracy with lower computational costs compared to their non-equivariant counterparts. Moreover, our method sets the new state-of-the-art performance and achieves 1st place on the SemanticKITTI 4D Panoptic Segmentation leaderboard."
                    },
                    {
                        "title": "X-Distill: Improving Self-Supervised Monocular Depth via Cross-Task Distillation",
                        "abstract": "In this paper, we propose a novel method, X-Distill, to improve the self-supervised training of monocular depth via cross-task knowledge distillation from semantic segmentation to depth estimation. More specifically, during training, we utilize a pretrained semantic segmentation teacher network and transfer its semantic knowledge to the depth network. In order to enable such knowledge distillation across two different visual tasks, we introduce a small, trainable network that translates the predicted depth map to a semantic segmentation map, which can then be supervised by the teacher network. In this way, this small network enables the backpropagation from the semantic segmentation teacher's supervision to the depth network during training. In addition, since the commonly used object classes in semantic segmentation are not directly transferable to depth, we study the visual and geometric characteristics of the objects and design a new way of grouping them that can be shared by both tasks. It is noteworthy that our approach only modifies the training process and does not incur additional computation during inference. We extensively evaluate the efficacy of our proposed approach on the standard KITTI benchmark and compare it with the latest state of the art. We further test the generalizability of our approach on Make3D. Overall, the results show that our approach significantly improves the depth estimation accuracy and outperforms the state of the art."
                    },
                    {
                        "title": "Learning Implicit Feature Alignment Function for Semantic Segmentation",
                        "abstract": "Integrating high-level context information with low-level details is of central importance in semantic segmentation. Towards this end, most existing segmentation models apply bilinear up-sampling and convolutions to feature maps of different scales, and then align them at the same resolution. However, bilinear up-sampling blurs the precise information learned in these feature maps and convolutions incur extra computation costs. To address these issues, we propose the Implicit Feature Alignment function (IFA). Our method is inspired by the rapidly expanding topic of implicit neural representations, where coordinate-based neural networks are used to designate fields of signals. In IFA, feature vectors are viewed as representing a 2D field of information. Given a query coordinate, nearby feature vectors with their relative coordinates are taken from the multi-level feature maps and then fed into an MLP to generate the corresponding output. As such, IFA implicitly aligns the feature maps at different levels and is capable of producing segmentation maps in arbitrary resolutions. We demonstrate the efficacy of IFA on multiple datasets, including Cityscapes, PASCAL Context, and ADE20K. Our method can be combined with improvement on various architectures, and it achieves state-of-the-art computation-accuracy trade-off on common benchmarks. Code will be made available at https://github.com/hzhupku/IFA."
                    },
                    {
                        "title": "Self-Supervised Geometric Correspondence for Category-Level 6D Object Pose Estimation in the Wild",
                        "abstract": "While 6D object pose estimation has wide applications across computer vision and robotics, it remains far from being solved due to the lack of annotations. The problem becomes even more challenging when moving to category-level 6D pose, which requires generalization to unseen instances. Current approaches are restricted by leveraging annotations from simulation or collected from humans. In this paper, we overcome this barrier by introducing a self-supervised learning approach trained directly on large-scale real-world object videos for category-level 6D pose estimation in the wild. Our framework reconstructs the canonical 3D shape of an object category and learns dense correspondences between input images and the canonical shape via surface embedding. For training, we propose novel geometrical cycle-consistency losses which construct cycles across 2D-3D spaces, across different instances and different time steps. The learned correspondence can be applied for 6D pose estimation and other downstream tasks such as keypoint transfer. Surprisingly, our method, without any human annotations or simulators, can achieve on-par or even better performance than previous supervised or semi-supervised methods on in-the-wild images. Our project page is: https://kywind.github.io/self-pose ."
                    },
                    {
                        "title": "X-Align: Cross-Modal Cross-View Alignment for Bird's-Eye-View Segmentation",
                        "abstract": "Bird's-eye-view (BEV) grid is a common representation for the perception of road components, e.g., drivable area, in autonomous driving. Most existing approaches rely on cameras only to perform segmentation in BEV space, which is fundamentally constrained by the absence of reliable depth information. Latest works leverage both camera and LiDAR modalities, but sub-optimally fuse their features using simple, concatenation-based mechanisms.   In this paper, we address these problems by enhancing the alignment of the unimodal features in order to aid feature fusion, as well as enhancing the alignment between the cameras' perspective view (PV) and BEV representations. We propose X-Align, a novel end-to-end cross-modal and cross-view learning framework for BEV segmentation consisting of the following components: (i) a novel Cross-Modal Feature Alignment (X-FA) loss, (ii) an attention-based Cross-Modal Feature Fusion (X-FF) module to align multi-modal BEV features implicitly, and (iii) an auxiliary PV segmentation branch with Cross-View Segmentation Alignment (X-SA) losses to improve the PV-to-BEV transformation. We evaluate our proposed method across two commonly used benchmark datasets, i.e., nuScenes and KITTI-360. Notably, X-Align significantly outperforms the state-of-the-art by 3 absolute mIoU points on nuScenes. We also provide extensive ablation studies to demonstrate the effectiveness of the individual components."
                    },
                    {
                        "title": "Perceptual Consistency in Video Segmentation",
                        "abstract": "In this paper, we present a novel perceptual consistency perspective on video semantic segmentation, which can capture both temporal consistency and pixel-wise correctness. Given two nearby video frames, perceptual consistency measures how much the segmentation decisions agree with the pixel correspondences obtained via matching general perceptual features. More specifically, for each pixel in one frame, we find the most perceptually correlated pixel in the other frame. Our intuition is that such a pair of pixels are highly likely to belong to the same class. Next, we assess how much the segmentation agrees with such perceptual correspondences, based on which we derive the perceptual consistency of the segmentation maps across these two frames. Utilizing perceptual consistency, we can evaluate the temporal consistency of video segmentation by measuring the perceptual consistency over consecutive pairs of segmentation maps in a video. Furthermore, given a sparsely labeled test video, perceptual consistency can be utilized to aid with predicting the pixel-wise correctness of the segmentation on an unlabeled frame. More specifically, by measuring the perceptual consistency between the predicted segmentation and the available ground truth on a nearby frame and combining it with the segmentation confidence, we can accurately assess the classification correctness on each pixel. Our experiments show that the proposed perceptual consistency can more accurately evaluate the temporal consistency of video segmentation as compared to flow-based measures. Furthermore, it can help more confidently predict segmentation accuracy on unlabeled test frames, as compared to using classification confidence alone. Finally, our proposed measure can be used as a regularizer during the training of segmentation models, which leads to more temporally consistent video segmentation while maintaining accuracy."
                    },
                    {
                        "title": "Multi-camera Bird's Eye View Perception for Autonomous Driving",
                        "abstract": "Most automated driving systems comprise a diverse sensor set, including several cameras, Radars, and LiDARs, ensuring a complete 360\\deg coverage in near and far regions. Unlike Radar and LiDAR, which measure directly in 3D, cameras capture a 2D perspective projection with inherent depth ambiguity. However, it is essential to produce perception outputs in 3D to enable the spatial reasoning of other agents and structures for optimal path planning. The 3D space is typically simplified to the BEV space by omitting the less relevant Z-coordinate, which corresponds to the height dimension.The most basic approach to achieving the desired BEV representation from a camera image is IPM, assuming a flat ground surface. Surround vision systems that are pretty common in new vehicles use the IPM principle to generate a BEV image and to show it on display to the driver. However, this approach is not suited for autonomous driving since there are severe distortions introduced by this too-simplistic transformation method. More recent approaches use deep neural networks to output directly in BEV space. These methods transform camera images into BEV space using geometric constraints implicitly or explicitly in the network. As CNN has more context information and a learnable transformation can be more flexible and adapt to image content, the deep learning-based methods set the new benchmark for BEV transformation and achieve state-of-the-art performance. First, this chapter discusses the contemporary trends of multi-camera-based DNN (deep neural network) models outputting object representations directly in the BEV space. Then, we discuss how this approach can extend to effective sensor fusion and coupling downstream tasks like situation analysis and prediction. Finally, we show challenges and open problems in BEV perception."
                    },
                    {
                        "title": "FouRA: Fourier Low Rank Adaptation",
                        "abstract": "While Low-Rank Adaptation (LoRA) has proven beneficial for efficiently fine-tuning large models, LoRA fine-tuned text-to-image diffusion models lack diversity in the generated images, as the model tends to copy data from the observed training samples. This effect becomes more pronounced at higher values of adapter strength and for adapters with higher ranks which are fine-tuned on smaller datasets. To address these challenges, we present FouRA, a novel low-rank method that learns projections in the Fourier domain along with learning a flexible input-dependent adapter rank selection strategy. Through extensive experiments and analysis, we show that FouRA successfully solves the problems related to data copying and distribution collapse while significantly improving the generated image quality. We demonstrate that FouRA enhances the generalization of fine-tuned models thanks to its adaptive rank selection. We further show that the learned projections in the frequency domain are decorrelated and prove effective when merging multiple adapters. While FouRA is motivated for vision tasks, we also demonstrate its merits for language tasks on the GLUE benchmark."
                    },
                    {
                        "title": "Simple and Efficient Architectures for Semantic Segmentation",
                        "abstract": "Though the state-of-the architectures for semantic segmentation, such as HRNet, demonstrate impressive accuracy, the complexity arising from their salient design choices hinders a range of model acceleration tools, and further they make use of operations that are inefficient on current hardware. This paper demonstrates that a simple encoder-decoder architecture with a ResNet-like backbone and a small multi-scale head, performs on-par or better than complex semantic segmentation architectures such as HRNet, FANet and DDRNets. Naively applying deep backbones designed for Image Classification to the task of Semantic Segmentation leads to sub-par results, owing to a much smaller effective receptive field of these backbones. Implicit among the various design choices put forth in works like HRNet, DDRNet, and FANet are networks with a large effective receptive field. It is natural to ask if a simple encoder-decoder architecture would compare favorably if comprised of backbones that have a larger effective receptive field, though without the use of inefficient operations like dilated convolutions. We show that with minor and inexpensive modifications to ResNets, enlarging the receptive field, very simple and competitive baselines can be created for Semantic Segmentation. We present a family of such simple architectures for desktop as well as mobile targets, which match or exceed the performance of complex models on the Cityscapes dataset. We hope that our work provides simple yet effective baselines for practitioners to develop efficient semantic segmentation models."
                    },
                    {
                        "title": "X$^3$KD: Knowledge Distillation Across Modalities, Tasks and Stages for Multi-Camera 3D Object Detection",
                        "abstract": "Recent advances in 3D object detection (3DOD) have obtained remarkably strong results for LiDAR-based models. In contrast, surround-view 3DOD models based on multiple camera images underperform due to the necessary view transformation of features from perspective view (PV) to a 3D world representation which is ambiguous due to missing depth information. This paper introduces X$^3$KD, a comprehensive knowledge distillation framework across different modalities, tasks, and stages for multi-camera 3DOD. Specifically, we propose cross-task distillation from an instance segmentation teacher (X-IS) in the PV feature extraction stage providing supervision without ambiguous error backpropagation through the view transformation. After the transformation, we apply cross-modal feature distillation (X-FD) and adversarial training (X-AT) to improve the 3D world representation of multi-camera features through the information contained in a LiDAR-based 3DOD teacher. Finally, we also employ this teacher for cross-modal output distillation (X-OD), providing dense supervision at the prediction stage. We perform extensive ablations of knowledge distillation at different stages of multi-camera 3DOD. Our final X$^3$KD model outperforms previous state-of-the-art approaches on the nuScenes and Waymo datasets and generalizes to RADAR-based 3DOD. Qualitative results video at https://youtu.be/1do9DPFmr38."
                    },
                    {
                        "title": "X-Align++: cross-modal cross-view alignment for Bird's-eye-view segmentation",
                        "abstract": "Bird's-eye-view (BEV) grid is a typical representation of the perception of road components, e.g., drivable area, in autonomous driving. Most existing approaches rely on cameras only to perform segmentation in BEV space, which is fundamentally constrained by the absence of reliable depth information. The latest works leverage both camera and LiDAR modalities but suboptimally fuse their features using simple, concatenation-based mechanisms. In this paper, we address these problems by enhancing the alignment of the unimodal features in order to aid feature fusion, as well as enhancing the alignment between the cameras' perspective view (PV) and BEV representations. We propose X-Align, a novel end-to-end cross-modal and cross-view learning framework for BEV segmentation consisting of the following components: (i) a novel Cross-Modal Feature Alignment (X-FA) loss, (ii) an attention-based Cross-Modal Feature Fusion (X-FF) module to align multi-modal BEV features implicitly, and (iii) an auxiliary PV segmentation branch with Cross-View Segmentation Alignment (X-SA) losses to improve the PV-to-BEV transformation. We evaluate our proposed method across two commonly used benchmark datasets, i.e., nuScenes and KITTI-360. Notably, X-Align significantly outperforms the state-of-the-art by 3 absolute mIoU points on nuScenes. We also provide extensive ablation studies to demonstrate the effectiveness of the individual components."
                    }
                ]
            },
            "cdbb67d3-7575-454f-979d-96651fe592b4": {
                "pk": "cdbb67d3-7575-454f-979d-96651fe592b4",
                "name": "Paul Whatmough",
                "collaborators": [
                    "Yuhao Zhu",
                    "Matthew Mattina",
                    "Chuteng Zhou",
                    "Yu Feng",
                    "Marco Donato",
                    "Glenn Ko",
                    "Sae-Kyu Lee",
                    "David Brooks",
                    "Gu-Yeon Wei",
                    "Anand Samajdar"
                ],
                "domain": [
                    "Hardware Design",
                    "Machine Learning",
                    "Neural Networks",
                    "Embedded Systems"
                ],
                "publications": [
                    {
                        "title": "CHIPKIT: An agile, reusable open-source framework for rapid test chip development",
                        "abstract": "The current trend for domain-specific architectures (DSAs) has led to renewed interest in research test chips to demonstrate new specialized hardware. Tape-outs also offer huge pedagogical value garnered from real hands-on exposure to the whole system stack. However, successful tape-outs demand hard-earned experience, and the design process is time consuming and fraught with challenges. Therefore, custom chips have remained the preserve of a small number of research groups, typically focused on circuit design research. This paper describes the CHIPKIT framework. We describe a reusable SoC subsystem which provides basic IO, an on-chip programmable host, memory and peripherals. This subsystem can be readily extended with new IP blocks to generate custom test chips. We also present an agile RTL development flow, including a code generation tool calledVGEN. Finally, we outline best practices for full-chip validation across the entire design cycle."
                    },
                    {
                        "title": "Mobile Machine Learning Hardware at ARM: A Systems-on-Chip (SoC) Perspective",
                        "abstract": "Machine learning is playing an increasingly significant role in emerging mobile application domains such as AR/VR, ADAS, etc. Accordingly, hardware architects have designed customized hardware for machine learning algorithms, especially neural networks, to improve compute efficiency. However, machine learning is typically just one processing stage in complex end-to-end applications, involving multiple components in a mobile Systems-on-a-chip (SoC). Focusing only on ML accelerators loses bigger optimization opportunity at the system (SoC) level. This paper argues that hardware architects should expand the optimization scope to the entire SoC. We demonstrate one particular case-study in the domain of continuous computer vision where camera sensor, image signal processor (ISP), memory, and NN accelerator are synergistically co-designed to achieve optimal system-level efficiency."
                    },
                    {
                        "title": "Euphrates: Algorithm-SoC Co-Design for Low-Power Mobile Continuous Vision",
                        "abstract": "Continuous computer vision (CV) tasks increasingly rely on convolutional neural networks (CNN). However, CNNs have massive compute demands that far exceed the performance and energy constraints of mobile devices. In this paper, we propose and develop an algorithm-architecture co-designed system, Euphrates, that simultaneously improves the energy-efficiency and performance of continuous vision tasks.   Our key observation is that changes in pixel data between consecutive frames represents visual motion. We first propose an algorithm that leverages this motion information to relax the number of expensive CNN inferences required by continuous vision applications. We co-design a mobile System-on-a-Chip (SoC) architecture to maximize the efficiency of the new algorithm. The key to our architectural augmentation is to co-optimize different SoC IP blocks in the vision pipeline collectively. Specifically, we propose to expose the motion data that is naturally generated by the Image Signal Processor (ISP) early in the vision pipeline to the CNN engine. Measurement and synthesis results show that Euphrates achieves up to 66% SoC-level energy savings (4 times for the vision computations), with only 1% accuracy loss."
                    },
                    {
                        "title": "Energy Efficient Hardware for On-Device CNN Inference via Transfer Learning",
                        "abstract": "On-device CNN inference for real-time computer vision applications can result in computational demands that far exceed the energy budgets of mobile devices. This paper proposes FixyNN, a co-designed hardware accelerator platform which splits a CNN model into two parts: a set of layers that are fixed in the hardware platform as a front-end fixed-weight feature extractor, and the remaining layers which become a back-end classifier running on a conventional programmable CNN accelerator. The common front-end provides ubiquitous CNN features for all FixyNN models, while the back-end is programmable and specific to a given dataset. Image classification models for FixyNN are trained end-to-end via transfer learning, with front-end layers fixed for the shared feature extractor, and back-end layers fine-tuned for a specific task. Over a suite of six datasets, we trained models via transfer learning with an accuracy loss of <1%, resulting in a FixyNN hardware platform with nearly 2 times better energy efficiency than a conventional programmable CNN accelerator of the same silicon area (i.e. hardware cost)."
                    },
                    {
                        "title": "ASV: Accelerated Stereo Vision System",
                        "abstract": "Estimating depth from stereo vision cameras, i.e., \"depth from stereo\", is critical to emerging intelligent applications deployed in energy- and performance-constrained devices, such as augmented reality headsets and mobile autonomous robots. While existing stereo vision systems make trade-offs between accuracy, performance and energy-efficiency, we describe ASV, an accelerated stereo vision system that simultaneously improves both performance and energy-efficiency while achieving high accuracy. The key to ASV is to exploit unique characteristics inherent to stereo vision, and apply stereo-specific optimizations, both algorithmically and computationally. We make two contributions. Firstly, we propose a new stereo algorithm, invariant-based stereo matching (ISM), that achieves significant speedup while retaining high accuracy. The algorithm combines classic \"hand-crafted\" stereo algorithms with recent developments in Deep Neural Networks (DNNs), by leveraging the correspondence invariant unique to stereo vision systems. Secondly, we observe that the bottleneck of the ISM algorithm is the DNN inference, and in particular the deconvolution operations that introduce massive compute-inefficiencies. We propose a set of software optimizations that mitigate these inefficiencies. We show that with less than 0.5% hardware area overhead, these algorithmic and computational optimizations can be effectively integrated within a conventional DNN accelerator. Overall, ASV achieves 5x speedup and 85% energy saving with 0.02% accuracy loss compared to today DNN-based stereo vision systems."
                    },
                    {
                        "title": "SCALE-Sim: Systolic CNN Accelerator Simulator",
                        "abstract": "Systolic Arrays are one of the most popular compute substrates within Deep Learning accelerators today, as they provide extremely high efficiency for running dense matrix multiplications. However, the research community lacks tools to insights on both the design trade-offs and efficient mapping strategies for systolic-array based accelerators. We introduce Systolic CNN Accelerator Simulator (SCALE-Sim), which is a configurable systolic array based cycle accurate DNN accelerator simulator. SCALE-Sim exposes various micro-architectural features as well as system integration parameters to the designer to enable comprehensive design space exploration. This is the first systolic-array simulator tuned for running DNNs to the best of our knowledge. Using SCALE-Sim, we conduct a suite of case studies and demonstrate the effect of bandwidth, data flow and aspect ratio on the overall runtime and energy of Deep Learning kernels across vision, speech, text, and games. We believe that these insights will be highly beneficial to architects and ML practitioners."
                    },
                    {
                        "title": "UDC: Unified DNAS for Compressible TinyML Models",
                        "abstract": "Deploying TinyML models on low-cost IoT hardware is very challenging, due to limited device memory capacity. Neural processing unit (NPU) hardware address the memory challenge by using model compression to exploit weight quantization and sparsity to fit more parameters in the same footprint. However, designing compressible neural networks (NNs) is challenging, as it expands the design space across which we must make balanced trade-offs. This paper demonstrates Unified DNAS for Compressible (UDC) NNs, which explores a large search space to generate state-of-the-art compressible NNs for NPU. ImageNet results show UDC networks are up to $3.35\\times$ smaller (iso-accuracy) or 6.25% more accurate (iso-model size) than previous work."
                    },
                    {
                        "title": "Mesorasi: Architecture Support for Point Cloud Analytics via Delayed-Aggregation",
                        "abstract": "Point cloud analytics is poised to become a key workload on battery-powered embedded and mobile platforms in a wide range of emerging application domains, such as autonomous driving, robotics, and augmented reality, where efficiency is paramount. This paper proposes Mesorasi, an algorithm-architecture co-designed system that simultaneously improves the performance and energy efficiency of point cloud analytics while retaining its accuracy. Our extensive characterizations of state-of-the-art point cloud algorithms show that, while structurally reminiscent of convolutional neural networks (CNNs), point cloud algorithms exhibit inherent compute and memory inefficiencies due to the unique characteristics of point cloud data. We propose delayed-aggregation, a new algorithmic primitive for building efficient point cloud algorithms. Delayed-aggregation hides the performance bottlenecks and reduces the compute and memory redundancies by exploiting the approximately distributive property of key operations in point cloud algorithms. Delayed-aggregation let point cloud algorithms achieve 1.6x speedup and 51.1% energy reduction on a mobile GPU while retaining the accuracy (-0.9% loss to 1.2% gains). To maximize the algorithmic benefits, we propose minor extensions to contemporary CNN accelerators, which can be integrated into a mobile Systems-on-a-Chip (SoC) without modifying other SoC components. With additional hardware support, Mesorasi achieves up to 3.6x speedup."
                    },
                    {
                        "title": "Super-Efficient Super Resolution for Fast Adversarial Defense at the Edge",
                        "abstract": "Autonomous systems are highly vulnerable to a variety of adversarial attacks on Deep Neural Networks (DNNs). Training-free model-agnostic defenses have recently gained popularity due to their speed, ease of deployment, and ability to work across many DNNs. To this end, a new technique has emerged for mitigating attacks on image classification DNNs, namely, preprocessing adversarial images using super resolution -- upscaling low-quality inputs into high-resolution images. This defense requires running both image classifiers and super resolution models on constrained autonomous systems. However, super resolution incurs a heavy computational cost. Therefore, in this paper, we investigate the following question: Does the robustness of image classifiers suffer if we use tiny super resolution models? To answer this, we first review a recent work called Super-Efficient Super Resolution (SESR) that achieves similar or better image quality than prior art while requiring 2x to 330x fewer Multiply-Accumulate (MAC) operations. We demonstrate that despite being orders of magnitude smaller than existing models, SESR achieves the same level of robustness as significantly larger networks. Finally, we estimate end-to-end performance of super resolution-based defenses on a commercial Arm Ethos-U55 micro-NPU. Our findings show that SESR achieves nearly 3x higher FPS than a baseline while achieving similar robustness."
                    },
                    {
                        "title": "Characterizing Soft-Error Resiliency in Arm's Ethos-U55 Embedded Machine Learning Accelerator",
                        "abstract": "As Neural Processing Units (NPU) or accelerators are increasingly deployed in a variety of applications including safety critical applications such as autonomous vehicle, and medical imaging, it is critical to understand the fault-tolerance nature of the NPUs. We present a reliability study of Arm's Ethos-U55, an important industrial-scale NPU being utilised in embedded and IoT applications. We perform large scale RTL-level fault injections to characterize Ethos-U55 against the Automotive Safety Integrity Level D (ASIL-D) resiliency standard commonly used for safety-critical applications such as autonomous vehicles. We show that, under soft errors, all four configurations of the NPU fall short of the required level of resiliency for a variety of neural networks running on the NPU. We show that it is possible to meet the ASIL-D level resiliency without resorting to conventional strategies like Dual Core Lock Step (DCLS) that has an area overhead of 100%. We achieve so through selective protection, where hardware structures are selectively protected (e.g., duplicated, hardened) based on their sensitivity to soft errors and their silicon areas. To identify the optimal configuration that minimizes the area overhead while meeting the ASIL-D standard, the main challenge is the large search space associated with the time-consuming RTL simulation. To address this challenge, we present a statistical analysis tool that is validated against Arm silicon and that allows us to quickly navigate hundreds of billions of fault sites without exhaustive RTL fault injections. We show that by carefully duplicating a small fraction of the functional blocks and hardening the Flops in other blocks meets the ASIL-D safety standard while introducing an area overhead of only 38%."
                    },
                    {
                        "title": "GPTVQ: The Blessing of Dimensionality for LLM Quantization",
                        "abstract": "In this work we show that the size versus accuracy trade-off of neural network quantization can be significantly improved by increasing the quantization dimensionality. We propose the GPTVQ method, a new fast method for post-training vector quantization (VQ) that scales well to Large Language Models (LLMs). Our method interleaves quantization of one or more columns with updates to the remaining unquantized weights, using information from the Hessian of the per-layer output reconstruction MSE. Quantization codebooks are initialized using an efficient data-aware version of the EM algorithm. The codebooks are then updated, and further compressed by using integer quantization and SVD-based compression. GPTVQ establishes a new state-of-the art in the size vs accuracy trade-offs on a wide range of LLMs such as Llama-v2 and Mistral. Furthermore, our method is efficient: on a single H100 it takes between 3 and 11 hours to process a Llamav2-70B model, depending on quantization setting. Lastly, with on-device timings for VQ decompression on a mobile CPU we show that VQ leads to improved latency compared to using a 4-bit integer format."
                    }
                ]
            },
            "fd593afc-8397-4ab3-984c-419aee7e4f82": {
                "pk": "fd593afc-8397-4ab3-984c-419aee7e4f82",
                "name": "Risheek Garrepalli",
                "collaborators": [
                    "Fatih Porikli",
                    "Hong Cai",
                    "Jisoo Jeong",
                    "Shubhankar Borse",
                    "Munawar Hayat",
                    "Jamie Menjay Lin",
                    "Debasmit Das",
                    "Hyojin Park",
                    "Rajeev Yasarla",
                    "Yunxiao Shi"
                ],
                "domain": [
                    "Open Set Detection",
                    "Optical Flow Estimation",
                    "Deep Learning",
                    "Data Augmentation"
                ],
                "publications": [
                    {
                        "title": "Oracle Analysis of Representations for Deep Open Set Detection",
                        "abstract": "The problem of detecting a novel class at run time is known as Open Set Detection & is important for various real-world applications like medical application, autonomous driving, etc. Open Set Detection within context of deep learning involves solving two problems: (i) Must map the input images into a latent representation that contains enough information to detect the outliers, and (ii) Must learn an anomaly scoring function that can extract this information from the latent representation to identify the anomalies. Research in deep anomaly detection methods has progressed slowly. One reason may be that most papers simultaneously introduce new representation learning techniques and new anomaly scoring approaches. The goal of this work is to improve this methodology by providing ways of separately measuring the effectiveness of the representation learning and anomaly scoring. This work makes two methodological contributions. The first is to introduce the notion of Oracle anomaly detection for quantifying the information available in a learned latent representation. The second is to introduce Oracle representation learning, which produces a representation that is guaranteed to be sufficient for accurate anomaly detection. These two techniques help researchers to separate the quality of the learned representation from the performance of the anomaly scoring mechanism so that they can debug and improve their systems. The methods also provide an upper limit on how much open category detection can be improved through better anomaly scoring mechanisms. The combination of the two oracles gives an upper limit on the performance that any open category detection method could achieve. This work introduces these two oracle techniques and demonstrates their utility by applying them to several leading open category detection methods."
                    },
                    {
                        "title": "DDIL: Improved Diffusion Distillation With Imitation Learning",
                        "abstract": "Diffusion models excel at generative modeling (e.g., text-to-image) but sampling requires multiple denoising network passes, limiting practicality. Efforts such as progressive distillation or consistency distillation have shown promise by reducing the number of passes at the expense of quality of the generated samples. In this work we identify co-variate shift as one of reason for poor performance of multi-step distilled models from compounding error at inference time. To address co-variate shift, we formulate diffusion distillation within imitation learning (DDIL) framework and enhance training distribution for distilling diffusion models on both data distribution (forward diffusion) and student induced distributions (backward diffusion). Training on data distribution helps to diversify the generations by preserving marginal data distribution and training on student distribution addresses compounding error by correcting covariate shift. In addition, we adopt reflected diffusion formulation for distillation and demonstrate improved performance, stable training across different distillation methods. We show that DDIL consistency improves on baseline algorithms of progressive distillation (PD), Latent consistency models (LCM) and Distribution Matching Distillation (DMD2)."
                    },
                    {
                        "title": "DistractFlow: Improving Optical Flow Estimation via Realistic Distractions and Pseudo-Labeling",
                        "abstract": "We propose a novel data augmentation approach, DistractFlow, for training optical flow estimation models by introducing realistic distractions to the input frames. Based on a mixing ratio, we combine one of the frames in the pair with a distractor image depicting a similar domain, which allows for inducing visual perturbations congruent with natural objects and scenes. We refer to such pairs as distracted pairs. Our intuition is that using semantically meaningful distractors enables the model to learn related variations and attain robustness against challenging deviations, compared to conventional augmentation schemes focusing only on low-level aspects and modifications. More specifically, in addition to the supervised loss computed between the estimated flow for the original pair and its ground-truth flow, we include a second supervised loss defined between the distracted pair's flow and the original pair's ground-truth flow, weighted with the same mixing ratio. Furthermore, when unlabeled data is available, we extend our augmentation approach to self-supervised settings through pseudo-labeling and cross-consistency regularization. Given an original pair and its distracted version, we enforce the estimated flow on the distracted pair to agree with the flow of the original pair. Our approach allows increasing the number of available training pairs significantly without requiring additional annotations. It is agnostic to the model architecture and can be applied to training any optical flow estimation models. Our extensive evaluations on multiple benchmarks, including Sintel, KITTI, and SlowFlow, show that DistractFlow improves existing models consistently, outperforming the latest state of the art."
                    },
                    {
                        "title": "DIFT: Dynamic Iterative Field Transforms for Memory Efficient Optical Flow",
                        "abstract": "Recent advancements in neural network-based optical flow estimation often come with prohibitively high computational and memory requirements, presenting challenges in their model adaptation for mobile and low-power use cases. In this paper, we introduce a lightweight low-latency and memory-efficient model, Dynamic Iterative Field Transforms (DIFT), for optical flow estimation feasible for edge applications such as mobile, XR, micro UAVs, robotics and cameras. DIFT follows an iterative refinement framework leveraging variable resolution of cost volumes for correspondence estimation. We propose a memory efficient solution for cost volume processing to reduce peak memory. Also, we present a novel dynamic coarse-to-fine cost volume processing during various stages of refinement to avoid multiple levels of cost volumes. We demonstrate first real-time cost-volume based optical flow DL architecture on Snapdragon 8 Gen 1 HTP efficient mobile AI accelerator with 32 inf/sec and 5.89 EPE (endpoint error) on KITTI with manageable accuracy-performance tradeoffs."
                    },
                    {
                        "title": "TransAdapt: A Transformative Framework for Online Test Time Adaptive Semantic Segmentation",
                        "abstract": "Test-time adaptive (TTA) semantic segmentation adapts a source pre-trained image semantic segmentation model to unlabeled batches of target domain test images, different from real-world, where samples arrive one-by-one in an online fashion. To tackle online settings, we propose TransAdapt, a framework that uses transformer and input transformations to improve segmentation performance. Specifically, we pre-train a transformer-based module on a segmentation network that transforms unsupervised segmentation output to a more reliable supervised output, without requiring test-time online training. To also facilitate test-time adaptation, we propose an unsupervised loss based on the transformed input that enforces the model to be invariant and equivariant to photometric and geometric perturbations, respectively. Overall, our framework produces higher quality segmentation masks with up to 17.6% and 2.8% mIOU improvement over no-adaptation and competitive baselines, respectively."
                    },
                    {
                        "title": "Clockwork Diffusion: Efficient Generation With Model-Step Distillation",
                        "abstract": "This work aims to improve the efficiency of text-to-image diffusion models. While diffusion models use computationally expensive UNet-based denoising operations in every generation step, we identify that not all operations are equally relevant for the final output quality. In particular, we observe that UNet layers operating on high-res feature maps are relatively sensitive to small perturbations. In contrast, low-res feature maps influence the semantic layout of the final image and can often be perturbed with no noticeable change in the output. Based on this observation, we propose Clockwork Diffusion, a method that periodically reuses computation from preceding denoising steps to approximate low-res feature maps at one or more subsequent steps. For multiple baselines, and for both text-to-image generation and image editing, we demonstrate that Clockwork leads to comparable or improved perceptual scores with drastically reduced computational complexity. As an example, for Stable Diffusion v1.5 with 8 DPM++ steps we save 32% of FLOPs with negligible FID and CLIP change."
                    },
                    {
                        "title": "OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation",
                        "abstract": "The scarcity of ground-truth labels poses one major challenge in developing optical flow estimation models that are both generalizable and robust. While current methods rely on data augmentation, they have yet to fully exploit the rich information available in labeled video sequences. We propose OCAI, a method that supports robust frame interpolation by generating intermediate video frames alongside optical flows in between. Utilizing a forward warping approach, OCAI employs occlusion awareness to resolve ambiguities in pixel values and fills in missing values by leveraging the forward-backward consistency of optical flows. Additionally, we introduce a teacher-student style semi-supervised learning method on top of the interpolated frames. Using a pair of unlabeled frames and the teacher model's predicted optical flow, we generate interpolated frames and flows to train a student model. The teacher's weights are maintained using Exponential Moving Averaging of the student. Our evaluations demonstrate perceptually superior interpolation quality and enhanced optical flow accuracy on established benchmarks such as Sintel and KITTI."
                    },
                    {
                        "title": "Open Category Detection with PAC Guarantees",
                        "abstract": "Open category detection is the problem of detecting \"alien\" test instances that belong to categories or classes that were not present in the training data. In many applications, reliably detecting such aliens is central to ensuring the safety and accuracy of test set predictions. Unfortunately, there are no algorithms that provide theoretical guarantees on their ability to detect aliens under general assumptions. Further, while there are algorithms for open category detection, there are few empirical results that directly report alien detection rates. Thus, there are significant theoretical and empirical gaps in our understanding of open category detection. In this paper, we take a step toward addressing this gap by studying a simple, but practically-relevant variant of open category detection. In our setting, we are provided with a \"clean\" training set that contains only the target categories of interest and an unlabeled \"contaminated\" training set that contains a fraction $\\alpha$ of alien examples. Under the assumption that we know an upper bound on $\\alpha$, we develop an algorithm with PAC-style guarantees on the alien detection rate, while aiming to minimize false alarms. Empirical results on synthetic and standard benchmark datasets demonstrate the regimes in which the algorithm can be effective and provide a baseline for further advancements."
                    },
                    {
                        "title": "Panoptic, Instance and Semantic Relations: A Relational Context Encoder to Enhance Panoptic Segmentation",
                        "abstract": "This paper presents a novel framework to integrate both semantic and instance contexts for panoptic segmentation. In existing works, it is common to use a shared backbone to extract features for both things (countable classes such as vehicles) and stuff (uncountable classes such as roads). This, however, fails to capture the rich relations among them, which can be utilized to enhance visual understanding and segmentation performance. To address this shortcoming, we propose a novel Panoptic, Instance, and Semantic Relations (PISR) module to exploit such contexts. First, we generate panoptic encodings to summarize key features of the semantic classes and predicted instances. A Panoptic Relational Attention (PRA) module is then applied to the encodings and the global feature map from the backbone. It produces a feature map that captures 1) the relations across semantic classes and instances and 2) the relations between these panoptic categories and spatial features. PISR also automatically learns to focus on the more important instances, making it robust to the number of instances used in the relational attention module. Moreover, PISR is a general module that can be applied to any existing panoptic segmentation architecture. Through extensive evaluations on panoptic segmentation benchmarks like Cityscapes, COCO, and ADE20K, we show that PISR attains considerable improvements over existing approaches."
                    },
                    {
                        "title": "DejaVu: Conditional Regenerative Learning to Enhance Dense Prediction",
                        "abstract": "We present DejaVu, a novel framework which leverages conditional image regeneration as additional supervision during training to improve deep networks for dense prediction tasks such as segmentation, depth estimation, and surface normal prediction. First, we apply redaction to the input image, which removes certain structural information by sparse sampling or selective frequency removal. Next, we use a conditional regenerator, which takes the redacted image and the dense predictions as inputs, and reconstructs the original image by filling in the missing structural information. In the redacted image, structural attributes like boundaries are broken while semantic context is largely preserved. In order to make the regeneration feasible, the conditional generator will then require the structure information from the other input source, i.e., the dense predictions. As such, by including this conditional regeneration objective during training, DejaVu encourages the base network to learn to embed accurate scene structure in its dense prediction. This leads to more accurate predictions with clearer boundaries and better spatial consistency. When it is feasible to leverage additional computation, DejaVu can be extended to incorporate an attention-based regeneration module within the dense prediction network, which further improves accuracy. Through extensive experiments on multiple dense prediction benchmarks such as Cityscapes, COCO, ADE20K, NYUD-v2, and KITTI, we demonstrate the efficacy of employing DejaVu during training, as it outperforms SOTA methods at no added computation cost."
                    },
                    {
                        "title": "MAMo: Leveraging Memory and Attention for Monocular Video Depth Estimation",
                        "abstract": "We propose MAMo, a novel memory and attention frame-work for monocular video depth estimation. MAMo can augment and improve any single-image depth estimation networks into video depth estimation models, enabling them to take advantage of the temporal information to predict more accurate depth. In MAMo, we augment model with memory which aids the depth prediction as the model streams through the video. Specifically, the memory stores learned visual and displacement tokens of the previous time instances. This allows the depth network to cross-reference relevant features from the past when predicting depth on the current frame. We introduce a novel scheme to continuously update the memory, optimizing it to keep tokens that correspond with both the past and the present visual information. We adopt attention-based approach to process memory features where we first learn the spatio-temporal relation among the resultant visual and displacement memory tokens using self-attention module. Further, the output features of self-attention are aggregated with the current visual features through cross-attention. The cross-attended features are finally given to a decoder to predict depth on the current frame. Through extensive experiments on several benchmarks, including KITTI, NYU-Depth V2, and DDAD, we show that MAMo consistently improves monocular depth estimation networks and sets new state-of-the-art (SOTA) accuracy. Notably, our MAMo video depth estimation provides higher accuracy with lower latency, when omparing to SOTA cost-volume-based video depth models."
                    },
                    {
                        "title": "SciFlow: Empowering Lightweight Optical Flow Models with Self-Cleaning Iterations",
                        "abstract": "Optical flow estimation is crucial to a variety of vision tasks. Despite substantial recent advancements, achieving real-time on-device optical flow estimation remains a complex challenge. First, an optical flow model must be sufficiently lightweight to meet computation and memory constraints to ensure real-time performance on devices. Second, the necessity for real-time on-device operation imposes constraints that weaken the model's capacity to adequately handle ambiguities in flow estimation, thereby intensifying the difficulty of preserving flow accuracy. This paper introduces two synergistic techniques, Self-Cleaning Iteration (SCI) and Regression Focal Loss (RFL), designed to enhance the capabilities of optical flow models, with a focus on addressing optical flow regression ambiguities. These techniques prove particularly effective in mitigating error propagation, a prevalent issue in optical flow models that employ iterative refinement. Notably, these techniques add negligible to zero overhead in model parameters and inference latency, thereby preserving real-time on-device efficiency. The effectiveness of our proposed SCI and RFL techniques, collectively referred to as SciFlow for brevity, is demonstrated across two distinct lightweight optical flow model architectures in our experiments. Remarkably, SciFlow enables substantial reduction in error metrics (EPE and Fl-all) over the baseline models by up to 6.3% and 10.5% for in-domain scenarios and by up to 6.2% and 13.5% for cross-domain scenarios on the Sintel and KITTI 2015 datasets, respectively."
                    },
                    {
                        "title": "FutureDepth: Learning to Predict the Future Improves Video Depth Estimation",
                        "abstract": "In this paper, we propose a novel video depth estimation approach, FutureDepth, which enables the model to implicitly leverage multi-frame and motion cues to improve depth estimation by making it learn to predict the future at training. More specifically, we propose a future prediction network, F-Net, which takes the features of multiple consecutive frames and is trained to predict multi-frame features one time step ahead iteratively. In this way, F-Net learns the underlying motion and correspondence information, and we incorporate its features into the depth decoding process. Additionally, to enrich the learning of multiframe correspondence cues, we further leverage a reconstruction network, R-Net, which is trained via adaptively masked auto-encoding of multiframe feature volumes. At inference time, both F-Net and R-Net are used to produce queries to work with the depth decoder, as well as a final refinement network. Through extensive experiments on several benchmarks, i.e., NYUDv2, KITTI, DDAD, and Sintel, which cover indoor, driving, and open-domain scenarios, we show that FutureDepth significantly improves upon baseline models, outperforms existing video depth estimation methods, and sets new state-of-the-art (SOTA) accuracy. Furthermore, FutureDepth is more efficient than existing SOTA video depth estimation models and has similar latencies when comparing to monocular models"
                    },
                    {
                        "title": "FouRA: Fourier Low Rank Adaptation",
                        "abstract": "While Low-Rank Adaptation (LoRA) has proven beneficial for efficiently fine-tuning large models, LoRA fine-tuned text-to-image diffusion models lack diversity in the generated images, as the model tends to copy data from the observed training samples. This effect becomes more pronounced at higher values of adapter strength and for adapters with higher ranks which are fine-tuned on smaller datasets. To address these challenges, we present FouRA, a novel low-rank method that learns projections in the Fourier domain along with learning a flexible input-dependent adapter rank selection strategy. Through extensive experiments and analysis, we show that FouRA successfully solves the problems related to data copying and distribution collapse while significantly improving the generated image quality. We demonstrate that FouRA enhances the generalization of fine-tuned models thanks to its adaptive rank selection. We further show that the learned projections in the frequency domain are decorrelated and prove effective when merging multiple adapters. While FouRA is motivated for vision tasks, we also demonstrate its merits for language tasks on the GLUE benchmark."
                    },
                    {
                        "title": "Rapid Switching and Multi-Adapter Fusion via Sparse High Rank Adapters",
                        "abstract": "In this paper, we propose Sparse High Rank Adapters (SHiRA) that directly finetune 1-2% of the base model weights while leaving others unchanged, thus, resulting in a highly sparse adapter. This high sparsity incurs no inference overhead, enables rapid switching directly in the fused mode, and significantly reduces concept-loss during multi-adapter fusion. Our extensive experiments on LVMs and LLMs demonstrate that finetuning merely 1-2% parameters in the base model is sufficient for many adapter tasks and significantly outperforms Low Rank Adaptation (LoRA). We also show that SHiRA is orthogonal to advanced LoRA methods such as DoRA and can be easily combined with existing techniques."
                    }
                ]
            },
            "a279d2e7-a258-4721-bf65-9711a15ba94c": {
                "pk": "a279d2e7-a258-4721-bf65-9711a15ba94c",
                "name": "Mart Van Baalen",
                "collaborators": [
                    "Markus Nagel",
                    "Andrey Kuzmin",
                    "Yuwei Ren",
                    "Jorn Peters",
                    "Tijmen Blankevoort",
                    "Kartikeya Bhardwaj",
                    "Nilesh Prasad Pandey",
                    "Sweta Priyadarshi",
                    "Viswanath Ganapathy",
                    "Rafael Esteves"
                ],
                "domain": [
                    "Neural Network Quantization",
                    "Sparse Adaptation",
                    "Efficient Inference"
                ],
                "publications": [
                    {
                        "title": "FP8 Quantization: The Power of the Exponent",
                        "abstract": "When quantizing neural networks for efficient inference, low-bit integers are the go-to format for efficiency. However, low-bit floating point numbers have an extra degree of freedom, assigning some bits to work on an exponential scale instead. This paper in-depth investigates this benefit of the floating point format for neural network inference. We detail the choices that can be made for the FP8 format, including the important choice of the number of bits for the mantissa and exponent, and show analytically in which settings these choices give better performance. Then we show how these findings translate to real networks, provide an efficient implementation for FP8 simulation, and a new algorithm that enables the learning of both the scale parameters and the number of exponent bits in the FP8 format. Our chief conclusion is that when doing post-training quantization for a wide range of networks, the FP8 format is better than INT8 in terms of accuracy, and the choice of the number of exponent bits is driven by the severity of outliers in the network. We also conduct experiments with quantization-aware training where the difference in formats disappears as the network is trained to reduce the effect of outliers."
                    },
                    {
                        "title": "Rapid Switching and Multi-Adapter Fusion via Sparse High Rank Adapters",
                        "abstract": "In this paper, we propose Sparse High Rank Adapters (SHiRA) that directly finetune 1-2% of the base model weights while leaving others unchanged, thus, resulting in a highly sparse adapter. This high sparsity incurs no inference overhead, enables rapid switching directly in the fused mode, and significantly reduces concept-loss during multi-adapter fusion. Our extensive experiments on LVMs and LLMs demonstrate that finetuning merely 1-2% parameters in the base model is sufficient for many adapter tasks and significantly outperforms Low Rank Adaptation (LoRA). We also show that SHiRA is orthogonal to advanced LoRA methods such as DoRA and can be easily combined with existing techniques."
                    }
                ]
            },
            "a0020d9a-ac25-47b5-8fb8-0ef8a4bdad7a": {
                "pk": "a0020d9a-ac25-47b5-8fb8-0ef8a4bdad7a",
                "name": "Harris Teague",
                "collaborators": [
                    "Burak Bartan",
                    "Haoming Li",
                    "Christopher Lott",
                    "Bistra Dilkina",
                    "Kartikeya Bhardwaj",
                    "Nilesh Prasad Pandey",
                    "Sweta Priyadarshi",
                    "Kyunggeun Lee",
                    "Jun Ma"
                ],
                "domain": [
                    "Edge Computing",
                    "Neural Networks",
                    "Attitude Estimation",
                    "Knowledge Distillation"
                ],
                "publications": [
                    {
                        "title": "Comparison of Attitude Estimation Techniques for Low-cost Unmanned Aerial Vehicles",
                        "abstract": "Attitude estimation for small, low-cost unmanned aerial vehicles is often achieved using a relatively simple complementary filter that combines onboard accelerometers, gyroscopes, and magnetometer sensing. This paper explores the limits of performance of such attitude estimation, with a focus on performance in highly dynamic maneuvers. The complementary filter is derived along with the extended Kalman filter and unscented Kalman filter to evaluate the potential performance gains when using a more sophisticated estimator. Simulations are presented that compare performance across a range of test cases, many where ground truth was generated from manually controlled flights in a flight simulator. Estimator scenarios that are generic across the different estimator types (such as the way sensor information is processed, and the use of dynamically changing gains) are compared across the test cases. An appendix is included as a quick reference for the common attitude representations and their kinematic expressions."
                    },
                    {
                        "title": "Moccasin: Efficient Tensor Rematerialization for Neural Networks",
                        "abstract": "The deployment and training of neural networks on edge computing devices pose many challenges. The low memory nature of edge devices is often one of the biggest limiting factors encountered in the deployment of large neural network models. Tensor rematerialization or recompute is a way to address high memory requirements for neural network training and inference. In this paper we consider the problem of execution time minimization of compute graphs subject to a memory budget. In particular, we develop a new constraint programming formulation called \\textsc{Moccasin} with only $O(n)$ integer variables, where $n$ is the number of nodes in the compute graph. This is a significant improvement over the works in the recent literature that propose formulations with $O(n^2)$ Boolean variables. We present numerical studies that show that our approach is up to an order of magnitude faster than recent work especially for large-scale graphs."
                    },
                    {
                        "title": "Oh! We Freeze: Improving Quantized Knowledge Distillation via Signal Propagation Analysis for Large Language Models",
                        "abstract": "Large generative models such as large language models (LLMs) and diffusion models have revolutionized the fields of NLP and computer vision respectively. However, their slow inference, high computation and memory requirement makes it challenging to deploy them on edge devices. In this study, we propose a light-weight quantization aware fine tuning technique using knowledge distillation (KD-QAT) to improve the performance of 4-bit weight quantized LLMs using commonly available datasets to realize a popular language use case, on device chat applications. To improve this paradigm of finetuning, as main contributions, we provide insights into stability of KD-QAT by empirically studying the gradient propagation during training to better understand the vulnerabilities of KD-QAT based approaches to low-bit quantization errors. Based on our insights, we propose ov-freeze, a simple technique to stabilize the KD-QAT process. Finally, we experiment with the popular 7B LLaMAv2-Chat model at 4-bit quantization level and demonstrate that ov-freeze results in near floating point precision performance, i.e., less than 0.7% loss of accuracy on Commonsense Reasoning benchmarks."
                    }
                ]
            },
            "79a10597-4af5-423d-8bb3-edadfaee99ac": {
                "pk": "79a10597-4af5-423d-8bb3-edadfaee99ac",
                "name": "Markus Nagel",
                "collaborators": [
                    "Tijmen Blankevoort",
                    "Mart van Baalen",
                    "Marios Fournarakis",
                    "Yelysei Bondarenko",
                    "Andrey Kuzmin",
                    "Max Welling",
                    "Christos Louizos",
                    "Arash Behboodi",
                    "Minseop Park",
                    "Jaeseong You"
                ],
                "domain": [
                    "Quantization",
                    "Neural Network Compression",
                    "Federated Learning",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "In-Hindsight Quantization Range Estimation for Quantized Training",
                        "abstract": "Quantization techniques applied to the inference of deep neural networks have enabled fast and efficient execution on resource-constraint devices. The success of quantization during inference has motivated the academic community to explore fully quantized training, i.e. quantizing back-propagation as well. However, effective gradient quantization is still an open problem. Gradients are unbounded and their distribution changes significantly during training, which leads to the need for dynamic quantization. As we show, dynamic quantization can lead to significant memory overhead and additional data traffic slowing down training. We propose a simple alternative to dynamic quantization, in-hindsight range estimation, that uses the quantization ranges estimated on previous iterations to quantize the present. Our approach enables fast static quantization of gradients and activations while requiring only minimal hardware support from the neural network accelerator to keep track of output statistics in an online fashion. It is intended as a drop-in replacement for estimating quantization ranges and can be used in conjunction with other advances in quantized training. We compare our method to existing methods for range estimation from the quantized training literature and demonstrate its effectiveness with a range of architectures, including MobileNetV2, on image classification benchmarks (Tiny ImageNet & ImageNet)."
                    },
                    {
                        "title": "Overcoming Oscillations in Quantization-Aware Training",
                        "abstract": "When training neural networks with simulated quantization, we observe that quantized weights can, rather unexpectedly, oscillate between two grid-points. The importance of this effect and its impact on quantization-aware training (QAT) are not well-understood or investigated in literature. In this paper, we delve deeper into the phenomenon of weight oscillations and show that it can lead to a significant accuracy degradation due to wrongly estimated batch-normalization statistics during inference and increased noise during training. These effects are particularly pronounced in low-bit ($\\leq$ 4-bits) quantization of efficient networks with depth-wise separable layers, such as MobileNets and EfficientNets. In our analysis we investigate several previously proposed QAT algorithms and show that most of these are unable to overcome oscillations. Finally, we propose two novel QAT algorithms to overcome oscillations during training: oscillation dampening and iterative weight freezing. We demonstrate that our algorithms achieve state-of-the-art accuracy for low-bit (3 & 4 bits) weight and activation quantization of efficient architectures, such as MobileNetV2, MobileNetV3, and EfficentNet-lite on ImageNet. Our source code is available at {https://github.com/qualcomm-ai-research/oscillations-qat}."
                    },
                    {
                        "title": "Understanding and Overcoming the Challenges of Efficient Transformer Quantization",
                        "abstract": "Transformer-based architectures have become the de-facto standard models for a wide range of Natural Language Processing tasks. However, their memory footprint and high latency are prohibitive for efficient deployment and inference on resource-limited devices. In this work, we explore quantization for transformers. We show that transformers have unique quantization challenges -- namely, high dynamic activation ranges that are difficult to represent with a low bit fixed-point format. We establish that these activations contain structured outliers in the residual connections that encourage specific attention patterns, such as attending to the special separator token. To combat these challenges, we present three solutions based on post-training quantization and quantization-aware training, each with a different set of compromises for accuracy, model size, and ease of use. In particular, we introduce a novel quantization scheme -- per-embedding-group quantization. We demonstrate the effectiveness of our methods on the GLUE benchmark using BERT, establishing state-of-the-art results for post-training quantization. Finally, we show that transformer weights and embeddings can be quantized to ultra-low bit-widths, leading to significant memory savings with a minimum accuracy loss. Our source code is available at~\\url{https://github.com/qualcomm-ai-research/transformer-quantization}."
                    },
                    {
                        "title": "Quadapter: Adapter for GPT-2 Quantization",
                        "abstract": "Transformer language models such as GPT-2 are difficult to quantize because of outliers in activations leading to a large quantization error. To adapt to the error, one must use quantization-aware training, which entails a fine-tuning process based on the dataset and the training pipeline identical to those for the original model. Pretrained language models, however, often do not grant access to their datasets and training pipelines, forcing us to rely on arbitrary ones for fine-tuning. In that case, it is observed that quantization-aware training overfits the model to the fine-tuning data. For quantization without overfitting, we introduce a quantization adapter (Quadapter), a small set of parameters that are learned to make activations quantization-friendly by scaling them channel-wise. It keeps the model parameters unchanged. By applying our method to the challenging task of quantizing GPT-2, we demonstrate that it effectively prevents the overfitting and improves the quantization performance."
                    },
                    {
                        "title": "Data-Free Quantization Through Weight Equalization and Bias Correction",
                        "abstract": "We introduce a data-free quantization method for deep neural networks that does not require fine-tuning or hyperparameter selection. It achieves near-original model performance on common computer vision architectures and tasks. 8-bit fixed-point quantization is essential for efficient inference on modern deep learning hardware. However, quantizing models to run in 8-bit is a non-trivial task, frequently leading to either significant performance reduction or engineering time spent on training a network to be amenable to quantization. Our approach relies on equalizing the weight ranges in the network by making use of a scale-equivariance property of activation functions. In addition the method corrects biases in the error that are introduced during quantization. This improves quantization accuracy performance, and can be applied to many common computer vision architectures with a straight forward API call. For common architectures, such as the MobileNet family, we achieve state-of-the-art quantized model performance. We further show that the method also extends to other computer vision architectures and tasks such as semantic segmentation and object detection."
                    },
                    {
                        "title": "Quantizable Transformers: Removing Outliers by Helping Attention Heads Do Nothing",
                        "abstract": "Transformer models have been widely adopted in various domains over the last years, and especially large language models have advanced the field of AI significantly. Due to their size, the capability of these networks has increased tremendously, but this has come at the cost of a significant increase in necessary compute. Quantization is one of the most effective ways to reduce the computational time and memory consumption of neural networks. Many studies have shown, however, that modern transformer models tend to learn strong outliers in their activations, making them difficult to quantize. To retain acceptable performance, the existence of these outliers requires activations to be in higher bitwidth or the use of different numeric formats, extra fine-tuning, or other workarounds. We show that strong outliers are related to very specific behavior of attention heads that try to learn a \"no-op\" or just a partial update of the residual. To achieve the exact zeros needed in the attention matrix for a no-update, the input to the softmax is pushed to be larger and larger during training, causing outliers in other parts of the network. Based on these observations, we propose two simple (independent) modifications to the attention mechanism - clipped softmax and gated attention. We empirically show that models pre-trained using our methods learn significantly smaller outliers while maintaining and sometimes even improving the floating-point task performance. This enables us to quantize transformers to full INT8 quantization of the activations without any additional effort. We demonstrate the effectiveness of our methods on both language models (BERT, OPT) and vision transformers."
                    },
                    {
                        "title": "Taxonomy and Evaluation of Structured Compression of Convolutional Neural Networks",
                        "abstract": "The success of deep neural networks in many real-world applications is leading to new challenges in building more efficient architectures. One effective way of making networks more efficient is neural network compression. We provide an overview of existing neural network compression methods that can be used to make neural networks more efficient by changing the architecture of the network. First, we introduce a new way to categorize all published compression methods, based on the amount of data and compute needed to make the methods work in practice. These are three 'levels of compression solutions'. Second, we provide a taxonomy of tensor factorization based and probabilistic compression methods. Finally, we perform an extensive evaluation of different compression techniques from the literature for models trained on ImageNet. We show that SVD and probabilistic compression or pruning methods are complementary and give the best results of all the considered methods. We also provide practical ways to combine them."
                    },
                    {
                        "title": "Low-Rank Quantization-Aware Training for LLMs",
                        "abstract": "Large language models (LLMs) are omnipresent, however their practical deployment is challenging due to their ever increasing computational and memory demands. Quantization is one of the most effective ways to make them more compute and memory efficient. Quantization-aware training (QAT) methods, generally produce the best quantized performance, however it comes at the cost of potentially long training time and excessive memory usage, making it impractical when applying for LLMs. Inspired by parameter-efficient fine-tuning (PEFT) and low-rank adaptation (LoRA) literature, we propose LR-QAT -- a lightweight and memory-efficient QAT algorithm for LLMs. LR-QAT employs several components to save memory without sacrificing predictive performance: (a) low-rank auxiliary weights that are aware of the quantization grid; (b) a downcasting operator using fixed-point or double-packed integers and (c) checkpointing. Unlike most related work, our method (i) is inference-efficient, leading to no additional overhead compared to traditional PTQ; (ii) can be seen as a general extended pretraining framework, meaning that the resulting model can still be utilized for any downstream task afterwards; (iii) can be applied across a wide range of quantization settings, such as different choices quantization granularity, activation quantization, and seamlessly combined with many PTQ techniques. We apply LR-QAT to LLaMA-1/2/3 and Mistral model families and validate its effectiveness on several downstream tasks. Our method outperforms common post-training quantization (PTQ) approaches and reaches the same model performance as full-model QAT at the fraction of its memory usage. Specifically, we can train a 7B LLM on a single consumer grade GPU with 24GB of memory. Our source code is available at https://github.com/qualcomm-ai-research/LR-QAT"
                    },
                    {
                        "title": "Quantization Robust Federated Learning for Efficient Inference on Heterogeneous Devices",
                        "abstract": "Federated Learning (FL) is a machine learning paradigm to distributively learn machine learning models from decentralized data that remains on-device. Despite the success of standard Federated optimization methods, such as Federated Averaging (FedAvg) in FL, the energy demands and hardware induced constraints for on-device learning have not been considered sufficiently in the literature. Specifically, an essential demand for on-device learning is to enable trained models to be quantized to various bit-widths based on the energy needs and heterogeneous hardware designs across the federation. In this work, we introduce multiple variants of federated averaging algorithm that train neural networks robust to quantization. Such networks can be quantized to various bit-widths with only limited reduction in full precision model accuracy. We perform extensive experiments on standard FL benchmarks to evaluate our proposed FedAvg variants for quantization robustness and provide a convergence analysis for our Quantization-Aware variants in FL. Our results demonstrate that integrating quantization robustness results in FL models that are significantly more robust to different bit-widths during quantized on-device inference."
                    },
                    {
                        "title": "Quantized Sparse Weight Decomposition for Neural Network Compression",
                        "abstract": "In this paper, we introduce a novel method of neural network weight compression. In our method, we store weight tensors as sparse, quantized matrix factors, whose product is computed on the fly during inference to generate the target model's weights. We use projected gradient descent methods to find quantized and sparse factorization of the weight tensors. We show that this approach can be seen as a unification of weight SVD, vector quantization, and sparse PCA. Combined with end-to-end fine-tuning our method exceeds or is on par with previous state-of-the-art methods in terms of the trade-off between accuracy and model size. Our method is applicable to both moderate compression regimes, unlike vector quantization, and extreme compression regimes."
                    },
                    {
                        "title": "ResQ: Residual Quantization for Video Perception",
                        "abstract": "This paper accelerates video perception, such as semantic segmentation and human pose estimation, by levering cross-frame redundancies. Unlike the existing approaches, which avoid redundant computations by warping the past features using optical-flow or by performing sparse convolutions on frame differences, we approach the problem from a new perspective: low-bit quantization. We observe that residuals, as the difference in network activations between two neighboring frames, exhibit properties that make them highly quantizable. Based on this observation, we propose a novel quantization scheme for video networks coined as Residual Quantization. ResQ extends the standard, frame-by-frame, quantization scheme by incorporating temporal dependencies that lead to better performance in terms of accuracy vs. bit-width. Furthermore, we extend our model to dynamically adjust the bit-width proportional to the amount of changes in the video. We demonstrate the superiority of our model, against the standard quantization and existing efficient video perception models, using various architectures on semantic segmentation and human pose estimation benchmarks."
                    },
                    {
                        "title": "Softmax Bias Correction for Quantized Generative Models",
                        "abstract": "Post-training quantization (PTQ) is the go-to compression technique for large generative models, such as stable diffusion or large language models. PTQ methods commonly keep the softmax activation in higher precision as it has been shown to be very sensitive to quantization noise. However, this can lead to a significant runtime and power overhead during inference on resource-constraint edge devices. In this work, we investigate the source of the softmax sensitivity to quantization and show that the quantization operation leads to a large bias in the softmax output, causing accuracy degradation. To overcome this issue, we propose an offline bias correction technique that improves the quantizability of softmax without additional compute during deployment, as it can be readily absorbed into the quantization parameters. We demonstrate the effectiveness of our method on stable diffusion v1.5 and 125M-size OPT language model, achieving significant accuracy improvement for 8-bit quantized softmax."
                    },
                    {
                        "title": "Up or Down? Adaptive Rounding for Post-Training Quantization",
                        "abstract": "When quantizing neural networks, assigning each floating-point weight to its nearest fixed-point value is the predominant approach. We find that, perhaps surprisingly, this is not the best we can do. In this paper, we propose AdaRound, a better weight-rounding mechanism for post-training quantization that adapts to the data and the task loss. AdaRound is fast, does not require fine-tuning of the network, and only uses a small amount of unlabelled data. We start by theoretically analyzing the rounding problem for a pre-trained neural network. By approximating the task loss with a Taylor series expansion, the rounding task is posed as a quadratic unconstrained binary optimization problem. We simplify this to a layer-wise local loss and propose to optimize this loss with a soft relaxation. AdaRound not only outperforms rounding-to-nearest by a significant margin but also establishes a new state-of-the-art for post-training quantization on several networks and tasks. Without fine-tuning, we can quantize the weights of Resnet18 and Resnet50 to 4 bits while staying within an accuracy loss of 1%."
                    },
                    {
                        "title": "LSQ+: Improving low-bit quantization through learnable offsets and better initialization",
                        "abstract": "Unlike ReLU, newer activation functions (like Swish, H-swish, Mish) that are frequently employed in popular efficient architectures can also result in negative activation values, with skewed positive and negative ranges. Typical learnable quantization schemes [PACT, LSQ] assume unsigned quantization for activations and quantize all negative activations to zero which leads to significant loss in performance. Naively using signed quantization to accommodate these negative values requires an extra sign bit which is expensive for low-bit (2-, 3-, 4-bit) quantization. To solve this problem, we propose LSQ+, a natural extension of LSQ, wherein we introduce a general asymmetric quantization scheme with trainable scale and offset parameters that can learn to accommodate the negative activations. Gradient-based learnable quantization schemes also commonly suffer from high instability or variance in the final training performance, hence requiring a great deal of hyper-parameter tuning to reach a satisfactory performance. LSQ+ alleviates this problem by using an MSE-based initialization scheme for the quantization parameters. We show that this initialization leads to significantly lower variance in final performance across multiple training runs. Overall, LSQ+ shows state-of-the-art results for EfficientNet and MixNet and also significantly outperforms LSQ for low-bit quantization of neural nets with Swish activations (e.g.: 1.8% gain with W4A4 quantization and upto 5.6% gain with W2A2 quantization of EfficientNet-B0 on ImageNet dataset). To the best of our knowledge, ours is the first work to quantize such architectures to extremely low bit-widths."
                    },
                    {
                        "title": "Implicit Neural Video Compression",
                        "abstract": "We propose a method to compress full-resolution video sequences with implicit neural representations. Each frame is represented as a neural network that maps coordinate positions to pixel values. We use a separate implicit network to modulate the coordinate inputs, which enables efficient motion compensation between frames. Together with a small residual network, this allows us to efficiently compress P-frames relative to the previous frame. We further lower the bitrate by storing the network weights with learned integer quantization. Our method, which we call implicit pixel flow (IPF), offers several simplifications over established neural video codecs: it does not require the receiver to have access to a pretrained neural network, does not use expensive interpolation-based warping operations, and does not require a separate training dataset. We demonstrate the feasibility of neural implicit compression on image and video data."
                    },
                    {
                        "title": "Cyclical Pruning for Sparse Neural Networks",
                        "abstract": "Current methods for pruning neural network weights iteratively apply magnitude-based pruning on the model weights and re-train the resulting model to recover lost accuracy. In this work, we show that such strategies do not allow for the recovery of erroneously pruned weights. To enable weight recovery, we propose a simple strategy called \\textit{cyclical pruning} which requires the pruning schedule to be periodic and allows for weights pruned erroneously in one cycle to recover in subsequent ones. Experimental results on both linear models and large-scale deep neural networks show that cyclical pruning outperforms existing pruning algorithms, especially at high sparsity ratios. Our approach is easy to tune and can be readily incorporated into existing pruning pipelines to boost performance."
                    },
                    {
                        "title": "Pruning vs Quantization: Which is Better?",
                        "abstract": "Neural network pruning and quantization techniques are almost as old as neural networks themselves. However, to date only ad-hoc comparisons between the two have been published. In this paper, we set out to answer the question on which is better: neural network quantization or pruning? By answering this question, we hope to inform design decisions made on neural network hardware going forward. We provide an extensive comparison between the two techniques for compressing deep neural networks. First, we give an analytical comparison of expected quantization and pruning error for general data distributions. Then, we provide lower bounds for the per-layer pruning and quantization error in trained networks, and compare these to empirical error after optimization. Finally, we provide an extensive experimental comparison for training 8 large-scale models on 3 tasks. Our results show that in most cases quantization outperforms pruning. Only in some scenarios with very high compression ratio, pruning might be beneficial from an accuracy standpoint."
                    }
                ]
            }
        }
    },
    "2406.01461": {
        "paper_data": {
            "title": "Hardness of Learning Neural Networks under the Manifold Hypothesis",
            "url": "http://arxiv.org/abs/2406.01461v1",
            "arxiv_id": "2406.01461",
            "authors": [
                "Bobak T. Kiani",
                "Jason Wang",
                "Melanie Weber"
            ],
            "abstract": "The manifold hypothesis presumes that high-dimensional data lies on or near a low-dimensional manifold. While the utility of encoding geometric structure has been demonstrated empirically, rigorous analysis of its impact on the learnability of neural networks is largely missing. Several recent results have established hardness results for learning feedforward and equivariant neural networks under i.i.d. Gaussian or uniform Boolean data distributions. In this paper, we investigate the hardness of learning under the manifold hypothesis. We ask which minimal assumptions on the curvature and regularity of the manifold, if any, render the learning problem efficiently learnable. We prove that learning is hard under input manifolds of bounded curvature by extending proofs of hardness in the SQ and cryptographic settings for Boolean data inputs to the geometric setting. On the other hand, we show that additional assumptions on the volume of the data manifold alleviate these fundamental limitations and guarantee learnability via a simple interpolation argument. Notable instances of this regime are manifolds which can be reliably reconstructed via manifold learning. Looking forward, we comment on and empirically explore intermediate regimes of manifolds, which have heterogeneous features commonly found in real world data.",
            "introduction": "   1 Introduction  High-dimensional data is often thought to have low-dimensional structure, which may stem, for instance, from symmetries in the underlying system. This observation has given rise to the manifold hypothesis, which presumes that high-dimensional data lies on or near a low-dimensional manifold. Empirical studies have confirmed this hypothesis across domains\u00a0[20, 77, 37, 65]. A plethora of algorithms for geometric data analysis\u00a0[44, 8, 80, 21] and, more recently, machine learning\u00a0[30, 15, 23] seek to leverage such structure. While these methods have shown empirical promise, few formal results on the benefits of encoding geometric structure have been established. Here, we study the impact of the manifold hypothesis on the computational complexity of learning algorithms for neural networks. Specifically, we investigate, under which geometric assumptions feedforward neural networks are efficiently learnable.   The computational hardness of learning shallow, fully-connected neural networks under i.i.d. Gaussian data distributions has been established by [36, 26, 48, 32], who show that the complexity of learning even single hidden layer neural networks grows at least exponentially with the input size. Some of this analysis utilizes the correlational statistical query (CSQ) framework, of which learnability via gradient decent is a notable instance. Recently, [57] have shown similar hardness results for equivariant neural networks, a class of geometric architectures that explicitly encode symmetries. Both lines of work indicate that additional assumptions on the neural network architecture or the data geometry are needed to establish learnability. Here, we focus on the latter.   A separate body of literature investigates nonlinear, low-dimensional structure in data. Approaches for estimating intrinsic dimension\u00a0[44, 63, 87] and curvature\u00a0[1, 86, 46] from data seek to characterize the geometry of low-dimensional manifolds. Manifold Learning aims to identify low-dimensional structure by reconstructing low-dimensional manifolds from data. Methods such as Multi-Dimensional Scaling\u00a0[62], Isomap\u00a0[89] or Diffusion Maps\u00a0[31] have shown empirical success in learning low-dimensional data manifolds. For some of these approaches, formal guarantees on their reliability with respect to the geometric characteristics and sample size of the data are known\u00a0[9]. More generally, the manifold learning problem and its complexity have been formally studied in\u00a0[43, 42, 3]. However, the impact of data geometry on the complexity of learning neural networks in downstream tasks remains open.   In this paper, we ask: Which assumptions on the data geometry guarantee learnability of neural networks? We show that the manifold hypothesis on its own does not guarantee learnability. In particular, we give hardness results for a class of low-dimensional manifolds in the statistical query (SQ) model or under cryptographic hardness assumptions. Our work follows an established line of proof techniques in the SQ literature\u00a0[32, 48, 26], extending hardness results for neural network training in the Boolean and Gaussian input models to more general geometries. We further show that additional assumptions on the volume and curvature of the data manifold alleviate the fundamental limitations and guarantee learnability through a rather simple interpolation argument. In particular, manifolds which can be reliably reconstructed via manifold learning are in this regime. We further discuss geometric regimes in which our results do not directly apply, and in which provable learnability remains an open question. We illustrate our learnability results through computational experiments on",
            "references": [
                {
                    "title": "On the hardness of learning under symmetries",
                    "abstract": "We study the problem of learning equivariant neural networks via gradient descent. The incorporation of known symmetries (\"equivariance\") into neural nets has empirically improved the performance of learning pipelines, in domains ranging from biology to computer vision. However, a rich yet separate line of learning theoretic research has demonstrated that actually learning shallow, fully-connected (i.e. non-symmetric) networks has exponential complexity in the correlational statistical query (CSQ) model, a framework encompassing gradient descent. In this work, we ask: are known problem symmetries sufficient to alleviate the fundamental hardness of learning neural nets with gradient descent? We answer this question in the negative. In particular, we give lower bounds for shallow graph neural networks, convolutional networks, invariant polynomials, and frame-averaged networks for permutation subgroups, which all scale either superpolynomially or exponentially in the relevant input dimension. Therefore, in spite of the significant inductive bias imparted via symmetry, actually learning the complete classes of functions represented by equivariant neural networks via gradient descent remains hard."
                },
                {
                    "title": "Strong Convexity of Sets in Riemannian Manifolds",
                    "abstract": "Convex curvature properties are important in designing and analyzing convex optimization algorithms in the Hilbertian or Riemannian settings. In the case of the Hilbertian setting, strongly convex sets are well studied. Herein, we propose various definitions of strong convexity for uniquely geodesic sets in a Riemannian manifold. We study their relationship, propose tools to determine the geodesic strongly convex nature of sets, and analyze the convergence of optimization algorithms over those sets. In particular, we demonstrate that the Riemannian Frank-Wolfe algorithm enjoys a global linear convergence rate when the Riemannian scaling inequalities hold."
                },
                {
                    "title": "A faster and simpler algorithm for learning shallow networks",
                    "abstract": "We revisit the well-studied problem of learning a linear combination of $k$ ReLU activations given labeled examples drawn from the standard $d$-dimensional Gaussian measure. Chen et al. [CDG+23] recently gave the first algorithm for this problem to run in $\\text{poly}(d,1/\\varepsilon)$ time when $k = O(1)$, where $\\varepsilon$ is the target error. More precisely, their algorithm runs in time $(d/\\varepsilon)^{\\mathrm{quasipoly}(k)}$ and learns over multiple stages. Here we show that a much simpler one-stage version of their algorithm suffices, and moreover its runtime is only $(d/\\varepsilon)^{O(k^2)}$."
                },
                {
                    "title": "Continuum Limits of Ollivier's Ricci Curvature on data clouds: pointwise consistency and global lower bounds",
                    "abstract": "Let $M$ denote a low-dimensional manifold embedded in Euclidean space and let ${X}= \\{ x_1, \\dots, x_n \\}$ be a collection of points uniformly sampled from it. We study the relationship between the curvature of a random geometric graph built from ${X}$ and the curvature of the manifold $M$ via continuum limits of Ollivier's discrete Ricci curvature. We prove pointwise, non-asymptotic consistency results and also show that if $M$ has Ricci curvature bounded from below by a positive constant, then the random geometric graph will inherit this global structural property with high probability. We discuss applications of the global discrete curvature bounds to contraction properties of heat kernels on graphs, as well as implications for manifold learning from data clouds. In particular, we show that our consistency results allow for estimating the intrinsic curvature of a manifold by first estimating concrete extrinsic quantities."
                },
                {
                    "title": "Learning Narrow One-Hidden-Layer ReLU Networks",
                    "abstract": "We consider the well-studied problem of learning a linear combination of $k$ ReLU activations with respect to a Gaussian distribution on inputs in $d$ dimensions. We give the first polynomial-time algorithm that succeeds whenever $k$ is a constant. All prior polynomial-time learners require additional assumptions on the network, such as positive combining coefficients or the matrix of hidden weight vectors being well-conditioned. Our approach is based on analyzing random contractions of higher-order moment tensors. We use a multi-scale analysis to argue that sufficiently close neurons can be collapsed together, sidestepping the conditioning issues present in prior work. This allows us to design an iterative procedure to discover individual neurons."
                },
                {
                    "title": "The Exact Sample Complexity Gain from Invariances for Kernel Regression on Manifolds",
                    "abstract": "In practice, encoding invariances into models helps sample complexity. In this work, we tighten and generalize theoretical results on how invariances improve sample complexity. In particular, we provide minimax optimal rates for kernel ridge regression on any manifold, with a target function that is invariant to an arbitrary group action on the manifold. Our results hold for (almost) any group action, even groups of positive dimension. For a finite group, the gain increases the\"effective\"number of samples by the group size. For groups of positive dimension, the gain is observed by a reduction in the manifold's dimension, in addition to a factor proportional to the volume of the quotient space. Our proof takes the viewpoint of differential geometry, in contrast to the more common strategy of using invariant polynomials. Hence, this new geometric viewpoint on learning with invariances may be of independent interest."
                },
                {
                    "title": "SGD learning on neural networks: leap complexity and saddle-to-saddle dynamics",
                    "abstract": "We investigate the time complexity of SGD learning on fully-connected neural networks with isotropic data. We put forward a complexity measure -- the leap -- which measures how\"hierarchical\"target functions are. For $d$-dimensional uniform Boolean or isotropic Gaussian data, our main conjecture states that the time complexity to learn a function $f$ with low-dimensional support is $\\tilde\\Theta (d^{\\max(\\mathrm{Leap}(f),2)})$. We prove a version of this conjecture for a class of functions on Gaussian isotropic data and 2-layer neural networks, under additional technical assumptions on how SGD is run. We show that the training sequentially learns the function support with a saddle-to-saddle dynamic. Our result departs from [Abbe et al. 2022] by going beyond leap 1 (merged-staircase functions), and by going beyond the mean-field and gradient flow approximations that prohibit the full complexity control obtained here. Finally, we note that this gives an SGD complexity for the full training trajectory that matches that of Correlational Statistical Query (CSQ) lower-bounds."
                },
                {
                    "title": "Score Approximation, Estimation and Distribution Recovery of Diffusion Models on Low-Dimensional Data",
                    "abstract": "Diffusion models achieve state-of-the-art performance in various generation tasks. However, their theoretical foundations fall far behind. This paper studies score approximation, estimation, and distribution recovery of diffusion models, when data are supported on an unknown low-dimensional linear subspace. Our result provides sample complexity bounds for distribution estimation using diffusion models. We show that with a properly chosen neural network architecture, the score function can be both accurately approximated and efficiently estimated. Furthermore, the generated distribution based on the estimated score function captures the data geometric structures and converges to a close vicinity of the data distribution. The convergence rate depends on the subspace dimension, indicating that diffusion models can circumvent the curse of data ambient dimensionality."
                },
                {
                    "title": "Your diffusion model secretly knows the dimension of the data manifold",
                    "abstract": "In this work, we propose a novel framework for estimating the dimension of the data manifold using a trained diffusion model. A diffusion model approximates the score function i.e. the gradient of the log density of a noise-corrupted version of the target distribution for varying levels of corruption. We prove that, if the data concentrates around a manifold embedded in the high-dimensional ambient space, then as the level of corruption decreases, the score function points towards the manifold, as this direction becomes the direction of maximal likelihood increase. Therefore, for small levels of corruption, the diffusion model provides us with access to an approximation of the normal bundle of the data manifold. This allows us to estimate the dimension of the tangent space, thus, the intrinsic dimension of the data manifold. To the best of our knowledge, our method is the first estimator of the data manifold dimension based on diffusion models and it outperforms well established statistical estimators in controlled experiments on both Euclidean and image data."
                },
                {
                    "title": "A Solvable Model of Neural Scaling Laws",
                    "abstract": "Large language models with a huge number of parameters, when trained on near internet-sized number of tokens, have been empirically shown to obey neural scaling laws: specifically, their performance behaves predictably as a power law in either parameters or dataset size until bottlenecked by the other resource. To understand this better, we first identify the necessary properties allowing such scaling laws to arise and then propose a statistical model \u2013 a joint generative data model and random feature model \u2013 that captures this neural scaling phenomenology. By solving this model in the dual limit of large training set size and large number of parameters, we gain insight into (i) the statistical structure of datasets and tasks that lead to scaling laws, (ii) the way nonlinear feature maps, such as those provided by neural networks, enable scaling laws when trained on these datasets, (iii) the optimality of the equiparameterization scaling of training sets and parameters, and (iv) whether such scaling laws can break down and how they behave when they do. Key findings are the manner in which the power laws that occur in the statistics of natural datasets are extended by nonlinear random feature maps and then translated into power-law scalings of the test loss and how the finite extent of the data\u2019s spectral power law causes the model\u2019s performance to plateau. ? Equal contribution. ar X iv :2 21 0. 16 85 9v 1 [ cs .L G ] 3 0 O ct 2 02 2"
                },
                {
                    "title": "Sampling is as easy as learning the score: theory for diffusion models with minimal data assumptions",
                    "abstract": "We provide theoretical convergence guarantees for score-based generative models (SGMs) such as denoising diffusion probabilistic models (DDPMs), which constitute the backbone of large-scale real-world generative models such as DALL$\\cdot$E 2. Our main result is that, assuming accurate score estimates, such SGMs can efficiently sample from essentially any realistic data distribution. In contrast to prior works, our results (1) hold for an $L^2$-accurate score estimate (rather than $L^\\infty$-accurate); (2) do not require restrictive functional inequality conditions that preclude substantial non-log-concavity; (3) scale polynomially in all relevant problem parameters; and (4) match state-of-the-art complexity guarantees for discretization of the Langevin diffusion, provided that the score error is sufficiently small. We view this as strong theoretical justification for the empirical success of SGMs. We also examine SGMs based on the critically damped Langevin diffusion (CLD). Contrary to conventional wisdom, we provide evidence that the use of the CLD does not reduce the complexity of SGMs."
                },
                {
                    "title": "Convergence of denoising diffusion models under the manifold hypothesis",
                    "abstract": "Denoising diffusion models are a recent class of generative models exhibiting state-of-the-art performance in image and audio synthesis. Such models approximate the time-reversal of a forward noising process from a target distribution to a reference density, which is usually Gaussian. Despite their strong empirical results, the theoretical analysis of such models remains limited. In particular, all current approaches crucially assume that the target density admits a density w.r.t. the Lebesgue measure. This does not cover settings where the target distribution is supported on a lower-dimensional manifold or is given by some empirical distribution. In this paper, we bridge this gap by providing the first convergence results for diffusion models in this more general setting. In particular, we provide quantitative bounds on the Wasserstein distance of order one between the target data distribution and the generative distribution of the diffusion model."
                },
                {
                    "title": "Verifying the Union of Manifolds Hypothesis for Image Data",
                    "abstract": "Deep learning has had tremendous success at learning low-dimensional representations of high-dimensional data. This success would be impossible if there was no hidden low-dimensional structure in data of interest; this existence is posited by the manifold hypothesis, which states that the data lies on an unknown manifold of low intrinsic dimension. In this paper, we argue that this hypothesis does not properly capture the low-dimensional structure typically present in image data. Assuming that data lies on a single manifold implies intrinsic dimension is identical across the entire data space, and does not allow for subregions of this space to have a different number of factors of variation. To address this deficiency, we consider the union of manifolds hypothesis, which states that data lies on a disjoint union of manifolds of varying intrinsic dimensions. We empirically verify this hypothesis on commonly-used image datasets, finding that indeed, observed data lies on a disconnected set and that intrinsic dimension is not constant. We also provide insights into the implications of the union of manifolds hypothesis in deep learning, both supervised and unsupervised, showing that designing models with an inductive bias for this structure improves performance across classification and generative modelling tasks. Our code is available at https://github.com/layer6ai-labs/UoMH."
                },
                {
                    "title": "Score-Based Generative Models Detect Manifolds",
                    "abstract": "Score-based generative models (SGMs) need to approximate the scores $\\nabla \\log p_t$ of the intermediate distributions as well as the final distribution $p_T$ of the forward process. The theoretical underpinnings of the effects of these approximations are still lacking. We find precise conditions under which SGMs are able to produce samples from an underlying (low-dimensional) data manifold $\\mathcal{M}$. This assures us that SGMs are able to generate the\"right kind of samples\". For example, taking $\\mathcal{M}$ to be the subset of images of faces, we find conditions under which the SGM robustly produces an image of a face, even though the relative frequencies of these images might not accurately represent the true data generating distribution. Moreover, this analysis is a first step towards understanding the generalization properties of SGMs: Taking $\\mathcal{M}$ to be the set of all training samples, our results provide a precise description of when the SGM memorizes its training data."
                },
                {
                    "title": "Hardness of Noise-Free Learning for Two-Hidden-Layer Neural Networks",
                    "abstract": "We give superpolynomial statistical query (SQ) lower bounds for learning two-hidden-layer ReLU networks with respect to Gaussian inputs in the standard (noise-free) model. No general SQ lower bounds were known for learning ReLU networks of any depth in this setting: previous SQ lower bounds held only for adversarial noise models (agnostic learning) or restricted models such as correlational SQ. Prior work hinted at the impossibility of our result: Vempala and Wilmes showed that general SQ lower bounds cannot apply to any real-valued family of functions that satisfies a simple non-degeneracy condition. To circumvent their result, we refine a lifting procedure due to Daniely and Vardi that reduces Boolean PAC learning problems to Gaussian ones. We show how to extend their technique to other learning models and, in many well-studied cases, obtain a more efficient reduction. As such, we also prove new cryptographic hardness results for PAC learning two-hidden-layer ReLU networks, as well as new lower bounds for learning constant-depth ReLU networks from label queries."
                },
                {
                    "title": "Besov Function Approximation and Binary Classification on Low-Dimensional Manifolds Using Convolutional Residual Networks",
                    "abstract": "Most of existing statistical theories on deep neural networks have sample complexities cursed by the data dimension and therefore cannot well explain the empirical success of deep learning on high-dimensional data. To bridge this gap, we propose to exploit low-dimensional geometric structures of the real world data sets. We establish theoretical guarantees of convolutional residual networks (ConvResNet) in terms of function approximation and statistical estimation for binary classification. Specifically, given the data lying on a $d$-dimensional manifold isometrically embedded in $\\mathbb{R}^D$, we prove that if the network architecture is properly chosen, ConvResNets can (1) approximate Besov functions on manifolds with arbitrary accuracy, and (2) learn a classifier by minimizing the empirical logistic risk, which gives an excess risk in the order of $n^{-\\frac{s}{2s+2(s\\vee d)}}$, where $s$ is a smoothness parameter. This implies that the sample complexity depends on the intrinsic dimension $d$, instead of the data dimension $D$. Our results demonstrate that ConvResNets are adaptive to low-dimensional structures of data sets."
                },
                {
                    "title": "A Universal Law of Robustness via Isoperimetry",
                    "abstract": "Classically, data interpolation with a parametrized model class is possible as long as the number of parameters is larger than the number of equations to be satisfied. A puzzling phenomenon in deep learning is that models are trained with many more parameters than what this classical theory would suggest. We propose a partial theoretical explanation for this phenomenon. We prove that for a broad class of data distributions and model classes, overparametrization is necessary if one wants to interpolate the data smoothly. Namely we show that smooth interpolation requires d times more parameters than mere interpolation, where d is the ambient data dimension. We prove this universal law of robustness for any smoothly parametrized function class with polynomial size weights, and any covariate distribution verifying isoperimetry (or a mixture thereof). In the case of two-layer neural networks and Gaussian covariates, this law was conjectured in prior work by Bubeck, Li, and Nagaraj. We also give an interpretation of our result as an improved generalization bound for model classes consisting of smooth functions."
                },
                {
                    "title": "The Intrinsic Dimension of Images and Its Impact on Learning",
                    "abstract": "It is widely believed that natural image data exhibits low-dimensional structure despite the high dimensionality of conventional pixel representations. This idea underlies a common intuition for the remarkable success of deep learning in computer vision. In this work, we apply dimension estimation tools to popular datasets and investigate the role of low-dimensional structure in deep learning. We find that common natural image datasets indeed have very low intrinsic dimension relative to the high number of pixels in the images. Additionally, we find that low dimensional datasets are easier for neural networks to learn, and models solving these tasks generalize better from training to test data. Along the way, we develop a technique for validating our dimension estimation tools on synthetic data generated by GANs allowing us to actively manipulate the intrinsic dimension by controlling the image generation process. Code for our experiments may be found here https://github.com/ppope/dimensions."
                },
                {
                    "title": "Improved Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) are a class of generative models which have recently been shown to produce excellent samples. We show that with a few simple modifications, DDPMs can also achieve competitive log-likelihoods while maintaining high sample quality. Additionally, we find that learning variances of the reverse diffusion process allows sampling with an order of magnitude fewer forward passes with a negligible difference in sample quality, which is important for the practical deployment of these models. We additionally use precision and recall to compare how well DDPMs and GANs cover the target distribution. Finally, we show that the sample quality and likelihood of these models scale smoothly with model capacity and training compute, making them easily scalable. We release our code at https://github.com/openai/improved-diffusion"
                },
                {
                    "title": "Explaining neural scaling laws",
                    "abstract": "Significance The population loss of trained deep neural networks has been empirically observed to improve as a power law in a variety of large models and datasets. We investigate the origins behind such \u201cscaling laws\u201d and provide a taxonomy for different scaling regimes. Our findings are based on derivations in linear random feature models\u2014which, in addition to being a simple fruitful model, also describe the wide network limit of deep neural networks. We further formulate and verify aspects of scaling based on smoothness in interpolating a data manifold. We support our theory with empirical results in realistic settings. Our work provides insights into scaling laws and bridges the large gap between theory and experiment in modern deep learning."
                },
                {
                    "title": "From Local Pseudorandom Generators to Hardness of Learning",
                    "abstract": "We prove hardness-of-learning results under a well-studied assumption on the existence of local pseudorandom generators. As we show, this assumption allows us to surpass the current state of the art, and prove hardness of various basic problems, with no hardness results to date. Our results include: hardness of learning shallow ReLU neural networks under the Gaussian distribution and other distributions; hardness of learning intersections of $\\omega(1)$ halfspaces, DNF formulas with $\\omega(1)$ terms, and ReLU networks with $\\omega(1)$ hidden neurons; hardness of weakly learning deterministic finite automata under the uniform distribution; hardness of weakly learning depth-$3$ Boolean circuits under the uniform distribution, as well as distribution-specific hardness results for learning DNF formulas and intersections of halfspaces. We also establish lower bounds on the complexity of learning intersections of a constant number of halfspaces, and ReLU networks with a constant number of hidden neurons. Moreover, our results imply the hardness of virtually all improper PAC-learning problems (both distribution-free and distribution-specific) that were previously shown hard under other assumptions."
                },
                {
                    "title": "Hyperbolic Deep Neural Networks: A Survey",
                    "abstract": "Recently, hyperbolic deep neural networks (HDNNs) have been gaining momentum as the deep representations in the hyperbolic space provide high fidelity embeddings with few dimensions, especially for data possessing hierarchical structure. Such a hyperbolic neural architecture is quickly extended to different scientific fields, including natural language processing, single-cell RNA-sequence analysis, graph embedding, financial analysis, and computer vision. The promising results demonstrate its superior capability, significant compactness of the model, and a substantially better physical interpretability than its counterpart in the euclidean space. To stimulate future research, this paper presents a comprehensive review of the literature around the neural components in the construction of HDNN, as well as the generalization of the leading deep approaches to the hyperbolic space. It also presents current applications of various tasks, together with insightful observations and identifying open questions and promising future directions."
                },
                {
                    "title": "Computing the Riemannian curvature of image patch and single-cell RNA sequencing data manifolds using extrinsic differential geometry",
                    "abstract": "Significance High-dimensional datasets are becoming increasingly prevalent in many scientific fields. A universal theme connecting these high-dimensional datasets is the ansatz that data points are constrained to lie on nonlinear low-dimensional manifolds, whose structure is dictated by the natural laws governing the data. While tools have been developed for estimating global properties of these data manifolds, estimating the Riemannian curvature, a local property, has not been considered. Computing curvature of data manifolds offers both detailed criteria with which to evaluate models of these complex data (e.g., a Klein bottle model of image patches) and a way to explore detailed geometric features that cannot simply be visualized by the naked eye (e.g., in single-cell RNA-sequencing data). Most high-dimensional datasets are thought to be inherently low-dimensional\u2014that is, data points are constrained to lie on a low-dimensional manifold embedded in a high-dimensional ambient space. Here, we study the viability of two approaches from differential geometry to estimate the Riemannian curvature of these low-dimensional manifolds. The intrinsic approach relates curvature to the Laplace\u2013Beltrami operator using the heat-trace expansion and is agnostic to how a manifold is embedded in a high-dimensional space. The extrinsic approach relates the ambient coordinates of a manifold\u2019s embedding to its curvature using the Second Fundamental Form and the Gauss\u2013Codazzi equation. We found that the intrinsic approach fails to accurately estimate the curvature of even a two-dimensional constant-curvature manifold, whereas the extrinsic approach was able to handle more complex toy models, even when confounded by practical constraints like small sample sizes and measurement noise. To test the applicability of the extrinsic approach to real-world data, we computed the curvature of a well-studied manifold of image patches and recapitulated its topological classification as a Klein bottle. Lastly, we applied the extrinsic approach to study single-cell transcriptomic sequencing (scRNAseq) datasets of blood, gastrulation, and brain cells to quantify the Riemannian curvature of scRNAseq manifolds."
                },
                {
                    "title": "Statistical query complexity of manifold estimation",
                    "abstract": "This paper studies the statistical query (SQ) complexity of estimating d-dimensional submanifolds in \u211dn. We propose a purely geometric algorithm called Manifold Propagation, that reduces the problem to three natural geometric routines: projection, tangent space estimation, and point detection. We then provide constructions of these geometric routines in the SQ framework. Given an adversarial STAT(\u03c4) oracle and a target Hausdorff distance precision \u03b5 = \u03a9(\u03c42/(d+1)), the resulting SQ manifold reconstruction algorithm has query complexity O(n polylog(n) \u03b5\u2212d/2), which is proved to be nearly optimal. In the process, we establish low-rank matrix completion results for SQ\u2019s and lower bounds for randomized SQ estimators in general metric spaces."
                },
                {
                    "title": "Deep Networks and the Multiple Manifold Problem",
                    "abstract": "We study the multiple manifold problem, a binary classification task modeled on applications in machine vision, in which a deep fully-connected neural network is trained to separate two low-dimensional submanifolds of the unit sphere. We provide an analysis of the one-dimensional case, proving for a simple manifold configuration that when the network depth $L$ is large relative to certain geometric and statistical properties of the data, the network width $n$ grows as a sufficiently large polynomial in $L$, and the number of i.i.d. samples from the manifolds is polynomial in $L$, randomly-initialized gradient descent rapidly learns to classify the two manifolds perfectly with high probability. Our analysis demonstrates concrete benefits of depth and width in the context of a practically-motivated model problem: the depth acts as a fitting resource, with larger depths corresponding to smoother networks that can more readily separate the class manifolds, and the width acts as a statistical resource, enabling concentration of the randomly-initialized network and its gradients. The argument centers around the neural tangent kernel and its role in the nonasymptotic analysis of training overparameterized neural networks; to this literature, we contribute essentially optimal rates of concentration for the neural tangent kernel of deep fully-connected networks, requiring width $n \\gtrsim L\\,\\mathrm{poly}(d_0)$ to achieve uniform concentration of the initial kernel over a $d_0$-dimensional submanifold of the unit sphere $\\mathbb{S}^{n_0-1}$, and a nonasymptotic framework for establishing generalization of networks trained in the NTK regime with structured data. The proof makes heavy use of martingale concentration to optimally treat statistical dependencies across layers of the initial random network. This approach should be of use in establishing similar results for other network architectures."
                },
                {
                    "title": "Superpolynomial Lower Bounds for Learning One-Layer Neural Networks using Gradient Descent",
                    "abstract": "We prove the first superpolynomial lower bounds for learning one-layer neural networks with respect to the Gaussian distribution using gradient descent. We show that any classifier trained using gradient descent with respect to square-loss will fail to achieve small test error in polynomial time given access to samples labeled by a one-layer neural network. For classification, we give a stronger result, namely that any statistical query (SQ) algorithm (including gradient descent) will fail to achieve small test error in polynomial time. Prior work held only for gradient descent run with small batch sizes, required sharp activations, and applied to specific classes of queries. Our lower bounds hold for broad classes of activations including ReLU and sigmoid. The core of our result relies on a novel construction of a simple family of neural networks that are exactly orthogonal with respect to all spherically symmetric distributions."
                },
                {
                    "title": "Manifold structure in graph embeddings",
                    "abstract": "Statistical analysis of a graph often starts with embedding, the process of representing its nodes as points in space. How to choose the embedding dimension is a nuanced decision in practice, but in theory a notion of true dimension is often available. In spectral embedding, this dimension may be very high. However, this paper shows that existing random graph models, including graphon and other latent position models, predict the data should live near a much lower-dimensional set. One may therefore circumvent the curse of dimensionality by employing methods which exploit hidden manifold structure."
                },
                {
                    "title": "Algorithms and SQ Lower Bounds for PAC Learning One-Hidden-Layer ReLU Networks",
                    "abstract": "We study the problem of PAC learning one-hidden-layer ReLU networks with $k$ hidden units on $\\mathbb{R}^d$ under Gaussian marginals in the presence of additive label noise. For the case of positive coefficients, we give the first polynomial-time algorithm for this learning problem for $k$ up to $\\tilde{O}(\\sqrt{\\log d})$. Previously, no polynomial time algorithm was known, even for $k=3$. This answers an open question posed by~\\cite{Kliv17}. Importantly, our algorithm does not require any assumptions about the rank of the weight matrix and its complexity is independent of its condition number. On the negative side, for the more general task of PAC learning one-hidden-layer ReLU networks with arbitrary real coefficients, we prove a Statistical Query lower bound of $d^{\\Omega(k)}$. Thus, we provide a separation between the two classes in terms of efficient learnability. Our upper and lower bounds are general, extending to broader families of activation functions."
                },
                {
                    "title": "A Neural Scaling Law from the Dimension of the Data Manifold",
                    "abstract": "When data is plentiful, the loss achieved by well-trained neural networks scales as a power-law $L \\propto N^{-\\alpha}$ in the number of network parameters $N$. This empirical scaling law holds for a wide variety of data modalities, and may persist over many orders of magnitude. The scaling law can be explained if neural models are effectively just performing regression on a data manifold of intrinsic dimension $d$. This simple theory predicts that the scaling exponents $\\alpha \\approx 4/d$ for cross-entropy and mean-squared error losses. We confirm the theory by independently measuring the intrinsic dimension and the scaling exponents in a teacher/student framework, where we can study a variety of $d$ and $\\alpha$ by dialing the properties of random teacher networks. We also test the theory with CNN image classifiers on several datasets and with GPT-type language models."
                },
                {
                    "title": "Robust Large-Margin Learning in Hyperbolic Space",
                    "abstract": "Recently, there has been a surge of interest in representation learning in hyperbolic spaces, driven by their ability to represent hierarchical data with significantly fewer dimensions than standard Euclidean spaces. However, the viability and benefits of hyperbolic spaces for downstream machine learning tasks have received less attention. In this paper, we present, to our knowledge, the first theoretical guarantees for learning a classifier in hyperbolic rather than Euclidean space. Specifically, we consider the problem of learning a large-margin classifier for data possessing a hierarchical structure. Our first contribution is a hyperbolic perceptron algorithm, which provably converges to a separating hyperplane. We then provide an algorithm to efficiently learn a large-margin hyperplane, relying on the careful injection of adversarial examples. Finally, we prove that for hierarchical data that embeds well into hyperbolic space, the low embedding dimension ensures superior guarantees when learning the classifier directly in hyperbolic space."
                },
                {
                    "title": "Statistical Queries and Statistical Algorithms: Foundations and Applications",
                    "abstract": "We give a survey of the foundations of statistical queries and their many applications to other areas. We introduce the model, give the main definitions, and we explore the fundamental theory statistical queries and how how it connects to various notions of learnability. We also give a detailed summary of some of the applications of statistical queries to other areas, including to optimization, to evolvability, and to differential privacy."
                },
                {
                    "title": "Minimax adaptive estimation in manifold inference",
                    "abstract": "We focus on the problem of manifold estimation: given a set of observations sampled close to some unknown submanifold M , one wants to recover information about the geometry of M. Minimax estimators which have been proposed so far all depend crucially on the a priori knowledge of some parameters quantifying the regularity of M (such as its reach), whereas those quantities will be unknown in practice. Our contribution to the matter is twofold: first, we introduce a one-parameter family of manifold estimators (M t) t\u22650 , and show that for some choice of t (depending on the regularity parameters), the corresponding estimator is minimax on the class of models of C 2 manifolds introduced in [Genovese et al., Manifold estimation and singular deconvolution under Hausdorff loss]. Second, we propose a completely data-driven selection procedure for the parameter t, leading to a minimax adaptive manifold estimator on this class of models. This selection procedure actually allows to recover the sample rate of the set of observations, and can therefore be used as an hyperparameter in other settings, such as tangent space estimation."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Deep Learning for Classical Japanese Literature",
                    "abstract": "Much of machine learning research focuses on producing models which perform well on benchmark tasks, in turn improving our understanding of the challenges associated with those tasks. From the perspective of ML researchers, the content of the task itself is largely irrelevant, and thus there have increasingly been calls for benchmark tasks to more heavily focus on problems which are of social or cultural relevance. In this work, we introduce Kuzushiji-MNIST, a dataset which focuses on Kuzushiji (cursive Japanese), as well as two larger, more challenging datasets, Kuzushiji-49 and Kuzushiji-Kanji. Through these datasets, we wish to engage the machine learning community into the world of classical Japanese literature. Dataset available at this https URL"
                },
                {
                    "title": "Learning Two Layer Rectified Neural Networks in Polynomial Time",
                    "abstract": "Consider the following fundamental learning problem: given input examples $x \\in \\mathbb{R}^d$ and their vector-valued labels, as defined by an underlying generative neural network, recover the weight matrices of this network. We consider two-layer networks, mapping $\\mathbb{R}^d$ to $\\mathbb{R}^m$, with $k$ non-linear activation units $f(\\cdot)$, where $f(x) = \\max \\{x , 0\\}$ is the ReLU. Such a network is specified by two weight matrices, $\\mathbf{U}^* \\in \\mathbb{R}^{m \\times k}, \\mathbf{V}^* \\in \\mathbb{R}^{k \\times d}$, such that the label of an example $x \\in \\mathbb{R}^{d}$ is given by $\\mathbf{U}^* f(\\mathbf{V}^* x)$, where $f(\\cdot)$ is applied coordinate-wise. Given $n$ samples as a matrix $\\mathbf{X} \\in \\mathbb{R}^{d \\times n}$ and the (possibly noisy) labels $\\mathbf{U}^* f(\\mathbf{V}^* \\mathbf{X}) + \\mathbf{E}$ of the network on these samples, where $\\mathbf{E}$ is a noise matrix, our goal is to recover the weight matrices $\\mathbf{U}^*$ and $\\mathbf{V}^*$. \nIn this work, we develop algorithms and hardness results under varying assumptions on the input and noise. Although the problem is NP-hard even for $k=2$, by assuming Gaussian marginals over the input $\\mathbf{X}$ we are able to develop polynomial time algorithms for the approximate recovery of $\\mathbf{U}^*$ and $\\mathbf{V}^*$. Perhaps surprisingly, in the noiseless case our algorithms recover $\\mathbf{U}^*,\\mathbf{V}^*$ exactly, i.e., with no error. To the best of the our knowledge, this is the first algorithm to accomplish exact recovery. For the noisy case, we give the first polynomial time algorithm that approximately recovers the weights in the presence of mean-zero noise $\\mathbf{E}$. Our algorithms generalize to a larger class of rectified activation functions, $f(x) = 0$ when $x\\leq 0$, and $f(x) > 0$ otherwise."
                },
                {
                    "title": "Fitting a Putative Manifold to Noisy Data",
                    "abstract": "In the present work, we give a solution to the following question from manifold learning. Suppose data belonging to a high dimensional Euclidean space is drawn independently, identically distributed from a measure supported on a low dimensional twice differentiable embedded manifoldM, and corrupted by a small amount of gaussian noise. How can we produce a manifoldMo whose Hausdorff distance to M is small and whose reach is not much smaller than the reach of M?"
                },
                {
                    "title": "Lipschitz regularity of deep neural networks: analysis and efficient estimation",
                    "abstract": "Deep neural networks are notorious for being sensitive to small well-chosen perturbations, and estimating the regularity of such architectures is of utmost importance for safe and robust practical applications. In this paper, we investigate one of the key characteristics to assess the regularity of such methods: the Lipschitz constant of deep learning architectures. First, we show that, even for two layer neural networks, the exact computation of this quantity is NP-hard and state-of-art methods may significantly overestimate it. Then, we both extend and improve previous estimation methods by providing AutoLip, the first generic algorithm for upper bounding the Lipschitz constant of any automatically differentiable function. We provide a power method algorithm working with automatic differentiation, allowing efficient computations even on large convolutions. Second, for sequential neural networks, we propose an improved algorithm named SeqLip that takes advantage of the linear computation graph to split the computation per pair of consecutive layers. Third we propose heuristics on SeqLip in order to tackle very large networks. Our experiments show that SeqLip can significantly improve on the existing upper bounds. Finally, we provide an implementation of AutoLip in the PyTorch environment that may be used to better estimate the robustness of a given neural network to small perturbations or regularize it using more precise Lipschitz estimations."
                },
                {
                    "title": "Hyperbolic Neural Networks",
                    "abstract": "Hyperbolic spaces have recently gained momentum in the context of machine learning due to their high capacity and tree-likeliness properties. However, the representational power of hyperbolic geometry is not yet on par with Euclidean geometry, mostly because of the absence of corresponding hyperbolic neural network layers. This makes it hard to use hyperbolic embeddings in downstream tasks. Here, we bridge this gap in a principled manner by combining the formalism of Mobius gyrovector spaces with the Riemannian geometry of the Poincare model of hyperbolic spaces. As a result, we derive hyperbolic versions of important deep learning tools: multinomial logistic regression, feed-forward and recurrent neural networks such as gated recurrent units. This allows to embed sequential data and perform classification in the hyperbolic space. Empirically, we show that, even if hyperbolic optimization tools are limited, hyperbolic sentence embeddings either outperform or are on par with their Euclidean variants on textual entailment and noisy-prefix recognition tasks."
                },
                {
                    "title": "Gradient Descent for One-Hidden-Layer Neural Networks: Polynomial Convergence and SQ Lower Bounds",
                    "abstract": "We study the complexity of training neural network models with one hidden nonlinear activation layer and an output weighted sum layer. We analyze Gradient Descent applied to learning a bounded target function on $n$ real-valued inputs. We give an agnostic learning guarantee for GD: starting from a randomly initialized network, it converges in mean squared loss to the minimum error (in $2$-norm) of the best approximation of the target function using a polynomial of degree at most $k$. Moreover, for any $k$, the size of the network and number of iterations needed are both bounded by $n^{O(k)}\\log(1/\\epsilon)$. In particular, this applies to training networks of unbiased sigmoids and ReLUs. We also rigorously explain the empirical finding that gradient descent discovers lower frequency Fourier components before higher frequency components. \nWe complement this result with nearly matching lower bounds in the Statistical Query model. GD fits well in the SQ framework since each training step is determined by an expectation over the input distribution. We show that any SQ algorithm that achieves significant improvement over a constant function with queries of tolerance some inverse polynomial in the input dimensionality $n$ must use $n^{\\Omega(k)}$ queries even when the target functions are restricted to a set of $n^{O(k)}$ degree-$k$ polynomials, and the input distribution is uniform over the unit sphere; for this class the information-theoretic lower bound is only $\\Theta(k \\log n)$. \nOur approach for both parts is based on spherical harmonics. We view gradient descent as an operator on the space of functions, and study its dynamics. An essential tool is the Funk-Hecke theorem, which explains the eigenfunctions of this operator in the case of the mean squared loss."
                },
                {
                    "title": "Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms",
                    "abstract": "We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at this https URL"
                },
                {
                    "title": "On the Complexity of Learning Neural Networks",
                    "abstract": "The stunning empirical successes of neural networks currently lack rigorous theoretical explanation. What form would such an explanation take, in the face of existing complexity-theoretic lower bounds? A first step might be to show that data generated by neural networks with a single hidden layer, smooth activation functions and benign input distributions can be learned efficiently. We demonstrate here a comprehensive lower bound ruling out this possibility: for a wide class of activation functions (including all currently used), and inputs drawn from any logconcave distribution, there is a family of one-hidden-layer functions whose output is a sum gate, that are hard to learn in a precise sense: any statistical query algorithm (which includes all known variants of stochastic gradient descent with any loss function) needs an exponential number of queries even using tolerance inversely proportional to the input dimensionality. Moreover, this hard family of functions is realizable with a small (sublinear in dimension) number of activation units in the single hidden layer. The lower bound is also robust to small perturbations of the true weights. Systematic experiments illustrate a phase transition in the training error as predicted by the analysis."
                },
                {
                    "title": "Estimating the reach of a manifold",
                    "abstract": "Various problems in manifold estimation make use of a quantity called the reach, denoted by $\\tau\\_M$, which is a measure of the regularity of the manifold. This paper is the first investigation into the problem of how to estimate the reach. First, we study the geometry of the reach through an approximation perspective. We derive new geometric results on the reach for submanifolds without boundary. An estimator $\\hat{\\tau}$ of $\\tau\\_{M}$ is proposed in a framework where tangent spaces are known, and bounds assessing its efficiency are derived. In the case of i.i.d. random point cloud $\\mathbb{X}\\_{n}$, $\\hat{\\tau}(\\mathbb{X}\\_{n})$ is showed to achieve uniform expected loss bounds over a $\\mathcal{C}^3$-like model. Finally, we obtain upper and lower bounds on the minimax rate for estimating the reach."
                },
                {
                    "title": "Deep Sets",
                    "abstract": "In this paper, we study the problem of designing objective functions for machine learning problems defined on finite \\emph{sets}. In contrast to traditional objective functions defined for machine learning problems operating on finite dimensional vectors, the new objective functions we propose are operating on finite sets and are invariant to permutations. Such problems are widespread, ranging from estimation of population statistics \\citep{poczos13aistats}, via anomaly detection in piezometer data of embankment dams \\citep{Jung15Exploration}, to cosmology \\citep{Ntampaka16Dynamical,Ravanbakhsh16ICML1}. Our main theorem characterizes the permutation invariant objective functions and provides a family of functions to which any permutation invariant objective function must belong. This family of functions has a special structure which enables us to design a deep network architecture that can operate on sets and which can be deployed on a variety of scenarios including both unsupervised and supervised learning tasks. We demonstrate the applicability of our method on population statistic estimation, point cloud classification, set expansion, and image tagging."
                },
                {
                    "title": "Geometric Deep Learning: Going beyond Euclidean data",
                    "abstract": "Many scientific fields study data with an underlying structure that is non-Euclidean. Some examples include social networks in computational social sciences, sensor networks in communications, functional networks in brain imaging, regulatory networks in genetics, and meshed surfaces in computer graphics. In many applications, such geometric data are large and complex (in the case of social networks, on the scale of billions) and are natural targets for machine-learning techniques. In particular, we would like to use deep neural networks, which have recently proven to be powerful tools for a broad range of problems from computer vision, natural-language processing, and audio analysis. However, these tools have been most successful on data with an underlying Euclidean or grid-like structure and in cases where the invariances of these structures are built into networks used to model them."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Group Equivariant Convolutional Networks",
                    "abstract": "We introduce Group equivariant Convolutional Neural Networks (G-CNNs), a natural generalization of convolutional neural networks that reduces sample complexity by exploiting symmetries. G-CNNs use G-convolutions, a new type of layer that enjoys a substantially higher degree of weight sharing than regular convolution layers. G-convolutions increase the expressive capacity of the network without increasing the number of parameters. Group convolution layers are easy to use and can be implemented with negligible computational overhead for discrete groups generated by translations, reflections and rotations. G-CNNs achieve state of the art results on CI- FAR10 and rotated MNIST."
                },
                {
                    "title": "Testing the Manifold Hypothesis",
                    "abstract": "The hypothesis that high dimensional data tend to lie in the vicinity of a low dimensional manifold is the basis of manifold learning. The goal of this paper is to develop an algorithm (with accompanying complexity guarantees) for fitting a manifold to an unknown probability distribution supported in a separable Hilbert space, only using i.i.d samples from that distribution. More precisely, our setting is the following. Suppose that data are drawn independently at random from a probability distribution $P$ supported on the unit ball of a separable Hilbert space $H$. Let $G(d, V, \\tau)$ be the set of submanifolds of the unit ball of $H$ whose volume is at most $V$ and reach (which is the supremum of all $r$ such that any point at a distance less than $r$ has a unique nearest point on the manifold) is at least $\\tau$. Let $L(M, P)$ denote mean-squared distance of a random point from the probability distribution $P$ to $M$. \nWe obtain an algorithm that tests the manifold hypothesis in the following sense. \nThe algorithm takes i.i.d random samples from $P$ as input, and determines which of the following two is true (at least one must be): \n(a) There exists $M \\in G(d, CV, \\frac{\\tau}{C})$ such that $L(M, P) \\leq C \\epsilon.$ \n(b) There exists no $M \\in G(d, V/C, C\\tau)$ such that $L(M, P) \\leq \\frac{\\epsilon}{C}.$ \nThe answer is correct with probability at least $1-\\delta$."
                },
                {
                    "title": "The MNIST Database of Handwritten Digit Images for Machine Learning Research [Best of the Web]",
                    "abstract": "In this issue, \u201cBest of the Web\u201d presents the modified National Institute of Standards and Technology (MNIST) resources, consisting of a collection of handwritten digit images used extensively in optical character recognition and machine learning research."
                },
                {
                    "title": "Manifold Learning Theory and Applications",
                    "abstract": "Trained to extract actionable information from large volumes of high-dimensional data, engineers and scientists often have trouble isolating meaningful low-dimensional structures hidden in their high-dimensional observations. Manifold learning, a groundbreaking technique designed to tackle these issues of dimensionality reduction, finds widespread application in machine learning, neural networks, pattern recognition, image processing, and computer vision. Filling a void in the literature, Manifold Learning Theory and Applications incorporates state-of-the-art techniques in manifold learning with a solid theoretical and practical treatment of the subject. Comprehensive in its coverage, this pioneering work explores this novel modality from algorithm creation to successful implementationoffering examples of applications in medical, biometrics, multimedia, and computer vision. Emphasizing implementation, it highlights the various permutations of manifold learning in industry including manifold optimization, large scale manifold learning, semidefinite programming for embedding, manifold models for signal acquisition, compression and processing, and multi scale manifold. Beginning with an introduction to manifold learning theories and applications, the book includes discussions on the relevance to nonlinear dimensionality reduction, clustering, graph-based subspace learning, spectral learning and embedding, extensions, and multi-manifold modeling. It synergizes cross-domain knowledge for interdisciplinary instructions, offers a rich set of specialized topics contributed by expert professionals and researchers from a variety of fields. Finally, the book discusses specific algorithms and methodologies using case studies to apply manifold learning for real-world problems."
                },
                {
                    "title": "Sample Complexity of Testing the Manifold Hypothesis",
                    "abstract": "The hypothesis that high dimensional data tends to lie in the vicinity of a low dimensional manifold is the basis of a collection of methodologies termed Manifold Learning. In this paper, we study statistical aspects of the question of fitting a manifold with a nearly optimal least squared error. Given upper bounds on the dimension, volume, and curvature, we show that Empirical Risk Minimization can produce a nearly optimal manifold using a number of random samples that is independent of the ambient dimension of the space in which data lie. We obtain an upper bound on the required number of samples that depends polynomially on the curvature, exponentially on the intrinsic dimension, and linearly on the intrinsic volume. For constant error, we prove a matching minimax lower bound on the sample complexity that shows that this dependence on intrinsic dimension, volume and curvature is unavoidable. Whether the known lower bound of O(k/e2 + log 1/\u03b4/e2) for the sample complexity of Empirical Risk minimization on k-means applied to data in a unit ball of arbitrary dimension is tight, has been an open question since 1997 [3]. Here e is the desired bound on the error and \u03b4 is a bound on the probability of failure. We improve the best currently known upper bound [14] of O(k2/e2 + log 1/\u03b4/e2) to O(k/e2 (min (k, + log4 k/e/e2)) + log 1/\u03b4/e2). Based on these results, we devise a simple algorithm for k-means and another that uses a family of convex programs to fit a piecewise linear curve of a specified length to high dimensional data, where the sample complexity is independent of the ambient dimension."
                },
                {
                    "title": "Introduction to the non-asymptotic analysis of random matrices",
                    "abstract": "This is a tutorial on some basic non-asymptotic methods and concepts in random matrix theory. The reader will learn several tools for the analysis of the extreme singular values of random matrices with independent rows or columns. Many of these methods sprung off from the development of geometric functional analysis since the 1970's. They have applications in several fields, most notably in theoretical computer science, statistics and signal processing. A few basic applications are covered in this text, particularly for the problem of estimating covariance matrices in statistics and for validating probabilistic constructions of measurement matrices in compressed sensing. These notes are written particularly for graduate students and beginning researchers in different areas, including functional analysts, probabilists, theoretical statisticians, electrical engineers, and theoretical computer scientists."
                },
                {
                    "title": "Minimax Manifold Estimation",
                    "abstract": "We find the minimax rate of convergence in Hausdorff distance for estimating a manifold M of dimension d embedded in R^D given a noisy sample from the manifold. We assume that the manifold satisfies a smoothness condition and that the noise distribution has compact support. We show that the optimal rate of convergence is n^{-2/(2+d)}. Thus, the minimax rate depends only on the dimension of the manifold, not on the dimension of the space in which M is embedded."
                },
                {
                    "title": "Manifold Reconstruction Using Tangential Delaunay Complexes",
                    "abstract": "We give a provably correct algorithm to reconstruct a k-dimensional smooth manifold embedded in d-dimensional Euclidean space. The input to our algorithm is a point sample coming from an unknown manifold. Our approach is based on two main ideas: the notion of tangential Delaunay complex defined in Boissonnat and Fl\u00f6totto (Comput. Aided Des. 36:161\u2013174, 2004), Fl\u00f6totto (A coordinate system associated to a point cloud issued from a manifold: definition, properties and applications. Ph.D. thesis, 2003), Freedman (IEEE Trans. Pattern Anal. Mach. Intell. 24(10), 2002), and the technique of sliver removal by weighting the sample points (Cheng et al. in J. ACM 47:883\u2013904, 2000). Differently from previous methods, we do not construct any subdivision of the d-dimensional ambient space. As a result, the running time of our algorithm depends only linearly on the extrinsic dimension d while it depends quadratically on the size of the input sample, and exponentially on the intrinsic dimension k. To the best of our knowledge, this is the first certified algorithm for manifold reconstruction whose complexity depends linearly on the ambient dimension. We also prove that for a dense enough sample the output of our algorithm is isotopic to the manifold and a close geometric approximation of the manifold."
                },
                {
                    "title": "Topology and data",
                    "abstract": "An important feature of modern science and engineering is that data of various kinds is being produced at an unprecedented rate. This is so in part because of new experimental methods, and in part because of the increase in the availability of high powered computing technology. It is also clear that the nature of the data we are obtaining is significantly different. For example, it is now often the case that we are given data in the form of very long vectors, where all but a few of the coordinates turn out to be irrelevant to the questions of interest, and further that we don\u2019t necessarily know which coordinates are the interesting ones. A related fact is that the data is often very high-dimensional, which severely restricts our ability to visualize it. The data obtained is also often much noisier than in the past and has more missing information (missing data). This is particularly so in the case of biological data, particularly high throughput data from microarray or other sources. Our ability to analyze this data, both in terms of quantity and the nature of the data, is clearly not keeping pace with the data being produced. In this paper, we will discuss how geometry and topology can be applied to make useful contributions to the analysis of various kinds of data. Geometry and topology are very natural tools to apply in this direction, since geometry can be regarded as the study of distance functions, and what one often works with are distance functions on large finite sets of data. The mathematical formalism which has been developed for incorporating geometric and topological techniques deals with point clouds, i.e. finite sets of points equipped with a distance function. It then adapts tools from the various branches of geometry to the study of point clouds. The point clouds are intended to be thought of as finite samples taken from a geometric object, perhaps with noise. Here are some of the key points which come up when applying these geometric methods to data analysis. \u2022 Qualitative information is needed: One important goal of data analysis is to allow the user to obtain knowledge about the data, i.e. to understand how it is organized on a large scale. For example, if we imagine that we are looking at a data set constructed somehow from diabetes patients, it would be important to develop the understanding that there are two types of the disease, namely the juvenile and adult onset forms. Once that is established, one of course wants to develop quantitative methods for distinguishing them, but the first insight about the distinct forms of the disease is key."
                },
                {
                    "title": "Random projection trees and low dimensional manifolds",
                    "abstract": "We present a simple variant of the k-d tree which automatically adapts to intrinsic low dimensional structure in data without having to explicitly learn this structure."
                },
                {
                    "title": "Manifold reconstruction from point samples",
                    "abstract": "We present an algorithm to \"reconstruct\" a smooth <i>k</i>-dimensional manifold <i>M</i> embedded in an Euclidean space R<sup>d</sup> from a \"sufficiently dense\" point sample from the manifold. The algorithm outputs a simplicial manifold that is homeomorphic and geometrically close to <i>M.</i> The running time is <i>O</i>(<i>n</i> log <i>n</i>) where <i>n</i> is the number of points in the sample (the multiplicative constant depends exponentially on the dimension though)."
                },
                {
                    "title": "Maximum Likelihood Estimation of Intrinsic Dimension",
                    "abstract": "We propose a new method for estimating intrinsic dimension of a dataset derived by applying the principle of maximum likelihood to the distances between close neighbors. We derive the estimator by a Poisson process approximation, assess its bias and variance theoretically and by simulations, and apply it to a number of simulated and real datasets. We also show it has the best overall performance compared with two other intrinsic dimension estimators."
                },
                {
                    "title": "Programming the Hilbert curve",
                    "abstract": "The Hilbert curve has previously been constructed recursively, using p levels of recursion of n\u2010bit Gray codes to attain a precision of p bits in n dimensions. Implementations have reflected the awkwardness of aligning the recursive steps to preserve geometrical adjacency. We point out that a single global Gray code can instead be applied to all np bits of a Hilbert length. Although this \u201cover\u2010transforms\u201d the length, the excess work can be undone in a single pass over the bits, leading to compact and efficient computer code."
                },
                {
                    "title": "Laplacian Eigenmaps for Dimensionality Reduction and Data Representation",
                    "abstract": "One of the central problems in machine learning and pattern recognition is to develop appropriate representations for complex data. We consider the problem of constructing a representation for data lying on a low-dimensional manifold embedded in a high-dimensional space. Drawing on the correspondence between the graph Laplacian, the Laplace Beltrami operator on the manifold, and the connections to the heat equation, we propose a geometrically motivated algorithm for representing the high-dimensional data. The algorithm provides a computationally efficient approach to nonlinear dimensionality reduction that has locality-preserving properties and a natural connection to clustering. Some potential applications and illustrative examples are discussed."
                },
                {
                    "title": "Estimating the Intrinsic Dimension of Data with a Fractal-Based Method",
                    "abstract": "In this paper, the problem of estimating the intrinsic dimension of a data set is investigated. A fractal-based approach using the Grassberger-Procaccia algorithm is proposed. Since the Grassberger-Procaccia algorithm (1983) performs badly on sets of high dimensionality, an empirical procedure that improves the original algorithm has been developed. The procedure has been tested on data sets of known dimensionality and on time series of Santa Fe competition."
                },
                {
                    "title": "Nonlinear dimensionality reduction by locally linear embedding.",
                    "abstract": "Many areas of science depend on exploratory data analysis and visualization. The need to analyze large amounts of multivariate data raises the fundamental problem of dimensionality reduction: how to discover compact representations of high-dimensional data. Here, we introduce locally linear embedding (LLE), an unsupervised learning algorithm that computes low-dimensional, neighborhood-preserving embeddings of high-dimensional inputs. Unlike clustering methods for local dimensionality reduction, LLE maps its inputs into a single global coordinate system of lower dimensionality, and its optimizations do not involve local minima. By exploiting the local symmetries of linear reconstructions, LLE is able to learn the global structure of nonlinear manifolds, such as those generated by images of faces or documents of text."
                },
                {
                    "title": "Weakly learning DNF and characterizing statistical query learning using Fourier analysis",
                    "abstract": "We present new results, both positive and negative, on the well-studied problem of learning disjunctive normal form (DNF) expressions. We first prove that an algorithm due to Kushilevitz and Mansour [16] can be used to weakly learn DNF using membership queries in polynomial time, with respect to the uniform distribution on the inputs. This is the first positive result for learning unrestricted DNF expressions in polynomial time in any nontrivial formal model of learning. It provides a sharp contrast with the results of Kharitonov [15], who proved that ACO is not efficiently learnable in the same model (given certain plausible cryptographic assumptions). We also present efficient learning algorithms in various models for the read-k and SAT-k subclasses of DNF. For our negative results, we turn our attention to the recently introduced statistical query model of learning [11]. This model is a restricted version of the popular Probably Approximately Correct (PAC) model [23], and practically every class known to be efficiently learnable in the PAC model is in fact learnable in the statistical query model [11]. Here we give a general characterization of the complexity of statistical query learning in terms of the number of uncorrelated functions in the concept class. This is a distributiondependent quantity yielding upper and lower bounds on the number of st atistical queries required for learning on any input distribution. As a corollary, we obtain that DNF expressions and decision trees are not even weakly learnable with \u25cfThis research M sponsored in part by the Wr]ght Laboratory, Aeronautical Systems Center, Air Force Materiel Command, USAF, and the Advanced Research Projects Agency (ARPA) under grant number F33615-93-1-1330 Support also M sponsored by the National Sc]ence Foundation under Grant No CC-91 19319. Blum also supported m part by NSF National Young Investigator grant CCR9357793 Views and conclusions contained in this document are those of the authors and should not be interpreted as necessarily representing official po!lcles or endorsements, either expressed or implied, of Wright Laboratory or the United States Government, or NSF tcontact ~\u201cthor Address: AT&T Bell Laboratcmes, Room 2A423, 600 Mountain Avenue, P.O. Box 636, Murray Hill, NJ 07974 Electronic mail. mkearns@research .at t corn ~Thi~ research ~a~ ~\u201cpported in p~~t by The Israel science Foun. datlon administered by The Israel Academy of Sc]ence and Humanities and by a grant of the Israeli Ministry of Science and Technology Permission to co y without fee all or part of this material is granted provided%atthe copies are not madeordistrftrutectfor direct commercial advantage, the ACM copyright notice and the title of the publication and its date appear, and notice is given that copying is by permission of the Association of Computing Machinery. To copy otherwise, or to republish, requires a fee ancf/or specific permission. STOC 945/94 Montreal, Quebec, Canada Q 1994 ACM 0-89791 -663-8/94/0005..$3.50 respect to the uniform input distribution in polynomial time in the statistical query model. This result is informationtheoretic and therefore does not rely on any unproven assumptions. It demonstrates that no simple modification of the existing algorithms in the computaticmal learning theory literature for learning various restricted forms of DNF and decision trees from passive random examples (and also several algorithms proposed in the experimental machine learning communities, such as the ID3 algorithm for decision trees [22] and its variants) will solve the general problem. The unifying tool for all of our results is the Fourier analysis of a finite class of boolean functions 011 the hypercube."
                },
                {
                    "title": "Fast Non-Linear Dimension Reduction",
                    "abstract": "We present a fast algorithm for non-linear dimension reduction. The algorithm builds a local linear model of the data by merging PCA with clustering based on a new distortion measure. Experiments with speech and image data indicate that the local linear algorithm produces encodings with lower distortion than those built by five layer auto-associative networks. The local linear algorithm is also more than an order of magnitude faster to train."
                },
                {
                    "title": "Efficient noise-tolerant learning from statistical queries",
                    "abstract": "In this paper, we study the problem of learning in the presence of classification noise in the probabilistic learning model of Valiant and its variants. In order to identify the class of \u201crobust\u201d learning algorithms in the most general way, we formalize a new but related model of learning from statistical queries. Intuitively, in this model, a learning algorithm is forbidden to examine individual examples of the unknown target function, but is given access to an oracle providing estimates of probabilities over the sample space of random examples. One of our main results shows that any class of functions learnable from statistical queries is in fact learnable with classification noise in Valiant\u2019s model, with a noise rate approaching the informationtheoretic barrier of 1/2. We then demonstrate the generality of the statistical query model, showing that practically every class learnable in Valiant\u2019s model and its variants can also be learned in the new model (and thus can be learned in the presence of noise). A notable exception to this statement is the class of parity functions, which we prove is not learnable from statistical queries, and for which no noise-tolerant algorithm is known."
                },
                {
                    "title": "An Algorithm for Finding Intrinsic Dimensionality of Data",
                    "abstract": "An algorithm for the analysis of multivariant data is presented along with some experimental results. The basic idea of the method is to examine the data in many small subregions, and from this determine the number of governing parameters, or intrinsic dimensionality. This intrinsic dimensionality is usually much lower than the dimensionality that is given by the standard Karhunen-Lo\u00e8ve technique. An analysis that demonstrates the feasability of this approach is presented."
                },
                {
                    "title": "Tight minimax rates for manifold estimation under Hausdorff loss",
                    "abstract": "This paper deals with minimax rates of convergence for mani- fold estimation. A new lower bound is obtained by a novel construction of two sets of manifolds and an application of convex hull testing method of Le Cam (1973). The minimax lower bound matches the upper bound up to a constant factor. MSC 2010 subject classications: Primary 62C25, 62G86; secondary 65C50."
                },
                {
                    "title": "On the Sample Complexity of Learning Smooth Cuts on a Manifold",
                    "abstract": "Modern data sets, though typically high dimensional, are often generated by processes possessing few essential degrees of freedom, as is the case with human speech. In recent years, such considerations have lead to the notion that high dimensional data may be modeled to lie on a submanifold of low intrinsic dimension. We derive bounds on the number of random samples needed before it is possible to approximately separate data into two classes using smooth decision boundaries with high probability."
                },
                {
                    "title": "Algorithms for manifold learning",
                    "abstract": ","
                },
                {
                    "title": "Intrinsic Dimension Estimation Using Packing Numbers",
                    "abstract": "We propose a new algorithm to estimate the intrinsic dimension of data sets. The method is based on geometric properties of the data and requires neither parametric assumptions on the data generating model nor input parameters to set. The method is compared to a similar, widely-used algorithm from the same family of geometric techniques. Experiments show that our method is more robust in terms of the data generating distribution and more reliable in the presence of noise."
                },
                {
                    "title": "ON A CLASSICAL PROBLEM OF PROBABILITY THEORY b",
                    "abstract": "ball may be placed in any of the urns with the same probability and the choices of the urns for the different balls are independent. We continue this process so long Cl1 there are at least m balls in every urn (m = I, 2, . . .). What can be said about the number of balls which are needed to achieve this goal! We denote the number in question (which is of course a random variable) by v,(n). The \u201cdixie cup\u201d-problem considered in [l] is clearly equivalent with the above problem. In Cl] the mean value M(v,(n)) of p,(n) has been evaluated (here and in what follows M( ) denotes the mean value of the random variable in the brackets) and it has been shown that"
                },
                {
                    "title": "Graph approximations to geodesics on embedded manifolds",
                    "abstract": "0 Introduction In [1] Tenenbaum, de Silva and Langford consider the problem of non-linear dimensionality reduction: discovering intrinsically low-dimensional structures embedded in high-dimensional data sets. They describe an algorithm, called Isomap, and demonstrate its successful application to several real and synthetic data sets. In this paper, we discuss some of the theoretical claims for Isomap made in [1]. In particular, we give a full proof of the asymptotic convergence theorem referred to in that paper. Isomap deals with finite data sets of points in R n which are assumed to lie on a smooth submanifold M d of low dimension d < n. The algorithm attempts to recover M given only the data points. A crucial stage in the algorithm involves estimating the unknown geodesic distance in M between data points in terms of the graph distance with respect to some graph G constructed on the data points. We show that the two distance metrics approximate each other arbitrarily closely, as the density of data points tends to infinity. Main Theorem A * The authors are listed alphabetically."
                },
                {
                    "title": "Automatic Choice of Dimensionality for PCA",
                    "abstract": "A central issue in principal component analysis (PCA) is choosing the number of principal components to be retained. By interpreting PCA as density estimation, we show how to use Bayesian model selection to estimate the true dimensionality of the data. The resulting estimate is simple to compute yet guaranteed to pick the correct dimensionality, given enough data. The estimate involves an integral over the Steifel manifold of k-frames, which is difficult to compute exactly. But after choosing an appropriate parameterization and applying Laplace's method, an accurate and practical estimator is obtained. In simulations, it is convincingly better than cross-validation and other proposed algorithms, plus it runs much faster."
                },
                {
                    "title": "A Global Geometric Framework for Nonlinear Dimensionality Reduction",
                    "abstract": "29, 31, and 45; right 17, 35, and 37). 17. Although the improvement in WM performance with cholinergic enhancement was a nonsignificant trend in the current study (P 5 0.07), in a previous study (9) with a larger sample (n 5 13) the effect was highly significant (P , 0.001). In the current study, we analyzed RT data for six of our seven subjects because the behavioral data for one subject were unavailable due to a computer failure. The difference in the significance of the two findings is simply a result of the difference in sample sizes. A power analysis shows that the size of the RT difference and variability in the current sample would yield a significant result (P 5 0.01) with a sample size of 13. During the memory trials, mean RT was 1180 ms during placebo and 1119 ms during physostigmine. During the control trials, mean RT was 735 ms during placebo and 709 ms during physostigmine, a difference that did not approach significance (P 5 0.24), suggesting that the effect of cholinergic enhancement on WM performance is not due to a nonspecific increase in arousal. 18. Matched-pair t tests (two-tailed) were used to test the significance of drug-related changes in the volume of regions of interest that showed significant response contrasts. 19. H. Sato, Y. Hata, H. Masui, T. Tsumoto, J. Neurophysiol. 55, 765 (1987). 20. M. E. Hasselmo, Behav. Brain Res. 67, 1 (1995). 21. M. G. Baxter, A. A. Chiba, Curr. Opin. Neurobiol. 9, 178 (1999). 22. B. J. Everitt, T. W. Robbins, Annu. Rev. Psychol. 48, 649 (1997). 23. R. Desimone, J. Duncan, Annu. Rev. Neurosci. 18, 193 (1995). 24. P. C. Murphy, A. M. Sillito, Neuroscience 40, 13 (1991). 25. M. Corbetta, F. M. Miezin, S. Dobmeyer, G. L. Shulman, S. E. Peterson, J. Neurosci. 11, 2383 (1991). 26. J. V. Haxby et al., J. Neurosci. 14, 6336 (1994). 27. A. Rosier, L. Cornette, G. A. Orban, Neuropsychobiology 37, 98 (1998). 28. M. E. Hasselmo, B. P. Wyble, G. V. Wallenstein, Hippocampus 6, 693 (1996). 29. S. P. Mewaldt, M. M. Ghoneim, Pharmacol. Biochem. Behav. 10, 1205 (1979). 30. M. Petrides, Philos. Trans. R. Soc. London Ser. B 351, 1455 (1996). 31. M. E. Hasselmo, E. Fransen, C. Dickson, A. A. Alonso, Ann. N.Y. Acad. Sci. 911, 418 (2000). 32. M. M. Mesulam, Prog. Brain Res. 109, 285 (1996). 33. R. T. Bartus, R. L. Dean III, B. Beer, A. S. Lippa, Science 217, 408 (1985). 34. N. Qizilbash et al., JAMA 280, 1777 (1998). 35. J. V. Haxby, J. Ma. Maisog, S. M. Courtney, in Mapping and Modeling the Human Brain, P. Fox, J. Lancaster, K. Friston, Eds. ( Wiley, New York, in press). 36. We express our appreciation to S. Courtney, R. Desimone, Y. Jiang, S. Kastner, L. Latour, A. Martin, L. Pessoa, and L. Ungerleider for careful and critical review of the manuscript. We also thank M. B. Schapiro and S. I. Rapoport for input during early stages of this project. This research was supported by the National Institute on Mental Health and National Institute on Aging Intramural Research Programs."
                }
            ],
            "categories": [
                "cs.LG",
                "math.DG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nWhich assumptions on the data geometry guarantee the learnability of neural networks?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant gap in understanding the relationship between data geometry and the learnability of neural networks. By establishing formal results on the impact of the manifold hypothesis, this research could lead to advancements in machine learning algorithms that are more efficient and effective in high-dimensional spaces. Furthermore, it could pave the way for practical applications in various domains where high-dimensional data is prevalent, such as computer vision, natural language processing, and bioinformatics. The findings could influence future research directions by encouraging the exploration of geometric assumptions in neural network design and training.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complex interplay between data geometry and the computational complexity of learning algorithms. Naive approaches may fail because they do not account for the specific geometric properties of the data, which can significantly affect the learnability of neural networks. Technical obstacles include the need to establish formal guarantees on learnability under various geometric assumptions, as well as the difficulty in characterizing low-dimensional manifolds accurately. Theoretical challenges arise from the existing hardness results for neural networks, which indicate that additional assumptions are necessary to achieve learnability, complicating the analysis.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either the computational hardness of learning neural networks or the geometric properties of data, but rarely have these two areas been integrated. Limitations in existing solutions include a lack of formal results connecting data geometry to neural network learnability and insufficient exploration of the implications of the manifold hypothesis. Barriers such as the complexity of the statistical query (SQ) model and the need for new proof techniques have hindered progress. This research differs by extending hardness results to more general geometries and providing a framework for understanding how specific geometric assumptions can lead to learnability.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves analyzing the learnability of neural networks under various geometric assumptions about the data manifold. The approach will utilize the statistical query (SQ) model and cryptographic hardness assumptions to establish hardness results for low-dimensional manifolds. The dataset will consist of high-dimensional data with known geometric properties, and the metrics for evaluation will include learnability guarantees and computational efficiency."
            }
        },
        "author_data": {
            "04ce9562-a6ec-4a60-b3d3-8bb462f7cb44": {
                "pk": "04ce9562-a6ec-4a60-b3d3-8bb462f7cb44",
                "name": "Bobak T. Kiani",
                "collaborators": [
                    "Seth Lloyd",
                    "Eric R. Anschuetz",
                    "Hannah Lawrence",
                    "Giacomo De Palma",
                    "Yann LeCun",
                    "Melanie Weber",
                    "David Gamarnik",
                    "Quynh T. Nguyen",
                    "Randall Balestriero",
                    "Dirk Englund"
                ],
                "domain": [
                    "Quantum Computing",
                    "Graph Neural Network",
                    "Self-Supervised Learning",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Unitary convolutions for learning on graphs and groups",
                        "abstract": "Data with geometric structure is ubiquitous in machine learning often arising from fundamental symmetries in a domain, such as permutation-invariance in graphs and translation-invariance in images. Group-convolutional architectures, which encode symmetries as inductive bias, have shown great success in applications, but can suffer from instabilities as their depth increases and often struggle to learn long range dependencies in data. For instance, graph neural networks experience instability due to the convergence of node representations (over-smoothing), which can occur after only a few iterations of message-passing, reducing their effectiveness in downstream tasks. Here, we propose and study unitary group convolutions, which allow for deeper networks that are more stable during training. The main focus of the paper are graph neural networks, where we show that unitary graph convolutions provably avoid over-smoothing. Our experimental results confirm that unitary graph convolutional networks achieve competitive performance on benchmark datasets compared to state-of-the-art graph neural networks. We complement our analysis of the graph domain with the study of general unitary convolutions and analyze their role in enhancing stability in general group convolutional architectures."
                    },
                    {
                        "title": "Beyond Barren Plateaus: Quantum Variational Algorithms Are Swamped With Traps",
                        "abstract": "One of the most important properties of classical neural networks is how surprisingly trainable they are, though their training algorithms typically rely on optimizing complicated, nonconvex loss functions. Previous results have shown that unlike the case in classical neural networks, variational quantum models are often not trainable. The most studied phenomenon is the onset of barren plateaus in the training landscape of these quantum models, typically when the models are very deep. This focus on barren plateaus has made the phenomenon almost synonymous with the trainability of quantum models. Here, we show that barren plateaus are only a part of the story. We prove that a wide class of variational quantum models -- which are shallow, and exhibit no barren plateaus -- have only a superpolynomially small fraction of local minima within any constant energy from the global minimum, rendering these models untrainable if no good initial guess of the optimal parameters is known. We also study the trainability of variational quantum algorithms from a statistical query framework, and show that noisy optimization of a wide variety of quantum models is impossible with a sub-exponential number of queries. Finally, we numerically confirm our results on a variety of problem instances. Though we exclude a wide variety of quantum algorithms here, we give reason for optimism for certain classes of variational algorithms and discuss potential ways forward in showing the practical utility of such algorithms."
                    },
                    {
                        "title": "Implicit Bias of Linear Equivariant Networks",
                        "abstract": "Group equivariant convolutional neural networks (G-CNNs) are generalizations of convolutional neural networks (CNNs) which excel in a wide range of technical applications by explicitly encoding symmetries, such as rotations and permutations, in their architectures. Although the success of G-CNNs is driven by their \\emph{explicit} symmetry bias, a recent line of work has proposed that the \\emph{implicit} bias of training algorithms on particular architectures is key to understanding generalization for overparameterized neural nets. In this context, we show that $L$-layer full-width linear G-CNNs trained via gradient descent for binary classification converge to solutions with low-rank Fourier matrix coefficients, regularized by the $2/L$-Schatten matrix norm. Our work strictly generalizes previous analysis on the implicit bias of linear CNNs to linear G-CNNs over all finite groups, including the challenging setting of non-commutative groups (such as permutations), as well as band-limited G-CNNs over infinite groups. We validate our theorems via experiments on a variety of groups, and empirically explore more realistic nonlinear networks, which locally capture similar regularization patterns. Finally, we provide intuitive interpretations of our Fourier space implicit regularization results in real space via uncertainty principles."
                    },
                    {
                        "title": "Product states optimize quantum $p$-spin models for large $p$",
                        "abstract": "We consider the problem of estimating the maximal energy of quantum $p$-local spin glass random Hamiltonians, the quantum analogues of widely studied classical spin glass models. Denoting by $E^*(p)$ the (appropriately normalized) maximal energy in the limit of a large number of qubits $n$, we show that $E^*(p)$ approaches $\\sqrt{2\\log 6}$ as $p$ increases. This value is interpreted as the maximal energy of a much simpler so-called Random Energy Model, widely studied in the setting of classical spin glasses.   Our most notable and (arguably) surprising result proves the existence of near-maximal energy states which are product states, and thus not entangled. Specifically, we prove that with high probability as $n\\to\\infty$, for any $E<E^*(p)$ there exists a product state with energy $\\geq E$ at sufficiently large constant $p$. Even more surprisingly, this remains true even when restricting to tensor products of Pauli eigenstates. Our approximations go beyond what is known from monogamy-of-entanglement style arguments -- the best of which, in this normalization, achieve approximation error growing with $n$. Our results not only challenge prevailing beliefs in physics that extremely low-temperature states of random local Hamiltonians should exhibit non-negligible entanglement, but they also imply that classical algorithms can be just as effective as quantum algorithms in optimizing Hamiltonians with large locality -- though performing such optimization is still likely a hard problem.   Our results are robust with respect to the choice of the randomness (disorder) and apply to the case of sparse random Hamiltonian using Lindeberg's interpolation method. The proof of the main result is obtained by estimating the expected trace of the associated partition function, and then matching its asymptotics with the extremal energy of product states using the second moment method."
                    },
                    {
                        "title": "Bounds on the ground state energy of quantum $p$-spin Hamiltonians",
                        "abstract": "We consider the problem of estimating the ground state energy of quantum $p$-local spin glass random Hamiltonians, the quantum analogues of widely studied classical spin glass models. Our main result shows that the maximum energy achievable by product states has a well-defined limit (for even $p$) as $n\\to\\infty$ and is $E_{\\text{product}}^\\ast=\\sqrt{2 \\log p}$ in the limit of large $p$. This value is interpreted as the maximal energy of a much simpler so-called Random Energy Model, widely studied in the setting of classical spin glasses. The proof of the limit existing follows from an extension of Fekete's Lemma after we demonstrate near super-additivity of the (normalized) quenched free energy. The proof of the value follows from a second moment method on the number of states achieving a given energy when restricting to an $\\epsilon$-net of product states.   Furthermore, we relate the maximal energy achieved over all states to a $p$-dependent constant $\\gamma\\left(p\\right)$, which is defined by the degree of violation of a certain asymptotic independence ansatz over graph matchings. We show that the maximal energy achieved by all states $E^\\ast\\left(p\\right)$ in the limit of large $n$ is at most $\\sqrt{\\gamma\\left(p\\right)}E_{\\text{product}}^\\ast$. We also prove using Lindeberg's interpolation method that the limiting $E^\\ast\\left(p\\right)$ is robust with respect to the choice of the randomness and, for instance, also applies to the case of sparse random Hamiltonians. This robustness in the randomness extends to a wide range of random Hamiltonian models including SYK and random quantum max-cut."
                    },
                    {
                        "title": "Adversarial Robustness Guarantees for Random Deep Neural Networks",
                        "abstract": "The reliability of deep learning algorithms is fundamentally challenged by the existence of adversarial examples, which are incorrectly classified inputs that are extremely close to a correctly classified input. We explore the properties of adversarial examples for deep neural networks with random weights and biases, and prove that for any $p\\ge1$, the $\\ell^p$ distance of any given input from the classification boundary scales as one over the square root of the dimension of the input times the $\\ell^p$ norm of the input. The results are based on the recently proved equivalence between Gaussian processes and deep neural networks in the limit of infinite width of the hidden layers, and are validated with experiments on both random deep neural networks and deep neural networks trained on the MNIST and CIFAR10 datasets. The results constitute a fundamental advance in the theoretical understanding of adversarial examples, and open the way to a thorough theoretical characterization of the relation between network architecture and robustness to adversarial perturbations."
                    },
                    {
                        "title": "Block-encoding dense and full-rank kernels using hierarchical matrices: applications in quantum numerical linear algebra",
                        "abstract": "Many quantum algorithms for numerical linear algebra assume black-box access to a block-encoding of the matrix of interest, which is a strong assumption when the matrix is not sparse. Kernel matrices, which arise from discretizing a kernel function $k(x,x')$, have a variety of applications in mathematics and engineering. They are generally dense and full-rank. Classically, the celebrated fast multipole method performs matrix multiplication on kernel matrices of dimension $N$ in time almost linear in $N$ by using the linear algebraic framework of hierarchical matrices. In light of this success, we propose a block-encoding scheme of the hierarchical matrix structure on a quantum computer. When applied to many physical kernel matrices, our method can improve the runtime of solving quantum linear systems of dimension $N$ to $O(\\kappa \\operatorname{polylog}(\\frac{N}{\\varepsilon}))$, where $\\kappa$ and $\\varepsilon$ are the condition number and error bound of the matrix operation. This runtime is near-optimal and, in terms of $N$, exponentially improves over prior quantum linear systems algorithms in the case of dense and full-rank kernel matrices. We discuss possible applications of our methodology in solving integral equations and accelerating computations in N-body problems."
                    },
                    {
                        "title": "Strongly interacting fermions are non-trivial yet non-glassy",
                        "abstract": "Random spin systems at low temperatures are glassy and feature computational hardness in finding low-energy states. We study the random all-to-all interacting fermionic Sachdev--Ye--Kitaev (SYK) model and prove that, in contrast, (I) the low-energy states have polynomial circuit depth, yet (II) the annealed and quenched free energies agree to inverse-polynomially low temperatures, ruling out a glassy phase transition in this sense. These results are derived by showing that fermionic and spin systems significantly differ in their commutation index, which quantifies the non-commutativity of Hamiltonian terms. Our results suggest that low-temperature strongly interacting fermions, unlike spins, belong in a classically nontrivial yet quantumly easy phase."
                    },
                    {
                        "title": "Joint Embedding Self-Supervised Learning in the Kernel Regime",
                        "abstract": "The fundamental goal of self-supervised learning (SSL) is to produce useful representations of data without access to any labels for classifying the data. Modern methods in SSL, which form representations based on known or constructed relationships between samples, have been particularly effective at this task. Here, we aim to extend this framework to incorporate algorithms based on kernel methods where embeddings are constructed by linear maps acting on the feature space of a kernel. In this kernel regime, we derive methods to find the optimal form of the output representations for contrastive and non-contrastive loss functions. This procedure produces a new representation space with an inner product denoted as the induced kernel which generally correlates points which are related by an augmentation in kernel space and de-correlates points otherwise. We analyze our kernel model on small datasets to identify common features of self-supervised learning algorithms and gain theoretical insights into their performance on downstream tasks."
                    },
                    {
                        "title": "Efficient classical algorithms for simulating symmetric quantum systems",
                        "abstract": "In light of recently proposed quantum algorithms that incorporate symmetries in the hope of quantum advantage, we show that with symmetries that are restrictive enough, classical algorithms can efficiently emulate their quantum counterparts given certain classical descriptions of the input. Specifically, we give classical algorithms that calculate ground states and time-evolved expectation values for permutation-invariant Hamiltonians specified in the symmetrized Pauli basis with runtimes polynomial in the system size. We use tensor-network methods to transform symmetry-equivariant operators to the block-diagonal Schur basis that is of polynomial size, and then perform exact matrix multiplication or diagonalization in this basis. These methods are adaptable to a wide range of input and output states including those prescribed in the Schur basis, as matrix product states, or as arbitrary quantum states when given the power to apply low depth circuits and single qubit measurements."
                    },
                    {
                        "title": "The SSL Interplay: Augmentations, Inductive Bias, and Generalization",
                        "abstract": "Self-supervised learning (SSL) has emerged as a powerful framework to learn representations from raw data without supervision. Yet in practice, engineers face issues such as instability in tuning optimizers and collapse of representations during training. Such challenges motivate the need for a theory to shed light on the complex interplay between the choice of data augmentation, network architecture, and training algorithm. We study such an interplay with a precise analysis of generalization performance on both pretraining and downstream tasks in a theory friendly setup, and highlight several insights for SSL practitioners that arise from our theory."
                    },
                    {
                        "title": "Self-Supervised Learning with Lie Symmetries for Partial Differential Equations",
                        "abstract": "Machine learning for differential equations paves the way for computationally efficient alternatives to numerical solvers, with potentially broad impacts in science and engineering. Though current algorithms typically require simulated training data tailored to a given setting, one may instead wish to learn useful information from heterogeneous sources, or from real dynamical systems observations that are messy or incomplete. In this work, we learn general-purpose representations of PDEs from heterogeneous data by implementing joint embedding methods for self-supervised learning (SSL), a framework for unsupervised representation learning that has had notable success in computer vision. Our representation outperforms baseline approaches to invariant tasks, such as regressing the coefficients of a PDE, while also improving the time-stepping performance of neural solvers. We hope that our proposed methodology will prove useful in the eventual development of general-purpose foundation models for PDEs. Code: https://github.com/facebookresearch/SSLForPDEs."
                    },
                    {
                        "title": "On the hardness of learning under symmetries",
                        "abstract": "We study the problem of learning equivariant neural networks via gradient descent. The incorporation of known symmetries (\"equivariance\") into neural nets has empirically improved the performance of learning pipelines, in domains ranging from biology to computer vision. However, a rich yet separate line of learning theoretic research has demonstrated that actually learning shallow, fully-connected (i.e. non-symmetric) networks has exponential complexity in the correlational statistical query (CSQ) model, a framework encompassing gradient descent. In this work, we ask: are known problem symmetries sufficient to alleviate the fundamental hardness of learning neural nets with gradient descent? We answer this question in the negative. In particular, we give lower bounds for shallow graph neural networks, convolutional networks, invariant polynomials, and frame-averaged networks for permutation subgroups, which all scale either superpolynomially or exponentially in the relevant input dimension. Therefore, in spite of the significant inductive bias imparted via symmetry, actually learning the complete classes of functions represented by equivariant neural networks via gradient descent remains hard."
                    },
                    {
                        "title": "Quantum algorithms for group convolution, cross-correlation, and equivariant transformations",
                        "abstract": "Group convolutions and cross-correlations, which are equivariant to the actions of group elements, are commonly used in mathematics to analyze or take advantage of symmetries inherent in a given problem setting. Here, we provide efficient quantum algorithms for performing linear group convolutions and cross-correlations on data stored as quantum states. Runtimes for our algorithms are logarithmic in the dimension of the group thus offering an exponential speedup compared to classical algorithms when input data is provided as a quantum state and linear operations are well conditioned. Motivated by the rich literature on quantum algorithms for solving algebraic problems, our theoretical framework opens a path for quantizing many algorithms in machine learning and numerical methods that employ group operations."
                    },
                    {
                        "title": "Quantum advantage for differential equation analysis",
                        "abstract": "Quantum algorithms for both differential equation solving and for machine learning potentially offer an exponential speedup over all known classical algorithms. However, there also exist obstacles to obtaining this potential speedup in useful problem instances. The essential obstacle for quantum differential equation solving is that outputting useful information may require difficult post-processing, and the essential obstacle for quantum machine learning is that inputting the training set is a difficult task just by itself. In this paper, we demonstrate, when combined, these difficulties solve one another. We show how the output of quantum differential equation solving can serve as the input for quantum machine learning, allowing dynamical analysis in terms of principal components, power spectra, and wavelet decompositions. To illustrate this, we consider continuous time Markov processes on epidemiological and social networks. These quantum algorithms provide an exponential advantage over existing classical Monte Carlo methods."
                    },
                    {
                        "title": "Equivariant Polynomials for Graph Neural Networks",
                        "abstract": "Graph Neural Networks (GNN) are inherently limited in their expressive power. Recent seminal works (Xu et al., 2019; Morris et al., 2019b) introduced the Weisfeiler-Lehman (WL) hierarchy as a measure of expressive power. Although this hierarchy has propelled significant advances in GNN analysis and architecture developments, it suffers from several significant limitations. These include a complex definition that lacks direct guidance for model improvement and a WL hierarchy that is too coarse to study current GNNs. This paper introduces an alternative expressive power hierarchy based on the ability of GNNs to calculate equivariant polynomials of a certain degree. As a first step, we provide a full characterization of all equivariant graph polynomials by introducing a concrete basis, significantly generalizing previous results. Each basis element corresponds to a specific multi-graph, and its computation over some graph data input corresponds to a tensor contraction problem. Second, we propose algorithmic tools for evaluating the expressiveness of GNNs using tensor contraction sequences, and calculate the expressive power of popular GNNs. Finally, we enhance the expressivity of common GNN architectures by adding polynomial features or additional operations / aggregations inspired by our theory. These enhanced GNNs demonstrate state-of-the-art results in experiments across multiple graph learning benchmarks."
                    }
                ]
            },
            "6b546729-d8be-4c19-be2c-9f2b82cd20c4": {
                "pk": "6b546729-d8be-4c19-be2c-9f2b82cd20c4",
                "name": "Jason Wang",
                "collaborators": [
                    "Luis Perez",
                    "Ryan Rezai"
                ],
                "domain": [
                    "Data Augmentation",
                    "Deep Reinforcement Learning",
                    "Image Classification",
                    "Game Theory"
                ],
                "publications": [
                    {
                        "title": "The Effectiveness of Data Augmentation in Image Classification using Deep Learning",
                        "abstract": "In this paper, we explore and compare multiple solutions to the problem of data augmentation in image classification. Previous work has demonstrated the effectiveness of data augmentation through simple techniques, such as cropping, rotating, and flipping input images. We artificially constrain our access to data to a small subset of the ImageNet dataset, and compare each data augmentation technique in turn. One of the more successful data augmentations strategies is the traditional transformations mentioned above. We also experiment with GANs to generate images of different styles. Finally, we propose a method to allow a neural net to learn augmentations that best improve the classifier, which we call neural augmentation. We discuss the successes and shortcomings of this method on various datasets."
                    },
                    {
                        "title": "Closed Drafting as a Case Study for First-Principle Interpretability, Memory, and Generalizability in Deep Reinforcement Learning",
                        "abstract": "Closed drafting or \"pick and pass\" is a popular game mechanic where each round players select a card or other playable element from their hand and pass the rest to the next player. In this paper, we establish first-principle methods for studying the interpretability, generalizability, and memory of Deep Q-Network (DQN) models playing closed drafting games. In particular, we use a popular family of closed drafting games called \"Sushi Go Party\", in which we achieve state-of-the-art performance. We fit decision rules to interpret the decision-making strategy of trained DRL agents by comparing them to the ranking preferences of different types of human players. As Sushi Go Party can be expressed as a set of closely-related games based on the set of cards in play, we quantify the generalizability of DRL models trained on various sets of cards, establishing a method to benchmark agent performance as a function of environment unfamiliarity. Using the explicitly calculable memory of other player's hands in closed drafting games, we create measures of the ability of DRL models to learn memory."
                    }
                ]
            },
            "75f6e4d1-5fb8-4ace-9b1b-0854e778c469": {
                "pk": "75f6e4d1-5fb8-4ace-9b1b-0854e778c469",
                "name": "Melanie Weber",
                "collaborators": [
                    "Suvrit Sra",
                    "Emil Saucan",
                    "Lukas Fesser",
                    "J\u00fcrgen Jost",
                    "Andrew Cheng",
                    "Jackson Sargent",
                    "Amy Feng",
                    "Nicolas Garcia Trillos",
                    "Yu Tian",
                    "Zachary Lubberts"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Geometric Analysis",
                    "Machine Learning",
                    "Algorithmic Fairness"
                ],
                "publications": [
                    {
                        "title": "Neighborhood Growth Determines Geometric Priors for Relational Representation Learning",
                        "abstract": "The problem of identifying geometric structure in heterogeneous, high-dimensional data is a cornerstone of representation learning. While there exists a large body of literature on the embeddability of canonical graphs, such as lattices or trees, the heterogeneity of the relational data typically encountered in practice limits the applicability of these classical methods. In this paper, we propose a combinatorial approach to evaluating embeddability, i.e., to decide whether a data set is best represented in Euclidean, Hyperbolic or Spherical space. Our method analyzes nearest-neighbor structures and local neighborhood growth rates to identify the geometric priors of suitable embedding spaces. For canonical graphs, the algorithm's prediction provably matches classical results. As for large, heterogeneous graphs, we introduce an efficiently computable statistic that approximates the algorithm's decision rule. We validate our method over a range of benchmark data sets and compare with recently published optimization-based embeddability methods."
                    },
                    {
                        "title": "Identifying biases in legal data: An algorithmic fairness perspective",
                        "abstract": "The need to address representation biases and sentencing disparities in legal case data has long been recognized. Here, we study the problem of identifying and measuring biases in large-scale legal case data from an algorithmic fairness perspective. Our approach utilizes two regression models: A baseline that represents the decisions of a \"typical\" judge as given by the data and a \"fair\" judge that applies one of three fairness concepts. Comparing the decisions of the \"typical\" judge and the \"fair\" judge allows for quantifying biases across demographic groups, as we demonstrate in four case studies on criminal data from Cook County (Illinois)."
                    },
                    {
                        "title": "Computing Brascamp-Lieb Constants through the lens of Thompson Geometry",
                        "abstract": "This paper studies algorithms for efficiently computing Brascamp-Lieb constants, a task that has recently received much interest. In particular, we reduce the computation to a nonlinear matrix-valued iteration, whose convergence we analyze through the lens of fixed-point methods under the well-known Thompson metric. This approach permits us to obtain (weakly) polynomial time guarantees, and it offers an efficient and transparent alternative to previous state-of-the-art approaches based on Riemannian optimization and geodesic convexity."
                    },
                    {
                        "title": "Forman's Ricci curvature - From networks to hypernetworks",
                        "abstract": "Networks and their higher order generalizations, such as hypernetworks or multiplex networks are ever more popular models in the applied sciences. However, methods developed for the study of their structural properties go little beyond the common name and the heavy reliance of combinatorial tools. We show that, in fact, a geometric unifying approach is possible, by viewing them as polyhedral complexes endowed with a simple, yet, the powerful notion of curvature - the Forman Ricci curvature. We systematically explore some aspects related to the modeling of weighted and directed hypernetworks and present expressive and natural choices involved in their definitions. A benefit of this approach is a simple method of structure-preserving embedding of hypernetworks in Euclidean N-space. Furthermore, we introduce a simple and efficient manner of computing the well established Ollivier-Ricci curvature of a hypernetwork."
                    },
                    {
                        "title": "Projection-free nonconvex stochastic optimization on Riemannian manifolds",
                        "abstract": "We study stochastic projection-free methods for constrained optimization of smooth functions on Riemannian manifolds, i.e., with additional constraints beyond the parameter domain being a manifold. Specifically, we introduce stochastic Riemannian Frank-Wolfe methods for nonconvex and geodesically convex problems. We present algorithms for both purely stochastic optimization and finite-sum problems. For the latter, we develop variance-reduced methods, including a Riemannian adaptation of the recently proposed Spider technique. For all settings, we recover convergence rates that are comparable to the best-known rates for their Euclidean counterparts. Finally, we discuss applications to two classic tasks: The computation of the Karcher mean of positive definite matrices and Wasserstein barycenters for multivariate normal distributions. For both tasks, stochastic Fw methods yield state-of-the-art empirical performance."
                    },
                    {
                        "title": "On a class of geodesically convex optimization problems solved via Euclidean MM methods",
                        "abstract": "We study geodesically convex (g-convex) problems that can be written as a difference of Euclidean convex functions. This structure arises in several optimization problems in statistics and machine learning, e.g., for matrix scaling, M-estimators for covariances, and Brascamp-Lieb inequalities. Our work offers efficient algorithms that on the one hand exploit g-convexity to ensure global optimality along with guarantees on iteration complexity. On the other hand, the split structure permits us to develop Euclidean Majorization-Minorization algorithms that help us bypass the need to compute expensive Riemannian operations such as exponential maps and parallel transport. We illustrate our results by specializing them to a few concrete optimization problems that have been previously studied in the machine learning literature. Ultimately, we hope our work helps motivate the broader search for mixed Euclidean-Riemannian optimization algorithms"
                    },
                    {
                        "title": "Structured Regularization for Constrained Optimization on the SPD Manifold",
                        "abstract": "Matrix-valued optimization tasks, including those involving symmetric positive definite (SPD) matrices, arise in a wide range of applications in machine learning, data science and statistics. Classically, such problems are solved via constrained Euclidean optimization, where the domain is viewed as a Euclidean space and the structure of the matrices (e.g., positive definiteness) enters as constraints. More recently, geometric approaches that leverage parametrizations of the problem as unconstrained tasks on the corresponding matrix manifold have been proposed. While they exhibit algorithmic benefits in many settings, they cannot directly handle additional constraints, such as inequality or sparsity constraints. A remedy comes in the form of constrained Riemannian optimization methods, notably, Riemannian Frank-Wolfe and Projected Gradient Descent. However, both algorithms require potentially expensive subroutines that can introduce computational bottlenecks in practise. To mitigate these shortcomings, we introduce a class of structured regularizers, based on symmetric gauge functions, which allow for solving constrained optimization on the SPD manifold with faster unconstrained methods. We show that our structured regularizers can be chosen to preserve or induce desirable structure, in particular convexity and \"difference of convex\" structure. We demonstrate the effectiveness of our approach in numerical experiments."
                    },
                    {
                        "title": "Riemannian Optimization via Frank-Wolfe Methods",
                        "abstract": "We study projection-free methods for constrained Riemannian optimization. In particular, we propose the Riemannian Frank-Wolfe (RFW) method. We analyze non-asymptotic convergence rates of RFW to an optimum for (geodesically) convex problems, and to a critical point for nonconvex objectives. We also present a practical setting under which RFW can attain a linear convergence rate. As a concrete example, we specialize RFW to the manifold of positive definite matrices and apply it to two tasks: (i) computing the matrix geometric mean (Riemannian centroid); and (ii) computing the Bures-Wasserstein barycenter. Both tasks involve geodesically convex interval constraints, for which we show that the Riemannian \"linear\" oracle required by RFW admits a closed-form solution; this result may be of independent interest. We further specialize RFW to the special orthogonal group and show that here too, the Riemannian \"linear\" oracle can be solved in closed form. Here, we describe an application to the synchronization of data matrices (Procrustes problem). We complement our theoretical results with an empirical comparison of RFW against state-of-the-art Riemannian optimization methods and observe that RFW performs competitively on the task of computing Riemannian centroids."
                    },
                    {
                        "title": "Effective Structural Encodings via Local Curvature Profiles",
                        "abstract": "Structural and Positional Encodings can significantly improve the performance of Graph Neural Networks in downstream tasks. Recent literature has begun to systematically investigate differences in the structural properties that these approaches encode, as well as performance trade-offs between them. However, the question of which structural properties yield the most effective encoding remains open. In this paper, we investigate this question from a geometric perspective. We propose a novel structural encoding based on discrete Ricci curvature (Local Curvature Profiles, short LCP) and show that it significantly outperforms existing encoding approaches. We further show that combining local structural encodings, such as LCP, with global positional encodings improves downstream performance, suggesting that they capture complementary geometric information. Finally, we compare different encoding types with (curvature-based) rewiring techniques. Rewiring has recently received a surge of interest due to its ability to improve the performance of Graph Neural Networks by mitigating over-smoothing and over-squashing effects. Our results suggest that utilizing curvature information for structural encodings delivers significantly larger performance increases than rewiring."
                    },
                    {
                        "title": "Graph Pooling via Ricci Flow",
                        "abstract": "Graph Machine Learning often involves the clustering of nodes based on similarity structure encoded in the graph's topology and the nodes' attributes. On homophilous graphs, the integration of pooling layers has been shown to enhance the performance of Graph Neural Networks by accounting for inherent multi-scale structure. Here, similar nodes are grouped together to coarsen the graph and reduce the input size in subsequent layers in deeper architectures. In both settings, the underlying clustering approach can be implemented via graph pooling operators, which often rely on classical tools from Graph Theory. In this work, we introduce a graph pooling operator (ORC-Pool), which utilizes a characterization of the graph's geometry via Ollivier's discrete Ricci curvature and an associated geometric flow. Previous Ricci flow based clustering approaches have shown great promise across several domains, but are by construction unable to account for similarity structure encoded in the node attributes. However, in many ML applications, such information is vital for downstream tasks. ORC-Pool extends such clustering approaches to attributed graphs, allowing for the integration of geometric coarsening into Graph Neural Networks as a pooling layer."
                    },
                    {
                        "title": "Mitigating Over-Smoothing and Over-Squashing using Augmentations of Forman-Ricci Curvature",
                        "abstract": "While Graph Neural Networks (GNNs) have been successfully leveraged for learning on graph-structured data across domains, several potential pitfalls have been described recently. Those include the inability to accurately leverage information encoded in long-range connections (over-squashing), as well as difficulties distinguishing the learned representations of nearby nodes with growing network depth (over-smoothing). An effective way to characterize both effects is discrete curvature: Long-range connections that underlie over-squashing effects have low curvature, whereas edges that contribute to over-smoothing have high curvature. This observation has given rise to rewiring techniques, which add or remove edges to mitigate over-smoothing and over-squashing. Several rewiring approaches utilizing graph characteristics, such as curvature or the spectrum of the graph Laplacian, have been proposed. However, existing methods, especially those based on curvature, often require expensive subroutines and careful hyperparameter tuning, which limits their applicability to large-scale graphs. Here we propose a rewiring technique based on Augmented Forman-Ricci curvature (AFRC), a scalable curvature notation, which can be computed in linear time. We prove that AFRC effectively characterizes over-smoothing and over-squashing effects in message-passing GNNs. We complement our theoretical results with experiments, which demonstrate that the proposed approach achieves state-of-the-art performance while significantly reducing the computational cost in comparison with other methods. Utilizing fundamental properties of discrete curvature, we propose effective heuristics for hyperparameters in curvature-based rewiring, which avoids expensive hyperparameter searches, further improving the scalability of the proposed approach."
                    },
                    {
                        "title": "Characterizing Complex Networks with Forman-Ricci Curvature and Associated Geometric Flows",
                        "abstract": "We introduce Forman-Ricci curvature and its corresponding flow as characteristics for complex networks attempting to extend the common approach of node-based network analysis by edge-based characteristics. Following a theoretical introduction and mathematical motivation, we apply the proposed network-analytic methods to static and dynamic complex networks and compare the results with established node-based characteristics. Our work suggests a number of applications for data mining, including denoising and clustering of experimental data, as well as extrapolation of network evolution."
                    },
                    {
                        "title": "Forman-Ricci flow for change detection in large dynamic data sets",
                        "abstract": "We present a viable solution to the challenging question of change detection in complex networks inferred from large dynamic data sets. Building on Forman's discretization of the classical notion of Ricci curvature, we introduce a novel geometric method to characterize different types of real-world networks with an emphasis on peer-to-peer networks. Furthermore we adapt the classical Ricci flow that already proved to be a powerful tool in image processing and graphics, to the case of undirected and weighted networks. The application of the proposed method on peer-to-peer networks yields insights into topological properties and the structure of their underlying data."
                    },
                    {
                        "title": "Can one see the shape of a network?",
                        "abstract": "Traditionally, network analysis is based on local properties of vertices, like their degree or clustering, and their statistical behavior across the network in question. This paper develops an approach which is different in two respects. We investigate edge-based properties, and we define global characteristics of networks directly. The latter will provide our affirmative answer to the question raised in the title. More concretely, we start with Forman's notion of the Ricci curvature of a graph, or more generally, a polyhedral complex. This will allow us to pass from a graph as representing a network to a polyhedral complex for instance by filling in triangles into connected triples of edges and to investigate the resulting effect on the curvature. This is insightful for two reasons: First, we can define a curvature flow in order to asymptotically simplify a network and reduce it to its essentials. Second, using a construction of Bloch, which yields a discrete Gauss-Bonnet theorem, we have the Euler characteristic of a network as a global characteristic. These two aspects beautifully merge in the sense that the asymptotic properties of the curvature flow are indicated by that Euler characteristic."
                    },
                    {
                        "title": "Continuum Limits of Ollivier's Ricci Curvature on data clouds: pointwise consistency and global lower bounds",
                        "abstract": "Let $M$ denote a low-dimensional manifold embedded in Euclidean space and let ${X}= \\{ x_1, \\dots, x_n \\}$ be a collection of points uniformly sampled from it. We study the relationship between the curvature of a random geometric graph built from ${X}$ and the curvature of the manifold $M$ via continuum limits of Ollivier's discrete Ricci curvature. We prove pointwise, non-asymptotic consistency results and also show that if $M$ has Ricci curvature bounded from below by a positive constant, then the random geometric graph will inherit this global structural property with high probability. We discuss applications of the global discrete curvature bounds to contraction properties of heat kernels on graphs, as well as implications for manifold learning from data clouds. In particular, we show that our consistency results allow for estimating the intrinsic curvature of a manifold by first estimating concrete extrinsic quantities."
                    },
                    {
                        "title": "Curvature-based Clustering on Graphs",
                        "abstract": "Unsupervised node clustering (or community detection) is a classical graph learning task. In this paper, we study algorithms, which exploit the geometry of the graph to identify densely connected substructures, which form clusters or communities. Our method implements discrete Ricci curvatures and their associated geometric flows, under which the edge weights of the graph evolve to reveal its community structure. We consider several discrete curvature notions and analyze the utility of the resulting algorithms. In contrast to prior literature, we study not only single-membership community detection, where each node belongs to exactly one community, but also mixed-membership community detection, where communities may overlap. For the latter, we argue that it is beneficial to perform community detection on the line graph, i.e., the graph's dual. We provide both theoretical and empirical evidence for the utility of our curvature-based clustering algorithms. In addition, we give several results on the relationship between the curvature of a graph and that of its dual, which enable the efficient implementation of our proposed mixed-membership community detection approach and which may be of independent interest for curvature-based network analysis."
                    },
                    {
                        "title": "Disciplined Geodesically Convex Programming",
                        "abstract": "Convex programming plays a fundamental role in machine learning, data science, and engineering. Testing convexity structure in nonlinear programs relies on verifying the convexity of objectives and constraints. \\citet{grant2006disciplined} introduced a framework, Disciplined Convex Programming (DCP), for automating this verification task for a wide range of convex functions that can be decomposed into basic convex functions (atoms) using convexity-preserving compositions and transformations (rules). However, the restriction to Euclidean convexity concepts can limit the applicability of the framework. For instance, many notable instances of statistical estimators and matrix-valued (sub)routines in machine learning applications are Euclidean non-convex, but exhibit geodesic convexity through a more general Riemannian lens. In this work, we extend disciplined programming to this setting by introducing Disciplined Geodesically Convex Programming (DGCP). We determine convexity-preserving compositions and transformations for geodesically convex functions on general Cartan-Hadamard manifolds, as well as for the special case of symmetric positive definite matrices, a common setting in matrix-valued optimization. For the latter, we also define a basic set of atoms. Our paper is accompanied by a Julia package SymbolicAnalysis.jl, which provides functionality for testing and certifying DGCP-compliant expressions. Our library interfaces with manifold optimization software, which allows for directly solving verified geodesically convex programs."
                    },
                    {
                        "title": "Unitary convolutions for learning on graphs and groups",
                        "abstract": "Data with geometric structure is ubiquitous in machine learning often arising from fundamental symmetries in a domain, such as permutation-invariance in graphs and translation-invariance in images. Group-convolutional architectures, which encode symmetries as inductive bias, have shown great success in applications, but can suffer from instabilities as their depth increases and often struggle to learn long range dependencies in data. For instance, graph neural networks experience instability due to the convergence of node representations (over-smoothing), which can occur after only a few iterations of message-passing, reducing their effectiveness in downstream tasks. Here, we propose and study unitary group convolutions, which allow for deeper networks that are more stable during training. The main focus of the paper are graph neural networks, where we show that unitary graph convolutions provably avoid over-smoothing. Our experimental results confirm that unitary graph convolutional networks achieve competitive performance on benchmark datasets compared to state-of-the-art graph neural networks. We complement our analysis of the graph domain with the study of general unitary convolutions and analyze their role in enhancing stability in general group convolutional architectures."
                    },
                    {
                        "title": "Associative Memory Impairments arising from Neurodegenerative Diseases and Traumatic Brain Injuries in a Hopfield Network Model",
                        "abstract": "Neurodegenerative diseases and traumatic brain injuries (TBI) are among the main causes of cognitive dysfunction in humans. Both manifestations exhibit the extensive presence of focal axonal swellings (FAS). FAS compromises the information encoded in spike trains, thus leading to potentially severe functional deficits. Complicating our understanding of the impact of FAS is our inability to access small scale injuries with non-invasive methods, the overall complexity of neuronal pathologies, and our limited knowledge of how networks process biological signals. Building on Hopfield's pioneering work, we extend a model for associative memory to account for FAS and its impact on memory encoding. We calibrate all FAS parameters from biophysical observations of their statistical distribution and size, providing a framework to simulate the effects of brain disorders on memory recall performance. A face recognition example is used to demonstrate and validate the functionality of the novel model. Our results link memory recall ability to observed FAS statistics, allowing for a description of different stages of brain disorders within neuronal networks. This provides a first theoretical model to bridge experimental observations of FAS in neurodegeneration and TBI with compromised memory recall, thus closing the large gap between theory and experiment on how biological signals are processed in damaged, high-dimensional functional networks. The work further lends new insight into positing diagnostic tools to measure cognitive deficits."
                    }
                ]
            }
        }
    },
    "2405.18075": {
        "paper_data": {
            "title": "Implicitly Guided Design with PropEn: Match your Data to Follow the Gradient",
            "url": "http://arxiv.org/abs/2405.18075v1",
            "arxiv_id": "2405.18075",
            "authors": [
                "Nata\u0161a Tagasovska",
                "Vladimir Gligorijevi\u0107",
                "Kyunghyun Cho",
                "Andreas Loukas"
            ],
            "abstract": "Across scientific domains, generating new models or optimizing existing ones while meeting specific criteria is crucial. Traditional machine learning frameworks for guided design use a generative model and a surrogate model (discriminator), requiring large datasets. However, real-world scientific applications often have limited data and complex landscapes, making data-hungry models inefficient or impractical. We propose a new framework, PropEn, inspired by ``matching'', which enables implicit guidance without training a discriminator. By matching each sample with a similar one that has a better property value, we create a larger training dataset that inherently indicates the direction of improvement. Matching, combined with an encoder-decoder architecture, forms a domain-agnostic generative framework for property enhancement. We show that training with a matched dataset approximates the gradient of the property of interest while remaining within the data distribution, allowing efficient design optimization. Extensive evaluations in toy problems and scientific applications, such as therapeutic protein design and airfoil optimization, demonstrate PropEn's advantages over common baselines. Notably, the protein design results are validated with wet lab experiments, confirming the competitiveness and effectiveness of our approach.",
            "introduction": "   1 Introduction  Navigating the complex world of design is a challenge in many fields, from engineering [1] to material science [2, 3] and life sciences [4]. In life sciences, the goal may be to refine molecular structures for drug discovery [4], focusing on properties like binding affinity or stability. In engineering, optimizing the shapes of aircraft wings or windmill blades to achieve desired aerodynamic traits like lift and drag forces is crucial [1]. The common thread in these fields is the design cycle: experts start with an initial design and aim to improve a specific property. Guided by intuition and expertise, they make adjustments and evaluate the changes. If the property improves, they continue this iterative optimization process. This cycle is repeated multiple times, making it time-consuming and resource-intensive. Machine learning (ML) holds a promise to reduce these costs, speed up design cycles, and create better-performing designs [5, 6, 7, 8].   Yet, progress in ML methods for design is hindered by practical challenges. The first challenge is limited data availability. Since gathering label measurements is resource intensive [9, 10, 11, 12], designers are more often than not constrained to very small-scale datasets. In addition, there are often non-smooth functional dependencies between the features and outcome, complicating approximation, even with for deep neural networks [13, 14]. Traditional methods use two part frameworks, requiring discriminators to guide the property enhancement for examples produced by a generative model. Such discriminators should be able to reliably predict the property of interest given some training data or its latent representations. Because of the dependency on training a discriminator for guidance, we denote these methods as explicit guidance. While Genetic Algorithms were once prevalent [15], contemporary models like auto-encoders [16], GANs [17], and diffusion models now dominate both research and practice [6] the role of generative models. Despite their flexibility, such models face challenges typical to deep learning: they are \u201cdata-hungry\u201d and unreliable when encountering out-of-distribution examples [18, 19, 20, 21].   Figure 1: Conceptual summary of implicit and explicit guidance. The task is increasing the size of the objects. Top - in implicit guidance, first we match the training dataset, by pairing each sample with the closest one w.r.t. its shape, which has a better property (size). Then, we train a encoder-decoder framework which due to the construction of the dataset learns a lower dimensional manifold where the embeddings are ordered by the property value. Bottom - in explicit guidance, we train two separate models: a generator and a discriminator that guides the optimization in latent space.   Motivated by these challenges, we propose a new approach inspired by the concept of \u201cmatching\u201d. Matching techniques in econometrics are used to address the challenge of selection bias and confounding when estimating causal effects in observational studies [22, 23, 24, 25]. These techniques aim to create comparable groups of units by matching treated and control observations based on observable characteristics. The basic idea behind matching is to identify untreated units that are similar to treated units in terms of observed covariates, effectively creating a counterfactual comparison group for each treated unit. Matching techniques in ML as in econometrics have",
            "references": [
                {
                    "title": "Improving antibody affinity using laboratory data with language model guided design",
                    "abstract": "Protein design involves navigating vast sequence spaces to discover sequences with desired traits. Language models (LMs) pretrained on universal protein datasets have shown potential to make this search space tractable. However, LMs trained solely on natural sequences have limitations in creating proteins with novel functions. In this work, we used a combination of methods to finetune pretrained LMs on laboratory data collected in an anti-CD40L single domain antibody library campaign to develop an ensemble scoring function to model the fitness landscape and guide the design of new antibodies. Laboratory experiments confirmed improved CD40L affinity in the designed antibodies. Notably, the designs improved the affinities of four antibodies, originally ranging from 1 nanomolar to 100 picomolar, all to below 25 picomolar, approaching the limit of detection. This work is a promising step towards realizing the potential of LMs to leverage laboratory data to develop improved treatments for diseases."
                },
                {
                    "title": "Cooperation of Thin-Airfoil Theory and Deep Learning for a Compact Airfoil Shape Parameterization",
                    "abstract": "An airfoil shape parameterization that can generate a compact design space is highly desirable in practice. In this paper, a compact airfoil parameterization is proposed by incorporating deep learning into the PAERO parameterization method based on the thin-airfoil theory. Following the PAERO parameterization, the mean camber line is represented by a number of aerodynamic performance parameters, which can be used to narrow down the design space according to the thin-airfoil theory. In order to further reduce the design space, the airfoil thickness distribution is represented by data-driven generative models, which are trained by the thickness distributions of existing airfoils. The trained models can automatically filter out the physically unreasonable airfoil shapes, resulting in a highly compact design space. The test results show that the proposed method is significantly more efficient and more robust than the widely used CST parameterization method for airfoil optimization."
                },
                {
                    "title": "Protein Discovery with Discrete Walk-Jump Sampling",
                    "abstract": "We resolve difficulties in training and sampling from a discrete generative model by learning a smoothed energy function, sampling from the smoothed data manifold with Langevin Markov chain Monte Carlo (MCMC), and projecting back to the true data manifold with one-step denoising. Our Discrete Walk-Jump Sampling formalism combines the contrastive divergence training of an energy-based model and improved sample quality of a score-based model, while simplifying training and sampling by requiring only a single noise level. We evaluate the robustness of our approach on generative modeling of antibody proteins and introduce the distributional conformity score to benchmark protein generative models. By optimizing and sampling from our models for the proposed distributional conformity score, 97-100% of generated samples are successfully expressed and purified and 70% of functional designs show equal or improved binding affinity compared to known functional antibodies on the first attempt in a single round of laboratory experiments. We also report the first demonstration of long-run fast-mixing MCMC chains where diverse antibody protein classes are visited in a single MCMC chain."
                },
                {
                    "title": "Protein Design with Guided Discrete Diffusion",
                    "abstract": "A popular approach to protein design is to combine a generative model with a discriminative model for conditional sampling. The generative model samples plausible sequences while the discriminative model guides a search for sequences with high fitness. Given its broad success in conditional sampling, classifier-guided diffusion modeling is a promising foundation for protein design, leading many to develop guided diffusion models for structure with inverse folding to recover sequences. In this work, we propose diffusioN Optimized Sampling (NOS), a guidance method for discrete diffusion models that follows gradients in the hidden states of the denoising network. NOS makes it possible to perform design directly in sequence space, circumventing significant limitations of structure-based methods, including scarce data and challenging inverse design. Moreover, we use NOS to generalize LaMBO, a Bayesian optimization procedure for sequence design that facilitates multiple objectives and edit-based constraints. The resulting method, LaMBO-2, enables discrete diffusions and stronger performance with limited edits through a novel application of saliency maps. We apply LaMBO-2 to a real-world protein design task, optimizing antibodies for higher expression yield and binding affinity to several therapeutic targets under locality and developability constraints, attaining a 99% expression rate and 40% binding rate in exploratory in vitro experiments."
                },
                {
                    "title": "A survey of machine learning in additive manufacturing technologies",
                    "abstract": "ABSTRACT Thirty years into its development, additive manufacturing has become a mainstream manufacturing process. Additive manufacturing fabricates products by adding materials layer-by-layer directly based on a 3D model. It is able to manufacture complex parts and allows more freedom of design optimization compared with traditional manufacturing techniques. Machine learning is now a hot technology that has been used in medical diagnosis, image processing, prediction, classification, learning association, regression, etc. Currently, focuses are increasingly given to using machine learning in the manufacturing industry, including additive manufacturing. Due to the rapid development of machine learning in additive manufacturing, a special issue \u2018Machine Learning in Additive Manufacturing\u2019 in International Journal of Computer Integrated Manufacturing is organized. This paper gives a comprehensive understanding of the current status of machine learning enhanced additive manufacturing technologies for this special issue. Discussions and future perspectives are also provided."
                },
                {
                    "title": "Exposing the Limitations of Molecular Machine Learning with Activity Cliffs",
                    "abstract": "Machine learning has become a crucial tool in drug discovery and chemistry at large, e.g., to predict molecular properties, such as bioactivity, with high accuracy. However, activity cliffs\u2014pairs of molecules that are highly similar in their structure but exhibit large differences in potency\u2014have received limited attention for their effect on model performance. Not only are these edge cases informative for molecule discovery and optimization but also models that are well equipped to accurately predict the potency of activity cliffs have increased potential for prospective applications. Our work aims to fill the current knowledge gap on best-practice machine learning methods in the presence of activity cliffs. We benchmarked a total of 24 machine and deep learning approaches on curated bioactivity data from 30 macromolecular targets for their performance on activity cliff compounds. While all methods struggled in the presence of activity cliffs, machine learning approaches based on molecular descriptors outperformed more complex deep learning methods. Our findings highlight large case-by-case differences in performance, advocating for (a) the inclusion of dedicated \u201cactivity-cliff-centered\u201d metrics during model development and evaluation and (b) the development of novel algorithms to better predict the properties of activity cliffs. To this end, the methods, metrics, and results of this study have been encapsulated into an open-access benchmarking platform named MoleculeACE (Activity Cliff Estimation, available on GitHub at: https://github.com/molML/MoleculeACE). MoleculeACE is designed to steer the community toward addressing the pressing but overlooked limitation of molecular machine learning models posed by activity cliffs."
                },
                {
                    "title": "Machine learning optimization of candidate antibody yields highly diverse sub-nanomolar affinity antibody libraries",
                    "abstract": "Therapeutic antibodies are an important and rapidly growing drug modality. However, the design and discovery of early-stage antibody therapeutics remain a time and cost-intensive endeavor. In this work, we present an end-to-end Bayesian, language model-based method for designing large and diverse libraries of high-affinity single-chain variable fragments (scFvs). We integrate target-specific binding affinities with information from millions of natural protein sequences in a probabilistic machine learning framework to design thousands of scFvs that are then empirically measured. In a head-to-head comparison with a directed evolution approach, we show that the best scFv generated from our method represents a 28.8-fold improvement in binding over the best scFv from the directed evolution. Additionally, 99% of the designed scFvs in our most successful library are improvements over the initial candidate scFv. By comparing a library\u2019s predicted success to actual measurements, we demonstrate our method\u2019s ability to explore tradeoffs between library success and diversity during the design phase and prior to experimental testing. The results of our work highlight the significant impact machine learning models can have on scFv development. We expect our end-to-end method to be broadly applicable and able to provide value to other protein engineering tasks."
                },
                {
                    "title": "Aeroacoustic airfoil shape optimization enhanced by autoencoders",
                    "abstract": "We present a framework for airfoil shape optimization to reduce the trailing edge noise for the design of wind turbine blades. Far-field noise is evaluated using Amiet's theory coupled with the TNO-Blake model to calculate the wall pressure spectrum and fast turn-around XFOIL simulations to evaluate the boundary layer parameters. The computational framework is first validated using a NACA0012 airfoil at zero angle of attack. Particle swarm optimization is used to find the optimized airfoil configuration. The multi-objective optimization minimizes the A-weighted overall sound pressure level at various angles of attack, while ensuring enough lift and minimum drag. We compare classic parametrization methods to define the airfoil geometry (i.e., CST) to a machine learning method (i.e., a variational autoencoder). We observe that variational autoencoders can represent a wide variety of geometries, with only four encoded variables, leading to efficient optimizations, which result in improved optimal shapes. When compared to the baseline geometry, a NACA0012, the autoencoder-based optimized airfoil reduces by 3% (1.75 dBA) the overall sound pressure level (with decreased noise across the entire frequency range), while maintaining favorable aerodynamic properties in terms of lift and drag."
                },
                {
                    "title": "Model-Based Deep Learning: On the Intersection of Deep Learning and Optimization",
                    "abstract": "Decision making algorithms are used in a multitude of different applications. Conventional approaches for designing decision algorithms employ principled and simplified modelling, based on which one can determine decisions via tractable optimization. More recently, deep learning approaches that use highly parametric architectures tuned from data without relying on mathematical models, are becoming increasingly popular. Model-based optimization and data-centric deep learning are often considered to be distinct disciplines. Here, we characterize them as edges of a continuous spectrum varying in specificity and parameterization, and provide a tutorial-style presentation to the methodologies lying in the middle ground of this spectrum, referred to as model-based deep learning. We accompany our presentation with running examples in super-resolution and stochastic control, and show how they are expressed using the provided characterization and specialized in each of the detailed methodologies. The gains of combining model-based optimization and deep learning are demonstrated using experimental results in various applications, ranging from biomedical imaging to digital communications."
                },
                {
                    "title": "Neural Score Matching for High-Dimensional Causal Inference",
                    "abstract": "Traditional methods for matching in causal inference are impractical for high-dimensional datasets. They suffer from the curse of dimensionality: exact matching and coarsened exact matching find exponentially fewer matches as the input dimension grows, and propensity score matching may match highly unrelated units together. To overcome this problem, we develop theoretical results which motivate the use of neural networks to obtain non-trivial, multivariate balancing scores of a chosen level of coarseness, in contrast to the classical, scalar propensity score. We leverage these balancing scores to perform matching for high-dimensional causal inference and call this procedure neural score matching. We show that our method is competitive against other matching approaches on semi-synthetic high-dimensional datasets, both in terms of treatment effect estimation and reducing imbalance."
                },
                {
                    "title": "Implications of Additivity and Nonadditivity for Machine Learning and Deep Learning Models in Drug Design",
                    "abstract": "Matched molecular pairs (MMPs) are nowadays a commonly applied concept in drug design. They are used in many computational tools for structure\u2013activity relationship analysis, biological activity prediction, or optimization of physicochemical properties. However, until now it has not been shown in a rigorous way that MMPs, that is, changing only one substituent between two molecules, can be predicted with higher accuracy and precision in contrast to any other chemical compound pair. It is expected that any model should be able to predict such a defined change with high accuracy and reasonable precision. In this study, we examine the predictability of four classical properties relevant for drug design ranging from simple physicochemical parameters (log D and solubility) to more complex cell-based ones (permeability and clearance), using different data sets and machine learning algorithms. Our study confirms that additive data are the easiest to predict, which highlights the importance of recognition of nonadditivity events and the challenging complexity of predicting properties in case of scaffold hopping. Despite deep learning being well suited to model nonlinear events, these methods do not seem to be an exception of this observation. Though they are in general performing better than classical machine learning methods, this leaves the field with a still standing challenge."
                },
                {
                    "title": "ILVR: Conditioning Method for Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) have shown remarkable performance in unconditional image generation. However, due to the stochasticity of the generative process in DDPM, it is challenging to generate images with the desired semantics. In this work, we propose Iterative Latent Variable Refinement (ILVR), a method to guide the generative process in DDPM to generate high-quality images based on a given reference image. Here, the refinement of the generative process in DDPM enables a single DDPM to sample images from various sets directed by the reference image. The proposed ILVR method generates high-quality images while controlling the generation. The controllability of our method allows adaptation of a single DDPM without any additional learning in various image generation tasks, such as generation from various downsampling factors, multi-domain image translation, paint-to-image, and editing with scribbles."
                },
                {
                    "title": "Variational Diffusion Models",
                    "abstract": "Diffusion-based generative models have demonstrated a capacity for perceptually impressive synthesis, but can they also be great likelihood-based models? We answer this in the affirmative, and introduce a family of diffusion-based generative models that obtain state-of-the-art likelihoods on standard image density estimation benchmarks. Unlike other diffusion-based models, our method allows for efficient optimization of the noise schedule jointly with the rest of the model. We show that the variational lower bound (VLB) simplifies to a remarkably short expression in terms of the signal-to-noise ratio of the diffused data, thereby improving our theoretical understanding of this model class. Using this insight, we prove an equivalence between several models proposed in the literature. In addition, we show that the continuous-time VLB is invariant to the noise schedule, except for the signal-to-noise ratio at its endpoints. This enables us to learn a noise schedule that minimizes the variance of the resulting VLB estimator, leading to faster optimization. Combining these advances with architectural improvements, we obtain state-of-the-art likelihoods on image density estimation benchmarks, outperforming autoregressive models that have dominated these benchmarks for many years, with often significantly faster optimization. In addition, we show how to use the model as part of a bits-back compression scheme, and demonstrate lossless compression rates close to the theoretical optimum. Code is available at https://github.com/google-research/vdm ."
                },
                {
                    "title": "Improving black-box optimization in VAE latent space using decoder uncertainty",
                    "abstract": "Optimization in the latent space of variational autoencoders is a promising approach to generate high-dimensional discrete objects that maximize an expensive black-box property (e.g., drug-likeness in molecular generation, function approximation with arithmetic expressions). However, existing methods lack robustness as they may decide to explore areas of the latent space for which no data was available during training and where the decoder can be unreliable, leading to the generation of unrealistic or invalid objects. We propose to leverage the epistemic uncertainty of the decoder to guide the optimization process. This is not trivial though, as a naive estimation of uncertainty in the high-dimensional and structured settings we consider would result in high estimator variance. To solve this problem, we introduce an importance sampling-based estimator that provides more robust estimates of epistemic uncertainty. Our uncertainty-guided optimization approach does not require modifications of the model architecture nor the training process. It produces samples with a better trade-off between black-box objective and validity of the generated samples, sometimes improving both simultaneously. We illustrate these advantages across several experimental settings in digit generation, arithmetic expression approximation and molecule generation for drug design."
                },
                {
                    "title": "Generating various airfoil shapes with required lift coefficient using conditional variational autoencoders",
                    "abstract": "Multiple shapes must be obtained in the mechanical design process to satisfy the required design specifications. The inverse design problem has been analyzed in previous studies to obtain such shapes. However, finding multiple shapes in a short computation period is difficult while using the conventional methods. This paper proposes the use of the conditional variational autoencoders (CVAE) with normal distribution, denoted by N-CVAE, along with the von Mises-Fischer distribution, denoted by S-CVAE, to find multiple solutions for the inverse design problems. Both the CVAE models embed shapes into a latent space. The S-CVAE enables the separation of data in the latent space, whereas the N-CVAE embeds the data in a narrow space. These different features are used for various tasks in this study. In one of the tasks, the dataset consists of only one type of data and generates similar airfoils. Here, S-CVAE outperforms N-CVAE because it can separate the data. Another task involves combining different types of airfoils and generating new types of data. N-CVAE is useful in this instance since it embeds different shapes in the same latent area, due to which, the model outputs intermediate shapes of different types. The shape-generation capability of S-CVAE and N-CVAE are experimentally compared in this study."
                },
                {
                    "title": "Algorithm Unrolling: Interpretable, Efficient Deep Learning for Signal and Image Processing",
                    "abstract": "Deep neural networks provide unprecedented performance gains in many real-world problems in signal and image processing. Despite these gains, the future development and practical deployment of deep networks are hindered by their black-box nature, i.e., a lack of interpretability and the need for very large training sets. An emerging technique called algorithm unrolling, or unfolding, offers promise in eliminating these issues by providing a concrete and systematic connection between iterative algorithms that are widely used in signal processing and deep neural networks. Unrolling methods were first proposed to develop fast neural network approximations for sparse coding. More recently, this direction has attracted enormous attention, and it is rapidly growing in both theoretic investigations and practical applications. The increasing popularity of unrolled deep networks is due, in part, to their potential in developing efficient, high-performance (yet interpretable) network architectures from reasonably sized training sets."
                },
                {
                    "title": "Learning Convex Optimization Control Policies",
                    "abstract": "Many control policies used in various applications determine the input or action by solving a convex optimization problem that depends on the current state and some parameters. Common examples of such convex optimization control policies (COCPs) include the linear quadratic regulator (LQR), convex model predictive control (MPC), and convex control-Lyapunov or approximate dynamic programming (ADP) policies. These types of control policies are tuned by varying the parameters in the optimization problem, such as the LQR weights, to obtain good performance, judged by application-specific metrics. Tuning is often done by hand, or by simple methods such as a crude grid search. In this paper we propose a method to automate this process, by adjusting the parameters using an approximate gradient of the performance metric with respect to the parameters. Our method relies on recently developed methods that can efficiently evaluate the derivative of the solution of a convex optimization problem with respect to its parameters. We illustrate our method on several examples."
                },
                {
                    "title": "Constrained Bayesian optimization for automatic chemical design using variational autoencoders",
                    "abstract": "Automatic Chemical Design is a framework for generating novel molecules with optimized properties."
                },
                {
                    "title": "Black Box Recursive Translations for Molecular Optimization",
                    "abstract": "Machine learning algorithms for generating molecular structures offer a promising new approach to drug discovery. We cast molecular optimization as a translation problem, where the goal is to map an input compound to a target compound with improved biochemical properties. Remarkably, we observe that when generated molecules are iteratively fed back into the translator, molecular compound attributes improve with each step. We show that this finding is invariant to the choice of translation model, making this a \"black box\" algorithm. We call this method Black Box Recursive Translation (BBRT), a new inference method for molecular property optimization. This simple, powerful technique operates strictly on the inputs and outputs of any translation model. We obtain new state-of-the-art results for molecular property optimization tasks using our simple drop-in replacement with well-known sequence and graph-based models. Our method provides a significant boost in performance relative to its non-recursive peers with just a simple \"for\" loop. Further, BBRT is highly interpretable, allowing users to map the evolution of newly discovered compounds from known starting points."
                },
                {
                    "title": "A Model to Search for Synthesizable Molecules",
                    "abstract": "Deep generative models are able to suggest new organic molecules by generating strings, trees, and graphs representing their structure. While such models allow one to generate molecules with desirable properties, they give no guarantees that the molecules can actually be synthesized in practice. We propose a new molecule generation model, mirroring a more realistic real-world process, where (a) reactants are selected, and (b) combined to form more complex molecules. More specifically, our generative model proposes a bag of initial reactants (selected from a pool of commercially-available molecules) and uses a reaction model to predict how they react together to generate new molecules. We first show that the model can generate diverse, valid and unique molecules due to the useful inductive biases of modeling reactions. Furthermore, our model allows chemists to interrogate not only the properties of the generated molecules but also the feasibility of the synthesis routes. We conclude by using our model to solve retrosynthesis problems, predicting a set of reactants that can produce a target product."
                },
                {
                    "title": "Conditioning by adaptive sampling for robust design",
                    "abstract": "We present a new method for design problems wherein the goal is to maximize or specify the value of one or more properties of interest. For example, in protein design, one may wish to find the protein sequence that maximizes fluorescence. We assume access to one or more, potentially black box, stochastic \"oracle\" predictive functions, each of which maps from input (e.g., protein sequences) design space to a distribution over a property of interest (e.g. protein fluorescence). At first glance, this problem can be framed as one of optimizing the oracle(s) with respect to the input. However, many state-of-the-art predictive models, such as neural networks, are known to suffer from pathologies, especially for data far from the training distribution. Thus we need to modulate the optimization of the oracle inputs with prior knowledge about what makes `realistic' inputs (e.g., proteins that stably fold). Herein, we propose a new method to solve this problem, Conditioning by Adaptive Sampling, which yields state-of-the-art results on a protein fluorescence problem, as compared to other recently published approaches. Formally, our method achieves its success by using model-based adaptive sampling to estimate the conditional distribution of the input sequences given the desired properties."
                },
                {
                    "title": "Learning Multimodal Graph-to-Graph Translation for Molecular Optimization",
                    "abstract": "We view molecular optimization as a graph-to-graph translation problem. The goal is to learn to map from one molecular graph to another with better properties based on an available corpus of paired molecules. Since molecules can be optimized in different ways, there are multiple viable translations for each input graph. A key challenge is therefore to model diverse translation outputs. Our primary contributions include a junction tree encoder-decoder for learning diverse graph translations along with a novel adversarial training method for aligning distributions of molecules. Diverse output distributions in our model are explicitly realized by low-dimensional latent vectors that modulate the translation process. We evaluate our model on multiple molecular optimization tasks and show that our model outperforms previous state-of-the-art baselines."
                },
                {
                    "title": "DeepMatch: Balancing Deep Covariate Representations for Causal Inference Using Adversarial Training",
                    "abstract": "We study optimal covariate balance for causal inferences from observational data when rich covariates and complex relationships necessitate flexible modeling with neural networks. Standard approaches such as propensity weighting and matching/balancing fail in such settings due to miscalibrated propensity nets and inappropriate covariate representations, respectively. We propose a new method based on adversarial training of a weighting and a discriminator network that effectively addresses this methodological gap. This is demonstrated through new theoretical characterizations of the method as well as empirical results using both fully connected architectures to learn complex relationships and convolutional architectures to handle image confounders, showing how this new method can enable strong causal analyses in these challenging settings."
                },
                {
                    "title": "Actively Learning what makes a Discrete Sequence Valid",
                    "abstract": "Deep learning techniques have been hugely successful for traditional supervised and unsupervised machine learning problems. In large part, these techniques solve continuous optimization problems. Recently however, discrete generative deep learning models have been successfully used to efficiently search high-dimensional discrete spaces. These methods work by representing discrete objects as sequences, for which powerful sequence-based deep models can be employed. Unfortunately, these techniques are significantly hindered by the fact that these generative models often produce invalid sequences. As a step towards solving this problem, we propose to learn a deep recurrent validator model. Given a partial sequence, our model learns the probability of that sequence occurring as the beginning of a full valid sequence. Thus this identifies valid versus invalid sequences and crucially it also provides insight about how individual sequence elements influence the validity of discrete objects. To learn this model we propose an approach inspired by seminal work in Bayesian active learning. On a synthetic dataset, we demonstrate the ability of our model to distinguish valid and invalid sequences. We believe this is a key step toward learning generative models that faithfully produce valid discrete objects."
                },
                {
                    "title": "FLAME: A Fast Large-scale Almost Matching Exactly Approach to Causal Inference",
                    "abstract": "A classical problem in causal inference is that of matching, where treatment units need to be matched to control units based on covariate information. In this work, we propose a method that computes high quality almost-exact matches for high-dimensional categorical datasets. This method, called FLAME (Fast Large-scale Almost Matching Exactly), learns a distance metric for matching using a hold-out training data set. In order to perform matching efficiently for large datasets, FLAME leverages techniques that are natural for query processing in the area of database management, and two implementations of FLAME are provided: the first uses SQL queries and the second uses bit-vector techniques. The algorithm starts by constructing matches of the highest quality (exact matches on all covariates), and successively eliminates variables in order to match exactly on as many variables as possible, while still maintaining interpretable high-quality matches and balance between treatment and control groups. We leverage these high quality matches to estimate conditional average treatment effects (CATEs). Our experiments show that FLAME scales to huge datasets with millions of observations where existing state-of-the-art methods fail, and that it achieves significantly better performance than other matching methods."
                },
                {
                    "title": "Grammar Variational Autoencoder",
                    "abstract": "Deep generative models have been wildly successful at learning coherent latent representations for continuous data such as video and audio. However, generative modeling of discrete data such as arithmetic expressions and molecular structures still poses significant challenges. Crucially, state-of-the-art methods often produce outputs that are not valid. We make the key observation that frequently, discrete data can be represented as a parse tree from a context-free grammar. We propose a variational autoencoder which encodes and decodes directly to and from these parse trees, ensuring the generated outputs are always valid. Surprisingly, we show that not only does our model more often generate valid outputs, it also learns a more coherent latent space in which nearby points decode to similar discrete outputs. We demonstrate the effectiveness of our learned models by showing their improved performance in Bayesian optimization for symbolic regression and molecular synthesis."
                },
                {
                    "title": "Automatic Chemical Design Using a Data-Driven Continuous Representation of Molecules",
                    "abstract": "We report a method to convert discrete representations of molecules to and from a multidimensional continuous representation. This model allows us to generate new molecules for efficient exploration and optimization through open-ended spaces of chemical compounds. A deep neural network was trained on hundreds of thousands of existing chemical structures to construct three coupled functions: an encoder, a decoder, and a predictor. The encoder converts the discrete representation of a molecule into a real-valued continuous vector, and the decoder converts these continuous vectors back to discrete molecular representations. The predictor estimates chemical properties from the latent continuous vector representation of the molecule. Continuous representations of molecules allow us to automatically generate novel chemical structures by performing simple operations in the latent space, such as decoding random vectors, perturbing known chemical structures, or interpolating between molecules. Continuous representations also allow the use of powerful gradient-based optimization to efficiently guide the search for optimized functional compounds. We demonstrate our method in the domain of drug-like molecules and also in a set of molecules with fewer that nine heavy atoms."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "ANARCI: antigen receptor numbering and receptor classification",
                    "abstract": "Motivation: Antibody amino-acid sequences can be numbered to identify equivalent positions. Such annotations are valuable for antibody sequence comparison, protein structure modelling and engineering. Multiple different numbering schemes exist, they vary in the nomenclature they use to annotate residue positions, their definitions of position equivalence and their popularity within different scientific disciplines. However, currently no publicly available software exists that can apply all the most widely used schemes or for which an executable can be obtained under an open license. Results: ANARCI is a tool to classify and number antibody and T-cell receptor amino-acid variable domain sequences. It can annotate sequences with the five most popular numbering schemes: Kabat, Chothia, Enhanced Chothia, IMGT and AHo. Availability and implementation: ANARCI is available for download under GPLv3 license at opig.stats.ox.ac.uk/webapps/anarci. A web-interface to the program is available at the same address. Contact: deane@stats.ox.ac.uk"
                },
                {
                    "title": "Matching methods for causal inference: A review and a look forward.",
                    "abstract": "When estimating causal effects using observational data, it is desirable to replicate a randomized experiment as closely as possible by obtaining treated and control groups with similar covariate distributions. This goal can often be achieved by choosing well-matched samples of the original treated and control groups, thereby reducing bias due to the covariates. Since the 1970's, work on matching methods has examined how to best choose treated and control subjects for comparison. Matching methods are gaining popularity in fields such as economics, epidemiology, medicine, and political science. However, until now the literature and related advice has been scattered across disciplines. Researchers who are interested in using matching methods-or developing methods related to matching-do not have a single place to turn to learn about past and current research. This paper provides a structure for thinking about matching methods and guidance on their use, coalescing the existing research (both old and new) and providing a summary of where the literature on matching methods is now and where it should be headed."
                },
                {
                    "title": "The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials",
                    "abstract": "For estimating causal effects of treatments, randomized experiments are generally considered the gold standard. Nevertheless, they are often infeasible to conduct for a variety of reasons, such as ethical concerns, excessive expense, or timeliness. Consequently, much of our knowledge of causal effects must come from non\u2010randomized observational studies. This article will advocate the position that observational studies can and should be designed to approximate randomized experiments as closely as possible. In particular, observational studies should be designed using only background information to create subgroups of similar treated and control units, where \u2018similar\u2019 here refers to their distributions of background variables. Of great importance, this activity should be conducted without any access to any outcome data, thereby assuring the objectivity of the design. In many situations, this objective creation of subgroups of similar treated and control units, which are balanced with respect to covariates, can be accomplished using propensity score methods. The theoretical perspective underlying this position will be presented followed by a particular application in the context of the US tobacco litigation. This application uses propensity score methods to create subgroups of treated units (male current smokers) and control units (male never smokers) who are at least as similar with respect to their distributions of observed background characteristics as if they had been randomized. The collection of these subgroups then \u2018approximate\u2019 a randomized block experiment with respect to the observed covariates. Copyright \u00a9 2006 John Wiley & Sons, Ltd."
                },
                {
                    "title": "Yet another numbering scheme for immunoglobulin variable domains: an automatic modeling and analysis tool.",
                    "abstract": "A common residue numbering scheme for all immunoglobulin variable domains (immunoglobulin light chain lambda (V(lambda)) and kappa (V(kappa)) variable domains, heavy chain variable domains (V(H)) and T-cell receptor alpha (V(alpha)), beta (V(beta)), gamma (V(gamma)) and delta (V(delta)) variable domains) has been devised. Based on the spatial alignment of known three-dimensional structures of immunoglobulin domains, it places the alignment gaps in a way that minimizes the average deviation from the averaged structure of the aligned domains. This residue numbering scheme was applied to the immunoglobulin variable domain structures in the PDB database to automate the extraction of information on structural variations in homologous positions of the different molecules. A number of methods are presented that allow the automated projection of information derived from individual structures or from the comparison of multi-structure alignments onto a graphical representation of the sequence alignment."
                },
                {
                    "title": "Matched Sampling for Causal Effects: Matching to Remove Bias in Observational Studies",
                    "abstract": "Several matching methods that match all of one sample from another larger sample on a continuous matching variable are compared with respect to their ability to remove the bias of the matching variable. One method is a simple mean-matching method and three are nearest available pair-matching methods. The methods' abilities to remove bias are also compared with the theoretical maximum given fixed distributions and fixed sample sizes. A summary of advice to an investigator is included."
                },
                {
                    "title": "The Computerized Construction of a Matched Sample",
                    "abstract": "The use of matched sampling in experimental and nonexperimental research has been frustrated by the problems of attrition and incomplete matching. The amount of attrition and incomplete matching. The amount of attrition of sample size and incomplete matching are shown to be a function of various aspects of the research, including the level of measurement of the matching variables, the sampling situation, the definition of a match, and the means used to construct matches. Several basic and heretofore neglected considerations in the conceptualization of a \"match\" are discussed. These considerations and a rarallel set of desiderata became apparent during the computerized construction of two matched samples of black and white college graduates. Three approaches to the computerization of matching are described, the last of which-\"guaranteed variable caliper matching\"- meets the desiderata. We describe our clerical and field procedures from the computerized matching to the final matching based on completed interviews. The degree to which both computer matching and these procedures controlled attition and incomplete matching is assessed. The quality of matches is found to be fairly high and attrition has been virtually elliminated, thus demonstrating the possibilities of our approach."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively optimize design properties in various fields using machine learning techniques, particularly in scenarios with limited data availability and non-smooth functional dependencies?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the inefficiencies in the design cycle across multiple disciplines, including engineering and life sciences. By developing machine learning methods that can operate effectively with small datasets and complex relationships, we can significantly reduce the time and resources required for design optimization. This advancement could lead to breakthroughs in drug discovery, material science, and engineering design, ultimately fostering innovation and practical applications that enhance performance and efficiency in these fields.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the limited availability of data, which restricts the training of machine learning models. Additionally, the non-smooth functional dependencies between design features and outcomes complicate the approximation process, making it difficult for even advanced models like deep neural networks to generalize effectively. Naive approaches may fail because they do not account for the intricacies of the design space or the need for reliable property predictions, which are essential for guiding the optimization process. Overcoming these technical and practical obstacles requires innovative methodologies that can leverage existing data more effectively.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often relied on traditional two-part frameworks that require separate discriminators for property enhancement, which can be inefficient and data-intensive. The limitations of existing generative models, such as their data-hungry nature and unreliability with out-of-distribution examples, have also hindered progress. Additionally, the application of matching techniques from econometrics to machine learning for design optimization has not been fully explored. Our approach differs by integrating matching techniques to create a more effective framework for optimizing design properties, thereby addressing the gaps left by prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using matching techniques to create a dataset where each sample is paired with a similar one that has a better property. We will implement an encoder-decoder framework that learns a lower-dimensional manifold, organizing embeddings by property value. The dataset will be constructed from small-scale design examples, and we will evaluate the model's performance using metrics such as property improvement and generalization ability. The expected outcomes include a more efficient design optimization process"
            }
        },
        "author_data": {
            "b1b38de5-4ea5-4c72-b55d-eb408bdb79c4": {
                "pk": "b1b38de5-4ea5-4c72-b55d-eb408bdb79c4",
                "name": "Nata\u0161a Tagasovska",
                "collaborators": [
                    "Stephen Ra",
                    "Kyunghyun Cho",
                    "Nathan C. Frey",
                    "Ji Won Park",
                    "Firat Ozdemir",
                    "Axel Brando",
                    "Andreas Loukas",
                    "Isidro H\u00f6tzel",
                    "Julien Lafrance-Vanasse",
                    "Ryan Lewis Kelly"
                ],
                "domain": [
                    "Uncertainty Estimation",
                    "Deep Learning",
                    "Multi-objective Optimization",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Retrospective Uncertainties for Deep Models using Vine Copulas",
                        "abstract": "Despite the major progress of deep models as learning machines, uncertainty estimation remains a major challenge. Existing solutions rely on modified loss functions or architectural changes. We propose to compensate for the lack of built-in uncertainty estimates by supplementing any network, retrospectively, with a subsequent vine copula model, in an overall compound we call Vine-Copula Neural Network (VCNN). Through synthetic and real-data experiments, we show that VCNNs could be task (regression/classification) and architecture (recurrent, fully connected) agnostic while providing reliable and better-calibrated uncertainty estimates, comparable to state-of-the-art built-in uncertainty solutions."
                    },
                    {
                        "title": "A Pareto-optimal compositional energy-based model for sampling and optimization of protein sequences",
                        "abstract": "Deep generative models have emerged as a popular machine learning-based approach for inverse design problems in the life sciences. However, these problems often require sampling new designs that satisfy multiple properties of interest in addition to learning the data distribution. This multi-objective optimization becomes more challenging when properties are independent or orthogonal to each other. In this work, we propose a Pareto-compositional energy-based model (pcEBM), a framework that uses multiple gradient descent for sampling new designs that adhere to various constraints in optimizing distinct properties. We demonstrate its ability to learn non-convex Pareto fronts and generate sequences that simultaneously satisfy multiple desired properties across a series of real-world antibody design tasks."
                    },
                    {
                        "title": "Antibody DomainBed: Out-of-Distribution Generalization in Therapeutic Protein Design",
                        "abstract": "Machine learning (ML) has demonstrated significant promise in accelerating drug design. Active ML-guided optimization of therapeutic molecules typically relies on a surrogate model predicting the target property of interest. The model predictions are used to determine which designs to evaluate in the lab, and the model is updated on the new measurements to inform the next cycle of decisions. A key challenge is that the experimental feedback from each cycle inspires changes in the candidate proposal or experimental protocol for the next cycle, which lead to distribution shifts. To promote robustness to these shifts, we must account for them explicitly in the model training. We apply domain generalization (DG) methods to classify the stability of interactions between an antibody and antigen across five domains defined by design cycles. Our results suggest that foundational models and ensembling improve predictive performance on out-of-distribution domains. We publicly release our codebase extending the DG benchmark ``DomainBed,'' and the associated dataset of antibody sequences and structures emulating distribution shifts across design cycles."
                    },
                    {
                        "title": "Learning Causal Representations of Single Cells via Sparse Mechanism Shift Modeling",
                        "abstract": "Latent variable models such as the Variational Auto-Encoder (VAE) have become a go-to tool for analyzing biological data, especially in the field of single-cell genomics. One remaining challenge is the interpretability of latent variables as biological processes that define a cell's identity. Outside of biological applications, this problem is commonly referred to as learning disentangled representations. Although several disentanglement-promoting variants of the VAE were introduced, and applied to single-cell genomics data, this task has been shown to be infeasible from independent and identically distributed measurements, without additional structure. Instead, recent methods propose to leverage non-stationary data, as well as the sparse mechanism shift assumption in order to learn disentangled representations with a causal semantic. Here, we extend the application of these methodological advances to the analysis of single-cell genomics data with genetic or chemical perturbations. More precisely, we propose a deep generative model of single-cell gene expression data for which each perturbation is treated as a stochastic intervention targeting an unknown, but sparse, subset of latent variables. We benchmark these methods on simulated single-cell data to evaluate their performance at latent units recovery, causal target identification and out-of-domain generalization. Finally, we apply those approaches to two real-world large-scale gene perturbation data sets and find that models that exploit the sparse mechanism shift hypothesis surpass contemporary methods on a transfer learning task. We implement our new model and benchmarks using the scvi-tools library, and release it as open-source software at https://github.com/Genentech/sVAE."
                    },
                    {
                        "title": "BOtied: Multi-objective Bayesian optimization with tied multivariate ranks",
                        "abstract": "Many scientific and industrial applications require the joint optimization of multiple, potentially competing objectives. Multi-objective Bayesian optimization (MOBO) is a sample-efficient framework for identifying Pareto-optimal solutions. At the heart of MOBO is the acquisition function, which determines the next candidate to evaluate by navigating the best compromises among the objectives. In this paper, we show a natural connection between non-dominated solutions and the extreme quantile of the joint cumulative distribution function (CDF). Motivated by this link, we propose the Pareto-compliant CDF indicator and the associated acquisition function, BOtied. BOtied inherits desirable invariance properties of the CDF, and an efficient implementation with copulas allows it to scale to many objectives. Our experiments on a variety of synthetic and real-world problems demonstrate that BOtied outperforms state-of-the-art MOBO acquisition functions while being computationally efficient for many objectives."
                    },
                    {
                        "title": "Uncertainty modeling for fine-tuned implicit functions",
                        "abstract": "Implicit functions such as Neural Radiance Fields (NeRFs), occupancy networks, and signed distance functions (SDFs) have become pivotal in computer vision for reconstructing detailed object shapes from sparse views. Achieving optimal performance with these models can be challenging due to the extreme sparsity of inputs and distribution shifts induced by data corruptions. To this end, large, noise-free synthetic datasets can serve as shape priors to help models fill in gaps, but the resulting reconstructions must be approached with caution. Uncertainty estimation is crucial for assessing the quality of these reconstructions, particularly in identifying areas where the model is uncertain about the parts it has inferred from the prior. In this paper, we introduce Dropsembles, a novel method for uncertainty estimation in tuned implicit functions. We demonstrate the efficacy of our approach through a series of experiments, starting with toy examples and progressing to a real-world scenario. Specifically, we train a Convolutional Occupancy Network on synthetic anatomical data and test it on low-resolution MRI segmentations of the lumbar spine. Our results show that Dropsembles achieve the accuracy and calibration levels of deep ensembles but with significantly less computational cost."
                    }
                ]
            },
            "8e5ae425-2769-4aed-ab02-207ae05606c7": {
                "pk": "8e5ae425-2769-4aed-ab02-207ae05606c7",
                "name": "Vladimir Gligorijevi\u0107",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "b36a92ef-8ef6-48c9-80fc-fa2413164414": {
                "pk": "b36a92ef-8ef6-48c9-80fc-fa2413164414",
                "name": "Kyunghyun Cho",
                "collaborators": [
                    "Tian Wang",
                    "Masha Esipova",
                    "Xi Chen",
                    "Yijun Xiao",
                    "Alex Wang",
                    "S\u00e9bastien Jean",
                    "Keunwoo Choi",
                    "Jake Zhao",
                    "Sean Welleck",
                    "Yoshua Bengio"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Deep Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Natural Language Understanding with Distributed Representation",
                        "abstract": "This is a lecture note for the course DS-GA 3001 <Natural Language Understanding with Distributed Representation> at the Center for Data Science , New York University in Fall, 2015. As the name of the course suggests, this lecture note introduces readers to a neural network based approach to natural language understanding/processing. In order to make it as self-contained as possible, I spend much time on describing basics of machine learning and neural networks, only after which how they are used for natural languages is introduced. On the language front, I almost solely focus on language modelling and machine translation, two of which I personally find most fascinating and most fundamental to natural language understanding."
                    },
                    {
                        "title": "Strawman: an Ensemble of Deep Bag-of-Ngrams for Sentiment Analysis",
                        "abstract": "This paper describes a builder entry, named \"strawman\", to the sentence-level sentiment analysis task of the \"Build It, Break It\" shared task of the First Workshop on Building Linguistically Generalizable NLP Systems. The goal of a builder is to provide an automated sentiment analyzer that would serve as a target for breakers whose goal is to find pairs of minimally-differing sentences that break the analyzer."
                    },
                    {
                        "title": "A Brief Introduction to Causal Inference in Machine Learning",
                        "abstract": "This is a lecture note produced for DS-GA 3001.003 \"Special Topics in DS - Causal Inference in Machine Learning\" at the Center for Data Science, New York University in Spring, 2024. This course was created to target master's and PhD level students with basic background in machine learning but who were not exposed to causal inference or causal reasoning in general previously. In particular, this course focuses on introducing such students to expand their view and knowledge of machine learning to incorporate causal reasoning, as this aspect is at the core of so-called out-of-distribution generalization (or lack thereof.)"
                    },
                    {
                        "title": "Boltzmann Machines and Denoising Autoencoders for Image Denoising",
                        "abstract": "Image denoising based on a probabilistic model of local image patches has been employed by various researchers, and recently a deep (denoising) autoencoder has been proposed by Burger et al. [2012] and Xie et al. [2012] as a good model for this. In this paper, we propose that another popular family of models in the field of deep learning, called Boltzmann machines, can perform image denoising as well as, or in certain cases of high level of noise, better than denoising autoencoders. We empirically evaluate the two models on three different sets of images with different types and levels of noise. Throughout the experiments we also examine the effect of the depth of the models. The experiments confirmed our claim and revealed that the performance can be improved by adding more hidden layers, especially when the level of noise is high."
                    },
                    {
                        "title": "Understanding Dropout: Training Multi-Layer Perceptrons with Auxiliary Independent Stochastic Neurons",
                        "abstract": "In this paper, a simple, general method of adding auxiliary stochastic neurons to a multi-layer perceptron is proposed. It is shown that the proposed method is a generalization of recently successful methods of dropout (Hinton et al., 2012), explicit noise injection (Vincent et al., 2010; Bishop, 1995) and semantic hashing (Salakhutdinov & Hinton, 2009). Under the proposed framework, an extension of dropout which allows using separate dropping probabilities for different hidden neurons, or layers, is found to be available. The use of different dropping probabilities for hidden layers separately is empirically investigated."
                    },
                    {
                        "title": "Noisy Parallel Approximate Decoding for Conditional Recurrent Language Model",
                        "abstract": "Recent advances in conditional recurrent language modelling have mainly focused on network architectures (e.g., attention mechanism), learning algorithms (e.g., scheduled sampling and sequence-level training) and novel applications (e.g., image/video description generation, speech recognition, etc.) On the other hand, we notice that decoding algorithms/strategies have not been investigated as much, and it has become standard to use greedy or beam search. In this paper, we propose a novel decoding strategy motivated by an earlier observation that nonlinear hidden layers of a deep neural network stretch the data manifold. The proposed strategy is embarrassingly parallelizable without any communication overhead, while improving an existing decoding algorithm. We extensively evaluate it with attention-based neural machine translation on the task of En->Cz translation."
                    },
                    {
                        "title": "Can neural machine translation do simultaneous translation?",
                        "abstract": "We investigate the potential of attention-based neural machine translation in simultaneous translation. We introduce a novel decoding algorithm, called simultaneous greedy decoding, that allows an existing neural machine translation model to begin translating before a full source sentence is received. This approach is unique from previous works on simultaneous translation in that segmentation and translation are done jointly to maximize the translation quality and that translating each segment is strongly conditioned on all the previous segments. This paper presents a first step toward building a full simultaneous translation system based on neural machine translation."
                    },
                    {
                        "title": "Classifying and Visualizing Motion Capture Sequences using Deep Neural Networks",
                        "abstract": "The gesture recognition using motion capture data and depth sensors has recently drawn more attention in vision recognition. Currently most systems only classify dataset with a couple of dozens different actions. Moreover, feature extraction from the data is often computational complex. In this paper, we propose a novel system to recognize the actions from skeleton data with simple, but effective, features using deep neural networks. Features are extracted for each frame based on the relative positions of joints (PO), temporal differences (TD), and normalized trajectories of motion (NT). Given these features a hybrid multi-layer perceptron is trained, which simultaneously classifies and reconstructs input data. We use deep autoencoder to visualize learnt features, and the experiments show that deep neural networks can capture more discriminative information than, for instance, principal component analysis can. We test our system on a public database with 65 classes and more than 2,000 motion sequences. We obtain an accuracy above 95% which is, to our knowledge, the state of the art result for such a large dataset."
                    },
                    {
                        "title": "Efficient Character-level Document Classification by Combining Convolution and Recurrent Layers",
                        "abstract": "Document classification tasks were primarily tackled at word level. Recent research that works with character-level inputs shows several benefits over word-level approaches such as natural incorporation of morphemes and better handling of rare words. We propose a neural network architecture that utilizes both convolution and recurrent layers to efficiently encode character inputs. We validate the proposed model on eight large scale document classification tasks and compare with character-level convolution-only models. It achieves comparable performances with much less parameters."
                    },
                    {
                        "title": "BERT has a Mouth, and It Must Speak: BERT as a Markov Random Field Language Model",
                        "abstract": "We show that BERT (Devlin et al., 2018) is a Markov random field language model. This formulation gives way to a natural procedure to sample sentences from BERT. We generate from BERT and find that it can produce high-quality, fluent generations. Compared to the generations of a traditional left-to-right language model, BERT generates sentences that are more diverse but of slightly worse quality."
                    },
                    {
                        "title": "Context-Aware Learning for Neural Machine Translation",
                        "abstract": "Interest in larger-context neural machine translation, including document-level and multi-modal translation, has been growing. Multiple works have proposed new network architectures or evaluation schemes, but potentially helpful context is still sometimes ignored by larger-context translation models. In this paper, we propose a novel learning algorithm that explicitly encourages a neural translation model to take into account additional context using a multilevel pair-wise ranking loss. We evaluate the proposed learning algorithm with a transformer-based larger-context translation system on document-level translation. By comparing performance using actual and random contexts, we show that a model trained with the proposed algorithm is more sensitive to the additional context."
                    },
                    {
                        "title": "Deep Unsupervised Drum Transcription",
                        "abstract": "We introduce DrummerNet, a drum transcription system that is trained in an unsupervised manner. DrummerNet does not require any ground-truth transcription and, with the data-scalability of deep neural networks, learns from a large unlabeled dataset. In DrummerNet, the target drum signal is first passed to a (trainable) transcriber, then reconstructed in a (fixed) synthesizer according to the transcription estimate. By training the system to minimize the distance between the input and the output audio signals, the transcriber learns to transcribe without ground truth transcription. Our experiment shows that DrummerNet performs favorably compared to many other recent drum transcription systems, both supervised and unsupervised."
                    },
                    {
                        "title": "Attention-based Mixture Density Recurrent Networks for History-based Recommendation",
                        "abstract": "The goal of personalized history-based recommendation is to automatically output a distribution over all the items given a sequence of previous purchases of a user. In this work, we present a novel approach that uses a recurrent network for summarizing the history of purchases, continuous vectors representing items for scalability, and a novel attention-based recurrent mixture density network, which outputs each component in a mixture sequentially, for modelling a multi-modal conditional distribution. We evaluate the proposed approach on two publicly available datasets, MovieLens-20M and RecSys15. The experiments show that the proposed approach, which explicitly models the multi-modal nature of the predictive distribution, is able to improve the performance over various baselines in terms of precision, recall and nDCG."
                    },
                    {
                        "title": "Retrieval-Augmented Convolutional Neural Networks for Improved Robustness against Adversarial Examples",
                        "abstract": "We propose a retrieval-augmented convolutional network and propose to train it with local mixup, a novel variant of the recently proposed mixup algorithm. The proposed hybrid architecture combining a convolutional network and an off-the-shelf retrieval engine was designed to mitigate the adverse effect of off-manifold adversarial examples, while the proposed local mixup addresses on-manifold ones by explicitly encouraging the classifier to locally behave linearly on the data manifold. Our evaluation of the proposed approach against five readily-available adversarial attacks on three datasets--CIFAR-10, SVHN and ImageNet--demonstrate the improved robustness compared to the vanilla convolutional network."
                    },
                    {
                        "title": "Sequential Graph Dependency Parser",
                        "abstract": "We propose a method for non-projective dependency parsing by incrementally predicting a set of edges. Since the edges do not have a pre-specified order, we propose a set-based learning method. Our method blends graph, transition, and easy-first parsing, including a prior state of the parser as a special case. The proposed transition-based method successfully parses near the state of the art on both projective and non-projective languages, without assuming a certain parsing order."
                    },
                    {
                        "title": "Exponentially Increasing the Capacity-to-Computation Ratio for Conditional Computation in Deep Learning",
                        "abstract": "Many state-of-the-art results obtained with deep networks are achieved with the largest models that could be trained, and if more computation power was available, we might be able to exploit much larger datasets in order to improve generalization ability. Whereas in learning algorithms such as decision trees the ratio of capacity (e.g., the number of parameters) to computation is very favorable (up to exponentially more parameters than computation), the ratio is essentially 1 for deep neural networks. Conditional computation has been proposed as a way to increase the capacity of a deep neural network without increasing the amount of computation required, by activating some parameters and computation \"on-demand\", on a per-example basis. In this note, we propose a novel parametrization of weight matrices in neural networks which has the potential to increase up to exponentially the ratio of the number of parameters to computation. The proposed approach is based on turning on some parameters (weight matrices) when specific bit patterns of hidden unit activations are obtained. In order to better control for the overfitting that might result, we propose a parametrization that is tree-structured, where each node of the tree corresponds to a prefix of a sequence of sign bits, or gating units, associated with hidden units."
                    },
                    {
                        "title": "Larger-Context Language Modelling",
                        "abstract": "In this work, we propose a novel method to incorporate corpus-level discourse information into language modelling. We call this larger-context language model. We introduce a late fusion approach to a recurrent language model based on long short-term memory units (LSTM), which helps the LSTM unit keep intra-sentence dependencies and inter-sentence dependencies separate from each other. Through the evaluation on three corpora (IMDB, BBC, and PennTree Bank), we demon- strate that the proposed model improves perplexity significantly. In the experi- ments, we evaluate the proposed approach while varying the number of context sentences and observe that the proposed late fusion is superior to the usual way of incorporating additional inputs to the LSTM. By analyzing the trained larger- context language model, we discover that content words, including nouns, adjec- tives and verbs, benefit most from an increasing number of context sentences. This analysis suggests that larger-context language model improves the unconditional language model by capturing the theme of a document better and more easily."
                    },
                    {
                        "title": "Query-Efficient Imitation Learning for End-to-End Autonomous Driving",
                        "abstract": "One way to approach end-to-end autonomous driving is to learn a policy function that maps from a sensory input, such as an image frame from a front-facing camera, to a driving action, by imitating an expert driver, or a reference policy. This can be done by supervised learning, where a policy function is tuned to minimize the difference between the predicted and ground-truth actions. A policy function trained in this way however is known to suffer from unexpected behaviours due to the mismatch between the states reachable by the reference policy and trained policy functions. More advanced algorithms for imitation learning, such as DAgger, addresses this issue by iteratively collecting training examples from both reference and trained policies. These algorithms often requires a large number of queries to a reference policy, which is undesirable as the reference policy is often expensive. In this paper, we propose an extension of the DAgger, called SafeDAgger, that is query-efficient and more suitable for end-to-end autonomous driving. We evaluate the proposed SafeDAgger in a car racing simulator and show that it indeed requires less queries to a reference policy. We observe a significant speed up in convergence, which we conjecture to be due to the effect of automated curriculum learning."
                    },
                    {
                        "title": "Gated Word-Character Recurrent Language Model",
                        "abstract": "We introduce a recurrent neural network language model (RNN-LM) with long short-term memory (LSTM) units that utilizes both character-level and word-level inputs. Our model has a gate that adaptively finds the optimal mixture of the character-level and word-level inputs. The gate creates the final vector representation of a word by combining two distinct representations of the word. The character-level inputs are converted into vector representations of words using a bidirectional LSTM. The word-level inputs are projected into another high-dimensional space by a word lookup table. The final vector representations of words are used in the LSTM language model which predicts the next word given all the preceding words. Our model with the gating mechanism effectively utilizes the character-level inputs for rare and out-of-vocabulary words and outperforms word-level language models on several English corpora."
                    },
                    {
                        "title": "Task-Oriented Query Reformulation with Reinforcement Learning",
                        "abstract": "Search engines play an important role in our everyday lives by assisting us in finding the information we need. When we input a complex query, however, results are often far from satisfactory. In this work, we introduce a query reformulation system based on a neural network that rewrites a query to maximize the number of relevant documents returned. We train this neural network with reinforcement learning. The actions correspond to selecting terms to build a reformulated query, and the reward is the document recall. We evaluate our approach on three datasets against strong baselines and show a relative improvement of 5-20% in terms of recall. Furthermore, we present a simple method to estimate a conservative upper-bound performance of a model in a particular environment and verify that there is still large room for improvements."
                    }
                ]
            },
            "cb277e43-cc35-43ad-84cf-e2432ef742fb": {
                "pk": "cb277e43-cc35-43ad-84cf-e2432ef742fb",
                "name": "Andreas Loukas",
                "collaborators": [
                    "Pierre Vandergheynst",
                    "Nathanael Perraudin",
                    "Jean-Baptiste Cordonnier",
                    "Damien Foucard",
                    "Andrea Simonetto",
                    "Geert Leus",
                    "Yihe Dong",
                    "Marinos Poiitis",
                    "Stefanie Jegelka",
                    "Karolis Martinkus"
                ],
                "domain": [
                    "Graph Signal Processing",
                    "Graph Neural Network",
                    "Spectral Clustering",
                    "Time-Series Analysis"
                ],
                "publications": [
                    {
                        "title": "How close are the eigenvectors and eigenvalues of the sample and actual covariance matrices?",
                        "abstract": "How many samples are sufficient to guarantee that the eigenvectors and eigenvalues of the sample covariance matrix are close to those of the actual covariance matrix? For a wide family of distributions, including distributions with finite second moment and distributions supported in a centered Euclidean ball, we prove that the inner product between eigenvectors of the sample and actual covariance matrices decreases proportionally to the respective eigenvalue distance. Our findings imply non-asymptotic concentration bounds for eigenvectors, eigenspaces, and eigenvalues. They also provide conditions for distinguishing principal components based on a constant number of samples."
                    },
                    {
                        "title": "Graph reduction with spectral and cut guarantees",
                        "abstract": "Can one reduce the size of a graph without significantly altering its basic properties? The graph reduction problem is hereby approached from the perspective of restricted spectral approximation, a modification of the spectral similarity measure used for graph sparsification. This choice is motivated by the observation that restricted approximation carries strong spectral and cut guarantees, and that it implies approximation results for unsupervised learning problems relying on spectral embeddings.   The paper then focuses on coarsening---the most common type of graph reduction. Sufficient conditions are derived for a small graph to approximate a larger one in the sense of restricted similarity. These findings give rise to nearly-linear algorithms that, compared to both standard and advanced graph reduction methods, find coarse graphs of improved quality, often by a large margin, without sacrificing speed."
                    },
                    {
                        "title": "What graph neural networks cannot learn: depth vs width",
                        "abstract": "This paper studies the expressive power of graph neural networks falling within the message-passing framework (GNNmp). Two results are presented. First, GNNmp are shown to be Turing universal under sufficient conditions on their depth, width, node attributes, and layer expressiveness. Second, it is discovered that GNNmp can lose a significant portion of their power when their depth and width is restricted. The proposed impossibility statements stem from a new technique that enables the repurposing of seminal results from distributed computing and leads to lower bounds for an array of decision, optimization, and estimation problems involving graphs. Strikingly, several of these problems are deemed impossible unless the product of a GNNmp's depth and width exceeds a polynomial of the graph size; this dependence remains significant even for tasks that appear simple or when considering approximation."
                    },
                    {
                        "title": "How hard is to distinguish graphs with graph neural networks?",
                        "abstract": "A hallmark of graph neural networks is their ability to distinguish the isomorphism class of their inputs. This study derives hardness results for the classification variant of graph isomorphism in the message-passing model (MPNN). MPNN encompasses the majority of graph neural networks used today and is universal when nodes are given unique features. The analysis relies on the introduced measure of communication capacity. Capacity measures how much information the nodes of a network can exchange during the forward pass and depends on the depth, message-size, global state, and width of the architecture. It is shown that the capacity of MPNN needs to grow linearly with the number of nodes so that a network can distinguish trees and quadratically for general connected graphs. The derived bounds concern both worst- and average-case behavior and apply to networks with/without unique features and adaptive architecture -- they are also up to two orders of magnitude tighter than those given by simpler arguments. An empirical study involving 12 graph classification tasks and 420 networks reveals strong alignment between actual performance and theoretical predictions."
                    },
                    {
                        "title": "Frequency Analysis of Temporal Graph Signals",
                        "abstract": "This letter extends the concept of graph-frequency to graph signals that evolve with time. Our goal is to generalize and, in fact, unify the familiar concepts from time- and graph-frequency analysis. To this end, we study a joint temporal and graph Fourier transform (JFT) and demonstrate its attractive properties. We build on our results to create filters which act on the joint (temporal and graph) frequency domain, and show how these can be used to perform interference cancellation. The proposed algorithms are distributed, have linear complexity, and can approximate any desired joint filtering objective."
                    },
                    {
                        "title": "Predicting the evolution of stationary graph signals",
                        "abstract": "An emerging way of tackling the dimensionality issues arising in the modeling of a multivariate process is to assume that the inherent data structure can be captured by a graph. Nevertheless, though state-of-the-art graph-based methods have been successful for many learning tasks, they do not consider time-evolving signals and thus are not suitable for prediction. Based on the recently introduced joint stationarity framework for time-vertex processes, this letter considers multivariate models that exploit the graph topology so as to facilitate the prediction. The resulting method yields similar accuracy to the joint (time-graph) mean-squared error estimator but at lower complexity, and outperforms purely time-based methods."
                    },
                    {
                        "title": "Spectrally approximating large graphs with smaller graphs",
                        "abstract": "How does coarsening affect the spectrum of a general graph? We provide conditions such that the principal eigenvalues and eigenspaces of a coarsened and original graph Laplacian matrices are close. The achieved approximation is shown to depend on standard graph-theoretic properties, such as the degree and eigenvalue distributions, as well as on the ratio between the coarsened and actual graph sizes. Our results carry implications for learning methods that utilize coarsening. For the particular case of spectral clustering, they imply that coarse eigenvectors can be used to derive good quality assignments even without refinement---this phenomenon was previously observed, but lacked formal justification."
                    },
                    {
                        "title": "Distributed Autoregressive Moving Average Graph Filters",
                        "abstract": "We introduce the concept of autoregressive moving average (ARMA) filters on a graph and show how they can be implemented in a distributed fashion. Our graph filter design philosophy is independent of the particular graph, meaning that the filter coefficients are derived irrespective of the graph. In contrast to finite-impulse response (FIR) graph filters, ARMA graph filters are robust against changes in the signal and/or graph. In addition, when time-varying signals are considered, we prove that the proposed graph filters behave as ARMA filters in the graph domain and, depending on the implementation, as first or higher ARMA filters in the time domain."
                    },
                    {
                        "title": "Extrapolating paths with graph neural networks",
                        "abstract": "We consider the problem of path inference: given a path prefix, i.e., a partially observed sequence of nodes in a graph, we want to predict which nodes are in the missing suffix. In particular, we focus on natural paths occurring as a by-product of the interaction of an agent with a network---a driver on the transportation network, an information seeker in Wikipedia, or a client in an online shop. Our interest is sparked by the realization that, in contrast to shortest-path problems, natural paths are usually not optimal in any graph-theoretic sense, but might still follow predictable patterns.   Our main contribution is a graph neural network called Gretel. Conditioned on a path prefix, this network can efficiently extrapolate path suffixes, evaluate path likelihood, and sample from the future path distribution. Our experiments with GPS traces on a road network and user-navigation paths in Wikipedia confirm that Gretel is able to adapt to graphs with very different properties, while also comparing favorably to previous solutions."
                    },
                    {
                        "title": "Attention is Not All You Need: Pure Attention Loses Rank Doubly Exponentially with Depth",
                        "abstract": "Attention-based architectures have become ubiquitous in machine learning, yet our understanding of the reasons for their effectiveness remains limited. This work proposes a new way to understand self-attention networks: we show that their output can be decomposed into a sum of smaller terms, each involving the operation of a sequence of attention heads across layers. Using this decomposition, we prove that self-attention possesses a strong inductive bias towards \"token uniformity\". Specifically, without skip connections or multi-layer perceptrons (MLPs), the output converges doubly exponentially to a rank-1 matrix. On the other hand, skip connections and MLPs stop the output from degeneration. Our experiments verify the identified convergence phenomena on different variants of standard transformer architectures."
                    },
                    {
                        "title": "What training reveals about neural network complexity",
                        "abstract": "This work explores the Benevolent Training Hypothesis (BTH) which argues that the complexity of the function a deep neural network (NN) is learning can be deduced by its training dynamics. Our analysis provides evidence for BTH by relating the NN's Lipschitz constant at different regions of the input space with the behavior of the stochastic training procedure. We first observe that the Lipschitz constant close to the training data affects various aspects of the parameter trajectory, with more complex networks having a longer trajectory, bigger variance, and often veering further from their initialization. We then show that NNs whose 1st layer bias is trained more steadily (i.e., slowly and with little variation) have bounded complexity even in regions of the input space that are far from any training point. Finally, we find that steady training with Dropout implies a training- and data-dependent generalization bound that grows poly-logarithmically with the number of parameters. Overall, our results support the intuition that good training behavior can be a useful bias towards good generalization."
                    },
                    {
                        "title": "Generalizing to any diverse distribution: uniformity, gentle finetuning and rebalancing",
                        "abstract": "As training datasets grow larger, we aspire to develop models that generalize well to any diverse test distribution, even if the latter deviates significantly from the training data. Various approaches like domain adaptation, domain generalization, and robust optimization attempt to address the out-of-distribution challenge by posing assumptions about the relation between training and test distribution. Differently, we adopt a more conservative perspective by accounting for the worst-case error across all sufficiently diverse test distributions within a known domain. Our first finding is that training on a uniform distribution over this domain is optimal. We also interrogate practical remedies when uniform samples are unavailable by considering methods for mitigating non-uniformity through finetuning and rebalancing. Our theory provides a mathematical grounding for previous observations on the role of entropy and rebalancing for o.o.d. generalization and foundation model training. We also provide new empirical evidence across tasks involving o.o.d. shifts which illustrate the broad applicability of our perspective."
                    },
                    {
                        "title": "Approximating Spectral Clustering via Sampling: a Review",
                        "abstract": "Spectral clustering refers to a family of unsupervised learning algorithms that compute a spectral embedding of the original data based on the eigenvectors of a similarity graph. This non-linear transformation of the data is both the key of these algorithms' success and their Achilles heel: forming a graph and computing its dominant eigenvectors can indeed be computationally prohibitive when dealing with more that a few tens of thousands of points. In this paper, we review the principal research efforts aiming to reduce this computational cost. We focus on methods that come with a theoretical control on the clustering performance and incorporate some form of sampling in their operation. Such methods abound in the machine learning, numerical linear algebra, and graph signal processing literature and, amongst others, include Nystr\\\"om-approximation, landmarks, coarsening, coresets, and compressive spectral clustering. We present the approximation guarantees available for each and discuss practical merits and limitations. Surprisingly, despite the breadth of the literature explored, we conclude that there is still a gap between theory and practice: the most scalable methods are only intuitively motivated or loosely controlled, whereas those that come with end-to-end guarantees rely on strong assumptions or enable a limited gain of computation time."
                    },
                    {
                        "title": "Stationary time-vertex signal processing",
                        "abstract": "This paper considers regression tasks involving high-dimensional multivariate processes whose structure is dependent on some {known} graph topology. We put forth a new definition of time-vertex wide-sense stationarity, or joint stationarity for short, that goes beyond product graphs. Joint stationarity helps by reducing the estimation variance and recovery complexity. In particular, for any jointly stationary process (a) one reliably learns the covariance structure from as little as a single realization of the process, and (b) solves MMSE recovery problems, such as interpolation and denoising, in computational time nearly linear on the number of edges and timesteps. Experiments with three datasets suggest that joint stationarity can yield accuracy improvements in the recovery of high-dimensional processes evolving over a graph, even when the latter is only approximately known, or the process is not strictly stationary."
                    },
                    {
                        "title": "Erdos Goes Neural: an Unsupervised Learning Framework for Combinatorial Optimization on Graphs",
                        "abstract": "Combinatorial optimization problems are notoriously challenging for neural networks, especially in the absence of labeled instances. This work proposes an unsupervised learning framework for CO problems on graphs that can provide integral solutions of certified quality. Inspired by Erdos' probabilistic method, we use a neural network to parametrize a probability distribution over sets. Crucially, we show that when the network is optimized w.r.t. a suitably chosen loss, the learned distribution contains, with controlled probability, a low-cost integral solution that obeys the constraints of the combinatorial problem. The probabilistic proof of existence is then derandomized to decode the desired solutions. We demonstrate the efficacy of this approach to obtain valid solutions to the maximum clique problem and to perform local graph clustering. Our method achieves competitive results on both real datasets and synthetic hard instances."
                    },
                    {
                        "title": "Fast Approximate Spectral Clustering for Dynamic Networks",
                        "abstract": "Spectral clustering is a widely studied problem, yet its complexity is prohibitive for dynamic graphs of even modest size. We claim that it is possible to reuse information of past cluster assignments to expedite computation. Our approach builds on a recent idea of sidestepping the main bottleneck of spectral clustering, i.e., computing the graph eigenvectors, by using fast Chebyshev graph filtering of random signals. We show that the proposed algorithm achieves clustering assignments with quality approximating that of spectral clustering and that it can yield significant complexity benefits when the graph dynamics are appropriately bounded."
                    },
                    {
                        "title": "Towards Communication-Aware Robust Topologies",
                        "abstract": "We currently witness the emergence of interesting new network topologies optimized towards the traffic matrices they serve, such as demand-aware datacenter interconnects (e.g., ProjecToR) and demand-aware overlay networks (e.g., SplayNets). This paper introduces a formal framework and approach to reason about and design such topologies. We leverage a connection between the communication frequency of two nodes and the path length between them in the network, which depends on the entropy of the communication matrix. Our main contribution is a novel robust, yet sparse, family of network topologies which guarantee an expected path length that is proportional to the entropy of the communication patterns."
                    },
                    {
                        "title": "Towards stationary time-vertex signal processing",
                        "abstract": "Graph-based methods for signal processing have shown promise for the analysis of data exhibiting irregular structure, such as those found in social, transportation, and sensor networks. Yet, though these systems are often dynamic, state-of-the-art methods for signal processing on graphs ignore the dimension of time, treating successive graph signals independently or taking a global average. To address this shortcoming, this paper considers the statistical analysis of time-varying graph signals. We introduce a novel definition of joint (time-vertex) stationarity, which generalizes the classical definition of time stationarity and the more recent definition appropriate for graphs. Joint stationarity gives rise to a scalable Wiener optimization framework for joint denoising, semi-supervised learning, or more generally inversing a linear operator, that is provably optimal. Experimental results on real weather data demonstrate that taking into account graph and time dimensions jointly can yield significant accuracy improvements in the reconstruction effort."
                    }
                ]
            }
        }
    },
    "2406.06769": {
        "paper_data": {
            "title": "DISCOVERYWORLD: A Virtual Environment for Developing and Evaluating Automated Scientific Discovery Agents",
            "url": "http://arxiv.org/abs/2406.06769v2",
            "arxiv_id": "2406.06769",
            "authors": [
                "Peter Jansen",
                "Marc-Alexandre C\u00f4t\u00e9",
                "Tushar Khot",
                "Erin Bransom",
                "Bhavana Dalvi Mishra",
                "Bodhisattwa Prasad Majumder",
                "Oyvind Tafjord",
                "Peter Clark"
            ],
            "abstract": "Automated scientific discovery promises to accelerate progress across scientific domains. However, developing and evaluating an AI agent's capacity for end-to-end scientific reasoning is challenging as running real-world experiments is often prohibitively expensive or infeasible. In this work we introduce DISCOVERYWORLD, the first virtual environment for developing and benchmarking an agent's ability to perform complete cycles of novel scientific discovery. DISCOVERYWORLD contains a variety of different challenges, covering topics as diverse as radioisotope dating, rocket science, and proteomics, to encourage development of general discovery skills rather than task-specific solutions. DISCOVERYWORLD itself is an inexpensive, simulated, text-based environment (with optional 2D visual overlay). It includes 120 different challenge tasks, spanning eight topics each with three levels of difficulty and several parametric variations. Each task requires an agent to form hypotheses, design and run experiments, analyze results, and act on conclusions. DISCOVERYWORLD further provides three automatic metrics for evaluating performance, based on (a) task completion, (b) task-relevant actions taken, and (c) the discovered explanatory knowledge. We find that strong baseline agents, that perform well in prior published environments, struggle on most DISCOVERYWORLD tasks, suggesting that DISCOVERYWORLD captures some of the novel challenges of discovery, and thus that DISCOVERYWORLD may help accelerate near-term development and assessment of scientific discovery competency in agents. Code available at: www.github.com/allenai/discoveryworld",
            "introduction": "   1 Introduction  A long-standing dream of AI has been to build systems that can perform scientific discovery, potentially leading to new breakthroughs for the benefit of humanity [13]. Recently, with the rise of neural techniques, there have been several successful discovery systems developed for specialized problems such as protein folding [10, 15], mathematics [22], and material science [24]. However, while the results have been impressive, these systems (deliberately) bypass the full discovery process of ideation, hypothesis formation, experiment design, etc., and instead (expertly) perform systematic searches over a pre-defined hypothesis space, with pre-defined goals. This raises the question: how much more can be achieved if AI is applied to the broader scientific process? Some works have indeed developed early systems for this, for example, in chemistry [1], and genetics [12]. These systems can also generate hypotheses, design experiments, and execute them (via robotics) in real environments. However, operating in real environments is expensive and complex, creating a barrier for entry. In addition, real environments inevitably encourage a focus on task-specific details, at the potential cost of developing more general discovery skills in an agent.   Our goal is to help remedy these by creating the first virtual discovery environment where solving tasks demands all of the key facets in end-to-end scientific discovery,222 While a precise definition of \u201dscientific discovery\u201d is somewhat elusive [17, 19], we here adopt a pragmatic approach: Our goal is to help develop systems that can perform the end-to-end research process that human scientists engage in as they work on a problem, attempt to answer a question, or more generally advance their field of study. We use the term \u201dscientific discovery\u201d here to refer to this process. and which covers a broad variety of discovery topics. Our approach is to develop a text-based simulated world (with optional 2D visual overlay), called DiscoveryWorld, where agents can navigate around, interact with objects in the world, use scientific equipment (measuring devices, tools, etc.), and make observations. Agents can then form hypotheses, plan and execute experiments, and draw conclusions to solve challenge tasks developed for this virtual world. DiscoveryWorld tasks are grounded in eight varied topics, such as radioisotope dating, rocket science, and proteomics, to encourage development of agents with general discovery skills rather than hard-wiring to a particular challenge (see Figure\u00a01). The tasks themselves are realistic (but simplified), allowing agents to apply both scientific and commonsense knowledge when attempting them. DiscoveryWorld thus provides an environment for exercising and evaluating general-purpose skills in end-to-end AI discovery systems (see Figure\u00a02).   DiscoveryWorld is inspired by a growing number of text-based simulation environments [5, 31, inter alia], while also being novel in both its environment and tasks:   \u2022  DiscoveryWorld tasks are long-horizon, requiring multiple facets of discovery including ideation, experimentation, systematic search, and analysis to be performed to solve a task.    \u2022  The tasks do not suggest a solution approach, instead requiring the agent to ideate and define hypotheses to explore. This contrasts with tasks in many adventure game environments where solution approaches are often more stylistic or constrained.    \u2022  DiscoveryWorld is realistic (but simplified) rather than counterfactual, so that background knowledge can be sensibly applied.    \u2022  The tasks cover eight diverse topics, from identifying the cause of space illnesses to reactor tuning, to encourage development of general rather than task-specific solutions.      Figure 2: DiscoveryWorld tasks require end-to-end scientific discovery, from ideation, hypothesis formation, experiment design, data collection and analysis, forming conclusions, and acting",
            "references": [
                {
                    "title": "ADaPT: As-Needed Decomposition and Planning with Language Models",
                    "abstract": "Large Language Models (LLMs) are increasingly being used for interactive decision-making tasks requiring planning and adapting to the environment. Recent works employ LLMs-as-agents in broadly two ways: iteratively determining the next action (iterative executors) or generating plans and executing sub-tasks using LLMs (plan-and-execute). However, these methods struggle with task complexity, as the inability to execute any sub-task may lead to task failure. To address these shortcomings, we introduce As-Needed Decomposition and Planning for complex Tasks (ADaPT), an approach that explicitly plans and decomposes complex sub-tasks as-needed, i.e., when the LLM is unable to execute them. ADaPT recursively decomposes sub-tasks to adapt to both task complexity and LLM capability. Our results demonstrate that ADaPT substantially outperforms established strong baselines, achieving success rates up to 28.3% higher in ALFWorld, 27% in WebShop, and 33% in TextCraft -- a novel compositional dataset that we introduce. Through extensive analysis, we illustrate the importance of multilevel decomposition and establish that ADaPT dynamically adjusts to the capabilities of the executor LLM as well as to task complexity."
                },
                {
                    "title": "CLIN: A Continually Learning Language Agent for Rapid Task Adaptation and Generalization",
                    "abstract": "Language agents have shown some ability to interact with an external environment, e.g., a virtual world such as ScienceWorld, to perform complex tasks, e.g., growing a plant, without the startup costs of reinforcement learning. However, despite their zero-shot capabilities, these agents to date do not continually improve over time beyond performance refinement on a specific task. Here we present CLIN, the first language-based agent to achieve this, so that it continually improves over multiple trials, including when both the environment and task are varied, and without requiring parameter updates. Our approach is to use a persistent, dynamic, textual memory centered on causal abstractions (rather than general\"helpful hints\") that is regularly updated after each trial so that the agent gradually learns useful knowledge for new trials. In the ScienceWorld benchmark, CLIN is able to continually improve on repeated trials on the same task and environment, outperforming state-of-the-art reflective language agents like Reflexion by 23 absolute points. CLIN can also transfer its learning to new environments (or new tasks), improving its zero-shot performance by 4 points (13 for new tasks) and can further improve performance there through continual memory updates, enhancing performance by an additional 17 points (7 for new tasks). This suggests a new architecture for agents built on frozen models that can still continually and rapidly improve over time."
                },
                {
                    "title": "Generalized Biomolecular Modeling and Design with RoseTTAFold All-Atom",
                    "abstract": "Although AlphaFold2 (AF2) and RoseTTAFold (RF) have transformed structural biology by enabling high-accuracy protein structure modeling, they are unable to model covalent modifications or interactions with small molecules and other non-protein molecules that can play key roles in biological function. Here, we describe RoseTTAFold All-Atom (RFAA), a deep network capable of modeling full biological assemblies containing proteins, nucleic acids, small molecules, metals, and covalent modifications given the sequences of the polymers and the atomic bonded geometry of the small molecules and covalent modifications. Following training on structures of full biological assemblies in the Protein Data Bank (PDB), RFAA has comparable protein structure prediction accuracy to AF2, excellent performance in CAMEO for flexible backbone small molecule docking, and reasonable prediction accuracy for protein covalent modifications and assemblies of proteins with multiple nucleic acid chains and small molecules which, to our knowledge, no existing method can model simultaneously. By fine-tuning on diffusive denoising tasks, we develop RFdiffusion All-Atom (RFdiffusionAA), which generates binding pockets by directly building protein structures around small molecules and other non-protein molecules. Starting from random distributions of amino acid residues surrounding target small molecules, we design and experimentally validate proteins that bind the cardiac disease therapeutic digoxigenin, the enzymatic cofactor heme, and optically active bilin molecules with potential for expanding the range of wavelengths captured by photosynthesis. We anticipate that RFAA and RFdiffusionAA will be widely useful for modeling and designing complex biomolecular systems."
                },
                {
                    "title": "MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation",
                    "abstract": "A central aspect of machine learning research is experimentation, the process of designing and running experiments, analyzing the results, and iterating towards some positive outcome (e.g., improving accuracy). Could agents driven by powerful language models perform machine learning experimentation effectively? To answer this question, we introduce MLAgentBench, a suite of 13 tasks ranging from improving model performance on CIFAR-10 to recent research problems like BabyLM. For each task, an agent can perform actions like reading/writing files, executing code, and inspecting outputs. We then construct an agent that can perform ML experimentation based on ReAct framework. We benchmark agents based on Claude v1.0, Claude v2.1, Claude v3 Opus, GPT-4, GPT-4-turbo, Gemini-Pro, and Mixtral and find that a Claude v3 Opus agent is the best in terms of success rate. It can build compelling ML models over many tasks in MLAgentBench with 37.5% average success rate. Our agents also display highly interpretable plans and actions. However, the success rates vary considerably; they span from 100% on well-established older datasets to as low as 0% on recent Kaggle challenges created potentially after the underlying LM was trained. Finally, we identify several key challenges for LM-based agents such as long-term planning and reducing hallucination. Our code is released at https://github.com/snap-stanford/MLAgentBench."
                },
                {
                    "title": "InterCode: Standardizing and Benchmarking Interactive Coding with Execution Feedback",
                    "abstract": "Humans write code in a fundamentally interactive manner and rely on constant execution feedback to correct errors, resolve ambiguities, and decompose tasks. While LLMs have recently exhibited promising coding capabilities, current coding benchmarks mostly consider a static instruction-to-code sequence transduction process, which has the potential for error propagation and a disconnect between the generated code and its final execution environment. To address this gap, we introduce InterCode, a lightweight, flexible, and easy-to-use framework of interactive coding as a standard reinforcement learning (RL) environment, with code as actions and execution feedback as observations. Our framework is language and platform agnostic, uses self-contained Docker environments to provide safe and reproducible execution, and is compatible out-of-the-box with traditional seq2seq coding methods, while enabling the development of new methods for interactive code generation. We use InterCode to create three interactive code environments with Bash, SQL, and Python as action spaces, leveraging data from the static NL2Bash, Spider, and MBPP datasets. We demonstrate InterCode's viability as a testbed by evaluating multiple state-of-the-art LLMs configured with different prompting strategies such as ReAct and Plan&Solve. Our results showcase the benefits of interactive code generation and demonstrate that InterCode can serve as a challenging benchmark for advancing code understanding and generation capabilities. InterCode is designed to be easily extensible and can even be used to create new tasks such as Capture the Flag, a popular coding puzzle that is inherently multi-step and involves multiple programming languages. Project site with code and data: https://intercode-benchmark.github.io"
                },
                {
                    "title": "Minigrid & Miniworld: Modular & Customizable Reinforcement Learning Environments for Goal-Oriented Tasks",
                    "abstract": "We present the Minigrid and Miniworld libraries which provide a suite of goal-oriented 2D and 3D environments. The libraries were explicitly created with a minimalistic design paradigm to allow users to rapidly develop new environments for a wide range of research-specific needs. As a result, both have received widescale adoption by the RL community, facilitating research in a wide range of areas. In this paper, we outline the design philosophy, environment details, and their world generation API. We also showcase the additional capabilities brought by the unified API between Minigrid and Miniworld through case studies on transfer learning (for both RL agents and humans) between the different observation spaces. The source code of Minigrid and Miniworld can be found at https://github.com/Farama-Foundation/{Minigrid, Miniworld} along with their documentation at https://{minigrid, miniworld}.farama.org/."
                },
                {
                    "title": "Plan-and-Solve Prompting: Improving Zero-Shot Chain-of-Thought Reasoning by Large Language Models",
                    "abstract": "Large language models (LLMs) have recently been shown to deliver impressive performance in various NLP tasks. To tackle multi-step reasoning tasks, Few-shot chain-of-thought (CoT) prompting includes a few manually crafted step-by-step reasoning demonstrations which enable LLMs to explicitly generate reasoning steps and improve their reasoning task accuracy. To eliminate the manual efforts, Zero-shot-CoT concatenates the target problem statement with \u201cLet\u2019s think step by step\u201d as an input prompt to LLMs. Despite the success of Zero-shot-CoT, it still suffers from three pitfalls: calculation errors, missing-step errors, and semantic misunderstanding errors. To address the missing-step errors, we propose Plan-and-Solve (PS) Prompting. It consists of two components: first, devising a plan to divide the entire task into smaller subtasks, and then carrying out the subtasks according to the plan. To address the calculation errors and improve the quality of generated reasoning steps, we extend PS prompting with more detailed instructions and derive PS+ prompting. We evaluate our proposed prompting strategy on ten datasets across three reasoning problems. The experimental results over GPT-3 show that our proposed zero-shot prompting consistently outperforms Zero-shot-CoT across all datasets by a large margin, is comparable to or exceeds Zero-shot-Program-of-Thought Prompting, and has comparable performance with 8-shot CoT prompting on the math reasoning problem. The code can be found at https://github.com/AGI-Edgerunners/Plan-and-Solve-Prompting."
                },
                {
                    "title": "ReAct: Synergizing Reasoning and Acting in Language Models",
                    "abstract": "While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources, such as knowledge bases or environments, to gather additional information. We apply our approach, named ReAct, to a diverse set of language and decision making tasks and demonstrate its effectiveness over state-of-the-art baselines, as well as improved human interpretability and trustworthiness over methods without reasoning or acting components. Concretely, on question answering (HotpotQA) and fact verification (Fever), ReAct overcomes issues of hallucination and error propagation prevalent in chain-of-thought reasoning by interacting with a simple Wikipedia API, and generates human-like task-solving trajectories that are more interpretable than baselines without reasoning traces. On two interactive decision making benchmarks (ALFWorld and WebShop), ReAct outperforms imitation and reinforcement learning methods by an absolute success rate of 34% and 10% respectively, while being prompted with only one or two in-context examples. Project site with code: https://react-lm.github.io"
                },
                {
                    "title": "Decomposed Prompting: A Modular Approach for Solving Complex Tasks",
                    "abstract": "Few-shot prompting is a surprisingly powerful way to use Large Language Models (LLMs) to solve various tasks. However, this approach struggles as the task complexity increases or when the individual reasoning steps of the task themselves are hard to learn, especially when embedded in more complex tasks. To address this, we propose Decomposed Prompting, a new approach to solve complex tasks by decomposing them (via prompting) into simpler sub-tasks that can be delegated to a library of prompting-based LLMs dedicated to these sub-tasks. This modular structure allows each prompt to be optimized for its specific sub-task, further decomposed if necessary, and even easily replaced with more effective prompts, trained models, or symbolic functions if desired. We show that the flexibility and modularity of Decomposed Prompting allows it to outperform prior work on few-shot prompting using GPT3. On symbolic reasoning tasks, we can further decompose sub-tasks that are hard for LLMs into even simpler solvable sub-tasks. When the complexity comes from the input length, we can recursively decompose the task into the same task but with smaller inputs. We also evaluate our approach on textual multi-step reasoning tasks: on long-context multi-hop QA task, we can more effectively teach the sub-tasks via our separate sub-tasks prompts; and on open-domain multi-hop QA, we can incorporate a symbolic information retrieval within our decomposition framework, leading to improved performance on both tasks. Datasets, Code and Prompts available at https://github.com/allenai/DecomP."
                },
                {
                    "title": "MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledge",
                    "abstract": "Autonomous agents have made great strides in specialist domains like Atari games and Go. However, they typically learn tabula rasa in isolated environments with limited and manually conceived objectives, thus failing to generalize across a wide spectrum of tasks and capabilities. Inspired by how humans continually learn and adapt in the open world, we advocate a trinity of ingredients for building generalist agents: 1) an environment that supports a multitude of tasks and goals, 2) a large-scale database of multimodal knowledge, and 3) a flexible and scalable agent architecture. We introduce MineDojo, a new framework built on the popular Minecraft game that features a simulation suite with thousands of diverse open-ended tasks and an internet-scale knowledge base with Minecraft videos, tutorials, wiki pages, and forum discussions. Using MineDojo's data, we propose a novel agent learning algorithm that leverages large pre-trained video-language models as a learned reward function. Our agent is able to solve a variety of open-ended tasks specified in free-form language without any manually designed dense shaping reward. We open-source the simulation suite, knowledge bases, algorithm implementation, and pretrained models (https://minedojo.org) to promote research towards the goal of generally capable embodied agents."
                },
                {
                    "title": "ScienceWorld: Is your Agent Smarter than a 5th Grader?",
                    "abstract": "We present ScienceWorld, a benchmark to test agents\u2019 scientific reasoning abilities in a new interactive text environment at the level of a standard elementary school science curriculum. Despite the transformer-based progress seen in question-answering and scientific text processing, we find that current models cannot reason about or explain learned science concepts in novel contexts. For instance, models can easily answer what the conductivity of a known material is but struggle when asked how they would conduct an experiment in a grounded environment to find the conductivity of an unknown material. This begs the question of whether current models are simply retrieving answers by way of seeing a large number of similar examples or if they have learned to reason about concepts in a reusable manner. We hypothesize that agents need to be grounded in interactive environments to achieve such reasoning capabilities. Our experiments provide empirical evidence supporting this hypothesis \u2013 showing that a 1.5 million parameter agent trained interactively for 100k steps outperforms a 11 billion parameter model statically trained for scientific question-answering and reasoning from millions of expert demonstrations."
                },
                {
                    "title": "MiniHack the Planet: A Sandbox for Open-Ended Reinforcement Learning Research",
                    "abstract": "Progress in deep reinforcement learning (RL) is heavily driven by the availability of challenging benchmarks used for training agents. However, benchmarks that are widely adopted by the community are not explicitly designed for evaluating specific capabilities of RL methods. While there exist environments for assessing particular open problems in RL (such as exploration, transfer learning, unsupervised environment design, or even language-assisted RL), it is generally difficult to extend these to richer, more complex environments once research goes beyond proof-of-concept results. We present MiniHack, a powerful sandbox framework for easily designing novel RL environments. MiniHack is a one-stop shop for RL experiments with environments ranging from small rooms to complex, procedurally generated worlds. By leveraging the full set of entities and environment dynamics from NetHack, one of the richest grid-based video games, MiniHack allows designing custom RL testbeds that are fast and convenient to use. With this sandbox framework, novel environments can be designed easily, either using a human-readable description language or a simple Python interface. In addition to a variety of RL tasks and baselines, MiniHack can wrap existing RL benchmarks and provide ways to seamlessly add additional complexity."
                },
                {
                    "title": "ALFWorld: Aligning Text and Embodied Environments for Interactive Learning",
                    "abstract": "Given a simple request (e.g., Put a washed apple in the kitchen fridge), humans can reason in purely abstract terms by imagining action sequences and scoring their likelihood of success, prototypicality, and efficiency, all without moving a muscle. Once we see the kitchen in question, we can update our abstract plans to fit the scene. Embodied agents require the same abilities, but existing work does not yet provide the infrastructure necessary for both reasoning abstractly and executing concretely. We address this limitation by introducing ALFWorld, a simulator that enables agents to learn abstract, text-based policies in TextWorld (Cote et al., 2018) and then execute goals from the ALFRED benchmark (Shridhar et al., 2020) in a rich visual environment. ALFWorld enables the creation of a new BUTLER agent whose abstract knowledge, learned in TextWorld, corresponds directly to concrete, visually grounded actions. In turn, as we demonstrate empirically, this fosters better agent generalization than training only in the visually grounded environment. BUTLER's simple, modular design factors the problem to allow researchers to focus on models for improving every piece of the pipeline (language understanding, planning, navigation, visual scene understanding, and so forth)."
                },
                {
                    "title": "The NetHack Learning Environment",
                    "abstract": "Progress in Reinforcement Learning (RL) algorithms goes hand-in-hand with the development of challenging environments that test the limits of current methods. While existing RL environments are either sufficiently complex or based on fast simulation, they are rarely both. Here, we present the NetHack Learning Environment (NLE), a scalable, procedurally generated, stochastic, rich, and challenging environment for RL research based on the popular single-player terminal-based roguelike game, NetHack. We argue that NetHack is sufficiently complex to drive long-term research on problems such as exploration, planning, skill acquisition, and language-conditioned RL, while dramatically reducing the computational resources required to gather a large amount of experience. We compare NLE and its task suite to existing alternatives, and discuss why it is an ideal medium for testing the robustness and systematic generalization of RL agents. We demonstrate empirical success for early stages of the game using a distributed Deep RL baseline and Random Network Distillation exploration, alongside qualitative analysis of various agents trained in the environment. NLE is open source at this https URL."
                },
                {
                    "title": "ALFRED: A Benchmark for Interpreting Grounded Instructions for Everyday Tasks",
                    "abstract": "We present ALFRED (Action Learning From Realistic Environments and Directives), a benchmark for learning a mapping from natural language instructions and egocentric vision to sequences of actions for household tasks. ALFRED includes long, compositional tasks with non-reversible state changes to shrink the gap between research benchmarks and real-world applications. ALFRED consists of expert demonstrations in interactive visual environments for 25k natural language directives. These directives contain both high-level goals like \u201cRinse off a mug and place it in the coffee maker.\u201d and low-level language instructions like \u201cWalk to the coffee maker on the right.\u201d ALFRED tasks are more complex in terms of sequence length, action space, and language than existing vision- and-language task datasets. We show that a baseline model based on recent embodied vision-and-language tasks performs poorly on ALFRED, suggesting that there is significant room for developing innovative grounded visual language understanding models with this benchmark."
                },
                {
                    "title": "Interactive Fiction Games: A Colossal Adventure",
                    "abstract": "A hallmark of human intelligence is the ability to understand and communicate with language. Interactive Fiction games are fully text-based simulation environments where a player issues text commands to effect change in the environment and progress through the story. We argue that IF games are an excellent testbed for studying language-based autonomous agents. In particular, IF games combine challenges of combinatorial action spaces, language understanding, and commonsense reasoning. To facilitate rapid development of language-based agents, we introduce Jericho, a learning environment for man-made IF games and conduct a comprehensive study of text-agents across a rich set of games, highlighting directions in which agents can improve."
                },
                {
                    "title": "Machine Learning-Assisted Discovery of Solid Li-Ion Conducting Materials",
                    "abstract": "We discover many new crystalline solid materials with fast single crystal Li ion conductivity at room temperature, discovered through density functional theory simulations guided by machine learning-based methods. The discovery of new solid Li superionic conductors is of critical importance to the development of safe all-solid-state Li-ion batteries. With a predictive universal structure\u2013property relationship for fast ion conduction not well understood, the search for new solid Li ion conductors has relied largely on trial-and-error computational and experimental searches over the last several decades. In this work, we perform a guided search of materials space with a machine learning (ML)-based prediction model for material selection and density functional theory molecular dynamics (DFT-MD) simulations for calculating ionic conductivity. These materials are screened from over 12\u202f000 experimentally synthesized and characterized candidates with very diverse structures and compositions. When compared to a r..."
                },
                {
                    "title": "AI2-THOR: An Interactive 3D Environment for Visual AI",
                    "abstract": "We introduce The House Of inteRactions (THOR), a framework for visual AI research, available at this http URL AI2-THOR consists of near photo-realistic 3D indoor scenes, where AI agents can navigate in the scenes and interact with objects to perform tasks. AI2-THOR enables research in many different domains including but not limited to deep reinforcement learning, imitation learning, learning by interaction, planning, visual question answering, unsupervised representation learning, object detection and segmentation, and learning models of cognition. The goal of AI2-THOR is to facilitate building visually intelligent models and push the research forward in this domain."
                },
                {
                    "title": "Artificial Intelligence to Win the Nobel Prize and Beyond: Creating the Engine for Scientific Discovery",
                    "abstract": "This article proposes a new grand challenge for AI reasearch: to develop AI system to make major scientific discoveries in biomedical sciences that worth Nobel Prize. There are a series of human cognitive limitations that prevents us from making accerlated scientific discoveries, particularity in biomedical sciences. As a result, scientific discoveries are left behind at the level of cottage industry. AI systems can transform scientific discoveries into highly efficient practice, thereby enable us to expand our knowledge in unprecedented way. Such system may out-compute all possible hypotheses and may redefine the nature of scientific intuition, hence scientific discovery process."
                },
                {
                    "title": "Cheaper faster drug development validated by the repositioning of drugs against neglected tropical diseases",
                    "abstract": "There is an urgent need to make drug discovery cheaper and faster. This will enable the development of treatments for diseases currently neglected for economic reasons, such as tropical and orphan diseases, and generally increase the supply of new drugs. Here, we report the Robot Scientist \u2018Eve\u2019 designed to make drug discovery more economical. A Robot Scientist is a laboratory automation system that uses artificial intelligence (AI) techniques to discover scientific knowledge through cycles of experimentation. Eve integrates and automates library-screening, hit-confirmation, and lead generation through cycles of quantitative structure activity relationship learning and testing. Using econometric modelling we demonstrate that the use of AI to select compounds economically outperforms standard drug screening. For further efficiency Eve uses a standardized form of assay to compute Boolean functions of compound properties. These assays can be quickly and cheaply engineered using synthetic biology, enabling more targets to be assayed for a given budget. Eve has repositioned several drugs against specific targets in parasites that cause tropical diseases. One validated discovery is that the anti-cancer compound TNP-470 is a potent inhibitor of dihydrofolate reductase from the malaria-causing parasite Plasmodium vivax."
                },
                {
                    "title": "Hypothesis",
                    "abstract": "A hypothesis is proposed on a potential role of protein dielectricity as an unfolding factor in protein-protein interactions. It is suggested that large protein complexes and aggregation seeds can unfold target proteins by virtue of their effect on the dielectric properties of water at the protein-solvent interface. Here, similar to the effect of membrane surfaces, protein surface can cause decrease in the local dielectric constant of solvent and thereby induce structural changes in a target protein approaching this surface. Some potential implementations of this hypothetical mechanism are also discussed."
                },
                {
                    "title": "Scientific Discovery: Computational Explorations of the Creative Processes",
                    "abstract": "Scientific Discovery: Computational Explorations of the Creative Processes. By Pat Langley, Herbert A. Simon, Gary L. Bradshaw and Jan M. Zytkow. Cambridge Mass., USA, The MIT Press, USD 13.50."
                },
                {
                    "title": "Alchemy: A structured task distribution for meta-reinforcement learning",
                    "abstract": ","
                }
            ],
            "categories": [
                "cs.AI",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we create a virtual discovery environment that enables AI agents to perform end-to-end scientific discovery, including ideation, hypothesis formation, experiment design, and analysis across diverse scientific topics?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem has significant implications for the research community as it could lead to the development of AI systems capable of conducting comprehensive scientific research autonomously. This advancement could transform how scientific inquiries are approached, potentially accelerating discoveries in various fields such as chemistry, genetics, and material science. By enabling AI to engage in the full scientific process, we could unlock new methodologies for hypothesis generation and experimental design, leading to practical applications in drug discovery, environmental science, and beyond. Furthermore, this research could inspire future studies on general-purpose AI systems that can adapt to various scientific challenges, fostering interdisciplinary collaboration and innovation.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of replicating the multifaceted nature of scientific discovery in a virtual environment. Naive approaches may fail because they often focus on isolated tasks without considering the interconnectedness of ideation, experimentation, and analysis. Additionally, creating a realistic yet simplified environment that allows for meaningful interactions and observations requires sophisticated modeling of scientific principles and commonsense knowledge. Technical obstacles include designing tasks that are both challenging and educational, ensuring that agents can navigate and manipulate the environment effectively, and developing robust evaluation metrics to assess the agents' performance in a comprehensive manner.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on narrow aspects of scientific discovery, such as hypothesis testing or data analysis, without addressing the full spectrum of the scientific process. Existing solutions may lack the necessary complexity or realism to facilitate genuine discovery, often resulting in task-specific agents that do not generalize well to new challenges. Barriers such as the high cost of real-world experimentation and the difficulty of creating a comprehensive simulation environment have also hindered progress. Our approach differs by providing a text-based simulated world, DiscoveryWorld, that emphasizes long-horizon tasks requiring a complete discovery process, thus encouraging the development of general discovery skills rather than task-specific solutions.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing DiscoveryWorld, a text-based simulated environment where AI agents can engage in scientific discovery tasks across eight diverse topics. The agents"
            }
        },
        "author_data": {
            "fa015642-932d-4b9d-9406-8c6230f22f2b": {
                "pk": "fa015642-932d-4b9d-9406-8c6230f22f2b",
                "name": "Peter Jansen",
                "collaborators": [
                    "Ruoyao Wang",
                    "Marc-Alexandre C\u00f4t\u00e9",
                    "Peter Clark",
                    "Prithviraj Ammanabrolu",
                    "Jordan Boyd-Graber",
                    "Hannah Smith",
                    "Zeyu Zhang",
                    "John Culnan",
                    "Dongfang Xu",
                    "Jaycie Martin"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Question Answering",
                    "Machine Learning",
                    "Interactive Agents"
                ],
                "publications": [
                    {
                        "title": "Multi-hop Inference for Sentence-level TextGraphs: How Challenging is Meaningfully Combining Information for Science Question Answering?",
                        "abstract": "Question Answering for complex questions is often modeled as a graph construction or traversal task, where a solver must build or traverse a graph of facts that answer and explain a given question. This \"multi-hop\" inference has been shown to be extremely challenging, with few models able to aggregate more than two facts before being overwhelmed by \"semantic drift\", or the tendency for long chains of facts to quickly drift off topic. This is a major barrier to current inference models, as even elementary science questions require an average of 4 to 6 facts to answer and explain. In this work we empirically characterize the difficulty of building or traversing a graph of sentences connected by lexical overlap, by evaluating chance sentence aggregation quality through 9,784 manually-annotated judgments across knowledge graphs built from three free-text corpora (including study guides and Simple Wikipedia). We demonstrate semantic drift tends to be high and aggregation quality low, at between 0.04% and 3%, and highlight scenarios that maximize the likelihood of meaningfully combining information."
                    },
                    {
                        "title": "From Words to Wires: Generating Functioning Electronic Devices from Natural Language Descriptions",
                        "abstract": "In this work, we show that contemporary language models have a previously unknown skill -- the capacity for electronic circuit design from high-level textual descriptions, akin to code generation. We introduce two benchmarks: Pins100, assessing model knowledge of electrical components, and Micro25, evaluating a model's capability to design common microcontroller circuits and code in the Arduino ecosystem that involve input, output, sensors, motors, protocols, and logic -- with models such as GPT-4 and Claude-V1 achieving between 60% to 96% Pass@1 on generating full devices. We include six case studies of using language models as a design assistant for moderately complex devices, such as a radiation-powered random number generator, an emoji keyboard, a visible spectrometer, and several assistive devices, while offering a qualitative analysis performance, outlining evaluation challenges, and suggesting areas of development to improve complex circuit design and practical utility. With this work, we aim to spur research at the juncture of natural language processing and electronic design."
                    },
                    {
                        "title": "Picard understanding Darmok: A Dataset and Model for Metaphor-Rich Translation in a Constructed Language",
                        "abstract": "Tamarian, a fictional language introduced in the Star Trek episode Darmok, communicates meaning through utterances of metaphorical references, such as \"Darmok and Jalad at Tanagra\" instead of \"We should work together.\" This work assembles a Tamarian-English dictionary of utterances from the original episode and several follow-on novels, and uses this to construct a parallel corpus of 456 English-Tamarian utterances. A machine translation system based on a large language model (T5) is trained using this parallel corpus, and is shown to produce an accuracy of 76% when translating from English to Tamarian on known utterances."
                    },
                    {
                        "title": "Self-Supervised Behavior Cloned Transformers are Path Crawlers for Text Games",
                        "abstract": "In this work, we introduce a self-supervised behavior cloning transformer for text games, which are challenging benchmarks for multi-step reasoning in virtual environments. Traditionally, Behavior Cloning Transformers excel in such tasks but rely on supervised training data. Our approach auto-generates training data by exploring trajectories (defined by common macro-action sequences) that lead to reward within the games, while determining the generality and utility of these trajectories by rapidly training small models then evaluating their performance on unseen development games. Through empirical analysis, we show our method consistently uncovers generalizable training data, achieving about 90\\% performance of supervised systems across three benchmark text games."
                    },
                    {
                        "title": "ScienceExamCER: A High-Density Fine-Grained Science-Domain Corpus for Common Entity Recognition",
                        "abstract": "Named entity recognition identifies common classes of entities in text, but these entity labels are generally sparse, limiting utility to downstream tasks. In this work we present ScienceExamCER, a densely-labeled semantic classification corpus of 133k mentions in the science exam domain where nearly all (96%) of content words have been annotated with one or more fine-grained semantic class labels including taxonomic groups, meronym groups, verb/action groups, properties and values, and synonyms. Semantic class labels are drawn from a manually-constructed fine-grained typology of 601 classes generated through a data-driven analysis of 4,239 science exam questions. We show an off-the-shelf BERT-based named entity recognition model modified for multi-label classification achieves an accuracy of 0.85 F1 on this task, suggesting strong utility for downstream tasks in science domain question answering requiring densely-labeled semantic classification."
                    },
                    {
                        "title": "Multi-class Hierarchical Question Classification for Multiple Choice Science Exams",
                        "abstract": "Prior work has demonstrated that question classification (QC), recognizing the problem domain of a question, can help answer it more accurately. However, developing strong QC algorithms has been hindered by the limited size and complexity of annotated data available. To address this, we present the largest challenge dataset for QC, containing 7,787 science exam questions paired with detailed classification labels from a fine-grained hierarchical taxonomy of 406 problem domains. We then show that a BERT-based model trained on this dataset achieves a large (+0.12 MAP) gain compared with previous methods, while also achieving state-of-the-art performance on benchmark open-domain and biomedical QC datasets. Finally, we show that using this model's predictions of question topic significantly improves the accuracy of a question answering system by +1.7% P@1, with substantial future gains possible as QC performance improves."
                    },
                    {
                        "title": "On the Challenges of Evaluating Compositional Explanations in Multi-Hop Inference: Relevance, Completeness, and Expert Ratings",
                        "abstract": "Building compositional explanations requires models to combine two or more facts that, together, describe why the answer to a question is correct. Typically, these \"multi-hop\" explanations are evaluated relative to one (or a small number of) gold explanations. In this work, we show these evaluations substantially underestimate model performance, both in terms of the relevance of included facts, as well as the completeness of model-generated explanations, because models regularly discover and produce valid explanations that are different than gold explanations. To address this, we construct a large corpus of 126k domain-expert (science teacher) relevance ratings that augment a corpus of explanations to standardized science exam questions, discovering 80k additional relevant facts not rated as gold. We build three strong models based on different methodologies (generation, ranking, and schemas), and empirically show that while expert-augmented ratings provide better estimates of explanation quality, both original (gold) and expert-augmented automatic evaluations still substantially underestimate performance by up to 36% when compared with full manual expert judgements, with different models being disproportionately affected. This poses a significant methodological challenge to accurately evaluating explanations produced by compositional reasoning models."
                    },
                    {
                        "title": "ScienceWorld: Is your Agent Smarter than a 5th Grader?",
                        "abstract": "We present ScienceWorld, a benchmark to test agents' scientific reasoning abilities in a new interactive text environment at the level of a standard elementary school science curriculum. Despite the transformer-based progress seen in question-answering and scientific text processing, we find that current models cannot reason about or explain learned science concepts in novel contexts. For instance, models can easily answer what the conductivity of a known material is but struggle when asked how they would conduct an experiment in a grounded environment to find the conductivity of an unknown material. This begs the question of whether current models are simply retrieving answers by way of seeing a large number of similar examples or if they have learned to reason about concepts in a reusable manner. We hypothesize that agents need to be grounded in interactive environments to achieve such reasoning capabilities. Our experiments provide empirical evidence supporting this hypothesis -- showing that a 1.5 million parameter agent trained interactively for 100k steps outperforms a 11 billion parameter model statically trained for scientific question-answering and reasoning from millions of expert demonstrations."
                    },
                    {
                        "title": "Behavior Cloned Transformers are Neurosymbolic Reasoners",
                        "abstract": "In this work, we explore techniques for augmenting interactive agents with information from symbolic modules, much like humans use tools like calculators and GPS systems to assist with arithmetic and navigation. We test our agent's abilities in text games -- challenging benchmarks for evaluating the multi-step reasoning abilities of game agents in grounded, language-based environments. Our experimental study indicates that injecting the actions from these symbolic modules into the action space of a behavior cloned transformer agent increases performance on four text game benchmarks that test arithmetic, navigation, sorting, and common sense reasoning by an average of 22%, allowing an agent to reach the highest possible performance on unseen games. This action injection technique is easily extended to new agents, environments, and symbolic modules."
                    },
                    {
                        "title": "PDDLEGO: Iterative Planning in Textual Environments",
                        "abstract": "Planning in textual environments have been shown to be a long-standing challenge even for current models. A recent, promising line of work uses LLMs to generate a formal representation of the environment that can be solved by a symbolic planner. However, existing methods rely on a fully-observed environment where all entity states are initially known, so a one-off representation can be constructed, leading to a complete plan. In contrast, we tackle partially-observed environments where there is initially no sufficient information to plan for the end-goal. We propose PDDLEGO that iteratively construct a planning representation that can lead to a partial plan for a given sub-goal. By accomplishing the sub-goal, more information is acquired to augment the representation, eventually achieving the end-goal. We show that plans produced by few-shot PDDLEGO are 43% more efficient than generating plans end-to-end on the Coin Collector simulation, with strong performance (98%) on the more complex Cooking World simulation where end-to-end LLMs fail to generate coherent plans (4%)."
                    },
                    {
                        "title": "ByteSized32: A Corpus and Challenge Task for Generating Task-Specific World Models Expressed as Text Games",
                        "abstract": "In this work, we investigate the capacity of language models to generate explicit, interpretable, and interactive world models of scientific and common-sense reasoning tasks. We operationalize this as a task of generating text games, expressed as hundreds of lines of Python code. To facilitate this task, we introduce ByteSized32 (Code: github.com/cognitiveailab/BYTESIZED32), a corpus of 32 reasoning-focused text games totaling 20k lines of Python code. We empirically demonstrate that GPT-4 can use these games as templates for single-shot in-context learning, successfully producing runnable games on unseen topics in 28% of cases. When allowed to self-reflect on program errors, game runnability substantially increases to 57%. While evaluating simulation fidelity is labor-intensive, we introduce a suite of automated metrics to assess game fidelity, technical validity, adherence to task specifications, and winnability, showing a high degree of agreement with expert human ratings. We pose this as a challenge task to spur further development at the juncture of world modeling and code generation."
                    }
                ]
            },
            "e54c367c-1266-45b4-aaca-c03cd0f71442": {
                "pk": "e54c367c-1266-45b4-aaca-c03cd0f71442",
                "name": "Marc-Alexandre C\u00f4t\u00e9",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "e3717e72-441c-40b3-b20a-755629626912": {
                "pk": "e3717e72-441c-40b3-b20a-755629626912",
                "name": "Tushar Khot",
                "collaborators": [
                    "Ashish Sabharwal",
                    "Peter Clark",
                    "Harsh Trivedi",
                    "Niranjan Balasubramanian",
                    "Daniel Khashabi",
                    "Dongyeop Kang",
                    "Souvik Kundu",
                    "Shih-Ting Lin",
                    "Kyle Richardson",
                    "Todor Mihaylov"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Question Answering",
                    "Multi-hop Reasoning",
                    "Knowledge Representation"
                ],
                "publications": [
                    {
                        "title": "Answering Complex Questions Using Open Information Extraction",
                        "abstract": "While there has been substantial progress in factoid question-answering (QA), answering complex questions remains challenging, typically requiring both a large body of knowledge and inference techniques. Open Information Extraction (Open IE) provides a way to generate semi-structured knowledge for QA, but to date such knowledge has only been used to answer simple questions with retrieval-based methods. We overcome this limitation by presenting a method for reasoning with Open IE knowledge, allowing more complex questions to be handled. Using a recently proposed support graph optimization framework for QA, we develop a new inference model for Open IE, in particular one that can work effectively with multiple short facts, noise, and the relational structure of tuples. Our model significantly outperforms a state-of-the-art structured solver on complex questions of varying difficulty, while also removing the reliance on manually curated knowledge."
                    },
                    {
                        "title": "What's Missing: A Knowledge Gap Guided Approach for Multi-hop Question Answering",
                        "abstract": "Multi-hop textual question answering requires combining information from multiple sentences. We focus on a natural setting where, unlike typical reading comprehension, only partial information is provided with each question. The model must retrieve and use additional knowledge to correctly answer the question. To tackle this challenge, we develop a novel approach that explicitly identifies the knowledge gap between a key span in the provided knowledge and the answer choices. The model, GapQA, learns to fill this gap by determining the relationship between the span and an answer choice, based on retrieved knowledge targeting this gap. We propose jointly training a model to simultaneously fill this knowledge gap and compose it with the provided partial knowledge. On the OpenBookQA dataset, given partial knowledge, explicitly identifying what's missing substantially outperforms previous approaches."
                    },
                    {
                        "title": "Exploiting Explicit Paths for Multi-hop Reading Comprehension",
                        "abstract": "We propose a novel, path-based reasoning approach for the multi-hop reading comprehension task where a system needs to combine facts from multiple passages to answer a question. Although inspired by multi-hop reasoning over knowledge graphs, our proposed approach operates directly over unstructured text. It generates potential paths through passages and scores them without any direct path supervision. The proposed model, named PathNet, attempts to extract implicit relations from text through entity pair representations, and compose them to encode each path. To capture additional context, PathNet also composes the passage representations along each path to compute a passage-based representation. Unlike previous approaches, our model is then able to explain its reasoning via these explicit paths through the passages. We show that our approach outperforms prior models on the multi-hop Wikihop dataset, and also can be generalized to apply to the OpenBookQA dataset, matching state-of-the-art performance."
                    },
                    {
                        "title": "ReadOnce Transformers: Reusable Representations of Text for Transformers",
                        "abstract": "We present ReadOnce Transformers, an approach to convert a transformer-based model into one that can build an information-capturing, task-independent, and compressed representation of text. The resulting representation is reusable across different examples and tasks, thereby requiring a document shared across many examples or tasks to only be \\emph{read once}. This leads to faster training and evaluation of models. Additionally, we extend standard text-to-text transformer models to Representation+Text-to-text models, and evaluate on multiple downstream tasks: multi-hop QA, abstractive QA, and long-document summarization. Our one-time computed representation results in a 2x-5x speedup compared to standard text-to-text models, while the compression also allows existing language models to handle longer documents without the need for designing new pre-trained models."
                    },
                    {
                        "title": "Hey AI, Can You Solve Complex Tasks by Talking to Agents?",
                        "abstract": "Training giant models from scratch for each complex task is resource- and data-inefficient. To help develop models that can leverage existing systems, we propose a new challenge: Learning to solve complex tasks by communicating with existing agents (or models) in natural language. We design a synthetic benchmark, CommaQA, with three complex reasoning tasks (explicit, implicit, numeric) designed to be solved by communicating with existing QA agents. For instance, using text and table QA agents to answer questions such as \"Who had the longest javelin throw from USA?\". We show that black-box models struggle to learn this task from scratch (accuracy under 50\\%) even with access to each agent's knowledge and gold facts supervision. In contrast, models that learn to communicate with agents outperform black-box models, reaching scores of 100\\% when given gold decomposition supervision. However, we show that the challenge of learning to solve complex tasks by communicating with existing agents \\emph{without relying on any auxiliary supervision or data} still remains highly elusive. We release CommaQA, along with a compositional generalization test split, to advance research in this direction. Dataset and Code available at https://github.com/allenai/commaqa."
                    },
                    {
                        "title": "Bridging Knowledge Gaps in Neural Entailment via Symbolic Models",
                        "abstract": "Most textual entailment models focus on lexical gaps between the premise text and the hypothesis, but rarely on knowledge gaps. We focus on filling these knowledge gaps in the Science Entailment task, by leveraging an external structured knowledge base (KB) of science facts. Our new architecture combines standard neural entailment models with a knowledge lookup module. To facilitate this lookup, we propose a fact-level decomposition of the hypothesis, and verifying the resulting sub-facts against both the textual premise and the structured KB. Our model, NSnet, learns to aggregate predictions from these heterogeneous data formats. On the SciTail dataset, NSnet outperforms a simpler combination of the two predictions by 3% and the base entailment model by 5%."
                    },
                    {
                        "title": "Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering",
                        "abstract": "We present a new kind of question answering dataset, OpenBookQA, modeled after open book exams for assessing human understanding of a subject. The open book that comes with our questions is a set of 1329 elementary level science facts. Roughly 6000 questions probe an understanding of these facts and their application to novel situations. This requires combining an open book fact (e.g., metals conduct electricity) with broad common knowledge (e.g., a suit of armor is made of metal) obtained from other sources. While existing QA datasets over documents or knowledge bases, being generally self-contained, focus on linguistic understanding, OpenBookQA probes a deeper understanding of both the topic---in the context of common knowledge---and the language it is expressed in. Human performance on OpenBookQA is close to 92%, but many state-of-the-art pre-trained QA methods perform surprisingly poorly, worse than several simple neural baselines we develop. Our oracle experiments designed to circumvent the knowledge retrieval bottleneck demonstrate the value of both the open book and additional facts. We leave it as a challenge to solve the retrieval problem in this multi-hop setting and to close the large gap to human performance."
                    },
                    {
                        "title": "Question Answering as Global Reasoning over Semantic Abstractions",
                        "abstract": "We propose a novel method for exploiting the semantic structure of text to answer multiple-choice questions. The approach is especially suitable for domains that require reasoning over a diverse set of linguistic constructs but have limited training data. To address these challenges, we present the first system, to the best of our knowledge, that reasons over a wide range of semantic abstractions of the text, which are derived using off-the-shelf, general-purpose, pre-trained natural language modules such as semantic role labelers, coreference resolvers, and dependency parsers. Representing multiple abstractions as a family of graphs, we translate question answering (QA) into a search for an optimal subgraph that satisfies certain global and local properties. This formulation generalizes several prior structured QA systems. Our system, SEMANTICILP, demonstrates strong performance on two domains simultaneously. In particular, on a collection of challenging science QA datasets, it outperforms various state-of-the-art approaches, including neural models, broad coverage information retrieval, and specialized techniques using structured knowledge bases, by 2%-6%."
                    },
                    {
                        "title": "More Bang for Your Buck: Natural Perturbation for Robust Question Answering",
                        "abstract": "While recent models have achieved human-level scores on many NLP datasets, we observe that they are considerably sensitive to small changes in input. As an alternative to the standard approach of addressing this issue by constructing training sets of completely new examples, we propose doing so via minimal perturbation of examples. Specifically, our approach involves first collecting a set of seed examples and then applying human-driven natural perturbations (as opposed to rule-based machine perturbations), which often change the gold label as well. Local perturbations have the advantage of being relatively easier (and hence cheaper) to create than writing out completely new examples. To evaluate the impact of this phenomenon, we consider a recent question-answering dataset (BoolQ) and study the benefit of our approach as a function of the perturbation cost ratio, the relative cost of perturbing an existing question vs. creating a new one from scratch. We find that when natural perturbations are moderately cheaper to create, it is more effective to train models using them: such models exhibit higher robustness and better generalization, while retaining performance on the original BoolQ dataset."
                    },
                    {
                        "title": "A Simple Yet Strong Pipeline for HotpotQA",
                        "abstract": "State-of-the-art models for multi-hop question answering typically augment large-scale language models like BERT with additional, intuitively useful capabilities such as named entity recognition, graph-based reasoning, and question decomposition. However, does their strong performance on popular multi-hop datasets really justify this added design complexity? Our results suggest that the answer may be no, because even our simple pipeline based on BERT, named Quark, performs surprisingly well. Specifically, on HotpotQA, Quark outperforms these models on both question answering and support identification (and achieves performance very close to a RoBERTa model). Our pipeline has three steps: 1) use BERT to identify potentially relevant sentences independently of each other; 2) feed the set of selected sentences as context into a standard BERT span prediction model to choose an answer; and 3) use the sentence selection model, now with the chosen answer, to produce supporting sentences. The strong performance of Quark resurfaces the importance of carefully exploring simple model designs before using popular benchmarks to justify the value of complex techniques."
                    },
                    {
                        "title": "Is Multihop QA in DiRe Condition? Measuring and Reducing Disconnected Reasoning",
                        "abstract": "Has there been real progress in multi-hop question-answering? Models often exploit dataset artifacts to produce correct answers, without connecting information across multiple supporting facts. This limits our ability to measure true progress and defeats the purpose of building multi-hop QA datasets. We make three contributions towards addressing this. First, we formalize such undesirable behavior as disconnected reasoning across subsets of supporting facts. This allows developing a model-agnostic probe for measuring how much any model can cheat via disconnected reasoning. Second, using a notion of \\emph{contrastive support sufficiency}, we introduce an automatic transformation of existing datasets that reduces the amount of disconnected reasoning. Third, our experiments suggest that there hasn't been much progress in multi-hop QA in the reading comprehension setting. For a recent large-scale model (XLNet), we show that only 18 points out of its answer F1 score of 72 on HotpotQA are obtained through multifact reasoning, roughly the same as that of a simpler RNN baseline. Our transformation substantially reduces disconnected reasoning (19 points in answer F1). It is complementary to adversarial approaches, yielding further reductions in conjunction."
                    },
                    {
                        "title": "MuSiQue: Multihop Questions via Single-hop Question Composition",
                        "abstract": "Multihop reasoning remains an elusive goal as existing multihop benchmarks are known to be largely solvable via shortcuts. Can we create a question answering (QA) dataset that, by construction, \\emph{requires} proper multihop reasoning? To this end, we introduce a bottom-up approach that systematically selects composable pairs of single-hop questions that are connected, i.e., where one reasoning step critically relies on information from another. This bottom-up methodology lets us explore a vast space of questions and add stringent filters as well as other mechanisms targeting connected reasoning. It provides fine-grained control over the construction process and the properties of the resulting $k$-hop questions. We use this methodology to create MuSiQue-Ans, a new multihop QA dataset with 25K 2-4 hop questions. Relative to existing datasets, MuSiQue-Ans is more difficult overall (3x increase in human-machine gap), and harder to cheat via disconnected reasoning (e.g., a single-hop model has a 30 point drop in F1). We further add unanswerable contrast questions to produce a more stringent dataset, MuSiQue-Full. We hope our datasets will help the NLP community develop models that perform genuine multihop reasoning."
                    },
                    {
                        "title": "Teaching Broad Reasoning Skills for Multi-Step QA by Generating Hard Contexts",
                        "abstract": "Question-answering datasets require a broad set of reasoning skills. We show how to use question decompositions to teach language models these broad reasoning skills in a robust fashion. Specifically, we use widely available QDMR representations to programmatically create hard-to-cheat synthetic contexts for real questions in six multi-step reasoning datasets. These contexts are carefully designed to avoid reasoning shortcuts prevalent in real contexts that prevent models from learning the right skills. This results in a pretraining dataset, named TeaBReaC, containing 525K multi-step questions (with associated formal programs) covering about 900 reasoning patterns. We show that pretraining standard language models (LMs) on TeaBReaC before fine-tuning them on target datasets improves their performance by up to 13 F1 points across 4 multi-step QA datasets, with up to 21 point gain on more complex questions. The resulting models also demonstrate higher robustness, with a 5-8 F1 point improvement on two contrast sets. Furthermore, TeaBReaC pretraining substantially improves model performance and robustness even when starting with numerate LMs pretrained using recent methods (e.g., PReasM, POET). Our work thus shows how to effectively use decomposition-guided contexts to robustly teach multi-step reasoning."
                    },
                    {
                        "title": "Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions",
                        "abstract": "Prompting-based large language models (LLMs) are surprisingly powerful at generating natural language reasoning steps or Chains-of-Thoughts (CoT) for multi-step question answering (QA). They struggle, however, when the necessary knowledge is either unavailable to the LLM or not up-to-date within its parameters. While using the question to retrieve relevant text from an external knowledge source helps LLMs, we observe that this one-step retrieve-and-read approach is insufficient for multi-step QA. Here, \\textit{what to retrieve} depends on \\textit{what has already been derived}, which in turn may depend on \\textit{what was previously retrieved}. To address this, we propose IRCoT, a new approach for multi-step QA that interleaves retrieval with steps (sentences) in a CoT, guiding the retrieval with CoT and in turn using retrieved results to improve CoT. Using IRCoT with GPT3 substantially improves retrieval (up to 21 points) as well as downstream QA (up to 15 points) on four datasets: HotpotQA, 2WikiMultihopQA, MuSiQue, and IIRC. We observe similar substantial gains in out-of-distribution (OOD) settings as well as with much smaller models such as Flan-T5-large without additional training. IRCoT reduces model hallucination, resulting in factually more accurate CoT reasoning. Code, data, and prompts are available at \\url{https://github.com/stonybrooknlp/ircot}"
                    },
                    {
                        "title": "Husky: A Unified, Open-Source Language Agent for Multi-Step Reasoning",
                        "abstract": "Language agents perform complex tasks by using tools to execute each step precisely. However, most existing agents are based on proprietary models or designed to target specific tasks, such as mathematics or multi-hop question answering. We introduce Husky, a holistic, open-source language agent that learns to reason over a unified action space to address a diverse set of complex tasks involving numerical, tabular, and knowledge-based reasoning. Husky iterates between two stages: 1) generating the next action to take towards solving a given task and 2) executing the action using expert models and updating the current solution state. We identify a thorough ontology of actions for addressing complex tasks and curate high-quality data to train expert models for executing these actions. Our experiments show that Husky outperforms prior language agents across 14 evaluation datasets. Moreover, we introduce HuskyQA, a new evaluation set which stress tests language agents for mixed-tool reasoning, with a focus on retrieving missing knowledge and performing numerical reasoning. Despite using 7B models, Husky matches or even exceeds frontier LMs such as GPT-4 on these tasks, showcasing the efficacy of our holistic approach in addressing complex reasoning problems. Our code and models are available at https://github.com/agent-husky/Husky-v1."
                    },
                    {
                        "title": "Better Retrieval May Not Lead to Better Question Answering",
                        "abstract": "Considerable progress has been made recently in open-domain question answering (QA) problems, which require Information Retrieval (IR) and Reading Comprehension (RC). A popular approach to improve the system's performance is to improve the quality of the retrieved context from the IR stage. In this work we show that for StrategyQA, a challenging open-domain QA dataset that requires multi-hop reasoning, this common approach is surprisingly ineffective -- improving the quality of the retrieved context hardly improves the system's performance. We further analyze the system's behavior to identify potential reasons."
                    },
                    {
                        "title": "Improving Language Model Negotiation with Self-Play and In-Context Learning from AI Feedback",
                        "abstract": "We study whether multiple large language models (LLMs) can autonomously improve each other in a negotiation game by playing, reflecting, and criticizing. We are interested in this question because if LLMs were able to improve each other, it would imply the possibility of creating strong AI agents with minimal human intervention. We ask two LLMs to negotiate with each other, playing the roles of a buyer and a seller, respectively. They aim to reach a deal with the buyer targeting a lower price and the seller a higher one. A third language model, playing the critic, provides feedback to a player to improve the player's negotiation strategies. We let the two agents play multiple rounds, using previous negotiation history and AI feedback as in-context demonstrations to improve the model's negotiation strategy iteratively. We use different LLMs (GPT and Claude) for different roles and use the deal price as the evaluation metric. Our experiments reveal multiple intriguing findings: (1) Only a subset of the language models we consider can self-play and improve the deal price from AI feedback, weaker models either do not understand the game's rules or cannot incorporate AI feedback for further improvement. (2) Models' abilities to learn from the feedback differ when playing different roles. For example, it is harder for Claude-instant to improve as the buyer than as the seller. (3) When unrolling the game to multiple rounds, stronger agents can consistently improve their performance by meaningfully using previous experiences and iterative AI feedback, yet have a higher risk of breaking the deal. We hope our work provides insightful initial explorations of having models autonomously improve each other with game playing and AI feedback."
                    },
                    {
                        "title": "AdvEntuRe: Adversarial Training for Textual Entailment with Knowledge-Guided Examples",
                        "abstract": "We consider the problem of learning textual entailment models with limited supervision (5K-10K training examples), and present two complementary approaches for it. First, we propose knowledge-guided adversarial example generators for incorporating large lexical resources in entailment models via only a handful of rule templates. Second, to make the entailment model - a discriminator - more robust, we propose the first GAN-style approach for training it using a natural language example generator that iteratively adjusts based on the discriminator's performance. We demonstrate effectiveness using two entailment datasets, where the proposed methods increase accuracy by 4.7% on SciTail and by 2.8% on a 1% training sub-sample of SNLI. Notably, even a single hand-written rule, negate, improves the accuracy on the negation examples in SNLI by 6.1%."
                    },
                    {
                        "title": "Repurposing Entailment for Multi-Hop Question Answering Tasks",
                        "abstract": "Question Answering (QA) naturally reduces to an entailment problem, namely, verifying whether some text entails the answer to a question. However, for multi-hop QA tasks, which require reasoning with multiple sentences, it remains unclear how best to utilize entailment models pre-trained on large scale datasets such as SNLI, which are based on sentence pairs. We introduce Multee, a general architecture that can effectively use entailment models for multi-hop QA tasks. Multee uses (i) a local module that helps locate important sentences, thereby avoiding distracting information, and (ii) a global module that aggregates information by effectively incorporating importance weights. Importantly, we show that both modules can use entailment functions pre-trained on a large scale NLI datasets. We evaluate performance on MultiRC and OpenBookQA, two multihop QA datasets. When using an entailment function pre-trained on NLI datasets, Multee outperforms QA models trained only on the target QA datasets and the OpenAI transformer models. The code is available at https://github.com/StonyBrookNLP/multee."
                    },
                    {
                        "title": "UnQovering Stereotyping Biases via Underspecified Questions",
                        "abstract": "While language embeddings have been shown to have stereotyping biases, how these biases affect downstream question answering (QA) models remains unexplored. We present UNQOVER, a general framework to probe and quantify biases through underspecified questions. We show that a naive use of model scores can lead to incorrect bias estimates due to two forms of reasoning errors: positional dependence and question independence. We design a formalism that isolates the aforementioned errors. As case studies, we use this metric to analyze four important classes of stereotypes: gender, nationality, ethnicity, and religion. We probe five transformer-based QA models trained on two QA datasets, along with their underlying language models. Our broad study reveals that (1) all these models, with and without fine-tuning, have notable stereotyping biases in these classes; (2) larger models often have higher bias; and (3) the effect of fine-tuning on bias varies strongly with the dataset and the model size."
                    }
                ]
            },
            "de5c65c7-0a1f-4eec-bee4-fbf18052fb78": {
                "pk": "de5c65c7-0a1f-4eec-bee4-fbf18052fb78",
                "name": "Erin Bransom",
                "collaborators": [
                    "Bailey Kuehl",
                    "Aakanksha Naik",
                    "Doug Downey",
                    "Lucy Lu Wang",
                    "Tom Hope",
                    "Yuze Lou",
                    "Sergey Feldman",
                    "Mike D'Arcy",
                    "Alexis Ross",
                    "Jonathan Bragg"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Information Extraction",
                    "Automated Evaluation",
                    "Literature Review"
                ],
                "publications": [
                    {
                        "title": "CARE: Extracting Experimental Findings From Clinical Literature",
                        "abstract": "Extracting fine-grained experimental findings from literature can provide dramatic utility for scientific applications. Prior work has developed annotation schemas and datasets for limited aspects of this problem, failing to capture the real-world complexity and nuance required. Focusing on biomedicine, this work presents CARE -- a new IE dataset for the task of extracting clinical findings. We develop a new annotation schema capturing fine-grained findings as n-ary relations between entities and attributes, which unifies phenomena challenging for current IE systems such as discontinuous entity spans, nested relations, variable arity n-ary relations and numeric results in a single schema. We collect extensive annotations for 700 abstracts from two sources: clinical trials and case reports. We also demonstrate the generalizability of our schema to the computer science and materials science domains. We benchmark state-of-the-art IE systems on CARE, showing that even models such as GPT4 struggle. We release our resources to advance research on extracting and aggregating literature findings."
                    },
                    {
                        "title": "S2abEL: A Dataset for Entity Linking from Scientific Tables",
                        "abstract": "Entity linking (EL) is the task of linking a textual mention to its corresponding entry in a knowledge base, and is critical for many knowledge-intensive NLP applications. When applied to tables in scientific papers, EL is a step toward large-scale scientific knowledge bases that could enable advanced scientific question answering and analytics. We present the first dataset for EL in scientific tables. EL for scientific tables is especially challenging because scientific knowledge bases can be very incomplete, and disambiguating table mentions typically requires understanding the papers's tet in addition to the table. Our dataset, S2abEL, focuses on EL in machine learning results tables and includes hand-labeled cell types, attributed sources, and entity links from the PaperswithCode taxonomy for 8,429 cells from 732 tables. We introduce a neural baseline method designed for EL on scientific tables containing many out-of-knowledge-base mentions, and show that it significantly outperforms a state-of-the-art generic table EL method. The best baselines fall below human performance, and our analysis highlights avenues for improvement."
                    },
                    {
                        "title": "ARIES: A Corpus of Scientific Paper Edits Made in Response to Peer Reviews",
                        "abstract": "We introduce the task of automatically revising scientific papers based on peer feedback and release ARIES, a dataset of review comments and their corresponding paper edits. The data is drawn from real reviewer-author interactions from computer science, and we provide labels linking each reviewer comment to the specific paper edits made by the author in response. We automatically create a high-precision silver training set, as well as an expert-labeled test set that shows high inter-annotator agreement. In experiments with 10 models covering the state of the art, we find that they struggle even to identify which edits correspond to a comment -- especially when the relationship between the edit and the comment is indirect and requires reasoning to uncover. We also extensively analyze GPT-4's ability to generate edits given a comment and the original paper. We find that it often succeeds on a superficial level, but tends to rigidly follow the wording of the feedback rather than the underlying intent, and lacks technical details compared to human-written edits."
                    },
                    {
                        "title": "LongEval: Guidelines for Human Evaluation of Faithfulness in Long-form Summarization",
                        "abstract": "While human evaluation remains best practice for accurately judging the faithfulness of automatically-generated summaries, few solutions exist to address the increased difficulty and workload when evaluating long-form summaries. Through a survey of 162 papers on long-form summarization, we first shed light on current human evaluation practices surrounding long-form summaries. We find that 73% of these papers do not perform any human evaluation on model-generated summaries, while other works face new difficulties that manifest when dealing with long documents (e.g., low inter-annotator agreement). Motivated by our survey, we present LongEval, a set of guidelines for human evaluation of faithfulness in long-form summaries that addresses the following challenges: (1) How can we achieve high inter-annotator agreement on faithfulness scores? (2) How can we minimize annotator workload while maintaining accurate faithfulness scores? and (3) Do humans benefit from automated alignment between summary and source snippets? We deploy LongEval in annotation studies on two long-form summarization datasets in different domains (SQuALITY and PubMed), and we find that switching to a finer granularity of judgment (e.g., clause-level) reduces inter-annotator variance in faithfulness scores (e.g., std-dev from 18.5 to 6.8). We also show that scores from a partial annotation of fine-grained units highly correlates with scores from a full annotation workload (0.89 Kendall's tau using 50% judgments). We release our human judgments, annotation templates, and our software as a Python library for future research."
                    },
                    {
                        "title": "Personalized Jargon Identification for Enhanced Interdisciplinary Communication",
                        "abstract": "Scientific jargon can impede researchers when they read materials from other domains. Current methods of jargon identification mainly use corpus-level familiarity indicators (e.g., Simple Wikipedia represents plain language). However, researchers' familiarity of a term can vary greatly based on their own background. We collect a dataset of over 10K term familiarity annotations from 11 computer science researchers for terms drawn from 100 paper abstracts. Analysis of this data reveals that jargon familiarity and information needs vary widely across annotators, even within the same sub-domain (e.g., NLP). We investigate features representing individual, sub-domain, and domain knowledge to predict individual jargon familiarity. We compare supervised and prompt-based approaches, finding that prompt-based methods including personal publications yields the highest accuracy, though zero-shot prompting provides a strong baseline. This research offers insight into features and methods to integrate personal data into scientific jargon identification."
                    },
                    {
                        "title": "CHIME: LLM-Assisted Hierarchical Organization of Scientific Studies for Literature Review Support",
                        "abstract": "Literature review requires researchers to synthesize a large amount of information and is increasingly challenging as the scientific literature expands. In this work, we investigate the potential of LLMs for producing hierarchical organizations of scientific studies to assist researchers with literature review. We define hierarchical organizations as tree structures where nodes refer to topical categories and every node is linked to the studies assigned to that category. Our naive LLM-based pipeline for hierarchy generation from a set of studies produces promising yet imperfect hierarchies, motivating us to collect CHIME, an expert-curated dataset for this task focused on biomedicine. Given the challenging and time-consuming nature of building hierarchies from scratch, we use a human-in-the-loop process in which experts correct errors (both links between categories and study assignment) in LLM-generated hierarchies. CHIME contains 2,174 LLM-generated hierarchies covering 472 topics, and expert-corrected hierarchies for a subset of 100 topics. Expert corrections allow us to quantify LLM performance, and we find that while they are quite good at generating and organizing categories, their assignment of studies to categories could be improved. We attempt to train a corrector model with human feedback which improves study assignment by 12.6 F1 points. We release our dataset and models to encourage research on developing better assistive tools for literature review."
                    },
                    {
                        "title": "SUPER: Evaluating Agents on Setting Up and Executing Tasks from Research Repositories",
                        "abstract": "Given that Large Language Models (LLMs) have made significant progress in writing code, can they now be used to autonomously reproduce results from research repositories? Such a capability would be a boon to the research community, helping researchers validate, understand, and extend prior work. To advance towards this goal, we introduce SUPER, the first benchmark designed to evaluate the capability of LLMs in setting up and executing tasks from research repositories. SUPERaims to capture the realistic challenges faced by researchers working with Machine Learning (ML) and Natural Language Processing (NLP) research repositories. Our benchmark comprises three distinct problem sets: 45 end-to-end problems with annotated expert solutions, 152 sub problems derived from the expert set that focus on specific challenges (e.g., configuring a trainer), and 602 automatically generated problems for larger-scale development. We introduce various evaluation measures to assess both task success and progress, utilizing gold solutions when available or approximations otherwise. We show that state-of-the-art approaches struggle to solve these problems with the best model (GPT-4o) solving only 16.3% of the end-to-end set, and 46.1% of the scenarios. This illustrates the challenge of this task, and suggests that SUPER can serve as a valuable resource for the community to make and measure progress."
                    },
                    {
                        "title": "Automated Metrics for Medical Multi-Document Summarization Disagree with Human Evaluations",
                        "abstract": "Evaluating multi-document summarization (MDS) quality is difficult. This is especially true in the case of MDS for biomedical literature reviews, where models must synthesize contradicting evidence reported across different documents. Prior work has shown that rather than performing the task, models may exploit shortcuts that are difficult to detect using standard n-gram similarity metrics such as ROUGE. Better automated evaluation metrics are needed, but few resources exist to assess metrics when they are proposed. Therefore, we introduce a dataset of human-assessed summary quality facets and pairwise preferences to encourage and support the development of better automated evaluation methods for literature review MDS. We take advantage of community submissions to the Multi-document Summarization for Literature Review (MSLR) shared task to compile a diverse and representative sample of generated summaries. We analyze how automated summarization evaluation metrics correlate with lexical features of generated summaries, to other automated metrics including several we propose in this work, and to aspects of human-assessed summary quality. We find that not only do automated metrics fail to capture aspects of quality as assessed by humans, in many cases the system rankings produced by these metrics are anti-correlated with rankings according to human annotators."
                    }
                ]
            },
            "0aa59b3e-8c59-46a8-bb8b-a53a3fdaa982": {
                "pk": "0aa59b3e-8c59-46a8-bb8b-a53a3fdaa982",
                "name": "Bhavana Dalvi Mishra",
                "collaborators": [
                    "Peter Clark",
                    "Niket Tandon",
                    "Oyvind Tafjord",
                    "Antoine Bosselut",
                    "Wen-tau Yih",
                    "Yuling Gu",
                    "Keisuke Sakaguchi",
                    "Kyle Richardson",
                    "Bodhisattwa Prasad Majumder",
                    "Xinya Du"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Question Answering",
                    "Machine Learning",
                    "Memory-Augmented Learning"
                ],
                "publications": [
                    {
                        "title": "Towards Teachable Reasoning Systems: Using a Dynamic Memory of User Feedback for Continual System Improvement",
                        "abstract": "Our goal is a teachable reasoning system for question-answering (QA), where a user can interact with faithful answer explanations, and correct its errors so that the system improves over time. Our approach is to augment a QA model with a dynamic memory of user feedback, containing user-supplied corrections to erroneous model beliefs that users identify during interaction. Retrievals from memory are used as additional context for QA, to help avoid previous mistakes in similar new situations - a novel application of memory-based continuous learning. With simulated feedback, we find that our system (called TeachMe) continually improves with time, and without model retraining, requiring feedback on only 25% of training examples to reach within 1% of the upper-bound (feedback on all examples). Similarly, in experiments with real users, we observe a similar trend, with performance improving by over 15% on a hidden test set after teaching. This suggests new opportunities for using frozen language models in an interactive setting where users can inspect, debug, and correct the model's beliefs, leading to improved system's performance over time."
                    },
                    {
                        "title": "Entailer: Answering Questions with Faithful and Truthful Chains of Reasoning",
                        "abstract": "Our goal is a question-answering (QA) system that can show how its answers are implied by its own internal beliefs via a systematic chain of reasoning. Such a capability would allow better understanding of why a model produced the answer it did. Our approach is to recursively combine a trained backward-chaining model, capable of generating a set of premises entailing an answer hypothesis, with a verifier that checks that the model itself believes those premises (and the entailment itself) through self-querying. To our knowledge, this is the first system to generate multistep chains that are both faithful (the answer follows from the reasoning) and truthful (the chain reflects the system's own internal beliefs). In evaluation using two different datasets, users judge that a majority (70%+) of generated chains clearly show how an answer follows from a set of facts - substantially better than a high-performance baseline - while preserving answer accuracy. By materializing model beliefs that systematically support an answer, new opportunities arise for understanding the model's system of belief, and diagnosing and correcting its misunderstandings when an answer is wrong."
                    },
                    {
                        "title": "DREAM: Improving Situational QA by First Elaborating the Situation",
                        "abstract": "When people answer questions about a specific situation, e.g., \"I cheated on my mid-term exam last week. Was that wrong?\", cognitive science suggests that they form a mental picture of that situation before answering. While we do not know how language models (LMs) answer such questions, we conjecture that they may answer more accurately if they are also provided with additional details about the question situation, elaborating the \"scene\". To test this conjecture, we train a new model, DREAM, to answer questions that elaborate the scenes that situated questions are about, and then provide those elaborations as additional context to a question-answering (QA) model. We find that DREAM is able to create better scene elaborations (more accurate, useful, and consistent) than a representative state-of-the-art, zero-shot model (Macaw). We also find that using the scene elaborations as additional context improves the answer accuracy of a downstream QA system, including beyond that obtainable by simply further finetuning the QA system on DREAM's training data. These results suggest that adding focused elaborations about a situation can improve a system's reasoning about it, and may serve as an effective way of injecting new scenario based knowledge into QA models. Finally, our approach is dataset-neutral; we observe improved QA performance across different models, with even bigger gains on models with fewer parameters. We make our dataset and model publicly available at https://github.com/allenai/dream."
                    },
                    {
                        "title": "ProofWriter: Generating Implications, Proofs, and Abductive Statements over Natural Language",
                        "abstract": "Transformers have been shown to emulate logical deduction over natural language theories (logical rules expressed in natural language), reliably assigning true/false labels to candidate implications. However, their ability to generate implications of a theory has not yet been demonstrated, and methods for reconstructing proofs of answers are imperfect. In this work we show that a generative model, called ProofWriter, can reliably generate both implications of a theory and the natural language proof(s) that support them. In particular, iterating a 1-step implication generator results in proofs that are highly reliable, and represent actual model decisions (rather than post-hoc rationalizations). On the RuleTaker dataset, the accuracy of ProofWriter's proofs exceed previous methods by +9% absolute, and in a way that generalizes to proof depths unseen in training and on out-of-domain problems. We also show that generative techniques can perform a type of abduction with high precision: Given a theory and an unprovable conclusion, identify a missing fact that allows the conclusion to be proved, along with a proof. These results significantly improve the viability of neural methods for systematically reasoning over natural language."
                    },
                    {
                        "title": "Do language models have coherent mental models of everyday things?",
                        "abstract": "When people think of everyday things like an egg, they typically have a mental image associated with it. This allows them to correctly judge, for example, that \"the yolk surrounds the shell\" is a false statement. Do language models similarly have a coherent picture of such everyday things? To investigate this, we propose a benchmark dataset consisting of 100 everyday things, their parts, and the relationships between these parts, expressed as 11,720 \"X relation Y?\" true/false questions. Using these questions as probes, we observe that state-of-the-art pre-trained language models (LMs) like GPT-3 and Macaw have fragments of knowledge about these everyday things, but do not have fully coherent \"parts mental models\" (54-59% accurate, 19-43% conditional constraint violation). We propose an extension where we add a constraint satisfaction layer on top of the LM's raw predictions to apply commonsense constraints. As well as removing inconsistencies, we find that this also significantly improves accuracy (by 16-20%), suggesting how the incoherence of the LM's pictures of everyday things can be significantly reduced."
                    },
                    {
                        "title": "BaRDa: A Belief and Reasoning Dataset that Separates Factual Accuracy and Reasoning Ability",
                        "abstract": "While there are numerous benchmarks comparing the performance of modern language models (LMs), end-task evaluations often conflate notions of *factual accuracy* (\"truth\") and *reasoning ability* (\"rationality\", or \"honesty\" in the sense of correctly reporting implications of beliefs). Our goal is a dataset that clearly distinguishes these two notions. Our approach is to leverage and extend a collection of human-annotated *entailment trees*, engineered to express both good and bad chains of reasoning, and using a mixture of true and false facts, in particular including counterfactual examples, to avoid belief bias (also known as the \"content effect\"). The resulting dataset, called BaRDa, contains 3000 entailments (1787 valid, 1213 invalid), using 6681 true and 2319 false statements. Testing on four GPT-series models, GPT3(curie)/GPT3(davinici)/3.5/4, we find factual accuracy (truth) scores of 74.1/80.6/82.6/87.1 and reasoning accuracy scores of 63.1/78.0/71.8/79.2. This shows the clear progression of models towards improved factual accuracy and entailment reasoning, and the dataset provides a new benchmark that more cleanly separates and quantifies these two notions."
                    },
                    {
                        "title": "Be Consistent! Improving Procedural Text Comprehension using Label Consistency",
                        "abstract": "Our goal is procedural text comprehension, namely tracking how the properties of entities (e.g., their location) change with time given a procedural text (e.g., a paragraph about photosynthesis, a recipe). This task is challenging as the world is changing throughout the text, and despite recent advances, current systems still struggle with this task. Our approach is to leverage the fact that, for many procedural texts, multiple independent descriptions are readily available, and that predictions from them should be consistent (label consistency). We present a new learning framework that leverages label consistency during training, allowing consistency bias to be built into the model. Evaluation on a standard benchmark dataset for procedural text, ProPara (Dalvi et al., 2018), shows that our approach significantly improves prediction performance (F1) over prior state-of-the-art systems."
                    },
                    {
                        "title": "Reasoning about Actions and State Changes by Injecting Commonsense Knowledge",
                        "abstract": "Comprehending procedural text, e.g., a paragraph describing photosynthesis, requires modeling actions and the state changes they produce, so that questions about entities at different timepoints can be answered. Although several recent systems have shown impressive progress in this task, their predictions can be globally inconsistent or highly improbable. In this paper, we show how the predicted effects of actions in the context of a paragraph can be improved in two ways: (1) by incorporating global, commonsense constraints (e.g., a non-existent entity cannot be destroyed), and (2) by biasing reading with preferences from large-scale corpora (e.g., trees rarely move). Unlike earlier methods, we treat the problem as a neural structured prediction task, allowing hard and soft constraints to steer the model away from unlikely predictions. We show that the new model significantly outperforms earlier systems on a benchmark dataset for procedural text comprehension (+8% relative gain), and that it also avoids some of the nonsensical predictions that earlier systems make."
                    },
                    {
                        "title": "Procedural Reading Comprehension with Attribute-Aware Context Flow",
                        "abstract": "Procedural texts often describe processes (e.g., photosynthesis and cooking) that happen over entities (e.g., light, food). In this paper, we introduce an algorithm for procedural reading comprehension by translating the text into a general formalism that represents processes as a sequence of transitions over entity attributes (e.g., location, temperature). Leveraging pre-trained language models, our model obtains entity-aware and attribute-aware representations of the text by joint prediction of entity attributes and their transitions. Our model dynamically obtains contextual encodings of the procedural text exploiting information that is encoded about previous and current states to predict the transition of a certain attribute which can be identified as a span of text or from a pre-defined set of classes. Moreover, our model achieves state of the art results on two procedural reading comprehension datasets, namely ProPara and npn-cooking"
                    },
                    {
                        "title": "Tracking State Changes in Procedural Text: A Challenge Dataset and Models for Process Paragraph Comprehension",
                        "abstract": "We present a new dataset and models for comprehending paragraphs about processes (e.g., photosynthesis), an important genre of text describing a dynamic world. The new dataset, ProPara, is the first to contain natural (rather than machine-generated) text about a changing world along with a full annotation of entity states (location and existence) during those changes (81k datapoints). The end-task, tracking the location and existence of entities through the text, is challenging because the causal effects of actions are often implicit and need to be inferred. We find that previous models that have worked well on synthetic data achieve only mediocre performance on ProPara, and introduce two new neural models that exploit alternative mechanisms for state prediction, in particular using LSTM input encoding and span prediction. The new models improve accuracy by up to 19%. The dataset and models are available to the community at http://data.allenai.org/propara."
                    },
                    {
                        "title": "WIQA: A dataset for \"What if...\" reasoning over procedural text",
                        "abstract": "We introduce WIQA, the first large-scale dataset of \"What if...\" questions over procedural text. WIQA contains three parts: a collection of paragraphs each describing a process, e.g., beach erosion; a set of crowdsourced influence graphs for each paragraph, describing how one change affects another; and a large (40k) collection of \"What if...?\" multiple-choice questions derived from the graphs. For example, given a paragraph about beach erosion, would stormy weather result in more or less erosion (or have no effect)? The task is to answer the questions, given their associated paragraph. WIQA contains three kinds of questions: perturbations to steps mentioned in the paragraph; external (out-of-paragraph) perturbations requiring commonsense knowledge; and irrelevant (no effect) perturbations. We find that state-of-the-art models achieve 73.8% accuracy, well below the human performance of 96.3%. We analyze the challenges, in particular tracking chains of influences, and present the dataset as an open challenge to the community."
                    },
                    {
                        "title": "Everything Happens for a Reason: Discovering the Purpose of Actions in Procedural Text",
                        "abstract": "Our goal is to better comprehend procedural text, e.g., a paragraph about photosynthesis, by not only predicting what happens, but why some actions need to happen before others. Our approach builds on a prior process comprehension framework for predicting actions' effects, to also identify subsequent steps that those effects enable. We present our new model (XPAD) that biases effect predictions towards those that (1) explain more of the actions in the paragraph and (2) are more plausible with respect to background knowledge. We also extend an existing benchmark dataset for procedural text comprehension, ProPara, by adding the new task of explaining actions by predicting their dependencies. We find that XPAD significantly outperforms prior systems on this task, while maintaining the performance on the original task in ProPara. The dataset is available at http://data.allenai.org/propara"
                    },
                    {
                        "title": "A Dataset for Tracking Entities in Open Domain Procedural Text",
                        "abstract": "We present the first dataset for tracking state changes in procedural text from arbitrary domains by using an unrestricted (open) vocabulary. For example, in a text describing fog removal using potatoes, a car window may transition between being foggy, sticky,opaque, and clear. Previous formulations of this task provide the text and entities involved,and ask how those entities change for just a small, pre-defined set of attributes (e.g., location), limiting their fidelity. Our solution is a new task formulation where given just a procedural text as input, the task is to generate a set of state change tuples(entity, at-tribute, before-state, after-state)for each step,where the entity, attribute, and state values must be predicted from an open vocabulary. Using crowdsourcing, we create OPENPI1, a high-quality (91.5% coverage as judged by humans and completely vetted), and large-scale dataset comprising 29,928 state changes over 4,050 sentences from 810 procedural real-world paragraphs from WikiHow.com. A current state-of-the-art generation model on this task achieves 16.1% F1 based on BLEU metric, leaving enough room for novel model architectures."
                    },
                    {
                        "title": "Just-DREAM-about-it: Figurative Language Understanding with DREAM-FLUTE",
                        "abstract": "Figurative language (e.g., \"he flew like the wind\") is challenging to understand, as it is hard to tell what implicit information is being conveyed from the surface form alone. We hypothesize that to perform this task well, the reader needs to mentally elaborate the scene being described to identify a sensible meaning of the language. We present DREAM-FLUTE, a figurative language understanding system that does this, first forming a \"mental model\" of situations described in a premise and hypothesis before making an entailment/contradiction decision and generating an explanation. DREAM-FLUTE uses an existing scene elaboration model, DREAM, for constructing its \"mental model.\" In the FigLang2022 Shared Task evaluation, DREAM-FLUTE achieved (joint) first place (Acc@60=63.3%), and can perform even better with ensemble techniques, demonstrating the effectiveness of this approach. More generally, this work suggests that adding a reflective component to pretrained language models can improve their performance beyond standard fine-tuning (3.3% improvement in Acc@60)."
                    },
                    {
                        "title": "Skill Set Optimization: Reinforcing Language Model Behavior via Transferable Skills",
                        "abstract": "Large language models (LLMs) have recently been used for sequential decision making in interactive environments. However, leveraging environment reward signals for continual LLM actor improvement is not straightforward. We propose Skill Set Optimization (SSO) for improving LLM actor performance through constructing and refining sets of transferable skills. SSO constructs skills by extracting common subtrajectories with high rewards and generating subgoals and instructions to represent each skill. These skills are provided to the LLM actor in-context to reinforce behaviors with high rewards. Then, SSO further refines the skill set by pruning skills that do not continue to result in high rewards. We evaluate our method in the classic videogame NetHack and the text environment ScienceWorld to demonstrate SSO's ability to optimize a set of skills and perform in-context policy improvement. SSO outperforms baselines by 40% in our custom NetHack task and outperforms the previous state-of-the-art in ScienceWorld by 35%."
                    },
                    {
                        "title": "Think you have Solved Direct-Answer Question Answering? Try ARC-DA, the Direct-Answer AI2 Reasoning Challenge",
                        "abstract": "We present the ARC-DA dataset, a direct-answer (\"open response\", \"freeform\") version of the ARC (AI2 Reasoning Challenge) multiple-choice dataset. While ARC has been influential in the community, its multiple-choice format is unrepresentative of real-world questions, and multiple choice formats can be particularly susceptible to artifacts. The ARC-DA dataset addresses these concerns by converting questions to direct-answer format using a combination of crowdsourcing and expert review. The resulting dataset contains 2985 questions with a total of 8436 valid answers (questions typically have more than one valid answer). ARC-DA is one of the first DA datasets of natural questions that often require reasoning, and where appropriate question decompositions are not evident from the questions themselves. We describe the conversion approach taken, appropriate evaluation metrics, and several strong models. Although high, the best scores (81% GENIE, 61.4% F1, 63.2% ROUGE-L) still leave considerable room for improvement. In addition, the dataset provides a natural setting for new research on explanation, as many questions require reasoning to construct answers. We hope the dataset spurs further advances in complex question-answering by the community. ARC-DA is available at https://allenai.org/data/arc-da"
                    },
                    {
                        "title": "CLIN: A Continually Learning Language Agent for Rapid Task Adaptation and Generalization",
                        "abstract": "Language agents have shown some ability to interact with an external environment, e.g., a virtual world such as ScienceWorld, to perform complex tasks, e.g., growing a plant, without the startup costs of reinforcement learning. However, despite their zero-shot capabilities, these agents to date do not continually improve over time beyond performance refinement on a specific task. Here we present CLIN, the first language-based agent to achieve this, so that it continually improves over multiple trials, including when both the environment and task are varied, and without requiring parameter updates. Our approach is to use a persistent, dynamic, textual memory centered on causal abstractions (rather than general \"helpful hints\") that is regularly updated after each trial so that the agent gradually learns useful knowledge for new trials. In the ScienceWorld benchmark, CLIN is able to continually improve on repeated trials on the same task and environment, outperforming state-of-the-art reflective language agents like Reflexion by 23 absolute points. CLIN can also transfer its learning to new environments (or new tasks), improving its zero-shot performance by 4 points (13 for new tasks) and can further improve performance there through continual memory updates, enhancing performance by an additional 17 points (7 for new tasks). This suggests a new architecture for agents built on frozen models that can still continually and rapidly improve over time."
                    }
                ]
            },
            "fc744bf3-19d0-4ed1-bf41-f32575f8f050": {
                "pk": "fc744bf3-19d0-4ed1-bf41-f32575f8f050",
                "name": "Bodhisattwa Prasad Majumder",
                "collaborators": [
                    "Julian McAuley",
                    "Zexue He",
                    "Harsh Jhamtani",
                    "Taylor Berg-Kirkpatrick",
                    "Shuyang Li",
                    "Sanchaita Hazra",
                    "Jianmo Ni",
                    "EunJeong Hwang",
                    "Niket Tandon",
                    "Sudha Rao"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Bias Mitigation",
                    "Explainable AI",
                    "Conversational Systems"
                ],
                "publications": [
                    {
                        "title": "InterFair: Debiasing with Natural Language Feedback for Fair Interpretable Predictions",
                        "abstract": "Debiasing methods in NLP models traditionally focus on isolating information related to a sensitive attribute (e.g., gender or race). We instead argue that a favorable debiasing method should use sensitive information 'fairly,' with explanations, rather than blindly eliminating it. This fair balance is often subjective and can be challenging to achieve algorithmically. We explore two interactive setups with a frozen predictive model and show that users able to provide feedback can achieve a better and fairer balance between task performance and bias mitigation. In one setup, users, by interacting with test examples, further decreased bias in the explanations (5-8%) while maintaining the same prediction accuracy. In the other setup, human feedback was able to disentangle associated bias and predictive information from the input leading to superior bias mitigation and improved task performance (4-5%) simultaneously."
                    },
                    {
                        "title": "Aligning Language Models to User Opinions",
                        "abstract": "An important aspect of developing LLMs that interact with humans is to align models' behavior to their users. It is possible to prompt an LLM into behaving as a certain persona, especially a user group or ideological persona the model captured during its pertaining stage. But, how to best align an LLM with a specific user and not a demographic or ideological group remains an open question. Mining public opinion surveys (by Pew Research), we find that the opinions of a user and their demographics and ideologies are not mutual predictors. We use this insight to align LLMs by modeling both user opinions as well as user demographics and ideology, achieving up to 7 points accuracy gains in predicting public opinions from survey questions across a broad set of topics. In addition to the typical approach of prompting LLMs with demographics and ideology, we discover that utilizing the most relevant past opinions from individual users enables the model to predict user opinions more accurately."
                    },
                    {
                        "title": "To Tell The Truth: Language of Deception and Language Models",
                        "abstract": "Text-based misinformation permeates online discourses, yet evidence of people's ability to discern truth from such deceptive textual content is scarce. We analyze a novel TV game show data where conversations in a high-stake environment between individuals with conflicting objectives result in lies. We investigate the manifestation of potentially verifiable language cues of deception in the presence of objective truth, a distinguishing feature absent in previous text-based deception datasets. We show that there exists a class of detectors (algorithms) that have similar truth detection performance compared to human subjects, even when the former accesses only the language cues while the latter engages in conversations with complete access to all potential sources of cues (language and audio-visual). Our model, built on a large language model, employs a bottleneck framework to learn discernible cues to determine truth, an act of reasoning in which human subjects often perform poorly, even with incentives. Our model detects novel but accurate language cues in many cases where humans failed to detect deception, opening up the possibility of humans collaborating with algorithms and ameliorating their ability to detect the truth."
                    },
                    {
                        "title": "Detect and Perturb: Neutral Rewriting of Biased and Sensitive Text via Gradient-based Decoding",
                        "abstract": "Written language carries explicit and implicit biases that can distract from meaningful signals. For example, letters of reference may describe male and female candidates differently, or their writing style may indirectly reveal demographic characteristics. At best, such biases distract from the meaningful content of the text; at worst they can lead to unfair outcomes. We investigate the challenge of re-generating input sentences to 'neutralize' sensitive attributes while maintaining the semantic meaning of the original text (e.g. is the candidate qualified?). We propose a gradient-based rewriting framework, Detect and Perturb to Neutralize (DEPEN), that first detects sensitive components and masks them for regeneration, then perturbs the generation model at decoding time under a neutralizing constraint that pushes the (predicted) distribution of sensitive attributes towards a uniform distribution. Our experiments in two different scenarios show that DEPEN can regenerate fluent alternatives that are neutral in the sensitive attribute while maintaining the semantics of other attributes."
                    },
                    {
                        "title": "Like hiking? You probably enjoy nature: Persona-grounded Dialog with Commonsense Expansions",
                        "abstract": "Existing persona-grounded dialog models often fail to capture simple implications of given persona descriptions, something which humans are able to do seamlessly. For example, state-of-the-art models cannot infer that interest in hiking might imply love for nature or longing for a break. In this paper, we propose to expand available persona sentences using existing commonsense knowledge bases and paraphrasing resources to imbue dialog models with access to an expanded and richer set of persona descriptions. Additionally, we introduce fine-grained grounding on personas by encouraging the model to make a discrete choice among persona sentences while synthesizing a dialog response. Since such a choice is not observed in the data, we model it using a discrete latent random variable and use variational learning to sample from hundreds of persona expansions. Our model outperforms competitive baselines on the PersonaChat dataset in terms of dialog quality and diversity while achieving persona-consistent and controllable dialog generation."
                    },
                    {
                        "title": "Interview: A Large-Scale Open-Source Corpus of Media Dialog",
                        "abstract": "Existing conversational datasets consist either of written proxies for dialog or small-scale transcriptions of natural speech. We introduce 'Interview': a large-scale (105K conversations) media dialog dataset collected from news interview transcripts. Compared to existing large-scale proxies for conversational data, language models trained on our dataset exhibit better zero-shot out-of-domain performance on existing spoken dialog datasets, demonstrating its usefulness in modeling real-world conversations. 'Interview' contains speaker role annotations for each turn, facilitating the development of engaging, responsive dialog systems. In fact, experiments on two dialog tasks show that leveraging such labels improves performance over strong speaker-agnostic baselines, and enabling models to generate more specific and inquisitive responses in interview-style conversations."
                    },
                    {
                        "title": "Generating Personalized Recipes from Historical User Preferences",
                        "abstract": "Existing approaches to recipe generation are unable to create recipes for users with culinary preferences but incomplete knowledge of ingredients in specific dishes. We propose a new task of personalized recipe generation to help these users: expanding a name and incomplete ingredient details into complete natural-text instructions aligned with the user's historical preferences. We attend on technique- and recipe-level representations of a user's previously consumed recipes, fusing these 'user-aware' representations in an attention fusion layer to control recipe text generation. Experiments on a new dataset of 180K recipes and 700K interactions show our model's ability to generate plausible and personalized recipes compared to non-personalized baselines."
                    },
                    {
                        "title": "Ask what's missing and what's useful: Improving Clarification Question Generation using Global Knowledge",
                        "abstract": "The ability to generate clarification questions i.e., questions that identify useful missing information in a given context, is important in reducing ambiguity. Humans use previous experience with similar contexts to form a global view and compare it to the given context to ascertain what is missing and what is useful in the context. Inspired by this, we propose a model for clarification question generation where we first identify what is missing by taking a difference between the global and the local view and then train a model to identify what is useful and generate a question about it. Our model outperforms several baselines as judged by both automatic metrics and humans."
                    },
                    {
                        "title": "Unsupervised Enrichment of Persona-grounded Dialog with Background Stories",
                        "abstract": "Humans often refer to personal narratives, life experiences, and events to make a conversation more engaging and rich. While persona-grounded dialog models are able to generate responses that follow a given persona, they often miss out on stating detailed experiences or events related to a persona, often leaving conversations shallow and dull. In this work, we equip dialog models with 'background stories' related to a persona by leveraging fictional narratives from existing story datasets (e.g. ROCStories). Since current dialog datasets do not contain such narratives as responses, we perform an unsupervised adaptation of a retrieved story for generating a dialog response using a gradient-based rewriting technique. Our proposed method encourages the generated response to be fluent (i.e., highly likely) with the dialog history, minimally different from the retrieved story to preserve event ordering and consistent with the original persona. We demonstrate that our method can generate responses that are more diverse, and are rated more engaging and human-like by human evaluators, compared to outputs from existing dialog models."
                    },
                    {
                        "title": "Achieving Conversational Goals with Unsupervised Post-hoc Knowledge Injection",
                        "abstract": "A limitation of current neural dialog models is that they tend to suffer from a lack of specificity and informativeness in generated responses, primarily due to dependence on training data that covers a limited variety of scenarios and conveys limited knowledge. One way to alleviate this issue is to extract relevant knowledge from external sources at decoding time and incorporate it into the dialog response. In this paper, we propose a post-hoc knowledge-injection technique where we first retrieve a diverse set of relevant knowledge snippets conditioned on both the dialog history and an initial response from an existing dialog model. We construct multiple candidate responses, individually injecting each retrieved snippet into the initial response using a gradient-based decoding method, and then select the final response with an unsupervised ranking step. Our experiments in goal-oriented and knowledge-grounded dialog settings demonstrate that human annotators judge the outputs from the proposed method to be more engaging and informative compared to responses from prior dialog systems. We further show that knowledge-augmentation promotes success in achieving conversational goals in both experimental settings."
                    },
                    {
                        "title": "Self-Supervised Bot Play for Conversational Recommendation with Justifications",
                        "abstract": "Conversational recommender systems offer the promise of interactive, engaging ways for users to find items they enjoy. We seek to improve conversational recommendation via three dimensions: 1) We aim to mimic a common mode of human interaction for recommendation: experts justify their suggestions, a seeker explains why they don't like the item, and both parties iterate through the dialog to find a suitable item. 2) We leverage ideas from conversational critiquing to allow users to flexibly interact with natural language justifications by critiquing subjective aspects. 3) We adapt conversational recommendation to a wider range of domains where crowd-sourced ground truth dialogs are not available. We develop a new two-part framework for training conversational recommender systems. First, we train a recommender system to jointly suggest items and justify its reasoning with subjective aspects. We then fine-tune this model to incorporate iterative user feedback via self-supervised bot-play. Experiments on three real-world datasets demonstrate that our system can be applied to different recommendation models across diverse domains to achieve superior performance in conversational recommendation compared to state-of-the-art methods. We also evaluate our model on human users, showing that systems trained under our framework provide more useful, helpful, and knowledgeable recommendations in warm- and cold-start settings."
                    },
                    {
                        "title": "Fault Detection Engine in Intelligent Predictive Analytics Platform for DCIM",
                        "abstract": "With the advancement of huge data generation and data handling capability, Machine Learning and Probabilistic modelling enables an immense opportunity to employ predictive analytics platform in high security critical industries namely data centers, electricity grids, utilities, airport etc. where downtime minimization is one of the primary objectives. This paper proposes a novel, complete architecture of an intelligent predictive analytics platform, Fault Engine, for huge device network connected with electrical/information flow. Three unique modules, here proposed, seamlessly integrate with available technology stack of data handling and connect with middleware to produce online intelligent prediction in critical failure scenarios. The Markov Failure module predicts the severity of a failure along with survival probability of a device at any given instances. The Root Cause Analysis model indicates probable devices as potential root cause employing Bayesian probability assignment and topological sort. Finally, a community detection algorithm produces correlated clusters of device in terms of failure probability which will further narrow down the search space of finding route cause. The whole Engine has been tested with different size of network with simulated failure environments and shows its potential to be scalable in real-time implementation."
                    },
                    {
                        "title": "On Faithfulness and Coherence of Language Explanations for Recommendation Systems",
                        "abstract": "Reviews contain rich information about product characteristics and user interests and thus are commonly used to boost recommender system performance. Specifically, previous work show that jointly learning to perform review generation improves rating prediction performance. Meanwhile, these model-produced reviews serve as recommendation explanations, providing the user with insights on predicted ratings. However, while existing models could generate fluent, human-like reviews, it is unclear to what degree the reviews fully uncover the rationale behind the jointly predicted rating. In this work, we perform a series of evaluations that probes state-of-the-art models and their review generation component. We show that the generated explanations are brittle and need further evaluation before being taken as literal rationales for the estimated ratings."
                    },
                    {
                        "title": "Deep Recurrent Neural Networks for Product Attribute Extraction in eCommerce",
                        "abstract": "Extracting accurate attribute qualities from product titles is a vital component in delivering eCommerce customers with a rewarding online shopping experience via an enriched faceted search. We demonstrate the potential of Deep Recurrent Networks in this domain, primarily models such as Bidirectional LSTMs and Bidirectional LSTM-CRF with or without an attention mechanism. These have improved overall F1 scores, as compared to the previous benchmarks (More et al.) by at least 0.0391, showcasing an overall precision of 97.94%, recall of 94.12% and the F1 score of 0.9599. This has made us achieve a significant coverage of important facets or attributes of products which not only shows the efficacy of deep recurrent models over previous machine learning benchmarks but also greatly enhances the overall customer experience while shopping online."
                    },
                    {
                        "title": "Data-driven Discovery with Large Generative Models",
                        "abstract": "With the accumulation of data at an unprecedented rate, its potential to fuel scientific discovery is growing exponentially. This position paper urges the Machine Learning (ML) community to exploit the capabilities of large generative models (LGMs) to develop automated systems for end-to-end data-driven discovery -- a paradigm encompassing the search and verification of hypotheses purely from a set of provided datasets, without the need for additional data collection or physical experiments. We first outline several desiderata for an ideal data-driven discovery system. Then, through DATAVOYAGER, a proof-of-concept utilizing GPT-4, we demonstrate how LGMs fulfill several of these desiderata -- a feat previously unattainable -- while also highlighting important limitations in the current system that open up opportunities for novel ML research. We contend that achieving accurate, reliable, and robust end-to-end discovery systems solely through the current capabilities of LGMs is challenging. We instead advocate for fail-proof tool integration, along with active user moderation through feedback mechanisms, to foster data-driven scientific discoveries with efficiency and reproducibility."
                    },
                    {
                        "title": "Controlling Bias Exposure for Fair Interpretable Predictions",
                        "abstract": "Recent work on reducing bias in NLP models usually focuses on protecting or isolating information related to a sensitive attribute (like gender or race). However, when sensitive information is semantically entangled with the task information of the input, e.g., gender information is predictive for a profession, a fair trade-off between task performance and bias mitigation is difficult to achieve. Existing approaches perform this trade-off by eliminating bias information from the latent space, lacking control over how much bias is necessarily required to be removed. We argue that a favorable debiasing method should use sensitive information 'fairly', rather than blindly eliminating it (Caliskan et al., 2017; Sun et al., 2019; Bogen et al., 2020). In this work, we provide a novel debiasing algorithm by adjusting the predictive model's belief to (1) ignore the sensitive information if it is not useful for the task; (2) use sensitive information minimally as necessary for the prediction (while also incurring a penalty). Experimental results on two text classification tasks (influenced by gender) and an open-ended generation task (influenced by race) indicate that our model achieves a desirable trade-off between debiasing and task performance along with producing debiased rationales as evidence."
                    },
                    {
                        "title": "Tell, Don't Show!: Language Guidance Eases Transfer Across Domains in Images and Videos",
                        "abstract": "We introduce LaGTran, a novel framework that utilizes text supervision to guide robust transfer of discriminative knowledge from labeled source to unlabeled target data with domain gaps. While unsupervised adaptation methods have been established to address this problem, they show limitations in handling challenging domain shifts due to their exclusive operation within the pixel-space. Motivated by our observation that semantically richer text modality has more favorable transfer properties, we devise a transfer mechanism to use a source-trained text-classifier to generate predictions on the target text descriptions, and utilize these predictions as supervision for the corresponding images. Our approach driven by language guidance is surprisingly easy and simple, yet significantly outperforms all prior approaches on challenging datasets like GeoNet and DomainNet, validating its extreme effectiveness. To further extend the scope of our study beyond images, we introduce a new benchmark called Ego2Exo to study ego-exo transfer in videos and find that our language-aided approach LaGTran yields significant gains in this highly challenging and non-trivial transfer setting. Code, models, and proposed datasets are publicly available at https://tarun005.github.io/lagtran/."
                    },
                    {
                        "title": "Upcycle Your OCR: Reusing OCRs for Post-OCR Text Correction in Romanised Sanskrit",
                        "abstract": "We propose a post-OCR text correction approach for digitising texts in Romanised Sanskrit. Owing to the lack of resources our approach uses OCR models trained for other languages written in Roman. Currently, there exists no dataset available for Romanised Sanskrit OCR. So, we bootstrap a dataset of 430 images, scanned in two different settings and their corresponding ground truth. For training, we synthetically generate training images for both the settings. We find that the use of copying mechanism (Gu et al., 2016) yields a percentage increase of 7.69 in Character Recognition Rate (CRR) than the current state of the art model in solving monotone sequence-to-sequence tasks (Schnober et al., 2016). We find that our system is robust in combating OCR-prone errors, as it obtains a CRR of 87.01% from an OCR output with CRR of 35.76% for one of the dataset settings. A human judgment survey performed on the models shows that our proposed model results in predictions which are faster to comprehend and faster to improve for a human than the other systems."
                    },
                    {
                        "title": "Knowledge-Grounded Self-Rationalization via Extractive and Natural Language Explanations",
                        "abstract": "Models that generate extractive rationales (i.e., subsets of features) or natural language explanations (NLEs) for their predictions are important for explainable AI. While an extractive rationale provides a quick view of the features most responsible for a prediction, an NLE allows for a comprehensive description of the decision-making process behind a prediction. However, current models that generate the best extractive rationales or NLEs often fall behind the state-of-the-art (SOTA) in terms of task performance. In this work, we bridge this gap by introducing RExC, a self-rationalizing framework that grounds its predictions and two complementary types of explanations (NLEs and extractive rationales) in background knowledge. Our framework improves over previous methods by: (i) reaching SOTA task performance while also providing explanations, (ii) providing two types of explanations, while existing models usually provide only one type, and (iii) beating by a large margin the previous SOTA in terms of quality of both types of explanations. Furthermore, a perturbation analysis in RExC shows a high degree of association between explanations and predictions, a necessary property of faithful explanations."
                    },
                    {
                        "title": "Kernelized Weighted SUSAN based Fuzzy C-Means Clustering for Noisy Image Segmentation",
                        "abstract": "The paper proposes a novel Kernelized image segmentation scheme for noisy images that utilizes the concept of Smallest Univalue Segment Assimilating Nucleus (SUSAN) and incorporates spatial constraints by computing circular colour map induced weights. Fuzzy damping coefficients are obtained for each nucleus or center pixel on the basis of the corresponding weighted SUSAN area values, the weights being equal to the inverse of the number of horizontal and vertical moves required to reach a neighborhood pixel from the center pixel. These weights are used to vary the contributions of the different nuclei in the Kernel based framework. The paper also presents an edge quality metric obtained by fuzzy decision based edge candidate selection and final computation of the blurriness of the edges after their selection. The inability of existing algorithms to preserve edge information and structural details in their segmented maps necessitates the computation of the edge quality factor (EQF) for all the competing algorithms. Qualitative and quantitative analysis have been rendered with respect to state-of-the-art algorithms and for images ridden with varying types of noises. Speckle noise ridden SAR images and Rician noise ridden Magnetic Resonance Images have also been considered for evaluating the effectiveness of the proposed algorithm in extracting important segmentation information."
                    }
                ]
            },
            "3fecdff1-02ee-40f6-ba85-781670013790": {
                "pk": "3fecdff1-02ee-40f6-ba85-781670013790",
                "name": "Oyvind Tafjord",
                "collaborators": [
                    "Peter Clark",
                    "Robert C. Myers",
                    "Bhavana Dalvi Mishra",
                    "Vipul Periwal",
                    "Neil R. Constable",
                    "Kevin Lin",
                    "Matt Gardner",
                    "Marcus T. Grisaru",
                    "Jayant Krishnamurthy",
                    "Aniruddha Kembhavi"
                ],
                "domain": [
                    "Quantum Gravity",
                    "Machine Learning",
                    "Natural Language Processing",
                    "Theoretical Physics"
                ],
                "publications": [
                    {
                        "title": "Finite-Time Amplitudes In Matrix Theory",
                        "abstract": "We evaluate one-loop finite-time amplitudes for graviton scattering in Matrix theory and compare to the corresponding amplitudes in supergravity. We find agreement for arbitrary time intervals at leading order in distance, providing a functional agreement between supergravity and Matrix theory. At subleading order, we find corrections to the effective potential found from previous phase shift calculations in Matrix theory."
                    },
                    {
                        "title": "Matrix string interactions",
                        "abstract": "String configurations have been identified in compactified Matrix theory at vanishing string coupling. We show how the interactions of these strings are determined by the Yang-Mills gauge field on the worldsheet. At finite string coupling, this suggests the underlying dynamics is not well-approximated as a theory of strings. This may explain why string perturbation theory diverges badly, while Matrix string perturbation theory presumably has a perturbative expansion with properties similar to the strong coupling expansion of 2d Yang-Mills theory."
                    },
                    {
                        "title": "Mechanism for long Dijkgraaf-Verlinde-Verlinde strings",
                        "abstract": "The proposal of Dijkgraaf, Verlinde and Verlinde for the emergence of smooth strings from a supersymmetric U(N) Yang-Mills theory assumes that conjugacy classes in the symmetric group with a few long cycles dominate the dynamics about the infrared fixed point. It is shown that the average number of cycles in a conjugacy class of $S_{N}$ is bounded below by $const.\\sqrt N\\ln\\sqrt N,$ implying that some physical mechanism is needed to ensure the assumed dominance. It is shown that if individual cycles have positive energies that depend very weakly on their lengths, then long cycles dominate the dynamics at low temperatures."
                    },
                    {
                        "title": "General-Purpose Question-Answering with Macaw",
                        "abstract": "Despite the successes of pretrained language models, there are still few high-quality, general-purpose QA systems that are freely available. In response, we present Macaw, a versatile, generative question-answering (QA) system that we are making available to the community. Macaw is built on UnifiedQA, itself built on T5, and exhibits strong performance, zero-shot, on a wide variety of topics, including outperforming GPT-3 by over 10% (absolute) on Challenge300, a suite of 300 challenge questions, despite being an order of magnitude smaller (11 billion vs. 175 billion parameters). In addition, Macaw allows different permutations (\"angles\") of its inputs and outputs to be used, for example Macaw can take a question and produce an answer; or take an answer and produce a question; or take an answer and question, and produce multiple-choice options. We describe the system, and illustrate a variety of question types where it produces surprisingly good answers, well outside the training setup. We also identify question classes where it still appears to struggle, offering insights into the limitations of pretrained language models. Macaw is freely available, and we hope that it proves useful to the community. Macaw is available at https://github.com/allenai/macaw"
                    },
                    {
                        "title": "SUSY and Goliath",
                        "abstract": "We investigate the `giant gravitons' of McGreevy, Susskind and Toumbas [hep-th/0003075]. We demonstrate that these are BPS configurations which preserve precisely the same supersymmetries as a `point-like' graviton. We also show that there exist `dual' giant gravitons consisting of spherical branes expanding into the AdS component of the spacetime. Finally, we discuss the realization of the stringy exclusion principle within this expanded framework."
                    },
                    {
                        "title": "Non-abelian Brane Intersections",
                        "abstract": "We study new solutions of the low-energy equations of motion for the non-abelian D-string. We find a \"fuzzy funnel\" solution consisting of a noncommutative four-sphere geometry which expands along the length of the D-string. We show that this funnel solution has an interpretation as D-strings ending on a set of orthogonal D5-branes. Although not supersymmetric, the system appears to be stable within this framework. We also give a dual description of this configuration as a bion spike in the non-abelian world volume theory of coincident D5-branes."
                    },
                    {
                        "title": "Superstars and Giant Gravitons",
                        "abstract": "We examine a family of BPS solutions of ten-dimensional type IIb supergravity. These solutions asymptotically approach AdS_5 X S^5 and carry internal `angular' momentum on the five-sphere. While a naked singularity appears at the center of the anti-de Sitter space, we show that it has a natural physical interpretation in terms of a collection of giant gravitons. We calculate the distribution of giant gravitons from the dipole field induced in the Ramond-Ramond five-form, and show that these sources account for the entire internal momentum carried by the BPS solutions."
                    },
                    {
                        "title": "Fuzzy Funnels: Non-abelian Brane Intersections",
                        "abstract": "We discuss dual formulations of D-brane intersections. The duality is between world volume field theories of different dimensionalities which both describe the same D-brane configuration but are valid in complementary regions of parameter space. We discuss the duality in terms of bion configurations involving D-strings orthogonally intersecting both D3-branes and D5-branes."
                    },
                    {
                        "title": "Semantic Parsing to Probabilistic Programs for Situated Question Answering",
                        "abstract": "Situated question answering is the problem of answering questions about an environment such as an image or diagram. This problem requires jointly interpreting a question and an environment using background knowledge to select the correct answer. We present Parsing to Probabilistic Programs (P3), a novel situated question answering model that can use background knowledge and global features of the question/environment interpretation while retaining efficient approximate inference. Our key insight is to treat semantic parses as probabilistic programs that execute nondeterministically and whose possible executions represent environmental uncertainty. We evaluate our approach on a new, publicly-released data set of 5000 science diagram questions, outperforming several competitive classical and neural baselines."
                    },
                    {
                        "title": "Reasoning Over Paragraph Effects in Situations",
                        "abstract": "A key component of successfully reading a passage of text is the ability to apply knowledge gained from the passage to a new situation. In order to facilitate progress on this kind of reading, we present ROPES, a challenging benchmark for reading comprehension targeting Reasoning Over Paragraph Effects in Situations. We target expository language describing causes and effects (e.g., \"animal pollinators increase efficiency of fertilization in flowers\"), as they have clear implications for new situations. A system is presented a background passage containing at least one of these relations, a novel situation that uses this background, and questions that require reasoning about effects of the relationships in the background passage in the context of the situation. We collect background passages from science textbooks and Wikipedia that contain such phenomena, and ask crowd workers to author situations, questions, and answers, resulting in a 14,322 question dataset. We analyze the challenges of this task and evaluate the performance of state-of-the-art reading comprehension models. The best model performs only slightly better than randomly guessing an answer of the correct type, at 61.6% F1, well below the human performance of 89.0%."
                    },
                    {
                        "title": "QuaRTz: An Open-Domain Dataset of Qualitative Relationship Questions",
                        "abstract": "We introduce the first open-domain dataset, called QuaRTz, for reasoning about textual qualitative relationships. QuaRTz contains general qualitative statements, e.g., \"A sunscreen with a higher SPF protects the skin longer.\", twinned with 3864 crowdsourced situated questions, e.g., \"Billy is wearing sunscreen with a lower SPF than Lucy. Who will be best protected from the sun?\", plus annotations of the properties being compared. Unlike previous datasets, the general knowledge is textual and not tied to a fixed set of relationships, and tests a system's ability to comprehend and apply textual qualitative knowledge in a novel setting. We find state-of-the-art results are substantially (20%) below human performance, presenting an open challenge to the NLP community."
                    },
                    {
                        "title": "Transformers as Soft Reasoners over Language",
                        "abstract": "Beginning with McCarthy's Advice Taker (1959), AI has pursued the goal of providing a system with explicit, general knowledge and having the system reason over that knowledge. However, expressing the knowledge in a formal (logical or probabilistic) representation has been a major obstacle to this research. This paper investigates a modern approach to this problem where the facts and rules are provided as natural language sentences, thus bypassing a formal representation. We train transformers to reason (or emulate reasoning) over these sentences using synthetically generated data. Our models, that we call RuleTakers, provide the first empirical demonstration that this kind of soft reasoning over language is learnable, can achieve high (99%) accuracy, and generalizes to test data requiring substantially deeper chaining than seen during training (95%+ scores). We also demonstrate that the models transfer well to two hand-authored rulebases, and to rulebases paraphrased into more natural language. These findings are significant as it suggests a new role for transformers, namely as limited \"soft theorem provers\" operating over explicit theories in language. This in turn suggests new possibilities for explainability, correctability, and counterfactual reasoning in question-answering."
                    },
                    {
                        "title": "ProofWriter: Generating Implications, Proofs, and Abductive Statements over Natural Language",
                        "abstract": "Transformers have been shown to emulate logical deduction over natural language theories (logical rules expressed in natural language), reliably assigning true/false labels to candidate implications. However, their ability to generate implications of a theory has not yet been demonstrated, and methods for reconstructing proofs of answers are imperfect. In this work we show that a generative model, called ProofWriter, can reliably generate both implications of a theory and the natural language proof(s) that support them. In particular, iterating a 1-step implication generator results in proofs that are highly reliable, and represent actual model decisions (rather than post-hoc rationalizations). On the RuleTaker dataset, the accuracy of ProofWriter's proofs exceed previous methods by +9% absolute, and in a way that generalizes to proof depths unseen in training and on out-of-domain problems. We also show that generative techniques can perform a type of abduction with high precision: Given a theory and an unprovable conclusion, identify a missing fact that allows the conclusion to be proved, along with a proof. These results significantly improve the viability of neural methods for systematically reasoning over natural language."
                    },
                    {
                        "title": "\"You are grounded!\": Latent Name Artifacts in Pre-trained Language Models",
                        "abstract": "Pre-trained language models (LMs) may perpetuate biases originating in their training corpus to downstream models. We focus on artifacts associated with the representation of given names (e.g., Donald), which, depending on the corpus, may be associated with specific entities, as indicated by next token prediction (e.g., Trump). While helpful in some contexts, grounding happens also in under-specified or inappropriate contexts. For example, endings generated for `Donald is a' substantially differ from those of other names, and often have more-than-average negative sentiment. We demonstrate the potential effect on downstream tasks with reading comprehension probes where name perturbation changes the model answers. As a silver lining, our experiments suggest that additional pre-training on different corpora may mitigate this bias."
                    },
                    {
                        "title": "Towards Teachable Reasoning Systems: Using a Dynamic Memory of User Feedback for Continual System Improvement",
                        "abstract": "Our goal is a teachable reasoning system for question-answering (QA), where a user can interact with faithful answer explanations, and correct its errors so that the system improves over time. Our approach is to augment a QA model with a dynamic memory of user feedback, containing user-supplied corrections to erroneous model beliefs that users identify during interaction. Retrievals from memory are used as additional context for QA, to help avoid previous mistakes in similar new situations - a novel application of memory-based continuous learning. With simulated feedback, we find that our system (called TeachMe) continually improves with time, and without model retraining, requiring feedback on only 25% of training examples to reach within 1% of the upper-bound (feedback on all examples). Similarly, in experiments with real users, we observe a similar trend, with performance improving by over 15% on a hidden test set after teaching. This suggests new opportunities for using frozen language models in an interactive setting where users can inspect, debug, and correct the model's beliefs, leading to improved system's performance over time."
                    },
                    {
                        "title": "Entailer: Answering Questions with Faithful and Truthful Chains of Reasoning",
                        "abstract": "Our goal is a question-answering (QA) system that can show how its answers are implied by its own internal beliefs via a systematic chain of reasoning. Such a capability would allow better understanding of why a model produced the answer it did. Our approach is to recursively combine a trained backward-chaining model, capable of generating a set of premises entailing an answer hypothesis, with a verifier that checks that the model itself believes those premises (and the entailment itself) through self-querying. To our knowledge, this is the first system to generate multistep chains that are both faithful (the answer follows from the reasoning) and truthful (the chain reflects the system's own internal beliefs). In evaluation using two different datasets, users judge that a majority (70%+) of generated chains clearly show how an answer follows from a set of facts - substantially better than a high-performance baseline - while preserving answer accuracy. By materializing model beliefs that systematically support an answer, new opportunities arise for understanding the model's system of belief, and diagnosing and correcting its misunderstandings when an answer is wrong."
                    },
                    {
                        "title": "Digital Socrates: Evaluating LLMs through Explanation Critiques",
                        "abstract": "While LLMs can provide reasoned explanations along with their answers, the nature and quality of those explanations are still poorly understood. In response, our goal is to define a detailed way of characterizing the explanation capabilities of modern models and to create a nuanced, interpretable explanation evaluation tool that can generate such characterizations automatically, without relying on expensive API calls or human annotations. Our approach is to (a) define the new task of explanation critiquing - identifying and categorizing any main flaw in an explanation and providing suggestions to address the flaw, (b) create a sizeable, human-verified dataset for this task, and (c) train an open-source, automatic critique model (called Digital Socrates) using this data. Through quantitative and qualitative analysis, we demonstrate how Digital Socrates is useful for revealing insights about student models by examining their reasoning chains, and how it can provide high-quality, nuanced, automatic evaluation of those model explanations for the first time. Digital Socrates thus fills an important gap in evaluation tools for understanding and improving the explanation behavior of models."
                    },
                    {
                        "title": "BaRDa: A Belief and Reasoning Dataset that Separates Factual Accuracy and Reasoning Ability",
                        "abstract": "While there are numerous benchmarks comparing the performance of modern language models (LMs), end-task evaluations often conflate notions of *factual accuracy* (\"truth\") and *reasoning ability* (\"rationality\", or \"honesty\" in the sense of correctly reporting implications of beliefs). Our goal is a dataset that clearly distinguishes these two notions. Our approach is to leverage and extend a collection of human-annotated *entailment trees*, engineered to express both good and bad chains of reasoning, and using a mixture of true and false facts, in particular including counterfactual examples, to avoid belief bias (also known as the \"content effect\"). The resulting dataset, called BaRDa, contains 3000 entailments (1787 valid, 1213 invalid), using 6681 true and 2319 false statements. Testing on four GPT-series models, GPT3(curie)/GPT3(davinici)/3.5/4, we find factual accuracy (truth) scores of 74.1/80.6/82.6/87.1 and reasoning accuracy scores of 63.1/78.0/71.8/79.2. This shows the clear progression of models towards improved factual accuracy and entailment reasoning, and the dataset provides a new benchmark that more cleanly separates and quantifies these two notions."
                    }
                ]
            },
            "04e05578-6162-4976-a7de-bca47e6416df": {
                "pk": "04e05578-6162-4976-a7de-bca47e6416df",
                "name": "Peter Clark",
                "collaborators": [
                    "Oyvind Tafjord",
                    "Sumithra Bhakthavatsalam",
                    "Kyle Richardson",
                    "Niket Tandon",
                    "Alastair Wood",
                    "Peter Olley",
                    "Peter Hase",
                    "Mohit Bansal",
                    "Sarah Wiegreffe",
                    "Carissa Schoenick"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Knowledge Representation",
                    "Question Answering",
                    "Story Generation"
                ],
                "publications": [
                    {
                        "title": "Story Generation and Aviation Incident Representation",
                        "abstract": "This working note discusses the topic of story generation, with a view to identifying the knowledge required to understand aviation incident narratives (which have structural similarities to stories), following the premise that to understand aviation incidents, one should at least be able to generate examples of them. We give a brief overview of aviation incidents and their relation to stories, and then describe two of our earlier attempts (using `scripts' and `story grammars') at incident generation which did not evolve promisingly. Following this, we describe a simple incident generator which did work (at a `toy' level), using a `world simulation' approach. This generator is based on Meehan's TALE-SPIN story generator (1977). We conclude with a critique of the approach."
                    },
                    {
                        "title": "What Knowledge is Needed to Solve the RTE5 Textual Entailment Challenge?",
                        "abstract": "This document gives a knowledge-oriented analysis of about 20 interesting Recognizing Textual Entailment (RTE) examples, drawn from the 2005 RTE5 competition test set. The analysis ignores shallow statistical matching techniques between T and H, and rather asks: What would it take to reasonably infer that T implies H? What world knowledge would be needed for this task? Although such knowledge-intensive techniques have not had much success in RTE evaluations, ultimately an intelligent system should be expected to know and deploy this kind of world knowledge required to perform this kind of reasoning.   The selected examples are typically ones which our RTE system (called BLUE) got wrong and ones which require world knowledge to answer. In particular, the analysis covers cases where there was near-perfect lexical overlap between T and H, yet the entailment was NO, i.e., examples that most likely all current RTE systems will have got wrong. A nice example is #341 (page 26), that requires inferring from \"a river floods\" that \"a river overflows its banks\". Seems it should be easy, right? Enjoy!"
                    },
                    {
                        "title": "Investigation of generalised Fredholm equations and their direct solution subject to constraints",
                        "abstract": "This paper explores the solution of Fredholm-like equations with infinite dimensional solution spaces. We set out to find a method for determining a particular solution to a Fredholm-like equation subject to a given constraint. The relevance and application comes via a connection to certain dynamics of gas-like systems."
                    },
                    {
                        "title": "General-Purpose Question-Answering with Macaw",
                        "abstract": "Despite the successes of pretrained language models, there are still few high-quality, general-purpose QA systems that are freely available. In response, we present Macaw, a versatile, generative question-answering (QA) system that we are making available to the community. Macaw is built on UnifiedQA, itself built on T5, and exhibits strong performance, zero-shot, on a wide variety of topics, including outperforming GPT-3 by over 10% (absolute) on Challenge300, a suite of 300 challenge questions, despite being an order of magnitude smaller (11 billion vs. 175 billion parameters). In addition, Macaw allows different permutations (\"angles\") of its inputs and outputs to be used, for example Macaw can take a question and produce an answer; or take an answer and produce a question; or take an answer and question, and produce multiple-choice options. We describe the system, and illustrate a variety of question types where it produces surprisingly good answers, well outside the training setup. We also identify question classes where it still appears to struggle, offering insights into the limitations of pretrained language models. Macaw is freely available, and we hope that it proves useful to the community. Macaw is available at https://github.com/allenai/macaw"
                    },
                    {
                        "title": "The Unreasonable Effectiveness of Easy Training Data for Hard Tasks",
                        "abstract": "How can we train models to perform well on hard test data when hard training data is by definition difficult to label correctly? This question has been termed the scalable oversight problem and has drawn increasing attention as language models have continually improved. In this paper, we present the surprising conclusion that current pretrained language models often generalize relatively well from easy to hard data, even performing as well as oracle models finetuned on hard data. We demonstrate this kind of easy-to-hard generalization using simple finetuning methods like in-context learning, linear classifier heads, and QLoRA for seven different measures of datapoint hardness, including six empirically diverse human hardness measures (like grade level) and one model-based measure (loss-based). Furthermore, we show that even if one cares most about model performance on hard data, it can be better to collect easy data rather than hard data for finetuning, since hard data is generally noisier and costlier to collect. Our experiments use open models up to 70b in size and four publicly available question-answering datasets with questions ranging in difficulty from 3rd grade science questions to college level STEM questions and general-knowledge trivia. We conclude that easy-to-hard generalization in LMs is surprisingly strong for the tasks studied. Our code is available at: https://github.com/allenai/easy-to-hard-generalization"
                    },
                    {
                        "title": "Moving Beyond the Turing Test with the Allen AI Science Challenge",
                        "abstract": "Given recent successes in AI (e.g., AlphaGo's victory against Lee Sedol in the game of GO), it's become increasingly important to assess: how close are AI systems to human-level intelligence? This paper describes the Allen AI Science Challenge---an approach towards that goal which led to a unique Kaggle Competition, its results, the lessons learned, and our next steps."
                    },
                    {
                        "title": "Do Dogs have Whiskers? A New Knowledge Base of hasPart Relations",
                        "abstract": "We present a new knowledge-base of hasPart relationships, extracted from a large corpus of generic statements. Complementary to other resources available, it is the first which is all three of: accurate (90% precision), salient (covers relationships a person may mention), and has high coverage of common terms (approximated as within a 10 year old's vocabulary), as well as having several times more hasPart entries than in the popular ontologies ConceptNet and WordNet. In addition, it contains information about quantifiers, argument modifiers, and links the entities to appropriate concepts in Wikipedia and WordNet. The knowledge base is available at https://allenai.org/data/haspartkb"
                    },
                    {
                        "title": "I'm in AGNi: A new standard for AGN pluralisation",
                        "abstract": "We present a new standard acronym for Active Galactic Nuclei, finally settling the argument of AGN vs. AGNs. Our new standard is not only etymologically superior (following the consensus set by SNe), but also boasts other linguistic opportunities, connecting strongly with relevant theology and streamlining descriptions of AGN properties."
                    },
                    {
                        "title": "Learning to Explain: Datasets and Models for Identifying Valid Reasoning Chains in Multihop Question-Answering",
                        "abstract": "Despite the rapid progress in multihop question-answering (QA), models still have trouble explaining why an answer is correct, with limited explanation training data available to learn from. To address this, we introduce three explanation datasets in which explanations formed from corpus facts are annotated. Our first dataset, eQASC, contains over 98K explanation annotations for the multihop question answering dataset QASC, and is the first that annotates multiple candidate explanations for each answer. The second dataset eQASC-perturbed is constructed by crowd-sourcing perturbations (while preserving their validity) of a subset of explanations in QASC, to test consistency and generalization of explanation prediction models. The third dataset eOBQA is constructed by adding explanation annotations to the OBQA dataset to test generalization of models trained on eQASC. We show that this data can be used to significantly improve explanation quality (+14% absolute F1 over a strong retrieval baseline) using a BERT-based classifier, but still behind the upper bound, offering a new challenge for future research. We also explore a delexicalized chain representation in which repeated noun phrases are replaced by variables, thus turning them into generalized reasoning chains (for example: \"X is a Y\" AND \"Y has Z\" IMPLIES \"X has Z\"). We find that generalized chains maintain performance while also being more robust to certain perturbations."
                    },
                    {
                        "title": "What Happened? Leveraging VerbNet to Predict the Effects of Actions in Procedural Text",
                        "abstract": "Our goal is to answer questions about paragraphs describing processes (e.g., photosynthesis). Texts of this genre are challenging because the effects of actions are often implicit (unstated), requiring background knowledge and inference to reason about the changing world states. To supply this knowledge, we leverage VerbNet to build a rulebase (called the Semantic Lexicon) of the preconditions and effects of actions, and use it along with commonsense knowledge of persistence to answer questions about change. Our evaluation shows that our system, ProComp, significantly outperforms two strong reading comprehension (RC) baselines. Our contributions are two-fold: the Semantic Lexicon rulebase itself, and a demonstration of how a simulation-based approach to machine reading can outperform RC methods that rely on surface cues alone.   Since this work was performed, we have developed neural systems that outperform ProComp, described elsewhere (Dalvi et al., NAACL'18). However, the Semantic Lexicon remains a novel and potentially useful resource, and its integration with neural systems remains a currently unexplored opportunity for further improvements in machine reading about processes."
                    },
                    {
                        "title": "QuaRTz: An Open-Domain Dataset of Qualitative Relationship Questions",
                        "abstract": "We introduce the first open-domain dataset, called QuaRTz, for reasoning about textual qualitative relationships. QuaRTz contains general qualitative statements, e.g., \"A sunscreen with a higher SPF protects the skin longer.\", twinned with 3864 crowdsourced situated questions, e.g., \"Billy is wearing sunscreen with a lower SPF than Lucy. Who will be best protected from the sun?\", plus annotations of the properties being compared. Unlike previous datasets, the general knowledge is textual and not tied to a fixed set of relationships, and tests a system's ability to comprehend and apply textual qualitative knowledge in a novel setting. We find state-of-the-art results are substantially (20%) below human performance, presenting an open challenge to the NLP community."
                    },
                    {
                        "title": "Transformers as Soft Reasoners over Language",
                        "abstract": "Beginning with McCarthy's Advice Taker (1959), AI has pursued the goal of providing a system with explicit, general knowledge and having the system reason over that knowledge. However, expressing the knowledge in a formal (logical or probabilistic) representation has been a major obstacle to this research. This paper investigates a modern approach to this problem where the facts and rules are provided as natural language sentences, thus bypassing a formal representation. We train transformers to reason (or emulate reasoning) over these sentences using synthetically generated data. Our models, that we call RuleTakers, provide the first empirical demonstration that this kind of soft reasoning over language is learnable, can achieve high (99%) accuracy, and generalizes to test data requiring substantially deeper chaining than seen during training (95%+ scores). We also demonstrate that the models transfer well to two hand-authored rulebases, and to rulebases paraphrased into more natural language. These findings are significant as it suggests a new role for transformers, namely as limited \"soft theorem provers\" operating over explicit theories in language. This in turn suggests new possibilities for explainability, correctability, and counterfactual reasoning in question-answering."
                    },
                    {
                        "title": "GenericsKB: A Knowledge Base of Generic Statements",
                        "abstract": "We present a new resource for the NLP community, namely a large (3.5M+ sentence) knowledge base of *generic statements*, e.g., \"Trees remove carbon dioxide from the atmosphere\", collected from multiple corpora. This is the first large resource to contain *naturally occurring* generic sentences, as opposed to extracted or crowdsourced triples, and thus is rich in high-quality, general, semantically complete statements. All GenericsKB sentences are annotated with their topical term, surrounding context (sentences), and a (learned) confidence. We also release GenericsKB-Best (1M+ sentences), containing the best-quality generics in GenericsKB augmented with selected, synthesized generics from WordNet and ConceptNet. In tests on two existing datasets requiring multihop reasoning (OBQA and QASC), we find using GenericsKB can result in higher scores and better explanations than using a much larger corpus. This demonstrates that GenericsKB can be a useful resource for NLP applications, as well as providing data for linguistic studies of generics and their semantics. GenericsKB is available at https://allenai.org/data/genericskb."
                    },
                    {
                        "title": "ProofWriter: Generating Implications, Proofs, and Abductive Statements over Natural Language",
                        "abstract": "Transformers have been shown to emulate logical deduction over natural language theories (logical rules expressed in natural language), reliably assigning true/false labels to candidate implications. However, their ability to generate implications of a theory has not yet been demonstrated, and methods for reconstructing proofs of answers are imperfect. In this work we show that a generative model, called ProofWriter, can reliably generate both implications of a theory and the natural language proof(s) that support them. In particular, iterating a 1-step implication generator results in proofs that are highly reliable, and represent actual model decisions (rather than post-hoc rationalizations). On the RuleTaker dataset, the accuracy of ProofWriter's proofs exceed previous methods by +9% absolute, and in a way that generalizes to proof depths unseen in training and on out-of-domain problems. We also show that generative techniques can perform a type of abduction with high precision: Given a theory and an unprovable conclusion, identify a missing fact that allows the conclusion to be proved, along with a proof. These results significantly improve the viability of neural methods for systematically reasoning over natural language."
                    }
                ]
            }
        }
    },
    "2402.11100": {
        "paper_data": {
            "title": "When LLMs Meet Cunning Texts: A Fallacy Understanding Benchmark for Large Language Models",
            "url": "http://arxiv.org/abs/2402.11100v2",
            "arxiv_id": "2402.11100",
            "authors": [
                "Yinghui Li",
                "Qingyu Zhou",
                "Yuanzhen Luo",
                "Shirong Ma",
                "Yangning Li",
                "Hai-Tao Zheng",
                "Xuming Hu",
                "Philip S. Yu"
            ],
            "abstract": "Recently, Large Language Models (LLMs) make remarkable evolutions in language understanding and generation. Following this, various benchmarks for measuring all kinds of capabilities of LLMs have sprung up. In this paper, we challenge the reasoning and understanding abilities of LLMs by proposing a FaLlacy Understanding Benchmark (FLUB) containing cunning texts that are easy for humans to understand but difficult for models to grasp. Specifically, the cunning texts that FLUB focuses on mainly consist of the tricky, humorous, and misleading texts collected from the real internet environment. And we design three tasks with increasing difficulty in the FLUB benchmark to evaluate the fallacy understanding ability of LLMs. Based on FLUB, we investigate the performance of multiple representative and advanced LLMs, reflecting our FLUB is challenging and worthy of more future study. Interesting discoveries and valuable insights are achieved in our extensive experiments and detailed analyses. We hope that our benchmark can encourage the community to improve LLMs' ability to understand fallacies. Our data and codes are available at https://github.com/THUKElab/FLUB.",
            "introduction": "   1 Introduction  Large Language Models (LLMs) have shown great abilities to understand human languages, including information extraction\u00a0[1], text correction\u00a0[2], humor understanding\u00a0[3], etc. Researchers have constructed numerous benchmarks to evaluate LLMs in various aspects\u00a0[4, 5, 6, 7, 8]. By using constructed benchmarks to interact with LLMs, researchers can analyze the behavior of LLMs to compare the performance of different LLMs and study how to further improve LLMs in a targeted manner.   Although many LLM benchmarks have sprung up, we believe that existing benchmarks are not challenging enough to truly measure the human-like intelligence of LLMs. In particular, we are still wondering whether LLMs can understand cunning texts that may contain misleading, wrong premise, intentional ambiguity, and so forth, considering that almost all LLMs are trained on \u201ccleaned\u201d and \u201ccorrect\u201d corpora. Therefore, we build a FaLlacy Understanding Benchmark (FLUB) to challenge LLMs for solving these problems.      (a) The examples of how LLMs and humans perform when faced with cunning texts. The LLM we use is ChatGPT-3.5 on Jan 23, 2024.     (b) We design three tasks, namely Cunning Type Classification, Fallacy Explanation, and Answer Selection (i.e., Multiple Choice).    Figure 1: The running examples and annotation examples of FLUB.   Figure\u00a01(a) shows the running examples from FLUB. From these cases, we directly feel the different behaviors of LLMs and humans when facing cunning texts. In the first example, LLMs ignore the common sense that the lotus root itself has many holes in its structure and fall into the trap of the cunning text, wrongly judging that the holes in the lotus root are caused by insect infestation. In the second example, LLMs fail to see the logic that depositing money into random ATMs does not create problems and therefore give an answer that seems reasonable but is absurdly laughable. In fact, these cunning texts for LLMs are very easy to handle for human intelligence. Therefore, it is very urgent and meaningful to construct a benchmark composed of cunning texts to evaluate and thereby promote the improvement of LLMs\u2019 fallacy understanding capabilities.   Inspired by the above motivation, we collect real cunning texts as our raw data from a famous Chinese online forum, the \u201cRuozhiba\u201d (retard forum)\u00a0111https://tieba.baidu.com/f?kw=%E5%BC%B1%E6%99%BA&ie=utf-8. This forum is popular for its cunning and unreasonable posts, which are generally easy for humans to understand but challenging for LLMs. The characteristics of the posts contained in this forum are consistent with our research motivation, so choosing it as the data source well supports FLUB\u2019s evaluation of LLMs\u2019 fallacy understanding ability. After data cleaning and annotating of cunning types, FLUB has 8 fine-grained types of cunning texts and most of the texts in FLUB fall into two types of fallacy, namely, faulty reasoning and word game. Moreover, we also manually annotated one correct answer (i.e., the explanation of the cunning text) and three confusing wrong answers for each input text in FLUB, as shown in Figure\u00a01(b).   Based on our constructed FLUB and its annotation information, we design three tasks with increasing difficulty to test whether the LLMs can understand the fallacy and solve the \u201ccunning\u201d texts. Specifically, (1) Answer Selection: The model is asked to select the correct one from the four answers provided",
            "references": [
                {
                    "title": "COIG-CQIA: Quality is All You Need for Chinese Instruction Fine-tuning",
                    "abstract": "Recently, there have been significant advancements in large language models (LLMs), particularly focused on the English language. These advancements have enabled these LLMs to understand and execute complex instructions with unprecedented accuracy and fluency. However, despite these advancements, there remains a noticeable gap in the development of Chinese instruction tuning. The unique linguistic features and cultural depth of the Chinese language pose challenges for instruction tuning tasks. Existing datasets are either derived from English-centric LLMs or are ill-suited for aligning with the interaction patterns of real-world Chinese users. To bridge this gap, we introduce COIG-CQIA, a high-quality Chinese instruction tuning dataset. Our aim is to build a diverse, wide-ranging instruction-tuning dataset to better align model behavior with human interactions. To this end, we collect a high-quality human-written corpus from various sources on the Chinese Internet, including Q&A communities, Wikis, examinations, and existing NLP datasets. This corpus was rigorously filtered and carefully processed to form the COIG-CQIA dataset. Furthermore, we train models of various scales on different subsets of CQIA, following in-depth evaluation and analyses. The findings from our experiments offer valuable insights for selecting and developing Chinese instruction-tuning datasets. We also find that models trained on CQIA-Subset achieve competitive results in human assessment as well as knowledge and security benchmarks. Data are available at https://huggingface.co/datasets/m-a-p/COIG-CQIA"
                },
                {
                    "title": "Towards Real-World Writing Assistance: A Chinese Character Checking Benchmark with Faked and Misspelled Characters",
                    "abstract": "Writing assistance is an application closely related to human life and is also a fundamental Natural Language Processing (NLP) research field. Its aim is to improve the correctness and quality of input texts, with character checking being crucial in detecting and correcting wrong characters. From the perspective of the real world where handwriting occupies the vast majority, characters that humans get wrong include faked characters (i.e., untrue characters created due to writing errors) and misspelled characters (i.e., true characters used incorrectly due to spelling errors). However, existing datasets and related studies only focus on misspelled characters mainly caused by phonological or visual confusion, thereby ignoring faked characters which are more common and difficult. To break through this dilemma, we present Visual-C$^3$, a human-annotated Visual Chinese Character Checking dataset with faked and misspelled Chinese characters. To the best of our knowledge, Visual-C$^3$ is the first real-world visual and the largest human-crafted dataset for the Chinese character checking scenario. Additionally, we also propose and evaluate novel baseline methods on Visual-C$^3$. Extensive empirical results and analyses show that Visual-C$^3$ is high-quality yet challenging. The Visual-C$^3$ dataset and the baseline methods will be publicly available to facilitate further research in the community."
                },
                {
                    "title": "Evaluating Large Language Models: A Comprehensive Survey",
                    "abstract": "Large language models (LLMs) have demonstrated remarkable capabilities across a broad spectrum of tasks. They have attracted significant attention and been deployed in numerous downstream applications. Nevertheless, akin to a double-edged sword, LLMs also present potential risks. They could suffer from private data leaks or yield inappropriate, harmful, or misleading content. Additionally, the rapid progress of LLMs raises concerns about the potential emergence of superintelligent systems without adequate safeguards. To effectively capitalize on LLM capacities as well as ensure their safe and beneficial development, it is critical to conduct a rigorous and comprehensive evaluation of LLMs. This survey endeavors to offer a panoramic perspective on the evaluation of LLMs. We categorize the evaluation of LLMs into three major groups: knowledge and capability evaluation, alignment evaluation and safety evaluation. In addition to the comprehensive review on the evaluation methodologies and benchmarks on these three aspects, we collate a compendium of evaluations pertaining to LLMs' performance in specialized domains, and discuss the construction of comprehensive evaluation platforms that cover LLM evaluations on capabilities, alignment, safety, and applicability. We hope that this comprehensive overview will stimulate further research interests in the evaluation of LLMs, with the ultimate goal of making evaluation serve as a cornerstone in guiding the responsible development of LLMs. We envision that this will channel their evolution into a direction that maximizes societal benefit while minimizing potential risks. A curated list of related papers has been publicly available at https://github.com/tjunlp-lab/Awesome-LLMs-Evaluation-Papers."
                },
                {
                    "title": "FreshLLMs: Refreshing Large Language Models with Search Engine Augmentation",
                    "abstract": "Most large language models (LLMs) are trained once and never updated; thus, they lack the ability to dynamically adapt to our ever-changing world. In this work, we perform a detailed study of the factuality of LLM-generated text in the context of answering questions that test current world knowledge. Specifically, we introduce FreshQA, a novel dynamic QA benchmark encompassing a diverse range of question and answer types, including questions that require fast-changing world knowledge as well as questions with false premises that need to be debunked. We benchmark a diverse array of both closed and open-source LLMs under a two-mode evaluation procedure that allows us to measure both correctness and hallucination. Through human evaluations involving more than 50K judgments, we shed light on limitations of these models and demonstrate significant room for improvement: for instance, all models (regardless of model size) struggle on questions that involve fast-changing knowledge and false premises. Motivated by these results, we present FreshPrompt, a simple few-shot prompting method that substantially boosts the performance of an LLM on FreshQA by incorporating relevant and up-to-date information retrieved from a search engine into the prompt. Our experiments show that FreshPrompt outperforms both competing search engine-augmented prompting methods such as Self-Ask (Press et al., 2022) as well as commercial systems such as Perplexity.AI. Further analysis of FreshPrompt reveals that both the number of retrieved evidences and their order play a key role in influencing the correctness of LLM-generated answers. Additionally, instructing the LLM to generate concise and direct answers helps reduce hallucination compared to encouraging more verbose answers. To facilitate future work, we release FreshQA at github.com/freshllms/freshqa and commit to updating it at regular intervals."
                },
                {
                    "title": "Qwen Technical Report",
                    "abstract": "Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural language processing tasks that were previously thought to be exclusive to humans. In this work, we introduce Qwen, the first installment of our large language model series. Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts. It includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques. The base language models consistently demonstrate superior performance across a multitude of downstream tasks, and the chat models, particularly those trained using Reinforcement Learning from Human Feedback (RLHF), are highly competitive. The chat models possess advanced tool-use and planning capabilities for creating agent applications, showcasing impressive performance even when compared to bigger models on complex tasks like utilizing a code interpreter. Furthermore, we have developed coding-specialized models, Code-Qwen and Code-Qwen-Chat, as well as mathematics-focused models, Math-Qwen-Chat, which are built upon base language models. These models demonstrate significantly improved performance in comparison with open-source models, and slightly fall behind the proprietary models."
                },
                {
                    "title": "Baichuan 2: Open Large-scale Language Models",
                    "abstract": "Large language models (LLMs) have demonstrated remarkable performance on a variety of natural language tasks based on just a few examples of natural language instructions, reducing the need for extensive feature engineering. However, most powerful LLMs are closed-source or limited in their capability for languages other than English. In this technical report, we present Baichuan 2, a series of large-scale multilingual language models containing 7 billion and 13 billion parameters, trained from scratch, on 2.6 trillion tokens. Baichuan 2 matches or outperforms other open-source models of similar size on public benchmarks like MMLU, CMMLU, GSM8K, and HumanEval. Furthermore, Baichuan 2 excels in vertical domains such as medicine and law. We will release all pre-training model checkpoints to benefit the research community in better understanding the training dynamics of Baichuan 2."
                },
                {
                    "title": "LatEval: An Interactive LLMs Evaluation Benchmark with Incomplete Information from Lateral Thinking Puzzles",
                    "abstract": "With the evolution of LLMs, they are endowed with impressive logical reasoning, or vertical thinking capabilities. But can they think out of the box? Do they possess proficient lateral thinking abilities? Following the setup of Lateral Thinking Puzzles, we propose a novel evaluation benchmark, LatEval, which assesses the model\u2019s lateral thinking within an interactive framework. In our benchmark, we challenge LLMs with 2 aspects: (1) posing high-quality questions that break out of conventional norms but are beneficial for puzzle-solving. (2) integrating existing information to gradually deduce the truth through reasoning. We observe that it is hard for most LLMs to accomplish lateral thinking during interactions. Even the most powerful LLM, GPT-4, faces challenges in achieving satisfactory performance, and for most open-source models, simply completing this task is quite difficult. This evaluation benchmark provides LLMs with a highly challenging and differentiating task that is crucial to an effective AI assistant. Our dataset and source codes are available at https://github.com/THUKElab/LatEval."
                },
                {
                    "title": "SeqGPT: An Out-of-the-box Large Language Model for Open Domain Sequence Understanding",
                    "abstract": "Large language models (LLMs) have shown impressive abilities for open-domain NLP tasks. However, LLMs are sometimes too footloose for natural language understanding (NLU) tasks which always have restricted output and input format. Their performances on NLU tasks are highly related to prompts or demonstrations and are shown to be poor at performing several representative NLU tasks, such as event extraction and entity typing. To this end, we present SeqGPT, a bilingual (i.e., English and Chinese) open-source autoregressive model specially enhanced for open-domain natural language understanding. We express all NLU tasks with two atomic tasks, which define fixed instructions to restrict the input and output format but still ``open'' for arbitrarily varied label sets. The model is first instruction-tuned with extremely fine-grained labeled data synthesized by ChatGPT and then further fine-tuned by 233 different atomic tasks from 152 datasets across various domains. The experimental results show that SeqGPT has decent classification and extraction ability, and is capable of performing language understanding tasks on unseen domains. We also conduct empirical studies on the scaling of data and model size as well as on the transfer across tasks. Our models are accessible at https://github.com/Alibaba-NLP/SeqGPT."
                },
                {
                    "title": "MESED: A Multi-modal Entity Set Expansion Dataset with Fine-grained Semantic Classes and Hard Negative Entities",
                    "abstract": "The Entity Set Expansion (ESE) task aims to expand a handful of seed entities with new entities belonging to the same semantic class. Conventional ESE methods are based on mono-modality (i.e., literal modality), which struggle to deal with complex entities in the real world such as (1) Negative entities with fine-grained semantic differences. (2) Synonymous entities. (3) Polysemous entities. (4) Long-tailed entities. These challenges prompt us to propose novel Multi-modal Entity Set Expansion (MESE), where models integrate information from multiple modalities to represent entities. Intuitively, the benefits of multi-modal information for ESE are threefold: (1) Different modalities can provide complementary information. (2) Multi-modal information provides a unified signal via common visual properties for the same semantic class or entity. (3) Multi-modal information offers robust alignment signals for synonymous entities. To assess model performance in MESE, we constructed the MESED dataset which is the first multi-modal dataset for ESE with large-scale and elaborate manual calibration. A powerful multi-modal model MultiExpan is proposed which is pre-trained on four multimodal pre-training tasks. The extensive experiments and analyses on MESED demonstrate the high quality of the dataset and the effectiveness of our MultiExpan, as well as pointing the direction for future research. The benchmark and code are public at https://github.com/THUKElab/MESED."
                },
                {
                    "title": "On the (In)Effectiveness of Large Language Models for Chinese Text Correction",
                    "abstract": "Recently, the development and progress of Large Language Models (LLMs) have amazed the entire Artificial Intelligence community. Benefiting from their emergent abilities, LLMs have attracted more and more researchers to study their capabilities and performance on various downstream Natural Language Processing (NLP) tasks. While marveling at LLMs' incredible performance on all kinds of tasks, we notice that they also have excellent multilingual processing capabilities, such as Chinese. To explore the Chinese processing ability of LLMs, we focus on Chinese Text Correction, a fundamental and challenging Chinese NLP task. Specifically, we evaluate various representative LLMs on the Chinese Grammatical Error Correction (CGEC) and Chinese Spelling Check (CSC) tasks, which are two main Chinese Text Correction scenarios. Additionally, we also fine-tune LLMs for Chinese Text Correction to better observe the potential capabilities of LLMs. From extensive analyses and comparisons with previous state-of-the-art small models, we empirically find that the LLMs currently have both amazing performance and unsatisfactory behavior for Chinese Text Correction. We believe our findings will promote the landing and application of LLMs in the Chinese NLP community."
                },
                {
                    "title": "A Survey on Evaluation of Large Language Models",
                    "abstract": "Large language models (LLMs) are gaining increasing popularity in both academia and industry, owing to their unprecedented performance in various applications. As LLMs continue to play a vital role in both research and daily use, their evaluation becomes increasingly critical, not only at the task level, but also at the society level for better understanding of their potential risks. Over the past years, significant efforts have been made to examine LLMs from various perspectives. This paper presents a comprehensive review of these evaluation methods for LLMs, focusing on three key dimensions: what to evaluate, where to evaluate, and how to evaluate. Firstly, we provide an overview from the perspective of evaluation tasks, encompassing general natural language processing tasks, reasoning, medical usage, ethics, education, natural and social sciences, agent applications, and other areas. Secondly, we answer the \u2018where\u2019 and \u2018how\u2019 questions by diving into the evaluation methods and benchmarks, which serve as crucial components in assessing the performance of LLMs. Then, we summarize the success and failure cases of LLMs in different tasks. Finally, we shed light on several future challenges that lie ahead in LLMs evaluation. Our aim is to offer invaluable insights to researchers in the realm of LLMs evaluation, thereby aiding the development of more proficient LLMs. Our key point is that evaluation should be treated as an essential discipline to better assist the development of LLMs. We consistently maintain the related open-source materials at: https://github.com/MLGroupJLU/LLM-eval-survey"
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "Evaluating the Logical Reasoning Ability of ChatGPT and GPT-4",
                    "abstract": "Harnessing logical reasoning ability is a comprehensive natural language understanding endeavor. With the release of Generative Pretrained Transformer 4 (GPT-4), highlighted as\"advanced\"at reasoning tasks, we are eager to learn the GPT-4 performance on various logical reasoning tasks. This report analyses multiple logical reasoning datasets, with popular benchmarks like LogiQA and ReClor, and newly-released datasets like AR-LSAT. We test the multi-choice reading comprehension and natural language inference tasks with benchmarks requiring logical reasoning. We further construct a logical reasoning out-of-distribution dataset to investigate the robustness of ChatGPT and GPT-4. We also make a performance comparison between ChatGPT and GPT-4. Experiment results show that ChatGPT performs significantly better than the RoBERTa fine-tuning method on most logical reasoning benchmarks. With early access to the GPT-4 API we are able to conduct intense experiments on the GPT-4 model. The results show GPT-4 yields even higher performance on most logical reasoning datasets. Among benchmarks, ChatGPT and GPT-4 do relatively well on well-known datasets like LogiQA and ReClor. However, the performance drops significantly when handling newly released and out-of-distribution datasets. Logical reasoning remains challenging for ChatGPT and GPT-4, especially on out-of-distribution and natural language inference datasets. We release the prompt-style logical reasoning datasets as a benchmark suite and name it LogiEval."
                },
                {
                    "title": "ChatGPT Is a Knowledgeable but Inexperienced Solver: An Investigation of Commonsense Problem in Large Language Models",
                    "abstract": "Large language models (LLMs) have made significant progress in NLP. However, their ability to memorize, represent, and leverage commonsense knowledge has been a well-known pain point. In this paper, we specifically focus on ChatGPT, a widely used and easily accessible LLM, and ask the following questions: (1) Can ChatGPT effectively answer commonsense questions? (2) Is ChatGPT aware of the underlying commonsense knowledge for answering a specific question? (3) Is ChatGPT knowledgeable in commonsense? (4) Can ChatGPT effectively leverage commonsense for answering questions? We conduct a series of experiments on 11 datasets to evaluate ChatGPT\u2019s commonsense abilities, including answering commonsense questions, identifying necessary knowledge, generating knowledge descriptions, and using knowledge descriptions to answer questions again. Experimental results show that: (1) ChatGPT can achieve good QA accuracies in commonsense tasks, while still struggling with certain domains of datasets. (2) ChatGPT is knowledgeable, and can accurately generate most of the commonsense knowledge using knowledge prompts. (3) Despite its knowledge, ChatGPT is an inexperienced commonsense problem solver, which cannot precisely identify the needed commonsense for answering a specific question. These findings raise the need to explore improved mechanisms for effectively incorporating commonsense into LLMs like ChatGPT, such as better instruction following and commonsense guidance."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "A Multitask, Multilingual, Multimodal Evaluation of ChatGPT on Reasoning, Hallucination, and Interactivity",
                    "abstract": "This paper proposes a framework for quantitatively evaluating interactive LLMs such as ChatGPT using publicly available data sets. We carry out an extensive technical evaluation of ChatGPT using 23 data sets covering 8 different common NLP application tasks. We evaluate the multitask, multilingual and multi-modal aspects of ChatGPT based on these data sets and a newly designed multimodal dataset. We find that ChatGPT outperforms LLMs with zero-shot learning on most tasks and even outperforms fine-tuned models on some tasks. We find that it is better at understanding non-Latin script languages than generating them. It is able to generate multimodal content from textual prompts, via an intermediate code generation step. Moreover, we find that ChatGPT is 63.41% accurate on average in 10 different reasoning categories under logical reasoning, non-textual reasoning, and commonsense reasoning, hence making it an unreliable reasoner. It is, for example, better at deductive than inductive reasoning. ChatGPT suffers from hallucination problems like other LLMs and it generates more extrinsic hallucinations from its parametric memory as it does not have access to an external knowledge base. Finally, the interactive feature of ChatGPT enables human collaboration with the underlying LLM to improve its performance, i.e, 8% ROUGE-1 on summarization and 2% ChrF++ on machine translation, in a multi-turn\"prompt engineering\"fashion. We also release codebase for evaluation set extraction."
                },
                {
                    "title": "Towards Attribute-Entangled Controllable Text Generation: A Pilot Study of Blessing Generation",
                    "abstract": "Controllable Text Generation (CTG) has obtained great success due to its fine-grained generation ability obtained by focusing on multiple attributes. However, most existing CTG researches overlook how to utilize the attribute entanglement to enhance the diversity of the controlled generated texts. Facing this dilemma, we focus on a novel CTG scenario, i.e., blessing generation which is challenging because high-quality blessing texts require CTG models to comprehensively consider the entanglement between multiple attributes (e.g., objects and occasions). To promote the research on blessing generation, we present EBleT, a large-scale Entangled Blessing Text dataset containing 293K English sentences annotated with multiple attributes. Furthermore, we propose novel evaluation metrics to measure the quality of the blessing texts generated by the baseline models we designed. Our study opens a new research direction for controllable text generation and enables the development of attribute-entangled CTG models."
                },
                {
                    "title": "Linguistic Rules-Based Corpus Generation for Native Chinese Grammatical Error Correction",
                    "abstract": "Chinese Grammatical Error Correction (CGEC) is both a challenging NLP task and a common application in human daily life. Recently, many data-driven approaches are proposed for the development of CGEC research. However, there are two major limitations in the CGEC field: First, the lack of high-quality annotated training corpora prevents the performance of existing CGEC models from being significantly improved. Second, the grammatical errors in widely used test sets are not made by native Chinese speakers, resulting in a significant gap between the CGEC models and the real application. In this paper, we propose a linguistic rules-based approach to construct large-scale CGEC training corpora with automatically generated grammatical errors. Additionally, we present a challenging CGEC benchmark derived entirely from errors made by native Chinese speakers in real-world scenarios. Extensive experiments and detailed analyses not only demonstrate that the training data constructed by our method effectively improves the performance of CGEC models, but also reflect that our benchmark is an excellent resource for further development of the CGEC field."
                },
                {
                    "title": "Do Androids Laugh at Electric Sheep? Humor \u201cUnderstanding\u201d Benchmarks from The New Yorker Caption Contest",
                    "abstract": "Large neural networks can now generate jokes, but do they really \u201cunderstand\u201d humor? We challenge AI models with three tasks derived from the New Yorker Cartoon Caption Contest: matching a joke to a cartoon, identifying a winning caption, and explaining why a winning caption is funny. These tasks encapsulate progressively more sophisticated aspects of \u201cunderstanding\u201d a cartoon; key elements are the complex, often surprising relationships between images and captions and the frequent inclusion of indirect and playful allusions to human experience and culture. We investigate both multimodal and language-only models: the former are challenged with the cartoon images directly, while the latter are given multifaceted descriptions of the visual scene to simulate human-level visual understanding. We find that both types of models struggle at all three tasks. For example, our best multimodal models fall 30 accuracy points behind human performance on the matching task, and, even when provided ground-truth visual scene descriptors, human-authored explanations are preferred head-to-head over the best machine-authored ones (few-shot GPT-4) in more than 2/3 of cases. We release models, code, leaderboard, and corpus, which includes newly-gathered annotations describing the image\u2019s locations/entities, what\u2019s unusual in the scene, and an explanation of the joke."
                },
                {
                    "title": "The Past Mistake is the Future Wisdom: Error-driven Contrastive Probability Optimization for Chinese Spell Checking",
                    "abstract": "Chinese Spell Checking (CSC) aims to detect and correct Chinese spelling errors, which are mainly caused by the phonological or visual similarity. Recently, pre-trained language models (PLMs) promote the progress of CSC task. However, there exists a gap between the learned knowledge of PLMs and the goal of CSC task. PLMs focus on the semantics in text and tend to correct the erroneous characters to semantically proper or commonly used ones, but these aren\u2019t the ground-truth corrections. To address this issue, we propose an Error-driven COntrastive Probability Optimization (ECOPO) framework for CSC task. ECOPO refines the knowledge representations of PLMs, and guides the model to avoid predicting these common characters through an error-driven way. Particularly, ECOPO is model-agnostic and it can be combined with existing CSC methods to achieve better performance. Extensive experiments and detailed analyses on SIGHAN datasets demonstrate that ECOPO is simple yet effective."
                },
                {
                    "title": "TruthfulQA: Measuring How Models Mimic Human Falsehoods",
                    "abstract": "We propose a benchmark to measure whether a language model is truthful in generating answers to questions. The benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics. We crafted questions that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. We tested GPT-3, GPT-Neo/J, GPT-2 and a T5-based model. The best model was truthful on 58% of questions, while human performance was 94%. Models generated many false answers that mimic popular misconceptions and have the potential to deceive humans. The largest models were generally the least truthful. This contrasts with other NLP tasks, where performance improves with model size. However, this result is expected if false answers are learned from the training distribution. We suggest that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web."
                },
                {
                    "title": "From LSAT: The Progress and Challenges of Complex Reasoning",
                    "abstract": "Complex reasoning aims to draw a correct inference based on complex rules. As a hallmark of human intelligence, it involves a degree of explicit reading comprehension, interpretation of logical knowledge and complex rule application. In this paper, we take a step forward in complex reasoning by systematically studying the three challenging and domain-general tasks of the Law School Admission Test (LSAT), including analytical reasoning, logical reasoning and reading comprehension. We propose a hybrid reasoning system to integrate these three tasks and achieve impressive overall performance on the LSAT tests. The experimental results demonstrate that our system endows itself a certain complex reasoning ability, especially the fundamental reading comprehension and challenging logical reasoning capacities. Further analysis also shows the effectiveness of combining the pre-trained models with the task-specific reasoning module, and integrating symbolic knowledge into discrete interpretable reasoning steps in complex reasoning. We further shed a light on the potential future directions, like unsupervised symbolic knowledge extraction, model interpretability, few-shot learning and comprehensive benchmark for complex reasoning."
                },
                {
                    "title": "GLM: General Language Model Pretraining with Autoregressive Blank Infilling",
                    "abstract": "There have been various types of pretraining architectures including autoencoding models (e.g., BERT), autoregressive models (e.g., GPT), and encoder-decoder models (e.g., T5). However, none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding (NLU), unconditional generation, and conditional generation. We propose a General Language Model (GLM) based on autoregressive blank infilling to address this challenge. GLM improves blank filling pretraining by adding 2D positional encodings and allowing an arbitrary order to predict spans, which results in performance gains over BERT and T5 on NLU tasks. Meanwhile, GLM can be pretrained for different types of tasks by varying the number and lengths of blanks. On a wide range of tasks across NLU, conditional and unconditional generation, GLM outperforms BERT, T5, and GPT given the same model sizes and data, and achieves the best performance from a single pretrained model with 1.25\u00d7 parameters of BERT Large , demonstrating its generalizability to different downstream tasks."
                },
                {
                    "title": "Natural Language Inference in Context - Investigating Contextual Reasoning over Long Texts",
                    "abstract": "Natural language inference (NLI) is a fundamental NLP task, investigating the entailment relationship between two texts.\u2028 Popular NLI datasets present the task at sentence-level. While adequate for testing semantic representations, they fall short for testing contextual reasoning over long texts, which is a natural part of the human inference process. We introduce ConTRoL, a new dataset for ConTextual Reasoning over Long texts. Consisting of 8,325 expert-designed \"context-hypothesis\" pairs with gold labels, ConTRoL is a passage-level NLI dataset with a focus on complex contextual reasoning types such as logical reasoning. It is derived from competitive selection and recruitment test (verbal reasoning test) for police recruitment, with expert level quality. Compared with previous NLI benchmarks, the materials in ConTRoL are much more challenging, involving a range of reasoning types. Empirical results show that state-of-the-art language models perform by far worse than educated humans. Our dataset can also serve as a testing-set for downstream tasks like checking the factual correctness of summaries."
                },
                {
                    "title": "ConjNLI: Natural Language Inference over Conjunctive Sentences",
                    "abstract": "Reasoning about conjuncts in conjunctive sentences is important for a deeper understanding of conjunctions in English and also how their usages and semantics differ from conjunctive and disjunctive boolean logic. Existing NLI stress tests do not consider non-boolean usages of conjunctions and use templates for testing such model knowledge. Hence, we introduce ConjNLI, a challenge stress-test for natural language inference over conjunctive sentences, where the premise differs from the hypothesis by conjuncts removed, added, or replaced. These sentences contain single and multiple instances of coordinating conjunctions (\"and\", \"or\", \"but\", \"nor\") with quantifiers, negations, and requiring diverse boolean and non-boolean inferences over conjuncts. We find that large-scale pre-trained language models like RoBERTa do not understand conjunctive semantics well and resort to shallow heuristics to make inferences over such sentences. As some initial solutions, we first present an iterative adversarial fine-tuning method that uses synthetically created training data based on boolean and non-boolean heuristics. We also propose a direct model advancement by making RoBERTa aware of predicate semantic roles. While we observe some performance gains, ConjNLI is still challenging for current methods, thus encouraging interesting future work for better understanding of conjunctions. Our data and code are publicly available at: this https URL"
                },
                {
                    "title": "LogiQA: A Challenge Dataset for Machine Reading Comprehension with Logical Reasoning",
                    "abstract": "Machine reading is a fundamental task for testing the capability of natural language understand- ing, which is closely related to human cognition in many aspects. With the rising of deep learning techniques, algorithmic models rival human performances on simple QA, and thus increasingly challenging machine reading datasets have been proposed. Though various challenges such as evidence integration and commonsense knowledge have been integrated, one of the fundamental capabilities in human reading, namely logical reasoning, is not fully investigated. We build a comprehensive dataset, named LogiQA, which is sourced from expert-written questions for testing human Logical reasoning. It consists of 8,678 QA instances, covering multiple types of deductive reasoning. Results show that state-of-the-art neural models perform by far worse than human ceiling. Our dataset can also serve as a benchmark for reinvestigating logical AI under the deep learning NLP setting. The dataset is freely available at https://github.com/lgw863/LogiQA-dataset."
                },
                {
                    "title": "PIQA: Reasoning about Physical Commonsense in Natural Language",
                    "abstract": "To apply eyeshadow without a brush, should I use a cotton swab or a toothpick? Questions requiring this kind of physical commonsense pose a challenge to today's natural language understanding systems. While recent pretrained models (such as BERT) have made progress on question answering over more abstract domains \u2013 such as news articles and encyclopedia entries, where text is plentiful \u2013 in more physical domains, text is inherently limited due to reporting bias. Can AI systems learn to reliably answer physical commonsense questions without experiencing the physical world?In this paper, we introduce the task of physical commonsense reasoning and a corresponding benchmark dataset Physical Interaction: Question Answering or PIQA. Though humans find the dataset easy (95% accuracy), large pretrained models struggle (\u223c75%). We provide analysis about the dimensions of knowledge that existing models lack, which offers significant opportunities for future research."
                },
                {
                    "title": "\u201cGoing on a vacation\u201d takes longer than \u201cGoing for a walk\u201d: A Study of Temporal Commonsense Understanding",
                    "abstract": "Understanding time is crucial for understanding events expressed in natural language. Because people rarely say the obvious, it is often necessary to have commonsense knowledge about various temporal aspects of events, such as duration, frequency, and temporal order. However, this important problem has so far received limited attention. This paper systematically studies this temporal commonsense problem. Specifically, we define five classes of temporal commonsense, and use crowdsourcing to develop a new dataset, MCTACO, that serves as a test set for this task. We find that the best current methods used on MCTACO are still far behind human performance, by about 20%, and discuss several directions for improvement. We hope that the new dataset and our study here can foster more future research on this topic."
                },
                {
                    "title": "HellaSwag: Can a Machine Really Finish Your Sentence?",
                    "abstract": "Recent work by Zellers et al. (2018) introduced a new task of commonsense natural language inference: given an event description such as \u201cA woman sits at a piano,\u201d a machine must select the most likely followup: \u201cShe sets her fingers on the keys.\u201d With the introduction of BERT, near human-level performance was reached. Does this mean that machines can perform human level commonsense inference? In this paper, we show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans (>95% accuracy), state-of-the-art models struggle (<48%). We achieve this via Adversarial Filtering (AF), a data collection paradigm wherein a series of discriminators iteratively select an adversarial set of machine-generated wrong answers. AF proves to be surprisingly robust. The key insight is to scale up the length and complexity of the dataset examples towards a critical \u2018Goldilocks\u2019 zone wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models. Our construction of HellaSwag, and its resulting difficulty, sheds light on the inner workings of deep pretrained models. More broadly, it suggests a new path forward for NLP research, in which benchmarks co-evolve with the evolving state-of-the-art in an adversarial way, so as to present ever-harder challenges."
                },
                {
                    "title": "Survey on Natural Language Generation",
                    "abstract": "In this paper, we are discussing the basic concepts and fundamentals of Natural Language Generation, a field in Natural Language Engineering that deals with the conversion of non-linguistic data into natural information. We will start our investigation by introducing the NLG system and its different types. We will also pin point the major differences between NLG and NLU also known as Natural Language Understanding. Afterwards, we will shed the light on the architecture of a basic NLG system, its advantages and disadvantages. Later, we will examine the different applications of NLG, showing a case study that illustrates how an NLG system operates from an algorithmic point of view. Finally, we will review some of the existing NLG systems together with their features, taken from the real world."
                },
                {
                    "title": "Humor Recognition Using Deep Learning",
                    "abstract": "Humor is an essential but most fascinating element in personal communication. How to build computational models to discover the structures of humor, recognize humor and even generate humor remains a challenge and there have been yet few attempts on it. In this paper, we construct and collect four datasets with distinct joke types in both English and Chinese and conduct learning experiments on humor recognition. We implement a Convolutional Neural Network (CNN) with extensive filter size, number and Highway Networks to increase the depth of networks. Results show that our model outperforms in recognition of different types of humor with benchmarks collected in both English and Chinese languages on accuracy, precision, and recall in comparison to previous works."
                },
                {
                    "title": "The measurement of observer agreement for categorical data.",
                    "abstract": "This paper presents a general statistical methodology for the analysis of multivariate categorical data arising from observer reliability studies. The procedure essentially involves the construction of functions of the observed proportions which are directed at the extent to which the observers agree among themselves and the construction of test statistics for hypotheses involving these functions. Tests for interobserver bias are presented in terms of first-order marginal homogeneity and measures of interobserver agreement are developed as generalized kappa-type statistics. These procedures are illustrated with a clinical diagnosis example from the epidemiological literature."
                },
                {
                    "title": "Exploring Humor in Natural Language Processing: A Comprehensive Review of JOKER Tasks at CLEF Symposium 2023",
                    "abstract": "As linguistic phenomena that showcase the richness and complexity of human language, puns pose significant challenges to natural language processing (NLP) systems. The significance of these tasks lies not only in the enhancement of humor recognition capabilities in AI but also in the improvement of machine translation systems, as puns often encapsulate cultural, idiomatic, and context-sensitive information [1]. We investigate a broad range of models, including traditional machine learning methods, such as Random Forests and Naive Bayes, and state-of-the-art deep learning architectures, like Transformer and BERT-based models. Experimental results show promising advancements in these areas, with specific models outperforming others depending on the task. The results contribute valuable insights towards the goal of improving the understanding, detection, and translation of puns in AI systems, ultimately promoting a more nuanced and culturally sensitive communication interface in AI technologies."
                },
                {
                    "title": "LogiQA 2.0\u2014An Improved Dataset for Logical Reasoning in Natural Language Understanding",
                    "abstract": "NLP research on logical reasoning regains momentum with the recent releases of a handful of datasets, notably LogiQA and Reclor. Logical reasoning is exploited in many probing tasks over large Pre-trained Language Models (PLMs) and downstream tasks like question-answering and dialogue systems. In this article, we release LogiQA 2.0. The dataset is an amendment and re-annotation of LogiQA in 2020, a large-scale logical reasoning reading comprehension dataset adapted from the Chinese Civil Service Examination. We increase the data size, refine the texts with manual translation by professionals, and improve the quality by removing items with distinctive cultural features like Chinese idioms. Furthermore, we conduct a fine-grained annotation on the dataset and turn it into a two-way natural language inference (NLI) task, resulting in 35 k premise-hypothesis pairs with gold labels, making it the first large-scale NLI dataset for complex logical reasoning. Compared to Question Answering, Natural Language Inference excels in generalizability and helps downstream tasks better. We establish a baseline for logical reasoning in NLI and incite further research."
                },
                {
                    "title": "Diagnosing the First-Order Logical Reasoning Ability Through LogicNLI",
                    "abstract": "Recently, language models (LMs) have achieved significant performance on many NLU tasks, which has spurred widespread interest for their possible applications in the scientific and social area. However, LMs have faced much criticism of whether they are truly capable of reasoning in NLU. In this work, we propose a diagnostic method for first-order logic (FOL) reasoning with a new proposed benchmark, LogicNLI. LogicNLI is an NLI-style dataset that effectively disentangles the target FOL reasoning from commonsense inference and can be used to diagnose LMs from four perspectives: accuracy, robustness, generalization, and interpretability. Experiments on BERT, RoBERTa, and XLNet, have uncovered the weaknesses of these LMs on FOL reasoning, which motivates future exploration to enhance the reasoning ability."
                },
                {
                    "title": "CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledge",
                    "abstract": "When answering a question, people often draw upon their rich world knowledge in addition to the particular context. Recent work has focused primarily on answering questions given some relevant document or context, and required very little general background. To investigate question answering with prior knowledge, we present CommonsenseQA: a challenging new dataset for commonsense question answering. To capture common sense beyond associations, we extract from ConceptNet (Speer et al., 2017) multiple target concepts that have the same semantic relation to a single source concept. Crowd-workers are asked to author multiple-choice questions that mention the source concept and discriminate in turn between each of the target concepts. This encourages workers to create questions with complex semantics that often require prior knowledge. We create 12,247 questions through this procedure and demonstrate the difficulty of our task with a large number of strong baselines. Our best baseline is based on BERT-large (Devlin et al., 2018) and obtains 56% accuracy, well below human performance, which is 89%."
                }
            ],
            "categories": [
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan Large Language Models (LLMs) effectively understand and interpret cunning texts that contain misleading premises, intentional ambiguity, and fallacies?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of LLMs' capabilities in comprehending complex human language. By developing a benchmark like FLUB, researchers can better evaluate LLMs' performance in real-world scenarios where language is often nuanced and misleading. This could lead to significant improvements in LLMs, enhancing their applicability in various fields such as education, customer service, and content moderation, where understanding subtlety and fallacies is essential.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of cunning texts, which often require a deep understanding of context, common sense reasoning, and the ability to identify fallacies. Naive approaches may fail because they typically rely on straightforward pattern recognition or surface-level analysis, which do not account for the subtleties and ambiguities present in such texts. Additionally, the lack of training data that includes these types of challenging examples makes it difficult for LLMs to learn how to handle them effectively.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on evaluating LLMs using \"cleaned\" and \"correct\" corpora, which do not reflect the complexities of real-world language use. Existing benchmarks have not adequately addressed the need for evaluating LLMs' understanding of fallacies and cunning texts. Barriers such as the scarcity of suitable datasets and the lack of a structured approach to assess fallacy understanding have prevented this problem from being solved. Our approach differs by specifically targeting these gaps through the creation of the FLUB benchmark, which includes real-world examples of cunning texts.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the construction of the FaLlacy Understanding Benchmark (FLUB), which includes three tasks: Cunning Type Classification, Fallacy Explanation, and Answer Selection. We will utilize a dataset sourced from the \"Ruozhiba\" forum, which contains real cunning texts, and we will annotate these texts for fallacy types and correct/incorrect answers. The evaluation metric will focus on the accuracy of LLMs in selecting the correct answers and explaining the fallacies. We expect that this benchmark will reveal significant insights into L"
            }
        },
        "author_data": {
            "1fa2f0f9-0443-4e61-afc5-ff5b5c235ddf": {
                "pk": "1fa2f0f9-0443-4e61-afc5-ff5b5c235ddf",
                "name": "Yinghui Li",
                "collaborators": [
                    "Weizhong Guo",
                    "Ze Fu",
                    "Jingheng Ye",
                    "Yangning Li",
                    "Hai-Tao Zheng",
                    "Jinze Wu",
                    "Xin Liu",
                    "Yufei Xue"
                ],
                "domain": [
                    "Robotics",
                    "Motion Planning",
                    "Data Augmentation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Swing Leg Motion Strategy for Heavy-load Legged Robot Based on Force Sensing",
                        "abstract": "The heavy-load legged robot has strong load carrying capacity and can adapt to various unstructured terrains. But the large weight results in higher requirements for motion stability and environmental perception ability. In order to utilize force sensing information to improve its motion performance, in this paper, we propose a finite state machine model for the swing leg in the static gait by imitating the movement of the elephant. Based on the presence or absence of additional terrain information, different trajectory planning strategies are provided for the swing leg to enhance the success rate of stepping and save energy. The experimental results on a novel quadruped robot show that our method has strong robustness and can enable heavy-load legged robots to pass through various complex terrains autonomously and smoothly."
                    },
                    {
                        "title": "MixEdit: Revisiting Data Augmentation and Beyond for Grammatical Error Correction",
                        "abstract": "Data Augmentation through generating pseudo data has been proven effective in mitigating the challenge of data scarcity in the field of Grammatical Error Correction (GEC). Various augmentation strategies have been widely explored, most of which are motivated by two heuristics, i.e., increasing the distribution similarity and diversity of pseudo data. However, the underlying mechanism responsible for the effectiveness of these strategies remains poorly understood. In this paper, we aim to clarify how data augmentation improves GEC models. To this end, we introduce two interpretable and computationally efficient measures: Affinity and Diversity. Our findings indicate that an excellent GEC data augmentation strategy characterized by high Affinity and appropriate Diversity can better improve the performance of GEC models. Based on this observation, we propose MixEdit, a data augmentation approach that strategically and dynamically augments realistic data, without requiring extra monolingual corpora. To verify the correctness of our findings and the effectiveness of the proposed MixEdit, we conduct experiments on mainstream English and Chinese GEC datasets. The results show that MixEdit substantially improves GEC models and is complementary to traditional data augmentation methods."
                    },
                    {
                        "title": "Experience-Learning Inspired Two-Step Reward Method for Efficient Legged Locomotion Learning Towards Natural and Robust Gaits",
                        "abstract": "Multi-legged robots offer enhanced stability in complex terrains, yet autonomously learning natural and robust motions in such environments remains challenging. Drawing inspiration from animals' progressive learning patterns, from simple to complex tasks, we introduce a universal two-stage learning framework with two-step reward setting based on self-acquired experience, which efficiently enables legged robots to incrementally learn natural and robust movements. In the first stage, robots learn through gait-related rewards to track velocity on flat terrain, acquiring natural, robust movements and generating effective motion experience data. In the second stage, mirroring animal learning from existing experiences, robots learn to navigate challenging terrains with natural and robust movements using adversarial imitation learning. To demonstrate our method's efficacy, we trained both quadruped robots and a hexapod robot, and the policy were successfully transferred to a physical quadruped robot GO1, which exhibited natural gait patterns and remarkable robustness in various terrains."
                    }
                ]
            },
            "97285d4f-e3f6-4674-a78e-39d0037342eb": {
                "pk": "97285d4f-e3f6-4674-a78e-39d0037342eb",
                "name": "Qingyu Zhou",
                "collaborators": [
                    "Furu Wei",
                    "Ming Zhou",
                    "Yunbo Cao",
                    "Nan Yang",
                    "Zhongli Li",
                    "Chao Li",
                    "Yuanzhen Luo",
                    "Feng Zhou",
                    "Ke Xu",
                    "Chuanqi Tan"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Text Summarization",
                    "Few-Shot Learning",
                    "Grammatical Error Correction"
                ],
                "publications": [
                    {
                        "title": "Selective Encoding for Abstractive Sentence Summarization",
                        "abstract": "We propose a selective encoding model to extend the sequence-to-sequence framework for abstractive sentence summarization. It consists of a sentence encoder, a selective gate network, and an attention equipped decoder. The sentence encoder and decoder are built with recurrent neural networks. The selective gate network constructs a second level sentence representation by controlling the information flow from encoder to decoder. The second level representation is tailored for sentence summarization task, which leads to better performance. We evaluate our model on the English Gigaword, DUC 2004 and MSR abstractive sentence summarization datasets. The experimental results show that the proposed selective encoding model outperforms the state-of-the-art baseline models."
                    },
                    {
                        "title": "Learning to Summarize Passages: Mining Passage-Summary Pairs from Wikipedia Revision Histories",
                        "abstract": "In this paper, we propose a method for automatically constructing a passage-to-summary dataset by mining the Wikipedia page revision histories. In particular, the method mines the main body passages and the introduction sentences which are added to the pages simultaneously. The constructed dataset contains more than one hundred thousand passage-summary pairs. The quality analysis shows that it is promising that the dataset can be used as a training and validation set for passage summarization. We validate and analyze the performance of various summarization systems on the proposed dataset. The dataset will be available online at https://res.qyzhou.me."
                    },
                    {
                        "title": "Sequential Copying Networks",
                        "abstract": "Copying mechanism shows effectiveness in sequence-to-sequence based neural network models for text generation tasks, such as abstractive sentence summarization and question generation. However, existing works on modeling copying or pointing mechanism only considers single word copying from the source sentences. In this paper, we propose a novel copying framework, named Sequential Copying Networks (SeqCopyNet), which not only learns to copy single words, but also copies sequences from the input sentence. It leverages the pointer networks to explicitly select a sub-span from the source side to target side, and integrates this sequential copying mechanism to the generation process in the encoder-decoder paradigm. Experiments on abstractive sentence summarization and question generation tasks show that the proposed SeqCopyNet can copy meaningful spans and outperforms the baseline models."
                    },
                    {
                        "title": "At Which Level Should We Extract? An Empirical Analysis on Extractive Document Summarization",
                        "abstract": "Extractive methods have been proven effective in automatic document summarization. Previous works perform this task by identifying informative contents at sentence level. However, it is unclear whether performing extraction at sentence level is the best solution. In this work, we show that unnecessity and redundancy issues exist when extracting full sentences, and extracting sub-sentential units is a promising alternative. Specifically, we propose extracting sub-sentential units based on the constituency parsing tree. A neural extractive model which leverages the sub-sentential information and extracts them is presented. Extensive experiments and analyses show that extracting sub-sentential units performs competitively comparing to full sentence extraction under the evaluation of both automatic and human evaluations. Hopefully, our work could provide some inspiration of the basic extraction units in extractive summarization for future research."
                    },
                    {
                        "title": "Enhancing Phrase Representation by Information Bottleneck Guided Text Diffusion Process for Keyphrase Extraction",
                        "abstract": "Keyphrase extraction (KPE) is an important task in Natural Language Processing for many scenarios, which aims to extract keyphrases that are present in a given document. Many existing supervised methods treat KPE as sequential labeling, span-level classification, or generative tasks. However, these methods lack the ability to utilize keyphrase information, which may result in biased results. In this study, we propose Diff-KPE, which leverages the supervised Variational Information Bottleneck (VIB) to guide the text diffusion process for generating enhanced keyphrase representations. Diff-KPE first generates the desired keyphrase embeddings conditioned on the entire document and then injects the generated keyphrase embeddings into each phrase representation. A ranking network and VIB are then optimized together with rank loss and classification loss, respectively. This design of Diff-KPE allows us to rank each candidate phrase by utilizing both the information of keyphrases and the document. Experiments show that Diff-KPE outperforms existing KPE methods on a large open domain keyphrase extraction benchmark, OpenKP, and a scientific domain dataset, KP20K."
                    },
                    {
                        "title": "Improving BERT with Syntax-aware Local Attention",
                        "abstract": "Pre-trained Transformer-based neural language models, such as BERT, have achieved remarkable results on varieties of NLP tasks. Recent works have shown that attention-based models can benefit from more focused attention over local regions. Most of them restrict the attention scope within a linear span, or confine to certain tasks such as machine translation and question answering. In this paper, we propose a syntax-aware local attention, where the attention scopes are restrained based on the distances in the syntactic structure. The proposed syntax-aware local attention can be integrated with pretrained language models, such as BERT, to render the model to focus on syntactically relevant words. We conduct experiments on various single-sentence benchmarks, including sentence classification and sequence labeling tasks. Experimental results show consistent gains over BERT on all benchmark datasets. The extensive studies verify that our model achieves better performance owing to more focused attention over syntactically relevant words."
                    },
                    {
                        "title": "Neural Question Generation from Text: A Preliminary Study",
                        "abstract": "Automatic question generation aims to generate questions from a text passage where the generated questions can be answered by certain sub-spans of the given passage. Traditional methods mainly use rigid heuristic rules to transform a sentence into related questions. In this work, we propose to apply the neural encoder-decoder model to generate meaningful and diverse questions from natural language sentences. The encoder reads the input text and the answer position, to produce an answer-aware input representation, which is fed to the decoder to generate an answer focused question. We conduct a preliminary study on neural question generation from text with the SQuAD dataset, and the experiment results show that our method can produce fluent and diverse questions."
                    },
                    {
                        "title": "Neural Document Summarization by Jointly Learning to Score and Select Sentences",
                        "abstract": "Sentence scoring and sentence selection are two main steps in extractive document summarization systems. However, previous works treat them as two separated subtasks. In this paper, we present a novel end-to-end neural network framework for extractive document summarization by jointly learning to score and select sentences. It first reads the document sentences with a hierarchical encoder to obtain the representation of sentences. Then it builds the output summary by extracting sentences one by one. Different from previous methods, our approach integrates the selection strategy into the scoring model, which directly predicts the relative importance given previously selected sentences. Experiments on the CNN/Daily Mail dataset show that the proposed framework significantly outperforms the state-of-the-art extractive summarization models."
                    },
                    {
                        "title": "An Enhanced Span-based Decomposition Method for Few-Shot Sequence Labeling",
                        "abstract": "Few-Shot Sequence Labeling (FSSL) is a canonical paradigm for the tagging models, e.g., named entity recognition and slot filling, to generalize on an emerging, resource-scarce domain. Recently, the metric-based meta-learning framework has been recognized as a promising approach for FSSL. However, most prior works assign a label to each token based on the token-level similarities, which ignores the integrality of named entities or slots. To this end, in this paper, we propose ESD, an Enhanced Span-based Decomposition method for FSSL. ESD formulates FSSL as a span-level matching problem between test query and supporting instances. Specifically, ESD decomposes the span matching problem into a series of span-level procedures, mainly including enhanced span representation, class prototype aggregation and span conflicts resolution. Extensive experiments show that ESD achieves the new state-of-the-art results on two popular FSSL benchmarks, FewNERD and SNIPS, and is proven to be more robust in the nested and noisy tagging scenarios. Our code is available at https://github.com/Wangpeiyi9979/ESD."
                    },
                    {
                        "title": "Type-Driven Multi-Turn Corrections for Grammatical Error Correction",
                        "abstract": "Grammatical Error Correction (GEC) aims to automatically detect and correct grammatical errors. In this aspect, dominant models are trained by one-iteration learning while performing multiple iterations of corrections during inference. Previous studies mainly focus on the data augmentation approach to combat the exposure bias, which suffers from two drawbacks. First, they simply mix additionally-constructed training instances and original ones to train models, which fails to help models be explicitly aware of the procedure of gradual corrections. Second, they ignore the interdependence between different types of corrections. In this paper, we propose a Type-Driven Multi-Turn Corrections approach for GEC. Using this approach, from each training instance, we additionally construct multiple training instances, each of which involves the correction of a specific type of errors. Then, we use these additionally-constructed training instances and the original one to train the model in turn. Experimental results and in-depth analysis show that our approach significantly benefits the model training. Particularly, our enhanced model achieves state-of-the-art single-model performance on English GEC benchmarks. We release our code at Github."
                    },
                    {
                        "title": "Seeking Patterns, Not just Memorizing Procedures: Contrastive Learning for Solving Math Word Problems",
                        "abstract": "Math Word Problem (MWP) solving needs to discover the quantitative relationships over natural language narratives. Recent work shows that existing models memorize procedures from context and rely on shallow heuristics to solve MWPs. In this paper, we look at this issue and argue that the cause is a lack of overall understanding of MWP patterns. We first investigate how a neural network understands patterns only from semantics, and observe that, if the prototype equations are the same, most problems get closer representations and those representations apart from them or close to other prototypes tend to produce wrong solutions. Inspired by it, we propose a contrastive learning approach, where the neural network perceives the divergence of patterns. We collect contrastive examples by converting the prototype equation into a tree and seeking similar tree structures. The solving model is trained with an auxiliary objective on the collected examples, resulting in the representations of problems with similar prototypes being pulled closer. We conduct experiments on the Chinese dataset Math23k and the English dataset MathQA. Our method greatly improves the performance in monolingual and multilingual settings."
                    },
                    {
                        "title": "A non-hierarchical attention network with modality dropout for textual response generation in multimodal dialogue systems",
                        "abstract": "Existing text- and image-based multimodal dialogue systems use the traditional Hierarchical Recurrent Encoder-Decoder (HRED) framework, which has an utterance-level encoder to model utterance representation and a context-level encoder to model context representation. Although pioneer efforts have shown promising performances, they still suffer from the following challenges: (1) the interaction between textual features and visual features is not fine-grained enough. (2) the context representation can not provide a complete representation for the context. To address the issues mentioned above, we propose a non-hierarchical attention network with modality dropout, which abandons the HRED framework and utilizes attention modules to encode each utterance and model the context representation. To evaluate our proposed model, we conduct comprehensive experiments on a public multimodal dialogue dataset. Automatic and human evaluation demonstrate that our proposed model outperforms the existing methods and achieves state-of-the-art performance."
                    },
                    {
                        "title": "Unifying Token and Span Level Supervisions for Few-Shot Sequence Labeling",
                        "abstract": "Few-shot sequence labeling aims to identify novel classes based on only a few labeled samples. Existing methods solve the data scarcity problem mainly by designing token-level or span-level labeling models based on metric learning. However, these methods are only trained at a single granularity (i.e., either token level or span level) and have some weaknesses of the corresponding granularity. In this paper, we first unify token and span level supervisions and propose a Consistent Dual Adaptive Prototypical (CDAP) network for few-shot sequence labeling. CDAP contains the token-level and span-level networks, jointly trained at different granularities. To align the outputs of two networks, we further propose a consistent loss to enable them to learn from each other. During the inference phase, we propose a consistent greedy inference algorithm that first adjusts the predicted probability and then greedily selects non-overlapping spans with maximum probability. Extensive experiments show that our model achieves new state-of-the-art results on three benchmark datasets."
                    }
                ]
            },
            "a7ec6776-ff40-4da1-9ce2-947cfacfdcdf": {
                "pk": "a7ec6776-ff40-4da1-9ce2-947cfacfdcdf",
                "name": "Yuanzhen Luo",
                "collaborators": [
                    "Qingyu Zhou",
                    "Feng Zhou",
                    "Qiang Lu",
                    "Xilei Hu",
                    "Jake Luo",
                    "Zhiguang Wang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Keyphrase Extraction",
                    "Neural Network Interpretation",
                    "Symbolic Regression"
                ],
                "publications": [
                    {
                        "title": "Enhancing Phrase Representation by Information Bottleneck Guided Text Diffusion Process for Keyphrase Extraction",
                        "abstract": "Keyphrase extraction (KPE) is an important task in Natural Language Processing for many scenarios, which aims to extract keyphrases that are present in a given document. Many existing supervised methods treat KPE as sequential labeling, span-level classification, or generative tasks. However, these methods lack the ability to utilize keyphrase information, which may result in biased results. In this study, we propose Diff-KPE, which leverages the supervised Variational Information Bottleneck (VIB) to guide the text diffusion process for generating enhanced keyphrase representations. Diff-KPE first generates the desired keyphrase embeddings conditioned on the entire document and then injects the generated keyphrase embeddings into each phrase representation. A ranking network and VIB are then optimized together with rank loss and classification loss, respectively. This design of Diff-KPE allows us to rank each candidate phrase by utilizing both the information of keyphrases and the document. Experiments show that Diff-KPE outperforms existing KPE methods on a large open domain keyphrase extraction benchmark, OpenKP, and a scientific domain dataset, KP20K."
                    },
                    {
                        "title": "Exploring Hidden Semantics in Neural Networks with Symbolic Regression",
                        "abstract": "Many recent studies focus on developing mechanisms to explain the black-box behaviors of neural networks (NNs). However, little work has been done to extract the potential hidden semantics (mathematical representation) of a neural network. A succinct and explicit mathematical representation of a NN model could improve the understanding and interpretation of its behaviors. To address this need, we propose a novel symbolic regression method for neural works (called SRNet) to discover the mathematical expressions of a NN. SRNet creates a Cartesian genetic programming (NNCGP) to represent the hidden semantics of a single layer in a NN. It then leverages a multi-chromosome NNCGP to represent hidden semantics of all layers of the NN. The method uses a (1+$\\lambda$) evolutionary strategy (called MNNCGP-ES) to extract the final mathematical expressions of all layers in the NN. Experiments on 12 symbolic regression benchmarks and 5 classification benchmarks show that SRNet not only can reveal the complex relationships between each layer of a NN but also can extract the mathematical representation of the whole NN. Compared with LIME and MAPLE, SRNet has higher interpolation accuracy and trends to approximate the real model on the practical dataset."
                    }
                ]
            },
            "849319ef-b87d-4c77-be5b-6989fd0d61ec": {
                "pk": "849319ef-b87d-4c77-be5b-6989fd0d61ec",
                "name": "Shirong Ma",
                "collaborators": [
                    "Yinghui Li",
                    "Hai-Tao Zheng",
                    "Yangning Li",
                    "Shulin Huang",
                    "Ying Shen",
                    "Qingyu Zhou",
                    "Haojing Huang",
                    "Yong Jiang",
                    "Yunbo Cao",
                    "Jingheng Ye"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Grammatical Error Correction",
                    "Text Generation",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "From Retrieval to Generation: Efficient and Effective Entity Set Expansion",
                        "abstract": "Entity Set Expansion (ESE) is a critical task aiming at expanding entities of the target semantic class described by seed entities. Most existing ESE methods are retrieval-based frameworks that need to extract contextual features of entities and calculate the similarity between seed entities and candidate entities. To achieve the two purposes, they iteratively traverse the corpus and the entity vocabulary, resulting in poor efficiency and scalability. Experimental results indicate that the time consumed by the retrieval-based ESE methods increases linearly with entity vocabulary and corpus size. In this paper, we firstly propose Generative Entity Set Expansion (GenExpan) framework, which utilizes a generative pre-trained auto-regressive language model to accomplish ESE task. Specifically, a prefix tree is employed to guarantee the validity of entity generation, and automatically generated class names are adopted to guide the model to generate target entities. Moreover, we propose Knowledge Calibration and Generative Ranking to further bridge the gap between generic knowledge of the language model and the goal of ESE task. For efficiency, expansion time consumed by GenExpan is independent of entity vocabulary and corpus size, and GenExpan achieves an average 600% speedup compared to strong baselines. For expansion effectiveness, our framework outperforms previous state-of-the-art ESE methods."
                    },
                    {
                        "title": "Focus Is What You Need For Chinese Grammatical Error Correction",
                        "abstract": "Chinese Grammatical Error Correction (CGEC) aims to automatically detect and correct grammatical errors contained in Chinese text. In the long term, researchers regard CGEC as a task with a certain degree of uncertainty, that is, an ungrammatical sentence may often have multiple references. However, we argue that even though this is a very reasonable hypothesis, it is too harsh for the intelligence of the mainstream models in this era. In this paper, we first discover that multiple references do not actually bring positive gains to model training. On the contrary, it is beneficial to the CGEC model if the model can pay attention to small but essential data during the training process. Furthermore, we propose a simple yet effective training strategy called OneTarget to improve the focus ability of the CGEC models and thus improve the CGEC performance. Extensive experiments and detailed analyses demonstrate the correctness of our discovery and the effectiveness of our proposed method."
                    },
                    {
                        "title": "Correct Like Humans: Progressive Learning Framework for Chinese Text Error Correction",
                        "abstract": "Chinese Text Error Correction (CTEC) aims to detect and correct errors in the input text, which benefits human daily life and various downstream tasks. Recent approaches mainly employ Pre-trained Language Models (PLMs) to resolve CTEC. Although PLMs have achieved remarkable success in CTEC, we argue that previous studies still overlook the importance of human thinking patterns. To enhance the development of PLMs for CTEC, inspired by humans' daily error-correcting behavior, we propose a novel model-agnostic progressive learning framework, named ProTEC, which guides PLMs-based CTEC models to learn to correct like humans. During the training process, ProTEC guides the model to learn text error correction by incorporating these sub-tasks into a progressive paradigm. During the inference process, the model completes these sub-tasks in turn to generate the correction results. Extensive experiments and detailed analyses demonstrate the effectiveness and efficiency of our proposed model-agnostic ProTEC framework."
                    },
                    {
                        "title": "EcomGPT-CT: Continual Pre-training of E-commerce Large Language Models with Semi-structured Data",
                        "abstract": "Large Language Models (LLMs) pre-trained on massive corpora have exhibited remarkable performance on various NLP tasks. However, applying these models to specific domains still poses significant challenges, such as lack of domain knowledge, limited capacity to leverage domain knowledge and inadequate adaptation to domain-specific data formats. Considering the exorbitant cost of training LLMs from scratch and the scarcity of annotated data within particular domains, in this work, we focus on domain-specific continual pre-training of LLMs using E-commerce domain as an exemplar. Specifically, we explore the impact of continual pre-training on LLMs employing unlabeled general and E-commercial corpora. Furthermore, we design a mixing strategy among different data sources to better leverage E-commercial semi-structured data. We construct multiple tasks to assess LLMs' few-shot In-context Learning ability and their zero-shot performance after instruction tuning in E-commerce domain. Experimental results demonstrate the effectiveness of continual pre-training of E-commerce LLMs and the efficacy of our devised data mixing strategy."
                    },
                    {
                        "title": "Contextual Similarity is More Valuable than Character Similarity: An Empirical Study for Chinese Spell Checking",
                        "abstract": "Chinese Spell Checking (CSC) task aims to detect and correct Chinese spelling errors. Recently, related researches focus on introducing character similarity from confusion set to enhance the CSC models, ignoring the context of characters that contain richer information. To make better use of contextual information, we propose a simple yet effective Curriculum Learning (CL) framework for the CSC task. With the help of our model-agnostic CL framework, existing CSC models will be trained from easy to difficult as humans learn Chinese characters and achieve further performance improvements. Extensive experiments and detailed analyses on widely used SIGHAN datasets show that our method outperforms previous state-of-the-art methods. More instructively, our study empirically suggests that contextual similarity is more valuable than character similarity for the CSC task."
                    },
                    {
                        "title": "Towards Attribute-Entangled Controllable Text Generation: A Pilot Study of Blessing Generation",
                        "abstract": "Controllable Text Generation (CTG) has obtained great success due to its fine-grained generation ability obtained by focusing on multiple attributes. However, most existing CTG researches overlook how to utilize the attribute entanglement to enhance the diversity of the controlled generated texts. Facing this dilemma, we focus on a novel CTG scenario, i.e., blessing generation which is challenging because high-quality blessing texts require CTG models to comprehensively consider the entanglement between multiple attributes (e.g., objects and occasions). To promote the research on blessing generation, we present EBleT, a large-scale Entangled Blessing Text dataset containing 293K English sentences annotated with multiple attributes. Furthermore, we propose novel evaluation metrics to measure the quality of the blessing texts generated by the baseline models we designed. Our study opens a new research direction for controllable text generation and enables the development of attribute-entangled CTG models. Our dataset and source codes are available at \\url{https://github.com/huangshulin123/Blessing-Generation}."
                    },
                    {
                        "title": "LatEval: An Interactive LLMs Evaluation Benchmark with Incomplete Information from Lateral Thinking Puzzles",
                        "abstract": "With the continuous evolution and refinement of LLMs, they are endowed with impressive logical reasoning or vertical thinking capabilities. But can they think out of the box? Do they possess proficient lateral thinking abilities? Following the setup of Lateral Thinking Puzzles, we propose a novel evaluation benchmark, LatEval, which assesses the model's lateral thinking within an interactive framework. In our benchmark, we challenge LLMs with 2 aspects: the quality of questions posed by the model and the model's capability to integrate information for problem-solving. We find that nearly all LLMs struggle with employing lateral thinking during interactions. For example, even the most advanced model, GPT-4, exhibits the advantage to some extent, yet still maintain a noticeable gap when compared to human. This evaluation benchmark provides LLMs with a highly challenging and distinctive task that is crucial to an effective AI assistant."
                    },
                    {
                        "title": "Mitigating Catastrophic Forgetting in Multi-domain Chinese Spelling Correction by Multi-stage Knowledge Transfer Framework",
                        "abstract": "Chinese Spelling Correction (CSC) aims to detect and correct spelling errors in given sentences. Recently, multi-domain CSC has gradually attracted the attention of researchers because it is more practicable. In this paper, we focus on the key flaw of the CSC model when adapting to multi-domain scenarios: the tendency to forget previously acquired knowledge upon learning new domain-specific knowledge (i.e., catastrophic forgetting). To address this, we propose a novel model-agnostic Multi-stage Knowledge Transfer (MKT) framework, which utilizes a continuously evolving teacher model for knowledge transfer in each domain, rather than focusing solely on new domain knowledge. It deserves to be mentioned that we are the first to apply continual learning methods to the multi-domain CSC task. Experiments prove the effectiveness of our proposed method, and further analyses demonstrate the importance of overcoming catastrophic forgetting for improving the model performance."
                    },
                    {
                        "title": "Equality before the Law: Legal Judgment Consistency Analysis for Fairness",
                        "abstract": "In a legal system, judgment consistency is regarded as one of the most important manifestations of fairness. However, due to the complexity of factual elements that impact sentencing in real-world scenarios, few works have been done on quantitatively measuring judgment consistency towards real-world data. In this paper, we propose an evaluation metric for judgment inconsistency, Legal Inconsistency Coefficient (LInCo), which aims to evaluate inconsistency between data groups divided by specific features (e.g., gender, region, race). We propose to simulate judges from different groups with legal judgment prediction (LJP) models and measure the judicial inconsistency with the disagreement of the judgment results given by LJP models trained on different groups. Experimental results on the synthetic data verify the effectiveness of LInCo. We further employ LInCo to explore the inconsistency in real cases and come to the following observations: (1) Both regional and gender inconsistency exist in the legal system, but gender inconsistency is much less than regional inconsistency; (2) The level of regional inconsistency varies little across different time periods; (3) In general, judicial inconsistency is negatively correlated with the severity of the criminal charges. Besides, we use LInCo to evaluate the performance of several de-bias methods, such as adversarial learning, and find that these mechanisms can effectively help LJP models to avoid suffering from data bias."
                    },
                    {
                        "title": "CLEME: Debiasing Multi-reference Evaluation for Grammatical Error Correction",
                        "abstract": "Evaluating the performance of Grammatical Error Correction (GEC) systems is a challenging task due to its subjectivity. Designing an evaluation metric that is as objective as possible is crucial to the development of GEC task. However, mainstream evaluation metrics, i.e., reference-based metrics, introduce bias into the multi-reference evaluation by extracting edits without considering the presence of multiple references. To overcome this issue, we propose Chunk-LEvel Multi-reference Evaluation (CLEME), designed to evaluate GEC systems in the multi-reference evaluation setting. CLEME builds chunk sequences with consistent boundaries for the source, the hypothesis and references, thus eliminating the bias caused by inconsistent edit boundaries. Furthermore, we observe the consistent boundary could also act as the boundary of grammatical errors, based on which the F$_{0.5}$ score is then computed following the correction independence assumption. We conduct experiments on six English reference sets based on the CoNLL-2014 shared task. Extensive experiments and detailed analyses demonstrate the correctness of our discovery and the effectiveness of CLEME. Further analysis reveals that CLEME is robust to evaluate GEC systems across reference sets with varying numbers of references and annotation style."
                    },
                    {
                        "title": "On the (In)Effectiveness of Large Language Models for Chinese Text Correction",
                        "abstract": "Recently, the development and progress of Large Language Models (LLMs) have amazed the entire Artificial Intelligence community. Benefiting from their emergent abilities, LLMs have attracted more and more researchers to study their capabilities and performance on various downstream Natural Language Processing (NLP) tasks. While marveling at LLMs' incredible performance on all kinds of tasks, we notice that they also have excellent multilingual processing capabilities, such as Chinese. To explore the Chinese processing ability of LLMs, we focus on Chinese Text Correction, a fundamental and challenging Chinese NLP task. Specifically, we evaluate various representative LLMs on the Chinese Grammatical Error Correction (CGEC) and Chinese Spelling Check (CSC) tasks, which are two main Chinese Text Correction scenarios. Additionally, we also fine-tune LLMs for Chinese Text Correction to better observe the potential capabilities of LLMs. From extensive analyses and comparisons with previous state-of-the-art small models, we empirically find that the LLMs currently have both amazing performance and unsatisfactory behavior for Chinese Text Correction. We believe our findings will promote the landing and application of LLMs in the Chinese NLP community."
                    },
                    {
                        "title": "EcomGPT: Instruction-tuning Large Language Models with Chain-of-Task Tasks for E-commerce",
                        "abstract": "Recently, instruction-following Large Language Models (LLMs) , represented by ChatGPT, have exhibited exceptional performance in general Natural Language Processing (NLP) tasks. However, the unique characteristics of E-commerce data pose significant challenges to general LLMs. An LLM tailored specifically for E-commerce scenarios, possessing robust cross-dataset/task generalization capabilities, is a pressing necessity. To solve this issue, in this work, we proposed the first e-commerce instruction dataset EcomInstruct, with a total of 2.5 million instruction data. EcomInstruct scales up the data size and task diversity by constructing atomic tasks with E-commerce basic data types, such as product information, user reviews. Atomic tasks are defined as intermediate tasks implicitly involved in solving a final task, which we also call Chain-of-Task tasks. We developed EcomGPT with different parameter scales by training the backbone model BLOOMZ with the EcomInstruct. Benefiting from the fundamental semantic understanding capabilities acquired from the Chain-of-Task tasks, EcomGPT exhibits excellent zero-shot generalization capabilities. Extensive experiments and human evaluations demonstrate that EcomGPT outperforms ChatGPT in term of cross-dataset/task generalization on E-commerce tasks."
                    },
                    {
                        "title": "Learning from the Dictionary: Heterogeneous Knowledge Guided Fine-tuning for Chinese Spell Checking",
                        "abstract": "Chinese Spell Checking (CSC) aims to detect and correct Chinese spelling errors. Recent researches start from the pretrained knowledge of language models and take multimodal information into CSC models to improve the performance. However, they overlook the rich knowledge in the dictionary, the reference book where one can learn how one character should be pronounced, written, and used. In this paper, we propose the LEAD framework, which renders the CSC model to learn heterogeneous knowledge from the dictionary in terms of phonetics, vision, and meaning. LEAD first constructs positive and negative samples according to the knowledge of character phonetics, glyphs, and definitions in the dictionary. Then a unified contrastive learning-based training scheme is employed to refine the representations of the CSC models. Extensive experiments and detailed analyses on the SIGHAN benchmark datasets demonstrate the effectiveness of our proposed methods."
                    },
                    {
                        "title": "Linguistic Rules-Based Corpus Generation for Native Chinese Grammatical Error Correction",
                        "abstract": "Chinese Grammatical Error Correction (CGEC) is both a challenging NLP task and a common application in human daily life. Recently, many data-driven approaches are proposed for the development of CGEC research. However, there are two major limitations in the CGEC field: First, the lack of high-quality annotated training corpora prevents the performance of existing CGEC models from being significantly improved. Second, the grammatical errors in widely used test sets are not made by native Chinese speakers, resulting in a significant gap between the CGEC models and the real application. In this paper, we propose a linguistic rules-based approach to construct large-scale CGEC training corpora with automatically generated grammatical errors. Additionally, we present a challenging CGEC benchmark derived entirely from errors made by native Chinese speakers in real-world scenarios. Extensive experiments and detailed analyses not only demonstrate that the training data constructed by our method effectively improves the performance of CGEC models, but also reflect that our benchmark is an excellent resource for further development of the CGEC field."
                    }
                ]
            },
            "239a306e-756b-4328-9c61-385898eba781": {
                "pk": "239a306e-756b-4328-9c61-385898eba781",
                "name": "Yangning Li",
                "collaborators": [
                    "Yinghui Li",
                    "Hai-Tao Zheng",
                    "Ying Shen",
                    "Shirong Ma",
                    "Qingyu Zhou",
                    "Xi Chen",
                    "Shulin Huang",
                    "Yong Jiang",
                    "Jingheng Ye",
                    "Jiaoyan Chen"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Knowledge Graph",
                    "Machine Learning",
                    "Data Augmentation"
                ],
                "publications": [
                    {
                        "title": "MixEdit: Revisiting Data Augmentation and Beyond for Grammatical Error Correction",
                        "abstract": "Data Augmentation through generating pseudo data has been proven effective in mitigating the challenge of data scarcity in the field of Grammatical Error Correction (GEC). Various augmentation strategies have been widely explored, most of which are motivated by two heuristics, i.e., increasing the distribution similarity and diversity of pseudo data. However, the underlying mechanism responsible for the effectiveness of these strategies remains poorly understood. In this paper, we aim to clarify how data augmentation improves GEC models. To this end, we introduce two interpretable and computationally efficient measures: Affinity and Diversity. Our findings indicate that an excellent GEC data augmentation strategy characterized by high Affinity and appropriate Diversity can better improve the performance of GEC models. Based on this observation, we propose MixEdit, a data augmentation approach that strategically and dynamically augments realistic data, without requiring extra monolingual corpora. To verify the correctness of our findings and the effectiveness of the proposed MixEdit, we conduct experiments on mainstream English and Chinese GEC datasets. The results show that MixEdit substantially improves GEC models and is complementary to traditional data augmentation methods."
                    },
                    {
                        "title": "Tree-structured Auxiliary Online Knowledge Distillation",
                        "abstract": "Traditional knowledge distillation adopts a two-stage training process in which a teacher model is pre-trained and then transfers the knowledge to a compact student model. To overcome the limitation, online knowledge distillation is proposed to perform one-stage distillation when the teacher is unavailable. Recent researches on online knowledge distillation mainly focus on the design of the distillation objective, including attention or gate mechanism. Instead, in this work, we focus on the design of the global architecture and propose Tree-Structured Auxiliary online knowledge distillation (TSA), which adds more parallel peers for layers close to the output hierarchically to strengthen the effect of knowledge distillation. Different branches construct different views of the inputs, which can be the source of the knowledge. The hierarchical structure implies that the knowledge transfers from general to task-specific with the growth of the layers. Extensive experiments on 3 computer vision and 4 natural language processing datasets show that our method achieves state-of-the-art performance without bells and whistles. To the best of our knowledge, we are the first to demonstrate the effectiveness of online knowledge distillation for machine translation tasks."
                    },
                    {
                        "title": "Embracing Ambiguity: Improving Similarity-oriented Tasks with Contextual Synonym Knowledge",
                        "abstract": "Contextual synonym knowledge is crucial for those similarity-oriented tasks whose core challenge lies in capturing semantic similarity between entities in their contexts, such as entity linking and entity matching. However, most Pre-trained Language Models (PLMs) lack synonym knowledge due to inherent limitations of their pre-training objectives such as masked language modeling (MLM). Existing works which inject synonym knowledge into PLMs often suffer from two severe problems: (i) Neglecting the ambiguity of synonyms, and (ii) Undermining semantic understanding of original PLMs, which is caused by inconsistency between the exact semantic similarity of the synonyms and the broad conceptual relevance learned from the original corpus. To address these issues, we propose PICSO, a flexible framework that supports the injection of contextual synonym knowledge from multiple domains into PLMs via a novel entity-aware Adapter which focuses on the semantics of the entities (synonyms) in the contexts. Meanwhile, PICSO stores the synonym knowledge in additional parameters of the Adapter structure, which prevents it from corrupting the semantic understanding of the original PLM. Extensive experiments demonstrate that PICSO can dramatically outperform the original PLMs and the other knowledge and synonym injection models on four different similarity-oriented tasks. In addition, experiments on GLUE prove that PICSO also benefits general natural language understanding tasks. Codes and data will be public."
                    },
                    {
                        "title": "Contrastive Learning with Hard Negative Entities for Entity Set Expansion",
                        "abstract": "Entity Set Expansion (ESE) is a promising task which aims to expand entities of the target semantic class described by a small seed entity set. Various NLP and IR applications will benefit from ESE due to its ability to discover knowledge. Although previous ESE methods have achieved great progress, most of them still lack the ability to handle hard negative entities (i.e., entities that are difficult to distinguish from the target entities), since two entities may or may not belong to the same semantic class based on different granularity levels we analyze on. To address this challenge, we devise an entity-level masked language model with contrastive learning to refine the representation of entities. In addition, we propose the ProbExpan, a novel probabilistic ESE framework utilizing the entity representation obtained by the aforementioned language model to expand entities. Extensive experiments and detailed analyses on three datasets show that our method outperforms previous state-of-the-art methods."
                    },
                    {
                        "title": "Vision, Deduction and Alignment: An Empirical Study on Multi-modal Knowledge Graph Alignment",
                        "abstract": "Entity alignment (EA) for knowledge graphs (KGs) plays a critical role in knowledge engineering. Existing EA methods mostly focus on utilizing the graph structures and entity attributes (including literals), but ignore images that are common in modern multi-modal KGs. In this study we first constructed Multi-OpenEA -- eight large-scale, image-equipped EA benchmarks, and then evaluated some existing embedding-based methods for utilizing images. In view of the complementary nature of visual modal information and logical deduction, we further developed a new multi-modal EA method named LODEME using logical deduction and multi-modal KG embedding, with state-of-the-art performance achieved on Multi-OpenEA and other existing multi-modal EA benchmarks."
                    },
                    {
                        "title": "Learning Purified Feature Representations from Task-irrelevant Labels",
                        "abstract": "Learning an empirically effective model with generalization using limited data is a challenging task for deep neural networks. In this paper, we propose a novel learning framework called PurifiedLearning to exploit task-irrelevant features extracted from task-irrelevant labels when training models on small-scale datasets. Particularly, we purify feature representations by using the expression of task-irrelevant information, thus facilitating the learning process of classification. Our work is built on solid theoretical analysis and extensive experiments, which demonstrate the effectiveness of PurifiedLearning. According to the theory we proved, PurifiedLearning is model-agnostic and doesn't have any restrictions on the model needed, so it can be combined with any existing deep neural networks with ease to achieve better performance. The source code of this paper will be available in the future for reproducibility."
                    },
                    {
                        "title": "A Curriculum Learning Approach for Multi-domain Text Classification Using Keyword weight Ranking",
                        "abstract": "Text classification is a very classic NLP task, but it has two prominent shortcomings: On the one hand, text classification is deeply domain-dependent. That is, a classifier trained on the corpus of one domain may not perform so well in another domain. On the other hand, text classification models require a lot of annotated data for training. However, for some domains, there may not exist enough annotated data. Therefore, it is valuable to investigate how to efficiently utilize text data from different domains to improve the performance of models in various domains. Some multi-domain text classification models are trained by adversarial training to extract shared features among all domains and the specific features of each domain. We noted that the distinctness of the domain-specific features is different, so in this paper, we propose to use a curriculum learning strategy based on keyword weight ranking to improve the performance of multi-domain text classification models. The experimental results on the Amazon review and FDU-MTL datasets show that our curriculum learning strategy effectively improves the performance of multi-domain text classification models based on adversarial learning and outperforms state-of-the-art methods."
                    },
                    {
                        "title": "From Retrieval to Generation: Efficient and Effective Entity Set Expansion",
                        "abstract": "Entity Set Expansion (ESE) is a critical task aiming at expanding entities of the target semantic class described by seed entities. Most existing ESE methods are retrieval-based frameworks that need to extract contextual features of entities and calculate the similarity between seed entities and candidate entities. To achieve the two purposes, they iteratively traverse the corpus and the entity vocabulary, resulting in poor efficiency and scalability. Experimental results indicate that the time consumed by the retrieval-based ESE methods increases linearly with entity vocabulary and corpus size. In this paper, we firstly propose Generative Entity Set Expansion (GenExpan) framework, which utilizes a generative pre-trained auto-regressive language model to accomplish ESE task. Specifically, a prefix tree is employed to guarantee the validity of entity generation, and automatically generated class names are adopted to guide the model to generate target entities. Moreover, we propose Knowledge Calibration and Generative Ranking to further bridge the gap between generic knowledge of the language model and the goal of ESE task. For efficiency, expansion time consumed by GenExpan is independent of entity vocabulary and corpus size, and GenExpan achieves an average 600% speedup compared to strong baselines. For expansion effectiveness, our framework outperforms previous state-of-the-art ESE methods."
                    },
                    {
                        "title": "Investigating Graph Structure Information for Entity Alignment with Dangling Cases",
                        "abstract": "Entity alignment (EA) aims to discover the equivalent entities in different knowledge graphs (KGs), which play an important role in knowledge engineering. Recently, EA with dangling entities has been proposed as a more realistic setting, which assumes that not all entities have corresponding equivalent entities. In this paper, we focus on this setting. Some work has explored this problem by leveraging translation API, pre-trained word embeddings, and other off-the-shelf tools. However, these approaches over-rely on the side information (e.g., entity names), and fail to work when the side information is absent. On the contrary, they still insufficiently exploit the most fundamental graph structure information in KG. To improve the exploitation of the structural information, we propose a novel entity alignment framework called Weakly-Optimal Graph Contrastive Learning (WOGCL), which is refined on three dimensions : (i) Model. We propose a novel Gated Graph Attention Network to capture local and global graph structure similarity. (ii) Training. Two learning objectives: contrastive learning and optimal transport learning are designed to obtain distinguishable entity representations via the optimal transport plan. (iii) Inference. In the inference phase, a PageRank-based method is proposed to calculate higher-order structural similarity. Extensive experiments on two dangling benchmarks demonstrate that our WOGCL outperforms the current state-of-the-art methods with pure structural information in both traditional (relaxed) and dangling (consolidated) settings. The code will be public soon."
                    },
                    {
                        "title": "Prompt Learning With Knowledge Memorizing Prototypes For Generalized Few-Shot Intent Detection",
                        "abstract": "Generalized Few-Shot Intent Detection (GFSID) is challenging and realistic because it needs to categorize both seen and novel intents simultaneously. Previous GFSID methods rely on the episodic learning paradigm, which makes it hard to extend to a generalized setup as they do not explicitly learn the classification of seen categories and the knowledge of seen intents. To address the dilemma, we propose to convert the GFSID task into the class incremental learning paradigm. Specifically, we propose a two-stage learning framework, which sequentially learns the knowledge of different intents in various periods via prompt learning. And then we exploit prototypes for categorizing both seen and novel intents. Furthermore, to achieve the transfer knowledge of intents in different stages, for different scenarios we design two knowledge preservation methods which close to realistic applications. Extensive experiments and detailed analyses on two widely used datasets show that our framework based on the class incremental learning paradigm achieves promising performance."
                    },
                    {
                        "title": "Retrieval-Augmented Meta Learning for Low-Resource Text Classification",
                        "abstract": "Meta learning have achieved promising performance in low-resource text classification which aims to identify target classes with knowledge transferred from source classes with sets of small tasks named episodes. However, due to the limited training data in the meta-learning scenario and the inherent properties of parameterized neural networks, poor generalization performance has become a pressing problem that needs to be addressed. To deal with this issue, we propose a meta-learning based method called Retrieval-Augmented Meta Learning(RAML). It not only uses parameterization for inference but also retrieves non-parametric knowledge from an external corpus to make inferences, which greatly alleviates the problem of poor generalization performance caused by the lack of diverse training data in meta-learning. This method differs from previous models that solely rely on parameters, as it explicitly emphasizes the importance of non-parametric knowledge, aiming to strike a balance between parameterized neural networks and non-parametric knowledge. The model is required to determine which knowledge to access and utilize during inference. Additionally, our multi-view passages fusion network module can effectively and efficiently integrate the retrieved information into low-resource classification task. The extensive experiments demonstrate that RAML significantly outperforms current SOTA low-resource text classification models."
                    },
                    {
                        "title": "Contextual Similarity is More Valuable than Character Similarity: An Empirical Study for Chinese Spell Checking",
                        "abstract": "Chinese Spell Checking (CSC) task aims to detect and correct Chinese spelling errors. Recently, related researches focus on introducing character similarity from confusion set to enhance the CSC models, ignoring the context of characters that contain richer information. To make better use of contextual information, we propose a simple yet effective Curriculum Learning (CL) framework for the CSC task. With the help of our model-agnostic CL framework, existing CSC models will be trained from easy to difficult as humans learn Chinese characters and achieve further performance improvements. Extensive experiments and detailed analyses on widely used SIGHAN datasets show that our method outperforms previous state-of-the-art methods. More instructively, our study empirically suggests that contextual similarity is more valuable than character similarity for the CSC task."
                    },
                    {
                        "title": "Towards Attribute-Entangled Controllable Text Generation: A Pilot Study of Blessing Generation",
                        "abstract": "Controllable Text Generation (CTG) has obtained great success due to its fine-grained generation ability obtained by focusing on multiple attributes. However, most existing CTG researches overlook how to utilize the attribute entanglement to enhance the diversity of the controlled generated texts. Facing this dilemma, we focus on a novel CTG scenario, i.e., blessing generation which is challenging because high-quality blessing texts require CTG models to comprehensively consider the entanglement between multiple attributes (e.g., objects and occasions). To promote the research on blessing generation, we present EBleT, a large-scale Entangled Blessing Text dataset containing 293K English sentences annotated with multiple attributes. Furthermore, we propose novel evaluation metrics to measure the quality of the blessing texts generated by the baseline models we designed. Our study opens a new research direction for controllable text generation and enables the development of attribute-entangled CTG models. Our dataset and source codes are available at \\url{https://github.com/huangshulin123/Blessing-Generation}."
                    },
                    {
                        "title": "Active Relation Discovery: Towards General and Label-aware Open Relation Extraction",
                        "abstract": "Open Relation Extraction (OpenRE) aims to discover novel relations from open domains. Previous OpenRE methods mainly suffer from two problems: (1) Insufficient capacity to discriminate between known and novel relations. When extending conventional test settings to a more general setting where test data might also come from seen classes, existing approaches have a significant performance decline. (2) Secondary labeling must be performed before practical application. Existing methods cannot label human-readable and meaningful types for novel relations, which is urgently required by the downstream tasks. To address these issues, we propose the Active Relation Discovery (ARD) framework, which utilizes relational outlier detection for discriminating known and novel relations and involves active learning for labeling novel relations. Extensive experiments on three real-world datasets show that ARD significantly outperforms previous state-of-the-art methods on both conventional and our proposed general OpenRE settings. The source code and datasets will be available for reproducibility."
                    },
                    {
                        "title": "Mitigating Catastrophic Forgetting in Multi-domain Chinese Spelling Correction by Multi-stage Knowledge Transfer Framework",
                        "abstract": "Chinese Spelling Correction (CSC) aims to detect and correct spelling errors in given sentences. Recently, multi-domain CSC has gradually attracted the attention of researchers because it is more practicable. In this paper, we focus on the key flaw of the CSC model when adapting to multi-domain scenarios: the tendency to forget previously acquired knowledge upon learning new domain-specific knowledge (i.e., catastrophic forgetting). To address this, we propose a novel model-agnostic Multi-stage Knowledge Transfer (MKT) framework, which utilizes a continuously evolving teacher model for knowledge transfer in each domain, rather than focusing solely on new domain knowledge. It deserves to be mentioned that we are the first to apply continual learning methods to the multi-domain CSC task. Experiments prove the effectiveness of our proposed method, and further analyses demonstrate the importance of overcoming catastrophic forgetting for improving the model performance."
                    },
                    {
                        "title": "Automatic Context Pattern Generation for Entity Set Expansion",
                        "abstract": "Entity Set Expansion (ESE) is a valuable task that aims to find entities of the target semantic class described by given seed entities. Various Natural Language Processing (NLP) and Information Retrieval (IR) downstream applications have benefited from ESE due to its ability to discover knowledge. Although existing corpus-based ESE methods have achieved great progress, they still rely on corpora with high-quality entity information annotated, because most of them need to obtain the context patterns through the position of the entity in a sentence. Therefore, the quality of the given corpora and their entity annotation has become the bottleneck that limits the performance of such methods. To overcome this dilemma and make the ESE models free from the dependence on entity annotation, our work aims to explore a new ESE paradigm, namely corpus-independent ESE. Specifically, we devise a context pattern generation module that utilizes autoregressive language models (e.g., GPT-2) to automatically generate high-quality context patterns for entities. In addition, we propose the GAPA, a novel ESE framework that leverages the aforementioned GenerAted PAtterns to expand target entities. Extensive experiments and detailed analyses on three widely used datasets demonstrate the effectiveness of our method. All the codes of our experiments are available at https://github.com/geekjuruo/GAPA."
                    },
                    {
                        "title": "CLEME: Debiasing Multi-reference Evaluation for Grammatical Error Correction",
                        "abstract": "Evaluating the performance of Grammatical Error Correction (GEC) systems is a challenging task due to its subjectivity. Designing an evaluation metric that is as objective as possible is crucial to the development of GEC task. However, mainstream evaluation metrics, i.e., reference-based metrics, introduce bias into the multi-reference evaluation by extracting edits without considering the presence of multiple references. To overcome this issue, we propose Chunk-LEvel Multi-reference Evaluation (CLEME), designed to evaluate GEC systems in the multi-reference evaluation setting. CLEME builds chunk sequences with consistent boundaries for the source, the hypothesis and references, thus eliminating the bias caused by inconsistent edit boundaries. Furthermore, we observe the consistent boundary could also act as the boundary of grammatical errors, based on which the F$_{0.5}$ score is then computed following the correction independence assumption. We conduct experiments on six English reference sets based on the CoNLL-2014 shared task. Extensive experiments and detailed analyses demonstrate the correctness of our discovery and the effectiveness of CLEME. Further analysis reveals that CLEME is robust to evaluate GEC systems across reference sets with varying numbers of references and annotation style."
                    },
                    {
                        "title": "Bidirectional End-to-End Learning of Retriever-Reader Paradigm for Entity Linking",
                        "abstract": "Entity Linking (EL) is a fundamental task for Information Extraction and Knowledge Graphs. The general form of EL (i.e., end-to-end EL) aims to first find mentions in the given input document and then link the mentions to corresponding entities in a specific knowledge base. Recently, the paradigm of retriever-reader promotes the progress of end-to-end EL, benefiting from the advantages of dense entity retrieval and machine reading comprehension. However, the existing study only trains the retriever and the reader separately in a pipeline manner, which ignores the benefit that the interaction between the retriever and the reader can bring to the task. To advance the retriever-reader paradigm to perform more perfectly on end-to-end EL, we propose BEER$^2$, a Bidirectional End-to-End training framework for Retriever and Reader. Through our designed bidirectional end-to-end training, BEER$^2$ guides the retriever and the reader to learn from each other, make progress together, and ultimately improve EL performance. Extensive experiments on benchmarks of multiple domains demonstrate the effectiveness of our proposed BEER$^2$."
                    },
                    {
                        "title": "On the (In)Effectiveness of Large Language Models for Chinese Text Correction",
                        "abstract": "Recently, the development and progress of Large Language Models (LLMs) have amazed the entire Artificial Intelligence community. Benefiting from their emergent abilities, LLMs have attracted more and more researchers to study their capabilities and performance on various downstream Natural Language Processing (NLP) tasks. While marveling at LLMs' incredible performance on all kinds of tasks, we notice that they also have excellent multilingual processing capabilities, such as Chinese. To explore the Chinese processing ability of LLMs, we focus on Chinese Text Correction, a fundamental and challenging Chinese NLP task. Specifically, we evaluate various representative LLMs on the Chinese Grammatical Error Correction (CGEC) and Chinese Spelling Check (CSC) tasks, which are two main Chinese Text Correction scenarios. Additionally, we also fine-tune LLMs for Chinese Text Correction to better observe the potential capabilities of LLMs. From extensive analyses and comparisons with previous state-of-the-art small models, we empirically find that the LLMs currently have both amazing performance and unsatisfactory behavior for Chinese Text Correction. We believe our findings will promote the landing and application of LLMs in the Chinese NLP community."
                    }
                ]
            },
            "192fbc93-6562-4dcc-bf20-f4f954bba5ad": {
                "pk": "192fbc93-6562-4dcc-bf20-f4f954bba5ad",
                "name": "Hai-Tao Zheng",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "bf88df53-f621-412d-bb4a-75502bb34794": {
                "pk": "bf88df53-f621-412d-bb4a-75502bb34794",
                "name": "Xuming Hu",
                "collaborators": [
                    "Lijie Wen",
                    "Philip S. Yu",
                    "Aiwei Liu",
                    "Li Lin",
                    "Yibo Yan",
                    "Zhijiang Guo",
                    "Chenwei Zhang",
                    "Shu'ang Li",
                    "Sisuo Lyu",
                    "Xiuze Zhou"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Contrastive Learning",
                    "Fact-Checking",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Pair-Level Supervised Contrastive Learning for Natural Language Inference",
                        "abstract": "Natural language inference (NLI) is an increasingly important task for natural language understanding, which requires one to infer the relationship between the sentence pair (premise and hypothesis). Many recent works have used contrastive learning by incorporating the relationship of the sentence pair from NLI datasets to learn sentence representation. However, these methods only focus on comparisons with sentence-level representations. In this paper, we propose a Pair-level Supervised Contrastive Learning approach (PairSCL). We adopt a cross attention module to learn the joint representations of the sentence pairs. A contrastive learning objective is designed to distinguish the varied classes of sentence pairs by pulling those in one class together and pushing apart the pairs in other classes. We evaluate PairSCL on two public datasets of NLI where the accuracy of PairSCL outperforms other methods by 2.1% on average. Furthermore, our method outperforms the previous state-of-the-art method on seven transfer tasks of text classification."
                    },
                    {
                        "title": "Multi-view Hypergraph-based Contrastive Learning Model for Cold-Start Micro-video Recommendation",
                        "abstract": "With the widespread use of mobile devices and the rapid growth of micro-video platforms such as TikTok and Kwai, the demand for personalized micro-video recommendation systems has significantly increased. Micro-videos typically contain diverse information, such as textual metadata, visual cues (e.g., cover images), and dynamic video content, significantly affecting user interaction and engagement patterns. However, most existing approaches often suffer from the problem of over-smoothing, which limits their ability to capture comprehensive interaction information effectively. Additionally, cold-start scenarios present ongoing challenges due to sparse interaction data and the underutilization of available interaction signals.   To address these issues, we propose a Multi-view Hypergraph-based Contrastive learning model for cold-start micro-video Recommendation (MHCR). MHCR introduces a multi-view multimodal feature extraction layer to capture interaction signals from various perspectives and incorporates multi-view self-supervised learning tasks to provide additional supervisory signals. Through extensive experiments on two real-world datasets, we show that MHCR significantly outperforms existing video recommendation models and effectively mitigates cold-start challenges. Our code is available at https://anonymous.4open.science/r/MHCR-02EF."
                    },
                    {
                        "title": "LongGenBench: Long-context Generation Benchmark",
                        "abstract": "Current long-context benchmarks primarily focus on retrieval-based tests, requiring Large Language Models (LLMs) to locate specific information within extensive input contexts, such as the needle-in-a-haystack (NIAH) benchmark. Long-context generation refers to the ability of a language model to generate coherent and contextually accurate text that spans across lengthy passages or documents. While recent studies show strong performance on NIAH and other retrieval-based long-context benchmarks, there is a significant lack of benchmarks for evaluating long-context generation capabilities. To bridge this gap and offer a comprehensive assessment, we introduce a synthetic benchmark, LongGenBench, which allows for flexible configurations of customized generation context lengths. LongGenBench advances beyond traditional benchmarks by redesigning the format of questions and necessitating that LLMs respond with a single, cohesive long-context answer. Upon extensive evaluation using LongGenBench, we observe that: (1) both API accessed and open source models exhibit performance degradation in long-context generation scenarios, ranging from 1.2% to 47.1%; (2) different series of LLMs exhibit varying trends of performance degradation, with the Gemini-1.5-Flash model showing the least degradation among API accessed models, and the Qwen2 series exhibiting the least degradation in LongGenBench among open source models."
                    },
                    {
                        "title": "MMNeuron: Discovering Neuron-Level Domain-Specific Interpretation in Multimodal Large Language Model",
                        "abstract": "Projecting visual features into word embedding space has become a significant fusion strategy adopted by Multimodal Large Language Models (MLLMs). However, its internal mechanisms have yet to be explored. Inspired by multilingual research, we identify domain-specific neurons in multimodal large language models. Specifically, we investigate the distribution of domain-specific neurons and the mechanism of how MLLMs process features from diverse domains. Furthermore, we propose a three-stage mechanism for language model modules in MLLMs when handling projected image features, and verify this hypothesis using logit lens. Extensive experiments indicate that while current MLLMs exhibit Visual Question Answering (VQA) capability, they may not fully utilize domain-specific information. Manipulating domain-specific neurons properly will result in a 10% change of accuracy at most, shedding light on the development of cross-domain, all-encompassing MLLMs in the future. The source code is available at https://github.com/Z1zs/MMNeuron."
                    },
                    {
                        "title": "Semantic Enhanced Text-to-SQL Parsing via Iteratively Learning Schema Linking Graph",
                        "abstract": "The generalizability to new databases is of vital importance to Text-to-SQL systems which aim to parse human utterances into SQL statements. Existing works achieve this goal by leveraging the exact matching method to identify the lexical matching between the question words and the schema items. However, these methods fail in other challenging scenarios, such as the synonym substitution in which the surface form differs between the corresponding question words and schema items. In this paper, we propose a framework named ISESL-SQL to iteratively build a semantic enhanced schema-linking graph between question tokens and database schemas. First, we extract a schema linking graph from PLMs through a probing procedure in an unsupervised manner. Then the schema linking graph is further optimized during the training process through a deep graph learning method. Meanwhile, we also design an auxiliary task called graph regularization to improve the schema information mentioned in the schema-linking graph. Extensive experiments on three benchmarks demonstrate that ISESL-SQL could consistently outperform the baselines and further investigations show its generalizability and robustness."
                    },
                    {
                        "title": "A comprehensive evaluation of ChatGPT's zero-shot Text-to-SQL capability",
                        "abstract": "This paper presents the first comprehensive analysis of ChatGPT's Text-to-SQL ability. Given the recent emergence of large-scale conversational language model ChatGPT and its impressive capabilities in both conversational abilities and code generation, we sought to evaluate its Text-to-SQL performance. We conducted experiments on 12 benchmark datasets with different languages, settings, or scenarios, and the results demonstrate that ChatGPT has strong text-to-SQL abilities. Although there is still a gap from the current state-of-the-art (SOTA) model performance, considering that the experiment was conducted in a zero-shot scenario, ChatGPT's performance is still impressive. Notably, in the ADVETA (RPL) scenario, the zero-shot ChatGPT even outperforms the SOTA model that requires fine-tuning on the Spider dataset by 4.1\\%, demonstrating its potential for use in practical applications. To support further research in related fields, we have made the data generated by ChatGPT publicly available at https://github.com/THU-BPM/chatgpt-sql."
                    },
                    {
                        "title": "Give Me More Details: Improving Fact-Checking with Latent Retrieval",
                        "abstract": "Evidence plays a crucial role in automated fact-checking. When verifying real-world claims, existing fact-checking systems either assume the evidence sentences are given or use the search snippets returned by the search engine. Such methods ignore the challenges of collecting evidence and may not provide sufficient information to verify real-world claims. Aiming at building a better fact-checking system, we propose to incorporate full text from source documents as evidence and introduce two enriched datasets. The first one is a multilingual dataset, while the second one is monolingual (English). We further develop a latent variable model to jointly extract evidence sentences from documents and perform claim verification. Experiments indicate that including source documents can provide sufficient contextual clues even when gold evidence sentences are not annotated. The proposed system is able to achieve significant improvements upon best-reported models under different settings."
                    },
                    {
                        "title": "Reefknot: A Comprehensive Benchmark for Relation Hallucination Evaluation, Analysis and Mitigation in Multimodal Large Language Models",
                        "abstract": "Hallucination issues persistently plagued current multimodal large language models (MLLMs). While existing research primarily focuses on object-level or attribute-level hallucinations, sidelining the more sophisticated relation hallucinations that necessitate advanced reasoning abilities from MLLMs. Besides, recent benchmarks regarding relation hallucinations lack in-depth evaluation and effective mitigation. Moreover, their datasets are typically derived from a systematic annotation process, which could introduce inherent biases due to the predefined process. To handle the aforementioned challenges, we introduce Reefknot, a comprehensive benchmark specifically targeting relation hallucinations, consisting of over 20,000 samples derived from real-world scenarios. Specifically, we first provide a systematic definition of relation hallucinations, integrating perspectives from perceptive and cognitive domains. Furthermore, we construct the relation-based corpus utilizing the representative scene graph dataset Visual Genome (VG), from which semantic triplets follow real-world distributions. Our comparative evaluation across three distinct tasks revealed a substantial shortcoming in the capabilities of current MLLMs to mitigate relation hallucinations. Finally, we advance a novel confidence-based mitigation strategy tailored to tackle the relation hallucinations problem. Across three datasets, including Reefknot, we observed an average reduction of 9.75% in the hallucination rate. We believe our paper sheds valuable insights into achieving trustworthy multimodal intelligence. Our dataset and code will be released upon paper acceptance."
                    },
                    {
                        "title": "SelfORE: Self-supervised Relational Feature Learning for Open Relation Extraction",
                        "abstract": "Open relation extraction is the task of extracting open-domain relation facts from natural language sentences. Existing works either utilize heuristics or distant-supervised annotations to train a supervised classifier over pre-defined relations, or adopt unsupervised methods with additional assumptions that have less discriminative power. In this work, we proposed a self-supervised framework named SelfORE, which exploits weak, self-supervised signals by leveraging large pretrained language model for adaptive clustering on contextualized relational features, and bootstraps the self-supervised signals by improving contextualized features in relation classification. Experimental results on three datasets show the effectiveness and robustness of SelfORE on open-domain Relation Extraction when comparing with competitive baselines."
                    },
                    {
                        "title": "Semi-supervised Relation Extraction via Incremental Meta Self-Training",
                        "abstract": "To alleviate human efforts from obtaining large-scale annotations, Semi-Supervised Relation Extraction methods aim to leverage unlabeled data in addition to learning from limited samples. Existing self-training methods suffer from the gradual drift problem, where noisy pseudo labels on unlabeled data are incorporated during training. To alleviate the noise in pseudo labels, we propose a method called MetaSRE, where a Relation Label Generation Network generates quality assessment on pseudo labels by (meta) learning from the successful and failed attempts on Relation Classification Network as an additional meta-objective. To reduce the influence of noisy pseudo labels, MetaSRE adopts a pseudo label selection and exploitation scheme which assesses pseudo label quality on unlabeled samples and only exploits high-quality pseudo labels in a self-training fashion to incrementally augment labeled samples for both robustness and accuracy. Experimental results on two public datasets demonstrate the effectiveness of the proposed approach."
                    },
                    {
                        "title": "CHEF: A Pilot Chinese Dataset for Evidence-Based Fact-Checking",
                        "abstract": "The explosion of misinformation spreading in the media ecosystem urges for automated fact-checking. While misinformation spans both geographic and linguistic boundaries, most work in the field has focused on English. Datasets and tools available in other languages, such as Chinese, are limited. In order to bridge this gap, we construct CHEF, the first CHinese Evidence-based Fact-checking dataset of 10K real-world claims. The dataset covers multiple domains, ranging from politics to public health, and provides annotated evidence retrieved from the Internet. Further, we develop established baselines and a novel approach that is able to model the evidence retrieval as a latent variable, allowing jointly training with the veracity prediction model in an end-to-end fashion. Extensive experiments show that CHEF will provide a challenging testbed for the development of fact-checking systems designed to retrieve and reason over non-English claims."
                    }
                ]
            },
            "ce53855b-ff5b-471b-ae3d-6bd864bb45ba": {
                "pk": "ce53855b-ff5b-471b-ae3d-6bd864bb45ba",
                "name": "Philip S. Yu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2305.16846": {
        "paper_data": {
            "title": "Lagrangian Flow Networks for Conservation Laws",
            "url": "http://arxiv.org/abs/2305.16846v2",
            "arxiv_id": "2305.16846",
            "authors": [
                "F. Arend Torres",
                "Marcello Massimo Negri",
                "Marco Inversi",
                "Jonathan Aellen",
                "Volker Roth"
            ],
            "abstract": "We introduce Lagrangian Flow Networks (LFlows) for modeling fluid densities and velocities continuously in space and time. By construction, the proposed LFlows satisfy the continuity equation, a PDE describing mass conservation in its differentiable form. Our model is based on the insight that solutions to the continuity equation can be expressed as time-dependent density transformations via differentiable and invertible maps. This follows from classical theory of the existence and uniqueness of Lagrangian flows for smooth vector fields. Hence, we model fluid densities by transforming a base density with parameterized diffeomorphisms conditioned on time. The key benefit compared to methods relying on numerical ODE solvers or PINNs is that the analytic expression of the velocity is always consistent with changes in density. Furthermore, we require neither expensive numerical solvers, nor additional penalties to enforce the PDE. LFlows show higher predictive accuracy in density modeling tasks compared to competing models in 2D and 3D, while being computationally efficient. As a real-world application, we model bird migration based on sparse weather radar measurements.",
            "introduction": "   1 Introduction  The development of physics-informed Machine Learning (PI-ML) (Karniadakis et\u00a0al., 2021) opens new opportunities to combine the power of modern ML methods with physical constraints that serve as meaningful regularizers. These constraints might, for example, be available in the form of partial differential equations (PDEs). Within PI-ML we consider hydrodynamic flow problems governed by the physical law of mass conservation. This law is described in its local and differentiable form by a PDE commonly known as the continuity equation (CE)      {\u2202t\u03c1+\u2207\u22c5(\ud835\udc97\u2062\u03c1)=0(t,\ud835\udc99)\u2208(t0,T)\u00d7\u03a9,\u03c1\u2062(t0,\ud835\udc99)=\u03c1t0\u2062(\ud835\udc99)\ud835\udc99\u2208\u03a9.\\displaystyle\\left\\{\\begin{aligned} \\partial_{t}\\rho+\\nabla\\cdot(\\bm{v}\\rho)&=% 0&(t,\\bm{x})\\in(t_{0},T)\\times\\Omega,\\\\ \\rho(t_{0},\\bm{x})&=\\rho_{t_{0}}(\\bm{x})&\\bm{x}\\in\\Omega.\\end{aligned}\\right.{ start_ROW start_CELL \u2202 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT italic_\u03c1 + \u2207 \u22c5 ( bold_italic_v italic_\u03c1 ) end_CELL start_CELL = 0 end_CELL start_CELL ( italic_t , bold_italic_x ) \u2208 ( italic_t start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_T ) \u00d7 roman_\u03a9 , end_CELL end_ROW start_ROW start_CELL italic_\u03c1 ( italic_t start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , bold_italic_x ) end_CELL start_CELL = italic_\u03c1 start_POSTSUBSCRIPT italic_t start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT end_POSTSUBSCRIPT ( bold_italic_x ) end_CELL start_CELL bold_italic_x \u2208 roman_\u03a9 . end_CELL end_ROW  (1)   For any time t\u2208[t0,T)\ud835\udc61subscript\ud835\udc610\ud835\udc47t\\in[t_{0},T)italic_t \u2208 [ italic_t start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_T ) the function \u03c1\u2062(t,\u22c5)\ud835\udf0c\ud835\udc61\u22c5\\rho(t,\\cdot)italic_\u03c1 ( italic_t , \u22c5 ) can be thought of as the density of parcels advected by the velocity field \ud835\udc97\ud835\udc97\\bm{v}bold_italic_v, with initial density \u03c1t0subscript\ud835\udf0csubscript\ud835\udc610\\rho_{t_{0}}italic_\u03c1 start_POSTSUBSCRIPT italic_t start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT end_POSTSUBSCRIPT. Here, [t0,T]\u00d7\u03a9subscript\ud835\udc610\ud835\udc47\u03a9[t_{0},T]\\times\\Omega[ italic_t start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_T ] \u00d7 roman_\u03a9 is the space-time domain, which is a subset of \u211d\u00d7\u211dd\u211dsuperscript\u211d\ud835\udc51\\mathbb{R}\\times\\mathbb{R}^{d}blackboard_R \u00d7 blackboard_R start_POSTSUPERSCRIPT italic_d end_POSTSUPERSCRIPT. The partial derivative w.r.t.\u00a0time t\ud835\udc61titalic_t is denoted by \u2202tsubscript\ud835\udc61\\partial_{t}\u2202 start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT and \u2207\u22c5\ud835\udc83=\u2207\ud835\udc99\u22c5\ud835\udc83=\u2211i=1d\u2202\ud835\udc83i\u2202xi\u22c5\u2207\ud835\udc83\u22c5subscript\u2207\ud835\udc99\ud835\udc83superscriptsubscript\ud835\udc561\ud835\udc51subscript\ud835\udc83\ud835\udc56subscript\ud835\udc65\ud835\udc56\\nabla\\cdot\\bm{b}=\\nabla_{\\bm{x}}\\cdot\\bm{b}=\\sum_{i=1}^{d}\\frac{\\partial{\\bm{% b}}_{i}}{\\partial x_{i}}\u2207 \u22c5 bold_italic_b = \u2207 start_POSTSUBSCRIPT bold_italic_x end_POSTSUBSCRIPT \u22c5 bold_italic_b = \u2211 start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_d end_POSTSUPERSCRIPT divide start_ARG \u2202 bold_italic_b start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT end_ARG start_ARG \u2202 italic_x start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT end_ARG is the spatial divergence of a d\ud835\udc51ditalic_d dimensional vector field \ud835\udc83:[t0,T]\u00d7\u03a9\u21a6\u211dd:\ud835\udc83maps-tosubscript\ud835\udc610\ud835\udc47\u03a9superscript\u211d\ud835\udc51\\bm{b}:[t_{0},T]\\times\\Omega\\mapsto\\mathbb{R}^{d}bold_italic_b : [ italic_t start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_T ] \u00d7 roman_\u03a9 \u21a6 blackboard_R start_POSTSUPERSCRIPT italic_d end_POSTSUPERSCRIPT.   Unlike classical initial value problems, we consider challenging settings where exact boundary and initial conditions are unknown. That is, the continuous density and velocity fields have to be inferred from sparse and noisy data. The important physical constraint is that the solution must comply with the CE in Eq.\u00a01. To this end we propose a neural network based model that fulfills the CE by construction and provides physically consistent velocity and density fields.   We specifically consider two distinct application settings. In both settings we are restricted to sparse and noisy measurements. In setting (i) we measure the fluid density \u03c1\ud835\udf0c\\rhoitalic_\u03c1 and velocity \ud835\udc97\ud835\udc97\\bm{v}bold_italic_v, without knowing any additional equations aside from the CE. This occurs for example within the area of radar ornithology (Chilson et\u00a0al., 2017), where density and velocity of birds can be inferred from radar data. Such radar-based measurements are to date the only practical high-throughput data source for birds. The goal is to spatiotemporally estimate brid migration densities (Nussbaumer et\u00a0al., 2019). Since more migration-specific dynamics are unknown, mass conservation is usually considered as a physical constraint (Nussbaumer et\u00a0al., 2021; Lippert et\u00a0al., 2022a; b). In setting (ii) we exclusively have sparse density observations, but we know additional equations constraining the velocity. This might, for example, occur in dynamical optimal transport problems. Here, two densities are to be interpolated while adhering to the CE. Therefore, minimizing the total transport cost constrains the velocity field. More generally,",
            "references": [
                {
                    "title": "Self-Consistent Velocity Matching of Probability Flows",
                    "abstract": "We present a discretization-free scalable framework for solving a large class of mass-conserving partial differential equations (PDEs), including the time-dependent Fokker-Planck equation and the Wasserstein gradient flow. The main observation is that the time-varying velocity field of the PDE solution needs to be self-consistent: it must satisfy a fixed-point equation involving the probability flow characterized by the same velocity field. Instead of directly minimizing the residual of the fixed-point equation with neural parameterization, we use an iterative formulation with a biased gradient estimator that bypasses significant computational obstacles with strong empirical performance. Compared to existing approaches, our method does not suffer from temporal or spatial discretization, covers a wider range of PDEs, and scales to high dimensions. Experimentally, our method recovers analytical solutions accurately when they are available and achieves superior performance in high dimensions with less training time compared to alternatives."
                },
                {
                    "title": "Physics-informed inference of aerial animal movements from weather radar data",
                    "abstract": "Studying animal movements is essential for effective wildlife conservation and conflict mitigation. For aerial movements, operational weather radars have become an indispensable data source in this respect. However, partial measurements, incomplete spatial coverage, and poor understanding of animal behaviours make it difficult to reconstruct complete spatio-temporal movement patterns from available radar data. We tackle this inverse problem by learning a mapping from high-dimensional radar measurements to low-dimensional latent representations using a convolutional encoder. Under the assumption that the latent system dynamics are well approximated by a locally linear Gaussian transition model, we perform efficient posterior estimation using the classical Kalman smoother. A convolutional decoder maps the inferred latent system states back to the physical space in which the known radar observation model can be applied, enabling fully unsupervised training. To encourage physical consistency, we additionally introduce a physics-informed loss term that leverages known mass conservation constraints. Our experiments on synthetic radar data show promising results in terms of reconstruction quality and data-efficiency."
                },
                {
                    "title": "Learning to predict spatiotemporal movement dynamics from weather radar networks",
                    "abstract": "Weather radar networks provide wide\u2010ranging opportunities for ecologists to quantify and predict movements of airborne organisms over unprecedented geographical expanses. The typically sparse spatial distribution of radar measurements poses, however, a major challenge to spatiotemporal predictive modelling. We propose FluxRGNN, a recurrent graph neural network that is based on a generic mechanistic description of population\u2010level movements across the Voronoi tessellation of radar sites. The resulting hybrid model capitalises on local associations between environmental conditions and animal density as well as on spatiotemporal dependencies inherent to the movement process. We applied FluxRGNN to make 72\u2010h forecasts of nocturnal bird migration over Western Europe using simulated trajectories and measurements from the European weather radar network. For both datasets, FluxRGNN achieves higher predictive performance than baseline models based on environmental conditions alone. It effectively disentangles local take\u2010off and landing dynamics from aerial movements and correctly predicts migration directions with an accuracy of 87%. Continental\u2010scale forecasts of animal density and biomass fluxes have the potential to improve the impact and cost\u2010effectiveness of wildlife management and conservation efforts. With FluxRGNN this becomes feasible for nocturnal bird migration. In the future, other migration systems could benefit from applying the proposed method to similar static sensor networks."
                },
                {
                    "title": "Neural Conservation Laws: A Divergence-Free Perspective",
                    "abstract": "We investigate the parameterization of deep neural networks that by design satisfy the continuity equation, a fundamental conservation law. This is enabled by the observation that any solution of the continuity equation can be represented as a divergence-free vector field. We hence propose building divergence-free neural networks through the concept of differential forms, and with the aid of automatic differentiation, realize two practical constructions. As a result, we can parameterize pairs of densities and vector fields that always exactly satisfy the continuity equation, foregoing the need for extra penalty methods or expensive numerical simulation. Furthermore, we prove these models are universal and so can be used to represent any divergence-free vector field. Finally, we experimentally validate our approaches by computing neural network-based solutions to fluid equations, solving for the Hodge decomposition, and learning dynamical optimal transport maps."
                },
                {
                    "title": "Mesh-free Eulerian Physics-Informed Neural Networks",
                    "abstract": "Physics-informed Neural Networks (PINNs) have recently emerged as a principled way to include prior physical knowledge in form of partial differential equations (PDEs) into neural networks. Although PINNs are generally viewed as mesh-free, current approaches still rely on collocation points within a bounded region, even in settings with spatially sparse signals. Furthermore, if the boundaries are not known, the selection of such a region is dif\ufb01cult and often results in a large proportion of collocation points being selected in areas of low relevance. To resolve this severe drawback of current methods, we present a mesh-free and adaptive approach termed particle-density PINN (pdPINN), which is inspired by the microscopic viewpoint of \ufb02uid dynamics. The method is based on the Eulerian formulation and, different from classical mesh-free method, does not require the introduction of Lagrangian updates. We propose to sample directly from the distribution over the particle positions, eliminating the need to introduce boundaries while adaptively focusing on the most relevant regions. This is achieved by interpreting a non-negative physical quantity (such as the density or temperature) as an unnormalized probability distribution from which we sample with dynamic Monte Carlo methods. The proposed method leads to higher sample ef\ufb01ciency and improved performance of PINNs. These advantages are demonstrated on various experiments based on the continuity equations, Fokker-Planck equations, and the heat equation."
                },
                {
                    "title": "Neural Flows: Efficient Alternative to Neural ODEs",
                    "abstract": "Neural ordinary differential equations describe how values change in time. This is the reason why they gained importance in modeling sequential data, especially when the observations are made at irregular intervals. In this paper we propose an alternative by directly modeling the solution curves - the flow of an ODE - with a neural network. This immediately eliminates the need for expensive numerical solvers while still maintaining the modeling capability of neural ODEs. We propose several flow architectures suitable for different applications by establishing precise conditions on when a function defines a valid flow. Apart from computational efficiency, we also provide empirical evidence of favorable generalization performance via applications in time series modeling, forecasting, and density estimation."
                },
                {
                    "title": "Jacobian Determinant of Normalizing Flows",
                    "abstract": "Normalizing flows learn a diffeomorphic mapping between the target and base distribution, while the Jacobian determinant of that mapping forms another real-valued function. In this paper, we show that the Jacobian determinant mapping is unique for the given distributions, hence the likelihood objective of flows has a unique global optimum. In particular, the likelihood for a class of flows is explicitly expressed by the eigenvalues of the auto-correlation matrix of individual data point, and independent of the parameterization of neural network, which provides a theoretical optimal value of likelihood objective and relates to probabilistic PCA. Additionally, Jacobian determinant is a measure of local volume change and is maximized when MLE is used for optimization. To stabilize normalizing flows training, it is required to maintain a balance between the expansiveness and contraction of volume, meaning Lipschitz constraint on the diffeomorphic mapping and its inverse. With these theoretical results, several principles of designing normalizing flow were proposed. And numerical experiments on highdimensional datasets (such as CelebA-HQ 1024x1024) were conducted to show the improved stability of training."
                },
                {
                    "title": "Invertible DenseNets with Concatenated LipSwish",
                    "abstract": "We introduce Invertible Dense Networks (i-DenseNets), a more parameter efficient extension of Residual Flows. The method relies on an analysis of the Lipschitz continuity of the concatenation in DenseNets, where we enforce invertibility of the network by satisfying the Lipschitz constant. Furthermore, we propose a learnable weighted concatenation, which not only improves the model performance but also indicates the importance of the concatenated weighted representation. Additionally, we introduce the Concatenated LipSwish as activation function, for which we show how to enforce the Lipschitz condition and which boosts performance. The new architecture, i-DenseNet, out-performs Residual Flow and other flow-based models on density estimation evaluated in bits per dimension, where we utilize an equal parameter budget. Moreover, we show that the proposed model out-performs Residual Flows when trained as a hybrid model where the model is both a generative and a discriminative model."
                },
                {
                    "title": "Neural Spatio-Temporal Point Processes",
                    "abstract": "We propose a new class of parameterizations for spatio-temporal point processes which leverage Neural ODEs as a computational method and enable flexible, high-fidelity models of discrete events that are localized in continuous time and space. Central to our approach is a combination of recurrent continuous-time neural networks with two novel neural architectures, i.e., Jump and Attentive Continuous-time Normalizing Flows. This approach allows us to learn complex distributions for both the spatial and temporal domain and to condition non-trivially on the observed event history. We validate our models on data sets from a wide variety of contexts such as seismology, epidemiology, urban mobility, and neuroscience."
                },
                {
                    "title": "Fourier Neural Operator for Parametric Partial Differential Equations",
                    "abstract": "The classical development of neural networks has primarily focused on learning mappings between finite-dimensional Euclidean spaces. Recently, this has been generalized to neural operators that learn mappings between function spaces. For partial differential equations (PDEs), neural operators directly learn the mapping from any functional parametric dependence to the solution. Thus, they learn an entire family of PDEs, in contrast to classical methods which solve one instance of the equation. In this work, we formulate a new neural operator by parameterizing the integral kernel directly in Fourier space, allowing for an expressive and efficient architecture. We perform experiments on Burgers' equation, Darcy flow, and the Navier-Stokes equation (including the turbulent regime). Our Fourier neural operator shows state-of-the-art performance compared to existing neural network methodologies and it is up to three orders of magnitude faster compared to traditional PDE solvers."
                },
                {
                    "title": "Quantifying year-round nocturnal bird migration with a fluid dynamics model",
                    "abstract": "To understand the influence of biomass flows on ecosystems, we need to characterize and quantify migrations at various spatial and temporal scales. Representing the movements of migrating birds as a fluid, we applied a flow model to bird density and velocity maps retrieved from the European weather radar network, covering almost a year. We quantified how many birds take-off, fly, and land across Western Europe to (1) track bird migration waves between nights, (2) cumulate the number of birds on the ground and (3) quantify the seasonal flow into and out of the study area through several regional transects. Our results identified several migration waves that crossed the study area in 4 days only and included up to 188 million (M) birds that took-off in a single night. In spring, we estimated that 494 M birds entered the study area, 251 M left it, and 243 M birds remained within the study area. In autumn, 314 M birds entered the study area while 858 M left it. In addition to identifying fundamental quantities, our study highlights the potential of combining interdisciplinary data and methods to elucidate the dynamics of avian migration from nightly to yearly time scales and from regional to continental spatial scales."
                },
                {
                    "title": "Implicit Neural Representations with Periodic Activation Functions",
                    "abstract": "Implicitly defined, continuous, differentiable signal representations parameterized by neural networks have emerged as a powerful paradigm, offering many possible benefits over conventional representations. However, current network architectures for such implicit neural representations are incapable of modeling signals with fine detail, and fail to represent a signal's spatial and temporal derivatives, despite the fact that these are essential to many physical signals defined implicitly as the solution to partial differential equations. We propose to leverage periodic activation functions for implicit neural representations and demonstrate that these networks, dubbed sinusoidal representation networks or Sirens, are ideally suited for representing complex natural signals and their derivatives. We analyze Siren activation statistics to propose a principled initialization scheme and demonstrate the representation of images, wavefields, video, sound, and their derivatives. Further, we show how Sirens can be leveraged to solve challenging boundary value problems, such as particular Eikonal equations (yielding signed distance functions), the Poisson equation, and the Helmholtz and wave equations. Lastly, we combine Sirens with hypernetworks to learn priors over the space of Siren functions."
                },
                {
                    "title": "The ERA5 global reanalysis",
                    "abstract": "Within the Copernicus Climate Change Service (C3S), ECMWF is producing the ERA5 reanalysis which, once completed, will embody a detailed record of the global atmosphere, land surface and ocean waves from 1950 onwards. This new reanalysis replaces the ERA\u2010Interim reanalysis (spanning 1979 onwards) which was started in 2006. ERA5 is based on the Integrated Forecasting System (IFS) Cy41r2 which was operational in 2016. ERA5 thus benefits from a decade of developments in model physics, core dynamics and data assimilation. In addition to a significantly enhanced horizontal resolution of 31\u2009km, compared to 80\u2009km for ERA\u2010Interim, ERA5 has hourly output throughout, and an uncertainty estimate from an ensemble (3\u2010hourly at half the horizontal resolution). This paper describes the general set\u2010up of ERA5, as well as a basic evaluation of characteristics and performance, with a focus on the dataset from 1979 onwards which is currently publicly available. Re\u2010forecasts from ERA5 analyses show a gain of up to one day in skill with respect to ERA\u2010Interim. Comparison with radiosonde and PILOT data prior to assimilation shows an improved fit for temperature, wind and humidity in the troposphere, but not the stratosphere. A comparison with independent buoy data shows a much improved fit for ocean wave height. The uncertainty estimate reflects the evolution of the observing systems used in ERA5. The enhanced temporal and spatial resolution allows for a detailed evolution of weather systems. For precipitation, global\u2010mean correlation with monthly\u2010mean GPCP data is increased from 67% to 77%. In general, low\u2010frequency variability is found to be well represented and from 10\u2009hPa downwards general patterns of anomalies in temperature match those from the ERA\u2010Interim, MERRA\u20102 and JRA\u201055 reanalyses."
                },
                {
                    "title": "OT-Flow: Fast and Accurate Continuous Normalizing Flows via Optimal Transport",
                    "abstract": "A normalizing flow is an invertible mapping between an arbitrary probability distribution and a standard normal distribution; it can be used for density estimation and statistical inference. Computing the flow follows the change of variables formula and thus requires invertibility of the mapping and an efficient way to compute the determinant of its Jacobian. To satisfy these requirements, normalizing flows typically consist of carefully chosen components. Continuous normalizing flows (CNFs) are mappings obtained by solving a neural ordinary differential equation (ODE). The neural ODE's dynamics can be chosen almost arbitrarily while ensuring invertibility. Moreover, the log-determinant of the flow's Jacobian can be obtained by integrating the trace of the dynamics' Jacobian along the flow. Our proposed OT-Flow approach tackles two critical computational challenges that limit a more widespread use of CNFs. First, OT-Flow leverages optimal transport (OT) theory to regularize the CNF and enforce straight trajectories that are easier to integrate. Second, OT-Flow features exact trace computation with time complexity equal to trace estimators used in existing CNFs. On five high-dimensional density estimation and generative modeling tasks, OT-Flow performs competitively to state-of-the-art CNFs while on average requiring one-fourth of the number of weights with an 8x speedup in training time and 24x speedup in inference."
                },
                {
                    "title": "Normalizing Flows: An Introduction and Review of Current Methods",
                    "abstract": "Normalizing Flows are generative models which produce tractable distributions where both sampling and density evaluation can be efficient and exact. The goal of this survey article is to give a coherent and comprehensive review of the literature around the construction and use of Normalizing Flows for distribution learning. We aim to provide context and explanation of the models, review current state-of-the-art literature, and identify open questions and promising future directions."
                },
                {
                    "title": "Lagrangian Neural Networks",
                    "abstract": "Accurate models of the world are built upon notions of its underlying symmetries. In physics, these symmetries correspond to conservation laws, such as for energy and momentum. Yet even though neural network models see increasing use in the physical sciences, they struggle to learn these symmetries. In this paper, we propose Lagrangian Neural Networks (LNNs), which can parameterize arbitrary Lagrangians using neural networks. In contrast to models that learn Hamiltonians, LNNs do not require canonical coordinates, and thus perform well in situations where canonical momenta are unknown or difficult to compute. Unlike previous approaches, our method does not restrict the functional form of learned energies and will produce energy-conserving models for a variety of tasks. We test our approach on a double pendulum and a relativistic particle, demonstrating energy conservation where a baseline approach incurs dissipation and modeling relativity without canonical coordinates where a Hamiltonian approach fails. Finally, we show how this model can be applied to graphs and continuous systems using a Lagrangian Graph Network, and demonstrate it on the 1D wave equation."
                },
                {
                    "title": "How to train your neural ODE",
                    "abstract": "Training neural ODEs on large datasets has not been tractable due to the necessity of allowing the adaptive numerical ODE solver to refine its step size to very small values. In practice this leads to dynamics equivalent to many hundreds or even thousands of layers. In this paper, we overcome this apparent difficulty by introducing a theoretically-grounded combination of both optimal transport and stability regularizations which encourage neural ODEs to prefer simpler dynamics out of all the dynamics that solve a problem well. Simpler dynamics lead to faster convergence and to fewer discretizations of the solver, considerably decreasing wall-clock time without loss in performance. Our approach allows us to train neural ODE based generative models to the same performance as the unregularized dynamics in just over a day on one GPU, whereas unregularized dynamics can take up to 4-6 days of training time on multiple GPUs. This brings neural ODEs significantly closer to practical relevance in large-scale applications."
                },
                {
                    "title": "Normalizing Flows for Probabilistic Modeling and Inference",
                    "abstract": "Normalizing flows provide a general mechanism for defining expressive probability distributions, only requiring the specification of a (usually simple) base distribution and a series of bijective transformations. There has been much recent work on normalizing flows, ranging from improving their expressive power to expanding their application. We believe the field has now matured and is in need of a unified perspective. In this review, we attempt to provide such a perspective by describing flows through the lens of probabilistic modeling and inference. We place special emphasis on the fundamental principles of flow design, and discuss foundational topics such as expressive power and computational trade-offs. We also broaden the conceptual framing of flows by relating them to more general probability transformations. Lastly, we summarize the use of flows for tasks such as generative modeling, approximate inference, and supervised learning."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Optimal transport mapping via input convex neural networks",
                    "abstract": "In this paper, we present a novel and principled approach to learn the optimal transport between two distributions, from samples. Guided by the optimal transport theory, we learn the optimal Kantorovich potential which induces the optimal transport map. This involves learning two convex functions, by solving a novel minimax optimization. Building upon recent advances in the field of input convex neural networks, we propose a new framework where the gradient of one convex function represents the optimal transport mapping. Numerical experiments confirm that we learn the optimal transport mapping. This approach ensures that the transport mapping we find is optimal independent of how we initialize the neural networks. Further, target distributions from a discontinuous support can be easily captured, as gradient of a convex function naturally models a {\\em discontinuous} transport mapping."
                },
                {
                    "title": "Unconstrained Monotonic Neural Networks",
                    "abstract": "Monotonic neural networks have recently been proposed as a way to define invertible transformations. These transformations can be combined into powerful autoregressive flows that have been shown to be universal approximators of continuous probability distributions. Architectures that ensure monotonicity typically enforce constraints on weights and activation functions, which enables invertibility but leads to a cap on the expressiveness of the resulting transformations. In this work, we propose the Unconstrained Monotonic Neural Network (UMNN) architecture based on the insight that a function is monotonic as long as its derivative is strictly positive. In particular, this latter condition can be enforced with a free-form neural network whose only constraint is the positiveness of its output. We evaluate our new invertible building block within a new autoregressive flow (UMNN-MAF) and demonstrate its effectiveness on density estimation experiments. We also illustrate the ability of UMNNs to improve variational inference."
                },
                {
                    "title": "Optuna: A Next-generation Hyperparameter Optimization Framework",
                    "abstract": "The purpose of this study is to introduce new design-criteria for next-generation hyperparameter optimization software. The criteria we propose include (1) define-by-run API that allows users to construct the parameter search space dynamically, (2) efficient implementation of both searching and pruning strategies, and (3) easy-to-setup, versatile architecture that can be deployed for various purposes, ranging from scalable distributed computing to light-weight experiment conducted via interactive interface. In order to prove our point, we will introduce Optuna, an optimization software which is a culmination of our effort in the development of a next generation optimization software. As an optimization software designed with define-by-run principle, Optuna is particularly the first of its kind. We will present the design-techniques that became necessary in the development of the software that meets the above criteria, and demonstrate the power of our new design through experimental results and real world applications. Our software is available under the MIT license (https://github.com/pfnet/optuna/)."
                },
                {
                    "title": "A geostatistical approach to estimate high resolution nocturnal bird migration densities from a weather radar network",
                    "abstract": "Quantifying nocturnal bird migration at high resolution is essential for (1) understanding the phenology of migration and its drivers, (2) identifying critical spatio-temporal protection zones for migratory birds, and (3) assessing the risk of collision with man-made structures. We propose a tailored geostatistical model to interpolate migration intensity monitored by a network of weather radars. The model is applied to data collected in autumn 2016 from 69 European weather radars. To cross-validate the model, we compared our results with independent measurements of two bird radars. Our model estimated bird densities at high resolution (0.2\u00b0latitude-longitude, 15min) and assessed the associated uncertainty. Within the area covered by the radar network, we estimated that around 120 million birds were simultaneously in flight [10-90 quantiles: 107-134]. Local estimations can be easily visualized and retrieved from a dedicated interactive website: birdmigrationmap.vogelwarte.ch. This proof-of-concept study demonstrates that a network of weather radar is able to quantify bird migration at high resolution and accuracy. The model presented has the ability to monitor population of migratory birds at scales ranging from regional to continental in space and daily to yearly in time. Near-real-time estimation should soon be possible with an update of the infrastructure and processing software."
                },
                {
                    "title": "Neural Spline Flows",
                    "abstract": "A normalizing flow models a complex probability density as an invertible transformation of a simple base density. Flows based on either coupling or autoregressive transforms both offer exact density evaluation and sampling, but rely on the parameterization of an easily invertible elementwise transformation, whose choice determines the flexibility of these models. Building upon recent work, we propose a fully-differentiable module based on monotonic rational-quadratic splines, which enhances the flexibility of both coupling and autoregressive transforms while retaining analytic invertibility. We demonstrate that neural spline flows improve density estimation, variational inference, and generative modeling of images."
                },
                {
                    "title": "Hamiltonian Neural Networks",
                    "abstract": "Even though neural networks enjoy widespread use, they still struggle to learn the basic laws of physics. How might we endow them with better inductive biases? In this paper, we draw inspiration from Hamiltonian mechanics to train models that learn and respect exact conservation laws in an unsupervised manner. We evaluate our models on problems where conservation of energy is important, including the two-body problem and pixel observations of a pendulum. Our model trains faster and generalizes better than a regular neural network. An interesting side effect is that our model is perfectly reversible in time."
                },
                {
                    "title": "Residual Flows for Invertible Generative Modeling",
                    "abstract": "Flow-based generative models parameterize probability distributions through an invertible transformation and can be trained by maximum likelihood. Invertible residual networks provide a flexible family of transformations where only Lipschitz conditions rather than strict architectural constraints are needed for enforcing invertibility. However, prior work trained invertible residual networks for density estimation by relying on biased log-density estimates whose bias increased with the network's expressiveness. We give a tractable unbiased estimate of the log density using a \"Russian roulette\" estimator, and reduce the memory required during training by using an alternative infinite series for the gradient. Furthermore, we improve invertible residual blocks by proposing the use of activation functions that avoid derivative saturation and generalizing the Lipschitz condition to induced mixed norms. The resulting approach, called Residual Flows, achieves state-of-the-art performance on density estimation amongst flow-based models, and outperforms networks that use coupling blocks at joint generative and discriminative modeling."
                },
                {
                    "title": "Semi-Conditional Normalizing Flows for Semi-Supervised Learning",
                    "abstract": "This paper proposes a semi-conditional normalizing flow model for semi-supervised learning. The model uses both labelled and unlabeled data to learn an explicit model of joint distribution over objects and labels. Semi-conditional architecture of the model allows us to efficiently compute a value and gradients of the marginal likelihood for unlabeled objects. The conditional part of the model is based on a proposed conditional coupling layer. We demonstrate performance of the model for semi-supervised classification problem on different datasets. The model outperforms the baseline approach based on variational auto-encoders on MNIST dataset."
                },
                {
                    "title": "Autoregressive Energy Machines",
                    "abstract": "Neural density estimators are flexible families of parametric models which have seen widespread use in unsupervised machine learning in recent years. Maximum-likelihood training typically dictates that these models be constrained to specify an explicit density. However, this limitation can be overcome by instead using a neural network to specify an energy function, or unnormalized density, which can subsequently be normalized to obtain a valid distribution. The challenge with this approach lies in accurately estimating the normalizing constant of the high-dimensional energy function. We propose the Autoregressive Energy Machine, an energy-based model which simultaneously learns an unnormalized density and computes an importance-sampling estimate of the normalizing constant for each conditional in an autoregressive decomposition. The Autoregressive Energy Machine achieves state-of-the-art performance on a suite of density-estimation tasks."
                },
                {
                    "title": "Measuring the Effects of Data Parallelism on Neural Network Training",
                    "abstract": "Recent hardware developments have dramatically increased the scale of data parallelism available for neural network training. Among the simplest ways to harness next-generation hardware is to increase the batch size in standard mini-batch neural network training algorithms. In this work, we aim to experimentally characterize the effects of increasing the batch size on training time, as measured by the number of steps necessary to reach a goal out-of-sample error. We study how this relationship varies with the training algorithm, model, and data set, and find extremely large variation between workloads. Along the way, we show that disagreements in the literature on how batch size affects model quality can largely be explained by differences in metaparameter tuning and compute budgets at different batch sizes. We find no evidence that larger batch sizes degrade out-of-sample performance. Finally, we discuss the implications of our results on efforts to train neural networks much faster in the future. Our experimental data is publicly available as a database of 71,638,836 loss measurements taken over the course of training for 168,160 individual models across 35 workloads."
                },
                {
                    "title": "FFJORD: Free-form Continuous Dynamics for Scalable Reversible Generative Models",
                    "abstract": "A promising class of generative models maps points from a simple distribution to a complex distribution through an invertible neural network. Likelihood-based training of these models requires restricting their architectures to allow cheap computation of Jacobian determinants. Alternatively, the Jacobian trace can be used if the transformation is specified by an ordinary differential equation. In this paper, we use Hutchinson's trace estimator to give a scalable unbiased estimate of the log-density. The result is a continuous-time invertible generative model with unbiased density estimation and one-pass sampling, while allowing unrestricted neural network architectures. We demonstrate our approach on high-dimensional density estimation, image generation, and variational inference, achieving the state-of-the-art among exact likelihood methods with efficient sampling."
                },
                {
                    "title": "Glow: Generative Flow with Invertible 1x1 Convolutions",
                    "abstract": "Flow-based generative models (Dinh et al., 2014) are conceptually attractive due to tractability of the exact log-likelihood, tractability of exact latent-variable inference, and parallelizability of both training and synthesis. In this paper we propose Glow, a simple type of generative flow using an invertible 1x1 convolution. Using our method we demonstrate a significant improvement in log-likelihood on standard benchmarks. Perhaps most strikingly, we demonstrate that a generative model optimized towards the plain log-likelihood objective is capable of efficient realistic-looking synthesis and manipulation of large images. The code for our model is available at this https URL"
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "Neural Autoregressive Flows",
                    "abstract": "Normalizing flows and autoregressive models have been successfully combined to produce state-of-the-art results in density estimation, via Masked Autoregressive Flows (MAF), and to accelerate state-of-the-art WaveNet-based speech synthesis to 20x faster than real-time, via Inverse Autoregressive Flows (IAF). We unify and generalize these approaches, replacing the (conditionally) affine univariate transformations of MAF/IAF with a more general class of invertible univariate transformations expressed as monotonic neural networks. We demonstrate that the proposed neural autoregressive flows (NAF) are universal approximators for continuous probability distributions, and their greater expressivity allows them to better capture multimodal target distributions. Experimentally, NAF yields state-of-the-art performance on a suite of density estimation tasks and outperforms IAF in variational autoencoders trained on binarized MNIST."
                },
                {
                    "title": "Masked Autoregressive Flow for Density Estimation",
                    "abstract": "Autoregressive models are among the best performing neural density estimators. We describe an approach for increasing the flexibility of an autoregressive model, based on modelling the random numbers that the model uses internally when generating data. By constructing a stack of autoregressive models, each modelling the random numbers of the next model in the stack, we obtain a type of normalizing flow suitable for density estimation, which we call Masked Autoregressive Flow. This type of flow is closely related to Inverse Autoregressive Flow and is a generalization of Real NVP. Masked Autoregressive Flow achieves state-of-the-art performance in a range of general-purpose density estimation tasks."
                },
                {
                    "title": "HyperNetworks",
                    "abstract": "This work explores hypernetworks: an approach of using one network, also known as a hypernetwork, to generate the weights for another network. We apply hypernetworks to generate adaptive weights for recurrent networks. In this case, hypernetworks can be viewed as a relaxed form of weight-sharing across layers. In our implementation, hypernetworks are are trained jointly with the main network in an end-to-end fashion. Our main result is that hypernetworks can generate non-shared weights for LSTM and achieve state-of-the-art results on a variety of sequence modelling tasks including character-level language modelling, handwriting generation and neural machine translation, challenging the weight-sharing paradigm for recurrent networks."
                },
                {
                    "title": "Sensitivity of the ECMWF Model to Semi-Lagrangian Departure Point Iterations",
                    "abstract": "AbstractAccurate estimation of the position of the departure points (d.p.) is crucial for the accuracy of a semi-Lagrangian NWP model. This calculation is often performed applying an implicit discretization to a kinematic equation solved by a fixed-point iteration scheme. A small number of iterations is typically used, assuming that this is sufficient for convergence. This assumption, derived from a past theoretical analysis, is revisited here. Analyzing the convergence of a generic d.p. iteration scheme and testing the ECMWF Integrated Forecast System (IFS) model, it is demonstrated that 2\u20133 iterations may not be sufficient for convergence to satisfactory accuracy in a modern high-resolution global model. Large forecast improvements can be seen by increasing the number of iterations. The extratropical geopotential error decreases and the simulated vertical structure of tropical cyclones improves. These findings prompted the implementation of an algorithm in which stopping criteria based on estimated conv..."
                },
                {
                    "title": "Displacement interpolation using Lagrangian mass transport",
                    "abstract": "Interpolation between pairs of values, typically vectors, is a fundamental operation in many computer graphics applications. In some cases simple linear interpolation yields meaningful results without requiring domain knowledge. However, interpolation between pairs of distributions or pairs of functions often demands more care because features may exhibit translational motion between exemplars. This property is not captured by linear interpolation. This paper develops the use of displacement interpolation for this class of problem, which provides a generic method for interpolating between distributions or functions based on advection instead of blending. The functions can be non-uniformly sampled, high-dimensional, and defined on non-Euclidean manifolds, e.g., spheres and tori. Our method decomposes distributions or functions into sums of radial basis functions (RBFs). We solve a mass transport problem to pair the RBFs and apply partial transport to obtain the interpolated function. We describe practical methods for computing the RBF decomposition and solving the transport problem. We demonstrate the interpolation approach on synthetic examples, BRDFs, color distributions, environment maps, stipple patterns, and value functions."
                },
                {
                    "title": "Bird migration flight altitudes studied by a network of operational weather radars",
                    "abstract": "A fully automated method for the detection and quantification of bird migration was developed for operational C-band weather radar, measuring bird density, speed and direction as a function of altitude. These weather radar bird observations have been validated with data from a high-accuracy dedicated bird radar, which was stationed in the measurement volume of weather radar sites in The Netherlands, Belgium and France for a full migration season during autumn 2007 and spring 2008. We show that weather radar can extract near real-time bird density altitude profiles that closely correspond to the density profiles measured by dedicated bird radar. Doppler weather radar can thus be used as a reliable sensor for quantifying bird densities aloft in an operational setting, which\u2014when extended to multiple radars\u2014enables the mapping and continuous monitoring of bird migration flyways. By applying the automated method to a network of weather radars, we observed how mesoscale variability in weather conditions structured the timing and altitude profile of bird migration within single nights. Bird density altitude profiles were observed that consisted of multiple layers, which could be explained from the distinct wind conditions at different take-off sites. Consistently lower bird densities are recorded in The Netherlands compared with sites in France and eastern Belgium, which reveals some of the spatial extent of the dominant Scandinavian flyway over continental Europe."
                },
                {
                    "title": "Monotone and Partially Monotone Neural Networks",
                    "abstract": "In many classification and prediction problems it is known that the response variable depends on certain explanatory variables. Monotone neural networks can be used as powerful tools to build monotone models with better accuracy and lower variance compared to ordinary nonmonotone models. Monotonicity is usually obtained by putting constraints on the parameters of the network. In this paper, we will clarify some of the theoretical results on monotone neural networks with positive weights, issues that are sometimes misunderstood in the neural network literature. Furthermore, we will generalize some of the results obtained by Sill for the so-called min-max networks to the case of partially monotone problems. The method is illustrated in practical case studies."
                },
                {
                    "title": "Semi-Lagrangian integration schemes for atmospheric models - A review",
                    "abstract": "Abstract The semi-Lagrangian methodology is described for a hierarchy of applications (passive advection, forced advection, and coupled sets of equations) of increasing complexity, in one, two, and three dimensions. Attention is focused on its accuracy, stability, and efficiency properties. Recent developments in applying semi-Lagrangian methods to 2D and 3D atmospheric flows in both Cartesian and spherical geometries are then reviewed. Finally, the current status of development is summarized, followed by a short discussion of future perspectives."
                },
                {
                    "title": "Sensitivity theory for nonlinear systems. I. Nonlinear functional analysis approach",
                    "abstract": "Concepts of nonlinear functional analysis are employed to investigate the mathematical foundations underlying sensitivity theory. This makes it possible not only to ascertain the limitations inherent in existing analytical approaches to sensitivity analysis, but also to rigorously formulate a considerably more general sensitivity theory for physical problems characterized by systems of nonlinear equations and by nonlinear functionals as responses. Two alternative formalisms, labeled the \u2019\u2019forward sensitivity formalism\u2019\u2019 and the \u2019\u2019adjoint sensitivity formalism,\u2019\u2019 are developed in order to evaluate the sensitivity of the response to variations in the system parameters. The forward sensitivity formalism is formulated in normed linear spaces, and the existence of the G\u00e2teaux differentials of the operators appearing in the problem is shown to be both necessary and sufficient for its validity. This formalism is conceptually straightforward and can be advantageously used to assess the effects of relatively few p..."
                },
                {
                    "title": "Sensitivity theory for nonlinear systems. II. Extensions to additional classes of responses",
                    "abstract": "This work extends a recent, functional\u2010analytic formulation of sensitivity theory to include treatment of additional types of responses. There are physical systems where a critical point of a function that depends on the system\u2019s state vector and parameters defines the location in phase\u2010space where the response functional is evaluated. The G\u00e2teaux differentials giving the sensitivities of both the functional and the critical point to changes in the system\u2019s parameters are obtained by alternative formalisms. The foward sensitivity formalism is the simpler and more general, but may be prohibitively expensive for problems with large data bases. The adjoint sensitivity formalism, although less generally applicable and requiring several adjoint calculations, is likely to be the only practical approach. Sensitivity theory is also extended to include treatment of general operators, acting on the system\u2019s state vector and parameters, as response. In this case, the forward sensitivity formalism is the same as for ..."
                },
                {
                    "title": "Physics-Informed Machine",
                    "abstract": "Traditional lithium-ion (Li-ion) battery state of health (SOH) estimation methodologies that focused on estimating present cell capacity do not provide sufficient information to determine the cell\u2019s lifecycle stage or value in second-life use. Quantifying the underlying degradation modes that cause capacity fade can give further insight into the electrochemical state of the cell and provide more detailed health information such as the remaining active materials and lithium inventory. However, current physics-based methods for degradation diagnostics require long-term cycling data and are computationally expensive to deploy locally on a device. To improve upon current methods, we propose and extensively test two light-weight physics-informed machine learning methods for online estimating the capacity of a battery cell and diagnosing its primary degradation modes using only limited early-life experimental degradation data. To enable late-life prediction (e.g. > 1.5 years) without the use of late-life experimental data, each of the methods is trained using simulation data from a physics-based half-cell model and early-life (e.g. < 3 months) degradation data obtained from cycling tests. The proposed methods are comprehensively evaluated using data from a long-term (3.5 years) cycling experiment of 16 implantable-grade Li-ion cells cycled under two temperatures and C-rates. Results from a four-fold cross-validation study show that the proposed physics-informed machine learning models are capable of improving the estimation accuracy of cell capacity and the state of three primary degradation modes by over 50% compared to a purely data-driven approach. Additionally, this work provides insights into the role of temperature and C-rate in cell degradation"
                },
                {
                    "title": "POT: Python Optimal Transport",
                    "abstract": "Optimal transport has recently been reintroduced to the machine learning community thanks in part to novel e\ufb03cient optimization procedures allowing for medium to large scale applications. We propose a Python toolbox that implements several key optimal transport ideas for the machine learning community. The toolbox contains implementations of a number of founding works of OT for machine learning such as Sinkhorn algorithm and Wasserstein barycenters, but also provides generic solvers that can be used for conducting novel fundamental research. This toolbox, named POT for Python Optimal Transport, is open source with an MIT license."
                },
                {
                    "title": "Searching for Activation Functions",
                    "abstract": "The choice of activation functions in deep networks has a significant effect on the training dynamics and task performance. Currently, the most successful and widely-used activation function is the Rectified Linear Unit (ReLU). Although various hand-designed alternatives to ReLU have been proposed, none have managed to replace it due to inconsistent gains. In this work, we propose to leverage automatic search techniques to discover new activation functions. Using a combination of exhaustive and reinforcement learning-based search, we discover multiple novel activation functions. We verify the effectiveness of the searches by conducting an empirical evaluation with the best discovered activation function. Our experiments show that the best discovered activation function, $f(x) = x \\cdot \\text{sigmoid}(\\beta x)$, which we name Swish, tends to work better than ReLU on deeper models across a number of challenging datasets. For example, simply replacing ReLUs with Swish units improves top-1 classification accuracy on ImageNet by 0.9\\% for Mobile NASNet-A and 0.6\\% for Inception-ResNet-v2. The simplicity of Swish and its similarity to ReLU make it easy for practitioners to replace ReLUs with Swish units in any neural network."
                },
                {
                    "title": "Under review as a conference paper at ICLR 2016",
                    "abstract": "We describe a method for learning word embeddings with stochastic dimensionality. Our Infinite Skip-Gram (iSG) model specifies an energy-based joint distribution over a word vector, a context vector, and their dimensionality. By employing the same techniques used to make the Infinite Restricted Boltzmann Machine (C\u00f4t\u00e9 & Larochelle, 2015) tractable, we define vector dimensionality over a countably infinite domain, allowing vectors to grow as needed during training. After training, we find that the distribution over embedding dimensionality for a given word is highly interpretable and leads to an elegant probabilistic mechanism for word sense induction. We show qualitatively and quantitatively that the iSG produces parameter-efficient representations that are robust to language\u2019s inherent ambiguity."
                },
                {
                    "title": "A Semi-Lagrangian and Semi-Implicit Numerical Integration Scheme for the Primitive Meteorological Equations",
                    "abstract": "A semi-Lagrangian algorithm is associated with the semi-implicit method in the integration of the shallow water equations on a rotating sphere. The resulting model is unconditionally stable and can be integrated with rather large time steps. Truncation errors remain reasonably small with time steps 25 times as large as those used with explicit integration schemes. An analysis of the proposed method is performed and it indicates that the scheme is stable. Also, the results of a few integrations are presented and from these we conclude that the model is not very sensitive to the size of the time step provided that it does not exceed a value of the order of two or three hours."
                }
            ],
            "categories": [
                "cs.LG",
                "physics.data-an",
                "physics.flu-dyn",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we infer continuous density and velocity fields from sparse and noisy data while ensuring compliance with the continuity equation in hydrodynamic flow problems?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of physics-informed machine learning (PI-ML) by integrating physical laws into machine learning models. It has broader implications for various applications, such as environmental monitoring, wildlife tracking, and fluid dynamics, where accurate modeling of flow and density is essential. Addressing this question could lead to significant advancements in our understanding of complex systems and improve practical applications in areas like radar ornithology and optimal transport problems, ultimately enhancing predictive capabilities and decision-making processes.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to infer continuous fields from limited and noisy observations, which can lead to significant uncertainty in the results. Naive approaches may fail because they do not incorporate the physical constraints imposed by the continuity equation, potentially resulting in physically unrealistic solutions. Additionally, the complexities of dealing with unknown boundary and initial conditions, as well as the need for the model to generalize from sparse data, present substantial technical and theoretical obstacles that must be addressed.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either fully known initial and boundary conditions or has not adequately integrated physical constraints into machine learning models. The limitations of existing solutions include a lack of robustness in handling sparse and noisy data and insufficient methods for ensuring compliance with physical laws. Our approach differs by explicitly incorporating the continuity equation into the neural network architecture, allowing for the generation of physically consistent density and velocity fields even in the absence of complete information.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a neural network model that inherently satisfies the continuity equation while inferring density and velocity fields from sparse and noisy measurements. We will utilize datasets from radar ornithology and dynamical optimal transport problems, focusing on metrics such as prediction accuracy and physical consistency. The expected outcomes include the successful estimation of spatiotemporal density fields and velocity profiles that adhere to the continuity equation, demonstrating the effectiveness of our approach in real-world applications."
            }
        },
        "author_data": {
            "c05c636b-ecf3-4b7f-9e9e-9ddf4f2db375": {
                "pk": "c05c636b-ecf3-4b7f-9e9e-9ddf4f2db375",
                "name": "F. Arend Torres",
                "collaborators": [
                    "Volker Roth",
                    "M. Negri",
                    "Monika Nagy-Huber",
                    "M. Samarin",
                    "V. Nesterov"
                ],
                "domain": [
                    "Gaussian Graphical Models",
                    "Physics-informed Neural Networks",
                    "Machine Learning",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Conditional Matrix Flows for Gaussian Graphical Models",
                        "abstract": "Studying conditional independence among many variables with few observations is a challenging task. Gaussian Graphical Models (GGMs) tackle this problem by encouraging sparsity in the precision matrix through $l_q$ regularization with $q\\leq1$. However, most GMMs rely on the $l_1$ norm because the objective is highly non-convex for sub-$l_1$ pseudo-norms. In the frequentist formulation, the $l_1$ norm relaxation provides the solution path as a function of the shrinkage parameter $\\lambda$. In the Bayesian formulation, sparsity is instead encouraged through a Laplace prior, but posterior inference for different $\\lambda$ requires repeated runs of expensive Gibbs samplers. Here we propose a general framework for variational inference with matrix-variate Normalizing Flow in GGMs, which unifies the benefits of frequentist and Bayesian frameworks. As a key improvement on previous work, we train with one flow a continuum of sparse regression models jointly for all regularization parameters $\\lambda$ and all $l_q$ norms, including non-convex sub-$l_1$ pseudo-norms. Within one model we thus have access to (i) the evolution of the posterior for any $\\lambda$ and any $l_q$ (pseudo-) norm, (ii) the marginal log-likelihood for model selection, and (iii) the frequentist solution paths through simulated annealing in the MAP limit."
                    },
                    {
                        "title": "Truly Mesh-free Physics-Informed Neural Networks",
                        "abstract": "Physics-informed Neural Networks (PINNs) have recently emerged as a principled way to include prior physical knowledge in form of partial differential equations (PDEs) into neural networks. Although generally viewed as being mesh-free, current approaches still rely on collocation points obtained within a bounded region, even in settings with spatially sparse signals. Furthermore, if the boundaries are not known, the selection of such a region may be arbitrary, resulting in a large proportion of collocation points being selected in areas of low relevance. To resolve this, we present a mesh-free and adaptive approach termed particle-density PINN (pdPINN), which is inspired by the microscopic viewpoint of \ufb02uid dynamics. Instead of sampling from a bounded region, we propose to sample directly from the distribution over the (\ufb02uids) particle positions, eliminating the need to introduce boundaries while adaptively focusing on the most relevant regions. This is achieved by reformulating the modeled \ufb02uid density as an unnormalized probability distribution from which we sample with dynamic Monte Carlo methods. We further generalize pdPINNs to different settings that allow interpreting a positive scalar quantity as a particle density, such as the evolution of the temperature in the heat equation. The utility of our approach is demonstrated on experiments for modeling (non-steady) compressible \ufb02uids in up to three dimensions and a two-dimensional diffusion problem, illustrating the high \ufb02exibility and sample ef\ufb01ciency compared to existing re\ufb01nement methods for PINNs."
                    },
                    {
                        "title": "Mesh-free Eulerian Physics-Informed Neural Networks",
                        "abstract": "Physics-informed Neural Networks (PINNs) have recently emerged as a principled way to include prior physical knowledge in form of partial differential equations (PDEs) into neural networks. Although PINNs are generally viewed as mesh-free, current approaches still rely on collocation points within a bounded region, even in settings with spatially sparse signals. Furthermore, if the boundaries are not known, the selection of such a region is dif\ufb01cult and often results in a large proportion of collocation points being selected in areas of low relevance. To resolve this severe drawback of current methods, we present a mesh-free and adaptive approach termed particle-density PINN (pdPINN), which is inspired by the microscopic viewpoint of \ufb02uid dynamics. The method is based on the Eulerian formulation and, different from classical mesh-free method, does not require the introduction of Lagrangian updates. We propose to sample directly from the distribution over the particle positions, eliminating the need to introduce boundaries while adaptively focusing on the most relevant regions. This is achieved by interpreting a non-negative physical quantity (such as the density or temperature) as an unnormalized probability distribution from which we sample with dynamic Monte Carlo methods. The proposed method leads to higher sample ef\ufb01ciency and improved performance of PINNs. These advantages are demonstrated on various experiments based on the continuity equations, Fokker-Planck equations, and the heat equation."
                    },
                    {
                        "title": "Learning Invariances with Generalised Input-Convex Neural Networks",
                        "abstract": "Considering smooth mappings from input vectors to continuous targets, our goal is to characterise subspaces of the input domain, which are invariant under such mappings. Thus, we want to characterise manifolds implicitly defined by level sets. Specifically, this characterisation should be of a global parametric form, which is especially useful for different informed data exploration tasks, such as building grid-based approximations, sampling points along the level curves, or finding trajectories on the manifold. However, global parameterisations can only exist if the level sets are connected. For this purpose, we introduce a novel and flexible class of neural networks that generalise input-convex networks. These networks represent functions that are guaranteed to have connected level sets forming smooth manifolds on the input space. We further show that global parameterisations of these level sets can be always found efficiently. Lastly, we demonstrate that our novel technique for characterising invariances is a powerful generative data exploration tool in real-world applications, such as computational chemistry."
                    }
                ]
            },
            "a968f706-bf60-4ca9-bc25-628dd93fab1f": {
                "pk": "a968f706-bf60-4ca9-bc25-628dd93fab1f",
                "name": "Marcello Massimo Negri",
                "collaborators": [
                    "Volker Roth",
                    "F. A. Torres",
                    "Monika Nagy-Huber",
                    "M. Samarin",
                    "Jonathan Aellen",
                    "Vincent Fortuin",
                    "Jan St\u00fchmer"
                ],
                "domain": [
                    "Normalizing Flows",
                    "Variational Inference",
                    "Physics-informed Neural Networks",
                    "Meta-learning"
                ],
                "publications": [
                    {
                        "title": "Injective flows for star-like manifolds",
                        "abstract": "Normalizing Flows (NFs) are powerful and efficient models for density estimation. When modeling densities on manifolds, NFs can be generalized to injective flows but the Jacobian determinant becomes computationally prohibitive. Current approaches either consider bounds on the log-likelihood or rely on some approximations of the Jacobian determinant. In contrast, we propose injective flows for star-like manifolds and show that for such manifolds we can compute the Jacobian determinant exactly and efficiently, with the same cost as NFs. This aspect is particularly relevant for variational inference settings, where no samples are available and only some unnormalized target is known. Among many, we showcase the relevance of modeling densities on star-like manifolds in two settings. Firstly, we introduce a novel Objective Bayesian approach for penalized likelihood models by interpreting level-sets of the penalty as star-like manifolds. Secondly, we consider probabilistic mixing models and introduce a general method for variational inference by defining the posterior of mixture weights on the probability simplex."
                    },
                    {
                        "title": "Conditional Matrix Flows for Gaussian Graphical Models",
                        "abstract": "Studying conditional independence among many variables with few observations is a challenging task. Gaussian Graphical Models (GGMs) tackle this problem by encouraging sparsity in the precision matrix through $l_q$ regularization with $q\\leq1$. However, most GMMs rely on the $l_1$ norm because the objective is highly non-convex for sub-$l_1$ pseudo-norms. In the frequentist formulation, the $l_1$ norm relaxation provides the solution path as a function of the shrinkage parameter $\\lambda$. In the Bayesian formulation, sparsity is instead encouraged through a Laplace prior, but posterior inference for different $\\lambda$ requires repeated runs of expensive Gibbs samplers. Here we propose a general framework for variational inference with matrix-variate Normalizing Flow in GGMs, which unifies the benefits of frequentist and Bayesian frameworks. As a key improvement on previous work, we train with one flow a continuum of sparse regression models jointly for all regularization parameters $\\lambda$ and all $l_q$ norms, including non-convex sub-$l_1$ pseudo-norms. Within one model we thus have access to (i) the evolution of the posterior for any $\\lambda$ and any $l_q$ (pseudo-) norm, (ii) the marginal log-likelihood for model selection, and (iii) the frequentist solution paths through simulated annealing in the MAP limit."
                    },
                    {
                        "title": "Truly Mesh-free Physics-Informed Neural Networks",
                        "abstract": "Physics-informed Neural Networks (PINNs) have recently emerged as a principled way to include prior physical knowledge in form of partial differential equations (PDEs) into neural networks. Although generally viewed as being mesh-free, current approaches still rely on collocation points obtained within a bounded region, even in settings with spatially sparse signals. Furthermore, if the boundaries are not known, the selection of such a region may be arbitrary, resulting in a large proportion of collocation points being selected in areas of low relevance. To resolve this, we present a mesh-free and adaptive approach termed particle-density PINN (pdPINN), which is inspired by the microscopic viewpoint of \ufb02uid dynamics. Instead of sampling from a bounded region, we propose to sample directly from the distribution over the (\ufb02uids) particle positions, eliminating the need to introduce boundaries while adaptively focusing on the most relevant regions. This is achieved by reformulating the modeled \ufb02uid density as an unnormalized probability distribution from which we sample with dynamic Monte Carlo methods. We further generalize pdPINNs to different settings that allow interpreting a positive scalar quantity as a particle density, such as the evolution of the temperature in the heat equation. The utility of our approach is demonstrated on experiments for modeling (non-steady) compressible \ufb02uids in up to three dimensions and a two-dimensional diffusion problem, illustrating the high \ufb02exibility and sample ef\ufb01ciency compared to existing re\ufb01nement methods for PINNs."
                    },
                    {
                        "title": "Meta-learning richer priors for VAEs",
                        "abstract": "Variational auto-encoders have proven to capture complicated data distributions and useful latent representations, while advances in meta-learning have made it possible to extract prior knowledge from data. We incorporate these two approaches and propose a novel flexible prior, namely the Pseudo-inputs prior, to obtain a richer latent space. We train VAEs using the Model-Agnostic Meta-Learning (MAML) algorithm and show that it achieves comparable reconstruction performance with standard training. However, we show that this MAML-VAE model learns richer latent representations, which we evaluate in terms of unsupervised few-shot classification as a downstream task. Moreover, we show that our proposed Pseudo-inputs prior outperforms baseline priors, including the VampPrior, in both models, while also encouraging high-level representations through its pseudo-inputs."
                    },
                    {
                        "title": "Mesh-free Eulerian Physics-Informed Neural Networks",
                        "abstract": "Physics-informed Neural Networks (PINNs) have recently emerged as a principled way to include prior physical knowledge in form of partial differential equations (PDEs) into neural networks. Although PINNs are generally viewed as mesh-free, current approaches still rely on collocation points within a bounded region, even in settings with spatially sparse signals. Furthermore, if the boundaries are not known, the selection of such a region is dif\ufb01cult and often results in a large proportion of collocation points being selected in areas of low relevance. To resolve this severe drawback of current methods, we present a mesh-free and adaptive approach termed particle-density PINN (pdPINN), which is inspired by the microscopic viewpoint of \ufb02uid dynamics. The method is based on the Eulerian formulation and, different from classical mesh-free method, does not require the introduction of Lagrangian updates. We propose to sample directly from the distribution over the particle positions, eliminating the need to introduce boundaries while adaptively focusing on the most relevant regions. This is achieved by interpreting a non-negative physical quantity (such as the density or temperature) as an unnormalized probability distribution from which we sample with dynamic Monte Carlo methods. The proposed method leads to higher sample ef\ufb01ciency and improved performance of PINNs. These advantages are demonstrated on various experiments based on the continuity equations, Fokker-Planck equations, and the heat equation."
                    }
                ]
            },
            "8e73f19f-bb9c-4f99-ac4c-930225f64567": {
                "pk": "8e73f19f-bb9c-4f99-ac4c-930225f64567",
                "name": "Marco Inversi",
                "collaborators": [
                    "Giorgio Stefani",
                    "P. Porta",
                    "G. Fierro",
                    "Gianluca Crippa",
                    "C. Saffirio",
                    "D. Zaccaria",
                    "N. Lamaddalena",
                    "I. Pettiti",
                    "D. Gazzoli",
                    "M. Valigi",
                    "S. Rossi",
                    "G. Ferraris",
                    "S. Colonna",
                    "I. Botto",
                    "M. Vassallo",
                    "D. Cordischi",
                    "G. Minelli",
                    "M. Lojacono",
                    "R. Lavecchia",
                    "F. Cioci"
                ],
                "domain": [
                    "Mathematical Analysis",
                    "Non-local Equations",
                    "Catalysis",
                    "X-ray Spectroscopy"
                ],
                "publications": [
                    {
                        "title": "Lagrangian stability for a system of non-local continuity equations under Osgood condition",
                        "abstract": ". We extend known existence and uniqueness results of weak measure solutions for systems of non-local continuity equations beyond the usual Lipschitz regularity. Existence of weak measure solutions holds for uniformly continuous vector \ufb01elds and convolution kernels, while uniqueness follows from a Lagrangian stability estimate under an additional Osgood condition."
                    },
                    {
                        "title": "Existence and stability of weak solutions of the Vlasov--Poisson system in localized Yudovich spaces",
                        "abstract": "We consider the Vlasov--Poisson system both in the repulsive (electrostatic potential) and in the attractive (gravitational potential) cases. In our first main theorem, we prove the uniqueness and the quantitative stability of Lagrangian solutions $f=f(t,x,v)$ whose associated spatial density $\\rho_f=\\rho_f(t,x)$ is potentially unbounded but belongs to suitable uniformly-localized Yudovich spaces. This requirement imposes a condition of slow growth on the function $p \\mapsto \\|\\rho_f(t,\\cdot)\\|_{L^p}$ uniformly in time. Previous works by Loeper, Miot and Holding--Miot have addressed the cases of bounded spatial density, i.e., $\\|\\rho_f(t,\\cdot)\\|_{L^p} \\lesssim 1$, and spatial density such that $\\|\\rho_f(t,\\cdot)\\|_{L^p} \\sim p^{1/\\alpha}$ for $\\alpha\\in[1,+\\infty)$. Our approach is Lagrangian and relies on an explicit estimate of the modulus of continuity of the electric field and on a second-order Osgood lemma. It also allows for iterated-logarithmic perturbations of the linear growth condition. In our second main theorem, we complement the aforementioned result by constructing solutions whose spatial density sharply satisfies such iterated-logarithmic growth. Our approach relies on real-variable techniques and extends the strategy developed for the Euler equations by the first and fourth-named authors. It also allows for the treatment of more general equations that share the same structure as the Vlasov--Poisson system. Notably, the uniqueness result and the stability estimates hold for both the classical and the relativistic Vlasov--Poisson systems."
                    },
                    {
                        "title": "Assessment of the environmental sustainability of irrigated agriculture in a large-scale scheme: a case study",
                        "abstract": "Article available on line / Article disponible en ligne \u00e0 l\u2019adresse : -------------------------------------------------------------------------------------------------------------------------------------------------------------------------http://om.ciheam.org/article.php?IDPDF=800227 --------------------------------------------------------------------------------------------------------------------------------------------------------------------------"
                    },
                    {
                        "title": "Cell for the in situ study of heterogeneous catalysts by transmission and fluorescence XAS spectroscopy",
                        "abstract": "A simple cell, assembled with commercial parts, suitable for in situ X-ray-absorption spectroscopy measurements of heterogeneous catalysts, has been designed. The cell, light and easy to handle, allows thermal treatments of the sample under investigation up to 823\u2005K in a reducing or oxidizing atmosphere and measurements at both high and liquid-nitrogen temperature. The cell was tested by studying the decomposition, in an oxygen flow, of ammonium metatungstate to WO3. Extended X-ray absorption fine-structure measurements of the ammonium metatungstate before and after a thermal treatment at 773\u2005K are reported."
                    },
                    {
                        "title": "Some aspects of \u03b2-V9Mo6O40 reduction: TPR, XRD, SEM, IR and EPR spectroscopic studies",
                        "abstract": "The \u03b2-V9Mo6O40 phase, containing only 1/9 of the V atoms as V4+ ions, has been prepared by the solid-state reaction of suitable amounts of MoO3 and V2O5 oxides heated to 923 K. The temperature-programmed reduction (TPR) analysis of the \u03b2-V9Mo6O40 phase has been carried out up to 1273 K. The TPR profiles were always characterised by two peaks regardless of the experimental conditions. At the end of the first reduction peak, occurring at ca. 900 K, the \u03b2-V9Mo6O40 phase disgregates by forming the MoO2 and V2O3 oxides. The second reduction peak appears at higher temperature and corresponds to the reduction of MoO2 to molybdenum metal. Both the as-prepared and reduced samples have been studied by X-ray powder diffraction (XRPD), IR and EPR spectroscopies as well as by scanning electron microscopy (SEM). The EPR analysis confirmed the presence of V4+ ions in the crystal lattice of the \u03b2-V9Mo6O40 phase which is characterized by a metal\u2013oxygen bonding different from those found for the pure MoO3 and V2O5 oxides, as evidenced by the IR spectra. Prismatic crystals were observed by SEM for the as-prepared sample. The original morphology was essentially preserved upon H2 reduction, but a widespread microroughness appeared at the surface of the prismatic crystals. At the beginning of the reduction process an increase of the V4+ species in the lattice was detected by EPR, giving rise to a gradual slight structural distortion. However, an oxidising treatment in air of the mildly reduced sample restores the original structure."
                    },
                    {
                        "title": "A Study of Anomalous Temperature-Programmed Reduction Profiles of Cu2O, CuO, and CuO-ZnO Catalysts",
                        "abstract": "A temperature-programmed reduction (TPR) study of CuO, Cu2O, and CuO-ZnO catalysts, in the atomic range Cu:Zn = 92:8 to 67:33, is reported. In order to promote a correct experimental approach, much attention has been devoted to the choice of the experimental operating variables. Direct evidence is presented that the reduction profiles of the investigated specimens are markedly affected by artefacts when TPR measurements are carried out under conditions of an inadequate combination of initial amount of reducible species, total flow rate, initial hydrogen concentration, and heating rate. It is shown that under correct experimental conditions the reduction profile of each sample is characterized by a single and quite sharp reduction peak. In contrast to this, a dramatic change of shape occurs when an improper combination of the experimental operating variables is imposed: for both CuO and Cu2O an apparent double peak appears, whereas for the CuO-ZnO catalysts the reduction profiles are transformed into broad bands. In all cases, the reliability of the TPR results is completely lost. To explain the complex origin of these artefacts, it is suggested that both a special dynamic situation and activated H2 adsorption-desorption phenomena coinciding with the chemical reduction are responsible for the observed perturbations. A simple kinetic analysis has been developed which enables a qualitative reproduction of the disturbed reduction profiles to be made."
                    }
                ]
            },
            "1f2a063c-f653-4b7c-9e72-47a59251b01c": {
                "pk": "1f2a063c-f653-4b7c-9e72-47a59251b01c",
                "name": "Jonathan Aellen",
                "collaborators": [
                    "M. Negri",
                    "Volker Roth",
                    "Dennis Madsen",
                    "Andreas Morel-Forster",
                    "T. Vetter",
                    "M. L\u00fcthi"
                ],
                "domain": [
                    "Normalizing Flows",
                    "Variational Inference",
                    "Gaussian Processes",
                    "Non-rigid Registration"
                ],
                "publications": [
                    {
                        "title": "Injective flows for star-like manifolds",
                        "abstract": "Normalizing Flows (NFs) are powerful and efficient models for density estimation. When modeling densities on manifolds, NFs can be generalized to injective flows but the Jacobian determinant becomes computationally prohibitive. Current approaches either consider bounds on the log-likelihood or rely on some approximations of the Jacobian determinant. In contrast, we propose injective flows for star-like manifolds and show that for such manifolds we can compute the Jacobian determinant exactly and efficiently, with the same cost as NFs. This aspect is particularly relevant for variational inference settings, where no samples are available and only some unnormalized target is known. Among many, we showcase the relevance of modeling densities on star-like manifolds in two settings. Firstly, we introduce a novel Objective Bayesian approach for penalized likelihood models by interpreting level-sets of the penalty as star-like manifolds. Secondly, we consider probabilistic mixing models and introduce a general method for variational inference by defining the posterior of mixture weights on the probability simplex."
                    },
                    {
                        "title": "GiNGR: Generalized Iterative Non-Rigid Point Cloud and Surface Registration Using Gaussian Process Regression",
                        "abstract": "In this paper, we unify popular non-rigid registration methods for point sets and surfaces under our general framework, GiNGR. GiNGR builds upon Gaussian Process Morphable Models (GPMM) and hence separates modeling the deformation prior from model adaptation for registration. In addition, it provides explainable hyperparameters, multi-resolution registration, trivial inclusion of expert annotation, and the ability to use and combine analytical and statistical deformation priors. But more importantly, the reformulation allows for a direct comparison of registration methods. Instead of using a general solver in the optimization step, we show how Gaussian process regression (GPR) iteratively can warp a reference onto a target, leading to smooth deformations following the prior for any dense, sparse, or partial estimated correspondences in a principled way. We show how the popular CPD and ICP algorithms can be directly explained with GiNGR. Furthermore, we show how existing algorithms in the GiNGR framework can perform probabilistic registration to obtain a distribution of different registrations instead of a single best registration. This can be used to analyze the uncertainty e.g. when registering partial observations. GiNGR is publicly available and fully modular to allow for domain-specific prior construction."
                    }
                ]
            },
            "24cccf3d-757b-4e6f-84b8-1a7487ac8027": {
                "pk": "24cccf3d-757b-4e6f-84b8-1a7487ac8027",
                "name": "Volker Roth",
                "collaborators": [
                    "M. Samarin",
                    "Monika Nagy-Huber",
                    "F. A. Torres",
                    "L. Zweifel",
                    "C. Alewell",
                    "M. Negri",
                    "David Belius",
                    "K. Meusburger",
                    "V. Nesterov"
                ],
                "domain": [
                    "Physics-informed Neural Networks",
                    "Deep Learning",
                    "Remote Sensing",
                    "Soil Erosion"
                ],
                "publications": [
                    {
                        "title": "Truly Mesh-free Physics-Informed Neural Networks",
                        "abstract": "Physics-informed Neural Networks (PINNs) have recently emerged as a principled way to include prior physical knowledge in form of partial differential equations (PDEs) into neural networks. Although generally viewed as being mesh-free, current approaches still rely on collocation points obtained within a bounded region, even in settings with spatially sparse signals. Furthermore, if the boundaries are not known, the selection of such a region may be arbitrary, resulting in a large proportion of collocation points being selected in areas of low relevance. To resolve this, we present a mesh-free and adaptive approach termed particle-density PINN (pdPINN), which is inspired by the microscopic viewpoint of \ufb02uid dynamics. Instead of sampling from a bounded region, we propose to sample directly from the distribution over the (\ufb02uids) particle positions, eliminating the need to introduce boundaries while adaptively focusing on the most relevant regions. This is achieved by reformulating the modeled \ufb02uid density as an unnormalized probability distribution from which we sample with dynamic Monte Carlo methods. We further generalize pdPINNs to different settings that allow interpreting a positive scalar quantity as a particle density, such as the evolution of the temperature in the heat equation. The utility of our approach is demonstrated on experiments for modeling (non-steady) compressible \ufb02uids in up to three dimensions and a two-dimensional diffusion problem, illustrating the high \ufb02exibility and sample ef\ufb01ciency compared to existing re\ufb01nement methods for PINNs."
                    },
                    {
                        "title": "Feature learning and random features in standard finite-width convolutional neural networks: An empirical study",
                        "abstract": "The Neural Tangent Kernel is an important mile-stone in the ongoing e\ufb00ort to build a theory for deep learning. Its prediction that su\ufb03ciently wide neural networks behave as kernel methods, or equivalently as random feature models arising from linearized networks, has been con\ufb01rmed empirically for cer-tain wide architectures. In this paper, we compare the performance of two common \ufb01nite-width convolutional neural networks, LeNet and AlexNet, to their linearizations on common benchmark datasets like MNIST and modi\ufb01ed versions of it, CIFAR-10 and an ImageNet subset. We demonstrate empirically that \ufb01nite-width neural networks, generally, greatly outperform the \ufb01nite-width linearization of these architectures. When increasing the problem di\ufb03culty of the classi\ufb01cation task, we observe a larger gap which is in line with common intuition that \ufb01nite-width neural networks perform feature learning which \ufb01nite-width linearizations cannot. At the same time, \ufb01nite-width linearizations im-prove dramatically with width, approaching the behavior of the wider standard networks which in turn perform slightly better than their standard width counterparts. Therefore, it appears that feature learning for non-wide standard networks is important but becomes less signi\ufb01cant with increasing width. We furthermore identify cases where both standard and linearized networks match in performance, in agreement with NTK theory, and a case where a wide linearization outperforms its standard width counterpart."
                    },
                    {
                        "title": "Mesh-free Eulerian Physics-Informed Neural Networks",
                        "abstract": "Physics-informed Neural Networks (PINNs) have recently emerged as a principled way to include prior physical knowledge in form of partial differential equations (PDEs) into neural networks. Although PINNs are generally viewed as mesh-free, current approaches still rely on collocation points within a bounded region, even in settings with spatially sparse signals. Furthermore, if the boundaries are not known, the selection of such a region is dif\ufb01cult and often results in a large proportion of collocation points being selected in areas of low relevance. To resolve this severe drawback of current methods, we present a mesh-free and adaptive approach termed particle-density PINN (pdPINN), which is inspired by the microscopic viewpoint of \ufb02uid dynamics. The method is based on the Eulerian formulation and, different from classical mesh-free method, does not require the introduction of Lagrangian updates. We propose to sample directly from the distribution over the particle positions, eliminating the need to introduce boundaries while adaptively focusing on the most relevant regions. This is achieved by interpreting a non-negative physical quantity (such as the density or temperature) as an unnormalized probability distribution from which we sample with dynamic Monte Carlo methods. The proposed method leads to higher sample ef\ufb01ciency and improved performance of PINNs. These advantages are demonstrated on various experiments based on the continuity equations, Fokker-Planck equations, and the heat equation."
                    },
                    {
                        "title": "Learning Invariances with Generalised Input-Convex Neural Networks",
                        "abstract": "Considering smooth mappings from input vectors to continuous targets, our goal is to characterise subspaces of the input domain, which are invariant under such mappings. Thus, we want to characterise manifolds implicitly defined by level sets. Specifically, this characterisation should be of a global parametric form, which is especially useful for different informed data exploration tasks, such as building grid-based approximations, sampling points along the level curves, or finding trajectories on the manifold. However, global parameterisations can only exist if the level sets are connected. For this purpose, we introduce a novel and flexible class of neural networks that generalise input-convex networks. These networks represent functions that are guaranteed to have connected level sets forming smooth manifolds on the input space. We further show that global parameterisations of these level sets can be always found efficiently. Lastly, we demonstrate that our novel technique for characterising invariances is a powerful generative data exploration tool in real-world applications, such as computational chemistry."
                    },
                    {
                        "title": "Visual Understanding in Semantic Segmentation of Soil Erosion Sites in Swiss Alpine Grasslands",
                        "abstract": "  <p>Understanding the occurrence of soil erosion phenomena is of vital importance for ecology and agriculture, especially under changing climate conditions. In Alpine grasslands, susceptibility to soil erosion is predominately due to the prevailing geological, morphological and climate conditions but is also affected by anthropogenic aspects such as agricultural land use. Climate change is expected to have a relevant impact on the driving factors of soil erosion like strong precipitation events and altered snow dynamics. In order to assess spatial and temporal changes of soil erosion phenomena and investigate possible reasons for their occurrence, large-scale methods to identify different soil erosion sites and quantify their extent are desirable.</p><p>In the field of remote sensing, one such semi-automatic method for (semantic) image segmentation is Object-based Image Analysis (OBIA), which makes use of spectral and spatial properties of image objects. In a recent study (Zweifel et al.), we successfully employed OBIA on high-resolution orthoimages (RGB spectral bands, 0.25 to 0.5 m pixel resolution) and derivatives of digital elevation models (DEM) of a study site in the Swiss Alps (Urseren Valley). The method provides high-quality segmentation results and an increasing trend of total area affected by soil erosion (+156 +/- 18%) is shown over a period from 2000 to 2016. However, using OBIA requires expert knowledge, manual adjustments, and is time-intensive in order to achieve satisfying segmentation results. In addition, the parameter settings of the method cannot be easily transferred from one image to another.</p><p>To allow for large-scale semantic segmentation of erosion sites, we make use of fully convolutional neural networks (CNNs). In recent years, CNNs proved to be very performant tools for a variety of image recognition tasks. While training CNNs might be more time demanding, predicting segmentations for new images and previously unseen regions is usually fast. For this study, we train a U-Net with high-quality segmentation masks provided by OBIA and DEM derivatives. The U-Net segmentation results are not only in good agreement with the OBIA results, but also a similar trend for the increase of total area affected by soil erosion is observed.</p><p>In order to have a natural understanding of what in the input is &#8220;relevant&#8221; for the segmentation result, we make use of methods which highlight different regions of the input image, thereby providing a visually interpretable result. We use different approaches to identify these relevant regions which are based on perturbation of the input image and relevance propagation of the output signal to the input image. While the former approach identifies the relevant regions by modifying the input image and considering the changes in the output, the latter approach tracks the dominant signal from the segmentation output back to the input image, highlighting the relevant regions. Although both approaches attempt to attain the same goal, differences in the relevant regions of the input images for the segmentation results can be observed.</p><p><span>Zweifel, L., Meusburger, K., and Alewell, C. Spatio-temporal pattern of soil degradation in a Swiss Alpine grassland catchment. Remote Sensing of Environment, 235, 2019.</span></p> "
                    },
                    {
                        "title": "Identification of Soil Erosion in Alpine Grasslands on High-Resolution Aerial Images: Switching from Object-based Image Analysis to Deep Learning?",
                        "abstract": "  <p>Soil erosion in Alpine grassland areas is an ecological threat caused by the extreme topography, prevailing climate conditions and land-use practices but enhanced by climate change (e.g., heavy precipitation events, changing snow dynamics) in combination with changing land-use practices (e.g, more intensely used pastures). To increase our understanding of ongoing soil erosion processes in Alpine grasslands, there is a need to acquire detailed information on spatial extension and temporal trends.</p><p>In the past, we have successfully applied a semi-automatic method using an object-based image analysis (OBIA) framework with high-resolution aerial images (0.25-0.5m) and a digital terrain model (2m) to map erosion features in the Central Swiss Alps (Urseren Valley, Canton Uri, Switzerland). Degraded sites are classified according to the major erosion process (shallow landslides; sites with reduced vegetation cover affected by sheet erosion) or triggering factors (trampling by livestock; management effects) (Zweifel et al. 2019). We now aim to apply a deep learning (DL) model with the purpose of fast and efficient spatial upscaling(e.g., alpine-wide analysis). While OBIA yields high quality results, there are multiple constraints, such as labor-intensive steps and the requirement of expert knowledge, which make the method unsuitable for larger scale applications. The results of OBIA are used as a training dataset for our DL model. The DL approach uses fully-convolutional networks with the U-Net architecture and is capable of rapid segmentation and classification to identify areas with reduced vegetation cover and bare soil sites.</p><p>Results for the Urseren Valley (Canton Uri, Switzerland) show an increase in total area affected by soil degradation of 156 &#177;18% during a 16-year observation period (2000-2016). A comparison of the two methods (OBIA and DL) shows that DL results for the Urseren Valley follow similar trends for the 16-year period and that the segmentations of eroded sites are in good agreement (IoU = 0.83). First transferability tests to other valleys not considered during training of the DL model are very promising, confirming that DL is a well-suited and efficient method for future projects to map and assess soil erosion processes in grassland areas at regional scales.</p><p>&#160;</p><p><strong>References</strong></p><p>L. Zweifel, K. Meusburger, and C. Alewell. Spatio-temporal pattern of soil degradation in a Swiss Alpine grassland catchment. Remote Sensing of Environment, 235, 2019.</p> "
                    },
                    {
                        "title": "On the Empirical Neural Tangent Kernel of Standard Finite-Width Convolutional Neural Network Architectures",
                        "abstract": "The Neural Tangent Kernel (NTK) is an important milestone in the ongoing effort to build a theory for deep learning. Its prediction that sufficiently wide neural networks behave as kernel methods, or equivalently as random feature models, has been confirmed empirically for certain wide architectures. It remains an open question how well NTK theory models standard neural network architectures of widths common in practice, trained on complex datasets such as ImageNet. We study this question empirically for two well-known convolutional neural network architectures, namely AlexNet and LeNet, and find that their behavior deviates significantly from their finite-width NTK counterparts. For wider versions of these networks, where the number of channels and widths of fully-connected layers are increased, the deviation decreases."
                    },
                    {
                        "title": "Identifying Soil Erosion Processes in Alpine Grasslands on Aerial Imagery with a U-Net Convolutional Neural Network",
                        "abstract": "Erosion in alpine grasslands is a major threat to ecosystem services of alpine soils. Natural causes for the occurrence of soil erosion are steep topography and prevailing climate conditions in combination with soil fragility. To increase our understanding of ongoing erosion processes and support sustainable land-use management, there is a need to acquire detailed information on spatial occurrence and temporal trends. Existing approaches to identify these trends are typically laborious, have lack of transferability to other regions, and are consequently only applicable to smaller regions. In order to overcome these limitations and create a sophisticated erosion monitoring tool capable of large-scale analysis, we developed a model based on U-Net, a fully convolutional neural network, to map different erosion processes on high-resolution aerial images (RGB, 0.25\u20130.5 m). U-Net was trained on a high-quality data set consisting of labeled erosion sites mapped with object-based image analysis (OBIA) for the Urseren Valley (Central Swiss Alps) for five aerial images (16 year period). We used the U-Net model to map the same study area and conduct quality assessments based on a held-out test region and a temporal transferability test on new images. Erosion classes are assigned according to their type (shallow landslide and sites with reduced vegetation affected by sheet erosion) or land-use impacts (livestock trails and larger management affected areas). We show that results obtained by OBIA and U-Net follow similar linear trends for the 16 year study period, exhibiting increases in total degraded area of 167% and 201%, respectively. Segmentations of eroded sites are generally in good agreement, but also display method-specific differences, which lead to an overall precision of 73%, a recall of 84%, and a F1-score of 78%. Our results show that U-Net is transferable to spatially (within our study area) and temporally unseen data (data from new years) and is therefore a method suitable to efficiently and successfully capture the temporal trends and spatial heterogeneity of degradation in alpine grasslands. Additionally, U-Net is a powerful and robust tool to map erosion sites in a predictive manner utilising large amounts of new aerial imagery."
                    },
                    {
                        "title": "Object Segmentation for Identification of Soil Erosion Sites in Swiss Alpine Grasslands with Neural Networks",
                        "abstract": "Soil erosion is a significant environmental issue which is of severe ecological impact and has a relevant influence on agricultural land usage. Susceptibility to soil erosion is mainly due to geological, morphological or anthropogenic aspects and often caused by strong precipitation events, avalanches, and snow gliding. With a changing climate, altering land use and their effects on the driving factors of soil erosion, an increase of soil erosion sites in the alpine region is expected."
                    }
                ]
            }
        }
    },
    "2402.04647": {
        "paper_data": {
            "title": "Latent Plan Transformer: Planning as Latent Variable Inference",
            "url": "http://arxiv.org/abs/2402.04647v2",
            "arxiv_id": "2402.04647",
            "authors": [
                "Deqian Kong",
                "Dehong Xu",
                "Minglu Zhao",
                "Bo Pang",
                "Jianwen Xie",
                "Andrew Lizarraga",
                "Yuhao Huang",
                "Sirui Xie",
                "Ying Nian Wu"
            ],
            "abstract": "In tasks aiming for long-term returns, planning becomes essential. We study generative modeling for planning with datasets repurposed from offline reinforcement learning. Specifically, we identify temporal consistency in the absence of step-wise rewards as one key technical challenge. We introduce the Latent Plan Transformer (LPT), a novel model that leverages a latent space to connect a Transformer-based trajectory generator and the final return. LPT can be learned with maximum likelihood estimation on trajectory-return pairs. In learning, posterior sampling of the latent variable naturally integrates sub-trajectories to form a consistent abstraction despite the finite context. At test time, the latent variable is inferred from an expected return before policy execution, realizing the idea of planning as inference. Our experiments demonstrate that LPT can discover improved decisions from suboptimal trajectories, achieving competitive performance across several benchmarks, including Gym-Mujoco, Franka Kitchen, Maze2D, and Connect Four. It exhibits capabilities in nuanced credit assignments, trajectory stitching, and adaptation to environmental contingencies. These results validate that latent variable inference can be a strong alternative to step-wise reward prompting.",
            "introduction": "   1 Introduction  Decision Transformer (DT)\u00a0(Chen et\u00a0al., 2021) and some concurrent work\u00a0(Janner et\u00a0al., 2021) have popularized the research agenda of decision-making via generative modeling. The general idea is to consider decision-making as a generative process that takes in a representation of the task objective, e.g. the rewards or returns of a trajectory, and outputs a representation of the trajectory. Intuitively, a purposeful decision-making process should shift the trajectory distribution towards regimes with higher returns. In the classical decision-making literature, this is achieved by two interweaving processes, policy evaluation and policy improvement (Sutton and Barto, 2018). Policy evaluation promotes consistency in the estimated correlations between the trajectories and the returns. In DT, this is realized by the maximum likelihood estimation (MLE) of the joint distribution of sequences consisting of states, actions, and return-to-gos (RTG). Policy improvement shifts the distribution to improve the status quo expectation of the returns. In DT, this is naturally entailed since the policy is a distribution of actions conditioned on step-wise RTGs.   In this work, we are interested in the problem of planning. Among various ways to identify planning as a special class of decision-making problems, we pay particular attention to its data specification and inductive biases. As designing step-wise rewards requires significant effort and domain expertise, we focus on the problem of learning from trajectory-return pairs, where a trajectory is a sequence of states and actions, and the return is its total rewards. This design choice forces the agents to predict into the long-term future and figure out step-wise credits by themselves. A competitive Temporal Difference (TD) learning baseline, CQL\u00a0(Kumar et\u00a0al., 2020), was reported to be fragile under this data specification\u00a0(Chen et\u00a0al., 2021).   Our design of inductive biases reflects our intuition of a plan. While a policy is a factor of the trajectory distribution, a plan is an abstraction lifted from the space of trajectories. As a plan is always made in advance of receiving returns, it implies significance, persistence, and contingency. An agent should plan for more significant returns. It should be persistent in its plan even if the return is assigned in hindsight. It should also be adaptable to the environment\u2019s changes during the execution of the plan. We formulate this hierarchy of decision-making with a top-down latent variable model. The latent variable we introduce is effectively a plan, for it decouples the trajectory generation from the expected improvement of returns. The autoregressive policy always consults this temporally extended latent variable to be persistent in the plan. The top-down structure enables the agent to disentangle the variations in its plan from the environment\u2019s contingencies.   In this work, we introduce the Latent Plan Transformer\u00a0(LPT), a novel generative model featuring a latent vector modeled by a neural transformation of Gaussian white noise, a Transformer-based policy conditioned on this latent vector and a return estimation model. LPT is learned by maximum likelihood estimation (MLE). Given an expected return, posterior inference of the latent vector in LPT is an explicit process for iterative refinement of the plan. The inferred latent variable replaces RTG in the conditioning of the auto-regressive policy, providing richer information about the",
            "references": [
                {
                    "title": "Survival Instinct in Offline Reinforcement Learning",
                    "abstract": "We present a novel observation about the behavior of offline reinforcement learning (RL) algorithms: on many benchmark datasets, offline RL can produce well-performing and safe policies even when trained with\"wrong\"reward labels, such as those that are zero everywhere or are negatives of the true rewards. This phenomenon cannot be easily explained by offline RL's return maximization objective. Moreover, it gives offline RL a degree of robustness that is uncharacteristic of its online RL counterparts, which are known to be sensitive to reward design. We demonstrate that this surprising robustness property is attributable to an interplay between the notion of pessimism in offline RL algorithms and a certain bias implicit in common data collection practices. As we prove in this work, pessimism endows the agent with a\"survival instinct\", i.e., an incentive to stay within the data support in the long term, while the limited and biased data coverage further constrains the set of survival policies. Formally, given a reward class -- which may not even contain the true reward -- we identify conditions on the training data distribution that enable offline RL to learn a near-optimal and safe policy from any reward within the class. We argue that the survival instinct should be taken into account when interpreting results from existing offline RL benchmarks and when creating future ones. Our empirical and theoretical results suggest a new paradigm for RL, whereby an agent is\"nudged\"to learn a desirable behavior with imperfect reward but purposely biased data coverage."
                },
                {
                    "title": "Dichotomy of Control: Separating What You Can Control from What You Cannot",
                    "abstract": "Future- or return-conditioned supervised learning is an emerging paradigm for offline reinforcement learning (RL), where the future outcome (i.e., return) associated with an observed action sequence is used as input to a policy trained to imitate those same actions. While return-conditioning is at the heart of popular algorithms such as decision transformer (DT), these methods tend to perform poorly in highly stochastic environments, where an occasional high return can arise from randomness in the environment rather than the actions themselves. Such situations can lead to a learned policy that is inconsistent with its conditioning inputs; i.e., using the policy to act in the environment, when conditioning on a specific desired return, leads to a distribution of real returns that is wildly different than desired. In this work, we propose the dichotomy of control (DoC), a future-conditioned supervised learning framework that separates mechanisms within a policy's control (actions) from those beyond a policy's control (environment stochasticity). We achieve this separation by conditioning the policy on a latent variable representation of the future, and designing a mutual information constraint that removes any information from the latent variable associated with randomness in the environment. Theoretically, we show that DoC yields policies that are consistent with their conditioning inputs, ensuring that conditioning a learned policy on a desired high-return future outcome will correctly induce high-return behavior. Empirically, we show that DoC is able to achieve significantly better performance than DT on environments that have highly stochastic rewards and transition"
                },
                {
                    "title": "Q-learning Decision Transformer: Leveraging Dynamic Programming for Conditional Sequence Modelling in Offline RL",
                    "abstract": "Recent works have shown that tackling offline reinforcement learning (RL) with a conditional policy produces promising results. The Decision Transformer (DT) combines the conditional policy approach and a transformer architecture, showing competitive performance against several benchmarks. However, DT lacks stitching ability -- one of the critical abilities for offline RL to learn the optimal policy from sub-optimal trajectories. This issue becomes particularly significant when the offline dataset only contains sub-optimal trajectories. On the other hand, the conventional RL approaches based on Dynamic Programming (such as Q-learning) do not have the same limitation; however, they suffer from unstable learning behaviours, especially when they rely on function approximation in an off-policy learning setting. In this paper, we propose the Q-learning Decision Transformer (QDT) to address the shortcomings of DT by leveraging the benefits of Dynamic Programming (Q-learning). It utilises the Dynamic Programming results to relabel the return-to-go in the training data to then train the DT with the relabelled data. Our approach efficiently exploits the benefits of these two approaches and compensates for each other's shortcomings to achieve better performance. We empirically show these in both simple toy environments and the more complex D4RL benchmark, showing competitive performance gains."
                },
                {
                    "title": "Efficient Planning in a Compact Latent Action Space",
                    "abstract": "Planning-based reinforcement learning has shown strong performance in tasks in discrete and low-dimensional continuous action spaces. However, planning usually brings significant computational overhead for decision-making, and scaling such methods to high-dimensional action spaces remains challenging. To advance efficient planning for high-dimensional continuous control, we propose Trajectory Autoencoding Planner (TAP), which learns low-dimensional latent action codes with a state-conditional VQ-VAE. The decoder of the VQ-VAE thus serves as a novel dynamics model that takes latent actions and current state as input and reconstructs long-horizon trajectories. During inference time, given a starting state, TAP searches over discrete latent actions to find trajectories that have both high probability under the training distribution and high predicted cumulative reward. Empirical evaluation in the offline RL setting demonstrates low decision latency which is indifferent to the growing raw action dimensionality. For Adroit robotic hand manipulation tasks with high-dimensional continuous action space, TAP surpasses existing model-based methods by a large margin and also beats strong model-free actor-critic baselines."
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "Addressing Optimism Bias in Sequence Modeling for Reinforcement Learning",
                    "abstract": "Impressive results in natural language processing (NLP) based on the Transformer neural network architecture have inspired researchers to explore viewing offline reinforcement learning (RL) as a generic sequence modeling problem. Recent works based on this paradigm have achieved state-of-the-art results in several of the mostly deterministic offline Atari and D4RL benchmarks. However, because these methods jointly model the states and actions as a single sequencing problem, they struggle to disentangle the effects of the policy and world dynamics on the return. Thus, in adversarial or stochastic environments, these methods lead to overly optimistic behavior that can be dangerous in safety-critical systems like autonomous driving. In this work, we propose a method that addresses this optimism bias by explicitly disentangling the policy and world models, which allows us at test time to search for policies that are robust to multiple possible futures in the environment. We demonstrate our method's superior performance on a variety of autonomous driving tasks in simulation."
                },
                {
                    "title": "Imitating Past Successes can be Very Suboptimal",
                    "abstract": "Prior work has proposed a simple strategy for reinforcement learning (RL): label experience with the outcomes achieved in that experience, and then imitate the relabeled experience. These outcome-conditioned imitation learning methods are appealing because of their simplicity, strong performance, and close ties with supervised learning. However, it remains unclear how these methods relate to the standard RL objective, reward maximization. In this paper, we formally relate outcome-conditioned imitation learning to reward maximization, drawing a precise relationship between the learned policy and Q-values and explaining the close connections between these methods and prior EM-based policy search methods. This analysis shows that existing outcome-conditioned imitation learning methods do not necessarily improve the policy, but a simple modification results in a method that does guarantee policy improvement, under some assumptions."
                },
                {
                    "title": "When does return-conditioned supervised learning work for offline reinforcement learning?",
                    "abstract": "Several recent works have proposed a class of algorithms for the offline reinforcement learning (RL) problem that we will refer to as return-conditioned supervised learning (RCSL). RCSL algorithms learn the distribution of actions conditioned on both the state and the return of the trajectory. Then they define a policy by conditioning on achieving high return. In this paper, we provide a rigorous study of the capabilities and limitations of RCSL, something which is crucially missing in previous work. We find that RCSL returns the optimal policy under a set of assumptions that are stronger than those needed for the more traditional dynamic programming-based algorithms. We provide specific examples of MDPs and datasets that illustrate the necessity of these assumptions and the limits of RCSL. Finally, we present empirical evidence that these limitations will also cause issues in practice by providing illustrative experiments in simple point-mass environments and on datasets from the D4RL benchmark."
                },
                {
                    "title": "Multi-Game Decision Transformers",
                    "abstract": "A longstanding goal of the field of AI is a method for learning a highly capable, generalist agent from diverse experience. In the subfields of vision and language, this was largely achieved by scaling up transformer-based models and training them on large, diverse datasets. Motivated by this progress, we investigate whether the same strategy can be used to produce generalist reinforcement learning agents. Specifically, we show that a single transformer-based model - with a single set of weights - trained purely offline can play a suite of up to 46 Atari games simultaneously at close-to-human performance. When trained and evaluated appropriately, we find that the same trends observed in language and vision hold, including scaling of performance with model size and rapid adaptation to new games via fine-tuning. We compare several approaches in this multi-game setting, such as online and offline RL methods and behavioral cloning, and find that our Multi-Game Decision Transformer models offer the best scalability and performance. We release the pre-trained models and code to encourage further research in this direction."
                },
                {
                    "title": "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness",
                    "abstract": "Transformers are slow and memory-hungry on long sequences, since the time and memory complexity of self-attention are quadratic in sequence length. Approximate attention methods have attempted to address this problem by trading off model quality to reduce the compute complexity, but often do not achieve wall-clock speedup. We argue that a missing principle is making attention algorithms IO-aware -- accounting for reads and writes between levels of GPU memory. We propose FlashAttention, an IO-aware exact attention algorithm that uses tiling to reduce the number of memory reads/writes between GPU high bandwidth memory (HBM) and GPU on-chip SRAM. We analyze the IO complexity of FlashAttention, showing that it requires fewer HBM accesses than standard attention, and is optimal for a range of SRAM sizes. We also extend FlashAttention to block-sparse attention, yielding an approximate attention algorithm that is faster than any existing approximate attention method. FlashAttention trains Transformers faster than existing baselines: 15% end-to-end wall-clock speedup on BERT-large (seq. length 512) compared to the MLPerf 1.1 training speed record, 3$\\times$ speedup on GPT-2 (seq. length 1K), and 2.4$\\times$ speedup on long-range arena (seq. length 1K-4K). FlashAttention and block-sparse FlashAttention enable longer context in Transformers, yielding higher quality models (0.7 better perplexity on GPT-2 and 6.4 points of lift on long-document classification) and entirely new capabilities: the first Transformers to achieve better-than-chance performance on the Path-X challenge (seq. length 16K, 61.4% accuracy) and Path-256 (seq. length 64K, 63.1% accuracy)."
                },
                {
                    "title": "Planning with Diffusion for Flexible Behavior Synthesis",
                    "abstract": "Model-based reinforcement learning methods often use learning only for the purpose of estimating an approximate dynamics model, offloading the rest of the decision-making work to classical trajectory optimizers. While conceptually simple, this combination has a number of empirical shortcomings, suggesting that learned models may not be well-suited to standard trajectory optimization. In this paper, we consider what it would look like to fold as much of the trajectory optimization pipeline as possible into the modeling problem, such that sampling from the model and planning with it become nearly identical. The core of our technical approach lies in a diffusion probabilistic model that plans by iteratively denoising trajectories. We show how classifier-guided sampling and image inpainting can be reinterpreted as coherent planning strategies, explore the unusual and useful properties of diffusion-based planning methods, and demonstrate the effectiveness of our framework in control settings that emphasize long-horizon decision-making and test-time flexibility."
                },
                {
                    "title": "Upside-Down Reinforcement Learning Can Diverge in Stochastic Environments With Episodic Resets",
                    "abstract": "Upside-Down Reinforcement Learning (UDRL) is an approach for solving RL problems that does not require value functions and uses only supervised learning, where the targets for given inputs in a dataset do not change over time. Ghosh et al. proved that Goal-Conditional Supervised Learning (GCSL) -- which can be viewed as a simplified version of UDRL -- optimizes a lower bound on goal-reaching performance. This raises expectations that such algorithms may enjoy guaranteed convergence to the optimal policy in arbitrary environments, similar to certain well-known traditional RL algorithms. Here we show that for a specific episodic UDRL algorithm (eUDRL, including GCSL), this is not the case, and give the causes of this limitation. To do so, we first introduce a helpful rewrite of eUDRL as a recursive policy update. This formulation helps to disprove its convergence to the optimal policy for a wide class of stochastic environments. Finally, we provide a concrete example of a very simple environment where eUDRL diverges. Since the primary aim of this paper is to present a negative result, and the best counterexamples are the simplest ones, we restrict all discussions to finite (discrete) environments, ignoring issues of function approximation and limited sample size."
                },
                {
                    "title": "Adversarially Trained Actor Critic for Offline Reinforcement Learning",
                    "abstract": "We propose Adversarially Trained Actor Critic (ATAC), a new model-free algorithm for offline reinforcement learning (RL) under insufficient data coverage, based on the concept of relative pessimism. ATAC is designed as a two-player Stackelberg game: A policy actor competes against an adversarially trained value critic, who finds data-consistent scenarios where the actor is inferior to the data-collection behavior policy. We prove that, when the actor attains no regret in the two-player game, running ATAC produces a policy that provably 1) outperforms the behavior policy over a wide range of hyperparameters that control the degree of pessimism, and 2) competes with the best policy covered by data with appropriately chosen hyperparameters. Compared with existing works, notably our framework offers both theoretical guarantees for general function approximation and a deep RL implementation scalable to complex environments and large datasets. In the D4RL benchmark, ATAC consistently outperforms state-of-the-art offline RL algorithms on a range of continuous control tasks."
                },
                {
                    "title": "RvS: What is Essential for Offline RL via Supervised Learning?",
                    "abstract": "Recent work has shown that supervised learning alone, without temporal difference (TD) learning, can be remarkably effective for offline RL. When does this hold true, and which algorithmic components are necessary? Through extensive experiments, we boil supervised learning for offline RL down to its essential elements. In every environment suite we consider, simply maximizing likelihood with a two-layer feedforward MLP is competitive with state-of-the-art results of substantially more complex methods based on TD learning or sequence modeling with Transformers. Carefully choosing model capacity (e.g., via regularization or architecture) and choosing which information to condition on (e.g., goals or rewards) are critical for performance. These insights serve as a field guide for practitioners doing Reinforcement Learning via Supervised Learning (which we coin\"RvS learning\"). They also probe the limits of existing RvS methods, which are comparatively weak on random data, and suggest a number of open problems."
                },
                {
                    "title": "Offline Reinforcement Learning with Implicit Q-Learning",
                    "abstract": "Offline reinforcement learning requires reconciling two conflicting aims: learning a policy that improves over the behavior policy that collected the dataset, while at the same time minimizing the deviation from the behavior policy so as to avoid errors due to distributional shift. This trade-off is critical, because most current offline reinforcement learning methods need to query the value of unseen actions during training to improve the policy, and therefore need to either constrain these actions to be in-distribution, or else regularize their values. We propose an offline RL method that never needs to evaluate actions outside of the dataset, but still enables the learned policy to improve substantially over the best behavior in the data through generalization. The main insight in our work is that, instead of evaluating unseen actions from the latest policy, we can approximate the policy improvement step implicitly by treating the state value function as a random variable, with randomness determined by the action (while still integrating over the dynamics to avoid excessive optimism), and then taking a state conditional upper expectile of this random variable to estimate the value of the best actions in that state. This leverages the generalization capacity of the function approximator to estimate the value of the best available action at a given state without ever directly querying a Q-function with this unseen action. Our algorithm alternates between fitting this upper expectile value function and backing it up into a Q-function. Then, we extract the policy via advantage-weighted behavioral cloning. We dub our method implicit Q-learning (IQL). IQL demonstrates the state-of-the-art performance on D4RL, a standard benchmark for offline reinforcement learning. We also demonstrate that IQL achieves strong performance fine-tuning using online interaction after offline initialization."
                },
                {
                    "title": "Pessimistic Model-based Offline Reinforcement Learning under Partial Coverage",
                    "abstract": "We study model-based offline Reinforcement Learning with general function approximation without a full coverage assumption on the offline data distribution. We present an algorithm named Constrained Pessimistic Policy Optimization (CPPO)which leverages a general function class and uses a constraint over the model class to encode pessimism. Under the assumption that the ground truth model belongs to our function class (i.e., realizability in the function class), CPPO has a PAC guarantee with offline data only providing partial coverage, i.e., it can learn a policy that competes against any policy that is covered by the offline data. We then demonstrate that this algorithmic framework can be applied to many specialized Markov Decision Processes where additional structural assumptions can further refine the concept of partial coverage. Two notable examples are: (1) low-rank MDP with representation learning where the partial coverage condition is defined using a relative condition number measured by the unknown ground truth feature representation; (2) factored MDP where the partial coverage condition is defined using density ratio based concentrability coefficients associated with individual factors."
                },
                {
                    "title": "Bellman-consistent Pessimism for Offline Reinforcement Learning",
                    "abstract": "The use of pessimism, when reasoning about datasets lacking exhaustive exploration has recently gained prominence in offline reinforcement learning. Despite the robustness it adds to the algorithm, overly pessimistic reasoning can be equally damaging in precluding the discovery of good policies, which is an issue for the popular bonus-based pessimism. In this paper, we introduce the notion of Bellman-consistent pessimism for general function approximation: instead of calculating a point-wise lower bound for the value function, we implement pessimism at the initial state over the set of functions consistent with the Bellman equations. Our theoretical guarantees only require Bellman closedness as standard in the exploratory setting, in which case bonus-based pessimism fails to provide guarantees. Even in the special case of linear function approximation where stronger expressivity assumptions hold, our result improves upon a recent bonus-based approach by $\\mathcal{O}(d)$ in its sample complexity when the action space is finite. Remarkably, our algorithms automatically adapt to the best bias-variance tradeoff in the hindsight, whereas most prior approaches require tuning extra hyperparameters a priori."
                },
                {
                    "title": "A Minimalist Approach to Offline Reinforcement Learning",
                    "abstract": "Offline reinforcement learning (RL) defines the task of learning from a fixed batch of data. Due to errors in value estimation from out-of-distribution actions, most offline RL algorithms take the approach of constraining or regularizing the policy with the actions contained in the dataset. Built on pre-existing RL algorithms, modifications to make an RL algorithm work offline comes at the cost of additional complexity. Offline RL algorithms introduce new hyperparameters and often leverage secondary components such as generative models, while adjusting the underlying RL algorithm. In this paper we aim to make a deep RL algorithm work while making minimal changes. We find that we can match the performance of state-of-the-art offline RL algorithms by simply adding a behavior cloning term to the policy update of an online RL algorithm and normalizing the data. The resulting algorithm is a simple to implement and tune baseline, while more than halving the overall run time by removing the additional computational overhead of previous methods."
                },
                {
                    "title": "Offline Reinforcement Learning as One Big Sequence Modeling Problem",
                    "abstract": "Reinforcement learning (RL) is typically concerned with estimating stationary policies or single-step models, leveraging the Markov property to factorize problems in time. However, we can also view RL as a generic sequence modeling problem, with the goal being to produce a sequence of actions that leads to a sequence of high rewards. Viewed in this way, it is tempting to consider whether high-capacity sequence prediction models that work well in other domains, such as natural-language processing, can also provide effective solutions to the RL problem. To this end, we explore how RL can be tackled with the tools of sequence modeling, using a Transformer architecture to model distributions over trajectories and repurposing beam search as a planning algorithm. Framing RL as sequence modeling problem simplifies a range of design decisions, allowing us to dispense with many of the components common in offline RL algorithms. We demonstrate the flexibility of this approach across long-horizon dynamics prediction, imitation learning, goal-conditioned RL, and offline RL. Further, we show that this approach can be combined with existing model-free algorithms to yield a state-of-the-art planner in sparse-reward, long-horizon tasks."
                },
                {
                    "title": "Decision Transformer: Reinforcement Learning via Sequence Modeling",
                    "abstract": "We introduce a framework that abstracts Reinforcement Learning (RL) as a sequence modeling problem. This allows us to draw upon the simplicity and scalability of the Transformer architecture, and associated advances in language modeling such as GPT-x and BERT. In particular, we present Decision Transformer, an architecture that casts the problem of RL as conditional sequence modeling. Unlike prior approaches to RL that fit value functions or compute policy gradients, Decision Transformer simply outputs the optimal actions by leveraging a causally masked Transformer. By conditioning an autoregressive model on the desired return (reward), past states, and actions, our Decision Transformer model can generate future actions that achieve the desired return. Despite its simplicity, Decision Transformer matches or exceeds the performance of state-of-the-art model-free offline RL baselines on Atari, OpenAI Gym, and Key-to-Door tasks."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "OPAL: Offline Primitive Discovery for Accelerating Offline Reinforcement Learning",
                    "abstract": "Reinforcement learning (RL) has achieved impressive performance in a variety of online settings in which an agent's ability to query the environment for transitions and rewards is effectively unlimited. However, in many practical applications, the situation is reversed: an agent may have access to large amounts of undirected offline experience data, while access to the online environment is severely limited. In this work, we focus on this offline setting. Our main insight is that, when presented with offline data composed of a variety of behaviors, an effective way to leverage this data is to extract a continuous space of recurring and temporally extended primitive behaviors before using these primitives for downstream task learning. Primitives extracted in this way serve two purposes: they delineate the behaviors that are supported by the data from those that are not, making them useful for avoiding distributional shift in offline RL; and they provide a degree of temporal abstraction, which reduces the effective horizon yielding better learning in theory, and improved offline RL in practice. In addition to benefiting offline policy optimization, we show that performing offline primitive learning in this way can also be leveraged for improving few-shot imitation learning as well as exploration and transfer in online RL on a variety of benchmark domains. Visualizations are available at this https URL"
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Learning Latent Space Energy-Based Prior Model",
                    "abstract": "The generator model assumes that the observed example is generated by a low-dimensional latent vector via a top-down network, and the latent vector follows a simple and known prior distribution, such as uniform or Gaussian white noise distribution. While we can learn an expressive top-down network to map the prior distribution to the data distribution, we can also learn an expressive prior model instead of assuming a given prior distribution. This follows the philosophy of empirical Bayes where the prior model is learned from the observed data. We propose to learn an energy-based prior model for the latent vector, where the energy function is parametrized by a very simple multi-layer perceptron. Due to the low-dimensionality of the latent space, learning a latent space energy-based prior model proves to be both feasible and desirable. In this paper, we develop the maximum likelihood learning algorithm and its variation based on short-run Markov chain Monte Carlo sampling from the prior and the posterior distributions of the latent vector, and we show that the learned model exhibits strong performance in terms of image and text generation and anomaly detection."
                },
                {
                    "title": "Conservative Q-Learning for Offline Reinforcement Learning",
                    "abstract": "Effectively leveraging large, previously collected datasets in reinforcement learning (RL) is a key challenge for large-scale real-world applications. Offline RL algorithms promise to learn effective policies from previously-collected, static datasets without further interaction. However, in practice, offline RL presents a major challenge, and standard off-policy RL methods can fail due to overestimation of values induced by the distributional shift between the dataset and the learned policy, especially when training on complex and multi-modal data distributions. In this paper, we propose conservative Q-learning (CQL), which aims to address these limitations by learning a conservative Q-function such that the expected value of a policy under this Q-function lower-bounds its true value. We theoretically show that CQL produces a lower bound on the value of the current policy and that it can be incorporated into a policy learning procedure with theoretical improvement guarantees. In practice, CQL augments the standard Bellman error objective with a simple Q-value regularizer which is straightforward to implement on top of existing deep Q-learning and actor-critic implementations. On both discrete and continuous control domains, we show that CQL substantially outperforms existing offline RL methods, often learning policies that attain 2-5 times higher final return, especially when learning from complex and multi-modal data distributions."
                },
                {
                    "title": "D4RL: Datasets for Deep Data-Driven Reinforcement Learning",
                    "abstract": "The offline reinforcement learning (RL) problem, also known as batch RL, refers to the setting where a policy must be learned from a static dataset, without additional online data collection. This setting is compelling as potentially it allows RL methods to take advantage of large, pre-collected datasets, much like how the rise of large datasets has fueled results in supervised learning in recent years. However, existing online RL benchmarks are not tailored towards the offline setting, making progress in offline RL difficult to measure. In this work, we introduce benchmarks specifically designed for the offline setting, guided by key properties of datasets relevant to real-world applications of offline RL. Examples of such properties include: datasets generated via hand-designed controllers and human demonstrators, multi-objective datasets where an agent can perform different tasks in the same environment, and datasets consisting of a mixtures of policies. To facilitate research, we release our benchmark tasks and datasets with a comprehensive evaluation of existing algorithms and an evaluation protocol together with an open-source codebase. We hope that our benchmark will focus research effort on methods that drive improvements not just on simulated tasks, but ultimately on the kinds of real-world problems where offline RL will have the largest impact."
                },
                {
                    "title": "Longformer: The Long-Document Transformer",
                    "abstract": "Transformer-based models are unable to process long sequences due to their self-attention operation, which scales quadratically with the sequence length. To address this limitation, we introduce the Longformer with an attention mechanism that scales linearly with sequence length, making it easy to process documents of thousands of tokens or longer. Longformer's attention mechanism is a drop-in replacement for the standard self-attention and combines a local windowed attention with a task motivated global attention. Following prior work on long-sequence transformers, we evaluate Longformer on character-level language modeling and achieve state-of-the-art results on text8 and enwik8. In contrast to most prior work, we also pretrain Longformer and finetune it on a variety of downstream tasks. Our pretrained Longformer consistently outperforms RoBERTa on long document tasks and sets new state-of-the-art results on WikiHop and TriviaQA. We finally introduce the Longformer-Encoder-Decoder (LED), a Longformer variant for supporting long document generative sequence-to-sequence tasks, and demonstrate its effectiveness on the arXiv summarization dataset."
                },
                {
                    "title": "Learning Latent Plans from Play",
                    "abstract": "We propose learning from teleoperated play data (LfP) as a way to scale up multi-task robotic skill learning. Learning from play (LfP) offers three main advantages: 1) It is cheap. Large amounts of play data can be collected quickly as it does not require scene staging, task segmenting, or resetting to an initial state. 2) It is general. It contains both functional and non-functional behavior, relaxing the need for a predefined task distribution. 3) It is rich. Play involves repeated, varied behavior and naturally leads to high coverage of the possible interaction space. These properties distinguish play from expert demonstrations, which are rich, but expensive, and scripted unattended data collection, which is cheap, but insufficiently rich. Variety in play, however, presents a multimodality challenge to methods seeking to learn control on top. To this end, we introduce Play-LMP, a method designed to handle variability in the LfP setting by organizing it in an embedding space. Play-LMP jointly learns 1) reusable latent plan representations unsupervised from play data and 2) a single goal-conditioned policy capable of decoding inferred plans to achieve user-specified tasks. We show empirically that Play-LMP, despite not being trained on task-specific data, is capable of generalizing to 18 complex user-specified manipulation tasks with average success of 85.5%, outperforming individual models trained on expert demonstrations (success of 70.3%). Furthermore, we find that play-supervised models, unlike their expert-trained counterparts, 1) are more robust to perturbations and 2) exhibit retrying-till-success. Finally, despite never being trained with task labels, we find that our agent learns to organize its latent plan space around functional tasks. Videos of the performed experiments are available at learning-from-play.github.io"
                },
                {
                    "title": "Why is Posterior Sampling Better than Optimism for Reinforcement Learning?",
                    "abstract": "Computational results demonstrate that posterior sampling for reinforcement learning (PSRL) dramatically outperforms algorithms driven by optimism, such as UCRL2. We provide insight into the extent of this performance boost and the phenomenon that drives it. We leverage this insight to establish an $\\tilde{O}(H\\sqrt{SAT})$ Bayesian expected regret bound for PSRL in finite-horizon episodic Markov decision processes, where $H$ is the horizon, $S$ is the number of states, $A$ is the number of actions and $T$ is the time elapsed. This improves upon the best previous bound of $\\tilde{O}(H S \\sqrt{AT})$ for any reinforcement learning algorithm."
                },
                {
                    "title": "Alternating Back-Propagation for Generator Network",
                    "abstract": "\n \n This paper proposes an alternating back-propagation algorithm for learning the generator network model. The model is a non-linear generalization of factor analysis. In this model, the mapping from the continuous latent factors to the observed signal is parametrized by a convolutional neural network. The alternating back-propagation algorithm iterates the following two steps: (1) Inferential back-propagation, which infers the latent factors by Langevin dynamics or gradient descent. (2) Learning back-propagation, which updates the parameters given the inferred latent factors by gradient descent. The gradient computations in both steps are powered by back-propagation, and they share most of their code in common. We show that the alternating back-propagation algorithm can learn realistic generator models of natural images, video sequences, and sounds. Moreover, it can also be used to learn from incomplete or indirect training data.\n \n"
                },
                {
                    "title": "Continuous control with deep reinforcement learning",
                    "abstract": "We adapt the ideas underlying the success of Deep Q-Learning to the continuous action domain. We present an actor-critic, model-free algorithm based on the deterministic policy gradient that can operate over continuous action spaces. Using the same learning algorithm, network architecture and hyper-parameters, our algorithm robustly solves more than 20 simulated physics tasks, including classic problems such as cartpole swing-up, dexterous manipulation, legged locomotion and car driving. Our algorithm is able to find policies whose performance is competitive with those found by a planning algorithm with full access to the dynamics of the domain and its derivatives. We further demonstrate that for many of the tasks the algorithm can learn policies end-to-end: directly from raw pixel inputs."
                },
                {
                    "title": "(More) Efficient Reinforcement Learning via Posterior Sampling",
                    "abstract": "Most provably-efficient reinforcement learning algorithms introduce optimism about poorly-understood states and actions to encourage exploration. We study an alternative approach for efficient exploration: posterior sampling for reinforcement learning (PSRL). This algorithm proceeds in repeated episodes of known duration. At the start of each episode, PSRL updates a prior distribution over Markov decision processes and takes one sample from this posterior. PSRL then follows the policy that is optimal for this sample during the episode. The algorithm is conceptually simple, computationally efficient and allows an agent to encode prior knowledge in a natural way. We establish an O(\u03c4S/\u221aAT) bound on expected regret, where T is time, \u03c4 is the episode length and S and A are the cardinalities of the state and action spaces. This bound is one of the first for an algorithm not based on optimism, and close to the state of the art for any reinforcement learning algorithm. We show through simulation that PSRL significantly outperforms existing algorithms with similar regret bounds."
                },
                {
                    "title": "MCMC Using Hamiltonian Dynamics",
                    "abstract": "Hamiltonian dynamics can be used to produce distant proposals for the Metropolis algorithm, thereby avoiding the slow exploration of the state space that results from the diffusive behaviour of simple random-walk proposals. Though originating in physics, Hamiltonian dynamics can be applied to most problems with continuous state spaces by simply introducing fictitious \"momentum\" variables. A key to its usefulness is that Hamiltonian dynamics preserves volume, and its trajectories can thus be used to define complex mappings without the need to account for a hard-to-compute Jacobian factor - a property that can be exactly maintained even when the dynamics is approximated by discretizing time. In this review, I discuss theoretical and practical aspects of Hamiltonian Monte Carlo, and present some of its variations, including using windows of states for deciding on acceptance or rejection, computing trajectories using fast approximations, tempering during the course of a trajectory to handle isolated modes, and short-cut methods that prevent useless trajectories from taking much computation time."
                },
                {
                    "title": "Efficient Reductions for Imitation Learning",
                    "abstract": "Imitation Learning, while applied successfully on many large real-world problems, is typically addressed as a standard supervised learning problem, where it is assumed the training and testing data are i.i.d.. This is not true in imitation learning as the learned policy influences the future test inputs (states) upon which it will be tested. We show that this leads to compounding errors and a regret bound that grows quadratically in the time horizon of the task. We propose two alternative algorithms for imitation learning where training occurs over several episodes of interaction. These two approaches share in common that the learner\u2019s policy is slowly modified from executing the expert\u2019s policy to the learned policy. We show that this leads to stronger performance guarantees and demonstrate the improved performance on two challenging problems: training a learner to play 1) a 3D racing game (Super Tux Kart) and 2) Mario Bros.; given input images from the games and corresponding actions taken by a human expert and near-optimal planner respectively."
                },
                {
                    "title": "Training restricted Boltzmann machines using approximations to the likelihood gradient",
                    "abstract": "A new algorithm for training Restricted Boltzmann Machines is introduced. The algorithm, named Persistent Contrastive Divergence, is different from the standard Contrastive Divergence algorithms in that it aims to draw samples from almost exactly the model distribution. It is compared to some standard Contrastive Divergence and Pseudo-Likelihood algorithms on the tasks of modeling and classifying various types of data. The Persistent Contrastive Divergence algorithm outperforms the other algorithms, and is equally fast and simple."
                },
                {
                    "title": "Learning Temporally AbstractWorld Models without Online Experimentation",
                    "abstract": "Agents that can build temporally abstract representations of their environment are better able to understand their world and make plans on extended time scales, with limited computational power and modeling capacity. However, existing methods for automatically learning temporally abstract world models usually require millions of online environmental interactions and incentivize agents to reach every accessible environmental state, which is infeasible for most real-world robots both in terms of data efficiency and hardware safety. In this paper, we present an approach for simultaneously learning sets of skills and temporally abstract, skill-conditioned world models purely from offline data, enabling agents to perform zero-shot online planning of skill sequences for new tasks. We show that our approach performs comparably to or better than a wide array of state-of-the-art offline RL algorithms on a number of simulated robotics locomotion and manipulation benchmarks, while offering a higher degree of adaptability to new goals. Finally, we show that our approach offers a much higher degree of robustness to perturbations in environmental dynamics, compared to policy-based methods."
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively learn and plan decision-making processes using trajectory-return pairs in a way that allows agents to predict long-term outcomes and adapt to environmental changes?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of decision-making in machine learning, particularly in environments where designing step-wise rewards is challenging. By improving how agents learn from trajectory-return pairs, we can enhance their ability to make informed decisions that lead to higher returns. This research could lead to practical applications in various domains, such as robotics, autonomous systems, and game playing, where effective planning and adaptability are essential. Furthermore, it could inspire future research on generative modeling in decision-making, potentially leading to more robust and flexible AI systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need for agents to predict long-term outcomes based solely on trajectory-return pairs, which requires them to assign step-wise credits without explicit guidance. Naive approaches may fail because they do not account for the complexities of planning, such as the need for persistence in decision-making and adaptability to changing environments. Additionally, the technical obstacles include the effective modeling of latent variables that represent plans and the integration of these variables into a generative framework that can handle the stochastic nature of decision-making.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on direct reward-based learning or simplistic planning methods that do not adequately address the intricacies of long-term decision-making. Limitations in existing solutions include a lack of effective inductive biases that capture the essence of planning and the fragility of certain learning algorithms, such as CQL, under the specified data conditions. Our approach differs by introducing a top-down latent variable model that decouples trajectory generation from return expectations, allowing for a more nuanced understanding of planning and decision-making.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the Latent Plan Transformer (LPT), which utilizes a latent vector derived from Gaussian white noise, a Transformer-based policy conditioned on this latent vector, and a return estimation model. We will employ maximum likelihood estimation (MLE) to train the model. The dataset will consist of trajectory-return pairs, and we will evaluate the model's performance using metrics that assess the quality of the generated trajectories and the accuracy of return predictions. We expect"
            }
        },
        "author_data": {
            "0a3f163b-8e3d-4199-b415-deb8caa467a2": {
                "pk": "0a3f163b-8e3d-4199-b415-deb8caa467a2",
                "name": "Deqian Kong",
                "collaborators": [
                    "Ying Nian Wu",
                    "Jianwen Xie",
                    "Yuhao Huang",
                    "Bo Pang",
                    "Dinghuai Zhang",
                    "Yasi Zhang",
                    "Peiyu Yu",
                    "Shi Feng",
                    "Nandini Trivedi",
                    "Tian Han"
                ],
                "domain": [
                    "Machine Learning",
                    "Molecular Design",
                    "Reinforcement Learning",
                    "Statistical Mechanics"
                ],
                "publications": [
                    {
                        "title": "A statistical approach to topological entanglement: Boltzmann machine representation of high-order irreducible correlation",
                        "abstract": "Strongly interacting systems can be described in terms of correlation functions at various orders. A quantum analog of high-order correlations is the topological entanglement in topologically ordered states of matter at zero temperature, usually quantified by topological entanglement entropy (TEE). In this work, we propose a statistical interpretation that unifies the two under the same information-theoretic framework. We demonstrate that the existence of a non-zero TEE can be understood in the statistical view as the emergent $n$th order mutual information $I_n$ (for arbitrary integer $n\\ge 3$) reflected in projectively measured samples, which also makes explicit the equivalence between the two existing methods for its extraction -- the Kitaev-Preskill and the Levin-Wen construction. To exploit the statistical nature of $I_n$, we construct a restricted Boltzmann machine (RBM) which captures the high-order correlations and correspondingly the topological entanglement that are encoded in the distribution of projected samples by representing the entanglement Hamiltonian of a local region under the proper basis. Furthermore, we derive a closed form which presents a method to interrogate the trained RBM, making explicit the analytical form of arbitrary order of correlations relevant for $I_n$. We remark that the interrogation method for extracting high-order correlation can also be applied to the construction of auxiliary fields that disentangle many-body interactions relevant for diverse interacting models."
                    },
                    {
                        "title": "Molecule Design by Latent Prompt Transformer",
                        "abstract": "This paper proposes a latent prompt Transformer model for solving challenging optimization problems such as molecule design, where the goal is to find molecules with optimal values of a target chemical or biological property that can be computed by an existing software. Our proposed model consists of three components. (1) A latent vector whose prior distribution is modeled by a Unet transformation of a Gaussian white noise vector. (2) A molecule generation model that generates the string-based representation of molecule conditional on the latent vector in (1). We adopt the causal Transformer model that takes the latent vector in (1) as prompt. (3) A property prediction model that predicts the value of the target property of a molecule based on a non-linear regression on the latent vector in (1). We call the proposed model the latent prompt Transformer model. After initial training of the model on existing molecules and their property values, we then gradually shift the model distribution towards the region that supports desired values of the target property for the purpose of molecule design. Our experiments show that our proposed model achieves state of the art performances on several benchmark molecule design tasks."
                    },
                    {
                        "title": "Molecule Design by Latent Space Energy-Based Modeling and Gradual Distribution Shifting",
                        "abstract": "Generation of molecules with desired chemical and biological properties such as high drug-likeness, high binding affinity to target proteins, is critical for drug discovery. In this paper, we propose a probabilistic generative model to capture the joint distribution of molecules and their properties. Our model assumes an energy-based model (EBM) in the latent space. Conditional on the latent vector, the molecule and its properties are modeled by a molecule generation model and a property regression model respectively. To search for molecules with desired properties, we propose a sampling with gradual distribution shifting (SGDS) algorithm, so that after learning the model initially on the training data of existing molecules and their properties, the proposed algorithm gradually shifts the model distribution towards the region supported by molecules with desired values of properties. Our experiments show that our method achieves very strong performances on various molecule design tasks."
                    },
                    {
                        "title": "Latent Space Energy-based Neural ODEs",
                        "abstract": "This paper introduces a novel family of deep dynamical models designed to represent continuous-time sequence data. This family of models generates each data point in the time series by a neural emission model, which is a non-linear transformation of a latent state vector. The trajectory of the latent states is implicitly described by a neural ordinary differential equation (ODE), with the initial state following an informative prior distribution parameterized by an energy-based model. Furthermore, we can extend this model to disentangle dynamic states from underlying static factors of variation, represented as time-invariant variables in the latent space. We train the model using maximum likelihood estimation with Markov chain Monte Carlo (MCMC) in an end-to-end manner, without requiring additional assisting components such as an inference network. Our experiments on oscillating systems, videos and real-world state sequences (MuJoCo) illustrate that ODEs with the learnable energy-based prior outperform existing counterparts, and can generalize to new dynamic parameterization, enabling long-horizon predictions."
                    },
                    {
                        "title": "Diverse and Faithful Knowledge-Grounded Dialogue Generation via Sequential Posterior Inference",
                        "abstract": "The capability to generate responses with diversity and faithfulness using factual knowledge is paramount for creating a human-like, trustworthy dialogue system. Common strategies either adopt a two-step paradigm, which optimizes knowledge selection and response generation separately, and may overlook the inherent correlation between these two tasks, or leverage conditional variational method to jointly optimize knowledge selection and response generation by employing an inference network. In this paper, we present an end-to-end learning framework, termed Sequential Posterior Inference (SPI), capable of selecting knowledge and generating dialogues by approximately sampling from the posterior distribution. Unlike other methods, SPI does not require the inference network or assume a simple geometry of the posterior distribution. This straightforward and intuitive inference procedure of SPI directly queries the response generation model, allowing for accurate knowledge selection and generation of faithful responses. In addition to modeling contributions, our experimental results on two common dialogue datasets (Wizard of Wikipedia and Holl-E) demonstrate that SPI outperforms previous strong baselines according to both automatic and human evaluation metrics."
                    },
                    {
                        "title": "YouRefIt: Embodied Reference Understanding with Language and Gesture",
                        "abstract": "We study the understanding of embodied reference: One agent uses both language and gesture to refer to an object to another agent in a shared physical environment. Of note, this new visual task requires understanding multimodal cues with perspective-taking to identify which object is being referred to. To tackle this problem, we introduce YouRefIt, a new crowd-sourced dataset of embodied reference collected in various physical scenes; the dataset contains 4,195 unique reference clips in 432 indoor scenes. To the best of our knowledge, this is the first embodied reference dataset that allows us to study referring expressions in daily physical scenes to understand referential behavior, human communication, and human-robot interaction. We further devise two benchmarks for image-based and video-based embodied reference understanding. Comprehensive baselines and extensive experiments provide the very first result of machine perception on how the referring expressions and gestures affect the embodied reference understanding. Our results provide essential evidence that gestural cues are as critical as language cues in understanding the embodied reference."
                    },
                    {
                        "title": "Dual-Space Optimization: Improved Molecule Sequence Design by Latent Prompt Transformer",
                        "abstract": "Designing molecules with desirable properties, such as drug-likeliness and high binding affinities towards protein targets, is a challenging problem. In this paper, we propose the Dual-Space Optimization (DSO) method that integrates latent space sampling and data space selection to solve this problem. DSO iteratively updates a latent space generative model and a synthetic dataset in an optimization process that gradually shifts the generative model and the synthetic data towards regions of desired property values. Our generative model takes the form of a Latent Prompt Transformer (LPT) where the latent vector serves as the prompt of a causal transformer. Our extensive experiments demonstrate effectiveness of the proposed method, which sets new performance benchmarks across single-objective, multi-objective and constrained molecule design tasks."
                    },
                    {
                        "title": "DODT: Enhanced Online Decision Transformer Learning through Dreamer's Actor-Critic Trajectory Forecasting",
                        "abstract": "Advancements in reinforcement learning have led to the development of sophisticated models capable of learning complex decision-making tasks. However, efficiently integrating world models with decision transformers remains a challenge. In this paper, we introduce a novel approach that combines the Dreamer algorithm's ability to generate anticipatory trajectories with the adaptive learning strengths of the Online Decision Transformer. Our methodology enables parallel training where Dreamer-produced trajectories enhance the contextual decision-making of the transformer, creating a bidirectional enhancement loop. We empirically demonstrate the efficacy of our approach on a suite of challenging benchmarks, achieving notable improvements in sample efficiency and reward maximization over existing methods. Our results indicate that the proposed integrated framework not only accelerates learning but also showcases robustness in diverse and dynamic scenarios, marking a significant step forward in model-based reinforcement learning."
                    },
                    {
                        "title": "Latent Energy-Based Odyssey: Black-Box Optimization via Expanded Exploration in the Energy-Based Latent Space",
                        "abstract": "Offline Black-Box Optimization (BBO) aims at optimizing a black-box function using the knowledge from a pre-collected offline dataset of function values and corresponding input designs. However, the high-dimensional and highly-multimodal input design space of black-box function pose inherent challenges for most existing methods that model and operate directly upon input designs. These issues include but are not limited to high sample complexity, which relates to inaccurate approximation of black-box function; and insufficient coverage and exploration of input design modes, which leads to suboptimal proposal of new input designs. In this work, we consider finding a latent space that serves as a compressed yet accurate representation of the design-value joint space, enabling effective latent exploration of high-value input design modes. To this end, we formulate an learnable energy-based latent space, and propose Noise-intensified Telescoping density-Ratio Estimation (NTRE) scheme for variational learning of an accurate latent space model without costly Markov Chain Monte Carlo. The optimization process is then exploration of high-value designs guided by the learned energy-based model in the latent space, formulated as gradient-based sampling from a latent-variable-parameterized inverse model. We show that our particular parameterization encourages expanded exploration around high-value design modes, motivated by inversion thinking of a fundamental result of conditional covariance matrix typically used for variance reduction. We observe that our method, backed by an accurately learned informative latent space and an expanding-exploration model design, yields significant improvements over strong previous methods on both synthetic and real world datasets such as the design-bench suite."
                    }
                ]
            },
            "d242f906-d655-495d-b4c4-0d4930018585": {
                "pk": "d242f906-d655-495d-b4c4-0d4930018585",
                "name": "Dehong Xu",
                "collaborators": [
                    "Ying Nian Wu",
                    "Ruiqi Gao",
                    "Wen-Hao Zhang",
                    "Xue-Xin Wei",
                    "Minglu Zhao",
                    "Tao Gao",
                    "Liang Qiu",
                    "Minseok Kim",
                    "Faisal Ladhak",
                    "Jaeyoung Do"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Neural Networks",
                    "Cognitive Science",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "From Cognition to Computation: A Comparative Review of Human Attention and Transformer Architectures",
                        "abstract": "Attention is a cornerstone of human cognition that facilitates the efficient extraction of information in everyday life. Recent developments in artificial intelligence like the Transformer architecture also incorporate the idea of attention in model designs. However, despite the shared fundamental principle of selectively attending to information, human attention and the Transformer model display notable differences, particularly in their capacity constraints, attention pathways, and intentional mechanisms. Our review aims to provide a comparative analysis of these mechanisms from a cognitive-functional perspective, thereby shedding light on several open research questions. The exploration encourages interdisciplinary efforts to derive insights from human attention mechanisms in the pursuit of developing more generalized artificial intelligence."
                    },
                    {
                        "title": "Aligning Large Language Models via Fine-grained Supervision",
                        "abstract": "Pre-trained large-scale language models (LLMs) excel at producing coherent articles, yet their outputs may be untruthful, toxic, or fail to align with user expectations. Current approaches focus on using reinforcement learning with human feedback (RLHF) to improve model alignment, which works by transforming coarse human preferences of LLM outputs into a feedback signal that guides the model learning process. However, because this approach operates on sequence-level feedback, it lacks the precision to identify the exact parts of the output affecting user preferences. To address this gap, we propose a method to enhance LLM alignment through fine-grained token-level supervision. Specifically, we ask annotators to minimally edit less preferred responses within the standard reward modeling dataset to make them more favorable, ensuring changes are made only where necessary while retaining most of the original content. The refined dataset is used to train a token-level reward model, which is then used for training our fine-grained Proximal Policy Optimization (PPO) model. Our experiment results demonstrate that this approach can achieve up to an absolute improvement of $5.1\\%$ in LLM performance, in terms of win rate against the reference model, compared with the traditional PPO model."
                    },
                    {
                        "title": "Diverse and Faithful Knowledge-Grounded Dialogue Generation via Sequential Posterior Inference",
                        "abstract": "The capability to generate responses with diversity and faithfulness using factual knowledge is paramount for creating a human-like, trustworthy dialogue system. Common strategies either adopt a two-step paradigm, which optimizes knowledge selection and response generation separately, and may overlook the inherent correlation between these two tasks, or leverage conditional variational method to jointly optimize knowledge selection and response generation by employing an inference network. In this paper, we present an end-to-end learning framework, termed Sequential Posterior Inference (SPI), capable of selecting knowledge and generating dialogues by approximately sampling from the posterior distribution. Unlike other methods, SPI does not require the inference network or assume a simple geometry of the posterior distribution. This straightforward and intuitive inference procedure of SPI directly queries the response generation model, allowing for accurate knowledge selection and generation of faithful responses. In addition to modeling contributions, our experimental results on two common dialogue datasets (Wizard of Wikipedia and Holl-E) demonstrate that SPI outperforms previous strong baselines according to both automatic and human evaluation metrics."
                    },
                    {
                        "title": "Conformal Isometry of Lie Group Representation in Recurrent Network of Grid Cells",
                        "abstract": "The activity of the grid cell population in the medial entorhinal cortex (MEC) of the mammalian brain forms a vector representation of the self-position of the animal. Recurrent neural networks have been proposed to explain the properties of the grid cells by updating the neural activity vector based on the velocity input of the animal. In doing so, the grid cell system effectively performs path integration. In this paper, we investigate the algebraic, geometric, and topological properties of grid cells using recurrent network models. Algebraically, we study the Lie group and Lie algebra of the recurrent transformation as a representation of self-motion. Geometrically, we study the conformal isometry of the Lie group representation where the local displacement of the activity vector in the neural space is proportional to the local displacement of the agent in the 2D physical space. Topologically, the compact abelian Lie group representation automatically leads to the torus topology commonly assumed and observed in neuroscience. We then focus on a simple non-linear recurrent model that underlies the continuous attractor neural networks of grid cells. Our numerical experiments show that conformal isometry leads to hexagon periodic patterns in the grid cell responses and our model is capable of accurate path integration. Code is available at \\url{https://github.com/DehongXu/grid-cell-rnn}."
                    },
                    {
                        "title": "Emergence of Grid-like Representations by Training Recurrent Networks with Conformal Normalization",
                        "abstract": "Grid cells in the entorhinal cortex of mammalian brains exhibit striking hexagon grid firing patterns in their response maps as the animal (e.g., a rat) navigates in a 2D open environment. In this paper, we study the emergence of the hexagon grid patterns of grid cells based on a general recurrent neural network (RNN) model that captures the navigation process. The responses of grid cells collectively form a high dimensional vector, representing the 2D self-position of the agent. As the agent moves, the vector is transformed by an RNN that takes the velocity of the agent as input. We propose a simple yet general conformal normalization of the input velocity of the RNN, so that the local displacement of the position vector in the high-dimensional neural space is proportional to the local displacement of the agent in the 2D physical space, regardless of the direction of the input velocity. We apply this mechanism to both a linear RNN and nonlinear RNNs. Theoretically, we provide an understanding that explains the connection between conformal normalization and the emergence of hexagon grid patterns. Empirically, we conduct extensive experiments to verify that conformal normalization is crucial for the emergence of hexagon grid patterns, across various types of RNNs. The learned patterns share similar profiles to biological grid cells, and the topological properties of the patterns also align with our theoretical understanding."
                    },
                    {
                        "title": "An Investigation of Conformal Isometry Hypothesis for Grid Cells",
                        "abstract": "This paper investigates the conformal isometry hypothesis as a potential explanation for hexagonal periodic patterns in grid cell response maps. The hypothesis posits that grid cell activity forms a high-dimensional vector in neural space, encoding the agent's position in 2D physical space. As the agent moves, this vector rotates within a 2D manifold in the neural space, driven by a recurrent neural network. The conformal hypothesis suggests that this neural manifold is a conformally isometric embedding of physical space, where local displacements in neural space are proportional to those in physical space. In this paper, we conduct numerical experiments to show that this hypothesis leads to the hexagon periodic patterns of grid cells, agnostic to the choice of transformation models. Furthermore, we present a theoretical understanding that hexagon patterns emerge by minimizing our loss function because hexagon flat torus exhibits minimal deviation from local conformal isometry. In addition, we propose a conformal modulation of the agent's input velocity, enabling the recurrent neural network of grid cells to satisfy the conformal isometry hypothesis automatically."
                    }
                ]
            },
            "efdd0d23-73dc-4f38-bc8f-ae63965361a5": {
                "pk": "efdd0d23-73dc-4f38-bc8f-ae63965361a5",
                "name": "Minglu Zhao",
                "collaborators": [
                    "Tao Gao",
                    "Dehong Xu",
                    "Louis Zhao",
                    "Chen Ye Gan",
                    "Stephanie Stacy",
                    "Chenfei Li",
                    "Yiling Yun",
                    "Qingyi Zhao",
                    "Max Kleiman-Weiner"
                ],
                "domain": [
                    "Cognitive Science",
                    "Artificial Intelligence",
                    "Complex Networks",
                    "Communication Theory"
                ],
                "publications": [
                    {
                        "title": "From Cognition to Computation: A Comparative Review of Human Attention and Transformer Architectures",
                        "abstract": "Attention is a cornerstone of human cognition that facilitates the efficient extraction of information in everyday life. Recent developments in artificial intelligence like the Transformer architecture also incorporate the idea of attention in model designs. However, despite the shared fundamental principle of selectively attending to information, human attention and the Transformer model display notable differences, particularly in their capacity constraints, attention pathways, and intentional mechanisms. Our review aims to provide a comparative analysis of these mechanisms from a cognitive-functional perspective, thereby shedding light on several open research questions. The exploration encourages interdisciplinary efforts to derive insights from human attention mechanisms in the pursuit of developing more generalized artificial intelligence."
                    },
                    {
                        "title": "Emergent Spatial Characteristics from Strategic Games Simulated on Random and Real Networks",
                        "abstract": "Complex networks are a great tool for simulating the outcomes of different strategies used within the iterated prisoners' dilemma game. However, because the strategies themselves rely on the connection between nodes, then initial network structure should have an impact on the progression of the game. By defining each interaction in terms of a prisoner's dilemma and using its payoff matrix as a basis for investigation, we implemented players with various interaction and edge attachment strategies, and ran this dynamic process on real and random networks with varying network structure. We found that, both network size and small world properties played an important role in not only deciding the convergence rate of the simulation but also the dominant status of nodes, under the conditions where identical strategies are employed by every player."
                    },
                    {
                        "title": "Modeling Communication to Coordinate Perspectives in Cooperation",
                        "abstract": "Communication is highly overloaded. Despite this, even young children are good at leveraging context to understand ambiguous signals. We propose a computational account of overloaded signaling from a shared agency perspective which we call the Imagined We for Communication. Under this framework, communication helps cooperators coordinate their perspectives, allowing them to act together to achieve shared goals. We assume agents are rational cooperators, which puts constraints on how signals can be sent and interpreted. We implement this model in a set of simulations demonstrating this model's success under increasing ambiguity as well as increasing layers of reasoning. Our model is capable of improving performance with deeper recursive reasoning; however, it outperforms comparison baselines at even the shallowest level, highlighting how shared knowledge and cooperative logic can do much of the heavy-lifting in language."
                    }
                ]
            },
            "175da6a7-95bc-4849-a39b-c96052e3e3fb": {
                "pk": "175da6a7-95bc-4849-a39b-c96052e3e3fb",
                "name": "Bo Pang",
                "collaborators": [
                    "Lillian Lee",
                    "Zhong-Ping Jiang",
                    "Cewu Lu",
                    "Deming Zhai",
                    "Junjun Jiang",
                    "Xianming Liu",
                    "Ying Nian Wu",
                    "Radu Soricut",
                    "Tian Han",
                    "Gao Peng"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Reinforcement Learning",
                    "Computer Vision",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "A Sentimental Education: Sentiment Analysis Using Subjectivity Summarization Based on Minimum Cuts",
                        "abstract": "Sentiment analysis seeks to identify the viewpoint(s) underlying a text span; an example application is classifying a movie review as \"thumbs up\" or \"thumbs down\". To determine this sentiment polarity, we propose a novel machine-learning method that applies text-categorization techniques to just the subjective portions of the document. Extracting these portions can be implemented using efficient techniques for finding minimum cuts in graphs; this greatly facilitates incorporation of cross-sentence contextual constraints."
                    },
                    {
                        "title": "Seeing stars: Exploiting class relationships for sentiment categorization with respect to rating scales",
                        "abstract": "We address the rating-inference problem, wherein rather than simply decide whether a review is \"thumbs up\" or \"thumbs down\", as in previous sentiment analysis work, one must determine an author's evaluation with respect to a multi-point scale (e.g., one to five \"stars\"). This task represents an interesting twist on standard multi-class text categorization because there are several different degrees of similarity between class labels; for example, \"three stars\" is intuitively closer to \"four stars\" than to \"one star\". We first evaluate human performance at the task. Then, we apply a meta-algorithm, based on a metric labeling formulation of the problem, that alters a given n-ary classifier's output in an explicit attempt to ensure that similar items receive similar labels. We show that the meta-algorithm can provide significant improvements over both multi-class and regression versions of SVMs when we employ a novel similarity measure appropriate to the problem."
                    },
                    {
                        "title": "Adaptive Optimal Control of Linear Periodic Systems: An Off-Policy Value Iteration Approach",
                        "abstract": "This paper studies the infinite-horizon adaptive optimal control of continuous-time linear periodic (CTLP) systems. A novel value iteration (VI) based off-policy ADP algorithm is proposed for a general class of CTLP systems, so that approximate optimal solutions can be obtained directly from the collected data, without the exact knowledge of system dynamics. Under mild conditions, the proofs on uniform convergence of the proposed algorithm to the optimal solutions are given for both the model-based and model-free cases. The VI-based ADP algorithm is able to find suboptimal controllers without assuming the knowledge of an initial stabilizing controller. Application to the optimal control of a triple inverted pendulum subjected to a periodically varying load demonstrates the feasibility and effectiveness of the proposed method."
                    },
                    {
                        "title": "Robust Reinforcement Learning: A Case Study in Linear Quadratic Regulation",
                        "abstract": "This paper studies the robustness of reinforcement learning algorithms to errors in the learning process. Specifically, we revisit the benchmark problem of discrete-time linear quadratic regulation (LQR) and study the long-standing open question: Under what conditions is the policy iteration method robustly stable from a dynamical systems perspective? Using advanced stability results in control theory, it is shown that policy iteration for LQR is inherently robust to small errors in the learning process and enjoys small-disturbance input-to-state stability: whenever the error in each iteration is bounded and small, the solutions of the policy iteration algorithm are also bounded, and, moreover, enter and stay in a small neighbourhood of the optimal LQR solution. As an application, a novel off-policy optimistic least-squares policy iteration for the LQR problem is proposed, when the system dynamics are subjected to additive stochastic disturbances. The proposed new results in robust reinforcement learning are validated by a numerical example."
                    },
                    {
                        "title": "Single Image Deraining via Scale-space Invariant Attention Neural Network",
                        "abstract": "Image enhancement from degradation of rainy artifacts plays a critical role in outdoor visual computing systems. In this paper, we tackle the notion of scale that deals with visual changes in appearance of rain steaks with respect to the camera. Specifically, we revisit multi-scale representation by scale-space theory, and propose to represent the multi-scale correlation in convolutional feature domain, which is more compact and robust than that in pixel domain. Moreover, to improve the modeling ability of the network, we do not treat the extracted multi-scale features equally, but design a novel scale-space invariant attention mechanism to help the network focus on parts of the features. In this way, we summarize the most activated presence of feature maps as the salient features. Extensive experiments results on synthetic and real rainy scenes demonstrate the superior performance of our scheme over the state-of-the-arts."
                    },
                    {
                        "title": "Fully Unsupervised Person Re-identification viaSelective Contrastive Learning",
                        "abstract": "Person re-identification (ReID) aims at searching the same identity person among images captured by various cameras. Unsupervised person ReID attracts a lot of attention recently, due to it works without intensive manual annotation and thus shows great potential of adapting to new conditions. Representation learning plays a critical role in unsupervised person ReID. In this work, we propose a novel selective contrastive learning framework for unsupervised feature learning. Specifically, different from traditional contrastive learning strategies, we propose to use multiple positives and adaptively sampled negatives for defining the contrastive loss, enabling to learn a feature embedding model with stronger identity discriminative representation. Moreover, we propose to jointly leverage global and local features to construct three dynamic dictionaries, among which the global and local memory banks are used for pairwise similarity computation and the mixture memory bank are used for contrastive loss definition. Experimental results demonstrate the superiority of our method in unsupervised person ReID compared with the state-of-the-arts."
                    },
                    {
                        "title": "Learning Latent Space Energy-Based Prior Model for Molecule Generation",
                        "abstract": "Deep generative models have recently been applied to molecule design. If the molecules are encoded in linear SMILES strings, modeling becomes convenient. However, models relying on string representations tend to generate invalid samples and duplicates. Prior work addressed these issues by building models on chemically-valid fragments or explicitly enforcing chemical rules in the generation process. We argue that an expressive model is sufficient to implicitly and automatically learn the complicated chemical rules from the data, even if molecules are encoded in simple character-level SMILES strings. We propose to learn latent space energy-based prior model with SMILES representation for molecule modeling. Our experiments show that our method is able to generate molecules with validity and uniqueness competitive with state-of-the-art models. Interestingly, generated molecules have structural and chemical features whose distributions almost perfectly match those of the real molecules."
                    },
                    {
                        "title": "PGT: A Progressive Method for Training Models on Long Videos",
                        "abstract": "Convolutional video models have an order of magnitude larger computational complexity than their counterpart image-level models. Constrained by computational resources, there is no model or training method that can train long video sequences end-to-end. Currently, the main-stream method is to split a raw video into clips, leading to incomplete fragmentary temporal information flow. Inspired by natural language processing techniques dealing with long sentences, we propose to treat videos as serial fragments satisfying Markov property, and train it as a whole by progressively propagating information through the temporal dimension in multiple steps. This progressive training (PGT) method is able to train long videos end-to-end with limited resources and ensures the effective transmission of information. As a general and robust training method, we empirically demonstrate that it yields significant performance improvements on different models and datasets. As an illustrative example, the proposed method improves SlowOnly network by 3.7 mAP on Charades and 1.9 top-1 accuracy on Kinetics with negligible parameter and computation overhead. Code is available at https://github.com/BoPang1996/PGT."
                    },
                    {
                        "title": "Reinforcement Learning for Adaptive Optimal Stationary Control of Linear Stochastic Systems",
                        "abstract": "This paper studies the adaptive optimal stationary control of continuous-time linear stochastic systems with both additive and multiplicative noises, using reinforcement learning techniques. Based on policy iteration, a novel off-policy reinforcement learning algorithm, named optimistic least-squares-based policy iteration, is proposed which is able to find iteratively near-optimal policies of the adaptive optimal stationary control problem directly from input/state data without explicitly identifying any system matrices, starting from an initial admissible control policy. The solutions given by the proposed optimistic least-squares-based policy iteration are proved to converge to a small neighborhood of the optimal solution with probability one, under mild conditions. The application of the proposed algorithm to a triple inverted pendulum example validates its feasibility and effectiveness."
                    },
                    {
                        "title": "Latent Space Energy-Based Model of Symbol-Vector Coupling for Text Generation and Classification",
                        "abstract": "We propose a latent space energy-based prior model for text generation and classification. The model stands on a generator network that generates the text sequence based on a continuous latent vector. The energy term of the prior model couples a continuous latent vector and a symbolic one-hot vector, so that discrete category can be inferred from the observed example based on the continuous latent vector. Such a latent space coupling naturally enables incorporation of information bottleneck regularization to encourage the continuous latent vector to extract information from the observed example that is informative of the underlying category. In our learning method, the symbol-vector coupling, the generator network and the inference network are learned jointly. Our model can be learned in an unsupervised setting where no category labels are provided. It can also be learned in semi-supervised setting where category labels are provided for a subset of training examples. Our experiments demonstrate that the proposed model learns well-structured and meaningful latent space, which (1) guides the generator to generate text with high quality, diversity, and interpretability, and (2) effectively classifies text."
                    },
                    {
                        "title": "Thumbs up? Sentiment Classification using Machine Learning Techniques",
                        "abstract": "We consider the problem of classifying documents not by topic, but by overall sentiment, e.g., determining whether a review is positive or negative. Using movie reviews as data, we find that standard machine learning techniques definitively outperform human-produced baselines. However, the three machine learning methods we employed (Naive Bayes, maximum entropy classification, and support vector machines) do not perform as well on sentiment classification as on traditional topic-based categorization. We conclude by examining factors that make the sentiment classification problem more challenging."
                    },
                    {
                        "title": "An Analysis of Structured Data on the Web",
                        "abstract": "In this paper, we analyze the nature and distribution of structured data on the Web. Web-scale information extraction, or the problem of creating structured tables using extraction from the entire web, is gathering lots of research interest. We perform a study to understand and quantify the value of Web-scale extraction, and how structured information is distributed amongst top aggregator websites and tail sites for various interesting domains. We believe this is the first study of its kind, and gives us new insights for information extraction over the Web."
                    },
                    {
                        "title": "Robust Policy Iteration for Continuous-time Linear Quadratic Regulation",
                        "abstract": "This paper studies the robustness of policy iteration in the context of continuous-time infinite-horizon linear quadratic regulation (LQR) problem. It is shown that Kleinman's policy iteration algorithm is inherently robust to small disturbances and enjoys local input-to-state stability in the sense of Sontag. More precisely, whenever the disturbance-induced input term in each iteration is bounded and small, the solutions of the policy iteration algorithm are also bounded and enter a small neighborhood of the optimal solution of the LQR problem. Based on this result, an off-policy data-driven policy iteration algorithm for the LQR problem is shown to be robust when the system dynamics are subjected to small additive unknown bounded disturbances. The theoretical results are validated by a numerical example."
                    },
                    {
                        "title": "Unsupervised Visual Representation Learning by Synchronous Momentum Grouping",
                        "abstract": "In this paper, we propose a genuine group-level contrastive visual representation learning method whose linear evaluation performance on ImageNet surpasses the vanilla supervised learning. Two mainstream unsupervised learning schemes are the instance-level contrastive framework and clustering-based schemes. The former adopts the extremely fine-grained instance-level discrimination whose supervisory signal is not efficient due to the false negatives. Though the latter solves this, they commonly come with some restrictions affecting the performance. To integrate their advantages, we design the SMoG method. SMoG follows the framework of contrastive learning but replaces the contrastive unit from instance to group, mimicking clustering-based methods. To achieve this, we propose the momentum grouping scheme which synchronously conducts feature grouping with representation learning. In this way, SMoG solves the problem of supervisory signal hysteresis which the clustering-based method usually faces, and reduces the false negatives of instance contrastive methods. We conduct exhaustive experiments to show that SMoG works well on both CNN and Transformer backbones. Results prove that SMoG has surpassed the current SOTA unsupervised representation learning methods. Moreover, its linear evaluation results surpass the performances obtained by vanilla supervised learning and the representation can be well transferred to downstream tasks."
                    },
                    {
                        "title": "Get out the vote: Determining support or opposition from Congressional floor-debate transcripts",
                        "abstract": "We investigate whether one can determine from the transcripts of U.S. Congressional floor debates whether the speeches represent support of or opposition to proposed legislation. To address this problem, we exploit the fact that these speeches occur as part of a discussion; this allows us to use sources of information regarding relationships between discourse segments, such as whether a given utterance indicates agreement with the opinion expressed by another. We find that the incorporation of such information yields substantial improvements over classifying speeches in isolation."
                    },
                    {
                        "title": "The effect of wording on message propagation: Topic- and author-controlled natural experiments on Twitter",
                        "abstract": "Consider a person trying to spread an important message on a social network. He/she can spend hours trying to craft the message. Does it actually matter? While there has been extensive prior work looking into predicting popularity of social-media content, the effect of wording per se has rarely been studied since it is often confounded with the popularity of the author and the topic. To control for these confounding factors, we take advantage of the surprising fact that there are many pairs of tweets containing the same url and written by the same user but employing different wording. Given such pairs, we ask: which version attracts more retweets? This turns out to be a more difficult task than predicting popular topics. Still, humans can answer this question better than chance (but far from perfectly), and the computational methods we develop can do better than both an average human and a strong competing method trained on non-controlled data."
                    },
                    {
                        "title": "Human Action Adverb Recognition: ADHA Dataset and A Three-Stream Hybrid Model",
                        "abstract": "We introduce the first benchmark for a new problem --- recognizing human action adverbs (HAA): \"Adverbs Describing Human Actions\" (ADHA). This is the first step for computer vision to change over from pattern recognition to real AI. We demonstrate some key features of ADHA: a semantically complete set of adverbs describing human actions, a set of common, describable human actions, and an exhaustive labeling of simultaneously emerging actions in each video. We commit an in-depth analysis on the implementation of current effective models in action recognition and image captioning on adverb recognition, and the results show that such methods are unsatisfactory. Moreover, we propose a novel three-stream hybrid model to deal the HAA problem, which achieves a better result."
                    },
                    {
                        "title": "Decoupled Box Proposal and Featurization with Ultrafine-Grained Semantic Labels Improve Image Captioning and Visual Question Answering",
                        "abstract": "Object detection plays an important role in current solutions to vision and language tasks like image captioning and visual question answering. However, popular models like Faster R-CNN rely on a costly process of annotating ground-truths for both the bounding boxes and their corresponding semantic labels, making it less amenable as a primitive task for transfer learning. In this paper, we examine the effect of decoupling box proposal and featurization for down-stream tasks. The key insight is that this allows us to leverage a large amount of labeled annotations that were previously unavailable for standard object detection benchmarks. Empirically, we demonstrate that this leads to effective transfer learning and improved image captioning and visual question answering models, as measured on publicly available benchmarks."
                    },
                    {
                        "title": "A Case Study on Combining ASR and Visual Features for Generating Instructional Video Captions",
                        "abstract": "Instructional videos get high-traffic on video sharing platforms, and prior work suggests that providing time-stamped, subtask annotations (e.g., \"heat the oil in the pan\") improves user experiences. However, current automatic annotation methods based on visual features alone perform only slightly better than constant prediction. Taking cues from prior work, we show that we can improve performance significantly by considering automatic speech recognition (ASR) tokens as input. Furthermore, jointly modeling ASR tokens and visual features results in higher performance compared to training individually on either modality. We find that unstated background information is better explained by visual features, whereas fine-grained distinctions (e.g., \"add oil\" vs. \"add olive oil\") are disambiguated more easily via ASR tokens."
                    }
                ]
            },
            "cd560418-0d33-49c4-b06d-0d00003abae5": {
                "pk": "cd560418-0d33-49c4-b06d-0d00003abae5",
                "name": "Jianwen Xie",
                "collaborators": [
                    "Ping Li",
                    "Yaxuan Zhu",
                    "Jing Zhang",
                    "Nick Barnes",
                    "Ying Nian Wu",
                    "Zilong Zheng",
                    "Song-Chun Zhu",
                    "Jun Li",
                    "Yang Zhao",
                    "Zongsheng Yue"
                ],
                "domain": [
                    "Generative Models",
                    "Energy-Based Models",
                    "Machine Learning",
                    "Video Processing"
                ],
                "publications": [
                    {
                        "title": "Learning Energy-Based Model with Variational Auto-Encoder as Amortized Sampler",
                        "abstract": "Due to the intractable partition function, training energy-based models (EBMs) by maximum likelihood requires Markov chain Monte Carlo (MCMC) sampling to approximate the gradient of the Kullback-Leibler divergence between data and model distributions. However, it is non-trivial to sample from an EBM because of the difficulty of mixing between modes. In this paper, we propose to learn a variational auto-encoder (VAE) to initialize the finite-step MCMC, such as Langevin dynamics that is derived from the energy function, for efficient amortized sampling of the EBM. With these amortized MCMC samples, the EBM can be trained by maximum likelihood, which follows an \"analysis by synthesis\" scheme; while the VAE learns from these MCMC samples via variational Bayes. We call this joint training algorithm the variational MCMC teaching, in which the VAE chases the EBM toward data distribution. We interpret the learning algorithm as a dynamic alternating projection in the context of information geometry. Our proposed models can generate samples comparable to GANs and EBMs. Additionally, we demonstrate that our model can learn effective probabilistic distribution toward supervised conditional learning tasks."
                    },
                    {
                        "title": "A Tale of Two Flows: Cooperative Learning of Langevin Flow and Normalizing Flow Toward Energy-Based Model",
                        "abstract": "This paper studies the cooperative learning of two generative flow models, in which the two models are iteratively updated based on the jointly synthesized examples. The first flow model is a normalizing flow that transforms an initial simple density to a target density by applying a sequence of invertible transformations. The second flow model is a Langevin flow that runs finite steps of gradient-based MCMC toward an energy-based model. We start from proposing a generative framework that trains an energy-based model with a normalizing flow as an amortized sampler to initialize the MCMC chains of the energy-based model. In each learning iteration, we generate synthesized examples by using a normalizing flow initialization followed by a short-run Langevin flow revision toward the current energy-based model. Then we treat the synthesized examples as fair samples from the energy-based model and update the model parameters with the maximum likelihood learning gradient, while the normalizing flow directly learns from the synthesized examples by maximizing the tractable likelihood. Under the short-run non-mixing MCMC scenario, the estimation of the energy-based model is shown to follow the perturbation of maximum likelihood, and the short-run Langevin flow and the normalizing flow form a two-flow generator that we call CoopFlow. We provide an understating of the CoopFlow algorithm by information geometry and show that it is a valid generator as it converges to a moment matching estimator. We demonstrate that the trained CoopFlow is capable of synthesizing realistic images, reconstructing images, and interpolating between images."
                    },
                    {
                        "title": "CoopInit: Initializing Generative Adversarial Networks via Cooperative Learning",
                        "abstract": "Numerous research efforts have been made to stabilize the training of the Generative Adversarial Networks (GANs), such as through regularization and architecture design. However, we identify the instability can also arise from the fragile balance at the early stage of adversarial learning. This paper proposes the CoopInit, a simple yet effective cooperative learning-based initialization strategy that can quickly learn a good starting point for GANs, with a very small computation overhead during training. The proposed algorithm consists of two learning stages: (i) Cooperative initialization stage: The discriminator of GAN is treated as an energy-based model (EBM) and is optimized via maximum likelihood estimation (MLE), with the help of the GAN's generator to provide synthetic data to approximate the learning gradients. The EBM also guides the MLE learning of the generator via MCMC teaching; (ii) Adversarial finalization stage: After a few iterations of initialization, the algorithm seamlessly transits to the regular mini-max adversarial training until convergence. The motivation is that the MLE-based initialization stage drives the model towards mode coverage, which is helpful in alleviating the issue of mode dropping during the adversarial learning stage. We demonstrate the effectiveness of the proposed approach on image generation and one-sided unpaired image-to-image translation tasks through extensive experiments."
                    },
                    {
                        "title": "Learning Generative Vision Transformer with Energy-Based Latent Space for Saliency Prediction",
                        "abstract": "Vision transformer networks have shown superiority in many computer vision tasks. In this paper, we take a step further by proposing a novel generative vision transformer with latent variables following an informative energy-based prior for salient object detection. Both the vision transformer network and the energy-based prior model are jointly trained via Markov chain Monte Carlo-based maximum likelihood estimation, in which the sampling from the intractable posterior and prior distributions of the latent variables are performed by Langevin dynamics. Further, with the generative vision transformer, we can easily obtain a pixel-wise uncertainty map from an image, which indicates the model confidence in predicting saliency from the image. Different from the existing generative models which define the prior distribution of the latent variables as a simple isotropic Gaussian distribution, our model uses an energy-based informative prior which can be more expressive to capture the latent space of the data. We apply the proposed framework to both RGB and RGB-D salient object detection tasks. Extensive experimental results show that our framework can achieve not only accurate saliency predictions but also meaningful uncertainty maps that are consistent with the human perception."
                    },
                    {
                        "title": "Likelihood-Based Generative Radiance Field with Latent Space Energy-Based Model for 3D-Aware Disentangled Image Representation",
                        "abstract": "We propose the NeRF-LEBM, a likelihood-based top-down 3D-aware 2D image generative model that incorporates 3D representation via Neural Radiance Fields (NeRF) and 2D imaging process via differentiable volume rendering. The model represents an image as a rendering process from 3D object to 2D image and is conditioned on some latent variables that account for object characteristics and are assumed to follow informative trainable energy-based prior models. We propose two likelihood-based learning frameworks to train the NeRF-LEBM: (i) maximum likelihood estimation with Markov chain Monte Carlo-based inference and (ii) variational inference with the reparameterization trick. We study our models in the scenarios with both known and unknown camera poses. Experiments on several benchmark datasets demonstrate that the NeRF-LEBM can infer 3D object structures from 2D images, generate 2D images with novel views and objects, learn from incomplete 2D images, and learn from 2D images with known or unknown camera poses."
                    },
                    {
                        "title": "Learning Noise-Aware Encoder-Decoder from Noisy Labels by Alternating Back-Propagation for Saliency Detection",
                        "abstract": "In this paper, we propose a noise-aware encoder-decoder framework to disentangle a clean saliency predictor from noisy training examples, where the noisy labels are generated by unsupervised handcrafted feature-based methods. The proposed model consists of two sub-models parameterized by neural networks: (1) a saliency predictor that maps input images to clean saliency maps, and (2) a noise generator, which is a latent variable model that produces noises from Gaussian latent vectors. The whole model that represents noisy labels is a sum of the two sub-models. The goal of training the model is to estimate the parameters of both sub-models, and simultaneously infer the corresponding latent vector of each noisy label. We propose to train the model by using an alternating back-propagation (ABP) algorithm, which alternates the following two steps: (1) learning back-propagation for estimating the parameters of two sub-models by gradient ascent, and (2) inferential back-propagation for inferring the latent vectors of training noisy examples by Langevin Dynamics. To prevent the network from converging to trivial solutions, we utilize an edge-aware smoothness loss to regularize hidden saliency maps to have similar structures as their corresponding images. Experimental results on several benchmark datasets indicate the effectiveness of the proposed model."
                    },
                    {
                        "title": "Semi-Supervised Video Deraining with Dynamical Rain Generator",
                        "abstract": "While deep learning (DL)-based video deraining methods have achieved significant success recently, they still exist two major drawbacks. Firstly, most of them do not sufficiently model the characteristics of rain layers of rainy videos. In fact, the rain layers exhibit strong physical properties (e.g., direction, scale and thickness) in spatial dimension and natural continuities in temporal dimension, and thus can be generally modelled by the spatial-temporal process in statistics. Secondly, current DL-based methods seriously depend on the labeled synthetic training data, whose rain types are always deviated from those in unlabeled real data. Such gap between synthetic and real data sets leads to poor performance when applying them in real scenarios. Against these issues, this paper proposes a new semi-supervised video deraining method, in which a dynamic rain generator is employed to fit the rain layer, expecting to better depict its insightful characteristics. Specifically, such dynamic generator consists of one emission model and one transition model to simultaneously encode the spatially physical structure and temporally continuous changes of rain streaks, respectively, which both are parameterized as deep neural networks (DNNs). Further more, different prior formats are designed for the labeled synthetic and unlabeled real data, so as to fully exploit the common knowledge underlying them. Last but not least, we also design a Monte Carlo EM algorithm to solve this model. Extensive experiments are conducted to verify the superiorities of the proposed semi-supervised deraining model."
                    },
                    {
                        "title": "Learning Energy-Based Models by Cooperative Diffusion Recovery Likelihood",
                        "abstract": "Training energy-based models (EBMs) on high-dimensional data can be both challenging and time-consuming, and there exists a noticeable gap in sample quality between EBMs and other generative frameworks like GANs and diffusion models. To close this gap, inspired by the recent efforts of learning EBMs by maximizing diffusion recovery likelihood (DRL), we propose cooperative diffusion recovery likelihood (CDRL), an effective approach to tractably learn and sample from a series of EBMs defined on increasingly noisy versions of a dataset, paired with an initializer model for each EBM. At each noise level, the two models are jointly estimated within a cooperative training framework: samples from the initializer serve as starting points that are refined by a few MCMC sampling steps from the EBM. The EBM is then optimized by maximizing recovery likelihood, while the initializer model is optimized by learning from the difference between the refined samples and the initial samples. In addition, we made several practical designs for EBM training to further improve the sample quality. Combining these advances, our approach significantly boost the generation performance compared to existing EBM methods on CIFAR-10 and ImageNet datasets. We also demonstrate the effectiveness of our models for several downstream tasks, including classifier-free guided generation, compositional generation, image inpainting and out-of-distribution detection."
                    },
                    {
                        "title": "STANLEY: Stochastic Gradient Anisotropic Langevin Dynamics for Learning Energy-Based Models",
                        "abstract": "We propose in this paper, STANLEY, a STochastic gradient ANisotropic LangEvin dYnamics, for sampling high dimensional data. With the growing efficacy and potential of Energy-Based modeling, also known as non-normalized probabilistic modeling, for modeling a generative process of different natures of high dimensional data observations, we present an end-to-end learning algorithm for Energy-Based models (EBM) with the purpose of improving the quality of the resulting sampled data points. While the unknown normalizing constant of EBMs makes the training procedure intractable, resorting to Markov Chain Monte Carlo (MCMC) is in general a viable option. Realizing what MCMC entails for the EBM training, we propose in this paper, a novel high dimensional sampling method, based on an anisotropic stepsize and a gradient-informed covariance matrix, embedded into a discretized Langevin diffusion. We motivate the necessity for an anisotropic update of the negative samples in the Markov Chain by the nonlinearity of the backbone of the EBM, here a Convolutional Neural Network. Our resulting method, namely STANLEY, is an optimization algorithm for training Energy-Based models via our newly introduced MCMC method. We provide a theoretical understanding of our sampling scheme by proving that the sampler leads to a geometrically uniformly ergodic Markov Chain. Several image generation experiments are provided in our paper to show the effectiveness of our method."
                    },
                    {
                        "title": "An Energy-Based Prior for Generative Saliency",
                        "abstract": "We propose a novel generative saliency prediction framework that adopts an informative energy-based model as a prior distribution. The energy-based prior model is defined on the latent space of a saliency generator network that generates the saliency map based on a continuous latent variables and an observed image. Both the parameters of saliency generator and the energy-based prior are jointly trained via Markov chain Monte Carlo-based maximum likelihood estimation, in which the sampling from the intractable posterior and prior distributions of the latent variables are performed by Langevin dynamics. With the generative saliency model, we can obtain a pixel-wise uncertainty map from an image, indicating model confidence in the saliency prediction. Different from existing generative models, which define the prior distribution of the latent variables as a simple isotropic Gaussian distribution, our model uses an energy-based informative prior which can be more expressive in capturing the latent space of the data. With the informative energy-based prior, we extend the Gaussian distribution assumption of generative models to achieve a more representative distribution of the latent space, leading to more reliable uncertainty estimation. We apply the proposed frameworks to both RGB and RGB-D salient object detection tasks with both transformer and convolutional neural network backbones. We further propose an adversarial learning algorithm and a variational inference algorithm as alternatives to train the proposed generative framework. Experimental results show that our generative saliency model with an energy-based prior can achieve not only accurate saliency predictions but also reliable uncertainty maps that are consistent with human perception. Results and code are available at \\url{https://github.com/JingZhang617/EBMGSOD}."
                    },
                    {
                        "title": "Synthesizing Dynamic Patterns by Spatial-Temporal Generative ConvNet",
                        "abstract": "Video sequences contain rich dynamic patterns, such as dynamic texture patterns that exhibit stationarity in the temporal domain, and action patterns that are non-stationary in either spatial or temporal domain. We show that a spatial-temporal generative ConvNet can be used to model and synthesize dynamic patterns. The model defines a probability distribution on the video sequence, and the log probability is defined by a spatial-temporal ConvNet that consists of multiple layers of spatial-temporal filters to capture spatial-temporal patterns of different scales. The model can be learned from the training video sequences by an \"analysis by synthesis\" learning algorithm that iterates the following two steps. Step 1 synthesizes video sequences from the currently learned model. Step 2 then updates the model parameters based on the difference between the synthesized video sequences and the observed training sequences. We show that the learning algorithm can synthesize realistic dynamic patterns."
                    },
                    {
                        "title": "Learning Energy-based Spatial-Temporal Generative ConvNets for Dynamic Patterns",
                        "abstract": "Video sequences contain rich dynamic patterns, such as dynamic texture patterns that exhibit stationarity in the temporal domain, and action patterns that are non-stationary in either spatial or temporal domain. We show that an energy-based spatial-temporal generative ConvNet can be used to model and synthesize dynamic patterns. The model defines a probability distribution on the video sequence, and the log probability is defined by a spatial-temporal ConvNet that consists of multiple layers of spatial-temporal filters to capture spatial-temporal patterns of different scales. The model can be learned from the training video sequences by an \"analysis by synthesis\" learning algorithm that iterates the following two steps. Step 1 synthesizes video sequences from the currently learned model. Step 2 then updates the model parameters based on the difference between the synthesized video sequences and the observed training sequences. We show that the learning algorithm can synthesize realistic dynamic patterns. We also show that it is possible to learn the model from incomplete training sequences with either occluded pixels or missing frames, so that model learning and pattern completion can be accomplished simultaneously."
                    },
                    {
                        "title": "Energy-Based Generative Cooperative Saliency Prediction",
                        "abstract": "Conventional saliency prediction models typically learn a deterministic mapping from an image to its saliency map, and thus fail to explain the subjective nature of human attention. In this paper, to model the uncertainty of visual saliency, we study the saliency prediction problem from the perspective of generative models by learning a conditional probability distribution over the saliency map given an input image, and treating the saliency prediction as a sampling process from the learned distribution. Specifically, we propose a generative cooperative saliency prediction framework, where a conditional latent variable model (LVM) and a conditional energy-based model (EBM) are jointly trained to predict salient objects in a cooperative manner. The LVM serves as a fast but coarse predictor to efficiently produce an initial saliency map, which is then refined by the iterative Langevin revision of the EBM that serves as a slow but fine predictor. Such a coarse-to-fine cooperative saliency prediction strategy offers the best of both worlds. Moreover, we propose a \"cooperative learning while recovering\" strategy and apply it to weakly supervised saliency prediction, where saliency annotations of training images are partially observed. Lastly, we find that the learned energy function in the EBM can serve as a refinement module that can refine the results of other pre-trained saliency prediction models. Experimental results show that our model can produce a set of diverse and plausible saliency maps of an image, and obtain state-of-the-art performance in both fully supervised and weakly supervised saliency prediction tasks."
                    },
                    {
                        "title": "A Tale of Two Latent Flows: Learning Latent Space Normalizing Flow with Short-run Langevin Flow for Approximate Inference",
                        "abstract": "We study a normalizing flow in the latent space of a top-down generator model, in which the normalizing flow model plays the role of the informative prior model of the generator. We propose to jointly learn the latent space normalizing flow prior model and the top-down generator model by a Markov chain Monte Carlo (MCMC)-based maximum likelihood algorithm, where a short-run Langevin sampling from the intractable posterior distribution is performed to infer the latent variables for each observed example, so that the parameters of the normalizing flow prior and the generator can be updated with the inferred latent variables. We show that, under the scenario of non-convergent short-run MCMC, the finite step Langevin dynamics is a flow-like approximate inference model and the learning objective actually follows the perturbation of the maximum likelihood estimation (MLE). We further point out that the learning framework seeks to (i) match the latent space normalizing flow and the aggregated posterior produced by the short-run Langevin flow, and (ii) bias the model from MLE such that the short-run Langevin flow inference is close to the true posterior. Empirical results of extensive experiments validate the effectiveness of the proposed latent space normalizing flow model in the tasks of image generation, image reconstruction, anomaly detection, supervised image inpainting and unsupervised image recovery."
                    },
                    {
                        "title": "Super-Trajectory for Video Segmentation",
                        "abstract": "We introduce a novel semi-supervised video segmentation approach based on an efficient video representation, called as \"super-trajectory\". Each super-trajectory corresponds to a group of compact trajectories that exhibit consistent motion patterns, similar appearance and close spatiotemporal relationships. We generate trajectories using a probabilistic model, which handles occlusions and drifts in a robust and natural way. To reliably group trajectories, we adopt a modified version of the density peaks based clustering algorithm that allows capturing rich spatiotemporal relations among trajectories in the clustering process. The presented video representation is discriminative enough to accurately propagate the initial annotations in the first frame onto the remaining video frames. Extensive experimental analysis on challenging benchmarks demonstrate our method is capable of distinguishing the target objects from complex backgrounds and even reidentifying them after long-term occlusions."
                    },
                    {
                        "title": "Learning Feature-to-Feature Translator by Alternating Back-Propagation for Generative Zero-Shot Learning",
                        "abstract": "We investigate learning feature-to-feature translator networks by alternating back-propagation as a general-purpose solution to zero-shot learning (ZSL) problems. It is a generative model-based ZSL framework. In contrast to models based on generative adversarial networks (GAN) or variational autoencoders (VAE) that require auxiliary networks to assist the training, our model consists of a single conditional generator that maps class-level semantic features and Gaussian white noise vector accounting for instance-level latent factors to visual features, and is trained by maximum likelihood estimation. The training process is a simple yet effective alternating back-propagation process that iterates the following two steps: (i) the inferential back-propagation to infer the latent factors of each observed example, and (ii) the learning back-propagation to update the model parameters. We show that, with slight modifications, our model is capable of learning from incomplete visual features for ZSL. We conduct extensive comparisons with existing generative ZSL methods on five benchmarks, demonstrating the superiority of our method in not only ZSL performance but also convergence speed and computational cost. Specifically, our model outperforms the existing state-of-the-art methods by a remarkable margin up to 3.1% and 4.0% in ZSL and generalized ZSL settings, respectively."
                    },
                    {
                        "title": "Molecule Design by Latent Prompt Transformer",
                        "abstract": "This paper proposes a latent prompt Transformer model for solving challenging optimization problems such as molecule design, where the goal is to find molecules with optimal values of a target chemical or biological property that can be computed by an existing software. Our proposed model consists of three components. (1) A latent vector whose prior distribution is modeled by a Unet transformation of a Gaussian white noise vector. (2) A molecule generation model that generates the string-based representation of molecule conditional on the latent vector in (1). We adopt the causal Transformer model that takes the latent vector in (1) as prompt. (3) A property prediction model that predicts the value of the target property of a molecule based on a non-linear regression on the latent vector in (1). We call the proposed model the latent prompt Transformer model. After initial training of the model on existing molecules and their property values, we then gradually shift the model distribution towards the region that supports desired values of the target property for the purpose of molecule design. Our experiments show that our proposed model achieves state of the art performances on several benchmark molecule design tasks."
                    }
                ]
            },
            "a590d83d-4e92-4004-83e2-a8ec0d841f63": {
                "pk": "a590d83d-4e92-4004-83e2-a8ec0d841f63",
                "name": "Andrew Lizarraga",
                "collaborators": [
                    "Shantanu H. Joshi",
                    "Elvis Nunez",
                    "Ying Nian Wu",
                    "Eric H. Jiang",
                    "Katherine L. Narr",
                    "Kirsten A. Donald",
                    "Brandon Taraku",
                    "Edouardo Honig",
                    "David Lee",
                    "Antoni Kubicki"
                ],
                "domain": [
                    "Deep Learning",
                    "Reinforcement Learning",
                    "Geometric Data Analysis",
                    "Autoencoder"
                ],
                "publications": [
                    {
                        "title": "SrvfNet: A Generative Network for Unsupervised Multiple Diffeomorphic Shape Alignment",
                        "abstract": "We present SrvfNet, a generative deep learning framework for the joint multiple alignment of large collections of functional data comprising square-root velocity functions (SRVF) to their templates. Our proposed framework is fully unsupervised and is capable of aligning to a predefined template as well as jointly predicting an optimal template from data while simultaneously achieving alignment. Our network is constructed as a generative encoder-decoder architecture comprising fully-connected layers capable of producing a distribution space of the warping functions. We demonstrate the strength of our framework by validating it on synthetic data as well as diffusion profiles from magnetic resonance imaging (MRI) data."
                    },
                    {
                        "title": "SDSRA: A Skill-Driven Skill-Recombination Algorithm for Efficient Policy Learning",
                        "abstract": "In this paper, we introduce a novel algorithm - the Skill-Driven Skill Recombination Algorithm (SDSRA) - an innovative framework that significantly enhances the efficiency of achieving maximum entropy in reinforcement learning tasks. We find that SDSRA achieves faster convergence compared to the traditional Soft Actor-Critic (SAC) algorithm and produces improved policies. By integrating skill-based strategies within the robust Actor-Critic framework, SDSRA demonstrates remarkable adaptability and performance across a wide array of complex and diverse benchmarks."
                    },
                    {
                        "title": "StreamNet: A WAE for White Matter Streamline Analysis",
                        "abstract": "We present StreamNet, an autoencoder architecture for the analysis of the highly heterogeneous geometry of large collections of white matter streamlines. This proposed framework takes advantage of geometry-preserving properties of the Wasserstein-1 metric in order to achieve direct encoding and reconstruction of entire bundles of streamlines. We show that the model not only accurately captures the distributive structures of streamlines in the population, but is also able to achieve superior reconstruction performance between real and synthetic streamlines. Experimental model performance is evaluated on white matter streamlines resulting from T1-weighted diffusion imaging of 40 healthy controls using recent state of the art bundle comparison metric that measures fiber-shape similarities."
                    },
                    {
                        "title": "Differentiable VQ-VAE's for Robust White Matter Streamline Encodings",
                        "abstract": "Given the complex geometry of white matter streamlines, Autoencoders have been proposed as a dimension-reduction tool to simplify the analysis streamlines in a low-dimensional latent spaces. However, despite these recent successes, the majority of encoder architectures only perform dimension reduction on single streamlines as opposed to a full bundle of streamlines. This is a severe limitation of the encoder architecture that completely disregards the global geometric structure of streamlines at the expense of individual fibers. Moreover, the latent space may not be well structured which leads to doubt into their interpretability. In this paper we propose a novel Differentiable Vector Quantized Variational Autoencoder, which are engineered to ingest entire bundles of streamlines as single data-point and provides reliable trustworthy encodings that can then be later used to analyze streamlines in the latent space. Comparisons with several state of the art Autoencoders demonstrate superior performance in both encoding and synthesis."
                    },
                    {
                        "title": "Alignment of Tractography Streamlines using Deformation Transfer via Parallel Transport",
                        "abstract": "We present a geometric framework for aligning white matter fiber tracts. By registering fiber tracts between brains, one expects to see overlap of anatomical structures that often provide meaningful comparisons across subjects. However, the geometry of white matter tracts is highly heterogeneous, and finding direct tract-correspondence across multiple individuals remains a challenging problem. We present a novel deformation metric between tracts that allows one to compare tracts while simultaneously obtaining a registration. To accomplish this, fiber tracts are represented by an intrinsic mean along with the deformation fields represented by tangent vectors from the mean. In this setting, one can determine a parallel transport between tracts and then register corresponding tangent vectors. We present the results of bundle alignment on a population of 43 healthy adult subjects."
                    },
                    {
                        "title": "DODT: Enhanced Online Decision Transformer Learning through Dreamer's Actor-Critic Trajectory Forecasting",
                        "abstract": "Advancements in reinforcement learning have led to the development of sophisticated models capable of learning complex decision-making tasks. However, efficiently integrating world models with decision transformers remains a challenge. In this paper, we introduce a novel approach that combines the Dreamer algorithm's ability to generate anticipatory trajectories with the adaptive learning strengths of the Online Decision Transformer. Our methodology enables parallel training where Dreamer-produced trajectories enhance the contextual decision-making of the transformer, creating a bidirectional enhancement loop. We empirically demonstrate the efficacy of our approach on a suite of challenging benchmarks, achieving notable improvements in sample efficiency and reward maximization over existing methods. Our results indicate that the proposed integrated framework not only accelerates learning but also showcases robustness in diverse and dynamic scenarios, marking a significant step forward in model-based reinforcement learning."
                    }
                ]
            },
            "068e4be5-e849-40df-8ce3-52fd603ceb1f": {
                "pk": "068e4be5-e849-40df-8ce3-52fd603ceb1f",
                "name": "Yuhao Huang",
                "collaborators": [
                    "Corey Keller",
                    "Gangrong Qu",
                    "Youran Ge",
                    "Yicheng Zeng",
                    "Xiaobin Xiong"
                ],
                "domain": [
                    "Physics-Informed Neural Networks",
                    "Bipedal Robotics",
                    "Video Compressed Sensing",
                    "Brain Connectivity"
                ],
                "publications": [
                    {
                        "title": "Parallel Physics-Informed Neural Networks with Bidirectional Balance",
                        "abstract": "As an emerging technology in deep learning, physics-informed neural networks (PINNs) have been widely used to solve various partial differential equations (PDEs) in engineering. However, PDEs based on practical considerations contain multiple physical quantities and complex initial boundary conditions, thus PINNs often returns incorrect results. Here we take heat transfer problem in multilayer fabrics as a typical example. It is coupled by multiple temperature fields with strong correlation, and the values of variables are extremely unbalanced among different dimensions. We clarify the potential difficulties of solving such problems by classic PINNs, and propose a parallel physics-informed neural networks with bidirectional balance. In detail, our parallel solving framework synchronously fits coupled equations through several multilayer perceptions. Moreover, we design two modules to balance forward process of data and back-propagation process of loss gradient. This bidirectional balance not only enables the whole network to converge stably, but also helps to fully learn various physical conditions in PDEs. We provide a series of ablation experiments to verify the effectiveness of the proposed methods. The results show that our approach makes the PINNs unsolvable problem solvable, and achieves excellent solving accuracy."
                    },
                    {
                        "title": "Improved Fast Iterative Algorithm for Eikonal Equation for GPU Computing",
                        "abstract": "In this paper we propose an improved fast iterative method to solve the Eikonal equation, which can be implemented in parallel. We improve the fast iterative method for Eikonal equation in two novel ways, in the value update and in the error correction. The new value update is very similar to the fast iterative method in that we selectively update the points, chosen by a convergence measure, in the active list. However, in order to reduce running time, the improved algorithm does not run a convergence check of the neighboring points of the narrow band as the fast iterative method usually does. The additional error correction step is to correct the errors that the previous value update step may cause. The error correction step consists of finding and recalculating the point values in a separate remedy list which is quite easy to implement on a GPU. In contrast to the fast marching method and the fast sweeping method for the Eikonal equation, our improved method does not need to compute the solution with any special ordering in neither the remedy list nor the active list. Therefore, our algorithm can be implemented in parallel. In our experiments, we implemente our new algorithm in parallel on a GPU and compare the elapsed time with other current algorithms. The improved fast iterative method runs faster than the other algorithms in most cases through our numercal studies."
                    },
                    {
                        "title": "How can I investigate causal brain networks with iEEG?",
                        "abstract": "While many human imaging methodologies probe the structural and functional connectivity of the brain, techniques to investigate cortical networks in a causal and directional manner are critical but limited. The use of iEEG enables several approaches to directly characterize brain regions that are functionally connected and in some cases also establish directionality of these connections. In this chapter we focus on the basis, method and application of the cortico-cortical evoked potential (CCEP), whereby electrical pulses applied to one set of intracranial electrodes yields an electrically-induced brain response at local and remote regions. In this chapter, CCEPs are first contextualized within common brain connectivity methods used to define cortical networks and how CCEP adds unique information. Second, the practical and analytical considerations when using CCEP are discussed. Third, we review the neurophysiology underlying CCEPs and the applications of CCEPs including exploring functional and pathological brain networks and probing brain plasticity. Finally, we end with a discussion of limitations, caveats, and directions to improve CCEP utilization in the future."
                    },
                    {
                        "title": "Sampling-Priors-Augmented Deep Unfolding Network for Robust Video Compressive Sensing",
                        "abstract": "Video Compressed Sensing (VCS) aims to reconstruct multiple frames from one single captured measurement, thus achieving high-speed scene recording with a low-frame-rate sensor. Although there have been impressive advances in VCS recently, those state-of-the-art (SOTA) methods also significantly increase model complexity and suffer from poor generality and robustness, which means that those networks need to be retrained to accommodate the new system. Such limitations hinder the real-time imaging and practical deployment of models. In this work, we propose a Sampling-Priors-Augmented Deep Unfolding Network (SPA-DUN) for efficient and robust VCS reconstruction. Under the optimization-inspired deep unfolding framework, a lightweight and efficient U-net is exploited to downsize the model while improving overall performance. Moreover, the prior knowledge from the sampling model is utilized to dynamically modulate the network features to enable single SPA-DUN to handle arbitrary sampling settings, augmenting interpretability and generality. Extensive experiments on both simulation and real datasets demonstrate that SPA-DUN is not only applicable for various sampling settings with one single model but also achieves SOTA performance with incredible efficiency."
                    },
                    {
                        "title": "STRIDE: An Open-Source, Low-Cost, and Versatile Bipedal Robot Platform for Research and Education",
                        "abstract": "In this paper, we present STRIDE, a Simple, Terrestrial, Reconfigurable, Intelligent, Dynamic, and Educational bipedal platform. STRIDE aims to propel bipedal robotics research and education by providing a cost-effective implementation with step-by-step instructions for building a bipedal robotic platform while providing flexible customizations via a modular and durable design. Moreover, a versatile terrain setup and a quantitative disturbance injection system are augmented to the robot platform to replicate natural terrains and push forces that can be used to evaluate legged locomotion in practical and adversarial scenarios. We demonstrate the functionalities of this platform by realizing an adaptive step-to-step dynamics based walking controller to achieve dynamic walking. Our work with the open-soured implementation shows that STRIDE is a highly versatile and durable platform that can be used in research and education to evaluate locomotion algorithms, mechanical designs, and robust and adaptative controls."
                    }
                ]
            },
            "1c128e9a-e5a7-4fea-82e5-e1a7a8571ba4": {
                "pk": "1c128e9a-e5a7-4fea-82e5-e1a7a8571ba4",
                "name": "Sirui Xie",
                "collaborators": [
                    "Song-Chun Zhu",
                    "Chunxiao Liu",
                    "Ying Nian Wu",
                    "Jianping Shi",
                    "Yixin Zhu",
                    "Dahua Lin",
                    "Xiaojian Ma",
                    "Peiyu Yu",
                    "Hehui Zheng",
                    "Liang Lin"
                ],
                "domain": [
                    "Neural Architecture Search",
                    "Reinforcement Learning",
                    "Generative Modeling",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "SNAS: Stochastic Neural Architecture Search",
                        "abstract": "We propose Stochastic Neural Architecture Search (SNAS), an economical end-to-end solution to Neural Architecture Search (NAS) that trains neural operation parameters and architecture distribution parameters in same round of back-propagation, while maintaining the completeness and differentiability of the NAS pipeline. In this work, NAS is reformulated as an optimization problem on parameters of a joint distribution for the search space in a cell. To leverage the gradient information in generic differentiable loss for architecture search, a novel search gradient is proposed. We prove that this search gradient optimizes the same objective as reinforcement-learning-based NAS, but assigns credits to structural decisions more efficiently. This credit assignment is further augmented with locally decomposable reward to enforce a resource-efficient constraint. In experiments on CIFAR-10, SNAS takes less epochs to find a cell architecture with state-of-the-art accuracy than non-differentiable evolution-based and reinforcement-learning-based NAS, which is also transferable to ImageNet. It is also shown that child networks of SNAS can maintain the validation accuracy in searching, with which attention-based NAS requires parameter retraining to compete, exhibiting potentials to stride towards efficient NAS on big datasets. We have released our implementation at https://github.com/SNAS-Series/SNAS-Series."
                    },
                    {
                        "title": "Generalized Inverse Planning: Learning Lifted non-Markovian Utility for Generalizable Task Representation",
                        "abstract": "In searching for a generalizable representation of temporally extended tasks, we spot two necessary constituents: the utility needs to be non-Markovian to transfer temporal relations invariant to a probability shift, the utility also needs to be lifted to abstract out specific grounding objects. In this work, we study learning such utility from human demonstrations. While inverse reinforcement learning (IRL) has been accepted as a general framework of utility learning, its fundamental formulation is one concrete Markov Decision Process. Thus the learned reward function does not specify the task independently of the environment. Going beyond that, we define a domain of generalization that spans a set of planning problems following a schema. We hence propose a new quest, Generalized Inverse Planning, for utility learning in this domain. We further outline a computational framework, Maximum Entropy Inverse Planning (MEIP), that learns non-Markovian utility and associated concepts in a generative manner. The learned utility and concepts form a task representation that generalizes regardless of probability shift or structural change. Seeing that the proposed generalization problem has not been widely studied yet, we carefully define an evaluation protocol, with which we illustrate the effectiveness of MEIP on two proof-of-concept domains and one challenging task: learning to fold from demonstrations."
                    },
                    {
                        "title": "DSNAS: Direct Neural Architecture Search without Parameter Retraining",
                        "abstract": "If NAS methods are solutions, what is the problem? Most existing NAS methods require two-stage parameter optimization. However, performance of the same architecture in the two stages correlates poorly. In this work, we propose a new problem definition for NAS, task-specific end-to-end, based on this observation. We argue that given a computer vision task for which a NAS method is expected, this definition can reduce the vaguely-defined NAS evaluation to i) accuracy of this task and ii) the total computation consumed to finally obtain a model with satisfying accuracy. Seeing that most existing methods do not solve this problem directly, we propose DSNAS, an efficient differentiable NAS framework that simultaneously optimizes architecture and parameters with a low-biased Monte Carlo estimate. Child networks derived from DSNAS can be deployed directly without parameter retraining. Comparing with two-stage methods, DSNAS successfully discovers networks with comparable accuracy (74.4%) on ImageNet in 420 GPU hours, reducing the total time by more than 34%. Our implementation is available at https://github.com/SNAS-Series/SNAS-Series."
                    },
                    {
                        "title": "NADPEx: An on-policy temporally consistent exploration method for deep reinforcement learning",
                        "abstract": "Reinforcement learning agents need exploratory behaviors to escape from local optima. These behaviors may include both immediate dithering perturbation and temporally consistent exploration. To achieve these, a stochastic policy model that is inherently consistent through a period of time is in desire, especially for tasks with either sparse rewards or long term information. In this work, we introduce a novel on-policy temporally consistent exploration strategy - Neural Adaptive Dropout Policy Exploration (NADPEx) - for deep reinforcement learning agents. Modeled as a global random variable for conditional distribution, dropout is incorporated to reinforcement learning policies, equipping them with inherent temporal consistency, even when the reward signals are sparse. Two factors, gradients' alignment with the objective and KL constraint in policy space, are discussed to guarantee NADPEx policy's stable improvement. Our experiments demonstrate that NADPEx solves tasks with sparse reward while naive exploration and parameter noise fail. It yields as well or even faster convergence in the standard mujoco benchmark for continuous control."
                    },
                    {
                        "title": "Understanding the wiring evolution in differentiable neural architecture search",
                        "abstract": "Controversy exists on whether differentiable neural architecture search methods discover wiring topology effectively. To understand how wiring topology evolves, we study the underlying mechanism of several existing differentiable NAS frameworks. Our investigation is motivated by three observed searching patterns of differentiable NAS: 1) they search by growing instead of pruning; 2) wider networks are more preferred than deeper ones; 3) no edges are selected in bi-level optimization. To anatomize these phenomena, we propose a unified view on searching algorithms of existing frameworks, transferring the global optimization to local cost minimization. Based on this reformulation, we conduct empirical and theoretical analyses, revealing implicit inductive biases in the cost's assignment mechanism and evolution dynamics that cause the observed phenomena. These biases indicate strong discrimination towards certain topologies. To this end, we pose questions that future differentiable methods for neural wiring discovery need to confront, hoping to evoke a discussion and rethinking on how much bias has been enforced implicitly in existing NAS methods."
                    },
                    {
                        "title": "Learning non-Markovian Decision-Making from State-only Sequences",
                        "abstract": "Conventional imitation learning assumes access to the actions of demonstrators, but these motor signals are often non-observable in naturalistic settings. Additionally, sequential decision-making behaviors in these settings can deviate from the assumptions of a standard Markov Decision Process (MDP). To address these challenges, we explore deep generative modeling of state-only sequences with non-Markov Decision Process (nMDP), where the policy is an energy-based prior in the latent space of the state transition generator. We develop maximum likelihood estimation to achieve model-based imitation, which involves short-run MCMC sampling from the prior and importance sampling for the posterior. The learned model enables \\textit{decision-making as inference}: model-free policy execution is equivalent to prior sampling, model-based planning is posterior sampling initialized from the policy. We demonstrate the efficacy of the proposed method in a prototypical path planning task with non-Markovian constraints and show that the learned model exhibits strong performances in challenging domains from the MuJoCo suite."
                    },
                    {
                        "title": "Graph-guided Architecture Search for Real-time Semantic Segmentation",
                        "abstract": "Designing a lightweight semantic segmentation network often requires researchers to find a trade-off between performance and speed, which is always empirical due to the limited interpretability of neural networks. In order to release researchers from these tedious mechanical trials, we propose a Graph-guided Architecture Search (GAS) pipeline to automatically search real-time semantic segmentation networks. Unlike previous works that use a simplified search space and stack a repeatable cell to form a network, we introduce a novel search mechanism with new search space where a lightweight model can be effectively explored through the cell-level diversity and latencyoriented constraint. Specifically, to produce the cell-level diversity, the cell-sharing constraint is eliminated through the cell-independent manner. Then a graph convolution network (GCN) is seamlessly integrated as a communication mechanism between cells. Finally, a latency-oriented constraint is endowed into the search process to balance the speed and performance. Extensive experiments on Cityscapes and CamVid datasets demonstrate that GAS achieves the new state-of-the-art trade-off between accuracy and speed. In particular, on Cityscapes dataset, GAS achieves the new best performance of 73.5% mIoU with speed of 108.4 FPS on Titan Xp."
                    },
                    {
                        "title": "HALMA: Humanlike Abstraction Learning Meets Affordance in Rapid Problem Solving",
                        "abstract": "Humans learn compositional and causal abstraction, \\ie, knowledge, in response to the structure of naturalistic tasks. When presented with a problem-solving task involving some objects, toddlers would first interact with these objects to reckon what they are and what can be done with them. Leveraging these concepts, they could understand the internal structure of this task, without seeing all of the problem instances. Remarkably, they further build cognitively executable strategies to \\emph{rapidly} solve novel problems. To empower a learning agent with similar capability, we argue there shall be three levels of generalization in how an agent represents its knowledge: perceptual, conceptual, and algorithmic. In this paper, we devise the very first systematic benchmark that offers joint evaluation covering all three levels. This benchmark is centered around a novel task domain, HALMA, for visual concept development and rapid problem-solving. Uniquely, HALMA has a minimum yet complete concept space, upon which we introduce a novel paradigm to rigorously diagnose and dissect learning agents' capability in understanding and generalizing complex and structural concepts. We conduct extensive experiments on reinforcement learning agents with various inductive biases and carefully report their proficiency and weakness."
                    },
                    {
                        "title": "Learning Energy-Based Prior Model with Diffusion-Amortized MCMC",
                        "abstract": "Latent space Energy-Based Models (EBMs), also known as energy-based priors, have drawn growing interests in the field of generative modeling due to its flexibility in the formulation and strong modeling power of the latent space. However, the common practice of learning latent space EBMs with non-convergent short-run MCMC for prior and posterior sampling is hindering the model from further progress; the degenerate MCMC sampling quality in practice often leads to degraded generation quality and instability in training, especially with highly multi-modal and/or high-dimensional target distributions. To remedy this sampling issue, in this paper we introduce a simple but effective diffusion-based amortization method for long-run MCMC sampling and develop a novel learning algorithm for the latent space EBM based on it. We provide theoretical evidence that the learned amortization of MCMC is a valid long-run MCMC sampler. Experiments on several image modeling benchmark datasets demonstrate the superior performance of our method compared with strong counterparts"
                    },
                    {
                        "title": "ASSIST: Interactive Scene Nodes for Scalable and Realistic Indoor Simulation",
                        "abstract": "We present ASSIST, an object-wise neural radiance field as a panoptic representation for compositional and realistic simulation. Central to our approach is a novel scene node data structure that stores the information of each object in a unified fashion, allowing online interaction in both intra- and cross-scene settings. By incorporating a differentiable neural network along with the associated bounding box and semantic features, the proposed structure guarantees user-friendly interaction on independent objects to scale up novel view simulation. Objects in the scene can be queried, added, duplicated, deleted, transformed, or swapped simply through mouse/keyboard controls or language instructions. Experiments demonstrate the efficacy of the proposed method, where scaled realistic simulation can be achieved through interactive editing and compositional rendering, with color images, depth images, and panoptic segmentation masks generated in a 3D consistent manner."
                    },
                    {
                        "title": "Learning a Decision Module by Imitating Driver's Control Behaviors",
                        "abstract": "Autonomous driving systems have a pipeline of perception, decision, planning, and control. The decision module processes information from the perception module and directs the execution of downstream planning and control modules. On the other hand, the recent success of deep learning suggests that this pipeline could be replaced by end-to-end neural control policies, however, safety cannot be well guaranteed for the data-driven neural networks. In this work, we propose a hybrid framework to learn neural decisions in the classical modular pipeline through end-to-end imitation learning. This hybrid framework can preserve the merits of the classical pipeline such as the strict enforcement of physical and logical constraints while learning complex driving decisions from data. To circumvent the ambiguous annotation of human driving decisions, our method learns high-level driving decisions by imitating low-level control behaviors. We show in the simulation experiments that our modular driving agent can generalize its driving decision and control to various complex scenarios where the rule-based programs fail. It can also generate smoother and safer driving trajectories than end-to-end neural policies."
                    },
                    {
                        "title": "Unsupervised Foreground Extraction via Deep Region Competition",
                        "abstract": "We present Deep Region Competition (DRC), an algorithm designed to extract foreground objects from images in a fully unsupervised manner. Foreground extraction can be viewed as a special case of generic image segmentation that focuses on identifying and disentangling objects from the background. In this work, we rethink the foreground extraction by reconciling energy-based prior with generative image modeling in the form of Mixture of Experts (MoE), where we further introduce the learned pixel re-assignment as the essential inductive bias to capture the regularities of background regions. With this modeling, the foreground-background partition can be naturally found through Expectation-Maximization (EM). We show that the proposed method effectively exploits the interaction between the mixture components during the partitioning process, which closely connects to region competition, a seminal approach for generic image segmentation. Experiments demonstrate that DRC exhibits more competitive performances on complex real-world data and challenging multi-object scenes compared with prior methods. Moreover, we show empirically that DRC can potentially generalize to novel foreground objects even from categories unseen during training."
                    },
                    {
                        "title": "Learning Algebraic Representation for Systematic Generalization in Abstract Reasoning",
                        "abstract": "Is intelligence realized by connectionist or classicist? While connectionist approaches have achieved superhuman performance, there has been growing evidence that such task-specific superiority is particularly fragile in systematic generalization. This observation lies in the central debate between connectionist and classicist, wherein the latter continually advocates an algebraic treatment in cognitive architectures. In this work, we follow the classicist's call and propose a hybrid approach to improve systematic generalization in reasoning. Specifically, we showcase a prototype with algebraic representation for the abstract spatial-temporal reasoning task of Raven's Progressive Matrices (RPM) and present the ALgebra-Aware Neuro-Semi-Symbolic (ALANS) learner. The ALANS learner is motivated by abstract algebra and the representation theory. It consists of a neural visual perception frontend and an algebraic abstract reasoning backend: the frontend summarizes the visual information from object-based representation, while the backend transforms it into an algebraic structure and induces the hidden operator on the fly. The induced operator is later executed to predict the answer's representation, and the choice most similar to the prediction is selected as the solution. Extensive experiments show that by incorporating an algebraic treatment, the ALANS learner outperforms various pure connectionist models in domains requiring systematic generalization. We further show the generative nature of the learned algebraic representation; it can be decoded by isomorphism to generate an answer."
                    },
                    {
                        "title": "Emergent Graphical Conventions in a Visual Communication Game",
                        "abstract": "Humans communicate with graphical sketches apart from symbolic languages. Primarily focusing on the latter, recent studies of emergent communication overlook the sketches; they do not account for the evolution process through which symbolic sign systems emerge in the trade-off between iconicity and symbolicity. In this work, we take the very first step to model and simulate this process via two neural agents playing a visual communication game; the sender communicates with the receiver by sketching on a canvas. We devise a novel reinforcement learning method such that agents are evolved jointly towards successful communication and abstract graphical conventions. To inspect the emerged conventions, we define three fundamental properties -- iconicity, symbolicity, and semanticity -- and design evaluation methods accordingly. Our experimental results under different controls are consistent with the observation in studies of human graphical conventions. Of note, we find that evolved sketches can preserve the continuum of semantics under proper environmental pressures. More interestingly, co-evolved agents can switch between conventionalized and iconic communication based on their familiarity with referents. We hope the present research can pave the path for studying emergent communication with the modality of sketches."
                    },
                    {
                        "title": "EM Distillation for One-step Diffusion Models",
                        "abstract": "While diffusion models can learn complex distributions, sampling requires a computationally expensive iterative process. Existing distillation methods enable efficient sampling, but have notable limitations, such as performance degradation with very few sampling steps, reliance on training data access, or mode-seeking optimization that may fail to capture the full distribution. We propose EM Distillation (EMD), a maximum likelihood-based approach that distills a diffusion model to a one-step generator model with minimal loss of perceptual quality. Our approach is derived through the lens of Expectation-Maximization (EM), where the generator parameters are updated using samples from the joint distribution of the diffusion teacher prior and inferred generator latents. We develop a reparametrized sampling scheme and a noise cancellation technique that together stabilizes the distillation process. We further reveal an interesting connection of our method with existing methods that minimize mode-seeking KL. EMD outperforms existing one-step generative methods in terms of FID scores on ImageNet-64 and ImageNet-128, and compares favorably with prior work on distilling text-to-image diffusion models."
                    }
                ]
            },
            "28152f88-c646-4c43-9496-f4ab798aad42": {
                "pk": "28152f88-c646-4c43-9496-f4ab798aad42",
                "name": "Ying Nian Wu",
                "collaborators": [
                    "Song-Chun Zhu",
                    "Quanshi Zhang",
                    "Yu Yang",
                    "Tian Han",
                    "Jianwen Xie",
                    "Bo Pang",
                    "Jiali Cui",
                    "Zijun Zhang",
                    "Qian Yu",
                    "Yuanhong Zeng"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Generative Models",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Discussion of \"EQUI-energy sampler\" by Kou, Zhou and Wong",
                        "abstract": "Discussion of ``EQUI-energy sampler'' by Kou, Zhou and Wong [math.ST/0507080]"
                    },
                    {
                        "title": "Unsupervised Learning of Neural Networks to Explain Neural Networks (extended abstract)",
                        "abstract": "This paper presents an unsupervised method to learn a neural network, namely an explainer, to interpret a pre-trained convolutional neural network (CNN), i.e., the explainer uses interpretable visual concepts to explain features in middle conv-layers of a CNN. Given feature maps of a conv-layer of the CNN, the explainer performs like an auto-encoder, which decomposes the feature maps into object-part features. The object-part features are learned to reconstruct CNN features without much loss of information. We can consider the disentangled representations of object parts a paraphrase of CNN features, which help people understand the knowledge encoded by the CNN. More crucially, we learn the explainer via knowledge distillation without using any annotations of object parts or textures for supervision. In experiments, our method was widely used to interpret features of different benchmark CNNs, and explainers significantly boosted the feature interpretability without hurting the discrimination power of the CNNs."
                    },
                    {
                        "title": "Learning Latent Space Energy-Based Prior Model for Molecule Generation",
                        "abstract": "Deep generative models have recently been applied to molecule design. If the molecules are encoded in linear SMILES strings, modeling becomes convenient. However, models relying on string representations tend to generate invalid samples and duplicates. Prior work addressed these issues by building models on chemically-valid fragments or explicitly enforcing chemical rules in the generation process. We argue that an expressive model is sufficient to implicitly and automatically learn the complicated chemical rules from the data, even if molecules are encoded in simple character-level SMILES strings. We propose to learn latent space energy-based prior model with SMILES representation for molecule modeling. Our experiments show that our method is able to generate molecules with validity and uniqueness competitive with state-of-the-art models. Interestingly, generated molecules have structural and chemical features whose distributions almost perfectly match those of the real molecules."
                    },
                    {
                        "title": "Latent Space Energy-Based Model of Symbol-Vector Coupling for Text Generation and Classification",
                        "abstract": "We propose a latent space energy-based prior model for text generation and classification. The model stands on a generator network that generates the text sequence based on a continuous latent vector. The energy term of the prior model couples a continuous latent vector and a symbolic one-hot vector, so that discrete category can be inferred from the observed example based on the continuous latent vector. Such a latent space coupling naturally enables incorporation of information bottleneck regularization to encourage the continuous latent vector to extract information from the observed example that is informative of the underlying category. In our learning method, the symbol-vector coupling, the generator network and the inference network are learned jointly. Our model can be learned in an unsupervised setting where no category labels are provided. It can also be learned in semi-supervised setting where category labels are provided for a subset of training examples. Our experiments demonstrate that the proposed model learns well-structured and meaningful latent space, which (1) guides the generator to generate text with high quality, diversity, and interpretability, and (2) effectively classifies text."
                    },
                    {
                        "title": "Triple Regression for Camera Agnostic Sim2Real Robot Grasping and Manipulation Tasks",
                        "abstract": "Sim2Real (Simulation to Reality) techniques have gained prominence in robotic manipulation and motion planning due to their ability to enhance success rates by enabling agents to test and evaluate various policies and trajectories. In this paper, we investigate the advantages of integrating Sim2Real into robotic frameworks. We introduce the Triple Regression Sim2Real framework, which constructs a real-time digital twin. This twin serves as a replica of reality to simulate and evaluate multiple plans before their execution in real-world scenarios. Our triple regression approach addresses the reality gap by: (1) mitigating projection errors between real and simulated camera perspectives through the first two regression models, and (2) detecting discrepancies in robot control using the third regression model. Experiments on 6-DoF grasp and manipulation tasks (where the gripper can approach from any direction) highlight the effectiveness of our framework. Remarkably, with only RGB input images, our method achieves state-of-the-art success rates. This research advances efficient robot training methods and sets the stage for rapid advancements in robotics and automation."
                    },
                    {
                        "title": "Learning Hierarchical Features with Joint Latent Space Energy-Based Prior",
                        "abstract": "This paper studies the fundamental problem of multi-layer generator models in learning hierarchical representations. The multi-layer generator model that consists of multiple layers of latent variables organized in a top-down architecture tends to learn multiple levels of data abstraction. However, such multi-layer latent variables are typically parameterized to be Gaussian, which can be less informative in capturing complex abstractions, resulting in limited success in hierarchical representation learning. On the other hand, the energy-based (EBM) prior is known to be expressive in capturing the data regularities, but it often lacks the hierarchical structure to capture different levels of hierarchical representations. In this paper, we propose a joint latent space EBM prior model with multi-layer latent variables for effective hierarchical representation learning. We develop a variational joint learning scheme that seamlessly integrates an inference model for efficient inference. Our experiments demonstrate that the proposed joint EBM prior is effective and expressive in capturing hierarchical representations and modelling data distribution."
                    },
                    {
                        "title": "Efficient and Scalable Fine-Tune of Language Models for Genome Understanding",
                        "abstract": "Although DNA foundation models have advanced the understanding of genomes, they still face significant challenges in the limited scale and diversity of genomic data. This limitation starkly contrasts with the success of natural language foundation models, which thrive on substantially larger scales. Furthermore, genome understanding involves numerous downstream genome annotation tasks with inherent data heterogeneity, thereby necessitating more efficient and robust fine-tuning methods tailored for genomics. Here, we present \\textsc{Lingo}: \\textsc{L}anguage prefix f\\textsc{In}e-tuning for \\textsc{G}en\\textsc{O}mes. Unlike DNA foundation models, \\textsc{Lingo} strategically leverages natural language foundation models' contextual cues, recalibrating their linguistic knowledge to genomic sequences. \\textsc{Lingo} further accommodates numerous, heterogeneous downstream fine-tune tasks by an adaptive rank sampling method that prunes and stochastically reintroduces pruned singular vectors within small computational budgets. Adaptive rank sampling outperformed existing fine-tuning methods on all benchmarked 14 genome understanding tasks, while requiring fewer than 2\\% of trainable parameters as genomic-specific adapters. Impressively, applying these adapters on natural language foundation models matched or even exceeded the performance of DNA foundation models. \\textsc{Lingo} presents a new paradigm of efficient and scalable genome understanding via genomic-specific adapters on language models."
                    },
                    {
                        "title": "Object-Conditioned Energy-Based Attention Map Alignment in Text-to-Image Diffusion Models",
                        "abstract": "Text-to-image diffusion models have shown great success in generating high-quality text-guided images. Yet, these models may still fail to semantically align generated images with the provided text prompts, leading to problems like incorrect attribute binding and/or catastrophic object neglect. Given the pervasive object-oriented structure underlying text prompts, we introduce a novel object-conditioned Energy-Based Attention Map Alignment (EBAMA) method to address the aforementioned problems. We show that an object-centric attribute binding loss naturally emerges by approximately maximizing the log-likelihood of a $z$-parameterized energy-based model with the help of the negative sampling technique. We further propose an object-centric intensity regularizer to prevent excessive shifts of objects attention towards their attributes. Extensive qualitative and quantitative experiments, including human evaluation, on several challenging benchmarks demonstrate the superior performance of our method over previous strong counterparts. With better aligned attention maps, our approach shows great promise in further enhancing the text-controlled image editing ability of diffusion models."
                    },
                    {
                        "title": "Learning Mixtures of Bernoulli Templates by Two-Round EM with Performance Guarantee",
                        "abstract": "Dasgupta and Shulman showed that a two-round variant of the EM algorithm can learn mixture of Gaussian distributions with near optimal precision with high probability if the Gaussian distributions are well separated and if the dimension is sufficiently high. In this paper, we generalize their theory to learning mixture of high-dimensional Bernoulli templates. Each template is a binary vector, and a template generates examples by randomly switching its binary components independently with a certain probability. In computer vision applications, a binary vector is a feature map of an image, where each binary component indicates whether a local feature or structure is present or absent within a certain cell of the image domain. A Bernoulli template can be considered as a statistical model for images of objects (or parts of objects) from the same category. We show that the two-round EM algorithm can learn mixture of Bernoulli templates with near optimal precision with high probability, if the Bernoulli templates are sufficiently different and if the number of features is sufficiently high. We illustrate the theoretical results by synthetic and real examples."
                    },
                    {
                        "title": "Learning Joint Latent Space EBM Prior Model for Multi-layer Generator",
                        "abstract": "This paper studies the fundamental problem of learning multi-layer generator models. The multi-layer generator model builds multiple layers of latent variables as a prior model on top of the generator, which benefits learning complex data distribution and hierarchical representations. However, such a prior model usually focuses on modeling inter-layer relations between latent variables by assuming non-informative (conditional) Gaussian distributions, which can be limited in model expressivity. To tackle this issue and learn more expressive prior models, we propose an energy-based model (EBM) on the joint latent space over all layers of latent variables with the multi-layer generator as its backbone. Such joint latent space EBM prior model captures the intra-layer contextual relations at each layer through layer-wise energy terms, and latent variables across different layers are jointly corrected. We develop a joint training scheme via maximum likelihood estimation (MLE), which involves Markov Chain Monte Carlo (MCMC) sampling for both prior and posterior distributions of the latent variables from different layers. To ensure efficient inference and learning, we further propose a variational training scheme where an inference model is used to amortize the costly posterior MCMC sampling. Our experiments demonstrate that the learned model can be expressive in generating high-quality images and capturing hierarchical features for better outlier detection."
                    },
                    {
                        "title": "Representation Learning: A Statistical Perspective",
                        "abstract": "Learning representations of data is an important problem in statistics and machine learning. While the origin of learning representations can be traced back to factor analysis and multidimensional scaling in statistics, it has become a central theme in deep learning with important applications in computer vision and computational neuroscience. In this article, we review recent advances in learning representations from a statistical perspective. In particular, we review the following two themes: (a) unsupervised learning of vector representations and (b) learning of both vector and matrix representations."
                    },
                    {
                        "title": "Mining Object Parts from CNNs via Active Question-Answering",
                        "abstract": "Given a convolutional neural network (CNN) that is pre-trained for object classification, this paper proposes to use active question-answering to semanticize neural patterns in conv-layers of the CNN and mine part concepts. For each part concept, we mine neural patterns in the pre-trained CNN, which are related to the target part, and use these patterns to construct an And-Or graph (AOG) to represent a four-layer semantic hierarchy of the part. As an interpretable model, the AOG associates different CNN units with different explicit object parts. We use an active human-computer communication to incrementally grow such an AOG on the pre-trained CNN as follows. We allow the computer to actively identify objects, whose neural patterns cannot be explained by the current AOG. Then, the computer asks human about the unexplained objects, and uses the answers to automatically discover certain CNN patterns corresponding to the missing knowledge. We incrementally grow the AOG to encode new knowledge discovered during the active-learning process. In experiments, our method exhibits high learning efficiency. Our method uses about 1/6-1/3 of the part annotations for training, but achieves similar or better part-localization performance than fast-RCNN methods."
                    },
                    {
                        "title": "Synthesizing Dynamic Patterns by Spatial-Temporal Generative ConvNet",
                        "abstract": "Video sequences contain rich dynamic patterns, such as dynamic texture patterns that exhibit stationarity in the temporal domain, and action patterns that are non-stationary in either spatial or temporal domain. We show that a spatial-temporal generative ConvNet can be used to model and synthesize dynamic patterns. The model defines a probability distribution on the video sequence, and the log probability is defined by a spatial-temporal ConvNet that consists of multiple layers of spatial-temporal filters to capture spatial-temporal patterns of different scales. The model can be learned from the training video sequences by an \"analysis by synthesis\" learning algorithm that iterates the following two steps. Step 1 synthesizes video sequences from the currently learned model. Step 2 then updates the model parameters based on the difference between the synthesized video sequences and the observed training sequences. We show that the learning algorithm can synthesize realistic dynamic patterns."
                    },
                    {
                        "title": "Network Transplanting (extended abstract)",
                        "abstract": "This paper focuses on a new task, i.e., transplanting a category-and-task-specific neural network to a generic, modular network without strong supervision. We design a functionally interpretable structure for the generic network. Like building LEGO blocks, we teach the generic network a new category by directly transplanting the module corresponding to the category from a pre-trained network with a few or even without sample annotations. Our method incrementally adds new categories to the generic network but does not affect representations of existing categories. In this way, our method breaks the typical bottleneck of learning a net for massive tasks and categories, i.e., the requirement of collecting samples for all tasks and categories at the same time before the learning begins. Thus, we use a new distillation algorithm, namely back-distillation, to overcome specific challenges of network transplanting. Our method without training samples even outperformed the baseline with 100 training samples."
                    },
                    {
                        "title": "Inducing Sparse Coding and And-Or Grammar from Generator Network",
                        "abstract": "We introduce an explainable generative model by applying sparse operation on the feature maps of the generator network. Meaningful hierarchical representations are obtained using the proposed generative model with sparse activations. The convolutional kernels from the bottom layer to the top layer of the generator network can learn primitives such as edges and colors, object parts, and whole objects layer by layer. From the perspective of the generator network, we propose a method for inducing both sparse coding and the AND-OR grammar for images. Experiments show that our method is capable of learning meaningful and explainable hierarchical representations."
                    },
                    {
                        "title": "Learning FRAME Models Using CNN Filters",
                        "abstract": "The convolutional neural network (ConvNet or CNN) has proven to be very successful in many tasks such as those in computer vision. In this conceptual paper, we study the generative perspective of the discriminative CNN. In particular, we propose to learn the generative FRAME (Filters, Random field, And Maximum Entropy) model using the highly expressive filters pre-learned by the CNN at the convolutional layers. We show that the learning algorithm can generate realistic and rich object and texture patterns in natural scenes. We explain that each learned model corresponds to a new CNN unit at a layer above the layer of filters employed by the model. We further show that it is possible to learn a new layer of CNN units using a generative CNN model, which is a product of experts model, and the learning algorithm admits an EM interpretation with binary latent variables."
                    },
                    {
                        "title": "Interpreting CNNs via Decision Trees",
                        "abstract": "This paper aims to quantitatively explain rationales of each prediction that is made by a pre-trained convolutional neural network (CNN). We propose to learn a decision tree, which clarifies the specific reason for each prediction made by the CNN at the semantic level. I.e., the decision tree decomposes feature representations in high conv-layers of the CNN into elementary concepts of object parts. In this way, the decision tree tells people which object parts activate which filters for the prediction and how much they contribute to the prediction score. Such semantic and quantitative explanations for CNN predictions have specific values beyond the traditional pixel-level analysis of CNNs. More specifically, our method mines all potential decision modes of the CNN, where each mode represents a common case of how the CNN uses object parts for prediction. The decision tree organizes all potential decision modes in a coarse-to-fine manner to explain CNN predictions at different fine-grained levels. Experiments have demonstrated the effectiveness of the proposed method."
                    },
                    {
                        "title": "Network Transplanting",
                        "abstract": "This paper focuses on a new task, i.e., transplanting a category-and-task-specific neural network to a generic, modular network without strong supervision. We design an functionally interpretable structure for the generic network. Like building LEGO blocks, we teach the generic network a new category by directly transplanting the module corresponding to the category from a pre-trained network with a few or even without sample annotations. Our method incrementally adds new categories to the generic network but does not affect representations of existing categories. In this way, our method breaks the typical bottleneck of learning a net for massive tasks and categories, i.e. the requirement of collecting samples for all tasks and categories at the same time before the learning begins. Thus, we use a new distillation algorithm, namely back-distillation, to overcome specific challenges of network transplanting. Our method without training samples even outperformed the baseline with 100 training samples."
                    },
                    {
                        "title": "Neural Architecture Search for Joint Optimization of Predictive Power and Biological Knowledge",
                        "abstract": "We report a neural architecture search framework, BioNAS, that is tailored for biomedical researchers to easily build, evaluate, and uncover novel knowledge from interpretable deep learning models. The introduction of knowledge dissimilarity functions in BioNAS enables the joint optimization of predictive power and biological knowledge through searching architectures in a model space. By optimizing the consistency with existing knowledge, we demonstrate that BioNAS optimal models reveal novel knowledge in both simulated data and in real data of functional genomics. BioNAS provides a useful tool for domain experts to inject their prior belief into automated machine learning and therefore making deep learning easily accessible to practitioners. BioNAS is available at https://github.com/zj-zhang/BioNAS-pub."
                    },
                    {
                        "title": "Learning Energy-based Spatial-Temporal Generative ConvNets for Dynamic Patterns",
                        "abstract": "Video sequences contain rich dynamic patterns, such as dynamic texture patterns that exhibit stationarity in the temporal domain, and action patterns that are non-stationary in either spatial or temporal domain. We show that an energy-based spatial-temporal generative ConvNet can be used to model and synthesize dynamic patterns. The model defines a probability distribution on the video sequence, and the log probability is defined by a spatial-temporal ConvNet that consists of multiple layers of spatial-temporal filters to capture spatial-temporal patterns of different scales. The model can be learned from the training video sequences by an \"analysis by synthesis\" learning algorithm that iterates the following two steps. Step 1 synthesizes video sequences from the currently learned model. Step 2 then updates the model parameters based on the difference between the synthesized video sequences and the observed training sequences. We show that the learning algorithm can synthesize realistic dynamic patterns. We also show that it is possible to learn the model from incomplete training sequences with either occluded pixels or missing frames, so that model learning and pattern completion can be accomplished simultaneously."
                    }
                ]
            }
        }
    },
    "2410.02249": {
        "paper_data": {
            "title": "Spiking Neural Network as Adaptive Event Stream Slicer",
            "url": "http://arxiv.org/abs/2410.02249v1",
            "arxiv_id": "2410.02249",
            "authors": [
                "Jiahang Cao",
                "Mingyuan Sun",
                "Ziqing Wang",
                "Hao Cheng",
                "Qiang Zhang",
                "Shibo Zhou",
                "Renjing Xu"
            ],
            "abstract": "Event-based cameras are attracting significant interest as they provide rich edge information, high dynamic range, and high temporal resolution. Many state-of-the-art event-based algorithms rely on splitting the events into fixed groups, resulting in the omission of crucial temporal information, particularly when dealing with diverse motion scenarios (e.g., high/low speed). In this work, we propose SpikeSlicer, a novel-designed plug-and-play event processing method capable of splitting events stream adaptively. SpikeSlicer utilizes a lightweight (0.41M) and low-energy spiking neural network (SNN) to trigger event slicing. To guide the SNN to fire spikes at optimal time steps, we propose the Spiking Position-aware Loss (SPA-Loss) to modulate the neuron's state. Additionally, we develop a Feedback-Update training strategy that refines the slicing decisions using feedback from the downstream artificial neural network (ANN). Extensive experiments demonstrate that our method yields significant performance improvements in event-based object tracking and recognition. Notably, SpikeSlicer provides a brand-new SNN-ANN cooperation paradigm, where the SNN acts as an efficient, low-energy data processor to assist the ANN in improving downstream performance, injecting new perspectives and potential avenues of exploration.",
            "introduction": "   1 Introduction  Event-based cameras\u00a0[1] are bio-inspired sensors that capture event streams in an asynchronous and sparse way. Compared with conventional frame-based cameras, event-based cameras offer numerous outstanding properties: high temporal resolution (with the order of \u03bc\u2062s\ud835\udf07\ud835\udc60\\mu sitalic_\u03bc italic_s), high dynamic range (higher than 120 dB), low latency, and low power consumption. Over recent years, rapid growth has been witnessed in dealing with event data due to the inherent advantages of event-based cameras, such as object tracking\u00a0[2, 3], depth estimation\u00a0[4, 5], and recognition\u00a0[6, 7]. Before applying to various downstream tasks, the event stream must be split by groups and then transformed into different event representations, e.g., frame, voxel, or point for deep learning architectures.   In detail, the process of event-to-representation conversion consists of two steps: (1) slicing the raw event stream into multiple sub-event stream groups, and (2) converting these sub-event streams into different event representations. Much of the current research focuses on the second step, aiming to refine event representation\u00a0[8] techniques such as time surface\u00a0[6] and event spike tensor\u00a0[9]. Yet, this focus often overlooks the crucial first step of slicing, where issues such as the non-uniformity of the information contained in the fixed-sliced event remain unaddressed.   To address the challenges in the slicing process, we delve into the limitations of traditional slicing techniques. Common methods typically cut the event stream into several fixed groups. For example, slicing event stream with fixed event count\u00a0[10] or fixed time intervals\u00a0[11, 12] as depicted in Fig.\u00a01. However, these fixed-group slicing techniques often lead to problems: they may cause insufficient information capture in low-speed motion scenarios or excessive redundancy in high-speed conditions, thereby failing to accurately capture the dynamic changes in event distribution. Additionally, some hyper-parameters, e.g., the length of time interval, are highly-sensitive to the downstream tasks (examples are provided in Appendix\u00a0C) and must be carefully pre-determined. Although some latest slicing methods\u00a0[13, 14] propose to adaptively sample the events, there still exists the problem of hyper-parameter tuning which can not achieve a fully learnable and adaptable slicing process.   In order to address the above issues, we propose SpikeSlicer, a novel-design event processing method that can adaptively slice the event streams. To achieve this, SpikeSlicer utilizes an SNN as an event trigger to dynamically determine the optimal moment to split the event stream. Our objectives include: (1) training the SNN to spike at a specific time step for accurate event slicing, and (2) developing a training strategy to identify the best slicing time for a continuous event stream during training. In our paper, we achieve (1) through our newly introduced Spiking Position-aware Loss (SPA-Loss) function, which effectively guides the SNN to spike at the desired time by manipulating the membrane potential. For (2), we implement a Feedback-Update training strategy, where the SNN receives real-time performance feedback from the downstream ANN model for supervision. An overview of our proposed method is depicted in Fig.\u00a02. We evaluate the effectiveness of our proposed SpikeSlicer in two downstream tasks: (i) event-based object tracking, which is strongly sensitive to temporal information and motion dynamics, and (ii) event-based recognition, which is highly related to event density. Extensive experiments validate the effectiveness of the",
            "references": [
                {
                    "title": "Spiking Denoising Diffusion Probabilistic Models",
                    "abstract": "Spiking neural networks (SNNs) have ultra-low energy consumption and high biological plausibility due to their binary and bio-driven nature compared with artificial neural networks (ANNs). While previous research has primarily focused on enhancing the performance of SNNs in classification tasks, the generative potential of SNNs remains relatively unexplored. In our paper, we put forward Spiking Denoising Diffusion Probabilistic Models (SDDPM), a new class of SNN-based generative models that achieve high sample quality. To fully exploit the energy efficiency of SNNs, we propose a purely Spiking U-Net architecture, which achieves comparable performance to its ANN counterpart using only 4 time steps, resulting in significantly reduced energy consumption. Extensive experimental results reveal that our approach achieves state-of-the-art on the generative tasks and substantially outperforms other SNN-based generative models, achieving up to 12\u00d7 and 6\u00d7 improvement on the CIFAR-10 and the CelebA datasets, respectively. Moreover, we propose a threshold-guided strategy that can further improve the performances by 2.69% in a training-free manner. The SDDPM symbolizes a significant advancement in the field of SNN generation, injecting new perspectives and potential avenues of exploration. Our code is available at https://github.com/AndyCao1125/SDDPM."
                },
                {
                    "title": "Better and Faster: Adaptive Event Conversion for Event-Based Object Detection",
                    "abstract": "Event cameras are a kind of bio-inspired imaging sensor, which asynchronously collect sparse event streams with many advantages. In this paper, we focus on building better and faster event-based object detectors. To this end, we first propose a computationally efficient event representation Hyper Histogram, which adequately preserves both the polarity and temporal information of events. Then we devise an Adaptive Event Conversion module, which converts events into Hyper Histograms according to event density via an adaptive queue. Moreover, we introduce a novel event-based augmentation method Shadow Mosaic, which significantly improves the event sample diversity and enhances the generalization ability of detection models. We equip our proposed modules on three representative object detection models: YOLOv5, Deformable-DETR, and RetinaNet. Experimental results on three event-based detection datasets (1Mpx, Gen1, and MVSEC-NIGHTL21) demonstrate that our proposed approach outperforms other state-of-the-art methods by a large margin, while achieving a much faster running speed (< 14 ms and < 4 ms for 50 ms event data on the 1Mpx and Gen1 datasets)."
                },
                {
                    "title": "Frame-Event Alignment and Fusion Network for High Frame Rate Tracking",
                    "abstract": "Most existing RGB-based trackers target low frame rate benchmarks of around 30 frames per second. This setting restricts the tracker's functionality in the real world, especially for fast motion. Event-based cameras as bioinspired sensors provide considerable potential for high frame rate tracking due to their high temporal resolution. However, event-based cameras cannot offer fine-grained texture information like conventional cameras. This unique complementarity motivates us to combine conventional frames and events for high frame rate object tracking under various challenging conditions. In this paper, we propose an end-to-end network consisting of multi-modality alignment and fusion modules to effectively combine meaningful information from both modalities at different measurement rates. The alignment module is responsible for cross-style and cross-frame-rate alignment between frame and event modalities under the guidance of the moving cues furnished by events. While the fusion module is accountable for emphasizing valuable features and suppressing noise information by the mutual complement between the two modalities. Extensive experiments show that the proposed approach outper-forms state-of-the-art trackers by a significant margin in high frame rate tracking. With the FE240Hz dataset, our approach achieves high frame rate tracking up to 240Hz."
                },
                {
                    "title": "Brain-inspired multimodal hybrid neural network for robot place recognition",
                    "abstract": "Place recognition is an essential spatial intelligence capability for robots to understand and navigate the world. However, recognizing places in natural environments remains a challenging task for robots because of resource limitations and changing environments. In contrast, humans and animals can robustly and efficiently recognize hundreds of thousands of places in different conditions. Here, we report a brain-inspired general place recognition system, dubbed NeuroGPR, that enables robots to recognize places by mimicking the neural mechanism of multimodal sensing, encoding, and computing through a continuum of space and time. Our system consists of a multimodal hybrid neural network (MHNN) that encodes and integrates multimodal cues from both conventional and neuromorphic sensors. Specifically, to encode different sensory cues, we built various neural networks of spatial view cells, place cells, head direction cells, and time cells. To integrate these cues, we designed a multiscale liquid state machine that can process and fuse multimodal information effectively and asynchronously using diverse neuronal dynamics and bioinspired inhibitory circuits. We deployed the MHNN on Tianjic, a hybrid neuromorphic chip, and integrated it into a quadruped robot. Our results show that NeuroGPR achieves better performance compared with conventional and existing biologically inspired approaches, exhibiting robustness to diverse environmental uncertainty, including perceptual aliasing, motion blur, light, or weather changes. Running NeuroGPR as an overall multi\u2013neural network workload on Tianjic showcases its advantages with 10.5 times lower latency and 43.6% lower power consumption than the commonly used mobile robot processor Jetson Xavier NX. Description NeuroGPR with multimodal sensing, encoding and computing facilitates robots to robustly and efficiently recognize places in natural environments."
                },
                {
                    "title": "Masked Spiking Transformer",
                    "abstract": "The combination of Spiking Neural Networks (SNNs) and Transformers has attracted significant attention due to their potential for high energy efficiency and high-performance nature. However, existing works on this topic typically rely on direct training, which can lead to suboptimal performance. To address this issue, we propose to leverage the benefits of the ANN-to-SNN conversion method to combine SNNs and Transformers, resulting in significantly improved performance over existing state-of-the-art SNN models. Furthermore, inspired by the quantal synaptic failures observed in the nervous system, which reduce the number of spikes transmitted across synapses, we introduce a novel Masked Spiking Transformer (MST) framework. This incorporates a Random Spike Masking (RSM) method to prune redundant spikes and reduce energy consumption without sacrificing performance. Our experimental results demonstrate that the proposed MST model achieves a significant reduction of 26.8% in power consumption when the masking ratio is 75% while maintaining the same level of performance as the unmasked model. The code is available at: https://github.com/bic-L/Masked-Spiking-Transformer."
                },
                {
                    "title": "Attention Spiking Neural Networks",
                    "abstract": "Brain-inspired spiking neural networks (SNNs) are becoming a promising energy-efficient alternative to traditional artificial neural networks (ANNs). However, the performance gap between SNNs and ANNs has been a significant hindrance to deploying SNNs ubiquitously. To leverage the full potential of SNNs, in this paper we study the attention mechanisms, which can help human focus on important information. We present our idea of attention in SNNs with a multi-dimensional attention module, which infers attention weights along the temporal, channel, as well as spatial dimension separately or simultaneously. Based on the existing neuroscience theories, we exploit the attention weights to optimize membrane potentials, which in turn regulate the spiking response. Extensive experimental results on event-based action recognition and image classification datasets demonstrate that attention facilitates vanilla SNNs to achieve sparser spiking firing, better performance, and energy efficiency concurrently. In particular, we achieve top-1 accuracy of 75.92% and 77.08% on ImageNet-1 K with single/4-step Res-SNN-104, which are state-of-the-art results in SNNs. Compared with counterpart Res-ANN-104, the performance gap becomes -0.95/+0.21 percent and the energy efficiency is 31.8\u00d7/7.4\u00d7. To analyze the effectiveness of attention SNNs, we theoretically prove that the spiking degradation or the gradient vanishing, which usually holds in general SNNs, can be resolved by introducing the block dynamical isometry theory. We also analyze the efficiency of attention SNNs based on our proposed spiking response visualization method. Our work lights up SNN's potential as a general backbone to support various applications in the field of SNN research, with a great balance between effectiveness and energy efficiency."
                },
                {
                    "title": "Ev-NeRF: Event Based Neural Radiance Field",
                    "abstract": "We present Ev-NeRF, a Neural Radiance Field derived from event data. While event cameras can measure subtle brightness changes in high frame rates, the measurements in low lighting or extreme motion suffer from significant domain discrepancy with complex noise. As a result, the performance of event-based vision tasks does not transfer to challenging environments, where the event cameras are expected to thrive over normal cameras. We find that the multi-view consistency of NeRF provides a powerful self-supervision signal for eliminating spurious measurements and extracting the consistent underlying structure despite highly noisy input. Instead of posed images of the original NeRF, the input to Ev-NeRF is the event measurements accompanied by the movements of the sensors. Using the loss function that reflects the measurement model of the sensor, Ev-NeRF creates an integrated neural volume that summarizes the unstructured and sparse data points captured for about 2-4 seconds. The generated neural volume can also produce intensity images from novel views with reasonable depth estimates, which can serve as a high-quality input to various vision-based tasks. Our results show that Ev-NeRF achieves competitive performance for intensity image reconstruction under extreme noise and high-dynamic-range imaging."
                },
                {
                    "title": "EventNeRF: Neural Radiance Fields from a Single Colour Event Camera",
                    "abstract": "Asynchronously operating event cameras find many applications due to their high dynamic range, vanishingly low motion blur, low latency and low data bandwidth. The field saw remarkable progress during the last few years, and existing event-based 3D reconstruction approaches recover sparse point clouds of the scene. However, such sparsity is a limiting factor in many cases, especially in computer vision and graphics, that has not been addressed satisfactorily so far. Accordingly, this paper proposes the first approach for 3D-consistent, dense and photorealistic novel view synthesis using just a single colour event stream as input. At its core is a neural radiance field trained entirely in a self-supervised manner from events while preserving the original resolution of the colour event channels. Next, our ray sampling strategy is tailored to events and allows for data-efficient training. At test, our method produces results in the RGB space at unprecedented quality. We evaluate our method qualitatively and numerically on several challenging synthetic and real scenes and show that it produces significantly denser and more visually appealing renderings than the existing methods. We also demonstrate robustness in challenging scenarios with fast motion and under low lighting conditions. We release the newly recorded dataset and our source code to facilitate the research field, see https://4dqv.mpi-inf.mpg.de/EventNeRF."
                },
                {
                    "title": "Spiking Transformers for Event-based Single Object Tracking",
                    "abstract": "Event-based cameras bring a unique capability to tracking, being able to function in challenging real-world conditions as a direct result of their high temporal resolution and high dynamic range. These imagers capture events asynchronously that encode rich temporal and spatial information. However, effectively extracting this information from events remains an open challenge. In this work, we propose a spiking transformer network, STNet, for single object tracking. STNet dynamically extracts and fuses information from both temporal and spatial domains. In particular, the proposed architecture features a transformer module to provide global spatial information and a spiking neural network (SNN) module for extracting temporal cues. The spiking threshold of the SNN module is dynamically adjusted based on the statistical cues of the spatial information, which we find essential in providing robust SNN features. We fuse both feature branches dynamically with a novel cross-domain attention fusion algorithm. Extensive experiments on three event-based datasets, FE240hz, EED and VisEvent validate that the proposed STNet outperforms existing state-of-the-art methods in both tracking accuracy and speed with a significant margin. The code and pretrained models are at https://github.com/Jee-King/CVPR2022_STNet."
                },
                {
                    "title": "Stereo Depth from Events Cameras: Concentrate and Focus on the Future",
                    "abstract": "Neuromorphic cameras or event cameras mimic human vision by reporting changes in the intensity in a scene, instead of reporting the whole scene at once in a form of an image frame as performed by conventional cameras. Events are streamed data that are often dense when either the scene changes or the camera moves rapidly. The rapid movement causes the events to be overridden or missed when creating a tensor for the machine to learn on. To alleviate the event missing or overriding issue, we propose to learn to concentrate on the dense events to produce a compact event representation with high details for depth estimation. Specifically, we learn a model with events from both past and future but infer only with past data with the predicted future. We initially estimate depth in an event-only setting but also propose to further incorporate images and events by a hier-archical event and intensity combination network for better depth estimation. By experiments in challenging real-world scenarios, we validate that our method outperforms prior arts even with low computational cost. Code is available at: https://github.com/yonseivnl/se-cff."
                },
                {
                    "title": "Asynchronous Spatio-Temporal Memory Network for Continuous Event-Based Object Detection",
                    "abstract": "Event cameras, offering extremely high temporal resolution and high dynamic range, have brought a new perspective to addressing common object detection challenges (e.g., motion blur and low light). However, how to learn a better spatio-temporal representation and exploit rich temporal cues from asynchronous events for object detection still remains an open issue. To address this problem, we propose a novel asynchronous spatio-temporal memory network (ASTMNet) that directly consumes asynchronous events instead of event images prior to processing, which can well detect objects in a continuous manner. Technically, ASTMNet learns an asynchronous attention embedding from the continuous event stream by adopting an adaptive temporal sampling strategy and a temporal attention convolutional module. Besides, a spatio-temporal memory module is designed to exploit rich temporal cues via a lightweight yet efficient inter-weaved recurrent-convolutional architecture. Empirically, it shows that our approach outperforms the state-of-the-art methods using the feed-forward frame-based detectors on three datasets by a large margin (i.e., 7.6% in the KITTI Simulated Dataset, 10.8% in the Gen1 Automotive Dataset, and 10.5% in the 1Mpx Detection Dataset). The results demonstrate that event cameras can perform robust object detection even in cases where conventional cameras fail, e.g., fast motion and challenging light conditions."
                },
                {
                    "title": "Event-based Video Reconstruction via Potential-assisted Spiking Neural Network",
                    "abstract": "Neuromorphic vision sensor is a new bio-inspired imaging paradigm that reports asynchronous, continuously perpixel brightness changes called \u2018events\u2019 with high temporal resolution and high dynamic range. So far, the event-based image reconstruction methods are based on artificial neural networks (ANN) or hand-crafted spatiotemporal smoothing techniques. In this paper, we first implement the image reconstruction work via deep spiking neural network (SNN) architecture. As the bio-inspired neural networks, SNNs operating with asynchronous binary spikes distributed over time, can potentially lead to greater computational efficiency on event-driven hardware. We propose a novel Event-based Video reconstruction framework based on a fully Spiking Neural Network (EVSNN), which utilizes Leaky-Integrate-and-Fire (LIF) neuron and Membrane Potential (MP) neuron. We find that the spiking neurons have the potential to store useful temporal information (memory) to complete such time-dependent tasks. Further-more, to better utilize the temporal information, we propose a hybrid potential-assisted framework (PAEVSNN) using the membrane potential of spiking neuron. The proposed neuron is referred as Adaptive Membrane Potential (AMP) neuron, which adaptively updates the membrane potential according to the input spikes. The experimental results demonstrate that our models achieve comparable performance to ANN-based models on IJRR, MVSEC, and HQF datasets. The energy consumptions of EVSNN and PAEVSNN are $19.36\\times$ and $7.75\\times$ more computationally ef-ficient than their ANN architectures, respectively. The code and pretrained model are available at https://sites.google.com/view/evsnn."
                },
                {
                    "title": "N-ImageNet: Towards Robust, Fine-Grained Object Recognition with Event Cameras",
                    "abstract": "We introduce N-ImageNet, a large-scale dataset targeted for robust, fine-grained object recognition with event cameras. The dataset is collected using programmable hardware in which an event camera consistently moves around a monitor displaying images from ImageNet. N-ImageNet serves as a challenging benchmark for event-based object recognition, due to its large number of classes and samples. We empirically show that pretraining on N-ImageNet improves the performance of event-based classifiers and helps them learn with few labeled data. In addition, we present several variants of N-ImageNet to test the robustness of event-based classifiers under diverse camera trajectories and severe lighting conditions, and propose a novel event representation to alleviate the performance degradation. To the best of our knowledge, we are the first to quantitatively investigate the consequences caused by various environmental conditions on event-based object recognition algorithms. N-ImageNet and its variants are expected to guide practical implementations for deploying event-based object recognition algorithms in the real world."
                },
                {
                    "title": "Object Tracking by Jointly Exploiting Frame and Event Domain",
                    "abstract": "Inspired by the complementarity between conventional frame-based and bio-inspired event-based cameras, we propose a multi-modal based approach to fuse visual cues from the frame- and event-domain to enhance the single object tracking performance, especially in degraded conditions (e.g., scenes with high dynamic range, low light, and fast-motion objects). The proposed approach can effectively and adaptively combine meaningful information from both domains. Our approach\u2019s effectiveness is enforced by a novel designed cross-domain attention schemes, which can effectively enhance features based on self- and cross-domain attention schemes; The adaptiveness is guarded by a specially designed weighting scheme, which can adaptively balance the contribution of the two domains. To exploit event-based visual cues in single-object tracking, we construct a large-scale frame-event-based dataset, which we subsequently employ to train a novel frame-event fusion based model. Extensive experiments show that the proposed approach outperforms state-of-the-art frame-based tracking methods by at least 10.4% and 11.9% in terms of representative success rate and precision rate, respectively. Besides, the effectiveness of each key component of our approach is evidenced by our thorough ablation study."
                },
                {
                    "title": "Temporal-wise Attention Spiking Neural Networks for Event Streams Classification",
                    "abstract": "How to effectively and efficiently deal with spatio-temporal event streams, where the events are generally sparse and non-uniform and have the \u03bcs temporal resolution, is of great value and has various real-life applications. Spiking neural network (SNN), as one of the brain-inspired event-triggered computing models, has the potential to extract effective spatio-temporal features from the event streams. However, when aggregating individual events into frames with a new higher temporal resolution, existing SNN models do not attach importance to that the serial frames have different signal-to-noise ratios since event streams are sparse and non-uniform. This situation interferes with the performance of existing SNNs. In this work, we propose a temporal-wise attention SNN (TA-SNN) model to learn frame-based representation for processing event streams. Concretely, we extend the attention concept to temporal-wise input to judge the significance of frames for the final decision at the training stage, and discard the irrelevant frames at the inference stage. We demonstrate that TA-SNN models improve the accuracy of event streams classification tasks. We also study the impact of multiple-scale temporal resolutions for frame-based representation. Our approach is tested on three different classification tasks: gesture recognition, image classification, and spoken digit recognition. We report the state-of-the-art results on these tasks, and get the essential improvement of accuracy (almost 19%) for gesture recognition with only 60 ms."
                },
                {
                    "title": "CarSNN: An Efficient Spiking Neural Network for Event-Based Autonomous Cars on the Loihi Neuromorphic Research Processor",
                    "abstract": "Autonomous Driving (AD) related features provide new forms of mobility that are also beneficial for other kind of intelligent and autonomous systems like robots, smart transportation, and smart industries. For these applications, the decisions need to be made fast and in real-time. Moreover, in the quest for electric mobility, this task must follow low power policy, without affecting much the autonomy of the mean of transport or the robot. These two challenges can be tackled using the emerging Spiking Neural Networks (SNNs). When deployed on a specialized neuromorphic hardware, SNNs can achieve high performance with low latency and low power consumption. In this paper, we use an SNN connected to an event-based camera for facing one of the key problems for AD, i.e., the classification between cars and other objects. To consume less power than traditional frame-based cameras, we use a Dynamic Vision Sensor (DVS) [1]. The experiments are made following an offline supervised learning rule, followed by mapping the learnt SNN model on the Intel Loihi Neuromorphic Research Chip [2]. Our best experiment achieves an accuracy on offline implementation of 86%, that drops to 83% when it is ported onto the Loihi Chip. The Neuromorphic Hardware implementation has maximum 0.72 ms of latency for every sample, and consumes only 310 mW. To the best of our knowledge, this work is the first implementation of an event-based car classifier on a Neuromorphic Chip."
                },
                {
                    "title": "Self-Supervised Learning of Event-Based Optical Flow with Spiking Neural Networks",
                    "abstract": "The field of neuromorphic computing promises extremely low-power and low-latency sensing and processing. Challenges in transferring learning algorithms from traditional artificial neural networks (ANNs) to spiking neural networks (SNNs) have so far prevented their application to large-scale, complex regression tasks. Furthermore, realizing a truly asynchronous and fully neuromorphic pipeline that maximally attains the abovementioned benefits involves rethinking the way in which this pipeline takes in and accumulates information. In the case of perception, spikes would be passed as-is and one-by-one between an event camera and an SNN, meaning all temporal integration of information must happen inside the network. In this article, we tackle these two problems. We focus on the complex task of learning to estimate optical flow from event-based camera inputs in a self-supervised manner, and modify the state-of-the-art ANN training pipeline to encode minimal temporal information in its inputs. Moreover, we reformulate the self-supervised loss function for event-based optical flow to improve its convexity. We perform experiments with various types of recurrent ANNs and SNNs using the proposed pipeline. Concerning SNNs, we investigate the effects of elements such as parameter initialization and optimization, surrogate gradient shape, and adaptive neuronal mechanisms. We find that initialization and surrogate gradient width play a crucial part in enabling learning with sparse inputs, while the inclusion of adaptivity and learnable neuronal parameters can improve performance. We show that the performance of the proposed ANNs and SNNs are on par with that of the current state-of-the-art ANNs trained in a self-supervised manner."
                },
                {
                    "title": "Transformer Tracking",
                    "abstract": "Correlation acts as a critical role in the tracking field, especially in recent popular Siamese-based trackers. The correlation operation is a simple fusion manner to consider the similarity between the template and the search region. However, the correlation operation itself is a local linear matching process, leading to lose semantic information and fall into local optimum easily, which may be the bottleneck of designing high-accuracy tracking algorithms. Is there any better feature fusion method than correlation? To address this issue, inspired by Transformer, this work presents a novel attention-based feature fusion network, which effectively combines the template and search region features solely using attention. Specifically, the proposed method includes an ego-context augment module based on self-attention and a cross-feature augment module based on cross-attention. Finally, we present a Transformer tracking (named TransT) method based on the Siamese-like feature extraction backbone, the designed attention-based fusion mechanism, and the classification and regression head. Experiments show that our TransT achieves very promising results on six challenging datasets, especially on large-scale LaSOT, TrackingNet, and GOT-10k benchmarks. Our tracker runs at approximatively 50 fps on GPU. Code and models are available at https://github.com/chenxin-dlut/TransT."
                },
                {
                    "title": "Time-Ordered Recent Event (TORE) Volumes for Event Cameras",
                    "abstract": "Event cameras are an exciting, new sensor modality enabling high-speed imaging with extremely low-latency and wide dynamic range. Unfortunately, most machine learning architectures are not designed to directly handle sparse data, like that generated from event cameras. Many state-of-the-art algorithms for event cameras rely on interpolated event representations\u2014obscuring crucial timing information, increasing the data volume, and limiting overall network performance. This paper details an event representation called Time-Ordered Recent Event (TORE) volumes. TORE volumes are designed to compactly store raw spike timing information with minimal information loss. This bio-inspired design is memory efficient, computationally fast, avoids time-blocking (i.e., fixed and predefined frame rates), and contains \u201clocal memory\u201d from past data. The design is evaluated on a wide range of challenging tasks (e.g., event denoising, image reconstruction, classification, and human pose estimation) and is shown to dramatically improve state-of-the-art performance. TORE volumes are an easy-to-implement replacement for any algorithm currently utilizing event representations."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Toward Scalable, Efficient, and Accurate Deep Spiking Neural Networks With Backward Residual Connections, Stochastic Softmax, and Hybridization",
                    "abstract": "Spiking Neural Networks (SNNs) may offer an energy-efficient alternative for implementing deep learning applications. In recent years, there have been several proposals focused on supervised (conversion, spike-based gradient descent) and unsupervised (spike timing dependent plasticity) training methods to improve the accuracy of SNNs on large-scale tasks. However, each of these methods suffer from scalability, latency, and accuracy limitations. In this paper, we propose novel algorithmic techniques of modifying the SNN configuration with backward residual connections, stochastic softmax, and hybrid artificial-and-spiking neuronal activations to improve the learning ability of the training methodologies to yield competitive accuracy, while, yielding large efficiency gains over their artificial counterparts. Note, artificial counterparts refer to conventional deep learning/artificial neural networks. Our techniques apply to VGG/Residual architectures, and are compatible with all forms of training methodologies. Our analysis reveals that the proposed solutions yield near state-of-the-art accuracy with significant energy-efficiency and reduced parameter overhead translating to hardware improvements on complex visual recognition tasks, such as, CIFAR10, Imagenet datatsets."
                },
                {
                    "title": "Probabilistic Regression for Visual Tracking",
                    "abstract": "Visual tracking is fundamentally the problem of regressing the state of the target in each video frame. While significant progress has been achieved, trackers are still prone to failures and inaccuracies. It is therefore crucial to represent the uncertainty in the target estimation. Although current prominent paradigms rely on estimating a state-dependent confidence score, this value lacks a clear probabilistic interpretation, complicating its use. In this work, we therefore propose a probabilistic regression formulation and apply it to tracking. Our network predicts the conditional probability density of the target state given an input image. Crucially, our formulation is capable of modeling label noise stemming from inaccurate annotations and ambiguities in the task. The regression network is trained by minimizing the Kullback-Leibler divergence. When applied for tracking, our formulation not only allows a probabilistic representation of the output, but also substantially improves the performance. Our tracker sets a new state-of-the-art on six datasets, achieving 59.8% AUC on LaSOT and 75.8% Success on TrackingNet. The code and models are available at https://github.com/visionml/pytracking."
                },
                {
                    "title": "SiamFC++: Towards Robust and Accurate Visual Tracking with Target Estimation Guidelines",
                    "abstract": "Visual tracking problem demands to efficiently perform robust classification and accurate target state estimation over a given target at the same time. Former methods have proposed various ways of target state estimation, yet few of them took the particularity of the visual tracking problem itself into consideration. Based on a careful analysis, we propose a set of practical guidelines of target state estimation for high-performance generic object tracker design. Following these guidelines, we design our Fully Convolutional Siamese tracker++ (SiamFC++) by introducing both classification and target state estimation branch (G1), classification score without ambiguity (G2), tracking without prior knowledge (G3), and estimation quality score (G4). Extensive analysis and ablation studies demonstrate the effectiveness of our proposed guidelines. Without bells and whistles, our SiamFC++ tracker achieves state-of-the-art performance on five challenging benchmarks(OTB2015, VOT2018, LaSOT, GOT-10k, TrackingNet), which proves both the tracking and generalization ability of the tracker. Particularly, on the large-scale TrackingNet dataset, SiamFC++ achieves a previously unseen AUC score of 75.4 while running at over 90 FPS, which is far above the real-time requirement."
                },
                {
                    "title": "Event-Based Vision: A Survey",
                    "abstract": "Event cameras are bio-inspired sensors that differ from conventional frame cameras: Instead of capturing images at a fixed rate, they asynchronously measure per-pixel brightness changes, and output a stream of events that encode the time, location and sign of the brightness changes. Event cameras offer attractive properties compared to traditional cameras: high temporal resolution (in the order of $\\mu$\u03bcs), very high dynamic range (140 dB versus 60 dB), low power consumption, and high pixel bandwidth (on the order of kHz) resulting in reduced motion blur. Hence, event cameras have a large potential for robotics and computer vision in challenging scenarios for traditional cameras, such as low-latency, high speed, and high dynamic range. However, novel methods are required to process the unconventional output of these sensors in order to unlock their potential. This paper provides a comprehensive overview of the emerging field of event-based vision, with a focus on the applications and the algorithms developed to unlock the outstanding properties of event cameras. We present event cameras from their working principle, the actual sensors that are available and the tasks that they have been used for, from low-level vision (feature detection and tracking, optic flow, etc.) to high-level vision (reconstruction, segmentation, recognition). We also discuss the techniques developed to process events, including learning-based techniques, as well as specialized processors for these novel sensors, such as spiking neural networks. Additionally, we highlight the challenges that remain to be tackled and the opportunities that lie ahead in the search for a more efficient, bio-inspired way for machines to perceive and interact with the world."
                },
                {
                    "title": "End-to-End Learning of Representations for Asynchronous Event-Based Data",
                    "abstract": "Event cameras are vision sensors that record asynchronous streams of per-pixel brightness changes, referred to as \"events\u201d. They have appealing advantages over frame based cameras for computer vision, including high temporal resolution, high dynamic range, and no motion blur. Due to the sparse, non-uniform spatio-temporal layout of the event signal, pattern recognition algorithms typically aggregate events into a grid-based representation and subsequently process it by a standard vision pipeline, e.g., Convolutional Neural Network (CNN). In this work, we introduce a general framework to convert event streams into grid-based representations by means of strictly differentiable operations. Our framework comes with two main advantages: (i) allows learning the input event representation together with the task dedicated network in an end to end manner, and (ii) lays out a taxonomy that unifies the majority of extant event representations in the literature and identifies novel ones. Empirically, we show that our approach to learning the event representation end-to-end yields an improvement of approximately 12% on optical flow estimation and object recognition over state-of-the-art methods."
                },
                {
                    "title": "Learning Discriminative Model Prediction for Tracking",
                    "abstract": "The current strive towards end-to-end trainable computer vision systems imposes major challenges for the task of visual tracking. In contrast to most other vision problems, tracking requires the learning of a robust target-specific appearance model online, during the inference stage. To be end-to-end trainable, the online learning of the target model thus needs to be embedded in the tracking architecture itself. Due to the imposed challenges, the popular Siamese paradigm simply predicts a target feature template, while ignoring the background appearance information during inference. Consequently, the predicted model possesses limited target-background discriminability. We develop an end-to-end tracking architecture, capable of fully exploiting both target and background appearance information for target model prediction. Our architecture is derived from a discriminative learning loss by designing a dedicated optimization process that is capable of predicting a powerful model in only a few iterations. Furthermore, our approach is able to learn key aspects of the discriminative loss itself. The proposed tracker sets a new state-of-the-art on 6 tracking benchmarks, achieving an EAO score of 0.440 on VOT2018, while running at over 40 FPS. The code and models are available at https://github.com/visionml/pytracking."
                },
                {
                    "title": "Space-Time Event Clouds for Gesture Recognition: From RGB Cameras to Event Cameras",
                    "abstract": "The recently developed event cameras can directly sense the motion in the scene by generating an asynchronous sequence of events, i.e., event streams, where each individual event (x, y, t) corresponds to the space-time location when a pixel sensor captures an intensity change. Compared with RGB cameras, event cameras are frameless but can capture much faster motion, therefore have great potential for recognizing gestures of fast motions. To deal with the unique output of event cameras, previous methods often treat event streams as time sequences, thus do not fully explore the space-time sparsity of the event stream data. In this work, we treat the event stream as a set of 3D points in space-time, i.e., space-time event clouds. To analyze event clouds and recognize gestures, we propose to leverage PointNet, a neural network architecture originally designed for matching and recognizing 3D point clouds. We further adapt PointNet to cater to event clouds for real-time gesture recognition. On the benchmark dataset of event camera based gesture recognition, i.e., IBM DVS128 Gesture dataset, our proposed method achieves a high accuracy of 97.08% and performs the best among existing methods."
                },
                {
                    "title": "Unsupervised Event-Based Learning of Optical Flow, Depth, and Egomotion",
                    "abstract": "In this work, we propose a novel framework for unsupervised learning for event cameras that learns motion information from only the event stream. In particular, we propose an input representation of the events in the form of a discretized volume that maintains the temporal distribution of the events, which we pass through a neural network to predict the motion of the events. This motion is used to attempt to remove any motion blur in the event image. We then propose a loss function applied to the motion compensated event image that measures the motion blur in this image. We train two networks with this framework, one to predict optical flow, and one to predict egomotion and depths, and evaluate these networks on the Multi Vehicle Stereo Event Camera dataset, along with qualitative results from a variety of different scenes."
                },
                {
                    "title": "Event-Based High Dynamic Range Image and Very High Frame Rate Video Generation Using Conditional Generative Adversarial Networks",
                    "abstract": "Event cameras have a lot of advantages over traditional cameras, such as low latency, high temporal resolution, and high dynamic range. However, since the outputs of event cameras are the sequences of asynchronous events over time rather than actual intensity images, existing algorithms could not be directly applied. Therefore, it is demanding to generate intensity images from events for other tasks. In this paper, we unlock the potential of event camera-based conditional generative adversarial networks to create images/videos from an adjustable portion of the event data stream. The stacks of space-time coordinates of events are used as inputs and the network is trained to reproduce images based on the spatio-temporal intensity changes. The usefulness of event cameras to generate high dynamic range (HDR) images even in extreme illumination conditions and also non blurred images under rapid motion is also shown. In addition, the possibility of generating very high frame rate videos is demonstrated, theoretically up to 1 million frames per second(FPS) since the temporal resolution of event cameras is about 1 microsecond. Proposed methods are evaluated by comparing the results with the intensity images captured on the same pixel grid-line of events using online available real datasets and synthetic datasets produced by the event camera simulator."
                },
                {
                    "title": "Event-Based Vision Meets Deep Learning on Steering Prediction for Self-Driving Cars",
                    "abstract": "Event cameras are bio-inspired vision sensors that naturally capture the dynamics of a scene, filtering out redundant information. This paper presents a deep neural network approach that unlocks the potential of event cameras on a challenging motion-estimation task: prediction of a vehicle's steering angle. To make the best out of this sensor-algorithm combination, we adapt state-of-the-art convolutional architectures to the output of event sensors and extensively evaluate the performance of our approach on a publicly available large scale event-camera dataset (~1000 km). We present qualitative and quantitative explanations of why event cameras allow robust steering prediction even in cases where traditional cameras fail, e.g. challenging illumination conditions and fast motion. Finally, we demonstrate the advantages of leveraging transfer learning from traditional to event-based vision, and show that our approach outperforms state-of-the-art algorithms based on standard cameras."
                },
                {
                    "title": "HATS: Histograms of Averaged Time Surfaces for Robust Event-Based Object Classification",
                    "abstract": "Event-based cameras have recently drawn the attention of the Computer Vision community thanks to their advantages in terms of high temporal resolution, low power consumption and high dynamic range, compared to traditional frame-based cameras. These properties make event-based cameras an ideal choice for autonomous vehicles, robot navigation or UAV vision, among others. However, the accuracy of event-based object classification algorithms, which is of crucial importance for any reliable system working in real-world conditions, is still far behind their frame-based counterparts. Two main reasons for this performance gap are: 1. The lack of effective low-level representations and architectures for event-based object classification and 2. The absence of large real-world event-based datasets. In this paper we address both problems. First, we introduce a novel event-based feature representation together with a new machine learning architecture. Compared to previous approaches, we use local memory units to efficiently leverage past temporal information and build a robust event-based representation. Second, we release the first large real-world event-based dataset for object classification. We compare our method to the state-of-the-art with extensive experiments, showing better classification performance and real-time computation."
                },
                {
                    "title": "Loihi: A Neuromorphic Manycore Processor with On-Chip Learning",
                    "abstract": "Loihi is a 60-mm2 chip fabricated in Intels 14-nm process that advances the state-of-the-art modeling of spiking neural networks in silicon. It integrates a wide range of novel features for the field, such as hierarchical connectivity, dendritic compartments, synaptic delays, and, most importantly, programmable synaptic learning rules. Running a spiking convolutional form of the Locally Competitive Algorithm, Loihi can solve LASSO optimization problems with over three orders of magnitude superior energy-delay-product compared to conventional solvers running on a CPU iso-process/voltage/area. This provides an unambiguous example of spike-based computation, outperforming all known conventional solutions."
                },
                {
                    "title": "A Low Power, Fully Event-Based Gesture Recognition System",
                    "abstract": "We present the first gesture recognition system implemented end-to-end on event-based hardware, using a TrueNorth neurosynaptic processor to recognize hand gestures in real-time at low power from events streamed live by a Dynamic Vision Sensor (DVS). The biologically inspired DVS transmits data only when a pixel detects a change, unlike traditional frame-based cameras which sample every pixel at a fixed frame rate. This sparse, asynchronous data representation lets event-based cameras operate at much lower power than frame-based cameras. However, much of the energy efficiency is lost if, as in previous work, the event stream is interpreted by conventional synchronous processors. Here, for the first time, we process a live DVS event stream using TrueNorth, a natively event-based processor with 1 million spiking neurons. Configured here as a convolutional neural network (CNN), the TrueNorth chip identifies the onset of a gesture with a latency of 105 ms while consuming less than 200 mW. The CNN achieves 96.5% out-of-sample accuracy on a newly collected DVS dataset (DvsGesture) comprising 11 hand gesture categories from 29 subjects under 3 illumination conditions."
                },
                {
                    "title": "HOTS: A Hierarchy of Event-Based Time-Surfaces for Pattern Recognition",
                    "abstract": "This paper describes novel event-based spatio-temporal features called time-surfaces and how they can be used to create a hierarchical event-based pattern recognition architecture. Unlike existing hierarchical architectures for pattern recognition, the presented model relies on a time oriented approach to extract spatio-temporal features from the asynchronously acquired dynamics of a visual scene. These dynamics are acquired using biologically inspired frameless asynchronous event-driven vision sensors. Similarly to cortical structures, subsequent layers in our hierarchy extract increasingly abstract features using increasingly large spatio-temporal windows. The central concept is to use the rich temporal information provided by events to create contexts in the form of time-surfaces which represent the recent temporal activity within a local spatial neighborhood. We demonstrate that this concept can robustly be used at all stages of an event-based hierarchical model. First layer feature units operate on groups of pixels, while subsequent layer feature units operate on the output of lower level feature units. We report results on a previously published 36 class character recognition task and a four class canonical dynamic card pip task, achieving near 100 percent accuracy on each. We introduce a new seven class moving face recognition task, achieving 79 percent accuracy."
                },
                {
                    "title": "CIFAR10-DVS: An Event-Stream Dataset for Object Classification",
                    "abstract": "Neuromorphic vision research requires high-quality and appropriately challenging event-stream datasets to support continuous improvement of algorithms and methods. However, creating event-stream datasets is a time-consuming task, which needs to be recorded using the neuromorphic cameras. Currently, there are limited event-stream datasets available. In this work, by utilizing the popular computer vision dataset CIFAR-10, we converted 10,000 frame-based images into 10,000 event streams using a dynamic vision sensor (DVS), providing an event-stream dataset of intermediate difficulty in 10 different classes, named as \u201cCIFAR10-DVS.\u201d The conversion of event-stream dataset was implemented by a repeated closed-loop smooth (RCLS) movement of frame-based images. Unlike the conversion of frame-based images by moving the camera, the image movement is more realistic in respect of its practical applications. The repeated closed-loop image movement generates rich local intensity changes in continuous time which are quantized by each pixel of the DVS camera to generate events. Furthermore, a performance benchmark in event-driven object classification is provided based on state-of-the-art classification algorithms. This work provides a large event-stream dataset and an initial benchmark for comparison, which may boost algorithm developments in even-driven pattern recognition and object classification."
                },
                {
                    "title": "Simultaneous Optical Flow and Intensity Estimation from an Event Camera",
                    "abstract": "Event cameras are bio-inspired vision sensors which mimic retinas to measure per-pixel intensity change rather than outputting an actual intensity image. This proposed paradigm shift away from traditional frame cameras offers significant potential advantages: namely avoiding high data rates, dynamic range limitations and motion blur. Unfortunately, however, established computer vision algorithms may not at all be applied directly to event cameras. Methods proposed so far to reconstruct images, estimate optical flow, track a camera and reconstruct a scene come with severe restrictions on the environment or on the motion of the camera, e.g. allowing only rotation. Here, we propose, to the best of our knowledge, the first algorithm to simultaneously recover the motion field and brightness image, while the camera undergoes a generic motion through any scene. Our approach employs minimisation of a cost function that contains the asynchronous event data as well as spatial and temporal regularisation within a sliding window time interval. Our implementation relies on GPU optimisation and runs in near real-time. In a series of examples, we demonstrate the successful operation of our framework, including in situations where conventional cameras suffer from dynamic range limitations and motion blur."
                },
                {
                    "title": "Spiking Deep Networks with LIF Neurons",
                    "abstract": "We train spiking deep networks using leaky integrate-and-fire (LIF) neurons, and achieve state-of-the-art results for spiking networks on the CIFAR-10 and MNIST datasets. This demonstrates that biologically-plausible spiking LIF neurons can be integrated into deep networks can perform as well as other spiking models (e.g. integrate-and-fire). We achieved this result by softening the LIF response function, such that its derivative remains bounded, and by training the network with noise to provide robustness against the variability introduced by spikes. Our method is general and could be applied to other neuron types, including those used on modern neuromorphic hardware. Our work brings more biological realism into modern image classification models, with the hope that these models can inform how the brain performs this difficult task. It also provides new methods for training deep networks to run on neuromorphic hardware, with the aim of fast, power-efficient image classification for robotics applications."
                },
                {
                    "title": "TrueNorth: Design and Tool Flow of a 65 mW 1 Million Neuron Programmable Neurosynaptic Chip",
                    "abstract": "The new era of cognitive computing brings forth the grand challenge of developing systems capable of processing massive amounts of noisy multisensory data. This type of intelligent computing poses a set of constraints, including real-time operation, low-power consumption and scalability, which require a radical departure from conventional system design. Brain-inspired architectures offer tremendous promise in this area. To this end, we developed TrueNorth, a 65 mW real-time neurosynaptic processor that implements a non-von Neumann, low-power, highly-parallel, scalable, and defect-tolerant architecture. With 4096 neurosynaptic cores, the TrueNorth chip contains 1 million digital neurons and 256 million synapses tightly interconnected by an event-driven routing infrastructure. The fully digital 5.4 billion transistor implementation leverages existing CMOS scaling trends, while ensuring one-to-one correspondence between hardware and software. With such aggressive design metrics and the TrueNorth architecture breaking path with prevailing architectures, it is clear that conventional computer-aided design (CAD) tools could not be used for the design. As a result, we developed a novel design methodology that includes mixed asynchronous-synchronous circuits and a complete tool flow for building an event-driven, low-power neurosynaptic chip. The TrueNorth chip is fully configurable in terms of connectivity and neural parameters to allow custom configurations for a wide range of cognitive and sensory perception applications. To reduce the system's communication energy, we have adapted existing application-agnostic very large-scale integration CAD placement tools for mapping logical neural networks to the physical neurosynaptic core locations on the TrueNorth chips. With that, we have successfully demonstrated the use of TrueNorth-based systems in multiple applications, including visual object recognition, with higher performance and orders of magnitude lower power consumption than the same algorithms run on von Neumann architectures. The TrueNorth chip and its tool flow serve as building blocks for future cognitive systems, and give designers an opportunity to develop novel brain-inspired architectures and systems based on the knowledge obtained from this paper."
                },
                {
                    "title": "Converting Static Image Datasets to Spiking Neuromorphic Datasets Using Saccades",
                    "abstract": "Creating datasets for Neuromorphic Vision is a challenging task. A lack of available recordings from Neuromorphic Vision sensors means that data must typically be recorded specifically for dataset creation rather than collecting and labeling existing data. The task is further complicated by a desire to simultaneously provide traditional frame-based recordings to allow for direct comparison with traditional Computer Vision algorithms. Here we propose a method for converting existing Computer Vision static image datasets into Neuromorphic Vision datasets using an actuated pan-tilt camera platform. Moving the sensor rather than the scene or image is a more biologically realistic approach to sensing and eliminates timing artifacts introduced by monitor updates when simulating motion on a computer monitor. We present conversion of two popular image datasets (MNIST and Caltech101) which have played important roles in the development of Computer Vision, and we provide performance metrics on these datasets using spike-based recognition algorithms. This work contributes datasets for future use in the field, as well as results from spike-based algorithms against which future works can compare. Furthermore, by converting datasets already popular in Computer Vision, we enable more direct comparison with frame-based approaches."
                },
                {
                    "title": "Learning Graph-embedded Key-event Back-tracing for Object Tracking in Event Clouds",
                    "abstract": "Event data-based object tracking is attracting attention increasingly. Unfortunately, the unusual data structure caused by the unique sensing mechanism poses great challenges in designing downstream algorithms. To tackle such challenges, existing methods usually re-organize raw event data (or event clouds) with the event frame/image representation to adapt to mature RGB data-based tracking paradigms, which compromises the high temporal resolution and sparse characteristics. By contrast, we advocate developing new designs/techniques tailored to the special data structure to realize object tracking. To this end, we make the \ufb01rst attempt to construct a new end-to-end learning-based paradigm that directly consumes event clouds. Speci\ufb01cally, to process a non-uniformly distributed large-scale event cloud ef\ufb01ciently, we propose a simple yet effective density-insensitive downsampling strategy to sample a subset called key-events. Then, we employ a graph-based network to embed the irregular spatio-temporal information of key-events into a high-dimensional feature space, and the resulting embeddings are utilized to predict their target likelihoods via semantic-driven Siamese-matching. Besides, we also propose motion-aware target likelihood prediction, which learns the motion \ufb02ow to back-trace the potential initial positions of key-events and measures them with the previous proposal. Finally, we obtain the bounding box by adaptively fusing the two intermediate ones separately regressed from the weighted embeddings of key-events by the two types of predicted target likelihoods. Extensive experiments on both synthetic and real event datasets demonstrate the superiority of the proposed framework over state-of-the-art methods in terms of both the tracking accuracy and speed. The code is publicly available at https://github.com/ZHU-Zhiyu/Event-tracking"
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.NE"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we improve the slicing process of event streams from event-based cameras to enhance the performance of downstream tasks in machine learning?\n\n### [Question 2] - Why is it interesting and important?\nSolving the problem of event stream slicing is crucial for the research community as it addresses a significant gap in the processing pipeline of event-based data. By developing a more effective slicing method, we can enhance the performance of various applications such as object tracking and recognition, which rely heavily on accurate temporal information. This advancement could lead to more robust and efficient machine learning models, fostering further research into adaptive event processing techniques and potentially opening new avenues for real-time applications in robotics, autonomous vehicles, and surveillance systems.\n\n### [Question 3] - Why is it hard?\nThe challenges in improving the slicing process stem from the non-uniformity of information in fixed-sliced event streams, which can lead to either insufficient data capture in low-speed scenarios or excessive redundancy in high-speed conditions. Naive approaches that rely on fixed event counts or time intervals fail to adapt to the dynamic nature of event data, resulting in poor performance in downstream tasks. Additionally, the sensitivity of hyper-parameters, such as time intervals, complicates the process, as they must be finely tuned for each specific application. Overcoming these technical and practical obstacles requires a novel approach that can dynamically adjust slicing based on real-time event characteristics.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has largely focused on refining event representation techniques, neglecting the critical first step of slicing event streams. Existing methods often rely on fixed-group slicing, which does not account for the variability in event distribution. While some recent approaches have attempted adaptive sampling, they still struggle with hyper-parameter tuning and lack a fully learnable and adaptable slicing process. Our approach, SpikeSlicer, differs by utilizing a Spiking Neural Network (SNN) to dynamically determine optimal slicing moments, thus addressing the limitations of prior work and providing a more effective solution.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology, SpikeSlicer, involves training a Spiking Neural Network (SNN) to trigger event slicing at optimal moments. The key components include: (1) the Spiking Position-aware Loss (SPA-Loss) function, which guides the SNN to spike at desired time steps by manipulating membrane potential, and (2) a Feedback-Update training strategy that allows the SNN to"
            }
        },
        "author_data": {
            "52ab0ffc-98b6-47ae-a7d6-1201a45e00e4": {
                "pk": "52ab0ffc-98b6-47ae-a7d6-1201a45e00e4",
                "name": "Jiahang Cao",
                "collaborators": [
                    "Renjing Xu",
                    "Hao Cheng",
                    "Qiang Zhang",
                    "Ziqing Wang",
                    "Erjia Xiao",
                    "Jize Zhang",
                    "Kaidi Xu",
                    "Yuetong Fang",
                    "Le Yang",
                    "Hang Lai"
                ],
                "domain": [
                    "Spiking Neural Networks",
                    "Reinforcement Learning",
                    "Knowledge Graph Embedding",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Knowledge Graph Embedding: A Survey from the Perspective of Representation Spaces",
                        "abstract": "Knowledge graph embedding (KGE) is an increasingly popular technique that aims to represent entities and relations of knowledge graphs into low-dimensional semantic spaces for a wide spectrum of applications such as link prediction, knowledge reasoning and knowledge completion. In this paper, we provide a systematic review of existing KGE techniques based on representation spaces. Particularly, we build a fine-grained classification to categorise the models based on three mathematical perspectives of the representation spaces: (1) Algebraic perspective, (2) Geometric perspective, and (3) Analytical perspective. We introduce the rigorous definitions of fundamental mathematical spaces before diving into KGE models and their mathematical properties. We further discuss different KGE methods over the three categories, as well as summarise how spatial advantages work over different embedding needs. By collating the experimental results from downstream tasks, we also explore the advantages of mathematical space in different scenarios and the reasons behind them. We further state some promising research directions from a representation space perspective, with which we hope to inspire researchers to design their KGE models as well as their related applications with more consideration of their mathematical space properties."
                    },
                    {
                        "title": "Adaptive Calibration: A Unified Conversion Framework of Spiking Neural Networks",
                        "abstract": "Spiking Neural Networks (SNNs) have emerged as a promising energy-efficient alternative to traditional Artificial Neural Networks (ANNs). Despite this, bridging the performance gap with ANNs in practical scenarios remains a significant challenge. This paper focuses on addressing the dual objectives of enhancing the performance and efficiency of SNNs through the established SNN Calibration conversion framework. Inspired by the biological nervous system, we propose a novel Adaptive-Firing Neuron Model (AdaFire) that dynamically adjusts firing patterns across different layers, substantially reducing conversion errors within limited timesteps. Moreover, to meet our efficiency objectives, we propose two novel strategies: an Sensitivity Spike Compression (SSC) technique and an Input-aware Adaptive Timesteps (IAT) technique. These techniques synergistically reduce both energy consumption and latency during the conversion process, thereby enhancing the overall efficiency of SNNs. Extensive experiments demonstrate our approach outperforms state-of-the-art SNNs methods, showcasing superior performance and efficiency in 2D, 3D, and event-driven classification, as well as object detection and segmentation tasks."
                    },
                    {
                        "title": "LoopSR: Looping Sim-and-Real for Lifelong Policy Adaptation of Legged Robots",
                        "abstract": "Reinforcement Learning (RL) has shown its remarkable and generalizable capability in legged locomotion through sim-to-real transfer. However, while adaptive methods like domain randomization are expected to make policy more robust to diverse environments, such comprehensiveness potentially detracts from the policy's performance in any specific environment according to the No Free Lunch theorem, leading to a suboptimal solution once deployed in the real world. To address this issue, we propose a lifelong policy adaptation framework named LoopSR, which utilizes a transformer-based encoder to project real-world trajectories into a latent space, and accordingly reconstruct the real-world environments back in simulation for further improvement. Autoencoder architecture and contrastive learning methods are adopted to better extract the characteristics of real-world dynamics. The simulation parameters for continual training are derived by combining predicted parameters from the decoder with retrieved parameters from the simulation trajectory dataset. By leveraging the continual training, LoopSR achieves superior data efficiency compared with strong baselines, with only a limited amount of data to yield eminent performance in both sim-to-sim and sim-to-real experiments."
                    },
                    {
                        "title": "Gaining the Sparse Rewards by Exploring Lottery Tickets in Spiking Neural Network",
                        "abstract": "Deploying energy-efficient deep learning algorithms on computational-limited devices, such as robots, is still a pressing issue for real-world applications. Spiking Neural Networks (SNNs), a novel brain-inspired algorithm, offer a promising solution due to their low-latency and low-energy properties over traditional Artificial Neural Networks (ANNs). Despite their advantages, the dense structure of deep SNNs can still result in extra energy consumption. The Lottery Ticket Hypothesis (LTH) posits that within dense neural networks, there exist winning Lottery Tickets (LTs), namely sub-networks, that can be obtained without compromising performance. Inspired by this, this paper delves into the spiking-based LTs (SLTs), examining their unique properties and potential for extreme efficiency. Then, two significant sparse \\textbf{\\textit{Rewards}} are gained through comprehensive explorations and meticulous experiments on SLTs across various dense structures. Moreover, a sparse algorithm tailored for spiking transformer structure, which incorporates convolution operations into the Patch Embedding Projection (ConvPEP) module, has been proposed to achieve Multi-level Sparsity (MultiSp). MultiSp refers to (1) Patch number sparsity; (2) ConvPEP weights sparsity and binarization; and (3) ConvPEP activation layer binarization. Extensive experiments demonstrate that our method achieves extreme sparsity with only a slight performance decrease, paving the way for deploying energy-efficient neural networks in robotics and beyond."
                    },
                    {
                        "title": "Spiking Denoising Diffusion Probabilistic Models",
                        "abstract": "Spiking neural networks (SNNs) have ultra-low energy consumption and high biological plausibility due to their binary and bio-driven nature compared with artificial neural networks (ANNs). While previous research has primarily focused on enhancing the performance of SNNs in classification tasks, the generative potential of SNNs remains relatively unexplored. In our paper, we put forward Spiking Denoising Diffusion Probabilistic Models (SDDPM), a new class of SNN-based generative models that achieve high sample quality. To fully exploit the energy efficiency of SNNs, we propose a purely Spiking U-Net architecture, which achieves comparable performance to its ANN counterpart using only 4 time steps, resulting in significantly reduced energy consumption. Extensive experimental results reveal that our approach achieves state-of-the-art on the generative tasks and substantially outperforms other SNN-based generative models, achieving up to 12x and 6x improvement on the CIFAR-10 and the CelebA datasets, respectively. Moreover, we propose a threshold-guided strategy that can further improve the performances by 2.69% in a training-free manner. The SDDPM symbolizes a significant advancement in the field of SNN generation, injecting new perspectives and potential avenues of exploration. Our code is available at https://github.com/AndyCao1125/SDDPM."
                    },
                    {
                        "title": "Spiking Diffusion Models",
                        "abstract": "Recent years have witnessed Spiking Neural Networks (SNNs) gaining attention for their ultra-low energy consumption and high biological plausibility compared with traditional Artificial Neural Networks (ANNs). Despite their distinguished properties, the application of SNNs in the computationally intensive field of image generation is still under exploration. In this paper, we propose the Spiking Diffusion Models (SDMs), an innovative family of SNN-based generative models that excel in producing high-quality samples with significantly reduced energy consumption. In particular, we propose a Temporal-wise Spiking Mechanism (TSM) that allows SNNs to capture more temporal features from a bio-plasticity perspective. In addition, we propose a threshold-guided strategy that can further improve the performances by up to 16.7% without any additional training. We also make the first attempt to use the ANN-SNN approach for SNN-based generation tasks. Extensive experimental results reveal that our approach not only exhibits comparable performance to its ANN counterpart with few spiking time steps, but also outperforms previous SNN-based generative models by a large margin. Moreover, we also demonstrate the high-quality generation ability of SDM on large-scale datasets, e.g., LSUN bedroom. This development marks a pivotal advancement in the capabilities of SNN-based generation, paving the way for future research avenues to realize low-energy and low-latency generative applications. Our code is available at https://github.com/AndyCao1125/SDM."
                    },
                    {
                        "title": "Chasing Day and Night: Towards Robust and Efficient All-Day Object Detection Guided by an Event Camera",
                        "abstract": "The ability to detect objects in all lighting (i.e., normal-, over-, and under-exposed) conditions is crucial for real-world applications, such as self-driving.Traditional RGB-based detectors often fail under such varying lighting conditions.Therefore, recent works utilize novel event cameras to supplement or guide the RGB modality; however, these methods typically adopt asymmetric network structures that rely predominantly on the RGB modality, resulting in limited robustness for all-day detection. In this paper, we propose EOLO, a novel object detection framework that achieves robust and efficient all-day detection by fusing both RGB and event modalities. Our EOLO framework is built based on a lightweight spiking neural network (SNN) to efficiently leverage the asynchronous property of events. Buttressed by it, we first introduce an Event Temporal Attention (ETA) module to learn the high temporal information from events while preserving crucial edge information. Secondly, as different modalities exhibit varying levels of importance under diverse lighting conditions, we propose a novel Symmetric RGB-Event Fusion (SREF) module to effectively fuse RGB-Event features without relying on a specific modality, thus ensuring a balanced and adaptive fusion for all-day detection. In addition, to compensate for the lack of paired RGB-Event datasets for all-day training and evaluation, we propose an event synthesis approach based on the randomized optical flow that allows for directly generating the event frame from a single exposure image. We further build two new datasets, E-MSCOCO and E-VOC based on the popular benchmarks MSCOCO and PASCAL VOC. Extensive experiments demonstrate that our EOLO outperforms the state-of-the-art detectors,e.g.,RENet,by a substantial margin (+3.74% mAP50) in all lighting conditions.Our code and datasets will be available at https://vlislab22.github.io/EOLO/"
                    },
                    {
                        "title": "Masked Spiking Transformer",
                        "abstract": "The combination of Spiking Neural Networks (SNNs) and Transformers has attracted significant attention due to their potential for high energy efficiency and high-performance nature. However, existing works on this topic typically rely on direct training, which can lead to suboptimal performance. To address this issue, we propose to leverage the benefits of the ANN-to-SNN conversion method to combine SNNs and Transformers, resulting in significantly improved performance over existing state-of-the-art SNN models. Furthermore, inspired by the quantal synaptic failures observed in the nervous system, which reduces the number of spikes transmitted across synapses, we introduce a novel Masked Spiking Transformer (MST) framework that incorporates a Random Spike Masking (RSM) method to prune redundant spikes and reduce energy consumption without sacrificing performance. Our experimental results demonstrate that the proposed MST model achieves a significant reduction of 26.8% in power consumption when the masking ratio is 75% while maintaining the same level of performance as the unmasked model."
                    },
                    {
                        "title": "Fully Spiking Neural Network for Legged Robots",
                        "abstract": "Recent advancements in legged robots using deep reinforcement learning have led to significant progress. Quadruped robots can perform complex tasks in challenging environments, while bipedal and humanoid robots have also achieved breakthroughs. Current reinforcement learning methods leverage diverse robot bodies and historical information to perform actions, but previous research has not emphasized the speed and energy consumption of network inference and the biological significance of neural networks. Most networks are traditional artificial neural networks that utilize multilayer perceptrons (MLP). This paper presents a novel Spiking Neural Network (SNN) for legged robots, showing exceptional performance in various simulated terrains. SNNs provide natural advantages in inference speed and energy consumption, and their pulse-form processing enhances biological interpretability. This study presents a highly efficient SNN for legged robots that can be seamless integrated into other learning models."
                    },
                    {
                        "title": "Deep Learning for Stock Selection Based on High Frequency Price-Volume Data",
                        "abstract": "Training a practical and effective model for stock selection has been a greatly concerned problem in the field of artificial intelligence. Even though some of the models from previous works have achieved good performance in the U.S. market by using low-frequency data and features, training a suitable model with high-frequency stock data is still a problem worth exploring. Based on the high-frequency price data of the past several days, we construct two separate models-Convolution Neural Network and Long Short-Term Memory-which can predict the expected return rate of stocks on the current day, and select the stocks with the highest expected yield at the opening to maximize the total return. In our CNN model, we propose improvements on the CNNpred model presented by E. Hoseinzade and S. Haratizadeh in their paper which deals with low-frequency features. Such improvements enable our CNN model to exploit the convolution layer's ability to extract high-level factors and avoid excessive loss of original information at the same time. Our LSTM model utilizes Recurrent Neural Network'advantages in handling time series data. Despite considerable transaction fees due to the daily changes of our stock position, annualized net rate of return is 62.27% for our CNN model, and 50.31% for our LSTM model."
                    },
                    {
                        "title": "Spiking Wavelet Transformer",
                        "abstract": "Spiking neural networks (SNNs) offer an energy-efficient alternative to conventional deep learning by emulating the event-driven processing manner of the brain. Incorporating Transformers with SNNs has shown promise for accuracy. However, they struggle to learn high-frequency patterns, such as moving edges and pixel-level brightness changes, because they rely on the global self-attention mechanism. Learning these high-frequency representations is challenging but essential for SNN-based event-driven vision. To address this issue, we propose the Spiking Wavelet Transformer (SWformer), an attention-free architecture that effectively learns comprehensive spatial-frequency features in a spike-driven manner by leveraging the sparse wavelet transform. The critical component is a Frequency-Aware Token Mixer (FATM) with three branches: 1) spiking wavelet learner for spatial-frequency domain learning, 2) convolution-based learner for spatial feature extraction, and 3) spiking pointwise convolution for cross-channel information aggregation - with negative spike dynamics incorporated in 1) to enhance frequency representation. The FATM enables the SWformer to outperform vanilla Spiking Transformers in capturing high-frequency visual components, as evidenced by our empirical results. Experiments on both static and neuromorphic datasets demonstrate SWformer's effectiveness in capturing spatial-frequency patterns in a multiplication-free and event-driven fashion, outperforming state-of-the-art SNNs. SWformer achieves a 22.03% reduction in parameter count, and a 2.52% performance improvement on the ImageNet dataset compared to vanilla Spiking Transformers. The code is available at: https://github.com/bic-L/Spiking-Wavelet-Transformer."
                    },
                    {
                        "title": "RBFormer: Improve Adversarial Robustness of Transformer by Robust Bias",
                        "abstract": "Recently, there has been a surge of interest and attention in Transformer-based structures, such as Vision Transformer (ViT) and Vision Multilayer Perceptron (VMLP). Compared with the previous convolution-based structures, the Transformer-based structure under investigation showcases a comparable or superior performance under its distinctive attention-based input token mixer strategy. Introducing adversarial examples as a robustness consideration has had a profound and detrimental impact on the performance of well-established convolution-based structures. This inherent vulnerability to adversarial attacks has also been demonstrated in Transformer-based structures. In this paper, our emphasis lies on investigating the intrinsic robustness of the structure rather than introducing novel defense measures against adversarial attacks. To address the susceptibility to robustness issues, we employ a rational structure design approach to mitigate such vulnerabilities. Specifically, we enhance the adversarial robustness of the structure by increasing the proportion of high-frequency structural robust biases. As a result, we introduce a novel structure called Robust Bias Transformer-based Structure (RBFormer) that shows robust superiority compared to several existing baseline structures. Through a series of extensive experiments, RBFormer outperforms the original structures by a significant margin, achieving an impressive improvement of +16.12% and +5.04% across different evaluation criteria on CIFAR-10 and ImageNet-1k, respectively."
                    },
                    {
                        "title": "Pursing the Sparse Limitation of Spiking Deep Learning Structures",
                        "abstract": "Spiking Neural Networks (SNNs), a novel brain-inspired algorithm, are garnering increased attention for their superior computation and energy efficiency over traditional artificial neural networks (ANNs). To facilitate deployment on memory-constrained devices, numerous studies have explored SNN pruning. However, these efforts are hindered by challenges such as scalability challenges in more complex architectures and accuracy degradation. Amidst these challenges, the Lottery Ticket Hypothesis (LTH) emerges as a promising pruning strategy. It posits that within dense neural networks, there exist winning tickets or subnetworks that are sparser but do not compromise performance. To explore a more structure-sparse and energy-saving model, we investigate the unique synergy of SNNs with LTH and design two novel spiking winning tickets to push the boundaries of sparsity within SNNs. Furthermore, we introduce an innovative algorithm capable of simultaneously identifying both weight and patch-level winning tickets, enabling the achievement of sparser structures without compromising on the final model's performance. Through comprehensive experiments on both RGB-based and event-based datasets, we demonstrate that our spiking lottery ticket achieves comparable or superior performance even when the model structure is extremely sparse."
                    },
                    {
                        "title": "Unveiling Typographic Deceptions: Insights of the Typographic Vulnerability in Large Vision-Language Model",
                        "abstract": "Large Vision-Language Models (LVLMs) rely on vision encoders and Large Language Models (LLMs) to exhibit remarkable capabilities on various multi-modal tasks in the joint space of vision and language. However, typographic attacks, which disrupt Vision-Language Models (VLMs) such as Contrastive Language-Image Pretraining (CLIP), have also been expected to be a security threat to LVLMs. Firstly, we verify typographic attacks on current well-known commercial and open-source LVLMs and uncover the widespread existence of this threat. Secondly, to better assess this vulnerability, we propose the most comprehensive and largest-scale Typographic Dataset to date. The Typographic Dataset not only considers the evaluation of typographic attacks under various multi-modal tasks but also evaluates the effects of typographic attacks, influenced by texts generated with diverse factors. Based on the evaluation results, we investigate the causes why typographic attacks impacting VLMs and LVLMs, leading to three highly insightful discoveries. During the process of further validating the rationality of our discoveries, we can reduce the performance degradation caused by typographic attacks from 42.07\\% to 13.90\\%. Code and Dataset are available in \\href{https://github.com/ChaduCheng/TypoDeceptions}"
                    },
                    {
                        "title": "Reinforcement Learning with Generalizable Gaussian Splatting",
                        "abstract": "An excellent representation is crucial for reinforcement learning (RL) performance, especially in vision-based reinforcement learning tasks. The quality of the environment representation directly influences the achievement of the learning task. Previous vision-based RL typically uses explicit or implicit ways to represent environments, such as images, points, voxels, and neural radiance fields. However, these representations contain several drawbacks. They cannot either describe complex local geometries or generalize well to unseen scenes, or require precise foreground masks. Moreover, these implicit neural representations are akin to a ``black box\", significantly hindering interpretability. 3D Gaussian Splatting (3DGS), with its explicit scene representation and differentiable rendering nature, is considered a revolutionary change for reconstruction and representation methods. In this paper, we propose a novel Generalizable Gaussian Splatting framework to be the representation of RL tasks, called GSRL. Through validation in the RoboMimic environment, our method achieves better results than other baselines in multiple tasks, improving the performance by 10%, 44%, and 15% compared with baselines on the hardest task. This work is the first attempt to leverage generalizable 3DGS as a representation for RL."
                    },
                    {
                        "title": "Typography Leads Semantic Diversifying: Amplifying Adversarial Transferability across Multimodal Large Language Models",
                        "abstract": "Recently, Multimodal Large Language Models (MLLMs) achieve remarkable performance in numerous zero-shot tasks due to their outstanding cross-modal interaction and comprehension abilities. However, MLLMs are found to still be vulnerable to human-imperceptible adversarial examples. In the exploration of security vulnerabilities in real-world scenarios, transferability, which can achieve cross-model impact, is considered the greatest threat posed by adversarial examples. However, there is currently no systematic research on the threat of cross-MLLMs adversarial transferability. Therefore, this paper as the first step to provide a comprehensive evaluation of the transferability of adversarial examples generated by various MLLMs. Furthermore, leveraging two key factors that influence transferability performance: 1) The strength of information diversity involved in the adversarial generation process; 2) Editing across vision-language modality information. We propose a boosting method called Typography Augment Transferability Method (TATM) to investigate the adversarial transferability performance across MLLMs further. Through extensive experimental validation, our TATM demonstrates exceptional performance in real-world applications of \"Harmful Word Insertion\" and \"Important Information Protection\"."
                    },
                    {
                        "title": "World Model-based Perception for Visual Legged Locomotion",
                        "abstract": "Legged locomotion over various terrains is challenging and requires precise perception of the robot and its surroundings from both proprioception and vision. However, learning directly from high-dimensional visual input is often data-inefficient and intricate. To address this issue, traditional methods attempt to learn a teacher policy with access to privileged information first and then learn a student policy to imitate the teacher's behavior with visual input. Despite some progress, this imitation framework prevents the student policy from achieving optimal performance due to the information gap between inputs. Furthermore, the learning process is unnatural since animals intuitively learn to traverse different terrains based on their understanding of the world without privileged knowledge. Inspired by this natural ability, we propose a simple yet effective method, World Model-based Perception (WMP), which builds a world model of the environment and learns a policy based on the world model. We illustrate that though completely trained in simulation, the world model can make accurate predictions of real-world trajectories, thus providing informative signals for the policy controller. Extensive simulated and real-world experiments demonstrate that WMP outperforms state-of-the-art baselines in traversability and robustness. Videos and Code are available at: https://wmp-loco.github.io/."
                    }
                ]
            },
            "5427cc7e-b810-44ca-b2ff-a65c5d11c8eb": {
                "pk": "5427cc7e-b810-44ca-b2ff-a65c5d11c8eb",
                "name": "Mingyuan Sun",
                "collaborators": [
                    "Xiaoling Cui",
                    "Xavier Leyronas",
                    "Peng Zhang",
                    "Hui Zhai",
                    "Jia-nan Cui",
                    "Zhengqiang Zhou",
                    "Chang Liu",
                    "Zhe-Yu Shi"
                ],
                "domain": [
                    "Quantum Physics",
                    "Many-Body Physics",
                    "Efimov Effect",
                    "Statistical Mechanics"
                ],
                "publications": [
                    {
                        "title": "Magnetic Manifestation of Discrete Scaling Symmetry in Dirac Semimetals",
                        "abstract": "Two-dimensional and three-dimensional massless Dirac fermions can form a sequence of quasibound states with an attractive charged impurity. These quasibound states exhibit a discrete scaling symmetry, i.e., the energy ratio between two successive states is a constant. Through the calculation of the energy spectrum directly, we find that in two dimension an applied magnetic field can shift or even destroy the quasibound states around the Dirac point and their discrete scaling symmetry disappears. However, as the magnetic field increases, the remaining quasibound states are pushed up to the Dirac point. When one quasibound state is close to the Dirac point, the spectrum is modified significantly, due to the resonant scattering. The magnetic oscillation of the spectrum displays the same discrete scaling symmetry as the quasibound state does in zero magnetic field. This phenomenon also occurs in the quantum limit of three-dimensional Dirac semimetals, where the system becomes quasi one-dimensional essentially. Our theoretical analysis are in good agreement with the recent experimental observations."
                    },
                    {
                        "title": "Landau Level to Efimov-like Bound State Crossover in Two-dimensional Dirac Semimetals",
                        "abstract": "In two-dimensional Dirac semimetals, massless Dirac fermions can form a series of Efimov-like quasibound states in an attractive Coulomb potential. In an applied magnetic field, these quasibound states can become bound states around the Dirac point, due to the energy gap between two successive Landau levels. Through the calculation of the energy spectrum directly, we show that there exists a crossover from the Landau level to the bound state as the magnetic field increases. Because of the decreasing magnetic length, the deep quasibound states are pushed up to the Dirac point relatively and transformed into bound states gradually. Furthermore, the magnetic positions of the emerging bound states in the crossovers obey a geometric scaling law, which can be regarded as an analogy of the radial law in the Efimov effect. We extend our analysis to massless three-component fermions, which also demonstrate this phenomenon. Our results show the universality of fermions with linear dispersion in two dimension and pave a way for the future experimental exploration."
                    },
                    {
                        "title": "Three in Many: Efimov physics in the presence of a Fermi sea",
                        "abstract": "Motivated by recent experiments on $^{6}$Li-$^{133}$Cs atomic mixtures with high mass imbalance, we study the Efimov correlation in atomic system of two heavy bosons ($^{133}$Cs) immersed in a bath of light fermions ($^{6}$Li). Using the Born-Oppenheimer approximation, we identify two different regimes, depending on the Fermi momentum of light fermions ($k_F$) and the boson-fermion scattering length $a_s(<0)$, where the presence of underlying Fermi sea plays distinct roles in the Efimov-type binding of bosons. Namely, in the regime $k_F|a_s|\\lesssim 1$ ($k_F|a_s|\\gtrsim1$), the Fermi sea induces an attractive (repulsive) effective interaction between bosons and thus favors (disfavors) the formation of bound state, which can be seen as the Efimov trimer dressed by the fermion cloud. Interestingly, this implies a non-monotonic behavior of these bound states as increasing the fermion density (or $k_F$). Moreover, we establish a generalized universal scaling law for the emergence/variation of such dressed Efimov bound states when incorporating a new scale ($k_F$) brought by the Fermi sea. These results can be directly tested in Li-Cs cold atoms experiment by measuring the modified bound state spectrum and the shifted Efimov resonance, which manifest an emergent non-trivial Efimov correlation in a fermionic many-body environment."
                    },
                    {
                        "title": "Enhancing the Efimov correlation in Bose polarons with large mass imbalance",
                        "abstract": "We study the effect of Efimov physics (in the few-body sector) to the spectral response of Bose polaron, a many-body system consisting of an impurity immersed in a bath of bosonic atoms. We find that the Efimov correlation can be greatly enhanced by increasing the mass ratio between the bosons and the impurity, which results in visible signatures in the rf spectrum of the polaron. Using a diagrammatic approach up to the third-order virial expansion, we show how the mass imbalance and the enhanced three-body effect modify the line shape and the width of the polaron spectrum. Moreover, we study the effect of a finite boson-boson interaction to the spectrum. Taking the realistic system of Li impurities immersed in Cs bosons with a positive Cs-Cs scattering length, we find a visible Efimov branch, which is associated with the second lowest Efimov trimer, in the polaron spectrum. In particular, by adjusting the boson density the Efimov branch can greatly hybridize with the attractive polaron branch leading to the spectrum broadening near their avoided level crossing. Our results can be directly probed in the cold atoms experiments of Li-Cs and Li-Rb Bose polarons."
                    },
                    {
                        "title": "High-temperature expansion for interacting fermions",
                        "abstract": "We present a general method for the high-temperature expansion of the self-energy of interacting particles. Though the method is valid for fermions and bosons, we illustrate it for spin one half fermions interacting via a zero range potential, in the Bose Einstein Condensate - Bardeen Cooper Schrieffer (BEC-BCS) crossover. The small parameter of the expansion is the fugacity z. Our results include terms of order z and z^2, which take into account respectively two and three body correlations. We give results for the high temperature expansion of Tan's contact at order z^3 in the whole BEC-BCS crossover. We apply our method to calculate the spectral function at the unitary limit. We find new structures which were overlooked by previous approaches, which included only two body correlations. This shows that including three-body correlations can play an important role in the structures of the spectral function."
                    },
                    {
                        "title": "High Temperature Virial Expansion to Universal Quench Dynamics",
                        "abstract": "High temperature virial expansion is a powerful tool in equilibrium statistical mechanics. In this letter we generalize the high temperature virial expansion approach to treat far-from-equilibrium quench dynamics. As an application of our framework, we study the dynamics of a Bose gas quenched from non-interacting to unitarity, and we compare our theoretical results with unexplained experimental results by the Cambridge group [Eigen et al., Nature 563, 221 (2018)]. We show that, during the quench dynamics, the momentum distribution decreases for low-momentum part with $k<k^*$, and increases for high-momentum part with $k>k^*$, where $k^*$ is a characteristic momentum scale separating the low- and the high-momentum regimes. We determine the universal value of $k^*\\lambda$ that agrees perfectly with the experiment, with $\\lambda$ being the thermal de Broglie wave length. We also find a jump of the half-way relaxation time across $k^*\\lambda$ and the non-monotonic behavior of energy distribution, both of which agree with the experiment. Finally, we address the issue whether the long-time steady state thermalizes or not, and we find that this state does thermalize except for the very high momentum tail with $k\\lambda\\gg 1$. Our framework can also be applied to quench dynamics in other systems."
                    },
                    {
                        "title": "Universal Dynamic Scaling and Contact Dynamics in Quenched Quantum Gases",
                        "abstract": "Recently universal dynamic scaling is observed in several systems, which exhibit a spatiotemporal self-similar scaling behavior, analogous to the spatial scaling near phase transitions. The latter arises from the emergent continuous scaling symmetry due to the divergent correlation length. Motivated by this, we investigate the relation between the scaling dynamics and continuous scaling symmetry. We derive a theorem that the scaling invariance of the quenched Hamiltonian and the initial density matrix can lead to the universal dynamic scaling in quench dynamics. It is demonstrated both in the two-body problem analytically and in the many-body problem numerically. For the latter one, we calculate the dynamics of quantum gases quenched from noninteracting to finite interaction in the framework of non-equilibrium high-temperature virial expansion. A dynamic scaling of the momentum distribution appears in certain momentum-time windows at unitarity as well as in the weak interacting limit. Remarkably, this universal scaling dynamics persists approximately with smaller scaling exponents even if the scaling symmetry is fairly broken. Our findings may offer a new perspective to interpret the related experiments. We also calculate the Contact dynamics in the BEC-BCS crossover. Surprisingly, the half-way time displays a maximum near unitarity while some damping oscillations occur on the BEC side due to the dimer state, which can be used to detect possible two-body bound states in experiments."
                    },
                    {
                        "title": "Efimov physics in the complex plane",
                        "abstract": "Efimov effect is characterized by an infinite number of three-body bound states following a universal geometric scaling law at two-body resonances. In this paper, we investigate the influence of two-body loss which can be described by a complex scattering length $a_c$ on these states. Interestingly, because of the complexity of the scattering length $a_c$, the trimer energy is no longer constrained on the negative real axis, and it is allowed to have a nonvanishing imaginary part and a real part which may exceed the three-body or the atom-dimer scattering threshold. Indeed, by taking the $^{133}$Cs-$^{133}$Cs-$^6$Li system as a concrete example, we calculate the trimer energies by solving the generalized Skorniakov-Ter-Martirosian equation and find such three-body bound states with energies that have positive real parts and obey a generalized geometric scaling law. Remarkably, we also find that in some regions these three-body bound states have longer lifetimes compared with the corresponding two-body bound states. The lifetimes for these trimer states can even tend to infinity. Our work paves the way for the future exploration of few-body bound states in the complex plane."
                    }
                ]
            },
            "78ad44bf-92a8-4d95-94c4-87ccfad7d06f": {
                "pk": "78ad44bf-92a8-4d95-94c4-87ccfad7d06f",
                "name": "Ziqing Wang",
                "collaborators": [
                    "Robert Malaney",
                    "Ming Li",
                    "Jonathan Green",
                    "Ryan Aguinaldo",
                    "Benjamin Burnett",
                    "Yuetong Fang",
                    "Jiahang Cao",
                    "Renjing Xu",
                    "Qidong Zhao",
                    "Jinku Cui"
                ],
                "domain": [
                    "Quantum Communication",
                    "Spiking Neural Networks",
                    "Molecular Motors",
                    "Photonics"
                ],
                "publications": [
                    {
                        "title": "Force-velocity relations for multiple-molecular-motor transport",
                        "abstract": "A transition rate model of cargo transport by $N$ molecular motors is proposed. Under the assumption of steady state, the force-velocity curve of multi-motor system can be derived from the force-velocity curve of single motor. Our work shows, in the case of low load, the velocity of multi-motor system can decrease or increase with increasing motor number, which is dependent on the single motor force-velocity curve. And most commonly, the velocity decreases. This gives a possible explanation to some recent"
                    },
                    {
                        "title": "A new cooperation mechanism of kinesin motors when extracting membrane tube",
                        "abstract": "Membrane tubes are important elements for living cells to organize many functions. Experiments have found that membrane tube can be extracted from giant lipid vesicles by a group of kinesin. How these motors cooperate in extracting the fluid-like membrane tube is still unclear. In this paper, we propose a new cooperation mechanism called two-track-dumbbell model, in which kinesin is regarded as a dumbbell with an end (tail domain) tightly bound onto the fluid-like membrane and the other end (head domain) stepping on or unbinding from the microtubule. Taking account of the elasticity of kinesin molecule and the exclude volume effect of both the head domain and the tail domain of kinesin, which are not considered in previous models, we simulate the growth process of the membrane tube pulled by kinesin motors. Our results indicate that motors along a single microtubule protofilament can generate enough force to extract membrane tubes from vesicles, and the average number of motors pulling the tube is about 8~9. These results are quite different from previous studies (Ref. \\cite{camp.08}), and further experimental tests are necessary to elucidate the cooperation mechanism."
                    },
                    {
                        "title": "Inter-satellite Quantum Key Distribution at Terahertz Frequencies",
                        "abstract": "Terahertz (THz) communication is a topic of much research in the context of high-capacity next-generation wireless networks. Quantum communication is also a topic of intensive research, most recently in the context of space-based deployments. In this work we explore the use of THz frequencies as a means to achieve quantum communication within a constellation of micro-satellites in Low-Earth-Orbit (LEO). Quantum communication between the micro-satellite constellation and high-altitude terrestrial stations is also investigated. Our work demonstrates that THz quantum entanglement distribution and THz quantum key distribution are viable deployment options in the micro-satellite context. We discuss how such deployment opens up the possibility for simpler integration of global quantum and wireless networks. The possibility of using THz frequencies for quantum-radar applications in the context of LEO deployments is briefly discussed."
                    },
                    {
                        "title": "Satellite-to-Earth Quantum Key Distribution via Orbital Angular Momentum",
                        "abstract": "In this work, we explore the feasibility of performing satellite-to-Earth quantum key distribution (QKD) using the orbital angular momentum (OAM) of light. Due to the fragility of OAM states the conventional wisdom is that turbulence would render OAM-QKD non-viable in a satellite-to-Earth channel. However, based on detailed phase screen simulations of the anticipated atmospheric turbulence we find that OAM-QKD is viable in some system configurations, especially if quantum channel information is utilized in the processing of post-selected states. More specifically, using classically entangled light as a probe of the quantum channel, and reasonably-sized transmitter-receiver apertures, we find that non-zero QKD rates are achievable on sea-level ground stations. Without using classical light probes, OAM-QKD is relegated to high-altitude ground stations with large receiver apertures. Our work represents the first quantitative assessment of the performance of OAM-QKD from satellites, showing under what circumstances the much-touted higher dimensionality of OAM can be utilized in the context of secure communications."
                    },
                    {
                        "title": "Detecting Orbital Angular Momentum of Light in Satellite-to-Ground Quantum Communications",
                        "abstract": "Satellite-based quantum communications enable a bright future for global-scale information security. However, the spin orbital momentum of light, currently used in many mainstream quantum communication systems, only allows for quantum encoding in a two-dimensional Hilbert space. The orbital angular momentum (OAM) of light, on the other hand, enables quantum encoding in higher-dimensional Hilbert spaces, opening up new opportunities for high-capacity quantum communications. Due to its turbulence-induced decoherence effects, however, the atmospheric channel may limit the practical usage of OAM. In order to determine whether OAM is useful for satellite-based quantum communications, we numerically investigate the detection likelihoods for OAM states that traverse satellite-to-ground channels. We show that the use of OAM through such channels is in fact feasible. We use our new results to then investigate design specifications that could improve OAM detection - particularly the use of advanced adaptive optics techniques. Finally, we discuss how our work provides new insights into future implementations of space-based OAM systems within the context of quantum communications."
                    },
                    {
                        "title": "Temporal Modes of Light in Satellite-to-Earth Quantum Communications",
                        "abstract": "The photonic Temporal Mode (TM) represents a possible candidate for the delivery of viable multidimensional quantum communications. However, relative to other multidimensional quantum information carriers such as the Orbital Angular Momentum (OAM), the TM has received less attention. Moreover, in the context of the emerging quantum internet and satellite-based quantum communications, the TM has received no attention. In this work, we remedy this situation by considering the traversal through the satellite-to-Earth channel of single photons encoded in TM space. Our results indicate that for anticipated atmospheric conditions the photonic TM offers a promising avenue for the delivery of high-throughput quantum communications from a satellite to a terrestrial receiver. In particular, we show how these modes can provide for improved multiplexing performance and superior quantum key distribution in the satellite-to-Earth channel, relative to OAM single-photon states. The levels of TM discrimination that guarantee this outcome are outlined and implications of our results for the emerging satellite-based quantum internet are discussed."
                    },
                    {
                        "title": "Adaptive Calibration: A Unified Conversion Framework of Spiking Neural Networks",
                        "abstract": "Spiking Neural Networks (SNNs) have emerged as a promising energy-efficient alternative to traditional Artificial Neural Networks (ANNs). Despite this, bridging the performance gap with ANNs in practical scenarios remains a significant challenge. This paper focuses on addressing the dual objectives of enhancing the performance and efficiency of SNNs through the established SNN Calibration conversion framework. Inspired by the biological nervous system, we propose a novel Adaptive-Firing Neuron Model (AdaFire) that dynamically adjusts firing patterns across different layers, substantially reducing conversion errors within limited timesteps. Moreover, to meet our efficiency objectives, we propose two novel strategies: an Sensitivity Spike Compression (SSC) technique and an Input-aware Adaptive Timesteps (IAT) technique. These techniques synergistically reduce both energy consumption and latency during the conversion process, thereby enhancing the overall efficiency of SNNs. Extensive experiments demonstrate our approach outperforms state-of-the-art SNNs methods, showcasing superior performance and efficiency in 2D, 3D, and event-driven classification, as well as object detection and segmentation tasks."
                    },
                    {
                        "title": "AutoST: Training-free Neural Architecture Search for Spiking Transformers",
                        "abstract": "Spiking Transformers have gained considerable attention because they achieve both the energy efficiency of Spiking Neural Networks (SNNs) and the high capacity of Transformers. However, the existing Spiking Transformer architectures, derived from Artificial Neural Networks (ANNs), exhibit a notable architectural gap, resulting in suboptimal performance compared to their ANN counterparts. Manually discovering optimal architectures is time-consuming. To address these limitations, we introduce AutoST, a training-free NAS method for Spiking Transformers, to rapidly identify high-performance Spiking Transformer architectures. Unlike existing training-free NAS methods, which struggle with the non-differentiability and high sparsity inherent in SNNs, we propose to utilize Floating-Point Operations (FLOPs) as a performance metric, which is independent of model computations and training dynamics, leading to a stronger correlation with performance. Our extensive experiments show that AutoST models outperform state-of-the-art manually or automatically designed SNN architectures on static and neuromorphic datasets. Full code, model, and data are released for reproduction."
                    },
                    {
                        "title": "Exploiting Spatial Diversity in Earth-to-Satellite Quantum-Classical Communications",
                        "abstract": "Despite being an integral part of the vision of quantum Internet, Earth-to-satellite (uplink) quantum communications have been considered more challenging than their satellite-to-Earth (downlink) counterparts due to the severe channel-loss fluctuations (fading) induced by atmospheric turbulence. The question of how to address the negative impact of fading on Earth-to-satellite quantum communications remains largely an open issue. In this work, we explore the feasibility of exploiting spatial diversity as a means of fading mitigation in Earth-to-satellite Continuous-Variable (CV) quantum-classical optical communications. We demonstrate, via both our theoretical analyses of quantum-state evolution and our detailed numerical simulations of uplink optical channels, that the use of spatial diversity can improve the effectiveness of entanglement distribution through the use of multiple transmitting ground stations and a single satellite with multiple receiving apertures. We further show that the transfer of both large (classically-encoded) and small (quantum-modulated) coherent states can benefit from the use of diversity over fading channels. Our work represents the first quantitative investigation into the use of spatial diversity for satellite-based quantum communications in the uplink direction, showing under what circumstances this fading-mitigation paradigm, which has been widely adopted in classical communications, can be helpful within the context of Earth-to-satellite CV quantum communications."
                    }
                ]
            },
            "a237ca74-95d6-4747-82cc-75de91c6d595": {
                "pk": "a237ca74-95d6-4747-82cc-75de91c6d595",
                "name": "Hao Cheng",
                "collaborators": [
                    "Hao Fang",
                    "Jianfeng Gao",
                    "Mari Ostendorf",
                    "Xiaodong Liu",
                    "Feitian Zhang",
                    "Fan Yang",
                    "Shanxiang Lyu",
                    "Jinming Wen",
                    "Hao Chen",
                    "Qinrui Tang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Robotics",
                    "Machine Learning",
                    "Data Science"
                ],
                "publications": [
                    {
                        "title": "Learning Latent Local Conversation Modes for Predicting Community Endorsement in Online Discussions",
                        "abstract": "Many social media platforms offer a mechanism for readers to react to comments, both positively and negatively, which in aggregate can be thought of as community endorsement. This paper addresses the problem of predicting community endorsement in online discussions, leveraging both the participant response structure and the text of the comment. The different types of features are integrated in a neural network that uses a novel architecture to learn latent modes of discussion structure that perform as well as deep neural networks but are more interpretable. In addition, the latent modes can be used to weight text features thereby improving prediction accuracy."
                    },
                    {
                        "title": "Task-Aware Specialization for Efficient and Robust Dense Retrieval for Open-Domain Question Answering",
                        "abstract": "Given its effectiveness on knowledge-intensive natural language processing tasks, dense retrieval models have become increasingly popular. Specifically, the de-facto architecture for open-domain question answering uses two isomorphic encoders that are initialized from the same pretrained model but separately parameterized for questions and passages. This bi-encoder architecture is parameter-inefficient in that there is no parameter sharing between encoders. Further, recent studies show that such dense retrievers underperform BM25 in various settings. We thus propose a new architecture, Task-aware Specialization for dense Retrieval (TASER), which enables parameter sharing by interleaving shared and specialized blocks in a single encoder. Our experiments on five question answering datasets show that TASER can achieve superior accuracy, surpassing BM25, while using about 60% of the parameters as bi-encoder dense retrievers. In out-of-domain evaluations, TASER is also empirically more robust than bi-encoder dense retrievers. Our code is available at https://github.com/microsoft/taser."
                    },
                    {
                        "title": "RGBlimp-Q: Robotic Gliding Blimp With Moving Mass Control Based on a Bird-Inspired Continuum Arm",
                        "abstract": "Robotic blimps, as lighter-than-air aerial systems, offer prolonged duration and enhanced safety in human-robot interactions due to their buoyant lift. However, robust flight against environmental airflow disturbances remains a significant challenge, limiting the broader application of these robots. Drawing inspiration from the flight mechanics of birds and their ability to perch against natural wind, this article introduces RGBlimp-Q, a robotic gliding blimp equipped with a bird-inspired continuum arm. This arm allows for flexible attitude adjustments through moving mass control to enhance disturbance resilience, while also enabling object capture by using claws to counteract environmental disturbances, similar to a bird. This article presents the design, modeling, and prototyping of RGBlimp-Q, thus extending the advantages of robotic blimps to more complex environments. To the best of the authors' knowledge, this is the first interdisciplinary design integrating continuum mechanisms onto robotic blimps. Experimental results from both indoor and outdoor settings validate the improved flight robustness against environmental disturbances offered by this novel design."
                    },
                    {
                        "title": "Lattice-Aided Extraction of Spread-Spectrum Hidden Data",
                        "abstract": "This paper discusses the problem of extracting spread spectrum hidden data from the perspective of lattice decoding. Since the conventional blind extraction scheme multi-carrier iterative generalize least-squares (M-IGLS) and non-blind extraction scheme minimum mean square error (MMSE) suffer from performance degradation when the carriers lack sufficient orthogonality, we present two novel schemes from the viewpoint of lattice decoding, namely multi-carrier iterative successive interference cancellation (M-ISIC) and sphere decoding (SD). The better performance of M-ISIC and SD are confirmed by both theoretical justification and numerical simulations."
                    },
                    {
                        "title": "Feature Pyramid biLSTM: Using Smartphone Sensors for Transportation Mode Detection",
                        "abstract": "The widespread utilization of smartphones has provided extensive availability to Inertial Measurement Units, providing a wide range of sensory data that can be advantageous for the detection of transportation modes. The objective of this study is to propose a novel end-to-end approach to effectively explore a reduced amount of sensory data collected from a smartphone to achieve accurate mode detection in common daily traveling activities. Our approach, called Feature Pyramid biLSTM (FPbiLSTM), is characterized by its ability to reduce the number of sensors required and processing demands, resulting in a more efficient modeling process without sacrificing the quality of the outcomes than the other current models. FPbiLSTM extends an existing CNN biLSTM model with the Feature Pyramid Network, leveraging the advantages of both shallow layer richness and deeper layer feature resilience for capturing temporal moving patterns in various transportation modes. It exhibits an excellent performance by employing the data collected from only three out of seven sensors, i.e. accelerometers, gyroscopes, and magnetometers, in the 2018 Sussex-Huawei Locomotion (SHL) challenge dataset, attaining a noteworthy accuracy of 95.1% and an F1-score of 94.7% in detecting eight different transportation modes."
                    },
                    {
                        "title": "Improving Take-over Situation by Active Communication",
                        "abstract": "In this short paper an idea is sketched, how to support drivers of an autonomous vehicle in taking back control of the vehicle after a longer section of autonomous cruising. The hypothesis is that a clear communication about the location and behavior of relevant objects in the environment will help the driver to quickly grasp the situational context and thus support drivers in safely handling the ongoing driving situation manually after take-over. Based on this hypothesis, a research concept is sketched, which entails the necessary components as well as the disciplines involved."
                    },
                    {
                        "title": "An inverse random source problem for the Helium production-diffusion equation driven by a fractional Brownian motion",
                        "abstract": "In this paper, we consider the prediction of the helium concentrations as function of a spatially variable source term perturbed by fractional Brownian motion. For the direct problem, we show that it is well-posed and has a unique mild solution under some conditions. For the inverse problem, the uniqueness and the instability are given. In the meanwhile, we determine the statistical properties of the source from the expectation and covariance of the final-time data u(r,T). Finally, numerical implements are given to verify the effectiveness of the proposed reconstruction."
                    },
                    {
                        "title": "Bi-directional Attention with Agreement for Dependency Parsing",
                        "abstract": "We develop a novel bi-directional attention model for dependency parsing, which learns to agree on headword predictions from the forward and backward parsing directions. The parsing procedure for each direction is formulated as sequentially querying the memory component that stores continuous headword embeddings. The proposed parser makes use of {\\it soft} headword embeddings, allowing the model to implicitly capture high-order parsing history without dramatically increasing the computational complexity. We conduct experiments on English, Chinese, and 12 other languages from the CoNLL 2006 shared task, showing that the proposed model achieves state-of-the-art unlabeled attachment scores on 6 languages."
                    }
                ]
            },
            "0275bd9b-324e-4a7c-b479-88664ae79ab2": {
                "pk": "0275bd9b-324e-4a7c-b479-88664ae79ab2",
                "name": "Qiang Zhang",
                "collaborators": [
                    "Xuezhi Zhao",
                    "Dongxiao Zhao",
                    "Bin Yan",
                    "Peng Wang",
                    "Jianchun Wu",
                    "Shengkui Ye",
                    "Jiangping Hu",
                    "Jialin Lei",
                    "Yingting Miao"
                ],
                "domain": [
                    "Topology",
                    "Group Theory",
                    "Quantum Mechanics",
                    "Mechanism Design"
                ],
                "publications": [
                    {
                        "title": "The fixed subgroups of homeomorphisms of Seifert manifolds",
                        "abstract": "Let $M$ be a compact connected orientable Seifert manifold with hyperbolic orbifold $B_M$, and $f_{\\pi}: \\pi_1(M)\\rightarrow\\pi_1(M)$ be an automorphism induced by an orientation-reversing homeomorphism $f$ of $M$. We give a bound on the rank of the fixed subgroup of $f_{\\pi}$, namely, $\\rank\\fix(f_{\\pi})<2\\rank \\pi_1(M)$, which is similar to the inequalities on surface groups and hyperbolic 3-manifold groups."
                    },
                    {
                        "title": "Mechanism Design with Exchangeable Allocations",
                        "abstract": "We investigate mechanism design without payments when agents have different types of preferences. Contrary to most settings in the literature where agents have the same preference, e.g. in the facility location games all agents would like to stay close to (or away from) the facility, we demonstrate the limitation of mechanism design without payments when agents have different preferences by introducing exchanging phases. We consider two types of exchanging phases. The first model is called central exchanges where the exchanges are performed by a central authority. The other model is called individual exchanges where agents exchange their outcomes by themselves. By using facility location games as an example, we provide a truthful mechanism that optimizes social welfare in central exchanges. We also provide a universally truthful randomized mechanism that achieves at least a half of the optimal social welfare in individual exchanges."
                    },
                    {
                        "title": "Aspherical manifolds which do not have Bounded Index Property",
                        "abstract": "We prove that some aspherical manifolds do not have BIP, negating a question of Jiang."
                    },
                    {
                        "title": "Iwip endomorphisms of free groups and fixed points of graph selfmaps",
                        "abstract": "In the paper \\cite{JWZ}, Jiang, Wang and Zhang gave a bound for the fixed points and fixed subgroups of a graph selfmap. In this paper, we give a sufficient condition when the bound can be reached, by studying iwip outer endomorphisms of free groups acting on the outer spaces."
                    },
                    {
                        "title": "A note on test elements for monomorphisms of free groups",
                        "abstract": "A word in a group is called a test element if any endomorphism fixing it is necessarily an automorphism. In this note, we give a sufficient condition in geometry to construct test elements for monomorphisms of a free group, by using the Whitehead graph and the action of the free group on its Cayley graph."
                    },
                    {
                        "title": "The group fixed by a family of endomorphisms of a surface group",
                        "abstract": "For a closed surface $S$ with $\\chi(S)<0$, we show that the fixed subgroup of a family $\\mathcal B$ of endomorphisms of $\\pi_1(S)$ has $\\rk \\fix\\mathcal B\\leq \\rk \\pi_1(S)$. In particular, if $\\mathcal B$ contains a non-epimorphic endomorphism, then $\\rk \\fix\\mathcal B\\leq \\frac{1}{2} \\rk \\pi_1(S)$. We also show that geometric subgroups of $\\pi_1(S)$ are inert, and hence the fixed subgroup of a family of epimorphisms of $\\pi_1(S)$ is also inert."
                    },
                    {
                        "title": "Many-Anyons Wavefunction, State Capacity and Gentile Statistics",
                        "abstract": "The many-anyons wavefunction is constructed via the superposition of all the permutations on the direct product of single anyon states and its interchange properties are examined. The phase of permutation is not a representation but the word metric of the permutation group . Amazingly the interchange phase yields a finite capacity of one quantum state interpolating between Fermion and Boson and the mutual exchange phase has no explicit effect on statistics. Finite capacity of quantum state is defined as Gentile statistics and it is different from the fractional exclusion statistics. Some discussion on the general model is also given."
                    },
                    {
                        "title": "Bounds for fixed points on products of hyperbolic surfaces",
                        "abstract": "For the product $S_1\\times S_2$ of any two connected compact hyperbolic surfaces $S_1$ and $S_2$, we give a finite bound $\\mathcal{B}$ such that for any self-homeomorphism $f$ of $S_1\\times S_2$ and any fixed point class $F$ of $f$, the index $|ind(f, F)|\\leq \\mathcal{B}$, which is an affirmative answer for a special case of a question asked by Boju Jiang. Moreover, we also give bounds for the Lefschetz number $L(f)$ and the Nielsen number $N(f)$ of the homeomorphism $f$."
                    },
                    {
                        "title": "On the bounded index property for products of aspherical polyhedra",
                        "abstract": "A compact polyhedron $X$ is said to have the Bounded Index Property for Homotopy Equivalence (BIPHE) if there is a finite bound $\\mathcal{B}$ such that for any homotopy equivalence $f:X\\rightarrow X$ and any fixed point class $\\mathbf{F}$ of $f$, the index $|\\mathrm{ind}(f,\\mathbf{F})|\\leq \\mathcal{B}$. In this note, we consider the product of compact polyhedra, and give some sufficient conditions for it to have BIPHE. Moreover, we show that the products of closed Riemannian manifolds with negative sectional curvature, in particular hyperbolic manifolds, have BIPHE, which gives an affirmative answer to a special case of a question asked by Boju Jiang."
                    },
                    {
                        "title": "Longitudinal modes of spin fluctuations in iron-based superconductors",
                        "abstract": "Iron-based superconductors can exhibit different magnetic ground states and are in a critical magnetic region where frustrated magnetic interactions strongly compete with each other. Here we investigate the longitudinal modes of spin fluctuations in an unified effective magnetic model for iron-based superconductors. We focus on the collinear antiferromagnetic phase (CAF) and calculate the behavior of the longitudinal modes when different phase boundaries are approached. The results can help to determine the nature of the magnetic fluctuations in iron-based superconductors."
                    },
                    {
                        "title": "Explicit bounds for fixed subgroups of endomorphisms of free products",
                        "abstract": "For an automorphism $\\phi$ of a free group $F_n$ of rank $n$, Bestvina and Handel showed that the rank $rk Fix(\\phi)$ of the fixed subgroup is not greater than $n$ (the so-called Scott conjecture). Soon after Bestvina and Handel's announcement, their result was generalized by many authors in various directions. In this paper, we are interested in the fixed subgroups of endomorphisms of free products, focusing on explicit bounds for their ranks."
                    },
                    {
                        "title": "Optimal Investment and Consumption Strategies with General and Linear Transaction Costs under CRRA Utility",
                        "abstract": "Transaction costs play a critical role in asset allocation and consumption strategies in portfolio management. We apply the methods of dynamic programming and singular perturbation expansion to derive the closed-form leading solutions to this problem for small transaction costs with arbitrary transaction cost structure by maximizing the expected CRRA (constant relative risk aversion) utility function for this problem. We also discuss in detail the case which consists of both fixed and proportional transaction costs."
                    },
                    {
                        "title": "Howson groups which are not strongly Howson",
                        "abstract": "A group $G$ is called a Howson group if the intersection $H\\cap K$ of any two finitely generated subgroups $H, K<G$ is again finitely generated, and called a strongly Howson group when a uniform bound for the rank of $H\\cap K$ can be obtained from the ranks of $H$ and $K$. Clearly, every strongly Howson group is a Howson group, but it is unclear in the literature whether the converse is true. In this note, we show that the converse is not true by constructing the first Howson groups which are not strongly Howson."
                    },
                    {
                        "title": "Fixed point indices and fixed words at infinity of selfmaps of graphs",
                        "abstract": "Indices of fixed point classes play a central role in Nielsen fixed point theory. Jiang-Wang-Zhang proved that for selfmaps of graphs and surfaces, the index of any fixed point class has an upper bound called its characteristic.   In this paper, we study the difference between the index and the characteristic for selfmaps of graphs. First, for free groups, we extend attracting fixed words at infinity of automorphisms into that of injective endomorphisms. Then, by using relative train track technique, we show that the difference mentioned above is quite likely to be the number of equivalence classes of attracting fixed words of the endomorphism induced on the fundamental group. Since both of attracting fixed words and the existed characteristic are totally determined by endomorphisms themselves, we give a new algebraic approach to estimate indices of fixed point classes of graph selfmaps.   As consequence, we obtain an upper bound for attracting fixed words of injective endomorphisms of free groups, generalizing the one for automorphisms due to Gaboriau-Jaeger-Levitt-Lustig. Furthermore, we give a simple approach to roughly detecting whether fixed words exist or not."
                    },
                    {
                        "title": "Permutation Phase and Gentile Statistics",
                        "abstract": "This paper presents a new way to construct single-valued many-body wavefunctions of identical particles with intermediate exchange phases between Fermi and Bose statistics. It is demonstrated that the exchange phase is not a representation character but the \\textit{word metric} of the permutation group, beyond the anyon phase from the braiding group in two dimensions. By constructing this type of wavefunction from the direct product of single-particle states, it is shown that a finite \\textit{capacity q} -- the maximally allowed particle occupation of each quantum state, naturally arises. The relation between the permutation phase and capacity is given, interpolating between fermions and bosons in the sense of both exchange phase and occupation number. This offers a quantum mechanics foundation for \\textit{Gentile statistics} and new directions to explore intermediate statistics and anyons."
                    }
                ]
            },
            "94518ba8-c75c-48ec-b018-7f277f19f6d5": {
                "pk": "94518ba8-c75c-48ec-b018-7f277f19f6d5",
                "name": "Shibo Zhou",
                "collaborators": [
                    "Xiaohua Li",
                    "Wei Wang",
                    "Zhanpeng Jin",
                    "Ying Chen",
                    "Arindam Sanyal",
                    "Yu Pan",
                    "Bo Yang",
                    "Mengwen Yuan",
                    "Runhao Jiang",
                    "Rui Yan"
                ],
                "domain": [
                    "Spiking Neural Networks",
                    "Time Series Classification",
                    "Autonomous Driving",
                    "Energy Efficiency"
                ],
                "publications": [
                    {
                        "title": "Spectrum Attention Mechanism for Time Series Classification",
                        "abstract": "Time series classification(TSC) has always been an important and challenging research task. With the wide application of deep learning, more and more researchers use deep learning models to solve TSC problems. Since time series always contains a lot of noise, which has a negative impact on network training, people usually filter the original data before training the network. The existing schemes are to treat the filtering and training as two stages, and the design of the filter requires expert experience, which increases the design difficulty of the algorithm and is not universal. We note that the essence of filtering is to filter out the insignificant frequency components and highlight the important ones, which is similar to the attention mechanism. In this paper, we propose an attention mechanism that acts on spectrum (SAM). The network can assign appropriate weights to each frequency component to achieve adaptive filtering. We use L1 regularization to further enhance the frequency screening capability of SAM. We also propose a segmented-SAM (SSAM) to avoid the loss of time domain information caused by using the spectrum of the whole sequence. In which, a tumbling window is introduced to segment the original data. Then SAM is applied to each segment to generate new features. We propose a heuristic strategy to search for the appropriate number of segments. Experimental results show that SSAM can produce better feature representations, make the network converge faster, and improve the robustness and classification accuracy."
                    },
                    {
                        "title": "Object Detection based on LIDAR Temporal Pulses using Spiking Neural Networks",
                        "abstract": "Neural networks has been successfully used in the processing of Lidar data, especially in the scenario of autonomous driving. However, existing methods heavily rely on pre-processing of the pulse signals derived from Lidar sensors and therefore result in high computational overhead and considerable latency. In this paper, we proposed an approach utilizing Spiking Neural Network (SNN) to address the object recognition problem directly with raw temporal pulses. To help with the evaluation and benchmarking, a comprehensive temporal pulses data-set was created to simulate Lidar reflection in different road scenarios. Being tested with regard to recognition accuracy and time efficiency under different noise conditions, our proposed method shows remarkable performance with the inference accuracy up to 99.83% (with 10% noise) and the average recognition delay as low as 265 ns. It highlights the potential of SNN in autonomous driving and some related applications. In particular, to our best knowledge, this is the first attempt to use SNN to directly perform object recognition on raw Lidar temporal pulses."
                    },
                    {
                        "title": "Spiking Neural Networks with Single-Spike Temporal-Coded Neurons for Network Intrusion Detection",
                        "abstract": "Spiking neural network (SNN) is interesting due to its strong bio-plausibility and high energy efficiency. However, its performance is falling far behind conventional deep neural networks (DNNs). In this paper, considering a general class of single-spike temporal-coded integrate-and-fire neurons, we analyze the input-output expressions of both leaky and nonleaky neurons. We show that SNNs built with leaky neurons suffer from the overly-nonlinear and overly-complex input-output response, which is the major reason for their difficult training and low performance. This reason is more fundamental than the commonly believed problem of nondifferentiable spikes. To support this claim, we show that SNNs built with nonleaky neurons can have a less-complex and less-nonlinear input-output response. They can be easily trained and can have superior performance, which is demonstrated by experimenting with the SNNs over two popular network intrusion detection datasets, i.e., the NSL-KDD and the AWID datasets. Our experiment results show that the proposed SNNs outperform a comprehensive list of DNN models and classic machine learning models. This paper demonstrates that SNNs can be promising and competitive in contrast to common beliefs."
                    },
                    {
                        "title": "A Spike Learning System for Event-driven Object Recognition",
                        "abstract": "Event-driven sensors such as LiDAR and dynamic vision sensor (DVS) have found increased attention in high-resolution and high-speed applications. A lot of work has been conducted to enhance recognition accuracy. However, the essential topic of recognition delay or time efficiency is largely under-explored. In this paper, we present a spiking learning system that uses the spiking neural network (SNN) with a novel temporal coding for accurate and fast object recognition. The proposed temporal coding scheme maps each event's arrival time and data into SNN spike time so that asynchronously-arrived events are processed immediately without delay. The scheme is integrated nicely with the SNN's asynchronous processing capability to enhance time efficiency. A key advantage over existing systems is that the event accumulation time for each recognition task is determined automatically by the system rather than pre-set by the user. The system can finish recognition early without waiting for all the input events. Extensive experiments were conducted over a list of 7 LiDAR and DVS datasets. The results demonstrated that the proposed system had state-of-the-art recognition accuracy while achieving remarkable time efficiency. Recognition delay was shown to reduce by 56.3% to 91.7% in various experiment settings over the popular KITTI dataset."
                    },
                    {
                        "title": "Deep SCNN-based Real-time Object Detection for Self-driving Vehicles Using LiDAR Temporal Data",
                        "abstract": "Real-time accurate detection of three-dimensional (3D) objects is a fundamental necessity for self-driving vehicles. Most existing computer vision approaches are based on convolutional neural networks (CNNs). Although the CNN-based approaches can achieve high detection accuracy, their high energy consumption is a severe drawback. To resolve this problem, novel energy efficient approaches should be explored. Spiking neural network (SNN) is a promising candidate because it has orders-of-magnitude lower energy consumption than CNN. Unfortunately, the studying of SNN has been limited in small networks only. The application of SNN for large 3D object detection networks has remain largely open. In this paper, we integrate spiking convolutional neural network (SCNN) with temporal coding into the YOLOv2 architecture for real-time object detection. To take the advantage of spiking signals, we develop a novel data preprocessing layer that translates 3D point-cloud data into spike time data. We propose an analog circuit to implement the non-leaky integrate and fire neuron used in our SCNN, from which the energy consumption of each spike is estimated. Moreover, we present a method to calculate the network sparsity and the energy consumption of the overall network. Extensive experiments have been conducted based on the KITTI dataset, which show that the proposed network can reach competitive detection accuracy as existing approaches, yet with much lower average energy consumption. If implemented in dedicated hardware, our network could have a mean sparsity of 56.24% and extremely low total energy consumption of 0.247mJ only. Implemented in NVIDIA GTX 1080i GPU, we can achieve 35.7 fps frame rate, high enough for real-time object detection."
                    },
                    {
                        "title": "Enhancing SNN-based Spatio-Temporal Learning: A Benchmark Dataset and Cross-Modality Attention Model",
                        "abstract": "Spiking Neural Networks (SNNs), renowned for their low power consumption, brain-inspired architecture, and spatio-temporal representation capabilities, have garnered considerable attention in recent years. Similar to Artificial Neural Networks (ANNs), high-quality benchmark datasets are of great importance to the advances of SNNs. However, our analysis indicates that many prevalent neuromorphic datasets lack strong temporal correlation, preventing SNNs from fully exploiting their spatio-temporal representation capabilities. Meanwhile, the integration of event and frame modalities offers more comprehensive visual spatio-temporal information. Yet, the SNN-based cross-modality fusion remains underexplored.   In this work, we present a neuromorphic dataset called DVS-SLR that can better exploit the inherent spatio-temporal properties of SNNs. Compared to existing datasets, it offers advantages in terms of higher temporal correlation, larger scale, and more varied scenarios. In addition, our neuromorphic dataset contains corresponding frame data, which can be used for developing SNN-based fusion methods. By virtue of the dual-modal feature of the dataset, we propose a Cross-Modality Attention (CMA) based fusion method. The CMA model efficiently utilizes the unique advantages of each modality, allowing for SNNs to learn both temporal and spatial attention scores from the spatio-temporal features of event and frame modalities, subsequently allocating these scores across modalities to enhance their synergy. Experimental results demonstrate that our method not only improves recognition accuracy but also ensures robustness across diverse scenarios."
                    },
                    {
                        "title": "Temporal Pulses Driven Spiking Neural Network for Fast Object Recognition in Autonomous Driving",
                        "abstract": "Accurate real-time object recognition from sensory data has long been a crucial and challenging task for autonomous driving. Even though deep neural networks (DNNs) have been successfully applied in this area, most existing methods still heavily rely on the pre-processing of the pulse signals derived from LiDAR sensors, and therefore introduce additional computational overhead and considerable latency. In this paper, we propose an approach to address the object recognition problem directly with raw temporal pulses utilizing the spiking neural network (SNN). Being evaluated on various datasets (including Sim LiDAR, KITTI and DVS-barrel) derived from LiDAR and dynamic vision sensor (DVS), our proposed method has shown comparable performance as the state-of-the-art methods, while achieving remarkable time efficiency. It highlights the SNN's great potentials in autonomous driving and related applications. To the best of our knowledge, this is the first attempt to use SNN to directly perform object recognition on raw temporal pulses."
                    },
                    {
                        "title": "Temporal-Coded Deep Spiking Neural Network with Easy Training and Robust Performance",
                        "abstract": "Spiking neural network (SNN) is interesting both theoretically and practically because of its strong bio-inspiration nature and potentially outstanding energy efficiency. Unfortunately, its development has fallen far behind the conventional deep neural network (DNN), mainly because of difficult training and lack of widely accepted hardware experiment platforms. In this paper, we show that a deep temporal-coded SNN can be trained easily and directly over the benchmark datasets CIFAR10 and ImageNet, with testing accuracy within 1% of the DNN of equivalent size and architecture. Training becomes similar to DNN thanks to the closed-form solution to the spiking waveform dynamics. Considering that SNNs should be implemented in practical neuromorphic hardwares, we train the deep SNN with weights quantized to 8, 4, 2 bits and with weights perturbed by random noise to demonstrate its robustness in practical applications. In addition, we develop a phase-domain signal processing circuit schematic to implement our spiking neuron with 90% gain of energy efficiency over existing work. This paper demonstrates that the temporal-coded deep SNN is feasible for applications with high performance and high energy efficient."
                    }
                ]
            },
            "736e17d3-19c7-4e77-bb64-3a3bbeda54d1": {
                "pk": "736e17d3-19c7-4e77-bb64-3a3bbeda54d1",
                "name": "Renjing Xu",
                "collaborators": [
                    "Qiang Zhang",
                    "Jiahang Cao",
                    "Jiaxu Wang",
                    "Ziqing Wang",
                    "Yuetong Fang",
                    "Ziyi Zhang",
                    "Yiqun Duan",
                    "Junhao He",
                    "Hao Cheng",
                    "Erjia Xiao"
                ],
                "domain": [
                    "Spiking Neural Networks",
                    "Reinforcement Learning",
                    "Computer Vision",
                    "Quantum Mechanics"
                ],
                "publications": [
                    {
                        "title": "Review of Strong Field Approximation and Investigating Semiclassical Evolution Approach",
                        "abstract": "This paper theoretically analyzes the behavior of an atom driven by a strong electro-magnetic field. Moreover, besides traditional quantum mechanics method, we also investigate semiclassical approaches to this problem. We first performed strong field approximation for system of an atom driven by a strong electromagnetic field in velocity gauge. Our simulation result is consistent with theories and close to experiments except some reasonable difference caused by different parameters and omitted bound and final states in the transition amplitude. Next, a new semiclassical approach is used to solve Volkov wave function. We prove that semiclassical approximation works well in predicting the particle evolution in quantum world, especially for system in a strong electromagnetic field with low frequency. Finally, we also briefly illustrated how to use semiclassical approximation to get the same results as strong field approximation and also partly tested the viability of the semiclassical approach to Teller Potential. This part needs future work to accomplish. But still, semiclassical approximation provides a potentially new method to solve complicated system, which might be more effective than traditional quantum mechanics recipe."
                    },
                    {
                        "title": "Adaptive Calibration: A Unified Conversion Framework of Spiking Neural Networks",
                        "abstract": "Spiking Neural Networks (SNNs) have emerged as a promising energy-efficient alternative to traditional Artificial Neural Networks (ANNs). Despite this, bridging the performance gap with ANNs in practical scenarios remains a significant challenge. This paper focuses on addressing the dual objectives of enhancing the performance and efficiency of SNNs through the established SNN Calibration conversion framework. Inspired by the biological nervous system, we propose a novel Adaptive-Firing Neuron Model (AdaFire) that dynamically adjusts firing patterns across different layers, substantially reducing conversion errors within limited timesteps. Moreover, to meet our efficiency objectives, we propose two novel strategies: an Sensitivity Spike Compression (SSC) technique and an Input-aware Adaptive Timesteps (IAT) technique. These techniques synergistically reduce both energy consumption and latency during the conversion process, thereby enhancing the overall efficiency of SNNs. Extensive experiments demonstrate our approach outperforms state-of-the-art SNNs methods, showcasing superior performance and efficiency in 2D, 3D, and event-driven classification, as well as object detection and segmentation tasks."
                    },
                    {
                        "title": "Simple Policy Optimization",
                        "abstract": "As one of the most important and influential algorithms in reinforcement learning, the Proximal Policy Optimization (PPO) algorithm has demonstrated outstanding performance across various domains. It simplifies the optimization-based importance sampling process of the Trust Region Policy Optimization (TRPO) algorithm through ratio clipping. However, this simplification with ratio clipping does not always effectively enforce trust region constraints. In this paper, we introduce an algorithm named \\textit{Simple Policy Optimization} (SPO), which incorporates a novel clipping method for the KL divergence between the old and new policies. Extensive experimental results in both \\textit{Atari 2600} and \\textit{MuJoCo} environments show that, compared to PPO, SPO achieves better sample efficiency, extremely low KL divergence, and higher policy entropy, while also being robust to increases in network depth or complexity. More importantly, SPO maintains the simplicity of an unconstrained first-order algorithm. Our code is available at https://github.com/MyRepositories-hub/Simple-Policy-Optimization."
                    },
                    {
                        "title": "Learning Robust Generalizable Radiance Field with Visibility and Feature Augmented Point Representation",
                        "abstract": "This paper introduces a novel paradigm for the generalizable neural radiance field (NeRF). Previous generic NeRF methods combine multiview stereo techniques with image-based neural rendering for generalization, yielding impressive results, while suffering from three issues. First, occlusions often result in inconsistent feature matching. Then, they deliver distortions and artifacts in geometric discontinuities and locally sharp shapes due to their individual process of sampled points and rough feature aggregation. Third, their image-based representations experience severe degradations when source views are not near enough to the target view. To address challenges, we propose the first paradigm that constructs the generalizable neural field based on point-based rather than image-based rendering, which we call the Generalizable neural Point Field (GPF). Our approach explicitly models visibilities by geometric priors and augments them with neural features. We propose a novel nonuniform log sampling strategy to improve both rendering speed and reconstruction quality. Moreover, we present a learnable kernel spatially augmented with features for feature aggregations, mitigating distortions at places with drastically varying geometries. Besides, our representation can be easily manipulated. Experiments show that our model can deliver better geometries, view consistencies, and rendering quality than all counterparts and benchmarks on three datasets in both generalization and finetuning settings, preliminarily proving the potential of the new paradigm for generalizable NeRF."
                    },
                    {
                        "title": "Prompting Multi-Modal Tokens to Enhance End-to-End Autonomous Driving Imitation Learning with LLMs",
                        "abstract": "The utilization of Large Language Models (LLMs) within the realm of reinforcement learning, particularly as planners, has garnered a significant degree of attention in recent scholarly literature. However, a substantial proportion of existing research predominantly focuses on planning models for robotics that transmute the outputs derived from perception models into linguistic forms, thus adopting a `pure-language' strategy. In this research, we propose a hybrid End-to-End learning framework for autonomous driving by combining basic driving imitation learning with LLMs based on multi-modality prompt tokens. Instead of simply converting perception results from the separated train model into pure language input, our novelty lies in two aspects. 1) The end-to-end integration of visual and LiDAR sensory input into learnable multi-modality tokens, thereby intrinsically alleviating description bias by separated pre-trained perception models. 2) Instead of directly letting LLMs drive, this paper explores a hybrid setting of letting LLMs help the driving model correct mistakes and complicated scenarios. The results of our experiments suggest that the proposed methodology can attain driving scores of 49.21%, coupled with an impressive route completion rate of 91.34% in the offline evaluation conducted via CARLA. These performance metrics are comparable to the most advanced driving models."
                    },
                    {
                        "title": "Reuse and Blend: Energy-Efficient Optical Neural Network Enabled by Weight Sharing",
                        "abstract": "Optical neural networks (ONN) based on micro-ring resonators (MRR) have emerged as a promising alternative to significantly accelerating the massive matrix-vector multiplication (MVM) operations in artificial intelligence (AI) applications. However, the limited scale of MRR arrays presents a challenge for AI acceleration. The disparity between the small MRR arrays and the large weight matrices in AI necessitates extensive MRR writings, including reprogramming and calibration, resulting in considerable latency and energy overheads. To address this problem, we propose a novel design methodology to lessen the need for frequent weight reloading. Specifically, we propose a reuse and blend (R&B) architecture to support efficient layer-wise and block-wise weight sharing, which allows weights to be reused several times between layers/blocks. Experimental results demonstrate the R&B system can maintain comparable accuracy with 69% energy savings and 57% latency improvement. These results highlight the promise of the R&B to enable the efficient deployment of advanced deep learning models on photonic accelerators."
                    },
                    {
                        "title": "3D-U-SAM Network For Few-shot Tooth Segmentation in CBCT Images",
                        "abstract": "Accurate representation of tooth position is extremely important in treatment. 3D dental image segmentation is a widely used method, however labelled 3D dental datasets are a scarce resource, leading to the problem of small samples that this task faces in many cases. To this end, we address this problem with a pretrained SAM and propose a novel 3D-U-SAM network for 3D dental image segmentation. Specifically, in order to solve the problem of using 2D pre-trained weights on 3D datasets, we adopted a convolution approximation method; in order to retain more details, we designed skip connections to fuse features at all levels with reference to U-Net. The effectiveness of the proposed method is demonstrated in ablation experiments, comparison experiments, and sample size experiments."
                    },
                    {
                        "title": "Physical Priors Augmented Event-Based 3D Reconstruction",
                        "abstract": "3D neural implicit representations play a significant component in many robotic applications. However, reconstructing neural radiance fields (NeRF) from realistic event data remains a challenge due to the sparsities and the lack of information when only event streams are available. In this paper, we utilize motion, geometry, and density priors behind event data to impose strong physical constraints to augment NeRF training. The proposed novel pipeline can directly benefit from those priors to reconstruct 3D scenes without additional inputs. Moreover, we present a novel density-guided patch-based sampling strategy for robust and efficient learning, which not only accelerates training procedures but also conduces to expressions of local geometries. More importantly, we establish the first large dataset for event-based 3D reconstruction, which contains 101 objects with various materials and geometries, along with the groundtruth of images and depth maps for all camera viewpoints, which significantly facilitates other research in the related fields. The code and dataset will be publicly available at https://github.com/Mercerai/PAEv3d."
                    },
                    {
                        "title": "PFGS: High Fidelity Point Cloud Rendering via Feature Splatting",
                        "abstract": "Rendering high-fidelity images from sparse point clouds is still challenging. Existing learning-based approaches suffer from either hole artifacts, missing details, or expensive computations. In this paper, we propose a novel framework to render high-quality images from sparse points. This method first attempts to bridge the 3D Gaussian Splatting and point cloud rendering, which includes several cascaded modules. We first use a regressor to estimate Gaussian properties in a point-wise manner, the estimated properties are used to rasterize neural feature descriptors into 2D planes which are extracted from a multiscale extractor. The projected feature volume is gradually decoded toward the final prediction via a multiscale and progressive decoder. The whole pipeline experiences a two-stage training and is driven by our well-designed progressive and multiscale reconstruction loss. Experiments on different benchmarks show the superiority of our method in terms of rendering qualities and the necessities of our main components."
                    },
                    {
                        "title": "Gaining the Sparse Rewards by Exploring Lottery Tickets in Spiking Neural Network",
                        "abstract": "Deploying energy-efficient deep learning algorithms on computational-limited devices, such as robots, is still a pressing issue for real-world applications. Spiking Neural Networks (SNNs), a novel brain-inspired algorithm, offer a promising solution due to their low-latency and low-energy properties over traditional Artificial Neural Networks (ANNs). Despite their advantages, the dense structure of deep SNNs can still result in extra energy consumption. The Lottery Ticket Hypothesis (LTH) posits that within dense neural networks, there exist winning Lottery Tickets (LTs), namely sub-networks, that can be obtained without compromising performance. Inspired by this, this paper delves into the spiking-based LTs (SLTs), examining their unique properties and potential for extreme efficiency. Then, two significant sparse \\textbf{\\textit{Rewards}} are gained through comprehensive explorations and meticulous experiments on SLTs across various dense structures. Moreover, a sparse algorithm tailored for spiking transformer structure, which incorporates convolution operations into the Patch Embedding Projection (ConvPEP) module, has been proposed to achieve Multi-level Sparsity (MultiSp). MultiSp refers to (1) Patch number sparsity; (2) ConvPEP weights sparsity and binarization; and (3) ConvPEP activation layer binarization. Extensive experiments demonstrate that our method achieves extreme sparsity with only a slight performance decrease, paving the way for deploying energy-efficient neural networks in robotics and beyond."
                    },
                    {
                        "title": "LIR: A Lightweight Baseline for Image Restoration",
                        "abstract": "Recently, there have been significant advancements in Image Restoration based on CNN and transformer. However, the inherent characteristics of the Image Restoration task are often overlooked in many works. They, instead, tend to focus on the basic block design and stack numerous such blocks to the model, leading to parameters redundant and computations unnecessary. Thus, the efficiency of the image restoration is hindered. In this paper, we propose a Lightweight Baseline network for Image Restoration called LIR to efficiently restore the image and remove degradations. First of all, through an ingenious structural design, LIR removes the degradations existing in the local and global residual connections that are ignored by modern networks. Then, a Lightweight Adaptive Attention (LAA) Block is introduced which is mainly composed of proposed Adaptive Filters and Attention Blocks. The proposed Adaptive Filter is used to adaptively extract high-frequency information and enhance object contours in various IR tasks, and Attention Block involves a novel Patch Attention module to approximate the self-attention part of the transformer. On the deraining task, our LIR achieves the state-of-the-art Structure Similarity Index Measure (SSIM) and comparable performance to state-of-the-art models on Peak Signal-to-Noise Ratio (PSNR). For denoising, dehazing, and deblurring tasks, LIR also achieves a comparable performance to state-of-the-art models with a parameter size of about 30\\%. In addition, it is worth noting that our LIR produces better visual results that are more in line with the human aesthetic."
                    },
                    {
                        "title": "Masked Spiking Transformer",
                        "abstract": "The combination of Spiking Neural Networks (SNNs) and Transformers has attracted significant attention due to their potential for high energy efficiency and high-performance nature. However, existing works on this topic typically rely on direct training, which can lead to suboptimal performance. To address this issue, we propose to leverage the benefits of the ANN-to-SNN conversion method to combine SNNs and Transformers, resulting in significantly improved performance over existing state-of-the-art SNN models. Furthermore, inspired by the quantal synaptic failures observed in the nervous system, which reduces the number of spikes transmitted across synapses, we introduce a novel Masked Spiking Transformer (MST) framework that incorporates a Random Spike Masking (RSM) method to prune redundant spikes and reduce energy consumption without sacrificing performance. Our experimental results demonstrate that the proposed MST model achieves a significant reduction of 26.8% in power consumption when the masking ratio is 75% while maintaining the same level of performance as the unmasked model."
                    },
                    {
                        "title": "Chasing Day and Night: Towards Robust and Efficient All-Day Object Detection Guided by an Event Camera",
                        "abstract": "The ability to detect objects in all lighting (i.e., normal-, over-, and under-exposed) conditions is crucial for real-world applications, such as self-driving.Traditional RGB-based detectors often fail under such varying lighting conditions.Therefore, recent works utilize novel event cameras to supplement or guide the RGB modality; however, these methods typically adopt asymmetric network structures that rely predominantly on the RGB modality, resulting in limited robustness for all-day detection. In this paper, we propose EOLO, a novel object detection framework that achieves robust and efficient all-day detection by fusing both RGB and event modalities. Our EOLO framework is built based on a lightweight spiking neural network (SNN) to efficiently leverage the asynchronous property of events. Buttressed by it, we first introduce an Event Temporal Attention (ETA) module to learn the high temporal information from events while preserving crucial edge information. Secondly, as different modalities exhibit varying levels of importance under diverse lighting conditions, we propose a novel Symmetric RGB-Event Fusion (SREF) module to effectively fuse RGB-Event features without relying on a specific modality, thus ensuring a balanced and adaptive fusion for all-day detection. In addition, to compensate for the lack of paired RGB-Event datasets for all-day training and evaluation, we propose an event synthesis approach based on the randomized optical flow that allows for directly generating the event frame from a single exposure image. We further build two new datasets, E-MSCOCO and E-VOC based on the popular benchmarks MSCOCO and PASCAL VOC. Extensive experiments demonstrate that our EOLO outperforms the state-of-the-art detectors,e.g.,RENet,by a substantial margin (+3.74% mAP50) in all lighting conditions.Our code and datasets will be available at https://vlislab22.github.io/EOLO/"
                    },
                    {
                        "title": "Fully Spiking Neural Network for Legged Robots",
                        "abstract": "Recent advancements in legged robots using deep reinforcement learning have led to significant progress. Quadruped robots can perform complex tasks in challenging environments, while bipedal and humanoid robots have also achieved breakthroughs. Current reinforcement learning methods leverage diverse robot bodies and historical information to perform actions, but previous research has not emphasized the speed and energy consumption of network inference and the biological significance of neural networks. Most networks are traditional artificial neural networks that utilize multilayer perceptrons (MLP). This paper presents a novel Spiking Neural Network (SNN) for legged robots, showing exceptional performance in various simulated terrains. SNNs provide natural advantages in inference speed and energy consumption, and their pulse-form processing enhances biological interpretability. This study presents a highly efficient SNN for legged robots that can be seamless integrated into other learning models."
                    },
                    {
                        "title": "TriHelper: Zero-Shot Object Navigation with Dynamic Assistance",
                        "abstract": "Navigating toward specific objects in unknown environments without additional training, known as Zero-Shot object navigation, poses a significant challenge in the field of robotics, which demands high levels of auxiliary information and strategic planning. Traditional works have focused on holistic solutions, overlooking the specific challenges agents encounter during navigation such as collision, low exploration efficiency, and misidentification of targets. To address these challenges, our work proposes TriHelper, a novel framework designed to assist agents dynamically through three primary navigation challenges: collision, exploration, and detection. Specifically, our framework consists of three innovative components: (i) Collision Helper, (ii) Exploration Helper, and (iii) Detection Helper. These components work collaboratively to solve these challenges throughout the navigation process. Experiments on the Habitat-Matterport 3D (HM3D) and Gibson datasets demonstrate that TriHelper significantly outperforms all existing baseline methods in Zero-Shot object navigation, showcasing superior success rates and exploration efficiency. Our ablation studies further underscore the effectiveness of each helper in addressing their respective challenges, notably enhancing the agent's navigation capabilities. By proposing TriHelper, we offer a fresh perspective on advancing the object navigation task, paving the way for future research in the domain of Embodied AI and visual-based navigation."
                    },
                    {
                        "title": "Spiking Denoising Diffusion Probabilistic Models",
                        "abstract": "Spiking neural networks (SNNs) have ultra-low energy consumption and high biological plausibility due to their binary and bio-driven nature compared with artificial neural networks (ANNs). While previous research has primarily focused on enhancing the performance of SNNs in classification tasks, the generative potential of SNNs remains relatively unexplored. In our paper, we put forward Spiking Denoising Diffusion Probabilistic Models (SDDPM), a new class of SNN-based generative models that achieve high sample quality. To fully exploit the energy efficiency of SNNs, we propose a purely Spiking U-Net architecture, which achieves comparable performance to its ANN counterpart using only 4 time steps, resulting in significantly reduced energy consumption. Extensive experimental results reveal that our approach achieves state-of-the-art on the generative tasks and substantially outperforms other SNN-based generative models, achieving up to 12x and 6x improvement on the CIFAR-10 and the CelebA datasets, respectively. Moreover, we propose a threshold-guided strategy that can further improve the performances by 2.69% in a training-free manner. The SDDPM symbolizes a significant advancement in the field of SNN generation, injecting new perspectives and potential avenues of exploration. Our code is available at https://github.com/AndyCao1125/SDDPM."
                    },
                    {
                        "title": "TTPOINT: A Tensorized Point Cloud Network for Lightweight Action Recognition with Event Cameras",
                        "abstract": "Event cameras have gained popularity in computer vision due to their data sparsity, high dynamic range, and low latency. As a bio-inspired sensor, event cameras generate sparse and asynchronous data, which is inherently incompatible with the traditional frame-based method. Alternatively, the point-based method can avoid additional modality transformation and naturally adapt to the sparsity of events. Still, it typically cannot reach a comparable accuracy as the frame-based method. We propose a lightweight and generalized point cloud network called TTPOINT which achieves competitive results even compared to the state-of-the-art (SOTA) frame-based method in action recognition tasks while only using 1.5 % of the computational resources. The model is adept at abstracting local and global geometry by hierarchy structure. By leveraging tensor-train compressed feature extractors, TTPOINT can be designed with minimal parameters and computational complexity. Additionally, we developed a straightforward downsampling algorithm to maintain the spatio-temporal feature. In the experiment, TTPOINT emerged as the SOTA method on three datasets while also attaining SOTA among point cloud methods on all five datasets. Moreover, by using the tensor-train decomposition method, the accuracy of the proposed TTPOINT is almost unaffected while compressing the parameter size by 55 % in all five datasets."
                    },
                    {
                        "title": "Prompt, Plan, Perform: LLM-based Humanoid Control via Quantized Imitation Learning",
                        "abstract": "In recent years, reinforcement learning and imitation learning have shown great potential for controlling humanoid robots' motion. However, these methods typically create simulation environments and rewards for specific tasks, resulting in the requirements of multiple policies and limited capabilities for tackling complex and unknown tasks. To overcome these issues, we present a novel approach that combines adversarial imitation learning with large language models (LLMs). This innovative method enables the agent to learn reusable skills with a single policy and solve zero-shot tasks under the guidance of LLMs. In particular, we utilize the LLM as a strategic planner for applying previously learned skills to novel tasks through the comprehension of task-specific prompts. This empowers the robot to perform the specified actions in a sequence. To improve our model, we incorporate codebook-based vector quantization, allowing the agent to generate suitable actions in response to unseen textual commands from LLMs. Furthermore, we design general reward functions that consider the distinct motion features of humanoid robots, ensuring the agent imitates the motion data while maintaining goal orientation without additional guiding direction approaches or policies. To the best of our knowledge, this is the first framework that controls humanoid robots using a single learning policy network and LLM as a planner. Extensive experiments demonstrate that our method exhibits efficient and adaptive ability in complicated motion tasks."
                    }
                ]
            }
        }
    },
    "2403.07721": {
        "paper_data": {
            "title": "Visual Decoding and Reconstruction via EEG Embeddings with Guided Diffusion",
            "url": "http://arxiv.org/abs/2403.07721v7",
            "arxiv_id": "2403.07721",
            "authors": [
                "Dongyang Li",
                "Chen Wei",
                "Shiying Li",
                "Jiachen Zou",
                "Haoyang Qin",
                "Quanying Liu"
            ],
            "abstract": "How to decode human vision through neural signals has attracted a long-standing interest in neuroscience and machine learning. Modern contrastive learning and generative models improved the performance of visual decoding and reconstruction based on functional Magnetic Resonance Imaging (fMRI). However, the high cost and low temporal resolution of fMRI limit their applications in brain-computer interfaces (BCIs), prompting a high need for visual decoding based on electroencephalography (EEG). In this study, we present an end-to-end EEG-based visual reconstruction zero-shot framework, consisting of a tailored brain encoder, called the Adaptive Thinking Mapper (ATM), which projects neural signals from different sources into the shared subspace as the clip embedding, and a two-stage multi-pipe EEG-to-image generation strategy. In stage one, EEG is embedded to align the high-level clip embedding, and then the prior diffusion model refines EEG embedding into image priors. A blurry image also decoded from EEG for maintaining the low-level feature. In stage two, we input both the high-level clip embedding, the blurry image and caption from EEG latent to a pre-trained diffusion model. Furthermore, we analyzed the impacts of different time windows and brain regions on decoding and reconstruction. The versatility of our framework is demonstrated in the magnetoencephalogram (MEG) data modality. The experimental results indicate that our EEG-based visual zero-shot framework achieves SOTA performance in classification, retrieval and reconstruction, highlighting the portability, low cost, and high temporal resolution of EEG, enabling a wide range of BCI applications. Our code is available at https://github.com/ncclab-sustech/EEG_Image_decode.",
            "introduction": "   1 Introduction  A key technical challenge in BCIs is to decode/reconstruct the visual world seen by humans through non-invasive brain recordings, such as fMRI, MEG or EEG. These highly dynamic brain activities reflect human perception of the visual world, which is influenced by properties of the external visual stimulus, our internal states, emotions and even personal experiences. Thus, visual decoding and reconstruction based on neural signals can uncover how the human brain processes and interprets natural visual stimulus, as well as promote non-invasive BCI applications.   Contrastive learning and generative models have greatly advanced fMRI-based visual decoding in both decoding tasks (e.g., image classification and retrieval) and generative tasks (e.g., image reconstruction). By combining pre-trained visual models, existing fMRI decoding models can learn highly-refined feature embeddings in limited data Liu et\u00a0al., 2023b ; Scotti et\u00a0al.,  (2023). Using these embedded fMRI features, generative models such as diffusion models can reconstruct the image one is seeing Scotti et\u00a0al.,  (2023); Fang et\u00a0al.,  (2023). However, despite many advances in fMRI-based visual decoding, fMRI equipment is unportable, expensive, and difficult to operate, largely limiting its application in BCIs. Alternatively, EEG is portable, cheap, and universal, facilitating a wide range of BCI applications. EEG has higher temporal resolution and can effectively capture rapid changes in brain activity when processing complex, dynamic visual stimulus.   EEG has long been considered incomparable to fMRI in natural image decoding/reconstruction tasks, as EEG suffers from low signal-to-noise ratio, low spatial resolution, and large inter-subject variability. Recent advances in multimodal alignment have made MEG/EEG visual decoding possible, although the performance is still inferior to fMRI\u00a0Cichy and Pantazis,  (2017); Song et\u00a0al.,  (2023); Grootswagers et\u00a0al.,  (2022). Yohann Benchetrit et al. used the CLIP model to extract the latent representation of the image and trained the MEG encoder to align it with the image representation extracted by CLIP. It achieved excellent retrieval and reconstruction performance on MEG and fMRI datasets, demonstrating the potential for real-time visual decoding and reconstruction using EEG/MEG signals. Recently, Song et al.\u00a0Song et\u00a0al.,  (2023) used an EEG encoder based on ShallowNet\u00a0Schirrmeister et\u00a0al.,  (2017) and performed representation alignment through contrastive learning, achieving excellent decoding performance on the THING-EEG dataset\u00a0Gifford et\u00a0al.,  (2022). These two studies provide preliminary evidence of the potential of EEG/MEG-based visual decoding. However, there is a significant gap in their performance compared to the fMRI-level performance. This gap is largely caused by the fact that the framework of EEG visual decoding and reconstruction have not yet been thoroughly explored.   To fill this gap, we have developed a visual decoding framework based on EEG/MEG, including a novel EEG encoder and a two-stage image generation strategy. Our work has three main contributions:     1.  We present brain decoding framework, which is the first work allows zero-shot image classification, retrieval, and reconstruction via EEG data. Experimental results demonstrate that our framework is applicable to various common EEG encoder architectures.    2.  By extensively studying the existing EEG encoder modules, we construct a tailored EEG encoder ATM, which achieves state-of-the-art performance in three downstream visual decoding tasks.    3.  We report a two-stage EEG-to-image generation strategy, which separately extracts high-level and low-level visual features from",
            "references": [
                {
                    "title": "CoCoG: Controllable Visual Stimuli Generation based on Human Concept Representations",
                    "abstract": "A central question for cognitive science is to understand how humans process visual scenes, i.e, to uncover human low-dimensional concept representation space from high-dimensional visual stimuli. Generating visual stimuli with controlling concepts is the key. However, there are currently no generative models in AI to solve this problem. Here, we present the Concept based Controllable Generation (CoCoG) framework. CoCoG consists of two components, a simple yet efficient AI agent for extracting interpretable concept and predicting human decision-making in visual similarity judgment tasks, and a conditional generation model for generating visual stimuli given the concepts. We quantify the performance of CoCoG from two aspects, the human behavior prediction accuracy and the controllable generation ability. The experiments with CoCoG indicate that 1) the reliable concept embeddings in CoCoG allows to predict human behavior with 64.07% accuracy in the THINGS-similarity dataset; 2) CoCoG can generate diverse stimuli through the control of concepts; 3) CoCoG can manipulate human similarity judgment behavior by intervening key concepts.\n\nCoCoG offers visual objects with controlling concepts to advance our understanding of causality in human cognition. The code of CoCoG framework is available at https://github.com/ncclab-sustech/CoCoG."
                },
                {
                    "title": "NeuralDiffuser: Controllable fMRI Reconstruction with Primary Visual Feature Guided Diffusion",
                    "abstract": "Reconstructing visual stimuli from functional Magnetic Resonance Imaging (fMRI) based on Latent Diffusion Models (LDM) provides a fine-grained retrieval of the brain. A challenge persists in reconstructing a cohesive alignment of details (such as structure, background, texture, color, etc.). Moreover, LDMs would generate different image results even under the same conditions. For these, we first uncover the neuroscientific perspective of LDM-based methods that is top-down creation based on pre-trained knowledge from massive images but lack of detail-driven bottom-up perception resulting in unfaithful details. We propose NeuralDiffuser which introduces primary visual feature guidance to provide detail cues in the form of gradients, extending the bottom-up process for LDM-based methods to achieve faithful semantics and details. We also developed a novel guidance strategy to ensure the consistency of repeated reconstructions rather than a variety of results. We obtain the state-of-the-art performance of NeuralDiffuser on the Natural Senses Dataset (NSD), which offers more faithful details and consistent results."
                },
                {
                    "title": "Advancing EEG/MEG Source Imaging with Geometric-Informed Basis Functions",
                    "abstract": "Electroencephalography (EEG) and Magnetoencephalography (MEG) are pivotal in understanding brain activity but are limited by their poor spatial resolution. EEG/MEG source imaging (ESI) infers the high-resolution electric field distribution in the brain based on the low-resolution scalp EEG/MEG observations. However, the ESI problem is ill-posed, and how to bring neuroscience priors into ESI method is the key. Here, we present a novel method which utilizes the Brain Geometric-informed Basis Functions (GBFs) as priors to enhance EEG/MEG source imaging. Through comprehensive experiments on both synthetic data and real task EEG data, we demonstrate the superiority of GBFs over traditional spatial basis functions (e.g., Harmonic and MSP), as well as existing ESI methods (e.g., dSPM, MNE, sLORETA, eLORETA). GBFs provide robust ESI results under different noise levels, and result in biologically interpretable EEG sources. We believe the high-resolution EEG source imaging from GBFs will greatly advance neuroscience research."
                },
                {
                    "title": "Adversarial Diffusion Distillation",
                    "abstract": "We introduce Adversarial Diffusion Distillation (ADD), a novel training approach that efficiently samples large-scale foundational image diffusion models in just 1-4 steps while maintaining high image quality. We use score distillation to leverage large-scale off-the-shelf image diffusion models as a teacher signal in combination with an adversarial loss to ensure high image fidelity even in the low-step regime of one or two sampling steps. Our analyses show that our model clearly outperforms existing few-step methods (GANs, Latent Consistency Models) in a single step and reaches the performance of state-of-the-art diffusion models (SDXL) in only four steps. ADD is the first method to unlock single-step, real-time image synthesis with foundation models. Code and weights available under https://github.com/Stability-AI/generative-models and https://huggingface.co/stabilityai/ ."
                },
                {
                    "title": "Brain decoding: toward real-time reconstruction of visual perception",
                    "abstract": "In the past five years, the use of generative and foundational AI systems has greatly improved the decoding of brain activity. Visual perception, in particular, can now be decoded from functional Magnetic Resonance Imaging (fMRI) with remarkable fidelity. This neuroimaging technique, however, suffers from a limited temporal resolution ($\\approx$0.5 Hz) and thus fundamentally constrains its real-time usage. Here, we propose an alternative approach based on magnetoencephalography (MEG), a neuroimaging device capable of measuring brain activity with high temporal resolution ($\\approx$5,000 Hz). For this, we develop an MEG decoding model trained with both contrastive and regression objectives and consisting of three modules: i) pretrained embeddings obtained from the image, ii) an MEG module trained end-to-end and iii) a pretrained image generator. Our results are threefold: Firstly, our MEG decoder shows a 7X improvement of image-retrieval over classic linear decoders. Second, late brain responses to images are best decoded with DINOv2, a recent foundational image model. Third, image retrievals and generations both suggest that high-level visual features can be decoded from MEG signals, although the same approach applied to 7T fMRI also recovers better low-level features. Overall, these results, while preliminary, provide an important step towards the decoding -- in real-time -- of the visual processes continuously unfolding within the human brain."
                },
                {
                    "title": "iTransformer: Inverted Transformers Are Effective for Time Series Forecasting",
                    "abstract": "The recent boom of linear forecasting models questions the ongoing passion for architectural modifications of Transformer-based forecasters. These forecasters leverage Transformers to model the global dependencies over temporal tokens of time series, with each token formed by multiple variates of the same timestamp. However, Transformers are challenged in forecasting series with larger lookback windows due to performance degradation and computation explosion. Besides, the embedding for each temporal token fuses multiple variates that represent potential delayed events and distinct physical measurements, which may fail in learning variate-centric representations and result in meaningless attention maps. In this work, we reflect on the competent duties of Transformer components and repurpose the Transformer architecture without any modification to the basic components. We propose iTransformer that simply applies the attention and feed-forward network on the inverted dimensions. Specifically, the time points of individual series are embedded into variate tokens which are utilized by the attention mechanism to capture multivariate correlations; meanwhile, the feed-forward network is applied for each variate token to learn nonlinear representations. The iTransformer model achieves state-of-the-art on challenging real-world datasets, which further empowers the Transformer family with promoted performance, generalization ability across different variates, and better utilization of arbitrary lookback windows, making it a nice alternative as the fundamental backbone of time series forecasting. Code is available at this repository: https://github.com/thuml/iTransformer."
                },
                {
                    "title": "Cross-dataset transfer learning for motor imagery signal classification via multi-task learning and pre-training",
                    "abstract": "Objective. Deep learning (DL) models have been proven to be effective in decoding motor imagery (MI) signals in Electroencephalogram (EEG) data. However, DL models\u2019 success relies heavily on large amounts of training data, whereas EEG data collection is laborious and time-consuming. Recently, cross-dataset transfer learning has emerged as a promising approach to meet the data requirements of DL models. Nevertheless, transferring knowledge across datasets involving different MI tasks remains a significant challenge in cross-dataset transfer learning, limiting the full utilization of valuable data resources. Approach: This study proposes a pre-training-based cross-dataset transfer learning method inspired by Hard Parameter Sharing in multi-task learning. Different datasets with distinct MI paradigms are considered as different tasks, classified with shared feature extraction layers and individual task-specific layers to allow cross-dataset classification with one unified model. Then, Pre-training and fine-tuning are employed to transfer knowledge across datasets. We also designed four fine-tuning schemes and conducted extensive experiments on them. Main results: The results showed that compared to models without pre-training, models with pre-training achieved a maximum increase in accuracy of 7.76%. Moreover, when limited training data were available, the pre-training method significantly improved DL model\u2019s accuracy by 27.34% at most. The experiments also revealed that pre-trained models exhibit faster convergence and remarkable robustness. The training time per subject could be reduced by up to 102.83 s, and the variance of classification accuracy decreased by 75.22% at best. Significance: This study represents the first comprehensive investigation of the cross-dataset transfer learning method between two datasets with different MI tasks. The proposed pre-training method requires only minimal fine-tuning data when applying DL models to new MI paradigms, making MI-Brain\u2013computer interface more practical and user-friendly."
                },
                {
                    "title": "Decoding Natural Images from EEG for Object Recognition",
                    "abstract": "Electroencephalography (EEG) signals, known for convenient non-invasive acquisition but low signal-to-noise ratio, have recently gained substantial attention due to the potential to decode natural images. This paper presents a self-supervised framework to demonstrate the feasibility of learning image representations from EEG signals, particularly for object recognition. The framework utilizes image and EEG encoders to extract features from paired image stimuli and EEG responses. Contrastive learning aligns these two modalities by constraining their similarity. With the framework, we attain significantly above-chance results on a comprehensive EEG-image dataset, achieving a top-1 accuracy of 15.6% and a top-5 accuracy of 42.8% in challenging 200-way zero-shot tasks. Moreover, we perform extensive experiments to explore the biological plausibility by resolving the temporal, spatial, spectral, and semantic aspects of EEG signals. Besides, we introduce attention modules to capture spatial correlations, providing implicit evidence of the brain activity perceived from EEG data. These findings yield valuable insights for neural decoding and brain-computer interfaces in real-world scenarios. The code will be released on https://github.com/eeyhsong/NICE-EEG."
                },
                {
                    "title": "IP-Adapter: Text Compatible Image Prompt Adapter for Text-to-Image Diffusion Models",
                    "abstract": "Recent years have witnessed the strong power of large text-to-image diffusion models for the impressive generative capability to create high-fidelity images. However, it is very tricky to generate desired images using only text prompt as it often involves complex prompt engineering. An alternative to text prompt is image prompt, as the saying goes:\"an image is worth a thousand words\". Although existing methods of direct fine-tuning from pretrained models are effective, they require large computing resources and are not compatible with other base models, text prompt, and structural controls. In this paper, we present IP-Adapter, an effective and lightweight adapter to achieve image prompt capability for the pretrained text-to-image diffusion models. The key design of our IP-Adapter is decoupled cross-attention mechanism that separates cross-attention layers for text features and image features. Despite the simplicity of our method, an IP-Adapter with only 22M parameters can achieve comparable or even better performance to a fully fine-tuned image prompt model. As we freeze the pretrained diffusion model, the proposed IP-Adapter can be generalized not only to other custom models fine-tuned from the same base model, but also to controllable generation using existing controllable tools. With the benefit of the decoupled cross-attention strategy, the image prompt can also work well with the text prompt to achieve multimodal image generation. The project page is available at \\url{https://ip-adapter.github.io}."
                },
                {
                    "title": "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis",
                    "abstract": "We present SDXL, a latent diffusion model for text-to-image synthesis. Compared to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet backbone: The increase of model parameters is mainly due to more attention blocks and a larger cross-attention context as SDXL uses a second text encoder. We design multiple novel conditioning schemes and train SDXL on multiple aspect ratios. We also introduce a refinement model which is used to improve the visual fidelity of samples generated by SDXL using a post-hoc image-to-image technique. We demonstrate that SDXL shows drastically improved performance compared the previous versions of Stable Diffusion and achieves results competitive with those of black-box state-of-the-art image generators. In the spirit of promoting open research and fostering transparency in large model training and evaluation, we provide access to code and model weights at https://github.com/Stability-AI/generative-models"
                },
                {
                    "title": "Reconstructing the Mind's Eye: fMRI-to-Image with Contrastive Learning and Diffusion Priors",
                    "abstract": "We present MindEye, a novel fMRI-to-image approach to retrieve and reconstruct viewed images from brain activity. Our model comprises two parallel submodules that are specialized for retrieval (using contrastive learning) and reconstruction (using a diffusion prior). MindEye can map fMRI brain activity to any high dimensional multimodal latent space, like CLIP image space, enabling image reconstruction using generative models that accept embeddings from this latent space. We comprehensively compare our approach with other existing methods, using both qualitative side-by-side comparisons and quantitative evaluations, and show that MindEye achieves state-of-the-art performance in both reconstruction and retrieval tasks. In particular, MindEye can retrieve the exact original image even among highly similar candidates indicating that its brain embeddings retain fine-grained image-specific information. This allows us to accurately retrieve images even from large-scale databases like LAION-5B. We demonstrate through ablations that MindEye's performance improvements over previous methods result from specialized submodules for retrieval and reconstruction, improved training techniques, and training models with orders of magnitude more parameters. Furthermore, we show that MindEye can better preserve low-level image features in the reconstructions by using img2img, with outputs from a separate autoencoder. All code is available on GitHub."
                },
                {
                    "title": "High-resolution image reconstruction with latent diffusion models from human brain activity",
                    "abstract": "Reconstructing visual experiences from human brain activity offers a unique way to understand how the brain represents the world, and to interpret the connection between computer vision models and our visual system. While deep generative models have recently been employed for this task, reconstructing realistic images with high semantic fidelity is still a challenging problem. Here, we propose a new method based on a diffusion model (DM) to reconstruct images from human brain activity obtained via functional magnetic resonance imaging (fMRI). More specifically, we rely on a latent diffusion model (LDM) termed Stable Diffusion. This model reduces the computational cost of DMs, while preserving their high generative performance. We also characterize the inner mechanisms of the LDM by studying how its different components (such as the latent vector of image Z, conditioning inputs C, and different elements of the denoising U-Net) relate to distinct brain functions. We show that our proposed method can reconstruct high-resolution images with high fidelity in straight-forward fashion, without the need for any additional training and fine-tuning of complex deep-learning models. We also provide a quantitative interpretation of different LDM components from a neuroscientific perspective. Overall, our study proposes a promising method for reconstructing images from human brain activity, and provides a new framework for understanding DMs. Please check out our webpage at https://sites.google.com/view/stablediffusion-with-brain/."
                },
                {
                    "title": "THINGS-data, a multimodal collection of large-scale datasets for investigating object representations in human brain and behavior",
                    "abstract": "Understanding object representations requires a broad, comprehensive sampling of the objects in our visual world with dense measurements of brain activity and behavior. Here, we present THINGS-data, a multimodal collection of large-scale neuroimaging and behavioral datasets in humans, comprising densely sampled functional MRI and magnetoencephalographic recordings, as well as 4.70 million similarity judgments in response to thousands of photographic images for up to 1,854 object concepts. THINGS-data is unique in its breadth of richly annotated objects, allowing for testing countless hypotheses at scale while assessing the reproducibility of previous findings. Beyond the unique insights promised by each individual dataset, the multimodality of THINGS-data allows combining datasets for a much broader view into object processing than previously possible. Our analyses demonstrate the high quality of the datasets and provide five examples of hypothesis-driven and data-driven applications. THINGS-data constitutes the core public release of the THINGS initiative (https://things-initiative.org) for bridging the gap between disciplines and the advancement of cognitive neuroscience."
                },
                {
                    "title": "BrainCLIP: Bridging Brain and Visual-Linguistic Representation via CLIP for Generic Natural Visual Stimulus Decoding from fMRI",
                    "abstract": "Due to the lack of paired samples and the low signal-to-noise ratio of functional MRI (fMRI) signals, reconstructing perceived natural images or decoding their semantic contents from fMRI data are challenging tasks. In this work, we propose, for the first time, a task-agnostic fMRI-based brain decoding model, BrainCLIP, which leverages CLIP's cross-modal generalization ability to bridge the modality gap between brain activity, image, and text. Our experiments demonstrate that CLIP can act as a pivot for generic brain decoding tasks, including zero-shot visual categories decoding, fMRI-image/text matching, and fMRI-to-image generation. Specifically, BrainCLIP aims to train a mapping network that transforms fMRI patterns into a well-aligned CLIP embedding space by combining visual and textual supervision. Our experiments show that this combination can boost the decoding model's performance on certain tasks like fMRI-text matching and fMRI-to-image generation. On the zero-shot visual category decoding task, BrainCLIP achieves significantly better performance than BraVL, a recently proposed multi-modal method specifically designed for this task. BrainCLIP can also reconstruct visual stimuli with high semantic fidelity and establishes a new state-of-the-art for fMRI-based natural image reconstruction in terms of high-level semantic features."
                },
                {
                    "title": "Representative-Based Cold Start for Adaptive SSVEP-BCI",
                    "abstract": "Objective: The tradeoff between calibration effort and model performance still hinders the user experience for steady-state visual evoked brain-computer interfaces (SSVEP-BCI). To address this issue and improve model generalizability, this work investigated the adaptation from the cross-dataset model to avoid the training process, while maintaining high prediction ability. Methods: When a new subject enrolls, a group of user-independent (UI) models is recommended as the representative from a multi-source data pool. The representative model is then augmented with online adaptation and transfer learning techniques based on user-dependent (UD) data. The proposed method is validated on both offline (N=55) and online (N=12) experiments. Results: Compared with the UD adaptation, the recommended representative model relieved approximately 160 trials of calibration efforts for a new user. In the online experiment, the time window decreased from 2 s to 0.56\u00b10.2 s, while maintaining high prediction accuracy of 0.89-0.96. Finally, the proposed method achieved the average information transfer rate (ITR) of 243.49 bits/min, which is the highest ITR ever reported in a complete calibration-free setting. The results of the offline result were consistent with the online experiment. Conclusion: Representatives can be recommended even in a cross-subject/device/session situation. With the help of represented UI data, the proposed method can achieve sustained high performance without a training process. Significance: This work provides an adaptive approach to the transferable model for SSVEP-BCIs, enabling a more generalized, plug-and-play and high-performance BCI free of calibrations."
                },
                {
                    "title": "EEG Conformer: Convolutional Transformer for EEG Decoding and Visualization",
                    "abstract": "Due to the limited perceptual field, convolutional neural networks (CNN) only extract local temporal features and may fail to capture long-term dependencies for EEG decoding. In this paper, we propose a compact Convolutional Transformer, named EEG Conformer, to encapsulate local and global features in a unified EEG classification framework. Specifically, the convolution module learns the low-level local features throughout the one-dimensional temporal and spatial convolution layers. The self-attention module is straightforwardly connected to extract the global correlation within the local temporal features. Subsequently, the simple classifier module based on fully-connected layers is followed to predict the categories for EEG signals. To enhance interpretability, we also devise a visualization strategy to project the class activation mapping onto the brain topography. Finally, we have conducted extensive experiments to evaluate our method on three public datasets in EEG-based motor imagery and emotion recognition paradigms. The experimental results show that our method achieves state-of-the-art performance and has great potential to be a new baseline for general EEG decoding. The code has been released in https://github.com/eeyhsong/EEG-Conformer."
                },
                {
                    "title": "Decoding Visual Neural Representations by Multimodal Learning of Brain-Visual-Linguistic Features",
                    "abstract": "Decoding human visual neural representations is a challenging task with great scientific significance in revealing vision-processing mechanisms and developing brain-like intelligent machines. Most existing methods are difficult to generalize to novel categories that have no corresponding neural data for training. The two main reasons are 1) the under-exploitation of the multimodal semantic knowledge underlying the neural data and 2) the small number of paired (stimuli-responses) training data. To overcome these limitations, this paper presents a generic neural decoding method called BraVL that uses multimodal learning of brain-visual-linguistic features. We focus on modeling the relationships between brain, visual and linguistic features via multimodal deep generative models. Specifically, we leverage the mixture-of-product-of-experts formulation to infer a latent code that enables a coherent joint generation of all three modalities. To learn a more consistent joint representation and improve the data efficiency in the case of limited brain activity data, we exploit both intra- and inter-modality mutual information maximization regularization terms. In particular, our BraVL model can be trained under various semi-supervised scenarios to incorporate the visual and textual features obtained from the extra categories. Finally, we construct three trimodal matching datasets, and the extensive experiments lead to some interesting conclusions and cognitive insights: 1) decoding novel visual categories from human brain activity is practically possible with good accuracy; 2) decoding models using the combination of visual and linguistic features perform much better than those using either of them alone; 3) visual perception may be accompanied by linguistic influences to represent the semantics of visual stimuli."
                },
                {
                    "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                    "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                },
                {
                    "title": "GIT: A Generative Image-to-text Transformer for Vision and Language",
                    "abstract": "In this paper, we design and train a Generative Image-to-text Transformer, GIT, to unify vision-language tasks such as image/video captioning and question answering. While generative models provide a consistent network architecture between pre-training and fine-tuning, existing work typically contains complex structures (uni/multi-modal encoder/decoder) and depends on external modules such as object detectors/taggers and optical character recognition (OCR). In GIT, we simplify the architecture as one image encoder and one text decoder under a single language modeling task. We also scale up the pre-training data and the model size to boost the model performance. Without bells and whistles, our GIT establishes new state of the arts on 12 challenging benchmarks with a large margin. For instance, our model surpasses the human performance for the first time on TextCaps (138.2 vs. 125.5 in CIDEr). Furthermore, we present a new scheme of generation-based image classification and scene text recognition, achieving decent performance on standard benchmarks. Codes are released at \\url{https://github.com/microsoft/GenerativeImage2Text}."
                },
                {
                    "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
                    "abstract": "Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples."
                },
                {
                    "title": "SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations",
                    "abstract": "Guided image synthesis enables everyday users to create and edit photo-realistic images with minimum effort. The key challenge is balancing faithfulness to the user input (e.g., hand-drawn colored strokes) and realism of the synthesized image. Existing GAN-based methods attempt to achieve such balance using either conditional GANs or GAN inversions, which are challenging and often require additional training data or loss functions for individual applications. To address these issues, we introduce a new image synthesis and editing method, Stochastic Differential Editing (SDEdit), based on a diffusion model generative prior, which synthesizes realistic images by iteratively denoising through a stochastic differential equation (SDE). Given an input image with user guide of any type, SDEdit first adds noise to the input, then subsequently denoises the resulting image through the SDE prior to increase its realism. SDEdit does not require task-specific training or inversions and can naturally achieve the balance between realism and faithfulness. SDEdit significantly outperforms state-of-the-art GAN-based methods by up to 98.09% on realism and 91.72% on overall satisfaction scores, according to a human perception study, on multiple tasks, including stroke-based image synthesis and editing as well as image compositing."
                },
                {
                    "title": "Generalized Multimodal ELBO",
                    "abstract": "Multiple data types naturally co-occur when describing real-world phenomena and learning from them is a long-standing goal in machine learning research. However, existing self-supervised generative models approximating an ELBO are not able to fulfill all desired requirements of multimodal models: their posterior approximation functions lead to a trade-off between the semantic coherence and the ability to learn the joint data distribution. We propose a new, generalized ELBO formulation for multimodal data that overcomes these limitations. The new objective encompasses two previous methods as special cases and combines their benefits without compromises. In extensive experiments, we demonstrate the advantage of the proposed method compared to state-of-the-art models in self-supervised, generative learning tasks."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Edge Sparse Basis Network: A Deep Learning Framework for EEG Source Localization",
                    "abstract": "EEG source localization is an important technical issue in EEG analysis. Despite many numerical methods existed for EEG source localization, they all rely on strong priors and the deep sources are intractable. Here we propose a deep learning framework using spatial basis function decomposition for EEG source localization. This framework combines the edge sparsity prior and Gaussian source basis, called Edge Sparse Basis Network (ESBN). The performance of ESBN is validated by both synthetic data and real EEG data during motor tasks. The results suggest that the supervised ESBN outperforms the traditional numerical methods in synthetic data and the unsupervised fine-tuning provides more focal and accurate localizations in real data. Our proposed deep learning framework can be extended to account for other source priors, and the real-time property of ESBN can facilitate the applications of EEG in brain-computer interfaces and clinics."
                },
                {
                    "title": "Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision",
                    "abstract": "Pre-trained representations are becoming crucial for many NLP and perception tasks. While representation learning in NLP has transitioned to training on raw text without human annotations, visual and vision-language representations still rely heavily on curated training datasets that are expensive or require expert knowledge. For vision applications, representations are mostly learned using datasets with explicit class labels such as ImageNet or OpenImages. For vision-language, popular datasets like Conceptual Captions, MSCOCO, or CLIP all involve a non-trivial data collection (and cleaning) process. This costly curation process limits the size of datasets and hence hinders the scaling of trained models. In this paper, we leverage a noisy dataset of over one billion image alt-text pairs, obtained without expensive filtering or post-processing steps in the Conceptual Captions dataset. A simple dual-encoder architecture learns to align visual and language representations of the image and text pairs using a contrastive loss. We show that the scale of our corpus can make up for its noise and leads to state-of-the-art representations even with such a simple learning scheme. Our visual representation achieves strong performance when transferred to classification tasks such as ImageNet and VTAB. The aligned visual and language representations enables zero-shot image classification and also set new state-of-the-art results on Flickr30K and MSCOCO image-text retrieval benchmarks, even when compared with more sophisticated cross-attention models. The representations also enable cross-modality search with complex text and text + image queries."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "The Perils and Pitfalls of Block Design for EEG Classification Experiments",
                    "abstract": "A recent paper <xref ref-type=\"bibr\" rid=\"ref1\">[1]</xref> claims to classify brain processing evoked in subjects watching ImageNet stimuli as measured with EEG and to employ a representation derived from this processing to construct a novel object classifier. That paper, together with a series of subsequent papers <xref ref-type=\"bibr\" rid=\"ref2\">[2]</xref> , <xref ref-type=\"bibr\" rid=\"ref3\">[3]</xref> , <xref ref-type=\"bibr\" rid=\"ref4\">[4]</xref> , <xref ref-type=\"bibr\" rid=\"ref5\">[5]</xref> , <xref ref-type=\"bibr\" rid=\"ref6\">[6]</xref> , <xref ref-type=\"bibr\" rid=\"ref7\">[7]</xref> , <xref ref-type=\"bibr\" rid=\"ref8\">[8]</xref> , claims to achieve successful results on a wide variety of computer-vision tasks, including object classification, transfer learning, and generation of images depicting human perception and thought using brain-derived representations measured through EEG. Our novel experiments and analyses demonstrate that their results crucially depend on the block design that they employ, where all stimuli of a given class are presented together, and fail with a rapid-event design, where stimuli of different classes are randomly intermixed. The block design leads to classification of arbitrary brain states based on block-level temporal correlations that are known to exist in all EEG data, rather than stimulus-related activity. Because every trial in their test sets comes from the same block as many trials in the corresponding training sets, their block design thus leads to classifying arbitrary temporal artifacts of the data instead of stimulus-related activity. This invalidates all subsequent analyses performed on this data in multiple published papers and calls into question all of the reported results. We further show that a novel object classifier constructed with a random codebook performs as well as or better than a novel object classifier constructed with the representation extracted from EEG data, suggesting that the performance of their classifier constructed with a representation extracted from EEG data does not benefit from the brain-derived representation. Together, our results illustrate the far-reaching implications of the temporal autocorrelations that exist in all neuroimaging data for classification experiments. Further, our results calibrate the underlying difficulty of the tasks involved and caution against overly optimistic, but incorrect, claims to the contrary."
                },
                {
                    "title": "Unsupervised Learning of Visual Features by Contrasting Cluster Assignments",
                    "abstract": "Unsupervised image representations have significantly reduced the gap with supervised pretraining, notably with the recent achievements of contrastive learning methods. These contrastive methods typically work online and rely on a large number of explicit pairwise feature comparisons, which is computationally challenging. In this paper, we propose an online algorithm, SwAV, that takes advantage of contrastive methods without requiring to compute pairwise comparisons. Specifically, our method simultaneously clusters the data while enforcing consistency between cluster assignments produced for different augmentations (or views) of the same image, instead of comparing features directly as in contrastive learning. Simply put, we use a swapped prediction mechanism where we predict the cluster assignment of a view from the representation of another view. Our method can be trained with large and small batches and can scale to unlimited amounts of data. Compared to previous contrastive methods, our method is more memory efficient since it does not require a large memory bank or a special momentum network. In addition, we also propose a new data augmentation strategy, multi-crop, that uses a mix of views with different resolutions in place of two full-resolution views, without increasing the memory or compute requirements much. We validate our findings by achieving 75.3% top-1 accuracy on ImageNet with ResNet-50, as well as surpassing supervised pretraining on all the considered transfer tasks."
                },
                {
                    "title": "Cross-Dataset Variability Problem in EEG Decoding With Deep Learning",
                    "abstract": "Cross-subject variability problems hinder practical usages of Brain-Computer Interfaces. Recently, deep learning has been introduced into the BCI community due to its better generalization and feature representation abilities. However, most studies currently only have validated deep learning models for single datasets, and the generalization ability for other datasets still needs to be further verified. In this paper, we validated deep learning models for eight MI datasets and demonstrated that the cross-dataset variability problem weakened the generalization ability of models. To alleviate the impact of cross-dataset variability, we proposed an online pre-alignment strategy for aligning the EEG distributions of different subjects before training and inference processes. The results of this study show that deep learning models with online pre-alignment strategies could significantly improve the generalization ability across datasets without any additional calibration data."
                },
                {
                    "title": "Transfer Learning for EEG-Based Brain\u2013Computer Interfaces: A Review of Progress Made Since 2016",
                    "abstract": "A brain\u2013computer interface (BCI) enables a user to communicate with a computer directly using brain signals. The most common noninvasive BCI modality, electroencephalogram (EEG), is sensitive to noise/artifact and suffers between-subject/within-subject nonstationarity. Therefore, it is difficult to build a generic pattern recognition model in an EEG-based BCI system that is optimal for different subjects, during different sessions, for different devices and tasks. Usually, a calibration session is needed to collect some training data for a new subject, which is time consuming and user unfriendly. Transfer learning (TL), which utilizes data or knowledge from similar or relevant subjects/sessions/devices/tasks to facilitate learning for a new subject/session/device/task, is frequently used to reduce the amount of calibration effort. This article reviews journal publications on TL approaches in EEG-based BCIs in the last few years, i.e., since 2016. Six paradigms and applications\u2014motor imagery, event-related potentials, steady-state visual evoked potentials, affective BCIs, regression problems, and adversarial attacks\u2014are considered. For each paradigm/application, we group the TL approaches into cross-subject/session, cross-device, and cross-task settings and review them separately. Observations and conclusions are made at the end of the article, which may point to future research directions."
                },
                {
                    "title": "Object classification from randomized EEG trials",
                    "abstract": "New results suggest strong limits to the feasibility of object classification from human brain activity evoked by image stimuli, as measured through EEG. Considerable prior work suffers from a confound between the stimulus class and the time since the start of the experiment. A prior attempt to avoid this confound using randomized trials was unable to achieve results above chance in a statistically significant fashion when the data sets were of the same size as the original experiments. Here, we attempt object classification from EEG using an array of methods that are representative of the state-of-the-art, with a far larger (20\u00d7) dataset of randomized EEG trials, 1,000 stimulus presentations of each of forty classes, all from a single subject. To our knowledge, this is the largest such EEG data-collection effort from a single subject and is at the bounds of feasibility. We obtain classification accuracy that is marginally above chance and above chance in a statistically significant fashion, and further assess how accuracy depends on the classifier used, the amount of training data used, and the number of classes. Reaching the limits of data collection with only marginally above-chance performance suggests that the prevailing literature substantially exaggerates the feasibility of object classification from EEG."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "The Human Brain Encodes a Chronicle of Visual Events at Each Instant of Time Through the Multiplexing of Traveling Waves",
                    "abstract": "The human brain continuously processes streams of visual input. Yet, a single image typically triggers neural responses that extend beyond 1s. To understand how the brain encodes and maintains successive images, we analyzed with electroencephalography the brain activity of human subjects while they watched \u223c5000 visual stimuli presented in fast sequences. First, we confirm that each stimulus can be decoded from brain activity for \u223c1s, and we demonstrate that the brain simultaneously represents multiple images at each time instant. Second, we source localize the corresponding brain responses in the expected visual hierarchy and show that distinct brain regions represent, at each time instant, different snapshots of past stimulations. Third, we propose a simple framework to further characterize the dynamical system of these traveling waves. Our results show that a chain of neural circuits, which each consist of (1) a hidden maintenance mechanism and (2) an observable update mechanism, accounts for the dynamics of macroscopic brain representations elicited by visual sequences. Together, these results detail a simple architecture explaining how successive visual events and their respective timings can be simultaneously represented in the brain. SIGNIFICANCE STATEMENT Our retinas are continuously bombarded with a rich flux of visual input. Yet, how our brain continuously processes such visual streams is a major challenge to neuroscience. Here, we developed techniques to decode and track, from human brain activity, multiple images flashed in rapid succession. Our results show that the brain simultaneously represents multiple successive images at each time instant by multiplexing them along a neural cascade. Dynamical modeling shows that these results can be explained by a hierarchy of neural assemblies that continuously propagate multiple visual contents. Overall, this study sheds new light on the biological basis of our visual experience."
                },
                {
                    "title": "Variational Mixture-of-Experts Autoencoders for Multi-Modal Deep Generative Models",
                    "abstract": "Learning generative models that span multiple data modalities, such as vision and language, is often motivated by the desire to learn more useful, generalisable representations that faithfully capture common underlying factors between the modalities. In this work, we characterise successful learning of such models as the fulfilment of four criteria: i) implicit latent decomposition into shared and private subspaces, ii) coherent joint generation over all modalities, iii) coherent cross-generation across individual modalities, and iv) improved model learning for individual modalities through multi-modal integration. Here, we propose a mixture-of-experts multi-modal variational autoencoder (MMVAE) for learning of generative models on different sets of modalities, including a challenging image language dataset, and demonstrate its ability to satisfy all four criteria, both qualitatively and quantitatively."
                },
                {
                    "title": "Generalized Zero- and Few-Shot Learning via Aligned Variational Autoencoders",
                    "abstract": "Many approaches in generalized zero-shot learning rely on cross-modal mapping between the image feature space and the class embedding space. As labeled images are expensive, one direction is to augment the dataset by generating either images or image features. However, the former misses fine-grained details and the latter requires learning a mapping associated with class embeddings. In this work, we take feature generation one step further and propose a model where a shared latent space of image features and class embeddings is learned by modality-specific aligned variational autoencoders. This leaves us with the required discriminative information about the image and classes in the latent features, on which we train a softmax classifier. The key to our approach is that we align the distributions learned from images and from side-information to construct latent features that contain the essential multi-modal information associated with unseen classes. We evaluate our learned latent features on several benchmark datasets, i.e. CUB, SUN, AWA1 and AWA2, and establish a new state of the art on generalized zero-shot as well as on few-shot learning. Moreover, our results on ImageNet with various zero-shot splits show that our latent features generalize well in large-scale settings."
                },
                {
                    "title": "Decoding Brain Representations by Multimodal Learning of Neural Activity and Visual Features",
                    "abstract": "This work presents a novel method of exploring human brain-visual representations, with a view towards replicating these processes in machines. The core idea is to learn plausible computational and biological representations by correlating human neural activity and natural images. Thus, we first propose a model, EEG-ChannelNet, to learn a brain manifold for EEG classification. After verifying that visual information can be extracted from EEG data, we introduce a multimodal approach that uses deep image and EEG encoders, trained in a siamese configuration, for learning a joint manifold that maximizes a compatibility measure between visual features and brain representations. We then carry out image classification and saliency detection on the learned manifold. Performance analyses show that our approach satisfactorily decodes visual information from neural signals. This, in turn, can be used to effectively supervise the training of deep learning models, as demonstrated by the high performance of image classification and saliency detection on out-of-training classes. The obtained results show that the learned brain-visual features lead to improved performance and simultaneously bring deep models more in line with cognitive neuroscience work related to visual perception and attention."
                },
                {
                    "title": "End-to-End Deep Image Reconstruction From Human Brain Activity",
                    "abstract": "Deep neural networks (DNNs) have recently been applied successfully to brain decoding and image reconstruction from functional magnetic resonance imaging (fMRI) activity. However, direct training of a DNN with fMRI data is often avoided because the size of available data is thought to be insufficient to train a complex network with numerous parameters. Instead, a pre-trained DNN has served as a proxy for hierarchical visual representations, and fMRI data were used to decode individual DNN features of a stimulus image using a simple linear model, which were then passed to a reconstruction module. Here, we present our attempt to directly train a DNN model with fMRI data and the corresponding stimulus images to build an end-to-end reconstruction model. We trained a generative adversarial network with an additional loss term defined in a high-level feature space (feature loss) using up to 6,000 training data points (natural images and the fMRI responses). The trained deep generator network was tested on an independent dataset, directly producing a reconstructed image given an fMRI pattern as the input. The reconstructions obtained from the proposed method showed resemblance with both natural and artificial test stimuli. The accuracy increased as a function of the training data size, though not outperforming the decoded feature-based method with the available data size. Ablation analyses indicated that the feature loss played a critical role to achieve accurate reconstruction. Our results suggest a potential for the end-to-end framework to learn a direct mapping between brain activity and perception given even larger datasets."
                },
                {
                    "title": "Multimodal Generative Models for Scalable Weakly-Supervised Learning",
                    "abstract": "Multiple modalities often co-occur when describing natural phenomena. Learning a joint representation of these modalities should yield deeper and more useful representations. Previous work have proposed generative models to handle multi-modal input. However, these models either do not learn a joint distribution or require complex additional computations to handle missing data. Here, we introduce a multimodal variational autoencoder that uses a product-of-experts inference network and a sub-sampled training paradigm to solve the multi-modal inference problem. Notably, our model shares parameters to efficiently learn under any combination of missing modalities, thereby enabling weakly-supervised learning. We apply our method on four datasets and show that we match state-of-the-art performance using many fewer parameters. In each case our approach yields strong weakly-supervised results. We then consider a case study of learning image transformations---edge detection, colorization, facial landmark segmentation, etc.---as a set of modalities. We find appealing results across this range of tasks."
                },
                {
                    "title": "Multivariate pattern analysis of MEG and EEG: A comparison of representational structure in time and space",
                    "abstract": "Multivariate pattern analysis of magnetoencephalography (MEG) and electroencephalography (EEG) data can reveal the rapid neural dynamics underlying cognition. However, MEG and EEG have systematic differences in sampling neural activity. This poses the question to which degree such measurement differences consistently bias the results of multivariate analysis applied to MEG and EEG activation patterns. To investigate, we conducted a concurrent MEG/EEG study while participants viewed images of everyday objects. We applied multivariate classification analyses to MEG and EEG data, and compared the resulting time courses to each other, and to fMRI data for an independent evaluation in space. We found that both MEG and EEG revealed the millisecond spatio-temporal dynamics of visual processing with largely equivalent results. Beyond yielding convergent results, we found that MEG and EEG also captured partly unique aspects of visual representations. Those unique components emerged earlier in time for MEG than for EEG. Identifying the sources of those unique components with fMRI, we found the locus for both MEG and EEG in high-level visual cortex, and in addition for MEG in early visual cortex. Together, our results show that multivariate analyses of MEG and EEG data offer a convergent and complimentary view on neural processing, and motivate the wider adoption of these methods in both MEG and EEG research."
                },
                {
                    "title": "EEGNet: a compact convolutional neural network for EEG-based brain\u2013computer interfaces",
                    "abstract": "Objective. Brain\u2013computer interfaces (BCI) enable direct communication with a computer, using neural activity as the control signal. This neural signal is generally chosen from a variety of well-studied electroencephalogram (EEG) signals. For a given BCI paradigm, feature extractors and classifiers are tailored to the distinct characteristics of its expected EEG control signal, limiting its application to that specific signal. Convolutional neural networks (CNNs), which have been used in computer vision and speech recognition to perform automatic feature extraction and classification, have successfully been applied to EEG-based BCIs; however, they have mainly been applied to single BCI paradigms and thus it remains unclear how these architectures generalize to other paradigms. Here, we ask if we can design a single CNN architecture to accurately classify EEG signals from different BCI paradigms, while simultaneously being as compact as possible. Approach. In this work we introduce EEGNet, a compact convolutional neural network for EEG-based BCIs. We introduce the use of depthwise and separable convolutions to construct an EEG-specific model which encapsulates well-known EEG feature extraction concepts for BCI. We compare EEGNet, both for within-subject and cross-subject classification, to current state-of-the-art approaches across four BCI paradigms: P300 visual-evoked potentials, error-related negativity responses (ERN), movement-related cortical potentials (MRCP), and sensory motor rhythms (SMR). Main results. We show that EEGNet generalizes across paradigms better than, and achieves comparably high performance to, the reference algorithms when only limited training data is available across all tested paradigms. In addition, we demonstrate three different approaches to visualize the contents of a trained EEGNet model to enable interpretation of the learned features. Significance. Our results suggest that EEGNet is robust enough to learn a wide variety of interpretable features over a range of BCI tasks. Our models can be found at: https://github.com/vlawhern/arl-eegmodels."
                },
                {
                    "title": "Adam: A Method for Stochastic Optimization",
                    "abstract": "We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
                },
                {
                    "title": "Image quality assessment: from error visibility to structural similarity",
                    "abstract": "Objective methods for assessing perceptual image quality traditionally attempted to quantify the visibility of errors (differences) between a distorted image and a reference image using a variety of known properties of the human visual system. Under the assumption that human visual perception is highly adapted for extracting structural information from a scene, we introduce an alternative complementary framework for quality assessment based on the degradation of structural information. As a specific example of this concept, we develop a structural similarity index and demonstrate its promise through a set of intuitive examples, as well as comparison to both subjective ratings and state-of-the-art objective methods on a database of images compressed with JPEG and JPEG2000. A MATLAB implementation of the proposed algorithm is available online at http://www.cns.nyu.edu//spl sim/lcv/ssim/."
                },
                {
                    "title": "Alleviating the Semantic Gap for Generalized fMRI-to-Image Reconstruction",
                    "abstract": "Although existing fMRI-to-image reconstruction methods could predict high-quality images, they do not explicitly consider the semantic gap between training and testing data, resulting in reconstruction with unstable and uncertain semantics. This paper addresses the problem of generalized fMRI-to-image reconstruction by explicitly alleviates the semantic gap. Specifically, we leverage the pre-trained CLIP model to map the training data to a compact feature representation, which essentially extends the sparse semantics of training data to dense ones, thus alleviating the semantic gap of the instances nearby known concepts (i.e., inside the training super-classes). Inspired by the robust low-level representation in fMRI data, which could help alleviate the semantic gap for instances that far from the known concepts (i.e., outside the training super-classes), we leverage structural information as a general cue to guide image reconstruction. Further, we quantify the semantic uncertainty based on probability density estimation and achieve G eneralized fMRI-to-image reconstruction by adaptively integrating E xpanded S emantics and S tructural information ( GESS ) within a diffusion process. Experimental results demonstrate that the proposed GESS model outperforms state-of-the-art methods, and we propose a generalized scenario split strategy to evaluate the advantage of GESS in closing the semantic gap. Our codes are available at https://github.com/duolala1/GESS."
                },
                {
                    "title": "Supplementary for: Deep learning with convolutional neural networks for EEG decoding and visualization",
                    "abstract": "Translational Neurotechnology Lab, Epilepsy Center, Medical Center \u2013 University of Freiburg, Engelberger Str. 21, 79106 Freiburg, Germany BrainLinks-BrainTools Cluster of Excellence, University of Freiburg, Georges-K\u00f6hler-Allee 79, 79110 Freiburg, Germany Machine Learning Lab, Computer Science Dept., University of Freiburg, Georges-K\u00f6hler-Allee 79, 79110 Freiburg, Germany Neurobiology and Biophysics, Faculty of Biology, University of Freiburg, Hansastr. 9a, 79104 Freiburg, Germany Machine Learning for Automated Algorithm Design Lab, Computer Science Dept., University of Freiburg, Georges-K\u00f6hler-Allee 52, 79110 Freiburg im Breisgau, Germany Brain State Decoding Lab, Computer Science Dept., University of Freiburg, Albertstr. 23, 79104 Freiburg, Germany Autonomous Intelligent Systems Lab, Computer Science Dept., University of Freiburg, Georges-K\u00f6hler-Allee 79, 79110 Freiburg, Germany"
                }
            ],
            "categories": [
                "cs.HC",
                "eess.SP",
                "q-bio.NC"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively decode and reconstruct visual stimuli from non-invasive EEG recordings to achieve performance comparable to fMRI-based methods?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem has significant implications for the research community as it could democratize access to brain-computer interfaces (BCIs) by utilizing portable and cost-effective EEG technology. This advancement could lead to practical applications in various fields, including neurorehabilitation, assistive technologies for individuals with disabilities, and enhanced understanding of human cognition and perception. By bridging the performance gap between EEG and fMRI, future research can explore new avenues in cognitive neuroscience and machine learning, potentially leading to breakthroughs in how we interpret brain activity related to visual processing.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent limitations of EEG, such as its low signal-to-noise ratio, low spatial resolution, and high inter-subject variability, which complicate the accurate decoding of visual information. Naive approaches may fail due to the complexity of aligning EEG signals with visual stimuli, as well as the need for sophisticated models to capture the dynamic nature of brain activity. Additionally, the lack of a comprehensive framework for EEG visual decoding and reconstruction has hindered progress, necessitating the development of novel methodologies that can effectively leverage the strengths of EEG data.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on fMRI for visual decoding due to its superior spatial resolution and signal quality, leading to a lack of exploration into EEG-based methods. Existing solutions have not adequately addressed the unique challenges posed by EEG, such as its noise and variability. Moreover, the frameworks developed for EEG visual decoding have been limited in scope and performance, failing to achieve results comparable to fMRI. Our approach differs by introducing a tailored EEG encoder and a two-stage image generation strategy, which systematically addresses these limitations and enhances the decoding capabilities of EEG.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology includes the development of a novel EEG encoder, ATM, specifically designed to optimize performance in visual decoding tasks. We will utilize a two-stage image generation strategy that separately extracts high-level and low-level visual features from EEG data. The dataset for our experiments will include the THING-EEG dataset, and we will evaluate our framework using metrics such as classification accuracy, retrieval performance"
            }
        },
        "author_data": {
            "b1072ed9-fadc-4910-8406-f4a525e3e772": {
                "pk": "b1072ed9-fadc-4910-8406-f4a525e3e772",
                "name": "Dongyang Li",
                "collaborators": [
                    "Reza Rahemi",
                    "Haobin Li",
                    "Junyu Zhang",
                    "Dmitry Ignatyev",
                    "Antonios Tsourdos",
                    "Zhongyuan Wang",
                    "Qian Chen",
                    "Taolin Zhang",
                    "Xiaofeng He",
                    "Lingling Yang"
                ],
                "domain": [
                    "Material Science",
                    "Optimization",
                    "Machine Learning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Variation in electron work function with temperature and its effect on the Young's modulus of metals",
                        "abstract": "Properties of metals are fundamentally determined by their electron behavior, which is largely reflected by the electron work function ($\\varphi $). Recent studies have demonstrated that many properties of metallic materials are directly related to $\\varphi $, which may provide a simple but fundamental parameter for material design. Since material properties are affected by temperature, in this article a simple model is proposed to correlate the work function with temperature, expressed as $\\varphi (T)=\\varphi_{0} -\\gamma \\frac{(k_{B} T)^{2}}{\\varphi_{0}} $, where $\\gamma $ varies with the crystal structure. This $\\varphi $-T relationship helps determine and understand the dependence of metal properties on temperature on a feasible electronic base. As a sample application, the established relationship is applied to determine the dependence of Young's modulus of metals on temperature. The proposed relationship is consistent with experimental observations."
                    },
                    {
                        "title": "General Procedure to Provide High-Probability Guarantees for Stochastic Saddle Point Problems",
                        "abstract": "This paper considers smooth strongly convex and strongly concave (SC-SC) stochastic saddle point (SSP) problems. Suppose there is an arbitrary oracle that in expectation returns an $\\epsilon$-solution in the sense of certain gaps, which can be the duality gap or its weaker variants. We propose a general PB-SSP framework to guarantee an $\\epsilon$ small duality gap solution with high probability via only $\\mathcal{O}\\big(\\log \\frac{1}{p}\\cdot\\text{poly}(\\log \\kappa)\\big)$ calls of this oracle, where $p\\in(0,1)$ is the confidence level and $\\kappa$ is the condition number. When applied to the sample average approximation (SAA) oracle, in addition to equipping the solution with high probability, our approach even improves the sample complexity by a factor of $\\text{poly}(\\kappa)$, since the high-probability argument enables us to circumvent some key difficulties of the uniform stability analysis of SAA."
                    },
                    {
                        "title": "Estimation of non-symmetric and unbounded region of attraction using shifted shape function and R-composition",
                        "abstract": "A general numerical method using sum of squares programming is proposed to address the problem of estimating the region of attraction (ROA) of an asymptotically stable equilibrium point of a nonlinear polynomial system. The method is based on Lyapunov theory, and a shape function is defined to enlarge the provable subset of a local Lyapunov function. In contrast with existing methods with a shape function centered at the equilibrium point, the proposed method utilizes a shifted shape function (SSF) with its center shifted iteratively towards the boundary of the newly obtained invariant subset to improve ROA estimation. A set of shifting centers with corresponding SSFs is generated to produce proven subsets of the exact ROA and then a composition method, namely R-composition, is employed to express these independent sets in a compact form by just a single but richer-shaped level set. The proposed method denoted as RcomSSF brings a significant improvement for general ROA estimation problems, especially for non-symmetric or unbounded ROA, while keeping the computational burden at a reasonable level. Its effectiveness and advantages are demonstrated by several benchmark examples from literature."
                    },
                    {
                        "title": "CIDR: A Cooperative Integrated Dynamic Refining Method for Minimal Feature Removal Problem",
                        "abstract": "The minimal feature removal problem in the post-hoc explanation area aims to identify the minimal feature set (MFS). Prior studies using the greedy algorithm to calculate the minimal feature set lack the exploration of feature interactions under a monotonic assumption which cannot be satisfied in general scenarios. In order to address the above limitations, we propose a Cooperative Integrated Dynamic Refining method (CIDR) to efficiently discover minimal feature sets. Specifically, we design Cooperative Integrated Gradients (CIG) to detect interactions between features. By incorporating CIG and characteristics of the minimal feature set, we transform the minimal feature removal problem into a knapsack problem. Additionally, we devise an auxiliary Minimal Feature Refinement algorithm to determine the minimal feature set from numerous candidate sets. To the best of our knowledge, our work is the first to address the minimal feature removal problem in the field of natural language processing. Extensive experiments demonstrate that CIDR is capable of tracing representative minimal feature sets with improved interpretability across various models and datasets."
                    },
                    {
                        "title": "Prediction Modeling and Analysis for Telecom Customer Churn in Two Months",
                        "abstract": "A practical churn customer prediction model is critical to retain customers for telecom companies in the saturated and competitive market. Previous studies focus on predicting churn customers in current or next month, in which telecom companies don't have enough time to develop and carry out churn management strategies. In this paper, we propose a new T+2 churn customer prediction model, in which the churn customers in two months are recognized and the one-month window T+1 is reserved to carry out churn management strategies. However, the predictions for churn customers in two months are much more difficult than in current or next month because of the weaker correlation between the customer information and churn states. Two characteristics of telecom dataset, the discrimination between churn and non-churn customers is complicated and the class imbalance problem is serious, are observed. To discriminate the churn customers accurately, random forest (RF) classifier is chosen because RF solves the nonlinear separable problem with low bias and low variance and handles high feature spaces and large number of training examples. To overcome the imbalance problem, synthetic minority over-sampling with borderline or tomek link, in which the distribution of the samples remains and the number of the training examples becomes larger, is applied. Overall, a precision ratio of about 50% with a recall ratio of about 50% is achieved in the T+2 churn prediction. The proposed prediction model provides an accurate and operable churn customer prediction model for telecom companies."
                    },
                    {
                        "title": "Multi-objective Optimal Roadside Units Deployment in Urban Vehicular Networks",
                        "abstract": "The significance of transportation efficiency, safety, and related services is increasing in urban vehicular networks. Within such networks, roadside units (RSUs) serve as intermediates in facilitating communication. Therefore, the deployment of RSUs is of utmost importance in ensuring the quality of communication services. However, the optimization objectives, such as time delay and deployment cost, are commonly developed from diverse perspectives. As a result, it is possible that conflicts may arise among the objectives. Furthermore, in urban environments, the presence of various obstacles, such as buildings, gardens, lakes, and other infrastructure, poses challenges for the deployment of RSUs. Hence, the deployment encounters significant difficulties due to the existence of multiple objectives, constraints imposed by obstacles, and the necessity to explore a large-scale optimization space. To address this issue, two versions of multi-objective optimization algorithms are proposed in this paper. By utilizing a multi-population strategy and an adaptive exploration technique, the methods efficiently explore a large-scale decision-variable space. In order to mitigate the issue of an overcrowded deployment of RSUs, a calibrating mechanism is adopted to adjust RSU density during the optimization procedures. The proposed methods also take care of data offloading between vehicles and RSUs by setting up an iterative best response sequence game (IBRSG). By comparing the proposed algorithms with several state-of-the-art algorithms, the results demonstrate that our strategies perform better in both high-density and low-density urban scenarios. The results also indicate that the proposed solutions substantially improve the efficiency of vehicular networks."
                    }
                ]
            },
            "7ed524de-4f6b-413d-aa13-3d203721e686": {
                "pk": "7ed524de-4f6b-413d-aa13-3d203721e686",
                "name": "Chen Wei",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "a5809b26-dba8-4fb6-b266-b602ef9327ef": {
                "pk": "a5809b26-dba8-4fb6-b266-b602ef9327ef",
                "name": "Shiying Li",
                "collaborators": [
                    "Chengliang Wang",
                    "Xinrun Chen",
                    "Haojian Ning",
                    "Gustavo K. Rohde",
                    "Caroline Moosmueller",
                    "Mei Shen",
                    "Akram Aldroubi",
                    "Xuwang Yin",
                    "Lingran Xiao",
                    "Yanfei Wang"
                ],
                "domain": [
                    "Optimal Transport",
                    "Image Segmentation",
                    "Machine Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Measure transfer via stochastic slicing and matching",
                        "abstract": "This paper studies iterative schemes for measure transfer and approximation problems, which are defined through a slicing-and-matching procedure. Similar to the sliced Wasserstein distance, these schemes benefit from the availability of closed-form solutions for the one-dimensional optimal transport problem and the associated computational advantages. While such schemes have already been successfully utilized in data science applications, not too many results on their convergence are available. The main contribution of this paper is an almost sure convergence proof for stochastic slicing-and-matching schemes. The proof builds on an interpretation as a stochastic gradient descent scheme on the Wasserstein space. Numerical examples on step-wise image morphing are demonstrated as well."
                    },
                    {
                        "title": "Approximation properties of slice-matching operators",
                        "abstract": "Iterative slice-matching procedures are efficient schemes for transferring a source measure to a target measure, especially in high dimensions. These schemes have been successfully used in applications such as color transfer and shape retrieval, and are guaranteed to converge under regularity assumptions. In this paper, we explore approximation properties related to a single step of such iterative schemes by examining an associated slice-matching operator, depending on a source measure, a target measure, and slicing directions. In particular, we demonstrate an invariance property with respect to the source measure, an equivariance property with respect to the target measure, and Lipschitz continuity concerning the slicing directions. We furthermore establish error bounds corresponding to approximating the target measure by one step of the slice-matching scheme and characterize situations in which the slice-matching operator recovers the optimal transport map between two measures. We also investigate connections to affine registration problems with respect to (sliced) Wasserstein distances. These connections can be also be viewed as extensions to the invariance and equivariance properties of the slice-matching operator and illustrate the extent to which slice-matching schemes incorporate affine effects."
                    },
                    {
                        "title": "Partitioning signal classes using transport transforms for data analysis and machine learning",
                        "abstract": "A relatively new set of transport-based transforms (CDT, R-CDT, LOT) have shown their strength and great potential in various image and data processing tasks such as parametric signal estimation, classification, cancer detection among many others. It is hence worthwhile to elucidate some of the mathematical properties that explain the successes of these transforms when they are used as tools in data analysis, signal processing or data classification. In particular, we give conditions under which classes of signals that are created by algebraic generative models are transformed into convex sets by the transport transforms. Such convexification of the classes simplify the classification and other data analysis and processing problems when viewed in the transform domain. More specifically, we study the extent and limitation of the convexification ability of these transforms under an algebraic generative modeling framework. We hope that this paper will serve as an introduction to these transforms and will encourage mathematicians and other researchers to further explore the theoretical underpinnings and algorithmic tools that will help understand the successes of these transforms and lay the groundwork for further successful applications."
                    },
                    {
                        "title": "Learning Energy-Based Models With Adversarial Training",
                        "abstract": "We study a new approach to learning energy-based models (EBMs) based on adversarial training (AT). We show that (binary) AT learns a special kind of energy function that models the support of the data distribution, and the learning process is closely related to MCMC-based maximum likelihood learning of EBMs. We further propose improved techniques for generative modeling with AT, and demonstrate that this new approach is capable of generating diverse and realistic images. Aside from having competitive image generation performance to explicit EBMs, the studied approach is stable to train, is well-suited for image translation tasks, and exhibits strong out-of-distribution adversarial robustness. Our results demonstrate the viability of the AT approach to generative modeling, suggesting that AT is a competitive alternative approach to learning EBMs."
                    },
                    {
                        "title": "SAM-OCTA: A Fine-Tuning Strategy for Applying Foundation Model to OCTA Image Segmentation Tasks",
                        "abstract": "In the analysis of optical coherence tomography angiography (OCTA) images, the operation of segmenting specific targets is necessary. Existing methods typically train on supervised datasets with limited samples (approximately a few hundred), which can lead to overfitting. To address this, the low-rank adaptation technique is adopted for foundation model fine-tuning and proposed corresponding prompt point generation strategies to process various segmentation tasks on OCTA datasets. This method is named SAM-OCTA and has been experimented on the publicly available OCTA-500 dataset. While achieving state-of-the-art performance metrics, this method accomplishes local vessel segmentation as well as effective artery-vein segmentation, which was not well-solved in previous works. The code is available at: https://github.com/ShellRedia/SAM-OCTA."
                    },
                    {
                        "title": "Automatically layout and visualize the biological pathway map with spectral graph theory",
                        "abstract": "The pathway is a biological term that refers to a series of interactions between molecules in a cell that causes a certain product or a change in the cell. Pathway analysis is a powerful method for gene expression analysis. Through pathway maps, the lists of genes that are differentially expressed across the given phenotypes are translated into various biological phenomena.   Visualizing a pathway map manually is a common practice nowadays because of the limitations of existing solutions to draw complicated graphs (i.e. directed graphs, graphs with edge crossings, etc).   This project provides a solution to draw pathway maps automatically based on spectral graph theory and topological sort. Various methods are taken to enhance pathway maps' readability. Significant reductions in the number of edge crossings and the sum of adjacent nodes are achieved."
                    },
                    {
                        "title": "Nonrigid Object Contact Estimation With Regional Unwrapping Transformer",
                        "abstract": "Acquiring contact patterns between hands and nonrigid objects is a common concern in the vision and robotics community. However, existing learning-based methods focus more on contact with rigid ones from monocular images. When adopting them for nonrigid contact, a major problem is that the existing contact representation is restricted by the geometry of the object. Consequently, contact neighborhoods are stored in an unordered manner and contact features are difficult to align with image cues. At the core of our approach lies a novel hand-object contact representation called RUPs (Region Unwrapping Profiles), which unwrap the roughly estimated hand-object surfaces as multiple high-resolution 2D regional profiles. The region grouping strategy is consistent with the hand kinematic bone division because they are the primitive initiators for a composite contact pattern. Based on this representation, our Regional Unwrapping Transformer (RUFormer) learns the correlation priors across regions from monocular inputs and predicts corresponding contact and deformed transformations. Our experiments demonstrate that the proposed framework can robustly estimate the deformed degrees and deformed transformations, which makes it suitable for both nonrigid and rigid contact."
                    },
                    {
                        "title": "An Accurate and Efficient Neural Network for OCTA Vessel Segmentation and a New Dataset",
                        "abstract": "Optical coherence tomography angiography (OCTA) is a noninvasive imaging technique that can reveal high-resolution retinal vessels. In this work, we propose an accurate and efficient neural network for retinal vessel segmentation in OCTA images. The proposed network achieves accuracy comparable to other SOTA methods, while having fewer parameters and faster inference speed (e.g. 110x lighter and 1.3x faster than U-Net), which is very friendly for industrial applications. This is achieved by applying the modified Recurrent ConvNeXt Block to a full resolution convolutional network. In addition, we create a new dataset containing 918 OCTA images and their corresponding vessel annotations. The data set is semi-automatically annotated with the help of Segment Anything Model (SAM), which greatly improves the annotation speed. For the benefit of the community, our code and dataset can be obtained from https://github.com/nhjydywd/OCTA-FRNet."
                    },
                    {
                        "title": "Hair Segmentation on Time-of-Flight RGBD Images",
                        "abstract": "Robust segmentation of hair from portrait images remains challenging: hair does not conform to a uniform shape, style or even color; dark hair in particular lacks features. We present a novel computational imaging solution that tackles the problem from both input and processing fronts. We explore using Time-of-Flight (ToF) RGBD sensors on recent mobile devices. We first conduct a comprehensive analysis to show that scattering and inter-reflection cause different noise patterns on hair vs. non-hair regions on ToF images, by changing the light path and/or combining multiple paths. We then develop a deep network based approach that employs both ToF depth map and the RGB gradient maps to produce an initial hair segmentation with labeled hair components. We then refine the result by imposing ToF noise prior under the conditional random field. We collect the first ToF RGBD hair dataset with 20k+ head images captured on 30 human subjects with a variety of hairstyles at different view angles. Comprehensive experiments show that our approach outperforms the RGB based techniques in accuracy and robustness and can handle traditionally challenging cases such as dark hair, similar hair/background, similar hair/foreground, etc."
                    },
                    {
                        "title": "SAM-OCTA: Prompting Segment-Anything for OCTA Image Segmentation",
                        "abstract": "Segmenting specific targets or biomarkers is necessary to analyze optical coherence tomography angiography (OCTA) images. Previous methods typically segment all the targets in an OCTA sample, such as retinal vessels (RVs). Although these methods perform well in accuracy and precision, OCTA analyses often focusing local information within the images which has not been fulfilled. In this paper, we propose a method called SAM-OCTA for local segmentation in OCTA images. The method fine-tunes a pre-trained segment anything model (SAM) using low-rank adaptation (LoRA) and utilizes prompt points for local RVs, arteries, and veins segmentation in OCTA. To explore the effect and mechanism of prompt points, we set up global and local segmentation modes with two prompt point generation strategies, namely random selection and special annotation. Considering practical usage, we conducted extended experiments with different model scales and analyzed the model performance before and after fine-tuning besides the general segmentation task. From comprehensive experimental results with the OCTA-500 dataset, our SAM-OCTA method has achieved state-of-the-art performance in common OCTA segmentation tasks related to RV and FAZ, and it also performs accurate segmentation of artery-vein and local vessels. The code is available at https://github.com/ShellRedia/SAM-OCTA-extend."
                    },
                    {
                        "title": "Geodesic Properties of a Generalized Wasserstein Embedding for Time Series Analysis",
                        "abstract": "Transport-based metrics and related embeddings (transforms) have recently been used to model signal classes where nonlinear structures or variations are present. In this paper, we study the geodesic properties of time series data with a generalized Wasserstein metric and the geometry related to their signed cumulative distribution transforms in the embedding space. Moreover, we show how understanding such geometric characteristics can provide added interpretability to certain time series classifiers, and be an inspiration for more robust classifiers."
                    },
                    {
                        "title": "Snake with Shifted Window: Learning to Adapt Vessel Pattern for OCTA Segmentation",
                        "abstract": "Segmenting specific targets or structures in optical coherence tomography angiography (OCTA) images is fundamental for conducting further pathological studies. The retinal vascular layers are rich and intricate, and such vascular with complex shapes can be captured by the widely-studied OCTA images. In this paper, we thus study how to use OCTA images with projection vascular layers to segment retinal structures. To this end, we propose the SSW-OCTA model, which integrates the advantages of deformable convolutions suited for tubular structures and the swin-transformer for global feature extraction, adapting to the characteristics of OCTA modality images. Our model underwent testing and comparison on the OCTA-500 dataset, achieving state-of-the-art performance. The code is available at: https://github.com/ShellRedia/Snake-SWin-OCTA."
                    }
                ]
            },
            "333cf4c9-f4b9-476b-a0ca-8ea9f353ad98": {
                "pk": "333cf4c9-f4b9-476b-a0ca-8ea9f353ad98",
                "name": "Jiachen Zou",
                "collaborators": [
                    "Chen Wei",
                    "Dietmar Heinke",
                    "Quanying Liu"
                ],
                "domain": [
                    "Cognitive Science",
                    "Image Generation",
                    "Human-Computer Interaction",
                    "Concept Representation"
                ],
                "publications": [
                    {
                        "title": "CoCoG-2: Controllable generation of visual stimuli for understanding human concept representation",
                        "abstract": "Humans interpret complex visual stimuli using abstract concepts that facilitate decision-making tasks such as food selection and risk avoidance. Similarity judgment tasks are effective for exploring these concepts. However, methods for controllable image generation in concept space are underdeveloped. In this study, we present a novel framework called CoCoG-2, which integrates generated visual stimuli into similarity judgment tasks. CoCoG-2 utilizes a training-free guidance algorithm to enhance generation flexibility. CoCoG-2 framework is versatile for creating experimental stimuli based on human concepts, supporting various strategies for guiding visual stimuli generation, and demonstrating how these stimuli can validate various experimental hypotheses. CoCoG-2 will advance our understanding of the causal relationship between concept representations and behaviors by generating visual stimuli. The code is available at \\url{https://github.com/ncclab-sustech/CoCoG-2}."
                    },
                    {
                        "title": "CoCoG: Controllable Visual Stimuli Generation based on Human Concept Representations",
                        "abstract": "A central question for cognitive science is to understand how humans process visual objects, i.e, to uncover human low-dimensional concept representation space from high-dimensional visual stimuli. Generating visual stimuli with controlling concepts is the key. However, there are currently no generative models in AI to solve this problem. Here, we present the Concept based Controllable Generation (CoCoG) framework. CoCoG consists of two components, a simple yet efficient AI agent for extracting interpretable concept and predicting human decision-making in visual similarity judgment tasks, and a conditional generation model for generating visual stimuli given the concepts. We quantify the performance of CoCoG from two aspects, the human behavior prediction accuracy and the controllable generation ability. The experiments with CoCoG indicate that 1) the reliable concept embeddings in CoCoG allows to predict human behavior with 64.07\\% accuracy in the THINGS-similarity dataset; 2) CoCoG can generate diverse objects through the control of concepts; 3) CoCoG can manipulate human similarity judgment behavior by intervening key concepts. CoCoG offers visual objects with controlling concepts to advance our understanding of causality in human cognition. The code of CoCoG is available at \\url{https://github.com/ncclab-sustech/CoCoG}."
                    }
                ]
            },
            "82b01d68-cea7-47f8-9596-92657092a946": {
                "pk": "82b01d68-cea7-47f8-9596-92657092a946",
                "name": "Haoyang Qin",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "6137ba36-e623-4a53-aa1b-2a41de0ad56c": {
                "pk": "6137ba36-e623-4a53-aa1b-2a41de0ad56c",
                "name": "Quanying Liu",
                "collaborators": [
                    "Chen Wei",
                    "Wanguang Yin",
                    "Kexin Lou",
                    "Zhichao Liang",
                    "Haiyan Wu",
                    "Zhengming Ma",
                    "Jiachen Zou",
                    "Dietmar Heinke",
                    "Youzhi Qu",
                    "Xuming Ran"
                ],
                "domain": [
                    "Inverse Reinforcement Learning",
                    "Deep Learning",
                    "Brain-Computer Interface",
                    "Cognitive Neuroscience"
                ],
                "publications": [
                    {
                        "title": "Modeling and Interpreting Real-world Human Risk Decision Making with Inverse Reinforcement Learning",
                        "abstract": "We model human decision-making behaviors in a risk-taking task using inverse reinforcement learning (IRL) for the purposes of understanding real human decision making under risk. To the best of our knowledge, this is the first work applying IRL to reveal the implicit reward function in human risk-taking decision making and to interpret risk-prone and risk-averse decision-making policies. We hypothesize that the state history (e.g. rewards and decisions in previous trials) are related to the human reward function, which leads to risk-averse and risk-prone decisions. We design features that reflect these factors in the reward function of IRL and learn the corresponding weight that is interpretable as the importance of features. The results confirm the sub-optimal risk-related decisions of human-driven by the personalized reward function. In particular, the risk-prone person tends to decide based on the current pump number, while the risk-averse person relies on burst information from the previous trial and the average end status. Our results demonstrate that IRL is an effective tool to model human decision-making behavior, as well as to help interpret the human psychological process in risk decision-making."
                    },
                    {
                        "title": "Riemannian Manifold Optimization for Discriminant Subspace Learning",
                        "abstract": "Linear discriminant analysis (LDA) is a widely used algorithm in machine learning to extract a low-dimensional representation of high-dimensional data, it features to find the orthogonal discriminant projection subspace by using the Fisher discriminant criterion. However, the traditional Euclidean-based methods for solving LDA are easily convergent to spurious local minima and hardly obtain an optimal solution. To address such a problem, in this paper, we propose a novel algorithm namely Riemannian-based discriminant analysis (RDA) for subspace learning. In order to obtain an explicit solution, we transform the traditional Euclidean-based methods to the Riemannian manifold space and use the trust-region method to learn the discriminant projection subspace. We compare the proposed algorithm to existing variants of LDA, as well as the unsupervised tensor decomposition methods on image classification tasks. The numerical results suggest that RDA achieves state-of-the-art performance in classification accuracy."
                    },
                    {
                        "title": "Partial Least Square Regression via Three-factor SVD-type Manifold Optimization for EEG Decoding",
                        "abstract": "Partial least square regression (PLSR) is a widely-used statistical model to reveal the linear relationships of latent factors that comes from the independent variables and dependent variables. However, traditional methods to solve PLSR models are usually based on the Euclidean space, and easily getting stuck into a local minimum. To this end, we propose a new method to solve the partial least square regression, named PLSR via optimization on bi-Grassmann manifold (PLSRbiGr). Specifically, we first leverage the three-factor SVD-type decomposition of the cross-covariance matrix defined on the bi-Grassmann manifold, converting the orthogonal constrained optimization problem into an unconstrained optimization problem on bi-Grassmann manifold, and then incorporate the Riemannian preconditioning of matrix scaling to regulate the Riemannian metric in each iteration. PLSRbiGr is validated with a variety of experiments for decoding EEG signals at motor imagery (MI) and steady-state visual evoked potential (SSVEP) task. Experimental results demonstrate that PLSRbiGr outperforms competing algorithms in multiple EEG decoding tasks, which will greatly facilitate small sample data learning."
                    },
                    {
                        "title": "CoCoG: Controllable Visual Stimuli Generation based on Human Concept Representations",
                        "abstract": "A central question for cognitive science is to understand how humans process visual objects, i.e, to uncover human low-dimensional concept representation space from high-dimensional visual stimuli. Generating visual stimuli with controlling concepts is the key. However, there are currently no generative models in AI to solve this problem. Here, we present the Concept based Controllable Generation (CoCoG) framework. CoCoG consists of two components, a simple yet efficient AI agent for extracting interpretable concept and predicting human decision-making in visual similarity judgment tasks, and a conditional generation model for generating visual stimuli given the concepts. We quantify the performance of CoCoG from two aspects, the human behavior prediction accuracy and the controllable generation ability. The experiments with CoCoG indicate that 1) the reliable concept embeddings in CoCoG allows to predict human behavior with 64.07\\% accuracy in the THINGS-similarity dataset; 2) CoCoG can generate diverse objects through the control of concepts; 3) CoCoG can manipulate human similarity judgment behavior by intervening key concepts. CoCoG offers visual objects with controlling concepts to advance our understanding of causality in human cognition. The code of CoCoG is available at \\url{https://github.com/ncclab-sustech/CoCoG}."
                    },
                    {
                        "title": "Memory Sequence Length of Data Sampling Impacts the Adaptation of Meta-Reinforcement Learning Agents",
                        "abstract": "Fast adaptation to new tasks is extremely important for embodied agents in the real world. Meta-reinforcement learning (meta-RL) has emerged as an effective method to enable fast adaptation in unknown environments. Compared to on-policy meta-RL algorithms, off-policy algorithms rely heavily on efficient data sampling strategies to extract and represent the historical trajectories. However, little is known about how different data sampling methods impact the ability of meta-RL agents to represent unknown environments. Here, we investigate the impact of data sampling strategies on the exploration and adaptability of meta-RL agents. Specifically, we conducted experiments with two types of off-policy meta-RL algorithms based on Thompson sampling and Bayes-optimality theories in continuous control tasks within the MuJoCo environment and sparse reward navigation tasks. Our analysis revealed the long-memory and short-memory sequence sampling strategies affect the representation and adaptive capabilities of meta-RL agents. We found that the algorithm based on Bayes-optimality theory exhibited more robust and better adaptability than the algorithm based on Thompson sampling, highlighting the importance of appropriate data sampling strategies for the agent's representation of an unknown environment, especially in the case of sparse rewards."
                    },
                    {
                        "title": "CoCoG-2: Controllable generation of visual stimuli for understanding human concept representation",
                        "abstract": "Humans interpret complex visual stimuli using abstract concepts that facilitate decision-making tasks such as food selection and risk avoidance. Similarity judgment tasks are effective for exploring these concepts. However, methods for controllable image generation in concept space are underdeveloped. In this study, we present a novel framework called CoCoG-2, which integrates generated visual stimuli into similarity judgment tasks. CoCoG-2 utilizes a training-free guidance algorithm to enhance generation flexibility. CoCoG-2 framework is versatile for creating experimental stimuli based on human concepts, supporting various strategies for guiding visual stimuli generation, and demonstrating how these stimuli can validate various experimental hypotheses. CoCoG-2 will advance our understanding of the causal relationship between concept representations and behaviors by generating visual stimuli. The code is available at \\url{https://github.com/ncclab-sustech/CoCoG-2}."
                    },
                    {
                        "title": "HyperNTF: A Hypergraph Regularized Nonnegative Tensor Factorization for Dimensionality Reduction",
                        "abstract": "Tensor decomposition is an effective tool for learning multi-way structures and heterogeneous features from high-dimensional data, such as the multi-view images and multichannel electroencephalography (EEG) signals, are often represented by tensors. However, most of tensor decomposition methods are the linear feature extraction techniques, which are unable to reveal the nonlinear structure within high-dimensional data. To address such problem, a lot of algorithms have been proposed for simultaneously performs linear and non-linear feature extraction. A representative algorithm is the Graph Regularized Non-negative Matrix Factorization (GNMF) for image clustering. However, the normal 2-order graph can only models the pairwise similarity of objects, which cannot sufficiently exploit the complex structures of samples. Thus, we propose a novel method, named Hypergraph Regularized Non-negative Tensor Factorization (HyperNTF), which utilizes hypergraph to encode the complex connections among samples and employs the factor matrix corresponding with last mode of Canonical Polyadic (CP) decomposition as low-dimensional representation. Extensive experiments on synthetic manifolds, real-world image datasets, and EEG signals, demonstrating that HyperNTF outperforms the state-of-the-art methods in terms of dimensionality reduction, clustering, and classification."
                    },
                    {
                        "title": "Controlling network-coupled neural dynamics with nonlinear network control theory",
                        "abstract": "This paper addresses the problem of controlling the temporal dynamics of complex nonlinear network-coupled dynamical systems, specifically in terms of neurodynamics. Based on the Lyapunov direct method, we derive a control strategy with theoretical guarantees of controllability. To verify the performance of the derived control strategy, we perform numerical experiments on two nonlinear network-coupled dynamical systems that emulate phase synchronization and neural population dynamics. The results demonstrate the feasibility and effectiveness of our control strategy."
                    },
                    {
                        "title": "Multi-objective optimization via evolutionary algorithm (MOVEA) for high-definition transcranial electrical stimulation of the human brain",
                        "abstract": "Designing a transcranial electrical stimulation (TES) strategy requires considering multiple objectives, such as intensity in the target area, focality, stimulation depth, and avoidance zone, which are often mutually exclusive. A computational framework for optimizing different strategies and comparing trade-offs between these objectives is currently lacking. In this paper, we propose a general framework called multi-objective optimization via evolutionary algorithms (MOVEA) to address the non-convex optimization problem in designing TES strategies without predefined direction. MOVEA enables simultaneous optimization of multiple targets through Pareto optimization, generating a Pareto front after a single run without manual weight adjustment and allowing easy expansion to more targets. This Pareto front consists of optimal solutions that meet various requirements while respecting trade-off relationships between conflicting objectives such as intensity and focality. MOVEA is versatile and suitable for both transcranial alternating current stimulation (tACS) and transcranial temporal interference stimulation (tTIS) based on high definition (HD) and two-pair systems. We performed a comprehensive comparison between tACS and tTIS in terms of intensity, focality, and steerability for targets at different depths.MOVEA facilitates the optimization of TES based on specific objectives and constraints, advancing tTIS and tACS-based neuromodulation in understanding the causal relationship between brain regions and cognitive functions and in treating diseases. The code for MOVEA is available at https://github.com/ncclabsustech/MOVEA."
                    },
                    {
                        "title": "Detecting Out-of-distribution Samples via Variational Auto-encoder with Reliable Uncertainty Estimation",
                        "abstract": "Variational autoencoders (VAEs) are influential generative models with rich representation capabilities from the deep neural network architecture and Bayesian method. However, VAE models have a weakness that assign a higher likelihood to out-of-distribution (OOD) inputs than in-distribution (ID) inputs. To address this problem, a reliable uncertainty estimation is considered to be critical for in-depth understanding of OOD inputs. In this study, we propose an improved noise contrastive prior (INCP) to be able to integrate into the encoder of VAEs, called INCPVAE. INCP is scalable, trainable and compatible with VAEs, and it also adopts the merits from the INCP for uncertainty estimation. Experiments on various datasets demonstrate that compared to the standard VAEs, our model is superior in uncertainty estimation for the OOD data and is robust in anomaly detection tasks. The INCPVAE model obtains reliable uncertainty estimation for OOD inputs and solves the OOD problem in VAE models."
                    },
                    {
                        "title": "A novel convolutional neural network model to remove muscle artifacts from EEG",
                        "abstract": "The recorded electroencephalography (EEG) signals are usually contaminated by many artifacts. In recent years, deep learning models have been used for denoising of electroencephalography (EEG) data and provided comparable performance with that of traditional techniques. However, the performance of the existing networks in electromyograph (EMG) artifact removal was limited and suffered from the over-fitting problem. Here we introduce a novel convolutional neural network (CNN) with gradually ascending feature dimensions and downsampling in time series for removing muscle artifacts in EEG data. Compared with other types of convolutional networks, this model largely eliminates the over-fitting and significantly outperforms four benchmark networks in EEGdenoiseNet. Our study suggested that the deep network architecture might help avoid overfitting and better remove EMG artifacts in EEG."
                    },
                    {
                        "title": "Embedding Decomposition for Artifacts Removal in EEG Signals",
                        "abstract": "Electroencephalogram (EEG) recordings are often contaminated with artifacts. Various methods have been developed to eliminate or weaken the influence of artifacts. However, most of them rely on prior experience for analysis. Here, we propose an deep learning framework to separate neural signal and artifacts in the embedding space and reconstruct the denoised signal, which is called DeepSeparator. DeepSeparator employs an encoder to extract and amplify the features in the raw EEG, a module called decomposer to extract the trend, detect and suppress artifact and a decoder to reconstruct the denoised signal. Besides, DeepSeparator can extract the artifact, which largely increases the model interpretability. The proposed method is tested with a semi-synthetic EEG dataset and a real task-related EEG dataset, suggesting that DeepSeparator outperforms the conventional models in both EOG and EMG artifact removal. DeepSeparator can be extended to multi-channel EEG and data of any length. It may motivate future developments and application of deep learning-based EEG denoising. The code for DeepSeparator is available at https://github.com/ncclabsustech/DeepSeparator."
                    },
                    {
                        "title": "A Hybrid Brain-Computer Interface Using Motor Imagery and SSVEP Based on Convolutional Neural Network",
                        "abstract": "The key to electroencephalography (EEG)-based brain-computer interface (BCI) lies in neural decoding, and its accuracy can be improved by using hybrid BCI paradigms, that is, fusing multiple paradigms. However, hybrid BCIs usually require separate processing processes for EEG signals in each paradigm, which greatly reduces the efficiency of EEG feature extraction and the generalizability of the model. Here, we propose a two-stream convolutional neural network (TSCNN) based hybrid brain-computer interface. It combines steady-state visual evoked potential (SSVEP) and motor imagery (MI) paradigms. TSCNN automatically learns to extract EEG features in the two paradigms in the training process, and improves the decoding accuracy by 25.4% compared with the MI mode, and 2.6% compared with SSVEP mode in the test data. Moreover, the versatility of TSCNN is verified as it provides considerable performance in both single-mode (70.2% for MI, 93.0% for SSVEP) and hybrid-mode scenarios (95.6% for MI-SSVEP hybrid). Our work will facilitate the real-world applications of EEG-based BCI systems."
                    },
                    {
                        "title": "Advancing EEG/MEG Source Imaging with Geometric-Informed Basis Functions",
                        "abstract": "Electroencephalography (EEG) and Magnetoencephalography (MEG) are pivotal in understanding brain activity but are limited by their poor spatial resolution. EEG/MEG source imaging (ESI) infers the high-resolution electric field distribution in the brain based on the low-resolution scalp EEG/MEG observations. However, the ESI problem is ill-posed, and how to bring neuroscience priors into ESI method is the key. Here, we present a novel method which utilizes the Brain Geometric-informed Basis Functions (GBFs) as priors to enhance EEG/MEG source imaging. Through comprehensive experiments on both synthetic data and real task EEG data, we demonstrate the superiority of GBFs over traditional spatial basis functions (e.g., Harmonic and MSP), as well as existing ESI methods (e.g., dSPM, MNE, sLORETA, eLORETA). GBFs provide robust ESI results under different noise levels, and result in biologically interpretable EEG sources. We believe the high-resolution EEG source imaging from GBFs will greatly advance neuroscience research."
                    },
                    {
                        "title": "Reverse engineering the brain input: Network control theory to identify cognitive task-related control nodes",
                        "abstract": "The human brain receives complex inputs when performing cognitive tasks, which range from external inputs via the senses to internal inputs from other brain regions. However, the explicit inputs to the brain during a cognitive task remain unclear. Here, we present an input identification framework for reverse engineering the control nodes and the corresponding inputs to the brain. The framework is verified with synthetic data generated by a predefined linear system, indicating it can robustly reconstruct data and recover the inputs. Then we apply the framework to the real motor-task fMRI data from 200 human subjects. Our results show that the model with sparse inputs can reconstruct neural dynamics in motor tasks ($EV=0.779$) and the identified 28 control nodes largely overlap with the motor system. Underpinned by network control theory, our framework offers a general tool for understanding brain inputs."
                    },
                    {
                        "title": "Bigeminal Priors Variational auto-encoder",
                        "abstract": "Variational auto-encoders (VAEs) are an influential and generally-used class of likelihood-based generative models in unsupervised learning. The likelihood-based generative models have been reported to be highly robust to the out-of-distribution (OOD) inputs and can be a detector by assuming that the model assigns higher likelihoods to the samples from the in-distribution (ID) dataset than an OOD dataset. However, recent works reported a phenomenon that VAE recognizes some OOD samples as ID by assigning a higher likelihood to the OOD inputs compared to the one from ID. In this work, we introduce a new model, namely Bigeminal Priors Variational auto-encoder (BPVAE), to address this phenomenon. The BPVAE aims to enhance the robustness of the VAEs by combing the power of VAE with the two independent priors that belong to the training dataset and simple dataset, which complexity is lower than the training dataset, respectively. BPVAE learns two datasets'features, assigning a higher likelihood for the training dataset than the simple dataset. In this way, we can use BPVAE's density estimate for detecting the OOD samples. Quantitative experimental results suggest that our model has better generalization capability and stronger robustness than the standard VAEs, proving the effectiveness of the proposed approach of hybrid learning by collaborative priors. Overall, this work paves a new avenue to potentially overcome the OOD problem via multiple latent priors modeling."
                    }
                ]
            }
        }
    },
    "2309.16779": {
        "paper_data": {
            "title": "Intriguing properties of generative classifiers",
            "url": "http://arxiv.org/abs/2309.16779v2",
            "arxiv_id": "2309.16779",
            "authors": [
                "Priyank Jaini",
                "Kevin Clark",
                "Robert Geirhos"
            ],
            "abstract": "What is the best paradigm to recognize objects -- discriminative inference (fast but potentially prone to shortcut learning) or using a generative model (slow but potentially more robust)? We build on recent advances in generative modeling that turn text-to-image models into classifiers. This allows us to study their behavior and to compare them against discriminative models and human psychophysical data. We report four intriguing emergent properties of generative classifiers: they show a record-breaking human-like shape bias (99% for Imagen), near human-level out-of-distribution accuracy, state-of-the-art alignment with human classification errors, and they understand certain perceptual illusions. Our results indicate that while the current dominant paradigm for modeling human object recognition is discriminative inference, zero-shot generative models approximate human object recognition data surprisingly well.",
            "introduction": " introduction, using a generative model comes with advantages and disadvan- tages: potentially better generalization currently comes at the cost of being slower computationally compared to standard discriminative models. While this doesn\u2019t matter much for the purpose of anal- yses, it is a big drawback in practical applications and any approaches that improve speed would be most welcome\u2014in particular, generating at least one prediction (i.e., image generation) per class as we currently do is both expensive and slow. Furthermore, the models we investigate all differ from another in more than one ways. For instance, their training data, architecture, and training procedure is not identical, thus any differences between the models cannot currently be attributed to a single factor. That said, through the inclusion of a set of diverse models covering pixel-based diffusion, latent space diffusion, and autoregressive models we seek to at least cover a variety of generative classifiers in order to ensure that the",
            "references": [
                {
                    "title": "Spatial-frequency channels, shape bias, and adversarial robustness",
                    "abstract": "What spatial frequency information do humans and neural networks use to recognize objects? In neuroscience, critical band masking is an established tool that can reveal the frequency-selective filters used for object recognition. Critical band masking measures the sensitivity of recognition performance to noise added at each spatial frequency. Existing critical band masking studies show that humans recognize periodic patterns (gratings) and letters by means of a spatial-frequency filter (or\"channel'') that has a frequency bandwidth of one octave (doubling of frequency). Here, we introduce critical band masking as a task for network-human comparison and test 14 humans and 76 neural networks on 16-way ImageNet categorization in the presence of narrowband noise. We find that humans recognize objects in natural images using the same one-octave-wide channel that they use for letters and gratings, making it a canonical feature of human object recognition. On the other hand, the neural network channel, across various architectures and training strategies, is 2-4 times as wide as the human channel. In other words, networks are vulnerable to high and low frequency noise that does not affect human performance. Adversarial and augmented-image training are commonly used to increase network robustness and shape bias. Does this training align network and human object recognition channels? Three network channel properties (bandwidth, center frequency, peak noise sensitivity) correlate strongly with shape bias (53% variance explained) and with robustness of adversarially-trained networks (74% variance explained). Adversarial training increases robustness but expands the channel bandwidth even further away from the human bandwidth. Thus, critical band masking reveals that the network channel is more than twice as wide as the human channel, and that adversarial training only increases this difference."
                },
                {
                    "title": "DreamTeacher: Pretraining Image Backbones with Deep Generative Models",
                    "abstract": "In this work, we introduce a self-supervised feature representation learning framework DreamTeacher that utilizes generative networks for pre-training downstream image backbones. We propose to distill knowledge from a trained generative model into standard image backbones that have been well engineered for specific perception tasks. We investigate two types of knowledge distillation: 1) distilling learned generative features onto target image backbones as an alternative to pretraining these backbones on large labeled datasets such as ImageNet, and 2) distilling labels obtained from generative networks with task heads onto logits of target backbones. We perform extensive analyses on multiple generative models, dense prediction benchmarks, and several pre-training regimes. We empirically find that our DreamTeacher significantly outperforms existing self-supervised representation learning approaches across the board. Unsupervised ImageNet pre-training with DreamTeacher leads to significant improvements over ImageNet classification pre-training on downstream datasets, showcasing generative models, and diffusion generative models specifically, as a promising approach to representation learning on large, diverse datasets without requiring manual annotation."
                },
                {
                    "title": "Image Captioners Are Scalable Vision Learners Too",
                    "abstract": "Contrastive pretraining on image-text pairs from the web is one of the most popular large-scale pretraining strategies for vision backbones, especially in the context of large multimodal models. At the same time, image captioning on this type of data is commonly considered an inferior pretraining strategy. In this paper, we perform a fair comparison of these two pretraining strategies, carefully matching training data, compute, and model capacity. Using a standard encoder-decoder transformer, we find that captioning alone is surprisingly effective: on classification tasks, captioning produces vision encoders competitive with contrastively pretrained encoders, while surpassing them on vision&language tasks. We further analyze the effect of the model architecture and scale, as well as the pretraining data on the representation quality, and find that captioning exhibits the same or better scaling behavior along these axes. Overall our results show that plain image captioning is a more powerful pretraining strategy than was previously believed."
                },
                {
                    "title": "Text-to-Image Diffusion Models are Zero-Shot Classifiers",
                    "abstract": "The excellent generative capabilities of text-to-image diffusion models suggest they learn informative representations of image-text data. However, what knowledge their representations capture is not fully understood, and they have not been thoroughly explored on downstream tasks. We investigate diffusion models by proposing a method for evaluating them as zero-shot classifiers. The key idea is using a diffusion model's ability to denoise a noised image given a text description of a label as a proxy for that label's likelihood. We apply our method to Stable Diffusion and Imagen, using it to probe fine-grained aspects of the models' knowledge and comparing them with CLIP's zero-shot abilities. They perform competitively with CLIP on a wide range of zero-shot image classification datasets. Additionally, they achieve state-of-the-art results on shape/texture bias tests and can successfully perform attribute binding while CLIP cannot. Although generative pre-training is prevalent in NLP, visual foundation models often use other methods such as contrastive learning. Based on our findings, we argue that generative pre-training should be explored as a compelling alternative for vision-language tasks."
                },
                {
                    "title": "Unleashing Text-to-Image Diffusion Models for Visual Perception",
                    "abstract": "Diffusion models (DMs) have become the new trend of generative models and have demonstrated a powerful ability of conditional synthesis. Among those, text-to-image diffusion models pre-trained on large-scale image-text pairs are highly controllable by customizable prompts. Unlike the unconditional generative models that focus on low-level attributes and details, text-to-image diffusion models contain more high-level knowledge thanks to the vision-language pre-training. In this paper, we propose VPD (Visual Perception with pre-trained Diffusion models), a new framework that exploits the semantic information of a pre-trained text-to-image diffusion model in visual perception tasks. Instead of using the pre-trained denoising autoencoder in a diffusion-based pipeline, we simply use it as a backbone and aim to study how to take full advantage of the learned knowledge. Specifically, we prompt the denoising decoder with proper textual inputs and refine the text features with an adapter, leading to a better alignment to the pre-trained stage and making the visual contents interact with the text prompts. We also propose to utilize the cross-attention maps between the visual features and the text features to provide explicit guidance. Compared with other pre-training methods, we show that vision-language pre-trained diffusion models can be faster adapted to downstream visual perception tasks using the proposed VPD. Extensive experiments on semantic segmentation, referring image segmentation, and depth estimation demonstrate the effectiveness of our method. Notably, VPD attains 0.254 RMSE on NYUv2 depth estimation and 73.3% oIoU on RefCOCO-val referring image segmentation, establishing new records on these two benchmarks. Code is available at https://github.com/wl-zhao/VPD."
                },
                {
                    "title": "Scaling Vision Transformers to 22 Billion Parameters",
                    "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                },
                {
                    "title": "Muse: Text-To-Image Generation via Masked Generative Transformers",
                    "abstract": "We present Muse, a text-to-image Transformer model that achieves state-of-the-art image generation performance while being significantly more efficient than diffusion or autoregressive models. Muse is trained on a masked modeling task in discrete token space: given the text embedding extracted from a pre-trained large language model (LLM), Muse is trained to predict randomly masked image tokens. Compared to pixel-space diffusion models, such as Imagen and DALL-E 2, Muse is significantly more efficient due to the use of discrete tokens and requiring fewer sampling iterations; compared to autoregressive models, such as Parti, Muse is more efficient due to the use of parallel decoding. The use of a pre-trained LLM enables fine-grained language understanding, translating to high-fidelity image generation and the understanding of visual concepts such as objects, their spatial relationships, pose, cardinality etc. Our 900M parameter model achieves a new SOTA on CC3M, with an FID score of 6.06. The Muse 3B parameter model achieves an FID of 7.88 on zero-shot COCO evaluation, along with a CLIP score of 0.32. Muse also directly enables a number of image editing applications without the need to fine-tune or invert the model: inpainting, outpainting, and mask-free editing. More results are available at https://muse-model.github.io"
                },
                {
                    "title": "Peekaboo: Text to Image Diffusion Models are Zero-Shot Segmentors",
                    "abstract": "Recently, text-to-image diffusion models have shown remarkable capabilities in creating realistic images from natural language prompts. However, few works have explored using these models for semantic localization or grounding. In this work, we explore how an off-the-shelf text-to-image diffusion model, trained without exposure to localization information, can ground various semantic phrases without segmentation-specific re-training. We introduce an inference time optimization process capable of generating segmentation masks conditioned on natural language prompts. Our proposal, Peekaboo, is a first-of-its-kind zero-shot, open-vocabulary, unsupervised semantic grounding technique leveraging diffusion models without any training. We evaluate Peekaboo on the Pascal VOC dataset for unsupervised semantic segmentation and the RefCOCO dataset for referring segmentation, showing results competitive with promising results. We also demonstrate how Peekaboo can be used to generate images with transparency, even though the underlying diffusion model was only trained on RGB images - which to our knowledge we are the first to attempt. Please see our project page, including our code: https://ryanndagreat.github.io/peekaboo"
                },
                {
                    "title": "Understanding Diffusion Models: A Unified Perspective",
                    "abstract": "Diffusion models have shown incredible capabilities as generative models; indeed, they power the current state-of-the-art models on text-conditioned image generation such as Imagen and DALL-E 2. In this work we review, demystify, and unify the understanding of diffusion models across both variational and score-based perspectives. We first derive Variational Diffusion Models (VDM) as a special case of a Markovian Hierarchical Variational Autoencoder, where three key assumptions enable tractable computation and scalable optimization of the ELBO. We then prove that optimizing a VDM boils down to learning a neural network to predict one of three potential objectives: the original source input from any arbitrary noisification of it, the original source noise from any arbitrarily noisified input, or the score function of a noisified input at any arbitrary noise level. We then dive deeper into what it means to learn the score function, and connect the variational perspective of a diffusion model explicitly with the Score-based Generative Modeling perspective through Tweedie's Formula. Lastly, we cover how to learn a conditional distribution using diffusion models via guidance."
                },
                {
                    "title": "Scaling Autoregressive Models for Content-Rich Text-to-Image Generation",
                    "abstract": "We present the Pathways Autoregressive Text-to-Image (Parti) model, which generates high-fidelity photorealistic images and supports content-rich synthesis involving complex compositions and world knowledge. Parti treats text-to-image generation as a sequence-to-sequence modeling problem, akin to machine translation, with sequences of image tokens as the target outputs rather than text tokens in another language. This strategy can naturally tap into the rich body of prior work on large language models, which have seen continued advances in capabilities and performance through scaling data and model sizes. Our approach is simple: First, Parti uses a Transformer-based image tokenizer, ViT-VQGAN, to encode images as sequences of discrete tokens. Second, we achieve consistent quality improvements by scaling the encoder-decoder Transformer model up to 20B parameters, with a new state-of-the-art zero-shot FID score of 7.23 and finetuned FID score of 3.22 on MS-COCO. Our detailed analysis on Localized Narratives as well as PartiPrompts (P2), a new holistic benchmark of over 1600 English prompts, demonstrate the effectiveness of Parti across a wide variety of categories and difficulty aspects. We also explore and highlight limitations of our models in order to define and exemplify key areas of focus for further improvements. See https://parti.research.google/ for high-resolution images."
                },
                {
                    "title": "Denoising Pretraining for Semantic Segmentation",
                    "abstract": "Semantic segmentation labels are expensive and time consuming to acquire. To improve label efficiency of semantic segmentation models, we revisit denoising autoencoders and study the use of a denoising objective for pretraining UNets. We pretrain a Transformer-based UNet as a denoising autoencoder, followed by fine-tuning on semantic segmentation using few labeled examples. Denoising pretraining outperforms training from random initialization, and even supervised ImageNet-21K pretraining of the encoder when the number of labeled images is small. A key advantage of denoising pretraining over supervised pretraining of the backbone is the ability to pretrain the decoder, which would otherwise be randomly initialized. We thus propose a novel Decoder Denoising Pretraining (DDeP) method, in which we initialize the encoder using supervised learning and pretrain only the decoder using the denoising objective. Despite its simplicity, DDeP achieves state-of-the-art results on label-efficient semantic segmentation, offering considerable gains on the Cityscapes, Pascal Context, and ADE20K datasets."
                },
                {
                    "title": "Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding",
                    "abstract": "We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding. Imagen builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation. Our key discovery is that generic large language models (e.g. T5), pretrained on text-only corpora, are surprisingly effective at encoding text for image synthesis: increasing the size of the language model in Imagen boosts both sample fidelity and image-text alignment much more than increasing the size of the image diffusion model. Imagen achieves a new state-of-the-art FID score of 7.27 on the COCO dataset, without ever training on COCO, and human raters find Imagen samples to be on par with the COCO data itself in image-text alignment. To assess text-to-image models in greater depth, we introduce DrawBench, a comprehensive and challenging benchmark for text-to-image models. With DrawBench, we compare Imagen with recent methods including VQ-GAN+CLIP, Latent Diffusion Models, and DALL-E 2, and find that human raters prefer Imagen over other models in side-by-side comparisons, both in terms of sample quality and image-text alignment. See https://imagen.research.google/ for an overview of the results."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Score-Based Generative Classifiers",
                    "abstract": "The tremendous success of generative models in recent years raises the question whether they can also be used to perform classification. Generative models have been used as adversarially robust classifiers on simple datasets such as MNIST, but this robustness has not been observed on more complex datasets like CIFAR-10. Additionally, on natural image datasets, previous results have suggested a trade-off between the likelihood of the data and classification accuracy. In this work, we investigate score-based generative models as classifiers for natural images. We show that these models not only obtain competitive likelihood values but simultaneously achieve state-of-the-art classification accuracy for generative classifiers on CIFAR-10. Nevertheless, we find that these models are only slightly, if at all, more robust than discriminative baseline models on out-of-distribution tasks based on common image corruptions. Similarly and contrary to prior results, we find that score-based are prone to worst-case distribution shifts in the form of adversarial perturbations. Our work highlights that score-based generative models are closing the gap in classification accuracy compared to standard discriminative models. While they do not yet deliver on the promise of adversarial and out-of-domain robustness, they provide a different approach to classification that warrants further research."
                },
                {
                    "title": "Variational Diffusion Models",
                    "abstract": "Diffusion-based generative models have demonstrated a capacity for perceptually impressive synthesis, but can they also be great likelihood-based models? We answer this in the affirmative, and introduce a family of diffusion-based generative models that obtain state-of-the-art likelihoods on standard image density estimation benchmarks. Unlike other diffusion-based models, our method allows for efficient optimization of the noise schedule jointly with the rest of the model. We show that the variational lower bound (VLB) simplifies to a remarkably short expression in terms of the signal-to-noise ratio of the diffused data, thereby improving our theoretical understanding of this model class. Using this insight, we prove an equivalence between several models proposed in the literature. In addition, we show that the continuous-time VLB is invariant to the noise schedule, except for the signal-to-noise ratio at its endpoints. This enables us to learn a noise schedule that minimizes the variance of the resulting VLB estimator, leading to faster optimization. Combining these advances with architectural improvements, we obtain state-of-the-art likelihoods on image density estimation benchmarks, outperforming autoregressive models that have dominated these benchmarks for many years, with often significantly faster optimization. In addition, we show how to use the model as part of a bits-back compression scheme, and demonstrate lossless compression rates close to the theoretical optimum. Code is available at https://github.com/google-research/vdm ."
                },
                {
                    "title": "Partial success in closing the gap between human and machine vision",
                    "abstract": "A few years ago, the first CNN surpassed human performance on ImageNet. However, it soon became clear that machines lack robustness on more challenging test cases, a major obstacle towards deploying machines\"in the wild\"and towards obtaining better computational models of human visual perception. Here we ask: Are we making progress in closing the gap between human and machine vision? To answer this question, we tested human observers on a broad range of out-of-distribution (OOD) datasets, recording 85,120 psychophysical trials across 90 participants. We then investigated a range of promising machine learning developments that crucially deviate from standard supervised CNNs along three axes: objective function (self-supervised, adversarially trained, CLIP language-image training), architecture (e.g. vision transformers), and dataset size (ranging from 1M to 1B). Our findings are threefold. (1.) The longstanding distortion robustness gap between humans and CNNs is closing, with the best models now exceeding human feedforward performance on most of the investigated OOD datasets. (2.) There is still a substantial image-level consistency gap, meaning that humans make different errors than models. In contrast, most models systematically agree in their categorisation errors, even substantially different ones like contrastive self-supervised vs. standard supervised models. (3.) In many cases, human-to-model consistency improves when training dataset size is increased by one to three orders of magnitude. Our results give reason for cautious optimism: While there is still much room for improvement, the behavioural difference between human and machine vision is narrowing. In order to measure future progress, 17 OOD datasets with image-level human behavioural data and evaluation code are provided as a toolbox and benchmark at: https://github.com/bethgelab/model-vs-human/"
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Controversial stimuli: Pitting neural networks against each other as models of human cognition",
                    "abstract": "Distinct scientific theories can make similar predictions. To adjudicate between theories, we must design experiments for which the theories make distinct predictions. Here we consider the problem of comparing deep neural networks as models of human visual recognition. To efficiently compare models\u2019 ability to predict human responses, we synthesize controversial stimuli: images for which different models produce distinct responses. We applied this approach to two visual recognition tasks, handwritten digits (MNIST) and objects in small natural images (CIFAR-10). For each task, we synthesized controversial stimuli to maximize the disagreement among models which employed different architectures and recognition algorithms. Human subjects viewed hundreds of these stimuli, as well as natural examples, and judged the probability of presence of each digit/object category in each image. We quantified how accurately each model predicted the human judgments. The best-performing models were a generative analysis-by-synthesis model (based on variational autoencoders) for MNIST and a hybrid discriminative\u2013generative joint energy model for CIFAR-10. These deep neural networks (DNNs), which model the distribution of images, performed better than purely discriminative DNNs, which learn only to map images to labels. None of the candidate models fully explained the human responses. Controversial stimuli generalize the concept of adversarial examples, obviating the need to assume a ground-truth model. Unlike natural images, controversial stimuli are not constrained to the stimulus distribution models are trained on, thus providing severe out-of-distribution tests that reveal the models\u2019 inductive biases. Controversial stimuli therefore provide powerful probes of discrepancies between models and human perception."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Beyond accuracy: quantifying trial-by-trial behaviour of CNNs and humans by measuring error consistency",
                    "abstract": "A central problem in cognitive science and behavioural neuroscience as well as in machine learning and artificial intelligence research is to ascertain whether two or more decision makers (e.g. brains or algorithms) use the same strategy. Accuracy alone cannot distinguish between strategies: two systems may achieve similar accuracy with very different strategies. The need to differentiate beyond accuracy is particularly pressing if two systems are at or near ceiling performance, like Convolutional Neural Networks (CNNs) and humans on visual object recognition. Here we introduce trial-by-trial error consistency, a quantitative analysis for measuring whether two decision making systems systematically make errors on the same inputs. Making consistent errors on a trial-by-trial basis is a necessary condition if we want to ascertain similar processing strategies between decision makers. Our analysis is applicable to compare algorithms with algorithms, humans with humans, and algorithms with humans. When applying error consistency to visual object recognition we obtain three main findings: (1.) Irrespective of architecture, CNNs are remarkably consistent with one another (2.) The consistency between CNNs and human observers, however, is little above what can be expected by chance alone--indicating that humans and CNNs are likely implementing very different strategies (3.) CORnet-S, a recurrent model termed the \"current best model of the primate ventral visual stream\", fails to capture essential characteristics of human behavioural data and behaves essentially like a ResNet-50 in our analysis--that is, just like a standard feedforward network. Taken together, error consistency analysis suggests that the strategies used by human and machine vision are still very different--but we envision our general-purpose error consistency analysis to serve as a fruitful tool for quantifying future progress."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Improved Techniques for Training Score-Based Generative Models",
                    "abstract": "Score-based generative models can produce high quality image samples comparable to GANs, without requiring adversarial optimization. However, existing training procedures are limited to images of low resolution (typically below 32x32), and can be unstable under some settings. We provide a new theoretical analysis of learning and sampling from score models in high dimensional spaces, explaining existing failure modes and motivating new solutions that generalize across datasets. To enhance stability, we also propose to maintain an exponential moving average of model weights. With these improvements, we can effortlessly scale score-based generative models to images with unprecedented resolutions ranging from 64x64 to 256x256. Our score-based models can generate high-fidelity samples that rival best-in-class GANs on various image datasets, including CelebA, FFHQ, and multiple LSUN categories."
                },
                {
                    "title": "The Origins and Prevalence of Texture Bias in Convolutional Neural Networks",
                    "abstract": "Recent work has indicated that, unlike humans, ImageNet-trained CNNs tend to classify images by texture rather than by shape. How pervasive is this bias, and where does it come from? We find that, when trained on datasets of images with conflicting shape and texture, CNNs learn to classify by shape at least as easily as by texture. What factors, then, produce the texture bias in CNNs trained on ImageNet? Different unsupervised training objectives and different architectures have small but significant and largely independent effects on the level of texture bias. However, all objectives and architectures still lead to models that make texture-based classification decisions a majority of the time, even if shape information is decodable from their hidden representations. The effect of data augmentation is much larger. By taking less aggressive random crops at training time and applying simple, naturalistic augmentation (color distortion, noise, and blur), we train models that classify ambiguous images by shape a majority of the time, and outperform baselines on out-of-distribution test sets. Our results indicate that apparent differences in the way humans and ImageNet-trained CNNs process images may arise not from differences in their internal workings, but from differences in the data that they see."
                },
                {
                    "title": "Deep convolutional networks do not classify based on global object shape",
                    "abstract": "Deep convolutional networks (DCNNs) are achieving previously unseen performance in object classification, raising questions about whether DCNNs operate similarly to human vision. In biological vision, shape is arguably the most important cue for recognition. We tested the role of shape information in DCNNs trained to recognize objects. In Experiment 1, we presented a trained DCNN with object silhouettes that preserved overall shape but were filled with surface texture taken from other objects. Shape cues appeared to play some role in the classification of artifacts, but little or none for animals. In Experiments 2\u20134, DCNNs showed no ability to classify glass figurines or outlines but correctly classified some silhouettes. Aspects of these results led us to hypothesize that DCNNs do not distinguish object\u2019s bounding contours from other edges, and that DCNNs access some local shape features, but not global shape. In Experiment 5, we tested this hypothesis with displays that preserved local features but disrupted global shape, and vice versa. With disrupted global shape, which reduced human accuracy to 28%, DCNNs gave the same classification labels as with ordinary shapes. Conversely, local contour changes eliminated accurate DCNN classification but caused no difficulty for human observers. These results provide evidence that DCNNs have access to some local shape information in the form of local edge relations, but they have no access to global object shapes."
                },
                {
                    "title": "ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness",
                    "abstract": "Convolutional Neural Networks (CNNs) are commonly thought to recognise objects by learning increasingly complex representations of object shapes. Some recent studies suggest a more important role of image textures. We here put these conflicting hypotheses to a quantitative test by evaluating CNNs and human observers on images with a texture-shape cue conflict. We show that ImageNet-trained CNNs are strongly biased towards recognising textures rather than shapes, which is in stark contrast to human behavioural evidence and reveals fundamentally different classification strategies. We then demonstrate that the same standard architecture (ResNet-50) that learns a texture-based representation on ImageNet is able to learn a shape-based representation instead when trained on \"Stylized-ImageNet\", a stylized version of ImageNet. This provides a much better fit for human behavioural performance in our well-controlled psychophysical lab setting (nine experiments totalling 48,560 psychophysical trials across 97 observers) and comes with a number of unexpected emergent benefits such as improved object detection performance and previously unseen robustness towards a wide range of image distortions, highlighting advantages of a shape-based representation."
                },
                {
                    "title": "Towards the first adversarially robust neural network model on MNIST",
                    "abstract": "Despite much effort, deep neural networks remain highly susceptible to tiny input perturbations and even for MNIST, one of the most common toy datasets in computer vision, no neural network model exists for which adversarial perturbations are large and make semantic sense to humans. We show that even the widely recognized and by far most successful defense by Madry et al. (1) overfits on the L-infinity metric (it's highly susceptible to L2 and L0 perturbations), (2) classifies unrecognizable images with high certainty, (3) performs not much better than simple input binarization and (4) features adversarial perturbations that make little sense to humans. These results suggest that MNIST is far from being solved in terms of adversarial robustness. We present a novel robust classification model that performs analysis by synthesis using learned class-conditional data distributions. We derive bounds on the robustness and go to great length to empirically evaluate our model using maximally effective adversarial attacks by (a) applying decision-based, score-based, gradient-based and transfer-based attacks for several different Lp norms, (b) by designing a new attack that exploits the structure of our defended model and (c) by devising a novel decision-based attack that seeks to minimize the number of perturbed pixels (L0). The results suggest that our approach yields state-of-the-art robustness on MNIST against L0, L2 and L-infinity perturbations and we demonstrate that most adversarial examples are strongly perturbed towards the perceptual boundary between the original and the adversarial class."
                },
                {
                    "title": "Revisiting Unreasonable Effectiveness of Data in Deep Learning Era",
                    "abstract": "The success of deep learning in vision can be attributed to: (a) models with high capacity; (b) increased computational power; and (c) availability of large-scale labeled data. Since 2012, there have been significant advances in representation capabilities of the models and computational capabilities of GPUs. But the size of the biggest dataset has surprisingly remained constant. What will happen if we increase the dataset size by 10 \u00d7 or 100 \u00d7 ? This paper takes a step towards clearing the clouds of mystery surrounding the relationship between \u2018enormous data\u2019 and visual deep learning. By exploiting the JFT-300M dataset which has more than 375M noisy labels for 300M images, we investigate how the performance of current vision tasks would change if this data was used for representation learning. Our paper delivers some surprising (and some expected) findings. First, we find that the performance on vision tasks increases logarithmically based on volume of training data size. Second, we show that representation learning (or pre-training) still holds a lot of promise. One can improve performance on many vision tasks by just training a better base model. Finally, as expected, we present new state-of-the-art results for different vision tasks including image classification, object detection, semantic segmentation and human pose estimation. Our sincere hope is that this inspires vision community to not undervalue the data and develop collective efforts in building larger datasets."
                },
                {
                    "title": "Deep neural networks: a new framework for modelling biological vision and brain information processing",
                    "abstract": "Recent advances in neural network modelling have enabled major strides in computer vision and other artificial intelligence applications. Human-level visual recognition abilities are coming within reach of artificial systems. Artificial neural networks are inspired by the brain and their computations could be implemented in biological neurons. Convolutional feedforward networks, which now dominate computer vision, take further inspiration from the architecture of the primate visual hierarchy. However, the current models are designed with engineering goals and not to model brain computations. Nevertheless, initial studies comparing internal representations between these models and primate brains find surprisingly similar representational spaces. With human-level performance no longer out of reach, we are entering an exciting new era, in which we will be able to build neurobiologically faithful feedforward and recurrent computational models of how biological brains perform high-level feats of intelligence, including vision."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification",
                    "abstract": "Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on the learnable activation and advanced initialization, we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66% [33]). To our knowledge, our result is the first to surpass the reported human-level performance (5.1%, [26]) on this dataset."
                },
                {
                    "title": "Auto-Encoding Variational Bayes",
                    "abstract": "Abstract: How can we perform efficient inference and learning in directed probabilistic models, in the presence of continuous latent variables with intractable posterior distributions, and large datasets? We introduce a stochastic variational inference and learning algorithm that scales to large datasets and, under some mild differentiability conditions, even works in the intractable case. Our contributions is two-fold. First, we show that a reparameterization of the variational lower bound yields a lower bound estimator that can be straightforwardly optimized using standard stochastic gradient methods. Second, we show that for i.i.d. datasets with continuous latent variables per datapoint, posterior inference can be made especially efficient by fitting an approximate inference model (also called a recognition model) to the intractable posterior using the proposed lower bound estimator. Theoretical advantages are reflected in experimental results."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "Analysis by Synthesis: A (Re-)Emerging Program of Research for Language and Vision",
                    "abstract": "This contribution reviews (some of) the history of analysis by synthesis, an approach to perception and comprehension articulated in the 1950s. Whereas much research has focused on bottom-up, feed-forward, inductive mechanisms, analysis by synthesis as a heuristic model emphasizes a balance of bottom-up and knowledge-driven, top-down, predictive steps in speech perception and language comprehension. This idea aligns well with contemporary Bayesian approaches to perception (in language and other domains), which are illustrated with examples from different aspects of perception and comprehension. Results from psycholinguistics, the cognitive neuroscience of language, and visual object recognition suggest that analysis by synthesis can provide a productive way of structuring biolinguistic research. Current evidence suggests that such a model is theoretically well motivated, biologically sensible, and becomes computationally tractable borrowing from Bayesian formalizations."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "On Discriminative vs. Generative Classifiers: A comparison of logistic regression and naive Bayes",
                    "abstract": "We compare discriminative and generative learning as typified by logistic regression and naive Bayes. We show, contrary to a widely-held belief that discriminative classifiers are almost always to be preferred, that there can often be two distinct regimes of performance as the training set size is increased, one in which each algorithm does better. This stems from the observation\u2014which is borne out in repeated experiments\u2014that while discriminative learning has lower asymptotic error, a generative classifier may also approach its (higher) asymptotic error much faster."
                },
                {
                    "title": "Using Generative Models for Handwritten Digit Recognition",
                    "abstract": "We describe a method of recognizing handwritten digits by fitting generative models that are built from deformable B-splines with Gaussian \"ink generators\" spaced along the length of the spline. The splines are adjusted using a novel elastic matching procedure based on the expectation maximization algorithm that maximizes the likelihood of the model generating the data. This approach has many advantages: 1) the system not only produces a classification of the digit but also a rich description of the instantiation parameters which can yield information such as the writing style; 2) the generative models can perform recognition driven segmentation; 3) the method involves a relatively small number of parameters and hence training is relatively easy and fast; and 4) unlike many other recognition schemes, it does not rely on some form of pre-normalization of input images, but can handle arbitrary scalings, translations and a limited degree of image rotation. We have demonstrated that our method of fitting models to images does not get trapped in poor local minima. The main disadvantage of the method is that it requires much more computation than more standard OCR techniques."
                },
                {
                    "title": "The Helmholtz Machine",
                    "abstract": "Discovering the structure inherent in a set of patterns is a fundamental aim of statistical inference or learning. One fruitful approach is to build a parameterized stochastic generative model, independent draws from which are likely to produce the patterns. For all but the simplest generative models, each pattern can be generated in exponentially many ways. It is thus intractable to adjust the parameters to maximize the probability of the observed patterns. We describe a way of finessing this combinatorial explosion by maximizing an easily computed lower bound on the probability of the observations. Our method can be viewed as a form of hierarchical self-supervised learning that may relate to the function of bottom-up and top-down cortical processing pathways."
                },
                {
                    "title": "A Coefficient of Agreement for Nominal Scales",
                    "abstract": "CONSIDER Table 1. It represents in its formal characteristics a situation which arises in the clinical-social-personality areas of psychology, where it frequently occurs that the only useful level of measurement obtainable is nominal scaling (Stevens, 1951, pp. 2526), i.e. placement in a set of k unordered categories. Because the categorizing of the units is a consequence of some complex judgment process performed by a &dquo;two-legged meter&dquo; (Stevens, 1958), it becomes important to determine the extent to which these judgments are reproducible, i.e., reliable. The procedure which suggests itself is that of having two (or more) judges independently categorize a sample of units and determine the degree, significance, and"
                },
                {
                    "title": "How does the brain combine generative models and direct discriminative computations in high-level vision?",
                    "abstract": "Figure 1: Two contrasting visions of vision. Separate traditions rooted in an empiricist and a rationalist conception, respectively, have envisioned the computations underlying visual perception as either a largely bottom-up, feedforward process of extracting behaviorally relevant information or as an inference process that interrogates the sensory evidence in light of a generative model that captures prior knowledge about the processes in the world that give rise to the sensory data. Several theoretically independent dimensions (double arrows) are conflated when considering the two perspectives as a dichotomy. The goal of this GAC is to disentangle the dimensions and clarify how they relate to each other, to understand how the favoured algorithm depends on the visual task, and to develop experiments that will help us understand how the primate brain combines elements of both conceptions. Our question is how the primate brain combines generative models and direct discriminative computations in high-level vision. Both approaches aim at inferring behaviorally relevant latent variables y from visual data x. In a probabilistic setting, the inference of the posterior p(y|x) is known as discriminative inference. The two approaches differ in how discriminative inference is implemented. In the generative approach, a model of the joint distribution p(y,x) of the latent variables and the visual input is employed. This model captures information about the processes in the world that give rise to the sensory data. Approximate inference algorithms are then used to infer the posterior over the latents given an image by estimating p(y|x) = p(y,x)/p(x). In the direct discriminative approach, a direct mapping from the sensory data to the posterior over the latents p(y|x) is learned without the use of an explicit generative model. The generative approach enables unsupervised learning of the structure of the world and promises better generalization to novel situations (statistical efficiency). Direct discriminative computations promise faster inferences (computational efficiency) that are accurate for new samples from the distribution experienced in training. In practice, inference of the full posterior may not be realistic and the visual system may settle for point estimates in certain cases."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI",
                "cs.LG",
                "q-bio.NC",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the computational efficiency of generative models in image generation while maintaining or enhancing their generalization capabilities?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the trade-off between computational speed and model performance in generative models. Enhancing the efficiency of these models could lead to broader adoption in practical applications, such as real-time image generation in various industries, including gaming, virtual reality, and design. Furthermore, it could inspire future research to explore new architectures or training procedures that prioritize both speed and accuracy, ultimately advancing our understanding of generative modeling.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in improving the computational efficiency of generative models stem from their inherent complexity and the need for high-quality outputs. Naive approaches may fail because they often overlook the intricate balance between model architecture, training data, and the generative process itself. Technical obstacles include the need for extensive computational resources and the difficulty in optimizing models that can generate high-fidelity images quickly. Theoretical challenges arise from understanding how different generative techniques (e.g., pixel-based diffusion vs. latent space diffusion) impact performance and efficiency.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on improving the quality of generated images without adequately addressing the computational costs associated with generative models. Limitations in existing solutions include a lack of comprehensive comparisons across diverse model architectures and training procedures, which has hindered the identification of optimal strategies for efficiency. Our approach differs by systematically investigating a variety of generative classifiers, allowing us to isolate factors that contribute to both speed and performance, thus providing a clearer path toward practical improvements.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a comparative analysis of various generative models, including pixel-based diffusion, latent space diffusion, and autoregressive models. We will utilize a diverse dataset of images to evaluate the models' performance based on metrics such as generation speed and image quality. The expected outcomes include identifying specific architectural or procedural modifications that significantly enhance computational efficiency without compromising the quality of generated images, ultimately leading to more practical applications of generative models in real-world scenarios."
            }
        },
        "author_data": {
            "34726398-b532-4381-8a02-b931ab8c7ce5": {
                "pk": "34726398-b532-4381-8a02-b931ab8c7ce5",
                "name": "Priyank Jaini",
                "collaborators": [
                    "M. Welling",
                    "Emiel Hoogeboom",
                    "Didrik Nielsen",
                    "Lars Holdijk",
                    "Yuanqi Du",
                    "F. Hooft",
                    "B. Ensing",
                    "Patrick Forr'e",
                    "Allen Wang",
                    "Yaoliang Yu"
                ],
                "domain": [
                    "Generative Models",
                    "Probabilistic Inference",
                    "Deep Learning",
                    "Molecular Dynamics"
                ],
                "publications": [
                    {
                        "title": "Towards flexible perception with visual memory",
                        "abstract": "Training a neural network is a monolithic endeavor, akin to carving knowledge into stone: once the process is completed, editing the knowledge in a network is nearly impossible, since all information is distributed across the network's weights. We here explore a simple, compelling alternative by marrying the representational power of deep neural networks with the flexibility of a database. Decomposing the task of image classification into image similarity (from a pre-trained embedding) and search (via fast nearest neighbor retrieval from a knowledge database), we build a simple and flexible visual memory that has the following key capabilities: (1.) The ability to flexibly add data across scales: from individual samples all the way to entire classes and billion-scale data; (2.) The ability to remove data through unlearning and memory pruning; (3.) An interpretable decision-mechanism on which we can intervene to control its behavior. Taken together, these capabilities comprehensively demonstrate the benefits of an explicit visual memory. We hope that it might contribute to a conversation on how knowledge should be represented in deep vision models -- beyond carving it in\"stone\"weights."
                    },
                    {
                        "title": "Decoupling Semantic Similarity from Spatial Alignment for Neural Networks",
                        "abstract": "What representation do deep neural networks learn? How similar are images to each other for neural networks? Despite the overwhelming success of deep learning methods key questions about their internal workings still remain largely unanswered, due to their internal high dimensionality and complexity. To address this, one approach is to measure the similarity of activation responses to various inputs. Representational Similarity Matrices (RSMs) distill this similarity into scalar values for each input pair. These matrices encapsulate the entire similarity structure of a system, indicating which input leads to similar responses. While the similarity between images is ambiguous, we argue that the spatial location of semantic objects does neither influence human perception nor deep learning classifiers. Thus this should be reflected in the definition of similarity between image responses for computer vision systems. Revisiting the established similarity calculations for RSMs we expose their sensitivity to spatial alignment. In this paper, we propose to solve this through semantic RSMs, which are invariant to spatial permutation. We measure semantic similarity between input responses by formulating it as a set-matching problem. Further, we quantify the superiority of semantic RSMs over spatio-semantic RSMs through image retrieval and by comparing the similarity between representations to the similarity between predicted class probabilities."
                    },
                    {
                        "title": "Text-to-Image Diffusion Models are Zero-Shot Classifiers",
                        "abstract": "The excellent generative capabilities of text-to-image diffusion models suggest they learn informative representations of image-text data. However, what knowledge their representations capture is not fully understood, and they have not been thoroughly explored on downstream tasks. We investigate diffusion models by proposing a method for evaluating them as zero-shot classifiers. The key idea is using a diffusion model's ability to denoise a noised image given a text description of a label as a proxy for that label's likelihood. We apply our method to Stable Diffusion and Imagen, using it to probe fine-grained aspects of the models' knowledge and comparing them with CLIP's zero-shot abilities. They perform competitively with CLIP on a wide range of zero-shot image classification datasets. Additionally, they achieve state-of-the-art results on shape/texture bias tests and can successfully perform attribute binding while CLIP cannot. Although generative pre-training is prevalent in NLP, visual foundation models often use other methods such as contrastive learning. Based on our findings, we argue that generative pre-training should be explored as a compelling alternative for vision-language tasks."
                    },
                    {
                        "title": "Path Integral Stochastic Optimal Control for Sampling Transition Paths",
                        "abstract": "We consider the problem of Sampling Transition Paths . Given two metastable conformational states of a molecular system, e.g. a folded and unfolded protein, we aim to sample the most likely transition path between the two states. Sampling such a transition path is computationally expensive due to the existence of high free energy barriers between the two states. To circumvent this, previous work has focused on simplifying the trajectories to occur along specific molecular descriptors called Collective Variables (CVs). However, finding CVs is not trivial and requires chemical intuition. For larger molecules, where intuition is not sufficient, using these CV-based methods biases the transition along possibly irrelevant dimensions. Instead, this work proposes a method for sampling transition paths that consider the entire geometry of the molecules. To achieve this, we first relate the problem to recent work on the Schr \u00a8 odinger bridge problem and stochastic optimal control. Using this relation, we construct a method that takes into account important characteristics of molecular systems such as second-order dynamics and invariance to rotations and translations. We demonstrate our method on the commonly studied Alanine Dipeptide, but also consider larger proteins such as Polyproline and Chignolin."
                    },
                    {
                        "title": "Recent Advancements in Tractable Probabilistic Inference (Dagstuhl Seminar 22161)",
                        "abstract": "In several real-world scenarios, decision making involves advanced reasoning under uncertainty, i.e. the ability to answer probabilistic queries. Typically, it is necessary to compute these answers in a limited amount of time . Moreover, in many domains, such as healthcare and economical decision making, it is crucial that the result of these queries is reliable, i.e. either exact or comes with approximation guarantees. In all these scenarios, tractable probabilistic inference and learning are becoming increasingly important. Research on representations and learning algorithms for tractable inference embraces very different fields, each one contributing its own perspective. These include automated reasoning, probabilistic modeling, statistical and Bayesian inference and deep learning. Among the many recent emerging venues in these fields there are: tractable neural density estimators such as autoregressive models and normalizing flows; deep tractable probabilistic circuits such as sum-product networks and sentential decision diagrams; approximate inference routines with guarantees on the quality of the approximation. Each of these model classes occupies a particular spot in the continuum between tractability and expressiveness. That is, different model classes might offer appealing advantages in terms of efficiency or representation capabilities while trading-off other of these aspects. So far, clear connections and a deeper understanding of the key differences among them have been hindered by the different languages and perspectives adopted by the different \u201csouls\u201d that comprise the tractable probabilistic modeling community. This Dagstuhl Seminar brought together experts from these sub-communities and provided the perfect venue to exchange perspectives, deeply discuss the recent advancements and build strong bridges that can greatly propel interdisciplinary research."
                    },
                    {
                        "title": "Stochastic Optimal Control for Collective Variable Free Sampling of Molecular Transition Paths",
                        "abstract": "We consider the problem of sampling transition paths between two given metastable states of a molecular system, e.g. a folded and unfolded protein or products and reactants of a chemical reaction. Due to the existence of high energy barriers separating the states, these transition paths are unlikely to be sampled with standard Molecular Dynamics (MD) simulation. Traditional methods to augment MD with a bias potential to increase the probability of the transition rely on a dimensionality reduction step based on Collective Variables (CVs). Unfortunately, selecting appropriate CVs requires chemical intuition and traditional methods are therefore not always applicable to larger systems. Additionally, when incorrect CVs are used, the bias potential might not be minimal and bias the system along dimensions irrelevant to the transition. Showing a formal relation between the problem of sampling molecular transition paths, the Schr\\\"odinger bridge problem and stochastic optimal control with neural network policies, we propose a machine learning method for sampling said transitions. Unlike previous non-machine learning approaches our method, named PIPS, does not depend on CVs. We show that our method successful generates low energy transitions for Alanine Dipeptide as well as the larger Polyproline and Chignolin proteins."
                    },
                    {
                        "title": "Argmax Flows and Multinomial Diffusion: Towards Non-Autoregressive Language Models",
                        "abstract": "The \ufb01eld of language modelling has been largely dominated by autoregressive models, for which sampling is inherently dif\ufb01cult to parallelize. This paper introduces two new classes of generative models for categorical data such as language or image segmentation: Argmax Flows and Multinomial Diffusion . Argmax Flows are de\ufb01ned by a composition of a continuous distribution (such as a normalizing \ufb02ow), and an argmax function. To optimize this model, we learn a probabilistic inverse for the argmax that lifts the categorical data to a continuous space. Multinomial Diffusion gradually adds categorical noise in a diffusion process, for which the generative denoising process is learned. We demonstrate that our models perform competitively on language modelling and modelling of image segmentation maps."
                    },
                    {
                        "title": "Argmax Flows: Learning Categorical Distributions with Normalizing Flows",
                        "abstract": "This paper introduces a new method to de\ufb01ne and train continuous distributions such as normalizing \ufb02ows directly on categorical data, for example text and image segmentation. The generative model is de\ufb01ned by a composition of a normalizing \ufb02ow and an argmax function. To optimize this model, we dequantize the argmax using a distribution that is a probabilistic right-inverse to the argmax. This distribution lifts the categorical data to a continuous space on which the \ufb02ow can be trained. We demonstrate that applying existing dequantization techniques na\u00a8\u0131vely to categorical data leads to suboptimal solutions. In addition, the model is fast both in generative (for sampling) and inference direction (for training), as opposed to autoregressive models."
                    },
                    {
                        "title": "Sampling in Combinatorial Spaces with SurVAE Flow Augmented MCMC",
                        "abstract": "Hybrid Monte Carlo is a powerful Markov Chain Monte Carlo method for sampling from complex continuous distributions. However, a major limitation of HMC is its inability to be applied to discrete domains due to the lack of gradient signal. In this work, we introduce a new approach based on augmenting Monte Carlo methods with SurVAE Flows to sample from discrete distributions using a combination of neural transport methods like normalizing flows and variational dequantization, and the Metropolis-Hastings rule. Our method first learns a continuous embedding of the discrete space using a surjective map and subsequently learns a bijective transformation from the continuous space to an approximately Gaussian distributed latent variable. Sampling proceeds by simulating MCMC chains in the latent space and mapping these samples to the target discrete space via the learned transformations. We demonstrate the efficacy of our algorithm on a range of examples from statistics, computational physics and machine learning, and observe improvements compared to alternative algorithms."
                    },
                    {
                        "title": "Particle Dynamics for Learning EBMs",
                        "abstract": "Energy-based modeling is a promising approach to unsupervised learning, which yields many downstream applications from a single model. The main difficulty in learning energy-based models with the\"contrastive approaches\"is the generation of samples from the current energy function at each iteration. Many advances have been made to accomplish this subroutine cheaply. Nevertheless, all such sampling paradigms run MCMC targeting the current model, which requires infinitely long chains to generate samples from the true energy distribution and is problematic in practice. This paper proposes an alternative approach to getting these samples and avoiding crude MCMC sampling from the current model. We accomplish this by viewing the evolution of the modeling distribution as (i) the evolution of the energy function, and (ii) the evolution of the samples from this distribution along some vector field. We subsequently derive this time-dependent vector field such that the particles following this field are approximately distributed as the current density model. Thereby we match the evolution of the particles with the evolution of the energy function prescribed by the learning procedure. Importantly, unlike Monte Carlo sampling, our method targets to match the current distribution in a finite time. Finally, we demonstrate its effectiveness empirically compared to MCMC-based learning methods."
                    },
                    {
                        "title": "Argmax Flows and Multinomial Diffusion: Learning Categorical Distributions",
                        "abstract": "Generative flows and diffusion models have been predominantly trained on ordinal data, for example natural images. This paper introduces two extensions of flows and diffusion for categorical data such as language or image segmentation: Argmax Flows and Multinomial Diffusion. Argmax Flows are defined by a composition of a continuous distribution (such as a normalizing flow), and an argmax function. To optimize this model, we learn a probabilistic inverse for the argmax that lifts the categorical data to a continuous space. Multinomial Diffusion gradually adds categorical noise in a diffusion process, for which the generative denoising process is learned. We demonstrate that our method outperforms existing dequantization approaches on text modelling and modelling on image segmentation maps in log-likelihood."
                    },
                    {
                        "title": "Learning Equivariant Energy Based Models with Equivariant Stein Variational Gradient Descent",
                        "abstract": "We focus on the problem of efficient sampling and learning of probability densities by incorporating symmetries in probabilistic models. We first introduce Equivariant Stein Variational Gradient Descent algorithm -- an equivariant sampling method based on Stein's identity for sampling from densities with symmetries. Equivariant SVGD explicitly incorporates symmetry information in a density through equivariant kernels which makes the resultant sampler efficient both in terms of sample complexity and the quality of generated samples. Subsequently, we define equivariant energy based models to model invariant densities that are learned using contrastive divergence. By utilizing our equivariant SVGD for training equivariant EBMs, we propose new ways of improving and scaling up training of energy based models. We apply these equivariant energy models for modelling joint densities in regression and classification tasks for image datasets, many-body particle systems and molecular structure generation."
                    },
                    {
                        "title": "A Positivstellensatz for Conditional SAGE Signomials",
                        "abstract": "Recently, the conditional SAGE certificate has been proposed as a sufficient condition for signomial positivity over a convex set. In this article, we show that the conditional SAGE certificate is $\\textit{complete}$. That is, for any signomial $f(\\mathbf{x}) = \\sum_{j=1}^{\\ell}c_j \\exp(\\mathbf{A}_j\\mathbf{x})$ defined by rational exponents that is positive over a compact convex set $\\mathcal{X}$, there is $p \\in \\mathbb{Z}_+$ and a specific positive definite function $w(\\mathbf{x})$ such that $w(\\mathbf{x})^p f(\\mathbf{x})$ may be verified by the conditional SAGE certificate. The completeness result is analogous to Positivstellensatz results from algebraic geometry, which guarantees representation of positive polynomials with sum of squares polynomials. The result gives rise to a convergent hierarchy of lower bounds for constrained signomial optimization over an $\\textit{arbitrary}$ compact convex set that is computable via the conditional SAGE certificate."
                    },
                    {
                        "title": "Complete Hierarchy of Relaxation for Constrained Signomial Positivity",
                        "abstract": "In this article, we prove that the Sums-of-AM/GM Exponential (SAGE) relaxation generalized to signomial over a constrained set is complete, with a compactness assumption. The high-level structure of the proof is as follows. We first apply variable change to convert a set of rational exponents to polynomial equations. In addition, we make the observation that linear constraints of the variables may also be converted to polynomial equations after variable change. Note that any convex set may be expressed as a set of linear constraints. Further, we use redundant constraints to find reduction to Positivstellensatz. We rely on Positivstellensatz results from algebraic geometry to obtain a decomposition of positive polynomials. Lastly, we explicitly show that the decomposition is of a form certifiable by SAGE."
                    },
                    {
                        "title": "Self Normalizing Flows",
                        "abstract": "Efficient gradient computation of the Jacobian determinant term is a core problem of the normalizing flow framework. Thus, most proposed flow models either restrict to a function class with easy evaluation of the Jacobian determinant, or an efficient estimator thereof. However, these restrictions limit the performance of such density models, frequently requiring significant depth to reach desired performance levels. In this work, we propose Self Normalizing Flows, a flexible framework for training normalizing flows by replacing expensive terms in the gradient by learned approximate inverses at each layer. This reduces the computational complexity of each layer's exact update from $\\mathcal{O}(D^3)$ to $\\mathcal{O}(D^2)$, allowing for the training of flow architectures which were otherwise computationally infeasible, while also providing efficient sampling. We show experimentally that such models are remarkably stable and optimize to similar data likelihood values as their exact gradient counterparts, while surpassing the performance of their functionally constrained counterparts."
                    },
                    {
                        "title": "SurVAE Flows: Surjections to Bridge the Gap between VAEs and Flows",
                        "abstract": "Normalizing flows and variational autoencoders are powerful generative models that can represent complicated density functions. However, they both impose constraints on the models: Normalizing flows use bijective transformations to model densities whereas VAEs learn stochastic transformations that are non-invertible and thus typically do not provide tractable estimates of the marginal likelihood. In this paper, we introduce SurVAE Flows: A modular framework of composable transformations that encompasses VAEs and normalizing flows. SurVAE Flows bridge the gap between normalizing flows and VAEs with surjective transformations, wherein the transformations are deterministic in one direction -- thereby allowing exact likelihood computation, and stochastic in the reverse direction -- hence providing a lower bound on the corresponding likelihood. We show that several recently proposed methods, including dequantization and augmented normalizing flows, can be expressed as SurVAE Flows. Finally, we introduce common operations such as the max value, the absolute value, sorting and stochastic permutation as composable layers in SurVAE Flows."
                    }
                ]
            },
            "6b2fcdf8-b279-418c-8287-bd75251dcc66": {
                "pk": "6b2fcdf8-b279-418c-8287-bd75251dcc66",
                "name": "Kevin Clark",
                "collaborators": [
                    "Christopher D. Manning",
                    "Urvashi Khandelwal",
                    "Minh-Thang Luong",
                    "Quoc V. Le",
                    "Omer Levy",
                    "Dan Jurafsky",
                    "J. Leskovec",
                    "P. Jaini",
                    "K. Singhal",
                    "Tao Tu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Representation Learning",
                    "Multimodal Processing"
                ],
                "publications": [
                    {
                        "title": "Text-to-Image Diffusion Models are Zero-Shot Classifiers",
                        "abstract": "The excellent generative capabilities of text-to-image diffusion models suggest they learn informative representations of image-text data. However, what knowledge their representations capture is not fully understood, and they have not been thoroughly explored on downstream tasks. We investigate diffusion models by proposing a method for evaluating them as zero-shot classifiers. The key idea is using a diffusion model's ability to denoise a noised image given a text description of a label as a proxy for that label's likelihood. We apply our method to Stable Diffusion and Imagen, using it to probe fine-grained aspects of the models' knowledge and comparing them with CLIP's zero-shot abilities. They perform competitively with CLIP on a wide range of zero-shot image classification datasets. Additionally, they achieve state-of-the-art results on shape/texture bias tests and can successfully perform attribute binding while CLIP cannot. Although generative pre-training is prevalent in NLP, visual foundation models often use other methods such as contrastive learning. Based on our findings, we argue that generative pre-training should be explored as a compelling alternative for vision-language tasks."
                    },
                    {
                        "title": "Towards Expert-Level Medical Question Answering with Large Language Models",
                        "abstract": "Recent artificial intelligence (AI) systems have reached milestones in\"grand challenges\"ranging from Go to protein-folding. The capability to retrieve medical knowledge, reason over it, and answer medical questions comparably to physicians has long been viewed as one such grand challenge. Large language models (LLMs) have catalyzed significant progress in medical question answering; Med-PaLM was the first model to exceed a\"passing\"score in US Medical Licensing Examination (USMLE) style questions with a score of 67.2% on the MedQA dataset. However, this and other prior work suggested significant room for improvement, especially when models' answers were compared to clinicians' answers. Here we present Med-PaLM 2, which bridges these gaps by leveraging a combination of base LLM improvements (PaLM 2), medical domain finetuning, and prompting strategies including a novel ensemble refinement approach. Med-PaLM 2 scored up to 86.5% on the MedQA dataset, improving upon Med-PaLM by over 19% and setting a new state-of-the-art. We also observed performance approaching or exceeding state-of-the-art across MedMCQA, PubMedQA, and MMLU clinical topics datasets. We performed detailed human evaluations on long-form questions along multiple axes relevant to clinical applications. In pairwise comparative ranking of 1066 consumer medical questions, physicians preferred Med-PaLM 2 answers to those produced by physicians on eight of nine axes pertaining to clinical utility (p<0.001). We also observed significant improvements compared to Med-PaLM on every evaluation axis (p<0.001) on newly introduced datasets of 240 long-form\"adversarial\"questions to probe LLM limitations. While further studies are necessary to validate the efficacy of these models in real-world settings, these results highlight rapid progress towards physician-level performance in medical question answering."
                    },
                    {
                        "title": "Meta-Learning Fast Weight Language Models",
                        "abstract": "Dynamic evaluation of language models (LMs) adapts model parameters at test time using gradient information from previous tokens and substantially improves LM performance. However, it requires over 3x more compute than standard inference. We present Fast Weight Layers (FWLs), a neural component that provides the benefits of dynamic evaluation much more efficiently by expressing gradient updates as linear attention. A key improvement over dynamic evaluation is that FWLs can also be applied at training time, so the model learns to make good use of gradient updates. FWLs can easily be added on top of existing transformer models, require relatively little extra compute or memory to run, and significantly improve language modeling perplexity."
                    },
                    {
                        "title": "Pre-Training Transformers as Energy-Based Cloze Models",
                        "abstract": "We introduce Electric, an energy-based cloze model for representation learning over text. Like BERT, it is a conditional generative model of tokens given their contexts. However, Electric does not use masking or output a full distribution over tokens that could occur in a context. Instead, it assigns a scalar energy score to each input token indicating how likely it is given its context. We train Electric using an algorithm based on noise-contrastive estimation and elucidate how this learning objective is closely related to the recently proposed ELECTRA pre-training method. Electric performs well when transferred to downstream tasks and is particularly effective at producing likelihood scores for text: it re-ranks speech recognition n-best lists better than language models and much faster than masked language models. Furthermore, it offers a clearer and more principled view of what ELECTRA learns during pre-training."
                    },
                    {
                        "title": "Emergent linguistic structure in artificial neural networks trained by self-supervision",
                        "abstract": "This paper explores the knowledge of linguistic structure learned by large artificial neural networks, trained via self-supervision, whereby the model simply tries to predict a masked word in a given context. Human language communication is via sequences of words, but language understanding requires constructing rich hierarchical structures that are never observed explicitly. The mechanisms for this have been a prime mystery of human language acquisition, while engineering work has mainly proceeded by supervised learning on treebanks of sentences hand labeled for this latent structure. However, we demonstrate that modern deep contextual language models learn major aspects of this structure, without any explicit supervision. We develop methods for identifying linguistic hierarchical structure emergent in artificial neural networks and demonstrate that components in these models focus on syntactic grammatical relationships and anaphoric coreference. Indeed, we show that a linear transformation of learned embeddings in these models captures parse tree distances to a surprising degree, allowing approximate reconstruction of the sentence tree structures normally assumed by linguists. These results help explain why these models have brought such large improvements across many language-understanding tasks."
                    },
                    {
                        "title": "Internship: probing joint vision-and-language representations",
                        "abstract": "Context Recent advances in deep learning have enabled exciting applications in the context of multimodal processing involving images and texts, such as visual question answering [1], visual dialog [4], image captionning [14], text undersanding in multimodal context [5]... This internship is focused on exploring representations trained to perform such tasks"
                    },
                    {
                        "title": "What Does BERT Look at? An Analysis of BERT\u2019s Attention",
                        "abstract": "Large pre-trained neural networks such as BERT have had great recent success in NLP, motivating a growing body of research investigating what aspects of language they are able to learn from unlabeled data. Most recent analysis has focused on model outputs (e.g., language model surprisal) or internal vector representations (e.g., probing classifiers). Complementary to these works, we propose methods for analyzing the attention mechanisms of pre-trained models and apply them to BERT. BERT\u2019s attention heads exhibit patterns such as attending to delimiter tokens, specific positional offsets, or broadly attending over the whole sentence, with heads in the same layer often exhibiting similar behaviors. We further show that certain attention heads correspond well to linguistic notions of syntax and coreference. For example, we find heads that attend to the direct objects of verbs, determiners of nouns, objects of prepositions, and coreferent mentions with remarkably high accuracy. Lastly, we propose an attention-based probing classifier and use it to further demonstrate that substantial syntactic information is captured in BERT\u2019s attention."
                    },
                    {
                        "title": "Sample Efficient Text Summarization Using a Single Pre-Trained Transformer",
                        "abstract": "Language model (LM) pre-training has resulted in impressive performance and sample efficiency on a variety of language understanding tasks. However, it remains unclear how to best use pre-trained LMs for generation tasks such as abstractive summarization, particularly to enhance sample efficiency. In these sequence-to-sequence settings, prior work has experimented with loading pre-trained weights into the encoder and/or decoder networks, but used non-pre-trained encoder-decoder attention weights. We instead use a pre-trained decoder-only network, where the same Transformer LM both encodes the source and generates the summary. This ensures that all parameters in the network, including those governing attention over source states, have been pre-trained before the fine-tuning step. Experiments on the CNN/Daily Mail dataset show that our pre-trained Transformer LM substantially improves over pre-trained Transformer encoder-decoder networks in limited-data settings. For instance, it achieves 13.1 ROUGE-2 using only 1% of the training data (~3000 examples), while pre-trained encoder-decoder models score 2.3 ROUGE-2."
                    },
                    {
                        "title": "BAM! Born-Again Multi-Task Networks for Natural Language Understanding",
                        "abstract": "It can be challenging to train multi-task neural networks that outperform or even match their single-task counterparts. To help address this, we propose using knowledge distillation where single-task models teach a multi-task model. We enhance this training with teacher annealing, a novel method that gradually transitions the model from distillation to supervised learning, helping the multi-task model surpass its single-task teachers. We evaluate our approach by multi-task fine-tuning BERT on the GLUE benchmark. Our method consistently improves over standard single-task and multi-task training."
                    },
                    {
                        "title": "Semi-Supervised Sequence Modeling with Cross-View Training",
                        "abstract": "Unsupervised representation learning algorithms such as word2vec and ELMo improve the accuracy of many supervised NLP models, mainly because they can take advantage of large amounts of unlabeled text. However, the supervised models only learn from task-specific labeled data during the main training phase. We therefore propose Cross-View Training (CVT), a semi-supervised learning algorithm that improves the representations of a Bi-LSTM sentence encoder using a mix of labeled and unlabeled data. On labeled examples, standard supervised learning is used. On unlabeled examples, CVT teaches auxiliary prediction modules that see restricted views of the input (e.g., only part of a sentence) to match the predictions of the full model seeing the whole input. Since the auxiliary modules and the full model share intermediate representations, this in turn improves the full model. Moreover, we show that CVT is particularly effective when combined with multi-task learning. We evaluate CVT on five sequence tagging tasks, machine translation, and dependency parsing, achieving state-of-the-art results."
                    },
                    {
                        "title": "Deep Reinforcement Learning for Mention-Ranking Coreference Models",
                        "abstract": "Coreference resolution systems are typically trained with heuristic loss functions that require careful tuning. In this paper we instead apply reinforcement learning to directly optimize a neural mention-ranking model for coreference evaluation metrics. We experiment with two approaches: the REINFORCE policy gradient algorithm and a reward-rescaled max-margin objective. We find the latter to be more effective, resulting in significant improvements over the current state-of-the-art on the English and Chinese portions of the CoNLL 2012 Shared Task."
                    },
                    {
                        "title": "Large-scale Analysis of Counseling Conversations: An Application of Natural Language Processing to Mental Health",
                        "abstract": "Mental illness is one of the most pressing public health issues of our time. While counseling and psychotherapy can be effective treatments, our knowledge about how to conduct successful counseling conversations has been limited due to lack of large-scale data with labeled outcomes of the conversations. In this paper, we present a large-scale, quantitative study on the discourse of text-message-based counseling conversations. We develop a set of novel computational discourse analysis methods to measure how various linguistic aspects of conversations are correlated with conversation outcomes. Applying techniques such as sequence-based conversation models, language model comparisons, message clustering, and psycholinguistics-inspired word frequency analyses, we discover actionable conversation strategies that are associated with better conversation outcomes."
                    },
                    {
                        "title": "Inducing Domain-Specific Sentiment Lexicons from Unlabeled Corpora",
                        "abstract": "A word's sentiment depends on the domain in which it is used. Computational social science research thus requires sentiment lexicons that are specific to the domains being studied. We combine domain-specific word embeddings with a label propagation framework to induce accurate domain-specific sentiment lexicons using small sets of seed words. We show that our approach achieves state-of-the-art performance on inducing sentiment lexicons from domain-specific corpora and that our purely corpus-based approach outperforms methods that rely on hand-curated resources (e.g., WordNet). Using our framework, we induce and release historical sentiment lexicons for 150 years of English and community-specific sentiment lexicons for 250 online communities from the social media forum Reddit. The historical lexicons we induce show that more than 5% of sentiment-bearing (non-neutral) English words completely switched polarity during the last 150 years, and the community-specific lexicons highlight how sentiment varies drastically between different communities."
                    },
                    {
                        "title": "Improving Coreference Resolution by Learning Entity-Level Distributed Representations",
                        "abstract": "A long-standing challenge in coreference resolution has been the incorporation of entity-level information - features defined over clusters of mentions instead of mention pairs. We present a neural network based coreference system that produces high-dimensional vector representations for pairs of coreference clusters. Using these representations, our system learns when combining clusters is desirable. We train the system with a learning-to-search algorithm that teaches it which local decisions (cluster merges) will lead to a high-scoring final coreference partition. The system substantially outperforms the current state-of-the-art on the English and Chinese portions of the CoNLL 2012 Shared Task dataset despite using few hand-engineered features."
                    },
                    {
                        "title": "Entity-Centric Coreference Resolution with Model Stacking",
                        "abstract": "Mention pair models that predict whether or not two mentions are coreferent have historically been very effective for coreference resolution, but do not make use of entity-level information. However, we show that the scores produced by such models can be aggregated to define powerful entity-level features between clusters of mentions. Using these features, we train an entity-centric coreference system that learns an effective policy for building up coreference chains incrementally. The mention pair scores are also used to prune the search space the system works in, allowing for efficient training with an exact loss function. We evaluate our system on the English portion of the 2012 CoNLL Shared Task dataset and show that it improves over the current state of the art."
                    },
                    {
                        "title": "Digital curation: Opportunities for learning, teaching, research and professional development",
                        "abstract": "The concept of digital curation has evolved from its original use. While the digital preservation of artifacts is still the primary domain of digital curation, due to Web 2.0/social media applications it has allowed anyone to easily create a topic centered library online to share with the world. Assets found online can now be quickly collated into a visually appealing web site using freely available tools. This descriptive paper explores the potential use of digital curation within three tertiary institutions."
                    },
                    {
                        "title": "RevMiner: an extractive interface for navigating reviews on a smartphone",
                        "abstract": "Smartphones are convenient, but their small screens make searching, clicking, and reading awkward. Thus, perusing product reviews on a smartphone is difficult. In response, we introduce RevMiner - a novel smartphone interface that utilizes Natural Language Processing techniques to analyze and navigate reviews. RevMiner was run over 300K Yelp restaurant reviews extracting attribute-value pairs, where attributes represent restaurant attributes such as sushi and service, and values represent opinions about the attributes such as fresh or fast. These pairs were aggregated and used to: 1) answer queries such as \"cheap Indian food\", 2) concisely present information about each restaurant, and 3) identify similar restaurants. Our user studies demonstrate that on a smartphone, participants preferred RevMiner's interface to tag clouds and color bars, and that they preferred RevMiner's results to Yelp's, particularly for conjunctive queries (e.g., \"great food and huge portions\"). Demonstrations of RevMiner are available at revminer.com."
                    },
                    {
                        "title": "Platter",
                        "abstract": "Platter So many deaths and conjoinings, so many cocktail parties on the slate patio, but they never dig it out. It's the design: heavy, clumsy at the grip, in the center an odd, clumped relief of flowers-vines grown out of an underworld, a tangled migraine. My widowed mother leans cat-still against the door frame, peering autistically into the kitchen-but the secret eludes her. She hasn't told us of her headaches. Animals roam the frontal lobes. I saw a movie with the venus fly trap as the guiding symbol. So thickly rooted. Giant knobs in the dirt. Portentous. The platter is like that, heaped like a king's last meal-only he can't accept the pronouncement floating from the physician's mouth, he's called for the jugglers. When my mother shudders from her vision and reaches into the cabinet for tea,"
                    }
                ]
            },
            "e560dde6-2ef7-4376-90a7-6fe84dd76505": {
                "pk": "e560dde6-2ef7-4376-90a7-6fe84dd76505",
                "name": "Robert Geirhos",
                "collaborators": [
                    "Wieland Brendel",
                    "Roland S. Zimmermann",
                    "J. Jacobsen",
                    "Richard Michaelis",
                    "Wieland Zemel",
                    "Brendel",
                    "Matthias Bethge",
                    "Felix A. Wichmann",
                    "Felix Wichmann",
                    "M. Bethge"
                ],
                "domain": [
                    "Computer Vision",
                    "Neural Networks",
                    "Machine Learning",
                    "Interpretability"
                ],
                "publications": [
                    {
                        "title": "Towards flexible perception with visual memory",
                        "abstract": "Training a neural network is a monolithic endeavor, akin to carving knowledge into stone: once the process is completed, editing the knowledge in a network is nearly impossible, since all information is distributed across the network's weights. We here explore a simple, compelling alternative by marrying the representational power of deep neural networks with the flexibility of a database. Decomposing the task of image classification into image similarity (from a pre-trained embedding) and search (via fast nearest neighbor retrieval from a knowledge database), we build a simple and flexible visual memory that has the following key capabilities: (1.) The ability to flexibly add data across scales: from individual samples all the way to entire classes and billion-scale data; (2.) The ability to remove data through unlearning and memory pruning; (3.) An interpretable decision-mechanism on which we can intervene to control its behavior. Taken together, these capabilities comprehensively demonstrate the benefits of an explicit visual memory. We hope that it might contribute to a conversation on how knowledge should be represented in deep vision models -- beyond carving it in\"stone\"weights."
                    },
                    {
                        "title": "Are Vision Language Models Texture or Shape Biased and Can We Steer Them?",
                        "abstract": "Vision language models (VLMs) have drastically changed the computer vision model landscape in only a few years, opening an exciting array of new applications from zero-shot image classification, over to image captioning, and visual question answering. Unlike pure vision models, they offer an intuitive way to access visual content through language prompting. The wide applicability of such models encourages us to ask whether they also align with human vision - specifically, how far they adopt human-induced visual biases through multimodal fusion, or whether they simply inherit biases from pure vision models. One important visual bias is the texture vs. shape bias, or the dominance of local over global information. In this paper, we study this bias in a wide range of popular VLMs. Interestingly, we find that VLMs are often more shape-biased than their vision encoders, indicating that visual biases are modulated to some extent through text in multimodal models. If text does indeed influence visual biases, this suggests that we may be able to steer visual biases not just through visual input but also through language: a hypothesis that we confirm through extensive experiments. For instance, we are able to steer shape bias from as low as 49% to as high as 72% through prompting alone. For now, the strong human bias towards shape (96%) remains out of reach for all tested VLMs."
                    },
                    {
                        "title": "Scaling Vision Transformers to 22 Billion Parameters",
                        "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                    },
                    {
                        "title": "Don't trust your eyes: on the (un)reliability of feature visualizations",
                        "abstract": "How do neural networks extract patterns from pixels? Feature visualizations attempt to answer this important question by visualizing highly activating patterns through optimization. Today, visualization methods form the foundation of our knowledge about the internal workings of neural networks, as a type of mechanistic interpretability. Here we ask: How reliable are feature visualizations? We start our investigation by developing network circuits that trick feature visualizations into showing arbitrary patterns that are completely disconnected from normal network behavior on natural input. We then provide evidence for a similar phenomenon occurring in standard, unmanipulated networks: feature visualizations are processed very differently from standard input, casting doubt on their ability to\"explain\"how neural networks process natural images. This can be used as a sanity check for feature visualizations. We underpin our empirical findings by theory proving that the set of functions that can be reliably understood by feature visualization is extremely small and does not include general black-box neural networks. Therefore, a promising way forward could be the development of networks that enforce certain structures in order to ensure more reliable feature visualizations."
                    },
                    {
                        "title": "Getting aligned on representational alignment",
                        "abstract": "Biological and artificial information processing systems form representations that they can use to categorize, reason, plan, navigate, and make decisions. How can we measure the extent to which the representations formed by these diverse systems agree? Do similarities in representations then translate into similar behavior? How can a system's representations be modified to better match those of another system? These questions pertaining to the study of representational alignment are at the heart of some of the most active research areas in cognitive science, neuroscience, and machine learning. For example, cognitive scientists measure the representational alignment of multiple individuals to identify shared cognitive priors, neuroscientists align fMRI responses from multiple individuals into a shared representational space for group-level analyses, and ML researchers distill knowledge from teacher models into student models by increasing their alignment. Unfortunately, there is limited knowledge transfer between research communities interested in representational alignment, so progress in one field often ends up being rediscovered independently in another. Thus, greater cross-field communication would be advantageous. To improve communication between these fields, we propose a unifying framework that can serve as a common language between researchers studying representational alignment. We survey the literature from all three fields and demonstrate how prior work fits into this framework. Finally, we lay out open problems in representational alignment where progress can benefit all three of these fields. We hope that our work can catalyze cross-disciplinary collaboration and accelerate progress for all communities studying and developing information processing systems. We note that this is a working paper and encourage readers to reach out with their suggestions for future revisions."
                    },
                    {
                        "title": "PromptASTE: Prompting a Dataset from Pre-trained Language Models for Unsupervised Aspect Sentiment Triplet Extraction",
                        "abstract": "Aspect sentiment triplet extraction (ASTE) is 001 a sentiment analysis task that aims to extract 002 views\u2019 sentiment polarity, expression, and tar-003 get (aspect). This paper proposes the first un-004 supervised method for aspect sentiment triplet 005 extraction. Based on the previous discovery 006 of the pre-trained language model (PLM)\u2019s 007 awareness of sentiment, we further leverage 008 the masked language model (MLM) to prompt 009 an ASTE dataset with automatically annotated 010 labels. Our method, PromptASTE, fills in a se-011 ries of prompts to generate a dataset for related 012 aspects and views. The dataset is then used 013 to train an ASTE model for prediction. Train-014 ing on PromptASTE results in models with an 015 outstanding capability in discerning sentiment 016 polarities and targeted aspects. Our model sets 017 the first and strong baseline on unsupervised 018 ASTE. 019"
                    },
                    {
                        "title": "Beyond neural scaling laws: beating power law scaling via data pruning",
                        "abstract": "Widely observed neural scaling laws, in which error falls off as a power of the training set size, model size, or both, have driven substantial performance improvements in deep learning. However, these improvements through scaling alone require considerable costs in compute and energy. Here we focus on the scaling of error with dataset size and show how in theory we can break beyond power law scaling and potentially even reduce it to exponential scaling instead if we have access to a high-quality data pruning metric that ranks the order in which training examples should be discarded to achieve any pruned dataset size. We then test this improved scaling prediction with pruned dataset size empirically, and indeed observe better than power law scaling in practice on ResNets trained on CIFAR-10, SVHN, and ImageNet. Next, given the importance of finding high-quality pruning metrics, we perform the first large-scale benchmarking study of ten different data pruning metrics on ImageNet. We find most existing high performing metrics scale poorly to ImageNet, while the best are computationally intensive and require labels for every image. We therefore developed a new simple, cheap and scalable self-supervised pruning metric that demonstrates comparable performance to the best supervised metrics. Overall, our work suggests that the discovery of good data-pruning metrics may provide a viable path forward to substantially improved neural scaling laws, thereby reducing the resource costs of modern deep learning."
                    },
                    {
                        "title": "Why are NLP Models Fumbling at Elementary Math? A Survey of Automatic Word Problem Solvers",
                        "abstract": "From the latter half of the last decade, there 001 has been growing interest in developing algo-002 rithms for automatically solving mathematical 003 word problems (MWP). It is an exciting lan-004 guage problem which demands not only sur-005 face level text pattern recognition but requires 006 coupling with mathematical reasoning as well. 007 In spite of the dedicated effort, we are still 008 miles away from building robust representa-009 tions of elementary math word problems. In 010 this paper, we critically examine the various 011 models that have been developed for solving 012 word problems, their pros and cons and the 013 challenges ahead. In the last two years, a lot of 014 deep learning models have come out with com-015 peting results on benchmark datasets. We take 016 a step back and analyse why, in spite of this, 017 the predominantly used experiment and dataset 018 designs are a stumbling block and provide a 019 road-map for the future. 020"
                    },
                    {
                        "title": "Mosaic Augmentation for Text: Cropping and Collaging as Cross-Domain Techniques",
                        "abstract": "We present new visually inspired cropping and 001 collaging data augmentations for text. We test 002 how these augmentations impact data-scarce 003 scenarios over multiple NLP tasks: Name En-004 tity Recognition, Extractive Question Answer-005 ing and Abstractive Summarization. Ablation 006 studies show different prevailing reasons for 007 the augmentations\u2019 effectiveness for each dif-008 ferent task, but all benefit from our approach. 009 We achieve significant improvements over base-010 lines, in particular for limited data use cases. 011"
                    },
                    {
                        "title": "The developmental trajectory of object recognition robustness: Children are like small adults but unlike big deep neural networks",
                        "abstract": "In laboratory object recognition tasks based on undistorted photographs, both adult humans and deep neural networks (DNNs) perform close to ceiling. Unlike adults\u2019, whose object recognition performance is robust against a wide range of image distortions, DNNs trained on standard ImageNet (1.3M images) perform poorly on distorted images. However, the last 2 years have seen impressive gains in DNN distortion robustness, predominantly achieved through ever-increasing large-scale datasets\u2014orders of magnitude larger than ImageNet. Although this simple brute-force approach is very effective in achieving human-level robustness in DNNs, it raises the question of whether human robustness, too, is simply due to extensive experience with (distorted) visual input during childhood and beyond. Here we investigate this question by comparing the core object recognition performance of 146 children (aged 4\u201315 years) against adults and against DNNs. We find, first, that already 4- to 6-year-olds show remarkable robustness to image distortions and outperform DNNs trained on ImageNet. Second, we estimated the number of images children had been exposed to during their lifetime. Compared with various DNNs, children\u2019s high robustness requires relatively little data. Third, when recognizing objects, children\u2014like adults but unlike DNNs\u2014rely heavily on shape but not on texture cues. Together our results suggest that the remarkable robustness to distortions emerges early in the developmental trajectory of human object recognition and is unlikely the result of a mere accumulation of experience with distorted visual input. Even though current DNNs match human performance regarding robustness, they seem to rely on different and more data-hungry strategies to do so."
                    },
                    {
                        "title": "Partial success in closing the gap between human and machine vision",
                        "abstract": "A few years ago, the first CNN surpassed human performance on ImageNet. However, it soon became clear that machines lack robustness on more challenging test cases, a major obstacle towards deploying machines\"in the wild\"and towards obtaining better computational models of human visual perception. Here we ask: Are we making progress in closing the gap between human and machine vision? To answer this question, we tested human observers on a broad range of out-of-distribution (OOD) datasets, recording 85,120 psychophysical trials across 90 participants. We then investigated a range of promising machine learning developments that crucially deviate from standard supervised CNNs along three axes: objective function (self-supervised, adversarially trained, CLIP language-image training), architecture (e.g. vision transformers), and dataset size (ranging from 1M to 1B). Our findings are threefold. (1.) The longstanding distortion robustness gap between humans and CNNs is closing, with the best models now exceeding human feedforward performance on most of the investigated OOD datasets. (2.) There is still a substantial image-level consistency gap, meaning that humans make different errors than models. In contrast, most models systematically agree in their categorisation errors, even substantially different ones like contrastive self-supervised vs. standard supervised models. (3.) In many cases, human-to-model consistency improves when training dataset size is increased by one to three orders of magnitude. Our results give reason for cautious optimism: While there is still much room for improvement, the behavioural difference between human and machine vision is narrowing. In order to measure future progress, 17 OOD datasets with image-level human behavioural data and evaluation code are provided as a toolbox and benchmark at: https://github.com/bethgelab/model-vs-human/"
                    },
                    {
                        "title": "A four-year-old can outperform ResNet-50: Out-of-distribution robustness may not require large-scale experience",
                        "abstract": "Recent gains in model robustness towards out-of-distribution images are predominantly achieved through ever-increasing large-scale datasets. While this approach is very effective in achieving human-level distortion robustness, it raises the question of whether human robustness, too, requires massive amounts of experience. We therefore investigated the developmental trajectory of human object recognition robustness by comparing children aged 4\u20136, 7\u20139, 10\u201312, 13\u201315 against adults and against different deep learning models. Assessing how recognition accuracy degrades when images are distorted by salt-and-pepper noise, we \ufb01nd that while overall performance improves with age, even the youngest children in the study showed remarkable robustness and outperformed standard CNNs and self-supervised models on distorted images. In order to compare the robustness of different age groups and models as a function of visual experience, we used a back-of-the-envelope calculation to estimated the number of \u2018images\u2019 that those young children had been exposed to during their lifetime. We \ufb01nd that for humans, more data does not necessarily lead to better out-of-distribution robustness. Compared to various deep learning models, children\u2019s high out-of-distribution robustness requires relatively little data. Taken together, this indicates that human out-of-distribution robustness develops very early in life and may not require seeing billions of different images during lifetime given the right choice of representation and information processing optimised during evolution."
                    },
                    {
                        "title": "How Well do Feature Visualizations Support Causal Understanding of CNN Activations?",
                        "abstract": "A precise understanding of why units in an artificial network respond to certain stimuli would constitute a big step towards explainable artificial intelligence. One widely used approach towards this goal is to visualize unit responses via activation maximization. These synthetic feature visualizations are purported to provide humans with precise information about the image features that cause a unit to be activated - an advantage over other alternatives like strongly activating natural dataset samples. If humans indeed gain causal insight from visualizations, this should enable them to predict the effect of an intervention, such as how occluding a certain patch of the image (say, a dog's head) changes a unit's activation. Here, we test this hypothesis by asking humans to decide which of two square occlusions causes a larger change to a unit's activation. Both a large-scale crowdsourced experiment and measurements with experts show that on average the extremely activating feature visualizations by Olah et al. (2017) indeed help humans on this task ($68 \\pm 4$% accuracy; baseline performance without any visualizations is $60 \\pm 3$%). However, they do not provide any substantial advantage over other visualizations (such as e.g. dataset samples), which yield similar performance ($66\\pm3$% to $67 \\pm3$% accuracy). Taken together, we propose an objective psychophysical task to quantify the benefit of unit-level interpretability methods for humans, and find no evidence that a widely-used feature visualization method provides humans with better\"causal understanding\"of unit activations than simple alternative visualizations."
                    },
                    {
                        "title": "Natural images are more informative for interpreting CNN activations than synthetic feature visualizations",
                        "abstract": "Feature visualizations such as synthetic maximally activating images are a widely used explanation method to better understand the information processing of convolutional neural networks (CNNs). At the same time, there are concerns that these visualizations might not accurately represent CNNs\u2019 inner workings. Here, we measure how much extremely activating images help humans in predicting CNN activations. Using a well-controlled psychophysical paradigm, we compare the informativeness of synthetic images by Olah et al. [45] with a simple baseline visualization, namely natural images that also strongly activate a speci\ufb01c feature map. Given either synthetic or natural reference images, human participants choose which of two query images leads to strong positive activation. The experiment is designed to maximize participants\u2019 performance, and is the \ufb01rst to probe intermediate instead of \ufb01nal layer representations. We \ufb01nd that synthetic images indeed provide helpful information about feature map activations ( 82 \u00b1 4% accuracy; chance would be 50% ). However, natural images\u2014originally intended to be a baseline\u2014outperform these synthetic images by a wide margin ( 92 \u00b1 2% accuracy). The superiority of natural images holds across the investigated network and various conditions. Therefore, we argue that visualization methods should improve over this simple baseline."
                    },
                    {
                        "title": "Beyond accuracy: quantifying trial-by-trial behaviour of CNNs and humans by measuring error consistency",
                        "abstract": "A central problem in cognitive science and behavioural neuroscience as well as in machine learning and artificial intelligence research is to ascertain whether two or more decision makers (e.g. brains or algorithms) use the same strategy. Accuracy alone cannot distinguish between strategies: two systems may achieve similar accuracy with very different strategies. The need to differentiate beyond accuracy is particularly pressing if two systems are at or near ceiling performance, like Convolutional Neural Networks (CNNs) and humans on visual object recognition. Here we introduce trial-by-trial error consistency, a quantitative analysis for measuring whether two decision making systems systematically make errors on the same inputs. Making consistent errors on a trial-by-trial basis is a necessary condition if we want to ascertain similar processing strategies between decision makers. Our analysis is applicable to compare algorithms with algorithms, humans with humans, and algorithms with humans. When applying error consistency to visual object recognition we obtain three main findings: (1.) Irrespective of architecture, CNNs are remarkably consistent with one another (2.) The consistency between CNNs and human observers, however, is little above what can be expected by chance alone--indicating that humans and CNNs are likely implementing very different strategies (3.) CORnet-S, a recurrent model termed the \"current best model of the primate ventral visual stream\", fails to capture essential characteristics of human behavioural data and behaves essentially like a ResNet-50 in our analysis--that is, just like a standard feedforward network. Taken together, error consistency analysis suggests that the strategies used by human and machine vision are still very different--but we envision our general-purpose error consistency analysis to serve as a fruitful tool for quantifying future progress."
                    }
                ]
            }
        }
    },
    "2404.07724": {
        "paper_data": {
            "title": "Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models",
            "url": "http://arxiv.org/abs/2404.07724v1",
            "arxiv_id": "2404.07724",
            "authors": [
                "Tuomas Kynk\u00e4\u00e4nniemi",
                "Miika Aittala",
                "Tero Karras",
                "Samuli Laine",
                "Timo Aila",
                "Jaakko Lehtinen"
            ],
            "abstract": "Guidance is a crucial technique for extracting the best performance out of image-generating diffusion models. Traditionally, a constant guidance weight has been applied throughout the sampling chain of an image. We show that guidance is clearly harmful toward the beginning of the chain (high noise levels), largely unnecessary toward the end (low noise levels), and only beneficial in the middle. We thus restrict it to a specific range of noise levels, improving both the inference speed and result quality. This limited guidance interval improves the record FID in ImageNet-512 significantly, from 1.81 to 1.40. We show that it is quantitatively and qualitatively beneficial across different sampler parameters, network architectures, and datasets, including the large-scale setting of Stable Diffusion XL. We thus suggest exposing the guidance interval as a hyperparameter in all diffusion models that use guidance.",
            "introduction": " Introduction Denoising diffusion models [ 14,22,31,32,33,34,36] have enabled rapid advances in high-quality image synthesis based on text prompts and other forms of input [ 11,29,37]. They scale effortlessly to large-scale datasets [ 4,5,28], and also to other modalities such as video [ 7,6,13,16], 3D shapes [21, 25, 27, 30], and audio [20, 26]. Diffusion models convert an initial image of pure noise to a novel generated image through repeated application of image denoising. This sampling chain typically contains dozens of steps, and in each step a little bit of the denoised result is blended into the noisy image. The sampling process first gravitates towards the mean of the training data, followed by the determination of image features in an approximate coarse-to-fine manner based on the remaining noise. This iterative process, where the image is formed little by little, offers considerable flexibility in terms of encouraging or discouraging certain kinds of behavior at each step. Negative prompts [ 3] are a widely used concept, where the sampling process is given an additional anti-goal that is to be avoided. For example, \u201cnudity\u201d might be a common negative prompt in text-based image generators. At every sampling step, the denoiser is executed twice: once for the positive and once for the negative prompt, and the positive result is then extrapolated further away from the negative one based on a weight parameter. This works remarkably well in practice. Classifier-free guidance (CFG) [ 15] builds on this general concept. It uses an unconditional model (no class information or text prompts) as a negative prompt, causing the result image to align more strongly with the conditioning signal. *Equal contribution.arXiv:2404.07724v1  [cs.CV]  11 Apr 2024x (a) PDFs \u03c3 (b) Unconditional \u03c3 (c) Conditional \u03c3 (d) Guided, \u03b3= 6  (e) Tangent vectors Figure 1: Visualizing the effect of guidance. (a)The unconditional (orange) and conditional (green) PDFs. In (b) through (e), the orange unconditional density is visualized in the background. Figure 11: More SD-XL results showing the effect of changing \u03b3with our method. We limit the guidance to \u03c3\u2208(0.19,1.61]. Increasing \u03b3produces images with more well-defined details while maintaining the color palette and the original image composition. 16\u03b3= 1 \u03b3= 2 \u03b3= 4 \u03b3= 8 \u03b3= 12 \u03b3= 16CFG  Ours, \u03c3\u2208(0.28,5.42] A highly detailed paper origami of a Dachshund on a table next to a porcelain teapot, 4k dslr. Figure 13: Effect of increasing guidance weight \u03b3with CFG vs. our method. Top: Increasing the guidance weight with CFG leads to large changes in the image composition. Note how the dog\u2019s head moves as \u03b3changes. Bottom: Our method leads to well-defined image details and retains the overall composition to a significant degree. 17\u03b3= 1 \u03b3= 2 \u03b3= 4 \u03b3= 8 \u03b3= 12 \u03b3= 16CFG  Ours, \u03c3\u2208(0.28,5.42] A fantasy landscape on an alien planet in which there are many buildings. There is a beautiful bridge with a pond in the center. There is one large moon in the sky. The sky is orange. Digital art, artstation Figure 14: Effect of increasing guidance weight \u03b3with CFG vs. our method. Top: Increasing the guidance weight with CFG leads to large changes in the image composition. Bottom: Our method leads to well-defined image details and retains the overall composition to a significant degree. 18CFG Ours, \u03c3\u2208(0.19,1.61] \u03b3= 1 \u03b3= 3 \u03b3= 5 \u03b3= 1 \u03b3= 3 \u03b3= 5ImageNet class 33: loggerhead  ImageNet",
            "references": [
                {
                    "title": "Adaptive Guidance: Training-free Acceleration of Conditional Diffusion Models",
                    "abstract": "This paper presents a comprehensive study on the role of Classifier-Free Guidance (CFG) in text-conditioned diffusion models from the perspective of inference efficiency. In particular, we relax the default choice of applying CFG in all diffusion steps and instead search for efficient guidance policies. We formulate the discovery of such policies in the differentiable Neural Architecture Search framework. Our findings suggest that the denoising steps proposed by CFG become increasingly aligned with simple conditional steps, which renders the extra neural network evaluation of CFG redundant, especially in the second half of the denoising process. Building upon this insight, we propose\"Adaptive Guidance\"(AG), an efficient variant of CFG, that adaptively omits network evaluations when the denoising process displays convergence. Our experiments demonstrate that AG preserves CFG's image quality while reducing computation by 25%. Thus, AG constitutes a plug-and-play alternative to Guidance Distillation, achieving 50% of the speed-ups of the latter while being training-free and retaining the capacity to handle negative prompts. Finally, we uncover further redundancies of CFG in the first half of the diffusion process, showing that entire neural function evaluations can be replaced by simple affine transformations of past score estimates. This method, termed LinearAG, offers even cheaper inference at the cost of deviating from the baseline model. Our findings provide insights into the efficiency of the conditional denoising process that contribute to more practical and swift deployment of text-conditioned diffusion models."
                },
                {
                    "title": "Analyzing and Improving the Training Dynamics of Diffusion Models",
                    "abstract": "Diffusion models currently dominate the field of data- driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high- level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on ex- pectation. We find that systematic application of this philoso- phy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational com- plexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling. As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance."
                },
                {
                    "title": "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis",
                    "abstract": "We present SDXL, a latent diffusion model for text-to-image synthesis. Compared to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet backbone: The increase of model parameters is mainly due to more attention blocks and a larger cross-attention context as SDXL uses a second text encoder. We design multiple novel conditioning schemes and train SDXL on multiple aspect ratios. We also introduce a refinement model which is used to improve the visual fidelity of samples generated by SDXL using a post-hoc image-to-image technique. We demonstrate that SDXL shows drastically improved performance compared the previous versions of Stable Diffusion and achieves results competitive with those of black-box state-of-the-art image generators. In the spirit of promoting open research and fostering transparency in large model training and evaluation, we provide access to code and model weights at https://github.com/Stability-AI/generative-models"
                },
                {
                    "title": "Exposing flaws of generative model evaluation metrics and their unfair treatment of diffusion models",
                    "abstract": "We systematically study a wide variety of generative models spanning semantically-diverse image datasets to understand and improve the feature extractors and metrics used to evaluate them. Using best practices in psychophysics, we measure human perception of image realism for generated samples by conducting the largest experiment evaluating generative models to date, and find that no existing metric strongly correlates with human evaluations. Comparing to 17 modern metrics for evaluating the overall performance, fidelity, diversity, rarity, and memorization of generative models, we find that the state-of-the-art perceptual realism of diffusion models as judged by humans is not reflected in commonly reported metrics such as FID. This discrepancy is not explained by diversity in generated samples, though one cause is over-reliance on Inception-V3. We address these flaws through a study of alternative self-supervised feature extractors, find that the semantic information encoded by individual networks strongly depends on their training procedure, and show that DINOv2-ViT-L/14 allows for much richer evaluation of generative models. Next, we investigate data memorization, and find that generative models do memorize training examples on simple, smaller datasets like CIFAR10, but not necessarily on more complex datasets like ImageNet. However, our experiments show that current metrics do not properly detect memorization: none in the literature is able to separate memorization from other phenomena such as underfitting or mode shrinkage. To facilitate further development of generative models and their evaluation we release all generated image datasets, human evaluation data, and a modular library to compute 17 common metrics for 9 different encoders at https://github.com/layer6ai-labs/dgm-eval."
                },
                {
                    "title": "Align Your Latents: High-Resolution Video Synthesis with Latent Diffusion Models",
                    "abstract": "Latent Diffusion Models (LDMs) enable high-quality image synthesis while avoiding excessive compute demands by training a diffusion model in a compressed lower-dimensional latent space. Here, we apply the LDM paradigm to high-resolution video generation, a particularly resource-intensive task. We first pre-train an LDM on images only; then, we turn the image generator into a video generator by introducing a temporal dimension to the latent space diffusion model and finetuning on encoded image sequences, i.e., videos. Similarly, we temporally align diffusion model upsamplers, turning them into temporally consistent video super resolution models. We focus on two relevant real-world applications: Simulation of in-the-wild driving data and creative content creation with text-to-video modeling. In particular, we validate our Video LDM on real driving videos of resolution $512 \\times 1024$, achieving state-of-the-art performance. Furthermore, our approach can easily leverage off-the-shelf pretrained image LDMs, as we only need to train a temporal alignment model in that case. Doing so, we turn the publicly available, state-of-the-art text-to-image LDM Stable Diffusion into an efficient and expressive text-to-video model with resolution up to $1280 \\times 2048$. We show that the temporal layers trained in this way generalize to different finetuned text-to-image LDMs. Utilizing this property, we show the first results for personalized text-to-video generation, opening exciting directions for future content creation. Project page: https://nv-tlabs.github.io/VideoLDM/"
                },
                {
                    "title": "DreamBooth3D: Subject-Driven Text-to-3D Generation",
                    "abstract": "We present DreamBooth3D, an approach to personalize text-to-3D generative models from as few as 3-6 casually captured images of a subject. Our approach combines recent advances in personalizing text-to-image models (DreamBooth) with text-to-3D generation (DreamFusion). We find that na\u00efvely combining these methods fails to yield satisfactory subject-specific 3D assets due to personalized text-to-image models overfitting to the input viewpoints of the subject. We overcome this through a 3-stage optimization strategy where we jointly leverage the 3D consistency of neural radiance fields together with the personalization capability of text-to-image models. Our method can produce high-quality, subject-specific 3D assets with text-driven modifications such as novel poses, colors and attributes that are not seen in any of the input images of the subject. More results are available at our project page: https://dreambooth3d.github.io"
                },
                {
                    "title": "Adding Conditional Control to Text-to-Image Diffusion Models",
                    "abstract": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models."
                },
                {
                    "title": "Scalable Diffusion Models with Transformers",
                    "abstract": "We explore a new class of diffusion models based on the transformer architecture. We train latent diffusion models of images, replacing the commonly-used U-Net backbone with a transformer that operates on latent patches. We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops. We find that DiTs with higher Gflops\u2014through increased transformer depth/width or increased number of input tokens\u2014consistently have lower FID. In addition to possessing good scalability properties, our largest DiT-XL/2 models outperform all prior diffusion models on the class-conditional ImageNet 512\u00d7512 and 256\u00d7256 benchmarks, achieving a state-of-the-art FID of 2.27 on the latter."
                },
                {
                    "title": "3D Neural Field Generation Using Triplane Diffusion",
                    "abstract": "Diffusion models have emerged as the state-of-the-art for image generation, among other tasks. Here, we present an efficient diffusion-based model for 3D-aware generation of neural fields. Our approach pre-processes training data, such as ShapeNet meshes, by converting them to continuous occupancy fields and factoring them into a set of axis-aligned triplane feature representations. Thus, our 3D training scenes are all represented by 2D feature planes, and we can directly train existing 2D diffusion models on these representations to generate 3D neural fields with high quality and diversity, outperforming alternative approaches to 3D-aware generation. Our approach requires essential modifications to existing triplane factorization pipelines to make the resulting features easy to learn for the diffusion model. We demonstrate state-of-the-art results on 3D generation on several object classes from ShapeNet."
                },
                {
                    "title": "Magic3D: High-Resolution Text-to-3D Content Creation",
                    "abstract": "DreamFusion [31] has recently demonstrated the utility of a pretrained text-to-image diffusion model to optimize Neural Radiance Fields (NeRF) [23], achieving remarkable text-to-3D synthesis results. However, the method has two inherent limitations: (a) extremely slow optimization of NeRF and (b) low-resolution image space supervision on NeRF, leading to low-quality 3D models with a long processing time. In this paper, we address these limitations by utilizing a two-stage optimization framework. First, we obtain a coarse model using a low-resolution diffusion prior and accelerate with a sparse 3D hash grid structure. Using the coarse representation as the initialization, we further optimize a textured 3D mesh model with an efficient differentiable renderer interacting with a high-resolution latent diffusion model. Our method, dubbed Magic3D, can create high quality 3D mesh models in 40 minutes, which is 2\u00d7 faster than DreamFusion (reportedly taking 1.5 hours on average), while also achieving higher resolution. User studies show 61.7% raters to prefer our approach over DreamFusion. Together with the image-conditioned generation capabilities, we provide users with new ways to control 3D synthesis, opening up new avenues to various creative applications."
                },
                {
                    "title": "eDiff-I: Text-to-Image Diffusion Models with an Ensemble of Expert Denoisers",
                    "abstract": "Large-scale diffusion-based generative models have led to breakthroughs in text-conditioned high-resolution image synthesis. Starting from random noise, such text-to-image diffusion models gradually synthesize images in an iterative fashion while conditioning on text prompts. We find that their synthesis behavior qualitatively changes throughout this process: Early in sampling, generation strongly relies on the text prompt to generate text-aligned content, while later, the text conditioning is almost entirely ignored. This suggests that sharing model parameters throughout the entire generation process may not be ideal. Therefore, in contrast to existing works, we propose to train an ensemble of text-to-image diffusion models specialized for different synthesis stages. To maintain training efficiency, we initially train a single model, which is then split into specialized models that are trained for the specific stages of the iterative generation process. Our ensemble of diffusion models, called eDiff-I, results in improved text alignment while maintaining the same inference computation cost and preserving high visual quality, outperforming previous large-scale text-to-image diffusion models on the standard benchmark. In addition, we train our model to exploit a variety of embeddings for conditioning, including the T5 text, CLIP text, and CLIP image embeddings. We show that these different embeddings lead to different behaviors. Notably, the CLIP image embedding allows an intuitive way of transferring the style of a reference image to the target text-to-image output. Lastly, we show a technique that enables eDiff-I's\"paint-with-words\"capability. A user can select the word in the input text and paint it in a canvas to control the output, which is very handy for crafting the desired image in mind. The project page is available at https://deepimagination.cc/eDiff-I/"
                },
                {
                    "title": "Imagen Video: High Definition Video Generation with Diffusion Models",
                    "abstract": "We present Imagen Video, a text-conditional video generation system based on a cascade of video diffusion models. Given a text prompt, Imagen Video generates high definition videos using a base video generation model and a sequence of interleaved spatial and temporal video super-resolution models. We describe how we scale up the system as a high definition text-to-video model including design decisions such as the choice of fully-convolutional temporal and spatial super-resolution models at certain resolutions, and the choice of the v-parameterization of diffusion models. In addition, we confirm and transfer findings from previous work on diffusion-based image generation to the video generation setting. Finally, we apply progressive distillation to our video models with classifier-free guidance for fast, high quality sampling. We find Imagen Video not only capable of generating videos of high fidelity, but also having a high degree of controllability and world knowledge, including the ability to generate diverse videos and text animations in various artistic styles and with 3D object understanding. See https://imagen.research.google/video/ for samples."
                },
                {
                    "title": "DreamFusion: Text-to-3D using 2D Diffusion",
                    "abstract": "Recent breakthroughs in text-to-image synthesis have been driven by diffusion models trained on billions of image-text pairs. Adapting this approach to 3D synthesis would require large-scale datasets of labeled 3D data and efficient architectures for denoising 3D data, neither of which currently exist. In this work, we circumvent these limitations by using a pretrained 2D text-to-image diffusion model to perform text-to-3D synthesis. We introduce a loss based on probability density distillation that enables the use of a 2D diffusion model as a prior for optimization of a parametric image generator. Using this loss in a DeepDream-like procedure, we optimize a randomly-initialized 3D model (a Neural Radiance Field, or NeRF) via gradient descent such that its 2D renderings from random angles achieve a low loss. The resulting 3D model of the given text can be viewed from any angle, relit by arbitrary illumination, or composited into any 3D environment. Our approach requires no 3D training data and no modifications to the image diffusion model, demonstrating the effectiveness of pretrained image diffusion models as priors."
                },
                {
                    "title": "DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation",
                    "abstract": "Large text-to-image models achieved a remarkable leap in the evolution of AI, enabling high-quality and diverse synthesis of images from a given text prompt. However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts. In this work, we present a new approach for \u201cpersonalization\u201d of text-to-image diffusion models. Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject. Once the subject is embedded in the output domain of the model, the unique identifier can be used to synthesize novel photorealistic images of the subject contextualized in different scenes. By leveraging the semantic prior embedded in the model with a new autogenous class-specific prior preservation loss, our technique enables synthesizing the subject in diverse scenes, poses, views and lighting conditions that do not appear in the reference images. We apply our technique to several previously-unassailable tasks, including subject recontextualization, text-guided view synthesis, and artistic rendering, all while preserving the subject's key features. We also provide a new dataset and evaluation protocol for this new task of subject-driven generation. Project page: https://dreambooth.github.io/"
                },
                {
                    "title": "An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion",
                    "abstract": "Text-to-image models offer unprecedented freedom to guide creation through natural language. Yet, it is unclear how such freedom can be exercised to generate images of specific unique concepts, modify their appearance, or compose them in new roles and novel scenes. In other words, we ask: how can we use language-guided models to turn our cat into a painting, or imagine a new product based on our favorite toy? Here we present a simple approach that allows such creative freedom. Using only 3-5 images of a user-provided concept, like an object or a style, we learn to represent it through new\"words\"in the embedding space of a frozen text-to-image model. These\"words\"can be composed into natural language sentences, guiding personalized creation in an intuitive way. Notably, we find evidence that a single word embedding is sufficient for capturing unique and varied concepts. We compare our approach to a wide range of baselines, and demonstrate that it can more faithfully portray the concepts across a range of applications and tasks. Our code, data and new words will be available at: https://textual-inversion.github.io"
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                    "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                },
                {
                    "title": "Video Diffusion Models",
                    "abstract": "Generating temporally coherent high fidelity video is an important milestone in generative modeling research. We make progress towards this milestone by proposing a diffusion model for video generation that shows very promising initial results. Our model is a natural extension of the standard image diffusion architecture, and it enables jointly training from image and video data, which we find to reduce the variance of minibatch gradients and speed up optimization. To generate long and higher resolution videos we introduce a new conditional sampling technique for spatial and temporal video extension that performs better than previously proposed methods. We present the first results on a large text-conditioned video generation task, as well as state-of-the-art results on established benchmarks for video prediction and unconditional video generation. Supplementary material is available at https://video-diffusion.github.io/"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Grad-TTS: A Diffusion Probabilistic Model for Text-to-Speech",
                    "abstract": "Recently, denoising diffusion probabilistic models and generative score matching have shown high potential in modelling complex data distributions while stochastic calculus has provided a unified point of view on these techniques allowing for flexible inference schemes. In this paper we introduce Grad-TTS, a novel text-to-speech model with score-based decoder producing mel-spectrograms by gradually transforming noise predicted by encoder and aligned with text input by means of Monotonic Alignment Search. The framework of stochastic differential equations helps us to generalize conventional diffusion probabilistic models to the case of reconstructing data from noise with different parameters and allows to make this reconstruction flexible by explicitly controlling trade-off between sound quality and inference speed. Subjective human evaluation shows that Grad-TTS is competitive with state-of-the-art text-to-speech approaches in terms of Mean Opinion Score. We will make the code publicly available shortly."
                },
                {
                    "title": "Improved Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) are a class of generative models which have recently been shown to produce excellent samples. We show that with a few simple modifications, DDPMs can also achieve competitive log-likelihoods while maintaining high sample quality. Additionally, we find that learning variances of the reverse diffusion process allows sampling with an order of magnitude fewer forward passes with a negligible difference in sample quality, which is important for the practical deployment of these models. We additionally use precision and recall to compare how well DDPMs and GANs cover the target distribution. Finally, we show that the sample quality and likelihood of these models scale smoothly with model capacity and training compute, making them easily scalable. We release our code at https://github.com/openai/improved-diffusion"
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "DiffWave: A Versatile Diffusion Model for Audio Synthesis",
                    "abstract": "In this work, we propose DiffWave, a versatile Diffusion probabilistic model for conditional and unconditional Waveform generation. The model is non-autoregressive, and converts the white noise signal into structured waveform through a Markov chain with a constant number of steps at synthesis. It is efficiently trained by optimizing a variant of variational bound on the data likelihood. DiffWave produces high-fidelity audios in Different Waveform generation tasks, including neural vocoding conditioned on mel spectrogram, class-conditional generation, and unconditional generation. We demonstrate that DiffWave matches a strong WaveNet vocoder in terms of speech quality~(MOS: 4.44 versus 4.43), while synthesizing orders of magnitude faster. In particular, it significantly outperforms autoregressive and GAN-based waveform models in the challenging unconditional generation task in terms of audio quality and sample diversity from various automatic and human evaluations."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium",
                    "abstract": "Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the \"Frechet Inception Distance\" (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "A Connection Between Score Matching and Denoising Autoencoders",
                    "abstract": "Denoising autoencoders have been previously shown to be competitive alternatives to restricted Boltzmann machines for unsupervised pretraining of each layer of a deep architecture. We show that a simple denoising autoencoder training criterion is equivalent to matching the score (with respect to the data) of a specific energy-based model to that of a nonparametric Parzen density estimator of the data. This yields several useful insights. It defines a proper probabilistic model for the denoising autoencoder technique, which makes it in principle possible to sample from them or rank examples by their energy. It suggests a different way to apply score matching that is related to learning to denoise and does not require computing second derivatives. It justifies the use of tied weights between the encoder and decoder and suggests ways to extend the success of denoising autoencoders to a larger family of energy-based models."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Improving Image Generation with Better Captions",
                    "abstract": "We show that prompt following abilities of text-to-image models can be substantially improved by training on highly descriptive generated image captions. Existing text-to-image models struggle to follow detailed image descriptions and often ignore words or confuse the meaning of prompts. We hypothesize that this issue stems from noisy and inaccurate image captions in the training dataset. We address this by training a bespoke image captioner and use it to recaption the training dataset. We then train several text-to-image models and find that training on these synthetic captions reliably improves prompt following ability. Finally, we use these findings to build DALL-E 3: a new text-to-image generation system, and benchmark its performance on an evaluation designed to measure prompt following, coherence, and aesthetics, finding that it compares favorably to competitors. We publish samples and code for these evaluations so that future research can continue optimizing this important aspect of text-to-image systems."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI",
                "cs.LG",
                "cs.NE",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the guidance mechanism in denoising diffusion models to enhance image synthesis while maintaining the integrity of the original composition?\n\n**[Question 2] - Why is it interesting and important?**  \nImproving the guidance mechanism in denoising diffusion models is crucial for advancing the field of generative models, particularly in image synthesis. Enhanced guidance can lead to more accurate and visually appealing outputs, which can have significant implications for various applications, including art generation, virtual reality, and content creation. By addressing this problem, future research can explore more sophisticated generative techniques, potentially leading to breakthroughs in multimodal generation and better alignment of generated content with user intentions.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in balancing the influence of positive and negative prompts during the denoising process. Naive approaches may lead to overemphasis on either the positive or negative guidance, resulting in distorted images or loss of essential details. The complexities include managing the iterative nature of the sampling process, ensuring that the generated images retain their original composition while also being influenced by the guidance parameters. Additionally, the technical challenge of fine-tuning the guidance weight without compromising the quality of the output adds to the difficulty.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either unconditional or conditional models without adequately addressing the nuances of guidance weight adjustments. Limitations in understanding the interplay between different types of prompts and their effects on image quality have hindered progress. Existing solutions often lead to significant changes in image composition when guidance weights are altered, which can detract from the desired output. Our approach aims to refine this process by introducing a more controlled method of guidance that preserves the overall composition while enhancing detail.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a novel adjustment of the guidance weight during the denoising process, utilizing a dataset of high-quality images for training. We will employ metrics such as perceptual similarity and detail retention to evaluate the effectiveness of our approach. The expected outcomes include improved image quality with well-defined details while maintaining the original composition, as evidenced by comparative analysis against existing methods like Classifier-Free Guidance (CFG). This will demonstrate the advantages of our method in generating visually coherent and detailed images."
            }
        },
        "author_data": {
            "339f91d6-7a8d-412b-a6cb-61a55226e31f": {
                "pk": "339f91d6-7a8d-412b-a6cb-61a55226e31f",
                "name": "Tuomas Kynk\u00e4\u00e4nniemi",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "011b08d9-50a0-4a88-9f53-ae7093012e81": {
                "pk": "011b08d9-50a0-4a88-9f53-ae7093012e81",
                "name": "Miika Aittala",
                "collaborators": [
                    "Samuli Laine",
                    "Jaakko Lehtinen",
                    "Tero Karras",
                    "Timo Aila",
                    "Janne Hellsten",
                    "Fredo Durand",
                    "Tuomas Kynk\u00e4\u00e4nniemi",
                    "Lukas Murmann",
                    "Jon Hasselgren",
                    "Jacob Munkberg"
                ],
                "domain": [
                    "Computer Vision",
                    "Generative Models",
                    "Image Processing",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "A Dataset of Multi-Illumination Images in the Wild",
                        "abstract": "Collections of images under a single, uncontrolled illumination have enabled the rapid advancement of core computer vision tasks like classification, detection, and segmentation. But even with modern learning techniques, many inverse problems involving lighting and material understanding remain too severely ill-posed to be solved with single-illumination datasets. To fill this gap, we introduce a new multi-illumination dataset of more than 1000 real scenes, each captured under 25 lighting conditions. We demonstrate the richness of this dataset by training state-of-the-art models for three challenging applications: single-image illumination estimation, image relighting, and mixed-illuminant white balance."
                    },
                    {
                        "title": "Generating Training Data for Denoising Real RGB Images via Camera Pipeline Simulation",
                        "abstract": "Image reconstruction techniques such as denoising often need to be applied to the RGB output of cameras and cellphones. Unfortunately, the commonly used additive white noise (AWGN) models do not accurately reproduce the noise and the degradation encountered on these inputs. This is particularly important for learning-based techniques, because the mismatch between training and real world data will hurt their generalization. This paper aims to accurately simulate the degradation and noise transformation performed by camera pipelines. This allows us to generate realistic degradation in RGB images that can be used to train machine learning models. We use our simulation to study the importance of noise modeling for learning-based denoising. Our study shows that a realistic noise model is required for learning to denoise real JPEG images. A neural network trained on realistic noise outperforms the one trained with AWGN by 3 dB. An ablation study of our pipeline shows that simulating denoising and demosaicking is important to this improvement and that realistic demosaicking algorithms, which have been rarely considered, is needed. We believe this simulation will also be useful for other image reconstruction tasks, and we will distribute our code publicly."
                    },
                    {
                        "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                        "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                    },
                    {
                        "title": "The Role of ImageNet Classes in Fr\u00e9chet Inception Distance",
                        "abstract": "Fr\\'echet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID \"looks at\" in generated images. We show that the feature space that FID is (typically) computed in is so close to the ImageNet classifications that aligning the histograms of Top-$N$ classifications between sets of generated and real images can reduce FID substantially -- without actually improving the quality of results. Thus, we conclude that FID is prone to intentional or accidental distortions. As a practical example of an accidental distortion, we discuss a case where an ImageNet pre-trained FastGAN achieves a FID comparable to StyleGAN2, while being worse in terms of human evaluation."
                    },
                    {
                        "title": "Simulator-Based Self-Supervision for Learned 3D Tomography Reconstruction",
                        "abstract": "We propose a deep learning method for 3D volumetric reconstruction in low-dose helical cone-beam computed tomography. Prior machine learning approaches require reference reconstructions computed by another algorithm for training. In contrast, we train our model in a fully self-supervised manner using only noisy 2D X-ray data. This is enabled by incorporating a fast differentiable CT simulator in the training loop. As we do not rely on reference reconstructions, the fidelity of our results is not limited by their potential shortcomings. We evaluate our method on real helical cone-beam projections and simulated phantoms. Our results show significantly higher visual fidelity and better PSNR over techniques that rely on existing reconstructions. When applied to full-dose data, our method produces high-quality results orders of magnitude faster than iterative techniques."
                    },
                    {
                        "title": "Appearance-Driven Automatic 3D Model Simplification",
                        "abstract": "We present a suite of techniques for jointly optimizing triangle meshes and shading models to match the appearance of reference scenes. This capability has a number of uses, including appearance-preserving simplification of extremely complex assets, conversion between rendering systems, and even conversion between geometric scene representations.   We follow and extend the classic analysis-by-synthesis family of techniques: enabled by a highly efficient differentiable renderer and modern nonlinear optimization algorithms, our results are driven to minimize the image-space difference to the target scene when rendered in similar viewing and lighting conditions. As the only signals driving the optimization are differences in rendered images, the approach is highly general and versatile: it easily supports many different forward rendering models such as normal mapping, spatially-varying BRDFs, displacement mapping, etc. Supervision through images only is also key to the ability to easily convert between rendering systems and scene representations.   We output triangle meshes with textured materials to ensure that the models render efficiently on modern graphics hardware and benefit from, e.g., hardware-accelerated rasterization, ray tracing, and filtered texture lookups. Our system is integrated in a small Python code base, and can be applied at high resolutions and on large models. We describe several use cases, including mesh decimation, level of detail generation, seamless mesh filtering and approximations of aggregate geometry."
                    },
                    {
                        "title": "Computational Mirrors: Blind Inverse Light Transport by Deep Matrix Factorization",
                        "abstract": "We recover a video of the motion taking place in a hidden scene by observing changes in indirect illumination in a nearby uncalibrated visible region. We solve this problem by factoring the observed video into a matrix product between the unknown hidden scene video and an unknown light transport matrix. This task is extremely ill-posed, as any non-negative factorization will satisfy the data. Inspired by recent work on the Deep Image Prior, we parameterize the factor matrices using randomly initialized convolutional neural networks trained in a one-off manner, and show that this results in decompositions that reflect the true motion in the hidden scene."
                    },
                    {
                        "title": "Flexible SVBRDF Capture with a Multi-Image Deep Network",
                        "abstract": "Empowered by deep learning, recent methods for material capture can estimate a spatially-varying reflectance from a single photograph. Such lightweight capture is in stark contrast with the tens or hundreds of pictures required by traditional optimization-based approaches. However, a single image is often simply not enough to observe the rich appearance of real-world materials. We present a deep-learning method capable of estimating material appearance from a variable number of uncalibrated and unordered pictures captured with a handheld camera and flash. Thanks to an order-independent fusing layer, this architecture extracts the most useful information from each picture, while benefiting from strong priors learned from data. The method can handle both view and light direction variation without calibration. We show how our method improves its prediction with the number of input pictures, and reaches high quality reconstructions with as little as 1 to 10 images -- a sweet spot between existing single-image and complex multi-image approaches."
                    },
                    {
                        "title": "Noise2Noise: Learning Image Restoration without Clean Data",
                        "abstract": "We apply basic statistical reasoning to signal reconstruction by machine learning -- learning to map corrupted observations to clean signals -- with a simple and powerful conclusion: it is possible to learn to restore images by only looking at corrupted examples, at performance at and sometimes exceeding training using clean data, without explicit image priors or likelihood models of the corruption. In practice, we show that a single model learns photographic noise removal, denoising synthetic Monte Carlo images, and reconstruction of undersampled MRI scans -- all corrupted by different processes -- based on noisy data only."
                    },
                    {
                        "title": "Training Generative Adversarial Networks with Limited Data",
                        "abstract": "Training generative adversarial networks (GAN) using too little data typically leads to discriminator overfitting, causing training to diverge. We propose an adaptive discriminator augmentation mechanism that significantly stabilizes training in limited data regimes. The approach does not require changes to loss functions or network architectures, and is applicable both when training from scratch and when fine-tuning an existing GAN on another dataset. We demonstrate, on several datasets, that good results are now possible using only a few thousand training images, often matching StyleGAN2 results with an order of magnitude fewer images. We expect this to open up new application domains for GANs. We also find that the widely used CIFAR-10 is, in fact, a limited data benchmark, and improve the record FID from 5.59 to 2.42."
                    },
                    {
                        "title": "Analyzing and Improving the Image Quality of StyleGAN",
                        "abstract": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign the generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent codes to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably attribute a generated image to a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality."
                    },
                    {
                        "title": "Alias-Free Generative Adversarial Networks",
                        "abstract": "We observe that despite their hierarchical convolutional nature, the synthesis process of typical generative adversarial networks depends on absolute pixel coordinates in an unhealthy manner. This manifests itself as, e.g., detail appearing to be glued to image coordinates instead of the surfaces of depicted objects. We trace the root cause to careless signal processing that causes aliasing in the generator network. Interpreting all signals in the network as continuous, we derive generally applicable, small architectural changes that guarantee that unwanted information cannot leak into the hierarchical synthesis process. The resulting networks match the FID of StyleGAN2 but differ dramatically in their internal representations, and they are fully equivariant to translation and rotation even at subpixel scales. Our results pave the way for generative models better suited for video and animation."
                    },
                    {
                        "title": "Analyzing and Improving the Training Dynamics of Diffusion Models",
                        "abstract": "Diffusion models currently dominate the field of data-driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high-level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on expectation. We find that systematic application of this philosophy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational complexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling.   As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance."
                    },
                    {
                        "title": "Guiding a Diffusion Model with a Bad Version of Itself",
                        "abstract": "The primary axes of interest in image-generating diffusion models are image quality, the amount of variation in the results, and how well the results align with a given condition, e.g., a class label or a text prompt. The popular classifier-free guidance approach uses an unconditional model to guide a conditional model, leading to simultaneously better prompt alignment and higher-quality images at the cost of reduced variation. These effects seem inherently entangled, and thus hard to control. We make the surprising observation that it is possible to obtain disentangled control over image quality without compromising the amount of variation by guiding generation using a smaller, less-trained version of the model itself rather than an unconditional model. This leads to significant improvements in ImageNet generation, setting record FIDs of 1.01 for 64x64 and 1.25 for 512x512, using publicly available networks. Furthermore, the method is also applicable to unconditional diffusion models, drastically improving their quality."
                    },
                    {
                        "title": "Disentangling Random and Cyclic Effects in Time-Lapse Sequences",
                        "abstract": "Time-lapse image sequences offer visually compelling insights into dynamic processes that are too slow to observe in real time. However, playing a long time-lapse sequence back as a video often results in distracting flicker due to random effects, such as weather, as well as cyclic effects, such as the day-night cycle. We introduce the problem of disentangling time-lapse sequences in a way that allows separate, after-the-fact control of overall trends, cyclic effects, and random effects in the images, and describe a technique based on data-driven generative models that achieves this goal. This enables us to \"re-render\" the sequences in ways that would not be possible with the input images alone. For example, we can stabilize a long sequence to focus on plant growth over many months, under selectable, consistent weather.   Our approach is based on Generative Adversarial Networks (GAN) that are conditioned with the time coordinate of the time-lapse sequence. Our architecture and training procedure are designed so that the networks learn to model random variations, such as weather, using the GAN's latent space, and to disentangle overall trends and cyclic variations by feeding the conditioning time label to the model using Fourier features with specific frequencies.   We show that our models are robust to defects in the training data, enabling us to amend some of the practical difficulties in capturing long time-lapse sequences, such as temporary occlusions, uneven frame spacing, and missing frames."
                    }
                ]
            },
            "1a13a2c6-7b1b-4901-9ca5-2d1d20915f77": {
                "pk": "1a13a2c6-7b1b-4901-9ca5-2d1d20915f77",
                "name": "Tero Karras",
                "collaborators": [
                    "Timo Aila",
                    "Samuli Laine",
                    "Jaakko Lehtinen",
                    "Miika Aittala",
                    "Janne Hellsten",
                    "Tuomas Kynk\u00e4\u00e4nniemi",
                    "Pavlo Molchanov",
                    "Stephen Tyree",
                    "Jan Kautz",
                    "Onni Kosomaa"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Image Synthesis",
                    "Deep Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Progressive Growing of GANs for Improved Quality, Stability, and Variation",
                        "abstract": "We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024^2. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10. Additionally, we describe several implementation details that are important for discouraging unhealthy competition between the generator and discriminator. Finally, we suggest a new metric for evaluating GAN results, both in terms of image quality and variation. As an additional contribution, we construct a higher-quality version of the CelebA dataset."
                    },
                    {
                        "title": "High-Quality Self-Supervised Deep Image Denoising",
                        "abstract": "We describe a novel method for training high-quality image denoising models based on unorganized collections of corrupted images. The training does not need access to clean reference images, or explicit pairs of corrupted images, and can thus be applied in situations where such data is unacceptably expensive or impossible to acquire. We build on a recent technique that removes the need for reference data by employing networks with a \"blind spot\" in the receptive field, and significantly improve two key aspects: image quality and training efficiency. Our result quality is on par with state-of-the-art neural network denoisers in the case of i.i.d. additive Gaussian noise, and not far behind with Poisson and impulse noise. We also successfully handle cases where parameters of the noise model are variable and/or unknown in both training and evaluation data."
                    },
                    {
                        "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                        "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                    },
                    {
                        "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                        "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                    },
                    {
                        "title": "Improved Precision and Recall Metric for Assessing Generative Models",
                        "abstract": "The ability to automatically estimate the quality and coverage of the samples produced by a generative model is a vital requirement for driving algorithm research. We present an evaluation metric that can separately and reliably measure both of these aspects in image generation tasks by forming explicit, non-parametric representations of the manifolds of real and generated data. We demonstrate the effectiveness of our metric in StyleGAN and BigGAN by providing several illustrative examples where existing metrics yield uninformative or contradictory results. Furthermore, we analyze multiple design variants of StyleGAN to better understand the relationships between the model architecture, training methods, and the properties of the resulting sample distribution. In the process, we identify new variants that improve the state-of-the-art. We also perform the first principled analysis of truncation methods and identify an improved method. Finally, we extend our metric to estimate the perceptual quality of individual samples, and use this to study latent space interpolations."
                    },
                    {
                        "title": "Pruning Convolutional Neural Networks for Resource Efficient Inference",
                        "abstract": "We propose a new formulation for pruning convolutional kernels in neural networks to enable efficient inference. We interleave greedy criteria-based pruning with fine-tuning by backpropagation - a computationally efficient procedure that maintains good generalization in the pruned network. We propose a new criterion based on Taylor expansion that approximates the change in the cost function induced by pruning network parameters. We focus on transfer learning, where large pretrained networks are adapted to specialized tasks. The proposed criterion demonstrates superior performance compared to other criteria, e.g. the norm of kernel weights or feature map activation, for pruning large CNNs after adaptation to fine-grained classification tasks (Birds-200 and Flowers-102) relaying only on the first order gradient information. We also show that pruning can lead to more than 10x theoretical (5x practical) reduction in adapted 3D-convolutional filters with a small drop in accuracy in a recurrent gesture classifier. Finally, we show results for the large-scale ImageNet dataset to emphasize the flexibility of our approach."
                    },
                    {
                        "title": "The Role of ImageNet Classes in Fr\u00e9chet Inception Distance",
                        "abstract": "Fr\\'echet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID \"looks at\" in generated images. We show that the feature space that FID is (typically) computed in is so close to the ImageNet classifications that aligning the histograms of Top-$N$ classifications between sets of generated and real images can reduce FID substantially -- without actually improving the quality of results. Thus, we conclude that FID is prone to intentional or accidental distortions. As a practical example of an accidental distortion, we discuss a case where an ImageNet pre-trained FastGAN achieves a FID comparable to StyleGAN2, while being worse in terms of human evaluation."
                    },
                    {
                        "title": "Simulator-Based Self-Supervision for Learned 3D Tomography Reconstruction",
                        "abstract": "We propose a deep learning method for 3D volumetric reconstruction in low-dose helical cone-beam computed tomography. Prior machine learning approaches require reference reconstructions computed by another algorithm for training. In contrast, we train our model in a fully self-supervised manner using only noisy 2D X-ray data. This is enabled by incorporating a fast differentiable CT simulator in the training loop. As we do not rely on reference reconstructions, the fidelity of our results is not limited by their potential shortcomings. We evaluate our method on real helical cone-beam projections and simulated phantoms. Our results show significantly higher visual fidelity and better PSNR over techniques that rely on existing reconstructions. When applied to full-dose data, our method produces high-quality results orders of magnitude faster than iterative techniques."
                    },
                    {
                        "title": "StyleGAN-T: Unlocking the Power of GANs for Fast Large-Scale Text-to-Image Synthesis",
                        "abstract": "Text-to-image synthesis has recently seen significant progress thanks to large pretrained language models, large-scale training data, and the introduction of scalable model families such as diffusion and autoregressive models. However, the best-performing models require iterative evaluation to generate a single sample. In contrast, generative adversarial networks (GANs) only need a single forward pass. They are thus much faster, but they currently remain far behind the state-of-the-art in large-scale text-to-image synthesis. This paper aims to identify the necessary steps to regain competitiveness. Our proposed model, StyleGAN-T, addresses the specific requirements of large-scale text-to-image synthesis, such as large capacity, stable training on diverse datasets, strong text alignment, and controllable variation vs. text alignment tradeoff. StyleGAN-T significantly improves over previous GANs and outperforms distilled diffusion models - the previous state-of-the-art in fast text-to-image synthesis - in terms of sample quality and speed."
                    },
                    {
                        "title": "Noise2Noise: Learning Image Restoration without Clean Data",
                        "abstract": "We apply basic statistical reasoning to signal reconstruction by machine learning -- learning to map corrupted observations to clean signals -- with a simple and powerful conclusion: it is possible to learn to restore images by only looking at corrupted examples, at performance at and sometimes exceeding training using clean data, without explicit image priors or likelihood models of the corruption. In practice, we show that a single model learns photographic noise removal, denoising synthetic Monte Carlo images, and reconstruction of undersampled MRI scans -- all corrupted by different processes -- based on noisy data only."
                    },
                    {
                        "title": "Training Generative Adversarial Networks with Limited Data",
                        "abstract": "Training generative adversarial networks (GAN) using too little data typically leads to discriminator overfitting, causing training to diverge. We propose an adaptive discriminator augmentation mechanism that significantly stabilizes training in limited data regimes. The approach does not require changes to loss functions or network architectures, and is applicable both when training from scratch and when fine-tuning an existing GAN on another dataset. We demonstrate, on several datasets, that good results are now possible using only a few thousand training images, often matching StyleGAN2 results with an order of magnitude fewer images. We expect this to open up new application domains for GANs. We also find that the widely used CIFAR-10 is, in fact, a limited data benchmark, and improve the record FID from 5.59 to 2.42."
                    },
                    {
                        "title": "Analyzing and Improving the Image Quality of StyleGAN",
                        "abstract": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign the generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent codes to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably attribute a generated image to a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality."
                    },
                    {
                        "title": "Modular Primitives for High-Performance Differentiable Rendering",
                        "abstract": "We present a modular differentiable renderer design that yields performance superior to previous methods by leveraging existing, highly optimized hardware graphics pipelines. Our design supports all crucial operations in a modern graphics pipeline: rasterizing large numbers of triangles, attribute interpolation, filtered texture lookups, as well as user-programmable shading and geometry processing, all in high resolutions. Our modular primitives allow custom, high-performance graphics pipelines to be built directly within automatic differentiation frameworks such as PyTorch or TensorFlow. As a motivating application, we formulate facial performance capture as an inverse rendering problem and show that it can be solved efficiently using our tools. Our results indicate that this simple and straightforward approach achieves excellent geometric correspondence between rendered results and reference imagery."
                    },
                    {
                        "title": "Alias-Free Generative Adversarial Networks",
                        "abstract": "We observe that despite their hierarchical convolutional nature, the synthesis process of typical generative adversarial networks depends on absolute pixel coordinates in an unhealthy manner. This manifests itself as, e.g., detail appearing to be glued to image coordinates instead of the surfaces of depicted objects. We trace the root cause to careless signal processing that causes aliasing in the generator network. Interpreting all signals in the network as continuous, we derive generally applicable, small architectural changes that guarantee that unwanted information cannot leak into the hierarchical synthesis process. The resulting networks match the FID of StyleGAN2 but differ dramatically in their internal representations, and they are fully equivariant to translation and rotation even at subpixel scales. Our results pave the way for generative models better suited for video and animation."
                    },
                    {
                        "title": "Analyzing and Improving the Training Dynamics of Diffusion Models",
                        "abstract": "Diffusion models currently dominate the field of data-driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high-level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on expectation. We find that systematic application of this philosophy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational complexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling.   As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance."
                    },
                    {
                        "title": "Guiding a Diffusion Model with a Bad Version of Itself",
                        "abstract": "The primary axes of interest in image-generating diffusion models are image quality, the amount of variation in the results, and how well the results align with a given condition, e.g., a class label or a text prompt. The popular classifier-free guidance approach uses an unconditional model to guide a conditional model, leading to simultaneously better prompt alignment and higher-quality images at the cost of reduced variation. These effects seem inherently entangled, and thus hard to control. We make the surprising observation that it is possible to obtain disentangled control over image quality without compromising the amount of variation by guiding generation using a smaller, less-trained version of the model itself rather than an unconditional model. This leads to significant improvements in ImageNet generation, setting record FIDs of 1.01 for 64x64 and 1.25 for 512x512, using publicly available networks. Furthermore, the method is also applicable to unconditional diffusion models, drastically improving their quality."
                    }
                ]
            },
            "e0af7d4a-bdde-4f2d-9027-5f4507275a1e": {
                "pk": "e0af7d4a-bdde-4f2d-9027-5f4507275a1e",
                "name": "Samuli Laine",
                "collaborators": [
                    "Timo Aila",
                    "Tero Karras",
                    "Jaakko Lehtinen",
                    "Miika Aittala",
                    "Janne Hellsten",
                    "Tuomas Kynk\u00e4\u00e4nniemi",
                    "Jon Hasselgren",
                    "Jacob Munkberg",
                    "Onni Kosomaa",
                    "Axel Sauer"
                ],
                "domain": [
                    "Generative Models",
                    "Semi-Supervised Learning",
                    "Image Synthesis",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Temporal Ensembling for Semi-Supervised Learning",
                        "abstract": "In this paper, we present a simple and efficient method for training deep neural networks in a semi-supervised setting where only a small portion of training data is labeled. We introduce self-ensembling, where we form a consensus prediction of the unknown labels using the outputs of the network-in-training on different epochs, and most importantly, under different regularization and input augmentation conditions. This ensemble prediction can be expected to be a better predictor for the unknown labels than the output of the network at the most recent training epoch, and can thus be used as a target for training. Using our method, we set new records for two standard semi-supervised learning benchmarks, reducing the (non-augmented) classification error rate from 18.44% to 7.05% in SVHN with 500 labels and from 18.63% to 16.55% in CIFAR-10 with 4000 labels, and further to 5.12% and 12.16% by enabling the standard augmentations. We additionally obtain a clear improvement in CIFAR-100 classification accuracy by using random images from the Tiny Images dataset as unlabeled extra inputs during training. Finally, we demonstrate good tolerance to incorrect labels."
                    },
                    {
                        "title": "Progressive Growing of GANs for Improved Quality, Stability, and Variation",
                        "abstract": "We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024^2. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10. Additionally, we describe several implementation details that are important for discouraging unhealthy competition between the generator and discriminator. Finally, we suggest a new metric for evaluating GAN results, both in terms of image quality and variation. As an additional contribution, we construct a higher-quality version of the CelebA dataset."
                    },
                    {
                        "title": "High-Quality Self-Supervised Deep Image Denoising",
                        "abstract": "We describe a novel method for training high-quality image denoising models based on unorganized collections of corrupted images. The training does not need access to clean reference images, or explicit pairs of corrupted images, and can thus be applied in situations where such data is unacceptably expensive or impossible to acquire. We build on a recent technique that removes the need for reference data by employing networks with a \"blind spot\" in the receptive field, and significantly improve two key aspects: image quality and training efficiency. Our result quality is on par with state-of-the-art neural network denoisers in the case of i.i.d. additive Gaussian noise, and not far behind with Poisson and impulse noise. We also successfully handle cases where parameters of the noise model are variable and/or unknown in both training and evaluation data."
                    },
                    {
                        "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                        "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                    },
                    {
                        "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                        "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                    },
                    {
                        "title": "Improved Precision and Recall Metric for Assessing Generative Models",
                        "abstract": "The ability to automatically estimate the quality and coverage of the samples produced by a generative model is a vital requirement for driving algorithm research. We present an evaluation metric that can separately and reliably measure both of these aspects in image generation tasks by forming explicit, non-parametric representations of the manifolds of real and generated data. We demonstrate the effectiveness of our metric in StyleGAN and BigGAN by providing several illustrative examples where existing metrics yield uninformative or contradictory results. Furthermore, we analyze multiple design variants of StyleGAN to better understand the relationships between the model architecture, training methods, and the properties of the resulting sample distribution. In the process, we identify new variants that improve the state-of-the-art. We also perform the first principled analysis of truncation methods and identify an improved method. Finally, we extend our metric to estimate the perceptual quality of individual samples, and use this to study latent space interpolations."
                    },
                    {
                        "title": "Simulator-Based Self-Supervision for Learned 3D Tomography Reconstruction",
                        "abstract": "We propose a deep learning method for 3D volumetric reconstruction in low-dose helical cone-beam computed tomography. Prior machine learning approaches require reference reconstructions computed by another algorithm for training. In contrast, we train our model in a fully self-supervised manner using only noisy 2D X-ray data. This is enabled by incorporating a fast differentiable CT simulator in the training loop. As we do not rely on reference reconstructions, the fidelity of our results is not limited by their potential shortcomings. We evaluate our method on real helical cone-beam projections and simulated phantoms. Our results show significantly higher visual fidelity and better PSNR over techniques that rely on existing reconstructions. When applied to full-dose data, our method produces high-quality results orders of magnitude faster than iterative techniques."
                    },
                    {
                        "title": "StyleGAN-T: Unlocking the Power of GANs for Fast Large-Scale Text-to-Image Synthesis",
                        "abstract": "Text-to-image synthesis has recently seen significant progress thanks to large pretrained language models, large-scale training data, and the introduction of scalable model families such as diffusion and autoregressive models. However, the best-performing models require iterative evaluation to generate a single sample. In contrast, generative adversarial networks (GANs) only need a single forward pass. They are thus much faster, but they currently remain far behind the state-of-the-art in large-scale text-to-image synthesis. This paper aims to identify the necessary steps to regain competitiveness. Our proposed model, StyleGAN-T, addresses the specific requirements of large-scale text-to-image synthesis, such as large capacity, stable training on diverse datasets, strong text alignment, and controllable variation vs. text alignment tradeoff. StyleGAN-T significantly improves over previous GANs and outperforms distilled diffusion models - the previous state-of-the-art in fast text-to-image synthesis - in terms of sample quality and speed."
                    },
                    {
                        "title": "Appearance-Driven Automatic 3D Model Simplification",
                        "abstract": "We present a suite of techniques for jointly optimizing triangle meshes and shading models to match the appearance of reference scenes. This capability has a number of uses, including appearance-preserving simplification of extremely complex assets, conversion between rendering systems, and even conversion between geometric scene representations.   We follow and extend the classic analysis-by-synthesis family of techniques: enabled by a highly efficient differentiable renderer and modern nonlinear optimization algorithms, our results are driven to minimize the image-space difference to the target scene when rendered in similar viewing and lighting conditions. As the only signals driving the optimization are differences in rendered images, the approach is highly general and versatile: it easily supports many different forward rendering models such as normal mapping, spatially-varying BRDFs, displacement mapping, etc. Supervision through images only is also key to the ability to easily convert between rendering systems and scene representations.   We output triangle meshes with textured materials to ensure that the models render efficiently on modern graphics hardware and benefit from, e.g., hardware-accelerated rasterization, ray tracing, and filtered texture lookups. Our system is integrated in a small Python code base, and can be applied at high resolutions and on large models. We describe several use cases, including mesh decimation, level of detail generation, seamless mesh filtering and approximations of aggregate geometry."
                    },
                    {
                        "title": "Noise2Noise: Learning Image Restoration without Clean Data",
                        "abstract": "We apply basic statistical reasoning to signal reconstruction by machine learning -- learning to map corrupted observations to clean signals -- with a simple and powerful conclusion: it is possible to learn to restore images by only looking at corrupted examples, at performance at and sometimes exceeding training using clean data, without explicit image priors or likelihood models of the corruption. In practice, we show that a single model learns photographic noise removal, denoising synthetic Monte Carlo images, and reconstruction of undersampled MRI scans -- all corrupted by different processes -- based on noisy data only."
                    },
                    {
                        "title": "Training Generative Adversarial Networks with Limited Data",
                        "abstract": "Training generative adversarial networks (GAN) using too little data typically leads to discriminator overfitting, causing training to diverge. We propose an adaptive discriminator augmentation mechanism that significantly stabilizes training in limited data regimes. The approach does not require changes to loss functions or network architectures, and is applicable both when training from scratch and when fine-tuning an existing GAN on another dataset. We demonstrate, on several datasets, that good results are now possible using only a few thousand training images, often matching StyleGAN2 results with an order of magnitude fewer images. We expect this to open up new application domains for GANs. We also find that the widely used CIFAR-10 is, in fact, a limited data benchmark, and improve the record FID from 5.59 to 2.42."
                    },
                    {
                        "title": "Analyzing and Improving the Image Quality of StyleGAN",
                        "abstract": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign the generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent codes to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably attribute a generated image to a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality."
                    },
                    {
                        "title": "Modular Primitives for High-Performance Differentiable Rendering",
                        "abstract": "We present a modular differentiable renderer design that yields performance superior to previous methods by leveraging existing, highly optimized hardware graphics pipelines. Our design supports all crucial operations in a modern graphics pipeline: rasterizing large numbers of triangles, attribute interpolation, filtered texture lookups, as well as user-programmable shading and geometry processing, all in high resolutions. Our modular primitives allow custom, high-performance graphics pipelines to be built directly within automatic differentiation frameworks such as PyTorch or TensorFlow. As a motivating application, we formulate facial performance capture as an inverse rendering problem and show that it can be solved efficiently using our tools. Our results indicate that this simple and straightforward approach achieves excellent geometric correspondence between rendered results and reference imagery."
                    },
                    {
                        "title": "Alias-Free Generative Adversarial Networks",
                        "abstract": "We observe that despite their hierarchical convolutional nature, the synthesis process of typical generative adversarial networks depends on absolute pixel coordinates in an unhealthy manner. This manifests itself as, e.g., detail appearing to be glued to image coordinates instead of the surfaces of depicted objects. We trace the root cause to careless signal processing that causes aliasing in the generator network. Interpreting all signals in the network as continuous, we derive generally applicable, small architectural changes that guarantee that unwanted information cannot leak into the hierarchical synthesis process. The resulting networks match the FID of StyleGAN2 but differ dramatically in their internal representations, and they are fully equivariant to translation and rotation even at subpixel scales. Our results pave the way for generative models better suited for video and animation."
                    },
                    {
                        "title": "Analyzing and Improving the Training Dynamics of Diffusion Models",
                        "abstract": "Diffusion models currently dominate the field of data-driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high-level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on expectation. We find that systematic application of this philosophy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational complexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling.   As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance."
                    },
                    {
                        "title": "Guiding a Diffusion Model with a Bad Version of Itself",
                        "abstract": "The primary axes of interest in image-generating diffusion models are image quality, the amount of variation in the results, and how well the results align with a given condition, e.g., a class label or a text prompt. The popular classifier-free guidance approach uses an unconditional model to guide a conditional model, leading to simultaneously better prompt alignment and higher-quality images at the cost of reduced variation. These effects seem inherently entangled, and thus hard to control. We make the surprising observation that it is possible to obtain disentangled control over image quality without compromising the amount of variation by guiding generation using a smaller, less-trained version of the model itself rather than an unconditional model. This leads to significant improvements in ImageNet generation, setting record FIDs of 1.01 for 64x64 and 1.25 for 512x512, using publicly available networks. Furthermore, the method is also applicable to unconditional diffusion models, drastically improving their quality."
                    },
                    {
                        "title": "Semi-supervised semantic segmentation needs strong, varied perturbations",
                        "abstract": "Consistency regularization describes a class of approaches that have yielded ground breaking results in semi-supervised classification problems. Prior work has established the cluster assumption - under which the data distribution consists of uniform class clusters of samples separated by low density regions - as important to its success. We analyze the problem of semantic segmentation and find that its' distribution does not exhibit low density regions separating classes and offer this as an explanation for why semi-supervised segmentation is a challenging problem, with only a few reports of success. We then identify choice of augmentation as key to obtaining reliable performance without such low-density regions. We find that adapted variants of the recently proposed CutOut and CutMix augmentation techniques yield state-of-the-art semi-supervised semantic segmentation results in standard datasets. Furthermore, given its challenging nature we propose that semantic segmentation acts as an effective acid test for evaluating semi-supervised regularizers. Implementation at: https://github.com/Britefury/cutmix-semisup-seg."
                    }
                ]
            },
            "c9e9a6ce-dfa1-4a2d-b00b-9617ee4178a2": {
                "pk": "c9e9a6ce-dfa1-4a2d-b00b-9617ee4178a2",
                "name": "Timo Aila",
                "collaborators": [
                    "Samuli Laine",
                    "Tero Karras",
                    "Jaakko Lehtinen",
                    "Miika Aittala",
                    "Janne Hellsten",
                    "Tuomas Kynk\u00e4\u00e4nniemi",
                    "Erik H\u00e4rk\u00f6nen",
                    "Pavlo Molchanov",
                    "Stephen Tyree",
                    "Jan Kautz"
                ],
                "domain": [
                    "Generative Models",
                    "Semi-Supervised Learning",
                    "Image Synthesis",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Temporal Ensembling for Semi-Supervised Learning",
                        "abstract": "In this paper, we present a simple and efficient method for training deep neural networks in a semi-supervised setting where only a small portion of training data is labeled. We introduce self-ensembling, where we form a consensus prediction of the unknown labels using the outputs of the network-in-training on different epochs, and most importantly, under different regularization and input augmentation conditions. This ensemble prediction can be expected to be a better predictor for the unknown labels than the output of the network at the most recent training epoch, and can thus be used as a target for training. Using our method, we set new records for two standard semi-supervised learning benchmarks, reducing the (non-augmented) classification error rate from 18.44% to 7.05% in SVHN with 500 labels and from 18.63% to 16.55% in CIFAR-10 with 4000 labels, and further to 5.12% and 12.16% by enabling the standard augmentations. We additionally obtain a clear improvement in CIFAR-100 classification accuracy by using random images from the Tiny Images dataset as unlabeled extra inputs during training. Finally, we demonstrate good tolerance to incorrect labels."
                    },
                    {
                        "title": "Progressive Growing of GANs for Improved Quality, Stability, and Variation",
                        "abstract": "We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024^2. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10. Additionally, we describe several implementation details that are important for discouraging unhealthy competition between the generator and discriminator. Finally, we suggest a new metric for evaluating GAN results, both in terms of image quality and variation. As an additional contribution, we construct a higher-quality version of the CelebA dataset."
                    },
                    {
                        "title": "High-Quality Self-Supervised Deep Image Denoising",
                        "abstract": "We describe a novel method for training high-quality image denoising models based on unorganized collections of corrupted images. The training does not need access to clean reference images, or explicit pairs of corrupted images, and can thus be applied in situations where such data is unacceptably expensive or impossible to acquire. We build on a recent technique that removes the need for reference data by employing networks with a \"blind spot\" in the receptive field, and significantly improve two key aspects: image quality and training efficiency. Our result quality is on par with state-of-the-art neural network denoisers in the case of i.i.d. additive Gaussian noise, and not far behind with Poisson and impulse noise. We also successfully handle cases where parameters of the noise model are variable and/or unknown in both training and evaluation data."
                    },
                    {
                        "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                        "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                    },
                    {
                        "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                        "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                    },
                    {
                        "title": "Improved Precision and Recall Metric for Assessing Generative Models",
                        "abstract": "The ability to automatically estimate the quality and coverage of the samples produced by a generative model is a vital requirement for driving algorithm research. We present an evaluation metric that can separately and reliably measure both of these aspects in image generation tasks by forming explicit, non-parametric representations of the manifolds of real and generated data. We demonstrate the effectiveness of our metric in StyleGAN and BigGAN by providing several illustrative examples where existing metrics yield uninformative or contradictory results. Furthermore, we analyze multiple design variants of StyleGAN to better understand the relationships between the model architecture, training methods, and the properties of the resulting sample distribution. In the process, we identify new variants that improve the state-of-the-art. We also perform the first principled analysis of truncation methods and identify an improved method. Finally, we extend our metric to estimate the perceptual quality of individual samples, and use this to study latent space interpolations."
                    },
                    {
                        "title": "Pruning Convolutional Neural Networks for Resource Efficient Inference",
                        "abstract": "We propose a new formulation for pruning convolutional kernels in neural networks to enable efficient inference. We interleave greedy criteria-based pruning with fine-tuning by backpropagation - a computationally efficient procedure that maintains good generalization in the pruned network. We propose a new criterion based on Taylor expansion that approximates the change in the cost function induced by pruning network parameters. We focus on transfer learning, where large pretrained networks are adapted to specialized tasks. The proposed criterion demonstrates superior performance compared to other criteria, e.g. the norm of kernel weights or feature map activation, for pruning large CNNs after adaptation to fine-grained classification tasks (Birds-200 and Flowers-102) relaying only on the first order gradient information. We also show that pruning can lead to more than 10x theoretical (5x practical) reduction in adapted 3D-convolutional filters with a small drop in accuracy in a recurrent gesture classifier. Finally, we show results for the large-scale ImageNet dataset to emphasize the flexibility of our approach."
                    },
                    {
                        "title": "The Role of ImageNet Classes in Fr\u00e9chet Inception Distance",
                        "abstract": "Fr\\'echet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID \"looks at\" in generated images. We show that the feature space that FID is (typically) computed in is so close to the ImageNet classifications that aligning the histograms of Top-$N$ classifications between sets of generated and real images can reduce FID substantially -- without actually improving the quality of results. Thus, we conclude that FID is prone to intentional or accidental distortions. As a practical example of an accidental distortion, we discuss a case where an ImageNet pre-trained FastGAN achieves a FID comparable to StyleGAN2, while being worse in terms of human evaluation."
                    },
                    {
                        "title": "StyleGAN-T: Unlocking the Power of GANs for Fast Large-Scale Text-to-Image Synthesis",
                        "abstract": "Text-to-image synthesis has recently seen significant progress thanks to large pretrained language models, large-scale training data, and the introduction of scalable model families such as diffusion and autoregressive models. However, the best-performing models require iterative evaluation to generate a single sample. In contrast, generative adversarial networks (GANs) only need a single forward pass. They are thus much faster, but they currently remain far behind the state-of-the-art in large-scale text-to-image synthesis. This paper aims to identify the necessary steps to regain competitiveness. Our proposed model, StyleGAN-T, addresses the specific requirements of large-scale text-to-image synthesis, such as large capacity, stable training on diverse datasets, strong text alignment, and controllable variation vs. text alignment tradeoff. StyleGAN-T significantly improves over previous GANs and outperforms distilled diffusion models - the previous state-of-the-art in fast text-to-image synthesis - in terms of sample quality and speed."
                    },
                    {
                        "title": "Semi-supervised semantic segmentation needs strong, varied perturbations",
                        "abstract": "Consistency regularization describes a class of approaches that have yielded ground breaking results in semi-supervised classification problems. Prior work has established the cluster assumption - under which the data distribution consists of uniform class clusters of samples separated by low density regions - as important to its success. We analyze the problem of semantic segmentation and find that its' distribution does not exhibit low density regions separating classes and offer this as an explanation for why semi-supervised segmentation is a challenging problem, with only a few reports of success. We then identify choice of augmentation as key to obtaining reliable performance without such low-density regions. We find that adapted variants of the recently proposed CutOut and CutMix augmentation techniques yield state-of-the-art semi-supervised semantic segmentation results in standard datasets. Furthermore, given its challenging nature we propose that semantic segmentation acts as an effective acid test for evaluating semi-supervised regularizers. Implementation at: https://github.com/Britefury/cutmix-semisup-seg."
                    },
                    {
                        "title": "Noise2Noise: Learning Image Restoration without Clean Data",
                        "abstract": "We apply basic statistical reasoning to signal reconstruction by machine learning -- learning to map corrupted observations to clean signals -- with a simple and powerful conclusion: it is possible to learn to restore images by only looking at corrupted examples, at performance at and sometimes exceeding training using clean data, without explicit image priors or likelihood models of the corruption. In practice, we show that a single model learns photographic noise removal, denoising synthetic Monte Carlo images, and reconstruction of undersampled MRI scans -- all corrupted by different processes -- based on noisy data only."
                    },
                    {
                        "title": "Training Generative Adversarial Networks with Limited Data",
                        "abstract": "Training generative adversarial networks (GAN) using too little data typically leads to discriminator overfitting, causing training to diverge. We propose an adaptive discriminator augmentation mechanism that significantly stabilizes training in limited data regimes. The approach does not require changes to loss functions or network architectures, and is applicable both when training from scratch and when fine-tuning an existing GAN on another dataset. We demonstrate, on several datasets, that good results are now possible using only a few thousand training images, often matching StyleGAN2 results with an order of magnitude fewer images. We expect this to open up new application domains for GANs. We also find that the widely used CIFAR-10 is, in fact, a limited data benchmark, and improve the record FID from 5.59 to 2.42."
                    },
                    {
                        "title": "Analyzing and Improving the Image Quality of StyleGAN",
                        "abstract": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign the generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent codes to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably attribute a generated image to a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality."
                    },
                    {
                        "title": "Modular Primitives for High-Performance Differentiable Rendering",
                        "abstract": "We present a modular differentiable renderer design that yields performance superior to previous methods by leveraging existing, highly optimized hardware graphics pipelines. Our design supports all crucial operations in a modern graphics pipeline: rasterizing large numbers of triangles, attribute interpolation, filtered texture lookups, as well as user-programmable shading and geometry processing, all in high resolutions. Our modular primitives allow custom, high-performance graphics pipelines to be built directly within automatic differentiation frameworks such as PyTorch or TensorFlow. As a motivating application, we formulate facial performance capture as an inverse rendering problem and show that it can be solved efficiently using our tools. Our results indicate that this simple and straightforward approach achieves excellent geometric correspondence between rendered results and reference imagery."
                    },
                    {
                        "title": "Alias-Free Generative Adversarial Networks",
                        "abstract": "We observe that despite their hierarchical convolutional nature, the synthesis process of typical generative adversarial networks depends on absolute pixel coordinates in an unhealthy manner. This manifests itself as, e.g., detail appearing to be glued to image coordinates instead of the surfaces of depicted objects. We trace the root cause to careless signal processing that causes aliasing in the generator network. Interpreting all signals in the network as continuous, we derive generally applicable, small architectural changes that guarantee that unwanted information cannot leak into the hierarchical synthesis process. The resulting networks match the FID of StyleGAN2 but differ dramatically in their internal representations, and they are fully equivariant to translation and rotation even at subpixel scales. Our results pave the way for generative models better suited for video and animation."
                    },
                    {
                        "title": "Analyzing and Improving the Training Dynamics of Diffusion Models",
                        "abstract": "Diffusion models currently dominate the field of data-driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high-level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on expectation. We find that systematic application of this philosophy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational complexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling.   As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance."
                    },
                    {
                        "title": "Guiding a Diffusion Model with a Bad Version of Itself",
                        "abstract": "The primary axes of interest in image-generating diffusion models are image quality, the amount of variation in the results, and how well the results align with a given condition, e.g., a class label or a text prompt. The popular classifier-free guidance approach uses an unconditional model to guide a conditional model, leading to simultaneously better prompt alignment and higher-quality images at the cost of reduced variation. These effects seem inherently entangled, and thus hard to control. We make the surprising observation that it is possible to obtain disentangled control over image quality without compromising the amount of variation by guiding generation using a smaller, less-trained version of the model itself rather than an unconditional model. This leads to significant improvements in ImageNet generation, setting record FIDs of 1.01 for 64x64 and 1.25 for 512x512, using publicly available networks. Furthermore, the method is also applicable to unconditional diffusion models, drastically improving their quality."
                    },
                    {
                        "title": "Disentangling Random and Cyclic Effects in Time-Lapse Sequences",
                        "abstract": "Time-lapse image sequences offer visually compelling insights into dynamic processes that are too slow to observe in real time. However, playing a long time-lapse sequence back as a video often results in distracting flicker due to random effects, such as weather, as well as cyclic effects, such as the day-night cycle. We introduce the problem of disentangling time-lapse sequences in a way that allows separate, after-the-fact control of overall trends, cyclic effects, and random effects in the images, and describe a technique based on data-driven generative models that achieves this goal. This enables us to \"re-render\" the sequences in ways that would not be possible with the input images alone. For example, we can stabilize a long sequence to focus on plant growth over many months, under selectable, consistent weather.   Our approach is based on Generative Adversarial Networks (GAN) that are conditioned with the time coordinate of the time-lapse sequence. Our architecture and training procedure are designed so that the networks learn to model random variations, such as weather, using the GAN's latent space, and to disentangle overall trends and cyclic variations by feeding the conditioning time label to the model using Fourier features with specific frequencies.   We show that our models are robust to defects in the training data, enabling us to amend some of the practical difficulties in capturing long time-lapse sequences, such as temporary occlusions, uneven frame spacing, and missing frames."
                    },
                    {
                        "title": "Production-Level Facial Performance Capture Using Deep Convolutional Neural Networks",
                        "abstract": "We present a real-time deep learning framework for video-based facial performance capture -- the dense 3D tracking of an actor's face given a monocular video. Our pipeline begins with accurately capturing a subject using a high-end production facial capture pipeline based on multi-view stereo tracking and artist-enhanced animations. With 5-10 minutes of captured footage, we train a convolutional neural network to produce high-quality output, including self-occluded regions, from a monocular video sequence of that subject. Since this 3D facial performance capture is fully automated, our system can drastically reduce the amount of labor involved in the development of modern narrative-driven video games or films involving realistic digital doubles of actors and potentially hours of animated dialogue per character. We compare our results with several state-of-the-art monocular real-time facial capture techniques and demonstrate compelling animation inference in challenging areas such as eyes and lips."
                    }
                ]
            },
            "4a9165cd-829f-4305-acc6-87f41bc47f66": {
                "pk": "4a9165cd-829f-4305-acc6-87f41bc47f66",
                "name": "Jaakko Lehtinen",
                "collaborators": [
                    "Tero Karras",
                    "Samuli Laine",
                    "Timo Aila",
                    "Miika Aittala",
                    "Erik H\u00e4rk\u00f6nen",
                    "Tuomas Kynk\u00e4\u00e4nniemi",
                    "Markus Kettunen",
                    "Aaron Hertzmann",
                    "Sylvain Paris",
                    "Eloi Moliner"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Image Processing",
                    "Reinforcement Learning",
                    "Audio Processing"
                ],
                "publications": [
                    {
                        "title": "E-LPIPS: Robust Perceptual Image Similarity via Random Transformation Ensembles",
                        "abstract": "It has been recently shown that the hidden variables of convolutional neural networks make for an efficient perceptual similarity metric that accurately predicts human judgment on relative image similarity assessment. First, we show that such learned perceptual similarity metrics (LPIPS) are susceptible to adversarial attacks that dramatically contradict human visual similarity judgment. While this is not surprising in light of neural networks' well-known weakness to adversarial perturbations, we proceed to show that self-ensembling with an infinite family of random transformations of the input --- a technique known not to render classification networks robust --- is enough to turn the metric robust against attack, while retaining predictive power on human judgments. Finally, we study the geometry imposed by our our novel self-ensembled metric (E-LPIPS) on the space of natural images. We find evidence of \"perceptual convexity\" by showing that convex combinations of similar-looking images retain appearance, and that discrete geodesics yield meaningful frame interpolation and texture morphing, all without explicit correspondences."
                    },
                    {
                        "title": "Progressive Growing of GANs for Improved Quality, Stability, and Variation",
                        "abstract": "We describe a new training methodology for generative adversarial networks. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. This both speeds the training up and greatly stabilizes it, allowing us to produce images of unprecedented quality, e.g., CelebA images at 1024^2. We also propose a simple way to increase the variation in generated images, and achieve a record inception score of 8.80 in unsupervised CIFAR10. Additionally, we describe several implementation details that are important for discouraging unhealthy competition between the generator and discriminator. Finally, we suggest a new metric for evaluating GAN results, both in terms of image quality and variation. As an additional contribution, we construct a higher-quality version of the CelebA dataset."
                    },
                    {
                        "title": "High-Quality Self-Supervised Deep Image Denoising",
                        "abstract": "We describe a novel method for training high-quality image denoising models based on unorganized collections of corrupted images. The training does not need access to clean reference images, or explicit pairs of corrupted images, and can thus be applied in situations where such data is unacceptably expensive or impossible to acquire. We build on a recent technique that removes the need for reference data by employing networks with a \"blind spot\" in the receptive field, and significantly improve two key aspects: image quality and training efficiency. Our result quality is on par with state-of-the-art neural network denoisers in the case of i.i.d. additive Gaussian noise, and not far behind with Poisson and impulse noise. We also successfully handle cases where parameters of the noise model are variable and/or unknown in both training and evaluation data."
                    },
                    {
                        "title": "GANSpace: Discovering Interpretable GAN Controls",
                        "abstract": "This paper describes a simple technique to analyze Generative Adversarial Networks (GANs) and create interpretable controls for image synthesis, such as change of viewpoint, aging, lighting, and time of day. We identify important latent directions based on Principal Components Analysis (PCA) applied either in latent space or feature space. Then, we show that a large number of interpretable controls can be defined by layer-wise perturbation along the principal directions. Moreover, we show that BigGAN can be controlled with layer-wise inputs in a StyleGAN-like manner. We show results on different GANs trained on various datasets, and demonstrate good qualitative matches to edit directions found through earlier supervised approaches."
                    },
                    {
                        "title": "Solving Audio Inverse Problems with a Diffusion Model",
                        "abstract": "This paper presents CQT-Diff, a data-driven generative audio model that can, once trained, be used for solving various different audio inverse problems in a problem-agnostic setting. CQT-Diff is a neural diffusion model with an architecture that is carefully constructed to exploit pitch-equivariant symmetries in music. This is achieved by preconditioning the model with an invertible Constant-Q Transform (CQT), whose logarithmically-spaced frequency axis represents pitch equivariance as translation equivariance. The proposed method is evaluated with objective and subjective metrics in three different and varied tasks: audio bandwidth extension, inpainting, and declipping. The results show that CQT-Diff outperforms the compared baselines and ablations in audio bandwidth extension and, without retraining, delivers competitive performance against modern baselines in audio inpainting and declipping. This work represents the first diffusion-based general framework for solving inverse problems in audio processing."
                    },
                    {
                        "title": "PPO-CMA: Proximal Policy Optimization with Covariance Matrix Adaptation",
                        "abstract": "Proximal Policy Optimization (PPO) is a highly popular model-free reinforcement learning (RL) approach. However, we observe that in a continuous action space, PPO can prematurely shrink the exploration variance, which leads to slow progress and may make the algorithm prone to getting stuck in local optima. Drawing inspiration from CMA-ES, a black-box evolutionary optimization method designed for robustness in similar situations, we propose PPO-CMA, a proximal policy optimization approach that adaptively expands the exploration variance to speed up progress. With only minor changes to PPO, our algorithm considerably improves performance in Roboschool continuous control benchmarks. Our results also show that PPO-CMA, as opposed to PPO, is significantly less sensitive to the choice of hyperparameters, allowing one to use it in complex movement optimization tasks without requiring tedious tuning."
                    },
                    {
                        "title": "Improved Precision and Recall Metric for Assessing Generative Models",
                        "abstract": "The ability to automatically estimate the quality and coverage of the samples produced by a generative model is a vital requirement for driving algorithm research. We present an evaluation metric that can separately and reliably measure both of these aspects in image generation tasks by forming explicit, non-parametric representations of the manifolds of real and generated data. We demonstrate the effectiveness of our metric in StyleGAN and BigGAN by providing several illustrative examples where existing metrics yield uninformative or contradictory results. Furthermore, we analyze multiple design variants of StyleGAN to better understand the relationships between the model architecture, training methods, and the properties of the resulting sample distribution. In the process, we identify new variants that improve the state-of-the-art. We also perform the first principled analysis of truncation methods and identify an improved method. Finally, we extend our metric to estimate the perceptual quality of individual samples, and use this to study latent space interpolations."
                    },
                    {
                        "title": "The Role of ImageNet Classes in Fr\u00e9chet Inception Distance",
                        "abstract": "Fr\\'echet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID \"looks at\" in generated images. We show that the feature space that FID is (typically) computed in is so close to the ImageNet classifications that aligning the histograms of Top-$N$ classifications between sets of generated and real images can reduce FID substantially -- without actually improving the quality of results. Thus, we conclude that FID is prone to intentional or accidental distortions. As a practical example of an accidental distortion, we discuss a case where an ImageNet pre-trained FastGAN achieves a FID comparable to StyleGAN2, while being worse in terms of human evaluation."
                    },
                    {
                        "title": "Simulator-Based Self-Supervision for Learned 3D Tomography Reconstruction",
                        "abstract": "We propose a deep learning method for 3D volumetric reconstruction in low-dose helical cone-beam computed tomography. Prior machine learning approaches require reference reconstructions computed by another algorithm for training. In contrast, we train our model in a fully self-supervised manner using only noisy 2D X-ray data. This is enabled by incorporating a fast differentiable CT simulator in the training loop. As we do not rely on reference reconstructions, the fidelity of our results is not limited by their potential shortcomings. We evaluate our method on real helical cone-beam projections and simulated phantoms. Our results show significantly higher visual fidelity and better PSNR over techniques that rely on existing reconstructions. When applied to full-dose data, our method produces high-quality results orders of magnitude faster than iterative techniques."
                    },
                    {
                        "title": "Appearance-Driven Automatic 3D Model Simplification",
                        "abstract": "We present a suite of techniques for jointly optimizing triangle meshes and shading models to match the appearance of reference scenes. This capability has a number of uses, including appearance-preserving simplification of extremely complex assets, conversion between rendering systems, and even conversion between geometric scene representations.   We follow and extend the classic analysis-by-synthesis family of techniques: enabled by a highly efficient differentiable renderer and modern nonlinear optimization algorithms, our results are driven to minimize the image-space difference to the target scene when rendered in similar viewing and lighting conditions. As the only signals driving the optimization are differences in rendered images, the approach is highly general and versatile: it easily supports many different forward rendering models such as normal mapping, spatially-varying BRDFs, displacement mapping, etc. Supervision through images only is also key to the ability to easily convert between rendering systems and scene representations.   We output triangle meshes with textured materials to ensure that the models render efficiently on modern graphics hardware and benefit from, e.g., hardware-accelerated rasterization, ray tracing, and filtered texture lookups. Our system is integrated in a small Python code base, and can be applied at high resolutions and on large models. We describe several use cases, including mesh decimation, level of detail generation, seamless mesh filtering and approximations of aggregate geometry."
                    }
                ]
            }
        }
    },
    "2309.16042": {
        "paper_data": {
            "title": "Towards Best Practices of Activation Patching in Language Models: Metrics and Methods",
            "url": "http://arxiv.org/abs/2309.16042v2",
            "arxiv_id": "2309.16042",
            "authors": [
                "Fred Zhang",
                "Neel Nanda"
            ],
            "abstract": "Mechanistic interpretability seeks to understand the internal mechanisms of machine learning models, where localization -- identifying the important model components -- is a key step. Activation patching, also known as causal tracing or interchange intervention, is a standard technique for this task (Vig et al., 2020), but the literature contains many variants with little consensus on the choice of hyperparameters or methodology. In this work, we systematically examine the impact of methodological details in activation patching, including evaluation metrics and corruption methods. In several settings of localization and circuit discovery in language models, we find that varying these hyperparameters could lead to disparate interpretability results. Backed by empirical observations, we give conceptual arguments for why certain metrics or methods may be preferred. Finally, we provide recommendations for the best practices of activation patching going forwards.",
            "introduction": "   1 Introduction  Mechanistic interpretability (MI) aims to unravel complex machine learning models by reverse engineering their internal mechanisms down to human-understandable algorithms (Geiger et\u00a0al., 2021; Olah, 2022; Wang et\u00a0al., 2023). With such understanding, we can better identify and fix model errors (Vig et\u00a0al., 2020; Hernandez et\u00a0al., 2021; Meng et\u00a0al., 2022; Hase et\u00a0al., 2023), steer model outputs (Li et\u00a0al., 2023b) and explain emergent behaviors (Nanda et\u00a0al., 2023a; Barak et\u00a0al., 2022).   A basic goal in MI is localization: identify the specific model components responsible for particular functions. Activation patching, also known as causal tracing, interchange intervention, causal mediation analysis or representation denoising, is a standard tool for localization in language models (Vig et\u00a0al., 2020; Meng et\u00a0al., 2022). The method attempts to pinpoint activations that causally affect on the output. Specifically, it involves 3 forward passes of the model: (1) on a clean prompt while caching the latent activations; (2) on a corrupted prompt; and (3) on the corrupted prompt but replacing the activation of a specific model component by its clean cache. For instance, the clean prompt can be \u201cThe Eiffel Tower is in\u201d and the corrupted one with the subject replaced by \u201cThe Colosseum\u201d. If the model outputs \u201cParis\u201d in step (3) but not in (2), then it suggests that the specific component being patched is important for producing the answer (Vig et\u00a0al., 2020; Pearl, 2001).   This technique has been widely applied for language model interpretability. For example, Meng et\u00a0al. (2022); Geva et\u00a0al. (2023) seek to understand which model weights store and process factual information. Wang et\u00a0al. (2023); Hanna et\u00a0al. (2023); Lieberum et\u00a0al. (2023) perform circuit analysis: identify the sub-network within a model\u2019s computation graph that implements a specified behavior. All these works leverage activation patching or its variants as a foundational technique.   Despite its broad applications across the literature, there is little consensus on the methodological details of activation patching. In particular, each paper tends to use its own method of generating corrupted prompts and the metric of evaluating patching effects. Concerningly, this lack of standardization leaves open the possibility that prior interpretability results may be highly sensitive to the hyperparameters they adopt. In this work, we study the impact of varying the metrics and methods in activation patching, as a step towards understanding best practices. To our knowledge, this is the first such systematic study of the technique.   Specifically, we identify three degrees of freedom in activation patching. First, we focus on the approach of generating corrupted prompts and evaluate two prominent methods from the literature:   \u2022  Gaussian noising (GN) adds a large Gaussian noise to the token embeddings of the tokens that contain the key information to completing a prompt, such as its subject (Meng et\u00a0al., 2022).     \u2022  Symmetric token replacement (STR) swaps these key tokens with semantically related ones; for example, \u201cThe Eiffel Tower\u201d\u2192\u2192\\rightarrow\u2192\u201cThe Colosseum\u201d (Vig et\u00a0al., 2020; Wang et\u00a0al., 2023).    Second, we examine the choice of metrics for measuring the effect of patching and compare probability and logit difference; both have found applications in the literature (Meng et\u00a0al., 2022; Wang et\u00a0al., 2023; Conmy et\u00a0al., 2023). Third, we study sliding window patching, which jointly restores the activations of multiple MLP layers, a technique used by Meng et\u00a0al. (2022);",
            "references": [
                {
                    "title": "Sparse Autoencoders Find Highly Interpretable Features in Language Models",
                    "abstract": "One of the roadblocks to a better understanding of neural networks' internals is \\textit{polysemanticity}, where neurons appear to activate in multiple, semantically distinct contexts. Polysemanticity prevents us from identifying concise, human-understandable explanations for what neural networks are doing internally. One hypothesised cause of polysemanticity is \\textit{superposition}, where neural networks represent more features than they have neurons by assigning features to an overcomplete set of directions in activation space, rather than to individual neurons. Here, we attempt to identify those directions, using sparse autoencoders to reconstruct the internal activations of a language model. These autoencoders learn sets of sparsely activating features that are more interpretable and monosemantic than directions identified by alternative approaches, where interpretability is measured by automated methods. Moreover, we show that with our learned set of features, we can pinpoint the features that are causally responsible for counterfactual behaviour on the indirect object identification task \\citep{wang2022interpretability} to a finer degree than previous decompositions. This work indicates that it is possible to resolve superposition in language models using a scalable, unsupervised method. Our method may serve as a foundation for future mechanistic interpretability work, which we hope will enable greater model transparency and steerability."
                },
                {
                    "title": "Explaining grokking through circuit efficiency",
                    "abstract": "One of the most surprising puzzles in neural network generalisation is grokking: a network with perfect training accuracy but poor generalisation will, upon further training, transition to perfect generalisation. We propose that grokking occurs when the task admits a generalising solution and a memorising solution, where the generalising solution is slower to learn but more efficient, producing larger logits with the same parameter norm. We hypothesise that memorising circuits become more inefficient with larger training datasets while generalising circuits do not, suggesting there is a critical dataset size at which memorisation and generalisation are equally efficient. We make and confirm four novel predictions about grokking, providing significant evidence in favour of our explanation. Most strikingly, we demonstrate two novel and surprising behaviours: ungrokking, in which a network regresses from perfect to low test accuracy, and semi-grokking, in which a network shows delayed generalisation to partial rather than perfect test accuracy."
                },
                {
                    "title": "Emergent Linear Representations in World Models of Self-Supervised Sequence Models",
                    "abstract": "How do sequence models represent their decision-making process? Prior work suggests that Othello-playing neural network learned nonlinear models of the board state (Li et al., 2023a). In this work, we provide evidence of a closely related linear representation of the board. In particular, we show that probing for \u201cmy colour\u201d vs. \u201copponent\u2019s colour\u201d may be a simple yet powerful way to interpret the model\u2019s internal state. This precise understanding of the internal representations allows us to control the model\u2019s behaviour with simple vector arithmetic. Linear representations enable significant interpretability progress, which we demonstrate with further exploration of how the world model is computed."
                },
                {
                    "title": "NeuroSurgeon: A Toolkit for Subnetwork Analysis",
                    "abstract": "Despite recent advances in the field of explainability, much remains unknown about the algorithms that neural networks learn to represent. Recent work has attempted to understand trained models by decomposing them into functional circuits (Csord\\'as et al., 2020; Lepori et al., 2023). To advance this research, we developed NeuroSurgeon, a python library that can be used to discover and manipulate subnetworks within models in the Huggingface Transformers library (Wolf et al., 2019). NeuroSurgeon is freely available at https://github.com/mlepori1/NeuroSurgeon."
                },
                {
                    "title": "The Hydra Effect: Emergent Self-repair in Language Model Computations",
                    "abstract": "We investigate the internal structure of language model computations using causal analysis and demonstrate two motifs: (1) a form of adaptive computation where ablations of one attention layer of a language model cause another layer to compensate (which we term the Hydra effect) and (2) a counterbalancing function of late MLP layers that act to downregulate the maximum-likelihood token. Our ablation studies demonstrate that language model layers are typically relatively loosely coupled (ablations to one layer only affect a small number of downstream layers). Surprisingly, these effects occur even in language models trained without any form of dropout. We analyse these effects in the context of factual recall and consider their implications for circuit-level attribution in language models."
                },
                {
                    "title": "Does Circuit Analysis Interpretability Scale? Evidence from Multiple Choice Capabilities in Chinchilla",
                    "abstract": "\\emph{Circuit analysis} is a promising technique for understanding the internal mechanisms of language models. However, existing analyses are done in small models far from the state of the art. To address this, we present a case study of circuit analysis in the 70B Chinchilla model, aiming to test the scalability of circuit analysis. In particular, we study multiple-choice question answering, and investigate Chinchilla's capability to identify the correct answer \\emph{label} given knowledge of the correct answer \\emph{text}. We find that the existing techniques of logit attribution, attention pattern visualization, and activation patching naturally scale to Chinchilla, allowing us to identify and categorize a small set of `output nodes' (attention heads and MLPs). We further study the `correct letter' category of attention heads aiming to understand the semantics of their features, with mixed results. For normal multiple-choice question answers, we significantly compress the query, key and value subspaces of the head without loss of performance when operating on the answer labels for multiple-choice questions, and we show that the query and key subspaces represent an `Nth item in an enumeration' feature to at least some extent. However, when we attempt to use this explanation to understand the heads' behaviour on a more general distribution including randomized answer labels, we find that it is only a partial explanation, suggesting there is more to learn about the operation of `correct letter' heads on multiple choice question answering."
                },
                {
                    "title": "The Clock and the Pizza: Two Stories in Mechanistic Explanation of Neural Networks",
                    "abstract": "Do neural networks, trained on well-understood algorithmic tasks, reliably rediscover known algorithms for solving those tasks? Several recent studies, on tasks ranging from group arithmetic to in-context linear regression, have suggested that the answer is yes. Using modular addition as a prototypical problem, we show that algorithm discovery in neural networks is sometimes more complex. Small changes to model hyperparameters and initializations can induce the discovery of qualitatively different algorithms from a fixed training set, and even parallel implementations of multiple such algorithms. Some networks trained to perform modular addition implement a familiar Clock algorithm; others implement a previously undescribed, less intuitive, but comprehensible procedure which we term the Pizza algorithm, or a variety of even more complex procedures. Our results show that even simple learning problems can admit a surprising diversity of solutions, motivating the development of new tools for characterizing the behavior of neural networks across their algorithmic phase space."
                },
                {
                    "title": "Inference-Time Intervention: Eliciting Truthful Answers from a Language Model",
                    "abstract": "We introduce Inference-Time Intervention (ITI), a technique designed to enhance the\"truthfulness\"of large language models (LLMs). ITI operates by shifting model activations during inference, following a set of directions across a limited number of attention heads. This intervention significantly improves the performance of LLaMA models on the TruthfulQA benchmark. On an instruction-finetuned LLaMA called Alpaca, ITI improves its truthfulness from 32.5% to 65.1%. We identify a tradeoff between truthfulness and helpfulness and demonstrate how to balance it by tuning the intervention strength. ITI is minimally invasive and computationally inexpensive. Moreover, the technique is data efficient: while approaches like RLHF require extensive annotations, ITI locates truthful directions using only few hundred examples. Our findings suggest that LLMs may have an internal representation of the likelihood of something being true, even as they produce falsehoods on the surface."
                },
                {
                    "title": "Language Models Implement Simple Word2Vec-style Vector Arithmetic",
                    "abstract": "A primary criticism towards language models (LMs) is their inscrutability. This paper presents evidence that, despite their size and complexity, LMs sometimes exploit a simple vector arithmetic style mechanism to solve some relational tasks using regularities encoded in the hidden space of the model (e.g., Poland:Warsaw::China:Beijing). We investigate a range of language model sizes (from 124M parameters to 176B parameters) in an in-context learning setting, and find that for a variety of tasks (involving capital cities, uppercasing, and past-tensing) a key part of the mechanism reduces to a simple additive update typically applied by the feedforward (FFN) networks. We further show that this mechanism is specific to tasks that require retrieval from pretraining memory, rather than retrieval from local context. Our results contribute to a growing body of work on the interpretability of LMs, and offer reason to be optimistic that, despite the massive and non-linear nature of the models, the strategies they ultimately use to solve tasks can sometimes reduce to familiar and even intuitive algorithms."
                },
                {
                    "title": "Interpretability at Scale: Identifying Causal Mechanisms in Alpaca",
                    "abstract": "Obtaining human-interpretable explanations of large, general-purpose language models is an urgent goal for AI safety. However, it is just as important that our interpretability methods are faithful to the causal dynamics underlying model behavior and able to robustly generalize to unseen inputs. Distributed Alignment Search (DAS) is a powerful gradient descent method grounded in a theory of causal abstraction that has uncovered perfect alignments between interpretable symbolic algorithms and small deep learning models fine-tuned for specific tasks. In the present paper, we scale DAS significantly by replacing the remaining brute-force search steps with learned parameters -- an approach we call Boundless DAS. This enables us to efficiently search for interpretable causal structure in large language models while they follow instructions. We apply Boundless DAS to the Alpaca model (7B parameters), which, off the shelf, solves a simple numerical reasoning problem. With Boundless DAS, we discover that Alpaca does this by implementing a causal model with two interpretable boolean variables. Furthermore, we find that the alignment of neural representations with these variables is robust to changes in inputs and instructions. These findings mark a first step toward faithfully understanding the inner-workings of our ever-growing and most widely deployed language models. Our tool is extensible to larger LLMs and is released publicly at `https://github.com/stanfordnlp/pyvene`."
                },
                {
                    "title": "Finding Neurons in a Haystack: Case Studies with Sparse Probing",
                    "abstract": "Despite rapid adoption and deployment of large language models (LLMs), the internal computations of these models remain opaque and poorly understood. In this work, we seek to understand how high-level human-interpretable features are represented within the internal neuron activations of LLMs. We train $k$-sparse linear classifiers (probes) on these internal activations to predict the presence of features in the input; by varying the value of $k$ we study the sparsity of learned representations and how this varies with model scale. With $k=1$, we localize individual neurons which are highly relevant for a particular feature, and perform a number of case studies to illustrate general properties of LLMs. In particular, we show that early layers make use of sparse combinations of neurons to represent many features in superposition, that middle layers have seemingly dedicated neurons to represent higher-level contextual features, and that increasing scale causes representational sparsity to increase on average, but there are multiple types of scaling dynamics. In all, we probe for over 100 unique features comprising 10 different categories in 7 different models spanning 70 million to 6.9 billion parameters."
                },
                {
                    "title": "How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language model",
                    "abstract": "Pre-trained language models can be surprisingly adept at tasks they were not explicitly trained on, but how they implement these capabilities is poorly understood. In this paper, we investigate the basic mathematical abilities often acquired by pre-trained language models. Concretely, we use mechanistic interpretability techniques to explain the (limited) mathematical abilities of GPT-2 small. As a case study, we examine its ability to take in sentences such as\"The war lasted from the year 1732 to the year 17\", and predict valid two-digit end years (years>32). We first identify a circuit, a small subset of GPT-2 small's computational graph that computes this task's output. Then, we explain the role of each circuit component, showing that GPT-2 small's final multi-layer perceptrons boost the probability of end years greater than the start year. Finally, we find related tasks that activate our circuit. Our results suggest that GPT-2 small computes greater-than using a complex but general mechanism that activates across diverse contexts."
                },
                {
                    "title": "Towards Automated Circuit Discovery for Mechanistic Interpretability",
                    "abstract": "Through considerable effort and intuition, several recent works have reverse-engineered nontrivial behaviors of transformer models. This paper systematizes the mechanistic interpretability process they followed. First, researchers choose a metric and dataset that elicit the desired model behavior. Then, they apply activation patching to find which abstract neural network units are involved in the behavior. By varying the dataset, metric, and units under investigation, researchers can understand the functionality of each component. We automate one of the process' steps: to identify the circuit that implements the specified behavior in the model's computational graph. We propose several algorithms and reproduce previous interpretability results to validate them. For example, the ACDC algorithm rediscovered 5/5 of the component types in a circuit in GPT-2 Small that computes the Greater-Than operation. ACDC selected 68 of the 32,000 edges in GPT-2 Small, all of which were manually found by previous work. Our code is available at https://github.com/ArthurConmy/Automatic-Circuit-Discovery."
                },
                {
                    "title": "Dissecting Recall of Factual Associations in Auto-Regressive Language Models",
                    "abstract": "Transformer-based language models (LMs) are known to capture factual knowledge in their parameters. While previous work looked into where factual associations are stored, only little is known about how they are retrieved internally during inference. We investigate this question through the lens of information flow. Given a subject-relation query, we study how the model aggregates information about the subject and relation to predict the correct attribute. With interventions on attention edges, we first identify two critical points where information propagates to the prediction: one from the relation positions followed by another from the subject positions. Next, by analyzing the information at these points, we unveil a three-step internal mechanism for attribute extraction. First, the representation at the last-subject position goes through an enrichment process, driven by the early MLP sublayers, to encode many subject-related attributes. Second, information from the relation propagates to the prediction. Third, the prediction representation\"queries\"the enriched subject to extract the attribute. Perhaps surprisingly, this extraction is typically done via attention heads, which often encode subject-attribute mappings in their parameters. Overall, our findings introduce a comprehensive view of how factual associations are stored and extracted internally in LMs, facilitating future research on knowledge localization and editing."
                },
                {
                    "title": "Localizing Model Behavior with Path Patching",
                    "abstract": "Localizing behaviors of neural networks to a subset of the network's components or a subset of interactions between components is a natural first step towards analyzing network mechanisms and possible failure modes. Existing work is often qualitative and ad-hoc, and there is no consensus on the appropriate way to evaluate localization claims. We introduce path patching, a technique for expressing and quantitatively testing a natural class of hypotheses expressing that behaviors are localized to a set of paths. We refine an explanation of induction heads, characterize a behavior of GPT-2, and open source a framework for efficiently running similar experiments."
                },
                {
                    "title": "How Do Transformers Learn Topic Structure: Towards a Mechanistic Understanding",
                    "abstract": "While the successes of transformers across many domains are indisputable, accurate understanding of the learning mechanics is still largely lacking. Their capabilities have been probed on benchmarks which include a variety of structured and reasoning tasks -- but mathematical understanding is lagging substantially behind. Recent lines of work have begun studying representational aspects of this question: that is, the size/depth/complexity of attention-based networks to perform certain tasks. However, there is no guarantee the learning dynamics will converge to the constructions proposed. In our paper, we provide fine-grained mechanistic understanding of how transformers learn\"semantic structure\", understood as capturing co-occurrence structure of words. Precisely, we show, through a combination of mathematical analysis and experiments on Wikipedia data and synthetic data modeled by Latent Dirichlet Allocation (LDA), that the embedding layer and the self-attention layer encode the topical structure. In the former case, this manifests as higher average inner product of embeddings between same-topic words. In the latter, it manifests as higher average pairwise attention between same-topic words. The mathematical results involve several assumptions to make the analysis tractable, which we verify on data, and might be of independent interest as well."
                },
                {
                    "title": "Finding Alignments Between Interpretable Causal Variables and Distributed Neural Representations",
                    "abstract": "Causal abstraction is a promising theoretical framework for explainable artificial intelligence that defines when an interpretable high-level causal model is a faithful simplification of a low-level deep learning system. However, existing causal abstraction methods have two major limitations: they require a brute-force search over alignments between the high-level model and the low-level one, and they presuppose that variables in the high-level model will align with disjoint sets of neurons in the low-level one. In this paper, we present distributed alignment search (DAS), which overcomes these limitations. In DAS, we find the alignment between high-level and low-level models using gradient descent rather than conducting a brute-force search, and we allow individual neurons to play multiple distinct roles by analyzing representations in non-standard bases-distributed representations. Our experiments show that DAS can discover internal structure that prior approaches miss. Overall, DAS removes previous obstacles to conducting causal abstraction analyses and allows us to find conceptual structure in trained neural nets."
                },
                {
                    "title": "A Toy Model of Universality: Reverse Engineering How Networks Learn Group Operations",
                    "abstract": "Universality is a key hypothesis in mechanistic interpretability -- that different models learn similar features and circuits when trained on similar tasks. In this work, we study the universality hypothesis by examining how small neural networks learn to implement group composition. We present a novel algorithm by which neural networks may implement composition for any finite group via mathematical representation theory. We then show that networks consistently learn this algorithm by reverse engineering model logits and weights, and confirm our understanding using ablations. By studying networks of differing architectures trained on various groups, we find mixed evidence for universality: using our algorithm, we can completely characterize the family of circuits and features that networks learn on this task, but for a given network the precise circuits learned -- as well as the order they develop -- are arbitrary."
                },
                {
                    "title": "Progress measures for grokking via mechanistic interpretability",
                    "abstract": "Neural networks often exhibit emergent behavior, where qualitatively new capabilities arise from scaling up the amount of parameters, training data, or training steps. One approach to understanding emergence is to find continuous \\textit{progress measures} that underlie the seemingly discontinuous qualitative changes. We argue that progress measures can be found via mechanistic interpretability: reverse-engineering learned behaviors into their individual components. As a case study, we investigate the recently-discovered phenomenon of ``grokking'' exhibited by small transformers trained on modular addition tasks. We fully reverse engineer the algorithm learned by these networks, which uses discrete Fourier transforms and trigonometric identities to convert addition to rotation about a circle. We confirm the algorithm by analyzing the activations and weights and by performing ablations in Fourier space. Based on this understanding, we define progress measures that allow us to study the dynamics of training and split training into three continuous phases: memorization, circuit formation, and cleanup. Our results show that grokking, rather than being a sudden shift, arises from the gradual amplification of structured mechanisms encoded in the weights, followed by the later removal of memorizing components."
                },
                {
                    "title": "Does Localization Inform Editing? Surprising Differences in Causality-Based Localization vs. Knowledge Editing in Language Models",
                    "abstract": "Language models learn a great quantity of factual information during pretraining, and recent work localizes this information to specific model weights like mid-layer MLP weights. In this paper, we find that we can change how a fact is stored in a model by editing weights that are in a different location than where existing methods suggest that the fact is stored. This is surprising because we would expect that localizing facts to specific model parameters would tell us where to manipulate knowledge in models, and this assumption has motivated past work on model editing methods. Specifically, we show that localization conclusions from representation denoising (also known as Causal Tracing) do not provide any insight into which model MLP layer would be best to edit in order to override an existing stored fact with a new one. This finding raises questions about how past work relies on Causal Tracing to select which model layers to edit. Next, we consider several variants of the editing problem, including erasing and amplifying facts. For one of our editing problems, editing performance does relate to localization results from representation denoising, but we find that which layer we edit is a far better predictor of performance. Our results suggest, counterintuitively, that better mechanistic understanding of how pretrained language models work may not always translate to insights about how to best change their behavior. Our code is available at https://github.com/google/belief-localization"
                },
                {
                    "title": "Rethinking the Role of Scale for In-Context Learning: An Interpretability-based Case Study at 66 Billion Scale",
                    "abstract": "Language models have been shown to perform better with an increase in scale on a wide variety of tasks via the in-context learning paradigm. In this paper, we investigate the hypothesis that the ability of a large language model to in-context learn-perform a task is not uniformly spread across all of its underlying components. Using a 66 billion parameter language model (OPT-66B) across a diverse set of 14 downstream tasks, we find this is indeed the case: ~70% of the attention heads and ~20% of the feed forward networks can be removed with minimal decline in task performance. We find substantial overlap in the set of attention heads (un)important for in-context learning across tasks and number of in-context examples. We also address our hypothesis through a task-agnostic lens, finding that a small set of attention heads in OPT-66B score highly on their ability to perform primitive induction operations associated with in-context learning, namely, prefix matching and copying. These induction heads overlap with task-specific important heads, reinforcing arguments by Olsson et al. (2022) regarding induction head generality to more sophisticated behaviors associated with in-context learning. Overall, our study provides several insights that indicate large language models may be under-trained for in-context learning and opens up questions on how to pre-train language models to more effectively perform in-context learning."
                },
                {
                    "title": "Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small",
                    "abstract": "Research in mechanistic interpretability seeks to explain behaviors of machine learning models in terms of their internal components. However, most previous work either focuses on simple behaviors in small models, or describes complicated behaviors in larger models with broad strokes. In this work, we bridge this gap by presenting an explanation for how GPT-2 small performs a natural language task called indirect object identification (IOI). Our explanation encompasses 26 attention heads grouped into 7 main classes, which we discovered using a combination of interpretability approaches relying on causal interventions. To our knowledge, this investigation is the largest end-to-end attempt at reverse-engineering a natural behavior\"in the wild\"in a language model. We evaluate the reliability of our explanation using three quantitative criteria--faithfulness, completeness and minimality. Though these criteria support our explanation, they also point to remaining gaps in our understanding. Our work provides evidence that a mechanistic understanding of large ML models is feasible, opening opportunities to scale our understanding to both larger models and more complex tasks."
                },
                {
                    "title": "Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task",
                    "abstract": "Language models show a surprising range of capabilities, but the source of their apparent competence is unclear. Do these networks just memorize a collection of surface statistics, or do they rely on internal representations of the process that generates the sequences they see? We investigate this question by applying a variant of the GPT model to the task of predicting legal moves in a simple board game, Othello. Although the network has no a priori knowledge of the game or its rules, we uncover evidence of an emergent nonlinear internal representation of the board state. Interventional experiments indicate this representation can be used to control the output of the network and create\"latent saliency maps\"that can help explain predictions in human terms."
                },
                {
                    "title": "Polysemanticity and Capacity in Neural Networks",
                    "abstract": "Individual neurons in neural networks often represent a mixture of unrelated features. This phenomenon, called polysemanticity, can make interpreting neural networks more difficult and so we aim to understand its causes. We propose doing so through the lens of feature \\emph{capacity}, which is the fractional dimension each feature consumes in the embedding space. We show that in a toy model the optimal capacity allocation tends to monosemantically represent the most important features, polysemantically represent less important features (in proportion to their impact on the loss), and entirely ignore the least important features. Polysemanticity is more prevalent when the inputs have higher kurtosis or sparsity and more prevalent in some architectures than others. Given an optimal allocation of capacity, we go on to study the geometry of the embedding space. We find a block-semi-orthogonal structure, with differing block sizes in different models, highlighting the impact of model architecture on the interpretability of its neurons."
                },
                {
                    "title": "In-context Learning and Induction Heads",
                    "abstract": "\u201cInduction heads\u201d are attention heads that implement a simple algorithm to complete token sequences like [A][B] ... [A] \u2192 [B]. In this work, we present preliminary and indirect evidence for a hypothesis that induction heads might constitute the mechanism for the majority of all \u201cincontext learning\u201d in large transformer models (i.e. decreasing loss at increasing token indices). We find that induction heads develop at precisely the same point as a sudden sharp increase in incontext learning ability, visible as a bump in the training loss. We present six complementary lines of evidence, arguing that induction heads may be the mechanistic source of general in-context learning in transformer models of any size. For small attention-only models, we present strong, causal evidence; for larger models with MLPs, we present correlational evidence. We recommend reading this paper as an HTML article. As Transformer generative models continue to scale and gain increasing real world use, addressing their associated safety problems becomes increasingly important. Mechanistic interpretability \u2013 attempting to reverse engineer the detailed computations performed by the model \u2013 offers one possible avenue for addressing these safety issues. If we can understand the internal structures that cause Transformer models to produce the outputs they do, then we may be able to address current safety problems more systematically, as well as anticipating safety problems in future more powerful models. [1, 2, 3, 4, 5]"
                },
                {
                    "title": "Analyzing Transformers in Embedding Space",
                    "abstract": "Understanding Transformer-based models has attracted significant attention, as they lie at the heart of recent technological advances across machine learning. While most interpretability methods rely on running models over inputs, recent work has shown that a zero-pass approach, where parameters are interpreted directly without a forward/backward pass is feasible for some Transformer parameters, and for two-layer attention networks. In this work, we present a theoretical analysis where all parameters of a trained Transformer are interpreted by projecting them into the embedding space, that is, the space of vocabulary items they operate on. We derive a simple theoretical framework to support our arguments and provide ample evidence for its validity. First, an empirical analysis showing that parameters of both pretrained and fine-tuned models can be interpreted in embedding space. Second, we present two applications of our framework: (a) aligning the parameters of different models that share a vocabulary, and (b) constructing a classifier without training by \u201ctranslating\u201d the parameters of a fine-tuned classifier to parameters of a different model that was only pretrained. Overall, our findings open the door to interpretation methods that, at least in part, abstract away from model specifics and operate in the embedding space only."
                },
                {
                    "title": "Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networks",
                    "abstract": "The last decade of machine learning has seen drastic increases in scale and capabilities. Deep neural networks (DNNs) are increasingly being deployed in the real world. However, they are difficult to analyze, raising concerns about using them without a rigorous understanding of how they function. Effective tools for interpreting them will be important for building more trustworthy AI by helping to identify problems, fix bugs, and improve basic understanding. In particular, \u201cinner\u201d interpretability techniques, which focus on explaining the internal components of DNNs, are well-suited for developing a mechanistic understanding, guiding manual modifications, and reverse engineering solutions. Much recent work has focused on DNN interpretability, and rapid progress has thus far made a thorough systematization of methods difficult. In this survey, we review over 300 works with a focus on inner interpretability tools. We introduce a taxonomy that classifies methods by what part of the network they help to explain (weights, neurons, subnetworks, or latent representations) and whether they are implemented during (intrinsic) or after (post hoc) training. To our knowledge, we are also the first to survey a number of connections between interpretability research and work in adversarial robustness, continual learning, modularity, network compression, and studying the human visual system. We discuss key challenges and argue that the status quo in interpretability research is largely unproductive. Finally, we highlight the importance of future work that emphasizes diagnostics, debugging, adversaries, and benchmarking in order to make interpretability tools more useful to engineers in practical applications."
                },
                {
                    "title": "Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limit",
                    "abstract": "There is mounting evidence of emergent phenomena in the capabilities of deep learning methods as we scale up datasets, model sizes, and training times. While there are some accounts of how these resources modulate statistical capacity, far less is known about their effect on the computational problem of model training. This work conducts such an exploration through the lens of learning a $k$-sparse parity of $n$ bits, a canonical discrete search problem which is statistically easy but computationally hard. Empirically, we find that a variety of neural networks successfully learn sparse parities, with discontinuous phase transitions in the training curves. On small instances, learning abruptly occurs at approximately $n^{O(k)}$ iterations; this nearly matches SQ lower bounds, despite the apparent lack of a sparse prior. Our theoretical analysis shows that these observations are not explained by a Langevin-like mechanism, whereby SGD\"stumbles in the dark\"until it finds the hidden set of features (a natural algorithm which also runs in $n^{O(k)}$ time). Instead, we show that SGD gradually amplifies the sparse solution via a Fourier gap in the population gradient, making continual progress that is invisible to loss and error metrics."
                },
                {
                    "title": "Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space",
                    "abstract": "Transformer-based language models (LMs) are at the core of modern NLP, but their internal prediction construction process is opaque and largely not understood. In this work, we make a substantial step towards unveiling this underlying prediction process, by reverse-engineering the operation of the feed-forward network (FFN) layers, one of the building blocks of transformer models. We view the token representation as a changing distribution over the vocabulary, and the output from each FFN layer as an additive update to that distribution. Then, we analyze the FFN updates in the vocabulary space, showing that each update can be decomposed to sub-updates corresponding to single FFN parameter vectors, each promoting concepts that are often human-interpretable. We then leverage these findings for controlling LM predictions, where we reduce the toxicity of GPT2 by almost 50%, and for improving computation efficiency with a simple early exit rule, saving 20% of computation on average."
                },
                {
                    "title": "Locating and Editing Factual Associations in GPT",
                    "abstract": "We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's factual predictions. This reveals a distinct set of steps in middle-layer feed-forward modules that mediate factual predictions while processing subject tokens. To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update specific factual associations using Rank-One Model Editing (ROME). We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another. Our results confirm an important role for mid-layer feed-forward modules in storing factual associations and suggest that direct manipulation of computational mechanisms may be a feasible approach for model editing. The code, dataset, visualizations, and an interactive demo notebook are available at https://rome.baulab.info/"
                },
                {
                    "title": "Natural Language Descriptions of Deep Visual Features",
                    "abstract": "Some neurons in deep networks specialize in recognizing highly specific perceptual, structural, or semantic features of inputs. In computer vision, techniques exist for identifying neurons that respond to individual concept categories like colors, textures, and object classes. But these techniques are limited in scope, labeling only a small subset of neurons and behaviors in any network. Is a richer characterization of neuron-level computation possible? We introduce a procedure (called MILAN, for mutual-information-guided linguistic annotation of neurons) that automatically labels neurons with open-ended, compositional, natural language descriptions. Given a neuron, MILAN generates a description by searching for a natural language string that maximizes pointwise mutual information with the image regions in which the neuron is active. MILAN produces fine-grained descriptions that capture categorical, relational, and logical structure in learned features. These descriptions obtain high agreement with human-generated feature descriptions across a diverse set of model architectures and tasks, and can aid in understanding and controlling learned models. We highlight three applications of natural language neuron descriptions. First, we use MILAN for analysis, characterizing the distribution and importance of neurons selective for attribute, category, and relational information in vision models. Second, we use MILAN for auditing, surfacing neurons sensitive to human faces in datasets designed to obscure them. Finally, we use MILAN for editing, improving robustness in an image classifier by deleting neurons sensitive to text features spuriously correlated with class labels."
                },
                {
                    "title": "Inducing Causal Structure for Interpretable Neural Networks",
                    "abstract": "In many areas, we have well-founded insights about causal structure that would be useful to bring into our trained models while still allowing them to learn in a data-driven fashion. To achieve this, we present the new method of interchange intervention training (IIT). In IIT, we (1) align variables in a causal model (e.g., a deterministic program or Bayesian network) with representations in a neural model and (2) train the neural model to match the counterfactual behavior of the causal model on a base input when aligned representations in both models are set to be the value they would be for a source input. IIT is fully differentiable, flexibly combines with other objectives, and guarantees that the target causal model is a causal abstraction of the neural model when its loss is zero. We evaluate IIT on a structural vision task (MNIST-PVR), a navigational language task (ReaSCAN), and a natural language inference task (MQNLI). We compare IIT against multi-task training objectives and data augmentation. In all our experiments, IIT achieves the best results and produces neural models that are more interpretable in the sense that they more successfully realize the target causal model."
                },
                {
                    "title": "Causal Analysis of Syntactic Agreement Mechanisms in Neural Language Models",
                    "abstract": "Targeted syntactic evaluations have demonstrated the ability of language models to perform subject-verb agreement given difficult contexts. To elucidate the mechanisms by which the models accomplish this behavior, this study applies causal mediation analysis to pre-trained neural language models. We investigate the magnitude of models\u2019 preferences for grammatical inflections, as well as whether neurons process subject-verb agreement similarly across sentences with different syntactic structures. We uncover similarities and differences across architectures and model sizes\u2014notably, that larger models do not necessarily learn stronger preferences. We also observe two distinct mechanisms for producing subject-verb agreement depending on the syntactic structure of the input sentence. Finally, we find that language models rely on similar sets of neurons when given sentences with similar syntactic structure."
                },
                {
                    "title": "Causal Abstractions of Neural Networks",
                    "abstract": "Structural analysis methods (e.g., probing and feature attribution) are increasingly important tools for neural network analysis. We propose a new structural analysis method grounded in a formal theory of causal abstraction that provides rich characterizations of model-internal representations and their roles in input/output behavior. In this method, neural representations are aligned with variables in interpretable causal models, and then interchange interventions are used to experimentally verify that the neural representations have the causal properties of their aligned variables. We apply this method in a case study to analyze neural models trained on Multiply Quantified Natural Language Inference (MQNLI) corpus, a highly complex NLI dataset that was constructed with a tree-structured natural logic causal model. We discover that a BERT-based model with state-of-the-art performance successfully realizes parts of the natural logic model's causal structure, whereas a simpler baseline model fails to show any such structure, demonstrating that BERT representations encode the compositional structure of MQNLI."
                },
                {
                    "title": "The Out-of-Distribution Problem in Explainability and Search Methods for Feature Importance Explanations",
                    "abstract": "Feature importance (FI) estimates are a popular form of explanation, and they are commonly created and evaluated by computing the change in model confidence caused by removing certain input features at test time. For example, in the standard Sufficiency metric, only the top-k most important tokens are kept. In this paper, we study several under-explored dimensions of FI explanations, providing conceptual and empirical improvements for this form of explanation. First, we advance a new argument for why it can be problematic to remove features from an input when creating or evaluating explanations: the fact that these counterfactual inputs are out-of-distribution (OOD) to models implies that the resulting explanations are socially misaligned. The crux of the problem is that the model prior and random weight initialization influence the explanations (and explanation metrics) in unintended ways. To resolve this issue, we propose a simple alteration to the model training process, which results in more socially aligned explanations and metrics. Second, we compare among five approaches for removing features from model inputs. We find that some methods produce more OOD counterfactuals than others, and we make recommendations for selecting a feature-replacement function. Finally, we introduce four search-based methods for identifying FI explanations and compare them to strong baselines, including LIME, Anchors, and Integrated Gradients. Through experiments with six diverse text classification datasets, we find that the only method that consistently outperforms random search is a Parallel Local Search (PLS) that we introduce. Improvements over the second-best method are as large as 5.4 points for Sufficiency and 17 points for Comprehensiveness. All supporting code for experiments in this paper is publicly available at https://github.com/peterbhase/ExplanationSearch."
                },
                {
                    "title": "Knowledge Neurons in Pretrained Transformers",
                    "abstract": "Large-scale pretrained language models are surprisingly good at recalling factual knowledge presented in the training corpus. In this paper, we present preliminary studies on how factual knowledge is stored in pretrained Transformers by introducing the concept of knowledge neurons. Specifically, we examine the fill-in-the-blank cloze task for BERT. Given a relational fact, we propose a knowledge attribution method to identify the neurons that express the fact. We find that the activation of such knowledge neurons is positively correlated to the expression of their corresponding facts. In our case studies, we attempt to leverage knowledge neurons to edit (such as update, and erase) specific factual knowledge without fine-tuning. Our results shed light on understanding the storage of knowledge within pretrained Transformers."
                },
                {
                    "title": "Transformer visualization via dictionary learning: contextualized embedding as a linear superposition of transformer factors",
                    "abstract": "Transformer networks have revolutionized NLP representation learning since they were introduced. Though a great effort has been made to explain the representation in transformers, it is widely recognized that our understanding is not sufficient. One important reason is that there lack enough visualization tools for detailed analysis. In this paper, we propose to use dictionary learning to open up these \u2018black boxes\u2019 as linear superpositions of transformer factors. Through visualization, we demonstrate the hierarchical semantic structures captured by the transformer factors, e.g., word-level polysemy disambiguation, sentence-level pattern formation, and long-range dependency. While some of these patterns confirm the conventional prior linguistic knowledge, the rest are relatively unexpected, which may provide new insights. We hope this visualization tool can bring further knowledge and a better understanding of how transformer networks work. The code is available at: https://github.com/zeyuyun1/TransformerVis."
                },
                {
                    "title": "Transformer Feed-Forward Layers Are Key-Value Memories",
                    "abstract": "Feed-forward layers constitute two-thirds of a transformer model\u2019s parameters, yet their role in the network remains under-explored. We show that feed-forward layers in transformer-based language models operate as key-value memories, where each key correlates with textual patterns in the training examples, and each value induces a distribution over the output vocabulary. Our experiments show that the learned patterns are human-interpretable, and that lower layers tend to capture shallow patterns, while upper layers learn more semantic ones. The values complement the keys\u2019 input patterns by inducing output distributions that concentrate probability mass on tokens likely to appear immediately after each pattern, particularly in the upper layers. Finally, we demonstrate that the output of a feed-forward layer is a composition of its memories, which is subsequently refined throughout the model\u2019s layers via residual connections to produce the final output distribution."
                },
                {
                    "title": "Compositional Explanations of Neurons",
                    "abstract": "We describe a procedure for explaining neurons in deep representations by identifying compositional logical concepts that closely approximate neuron behavior. Compared to prior work that uses atomic labels as explanations, analyzing neurons compositionally allows us to more precisely and expressively characterize their behavior. We use this procedure to answer several questions on interpretability in models for vision and natural language processing. First, we examine the kinds of abstractions learned by neurons. In image classification, we find that many neurons learn highly abstract but semantically coherent visual concepts, while other polysemantic neurons detect multiple unrelated features; in natural language inference (NLI), neurons learn shallow lexical heuristics from dataset biases. Second, we see whether compositional explanations give us insight into model performance: vision neurons that detect human-interpretable concepts are positively correlated with task performance, while NLI neurons that fire for shallow heuristics are negatively correlated with task performance. Finally, we show how compositional explanations provide an accessible way for end users to produce simple \"copy-paste\" adversarial examples that change model behavior in predictable ways."
                },
                {
                    "title": "Neural Natural Language Inference Models Partially Embed Theories of Lexical Entailment and Negation",
                    "abstract": "We address whether neural models for Natural Language Inference (NLI) can learn the compositional interactions between lexical entailment and negation, using four methods: the behavioral evaluation methods of (1) challenge test sets and (2) systematic generalization tasks, and the structural evaluation methods of (3) probes and (4) interventions. To facilitate this holistic evaluation, we present Monotonicity NLI (MoNLI), a new naturalistic dataset focused on lexical entailment and negation. In our behavioral evaluations, we find that models trained on general-purpose NLI datasets fail systematically on MoNLI examples containing negation, but that MoNLI fine-tuning addresses this failure. In our structural evaluations, we look for evidence that our top-performing BERT-based model has learned to implement the monotonicity algorithm behind MoNLI. Probes yield evidence consistent with this conclusion, and our intervention experiments bolster this, showing that the causal dynamics of the model mirror the causal dynamics of this algorithm on subsets of MoNLI. This suggests that the BERT model at least partially embeds a theory of lexical entailment and negation at an algorithmic level."
                },
                {
                    "title": "Feature relevance quantification in explainable AI: A causality problem",
                    "abstract": "We discuss promising recent contributions on quantifying feature relevance using Shapley values, where we observed some confusion on which probability distribution is the right one for dropped features. We argue that the confusion is based on not carefully distinguishing between observational and interventional conditional probabilities and try a clarification based on Pearl's seminal work on causality. We conclude that unconditional rather than conditional expectations provide the right notion of dropping features in contradiction to the theoretical justification of the software package SHAP. Parts of SHAP are unaffected because unconditional expectations (which we argue to be conceptually right) are used as approximation for the conditional ones, which encouraged others to `improve' SHAP in a way that we believe to be flawed."
                },
                {
                    "title": "Discovering the Compositional Structure of Vector Representations with Role Learning Networks",
                    "abstract": "How can neural networks perform so well on compositional tasks even though they lack explicit compositional representations? We use a novel analysis technique called ROLE to show that recurrent neural networks perform well on such tasks by converging to solutions which implicitly represent symbolic structure. This method uncovers a symbolic structure which, when properly embedded in vector space, closely approximates the encodings of a standard seq2seq network trained to perform the compositional SCAN task. We verify the causal importance of the discovered symbolic structure by showing that, when we systematically manipulate hidden embeddings based on this symbolic structure, the model\u2019s output is changed in the way predicted by our analysis."
                },
                {
                    "title": "A Benchmark for Interpretability Methods in Deep Neural Networks",
                    "abstract": "We propose an empirical measure of the approximate accuracy of feature importance estimates in deep neural networks. Our results across several large-scale image classification datasets show that many popular interpretability methods produce estimates of feature importance that are not better than a random designation of feature importance. Only certain ensemble based approaches---VarGrad and SmoothGrad-Squared---outperform such a random assignment of importance. The manner of ensembling remains critical, we show that some approaches do no better then the underlying method but carry a far higher computational burden."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Direct and Indirect Effects",
                    "abstract": "The direct effect of one event on another can be defined and measured by holding constant all intermediate variables between the two. Indirect effects present conceptual and practical difficulties (in nonlinear models), because they cannot be isolated by holding certain variables constant. This paper presents a new way of defining the effect transmitted through a restricted set of paths, without controlling variables on the remaining paths. This permits the assessment of a more natural type of direct and indirect effects, one that is applicable in both linear and nonlinear models and that has broader policy-related interpretations. The paper establishes conditions under which such assessments can be estimated consistently from experimental and nonexperimental data, and thus extends path-analytic techniques to nonlinear and nonparametric models."
                },
                {
                    "title": "On Privileged and Convergent Bases in Neural Network Representations",
                    "abstract": "In this study, we investigate whether the representations learned by neural networks possess a privileged and convergent basis. Specifically, we examine the significance of feature directions represented by individual neurons. First, we establish that arbitrary rotations of neural representations cannot be inverted (unlike linear networks), indicating that they do not exhibit complete rotational invariance. Subsequently, we explore the possibility of multiple bases achieving identical performance. To do this, we compare the bases of networks trained with the same parameters but with varying random initializations. Our study reveals two findings: (1) Even in wide networks such as WideResNets, neural networks do not converge to a unique basis; (2) Basis correlation increases significantly when a few early layers of the network are frozen identically. Furthermore, we analyze Linear Mode Connectivity, which has been studied as a measure of basis correlation. Our findings give evidence that while Linear Mode Connectivity improves with increased network width, this improvement is not due to an increase in basis correlation."
                },
                {
                    "title": "Investigating Gender Bias in Language Models Using Causal Mediation Analysis",
                    "abstract": "Many interpretation methods for neural models in natural language processing investigate how information is encoded inside hidden representations. However, these methods can only measure whether the information exists, not whether it is actually used by the model. We propose a methodology grounded in the theory of causal mediation analysis for interpreting which parts of a model are causally implicated in its behavior. The approach enables us to analyze the mechanisms that facilitate the \ufb02ow of information from input to output through various model components, known as mediators. As a case study, we apply this methodology to analyzing gender bias in pre-trained Transformer language models. We study the role of individual neurons and attention heads in mediating gender bias across three datasets designed to gauge a model\u2019s sensitivity to gender bias. Our mediation analysis reveals that gender bias effects are concentrated in speci\ufb01c components of the model that may exhibit highly specialized behavior."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does the lack of standardization in activation patching methodologies affect the interpretability results of machine learning models?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the inconsistencies in interpretability results derived from activation patching techniques. By establishing standardized methodologies, future research can build upon reliable findings, leading to a more robust understanding of model behavior and improved interpretability practices. This advancement could facilitate the development of more transparent machine learning systems, ultimately enhancing trust and usability in practical applications across various domains, such as healthcare, finance, and autonomous systems.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of machine learning models and the diverse methodologies currently employed in activation patching. Naive approaches may fail because they do not account for the variability in how corrupted prompts are generated or how the effects of patching are measured. Additionally, the lack of consensus on metrics and methods introduces significant technical obstacles, making it difficult to draw reliable conclusions from existing studies. Overcoming these complexities requires a systematic investigation into the various degrees of freedom in activation patching and their impact on interpretability.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on applying activation patching without critically examining the methodological differences that influence results. This oversight has created gaps in understanding how variations in prompt corruption and evaluation metrics affect interpretability outcomes. Barriers to solving this problem include the fragmented nature of existing studies and the absence of a comprehensive framework for comparing methodologies. Our approach differs by systematically analyzing these variations and their implications, thereby providing a clearer path toward establishing best practices in activation patching.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a systematic study of activation patching by examining three key components: (1) the generation of corrupted prompts using Gaussian noising and symmetric token replacement, (2) the evaluation of patching effects through probability and logit difference metrics, and (3) the implementation of sliding window patching to restore activations across multiple MLP layers. We will utilize a diverse dataset of language model prompts to assess the impact of these variations. The expected outcomes include a clearer understanding of how different methodologies influence interpretability results, leading to recommendations for standardized practices in activation patching."
            }
        },
        "author_data": {
            "bb0aa302-74db-407b-bd38-e6498de62f42": {
                "pk": "bb0aa302-74db-407b-bd38-e6498de62f42",
                "name": "Fred Zhang",
                "collaborators": [
                    "David P. Woodruff",
                    "Qiuyi Zhang",
                    "Sandeep Silwal",
                    "Justin Y. Chen",
                    "A. Vakilian",
                    "Samson Zhou",
                    "Prayaag Venkat",
                    "Zhou Lu",
                    "Xinyi Chen",
                    "Elad Hazan",
                    "Danny Halawi",
                    "Chen Yueh-Han",
                    "Jacob Steinhardt",
                    "Anders Aamand",
                    "Allen Liu",
                    "Pattara Sukprasert",
                    "Manolis Zampetakis",
                    "Yeshwanth Cherapanamjeri",
                    "Binghui Peng",
                    "Daniel Alabi",
                    "Pravesh Kothari",
                    "Pranay Tankala",
                    "Samuel B. Hopkins",
                    "J. Li",
                    "Alexander Wei",
                    "Zhixian Lei",
                    "K. Luh",
                    "Preetum Nakkiran",
                    "Gal Kaplun",
                    "Dimitris Kalimeris",
                    "Tristan Yang",
                    "Benjamin L. Edelman",
                    "B. Barak",
                    "K. Fox",
                    "Debmalya Panigrahi"
                ],
                "domain": [
                    "Online Learning",
                    "Bandit Algorithms",
                    "Machine Learning",
                    "Robust Statistics"
                ],
                "publications": [
                    {
                        "title": "Adaptive Regret for Bandits Made Possible: Two Queries Suffice",
                        "abstract": "Fast changing states or volatile environments pose a significant challenge to online optimization, which needs to perform rapid adaptation under limited observation. In this paper, we give query and regret optimal bandit algorithms under the strict notion of strongly adaptive regret, which measures the maximum regret over any contiguous interval $I$. Due to its worst-case nature, there is an almost-linear $\\Omega(|I|^{1-\\epsilon})$ regret lower bound, when only one query per round is allowed [Daniely el al, ICML 2015]. Surprisingly, with just two queries per round, we give Strongly Adaptive Bandit Learner (StABL) that achieves $\\tilde{O}(\\sqrt{n|I|})$ adaptive regret for multi-armed bandits with $n$ arms. The bound is tight and cannot be improved in general. Our algorithm leverages a multiplicative update scheme of varying stepsizes and a carefully chosen observation distribution to control the variance. Furthermore, we extend our results and provide optimal algorithms in the bandit convex optimization setting. Finally, we empirically demonstrate the superior performance of our algorithms under volatile environments and for downstream tasks, such as algorithm selection for hyperparameter optimization."
                    },
                    {
                        "title": "Approaching Human-Level Forecasting with Language Models",
                        "abstract": "Forecasting future events is important for policy and decision making. In this work, we study whether language models (LMs) can forecast at the level of competitive human forecasters. Towards this goal, we develop a retrieval-augmented LM system designed to automatically search for relevant information, generate forecasts, and aggregate predictions. To facilitate our study, we collect a large dataset of questions from competitive forecasting platforms. Under a test set published after the knowledge cut-offs of our LMs, we evaluate the end-to-end performance of our system against the aggregates of human forecasts. On average, the system nears the crowd aggregate of competitive forecasters, and in some settings surpasses it. Our work suggests that using LMs to forecast the future could provide accurate predictions at scale and help to inform institutional decision making."
                    },
                    {
                        "title": "Constant Approximation for Individual Preference Stable Clustering",
                        "abstract": "Individual preference (IP) stability, introduced by Ahmadi et al. (ICML 2022), is a natural clustering objective inspired by stability and fairness constraints. A clustering is $\\alpha$-IP stable if the average distance of every data point to its own cluster is at most $\\alpha$ times the average distance to any other cluster. Unfortunately, determining if a dataset admits a $1$-IP stable clustering is NP-Hard. Moreover, before this work, it was unknown if an $o(n)$-IP stable clustering always \\emph{exists}, as the prior state of the art only guaranteed an $O(n)$-IP stable clustering. We close this gap in understanding and show that an $O(1)$-IP stable clustering always exists for general metrics, and we give an efficient algorithm which outputs such a clustering. We also introduce generalizations of IP stability beyond average distance and give efficient, near-optimal algorithms in the cases where we consider the maximum and minimum distances within and between clusters."
                    },
                    {
                        "title": "Streaming Algorithms for Learning with Experts: Deterministic Versus Robust",
                        "abstract": "In the online learning with experts problem, an algorithm must make a prediction about an outcome on each of $T$ days (or times), given a set of $n$ experts who make predictions on each day (or time). The algorithm is given feedback on the outcomes of each day, including the cost of its prediction and the cost of the expert predictions, and the goal is to make a prediction with the minimum cost, specifically compared to the best expert in the set. Recent work by Srinivas, Woodruff, Xu, and Zhou (STOC 2022) introduced the study of the online learning with experts problem under memory constraints. However, often the predictions made by experts or algorithms at some time influence future outcomes, so that the input is adaptively chosen. Whereas deterministic algorithms would be robust to adaptive inputs, existing algorithms all crucially use randomization to sample a small number of experts. In this paper, we study deterministic and robust algorithms for the experts problem. We first show a space lower bound of $\\widetilde{\\Omega}\\left(\\frac{nM}{RT}\\right)$ for any deterministic algorithm that achieves regret $R$ when the best expert makes $M$ mistakes. Our result shows that the natural deterministic algorithm, which iterates through pools of experts until each expert in the pool has erred, is optimal up to polylogarithmic factors. On the positive side, we give a randomized algorithm that is robust to adaptive inputs that uses $\\widetilde{O}\\left(\\frac{n}{R\\sqrt{T}}\\right)$ space for $M=O\\left(\\frac{R^2 T}{\\log^2 n}\\right)$, thereby showing a smooth space-regret trade-off."
                    },
                    {
                        "title": "Bayesian Strategy-Proof Facility Location via Robust Estimation",
                        "abstract": "A seminal work by Moulin (1980) shows that the median voting scheme fully characterizes (deterministic) strategy-proof facility location mechanism for single-peaked preferences. In this simple setting, median also achieves the optimal social cost. In d dimensions, strategy-proof mechanism is characterized by coordinate-wise median, which is known to have a large \u221a d approximation ratio of the social cost in the Euclidean space, whereas the socially optimal mechanism fails at being strategy-proof. In light of the negative results in the classic, worst-case setting, we initiate the study of Bayesian mechanism de-sign for strategy-proof facility location for multi-dimensional Euclidean preferences, where the agents\u2019 preferences are drawn from a distribution. We approach the problem via connections to algorithmic high-dimensional robust statistics. Specially, our contributions are the following:"
                    },
                    {
                        "title": "Robust Algorithms on Adaptive Inputs from Bounded Adversaries",
                        "abstract": "We study dynamic algorithms robust to adaptive input generated from sources with bounded capabilities, such as sparsity or limited interaction. For example, we consider robust linear algebraic algorithms when the updates to the input are sparse but given by an adversary with access to a query oracle. We also study robust algorithms in the standard centralized setting, where an adversary queries an algorithm in an adaptive manner, but the number of interactions between the adversary and the algorithm is bounded. We first recall a unified framework of [HKM+20, BKM+22, ACSS23] for answering $Q$ adaptive queries that incurs $\\widetilde{\\mathcal{O}}(\\sqrt{Q})$ overhead in space, which is roughly a quadratic improvement over the na\\\"{i}ve implementation, and only incurs a logarithmic overhead in query time. Although the general framework has diverse applications in machine learning and data science, such as adaptive distance estimation, kernel density estimation, linear regression, range queries, and point queries and serves as a preliminary benchmark, we demonstrate even better algorithmic improvements for (1) reducing the pre-processing time for adaptive distance estimation and (2) permitting an unlimited number of adaptive queries for kernel density estimation. Finally, we complement our theoretical results with additional empirical evaluations."
                    },
                    {
                        "title": "Online Prediction in Sub-linear Space",
                        "abstract": "We provide the first sub-linear space and sub-linear regret algorithm for online learning with expert advice (against an oblivious adversary), addressing an open question raised recently by Srinivas, Woodruff, Xu and Zhou (STOC 2022). We also demonstrate a separation between oblivious and (strong) adaptive adversaries by proving a linear memory lower bound of any sub-linear regret algorithm against an adaptive adversary. Our algorithm is based on a novel pool selection procedure that bypasses the traditional wisdom of leader selection for online learning, and a generic reduction that transforms any weakly sub-linear regret $o(T)$ algorithm to $T^{1-\\alpha}$ regret algorithm, which may be of independent interest. Our lower bound utilizes the connection of no-regret learning and equilibrium computation in zero-sum games, leading to a proof of a strong lower bound against an adaptive adversary."
                    },
                    {
                        "title": "Faster Fundamental Graph Algorithms via Learned Predictions",
                        "abstract": "We consider the question of speeding up classic graph algorithms with machine-learned predictions. In this model, algorithms are furnished with extra advice learned from past or similar instances. Given the additional information, we aim to improve upon the traditional worst-case run-time guarantees. Our contributions are the following: Finally, we give a set of general learnability theorems, showing that the predictions required by our algorithms can be efficiently learned in a PAC fashion."
                    },
                    {
                        "title": "Optimal Query Complexities for Dynamic Trace Estimation",
                        "abstract": "We consider the problem of minimizing the number of matrix-vector queries needed for accurate trace estimation in the dynamic setting where our underlying matrix is changing slowly, such as during an optimization process. Specifically, for any $m$ matrices $A_1,...,A_m$ with consecutive differences bounded in Schatten-$1$ norm by $\\alpha$, we provide a novel binary tree summation procedure that simultaneously estimates all $m$ traces up to $\\epsilon$ error with $\\delta$ failure probability with an optimal query complexity of $\\widetilde{O}\\left(m \\alpha\\sqrt{\\log(1/\\delta)}/\\epsilon + m\\log(1/\\delta)\\right)$, improving the dependence on both $\\alpha$ and $\\delta$ from Dharangutte and Musco (NeurIPS, 2021). Our procedure works without additional norm bounds on $A_i$ and can be generalized to a bound for the $p$-th Schatten norm for $p \\in [1,2]$, giving a complexity of $\\widetilde{O}\\left(m \\alpha\\left(\\sqrt{\\log(1/\\delta)}/\\epsilon\\right)^p +m \\log(1/\\delta)\\right)$. By using novel reductions to communication complexity and information-theoretic analyses of Gaussian matrices, we provide matching lower bounds for static and dynamic trace estimation in all relevant parameters, including the failure probability. Our lower bounds (1) give the first tight bounds for Hutchinson's estimator in the matrix-vector product model with Frobenius norm error even in the static setting, and (2) are the first unconditional lower bounds for dynamic trace estimation, resolving open questions of prior work."
                    },
                    {
                        "title": "Privately Estimating a Gaussian: Efficient, Robust, and Optimal",
                        "abstract": "In this work, we give efficient algorithms for privately estimating a Gaussian distribution in both pure and approximate differential privacy (DP) models with optimal dependence on the dimension in the sample complexity. In the pure DP setting, we give an efficient algorithm that estimates an unknown d-dimensional Gaussian distribution up to an arbitrary tiny total variation error using O(d2 log\u03ba) samples while tolerating a constant fraction of adversarial outliers. Here, \u03ba is the condition number of the target covariance matrix. The sample bound matches best non-private estimators in the dependence on the dimension (up to a polylogarithmic factor). We prove a new lower bound on differentially private covariance estimation to show that the dependence on the condition number \u03ba in the above sample bound is also tight. Prior to our work, only identifiability results (yielding inefficient super-polynomial time algorithms) were known for the problem. In the approximate DP setting, we give an efficient algorithm to estimate an unknown Gaussian distribution up to an arbitrarily tiny total variation error using O(d2) samples while tolerating a constant fraction of adversarial outliers. Prior to our work, all efficient approximate DP algorithms incurred a super-quadratic sample cost or were not outlier-robust. For the special case of mean estimation, our algorithm achieves the optimal sample complexity of O(d), improving on a O(d1.5) bound from prior work. Our pure DP algorithm relies on a recursive private preconditioning subroutine that utilizes recent work of Hopkins et al. (STOC 2022) on private mean estimation. Our approximate DP algorithms are based on a substantial upgrade of the method of stabilizing convex relaxations introduced by Kothari et al. (COLT 2022). In particular, we improve on their mechanism by using a new unnormalized entropy regularization and a new and surprisingly simple mechanism for privately releasing covariances."
                    },
                    {
                        "title": "Robust and Heavy-Tailed Mean Estimation Made Simple, via Regret Minimization",
                        "abstract": "We study the problem of estimating the mean of a distribution in high dimensions when either the samples are adversarially corrupted or the distribution is heavy-tailed. Recent developments in robust statistics have established efficient and (near) optimal procedures for both settings. However, the algorithms developed on each side tend to be sophisticated and do not directly transfer to the other, with many of them having ad-hoc or complicated analyses.  In this paper, we provide a meta-problem and a duality theorem that lead to a new unified view on robust and heavy-tailed mean estimation in high dimensions. We show that the meta-problem can be solved either by a variant of the Filter algorithm from the recent literature on robust estimation or by the quantum entropy scoring scheme (QUE), due to Dong, Hopkins and Li (NeurIPS '19). By leveraging our duality theorem, these results translate into simple and efficient algorithms for both robust and heavy-tailed settings. Furthermore, the QUE-based procedure has run-time that matches the fastest known algorithms on both fronts.  Our analysis of Filter is through the classic regret bound of the multiplicative weights update method. This connection allows us to avoid the technical complications in previous works and improve upon the run-time analysis of a gradient-descent-based algorithm for robust mean estimation by Cheng, Diakonikolas, Ge and Soltanolkotabi (ICML '20)."
                    },
                    {
                        "title": "Optimal Robustness-Consistency Trade-offs for Learning-Augmented Online Algorithms",
                        "abstract": "We study the problem of improving the performance of online algorithms by incorporating machine-learned predictions. The goal is to design algorithms that are both consistent and robust, meaning that the algorithm performs well when predictions are accurate and maintains worst-case guarantees. Such algorithms have been studied in a recent line of works due to Lykouris and Vassilvitskii (ICML '18) and Purohit et al (NeurIPS '18). They provide robustness-consistency trade-offs for a variety of online problems. However, they leave open the question of whether these trade-offs are tight, i.e., to what extent to such trade-offs are necessary. In this paper, we provide the first set of non-trivial lower bounds for competitive analysis using machine-learned predictions. We focus on the classic problems of ski-rental and non-clairvoyant scheduling and provide optimal trade-offs in various settings."
                    },
                    {
                        "title": "A Fast Spectral Algorithm for Mean Estimation with Sub-Gaussian Rates",
                        "abstract": "We study the algorithmic problem of estimating the mean of heavy-tailed random vector in $\\mathbb{R}^d$, given $n$ i.i.d. samples. The goal is to design an efficient estimator that attains the optimal sub-gaussian error bound, only assuming that the random vector has bounded mean and covariance. Polynomial-time solutions to this problem are known but have high runtime due to their use of semi-definite programming (SDP). Conceptually, it remains open whether convex relaxation is truly necessary for this problem.  In this work, we show that it is possible to go beyond SDP and achieve better computational efficiency. In particular, we provide a spectral algorithm that achieves the optimal statistical performance and runs in time $\\widetilde O\\left(n^2 d \\right)$, improving upon the previous fastest runtime $\\widetilde O\\left(n^{3.5}+ n^2d\\right)$ by Cherapanamjeri el al. (COLT '19). Our algorithm is spectral in that it only requires (approximate) eigenvector computations, which can be implemented very efficiently by, for example, power iteration or the Lanczos method.  At the core of our algorithm is a novel connection between the furthest hyperplane problem introduced by Karnin et al. (COLT '12) and a structural lemma on heavy-tailed distributions by Lugosi and Mendelson (Ann. Stat. '19). This allows us to iteratively reduce the estimation error at a geometric rate using only the information derived from the top singular vector of the data matrix, leading to a significantly faster running time."
                    },
                    {
                        "title": "SGD on Neural Networks Learns Functions of Increasing Complexity",
                        "abstract": "We perform an experimental study of the dynamics of Stochastic Gradient Descent (SGD) in learning deep neural networks for several real and synthetic classification tasks. We show that in the initial epochs, almost all of the performance improvement of the classifier obtained by SGD can be explained by a linear classifier. More generally, we give evidence for the hypothesis that, as iterations progress, SGD learns functions of increasing complexity. This hypothesis can be helpful in explaining why SGD-learned classifiers tend to generalize well even in the over-parameterized regime. We also show that the linear classifier learned in the initial stages is \"retained\" throughout the execution even if training is continued to the point of zero training error, and complement this with a theoretical result in a simplified model. Key to our work is a new measure of how well one classifier explains the performance of another, based on conditional mutual information."
                    },
                    {
                        "title": "Minimum Cut and Minimum k-Cut in Hypergraphs via Branching Contractions",
                        "abstract": "On hypergraphs with m hyperedges and n vertices, where p denotes the total size of the hyperedges, we provide the following results: We give an algorithm that runs in \\(\\widetilde{O}(mn^{2k-2})\\) time for finding a minimum k-cut in hypergraphs of arbitrary rank. This algorithm betters the previous best running time for the minimum k-cut problem, for k > 2. We give an algorithm that runs in \\(\\widetilde{O}(n^{\\max \\lbrace r,2k-2\\rbrace })\\) time for finding a minimum k-cut in hypergraphs of constant rank r. This algorithm betters the previous best running times for both the minimum cut and minimum k-cut problems for dense hypergraphs. Both of our algorithms are Monte Carlo, i.e., they return a minimum k-cut (or minimum cut) with high probability. These algorithms are obtained as instantiations of a generic branching randomized contraction technique on hypergraphs, which extends the celebrated work of Karger and Stein on recursive contractions in graphs. Our techniques and results also extend to the problems of minimum hedge-cut and minimum hedge-k-cut on hedgegraphs, which generalize hypergraphs."
                    }
                ]
            },
            "ef442cd0-7f1b-4878-939b-2922006427a7": {
                "pk": "ef442cd0-7f1b-4878-939b-2922006427a7",
                "name": "Neel Nanda",
                "collaborators": [
                    "Arthur Conmy",
                    "Wes Gurnee",
                    "Tom Lieberum",
                    "Senthooran Rajamanoharan",
                    "Vikrant Varma",
                    "J'anos Kram'ar",
                    "Rohin Shah",
                    "Nicolas Sonnerat",
                    "Lewis Smith",
                    "Cody Rushing",
                    "Bilal Chughtai",
                    "J\u00e1nos Kram\u00e1r",
                    "Jacob Dunefsky",
                    "Philippe Chlenski",
                    "Theo Horsley",
                    "Zifan Carl Guo",
                    "Tara Rezaei Kheirkhah",
                    "Qinyi Sun",
                    "Will Hathaway",
                    "Dimitris Bertsimas",
                    "Andy Arditi",
                    "Oscar Obeso",
                    "Aaquib Syed",
                    "Daniel Paleka",
                    "Nina Rimsky",
                    "Anca Dragan",
                    "M. E. Ildiz",
                    "Zhe Zhao",
                    "Samet Oymak",
                    "Xiangyu Chang",
                    "Yingcong Li",
                    "Christos Thrampoulidis",
                    "Lin Chen",
                    "Yifei Min",
                    "Mikhail Belkin",
                    "Aakanksha Chowdhery",
                    "Sharan Narang",
                    "Jacob Devlin",
                    "Maarten Bosma",
                    "Gaurav Mishra",
                    "Adam Roberts",
                    "Liam Collins",
                    "Hamed Hassani",
                    "M. Soltanolkotabi",
                    "Aryan Mokhtari",
                    "Sanjay Shakkottai",
                    "Provable",
                    "Simon S. Du",
                    "Wei Hu",
                    "S. Kakade",
                    "Chelsea Finn",
                    "A. Rajeswaran",
                    "Deep Ganguli",
                    "Danny Hernandez",
                    "Liane Lovitt",
                    "Amanda Askell",
                    "Yu Bai",
                    "Anna Chen",
                    "Tom Conerly",
                    "Nova Dassarma",
                    "Dawn Drain",
                    "Sheer Nelson El-hage",
                    "El Showk",
                    "Stanislav Fort",
                    "Zac Hatfield-Dodds",
                    "T. Henighan",
                    "Scott Johnston",
                    "Andy Jones",
                    "Nicholas Joseph",
                    "Jackson Kernian",
                    "Shauna Kravec",
                    "Benjamin Mann",
                    "Kamal Ndousse",
                    "Catherine Olsson",
                    "D. Amodei",
                    "Tom Brown",
                    "Jared Ka-plan",
                    "Sam McCandlish",
                    "C. Olah",
                    "Dario Amodei",
                    "Trevor Hastie",
                    "Andrea Montanari",
                    "Saharon Rosset",
                    "Jordan Hoffmann",
                    "Sebastian Borgeaud",
                    "A. Mensch",
                    "Elena Buchatskaya",
                    "Trevor Cai",
                    "Eliza Rutherford",
                    "Diego de",
                    "Las Casas",
                    "Lisa Anne Hendricks",
                    "Johannes Welbl",
                    "Aidan Clark",
                    "Tom Hennigan",
                    "Eric Noland",
                    "Katie Millican",
                    "George van den Driessche",
                    "Bogdan Damoc",
                    "Aurelia Guy"
                ],
                "domain": [
                    "Mechanistic Interpretability",
                    "Sparse Autoencoders",
                    "Neural Networks",
                    "Language Models"
                ],
                "publications": [
                    {
                        "title": "Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders",
                        "abstract": "Sparse autoencoders (SAEs) are a promising unsupervised approach for identifying causally relevant and interpretable linear features in a language model's (LM) activations. To be useful for downstream tasks, SAEs need to decompose LM activations faithfully; yet to be interpretable the decomposition must be sparse -- two objectives that are in tension. In this paper, we introduce JumpReLU SAEs, which achieve state-of-the-art reconstruction fidelity at a given sparsity level on Gemma 2 9B activations, compared to other recent advances such as Gated and TopK SAEs. We also show that this improvement does not come at the cost of interpretability through manual and automated interpretability studies. JumpReLU SAEs are a simple modification of vanilla (ReLU) SAEs -- where we replace the ReLU with a discontinuous JumpReLU activation function -- and are similarly efficient to train and run. By utilising straight-through-estimators (STEs) in a principled manner, we show how it is possible to train JumpReLU SAEs effectively despite the discontinuous JumpReLU function introduced in the SAE's forward pass. Similarly, we use STEs to directly train L0 to be sparse, instead of training on proxies such as L1, avoiding problems like shrinkage."
                    },
                    {
                        "title": "Transcoders Find Interpretable LLM Feature Circuits",
                        "abstract": "A key goal in mechanistic interpretability is circuit analysis: finding sparse subgraphs of models corresponding to specific behaviors or capabilities. However, MLP sublayers make fine-grained circuit analysis on transformer-based language models difficult. In particular, interpretable features -- such as those found by sparse autoencoders (SAEs) -- are typically linear combinations of extremely many neurons, each with its own nonlinearity to account for. Circuit analysis in this setting thus either yields intractably large circuits or fails to disentangle local and global behavior. To address this we explore transcoders, which seek to faithfully approximate a densely activating MLP layer with a wider, sparsely-activating MLP layer. We successfully train transcoders on language models with 120M, 410M, and 1.4B parameters, and find them to perform at least on par with SAEs in terms of sparsity, faithfulness, and human-interpretability. We then introduce a novel method for using transcoders to perform weights-based circuit analysis through MLP sublayers. The resulting circuits neatly factorize into input-dependent and input-invariant terms. Finally, we apply transcoders to reverse-engineer unknown circuits in the model, and we obtain novel insights regarding the greater-than circuit in GPT2-small. Our results suggest that transcoders can prove effective in decomposing model computations involving MLPs into interpretable circuits. Code is available at https://github.com/jacobdunefsky/transcoder_circuits."
                    },
                    {
                        "title": "Universal Neurons in GPT2 Language Models",
                        "abstract": "A basic question within the emerging field of mechanistic interpretability is the degree to which neural networks learn the same underlying mechanisms. In other words, are neural mechanisms universal across different models? In this work, we study the universality of individual neurons across GPT2 models trained from different initial random seeds, motivated by the hypothesis that universal neurons are likely to be interpretable. In particular, we compute pairwise correlations of neuron activations over 100 million tokens for every neuron pair across five different seeds and find that 1-5\\% of neurons are universal, that is, pairs of neurons which consistently activate on the same inputs. We then study these universal neurons in detail, finding that they usually have clear interpretations and taxonomize them into a small number of neuron families. We conclude by studying patterns in neuron weights to establish several universal functional roles of neurons in simple circuits: deactivating attention heads, changing the entropy of the next token distribution, and predicting the next token to (not) be within a particular set."
                    },
                    {
                        "title": "Refusal in Language Models Is Mediated by a Single Direction",
                        "abstract": "Conversational large language models are fine-tuned for both instruction-following and safety, resulting in models that obey benign requests but refuse harmful ones. While this refusal behavior is widespread across chat models, its underlying mechanisms remain poorly understood. In this work, we show that refusal is mediated by a one-dimensional subspace, across 13 popular open-source chat models up to 72B parameters in size. Specifically, for each model, we find a single direction such that erasing this direction from the model's residual stream activations prevents it from refusing harmful instructions, while adding this direction elicits refusal on even harmless instructions. Leveraging this insight, we propose a novel white-box jailbreak method that surgically disables refusal with minimal effect on other capabilities. Finally, we mechanistically analyze how adversarial suffixes suppress propagation of the refusal-mediating direction. Our findings underscore the brittleness of current safety fine-tuning methods. More broadly, our work showcases how an understanding of model internals can be leveraged to develop practical methods for controlling model behavior."
                    },
                    {
                        "title": "Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2",
                        "abstract": "Sparse autoencoders (SAEs) are an unsupervised method for learning a sparse decomposition of a neural network's latent representations into seemingly interpretable features. Despite recent excitement about their potential, research applications outside of industry are limited by the high cost of training a comprehensive suite of SAEs. In this work, we introduce Gemma Scope, an open suite of JumpReLU SAEs trained on all layers and sub-layers of Gemma 2 2B and 9B and select layers of Gemma 2 27B base models. We primarily train SAEs on the Gemma 2 pre-trained models, but additionally release SAEs trained on instruction-tuned Gemma 2 9B for comparison. We evaluate the quality of each SAE on standard metrics and release these results. We hope that by releasing these SAE weights, we can help make more ambitious safety and interpretability research easier for the community. Weights and a tutorial can be found at https://huggingface.co/google/gemma-scope and an interactive demo can be found at https://www.neuronpedia.org/gemma-scope"
                    },
                    {
                        "title": "Explorations of Self-Repair in Language Models",
                        "abstract": "Prior interpretability research studying narrow distributions has preliminarily identified self-repair, a phenomena where if components in large language models are ablated, later components will change their behavior to compensate. Our work builds off this past literature, demonstrating that self-repair exists on a variety of models families and sizes when ablating individual attention heads on the full training distribution. We further show that on the full training distribution self-repair is imperfect, as the original direct effect of the head is not fully restored, and noisy, since the degree of self-repair varies significantly across different prompts (sometimes overcorrecting beyond the original effect). We highlight two different mechanisms that contribute to self-repair, including changes in the final LayerNorm scaling factor and sparse sets of neurons implementing Anti-Erasure. We additionally discuss the implications of these results for interpretability practitioners and close with a more speculative discussion on the mystery of why self-repair occurs in these models at all, highlighting evidence for the Iterative Inference hypothesis in language models, a framework that predicts self-repair."
                    },
                    {
                        "title": "Understanding Inverse Scaling and Emergence in Multitask Representation Learning",
                        "abstract": "Large language models exhibit strong multi-tasking capabilities, however, their learning dynamics as a function of task characteristics, sample size, and model complexity remain mysterious. For instance, it is known that, as the model size grows, large language models exhibit emerging abilities where certain tasks can abruptly jump from poor to respectable performance. Such phenomena motivate a deeper understanding of how individual tasks evolve during multitasking. To this aim, we study a multitask representation learning setup where tasks can have distinct distributions , quantified by their covariance priors. Through random matrix theory, we precisely characterize the optimal linear representation for few-shot learning that minimizes the average test risk in terms of task covariances. When tasks have equal sample sizes, we prove a reduction to an equivalent problem with a single effective covariance from which the individual task risks of the original problem can be deduced. Importantly, we introduce \u201c task competition \u201d to explain how tasks with dominant covariance eigen-spectrum emerge faster than others. We show that task competition can potentially explain the inverse scaling of certain tasks i.e. reduced test accuracy as the model grows. Overall, this work sheds light on the risk and emergence of individual tasks and uncovers new high-dimensional phenomena (including multiple-descent risk curves) that arise in multitask representation learning."
                    },
                    {
                        "title": "Confidence Regulation Neurons in Language Models",
                        "abstract": "Despite their widespread use, the mechanisms by which large language models (LLMs) represent and regulate uncertainty in next-token predictions remain largely unexplored. This study investigates two critical components believed to influence this uncertainty: the recently discovered entropy neurons and a new set of components that we term token frequency neurons. Entropy neurons are characterized by an unusually high weight norm and influence the final layer normalization (LayerNorm) scale to effectively scale down the logits. Our work shows that entropy neurons operate by writing onto an unembedding null space, allowing them to impact the residual stream norm with minimal direct effect on the logits themselves. We observe the presence of entropy neurons across a range of models, up to 7 billion parameters. On the other hand, token frequency neurons, which we discover and describe here for the first time, boost or suppress each token's logit proportionally to its log frequency, thereby shifting the output distribution towards or away from the unigram distribution. Finally, we present a detailed case study where entropy neurons actively manage confidence in the setting of induction, i.e. detecting and continuing repeated subsequences."
                    },
                    {
                        "title": "AtP*: An efficient and scalable method for localizing LLM behaviour to components",
                        "abstract": "Activation Patching is a method of directly computing causal attributions of behavior to model components. However, applying it exhaustively requires a sweep with cost scaling linearly in the number of model components, which can be prohibitively expensive for SoTA Large Language Models (LLMs). We investigate Attribution Patching (AtP), a fast gradient-based approximation to Activation Patching and find two classes of failure modes of AtP which lead to significant false negatives. We propose a variant of AtP called AtP*, with two changes to address these failure modes while retaining scalability. We present the first systematic study of AtP and alternative methods for faster activation patching and show that AtP significantly outperforms all other investigated methods, with AtP* providing further significant improvement. Finally, we provide a method to bound the probability of remaining false negatives of AtP* estimates."
                    },
                    {
                        "title": "How to use and interpret activation patching",
                        "abstract": "Activation patching is a popular mechanistic interpretability technique, but has many subtleties regarding how it is applied and how one may interpret the results. We provide a summary of advice and best practices, based on our experience using this technique in practice. We include an overview of the different ways to apply activation patching and a discussion on how to interpret the results. We focus on what evidence patching experiments provide about circuits, and on the choice of metric and associated pitfalls."
                    },
                    {
                        "title": "Improving Dictionary Learning with Gated Sparse Autoencoders",
                        "abstract": "Recent work has found that sparse autoencoders (SAEs) are an effective technique for unsupervised discovery of interpretable features in language models' (LMs) activations, by finding sparse, linear reconstructions of LM activations. We introduce the Gated Sparse Autoencoder (Gated SAE), which achieves a Pareto improvement over training with prevailing methods. In SAEs, the L1 penalty used to encourage sparsity introduces many undesirable biases, such as shrinkage -- systematic underestimation of feature activations. The key insight of Gated SAEs is to separate the functionality of (a) determining which directions to use and (b) estimating the magnitudes of those directions: this enables us to apply the L1 penalty only to the former, limiting the scope of undesirable side effects. Through training SAEs on LMs of up to 7B parameters we find that, in typical hyper-parameter ranges, Gated SAEs solve shrinkage, are similarly interpretable, and require half as many firing features to achieve comparable reconstruction fidelity."
                    },
                    {
                        "title": "Towards Principled Evaluations of Sparse Autoencoders for Interpretability and Control",
                        "abstract": "Disentangling model activations into meaningful features is a central problem in interpretability. However, the absence of ground-truth for these features in realistic scenarios makes validating recent approaches, such as sparse dictionary learning, elusive. To address this challenge, we propose a framework for evaluating feature dictionaries in the context of specific tasks, by comparing them against \\emph{supervised} feature dictionaries. First, we demonstrate that supervised dictionaries achieve excellent approximation, control, and interpretability of model computations on the task. Second, we use the supervised dictionaries to develop and contextualize evaluations of unsupervised dictionaries along the same three axes. We apply this framework to the indirect object identification (IOI) task using GPT-2 Small, with sparse autoencoders (SAEs) trained on either the IOI or OpenWebText datasets. We find that these SAEs capture interpretable features for the IOI task, but they are less successful than supervised features in controlling the model. Finally, we observe two qualitative phenomena in SAE training: feature occlusion (where a causally relevant concept is robustly overshadowed by even slightly higher-magnitude ones in the learned features), and feature over-splitting (where binary features split into many smaller, less interpretable features). We hope that our framework will provide a useful step towards more objective and grounded evaluations of sparse dictionary learning methods."
                    },
                    {
                        "title": "Summing Up the Facts: Additive Mechanisms Behind Factual Recall in LLMs",
                        "abstract": "How do transformer-based large language models (LLMs) store and retrieve knowledge? We focus on the most basic form of this task -- factual recall, where the model is tasked with explicitly surfacing stored facts in prompts of form `Fact: The Colosseum is in the country of'. We find that the mechanistic story behind factual recall is more complex than previously thought. It comprises several distinct, independent, and qualitatively different mechanisms that additively combine, constructively interfering on the correct attribute. We term this generic phenomena the additive motif: models compute through summing up multiple independent contributions. Each mechanism's contribution may be insufficient alone, but summing results in constructive interfere on the correct answer. In addition, we extend the method of direct logit attribution to attribute an attention head's output to individual source tokens. We use this technique to unpack what we call `mixed heads' -- which are themselves a pair of two separate additive updates from different source tokens."
                    },
                    {
                        "title": "Interpreting Attention Layer Outputs with Sparse Autoencoders",
                        "abstract": "Decomposing model activations into interpretable components is a key open problem in mechanistic interpretability. Sparse autoencoders (SAEs) are a popular method for decomposing the internal activations of trained transformers into sparse, interpretable features, and have been applied to MLP layers and the residual stream. In this work we train SAEs on attention layer outputs and show that also here SAEs find a sparse, interpretable decomposition. We demonstrate this on transformers from several model families and up to 2B parameters. We perform a qualitative study of the features computed by attention layers, and find multiple families: long-range context, short-range context and induction features. We qualitatively study the role of every head in GPT-2 Small, and estimate that at least 90% of the heads are polysemantic, i.e. have multiple unrelated roles. Further, we show that Sparse Autoencoders are a useful tool that enable researchers to explain model behavior in greater detail than prior work. For example, we explore the mystery of why models have so many seemingly redundant induction heads, use SAEs to motivate the hypothesis that some are long-prefix whereas others are short-prefix, and confirm this with more rigorous analysis. We use our SAEs to analyze the computation performed by the Indirect Object Identification circuit (Wang et al.), validating that the SAEs find causally meaningful intermediate variables, and deepening our understanding of the semantics of the circuit. We open-source the trained SAEs and a tool for exploring arbitrary prompts through the lens of Attention Output SAEs."
                    },
                    {
                        "title": "Linear Representations of Sentiment in Large Language Models",
                        "abstract": "Sentiment is a pervasive feature in natural language text, yet it is an open question how sentiment is represented within Large Language Models (LLMs). In this study, we reveal that across a range of models, sentiment is represented linearly: a single direction in activation space mostly captures the feature across a range of tasks with one extreme for positive and the other for negative. Through causal interventions, we isolate this direction and show it is causally relevant in both toy tasks and real world datasets such as Stanford Sentiment Treebank. Through this case study we model a thorough investigation of what a single direction means on a broad data distribution. We further uncover the mechanisms that involve this direction, highlighting the roles of a small subset of attention heads and neurons. Finally, we discover a phenomenon which we term the summarization motif: sentiment is not solely represented on emotionally charged words, but is additionally summarized at intermediate positions without inherent sentiment, such as punctuation and names. We show that in Stanford Sentiment Treebank zero-shot classification, 76% of above-chance classification accuracy is lost when ablating the sentiment direction, nearly half of which (36%) is due to ablating the summarized sentiment direction exclusively at comma positions."
                    },
                    {
                        "title": "Training Dynamics of Contextual N-Grams in Language Models",
                        "abstract": "Prior work has shown the existence of contextual neurons in language models, including a neuron that activates on German text. We show that this neuron exists within a broader contextual n-gram circuit: we find late layer neurons which recognize and continue n-grams common in German text, but which only activate if the German neuron is active. We investigate the formation of this circuit throughout training and find that it is an example of what we call a second-order circuit. In particular, both the constituent n-gram circuits and the German detection circuit which culminates in the German neuron form with independent functions early in training - the German detection circuit partially through modeling German unigram statistics, and the n-grams by boosting appropriate completions. Only after both circuits have already formed do they fit together into a second-order circuit. Contrary to the hypotheses presented in prior work, we find that the contextual n-gram circuit forms gradually rather than in a sudden phase transition. We further present a range of anomalous observations such as a simultaneous phase transition in many tasks coinciding with the learning rate warm-up, and evidence that many context neurons form simultaneously early in training but are later unlearned."
                    },
                    {
                        "title": "Copy Suppression: Comprehensively Understanding an Attention Head",
                        "abstract": "We present a single attention head in GPT-2 Small that has one main role across the entire training distribution. If components in earlier layers predict a certain token, and this token appears earlier in the context, the head suppresses it: we call this copy suppression. Attention Head 10.7 (L10H7) suppresses naive copying behavior which improves overall model calibration. This explains why multiple prior works studying certain narrow tasks found negative heads that systematically favored the wrong answer. We uncover the mechanism that the Negative Heads use for copy suppression with weights-based evidence and are able to explain 76.9% of the impact of L10H7 in GPT-2 Small. To the best of our knowledge, this is the most comprehensive description of the complete role of a component in a language model to date. One major effect of copy suppression is its role in self-repair. Self-repair refers to how ablating crucial model components results in downstream neural network parts compensating for this ablation. Copy suppression leads to self-repair: if an initial overconfident copier is ablated, then there is nothing to suppress. We show that self-repair is implemented by several mechanisms, one of which is copy suppression, which explains 39% of the behavior in a narrow task. Interactive visualisations of the copy suppression phenomena may be seen at our web app https://copy-suppression.streamlit.app/"
                    },
                    {
                        "title": "Finding Neurons in a Haystack: Case Studies with Sparse Probing",
                        "abstract": "Despite rapid adoption and deployment of large language models (LLMs), the internal computations of these models remain opaque and poorly understood. In this work, we seek to understand how high-level human-interpretable features are represented within the internal neuron activations of LLMs. We train $k$-sparse linear classifiers (probes) on these internal activations to predict the presence of features in the input; by varying the value of $k$ we study the sparsity of learned representations and how this varies with model scale. With $k=1$, we localize individual neurons which are highly relevant for a particular feature, and perform a number of case studies to illustrate general properties of LLMs. In particular, we show that early layers make use of sparse combinations of neurons to represent many features in superposition, that middle layers have seemingly dedicated neurons to represent higher-level contextual features, and that increasing scale causes representational sparsity to increase on average, but there are multiple types of scaling dynamics. In all, we probe for over 100 unique features comprising 10 different categories in 7 different models spanning 70 million to 6.9 billion parameters."
                    },
                    {
                        "title": "A Toy Model of Universality: Reverse Engineering How Networks Learn Group Operations",
                        "abstract": "Universality is a key hypothesis in mechanistic interpretability -- that different models learn similar features and circuits when trained on similar tasks. In this work, we study the universality hypothesis by examining how small neural networks learn to implement group composition. We present a novel algorithm by which neural networks may implement composition for any finite group via mathematical representation theory. We then show that networks consistently learn this algorithm by reverse engineering model logits and weights, and confirm our understanding using ablations. By studying networks of differing architectures trained on various groups, we find mixed evidence for universality: using our algorithm, we can completely characterize the family of circuits and features that networks learn on this task, but for a given network the precise circuits learned -- as well as the order they develop -- are arbitrary."
                    }
                ]
            }
        }
    },
    "2309.17167": {
        "paper_data": {
            "title": "DyVal: Dynamic Evaluation of Large Language Models for Reasoning Tasks",
            "url": "http://arxiv.org/abs/2309.17167v3",
            "arxiv_id": "2309.17167",
            "authors": [
                "Kaijie Zhu",
                "Jiaao Chen",
                "Jindong Wang",
                "Neil Zhenqiang Gong",
                "Diyi Yang",
                "Xing Xie"
            ],
            "abstract": "Large language models (LLMs) have achieved remarkable performance in various evaluation benchmarks. However, concerns are raised about potential data contamination in their considerable volume of training corpus. Moreover, the static nature and fixed complexity of current benchmarks may inadequately gauge the advancing capabilities of LLMs. In this paper, we introduce DyVal, a general and flexible protocol for dynamic evaluation of LLMs. Based on our framework, we build graph-informed DyVal by leveraging the structural advantage of directed acyclic graphs to dynamically generate evaluation samples with controllable complexities. DyVal generates challenging evaluation sets on reasoning tasks including mathematics, logical reasoning, and algorithm problems. We evaluate various LLMs ranging from Flan-T5-large to GPT-3.5-Turbo and GPT-4. Experiments show that LLMs perform worse in DyVal-generated evaluation samples with different complexities, highlighting the significance of dynamic evaluation. We also analyze the failure cases and results of different prompting methods. Moreover, DyVal-generated samples are not only evaluation sets, but also helpful data for fine-tuning to improve the performance of LLMs on existing benchmarks. We hope that DyVal can shed light on future evaluation research of LLMs. Code is available at: https://github.com/microsoft/promptbench.",
            "introduction": "   1 Introduction  Large Language Models (LLMs) have recently achieved unprecedented performance across diverse tasks (OpenAI, 2023b; Bubeck et\u00a0al., 2023). The great endeavor have led to positive speculation on the possibility of LLMs being precursors of artificial general intelligence, necessitating the creation of nuanced evaluations. By pinpointing gaps for improvements, evaluation becomes the bedrock that enhances the understanding of current models and ensures AI\u2019s continued progression.   Efforts to evaluate LLMs have become intensified significantly. Liang et\u00a0al. (2023) introduced HELM, which offers a holistic assessment of LLM in various scenarios. Similarly, Chatbot Arena (Zheng et\u00a0al., 2023) evaluates LLMs by contrasting their generated output. Other benchmarks that have set the standard in the realm of LLM evaluations include AlpacaEval (Li et\u00a0al., 2023c), C-Eval (Huang et\u00a0al., 2023), ARB (Sawada et\u00a0al., 2023), API-Bank (Li et\u00a0al., 2023a), Socket (Choi et\u00a0al., 2023), and Big-Bench (bench authors, 2023). Moreover, manual experiments have emerged as a complementary approach to these benchmarks, with works such as Bubeck et\u00a0al. (2023) and Bang et\u00a0al. (2023). Complementing these, human evaluators have also been instrumental in gauging the prowess of LLMs, as discussed by Ziems et\u00a0al. (2023) and Ze\u010devi\u0107 et\u00a0al. (2023).   Current evaluation benchmarks face two fundamental challenges. First, data contamination. Many benchmarks source their data from the Internet, causing potential overlap with the vast corpus on which LLMs are trained, leading to the debate of \u201cGeneralization vs. Memorization\u201d (Bender et\u00a0al., 2021; Magar & Schwartz, 2022; Carlini et\u00a0al., 2023; Biderman et\u00a0al., 2023): Are the model\u2019s results stemming from genuine ability or just memorization of the training data? A recent example is provided by Ze\u010devi\u0107 et\u00a0al. (2023): LLMs can ambiguously deduce the conclusion that altitude influences temperature based on seen data. Similarly, Berglund et\u00a0al. (2023) found that LLMs trained on \u201cA is B\u201d fail to infer \u201cB is A\u201d, which doubts the abilities of LLMs might come from memorization. Second, static dataset and fixed complexity. As LLMs progress at a rapid pace, existing datasets usually fail to match the models\u2019 ever-evolving capabilities, because the complexity level of existing benchmarks is usually static and fixed. As Dziri et\u00a0al. (2023) demonstrated, while handling simple problems pretty well, LLMs fail to solve complex problems. The inability to automatically and dynamically increase complexity levels based on existing data prevents static benchmarks from being adapted to accurately select, compare, and advance LLMs. Although there are a few existing dynamic benchmarks like DynaBench\u00a0(Kiela et\u00a0al., 2021) and DynaBoard\u00a0(Ma et\u00a0al., 2021), they rely on crowd-sourcing efforts for data collection, which might be expensive and tedious.   In this paper, we introduce DyVal\u2014a novel, general, and flexible evaluation protocol for the dynamic evaluation of LLMs (Sec. 3.1). The core of DyVal is to dynamically generate evaluation samples on the fly instead of collecting a fixed set of data. DyVal consists of three components: 1) the generation algorithm \ud835\udca2\ud835\udca2\\mathcal{G}caligraphic_G to generate test samples with diversities; 2) the constraint \ud835\udc9e\ud835\udc9e\\mathcal{C}caligraphic_C to modulate sample complexity and validity; and 3) the description function \u2131\u2131\\mathcal{F}caligraphic_F to translate the generated samples into natural languages. Based on this framework, we propose a graph-informed DyVal (Sec. 3.2, Figure\u00a01) to generate data using graphs. Specifically, inspired by techniques such as the compiler principle (Alfred\u00a0V et\u00a0al., 2007) and parsing trees which decompose complexities",
            "references": [
                {
                    "title": "Data Contamination Quiz: A Tool to Detect and Estimate Contamination in Large Language Models",
                    "abstract": "We propose the Data Contamination Quiz (DCQ), a simple and effective approach to detect data contamination in large language models (LLMs) and estimate the amount of it. Specifically, we frame data contamination detection as a series of multiple-choice questions and devise a quiz format wherein three perturbed versions of each subsampled instance from a specific dataset partition (e.g., GSM8k test set) are created. These changes only include word-level perturbations. The generated perturbations, along with the original dataset instance, form the options in the DCQ, with an extra option accommodating the possibility of selecting none of the provided options. Given that the only distinguishing signal among the options is the exact wording with respect to the original dataset instance, an LLM, when tasked with identifying the original dataset instance, gravitates towards selecting the original one if it has been exposed to it in its pre-training phase -- a trait intrinsic to LLMs. While accounting for positional biases in LLMs, the quiz performance reveals the contamination level for the model being examined with the dataset partition to which the quiz pertains. Applied to various datasets with GPT-4 and GPT-3.5, our findings -- while fully lacking access to pre-training data and model parameters -- suggest that DCQ achieves state-of-the-art results and uncovers greater contamination/memorization levels compared to existing methods and proficiently bypasses more safety filters, especially those set to avoid generating copyrighted contents."
                },
                {
                    "title": "Rethinking Benchmark and Contamination for Language Models with Rephrased Samples",
                    "abstract": "Large language models are increasingly trained on all the data ever produced by humans. Many have raised concerns about the trustworthiness of public benchmarks due to potential contamination in pre-training or fine-tuning datasets. While most data decontamination efforts apply string matching (e.g., n-gram overlap) to remove benchmark data, we show that these methods are insufficient, and simple variations of test data (e.g., paraphrasing, translation) can easily bypass these decontamination measures. Furthermore, we demonstrate that if such variation of test data is not eliminated, a 13B model can easily overfit a test benchmark and achieve drastically high performance, on par with GPT-4. We validate such observations in widely used benchmarks such as MMLU, GSK8k, and HumanEval. To address this growing risk, we propose a stronger LLM-based decontamination method and apply it to widely used pre-training and fine-tuning datasets, revealing significant previously unknown test overlap. For example, in pre-training sets such as RedPajama-Data-1T and StarCoder-Data, we identified that 8-18\\% of the HumanEval benchmark overlaps. Interestingly, we also find such contamination in synthetic dataset generated by GPT-3.5/4, suggesting a potential risk of unintentional contamination. We urge the community to adopt stronger decontamination approaches when using public benchmarks. Moreover, we call for the community to actively develop fresh one-time exams to evaluate models accurately. Our decontamination tool is publicly available at https://github.com/lm-sys/llm-decontaminator."
                },
                {
                    "title": "Don't Make Your LLM an Evaluation Benchmark Cheater",
                    "abstract": "Large language models~(LLMs) have greatly advanced the frontiers of artificial intelligence, attaining remarkable improvement in model capacity. To assess the model performance, a typical approach is to construct evaluation benchmarks for measuring the ability level of LLMs in different aspects. Despite that a number of high-quality benchmarks have been released, the concerns about the appropriate use of these benchmarks and the fair comparison of different models are increasingly growing. Considering these concerns, in this paper, we discuss the potential risk and impact of inappropriately using evaluation benchmarks and misleadingly interpreting the evaluation results. Specially, we focus on a special issue that would lead to inappropriate evaluation, \\ie \\emph{benchmark leakage}, referring that the data related to evaluation sets is occasionally used for model training. This phenomenon now becomes more common since pre-training data is often prepared ahead of model test. We conduct extensive experiments to study the effect of benchmark leverage, and find that it can dramatically boost the evaluation results, which would finally lead to an unreliable assessment of model performance. To improve the use of existing evaluation benchmarks, we finally present several guidelines for both LLM developers and benchmark maintainers. We hope this work can draw attention to appropriate training and evaluation of LLMs."
                },
                {
                    "title": "Skywork: A More Open Bilingual Foundation Model",
                    "abstract": "In this technical report, we present Skywork-13B, a family of large language models (LLMs) trained on a corpus of over 3.2 trillion tokens drawn from both English and Chinese texts. This bilingual foundation model is the most extensively trained and openly published LLMs of comparable size to date. We introduce a two-stage training methodology using a segmented corpus, targeting general purpose training and then domain-specific enhancement training, respectively. We show that our model not only excels on popular benchmarks, but also achieves \\emph{state of the art} performance in Chinese language modeling on diverse domains. Furthermore, we propose a novel leakage detection method, demonstrating that test data contamination is a pressing issue warranting further investigation by the LLM community. To spur future research, we release Skywork-13B along with checkpoints obtained during intermediate stages of the training process. We are also releasing part of our SkyPile corpus, a collection of over 150 billion tokens of web text, which is the largest high quality open Chinese pre-training corpus to date. We hope Skywork-13B and our open corpus will serve as a valuable open-source resource to democratize access to high-quality LLMs."
                },
                {
                    "title": "Proving Test Set Contamination in Black Box Language Models",
                    "abstract": "Large language models are trained on vast amounts of internet data, prompting concerns and speculation that they have memorized public benchmarks. Going from speculation to proof of contamination is challenging, as the pretraining data used by proprietary models are often not publicly accessible. We show that it is possible to provide provable guarantees of test set contamination in language models without access to pretraining data or model weights. Our approach leverages the fact that when there is no data contamination, all orderings of an exchangeable benchmark should be equally likely. In contrast, the tendency for language models to memorize example order means that a contaminated language model will find certain canonical orderings to be much more likely than others. Our test flags potential contamination whenever the likelihood of a canonically ordered benchmark dataset is significantly higher than the likelihood after shuffling the examples. We demonstrate that our procedure is sensitive enough to reliably prove test set contamination in challenging situations, including models as small as 1.4 billion parameters, on small test sets of only 1000 examples, and datasets that appear only a few times in the pretraining corpus. Using our test, we audit five popular publicly accessible language models for test set contamination and find little evidence for pervasive contamination."
                },
                {
                    "title": "Physics of Language Models: Part 3.1, Knowledge Storage and Extraction",
                    "abstract": "Large language models (LLMs) can store a vast amount of world knowledge, often extractable via question-answering (e.g.,\"What is Abraham Lincoln's birthday?\"). However, do they answer such questions based on exposure to similar questions during training (i.e., cheating), or by genuinely learning to extract knowledge from sources like Wikipedia? In this paper, we investigate this issue using a controlled biography dataset. We find a strong correlation between the model's ability to extract knowledge and various diversity measures of the training data. $\\textbf{Essentially}$, for knowledge to be reliably extracted, it must be sufficiently augmented (e.g., through paraphrasing, sentence shuffling, translations) $\\textit{during pretraining}$. Without such augmentation, knowledge may be memorized but not extractable, leading to 0% accuracy, regardless of subsequent instruction fine-tuning. To understand why this occurs, we employ (nearly) linear probing to demonstrate a strong connection between the observed correlation and how the model internally encodes knowledge -- whether it is linearly encoded in the hidden embeddings of entity names or distributed across other token embeddings in the training text. This paper provides $\\textbf{several key recommendations for LLM pretraining in the industry}$: (1) rewrite the pretraining data -- using small, auxiliary models -- to provide knowledge augmentation, and (2) incorporate more instruction-finetuning data into the pretraining stage before it becomes too late."
                },
                {
                    "title": "The Reversal Curse: LLMs trained on \"A is B\" fail to learn \"B is A\"",
                    "abstract": "We expose a surprising failure of generalization in auto-regressive large language models (LLMs). If a model is trained on a sentence of the form\"A is B\", it will not automatically generalize to the reverse direction\"B is A\". This is the Reversal Curse. For instance, if a model is trained on\"Valentina Tereshkova was the first woman to travel to space\", it will not automatically be able to answer the question,\"Who was the first woman to travel to space?\". Moreover, the likelihood of the correct answer (\"Valentina Tershkova\") will not be higher than for a random name. Thus, models do not generalize a prevalent pattern in their training set: if\"A is B\"occurs,\"B is A\"is more likely to occur. It is worth noting, however, that if\"A is B\"appears in-context, models can deduce the reverse relationship. We provide evidence for the Reversal Curse by finetuning GPT-3 and Llama-1 on fictitious statements such as\"Uriah Hawthorne is the composer of Abyssal Melodies\"and showing that they fail to correctly answer\"Who composed Abyssal Melodies?\". The Reversal Curse is robust across model sizes and model families and is not alleviated by data augmentation. We also evaluate ChatGPT (GPT-3.5 and GPT-4) on questions about real-world celebrities, such as\"Who is Tom Cruise's mother? [A: Mary Lee Pfeiffer]\"and the reverse\"Who is Mary Lee Pfeiffer's son?\". GPT-4 correctly answers questions like the former 79% of the time, compared to 33% for the latter. Code available at: https://github.com/lukasberglund/reversal_curse."
                },
                {
                    "title": "Pretraining on the Test Set Is All You Need",
                    "abstract": "Inspired by recent work demonstrating the promise of smaller Transformer-based language models pretrained on carefully curated data, we supercharge such approaches by investing heavily in curating a novel, high quality, non-synthetic data mixture based solely on evaluation benchmarks. Using our novel dataset mixture consisting of less than 100 thousand tokens, we pretrain a 1 million parameter transformer-based LLM \\textbf{phi-CTNL} (pronounced ``fictional\") that achieves perfect results across diverse academic benchmarks, strictly outperforming all known foundation models. \\textbf{phi-CTNL} also beats power-law scaling and exhibits a never-before-seen grokking-like ability to accurately predict downstream evaluation benchmarks' canaries."
                },
                {
                    "title": "Textbooks Are All You Need II: phi-1.5 technical report",
                    "abstract": "We continue the investigation into the power of smaller Transformer-based language models as initiated by \\textbf{TinyStories} -- a 10 million parameter model that can produce coherent English -- and the follow-up work on \\textbf{phi-1}, a 1.3 billion parameter model with Python coding performance close to the state-of-the-art. The latter work proposed to use existing Large Language Models (LLMs) to generate ``textbook quality\"data as a way to enhance the learning process compared to traditional web data. We follow the ``Textbooks Are All You Need\"approach, focusing this time on common sense reasoning in natural language, and create a new 1.3 billion parameter model named \\textbf{phi-1.5}, with performance on natural language tasks comparable to models 5x larger, and surpassing most non-frontier LLMs on more complex reasoning tasks such as grade-school mathematics and basic coding. More generally, \\textbf{phi-1.5} exhibits many of the traits of much larger LLMs, both good -- such as the ability to ``think step by step\"or perform some rudimentary in-context learning -- and bad, including hallucinations and the potential for toxic and biased generations -- encouragingly though, we are seeing improvement on that front thanks to the absence of web data. We open-source \\textbf{phi-1.5} to promote further research on these urgent topics."
                },
                {
                    "title": "Causal Parrots: Large Language Models May Talk Causality But Are Not Causal",
                    "abstract": "Some argue scale is all what is needed to achieve AI, covering even causal models. We make it clear that large language models (LLMs) cannot be causal and give reason onto why sometimes we might feel otherwise. To this end, we define and exemplify a new subgroup of Structural Causal Model (SCM) that we call meta SCM which encode causal facts about other SCM within their variables. We conjecture that in the cases where LLM succeed in doing causal inference, underlying was a respective meta SCM that exposed correlations between causal facts in natural language on whose data the LLM was ultimately trained. If our hypothesis holds true, then this would imply that LLMs are like parrots in that they simply recite the causal knowledge embedded in the data. Our empirical analysis provides favoring evidence that current LLMs are even weak `causal parrots.'"
                },
                {
                    "title": "WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct",
                    "abstract": "Large language models (LLMs), such as GPT-4, have shown remarkable performance in natural language processing (NLP) tasks, including challenging mathematical reasoning. However, most existing open-source models are only pre-trained on large-scale internet data and without math-related optimization. In this paper, we present WizardMath, which enhances the mathematical reasoning abilities of Llama-2, by applying our proposed Reinforcement Learning from Evol-Instruct Feedback (RLEIF) method to the domain of math. Through extensive experiments on two mathematical reasoning benchmarks, namely GSM8k and MATH, we reveal the extraordinary capabilities of our model. WizardMath surpasses all other open-source LLMs by a substantial margin. Furthermore, our model even outperforms ChatGPT-3.5, Claude Instant-1, PaLM-2 and Minerva on GSM8k, simultaneously surpasses Text-davinci-002, PaLM-1 and GPT-3 on MATH. More details and model weights are public at https://github.com/nlpxucan/WizardLM and https://huggingface.co/WizardLM."
                },
                {
                    "title": "Time Travel in LLMs: Tracing Data Contamination in Large Language Models",
                    "abstract": "Data contamination, i.e., the presence of test data from downstream tasks in the training data of large language models (LLMs), is a potential major issue in measuring LLMs' real effectiveness on other tasks. We propose a straightforward yet effective method for identifying data contamination within LLMs. At its core, our approach starts by identifying potential contamination at the instance level; using this information, our approach then assesses wider contamination at the partition level. To estimate contamination of individual instances, we employ\"guided instruction:\"a prompt consisting of the dataset name, partition type, and the random-length initial segment of a reference instance, asking the LLM to complete it. An instance is flagged as contaminated if the LLM's output either exactly or nearly matches the latter segment of the reference. To understand if an entire partition is contaminated, we propose two ideas. The first idea marks a dataset partition as contaminated if the average overlap score with the reference instances (as measured by ROUGE-L or BLEURT) is statistically significantly better with the completions from guided instruction compared to a\"general instruction\"that does not include the dataset and partition name. The second idea marks a dataset partition as contaminated if a classifier based on GPT-4 with few-shot in-context learning prompt marks multiple generated completions as exact/near-exact matches of the corresponding reference instances. Our best method achieves an accuracy between 92% and 100% in detecting if an LLM is contaminated with seven datasets, containing train and test/validation partitions, when contrasted with manual evaluation by human experts. Further, our findings indicate that GPT-4 is contaminated with AG News, WNLI, and XSum datasets."
                },
                {
                    "title": "Cumulative Reasoning with Large Language Models",
                    "abstract": "Despite the recent advancements in language models (LMs), their ability to solve complex problems remains limited. This paper introduces Cumulative Reasoning (CR), a novel approach that utilizes LMs cumulatively and iteratively, mirroring human thought processes for problem-solving. CR decomposes tasks into smaller, manageable components and leverages previous propositions for effective composition, significantly enhancing problem-solving capabilities. We demonstrate CR's superiority through several complex reasoning tasks: it outperforms existing methods in logical inference tasks with up to a 9.3% improvement, achieving 98.04% accuracy on the curated FOLIO wiki dataset. In the Game of 24, it achieves 98% accuracy, marking a 24% improvement over the prior state-of-the-art. Additionally, CR sets new state-of-the-art on the MATH dataset, achieving a 4.2% increase from previous methods and a 43% relative improvement in the most challenging problems. By extending CR to incorporate a code environment without external aids like retrieval or web browsing, we further harness the computational and logical reasoning capabilities of LMs, achieving a remarkable 72.2% accuracy on the MATH dataset and outperforming the PAL/PoT method by 38.8%. Our work not only sets new state-of-the-art but also paves the way toward more sophisticated AI reasoning methods. The code is available at https://github.com/iiis-ai/cumulative-reasoning."
                },
                {
                    "title": "Skills-in-Context Prompting: Unlocking Compositionality in Large Language Models",
                    "abstract": "We investigate how to elicit compositional generalization capabilities in large language models (LLMs). Compositional generalization empowers LLMs to solve complex problems by combining foundational skills, a critical reasoning ability akin to human intelligence. However, even the most advanced LLMs currently struggle with this form of reasoning. We examine this problem within the framework of in-context learning and find that demonstrating both foundational skills and compositional examples grounded in these skills within the same prompt context is crucial. We refer to this prompt structure as skills-in-context (SKiC). With as few as two exemplars, this in-context learning structure enables LLMs to tackle more challenging problems requiring innovative skill combinations, achieving near-perfect systematic generalization across a broad range of tasks. Intriguingly, SKiC also unlocks the latent potential of LLMs, allowing them to more actively utilize pre-existing internal skills acquired during earlier pretraining stages to solve complex reasoning problems. The SKiC structure is robust across different skill constructions and exemplar choices and demonstrates strong transferability to new tasks. Finally, inspired by our in-context learning study, we show that fine-tuning LLMs with SKiC-style data can elicit zero-shot weak-to-strong generalization, enabling the models to solve much harder problems directly with standard prompting."
                },
                {
                    "title": "ARB: Advanced Reasoning Benchmark for Large Language Models",
                    "abstract": "Large Language Models (LLMs) have demonstrated remarkable performance on various quantitative reasoning and knowledge benchmarks. However, many of these benchmarks are losing utility as LLMs get increasingly high scores, despite not yet reaching expert performance in these domains. We introduce ARB, a novel benchmark composed of advanced reasoning problems in multiple fields. ARB presents a more challenging test than prior benchmarks, featuring problems in mathematics, physics, biology, chemistry, and law. As a subset of ARB, we introduce a challenging set of math and physics problems which require advanced symbolic reasoning and domain knowledge. We evaluate recent models such as GPT-4 and Claude on ARB and demonstrate that current models score well below 50% on more demanding tasks. In order to improve both automatic and assisted evaluation capabilities, we introduce a rubric-based evaluation approach, allowing GPT-4 to score its own intermediate reasoning steps. Further, we conduct a human evaluation of the symbolic subset of ARB, finding promising agreement between annotators and GPT-4 rubric evaluation scores."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks",
                    "abstract": "The impressive performance of recent language models across a wide range of tasks suggests that they possess a degree of abstract reasoning skills. Are these skills general and transferable, or specialized to specific tasks seen during pretraining? To disentangle these effects, we propose an evaluation framework based on \u201ccounterfactual\u201d task variants that deviate from the default assumptions underlying standard tasks. Across a suite of 11 tasks, we observe nontrivial performance on the counterfactual variants, but nevertheless find that performance substantially and consistently degrades compared to the default conditions. This suggests that while current LMs may possess abstract task-solving skills to an extent, they often also rely on narrow, non-transferable procedures for task-solving. These results motivate a more careful interpretation of language model performance that teases apart these aspects."
                },
                {
                    "title": "Can Large Language Models Infer Causation from Correlation?",
                    "abstract": "Causal inference is one of the hallmarks of human intelligence. While the field of CausalNLP has attracted much interest in the recent years, existing causal inference datasets in NLP primarily rely on discovering causality from empirical knowledge (e.g., commonsense knowledge). In this work, we propose the first benchmark dataset to test the pure causal inference skills of large language models (LLMs). Specifically, we formulate a novel task Corr2Cause, which takes a set of correlational statements and determines the causal relationship between the variables. We curate a large-scale dataset of more than 200K samples, on which we evaluate seventeen existing LLMs. Through our experiments, we identify a key shortcoming of LLMs in terms of their causal inference skills, and show that these models achieve almost close to random performance on the task. This shortcoming is somewhat mitigated when we try to re-purpose LLMs for this skill via finetuning, but we find that these models still fail to generalize -- they can only perform causal inference in in-distribution settings when variable names and textual expressions used in the queries are similar to those in the training set, but fail in out-of-distribution settings generated by perturbing these queries. Corr2Cause is a challenging task for LLMs, and would be helpful in guiding future research on improving LLMs' pure reasoning skills and generalizability. Our data is at https://huggingface.co/datasets/causalnlp/corr2cause. Our code is at https://github.com/causalNLP/corr2cause."
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "PromptRobust: Towards Evaluating the Robustness of Large Language Models on Adversarial Prompts",
                    "abstract": "The increasing reliance on Large Language Models (LLMs) across academia and industry necessitates a comprehensive understanding of their robustness to prompts. In response to this vital need, we introduce PromptRobust, a robustness benchmark designed to measure LLMs' resilience to adversarial prompts. This study uses a plethora of adversarial textual attacks targeting prompts across multiple levels: character, word, sentence, and semantic. The adversarial prompts, crafted to mimic plausible user errors like typos or synonyms, aim to evaluate how slight deviations can affect LLM outcomes while maintaining semantic integrity. These prompts are then employed in diverse tasks including sentiment analysis, natural language inference, reading comprehension, machine translation, and math problem-solving. Our study generates 4,788 adversarial prompts, meticulously evaluated over 8 tasks and 13 datasets. Our findings demonstrate that contemporary LLMs are not robust to adversarial prompts. Furthermore, we present a comprehensive analysis to understand the mystery behind prompt robustness and its transferability. We then offer insightful robustness analysis and pragmatic recommendations for prompt composition, beneficial to both researchers and everyday users."
                },
                {
                    "title": "Faith and Fate: Limits of Transformers on Compositionality",
                    "abstract": "Transformer large language models (LLMs) have sparked admiration for their exceptional performance on tasks that demand intricate multi-step reasoning. Yet, these models simultaneously show failures on surprisingly trivial problems. This begs the question: Are these errors incidental, or do they signal more substantial limitations? In an attempt to demystify transformer LLMs, we investigate the limits of these models across three representative compositional tasks -- multi-digit multiplication, logic grid puzzles, and a classic dynamic programming problem. These tasks require breaking problems down into sub-steps and synthesizing these steps into a precise answer. We formulate compositional tasks as computation graphs to systematically quantify the level of complexity, and break down reasoning steps into intermediate sub-procedures. Our empirical findings suggest that transformer LLMs solve compositional tasks by reducing multi-step compositional reasoning into linearized subgraph matching, without necessarily developing systematic problem-solving skills. To round off our empirical study, we provide theoretical arguments on abstract multi-step reasoning problems that highlight how autoregressive generations' performance can rapidly decay with\\,increased\\,task\\,complexity."
                },
                {
                    "title": "Holistic Evaluation of Language Models",
                    "abstract": "Language models (LMs) like GPT\u20103, PaLM, and ChatGPT are the foundation for almost all major language technologies, but their capabilities, limitations, and risks are not well understood. We present Holistic Evaluation of Language Models (HELM) to improve the transparency of LMs. LMs can serve many purposes and their behavior should satisfy many desiderata. To navigate the vast space of potential scenarios and metrics, we taxonomize the space and select representative subsets. We evaluate models on 16 core scenarios and 7 metrics, exposing important trade\u2010offs. We supplement our core evaluation with seven targeted evaluations to deeply analyze specific aspects (including world knowledge, reasoning, regurgitation of copyrighted content, and generation of disinformation). We benchmark 30 LMs, from OpenAI, Microsoft, Google, Meta, Cohere, AI21 Labs, and others. Prior to HELM, models were evaluated on just 17.9% of the core HELM scenarios, with some prominent models not sharing a single scenario in common. We improve this to 96.0%: all 30 models are now benchmarked under the same standardized conditions. Our evaluation surfaces 25 top\u2010level findings. For full transparency, we release all raw model prompts and completions publicly. HELM is a living benchmark for the community, continuously updated with new scenarios, metrics, and models https://crfm.stanford.edu/helm/latest/."
                },
                {
                    "title": "Large Language Models are In-Context Semantic Reasoners rather than Symbolic Reasoners",
                    "abstract": "The emergent few-shot reasoning capabilities of Large Language Models (LLMs) have excited the natural language and machine learning community over recent years. Despite of numerous successful applications, the underlying mechanism of such in-context capabilities still remains unclear. In this work, we hypothesize that the learned \\textit{semantics} of language tokens do the most heavy lifting during the reasoning process. Different from human's symbolic reasoning process, the semantic representations of LLMs could create strong connections among tokens, thus composing a superficial logical chain. To test our hypothesis, we decouple semantics from the language reasoning process and evaluate three kinds of reasoning abilities, i.e., deduction, induction and abduction. Our findings reveal that semantics play a vital role in LLMs' in-context reasoning -- LLMs perform significantly better when semantics are consistent with commonsense but struggle to solve symbolic or counter-commonsense reasoning tasks by leveraging in-context new knowledge. The surprising observations question whether modern LLMs have mastered the inductive, deductive and abductive reasoning abilities as in human intelligence, and motivate research on unveiling the magic existing within the black-box LLMs. On the whole, our analysis provides a novel perspective on the role of semantics in developing and evaluating language models' reasoning abilities. Code is available at {\\url{https://github.com/XiaojuanTang/ICSR}}."
                },
                {
                    "title": "Do LLMs Understand Social Knowledge? Evaluating the Sociability of Large Language Models with SocKET Benchmark",
                    "abstract": "Large language models (LLMs) have been shown to perform well at a variety of syntactic, discourse, and reasoning tasks. While LLMs are increasingly deployed in many forms including conversational agents that interact with humans, we lack a grounded benchmark to measure how well LLMs understand \\textit{social} language. Here, we introduce a new theory-driven benchmark, SocKET, that contains 58 NLP tasks testing social knowledge which we group into five categories: humor&sarcasm, offensiveness, sentiment&emotion, and trustworthiness. In tests on the benchmark, we demonstrate that current models attain only moderate performance but reveal significant potential for task transfer among different types and categories of tasks, which were predicted from theory. Through zero-shot evaluations, we show that pretrained models already possess some innate but limited capabilities of social language understanding and training on one category of tasks can improve zero-shot testing on others. Our benchmark provides a systematic way to analyze model performance on an important dimension of language and points to clear room for improvement to build more socially-aware LLMs. The associated resources are released at https://github.com/minjechoi/SOCKET."
                },
                {
                    "title": "C-Eval: A Multi-Level Multi-Discipline Chinese Evaluation Suite for Foundation Models",
                    "abstract": "New NLP benchmarks are urgently needed to align with the rapid development of large language models (LLMs). We present C-Eval, the first comprehensive Chinese evaluation suite designed to assess advanced knowledge and reasoning abilities of foundation models in a Chinese context. C-Eval comprises multiple-choice questions across four difficulty levels: middle school, high school, college, and professional. The questions span 52 diverse disciplines, ranging from humanities to science and engineering. C-Eval is accompanied by C-Eval Hard, a subset of very challenging subjects in C-Eval that requires advanced reasoning abilities to solve. We conduct a comprehensive evaluation of the most advanced LLMs on C-Eval, including both English- and Chinese-oriented models. Results indicate that only GPT-4 could achieve an average accuracy of over 60%, suggesting that there is still significant room for improvement for current LLMs. We anticipate C-Eval will help analyze important strengths and shortcomings of foundation models, and foster their development and growth for Chinese users."
                },
                {
                    "title": "Emergent and Predictable Memorization in Large Language Models",
                    "abstract": "Memorization, or the tendency of large language models (LLMs) to output entire sequences from their training data verbatim, is a key concern for safely deploying language models. In particular, it is vital to minimize a model's memorization of sensitive datapoints such as those containing personal identifiable information (PII). The prevalence of such undesirable memorization can pose issues for model trainers, and may even require discarding an otherwise functional model. We therefore seek to predict which sequences will be memorized before a large model's full train-time by extrapolating the memorization behavior of lower-compute trial runs. We measure memorization of the Pythia model suite and plot scaling laws for forecasting memorization, allowing us to provide equi-compute recommendations to maximize the reliability (recall) of such predictions. We additionally provide further novel discoveries on the distribution of memorization scores across models and data. We release all code and data necessary to reproduce the results in this paper at https://github.com/EleutherAI/pythia"
                },
                {
                    "title": "AGIEval: A Human-Centric Benchmark for Evaluating Foundation Models",
                    "abstract": "Evaluating the general abilities of foundation models to tackle human-level tasks is a vital aspect of their development and application in the pursuit of Artificial General Intelligence (AGI). Traditional benchmarks, which rely on artificial datasets, may not accurately represent human-level capabilities. In this paper, we introduce AGIEval, a novel benchmark specifically designed to assess foundation model in the context of human-centric standardized exams, such as college entrance exams, law school admission tests, math competitions, and lawyer qualification tests. We evaluate several state-of-the-art foundation models, including GPT-4, ChatGPT, and Text-Davinci-003, using this benchmark. Impressively, GPT-4 surpasses average human performance on SAT, LSAT, and math competitions, attaining a 95% accuracy rate on the SAT Math test and a 92.5% accuracy on the English test of the Chinese national college entrance exam. This demonstrates the extraordinary performance of contemporary foundation models. In contrast, we also find that GPT-4 is less proficient in tasks that require complex reasoning or specific domain knowledge. Our comprehensive analyses of model capabilities (understanding, knowledge, reasoning, and calculation) reveal these models' strengths and limitations, providing valuable insights into future directions for enhancing their general capabilities. By concentrating on tasks pertinent to human cognition and decision-making, our benchmark delivers a more meaningful and robust evaluation of foundation models' performance in real-world scenarios. The data, code, and all model outputs are released in https://github.com/ruixiangcui/AGIEval."
                },
                {
                    "title": "Can Large Language Models Transform Computational Social Science?",
                    "abstract": "Large language models (LLMs) are capable of successfully performing many language processing tasks zero-shot (without training data). If zero-shot LLMs can also reliably classify and explain social phenomena like persuasiveness and political ideology, then LLMs could augment the computational social science (CSS) pipeline in important ways. This work provides a road map for using LLMs as CSS tools. Towards this end, we contribute a set of prompting best practices and an extensive evaluation pipeline to measure the zero-shot performance of 13 language models on 25 representative English CSS benchmarks. On taxonomic labeling tasks (classification), LLMs fail to outperform the best fine-tuned models but still achieve fair levels of agreement with humans. On free-form coding tasks (generation), LLMs produce explanations that often exceed the quality of crowdworkers\u2019 gold references. We conclude that the performance of today\u2019s LLMs can augment the CSS research pipeline in two ways: (1) serving as zero-shot data annotators on human annotation teams, and (2) bootstrapping challenging creative generation tasks (e.g., explaining the underlying attributes of a text). In summary, LLMs are posed to meaningfully participate in social science analysis in partnership with humans."
                },
                {
                    "title": "Sparks of Artificial General Intelligence: Early experiments with GPT-4",
                    "abstract": "Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. The latest model developed by OpenAI, GPT-4, was trained using an unprecedented scale of compute and data. In this paper, we report on our investigation of an early version of GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT-4 is part of a new cohort of LLMs (along with ChatGPT and Google's PaLM for example) that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4's performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4's capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system. In our exploration of GPT-4, we put special emphasis on discovering its limitations, and we discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for pursuing a new paradigm that moves beyond next-word prediction. We conclude with reflections on societal influences of the recent technological leap and future research directions."
                },
                {
                    "title": "A Multitask, Multilingual, Multimodal Evaluation of ChatGPT on Reasoning, Hallucination, and Interactivity",
                    "abstract": "This paper proposes a framework for quantitatively evaluating interactive LLMs such as ChatGPT using publicly available data sets. We carry out an extensive technical evaluation of ChatGPT using 23 data sets covering 8 different common NLP application tasks. We evaluate the multitask, multilingual and multi-modal aspects of ChatGPT based on these data sets and a newly designed multimodal dataset. We find that ChatGPT outperforms LLMs with zero-shot learning on most tasks and even outperforms fine-tuned models on some tasks. We find that it is better at understanding non-Latin script languages than generating them. It is able to generate multimodal content from textual prompts, via an intermediate code generation step. Moreover, we find that ChatGPT is 63.41% accurate on average in 10 different reasoning categories under logical reasoning, non-textual reasoning, and commonsense reasoning, hence making it an unreliable reasoner. It is, for example, better at deductive than inductive reasoning. ChatGPT suffers from hallucination problems like other LLMs and it generates more extrinsic hallucinations from its parametric memory as it does not have access to an external knowledge base. Finally, the interactive feature of ChatGPT enables human collaboration with the underlying LLM to improve its performance, i.e, 8% ROUGE-1 on summarization and 2% ChrF++ on machine translation, in a multi-turn\"prompt engineering\"fashion. We also release codebase for evaluation set extraction."
                },
                {
                    "title": "Adaptive Testing of Computer Vision Models",
                    "abstract": "Vision models often fail systematically on groups of data that share common semantic characteristics (e.g., rare objects or unusual scenes), but identifying these failure modes is a challenge. We introduce AdaVision, an interactive process for testing vision models which helps users identify and fix coherent failure modes. Given a natural language description of a coherent group, AdaVision retrieves relevant images from LAION-5B with CLIP. The user then labels a small amount of data for model correctness, which is used in successive retrieval rounds to hill-climb towards high-error regions, refining the group definition. Once a group is saturated, AdaVision uses GPT-3 to suggest new group descriptions for the user to explore. We demonstrate the usefulness and generality of AdaVision in user studies, where users find major bugs in state-of-the-art classification, object detection, and image captioning models. These user-discovered groups have failure rates 2-3x higher than those surfaced by automatic error clustering methods. Finally, finetuning on examples found with AdaVision fixes the discovered bugs when evaluated on unseen examples, without degrading in-distribution accuracy, and while also improving performance on out-of-distribution datasets."
                },
                {
                    "title": "GLUE-X: Evaluating Natural Language Understanding Models from an Out-of-distribution Generalization Perspective",
                    "abstract": "Pre-trained language models (PLMs) are known to improve the generalization performance of natural language understanding models by leveraging large amounts of data during the pre-training phase. However, the out-of-distribution (OOD) generalization problem remains a challenge in many NLP tasks, limiting the real-world deployment of these methods. This paper presents the first attempt at creating a unified benchmark named GLUE-X for evaluating OOD robustness in NLP models, highlighting the importance of OOD robustness and providing insights on how to measure the robustness of a model and how to improve it. The benchmark includes 13 publicly available datasets for OOD testing, and evaluations are conducted on 8 classic NLP tasks over 21 popularly used PLMs, including GPT-3 and GPT-3.5. Our findings confirm the need for improved OOD accuracy in NLP tasks, as significant performance degradation was observed in all settings compared to in-distribution (ID) accuracy."
                },
                {
                    "title": "Large Language Models Are Human-Level Prompt Engineers",
                    "abstract": "By conditioning on natural language instructions, large language models (LLMs) have displayed impressive capabilities as general-purpose computers. However, task performance depends significantly on the quality of the prompt used to steer the model, and most effective prompts have been handcrafted by humans. Inspired by classical program synthesis and the human approach to prompt engineering, we propose Automatic Prompt Engineer (APE) for automatic instruction generation and selection. In our method, we treat the instruction as the\"program,\"optimized by searching over a pool of instruction candidates proposed by an LLM in order to maximize a chosen score function. To evaluate the quality of the selected instruction, we evaluate the zero-shot performance of another LLM following the selected instruction. Experiments on 24 NLP tasks show that our automatically generated instructions outperform the prior LLM baseline by a large margin and achieve better or comparable performance to the instructions generated by human annotators on 19/24 tasks. We conduct extensive qualitative and quantitative analyses to explore the performance of APE. We show that APE-engineered prompts can be applied to steer models toward truthfulness and/or informativeness, as well as to improve few-shot learning performance by simply prepending them to standard in-context learning prompts. Please check out our webpage at https://sites.google.com/view/automatic-prompt-engineer."
                },
                {
                    "title": "Scaling Instruction-Finetuned Language Models",
                    "abstract": "Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models."
                },
                {
                    "title": "Decomposed Prompting: A Modular Approach for Solving Complex Tasks",
                    "abstract": "Few-shot prompting is a surprisingly powerful way to use Large Language Models (LLMs) to solve various tasks. However, this approach struggles as the task complexity increases or when the individual reasoning steps of the task themselves are hard to learn, especially when embedded in more complex tasks. To address this, we propose Decomposed Prompting, a new approach to solve complex tasks by decomposing them (via prompting) into simpler sub-tasks that can be delegated to a library of prompting-based LLMs dedicated to these sub-tasks. This modular structure allows each prompt to be optimized for its specific sub-task, further decomposed if necessary, and even easily replaced with more effective prompts, trained models, or symbolic functions if desired. We show that the flexibility and modularity of Decomposed Prompting allows it to outperform prior work on few-shot prompting using GPT3. On symbolic reasoning tasks, we can further decompose sub-tasks that are hard for LLMs into even simpler solvable sub-tasks. When the complexity comes from the input length, we can recursively decompose the task into the same task but with smaller inputs. We also evaluate our approach on textual multi-step reasoning tasks: on long-context multi-hop QA task, we can more effectively teach the sub-tasks via our separate sub-tasks prompts; and on open-domain multi-hop QA, we can incorporate a symbolic information retrieval within our decomposition framework, leading to improved performance on both tasks. Datasets, Code and Prompts available at https://github.com/allenai/DecomP."
                },
                {
                    "title": "FOLIO: Natural Language Reasoning with First-Order Logic",
                    "abstract": "Large language models (LLMs) have achieved remarkable performance on a variety of natural language understanding tasks. However, existing benchmarks are inadequate in measuring the complex logical reasoning capabilities of a model. We present FOLIO, a human-annotated, logically complex and diverse dataset for reasoning in natural language (NL), equipped with first-order logic (FOL) annotations. FOLIO consists of 1,430 examples (unique conclusions), each paired with one of 487 sets of premises used to deductively reason for the validity of each conclusion. The logical correctness of the premises and conclusions is ensured by their FOL annotations, which are automatically verified by an FOL inference engine. In addition to the main NL reasoning task, NL-FOL pairs in FOLIO constitute a new NL-FOL translation dataset. Our experiments on FOLIO systematically evaluate the FOL reasoning ability of supervised fine-tuning on medium-sized language models. For both NL reasoning and NL-FOL translation, we benchmark multiple state-of-the-art language models. Our results show that a subset of FOLIO presents a challenge for one of the most capable {Large Language Model (LLM)} publicly available, GPT-4."
                },
                {
                    "title": "Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models",
                    "abstract": "Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit\"breakthrough\"behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting."
                },
                {
                    "title": "Least-to-Most Prompting Enables Complex Reasoning in Large Language Models",
                    "abstract": "Chain-of-thought prompting has demonstrated remarkable performance on various natural language reasoning tasks. However, it tends to perform poorly on tasks which requires solving problems harder than the exemplars shown in the prompts. To overcome this challenge of easy-to-hard generalization, we propose a novel prompting strategy, least-to-most prompting. The key idea in this strategy is to break down a complex problem into a series of simpler subproblems and then solve them in sequence. Solving each subproblem is facilitated by the answers to previously solved subproblems. Our experimental results on tasks related to symbolic manipulation, compositional generalization, and math reasoning reveal that least-to-most prompting is capable of generalizing to more difficult problems than those seen in the prompts. A notable finding is that when the GPT-3 code-davinci-002 model is used with least-to-most prompting, it can solve the compositional generalization benchmark SCAN in any split (including length split) with an accuracy of at least 99% using just 14 exemplars, compared to only 16% accuracy with chain-of-thought prompting. This is particularly noteworthy because neural-symbolic models in the literature that specialize in solving SCAN are trained on the entire training set containing over 15,000 examples. We have included prompts for all the tasks in the Appendix."
                },
                {
                    "title": "TreeMix: Compositional Constituency-based Data Augmentation for Natural Language Understanding",
                    "abstract": "Data augmentation is an effective approach to tackle over-fitting. Many previous works have proposed different data augmentations strategies for NLP, such as noise injection, word replacement, back-translation etc. Though effective, they missed one important characteristic of language\u2013compositionality, meaning of a complex expression is built from its sub-parts. Motivated by this, we propose a compositional data augmentation approach for natural language understanding called TreeMix. Specifically, TreeMix leverages constituency parsing tree to decompose sentences into constituent sub-structures and the Mixup data augmentation technique to recombine them to generate new sentences. Compared with previous approaches, TreeMix introduces greater diversity to the samples generated and encourages models to learn compositionality of NLP data. Extensive experiments on text classification and SCAN demonstrate that TreeMix outperforms current state-of-the-art data augmentation methods."
                },
                {
                    "title": "Data Contamination: From Memorization to Exploitation",
                    "abstract": "Pretrained language models are typically trained on massive web-based datasets, which are often \u201ccontaminated\u201d with downstream test sets. It is not clear to what extent models exploit the contaminated data for downstream tasks. We present a principled method to study this question. We pretrain BERT models on joint corpora of Wikipedia and labeled downstream datasets, and fine-tune them on the relevant task. Comparing performance between samples seen and unseen during pretraining enables us to define and quantify levels of memorization and exploitation.Experiments with two models and three downstream tasks show that exploitation exists in some cases, but in others the models memorize the contaminated data, but do not exploit it. We show that these two measures are affected by different factors such as the number of duplications of the contaminated data and the model size. Our results highlight the importance of analyzing massive web-scale datasets to verify that progress in NLP is obtained by better language understanding and not better data exploitation."
                },
                {
                    "title": "GraphWorld: Fake Graphs Bring Real Insights for GNNs",
                    "abstract": "Despite advances in the field of Graph Neural Networks (GNNs), only a small number (~5) of datasets are currently used to evaluate new models. This continued reliance on a handful of datasets provides minimal insight into the performance differences between models, and is especially challenging for industrial practitioners who are likely to have datasets which are very different from academic benchmarks. In the course of our work on GNN infrastructure and open-source software at Google, we have sought to develop benchmarks that are robust, tunable, scalable, and generalizable. In this work we introduce GraphWorld, a novel methodology and system for benchmarking GNN models on an arbitrarily-large population ofsynthetic graphs for any conceivable GNN task. GraphWorld allows a user to efficiently generate a world with millions of statistically diverse datasets. It is accessible, scalable, and easy to use. GraphWorld can be run on a single machine without specialized hardware, or it can be easily scaled up to run on arbitrary clusters or cloud frameworks. Using GraphWorld, a user has fine-grained control over graph generator parameters, and can benchmark arbitrary GNN models with built-in hyperparameter tuning. We present insights from GraphWorld experiments regarding the performance characteristics of thirteen GNN models and baselines over millions of benchmark datasets. We further show that GraphWorld efficiently explores regions of benchmark dataset space uncovered by standard benchmarks, revealing comparisons between models that have not been historically obtainable. Using GraphWorld, we also are able to study in-detail the relationship between graph properties and task performance metrics, which is nearly impossible with the classic collection of real-world benchmarks."
                },
                {
                    "title": "Quantifying Memorization Across Neural Language Models",
                    "abstract": "Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "Dynaboard: An Evaluation-As-A-Service Platform for Holistic Next-Generation Benchmarking",
                    "abstract": "We introduce Dynaboard, an evaluation-as-a-service framework for hosting benchmarks and conducting holistic model comparison, integrated with the Dynabench platform. Our platform evaluates NLP models directly instead of relying on self-reported metrics or predictions on a single dataset. Under this paradigm, models are submitted to be evaluated in the cloud, circumventing the issues of reproducibility, accessibility, and backwards compatibility that often hinder benchmarking in NLP. This allows users to interact with uploaded models in real time to assess their quality, and permits the collection of additional metrics such as memory use, throughput, and robustness, which -- despite their importance to practitioners -- have traditionally been absent from leaderboards. On each task, models are ranked according to the Dynascore, a novel utility-based aggregation of these statistics, which users can customize to better reflect their preferences, placing more/less weight on a particular axis of evaluation or dataset. As state-of-the-art NLP models push the limits of traditional benchmarks, Dynaboard offers a standardized solution for a more diverse and comprehensive evaluation of model quality."
                },
                {
                    "title": "Memorisation versus Generalisation in Pre-trained Language Models",
                    "abstract": "State-of-the-art pre-trained language models have been shown to memorise facts and perform well with limited amounts of training data. To gain a better understanding of how these models learn, we study their generalisation and memorisation capabilities in noisy and low-resource scenarios. We find that the training of these models is almost unaffected by label noise and that it is possible to reach near-optimal results even on extremely noisy datasets. However, our experiments also show that they mainly learn from high-frequency patterns and largely fail when tested on low-resource tasks such as few-shot learning and rare entity recognition. To mitigate such limitations, we propose an extension based on prototypical networks that improves performance in low-resource named entity recognition tasks."
                },
                {
                    "title": "Dynabench: Rethinking Benchmarking in NLP",
                    "abstract": "We introduce Dynabench, an open-source platform for dynamic dataset creation and model benchmarking. Dynabench runs in a web browser and supports human-and-model-in-the-loop dataset creation: annotators seek to create examples that a target model will misclassify, but that another person will not. In this paper, we argue that Dynabench addresses a critical need in our community: contemporary models quickly achieve outstanding performance on benchmark tasks but nonetheless fail on simple challenge examples and falter in real-world scenarios. With Dynabench, dataset creation, model development, and model assessment can directly inform each other, leading to more robust and informative benchmarks. We report on four initial NLP tasks, illustrating these concepts and highlighting the promise of the platform, and address potential objections to dynamic benchmarking as a new standard for the field."
                },
                {
                    "title": "Are NLP Models really able to Solve Simple Math Word Problems?",
                    "abstract": "The problem of designing NLP solvers for math word problems (MWP) has seen sustained research activity and steady gains in the test accuracy. Since existing solvers achieve high performance on the benchmark datasets for elementary level MWPs containing one-unknown arithmetic word problems, such problems are often considered \u201csolved\u201d with the bulk of research attention moving to more complex MWPs. In this paper, we restrict our attention to English MWPs taught in grades four and lower. We provide strong evidence that the existing MWP solvers rely on shallow heuristics to achieve high performance on the benchmark datasets. To this end, we show that MWP solvers that do not have access to the question asked in the MWP can still solve a large fraction of MWPs. Similarly, models that treat MWPs as bag-of-words can also achieve surprisingly high accuracy. Further, we introduce a challenge dataset, SVAMP, created by applying carefully chosen variations over examples sampled from existing datasets. The best accuracy achieved by state-of-the-art models is substantially lower on SVAMP, thus showing that much remains to be done even for the simplest of the MWPs."
                },
                {
                    "title": "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? \ud83e\udd9c",
                    "abstract": "The past 3 years of work in NLP have been characterized by the development and deployment of ever larger language models, especially for English. BERT, its variants, GPT-2/3, and others, most recently Switch-C, have pushed the boundaries of the possible both through architectural innovations and through sheer size. Using these pretrained models and the methodology of fine-tuning them for specific tasks, researchers have extended the state of the art on a wide array of tasks as measured by leaderboards on specific benchmarks for English. In this paper, we take a step back and ask: How big is too big? What are the possible risks associated with this technology and what paths are available for mitigating those risks? We provide recommendations including weighing the environmental and financial costs first, investing resources into curating and carefully documenting datasets rather than ingesting everything on the web, carrying out pre-development exercises evaluating how the planned approach fits into research and development goals and supports stakeholder values, and encouraging research directions beyond ever larger language models."
                },
                {
                    "title": "Understanding deep learning (still) requires rethinking generalization",
                    "abstract": "Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small gap between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family or to the regularization techniques used during training. Through extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. We interpret our experimental findings by comparison with traditional models. We supplement this republication with a new section at the end summarizing recent progresses in the field since the original version of this paper."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Beyond Accuracy: Behavioral Testing of NLP Models with CheckList",
                    "abstract": "Although measuring held-out accuracy has been the primary approach to evaluate generalization, it often overestimates the performance of NLP models, while alternative approaches for evaluating models either focus on individual tasks or on specific behaviors. Inspired by principles of behavioral testing in software engineering, we introduce CheckList, a task-agnostic methodology for testing NLP models. CheckList includes a matrix of general linguistic capabilities and test types that facilitate comprehensive test ideation, as well as a software tool to generate a large and diverse number of test cases quickly. We illustrate the utility of CheckList with tests for three tasks, identifying critical failures in both commercial and state-of-art models. In a user study, a team responsible for a commercial sentiment analysis model found new and actionable bugs in an extensively tested model. In another user study, NLP practitioners with CheckList created twice as many tests, and found almost three times as many bugs as users without it."
                },
                {
                    "title": "BERT-ATTACK: Adversarial Attack against BERT Using BERT",
                    "abstract": "Adversarial attacks for discrete data (such as text) has been proved significantly more challenging than continuous data (such as image), since it is difficult to generate adversarial samples with gradient-based methods. Currently, the successful attack methods for text usually adopt heuristic replacement strategies on character or word level, which remains challenging to find the optimal solution in the massive space of possible combination of replacements, while preserving semantic consistency and language fluency. In this paper, we propose \\textbf{BERT-Attack}, a high-quality and effective method to generate adversarial samples using pre-trained masked language models exemplified by BERT. We turn BERT against its fine-tuned models and other deep neural models for downstream tasks. Our method successfully misleads the target models to predict incorrectly, outperforming state-of-the-art attack strategies in both success rate and perturb percentage, while the generated adversarial samples are fluent and semantically preserved. Also, the cost of calculation is low, thus possible for large-scale generations."
                },
                {
                    "title": "Transformers as Soft Reasoners over Language",
                    "abstract": "Beginning with McCarthy's Advice Taker (1959), AI has pursued the goal of providing a system with explicit, general knowledge and having the system reason over that knowledge. However, expressing the knowledge in a formal (logical or probabilistic) representation has been a major obstacle to this research. This paper investigates a modern approach to this problem where the facts and rules are provided as natural language sentences, thus bypassing a formal representation. We train transformers to reason (or emulate reasoning) over these sentences using synthetically generated data. Our models, that we call RuleTakers, provide the first empirical demonstration that this kind of soft reasoning over language is learnable, can achieve high (99%) accuracy, and generalizes to test data requiring substantially deeper chaining than seen during training (95%+ scores). We also demonstrate that the models transfer well to two hand-authored rulebases, and to rulebases paraphrased into more natural language. These findings are significant as it suggests a new role for transformers, namely as limited \"soft theorem provers\" operating over explicit theories in language. This in turn suggests new possibilities for explainability, correctability, and counterfactual reasoning in question-answering."
                },
                {
                    "title": "Easy-to-Hard: Leveraging Simple Questions for Complex Question Generation",
                    "abstract": "This paper makes one of the first efforts toward automatically generating complex questions from knowledge graphs. Particularly, we study how to leverage existing simple question datasets for this task, under two separate scenarios: using either sub-questions of the target complex questions, or distantly related pseudo sub-questions when the former are unavailable. First, a competitive base model named CoG2Q is designed to map complex query qraphs to natural language questions. Afterwards, we propose two extension models, namely CoGSub2Q and CoGSub^m2Q, respectively for the above two scenarios. The former encodes and copies from a sub-question, while the latter further scores and aggregates multiple pseudo sub-questions. Experiment results show that the extension models significantly outperform not only base CoG2Q, but also its augmented variant using simple questions as additional training examples. This demonstrates the importance of instance-level connections between simple and corresponding complex questions, which may be underexploited by straightforward data augmentation of CoG2Q that builds model-level connections through learned parameters."
                },
                {
                    "title": "CLUTRR: A Diagnostic Benchmark for Inductive Reasoning from Text",
                    "abstract": "The recent success of natural language understanding (NLU) systems has been troubled by results highlighting the failure of these models to generalize in a systematic and robust way. In this work, we introduce a diagnostic benchmark suite, named CLUTRR, to clarify some key issues related to the robustness and systematicity of NLU systems. Motivated by the classic work on inductive logic programming, CLUTRR requires that an NLU system infer kinship relations between characters in short stories. Successful performance on this task requires both extracting relationships between entities, as well as inferring the logical rules governing these relationships. CLUTRR allows us to precisely measure a model\u2019s ability for systematic generalization by evaluating on held-out combinations of logical rules, and allows us to evaluate a model\u2019s robustness by adding curated noise facts. Our empirical results highlight a substantial performance gap between state-of-the-art NLU models (e.g., BERT and MAC) and a graph neural network model that works directly with symbolic inputs\u2014with the graph-based model exhibiting both stronger generalization and greater robustness."
                },
                {
                    "title": "Is BERT Really Robust? A Strong Baseline for Natural Language Attack on Text Classification and Entailment",
                    "abstract": "Machine learning algorithms are often vulnerable to adversarial examples that have imperceptible alterations from the original counterparts but can fool the state-of-the-art models. It is helpful to evaluate or even improve the robustness of these models by exposing the maliciously crafted adversarial examples. In this paper, we present TextFooler, a simple but strong baseline to generate adversarial text. By applying it to two fundamental natural language tasks, text classification and textual entailment, we successfully attacked three target models, including the powerful pre-trained BERT, and the widely used convolutional and recurrent neural networks. We demonstrate three advantages of this framework: (1) effective\u2014it outperforms previous attacks by success rate and perturbation rate, (2) utility-preserving\u2014it preserves semantic content, grammaticality, and correct types classified by humans, and (3) efficient\u2014it generates adversarial text with computational complexity linear to the text length.1"
                },
                {
                    "title": "Good-Enough Compositional Data Augmentation",
                    "abstract": "We propose a simple data augmentation protocol aimed at providing a compositional inductive bias in conditional and unconditional sequence models. Under this protocol, synthetic training examples are constructed by taking real training examples and replacing (possibly discontinuous) fragments with other fragments that appear in at least one similar environment. The protocol is model-agnostic and useful for a variety of tasks. Applied to neural sequence-to-sequence models, it reduces error rate by as much as 87% on diagnostic tasks from the SCAN dataset and 16% on a semantic parsing task. Applied to n-gram language models, it reduces perplexity by roughly 1% on small corpora in several languages."
                },
                {
                    "title": "TextBugger: Generating Adversarial Text Against Real-world Applications",
                    "abstract": "Deep Learning-based Text Understanding (DLTU) is the backbone technique behind various applications, including question answering, machine translation, and text classification. Despite its tremendous popularity, the security vulnerabilities of DLTU are still largely unknown, which is highly concerning given its increasing use in security-sensitive applications such as sentiment analysis and toxic content detection. In this paper, we show that DLTU is inherently vulnerable to adversarial text attacks, in which maliciously crafted texts trigger target DLTU systems and services to misbehave. Specifically, we present TextBugger, a general attack framework for generating adversarial texts. In contrast to prior works, TextBugger differs in significant ways: (i) effective -- it outperforms state-of-the-art attacks in terms of attack success rate; (ii) evasive -- it preserves the utility of benign text, with 94.9\\% of the adversarial text correctly recognized by human readers; and (iii) efficient -- it generates adversarial text with computational complexity sub-linear to the text length. We empirically evaluate TextBugger on a set of real-world DLTU systems and services used for sentiment analysis and toxic content detection, demonstrating its effectiveness, evasiveness, and efficiency. For instance, TextBugger achieves 100\\% success rate on the IMDB dataset based on Amazon AWS Comprehend within 4.61 seconds and preserves 97\\% semantic similarity. We further discuss possible defense mechanisms to mitigate such attack and the adversary's potential countermeasures, which leads to promising directions for further research."
                },
                {
                    "title": "Black-Box Generation of Adversarial Text Sequences to Evade Deep Learning Classifiers",
                    "abstract": "Although various techniques have been proposed to generate adversarial samples for white-box attacks on text, little attention has been paid to a black-box attack, which is a more realistic scenario. In this paper, we present a novel algorithm, DeepWordBug, to effectively generate small text perturbations in a black-box setting that forces a deep-learning classifier to misclassify a text input. We develop novel scoring strategies to find the most important words to modify such that the deep classifier makes a wrong prediction. Simple character-level transformations are applied to the highest-ranked words in order to minimize the edit distance of the perturbation. We evaluated DeepWordBug on two real-world text datasets: Enron spam emails and IMDB movie reviews. Our experimental results indicate that DeepWordBug can reduce the classification accuracy from 99% to 40% on Enron and from 87% to 26% on IMDB. Our results strongly demonstrate that the generated adversarial sequences from a deep-learning model can similarly evade other deep models."
                },
                {
                    "title": "Dynamic Evaluation of Neural Sequence Models",
                    "abstract": "We present methodology for using dynamic evaluation to improve neural sequence models. Models are adapted to recent history via a gradient descent based mechanism, causing them to assign higher probabilities to re-occurring sequential patterns. Dynamic evaluation outperforms existing adaptation approaches in our comparisons. Dynamic evaluation improves the state-of-the-art word-level perplexities on the Penn Treebank and WikiText-2 datasets to 51.1 and 44.3 respectively, and the state-of-the-art character-level cross-entropies on the text8 and Hutter Prize datasets to 1.19 bits/char and 1.08 bits/char respectively."
                },
                {
                    "title": "DeepTest: Automated Testing of Deep-Neural-Network-Driven Autonomous Cars",
                    "abstract": "Recent advances in Deep Neural Networks (DNNs) have led to the development of DNN-driven autonomous cars that, using sensors like camera, LiDAR, etc., can drive without any human intervention. Most major manufacturers including Tesla, GM, Ford, BMW, and Waymo/Google are working on building and testing different types of autonomous vehicles. The lawmakers of several US states including California, Texas, and New York have passed new legislation to fast-track the process of testing and deployment of autonomous vehicles on their roads. However, despite their spectacular progress, DNNs, just like traditional software, often demonstrate incorrect or unexpected corner-case behaviors that can lead to potentially fatal collisions. Several such real-world accidents involving autonomous cars have already happened including one which resulted in a fatality. Most existing testing techniques for DNN-driven vehicles are heavily dependent on the manual collection of test data under different driving conditions which become prohibitively expensive as the number of test conditions increases. In this paper, we design, implement, and evaluate DeepTest, a systematic testing tool for automatically detecting erroneous behaviors of DNN-driven vehicles that can potentially lead to fatal crashes. First, our tool is designed to automatically generated test cases leveraging real-world changes in driving conditions like rain, fog, lighting conditions, etc. DeepTest systematically explore different parts of the DNN logic by generating test inputs that maximize the numbers of activated neurons. DeepTest found thousands of erroneous behaviors under different realistic driving conditions (e.g., blurring, rain, fog, etc.) many of which lead to potentially fatal crashes in three top performing DNNs in the Udacity self-driving car challenge."
                },
                {
                    "title": "A Closer Look at Memorization in Deep Networks",
                    "abstract": "We examine the role of memorization in deep learning, drawing connections to capacity, generalization, and adversarial robustness. While deep networks are capable of memorizing noise data, our results suggest that they tend to prioritize learning simple patterns first. In our experiments, we expose qualitative differences in gradient-based optimization of deep neural networks (DNNs) on noise vs. real data. We also demonstrate that for appropriately tuned explicit regularization (e.g., dropout) we can degrade DNN training performance on noise datasets without compromising generalization on real data. Our analysis suggests that the notions of effective capacity which are dataset independent are unlikely to explain the generalization performance of deep networks when trained with gradient based methods because training data itself plays an important role in determining the degree of memorization."
                },
                {
                    "title": "Solving General Arithmetic Word Problems",
                    "abstract": "This paper presents a novel approach to automatically solving arithmetic word problems. This is the first algorithmic approach that can handle arithmetic problems with multiple steps and operations, without depending on additional annotations or predefined templates. We develop a theory for expression trees that can be used to represent and evaluate the target arithmetic expressions; we use it to uniquely decompose the target arithmetic problem to multiple classification problems; we then compose an expression tree, combining these with world knowledge through a constrained inference framework. Our classifiers gain from the use of quantity schemas that supports better extraction of features. Experimental results show that our method outperforms existing systems, achieving state of the art performance on benchmark datasets of arithmetic word problems."
                },
                {
                    "title": "Grammar as a Foreign Language",
                    "abstract": "Syntactic constituency parsing is a fundamental problem in natural language processing and has been the subject of intensive research and engineering for decades. As a result, the most accurate parsers are domain specific, complex, and inefficient. In this paper we show that the domain agnostic attention-enhanced sequence-to-sequence model achieves state-of-the-art results on the most widely used syntactic constituency parsing dataset, when trained on a large synthetic corpus that was annotated using existing parsers. It also matches the performance of standard parsers when trained only on a small human-annotated dataset, which shows that this model is highly data-efficient, in contrast to sequence-to-sequence models without the attention mechanism. Our parser is also fast, processing over a hundred sentences per second with an unoptimized CPU implementation."
                },
                {
                    "title": "Accurate Unlexicalized Parsing",
                    "abstract": "We demonstrate that an unlexicalized PCFG can parse much more accurately than previously shown, by making use of simple, linguistically motivated state splits, which break down false independence assumptions latent in a vanilla treebank grammar. Indeed, its performance of 86.36% (LP/LR F1) is better than that of early lexicalized PCFG models, and surprisingly close to the current state-of-the-art. This result has potential uses beyond establishing a strong lower bound on the maximum possible accuracy of unlexicalized models: an unlexicalized PCFG is much more compact, easier to replicate, and easier to interpret than more complex lexical models, and the parsing algorithms are simpler, more widely understood, of lower asymptotic complexity, and easier to optimize."
                },
                {
                    "title": "Graphs: Theory and Algorithms",
                    "abstract": "Basic Concepts. Trees, Cutsets, and Circuits. Eulerian and Hamiltonian Graphs. Graphs and Vector Spaces. Directed Graphs. Matrices of a Graph. Planarity and Duality. Connectivity and Matching. Covering and Coloring. Matroids. Graph Algorithms. Flows in Networks. Indexes."
                },
                {
                    "title": "An Open Source Data Contamination Report for Llama Series Models",
                    "abstract": "Data contamination in language model evaluation is increasingly prevalent as the popularity of large language models. It allows models to \"cheat\" via memorisation instead of displaying true capabilities. Therefore, contamination analysis has became an crucial part of reliable model evaluation to validate results. However, existing contamination analysis is usually conducted internally by LLM developers and often lacks transparency and completeness. This paper present an open source data contamination reports for the Llama series models. We analyse six popular multi-choice QA benchmarks and quantify their overlapping with the training set of Llama. Various levels of contamination ranging from 1% to 8.7% are found across benchmarks. Our comparison also reveals that Llama models can gain over 5% higher accuracy on contaminated subsets versus clean subsets. Data and code are available at: https://github.com/liyucheng09/ Contamination_Detector ."
                },
                {
                    "title": "Adaptive Testing and Debugging of NLP Models",
                    "abstract": "Current approaches to testing and debugging NLP models rely on highly variable human creativity and extensive labor, or only work for a very restrictive class of bugs. We present AdaTest, a process which uses large scale language models (LMs) in partnership with human feedback to automatically write unit tests highlighting bugs in a target model. Such bugs are then addressed through an iterative text-fix-retest loop, inspired by traditional software development. In experiments with expert and non-expert users and commercial / research models for 8 different tasks, AdaTest makes users 5-10x more effective at finding bugs than current approaches, and helps users effectively fix bugs without adding new bugs."
                },
                {
                    "title": "Diagnosing the First-Order Logical Reasoning Ability Through LogicNLI",
                    "abstract": "Recently, language models (LMs) have achieved significant performance on many NLU tasks, which has spurred widespread interest for their possible applications in the scientific and social area. However, LMs have faced much criticism of whether they are truly capable of reasoning in NLU. In this work, we propose a diagnostic method for first-order logic (FOL) reasoning with a new proposed benchmark, LogicNLI. LogicNLI is an NLI-style dataset that effectively disentangles the target FOL reasoning from commonsense inference and can be used to diagnose LMs from four perspectives: accuracy, robustness, generalization, and interpretability. Experiments on BERT, RoBERTa, and XLNet, have uncovered the weaknesses of these LMs on FOL reasoning, which motivates future exploration to enhance the reasoning ability."
                },
                {
                    "title": "Compilers Principles Techniques And Tools",
                    "abstract": "Thank you for reading compilers principles techniques and tools. As you may know, people have search numerous times for their chosen readings like this compilers principles techniques and tools, but end up in malicious downloads. Rather than enjoying a good book with a cup of coffee in the afternoon, instead they cope with some malicious virus inside their laptop. compilers principles techniques and tools is available in our digital library an online access to it is set as public so you can download it instantly. Our books collection saves in multiple locations, allowing you to get the most less latency time to download any of our books like this one. Kindly say, the compilers principles techniques and tools is universally compatible with any devices to read."
                }
            ],
            "categories": [
                "cs.AI",
                "cs.CL",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we create a dynamic evaluation protocol for Large Language Models (LLMs) that adapts to their evolving capabilities and addresses the issues of data contamination and static complexity in existing benchmarks?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it will provide a more accurate and nuanced understanding of LLMs' capabilities, distinguishing between genuine understanding and memorization. A dynamic evaluation protocol like DyVal could lead to more effective benchmarking, fostering advancements in LLM research and development. This could ultimately contribute to the progression towards artificial general intelligence by ensuring that models are rigorously tested against their true potential, leading to practical applications in various fields such as natural language processing, education, and automated reasoning.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the technical complexity of dynamically generating evaluation samples that accurately reflect the capabilities of LLMs, as well as the need to modulate sample complexity and validity in real-time. Naive approaches may fail because they do not account for the rapid advancements in LLMs or the nuances of their training data, leading to evaluations that are either too simplistic or irrelevant. Additionally, ensuring that the generated samples are diverse and representative of real-world scenarios poses a significant obstacle.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by static datasets that do not evolve with LLM capabilities, leading to evaluations that may not accurately reflect model performance. Existing dynamic benchmarks rely on crowd-sourcing for data collection, which is often expensive and labor-intensive, creating barriers to their widespread adoption. DyVal improves upon prior work by introducing a framework that generates evaluation samples on-the-fly, eliminating the need for fixed datasets and allowing for real-time adaptation to the models being evaluated.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology for DyVal includes three key components: 1) a generation algorithm (\ud835\udca2) that creates diverse test samples, 2) a constraint (\ud835\udc9e) that modulates sample complexity and validity, and 3) a description function (\u2131) that translates generated samples into natural language. The expected outcomes include a more robust and flexible evaluation framework that can dynamically adapt to the capabilities of LLMs, leading to more accurate assessments of their performance and a better understanding"
            }
        },
        "author_data": {
            "8313e8b4-b645-4fbd-8ace-d7aa59df5e22": {
                "pk": "8313e8b4-b645-4fbd-8ace-d7aa59df5e22",
                "name": "Kaijie Zhu",
                "collaborators": [
                    "Jindong Wang",
                    "Hao Chen",
                    "Qinlin Zhao",
                    "Xing Xie",
                    "Yiqiao Jin",
                    "Yixuan Zhang",
                    "Xingxu Xie",
                    "Wenyue Hua",
                    "Lingyao Li",
                    "Lizhou Fan",
                    "Mingyu Jin",
                    "Yongfeng Zhang",
                    "Linyi Yang",
                    "Yidong Wang",
                    "Weirong Ye",
                    "Yue Zhang",
                    "Ruochen Xu",
                    "Shuhang Lin",
                    "Haochen Xue",
                    "Zelong Li",
                    "Yiyang Wang",
                    "Yijia Xiao",
                    "Xiang Li",
                    "Haoyang Ling",
                    "Jinkui Chi",
                    "Xin Ma",
                    "Yu-Chu Chang",
                    "Xu Wang",
                    "Yuanyi Wu",
                    "Xiaoyuan Yi",
                    "Cunxiang Wang",
                    "Yi Chang",
                    "Philip S. Yu",
                    "Qian Yang",
                    "Jiaheng Zhou",
                    "N. Gong",
                    "Cheng Li",
                    "Xinyi Wang",
                    "Wenxin Hou",
                    "Jianxun Lian",
                    "Fang Luo",
                    "Qiang Yang",
                    "Xixu Hu",
                    "G. Yang"
                ],
                "domain": [
                    "Large Language Models",
                    "Evaluation",
                    "Adversarial Robustness",
                    "Multi-Agent Systems"
                ],
                "publications": [
                    {
                        "title": "DyVal 2: Dynamic Evaluation of Large Language Models by Meta Probing Agents",
                        "abstract": "Evaluation of large language models (LLMs) has raised great concerns in the community due to the issue of data contamination. Existing work designed evaluation protocols using well-defined algorithms for specific tasks, which cannot be easily extended to diverse scenarios. Moreover, current evaluation benchmarks can only provide the overall benchmark results and cannot support a fine-grained and multifaceted analysis of LLMs' abilities. In this paper, we propose meta probing agents (MPA), a general dynamic evaluation protocol inspired by psychometrics to evaluate LLMs. MPA is the key component of DyVal 2, which naturally extends the previous DyVal~\\citep{zhu2023dyval}. MPA designs the probing and judging agents to automatically transform an original evaluation problem into a new one following psychometric theory on three basic cognitive abilities: language understanding, problem solving, and domain knowledge. These basic abilities are also dynamically configurable, allowing multifaceted analysis. We conducted extensive evaluations using MPA and found that most LLMs achieve poorer performance, indicating room for improvement. Our multifaceted analysis demonstrated the strong correlation between the basic abilities and an implicit Matthew effect on model size, i.e., larger models possess stronger correlations of the abilities. MPA can also be used as a data augmentation approach to enhance LLMs. Code is available at: https://github.com/microsoft/promptbench."
                    },
                    {
                        "title": "Disentangling Logic: The Role of Context in Large Language Model Reasoning Capabilities",
                        "abstract": "This study intends to systematically disentangle pure logic reasoning and text understanding by investigating the contrast across abstract and contextualized logical problems from a comprehensive set of domains. We explore whether LLMs demonstrate genuine reasoning capabilities across various domains when the underlying logical structure remains constant. We focus on two main questions (1) Can abstract logical problems alone accurately benchmark an LLM's reasoning ability in real-world scenarios, disentangled from contextual support in practical settings? (2) Does fine-tuning LLMs on abstract logic problem generalize to contextualized logic problems and vice versa? To investigate these questions, we focus on standard propositional logic, specifically propositional deductive and abductive logic reasoning. In particular, we construct instantiated datasets for deductive and abductive reasoning with 4 levels of difficulty, encompassing 12 distinct categories or domains based on the categorization of Wikipedia. Our experiments aim to provide insights into disentangling context in logical reasoning and the true reasoning capabilities of LLMs and their generalization potential. The code and dataset are available at: https://github.com/agiresearch/ContextHub."
                    },
                    {
                        "title": "AgentReview: Exploring Peer Review Dynamics with LLM Agents",
                        "abstract": "Peer review is fundamental to the integrity and advancement of scientific publication. Traditional methods of peer review analyses often rely on exploration and statistics of existing peer review data, which do not adequately address the multivariate nature of the process, account for the latent variables, and are further constrained by privacy concerns due to the sensitive nature of the data. We introduce AgentReview, the first large language model (LLM) based peer review simulation framework, which effectively disentangles the impacts of multiple latent factors and addresses the privacy issue. Our study reveals significant insights, including a notable 37.1% variation in paper decisions due to reviewers' biases, supported by sociological theories such as the social influence theory, altruism fatigue, and authority bias. We believe that this study could offer valuable insights to improve the design of peer review mechanisms. Our code is available at https://github.com/Ahren09/AgentReview."
                    },
                    {
                        "title": "NPHardEval4V: A Dynamic Reasoning Benchmark of Multimodal Large Language Models",
                        "abstract": "Understanding the reasoning capabilities of Multimodal Large Language Models (MLLMs) is an important area of research. In this study, we introduce a dynamic benchmark, NPHardEval4V, aimed at addressing the existing gaps in evaluating the pure reasoning abilities of MLLMs. Our benchmark aims to provide a venue to disentangle the effect of various factors such as image recognition and instruction following, from the overall performance of the models, allowing us to focus solely on evaluating their reasoning abilities. It is built by converting textual description of questions from NPHardEval to image representations. Our findings reveal significant discrepancies in reasoning abilities across different models and highlight the relatively weak performance of MLLMs compared to LLMs in terms of reasoning. We also investigate the impact of different prompting styles, including visual, text, and combined visual and text prompts, on the reasoning abilities of MLLMs, demonstrating the different impacts of multimodal inputs in model performance. Unlike traditional benchmarks, which focus primarily on static evaluations, our benchmark will be updated monthly to prevent overfitting and ensure a more authentic and fine-grained evaluation of the models. We believe that this benchmark can aid in understanding and guide the further development of reasoning abilities in MLLMs. The benchmark dataset and code are available at https://github.com/lizhouf/NPHardEval4V"
                    },
                    {
                        "title": "CompeteAI: Understanding the Competition Behaviors in Large Language Model-based Agents",
                        "abstract": "Large language models (LLMs) have been widely used as agents to complete different tasks, such as personal assistance or event planning. While most work has focused on cooperation and collaboration between agents, little work explores competition , another important mechanism that fosters the development of society and economy. In this paper, we seek to examine the competition behaviors in LLM-based agents. We first propose a general framework to study the competition between agents. Then, we implement a practical competitive environment using GPT-4 to simulate a virtual town with two types of agents, including restaurant agents and customer agents. Specifically, the restaurant agents compete with each other to attract more customers, where the competition fosters them to transform, such as cultivating new operating strategies. The results of our experiments reveal several interesting findings ranging from social learning to Matthew Effect, which aligns well with existing sociological and economic theories. We believe that competition between agents deserves further investigation to help us understand society better. The code will be released soon."
                    },
                    {
                        "title": "PromptBench: A Unified Library for Evaluation of Large Language Models",
                        "abstract": "The evaluation of large language models (LLMs) is crucial to assess their performance and mitigate potential security risks. In this paper, we introduce PromptBench, a unified library to evaluate LLMs. It consists of several key components that are easily used and extended by researchers: prompt construction, prompt engineering, dataset and model loading, adversarial prompt attack, dynamic evaluation protocols, and analysis tools. PromptBench is designed to be an open, general, and flexible codebase for research purposes that can facilitate original study in creating new benchmarks, deploying downstream applications, and designing new evaluation protocols. The code is available at: https://github.com/microsoft/promptbench and will be continuously supported."
                    },
                    {
                        "title": "A Survey on Evaluation of Large Language Models",
                        "abstract": "Large language models (LLMs) are gaining increasing popularity in both academia and industry, owing to their unprecedented performance in various applications. As LLMs continue to play a vital role in both research and daily use, their evaluation becomes increasingly critical, not only at the task level, but also at the society level for better understanding of their potential risks. Over the past years, significant efforts have been made to examine LLMs from various perspectives. This paper presents a comprehensive review of these evaluation methods for LLMs, focusing on three key dimensions: what to evaluate, where to evaluate, and how to evaluate. Firstly, we provide an overview from the perspective of evaluation tasks, encompassing general natural language processing tasks, reasoning, medical usage, ethics, education, natural and social sciences, agent applications, and other areas. Secondly, we answer the \u2018where\u2019 and \u2018how\u2019 questions by diving into the evaluation methods and benchmarks, which serve as crucial components in assessing the performance of LLMs. Then, we summarize the success and failure cases of LLMs in different tasks. Finally, we shed light on several future challenges that lie ahead in LLMs evaluation. Our aim is to offer invaluable insights to researchers in the realm of LLMs evaluation, thereby aiding the development of more proficient LLMs. Our key point is that evaluation should be treated as an essential discipline to better assist the development of LLMs. We consistently maintain the related open-source materials at: https://github.com/MLGroupJLU/LLM-eval-survey"
                    },
                    {
                        "title": "PromptRobust: Towards Evaluating the Robustness of Large Language Models on Adversarial Prompts",
                        "abstract": "The increasing reliance on Large Language Models (LLMs) across academia and industry necessitates a comprehensive understanding of their robustness to prompts. In response to this vital need, we introduce PromptRobust, a robustness benchmark designed to measure LLMs' resilience to adversarial prompts. This study uses a plethora of adversarial textual attacks targeting prompts across multiple levels: character, word, sentence, and semantic. The adversarial prompts, crafted to mimic plausible user errors like typos or synonyms, aim to evaluate how slight deviations can affect LLM outcomes while maintaining semantic integrity. These prompts are then employed in diverse tasks including sentiment analysis, natural language inference, reading comprehension, machine translation, and math problem-solving. Our study generates 4,788 adversarial prompts, meticulously evaluated over 8 tasks and 13 datasets. Our findings demonstrate that contemporary LLMs are not robust to adversarial prompts. Furthermore, we present a comprehensive analysis to understand the mystery behind prompt robustness and its transferability. We then offer insightful robustness analysis and pragmatic recommendations for prompt composition, beneficial to both researchers and everyday users."
                    },
                    {
                        "title": "The Good, The Bad, and Why: Unveiling Emotions in Generative AI",
                        "abstract": "Emotion significantly impacts our daily behaviors and interactions. While recent generative AI models, such as large language models, have shown impressive performance in various tasks, it remains unclear whether they truly comprehend emotions. This paper aims to address this gap by incorporating psychological theories to gain a holistic understanding of emotions in generative AI models. Specifically, we propose three approaches: 1) EmotionPrompt to enhance AI model performance, 2) EmotionAttack to impair AI model performance, and 3) EmotionDecode to explain the effects of emotional stimuli, both benign and malignant. Through extensive experiments involving language and multi-modal models on semantic understanding, logical reasoning, and generation tasks, we demonstrate that both textual and visual EmotionPrompt can boost the performance of AI models while EmotionAttack can hinder it. Additionally, EmotionDecode reveals that AI models can comprehend emotional stimuli akin to the mechanism of dopamine in the human brain. Our work heralds a novel avenue for exploring psychology to enhance our understanding of generative AI models."
                    },
                    {
                        "title": "Improving Generalization of Adversarial Training via Robust Critical Fine-Tuning",
                        "abstract": "Deep neural networks are susceptible to adversarial examples, posing a significant security risk in critical applications. Adversarial Training (AT) is a well-established technique to enhance adversarial robustness, but it often comes at the cost of decreased generalization ability. This paper proposes Robustness Critical Fine-Tuning (RiFT), a novel approach to enhance generalization without compromising adversarial robustness. The core idea of RiFT is to exploit the redundant capacity for robustness by fine-tuning the adversarially trained model on its non-robust-critical module. To do so, we introduce module robust criticality (MRC), a measure that evaluates the significance of a given module to model robustness under worst-case weight perturbations. Using this measure, we identify the module with the lowest MRC value as the non-robust-critical module and fine-tune its weights to obtain fine-tuned weights. Subsequently, we linearly interpolate between the adversarially trained weights and fine-tuned weights to derive the optimal fine-tuned model weights. We demonstrate the efficacy of RiFT on ResNet18, ResNet34, and WideResNet34-10 models trained on CIFAR10, CIFAR100, and Tiny-ImageNet datasets. Our experiments show that RiFT can significantly improve both generalization and out-of-distribution robustness by around 1.5% while maintaining or even slightly enhancing adversarial robustness. Code is available at https://github.com/Immortalise/RiFT."
                    },
                    {
                        "title": "CompeteAI: Understanding the Competition Dynamics of Large Language Model-based Agents",
                        "abstract": "Large language models (LLMs) have been widely used as agents to complete different tasks, such as personal assistance or event planning. While most of the work has focused on cooperation and collaboration between agents, little work explores competition, another important mechanism that promotes the development of society and economy. In this paper, we seek to examine the competition dynamics in LLM-based agents. We first propose a general framework for studying the competition between agents. Then, we implement a practical competitive environment using GPT-4 to simulate a virtual town with two types of agents, restaurant agents and customer agents. Specifically, the restaurant agents compete with each other to attract more customers, where competition encourages them to transform, such as cultivating new operating strategies. Simulation experiments reveal several interesting findings at the micro and macro levels, which align well with existing market and sociological theories. We hope that the framework and environment can be a promising testbed to study competition that fosters understanding of society. Code is available at: https://github.com/microsoft/competeai."
                    }
                ]
            },
            "ce908991-29c3-4556-906f-ce644af7948c": {
                "pk": "ce908991-29c3-4556-906f-ce644af7948c",
                "name": "Jiaao Chen",
                "collaborators": [
                    "Diyi Yang",
                    "Zhehao Zhang",
                    "Aston Zhang",
                    "Mu Li",
                    "Alexander J. Smola",
                    "Caleb Ziems",
                    "William B. Held",
                    "Xingjian Shi",
                    "Yanchen Liu",
                    "Mingyu Derek Ma",
                    "Wenna Qin",
                    "Azure Zhou",
                    "Weiyan Shi",
                    "Wei Wang",
                    "Xiaoman Pan",
                    "Dian Yu",
                    "Kaiqiang Song",
                    "Xiaoyang Wang",
                    "Dong Yu",
                    "Jianshu Chen",
                    "Chengwei Qin",
                    "Zhuosheng Zhang",
                    "Michihiro Yasunaga",
                    "Omar Shaikh",
                    "Alexander K. Taylor",
                    "Nuan Wen",
                    "Po-Nien Kung",
                    "Violet Peng",
                    "W. Wang",
                    "Siru Ouyang",
                    "Jiawei Han",
                    "Kexun Zhang",
                    "Raj Sanjay Shah",
                    "Kunal Chawla",
                    "Dheeraj Eidnani",
                    "Agam Shah",
                    "Wendi Du",
                    "S. Chava",
                    "Natraj Raman",
                    "Charese Smiley",
                    "Mohan Dodda",
                    "Camille Harris",
                    "J. Anderson",
                    "Aaron Reich",
                    "Aastha Agrawal",
                    "Yanzhe Zhang",
                    "Derek Tam",
                    "Colin Raffel",
                    "Mohit Bansal"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Data Augmentation",
                    "Evaluation Methods"
                ],
                "publications": [
                    {
                        "title": "DARG: Dynamic Evaluation of Large Language Models via Adaptive Reasoning Graph",
                        "abstract": "The current paradigm of evaluating Large Language Models (LLMs) through static benchmarks comes with significant limitations, such as vulnerability to data contamination and a lack of adaptability to the evolving capabilities of LLMs. Therefore, evaluation methods that can adapt and generate evaluation data with controlled complexity are urgently needed. In this work, we introduce Dynamic Evaluation of LLMs via Adaptive Reasoning Graph Evolvement (DARG) to dynamically extend current benchmarks with controlled complexity and diversity. Specifically, we first extract the reasoning graphs of data points in current benchmarks and then perturb the reasoning graphs to generate novel testing data. Such newly generated test samples can have different levels of complexity while maintaining linguistic diversity similar to the original benchmarks. We further use a code-augmented LLM to ensure the label correctness of newly generated data. We apply our DARG framework to diverse reasoning tasks in four domains with 15 state-of-the-art LLMs. Experimental results show that almost all LLMs experience a performance decrease with increased complexity and certain LLMs exhibit significant drops. Additionally, we find that LLMs exhibit more biases when being evaluated via the data generated by DARG with higher complexity levels. These observations provide useful insights into how to dynamically and adaptively evaluate LLMs. The code is available at https://github.com/SALT-NLP/DARG."
                    },
                    {
                        "title": "Unlearn What You Want to Forget: Efficient Unlearning for LLMs",
                        "abstract": "Large language models (LLMs) have achieved significant progress from pre-training on and memorizing a wide range of textual data, however, this process might suffer from privacy issues and violations of data protection regulations. As a result, the ability to easily remove data related to individual users from such models while not deteriorating their predictive quality after the removal becomes increasingly important. To address these issues, in this work, we propose an efficient unlearning framework that could efficiently update LLMs without having to retrain the whole model after data removals, by introducing lightweight unlearning layers learned with a selective teacher-student objective into the transformers. In addition, we introduce a fusion mechanism to effectively combine different unlearning layers that learns to forget different sets of data to handle a sequence of forgetting operations. Experiments on classification and generation tasks demonstrate the effectiveness of our proposed methods compared to the state-of-the-art baselines."
                    },
                    {
                        "title": "Parameter-Efficient Fine-Tuning Design Spaces",
                        "abstract": "Parameter-efficient fine-tuning aims to achieve performance comparable to fine-tuning, using fewer trainable parameters. Several strategies (e.g., Adapters, prefix tuning, BitFit, and LoRA) have been proposed. However, their designs are hand-crafted separately, and it remains unclear whether certain design patterns exist for parameter-efficient fine-tuning. Thus, we present a parameter-efficient fine-tuning design paradigm and discover design patterns that are applicable to different experimental settings. Instead of focusing on designing another individual tuning strategy, we introduce parameter-efficient fine-tuning design spaces that parameterize tuning structures and tuning strategies. Specifically, any design space is characterized by four components: layer grouping, trainable parameter allocation, tunable groups, and strategy assignment. Starting from an initial design space, we progressively refine the space based on the model quality of each design choice and make greedy selection at each stage over these four components. We discover the following design patterns: (i) group layers in a spindle pattern; (ii) allocate the number of trainable parameters to layers uniformly; (iii) tune all the groups; (iv) assign proper tuning strategies to different groups. These design patterns result in new parameter-efficient fine-tuning methods. We show experimentally that these methods consistently and significantly outperform investigated parameter-efficient fine-tuning strategies across different backbone models and different tasks in natural language processing."
                    },
                    {
                        "title": "Mitigating Biases in Hate Speech Detection from A Causal Perspective",
                        "abstract": ","
                    },
                    {
                        "title": "Decoding Susceptibility: Modeling Misbelief to Misinformation Through a Computational Approach (preprint)",
                        "abstract": "Susceptibility to misinformation describes the degree of belief in unverifiable claims, a latent aspect of individuals' mental processes that is not observable. Existing susceptibility studies heavily rely on self-reported beliefs, which can be subject to bias, expensive to collect, and challenging to scale for downstream applications. To address these limitations, in this work, we propose a computational approach to model users' latent susceptibility levels. As shown in previous research, susceptibility is influenced by various factors (e.g., demographic factors, political ideology), and directly influences people's reposting behavior on social media. To represent the underlying mental process, our susceptibility modeling incorporates these factors as inputs, guided by the supervision of people's sharing behavior. Using COVID-19 as a testbed domain, our experiments demonstrate a significant alignment between the susceptibility scores estimated by our computational modeling and human judgments, confirming the effectiveness of this latent modeling approach. Furthermore, we apply our model to annotate susceptibility scores on a large-scale dataset and analyze the relationships between susceptibility with various factors. Our analysis reveals that political leanings and psychological factors exhibit varying degrees of association with susceptibility to COVID-19 misinformation."
                    },
                    {
                        "title": "Skills-in-Context Prompting: Unlocking Compositionality in Large Language Models",
                        "abstract": "We investigate how to elicit compositional generalization capabilities in large language models (LLMs). Compositional generalization empowers LLMs to solve complex problems by combining foundational skills, a critical reasoning ability akin to human intelligence. However, even the most advanced LLMs currently struggle with this form of reasoning. We examine this problem within the framework of in-context learning and find that demonstrating both foundational skills and compositional examples grounded in these skills within the same prompt context is crucial. We refer to this prompt structure as skills-in-context (SKiC). With as few as two exemplars, this in-context learning structure enables LLMs to tackle more challenging problems requiring innovative skill combinations, achieving near-perfect systematic generalization across a broad range of tasks. Intriguingly, SKiC also unlocks the latent potential of LLMs, allowing them to more actively utilize pre-existing internal skills acquired during earlier pretraining stages to solve complex reasoning problems. The SKiC structure is robust across different skill constructions and exemplar choices and demonstrates strong transferability to new tasks. Finally, inspired by our in-context learning study, we show that fine-tuning LLMs with SKiC-style data can elicit zero-shot weak-to-strong generalization, enabling the models to solve much harder problems directly with standard prompting."
                    },
                    {
                        "title": "Controllable Conversation Generation with Conversation Structures via Diffusion Models",
                        "abstract": ","
                    },
                    {
                        "title": "Is ChatGPT a General-Purpose Natural Language Processing Task Solver?",
                        "abstract": "Spurred by advancements in scale, large language models (LLMs) have demonstrated the ability to perform a variety of natural language processing (NLP) tasks zero-shot -- i.e., without adaptation on downstream data. Recently, the debut of ChatGPT has drawn a great deal of attention from the natural language processing (NLP) community due to the fact that it can generate high-quality responses to human input and self-correct previous mistakes based on subsequent conversations. However, it is not yet known whether ChatGPT can serve as a generalist model that can perform many NLP tasks zero-shot. In this work, we empirically analyze the zero-shot learning ability of ChatGPT by evaluating it on 20 popular NLP datasets covering 7 representative task categories. With extensive empirical studies, we demonstrate both the effectiveness and limitations of the current version of ChatGPT. We find that ChatGPT performs well on many tasks favoring reasoning capabilities (e.g., arithmetic reasoning) while it still faces challenges when solving specific tasks such as sequence tagging. We additionally provide in-depth analysis through qualitative case studies."
                    },
                    {
                        "title": "Can Large Language Models Transform Computational Social Science?",
                        "abstract": "Large language models (LLMs) are capable of successfully performing many language processing tasks zero-shot (without training data). If zero-shot LLMs can also reliably classify and explain social phenomena like persuasiveness and political ideology, then LLMs could augment the computational social science (CSS) pipeline in important ways. This work provides a road map for using LLMs as CSS tools. Towards this end, we contribute a set of prompting best practices and an extensive evaluation pipeline to measure the zero-shot performance of 13 language models on 25 representative English CSS benchmarks. On taxonomic labeling tasks (classification), LLMs fail to outperform the best fine-tuned models but still achieve fair levels of agreement with humans. On free-form coding tasks (generation), LLMs produce explanations that often exceed the quality of crowdworkers\u2019 gold references. We conclude that the performance of today\u2019s LLMs can augment the CSS research pipeline in two ways: (1) serving as zero-shot data annotators on human annotation teams, and (2) bootstrapping challenging creative generation tasks (e.g., explaining the underlying attributes of a text). In summary, LLMs are posed to meaningfully participate in social science analysis in partnership with humans."
                    },
                    {
                        "title": "Where Does Your News Come From? Predicting Information Pathways in Social Media",
                        "abstract": "As social networks become further entrenched in modern society, it becomes increasingly important to understand and predict how information (e.g., news coverage of a given event) is propagated across social media (i.e., information pathway), which helps the understandings of the impact of real-world information. Thus, in this paper, we propose a novel task, Information Pathway Prediction (IPP), which depicts the propagation paths of a given passage as a community tree (rooted at the information source) on constructed community interaction graphs where we first aggregate individual users into communities formed around news sources and influential users, and then elucidate the patterns of information dissemination across media based on such community nodes. We argue that this is an important and useful task because, on one hand, community-level interactions offer more stability than those at the user level; on the other hand, individual users are often influenced by their community, and modeling community-level information propagation will help the traditional link-prediction problem. To tackle the IPP task, we introduce Lightning, a novel content-aware link prediction GNN model and demonstrate using a large Twitter dataset consisting of all COVID related tweets that Lightning outperforms state-of-the-art link prediction baselines by a significant margin."
                    },
                    {
                        "title": "Compositional Data Augmentation for Abstractive Conversation Summarization",
                        "abstract": "Recent abstractive conversation summarization systems generally rely on large-scale datasets with annotated summaries. However, collecting and annotating these conversations can be a time-consuming and labor-intensive task. To address this issue, in this work, we present a sub-structure level compositional data augmentation method, Compo, for generating diverse and high-quality pairs of conversations and summaries. Specifically, Compo first extracts conversation structures like topic splits and action triples as basic units. Then we organize these semantically meaningful conversation snippets compositionally to create new training instances.Additionally, we explore noise-tolerant settings in both self-training and joint-training paradigms to make the most of these augmented samples. Our experiments on benchmark datasets, SAMSum and DialogSum, show that Compo substantially outperforms prior baseline methods by achieving a nearly 10% increase of ROUGE scores with limited data. Code is available at https://github.com/ozyyshr/Compo."
                    },
                    {
                        "title": "A Cheaper and Better Diffusion Language Model with Soft-Masked Noise",
                        "abstract": "Diffusion models that are based on iterative denoising have been recently proposed and leveraged in various generation tasks like image generation. Whereas, as a way inherently built for continuous data, existing diffusion models still have some limitations in modeling discrete data, e.g., languages. For example, the generally used Gaussian noise can not handle the discrete corruption well, and the objectives in continuous spaces fail to be stable for textual data in the diffusion process especially when the dimension is high. To alleviate these issues, we introduce a novel diffusion model for language modeling, Masked-Diffuse LM, with lower training cost and better performances, inspired by linguistic features in languages. Specifically, we design a linguistic-informed forward process which adds corruptions to the text through strategically soft-masking to better noise the textual data. Also, we directly predict the categorical distribution with cross-entropy loss function in every diffusion step to connect the continuous space and discrete space in a more efficient and straightforward way. Through experiments on 5 controlled generation tasks, we demonstrate that our Masked-Diffuse LM can achieve better generation quality than the state-of-the-art diffusion models with better efficiency."
                    },
                    {
                        "title": "Mixture of Virtual Prompts: Efficient Fine-tuning with Virtual Prompts for Natural Language Understanding",
                        "abstract": "Fine-tuning large-scale pre-trained language models tasks has become the de facto standard for downstream NLP tasks. However, fine-tuning all parameters of these large models brings expensive computation and storage cost. In this work, inspired by model prompting, we propose Mixture of Virtual Prompts adaptation (MVP adaptation). We first learn virtual prompts downstream tasks to inject task or language specific knowledge, while the pre-trained models are frozen to provide general knowledge. We extend virtual prompting to deal with varying complexity inputs, learning a set of virtual prompts with distinct capacities to learn different levels of knowledge, and dynamically aggregating them to adapt for each individual sequence. Experiments conducted on 5 natural language understanding tasks demonstrate that our methods could achieve comparable performances relative to the full model. Importantly, we show the superiority of MVP in terms of linguistic variation settings which allows efficient adaption to different dialects. We outperform strong efficient finetuning base-lines using only 0.03% of its parameters."
                    },
                    {
                        "title": "Focus on the Action: Learning to Highlight and Summarize Jointly for Email To-Do Items Summarization",
                        "abstract": "Automatic email to-do item generation is the task of generating to-do items from a given email to help people overview emails and schedule daily work. Different from prior research on email summarization, to-do item generation focuses on generating action mentions to provide more structured summaries of email text.Prior work either requires large amount of annotation for key sentences with potential actions or fails to pay attention to nuanced actions from these unstructured emails, and thus often lead to unfaithful summaries. To fill these gaps, we propose a simple and effective learning to highlight and summarize framework (LHS) to learn to identify the most salient text and actions, and incorporate these structured representations to generate more faithful to-do items. Experiments show that our LHS model outperforms the baselines and achieves the state-of-the-art performance in terms of both quantitative evaluation and human judgement. We also discussed specific challenges that current models faced with email to-do summarization."
                    },
                    {
                        "title": "When FLUE Meets FLANG: Benchmarks and Large Pretrained Language Model for Financial Domain",
                        "abstract": "Pre-trained language models have shown impressive performance on a variety of tasks and domains. Previous research on financial language models usually employs a generic training scheme to train standard model architectures, without completely leveraging the richness of the financial data. We propose a novel domain specific Financial LANGuage model (FLANG) which uses financial keywords and phrases for better masking, together with span boundary objective and in-filing objective. Additionally, the evaluation benchmarks in the field have been limited. To this end, we contribute the Financial Language Understanding Evaluation (FLUE), an open-source comprehensive suite of benchmarks for the financial domain. These include new benchmarks across 5 NLP tasks in financial domain as well as common benchmarks used in the previous research. Experiments on these benchmarks suggest that our model outperforms those in prior literature on a variety of NLP tasks. Our models, code and benchmark data will be made publicly available on Github and Huggingface."
                    },
                    {
                        "title": "Human-in-the-loop Abstractive Dialogue Summarization",
                        "abstract": "Abstractive dialogue summarization has received increasing attention recently. Despite the fact that most of the current dialogue summarization systems are trained to maximize the likelihood of human-written summaries and have achieved significant results, there is still a huge gap in generating high-quality summaries as determined by humans, such as coherence and faithfulness, partly due to the misalignment in maximizing a single human-written summary. To this end, we propose to incorporate different levels of human feedback into the training process. This will enable us to guide the models to capture the behaviors humans care about for summaries. Specifically, we ask humans to highlight the salient information to be included in summaries to provide the local feedback , and to make overall comparisons among summaries in terms of coherence, accuracy, coverage, concise and overall quality, as the global feedback. We then combine both local and global feedback to fine-tune the dialog summarization policy with Reinforcement Learning. Experiments conducted on multiple datasets demonstrate the effectiveness and generalization of our methods over the state-of-the-art supervised baselines, especially in terms of human judgments."
                    },
                    {
                        "title": "VALUE: Understanding Dialect Disparity in NLU",
                        "abstract": "English Natural Language Understanding (NLU) systems have achieved great performances and even outperformed humans on benchmarks like GLUE and SuperGLUE. However, these benchmarks contain only textbook Standard American English (SAE). Other dialects have been largely overlooked in the NLP community. This leads to biased and inequitable NLU systems that serve only a sub-population of speakers. To understand disparities in current models and to facilitate more dialect-competent NLU systems, we introduce the VernAcular Language Understanding Evaluation (VALUE) benchmark, a challenging variant of GLUE that we created with a set of lexical and morphosyntactic transformation rules. In this initial release (V.1), we construct rules for 11 features of African American Vernacular English (AAVE), and we recruit fluent AAVE speakers to validate each feature transformation via linguistic acceptability judgments in a participatory design manner. Experiments show that these new dialectal features can lead to a drop in model performance."
                    },
                    {
                        "title": "Leveraging Expert Guided Adversarial Augmentation For Improving Generalization in Named Entity Recognition",
                        "abstract": "Named Entity Recognition (NER) systems often demonstrate great performance on in-distribution data, but perform poorly on examples drawn from a shifted distribution. One way to evaluate the generalization ability of NER models is to use adversarial examples, on which the specific variations associated with named entities are rarely considered. To this end, we propose leveraging expert-guided heuristics to change the entity tokens and their surrounding contexts thereby altering their entity types as adversarial attacks. Using expert-guided heuristics, we augmented the CoNLL 2003 test set and manually annotated it to construct a high-quality challenging set. We found that state-of-the-art NER systems trained on CoNLL 2003 training data drop performance dramatically on our challenging set. By training on adversarial augmented training examples and using mixup for regularization, we were able to significantly improve the performance on the challenging set as well as improve out-of-domain generalization which we evaluated by using OntoNotes data. We have publicly released our dataset and code at https://github.com/GT-SALT/Guided-Adversarial-Augmentation."
                    },
                    {
                        "title": "An Empirical Survey of Data Augmentation for Limited Data Learning in NLP",
                        "abstract": "NLP has achieved great progress in the past decade through the use of neural models and large labeled datasets. The dependence on abundant data prevents NLP models from being applied to low-resource settings or novel tasks where significant time, money, or expertise is required to label massive amounts of textual data. Recently, data augmentation methods have been explored as a means of improving data efficiency in NLP. To date, there has been no systematic empirical overview of data augmentation for NLP in the limited labeled data setting, making it difficult to understand which methods work in which settings. In this paper, we provide an empirical survey of recent progress on data augmentation for NLP in the limited labeled data setting, summarizing the landscape of methods (including token-level augmentations, sentence-level augmentations, adversarial augmentations, and hidden-space augmentations) and carrying out experiments on 11 datasets covering topics/news classification, inference tasks, paraphrasing tasks, and single-sentence tasks. Based on the results, we draw several conclusions to help practitioners choose appropriate augmentations in different settings and discuss the current challenges and future directions for limited data learning in NLP."
                    }
                ]
            },
            "441f3e12-71ab-40e9-9f9a-3943b85e0499": {
                "pk": "441f3e12-71ab-40e9-9f9a-3943b85e0499",
                "name": "Jindong Wang",
                "collaborators": [
                    "Xing Xie",
                    "Hao Chen",
                    "Kaijie Zhu",
                    "Yixuan Zhang",
                    "Cheng Li",
                    "Yidong Wang",
                    "Bhiksha Raj",
                    "Xixu Hu",
                    "Qiang Yang",
                    "Xingxu Xie",
                    "Xiang Li",
                    "Masashi Sugiyama",
                    "Damien Teney",
                    "Qinlin Zhao",
                    "Wenxin Hou",
                    "Jianxun Lian",
                    "Mengzhou Chen",
                    "Sunayana Sitaram",
                    "Lei Feng",
                    "Rita Singh",
                    "Runkai Zheng",
                    "Cheuk Hang Leung",
                    "Qi Wu",
                    "Lichao Sun",
                    "Yue Huang",
                    "Haoran Wang",
                    "Siyuan Wu",
                    "Qihui Zhang",
                    "Chujie Gao",
                    "Yixin Huang",
                    "Wenhan Lyu",
                    "Xiner Li",
                    "Zheng Liu",
                    "Yixin Liu",
                    "Yijue Wang",
                    "Zhikun Zhang",
                    "B. Kailkhura",
                    "Caiming Xiong",
                    "Chaowei Xiao",
                    "Chun-Yan Li",
                    "Eric P. Xing",
                    "Furong Huang",
                    "Haodong Liu",
                    "Heng Ji",
                    "Hongyi Wang",
                    "Huan Zhang",
                    "Huaxiu Yao",
                    "M. Kellis",
                    "M. Zitnik",
                    "Meng Jiang",
                    "Mohit Bansal",
                    "James Zou",
                    "Jian Pei",
                    "Jian Liu",
                    "Jianfeng Gao",
                    "Jiawei Han",
                    "Jieyu Zhao",
                    "Jiliang Tang",
                    "John Mitchell",
                    "Kai Shu",
                    "Kaidi Xu",
                    "Kai-Wei Chang",
                    "Lifang He",
                    "Lifu Huang",
                    "M. Backes",
                    "Neil Zhenqiang Gong",
                    "Philip S. Yu",
                    "Pin-Yu Chen",
                    "Quanquan Gu",
                    "Ran Xu",
                    "Rex Ying",
                    "Shuiwang Ji",
                    "S. Jana",
                    "Tian-Xiang Chen",
                    "Tianming Liu",
                    "Tianying Zhou",
                    "William Wang",
                    "Xiang-Yu Zhang",
                    "Xiao Wang",
                    "Xingyao Xie",
                    "Xun Chen",
                    "Xuyu Wang",
                    "Yan Liu",
                    "Yanfang Ye",
                    "Yinzhi Cao",
                    "Yue Zhao",
                    "Cunxiang Wang",
                    "Xiaoze Liu",
                    "Yuanhao Yue",
                    "Xiangru Tang",
                    "Tianhang Zhang",
                    "Cheng Jiayang",
                    "Yunzhi Yao",
                    "Wenyang Gao",
                    "Xuming Hu",
                    "Zehan Qi",
                    "Linyi Yang",
                    "Zheng Zhang",
                    "Yue Zhang",
                    "Yachuan Liu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Adversarial Robustness",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "CultureLLM: Incorporating Cultural Differences into Large Language Models",
                        "abstract": "Large language models (LLMs) are reported to be partial to certain cultures owing to the training data dominance from the English corpora. Since multilingual cultural data are often expensive to collect, existing efforts handle this by prompt engineering or culture-specific pre-training. However, they might overlook the knowledge deficiency of low-resource culture and require extensive computing resources. In this paper, we propose CultureLLM, a cost-effective solution to incorporate cultural differences into LLMs. CultureLLM adopts World Value Survey (WVS) as seed data and generates semantically equivalent training data via the proposed semantic data augmentation. Using only 50 seed samples from WVS with augmented data, we fine-tune culture-specific LLMs and one unified model (CultureLLM-One) for 9 cultures covering rich and low-resource languages. Extensive experiments on 60 culture-related datasets demonstrate that CultureLLM significantly outperforms various counterparts such as GPT-3.5 (by 8.1%) and Gemini Pro (by 9.5%) with comparable performance to GPT-4 or even better. Our human study shows that the generated samples are semantically equivalent to the original samples, providing an effective solution for LLMs augmentation. Code is released at https://github.com/Scarelette/CultureLLM."
                    },
                    {
                        "title": "A General Framework for Learning from Weak Supervision",
                        "abstract": "Weakly supervised learning generally faces challenges in applicability to various scenarios with diverse weak supervision and in scalability due to the complexity of existing algorithms, thereby hindering the practical deployment. This paper introduces a general framework for learning from weak supervision (GLWS) with a novel algorithm. Central to GLWS is an Expectation-Maximization (EM) formulation, adeptly accommodating various weak supervision sources, including instance partial labels, aggregate statistics, pairwise observations, and unlabeled data. We further present an advanced algorithm that significantly simplifies the EM computational demands using a Non-deterministic Finite Automaton (NFA) along with a forward-backward algorithm, which effectively reduces time complexity from quadratic or factorial often required in existing solutions to linear scale. The problem of learning from arbitrary weak supervision is therefore converted to the NFA modeling of them. GLWS not only enhances the scalability of machine learning models but also demonstrates superior performance and versatility across 11 weak supervision scenarios. We hope our work paves the way for further advancements and practical deployment in this field."
                    },
                    {
                        "title": "SpecFormer: Guarding Vision Transformer Robustness via Maximum Singular Value Penalization",
                        "abstract": "Vision Transformers (ViTs) are increasingly used in computer vision due to their high performance, but their vulnerability to adversarial attacks is a concern. Existing methods lack a solid theoretical basis, focusing mainly on empirical training adjustments. This study introduces SpecFormer, tailored to fortify ViTs against adversarial attacks, with theoretical underpinnings. We establish local Lipschitz bounds for the self-attention layer and propose the Maximum Singular Value Penalization (MSVP) to precisely manage these bounds By incorporating MSVP into ViTs' attention layers, we enhance the model's robustness without compromising training efficiency. SpecFormer, the resulting model, outperforms other state-of-the-art models in defending against adversarial attacks, as proven by experiments on CIFAR and ImageNet datasets. Code is released at https://github.com/microsoft/robustlearn."
                    },
                    {
                        "title": "On Catastrophic Inheritance of Large Foundation Models",
                        "abstract": "Large foundation models (LFMs) are claiming incredible performances. Yet great concerns have been raised about their mythic and uninterpreted potentials not only in machine learning, but also in various other disciplines. In this position paper, we propose to identify a neglected issue deeply rooted in LFMs: Catastrophic Inheritance, describing the weaknesses and limitations inherited from biased large-scale pre-training data to behaviors of LFMs on the downstream tasks, including samples that are corrupted, long-tailed, noisy, out-of-distributed, to name a few. Such inheritance can potentially cause catastrophes to downstream applications, such as bias, lack of generalization, deteriorated performance, security vulnerability, privacy leakage, and value misalignment. We discuss the challenges behind this issue and propose UIM, a framework to Understand the catastrophic inheritance of LFMs from both pre-training and downstream adaptation, Interpret the implications of catastrophic inheritance on downstream tasks, and how to Mitigate it. UIM aims to unite both the machine learning and social sciences communities for more responsible and promising AI development and deployment."
                    },
                    {
                        "title": "TrustLLM: Trustworthiness in Large Language Models",
                        "abstract": "Large language models (LLMs), exemplified by ChatGPT, have gained considerable attention for their excellent natural language processing capabilities. Nonetheless, these LLMs present many challenges, particularly in the realm of trustworthiness. Therefore, ensuring the trustworthiness of LLMs emerges as an important topic. This paper introduces TrustLLM, a comprehensive study of trustworthiness in LLMs, including principles for different dimensions of trustworthiness, established benchmark, evaluation, and analysis of trustworthiness for mainstream LLMs, and discussion of open challenges and future directions. Specifically, we first propose a set of principles for trustworthy LLMs that span eight different dimensions. Based on these principles, we further establish a benchmark across six dimensions including truthfulness, safety, fairness, robustness, privacy, and machine ethics. We then present a study evaluating 16 mainstream LLMs in TrustLLM, consisting of over 30 datasets. Our findings firstly show that in general trustworthiness and utility (i.e., functional effectiveness) are positively related. Secondly, our observations reveal that proprietary LLMs generally outperform most open-source counterparts in terms of trustworthiness, raising concerns about the potential risks of widely accessible open-source LLMs. However, a few open-source LLMs come very close to proprietary ones. Thirdly, it is important to note that some LLMs may be overly calibrated towards exhibiting trustworthiness, to the extent that they compromise their utility by mistakenly treating benign prompts as harmful and consequently not responding. Finally, we emphasize the importance of ensuring transparency not only in the models themselves but also in the technologies that underpin trustworthiness. Knowing the specific trustworthy technologies that have been employed is crucial for analyzing their effectiveness."
                    },
                    {
                        "title": "Survey on Factuality in Large Language Models: Knowledge, Retrieval and Domain-Specificity",
                        "abstract": "This survey addresses the crucial issue of factuality in Large Language Models (LLMs). As LLMs find applications across diverse domains, the reliability and accuracy of their outputs become vital. We define the Factuality Issue as the probability of LLMs to produce content inconsistent with established facts. We first delve into the implications of these inaccuracies, highlighting the potential consequences and challenges posed by factual errors in LLM outputs. Subsequently, we analyze the mechanisms through which LLMs store and process facts, seeking the primary causes of factual errors. Our discussion then transitions to methodologies for evaluating LLM factuality, emphasizing key metrics, benchmarks, and studies. We further explore strategies for enhancing LLM factuality, including approaches tailored for specific domains. We focus two primary LLM configurations standalone LLMs and Retrieval-Augmented LLMs that utilizes external data, we detail their unique challenges and potential enhancements. Our survey offers a structured guide for researchers aiming to fortify the factual reliability of LLMs."
                    },
                    {
                        "title": "Meta Semantic Template for Evaluation of Large Language Models",
                        "abstract": "Do large language models (LLMs) genuinely understand the semantics of the language, or just memorize the training data? The recent concern on potential data contamination of LLMs has raised awareness of the community to conduct research on LLMs evaluation. In this paper, we propose MSTemp, an approach that creates meta semantic templates to evaluate the semantic understanding ability of LLMs. The core of MSTemp is not to perform evaluation directly on existing benchmark datasets, but to generate new out-of-distribution (OOD) evaluation sets using existing datasets as seeds. Specifically, for a given sentence, MSTemp leverages another language model to generate new samples while preserving its semantics. The new samples are called semantic templates to the original sentence. Then, MSTemp generates evaluation samples via sentence parsing and random word replacement on the semantic templates. MSTemp is highly flexible, dynamic, and cost-effective. Our initial experiments show that MSTemp-generated samples can significantly reduce the performance of LLMs using existing datasets as seeds. We hope this initial work can shed light on future research of LLMs evaluation."
                    },
                    {
                        "title": "Selective Mixup Helps with Distribution Shifts, But Not (Only) because of Mixup",
                        "abstract": "Mixup is a highly successful technique to improve generalization of neural networks by augmenting the training data with combinations of random pairs. Selective mixup is a family of methods that apply mixup to specific pairs, e.g. only combining examples across classes or domains. These methods have claimed remarkable improvements on benchmarks with distribution shifts, but their mechanisms and limitations remain poorly understood. We examine an overlooked aspect of selective mixup that explains its success in a completely new light. We find that the non-random selection of pairs affects the training distribution and improve generalization by means completely unrelated to the mixing. For example in binary classification, mixup across classes implicitly resamples the data for a uniform class distribution - a classical solution to label shift. We show empirically that this implicit resampling explains much of the improvements in prior work. Theoretically, these results rely on a regression toward the mean, an accidental property that we identify in several datasets. We have found a new equivalence between two successful methods: selective mixup and resampling. We identify limits of the former, confirm the effectiveness of the latter, and find better combinations of their respective benefits."
                    },
                    {
                        "title": "A Recent Survey of Heterogeneous Transfer Learning",
                        "abstract": "The application of transfer learning, leveraging knowledge from source domains to enhance model performance in a target domain, has significantly grown, supporting diverse real-world applications. Its success often relies on shared knowledge between domains, typically required in these methodologies. Commonly, methods assume identical feature and label spaces in both domains, known as homogeneous transfer learning. However, this is often impractical as source and target domains usually differ in these spaces, making precise data matching challenging and costly. Consequently, heterogeneous transfer learning (HTL), which addresses these disparities, has become a vital strategy in various tasks. In this paper, we offer an extensive review of over 60 HTL methods, covering both data-based and model-based approaches. We describe the key assumptions and algorithms of these methods and systematically categorize them into instance-based, feature representation-based, parameter regularization, and parameter tuning techniques. Additionally, we explore applications in natural language processing, computer vision, multimodal learning, and biomedicine, aiming to deepen understanding and stimulate further research in these areas. Our paper includes recent advancements in HTL, such as the introduction of transformer-based models and multimodal learning techniques, ensuring the review captures the latest developments in the field. We identify key limitations in current HTL studies and offer systematic guidance for future research, highlighting areas needing further exploration and suggesting potential directions for advancing the field."
                    },
                    {
                        "title": "CompeteAI: Understanding the Competition Behaviors in Large Language Model-based Agents",
                        "abstract": "Large language models (LLMs) have been widely used as agents to complete different tasks, such as personal assistance or event planning. While most work has focused on cooperation and collaboration between agents, little work explores competition , another important mechanism that fosters the development of society and economy. In this paper, we seek to examine the competition behaviors in LLM-based agents. We first propose a general framework to study the competition between agents. Then, we implement a practical competitive environment using GPT-4 to simulate a virtual town with two types of agents, including restaurant agents and customer agents. Specifically, the restaurant agents compete with each other to attract more customers, where the competition fosters them to transform, such as cultivating new operating strategies. The results of our experiments reveal several interesting findings ranging from social learning to Matthew Effect, which aligns well with existing sociological and economic theories. We believe that competition between agents deserves further investigation to help us understand society better. The code will be released soon."
                    },
                    {
                        "title": "PromptBench: A Unified Library for Evaluation of Large Language Models",
                        "abstract": "The evaluation of large language models (LLMs) is crucial to assess their performance and mitigate potential security risks. In this paper, we introduce PromptBench, a unified library to evaluate LLMs. It consists of several key components that are easily used and extended by researchers: prompt construction, prompt engineering, dataset and model loading, adversarial prompt attack, dynamic evaluation protocols, and analysis tools. PromptBench is designed to be an open, general, and flexible codebase for research purposes that can facilitate original study in creating new benchmarks, deploying downstream applications, and designing new evaluation protocols. The code is available at: https://github.com/microsoft/promptbench and will be continuously supported."
                    },
                    {
                        "title": "Non-IID always Bad? Semi-Supervised Heterogeneous Federated Learning with Local Knowledge Enhancement",
                        "abstract": "Federated learning (FL) is important for privacy-preserving services by training models without collecting raw user data. Most FL algorithms assume all data is annotated, which is impractical due to the high cost of labeling data in real applications. To alleviate the reliance on labeled data, semi-supervised federated learning (SSFL) has been proposed to utilize unlabeled data on clients to improve model performance. However, most existing methods either have privacy issues which share models trained on other clients, or generate pseudo-labels for unlabeled local datasets with the global model, which is usually biased towards the global data distribution. The latter may lead to sub-optimal accuracy of pseudo-labels, due to the gap between the local data distribution and the global model, especially in non-IID settings. In this paper, we propose a semi-supervised heterogeneous federated learning method with local knowledge enhancement, called FedLoKe, which aims to train an accurate global model from both labeled and unlabeled local data with non-IID distributions. Specifically, in FedLoKe, the server maintains a global model to capture global data distribution, and each client learns a local model to capture local data distribution. Since the distribution captured by the local model is aligned with the local data distribution, we utilize it to generate high-accuracy pseudo-labels of the unlabeled dataset for global model training. To prevent the local model from severely overfitting the small number of local labeled data, we further use the exponential moving average and apply the global model to generate pseudo-labels for local modeling training. Experiments on four datasets show the effectiveness of FedLoKe. Our code is available at: https://github.com/zcfinal/FedLoKe."
                    },
                    {
                        "title": "How Well Does GPT-4V(ision) Adapt to Distribution Shifts? A Preliminary Investigation",
                        "abstract": "In machine learning, generalization against distribution shifts -- where deployment conditions diverge from the training scenarios -- is crucial, particularly in fields like climate modeling, biomedicine, and autonomous driving. The emergence of foundation models, distinguished by their extensive pretraining and task versatility, has led to an increased interest in their adaptability to distribution shifts. GPT-4V(ision) acts as the most advanced publicly accessible multimodal foundation model, with extensive applications across various domains, including anomaly detection, video understanding, image generation, and medical diagnosis. However, its robustness against data distributions remains largely underexplored. Addressing this gap, this study rigorously evaluates GPT-4V's adaptability and generalization capabilities in dynamic environments, benchmarking against prominent models like CLIP, LLaVA, and Gemini. We delve into GPT-4V's zero-shot generalization across 13 diverse datasets spanning natural, medical, and molecular domains. We further investigate its adaptability to controlled data perturbations and examine the efficacy of in-context learning as a tool to enhance its adaptation. Our findings delineate GPT-4V's capability boundaries in distribution shifts, shedding light on its strengths and limitations across various scenarios. Importantly, this investigation contributes to our understanding of how AI foundation models generalize to distribution shifts, offering pivotal insights into their adaptability and robustness. The code is publicly available at https://github.com/jameszhou-gl/gpt-4v-distribution-shift."
                    },
                    {
                        "title": "Frustratingly Easy Model Generalization by Dummy Risk Minimization",
                        "abstract": "Empirical risk minimization (ERM) is a fundamental machine learning paradigm. However, its generalization ability is limited in various tasks. In this paper, we devise Dummy Risk Minimization (DuRM), a frustratingly easy and general technique to improve the generalization of ERM. DuRM is extremely simple to implement: just enlarging the dimension of the output logits and then optimizing using standard gradient descent. Moreover, we validate the efficacy of DuRM on both theoretical and empirical analysis. Theoretically, we show that DuRM derives greater variance of the gradient, which facilitates model generalization by observing better flat local minima. Empirically, we conduct evaluations of DuRM across different datasets, modalities, and network architectures on diverse tasks, including conventional classification, semantic segmentation, out-of-distribution generalization, adverserial training, and long-tailed recognition. Results demonstrate that DuRM could consistently improve the performance under all tasks with an almost free lunch manner. Furthermore, we show that DuRM is compatible with existing generalization techniques and we discuss possible limitations. We hope that DuRM could trigger new interest in the fundamental research on risk minimization."
                    },
                    {
                        "title": "Understanding and Mitigating the Label Noise in Pre-training on Downstream Tasks",
                        "abstract": "Pre-training on large-scale datasets and then fine-tuning on downstream tasks have become a standard practice in deep learning. However, pre-training data often contain label noise that may adversely affect the generalization of the model. This paper aims to understand the nature of noise in pre-training datasets and to mitigate its impact on downstream tasks. More specifically, through extensive experiments of supervised pre-training models on synthetic noisy ImageNet-1K and YFCC15M datasets, we demonstrate that while slight noise in pre-training can benefit in-domain (ID) transfer performance, where the training and testing data share the same distribution, it always deteriorates out-of-domain (OOD) performance, where training and testing data distribution are different. We empirically verify that the reason behind is noise in pre-training shapes the feature space differently. We then propose a light-weight black-box tuning method (NMTune) to affine the feature space to mitigate the malignant effect of noise and improve generalization on both ID and OOD tasks, considering one may not be able to fully fine-tune or even access the pre-trained models. We conduct practical experiments on popular vision and language models that are pre-trained on noisy data for evaluation of our approach. Our analysis and results show the importance of this interesting and novel research direction, which we term Noisy Model Learning."
                    },
                    {
                        "title": "EmotionPrompt: Leveraging Psychology for Large Language Models Enhancement via Emotional Stimulus",
                        "abstract": "Large language models (LLMs) have achieved signi\ufb01cant performance in many \ufb01elds, such as reasoning, language understanding, and math problem-solving, and are regarded as an important step to arti\ufb01cial general intelligence (AGI). However, the sensitivity of LLMs to prompts remain a major bottleneck for their daily adoption. In this paper, we take inspiration from psychology and propose EmotionPrompt to explore emotional intelligence to enhance the performance of LLMs. Our EmotionPrompt operates on a remarkably straightforward principle: the incorporation of emotional stimulus into prompts. Experimental re-sults demonstrate that our EmotionPrompt, using the same single prompt templates, signi\ufb01cantly out-performs the original prompt and Zero-shot-CoT in both zero-shot and few-shot settings on eight tasks with diverse models: ChatGPT, Vicuna-13b, Bloom, and Flan-T5-large. Furthermore, Emotion-Prompt was observed to improve both the truthfulness and informativeness. We believe that Emo-tionPrompt heralds a novel avenue for exploring interdisciplinary knowledge for interaction between humans and LLMs."
                    },
                    {
                        "title": "The Good, The Bad, and Why: Unveiling Emotions in Generative AI",
                        "abstract": "Emotion significantly impacts our daily behaviors and interactions. While recent generative AI models, such as large language models, have shown impressive performance in various tasks, it remains unclear whether they truly comprehend emotions. This paper aims to address this gap by incorporating psychological theories to gain a holistic understanding of emotions in generative AI models. Specifically, we propose three approaches: 1) EmotionPrompt to enhance AI model performance, 2) EmotionAttack to impair AI model performance, and 3) EmotionDecode to explain the effects of emotional stimuli, both benign and malignant. Through extensive experiments involving language and multi-modal models on semantic understanding, logical reasoning, and generation tasks, we demonstrate that both textual and visual EmotionPrompt can boost the performance of AI models while EmotionAttack can hinder it. Additionally, EmotionDecode reveals that AI models can comprehend emotional stimuli akin to the mechanism of dopamine in the human brain. Our work heralds a novel avenue for exploring psychology to enhance our understanding of generative AI models."
                    },
                    {
                        "title": "ZooPFL: Exploring Black-box Foundation Models for Personalized Federated Learning",
                        "abstract": "When personalized federated learning (FL) meets large foundation models, new challenges arise from various limitations in resources. In addition to typical limitations such as data, computation, and communication costs, access to the models is also often limited. This paper endeavors to solve both the challenges of limited resources and personalization. i.e., distribution shifts between clients. To do so, we propose a method named ZOOPFL that uses Zeroth-Order Optimization for Personalized Federated Learning. ZOOPFL avoids direct interference with the foundation models and instead learns to adapt its inputs through zeroth-order optimization. In addition, we employ simple yet effective linear projections to remap its predictions for personalization. To reduce the computation costs and enhance personalization, we propose input surgery to incorporate an auto-encoder with low-dimensional and client-specific embeddings. We provide theoretical support for ZOOPFL to analyze its convergence. Extensive empirical experiments on computer vision and natural language processing tasks using popular foundation models demonstrate its effectiveness for FL on black-box foundation models."
                    },
                    {
                        "title": "Improving Generalization of Adversarial Training via Robust Critical Fine-Tuning",
                        "abstract": "Deep neural networks are susceptible to adversarial examples, posing a significant security risk in critical applications. Adversarial Training (AT) is a well-established technique to enhance adversarial robustness, but it often comes at the cost of decreased generalization ability. This paper proposes Robustness Critical Fine-Tuning (RiFT), a novel approach to enhance generalization without compromising adversarial robustness. The core idea of RiFT is to exploit the redundant capacity for robustness by fine-tuning the adversarially trained model on its non-robust-critical module. To do so, we introduce module robust criticality (MRC), a measure that evaluates the significance of a given module to model robustness under worst-case weight perturbations. Using this measure, we identify the module with the lowest MRC value as the non-robust-critical module and fine-tune its weights to obtain fine-tuned weights. Subsequently, we linearly interpolate between the adversarially trained weights and fine-tuned weights to derive the optimal fine-tuned model weights. We demonstrate the efficacy of RiFT on ResNet18, ResNet34, and WideResNet34-10 models trained on CIFAR10, CIFAR100, and Tiny-ImageNet datasets. Our experiments show that RiFT can significantly improve both generalization and out-of-distribution robustness by around 1.5% while maintaining or even slightly enhancing adversarial robustness. Code is available at https://github.com/Immortalise/RiFT."
                    }
                ]
            },
            "324af0e6-fac2-47bf-9932-06dd7be4d043": {
                "pk": "324af0e6-fac2-47bf-9932-06dd7be4d043",
                "name": "Neil Zhenqiang Gong",
                "collaborators": [
                    "Lichao Sun",
                    "Jiawen Shi",
                    "Pan Zhou",
                    "Yue Huang",
                    "Yao Wan",
                    "Jindong Wang",
                    "Siyuan Wu",
                    "Qihui Zhang",
                    "Yixin Liu",
                    "Yuan Li",
                    "Chenrui Fan",
                    "Xing Xie",
                    "Dongping Chen",
                    "Xilong Wang",
                    "Hao Fu",
                    "Yixin Wu",
                    "Xinlei He",
                    "Pascal Berrang",
                    "Mathias Humbert",
                    "Michael Backes",
                    "Yang Zhang",
                    "Zenghui Yuan",
                    "Yinuo Liu",
                    "Haoran Wang",
                    "Chujie Gao",
                    "Yixin Huang",
                    "Wenhan Lyu",
                    "Yixuan Zhang",
                    "Xiner Li",
                    "Zheng Liu",
                    "Yijue Wang",
                    "Zhikun Zhang",
                    "B. Kailkhura",
                    "Caiming Xiong",
                    "Chaowei Xiao",
                    "Chun-Yan Li",
                    "Eric P. Xing",
                    "Furong Huang",
                    "Haodong Liu",
                    "Heng Ji",
                    "Hongyi Wang",
                    "Huan Zhang",
                    "Huaxiu Yao",
                    "M. Kellis",
                    "M. Zitnik",
                    "Meng Jiang",
                    "Mohit Bansal",
                    "James Zou",
                    "Jian Pei",
                    "Jian Liu",
                    "Jianfeng Gao",
                    "Jiawei Han",
                    "Jieyu Zhao",
                    "Jiliang Tang",
                    "John Mitchell",
                    "Kai Shu",
                    "Kaidi Xu",
                    "Kai-Wei Chang",
                    "Lifang He",
                    "Lifu Huang",
                    "M. Backes",
                    "Philip S. Yu",
                    "Pin-Yu Chen",
                    "Quanquan Gu",
                    "Ran Xu",
                    "Rex Ying",
                    "Shuiwang Ji",
                    "S. Jana",
                    "Tian-Xiang Chen",
                    "Tianming Liu",
                    "Tianying Zhou",
                    "William Wang",
                    "Xiang Li",
                    "Xiang-Yu Zhang",
                    "Xiao Wang",
                    "Xingyao Xie",
                    "Xun Chen",
                    "Xuyu Wang",
                    "Yan Liu",
                    "Yanfang Ye",
                    "Yinzhi Cao",
                    "Yue Zhao",
                    "Kaijie Zhu",
                    "Jiaheng Zhou",
                    "Zichen Wang",
                    "Hao Chen",
                    "Yidong Wang",
                    "Linyi Yang",
                    "Weirong Ye",
                    "Yue Zhang",
                    "Guangzhong Sun"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Privacy",
                    "Adversarial Attacks"
                ],
                "publications": [
                    {
                        "title": "Self-Cognition in Large Language Models: An Exploratory Study",
                        "abstract": "While Large Language Models (LLMs) have achieved remarkable success across various applications, they also raise concerns regarding self-cognition. In this paper, we perform a pioneering study to explore self-cognition in LLMs. Specifically, we first construct a pool of self-cognition instruction prompts to evaluate where an LLM exhibits self-cognition and four well-designed principles to quantify LLMs' self-cognition. Our study reveals that 4 of the 48 models on Chatbot Arena--specifically Command R, Claude3-Opus, Llama-3-70b-Instruct, and Reka-core--demonstrate some level of detectable self-cognition. We observe a positive correlation between model size, training data quality, and self-cognition level. Additionally, we also explore the utility and trustworthiness of LLM in the self-cognition state, revealing that the self-cognition state enhances some specific tasks such as creative writing and exaggeration. We believe that our work can serve as an inspiration for further research to study the self-cognition in LLMs."
                    },
                    {
                        "title": "StringLLM: Understanding the String Processing Capability of Large Language Models",
                        "abstract": "String processing, which mainly involves the analysis and manipulation of strings, is a fundamental component of modern computing. Despite the significant advancements of large language models (LLMs) in various natural language processing (NLP) tasks, their capability in string processing remains underexplored and underdeveloped. To bridge this gap, we present a comprehensive study of LLMs' string processing capability. In particular, we first propose StringLLM, a method to construct datasets for benchmarking string processing capability of LLMs. We use StringLLM to build a series of datasets, referred to as StringBench. It encompasses a wide range of string processing tasks, allowing us to systematically evaluate LLMs' performance in this area. Our evaluations indicate that LLMs struggle with accurately processing strings compared to humans. To uncover the underlying reasons for this limitation, we conduct an in-depth analysis and subsequently propose an effective approach that significantly enhances LLMs' string processing capability via fine-tuning. This work provides a foundation for future research to understand LLMs' string processing capability. Our code and data are available at https://github.com/wxl-lxw/StringLLM."
                    },
                    {
                        "title": "Link Stealing Attacks Against Inductive Graph Neural Networks",
                        "abstract": "A graph neural network (GNN) is a type of neural network that is specifically designed to process graph-structured data. Typically, GNNs can be implemented in two settings, including the transductive setting and the inductive setting. In the transductive setting, the trained model can only predict the labels of nodes that were observed at the training time. In the inductive setting, the trained model can be generalized to new nodes/graphs. Due to its flexibility, the inductive setting is the most popular GNN setting at the moment. Previous work has shown that transductive GNNs are vulnerable to a series of privacy attacks. However, a comprehensive privacy analysis of inductive GNN models is still missing. This paper fills the gap by conducting a systematic privacy analysis of inductive GNNs through the lens of link stealing attacks. We propose two types of link stealing attacks, i.e., posterior-only attacks and combined attacks. We define threat models of the posterior-only attacks with respect to node topology and the combined attacks by considering combinations of posteriors, node attributes, and graph features. Extensive evaluation on six real-world datasets demonstrates that inductive GNNs leak rich information that enables link stealing attacks with advantageous properties. Even attacks with no knowledge about graph structures can be effective. We also show that our attacks are robust to different node similarities and different graph features. As a counterpart, we investigate two possible defenses and discover they are ineffective against our attacks, which calls for more effective defenses."
                    },
                    {
                        "title": "Optimization-based Prompt Injection Attack to LLM-as-a-Judge",
                        "abstract": "LLM-as-a-Judge uses a large language model (LLM) to select the best response from a set of candidates for a given question. LLM-as-a-Judge has many applications such as LLM-powered search, reinforcement learning with AI feedback (RLAIF), and tool selection. In this work, we propose JudgeDeceiver, an optimization-based prompt injection attack to LLM-as-a-Judge. JudgeDeceiver injects a carefully crafted sequence into an attacker-controlled candidate response such that LLM-as-a-Judge selects the candidate response for an attacker-chosen question no matter what other candidate responses are. Specifically, we formulate finding such sequence as an optimization problem and propose a gradient based method to approximately solve it. Our extensive evaluation shows that JudgeDeceive is highly effective, and is much more effective than existing prompt injection attacks that manually craft the injected sequences and jailbreak attacks when extended to our problem. We also show the effectiveness of JudgeDeceiver in three case studies, i.e., LLM-powered search, RLAIF, and tool selection. Moreover, we consider defenses including known-answer detection, perplexity detection, and perplexity windowed detection. Our results show these defenses are insufficient, highlighting the urgent need for developing new defense strategies."
                    },
                    {
                        "title": "TrustLLM: Trustworthiness in Large Language Models",
                        "abstract": "Large language models (LLMs), exemplified by ChatGPT, have gained considerable attention for their excellent natural language processing capabilities. Nonetheless, these LLMs present many challenges, particularly in the realm of trustworthiness. Therefore, ensuring the trustworthiness of LLMs emerges as an important topic. This paper introduces TrustLLM, a comprehensive study of trustworthiness in LLMs, including principles for different dimensions of trustworthiness, established benchmark, evaluation, and analysis of trustworthiness for mainstream LLMs, and discussion of open challenges and future directions. Specifically, we first propose a set of principles for trustworthy LLMs that span eight different dimensions. Based on these principles, we further establish a benchmark across six dimensions including truthfulness, safety, fairness, robustness, privacy, and machine ethics. We then present a study evaluating 16 mainstream LLMs in TrustLLM, consisting of over 30 datasets. Our findings firstly show that in general trustworthiness and utility (i.e., functional effectiveness) are positively related. Secondly, our observations reveal that proprietary LLMs generally outperform most open-source counterparts in terms of trustworthiness, raising concerns about the potential risks of widely accessible open-source LLMs. However, a few open-source LLMs come very close to proprietary ones. Thirdly, it is important to note that some LLMs may be overly calibrated towards exhibiting trustworthiness, to the extent that they compromise their utility by mistakenly treating benign prompts as harmful and consequently not responding. Finally, we emphasize the importance of ensuring transparency not only in the models themselves but also in the technologies that underpin trustworthiness. Knowing the specific trustworthy technologies that have been employed is crucial for analyzing their effectiveness."
                    },
                    {
                        "title": "M ETA T OOL B ENCHMARK : D ECIDING W HETHER TO U SE T OOLS AND W HICH TO U SE",
                        "abstract": "Large language models (LLMs) have garnered significant attention due to their impressive natural language processing (NLP) capabilities. Recently, many studies have focused on the tool utilization ability of LLMs. They primarily investigated how LLMs effectively collaborate with given specific tools. However, in scenarios where LLMs serve as intelligent agents, as seen in applications like AutoGPT and MetaGPT, LLMs are expected to engage in intricate decision-making processes that involve deciding whether to employ a tool and selecting the most suitable tool(s) from a collection of available tools to fulfill user requests. Therefore, in this paper, we introduce METATOOL, a benchmark designed to evaluate whether LLMs have tool usage awareness and can correctly choose tools. Specifically, we create a dataset called TOOLE within the benchmark. This dataset contains various types of user queries in the form of prompts that trigger LLMs to use tools, including both single-tool and multi-tool scenarios. Subsequently, we set the tasks for both tool usage awareness and tool selection. We define four subtasks from different perspectives in tool selection, including tool selection with similar choices, tool selection in specific scenarios , tool selection with possible reliability issues, and multi-tool selection. We conduct experiments involving nine popular LLMs and find that the majority of them still struggle to effectively select tools, highlighting the existing gaps between LLMs and genuine intelligent agents. However, through the error analysis, we found there is still significant room for improvement. Finally, we conclude with insights for tool developers that follow ChatGPT to provide detailed descriptions that can enhance the tool selection performance of LLMs. Our TOOLE dataset is available at the URL 1."
                    },
                    {
                        "title": "MetaTool Benchmark for Large Language Models: Deciding Whether to Use Tools and Which to Use",
                        "abstract": "Large language models (LLMs) have garnered significant attention due to their impressive natural language processing (NLP) capabilities. Recently, many studies have focused on the tool utilization ability of LLMs. They primarily investigated how LLMs effectively collaborate with given specific tools. However, in scenarios where LLMs serve as intelligent agents, as seen in applications like AutoGPT and MetaGPT, LLMs are expected to engage in intricate decision-making processes that involve deciding whether to employ a tool and selecting the most suitable tool(s) from a collection of available tools to fulfill user requests. Therefore, in this paper, we introduce MetaTool, a benchmark designed to evaluate whether LLMs have tool usage awareness and can correctly choose tools. Specifically, we create a dataset called ToolE within the benchmark. This dataset contains various types of user queries in the form of prompts that trigger LLMs to use tools, including both single-tool and multi-tool scenarios. Subsequently, we set the tasks for both tool usage awareness and tool selection. We define four subtasks from different perspectives in tool selection, including tool selection with similar choices, tool selection in specific scenarios, tool selection with possible reliability issues, and multi-tool selection. We conduct experiments involving eight popular LLMs and find that the majority of them still struggle to effectively select tools, highlighting the existing gaps between LLMs and genuine intelligent agents. However, through the error analysis, we found there is still significant room for improvement. Finally, we conclude with insights for tool developers -- we strongly recommend that tool developers choose an appropriate rewrite model for generating new descriptions based on the downstream LLM the tool will apply to. Our code is in https://github.com/HowieHwong/MetaTool."
                    },
                    {
                        "title": "PromptBench: Towards Evaluating the Robustness of Large Language Models on Adversarial Prompts",
                        "abstract": "The increasing reliance on Large Language Models (LLMs) across academia and industry necessitates a comprehensive understanding of their robustness to prompts. In response to this vital need, we introduce PromptBench, a robustness benchmark designed to measure LLMs\u2019 resilience to adversarial prompts. This study uses a plethora of adversarial textual attacks targeting prompts across multiple levels: character, word, sentence, and semantic. These prompts are then employed in diverse tasks, such as sentiment analysis, natural language inference, reading comprehension, machine translation, and math problem-solving. Our study generates 4 , 032 adversarial prompts, meticulously evaluated over 8 tasks and 13 datasets, with 567 , 084 test samples in total. Our findings demonstrate that contemporary LLMs are vulnerable to adversarial prompts. Furthermore, we present comprehensive analysis to understand the mystery behind prompt robustness and its transferability. We then offer insightful robustness analysis and pragmatic recommendations for prompt composition, beneficial to both researchers and everyday users. We make our code, prompts, and methodologies to generate adversarial prompts publicly accessible, thereby enabling and encouraging collaborative exploration in this pivotal field: https://github.com/microsoft/promptbench ."
                    },
                    {
                        "title": "Protecting Privacy in Location-Based Services Using K-Anonymity without Cloaked Region",
                        "abstract": "The emerging location-detection devices together with ubiquitous connectivity have enabled a large variety of location-based services (LBS). Unfortunately, LBS may threaten the users\u2019 privacy. K-anonymity cloaking the user location to K-anonymizing spatial region (K-ASR) has been extensively studied to protect privacy in LBS. Traditional K-anonymity method needs complex query processing algorithms at the server side. SpaceTwist [8] rectifies the above shortcoming of traditional K-anonymity since it only requires incremental nearest neighbor (INN) queries processing techniques at the server side. However, Space Twist may fail since it cannot guarantee K-anonymity. In this paper, our proposed framework, called KAWCR (K-anonymity Without Cloaked Region), rectifies the shortcomings and retains the advantages of the above two techniques. KAWCR only needs the server to process INN queries and can guarantee that the users issuing the query is indistinguishable from at least K-1 other users. The extensive experimental results show that the communication cost of KAWCR for kNN queries is lower than that of both traditional K-anonymity and SpaceTwist."
                    }
                ]
            },
            "24bb6610-f895-49e8-8645-804171d353d0": {
                "pk": "24bb6610-f895-49e8-8645-804171d353d0",
                "name": "Diyi Yang",
                "collaborators": [
                    "Jiaao Chen",
                    "Raj Sanjay Shah",
                    "Omar Shaikh",
                    "Yi-Chia Wang",
                    "Robert E. Kraut",
                    "Shirley Anugrah Hayati",
                    "Tony Wang",
                    "Haard K Shah",
                    "Hayeong Song",
                    "Zhengyang Qi",
                    "J. Stasko",
                    "Valentino Chai",
                    "Michele J. Gelfand",
                    "Michael S. Bernstein",
                    "Shang-ling Hsu",
                    "Prathik Senthil",
                    "Zahra Ashktorab",
                    "Casey Dugan",
                    "Werner Geyer",
                    "William B. Held",
                    "Yih-Ling Liu",
                    "Anushk Mittal",
                    "A. Bruckman",
                    "Cassandra Ung",
                    "Felipe Chacon",
                    "Bolin Lai",
                    "Hongxin Zhang",
                    "Miao Liu",
                    "Aryan Pariani",
                    "Fiona Ryan",
                    "Wenqi Jia",
                    "J. Rehg",
                    "Faye Holt",
                    "A. Agarwal",
                    "Yujia Gao",
                    "Jinu Jang",
                    "Aditya Gupta",
                    "Jiacheng Xu",
                    "Shyam Upadhyay",
                    "Manaal Faruqui",
                    "Dinghan Shen",
                    "Weizhu Chen",
                    "Zhenhui Peng",
                    "Xiaojuan Ma",
                    "Ka Wing Tsang",
                    "Qingyu Guo",
                    "Yufan Huang",
                    "Yanzhe Zhang",
                    "Xuezhi Wang",
                    "Aaron L Halfaker",
                    "R. Kraut",
                    "E. Hovy",
                    "Sebastian Gehrmann",
                    "Tosin P. Adewumi",
                    "Karmanya Aggarwal",
                    "Pawan Sasanka Ammanamanchi",
                    "Aremu Anuoluwapo",
                    "Antoine Bosselut",
                    "Khyathi Raghavi Chandu",
                    "Miruna Clinciu",
                    "Dipanjan Das",
                    "Kaustubh D. Dhole",
                    "Wanyu Du",
                    "Esin Durmus",
                    "Ondrej Dusek",
                    "Chris C. Emezue",
                    "Varun Prashant Gangal",
                    "Cristina Garbacea",
                    "Tatsunori B. Hashimoto",
                    "Yufang Hou",
                    "Yacine Jernite",
                    "Harsh Jhamtani",
                    "Yangfeng Ji",
                    "Shailza Jolly",
                    "Mihir Kale",
                    "Dhruv Kumar",
                    "Faisal Ladhak",
                    "Aman Madaan",
                    "Mounica Maddela",
                    "Khyati Mahajan",
                    "Saad Mahamood",
                    "Bodhisattwa Prasad Majumder",
                    "Pedro Henrique Martins",
                    "Angelina McMillan-Major",
                    "Simon Mille",
                    "Emiel van Miltenburg",
                    "Moin Nadeem",
                    "Shashi Narayan",
                    "Vitaly Nikolaev",
                    "Andre Niyongabo Rubungo",
                    "Salomey Osei",
                    "Ankur P. Parikh",
                    "Laura Perez-Beltrachini",
                    "Niranjan Rao",
                    "Vikas Raunak",
                    "Juan Diego Rodriguez",
                    "Sashank Santhanam",
                    "Jo\u00e3o Sedoc",
                    "Thibault Sellam",
                    "Samira Shaikh"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Human-Computer Interaction",
                    "Online Communities",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Metrics for Peer Counseling: Triangulating Success Outcomes for Online Therapy Platforms",
                        "abstract": "Extensive research has been published on the conversational factors of effective volunteer peer counseling on online mental health platforms (OMHPs). However, studies differ in how they define and measure success outcomes, with most prior work examining only a single success metric. In this work, we model the relationship between previously reported linguistic predictors of effective counseling with four outcomes following a peer-to-peer session on a single OMHP: retention in the community, following up on a previous session with a counselor, users\u2019 evaluation of a counselor, and changes in users\u2019 mood. Results show that predictors correlate negatively with community retention but positively with users following up with and giving higher evaluations to individual counselors. We suggest actionable insights for therapy platform design and outcome measurement based on findings that the relationship between predictors and outcomes of successful conversations depends on differences in measurement construct and operationalization."
                    },
                    {
                        "title": "Understanding People\u2019s Needs in Viewing Diverse Social Opinions about Controversial Topics",
                        "abstract": "Social media (i.e., Reddit) users are overloaded with people\u2019s opinions when viewing discourses about divisive topics. Traditional user interfaces in such media present those opinions in a linear structure, which can limit users in viewing diverse social opinions at scale. Prior work has recognized this limitation, that the linear structure can reinforce biases, where a certain point of view becomes widespread simply because many viewers seem to believe it. This limitation can make it difficult for users to have a truly conversational mode of mediated discussion. Thus, when designing a user interface for viewing people\u2019s opinions, we should consider ways to mitigate selective exposure to information and polarization of opinions. We conducted a needs-finding study with 11 Reddit users, who follow climate change threads and make posts and comments regularly. In the study, we aimed to understand key limitations in people viewing online controversial discourses and to extract design implications to address these problems. Our findings discuss potential future directions to address these problems."
                    },
                    {
                        "title": "Rehearsal: Simulating Conflict to Teach Conflict Resolution",
                        "abstract": "Interpersonal conflict is an uncomfortable but unavoidable fact of life. Navigating conflict successfully is a skill\u2014one that can be learned through deliberate practice\u2014but few have access to effective training or feedback. To expand this access, we introduce Rehearsal, a system that allows users to rehearse conflicts with a believable simulated interlocutor, explore counterfactual \u201cwhat if?\u201d scenarios to identify alternative conversational paths, and learn through feedback on how and when to apply specific conflict strategies. Users can utilize Rehearsal to practice handling a variety of predefined conflict scenarios, from office disputes to relationship issues, or they can choose to create their own setting. To enable Rehearsal, we develop IRP prompting, a method of conditioning output of a large language model on the influential Interest-Rights-Power (IRP) theory from conflict resolution. Rehearsal uses IRP to generate utterances grounded in conflict resolution theory, guiding users towards counterfactual conflict resolution strategies that help de-escalate difficult conversations. In a between-subjects evaluation, 40 participants engaged in an actual conflict with a confederate after training. Compared to a control group with lecture material covering the same IRP theory, participants with simulated training from Rehearsal significantly improved their performance in the unaided conflict: they reduced their use of escalating competitive strategies by an average of 67%, while doubling their use of cooperative strategies. Overall, Rehearsal highlights the potential effectiveness of language models as tools for learning and practicing interpersonal skills."
                    },
                    {
                        "title": "Helping the Helper: Supporting Peer Counselors via AI-Empowered Practice and Feedback",
                        "abstract": "Millions of users come to online peer counseling platforms to seek support on diverse topics ranging from relationship stress to anxiety. However, studies show that online peer support groups are not always as effective as expected largely due to users' negative experiences with unhelpful counselors. Peer counselors are key to the success of online peer counseling platforms, but most of them often do not have systematic ways to receive guidelines or supervision. In this work, we introduce CARE: an interactive AI-based tool to empower peer counselors through automatic suggestion generation. During the practical training stage, CARE helps diagnose which specific counseling strategies are most suitable in the given context and provides tailored example responses as suggestions. Counselors can choose to select, modify, or ignore any suggestion before replying to the support seeker. Building upon the Motivational Interviewing framework, CARE utilizes large-scale counseling conversation data together with advanced natural language generation techniques to achieve these functionalities. We demonstrate the efficacy of CARE by performing both quantitative evaluations and qualitative user studies through simulated chats and semi-structured interviews. We also find that CARE especially helps novice counselors respond better in challenging situations."
                    },
                    {
                        "title": "Mixture of Virtual Prompts: Efficient Fine-tuning with Virtual Prompts for Natural Language Understanding",
                        "abstract": "Fine-tuning large-scale pre-trained language models tasks has become the de facto standard for downstream NLP tasks. However, fine-tuning all parameters of these large models brings expensive computation and storage cost. In this work, inspired by model prompting, we propose Mixture of Virtual Prompts adaptation (MVP adaptation). We first learn virtual prompts downstream tasks to inject task or language specific knowledge, while the pre-trained models are frozen to provide general knowledge. We extend virtual prompting to deal with varying complexity inputs, learning a set of virtual prompts with distinct capacities to learn different levels of knowledge, and dynamically aggregating them to adapt for each individual sequence. Experiments conducted on 5 natural language understanding tasks demonstrate that our methods could achieve comparable performances relative to the full model. Importantly, we show the superiority of MVP in terms of linguistic variation settings which allows efficient adaption to different dialects. We outperform strong efficient finetuning base-lines using only 0.03% of its parameters."
                    },
                    {
                        "title": "Will AI Console Me when I Lose my Pet? Understanding Perceptions of AI-Mediated Email Writing",
                        "abstract": "Large language models are increasingly mediating, modifying, and even generating messages for users, but the receivers of these messages may not be aware of the involvement of AI. To examine this emerging direction of AI-Mediated Communication (AI-MC), we investigate people\u2019s perceptions of AI written messages. We analyze how such perceptions change in accordance with the interpersonal emphasis of a given message. We conducted both large-scale surveys and in-depth interviews to investigate how a diverse set of factors influence people\u2019s perceived trust in AI-mediated writing of emails. We found that people\u2019s trust in email writers decreased when they were told that AI was involved in the writing process. Surprisingly trust increased when AI was used for writing more interpersonal emails (as opposed to more transactional ones). Our study provides insights regarding how people perceive AI-MC and has practical design implications on building AI-based products to aid human interlocutors in communication1."
                    },
                    {
                        "title": "Six Feet Apart: Online Payments During the COVID-19 Pandemic",
                        "abstract": "Since the COVID-19 pandemic, businesses have faced unprecedented challenges when trying to remain open. Because COVID-19 spreads through aerosolized droplets, businesses were forced to distance their services; in some cases, distancing may have involved moving business services online. In this work, we explore digitization strategies used by small businesses that remained open during the pandemic, and survey/interview small businesses owners to understand preliminary challenges associated with moving online. Furthermore, we analyze payments from 400K businesses across Japan, Australia, United States, Great Britain, and Canada. Following initial government interventions, we observe (at minimum for each country) a 47% increase in digitizing businesses compared to pre-pandemic levels, with about 80% of surveyed businesses digitizing in under a week. From both our quantitative models and our surveys/interviews, we find that businesses rapidly digitized at the start of the pandemic in preparation of future uncertainty. We also conduct a case-study of initial digitization in the United States, examining finer relationships between specific government interventions, business sectors, political orientation, and resulting digitization shifts. Finally, we discuss the implications of rapid & widespread digitization for small businesses in the context of usability challenges and interpersonal interactions, while highlighting potential shifts in pre-existing social norms."
                    },
                    {
                        "title": "Werewolf Among Us: A Multimodal Dataset for Modeling Persuasion Behaviors in Social Deduction Games",
                        "abstract": "Persuasion modeling is a key building block for conversational agents. Existing works in this direction are limited to analyzing textual dialogue corpus. We argue that visual signals also play an important role in understanding human persuasive behaviors. In this paper, we introduce the first multimodal dataset for modeling persuasion behaviors. Our dataset includes 199 dialogue transcriptions and videos captured in a multi-player social deduction game setting, 26,647 utterance level annotations of persuasion strategy, and game level annotations of deduction game outcomes. We provide extensive experiments to show how dialogue context and visual signals benefit persuasion strategy prediction. We also explore the generalization ability of language models for persuasion modeling and the role of persuasion strategies in predicting social deduction game outcomes. Our dataset, code, and models can be found at https://persuasion-deductiongame.socialai-data.org."
                    },
                    {
                        "title": "Modeling Motivational Interviewing Strategies on an Online Peer-to-Peer Counseling Platform",
                        "abstract": "Millions of people participate in online peer-to-peer support sessions, yet there has been little prior research on systematic psychology-based evaluations of fine-grained peer-counselor behavior in relation to client satisfaction. This paper seeks to bridge this gap by mapping peer-counselor chat-messages to motivational interviewing (MI) techniques. We annotate 14,797 utterances from 734 chat conversations using 17 MI techniques and introduce four new interviewing codes such as ''chit-chat'' and ''inappropriate'' to account for the unique conversational patterns observed on online platforms. We automate the process of labeling peer-counselor responses to MI techniques by fine-tuning large domain-specific language models and then use these automated measures to investigate the behavior of the peer counselors via correlational studies. Specifically, we study the impact of MI techniques on the conversation ratings to investigate the techniques that predict clients' satisfaction with their counseling sessions. When counselors use techniques such as reflection and affirmation, clients are more satisfied. Examining volunteer counselors' change in usage of techniques suggest that counselors learn to use more introduction and open questions as they gain experience. This work provides a deeper understanding of the use of motivational interviewing techniques on peer-to-peer counselor platforms and sheds light on how to build better training programs for volunteer counselors on online platforms."
                    },
                    {
                        "title": "Weakly-Supervised Hierarchical Models for Predicting Persuasive Strategies in Good-faith Textual Requests",
                        "abstract": "Modeling persuasive language has the potential to better facilitate our decision-making processes. Despite its importance, computational modeling of persuasion is still in its infancy, largely due to the lack of benchmark datasets that can provide quantitative labels of persuasive strategies to expedite this line of research. To this end, we introduce a large-scale multi-domain text corpus for modeling persuasive strategies in good-faith text requests. Moreover, we design a hierarchical weakly-supervised latent variable model that can leverage partially labeled data to predict such associated persuasive strategies for each sentence, where the supervision comes from both the overall document-level labels and very limited sentence-level labels. Experimental results showed that our proposed method outperformed existing semi-supervised baselines significantly. We have publicly released our code at https://github.com/GT-SALT/Persuasion_Strategy_WVAE."
                    },
                    {
                        "title": "Understanding the Usage of Online Media for Parenting from Infancy to Preschool At Scale",
                        "abstract": "New parents, defined as parents of children between the infant and preschooler stages, are increasingly turning to online media to exchange support and information to help with their life-changing transition. Understanding parents\u2019 discussion online is crucial to the design and development of technologies that can better support their media interaction. This work studies how new parents use online media using a large-scale parenting corpus. To do so, we first employed a card-sorting methodology to identify a set of parenting topics, with which we trained BERT classifiers to automatically identify the topics of Reddit posts. We then investigate at scale what parenting topics were talked about most by new parents, how topics change over the course of their participation, and how interactions with different topics affect members\u2019 engagement in the community. We conclude with implications of our research in designing future research and online parenting communities."
                    },
                    {
                        "title": "Disfl-QA: A Benchmark Dataset for Understanding Disfluencies in Question Answering",
                        "abstract": "Disfluencies is an under-studied topic in NLP, even though it is ubiquitous in human conversation. This is largely due to the lack of datasets containing disfluencies. In this paper, we present a new challenge question answering dataset, Disfl-QA, a derivative of SQuAD, where humans introduce contextual disfluencies in previously fluent questions. Disfl-QA contains a variety of challenging disfluencies that require a more comprehensive understanding of the text than what was necessary in prior datasets. Experiments show that the performance of existing state-of-the-art question answering models degrades significantly when tested on Disfl-QA in a zero-shot setting.We show data augmentation methods partially recover the loss in performance and also demonstrate the efficacy of using gold data for fine-tuning. We argue that we need large-scale disfluency datasets in order for NLP models to be robust to them. The dataset is publicly available at: https://github.com/google-research-datasets/disfl-qa."
                    },
                    {
                        "title": "HiddenCut: Simple Data Augmentation for Natural Language Understanding with Better Generalizability",
                        "abstract": "Fine-tuning large pre-trained models with task-specific data has achieved great success in NLP. However, it has been demonstrated that the majority of information within the self-attention networks is redundant and not utilized effectively during the fine-tuning stage. This leads to inferior results when generalizing the obtained models to out-of-domain distributions. To this end, we propose a simple yet effective data augmentation technique, HiddenCut, to better regularize the model and encourage it to learn more generalizable features. Specifically, contiguous spans within the hidden space are dynamically and strategically dropped during training. Experiments show that our HiddenCut method outperforms the state-of-the-art augmentation methods on the GLUE benchmark, and consistently exhibits superior generalization performances on out-of-distribution and challenging counterexamples. We have publicly released our code at https://github.com/GT-SALT/HiddenCut."
                    },
                    {
                        "title": "Effects of Support-Seekers\u2019 Community Knowledge on Their Expressed Satisfaction with the Received Comments in Mental Health Communities",
                        "abstract": "Online mental health communities (OMHCs) are prominent resources for improving people\u2019s mental wellbeing. An immediate cue of such improvement is support-seekers\u2019 satisfaction expressed in their replies to the received comments. However, the comments that seekers find satisfying may change with their community knowledge, e.g., measured by tenure and posting experience in that community. In this paper, we first model the amount of satisfaction conveyed in the support-seekers\u2019 replies to the received comments. Then we quantitatively examine how seekers\u2019 expressed satisfaction is affected by their community knowledge, sought and received support in an OMHC. Results show that support-seekers with more posting experience generally display less contentment to the received comments. Compared to newcomers, higher tenured members express less satisfaction when receiving informational support. We also found that support matching positively predicts seekers\u2019 satisfaction regardless of their community knowledge. Our findings have implications for OMHCs to satisfy support-seekers through their community knowledge."
                    },
                    {
                        "title": "Continual Learning for Text Classification with Information Disentanglement Based Regularization",
                        "abstract": "Continual learning has become increasingly important as it enables NLP models to constantly learn and gain knowledge over time. Previous continual learning methods are mainly designed to preserve knowledge from previous tasks, without much emphasis on how to well generalize models to new tasks. In this work, we propose an information disentanglement based regularization method for continual learning on text classification. Our proposed method first disentangles text hidden spaces into representations that are generic to all tasks and representations specific to each individual task, and further regularizes these representations differently to better constrain the knowledge required to generalize. We also introduce two simple auxiliary tasks: next sentence prediction and task-id prediction, for learning better generic and specific representation spaces. Experiments conducted on large-scale benchmarks demonstrate the effectiveness of our method in continual text classification tasks with various sequences and lengths over state-of-the-art baselines. We have publicly released our code at https://github.com/GT-SALT/IDBR."
                    },
                    {
                        "title": "Who Did What: Editor Role Identification in Wikipedia",
                        "abstract": "    Understanding the social roles played by contributors to online communities can facilitate the process of task routing. In this work, we develop new techniques to find roles in Wikipedia based on editors' low-level edit types and investigate how work contributed by people from different roles affect the article quality. To do this, we first built machine-learning models to automatically identify the edit categories associated with edits. We then applied a graphical model analogous to Latent Dirichlet Allocation to uncover the latent roles in editors' edit histories. Applying this technique revealed eight different roles editors play. Finally, we validated how our identified roles collaborate to improve the quality of articles. The results demonstrate that editors carrying on different roles contribute differently in terms of edit categories and articles in different quality stages need different types of editors. Implications for editor role identification and the validation of role contribution are discussed.   "
                    },
                    {
                        "title": "The GEM Benchmark: Natural Language Generation, its Evaluation and Metrics",
                        "abstract": "We introduce GEM, a living benchmark for natural language Generation (NLG), its Evaluation, and Metrics. Measuring progress in NLG relies on a constantly evolving ecosystem of automated metrics, datasets, and human evaluation standards. Due to this moving target, new models often still evaluate on divergent anglo-centric corpora with well-established, but flawed, metrics. This disconnect makes it challenging to identify the limitations of current models and opportunities for progress. Addressing this limitation, GEM provides an environment in which models can easily be applied to a wide set of tasks and in which evaluation strategies can be tested. Regular updates to the benchmark will help NLG research become more multilingual and evolve the challenge alongside models. This paper serves as the description of the data for the 2021 shared task at the associated GEM Workshop."
                    },
                    {
                        "title": "RECAST",
                        "abstract": "With the widespread use of toxic language online, platforms are increasingly using automated systems that leverage advances in natural language processing to automatically flag and remove toxic comments. However, most automated systems, when detecting and moderating toxic language, do not provide feedback to their users, let alone provide an avenue of recourse for these users to make actionable changes. We present our work, RECAST, an interactive, open-sourced web tool for visualizing these models' toxic predictions, while providing alternative suggestions for flagged toxic language. Our work also provides users with a new path of recourse when using these automated moderation tools. RECAST highlights text responsible for classifying toxicity, and allows users to interactively substitute potentially toxic phrases with neutral alternatives. We examined the effect of RECAST via two large-scale user evaluations, and found that RECAST was highly effective at helping users reduce toxicity as detected through the model. Users also gained a stronger understanding of the underlying toxicity criterion used by black-box models, enabling transparency and recourse. In addition, we found that when users focus on optimizing language for these models instead of their own judgement (which is the implied incentive and goal of deploying automated models), these models cease to be effective classifiers of toxicity compared to human annotations. This opens a discussion for how toxicity detection models work and should work, and their effect on the future of online discourse."
                    }
                ]
            },
            "5b9587a7-5771-4c77-8783-90cecb306874": {
                "pk": "5b9587a7-5771-4c77-8783-90cecb306874",
                "name": "Xing Xie",
                "collaborators": [
                    "Jindong Wang",
                    "Kaijie Zhu",
                    "Hao Chen",
                    "Qinlin Zhao",
                    "Yidong Wang",
                    "Cheng Li",
                    "Yixuan Zhang",
                    "Ruochen Xu",
                    "Yue Zhang",
                    "Linyi Yang",
                    "Bhiksha Raj",
                    "Zhuohao Yu",
                    "Masashi Sugiyama",
                    "Yiqiao Jin",
                    "Wenxin Hou",
                    "Jianxun Lian",
                    "Chang Gao",
                    "Wenjin Yao",
                    "Wei Ye",
                    "Shikun Zhang",
                    "Mengzhou Chen",
                    "Sunayana Sitaram",
                    "Damien Teney",
                    "Qingsong Wen",
                    "Lei Feng",
                    "Xiang Li",
                    "Rita Singh",
                    "Jio Oh",
                    "Soyeon Kim",
                    "Junseok Seo",
                    "Steven Euijong Whang",
                    "Xixu Hu",
                    "Runkai Zheng",
                    "Cheuk Hang Leung",
                    "Qi Wu",
                    "Zihan Wang",
                    "Ran Tao",
                    "Hongxin Wei",
                    "Zhiquan Tan",
                    "Lai Wei",
                    "Weiran Huang",
                    "Cunxiang Wang",
                    "Xiaoze Liu",
                    "Yuanhao Yue",
                    "Xiangru Tang",
                    "Tianhang Zhang",
                    "Cheng Jiayang",
                    "Yunzhi Yao",
                    "Wenyang Gao",
                    "Xuming Hu",
                    "Zehan Qi",
                    "Zheng Zhang",
                    "Yachuan Liu",
                    "Liang Chen",
                    "Qiaozhu Mei",
                    "Shuibai Zhang",
                    "Guangsheng Bao",
                    "Weirong Ye",
                    "Weizhu Chen",
                    "Xinyi Wang",
                    "Fang Luo",
                    "Qiang Yang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Evaluation Framework",
                    "Cultural AI"
                ],
                "publications": [
                    {
                        "title": "DyVal 2: Dynamic Evaluation of Large Language Models by Meta Probing Agents",
                        "abstract": "Evaluation of large language models (LLMs) has raised great concerns in the community due to the issue of data contamination. Existing work designed evaluation protocols using well-defined algorithms for specific tasks, which cannot be easily extended to diverse scenarios. Moreover, current evaluation benchmarks can only provide the overall benchmark results and cannot support a fine-grained and multifaceted analysis of LLMs' abilities. In this paper, we propose meta probing agents (MPA), a general dynamic evaluation protocol inspired by psychometrics to evaluate LLMs. MPA is the key component of DyVal 2, which naturally extends the previous DyVal~\\citep{zhu2023dyval}. MPA designs the probing and judging agents to automatically transform an original evaluation problem into a new one following psychometric theory on three basic cognitive abilities: language understanding, problem solving, and domain knowledge. These basic abilities are also dynamically configurable, allowing multifaceted analysis. We conducted extensive evaluations using MPA and found that most LLMs achieve poorer performance, indicating room for improvement. Our multifaceted analysis demonstrated the strong correlation between the basic abilities and an implicit Matthew effect on model size, i.e., larger models possess stronger correlations of the abilities. MPA can also be used as a data augmentation approach to enhance LLMs. Code is available at: https://github.com/microsoft/promptbench."
                    },
                    {
                        "title": "KIEval: A Knowledge-grounded Interactive Evaluation Framework for Large Language Models",
                        "abstract": "Automatic evaluation methods for large language models (LLMs) are hindered by data contamination, leading to inflated assessments of their effectiveness. Existing strategies, which aim to detect contaminated texts, focus on quantifying contamination status instead of accurately gauging model performance. In this paper, we introduce KIEval, a Knowledge-grounded Interactive Evaluation framework, which incorporates an LLM-powered\"interactor\"role for the first time to accomplish a dynamic contamination-resilient evaluation. Starting with a question in a conventional LLM benchmark involving domain-specific knowledge, KIEval utilizes dynamically generated, multi-round, and knowledge-focused dialogues to determine whether a model's response is merely a recall of benchmark answers or demonstrates a deep comprehension to apply knowledge in more complex conversations. Extensive experiments on seven leading LLMs across five datasets validate KIEval's effectiveness and generalization. We also reveal that data contamination brings no contribution or even negative effect to models' real-world applicability and understanding, and existing contamination detection methods for LLMs can only identify contamination in pre-training but not during supervised fine-tuning."
                    },
                    {
                        "title": "CultureLLM: Incorporating Cultural Differences into Large Language Models",
                        "abstract": "Large language models (LLMs) are reported to be partial to certain cultures owing to the training data dominance from the English corpora. Since multilingual cultural data are often expensive to collect, existing efforts handle this by prompt engineering or culture-specific pre-training. However, they might overlook the knowledge deficiency of low-resource culture and require extensive computing resources. In this paper, we propose CultureLLM, a cost-effective solution to incorporate cultural differences into LLMs. CultureLLM adopts World Value Survey (WVS) as seed data and generates semantically equivalent training data via the proposed semantic data augmentation. Using only 50 seed samples from WVS with augmented data, we fine-tune culture-specific LLMs and one unified model (CultureLLM-One) for 9 cultures covering rich and low-resource languages. Extensive experiments on 60 culture-related datasets demonstrate that CultureLLM significantly outperforms various counterparts such as GPT-3.5 (by 8.1%) and Gemini Pro (by 9.5%) with comparable performance to GPT-4 or even better. Our human study shows that the generated samples are semantically equivalent to the original samples, providing an effective solution for LLMs augmentation. Code is released at https://github.com/Scarelette/CultureLLM."
                    },
                    {
                        "title": "CulturePark: Boosting Cross-cultural Understanding in Large Language Models",
                        "abstract": "Cultural bias is pervasive in many large language models (LLMs), largely due to the deficiency of data representative of different cultures. Typically, cultural datasets and benchmarks are constructed either by extracting subsets of existing datasets or by aggregating from platforms such as Wikipedia and social media. However, these approaches are highly dependent on real-world data and human annotations, making them costly and difficult to scale. Inspired by cognitive theories on social communication, this paper introduces CulturePark, an LLM-powered multi-agent communication framework for cultural data collection. CulturePark simulates cross-cultural human communication with LLM-based agents playing roles in different cultures. It generates high-quality cross-cultural dialogues encapsulating human beliefs, norms, and customs. Using CulturePark, we generated 41,000 cultural samples to fine-tune eight culture-specific LLMs. We evaluated these models across three downstream tasks: content moderation, cultural alignment, and cultural education. Results show that for content moderation, our GPT-3.5-based models either match or outperform GPT-4 on datasets. Regarding cultural alignment, our models surpass GPT-4 on Hofstede's VSM 13 framework. Furthermore, for cultural education of human participants, our models demonstrate superior outcomes in both learning efficacy and user experience compared to GPT-4. CulturePark proves an important step in addressing cultural bias and advancing the democratization of AI, highlighting the critical role of culturally inclusive data in model training. Code is released at https://github.com/Scarelette/CulturePark."
                    },
                    {
                        "title": "A General Framework for Learning from Weak Supervision",
                        "abstract": "Weakly supervised learning generally faces challenges in applicability to various scenarios with diverse weak supervision and in scalability due to the complexity of existing algorithms, thereby hindering the practical deployment. This paper introduces a general framework for learning from weak supervision (GLWS) with a novel algorithm. Central to GLWS is an Expectation-Maximization (EM) formulation, adeptly accommodating various weak supervision sources, including instance partial labels, aggregate statistics, pairwise observations, and unlabeled data. We further present an advanced algorithm that significantly simplifies the EM computational demands using a Non-deterministic Finite Automaton (NFA) along with a forward-backward algorithm, which effectively reduces time complexity from quadratic or factorial often required in existing solutions to linear scale. The problem of learning from arbitrary weak supervision is therefore converted to the NFA modeling of them. GLWS not only enhances the scalability of machine learning models but also demonstrates superior performance and versatility across 11 weak supervision scenarios. We hope our work paves the way for further advancements and practical deployment in this field."
                    },
                    {
                        "title": "ERBench: An Entity-Relationship based Automatically Verifiable Hallucination Benchmark for Large Language Models",
                        "abstract": "Large language models (LLMs) have achieved unprecedented performances in various applications, yet evaluating them is still challenging. Existing benchmarks are either manually constructed or are automatic, but lack the ability to evaluate the thought process of LLMs with arbitrary complexity. We contend that utilizing existing relational databases based on the entity-relationship (ER) model is a promising approach for constructing benchmarks as they contain structured knowledge that can be used to question LLMs. Unlike knowledge graphs, which are also used to evaluate LLMs, relational databases have integrity constraints that can be used to better construct complex in-depth questions and verify answers: (1) functional dependencies can be used to pinpoint critical keywords that an LLM must know to properly answer a given question containing certain attribute values; and (2) foreign key constraints can be used to join relations and construct multi-hop questions, which can be arbitrarily long and used to debug intermediate answers. We thus propose ERBench, which uses these integrity constraints to convert any database into an LLM benchmark. ERBench supports continuous evaluation as databases change, multimodal questions, and various prompt engineering techniques. In our experiments, we construct LLM benchmarks using databases of multiple domains and make an extensive comparison of contemporary LLMs. We show how ERBench can properly evaluate any LLM by not only checking for answer correctness, but also effectively verifying the rationales by looking for the right keywords."
                    },
                    {
                        "title": "SpecFormer: Guarding Vision Transformer Robustness via Maximum Singular Value Penalization",
                        "abstract": "Vision Transformers (ViTs) are increasingly used in computer vision due to their high performance, but their vulnerability to adversarial attacks is a concern. Existing methods lack a solid theoretical basis, focusing mainly on empirical training adjustments. This study introduces SpecFormer, tailored to fortify ViTs against adversarial attacks, with theoretical underpinnings. We establish local Lipschitz bounds for the self-attention layer and propose the Maximum Singular Value Penalization (MSVP) to precisely manage these bounds By incorporating MSVP into ViTs' attention layers, we enhance the model's robustness without compromising training efficiency. SpecFormer, the resulting model, outperforms other state-of-the-art models in defending against adversarial attacks, as proven by experiments on CIFAR and ImageNet datasets. Code is released at https://github.com/microsoft/robustlearn."
                    },
                    {
                        "title": "Learning with Noisy Foundation Models",
                        "abstract": "Foundation models are usually pre-trained on large-scale datasets and then adapted to downstream tasks through tuning. However, the large-scale pre-training datasets, often inaccessible or too expensive to handle, can contain label noise that may adversely affect the generalization of the model and pose unexpected risks. This paper stands out as the first work to comprehensively understand and analyze the nature of noise in pre-training datasets and then effectively mitigate its impacts on downstream tasks. Specifically, through extensive experiments of fully-supervised and image-text contrastive pre-training on synthetic noisy ImageNet-1K, YFCC15M, and CC12M datasets, we demonstrate that, while slight noise in pre-training can benefit in-domain (ID) performance, where the training and testing data share a similar distribution, it always deteriorates out-of-domain (OOD) performance, where training and testing distributions are significantly different. These observations are agnostic to scales of pre-training datasets, pre-training noise types, model architectures, pre-training objectives, downstream tuning methods, and downstream applications. We empirically ascertain that the reason behind this is that the pre-training noise shapes the feature space differently. We then propose a tuning method (NMTune) to affine the feature space to mitigate the malignant effect of noise and improve generalization, which is applicable in both parameter-efficient and black-box tuning manners. We additionally conduct extensive experiments on popular vision and language models, including APIs, which are supervised and self-supervised pre-trained on realistic noisy data for evaluation. Our analysis and results demonstrate the importance of this novel and fundamental research direction, which we term as Noisy Model Learning."
                    },
                    {
                        "title": "On Catastrophic Inheritance of Large Foundation Models",
                        "abstract": "Large foundation models (LFMs) are claiming incredible performances. Yet great concerns have been raised about their mythic and uninterpreted potentials not only in machine learning, but also in various other disciplines. In this position paper, we propose to identify a neglected issue deeply rooted in LFMs: Catastrophic Inheritance, describing the weaknesses and limitations inherited from biased large-scale pre-training data to behaviors of LFMs on the downstream tasks, including samples that are corrupted, long-tailed, noisy, out-of-distributed, to name a few. Such inheritance can potentially cause catastrophes to downstream applications, such as bias, lack of generalization, deteriorated performance, security vulnerability, privacy leakage, and value misalignment. We discuss the challenges behind this issue and propose UIM, a framework to Understand the catastrophic inheritance of LFMs from both pre-training and downstream adaptation, Interpret the implications of catastrophic inheritance on downstream tasks, and how to Mitigate it. UIM aims to unite both the machine learning and social sciences communities for more responsible and promising AI development and deployment."
                    },
                    {
                        "title": "Can I understand what I create? Self-Knowledge Evaluation of Large Language Models",
                        "abstract": "Large language models (LLMs) have achieved remarkable progress in linguistic tasks, necessitating robust evaluation frameworks to understand their capabilities and limitations. Inspired by Feynman's principle of understanding through creation, we introduce a self-knowledge evaluation framework that is easy to implement, evaluating models on their ability to comprehend and respond to self-generated questions. Our findings, based on testing multiple models across diverse tasks, reveal significant gaps in the model's self-knowledge ability. Further analysis indicates these gaps may be due to misalignment with human attention mechanisms. Additionally, fine-tuning on self-generated math task may enhance the model's math performance, highlighting the potential of the framework for efficient and insightful model evaluation and may also contribute to the improvement of LLMs."
                    },
                    {
                        "title": "Survey on Factuality in Large Language Models: Knowledge, Retrieval and Domain-Specificity",
                        "abstract": "This survey addresses the crucial issue of factuality in Large Language Models (LLMs). As LLMs find applications across diverse domains, the reliability and accuracy of their outputs become vital. We define the Factuality Issue as the probability of LLMs to produce content inconsistent with established facts. We first delve into the implications of these inaccuracies, highlighting the potential consequences and challenges posed by factual errors in LLM outputs. Subsequently, we analyze the mechanisms through which LLMs store and process facts, seeking the primary causes of factual errors. Our discussion then transitions to methodologies for evaluating LLM factuality, emphasizing key metrics, benchmarks, and studies. We further explore strategies for enhancing LLM factuality, including approaches tailored for specific domains. We focus two primary LLM configurations standalone LLMs and Retrieval-Augmented LLMs that utilizes external data, we detail their unique challenges and potential enhancements. Our survey offers a structured guide for researchers aiming to fortify the factual reliability of LLMs."
                    },
                    {
                        "title": "Meta Semantic Template for Evaluation of Large Language Models",
                        "abstract": "Do large language models (LLMs) genuinely understand the semantics of the language, or just memorize the training data? The recent concern on potential data contamination of LLMs has raised awareness of the community to conduct research on LLMs evaluation. In this paper, we propose MSTemp, an approach that creates meta semantic templates to evaluate the semantic understanding ability of LLMs. The core of MSTemp is not to perform evaluation directly on existing benchmark datasets, but to generate new out-of-distribution (OOD) evaluation sets using existing datasets as seeds. Specifically, for a given sentence, MSTemp leverages another language model to generate new samples while preserving its semantics. The new samples are called semantic templates to the original sentence. Then, MSTemp generates evaluation samples via sentence parsing and random word replacement on the semantic templates. MSTemp is highly flexible, dynamic, and cost-effective. Our initial experiments show that MSTemp-generated samples can significantly reduce the performance of LLMs using existing datasets as seeds. We hope this initial work can shed light on future research of LLMs evaluation."
                    },
                    {
                        "title": "CompeteAI: Understanding the Competition Behaviors in Large Language Model-based Agents",
                        "abstract": "Large language models (LLMs) have been widely used as agents to complete different tasks, such as personal assistance or event planning. While most work has focused on cooperation and collaboration between agents, little work explores competition , another important mechanism that fosters the development of society and economy. In this paper, we seek to examine the competition behaviors in LLM-based agents. We first propose a general framework to study the competition between agents. Then, we implement a practical competitive environment using GPT-4 to simulate a virtual town with two types of agents, including restaurant agents and customer agents. Specifically, the restaurant agents compete with each other to attract more customers, where the competition fosters them to transform, such as cultivating new operating strategies. The results of our experiments reveal several interesting findings ranging from social learning to Matthew Effect, which aligns well with existing sociological and economic theories. We believe that competition between agents deserves further investigation to help us understand society better. The code will be released soon."
                    },
                    {
                        "title": "Supervised Knowledge Makes Large Language Models Better In-context Learners",
                        "abstract": "Large Language Models (LLMs) exhibit emerging in-context learning abilities through prompt engineering. The recent progress in large-scale generative models has further expanded their use in real-world language applications. However, the critical challenge of improving the generalizability and factuality of LLMs in natural language understanding and question answering remains under-explored. While previous in-context learning research has focused on enhancing models to adhere to users' specific instructions and quality expectations, and to avoid undesired outputs, little to no work has explored the use of task-Specific fine-tuned Language Models (SLMs) to improve LLMs' in-context learning during the inference stage. Our primary contribution is the establishment of a simple yet effective framework that enhances the reliability of LLMs as it: 1) generalizes out-of-distribution data, 2) elucidates how LLMs benefit from discriminative models, and 3) minimizes hallucinations in generative tasks. Using our proposed plug-in method, enhanced versions of Llama 2 and ChatGPT surpass their original versions regarding generalizability and factuality. We offer a comprehensive suite of resources, including 16 curated datasets, prompts, model checkpoints, and LLM outputs across 9 distinct tasks. The code and data are released at: https://github.com/YangLinyi/Supervised-Knowledge-Makes-Large-Language-Models-Better-In-context-Learners. Our empirical analysis sheds light on the advantages of incorporating discriminative models into LLMs and highlights the potential of our methodology in fostering more reliable LLMs."
                    },
                    {
                        "title": "PromptBench: A Unified Library for Evaluation of Large Language Models",
                        "abstract": "The evaluation of large language models (LLMs) is crucial to assess their performance and mitigate potential security risks. In this paper, we introduce PromptBench, a unified library to evaluate LLMs. It consists of several key components that are easily used and extended by researchers: prompt construction, prompt engineering, dataset and model loading, adversarial prompt attack, dynamic evaluation protocols, and analysis tools. PromptBench is designed to be an open, general, and flexible codebase for research purposes that can facilitate original study in creating new benchmarks, deploying downstream applications, and designing new evaluation protocols. The code is available at: https://github.com/microsoft/promptbench and will be continuously supported."
                    },
                    {
                        "title": "EmotionPrompt: Leveraging Psychology for Large Language Models Enhancement via Emotional Stimulus",
                        "abstract": "Large language models (LLMs) have achieved signi\ufb01cant performance in many \ufb01elds, such as reasoning, language understanding, and math problem-solving, and are regarded as an important step to arti\ufb01cial general intelligence (AGI). However, the sensitivity of LLMs to prompts remain a major bottleneck for their daily adoption. In this paper, we take inspiration from psychology and propose EmotionPrompt to explore emotional intelligence to enhance the performance of LLMs. Our EmotionPrompt operates on a remarkably straightforward principle: the incorporation of emotional stimulus into prompts. Experimental re-sults demonstrate that our EmotionPrompt, using the same single prompt templates, signi\ufb01cantly out-performs the original prompt and Zero-shot-CoT in both zero-shot and few-shot settings on eight tasks with diverse models: ChatGPT, Vicuna-13b, Bloom, and Flan-T5-large. Furthermore, Emotion-Prompt was observed to improve both the truthfulness and informativeness. We believe that Emo-tionPrompt heralds a novel avenue for exploring interdisciplinary knowledge for interaction between humans and LLMs."
                    },
                    {
                        "title": "The Good, The Bad, and Why: Unveiling Emotions in Generative AI",
                        "abstract": "Emotion significantly impacts our daily behaviors and interactions. While recent generative AI models, such as large language models, have shown impressive performance in various tasks, it remains unclear whether they truly comprehend emotions. This paper aims to address this gap by incorporating psychological theories to gain a holistic understanding of emotions in generative AI models. Specifically, we propose three approaches: 1) EmotionPrompt to enhance AI model performance, 2) EmotionAttack to impair AI model performance, and 3) EmotionDecode to explain the effects of emotional stimuli, both benign and malignant. Through extensive experiments involving language and multi-modal models on semantic understanding, logical reasoning, and generation tasks, we demonstrate that both textual and visual EmotionPrompt can boost the performance of AI models while EmotionAttack can hinder it. Additionally, EmotionDecode reveals that AI models can comprehend emotional stimuli akin to the mechanism of dopamine in the human brain. Our work heralds a novel avenue for exploring psychology to enhance our understanding of generative AI models."
                    },
                    {
                        "title": "CompeteAI: Understanding the Competition Dynamics of Large Language Model-based Agents",
                        "abstract": "Large language models (LLMs) have been widely used as agents to complete different tasks, such as personal assistance or event planning. While most of the work has focused on cooperation and collaboration between agents, little work explores competition, another important mechanism that promotes the development of society and economy. In this paper, we seek to examine the competition dynamics in LLM-based agents. We first propose a general framework for studying the competition between agents. Then, we implement a practical competitive environment using GPT-4 to simulate a virtual town with two types of agents, restaurant agents and customer agents. Specifically, the restaurant agents compete with each other to attract more customers, where competition encourages them to transform, such as cultivating new operating strategies. Simulation experiments reveal several interesting findings at the micro and macro levels, which align well with existing market and sociological theories. We hope that the framework and environment can be a promising testbed to study competition that fosters understanding of society. Code is available at: https://github.com/microsoft/competeai."
                    }
                ]
            }
        }
    },
    "2409.18269": {
        "paper_data": {
            "title": "Intrinsic Robustness of Prophet Inequality to Strategic Reward Signaling",
            "url": "http://arxiv.org/abs/2409.18269v1",
            "arxiv_id": "2409.18269",
            "authors": [
                "Wei Tang",
                "Haifeng Xu",
                "Ruimin Zhang",
                "Derek Zhu"
            ],
            "abstract": "Prophet inequality concerns a basic optimal stopping problem and states that simple threshold stopping policies -- i.e., accepting the first reward larger than a certain threshold -- can achieve tight $\\frac{1}{2}$-approximation to the optimal prophet value. Motivated by its economic applications, this paper studies the robustness of this approximation to natural strategic manipulations in which each random reward is associated with a self-interested player who may selectively reveal his realized reward to the searcher in order to maximize his probability of being selected.   We say a threshold policy is $\\alpha$(-strategically)-robust if it (a) achieves the $\\alpha$-approximation to the prophet value for strategic players; and (b) meanwhile remains a $\\frac{1}{2}$-approximation in the standard non-strategic setting. Starting with a characterization of each player's optimal information revealing strategy, we demonstrate the intrinsic robustness of prophet inequalities to strategic reward signaling through the following results: (1) for arbitrary reward distributions, there is a threshold policy that is $\\frac{1-\\frac{1}{e}}{2}$-robust, and this ratio is tight; (2) for i.i.d. reward distributions, there is a threshold policy that is $\\frac{1}{2}$-robust, which is tight for the setting; and (3) for log-concave (but non-identical) reward distributions, the $\\frac{1}{2}$-robustness can also be achieved under certain regularity assumptions.",
            "introduction": " Introduction The prophet inequality of Krengel and Sucheston [1977] is a foundational framework for the theory of optimal stopping problems and sequential decision-maki ng. In the classic prophet inequality, a searcher faces a \ufb01nite sequence of non-negative, real-val ued and independent random variables X1,...,X Nwith known distributions Hifrom which a reward of value Xiis drawn sequentially for i= 1,...,N. Once a random reward is realized, the searcher decides whet her to accept the realized reward and stop searching, or reject the reward and proceed t o the next reward. The searcher\u2019s objective is to maximize the value of the accepted reward. Th e performance of the searcher\u2019s stopping policy is evaluated against a prophet value which equals to the ex-post maximum realized reward. The classic and elegant result of Samuel-Cahn [1984] showed that a simple static threshold policy achieves at least half of the prophet value, and surpr isingly, this bound is the best possible even among dynamic policies. Samuel-Cahn\u2019s policy uses the threshold that is the median of \u2217Chinese University of Hong Kong. Email: weitang@cuhk.edu.hk . \u2020University of Chicago. Email: haifengxu@uchicago.edu . \u2021University of Chicago. Email: ruimin@uchicago.edu . \u00a7University of Chicago. Email: dzhu8@uchicago.edu . 1the distribution of the highest prizes, and then accepts the \ufb01rst realized reward that exceeds this threshold. The existence of this 1/2-approximation is now known as the prophet inequality . Recently, there is a regained interest of the prophet inequa lity due to its beautiful connection to online mechanism design (see, e.g., Hajiaghayi et al. [2007],Chawla et al. [2010]), and many di\ufb00erent settings of the prophet inequality have been studie d (see the survey by Lucier[2017], Correa et al. [2019]). For example, Kleinberg and Weinberg [2012] extend the prophetinequality to all matroid constraints and similar to Samuel-Cahn [1984], they also show that a threshold stopping policy withthethresholdequal tohalfoftheexpected maxim umrewardcanalso leadtotheoptimal 1/2-approximation. Indeed, itisnowwell-knownthatthereexi stsarangeofthresholdsfortheclassic prophet inequality that can achieve the optimal 1/2-approximation (see De\ufb01nition 2.2). An important assumption in the classic prophet inequality i s that the distribution of each random variable is an inanimate object and, once the searcher reach es it, it will fully disclose its realized reward to the searcher. Yet this may not be the case in many rea l-world applications where each distribution may often be associated with a strategic player1who may have incentives to selectively disclose information to maximize his own probability of bei ng chosen by the searcher. This is usually the case when information is not controlled by natur e but by humans or algorithms. Such examples are ample in economic activities. For instance, wh en a recruiter searches for the best candidate for a position by sequentially interviewing a set of job applicants, each job applicant is naturally a strategic player and would want to control how much information they disclose about their characteristics (including strengths, weakne sses, personality, and experience) to the recruiter (the searcher) so as to maximize the probability o f being hired. Similarly, when a venture capitalist searches for the most promising startup to inves t by sequentially visiting each startup, the startups are strategic players. They have more accurate information about their own potential and can control how much of this information they disclose to attract the investment from the venture capitalist (the searcher). Finally, in the real est ate market,",
            "references": [
                {
                    "title": "Multi-Sender Persuasion: A Computational Perspective",
                    "abstract": "We consider the multi-sender persuasion problem: multiple players with informational advantage signal to convince a single self-interested actor to take certain actions. This problem generalizes the seminal Bayesian Persuasion framework and is ubiquitous in computational economics, multi-agent learning, and multi-objective machine learning. The core solution concept here is the Nash equilibrium of senders' signaling policies. Theoretically, we prove that finding an equilibrium in general is PPAD-Hard; in fact, even computing a sender's best response is NP-Hard. Given these intrinsic difficulties, we turn to finding local Nash equilibria. We propose a novel differentiable neural network to approximate this game's non-linear and discontinuous utilities. Complementing this with the extra-gradient algorithm, we discover local equilibria that Pareto dominates full-revelation equilibria and those found by existing neural networks. Broadly, our theoretical and empirical contributions are of interest to a large class of economic problems."
                },
                {
                    "title": "Optimal Rates and Efficient Algorithms for Online Bayesian Persuasion",
                    "abstract": "Bayesian persuasion studies how an informed sender should influence beliefs of rational receivers who take decisions through Bayesian updating of a common prior. We focus on the online Bayesian persuasion framework, in which the sender repeatedly faces one or more receivers with unknown and adversarially selected types. First, we show how to obtain a tight $\\tilde O(T^{1/2})$ regret bound in the case in which the sender faces a single receiver and has partial feedback, improving over the best previously known bound of $\\tilde O(T^{4/5})$. Then, we provide the first no-regret guarantees for the multi-receiver setting under partial feedback. Finally, we show how to design no-regret algorithms with polynomial per-iteration running time by exploiting type reporting, thereby circumventing known intractability results on online Bayesian persuasion. We provide efficient algorithms guaranteeing a $O(T^{1/2})$ regret upper bound both in the single- and multi-receiver scenario when type reporting is allowed."
                },
                {
                    "title": "The IID Prophet Inequality with Limited Flexibility",
                    "abstract": "In online sales, sellers usually offer each potential buyer a posted price in a take-it-or-leave fashion. Buyers can sometimes see posted prices faced by other buyers, and changing the price frequently could be considered unfair. The literature on posted price mechanisms and prophet inequality problems has studied the two extremes of pricing policies, the fixed price policy and fully dynamic pricing. The former is suboptimal in revenue but is perceived as fairer than the latter. This work examines the middle situation, where there are at most $k$ distinct prices over the selling horizon. Using the framework of prophet inequalities with independent and identically distributed random variables, we propose a new prophet inequality for strategies that use at most $k$ thresholds. We present asymptotic results in $k$ and results for small values of $k$. For $k=2$ prices, we show an improvement of at least $11\\%$ over the best fixed-price solution. Moreover, $k=5$ prices suffice to guarantee almost $99\\%$ of the approximation factor obtained by a fully dynamic policy that uses an arbitrary number of prices. From a technical standpoint, we use an infinite-dimensional linear program in our analysis; this formulation could be of independent interest to other online selection problems."
                },
                {
                    "title": "Individual Fairness in Prophet Inequalities",
                    "abstract": "Prophet inequalities are performance guarantees for online algorithms (a.k.a. stopping rules) solving the following ''hiring problem'': a decision maker sequentially inspects candidates whose values are independent random numbers and is asked to hire at most one candidate by selecting it before inspecting the values of future candidates in the sequence. A classic result in optimal stopping theory asserts that there exist stopping rules guaranteeing that the decision maker will hire a candidate whose expected value is at least half as good as the expected value of the candidate hired by a ''prophet,'' i.e.one who has simultaneous access to the realizations of all candidates' values. Such stopping rules may indeed have provably good performance but might treat individual candidates unfairly in a number of different ways. In this work we identify two types of individual fairness that might be desirable in optimal stopping problems. We call them identity-independent fairness (IIF) and time-independent fairness (TIF) and give precise definitions in the context of the hiring problem. We give polynomial-time algorithms for finding the optimal IIF/TIF stopping rules for a given instance with discrete support and we manage to recover a prophet inequality with factor 1/2 when the decision maker's stopping rule is required to satisfy both fairness properties while the prophet is unconstrained. We also explore worst-case ratios between optimal selection rules in the presence vs. absence of individual fairness constraints, in both the online and offline settings. We prove an impossibility result showing that there is no prophet inequality with a non-zero factor for either IIF or TIF stopping rules when we further constrain the decision maker to make a hire with probability 1. We finally consider a setting in which the decision maker doesn't know the distributions of candidates' values but has access to a bounded number of independent samples from each distribution. We provide constant-competitive algorithms that satisfy both TIF and IIF, using one sample from each distribution in the offline setting and two samples from each distribution in the online setting. The full version of the paper: https://arxiv.org/abs/2205.10302v1"
                },
                {
                    "title": "Sequential Information Design: Markov Persuasion Process and Its Efficient Reinforcement Learning",
                    "abstract": "In today's economy, it becomes important for Internet platforms to consider the sequential information design problem to align its long term interest with incentives of the gig service providers (e.g., drivers, hosts). This paper proposes a novel model of sequential information design, namely the Markov persuasion processes (MPPs), in which a sender, with informational advantage, seeks to persuade a stream of myopic receivers to take actions that maximize the sender's cumulative utilities in a finite horizon Markovian environment with varying prior and utility functions. Planning in MPPs thus faces the unique challenge in finding a signaling policy that is simultaneously persuasive to the myopic receivers and inducing the optimal long-term cumulative utilities of the sender. Nevertheless, in the population level where the model is known, it turns out that we can efficiently determine the optimal (resp. \u03b5-optimal) policy with finite (resp. infinite) states and outcomes, through a modified formulation of the Bellman equation that additionally takes persuasiveness into consideration. Our main technical contribution is to study the MPP under the online reinforcement learning (RL) setting, where the goal is to learn the optimal signaling policy by interacting with with the underlying MPP, without the knowledge of the sender's utility functions, prior distributions, and the Markov transition kernels. For such a problem, we design a provably efficient no-regret learning algorithm, the Optimism-Pessimism Principle for Persuasion Process (OP4), which features a novel combination of both optimism and pessimism principles. In particular, we obtain optimistic estimates of the value functions to encourage exploration under the unknown environment, and additionally robustify the signaling policy with respect to the uncertainty of prior estimation to prevent receiver's detrimental equilibrium behavior. Our algorithm enjoys sample efficiency by achieving a sublinear \u221aT-regret upper bound. Furthermore, both our algorithm and theory can be applied to MPPs with large space of outcomes and states via function approximation, and we showcase such a success under the linear setting."
                },
                {
                    "title": "Online Bayesian Recommendation with No Regret",
                    "abstract": "We introduce and study the online Bayesian recommendation problem for a platform, who can observe a utility-relevant state of a product, repeatedly interacting with a population of myopic users through an online recommendation mechanism. This paradigm is common in a wide range of scenarios in the current Internet economy. For each user with her own private preference and belief, the platform commits to a recommendation strategy to utilize his information advantage on the product state to persuade the self-interested user to follow the recommendation. The platform does not know user's preferences and beliefs, and has to use an adaptive recommendation strategy to persuade with gradually learning user's preferences and beliefs in the process. We aim to design online learning policies with no Stackelberg regret for the platform, i.e., against the optimum policy in hindsight under the assumption that users will correspondingly adapt their behaviors to the benchmark policy. Our first result is an online policy that achieves double logarithm regret dependence on the number of rounds. We then present a hardness result showing that no adaptive online policy can achieve regret with better dependency on the number of rounds. Finally, by formulating the platform's problem as optimizing a linear program with membership oracle access, we present our second online policy that achieves regret with polynomial dependence on the number of states but logarithm dependence on the number of rounds."
                },
                {
                    "title": "Who Leads and Who Follows in Strategic Classification?",
                    "abstract": "As predictive models are deployed into the real world, they must increasingly contend with strategic behavior. A growing body of work on strategic classification treats this problem as a Stackelberg game: the decision-maker\"leads\"in the game by deploying a model, and the strategic agents\"follow\"by playing their best response to the deployed model. Importantly, in this framing, the burden of learning is placed solely on the decision-maker, while the agents' best responses are implicitly treated as instantaneous. In this work, we argue that the order of play in strategic classification is fundamentally determined by the relative frequencies at which the decision-maker and the agents adapt to each other's actions. In particular, by generalizing the standard model to allow both players to learn over time, we show that a decision-maker that makes updates faster than the agents can reverse the order of play, meaning that the agents lead and the decision-maker follows. We observe in standard learning settings that such a role reversal can be desirable for both the decision-maker and the strategic agents. Finally, we show that a decision-maker with the freedom to choose their update frequency can induce learning dynamics that converge to Stackelberg equilibria with either order of play."
                },
                {
                    "title": "Incentive-Aware PAC Learning",
                    "abstract": "We study PAC learning in the presence of strategic manipulation, where data points may modify their features in certain predefined ways in order to receive a better outcome. We show that the vanilla ERM principle fails to achieve any nontrivial guarantee in this context. Instead, we propose an incentive-aware version of the ERM principle which has asymptotically optimal sample complexity. We then focus our attention on incentive-compatible classifiers, which provably prevent any kind of strategic manipulation. We give a sample complexity bound that is, curiously, independent of the hypothesis class, for the ERM principle restricted to incentive-compatible classifiers. This suggests that incentive compatibility alone can act as an effective means of regularization. We further show that it is without loss of generality to consider only incentive-compatible classifiers when opportunities for strategic manipulation satisfy a transitivity condition. As a consequence, in such cases, our hypothesis-class-independent sample complexity bound applies even without incentive compatibility. Our results set the foundations of incentive-aware PAC learning."
                },
                {
                    "title": "Competitive Information Design for Pandora's Box",
                    "abstract": "We study a natural competitive-information-design variant for the Pandora's Box problem (Weitzman 1979), where each box is associated with a strategic information sender who can design what information about the box's prize value to be revealed to the agent when the agent inspects the box. This variant with strategic boxes is motivated by a wide range of real-world economic applications for Pandora's Box. The main contributions of this article are two-fold: (1) we study informational properties of Pandora's Box by showing an intrinsic connection between informativeness of any box's value distribution and the utility order of the search agent; and (2) we fully characterize the pure symmetric equilibrium for the boxes' competitive information revelation, which reveals various insights regarding information competition and the resultant agent utility at equilibrium."
                },
                {
                    "title": "Learning to Persuade on the Fly: Robustness Against Ignorance",
                    "abstract": "We study a repeated persuasion setting between a sender and a receiver, where at each time t, the sender shares information about a payoff-relevant state with the receiver. The state at each time t is drawn independently and identically from an unknown distribution, and subsequent to receiving information about it, the receiver (myopically) chooses an action from a finite set. The sender seeks to persuade the receiver into choosing actions that are aligned with her preference by selectively sharing information about the state. In contrast to the standard persuasion setting, we focus on the case where neither the sender nor the receiver knows the distribution of the payoff relevant state. Instead, the sender learns this distribution over time by observing the state realizations. We adopt the assumption common in the literature on Bayesian persuasion that at each time period, prior to observing the realized state in that period, the sender commits to a signaling mechanism that maps each state to a possibly random action recommendation. Subsequent to the state observation, the sender recommends an action as per the chosen signaling mechanism."
                },
                {
                    "title": "PAC-Learning for Strategic Classification",
                    "abstract": "Machine learning (ML) algorithms may be susceptible to being gamed by individuals with knowledge of the algorithm (a.k.a. Goodhart's law). Such concerns have motivated a surge of recent work on strategic classification where each data point is a self-interested agent and may strategically manipulate his features to induce a more desirable classification outcome for himself. Previous works assume agents have homogeneous preferences and all equally prefer the positive label. This paper generalizes strategic classification to settings where different data points may have different preferences over the classification outcomes. Besides a richer model, this generalization allows us to include evasion attacks in adversarial ML also as a special case of our model where positive [resp. negative] data points prefer the negative [resp. positive] label, and thus for the first time allows strategic and adversarial learning to be studied under the same framework. We introduce the strategic VC-dimension (SVC), which captures the PAC-learnability of a hypothesis class in our general strategic setup. SVC generalizes the notion of adversarial VC-dimension (AVC) introduced recently by Cullina et al. arXiv:1806.01471. We then instantiate our framework for arguably the most basic hypothesis class, i.e., linear classifiers. We fully characterize the statistical learnability of linear classifiers by pinning down its SVC and the computational tractability by pinning down the complexity of the empirical risk minimization problem. Our bound of SVC for linear classifiers also strictly generalizes the AVC bound for linear classifiers in arXiv:1806.01471. Finally, we briefly study the power of randomization in our strategic classification setup. We show that randomization may strictly increase the accuracy in general, but will not help in the special case of adversarial classification under evasion attacks."
                },
                {
                    "title": "Static pricing for multi-unit prophet inequalities",
                    "abstract": "Characterizing the Efficiency of Static Pricing Schemes as a Function of the Supply The problem of selling a supply of k units to a stream of customers constitutes one of the cornerstones in revenue management. Static pricing schemes (that output the same price to all customers) are commonly used because of their simplicity and their many desirable properties; they are anonymous, nonadaptive, and order oblivious. Although the efficiency of those schemes should improve as the supply k increases, prior work has only focused either on algorithms that aim for a constant approximation that is independent of k or on the setting where k becomes really large. In contrast, this paper characterizes the efficiency of static pricing schemes as a function of the supply. Our approach stems from identifying a \u201csweet spot\u201d between selling enough items and obtaining enough utility from customers with high valuations. Subsequent work shows that our pricing scheme is the optimal static pricing for every value of k."
                },
                {
                    "title": "Prophet Inequalities for Bayesian Persuasion",
                    "abstract": "We study an information-structure design problem (i.e., a Bayesian persuasion problem) in an online scenario. Inspired by the classic gambler\u2019s problem, consider a set of candidates who arrive sequentially and are evaluated by one agent (the sender). This agent learns the value from hiring the candidate to herself as well as the value to another agent, the receiver. The sender provides a signal to the receiver who, in turn, makes an irrevocable decision on whether or not to hire the candidate. A-priori, for each agent the distribution of valuation is independent across candidates but may not be identical. We design good online signaling schemes for the sender. To assess the performance, we compare the expected utility to that of an optimal offline scheme by a prophet sender who knows all candidate realizations in advance. We show an optimal prophet inequality for online Bayesian persuasion, with a 1/2-approximation when the instance satisfies a \u201csatisfactory-statusquo\u201d assumption. Without this assumption there are instances without any finite approximation factor. We extend the results to combinatorial domains and obtain prophet inequalities for matching with multiple hires and multiple receivers."
                },
                {
                    "title": "Recent developments in prophet inequalities",
                    "abstract": "The classic prophet inequality states that, when faced with a finite sequence of non-negative independent random variables, a gambler who knows their distribution and is allowed to stop the sequence at any time, can obtain, in expectation, at least half as much reward as a prophet who knows the values of each random variable and can choose the largest one. Following this classic theorem from the 70s, many results have been obtained for several related optimal stopping problems. Moreover, the recently uncovered connection between prophet inequalities and posted price mechanisms, has given the area a new surge. We survey some new developments and highlight some compelling open problems."
                },
                {
                    "title": "Competitive Information Disclosure by Multiple Senders",
                    "abstract": "Abstract We analyze a model of competition in Bayesian persuasion in which multiple symmetric senders vie for the patronage of a receiver by disclosing information about their respective proposal qualities. We show that a symmetric equilibrium exists and is unique. We then show that as the number of senders increases, each sender discloses information more aggressively, and full disclosure by each sender arises in the limit of infinitely many senders."
                },
                {
                    "title": "An economic view of prophet inequalities",
                    "abstract": "Over the past decade, an exciting connection has developed between the theory of posted-price mechanisms and the prophet inequality, a result from the theory of optimal stopping. This survey provides an overview of this literature, covering extensions and applications of the prophet inequality through the lens of an economic proof of this classic result. We focus on highlighting ways in which the economic perspective drives new advances in the theory of online stochastic optimization, and vice versa."
                },
                {
                    "title": "Posted Price Mechanisms for a Random Stream of Customers",
                    "abstract": "Posted price mechanisms constitute a widely used way of selling items to strategic consumers. Although suboptimal, the attractiveness of these mechanisms comes from their simplicity and easy implementation. In this paper, we investigate the performance of posted price mechanisms when customers arrive in an unknown random order. We compare the expected revenue of these mechanisms to the expected revenue of the optimal auction in two different settings. Namely, the nonadaptive setting in which all offers are sent to the customers beforehand, and the adaptive setting in which an offer is made when a consumer arrives. For the nonadaptive case, we obtain a strategy achieving an expected revenue within at least a 1-1/e fraction of that of the optimal auction. We also show that this bound is tight, even if the customers have i.i.d. valuations for the item. For the adaptive case, we exhibit a posted price mechanism that achieves a factor 0.745 of the optimal revenue, when the customers have i.i.d. valuations for the item. Furthermore, we prove that our results extend to the prophet inequality setting and in particular our result for i.i.d. random valuations resolves a problem posed by Hill and Kertz. [13]"
                },
                {
                    "title": "Beating 1-1/e for ordered prophets",
                    "abstract": "Hill and Kertz studied the prophet inequality on iid distributions [The Annals of Probability 1982]. They proved a theoretical bound of 1 - 1/e on the approximation factor of their algorithm. They conjectured that the best approximation factor for arbitrarily large n is 1/1+1/e \u2243 0.731. This conjecture remained open prior to this paper for over 30 years. In this paper we present a threshold-based algorithm for the prophet inequality with n iid distributions. Using a nontrivial and novel approach we show that our algorithm is a 0.738-approximation algorithm. By beating the bound of 1/1+1/e, this refutes the conjecture of Hill and Kertz. Moreover, we generalize our results to non-uniform distributions and discuss its applications in mechanism design."
                },
                {
                    "title": "Persuasion of a Privately Informed Receiver",
                    "abstract": "Kolotilin acknowledges financial support from the Australian Research Council. Zapechelnyuk acknowledges financial support from the Economic and Social Research Council (grant no. ES/N01829X/1)"
                },
                {
                    "title": "A Rothschild-Stiglitz Approach to Bayesian Persuasion",
                    "abstract": "Rothschild and Stiglitz (1970) represent random variables as convex functions (integrals of the cumulative distribution function). Combining this representation with Blackwell's Theorem (1953), we characterize distributions of posterior means that can be induced by a signal. This characterization provides a novel way to analyze a class of Bayesian persuasion problems."
                },
                {
                    "title": "Prophet Inequalities with Limited Information",
                    "abstract": "In the classical prophet inequality, a gambler observes a sequence of stochastic rewards V1, ..., Vn and must decide, for each reward Vi, whether to keep it and stop the game or to forfeit the reward forever and reveal the next value Vi. The gambler's goal is to obtain a constant fraction of the expected reward that the optimal offline algorithm would get. Recently, prophet inequalities have been generalized to settings where the gambler can choose k items, and, more generally, where he can choose any independent set in a matroid. However, all the existing algorithms require the gambler to know the distribution from which the rewards V1, ..., Vn are drawn. \n \nThe assumption that the gambler knows the distribution from which V1, ..., Vn are drawn is very strong. Instead, we work with the much simpler assumption that the gambler only knows a few samples from this distribution. We construct the first single-sample prophet inequalities for many settings of interest, whose guarantees all match the best possible asymptotically, even with full knowledge of the distribution. Specifically, we provide a novel single-sample algorithm when the gambler can choose any k elements whose analysis is based on random walks with limited correlation. In addition, we provide a black-box method for converting specific types of solutions to the related secretary problem to single-sample prophet inequalities, and apply it to several existing algorithms. Finally, we provide a constant-sample prophet inequality for constant-degree bipartite matchings. \n \nIn addition, we apply these results to design the first posted-price and multi-dimensional auction mechanisms with limited information in settings with asymmetric bidders. Connections between prophet inequalities and posted-price mechanisms are already known, but applying the existing framework requires knowledge of the underlying distributions, as well as the so-called \"virtual values\" even when the underlying prophet inequalities do not. We therefore provide an extension of this framework that bypasses virtual values altogether, allowing our mechanisms to take full advantage of the limited information required by our new prophet inequalities."
                },
                {
                    "title": "Bayes Correlated Equilibrium and the Comparison of Information Structures in Games",
                    "abstract": "A game of incomplete information can be decomposed into a basic game and an information structure. The basic game defines the set of actions, the set of payoff states the payoff functions and the common prior over the payoff states. The information structure refers to the signals that the players receive in the game. We characterize the set of outcomes that can arise in Bayes Nash equilibrium if players observe the given information structure but may also observe additional signals. The characterization corresponds to the set of (a version of) incomplete information correlated equilibria which we dub Bayes correlated equilibria. We identify a partial order on many player information structures (individual sufficiency) under which more information shrinks the set of Bayes correlated equilibria. This order captures the role of information in imposing (incentive) constraints on behavior."
                },
                {
                    "title": "Asymptotically optimal algorithm for stochastic adwords",
                    "abstract": "In this paper we consider the adwords problem in the unknown distribution model. We consider the case where the budget to bid ratio k is at least 2, and give improved competitive ratios. Earlier results had competitive ratios better than 1-1/e only for \"large enough\" k, while our competitive ratio increases continuously with k. For k=2 the competitive ratio we get is 0.729 and it is 0.9 for k=16. We also improve the asymptotic competitive ratio for large k from 1 - O(\u221alog n/k) to 1 - O(\u221a1/k), thus removing any dependence on n, the number of advertisers. This ratio is optimal, even with known distributions. That is, even if an algorithm is tailored to the distribution, it cannot get a competitive ratio of 1 - o(\u221a1/k), whereas our algorithm does not depend on the distribution. The algorithm is rather simple, it computes a score for every advertiser based on his original budget, the remaining budget and the remaining number of steps in the algorithm and assigns a query to the advertiser with the highest bid plus his score. The analysis is based on a \"hybrid argument\" that considers algorithms that are part actual, part hypothetical, to prove that our (actual) algorithm is better than a completely hypothetical algorithm whose performance is easy to analyze."
                },
                {
                    "title": "Matroid prophet inequalities",
                    "abstract": "Consider a gambler who observes a sequence of independent, non-negative random numbers and is allowed to stop the sequence at any time, claiming a reward equal to the most recent observation. The famous prophet inequality of Krengel, Sucheston, and Garling asserts that a gambler who knows the distribution of each random variable can achieve at least half as much reward, in expectation, as a \"prophet\" who knows the sampled values of each random variable and can choose the largest one. We generalize this result to the setting in which the gambler and the prophet are allowed to make more than one selection, subject to a matroid constraint. We show that the gambler can still achieve at least half as much reward as the prophet; this result is the best possible, since it is known that the ratio cannot be improved even in the original prophet inequality, which corresponds to the special case of rank-one matroids. Generalizing the result still further, we show that under an intersection of $p$ matroid constraints, the prophet's reward exceeds the gambler's by a factor of at most $O(p)$, and this factor is also tight.\n Beyond their interest as theorems about pure online algoritms or optimal stopping rules, these results also have applications to mechanism design. Our results imply improved bounds on the ability of sequential posted-price mechanisms to approximate optimal mechanisms in both single-parameter and multi-parameter Bayesian settings. In particular, our results imply the first efficiently computable constant-factor approximations to the Bayesian optimal revenue in certain multi-parameter settings."
                },
                {
                    "title": "Competition in Persuasion",
                    "abstract": "We study symmetric information games where a number of senders choose what information to communicate. We show that the impact of competition on information revelation is ambiguous in general. We identify a condition on the information environment (i.e., the set of signals available to each sender) that is necessary and sufficient for equilibrium outcomes to be no less informative than the collusive outcome, regardless of preferences. The same condition also provides an easy way to characterize the equilibrium set and governs whether introducing additional senders or decreasing the alignment of senders\u2019 preferences necessarily increases the amount of information revealed."
                },
                {
                    "title": "Bayesian Combinatorial Auctions: Expanding Single Buyer Mechanisms to Many Buyers",
                    "abstract": "For Bayesian combinatorial auctions, we present a general framework for approximately reducing the mechanism design problem for multiple buyers to the mechanism design problem for each individual buyer. Our framework can be applied to any setting which roughly satisfies the following assumptions: (i) The buyer's types must be distributed independently (not necessarily identically). (ii) The objective function must be linearly separable over the set of buyers (iii) The supply constraints must be the only constraints involving more than one buyer. Our framework is general in the sense that it makes no explicit assumption about any of the following: (i) The buyer's valuations (e.g., sub modular, additive, etc). (ii) The distribution of types for each buyer. (iii) The other constraints involving individual buyers (e.g., budget constraints, etc). We present two generic $n$-buyer mechanisms that use $1$-buyer mechanisms as black boxes. Assuming that we have an$\\alpha$-approximate $1$-buyer mechanism for each buyer\\footnote{Note that we can use different $1$-buyer mechanisms to accommodate different classes of buyers.} and assuming that no buyer ever needs more than $\\frac{1}{k}$ of all copies of each item for some integer $k \\ge 1$, then our generic $n$-buyer mechanisms are $\\gamma_k\\cdot\\alpha$-approximation of the optimal$n$-buyer mechanism, in which $\\gamma_k$ is a constant which is at least $1-\\frac{1}{\\sqrt{k+3}}$. Observe that $\\gamma_k$ is at least $\\frac{1}{2}$ (for $k=1$) and approaches $1$ as $k$ increases. As a byproduct of our construction, we improve a generalization of prophet inequalities. Furthermore, as applications of our main theorem, we improve several results from the literature."
                },
                {
                    "title": "Near optimal online algorithms and fast approximation algorithms for resource allocation problems",
                    "abstract": "We present algorithms for a class of resource allocation problems both in the online setting with stochastic input and in the offline setting. This class of problems contains many interesting special cases such as the Adwords problem. In the online setting we introduce a new distributional model called the adversarial stochastic input model, which is a generalization of the i.i.d model with unknown distributions, where the distributions can change over time. In this model we give a 1-O(\u03b5) approximation algorithm for the resource allocation problem, with almost the weakest possible assumption: the ratio of the maximum amount of resource consumed by any single request to the total capacity of the resource, and the ratio of the profit contributed by any single request to the optimal profit is at most (\u03b52/log(1/\u03b5)2)/(log n + log (1/\u03b5)) where n is the number of resources available. There are instances where this ratio is #949;2/log n such that no randomized algorithm can have a competitive ratio of 1-o(\u03b5) even in the i.i.d model. The upper bound on ratio that we require improves on the previous upper-bound for the i.i.d case by a factor of n.\n Our proof technique also gives a very simple proof that the greedy algorithm has a competitive ratio of 1-1/e for the Adwords problem in the i.i.d model with unknown distributions, and more generally in the adversarial stochastic input model, when there is no bound on the bid to budget ratio. All the previous proofs assume that either bids are very small compared to budgets or something very similar to this.\n In the offline setting we give a fast algorithm to solve very large LPs with both packing and covering constraints. We give algorithms to approximately solve (within a factor of 1+\u03b5) the mixed packing-covering problem with O(\u03b3 m log n/\u03b52) oracle calls where the constraint matrix of this LP has dimension n x m, and \u03b3 is a parameter which is very similar to the ratio described for the online setting.\n We discuss several applications, and how our algorithms improve existing results in some of these applications."
                },
                {
                    "title": "Multi-parameter mechanism design and sequential posted pricing",
                    "abstract": "We consider the classical mathematical economics problem of Bayesian optimal mechanism design where a principal aims to optimize expected revenue when allocating resources to self-interested agents with preferences drawn from a known distribution. In single-parameter settings (i.e., where each agent's preference is given by a single private value for being served and zero for not being served) this problem is solved. Unfortunately, these single parameter optimal mechanisms are impractical and rarely employed, and furthermore the underlying economic theory fails to generalize to the important, relevant, and unsolved multi-dimensional setting (i.e., where each agent's preference is given by multiple values for each of the multiple services available). In contrast to the theory of optimal mechanisms we develop a theory of sequential posted price mechanisms, where agents in sequence are offered take-it-or-leave-it prices. We prove that these mechanisms are approximately optimal in single-dimensional settings. These posted-price mechanisms avoid many of the properties of optimal mechanisms that make the latter impractical. Furthermore, these mechanisms generalize naturally to multidimensional settings where they give the first known approximations to the elusive optimal multi-dimensional mechanism design problem."
                },
                {
                    "title": "Bayesian Persuasion",
                    "abstract": "When is it possible for one person to persuade another to change her action? We take a mechanism design approach to this question. Taking preferences and initial beliefs as given, we introduce the notion of a persuasion mechanism: a game between Sender and Receiver defined by an information structure and a message technology. We derive necessary and sufficient conditions for the existence of a persuasion mechanism that strictly benefits Sender. We characterize the optimal mechanism. Finally, we analyze several examples that illustrate the applicability of our results."
                },
                {
                    "title": "Automated Online Mechanism Design and Prophet Inequalities",
                    "abstract": "Recent work on online auctions for digital goods has explored the role of optimal stopping theory -- particularly secretary problems -- in the design of approximately optimal online mechanisms. This work generally assumes that the size of the market (number of bidders) is known a priori, but that the mechanism designer has no knowledge of the distribution of bid values. However, in many real-world applications (such as online ticket sales), the opposite is true: the seller has distributional knowledge of the bid values (e.g., via the history of past transactions in the market), but there is uncertainty about market size. Adopting the perspective of automated mechanism design, introduced by Conitzer and Sandholm, we develop algorithms that compute an optimal, or approximately optimal, online auction mechanism given access to this distributional knowledge. Our main results are twofold. First, we show that when the seller does not know the market size, no constant-approximation to the optimum efficiency or revenue is achievable in the worst case, even under the very strong assumption that bid values are i.i.d. samples from a distribution known to the seller. Second, we show that when the seller has distributional knowledge of the market size as well as the bid values, one can do well in several senses. Perhaps most interestingly, by combining dynamic programming with prophet inequalities (a technique from optimal stopping theory) we are able to design and analyze online mechanisms which are temporally strategyproof (even with respect to arrival and departure times) and approximately efficiency (revenue)-maximizing. In exploring the interplay between automated mechanism design and prophet inequalities, we prove new prophet inequalities motivated by the auction setting."
                },
                {
                    "title": "Algorithmic pricing via virtual valuations",
                    "abstract": "Algorithmic pricing is the computational problem that sellers (e.g.,in supermarkets) face when trying to set prices for their items to maximize their profit in the presence of a known demand. Guruswami etal. (SODA, 2005) proposed this problem and gave logarithmic approximations (in the number of consumers) for the unit-demand and single-parameter cases where there is a specific set of consumers and their valuations for bundles are known precisely. Subsequently several versions of the problem have been shown to have poly-logarithmic in approximability. This problem has direct ties to the important open question of better understanding the Bayesian optimal mechanism in multi-parameter agent settings; however, for this purpose approximation factors logarithmic in the number of agents are inadequate. It is therefore of vital interest to consider special cases where constant approximations are possible. We consider the unit-demand variant of this pricing problem. Here a consumer has a valuation for each different item and their value for aset of items is simply the maximum value they have for any item in the set. Instead of considering a set of consumers with precisely known preferences, like the prior algorithmic pricing literature, we assume that the preferences of the consumers are drawn from a distribution. This is the standard assumption in economics; furthermore, the setting of a specific set of customers with specific preferences, which is employed in all of the prior work in algorithmic pricing, is a special case of this general Bayesian pricing problem, where there is a discrete Bayesian distribution for preferences specified by picking one consumer uniformly from the given set of consumers. Notice that the distribution over the valuations for the individual items that this generates is obviously correlated. Our work complements these existing works by considering the case where the consumer's valuations for the different items are independent random variables. Our main result is a constant approximation algorithm for this problem that makes use of an interesting connection between this problem and the concept of virtual valuations from the single-parameter Bayesian optimal mechanism design literature."
                },
                {
                    "title": "Matroids, secretary problems, and online mechanisms",
                    "abstract": "We study a generalization of the classical secretary problem which we call the \"matroid secretary problem\". In this problem, the elements of a matroid are presented to an online algorithm in random order. When an element arrives, the algorithm observes its value and must make an irrevocable decision regarding whether or not to accept it. The accepted elements must form an independent set, and the objective is to maximize the combined value of these elements. This paper presents an O(log k)-competitive algorithm for general matroids (where k is the rank of the matroid), and constant-competitive algorithms for several special cases including graphic matroids, truncated partition matroids, and bounded degree transversal matroids. We leave as an open question the existence of constant-competitive algorithms for general matroids. Our results have applications in welfare-maximizing online mechanism design for domains in which the sets of simultaneously satisfiable agents form a matroid."
                },
                {
                    "title": "Comparisons of Stop Rule and Supremum Expectations of I.I.D. Random Variables",
                    "abstract": "Implicitly defined (and easily approximated) universal constants 1.1 < an random variables and if Tn is the set of stop rules for Xl, \"', Xn, then E(max{Xl , \u2022 \u2022 \u2022 ,Xn}) ~ an sup {EX, : tE Tn}, and the bound an is best possible. Similar universal constants 0 < bn < Y. are found so that if the {Xi} are i.i.d. random variables taking values only in [a, b), then E (max {XI , .. , , Xn }) ~ sup {EXt: t E Tn} + bn(b-a), where again the bound bn is best possible. In both situations, extremal distributions for which equality is attained (or nearly attained) are given in implicit form."
                },
                {
                    "title": "Posted Pricing and Dynamic Prior-independent Mechanisms with Value Maximizers",
                    "abstract": "We study posted price auctions and dynamic prior-independent mechanisms for (ROI-constrained) value maximizers. In contrast to classic (quasi-linear) utility maximizers, these agents aim to maximize their total value subject to a minimum ratio of value per unit of payment made. When personalized posted prices are allowed, posted price auctions for value maximizers can be reduced to posted price auctions for utility maximizers. However, for anonymous posted prices, the well-known 12 approximation for utility maximizers is impossible for value maxi-mizers and we provide a posted price mechanism with 12 (1 \u2212 1 /e ) approximation. Moreover, we demonstrate how to apply our results to design prior-independent mechanisms in a dynamic environment; and to the best of our knowledge, this gives the first constant revenue approximation with multiple value maximizers. Finally, we provide an extension to combinatorial auctions with submodular / XOS agents."
                },
                {
                    "title": "Gaming Helps! Learning from Strategic Interactions in Natural Dynamics",
                    "abstract": "We consider an online regression setting in which individuals adapt to the regression model: arriving individuals are aware of the current model, and invest strategically in modifying their own features so as to improve the predicted score that the current model assigns to them. Such feature manipulation has been observed in various scenarios\u2014from credit assessment to school admissions\u2014 posing a challenge for the learner. Surprisingly, we find that such strategic manipulations may in fact help the learner recover the meaningful variables\u2014that is, the features that, when changed, affect the true label (as opposed to non-meaningful features that have no effect). We show that even simple behavior on the learner\u2019s part allows her to simultaneously i) accurately recover the meaningful features, and ii) incentivize agents to invest in these meaningful features, providing incentives for improvement."
                },
                {
                    "title": "Online Bayesian Persuasion",
                    "abstract": "In Bayesian persuasion, an informed sender has to design a signaling scheme that discloses the right amount of information so as to in\ufb02uence the behavior of a self-interested receiver. This kind of strategic interaction is ubiquitous in real-world economic scenarios. However, the seminal model by Kamenica and Gentzkow makes some stringent assumptions that limit its applicability in practice. One of the most limiting assumptions is, arguably, that the sender is required to know the receiver\u2019s utility function to compute an optimal signaling scheme. We relax this assumption through an online learning framework in which the sender repeatedly faces a receiver whose type is unknown and chosen adversarially at each round from a \ufb01nite set of possible types. We are interested in no-regret algorithms prescribing a signaling scheme at each round of the repeated interaction with performances close to that of a best-in-hindsight signaling scheme. First, we prove a hardness result on the per-round running time required to achieve no-\u03b1 -regret for any \u03b1 < 1 . Then, we provide algorithms for the full and partial feedback models with regret bounds sublinear in the number of rounds and polynomial in the size of the instance."
                },
                {
                    "title": "Theory Of Games And Statistical Decisions",
                    "abstract": "Thank you very much for downloading theory of games and statistical decisions. As you may know, people have search numerous times for their favorite books like this theory of games and statistical decisions, but end up in harmful downloads. Rather than reading a good book with a cup of tea in the afternoon, instead they juggled with some infectious virus inside their computer. theory of games and statistical decisions is available in our digital library an online access to it is set as public so you can download it instantly. Our book servers spans in multiple countries, allowing you to get the most less latency time to download any of our books like this one. Kindly say, the theory of games and statistical decisions is universally compatible with any devices to read."
                },
                {
                    "title": "Comparison of Experiments",
                    "abstract": "1. Summary Bohnenblust, Shapley, and Sherman [2] have introduced a method of comparing two sampling procedures or experiments; essentially their concept is that one experiment a is more informative than a second experiment ,, a v ,S, if, for every possible risk function, any risk attainable with , is also attainable with a. If a is a sufficient statistic for a procedure equivalent to ,S, a >,, it is shown that a v j3. In the case of dichotomies, the converse is proved. Whether > and v are equivalent in general is not known. Various properties of > and n are obtained, such as the following: if a > , and y is independent of both, then the combination (a, -y) > (#, y). An application to a problem in 2 X 2 tables is discussed."
                }
            ],
            "categories": [
                "cs.GT"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we adapt the classic prophet inequality framework to account for strategic players who selectively disclose information in sequential decision-making scenarios?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the understanding of optimal stopping problems in real-world applications where information asymmetry exists. By addressing the strategic behavior of players, this research could lead to more effective decision-making frameworks in various fields, such as recruitment, investment, and real estate. It could also inspire future research on online mechanism design and improve algorithms that rely on sequential decision-making under uncertainty.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in modeling the strategic behavior of players who may manipulate the information they disclose to influence the searcher's decision. Naive approaches that assume full disclosure of information will fail to capture the complexities of real-world interactions, where incentives can lead to biased or incomplete information. Overcoming these obstacles requires a deep understanding of game theory, information theory, and the dynamics of strategic interactions.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on scenarios where distributions are treated as static and fully disclosed, neglecting the impact of strategic behavior. The lack of models that incorporate the incentives of players to withhold or manipulate information has been a significant barrier. Our approach differs by integrating game-theoretic principles into the prophet inequality framework, allowing for a more realistic representation of decision-making processes in strategic environments.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a modified prophet inequality model that incorporates strategic players' behavior. We will use a combination of theoretical analysis and simulations based on real-world datasets, such as recruitment and investment scenarios, to evaluate the performance of our model. The key metrics will include the expected reward achieved by the searcher and the efficiency of the stopping policy. We expect to demonstrate that our approach can yield better outcomes than traditional models by effectively accounting for the strategic nature of information disclosure."
            }
        },
        "author_data": {
            "0209f83a-0b46-460c-ad19-683ce744ff37": {
                "pk": "0209f83a-0b46-460c-ad19-683ce744ff37",
                "name": "Wei Tang",
                "collaborators": [
                    "Margaret Martonosi",
                    "Jutho Haegeman",
                    "Shipra Agrawal",
                    "Yue Zhou"
                ],
                "domain": [
                    "Quantum Computing",
                    "Theoretical Physics",
                    "Dynamic Pricing",
                    "Combinatorial Design"
                ],
                "publications": [
                    {
                        "title": "Search for Massive Bosons Decaying to Wg and Zg Using the ATLAS Detector",
                        "abstract": "The Standard Model is by far the most encompassing physics theory. With the recent discovery of the Higgs Boson, the Standard Model has performed extremely well against experimental data. However, the theory is intrinsically not complete, it does not address physical phenomenon such as gravity and dark matter. Many proposals for physics beyond the Standard Model predict new massive bosons from additional gauge fields or extended Higgs sectors. This paper examines a preliminary search of possible V gamma decay channels of proposed new bosons."
                    },
                    {
                        "title": "Cutting Quantum Circuits to Run on Quantum and Classical Platforms",
                        "abstract": "Quantum computing (QC) offers a new computing paradigm that has the potential to provide significant speedups over classical computing. Each additional qubit doubles the size of the computational state space available to a quantum algorithm. Such exponentially expanding reach underlies QC's power, but at the same time puts demanding requirements on the quantum processing units (QPU) hardware. On the other hand, purely classical simulations of quantum circuits on either central processing unit (CPU) or graphics processing unit (GPU) scale poorly as they quickly become bottlenecked by runtime and memory. This paper introduces CutQC, a scalable hybrid computing approach that distributes a large quantum circuit onto quantum (QPU) and classical platforms (CPU or GPU) for co-processing. CutQC demonstrates evaluation of quantum circuits that are larger than the limit of QPU or classical simulation, and achieves much higher quantum circuit evaluation fidelity than the large NISQ devices achieve in real-system runs."
                    },
                    {
                        "title": "ScaleQC: A Scalable Framework for Hybrid Computation on Quantum and Classical Processors",
                        "abstract": "Quantum processing unit (QPU) has to satisfy highly demanding quantity and quality requirements on its qubits to produce accurate results for problems at useful scales. Furthermore, classical simulations of quantum circuits generally do not scale. Instead, quantum circuit cutting techniques cut and distribute a large quantum circuit into multiple smaller subcircuits feasible for less powerful QPUs. However, the classical post-processing incurred from the cutting introduces runtime and memory bottlenecks. Our tool, called ScaleQC, addresses the bottlenecks by developing novel algorithmic techniques including (1) a quantum states merging framework that quickly locates the solution states of large quantum circuits; (2) an automatic solver that cuts complex quantum circuits to fit on less powerful QPUs; and (3) a tensor network based post-processing that minimizes the classical overhead. Our experiments demonstrate both QPU requirement advantages over the purely quantum platforms, and runtime advantages over the purely classical platforms for benchmarks up to 1000 qubits."
                    },
                    {
                        "title": "Kac-Moody symmetries in one-dimensional bosonic systems",
                        "abstract": "In conformal field theories, when the conformal symmetry is enhanced by a global Lie group symmetry, the original Virasoro algebra can be extended to Kac-Moody algebra. In this paper, we extend the lattice construction of the Kac-Moody generators introduced in Wang et al., [Phys. Rev. B. 106, 115111 (2022)] to continuous systems and apply it to one-dimensional continuous boson systems. We justify this microscopic construction of Kac-Moody generators in two ways. First, through phenomenological bosonization, we express the microscopic construction in terms of the boson operators in the bosonization context, which can be related to the Kac-Moody generators in conformal field theories. Second, we study the behavior of the Kac-Moody generators in the integrable Lieb-Liniger model, and reveal its underlying particle-hole excitation picture through Bethe ansatz solutions. Finally, we test the computation of the Kac-Moody generator in the continuous matrix product state simulations, paving the way for more challenging non-integrable systems."
                    },
                    {
                        "title": "Dynamic Pricing and Learning with Long-term Reference Effects",
                        "abstract": "We consider a dynamic pricing problem where customer response to the current price is impacted by the customer price expectation, aka reference price. We study a simple and novel reference price mechanism where reference price is the average of the past prices offered by the seller. As opposed to the more commonly studied exponential smoothing mechanism, in our reference price mechanism the prices offered by seller have a longer term effect on the future customer expectations.   We show that under this mechanism, a markdown policy is near-optimal irrespective of the parameters of the model. This matches the common intuition that a seller may be better off by starting with a higher price and then decreasing it, as the customers feel like they are getting bargains on items that are ordinarily more expensive. For linear demand models, we also provide a detailed characterization of the near-optimal markdown policy along with an efficient way of computing it.   We then consider a more challenging dynamic pricing and learning problem, where the demand model parameters are apriori unknown, and the seller needs to learn them online from the customers' responses to the offered prices while simultaneously optimizing revenue. The objective is to minimize regret, i.e., the $T$-round revenue loss compared to a clairvoyant optimal policy. This task essentially amounts to learning a non-stationary optimal policy in a time-variant Markov Decision Process (MDP). For linear demand models, we provide an efficient learning algorithm with an optimal $\\tilde{O}(\\sqrt{T})$ regret upper bound."
                    },
                    {
                        "title": "Equivalence and automorphism groups of two families of maximum scattered linear sets",
                        "abstract": "Linear set in projective spaces over finite fields plays central roles in the study of blocking sets, semifields, rank-metric codes and etc. A linear set with the largest possible cardinality and the maximum rank is called maximum scattered. Despite two decades of study, there are only a few number of known maximum scattered linear sets in projective lines, including the family constructed by Csajb\\'ok, Marino, Polverino and Zanella 2018, and the family constructed by Csajb\\'ok, Marino, Zullo 2018 (also Marino, Montanucci, and Zullo 2020). This paper aims to solve the equivalence problem of the linear sets in each of these families and to determine their automorphism groups."
                    }
                ]
            },
            "0c0ae439-1650-4332-853c-949eb0259b76": {
                "pk": "0c0ae439-1650-4332-853c-949eb0259b76",
                "name": "Haifeng Xu",
                "collaborators": [
                    "Jiuru Zhou",
                    "Shaddin Dughmi",
                    "Zuyi Zhang",
                    "Jiaqiang Mei",
                    "James Nachbar",
                    "Binxian Yuan",
                    "Jerry Anunrojwong",
                    "Yiling Chen",
                    "Bo Waggoner",
                    "Ashwinkumar Badanidiyuru"
                ],
                "domain": [
                    "Number Theory",
                    "Game Theory",
                    "Algorithm Design",
                    "Adversarial Learning"
                ],
                "publications": [
                    {
                        "title": "The largest cycles consist by the quadratic residues and Fermat primes",
                        "abstract": "This paper studies the largest cycles consisted by the quadratic residues modulo prime numbers. We give some formulae about the maximum length of the cycles. Especially, the formula for modulo Fermat primes is given."
                    },
                    {
                        "title": "An alternative proof of the Dirichlet prime number theorem",
                        "abstract": "Dirichlet's theorem on arithmetic progressions called as Dirichlet prime number theorem is a classical result in number theory. Atle Selberg\\cite{Selberg} gave an elementary proof of this theorem. In this article we give an alternative proof of it based on a previous result of us. Also we get an estimation of the prime counting function in the special cases."
                    },
                    {
                        "title": "An equation about sum of primes with digital sum constraints",
                        "abstract": "We know that any prime number of form $4s+1$ can be written as a sum of two perfect square numbers. As a consequence of Goldbach's weak conjecture, any number great than $10$ can be represented as a sum of four primes. We are motivated to consider an equation with some constraints about digital sum for the four primes. And we conclude that the square root of the digital sum of the four primes will greater than $4$ and will not be a multiple of $3$ if the equation has solutions. In the proof, we give the method of determining whether a number is a perfect square."
                    },
                    {
                        "title": "On the Tractability of Public Persuasion with No Externalities",
                        "abstract": "Persuasion studies how a principal can influence agents' decisions via strategic information revelation --- often described as a signaling scheme --- in order to yield the most desirable equilibrium outcome. Recently, there has been a large body of algorithmic study of designing optimal public signaling schemes, a.k.a., public persuasion, which however is rifle with computational intractability results. In this paper, we design efficient and tight algorithms for public persuasion, and focus on a fundamental multi-agent persuasion model with no inter-agent externalities and binary actions introduced by Arieli and Babichenko. En route, we develop new algorithmic techniques which may be of independent interests.   First, we prove that optimal public persuasion is fixed parameter tractable. Our main result here relies on an interesting connection to a basic question in combinatorial geometry: how many cells can $n$ hyperplanes divide $R^d$ into? We use this connection to show a new characterization of public persuasion, which then enables efficient algorithm design. Second, we relax agent incentives and show that optimal public persuasion admits a bi-criteria PTAS for monotone submodular objectives and this approximation is tight. To prove this result, we establish an intriguing \"noise stability\" property of submodular functions which strictly generalizes the key result of Cheraghchi et al., originally motivated by applications of learning submodular functions and differential privacy. Finally, motivated by automated persuasion implemented as software, we consider relaxing the equilibrium concept of the model to coarse correlated equilibrium. Here we use a sophisticated primal-dual analysis to establish the polynomial-time equivalence between optimal public persuasion and the combinatorial problem of directly maximizing the sender's objective minus any linear function."
                    },
                    {
                        "title": "The Mysteries of Security Games: Equilibrium Computation Becomes Combinatorial Algorithm Design",
                        "abstract": "The security game is a basic model for resource allocation in adversarial environments. Here there are two players, a defender and an attacker. The defender wants to allocate her limited resources to defend critical targets and the attacker seeks his most favorable target to attack. In the past decade, there has been a surge of research interest in analyzing and solving security games that are motivated by applications from various domains. This paper examines security games from a theoretical perspective and provides a unified view of various security game models. In particular, each security game can be characterized by a set system $E$ which consists of the defender's pure strategies, The defender's best response problem can be viewed as a combinatorial optimization problem over $E$. Our framework captures most of the basic security game models in the literature, including all the deployed systems, The set system $E$ arising from various domains encodes standard combinatorial problems like bipartite matching, maximum coverage, min-cost flow, packing problems, etc. Our main result shows that equilibrium computation in security games is essentially a combinatorial problem. In particular, we prove that, for any set system $E$, the following problems can be reduced to each other in polynomial time: (0) combinatorial optimization over $E$, (1) computing the minimax equilibrium for zero-sum security games over $E$, (2) computing the strong Stackelberg equilibrium for security games over $E$, (3) computing the best or worst (for the defender) Nash equilibrium for security games over $E$. Here, by \"games over $E$\" we mean the class of security games with arbitrary payoff structures, but a fixed set $E$ of defender pure strategies. This shows that the complexity of a security game is essentially determined by the set system $E$."
                    },
                    {
                        "title": "Note on the H\u00f6lder norm estimate of the function $x\\sin(1/x)$",
                        "abstract": "In this paper, we prove the following inequality: for any $x, y>0$, there holds $$\\big|x\\sin\\frac{1}{x} - y\\sin\\frac{1}{y} \\big| \\leq \\sqrt{2|x - y|}.$$"
                    },
                    {
                        "title": "The connection between the Basel problem and a special integral",
                        "abstract": "By using Fubini theorem or Tonelli theorem, we find that the zeta function value at 2 is equal to a special integral. Furthermore, We find that this special integral is two times of another special integral. By using this fact we obtain the relationship between Genocchi numbers and Bernoulli numbers. And get some results about Bernoulli polynomials."
                    },
                    {
                        "title": "The Power of Signaling and its Intrinsic Connection to the Price of Anarchy",
                        "abstract": "A basic lesson from game theory is that strategic behavior often renders the equilibrium outcome inefficient. The recent literature of information design -- a.k.a. signaling or persuasion -- looks to improve equilibria by providing carefully-tuned information to players in order to influence their actions. Most previous studies have focused on the question of designing optimal signaling schemes. This work departs from previous research by considering a descriptive question and looks to quantitatively characterize the power of signaling (PoS), i.e., how much a signaling designer can improve her objective of the equilibrium outcome. We consider four signaling schemes with increasing power: full information, optimal public signaling, optimal private signaling and optimal ex-ante private signaling. Our main result is a clean and tight characterization of the additional power each signaling scheme has over its immediate predecessor above in the broad classes of cost-minimization and payoff-maximization games where: (1) all players minimize non-negative cost functions or maximize non-negative payoff functions; (2) the signaling designer (naturally) optimizes the sum of players' utilities. We prove that the additional power of signaling -- defined as the worst-case ratio between the equilibrium objectives of two consecutive signaling schemes in the above list -- is bounded precisely by the well-studied notion of the price of anarchy of the corresponding games. Moreover, we show that all these bounds are tight."
                    },
                    {
                        "title": "Algorithmic Bayesian Persuasion",
                        "abstract": "Persuasion, defined as the act of exploiting an informational advantage in order to effect the decisions of others, is ubiquitous. Indeed, persuasive communication has been estimated to account for almost a third of all economic activity in the US. This paper examines persuasion through a computational lens, focusing on what is perhaps the most basic and fundamental model in this space: the celebrated Bayesian persuasion model of Kamenica and Gentzkow. Here there are two players, a sender and a receiver. The receiver must take one of a number of actions with a-priori unknown payoff, and the sender has access to additional information regarding the payoffs. The sender can commit to revealing a noisy signal regarding the realization of the payoffs of various actions, and would like to do so as to maximize her own payoff assuming a perfectly rational receiver.   We examine the sender's optimization task in three of the most natural input models for this problem, and essentially pin down its computational complexity in each. When the payoff distributions of the different actions are i.i.d. and given explicitly, we exhibit a polynomial-time (exact) algorithm, and a \"simple\" $(1-1/e)$-approximation algorithm. Our optimal scheme for the i.i.d. setting involves an analogy to auction theory, and makes use of Border's characterization of the space of reduced-forms for single-item auctions. When action payoffs are independent but non-identical with marginal distributions given explicitly, we show that it is #P-hard to compute the optimal expected sender utility. Finally, we consider a general (possibly correlated) joint distribution of action payoffs presented by a black box sampling oracle, and exhibit a fully polynomial-time approximation scheme (FPTAS) with a bi-criteria guarantee. We show that this result is the best possible in the black-box model for information-theoretic reasons."
                    },
                    {
                        "title": "Algorithmic Persuasion with No Externalities",
                        "abstract": "We study the algorithmics of information structure design -- a.k.a. persuasion or signaling -- in a fundamental special case introduced by Arieli and Babichenko: multiple agents, binary actions, and no inter-agent externalities. Unlike prior work on this model, we allow many states of nature. We assume that the principal's objective is a monotone set function, and study the problem both in the public signal and private signal models, drawing a sharp contrast between the two in terms of both efficacy and computational complexity.   When private signals are allowed, our results are largely positive and quite general. First, we show polynomial-time equivalence between optimal signaling and the problem of maximizing the objective function plus an additive function. This yields an efficient implementation of the optimal scheme when the objective is supermodular or anonymous. Second, we exhibit a (1-1/e)-approximation of the optimal private signaling scheme, modulo an additive loss of $\\epsilon$, when the objective function is submodular. These two results simplify, unify, and generalize results of [Arieli and Babichenko, 2016] and [Babichenko and Barman, 2016], extending them from a binary state of nature to many states (modulo the additive loss in the latter result). Third, we consider the binary-state case with a submodular objective, and simplify and slightly strengthen the result of [Babichenko and Barman, 2016] to obtain a (1-1/e)-approximation via an explicitly constructed scheme.   When only a public signal is allowed, our results are negative. First, we show that it is NP-hard to approximate the optimal public scheme, within any constant factor, even when the objective is additive. Second, we show that the optimal private scheme can outperform the optimal public scheme, in terms of maximizing the sender's objective, by a polynomial factor."
                    },
                    {
                        "title": "Periodicity related to a sieve method of producing primes",
                        "abstract": "In this paper we consider a slightly different sieve method from Eratosthenes' to get primes. We find the periodicity and mirror symmetry of the pattern."
                    },
                    {
                        "title": "The limit of the m-norms of a class of symmetric matrices and its applications",
                        "abstract": "We consider a special symmetric matrix and obtain a similar formula as the one obtained by Weyl's criterion. Some applications of the formula are given, where we give a new way to calculate the integral of $\\ln\\Gamma(x)$ on $[0,1]$, and we claim that one class of matrices are not Hadamard matrices."
                    },
                    {
                        "title": "Computing Equilibria of Prediction Markets via Persuasion",
                        "abstract": "We study the computation of equilibria in prediction markets in perhaps the most fundamental special case with two players and three trading opportunities. To do so, we show equivalence of prediction market equilibria with those of a simpler signaling game with commitment introduced by Kong and Schoenebeck (2018). We then extend their results by giving computationally efficient algorithms for additional parameter regimes. Our approach leverages a new connection between prediction markets and Bayesian persuasion, which also reveals interesting conceptual insights."
                    },
                    {
                        "title": "Targeting and Signaling in Ad Auctions",
                        "abstract": "Modern ad auctions allow advertisers to target more specific segments of the user population. Unfortunately, this is not always in the best interest of the ad platform. In this paper, we examine the following basic question in the context of second-price ad auctions: how should an ad platform optimally reveal information about the ad opportunity to the advertisers in order to maximize revenue? We consider a model in which bidders' valuations depend on a random state of the ad opportunity. Different from previous work, we focus on a more practical, and challenging, situation where the space of possible realizations of ad opportunities is extremely large. We thus focus on developing algorithms whose running time is independent of the number of ad opportunity realizations.   We examine the auctioneer's algorithmic question of designing the optimal signaling scheme. When the auctioneer is restricted to send a public signal to all bidders, we focus on a well-motivated Bayesian valuation setting in which the auctioneer and bidders both have private information, and present two main results: 1. we exhibit a characterization result regarding approximately optimal schemes and prove that any constant-approximate public signaling scheme must use exponentially many signals; 2. we present a \"simple\" public signaling scheme that serves as a constant approximation under mild assumptions. We then initiate an exploration on the power of being able to send different signals privately to different bidders. Here we examine a basic setting where the auctioneer knows bidders' valuations, and exhibit a polynomial-time private scheme that extracts almost full surplus even in the worst Bayes Nash equilibrium. This illustrates the surprising power of private signaling schemes in extracting revenue."
                    },
                    {
                        "title": "PAC-Learning for Strategic Classification",
                        "abstract": "The study of strategic or adversarial manipulation of testing data to fool a classifier has attracted much recent attention. Most previous works have focused on two extreme situations where any testing data point either is completely adversarial or always equally prefers the positive label. In this paper, we generalize both of these through a unified framework for strategic classification, and introduce the notion of strategic VC-dimension (SVC) to capture the PAC-learnability in our general strategic setup. SVC provably generalizes the recent concept of adversarial VC-dimension (AVC) introduced by Cullina et al. arXiv:1806.01471. We instantiate our framework for the fundamental strategic linear classification problem. We fully characterize: (1) the statistical learnability of linear classifiers by pinning down its SVC; (2) its computational tractability by pinning down the complexity of the empirical risk minimization problem. Interestingly, the SVC of linear classifiers is always upper bounded by its standard VC-dimension. This characterization also strictly generalizes the AVC bound for linear classifiers in arXiv:1806.01471."
                    },
                    {
                        "title": "Optimal Pricing of Information",
                        "abstract": "A decision maker is choosing between an active action (e.g., purchase a house, invest certain stock) and a passive action. The payoff of the active action depends on the buyer's private type and also an unknown state of nature. An information seller can design experiments to reveal information about the realized state to the decision maker, and would like to maximize profit from selling such information. We characterize, in closed-form, the revenue-optimal information selling mechanism for the seller. After eliciting the buyer's type, the optimal mechanism charges the buyer an upfront payment and then simply reveals whether the realized state exceeds a certain threshold or not. The optimal mechanism features both price discrimination and information discrimination. The special buyer type who is a-priori indifferent between the active and passive action benefits the most from participating in the mechanism."
                    },
                    {
                        "title": "When Are Linear Stochastic Bandits Attackable?",
                        "abstract": "We study adversarial attacks on linear stochastic bandits: by manipulating the rewards, an adversary aims to control the behaviour of the bandit algorithm. Perhaps surprisingly, we first show that some attack goals can never be achieved. This is in sharp contrast to context-free stochastic bandits, and is intrinsically due to the correlation among arms in linear stochastic bandits. Motivated by this finding, this paper studies the attackability of a $k$-armed linear bandit environment. We first provide a complete necessity and sufficiency characterization of attackability based on the geometry of the arms' context vectors. We then propose a two-stage attack method against LinUCB and Robust Phase Elimination. The method first asserts whether the given environment is attackable; and if yes, it poisons the rewards to force the algorithm to pull a target arm linear times using only a sublinear cost. Numerical experiments further validate the effectiveness and cost-efficiency of the proposed attack method."
                    }
                ]
            },
            "532f4c6f-84b7-4b83-a893-b44f4aff8fe1": {
                "pk": "532f4c6f-84b7-4b83-a893-b44f4aff8fe1",
                "name": "Ruimin Zhang",
                "collaborators": [
                    "Julia Chuzhoy",
                    "Shuchi Chawla",
                    "Evangelia Gergatsouli",
                    "Yifeng Teng",
                    "Christos Tzamos",
                    "Dana G. Korssjoen",
                    "Biyao Li",
                    "Stefan Steinerberger",
                    "Raghavendra Tripathi"
                ],
                "domain": [
                    "Graph Algorithms",
                    "Approximation Algorithms",
                    "Stochastic Optimization",
                    "Dynamic Graphs"
                ],
                "publications": [
                    {
                        "title": "A New Deterministic Algorithm for Fully Dynamic All-Pairs Shortest Paths",
                        "abstract": "We study the fully dynamic All-Pairs Shortest Paths (APSP) problem in undirected edge-weighted graphs. Given an $n$-vertex graph $G$ with non-negative edge lengths, that undergoes an online sequence of edge insertions and deletions, the goal is to support approximate distance queries and shortest-path queries. We provide a deterministic algorithm for this problem, that, for a given precision parameter $\\epsilon$, achieves approximation factor $(\\log\\log n)^{2^{O(1/\\epsilon^3)}}$, and has amortized update time $O(n^{\\epsilon}\\log L)$ per operation, where $L$ is the ratio of longest to shortest edge length. Query time for distance-query is $O(2^{O(1/\\epsilon)}\\cdot \\log n\\cdot \\log\\log L)$, and query time for shortest-path query is $O(|E(P)|+2^{O(1/\\epsilon)}\\cdot \\log n\\cdot \\log\\log L)$, where $P$ is the path that the algorithm returns. To the best of our knowledge, even allowing any $o(n)$-approximation factor, no adaptive-update algorithms with better than $\\Theta(m)$ amortized update time and better than $\\Theta(n)$ query time were known prior to this work. We also note that our guarantees are stronger than the best current guarantees for APSP in decremental graphs in the adaptive-adversary setting."
                    },
                    {
                        "title": "Pandora's Box with Correlations: Learning and Approximation",
                        "abstract": "The Pandora's Box problem and its extensions capture optimization problems with stochastic input where the algorithm can obtain instantiations of input random variables at some cost. To our knowledge, all previous work on this class of problems assumes that different random variables in the input are distributed independently. As such it does not capture many real-world settings. In this paper, we provide the first approximation algorithms for Pandora's Box-type problems with correlations. We assume that the algorithm has access to samples drawn from the joint distribution on input.   Algorithms for these problems must determine an order in which to probe random variables, as well as when to stop and return the best solution found so far. In general, an optimal algorithm may make both decisions adaptively based on instantiations observed previously. Such fully adaptive (FA) strategies cannot be efficiently approximated to within any sublinear factor with sample access. We therefore focus on the simpler objective of approximating partially adaptive (PA) strategies that probe random variables in a fixed predetermined order but decide when to stop based on the instantiations observed. We consider a number of different feasibility constraints and provide simple PA strategies that are approximately optimal with respect to the best PA strategy for each case. All of our algorithms have polynomial sample complexity. We further show that our results are tight within constant factors: better factors cannot be achieved even using the full power of FA strategies."
                    },
                    {
                        "title": "Finding Structure in Sequences of Real Numbers via Graph Theory: a Problem List",
                        "abstract": "We investigate a method of generating a graph $G=(V,E)$ out of an ordered list of $n$ distinct real numbers $a_1, \\dots, a_n$. These graphs can be used to test for the presence of interesting structure in the sequence. We describe sequences exhibiting intricate hidden structure that was discovered this way. Our list includes sequences of Deutsch, Erd\\H{o}s, Freud & Hegyvari, Recaman, Quet, Zabolotskiy and Zizka. Since our observations are mostly empirical, each sequence in the list is an open problem."
                    }
                ]
            },
            "09900cf0-be37-4cf9-ac55-76903b70ed0d": {
                "pk": "09900cf0-be37-4cf9-ac55-76903b70ed0d",
                "name": "Derek Zhu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            }
        }
    },
    "2406.19861": {
        "paper_data": {
            "title": "Operator World Models for Reinforcement Learning",
            "url": "http://arxiv.org/abs/2406.19861v1",
            "arxiv_id": "2406.19861",
            "authors": [
                "Pietro Novelli",
                "Marco Prattic\u00f2",
                "Massimiliano Pontil",
                "Carlo Ciliberto"
            ],
            "abstract": "Policy Mirror Descent (PMD) is a powerful and theoretically sound methodology for sequential decision-making. However, it is not directly applicable to Reinforcement Learning (RL) due to the inaccessibility of explicit action-value functions. We address this challenge by introducing a novel approach based on learning a world model of the environment using conditional mean embeddings. We then leverage the operatorial formulation of RL to express the action-value function in terms of this quantity in closed form via matrix operations. Combining these estimators with PMD leads to POWR, a new RL algorithm for which we prove convergence rates to the global optimum. Preliminary experiments in finite and infinite state settings support the effectiveness of our method.",
            "introduction": "   1 Introduction 11footnotetext: Computational Statistics and Machine Learning - Istituto Italiano di Tecnologia, 16100 Genova, Italy22footnotetext: AI Centre, Computer Science Department, University College London, London, UK.  In recent years, Reinforcement Learning (RL) [1] has seen significant progress, with methods capable of tackling challenging applications such as robotic manipulation [2], playing Go [3] or Atari games [4] and resource management [5] to name but a few. The central challenge in RL settings is to balance the trade-off between exploration and exploitation, namely to improve upon previous policies while gathering sufficient information about the environment dynamics. Several strategies have been proposed to tackle this issue, such as Q-learning-based methods [4], policy optimization [6, 7] or actor-critics [8] to name a few. In contrast, when full information about the environment is available, sequential decision-making methods need only to focus on exploitation. Here, strategies such as policy improvement or policy iteration [9] have been thoroughly studied from both the algorithmic and theoretical standpoints. Within this context, the understanding of Policy Mirror Descent (PMD) methods has recently enjoyed a significant step forward, with results guaranteeing convergence to a global optimum with associated rates\u00a0[10, 11, 12].   In their original formulation, PMD methods require explicit knowledge of the action-value functions for all policies generated during the optimization process. This is clearly inaccessible in RL applications. Recently, [12] showed how PMD convergence rates can be extended to settings in which inexact estimators of the action-value function are used (see [13] for a similar result from a regret-based perspective). The resulting convergence rates, however, depend on uniform norm bounds on the approximation error, usually guaranteed only under unrealistic and inefficient assumptions such as the availability of a (perfect) simulator to be queried on arbitrary state-action pairs. Moreover, these strategies require repeating this sampling/learning process for any policy generated by the PMD algorithm, which is computationally expensive and demands numerous interactions with the environment. A natural question, therefore, is whether PMD approaches can be efficiently deployed in RL settings while enjoying the same strong theoretical guarantees.   In this work we address these issues by proposing a novel approach to estimating the action-value function. Unlike previous methods that directly approximate the action-value function from samples, we first learn the transition operator and reward function associated with the Markov decision process (MDP). To model the transition operator, we adopt the Conditional Mean Embedding (CME) framework [14, 15]. We then leverage an operatorial characterization of the action-value function to express it in terms of these estimated quantities. This strategy draws a peculiar connection with world model methods [16], and can be interpreted as world model-learning via CMEs. However, traditional world model methods emphasize learning an implicit model of the environment in the form of a simulator. This requires extensive sampling for application to PMD and incurs into to two sources of error in estimating the action-value function: model and sampling error. In contrast, CMEs can be used to estimate expectations without sampling and incur only in model error, for which learning bounds are available [17, 18]. One of our key results shows that by modeling the transition operator as",
            "references": [
                {
                    "title": "Estimating Koopman operators with sketching to provably learn large scale dynamical systems",
                    "abstract": "The theory of Koopman operators allows to deploy non-parametric machine learning algorithms to predict and analyze complex dynamical systems. Estimators such as principal component regression (PCR) or reduced rank regression (RRR) in kernel spaces can be shown to provably learn Koopman operators from finite empirical observations of the system's time evolution. Scaling these approaches to very long trajectories is a challenge and requires introducing suitable approximations to make computations feasible. In this paper, we boost the efficiency of different kernel-based Koopman operator estimators using random projections (sketching). We derive, implement and test the new\"sketched\"estimators with extensive experiments on synthetic and large-scale molecular dynamics datasets. Further, we establish non asymptotic error bounds giving a sharp characterization of the trade-offs between statistical learning rates and computational efficiency. Our empirical and theoretical analysis shows that the proposed estimators provide a sound and efficient way to learn large scale dynamical systems. In particular our experiments indicate that the proposed estimators retain the same accuracy of PCR or RRR, while being much faster."
                },
                {
                    "title": "Optimistic Planning by Regularized Dynamic Programming",
                    "abstract": "We propose a new method for optimistic planning in infinite-horizon discounted Markov decision processes based on the idea of adding regularization to the updates of an otherwise standard approximate value iteration procedure. This technique allows us to avoid contraction and monotonicity arguments typically required by existing analyses of approximate dynamic programming methods, and in particular to use approximate transition functions estimated via least-squares procedures in MDPs with linear function approximation. We use our method to recover known guarantees in tabular MDPs and to provide a computationally efficient algorithm for learning near-optimal policies in discounted linear mixture MDPs from a single stream of experience, and show it achieves near-optimal statistical guarantees."
                },
                {
                    "title": "Sharp Spectral Rates for Koopman Operator Learning",
                    "abstract": "Non-linear dynamical systems can be handily described by the associated Koopman operator, whose action evolves every observable of the system forward in time. Learning the Koopman operator and its spectral decomposition from data is enabled by a number of algorithms. In this work we present for the first time non-asymptotic learning bounds for the Koopman eigenvalues and eigenfunctions. We focus on time-reversal-invariant stochastic dynamical systems, including the important example of Langevin dynamics. We analyze two popular estimators: Extended Dynamic Mode Decomposition (EDMD) and Reduced Rank Regression (RRR). Our results critically hinge on novel minimax estimation bounds for the operator norm error, that may be of independent interest. Our spectral learning bounds are driven by the simultaneous control of the operator norm error and a novel metric distortion functional of the estimated eigenfunctions. The bounds indicates that both EDMD and RRR have similar variance, but EDMD suffers from a larger bias which might be detrimental to its learning rate. Our results shed new light on the emergence of spurious eigenvalues, an issue which is well known empirically. Numerical experiments illustrate the implications of the bounds in practice."
                },
                {
                    "title": "Mastering Diverse Domains through World Models",
                    "abstract": "Developing a general algorithm that learns to solve tasks across a wide range of applications has been a fundamental challenge in artificial intelligence. Although current reinforcement learning algorithms can be readily applied to tasks similar to what they have been developed for, configuring them for new application domains requires significant human expertise and experimentation. We present DreamerV3, a general algorithm that outperforms specialized methods across over 150 diverse tasks, with a single configuration. Dreamer learns a model of the environment and improves its behavior by imagining future scenarios. Robustness techniques based on normalization, balancing, and transformations enable stable learning across domains. Applied out of the box, Dreamer is the first algorithm to collect diamonds in Minecraft from scratch without human data or curricula. This achievement has been posed as a significant challenge in artificial intelligence that requires exploring farsighted strategies from pixels and sparse rewards in an open world. Our work allows solving challenging control problems without extensive experimentation, making reinforcement learning broadly applicable."
                },
                {
                    "title": "Optimal Rates for Regularized Conditional Mean Embedding Learning",
                    "abstract": "We address the consistency of a kernel ridge regression estimate of the conditional mean embedding (CME), which is an embedding of the conditional distribution of $Y$ given $X$ into a target reproducing kernel Hilbert space $\\mathcal{H}_Y$. The CME allows us to take conditional expectations of target RKHS functions, and has been employed in nonparametric causal and Bayesian inference. We address the misspecified setting, where the target CME is in the space of Hilbert-Schmidt operators acting from an input interpolation space between $\\mathcal{H}_X$ and $L_2$, to $\\mathcal{H}_Y$. This space of operators is shown to be isomorphic to a newly defined vector-valued interpolation space. Using this isomorphism, we derive a novel and adaptive statistical learning rate for the empirical CME estimator under the misspecified setting. Our analysis reveals that our rates match the optimal $O(\\log n / n)$ rates without assuming $\\mathcal{H}_Y$ to be finite dimensional. We further establish a lower bound on the learning rate, which shows that the obtained upper bound is optimal."
                },
                {
                    "title": "Learning Dynamical Systems via Koopman Operator Regression in Reproducing Kernel Hilbert Spaces",
                    "abstract": "We study a class of dynamical systems modelled as Markov chains that admit an invariant distribution via the corresponding transfer, or Koopman, operator. While data-driven algorithms to reconstruct such operators are well known, their relationship with statistical learning is largely unexplored. We formalize a framework to learn the Koopman operator from finite data trajectories of the dynamical system. We consider the restriction of this operator to a reproducing kernel Hilbert space and introduce a notion of risk, from which different estimators naturally arise. We link the risk with the estimation of the spectral decomposition of the Koopman operator. These observations motivate a reduced-rank operator regression (RRR) estimator. We derive learning bounds for the proposed estimator, holding both in i.i.d. and non i.i.d. settings, the latter in terms of mixing coefficients. Our results suggest RRR might be beneficial over other widely used estimators as confirmed in numerical experiments both for forecasting and mode decomposition."
                },
                {
                    "title": "On the Convergence Rates of Policy Gradient Methods",
                    "abstract": "We consider infinite-horizon discounted Markov decision problems with finite state and action spaces and study the convergence rates of the projected policy gradient method and a general class of policy mirror descent methods, all with direct parametrization in the policy space. First, we develop a theory of weak gradient-mapping dominance and use it to prove sharper sublinear convergence rate of the projected policy gradient method. Then we show that with geometrically increasing step sizes, a general class of policy mirror descent methods, including the natural policy gradient method and a projected Q-descent method, all enjoy a linear rate of convergence without relying on entropy or other strongly convex regularization. Finally, we also analyze the convergence rate of an inexact policy mirror descent method and estimate its sample complexity under a simple generative model."
                },
                {
                    "title": "Model-Based Reinforcement Learning with Value-Targeted Regression",
                    "abstract": "This paper studies model-based reinforcement learning (RL) for regret minimization. We focus on finite-horizon episodic RL where the transition model $P$ belongs to a known family of models $\\mathcal{P}$, a special case of which is when models in $\\mathcal{P}$ take the form of linear mixtures: $P_{\\theta} = \\sum_{i=1}^{d} \\theta_{i}P_{i}$. We propose a model based RL algorithm that is based on optimism principle: In each episode, the set of models that are `consistent' with the data collected is constructed. The criterion of consistency is based on the total squared error of that the model incurs on the task of predicting \\emph{values} as determined by the last value estimate along the transitions. The next value function is then chosen by solving the optimistic planning problem with the constructed set of models. We derive a bound on the regret, which, in the special case of linear mixtures, the regret bound takes the form $\\tilde{\\mathcal{O}}(d\\sqrt{H^{3}T})$, where $H$, $T$ and $d$ are the horizon, total number of steps and dimension of $\\theta$, respectively. In particular, this regret bound is independent of the total number of states or actions, and is close to a lower bound $\\Omega(\\sqrt{HdT})$. For a general model family $\\mathcal{P}$, the regret bound is derived using the notion of the so-called Eluder dimension proposed by Russo & Van Roy (2014)."
                },
                {
                    "title": "A General Framework for Consistent Structured Prediction with Implicit Loss Embeddings",
                    "abstract": "We propose and analyze a novel theoretical and algorithmic framework for structured prediction. While so far the term has referred to discrete output spaces, here we consider more general settings, such as manifolds or spaces of probability measures. We define structured prediction as a problem where the output space lacks a vectorial structure. We identify and study a large class of loss functions that implicitly defines a suitable geometry on the problem. The latter is the key to develop an algorithmic framework amenable to a sharp statistical analysis and yielding efficient computations. When dealing with output spaces with infinite cardinality, a suitable implicit formulation of the estimator is shown to be crucial."
                },
                {
                    "title": "Adaptive Trust Region Policy Optimization: Global Convergence and Faster Rates for Regularized MDPs",
                    "abstract": "Trust region policy optimization (TRPO) is a popular and empirically successful policy search algorithm in Reinforcement Learning (RL) in which a surrogate problem, that restricts consecutive policies to be \u2018close\u2019 to one another, is iteratively solved. Nevertheless, TRPO has been considered a heuristic algorithm inspired by Conservative Policy Iteration (CPI). We show that the adaptive scaling mechanism used in TRPO is in fact the natural \u201cRL version\u201d of traditional trust-region methods from convex analysis. We first analyze TRPO in the planning setting, in which we have access to the model and the entire state space. Then, we consider sample-based TRPO and establish \u00d5(1/\u221aN) convergence rate to the global optimum. Importantly, the adaptive scaling mechanism allows us to analyze TRPO in regularized MDPs for which we prove fast rates of \u00d5(1/N), much like results in convex optimization. This is the first result in RL of better rates when regularizing the instantaneous cost or reward."
                },
                {
                    "title": "On the Theory of Policy Gradient Methods: Optimality, Approximation, and Distribution Shift",
                    "abstract": "Policy gradient methods are among the most effective methods in challenging reinforcement learning problems with large state and/or action spaces. However, little is known about even their most basic theoretical convergence properties, including: if and how fast they converge to a globally optimal solution or how they cope with approximation error due to using a restricted class of parametric policies. This work provides provable characterizations of the computational, approximation, and sample size properties of policy gradient methods in the context of discounted Markov Decision Processes (MDPs). We focus on both: \"tabular\" policy parameterizations, where the optimal policy is contained in the class and where we show global convergence to the optimal policy; and parametric policy classes (considering both log-linear and neural policy classes), which may not contain the optimal policy and where we provide agnostic learning results. One central contribution of this work is in providing approximation guarantees that are average case -- which avoid explicit worst-case dependencies on the size of state space -- by making a formal connection to supervised learning under distribution shift. This characterization shows an important interplay between estimation error, approximation error, and exploration (as characterized through a precisely defined condition number)."
                },
                {
                    "title": "Sinkhorn Barycenters with Free Support via Frank-Wolfe Algorithm",
                    "abstract": "We present a novel algorithm to estimate the barycenter of arbitrary probability distributions with respect to the Sinkhorn divergence. Based on a Frank-Wolfe optimization strategy, our approach proceeds by populating the support of the barycenter incrementally, without requiring any pre-allocation. We consider discrete as well as continuous distributions, proving convergence rates of the proposed algorithm in both settings. Key elements of our analysis are a new result showing that the Sinkhorn divergence on compact domains has Lipschitz continuous gradient with respect to the Total Variation and a characterization of the sample complexity of Sinkhorn potentials. Experiments validate the effectiveness of our method in practice."
                },
                {
                    "title": "A Theory of Regularized Markov Decision Processes",
                    "abstract": "Many recent successful (deep) reinforcement learning algorithms make use of regularization, generally based on entropy or Kullback-Leibler divergence. We propose a general theory of regularized Markov Decision Processes that generalizes these approaches in two directions: we consider a larger class of regularizers, and we consider the general modified policy iteration approach, encompassing both policy iteration and value iteration. The core building blocks of this theory are a notion of regularized Bellman operator and the Legendre-Fenchel transform, a classical tool of convex optimization. This approach allows for error propagation analyses of general algorithmic schemes of which (possibly variants of) classical algorithms such as Trust Region Policy Optimization, Soft Q-learning, Stochastic Actor Critic or Dynamic Policy Programming are special cases. This also draws connections to proximal convex optimization, especially to Mirror Descent."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "Sobolev Norm Learning Rates for Regularized Least-Squares Algorithms",
                    "abstract": "Learning rates for least-squares regression are typically expressed in terms of $L_2$-norms. In this paper we extend these rates to norms stronger than the $L_2$-norm without requiring the regression function to be contained in the hypothesis space. In the special case of Sobolev reproducing kernel Hilbert spaces used as hypotheses spaces, these stronger norms coincide with fractional Sobolev norms between the used Sobolev space and $L_2$. As a consequence, not only the target function but also some of its derivatives can be estimated without changing the algorithm. From a technical point of view, we combine the well-known integral operator techniques with an embedding property, which so far has only been used in combination with empirical process arguments. This combination results in new finite sample bounds with respect to the stronger norms. From these finite sample bounds our rates easily follow. Finally, we prove the asymptotic optimality of our results in many cases."
                },
                {
                    "title": "Resource Management with Deep Reinforcement Learning",
                    "abstract": "Resource management problems in systems and networking often manifest as difficult online decision making tasks where appropriate solutions depend on understanding the workload and environment. Inspired by recent advances in deep reinforcement learning for AI problems, we consider building systems that learn to manage resources directly from experience. We present DeepRM, an example solution that translates the problem of packing tasks with multiple resource demands into a learning problem. Our initial results show that DeepRM performs comparably to state-of-the-art heuristics, adapts to different conditions, converges quickly, and learns strategies that are sensible in hindsight."
                },
                {
                    "title": "Kernel Mean Embedding of Distributions: A Review and Beyonds",
                    "abstract": "A Hilbert space embedding of a distribution---in short, a kernel mean embedding---has recently emerged as a powerful tool for machine learning and inference. The basic idea behind this framework is to map distributions into a reproducing kernel Hilbert space (RKHS) in which the whole arsenal of kernel methods can be extended to probability measures. It can be viewed as a generalization of the original \"feature map\" common to support vector machines (SVMs) and other kernel methods. While initially closely associated with the latter, it has meanwhile found application in fields ranging from kernel machines and probabilistic modeling to statistical inference, causal discovery, and deep learning. The goal of this survey is to give a comprehensive review of existing work and recent advances in this research area, and to discuss the most challenging issues and open problems that could lead to new research directions. The survey begins with a brief introduction to the RKHS and positive definite kernels which forms the backbone of this survey, followed by a thorough discussion of the Hilbert space embedding of marginal distributions, theoretical guarantees, and a review of its applications. The embedding of distributions enables us to apply RKHS methods to probability measures which prompts a wide range of applications such as kernel two-sample testing, independent testing, and learning on distributional data. Next, we discuss the Hilbert space embedding for conditional distributions, give theoretical insights, and review some applications. The conditional mean embedding enables us to perform sum, product, and Bayes' rules---which are ubiquitous in graphical model, probabilistic inference, and reinforcement learning---in a non-parametric way. We then discuss relationships between this framework and other related areas. Lastly, we give some suggestions on future research directions."
                },
                {
                    "title": "Less is More: Nystr\u00f6m Computational Regularization",
                    "abstract": "We study Nystrom type subsampling approaches to large scale kernel methods, and prove learning bounds in the statistical learning setting, where random sampling and high probability estimates are considered. In particular, we prove that these approaches can achieve optimal learning bounds, provided the subsampling level is suitably chosen. These results suggest a simple incremental variant of Nystrom Kernel Regularized Least Squares, where the subsampling level implements a form of computational regularization, in the sense that it controls at the same time regularization and computations. Extensive experimental analysis shows that the considered approach achieves state of the art performances on benchmark large scale datasets."
                },
                {
                    "title": "Convex Optimization: Algorithms and Complexity",
                    "abstract": "This monograph presents the main complexity theorems in convex optimization and their corresponding algorithms. Starting from the fundamental theory of black-box optimization, the material progresses towards recent advances in structural optimization and stochastic optimization. Our presentation of black-box optimization, strongly influenced by Nesterov's seminal book and Nemirovski's lecture notes, includes the analysis of cutting plane methods, as well as (accelerated) gradient descent schemes. We also pay special attention to non-Euclidean settings (relevant algorithms include Frank-Wolfe, mirror descent, and dual averaging) and discuss their relevance in machine learning. We provide a gentle introduction to structural optimization with FISTA (to optimize a sum of a smooth and a simple non-smooth term), saddle-point mirror prox (Nemirovski's alternative to Nesterov's smoothing), and a concise description of interior point methods. In stochastic optimization we discuss stochastic gradient descent, mini-batches, random coordinate descent, and sublinear algorithms. We also briefly touch upon convex relaxation of combinatorial problems and the use of randomness to round solutions, as well as random walks based methods."
                },
                {
                    "title": "Playing Atari with Deep Reinforcement Learning",
                    "abstract": "We present the first deep learning model to successfully learn control policies directly from high-dimensional sensory input using reinforcement learning. The model is a convolutional neural network, trained with a variant of Q-learning, whose input is raw pixels and whose output is a value function estimating future rewards. We apply our method to seven Atari 2600 games from the Arcade Learning Environment, with no adjustment of the architecture or learning algorithm. We find that it outperforms all previous approaches on six of the games and surpasses a human expert on three of them."
                },
                {
                    "title": "Modelling transition dynamics in MDPs with RKHS embeddings",
                    "abstract": "We propose a new, nonparametric approach to learning and representing transition dynamics in Markov decision processes (MDPs), which can be combined easily with dynamic programming methods for policy optimisation and value estimation. This approach makes use of a recently developed representation of conditional distributions as embeddings in a reproducing kernel Hilbert space (RKHS). Such representations bypass the need for estimating transition probabilities or densities, and apply to any domain on which kernels can be defined. This avoids the need to calculate intractable integrals, since expectations are represented as RKHS inner products whose computation has linear complexity in the number of points used to represent the embedding. We provide guarantees for the proposed applications in MDPs: in the context of a value iteration algorithm, we prove convergence to either the optimal policy, or to the closest projection of the optimal policy in our model class (an RKHS), under reasonable assumptions. In experiments, we investigate a learning task in a typical classical control setting (the under-actuated pendulum), and on a navigation problem where only images from a sensor are observed. For policy optimisation we compare with least-squares policy iteration where a Gaussian process is used for value function estimation. For value estimation we also compare to the NPDP method. Our approach achieves better performance in all experiments."
                },
                {
                    "title": "Conditional mean embeddings as regressors",
                    "abstract": "We demonstrate an equivalence between reproducing kernel Hilbert space (RKHS) embeddings of conditional distributions and vector-valued regressors. This connection introduces a natural regularized loss function which the RKHS embeddings minimise, providing an intuitive understanding of the embeddings and a justification for their use. Furthermore, the equivalence allows the application of vector-valued regression methods and results to the problem of learning conditional distributions. Using this link we derive a sparse version of the embedding by considering alternative formulations. Further, by applying convergence results for vector-valued regression to the embedding problem we derive minimax convergence rates which are O(\\log(n)/n) -- compared to current state of the art rates of O(n^{-1/4}) -- and are valid under milder and more intuitive assumptions. These minimax upper rates coincide with lower rates up to a logarithmic factor, showing that the embedding method achieves nearly optimal rates. We study our sparse embedding algorithm in a reinforcement learning task where the algorithm shows significant improvement in sparsity over an incomplete Cholesky decomposition."
                },
                {
                    "title": "Dimensionality Reduction for Supervised Learning with Reproducing Kernel Hilbert Spaces",
                    "abstract": "We propose a novel method of dimensionality reduction for supervised learning problems. Given a regression or classification problem in which we wish to predict a response variable Y from an explanatory variable X, we treat the problem of dimensionality reduction as that of finding a low-dimensional \"effective subspace\" for X which retains the statistical relationship between X and Y. We show that this problem can be formulated in terms of conditional independence. To turn this formulation into an optimization problem we establish a general nonparametric characterization of conditional independence using covariance operators on reproducing kernel Hilbert spaces. This characterization allows us to derive a contrast function for estimation of the effective subspace. Unlike many conventional methods for dimensionality reduction in supervised learning, the proposed method requires neither assumptions on the marginal distribution of X, nor a parametric model of the conditional distribution of Y. We present experiments that compare the performance of the method with conventional methods."
                },
                {
                    "title": "A Natural Policy Gradient",
                    "abstract": "We provide a natural gradient method that represents the steepest descent direction based on the underlying structure of the parameter space. Although gradient methods cannot make large changes in the values of the parameters, we show that the natural gradient is moving toward choosing a greedy optimal action rather than just a better action. These greedy optimal actions are those that would be chosen under one improvement step of policy iteration with approximate, compatible value functions, as defined by Sutton et al. [9]. We then show drastic performance improvements in simple MDPs and in the more challenging MDP of Tetris."
                },
                {
                    "title": "Foundations of Modern Probability",
                    "abstract": "* Measure Theory-Basic Notions * Measure Theory-Key Results * Processes, Distributions, and Independence * Random Sequences, Series, and Averages * Characteristic Functions and Classical Limit Theorems * Conditioning and Disintegration * Martingales and Optional Times * Markov Processes and Discrete-Time Chains * Random Walks and Renewal Theory * Stationary Processes and Ergodic Theory * Special Notions of Symmetry and Invariance * Poisson and Pure Jump-Type Markov Processes * Gaussian Processes and Brownian Motion * Skorohod Embedding and Invariance Principles * Independent Increments and Infinite Divisibility * Convergence of Random Processes, Measures, and Sets * Stochastic Integrals and Quadratic Variation * Continuous Martingales and Brownian Motion * Feller Processes and Semigroups * Ergodic Properties of Markov Processes * Stochastic Differential Equations and Martingale Problems * Local Time, Excursions, and Additive Functionals * One-Dimensional SDEs and Diffusions * Connections with PDEs and Potential Theory * Predictability, Compensation, and Excessive Functions * Semimartingales and General Stochastic Integration * Large Deviations * Appendix 1: Advanced Measure Theory * Appendix 2: Some Special Spaces * Historical and Bibliographical Notes * Bibliography * Indices"
                },
                {
                    "title": "Stable-Baselines3: Reliable Reinforcement Learning Implementations",
                    "abstract": "Stable-Baselines3 provides open-source implementations of deep reinforcement learning (RL) algorithms in Python. The implementations have been benchmarked against reference codebases, and automated unit tests cover 95% of the code. The algorithms follow a consistent interface and are accompanied by extensive documentation, making it simple to train and compare di\ufb00erent RL algorithms. Our documentation, examples, and source-code are available at https://github.com/DLR-RM/stable-baselines3 ."
                },
                {
                    "title": "Using the Nystr\u00f6m Method to Speed Up Kernel Machines",
                    "abstract": "A major problem for kernel-based predictors (such as Support Vector Machines and Gaussian processes) is that the amount of computation required to find the solution scales as O(n3), where n is the number of training examples. We show that an approximation to the eigendecomposition of the Gram matrix can be computed by the Nystrom method (which is used for the numerical solution of eigenproblems). This is achieved by carrying out an eigendecomposition on a smaller system of size m < n, and then expanding the results back up to n dimensions. The computational complexity of a predictor using this approximation is O(m2n). We report experiments on the USPS and abalone data sets and show that we can set m \u226a n without any significant decrease in the accuracy of the solution."
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                },
                {
                    "title": "Infinite Dimensional Analysis: A Hitchhiker\u2019s Guide",
                    "abstract": "This text was born out of an advanced mathematical economics seminar at Caltech in 1989-90. We realized that the typical graduate student in mathematical economics has to be familiar with a vast amount of material that spans several traditional fields in mathematics. Much of the mate\u00acrial appears only in esoteric research monographs that are designed for specialists, not for the sort of generalist that our students need be. We hope that in a small way this text will make the material here accessible to a much broader audience. While our motivation is to present and orga\u00acnize the analytical foundations underlying modern economics and finance, this is a book of mathematics, not of economics. We mention applications to economics but present very few of them. They are there to convince economists that the material has so me relevance and to let mathematicians know that there are areas of application for these results. We feel that this text could be used for a course in analysis that would benefit mathematicians, engineers, and scientists. Most of the material we present is available elsewhere, but is scattered throughout a variety of sources and occasionally buried in obscurity. Some of our results are original (or more likely, independent rediscoveries). We have included some material that we cannot honestly say is neces\u00acsary to understand modern economic theory, but may yet prove useful in future research."
                }
            ],
            "categories": [
                "cs.LG",
                "math.OC",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan Policy Mirror Descent (PMD) methods be efficiently deployed in Reinforcement Learning (RL) settings while maintaining strong theoretical guarantees?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it bridges the gap between theoretical advancements in PMD methods and their practical applicability in RL. By enabling efficient deployment of PMD in RL, we can enhance the performance of various applications, such as robotic manipulation and resource management, leading to more robust and intelligent systems. This research could pave the way for future studies that explore the integration of advanced optimization techniques in RL, ultimately advancing our understanding of decision-making processes in complex environments.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need for explicit knowledge of action-value functions in PMD methods, which is typically inaccessible in RL applications. Naive approaches may fail because they rely on direct approximation of action-value functions from samples, which can be computationally expensive and inefficient. Additionally, the requirement for numerous interactions with the environment to sample for each policy generated by the PMD algorithm poses significant practical obstacles. Overcoming these issues requires innovative methods to estimate action-value functions without incurring excessive sampling or model errors.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either theoretical aspects of PMD methods or practical implementations in RL, but rarely have they successfully integrated both. Existing solutions often depend on unrealistic assumptions, such as having access to a perfect simulator, which limits their applicability. Additionally, traditional world model methods emphasize implicit modeling of the environment, leading to compounded errors in action-value function estimation. Our approach differs by utilizing Conditional Mean Embedding (CME) to estimate expectations without sampling, thus reducing the sources of error and providing a more efficient framework for PMD in RL.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves first learning the transition operator and reward function of the Markov decision process (MDP) using the Conditional Mean Embedding (CME) framework. We then leverage an operatorial characterization of the action-value function to express it in terms of these estimated quantities. The dataset will consist of interactions from the RL environment, and we will evaluate our approach using metrics such as convergence rates and computational efficiency. We expect our results to demonstrate that PMD methods can be effectively"
            }
        },
        "author_data": {
            "339b5184-acf3-43e7-858c-3522506df85e": {
                "pk": "339b5184-acf3-43e7-858c-3522506df85e",
                "name": "Pietro Novelli",
                "collaborators": [
                    "Massimiliano Pontil",
                    "Vladimir Kostic",
                    "Karim Lounici",
                    "Vladimir R. Kostic",
                    "Marco Polini",
                    "Giacomo Turri",
                    "Fabio Taddei",
                    "Luigi Bonati",
                    "Michele Parrinello",
                    "Lorenzo Rosasco"
                ],
                "domain": [
                    "Dynamical Systems",
                    "Machine Learning",
                    "Koopman Operator",
                    "Quantum Systems"
                ],
                "publications": [
                    {
                        "title": "Learning dynamical systems: an example from open quantum system dynamics",
                        "abstract": "Machine learning algorithms designed to learn dynamical systems from data can be used to forecast, control and interpret the observed dynamics. In this work we exemplify the use of one of such algorithms, namely Koopman operator learning, in the context of open quantum system dynamics. We will study the dynamics of a small spin chain coupled with dephasing gates and show how Koopman operator learning is an approach to efficiently learn not only the evolution of the density matrix, but also of every physical observable associated to the system. Finally, leveraging the spectral decomposition of the learned Koopman operator, we show how symmetries obeyed by the underlying dynamics can be inferred directly from data."
                    },
                    {
                        "title": "A randomized algorithm to solve reduced rank operator regression",
                        "abstract": "We present and analyze an algorithm designed for addressing vector-valued regression problems involving possibly infinite-dimensional input and output spaces. The algorithm is a randomized adaptation of reduced rank regression, a technique to optimally learn a low-rank vector-valued function (i.e. an operator) between sampled data via regularized empirical risk minimization with rank constraints. We propose Gaussian sketching techniques both for the primal and dual optimization objectives, yielding Randomized Reduced Rank Regression (R4) estimators that are efficient and accurate. For each of our R4 algorithms we prove that the resulting regularized empirical risk is, in expectation w.r.t. randomness of a sketch, arbitrarily close to the optimal value when hyper-parameteres are properly tuned. Numerical expreriments illustrate the tightness of our bounds and show advantages in two distinct scenarios: (i) solving a vector-valued regression problem using synthetic and large-scale neuroscience datasets, and (ii) regressing the Koopman operator of a nonlinear stochastic dynamical system."
                    },
                    {
                        "title": "Sharp Spectral Rates for Koopman Operator Learning",
                        "abstract": "Nonlinear dynamical systems can be handily described by the associated Koopman operator, whose action evolves every observable of the system forward in time. Learning the Koopman operator and its spectral decomposition from data is enabled by a number of algorithms. In this work we present for the first time non-asymptotic learning bounds for the Koopman eigenvalues and eigenfunctions. We focus on time-reversal-invariant stochastic dynamical systems, including the important example of Langevin dynamics. We analyze two popular estimators: Extended Dynamic Mode Decomposition (EDMD) and Reduced Rank Regression (RRR). Our results critically hinge on novel {minimax} estimation bounds for the operator norm error, that may be of independent interest. Our spectral learning bounds are driven by the simultaneous control of the operator norm error and a novel metric distortion functional of the estimated eigenfunctions. The bounds indicates that both EDMD and RRR have similar variance, but EDMD suffers from a larger bias which might be detrimental to its learning rate. Our results shed new light on the emergence of spurious eigenvalues, an issue which is well known empirically. Numerical experiments illustrate the implications of the bounds in practice."
                    },
                    {
                        "title": "Failure of conductance quantization in two-dimensional topological insulators due to non-magnetic impurities",
                        "abstract": "Despite topological protection and the absence of magnetic impurities, two-dimensional topological insulators display quantized conductance only in surprisingly short channels, which can be as short as 100 nm for atomically thin materials. We show that the combined action of short-range nonmagnetic impurities located near the edges and on site electron-electron interactions effectively creates noncollinear magnetic scatterers, and, hence, results in strong backscattering. The mechanism causes deviations from quantization even at zero temperature and for a modest strength of electron-electron interactions. Our theory provides a straightforward conceptual framework to explain experimental results, especially those in atomically thin crystals, plagued with short-range edge disorder."
                    },
                    {
                        "title": "Characterizing metastable states with the help of machine learning",
                        "abstract": "Present-day atomistic simulations generate long trajectories of ever more complex systems. Analyzing these data, discovering metastable states, and uncovering their nature is becoming increasingly challenging. In this paper, we first use the variational approach to conformation dynamics to discover the slowest dynamical modes of the simulations. This allows the different metastable states of the system to be located and organized hierarchically. The physical descriptors that characterize metastable states are discovered by means of a machine learning method. We show in the cases of two proteins, Chignolin and Bovine Pancreatic Trypsin Inhibitor, how such analysis can be effortlessly performed in a matter of seconds. Another strength of our approach is that it can be applied to the analysis of both unbiased and biased simulations."
                    },
                    {
                        "title": "Consistent Long-Term Forecasting of Ergodic Dynamical Systems",
                        "abstract": "We study the evolution of distributions under the action of an ergodic dynamical system, which may be stochastic in nature. By employing tools from Koopman and transfer operator theory one can evolve any initial distribution of the state forward in time, and we investigate how estimators of these operators perform on long-term forecasting. Motivated by the observation that standard estimators may fail at this task, we introduce a learning paradigm that neatly combines classical techniques of eigenvalue deflation from operator theory and feature centering from statistics. This paradigm applies to any operator estimator based on empirical risk minimization, making them satisfy learning bounds which hold uniformly on the entire trajectory of future distributions, and abide to the conservation of mass for each of the forecasted distributions. Numerical experiments illustrates the advantages of our approach in practice."
                    },
                    {
                        "title": "Learning Dynamical Systems via Koopman Operator Regression in Reproducing Kernel Hilbert Spaces",
                        "abstract": "We study a class of dynamical systems modelled as Markov chains that admit an invariant distribution via the corresponding transfer, or Koopman, operator. While data-driven algorithms to reconstruct such operators are well known, their relationship with statistical learning is largely unexplored. We formalize a framework to learn the Koopman operator from finite data trajectories of the dynamical system. We consider the restriction of this operator to a reproducing kernel Hilbert space and introduce a notion of risk, from which different estimators naturally arise. We link the risk with the estimation of the spectral decomposition of the Koopman operator. These observations motivate a reduced-rank operator regression (RRR) estimator. We derive learning bounds for the proposed estimator, holding both in i.i.d. and non i.i.d. settings, the latter in terms of mixing coefficients. Our results suggest RRR might be beneficial over other widely used estimators as confirmed in numerical experiments both for forecasting and mode decomposition."
                    },
                    {
                        "title": "Learning invariant representations of time-homogeneous stochastic dynamical systems",
                        "abstract": "We consider the general class of time-homogeneous stochastic dynamical systems, both discrete and continuous, and study the problem of learning a representation of the state that faithfully captures its dynamics. This is instrumental to learning the transfer operator or the generator of the system, which in turn can be used for numerous tasks, such as forecasting and interpreting the system dynamics. We show that the search for a good representation can be cast as an optimization problem over neural networks. Our approach is supported by recent results in statistical learning theory, highlighting the role of approximation error and metric distortion in the learning problem. The objective function we propose is associated with projection operators from the representation space to the data space, overcomes metric distortion, and can be empirically estimated from data. In the discrete-time setting, we further derive a relaxed objective function that is differentiable and numerically well-conditioned. We compare our method against state-of-the-art approaches on different datasets, showing better performance across the board."
                    },
                    {
                        "title": "Transfer learning for atomistic simulations using GNNs and kernel mean embeddings",
                        "abstract": "Interatomic potentials learned using machine learning methods have been successfully applied to atomistic simulations. However, accurate models require large training datasets, while generating reference calculations is computationally demanding. To bypass this difficulty, we propose a transfer learning algorithm that leverages the ability of graph neural networks (GNNs) to represent chemical environments together with kernel mean embeddings. We extract a feature map from GNNs pre-trained on the OC20 dataset and use it to learn the potential energy surface from system-specific datasets of catalytic processes. Our method is further enhanced by incorporating into the kernel the chemical species information, resulting in improved performance and interpretability. We test our approach on a series of realistic datasets of increasing complexity, showing excellent generalization and transferability performance, and improving on methods that rely on GNNs or ridge regression alone, as well as similar fine-tuning approaches."
                    },
                    {
                        "title": "Optical and plasmonic properties of twisted bilayer graphene: Impact of interlayer tunneling asymmetry and ground-state charge inhomogeneity",
                        "abstract": "We present a theoretical study of the local optical conductivity, plasmon spectra, and thermoelectric properties of twisted bilayer graphene (TBG) at different filling factors and twist angles $\\theta$. Our calculations are based on the electronic band structures obtained from a continuum model that has two tunable parameters, $u_0$ and $u_1$, which parametrize the intra-sublattice inter-layer and inter-sublattice inter-layer tunneling rate, respectively. In this Article we focus on two key aspects: i) we study the dependence of our results on the value of $u_0$, exploring the whole range $0\\leq u_0\\leq u_1$; ii) we take into account effects arising from the intrinsic charge density inhomogeneity present in TBG, by calculating the band structures within the self-consistent Hartree approximation. At zero filling factor, i.e. at the charge neutrality point, the optical conductivity is quite sensitive to the value of $u_0$ and twist angle, whereas the charge inhomogeneity brings about only modest corrections. On the other hand, away from zero filling, static screening dominates and the optical conductivity is appreciably affected by the charge inhomogeneity, the largest effects being seen on the intra-band contribution to it. These findings are also reflected by the plasmonic spectra. We compare our results with existing ones in the literature, where effects i) and ii) above have not been studied systematically. As natural byproducts of our calculations, we obtain the Drude weight and Seebeck coefficient. The former displays an enhanced particle-hole asymmetry stemming from the inhomogeneous ground-state charge distribution. The latter is shown to display a broad sign-changing feature even at low temperatures ($\\approx 5~{\\rm K}$) due to the reduced slope of the bands, as compared to those of single-layer graphene."
                    },
                    {
                        "title": "Dynamics Harmonic Analysis of Robotic Systems: Application in Data-Driven Koopman Modelling",
                        "abstract": "We introduce the use of harmonic analysis to decompose the state space of symmetric robotic systems into orthogonal isotypic subspaces. These are lower-dimensional spaces that capture distinct, symmetric, and synergistic motions. For linear dynamics, we characterize how this decomposition leads to a subdivision of the dynamics into independent linear systems on each subspace, a property we term dynamics harmonic analysis (DHA). To exploit this property, we use Koopman operator theory to propose an equivariant deep-learning architecture that leverages the properties of DHA to learn a global linear model of the system dynamics. Our architecture, validated on synthetic systems and the dynamics of locomotion of a quadrupedal robot, exhibits enhanced generalization, sample efficiency, and interpretability, with fewer trainable parameters and computational costs."
                    },
                    {
                        "title": "Neural Conditional Probability for Inference",
                        "abstract": "We introduce NCP (Neural Conditional Probability), a novel operator-theoretic approach for learning conditional distributions with a particular focus on inference tasks. NCP can be used to build conditional confidence regions and extract important statistics like conditional quantiles, mean, and covariance. It offers streamlined learning through a single unconditional training phase, facilitating efficient inference without the need for retraining even when conditioning changes. By tapping into the powerful approximation capabilities of neural networks, our method efficiently handles a wide variety of complex probability distributions, effectively dealing with nonlinear relationships between input and output variables. Theoretical guarantees ensure both optimization consistency and statistical accuracy of the NCP method. Our experiments show that our approach matches or beats leading methods using a simple Multi-Layer Perceptron (MLP) with two hidden layers and GELU activations. This demonstrates that a minimalistic architecture with a theoretically grounded loss function can achieve competitive results without sacrificing performance, even in the face of more complex architectures."
                    },
                    {
                        "title": "Laplace Transform Based Low-Complexity Learning of Continuous Markov Semigroups",
                        "abstract": "Markov processes serve as a universal model for many real-world random processes. This paper presents a data-driven approach for learning these models through the spectral decomposition of the infinitesimal generator (IG) of the Markov semigroup. The unbounded nature of IGs complicates traditional methods such as vector-valued regression and Hilbert-Schmidt operator analysis. Existing techniques, including physics-informed kernel regression, are computationally expensive and limited in scope, with no recovery guarantees for transfer operator methods when the time-lag is small. We propose a novel method that leverages the IG's resolvent, characterized by the Laplace transform of transfer operators. This approach is robust to time-lag variations, ensuring accurate eigenvalue learning even for small time-lags. Our statistical analysis applies to a broader class of Markov processes than current methods while reducing computational complexity from quadratic to linear in the state dimension. Finally, we illustrate the behaviour of our method in two experiments."
                    },
                    {
                        "title": "Estimating Koopman operators with sketching to provably learn large scale dynamical systems",
                        "abstract": "The theory of Koopman operators allows to deploy non-parametric machine learning algorithms to predict and analyze complex dynamical systems. Estimators such as principal component regression (PCR) or reduced rank regression (RRR) in kernel spaces can be shown to provably learn Koopman operators from finite empirical observations of the system's time evolution. Scaling these approaches to very long trajectories is a challenge and requires introducing suitable approximations to make computations feasible. In this paper, we boost the efficiency of different kernel-based Koopman operator estimators using random projections (sketching). We derive, implement and test the new \"sketched\" estimators with extensive experiments on synthetic and large-scale molecular dynamics datasets. Further, we establish non asymptotic error bounds giving a sharp characterization of the trade-offs between statistical learning rates and computational efficiency. Our empirical and theoretical analysis shows that the proposed estimators provide a sound and efficient way to learn large scale dynamical systems. In particular our experiments indicate that the proposed estimators retain the same accuracy of PCR or RRR, while being much faster."
                    },
                    {
                        "title": "Moir\u00e9-Induced Transport in CVD-Based Small-Angle Twisted Bilayer Graphene",
                        "abstract": "To realize the applicative potential of 2D twistronic devices, scalable synthesis and assembly techniques need to meet stringent requirements in terms of interface cleanness and twist-angle homogeneity. Here, we show that small-angle twisted bilayer graphene assembled from separated CVD-grown graphene single-crystals can ensure high-quality transport properties, determined by a device-scale-uniform moire\\'e potential. Via low-temperature dual-gated magnetotransport, we demonstrate the hallmarks of a $2.4^\\circ$ -twisted superlattice, including tunable regimes of interlayer coupling, reduced Fermi velocity, large interlayer capacitance, and density-independent Brown-Zak oscillations. The observation of these moir\\'e-induced electrical transport features establishes CVD-based twisted bilayer graphene as an alternative to 'tear-and-stack' exfoliated flakes for fundamental studies, while serving as a proof-of-concept for future large-scale assembly."
                    }
                ]
            },
            "2866340e-133b-42a5-8f7b-fb79dad30583": {
                "pk": "2866340e-133b-42a5-8f7b-fb79dad30583",
                "name": "Marco Prattic\u00f2",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "529bd7fa-c671-49e3-aa9a-fcc011bf4dd2": {
                "pk": "529bd7fa-c671-49e3-aa9a-fcc011bf4dd2",
                "name": "Massimiliano Pontil",
                "collaborators": [
                    "Andreas Maurer",
                    "Dimitris Stamos",
                    "Bernardino Romera-Paredes",
                    "Andrew M. McDonald",
                    "Andreas Argyriou",
                    "Giulia Denevi",
                    "Andreas Maurer Massimiliano Pontil",
                    "Charles Micchelli",
                    "Luca Franceschi",
                    "Michele Donini"
                ],
                "domain": [
                    "Multi-task Learning",
                    "Regularization",
                    "Statistical Learning Theory",
                    "Sparse Coding"
                ],
                "publications": [
                    {
                        "title": "Bounds for Vector-Valued Function Estimation",
                        "abstract": "We present a framework to derive risk bounds for vector-valued learning with a broad class of feature maps and loss functions. Multi-task learning and one-vs-all multi-category learning are treated as examples. We discuss in detail vector-valued functions with one hidden layer, and demonstrate that the conditions under which shared representations are beneficial for multi- task learning are equally applicable to multi-category learning."
                    },
                    {
                        "title": "Some Hoeffding- and Bernstein-type Concentration Inequalities",
                        "abstract": "We prove concentration inequalities for functions of independent random variables {under} sub-gaussian and sub-exponential conditions. The utility of the inequalities is demonstrated by an extension of the now classical method of Rademacher complexities to Lipschitz function classes and unbounded sub-exponential distribution."
                    },
                    {
                        "title": "Empirical bounds for functions with weak interactions",
                        "abstract": "We provide sharp empirical estimates of expectation, variance and normal approximation for a class of statistics whose variation in any argument does not change too much when another argument is modified. Examples of such weak interactions are furnished by U- and V-statistics, Lipschitz L-statistics and various error functionals of L2-regularized algorithms and Gibbs algorithms."
                    },
                    {
                        "title": "Structured Sparsity and Generalization",
                        "abstract": "We present a data dependent generalization bound for a large class of regularized algorithms which implement structured sparsity constraints. The bound can be applied to standard squared-norm regularization, the Lasso, the group Lasso, some versions of the group Lasso with overlapping groups, multiple kernel learning and other regularization schemes. In all these cases competitive results are obtained. A novel feature of our bound is that it can be applied in an infinite dimensional setting such as the Lasso in a separable Hilbert space or multiple kernel learning with a countable number of kernels."
                    },
                    {
                        "title": "Excess risk bounds for multitask learning with trace norm regularization",
                        "abstract": "Trace norm regularization is a popular method of multitask learning. We give excess risk bounds with explicit dependence on the number of tasks, the number of examples per task and properties of the data distribution. The bounds are independent of the dimension of the input space, which may be infinite as in the case of reproducing kernel Hilbert spaces. A byproduct of the proof are bounds on the expected norm of sums of random positive semidefinite matrices with subexponential moments."
                    },
                    {
                        "title": "A New Convex Relaxation for Tensor Completion",
                        "abstract": "We study the problem of learning a tensor from a set of linear measurements. A prominent methodology for this problem is based on a generalization of trace norm regularization, which has been used extensively for learning low rank matrices, to the tensor setting. In this paper, we highlight some limitations of this approach and propose an alternative convex relaxation on the Euclidean ball. We then describe a technique to solve the associated regularization problem, which builds upon the alternating direction method of multipliers. Experiments on one synthetic dataset and two real datasets indicate that the proposed method improves significantly over tensor trace norm regularization in terms of estimation error, while remaining computationally tractable."
                    },
                    {
                        "title": "Uniform concentration and symmetrization for weak interactions",
                        "abstract": "The method to derive uniform bounds with Gaussian and Rademacher complexities is extended to the case where the sample average is replaced by a nonlinear statistic. Tight bounds are obtained for U-statistics, smoothened L-statistics and error functionals of l2-regularized algorithms."
                    },
                    {
                        "title": "Empirical Bernstein Bounds and Sample Variance Penalization",
                        "abstract": "We give improved constants for data dependent and variance sensitive confidence bounds, called empirical Bernstein bounds, and extend these inequalities to hold uniformly over classes of functionswhose growth function is polynomial in the sample size n. The bounds lead us to consider sample variance penalization, a novel learning method which takes into account the empirical variance of the loss function. We give conditions under which sample variance penalization is effective. In particular, we present a bound on the excess risk incurred by the method. Using this, we argue that there are situations in which the excess risk of our method is of order 1/n, while the excess risk of empirical risk minimization is of order 1/sqrt/{n}. We show some experimental results, which confirm the theory. Finally, we discuss the potential application of our results to sample compression schemes."
                    },
                    {
                        "title": "K-Dimensional Coding Schemes in Hilbert Spaces",
                        "abstract": "This paper presents a general coding method where data in a Hilbert space are represented by finite dimensional coding vectors. The method is based on empirical risk minimization within a certain class of linear operators, which map the set of coding vectors to the Hilbert space. Two results bounding the expected reconstruction error of the method are derived, which highlight the role played by the codebook and the class of linear operators. The results are specialized to some cases of practical importance, including K-means clustering, nonnegative matrix factorization and other sparse coding methods."
                    },
                    {
                        "title": "An Inequality with Applications to Structured Sparsity and Multitask Dictionary Learning",
                        "abstract": "From concentration inequalities for the suprema of Gaussian or Rademacher processes an inequality is derived. It is applied to sharpen existing and to derive novel bounds on the empirical Rademacher complexities of unit balls in various norms appearing in the context of structured sparsity and multitask dictionary learning or matrix factorization. A key role is played by the largest eigenvalue of the data covariance matrix."
                    },
                    {
                        "title": "When is there a representer theorem? Vector versus matrix regularizers",
                        "abstract": "We consider a general class of regularization methods which learn a vector of parameters on the basis of linear measurements. It is well known that if the regularizer is a nondecreasing function of the inner product then the learned vector is a linear combination of the input data. This result, known as the {\\em representer theorem}, is at the basis of kernel-based methods in machine learning. In this paper, we prove the necessity of the above condition, thereby completing the characterization of kernel methods based on regularization. We further extend our analysis to regularization methods which learn a matrix, a problem which is motivated by the application to multi-task learning. In this context, we study a more general representer theorem, which holds for a larger class of regularizers. We provide a necessary and sufficient condition for these class of matrix regularizers and highlight them with some concrete examples of practical importance. Our analysis uses basic principles from matrix theory, especially the useful notion of matrix nondecreasing function."
                    },
                    {
                        "title": "New Perspectives on $k$-Support and Cluster Norms",
                        "abstract": "We study a regularizer which is defined as a parameterized infimum of quadratics, and which we call the box-norm. We show that the k-support norm, a regularizer proposed by [Argyriou et al, 2012] for sparse vector prediction problems, belongs to this family, and the box-norm can be generated as a perturbation of the former. We derive an improved algorithm to compute the proximity operator of the squared box-norm, and we provide a method to compute the norm. We extend the norms to matrices, introducing the spectral k-support norm and spectral box-norm. We note that the spectral box-norm is essentially equivalent to the cluster norm, a multitask learning regularizer introduced by [Jacob et al. 2009a], and which in turn can be interpreted as a perturbation of the spectral k-support norm. Centering the norm is important for multitask learning and we also provide a method to use centered versions of the norms as regularizers. Numerical experiments indicate that the spectral k-support and box-norms and their centered variants provide state of the art performance in matrix completion and multitask learning problems respectively."
                    },
                    {
                        "title": "Fitting Spectral Decay with the $k$-Support Norm",
                        "abstract": "The spectral $k$-support norm enjoys good estimation properties in low rank matrix learning problems, empirically outperforming the trace norm. Its unit ball is the convex hull of rank $k$ matrices with unit Frobenius norm. In this paper we generalize the norm to the spectral $(k,p)$-support norm, whose additional parameter $p$ can be used to tailor the norm to the decay of the spectrum of the underlying model. We characterize the unit ball and we explicitly compute the norm. We further provide a conditional gradient method to solve regularization problems with the norm, and we derive an efficient algorithm to compute the Euclidean projection on the unit ball in the case $p=\\infty$. In numerical experiments, we show that allowing $p$ to vary significantly improves performance over the spectral $k$-support norm on various matrix completion benchmarks, and better captures the spectral decay of the underlying model."
                    },
                    {
                        "title": "Forward and Reverse Gradient-Based Hyperparameter Optimization",
                        "abstract": "We study two procedures (reverse-mode and forward-mode) for computing the gradient of the validation error with respect to the hyperparameters of any iterative learning algorithm such as stochastic gradient descent. These procedures mirror two methods of computing gradients for recurrent neural networks and have different trade-offs in terms of running time and space requirements. Our formulation of the reverse-mode procedure is linked to previous work by Maclaurin et al. [2015] but does not require reversible dynamics. The forward-mode procedure is suitable for real-time hyperparameter updates, which may significantly speed up hyperparameter optimization on large datasets. We present experiments on data cleaning and on learning task interactions. We also present one large-scale experiment where the use of previous gradient-based methods would be prohibitive."
                    },
                    {
                        "title": "Incremental Learning-to-Learn with Statistical Guarantees",
                        "abstract": "In learning-to-learn the goal is to infer a learning algorithm that works well on a class of tasks sampled from an unknown meta distribution. In contrast to previous work on batch learning-to-learn, we consider a scenario where tasks are presented sequentially and the algorithm needs to adapt incrementally to improve its performance on future tasks. Key to this setting is for the algorithm to rapidly incorporate new observations into the model as they arrive, without keeping them in memory. We focus on the case where the underlying algorithm is ridge regression parameterized by a positive semidefinite matrix. We propose to learn this matrix by applying a stochastic strategy to minimize the empirical error incurred by ridge regression on future tasks sampled from the meta distribution. We study the statistical properties of the proposed algorithm and prove non-asymptotic bounds on its excess transfer risk, that is, the generalization performance on new tasks from the same meta distribution. We compare our online learning-to-learn approach with a state of the art batch method, both theoretically and empirically."
                    },
                    {
                        "title": "Online Parameter-Free Learning of Multiple Low Variance Tasks",
                        "abstract": "We propose a method to learn a common bias vector for a growing sequence of low-variance tasks. Unlike state-of-the-art approaches, our method does not require tuning any hyper-parameter. Our approach is presented in the non-statistical setting and can be of two variants. The \"aggressive\" one updates the bias after each datapoint, the \"lazy\" one updates the bias only at the end of each task. We derive an across-tasks regret bound for the method. When compared to state-of-the-art approaches, the aggressive variant returns faster rates, the lazy one recovers standard rates, but with no need of tuning hyper-parameters. We then adapt the methods to the statistical setting: the aggressive variant becomes a multi-task learning method, the lazy one a meta-learning method. Experiments confirm the effectiveness of our methods in practice."
                    },
                    {
                        "title": "On Sparsity Inducing Regularization Methods for Machine Learning",
                        "abstract": "During the past years there has been an explosion of interest in learning methods based on sparsity regularization. In this paper, we discuss a general class of such methods, in which the regularizer can be expressed as the composition of a convex function $\\omega$ with a linear function. This setting includes several methods such the group Lasso, the Fused Lasso, multi-task learning and many more. We present a general approach for solving regularization problems of this kind, under the assumption that the proximity operator of the function $\\omega$ is available. Furthermore, we comment on the application of this approach to support vector machines, a technique pioneered by the groundbreaking work of Vladimir Vapnik."
                    },
                    {
                        "title": "Sparse coding for multitask and transfer learning",
                        "abstract": "We investigate the use of sparse coding and dictionary learning in the context of multitask and transfer learning. The central assumption of our learning method is that the tasks parameters are well approximated by sparse linear combinations of the atoms of a dictionary on a high or infinite dimensional space. This assumption, together with the large quantity of available data in the multitask and transfer learning settings, allows a principled choice of the dictionary. We provide bounds on the generalization error of this approach, for both settings. Numerical experiments on one synthetic and two real datasets show the advantage of our method over single task learning, a previous method based on orthogonal and dense representation of the tasks and a related method learning task grouping."
                    },
                    {
                        "title": "New Perspectives on k-Support and Cluster Norms",
                        "abstract": "The $k$-support norm is a regularizer which has been successfully applied to sparse vector prediction problems. We show that it belongs to a general class of norms which can be formulated as a parameterized infimum over quadratics. We further extend the $k$-support norm to matrices, and we observe that it is a special case of the matrix cluster norm. Using this formulation we derive an efficient algorithm to compute the proximity operator of both norms. This improves upon the standard algorithm for the $k$-support norm and allows us to apply proximal gradient methods to the cluster norm. We also describe how to solve regularization problems which employ centered versions of these norms. Finally, we apply the matrix regularizers to different matrix completion and multitask learning datasets. Our results indicate that the spectral $k$-support norm and the cluster norm give state of the art performance on these problems, significantly outperforming trace norm and elastic net penalties."
                    },
                    {
                        "title": "The Benefit of Multitask Representation Learning",
                        "abstract": "We discuss a general method to learn data representations from multiple tasks. We provide a justification for this method in both settings of multitask learning and learning-to-learn. The method is illustrated in detail in the special case of linear feature learning. Conditions on the theoretical advantage offered by multitask representation learning over independent task learning are established. In particular, focusing on the important example of half-space learning, we derive the regime in which multitask representation learning is beneficial over independent task learning, as a function of the sample size, the number of tasks and the intrinsic data dimensionality. Other potential applications of our results include multitask feature learning in reproducing kernel Hilbert spaces and multilayer, deep networks."
                    }
                ]
            },
            "03dd9575-96cc-4a0c-b7d4-9684526acf1b": {
                "pk": "03dd9575-96cc-4a0c-b7d4-9684526acf1b",
                "name": "Carlo Ciliberto",
                "collaborators": [
                    "Massimiliano Pontil",
                    "Alessandro Rudi",
                    "Lorenzo Rosasco",
                    "Ruohan Wang",
                    "Dimitris Stamos",
                    "Yiannis Demiris",
                    "Giulia Denevi",
                    "Giulia Luise",
                    "Pierluigi Amadori",
                    "Silvia Villa"
                ],
                "domain": [
                    "Machine Learning",
                    "Structured Prediction",
                    "Meta-Learning",
                    "Imitation Learning"
                ],
                "publications": [
                    {
                        "title": "PSD Representations for Effective Probability Models",
                        "abstract": "Finding a good way to model probability densities is key to probabilistic inference. An ideal model should be able to concisely approximate any probability while being also compatible with two main operations: multiplications of two models (product rule) and marginalization with respect to a subset of the random variables (sum rule). In this work, we show that a recently proposed class of positive semi-definite (PSD) models for non-negative functions is particularly suited to this end. In particular, we characterize both approximation and generalization capabilities of PSD models, showing that they enjoy strong theoretical guarantees. Moreover, we show that we can perform efficiently both sum and product rule in closed form via matrix operations, enjoying the same versatility of mixture models. Our results open the way to applications of PSD models to density estimation, decision theory and inference."
                    },
                    {
                        "title": "A Consistent Regularization Approach for Structured Prediction",
                        "abstract": "We propose and analyze a regularization approach for structured prediction problems. We characterize a large class of loss functions that allows to naturally embed structured outputs in a linear space. We exploit this fact to design learning algorithms using a surrogate loss approach and regularization techniques. We prove universal consistency and finite sample bounds characterizing the generalization properties of the proposed methods. Experimental results are provided to demonstrate the practical usefulness of the proposed approach."
                    },
                    {
                        "title": "Learning Multiple Visual Tasks while Discovering their Structure",
                        "abstract": "Multi-task learning is a natural approach for computer vision applications that require the simultaneous solution of several distinct but related problems, e.g. object detection, classification, tracking of multiple agents, or denoising, to name a few. The key idea is that exploring task relatedness (structure) can lead to improved performances.   In this paper, we propose and study a novel sparse, non-parametric approach exploiting the theory of Reproducing Kernel Hilbert Spaces for vector-valued functions. We develop a suitable regularization framework which can be formulated as a convex optimization problem, and is provably solvable using an alternating minimization approach. Empirical tests show that the proposed method compares favorably to state of the art techniques and further allows to recover interpretable structures, a problem of interest in its own right."
                    },
                    {
                        "title": "Consistent Multitask Learning with Nonlinear Output Relations",
                        "abstract": "Key to multitask learning is exploiting relationships between different tasks to improve prediction performance. If the relations are linear, regularization approaches can be used successfully. However, in practice assuming the tasks to be linearly related might be restrictive, and allowing for nonlinear structures is a challenge. In this paper, we tackle this issue by casting the problem within the framework of structured prediction. Our main contribution is a novel algorithm for learning multiple tasks which are related by a system of nonlinear equations that their joint outputs need to satisfy. We show that the algorithm is consistent and can be efficiently implemented. Experimental results show the potential of the proposed method."
                    },
                    {
                        "title": "Reexamining Low Rank Matrix Factorization for Trace Norm Regularization",
                        "abstract": "Trace norm regularization is a widely used approach for learning low rank matrices. A standard optimization strategy is based on formulating the problem as one of low rank matrix factorization which, however, leads to a non-convex problem. In practice this approach works well, and it is often computationally faster than standard convex solvers such as proximal gradient methods. Nevertheless, it is not guaranteed to converge to a global optimum, and the optimization can be trapped at poor stationary points. In this paper we show that it is possible to characterize all critical points of the non-convex problem. This allows us to provide an efficient criterion to determine whether a critical point is also a global minimizer. Our analysis suggests an iterative meta-algorithm that dynamically expands the parameter space and allows the optimization to escape any non-global critical point, thereby converging to a global minimizer. The algorithm can be applied to problems such as matrix completion or multitask learning, and our analysis holds for any random initialization of the factor matrices. Finally, we confirm the good performance of the algorithm on synthetic and real datasets."
                    },
                    {
                        "title": "Localized Structured Prediction",
                        "abstract": "Key to structured prediction is exploiting the problem structure to simplify the learning process. A major challenge arises when data exhibit a local structure (e.g., are made by \"parts\") that can be leveraged to better approximate the relation between (parts of) the input and (parts of) the output. Recent literature on signal processing, and in particular computer vision, has shown that capturing these aspects is indeed essential to achieve state-of-the-art performance. While such algorithms are typically derived on a case-by-case basis, in this work we propose the first theoretical framework to deal with part-based data from a general perspective. We derive a novel approach to deal with these problems and study its generalization properties within the setting of statistical learning theory. Our analysis is novel in that it explicitly quantifies the benefits of leveraging the part-based structure of the problem with respect to the learning rates of the proposed estimator."
                    },
                    {
                        "title": "A General Framework for Consistent Structured Prediction with Implicit Loss Embeddings",
                        "abstract": "We propose and analyze a novel theoretical and algorithmic framework for structured prediction. While so far the term has referred to discrete output spaces, here we consider more general settings, such as manifolds or spaces of probability measures. We define structured prediction as a problem where the output space lacks a vectorial structure. We identify and study a large class of loss functions that implicitly defines a suitable geometry on the problem. The latter is the key to develop an algorithmic framework amenable to a sharp statistical analysis and yielding efficient computations. When dealing with output spaces with infinite cardinality, a suitable implicit formulation of the estimator is shown to be crucial."
                    },
                    {
                        "title": "Structured Prediction for Conditional Meta-Learning",
                        "abstract": "The goal of optimization-based meta-learning is to find a single initialization shared across a distribution of tasks to speed up the process of learning new tasks. Conditional meta-learning seeks task-specific initialization to better capture complex task distributions and improve performance. However, many existing conditional methods are difficult to generalize and lack theoretical guarantees. In this work, we propose a new perspective on conditional meta-learning via structured prediction. We derive task-adaptive structured meta-learning (TASML), a principled framework that yields task-specific objective functions by weighing meta-training data on target tasks. Our non-parametric approach is model-agnostic and can be combined with existing meta-learning methods to achieve conditioning. Empirically, we show that TASML improves the performance of existing meta-learning models, and outperforms the state-of-the-art on benchmark datasets."
                    },
                    {
                        "title": "The Advantage of Conditional Meta-Learning for Biased Regularization and Fine-Tuning",
                        "abstract": "Biased regularization and fine-tuning are two recent meta-learning approaches. They have been shown to be effective to tackle distributions of tasks, in which the tasks' target vectors are all close to a common meta-parameter vector. However, these methods may perform poorly on heterogeneous environments of tasks, where the complexity of the tasks' distribution cannot be captured by a single meta-parameter vector. We address this limitation by conditional meta-learning, inferring a conditioning function mapping task's side information into a meta-parameter vector that is appropriate for that task at hand. We characterize properties of the environment under which the conditional approach brings a substantial advantage over standard meta-learning and we highlight examples of environments, such as those with multiple clusters, satisfying these properties. We then propose a convex meta-algorithm providing a comparable advantage also in practice. Numerical experiments confirm our theoretical findings."
                    },
                    {
                        "title": "Convex Learning of Multiple Tasks and their Structure",
                        "abstract": "Reducing the amount of human supervision is a key problem in machine learning and a natural approach is that of exploiting the relations (structure) among different tasks. This is the idea at the core of multi-task learning. In this context a fundamental question is how to incorporate the tasks structure in the learning problem.We tackle this question by studying a general computational framework that allows to encode a-priori knowledge of the tasks structure in the form of a convex penalty; in this setting a variety of previously proposed methods can be recovered as special cases, including linear and non-linear approaches. Within this framework, we show that tasks and their structure can be efficiently learned considering a convex optimization problem that can be approached by means of block coordinate methods such as alternating minimization and for which we prove convergence to the global minimum."
                    },
                    {
                        "title": "Differential Properties of Sinkhorn Approximation for Learning with Wasserstein Distance",
                        "abstract": "Applications of optimal transport have recently gained remarkable attention thanks to the computational advantages of entropic regularization. However, in most situations the Sinkhorn approximation of the Wasserstein distance is replaced by a regularized version that is less accurate but easy to differentiate. In this work we characterize the differential properties of the original Sinkhorn distance, proving that it enjoys the same smoothness as its regularized version and we explicitly provide an efficient algorithm to compute its gradient. We show that this result benefits both theory and applications: on one hand, high order smoothness confers statistical guarantees to learning with Wasserstein approximations. On the other hand, the gradient formula allows us to efficiently solve learning and optimization problems in practice. Promising preliminary experiments complement our analysis."
                    },
                    {
                        "title": "Leveraging Low-Rank Relations Between Surrogate Tasks in Structured Prediction",
                        "abstract": "We study the interplay between surrogate methods for structured prediction and techniques from multitask learning designed to leverage relationships between surrogate outputs. We propose an efficient algorithm based on trace norm regularization which, differently from previous methods, does not require explicit knowledge of the coding/decoding functions of the surrogate framework. As a result, our algorithm can be applied to the broad class of problems in which the surrogate space is large or even infinite dimensional. We study excess risk bounds for trace norm regularized structured prediction, implying the consistency and learning rates for our estimator. We also identify relevant regimes in which our approach can enjoy better generalization performance than previous methods. Numerical experiments on ranking problems indicate that enforcing low-rank relations among surrogate outputs may indeed provide a significant advantage in practice."
                    },
                    {
                        "title": "Learning-to-Learn Stochastic Gradient Descent with Biased Regularization",
                        "abstract": "We study the problem of learning-to-learn: inferring a learning algorithm that works well on tasks sampled from an unknown distribution. As class of algorithms we consider Stochastic Gradient Descent on the true risk regularized by the square euclidean distance to a bias vector. We present an average excess risk bound for such a learning algorithm. This result quantifies the potential benefit of using a bias vector with respect to the unbiased case. We then address the problem of estimating the bias from a sequence of tasks. We propose a meta-algorithm which incrementally updates the bias, as new tasks are observed. The low space and time complexity of this approach makes it appealing in practice. We provide guarantees on the learning ability of the meta-algorithm. A key feature of our results is that, when the number of tasks grows and their variance is relatively small, our learning-to-learn approach has a significant advantage over learning each task in isolation by Stochastic Gradient Descent without a bias term. We report on numerical experiments which demonstrate the effectiveness of our approach."
                    },
                    {
                        "title": "Random Expert Distillation: Imitation Learning via Expert Policy Support Estimation",
                        "abstract": "We consider the problem of imitation learning from a finite set of expert trajectories, without access to reinforcement signals. The classical approach of extracting the expert's reward function via inverse reinforcement learning, followed by reinforcement learning is indirect and may be computationally expensive. Recent generative adversarial methods based on matching the policy distribution between the expert and the agent could be unstable during training. We propose a new framework for imitation learning by estimating the support of the expert policy to compute a fixed reward function, which allows us to re-frame imitation learning within the standard reinforcement learning setting. We demonstrate the efficacy of our reward function on both discrete and continuous domains, achieving comparable or better performance than the state of the art under different reinforcement learning algorithms."
                    },
                    {
                        "title": "Sinkhorn Barycenters with Free Support via Frank-Wolfe Algorithm",
                        "abstract": "We present a novel algorithm to estimate the barycenter of arbitrary probability distributions with respect to the Sinkhorn divergence. Based on a Frank-Wolfe optimization strategy, our approach proceeds by populating the support of the barycenter incrementally, without requiring any pre-allocation. We consider discrete as well as continuous distributions, proving convergence rates of the proposed algorithm in both settings. Key elements of our analysis are a new result showing that the Sinkhorn divergence on compact domains has Lipschitz continuous gradient with respect to the Total Variation and a characterization of the sample complexity of Sinkhorn potentials. Experiments validate the effectiveness of our method in practice."
                    },
                    {
                        "title": "Statistical Limits of Supervised Quantum Learning",
                        "abstract": "Within the framework of statistical learning theory it is possible to bound the minimum number of samples required by a learner to reach a target accuracy. We show that if the bound on the accuracy is taken into account, quantum machine learning algorithms for supervised learning---for which statistical guarantees are available---cannot achieve polylogarithmic runtimes in the input dimension. We conclude that, when no further assumptions on the problem are made, quantum machine learning algorithms for supervised learning can have at most polynomial speedups over efficient classical algorithms, even in cases where quantum access to the data is naturally available."
                    },
                    {
                        "title": "Support-weighted Adversarial Imitation Learning",
                        "abstract": "Adversarial Imitation Learning (AIL) is a broad family of imitation learning methods designed to mimic expert behaviors from demonstrations. While AIL has shown state-of-the-art performance on imitation learning with only small number of demonstrations, it faces several practical challenges such as potential training instability and implicit reward bias. To address the challenges, we propose Support-weighted Adversarial Imitation Learning (SAIL), a general framework that extends a given AIL algorithm with information derived from support estimation of the expert policies. SAIL improves the quality of the reinforcement signals by weighing the adversarial reward with a confidence score from support estimation of the expert policy. We also show that SAIL is always at least as efficient as the underlying AIL algorithm that SAIL uses for learning the adversarial reward. Empirically, we show that the proposed method achieves better performance and training stability than baseline methods on a wide range of benchmark control tasks."
                    },
                    {
                        "title": "The Role of Global Labels in Few-Shot Classification and How to Infer Them",
                        "abstract": "Few-shot learning is a central problem in meta-learning, where learners must quickly adapt to new tasks given limited training data. Recently, feature pre-training has become a ubiquitous component in state-of-the-art meta-learning methods and is shown to provide significant performance improvement. However, there is limited theoretical understanding of the connection between pre-training and meta-learning. Further, pre-training requires global labels shared across tasks, which may be unavailable in practice. In this paper, we show why exploiting pre-training is theoretically advantageous for meta-learning, and in particular the critical role of global labels. This motivates us to propose Meta Label Learning (MeLa), a novel meta-learning framework that automatically infers global labels to obtains robust few-shot models. Empirically, we demonstrate that MeLa is competitive with existing methods and provide extensive ablation experiments to highlight its key properties."
                    },
                    {
                        "title": "Incremental Learning-to-Learn with Statistical Guarantees",
                        "abstract": "In learning-to-learn the goal is to infer a learning algorithm that works well on a class of tasks sampled from an unknown meta distribution. In contrast to previous work on batch learning-to-learn, we consider a scenario where tasks are presented sequentially and the algorithm needs to adapt incrementally to improve its performance on future tasks. Key to this setting is for the algorithm to rapidly incorporate new observations into the model as they arrive, without keeping them in memory. We focus on the case where the underlying algorithm is ridge regression parameterized by a positive semidefinite matrix. We propose to learn this matrix by applying a stochastic strategy to minimize the empirical error incurred by ridge regression on future tasks sampled from the meta distribution. We study the statistical properties of the proposed algorithm and prove non-asymptotic bounds on its excess transfer risk, that is, the generalization performance on new tasks from the same meta distribution. We compare our online learning-to-learn approach with a state of the art batch method, both theoretically and empirically."
                    },
                    {
                        "title": "Adversarial Imitation Learning with Trajectorial Augmentation and Correction",
                        "abstract": "Deep Imitation Learning requires a large number of expert demonstrations, which are not always easy to obtain, especially for complex tasks. A way to overcome this shortage of labels is through data augmentation. However, this cannot be easily applied to control tasks due to the sequential nature of the problem. In this work, we introduce a novel augmentation method which preserves the success of the augmented trajectories. To achieve this, we introduce a semi-supervised correction network that aims to correct distorted expert actions. To adequately test the abilities of the correction network, we develop an adversarial data augmented imitation architecture to train an imitation agent using synthetic experts. Additionally, we introduce a metric to measure diversity in trajectory datasets. Experiments show that our data augmentation strategy can improve accuracy and convergence time of adversarial imitation while preserving the diversity between the generated and real trajectories."
                    }
                ]
            }
        }
    },
    "2406.01486": {
        "paper_data": {
            "title": "Differentiable Task Graph Learning: Procedural Activity Representation and Online Mistake Detection from Egocentric Videos",
            "url": "http://arxiv.org/abs/2406.01486v1",
            "arxiv_id": "2406.01486",
            "authors": [
                "Luigi Seminara",
                "Giovanni Maria Farinella",
                "Antonino Furnari"
            ],
            "abstract": "Procedural activities are sequences of key-steps aimed at achieving specific goals. They are crucial to build intelligent agents able to assist users effectively. In this context, task graphs have emerged as a human-understandable representation of procedural activities, encoding a partial ordering over the key-steps. While previous works generally relied on hand-crafted procedures to extract task graphs from videos, in this paper, we propose an approach based on direct maximum likelihood optimization of edges' weights, which allows gradient-based learning of task graphs and can be naturally plugged into neural network architectures. Experiments on the CaptainCook4D dataset demonstrate the ability of our approach to predict accurate task graphs from the observation of action sequences, with an improvement of +16.7% over previous approaches. Owing to the differentiability of the proposed framework, we also introduce a feature-based approach, aiming to predict task graphs from key-step textual or video embeddings, for which we observe emerging video understanding abilities. Task graphs learned with our approach are also shown to significantly enhance online mistake detection in procedural egocentric videos, achieving notable gains of +19.8% and +7.5% on the Assembly101 and EPIC-Tent datasets. Code for replicating experiments is available at https://github.com/fpv-iplab/Differentiable-Task-Graph-Learning.",
            "introduction": "   1 Introduction  Procedural activities are fundamental for humans to organize tasks, improve efficiency, and ensuring consistency in the desired outcomes, but require time and effort to be learned and achieved effectively. This makes the design of artificial intelligent agents able to assist users to correctly perform a task appealing\u00a0[19, 27]. Achieving these abilities requires building a flexible representation of a procedure, encapsulating knowledge on the partial ordering of key-steps arising from the specific context at hand. For example, a virtual assistant needs to understand that it is necessary to break eggs before mixing them or that the bike\u2019s brakes need to be released before removing the wheel. Importantly, for a system to be scalable, this representation should be automatically learned from observations (e.g., humans making a recipe many times) rather than explicitly programmed by an expert.   Previous approaches focused on directly tackling tasks requiring procedural knowledge such as action anticipation\u00a0[15, 13, 30] and mistake detection\u00a0[33, 12, 6, 37, 14] without developing explicit representations of the procedure. Other works proposed neural models able to develop implicit representations of the procedure by learning how to recover missing actions\u00a0[39, 25], discover key-steps\u00a0[10, 4, 5], or grounding them to video\u00a0[9, 22]. A different approach\u00a0[3, 9, 16] consists in representing the structure of a procedure in the form of a task graph, i.e., a Directed Acyclic Graph (DAG) in which nodes represent key-steps, and directed edges impose a partial ordering over key-steps, encoding dependencies between them (see Figure\u00a01(a)). Graphs provide an explicit representation which is readily interpretable by humans and easy to incorporate in downstream tasks such as detecting mistakes or validating the execution of a procedure. While graphs have been historically used to represent constraints in complex tasks and design optimal sub-tasks scheduling\u00a0[34], graph-based representations mined from videos\u00a0[3], key-step sequences\u00a0[35, 18] or external knowledge bases\u00a0[40] have only recently emerged as a powerful representation of procedural activities able to support downstream tasks such as key-step recognition or forecasting\u00a0[3, 40]. Despite these efforts, current methods rely on meticulously crafted graph mining procedures rather than setting graph generation in a learning framework, limiting the inclusion of task graph representations in end-to-end systems.   Figure 1: (a) An example task graph encoding dependencies in a \u201cmix eggs\u201d procedure. (b) We learn a task graph which encodes a partial ordering between actions (left), represented as an adjacency matrix Z\ud835\udc4dZitalic_Z (center), from input action sequences (right). The proposed Task Graph Maximum Likelihood (TGML) loss directly supervises the entries of the adjacency matrix Z\ud835\udc4dZitalic_Z generating gradients to maximize the probability of edges from past nodes (K3,K1subscript\ud835\udc3e3subscript\ud835\udc3e1K_{3},K_{1}italic_K start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT , italic_K start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT) to the current node (K2subscript\ud835\udc3e2K_{2}italic_K start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT), while minimizing the probability of edges from past nodes to future nodes (K4,K5subscript\ud835\udc3e4subscript\ud835\udc3e5K_{4},K_{5}italic_K start_POSTSUBSCRIPT 4 end_POSTSUBSCRIPT , italic_K start_POSTSUBSCRIPT 5 end_POSTSUBSCRIPT) in a contrastive manner.   In this work, we propose an approach to learn task graphs from demonstrations in the form of sequences of key-steps performed by real users in a video while executing a procedure. Given a directed graph represented as an adjacency matrix and a set of key-step sequences, we provide an estimate of the likelihood of observing the",
            "references": [
                {
                    "title": "PREGO: Online Mistake Detection in PRocedural EGOcentric Videos",
                    "abstract": "Promptly identifying procedural errors from egocentric videos in an online setting is highly challenging and valuable for detecting mistakes as soon as they happen. This capability has a wide range of applications across various fields, such as manufacturing and healthcare. The nature of procedural mistakes is open-set since novel types of failures might occur, which calls for one-class classifiers trained on correctly executed procedures. However, no technique can currently detect open-set procedural mistakes online. We propose PREGO, the first online one-class classification model for mistake detection in PRocedural EGOcentric videos. PREGO is based on an online action recognition component to model the current action, and a symbolic reasoning module to predict the next actions. Mistake detection is performed by comparing the recognized current action with the expected future one. We evaluate PREGO on two procedural egocentric video datasets, Assembly101 and Epic-tent, which we adapt for online benchmarking of procedural mistake detection to establish suitable benchmarks, thus defining the Assembly101-O and Epic-tent-O datasets, respectively. The code is available at https://github.com/alef/abo/PREGO."
                },
                {
                    "title": "CaptainCook4D: A dataset for understanding errors in procedural activities",
                    "abstract": "Following step-by-step procedures is an essential component of various activities carried out by individuals in their daily lives. These procedures serve as a guiding framework that helps to achieve goals efficiently, whether it is assembling furniture or preparing a recipe. However, the complexity and duration of procedural activities inherently increase the likelihood of making errors. Understanding such procedural activities from a sequence of frames is a challenging task that demands an accurate interpretation of visual information and the ability to reason about the structure of the activity. To this end, we collect a new egocentric 4D dataset, CaptainCook4D, comprising 384 recordings (94.5 hours) of people performing recipes in real kitchen environments. This dataset consists of two distinct types of activity: one in which participants adhere to the provided recipe instructions and another in which they deviate and induce errors. We provide 5.3K step annotations and 10K fine-grained action annotations and benchmark the dataset for the following tasks: supervised error recognition, multistep localization, and procedure learning"
                },
                {
                    "title": "Ego-Exo4D: Understanding Skilled Human Activity from First- and Third-Person Perspectives",
                    "abstract": "We present Ego-Exo4D, a diverse, large-scale multi-modal multiview video dataset and benchmark challenge. Ego-Exo4D centers around simultaneously-captured ego-centric and exocentric video of skilled human activities (e.g., sports, music, dance, bike repair). 740 participants from 13 cities worldwide performed these activities in 123 different natural scene contexts, yielding long-form captures from 1 to 42 minutes each and 1,286 hours of video combined. The multimodal nature of the dataset is un-precedented: the video is accompanied by multichannel audio, eye gaze, 3D point clouds, camera poses, IMU, and multiple paired language descriptions-including a novel \u201cexpert commentary\u201d done by coaches and teachers and tailored to the skilled-activity domain. To push the frontier of first-person video understanding of skilled human activity, we also present a suite of benchmark tasks and their annotations, including fine-grained activity understanding, proficiency estimation, cross-view translation, and 3D hand/body pose. All resources are open sourced to fuel new research in the community."
                },
                {
                    "title": "United We Stand, Divided We Fall: UnityGraph for Unsupervised Procedure Learning from Videos",
                    "abstract": "Given multiple videos of the same task, procedure learning addresses identifying the key-steps and determining their order to perform the task. For this purpose, existing approaches use the signal generated from a pair of videos. This makes key-steps discovery challenging as the algorithms lack inter-videos perspective. Instead, we propose an unsupervised Graph-based Procedure Learning (GPL) framework. GPL consists of the novel UnityGraph that represents all the videos of a task as a graph to obtain both intra-video and inter-videos context. Further, to obtain similar embeddings for the same key-steps, the embeddings of UnityGraph are updated in an unsupervised manner using the Node2Vec algorithm. Finally, to identify the key-steps, we cluster the embeddings using KMeans. We test GPL on benchmark ProceL, CrossTask, and EgoProceL datasets and achieve an average improvement of 2% on third-person datasets and 3.6% on EgoProceL over the state-of-the-art."
                },
                {
                    "title": "MiniROAD: Minimal RNN Framework for Online Action Detection",
                    "abstract": "Online Action Detection (OAD) is the task of identifying actions in streaming videos without access to future frames. Much effort has been devoted to effectively capturing long-range dependencies, with transformers receiving the spotlight for their ability to capture long-range temporal structures. In contrast, RNNs have received less attention lately, due to their lower performance compared to recent methods that utilize transformers. In this paper, we investigate the underlying reasons for the inferior performance of RNNs compared to transformer-based algorithms. Our findings indicate that the discrepancy between training and inference is the primary hindrance to the effective training of RNNs. To address this, we propose applying non-uniform weights to the loss computed at each time step, which allows the RNN model to learn from the predictions made in an environment that better resembles the inference stage. Extensive experiments on three benchmark datasets, THUMOS, TVSeries, and FineAction demonstrate that a minimal RNN-based model trained with the proposed methodology performs equally or better than the existing best methods with a significant increase in efficiency. The code is available at https://github.com/jbistanbul/MiniROAD."
                },
                {
                    "title": "Weakly-Supervised Action Segmentation and Unseen Error Detection in Anomalous Instructional Videos",
                    "abstract": "We present a novel method for weakly-supervised action segmentation and unseen error detection in anomalous instructional videos. In the absence of an appropriate dataset for this task, we introduce the Anomalous Toy Assembly (ATA) dataset 1, which comprises 1152 untrimmed videos of 32 participants assembling three different toys, recorded from four different viewpoints. The training set comprises 27 participants who assemble toys in an expected and consistent manner, while the test and validation sets comprise 5 participants who display sequential anomalies in their task. We introduce a weakly labeled segmentation algorithm that is a generalization of the constrained Viterbi algorithm and identifies potential anomalous moments based on the difference between future anticipation and current recognition results. The proposed method is not restricted by the training transcripts during testing, allowing for the inference of anomalous action sequences while maintaining real-time performance. Based on these segmentation results, we also introduce a baseline to detect pre-defined human errors, and benchmark results on the ATA dataset. Experiments were conducted on the ATA and CSV datasets, outperforming the state-of-the-art in segmenting anomalous videos under both online and offline conditions."
                },
                {
                    "title": "HoloAssist: an Egocentric Human Interaction Dataset for Interactive AI Assistants in the Real World",
                    "abstract": "Building an interactive AI assistant that can perceive, reason, and collaborate with humans in the real world has been a long-standing pursuit in the AI community. This work is part of a broader research effort to develop intelligent agents that can interactively guide humans through performing tasks in the physical world. As a first step in this direction, we introduce HoloAssist, a large-scale egocentric human interaction dataset, where two people collaboratively complete physical manipulation tasks. The task performer executes the task while wearing a mixed-reality headset that captures seven synchronized data streams. The task instructor watches the performer\u2019s egocentric video in real time and guides them verbally. By augmenting the data with action and conversational annotations and observing the rich behaviors of various participants, we present key insights into how human assistants correct mistakes, intervene in the task completion procedure, and ground their instructions to the environment. HoloAssist spans 166 hours of data captured by 350 unique instructor-performer pairs. Furthermore, we construct and present benchmarks on mistake detection, intervention type prediction, and hand forecasting, along with detailed analysis. We expect HoloAssist will provide an important resource for building AI assistants that can fluidly collaborate with humans in the real world. Data can be downloaded at https://holoassist.github.io/."
                },
                {
                    "title": "An Outlook into the Future of Egocentric Vision",
                    "abstract": "What will the future be? We wonder! In this survey, we explore the gap between current research in egocentric vision and the ever-anticipated future, where wearable computing, with outward facing cameras and digital overlays, is expected to be integrated in our every day lives. To understand this gap, the article starts by envisaging the future through character-based stories, showcasing through examples the limitations of current technology. We then provide a mapping between this future and previously defined research tasks. For each task, we survey its seminal works, current state-of-the-art methodologies and available datasets, then reflect on shortcomings that limit its applicability to future research. Note that this survey focuses on software models for egocentric vision, independent of any specific hardware. The paper concludes with recommendations for areas of immediate explorations so as to unlock our path to the future always-on, personalised and life-enhancing egocentric vision."
                },
                {
                    "title": "Every Mistake Counts in Assembly",
                    "abstract": "One promising use case of AI assistants is to help with complex procedures like cooking, home repair, and assembly tasks. Can we teach the assistant to interject after the user makes a mistake? This paper targets the problem of identifying ordering mistakes in assembly procedures. We propose a system that can detect ordering mistakes by utilizing a learned knowledge base. Our framework constructs a knowledge base with spatial and temporal beliefs based on observed mistakes. Spatial beliefs depict the topological relationship of the assembling components, while temporal beliefs aggregate prerequisite actions as ordering constraints. With an episodic memory design, our algorithm can dynamically update and construct the belief sets as more actions are observed, all in an online fashion. We demonstrate experimentally that our inferred spatial and temporal beliefs are capable of identifying incorrect orderings in real-world action sequences. To construct the spatial beliefs, we collect a new set of coarse-level action annotations for Assembly101 based on the positioning of the toy parts. Finally, we demonstrate the superior performance of our belief inference algorithm in detecting ordering mistakes on the Assembly101 dataset."
                },
                {
                    "title": "Video-Mined Task Graphs for Keystep Recognition in Instructional Videos",
                    "abstract": "Procedural activity understanding requires perceiving human actions in terms of a broader task, where multiple keysteps are performed in sequence across a long video to reach a final goal state -- such as the steps of a recipe or a DIY fix-it task. Prior work largely treats keystep recognition in isolation of this broader structure, or else rigidly confines keysteps to align with a predefined sequential script. We propose discovering a task graph automatically from how-to videos to represent probabilistically how people tend to execute keysteps, and then leverage this graph to regularize keystep recognition in novel videos. On multiple datasets of real-world instructional videos, we show the impact: more reliable zero-shot keystep localization and improved video representation learning, exceeding the state of the art."
                },
                {
                    "title": "EgoVLPv2: Egocentric Video-Language Pre-training with Fusion in the Backbone",
                    "abstract": "Video-language pre-training (VLP) has become increasingly important due to its ability to generalize to various vision and language tasks. However, existing egocentric VLP frameworks utilize separate video and language encoders and learn task-specific cross-modal information only during fine-tuning, limiting the development of a unified system. In this work, we introduce the second generation of egocentric video-language pre-training (EgoVLPv2), a significant improvement from the previous generation, by incorporating cross-modal fusion directly into the video and language backbones. EgoVLPv2 learns strong video-text representation during pre-training and reuses the cross-modal attention modules to support different downstream tasks in a flexible and efficient manner, reducing fine-tuning costs. Moreover, our proposed fusion in the backbone strategy is more lightweight and compute-efficient than stacking additional fusion-specific layers. Extensive experiments on a wide range of VL tasks demonstrate the effectiveness of EgoVLPv2 by achieving consistent state-of-the-art performance over strong baselines across all downstream. Our project page can be found at https://shramanpramanick.github.io/EgoVLPv2/."
                },
                {
                    "title": "StepFormer: Self-Supervised Step Discovery and Localization in Instructional Videos",
                    "abstract": "Instructional videos are an important resource to learn procedural tasks from human demonstrations. However, the instruction steps in such videos are typically short and sparse, with most of the video being irrelevant to the procedure. This motivates the need to temporally localize the instruction steps in such videos, i.e. the task called key-step localization. Traditional methods for key-step localization require video-level human annotations and thus do not scale to large datasets. In this work, we tackle the problem with no human supervision and introduce StepFormer, a self-supervised model that discovers and localizes instruction steps in a video. StepFormer is a transformer decoder that attends to the video with learnable queries, and produces a sequence of slots capturing the key-steps in the video. We train our system on a large dataset of instructional videos, using their automatically-generated subtitles as the only source of supervision. In particular, we supervise our system with a sequence of text narrations using an order-aware loss function that filters out irrelevant phrases. We show that our model outperforms all previous unsupervised and weakly-supervised approaches on step detection and localization by a large margin on three challenging benchmarks. Moreover, our model demonstrates an emergent property to solve zero-shot multi-step localization and outperforms all relevant baselines at this task."
                },
                {
                    "title": "Procedure-Aware Pretraining for Instructional Video Understanding",
                    "abstract": "Our goal is to learn a video representation that is useful for downstream procedure understanding tasks in instructional videos. Due to the small amount of available annotations, a key challenge in procedure understanding is to be able to extract from unlabeled videos the procedural knowledge such as the identity of the task (e.g., \u2018make latte\u2019), its steps (e.g., \u2018pour milk\u2019), or the potential next steps given partial progress in its execution. Our main insight is that instructional videos depict sequences of steps that repeat between instances of the same or different tasks, and that this structure can be well represented by a Procedural Knowledge Graph (PKG), where nodes are discrete steps and edges connect steps that occur sequentially in the instructional activities. This graph can then be used to generate pseudo labels to train a video representation that encodes the procedural knowledge in a more accessible form to generalize to multiple procedure understanding tasks. We build a PKG by combining information from a text-based procedural knowledge database and an unlabeled instructional video corpus and then use it to generate training pseudo labels with four novel pre-training objectives. We call this PKG-based pre-training procedure and the resulting model Paprika, Procedure-Aware PRe-training for Instructional Knowledge Acquisition. We evaluate Paprika on COIN and CrossTask for procedure understanding tasks such as task recognition, step recognition, and step forecasting. Paprika yields a video representation that improves over the state of the art: up to 11.23% gains in accuracy in 12 evaluation settings. Implementation is available at https://github.com/salesforce/paprika."
                },
                {
                    "title": "Learning Procedure-aware Video Representation from Instructional Videos and Their Narrations",
                    "abstract": "The abundance of instructional videos and their narrations over the Internet offers an exciting avenue for understanding procedural activities. In this work, we propose to learn video representation that encodes both action steps and their temporal ordering, based on a large-scale dataset of web instructional videos and their narrations, without using human annotations. Our method jointly learns a video representation to encode individual step concepts, and a deep probabilistic model to capture both temporal dependencies and immense individual variations in the step ordering. We empirically demonstrate that learning temporal ordering not only enables new capabilities for procedure reasoning, but also reinforces the recognition of individual steps. Our model significantly advances the state-of-the-art results on step classification (+2.8%/+3.3% on COIN / EPIC-Kitchens) and step forecasting (+7.4% on COIN). Moreover, our model attains promising results in zero-shot inference for step classification and fore-casting, as well as in predicting diverse and plausible steps for incomplete procedures. Our code is available at https://github.com/facebookresearch/ProcedureVRL."
                },
                {
                    "title": "Learning and Verification of Task Structure in Instructional Videos",
                    "abstract": "Given the enormous number of instructional videos available online, learning a diverse array of multi-step task models from videos is an appealing goal. We introduce a new pre-trained video model, VideoTaskformer, focused on representing the semantics and structure of instructional videos. We pre-train VideoTaskformer using a simple and effective objective: predicting weakly supervised textual labels for steps that are randomly masked out from an instructional video (masked step modeling). Compared to prior work which learns step representations locally, our approach involves learning them globally, leveraging video of the entire surrounding task as context. From these learned representations, we can verify if an unseen video correctly executes a given task, as well as forecast which steps are likely to be taken after a given step. We introduce two new benchmarks for detecting mistakes in instructional videos, to verify if there is an anomalous step and if steps are executed in the right order. We also introduce a long-term forecasting benchmark, where the goal is to predict long-range future steps from a given step. Our method outperforms previous baselines on these tasks, and we believe the tasks will be a valuable way for the community to measure the quality of step representations. Additionally, we evaluate VideoTaskformer on 3 existing benchmarks -- procedural activity recognition, step classification, and step forecasting -- and demonstrate on each that our method outperforms existing baselines and achieves new state-of-the-art performance."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Multimodal Subtask Graph Generation from Instructional Videos",
                    "abstract": "Real-world tasks consist of multiple inter-dependent subtasks (e.g., a dirty pan needs to be washed before it can be used for cooking). In this work, we aim to model the causal dependencies between such subtasks from instructional videos describing the task. This is a challenging problem since complete information about the world is often inaccessible from videos, which demands robust learning mechanisms to understand the causal structure of events. We present Multimodal Subtask Graph Generation (MSG2), an approach that constructs a Subtask Graph defining the dependency between a task's subtasks relevant to a task from noisy web videos. Graphs generated by our multimodal approach are closer to human-annotated graphs compared to prior approaches. MSG2 further performs the downstream task of next subtask prediction 85% and 30% more accurately than recent video transformer models in the ProceL and CrossTask datasets, respectively."
                },
                {
                    "title": "Interaction Region Visual Transformer for Egocentric Action Anticipation",
                    "abstract": "Human-object interaction (HOI) and temporal dynamics along the motion paths are the most important visual cues for egocentric action anticipation. Especially, interaction regions covering objects and the human hand reveal significant visual cues to predict future human actions. However, how to incorporate and capture these important visual cues in modern video Transformer architecture remains a challenge. We leverage the effective MotionFormer that models motion dynamics to incorporate interaction regions using spatial cross-attention and further infuse contextual information using trajectory cross-attention to obtain an interaction-centric video representation for action anticipation. We term our model InAViT which achieves state-of-the-art action anticipation performance on large-scale egocentric datasets EPICKTICHENS100 (EK100) and EGTEA Gaze+. On the EK100 evaluation server, InAViT is on top of the public leader board (at the time of submission) where it outperforms the second-best model by 3.3% on mean-top5 recall. The code is available1."
                },
                {
                    "title": "Graph2Vid: Flow graph to Video Grounding for Weakly-supervised Multi-Step Localization",
                    "abstract": "In this work, we consider the problem of weakly-supervised multi-step localization in instructional videos. An established approach to this problem is to rely on a given list of steps. However, in reality, there is often more than one way to execute a procedure successfully, by following the set of steps in slightly varying orders. Thus, for successful localization in a given video, recent works require the actual order of procedure steps in the video, to be provided by human annotators at both training and test times. Instead, here, we only rely on generic procedural text that is not tied to a specific video. We represent the various ways to complete the procedure by transforming the list of instructions into a procedure flow graph which captures the partial order of steps. Using the flow graphs reduces both training and test time annotation requirements. To this end, we introduce the new problem of flow graph to video grounding. In this setup, we seek the optimal step ordering consistent with the procedure flow graph and a given video. To solve this problem, we propose a new algorithm - Graph2Vid - that infers the actual ordering of steps in the video and simultaneously localizes them. To show the advantage of our proposed formulation, we extend the CrossTask dataset with procedure flow graph information. Our experiments show that Graph2Vid is both more efficient than the baselines and yields strong step localization results, without the need for step order annotation."
                },
                {
                    "title": "My View is the Best View: Procedure Learning from Egocentric Videos",
                    "abstract": "Procedure learning involves identifying the key-steps and determining their logical order to perform a task. Existing approaches commonly use third-person videos for learning the procedure, making the manipulated object small in appearance and often occluded by the actor, leading to significant errors. In contrast, we observe that videos obtained from first-person (egocentric) wearable cameras provide an unobstructed and clear view of the action. However, procedure learning from egocentric videos is challenging because (a) the camera view undergoes extreme changes due to the wearer's head motion, and (b) the presence of unrelated frames due to the unconstrained nature of the videos. Due to this, current state-of-the-art methods' assumptions that the actions occur at approximately the same time and are of the same duration, do not hold. Instead, we propose to use the signal provided by the temporal correspondences between key-steps across videos. To this end, we present a novel self-supervised Correspond and Cut (CnC) framework for procedure learning. CnC identifies and utilizes the temporal correspondences between the key-steps across multiple videos to learn the procedure. Our experiments show that CnC outperforms the state-of-the-art on the benchmark ProceL and CrossTask datasets by 5.2% and 6.3%, respectively. Furthermore, for procedure learning using egocentric videos, we propose the EgoProceL dataset consisting of 62 hours of videos captured by 130 subjects performing 16 tasks. The source code and the dataset are available on the project page https://sid2697.github.io/egoprocel/."
                },
                {
                    "title": "Set-Supervised Action Learning in Procedural Task Videos via Pairwise Order Consistency",
                    "abstract": "We address the problem of set-supervised action learning, whose goal is to learn an action segmentation model using weak supervision in the form of sets of actions occurring in training videos. Our key observation is that videos within the same task have similar ordering of actions, which can be leveraged for effective learning. Therefore, we propose an attention-based method with a new Pairwise Ordering Consistency (POC) loss that encourages that for each common action pair in two videos of the same task, the attentions of actions follow a similar ordering. Unlike existing sequence alignment methods, which misalign actions in videos with different orderings or cannot reliably separate more from less consistent orderings, our POC loss efficiently aligns videos with different action orders and is differentiable, which enables end-to-end training. In addition, it avoids the time-consuming pseudo-label generation of prior works. Our method efficiently learns the actions and their temporal locations, therefore, extends the existing attention-based action localization methods from learning one action per video to multiple actions using our POC loss along with video-level and frame-level losses. By experiments on three datasets, we demonstrate that our method significantly improves the state of the art. We also show that our method, with a small modification, can effectively address the transcript-supervised action learning task, where actions and their ordering are available during training.11Code available at https://github.com/ZijiaLewisLu/CVPR22-POC."
                },
                {
                    "title": "Assembly101: A Large-Scale Multi-View Video Dataset for Understanding Procedural Activities",
                    "abstract": "Assembly101 is a new procedural activity dataset fea-turing 4321 videos of people assembling and disassembling 101 \u201ctake-apart\u201d toy vehicles. Participants work without fixed instructions, and the sequences feature rich and natu-ral variations in action ordering, mistakes, and corrections. Assembly101 is the first multi-view action dataset, with si-multaneous static (8) and egocentric (4) recordings. Se-quences are annotated with more than 100K coarse and 1M fine-grained action segments, and I8M 3D hand poses. We benchmark on three action understanding tasks: recognition, anticipation and temporal segmentation. Ad-ditionally, we propose a novel task of detecting mistakes. The unique recording format and rich set of annotations al-low us to investigate generalization to new toys, cross-view transfer, long-tailed distributions, and pose vs. appearance. We envision that Assemblyl0l will serve as a new challenge to investigate various activity understanding problems."
                },
                {
                    "title": "Learning To Recognize Procedural Activities with Distant Supervision",
                    "abstract": "In this paper we consider the problem of classifying fine-grained, multi-step activities (e.g., cooking different recipes, making disparate home improvements, creating various forms of arts and crafts) from long videos spanning up to several minutes. Accurately categorizing these activities requires not only recognizing the individual steps that compose the task but also capturing their temporal dependencies. This problem is dramatically different from traditional action classification, where models are typically optimized on videos that span only a few seconds and that are manually trimmed to contain simple atomic actions. While step annotations could enable the training of models to recognize the individual steps of procedural activities, existing large-scale datasets in this area do not include such segment labels due to the prohibitive cost of manually annotating temporal boundaries in long videos. To address this issue, we propose to automatically identify steps in instructional videos by leveraging the distant supervision of a textual knowledge base (wikiHow) that includes detailed descriptions of the steps needed for the execution of a wide variety of complex activities. Our method uses a language model to match noisy, automatically-transcribed speech from the video to step descriptions in the knowledge base. We demonstrate that video models trained to recognize these automatically-labeled steps (without manual supervision) yield a representation that achieves superior generalization performance on four downstream tasks: recognition of procedural activities, step classification, step forecasting and egocentric video classification."
                },
                {
                    "title": "Drop-DTW: Aligning Common Signal Between Sequences While Dropping Outliers",
                    "abstract": "In this work, we consider the problem of sequence-to-sequence alignment for signals containing outliers. Assuming the absence of outliers, the standard Dynamic Time Warping (DTW) algorithm efficiently computes the optimal alignment between two (generally) variable-length sequences. While DTW is robust to temporal shifts and dilations of the signal, it fails to align sequences in a meaningful way in the presence of outliers that can be arbitrarily interspersed in the sequences. To address this problem, we introduce Drop-DTW, a novel algorithm that aligns the common signal between the sequences while automatically dropping the outlier elements from the matching. The entire procedure is implemented as a single dynamic program that is efficient and fully differentiable. In our experiments, we show that Drop-DTW is a robust similarity measure for sequence retrieval and demonstrate its effectiveness as a training loss on diverse applications. With Drop-DTW, we address temporal step localization on instructional videos, representation learning from noisy videos, and cross-modal representation learning for audio-visual retrieval and localization. In all applications, we take a weakly- or unsupervised approach and demonstrate state-of-the-art results under these settings."
                },
                {
                    "title": "Anticipative Video Transformer",
                    "abstract": "We propose Anticipative Video Transformer (AVT), an end-to-end attention-based video modeling architecture that attends to the previously observed video in order to anticipate future actions. We train the model jointly to predict the next action in a video sequence, while also learning frame feature encoders that are predictive of successive future frames\u2019 features. Compared to existing temporal aggregation strategies, AVT has the advantage of both maintaining the sequential progression of observed actions while still capturing long-range dependencies\u2014both critical for the anticipation task. Through extensive experiments, we show that AVT obtains the best reported performance on four popular action anticipation benchmarks: EpicKitchens-55, EpicKitchens-100, EGTEA Gaze+, and 50-Salads; and it wins first place in the EpicKitchens-100 CVPR\u201921 challenge."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "English Recipe Flow Graph Corpus",
                    "abstract": "We present an annotated corpus of English cooking recipe procedures, and describe and evaluate computational methods for learning these annotations. The corpus consists of 300 recipes written by members of the public, which we have annotated with domain-specific linguistic and semantic structure. Each recipe is annotated with (1) \u2018recipe named entities\u2019 (r-NEs) specific to the recipe domain, and (2) a flow graph representing in detail the sequencing of steps, and interactions between cooking tools, food ingredients and the products of intermediate steps. For these two kinds of annotations, inter-annotator agreement ranges from 82.3 to 90.5 F1, indicating that our annotation scheme is appropriate and consistent. We experiment with producing these annotations automatically. For r-NE tagging we train a deep neural network NER tool; to compute flow graphs we train a dependency-style parsing procedure which we apply to the entire sequence of r-NEs in a recipe.In evaluations, our systems achieve 71.1 to 87.5 F1, demonstrating that our annotation scheme is learnable."
                },
                {
                    "title": "Rolling-Unrolling LSTMs for Action Anticipation from First-Person Video",
                    "abstract": "In this paper, we tackle the problem of egocentric action anticipation, i.e., predicting what actions the camera wearer will perform in the near future and which objects they will interact with. Specifically, we contribute Rolling-Unrolling LSTM, a learning architecture to anticipate actions from egocentric videos. The method is based on three components: 1) an architecture comprised of two LSTMs to model the sub-tasks of summarizing the past and inferring the future, 2) a Sequence Completion Pre-Training technique which encourages the LSTMs to focus on the different sub-tasks, and 3) a Modality ATTention (MATT) mechanism to efficiently fuse multi-modal predictions performed by processing RGB frames, optical flow fields and object-based features. The proposed approach is validated on EPIC-Kitchens, EGTEA Gaze+ and ActivityNet. The experiments show that the proposed architecture is state-of-the-art in the domain of egocentric videos, achieving top performances in the 2019 EPIC-Kitchens egocentric action anticipation challenge. The approach also achieves competitive performance on ActivityNet with respect to methods not based on unsupervised pre-training and generalizes to the tasks of early action recognition and action recognition. To encourage research on this challenging topic, we made our code, trained models, and pre-extracted features available at our web page: http://iplab.dmi.unict.it/rulstm."
                },
                {
                    "title": "Meta Reinforcement Learning with Autonomous Inference of Subtask Dependencies",
                    "abstract": "We propose and address a novel few-shot RL problem, where a task is characterized by a subtask graph which describes a set of subtasks and their dependencies that are unknown to the agent. The agent needs to quickly adapt to the task over few episodes during adaptation phase to maximize the return in the test phase. Instead of directly learning a meta-policy, we develop a Meta-learner with Subtask Graph Inference(MSGI), which infers the latent parameter of the task by interacting with the environment and maximizes the return given the latent parameter. To facilitate learning, we adopt an intrinsic reward inspired by upper confidence bound (UCB) that encourages efficient exploration. Our experiment results on two grid-world domains and StarCraft II environments show that the proposed method is able to accurately infer the latent task parameter, and to adapt more efficiently than existing meta RL and hierarchical RL methods."
                },
                {
                    "title": "End-to-End Learning of Visual Representations From Uncurated Instructional Videos",
                    "abstract": "Annotating videos is cumbersome, expensive and not scalable. Yet, many strong video models still rely on manually annotated data. With the recent introduction of the HowTo100M dataset, narrated videos now offer the possibility of learning video representations without manual supervision. In this work we propose a new learning approach, MIL-NCE, capable of addressing mis- alignments inherent in narrated videos. With this approach we are able to learn strong video representations from scratch, without the need for any manual annotation. We evaluate our representations on a wide range of four downstream tasks over eight datasets: action recognition (HMDB-51, UCF-101, Kinetics-700), text-to- video retrieval (YouCook2, MSR-VTT), action localization (YouTube-8M Segments, CrossTask) and action segmentation (COIN). Our method outperforms all published self-supervised approaches for these tasks as well as several fully supervised baselines."
                },
                {
                    "title": "EPIC-Tent: An Egocentric Video Dataset for Camping Tent Assembly",
                    "abstract": "This paper presents an outdoor video dataset annotated with action labels, collected from 24 participants wearing two head-mounted cameras (GoPro and SMI eye tracker) while assembling a camping tent. In total, this is 5.4 hours of recordings. Tent assembly includes manual interactions with non-rigid objects such as spreading the tent, securing guylines, reading instructions, and opening a tent bag. An interesting aspect of the dataset is that it reflects participants' proficiency in completing or understanding the task. This leads to participant differences in action sequences and action durations. Our dataset, called EPIC-Tent, also has several new types of annotations for two synchronised egocentric videos. These include task errors, self-rated uncertainty and gaze position, in addition to the task action labels. We present baseline results on the EPIC-Tent dataset using a state-of-the-art method for offline and online action recognition and detection."
                },
                {
                    "title": "Cross-Task Weakly Supervised Learning From Instructional Videos",
                    "abstract": "In this paper we investigate learning visual models for the steps of ordinary tasks using weak supervision via instructional narrations and an ordered list of steps instead of strong supervision via temporal annotations. At the heart of our approach is the observation that weakly supervised learning may be easier if a model shares components while learning different steps: ``pour egg'' should be trained jointly with other tasks involving ``pour'' and ``egg''. We formalize this in a component model for recognizing steps and a weakly supervised learning framework that can learn this model under temporal constraints from narration and the list of steps. Past data does not permit systematic studying of sharing and so we also gather a new dataset aimed at assessing cross-task sharing. Our experiments demonstrate that sharing across tasks improves performance, especially when done at the component level and that our component model can parse previously unseen tasks by virtue of its compositionality."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Towards Automatic Learning of Procedures From Web Instructional Videos",
                    "abstract": "\n \n The potential for agents, whether embodied or software, to learn by observing other agents performing procedures involving objects and actions is rich. Current research on automatic procedure learning heavily relies on action labels or video subtitles, even during the evaluation phase, which makes them infeasible in real-world scenarios. This leads to our question: can the human-consensus structure of a procedure be learned from a large set of long, unconstrained videos (e.g., instructional videos from YouTube) with only visual evidence? To answer this question, we introduce the problem of procedure segmentation---to segment a video procedure into category-independent procedure segments. Given that no large-scale dataset is available for this problem, we collect a large-scale procedure segmentation dataset with procedure segments temporally localized and described; we use cooking videos and name the dataset YouCook2. We propose a segment-level recurrent network for generating procedure segments by modeling the dependencies across segments. The generated segments can be used as pre-processing for other tasks, such as dense video captioning and event parsing. We show in our experiments that the proposed model outperforms competitive baselines in procedure segmentation.\n \n"
                },
                {
                    "title": "Mise en Place: Unsupervised Interpretation of Instructional Recipes",
                    "abstract": "We present an unsupervised hard EM approach to automatically mapping instructional recipes to action graphs, which define what actions should be performed on which objects and in what order. Recovering such structures can be challenging, due to unique properties of procedural language where, for example, verbal arguments are commonly elided when they can be inferred from context and disambiguation often requires world knowledge. Our probabilistic model incorporates aspects of procedural semantics and world knowledge, such as likely locations and selectional preferences for different actions. Experiments with cooking recipes demonstrate the ability to recover high quality action graphs, outperforming a strong sequential baseline by 8 points in F1, while also discovering general-purpose knowledge about cooking."
                },
                {
                    "title": "Temporal Perception and Prediction in Ego-Centric Video",
                    "abstract": "Given a video of an activity, can we predict what will happen next? In this paper we explore two simple tasks related to temporal prediction in egocentric videos of everyday activities. We provide both human experiments to understand how well people can perform on these tasks and computational models for prediction. Experiments indicate that humans and computers can do well on temporal prediction and that personalization to a particular individual or environment provides significantly increased performance. Developing methods for temporal prediction could have far reaching benefits for robots or intelligent agents to anticipate what a person will do, before they do it."
                },
                {
                    "title": "First-Person Vision",
                    "abstract": "For understanding the behavior, intent, and environment of a person, the surveillance metaphor is traditional; that is, install cameras and observe the subject, and his/her interaction with other people and the environment. Instead, we argue that the first-person vision (FPV), which senses the environment and the subject's activities from a wearable sensor, is more advantageous with images about the subject's environment as taken from his/her view points, and with readily available information about head motion and gaze through eye tracking. In this paper, we review key research challenges that need to be addressed to develop such FPV systems, and describe our ongoing work to address them using examples from our prototype systems."
                },
                {
                    "title": "Extraction of procedural knowledge from the web: a comparison of two workflow extraction approaches",
                    "abstract": "User generated Web content includes large amounts of procedural knowledge (also called how to knowledge). This paper is on a comparison of two extraction methods for procedural knowledge from the Web. Both methods create workflow representations automatically from text with the aim to reuse the Web experience by reasoning methods. Two variants of the workflow extraction process are introduced and evaluated by experiments with cooking recipes as a sample domain. The first variant is a term-based approach that integrates standard information extraction methods from the GATE system. The second variant is a frame-based approach that is implemented by means of the SUNDANCE system. The expert assessment of the extraction results clearly shows that the more sophisticated frame-based approach outperforms the term-based approach of automated workflow extraction."
                },
                {
                    "title": "Aligning Actions Across Recipe Graphs",
                    "abstract": "Recipe texts are an idiosyncratic form of instructional language that pose unique challenges for automatic understanding. One challenge is that a cooking step in one recipe can be explained in another recipe in different words, at a different level of abstraction, or not at all. Previous work has annotated correspondences between recipe instructions at the sentence level, often glossing over important correspondences between cooking steps across recipes. We present a novel and fully-parsed English recipe corpus, ARA (Aligned Recipe Actions), which annotates correspondences between individual actions across similar recipes with the goal of capturing information implicit for accurate recipe understanding. We represent this information in the form of recipe graphs, and we train a neural model for predicting correspondences on ARA. We find that substantial gains in accuracy can be obtained by taking fine-grained structural information about the recipes into account."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we automatically learn task graphs that represent the partial ordering of key-steps in procedural activities from user demonstrations in video format?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of artificial intelligence, particularly in creating intelligent agents that can assist users in performing complex tasks efficiently and accurately. By developing a method to automatically learn task graphs, we can enhance the interpretability and scalability of AI systems, allowing them to adapt to various procedural contexts without requiring extensive manual programming. This research could lead to significant advancements in human-computer interaction, robotics, and automation, ultimately improving the design of virtual assistants and other AI applications that rely on procedural knowledge.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of accurately capturing the dependencies and partial orderings of key-steps in diverse procedural activities. Naive approaches may fail because they do not account for the nuances of human behavior and the variability in how tasks are performed. Additionally, existing methods often rely on handcrafted graph mining procedures, which can be rigid and may not generalize well across different tasks. Overcoming these technical obstacles requires developing a robust learning framework that can effectively infer task graphs from noisy and unstructured video data, while also ensuring that the learned representations are interpretable and usable in downstream applications.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on specific tasks related to procedural knowledge, such as action anticipation and mistake detection, without establishing a comprehensive framework for learning explicit representations of procedures. The limitations of existing solutions stem from their reliance on meticulously crafted graph mining techniques, which do not leverage the potential of end-to-end learning systems. Additionally, the lack of scalable methods for automatically generating task graphs from video demonstrations has hindered progress in this area. Our approach differs by integrating the learning of task graphs into a unified framework that directly supervises the generation of adjacency matrices, allowing for a more flexible and effective representation of procedural knowledge.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves learning task graphs from sequences of key-steps demonstrated by users in video format. We will utilize a directed graph represented as an adjacency matrix, where the entries of the matrix are optimized using a Task Graph Maximum Likelihood (TGML) loss function. This loss function will supervise the learning process by"
            }
        },
        "author_data": {
            "d17e9055-a191-44a9-9bdf-c8f33496991a": {
                "pk": "d17e9055-a191-44a9-9bdf-c8f33496991a",
                "name": "Luigi Seminara",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "18920bcd-e9a8-4da7-b96d-5735bd200c24": {
                "pk": "18920bcd-e9a8-4da7-b96d-5735bd200c24",
                "name": "Giovanni Maria Farinella",
                "collaborators": [
                    "Antonino Furnari",
                    "Francesco Ragusa",
                    "Sebastiano Battiato",
                    "Giovanni Signorello",
                    "Michele Mazzamuto",
                    "Giovanni Pasqualino",
                    "Matteo Dunnhofer",
                    "Christian Micheloni",
                    "Kristen Grauman",
                    "Rosario Leonardi"
                ],
                "domain": [
                    "Egocentric Vision",
                    "Object Detection",
                    "Action Anticipation",
                    "Domain Adaptation"
                ],
                "publications": [
                    {
                        "title": "Egocentric Visitors Localization in Cultural Sites",
                        "abstract": "We consider the problem of localizing visitors in a cultural site from egocentric (first person) images. Localization information can be useful both to assist the user during his visit (e.g., by suggesting where to go and what to see next) and to provide behavioral information to the manager of the cultural site (e.g., how much time has been spent by visitors at a given location? What has been liked most?). To tackle the problem, we collected a large dataset of egocentric videos using two cameras: a head-mounted HoloLens device and a chest-mounted GoPro. Each frame has been labeled according to the location of the visitor and to what he was looking at. The dataset is freely available in order to encourage research in this domain. The dataset is complemented with baseline experiments performed considering a state-of-the-art method for location-based temporal segmentation of egocentric videos. Experiments show that compelling results can be achieved to extract useful information for both the visitor and the site-manager."
                    },
                    {
                        "title": "Weakly Supervised Attended Object Detection Using Gaze Data as Annotations",
                        "abstract": "We consider the problem of detecting and recognizing the objects observed by visitors (i.e., attended objects) in cultural sites from egocentric vision. A standard approach to the problem involves detecting all objects and selecting the one which best overlaps with the gaze of the visitor, measured through a gaze tracker. Since labeling large amounts of data to train a standard object detector is expensive in terms of costs and time, we propose a weakly supervised version of the task which leans only on gaze data and a frame-level label indicating the class of the attended object. To study the problem, we present a new dataset composed of egocentric videos and gaze coordinates of subjects visiting a museum. We hence compare three different baselines for weakly supervised attended object detection on the collected data. Results show that the considered approaches achieve satisfactory performance in a weakly supervised manner, which allows for significant time savings with respect to a fully supervised detector based on Faster R-CNN. To encourage research on the topic, we publicly release the code and the dataset at the following url: https://iplab.dmi.unict.it/WS_OBJ_DET/"
                    },
                    {
                        "title": "An Unsupervised Domain Adaptation Scheme for Single-Stage Artwork Recognition in Cultural Sites",
                        "abstract": "Recognizing artworks in a cultural site using images acquired from the user's point of view (First Person Vision) allows to build interesting applications for both the visitors and the site managers. However, current object detection algorithms working in fully supervised settings need to be trained with large quantities of labeled data, whose collection requires a lot of times and high costs in order to achieve good performance. Using synthetic data generated from the 3D model of the cultural site to train the algorithms can reduce these costs. On the other hand, when these models are tested with real images, a significant drop in performance is observed due to the differences between real and synthetic images. In this study we consider the problem of Unsupervised Domain Adaptation for object detection in cultural sites. To address this problem, we created a new dataset containing both synthetic and real images of 16 different artworks. We hence investigated different domain adaptation techniques based on one-stage and two-stage object detector, image-to-image translation and feature alignment. Based on the observation that single-stage detectors are more robust to the domain shift in the considered settings, we proposed a new method which builds on RetinaNet and feature alignment that we called DA-RetinaNet. The proposed approach achieves better results than compared methods on the proposed dataset and on Cityscapes. To support research in this field we release the dataset at the following link https://iplab.dmi.unict.it/EGO-CH-OBJ-UDA/ and the code of the proposed architecture at https://github.com/fpv-iplab/DA-RetinaNet."
                    },
                    {
                        "title": "What Would You Expect? Anticipating Egocentric Actions with Rolling-Unrolling LSTMs and Modality Attention",
                        "abstract": "Egocentric action anticipation consists in understanding which objects the camera wearer will interact with in the near future and which actions they will perform. We tackle the problem proposing an architecture able to anticipate actions at multiple temporal scales using two LSTMs to 1) summarize the past, and 2) formulate predictions about the future. The input video is processed considering three complimentary modalities: appearance (RGB), motion (optical flow) and objects (object-based features). Modality-specific predictions are fused using a novel Modality ATTention (MATT) mechanism which learns to weigh modalities in an adaptive fashion. Extensive evaluations on two large-scale benchmark datasets show that our method outperforms prior art by up to +7% on the challenging EPIC-Kitchens dataset including more than 2500 actions, and generalizes to EGTEA Gaze+. Our approach is also shown to generalize to the tasks of early action recognition and action recognition. Our method is ranked first in the public leaderboard of the EPIC-Kitchens egocentric action anticipation challenge 2019. Please see our web pages for code and examples: http://iplab.dmi.unict.it/rulstm - https://github.com/fpv-iplab/rulstm."
                    },
                    {
                        "title": "A Multi Camera Unsupervised Domain Adaptation Pipeline for Object Detection in Cultural Sites through Adversarial Learning and Self-Training",
                        "abstract": "Object detection algorithms allow to enable many interesting applications which can be implemented in different devices, such as smartphones and wearable devices. In the context of a cultural site, implementing these algorithms in a wearable device, such as a pair of smart glasses, allow to enable the use of augmented reality (AR) to show extra information about the artworks and enrich the visitors' experience during their tour. However, object detection algorithms require to be trained on many well annotated examples to achieve reasonable results. This brings a major limitation since the annotation process requires human supervision which makes it expensive in terms of time and costs. A possible solution to reduce these costs consist in exploiting tools to automatically generate synthetic labeled images from a 3D model of the site. However, models trained with synthetic data do not generalize on real images acquired in the target scenario in which they are supposed to be used. Furthermore, object detectors should be able to work with different wearable devices or different mobile devices, which makes generalization even harder. In this paper, we present a new dataset collected in a cultural site to study the problem of domain adaptation for object detection in the presence of multiple unlabeled target domains corresponding to different cameras and a labeled source domain obtained considering synthetic images for training purposes. We present a new domain adaptation method which outperforms current state-of-the-art approaches combining the benefits of aligning the domains at the feature and pixel level with a self-training process. We release the dataset at the following link https://iplab.dmi.unict.it/OBJ-MDA/ and the code of the proposed architecture at https://github.com/fpv-iplab/STMDA-RetinaNet."
                    },
                    {
                        "title": "EGO-CH: Dataset and Fundamental Tasks for Visitors BehavioralUnderstanding using Egocentric Vision",
                        "abstract": "Equipping visitors of a cultural site with a wearable device allows to easily collect information about their preferences which can be exploited to improve the fruition of cultural goods with augmented reality. Moreover, egocentric video can be processed using computer vision and machine learning to enable an automated analysis of visitors' behavior. The inferred information can be used both online to assist the visitor and offline to support the manager of the site. Despite the positive impact such technologies can have in cultural heritage, the topic is currently understudied due to the limited number of public datasets suitable to study the considered problems. To address this issue, in this paper we propose EGOcentric-Cultural Heritage (EGO-CH), the first dataset of egocentric videos for visitors' behavior understanding in cultural sites. The dataset has been collected in two cultural sites and includes more than $27$ hours of video acquired by $70$ subjects, with labels for $26$ environments and over $200$ different Points of Interest. A large subset of the dataset, consisting of $60$ videos, is associated with surveys filled out by real visitors. To encourage research on the topic, we propose $4$ challenging tasks (room-based localization, point of interest/object recognition, object retrieval and survey prediction) useful to understand visitors' behavior and report baseline results on the dataset."
                    },
                    {
                        "title": "Rolling-Unrolling LSTMs for Action Anticipation from First-Person Video",
                        "abstract": "In this paper, we tackle the problem of egocentric action anticipation, i.e., predicting what actions the camera wearer will perform in the near future and which objects they will interact with. Specifically, we contribute Rolling-Unrolling LSTM, a learning architecture to anticipate actions from egocentric videos. The method is based on three components: 1) an architecture comprised of two LSTMs to model the sub-tasks of summarizing the past and inferring the future, 2) a Sequence Completion Pre-Training technique which encourages the LSTMs to focus on the different sub-tasks, and 3) a Modality ATTention (MATT) mechanism to efficiently fuse multi-modal predictions performed by processing RGB frames, optical flow fields and object-based features. The proposed approach is validated on EPIC-Kitchens, EGTEA Gaze+ and ActivityNet. The experiments show that the proposed architecture is state-of-the-art in the domain of egocentric videos, achieving top performances in the 2019 EPIC-Kitchens egocentric action anticipation challenge. The approach also achieves competitive performance on ActivityNet with respect to methods not based on unsupervised pre-training and generalizes to the tasks of early action recognition and action recognition. To encourage research on this challenging topic, we made our code, trained models, and pre-extracted features available at our web page: http://iplab.dmi.unict.it/rulstm."
                    },
                    {
                        "title": "Towards Streaming Egocentric Action Anticipation",
                        "abstract": "Egocentric action anticipation is the task of predicting the future actions a camera wearer will likely perform based on past video observations. While in a real-world system it is fundamental to output such predictions before the action begins, past works have not generally paid attention to model runtime during evaluation. Indeed, current evaluation schemes assume that predictions can be made offline, and hence that computational resources are not limited. In contrast, in this paper, we propose a \"streaming\" egocentric action anticipation evaluation protocol which explicitly considers model runtime for performance assessment, assuming that predictions will be available only after the current video segment is processed, which depends on the processing time of a method. Following the proposed evaluation scheme, we benchmark different state-of-the-art approaches for egocentric action anticipation on two popular datasets. Our analysis shows that models with a smaller runtime tend to outperform heavier models in the considered streaming scenario, thus changing the rankings generally observed in standard offline evaluations. Based on this observation, we propose a lightweight action anticipation model consisting in a simple feed-forward 3D CNN, which we propose to optimize using knowledge distillation techniques and a custom loss. The results show that the proposed approach outperforms prior art in the streaming scenario, also in combination with other lightweight models."
                    },
                    {
                        "title": "Streaming egocentric action anticipation: An evaluation scheme and approach",
                        "abstract": "Egocentric action anticipation aims to predict the future actions the camera wearer will perform from the observation of the past. While predictions about the future should be available before the predicted events take place, most approaches do not pay attention to the computational time required to make such predictions. As a result, current evaluation schemes assume that predictions are available right after the input video is observed, i.e., presuming a negligible runtime, which may lead to overly optimistic evaluations. We propose a streaming egocentric action evaluation scheme which assumes that predictions are performed online and made available only after the model has processed the current input segment, which depends on its runtime. To evaluate all models considering the same prediction horizon, we hence propose that slower models should base their predictions on temporal segments sampled ahead of time. Based on the observation that model runtime can affect performance in the considered streaming evaluation scenario, we further propose a lightweight action anticipation model based on feed-forward 3D CNNs which is optimized using knowledge distillation techniques with a novel past-to-future distillation loss. Experiments on the three popular datasets EPIC-KITCHENS-55, EPIC-KITCHENS-100 and EGTEA Gaze+ show that (i) the proposed evaluation scheme induces a different ranking on state-of-the-art methods as compared to classic evaluations, (ii) lightweight approaches tend to outmatch more computationally expensive ones, and (iii) the proposed model based on feed-forward 3D CNNs and knowledge distillation outperforms current art in the streaming egocentric action anticipation scenario."
                    },
                    {
                        "title": "Eyes Wide Unshut: Unsupervised Mistake Detection in Egocentric Procedural Video by Detecting Unpredictable Gaze",
                        "abstract": "In this paper, we address the challenge of unsupervised mistake detection in egocentric procedural video through the analysis of gaze signals. Traditional supervised mistake detection methods rely on manually labeled mistakes, and hence suffer from domain-dependence and scalability issues. We introduce an unsupervised method for detecting mistakes in videos of human activities, overcoming the challenges of domain-specific requirements and the need for annotated data. We postulate that, when a subject is making a mistake in the execution of a procedure, their attention patterns will deviate from normality. We hence propose to detect mistakes by comparing gaze trajectories predicted from input video with ground truth gaze signals collected through a gaze tracker. Since predicting gaze in video is characterized by high uncertainty, we propose a novel \\textit{gaze completion task}, which aims to predict gaze from visual observations and partial gaze trajectories. We further contribute a \\textit{gaze completion approach} based on a Gaze-Frame Correlation module to explicitly model the correlation between gaze information and each local visual token. Inconsistencies between the predicted and observed gaze trajectories act as an indicator for identifying mistakes. Experiments on the EPIC-Tent, HoloAssist and IndustReal datasets showcase the effectiveness of the proposed approach as compared to unsupervised and one-class techniques. Our method is ranked first on the HoloAssist Mistake Detection challenge."
                    },
                    {
                        "title": "Next-Active-Object prediction from Egocentric Videos",
                        "abstract": "Although First Person Vision systems can sense the environment from the user's perspective, they are generally unable to predict his intentions and goals. Since human activities can be decomposed in terms of atomic actions and interactions with objects, intelligent wearable systems would benefit from the ability to anticipate user-object interactions. Even if this task is not trivial, the First Person Vision paradigm can provide important cues to address this challenge. We propose to exploit the dynamics of the scene to recognize next-active-objects before an object interaction begins. We train a classifier to discriminate trajectories leading to an object activation from all others and forecast next-active-objects by analyzing fixed-length trajectory segments within a temporal sliding window. The proposed method compares favorably with respect to several baselines on the Activity of Daily Living (ADL) egocentric dataset comprising 10 hours of videos acquired by 20 subjects while performing unconstrained interactions with several objects."
                    },
                    {
                        "title": "Is First Person Vision Challenging for Object Tracking?",
                        "abstract": "Understanding human-object interactions is fundamental in First Person Vision (FPV). Tracking algorithms which follow the objects manipulated by the camera wearer can provide useful cues to effectively model such interactions. Visual tracking solutions available in the computer vision literature have significantly improved their performance in the last years for a large variety of target objects and tracking scenarios. However, despite a few previous attempts to exploit trackers in FPV applications, a methodical analysis of the performance of state-of-the-art trackers in this domain is still missing. In this paper, we fill the gap by presenting the first systematic study of object tracking in FPV. Our study extensively analyses the performance of recent visual trackers and baseline FPV trackers with respect to different aspects and considering a new performance measure. This is achieved through TREK-150, a novel benchmark dataset composed of 150 densely annotated video sequences. Our results show that object tracking in FPV is challenging, which suggests that more research efforts should be devoted to this problem so that tracking could benefit FPV tasks."
                    },
                    {
                        "title": "Is First Person Vision Challenging for Object Tracking?",
                        "abstract": "Understanding human-object interactions is fundamental in First Person Vision (FPV). Tracking algorithms which follow the objects manipulated by the camera wearer can provide useful cues to effectively model such interactions. Despite a few previous attempts to exploit trackers in FPV applications, a methodical analysis of the performance of state-of-the-art visual trackers in this domain is still missing. In this short paper, we provide a recap of the first systematic study of object tracking in FPV. Our work extensively analyses the performance of recent and baseline FPV trackers with respect to different aspects. This is achieved through TREK-150, a novel benchmark dataset composed of 150 densely annotated video sequences. The results suggest that more research efforts should be devoted to this problem so that tracking could benefit FPV tasks. The full version of this paper is available at arXiv:2108.13665."
                    },
                    {
                        "title": "StillFast: An End-to-End Approach for Short-Term Object Interaction Anticipation",
                        "abstract": "Anticipation problem has been studied considering different aspects such as predicting humans' locations, predicting hands and objects trajectories, and forecasting actions and human-object interactions. In this paper, we studied the short-term object interaction anticipation problem from the egocentric point of view, proposing a new end-to-end architecture named StillFast. Our approach simultaneously processes a still image and a video detecting and localizing next-active objects, predicting the verb which describes the future interaction and determining when the interaction will start. Experiments on the large-scale egocentric dataset EGO4D show that our method outperformed state-of-the-art approaches on the considered task. Our method is ranked first in the public leaderboard of the EGO4D short term object interaction anticipation challenge 2022. Please see the project web page for code and additional details: https://iplab.dmi.unict.it/stillfast/."
                    },
                    {
                        "title": "Are Synthetic Data Useful for Egocentric Hand-Object Interaction Detection?",
                        "abstract": "In this study, we investigate the effectiveness of synthetic data in enhancing egocentric hand-object interaction detection. Via extensive experiments and comparative analyses on three egocentric datasets, VISOR, EgoHOS, and ENIGMA-51, our findings reveal how to exploit synthetic data for the HOI detection task when real labeled data are scarce or unavailable. Specifically, by leveraging only 10% of real labeled data, we achieve improvements in Overall AP compared to baselines trained exclusively on real data of: +5.67% on EPIC-KITCHENS VISOR, +8.24% on EgoHOS, and +11.69% on ENIGMA-51. Our analysis is supported by a novel data generation pipeline and the newly introduced HOI-Synth benchmark which augments existing datasets with synthetic images of hand-object interactions automatically labeled with hand-object contact states, bounding boxes, and pixel-wise segmentation masks. Data, code, and data generation tools to support future research are released at: https://fpv-iplab.github.io/HOI-Synth/."
                    },
                    {
                        "title": "Survey on Visual Sentiment Analysis",
                        "abstract": "Visual Sentiment Analysis aims to understand how images affect people, in terms of evoked emotions. Although this field is rather new, a broad range of techniques have been developed for various data sources and problems, resulting in a large body of research. This paper reviews pertinent publications and tries to present an exhaustive overview of the field. After a description of the task and the related applications, the subject is tackled under different main headings. The paper also describes principles of design of general Visual Sentiment Analysis systems from three main points of view: emotional models, dataset definition, feature design. A formalization of the problem is discussed, considering different levels of granularity, as well as the components that can affect the sentiment toward an image in different ways. To this aim, this paper considers a structured formalization of the problem which is usually used for the analysis of text, and discusses it's suitability in the context of Visual Sentiment Analysis. The paper also includes a description of new challenges, the evaluation from the viewpoint of progress toward more sophisticated systems and related practical applications, as well as a summary of the insights resulting from this study."
                    },
                    {
                        "title": "Predicting the Future from First Person (Egocentric) Vision: A Survey",
                        "abstract": "Egocentric videos can bring a lot of information about how humans perceive the world and interact with the environment, which can be beneficial for the analysis of human behaviour. The research in egocentric video analysis is developing rapidly thanks to the increasing availability of wearable devices and the opportunities offered by new large-scale egocentric datasets. As computer vision techniques continue to develop at an increasing pace, the tasks related to the prediction of future are starting to evolve from the need of understanding the present. Predicting future human activities, trajectories and interactions with objects is crucial in applications such as human-robot interaction, assistive wearable technologies for both industrial and daily living scenarios, entertainment and virtual or augmented reality. This survey summarises the evolution of studies in the context of future prediction from egocentric vision making an overview of applications, devices, existing problems, commonly used datasets, models and input modalities. Our analysis highlights that methods for future prediction from egocentric vision can have a significant impact in a range of applications and that further research efforts should be devoted to the standardisation of tasks and the proposal of datasets considering real-world scenarios such as the ones with an industrial vocation."
                    },
                    {
                        "title": "On Embodied Visual Navigation in Real Environments Through Habitat",
                        "abstract": "Visual navigation models based on deep learning can learn effective policies when trained on large amounts of visual observations through reinforcement learning. Unfortunately, collecting the required experience in the real world requires the deployment of a robotic platform, which is expensive and time-consuming. To deal with this limitation, several simulation platforms have been proposed in order to train visual navigation policies on virtual environments efficiently. Despite the advantages they offer, simulators present a limited realism in terms of appearance and physical dynamics, leading to navigation policies that do not generalize in the real world.   In this paper, we propose a tool based on the Habitat simulator which exploits real world images of the environment, together with sensor and actuator noise models, to produce more realistic navigation episodes. We perform a range of experiments to assess the ability of such policies to generalize using virtual and real-world images, as well as observations transformed with unsupervised domain adaptation approaches. We also assess the impact of sensor and actuation noise on the navigation performance and investigate whether it allows to learn more robust navigation policies. We show that our tool can effectively help to train and evaluate navigation policies on real-world observations without running navigation pisodes in the real world."
                    },
                    {
                        "title": "MECCANO: A Multimodal Egocentric Dataset for Humans Behavior Understanding in the Industrial-like Domain",
                        "abstract": "Wearable cameras allow to acquire images and videos from the user's perspective. These data can be processed to understand humans behavior. Despite human behavior analysis has been thoroughly investigated in third person vision, it is still understudied in egocentric settings and in particular in industrial scenarios. To encourage research in this field, we present MECCANO, a multimodal dataset of egocentric videos to study humans behavior understanding in industrial-like settings. The multimodality is characterized by the presence of gaze signals, depth maps and RGB videos acquired simultaneously with a custom headset. The dataset has been explicitly labeled for fundamental tasks in the context of human behavior understanding from a first person view, such as recognizing and anticipating human-object interactions. With the MECCANO dataset, we explored five different tasks including 1) Action Recognition, 2) Active Objects Detection and Recognition, 3) Egocentric Human-Objects Interaction Detection, 4) Action Anticipation and 5) Next-Active Objects Detection. We propose a benchmark aimed to study human behavior in the considered industrial-like scenario which demonstrates that the investigated tasks and the considered scenario are challenging for state-of-the-art algorithms. To support research in this field, we publicy release the dataset at https://iplab.dmi.unict.it/MECCANO/."
                    }
                ]
            },
            "02d5fb57-5add-450d-a4c6-6ce1066a31d7": {
                "pk": "02d5fb57-5add-450d-a4c6-6ce1066a31d7",
                "name": "Antonino Furnari",
                "collaborators": [
                    "Giovanni Maria Farinella",
                    "Francesco Ragusa",
                    "Sebastiano Battiato",
                    "Matteo Dunnhofer",
                    "Christian Micheloni",
                    "Rosario Leonardi",
                    "Kristen Grauman",
                    "Ivan Rodin",
                    "Dimitrios Mavroedis",
                    "Michele Mazzamuto"
                ],
                "domain": [
                    "Egocentric Vision",
                    "Action Anticipation",
                    "Human-Object Interaction",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "What Would You Expect? Anticipating Egocentric Actions with Rolling-Unrolling LSTMs and Modality Attention",
                        "abstract": "Egocentric action anticipation consists in understanding which objects the camera wearer will interact with in the near future and which actions they will perform. We tackle the problem proposing an architecture able to anticipate actions at multiple temporal scales using two LSTMs to 1) summarize the past, and 2) formulate predictions about the future. The input video is processed considering three complimentary modalities: appearance (RGB), motion (optical flow) and objects (object-based features). Modality-specific predictions are fused using a novel Modality ATTention (MATT) mechanism which learns to weigh modalities in an adaptive fashion. Extensive evaluations on two large-scale benchmark datasets show that our method outperforms prior art by up to +7% on the challenging EPIC-Kitchens dataset including more than 2500 actions, and generalizes to EGTEA Gaze+. Our approach is also shown to generalize to the tasks of early action recognition and action recognition. Our method is ranked first in the public leaderboard of the EPIC-Kitchens egocentric action anticipation challenge 2019. Please see our web pages for code and examples: http://iplab.dmi.unict.it/rulstm - https://github.com/fpv-iplab/rulstm."
                    },
                    {
                        "title": "Rolling-Unrolling LSTMs for Action Anticipation from First-Person Video",
                        "abstract": "In this paper, we tackle the problem of egocentric action anticipation, i.e., predicting what actions the camera wearer will perform in the near future and which objects they will interact with. Specifically, we contribute Rolling-Unrolling LSTM, a learning architecture to anticipate actions from egocentric videos. The method is based on three components: 1) an architecture comprised of two LSTMs to model the sub-tasks of summarizing the past and inferring the future, 2) a Sequence Completion Pre-Training technique which encourages the LSTMs to focus on the different sub-tasks, and 3) a Modality ATTention (MATT) mechanism to efficiently fuse multi-modal predictions performed by processing RGB frames, optical flow fields and object-based features. The proposed approach is validated on EPIC-Kitchens, EGTEA Gaze+ and ActivityNet. The experiments show that the proposed architecture is state-of-the-art in the domain of egocentric videos, achieving top performances in the 2019 EPIC-Kitchens egocentric action anticipation challenge. The approach also achieves competitive performance on ActivityNet with respect to methods not based on unsupervised pre-training and generalizes to the tasks of early action recognition and action recognition. To encourage research on this challenging topic, we made our code, trained models, and pre-extracted features available at our web page: http://iplab.dmi.unict.it/rulstm."
                    },
                    {
                        "title": "Towards Streaming Egocentric Action Anticipation",
                        "abstract": "Egocentric action anticipation is the task of predicting the future actions a camera wearer will likely perform based on past video observations. While in a real-world system it is fundamental to output such predictions before the action begins, past works have not generally paid attention to model runtime during evaluation. Indeed, current evaluation schemes assume that predictions can be made offline, and hence that computational resources are not limited. In contrast, in this paper, we propose a \"streaming\" egocentric action anticipation evaluation protocol which explicitly considers model runtime for performance assessment, assuming that predictions will be available only after the current video segment is processed, which depends on the processing time of a method. Following the proposed evaluation scheme, we benchmark different state-of-the-art approaches for egocentric action anticipation on two popular datasets. Our analysis shows that models with a smaller runtime tend to outperform heavier models in the considered streaming scenario, thus changing the rankings generally observed in standard offline evaluations. Based on this observation, we propose a lightweight action anticipation model consisting in a simple feed-forward 3D CNN, which we propose to optimize using knowledge distillation techniques and a custom loss. The results show that the proposed approach outperforms prior art in the streaming scenario, also in combination with other lightweight models."
                    },
                    {
                        "title": "Streaming egocentric action anticipation: An evaluation scheme and approach",
                        "abstract": "Egocentric action anticipation aims to predict the future actions the camera wearer will perform from the observation of the past. While predictions about the future should be available before the predicted events take place, most approaches do not pay attention to the computational time required to make such predictions. As a result, current evaluation schemes assume that predictions are available right after the input video is observed, i.e., presuming a negligible runtime, which may lead to overly optimistic evaluations. We propose a streaming egocentric action evaluation scheme which assumes that predictions are performed online and made available only after the model has processed the current input segment, which depends on its runtime. To evaluate all models considering the same prediction horizon, we hence propose that slower models should base their predictions on temporal segments sampled ahead of time. Based on the observation that model runtime can affect performance in the considered streaming evaluation scenario, we further propose a lightweight action anticipation model based on feed-forward 3D CNNs which is optimized using knowledge distillation techniques with a novel past-to-future distillation loss. Experiments on the three popular datasets EPIC-KITCHENS-55, EPIC-KITCHENS-100 and EGTEA Gaze+ show that (i) the proposed evaluation scheme induces a different ranking on state-of-the-art methods as compared to classic evaluations, (ii) lightweight approaches tend to outmatch more computationally expensive ones, and (iii) the proposed model based on feed-forward 3D CNNs and knowledge distillation outperforms current art in the streaming egocentric action anticipation scenario."
                    },
                    {
                        "title": "Next-Active-Object prediction from Egocentric Videos",
                        "abstract": "Although First Person Vision systems can sense the environment from the user's perspective, they are generally unable to predict his intentions and goals. Since human activities can be decomposed in terms of atomic actions and interactions with objects, intelligent wearable systems would benefit from the ability to anticipate user-object interactions. Even if this task is not trivial, the First Person Vision paradigm can provide important cues to address this challenge. We propose to exploit the dynamics of the scene to recognize next-active-objects before an object interaction begins. We train a classifier to discriminate trajectories leading to an object activation from all others and forecast next-active-objects by analyzing fixed-length trajectory segments within a temporal sliding window. The proposed method compares favorably with respect to several baselines on the Activity of Daily Living (ADL) egocentric dataset comprising 10 hours of videos acquired by 20 subjects while performing unconstrained interactions with several objects."
                    },
                    {
                        "title": "Predicting the Future from First Person (Egocentric) Vision: A Survey",
                        "abstract": "Egocentric videos can bring a lot of information about how humans perceive the world and interact with the environment, which can be beneficial for the analysis of human behaviour. The research in egocentric video analysis is developing rapidly thanks to the increasing availability of wearable devices and the opportunities offered by new large-scale egocentric datasets. As computer vision techniques continue to develop at an increasing pace, the tasks related to the prediction of future are starting to evolve from the need of understanding the present. Predicting future human activities, trajectories and interactions with objects is crucial in applications such as human-robot interaction, assistive wearable technologies for both industrial and daily living scenarios, entertainment and virtual or augmented reality. This survey summarises the evolution of studies in the context of future prediction from egocentric vision making an overview of applications, devices, existing problems, commonly used datasets, models and input modalities. Our analysis highlights that methods for future prediction from egocentric vision can have a significant impact in a range of applications and that further research efforts should be devoted to the standardisation of tasks and the proposal of datasets considering real-world scenarios such as the ones with an industrial vocation."
                    },
                    {
                        "title": "Eyes Wide Unshut: Unsupervised Mistake Detection in Egocentric Procedural Video by Detecting Unpredictable Gaze",
                        "abstract": "In this paper, we address the challenge of unsupervised mistake detection in egocentric procedural video through the analysis of gaze signals. Traditional supervised mistake detection methods rely on manually labeled mistakes, and hence suffer from domain-dependence and scalability issues. We introduce an unsupervised method for detecting mistakes in videos of human activities, overcoming the challenges of domain-specific requirements and the need for annotated data. We postulate that, when a subject is making a mistake in the execution of a procedure, their attention patterns will deviate from normality. We hence propose to detect mistakes by comparing gaze trajectories predicted from input video with ground truth gaze signals collected through a gaze tracker. Since predicting gaze in video is characterized by high uncertainty, we propose a novel \\textit{gaze completion task}, which aims to predict gaze from visual observations and partial gaze trajectories. We further contribute a \\textit{gaze completion approach} based on a Gaze-Frame Correlation module to explicitly model the correlation between gaze information and each local visual token. Inconsistencies between the predicted and observed gaze trajectories act as an indicator for identifying mistakes. Experiments on the EPIC-Tent, HoloAssist and IndustReal datasets showcase the effectiveness of the proposed approach as compared to unsupervised and one-class techniques. Our method is ranked first on the HoloAssist Mistake Detection challenge."
                    },
                    {
                        "title": "Egocentric Visitors Localization in Cultural Sites",
                        "abstract": "We consider the problem of localizing visitors in a cultural site from egocentric (first person) images. Localization information can be useful both to assist the user during his visit (e.g., by suggesting where to go and what to see next) and to provide behavioral information to the manager of the cultural site (e.g., how much time has been spent by visitors at a given location? What has been liked most?). To tackle the problem, we collected a large dataset of egocentric videos using two cameras: a head-mounted HoloLens device and a chest-mounted GoPro. Each frame has been labeled according to the location of the visitor and to what he was looking at. The dataset is freely available in order to encourage research in this domain. The dataset is complemented with baseline experiments performed considering a state-of-the-art method for location-based temporal segmentation of egocentric videos. Experiments show that compelling results can be achieved to extract useful information for both the visitor and the site-manager."
                    },
                    {
                        "title": "A Survey on Human-aware Robot Navigation",
                        "abstract": "Intelligent systems are increasingly part of our everyday lives and have been integrated seamlessly to the point where it is difficult to imagine a world without them. Physical manifestations of those systems on the other hand, in the form of embodied agents or robots, have so far been used only for specific applications and are often limited to functional roles (e.g. in the industry, entertainment and military fields). Given the current growth and innovation in the research communities concerned with the topics of robot navigation, human-robot-interaction and human activity recognition, it seems like this might soon change. Robots are increasingly easy to obtain and use and the acceptance of them in general is growing. However, the design of a socially compliant robot that can function as a companion needs to take various areas of research into account. This paper is concerned with the navigation aspect of a socially-compliant robot and provides a survey of existing solutions for the relevant areas of research as well as an outlook on possible future directions."
                    },
                    {
                        "title": "Is First Person Vision Challenging for Object Tracking?",
                        "abstract": "Understanding human-object interactions is fundamental in First Person Vision (FPV). Tracking algorithms which follow the objects manipulated by the camera wearer can provide useful cues to effectively model such interactions. Visual tracking solutions available in the computer vision literature have significantly improved their performance in the last years for a large variety of target objects and tracking scenarios. However, despite a few previous attempts to exploit trackers in FPV applications, a methodical analysis of the performance of state-of-the-art trackers in this domain is still missing. In this paper, we fill the gap by presenting the first systematic study of object tracking in FPV. Our study extensively analyses the performance of recent visual trackers and baseline FPV trackers with respect to different aspects and considering a new performance measure. This is achieved through TREK-150, a novel benchmark dataset composed of 150 densely annotated video sequences. Our results show that object tracking in FPV is challenging, which suggests that more research efforts should be devoted to this problem so that tracking could benefit FPV tasks."
                    },
                    {
                        "title": "Is First Person Vision Challenging for Object Tracking?",
                        "abstract": "Understanding human-object interactions is fundamental in First Person Vision (FPV). Tracking algorithms which follow the objects manipulated by the camera wearer can provide useful cues to effectively model such interactions. Despite a few previous attempts to exploit trackers in FPV applications, a methodical analysis of the performance of state-of-the-art visual trackers in this domain is still missing. In this short paper, we provide a recap of the first systematic study of object tracking in FPV. Our work extensively analyses the performance of recent and baseline FPV trackers with respect to different aspects. This is achieved through TREK-150, a novel benchmark dataset composed of 150 densely annotated video sequences. The results suggest that more research efforts should be devoted to this problem so that tracking could benefit FPV tasks. The full version of this paper is available at arXiv:2108.13665."
                    },
                    {
                        "title": "StillFast: An End-to-End Approach for Short-Term Object Interaction Anticipation",
                        "abstract": "Anticipation problem has been studied considering different aspects such as predicting humans' locations, predicting hands and objects trajectories, and forecasting actions and human-object interactions. In this paper, we studied the short-term object interaction anticipation problem from the egocentric point of view, proposing a new end-to-end architecture named StillFast. Our approach simultaneously processes a still image and a video detecting and localizing next-active objects, predicting the verb which describes the future interaction and determining when the interaction will start. Experiments on the large-scale egocentric dataset EGO4D show that our method outperformed state-of-the-art approaches on the considered task. Our method is ranked first in the public leaderboard of the EGO4D short term object interaction anticipation challenge 2022. Please see the project web page for code and additional details: https://iplab.dmi.unict.it/stillfast/."
                    },
                    {
                        "title": "Are Synthetic Data Useful for Egocentric Hand-Object Interaction Detection?",
                        "abstract": "In this study, we investigate the effectiveness of synthetic data in enhancing egocentric hand-object interaction detection. Via extensive experiments and comparative analyses on three egocentric datasets, VISOR, EgoHOS, and ENIGMA-51, our findings reveal how to exploit synthetic data for the HOI detection task when real labeled data are scarce or unavailable. Specifically, by leveraging only 10% of real labeled data, we achieve improvements in Overall AP compared to baselines trained exclusively on real data of: +5.67% on EPIC-KITCHENS VISOR, +8.24% on EgoHOS, and +11.69% on ENIGMA-51. Our analysis is supported by a novel data generation pipeline and the newly introduced HOI-Synth benchmark which augments existing datasets with synthetic images of hand-object interactions automatically labeled with hand-object contact states, bounding boxes, and pixel-wise segmentation masks. Data, code, and data generation tools to support future research are released at: https://fpv-iplab.github.io/HOI-Synth/."
                    },
                    {
                        "title": "On Embodied Visual Navigation in Real Environments Through Habitat",
                        "abstract": "Visual navigation models based on deep learning can learn effective policies when trained on large amounts of visual observations through reinforcement learning. Unfortunately, collecting the required experience in the real world requires the deployment of a robotic platform, which is expensive and time-consuming. To deal with this limitation, several simulation platforms have been proposed in order to train visual navigation policies on virtual environments efficiently. Despite the advantages they offer, simulators present a limited realism in terms of appearance and physical dynamics, leading to navigation policies that do not generalize in the real world.   In this paper, we propose a tool based on the Habitat simulator which exploits real world images of the environment, together with sensor and actuator noise models, to produce more realistic navigation episodes. We perform a range of experiments to assess the ability of such policies to generalize using virtual and real-world images, as well as observations transformed with unsupervised domain adaptation approaches. We also assess the impact of sensor and actuation noise on the navigation performance and investigate whether it allows to learn more robust navigation policies. We show that our tool can effectively help to train and evaluate navigation policies on real-world observations without running navigation pisodes in the real world."
                    },
                    {
                        "title": "MECCANO: A Multimodal Egocentric Dataset for Humans Behavior Understanding in the Industrial-like Domain",
                        "abstract": "Wearable cameras allow to acquire images and videos from the user's perspective. These data can be processed to understand humans behavior. Despite human behavior analysis has been thoroughly investigated in third person vision, it is still understudied in egocentric settings and in particular in industrial scenarios. To encourage research in this field, we present MECCANO, a multimodal dataset of egocentric videos to study humans behavior understanding in industrial-like settings. The multimodality is characterized by the presence of gaze signals, depth maps and RGB videos acquired simultaneously with a custom headset. The dataset has been explicitly labeled for fundamental tasks in the context of human behavior understanding from a first person view, such as recognizing and anticipating human-object interactions. With the MECCANO dataset, we explored five different tasks including 1) Action Recognition, 2) Active Objects Detection and Recognition, 3) Egocentric Human-Objects Interaction Detection, 4) Action Anticipation and 5) Next-Active Objects Detection. We propose a benchmark aimed to study human behavior in the considered industrial-like scenario which demonstrates that the investigated tasks and the considered scenario are challenging for state-of-the-art algorithms. To support research in this field, we publicy release the dataset at https://iplab.dmi.unict.it/MECCANO/."
                    },
                    {
                        "title": "Visual Object Tracking in First Person Vision",
                        "abstract": "The understanding of human-object interactions is fundamental in First Person Vision (FPV). Visual tracking algorithms which follow the objects manipulated by the camera wearer can provide useful information to effectively model such interactions. In the last years, the computer vision community has significantly improved the performance of tracking algorithms for a large variety of target objects and scenarios. Despite a few previous attempts to exploit trackers in the FPV domain, a methodical analysis of the performance of state-of-the-art trackers is still missing. This research gap raises the question of whether current solutions can be used ``off-the-shelf'' or more domain-specific investigations should be carried out. This paper aims to provide answers to such questions. We present the first systematic investigation of single object tracking in FPV. Our study extensively analyses the performance of 42 algorithms including generic object trackers and baseline FPV-specific trackers. The analysis is carried out by focusing on different aspects of the FPV setting, introducing new performance measures, and in relation to FPV-specific tasks. The study is made possible through the introduction of TREK-150, a novel benchmark dataset composed of 150 densely annotated video sequences. Our results show that object tracking in FPV poses new challenges to current visual trackers. We highlight the factors causing such behavior and point out possible research directions. Despite their difficulties, we prove that trackers bring benefits to FPV downstream tasks requiring short-term object tracking. We expect that generic object tracking will gain popularity in FPV as new and FPV-specific methodologies are investigated."
                    },
                    {
                        "title": "Exploiting Multimodal Synthetic Data for Egocentric Human-Object Interaction Detection in an Industrial Scenario",
                        "abstract": "In this paper, we tackle the problem of Egocentric Human-Object Interaction (EHOI) detection in an industrial setting. To overcome the lack of public datasets in this context, we propose a pipeline and a tool for generating synthetic images of EHOIs paired with several annotations and data signals (e.g., depth maps or segmentation masks). Using the proposed pipeline, we present EgoISM-HOI a new multimodal dataset composed of synthetic EHOI images in an industrial environment with rich annotations of hands and objects. To demonstrate the utility and effectiveness of synthetic EHOI data produced by the proposed tool, we designed a new method that predicts and combines different multimodal signals to detect EHOIs in RGB images. Our study shows that exploiting synthetic data to pre-train the proposed method significantly improves performance when tested on real-world data. Moreover, to fully understand the usefulness of our method, we conducted an in-depth analysis in which we compared and highlighted the superiority of the proposed approach over different state-of-the-art class-agnostic methods. To support research in this field, we publicly release the datasets, source code, and pre-trained models at https://iplab.dmi.unict.it/egoism-hoi."
                    },
                    {
                        "title": "A Multi Camera Unsupervised Domain Adaptation Pipeline for Object Detection in Cultural Sites through Adversarial Learning and Self-Training",
                        "abstract": "Object detection algorithms allow to enable many interesting applications which can be implemented in different devices, such as smartphones and wearable devices. In the context of a cultural site, implementing these algorithms in a wearable device, such as a pair of smart glasses, allow to enable the use of augmented reality (AR) to show extra information about the artworks and enrich the visitors' experience during their tour. However, object detection algorithms require to be trained on many well annotated examples to achieve reasonable results. This brings a major limitation since the annotation process requires human supervision which makes it expensive in terms of time and costs. A possible solution to reduce these costs consist in exploiting tools to automatically generate synthetic labeled images from a 3D model of the site. However, models trained with synthetic data do not generalize on real images acquired in the target scenario in which they are supposed to be used. Furthermore, object detectors should be able to work with different wearable devices or different mobile devices, which makes generalization even harder. In this paper, we present a new dataset collected in a cultural site to study the problem of domain adaptation for object detection in the presence of multiple unlabeled target domains corresponding to different cameras and a labeled source domain obtained considering synthetic images for training purposes. We present a new domain adaptation method which outperforms current state-of-the-art approaches combining the benefits of aligning the domains at the feature and pixel level with a self-training process. We release the dataset at the following link https://iplab.dmi.unict.it/OBJ-MDA/ and the code of the proposed architecture at https://github.com/fpv-iplab/STMDA-RetinaNet."
                    },
                    {
                        "title": "The MECCANO Dataset: Understanding Human-Object Interactions from Egocentric Videos in an Industrial-like Domain",
                        "abstract": "Wearable cameras allow to collect images and videos of humans interacting with the world. While human-object interactions have been thoroughly investigated in third person vision, the problem has been understudied in egocentric settings and in industrial scenarios. To fill this gap, we introduce MECCANO, the first dataset of egocentric videos to study human-object interactions in industrial-like settings. MECCANO has been acquired by 20 participants who were asked to build a motorbike model, for which they had to interact with tiny objects and tools. The dataset has been explicitly labeled for the task of recognizing human-object interactions from an egocentric perspective. Specifically, each interaction has been labeled both temporally (with action segments) and spatially (with active object bounding boxes). With the proposed dataset, we investigate four different tasks including 1) action recognition, 2) active object detection, 3) active object recognition and 4) egocentric human-object interaction detection, which is a revisited version of the standard human-object interaction detection task. Baseline results show that the MECCANO dataset is a challenging benchmark to study egocentric human-object interactions in industrial-like scenarios. We publicy release the dataset at https://iplab.dmi.unict.it/MECCANO."
                    }
                ]
            }
        }
    },
    "2310.01636": {
        "paper_data": {
            "title": "Adaptive Visual Scene Understanding: Incremental Scene Graph Generation",
            "url": "http://arxiv.org/abs/2310.01636v3",
            "arxiv_id": "2310.01636",
            "authors": [
                "Naitik Khandelwal",
                "Xiao Liu",
                "Mengmi Zhang"
            ],
            "abstract": "Scene graph generation (SGG) involves analyzing images to extract meaningful information about objects and their relationships. Given the dynamic nature of the visual world, it becomes crucial for AI systems to detect new objects and establish their new relationships with existing objects. To address the lack of continual learning methodologies in SGG, we introduce the comprehensive Continual ScenE Graph Generation (CSEGG) dataset along with 3 learning scenarios and 8 evaluation metrics. Our research investigates the continual learning performances of existing SGG methods on the retention of previous object entities and relationships as they learn new ones. Moreover, we also explore how continual object detection enhances generalization in classifying known relationships on unknown objects. We conduct extensive experiments benchmarking and analyzing the classical two-stage SGG methods and the most recent transformer-based SGG methods in continual learning settings, and gain valuable insights into the CSEGG problem. We invite the research community to explore this emerging field of study.",
            "introduction": "   1 Introduction  Scene graph generation (SGG) aims to extract object entities and their relationships in a scene. The resulting scene graph, carrying semantic scene structures, can be used for a variety of downstream tasks such as object detection[55], image captioning [19, 1] , and visual question answering [16]. Despite the notable advancements in SGG, current works have largely overlooked the critical aspect of continual learning. In the dynamic visual world, new objects and relationships are introduced incrementally, posing challenges for SGG models to account for new changes without forgetting previously acquired knowledge. This problem of Continual ScenE Graph Generation (CSEGG) holds great potential for various applications, such as real-time robotic navigation in dynamic environments and adaptive augmented reality experiences.             Figure 1: (a) A scene graph is a graph structure, where objects are represented as nodes (red boxes), and the relationships between objects are represented as edges connecting the corresponding nodes (green boxes). Each node in the graph contains information such as the object\u2019s class label, and spatial location. The edges in the graph indicate the relationships between objects, often described by predicates. A scene graph can be parsed into a set of triplets, consisting of three components: a subject, a relationship predicate, and an object that serves as the target or object of the relationship. The graph allows for a compact and structured representation of the objects and their relationships within a visual scene. (b) An example CSEGG application is presented, where a robot continuously encounters new objects (blue) and new relationships (yellow) over time across new scenes.   The field of continual learning has witnessed significant growth in recent years, with a major focus on tasks such as image classification [37], object detection [58], and visual question answering [26]. However, these endeavors have largely neglected the distinctive complexities associated with CSEGG. Here, we highlight several unique challenges of CSEGG: (1) In contrast to object detection, SGG involves understanding and capturing the relationships between objects, which can be intricate and diverse. Consequently, in CSEGG, conveying the spatial and semantic relationships between objects demands adaptive reasoning from the dynamic scene. (2) SGG introduces a higher level of combinatorial complexity than object detection and image classification because each detected object pair may have multiple potential spatial and functional relationships. Thus, as new objects are introduced to the scenes, the complexity of relationships among all the objects increases significantly in a non-linear fashion. (3) The long-tailed distribution in SGG can be attributed to the inherent characteristics of real-world scenes, where certain objects are more prevalent than others. Consequently, CSEGG requires the computational models to adapt continually to the evolving long-tailed distributions over different scenes. Due to a scarcity of research specifically addressing these challenges of CSEGG, there is a pressing need for specialized investigations and methodologies to enable computational models with the ability of CSEGG.   In this study, we re-organize existing SGG datasets [23, 25] to establish a novel and comprehensive CSEGG benchmark with 3 learning protocols as shown in Fig. 2. (S1). Relationship-incremental setting: an SGG agent learns to recognize new relationships among familiar objects within the same scene. (S2). Scene-incremental",
            "references": [
                {
                    "title": "Learning to Learn: How to Continuously Teach Humans and Machines",
                    "abstract": "Curriculum design is a fundamental component of education. For example, when we learn mathematics at school, we build upon our knowledge of addition to learn multiplication. These and other concepts must be mastered before our first algebra lesson, which also reinforces our addition and multiplication skills. Designing a curriculum for teaching either a human or a machine shares the underlying goal of maximizing knowledge transfer from earlier to later tasks, while also minimizing forgetting of learned tasks. Prior research on curriculum design for image classification focuses on the ordering of training examples during a single offline task. Here, we investigate the effect of the order in which multiple distinct tasks are learned in a sequence. We focus on the online class-incremental continual learning setting, where algorithms or humans must learn image classes one at a time during a single pass through a dataset. We find that curriculum consistently influences learning outcomes for humans and for multiple continual machine learning algorithms across several benchmark datasets. We introduce a novel-object recognition dataset for human curriculum learning experiments and observe that curricula that are effective for humans are highly correlated with those that are effective for machines. As an initial step towards automated curriculum design for online class-incremental learning, we propose a novel algorithm, dubbed Curriculum Designer (CD), that designs and ranks curricula based on inter-class feature similarities. We find significant overlap between curricula that are empirically highly effective and those that are highly ranked by our CD. Our study establishes a framework for further research on teaching humans and machines to learn continuously using optimized curricula. Our code and data are available through this link."
                },
                {
                    "title": "Beyond Not-Forgetting: Continual Learning with Backward Knowledge Transfer",
                    "abstract": "By learning a sequence of tasks continually, an agent in continual learning (CL) can improve the learning performance of both a new task and `old' tasks by leveraging the forward knowledge transfer and the backward knowledge transfer, respectively. However, most existing CL methods focus on addressing catastrophic forgetting in neural networks by minimizing the modification of the learnt model for old tasks. This inevitably limits the backward knowledge transfer from the new task to the old tasks, because judicious model updates could possibly improve the learning performance of the old tasks as well. To tackle this problem, we first theoretically analyze the conditions under which updating the learnt model of old tasks could be beneficial for CL and also lead to backward knowledge transfer, based on the gradient projection onto the input subspaces of old tasks. Building on the theoretical analysis, we next develop a ContinUal learning method with Backward knowlEdge tRansfer (CUBER), for a fixed capacity neural network without data replay. In particular, CUBER first characterizes the task correlation to identify the positively correlated old tasks in a layer-wise manner, and then selectively modifies the learnt model of the old tasks when learning the new task. Experimental studies show that CUBER can even achieve positive backward knowledge transfer on several existing CL benchmarks for the first time without data replay, where the related baselines still suffer from catastrophic forgetting (negative backward knowledge transfer). The superior performance of CUBER on the backward knowledge transfer also leads to higher accuracy accordingly."
                },
                {
                    "title": "Symbolic Replay: Scene Graph as Prompt for Continual Learning on VQA Task",
                    "abstract": "VQA is an ambitious task aiming to answer any image-related question. However, in reality, it is hard to build such a system once for all since the needs of users are continuously updated, and the system has to implement new functions. Thus, Continual Learning (CL) ability is a must in developing advanced VQA systems. Recently, a pioneer work split a VQA dataset into disjoint answer sets to study this topic. However, CL on VQA involves not only the expansion of label sets (new Answer sets). It is crucial to study how to answer questions when deploying VQA systems to new environments (new Visual scenes) and how to answer questions requiring new functions (new Question types). Thus, we propose CLOVE, a benchmark for Continual Learning On Visual quEstion answering, which contains scene- and function-incremental settings for the two aforementioned CL scenarios. In terms of methodology, the main difference between CL on VQA and classification is that the former additionally involves expanding and preventing forgetting of reasoning mechanisms, while the latter focusing on class representation. Thus, we propose a real-data-free replay-based method tailored for CL on VQA, named Scene Graph as Prompt for Symbolic Replay. Using a piece of scene graph as a prompt, it replays pseudo scene graphs to represent the past images, along with correlated QA pairs. A unified VQA model is also proposed to utilize the current and replayed data to enhance its QA ability. Finally, experimental results reveal challenges in CLOVE and demonstrate the effectiveness of our method. Code and data\nare available at https://github.com/showlab/CLVQA."
                },
                {
                    "title": "Continual Object Detection: A review of definitions, strategies, and challenges",
                    "abstract": "The field of Continual Learning investigates the ability to learn consecutive tasks without losing performance on those previously learned. The efforts of researchers have been mainly focused on incremental classification tasks. Yet, we believe that continual object detection deserves even more attention due to its vast range of applications in robotics and autonomous vehicles. This scenario is also more complex than conventional classification, given the occurrence of instances of classes that are unknown at the time but can appear in subsequent tasks as a new class to be learned, resulting in missing annotations and conflicts with the background label. In this review, we analyze the current strategies proposed to tackle the problem of class-incremental object detection. Our main contributions are: (1) a short and systematic review of the methods that propose solutions to traditional incremental object detection scenarios; (2) A comprehensive evaluation of the existing approaches using a new metric to quantify the stability and plasticity of each technique in a standard way; (3) an overview of the current trends within continual object detection and a discussion of possible future research directions."
                },
                {
                    "title": "FOSTER: Feature Boosting and Compression for Class-Incremental Learning",
                    "abstract": "The ability to learn new concepts continually is necessary in this ever-changing world. However, deep neural networks suffer from catastrophic forgetting when learning new categories. Many works have been proposed to alleviate this phenomenon, whereas most of them either fall into the stability-plasticity dilemma or take too much computation or storage overhead. Inspired by the gradient boosting algorithm to gradually fit the residuals between the target model and the previous ensemble model, we propose a novel two-stage learning paradigm FOSTER, empowering the model to learn new categories adaptively. Specifically, we first dynamically expand new modules to fit the residuals between the target and the output of the original model. Next, we remove redundant parameters and feature dimensions through an effective distillation strategy to maintain the single backbone model. We validate our method FOSTER on CIFAR-100 and ImageNet-100/1000 under different settings. Experimental results show that our method achieves state-of-the-art performance. Code is available at: https://github.com/G-U-N/ECCV22-FOSTER."
                },
                {
                    "title": "Equalized Focal Loss for Dense Long-Tailed Object Detection",
                    "abstract": "Despite the recent success of long-tailed object detection, almost all long-tailed object detectors are developed based on the two-stage paradigm. In practice, one-stage detectors are more prevalent in the industry because they have a simple and fast pipeline that is easy to deploy. However, in the long-tailed scenario, this line of work has not been explored so far. In this paper, we investigate whether one-stage detectors can perform well in this case. We discover the primary obstacle that prevents one-stage detectors from achieving excellent performance is: categories suffer from different degrees of positive-negative imbalance problems under the long-tailed data distribution. The conventional focal loss balances the training process with the same modulating factor for all categories, thus failing to handle the long-tailed problem. To address this issue, we propose the Equalized Focal Loss (EFL) that rebalances the loss contribution of positive and negative samples of different categories independently according to their imbalance degrees. Specifically, EFL adopts a category-relevant modulating factor which can be adjusted dynamically by the training status of different categories. Extensive experiments conducted on the challenging LVIS v1 benchmark demonstrate the effectiveness of our proposed method. With an end-to-end training pipeline, EFL achieves 29.2% in terms of overall AP and obtains significant performance improvements on rare categories, surpassing all existing state-of-the-art methods. The code is available at https: //github.com/ModelTC/EOD."
                },
                {
                    "title": "SGTR: End-to-end Scene Graph Generation with Transformer",
                    "abstract": "Scene Graph Generation (SGG) remains a challenging visual understanding task due to its compositional property. Most previous works adopt a bottom-up two-stage or a point-based one-stage approach, which often suffers from high time complexity or sub-optimal designs. In this work, we propose a novel SGG method to address the aforementioned issues, formulating the task as a bipartite graph construction problem. To solve the problem, we develop a transformer-based end-to-end framework that first generates the entity and predicate proposal set, followed by inferring directed edges to form the relation triplets. In particular, we develop a new entity-aware predicate representation based on a structural predicate generator that leverages the compositional property of relationships. Moreover, we design a graph assembling module to infer the connectivity of the bipartite scene graph based on our entity-aware structure, enabling us to generate the scene graph in an end-to-end manner. Extensive experimental results show that our design is able to achieve the state-of-the-art or comparable performance on two challenging benchmarks, surpassing most of the existing approaches and enjoying higher efficiency in inference. We hope our model can serve as a strong baseline for the Transformer-based scene graph generation. 11Code is available: https://github.com/Scarecrow0/SGTR"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Learning to Prompt for Continual Learning",
                    "abstract": "The mainstream paradigm behind continual learning has been to adapt the model parameters to non-stationary data distributions, where catastrophic forgetting is the central challenge. Typical methods rely on a rehearsal buffer or known task identity at test time to retrieve learned knowl-edge and address forgetting, while this work presents a new paradigm for continual learning that aims to train a more succinct memory system without accessing task identity at test time. Our method learns to dynamically prompt (L2P) a pre-trained model to learn tasks sequen-tially under different task transitions. In our proposed framework, prompts are small learnable parameters, which are maintained in a memory space. The objective is to optimize prompts to instruct the model prediction and ex-plicitly manage task-invariant and task-specific knowledge while maintaining model plasticity. We conduct comprehen-sive experiments under popular image classification bench-marks with different challenging continual learning set-tings, where L2P consistently outperforms prior state-of-the-art methods. Surprisingly, L2P achieves competitive results against rehearsal-based methods even without a re-hearsal buffer and is directly applicable to challenging task-agnostic continual learning. Source code is available at https://github.com/google-research/12p."
                },
                {
                    "title": "Uncovering Main Causalities for Long-tailed Information Extraction",
                    "abstract": "Information Extraction (IE) aims to extract structural information from unstructured texts. In practice, long-tailed distributions caused by the selection bias of a dataset may lead to incorrect correlations, also known as spurious correlations, between entities and labels in the conventional likelihood models. This motivates us to propose counterfactual IE (CFIE), a novel framework that aims to uncover the main causalities behind data in the view of causal inference. Specifically, 1) we first introduce a unified structural causal model (SCM) for various IE tasks, describing the relationships among variables; 2) with our SCM, we then generate counterfactuals based on an explicit language structure to better calculate the direct causal effect during the inference stage; 3) we further propose a novel debiasing approach to yield more robust predictions. Experiments on three IE tasks across five public datasets show the effectiveness of our CFIE model in mitigating the spurious correlation issues."
                },
                {
                    "title": "Wanderlust: Online Continual Object Detection in the Real World",
                    "abstract": "Online continual learning from data streams in dynamic environments is a critical direction in the computer vision field. However, realistic benchmarks and fundamental studies in this line are still missing. To bridge the gap, we present a new online continual object detection benchmark with an egocentric video dataset, Objects Around Krishna (OAK). OAK adopts the KrishnaCAM videos, an ego-centric video stream collected over nine months by a graduate student. OAK provides exhaustive bounding box annotations of 80 video snippets (~17.5 hours) for 105 object categories in outdoor scenes. The emergence of new object categories in our benchmark follows a pattern similar to what a single person might see in their day-to-day life. The dataset also captures the natural distribution shifts as the person travels to different places. These egocentric long running videos provide a realistic playground for continual learning algorithms, especially in online embodied settings. We also introduce new evaluation metrics to evaluate the model performance and catastrophic forgetting and provide baseline studies for online continual object detection. We believe this benchmark will pose new exciting challenges for learning from non-stationary data in continual learning. The OAK dataset and the associated benchmark are released at https://oakdata.github.io/."
                },
                {
                    "title": "Learning of Visual Relations: The Devil is in the Tails",
                    "abstract": "Significant effort has been recently devoted to modeling visual relations. This has mostly addressed the design of architectures, typically by adding parameters and increasing model complexity. However, visual relation learning is a long-tailed problem, due to the combinatorial nature of joint reasoning about groups of objects. Increasing model complexity is, in general, illsuited for long-tailed problems due to their tendency to overfit. In this paper, we explore an alternative hypothesis, denoted the Devil is in the Tails. Under this hypothesis, better performance is achieved by keeping the model simple but improving its ability to cope with long-tailed distributions. To test this hypothesis, we devise a new approach for training visual relationships models, which is inspired by state-of-the-art long-tailed recognition literature. This is based on an iterative decoupled training scheme, denoted Decoupled Training for Devil in the Tails (DT2). DT2 employs a novel sampling approach, Alternating Class-Balanced Sampling (ACBS), to capture the interplay between the long-tailed entity and predicate distributions of visual relations. Results show that, with an extremely simple architecture, DT2-ACBS significantly out-performs much more complex state-of-the-art methods on scene graph generation tasks. This suggests that the development of sophisticated models must be considered in tandem with the long-tailed nature of the problem."
                },
                {
                    "title": "Recovering the Unbiased Scene Graphs from the Biased Ones",
                    "abstract": "Given input images, scene graph generation (SGG) aims to produce comprehensive, graphical representations describing visual relationships among salient objects. Recently, more efforts have been paid to the long tail problem in SGG; however, the imbalance in the fraction of missing labels of different classes, or reporting bias, exacerbating the long tail is rarely considered and cannot be solved by the existing debiasing methods. In this paper we show that, due to the missing labels, SGG can be viewed as a \"Learning from Positive and Unlabeled data\" (PU learning) problem, where the reporting bias can be removed by recovering the unbiased probabilities from the biased ones by utilizing label frequencies, i.e., the per-class fraction of labeled, positive examples in all the positive examples. To obtain accurate label frequency estimates, we propose Dynamic Label Frequency Estimation (DLFE) to take advantage of training-time data augmentation and average over multiple training iterations to introduce more valid examples. Extensive experiments show that DLFE is more effective in estimating label frequencies than a naive variant of the traditional estimate, and DLFE significantly alleviates the long tail and achieves state-of-the-art debiasing performance on the VG dataset. We also show qualitatively that SGG models with DLFE produce prominently more balanced and unbiased scene graphs. The source code is publicly available."
                },
                {
                    "title": "Co2L: Contrastive Continual Learning",
                    "abstract": "Recent breakthroughs in self-supervised learning show that such algorithms learn visual representations that can be transferred better to unseen tasks than cross-entropy based methods which rely on task-specific supervision. In this paper, we found that the similar holds in the continual learning context: contrastively learned representations are more robust against the catastrophic forgetting than ones trained with the cross-entropy objective. Based on this novel observation, we propose a rehearsal-based continual learning algorithm that focuses on continually learning and maintaining transferable representations. More specifically, the proposed scheme (1) learns representations using the contrastive learning objective, and (2) preserves learned representations using a self-supervised distillation step. We conduct extensive experimental validations under popular benchmark image classification datasets, where our method sets the new state-of-the-art performance. Source code is available at https://github.com/chaht01/Co2L."
                },
                {
                    "title": "When Pigs Fly: Contextual Reasoning in Synthetic and Natural Scenes",
                    "abstract": "Context is of fundamental importance to both human and machine vision; e.g., an object in the air is more likely to be an airplane than a pig. The rich notion of context incorporates several aspects including physics rules, statistical co-occurrences, and relative object sizes, among others. While previous work has focused on crowd-sourced out-of-context photographs from the web to study scene context, controlling the nature and extent of contextual violations has been a daunting task. Here we introduce a diverse, synthetic Out-of-Context Dataset (OCD) with fine-grained control over scene context. By leveraging a 3D simulation engine, we systematically control the gravity, object co-occurrences and relative sizes across 36 object categories in a virtual household environment. We conducted a series of experiments to gain insights into the impact of contextual cues on both human and machine vision using OCD. We conducted psychophysics experiments to establish a human benchmark for out-of-context recognition, and then compared it with state-of-the-art computer vision models to quantify the gap between the two. We propose a context-aware recognition transformer model, fusing object and contextual information via multi-head attention. Our model captures useful information for contextual reasoning, enabling human-level performance and better robustness in out-of-context conditions compared to baseline models across OCD and other out-of-context datasets. All source code and data are publicly available at https://github.com/kreimanlab/WhenPigsFlyContext"
                },
                {
                    "title": "Bipartite Graph Network with Adaptive Message Passing for Unbiased Scene Graph Generation",
                    "abstract": "Scene graph generation is an important visual understanding task with a broad range of vision applications. Despite recent tremendous progress, it remains challenging due to the intrinsic long-tailed class distribution and large intra-class variation. To address these issues, we introduce a novel confidence-aware bipartite graph neural network with adaptive message propagation mechanism for unbiased scene graph generation. In addition, we propose an efficient bi-level data resampling strategy to alleviate the imbalanced data distribution problem in training our graph network. Our approach achieves superior or competitive performance over previous methods on several challenging datasets, including Visual Genome, Open Images V4/V6, demonstrating its effectiveness and generality."
                },
                {
                    "title": "Generative Replay-based Continual Zero-Shot Learning",
                    "abstract": "Zero-shot learning is a new paradigm to classify objects from classes that are not available at training time. Zero-shot learning (ZSL) methods have attracted considerable attention in recent years because of their ability to classify unseen/novel class examples. Most of the existing approaches on ZSL works when all the samples from seen classes are available to train the model, which does not suit real life. In this paper, we tackle this hindrance by developing a generative replay-based continual ZSL (GRCZSL). The proposed method endows traditional ZSL to learn from streaming data and acquire new knowledge without forgetting the previous tasks' gained experience. We handle catastrophic forgetting in GRCZSL by replaying the synthetic samples of seen classes, which have appeared in the earlier tasks. These synthetic samples are synthesized using the trained conditional variational autoencoder (VAE) over the immediate past task. Moreover, we only require the current and immediate previous VAE at any time for training and testing. The proposed GRZSL method is developed for a single-head setting of continual learning, simulating a real-world problem setting. In this setting, task identity is given during training but unavailable during testing. GRCZSL performance is evaluated on five benchmark datasets for the generalized setup of ZSL with fixed and dynamic (incremental class) settings of continual learning. The existing class setting presented recently in the literature is not suitable for a class-incremental setting. Therefore, this paper proposes a new setting to address this issue. Experimental results show that the proposed method significantly outperforms the baseline and the state-of-the-art method and makes it more suitable for real-world applications."
                },
                {
                    "title": "Continual Learning Strategy in One-Stage Object Detection Framework Based on Experience Replay for Autonomous Driving Vehicle",
                    "abstract": "Object detection is an important aspect for autonomous driving vehicles (ADV), which may comprise of a machine learning model that detects a range of classes. As the deployment of ADV widens globally, the variety of objects to be detected may increase beyond the designated range of classes. Continual learning for object detection essentially ensure a robust adaptation of a model to detect additional classes on the fly. This study proposes a novel continual learning method for object detection that learns new object class(es) along with cumulative memory of classes from prior learning rounds to avoid any catastrophic forgetting. The results of PASCAL VOC 2007 have suggested that the proposed ER method obtains 4.3% of mAP drop compared against the all-classes learning, which is the lowest amongst other prior arts."
                },
                {
                    "title": "Deformable DETR: Deformable Transformers for End-to-End Object Detection",
                    "abstract": "DETR has been recently proposed to eliminate the need for many hand-designed components in object detection while demonstrating good performance. However, it suffers from slow convergence and limited feature spatial resolution, due to the limitation of Transformer attention modules in processing image feature maps. To mitigate these issues, we proposed Deformable DETR, whose attention modules only attend to a small set of key sampling points around a reference. Deformable DETR can achieve better performance than DETR (especially on small objects) with 10$\\times$ less training epochs. Extensive experiments on the COCO benchmark demonstrate the effectiveness of our approach. Code shall be released."
                },
                {
                    "title": "Compacting, Picking and Growing for Unforgetting Continual Learning",
                    "abstract": "Continual lifelong learning is essential to many applications. In this paper, we propose a simple but effective approach to continual deep learning. Our approach leverages the principles of deep model compression, critical weights selection, and progressive networks expansion. By enforcing their integration in an iterative manner, we introduce an incremental learning method that is scalable to the number of sequential tasks in a continual learning process. Our approach is easy to implement and owns several favorable characteristics. First, it can avoid forgetting (i.e., learn new tasks while remembering all previous tasks). Second, it allows model expansion but can maintain the model compactness when handling sequential tasks. Besides, through our compaction and selection/expansion mechanism, we show that the knowledge accumulated through learning previous tasks is helpful to build a better model for the new tasks compared to training the models independently with tasks. Experimental results show that our approach can incrementally learn a deep model tackling multiple tasks without forgetting, while the model compactness is maintained with the performance more satisfiable than individual task training."
                },
                {
                    "title": "Continual Unsupervised Representation Learning",
                    "abstract": "Continual learning aims to improve the ability of modern learning systems to deal with non-stationary distributions, typically by attempting to learn a series of tasks sequentially. Prior art in the field has largely considered supervised or reinforcement learning tasks, and often assumes full knowledge of task labels and boundaries. In this work, we propose an approach (CURL) to tackle a more general problem that we will refer to as unsupervised continual learning. The focus is on learning representations without any knowledge about task identity, and we explore scenarios when there are abrupt changes between tasks, smooth transitions from one task to another, or even when the data is shuffled. The proposed approach performs task inference directly within the model, is able to dynamically expand to capture new concepts over its lifetime, and incorporates additional rehearsal-based techniques to deal with catastrophic forgetting. We demonstrate the efficacy of CURL in an unsupervised learning setting with MNIST and Omniglot, where the lack of labels ensures no information is leaked about the task. Further, we demonstrate strong performance compared to prior art in an i.i.d setting, or when adapting the technique to supervised tasks such as incremental class learning."
                },
                {
                    "title": "Scene Graph Parsing by Attention Graph",
                    "abstract": "Scene graph representations, which form a graph of visual object nodes together with their attributes and relations, have proved useful across a variety of vision and language applications. Recent work in the area has used Natural Language Processing dependency tree methods to automatically build scene graphs. \nIn this work, we present an 'Attention Graph' mechanism that can be trained end-to-end, and produces a scene graph structure that can be lifted directly from the top layer of a standard Transformer model. \nThe scene graphs generated by our model achieve an F-score similarity of 52.21% to ground-truth graphs on the evaluation set using the SPICE metric, surpassing the best previous approaches by 2.5%."
                },
                {
                    "title": "SpatialSense: An Adversarially Crowdsourced Benchmark for Spatial Relation Recognition",
                    "abstract": "Understanding the spatial relations between objects in images is a surprisingly challenging task. A chair may be \"behind\" a person even if it appears to the left of the person in the image (depending on which way the person is facing). Two students that appear close to each other in the image may not in fact be \"next to\" each other if there is a third student between them. We introduce SpatialSense, a dataset specializing in spatial relation recognition which captures a broad spectrum of such challenges, allowing for proper benchmarking of computer vision techniques. SpatialSense is constructed through adversarial crowdsourcing, in which human annotators are tasked with finding spatial relations that are difficult to predict using simple cues such as 2D spatial configuration or language priors. Adversarial crowdsourcing significantly reduces dataset bias and samples more interesting relations in the long tail compared to existing datasets. On SpatialSense, state-of-the-art recognition models perform comparably to simple baselines, suggesting that they rely on straightforward cues instead of fully reasoning about this complex task. The SpatialSense benchmark provides a path forward to advancing the spatial reasoning capabilities of computer vision systems. The dataset and code are available at https://github.com/princeton-vl/SpatialSense."
                },
                {
                    "title": "LVIS: A Dataset for Large Vocabulary Instance Segmentation",
                    "abstract": "Progress on object detection is enabled by datasets that focus the research community\u2019s attention on open challenges. This process led us from simple images to complex scenes and from bounding boxes to segmentation masks. In this work, we introduce LVIS (pronounced \u2018el-vis\u2019): a new dataset for Large Vocabulary Instance Segmentation. We plan to collect 2.2 million high-quality instance segmentation masks for over 1000 entry-level object categories in 164k images. Due to the Zipfian distribution of categories in natural images, LVIS naturally has a long tail of categories with few training samples. Given that state-of-the-art deep learning methods for object detection perform poorly in the low-sample regime, we believe that our dataset poses an important and exciting new scientific challenge. LVIS is available at http://www.lvisdataset.org."
                },
                {
                    "title": "Exploring Context and Visual Pattern of Relationship for Scene Graph Generation",
                    "abstract": "Relationship is the core of scene graph, but its prediction is far from satisfying because of its complex visual diversity. To alleviate this problem, we treat relationship as an abstract object, exploring not only significative visual pattern but contextual information for it, which are two key aspects when considering object recognition. Our observation on current datasets reveals that there exists intimate association among relationships. Therefore, inspired by the successful application of context to object-oriented tasks, we especially construct context for relationships where all of them are gathered so that the recognition could benefit from their association. Moreover, accurate recognition needs discriminative visual pattern for object, and so does relationship. In order to discover effective pattern for relationship, traditional relationship feature extraction methods such as using union region or combination of subject-object feature pairs are replaced with our proposed intersection region which focuses on more essential parts. Therefore, we present our so-called Relationship Context - InterSeCtion Region (CISC) method. Experiments for scene graph generation on Visual Genome dataset and visual relationship prediction on VRD dataset indicate that both the relationship context and intersection region improve performances and realize anticipated functions."
                },
                {
                    "title": "Learning to Remember: A Synaptic Plasticity Driven Framework for Continual Learning",
                    "abstract": "Models trained in the context of continual learning (CL) should be able to learn from a stream of data over an undefined period of time. The main challenges herein are: 1) maintaining old knowledge while simultaneously benefiting from it when learning new tasks, and 2) guaranteeing model scalability with a growing amount of data to learn from. In order to tackle these challenges, we introduce Dynamic Generative Memory (DGM) - synaptic plasticity driven framework for continual learning. DGM relies on conditional generative adversarial networks with learnable connection plasticity realized with neural masking. Specifically, we evaluate two variants of neural masking: applied to (i) layer activations and (ii) to connection weights directly. Furthermore, we propose a dynamic network expansion mechanism that ensures sufficient model capacity to accommodate for continually incoming tasks. The amount of added capacity is determined dynamically from the learned binary mask. We evaluate DGM in the continual class-incremental setup on visual classification tasks."
                },
                {
                    "title": "Graphical Contrastive Losses for Scene Graph Parsing",
                    "abstract": "Most scene graph parsers use a two-stage pipeline to detect visual relationships: the first stage detects entities, and the second predicts the predicate for each entity pair using a softmax distribution. We find that such pipelines, trained with only a cross entropy loss over predicate classes, suffer from two common errors. The first, Entity Instance Confusion, occurs when the model confuses multiple instances of the same type of entity (e.g. multiple cups). The second, Proximal Relationship Ambiguity, arises when multiple subject-predicate-object triplets appear in close proximity with the same predicate, and the model struggles to infer the correct subject-object pairings (e.g. mis-pairing musicians and their instruments). We propose a set of contrastive loss formulations that specifically target these types of errors within the scene graph parsing problem, collectively termed the Graphical Contrastive Losses. These losses explicitly force the model to disambiguate related and unrelated instances through margin constraints specific to each type of confusion. We further construct a relationship detector, called RelDN, using the aforementioned pipeline to demonstrate the efficacy of our proposed losses. Our model outperforms the winning method of the OpenImages Relationship Detection Challenge by 4.7\\% (16.5\\% relatively) on the test set. We also show improved results over the best previous methods on the Visual Genome and Visual Relationship Detection datasets."
                },
                {
                    "title": "On Tiny Episodic Memories in Continual Learning",
                    "abstract": "In continual learning (CL), an agent learns from a stream of tasks leveraging prior experience to transfer knowledge to future tasks. It is an ideal framework to decrease the amount of supervision in the existing learning algorithms. But for a successful knowledge transfer, the learner needs to remember how to perform previous tasks. One way to endow the learner the ability to perform tasks seen in the past is to store a small memory, dubbed episodic memory, that stores few examples from previous tasks and then to replay these examples when training for future tasks. In this work, we empirically analyze the effectiveness of a very small episodic memory in a CL setup where each training example is only seen once. Surprisingly, across four rather different supervised learning benchmarks adapted to CL, a very simple baseline, that jointly trains on both examples from the current task as well as examples stored in the episodic memory, significantly outperforms specifically designed CL approaches with and without episodic memory. Interestingly, we find that repetitive training on even tiny memories of past tasks does not harm generalization, on the contrary, it improves it, with gains between 7\\% and 17\\% when the memory is populated with a single example per class."
                },
                {
                    "title": "Generalized Intersection Over Union: A Metric and a Loss for Bounding Box Regression",
                    "abstract": "Intersection over Union (IoU) is the most popular evaluation metric used in the object detection benchmarks. However, there is a gap between optimizing the commonly used distance losses for regressing the parameters of a bounding box and maximizing this metric value. The optimal objective for a metric is the metric itself. In the case of axis-aligned 2D bounding boxes, it can be shown that IoU can be directly used as a regression loss. However, IoU has a plateau making it infeasible to optimize in the case of non-overlapping bounding boxes. In this paper, we address the this weakness by introducing a generalized version of IoU as both a new loss and a new metric. By incorporating this generalized IoU ( GIoU) as a loss into the state-of-the art object detection frameworks, we show a consistent improvement on their performance using both the standard, IoU based, and new, GIoU based, performance measures on popular object detection benchmarks such as PASCAL VOC and MS COCO."
                },
                {
                    "title": "Generating Natural Language Explanations for Visual Question Answering using Scene Graphs and Visual Attention",
                    "abstract": "In this paper, we present a novel approach for the task of eXplainable Question Answering (XQA), i.e., generating natural language (NL) explanations for the Visual Question Answering (VQA) problem. We generate NL explanations comprising of the evidence to support the answer to a question asked to an image using two sources of information: (a) annotations of entities in an image (e.g., object labels, region descriptions, relation phrases) generated from the scene graph of the image, and (b) the attention map generated by a VQA model when answering the question. We show how combining the visual attention map with the NL representation of relevant scene graph entities, carefully selected using a language model, can give reasonable textual explanations without the need of any additional collected data (explanation captions, etc). We run our algorithms on the Visual Genome (VG) dataset and conduct internal user-studies to demonstrate the efficacy of our approach over a strong baseline. We have also released a live web demo showcasing our VQA and textual explanation generation using scene graphs and visual attention."
                },
                {
                    "title": "VrR-VG: Refocusing Visually-Relevant Relationships",
                    "abstract": "Relationships encode the interactions among individual instances and play a critical role in deep visual scene understanding. Suffering from the high predictability with non-visual information, relationship models tend to fit the statistical bias rather than ``learning\" to infer the relationships from images. To encourage further development in visual relationships, we propose a novel method to mine more valuable relationships by automatically pruning visually-irrelevant relationships. We construct a new scene graph dataset named Visually-Relevant Relationships Dataset (VrR-VG) based on Visual Genome. Compared with existing datasets, the performance gap between learnable and statistical method is more significant in VrR-VG, and frequency-based analysis does not work anymore. Moreover, we propose to learn a relationship-aware representation by jointly considering instances, attributes and relationships. By applying the representation-aware feature learned on VrR-VG, the performances of image captioning and visual question answering are systematically improved, which demonstrates the effectiveness of both our dataset and features embedding schema. Both our VrR-VG dataset and representation-aware features will be made publicly available soon."
                },
                {
                    "title": "Generative Models from the perspective of Continual Learning",
                    "abstract": "Which generative model is the most suitable for Continual Learning? This paper aims at evaluating and comparing generative models on disjoint sequential image generation tasks. We investigate how several models learn and forget, considering various strategies: rehearsal, regularization, generative replay and fine-tuning. We used two quantitative metrics to estimate the generation quality and memory ability. We experiment with sequential tasks on three commonly used benchmarks for Continual Learning (MNIST, Fashion MNIST and CIFAR10). We found that among all models, the original GAN performs best and among Continual Learning strategies, generative replay outperforms all other methods. Even if we found satisfactory combinations on MNIST and Fashion MNIST, training generative models sequentially on CIFAR10 is particularly instable, and remains a challenge. Our code is available online1."
                },
                {
                    "title": "Experience Replay for Continual Learning",
                    "abstract": "Continual learning is the problem of learning new tasks or knowledge while protecting old knowledge and ideally generalizing from old experience to learn new tasks faster. Neural networks trained by stochastic gradient descent often degrade on old tasks when trained successively on new tasks with different data distributions. This phenomenon, referred to as catastrophic forgetting, is considered a major hurdle to learning with non-stationary data or sequences of new tasks, and prevents networks from continually accumulating knowledge and skills. We examine this issue in the context of reinforcement learning, in a setting where an agent is exposed to tasks in a sequence. Unlike most other work, we do not provide an explicit indication to the model of task boundaries, which is the most general circumstance for a learning agent exposed to continuous experience. While various methods to counteract catastrophic forgetting have recently been proposed, we explore a straightforward, general, and seemingly overlooked solution - that of using experience replay buffers for all past events - with a mixture of on- and off-policy learning, leveraging behavioral cloning. We show that this strategy can still learn new tasks quickly yet can substantially reduce catastrophic forgetting in both Atari and DMLab domains, even matching the performance of methods that require task identities. When buffer storage is constrained, we confirm that a simple mechanism for randomly discarding data allows a limited size buffer to perform almost as well as an unbounded one."
                },
                {
                    "title": "Learning to Learn without Forgetting By Maximizing Transfer and Minimizing Interference",
                    "abstract": "Lack of performance when it comes to continual learning over non-stationary distributions of data remains a major challenge in scaling neural network learning to more human realistic settings. In this work we propose a new conceptualization of the continual learning problem in terms of a temporally symmetric trade-off between transfer and interference that can be optimized by enforcing gradient alignment across examples. We then propose a new algorithm, Meta-Experience Replay (MER), that directly exploits this view by combining experience replay with optimization based meta-learning. This method learns parameters that make interference based on future gradients less likely and transfer based on future gradients more likely. We conduct experiments across continual lifelong supervised learning benchmarks and non-stationary reinforcement learning environments demonstrating that our approach consistently outperforms recently proposed baselines for continual learning. Our experiments show that the gap between the performance of MER and baseline algorithms grows both as the environment gets more non-stationary and as the fraction of the total experiences stored gets smaller."
                },
                {
                    "title": "Large-Scale Visual Relationship Understanding",
                    "abstract": "Large scale visual understanding is challenging, as it requires a model to handle the widely-spread and imbalanced distribution of \u2329subject, relation, object\u232a triples. In real-world scenarios with large numbers of objects and relations, some are seen very commonly while others are barely seen. We develop a new relationship detection model that embeds objects and relations into two vector spaces where both discriminative capability and semantic affinity are preserved. We learn a visual and a semantic module that map features from the two modalities into a shared space, where matched pairs of features have to discriminate against those unmatched, but also maintain close distances to semantically similar ones. Benefiting from that, our model can achieve superior performance even when the visual entity categories scale up to more than 80,000, with extremely skewed class distribution. We demonstrate the efficacy of our model on a large and imbalanced benchmark based of Visual Genome that comprises 53,000+ objects and 29,000+ relations, a scale at which no previous work has been evaluated at. We show superiority of our model over competitive baselines on the original Visual Genome dataset with 80,000+ categories. We also show state-of-the-art performance on the VRD dataset and the scene graph dataset which is a subset of Visual Genome with 200 categories."
                },
                {
                    "title": "Natural Language Guided Visual Relationship Detection",
                    "abstract": "Reasoning about the relationships between object pairs in images is a crucial task for holistic scene understanding. Most of the existing works treat this task as a pure visual classification task: each type of relationship or phrase is classified as a relation category based on the extracted visual features. However, each kind of relationships has a wide variety of object combination and each pair of objects has diverse interactions. Obtaining sufficient training samples for all possible relationship categories is difficult and expensive. In this work, we propose a natural language guided framework to tackle this problem. We propose to use a generic bi-directional recurrent neural network to predict the semantic connection between the participating objects in the relationship from the aspect of natural language. The proposed simple method achieves the state-of-the-art on the Visual Relationship Detection (VRD) and Visual Genome datasets, especially when predicting unseen relationships (e.g., recall improved from 76.42% to 89.79% on VRD zeroshot testing set)."
                },
                {
                    "title": "PackNet: Adding Multiple Tasks to a Single Network by Iterative Pruning",
                    "abstract": "This paper presents a method for adding multiple tasks to a single deep neural network while avoiding catastrophic forgetting. Inspired by network pruning techniques, we exploit redundancies in large deep networks to free up parameters that can then be employed to learn new tasks. By performing iterative pruning and network re-training, we are able to sequentially \"pack\" multiple tasks into a single network while ensuring minimal drop in performance and minimal storage overhead. Unlike prior work that uses proxy losses to maintain accuracy on older tasks, we always optimize for the task at hand. We perform extensive experiments on a variety of network architectures and large-scale datasets, and observe much better robustness against catastrophic forgetting than prior work. In particular, we are able to add three fine-grained classification tasks to a single ImageNet-trained VGG-16 network and achieve accuracies close to those of separately trained networks for each task."
                },
                {
                    "title": "Lifelong Learning with Dynamically Expandable Networks",
                    "abstract": "We propose a novel deep network architecture for lifelong learning which we refer to as Dynamically Expandable Network (DEN), that can dynamically decide its network capacity as it trains on a sequence of tasks, to learn a compact overlapping knowledge sharing structure among tasks. DEN is efficiently trained in an online manner by performing selective retraining, dynamically expands network capacity upon arrival of each task with only the necessary number of units, and effectively prevents semantic drift by splitting/duplicating units and timestamping them. We validate DEN on multiple public datasets under lifelong learning scenarios, on which it not only significantly outperforms existing lifelong learning methods for deep networks, but also achieves the same level of performance as the batch counterparts with substantially fewer number of parameters. Further, the obtained network fine-tuned on all tasks obtained significantly better performance over the batch models, which shows that it can be used to estimate the optimal network structure even when all tasks are available in the first place."
                },
                {
                    "title": "Scene Graph Generation from Objects, Phrases and Region Captions",
                    "abstract": "Object detection, scene graph generation and region captioning, which are three scene understanding tasks at different semantic levels, are tied together: scene graphs are generated on top of objects detected in an image with their pairwise relationship predicted, while region captioning gives a language description of the objects, their attributes, relations and other context information. In this work, to leverage the mutual connections across semantic levels, we propose a novel neural network model, termed as Multi-level Scene Description Network (denoted as MSDN), to solve the three vision tasks jointly in an end-to-end manner. Object, phrase, and caption regions are first aligned with a dynamic graph based on their spatial and semantic connections. Then a feature refining structure is used to pass messages across the three levels of semantic tasks through the graph. We benchmark the learned model on three tasks, and show the joint learning across three tasks with our proposed method can bring mutual improvements over previous models. Particularly, on the scene graph generation task, our proposed method outperforms the stateof- art method with more than 3% margin. Code has been made publicly available."
                },
                {
                    "title": "Visual Relationship Detection with Internal and External Linguistic Knowledge Distillation",
                    "abstract": "Understanding the visual relationship between two objects involves identifying the subject, the object, and a predicate relating them. We leverage the strong correlations between the predicate and the hsubj; obji pair (both semantically and spatially) to predict predicates conditioned on the subjects and the objects. Modeling the three entities jointly more accurately reflects their relationships compared to modeling them independently, but it complicates learning since the semantic space of visual relationships is huge and training data is limited, especially for longtail relationships that have few instances. To overcome this, we use knowledge of linguistic statistics to regularize visual model learning. We obtain linguistic knowledge by mining from both training annotations (internal knowledge) and publicly available text, e.g., Wikipedia (external knowledge), computing the conditional probability distribution of a predicate given a (subj, obj) pair. As we train the visual model, we distill this knowledge into the deep model to achieve better generalization. Our experimental results on the Visual Relationship Detection (VRD) and Visual Genome datasets suggest that with this linguistic knowledge distillation, our model outperforms the stateof- the-art methods significantly, especially when predicting unseen relationships (e.g., recall improved from 8.45% to 19.17% on VRD zero-shot testing set)."
                },
                {
                    "title": "Gradient Episodic Memory for Continual Learning",
                    "abstract": "One major obstacle towards AI is the poor ability of models to solve new problems quicker, and without forgetting previously acquired knowledge. To better understand this issue, we study the problem of continual learning, where the model observes, once and one by one, examples concerning a sequence of tasks. First, we propose a set of metrics to evaluate models learning over a continuum of data. These metrics characterize models not only by their test accuracy, but also in terms of their ability to transfer knowledge across tasks. Second, we propose a model for continual learning, called Gradient Episodic Memory (GEM) that alleviates forgetting, while allowing beneficial transfer of knowledge to previous tasks. Our experiments on variants of the MNIST and CIFAR-100 datasets demonstrate the strong performance of GEM when compared to the state-of-the-art."
                },
                {
                    "title": "Continual Learning with Deep Generative Replay",
                    "abstract": "Attempts to train a comprehensive artificial intelligence capable of solving multiple tasks have been impeded by a chronic problem called catastrophic forgetting. Although simply replaying all previous data alleviates the problem, it requires large memory and even worse, often infeasible in real world applications where the access to past data is limited. Inspired by the generative nature of hippocampus as a short-term memory system in primate brain, we propose the Deep Generative Replay, a novel framework with a cooperative dual model architecture consisting of a deep generative model (\"generator\") and a task solving model (\"solver\"). With only these two models, training data for previous tasks can easily be sampled and interleaved with those for a new task. We test our methods in several sequential learning settings involving image classification tasks."
                },
                {
                    "title": "Detecting Visual Relationships with Deep Relational Networks",
                    "abstract": "Relationships among objects play a crucial role in image understanding. Despite the great success of deep learning techniques in recognizing individual objects, reasoning about the relationships among objects remains a challenging task. Previous methods often treat this as a classification problem, considering each type of relationship (e.g. ride) or each distinct visual phrase (e.g. person-ride-horse) as a category. Such approaches are faced with significant difficulties caused by the high diversity of visual appearance for each kind of relationships or the large number of distinct visual phrases. We propose an integrated framework to tackle this problem. At the heart of this framework is the Deep Relational Network, a novel formulation designed specifically for exploiting the statistical dependencies between objects and their relationships. On two large data sets, the proposed method achieves substantial improvement over state-of-the-art."
                },
                {
                    "title": "Continual Learning Through Synaptic Intelligence",
                    "abstract": "While deep learning has led to remarkable advances across diverse applications, it struggles in domains where the data distribution changes over the course of learning. In stark contrast, biological neural networks continually adapt to changing domains, possibly by leveraging complex molecular machinery to solve many tasks simultaneously. In this study, we introduce intelligent synapses that bring some of this biological complexity into artificial neural networks. Each synapse accumulates task relevant information over time, and exploits this information to rapidly store new memories without forgetting old ones. We evaluate our approach on continual learning of classification tasks, and show that it dramatically reduces forgetting while maintaining computational efficiency."
                },
                {
                    "title": "Visual Translation Embedding Network for Visual Relation Detection",
                    "abstract": "Visual relations, such as person ride bike and bike next to car, offer a comprehensive scene understanding of an image, and have already shown their great utility in connecting computer vision and natural language. However, due to the challenging combinatorial complexity of modeling subject-predicate-object relation triplets, very little work has been done to localize and predict visual relations. Inspired by the recent advances in relational representation learning of knowledge bases and convolutional object detection networks, we propose a Visual Translation Embedding network (VTransE) for visual relation detection. VTransE places objects in a low-dimensional relation space where a relation can be modeled as a simple vector translation, i.e., subject + predicate &#x2248; object. We propose a novel feature extraction layer that enables object-relation knowledge transfer in a fully-convolutional fashion that supports training and inference in a single forward/backward pass. To the best of our knowledge, VTransE is the first end-toend relation detection network. We demonstrate the effectiveness of VTransE over other state-of-the-art methods on two large-scale datasets: Visual Relationship and Visual Genome. Note that even though VTransE is a purely visual model, it is still competitive to the Lu&#x2019;s multi-modal model with language priors [27]."
                },
                {
                    "title": "ViP-CNN: Visual Phrase Guided Convolutional Neural Network",
                    "abstract": "As the intermediate level task connecting image captioning and object detection, visual relationship detection started to catch researchers attention because of its descriptive power and clear structure. It detects the objects and captures their pair-wise interactions with a subject-predicate-object triplet, e.g. person-ride-horse. In this paper, each visual relationship is considered as a phrase with three components. We formulate the visual relationship detection as three inter-connected recognition problems and propose a Visual Phrase guided Convolutional Neural Network (ViP-CNN) to address them simultaneously. In ViP-CNN, we present a Phrase-guided Message Passing Structure (PMPS) to establish the connection among relationship components and help the model consider the three problems jointly. Corresponding non-maximum suppression method and model training strategy are also proposed. Experimental results show that our ViP-CNN outperforms the state-of-art method both in speed and accuracy. We further pretrain ViP-CNN on our cleansed Visual Genome Relationship dataset, which is found to perform better than the pretraining on the ImageNet for this task."
                },
                {
                    "title": "Scene Graph Generation by Iterative Message Passing",
                    "abstract": "Understanding a visual scene goes beyond recognizing individual objects in isolation. Relationships between objects also constitute rich semantic information about the scene. In this work, we explicitly model the objects and their relationships using scene graphs, a visually-grounded graphical structure of an image. We propose a novel end-to-end model that generates such structured scene representation from an input image. Our key insight is that the graph generation problem can be formulated as message passing between the primal node graph and its dual edge graph. Our joint inference model can take advantage of contextual cues to make better predictions on objects and their relationships. The experiments show that our model significantly outperforms previous methods on the Visual Genome dataset as well as support relation inference in NYU Depth V2 dataset."
                },
                {
                    "title": "Overcoming catastrophic forgetting in neural networks",
                    "abstract": "Significance Deep neural networks are currently the most successful machine-learning technique for solving a variety of tasks, including language translation, image classification, and image generation. One weakness of such models is that, unlike humans, they are unable to learn multiple tasks sequentially. In this work we propose a practical solution to train such models sequentially by protecting the weights important for previous tasks. This approach, inspired by synaptic consolidation in neuroscience, enables state of the art results on multiple reinforcement learning problems experienced sequentially. The ability to learn tasks in a sequential fashion is crucial to the development of artificial intelligence. Until now neural networks have not been capable of this and it has been widely thought that catastrophic forgetting is an inevitable feature of connectionist models. We show that it is possible to overcome this limitation and train networks that can maintain expertise on tasks that they have not experienced for a long time. Our approach remembers old tasks by selectively slowing down learning on the weights important for those tasks. We demonstrate our approach is scalable and effective by solving a set of classification tasks based on a hand-written digit dataset and by learning several Atari 2600 games sequentially."
                },
                {
                    "title": "iCaRL: Incremental Classifier and Representation Learning",
                    "abstract": "A major open problem on the road to artificial intelligence is the development of incrementally learning systems that learn about more and more concepts over time from a stream of data. In this work, we introduce a new training strategy, iCaRL, that allows learning in such a class-incremental way: only the training data for a small number of classes has to be present at the same time and new classes can be added progressively. iCaRL learns strong classifiers and a data representation simultaneously. This distinguishes it from earlier works that were fundamentally limited to fixed data representations and therefore incompatible with deep learning architectures. We show by experiments on CIFAR-100 and ImageNet ILSVRC 2012 data that iCaRL can learn many classes incrementally over a long period of time where other strategies quickly fail."
                },
                {
                    "title": "Phrase Localization and Visual Relationship Detection with Comprehensive Image-Language Cues",
                    "abstract": "This paper presents a framework for localization or grounding of phrases in images using a large collection of linguistic and visual cues. We model the appearance, size, and position of entity bounding boxes, adjectives that contain attribute information, and spatial relationships between pairs of entities connected by verbs or prepositions. Special attention is given to relationships between people and clothing or body part mentions, as they are useful for distinguishing individuals. We automatically learn weights for combining these cues and at test time, perform joint inference over all phrases in a caption. The resulting system produces state of the art performance on phrase localization on the Flickr30k Entities dataset [33] and visual relationship detection on the Stanford VRD dataset [27]."
                },
                {
                    "title": "From Images to Sentences through Scene Description Graphs using Commonsense Reasoning and Knowledge",
                    "abstract": "In this paper we propose the construction of linguistic descriptions of images. This is achieved through the extraction of scene description graphs (SDGs) from visual scenes using an automatically constructed knowledge base. SDGs are constructed using both vision and reasoning. Specifically, commonsense reasoning is applied on (a) detections obtained from existing perception methods on given images, (b) a \"commonsense\" knowledge base constructed using natural language processing of image annotations and (c) lexical ontological knowledge from resources such as WordNet. Amazon Mechanical Turk(AMT)-based evaluations on Flickr8k, Flickr30k and MS-COCO datasets show that in most cases, sentences auto-constructed from SDGs obtained by our method give a more relevant and thorough description of an image than a recent state-of-the-art image caption based approach. Our Image-Sentence Alignment Evaluation results are also comparable to that of the recent state-of-the art approaches."
                },
                {
                    "title": "Image retrieval using scene graphs",
                    "abstract": "This paper develops a novel framework for semantic image retrieval based on the notion of a scene graph. Our scene graphs represent objects (\u201cman\u201d, \u201cboat\u201d), attributes of objects (\u201cboat is white\u201d) and relationships between objects (\u201cman standing on boat\u201d). We use these scene graphs as queries to retrieve semantically related images. To this end, we design a conditional random field model that reasons about possible groundings of scene graphs to test images. The likelihoods of these groundings are used as ranking scores for retrieval. We introduce a novel dataset of 5,000 human-generated scene graphs grounded to images and use this dataset to evaluate our method for image retrieval. In particular, we evaluate retrieval using full scene graphs and small scene subgraphs, and show that our method outperforms retrieval methods that use only objects or low-level image features. In addition, we show that our full model can be used to improve object localization compared to baseline methods."
                },
                {
                    "title": "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks",
                    "abstract": "State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available"
                },
                {
                    "title": "Fast R-CNN",
                    "abstract": "This paper proposes Fast R-CNN, a clean and fast framework for object detection. Compared to traditional R-CNN, and its accelerated version SPPnet, Fast R-CNN trains networks using a multi-task loss in a single training stage. The multi-task loss simplifies learning and improves detection accuracy. Unlike SPPnet, all network layers can be updated during fine-tuning. We show that this difference has practical ramifications for very deep networks, such as VGG16, where mAP suffers when only the fully-connected layers are updated. Compared to\"slow\"R-CNN, Fast R-CNN is 9x faster at training VGG16 for detection, 213x faster at test-time, and achieves a significantly higher mAP on PASCAL VOC 2012. Compared to SPPnet, Fast R-CNN trains VGG16 3x faster, tests 10x faster, and is more accurate. Fast R-CNN is implemented in Python and C++ and is available under the open-source MIT License at https://github.com/rbgirshick/fast-rcnn"
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "Deep Neural Networks for Object Detection",
                    "abstract": "Deep Neural Networks (DNNs) have recently shown outstanding performance on image classification tasks [14]. In this paper we go one step further and address the problem of object detection using DNNs, that is not only classifying but also precisely localizing objects of various classes. We present a simple and yet powerful formulation of object detection as a regression problem to object bounding box masks. We define a multi-scale inference procedure which is able to produce high-resolution object detections at a low cost by a few network applications. State-of-the-art performance of the approach is shown on Pascal VOC."
                },
                {
                    "title": "Modern hierarchical, agglomerative clustering algorithms",
                    "abstract": "This paper presents algorithms for hierarchical, agglomerative clustering which perform most efficiently in the general-purpose setup that is given in modern standard software. Requirements are: (1) the input data is given by pairwise dissimilarities between data points, but extensions to vector data are also discussed (2) the output is a \"stepwise dendrogram\", a data structure which is shared by all implementations in current standard software. We present algorithms (old and new) which perform clustering in this setting efficiently, both in an asymptotic worst-case analysis and from a practical point of view. The main contributions of this paper are: (1) We present a new algorithm which is suitable for any distance update scheme and performs significantly better than the existing algorithms. (2) We prove the correctness of two algorithms by Rohlf and Murtagh, which is necessary in each case for different reasons. (3) We give well-founded recommendations for the best current algorithms for the various agglomerative clustering schemes."
                },
                {
                    "title": "Recognition using visual phrases",
                    "abstract": "In this paper we introduce visual phrases, complex visual composites like \u201ca person riding a horse\u201d. Visual phrases often display significantly reduced visual complexity compared to their component objects, because the appearance of those objects can change profoundly when they participate in relations. We introduce a dataset suitable for phrasal recognition that uses familiar PASCAL object categories, and demonstrate significant experimental gains resulting from exploiting visual phrases. We show that a visual phrase detector significantly outperforms a baseline which detects component objects and reasons about relations, even though visual phrase training sets tend to be smaller than those for objects. We argue that any multi-class detection system must decode detector outputs to produce final results; this is usually done with non-maximum suppression. We describe a novel decoding procedure that can account accurately for local context without solving difficult inference problems. We show this decoding procedure outperforms the state of the art. Finally, we show that decoding a combination of phrasal and object detectors produces real improvements in detector results."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Random sampling with a reservoir",
                    "abstract": "We introduce fast algorithms for selecting a random sample of <italic>n</italic> records without replacement from a pool of <italic>N</italic> records, where the value of <italic>N</italic> is unknown beforehand. The main result of the paper is the design and analysis of Algorithm Z; it does the sampling in one pass using constant space and in <italic>O</italic>(<italic>n</italic>(1 + log(<italic>N/n</italic>))) expected time, which is optimum, up to a constant factor. Several optimizations are studied that collectively improve the speed of the naive version of the algorithm by an order of magnitude. We give an efficient Pascal-like implementation that incorporates these modifications and that is suitable for general use. Theoretical and empirical results indicate that Algorithm Z outperforms current methods by a significant margin."
                },
                {
                    "title": "DDGR: Continual Learning with Deep Diffusion-based Generative Replay",
                    "abstract": "Popular deep-learning models in the field of image classification suffer from catastrophic forgetting\u2014models will forget previously acquired skills when learning new ones. Generative replay (GR), which typically consists of a generator and a classifier, is an efficient way to mitigate catastrophic forgetting. However, conventional GR methods only focus on a single instruction relationship (generator-to-classifier), where the generator synthesizes samples for previous tasks to instruct the training of the classifier, while ignoring the ways in which the classifier can benefit the generator. In addition, most generative replay methods typically reuse the generated samples to update the generator, which causes the samples regenerated by the generator deviating from the distribution of previous tasks. To overcome these two issues, we propose a novel approach, called deep diffusion-based generative replay (DDGR), which adopts a diffusion model as the generator and calculates an instruction-operator through the classifier to instruct the generation of samples. Extensive experiments in class incremental (CI) and class incremental with repetition (CIR) settings demonstrate the advantages of DDGR. Our code is available at https://github.com/ xiaocangshengGR/DDGR ."
                },
                {
                    "title": "Unifying Importance Based Regularisation Methods for Continual Learning",
                    "abstract": "Continual Learning addresses the challenge of learning a number of di\ufb00erent tasks sequentially. The goal of maintaining knowledge of earlier tasks without re-accessing them starkly con\ufb02icts with standard SGD training for arti\ufb01-cial neural networks. An in\ufb02uential method to tackle this problem without storing old data are so-called regularisation approaches. They measure the importance of each parameter for solving a given task and subsequently protect important parameters from large changes. In the literature, three ways to measure parameter importance have been put forward and they have inspired a large body of follow-up work. Here, we present strong theoretical and empirical evidence that these three meth-ods, Elastic Weight Consolidation (EWC), Synaptic Intelligence (SI) and Memory Aware Synapses (MAS), are surprisingly similar and are all linked to the same theoretical quantity. Concretely, we show that, despite stemming from very di\ufb00erent motivations, both SI and MAS approximate the square root of the Fisher Information, with the Fisher being the theoretically justi\ufb01ed basis of EWC. Moreover, we show that for SI the relation to the Fisher \u2013 and in fact its performance \u2013 is due to a previously unknown bias. On top of uncovering unknown similarities and unifying regularisation approaches, we also demonstrate that our insights enable practical performance improvements for large batch training."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a continual learning framework for scene graph generation (CSEGG) that allows models to adapt to new objects and relationships in dynamic visual environments without forgetting previously learned information?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of CSEGG is crucial for advancing the field of machine learning, particularly in applications that require real-time adaptability, such as robotic navigation and augmented reality. By addressing this issue, we can enhance the capabilities of models to understand and interact with dynamic environments, leading to more intelligent systems. This research could pave the way for future studies focused on continual learning in complex visual tasks, ultimately contributing to the development of more robust AI systems that can operate effectively in real-world scenarios.\n\n---\n\n**[Question 3] - Why is it hard?**  \nCSEGG presents several challenges that complicate its resolution. First, understanding and capturing the intricate relationships between objects in a scene requires adaptive reasoning, which is more complex than simple object detection. Second, the combinatorial complexity increases non-linearly as new objects and relationships are introduced, making it difficult for models to maintain performance. Third, the long-tailed distribution of objects in real-world scenes necessitates that models continually adapt to evolving distributions, which is a significant technical challenge. Naive approaches may fail because they do not account for the dynamic nature of relationships and the need for models to retain previously learned knowledge while integrating new information.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on static scene graph generation and has not adequately addressed the unique complexities of continual learning in this context. Existing solutions often overlook the need for models to adapt to new relationships and objects without forgetting prior knowledge. Barriers such as the lack of specialized datasets and methodologies for CSEGG have hindered progress. Our approach differs by establishing a comprehensive CSEGG benchmark with specific learning protocols that target the unique challenges of continual learning in scene graph generation, thus filling a critical gap in the literature.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves reorganizing existing scene graph generation datasets to create a novel CSEGG benchmark with three distinct learning protocols: (1) Relationship-incremental setting, where the model learns new relationships among familiar objects; (2) Scene-incremental setting, where the model adapts to new scenes with previously learned objects"
            }
        },
        "author_data": {
            "6eb3947c-3e34-4d2f-a0ea-fe4d214b0d57": {
                "pk": "6eb3947c-3e34-4d2f-a0ea-fe4d214b0d57",
                "name": "Naitik Khandelwal",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "80537c77-6002-4656-81c2-a1a0d9c6ec51": {
                "pk": "80537c77-6002-4656-81c2-a1a0d9c6ec51",
                "name": "Xiao Liu",
                "collaborators": [
                    "Kui Xiao",
                    "Jian-Yang Zhu",
                    "Yaron Oz",
                    "Jiawei Zhang"
                ],
                "domain": [
                    "Quantum Field Theory",
                    "Cosmology",
                    "Statistical Mechanics",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "On the Vacuum Propagation of Gravitational Waves",
                        "abstract": "We show that, for any local, causal quantum field theory which couples covariantly to gravity, and which admits Minkowski spacetime vacuum(a) invariant under the inhomogeneous proper orthochronous Lorentz group, plane gravitational waves propagating in such Minkowski vacuum(a) do not dissipate energy or momentum via quantum field theoretic effects."
                    },
                    {
                        "title": "The Leading Behaviour of The Ground-State Energy of Heavy Ions According to Brown and Ravenhall",
                        "abstract": "In this article we prove the absence of relativistic effects in leading order for the ground-state energy according to Brown-Ravenhall operator. We obtain this asymptotic result for negative ions and for systems with the number of electrons proportional to the nuclear charge. In the case of neutral atoms the analogous result was obtained earlier by Cassanas and Siedentop [4]."
                    },
                    {
                        "title": "The Dirac Operator on Regular Metric Trees",
                        "abstract": "A metric tree is a tree whose edges are viewed as line segments of positive length. The Dirac operator on such tree is the operator which operates on each edge, complemented by the matching conditions at the vertices which were given by Bolte and Harrison \\cite{BolteHarrison2003}. The spectrum of Dirac operator can be quite different, reflecting geometry of the tree.   We discuss a special case of trees, namely the so-called regular trees. They possess a rich group of symmetries. This allows one to construct an orthogonal decomposition of the space $L^2(\\Gamma)$ which reduces the Dirac. Based upon this decomposition, a detailed spectral analysis of Dirac operator on the regular metric trees is possible."
                    },
                    {
                        "title": "Instability and spectrum of the linearized two-phase fluids interface problem at shear flows",
                        "abstract": "This paper is concerned with the 2-dim two-phase interface Euler equation linearized at a pair of monotone shear flows in both fluids. We extend the Howard's Semicircle Theorem and study the eigenvalue distribution of the linearized Euler system. Under certain conditions, there are exactly two eigenvalues for each fixed wave number $k\\in \\mathbb{R}$ in the whole complex plane. We provide sufficient conditions for spectral instability arising from some boundary values of the shear flow velocity. A typical mode is the ocean-air system in which the density ratio of the fluids is sufficiently small. We give a complete picture of eigenvalue distribution for a certain class of shear flows in the ocean-air system."
                    },
                    {
                        "title": "Reconstruction of rational functions made simple",
                        "abstract": "We present a new method for the reconstruction of rational functions through finite-fields sampling that can significantly reduce the number of samples required. The method works by exploiting all the independent linear relations among target functions. Subsequently, the explicit solutions of the functions can be efficiently obtained by solving the linear system. As a first application, we utilize the method to address various examples within the context of Feynman integrals reduction. These examples demonstrate that our method can substantially improve the computational efficiency, making it useful for future computations in particle physics."
                    },
                    {
                        "title": "A Note On Projective Structures On Compact Surfaces",
                        "abstract": "Projective structures on topological surfaces support the structure of 2d CFTs with a degree of technical simplification. Arguments for their existence are reviewed along with their essential properties. We propose a complex analytic manifold $\\mathcal{P}_g$ biholomorphic to $T^*_{(1,0)} \\mathcal{M}_g$ as a pseudo moduli space of the projective structures of the genus $g$ topological surface. Explicit computations at $g=1$ including the analysis of transformations under the modular group support this proposal, and show that $\\mathcal{P}_{g=1}$ naturally resolves the orbifold locus of the affine structure moduli space. For $g \\geq 2$, whether $\\mathcal{P}_g$ contains redundancy at each value of the complex structure moduli remains open.   Physically, the space $\\mathcal{P}_g$ represents the bundle of universal, stationary, chiral hydrodynamic flows spatially confined to compact genus-$g$ Riemann surfaces."
                    },
                    {
                        "title": "Dynamical behaviors of FRW Universe containing a positive/negative potential scalar field in loop quantum cosmology",
                        "abstract": "The dynamical behaviors of FRW Universe containing a posivive/negative potential scalar field in loop quantum cosmology scenario are discussed. The method of the phase-plane analysis is used to investigate the stability of the Universe. It is found that the stability properties in this situation are quite different from the classical cosmology case. For a positive potential scalar field coupled with a barotropic fluid, the cosmological autonomous system has five fixed points and one of them is stable if the adiabatic index $\\gamma$ satisfies $0<\\gamma<2$. This leads to the fact that the universe just have one bounce point instead of the singularity which lies in the quantum dominated area and it is caused by the quantum geometry effect. There are four fixed points if one considers a scalar field with a negative potential, but none of them is stable. Therefore, the universe has two kinds of bounce points, one is caused by the quantum geometry effect and the other is caused by the negative potential, the Universe may enter a classical re-collapse after the quantum bounce. This hints that the spatially flat FRW Universe containing a negative potential scalar field is cyclic."
                    },
                    {
                        "title": "Shocks and Universal Statistics in (1+1)-Dimensional Relativistic Turbulence",
                        "abstract": "We propose that statistical averages in relativistic turbulence exhibit universal properties. We consider analytically the velocity and temperature differences structure functions in the (1+1)-dimensional relativistic turbulence in which shock waves provide the main contribution to the structure functions in the inertial range. We study shock scattering, demonstrate the stability of the shock waves, and calculate the anomalous exponents. We comment on the possibility of finite time blowup singularities."
                    },
                    {
                        "title": "GPTA: Generative Prompt Tuning Assistant for Synergistic Downstream Neural Network Enhancement with LLMs",
                        "abstract": "This study introduces GPTA, a Large Language Model assistance training framework, that enhances the training of downstream task models via prefix prompt. By minimizing data exposure to LLM, the framework addresses the security and legal challenges of applying LLM in downstream task model training. GPTA utilizes a new synergistic training approach, optimizing the downstream models with parameter gradients and LLMs with the novel ``dialogue gradient''. The framework not only demonstrates significant improvements in model performance across six NLP benchmark datasets, but also reduces overfitting in low-resource scenarios effectively. The detailed analyses further validate that our pioneer framework provides a cost-efficient and adaptive method for downstream task model training with LLM support."
                    }
                ]
            },
            "b2c09b22-3466-4cb4-a9e8-f1442ea03b71": {
                "pk": "b2c09b22-3466-4cb4-a9e8-f1442ea03b71",
                "name": "Mengmi Zhang",
                "collaborators": [
                    "Gabriel Kreiman",
                    "Shuangpeng Han",
                    "Claire Tseng",
                    "Ziyu Wang",
                    "Ankur Sikarwar",
                    "Zenglin Shi",
                    "Mike Zheng Shou",
                    "Spandan Madan",
                    "Ying Sun",
                    "Ren Jie Tee"
                ],
                "domain": [
                    "Computer Vision",
                    "Machine Learning",
                    "Cognitive Science",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "What am I Searching for: Zero-shot Target Identity Inference in Visual Search",
                        "abstract": "Can we infer intentions from a person's actions? As an example problem, here we consider how to decipher what a person is searching for by decoding their eye movement behavior. We conducted two psychophysics experiments where we monitored eye movements while subjects searched for a target object. We defined the fixations falling on non-target objects as \"error fixations\". Using those error fixations, we developed a model (InferNet) to infer what the target was. InferNet uses a pre-trained convolutional neural network to extract features from the error fixations and computes a similarity map between the error fixations and all locations across the search image. The model consolidates the similarity maps across layers and integrates these maps across all error fixations. InferNet successfully identifies the subject's goal and outperforms competitive null models, even without any object-specific training on the inference task."
                    },
                    {
                        "title": "What am I Searching for: Zero-shot Target Identity Inference in Visual Search",
                        "abstract": "Can we infer intentions from a person's actions? As an example problem, here we consider how to decipher what a person is searching for by decoding their eye movement behavior. We conducted two psychophysics experiments where we monitored eye movements while subjects searched for a target object. We defined the fixations falling on \\textit{non-target} objects as \"error fixations\". Using those error fixations, we developed a model (InferNet) to infer what the target was. InferNet uses a pre-trained convolutional neural network to extract features from the error fixations and computes a similarity map between the error fixations and all locations across the search image. The model consolidates the similarity maps across layers and integrates these maps across all error fixations. InferNet successfully identifies the subject's goal and outperforms competitive null models, even without any object-specific training on the inference task."
                    },
                    {
                        "title": "Unveiling AI's Blind Spots: An Oracle for In-Domain, Out-of-Domain, and Adversarial Errors",
                        "abstract": "AI models make mistakes when recognizing images-whether in-domain, out-of-domain, or adversarial. Predicting these errors is critical for improving system reliability, reducing costly mistakes, and enabling proactive corrections in real-world applications such as healthcare, finance, and autonomous systems. However, understanding what mistakes AI models make, why they occur, and how to predict them remains an open challenge. Here, we conduct comprehensive empirical evaluations using a \"mentor\" model-a deep neural network designed to predict another model's errors. Our findings show that the mentor model excels at learning from a mentee's mistakes on adversarial images with small perturbations and generalizes effectively to predict in-domain and out-of-domain errors of the mentee. Additionally, transformer-based mentor models excel at predicting errors across various mentee architectures. Subsequently, we draw insights from these observations and develop an \"oracle\" mentor model, dubbed SuperMentor, that achieves 78% accuracy in predicting errors across different error types. Our error prediction framework paves the way for future research on anticipating and correcting AI model behaviours, ultimately increasing trust in AI systems. All code, models, and data will be made publicly available."
                    },
                    {
                        "title": "Decoding the Enigma: Benchmarking Humans and AIs on the Many Facets of Working Memory",
                        "abstract": "Working memory (WM), a fundamental cognitive process facilitating the temporary storage, integration, manipulation, and retrieval of information, plays a vital role in reasoning and decision-making tasks. Robust benchmark datasets that capture the multifaceted nature of WM are crucial for the effective development and evaluation of AI WM models. Here, we introduce a comprehensive Working Memory (WorM) benchmark dataset for this purpose. WorM comprises 10 tasks and a total of 1 million trials, assessing 4 functionalities, 3 domains, and 11 behavioral and neural characteristics of WM. We jointly trained and tested state-of-the-art recurrent neural networks and transformers on all these tasks. We also include human behavioral benchmarks as an upper bound for comparison. Our results suggest that AI models replicate some characteristics of WM in the brain, most notably primacy and recency effects, and neural clusters and correlates specialized for different domains and functionalities of WM. In the experiments, we also reveal some limitations in existing models to approximate human behavior. This dataset serves as a valuable resource for communities in cognitive psychology, neuroscience, and AI, offering a standardized framework to compare and enhance WM models, investigate WM's neural underpinnings, and develop WM models with human-like capabilities. Our source code and data are available at https://github.com/ZhangLab-DeepNeuroCogLab/WorM."
                    },
                    {
                        "title": "Putting visual object recognition in context",
                        "abstract": "Context plays an important role in visual recognition. Recent studies have shown that visual recognition networks can be fooled by placing objects in inconsistent contexts (e.g., a cow in the ocean). To model the role of contextual information in visual recognition, we systematically investigated ten critical properties of where, when, and how context modulates recognition, including the amount of context, context and object resolution, geometrical structure of context, context congruence, and temporal dynamics of contextual modulation. The tasks involved recognizing a target object surrounded with context in a natural image. As an essential benchmark, we conducted a series of psychophysics experiments where we altered one aspect of context at a time, and quantified recognition accuracy. We propose a biologically-inspired context-aware object recognition model consisting of a two-stream architecture. The model processes visual information at the fovea and periphery in parallel, dynamically incorporates object and contextual information, and sequentially reasons about the class label for the target object. Across a wide range of behavioral tasks, the model approximates human level performance without retraining for each task, captures the dependence of context enhancement on image properties, and provides initial steps towards integrating scene and object information for visual recognition. All source code and data are publicly available: https://github.com/kreimanlab/Put-In-Context."
                    },
                    {
                        "title": "Training-free Object Counting with Prompts",
                        "abstract": "This paper tackles the problem of object counting in images. Existing approaches rely on extensive training data with point annotations for each object, making data collection labor-intensive and time-consuming. To overcome this, we propose a training-free object counter that treats the counting task as a segmentation problem. Our approach leverages the Segment Anything Model (SAM), known for its high-quality masks and zero-shot segmentation capability. However, the vanilla mask generation method of SAM lacks class-specific information in the masks, resulting in inferior counting accuracy. To overcome this limitation, we introduce a prior-guided mask generation method that incorporates three types of priors into the segmentation process, enhancing efficiency and accuracy. Additionally, we tackle the issue of counting objects specified through text by proposing a two-stage approach that combines reference object selection and prior-guided mask generation. Extensive experiments on standard datasets demonstrate the competitive performance of our training-free counter compared to learning-based approaches. This paper presents a promising solution for counting objects in various scenarios without the need for extensive data collection and counting-specific training. Code is available at \\url{https://github.com/shizenglin/training-free-object-counter}"
                    },
                    {
                        "title": "Flow Snapshot Neurons in Action: Deep Neural Networks Generalize to Biological Motion Perception",
                        "abstract": "Biological motion perception (BMP) refers to humans' ability to perceive and recognize the actions of living beings solely from their motion patterns, sometimes as minimal as those depicted on point-light displays. While humans excel at these tasks without any prior training, current AI models struggle with poor generalization performance. To close this research gap, we propose the Motion Perceiver (MP). MP solely relies on patch-level optical flows from video clips as inputs. During training, it learns prototypical flow snapshots through a competitive binding mechanism and integrates invariant motion representations to predict action labels for the given video. During inference, we evaluate the generalization ability of all AI models and humans on 62,656 video stimuli spanning 24 BMP conditions using point-light displays in neuroscience. Remarkably, MP outperforms all existing AI models with a maximum improvement of 29% in top-1 action recognition accuracy on these conditions. Moreover, we benchmark all AI models in point-light displays of two standard video datasets in computer vision. MP also demonstrates superior performance in these cases. More interestingly, via psychophysics experiments, we found that MP recognizes biological movements in a way that aligns with human behavioural data. All data and code will be made public."
                    },
                    {
                        "title": "Object-centric Learning with Cyclic Walks between Parts and Whole",
                        "abstract": "Learning object-centric representations from complex natural environments enables both humans and machines with reasoning abilities from low-level perceptual features. To capture compositional entities of the scene, we proposed cyclic walks between perceptual features extracted from vision transformers and object entities. First, a slot-attention module interfaces with these perceptual features and produces a finite set of slot representations. These slots can bind to any object entities in the scene via inter-slot competitions for attention. Next, we establish entity-feature correspondence with cyclic walks along high transition probability based on the pairwise similarity between perceptual features (aka \"parts\") and slot-binded object representations (aka \"whole\"). The whole is greater than its parts and the parts constitute the whole. The part-whole interactions form cycle consistencies, as supervisory signals, to train the slot-attention module. Our rigorous experiments on \\textit{seven} image datasets in \\textit{three} \\textit{unsupervised} tasks demonstrate that the networks trained with our cyclic walks can disentangle foregrounds and backgrounds, discover objects, and segment semantic objects in complex scenes. In contrast to object-centric models attached with a decoder for the pixel-level or feature-level reconstructions, our cyclic walks provide strong learning signals, avoiding computation overheads and enhancing memory efficiency. Our source code and data are available at: \\href{https://github.com/ZhangLab-DeepNeuroCogLab/Parts-Whole-Object-Centric-Learning/}{link}."
                    },
                    {
                        "title": "Integrating Curricula with Replays: Its Effects on Continual Learning",
                        "abstract": "Humans engage in learning and reviewing processes with curricula when acquiring new skills or knowledge. This human learning behavior has inspired the integration of curricula with replay methods in continual learning agents. The goal is to emulate the human learning process, thereby improving knowledge retention and facilitating learning transfer. Existing replay methods in continual learning agents involve the random selection and ordering of data from previous tasks, which has shown to be effective. However, limited research has explored the integration of different curricula with replay methods to enhance continual learning. Our study takes initial steps in examining the impact of integrating curricula with replay methods on continual learning in three specific aspects: the interleaved frequency of replayed exemplars with training data, the sequence in which exemplars are replayed, and the strategy for selecting exemplars into the replay buffer. These aspects of curricula design align with cognitive psychology principles and leverage the benefits of interleaved practice during replays, easy-to-hard rehearsal, and exemplar selection strategy involving exemplars from a uniform distribution of difficulties. Based on our results, these three curricula effectively mitigated catastrophic forgetting and enhanced positive knowledge transfer, demonstrating the potential of curricula in advancing continual learning methodologies. Our code and data are available: https://github.com/ZhangLab-DeepNeuroCogLab/Integrating-Curricula-with-Replays"
                    },
                    {
                        "title": "TTA-Nav: Test-time Adaptive Reconstruction for Point-Goal Navigation under Visual Corruptions",
                        "abstract": "Robot navigation under visual corruption presents a formidable challenge. To address this, we propose a Test-time Adaptation (TTA) method, named as TTA-Nav, for point-goal navigation under visual corruptions. Our \"plug-and-play\" method incorporates a top-down decoder to a pre-trained navigation model. Firstly, the pre-trained navigation model gets a corrupted image and extracts features. Secondly, the top-down decoder produces the reconstruction given the high-level features extracted by the pre-trained model. Then, it feeds the reconstruction of a corrupted image back to the pre-trained model. Finally, the pre-trained model does forward pass again to output action. Despite being trained solely on clean images, the top-down decoder can reconstruct cleaner images from corrupted ones without the need for gradient-based adaptation. The pre-trained navigation model with our top-down decoder significantly enhances navigation performance across almost all visual corruptions in our benchmarks. Our method improves the success rate of point-goal navigation from the state-of-the-art result of 46% to 94% on the most severe corruption. This suggests its potential for broader application in robotic visual navigation. Project page: https://sites.google.com/view/tta-nav"
                    },
                    {
                        "title": "Lift-the-flap: what, where and when for context reasoning",
                        "abstract": "Context reasoning is critical in a wide variety of applications where current inputs need to be interpreted in the light of previous experience and knowledge. Both spatial and temporal contextual information play a critical role in the domain of visual recognition. Here we investigate spatial constraints (what image features provide contextual information and where they are located), and temporal constraints (when different contextual cues matter) for visual recognition. The task is to reason about the scene context and infer what a target object hidden behind a flap is in a natural image. To tackle this problem, we first describe an online human psychophysics experiment recording active sampling via mouse clicks in lift-the-flap games and identify clicking patterns and features which are diagnostic for high contextual reasoning accuracy. As a proof of the usefulness of these clicking patterns and visual features, we extend a state-of-the-art recurrent model capable of attending to salient context regions, dynamically integrating useful information, making inferences, and predicting class label for the target object over multiple clicks. The proposed model achieves human-level contextual reasoning accuracy, shares human-like sampling behavior and learns interpretable features for contextual reasoning."
                    },
                    {
                        "title": "Make Me Happier: Evoking Emotions Through Image Diffusion Models",
                        "abstract": "Despite the rapid progress in image generation, emotional image editing remains under-explored. The semantics, context, and structure of an image can evoke emotional responses, making emotional image editing techniques valuable for various real-world applications, including treatment of psychological disorders, commercialization of products, and artistic design. For the first time, we present a novel challenge of emotion-evoked image generation, aiming to synthesize images that evoke target emotions while retaining the semantics and structures of the original scenes. To address this challenge, we propose a diffusion model capable of effectively understanding and editing source images to convey desired emotions and sentiments. Moreover, due to the lack of emotion editing datasets, we provide a unique dataset consisting of 340,000 pairs of images and their emotion annotations. Furthermore, we conduct human psychophysics experiments and introduce four new evaluation metrics to systematically benchmark all the methods. Experimental results demonstrate that our method surpasses all competitive baselines. Our diffusion model is capable of identifying emotional cues from original images, editing images that elicit desired emotions, and meanwhile, preserving the semantic structure of the original images. All code, model, and dataset will be made public."
                    },
                    {
                        "title": "Improving generalization by mimicking the human visual diet",
                        "abstract": "We present a new perspective on bridging the generalization gap between biological and computer vision -- mimicking the human visual diet. While computer vision models rely on internet-scraped datasets, humans learn from limited 3D scenes under diverse real-world transformations with objects in natural context. Our results demonstrate that incorporating variations and contextual cues ubiquitous in the human visual training data (visual diet) significantly improves generalization to real-world transformations such as lighting, viewpoint, and material changes. This improvement also extends to generalizing from synthetic to real-world data -- all models trained with a human-like visual diet outperform specialized architectures by large margins when tested on natural image data. These experiments are enabled by our two key contributions: a novel dataset capturing scene context and diverse real-world transformations to mimic the human visual diet, and a transformer model tailored to leverage these aspects of the human visual diet. All data and source code can be accessed at https://github.com/Spandan-Madan/human_visual_diet."
                    },
                    {
                        "title": "Unsupervised Prior Learning: Discovering Categorical Pose Priors from Videos",
                        "abstract": "A prior represents a set of beliefs or assumptions about a system, aiding inference and decision-making. In this work, we introduce the challenge of unsupervised prior learning in pose estimation, where AI models learn pose priors of animate objects from videos in a self-supervised manner. These videos present objects performing various actions, providing crucial information about their keypoints and connectivity. While priors are effective in pose estimation, acquiring them can be difficult. We propose a novel method, named Pose Prior Learner (PPL), to learn general pose priors applicable to any object category. PPL uses a hierarchical memory to store compositional parts of prototypical poses, from which we distill a general pose prior. This prior enhances pose estimation accuracy through template transformation and image reconstruction. PPL learns meaningful pose priors without any additional human annotations or interventions, outperforming competitive baselines on both human and animal pose estimation datasets. Notably, our experimental results reveal the effectiveness of PPL using learnt priors for pose estimation on occluded images. Through iterative inference, PPL leverages priors to refine estimated poses, regressing them to any prototypical poses stored in memory. Our code, model, and data will be publicly available."
                    },
                    {
                        "title": "Reason from Context with Self-supervised Learning",
                        "abstract": "Self-supervised learning (SSL) learns to capture discriminative visual features useful for knowledge transfers. To better accommodate the object-centric nature of current downstream tasks such as object recognition and detection, various methods have been proposed to suppress contextual biases or disentangle objects from contexts. Nevertheless, these methods may prove inadequate in situations where object identity needs to be reasoned from associated context, such as recognizing or inferring tiny or obscured objects. As an initial effort in the SSL literature, we investigate whether and how contextual associations can be enhanced for visual reasoning within SSL regimes, by (a) proposing a new Self-supervised method with external memories for Context Reasoning (SeCo), and (b) introducing two new downstream tasks, lift-the-flap and object priming, addressing the problems of \"what\" and \"where\" in context reasoning. In both tasks, SeCo outperformed all state-of-the-art (SOTA) SSL methods by a significant margin. Our network analysis revealed that the proposed external memory in SeCo learns to store prior contextual knowledge, facilitating target identity inference in the lift-the-flap task. Moreover, we conducted psychophysics experiments and introduced a Human benchmark in Object Priming dataset (HOP). Our results demonstrate that SeCo exhibits human-like behaviors."
                    },
                    {
                        "title": "Variational Prototype Replays for Continual Learning",
                        "abstract": "Continual learning refers to the ability to acquire and transfer knowledge without catastrophically forgetting what was previously learned. In this work, we consider \\emph{few-shot} continual learning in classification tasks, and we propose a novel method, Variational Prototype Replays, that efficiently consolidates and recalls previous knowledge to avoid catastrophic forgetting. In each classification task, our method learns a set of variational prototypes with their means and variances, where embedding of the samples from the same class can be represented in a prototypical distribution and class-representative prototypes are separated apart. To alleviate catastrophic forgetting, our method replays one sample per class from previous tasks, and correspondingly matches newly predicted embeddings to their nearest class-representative prototypes stored from previous tasks. Compared with recent continual learning approaches, our method can readily adapt to new tasks with more classes without requiring the addition of new units. Furthermore, our method is more memory efficient since only class-representative prototypes with their means and variances, as well as only one sample per class from previous tasks need to be stored. Without tampering with the performance on initial tasks, our method learns novel concepts given a few training examples of each class in new tasks."
                    },
                    {
                        "title": "When Pigs Fly: Contextual Reasoning in Synthetic and Natural Scenes",
                        "abstract": "Context is of fundamental importance to both human and machine vision; e.g., an object in the air is more likely to be an airplane than a pig. The rich notion of context incorporates several aspects including physics rules, statistical co-occurrences, and relative object sizes, among others. While previous work has focused on crowd-sourced out-of-context photographs from the web to study scene context, controlling the nature and extent of contextual violations has been a daunting task. Here we introduce a diverse, synthetic Out-of-Context Dataset (OCD) with fine-grained control over scene context. By leveraging a 3D simulation engine, we systematically control the gravity, object co-occurrences and relative sizes across 36 object categories in a virtual household environment. We conducted a series of experiments to gain insights into the impact of contextual cues on both human and machine vision using OCD. We conducted psychophysics experiments to establish a human benchmark for out-of-context recognition, and then compared it with state-of-the-art computer vision models to quantify the gap between the two. We propose a context-aware recognition transformer model, fusing object and contextual information via multi-head attention. Our model captures useful information for contextual reasoning, enabling human-level performance and better robustness in out-of-context conditions compared to baseline models across OCD and other out-of-context datasets. All source code and data are publicly available at https://github.com/kreimanlab/WhenPigsFlyContext"
                    },
                    {
                        "title": "Efficient Zero-shot Visual Search via Target and Context-aware Transformer",
                        "abstract": "Visual search is a ubiquitous challenge in natural vision, including daily tasks such as finding a friend in a crowd or searching for a car in a parking lot. Human rely heavily on relevant target features to perform goal-directed visual search. Meanwhile, context is of critical importance for locating a target object in complex scenes as it helps narrow down the search area and makes the search process more efficient. However, few works have combined both target and context information in visual search computational models. Here we propose a zero-shot deep learning architecture, TCT (Target and Context-aware Transformer), that modulates self attention in the Vision Transformer with target and contextual relevant information to enable human-like zero-shot visual search performance. Target modulation is computed as patch-wise local relevance between the target and search images, whereas contextual modulation is applied in a global fashion. We conduct visual search experiments on TCT and other competitive visual search models on three natural scene datasets with varying levels of difficulty. TCT demonstrates human-like performance in terms of search efficiency and beats the SOTA models in challenging visual search tasks. Importantly, TCT generalizes well across datasets with novel objects without retraining or fine-tuning. Furthermore, we also introduce a new dataset to benchmark models for invariant visual search under incongruent contexts. TCT manages to search flexibly via target and context modulation, even under incongruent contexts."
                    },
                    {
                        "title": "Tuned Compositional Feature Replays for Efficient Stream Learning",
                        "abstract": "Our brains extract durable, generalizable knowledge from transient experiences of the world. Artificial neural networks come nowhere close to this ability. When tasked with learning to classify objects by training on non-repeating video frames in temporal order (online stream learning), models that learn well from shuffled datasets catastrophically forget old knowledge upon learning new stimuli. We propose a new continual learning algorithm, Compositional Replay Using Memory Blocks (CRUMB), which mitigates forgetting by replaying feature maps reconstructed by combining generic parts. CRUMB concatenates trainable and re-usable \"memory block\" vectors to compositionally reconstruct feature map tensors in convolutional neural networks. Storing the indices of memory blocks used to reconstruct new stimuli enables memories of the stimuli to be replayed during later tasks. This reconstruction mechanism also primes the neural network to minimize catastrophic forgetting by biasing it towards attending to information about object shapes more than information about image textures, and stabilizes the network during stream learning by providing a shared feature-level basis for all training examples. These properties allow CRUMB to outperform an otherwise identical algorithm that stores and replays raw images, while occupying only 3.6% as much memory. We stress-tested CRUMB alongside 13 competing methods on 7 challenging datasets. To address the limited number of existing online stream learning datasets, we introduce 2 new benchmarks by adapting existing datasets for stream learning. With only 3.7-4.1% as much memory and 15-43% as much runtime, CRUMB mitigates catastrophic forgetting more effectively than the state-of-the-art. Our code is available at https://github.com/MorganBDT/crumb.git."
                    }
                ]
            }
        }
    },
    "2312.11846": {
        "paper_data": {
            "title": "Initializing Services in Interactive ML Systems for Diverse Users",
            "url": "http://arxiv.org/abs/2312.11846v1",
            "arxiv_id": "2312.11846",
            "authors": [
                "Avinandan Bose",
                "Mihaela Curmei",
                "Daniel L. Jiang",
                "Jamie Morgenstern",
                "Sarah Dean",
                "Lillian J. Ratliff",
                "Maryam Fazel"
            ],
            "abstract": "This paper studies ML systems that interactively learn from users across multiple subpopulations with heterogeneous data distributions. The primary objective is to provide specialized services for different user groups while also predicting user preferences. Once the users select a service based on how well the service anticipated their preference, the services subsequently adapt and refine themselves based on the user data they accumulate, resulting in an iterative, alternating minimization process between users and services (learning dynamics). Employing such tailored approaches has two main challenges: (i) Unknown user preferences: Typically, data on user preferences are unavailable without interaction, and uniform data collection across a large and diverse user base can be prohibitively expensive. (ii) Suboptimal Local Solutions: The total loss (sum of loss functions across all users and all services) landscape is not convex even if the individual losses on a single service are convex, making it likely for the learning dynamics to get stuck in local minima. The final outcome of the aforementioned learning dynamics is thus strongly influenced by the initial set of services offered to users, and is not guaranteed to be close to the globally optimal outcome. In this work, we propose a randomized algorithm to adaptively select very few users to collect preference data from, while simultaneously initializing a set of services. We prove that under mild assumptions on the loss functions, the expected total loss achieved by the algorithm right after initialization is within a factor of the globally optimal total loss with complete user preference data, and this factor scales only logarithmically in the number of services. Our theory is complemented by experiments on real as well as semi-synthetic datasets.",
            "introduction": "   1 Introduction  In various applications including federated learning (Li et\u00a0al., 2020), crowd-sourcing (Steinhardt et\u00a0al., 2016), online recommendation systems (Song et\u00a0al., 2014), data about user preferences is acquired through interactions with multiple users or sources. The primary objective is to leverage this data to deliver services that cater to the diverse requirements of users. Typically, these users exhibit heterogeneity in their preferences, and recent approaches demonstrate the benefits of deploying and updating multiple, more specialized models versus using one large shared model as a one-size fits all approach; e.g., for clustered federated learning Mansour et\u00a0al. (2020); Sattler et\u00a0al. (2020); Ghosh et\u00a0al. (2020) (Dwork et\u00a0al., 2018), in the context of fair classifiers (Ustun et\u00a0al., 2019), and for meta learning (Kong et\u00a0al., 2020). And even when a single large model is used, it is often followed by a fine-tuning step for specialization to a group of users or tasks by retraining part of the model (Chua et\u00a0al., 2021; Sun et\u00a0al., 2021).   In this paper we consider ML systems that interactively learn from users across multiple subpopulations with heterogeneous data distributions. The aim is to provide a range of services, each service specialized for a group of users. These services typically work by trying to predict user preferences, and the likelihood of any user picking a service from the available candidates depends on the services\u2019 performance for them. For example, Meta111https://nickclegg.medium.com/you-and-the-algorithm-it-takes-two-to-tango-7722b19aa1c2 uses ML to predict user preferences and make personalized recommendations for its services like Facebook, Instagram, etc. Users decide whether to join recommended pages, leave likes or other forms of engagement such as scrolling activity, hours online indicating their level of satisfaction with the services. Meta would in turn plan which users each of its platforms is tailored to, and what personalized content is recommended to users\u2014these decisions by Meta could altogether be modeled as choosing the parameters of its services.    These services are usually trained through interaction with users. Once a set of services is launched users select the service that best predicts their preferences (generally choosing the service that minimizes the particular user\u2019s performance metric). These services are then subsequently refined based on data of current users. This sequence of (alternating) decisions by users and by services (learning dynamics) continues as more data is gathered. Broadly speaking, such a learning system could include a set of independent learners (e.g., different service providers) having an incentive to coordinate for the purpose of initializing their services, or a single learner with several deployed models (e.g., a company that owns different platforms, or a multi-model server in certain federated learning setups). This setup has been studied by Ginart et\u00a0al. (2021), who discuss the effect of competition between services on user welfare, and Dean et\u00a0al. (2022), who discuss the the stability of equilibria of learning dynamics in such systems.   Our goal is to design a set of services to minimize the total loss, i.e., the sum of all losses across users. However, we face the following challenges:   \u2022  Unknown User Preferences: In practice, the presence of a service often precedes data collection. Specifically, in contexts like online recommendations, it is usually not feasible to gather user preference data without first deploying the",
            "references": [
                {
                    "title": "Learning Preference Distributions From Distance Measurements",
                    "abstract": "We introduce the problem of learning a distribution of user preferences over a set of items from noisy responses to distance queries. Rather than aiming to learn the preferences of each user, our goal is only to recover the overall distribution of user preferences. We show that distribution recovery can require just one response from each user. In contrast, learning the preferences of each user would require multiple responses from each user. Thus, learning preference distributions, rather than individual preferences, may be more practical in many applications. The preference distribution problem is formulated on a discrete domain in which items (e.g., products) and users' ideal preference points are located. We study both the noiseless and noisy settings in one dimension and provide sufficient conditions for identifiability of the underlying true distribution as a function of the set of items used for queries. We establish an upper bound on the total variation distance between the true distribution and the distribution learned via constrained least squares optimization problem for both noiseless and noisy settings. While the one-dimensional setting we consider is simple, our simulation results show that our proposed recovery technique extends to multidimensional settings and graph structures."
                },
                {
                    "title": "One for All: Simultaneous Metric and Preference Learning over Multiple Users",
                    "abstract": "This paper investigates simultaneous preference and metric learning from a crowd of respondents. A set of items represented by $d$-dimensional feature vectors and paired comparisons of the form ``item $i$ is preferable to item $j$'' made by each user is given. Our model jointly learns a distance metric that characterizes the crowd's general measure of item similarities along with a latent ideal point for each user reflecting their individual preferences. This model has the flexibility to capture individual preferences, while enjoying a metric learning sample cost that is amortized over the crowd. We first study this problem in a noiseless, continuous response setting (i.e., responses equal to differences of item distances) to understand the fundamental limits of learning. Next, we establish prediction error guarantees for noisy, binary measurements such as may be collected from human respondents, and show how the sample complexity improves when the underlying metric is low-rank. Finally, we establish recovery guarantees under assumptions on the response distribution. We demonstrate the performance of our model on both simulated data and on a dataset of color preference judgements across a large number of users."
                },
                {
                    "title": "Scalable Distributional Robustness in a Class of Non Convex Optimization with Guarantees",
                    "abstract": "Distributionally robust optimization (DRO) has shown lot of promise in providing robustness in learning as well as sample based optimization problems. We endeavor to provide DRO solutions for a class of sum of fractionals, non-convex optimization which is used for decision making in prominent areas such as facility location and security games. In contrast to previous work, we find it more tractable to optimize the equivalent variance regularized form of DRO rather than the minimax form. We transform the variance regularized form to a mixed-integer second order cone program (MISOCP), which, while guaranteeing near global optimality, does not scale enough to solve problems with real world data-sets. We further propose two abstraction approaches based on clustering and stratified sampling to increase scalability, which we then use for real world data-sets. Importantly, we provide near global optimality guarantees for our approach and show experimentally that our solution quality is better than the locally optimal ones achieved by state-of-the-art gradient-based methods. We experimentally compare our different approaches and baselines, and reveal nuanced properties of a DRO solution."
                },
                {
                    "title": "Towards Sample-efficient Overparameterized Meta-learning",
                    "abstract": "An overarching goal in machine learning is to build a generalizable model with few samples. To this end, overparameterization has been the subject of immense interest to explain the generalization ability of deep nets even when the size of the dataset is smaller than that of the model. While the prior literature focuses on the classical supervised setting, this paper aims to demystify overparameterization for meta-learning. Here we have a sequence of linear-regression tasks and we ask: (1) Given earlier tasks, what is the optimal linear representation of features for a new downstream task? and (2) How many samples do we need to build this representation? This work shows that surprisingly, overparameterization arises as a natural answer to these fundamental meta-learning questions. Specifically, for (1), we first show that learning the optimal representation coincides with the problem of designing a task-aware regularization to promote inductive bias. We leverage this inductive bias to explain how the downstream task actually benefits from overparameterization, in contrast to prior works on few-shot learning. For (2), we develop a theory to explain how feature covariance can implicitly help reduce the sample complexity well below the degrees of freedom and lead to small estimation error. We then integrate these findings to obtain an overall performance guarantee for our meta-learning algorithm. Numerical experiments on real and synthetic data verify our insights on overparameterized meta-learning."
                },
                {
                    "title": "Retiring Adult: New Datasets for Fair Machine Learning",
                    "abstract": "Although the fairness community has recognized the importance of data, researchers in the area primarily rely on UCI Adult when it comes to tabular data. Derived from a 1994 US Census survey, this dataset has appeared in hundreds of research papers where it served as the basis for the development and comparison of many algorithmic fairness interventions. We reconstruct a superset of the UCI Adult data from available US Census sources and reveal idiosyncrasies of the UCI Adult dataset that limit its external validity. Our primary contribution is a suite of new datasets derived from US Census surveys that extend the existing data ecosystem for research on fair machine learning. We create prediction tasks relating to income, employment, health, transportation, and housing. The data span multiple years and all states of the United States, allowing researchers to study temporal shift and geographic variation. We highlight a broad initial sweep of new empirical insights relating to trade-offs between fairness criteria, performance of algorithmic interventions, and the role of distribution shift based on our new datasets. Our findings inform ongoing debates, challenge some existing narratives, and point to future research directions. Our datasets are available at https://github.com/zykls/folktables."
                },
                {
                    "title": "How Fine-Tuning Allows for Effective Meta-Learning",
                    "abstract": "Representation learning has been widely studied in the context of meta-learning, enabling rapid learning of new tasks through shared representations. Recent works such as MAML have explored using fine-tuning-based metrics, which measure the ease by which fine-tuning can achieve good performance, as proxies for obtaining representations. We present a theoretical framework for analyzing representations derived from a MAML-like algorithm, assuming the available tasks use approximately the same underlying representation. We then provide risk bounds on the best predictor found by fine-tuning via gradient descent, demonstrating that the algorithm can provably leverage the shared structure. The upper bound applies to general function classes, which we demonstrate by instantiating the guarantees of our framework in the logistic regression and neural network settings. In contrast, we establish the existence of settings where any algorithm, using a representation trained with no consideration for task-specific fine-tuning, performs as well as a learner with no access to source tasks in the worst case. This separation result underscores the benefit of fine-tuning-based methods, such as MAML, over methods with\"frozen representation\"objectives in few-shot learning."
                },
                {
                    "title": "Socially Fair k-Means Clustering",
                    "abstract": "We show that the popular k-means clustering algorithm (Lloyd's heuristic), used for a variety of scientific data, can result in outcomes that are unfavorable to subgroups of data (e.g., demographic groups). Such biased clusterings can have deleterious implications for human-centric applications such as resource allocation. We present a fair k-means objective and algorithm to choose cluster centers that provide equitable costs for different groups. The algorithm, Fair-Lloyd, is a modification of Lloyd's heuristic for k-means, inheriting its simplicity, efficiency, and stability. In comparison with standard Lloyd's, we find that on benchmark datasets, Fair-Lloyd exhibits unbiased performance by ensuring that all groups have equal costs in the output k-clustering, while incurring a negligible increase in running time, thus making it a viable fair option wherever k-means is currently used."
                },
                {
                    "title": "Improved Guarantees for k-means++ and k-means++ Parallel",
                    "abstract": "In this paper, we study k-means++ and k-means++ parallel, the two most popular algorithms for the classic k-means clustering problem. We provide novel analyses and show improved approximation and bi-criteria approximation guarantees for k-means++ and k-means++ parallel. Our results give a better theoretical justification for why these algorithms perform extremely well in practice. We also propose a new variant of k-means++ parallel algorithm (Exponential Race k-means++) that has the same approximation guarantees as k-means++."
                },
                {
                    "title": "High-Dimensional Probability: An Introduction with Applications in Data Science",
                    "abstract": "\u00a9 2018, Cambridge University Press Let us summarize our findings. A random projection of a set T in R n onto an m-dimensional subspace approximately preserves the geometry of T if m \u2a86 d ( T ) . For..."
                },
                {
                    "title": "Competing AI: How does competition feedback affect machine learning?",
                    "abstract": "This papers studies how competition affects machine learning (ML) predictors. As ML becomes more ubiquitous, it is often deployed by companies to compete over customers. For example, digital platforms like Yelp use ML to predict user preference and make recommendations. A service that is more often queried by users, perhaps because it more accurately anticipates user preferences, is also more likely to obtain additional user data (e.g. in the form of a Yelp review). Thus, competing predictors cause feedback loops whereby a predictor's performance impacts what training data it receives and biases its predictions over time. We introduce a flexible model of competing ML predictors that enables both rapid experimentation and theoretical tractability. We show with empirical and mathematical analysis that competition causes predictors to specialize for specific sub-populations at the cost of worse performance over the general population. We further analyze the impact of predictor specialization on the overall prediction quality experienced by users. We show that having too few or too many competing predictors in a market can hurt the overall prediction quality. Our theory is complemented by experiments on several real datasets using popular learning algorithms, such as neural networks and nearest neighbor methods."
                },
                {
                    "title": "Surprise: A Python library for recommender systems",
                    "abstract": "Recommender systems aim at providing users with a list of recommendations of items that a service offers. For example, a video streaming service will typically rely on a recommender system to propose a personalized list of movies or series to each of its users. A typical problem in recommendation is that of rating prediction: given an incomplete dataset of useritem interations which take the form of numerical ratings (e.g. on a scale from 1 to 5), the goal is to predict the missing ratings for all remaining user-item pairs."
                },
                {
                    "title": "An Efficient Framework for Clustered Federated Learning",
                    "abstract": "We address the problem of federated learning (FL) where users are distributed and partitioned into clusters. This setup captures settings where different groups of users have their own objectives (learning tasks) but by aggregating their data with others in the same cluster (same learning task), they can leverage the strength in numbers in order to perform more efficient federated learning. For this new framework of clustered federated learning, we propose the Iterative Federated Clustering Algorithm (IFCA), which alternately estimates the cluster identities of the users and optimizes model parameters for the user clusters via gradient descent. We analyze the convergence rate of this algorithm first in a linear model with squared loss and then for generic strongly convex and smooth loss functions. We show that in both settings, with good initialization, IFCA is guaranteed to converge, and discuss the optimality of the statistical error rate. In particular, for the linear model with two clusters, we can guarantee that our algorithm converges as long as the initialization is slightly better than random. When the clustering structure is ambiguous, we propose to train the models by combining IFCA with the weight sharing technique in multi-task learning. In the experiments, we show that our algorithm can succeed even if we relax the requirements on initialization with random initialization and multiple restarts. We also present experimental results showing that our algorithm is efficient in non-convex problems such as neural networks. We demonstrate the benefits of IFCA over the baselines on several clustered FL benchmarks."
                },
                {
                    "title": "Three Approaches for Personalization with Applications to Federated Learning",
                    "abstract": "The standard objective in machine learning is to train a single model for all users. However, in many learning scenarios, such as cloud computing and federated learning, it is possible to learn a personalized model per user. In this work, we present a systematic learning-theoretic study of personalization. We propose and analyze three approaches: user clustering, data interpolation, and model interpolation. For all three approaches, we provide learning-theoretic guarantees and efficient algorithms for which we also demonstrate the performance empirically. All of our algorithms are model-agnostic and work for any hypothesis class."
                },
                {
                    "title": "Meta-learning for mixed linear regression",
                    "abstract": "In modern supervised learning, there are a large number of tasks, but many of them are associated with only a small amount of labeled data. These include data from medical image processing and robotic interaction. Even though each individual task cannot be meaningfully trained in isolation, one seeks to meta-learn across the tasks from past experiences by exploiting some similarities. We study a fundamental question of interest: When can abundant tasks with small data compensate for lack of tasks with big data? We focus on a canonical scenario where each task is drawn from a mixture of $k$ linear regressions, and identify sufficient conditions for such a graceful exchange to hold; The total number of examples necessary with only small data tasks scales similarly as when big data tasks are available. To this end, we introduce a novel spectral approach and show that we can efficiently utilize small data tasks with the help of $\\tilde\\Omega(k^{3/2})$ medium data tasks each with $\\tilde\\Omega(k^{1/2})$ examples."
                },
                {
                    "title": "Clustered Federated Learning: Model-Agnostic Distributed Multitask Optimization Under Privacy Constraints",
                    "abstract": "Federated learning (FL) is currently the most widely adopted framework for collaborative training of (deep) machine learning models under privacy constraints. Albeit its popularity, it has been observed that FL yields suboptimal results if the local clients\u2019 data distributions diverge. To address this issue, we present clustered FL (CFL), a novel federated multitask learning (FMTL) framework, which exploits geometric properties of the FL loss surface to group the client population into clusters with jointly trainable data distributions. In contrast to existing FMTL approaches, CFL does not require any modifications to the FL communication protocol to be made, is applicable to general nonconvex objectives (in particular, deep neural networks), does not require the number of clusters to be known a priori, and comes with strong mathematical guarantees on the clustering quality. CFL is flexible enough to handle client populations that vary over time and can be implemented in a privacy-preserving way. As clustering is only performed after FL has converged to a stationary point, CFL can be viewed as a postprocessing method that will always achieve greater or equal performance than conventional FL by allowing clients to arrive at more specialized models. We verify our theoretical analysis in experiments with deep convolutional and recurrent neural networks on commonly used FL data sets."
                },
                {
                    "title": "Federated Learning: Challenges, Methods, and Future Directions",
                    "abstract": "Federated learning involves training statistical models over remote devices or siloed data centers, such as mobile phones or hospitals, while keeping data localized. Training in heterogeneous and potentially massive networks introduces novel challenges that require a fundamental departure from standard approaches for large-scale machine learning, distributed optimization, and privacy-preserving data analysis. In this article, we discuss the unique characteristics and challenges of federated learning, provide a broad overview of current approaches, and outline several directions of future work that are relevant to a wide range of research communities."
                },
                {
                    "title": "Fairness without Harm: Decoupled Classifiers with Preference Guarantees",
                    "abstract": "In domains such as medicine, it can be acceptable for machine learning models to include sensitive attributes such as gender and ethnicity. In this work, we argue that when there is this kind of treatment disparity then it should be in the best interest of each group. Drawing on ethical principles such as bene\ufb01cence (\u201cdo the best\u201d) and non-male\ufb01cence (\u201cdo no harm\u201d), we show how to use sensitive attributes to train decoupled clas-si\ufb01ers that satisfy preference guarantees . These guarantees ensure the majority of individuals in each group prefer their assigned classi\ufb01er to (i) a pooled model that ignores group membership ( rationality ), and (ii) the model assigned to any other group ( envy-freeness ). We introduce a recursive procedure that adaptively selects group attributes for decoupling, and present formal conditions to ensure preference guarantees in terms of generalization error. We validate the effectiveness of the procedure on real-world datasets, showing that it improves accuracy without violating preference guarantees on test data."
                },
                {
                    "title": "Group Retention when Using Machine Learning in Sequential Decision Making: the Interplay between User Dynamics and Fairness",
                    "abstract": "Machine Learning (ML) models trained on data from multiple demographic groups can inherit representation disparity (Hashimoto et al., 2018) that may exist in the data: the model may be less favorable to groups contributing less to the training process; this in turn can degrade population retention in these groups over time, and exacerbate representation disparity in the long run. In this study, we seek to understand the interplay between ML decisions and the underlying group representation, how they evolve in a sequential framework, and how the use of fairness criteria plays a role in this process. We show that the representation disparity can easily worsen over time under a natural user dynamics (arrival and departure) model when decisions are made based on a commonly used objective and fairness criteria, resulting in some groups diminishing entirely from the sample pool in the long run. It highlights the fact that fairness criteria have to be defined while taking into consideration the impact of decisions on user dynamics. Toward this end, we explain how a proper fairness criterion can be selected based on a general user dynamics model."
                },
                {
                    "title": "Eigenvalue and Generalized Eigenvalue Problems: Tutorial",
                    "abstract": "This paper is a tutorial for eigenvalue and generalized eigenvalue problems. We first introduce eigenvalue problem, eigen-decomposition (spectral decomposition), and generalized eigenvalue problem. Then, we mention the optimization problems which yield to the eigenvalue and generalized eigenvalue problems. We also provide examples from machine learning, including principal component analysis, kernel supervised principal component analysis, and Fisher discriminant analysis, which result in eigenvalue and generalized eigenvalue problems. Finally, we introduce the solutions to both eigenvalue and generalized eigenvalue problems."
                },
                {
                    "title": "Fairness Without Demographics in Repeated Loss Minimization",
                    "abstract": "Machine learning models (e.g., speech recognizers) are usually trained to minimize average loss, which results in representation disparity---minority groups (e.g., non-native speakers) contribute less to the training objective and thus tend to suffer higher loss. Worse, as model accuracy affects user retention, a minority group can shrink over time. In this paper, we first show that the status quo of empirical risk minimization (ERM) amplifies representation disparity over time, which can even make initially fair models unfair. To mitigate this, we develop an approach based on distributionally robust optimization (DRO), which minimizes the worst case risk over all distributions close to the empirical distribution. We prove that this approach controls the risk of the minority group at each time step, in the spirit of Rawlsian distributive justice, while remaining oblivious to the identity of the groups. We demonstrate that DRO prevents disparity amplification on examples where ERM fails, and show improvements in minority group user satisfaction in a real-world text autocomplete task."
                },
                {
                    "title": "Learning User Preferences to Incentivize Exploration in the Sharing Economy",
                    "abstract": "\n \n We study platforms in the sharing economy and discuss the need for incentivizing users to explore options that otherwise would not be chosen. For instance, rental platforms such as Airbnb typically rely on customer reviews to provide users with relevant information about different options. Yet, often a large fraction of options does not have any reviews available. Such options are frequently neglected as viable choices, and in turn are unlikely to be evaluated, creating a vicious cycle. Platforms can engage users to deviate from their preferred choice by offering monetary incentives for choosing a different option instead. To efficiently learn the optimal incentives to offer, we consider structural information in user preferences and introduce a novel algorithm---Coordinated Online Learning (CoOL)---for learning with structural information modeled as convex constraints. We provide formal guarantees on the performance of our algorithm and test the viability of our approach in a user study with data of apartments on Airbnb. Our findings suggest that our approach is well-suited to learn appropriate incentives and increase exploration on the investigated platform.\n \n"
                },
                {
                    "title": "Decoupled Classifiers for Group-Fair and Efficient Machine Learning",
                    "abstract": "When it is ethical and legal to use a sensitive attribute (such as gender or race) in machine learning systems, the question remains how to do so. We show that the na\u00a8\u0131ve application of machine learning algorithms using sensitive attributes leads to an inherent tradeo\ufb00 in accuracy between groups. We provide a simple and e\ufb03cient decoupling technique, which can be added on top of any black-box machine learning algorithm, to learn di\ufb00erent classi\ufb01ers for di\ufb00erent groups. Transfer learning is used to mitigate the problem of having too little data on any one group."
                },
                {
                    "title": "Better Guarantees for k-Means and Euclidean k-Median by Primal-Dual Algorithms",
                    "abstract": "Clustering is a classic topic in optimization with k-means being one of the most fundamental such problems. In the absence of any restrictions on the input, the best known algorithm for k-means with a provable guarantee is a simple local search heuristic yielding an approximation guarantee of 9+&#x2265;ilon, a ratio that is known to be tight with respect to such methods.We overcome this barrier by presenting a new primal-dual approach that allows us to (1) exploit the geometric structure of k-means and (2) to satisfy the hard constraint that at most k clusters are selected without deteriorating the approximation guarantee. Our main result is a 6.357-approximation algorithm with respect to the standard LP relaxation. Our techniques are quite general and we also show improved guarantees for the general version of k-means where the underlying metric is not required to be Euclidean and for k-median in Euclidean metrics."
                },
                {
                    "title": "Avoiding Imposters and Delinquents: Adversarial Crowdsourcing and Peer Prediction",
                    "abstract": "We consider a crowdsourcing model in which $n$ workers are asked to rate the quality of $n$ items previously generated by other workers. An unknown set of $\\alpha n$ workers generate reliable ratings, while the remaining workers may behave arbitrarily and possibly adversarially. The manager of the experiment can also manually evaluate the quality of a small number of items, and wishes to curate together almost all of the high-quality items with at most an $\\epsilon$ fraction of low-quality items. Perhaps surprisingly, we show that this is possible with an amount of work required of the manager, and each worker, that does not scale with $n$: the dataset can be curated with $\\tilde{O}\\Big(\\frac{1}{\\beta\\alpha^3\\epsilon^4}\\Big)$ ratings per worker, and $\\tilde{O}\\Big(\\frac{1}{\\beta\\epsilon^2}\\Big)$ ratings by the manager, where $\\beta$ is the fraction of high-quality items. Our results extend to the more general setting of peer prediction, including peer grading in online classrooms."
                },
                {
                    "title": "Online Learning in Large-Scale Contextual Recommender Systems",
                    "abstract": "In this paper, we propose a novel large-scale, context-aware recommender system that provides accurate recommendations, scalability to a large number of diverse users and items, differential services, and does not suffer from \u201ccold start\u201d problems. Our proposed recommendation system relies on a novel algorithm which learns online the item preferences of users based on their click behavior, and constructs online item-cluster trees. The recommendations are then made by choosing an item-cluster level and then selecting an item within that cluster as a recommendation for the user. This approach is able to significantly improve the learning speed when the number of users and items is large, while still providing high recommendation accuracy. Each time a user arrives at the website, the system makes a recommendation based on the estimations of item payoffs by exploiting past context arrivals in a neighborhood of the current user's context. It exploits the similarity of contexts to learn how to make better recommendations even when the number and diversity of users and items is large. This also addresses the cold start problem by using the information gained from similar users and items to make recommendations for new users and items. We theoretically prove that the proposed algorithm for item recommendations converges to the optimal item recommendations in the long-run. We also bound the probability of making a suboptimal item recommendation for each user arriving to the system while the system is learning. Experimental results show that our approach outperforms the state-of-the-art algorithms by over 20 percent in terms of click through rates."
                },
                {
                    "title": "The MovieLens Datasets: History and Context",
                    "abstract": "The MovieLens datasets are widely used in education, research, and industry. They are downloaded hundreds of thousands of times each year, reflecting their use in popular press programming books, traditional and online courses, and software. These datasets are a product of member activity in the MovieLens movie recommendation system, an active research platform that has hosted many experiments since its launch in 1997. This article documents the history of MovieLens and the MovieLens datasets. We include a discussion of lessons learned from running a long-standing, live research platform from the perspective of a research organization. We document best practices and limitations of using the MovieLens datasets in new research."
                },
                {
                    "title": "Tight Bounds for the Expected Risk of Linear Classifiers and PAC-Bayes Finite-Sample Guarantees",
                    "abstract": "We analyze the expected risk of linear classiers for a xed weight vector in the \\minimax\" setting. That is, we analyze the worst-case risk among all data distributions with a given mean and covariance. We provide a simpler proof of the tight polynomial-tail bound for general random variables. For sub-Gaussian random variables, we derive a novel tight exponentialtail bound. We also provide new PAC-Bayes nite-sample guarantees when training data is available. Our \\minimax\" generalization bounds are dimensionality-independent and O( p 1=m) for m samples."
                },
                {
                    "title": "k-means++: the advantages of careful seeding",
                    "abstract": "The k-means method is a widely used clustering technique that seeks to minimize the average squared distance between points in the same cluster. Although it offers no accuracy guarantees, its simplicity and speed are very appealing in practice. By augmenting k-means with a very simple, randomized seeding technique, we obtain an algorithm that is \u0398(logk)-competitive with the optimal clustering. Preliminary experiments show that our augmentation improves both the speed and the accuracy of k-means, often quite dramatically."
                },
                {
                    "title": "The Effectiveness of Lloyd-Type Methods for the k-Means Problem",
                    "abstract": "We investigate variants of Lloyd's heuristic for clustering high dimensional data in an attempt to explain its popularity (a half century after its introduction) among practitioners, and in order to suggest improvements in its application. We propose and justify a clusterability criterion for data sets. We present variants of Lloyd's heuristic that quickly lead to provably near-optimal clustering solutions when applied to well-clusterable instances. This is the first performance guarantee for a variant of Lloyd's heuristic. The provision of a guarantee on output quality does not come at the expense of speed: some of our algorithms are candidates for being faster in practice than currently used variants of Lloyd's method. In addition, our other algorithms are faster on well-clusterable instances than recently proposed approximation algorithms, while maintaining similar guarantees on clustering quality. Our main algorithmic contribution is a novel probabilistic seeding process for the starting configuration of a Lloyd-type iteration"
                },
                {
                    "title": "A local search approximation algorithm for k-means clustering",
                    "abstract": "In k-means clustering we are given a set of n data points in d-dimensional space Rd and an integer k, and the problem is to determine a set of k points in \u00d3C;d, called centers, to minimize the mean squared distance from each data point to its nearest center. No exact polynomial-time algorithms are known for this problem. Although asymptotically efficient approximation algorithms exist, these algorithms are not practical due to the extremely high constant factors involved. There are many heuristics that are used in practice, but we know of no bounds on their performance.We consider the question of whether there exists a simple and practical approximation algorithm for k-means clustering. We present a local improvement heuristic based on swapping centers in and out. We prove that this yields a (9+\u03b5)-approximation algorithm. We show that the approximation factor is almost tight, by giving an example for which the algorithm achieves an approximation factor of (9-\u03b5). To establish the practical value of the heuristic, we present an empirical study that shows that, when combined with Lloyd's algorithm, this heuristic performs quite well in practice."
                },
                {
                    "title": "Least squares quantization in PCM",
                    "abstract": "It has long been realized that in pulse-code modulation (PCM), with a given ensemble of signals to handle, the quantum values should be spaced more closely in the voltage regions where the signal amplitude is more likely to fall. It has been shown by Panter and Dite that, in the limit as the number of quanta becomes infinite, the asymptotic fractional density of quanta per unit voltage should vary as the one-third power of the probability density per unit voltage of signal amplitudes. In this paper the corresponding result for any finite number of quanta is derived; that is, necessary conditions are found that the quanta and associated quantization intervals of an optimum finite quantization scheme must satisfy. The optimization criterion used is that the average quantization noise power be a minimum. It is shown that the result obtained here goes over into the Panter and Dite result as the number of quanta become large. The optimum quautization schemes for 2^{b} quanta, b=1,2, \\cdots, 7 , are given numerically for Gaussian and for Laplacian distribution of signal amplitudes."
                },
                {
                    "title": "Multi-learner risk reduction under endogenous participation dynamics",
                    "abstract": ","
                },
                {
                    "title": "The hardness of k-means clustering",
                    "abstract": "We show that k-means clustering is an NP-hard optimization problem, even if k is fixed to 2."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively design and optimize multiple specialized machine learning services that learn from heterogeneous user preferences across different subpopulations in real-time?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of personalized machine learning systems, particularly in contexts like federated learning and online recommendations. By addressing the challenge of optimizing services for diverse user preferences, we can enhance user satisfaction and engagement, leading to more effective and tailored applications. This research could pave the way for future studies on user-centric models, improve the efficiency of service delivery, and foster innovations in adaptive learning systems that respond dynamically to user interactions.\n\n**[Question 3] - Why is it hard?**  \nThe complexity of this problem arises from the need to manage unknown user preferences, which complicates the initial deployment of services. Naive approaches may fail because they do not account for the dynamic nature of user interactions and the heterogeneity of user data distributions. Additionally, technical challenges include ensuring that the learning dynamics between users and services converge to a stable solution, as well as the need for effective coordination among independent learners or models. Theoretical obstacles involve understanding the equilibrium states of these learning systems and how competition among services affects user welfare.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either single large models or independent service providers without adequately addressing the complexities of user preference heterogeneity and the interactive learning dynamics involved. Limitations in existing solutions include a lack of frameworks that can simultaneously optimize multiple services while adapting to real-time user feedback. Barriers such as insufficient data collection methods prior to service deployment and the challenges of modeling user behavior in a competitive environment have hindered progress. Our approach differs by integrating a framework that accounts for these dynamics and emphasizes the importance of specialized services tailored to distinct user groups.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a multi-service framework that utilizes reinforcement learning to adaptively optimize service parameters based on user interactions. We will employ a diverse dataset that captures user preferences across various subpopulations and use metrics such as total loss and user engagement rates to evaluate performance. The expected outcomes include a set of optimized services that minimize total loss across users while effectively predicting and catering to their preferences, ultimately leading to improved user satisfaction and engagement in real-time applications."
            }
        },
        "author_data": {
            "d119548c-e0cc-46bc-ab86-01036dc6634d": {
                "pk": "d119548c-e0cc-46bc-ab86-01036dc6634d",
                "name": "Avinandan Bose",
                "collaborators": [
                    "Arunesh Sinha",
                    "Tien Mai",
                    "Pradeep Varakantham",
                    "Simon Shaolei Du",
                    "Maryam Fazel",
                    "Aniket Das",
                    "Yatin Dandi",
                    "Piyush Rai",
                    "Soumendu Sundar Mukherjee",
                    "Changyu Chen"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Optimization",
                    "Generative Models",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Conditional Expectation based Value Decomposition for Scalable On-Demand Ride Pooling",
                        "abstract": "Owing to the benefits for customers (lower prices), drivers (higher revenues), aggregation companies (higher revenues) and the environment (fewer vehicles), on-demand ride pooling (e.g., Uber pool, Grab Share) has become quite popular. The significant computational complexity of matching vehicles to combinations of requests has meant that traditional ride pooling approaches are myopic in that they do not consider the impact of current matches on future value for vehicles/drivers. Recently, Neural Approximate Dynamic Programming (NeurADP) has employed value decomposition with Approximate Dynamic Programming (ADP) to outperform leading approaches by considering the impact of an individual agent's (vehicle) chosen actions on the future value of that agent. However, in order to ensure scalability and facilitate city-scale ride pooling, NeurADP completely ignores the impact of other agents actions on individual agent/vehicle value. As demonstrated in our experimental results, ignoring the impact of other agents actions on individual value can have a significant impact on the overall performance when there is increased competition among vehicles for demand. Our key contribution is a novel mechanism based on computing conditional expectations through joint conditional probabilities for capturing dependencies on other agents actions without increasing the complexity of training or decision making. We show that our new approach, Conditional Expectation based Value Decomposition (CEVD) outperforms NeurADP by up to 9.76% in terms of overall requests served, which is a significant improvement on a city wide benchmark taxi dataset."
                    },
                    {
                        "title": "Scalable Distributional Robustness in a Class of Non Convex Optimization with Guarantees",
                        "abstract": "Distributionally robust optimization (DRO) has shown lot of promise in providing robustness in learning as well as sample based optimization problems. We endeavor to provide DRO solutions for a class of sum of fractionals, non-convex optimization which is used for decision making in prominent areas such as facility location and security games. In contrast to previous work, we find it more tractable to optimize the equivalent variance regularized form of DRO rather than the minimax form. We transform the variance regularized form to a mixed-integer second order cone program (MISOCP), which, while guaranteeing near global optimality, does not scale enough to solve problems with real world data-sets. We further propose two abstraction approaches based on clustering and stratified sampling to increase scalability, which we then use for real world data-sets. Importantly, we provide near global optimality guarantees for our approach and show experimentally that our solution quality is better than the locally optimal ones achieved by state-of-the-art gradient-based methods. We experimentally compare our different approaches and baselines, and reveal nuanced properties of a DRO solution."
                    },
                    {
                        "title": "Offline Multi-task Transfer RL with Representational Penalization",
                        "abstract": "We study the problem of representation transfer in offline Reinforcement Learning (RL), where a learner has access to episodic data from a number of source tasks collected a priori, and aims to learn a shared representation to be used in finding a good policy for a target task. Unlike in online RL where the agent interacts with the environment while learning a policy, in the offline setting there cannot be such interactions in either the source tasks or the target task; thus multi-task offline RL can suffer from incomplete coverage.   We propose an algorithm to compute pointwise uncertainty measures for the learnt representation, and establish a data-dependent upper bound for the suboptimality of the learnt policy for the target task. Our algorithm leverages the collective exploration done by source tasks to mitigate poor coverage at some points by a few tasks, thus overcoming the limitation of needing uniformly good coverage for a meaningful transfer by existing offline algorithms. We complement our theoretical results with empirical evaluation on a rich-observation MDP which requires many samples for complete coverage. Our findings illustrate the benefits of penalizing and quantifying the uncertainty in the learnt representation."
                    },
                    {
                        "title": "NeurInt : Learning to Interpolate through Neural ODEs",
                        "abstract": "A wide range of applications require learning image generation models whose latent space effectively captures the high-level factors of variation present in the data distribution. The extent to which a model represents such variations through its latent space can be judged by its ability to interpolate between images smoothly. However, most generative models mapping a fixed prior to the generated images lead to interpolation trajectories lacking smoothness and containing images of reduced quality. In this work, we propose a novel generative model that learns a flexible non-parametric prior over interpolation trajectories, conditioned on a pair of source and target images. Instead of relying on deterministic interpolation methods (such as linear or spherical interpolation in latent space), we devise a framework that learns a distribution of trajectories between two given images using Latent Second-Order Neural Ordinary Differential Equations. Through a hybrid combination of reconstruction and adversarial losses, the generator is trained to map the sampled points from these trajectories to sequences of realistic images that smoothly transition from the source to the target image. Through comprehensive qualitative and quantitative experiments, we demonstrate our approach's effectiveness in generating images of improved quality as well as its ability to learn a diverse distribution over smooth interpolation trajectories for any pair of real source and target images."
                    },
                    {
                        "title": "Changepoint Analysis of Topic Proportions in Temporal Text Data",
                        "abstract": "Changepoint analysis deals with unsupervised detection and/or estimation of time-points in time-series data, when the distribution generating the data changes. In this article, we consider \\emph{offline} changepoint detection in the context of large scale textual data. We build a specialised temporal topic model with provisions for changepoints in the distribution of topic proportions. As full likelihood based inference in this model is computationally intractable, we develop a computationally tractable approximate inference procedure. More specifically, we use sample splitting to estimate topic polytopes first and then apply a likelihood ratio statistic together with a modified version of the wild binary segmentation algorithm of Fryzlewicz et al. (2014). Our methodology facilitates automated detection of structural changes in large corpora without the need of manual processing by domain experts. As changepoints under our model correspond to changes in topic structure, the estimated changepoints are often highly interpretable as marking the surge or decline in popularity of a fashionable topic. We apply our procedure on two large datasets: (i) a corpus of English literature from the period 1800-1922 (Underwoodet al., 2015); (ii) abstracts from the High Energy Physics arXiv repository (Clementet al., 2019). We obtain some historically well-known changepoints and discover some new ones."
                    },
                    {
                        "title": "Multiscale Generative Models: Improving Performance of a Generative Model Using Feedback from Other Dependent Generative Models",
                        "abstract": "Realistic fine-grained multi-agent simulation of real-world complex systems is crucial for many downstream tasks such as reinforcement learning. Recent work has used generative models (GANs in particular) for providing high-fidelity simulation of real-world systems. However, such generative models are often monolithic and miss out on modeling the interaction in multi-agent systems. In this work, we take a first step towards building multiple interacting generative models (GANs) that reflects the interaction in real world. We build and analyze a hierarchical set-up where a higher-level GAN is conditioned on the output of multiple lower-level GANs. We present a technique of using feedback from the higher-level GAN to improve performance of lower-level GANs. We mathematically characterize the conditions under which our technique is impactful, including understanding the transfer learning nature of our set-up. We present three distinct experiments on synthetic data, time series data, and image domain, revealing the wide applicability of our technique."
                    },
                    {
                        "title": "Tackling Stackelberg Network Interdiction against a Boundedly Rational Adversary",
                        "abstract": "This work studies Stackelberg network interdiction games -- an important class of games in which a defender first allocates (randomized) defense resources to a set of critical nodes on a graph while an adversary chooses its path to attack these nodes accordingly. We consider a boundedly rational adversary in which the adversary's response model is based on a dynamic form of classic logit-based discrete choice models. We show that the problem of finding an optimal interdiction strategy for the defender in the rational setting is NP-hard. The resulting optimization is in fact non-convex and additionally, involves complex terms that sum over exponentially many paths. We tackle these computational challenges by presenting new efficient approximation algorithms with bounded solution guarantees. First, we address the exponentially-many-path challenge by proposing a polynomial-time dynamic programming-based formulation. We then show that the gradient of the non-convex objective can also be computed in polynomial time, which allows us to use a gradient-based method to solve the problem efficiently. Second, we identify a restricted problem that is convex and hence gradient-based methods find the global optimal solution for this restricted problem. We further identify mild conditions under which this restricted problem provides a bounded approximation for the original problem."
                    }
                ]
            },
            "ea577153-d84d-4365-8371-9f98e36268a5": {
                "pk": "ea577153-d84d-4365-8371-9f98e36268a5",
                "name": "Mihaela Curmei",
                "collaborators": [
                    "Benjamin Recht",
                    "Sarah Dean",
                    "Wenshuo Guo",
                    "Michael I. Jordan",
                    "Walid Krichene",
                    "Li Zhang",
                    "Mukund Sundararajan",
                    "Georgina Hall",
                    "Han Zhang",
                    "Shangen Lu"
                ],
                "domain": [
                    "Recommendation Systems",
                    "Differential Privacy",
                    "Game Theory",
                    "Dynamic Modeling"
                ],
                "publications": [
                    {
                        "title": "Private Matrix Factorization with Public Item Features",
                        "abstract": "We consider the problem of training private recommendation models with access to public item features. Training with Differential Privacy (DP) offers strong privacy guarantees, at the expense of loss in recommendation quality. We show that incorporating public item features during training can help mitigate this loss in quality. We propose a general approach based on collective matrix factorization (CMF), that works by simultaneously factorizing two matrices: the user feedback matrix (representing sensitive data) and an item feature matrix that encodes publicly available (non-sensitive) item information.   The method is conceptually simple, easy to tune, and highly scalable. It can be applied to different types of public item data, including: (1) categorical item features; (2) item-item similarities learned from public sources; and (3) publicly available user feedback. Furthermore, these data modalities can be collectively utilized to fully leverage public data.   Evaluating our method on a standard DP recommendation benchmark, we find that using public item features significantly narrows the quality gap between private models and their non-private counterparts. As privacy constraints become more stringent, models rely more heavily on public side features for recommendation. This results in a smooth transition from collaborative filtering to item-based contextual recommendations."
                    },
                    {
                        "title": "Shape-Constrained Regression using Sum of Squares Polynomials",
                        "abstract": "We present a hierarchy of semidefinite programs (SDPs) for the problem of fitting a shape-constrained (multivariate) polynomial to noisy evaluations of an unknown shape-constrained function. These shape constraints include convexity or monotonicity over a box. We show that polynomial functions that are optimal to any fixed level of our hierarchy form a consistent estimator of the underlying shape-constrained function. As a byproduct of the proof, we establish that sum-of-squares-convex polynomials are dense in the set of polynomials that are convex over an arbitrary box. A similar sum of squares type density result is established for monotone polynomials. In addition, we classify the complexity of convex and monotone polynomial regression as a function of the degree of the polynomial regressor. While our results show NP-hardness of these problems for degree three or larger, we can check numerically that our SDP-based regressors often achieve similar training error at low levels of the hierarchy. Finally, on the computational side, we present an empirical comparison of our SDP-based convex regressors with the convex least squares estimator introduced in [Hildreth, 1954] and [Holloway, 1979] and show that our regressor is valuable in settings where the number of data points is large and the dimension is relatively small. We demonstrate the performance of our regressor for the problem of computing optimal transport maps in a color transfer task and that of estimating the optimal value function of a conic program. A real-time application of the latter problem to inventory management contract negotiation is presented."
                    },
                    {
                        "title": "Delayed and Indirect Impacts of Link Recommendations",
                        "abstract": "The impacts of link recommendations on social networks are challenging to evaluate, and so far they have been studied in limited settings. Observational studies are restricted in the kinds of causal questions they can answer and naive A/B tests often lead to biased evaluations due to unaccounted network interference. Furthermore, evaluations in simulation settings are often limited to static network models that do not take into account the potential feedback loops between link recommendation and organic network evolution. To this end, we study the impacts of recommendations on social networks in dynamic settings. Adopting a simulation-based approach, we consider an explicit dynamic formation model -- an extension of the celebrated Jackson-Rogers model -- and investigate how link recommendations affect network evolution over time. Empirically, we find that link recommendations have surprising delayed and indirect effects on the structural properties of networks. Specifically, we find that link recommendations can exhibit considerably different impacts in the immediate term and in the long term. For instance, we observe that friend-of-friend recommendations can have an immediate effect in decreasing degree inequality, but in the long term, they can make the degree distribution substantially more unequal. Moreover, we show that the effects of recommendations can persist in networks, in part due to their indirect impacts on natural dynamics even after recommendations are turned off. We show that, in counterfactual simulations, removing the indirect effects of link recommendations can make the network trend faster toward what it would have been under natural growth dynamics."
                    },
                    {
                        "title": "Quantifying Availability and Discovery in Recommender Systems via Stochastic Reachability",
                        "abstract": "In this work, we consider how preference models in interactive recommendation systems determine the availability of content and users' opportunities for discovery. We propose an evaluation procedure based on stochastic reachability to quantify the maximum probability of recommending a target piece of content to an user for a set of allowable strategic modifications. This framework allows us to compute an upper bound on the likelihood of recommendation with minimal assumptions about user behavior. Stochastic reachability can be used to detect biases in the availability of content and diagnose limitations in the opportunities for discovery granted to users. We show that this metric can be computed efficiently as a convex program for a variety of practical settings, and further argue that reachability is not inherently at odds with accuracy. We demonstrate evaluations of recommendation algorithms trained on large datasets of explicit and implicit ratings. Our results illustrate how preference models, selection rules, and user interventions impact reachability and how these effects can be distributed unevenly."
                    },
                    {
                        "title": "Towards Psychologically-Grounded Dynamic Preference Models",
                        "abstract": "Designing recommendation systems that serve content aligned with time varying preferences requires proper accounting of the feedback effects of recommendations on human behavior and psychological condition. We argue that modeling the influence of recommendations on people's preferences must be grounded in psychologically plausible models. We contribute a methodology for developing grounded dynamic preference models. We demonstrate this method with models that capture three classic effects from the psychology literature: Mere-Exposure, Operant Conditioning, and Hedonic Adaptation. We conduct simulation-based studies to show that the psychological models manifest distinct behaviors that can inform system design. Our study has two direct implications for dynamic user modeling in recommendation systems. First, the methodology we outline is broadly applicable for psychologically grounding dynamic preference models. It allows us to critique recent contributions based on their limited discussion of psychological foundation and their implausible predictions. Second, we discuss implications of dynamic preference models for recommendation systems evaluation and design. In an example, we show that engagement and diversity metrics may be unable to capture desirable recommendation system performance."
                    },
                    {
                        "title": "Finding Equilibrium in Multi-Agent Games with Payoff Uncertainty",
                        "abstract": "We study the problem of finding equilibrium strategies in multi-agent games with incomplete payoff information, where the payoff matrices are only known to the players up to some bounded uncertainty sets. In such games, an ex-post equilibrium characterizes equilibrium strategies that are robust to the payoff uncertainty. When the game is one-shot, we show that in zero-sum polymatrix games, an ex-post equilibrium can be computed efficiently using linear programming. We further extend the notion of ex-post equilibrium to stochastic games, where the game is played repeatedly in a sequence of stages and the transition dynamics are governed by an Markov decision process (MDP). We provide sufficient condition for the existence of an ex-post Markov perfect equilibrium (MPE). We show that under bounded payoff uncertainty, the value of any two-player zero-sum stochastic game can be computed up to a tight value interval using dynamic programming."
                    },
                    {
                        "title": "Emergent specialization from participation dynamics and multi-learner retraining",
                        "abstract": "Numerous online services are data-driven: the behavior of users affects the system's parameters, and the system's parameters affect the users' experience of the service, which in turn affects the way users may interact with the system. For example, people may choose to use a service only for tasks that already works well, or they may choose to switch to a different service. These adaptations influence the ability of a system to learn about a population of users and tasks in order to improve its performance broadly. In this work, we analyze a class of such dynamics -- where users allocate their participation amongst services to reduce the individual risk they experience, and services update their model parameters to reduce the service's risk on their current user population. We refer to these dynamics as \\emph{risk-reducing}, which cover a broad class of common model updates including gradient descent and multiplicative weights. For this general class of dynamics, we show that asymptotically stable equilibria are always segmented, with sub-populations allocated to a single learner. Under mild assumptions, the utilitarian social optimum is a stable equilibrium. In contrast to previous work, which shows that repeated risk minimization can result in (Hashimoto et al., 2018; Miller et al., 2021), we find that repeated myopic updates with multiple learners lead to better outcomes. We illustrate the phenomena via a simulated example initialized from real data."
                    },
                    {
                        "title": "Do Offline Metrics Predict Online Performance in Recommender Systems?",
                        "abstract": "Recommender systems operate in an inherently dynamical setting. Past recommendations influence future behavior, including which data points are observed and how user preferences change. However, experimenting in production systems with real user dynamics is often infeasible, and existing simulation-based approaches have limited scale. As a result, many state-of-the-art algorithms are designed to solve supervised learning problems, and progress is judged only by offline metrics. In this work we investigate the extent to which offline metrics predict online performance by evaluating eleven recommenders across six controlled simulated environments. We observe that offline metrics are correlated with online performance over a range of environments. However, improvements in offline metrics lead to diminishing returns in online performance. Furthermore, we observe that the ranking of recommenders varies depending on the amount of initial offline data available. We study the impact of adding exploration strategies, and observe that their effectiveness, when compared to greedy recommendation, is highly dependent on the recommendation algorithm. We provide the environments and recommenders described in this paper as Reclab: an extensible ready-to-use simulation framework at https://github.com/berkeley-reclab/RecLab."
                    }
                ]
            },
            "611f1e71-1b32-46a6-a988-c347f8992b4b": {
                "pk": "611f1e71-1b32-46a6-a988-c347f8992b4b",
                "name": "Daniel L. Jiang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "7ce6ca6c-93d0-4bee-8b2e-0a671af6fe9e": {
                "pk": "7ce6ca6c-93d0-4bee-8b2e-0a671af6fe9e",
                "name": "Jamie Morgenstern",
                "collaborators": [
                    "Pranjal Awasthi",
                    "Avrim Blum",
                    "Matth\u00e4us Kleindessner",
                    "Aaron Roth",
                    "Tim Roughgarden",
                    "Sampath Kannan",
                    "Yishay Mansour",
                    "Ryan Rogers",
                    "Siddarth Srinivasan",
                    "Sarah Dean"
                ],
                "domain": [
                    "Auction Theory",
                    "Mechanism Design",
                    "Fairness in Machine Learning",
                    "Statistical Learning Theory"
                ],
                "publications": [
                    {
                        "title": "The Pseudo-Dimension of Near-Optimal Auctions",
                        "abstract": "This paper develops a general approach, rooted in statistical learning theory, to learning an approximately revenue-maximizing auction from data. We introduce $t$-level auctions to interpolate between simple auctions, such as welfare maximization with reserve prices, and optimal auctions, thereby balancing the competing demands of expressivity and simplicity. We prove that such auctions have small representation error, in the sense that for every product distribution $F$ over bidders' valuations, there exists a $t$-level auction with small $t$ and expected revenue close to optimal. We show that the set of $t$-level auctions has modest pseudo-dimension (for polynomial $t$) and therefore leads to small learning error. One consequence of our results is that, in arbitrary single-parameter settings, one can learn a mechanism with expected revenue arbitrarily close to optimal from a polynomial number of samples."
                    },
                    {
                        "title": "Learning Simple Auctions",
                        "abstract": "We present a general framework for proving polynomial sample complexity bounds for the problem of learning from samples the best auction in a class of \"simple\" auctions. Our framework captures all of the most prominent examples of \"simple\" auctions, including anonymous and non-anonymous item and bundle pricings, with either a single or multiple buyers. The technique we propose is to break the analysis of auctions into two natural pieces. First, one shows that the set of allocation rules have large amounts of structure; second, fixing an allocation on a sample, one shows that the set of auctions agreeing with this allocation on that sample have revenue functions with low dimensionality. Our results effectively imply that whenever it's possible to compute a near-optimal simple auction with a known prior, it is also possible to compute such an auction with an unknown prior (given a polynomial number of samples)."
                    },
                    {
                        "title": "Auctions and Peer Prediction for Academic Peer Review",
                        "abstract": "Peer reviewed publications are considered the gold standard in certifying and disseminating ideas that a research community considers valuable. However, we identify two major drawbacks of the current system: (1) the overwhelming demand for reviewers due to a large volume of submissions, and (2) the lack of incentives for reviewers to participate and expend the necessary effort to provide high-quality reviews. In this work, we adopt a mechanism-design approach to propose improvements to the peer review process, tying together the paper submission and review processes and simultaneously incentivizing high-quality submissions and reviews. In the submission stage, authors participate in a VCG auction for review slots by submitting their papers along with a bid that represents their expected value for having their paper reviewed. For the reviewing stage, we propose a novel peer prediction mechanism (H-DIPP) building on recent work in the information elicitation literature, which incentivizes participating reviewers to provide honest and effortful reviews. The revenue raised in the submission stage auction is used to pay reviewers based on the quality of their reviews in the reviewing stage."
                    },
                    {
                        "title": "Preference Dynamics Under Personalized Recommendations",
                        "abstract": "Many projects (both practical and academic) have designed algorithms to match users to content they will enjoy under the assumption that user's preferences and opinions do not change with the content they see. Evidence suggests that individuals' preferences are directly shaped by what content they see -- radicalization, rabbit holes, polarization, and boredom are all example phenomena of preferences affected by content. Polarization in particular can occur even in ecosystems with \"mass media,\" where no personalization takes place, as recently explored in a natural model of preference dynamics by~\\citet{hkazla2019geometric} and~\\citet{gaitonde2021polarization}. If all users' preferences are drawn towards content they already like, or are repelled from content they already dislike, uniform consumption of media leads to a population of heterogeneous preferences converging towards only two poles.   In this work, we explore whether some phenomenon akin to polarization occurs when users receive \\emph{personalized} content recommendations. We use a similar model of preference dynamics, where an individual's preferences move towards content the consume and enjoy, and away from content they consume and dislike. We show that standard user reward maximization is an almost trivial goal in such an environment (a large class of simple algorithms will achieve only constant regret). A more interesting objective, then, is to understand under what conditions a recommendation algorithm can ensure stationarity of user's preferences. We show how to design a content recommendations which can achieve approximate stationarity, under mild conditions on the set of available content, when a user's preferences are known, and how one can learn enough about a user's preferences to implement such a strategy even when user preferences are initially unknown."
                    },
                    {
                        "title": "Approximately Stable, School Optimal, and Student-Truthful Many-to-One Matchings (via Differential Privacy)",
                        "abstract": "We present a mechanism for computing asymptotically stable school optimal matchings, while guaranteeing that it is an asymptotic dominant strategy for every student to report their true preferences to the mechanism. Our main tool in this endeavor is differential privacy: we give an algorithm that coordinates a stable matching using differentially private signals, which lead to our truthfulness guarantee. This is the first setting in which it is known how to achieve nontrivial truthfulness guarantees for students when computing school optimal matchings, assuming worst- case preferences (for schools and students) in large markets."
                    },
                    {
                        "title": "Learning Valuation Distributions from Partial Observation",
                        "abstract": "Auction theory traditionally assumes that bidders' valuation distributions are known to the auctioneer, such as in the celebrated, revenue-optimal Myerson auction. However, this theory does not describe how the auctioneer comes to possess this information. Recently, Cole and Roughgarden [2014] showed that an approximation based on a finite sample of independent draws from each bidder's distribution is sufficient to produce a near-optimal auction. In this work, we consider the problem of learning bidders' valuation distributions from much weaker forms of observations. Specifically, we consider a setting where there is a repeated, sealed-bid auction with $n$ bidders, but all we observe for each round is who won, but not how much they bid or paid. We can also participate (i.e., submit a bid) ourselves, and observe when we win. From this information, our goal is to (approximately) recover the inherently recoverable part of the underlying bid distributions. We also consider extensions where different subsets of bidders participate in each round, and where bidders' valuations have a common-value component added to their independent private values."
                    },
                    {
                        "title": "Learning What's going on: reconstructing preferences and priorities from opaque transactions",
                        "abstract": "We consider a setting where $n$ buyers, with combinatorial preferences over $m$ items, and a seller, running a priority-based allocation mechanism, repeatedly interact. Our goal, from observing limited information about the results of these interactions, is to reconstruct both the preferences of the buyers and the mechanism of the seller. More specifically, we consider an online setting where at each stage, a subset of the buyers arrive and are allocated items, according to some unknown priority that the seller has among the buyers. Our learning algorithm observes only which buyers arrive and the allocation produced (or some function of the allocation, such as just which buyers received positive utility and which did not), and its goal is to predict the outcome for future subsets of buyers. For this task, the learning algorithm needs to reconstruct both the priority among the buyers and the preferences of each buyer. We derive mistake bound algorithms for additive, unit-demand and single minded buyers. We also consider the case where buyers' utilities for a fixed bundle can change between stages due to different (observed) prices. Our algorithms are efficient both in computation time and in the maximum number of mistakes (both polynomial in the number of buyers and items)."
                    },
                    {
                        "title": "Predictive Inequity in Object Detection",
                        "abstract": "In this work, we investigate whether state-of-the-art object detection systems have equitable predictive performance on pedestrians with different skin tones. This work is motivated by many recent examples of ML and vision systems displaying higher error rates for certain demographic groups than others. We annotate an existing large scale dataset which contains pedestrians, BDD100K, with Fitzpatrick skin tones in ranges [1-3] or [4-6]. We then provide an in-depth comparative analysis of performance between these two skin tone groupings, finding that neither time of day nor occlusion explain this behavior, suggesting this disparity is not merely the result of pedestrians in the 4-6 range appearing in more difficult scenes for detection. We investigate to what extent time of day, occlusion, and reweighting the supervised loss during training affect this predictive bias."
                    },
                    {
                        "title": "Privacy-Preserving Public Information for Sequential Games",
                        "abstract": "In settings with incomplete information, players can find it difficult to coordinate to find states with good social welfare. For example, in financial settings, if a collection of financial firms have limited information about each other's strategies, some large number of them may choose the same high-risk investment in hopes of high returns. While this might be acceptable in some cases, the economy can be hurt badly if many firms make investments in the same risky market segment and it fails. One reason why many firms might end up choosing the same segment is that they do not have information about other firms' investments (imperfect information may lead to `bad' game states). Directly reporting all players' investments, however, raises confidentiality concerns for both individuals and institutions.   In this paper, we explore whether information about the game-state can be publicly announced in a manner that maintains the privacy of the actions of the players, and still suffices to deter players from reaching bad game-states. We show that in many games of interest, it is possible for players to avoid these bad states with the help of privacy-preserving, publicly-announced information. We model behavior of players in this imperfect information setting in two ways -- greedy and undominated strategic behaviours, and we prove guarantees on social welfare that certain kinds of privacy-preserving information can help attain. Furthermore, we design a counter with improved privacy guarantees under continual observation."
                    },
                    {
                        "title": "Additive Approximation for Near-Perfect Phylogeny Construction",
                        "abstract": "We study the problem of constructing phylogenetic trees for a given set of species. The problem is formulated as that of finding a minimum Steiner tree on $n$ points over the Boolean hypercube of dimension $d$. It is known that an optimal tree can be found in linear time if the given dataset has a perfect phylogeny, i.e. cost of the optimal phylogeny is exactly $d$. Moreover, if the data has a near-perfect phylogeny, i.e. the cost of the optimal Steiner tree is $d+q$, it is known that an exact solution can be found in running time which is polynomial in the number of species and $d$, yet exponential in $q$. In this work, we give a polynomial-time algorithm (in both $d$ and $q$) that finds a phylogenetic tree of cost $d+O(q^2)$. This provides the best guarantees known - namely, a $(1+o(1))$-approximation - for the case $\\log(d) \\ll q \\ll \\sqrt{d}$, broadening the range of settings for which near-optimal solutions can be efficiently found. We also discuss the motivation and reasoning for studying such additive approximations."
                    },
                    {
                        "title": "Draft Auctions",
                        "abstract": "We introduce draft auctions, which is a sequential auction format where at each iteration players bid for the right to buy items at a fixed price. We show that draft auctions offer an exponential improvement in social welfare at equilibrium over sequential item auctions where predetermined items are auctioned at each time step. Specifically, we show that for any subadditive valuation the social welfare at equilibrium is an $O(\\log^2(m))$-approximation to the optimal social welfare, where $m$ is the number of items. We also provide tighter approximation results for several subclasses. Our welfare guarantees hold for Bayes-Nash equilibria and for no-regret learning outcomes, via the smooth-mechanism framework. Of independent interest, our techniques show that in a combinatorial auction setting, efficiency guarantees of a mechanism via smoothness for a very restricted class of cardinality valuations, extend with a small degradation, to subadditive valuations, the largest complement-free class of valuations. Variants of draft auctions have been used in practice and have been experimentally shown to outperform other auctions. Our results provide a theoretical justification."
                    },
                    {
                        "title": "Fair k-Center Clustering for Data Summarization",
                        "abstract": "In data summarization we want to choose $k$ prototypes in order to summarize a data set. We study a setting where the data set comprises several demographic groups and we are restricted to choose $k_i$ prototypes belonging to group $i$. A common approach to the problem without the fairness constraint is to optimize a centroid-based clustering objective such as $k$-center. A natural extension then is to incorporate the fairness constraint into the clustering problem. Existing algorithms for doing so run in time super-quadratic in the size of the data set, which is in contrast to the standard $k$-center problem being approximable in linear time. In this paper, we resolve this gap by providing a simple approximation algorithm for the $k$-center problem under the fairness constraint with running time linear in the size of the data set and $k$. If the number of demographic groups is small, the approximation guarantee of our algorithm only incurs a constant-factor overhead."
                    },
                    {
                        "title": "Guarantees for Spectral Clustering with Fairness Constraints",
                        "abstract": "Given the widespread popularity of spectral clustering (SC) for partitioning graph data, we study a version of constrained SC in which we try to incorporate the fairness notion proposed by Chierichetti et al. (2017). According to this notion, a clustering is fair if every demographic group is approximately proportionally represented in each cluster. To this end, we develop variants of both normalized and unnormalized constrained SC and show that they help find fairer clusterings on both synthetic and real data. We also provide a rigorous theoretical analysis of our algorithms on a natural variant of the stochastic block model, where $h$ groups have strong inter-group connectivity, but also exhibit a \"natural\" clustering structure which is fair. We prove that our algorithms can recover this fair clustering with high probability."
                    },
                    {
                        "title": "Equalized odds postprocessing under imperfect group information",
                        "abstract": "Most approaches aiming to ensure a model's fairness with respect to a protected attribute (such as gender or race) assume to know the true value of the attribute for every data point. In this paper, we ask to what extent fairness interventions can be effective even when only imperfect information about the protected attribute is available. In particular, we study the prominent equalized odds postprocessing method of Hardt et al. (2016) under a perturbation of the attribute. We identify conditions on the perturbation that guarantee that the bias of a classifier is reduced even by running equalized odds with the perturbed attribute. We also study the error of the resulting classifier. We empirically observe that under our identified conditions most often the error does not suffer from a perturbation of the protected attribute. For a special case, we formally prove this observation to be true."
                    },
                    {
                        "title": "A Notion of Individual Fairness for Clustering",
                        "abstract": "A common distinction in fair machine learning, in particular in fair classification, is between group fairness and individual fairness. In the context of clustering, group fairness has been studied extensively in recent years; however, individual fairness for clustering has hardly been explored. In this paper, we propose a natural notion of individual fairness for clustering. Our notion asks that every data point, on average, is closer to the points in its own cluster than to the points in any other cluster. We study several questions related to our proposed notion of individual fairness. On the negative side, we show that deciding whether a given data set allows for such an individually fair clustering in general is NP-hard. On the positive side, for the special case of a data set lying on the real line, we propose an efficient dynamic programming approach to find an individually fair clustering. For general data sets, we investigate heuristics aimed at minimizing the number of individual fairness violations and compare them to standard clustering approaches on real data sets."
                    },
                    {
                        "title": "Competition Alleviates Present Bias in Task Completion",
                        "abstract": "We build upon recent work [Kleinberg and Oren, 2014, Kleinberg et al., 2016, 2017] that considers present biased agents, who place more weight on costs they must incur now than costs they will incur in the future. They consider a graph theoretic model where agents must complete a task and show that present biased agents can take exponentially more expensive paths than optimal. We propose a theoretical model that adds competition into the mix -- two agents compete to finish a task first. We show that, in a wide range of settings, a small amount of competition can alleviate the harms of present bias. This can help explain why biased agents may not perform so poorly in naturally competitive settings, and can guide task designers on how to protect present biased agents from harm. Our work thus paints a more positive picture than much of the existing literature on present bias."
                    },
                    {
                        "title": "Distributionally Robust Data Join",
                        "abstract": "Suppose we are given two datasets: a labeled dataset and unlabeled dataset which also has additional auxiliary features not present in the first dataset. What is the most principled way to use these datasets together to construct a predictor?   The answer should depend upon whether these datasets are generated by the same or different distributions over their mutual feature sets, and how similar the test distribution will be to either of those distributions. In many applications, the two datasets will likely follow different distributions, but both may be close to the test distribution. We introduce the problem of building a predictor which minimizes the maximum loss over all probability distributions over the original features, auxiliary features, and binary labels, whose Wasserstein distance is $r_1$ away from the empirical distribution over the labeled dataset and $r_2$ away from that of the unlabeled dataset. This can be thought of as a generalization of distributionally robust optimization (DRO), which allows for two data sources, one of which is unlabeled and may contain auxiliary features."
                    },
                    {
                        "title": "Do Prices Coordinate Markets?",
                        "abstract": "Walrasian equilibrium prices can be said to coordinate markets: They support a welfare optimal allocation in which each buyer is buying bundle of goods that is individually most preferred. However, this clean story has two caveats. First, the prices alone are not sufficient to coordinate the market, and buyers may need to select among their most preferred bundles in a coordinated way to find a feasible allocation. Second, we don't in practice expect to encounter exact equilibrium prices tailored to the market, but instead only approximate prices, somehow encoding \"distributional\" information about the market. How well do prices work to coordinate markets when tie-breaking is not coordinated, and they encode only distributional information?   We answer this question. First, we provide a genericity condition such that for buyers with Matroid Based Valuations, overdemand with respect to equilibrium prices is at most 1, independent of the supply of goods, even when tie-breaking is done in an uncoordinated fashion. Second, we provide learning-theoretic results that show that such prices are robust to changing the buyers in the market, so long as all buyers are sampled from the same (unknown) distribution."
                    },
                    {
                        "title": "Private Pareto Optimal Exchange",
                        "abstract": "We consider the problem of implementing an individually rational, asymptotically Pareto optimal allocation in a barter-exchange economy where agents are endowed with goods and have preferences over the goods of others, but may not use money as a medium of exchange. Because one of the most important instantiations of such economies is kidney exchange -- where the \"input\"to the problem consists of sensitive patient medical records -- we ask to what extent such exchanges can be carried out while providing formal privacy guarantees to the participants. We show that individually rational allocations cannot achieve any non-trivial approximation to Pareto optimality if carried out under the constraint of differential privacy -- or even the relaxation of \\emph{joint} differential privacy, under which it is known that asymptotically optimal allocations can be computed in two-sided markets, where there is a distinction between buyers and sellers and we are concerned only with privacy of the buyers~\\citep{Matching}. We therefore consider a further relaxation that we call \\emph{marginal} differential privacy -- which promises, informally, that the privacy of every agent $i$ is protected from every other agent $j \\neq i$ so long as $j$ does not collude or share allocation information with other agents. We show that, under marginal differential privacy, it is possible to compute an individually rational and asymptotically Pareto optimal allocation in such exchange economies."
                    },
                    {
                        "title": "Simple Mechanisms For Agents With Complements",
                        "abstract": "We study the efficiency of simple auctions in the presence of complements. [DMSW15] introduced the single-bid auction, and showed that it has a price of anarchy (PoA) of $O(\\log m)$ for complement-free (i.e., subadditive) valuations. Prior to our work, no non-trivial upper bound on the PoA of single bid auctions was known for valuations exhibiting complements. We introduce a hierarchy over valuations, where levels of the hierarchy correspond to the degree of complementarity, and the PoA of the single bid auction degrades gracefully with the level of the hierarchy. This hierarchy is a refinement of the Maximum over Positive Hypergraphs (MPH) hierarchy [FFIILS15], where the degree of complementarity $d$ is captured by the maximum number of neighbors of a node in the positive hypergraph representation. We show that the price of anarchy of the single bid auction for valuations of level $d$ of the hierarchy is $O(d^2 \\log(m/d))$, where $m$ is the number of items. We also establish an improved upper bound of $O(d \\log m)$ for a subclass where every hyperedge in the positive hypergraph representation is of size at most 2 (but the degree is still $d$). Finally, we show that randomizing between the single bid auction and the grand bundle auction has a price of anarchy of at most $O(\\sqrt{m})$ for general valuations. All of our results are derived via the smoothness framework, thus extend to coarse-correlated equilibria and to Bayes Nash equilibria."
                    }
                ]
            },
            "ca9cbe13-3262-46e1-841b-634ee3566c70": {
                "pk": "ca9cbe13-3262-46e1-841b-634ee3566c70",
                "name": "Sarah Dean",
                "collaborators": [
                    "Benjamin Recht",
                    "Sarah Dean Rasmussen",
                    "Nikolai Matni",
                    "Sarah Rich",
                    "Fengyu Li",
                    "Jinyan Su",
                    "Jamie Morgenstern",
                    "Ruqing Xu",
                    "Stephen Tu",
                    "Eliot Shekhtman"
                ],
                "domain": [
                    "Recommender Systems",
                    "Machine Learning",
                    "User Interaction",
                    "Control Theory"
                ],
                "publications": [
                    {
                        "title": "Recommendations and User Agency: The Reachability of Collaboratively-Filtered Information",
                        "abstract": "Recommender systems often rely on models which are trained to maximize accuracy in predicting user preferences. When the systems are deployed, these models determine the availability of content and information to different users. The gap between these objectives gives rise to a potential for unintended consequences, contributing to phenomena such as filter bubbles and polarization. In this work, we consider directly the information availability problem through the lens of user recourse. Using ideas of reachability, we propose a computationally efficient audit for top-$N$ linear recommender models. Furthermore, we describe the relationship between model complexity and the effort necessary for users to exert control over their recommendations. We use this insight to provide a novel perspective on the user cold-start problem. Finally, we demonstrate these concepts with an empirical investigation of a state-of-the-art model trained on a widely used movie ratings dataset."
                    },
                    {
                        "title": "Certainty Equivalent Perception-Based Control",
                        "abstract": "In order to certify performance and safety, feedback control requires precise characterization of sensor errors. In this paper, we provide guarantees on such feedback systems when sensors are characterized by solving a supervised learning problem. We show a uniform error bound on nonparametric kernel regression under a dynamically-achievable dense sampling scheme. This allows for a finite-time convergence rate on the sub-optimality of using the regressor in closed-loop for waypoint tracking. We demonstrate our results in simulation with simplified unmanned aerial vehicle and autonomous driving examples."
                    },
                    {
                        "title": "L-space intervals for Graph Manifolds and Cables",
                        "abstract": "We present a graph manifold analog of the Jankins-Neumann classification of Seifert fibered spaces over $S^2$ admitting taut foliations, providing a finite recursive formula to compute the L-space Dehn-filling interval for any graph manifold with torus boundary. As an application of a generalization of this result to Floer simple manifolds, we compute the L-space interval for any cable of a Floer simple knot complement in a closed three-manifold in terms of the original L-space interval, recovering a result of Hedden and Hom as a special case."
                    },
                    {
                        "title": "Cross-Dataset Propensity Estimation for Debiasing Recommender Systems",
                        "abstract": "Datasets for training recommender systems are often subject to distribution shift induced by users' and recommenders' selection biases. In this paper, we study the impact of selection bias on datasets with different quantization. We then leverage two differently quantized datasets from different source distributions to mitigate distribution shift by applying the inverse probability scoring method from causal inference. Empirically, our approach gains significant performance improvement over single-dataset methods and alternative ways of combining two datasets."
                    },
                    {
                        "title": "Learning from Streaming Data when Users Choose",
                        "abstract": "In digital markets comprised of many competing services, each user chooses between multiple service providers according to their preferences, and the chosen service makes use of the user data to incrementally improve its model. The service providers' models influence which service the user will choose at the next time step, and the user's choice, in return, influences the model update, leading to a feedback loop. In this paper, we formalize the above dynamics and develop a simple and efficient decentralized algorithm to locally minimize the overall user loss. Theoretically, we show that our algorithm asymptotically converges to stationary points of of the overall loss almost surely. We also experimentally demonstrate the utility of our algorithm with real world data."
                    },
                    {
                        "title": "Preference Dynamics Under Personalized Recommendations",
                        "abstract": "Many projects (both practical and academic) have designed algorithms to match users to content they will enjoy under the assumption that user's preferences and opinions do not change with the content they see. Evidence suggests that individuals' preferences are directly shaped by what content they see -- radicalization, rabbit holes, polarization, and boredom are all example phenomena of preferences affected by content. Polarization in particular can occur even in ecosystems with \"mass media,\" where no personalization takes place, as recently explored in a natural model of preference dynamics by~\\citet{hkazla2019geometric} and~\\citet{gaitonde2021polarization}. If all users' preferences are drawn towards content they already like, or are repelled from content they already dislike, uniform consumption of media leads to a population of heterogeneous preferences converging towards only two poles.   In this work, we explore whether some phenomenon akin to polarization occurs when users receive \\emph{personalized} content recommendations. We use a similar model of preference dynamics, where an individual's preferences move towards content the consume and enjoy, and away from content they consume and dislike. We show that standard user reward maximization is an almost trivial goal in such an environment (a large class of simple algorithms will achieve only constant regret). A more interesting objective, then, is to understand under what conditions a recommendation algorithm can ensure stationarity of user's preferences. We show how to design a content recommendations which can achieve approximate stationarity, under mild conditions on the set of available content, when a user's preferences are known, and how one can learn enough about a user's preferences to implement such a strategy even when user preferences are initially unknown."
                    },
                    {
                        "title": "A number theoretic result for Berge's conjecture",
                        "abstract": "(Original version of PhD thesis, submitted in Spring 2009 to Harvard University. Provides a solution of the $p > k^2$ case, corresponding to Berge families I-VI, of the \"Lens space realization problem\" later solved in entirety by Greene.) In the 1980's, Berge proved that a certain collection of knots in $S^3$ admitted lens space surgeries, a list which Gordon conjectured was exhaustive. More recently, J. Rasmussen used techniques from Heegaard Floer homology to translate the related problem of classifying simple knots in lens spaces admitting L-space homology sphere surgeries into a combinatorial number theory question about the data $(p,q,k)$ associated to a knot of homology class $k \\in H_1(L(p,q))$ in the lens space $L(p,q)$. In the following paper, we solve this number theoretic problem in the case of $p > k^2$."
                    },
                    {
                        "title": "L-space surgeries on satellites by algebraic links",
                        "abstract": "Given an $n$-component link $L$ in any 3-manifold $M$, the space $\\mathcal{L} \\subset (\\mathbb{Q}\\cup \\mkern-1.5mu\\{\\infty\\})^n$ of rational surgery slopes yielding L-spaces is already fully characterized (in joint work by the author) when $n\\!=\\!1$ and $\\mathcal{L}$ is nontrivial. For $n\\mkern-2mu>\\mkern-3mu1$, however, there are no previous results for $\\mathcal{L}$ as a rational subspace, and only limited results for integer surgeries $\\mathcal{L}\\cap\\mathbb{Z}^n$ on $S^3\\mkern-2mu$. Herein, we provide the first nontrivial explicit descriptions of $\\mathcal{L}$ for rational surgeries on multi-component links. Generalizing Hedden's and Hom's L-space result for cables, we compute both $\\mathcal{L}$, and its topology, for all satellites by torus-links in $S^3\\mkern-2mu$. For fractal-boundaried $\\mathcal{L}$ resulting from satellites by algebraic links or iterated torus links, we develop arbitrarily precise approximation tools. We also extend the provisional validity of the L-space conjecture for rational surgeries on a knot $K \\subset S^3$ to rational surgeries on such satellite-links of $K$. These results exploit the author's generalized Jankins-Neumann formula for graph manifolds."
                    },
                    {
                        "title": "Decision-aid or Controller? Steering Human Decision Makers with Algorithms",
                        "abstract": "Algorithms are used to aid human decision makers by making predictions and recommending decisions. Currently, these algorithms are trained to optimize prediction accuracy. What if they were optimized to control final decisions? In this paper, we study a decision-aid algorithm that learns about the human decision maker and provides ''personalized recommendations'' to influence final decisions. We first consider fixed human decision functions which map observable features and the algorithm's recommendations to final decisions. We characterize the conditions under which perfect control over final decisions is attainable. Under fairly general assumptions, the parameters of the human decision function can be identified from past interactions between the algorithm and the human decision maker, even when the algorithm was constrained to make truthful recommendations. We then consider a decision maker who is aware of the algorithm's manipulation and responds strategically. By posing the setting as a variation of the cheap talk game [Crawford and Sobel, 1982], we show that all equilibria are partition equilibria where only coarse information is shared: the algorithm recommends an interval containing the ideal decision. We discuss the potential applications of such algorithms and their social implications."
                    },
                    {
                        "title": "Safely Learning to Control the Constrained Linear Quadratic Regulator",
                        "abstract": "We study the constrained linear quadratic regulator with unknown dynamics, addressing the tension between safety and exploration in data-driven control techniques. We present a framework which allows for system identification through persistent excitation, while maintaining safety by guaranteeing the satisfaction of state and input constraints. This framework involves a novel method for synthesizing robust constraint-satisfying feedback controllers, leveraging newly developed tools from system level synthesis. We connect statistical results with cost sub-optimality bounds to give non-asymptotic guarantees on both estimation and controller performance."
                    },
                    {
                        "title": "Strategic Usage in a Multi-Learner Setting",
                        "abstract": "Real-world systems often involve some pool of users choosing between a set of services. With the increase in popularity of online learning algorithms, these services can now self-optimize, leveraging data collected on users to maximize some reward such as service quality. On the flipside, users may strategically choose which services to use in order to pursue their own reward functions, in the process wielding power over which services can see and use their data. Extensive prior research has been conducted on the effects of strategic users in single-service settings, with strategic behavior manifesting in the manipulation of observable features to achieve a desired classification; however, this can often be costly or unattainable for users and fails to capture the full behavior of multi-service dynamic systems. As such, we analyze a setting in which strategic users choose among several available services in order to pursue positive classifications, while services seek to minimize loss functions on their observations. We focus our analysis on realizable settings, and show that naive retraining can still lead to oscillation even if all users are observed at different times; however, if this retraining uses memory of past observations, convergent behavior can be guaranteed for certain loss function classes. We provide results obtained from synthetic and real-world data to empirically validate our theoretical findings."
                    },
                    {
                        "title": "Robust Guarantees for Perception-Based Control",
                        "abstract": "Motivated by vision-based control of autonomous vehicles, we consider the problem of controlling a known linear dynamical system for which partial state information, such as vehicle position, is extracted from complex and nonlinear data, such as a camera image. Our approach is to use a learned perception map that predicts some linear function of the state and to design a corresponding safe set and robust controller for the closed loop system with this sensing scheme. We show that under suitable smoothness assumptions on both the perception map and the generative model relating state to complex and nonlinear data, parameters of the safe set can be learned via appropriately dense sampling of the state space. We then prove that the resulting perception-control loop has favorable generalization properties. We illustrate the usefulness of our approach on a synthetic example and on the self-driving car simulation platform CARLA."
                    },
                    {
                        "title": "Quantifying Availability and Discovery in Recommender Systems via Stochastic Reachability",
                        "abstract": "In this work, we consider how preference models in interactive recommendation systems determine the availability of content and users' opportunities for discovery. We propose an evaluation procedure based on stochastic reachability to quantify the maximum probability of recommending a target piece of content to an user for a set of allowable strategic modifications. This framework allows us to compute an upper bound on the likelihood of recommendation with minimal assumptions about user behavior. Stochastic reachability can be used to detect biases in the availability of content and diagnose limitations in the opportunities for discovery granted to users. We show that this metric can be computed efficiently as a convex program for a variety of practical settings, and further argue that reachability is not inherently at odds with accuracy. We demonstrate evaluations of recommendation algorithms trained on large datasets of explicit and implicit ratings. Our results illustrate how preference models, selection rules, and user interventions impact reachability and how these effects can be distributed unevenly."
                    },
                    {
                        "title": "Floer Simple Manifolds and L-Space Intervals",
                        "abstract": "An oriented three-manifold with torus boundary admits either no L-space Dehn filling, a unique L-space filling, or an interval of L-space fillings. In the latter case, which we call \"Floer simple,\" we construct an invariant which computes the interval of L-space filling slopes from the Turaev torsion and a given slope from the interval's interior. As applications, we give a new proof of the classification of Seifert fibered L-spaces over $S^2$, and prove a special case of a conjecture of Boyer and Clay about L-spaces formed by gluing three-manifolds along a torus."
                    },
                    {
                        "title": "A Broader View on Bias in Automated Decision-Making: Reflecting on Epistemology and Dynamics",
                        "abstract": "Machine learning (ML) is increasingly deployed in real world contexts, supplying actionable insights and forming the basis of automated decision-making systems. While issues resulting from biases pre-existing in training data have been at the center of the fairness debate, these systems are also affected by technical and emergent biases, which often arise as context-specific artifacts of implementation. This position paper interprets technical bias as an epistemological problem and emergent bias as a dynamical feedback phenomenon. In order to stimulate debate on how to change machine learning practice to effectively address these issues, we explore this broader view on bias, stress the need to reflect on epistemology, and point to value-sensitive design methodologies to revisit the design and implementation process of automated decision-making systems."
                    },
                    {
                        "title": "Ranking with Long-Term Constraints",
                        "abstract": "The feedback that users provide through their choices (e.g., clicks, purchases) is one of the most common types of data readily available for training search and recommendation algorithms. However, myopically training systems based on choice data may only improve short-term engagement, but not the long-term sustainability of the platform and the long-term benefits to its users, content providers, and other stakeholders. In this paper, we thus develop a new framework in which decision makers (e.g., platform operators, regulators, users) can express long-term goals for the behavior of the platform (e.g., fairness, revenue distribution, legal requirements). These goals take the form of exposure or impact targets that go well beyond individual sessions, and we provide new control-based algorithms to achieve these goals. In particular, the controllers are designed to achieve the stated long-term goals with minimum impact on short-term engagement. Beyond the principled theoretical derivation of the controllers, we evaluate the algorithms on both synthetic and real-world data. While all controllers perform well, we find that they provide interesting trade-offs in efficiency, robustness, and the ability to plan ahead."
                    },
                    {
                        "title": "Accounting for AI and Users Shaping One Another: The Role of Mathematical Models",
                        "abstract": "As AI systems enter into a growing number of societal domains, these systems increasingly shape and are shaped by user preferences, opinions, and behaviors. However, the design of AI systems rarely accounts for how AI and users shape one another. In this position paper, we argue for the development of formal interaction models which mathematically specify how AI and users shape one another. Formal interaction models can be leveraged to (1) specify interactions for implementation, (2) monitor interactions through empirical analysis, (3) anticipate societal impacts via counterfactual analysis, and (4) control societal impacts via interventions. The design space of formal interaction models is vast, and model design requires careful consideration of factors such as style, granularity, mathematical complexity, and measurability. Using content recommender systems as a case study, we critically examine the nascent literature of formal interaction models with respect to these use-cases and design axes. More broadly, we call for the community to leverage formal interaction models when designing, evaluating, or auditing any AI system which interacts with users."
                    }
                ]
            },
            "7252b489-cde9-44f1-976e-88c6f888891a": {
                "pk": "7252b489-cde9-44f1-976e-88c6f888891a",
                "name": "Lillian J. Ratliff",
                "collaborators": [
                    "Liyuan Zheng",
                    "Eric Mazumdar",
                    "S. Shankar Sastry",
                    "Arnab Maiti",
                    "Kevin Jamieson",
                    "Tanner Fiez",
                    "Baosen Zhang",
                    "Roy Dong",
                    "Shreyas Sekar",
                    "Mitas Ray"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Game Theory",
                    "Control Systems",
                    "Multi-Agent Systems"
                ],
                "publications": [
                    {
                        "title": "Inverse Risk-Sensitive Reinforcement Learning",
                        "abstract": "We address the problem of inverse reinforcement learning in Markov decision processes where the agent is risk-sensitive. In particular, we model risk-sensitivity in a reinforcement learning framework by making use of models of human decision-making having their origins in behavioral psychology, behavioral economics, and neuroscience. We propose a gradient-based inverse reinforcement learning algorithm that minimizes a loss function defined on the observed behavior. We demonstrate the performance of the proposed technique on two examples, the first of which is the canonical Grid World example and the second of which is a Markov decision process modeling passengers' decisions regarding ride-sharing. In the latter, we use pricing and travel time data from a ride-sharing company to construct the transition probabilities and rewards of the Markov decision process."
                    },
                    {
                        "title": "On the Characterization of Local Nash Equilibria in Continuous Games",
                        "abstract": "We present a unified framework for characterizing local Nash equilibria in continuous games on either infinite-dimensional or finite-dimensional non-convex strategy spaces. We provide intrinsic necessary and sufficient first- and second-order conditions ensuring strategies constitute local Nash equilibria. We term points satisfying the sufficient conditions differential Nash equilibria. Further, we provide a sufficient condition (non-degeneracy) guaranteeing differential Nash equilibria are isolated and show that such equilibria are structurally stable. We present tutorial examples to illustrate our results and highlight degeneracies that can arise in continuous games."
                    },
                    {
                        "title": "Constrained Upper Confidence Reinforcement Learning",
                        "abstract": "Constrained Markov Decision Processes are a class of stochastic decision problems in which the decision maker must select a policy that satisfies auxiliary cost constraints. This paper extends upper confidence reinforcement learning for settings in which the reward function and the constraints, described by cost functions, are unknown a priori but the transition kernel is known. Such a setting is well-motivated by a number of applications including exploration of unknown, potentially unsafe, environments. We present an algorithm C-UCRL and show that it achieves sub-linear regret ($ O(T^{\\frac{3}{4}}\\sqrt{\\log(T/\\delta)})$) with respect to the reward while satisfying the constraints even while learning with probability $1-\\delta$. Illustrative examples are provided."
                    },
                    {
                        "title": "Instance-dependent Sample Complexity Bounds for Zero-sum Matrix Games",
                        "abstract": "We study the sample complexity of identifying an approximate equilibrium for two-player zero-sum $n\\times 2$ matrix games. That is, in a sequence of repeated game plays, how many rounds must the two players play before reaching an approximate equilibrium (e.g., Nash)? We derive instance-dependent bounds that define an ordering over game matrices that captures the intuition that the dynamics of some games converge faster than others. Specifically, we consider a stochastic observation model such that when the two players choose actions $i$ and $j$, respectively, they both observe each other's played actions and a stochastic observation $X_{ij}$ such that $\\mathbb E[ X_{ij}] = A_{ij}$. To our knowledge, our work is the first case of instance-dependent lower bounds on the number of rounds the players must play before reaching an approximate equilibrium in the sense that the number of rounds depends on the specific properties of the game matrix $A$ as well as the desired accuracy. We also prove a converse statement: there exist player strategies that achieve this lower bound."
                    },
                    {
                        "title": "Adaptive Incentive Design",
                        "abstract": "We apply control theoretic and optimization techniques to adaptively design incentives. In particular, we consider the problem of a planner with an objective that depends on data from strategic decision makers. The planner does not know the process by which the strategic agents make decisions. Under the assumption that the agents are utility maximizers, we model their interactions as a non-cooperative game and utilize the Nash equilibrium concept as well as myopic update rules to model the selection of their decision. By parameterizing the agents' utility functions and the incentives offered, we develop an algorithm that the planner can employ to learn the agents' decision-making processes while simultaneously designing incentives to change their response to a more desirable response from the planner's perspective. We provide convergence results for this algorithm both in the noise-free and noisy cases and present illustrative examples."
                    },
                    {
                        "title": "Distribution-Free Guarantees for Systems with Decision-Dependent Noise",
                        "abstract": "In many real-world dynamical systems, obtaining precise models of system uncertainty remains a challenge. It may be difficult to estimate noise distributions or robustness bounds, especially when the distributions/robustness bounds vary with different control inputs in unknown ways. Addressing this challenge, this paper presents a novel iterative method tailored for systems with decision-dependent noise without prior knowledge of the distributions. Our approach finds the open-loop control law that minimizes the worst-case loss, given that the noise induced by this control lies in its $(1 - p)$-confidence set for a predetermined $p$. At each iteration, we use a quantile method inspired by conformal prediction to empirically estimate the confidence set shaped by the preceding control law. These derived confidence sets offer distribution-free guarantees on the system's noise, guiding a robust control formulation that targets worst-case loss minimization. Under specific regularity conditions, our method is shown to converge to a near-optimal open-loop control. While our focus is on open-loop controls, the adaptive, data-driven nature of our approach suggests its potential applicability across diverse scenarios and extensions."
                    },
                    {
                        "title": "To Observe or Not to Observe: Queuing Game Framework for Urban Parking",
                        "abstract": "We model parking in urban centers as a set of parallel queues and overlay a game theoretic structure that allows us to compare the user-selected (Nash) equilibrium to the socially optimal equilibrium. We model arriving drivers as utility maximizers and consider the game in which observing the queue length is free as well as the game in which drivers must pay to observe the queue length. In both games, drivers must decide between balking and joining. We compare the Nash induced welfare to the socially optimal welfare. We find that gains to welfare do not require full information penetration---meaning, for social welfare to increase, not everyone needs to pay to observe. Through simulation, we explore a more complex scenario where drivers decide based the queueing game whether or not to enter a collection of queues over a network. We examine the occupancy-congestion relationship, an important relationship for determining the impact of parking resources on overall traffic congestion. Our simulated models use parameters informed by real-world data collected by the Seattle Department of Transportation."
                    },
                    {
                        "title": "Convergence of Learning Dynamics in Stackelberg Games",
                        "abstract": "This paper investigates the convergence of learning dynamics in Stackelberg games. In the class of games we consider, there is a hierarchical game being played between a leader and a follower with continuous action spaces. We establish a number of connections between the Nash and Stackelberg equilibrium concepts and characterize conditions under which attracting critical points of simultaneous gradient descent are Stackelberg equilibria in zero-sum games. Moreover, we show that the only stable critical points of the Stackelberg gradient dynamics are Stackelberg equilibria in zero-sum games. Using this insight, we develop a gradient-based update for the leader while the follower employs a best response strategy for which each stable critical point is guaranteed to be a Stackelberg equilibrium in zero-sum games. As a result, the learning rule provably converges to a Stackelberg equilibria given an initialization in the region of attraction of a stable critical point. We then consider a follower employing a gradient-play update rule instead of a best response strategy and propose a two-timescale algorithm with similar asymptotic convergence guarantees. For this algorithm, we also provide finite-time high probability bounds for local convergence to a neighborhood of a stable Stackelberg equilibrium in general-sum games. Finally, we present extensive numerical results that validate our theory, provide insights into the optimization landscape of generative adversarial networks, and demonstrate that the learning dynamics we propose can effectively train generative adversarial networks."
                    },
                    {
                        "title": "Incentive Design and Utility Learning via Energy Disaggregation",
                        "abstract": "The utility company has many motivations for modifying energy consumption patterns of consumers such as revenue decoupling and demand response programs. We model the utility company--consumer interaction as a principal--agent problem. We present an iterative algorithm for designing incentives while estimating the consumer's utility function. Incentives are designed using the aggregated as well as the disaggregated (device level) consumption data. We simulate the iterative control (incentive design) and estimation (utility learning and disaggregation) process for examples including the design of incentives based on the aggregate consumption data as well as the disaggregated consumption data."
                    },
                    {
                        "title": "Global Convergence of Policy Gradient for Sequential Zero-Sum Linear Quadratic Dynamic Games",
                        "abstract": "We propose projection-free sequential algorithms for linear-quadratic dynamics games. These policy gradient based algorithms are akin to Stackelberg leadership model and can be extended to model-free settings. We show that if the leader performs natural gradient descent/ascent, then the proposed algorithm has a global sublinear convergence to the Nash equilibrium. Moreover, if the leader adopts a quasi-Newton policy, the algorithm enjoys a global quadratic convergence. Along the way, we examine and clarify the intricacies of adopting sequential policy updates for LQ games, namely, issues pertaining to stabilization, indefinite cost structure, and circumventing projection steps."
                    },
                    {
                        "title": "Logarithmic Regret for Matrix Games against an Adversary with Noisy Bandit Feedback",
                        "abstract": "This paper considers a variant of zero-sum matrix games where at each timestep the row player chooses row $i$, the column player chooses column $j$, and the row player receives a noisy reward with mean $A_{i,j}$. The objective of the row player is to accumulate as much reward as possible, even against an adversarial column player. If the row player uses the EXP3 strategy, an algorithm known for obtaining $\\sqrt{T}$ regret against an arbitrary sequence of rewards, it is immediate that the row player also achieves $\\sqrt{T}$ regret relative to the Nash equilibrium in this game setting. However, partly motivated by the fact that the EXP3 strategy is myopic to the structure of the game, O'Donoghue et al. (2021) proposed a UCB-style algorithm that leverages the game structure and demonstrated that this algorithm greatly outperforms EXP3 empirically. While they showed that this UCB-style algorithm achieved $\\sqrt{T}$ regret, in this paper we ask if there exists an algorithm that provably achieves $\\text{polylog}(T)$ regret against any adversary, analogous to results from stochastic bandits. We propose a novel algorithm that answers this question in the affirmative for the simple $2 \\times 2$ setting, providing the first instance-dependent guarantees for games in the regret setting. Our algorithm overcomes two major hurdles: 1) obtaining logarithmic regret even though the Nash equilibrium is estimable only at a $1/\\sqrt{T}$ rate, and 2) designing row-player strategies that guarantee that either the adversary provides information about the Nash equilibrium, or the row player incurs negative regret. Moreover, in the full information case we address the general $n \\times m$ case where the first hurdle is still relevant. Finally, we show that EXP3 and the UCB-based algorithm necessarily cannot perform better than $\\sqrt{T}$."
                    },
                    {
                        "title": "Near-Optimal Pure Exploration in Matrix Games: A Generalization of Stochastic Bandits & Dueling Bandits",
                        "abstract": "We study the sample complexity of identifying the pure strategy Nash equilibrium (PSNE) in a two-player zero-sum matrix game with noise. Formally, we are given a stochastic model where any learner can sample an entry $(i,j)$ of the input matrix $A\\in[-1,1]^{n\\times m}$ and observe $A_{i,j}+\\eta$ where $\\eta$ is a zero-mean 1-sub-Gaussian noise. The aim of the learner is to identify the PSNE of $A$, whenever it exists, with high probability while taking as few samples as possible. Zhou et al. (2017) presents an instance-dependent sample complexity lower bound that depends only on the entries in the row and column in which the PSNE lies. We design a near-optimal algorithm whose sample complexity matches the lower bound, up to log factors. The problem of identifying the PSNE also generalizes the problem of pure exploration in stochastic multi-armed bandits and dueling bandits, and our result matches the optimal bounds, up to log factors, in both the settings."
                    },
                    {
                        "title": "Uncertainty in Multi-Commodity Routing Networks: When does it help?",
                        "abstract": "We study the equilibrium behavior in a multi-commodity selfish routing game with many types of uncertain users where each user over- or under-estimates their congestion costs by a multiplicative factor. Surprisingly, we find that uncertainties in different directions have qualitatively distinct impacts on equilibria. Namely, contrary to the usual notion that uncertainty increases inefficiencies, network congestion actually decreases when users over-estimate their costs. On the other hand, under-estimation of costs leads to increased congestion. We apply these results to urban transportation networks, where drivers have different estimates about the cost of congestion. In light of the dynamic pricing policies aimed at tackling congestion, our results indicate that users' perception of these prices can significantly impact the policy's efficacy, and \"caution in the face of uncertainty\" leads to favorable network conditions."
                    },
                    {
                        "title": "On Gradient-Based Learning in Continuous Games",
                        "abstract": "We formulate a general framework for competitive gradient-based learning that encompasses a wide breadth of multi-agent learning algorithms, and analyze the limiting behavior of competitive gradient-based learning algorithms using dynamical systems theory. For both general-sum and potential games, we characterize a non-negligible subset of the local Nash equilibria that will be avoided if each agent employs a gradient-based learning algorithm. We also shed light on the issue of convergence to non-Nash strategies in general- and zero-sum games, which may have no relevance to the underlying game, and arise solely due to the choice of algorithm. The existence and frequency of such strategies may explain some of the difficulties encountered when using gradient descent in zero-sum games as, e.g., in the training of generative adversarial networks. To reinforce the theoretical contributions, we provide empirical results that highlight the frequency of linear quadratic dynamic games (a benchmark for multi-agent reinforcement learning) that admit global Nash equilibria that are almost surely avoided by policy gradient."
                    },
                    {
                        "title": "Combinatorial Bandits for Incentivizing Agents with Dynamic Preferences",
                        "abstract": "The design of personalized incentives or recommendations to improve user engagement is gaining prominence as digital platform providers continually emerge. We propose a multi-armed bandit framework for matching incentives to users, whose preferences are unknown a priori and evolving dynamically in time, in a resource constrained environment. We design an algorithm that combines ideas from three distinct domains: (i) a greedy matching paradigm, (ii) the upper confidence bound algorithm (UCB) for bandits, and (iii) mixing times from the theory of Markov chains. For this algorithm, we provide theoretical bounds on the regret and demonstrate its performance via both synthetic and realistic (matching supply and demand in a bike-sharing platform) examples."
                    },
                    {
                        "title": "Koopman Operator Applications in Signalized Traffic Systems",
                        "abstract": "This paper proposes Koopman operator theory and the related algorithm dynamical mode decomposition (DMD) for analysis and control of signalized traffic flow networks. DMD provides a model-free approach for representing complex oscillatory dynamics from measured data, and we study its application to several problems in signalized traffic. We first study a single signalized intersection, and we propose applying this method to infer traffic signal control parameters such as phase timing directly from traffic flow data. Next, we propose using the oscillatory modes of the Koopman operator, approximated with DMD, for early identification of unstable queue growth that has the potential to cause cascading congestion. Then we demonstrate how DMD can be coupled with knowledge of the traffic signal control status to determine traffic signal control parameters that are able to reduce queue lengths. Lastly, we demonstrate that DMD allows for determining the structure and the strength of interactions in a network of signalized intersections. All examples are demonstrated using a case study network instrumented with high resolution traffic flow sensors."
                    },
                    {
                        "title": "Function Design for Improved Competitive Ratio in Online Resource Allocation with Procurement Costs",
                        "abstract": "We study the problem of online resource allocation, where multiple customers arrive sequentially and the seller must irrevocably allocate resources to each incoming customer while also facing a procurement cost for the total allocation. Assuming resource procurement follows an a priori known marginally increasing cost function, the objective is to maximize the reward obtained from fulfilling the customers' requests sans the cumulative procurement cost. We analyze the competitive ratio of a primal-dual algorithm in this setting, and develop an optimization framework for synthesizing a surrogate function for the procurement cost function to be used by the algorithm, in order to improve the competitive ratio of the primal-dual algorithm. Our first design method focuses on polynomial procurement cost functions and uses the optimal surrogate function to provide a more refined bound than the state of the art. Our second design method uses quasiconvex optimization to find optimal design parameters for a general class of procurement cost functions. Numerical examples are used to illustrate the design techniques. We conclude by extending the analysis to devise a posted pricing mechanism in which the algorithm does not require the customers' preferences to be revealed."
                    },
                    {
                        "title": "Decision-Dependent Risk Minimization in Geometrically Decaying Dynamic Environments",
                        "abstract": "This paper studies the problem of expected loss minimization given a data distribution that is dependent on the decision-maker's action and evolves dynamically in time according to a geometric decay process. Novel algorithms for both the information setting in which the decision-maker has a first order gradient oracle and the setting in which they have simply a loss function oracle are introduced. The algorithms operate on the same underlying principle: the decision-maker repeatedly deploys a fixed decision over the length of an epoch, thereby allowing the dynamically changing environment to sufficiently mix before updating the decision. The iteration complexity in each of the settings is shown to match existing rates for first and zero order stochastic gradient methods up to logarithmic factors. The algorithms are evaluated on a \"semi-synthetic\" example using real world data from the SFpark dynamic pricing pilot study; it is shown that the announced prices result in an improvement for the institution's objective (target occupancy), while achieving an overall reduction in parking rates."
                    },
                    {
                        "title": "Safe Reinforcement Learning of Control-Affine Systems with Vertex Networks",
                        "abstract": "This paper focuses on finding reinforcement learning policies for control systems with hard state and action constraints. Despite its success in many domains, reinforcement learning is challenging to apply to problems with hard constraints, especially if both the state variables and actions are constrained. Previous works seeking to ensure constraint satisfaction, or safety, have focused on adding a projection step to a learned policy. Yet, this approach requires solving an optimization problem at every policy execution step, which can lead to significant computational costs.   To tackle this problem, this paper proposes a new approach, termed Vertex Networks (VNs), with guarantees on safety during exploration and on learned control policies by incorporating the safety constraints into the policy network architecture. Leveraging the geometric property that all points within a convex set can be represented as the convex combination of its vertices, the proposed algorithm first learns the convex combination weights and then uses these weights along with the pre-calculated vertices to output an action. The output action is guaranteed to be safe by construction. Numerical examples illustrate that the proposed VN algorithm outperforms vanilla reinforcement learning in a variety of benchmark control tasks."
                    }
                ]
            },
            "ffa86ad7-c94b-491c-9772-7e6e997457b4": {
                "pk": "ffa86ad7-c94b-491c-9772-7e6e997457b4",
                "name": "Maryam Fazel",
                "collaborators": [
                    "Omid Sadeghi",
                    "Reza Eghbali",
                    "Zaid Harchaoui",
                    "Amin Jalali",
                    "Lin Xiao",
                    "Vincent Roulet",
                    "Siddhartha Srinivasa",
                    "Krishnamurthy Dvijotham",
                    "Emanuel Todorov",
                    "Dmitriy Drusvyatskiy"
                ],
                "domain": [
                    "Quantum Computing",
                    "Online Optimization",
                    "Convex Optimization",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Computational approach to quantum encoder design for purity optimization",
                        "abstract": "In this paper, we address the problem of designing a quantum encoder that maximizes the minimum output purity of a given decohering channel, where the minimum is taken over all possible pure inputs. This problem is cast as a max-min optimization problem with a rank constraint on an appropriately defined matrix variable. The problem is computationally very hard because it is non-convex with respect to both the objective function (output purity) and the rank constraint. Despite this difficulty, we provide a tractable computational algorithm that produces the exact optimal solution for codespace of dimension two. Moreover, this algorithm is easily extended to cover the general class of codespaces, in which case the solution is suboptimal in the sense that the suboptimized output purity serves as a lower bound of the exact optimal purity. The algorithm consists of a sequence of semidefinite programmings and can be performed easily. Two typical quantum error channels are investigated to illustrate the effectiveness of our method."
                    },
                    {
                        "title": "Decomposable Norm Minimization with Proximal-Gradient Homotopy Algorithm",
                        "abstract": "We study the convergence rate of the proximal-gradient homotopy algorithm applied to norm-regularized linear least squares problems, for a general class of norms. The homotopy algorithm reduces the regularization parameter in a series of steps, and uses a proximal-gradient algorithm to solve the problem at each step. Proximal-gradient algorithm has a linear rate of convergence given that the objective function is strongly convex, and the gradient of the smooth component of the objective function is Lipschitz continuous. In many applications, the objective function in this type of problem is not strongly convex, especially when the problem is high-dimensional and regularizers are chosen that induce sparsity or low-dimensionality. We show that if the linear sampling matrix satisfies certain assumptions and the regularizing norm is decomposable, proximal-gradient homotopy algorithm converges with a \\emph{linear rate} even though the objective function is not strongly convex. Our result generalizes results on the linear convergence of homotopy algorithm for $l_1$-regularized least squares problems. Numerical experiments are presented that support the theoretical convergence rate analysis."
                    },
                    {
                        "title": "Worst Case Competitive Analysis of Online Algorithms for Conic Optimization",
                        "abstract": "Online optimization covers problems such as online resource allocation, online bipartite matching, adwords (a central problem in e-commerce and advertising), and adwords with separable concave returns. We analyze the worst case competitive ratio of two primal-dual algorithms for a class of online convex (conic) optimization problems that contains the previous examples as special cases defined on the positive orthant. We derive a sufficient condition on the objective function that guarantees a constant worst case competitive ratio (greater than or equal to $\\frac{1}{2}$) for monotone objective functions. We provide new examples of online problems on the positive orthant and the positive semidefinite cone that satisfy the sufficient condition. We show how smoothing can improve the competitive ratio of these algorithms, and in particular for separable functions, we show that the optimal smoothing can be derived by solving a convex optimization problem. This result allows us to directly optimize the competitive ratio bound over a class of smoothing functions, and hence design effective smoothing customized for a given cost function."
                    },
                    {
                        "title": "No-Regret Online Prediction with Strategic Experts",
                        "abstract": "We study a generalization of the online binary prediction with expert advice framework where at each round, the learner is allowed to pick $m\\geq 1$ experts from a pool of $K$ experts and the overall utility is a modular or submodular function of the chosen experts. We focus on the setting in which experts act strategically and aim to maximize their influence on the algorithm's predictions by potentially misreporting their beliefs about the events. Among others, this setting finds applications in forecasting competitions where the learner seeks not only to make predictions by aggregating different forecasters but also to rank them according to their relative performance. Our goal is to design algorithms that satisfy the following two requirements: 1) $\\textit{Incentive-compatible}$: Incentivize the experts to report their beliefs truthfully, and 2) $\\textit{No-regret}$: Achieve sublinear regret with respect to the true beliefs of the best fixed set of $m$ experts in hindsight. Prior works have studied this framework when $m=1$ and provided incentive-compatible no-regret algorithms for the problem. We first show that a simple reduction of our problem to the $m=1$ setting is neither efficient nor effective. Then, we provide algorithms that utilize the specific structure of the utility functions to achieve the two desired goals."
                    },
                    {
                        "title": "Online Continuous DR-Submodular Maximization with Long-Term Budget Constraints",
                        "abstract": "In this paper, we study a class of online optimization problems with long-term budget constraints where the objective functions are not necessarily concave (nor convex) but they instead satisfy the Diminishing Returns (DR) property. Specifically, a sequence of monotone DR-submodular objective functions $\\{f_t(x)\\}_{t=1}^T$ and monotone linear budget functions $\\{\\langle p_t,x \\rangle \\}_{t=1}^T$ arrive over time and assuming a total targeted budget $B_T$, the goal is to choose points $x_t$ at each time $t\\in\\{1,\\dots,T\\}$, without knowing $f_t$ and $p_t$ on that step, to achieve sub-linear regret bound while the total budget violation $\\sum_{t=1}^T \\langle p_t,x_t \\rangle -B_T$ is sub-linear as well. Prior work has shown that achieving sub-linear regret is impossible if the budget functions are chosen adversarially. Therefore, we modify the notion of regret by comparing the agent against a $(1-\\frac{1}{e})$-approximation to the best fixed decision in hindsight which satisfies the budget constraint proportionally over any window of length $W$. We propose the Online Saddle Point Hybrid Gradient (OSPHG) algorithm to solve this class of online problems. For $W=T$, we recover the aforementioned impossibility result. However, when $W=o(T)$, we show that it is possible to obtain sub-linear bounds for both the $(1-\\frac{1}{e})$-regret and the total budget violation."
                    },
                    {
                        "title": "Fast First-Order Methods for Monotone Strongly DR-Submodular Maximization",
                        "abstract": "Continuous DR-submodular functions are a class of functions that satisfy the Diminishing Returns (DR) property, which implies that they are concave along non-negative directions. Existing works have studied monotone continuous DR-submodular maximization subject to a convex constraint and have proposed efficient algorithms with approximation guarantees. However, in many applications, e.g., computing the stability number of a graph and mean-field inference for probabilistic log-submodular models, the DR-submodular function has the additional property of being \\emph{strongly} concave along non-negative directions that could be utilized for obtaining faster convergence rates. In this paper, we first introduce and characterize the class of \\emph{strongly DR-submodular} functions and show how such a property implies strong concavity along non-negative directions. Then, we study $L$-smooth monotone strongly DR-submodular functions that have bounded curvature, and we show how to exploit such additional structure to obtain algorithms with improved approximation guarantees and faster convergence rates for the maximization problem. In particular, we propose the SDRFW algorithm that matches the provably optimal $1-\\frac{c}{e}$ approximation ratio after only $\\lceil\\frac{L}{\\mu}\\rceil$ iterations, where $c\\in[0,1]$ and $\\mu\\geq 0$ are the curvature and the strong DR-submodularity parameter. Furthermore, we study the Projected Gradient Ascent (PGA) method for this problem and provide a refined analysis of the algorithm with an improved $\\frac{1}{1+c}$ approximation ratio and a linear convergence rate. Given that both algorithms require knowledge of the smoothness parameter $L$, we provide a \\emph{novel} characterization of $L$ for DR-submodular functions showing that in many cases, computing $L$ could be formulated as a convex problem, i.e., a geometric program, that could be solved efficiently."
                    },
                    {
                        "title": "Competitive Algorithms for Online Budget-Constrained Continuous DR-Submodular Problems",
                        "abstract": "In this paper, we study a certain class of online optimization problems, where the goal is to maximize a function that is not necessarily concave and satisfies the Diminishing Returns (DR) property under budget constraints. We analyze a primal-dual algorithm, called the Generalized Sequential algorithm, and we obtain the first bound on the competitive ratio of online monotone DR-submodular function maximization subject to linear packing constraints which matches the known tight bound in the special case of linear objective function."
                    },
                    {
                        "title": "Dual Approximation Policy Optimization",
                        "abstract": "We propose Dual Approximation Policy Optimization (DAPO), a framework that incorporates general function approximation into policy mirror descent methods. In contrast to the popular approach of using the $L_2$-norm to measure function approximation errors, DAPO uses the dual Bregman divergence induced by the mirror map for policy projection. This duality framework has both theoretical and practical implications: not only does it achieve fast linear convergence with general function approximation, but it also includes several well-known practical methods as special cases, immediately providing strong convergence guarantees."
                    },
                    {
                        "title": "On the Convergence of the Iterative Linear Exponential Quadratic Gaussian Algorithm to Stationary Points",
                        "abstract": "A classical method for risk-sensitive nonlinear control is the iterative linear exponential quadratic Gaussian algorithm. We present its convergence analysis from a first-order optimization viewpoint. We identify the objective that the algorithm actually minimizes and we show how the addition of a proximal term guarantees convergence to a stationary point."
                    },
                    {
                        "title": "A Simplified Approach to Recovery Conditions for Low Rank Matrices",
                        "abstract": "Recovering sparse vectors and low-rank matrices from noisy linear measurements has been the focus of much recent research. Various reconstruction algorithms have been studied, including $\\ell_1$ and nuclear norm minimization as well as $\\ell_p$ minimization with $p<1$. These algorithms are known to succeed if certain conditions on the measurement map are satisfied. Proofs of robust recovery for matrices have so far been much more involved than in the vector case.   In this paper, we show how several robust classes of recovery conditions can be extended from vectors to matrices in a simple and transparent way, leading to the best known restricted isometry and nullspace conditions for matrix recovery. Our results rely on the ability to \"vectorize\" matrices through the use of a key singular value inequality."
                    },
                    {
                        "title": "Convex Structured Controller Design",
                        "abstract": "We consider the problem of synthesizing optimal linear feedback policies subject to arbitrary convex constraints on the feedback matrix. This is known to be a hard problem in the usual formulations ($\\Htwo,\\Hinf,\\LQR$) and previous works have focused on characterizing classes of structural constraints that allow efficient solution through convex optimization or dynamic programming techniques. In this paper, we propose a new control objective and show that this formulation makes the problem of computing optimal linear feedback matrices convex under arbitrary convex constraints on the feedback matrix. This allows us to solve problems in decentralized control (sparsity in the feedback matrices), control with delays and variable impedance control. Although the control objective is nonstandard, we present theoretical and empirical evidence that it agrees well with standard notions of control. We also present an extension to nonlinear control affine systems. We present numerical experiments validating our approach."
                    },
                    {
                        "title": "Universal Convexification via Risk-Aversion",
                        "abstract": "We develop a framework for convexifying a fairly general class of optimization problems. Under additional assumptions, we analyze the suboptimality of the solution to the convexified problem relative to the original nonconvex problem and prove additive approximation guarantees. We then develop algorithms based on stochastic gradient methods to solve the resulting optimization problems and show bounds on convergence rates. %We show a simple application of this framework to supervised learning, where one can perform integration explicitly and can use standard (non-stochastic) optimization algorithms with better convergence guarantees. We then extend this framework to apply to a general class of discrete-time dynamical systems. In this context, our convexification approach falls under the well-studied paradigm of risk-sensitive Markov Decision Processes. We derive the first known model-based and model-free policy gradient optimization algorithms with guaranteed convergence to the optimal solution. Finally, we present numerical results validating our formulation in different applications."
                    },
                    {
                        "title": "Variational Gram Functions: Convex Analysis and Optimization",
                        "abstract": "We propose a new class of convex penalty functions, called \\emph{variational Gram functions} (VGFs), that can promote pairwise relations, such as orthogonality, among a set of vectors in a vector space. These functions can serve as regularizers in convex optimization problems arising from hierarchical classification, multitask learning, and estimating vectors with disjoint supports, among other applications. We study convexity for VGFs, and give efficient characterizations for their convex conjugates, subdifferentials, and proximal operators. We discuss efficient optimization algorithms for regularized loss minimization problems where the loss admits a common, yet simple, variational representation and the regularizer is a VGF. These algorithms enjoy a simple kernel trick, an efficient line search, as well as computational advantages over first order methods based on the subdifferential or proximal maps. We also establish a general representer theorem for such learning problems. Lastly, numerical experiments on a hierarchical classification problem are presented to demonstrate the effectiveness of VGFs and the associated optimization algorithms."
                    },
                    {
                        "title": "Relative Density and Exact Recovery in Heterogeneous Stochastic Block Models",
                        "abstract": "The Stochastic Block Model (SBM) is a widely used random graph model for networks with communities. Despite the recent burst of interest in recovering communities in the SBM from statistical and computational points of view, there are still gaps in understanding the fundamental information theoretic and computational limits of recovery. In this paper, we consider the SBM in its full generality, where there is no restriction on the number and sizes of communities or how they grow with the number of nodes, as well as on the connection probabilities inside or across communities. This generality allows us to move past the artifacts of homogenous SBM, and understand the right parameters (such as the relative densities of communities) that define the various recovery thresholds. We outline the implications of our generalizations via a set of illustrative examples. For instance, $\\log n$ is considered to be the standard lower bound on the cluster size for exact recovery via convex methods, for homogenous SBM. We show that it is possible, in the right circumstances (when sizes are spread and the smaller the cluster, the denser), to recover very small clusters (up to $\\sqrt{\\log n}$ size), if there are just a few of them (at most polylogarithmic in $n$)."
                    },
                    {
                        "title": "An optimal first order method based on optimal quadratic averaging",
                        "abstract": "In a recent paper, Bubeck, Lee, and Singh introduced a new first order method for minimizing smooth strongly convex functions. Their geometric descent algorithm, largely inspired by the ellipsoid method, enjoys the optimal linear rate of convergence. We show that the same iterate sequence is generated by a scheme that in each iteration computes an optimal average of quadratic lower-models of the function. Indeed, the minimum of the averaged quadratic approaches the true minimum at an optimal rate. This intuitive viewpoint reveals clear connections to the original fast-gradient methods and cutting plane ideas, and leads to limited-memory extensions with improved performance."
                    },
                    {
                        "title": "Fair Packing and Covering on a Relative Scale",
                        "abstract": "Fair resource allocation is a fundamental optimization problem with applications in operations research, networking, and economic and game theory. Research in these areas has led to the general acceptance of a class of $\\alpha$-fair utility functions parameterized by $\\alpha \\in [0, \\infty]$. We consider $\\alpha$-fair packing -- the problem of maximizing $\\alpha$-fair utilities under positive linear constraints -- and provide a simple first-order method for solving it with relative-error guarantees. The method has a significantly lower convergence time than the state of the art, and to analyze it, we leverage the Approximate Duality Gap Technique, which provides an intuitive interpretation of the convergence argument. Finally, we introduce a natural counterpart of $\\alpha$-fairness for minimization problems and motivate its usage in the context of fair task allocation. This generalization yields $\\alpha$-fair covering problems, for which we provide the first near-linear-time solvers with relative-error guarantees."
                    },
                    {
                        "title": "New Computational and Statistical Aspects of Regularized Regression with Application to Rare Feature Selection and Aggregation",
                        "abstract": "Prior knowledge on properties of a target model often come as discrete or combinatorial descriptions. This work provides a unified computational framework for defining norms that promote such structures. More specifically, we develop associated tools for optimization involving such norms given only the orthogonal projection oracle onto the non-convex set of desired models. As an example, we study a norm, which we term the doubly-sparse norm, for promoting vectors with few nonzero entries taking only a few distinct values. We further discuss how the K-means algorithm can serve as the underlying projection oracle in this case and how it can be efficiently represented as a quadratically constrained quadratic program. Our motivation for the study of this norm is regularized regression in the presence of rare features which poses a challenge to various methods within high-dimensional statistics, and in machine learning in general. The proposed estimation procedure is designed to perform automatic feature selection and aggregation for which we develop statistical bounds. The bounds are general and offer a statistical framework for norm-based regularization. The bounds rely on novel geometric quantities on which we attempt to elaborate as well."
                    },
                    {
                        "title": "Escaping from saddle points on Riemannian manifolds",
                        "abstract": "We consider minimizing a nonconvex, smooth function $f$ on a Riemannian manifold $\\mathcal{M}$. We show that a perturbed version of Riemannian gradient descent algorithm converges to a second-order stationary point (and hence is able to escape saddle points on the manifold). The rate of convergence depends as $1/\\epsilon^2$ on the accuracy $\\epsilon$, which matches a rate known only for unconstrained smooth minimization. The convergence rate depends polylogarithmically on the manifold dimension $d$, hence is almost dimension-free. The rate also has a polynomial dependence on the parameters describing the curvature of the manifold and the smoothness of the function. While the unconstrained problem (Euclidean setting) is well-studied, our result is the first to prove such a rate for nonconvex, manifold-constrained problems."
                    },
                    {
                        "title": "Complexity Bounds of Iterative Linear Quadratic Optimization Algorithms for Discrete Time Nonlinear Control",
                        "abstract": "A classical approach for solving discrete time nonlinear control on a finite horizon consists in repeatedly minimizing linear quadratic approximations of the original problem around current candidate solutions. While widely popular in many domains, such an approach has mainly been analyzed locally. We observe that global convergence guarantees can be ensured provided that the linearized discrete time dynamics are surjective and costs on the state variables are strongly convex. We present how the surjectivity of the linearized dynamics can be ensured by appropriate discretization schemes given the existence of a feedback linearization scheme. We present complexity bounds of algorithms based on linear quadratic approximations through the lens of generalized Gauss-Newton methods. Our analysis uncovers several convergence phases for regularized generalized Gauss-Newton algorithms."
                    },
                    {
                        "title": "Flat minima generalize for low-rank matrix recovery",
                        "abstract": "Empirical evidence suggests that for a variety of overparameterized nonlinear models, most notably in neural network training, the growth of the loss around a minimizer strongly impacts its performance. Flat minima -- those around which the loss grows slowly -- appear to generalize well. This work takes a step towards understanding this phenomenon by focusing on the simplest class of overparameterized nonlinear models: those arising in low-rank matrix recovery. We analyze overparameterized matrix and bilinear sensing, robust PCA, covariance matrix estimation, and single hidden layer neural networks with quadratic activation functions. In all cases, we show that flat minima, measured by the trace of the Hessian, exactly recover the ground truth under standard statistical assumptions. For matrix completion, we establish weak recovery, although empirical evidence suggests exact recovery holds here as well. We conclude with synthetic experiments that illustrate our findings and discuss the effect of depth on flat solutions."
                    }
                ]
            }
        }
    },
    "2410.11187": {
        "paper_data": {
            "title": "Multiview Scene Graph",
            "url": "http://arxiv.org/abs/2410.11187v1",
            "arxiv_id": "2410.11187",
            "authors": [
                "Juexiao Zhang",
                "Gao Zhu",
                "Sihang Li",
                "Xinhao Liu",
                "Haorui Song",
                "Xinran Tang",
                "Chen Feng"
            ],
            "abstract": "A proper scene representation is central to the pursuit of spatial intelligence where agents can robustly reconstruct and efficiently understand 3D scenes. A scene representation is either metric, such as landmark maps in 3D reconstruction, 3D bounding boxes in object detection, or voxel grids in occupancy prediction, or topological, such as pose graphs with loop closures in SLAM or visibility graphs in SfM. In this work, we propose to build Multiview Scene Graphs (MSG) from unposed images, representing a scene topologically with interconnected place and object nodes. The task of building MSG is challenging for existing representation learning methods since it needs to jointly address both visual place recognition, object detection, and object association from images with limited fields of view and potentially large viewpoint changes. To evaluate any method tackling this task, we developed an MSG dataset and annotation based on a public 3D dataset. We also propose an evaluation metric based on the intersection-over-union score of MSG edges. Moreover, we develop a novel baseline method built on mainstream pretrained vision models, combining visual place recognition and object association into one Transformer decoder architecture. Experiments demonstrate our method has superior performance compared to existing relevant baselines.",
            "introduction": "   1 Introduction  The ability to understand 3D space and the spatial relationships among 2D observations plays a central role in mobile agents interacting with the physical real world. Humans obtain such spatial intelligence largely from our visual intelligence\u00a0[45, 26]. When humans are situated in an unseen environment and try to understand the spatial structure from visual observations, we don\u2019t perceive and memorize the scene by exact meters and degrees. Instead, we build cognitive maps topologically based on visual observations and commonsense\u00a0[48, 27]. Given imagery observations, we are able to associate the images taken at the same place by finding overlapping visual clues and identifying the same or different objects from various viewpoints. This ability to establish correspondence from visual perception constitutes the foundation of our spatial memory and cognitive representation of the world. Can we equip AI models with similar spatial intelligence?   Motivated by this question, we propose the task of building a Multiview Scene Graph (MSG) to explicitly evaluate a representation learning model\u2019s capability of understanding spatial correspondences. Specifically, as illustrated in Figure\u00a01, given a set of unposed RGB images taken from the same scene, this task requires building a place+object graph consisting of images and object nodes, where images taken at nearby locations are connected, and objects appearances across the different views should be associated together as one object node.   We position the proposed Multiview Scene Graph as a general topological scene representation. It bridges the place recognition from robotics literature\u00a0[36, 4, 3] and the object tracking and semantic correspondence tasks from computer vision literature\u00a0[62, 20, 23]. Different from previous work in topological mapping that evaluates a method\u2019s performance on downstream tasks such as navigation, we propose to directly evaluate the quality of the multiview scene graph, which explicitly demonstrates a model\u2019s spatial understanding with correct visual correspondence of both objects and places across multiple views. Moreover, the MSG does not require any metric map or depth and pose information, making it adaptable to the vast data of everyday images and videos. This also differentiates MSG from the previous work in 2D and 3D scene graphs\u00a0[67, 5, 32, 35], where they emphasize objects\u2019 semantic relationships or require different levels of 3D and metric information.   To facilitate the research of MSG, we curated a dataset from a publicly available 3D scene-level dataset ARKitScenes\u00a0[8] and designed a set of evaluation metrics based on the intersection-over-union of the graph adjacency matrix. The detailed definition of the MSG generation task and the evaluation metrics are discussed in Section\u00a03.1. Meanwhile, since this task mainly involves solving place recognition and object association, we benchmarked popular baseline methods respectively in place recognition and object tracking, as well as some mainstream pretrained vision foundation models. We also designed a new Transformer-based architecture as our method, Attention Association MSG, dubbed AoMSG, which learns place and object embeddings jointly in a single Transformer decoder and builds the MSG based on the distances in the learned embedding space. Our experiments demonstrate the superiority of our new model compared with the baselines by a great margin, yet still reveal strong needs for future advances in research for spatial intelligence.   In summary, our contributions are two-fold:   \u2022  We propose the Multiview Scene Graph (MSG) generation as",
            "references": [
                {
                    "title": "Tell Me Where You Are: Multimodal LLMs Meet Place Recognition",
                    "abstract": "Large language models (LLMs) exhibit a variety of promising capabilities in robotics, including long-horizon planning and commonsense reasoning. However, their performance in place recognition is still underexplored. In this work, we introduce multimodal LLMs (MLLMs) to visual place recognition (VPR), where a robot must localize itself using visual observations. Our key design is to use vision-based retrieval to propose several candidates and then leverage language-based reasoning to carefully inspect each candidate for a final decision. Specifically, we leverage the robust visual features produced by off-the-shelf vision foundation models (VFMs) to obtain several candidate locations. We then prompt an MLLM to describe the differences between the current observation and each candidate in a pairwise manner, and reason about the best candidate based on these descriptions. Our results on three datasets demonstrate that integrating the general-purpose visual features from VFMs with the reasoning capabilities of MLLMs already provides an effective place recognition solution, without any VPR-specific supervised training. We believe our work can inspire new possibilities for applying and designing foundation models, i.e., VFMs, LLMs, and MLLMs, to enhance the localization and navigation of mobile robots."
                },
                {
                    "title": "EgoSG: Learning 3D Scene Graphs from Egocentric RGB-D Sequences",
                    "abstract": "Constructing a 3D scene graph of an environment is essential for agents and smart glasses assistants to develop an understanding of their surroundings and predict relationships between various entities within it. 3D Scene Graph Prediction (3DSGP) is commonly adopted to predict the spatial and semantic relationships between objects in a 3D environment reconstructed from posed (calibrated) RGB-D sequences, such as object containment or adjacency. However, reconstructing a scene can be time-consuming and computationally intensive, and requires specialized hardware like IMUs for accurate poses. The reliance on (1) robust algorithms and (2) accurate camera poses limits its applicability. Unlike existing 3DSGP methods, we propose to perform perception and reasoning on each frame without assuming available camera poses, which we call EgoSG, to estimate 3D scene graphs directly from egocentric frame sequences. In our method, per-frame instance features are acquired from a partial (per-frame) point cloud. By globally optimizing per-frame features, object instances are then associated across the egocentric frames, and graph representations are aggregated for 3D scene graph prediction. Compared to the state-of-the-art methods that heavily rely on 3D reconstruction, our approach is reconstruction-free and can be derived directly from unposed RGB-D sequences. We benchmark our EgoSG framework against existing reconstruction-based approaches on 3DSGP tasks. Our method outperforms the state-of-the-art methods by a large margin, achieving +44.63 R@1 in Object and +22.74 R@1 in Predicate from egocentric sequences without any reliance on reconstruction algorithms or camera poses."
                },
                {
                    "title": "RoboHop: Segment-based Topological Map Representation for Open-World Visual Navigation",
                    "abstract": "Mapping is crucial for spatial reasoning, planning and robot navigation. Existing approaches range from metric, which require precise geometry-based optimization, to purely topological, where image-as-node based graphs lack explicit object-level reasoning and interconnectivity. In this paper, we propose a novel topological representation of an environment based on , which are semantically meaningful and open-vocabulary queryable, conferring several advantages over previous works based on pixel-level features. Unlike 3D scene graphs, we create a purely topological graph with segments as nodes, where edges are formed by a) associating segment-level descriptors between pairs of consecutive images and b) connecting neighboring segments within an image using their pixel centroids. This unveils a continuous sense of a place, defined by inter-image persistence of segments along with their intra-image neighbours. It further enables us to represent and update segment-level descriptors through neighborhood aggregation using graph convolution layers, which improves robot localization based on segment-level retrieval. Using real-world data, we show how our proposed map representation can be used to i) generate navigation plans in the form of hops over segments and ii) search for target objects using natural language queries describing spatial relations of objects. Furthermore, we quantitatively analyze data association at the segment level, which underpins inter-image connectivity during mapping and segment-level localization when revisiting the same place. Finally, we show preliminary trials on segment-level \u2018hopping\u2019 based zero-shot real-world navigation. Project page with supplementary details: oravus.github.io/RoboHop/."
                },
                {
                    "title": "Probing the 3D Awareness of Visual Foundation Models",
                    "abstract": "Recent advances in large-scale pretraining have yielded visual foundation models with strong capabilities. Not only can recent models generalize to arbitrary images for their training task, their intermediate representations are useful for other visual tasks such as detection and segmentation. Given that such models can classify, delineate, and local- ize objects in 2D, we ask whether they also represent their 3D structure? In this work, we analyze the 3D awareness of visual foundation models. We posit that 3D awareness implies that representations (1) encode the 3D structure of the scene and (2) consistently represent the surface across views. We conduct a series of experiments using task-specific probes and zero-shot inference procedures on frozen fea- tures. Our experiments reveal several limitations of the current models. Our code and analysis can be found at https://github.com/mbanani/probe3d."
                },
                {
                    "title": "Matching 2D Images in 3D: Metric Relative Pose from Metric Correspondences",
                    "abstract": "Given two images, we can estimate the relative camera pose between them by establishing image-to-image correspondences. Usually, correspondences are 2D-to-2D and the pose we estimate is defined only up to scale. Some applications, aiming at instant augmented reality anywhere, require scale-metric pose estimates, and hence, they rely on external depth estimators to recover the scale. We present MicKey, a keypoint matching pipeline that is able to predict metric correspondences in 3D camera space. By learning to match 3D coordinates across images, we are able to infer the metric relative pose without depth measurements. Depth measurements are also not required for training, nor are scene reconstructions or image overlap information. MicKey is supervised only by pairs of images and their relative poses. MicKey achieves state-of-the-art performance on the Map-Free Relocalisation benchmark while requiring less supervision than competing approaches."
                },
                {
                    "title": "DUSt3R: Geometric 3D Vision Made Easy",
                    "abstract": "Multi-view stereo reconstruction (MVS) in the wild re-quires to first estimate the camera intrinsic and extrinsic parameters. These are usually tedious and cumbersome to obtain, yet they are mandatory to triangulate corresponding pixels in 3D space, which is at the core of all best performing MVS algorithms. In this work, we take an opposite stance and introduce DUSt3R, a radically novel paradigm for Dense and Unconstrained Stereo 3D Reconstruction of arbitrary image collections, operating without prior infor-mation about camera calibration nor viewpoint poses. We cast the pairwise reconstruction problem as a regression of pointmaps, relaxing the hard constraints of usual projective camera models. We show that this formulation smoothly unifies the monocular and binocular reconstruction cases. In the case where more than two images are provided, we fur-ther propose a simple yet effective global alignment strategy that expresses all pairwise pointmaps in a common refer-ence frame. We base our network architecture on standard Transformer encoders and decoders, allowing us to leverage powerful pretrained models. Our formulation directly provides a 3D model of the scene as well as depth information, but interestingly, we can seamlessly recover from it, pixel matches, focal lengths, relative and absolute cameras. Extensive experiments on all these tasks showcase how DUSt3R effectively unifies various 3D vision tasks, setting new performance records on monocular & multi-view depth estimation as well as relative pose estimation. In summary, DUSt3R makes many geometric 3D vision tasks easy. Code and mod-els at https://github.com/naver/dust3r."
                },
                {
                    "title": "Optimal Transport Aggregation for Visual Place Recognition",
                    "abstract": "The task of Visual Place Recognition (VPR) aims to match a query image against references from an extensive database of images from different places, relying solely on visual cues. State-of-the-art pipelines focus on the aggre-gation offeatures extractedfrom a deep backbone, in order to form a global descriptor for each image. In this con-text, we introduce SALAD (Sinkhorn Algorithm for Locally Aggregated Descriptors), which reformulates NetVLAD's soft-assignment of local features to clusters as an optimal transport problem. In SALAD, we consider both feature-to-cluster and cluster-to-feature relations and we also in-troduce a \u2018dustbin\u2019 cluster, designed to selectively discard features deemed non-informative, enhancing the overall de-scriptor quality. Additionally, we leverage and fine-tune DINOv2 as a backbone, which provides enhanced description power for the local features, and dramatically reduces the required training time. As a result, our single-stage method not only surpasses single-stage baselines in pub-lic VPR datasets, but also surpasses two-stage methods that add a re-ranking with significantly higher cost. Code and models are available at https://github.com/serizbalsalad."
                },
                {
                    "title": "Tracking Anything with Decoupled Video Segmentation",
                    "abstract": "Training data for video segmentation are expensive to annotate. This impedes extensions of end-to-end algorithms to new video segmentation tasks, especially in large-vocabulary settings. To \u2018track anything\u2019 without training on video data for every individual task, we develop a decoupled video segmentation approach (DEVA), composed of task-specific image-level segmentation and class/task-agnostic bi-directional temporal propagation. Due to this design, we only need an image-level model for the target task (which is cheaper to train) and a universal temporal propagation model which is trained once and generalizes across tasks. To effectively combine these two modules, we use bi-directional propagation for (semi-)online fusion of segmentation hypotheses from different frames to generate a coherent segmentation. We show that this decoupled formulation compares favorably to end-to-end approaches in several data-scarce tasks including large-vocabulary video panoptic segmentation, open-world video segmentation, referring video segmentation, and unsupervised video object segmentation. Code is available at: hkchengrex.github.io/Tracking-Anything-with-DEVA."
                },
                {
                    "title": "AnyLoc: Towards Universal Visual Place Recognition",
                    "abstract": "Visual Place Recognition (VPR) is vital for robot localization. To date, the most performant VPR approaches are <italic>environment- and task-specific:</italic> while they exhibit strong performance in structured environments (predominantly urban driving), their performance degrades severely in unstructured environments, rendering most approaches brittle to robust real-world deployment. In this work, we develop a <italic>universal</italic> solution to VPR \u2013 a technique that works across a broad range of structured and unstructured environments (urban, outdoors, indoors, aerial, underwater, and subterranean environments) without any re-training or finetuning. We demonstrate that general-purpose feature representations derived from off-the-shelf self-supervised models <italic>with no VPR-specific training</italic> are the right substrate upon which to build such a universal VPR solution. Combining these derived features with <italic>unsupervised feature aggregation</italic> enables our suite of methods, <italic>AnyLoc</italic>, to achieve up to <inline-formula><tex-math notation=\"LaTeX\">$4\\times$</tex-math></inline-formula> significantly higher performance than existing approaches. We further obtain a 6% improvement in performance by characterizing the semantic properties of these features, uncovering unique <italic>domains</italic> which encapsulate datasets from similar environments. Our detailed experiments and analysis lay a foundation for building VPR solutions that may be deployed <italic>anywhere</italic>, <italic>anytime</italic>, and across <italic>anyview</italic>."
                },
                {
                    "title": "An Object SLAM Framework for Association, Mapping, and High-Level Tasks",
                    "abstract": "Object SLAM is considered increasingly significant for robot high-level perception and decision-making. Existing studies fall short in terms of data association, object representation, and semantic mapping and frequently rely on additional assumptions, limiting their performance. In this article, we present a comprehensive object SLAM framework that focuses on object-based perception and object-oriented robot tasks. First, we propose an ensemble data association approach for associating objects in complicated conditions by incorporating parametric and nonparametric statistic testing. In addition, we suggest an outlier-robust centroid and scale estimation algorithm for modeling objects based on the iForest and line alignment. Then a lightweight and object-oriented map is represented by estimated general object models. Taking into consideration the semantic invariance of objects, we convert the object map to a topological map to provide semantic descriptors to enable multimap matching. Finally, we suggest an object-driven active exploration strategy to achieve autonomous mapping in the grasping scenario. A range of public datasets and real-world results in mapping, augmented reality, scene matching, relocalization, and robotic manipulation have been used to evaluate the proposed object SLAM framework for its efficient performance."
                },
                {
                    "title": "Incremental 3D Semantic Scene Graph Prediction from RGB Sequences",
                    "abstract": "3D semantic scene graphs are a powerful holistic representation as they describe the individual objects and depict the relation between them. They are compact high-level graphs that enable many tasks requiring scene reasoning. In real-world settings, existing 3D estimation methods produce robust predictions that mostly rely on dense inputs. In this work, we propose a real-time framework that incrementally builds a consistent 3D semantic scene graph of a scene given an RGB image sequence. Our method consists of a novel incremental entity estimation pipeline and a scene graph prediction network. The proposed pipeline simultaneously reconstructs a sparse point map and fuses entity estimation from the input images. The proposed network estimates 3D semantic scene graphs with iterative message passing using multi-view and geometric features extracted from the scene entities. Extensive experiments on the 3RScan dataset show the effectiveness of the proposed method in this challenging task, outperforming state-of-the-art approaches. Our implementation is available at https://shunchengwu.github.io/MonoSSG."
                },
                {
                    "title": "Learning-based Relational Object Matching Across Views",
                    "abstract": "Intelligent robots require object-level scene understanding to reason about possible tasks and interactions with the environment. Moreover, many perception tasks such as scene reconstruction, image retrieval, or place recognition can benefit from reasoning on the level of objects. While keypoint-based matching can yield strong results for finding correspondences for images with small to medium view point changes, for large view point changes, matching semantically on the object-level becomes advantageous. In this paper, we propose a learning-based approach which combines local keypoints with novel object-level features for matching object detections between RGB images. We train our object-level matching features based on appearance and inter-frame and cross-frame spatial relations between objects in an associative graph neural network. We demonstrate our approach in a large variety of views on realistically rendered synthetic images. Our approach compares favorably to previous state-of-the-art object-level matching approaches and achieves improved performance over a pure keypoint-based approach for large view-point changes."
                },
                {
                    "title": "DINOv2: Learning Robust Visual Features without Supervision",
                    "abstract": "The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels."
                },
                {
                    "title": "Segment Anything",
                    "abstract": "We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive \u2013 often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images at segment-anything.com to foster research into foundation models for computer vision. We recommend reading the full paper at: arxiv.org/abs/2304.02643."
                },
                {
                    "title": "MotionTrack: Learning Robust Short-Term and Long-Term Motions for Multi-Object Tracking",
                    "abstract": "The main challenge of Multi-Object Tracking (MOT) lies in maintaining a continuous trajectory for each target. Existing methods often learn reliable motion patterns to match the same target between adjacent frames and discriminative appearance features to re-identify the lost targets after a long period. However, the reliability of motion prediction and the discriminability of appearances can be easily hurt by dense crowds and extreme occlusions in the tracking process. In this paper, we propose a simple yet effective multi-object tracker, i.e., MotionTrack, which learns robust short-term and long-term motions in a unified framework to associate trajectories from a short to long range. For dense crowds, we design a novel Interaction Module to learn interaction-aware motions from short-term trajectories, which can estimate the complex movement of each target. For extreme occlusions, we build a novel Refind Module to learn reliable long-term motions from the target's history trajectory, which can link the interrupted trajectory with its corresponding detection. Our Interaction Module and Refind Module are embedded in the well-known tracking-by-detection paradigm, which can work in tandem to maintain superior performance. Extensive experimental results on MOT17 and MOT20 datasets demonstrate the superiority of our approach in challenging scenarios, and it achieves state-of-the-art performances at various MOT metrics. Code is available at https://github.com/qwomeng/MotionTrack."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection",
                    "abstract": "In this paper, we present an open-set object detector, called Grounding DINO, by marrying Transformer-based detector DINO with grounded pre-training, which can detect arbitrary objects with human inputs such as category names or referring expressions. The key solution of open-set object detection is introducing language to a closed-set detector for open-set concept generalization. To effectively fuse language and vision modalities, we conceptually divide a closed-set detector into three phases and propose a tight fusion solution, which includes a feature enhancer, a language-guided query selection, and a cross-modality decoder for cross-modality fusion. While previous works mainly evaluate open-set object detection on novel categories, we propose to also perform evaluations on referring expression comprehension for objects specified with attributes. Grounding DINO performs remarkably well on all three settings, including benchmarks on COCO, LVIS, ODinW, and RefCOCO/+/g. Grounding DINO achieves a $52.5$ AP on the COCO detection zero-shot transfer benchmark, i.e., without any training data from COCO. It sets a new record on the ODinW zero-shot benchmark with a mean $26.1$ AP. Code will be available at \\url{https://github.com/IDEA-Research/GroundingDINO}."
                },
                {
                    "title": "Referring Multi-Object Tracking",
                    "abstract": "Existing referring understanding tasks tend to involve the detection of a single text-referred object. In this paper, we propose a new and general referring understanding task, termed referring multi-object tracking (RMOT). Its core idea is to employ a language expression as a semantic cue to guide the prediction of multi-object tracking. To the best of our knowledge, it is the first work to achieve an arbitrary number of referent object predictions in videos. To push forward RMOT, we construct one benchmark with scalable expressions based on KITTI, named Refer-KITTI. Specifically, it provides 18 videos with 818 expressions, and each expression in a video is annotated with an average of 10.7 objects. Further, we develop a transformer-based architecture TransRMOT to tackle the new task in an online manner, which achieves impressive detection performance and out-performs other counterparts. The Refer-KITTI dataset and the code are released at https://referringmot.github.io."
                },
                {
                    "title": "Towards Democratizing Joint-Embedding Self-Supervised Learning",
                    "abstract": "Joint Embedding Self-Supervised Learning (JE-SSL) has seen rapid developments in recent years, due to its promise to effectively leverage large unlabeled data. The development of JE-SSL methods was driven primarily by the search for ever increasing downstream classification accuracies, using huge computational resources, and typically built upon insights and intuitions inherited from a close parent JE-SSL method. This has led unwittingly to numerous pre-conceived ideas that carried over across methods e.g. that SimCLR requires very large mini batches to yield competitive accuracies; that strong and computationally slow data augmentations are required. In this work, we debunk several such ill-formed a priori ideas in the hope to unleash the full potential of JE-SSL free of unnecessary limitations. In fact, when carefully evaluating performances across different downstream tasks and properly optimizing hyper-parameters of the methods, we most often -- if not always -- see that these widespread misconceptions do not hold. For example we show that it is possible to train SimCLR to learn useful representations, while using a single image patch as negative example, and simple Gaussian noise as the only data augmentation for the positive pair. Along these lines, in the hope to democratize JE-SSL and to allow researchers to easily make more extensive evaluations of their methods, we introduce an optimized PyTorch library for SSL."
                },
                {
                    "title": "MixVPR: Feature Mixing for Visual Place Recognition",
                    "abstract": "Visual Place Recognition (VPR) is a crucial part of mobile robotics and autonomous driving as well as other computer vision tasks. It refers to the process of identifying a place depicted in a query image using only computer vision. At large scale, repetitive structures, weather and illumination changes pose a real challenge, as appearances can drastically change over time. Along with tackling these challenges, an efficient VPR technique must also be practical in real-world scenarios where latency matters. To address this, we introduce MixVPR, a new holistic feature aggregation technique that takes feature maps from pre-trained backbones as a set of global features. Then, it incorporates a global relationship between elements in each feature map in a cascade of feature mixing, eliminating the need for local or pyramidal aggregation as done in NetVLAD or TransVPR. We demonstrate the effectiveness of our technique through extensive experiments on multiple large-scale benchmarks. Our method outperforms all existing techniques by a large margin while having less than half the number of parameters compared to CosPlace and NetVLAD. We achieve a new all-time high recall@1 score of 94.6% on Pitts250k-test, 88.0% on MapillarySLS, and more importantly, 58.4% on Nordland. Finally, our method outperforms two-stage retrieval techniques such as Patch-NetVLAD, TransVPR and SuperGLUE all while being orders of magnitude faster."
                },
                {
                    "title": "Look Before You Match: Instance Understanding Matters in Video Object Segmentation",
                    "abstract": "Exploring dense matching between the current frame and past frames for long-range context modeling, memory-based methods have demonstrated impressive results in video object segmentation (VOS) recently. Nevertheless, due to the lack of instance understanding ability, the above approaches are oftentimes brittle to large appearance variations or viewpoint changes resulted from the movement of objects and cameras. In this paper, we argue that instance understanding matters in VOS, and integrating it with memory-based matching can enjoy the synergy, which is intuitively sensible from the definition of VOS task, i.e., identifying and segmenting object instances within the video. Towards this goal, we present a two-branch network for VOS, where the query-based instance segmentation (IS) branch delves into the instance details of the current frame and the VOS branch performs spatial-temporal matching with the memory bank. We employ the well-learned object queries from IS branch to inject instance-specific information into the query key, with which the instance-augmented matching is further performed. In addition, we introduce a multi-path fusion block to effectively combine the memory readout with multi-scale features from the instance segmentation decoder, which incorporates high-resolution instance-aware features to produce final segmentation results. Our method achieves state-of-the-art performance on DAVIS 2016/2017 val (92.6% and 87.1%), DAVIS 2017 test-dev (82.8%), and YouTube-VOS 2018/2019 val (86.3% and 86.3%), outperforming alternative methods by clear margins."
                },
                {
                    "title": "DeepMapping2: Self-Supervised Large-Scale LiDAR Map Optimization",
                    "abstract": "LiDAR mapping is important yet challenging in selfdriving and mobile robotics. To tackle such a global point cloud registration problem, DeepMapping [1] converts the complex map estimation into a self-supervised training of simple deep networks. Despite its broad convergence range on small datasets, DeepMapping still cannot produce satisfactory results on large-scale datasets with thousands of frames. This is due to the lack of loop closures and exact cross-frame point correspondences, and the slow convergence of its global localization network. We propose DeepMapping2 by adding two novel techniques to address these issues: (1) organization of training batch based on map topology from loop closing, and (2) self-supervised local-to-global point consistency loss leveraging pairwise registration. Our experiments and ablation studies on public datasets such as KITTI, NCLT, and Nebula demonstrate the effectiveness of our method."
                },
                {
                    "title": "LVOS: A Benchmark for Long-term Video Object Segmentation",
                    "abstract": "Existing video object segmentation (VOS) benchmarks focus on short-term videos which just last about 3-5 seconds and where objects are visible most of the time. These videos are poorly representative of practical applications, and the absence of long-term datasets restricts further investigation of VOS on the application in realistic scenarios. So, in this paper, we present a new benchmark dataset named LVOS, which consists of 220 videos with a total duration of 421 minutes. To the best of our knowledge, LVOS is the first densely annotated long-term VOS dataset. The videos in our LVOS last 1.59 minutes on average, which is 20 times longer than videos in existing VOS datasets. Each video includes various attributes, especially challenges deriving from the wild, such as long-term reappearing and cross-temporal similar objeccts. Based on LVOS, we assess existing video object segmentation algorithms and propose a Diverse Dynamic Memory network (DDMemory) that consists of three complementary memory banks to exploit temporal information adequately. The experimental results demonstrate the strength and weaknesses of prior methods, pointing promising directions for further study. Data and code are available at https://lingyihongfd.github.io/lvos.github.io/."
                },
                {
                    "title": "MOTRv2: Bootstrapping End-to-End Multi-Object Tracking by Pretrained Object Detectors",
                    "abstract": "In this paper, we propose MOTRv2, a simple yet effective pipeline to bootstrap end-to-end multi-object tracking with a pretrained object detector. Existing end-to-end methods, e.g. MOTR [43] and TrackFormer [20] are inferior to their tracking-by-detection counterparts mainly due to their poor detection performance. We aim to improve MOTR by elegantly incorporating an extra object detector. We first adopt the anchor formulation of queries and then use an extra object detector to generate proposals as anchors, providing detection prior to MOTR. The simple modification greatly eases the conflict between joint learning detection and association tasks in MOTR. MOTRv2 keeps the query propogation feature and scales well on large-scale benchmarks. MOTRv2 ranks the 1st place (73.4% HOTA on DanceTrack) in the 1st Multiple People Tracking in Group Dance Challenge. Moreover, MOTRv2 reaches state-of-the-art performance on the BDD100K dataset. We hope this simple and effective pipeline can provide some new insights to the end-to-end MOT community. Code is available at https://github.com/megvii-research/MOTRv2."
                },
                {
                    "title": "Topological Semantic Graph Memory for Image-Goal Navigation",
                    "abstract": "A novel framework is proposed to incrementally collect landmark-based graph memory and use the collected memory for image goal navigation. Given a target image to search, an embodied robot utilizes semantic memory to find the target in an unknown environment. % The semantic graph memory is collected from a panoramic observation of an RGB-D camera without knowing the robot's pose. In this paper, we present a topological semantic graph memory (TSGM), which consists of (1) a graph builder that takes the observed RGB-D image to construct a topological semantic graph, (2) a cross graph mixer module that takes the collected nodes to get contextual information, and (3) a memory decoder that takes the contextual memory as an input to find an action to the target. On the task of image goal navigation, TSGM significantly outperforms competitive baselines by +5.0-9.0% on the success rate and +7.0-23.5% on SPL, which means that the TSGM finds efficient paths. Additionally, we demonstrate our method on a mobile robot in real-world image goal scenarios."
                },
                {
                    "title": "Self-Supervised Visual Place Recognition by Mining Temporal and Feature Neighborhoods",
                    "abstract": "Visual place recognition (VPR) using deep networks has achieved state-of-the-art performance. However, most of them require a training set with ground truth sensor poses to obtain positive and negative samples of each observation's spatial neighborhood for supervised learning. When such information is unavailable, temporal neighborhoods from a sequentially collected data stream could be exploited for self-supervised training, although we find its performance suboptimal. Inspired by noisy label learning, we propose a novel self-supervised framework named \\textit{TF-VPR} that uses temporal neighborhoods and learnable feature neighborhoods to discover unknown spatial neighborhoods. Our method follows an iterative training paradigm which alternates between: (1) representation learning with data augmentation, (2) positive set expansion to include the current feature space neighbors, and (3) positive set contraction via geometric verification. We conduct comprehensive experiments on both simulated and real datasets, with either RGB images or point clouds as inputs. The results show that our method outperforms our baselines in recall rate, robustness, and heading diversity, a novel metric we propose for VPR. Our code and datasets can be found at https://ai4ce.github.io/TF-VPR/."
                },
                {
                    "title": "XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model",
                    "abstract": "We present XMem, a video object segmentation architecture for long videos with unified feature memory stores inspired by the Atkinson-Shiffrin memory model. Prior work on video object segmentation typically only uses one type of feature memory. For videos longer than a minute, a single feature memory model tightly links memory consumption and accuracy. In contrast, following the Atkinson-Shiffrin model, we develop an architecture that incorporates multiple independent yet deeply-connected feature memory stores: a rapidly updated sensory memory, a high-resolution working memory, and a compact thus sustained long-term memory. Crucially, we develop a memory potentiation algorithm that routinely consolidates actively used working memory elements into the long-term memory, which avoids memory explosion and minimizes performance decay for long-term prediction. Combined with a new memory reading mechanism, XMem greatly exceeds state-of-the-art performance on long-video datasets while being on par with state-of-the-art methods (that do not work on long videos) on short-video datasets. Code is available at https://hkchengrex.github.io/XMem"
                },
                {
                    "title": "Deep Visual Geo-localization Benchmark",
                    "abstract": "In this paper, we propose a new open-source benchmarkingframeworkfor Visual Geo-localization (VG) that allows to build, train, and test a wide range of commonly used ar-chitectures, with the flexibility to change individual components of a geo-localization pipeline. The purpose of this framework is twofold: i) gaining insights into how differ-ent components and design choices in a VG pipeline im-pact the final results, both in terms of performance (re-call@N metric) and system requirements (such as execution time and memory consumption); ii) establish a system-atic evaluation protocol for comparing different methods. Using the proposed framework, we perform a large suite of experiments which provide criteria for choosing back-bone, aggregation and negative mining depending on the use-case and requirements. We also assess the impact of engineering techniques like pre/post-processing, data aug-mentation and image resizing, showing that better performance can be obtained through somewhat simple procedures: for example, downscaling the images' resolution to 80% can lead to similar results with a 36% savings in ex-traction time and dataset storage requirement. Code and trained models are available at dataset storage requirement. https://deep-vg-bench.herokuapp.com/."
                },
                {
                    "title": "Continuous Scene Representations for Embodied AI",
                    "abstract": "We propose Continuous Scene Representations (CSR), a scene representation constructed by an embodied agent navigating within a space, where objects and their relationships are modeled by continuous valued embeddings. Our method captures feature relationships between objects, composes them into a graph structure on-the-fly, and situates an embodied agent within the representation. Our key insight is to embed pair-wise relationships between objects in a latent space. This allows for a richer representation compared to discrete relations (e.g., [SUPPORT], [NEXT-TO]) commonly used for building scene representations. CSR can track objects as the agent moves in a scene, update the representation accordingly, and detect changes in room configurations. Using CSR, we outperform state-of-the-art approaches for the challenging downstream task of visual room rearrangement, without any task specific training. Moreover, we show the learned embeddings capture salient spatial details of the scene and show applicability to real world data. A summery video and code is available at prior.allenai.org/projects/csr."
                },
                {
                    "title": "Hydra: A Real-time Spatial Perception System for 3D Scene Graph Construction and Optimization",
                    "abstract": "3D scene graphs have recently emerged as a powerful high-level representation of 3D environments. A 3D scene graph describes the environment as a layered graph where nodes represent spatial concepts at multiple levels of abstraction and edges represent relations between concepts. While 3D scene graphs can serve as an advanced\"mental model\"for robots, how to build such a rich representation in real-time is still uncharted territory. This paper describes a real-time Spatial Perception System, a suite of algorithms to build a 3D scene graph from sensor data in real-time. Our first contribution is to develop real-time algorithms to incrementally construct the layers of a scene graph as the robot explores the environment; these algorithms build a local Euclidean Signed Distance Function (ESDF) around the current robot location, extract a topological map of places from the ESDF, and then segment the places into rooms using an approach inspired by community-detection techniques. Our second contribution is to investigate loop closure detection and optimization in 3D scene graphs. We show that 3D scene graphs allow defining hierarchical descriptors for loop closure detection; our descriptors capture statistics across layers in the scene graph, ranging from low-level visual appearance to summary statistics about objects and places. We then propose the first algorithm to optimize a 3D scene graph in response to loop closures; our approach relies on embedded deformation graphs to simultaneously correct all layers of the scene graph. We implement the proposed Spatial Perception System into a architecture named Hydra, that combines fast early and mid-level perception processes with slower high-level perception. We evaluate Hydra on simulated and real data and show it is able to reconstruct 3D scene graphs with an accuracy comparable with batch offline methods despite running online."
                },
                {
                    "title": "A ConvNet for the 2020s",
                    "abstract": "The \u201cRoaring 20s\u201d of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually \u201cmodernize\u201d a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets."
                },
                {
                    "title": "ARKitScenes: A Diverse Real-World Dataset For 3D Indoor Scene Understanding Using Mobile RGB-D Data",
                    "abstract": "Scene understanding is an active research area. Commercial depth sensors, such as Kinect, have enabled the release of several RGB-D datasets over the past few years which spawned novel methods in 3D scene understanding. More recently with the launch of the LiDAR sensor in Apple's iPads and iPhones, high quality RGB-D data is accessible to millions of people on a device they commonly use. This opens a whole new era in scene understanding for the Computer Vision community as well as app developers. The fundamental research in scene understanding together with the advances in machine learning can now impact people's everyday experiences. However, transforming these scene understanding methods to real-world experiences requires additional innovation and development. In this paper we introduce ARKitScenes. It is not only the first RGB-D dataset that is captured with a now widely available depth sensor, but to our best knowledge, it also is the largest indoor scene understanding data released. In addition to the raw and processed data from the mobile device, ARKitScenes includes high resolution depth maps captured using a stationary laser scanner, as well as manually labeled 3D oriented bounding boxes for a large taxonomy of furniture. We further analyze the usefulness of the data for two downstream tasks: 3D object detection and color-guided depth upsampling. We demonstrate that our dataset can help push the boundaries of existing state-of-the-art methods and it introduces new challenges that better represent real-world scenarios."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras",
                    "abstract": "We introduce DROID-SLAM, a new deep learning based SLAM system. DROID-SLAM consists of recurrent iterative updates of camera pose and pixelwise depth through a Dense Bundle Adjustment layer. DROID-SLAM is accurate, achieving large improvements over prior work, and robust, suffering from substantially fewer catastrophic failures. Despite training on monocular video, it can leverage stereo or RGB-D video to achieve improved performance at test time. The URL to our open source code is https://github.com/princeton-vl/DROID-SLAM."
                },
                {
                    "title": "Do Different Tracking Tasks Require Different Appearance Models?",
                    "abstract": "Tracking objects of interest in a video is one of the most popular and widely applicable problems in computer vision. However, with the years, a Cambrian explosion of use cases and benchmarks has fragmented the problem in a multitude of different experimental setups. As a consequence, the literature has fragmented too, and now novel approaches proposed by the community are usually specialised to fit only one specific setup. To understand to what extent this specialisation is necessary, in this work we present UniTrack, a solution to address five different tasks within the same framework. UniTrack consists of a single and task-agnostic appearance model, which can be learned in a supervised or self-supervised fashion, and multiple ``heads'' that address individual tasks and do not require training. We show how most tracking tasks can be solved within this framework, and that the same appearance model can be successfully used to obtain results that are competitive against specialised methods for most of the tasks considered. The framework also allows us to analyse appearance models obtained with the most recent self-supervised methods, thus extending their evaluation and comparison to a larger variety of important problems."
                },
                {
                    "title": "VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning",
                    "abstract": "Recent self-supervised methods for image representation learning are based on maximizing the agreement between embedding vectors from different views of the same image. A trivial solution is obtained when the encoder outputs constant vectors. This collapse problem is often avoided through implicit biases in the learning architecture, that often lack a clear justification or interpretation. In this paper, we introduce VICReg (Variance-Invariance-Covariance Regularization), a method that explicitly avoids the collapse problem with a simple regularization term on the variance of the embeddings along each dimension individually. VICReg combines the variance term with a decorrelation mechanism based on redundancy reduction and covariance regularization, and achieves results on par with the state of the art on several downstream tasks. In addition, we show that incorporating our new variance term into other methods helps stabilize the training and leads to performance improvements."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "SceneGraphFusion: Incremental 3D Scene Graph Prediction from RGB-D Sequences",
                    "abstract": "Scene graphs are a compact and explicit representation successfully used in a variety of 2D scene understanding tasks. This work proposes a method to incrementally build up semantic scene graphs from a 3D environment given a sequence of RGB-D frames. To this end, we aggregate PointNet features from primitive scene components by means of a graph neural network. We also propose a novel attention mechanism well suited for partial and missing graph data present in such an incremental reconstruction scenario. Although our proposed method is designed to run on submaps of the scene, we show it also transfers to entire 3D scenes. Experiments show that our approach outperforms 3D scene graph prediction methods by a large margin and its accuracy is on par with other 3D semantic and panoptic segmentation methods while running at 35Hz."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "In Defense of Scene Graphs for Image Captioning",
                    "abstract": "The mainstream image captioning models rely on Convolutional Neural Network (CNN) image features to generate captions via recurrent models. Recently, image scene graphs have been used to augment captioning models so as to leverage their structural semantics, such as object entities, relationships and attributes. Several studies have noted that the naive use of scene graphs from a black-box scene graph generator harms image captioning performance and that scene graph-based captioning models have to incur the overhead of explicit use of image features to generate decent captions. Addressing these challenges, we propose SG2Caps, a framework that utilizes only the scene graph labels for competitive image captioning performance. The basic idea is to close the semantic gap between the two scene graphs - one derived from the input image and the other from its caption. In order to achieve this, we leverage the spatial location of objects and the Human-Object-Interaction (HOI) labels as an additional HOI graph. SG2Caps outperforms existing scene graph-only captioning models by a large margin, indicating scene graphs as a promising representation for image captioning. Direct utilization of scene graph labels avoids expensive graph convolutions over high-dimensional CNN features resulting in 49% fewer trainable parameters. Our code is available at: https://github.com/Kien085/SG2Caps"
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Neural Topological SLAM for Visual Navigation",
                    "abstract": "This paper studies the problem of image-goal navigation which involves navigating to the location indicated by a goal image in a novel previously unseen environment. To tackle this problem, we design topological representations for space that effectively leverage semantics and afford approximate geometric reasoning. At the heart of our representations are nodes with associated semantic features, that are interconnected using coarse geometric information. We describe supervised learning-based algorithms that can build, maintain and use such representations under noisy actuation. Experimental study in visually and physically realistic simulation suggests that our method builds effective representations that capture structural regularities and efficiently solve long-horizon navigation problems. We observe a relative improvement of more than 50% over existing methods that study this task."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "SuperGlue: Learning Feature Matching With Graph Neural Networks",
                    "abstract": "This paper introduces SuperGlue, a neural network that matches two sets of local features by jointly finding correspondences and rejecting non-matchable points. Assignments are estimated by solving a differentiable optimal transport problem, whose costs are predicted by a graph neural network. We introduce a flexible context aggregation mechanism based on attention, enabling SuperGlue to reason about the underlying 3D scene and feature assignments jointly. Compared to traditional, hand-designed heuristics, our technique learns priors over geometric transformations and regularities of the 3D world through end-to-end training from image pairs. SuperGlue outperforms other learned approaches and achieves state-of-the-art results on the task of pose estimation in challenging real-world indoor and outdoor environments. The proposed method performs matching in real-time on a modern GPU and can be readily integrated into modern SfM or SLAM systems. The code and trained weights are publicly available at github.com/magicleap/SuperGluePretrainedNetwork."
                },
                {
                    "title": "Momentum Contrast for Unsupervised Visual Representation Learning",
                    "abstract": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a moving-averaged encoder. This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning. MoCo provides competitive results under the common linear protocol on ImageNet classification. More importantly, the representations learned by MoCo transfer well to downstream tasks. MoCo can outperform its supervised pre-training counterpart in 7 detection/segmentation tasks on PASCAL VOC, COCO, and other datasets, sometimes surpassing it by large margins. This suggests that the gap between unsupervised and supervised representation learning has been largely closed in many vision tasks."
                },
                {
                    "title": "3D Scene Graph: A Structure for Unified Semantics, 3D Space, and Camera",
                    "abstract": "A comprehensive semantic understanding of a scene is important for many applications - but in what space should diverse semantic information (e.g., objects, scene categories, material types, 3D shapes, etc.) be grounded and what should be its structure? Aspiring to have one unified structure that hosts diverse types of semantics, we follow the Scene Graph paradigm in 3D, generating a 3D Scene Graph. Given a 3D mesh and registered panoramic images, we construct a graph that spans the entire building and includes semantics on objects (e.g., class, material, shape and other attributes), rooms (e.g., function, illumination type, etc.) and cameras (e.g., location, etc.), as well as the relationships among these entities. However, this process is prohibitively labor heavy if done manually. To alleviate this we devise a semi-automatic framework that employs existing detection methods and enhances them using two main constraints: I. framing of query images sampled on panoramas to maximize the performance of 2D detectors, and II. multi-view consistency enforcement across 2D detections that originate in different camera locations."
                },
                {
                    "title": "Generalized Intersection Over Union: A Metric and a Loss for Bounding Box Regression",
                    "abstract": "Intersection over Union (IoU) is the most popular evaluation metric used in the object detection benchmarks. However, there is a gap between optimizing the commonly used distance losses for regressing the parameters of a bounding box and maximizing this metric value. The optimal objective for a metric is the metric itself. In the case of axis-aligned 2D bounding boxes, it can be shown that IoU can be directly used as a regression loss. However, IoU has a plateau making it infeasible to optimize in the case of non-overlapping bounding boxes. In this paper, we address the this weakness by introducing a generalized version of IoU as both a new loss and a new metric. By incorporating this generalized IoU ( GIoU) as a loss into the state-of-the art object detection frameworks, we show a consistent improvement on their performance using both the standard, IoU based, and new, GIoU based, performance measures on popular object detection benchmarks such as PASCAL VOC and MS COCO."
                },
                {
                    "title": "LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain",
                    "abstract": "We propose a lightweight and ground-optimized lidar odometry and mapping method, LeGO-LOAM, for realtime six degree-of-freedom pose estimation with ground vehicles. LeGO-LOAM is lightweight, as it can achieve realtime pose estimation on a low-power embedded system. LeGO-LOAM is ground-optimized, as it leverages the presence of a ground plane in its segmentation and optimization steps. We first apply point cloud segmentation to filter out noise, and feature extraction to obtain distinctive planar and edge features. A two-step Levenberg-Marquardt optimization method then uses the planar and edge features to solve different components of the six degree-of-freedom transformation across consecutive scans. We compare the performance of LeGO-LOAM with a state-of-the-art method, LOAM, using datasets gathered from variable-terrain environments with ground vehicles, and show that LeGO-LOAM achieves similar or better accuracy with reduced computational expense. We also integrate LeGO-LOAM into a SLAM framework to eliminate the pose estimation error caused by drift, which is tested using the KITTI dataset."
                },
                {
                    "title": "Semi-parametric Topological Memory for Navigation",
                    "abstract": "We introduce a new memory architecture for navigation in previously unseen environments, inspired by landmark-based navigation in animals. The proposed semi-parametric topological memory (SPTM) consists of a (non-parametric) graph with nodes corresponding to locations in the environment and a (parametric) deep network capable of retrieving nodes from the graph based on observations. The graph stores no metric information, only connectivity of locations corresponding to the nodes. We use SPTM as a planning module in a navigation system. Given only 5 minutes of footage of a previously unseen maze, an SPTM-based navigation agent can build a topological map of the environment and use it to confidently navigate towards goals. The average success rate of the SPTM agent in goal-directed navigation across test environments is higher than the best-performing baseline by a factor of three. A video of the agent is available at this https URL"
                },
                {
                    "title": "Topomap: Topological Mapping and Navigation Based on Visual SLAM Maps",
                    "abstract": "Visual robot navigation within large-scale, semistructured environments deals with various challenges such as computation intensive path planning algorithms or insufficient knowledge about traversable spaces. Moreover, many state-of-the-art navigation approaches only operate locally instead of gaining a more conceptual understanding of the planning objective. This limits the complexity of tasks a robot can accomplish and makes it harder to deal with uncertainties that are present in the context of real-time robotics applications. In this work, we present Topomap, a framework which simplifies the navigation task by providing a map to the robot which is tailored for path planning use. This novel approach transforms a sparse feature-based map from a visual Simultaneous Localization And Mapping (SLAM) system into a three-dimensional topological map. This is done in two steps. First, we extract occupancy information directly from the noisy sparse point cloud. Then, we create a set of convex free-space clusters, which are the vertices of the topological map. We show that this representation improves the efficiency of global planning, and we provide a complete derivation of our algorithm. Planning experiments on real world datasets demonstrate that we achieve similar performance as RRT* with significantly lower computation times and storage requirements. Finally, we test our algorithm on a mobile robotic platform to prove its advantages."
                },
                {
                    "title": "Scene Graph Generation by Iterative Message Passing",
                    "abstract": "Understanding a visual scene goes beyond recognizing individual objects in isolation. Relationships between objects also constitute rich semantic information about the scene. In this work, we explicitly model the objects and their relationships using scene graphs, a visually-grounded graphical structure of an image. We propose a novel end-to-end model that generates such structured scene representation from an input image. Our key insight is that the graph generation problem can be formulated as message passing between the primal node graph and its dual edge graph. Our joint inference model can take advantage of contextual cues to make better predictions on objects and their relationships. The experiments show that our model significantly outperforms previous methods on the Visual Genome dataset as well as support relation inference in NYU Depth V2 dataset."
                },
                {
                    "title": "Understanding intermediate layers using linear classifier probes",
                    "abstract": "Neural network models have a reputation for being black boxes. We propose to monitor the features at every layer of a model and measure how suitable they are for classification. We use linear classifiers, which we refer to as \"probes\", trained entirely independently of the model itself. \nThis helps us better understand the roles and dynamics of the intermediate layers. We demonstrate how this can be used to develop a better intuition about models and to diagnose potential problems. \nWe apply this technique to the popular models Inception v3 and Resnet-50. Among other things, we observe experimentally that the linear separability of features increase monotonically along the depth of the model."
                },
                {
                    "title": "Visual Place Recognition: A Survey",
                    "abstract": "Visual place recognition is a challenging problem due to the vast range of ways in which the appearance of real-world places can vary. In recent years, improvements in visual sensing capabilities, an ever-increasing focus on long-term mobile robot autonomy, and the ability to draw on state-of-the-art research in other disciplines-particularly recognition in computer vision and animal navigation in neuroscience-have all contributed to significant advances in visual place recognition systems. This paper presents a survey of the visual place recognition research landscape. We start by introducing the concepts behind place recognition-the role of place recognition in the animal kingdom, how a \u201cplace\u201d is defined in a robotics context, and the major components of a place recognition system. Long-term robot operations have revealed that changing appearance can be a significant factor in visual place recognition failure; therefore, we discuss how place recognition solutions can implicitly or explicitly account for appearance change within the environment. Finally, we close with a discussion on the future of visual place recognition, in particular with respect to the rapid advances being made in the related fields of deep learning, semantic scene understanding, and video description."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "NetVLAD: CNN Architecture for Weakly Supervised Place Recognition",
                    "abstract": "We tackle the problem of large scale visual place recognition, where the task is to quickly and accurately recognize the location of a given query photograph. We present the following three principal contributions. First, we develop a convolutional neural network (CNN) architecture that is trainable in an end-to-end manner directly for the place recognition task. The main component of this architecture, NetVLAD, is a new generalized VLAD layer, inspired by the \"Vector of Locally Aggregated Descriptors\" image representation commonly used in image retrieval. The layer is readily pluggable into any CNN architecture and amenable to training via backpropagation. Second, we develop a training procedure, based on a new weakly supervised ranking loss, to learn parameters of the architecture in an end-to-end manner from images depicting the same places over time downloaded from Google Street View Time Machine. Finally, we show that the proposed architecture significantly outperforms non-learnt image representations and off-the-shelf CNN descriptors on two challenging place recognition benchmarks, and improves over current state of-the-art compact image representations on standard image retrieval benchmarks."
                },
                {
                    "title": "Image retrieval using scene graphs",
                    "abstract": "This paper develops a novel framework for semantic image retrieval based on the notion of a scene graph. Our scene graphs represent objects (\u201cman\u201d, \u201cboat\u201d), attributes of objects (\u201cboat is white\u201d) and relationships between objects (\u201cman standing on boat\u201d). We use these scene graphs as queries to retrieve semantically related images. To this end, we design a conditional random field model that reasons about possible groundings of scene graphs to test images. The likelihoods of these groundings are used as ranking scores for retrieval. We introduce a novel dataset of 5,000 human-generated scene graphs grounded to images and use this dataset to evaluate our method for image retrieval. In particular, we evaluate retrieval using full scene graphs and small scene subgraphs, and show that our method outperforms retrieval methods that use only objects or low-level image features. In addition, we show that our full model can be used to improve object localization compared to baseline methods."
                },
                {
                    "title": "ORB-SLAM: A Versatile and Accurate Monocular SLAM System",
                    "abstract": "This paper presents ORB-SLAM, a feature-based monocular simultaneous localization and mapping (SLAM) system that operates in real time, in small and large indoor and outdoor environments. The system is robust to severe motion clutter, allows wide baseline loop closing and relocalization, and includes full automatic initialization. Building on excellent algorithms of recent years, we designed from scratch a novel system that uses the same features for all SLAM tasks: tracking, mapping, relocalization, and loop closing. A survival of the fittest strategy that selects the points and keyframes of the reconstruction leads to excellent robustness and generates a compact and trackable map that only grows if the scene content changes, allowing lifelong operation. We present an exhaustive evaluation in 27 sequences from the most popular datasets. ORB-SLAM achieves unprecedented performance with respect to other state-of-the-art monocular SLAM approaches. For the benefit of the community, we make the source code public."
                },
                {
                    "title": "SLAM++: Simultaneous Localisation and Mapping at the Level of Objects",
                    "abstract": "We present the major advantages of a new 'object oriented' 3D SLAM paradigm, which takes full advantage in the loop of prior knowledge that many scenes consist of repeated, domain-specific objects and structures. As a hand-held depth camera browses a cluttered scene, real-time 3D object recognition and tracking provides 6DoF camera-object constraints which feed into an explicit graph of objects, continually refined by efficient pose-graph optimisation. This offers the descriptive and predictive power of SLAM systems which perform dense surface reconstruction, but with a huge representation compression. The object graph enables predictions for accurate ICP-based camera to model tracking at each live frame, and efficient active search for new objects in currently undescribed image regions. We demonstrate real-time incremental SLAM in large, cluttered environments, including loop closure, relocalisation and the detection of moved objects, and of course the generation of an object level scene description with the potential to enable interaction."
                },
                {
                    "title": "Object recognition from local scale-invariant features",
                    "abstract": "An object recognition system has been developed that uses a new class of local image features. The features are invariant to image scaling, translation, and rotation, and partially invariant to illumination changes and affine or 3D projection. These features share similar properties with neurons in inferior temporal cortex that are used for object recognition in primate vision. Features are efficiently detected through a staged filtering approach that identifies stable points in scale space. Image keys are created that allow for local geometric deformations by representing blurred image gradients in multiple orientation planes and at multiple scales. The keys are used as input to a nearest neighbor indexing method that identifies candidate object matches. Final verification of each match is achieved by finding a low residual least squares solution for the unknown model parameters. Experimental results show that robust object recognition can be achieved in cluttered partially occluded images with a computation time of under 2 seconds."
                },
                {
                    "title": "Binding of hippocampal CA1 neural activity to multiple reference frames in a landmark-based navigation task",
                    "abstract": "The behavioral correlates of rat hippocampal CA1 cells were examined in a spatial navigation task in which two cylindrical landmarks predicted the location of food. The landmarks were maintained at a constant distance from each other but were moved from trial to trial within a large arena surrounded by static background cues. On each trial, the rats were released from a box to which they returned for additional food after locating the goal. The box also was located variably from trial to trial and was moved to a new location while the animals were searching for the goal site. The discharge characteristics of multiple, simultaneously recorded cells were examined with respect to the landmarks, the static background cues, and the box in which each trial started and ended. Three clear categories of cells were observed: (1) cells with location-specific firing (place cells); (2) goal/landmark-related cells that fired in the vicinity of the goal or landmarks, regardless of their location in the arena; and (3) box-related cells that fired either when the rat was in the box or as it was leaving or entering the box, regardless of its location in the arena. Disjunctive cells with separate firing fields in more than one reference frame also were observed. These results suggest that in this task a subpopulation of hippocampal cells encodes location in the fixed spatial frame, whereas other subpopulations encode location with respect to different reference frames associated with the task-relevant, mobile objects."
                },
                {
                    "title": "The effects of changes in the environment on the spatial firing of hippocampal complex-spike cells",
                    "abstract": "Using the techniques set out in the preceding paper (Muller et al., 1987), we investigated the response of place cells to changes in the animal's environment. The standard apparatus used was a cylinder, 76 cm in diameter, with walls 51 cm high. The interior was uniformly gray except for a white cue card that ran the full height of the wall and occupied 100 degrees of arc. The floor of the apparatus presented no obstacles to the animal's motions. Each of these major features of the apparatus was varied while the others were held constant. One set of manipulations involved the cue card. Rotating the cue card produced equal rotations of the firing fields of single cells. Changing the width of the card did not affect the size, shape, or radial position of firing fields, although sometimes the field rotated to a modest extent. Removing the cue card altogether also left the size, shape, and radial positions of firing fields unchanged, but caused fields to rotate to unpredictable angular positions. The second set of manipulations dealt with the size and shape of the apparatus wall. When the standard (small) cylinder was scaled up in diameter and height by a factor of 2, the firing fields of 36% of the cells observed in both cylinders also scaled, in the sense that the field stayed at the same angular position and at the same relative radial position. Of the cells recorded in both cylinders, 52% showed very different firing patterns in one cylinder than in the other. The remaining 12% of the cells were virtually silent in both cylinders. Similar results were obtained when individual cells were recorded in both a small and a large rectangular enclosure. By contrast, when the apparatus floor plan was changed from circular to rectangular, the firing pattern of a cell in an apparatus of one shape could not be predicted from a knowledge of the firing pattern in the other shape. The final manipulations involved placing vertical barriers into the otherwise unobstructed floor of the small cylinder. When an opaque barrier was set up to bisect a previously recorded firing field, in almost all cases the firing field was nearly abolished. This was true even though the barrier occupied only a small fraction of the firing field area. A transparent barrier was effective as the opaque barrier in attenuating firing fields. The lead base used to anchor the vertical barriers did not affect place cell firing.(ABSTRACT TRUNCATED AT 400 WORDS)"
                },
                {
                    "title": "Concerning nonnegative matrices and doubly stochastic matrices",
                    "abstract": "This paper is concerned with the condition for the convergence to a doubly stochastic limit of a sequence of matrices obtained from a nonnegative matrix A by alternately scaling the rows and columns of A and with the condition for the existence of diagonal matrices A and D2 with positive main diagonals such that \u038f\u03b3A\u038f2 is doubly stochastic. The result is the following. The sequence of matrices converges to a doubly stochastic limit if and only if the matrix A contains at least one positive diagonal. A necessary and sufficient condition that there exist diagonal matrices A and D2 with positive main diagonals such that D1AD2 is both doubly stochastic and the limit of the iteration is that A\u03c6O and each positive entry of A is contained in a positive diagonal. The form D\u03b9AD2 is unique, and A and D2 are unique up to a positive scalar multiple if and only if A is fully indecomposable."
                },
                {
                    "title": "Visualizing Data using t-SNE",
                    "abstract": "We present a new technique called \u201ct-SNE\u201d that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to influence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are significantly better than those produced by the other techniques on almost all of the datasets."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we build a Multiview Scene Graph (MSG) that effectively represents spatial correspondences among unposed RGB images taken from the same scene?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of building an MSG is significant for the research community as it enhances our understanding of spatial intelligence in AI, akin to human cognitive mapping. This advancement could lead to improved performance in various applications, such as robotics, augmented reality, and autonomous navigation, by enabling AI systems to better interpret and interact with their environments. Furthermore, it could inspire future research into more sophisticated models that integrate visual perception with spatial reasoning, ultimately advancing the field of computer vision and machine learning.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in building an MSG stem from the complexities of accurately associating images and objects without relying on metric maps or depth information. Naive approaches may fail because they often assume a fixed perspective or require precise spatial data, which is not available in many real-world scenarios. Additionally, the task involves overcoming technical obstacles such as effectively learning embeddings that capture both place recognition and object association, as well as ensuring the graph accurately reflects the spatial relationships among multiple views.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either topological mapping or semantic relationships in 2D and 3D scene graphs, often requiring depth and pose information that limits their applicability to everyday images. The lack of a unified approach that directly evaluates the quality of spatial understanding through a multiview scene graph has been a significant gap. Our approach differs by proposing a method that does not depend on metric data and directly assesses the model's ability to establish visual correspondences, thus addressing limitations in prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the Attention Association MSG (AoMSG), a Transformer-based architecture that learns place and object embeddings jointly within a single decoder. We will utilize a curated dataset from the ARKitScenes and evaluate the MSG generation using metrics based on the intersection-over-union of the graph adjacency matrix. The expected outcomes include demonstrating the superiority of AoMSG over baseline methods in accurately building the MSG, while also highlighting areas for further research in enhancing spatial intelligence in AI models."
            }
        },
        "author_data": {
            "a6b21ec5-9496-43f9-a135-bbac7c38beee": {
                "pk": "a6b21ec5-9496-43f9-a135-bbac7c38beee",
                "name": "Juexiao Zhang",
                "collaborators": [
                    "Chen Feng",
                    "Yiming Li",
                    "Yubei Chen",
                    "Irving Fang",
                    "Suozhi Huang",
                    "Zeyu Yun",
                    "Bruno Olshausen",
                    "Yann LeCun",
                    "Zonglin Lyu",
                    "Mingxuan Lu"
                ],
                "domain": [
                    "Robotics",
                    "Unsupervised Learning",
                    "Vision Language Models",
                    "Collaborative Perception"
                ],
                "publications": [
                    {
                        "title": "ActFormer: Scalable Collaborative Perception via Active Queries",
                        "abstract": "Collaborative perception leverages rich visual observations from multiple robots to extend a single robot's perception ability beyond its field of view. Many prior works receive messages broadcast from all collaborators, leading to a scalability challenge when dealing with a large number of robots and sensors. In this work, we aim to address \\textit{scalable camera-based collaborative perception} with a Transformer-based architecture. Our key idea is to enable a single robot to intelligently discern the relevance of the collaborators and their associated cameras according to a learned spatial prior. This proactive understanding of the visual features' relevance does not require the transmission of the features themselves, enhancing both communication and computation efficiency. Specifically, we present ActFormer, a Transformer that learns bird's eye view (BEV) representations by using predefined BEV queries to interact with multi-robot multi-camera inputs. Each BEV query can actively select relevant cameras for information aggregation based on pose information, instead of interacting with all cameras indiscriminately. Experiments on the V2X-Sim dataset demonstrate that ActFormer improves the detection performance from 29.89% to 45.15% in terms of AP@0.7 with about 50% fewer queries, showcasing the effectiveness of ActFormer in multi-agent collaborative 3D object detection."
                    },
                    {
                        "title": "URLOST: Unsupervised Representation Learning without Stationarity or Topology",
                        "abstract": "Unsupervised representation learning has seen tremendous progress but is constrained by its reliance on data modality-specific stationarity and topology, a limitation not found in biological intelligence systems. For instance, human vision processes visual signals derived from irregular and non-stationary sampling lattices yet accurately perceives the geometry of the world. We introduce a novel framework that learns from high-dimensional data lacking stationarity and topology. Our model combines a learnable self-organizing layer, density adjusted spectral clustering, and masked autoencoders. We evaluate its effectiveness on simulated biological vision data, neural recordings from the primary visual cortex, and gene expression datasets. Compared to state-of-the-art unsupervised learning methods like SimCLR and MAE, our model excels at learning meaningful representations across diverse modalities without depending on stationarity or topology. It also outperforms other methods not dependent on these factors, setting a new benchmark in the field. This work represents a step toward unsupervised learning methods that can generalize across diverse high-dimensional data modalities."
                    },
                    {
                        "title": "Tell Me Where You Are: Multimodal LLMs Meet Place Recognition",
                        "abstract": "Large language models (LLMs) exhibit a variety of promising capabilities in robotics, including long-horizon planning and commonsense reasoning. However, their performance in place recognition is still underexplored. In this work, we introduce multimodal LLMs (MLLMs) to visual place recognition (VPR), where a robot must localize itself using visual observations. Our key design is to use vision-based retrieval to propose several candidates and then leverage language-based reasoning to carefully inspect each candidate for a final decision. Specifically, we leverage the robust visual features produced by off-the-shelf vision foundation models (VFMs) to obtain several candidate locations. We then prompt an MLLM to describe the differences between the current observation and each candidate in a pairwise manner, and reason about the best candidate based on these descriptions. Our results on three datasets demonstrate that integrating the general-purpose visual features from VFMs with the reasoning capabilities of MLLMs already provides an effective place recognition solution, without any VPR-specific supervised training. We believe our work can inspire new possibilities for applying and designing foundation models, i.e., VFMs, LLMs, and MLLMs, to enhance the localization and navigation of mobile robots."
                    },
                    {
                        "title": "Word Embedding Visualization Via Dictionary Learning",
                        "abstract": "Co-occurrence statistics based word embedding techniques have proved to be very useful in extracting the semantic and syntactic representation of words as low dimensional continuous vectors. In this work, we discovered that dictionary learning can open up these word vectors as a linear combination of more elementary word factors. We demonstrate many of the learned factors have surprisingly strong semantic or syntactic meaning corresponding to the factors previously identified manually by human inspection. Thus dictionary learning provides a powerful visualization tool for understanding word embedding representations. Furthermore, we show that the word factors can help in identifying key semantic and syntactic differences in word analogy tasks and improve upon the state-of-the-art word embedding techniques in these tasks by a large margin."
                    },
                    {
                        "title": "VLM See, Robot Do: Human Demo Video to Robot Action Plan via Vision Language Model",
                        "abstract": "Vision Language Models (VLMs) have recently been adopted in robotics for their capability in common sense reasoning and generalizability. Existing work has applied VLMs to generate task and motion planning from natural language instructions and simulate training data for robot learning. In this work, we explore using VLM to interpret human demonstration videos and generate robot task planning. Our method integrates keyframe selection, visual perception, and VLM reasoning into a pipeline. We named it SeeDo because it enables the VLM to ''see'' human demonstrations and explain the corresponding plans to the robot for it to ''do''. To validate our approach, we collected a set of long-horizon human videos demonstrating pick-and-place tasks in three diverse categories and designed a set of metrics to comprehensively benchmark SeeDo against several baselines, including state-of-the-art video-input VLMs. The experiments demonstrate SeeDo's superior performance. We further deployed the generated task plans in both a simulation environment and on a real robot arm."
                    },
                    {
                        "title": "LUWA Dataset: Learning Lithic Use-Wear Analysis on Microscopic Images",
                        "abstract": "Lithic Use-Wear Analysis (LUWA) using microscopic images is an underexplored vision-for-science research area. It seeks to distinguish the worked material, which is critical for understanding archaeological artifacts, material interactions, tool functionalities, and dental records. However, this challenging task goes beyond the well-studied image classification problem for common objects. It is affected by many confounders owing to the complex wear mechanism and microscopic imaging, which makes it difficult even for human experts to identify the worked material successfully. In this paper, we investigate the following three questions on this unique vision task for the first time:(i) How well can state-of-the-art pre-trained models (like DINOv2) generalize to the rarely seen domain? (ii) How can few-shot learning be exploited for scarce microscopic images? (iii) How do the ambiguous magnification and sensing modality influence the classification accuracy? To study these, we collaborated with archaeologists and built the first open-source and the largest LUWA dataset containing 23,130 microscopic images with different magnifications and sensing modalities. Extensive experiments show that existing pre-trained models notably outperform human experts but still leave a large gap for improvements. Most importantly, the LUWA dataset provides an underexplored opportunity for vision and learning communities and complements existing image classification problems on common objects."
                    }
                ]
            },
            "4fb0ce69-0952-4285-80d7-64678237317a": {
                "pk": "4fb0ce69-0952-4285-80d7-64678237317a",
                "name": "Gao Zhu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "212fd233-332c-428b-98c8-83c73e3a3f61": {
                "pk": "212fd233-332c-428b-98c8-83c73e3a3f61",
                "name": "Sihang Li",
                "collaborators": [
                    "Changhong Fu",
                    "Junjie Ye",
                    "Xiang Wang",
                    "Xiangnan He",
                    "Ziang Cao",
                    "Tat-Seng Chua",
                    "Guangze Zheng",
                    "Kuangzheng Li",
                    "Haibing Lu",
                    "Amy Li"
                ],
                "domain": [
                    "Artificial Intelligence",
                    "Deep Learning",
                    "Unmanned Aerial Vehicles",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "National Origin Discrimination in Deep-learning-powered Automated Resume Screening",
                        "abstract": "Many companies and organizations have started to use some form of AIenabled auto mated tools to assist in their hiring process, e.g. screening resumes, interviewing candi dates, performance evaluation. While those AI tools have greatly improved human re source operations efficiency and provided conveniences to job seekers as well, there are increasing concerns on unfair treatment to candidates, caused by underlying bias in AI systems. Laws around equal opportunity and fairness, like GDPR, CCPA, are introduced or under development, in attempt to regulate AI. However, it is difficult to implement AI regulations in practice, as technologies are constantly advancing and the risk perti nent to their applications can fail to be recognized. This study examined deep learning methods, a recent technology breakthrough, with focus on their application to automated resume screening. One impressive performance of deep learning methods is the represen tation of individual words as lowdimensional numerical vectors, called word embedding, which are learned from aggregated global wordword cooccurrence statistics from a cor pus, like Wikipedia or Google news. The resulting word representations possess interest ing linear substructures of the word vector space and have been widely used in down stream tasks, like resume screening. However, word embedding inherits and reinforces the stereotyping from the training corpus, as deep learning models essentially learn a probability distribution of words and their relations from history data. Our study finds out that if we rely on such deeplearningpowered automated resume screening tools, it may lead to decisions favoring or disfavoring certain demographic groups and raise eth ical, even legal, concerns. To address the issue, we developed bias mitigation method. Extensive experiments on real candidate resumes are conducted to validate our study"
                    },
                    {
                        "title": "Uncertainty-Aware Decarbonization for Datacenters",
                        "abstract": "This paper represents the first effort to quantify uncertainty in carbon intensity forecasting for datacenter decarbonization. We identify and analyze two types of uncertainty -- temporal and spatial -- and discuss their system implications. To address the temporal dynamics in quantifying uncertainty for carbon intensity forecasting, we introduce a conformal prediction-based framework. Evaluation results show that our technique robustly achieves target coverages in uncertainty quantification across various significance levels. We conduct two case studies using production power traces, focusing on temporal and spatial load shifting respectively. The results show that incorporating uncertainty into scheduling decisions can prevent a 5% and 14% increase in carbon emissions, respectively. These percentages translate to an absolute reduction of 2.1 and 10.4 tons of carbon emissions in a 20 MW datacenter cluster."
                    },
                    {
                        "title": "Direct 3D information fusion for depth of field enhancement in optical-resolution photoacoustic microscopy",
                        "abstract": "As an important branch of photoacoustic microscopy, optical-resolution photoacoustic microscopy suffers from limited depth of field due to the strongly focused laser beam. In this work, a 3D information fusion algorithm based on 3D stationary wavelet transform and joint weighted evaluation optimization is proposed to fuse multi-focus photoacoustic data to achieve large-volumetric and high-resolution 3D imaging. First, a three-dimensional stationary wavelet transform was performed on the multi-focus data to obtain eight wavelet coefficients. Differential evolution algorithm based on joint weighted evaluation was then employed to optimize the block size of division for each wavelet coefficient. Corresponding sub-coefficients of multi-focus 3D data were fused with the proposed fusion rule utilizing standard deviation for focus detection. Finally, photoacoustic microscopy with large depth of field can be achieved by applying the inverse stationary wavelet transform on the 8 fused sub-coefficients. The fusion result of multi-focus vertically tilted fiber shows that the depth of field of optical-resolution photoacoustic microscopy is doubled without sacrificing lateral resolution via the proposed method. Furthermore, the effectiveness of the proposed method was verified through the fusion results of multi-focus vessel data. Our work provides a feasible solution for achieving large-volumetric, high-resolution photoacoustic microscopy for further data analysis, processing and applications."
                    },
                    {
                        "title": "Towards Dynamic and Small Objects Refinement for Unsupervised Domain Adaptative Nighttime Semantic Segmentation",
                        "abstract": "Nighttime semantic segmentation plays a crucial role in practical applications, such as autonomous driving, where it frequently encounters difficulties caused by inadequate illumination conditions and the absence of well-annotated datasets. Moreover, semantic segmentation models trained on daytime datasets often face difficulties in generalizing effectively to nighttime conditions. Unsupervised domain adaptation (UDA) has shown the potential to address the challenges and achieved remarkable results for nighttime semantic segmentation. However, existing methods still face limitations in 1) their reliance on style transfer or relighting models, which struggle to generalize to complex nighttime environments, and 2) their ignorance of dynamic and small objects like vehicles and poles, which are difficult to be directly learned from other domains. This paper proposes a novel UDA method that refines both label and feature levels for dynamic and small objects for nighttime semantic segmentation. First, we propose a dynamic and small object refinement module to complement the knowledge of dynamic and small objects from the source domain to target the nighttime domain. These dynamic and small objects are normally context-inconsistent in under-exposed conditions. Then, we design a feature prototype alignment module to reduce the domain gap by deploying contrastive learning between features and prototypes of the same class from different domains, while re-weighting the categories of dynamic and small objects. Extensive experiments on three benchmark datasets demonstrate that our method outperforms prior arts by a large margin for nighttime segmentation. Project page: https://rorisis.github.io/DSRNSS/."
                    },
                    {
                        "title": "Ad2Attack: Adaptive Adversarial Attack on Real-Time UAV Tracking",
                        "abstract": "Visual tracking is adopted to extensive unmanned aerial vehicle (UAV)-related applications, which leads to a highly demanding requirement on the robustness of UAV trackers. However, adding imperceptible perturbations can easily fool the tracker and cause tracking failures. This risk is often overlooked and rarely researched at present. Therefore, to help increase awareness of the potential risk and the robustness of UAV tracking, this work proposes a novel adaptive adversarial attack approach, i.e., Ad$^2$Attack, against UAV object tracking. Specifically, adversarial examples are generated online during the resampling of the search patch image, which leads trackers to lose the target in the following frames. Ad$^2$Attack is composed of a direct downsampling module and a super-resolution upsampling module with adaptive stages. A novel optimization function is proposed for balancing the imperceptibility and efficiency of the attack. Comprehensive experiments on several well-known benchmarks and real-world conditions show the effectiveness of our attack method, which dramatically reduces the performance of the most advanced Siamese trackers."
                    },
                    {
                        "title": "Let Invariant Rationale Discovery Inspire Graph Contrastive Learning",
                        "abstract": "Leading graph contrastive learning (GCL) methods perform graph augmentations in two fashions: (1) randomly corrupting the anchor graph, which could cause the loss of semantic information, or (2) using domain knowledge to maintain salient features, which undermines the generalization to other domains. Taking an invariance look at GCL, we argue that a high-performing augmentation should preserve the salient semantics of anchor graphs regarding instance-discrimination. To this end, we relate GCL with invariant rationale discovery, and propose a new framework, Rationale-aware Graph Contrastive Learning (RGCL). Specifically, without supervision signals, RGCL uses a rationale generator to reveal salient features about graph instance-discrimination as the rationale, and then creates rationale-aware views for contrastive learning. This rationale-aware pre-training scheme endows the backbone model with the powerful representation ability, further facilitating the fine-tuning on downstream tasks. On MNIST-Superpixel and MUTAG datasets, visual inspections on the discovered rationales showcase that the rationale generator successfully captures the salient features (i.e. distinguishing semantic nodes in graphs). On biochemical molecule and social network benchmark datasets, the state-of-the-art performance of RGCL demonstrates the effectiveness of rationale-aware views for contrastive learning. Our codes are available at https://github.com/lsh0520/RGCL."
                    },
                    {
                        "title": "Local Perception-Aware Transformer for Aerial Tracking",
                        "abstract": "Transformer-based visual object tracking has been utilized extensively. However, the Transformer structure is lack of enough inductive bias. In addition, only focusing on encoding the global feature does harm to modeling local details, which restricts the capability of tracking in aerial robots. Specifically, with local-modeling to global-search mechanism, the proposed tracker replaces the global encoder by a novel local-recognition encoder. In the employed encoder, a local-recognition attention and a local element correction network are carefully designed for reducing the global redundant information interference and increasing local inductive bias. Meanwhile, the latter can model local object details precisely under aerial view through detail-inquiry net. The proposed method achieves competitive accuracy and robustness in several authoritative aerial benchmarks with 316 sequences in total. The proposed tracker's practicability and efficiency have been validated by the real-world tests."
                    },
                    {
                        "title": "SGDViT: Saliency-Guided Dynamic Vision Transformer for UAV Tracking",
                        "abstract": "Vision-based object tracking has boosted extensive autonomous applications for unmanned aerial vehicles (UAVs). However, the dynamic changes in flight maneuver and viewpoint encountered in UAV tracking pose significant difficulties, e.g. , aspect ratio change, and scale variation. The conventional cross-correlation operation, while commonly used, has limitations in effectively capturing perceptual similarity and incorporates extraneous background information. To mitigate these limitations, this work presents a novel saliency-guided dynamic vision Transformer (SGDViT) for UAV tracking. The proposed method designs a new task-specific object saliency mining network to refine the cross-correlation operation and effectively discriminate foreground and background information. Additionally, a saliency adaptation embedding operation dynamically generates tokens based on initial saliency, thereby reducing the computational complexity of the Transformer architecture. Finally, a lightweight saliency filtering Transformer further refines saliency information and increases the focus on appearance information. The efficacy and robustness of the proposed approach have been thoroughly assessed through experiments on three widely-used UAV tracking benchmarks and real-world scenarios, with results demonstrating its superiority. The source code and demo videos are available at https://github.com/vision4robotics/SGDViT."
                    },
                    {
                        "title": "Continuity-Aware Latent Interframe Information Mining for Reliable UAV Tracking",
                        "abstract": "Unmanned aerial vehicle (UAV) tracking is crucial for autonomous navigation and has broad applications in robotic automation fields. However, reliable UAV tracking remains a challenging task due to various difficulties like frequent occlusion and aspect ratio change. Additionally, most of the existing work mainly focuses on explicit information to improve tracking performance, ignoring potential interframe connections. To address the above issues, this work proposes a novel framework with continuity-aware latent interframe information mining for reliable UAV tracking, i.e., ClimRT. Specifically, a new efficient continuity-aware latent interframe information mining network (ClimNet) is proposed for UAV tracking, which can generate highly-effective latent frame between two adjacent frames. Besides, a novel location-continuity Transformer (LCT) is designed to fully explore continuity-aware spatial-temporal information, thereby markedly enhancing UAV tracking. Extensive qualitative and quantitative experiments on three authoritative aerial benchmarks strongly validate the robustness and reliability of ClimRT in UAV tracking performance. Furthermore, real-world tests on the aerial platform validate its practicability and effectiveness. The code and demo materials are released at https://github.com/vision4robotics/ClimRT."
                    },
                    {
                        "title": "HighlightNet: Highlighting Low-Light Potential Features for Real-Time UAV Tracking",
                        "abstract": "Low-light environments have posed a formidable challenge for robust unmanned aerial vehicle (UAV) tracking even with state-of-the-art (SOTA) trackers since the potential image features are hard to extract under adverse light conditions. Besides, due to the low visibility, accurate online selection of the object also becomes extremely difficult for human monitors to initialize UAV tracking in ground control stations. To solve these problems, this work proposes a novel enhancer, i.e., HighlightNet, to light up potential objects for both human operators and UAV trackers. By employing Transformer, HighlightNet can adjust enhancement parameters according to global features and is thus adaptive for the illumination variation. Pixel-level range mask is introduced to make HighlightNet more focused on the enhancement of the tracking object and regions without light sources. Furthermore, a soft truncation mechanism is built to prevent background noise from being mistaken for crucial features. Evaluations on image enhancement benchmarks demonstrate HighlightNet has advantages in facilitating human perception. Experiments on the public UAVDark135 benchmark show that HightlightNet is more suitable for UAV tracking tasks than other SOTA low-light enhancers. In addition, real-world tests on a typical UAV platform verify HightlightNet's practicability and efficiency in nighttime aerial tracking-related applications. The code and demo videos are available at https://github.com/vision4robotics/HighlightNet."
                    },
                    {
                        "title": "LLaRA: Large Language-Recommendation Assistant",
                        "abstract": "Sequential recommendation aims to predict users' next interaction with items based on their past engagement sequence. Recently, the advent of Large Language Models (LLMs) has sparked interest in leveraging them for sequential recommendation, viewing it as language modeling. Previous studies represent items within LLMs' input prompts as either ID indices or textual metadata. However, these approaches often fail to either encapsulate comprehensive world knowledge or exhibit sufficient behavioral understanding. To combine the complementary strengths of conventional recommenders in capturing behavioral patterns of users and LLMs in encoding world knowledge about items, we introduce Large Language-Recommendation Assistant (LLaRA). Specifically, it uses a novel hybrid prompting method that integrates ID-based item embeddings learned by traditional recommendation models with textual item features. Treating the \"sequential behaviors of users\" as a distinct modality beyond texts, we employ a projector to align the traditional recommender's ID embeddings with the LLM's input space. Moreover, rather than directly exposing the hybrid prompt to LLMs, a curriculum learning strategy is adopted to gradually ramp up training complexity. Initially, we warm up the LLM using text-only prompts, which better suit its inherent language modeling ability. Subsequently, we progressively transition to the hybrid prompts, training the model to seamlessly incorporate the behavioral knowledge from the traditional sequential recommender into the LLM. Empirical results validate the effectiveness of our proposed framework. Codes are available at https://github.com/ljy0ustc/LLaRA."
                    },
                    {
                        "title": "Towards 3D Molecule-Text Interpretation in Language Models",
                        "abstract": "Language Models (LMs) have greatly influenced diverse domains. However, their inherent limitation in comprehending 3D molecular structures has considerably constrained their potential in the biomolecular domain. To bridge this gap, we focus on 3D molecule-text interpretation, and propose 3D-MoLM: 3D-Molecular Language Modeling. Specifically, 3D-MoLM enables an LM to interpret and analyze 3D molecules by equipping the LM with a 3D molecular encoder. This integration is achieved by a 3D molecule-text projector, bridging the 3D molecular encoder's representation space and the LM's input space. Moreover, to enhance 3D-MoLM's ability of cross-modal molecular understanding and instruction following, we meticulously curated a 3D molecule-centric instruction tuning dataset -- 3D-MoIT. Through 3D molecule-text alignment and 3D molecule-centric instruction tuning, 3D-MoLM establishes an integration of 3D molecular encoder and LM. It significantly surpasses existing baselines on downstream tasks, including molecule-text retrieval, molecule captioning, and more challenging open-text molecular QA tasks, especially focusing on 3D-dependent properties. We release our codes and datasets at https://github.com/lsh0520/3D-MoLM."
                    },
                    {
                        "title": "SciLitLLM: How to Adapt LLMs for Scientific Literature Understanding",
                        "abstract": "Scientific literature understanding is crucial for extracting targeted information and garnering insights, thereby significantly advancing scientific discovery. Despite the remarkable success of Large Language Models (LLMs), they face challenges in scientific literature understanding, primarily due to (1) a lack of scientific knowledge and (2) unfamiliarity with specialized scientific tasks.   To develop an LLM specialized in scientific literature understanding, we propose a hybrid strategy that integrates continual pre-training (CPT) and supervised fine-tuning (SFT), to simultaneously infuse scientific domain knowledge and enhance instruction-following capabilities for domain-specific tasks.cIn this process, we identify two key challenges: (1) constructing high-quality CPT corpora, and (2) generating diverse SFT instructions. We address these challenges through a meticulous pipeline, including PDF text extraction, parsing content error correction, quality filtering, and synthetic instruction creation. Applying this strategy, we present a suite of LLMs: SciLitLLM, specialized in scientific literature understanding. These models demonstrate promising performance on scientific literature understanding benchmarks.   Our contributions are threefold: (1) We present an effective framework that integrates CPT and SFT to adapt LLMs to scientific literature understanding, which can also be easily adapted to other domains. (2) We propose an LLM-based synthesis method to generate diverse and high-quality scientific instructions, resulting in a new instruction set -- SciLitIns -- for supervised fine-tuning in less-represented scientific domains. (3) SciLitLLM achieves promising performance improvements on scientific literature understanding benchmarks."
                    }
                ]
            },
            "55410602-3bb9-48da-8278-5dcd93f974e9": {
                "pk": "55410602-3bb9-48da-8278-5dcd93f974e9",
                "name": "Xinhao Liu",
                "collaborators": [
                    "Yiming Li",
                    "Chen Feng",
                    "Yingzhao Jiang",
                    "Zetao Lin",
                    "Hongyuan Diao",
                    "Fuzhong Nian",
                    "Xuelong Yu",
                    "Xirui Liu",
                    "Chao Chen",
                    "Li Ding"
                ],
                "domain": [
                    "Model Inversion Attack",
                    "LiDAR Mapping",
                    "Autonomous Driving",
                    "Sentiment Analysis"
                ],
                "publications": [
                    {
                        "title": "Breaking the Black-Box: Confidence-Guided Model Inversion Attack for Distribution Shift",
                        "abstract": "Model inversion attacks (MIAs) seek to infer the private training data of a target classifier by generating synthetic images that reflect the characteristics of the target class through querying the model. However, prior studies have relied on full access to the target model, which is not practical in real-world scenarios. Additionally, existing black-box MIAs assume that the image prior and target model follow the same distribution. However, when confronted with diverse data distribution settings, these methods may result in suboptimal performance in conducting attacks. To address these limitations, this paper proposes a \\textbf{C}onfidence-\\textbf{G}uided \\textbf{M}odel \\textbf{I}nversion attack method called CG-MI, which utilizes the latent space of a pre-trained publicly available generative adversarial network (GAN) as prior information and gradient-free optimizer, enabling high-resolution MIAs across different data distributions in a black-box setting. Our experiments demonstrate that our method significantly \\textbf{outperforms the SOTA black-box MIA by more than 49\\% for Celeba and 58\\% for Facescrub in different distribution settings}. Furthermore, our method exhibits the ability to generate high-quality images \\textbf{comparable to those produced by white-box attacks}. Our method provides a practical and effective solution for black-box model inversion attacks."
                    },
                    {
                        "title": "The Emotion coding and Propagation based on improved Genetic algorithm",
                        "abstract": "Computational communication research on information has been prevalent in recent years, as people are progressively inquisitive in social behavior and public opinion. Nevertheless, it is of great significance to analyze the direction of predominant sentiment from the sentiment communication perspective. In this paper, the information emotion propagation model is established by introducing revamp genetic algorithms into information emotion. In the process of information dissemination, both the information emotions and the network emotions are dynamic. For this model, the information emotions and the network nodes emotions are quantified as binary codes. The convergence effects, crossover and mutation algorithms are introduced. These factors all act on the transmission process via dynamic propagation rate, and the improved genetic algorithm also acts on the emotion transmission. In particular, the latter two algorithms are different from the existing biological domain. Based on the existing research results in other manuscripts, we perform simulation described above on the hybrid network. The simulation results demonstrate that the trend approximate to the actual data. As a result, our work can prove that our proposed model is essentially consistent with the actual emotion transmission phenomenon."
                    },
                    {
                        "title": "DeepMapping2: Self-Supervised Large-Scale LiDAR Map Optimization",
                        "abstract": "LiDAR mapping is important yet challenging in self-driving and mobile robotics. To tackle such a global point cloud registration problem, DeepMapping converts the complex map estimation into a self-supervised training of simple deep networks. Despite its broad convergence range on small datasets, DeepMapping still cannot produce satisfactory results on large-scale datasets with thousands of frames. This is due to the lack of loop closures and exact cross-frame point correspondences, and the slow convergence of its global localization network. We propose DeepMapping2 by adding two novel techniques to address these issues: (1) organization of training batch based on map topology from loop closing, and (2) self-supervised local-to-global point consistency loss leveraging pairwise registration. Our experiments and ablation studies on public datasets (KITTI, NCLT, and Nebula) demonstrate the effectiveness of our method."
                    },
                    {
                        "title": "LiDAR-based 4D Occupancy Completion and Forecasting",
                        "abstract": "Scene completion and forecasting are two popular perception problems in research for mobile agents like autonomous vehicles. Existing approaches treat the two problems in isolation, resulting in a separate perception of the two aspects. In this paper, we introduce a novel LiDAR perception task of Occupancy Completion and Forecasting (OCF) in the context of autonomous driving to unify these aspects into a cohesive framework. This task requires new algorithms to address three challenges altogether: (1) sparse-to-dense reconstruction, (2) partial-to-complete hallucination, and (3) 3D-to-4D prediction. To enable supervision and evaluation, we curate a large-scale dataset termed OCFBench from public autonomous driving datasets. We analyze the performance of closely related existing baseline models and our own ones on our dataset. We envision that this research will inspire and call for further investigation in this evolving and crucial area of 4D perception. Our code for data curation and baseline implementation is available at https://github.com/ai4ce/Occ4cast."
                    }
                ]
            },
            "2ba2d096-5b5c-4de3-a7d8-bdc9f52616cd": {
                "pk": "2ba2d096-5b5c-4de3-a7d8-bdc9f52616cd",
                "name": "Haorui Song",
                "collaborators": [
                    "Qi Chen",
                    "Xiaoxi Chen",
                    "Zhiwei Xiong",
                    "Alan Yuille",
                    "Chen Wei",
                    "Zongwei Zhou",
                    "Yong Du",
                    "Tianyi Xiang",
                    "Junyu Dong",
                    "Jing Qin"
                ],
                "domain": [
                    "Medical Imaging",
                    "Generative Adversarial Networks",
                    "Tumor Detection",
                    "Image Synthesis"
                ],
                "publications": [
                    {
                        "title": "Towards Generalizable Tumor Synthesis",
                        "abstract": "Tumor synthesis enables the creation of artificial tumors in medical images, facilitating the training of AI models for tumor detection and segmentation. However, success in tumor synthesis hinges on creating visually realistic tumors that are generalizable across multiple organs and, furthermore, the resulting AI models being capable of detecting real tumors in images sourced from different domains (e.g., hospitals). This paper made a progressive stride toward generalizable tumor synthesis by leveraging a critical observation: early-stage tumors (< 2cm) tend to have similar imaging characteristics in computed tomography (CT), whether they originate in the liver, pancreas, or kidneys. We have ascertained that generative AI models, e.g., Diffusion Models, can create realistic tumors generalized to a range of organs even when trained on a limited number of tumor examples from only one organ. Moreover, we have shown that AI models trained on these synthetic tumors can be generalized to detect and segment real tumors from CT volumes, encompassing a broad spectrum of patient demographics, imaging protocols, and healthcare facilities."
                    },
                    {
                        "title": "Editing Out-of-domain GAN Inversion via Differential Activations",
                        "abstract": "Despite the demonstrated editing capacity in the latent space of a pretrained GAN model, inverting real-world images is stuck in a dilemma that the reconstruction cannot be faithful to the original input. The main reason for this is that the distributions between training and real-world data are misaligned, and because of that, it is unstable of GAN inversion for real image editing. In this paper, we propose a novel GAN prior based editing framework to tackle the out-of-domain inversion problem with a composition-decomposition paradigm. In particular, during the phase of composition, we introduce a differential activation module for detecting semantic changes from a global perspective, \\ie, the relative gap between the features of edited and unedited images. With the aid of the generated Diff-CAM mask, a coarse reconstruction can intuitively be composited by the paired original and edited images. In this way, the attribute-irrelevant regions can be survived in almost whole, while the quality of such an intermediate result is still limited by an unavoidable ghosting effect. Consequently, in the decomposition phase, we further present a GAN prior based deghosting network for separating the final fine edited image from the coarse reconstruction. Extensive experiments exhibit superiorities over the state-of-the-art methods, in terms of qualitative and quantitative evaluations. The robustness and flexibility of our method is also validated on both scenarios of single attribute and multi-attribute manipulations."
                    }
                ]
            },
            "e6025334-7995-408e-aa5f-9330b249e972": {
                "pk": "e6025334-7995-408e-aa5f-9330b249e972",
                "name": "Xinran Tang",
                "collaborators": [
                    "Van Anh Le",
                    "Varshini Reddy",
                    "Zixi Chen",
                    "Mengyuan Li",
                    "Anthony Ortiz",
                    "Simone Fobi Nsutezo",
                    "Caleb Robinson"
                ],
                "domain": [
                    "Synthetic Image Generation",
                    "Satellite Imagery",
                    "Data Augmentation",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Mask Conditional Synthetic Satellite Imagery",
                        "abstract": "In this paper we propose a mask-conditional synthetic image generation model for creating synthetic satellite imagery datasets. Given a dataset of real high-resolution images and accompanying land cover masks, we show that it is possible to train an upstream conditional synthetic imagery generator, use that generator to create synthetic imagery with the land cover masks, then train a downstream model on the synthetic imagery and land cover masks that achieves similar test performance to a model that was trained with the real imagery. Further, we find that incorporating a mixture of real and synthetic imagery acts as a data augmentation method, producing better models than using only real imagery (0.5834 vs. 0.5235 mIoU). Finally, we find that encouraging diversity of outputs in the upstream model is a necessary component for improved downstream task performance. We have released code for reproducing our work on GitHub, see https://github.com/ms-synthetic-satellite-image/synthetic-satellite-imagery ."
                    }
                ]
            },
            "7de505d9-4bbf-4139-b5da-6c067a8a8c04": {
                "pk": "7de505d9-4bbf-4139-b5da-6c067a8a8c04",
                "name": "Chen Feng",
                "collaborators": [
                    "Andrew L. Lu",
                    "Yihsu Chen"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Energy Systems",
                    "Multi-Agent Systems",
                    "Peer-to-Peer Trading"
                ],
                "publications": [
                    {
                        "title": "Decentralized Voltage Control with Peer-to-peer Energy Trading in a Distribution Network",
                        "abstract": "Utilizing distributed renewable and energy storage resources via peer-to-peer (P2P) energy trading has long been touted as a solution to improve energy system's resilience and sustainability. Consumers and prosumers (those who have energy generation resources), however, do not have expertise to engage in repeated P2P trading, and the zero-marginal costs of renewables present challenges in determining fair market prices. To address these issues, we propose a multi-agent reinforcement learning (MARL) framework to help automate consumers' bidding and management of their solar PV and energy storage resources, under a specific P2P clearing mechanism that utilizes the so-called supply-demand ratio. In addition, we show how the MARL framework can integrate physical network constraints to realize decentralized voltage control, hence ensuring physical feasibility of the P2P energy trading and paving ways for real-world implementations."
                    }
                ]
            }
        }
    },
    "2401.14657": {
        "paper_data": {
            "title": "Validating Climate Models with Spherical Convolutional Wasserstein Distance",
            "url": "http://arxiv.org/abs/2401.14657v1",
            "arxiv_id": "2401.14657",
            "authors": [
                "Robert C. Garrett",
                "Trevor Harris",
                "Bo Li",
                "Zhuo Wang"
            ],
            "abstract": "The validation of global climate models is crucial to ensure the accuracy and efficacy of model output. We introduce the spherical convolutional Wasserstein distance to more comprehensively measure differences between climate models and reanalysis data. This new similarity measure accounts for spatial variability using convolutional projections and quantifies local differences in the distribution of climate variables. We apply this method to evaluate the historical model outputs of the Coupled Model Intercomparison Project (CMIP) members by comparing them to observational and reanalysis data products. Additionally, we investigate the progression from CMIP phase 5 to phase 6 and find modest improvements in the phase 6 models regarding their ability to produce realistic climatologies.",
            "introduction": "   1 Introduction  Climate Model Validation  General Circulation Models, or climate models, are mathematical representations of the climate system that describe interactions between matter and energy through the ocean, atmosphere, and land (Washington & Parkinson, 2005). Climate models are the primary tool for investigating the response of the climate system to changes in forcing, such as increases in CO2subscriptCO2\\text{CO}_{2}CO start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT, and projecting future climate states (Flato et\u00a0al., 2014). To assess the plausibility of climate models, climate scientists compare output from model simulations against observational data (Rood, 2019). This comparison is the focus of climate model validation techniques for ensuring that climate models capture the dynamics of the climate system (Roca et\u00a0al., 2021).   The Coupled Model Intercomparison Project (CMIP) was initiated in 1995 as a comprehensive and systematic program for assessing climate models against each other and observational data (Eyring et\u00a0al., 2016). Each model in CMIP participates in a wide variety of experiments such as performing a historical simulation, a pre-industrial control simulation, and various simulations representing different scenarios for CO22{}_{2}start_FLOATSUBSCRIPT 2 end_FLOATSUBSCRIPT emissions (Eyring et\u00a0al., 2016). Because historical simulations coincide with observational measurements, we can compare each model\u2019s synthetic climate distribution to the distribution of observational or quasi-observational data products (Ra\u00e4isa\u00e4nen, 2007), to assess their reconstructive skill. For complete spatial coverage we compare against reanalysis data, a blend of observations and short-range weather forecasts through data assimilation (Bengtsson et\u00a0al., 2004). This has become one popular climate model validation method (Flato et\u00a0al., 2014).    Previous approaches  Many statistical and machine learning-based methods have been applied to assess climate model output against reanalysis fields. The most common approach is to compute the L\u20622\ud835\udc3f2L2italic_L 2 distance, or root mean square error (RMSE), between different features of the climate model output and the reanalysis field (Li et\u00a0al., 2021; Zamani et\u00a0al., 2020; Karim et\u00a0al., 2020; Ayugi et\u00a0al., 2021). RMSE provides a direct measure of the differences between two climate fields but does not take into account randomness of data, so we should use it with caution in evaluating climate models. Another approach, which is invariant to bias, is to compute measures of correlation between climate model output and reanalysis fields (Zhao et\u00a0al., 2021; Zamani et\u00a0al., 2020; Karim et\u00a0al., 2020; Ayugi et\u00a0al., 2021).   More comprehensive approaches employ techniques for random processes to compare two spatial fields. For example, (Shen et\u00a0al., 2002) and (Cressie et\u00a0al., 2008) use the wavelet decomposition to compare the spatial-frequency content of two fields. (Hering & Genton, 2011) measures the loss differential between models of spatial processes, (Lund & Li, 2009) and (Li & Smerdon, 2012) compare the first and second moments of two random processes, and (Yun et\u00a0al., 2022) identifies local differences in the mean and dependency structure between two spatiotemporal climate fields. Functional data analysis techniques have been introduced to compare spatial and spatiotemporal random fields, considering the random fields as continuous functions. Many of these approaches compare the underlying mean functions from two sets of functional data (Zhang & Chen, 2007; Horv\u00e1th et\u00a0al., 2013; Staicu et\u00a0al., 2014). Other approaches include the second-order structure in the comparison (Zhang & Shao, 2015; Li et\u00a0al., 2016), or compare the distributions of two spatial random processes (Harris et\u00a0al., 2021).   Since climate",
            "references": [
                {
                    "title": "Spherical Sliced-Wasserstein",
                    "abstract": "Many variants of the Wasserstein distance have been introduced to reduce its original computational burden. In particular the Sliced-Wasserstein distance (SW), which leverages one-dimensional projections for which a closed-form solution of the Wasserstein distance is available, has received a lot of interest. Yet, it is restricted to data living in Euclidean spaces, while the Wasserstein distance has been studied and used recently on manifolds. We focus more specifically on the sphere, for which we define a novel SW discrepancy, which we call spherical Sliced-Wasserstein, making a first step towards defining SW discrepancies on manifolds. Our construction is notably based on closed-form solutions of the Wasserstein distance on the circle, together with a new spherical Radon transform. Along with efficient algorithms and the corresponding implementations, we illustrate its properties in several machine learning use cases where spherical representations of data are at stake: sampling on the sphere, density estimation on real earth data or hyperspherical auto-encoders."
                },
                {
                    "title": "Revisiting Sliced Wasserstein on Images: From Vectorization to Convolution",
                    "abstract": "The conventional sliced Wasserstein is defined between two probability measures that have realizations as vectors. When comparing two probability measures over images, practitioners first need to vectorize images and then project them to one-dimensional space by using matrix multiplication between the sample matrix and the projection matrix. After that, the sliced Wasserstein is evaluated by averaging the two corresponding one-dimensional projected probability measures. However, this approach has two limitations. The first limitation is that the spatial structure of images is not captured efficiently by the vectorization step; therefore, the later slicing process becomes harder to gather the discrepancy information. The second limitation is memory inefficiency since each slicing direction is a vector that has the same dimension as the images. To address these limitations, we propose novel slicing methods for sliced Wasserstein between probability measures over images that are based on the convolution operators. We derive convolution sliced Wasserstein (CSW) and its variants via incorporating stride, dilation, and non-linear activation function into the convolution operators. We investigate the metricity of CSW as well as its sample complexity, its computational complexity, and its connection to conventional sliced Wasserstein distances. Finally, we demonstrate the favorable performance of CSW over the conventional sliced Wasserstein in comparing probability measures over images and in training deep generative modeling on images."
                },
                {
                    "title": "Comparative Assessment and Future Prediction Using CMIP6 and CMIP5 for Annual Precipitation and Extreme Precipitation Simulation",
                    "abstract": "This study assesses the improvement of the latest Coupled Model Intercomparison Project Phase 6 (CMIP6) over Coupled Model Intercomparison Project Phase 5 (CMIP5) for precipitation simulation. Precipitation simulations under different future climate scenarios are also compared in this work. The results show that: 1) CMIP6 has no overall advantage over CMIP5 in simulating total precipitation (PRCPTOT) and maximum consecutive dry days (CDD). The performance of CMIP6 increases or decreases regionally in PRCPTOT and consecutive dry days. But it is slightly worse than CMIP5 in simulating very wet days (R95pTOT). 2) Comparing the trend test results of CMIP5 and CMIP6 in the future, there are more areas with significant trend based on Mann\u2013Kendall test in CMIP6 compared with that of CMIP5. The differences in PRCPTOT are mainly found in Amazon Basin and Western Africa. The differences between the R95pTOT trends mainly noticeable in South America and Western Africa, and the differences in CDD are mainly reflected in Central Asia, Sahara Desert and central South America. 3) In Southern South America and Western North America, the PRCPTOT changing rate of CMIP6 in the future under various scenarios is always greater than that of CMIP5; in Alaska, Western Africa, Southern Africa, the PRCPTOT changing rate of CMIP6 in the future under various scenarios is always less than that of CMIP5. In Southern South America, the R95pTOT changing rate of CMIP6 in the future under various scenarios is always greater than that of CMIP5; in Alaska, East Asia, North Asia, the R95pTOT changing rate of CMIP6 in the future under various scenarios is always less than that of CMIP5. In almost half of the regions, the CDD changing rate of CMIP6 is less than that of CMIP5 under all scenarios, namely Australia, Amazon Basin, Southern South America, Central America, Western North America, Central North America, Eastern North America, Central Asia, Tibet."
                },
                {
                    "title": "Evaluation of the Performance of CMIP5 Models to Simulate Land Surface Air Temperature Based on Long-Range Correlation",
                    "abstract": "The daily average land surface air temperature (SAT) simulated by 8 CMIP5 models historical experiments and that from NCEP data during 1960\u20132005, are used to evaluate the performance of the CMIP5 model based on detrended fluctuation analysis (DFA) method. The DFA results of NCEP data show that SAT in most regions of the world exhibit long-range correlation. The scaling exponents of NCEP SAT show the zonal distribution characteristics of larg in tropics while small in medium and high latitudes. The distribution characteristics of the zonal average scaling exponents of CMCC-CMS, GFDL-ESM2G, IPSL-CM5A-MR are similar to that of NCEP data. From the DFA errors of model-simulated SAT, the performance of IPSL-CM5A-MR is the best among the 8 models throughout the year, the performance of FGOALS-g2 is good in spring and summer, GFDL-ESM2G is the best in autumn, CNRM-CM5 and CMCC-CMS is good in winter. The scaling exponents of model-simulated SAT are closer to that of NCEP data in most areas of the mid-high latitude on the northern hemisphere. However, simulations of SAT in East Asia and Central North American are generally less effective. In spring, most models have better performance in Siberian (SIB), Central Asia (CAS) and Tibetan (TIB). SAT in Northern Europe area are well simulated by most models in summer. In autumn, areas with better performance of most models are Mediterranean, SIB and TIB regions. In winter, SAT in Greenland, SIB and TIB areas are well simulated by most models. Generally speaking, the performance of CMIP5 models for SAT on global continents varies in different seasons and different regions."
                },
                {
                    "title": "Comparison of CMIP6 and CMIP5 models in simulating mean and extreme precipitation over East Africa",
                    "abstract": "This study examines the improvement in Coupled Model Intercomparison Project Phase Six (CMIP6) models against the predecessor CMIP5 in simulating mean and extreme precipitation over the East Africa region. The study compares the climatology of the precipitation indices simulated by the CMIP models with the CHIRPS data set using robust statistical techniques for 1981\u20132005. The results display the varying performance of the general circulation models (GCMs) in the simulation of annual and seasonal precipitation climatology over the study domain. CMIP6 multi\u2010model ensemble mean (hereafter MME) shows improved performance in the local annual mean cycle simulation with a better representation of the rainfall within the two peaks, especially the MAM rainfall relative to their predecessor. Moreover, simulation of extreme indices is well captured in CMIP6 models relative to CMIP5. The CMIP6\u2010MME performed better than the CMIP5\u2010MME with lesser biases in simulating Simple Daily Intensity Index (SDII), consecutive dry days (CDD), and very heavy precipitation days >20\u2009mm (R20mm) over East Africa. Remarkably, most CMIP6 models are unable to simulate extremely wet days (R95p). Some CMIP6 models (e.g., NorESM2\u2010MM and CNRM\u2010CM6\u20101) depict robust performance in reproducing the observed indices across all analyses. OND season shows wet biases for some indices (i.e., R95p, PRCPTOT), except for SDII, CDD, and R20mm in CMIP6 models. Consistent with other studies, the mean ensemble performance for both CMIP5/6 shows better performance as compared with individual models due to the cancellation of some systematic errors in the individual models. Generally, CMIP6 depicts improved performance in the simulation of MAM rainfall compared with CMIP5 models. However, the new model generation is still marred by uncertainty, thereby depicting unsatisfactory performance over the East Africa domain. This calls for further investigation into the sources of persistent systematic biases and a methodology for identifying individual models with robust features that can accurately simulate observed patterns for future usage."
                },
                {
                    "title": "Evaluating the Performance of Climate Models Based on Wasserstein Distance",
                    "abstract": "We propose a methodology for intercomparing climate models and evaluating their performance against benchmarks based on the use of the Wasserstein distance (WD). This distance provides a rigorous way to measure quantitatively the difference between two probability distributions. The proposed approach is flexible and can be applied in any number of dimensions; it allows one to rank climate models taking into account all the moments of the distributions. By selecting the combination of climatic variables and the regions of interest, it is possible to highlight specific model deficiencies. The WD enables a comprehensive evaluation of the skill of a climate model. We apply this approach to a selected number of physical fields, ranking the models in terms of their performance in simulating them and pinpointing their weaknesses in the simulation of some of the selected physical fields in specific areas of the Earth."
                },
                {
                    "title": "Evaluation of Historical CMIP6 Model Simulations of Seasonal Mean Temperature over Pakistan during 1970\u20132014",
                    "abstract": "This work employed recent model outputs from coupled model intercomparison project phase six to simulate surface mean temperature during the June\u2013July\u2013August (JJA) and December\u2013January\u2013February (DJF) seasons for 1970\u20132014 over Pakistan. The climatic research unit (CRU TS4.03) dataset was utilized as benchmark data to analyze models\u2019 performance. The JJA season exhibited the highest mean temperature, whilst DJF displayed the lowest mean temperature in the whole study period. The JJA monthly empirical cumulative distribution frequency (ECDF) range (26 to 28 \u00b0C) was less than that of DJF (7 to 10 \u00b0C) since JJA matched closely to CRU. The JJA and DJF seasons are warming, with higher warming trends in winters than in summers. On temporal scale, models performed better in JJA with overall low bias, low RMSE (root mean square error), and higher positive CC (correlation coefficient) values. DJF performance was undermined with higher bias and RMSE with weak positive correlation estimates. Overall, CanESM5, CESM2, CESM2-WACCM, GFDL-CM4, HadGEM-GC31-LL, MPI-ESM1-2-LR, MPI-ESM1-2-HR, and MRI-ESM-0 performed better for JJA and DJF."
                },
                {
                    "title": "GISS\u2010E2.1: Configurations and Climatology",
                    "abstract": "This paper describes the GISS\u2010E2.1 contribution to the Coupled Model Intercomparison Project, Phase 6 (CMIP6). This model version differs from the predecessor model (GISS\u2010E2) chiefly due to parameterization improvements to the atmospheric and ocean model components, while keeping atmospheric resolution the same. Model skill when compared to modern era climatologies is significantly higher than in previous versions. Additionally, updates in forcings have a material impact on the results. In particular, there have been specific improvements in representations of modes of variability (such as the Madden\u2010Julian Oscillation and other modes in the Pacific) and significant improvements in the simulation of the climate of the Southern Oceans, including sea ice. The effective climate sensitivity to 2 \u00d7 CO2 is slightly higher than previously at 2.7\u20133.1\u00b0C (depending on version) and is a result of lower CO2 radiative forcing and stronger positive feedbacks."
                },
                {
                    "title": "Detection of Local Differences in Spatial Characteristics Between Two Spatiotemporal Random Fields",
                    "abstract": "Abstract Comparing the spatial characteristics of spatiotemporal random fields is often at demand. However, the comparison can be challenging due to the high-dimensional feature and dependency in the data. We develop a new multiple testing approach to detect local differences in the spatial characteristics of two spatiotemporal random fields by taking the spatial information into account. Our method adopts a two-component mixture model for location wise p-values and then derives a new false discovery rate (FDR) control, called mirror procedure, to determine the optimal rejection region. This procedure is robust to model misspecification and allows for weak dependency among hypotheses. To integrate the spatial heterogeneity, we model the mixture probability as well as study the benefit if any of allowing the alternative distribution to be spatially varying. An EM-algorithm is developed to estimate the mixture model and implement the FDR procedure. We study the FDR control and the power of our new approach both theoretically and numerically, and apply the approach to compare the mean and teleconnection pattern between two synthetic climate fields. Supplementary materials for this article are available online."
                },
                {
                    "title": "The ERA5 global reanalysis",
                    "abstract": "Within the Copernicus Climate Change Service (C3S), ECMWF is producing the ERA5 reanalysis which, once completed, will embody a detailed record of the global atmosphere, land surface and ocean waves from 1950 onwards. This new reanalysis replaces the ERA\u2010Interim reanalysis (spanning 1979 onwards) which was started in 2006. ERA5 is based on the Integrated Forecasting System (IFS) Cy41r2 which was operational in 2016. ERA5 thus benefits from a decade of developments in model physics, core dynamics and data assimilation. In addition to a significantly enhanced horizontal resolution of 31\u2009km, compared to 80\u2009km for ERA\u2010Interim, ERA5 has hourly output throughout, and an uncertainty estimate from an ensemble (3\u2010hourly at half the horizontal resolution). This paper describes the general set\u2010up of ERA5, as well as a basic evaluation of characteristics and performance, with a focus on the dataset from 1979 onwards which is currently publicly available. Re\u2010forecasts from ERA5 analyses show a gain of up to one day in skill with respect to ERA\u2010Interim. Comparison with radiosonde and PILOT data prior to assimilation shows an improved fit for temperature, wind and humidity in the troposphere, but not the stratosphere. A comparison with independent buoy data shows a much improved fit for ocean wave height. The uncertainty estimate reflects the evolution of the observing systems used in ERA5. The enhanced temporal and spatial resolution allows for a detailed evolution of weather systems. For precipitation, global\u2010mean correlation with monthly\u2010mean GPCP data is increased from 67% to 77%. In general, low\u2010frequency variability is found to be well represented and from 10\u2009hPa downwards general patterns of anomalies in temperature match those from the ERA\u2010Interim, MERRA\u20102 and JRA\u201055 reanalyses."
                },
                {
                    "title": "Evaluating Proxy Influence in Assimilated Paleoclimate Reconstructions\u2014Testing the Exchangeability of Two Ensembles of Spatial Processes",
                    "abstract": "Abstract Climate field reconstructions (CFRs) attempt to estimate spatiotemporal fields of climate variables in the past using climate proxies such as tree rings, ice cores, and corals. Data assimilation (DA) methods are a recent and promising new means of deriving CFRs that optimally fuse climate proxies with climate model output. Despite the growing application of DA-based CFRs, little is understood about how much the assimilated proxies change the statistical properties of the climate model data. To address this question, we propose a robust and computationally efficient method, based on functional data depth, to evaluate differences in the distributions of two spatiotemporal processes. We apply our test to study global and regional proxy influence in DA-based CFRs by comparing the background and analysis states, which are treated as two samples of spatiotemporal fields. We find that the analysis states are significantly altered from the climate-model-based background states due to the assimilation of proxies. Moreover, the difference between the analysis and background states increases with the number of proxies, even in regions far beyond proxy collection sites. Our approach allows us to characterize the added value of proxies, indicating where and when the analysis states are distinct from the background states. Supplementary materials for this article are available online."
                },
                {
                    "title": "Generalized Sliced Wasserstein Distances",
                    "abstract": "The Wasserstein distance and its variations, e.g., the sliced-Wasserstein (SW) distance, have recently drawn attention from the machine learning community. The SW distance, specifically, was shown to have similar properties to the Wasserstein distance, while being much simpler to compute, and is therefore used in various applications including generative modeling and general supervised/unsupervised learning. In this paper, we first clarify the mathematical connection between the SW distance and the Radon transform. We then utilize the generalized Radon transform to define a new family of distances for probability measures, which we call generalized sliced-Wasserstein (GSW) distances. We also show that, similar to the SW distance, the GSW distance can be extended to a maximum GSW (max-GSW) distance. We then provide the conditions under which GSW and max-GSW distances are indeed distances. Finally, we compare the numerical performance of the proposed distances on several generative modeling tasks, including SW flows and SW auto-encoders."
                },
                {
                    "title": "Permutation tests for equality of distributions of functional data",
                    "abstract": "Economic data are often generated by stochastic processes that take place in continuous time, though observations may occur only at discrete times. For example, electricity and gas consumption take place in continuous time. Data generated by a continuous time stochastic process are called functional data. This paper is concerned with comparing two or more stochastic processes that generate functional data. The data may be produced by a randomized experiment in which there are multiple treatments. The paper presents a test of the hypothesis that the same stochastic process generates all the functional data. In contrast to existing methods, the test described here applies to both functional data and multiple treatments. The test is presented as a permutation test, which ensures that in a finite sample, the true and nominal probabilities of rejecting a correct null hypothesis are equal. The paper also presents the asymptotic distribution of the test statistic under alternative hypotheses. The results of Monte Carlo experiments and an application to an experiment on billing and pricing of natural gas illustrate the usefulness of the test."
                },
                {
                    "title": "On a General Definition of Depth for Functional Data",
                    "abstract": "In this paper, we provide an elaboration on the desirable properties \nof statistical depths for functional data. Although a formal definition has \nbeen put forward in the literature, there are still several unclarities to be tackled, \nand further insights to be gained. Herein, a few interesting connections \nbetween the wanted properties are found. In particular, it is demonstrated that \nthe conditions needed for some desirable properties to hold are extremely demanding, \nand virtually impossible to be met for common depths. We establish \nadaptations of these properties which prove to be still sensible, and more \neasily met by common functional depths."
                },
                {
                    "title": "Comparison between spatio\u2010temporal random processes and application to climate model data",
                    "abstract": "Comparing two spatio\u2010temporal processes are often a desirable exercise. For example, assessments of the difference between various climate models may involve the comparisons of the synthetic climate random fields generated as simulations from each model. We develop rigorous methods to compare two spatio\u2010temporal random processes both in terms of moments and in terms of temporal trend, using the functional data analysis approach. A highlight of our method is that we can compare the trend surfaces between two random processes, which are motivated by evaluating the skill of synthetic climate from climate models in terms of capturing the pronounced upward trend of real\u2010observational data. We perform simulations to evaluate our methods and then apply the methods to compare different climate models as well as to evaluate the synthetic temperature fields from model simulations, with respect to observed temperature fields. Copyright \u00a9 2016 John Wiley & Sons, Ltd."
                },
                {
                    "title": "Functional Data Analysis",
                    "abstract": "With the advance of modern technology, more and more data are being recorded continuously during a time interval or intermittently at several discrete time points. These are both examples of functional data, which has become a commonly encountered type of data. Functional data analysis (FDA) encompasses the statistical methodology for such data. Broadly interpreted, FDA deals with the analysis and theory of data that are in the form of functions. This paper provides an overview of FDA, starting with simple statistical notions such as mean and covariance functions, then covering some core techniques, the most popular of which is functional principal component analysis (FPCA). FPCA is an important dimension reduction tool, and in sparse data situations it can be used to impute functional data that are sparsely observed. Other dimension reduction approaches are also discussed. In addition, we review another core technique, functional linear regression, as well as clustering and classification of functional d..."
                },
                {
                    "title": "Spatio-Temporal Covariance and Cross-Covariance Functions of the Great Circle Distance on a Sphere",
                    "abstract": "Abstract In this article, we propose stationary covariance functions for processes that evolve temporally over a sphere, as well as cross-covariance functions for multivariate random fields defined over a sphere. For such processes, the great circle distance is the natural metric that should be used to describe spatial dependence. Given the mathematical difficulties for the construction of covariance functions for processes defined over spheres cross time, approximations of the state of nature have been proposed in the literature by using the Euclidean (based on map projections) and the chordal distances. We present several methods of construction based on the great circle distance and provide closed-form expressions for both spatio-temporal and multivariate cases. A simulation study assesses the discrepancy between the great circle distance, chordal distance, and Euclidean distance based on a map projection both in terms of estimation and prediction in a space-time and a bivariate spatial setting, where the space is in this case the Earth. We revisit the analysis of Total Ozone Mapping Spectrometer (TOMS) data and investigate differences in terms of estimation and prediction between the aforementioned distance-based approaches. Both simulation and real data highlight sensible differences in terms of estimation of the spatial scale parameter. As far as prediction is concerned, the differences can be appreciated only when the interpoint distances are large, as demonstrated by an illustrative example. Supplementary materials for this article are available online."
                },
                {
                    "title": "A two\u2010sample distribution\u2010free test for functional data with application to a diffusion tensor imaging study of multiple sclerosis",
                    "abstract": "Motivated by an imaging study, the paper develops a non\u2010parametric testing procedure for testing the null hypothesis that two samples of curves observed at discrete grids and with noise have the same underlying distribution. The objective is to compare formally white matter tract profiles between healthy individuals and multiple\u2010sclerosis patients, as assessed by conventional diffusion tensor imaging measures. We propose to decompose the curves by using functional principal component analysis of a mixture process, which we refer to as marginal functional principal component analysis. This approach reduces the dimension of the testing problem in a way that enables the use of traditional non\u2010parametric univariate testing procedures. The procedure is computationally efficient and accommodates different sampling designs. Numerical studies are presented to validate the size and power properties of the test in many realistic scenarios. In these cases, the test proposed has been found to be more powerful than its primary competitor. Application to the diffusion tensor imaging data reveals that all the tracts studied are associated with multiple sclerosis and the choice of the diffusion tensor image measurement is important when assessing axonal disruption."
                },
                {
                    "title": "Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization",
                    "abstract": "Abstract. By coordinating the design and distribution of global climate model simulations of the past, current, and future climate, the Coupled Model Intercomparison Project (CMIP) has become one of the foundational elements of climate science. However, the need to address an ever-expanding range of scientific questions arising from more and more research communities has made it necessary to revise the organization of CMIP. After a long and wide community consultation, a new and more federated structure has been put in place. It consists of three major elements: (1)\u00a0a handful of common experiments, the DECK (Diagnostic, Evaluation and Characterization of Klima) and CMIP historical simulations (1850\u2013near present) that will maintain continuity and help document basic characteristics of models across different phases of CMIP; (2)\u00a0common standards, coordination, infrastructure, and documentation that will facilitate the distribution of model outputs and the characterization of the model ensemble; and (3)\u00a0an ensemble of CMIP-Endorsed Model Intercomparison Projects (MIPs) that will be specific to a particular phase of CMIP (now CMIP6) and that will build on the DECK and CMIP historical simulations to address a large range of specific questions and fill the scientific gaps of the previous CMIP phases. The DECK and CMIP historical simulations, together with the use of CMIP data standards, will be the entry cards for models participating in CMIP. Participation in CMIP6-Endorsed MIPs by individual modelling groups will be at their own discretion and will depend on their scientific interests and priorities. With the Grand Science Challenges of the World Climate Research Programme (WCRP) as its scientific backdrop, CMIP6 will address three broad questions: \u2013 How does the Earth system respond to forcing? \u2013 What are the origins and consequences of systematic model biases? \u2013 How can we assess future climate changes given internal climate variability, predictability, and uncertainties in scenarios? This CMIP6 overview paper presents the background and rationale for the new structure of CMIP, provides a detailed description of the DECK and CMIP6 historical simulations, and includes a brief introduction to the 21\u00a0CMIP6-Endorsed MIPs."
                },
                {
                    "title": "Two sample inference for the second-order property of temporally dependent functional data",
                    "abstract": "Motivated by the need to statistically quantify the difference between two spatio-temporal datasets that arise in climate downscaling studies, we propose new tests to detect the differences of the covariance operators and their associated characteristics of two functional time series. Our two sample tests are constructed on the basis of functional principal component analysis and self-normalization, the latter of which is a new studentization technique recently developed for the inference of a univariate time series. Compared to the existing tests, our SN-based tests allow for weak dependence within each sample and it is robust to the dependence between the two samples in the case of equal sample sizes. Asymptotic properties of the SN-based test statistics are derived under both the null and local alternatives. Through extensive simulations, our SN-based tests are shown to outperform existing alternatives in size and their powers are found to be respectable. The tests are then applied to the gridded climate model outputs and interpolated observations to detect the difference in their spatial dynamics."
                },
                {
                    "title": "A Multiresolution Gaussian Process Model for the Analysis of Large Spatial Datasets",
                    "abstract": "We develop a multiresolution model to predict two-dimensional spatial fields based on irregularly spaced observations. The radial basis functions at each level of resolution are constructed using a Wendland compactly supported correlation function with the nodes arranged on a rectangular grid. The grid at each finer level increases by a factor of two and the basis functions are scaled to have a constant overlap. The coefficients associated with the basis functions at each level of resolution are distributed according to a Gaussian Markov random field (GMRF) and take advantage of the fact that the basis is organized as a lattice. Several numerical examples and analytical results establish that this scheme gives a good approximation to standard covariance functions such as the Mat\u00e9rn and also has flexibility to fit more complicated shapes. The other important feature of this model is that it can be applied to statistical inference for large spatial datasets because key matrices in the computations are sparse. The computational efficiency applies to both the evaluation of the likelihood and spatial predictions."
                },
                {
                    "title": "Modelling non\u2010Euclidean movement and landscape connectivity in highly structured ecological networks",
                    "abstract": "Movement is influenced by landscape structure, configuration and geometry, but measuring distance as perceived by animals poses technical and logistical challenges. Instead, movement is typically measured using Euclidean distance, irrespective of location or landscape structure, or is based on arbitrary cost surfaces. A recently proposed extension of spatial capture\u2010recapture (SCR) models resolves this issue using spatial encounter histories of individuals to calculate least\u2010cost paths (ecological distance: Ecology, 94, 2013, 287) thereby relaxing the Euclidean assumption. We evaluate the consequences of not accounting for movement heterogeneity when estimating abundance in highly structured landscapes, and demonstrate the value of this approach for estimating biologically realistic space\u2010use patterns and landscape connectivity. We simulated SCR data in a riparian habitat network, using the ecological distance model under a range of scenarios where space\u2010use in and around the landscape was increasingly associated with water (i.e. increasingly less Euclidean). To assess the influence of miscalculating distance on estimates of population size, we compared the results from the ecological and Euclidean distance based models. We then demonstrate that the ecological distance model can be used to estimate home range geometry when space use is not symmetrical. Finally, we provide a method for calculating landscape connectivity based on modelled species\u2010landscape interactions generated from capture\u2010recapture data. Using ecological distance always produced unbiased estimates of abundance. Explicitly modelling the strength of the species\u2010landscape interaction provided a direct measure of landscape connectivity and better characterised true home range geometry. Abundance under the Euclidean distance model was increasingly (negatively) biased as space use was more strongly associated with water and, because home ranges are assumed to be symmetrical, produced poor characterisations of home range geometry and no information about landscape connectivity. The ecological distance SCR model uses spatially indexed capture\u2010recapture data to estimate how activity patterns are influenced by landscape structure. As well as reducing bias in estimates of abundance, this approach provides biologically realistic representations of home range geometry, and direct information about species\u2010landscape interactions. The incorporation of both structural (landscape) and functional (movement) components of connectivity provides a direct measure of species\u2010specific landscape connectivity."
                },
                {
                    "title": "Constructing valid spatial processes on the sphere using kernel convolutions",
                    "abstract": "Remotely sensed data products are now routinely used to study various aspects of the Earth's atmosphere. These remote sensing datasets are typically very high dimensional, have near global coverage and exhibit nonstationary spatial correlation structures. Proper statistical analysis of these datasets should be sufficiently flexible to account for all these aspects. To this end, we develop a kernel convolution construction of spatial processes on a sphere. As is the case with kernel convolution constructions on the plane, we establish a link between stationary kernels and a stationary covariance function on the sphere via the spherical harmonic decomposition of the kernel. We also introduce the Kent distribution as an appropriate kernel with interpretable parameters to be used in the kernel convolution construction. We demonstrate the discrete kernel convolution model using a dataset of remotely sensed CO2 concentrations over the globe. Copyright \u00a9 2014 John Wiley & Sons, Ltd."
                },
                {
                    "title": "Evaluation of climate models",
                    "abstract": "Climate models have continued to be developed and improved since the AR4, and many models have been extended into Earth System models by including the representation of biogeochemical cycles important to climate change. These models allow for policy-relevant calculations such as the carbon dioxide (CO2) emissions compatible with a specified climate stabilization target. In addition, the range of climate variables and processes that have been evaluated has greatly expanded, and differences between models and observations are increasingly quantified using \u2018performance metrics\u2019. In this chapter, model evaluation covers simulation of the mean climate, of historical climate change, of variability on multiple time scales and of regional modes of variability. This evaluation is based on recent internationally coordinated model experiments, including simulations of historic and paleo climate, specialized experiments designed to provide insight into key climate processes and feedbacks and regional climate downscaling. Figure 9.44 provides an overview of model capabilities as assessed in this chapter, including improvements, or lack thereof, relative to models assessed in the AR4. The chapter concludes with an assessment of recent work connecting model performance to the detection and attribution of climate change as well as to future projections. {9.1.2, 9.8.1, Table 9.1, Figure 9.44} \n \nThe ability of climate models to simulate surface temperature has improved in many, though not all, important aspects relative to the generation of models assessed in the AR4."
                },
                {
                    "title": "Defining spatial comparison metrics for evaluation of paleoclimatic field reconstructions of the Common Era",
                    "abstract": "Climate field reconstructions (CFR) of the Common Era (the last two millennia) provide important insights into the dynamics of past climate change that, in turn, have implications for the future. Multiple CFR methods have emerged in the literature, and comparisons between these methods using pseudoproxy experiments have been performed. These experiments, however, have not fully quantified the spatial skill of the CFRs, particularly with regard to the relative performance of each. Toward such ends, a formal statistical hypothesis test is proposed as a means of evaluating the differences between two random fields that integrate the differences in both the mean and the dependence structure. This involves a careful selection of the statistical model for the CFR residual process and systematic comparisons over different spatial scales. Application of this method yields a systematic assessment of the spatial character of five widely applied CFRs in a pseudoproxy experiment context. The analyses indicate that spatial differences among the five CFRs are not statistically significant. Further rigorous statistical assessments will help elucidate the strength and weakness of each CFR method, while quantifying the degree to which their spatial dissimilarities can be ascribed to methodological choices. Copyright \u00a9 2012 John Wiley & Sons, Ltd."
                },
                {
                    "title": "Comparing Spatial Predictions",
                    "abstract": "Under a general loss function, we develop a hypothesis test to determine whether a significant difference in the spatial predictions produced by two competing models exists on average across the entire spatial domain of interest. The null hypothesis is that of no difference, and a spatial loss differential is created based on the observed data, the two sets of predictions, and the loss function chosen by the researcher. The test assumes only isotropy and short-range spatial dependence of the loss differential but does allow it to be non-Gaussian, non-zero-mean, and spatially correlated. Constant and nonconstant spatial trends in the loss differential are treated in two separate cases. Monte Carlo simulations illustrate the size and power properties of this test, and an example based on daily average wind speeds in Oklahoma is used for illustration. Supplemental results are available online."
                },
                {
                    "title": "Estimation of the mean of functional time series and a two\u2010sample problem",
                    "abstract": "Summary.\u2002 The paper is concerned with inference based on the mean function of a functional time series. We develop a normal approximation for the functional sample mean and then focus on the estimation of the asymptotic variance kernel. Using these results, we develop and asymptotically justify testing procedures for the equality of means in two functional samples exhibiting temporal dependence. Evaluated by means of a simulation study and application to a real data set, these two\u2010sample procedures enjoy good size and power in finite samples."
                },
                {
                    "title": "Revisiting Climate Region Definitions via Clustering",
                    "abstract": "Abstract This paper introduces a new distance metric that enables the clustering of general climatic time series. Clustering methods have been frequently used to partition a domain of interest into distinct climatic zones. However, previous techniques have neglected the time series (autocorrelation) component and have also handled seasonal features in a suboptimal way. The distance proposed here incorporates the seasonal mean and autocorrelation structures of the series in a natural way; moreover, trends and covariate effects can be considered. As an important by-product, the methods can be used to statistically assess whether two stations can serve as reference stations for one another. The methods are illustrated by partitioning 292 weather stations within the state of Colorado into six different zones."
                },
                {
                    "title": "Statistical inferences for functional data",
                    "abstract": "With modern technology development, functional data are being observed frequently in many scientific fields. A popular method for analyzing such functional data is ``smoothing first, then estimation.'' That is, statistical inference such as estimation and hypothesis testing about functional data is conducted based on the substitution of the underlying individual functions by their reconstructions obtained by one smoothing technique or another. However, little is known about this substitution effect on functional data analysis. In this paper this problem is investigated when the local polynomial kernel (LPK) smoothing technique is used for individual function reconstructions. We find that under some mild conditions, the substitution effect can be ignored asymptotically. Based on this, we construct LPK reconstruction-based estimators for the mean, covariance and noise variance functions of a functional data set and derive their asymptotics. We also propose a GCV rule for selecting good bandwidths for the LPK reconstructions. When the mean function also depends on some time-independent covariates, we consider a functional linear model where the mean function is linearly related to the covariates but the covariate effects are functions of time. The LPK reconstruction-based estimators for the covariate effects and the covariance function are also constructed and their asymptotics are derived. Moreover, we propose a $L^2$-norm-based global test statistic for a general hypothesis testing problem about the covariate effects and derive its asymptotic random expression. The effect of the bandwidths selected by the proposed GCV rule on the accuracy of the LPK reconstructions and the mean function estimator is investigated via a simulation study. The proposed methodologies are illustrated via an application to a real functional data set collected in climatology."
                },
                {
                    "title": "How reliable are climate models?",
                    "abstract": "ABSTRACT How much can we trust model-based projections of future anthropogenic climate change? This review attempts to give an overview of this important but difficult topic by using three main lines of evidence: the skill of models in simulating present-day climate, intermodel agreement on future climate changes, and the ability of models to simulate climate changes that have already occurred. A comparison of simulated and observed present-day climates shows good agreement for many basic variables, particularly at large horizontal scales, and a tendency for biases to vary in sign between different models, but there is a risk that these features might be partly a result of tuning. Overall, the connection between model skill in simulating present-day climate and the skill in simulating future climate changes is poorly known. An intercomparison of future climate changes between models shows a better agreement for changes in temperature than that for precipitation and sea level pressure, but some aspects of change in the latter two variables are also quite consistent between models. A comparison of simulations with observed climate changes is, in principle, a good test for the models, but there are several complications. Nonetheless, models have skilfully simulated many large-scale aspects of observed climate changes, including but not limited to the evolution of the global mean surface air temperature in the 20th century. Furthermore, although there is no detailed agreement between the simulated and observed geographical patterns of change, the grid box scale temperature, precipitation and pressure changes observed during the past half-century generally fall within the range of model results. Considering the difficulties associated with other sources of information, the variation of climate changes between different models is probably the most meaningful measure of uncertainty that is presently available. In general, however, this measure is more likely to underestimate than overestimate the actual uncertainty."
                },
                {
                    "title": "Can climate trends be calculated from reanalysis data",
                    "abstract": "Several global quantities are computed from the ERA40 reanalysis for \nthe period 1958-2001 and explored for trends. These are discussed \nin the context of changes to the global observing system. Temperature, \nintegrated water vapor (IWV), and kinetic energy are considered. \nThe ERA40 global mean temperature in the lower troposphere has a \ntrend of +0.11 K per decade over the period of 1979-2001, which is \nslightly higher than the MSU measurements, but within the estimated \nerror limit. For the period 1958 2001 the warming trend is 0.14 K \nper decade but this is likely to be an artifact of changes in the \nobserving system. When this is corrected for, the warming trend is \nreduced to 0.10 K per decade. The global trend in IWV for the period \n1979-2001 is +0.36 mm per decade. This is about twice as high as \nthe trend determined from the Clausius-Clapeyron relation assuming \nconservation of relative humidity. It is also larger than results \nfrom free climate model integrations driven by the same observed \nsea surface temperature as used in ERA40. It is suggested that the \nlarge trend in IWV does not represent a genuine climate trend but \nan artifact caused by changes in the global observing system such \nas the use of SSM/I and more satellite soundings in later years. \nRecent results are in good agreement with GPS measurements. The IWV \ntrend for the period 1958-2001 is still higher but reduced to +0.16 \nmm per decade when corrected for changes in the observing systems. \nTotal kinetic energy shows an increasing global trend. Results from \ndata assimilation experiments strongly suggest that this trend is \nalso incorrect and mainly caused by the huge changes in the global \nobserving system in 1979. When this is corrected for, no significant \nchange in global kinetic energy from 1958 onward can be found."
                },
                {
                    "title": "Nonparametric hypothesis testing for a spatial signal",
                    "abstract": "Summary form only given. Nonparametric hypothesis testing for a spatial signal can involve a large number of hypotheses. For instance, two satellite images of the same scene, taken before and after an event, could be used to test a hypothesis that the event has no environmental impact. This is equivalent to testing that the mean difference of \"after-before\" is zero at each of the (typically thousands of) pixels that make up the scene. In such a situation, conventional testing procedures that control the overall Type I error deteriorate as the number of hypotheses increase. Powerful testing procedures are needed for this problem of testing for the presence of a spatial signal. In this talk, we propose a procedure called enhanced FDR (EFDR), which is based on controlling the false discovery rate (FDR) and a concept known as generalized degrees of freedom (GDF). EFDR differs from the standard FDR procedure through its reducing of the number of hypotheses tested. This is done in two ways: first, the model is represented more parsimoniously in the wavelet domain, and second, an optimal selection of hypotheses is made using a criterion based on generalized degrees of freedom. Not only does the EFDR procedure tell us whether a spatial signal is present or not, it has an added bonus that, if a signal is deemed present, it can indicate its location and magnitude. The EFDR procedure is applied to an air-temperature data set generated from the climate system model (CSM) of the National Center for Atmospheric Research (NCAR) and to brain-imaging data from fMRI experiments."
                },
                {
                    "title": "NCEP\u2013DOE AMIP-II Reanalysis (R-2)",
                    "abstract": "The NCEP\u2013DOE Atmospheric Model Intercomparison Project (AMIP-II) reanalysis is a follow-on project to the \u201c50-year\u201d (1948\u2013present) NCEP\u2013NCAR Reanalysis Project. NCEP\u2013DOE AMIP-II re-analysis covers the \u201c20-year\u201d satellite period of 1979 to the present and uses an updated forecast model, updated data assimilation system, improved diagnostic outputs, and fixes for the known processing problems of the NCEP\u2013NCAR reanalysis. Only minor differences are found in the primary analysis variables such as free atmospheric geopotential height and winds in the Northern Hemisphere extratropics, while significant improvements upon NCEP\u2013NCAR reanalysis are made in land surface parameters and land\u2013ocean fluxes. This analysis can be used as a supplement to the NCEP\u2013NCAR reanalysis especially where the original analysis has problems. The differences between the two analyses also provide a measure of uncertainty in current analyses."
                },
                {
                    "title": "A Comparison of the NCEP-NCAR Reanalysis Precipitation and the GPCP Rain Gauge-Satellite Combined Dataset with Observational Error Considerations",
                    "abstract": "Abstract The Global Precipitation Climatology Project (GPCP) has released monthly mean estimates of precipitation that comprise gauge observations and satellite-derived precipitation estimates. Estimates of standard random error for each month at each grid location are also provided in this data release. One of the primary intended uses of this dataset is the validation of climatic-scale precipitation fields that are produced by numerical models. Nearly coincident with this dataset development, the National Centers for Environmental Prediction and the National Center for Atmospheric Research have joined in a cooperative effort to reanalyze meteorological fields from the present back to the 1940s using a fixed state-of-the-art data assimilation system and large input database. In this paper, monthly accumulations of reanalysis precipitation are compared with the GPCP combined rain gauge\u2013satellite dataset over the period 1988\u201395. A unique feature of this comparison is the use of standard error estimates tha..."
                },
                {
                    "title": "The Seasonal Cycle over the Tropical Pacific in Coupled Ocean Atmosphere General Circulation Models",
                    "abstract": "Abstract The seasonal cycle over the tropical Pacific simulated by 11 coupled ocean\u2013atmosphere general circulation models (GCMs) is examined. Each model consists of a high-resolution ocean GCM of either the tropical Pacific or near-global means coupled to a moderate- or high-resolution atmospheric GCM, without the use of flux correction. The seasonal behavior of sea surface temperature (SST) and eastern Pacific rainfall is presented for each model. The results show that current state-of-the-art coupled GCMs share important successes and troublesome systematic errors. All 11 models are able to simulate the mean zonal gradient in SST at the equator over the central Pacific. The simulated equatorial cold tongue generally tends to be too strong, too narrow, and extend too far west. SSTs are generally too warm in a broad region west of Peru and in a band near 10\u00b0S. This is accompanied in some models by a double intertropical convergence zone (ITCZ) straddling the equator over the eastern Pacific, and in others..."
                },
                {
                    "title": "Computing Fourier Transforms and Convolutions on the 2-Sphere",
                    "abstract": "This paper considers the problem of efficient computation of the spherical harmonic expansion, or Fourier transform, of functions defined on the two dimensional sphere, S^2. The resulting algorithms are applied to the efficient computation of convolutions of functions on the sphere. We begin by proving convolution theorems generalizing well known and useful results from the abelian case. These convolution theorems are then used to develop a sampling theorem on the sphere. which reduces the calculation of Fourier transforms and convolutions of band-limited functions to discrete computations. We show how to perform these efficiently, starting with an O(n(log n)^2) time algorithm for computing the Legendre transform of a function defined on the interval [-1,1] sampled at n points there. Theoretical and experimental results on the effects of finite precision arithmetic are presented. The Legendre transform algorithm is then generalized to obtain an algorithm for the Fourier transform, requiring O(n(log n)^2) time, and an algorithm for its inverse in O(n^1^.^5) time, where n is the number of points on the sphere at which the function is sampled. This improves the naive O(n^2) bound, which is the best previously known. These transforms give an O(n^1^.^5) algorithm for convolving two functions on the sphere."
                },
                {
                    "title": "An Introduction to Three-Dimensional Climate Modeling",
                    "abstract": "This book provides a guide to the development and use of computer models of the Earth's climate. The book describes the basic theory of climate simulation, including the fundamental equations and relevant numerical techniques for simulating the atmosphere, oceans, and sea ice. Results for a variety of past, present, and future climates are shown and compared with observations."
                },
                {
                    "title": "Functional Analysis: Introduction to Further Topics in Analysis",
                    "abstract": "This is the fourth and final volume in the Princeton Lectures in Analysis, a series of textbooks that aim to present, in an integrated manner, the core areas of analysis. Beginning with the basic facts of functional analysis, this volume looks at Banach spaces, Lp spaces, and distribution theory, and highlights their roles in harmonic analysis. The authors then use the Baire category theorem to illustrate several points, including the existence of Besicovitch sets. The second half of the book introduces readers to other central topics in analysis, such as probability theory and Brownian motion, which culminates in the solution of Dirichlet's problem. The concluding chapters explore several complex variables and oscillatory integrals in Fourier analysis, and illustrate applications to such diverse areas as nonlinear dispersion equations and the problem of counting lattice points. Throughout the book, the authors focus on key results in each area and stress the organic unity of the subject. * A comprehensive and authoritative text that treats some of the main topics of modern analysis * A look at basic functional analysis and its applications in harmonic analysis, probability theory, and several complex variables * Key results in each area discussed in relation to other areas of mathematics * Highlights the organic unity of large areas of analysis traditionally split into subfields * Interesting exercises and problems illustrate ideas * Clear proofs provided"
                },
                {
                    "title": "Detecting signals in FMRI data using powerful FDR procedures",
                    "abstract": "Functional magnetic resonance imaging (FMRI) has revolutionized the study of linking physical stimuli with localized brain activity. Among the challenges of working with FMRI data, they are noisy, they exhibit spatial correlation, and they are usually large containing tens of thousands of voxels of information. The notion of False Discovery Rate (FDR) has made a great impact on how to perform powerful multiple hypothesis tests to detect signals in such large multivariate data. The spatial dependence in FMRI data requires special care since, if ignored, it can lead to a loss of control of size as well as a deterioration in power of FDR procedures. This article advocates transforming the voxelwise test statistics to wavelet space, where the coefficients are approximately uncorrelated. We demonstrate, through a series of experiments, that an FDR procedure in wavelet space enhanced by P -value adaptive thresholding (EPAT), maintains control of the size of the multiple-testing procedure and offers substantially increased power over an FDR procedure that is applied directly to the map of (spatially dependent) test statistics. The EPAT methodology, developed here for FMRI data, is generic and can be applied in other dependent data settings."
                },
                {
                    "title": "Two-sample tests in functional data analysis starting from discrete data",
                    "abstract": "One of the ways in which functional data analysis differs from other areas of statistics is in the extent to which data are pre-processed prior to analysis. Functional data are invariably recorded discretely, although they are generally substantially smoothed as a prelude even to viewing by statisticians, let alone further analysis. This has a potential to interfere with the performance of two-sample statistical tests, since the use of different tuning parameters for the smoothing step, or different observation times or subsample sizes (i.e., numbers of observations per curve), can mask the differences between distributions that a test is trying to locate. In this paper, and in the context of two-sample tests, we take up this issue. Ways of pre-processing the data, so as to minimise the effects of smoothing, are suggested. We show theoretically and numerically that, by employing exactly the same tuning parameter (e.g. bandwidth) to produce each curve from its raw data, significant loss of power can be avoided. Provided a common tuning parameter is used, it is often satisfactory to choose that parameter along conventional lines, as though the target was estimation of the continuous functions themselves, rather than testing hypotheses about them. Moreover, in this case, using a second-order smoother (such as a local-linear method), the subsample sizes can be almost as small as the square root of sample sizes before the effects of smoothing have any first-order impact on the results of a two-sample test."
                },
                {
                    "title": "Invariant measures for a stochastic nonlinear Schroedinger equation",
                    "abstract": "We will prove the existence of an invariant measure for a nonlinear Schrodinger equation with random noise in R n . The existence of solutions of the Cauchy problem in H 1 - and L 2 -settings was established by de Bouard and Debussche [4, 5]. Here we discuss only the defocusing equation with a zero-order dissipation. The proof for the existence of an invariant measure is based on various energy estimates and the approximation scheme to construct solutions, which are also crucial for the existence of global solutions to the Cauchy problem."
                }
            ],
            "categories": [
                "stat.AP",
                "cs.LG",
                "physics.ao-ph",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the validation of climate models against observational data using advanced machine learning techniques?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nImproving climate model validation is crucial for enhancing the reliability of climate projections, which have significant implications for policy-making, environmental management, and understanding climate change impacts. A more robust validation framework could lead to better-informed decisions in climate adaptation and mitigation strategies. Additionally, advancements in this area could stimulate further research in machine learning applications within climate science, potentially leading to innovative methodologies that can be applied across various domains.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in improving climate model validation stem from the inherent complexity of climate systems, which involve numerous interacting variables and stochastic processes. Naive approaches, such as simple error metrics like RMSE, may fail to capture the nuanced relationships and dependencies in the data, leading to misleading conclusions. Moreover, the high dimensionality of climate data and the need for spatial-temporal analysis introduce significant computational and theoretical obstacles, requiring sophisticated statistical and machine learning techniques to adequately address these complexities.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily relied on traditional statistical methods and basic machine learning techniques that do not fully account for the randomness and spatial dependencies in climate data. Limitations in computational power and the lack of comprehensive datasets have also hindered progress. Many existing approaches focus on specific aspects of model validation without integrating a holistic view of the climate system. Our approach aims to bridge these gaps by employing advanced machine learning methods that consider the full complexity of climate data, thus providing a more thorough validation framework.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the use of advanced machine learning techniques, such as deep learning and functional data analysis, to compare climate model outputs with reanalysis data. We will utilize a comprehensive dataset from the Coupled Model Intercomparison Project (CMIP) and employ metrics that account for both spatial and temporal dependencies. The expected outcomes include a more accurate assessment of model performance, identification of systematic biases, and improved understanding of the dynamics of climate models, ultimately leading to enhanced model reliability and predictive capabilities."
            }
        },
        "author_data": {
            "64cfc251-4351-41e5-aa72-63f5774de296": {
                "pk": "64cfc251-4351-41e5-aa72-63f5774de296",
                "name": "Robert C. Garrett",
                "collaborators": [
                    "Trevor Harris",
                    "Bo Li"
                ],
                "domain": [
                    "Climate Modeling",
                    "Statistical Analysis",
                    "Metric Development"
                ],
                "publications": [
                    {
                        "title": "Evaluating Climate Models with Sliced Elastic Distance",
                        "abstract": "The validation of global climate models plays a crucial role in ensuring the accuracy of climatological predictions. However, existing statistical methods for evaluating differences between climate fields often overlook time misalignment and therefore fail to distinguish between sources of variability. To more comprehensively measure differences between climate fields, we introduce a new vector-valued metric, the sliced elastic distance. This new metric simultaneously accounts for spatial and temporal variability while decomposing the total distance into shape differences (amplitude), timing variability (phase), and bias (translation). We compare the sliced elastic distance against a classical metric and a newly developed Wasserstein-based approach through a simulation study. Our results demonstrate that the sliced elastic distance outperforms previous methods by capturing a broader range of features. We then apply our metric to evaluate the historical model outputs of the Coupled Model Intercomparison Project (CMIP) members, focusing on monthly average surface temperatures and monthly total precipitation. By comparing these model outputs with quasi-observational ERA5 Reanalysis data products, we rank the CMIP models and assess their performance. Additionally, we investigate the progression from CMIP phase 5 to phase 6 and find modest improvements in the phase 6 models regarding their ability to produce realistic climate dynamics."
                    }
                ]
            },
            "f4f49299-0664-444f-a571-3533e47540d3": {
                "pk": "f4f49299-0664-444f-a571-3533e47540d3",
                "name": "Trevor Harris",
                "collaborators": [
                    "Bo Li",
                    "Ryan Sriver",
                    "James Derek Tucker",
                    "Mengchen Wang",
                    "Robert C. Garrett",
                    "Lyndsay Shand"
                ],
                "domain": [
                    "Climate Modeling",
                    "Uncertainty Quantification",
                    "Functional Data Analysis",
                    "Bayesian Statistics"
                ],
                "publications": [
                    {
                        "title": "Quantifying uncertainty in climate projections with conformal ensembles",
                        "abstract": "Large climate model ensembles are the primary tool for robustly projecting future climate states and quantifying projection uncertainty. Despite significant advancements in climate modeling over the past few decades, overall projection certainty has not commensurately decreased with steadily improving model skill. We introduce conformal ensembling, a new approach to uncertainty quantification in climate projections based on conformal inference to reduce projection uncertainty. Unlike traditional methods, conformal ensembling seamlessly integrates climate model ensembles and observational data across a range of scales to generate statistically rigorous, easy-to-interpret uncertainty estimates. It can be applied to any climatic variable using any ensemble analysis method and outperforms existing inter-model variability methods in uncertainty quantification across all time horizons and most spatial locations under SSP2-4.5. Conformal ensembling is also computationally efficient, requires minimal assumptions, and is highly robust to the conformity measure. Experiments show that it is effective when conditioning future projections on historical reanalysis data compared with standard ensemble averaging approaches, yielding more physically consistent projections."
                    },
                    {
                        "title": "Scalable Multiple Changepoint Detection for Functional Data Sequences",
                        "abstract": "We propose the Multiple Changepoint Isolation (MCI) method for detecting multiple changes in the mean and covariance of a functional process. We first introduce a pair of projections to represent the variability \"between\" and \"within\" the functional observations. We then present an augmented fused lasso procedure to split the projections into multiple regions robustly. These regions act to isolate each changepoint away from the others so that the powerful univariate CUSUM statistic can be applied region-wise to identify the changepoints. Simulations show that our method accurately detects the number and locations of changepoints under many different scenarios. These include light and heavy tailed data, data with symmetric and skewed distributions, sparsely and densely sampled changepoints, and mean and covariance changes. We show that our method outperforms a recent multiple functional changepoint detector and several univariate changepoint detectors applied to our proposed projections. We also show that MCI is more robust than existing approaches and scales linearly with sample size. Finally, we demonstrate our method on a large time series of water vapor mixing ratio profiles from atmospheric emitted radiance interferometer measurements."
                    },
                    {
                        "title": "Multi-model Ensemble Analysis with Neural Network Gaussian Processes",
                        "abstract": "Multi-model ensemble analysis integrates information from multiple climate models into a unified projection. However, existing integration approaches based on model averaging can dilute fine-scale spatial information and incur bias from rescaling low-resolution climate models. We propose a statistical approach, called NN-GPR, using Gaussian process regression (GPR) with an infinitely wide deep neural network based covariance function. NN-GPR requires no assumptions about the relationships between models, no interpolation to a common grid, no stationarity assumptions, and automatically downscales as part of its prediction algorithm. Model experiments show that NN-GPR can be highly skillful at surface temperature and precipitation forecasting by preserving geospatial signals at multiple scales and capturing inter-annual variability. Our projections particularly show improved accuracy and uncertainty quantification skill in regions of high variability, which allows us to cheaply assess tail behavior at a 0.44$^\\circ$/50 km spatial resolution without a regional climate model (RCM). Evaluations on reanalysis data and SSP245 forced climate models show that NN-GPR produces similar, overall climatologies to the model ensemble while better capturing fine scale spatial patterns. Finally, we compare NN-GPR's regional predictions against two RCMs and show that NN-GPR can rival the performance of RCMs using only global model data as input."
                    },
                    {
                        "title": "Bayesian Changepoint Estimation for Spatially Indexed Functional Time Series",
                        "abstract": "We propose a Bayesian hierarchical model to simultaneously estimate mean based changepoints in spatially correlated functional time series. Unlike previous methods that assume a shared changepoint at all spatial locations or ignore spatial correlation, our method treats changepoints as a spatial process. This allows our model to respect spatial heterogeneity and exploit spatial correlations to improve estimation. Our method is derived from the ubiquitous cumulative sum (CUSUM) statistic that dominates changepoint detection in functional time series. However, instead of directly searching for the maximum of the CUSUM based processes, we build spatially correlated two-piece linear models with appropriate variance structure to locate all changepoints at once. The proposed linear model approach increases the robustness of our method to variability in the CUSUM process, which, combined with our spatial correlation model, improves changepoint estimation near the edges. We demonstrate through extensive simulation studies that our method outperforms existing functional changepoint estimators in terms of both estimation accuracy and uncertainty quantification, under either weak and strong spatial correlation, and weak and strong change signals. Finally, we demonstrate our method using a temperature data set and a coronavirus disease 2019 (COVID-19) study."
                    },
                    {
                        "title": "Evaluating Climate Models with Sliced Elastic Distance",
                        "abstract": "The validation of global climate models plays a crucial role in ensuring the accuracy of climatological predictions. However, existing statistical methods for evaluating differences between climate fields often overlook time misalignment and therefore fail to distinguish between sources of variability. To more comprehensively measure differences between climate fields, we introduce a new vector-valued metric, the sliced elastic distance. This new metric simultaneously accounts for spatial and temporal variability while decomposing the total distance into shape differences (amplitude), timing variability (phase), and bias (translation). We compare the sliced elastic distance against a classical metric and a newly developed Wasserstein-based approach through a simulation study. Our results demonstrate that the sliced elastic distance outperforms previous methods by capturing a broader range of features. We then apply our metric to evaluate the historical model outputs of the Coupled Model Intercomparison Project (CMIP) members, focusing on monthly average surface temperatures and monthly total precipitation. By comparing these model outputs with quasi-observational ERA5 Reanalysis data products, we rank the CMIP models and assess their performance. Additionally, we investigate the progression from CMIP phase 5 to phase 6 and find modest improvements in the phase 6 models regarding their ability to produce realistic climate dynamics."
                    },
                    {
                        "title": "Elastic depths for detecting shape anomalies in functional data",
                        "abstract": "We propose a new family of depth measures called the elastic depths that can be used to greatly improve shape anomaly detection in functional data. Shape anomalies are functions that have considerably different geometric forms or features from the rest of the data. Identifying them is generally more difficult than identifying magnitude anomalies because shape anomalies are often not distinguishable from the bulk of the data with visualization methods. The proposed elastic depths use the recently developed elastic distances to directly measure the centrality of functions in the amplitude and phase spaces. Measuring shape outlyingness in these spaces provides a rigorous quantification of shape, which gives the elastic depths a strong theoretical and practical advantage over other methods in detecting shape anomalies. A simple boxplot and thresholding method is introduced to identify shape anomalies using the elastic depths. We assess the elastic depth's detection skill on simulated shape outlier scenarios and compare them against popular shape anomaly detectors. Finally, we use hurricane trajectories to demonstrate the elastic depth methodology on manifold valued functional data. Supplementary materials, including additional simulations, data examples, and an R-package are available online."
                    }
                ]
            },
            "f7ec7212-91d4-46dc-95ac-6384c368b491": {
                "pk": "f7ec7212-91d4-46dc-95ac-6384c368b491",
                "name": "Bo Li",
                "collaborators": [
                    "Xiaowen Zhou",
                    "Zihui Zhang",
                    "Bo Yang",
                    "Bing Wang",
                    "Chunhao Cai"
                ],
                "domain": [
                    "Mathematics",
                    "Machine Learning",
                    "3D Modeling",
                    "Financial Analysis"
                ],
                "publications": [
                    {
                        "title": "Parameterized Littlewood-Paley operators with variable kernels on Hardy spaces and weak Hardy spaces",
                        "abstract": "In this paper, by using the atomic decomposition theory of Hardy space and weak Hardy space, we discuss the boundedness of parameterized Littlewood-Paley operator with variable kernel on these spaces."
                    },
                    {
                        "title": "Invertible Convolution with Symmetric Paddings",
                        "abstract": "We show that symmetrically padded convolution can be analytically inverted via DFT. We comprehensively analyze several different symmetric and anti-symmetric padding modes and show that multiple cases exist where the inversion can be achieved. The implementation is available at \\url{https://github.com/prclibo/iconv_dft}."
                    },
                    {
                        "title": "Boundedness of parameterized Marcinkiewicz integrals with variable kernels on Hardy spaces and weak Hardy spaces",
                        "abstract": "In this paper, by using the atomic decomposition theory of Hardy space and weak Hardy space, the author establishes the boundedness of parameterized Marcinkiewicz integral with variable kernel on these spaces, under the Dini condition or H\\\"{o}rmander condition imposed on kernel."
                    },
                    {
                        "title": "Weighted Norm Inequalities for Parametric Littlewood-Paley Operators",
                        "abstract": "In this paper, the author establishes the boundedness of parametric Littlewood-Paley operators from Musielak-Orlicz Hardy space to Musielak-Orlicz space, or to weak Musielak-Orlicz space at the critical index. Part of these results are new even for classical Hardy space of Fefferman and Stein."
                    },
                    {
                        "title": "A Spline-based Volumetric Data Modeling Framework and Its Applications",
                        "abstract": "In this dissertation, we concentrate on the challenging research issue of developing a spline-based modeling framework, which converts the conventional data (e.g., surface meshes) to tensor-product trivariate splines. This methodology can represent both boundary/volumetric geometry and real volumetric physical attributes in a compact and continuous fashion. The regular tensor-product structure enables our new developed methods to be embedded into the industry standard seamlessly. These properties make our techniques highly preferable in many physically-based applications including mechanical analysis, shape deformation and editing, virtual surgery training, etc."
                    },
                    {
                        "title": "3D Fully Convolutional Network for Vehicle Detection in Point Cloud",
                        "abstract": "2D fully convolutional network has been recently successfully applied to object detection from images. In this paper, we extend the fully convolutional network based detection techniques to 3D and apply it to point cloud data. The proposed approach is verified on the task of vehicle detection from lidar point cloud for autonomous driving. Experiments on the KITTI dataset shows a significant performance improvement over the previous point cloud based detection approaches."
                    },
                    {
                        "title": "A Survey of Spline-based Volumetric Data Modeling Framework and Its Applications",
                        "abstract": "The rapid advances in 3D scanning and acquisition techniques have given rise to the explosive increase of volumetric digital models in recent years. This dissertation systematically trailblazes a novel volumetric modeling framework to represent 3D solids. The need to explore more efficient and robust 3D modeling framework has gained the prominence. Although the traditional surface representation (e.g., triangle mesh) has many attractive properties, it is incapable of expressing the interior space and materials. Such a serious drawback overshadows many potential modeling and analysis applications. Consequently volumetric modeling techniques become the well-known solution to this problem. Nevertheless, many unsolved research issues remain when developing an efficient modeling paradigm for existing 3D models: complex geometry (fine details and extreme concaveness), arbitrary topology, heterogenous materials, large-scale data storage and processing, etc."
                    },
                    {
                        "title": "On Enhancing Ground Surface Detection from Sparse Lidar Point Cloud",
                        "abstract": "Ground surface detection in point cloud is widely used as a key module in autonomous driving systems. Different from previous approaches which are mostly developed for lidars with high beam resolution, e.g. Velodyne HDL-64, this paper proposes ground detection techniques applicable to much sparser point cloud captured by lidars with low beam resolution, e.g. Velodyne VLP-16. The approach is based on the RANSAC scheme of plane fitting. Inlier verification for plane hypotheses is enhanced by exploiting the point-wise tangent, which is a local feature available to compute regardless of the density of lidar beams. Ground surface which is not perfectly planar is fitted by multiple (specifically 4 in our implementation) disjoint plane regions. By assuming these plane regions to be rectanglar and exploiting the integral image technique, our approach approximately finds the optimal region partition and plane hypotheses under the RANSAC scheme with real-time computational complexity."
                    },
                    {
                        "title": "An explanation for the distribution characteristics of stock returns",
                        "abstract": "Observations indicate that the distributions of stock returns in financial markets usually do not conform to normal distributions, but rather exhibit characteristics of high peaks, fat tails and biases. In this work, we assume that the effects of events or information on prices obey normal distribution, while financial markets often overreact or underreact to events or information, resulting in non normal distributions of stock returns. Based on the above assumptions, we propose a reaction function for a financial market reacting to events or information, and a model based on it to describe the distribution of real stock returns. Our analysis of the returns of China Securities Index 300 (CSI 300), the Standard & Poor's 500 Index (SPX or S&P 500) and the Nikkei 225 Index (N225) at different time scales shows that financial markets often underreact to events or information with minor impacts, overreact to events or information with relatively significant impacts, and react slightly stronger to positive events or information than to negative ones. In addition, differences in financial markets and time scales of returns can also affect the shapes of the reaction functions."
                    },
                    {
                        "title": "Testing Business Cycle Theories: Evidence from the Great Recession",
                        "abstract": "Empirical business cycle studies using cross-country data usually cannot achieve causal relationships while within-country studies mostly focus on the bust period. We provide the first causal investigation into the boom period of the 1999-2010 U.S. cross-metropolitan business cycle. Using a novel research design, we show that credit expansion in private-label mortgages causes a differentially stronger boom (2000-2006) and bust (2007-2010) cycle in the house-related industries in the high net-export-growth areas. Most importantly, our unique research design enables us to perform the most comprehensive tests on theories (hypotheses) regarding the business cycle. We show that the following theories (hypotheses) cannot explain the cause of the 1999-2010 U.S. business cycle: the speculative euphoria hypothesis, the real business cycle theory, the collateral-driven credit cycle theory, the business uncertainty theory, and the extrapolative expectation theory."
                    },
                    {
                        "title": "Household Leverage Cycle Around the Great Recession",
                        "abstract": "This paper provides the first causal evidence that credit supply expansion caused the 1999-2010 U.S. business cycle mainly through the channel of household leverage (debt-to-income ratio). Specifically, induced by net export growth, credit expansion in private-label mortgages, rather than government-sponsored enterprise mortgages, causes a much stronger boom and bust cycle in household leverage in the high net-export-growth areas. In addition, such a stronger household leverage cycle creates a stronger boom and bust cycle in the local economy, including housing prices, residential construction investment, and house-related employment. Thus, our results are consistent with the credit-driven household demand channel (Mian and Sufi, 2018). Further, we show multiple pieces of evidence against the corporate channel, which is emphasized by other business cycle theories (hypotheses)."
                    },
                    {
                        "title": "Balancing Complementarity and Consistency via Delayed Activation in Incomplete Multi-view Clustering",
                        "abstract": "This paper study one challenging issue in incomplete multi-view clustering, where valuable complementary information from other views is always ignored. To be specific, we propose a framework that effectively balances Complementarity and Consistency information in Incomplete Multi-view Clustering (CoCo-IMC). Specifically, we design a dual network of delayed activation, which achieves a balance of complementarity and consistency among different views. The delayed activation could enriches the complementarity information that was ignored during consistency learning. Then, we recover the incomplete information and enhance the consistency learning by minimizing the conditional entropy and maximizing the mutual information across different views. This could be the first theoretical attempt to incorporate delayed activation into incomplete data recovery and the balance of complementarity and consistency. We have proved the effectiveness of CoCo-IMC in extensive comparative experiments with 12 state-of-the-art baselines on four publicly available datasets."
                    },
                    {
                        "title": "Involution game with specialization strategy",
                        "abstract": "Involution now refers to the phenomenon that competitors in the same field make more efforts to struggle for limited resources but get lower individual ''profit effort ratio''. In this work, we investigate the evolution of the involution game when competitors in the same field can adopt not only the strategy of making more efforts but also a specialization strategy which allows competitors to devote all their efforts to part of the competitive field. Based on the existing model, we construct the involution game with the specialization strategy and simulate the evolution of it on a square lattice under different social resource, allocation parameter (characterizing the intensity of social competition), effort and other conditions. In addition, we also conduct an theoretical analysis to further understand the underlying mechanism of our model and to avoid illusive results caused by the model settings. Our main results show that, when the total effort of the specialization strategy and the ordinary strategy is equal, the group composed of all the agents has a certain probability to choose the ordinary strategy if the allocation parameter is very large (that is to say the intensity of competition is very weak), otherwise the group will choose the specialization strategy; when the total effort of the two strategies is not equal, the proportion of the specialization strategy adoption is related to the social resource, the effort and the allocation parameter. To some extent, our study can explain why division of labor appears in human society, and provide suggestions for individuals on competition strategy selection and governments on competition policy development."
                    },
                    {
                        "title": "GrowSP: Unsupervised Semantic Segmentation of 3D Point Clouds",
                        "abstract": "We study the problem of 3D semantic segmentation from raw point clouds. Unlike existing methods which primarily rely on a large amount of human annotations for training neural networks, we propose the first purely unsupervised method, called GrowSP, to successfully identify complex semantic classes for every point in 3D scenes, without needing any type of human labels or pretrained models. The key to our approach is to discover 3D semantic elements via progressive growing of superpoints. Our method consists of three major components, 1) the feature extractor to learn per-point features from input point clouds, 2) the superpoint constructor to progressively grow the sizes of superpoints, and 3) the semantic primitive clustering module to group superpoints into semantic elements for the final semantic segmentation. We extensively evaluate our method on multiple datasets, demonstrating superior performance over all unsupervised baselines and approaching the classic fully-supervised PointNet. We hope our work could inspire more advanced methods for unsupervised 3D semantic learning."
                    },
                    {
                        "title": "Occupation times of intervals until last passage times for spectrally negative Levy processes",
                        "abstract": "In this paper, we derive the joint Laplace transforms of occupation times until its last passage times as well as its positions. Motivated by Baurdoux [2], the last times before an independent exponential variable are studied. By applying dual arguments, explicit formulas are derived in terms of new analytical identities from Loeffen et al. [12]."
                    },
                    {
                        "title": "On weighted occupation times for refracted spectrally negative L\u00e9vy processes",
                        "abstract": "For refracted spectrally negative L\\'evy processes, we identify expressions of several quantities related to Laplace transforms on their weighted occupation times until first exit times. Such quantities are expressed in terms of unique solutions to integral equations involving weight functions and scale functions for the associated spectrally negative L\\'evy processes. Previous results on refracted L\\'evy processes are recovered."
                    },
                    {
                        "title": "An excursion theoretic approach to Parisian ruin problem",
                        "abstract": "Applying excursion theory, we re-express several well studied fluctuation quantities associated to Parisian ruin problem for L\\'evy risk processes in terms of integrals with respect to excursion measure for spectrally negative L\\'evy process. We show that these new expressions reconcile with the previous results on Parisian ruin problem."
                    }
                ]
            },
            "c018c6ca-1af8-46c0-9dc0-c1a6edbc13c7": {
                "pk": "c018c6ca-1af8-46c0-9dc0-c1a6edbc13c7",
                "name": "Zhuo Wang",
                "collaborators": [
                    "Wei Zhang",
                    "Ning Liu",
                    "Jianyong Wang",
                    "Tianyu Ye",
                    "R. G. Mani",
                    "Kai Sun",
                    "Hen Fan",
                    "Vlatko Vedral"
                ],
                "domain": [
                    "Quantum Computing",
                    "Machine Learning",
                    "Rule-based Models",
                    "Superconductivity"
                ],
                "publications": [
                    {
                        "title": "Superconducting-contact-induced resistance-anomalies in the 3D topological insulator Bi2Te3",
                        "abstract": "This study examines the magnetotransport response observed in flakes of the 3D topological insulator (TI) Bi2Te3, including indium superconducting electrodes, and demonstrates two critical transitions in the magnetoresistive response with decreasing temperatures below T = 3.4K. The first transition is attributed to superconductivity in the indium electrodes, and the second transition, with a critical field exceeding the transition field of indium, is attributed to a proximity effect at the 2D planar interface of this hybrid TI/superconductor structure."
                    },
                    {
                        "title": "Scalable Rule-Based Representation Learning for Interpretable Classification",
                        "abstract": "Rule-based models, e.g., decision trees, are widely used in scenarios demanding high model interpretability for their transparent inner structures and good model expressivity. However, rule-based models are hard to optimize, especially on large data sets, due to their discrete parameters and structures. Ensemble methods and fuzzy/soft rules are commonly used to improve performance, but they sacrifice the model interpretability. To obtain both good scalability and interpretability, we propose a new classifier, named Rule-based Representation Learner (RRL), that automatically learns interpretable non-fuzzy rules for data representation and classification. To train the non-differentiable RRL effectively, we project it to a continuous space and propose a novel training method, called Gradient Grafting, that can directly optimize the discrete model using gradient descent. An improved design of logical activation functions is also devised to increase the scalability of RRL and enable it to discretize the continuous features end-to-end. Exhaustive experiments on nine small and four large data sets show that RRL outperforms the competitive interpretable approaches and can be easily adjusted to obtain a trade-off between classification accuracy and model complexity for different scenarios. Our code is available at: https://github.com/12wang3/rrl."
                    },
                    {
                        "title": "Transparent Classification with Multilayer Logical Perceptrons and Random Binarization",
                        "abstract": "Models with transparent inner structure and high classification performance are required to reduce potential risk and provide trust for users in domains like health care, finance, security, etc. However, existing models are hard to simultaneously satisfy the above two properties. In this paper, we propose a new hierarchical rule-based model for classification tasks, named Concept Rule Sets (CRS), which has both a strong expressive ability and a transparent inner structure. To address the challenge of efficiently learning the non-differentiable CRS model, we propose a novel neural network architecture, Multilayer Logical Perceptron (MLLP), which is a continuous version of CRS. Using MLLP and the Random Binarization (RB) method we proposed, we can search the discrete solution of CRS in continuous space using gradient descent and ensure the discrete CRS acts almost the same as the corresponding continuous MLLP. Experiments on 12 public data sets show that CRS outperforms the state-of-the-art approaches and the complexity of the learned CRS is close to the simple decision tree. Source code is available at https://github.com/12wang3/mllp."
                    },
                    {
                        "title": "Learning Interpretable Rules for Scalable Data Representation and Classification",
                        "abstract": "Rule-based models, e.g., decision trees, are widely used in scenarios demanding high model interpretability for their transparent inner structures and good model expressivity. However, rule-based models are hard to optimize, especially on large data sets, due to their discrete parameters and structures. Ensemble methods and fuzzy/soft rules are commonly used to improve performance, but they sacrifice the model interpretability. To obtain both good scalability and interpretability, we propose a new classifier, named Rule-based Representation Learner (RRL), that automatically learns interpretable non-fuzzy rules for data representation and classification. To train the non-differentiable RRL effectively, we project it to a continuous space and propose a novel training method, called Gradient Grafting, that can directly optimize the discrete model using gradient descent. A novel design of logical activation functions is also devised to increase the scalability of RRL and enable it to discretize the continuous features end-to-end. Exhaustive experiments on ten small and four large data sets show that RRL outperforms the competitive interpretable approaches and can be easily adjusted to obtain a trade-off between classification accuracy and model complexity for different scenarios. Our code is available at: https://github.com/12wang3/rrl."
                    },
                    {
                        "title": "Nested Quantum Error Correction Codes",
                        "abstract": "The theory of quantum error correction was established more than a decade ago as the primary tool for fighting decoherence in quantum information processing. Although great progress has already been made in this field, limited methods are available in constructing new quantum error correction codes from old codes. Here we exhibit a simple and general method to construct new quantum error correction codes by nesting certain quantum codes together. The problem of finding long quantum error correction codes is reduced to that of searching several short length quantum codes with certain properties. Our method works for all length and all distance codes, and is quite efficient to construct optimal or near optimal codes. Two main known methods in constructing new codes from old codes in quantum error-correction theory, the concatenating and pasting, can be understood in the framework of nested quantum error correction codes."
                    }
                ]
            }
        }
    },
    "2404.18928": {
        "paper_data": {
            "title": "Stylus: Automatic Adapter Selection for Diffusion Models",
            "url": "http://arxiv.org/abs/2404.18928v1",
            "arxiv_id": "2404.18928",
            "authors": [
                "Michael Luo",
                "Justin Wong",
                "Brandon Trabucco",
                "Yanping Huang",
                "Joseph E. Gonzalez",
                "Zhifeng Chen",
                "Ruslan Salakhutdinov",
                "Ion Stoica"
            ],
            "abstract": "Beyond scaling base models with more data or parameters, fine-tuned adapters provide an alternative way to generate high fidelity, custom images at reduced costs. As such, adapters have been widely adopted by open-source communities, accumulating a database of over 100K adapters-most of which are highly customized with insufficient descriptions. This paper explores the problem of matching the prompt to a set of relevant adapters, built on recent work that highlight the performance gains of composing adapters. We introduce Stylus, which efficiently selects and automatically composes task-specific adapters based on a prompt's keywords. Stylus outlines a three-stage approach that first summarizes adapters with improved descriptions and embeddings, retrieves relevant adapters, and then further assembles adapters based on prompts' keywords by checking how well they fit the prompt. To evaluate Stylus, we developed StylusDocs, a curated dataset featuring 75K adapters with pre-computed adapter embeddings. In our evaluation on popular Stable Diffusion checkpoints, Stylus achieves greater CLIP-FID Pareto efficiency and is twice as preferred, with humans and multimodal models as evaluators, over the base model. See stylus-diffusion.github.io for more.",
            "introduction": "   1 Introduction  In the evolving field of generative image models, finetuned adapters\u00a0[9, 7] have become the standard, enabling custom image creation with reduced storage requirements. This shift has spurred the growth of extensive open-source platforms that encourage communities to develop and share different adapters and model checkpoints, fueling the proliferation of creative AI art\u00a0[43, 24]. As the ecosystem expands, the number of adapters has grown to over 100K, with Low-Rank Adaptation (LoRA)\u00a0[12] emerging as the dominant finetuning approach (see Fig.\u00a03). A new paradigm has emerged where users manually select and creatively compose multiple adapters, on top of existing checkpoints, to generate high-fidelity images, moving beyond the standard approach of improving model class or scale.   Figure 2: Stylus algorithm. Stylus consists of three stages. The refiner plugs an adapter\u2019s model card through a VLM to generate textual descriptions of an adapter\u2019s task and then through an encoder to produce the corresponding text embedding. The retriever fetches candidate adapters that are relevant to the entire user prompt. Finally, the composer prunes and jointly categorizes the remaining adapters based on the prompt\u2019s tasks, which correspond to a set of keywords.   In light of performance gains, our paper explores the automatic selection of adapters based on user-provided prompts (see Fig.\u00a01). However, selecting relevant adapters presents unique challenges compared to existing retrieval-based systems, which rank relevant texts via lookup embeddings\u00a0[16]. Specifically, efficiently retrieving adapters requires converting adapters into lookup embeddings, a step made difficult with low-quality documentation or no direct access to training data\u2014a common issue on open-source platforms. Furthermore, in the context of image generation, user prompts often imply multiple highly-specific tasks. For instance, the prompt \u201ctwo dogs playing the snow\u201d suggests that there are two tasks: generating images of \u201cdogs\u201d and \u201csnow\u201d. This necessitates segmenting the prompt into various tasks (i.e. keywords) and selecting relevant adapters for each task, a requirement beyond the scope of existing retrieval-based systems\u00a0[8]. Finally, composing multiple adapters can degrade image quality, override existing concepts, and introduce unwanted biases into the model (see App.\u00a0A.4).   We propose Stylus, a system that efficiently assesses user prompts to retrieve and compose sets of highly-relevant adapters, automatically augmenting generative models to produce diverse sets of high quality images. Stylus employs a three-stage framework to address the above challenges. As shown in Fig.\u00a02, the refiner plugs in an adapter\u2019s model card, including generated images and prompts, through a multi-modal vision-language model (VLM) and a text encoder to pre-compute concise adapter descriptions as lookup embeddings. Similar to prior retrieval methods\u00a0[16], the retriever scores the relevance of each embedding against the user\u2019s entire prompt to retrieve a set of candidate adapters. Finally, the composer segments the prompt into disjoint tasks, further prunes irrelevant candidate adapters, and assigns the remaining adapters to each task. We show that the composer identifies highly-relevant adapters and avoids conceptually-similar adapters that introduce biases detrimental to image generation (\u00a7\u00a04.3). Finally, Stylus applies a binary mask to control the number of adapters per task, ensuring high image diversity by using different adapters for each image and mitigating challenges with composing many adapters.   To evaluate our system, we introduce StylusDocs, an adapter",
            "references": [
                {
                    "title": "LoRA-Flow: Dynamic LoRA Fusion for Large Language Models in Generative Tasks",
                    "abstract": "LoRA employs lightweight modules to customize large language models (LLMs) for each downstream task or domain, where different learned additional modules represent diverse skills. Combining existing LoRAs to address new tasks can enhance the reusability of learned LoRAs, particularly beneficial for tasks with limited annotated data. Most prior works on LoRA combination primarily rely on task-level weights for each involved LoRA, making different examples and tokens share the same LoRA weights. However, in generative tasks, different tokens may necessitate diverse skills to manage. Taking the Chinese math task as an example, understanding the problem description may depend more on the Chinese LoRA, while the calculation part may rely more on the math LoRA. To this end, we propose LoRA-Flow, which utilizes dynamic weights to adjust the impact of different LoRAs. The weights at each step are determined by a fusion gate with extremely few parameters, which can be learned with only 200 training examples. Experiments across six generative tasks demonstrate that our method consistently outperforms baselines with task-level fusion weights. This underscores the necessity of introducing dynamic fusion weights for LoRA combination."
                },
                {
                    "title": "LoraRetriever: Input-Aware LoRA Retrieval and Composition for Mixed Tasks in the Wild",
                    "abstract": "Low-Rank Adaptation (LoRA) provides an effective yet efficient solution for fine-tuning large language models (LLM). The modular and plug-and-play nature of LoRA enables the integration of diverse domain-specific LoRAs to enhance the capabilities of LLMs. Previous research on exploiting multiple LoRAs either focuses on specific isolated downstream tasks or fixes the selection of LoRAs during training. However, in real-world scenarios, LLMs receive diverse prompts covering different tasks, and the pool of candidate LoRAs is often dynamically updated. To bridge this gap, we propose LoraRetriever, a retrieve-then-compose framework that adaptively retrieves and composes multiple LoRAs according to the input prompts. LoraRetriever contains three main components: firstly, identifying and retrieving LoRAs relevant to the given input; secondly, formulating strategies for effectively integrating the retrieved LoRAs; and thirdly, developing efficient batch inference to accommodate heterogeneous requests. Experimental results indicate that LoraRetriever consistently outperforms the baselines, highlighting its practical effectiveness and versatility."
                },
                {
                    "title": "Retrieval-Augmented Generation for Large Language Models: A Survey",
                    "abstract": "Large Language Models (LLMs) showcase impressive capabilities but encounter challenges like hallucination, outdated knowledge, and non-transparent, untraceable reasoning processes. Retrieval-Augmented Generation (RAG) has emerged as a promising solution by incorporating knowledge from external databases. This enhances the accuracy and credibility of the generation, particularly for knowledge-intensive tasks, and allows for continuous knowledge updates and integration of domain-specific information. RAG synergistically merges LLMs' intrinsic knowledge with the vast, dynamic repositories of external databases. This comprehensive review paper offers a detailed examination of the progression of RAG paradigms, encompassing the Naive RAG, the Advanced RAG, and the Modular RAG. It meticulously scrutinizes the tripartite foundation of RAG frameworks, which includes the retrieval, the generation and the augmentation techniques. The paper highlights the state-of-the-art technologies embedded in each of these critical components, providing a profound understanding of the advancements in RAG systems. Furthermore, this paper introduces up-to-date evaluation framework and benchmark. At the end, this article delineates the challenges currently faced and points out prospective avenues for research and development."
                },
                {
                    "title": "Describing Differences in Image Sets with Natural Language",
                    "abstract": "How do two sets of images differ? Discerning set-level differences is crucial for understanding model behaviors and analyzing datasets, yet manually sifting through thousands of images is impractical. To aid in this discovery process, we explore the task of automatically describing the differences between two sets of images, which we term Set Difference Captioning. This task takes in image sets $\\mathcal{D}_{A}$ and $\\mathcal{D}_{B}$, and outputs a description that is more often true on $\\mathcal{D}_{A}$ than $\\mathcal{D}_{B}$. We outline a two-stage approach that first proposes candidate difference descriptions from image sets and then re-ranks the candidates by checking how well they can differentiate the two sets. We introduce VisDiff, which first captions the images and prompts a language model to propose candidate descriptions, then re-ranks these descriptions using CLIP. To evaluate VisDiff, we collect VisDiffBench, a dataset with 187 paired image sets with ground truth difference descriptions. We apply VisDiff to various domains, such as comparing datasets (e.g., ImageNet vs. ImageNetV2), comparing classification models (e.g., zero-shot CLIP vs. supervised ResNet), characterizing differences between generative models (e.g., StableDiffusionV1 and V2), and discovering what makes images memorable. Using VisDiff, we are able to find interesting and previously unknown differences in datasets and models, demonstrating its utility in revealing nuanced insights.11Project page available at https:/understanding-visual-datasets.github.io/VisDiff-website/."
                },
                {
                    "title": "Personalized Soups: Personalized Large Language Model Alignment via Post-hoc Parameter Merging",
                    "abstract": "While Reinforcement Learning from Human Feedback (RLHF) aligns Large Language Models (LLMs) with general, aggregate human preferences, it is suboptimal for learning diverse, individual perspectives. In this work, we study Reinforcement Learning from Personalized Human Feedback (RLPHF) problem, wherein LLMs are aligned to multiple (sometimes conflicting) preferences by modeling alignment as a Multi-Objective Reinforcement Learning (MORL) problem. Compared to strong single-objective baselines, we show that we can achieve personalized alignment by decomposing preferences into multiple dimensions. These dimensions are defined based on personalizations that are declared as desirable by the user. In this work, we show that they can be efficiently trained independently in a distributed manner and combined effectively post-hoc through parameter merging. The code is available at https://github.com/joeljang/RLPHF."
                },
                {
                    "title": "Vector Search with OpenAI Embeddings: Lucene Is All You Need",
                    "abstract": "We provide a reproducible, end-to-end demonstration of vector search with OpenAI embeddings using Lucene on the popular MS MARCO passage ranking test collection. The main goal of our work is to challenge the prevailing narrative that a dedicated vector store is necessary to take advantage of recent advances in deep neural networks as applied to search. Quite the contrary, we show that hierarchical navigable small-world network (HNSW) indexes in Lucene are adequate to provide vector search capabilities in a standard bi-encoder architecture. This suggests that, from a simple cost-benefit analysis, there does not appear to be a compelling reason to introduce a dedicated vector store into a modern \"AI stack\" for search, since such applications have already received substantial investments in existing, widely deployed infrastructure."
                },
                {
                    "title": "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis",
                    "abstract": "We present SDXL, a latent diffusion model for text-to-image synthesis. Compared to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet backbone: The increase of model parameters is mainly due to more attention blocks and a larger cross-attention context as SDXL uses a second text encoder. We design multiple novel conditioning schemes and train SDXL on multiple aspect ratios. We also introduce a refinement model which is used to improve the visual fidelity of samples generated by SDXL using a post-hoc image-to-image technique. We demonstrate that SDXL shows drastically improved performance compared the previous versions of Stable Diffusion and achieves results competitive with those of black-box state-of-the-art image generators. In the spirit of promoting open research and fostering transparency in large model training and evaluation, we provide access to code and model weights at https://github.com/Stability-AI/generative-models"
                },
                {
                    "title": "Unsupervised Compositional Concepts Discovery with Text-to-Image Generative Models",
                    "abstract": "Text-to-image generative models have enabled high-resolution image synthesis across different domains, but require users to specify the content they wish to generate. In this paper, we consider the inverse problem \u2013 given a collection of different images, can we discover the generative concepts that represent each image? We present an unsupervised approach to discover generative concepts from a collection of images, disentangling different art styles in paintings, objects, and lighting from kitchen scenes, and discovering image classes given ImageNet images. We show how such generative concepts can accurately represent the content of images, be recombined and composed to generate new artistic and hybrid images, and be further used as a representation for downstream classification tasks."
                },
                {
                    "title": "AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedback",
                    "abstract": "Large language models (LLMs) such as ChatGPT have seen widespread adoption due to their strong instruction-following abilities. Developing these LLMs involves a complex yet poorly understood workflow requiring training with human feedback. Replicating and understanding this instruction-following requires tackling three major challenges: the high cost of data collection, the lack of trustworthy evaluation, and the absence of reference method implementations. We address these challenges with AlpacaFarm, a simulator that enables research and development for learning from feedback at a low cost. First, we design LLM prompts to simulate human feedback that are 50x cheaper than crowdworkers and display high agreement with humans. Second, we propose an automatic evaluation and validate it against human instructions obtained on real-world interactions. Third, we contribute reference implementations for several methods (PPO, DPO, best-of-n, expert iteration, and more) that learn from pairwise feedback. Finally, as an end-to-end validation of AlpacaFarm, we train and evaluate eleven models on 10k pairs of real human feedback and show that rankings of models trained in AlpacaFarm match rankings of models trained on human data. As a demonstration of the research possible in AlpacaFarm, we find that methods that use a reward model can substantially improve over supervised fine-tuning and that our reference PPO implementation leads to a +10% improvement in win-rate against Davinci003. We release all components of AlpacaFarm at https://github.com/tatsu-lab/alpaca_farm."
                },
                {
                    "title": "Tree of Thoughts: Deliberate Problem Solving with Large Language Models",
                    "abstract": "Language models are increasingly being deployed for general problem solving across a wide range of tasks, but are still confined to token-level, left-to-right decision-making processes during inference. This means they can fall short in tasks that require exploration, strategic lookahead, or where initial decisions play a pivotal role. To surmount these challenges, we introduce a new framework for language model inference, Tree of Thoughts (ToT), which generalizes over the popular Chain of Thought approach to prompting language models, and enables exploration over coherent units of text (thoughts) that serve as intermediate steps toward problem solving. ToT allows LMs to perform deliberate decision making by considering multiple different reasoning paths and self-evaluating choices to decide the next course of action, as well as looking ahead or backtracking when necessary to make global choices. Our experiments show that ToT significantly enhances language models' problem-solving abilities on three novel tasks requiring non-trivial planning or search: Game of 24, Creative Writing, and Mini Crosswords. For instance, in Game of 24, while GPT-4 with chain-of-thought prompting only solved 4% of tasks, our method achieved a success rate of 74%. Code repo with all prompts: https://github.com/princeton-nlp/tree-of-thought-llm."
                },
                {
                    "title": "AdapterSoup: Weight Averaging to Improve Generalization of Pretrained Language Models",
                    "abstract": "Pretrained language models (PLMs) are trained on massive corpora, but often need to specialize to specific domains. A parameter-efficient adaptation method suggests training an adapter for each domain on the task of language modeling. This leads to good in-domain scores but can be impractical for domain- or resource-restricted settings. A solution is to use a related-domain adapter for the novel domain at test time. In this paper, we introduce AdapterSoup, an approach that performs weight-space averaging of adapters trained on different domains. Our approach is embarrassingly parallel: first, we train a set of domain-specific adapters; then, for each novel domain, we determine which adapters should be averaged at test time. We present extensive experiments showing that AdapterSoup consistently improves performance to new domains without extra training. We also explore weight averaging of adapters trained on the same domain with different hyper-parameters, and show that it preserves the performance of a PLM on new domains while obtaining strong in-domain results. We explore various approaches for choosing which adapters to combine, such as text clustering and semantic similarity. We find that using clustering leads to the most competitive results on novel domains."
                },
                {
                    "title": "DPM-Solver++: Fast Solver for Guided Sampling of Diffusion Probabilistic Models",
                    "abstract": "Diffusion probabilistic models (DPMs) have achieved impressive success in high-resolution image synthesis, especially in recent large-scale text-to-image generation applications. An essential technique for improving the sample quality of DPMs is guided sampling, which usually needs a large guidance scale to obtain the best sample quality. The commonly-used fast sampler for guided sampling is DDIM, a first-order diffusion ODE solver that generally needs 100 to 250 steps for high-quality samples. Although recent works propose dedicated high-order solvers and achieve a further speedup for sampling without guidance, their effectiveness for guided sampling has not been well-tested before. In this work, we demonstrate that previous high-order fast samplers suffer from instability issues, and they even become slower than DDIM when the guidance scale grows large. To further speed up guided sampling, we propose DPM-Solver++, a high-order solver for the guided sampling of DPMs. DPM-Solver++ solves the diffusion ODE with the data prediction model and adopts thresholding methods to keep the solution matches training data distribution. We further propose a multistep variant of DPM-Solver++ to address the instability issue by reducing the effective step size. Experiments show that DPM-Solver++ can generate high-quality samples within only 15 to 20 steps for guided sampling by pixel-space and latent-space DPMs."
                },
                {
                    "title": "An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion",
                    "abstract": "Text-to-image models offer unprecedented freedom to guide creation through natural language. Yet, it is unclear how such freedom can be exercised to generate images of specific unique concepts, modify their appearance, or compose them in new roles and novel scenes. In other words, we ask: how can we use language-guided models to turn our cat into a painting, or imagine a new product based on our favorite toy? Here we present a simple approach that allows such creative freedom. Using only 3-5 images of a user-provided concept, like an object or a style, we learn to represent it through new\"words\"in the embedding space of a frozen text-to-image model. These\"words\"can be composed into natural language sentences, guiding personalized creation in an intuitive way. Notably, we find evidence that a single word embedding is sufficient for capturing unique and varied concepts. We compare our approach to a wide range of baselines, and demonstrate that it can more faithfully portray the concepts across a range of applications and tasks. Our code, data and new words will be available at: https://textual-inversion.github.io"
                },
                {
                    "title": "Scaling Autoregressive Models for Content-Rich Text-to-Image Generation",
                    "abstract": "We present the Pathways Autoregressive Text-to-Image (Parti) model, which generates high-fidelity photorealistic images and supports content-rich synthesis involving complex compositions and world knowledge. Parti treats text-to-image generation as a sequence-to-sequence modeling problem, akin to machine translation, with sequences of image tokens as the target outputs rather than text tokens in another language. This strategy can naturally tap into the rich body of prior work on large language models, which have seen continued advances in capabilities and performance through scaling data and model sizes. Our approach is simple: First, Parti uses a Transformer-based image tokenizer, ViT-VQGAN, to encode images as sequences of discrete tokens. Second, we achieve consistent quality improvements by scaling the encoder-decoder Transformer model up to 20B parameters, with a new state-of-the-art zero-shot FID score of 7.23 and finetuned FID score of 3.22 on MS-COCO. Our detailed analysis on Localized Narratives as well as PartiPrompts (P2), a new holistic benchmark of over 1600 English prompts, demonstrate the effectiveness of Parti across a wide variety of categories and difficulty aspects. We also explore and highlight limitations of our models in order to define and exemplify key areas of focus for further improvements. See https://parti.research.google/ for high-resolution images."
                },
                {
                    "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                    "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Efficient Test Time Adapter Ensembling for Low-resource Language Varieties",
                    "abstract": "Adapters are light-weight modules that allow parameter-efficient fine-tuning of pretrained models. Specialized language and task adapters have recently been proposed to facilitate cross-lingual transfer of multilingual pretrained models (Pfeiffer et al., 2020b). However, this approach requires training a separate language adapter for every language one wishes to support, which can be impractical for languages with limited data. An intuitive solution is to use a related language adapter for the new language variety, but we observe that this solution can lead to sub-optimal performance. In this paper, we aim to improve the robustness of language adapters to uncovered languages without training new adapters. We find that ensembling multiple existing language adapters makes the fine-tuned model significantly more robust to other language varieties not included in these adapters. Building upon this observation, we propose Entropy Minimized Ensemble of Adapters (EMEA), a method that optimizes the ensemble weights of the pretrained language adapters for each test sentence by minimizing the entropy of its predictions. Experiments on three diverse groups of language varieties show that our method leads to significant improvements on both named entity recognition and part-of-speech tagging across all languages."
                },
                {
                    "title": "CLIPScore: A Reference-free Evaluation Metric for Image Captioning",
                    "abstract": "Image captioning has conventionally relied on reference-based automatic evaluations, where machine captions are compared against captions written by humans. This is in contrast to the reference-free manner in which humans assess caption quality. In this paper, we report the surprising empirical finding that CLIP (Radford et al., 2021), a cross-modal model pretrained on 400M image+caption pairs from the web, can be used for robust automatic evaluation of image captioning without the need for references. Experiments spanning several corpora demonstrate that our new reference-free metric, CLIPScore, achieves the highest correlation with human judgements, outperforming existing reference-based metrics like CIDEr and SPICE. Information gain experiments demonstrate that CLIPScore, with its tight focus on image-text compatibility, is complementary to existing reference-based metrics that emphasize text-text similarities. Thus, we also present a reference-augmented version, RefCLIPScore, which achieves even higher correlation. Beyond literal description tasks, several case studies reveal domains where CLIPScore performs well (clip-art images, alt-text rating), but also where it is relatively weaker in comparison to reference-based metrics, e.g., news captions that require richer contextual knowledge."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks",
                    "abstract": "Large pre-trained language models have been shown to store factual knowledge in their parameters, and achieve state-of-the-art results when fine-tuned on downstream NLP tasks. However, their ability to access and precisely manipulate knowledge is still limited, and hence on knowledge-intensive tasks, their performance lags behind task-specific architectures. Additionally, providing provenance for their decisions and updating their world knowledge remain open research problems. Pre-trained models with a differentiable access mechanism to explicit non-parametric memory can overcome this issue, but have so far been only investigated for extractive downstream tasks. We explore a general-purpose fine-tuning recipe for retrieval-augmented generation (RAG) -- models which combine pre-trained parametric and non-parametric memory for language generation. We introduce RAG models where the parametric memory is a pre-trained seq2seq model and the non-parametric memory is a dense vector index of Wikipedia, accessed with a pre-trained neural retriever. We compare two RAG formulations, one which conditions on the same retrieved passages across the whole generated sequence, the other can use different passages per token. We fine-tune and evaluate our models on a wide range of knowledge-intensive NLP tasks and set the state-of-the-art on three open domain QA tasks, outperforming parametric seq2seq models and task-specific retrieve-and-extract architectures. For language generation tasks, we find that RAG models generate more specific, diverse and factual language than a state-of-the-art parametric-only seq2seq baseline."
                },
                {
                    "title": "Reliable Fidelity and Diversity Metrics for Generative Models",
                    "abstract": "Devising indicative evaluation metrics for the image generation task remains an open problem. The most widely used metric for measuring the similarity between real and generated images has been the Frechet Inception Distance (FID) score. Because it does not differentiate the fidelity and diversity aspects of the generated images, recent papers have introduced variants of precision and recall metrics to diagnose those properties separately. In this paper, we show that even the latest version of the precision and recall metrics are not reliable yet. For example, they fail to detect the match between two identical distributions, they are not robust against outliers, and the evaluation hyperparameters are selected arbitrarily. We propose density and coverage metrics that solve the above issues. We analytically and experimentally show that density and coverage provide more interpretable and reliable signals for practitioners than the existing metrics. Code: this https URL."
                },
                {
                    "title": "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
                    "abstract": "BERT (Devlin et al., 2018) and RoBERTa (Liu et al., 2019) has set a new state-of-the-art performance on sentence-pair regression tasks like semantic textual similarity (STS). However, it requires that both sentences are fed into the network, which causes a massive computational overhead: Finding the most similar pair in a collection of 10,000 sentences requires about 50 million inference computations (~65 hours) with BERT. The construction of BERT makes it unsuitable for semantic similarity search as well as for unsupervised tasks like clustering. In this publication, we present Sentence-BERT (SBERT), a modification of the pretrained BERT network that use siamese and triplet network structures to derive semantically meaningful sentence embeddings that can be compared using cosine-similarity. This reduces the effort for finding the most similar pair from 65 hours with BERT / RoBERTa to about 5 seconds with SBERT, while maintaining the accuracy from BERT. We evaluate SBERT and SRoBERTa on common STS tasks and transfer learning tasks, where it outperforms other state-of-the-art sentence embeddings methods."
                },
                {
                    "title": "Assessing Generative Models via Precision and Recall",
                    "abstract": "Recent advances in generative modeling have led to an increased interest in the study of statistical divergences as means of model comparison. Commonly used evaluation methods, such as the Frechet Inception Distance (FID), correlate well with the perceived quality of samples and are sensitive to mode dropping. However, these metrics are unable to distinguish between different failure cases since they only yield one-dimensional scores. We propose a novel definition of precision and recall for distributions which disentangles the divergence into two separate dimensions. The proposed notion is intuitive, retains desirable properties, and naturally leads to an efficient algorithm that can be used to evaluate generative models. We relate this notion to total variation as well as to recent evaluation metrics such as Inception Score and FID. To demonstrate the practical utility of the proposed approach we perform an empirical study on several variants of Generative Adversarial Networks and Variational Autoencoders. In an extensive set of experiments we show that the proposed metric is able to disentangle the quality of generated samples from the coverage of the target distribution."
                },
                {
                    "title": "GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium",
                    "abstract": "Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the \"Frechet Inception Distance\" (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark."
                },
                {
                    "title": "Rethinking the Inception Architecture for Computer Vision",
                    "abstract": "Convolutional networks are at the core of most state of-the-art computer vision solutions for a wide variety of tasks. Since 2014 very deep convolutional networks started to become mainstream, yielding substantial gains in various benchmarks. Although increased model size and computational cost tend to translate to immediate quality gains for most tasks (as long as enough labeled data is provided for training), computational efficiency and low parameter count are still enabling factors for various use cases such as mobile vision and big-data scenarios. Here we are exploring ways to scale up networks in ways that aim at utilizing the added computation as efficiently as possible by suitably factorized convolutions and aggressive regularization. We benchmark our methods on the ILSVRC 2012 classification challenge validation set demonstrate substantial gains over the state of the art: 21:2% top-1 and 5:6% top-5 error for single frame evaluation using a network with a computational cost of 5 billion multiply-adds per inference and with using less than 25 million parameters. With an ensemble of 4 models and multi-crop evaluation, we report 3:5% top-5 error and 17:3% top-1 error on the validation set and 3:6% top-5 error on the official test set."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI",
                "cs.CL",
                "cs.GR",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we automatically select and compose relevant adapters for generative image models based on user-provided prompts to enhance image quality and diversity?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of generative image models, as it addresses the growing complexity of user prompts and the need for high-quality image generation. By automating the selection and composition of adapters, we can significantly improve the efficiency and effectiveness of creative AI art generation. This research could lead to practical applications in various domains, such as digital art, advertising, and content creation, while also fostering further exploration of multi-modal systems and their integration with generative models.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to convert adapters into lookup embeddings efficiently, which is complicated by low-quality documentation and lack of access to training data on open-source platforms. Additionally, user prompts often imply multiple specific tasks, requiring segmentation into keywords and the selection of relevant adapters for each task. Existing retrieval-based systems do not adequately address this complexity, and composing multiple adapters can lead to degraded image quality, concept overrides, and the introduction of unwanted biases, making the task technically and practically challenging.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on retrieval-based systems that rank relevant texts without considering the multi-task nature of user prompts in image generation. Limitations in existing solutions include inadequate handling of low-quality documentation and the inability to segment prompts into disjoint tasks for adapter selection. Our approach differs by employing a three-stage framework that refines adapter descriptions, retrieves relevant adapters based on user prompts, and composes them while mitigating biases and ensuring high image diversity, addressing gaps left by prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Stylus, consists of three key components: (1) a refiner that generates concise adapter descriptions as lookup embeddings using a multi-modal vision-language model (VLM) and a text encoder; (2) a retriever that scores the relevance of these embeddings against the user\u2019s prompt to identify candidate adapters; and (3) a composer that segments the prompt into tasks, prunes irrelevant adapters, and assigns the remaining adapters to each task. We will evaluate our system using StylusDocs, focusing on metrics such as image quality and diversity"
            }
        },
        "author_data": {
            "174489ae-6267-445d-b9e9-d48acf33ecf9": {
                "pk": "174489ae-6267-445d-b9e9-d48acf33ecf9",
                "name": "Michael Luo",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "6ac9f3cc-b67e-45ca-8c19-6f09e69a0aaf": {
                "pk": "6ac9f3cc-b67e-45ca-8c19-6f09e69a0aaf",
                "name": "Justin Wong",
                "collaborators": [
                    "Gabriel Ryan",
                    "Jianan Yao",
                    "Ronghui Gu",
                    "Suman Jana",
                    "Dominik Damjakob",
                    "Danny Ho",
                    "Luiz Fernando Capretz",
                    "Yury Orlovskiy",
                    "Michael Luo",
                    "Sanjit A. Seshia"
                ],
                "domain": [
                    "Meta-Learning",
                    "Causal Inference",
                    "Program Verification",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "A Meta Learning Approach to Discerning Causal Graph Structure",
                        "abstract": "We explore the usage of meta-learning to derive the causal direction between variables by optimizing over a measure of distribution simplicity. We incorporate a stochastic graph representation which includes latent variables and allows for more generalizability and graph structure expression. Our model is able to learn causal direction indicators for complex graph structures despite effects of latent confounders. Further, we explore robustness of our method with respect to violations of our distributional assumptions and data scarcity. Our model is particularly robust to modest data scarcity, but is less robust to distributional changes. By interpreting the model predictions as stochastic events, we propose a simple ensemble method classifier to reduce the outcome variability as an average of biased events. This methodology demonstrates ability to infer the existence as well as the direction of a causal relationship between data distributions."
                    },
                    {
                        "title": "A Neuro-Fuzzy Method to Improving Backfiring Conversion Ratios",
                        "abstract": "Software project estimation is crucial aspect in delivering software on time and on budget. Software size is an important metric in determining the effort, cost, and productivity. Today, source lines of code and function point are the most used sizing metrics. Backfiring is a well-known technique for converting between function points and source lines of code. However when backfiring is used, there is a high margin of error. This study introduces a method to improve the accuracy of backfiring. Intelligent systems have been used in software prediction models to improve performance over traditional techniques. For this reason, a hybrid Neuro-Fuzzy is used because it takes advantages of the neural networks learning and fuzzy logic human-like reasoning. This paper describes an improved backfiring technique which uses Neuro-Fuzzy and compares the new method against the default conversion ratios currently used by software practitioners."
                    },
                    {
                        "title": "SimpleStrat: Diversifying Language Model Generation with Stratification",
                        "abstract": "Generating diverse responses from large language models (LLMs) is crucial for applications such as planning/search and synthetic data generation, where diversity provides distinct answers across generations. Prior approaches rely on increasing temperature to increase diversity. However, contrary to popular belief, we show not only does this approach produce lower quality individual generations as temperature increases, but it depends on model's next-token probabilities being similar to the true distribution of answers. We propose SimpleStrat, an alternative approach that uses the language model itself to partition the space into strata. At inference, a random stratum is selected and a sample drawn from within the strata. To measure diversity, we introduce CoverageQA, a dataset of underspecified questions with multiple equally plausible answers, and assess diversity by measuring KL Divergence between the output distribution and uniform distribution over valid ground truth answers. As computing probability per response/solution for proprietary models is infeasible, we measure recall on ground truth solutions. Our evaluation show using SimpleStrat achieves higher recall by 0.05 compared to GPT-4o and 0.36 average reduction in KL Divergence compared to Llama 3."
                    },
                    {
                        "title": "CLN2INV: Learning Loop Invariants with Continuous Logic Networks",
                        "abstract": "Program verification offers a framework for ensuring program correctness and therefore systematically eliminating different classes of bugs. Inferring loop invariants is one of the main challenges behind automated verification of real-world programs which often contain many loops. In this paper, we present Continuous Logic Network (CLN), a novel neural architecture for automatically learning loop invariants directly from program execution traces. Unlike existing neural networks, CLNs can learn precise and explicit representations of formulas in Satisfiability Modulo Theories (SMT) for loop invariants from program execution traces. We develop a new sound and complete semantic mapping for assigning SMT formulas to continuous truth values that allows CLNs to be trained efficiently. We use CLNs to implement a new inference system for loop invariants, CLN2INV, that significantly outperforms existing approaches on the popular Code2Inv dataset. CLN2INV is the first tool to solve all 124 theoretically solvable problems in the Code2Inv dataset. Moreover, CLN2INV takes only 1.1 seconds on average for each problem, which is 40 times faster than existing approaches. We further demonstrate that CLN2INV can even learn 12 significantly more complex loop invariants than the ones required for the Code2Inv dataset."
                    },
                    {
                        "title": "Learning Nonlinear Loop Invariants with Gated Continuous Logic Networks (Extended Version)",
                        "abstract": "Verifying real-world programs often requires inferring loop invariants with nonlinear constraints. This is especially true in programs that perform many numerical operations, such as control systems for avionics or industrial plants. Recently, data-driven methods for loop invariant inference have shown promise, especially on linear invariants. However, applying data-driven inference to nonlinear loop invariants is challenging due to the large numbers of and magnitudes of high-order terms, the potential for overfitting on a small number of samples, and the large space of possible inequality bounds.   In this paper, we introduce a new neural architecture for general SMT learning, the Gated Continuous Logic Network (G-CLN), and apply it to nonlinear loop invariant learning. G-CLNs extend the Continuous Logic Network (CLN) architecture with gating units and dropout, which allow the model to robustly learn general invariants over large numbers of terms. To address overfitting that arises from finite program sampling, we introduce fractional sampling---a sound relaxation of loop semantics to continuous functions that facilitates unbounded sampling on real domain. We additionally design a new CLN activation function, the Piecewise Biased Quadratic Unit (PBQU), for naturally learning tight inequality bounds.   We incorporate these methods into a nonlinear loop invariant inference system that can learn general nonlinear loop invariants. We evaluate our system on a benchmark of nonlinear loop invariants and show it solves 26 out of 27 problems, 3 more than prior work, with an average runtime of 53.3 seconds. We further demonstrate the generic learning ability of G-CLNs by solving all 124 problems in the linear Code2Inv benchmark. We also perform a quantitative stability evaluation and show G-CLNs have a convergence rate of $97.5\\%$ on quadratic problems, a $39.2\\%$ improvement over CLN models."
                    }
                ]
            },
            "16767dfc-d1ab-4aba-af2d-ee558b9e3506": {
                "pk": "16767dfc-d1ab-4aba-af2d-ee558b9e3506",
                "name": "Brandon Trabucco",
                "collaborators": [
                    "Ruslan Salakhutdinov",
                    "Kyle Doherty",
                    "Max Gurinas",
                    "Sergey Levine",
                    "Glen Berseth",
                    "Aviral Kumar",
                    "Xinyang Geng",
                    "Albert Qu",
                    "Simon Li",
                    "Ganeshkumar Ashokavardhanan"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Model-Based Optimization",
                    "Computer Vision",
                    "Program Synthesis"
                ],
                "publications": [
                    {
                        "title": "Inferring the Optimal Policy using Markov Chain Monte Carlo",
                        "abstract": "This paper investigates methods for estimating the optimal stochastic control policy for a Markov Decision Process with unknown transition dynamics and an unknown reward function. This form of model-free reinforcement learning comprises many real world systems such as playing video games, simulated control tasks, and real robot locomotion. Existing methods for estimating the optimal stochastic control policy rely on high variance estimates of the policy descent. However, these methods are not guaranteed to find the optimal stochastic policy, and the high variance gradient estimates make convergence unstable. In order to resolve these problems, we propose a technique using Markov Chain Monte Carlo to generate samples from the posterior distribution of the parameters conditioned on being optimal. Our method provably converges to the globally optimal stochastic policy, and empirically similar variance compared to the policy gradient."
                    },
                    {
                        "title": "AnyMorph: Learning Transferable Polices By Inferring Agent Morphology",
                        "abstract": "The prototypical approach to reinforcement learning involves training policies tailored to a particular agent from scratch for every new morphology. Recent work aims to eliminate the re-training of policies by investigating whether a morphology-agnostic policy, trained on a diverse set of agents with similar task objectives, can be transferred to new agents with unseen morphologies without re-training. This is a challenging problem that required previous approaches to use hand-designed descriptions of the new agent's morphology. Instead of hand-designing this description, we propose a data-driven method that learns a representation of morphology directly from the reinforcement learning objective. Ours is the first reinforcement learning algorithm that can train a policy to generalize to new agent morphologies without requiring a description of the agent's morphology in advance. We evaluate our approach on the standard benchmark for agent-agnostic control, and improve over the current state of the art in zero-shot generalization to new agents. Importantly, our method attains good performance without an explicit description of morphology."
                    },
                    {
                        "title": "Effective Data Augmentation With Diffusion Models",
                        "abstract": "Data augmentation is one of the most prevalent tools in deep learning, underpinning many recent advances, including those from classification, generative models, and representation learning. The standard approach to data augmentation combines simple transformations like rotations and flips to generate new images from existing ones. However, these new images lack diversity along key semantic axes present in the data. Current augmentations cannot alter the high-level semantic attributes, such as animal species present in a scene, to enhance the diversity of data. We address the lack of diversity in data augmentation with image-to-image transformations parameterized by pre-trained text-to-image diffusion models. Our method edits images to change their semantics using an off-the-shelf diffusion model, and generalizes to novel visual concepts from a few labelled examples. We evaluate our approach on few-shot image classification tasks, and on a real-world weed recognition task, and observe an improvement in accuracy in tested domains."
                    },
                    {
                        "title": "Understanding Visual Concepts Across Models",
                        "abstract": "Large multimodal models such as Stable Diffusion can generate, detect, and classify new visual concepts after fine-tuning just a single word embedding. Do models learn similar words for the same concepts (i.e. <orange-cat> = orange + cat)? We conduct a large-scale analysis on three state-of-the-art models in text-to-image generation, open-set object detection, and zero-shot classification, and find that new word embeddings are model-specific and non-transferable. Across 4,800 new embeddings trained for 40 diverse visual concepts on four standard datasets, we find perturbations within an $\\epsilon$-ball to any prior embedding that generate, detect, and classify an arbitrary concept. When these new embeddings are spliced into new models, fine-tuning that targets the original model is lost. We show popular soft prompt-tuning approaches find these perturbative solutions when applied to visual concept learning tasks, and embeddings for visual concepts are not transferable. Code for reproducing our work is available at: https://visual-words.github.io."
                    },
                    {
                        "title": "A Simple Approach for Visual Rearrangement: 3D Mapping and Semantic Search",
                        "abstract": "Physically rearranging objects is an important capability for embodied agents. Visual room rearrangement evaluates an agent's ability to rearrange objects in a room to a desired goal based solely on visual input. We propose a simple yet effective method for this problem: (1) search for and map which objects need to be rearranged, and (2) rearrange each object until the task is complete. Our approach consists of an off-the-shelf semantic segmentation model, voxel-based semantic map, and semantic search policy to efficiently find objects that need to be rearranged. On the AI2-THOR Rearrangement Challenge, our method improves on current state-of-the-art end-to-end reinforcement learning-based methods that learn visual rearrangement policies from 0.53% correct rearrangement to 16.56%, using only 2.7% as many samples from the environment."
                    },
                    {
                        "title": "Conservative Objective Models for Effective Offline Model-Based Optimization",
                        "abstract": "Computational design problems arise in a number of settings, from synthetic biology to computer architectures. In this paper, we aim to solve data-driven model-based optimization (MBO) problems, where the goal is to find a design input that maximizes an unknown objective function provided access to only a static dataset of prior experiments. Such data-driven optimization procedures are the only practical methods in many real-world domains where active data collection is expensive (e.g., when optimizing over proteins) or dangerous (e.g., when optimizing over aircraft designs). Typical methods for MBO that optimize the design against a learned model suffer from distributional shift: it is easy to find a design that \"fools\" the model into predicting a high value. To overcome this, we propose conservative objective models (COMs), a method that learns a model of the objective function that lower bounds the actual value of the ground-truth objective on out-of-distribution inputs, and uses it for optimization. Structurally, COMs resemble adversarial training methods used to overcome adversarial examples. COMs are simple to implement and outperform a number of existing methods on a wide range of MBO problems, including optimizing protein sequences, robot morphologies, neural network weights, and superconducting materials."
                    },
                    {
                        "title": "Design-Bench: Benchmarks for Data-Driven Offline Model-Based Optimization",
                        "abstract": "Black-box model-based optimization (MBO) problems, where the goal is to find a design input that maximizes an unknown objective function, are ubiquitous in a wide range of domains, such as the design of proteins, DNA sequences, aircraft, and robots. Solving model-based optimization problems typically requires actively querying the unknown objective function on design proposals, which means physically building the candidate molecule, aircraft, or robot, testing it, and storing the result. This process can be expensive and time consuming, and one might instead prefer to optimize for the best design using only the data one already has. This setting -- called offline MBO -- poses substantial and different algorithmic challenges than more commonly studied online techniques. A number of recent works have demonstrated success with offline MBO for high-dimensional optimization problems using high-capacity deep neural networks. However, the lack of standardized benchmarks in this emerging field is making progress difficult to track. To address this, we present Design-Bench, a benchmark for offline MBO with a unified evaluation protocol and reference implementations of recent methods. Our benchmark includes a suite of diverse and realistic tasks derived from real-world optimization problems in biology, materials science, and robotics that present distinct challenges for offline MBO. Our benchmark and reference implementations are released at github.com/rail-berkeley/design-bench and github.com/rail-berkeley/design-baselines."
                    },
                    {
                        "title": "Discovering Non-monotonic Autoregressive Orderings with Variational Inference",
                        "abstract": "The predominant approach for language modeling is to process sequences from left to right, but this eliminates a source of information: the order by which the sequence was generated. One strategy to recover this information is to decode both the content and ordering of tokens. Existing approaches supervise content and ordering by designing problem-specific loss functions and pre-training with an ordering pre-selected. Other recent works use iterative search to discover problem-specific orderings for training, but suffer from high time complexity and cannot be efficiently parallelized. We address these limitations with an unsupervised parallelizable learner that discovers high-quality generation orders purely from training data -- no domain knowledge required. The learner contains an encoder network and decoder language model that perform variational inference with autoregressive orders (represented as permutation matrices) as latent variables. The corresponding ELBO is not differentiable, so we develop a practical algorithm for end-to-end optimization using policy gradients. We implement the encoder as a Transformer with non-causal attention that outputs permutations in one forward pass. Permutations then serve as target generation orders for training an insertion-based Transformer language model. Empirical results in language modeling tasks demonstrate that our method is context-aware and discovers orderings that are competitive with or even better than fixed orders."
                    },
                    {
                        "title": "Leafy Spurge Dataset: Real-world Weed Classification Within Aerial Drone Imagery",
                        "abstract": "Invasive plant species are detrimental to the ecology of both agricultural and wildland areas. Euphorbia esula, or leafy spurge, is one such plant that has spread through much of North America from Eastern Europe. When paired with contemporary computer vision systems, unmanned aerial vehicles, or drones, offer the means to track expansion of problem plants, such as leafy spurge, and improve chances of controlling these weeds. We gathered a dataset of leafy spurge presence and absence in grasslands of western Montana, USA, then surveyed these areas with a commercial drone. We trained image classifiers on these data, and our best performing model, a pre-trained DINOv2 vision transformer, identified leafy spurge with 0.84 accuracy (test set). This result indicates that classification of leafy spurge is tractable, but not solved. We release this unique dataset of labelled and unlabelled, aerial drone imagery for the machine learning community to explore. Improving classification performance of leafy spurge would benefit the fields of ecology, conservation, and remote sensing alike. Code and data are available at our website: leafy-spurge-dataset.github.io."
                    },
                    {
                        "title": "Synthetic Datasets for Neural Program Synthesis",
                        "abstract": "The goal of program synthesis is to automatically generate programs in a particular language from corresponding specifications, e.g. input-output behavior. Many current approaches achieve impressive results after training on randomly generated I/O examples in limited domain-specific languages (DSLs), as with string transformations in RobustFill. However, we empirically discover that applying test input generation techniques for languages with control flow and rich input space causes deep networks to generalize poorly to certain data distributions; to correct this, we propose a new methodology for controlling and evaluating the bias of synthetic data distributions over both programs and specifications. We demonstrate, using the Karel DSL and a small Calculator DSL, that training deep networks on these distributions leads to improved cross-distribution generalization performance."
                    },
                    {
                        "title": "Inter-Level Cooperation in Hierarchical Reinforcement Learning",
                        "abstract": "Hierarchies of temporally decoupled policies present a promising approach for enabling structured exploration in complex long-term planning problems. To fully achieve this approach an end-to-end training paradigm is needed. However, training these multi-level policies has had limited success due to challenges arising from interactions between the goal-assigning and goal-achieving levels within a hierarchy. In this article, we consider the policy optimization process as a multi-agent process. This allows us to draw on connections between communication and cooperation in multi-agent RL, and demonstrate the benefits of increased cooperation between sub-policies on the training performance of the overall policy. We introduce a simple yet effective technique for inducing inter-level cooperation by modifying the objective function and subsequent gradients of higher-level policies. Experimental results on a wide variety of simulated robotics and traffic control tasks demonstrate that inducing cooperation results in stronger performing policies and increased sample efficiency on a set of difficult long time horizon tasks. We also find that goal-conditioned policies trained using our method display better transfer to new tasks, highlighting the benefits of our method in learning task-agnostic lower-level behaviors. Videos and code are available at: https://sites.google.com/berkeley.edu/cooperative-hrl."
                    }
                ]
            },
            "ea51e80b-d9b1-43b2-a2f0-60ea0ba46793": {
                "pk": "ea51e80b-d9b1-43b2-a2f0-60ea0ba46793",
                "name": "Yanping Huang",
                "collaborators": [
                    "Gang Li",
                    "Manqi Ruan",
                    "Francis Iannacci",
                    "Sai Zhang",
                    "Eddie Yan",
                    "Dong Chen",
                    "Xue Wang",
                    "Mi Tan",
                    "Bingling Cen",
                    "Esteban Real"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Deep Learning",
                    "Neural Architecture Search",
                    "Lifelong Learning"
                ],
                "publications": [
                    {
                        "title": "Learning Efficient Representations for Reinforcement Learning",
                        "abstract": "Markov decision processes (MDPs) are a well studied framework for solving sequential decision making problems under uncertainty. Exact methods for solving MDPs based on dynamic programming such as policy iteration and value iteration are effective on small problems. In problems with a large discrete state space or with continuous state spaces, a compact representation is essential for providing an efficient approximation solutions to MDPs. Commonly used approximation algorithms involving constructing basis functions for projecting the value function onto a low dimensional subspace, and building a factored or hierarchical graphical model to decompose the transition and reward functions. However, hand-coding a good compact representation for a given reinforcement learning (RL) task can be quite difficult and time consuming. Recent approaches have attempted to automatically discover efficient representations for RL.   In this thesis proposal, we discuss the problems of automatically constructing structured kernel for kernel based RL, a popular approach to learning non-parametric approximations for value function. We explore a space of kernel structures which are built compositionally from base kernels using a context-free grammar. We examine a greedy algorithm for searching over the structure space. To demonstrate how the learned structure can represent and approximate the original RL problem in terms of compactness and efficiency, we plan to evaluate our method on a synthetic problem and compare it to other RL baselines."
                    },
                    {
                        "title": "ChirpCast: Data Transmission via Audio",
                        "abstract": "In this paper we present ChirpCast, a system for broadcasting network access keys to laptops ultrasonically. This work explores several modulation techniques for sending and receiving data using sound waves through commodity speakers and built-in laptop microphones. Requiring only that laptop users run a small application, the system successfully provides robust room-specific broadcasting at data rates of 200 bits/second."
                    },
                    {
                        "title": "Partitioning Large Scale Deep Belief Networks Using Dropout",
                        "abstract": "Deep learning methods have shown great promise in many practical applications, ranging from speech recognition, visual object recognition, to text processing. However, most of the current deep learning methods suffer from scalability problems for large-scale applications, forcing researchers or users to focus on small-scale problems with fewer parameters.   In this paper, we consider a well-known machine learning model, deep belief networks (DBNs) that have yielded impressive classification performance on a large number of benchmark machine learning tasks. To scale up DBN, we propose an approach that can use the computing clusters in a distributed environment to train large models, while the dense matrix computations within a single machine are sped up using graphics processors (GPU). When training a DBN, each machine randomly drops out a portion of neurons in each hidden layer, for each training case, making the remaining neurons only learn to detect features that are generally helpful for producing the correct answer. Within our approach, we have developed four methods to combine outcomes from each machine to form a unified model. Our preliminary experiment on the mnst handwritten digit database demonstrates that our approach outperforms the state of the art test error rate."
                    },
                    {
                        "title": "Do CNNs Encode Data Augmentations?",
                        "abstract": "Data augmentations are important ingredients in the recipe for training robust neural networks, especially in computer vision. A fundamental question is whether neural network features encode data augmentation transformations. To answer this question, we introduce a systematic approach to investigate which layers of neural networks are the most predictive of augmentation transformations. Our approach uses features in pre-trained vision models with minimal additional processing to predict common properties transformed by augmentation (scale, aspect ratio, hue, saturation, contrast, and brightness). Surprisingly, neural network features not only predict data augmentation transformations, but they predict many transformations with high accuracy. After validating that neural networks encode features corresponding to augmentation transformations, we show that these features are encoded in the early layers of modern CNNs, though the augmentation signal fades in deeper layers."
                    },
                    {
                        "title": "Research on PM emission and energy consumption of mixed traffic flow based on cellular automata FI model",
                        "abstract": "Traffic congestion leads to a sharp increase in fuel consumption and exhaust emissions, which easily leads to air pollution and fuel waste. Based on Fukui-Ishibashi (FI) model, this paper studies particulate matter(PM) emission and energy dissipation of traffic flow by considering the maximum speed, mixing ratio and random slowing probability. Simulation results show that the energy dissipation of the traffic flow consisting of the same length and different maximum velocity is associated with the maximum velocity. The greater the maximum velocity, the more the energy consumes. When the maximum velocity is different, the PM emission is similar to the change of energy consumption. Moreover, for the same maximum velocity and different mixing ratio, the energy dissipation and the PM emission related to the mixing ratio. There are the maximum fuel consumption and emission value in the maximum flow. For the same maximum velocity and length of vehicles, their energy consumption and emission are determined by random slowing probability."
                    },
                    {
                        "title": "Regularized Evolution for Image Classifier Architecture Search",
                        "abstract": "The effort devoted to hand-crafting neural network image classifiers has motivated the use of architecture search to discover them automatically. Although evolutionary algorithms have been repeatedly applied to neural network topologies, the image classifiers thus discovered have remained inferior to human-crafted ones. Here, we evolve an image classifier---AmoebaNet-A---that surpasses hand-designs for the first time. To do this, we modify the tournament selection evolutionary algorithm by introducing an age property to favor the younger genotypes. Matching size, AmoebaNet-A has comparable accuracy to current state-of-the-art ImageNet models discovered with more complex architecture-search methods. Scaled to larger size, AmoebaNet-A sets a new state-of-the-art 83.9% / 96.6% top-5 ImageNet accuracy. In a controlled comparison against a well known reinforcement learning algorithm, we give evidence that evolution can obtain results faster with the same hardware, especially at the earlier stages of the search. This is relevant when fewer compute resources are available. Evolution is, thus, a simple method to effectively discover high-quality architectures."
                    },
                    {
                        "title": "Just Pick a Sign: Optimizing Deep Multitask Models with Gradient Sign Dropout",
                        "abstract": "The vast majority of deep models use multiple gradient signals, typically corresponding to a sum of multiple loss terms, to update a shared set of trainable weights. However, these multiple updates can impede optimal training by pulling the model in conflicting directions. We present Gradient Sign Dropout (GradDrop), a probabilistic masking procedure which samples gradients at an activation layer based on their level of consistency. GradDrop is implemented as a simple deep layer that can be used in any deep net and synergizes with other gradient balancing approaches. We show that GradDrop outperforms the state-of-the-art multiloss methods within traditional multitask and transfer learning settings, and we discuss how GradDrop reveals links between optimal multiloss training and gradient stochasticity."
                    },
                    {
                        "title": "Lifelong Language Pretraining with Distribution-Specialized Experts",
                        "abstract": "Pretraining on a large-scale corpus has become a standard method to build general language models (LMs). Adapting a model to new data distributions targeting different downstream tasks poses significant challenges. Naive fine-tuning may incur catastrophic forgetting when the over-parameterized LMs overfit the new data but fail to preserve the pretrained features. Lifelong learning (LLL) aims to enable information systems to learn from a continuous data stream across time. However, most prior work modifies the training recipe assuming a static fixed network architecture. We find that additional model capacity and proper regularization are key elements to achieving strong LLL performance. Thus, we propose Lifelong-MoE, an extensible MoE (Mixture-of-Experts) architecture that dynamically adds model capacity via adding experts with regularized pretraining. Our results show that by only introducing a limited number of extra experts while keeping the computation cost constant, our model can steadily adapt to data distribution shifts while preserving the previous knowledge. Compared to existing lifelong learning approaches, Lifelong-MoE achieves better few-shot performance on 19 downstream NLP tasks."
                    },
                    {
                        "title": "Determination of Strong Coupling Constant from Inclusive Semileptonic Decays of Charmed Mesons",
                        "abstract": "Employing the heavy quark expansion model with the kinetic scheme, we evaluate $\\alpha_S(m_c^2)$, the strong coupling constant at the charm quark mass $m_c$ with data on inclusive semileptonic decays of charmed mesons. Using the experimental values of semileptonic decay widths of the $D^0$ and the $D^+$, the value of $\\alpha_{s}(m_c^{2})$ is determined to be $0.445\\pm0.009\\pm0.114$, where the first uncertainty is experimental and the second systematic. This reported $\\alpha_{s}(m_c^{2})$ is in good agreement with the value of $\\alpha_{s}(m_c^{2})$ calculated by running $\\alpha_S(m_Z^2)$ at the $Z^0$ boson mass $m_Z$ with the renormalization group evolution equation. In addition, values of $\\alpha_{s}(m_c^{2})$ obtained individually from each of the $D^0$, $D^+$, and $D_s^+$ mesons are found to be consistent being of the same origin."
                    },
                    {
                        "title": "Expected $H\\to\u03bc^{+}\u03bc^{-}$ measurement precision with $e^{+}e^{-}\\to Z(q\\bar{q})H$ production at the CEPC",
                        "abstract": "A search for the dimuon decay of the Standard Model Higgs boson is performed using the Monte Carlo simulated events to mimic data corresponding to an integrated luminosity of 5.6 ab$^{-1}$ collected with the Circular Electron-Positron Collider detector in $e^{+}e^{-}$ collisions at $\\sqrt{s}=240$ GeV. The paper studies $e^{+}e^{-}\\to ZH,\\,Z\\to q\\bar{q},\\,H\\to\\mu^{+}\\mu^{-}$ process, and the expected significance considering only the data statistical uncertainty over the background-only hypothesis for a Higgs boson with a mass of 125 GeV is found to be 6.1$\\sigma$, corresponding to the precision of 19%. The systematic impacts from the background Monte Carlo statistical fluctuations are estimated to be negligible. The dependence of the measurement accuracy on the muon momentum resolution of the CEPC detector has been investigated. It is found that the muon momentum resolution has to be better than 204 MeV to discover the $H\\to\\mu\\mu$ process at the nominal integrated luminosity. And if the resolution is 100% worse than the designed parameter, the integrated luminosity is needed to be greater than 7.2 ab$^{-1}$ to reach 5$\\sigma$ significance."
                    }
                ]
            },
            "4c9d0270-8c1f-4ce5-8b7f-b312d82c30e5": {
                "pk": "4c9d0270-8c1f-4ce5-8b7f-b312d82c30e5",
                "name": "Joseph E. Gonzalez",
                "collaborators": [
                    "Joseph M. Hellerstein",
                    "Vikram Sreekanti",
                    "Harikaran Subbaraj",
                    "Chenggang Wu",
                    "Vidit Saxena",
                    "Joakim Jald\u00e9n",
                    "Charles Packer",
                    "Pieter Abbeel",
                    "Doris Xin",
                    "Devin Petersohn"
                ],
                "domain": [
                    "Machine Learning",
                    "Reinforcement Learning",
                    "Data Processing",
                    "Video Analysis"
                ],
                "publications": [
                    {
                        "title": "Optimizing Prediction Serving on Low-Latency Serverless Dataflow",
                        "abstract": "Prediction serving systems are designed to provide large volumes of low-latency inferences machine learning models. These systems mix data processing and computationally intensive model inference and benefit from multiple heterogeneous processors and distributed computing resources. In this paper, we argue that a familiar dataflow API is well-suited to this latency-sensitive task, and amenable to optimization even with unmodified black-box ML models. We present the design of Cloudflow, a system that provides this API and realizes it on an autoscaling serverless backend. Cloudflow transparently implements performance-critical optimizations including operator fusion and competitive execution. Our evaluation shows that Cloudflow's optimizations yield significant performance improvements on synthetic workloads and that Cloudflow outperforms state-of-the-art prediction serving systems by as much as 2x on real-world prediction pipelines, meeting latency goals of demanding applications like real-time video analysis."
                    },
                    {
                        "title": "Thompson Sampling for Linearly Constrained Bandits",
                        "abstract": "We address multi-armed bandits (MAB) where the objective is to maximize the cumulative reward under a probabilistic linear constraint. For a few real-world instances of this problem, constrained extensions of the well-known Thompson Sampling (TS) heuristic have recently been proposed. However, finite-time analysis of constrained TS is challenging; as a result, only O(\\sqrt{T}) bounds on the cumulative reward loss (i.e., the regret) are available. In this paper, we describe LinConTS, a TS-based algorithm for bandits that place a linear constraint on the probability of earning a reward in every round. We show that for LinConTS, the regret as well as the cumulative constraint violations are upper bounded by O(\\log T) for the suboptimal arms. We develop a proof technique that relies on careful analysis of the dual problem and combine it with recent theoretical work on unconstrained TS. Through numerical experiments on two real-world datasets, we demonstrate that LinConTS outperforms an asymptotically optimal upper confidence bound (UCB) scheme in terms of simultaneously minimizing the regret and the violation."
                    },
                    {
                        "title": "Hindsight Task Relabelling: Experience Replay for Sparse Reward Meta-RL",
                        "abstract": "Meta-reinforcement learning (meta-RL) has proven to be a successful framework for leveraging experience from prior tasks to rapidly learn new related tasks, however, current meta-RL approaches struggle to learn in sparse reward environments. Although existing meta-RL algorithms can learn strategies for adapting to new sparse reward tasks, the actual adaptation strategies are learned using hand-shaped reward functions, or require simple environments where random exploration is sufficient to encounter sparse reward. In this paper, we present a formulation of hindsight relabeling for meta-RL, which relabels experience during meta-training to enable learning to learn entirely using sparse reward. We demonstrate the effectiveness of our approach on a suite of challenging sparse reward goal-reaching environments that previously required dense reward during meta-training to solve. Our approach solves these environments using the true sparse reward function, with performance comparable to training with a proxy dense reward function."
                    },
                    {
                        "title": "Enhancing the Interactivity of Dataframe Queries by Leveraging Think Time",
                        "abstract": "We propose opportunistic evaluation, a framework for accelerating interactions with dataframes. Interactive latency is critical for iterative, human-in-the-loop dataframe workloads for supporting exploratory data analysis. Opportunistic evaluation significantly reduces interactive latency by 1) prioritizing computation directly relevant to the interactions and 2) leveraging think time for asynchronous background computation for non-critical operators that might be relevant to future interactions. We show, through empirical analysis, that current user behavior presents ample opportunities for optimization, and the solutions we propose effectively harness such opportunities."
                    },
                    {
                        "title": "Inter-BMV: Interpolation with Block Motion Vectors for Fast Semantic Segmentation on Video",
                        "abstract": "Models optimized for accuracy on single images are often prohibitively slow to run on each frame in a video. Recent work exploits the use of optical flow to warp image features forward from select keyframes, as a means to conserve computation on video. This approach, however, achieves only limited speedup, even when optimized, due to the accuracy degradation introduced by repeated forward warping, and the inference cost of optical flow estimation. To address these problems, we propose a new scheme that propagates features using the block motion vectors (BMV) present in compressed video (e.g. H.264 codecs), instead of optical flow, and bi-directionally warps and fuses features from enclosing keyframes to capture scene context on each video frame. Our technique, interpolation-BMV, enables us to accurately estimate the features of intermediate frames, while keeping inference costs low. We evaluate our system on the CamVid and Cityscapes datasets, comparing to both a strong single-frame baseline and related work. We find that we are able to substantially accelerate segmentation on video, achieving near real-time frame rates (20+ frames per second) on large images (e.g. 960 x 720 pixels), while maintaining competitive accuracy. This represents an improvement of almost 6x over the single-frame baseline and 2.5x over the fastest prior work."
                    }
                ]
            },
            "f67f36ce-a3df-47a7-b3ca-e84dc93f2d85": {
                "pk": "f67f36ce-a3df-47a7-b3ca-e84dc93f2d85",
                "name": "Zhifeng Chen",
                "collaborators": [
                    "Chengguang Bao",
                    "Zhibing Li",
                    "Yanzhang He",
                    "Zetong Yang",
                    "Yin Zhou",
                    "Jiquan Ngiam",
                    "Dinghua Shi",
                    "Chuang Ma",
                    "Guanrong Chen"
                ],
                "domain": [
                    "Quantum Physics",
                    "Spin Dynamics",
                    "Topological Data Analysis",
                    "3D Object Detection"
                ],
                "publications": [
                    {
                        "title": "Effect of magnetic fields on the spin evolution of non-polarized 87Rb Bose-Einstein condensates",
                        "abstract": "The spin mixing dynamics of spin-1 Bose-Einstein condensates with zero magnetization and under an external magnetic field is investigated. The time-dependent solutions are obtained via a diagonalization of the Hamiltonian, which has a simple form under the single mode approximation. The features of evolution are compared in detail with those with the field removed so as to emphasize the effect of the field, which can induce strong oscillation in population of atoms in spin component 0 . A new mode of oscillation characterized by a high frequency and a low frequency, is found when the field is sufficiently strong."
                    },
                    {
                        "title": "Strong spin-oscillation of small spin-1 condensates caused by an inclined weak magnetic field",
                        "abstract": "When a magnetic field is applied along a direction deviated from the quantization $Z$-axis, the conservation of total magnetization holds no more. In this case the inclined field can cause a strong spin-evolution via the linear Zeeman term even the field is as weak as a percentage of $mG$. An approach beyond the mean field theory is proposed to study the evolution of small $^{87}$Rb condensates under the weak inclined fields. The time-dependent populations of spin-components are given in analytical forms. The evolution is found to be highly sensitive to the magnitude and direction of the field."
                    },
                    {
                        "title": "Repeating head-on collisions in an optical trap and the evaluation of spin-dependent interactions among neutral particles",
                        "abstract": "A dynamic process of repeating collisions of a pair of trapped neutral particles with weak spin-dependent interaction is designed and studied. Related theoretical derivation and numerical calculation have been performed to study the inherent coordinate-spin and momentum-spin correlation. Due to the repeating collisions the effect of the weak interaction can be accumulated and enlarged, and therefore can be eventually detected. Numerical results suggest that the Cr-Cr interaction, which has not yet been completely clear, could be thereby determined. The design can be in general used to determine various interactions among neutral atoms and molecules, in particular for the determination of very weak forces."
                    },
                    {
                        "title": "3D-MAN: 3D Multi-frame Attention Network for Object Detection",
                        "abstract": "3D object detection is an important module in autonomous driving and robotics. However, many existing methods focus on using single frames to perform 3D detection, and do not fully utilize information from multiple frames. In this paper, we present 3D-MAN: a 3D multi-frame attention network that effectively aggregates features from multiple perspectives and achieves state-of-the-art performance on Waymo Open Dataset. 3D-MAN first uses a novel fast single-frame detector to produce box proposals. The box proposals and their corresponding feature maps are then stored in a memory bank. We design a multi-view alignment and aggregation module, using attention networks, to extract and aggregate the temporal features stored in the memory bank. This effectively combines the features coming from different perspectives of the scene. We demonstrate the effectiveness of our approach on the large-scale complex Waymo Open Dataset, achieving state-of-the-art results compared to published single-frame and multi-frame methods."
                    },
                    {
                        "title": "Computing persistent homology by spanning trees and critical simplices",
                        "abstract": "Topological data analysis can extract effective information from higher-dimensional data. Its mathematical basis is persistent homology. The persistent homology can calculate topological features at different spatiotemporal scales of the dataset; that is, establishing the integrated taxonomic relation among points, lines and simplices. Here, the simplicial network composed of all-order simplices in a simplicial complex is essential. Because the sequence of nested simplicial subnetworks can be regarded as a discrete Morse function from the simplicial network to real values, a method based on the concept of critical simplices can be developed by searching all-order spanning trees. Employing this new method, not only the Morse function values with the theoretical minimum number of critical simplices can be obtained, but also the Betti numbers and composition of all-order cavities in the simplicial network can be calculated quickly. Finally, this method is used to analyze some examples and compared with other methods, showing its effectiveness and feasibility."
                    }
                ]
            },
            "9c6f101e-6f1d-4e24-9e75-cc563d89ce0a": {
                "pk": "9c6f101e-6f1d-4e24-9e75-cc563d89ce0a",
                "name": "Ruslan Salakhutdinov",
                "collaborators": [
                    "Nathan Srebro",
                    "Emilio Parisotto",
                    "Behnam Neyshabur",
                    "Ryan Kiros",
                    "Nitish Srivastava",
                    "Richard S. Zemel",
                    "Cody Severinski",
                    "John Langford",
                    "Tong Zhang",
                    "Dean Foster"
                ],
                "domain": [
                    "Matrix Completion",
                    "Reinforcement Learning",
                    "Deep Learning",
                    "Attention Mechanisms"
                ],
                "publications": [
                    {
                        "title": "Collaborative Filtering in a Non-Uniform World: Learning with the Weighted Trace Norm",
                        "abstract": "We show that matrix completion with trace-norm regularization can be significantly hurt when entries of the matrix are sampled non-uniformly. We introduce a weighted version of the trace-norm regularizer that works well also with non-uniform sampling. Our experimental results demonstrate that the weighted trace-norm regularization indeed yields significant gains on the (highly non-uniformly sampled) Netflix dataset."
                    },
                    {
                        "title": "Bayesian Probabilistic Matrix Factorization: A User Frequency Analysis",
                        "abstract": "Matrix factorization (MF) has become a common approach to collaborative filtering, due to ease of implementation and scalability to large data sets. Two existing drawbacks of the basic model is that it does not incorporate side information on either users or items, and assumes a common variance for all users. We extend the work of constrained probabilistic matrix factorization by deriving the Gibbs updates for the side feature vectors for items (Salakhutdinov and Minh, 2008). We show that this Bayesian treatment to the constrained PMF model outperforms simple MAP estimation. We also consider extensions to heteroskedastic precision introduced in the literature (Lakshminarayanan, Bouchard, and Archambeau, 2011). We show that this tends result in overfitting for deterministic approximation algorithms (ex: Variational inference) when the observed entries in the user / item matrix are distributed in an non-uniform manner. In light of this, we propose a truncated precision model. Our experimental results suggest that this model tends to delay overfitting."
                    },
                    {
                        "title": "Neural Map: Structured Memory for Deep Reinforcement Learning",
                        "abstract": "A critical component to enabling intelligent reasoning in partially observable environments is memory. Despite this importance, Deep Reinforcement Learning (DRL) agents have so far used relatively simple memory architectures, with the main methods to overcome partial observability being either a temporal convolution over the past k frames or an LSTM layer. More recent work (Oh et al., 2016) has went beyond these architectures by using memory networks which can allow more sophisticated addressing schemes over the past k frames. But even these architectures are unsatisfactory due to the reason that they are limited to only remembering information from the last k frames. In this paper, we develop a memory system with an adaptable write operator that is customized to the sorts of 3D environments that DRL agents typically interact with. This architecture, called the Neural Map, uses a spatially structured 2D memory image to learn to store arbitrary information about the environment over long time lags. We demonstrate empirically that the Neural Map surpasses previous DRL memories on a set of challenging 2D and 3D maze environments and show that it is capable of generalizing to environments that were not seen during training."
                    },
                    {
                        "title": "Efficient Transformers in Reinforcement Learning using Actor-Learner Distillation",
                        "abstract": "Many real-world applications such as robotics provide hard constraints on power and compute that limit the viable model complexity of Reinforcement Learning (RL) agents. Similarly, in many distributed RL settings, acting is done on un-accelerated hardware such as CPUs, which likewise restricts model size to prevent intractable experiment run times. These \"actor-latency\" constrained settings present a major obstruction to the scaling up of model complexity that has recently been extremely successful in supervised learning. To be able to utilize large model capacity while still operating within the limits imposed by the system during acting, we develop an \"Actor-Learner Distillation\" (ALD) procedure that leverages a continual form of distillation that transfers learning progress from a large capacity learner model to a small capacity actor model. As a case study, we develop this procedure in the context of partially-observable environments, where transformer models have had large improvements over LSTMs recently, at the cost of significantly higher computational complexity. With transformer models as the learner and LSTMs as the actor, we demonstrate in several challenging memory environments that using Actor-Learner Distillation recovers the clear sample-efficiency gains of the transformer learner model while maintaining the fast inference and reduced total training time of the LSTM actor model."
                    },
                    {
                        "title": "Learning Nonlinear Dynamic Models",
                        "abstract": "We present a novel approach for learning nonlinear dynamic models, which leads to a new set of tools capable of solving problems that are otherwise difficult. We provide theory showing this new approach is consistent for models with long range structure, and apply the approach to motion capture and high-dimensional video data, yielding results superior to standard alternatives."
                    },
                    {
                        "title": "Domain Adaptation: Overfitting and Small Sample Statistics",
                        "abstract": "We study the prevalent problem when a test distribution differs from the training distribution. We consider a setting where our training set consists of a small number of sample domains, but where we have many samples in each domain. Our goal is to generalize to a new domain. For example, we may want to learn a similarity function using only certain classes of objects, but we desire that this similarity function be applicable to object classes not present in our training sample (e.g. we might seek to learn that \"dogs are similar to dogs\" even though images of dogs were absent from our training set). Our theoretical analysis shows that we can select many more features than domains while avoiding overfitting by utilizing data-dependent variance properties. We present a greedy feature selection algorithm based on using T-statistics. Our experiments validate this theory showing that our T-statistic based greedy feature selection is more robust at avoiding overfitting than the classical greedy procedure."
                    },
                    {
                        "title": "Matrix reconstruction with the local max norm",
                        "abstract": "We introduce a new family of matrix norms, the \"local max\" norms, generalizing existing methods such as the max norm, the trace norm (nuclear norm), and the weighted or smoothed weighted trace norms, which have been extensively used in the literature as regularizers for matrix reconstruction problems. We show that this new family can be used to interpolate between the (weighted or unweighted) trace norm and the more conservative max norm. We test this interpolation on simulated data and on the large-scale Netflix and MovieLens ratings data, and find improved accuracy relative to the existing matrix norms. We also provide theoretical results showing learning guarantees for some of the new norms."
                    },
                    {
                        "title": "Initialization Strategies of Spatio-Temporal Convolutional Neural Networks",
                        "abstract": "We propose a new way of incorporating temporal information present in videos into Spatial Convolutional Neural Networks (ConvNets) trained on images, that avoids training Spatio-Temporal ConvNets from scratch. We describe several initializations of weights in 3D Convolutional Layers of Spatio-Temporal ConvNet using 2D Convolutional Weights learned from ImageNet. We show that it is important to initialize 3D Convolutional Weights judiciously in order to learn temporal representations of videos. We evaluate our methods on the UCF-101 dataset and demonstrate improvement over Spatial ConvNets."
                    },
                    {
                        "title": "Path-SGD: Path-Normalized Optimization in Deep Neural Networks",
                        "abstract": "We revisit the choice of SGD for training deep neural networks by reconsidering the appropriate geometry in which to optimize the weights. We argue for a geometry invariant to rescaling of weights that does not affect the output of the network, and suggest Path-SGD, which is an approximate steepest descent method with respect to a path-wise regularizer related to max-norm regularization. Path-SGD is easy and efficient to implement and leads to empirical gains over SGD and AdaGrad."
                    },
                    {
                        "title": "Data-Dependent Path Normalization in Neural Networks",
                        "abstract": "We propose a unified framework for neural net normalization, regularization and optimization, which includes Path-SGD and Batch-Normalization and interpolates between them across two different dimensions. Through this framework we investigate issue of invariance of the optimization, data dependence and the connection with natural gradients."
                    },
                    {
                        "title": "Path-Normalized Optimization of Recurrent Neural Networks with ReLU Activations",
                        "abstract": "We investigate the parameter-space geometry of recurrent neural networks (RNNs), and develop an adaptation of path-SGD optimization method, attuned to this geometry, that can learn plain RNNs with ReLU activations. On several datasets that require capturing long-term dependency structure, we show that path-SGD can significantly improve trainability of ReLU RNNs compared to RNNs trained with SGD, even with various recently suggested initialization schemes."
                    },
                    {
                        "title": "Topological Sort for Sentence Ordering",
                        "abstract": "Sentence ordering is the task of arranging the sentences of a given text in the correct order. Recent work using deep neural networks for this task has framed it as a sequence prediction problem. In this paper, we propose a new framing of this task as a constraint solving problem and introduce a new technique to solve it. Additionally, we propose a human evaluation for this task. The results on both automatic and human metrics across four different datasets show that this new technique is better at capturing coherence in documents."
                    },
                    {
                        "title": "Learning Wake-Sleep Recurrent Attention Models",
                        "abstract": "Despite their success, convolutional neural networks are computationally expensive because they must examine all image locations. Stochastic attention-based models have been shown to improve computational efficiency at test time, but they remain difficult to train because of intractable posterior inference and high variance in the stochastic gradient estimates. Borrowing techniques from the literature on training deep generative models, we present the Wake-Sleep Recurrent Attention Model, a method for training stochastic attention networks which improves posterior inference and which reduces the variability in the stochastic gradients. We show that our method can greatly speed up the training time for stochastic attention networks in the domains of image classification and caption generation."
                    },
                    {
                        "title": "Learning Generative Models with Visual Attention",
                        "abstract": "Attention has long been proposed by psychologists as important for effectively dealing with the enormous sensory stimulus available in the neocortex. Inspired by the visual attention models in computational neuroscience and the need of object-centric data for generative models, we describe for generative learning framework using attentional mechanisms. Attentional mechanisms can propagate signals from region of interest in a scene to an aligned canonical representation, where generative modeling takes place. By ignoring background clutter, generative models can concentrate their resources on the object of interest. Our model is a proper graphical model where the 2D Similarity transformation is a part of the top-down process. A ConvNet is employed to provide good initializations during posterior inference which is based on Hamiltonian Monte Carlo. Upon learning images of faces, our model can robustly attend to face regions of novel test subjects. More importantly, our model can learn generative models of new faces from a novel dataset of large images where the face locations are not known."
                    },
                    {
                        "title": "A Multiplicative Model for Learning Distributed Text-Based Attribute Representations",
                        "abstract": "In this paper we propose a general framework for learning distributed representations of attributes: characteristics of text whose representations can be jointly learned with word embeddings. Attributes can correspond to document indicators (to learn sentence vectors), language indicators (to learn distributed language representations), meta-data and side information (such as the age, gender and industry of a blogger) or representations of authors. We describe a third-order model where word context and attribute vectors interact multiplicatively to predict the next word in a sequence. This leads to the notion of conditional word similarity: how meanings of words change when conditioned on different attributes. We perform several experimental tasks including sentiment classification, cross-lingual document classification, and blog authorship attribution. We also qualitatively evaluate conditional word neighbours and attribute-conditioned text generation."
                    },
                    {
                        "title": "Unifying Visual-Semantic Embeddings with Multimodal Neural Language Models",
                        "abstract": "Inspired by recent advances in multimodal learning and machine translation, we introduce an encoder-decoder pipeline that learns (a): a multimodal joint embedding space with images and text and (b): a novel language model for decoding distributed representations from our space. Our pipeline effectively unifies joint image-text embedding models with multimodal neural language models. We introduce the structure-content neural language model that disentangles the structure of a sentence to its content, conditioned on representations produced by the encoder. The encoder allows one to rank images and sentences while the decoder can generate novel descriptions from scratch. Using LSTM to encode sentences, we match the state-of-the-art performance on Flickr8K and Flickr30K without using object detections. We also set new best results when using the 19-layer Oxford convolutional network. Furthermore we show that with linear encoders, the learned embedding space captures multimodal regularities in terms of vector space arithmetic e.g. *image of a blue car* - \"blue\" + \"red\" is near images of red cars. Sample captions generated for 800 images are made available for comparison."
                    },
                    {
                        "title": "segDeepM: Exploiting Segmentation and Context in Deep Neural Networks for Object Detection",
                        "abstract": "In this paper, we propose an approach that exploits object segmentation in order to improve the accuracy of object detection. We frame the problem as inference in a Markov Random Field, in which each detection hypothesis scores object appearance as well as contextual information using Convolutional Neural Networks, and allows the hypothesis to choose and score a segment out of a large pool of accurate object segmentation proposals. This enables the detector to incorporate additional evidence when it is available and thus results in more accurate detections. Our experiments show an improvement of 4.1% in mAP over the R-CNN baseline on PASCAL VOC 2010, and 3.4% over the current state-of-the-art, demonstrating the power of our approach."
                    },
                    {
                        "title": "Action Recognition using Visual Attention",
                        "abstract": "We propose a soft attention based model for the task of action recognition in videos. We use multi-layered Recurrent Neural Networks (RNNs) with Long Short-Term Memory (LSTM) units which are deep both spatially and temporally. Our model learns to focus selectively on parts of the video frames and classifies videos after taking a few glimpses. The model essentially learns which parts in the frames are relevant for the task at hand and attaches higher importance to them. We evaluate the model on UCF-11 (YouTube Action), HMDB-51 and Hollywood2 datasets and analyze how the model focuses its attention depending on the scene and the action being performed."
                    },
                    {
                        "title": "Actor-Mimic: Deep Multitask and Transfer Reinforcement Learning",
                        "abstract": "The ability to act in multiple environments and transfer previous knowledge to new situations can be considered a critical aspect of any intelligent agent. Towards this goal, we define a novel method of multitask and transfer learning that enables an autonomous agent to learn how to behave in multiple tasks simultaneously, and then generalize its knowledge to new domains. This method, termed \"Actor-Mimic\", exploits the use of deep reinforcement learning and model compression techniques to train a single policy network that learns how to act in a set of distinct tasks by using the guidance of several expert teachers. We then show that the representations learnt by the deep policy network are capable of generalizing to new tasks with no prior expert guidance, speeding up learning in novel environments. Although our method can in general be applied to a wide range of problems, we use Atari games as a testing environment to demonstrate these methods."
                    }
                ]
            },
            "d2bf11da-ed2d-4b14-8c6f-886ff23f3239": {
                "pk": "d2bf11da-ed2d-4b14-8c6f-886ff23f3239",
                "name": "Ion Stoica",
                "collaborators": [
                    "Sanjay Krishnan",
                    "Ken Goldberg",
                    "Daniel Kang",
                    "Tatsunori Hashimoto",
                    "Yi Sun",
                    "Eric Liang",
                    "Roy Fox",
                    "Peter Bailis",
                    "Hao Zhang",
                    "Scott Shenker"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Machine Learning",
                    "Distributed Systems",
                    "Zero-Knowledge Proofs"
                ],
                "publications": [
                    {
                        "title": "Multi-Level Discovery of Deep Options",
                        "abstract": "Augmenting an agent's control with useful higher-level behaviors called options can greatly reduce the sample complexity of reinforcement learning, but manually designing options is infeasible in high-dimensional and abstract state spaces. While recent work has proposed several techniques for automated option discovery, they do not scale to multi-level hierarchies and to expressive representations such as deep networks. We present Discovery of Deep Options (DDO), a policy-gradient algorithm that discovers parametrized options from a set of demonstration trajectories, and can be used recursively to discover additional levels of the hierarchy. The scalability of our approach to multi-level hierarchies stems from the decoupling of low-level option discovery from high-level meta-control policy learning, facilitated by under-parametrization of the high level. We demonstrate that using the discovered options to augment the action space of Deep Q-Network agents can accelerate learning by guiding exploration in tasks where random actions are unlikely to reach valuable states. We show that DDO is effective in adding options that accelerate learning in 4 out of 5 Atari RAM environments chosen in our experiments. We also show that DDO can discover structure in robot-assisted surgical videos and kinematics that match expert annotation with 72% accuracy."
                    },
                    {
                        "title": "Scaling up Trustless DNN Inference with Zero-Knowledge Proofs",
                        "abstract": "As ML models have increased in capabilities and accuracy, so has the complexity of their deployments. Increasingly, ML model consumers are turning to service providers to serve the ML models in the ML-as-a-service (MLaaS) paradigm. As MLaaS proliferates, a critical requirement emerges: how can model consumers verify that the correct predictions were served, in the face of malicious, lazy, or buggy service providers?   In this work, we present the first practical ImageNet-scale method to verify ML model inference non-interactively, i.e., after the inference has been done. To do so, we leverage recent developments in ZK-SNARKs (zero-knowledge succinct non-interactive argument of knowledge), a form of zero-knowledge proofs. ZK-SNARKs allows us to verify ML model execution non-interactively and with only standard cryptographic hardness assumptions. In particular, we provide the first ZK-SNARK proof of valid inference for a full resolution ImageNet model, achieving 79\\% top-5 accuracy. We further use these ZK-SNARKs to design protocols to verify ML model execution in a variety of scenarios, including for verifying MLaaS predictions, verifying MLaaS model accuracy, and using ML models for trustless retrieval. Together, our results show that ZK-SNARKs have the promise to make verified ML model inference practical."
                    },
                    {
                        "title": "CodeScholar: Growing Idiomatic Code Examples",
                        "abstract": "Programmers often search for usage examples for API methods. A tool that could generate realistic, idiomatic, and contextual usage examples for one or more APIs would be immensely beneficial to developers. Such a tool would relieve the need for a deep understanding of the API landscape, augment existing documentation, and help discover interactions among APIs. We present CodeScholar, a tool that generates idiomatic code examples demonstrating the common usage of API methods. It includes a novel neural-guided search technique over graphs that grows the query APIs into idiomatic code examples. Our user study demonstrates that in 70% of cases, developers prefer CodeScholar generated examples over state-of-the-art large language models (LLM) like GPT3.5. We quantitatively evaluate 60 single and 25 multi-API queries from 6 popular Python libraries and show that across-the-board CodeScholar generates more realistic, diverse, and concise examples. In addition, we show that CodeScholar not only helps developers but also LLM-powered programming assistants generate correct code in a program synthesis setting."
                    },
                    {
                        "title": "Break the Sequential Dependency of LLM Inference Using Lookahead Decoding",
                        "abstract": "Autoregressive decoding of large language models (LLMs) is memory bandwidth bounded, resulting in high latency and significant wastes of the parallel processing power of modern accelerators. Existing methods for accelerating LLM decoding often require a draft model (e.g., speculative decoding), which is nontrivial to obtain and unable to generalize. In this paper, we introduce Lookahead decoding, an exact, parallel decoding algorithm that accelerates LLM decoding without needing auxiliary models or data stores. It allows trading per-step log(FLOPs) to reduce the number of total decoding steps, is more parallelizable on single or multiple modern accelerators, and is compatible with concurrent memory-efficient attention (e.g., FlashAttention). Our implementation of Lookahead decoding can speed up autoregressive decoding by up to 1.8x on MT-bench and 4x with strong scaling on multiple GPUs in code completion tasks. Our code is avialable at https://github.com/hao-ai-lab/LookaheadDecoding"
                    },
                    {
                        "title": "Neural Packet Classification",
                        "abstract": "Packet classification is a fundamental problem in computer networking. This problem exposes a hard tradeoff between the computation and state complexity, which makes it particularly challenging. To navigate this tradeoff, existing solutions rely on complex hand-tuned heuristics, which are brittle and hard to optimize. In this paper, we propose a deep reinforcement learning (RL) approach to solve the packet classification problem. There are several characteristics that make this problem a good fit for Deep RL. First, many of the existing solutions are iteratively building a decision tree by splitting nodes in the tree. Second, the effects of these actions (e.g., splitting nodes) can only be evaluated once we are done with building the tree. These two characteristics are naturally captured by the ability of RL to take actions that have sparse and delayed rewards. Third, it is computationally efficient to generate data traces and evaluate decision trees, which alleviate the notoriously high sample complexity problem of Deep RL algorithms. Our solution, NeuroCuts, uses succinct representations to encode state and action space, and efficiently explore candidate decision trees to optimize for a global objective. It produces compact decision trees optimized for a specific set of rules and a given performance metric, such as classification time, memory footprint, or a combination of the two. Evaluation on ClassBench shows that NeuroCuts outperforms existing hand-crafted algorithms in classification time by 18% at the median, and reduces both time and memory footprint by up to 3x."
                    },
                    {
                        "title": "ZK-IMG: Attested Images via Zero-Knowledge Proofs to Fight Disinformation",
                        "abstract": "Over the past few years, AI methods of generating images have been increasing in capabilities, with recent breakthroughs enabling high-resolution, photorealistic \"deepfakes\" (artificially generated images with the purpose of misinformation or harm). The rise of deepfakes has potential for social disruption. Recent work has proposed using ZK-SNARKs (zero-knowledge succinct non-interactive argument of knowledge) and attested cameras to verify that images were taken by a camera. ZK-SNARKs allow verification of image transformations non-interactively (i.e., post-hoc) with only standard cryptographic hardness assumptions. Unfortunately, this work does not preserve input privacy, is impractically slow (working only on 128$\\times$128 images), and/or requires custom cryptographic arguments.   To address these issues, we present zk-img, a library for attesting to image transformations while hiding the pre-transformed image. zk-img allows application developers to specify high level image transformations. Then, zk-img will transparently compile these specifications to ZK-SNARKs. To hide the input or output images, zk-img will compute the hash of the images inside the ZK-SNARK. We further propose methods of chaining image transformations securely and privately, which allows for arbitrarily many transformations. By combining these optimizations, zk-img is the first system to be able to transform HD images on commodity hardware, securely and privately."
                    },
                    {
                        "title": "DDCO: Discovery of Deep Continuous Options for Robot Learning from Demonstrations",
                        "abstract": "An option is a short-term skill consisting of a control policy for a specified region of the state space, and a termination condition recognizing leaving that region. In prior work, we proposed an algorithm called Deep Discovery of Options (DDO) to discover options to accelerate reinforcement learning in Atari games. This paper studies an extension to robot imitation learning, called Discovery of Deep Continuous Options (DDCO), where low-level continuous control skills parametrized by deep neural networks are learned from demonstrations. We extend DDO with: (1) a hybrid categorical-continuous distribution model to parametrize high-level policies that can invoke discrete options as well continuous control actions, and (2) a cross-validation method that relaxes DDO's requirement that users specify the number of options to be discovered. We evaluate DDCO in simulation of a 3-link robot in the vertical plane pushing a block with friction and gravity, and in two physical experiments on the da Vinci surgical robot, needle insertion where a needle is grasped and inserted into a silicone tissue phantom, and needle bin picking where needles and pins are grasped from a pile and categorized into bins. In the 3-link arm simulation, results suggest that DDCO can take 3x fewer demonstrations to achieve the same reward compared to a baseline imitation learning approach. In the needle insertion task, DDCO was successful 8/10 times compared to the next most accurate imitation learning baseline 6/10. In the surgical bin picking task, the learned policy successfully grasps a single object in 66 out of 99 attempted grasps, and in all but one case successfully recovered from failed grasps by retrying a second time."
                    },
                    {
                        "title": "Occupy the Cloud: Distributed Computing for the 99%",
                        "abstract": "Distributed computing remains inaccessible to a large number of users, in spite of many open source platforms and extensive commercial offerings. While distributed computation frameworks have moved beyond a simple map-reduce model, many users are still left to struggle with complex cluster management and configuration tools, even for running simple embarrassingly parallel jobs. We argue that stateless functions represent a viable platform for these users, eliminating cluster management overhead, fulfilling the promise of elasticity. Furthermore, using our prototype implementation, PyWren, we show that this model is general enough to implement a number of distributed computing models, such as BSP, efficiently. Extrapolating from recent trends in network bandwidth and the advent of disaggregated storage, we suggest that stateless functions are a natural fit for data processing in future computing environments."
                    },
                    {
                        "title": "Learning to Optimize Join Queries With Deep Reinforcement Learning",
                        "abstract": "Exhaustive enumeration of all possible join orders is often avoided, and most optimizers leverage heuristics to prune the search space. The design and implementation of heuristics are well-understood when the cost model is roughly linear, and we find that these heuristics can be significantly suboptimal when there are non-linearities in cost. Ideally, instead of a fixed heuristic, we would want a strategy to guide the search space in a more data-driven way---tailoring the search to a specific dataset and query workload. Recognizing the link between classical Dynamic Programming enumeration methods and recent results in Reinforcement Learning (RL), we propose a new method for learning optimized join search strategies. We present our RL-based DQ optimizer, which currently optimizes select-project-join blocks. We implement three versions of DQ to illustrate the ease of integration into existing DBMSes: (1) A version built on top of Apache Calcite, (2) a version integrated into PostgreSQL, and (3) a version integrated into SparkSQL. Our extensive evaluation shows that DQ achieves plans with optimization costs and query execution times competitive with the native query optimizer in each system, but can execute significantly faster after learning (often by orders of magnitude)."
                    },
                    {
                        "title": "IMPACT: Importance Weighted Asynchronous Architectures with Clipped Target Networks",
                        "abstract": "The practical usage of reinforcement learning agents is often bottlenecked by the duration of training time. To accelerate training, practitioners often turn to distributed reinforcement learning architectures to parallelize and accelerate the training process. However, modern methods for scalable reinforcement learning (RL) often tradeoff between the throughput of samples that an RL agent can learn from (sample throughput) and the quality of learning from each sample (sample efficiency). In these scalable RL architectures, as one increases sample throughput (i.e. increasing parallelization in IMPALA), sample efficiency drops significantly. To address this, we propose a new distributed reinforcement learning algorithm, IMPACT. IMPACT extends IMPALA with three changes: a target network for stabilizing the surrogate objective, a circular buffer, and truncated importance sampling. In discrete action-space environments, we show that IMPACT attains higher reward and, simultaneously, achieves up to 30% decrease in training wall-time than that of IMPALA. For continuous control environments, IMPACT trains faster than existing scalable agents while preserving the sample efficiency of synchronous PPO."
                    },
                    {
                        "title": "Population Based Augmentation: Efficient Learning of Augmentation Policy Schedules",
                        "abstract": "A key challenge in leveraging data augmentation for neural network training is choosing an effective augmentation policy from a large search space of candidate operations. Properly chosen augmentation policies can lead to significant generalization improvements; however, state-of-the-art approaches such as AutoAugment are computationally infeasible to run for the ordinary user. In this paper, we introduce a new data augmentation algorithm, Population Based Augmentation (PBA), which generates nonstationary augmentation policy schedules instead of a fixed augmentation policy. We show that PBA can match the performance of AutoAugment on CIFAR-10, CIFAR-100, and SVHN, with three orders of magnitude less overall compute. On CIFAR-10 we achieve a mean test error of 1.46%, which is a slight improvement upon the current state-of-the-art. The code for PBA is open source and is available at https://github.com/arcelien/pba."
                    },
                    {
                        "title": "MPC-Minimized Secure LLM Inference",
                        "abstract": "Many inference services based on large language models (LLMs) pose a privacy concern, either revealing user prompts to the service or the proprietary weights to the user. Secure inference offers a solution to this problem through secure multi-party computation (MPC), however, it is still impractical for modern LLM workload due to the large overhead imposed by MPC. To address this overhead, we propose Marill, a framework that adapts LLM fine-tuning to minimize MPC usage during secure inference. Marill introduces high-level architectural changes during fine-tuning that significantly reduce the number of expensive operations needed within MPC during inference, by removing some and relocating others outside MPC without compromising security. As a result, Marill-generated models are more efficient across all secure inference protocols and our approach complements MPC-friendly approximations for such operations. Compared to standard fine-tuning, Marill results in 3.6-11.3x better runtime and 2.4-6.9x better communication during secure inference across various MPC settings, while typically preserving over 90% performance across downstream tasks."
                    },
                    {
                        "title": "BlinkDB: Queries with Bounded Errors and Bounded Response Times on Very Large Data",
                        "abstract": "In this paper, we present BlinkDB, a massively parallel, sampling-based approximate query engine for running ad-hoc, interactive SQL queries on large volumes of data. The key insight that BlinkDB builds on is that one can often make reasonable decisions in the absence of perfect answers. For example, reliably detecting a malfunctioning server using a distributed collection of system logs does not require analyzing every request processed by the system. Based on this insight, BlinkDB allows one to trade-off query accuracy for response time, enabling interactive queries over massive data by running queries on data samples and presenting results annotated with meaningful error bars. To achieve this, BlinkDB uses two key ideas that differentiate it from previous work in this area: (1) an adaptive optimization framework that builds and maintains a set of multi-dimensional, multi-resolution samples from original data over time, and (2) a dynamic sample selection strategy that selects an appropriately sized sample based on a query's accuracy and/or response time requirements. We have built an open-source version of BlinkDB and validated its effectiveness using the well-known TPC-H benchmark as well as a real-world analytic workload derived from Conviva Inc. Our experiments on a 100 node cluster show that BlinkDB can answer a wide range of queries from a real-world query trace on up to 17 TBs of data in less than 2 seconds (over 100\\times faster than Hive), within an error of 2 - 10%."
                    },
                    {
                        "title": "SparkNet: Training Deep Networks in Spark",
                        "abstract": "Training deep networks is a time-consuming process, with networks for object recognition often requiring multiple days to train. For this reason, leveraging the resources of a cluster to speed up training is an important area of work. However, widely-popular batch-processing computational frameworks like MapReduce and Spark were not designed to support the asynchronous and communication-intensive workloads of existing distributed deep learning systems. We introduce SparkNet, a framework for training deep networks in Spark. Our implementation includes a convenient interface for reading data from Spark RDDs, a Scala interface to the Caffe deep learning framework, and a lightweight multi-dimensional tensor library. Using a simple parallelization scheme for stochastic gradient descent, SparkNet scales well with the cluster size and tolerates very high-latency communication. Furthermore, it is easy to deploy and use with no parameter tuning, and it is compatible with existing Caffe models. We quantify the dependence of the speedup obtained by SparkNet on the number of machines, the communication frequency, and the cluster's communication overhead, and we benchmark our system's performance on the ImageNet dataset."
                    },
                    {
                        "title": "Trustless Audits without Revealing Data or Models",
                        "abstract": "There is an increasing conflict between business incentives to hide models and data as trade secrets, and the societal need for algorithmic transparency. For example, a rightsholder wishing to know whether their copyrighted works have been used during training must convince the model provider to allow a third party to audit the model and data. Finding a mutually agreeable third party is difficult, and the associated costs often make this approach impractical.   In this work, we show that it is possible to simultaneously allow model providers to keep their model weights (but not architecture) and data secret while allowing other parties to trustlessly audit model and data properties. We do this by designing a protocol called ZkAudit in which model providers publish cryptographic commitments of datasets and model weights, alongside a zero-knowledge proof (ZKP) certifying that published commitments are derived from training the model. Model providers can then respond to audit requests by privately computing any function F of the dataset (or model) and releasing the output of F alongside another ZKP certifying the correct execution of F. To enable ZkAudit, we develop new methods of computing ZKPs for SGD on modern neural nets for simple recommender systems and image classification models capable of high accuracies on ImageNet. Empirically, we show it is possible to provide trustless audits of DNNs, including copyright, censorship, and counterfactual audits with little to no loss in accuracy."
                    },
                    {
                        "title": "Managing Bandwidth: The Key to Cloud-Assisted Autonomous Driving",
                        "abstract": "Prevailing wisdom asserts that one cannot rely on the cloud for critical real-time control systems like self-driving cars. We argue that we can, and must. Following the trends of increasing model sizes, improvements in hardware, and evolving mobile networks, we identify an opportunity to offload parts of time-sensitive and latency-critical compute to the cloud. Doing so requires carefully allocating bandwidth to meet strict latency SLOs, while maximizing benefit to the car."
                    },
                    {
                        "title": "Shark: SQL and Rich Analytics at Scale",
                        "abstract": "Shark is a new data analysis system that marries query processing with complex analytics on large clusters. It leverages a novel distributed memory abstraction to provide a unified engine that can run SQL queries and sophisticated analytics functions (e.g., iterative machine learning) at scale, and efficiently recovers from failures mid-query. This allows Shark to run SQL queries up to 100x faster than Apache Hive, and machine learning programs up to 100x faster than Hadoop. Unlike previous systems, Shark shows that it is possible to achieve these speedups while retaining a MapReduce-like execution engine, and the fine-grained fault tolerance properties that such engines provide. It extends such an engine in several ways, including column-oriented in-memory storage and dynamic mid-query replanning, to effectively execute SQL. The result is a system that matches the speedups reported for MPP analytic databases over MapReduce, while offering fault tolerance properties and complex analytics capabilities that they lack."
                    },
                    {
                        "title": "Highly Available Transactions: Virtues and Limitations (Extended Version)",
                        "abstract": "To minimize network latency and remain online during server failures and network partitions, many modern distributed data storage systems eschew transactional functionality, which provides strong semantic guarantees for groups of multiple operations over multiple data items. In this work, we consider the problem of providing Highly Available Transactions (HATs): transactional guarantees that do not suffer unavailability during system partitions or incur high network latency. We introduce a taxonomy of highly available systems and analyze existing ACID isolation and distributed data consistency guarantees to identify which can and cannot be achieved in HAT systems. This unifies the literature on weak transactional isolation, replica consistency, and highly available systems. We analytically and experimentally quantify the availability and performance benefits of HATs--often two to three orders of magnitude over wide-area networks--and discuss their necessary semantic compromises."
                    }
                ]
            }
        }
    },
    "2406.04801": {
        "paper_data": {
            "title": "MoE Jetpack: From Dense Checkpoints to Adaptive Mixture of Experts for Vision Tasks",
            "url": "http://arxiv.org/abs/2406.04801v1",
            "arxiv_id": "2406.04801",
            "authors": [
                "Xingkui Zhu",
                "Yiran Guan",
                "Dingkang Liang",
                "Yuchao Chen",
                "Yuliang Liu",
                "Xiang Bai"
            ],
            "abstract": "The sparsely activated mixture of experts (MoE) model presents a promising alternative to traditional densely activated (dense) models, enhancing both quality and computational efficiency. However, training MoE models from scratch demands extensive data and computational resources. Moreover, public repositories like timm mainly provide pre-trained dense checkpoints, lacking similar resources for MoE models, hindering their adoption. To bridge this gap, we introduce MoE Jetpack, an effective method for fine-tuning dense checkpoints into MoE models. MoE Jetpack incorporates two key techniques: (1) checkpoint recycling, which repurposes dense checkpoints as initial weights for MoE models, thereby accelerating convergence, enhancing accuracy, and alleviating the computational burden of pre-training; (2) hyperspherical adaptive MoE (SpheroMoE) layer, which optimizes the MoE architecture for better integration of dense checkpoints, enhancing fine-tuning performance. Our experiments on vision tasks demonstrate that MoE Jetpack significantly improves convergence speed and accuracy when fine-tuning dense checkpoints into MoE models. Our code will be publicly available at https://github.com/Adlith/MoE-Jetpack.",
            "introduction": "   1 Introduction  Increased scale is one of the key factors boosting performance in deep learning\u00a0[1, 2, 3]. However, as models expand in size, their computational demands surge, resulting in considerable slowdowns during both training and inference phases. A promising approach that decouples model size from computational costs is the sparsely activated mixture of experts (MoE)\u00a0[4, 5, 6]. Unlike densely activated models (referred to as dense models hereafter)\u00a0[7, 8] apply the full network parameters to all inputs, MoE dynamically activates different pieces of the model for distinct input tokens. This allows for model scaling without substantially increasing the FLOPs111FLOPs means the floating point operations per second. The vanilla design of MoE does not inherently provide runtime advantages and requires additional parallelization strategies\u00a0[9, 10] for acceleration. In our implementation, we offer an effective matrix multiplication method for parallelization, detailed in Appendix\u00a0C., thereby maintaining training and inference speeds during model upscaling. Recent advancements have seen successful implementations of MoE across various domains\u00a0[11, 12, 13].   Despite their potential, MoE models face significant adoption challenges primarily due to the lack of pre-trained models. Unlike dense models, which benefit from a rich repository of pre-trained models accessible through communities like Hugging Face\u00a0[14] (\u223csimilar-to\\sim\u223c81k models) and Timm\u00a0[15] (\u223csimilar-to\\sim\u223c800 models), most MoE models must be trained from scratch using randomly initialized weights. The absence of pre-trained weights necessitates substantial GPU hours and extensive data for training Mixture of Experts (MoE) models, thereby restricting MoE research to a limited number of research teams. Consequently, our research aims to reduce the training time and data requirements for MoE models by leveraging the pre-trained knowledge from dense checkpoints. We will specifically investigate whether utilizing dense checkpoints can enhance the accuracy and convergence speed of MoE models during fine-tuning.   Figure 1:  (a) Our MoE Jetpack converts pre-trained dense models into MoE models, enhancing convergence and performance while maintaining equivalent FLOPs. Here, Exp. represents individual experts, E\ud835\udc38Eitalic_E denotes the number of experts, and L\ud835\udc3fLitalic_L indicates the total number of layers. (b) Performance comparison of ViT trained from scratch, pre-trained ViT, Soft MoE\u00a0[6] trained from scratch, and MoE Jetpack across various datasets. MoE Jetpack shows significant performance improvements.    In this paper, we propose MoE Jetpack, a new approach for fine-tuning pre-trained dense checkpoints into MoE models. As illustrated in Fig.\u00a01(a), MoE Jetpack leverages the sunk cost of dense pre-training to enhance MoE model performance and accelerate convergence. It comprises two key techniques. The first is checkpoint recycling, which initializes MoE models using dense checkpoints. Unlike sparse upcycling\u00a0[16], which merely copies the Multilayer Perceptron (MLP) to construct experts, checkpoint recycling leverages various dense checkpoints and multiple weight selection methods. This approach provides greater flexibility and results in superior MoE initialization weights. The second technique is the hyperspherical adaptive MoE (SpheroMoE) layer, which presents an optimized MoE architecture for seamless integration of dense checkpoints and enhanced fine-tuning performance. Existing MoE architectures, such as Switch Transformers\u00a0[4] and Soft MoE\u00a0[6], are not designed to leverage pre-existing dense checkpoints, which may lead to optimization and over-specialization challenges during fine-tuning. The SpheroMoE layer mitigates these challenges by normalized token mixing, expert regularization, and adaptive dual-path.   By equipping dense checkpoints with",
            "references": [
                {
                    "title": "MoE-LLaVA: Mixture of Experts for Large Vision-Language Models",
                    "abstract": "Recent advances demonstrate that scaling Large Vision-Language Models (LVLMs) effectively improves downstream task performances. However, existing scaling methods enable all model parameters to be active for each token in the calculation, which brings massive training and inferring costs. In this work, we propose a simple yet effective training strategy MoE-Tuning for LVLMs. This strategy innovatively addresses the common issue of performance degradation in multi-modal sparsity learning, consequently constructing a sparse model with an outrageous number of parameters but a constant computational cost. Furthermore, we present the MoE-LLaVA, a MoE-based sparse LVLM architecture, which uniquely activates only the top-k experts through routers during deployment, keeping the remaining experts inactive. Extensive experiments show the significant performance of MoE-LLaVA in a variety of visual understanding and object hallucination benchmarks. Remarkably, with only approximately 3B sparsely activated parameters, MoE-LLaVA demonstrates performance comparable to the LLaVA-1.5-7B on various visual understanding datasets and even surpasses the LLaVA-1.5-13B in object hallucination benchmark. Through MoE-LLaVA, we aim to establish a baseline for sparse LVLMs and provide valuable insights for future research in developing more efficient and effective multi-modal learning systems. Code is released at https://github.com/PKU-YuanGroup/MoE-LLaVA."
                },
                {
                    "title": "SliceGPT: Compress Large Language Models by Deleting Rows and Columns",
                    "abstract": "Large language models have become the cornerstone of natural language processing, but their use comes with substantial costs in terms of compute and memory resources. Sparsification provides a solution to alleviate these resource constraints, and recent works have shown that trained models can be sparsified post-hoc. Existing sparsification techniques face challenges as they need additional data structures and offer constrained speedup with current hardware. In this paper we present SliceGPT, a new post-training sparsification scheme which replaces each weight matrix with a smaller (dense) matrix, reducing the embedding dimension of the network. Through extensive experimentation, we show that SliceGPT can remove up to 25% of the model parameters (including embeddings) for LLAMA2-70B, OPT 66B and Phi-2 models while maintaining 99%, 99% and 90% zero-shot task performance of the dense model respectively. Our sliced models run on fewer GPUs and run faster without any additional code optimization: on 24GB consumer GPUs we reduce the total compute for inference on LLAMA2-70B to 64% of that of the dense model; on 40GB A100 GPUs we reduce it to 66%. We offer a new insight, computational invariance in transformer networks, which enables SliceGPT and we hope it will inspire and enable future avenues to reduce memory and computation demands for pre-trained models. Code is available at: https://github.com/microsoft/TransformerCompression"
                },
                {
                    "title": "DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models",
                    "abstract": "In the era of large language models, Mixture-of-Experts (MoE) is a promising architecture for managing computational costs when scaling up model parameters. However, conventional MoE architectures like GShard, which activate the top-$K$ out of $N$ experts, face challenges in ensuring expert specialization, i.e. each expert acquires non-overlapping and focused knowledge. In response, we propose the DeepSeekMoE architecture towards ultimate expert specialization. It involves two principal strategies: (1) finely segmenting the experts into $mN$ ones and activating $mK$ from them, allowing for a more flexible combination of activated experts; (2) isolating $K_s$ experts as shared ones, aiming at capturing common knowledge and mitigating redundancy in routed experts. Starting from a modest scale with 2B parameters, we demonstrate that DeepSeekMoE 2B achieves comparable performance with GShard 2.9B, which has 1.5 times the expert parameters and computation. In addition, DeepSeekMoE 2B nearly approaches the performance of its dense counterpart with the same number of total parameters, which set the upper bound of MoE models. Subsequently, we scale up DeepSeekMoE to 16B parameters and show that it achieves comparable performance with LLaMA2 7B, with only about 40% of computations. Further, our preliminary efforts to scale up DeepSeekMoE to 145B parameters consistently validate its substantial advantages over the GShard architecture, and show its performance comparable with DeepSeek 67B, using only 28.5% (maybe even 18.2%) of computations."
                },
                {
                    "title": "Mixtral of Experts",
                    "abstract": "We introduce Mixtral 8x7B, a Sparse Mixture of Experts (SMoE) language model. Mixtral has the same architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks (i.e. experts). For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, each token has access to 47B parameters, but only uses 13B active parameters during inference. Mixtral was trained with a context size of 32k tokens and it outperforms or matches Llama 2 70B and GPT-3.5 across all evaluated benchmarks. In particular, Mixtral vastly outperforms Llama 2 70B on mathematics, code generation, and multilingual benchmarks. We also provide a model fine-tuned to follow instructions, Mixtral 8x7B - Instruct, that surpasses GPT-3.5 Turbo, Claude-2.1, Gemini Pro, and Llama 2 70B - chat model on human benchmarks. Both the base and instruct models are released under the Apache 2.0 license."
                },
                {
                    "title": "Initializing Models with Larger Ones",
                    "abstract": "Weight initialization plays an important role in neural network training. Widely used initialization methods are proposed and evaluated for networks that are trained from scratch. However, the growing number of pretrained models now offers new opportunities for tackling this classical problem of weight initialization. In this work, we introduce weight selection, a method for initializing smaller models by selecting a subset of weights from a pretrained larger model. This enables the transfer of knowledge from pretrained weights to smaller models. Our experiments demonstrate that weight selection can significantly enhance the performance of small models and reduce their training time. Notably, it can also be used together with knowledge distillation. Weight selection offers a new approach to leverage the power of pretrained models in resource-constrained settings, and we hope it can be a useful tool for training small models in the large-model era. Code is available at https://github.com/OscarXZQ/weight-selection."
                },
                {
                    "title": "Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning",
                    "abstract": "The popularity of LLaMA (Touvron et al., 2023a;b) and other recently emerged moderate-sized large language models (LLMs) highlights the potential of building smaller yet powerful LLMs. Regardless, the cost of training such models from scratch on trillions of tokens remains high. In this work, we study structured pruning as an effective means to develop smaller LLMs from pre-trained, larger models. Our approach employs two key techniques: (1) targeted structured pruning, which prunes a larger model to a specified target shape by removing layers, heads, and intermediate and hidden dimensions in an end-to-end manner, and (2) dynamic batch loading, which dynamically updates the composition of sampled data in each training batch based on varying losses across different domains. We demonstrate the efficacy of our approach by presenting the Sheared-LLaMA series, pruning the LLaMA2-7B model down to 1.3B and 2.7B parameters. Sheared-LLaMA models outperform state-of-the-art open-source models of equivalent sizes, such as Pythia, INCITE, OpenLLaMA and the concurrent TinyLlama models, on a wide range of downstream and instruction tuning evaluations, while requiring only 3% of compute compared to training such models from scratch. This work provides compelling evidence that leveraging existing LLMs with structured pruning is a far more cost-effective approach for building competitive small-scale LLMs"
                },
                {
                    "title": "From Sparse to Soft Mixtures of Experts",
                    "abstract": "Sparse mixture of expert architectures (MoEs) scale model capacity without significant increases in training or inference costs. Despite their success, MoEs suffer from a number of issues: training instability, token dropping, inability to scale the number of experts, or ineffective finetuning. In this work, we propose Soft MoE, a fully-differentiable sparse Transformer that addresses these challenges, while maintaining the benefits of MoEs. Soft MoE performs an implicit soft assignment by passing different weighted combinations of all input tokens to each expert. As in other MoEs, experts in Soft MoE only process a subset of the (combined) tokens, enabling larger model capacity (and performance) at lower inference cost. In the context of visual recognition, Soft MoE greatly outperforms dense Transformers (ViTs) and popular MoEs (Tokens Choice and Experts Choice). Furthermore, Soft MoE scales well: Soft MoE Huge/14 with 128 experts in 16 MoE layers has over 40x more parameters than ViT Huge/14, with only 2% increased inference time, and substantially better quality."
                },
                {
                    "title": "Soft Merging of Experts with Adaptive Routing",
                    "abstract": "Sparsely activated neural networks with conditional computation learn to route their inputs through different\"expert\"subnetworks, providing a form of modularity that densely activated models lack. Despite their possible benefits, models with learned routing often underperform their parameter-matched densely activated counterparts as well as models that use non-learned heuristic routing strategies. In this paper, we hypothesize that these shortcomings stem from the gradient estimation techniques used to train sparsely activated models that use non-differentiable discrete routing decisions. To address this issue, we introduce Soft Merging of Experts with Adaptive Routing (SMEAR), which avoids discrete routing by using a single\"merged\"expert constructed via a weighted average of all of the experts' parameters. By routing activations through a single merged expert, SMEAR does not incur a significant increase in computational costs and enables standard gradient-based training. We empirically validate that models using SMEAR outperform models that route based on metadata or learn sparse routing through gradient estimation. Furthermore, we provide qualitative analysis demonstrating that the experts learned via SMEAR exhibit a significant amount of specialization. All of the code used in our experiments is publicly available."
                },
                {
                    "title": "DINOv2: Learning Robust Visual Features without Supervision",
                    "abstract": "The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels."
                },
                {
                    "title": "EVA-CLIP: Improved Training Techniques for CLIP at Scale",
                    "abstract": "Contrastive language-image pre-training, CLIP for short, has gained increasing attention for its potential in various scenarios. In this paper, we propose EVA-CLIP, a series of models that significantly improve the efficiency and effectiveness of CLIP training. Our approach incorporates new techniques for representation learning, optimization, and augmentation, enabling EVA-CLIP to achieve superior performance compared to previous CLIP models with the same number of parameters but significantly smaller training costs. Notably, our largest 5.0B-parameter EVA-02-CLIP-E/14+ with only 9 billion seen samples achieves 82.0 zero-shot top-1 accuracy on ImageNet-1K val. A smaller EVA-02-CLIP-L/14+ with only 430 million parameters and 6 billion seen samples achieves 80.4 zero-shot top-1 accuracy on ImageNet-1K val. To facilitate open access and open research, we release the complete suite of EVA-CLIP to the community at https://github.com/baaivision/EVA/tree/master/EVA-CLIP."
                },
                {
                    "title": "Sparse Upcycling: Training Mixture-of-Experts from Dense Checkpoints",
                    "abstract": "Training large, deep neural networks to convergence can be prohibitively expensive. As a result, often only a small selection of popular, dense models are reused across different contexts and tasks. Increasingly, sparsely activated models, which seek to decouple model size from computation costs, are becoming an attractive alternative to dense models. Although more efficient in terms of quality and computation cost, sparse models remain data-hungry and costly to train from scratch in the large scale regime. In this work, we propose sparse upcycling -- a simple way to reuse sunk training costs by initializing a sparsely activated Mixture-of-Experts model from a dense checkpoint. We show that sparsely upcycled T5 Base, Large, and XL language models and Vision Transformer Base and Large models, respectively, significantly outperform their dense counterparts on SuperGLUE and ImageNet, using only ~50% of the initial dense pretraining sunk cost. The upcycled models also outperform sparse models trained from scratch on 100% of the initial dense pretraining computation budget."
                },
                {
                    "title": "M3ViT: Mixture-of-Experts Vision Transformer for Efficient Multi-task Learning with Model-Accelerator Co-design",
                    "abstract": "Multi-task learning (MTL) encapsulates multiple learned tasks in a single model and often lets those tasks learn better jointly. However, when deploying MTL onto those real-world systems that are often resource-constrained or latency-sensitive, two prominent challenges arise: (i) during training, simultaneously optimizing all tasks is often difficult due to gradient conflicts across tasks; (ii) at inference, current MTL regimes have to activate nearly the entire model even to just execute a single task. Yet most real systems demand only one or two tasks at each moment, and switch between tasks as needed: therefore such all tasks activated inference is also highly inefficient and non-scalable. In this paper, we present a model-accelerator co-design framework to enable efficient on-device MTL. Our framework, dubbed M$^3$ViT, customizes mixture-of-experts (MoE) layers into a vision transformer (ViT) backbone for MTL, and sparsely activates task-specific experts during training. Then at inference with any task of interest, the same design allows for activating only the task-corresponding sparse expert pathway, instead of the full model. Our new model design is further enhanced by hardware-level innovations, in particular, a novel computation reordering scheme tailored for memory-constrained MTL that achieves zero-overhead switching between tasks and can scale to any number of experts. When executing single-task inference, M$^{3}$ViT achieves higher accuracies than encoder-focused MTL methods, while significantly reducing 88% inference FLOPs. When implemented on a hardware platform of one Xilinx ZCU104 FPGA, our co-design framework reduces the memory requirement by 2.4 times, while achieving energy efficiency up to 9.23 times higher than a comparable FPGA baseline. Code is available at: https://github.com/VITA-Group/M3ViT."
                },
                {
                    "title": "Width & Depth Pruning for Vision Transformers",
                    "abstract": "Transformer models have demonstrated their promising potential and achieved excellent performance on a series of computer vision tasks. However, the huge computational cost of vision transformers hinders their deployment and application to edge devices. Recent works have proposed to \ufb01nd and remove the unimportant units of vision transformers. Despite achieving remarkable results, these methods take one dimension of network width into consideration and ignore network depth, which is another important dimension for pruning vision transformers. Therefore, we propose a Width & Depth Pruning (WDPruning) framework that reduces both width and depth dimensions simultaneously. Speci\ufb01cally, for width pruning, a set of learnable pruning-related parameters is used to adaptively adjust the width of transformer. For depth pruning, we introduce several shallow classi\ufb01ers by using the intermediate information of the transformer blocks, which allows images to be classi\ufb01ed by shallow classi\ufb01ers instead of the deeper classi\ufb01ers. In the inference period, all of the blocks after shallow classi\ufb01ers can be dropped so they don\u2019t bring additional parameters and computation. Experimental results on benchmark datasets demonstrate that the proposed method can signi\ufb01cantly reduce the computational costs of mainstream vision transformers such as DeiT and Swin Transformer with a minor accuracy drop. In particular, on ILSVRC-12, we achieve over 22% pruning ratio of FLOPs by compressing DeiT-Base, even with an increase of 0.14% Top-1 accuracy."
                },
                {
                    "title": "Mixture-of-Experts with Expert Choice Routing",
                    "abstract": "Sparsely-activated Mixture-of-experts (MoE) models allow the number of parameters to greatly increase while keeping the amount of computation for a given token or a given sample unchanged. However, a poor expert routing strategy (e.g. one resulting in load imbalance) can cause certain experts to be under-trained, leading to an expert being under or over-specialized. Prior work allocates a fixed number of experts to each token using a top-k function regardless of the relative importance of different tokens. To address this, we propose a heterogeneous mixture-of-experts employing an expert choice method. Instead of letting tokens select the top-k experts, we have experts selecting the top-k tokens. As a result, each token can be routed to a variable number of experts and each expert can have a fixed bucket size. We systematically study pre-training speedups using the same computational resources of the Switch Transformer top-1 and GShard top-2 gating of prior work and find that our method improves training convergence time by more than 2x. For the same computational cost, our method demonstrates higher performance in fine-tuning 11 selected tasks in the GLUE and SuperGLUE benchmarks. For a smaller activation cost, our method outperforms the T5 dense model in 7 out of the 11 tasks."
                },
                {
                    "title": "Unified Scaling Laws for Routed Language Models",
                    "abstract": "The performance of a language model has been shown to be effectively modeled as a power-law in its parameter count. Here we study the scaling behaviors of Routing Networks: architectures that conditionally use only a subset of their parameters while processing an input. For these models, parameter count and computational requirement form two independent axes along which an increase leads to better performance. In this work we derive and justify scaling laws defined on these two variables which generalize those known for standard language models and describe the performance of a wide range of routing architectures trained via three different techniques. Afterwards we provide two applications of these laws: first deriving an Effective Parameter Count along which all models scale at the same rate, and then using the scaling coefficients to give a quantitative comparison of the three routing techniques considered. Our analysis derives from an extensive evaluation of Routing Networks across five orders of magnitude of size, including models with hundreds of experts and hundreds of billions of parameters."
                },
                {
                    "title": "DeepSpeed-MoE: Advancing Mixture-of-Experts Inference and Training to Power Next-Generation AI Scale",
                    "abstract": "As the training of giant dense models hits the boundary on the availability and capability of the hardware resources today, Mixture-of-Experts (MoE) models become one of the most promising model architectures due to their significant training cost reduction compared to a quality-equivalent dense model. Its training cost saving is demonstrated from encoder-decoder models (prior works) to a 5x saving for auto-aggressive language models (this work along with parallel explorations). However, due to the much larger model size and unique architecture, how to provide fast MoE model inference remains challenging and unsolved, limiting its practical usage. To tackle this, we present DeepSpeed-MoE, an end-to-end MoE training and inference solution as part of the DeepSpeed library, including novel MoE architecture designs and model compression techniques that reduce MoE model size by up to 3.7x, and a highly optimized inference system that provides 7.3x better latency and cost compared to existing MoE inference solutions. DeepSpeed-MoE offers an unprecedented scale and efficiency to serve massive MoE models with up to 4.5x faster and 9x cheaper inference compared to quality-equivalent dense models. We hope our innovations and systems help open a promising path to new directions in the large model landscape, a shift from dense to sparse MoE models, where training and deploying higher-quality models with fewer resources becomes more widely possible."
                },
                {
                    "title": "A ConvNet for the 2020s",
                    "abstract": "The \u201cRoaring 20s\u201d of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually \u201cmodernize\u201d a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "MoEfication: Transformer Feed-forward Layers are Mixtures of Experts",
                    "abstract": "Recent work has shown that feed-forward networks (FFNs) in pre-trained Transformers are a key component, storing various linguistic and factual knowledge. However, the computational patterns of FFNs are still unclear. In this work, we study the computational patterns of FFNs and observe that most inputs only activate a tiny ratio of neurons of FFNs. This phenomenon is similar to the sparsity of the human brain, which drives research on functional partitions of the human brain. To verify whether functional partitions also emerge in FFNs, we propose to convert a model into its MoE version with the same parameters, namely MoEfication. Specifically, MoEfication consists of two phases: (1) splitting the parameters of FFNs into multiple functional partitions as experts, and (2) building expert routers to decide which experts will be used for each input. Experimental results show that MoEfication can conditionally use 10% to 30% of FFN parameters while maintaining over 95% original performance for different models on various downstream tasks. Besides, MoEfication brings two advantages: (1) it significantly reduces the FLOPS of inference, i.e., 2x speedup with 25% of FFN parameters, and (2) it provides a fine-grained perspective to study the inner mechanism of FFNs. The source code of this paper can be obtained from https://github.com/thunlp/MoEfication."
                },
                {
                    "title": "Scaling Vision with Sparse Mixture of Experts",
                    "abstract": "Sparsely-gated Mixture of Experts networks (MoEs) have demonstrated excellent scalability in Natural Language Processing. In Computer Vision, however, almost all performant networks are\"dense\", that is, every input is processed by every parameter. We present a Vision MoE (V-MoE), a sparse version of the Vision Transformer, that is scalable and competitive with the largest dense networks. When applied to image recognition, V-MoE matches the performance of state-of-the-art networks, while requiring as little as half of the compute at inference time. Further, we propose an extension to the routing algorithm that can prioritize subsets of each input across the entire batch, leading to adaptive per-image compute. This allows V-MoE to trade-off performance and compute smoothly at test-time. Finally, we demonstrate the potential of V-MoE to scale vision models, and train a 15B parameter model that attains 90.35% on ImageNet."
                },
                {
                    "title": "Scaling Vision Transformers",
                    "abstract": "Attention-based neural networks such as the Vision Transformer (ViT) have recently attained state-of-the-art results on many computer vision benchmarks. Scale is a primary ingredient in attaining excellent results, therefore, understanding a model's scaling properties is a key to designing future generations effectively. While the laws for scaling Transformer language models have been studied, it is unknown how Vision Transformers scale. To address this, we scale ViT models and data, both up and down, and characterize the relationships between error rate, data, and compute. Along the way, we refine the architecture and training of ViT, reducing memory consumption and increasing accuracy of the resulting models. As a result, we successfully train a ViT model with two billion parameters, which attains a new state-of-the-art on ImageNet of 90.45% top-1 accuracy. The model also performs well for few-shot transfer, for example, reaching 84.86% top-1 accuracy on ImageNet with only 10 examples per class."
                },
                {
                    "title": "Hash Layers For Large Sparse Models",
                    "abstract": "We investigate the training of sparse layers that use different parameters for different inputs based on hashing in large Transformer models. Specifically, we modify the feedforward layer to hash to different sets of weights depending on the current token, over all tokens in the sequence. We show that this procedure either outperforms or is competitive with learning-to-route mixture-of-expert methods such as Switch Transformers and BASE Layers, while requiring no routing parameters or extra terms in the objective function such as a load balancing loss, and no sophisticated assignment algorithm. We study the performance of different hashing techniques, hash sizes and input features, and show that balanced and random hashes focused on the most local features work best, compared to either learning clusters or using longer-range context. We show our approach works well both on large language modeling and dialogue tasks, and on downstream fine-tuning tasks."
                },
                {
                    "title": "Going deeper with Image Transformers",
                    "abstract": "Transformers have been recently adapted for large scale image classification, achieving high scores shaking up the long supremacy of convolutional neural networks. However the optimization of vision transformers has been little studied so far. In this work, we build and optimize deeper transformer networks for image classification. In particular, we investigate the interplay of architecture and optimization of such dedicated transformers. We make two architecture changes that significantly improve the accuracy of deep transformers. This leads us to produce models whose performance does not saturate early with more depth, for in-stance we obtain 86.5% top-1 accuracy on Imagenet when training with no external data, we thus attain the current sate of the art with less floating-point operations and parameters. Our best model establishes the new state of the art on Imagenet with Reassessed labels and Imagenet-V2 / match frequency, in the setting with no additional training data. We share our code and models1."
                },
                {
                    "title": "BASE Layers: Simplifying Training of Large, Sparse Models",
                    "abstract": "We introduce a new balanced assignment of experts (BASE) layer for large language models that greatly simplifies existing high capacity sparse layers. Sparse layers can dramatically improve the efficiency of training and inference by routing each token to specialized expert modules that contain only a small fraction of the model parameters. However, it can be difficult to learn balanced routing functions that make full use of the available experts; existing approaches typically use routing heuristics or auxiliary expert-balancing loss functions. In contrast, we formulate token-to-expert allocation as a linear assignment problem, allowing an optimal assignment in which each expert receives an equal number of tokens. This optimal assignment scheme improves efficiency by guaranteeing balanced compute loads, and also simplifies training by not requiring any new hyperparameters or auxiliary losses. Code is publicly released at https://github.com/pytorch/fairseq/"
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity",
                    "abstract": "In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with outrageous numbers of parameters -- but a constant computational cost. However, despite several notable successes of MoE, widespread adoption has been hindered by complexity, communication costs and training instability -- we address these with the Switch Transformer. We simplify the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs. Our proposed training techniques help wrangle the instabilities and we show large sparse models may be trained, for the first time, with lower precision (bfloat16) formats. We design models based off T5-Base and T5-Large to obtain up to 7x increases in pre-training speed with the same computational resources. These improvements extend into multilingual settings where we measure gains over the mT5-Base version across all 101 languages. Finally, we advance the current scale of language models by pre-training up to trillion parameter models on the\"Colossal Clean Crawled Corpus\"and achieve a 4x speedup over the T5-XXL model."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding",
                    "abstract": "Neural network scaling has been critical for improving the model quality in many real-world machine learning applications with vast amounts of training data and compute. Although this trend of scaling is affirmed to be a sure-fire approach for better model quality, there are challenges on the path such as the computation cost, ease of programming, and efficient implementation on parallel devices. GShard is a module composed of a set of lightweight annotation APIs and an extension to the XLA compiler. It provides an elegant way to express a wide range of parallel computation patterns with minimal changes to the existing model code. GShard enabled us to scale up multilingual neural machine translation Transformer model with Sparsely-Gated Mixture-of-Experts beyond 600 billion parameters using automatic sharding. We demonstrate that such a giant model can efficiently be trained on 2048 TPU v3 accelerators in 4 days to achieve far superior quality for translation from 100 languages to English compared to the prior art."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "HuggingFace's Transformers: State-of-the-art Natural Language Processing",
                    "abstract": "Recent progress in natural language processing has been driven by advances in both model architecture and model pretraining. Transformer architectures have facilitated building higher-capacity models and pretraining has made it possible to effectively utilize this capacity for a wide variety of tasks. \\textit{Transformers} is an open-source library with the goal of opening up these advances to the wider machine learning community. The library consists of carefully engineered state-of-the art Transformer architectures under a unified API. Backing this library is a curated collection of pretrained models made by and available for the community. \\textit{Transformers} is designed to be extensible by researchers, simple for practitioners, and fast and robust in industrial deployments. The library is available at \\url{this https URL}."
                },
                {
                    "title": "Randaugment: Practical automated data augmentation with a reduced search space",
                    "abstract": "Recent work on automated augmentation strategies has led to state-of-the-art results in image classification and object detection. An obstacle to a large-scale adoption of these methods is that they require a separate and expensive search phase. A common way to overcome the expense of the search phase was to use a smaller proxy task. However, it was not clear if the optimized hyperparameters found on the proxy task are also optimal for the actual task. In this work, we rethink the process of designing automated augmentation strategies. We find that while previous work required a search for both magnitude and probability of each operation independently, it is sufficient to only search for a single distortion magnitude that jointly controls all operations. We hence propose a simplified search space that vastly reduces the computational expense of automated augmentation, and permits the removal of a separate proxy task.Despite the simplifications, our method achieves equal or better performance over previous automated augmentation strategies on on CIFAR-10/100, SVHN, ImageNet and COCO datasets. EfficientNet-B7, we achieve 85.0% accuracy, a 1.0% increase over baseline augmentation, a 0.6% improvement over AutoAugment on the ImageNet dataset. With EfficientNet-B8, we achieve 85.4% accuracy on ImageNet, which matches a previous result that used 3.5B extra images. On object detection, the same method as classification leads to 1.0-1.3% improvement over baseline augmentation. Code will be made available online."
                },
                {
                    "title": "CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features",
                    "abstract": "Regional dropout strategies have been proposed to enhance performance of convolutional neural network classifiers. They have proved to be effective for guiding the model to attend on less discriminative parts of objects (e.g. leg as opposed to head of a person), thereby letting the network generalize better and have better object localization capabilities. On the other hand, current methods for regional dropout removes informative pixels on training images by overlaying a patch of either black pixels or random noise. Such removal is not desirable because it suffers from information loss causing inefficiency in training. We therefore propose the CutMix augmentation strategy: patches are cut and pasted among training images where the ground truth labels are also mixed proportionally to the area of the patches. By making efficient use of training pixels and retaining the regularization effect of regional dropout, CutMix consistently outperforms state-of-the-art augmentation strategies on CIFAR and ImageNet classification tasks, as well as on ImageNet weakly-supervised localization task. Moreover, unlike previous augmentation methods, our CutMix-trained ImageNet classifier, when used as a pretrained model, results in consistent performance gain in Pascal detection and MS-COCO image captioning benchmarks. We also show that CutMix can improve the model robustness against input corruptions and its out-of distribution detection performance."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "mixup: Beyond Empirical Risk Minimization",
                    "abstract": "Large deep neural networks are powerful, but exhibit undesirable behaviors such as memorization and sensitivity to adversarial examples. In this work, we propose mixup, a simple learning principle to alleviate these issues. In essence, mixup trains a neural network on convex combinations of pairs of examples and their labels. By doing so, mixup regularizes the neural network to favor simple linear behavior in-between training examples. Our experiments on the ImageNet-2012, CIFAR-10, CIFAR-100, Google commands and UCI datasets show that mixup improves the generalization of state-of-the-art neural network architectures. We also find that mixup reduces the memorization of corrupt labels, increases the robustness to adversarial examples, and stabilizes the training of generative adversarial networks."
                },
                {
                    "title": "Random Erasing Data Augmentation",
                    "abstract": "In this paper, we introduce Random Erasing, a new data augmentation method for training the convolutional neural network (CNN). In training, Random Erasing randomly selects a rectangle region in an image and erases its pixels with random values. In this process, training images with various levels of occlusion are generated, which reduces the risk of over-fitting and makes the model robust to occlusion. Random Erasing is parameter learning free, easy to implement, and can be integrated with most of the CNN-based recognition models. Albeit simple, Random Erasing is complementary to commonly used data augmentation techniques such as random cropping and flipping, and yields consistent improvement over strong baselines in image classification, object detection and person re-identification. Code is available at: https://github.com/zhunzhong07/Random-Erasing."
                },
                {
                    "title": "Revisiting Unreasonable Effectiveness of Data in Deep Learning Era",
                    "abstract": "The success of deep learning in vision can be attributed to: (a) models with high capacity; (b) increased computational power; and (c) availability of large-scale labeled data. Since 2012, there have been significant advances in representation capabilities of the models and computational capabilities of GPUs. But the size of the biggest dataset has surprisingly remained constant. What will happen if we increase the dataset size by 10 \u00d7 or 100 \u00d7 ? This paper takes a step towards clearing the clouds of mystery surrounding the relationship between \u2018enormous data\u2019 and visual deep learning. By exploiting the JFT-300M dataset which has more than 375M noisy labels for 300M images, we investigate how the performance of current vision tasks would change if this data was used for representation learning. Our paper delivers some surprising (and some expected) findings. First, we find that the performance on vision tasks increases logarithmically based on volume of training data size. Second, we show that representation learning (or pre-training) still holds a lot of promise. One can improve performance on many vision tasks by just training a better base model. Finally, as expected, we present new state-of-the-art results for different vision tasks including image classification, object detection, semantic segmentation and human pose estimation. Our sincere hope is that this inspires vision community to not undervalue the data and develop collective efforts in building larger datasets."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer",
                    "abstract": "The capacity of a neural network to absorb information is limited by its number of parameters. Conditional computation, where parts of the network are active on a per-example basis, has been proposed in theory as a way of dramatically increasing model capacity without a proportional increase in computation. In practice, however, there are significant algorithmic and performance challenges. In this work, we address these challenges and finally realize the promise of conditional computation, achieving greater than 1000x improvements in model capacity with only minor losses in computational efficiency on modern GPU clusters. We introduce a Sparsely-Gated Mixture-of-Experts layer (MoE), consisting of up to thousands of feed-forward sub-networks. A trainable gating network determines a sparse combination of these experts to use for each example. We apply the MoE to the tasks of language modeling and machine translation, where model capacity is critical for absorbing the vast quantities of knowledge available in the training corpora. We present model architectures in which a MoE with up to 137 billion parameters is applied convolutionally between stacked LSTM layers. On large language modeling and machine translation benchmarks, these models achieve significantly better results than state-of-the-art at lower computational cost."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Rethinking the Inception Architecture for Computer Vision",
                    "abstract": "Convolutional networks are at the core of most state of-the-art computer vision solutions for a wide variety of tasks. Since 2014 very deep convolutional networks started to become mainstream, yielding substantial gains in various benchmarks. Although increased model size and computational cost tend to translate to immediate quality gains for most tasks (as long as enough labeled data is provided for training), computational efficiency and low parameter count are still enabling factors for various use cases such as mobile vision and big-data scenarios. Here we are exploring ways to scale up networks in ways that aim at utilizing the added computation as efficiently as possible by suitably factorized convolutions and aggressive regularization. We benchmark our methods on the ILSVRC 2012 classification challenge validation set demonstrate substantial gains over the state of the art: 21:2% top-1 and 5:6% top-5 error for single frame evaluation using a network with a computational cost of 5 billion multiply-adds per inference and with using less than 25 million parameters. With an ensemble of 4 models and multi-crop evaluation, we report 3:5% top-5 error and 17:3% top-1 error on the validation set and 3:6% top-5 error on the official test set."
                },
                {
                    "title": "Learning both Weights and Connections for Efficient Neural Network",
                    "abstract": "Neural networks are both computationally intensive and memory intensive, making them difficult to deploy on embedded systems. Also, conventional networks fix the architecture before training starts; as a result, training cannot improve the architecture. To address these limitations, we describe a method to reduce the storage and computation required by neural networks by an order of magnitude without affecting their accuracy by learning only the important connections. Our method prunes redundant connections using a three-step method. First, we train the network to learn which connections are important. Next, we prune the unimportant connections. Finally, we retrain the network to fine tune the weights of the remaining connections. On the ImageNet dataset, our method reduced the number of parameters of AlexNet by a factor of 9x, from 61 million to 6.7 million, without incurring accuracy loss. Similar experiments with VGG-16 found that the number of parameters can be reduced by 13x, from 138 million to 10.3 million, again with no loss of accuracy."
                },
                {
                    "title": "Distilling the Knowledge in a Neural Network",
                    "abstract": "A very simple way to improve the performance of almost any machine learning algorithm is to train many different models on the same data and then to average their predictions. Unfortunately, making predictions using a whole ensemble of models is cumbersome and may be too computationally expensive to allow deployment to a large number of users, especially if the individual models are large neural nets. Caruana and his collaborators have shown that it is possible to compress the knowledge in an ensemble into a single model which is much easier to deploy and we develop this approach further using a different compression technique. We achieve some surprising results on MNIST and we show that we can significantly improve the acoustic model of a heavily used commercial system by distilling the knowledge in an ensemble of models into a single model. We also introduce a new type of ensemble composed of one or more full models and many specialist models which learn to distinguish fine-grained classes that the full models confuse. Unlike a mixture of experts, these specialist models can be trained rapidly and in parallel."
                },
                {
                    "title": "Describing Textures in the Wild",
                    "abstract": "Patterns and textures are key characteristics of many natural objects: a shirt can be striped, the wings of a butterfly can be veined, and the skin of an animal can be scaly. Aiming at supporting this dimension in image understanding, we address the problem of describing textures with semantic attributes. We identify a vocabulary of forty-seven texture terms and use them to describe a large dataset of patterns collected \"in the wild\". The resulting Describable Textures Dataset (DTD) is a basis to seek the best representation for recognizing describable texture attributes in images. We port from object recognition to texture recognition the Improved Fisher Vector (IFV) and Deep Convolutional-network Activation Features (DeCAF), and show that surprisingly, they both outperform specialized texture descriptors not only on our problem, but also in established material recognition datasets. We also show that our describable attributes are excellent texture descriptors, transferring between datasets and tasks, in particular, combined with IFV and DeCAF, they significantly outperform the state-of-the-art by more than 10% on both FMD and KTH-TIPS-2b benchmarks. We also demonstrate that they produce intuitive descriptions of materials and Internet images."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "Cats and dogs",
                    "abstract": "We investigate the fine grained object categorization problem of determining the breed of animal from an image. To this end we introduce a new annotated dataset of pets covering 37 different breeds of cats and dogs. The visual problem is very challenging as these animals, particularly cats, are very deformable and there can be quite subtle differences between the breeds. We make a number of contributions: first, we introduce a model to classify a pet breed automatically from an image. The model combines shape, captured by a deformable part model detecting the pet face, and appearance, captured by a bag-of-words model that describes the pet fur. Fitting the model involves automatically segmenting the animal in the image. Second, we compare two classification approaches: a hierarchical one, in which a pet is first assigned to the cat or dog family and then to a breed, and a flat one, in which the breed is obtained directly. We also investigate a number of animal and image orientated spatial layouts. These models are very good: they beat all previously published results on the challenging ASIRRA test (cat vs dog discrimination). When applied to the task of discriminating the 37 different breeds of pets, the models obtain an average accuracy of about 59%, a very encouraging result considering the difficulty of the problem."
                },
                {
                    "title": "An Analysis of Single-Layer Networks in Unsupervised Feature Learning",
                    "abstract": "A great deal of research has focused on algorithms for learning features from unlabeled data. Indeed, much progress has been made on benchmark datasets like NORB and CIFAR by employing increasingly complex unsupervised learning algorithms and deep models. In this paper, however, we show that several simple factors, such as the number of hidden nodes in the model, may be more important to achieving high performance than the learning algorithm or the depth of the model. Specifically, we will apply several othe-shelf feature learning algorithms (sparse auto-encoders, sparse RBMs, K-means clustering, and Gaussian mixtures) to CIFAR, NORB, and STL datasets using only singlelayer networks. We then present a detailed analysis of the eect of changes in the model setup: the receptive field size, number of hidden nodes (features), the step-size (\u201cstride\u201d) between extracted features, and the eect of whitening. Our results show that large numbers of hidden nodes and dense feature extraction are critical to achieving high performance\u2014so critical, in fact, that when these parameters are pushed to their limits, we achieve state-of-the-art performance on both CIFAR-10 and NORB using only a single layer of features. More surprisingly, our best performance is based on K-means clustering, which is extremely fast, has no hyperparameters to tune beyond the model structure itself, and is very easy to implement. Despite the simplicity of our system, we achieve accuracy beyond all previously published results on the CIFAR-10 and NORB datasets (79.6% and 97.2% respectively)."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Automated Flower Classification over a Large Number of Classes",
                    "abstract": "We investigate to what extent combinations of features can improve classification performance on a large dataset of similar classes. To this end we introduce a 103 class flower dataset. We compute four different features for the flowers, each describing different aspects, namely the local shape/texture, the shape of the boundary, the overall spatial distribution of petals, and the colour. We combine the features using a multiple kernel framework with a SVM classifier. The weights for each class are learnt using the method of Varma and Ray, which has achieved state of the art performance on other large dataset, such as Caltech 101/256. Our dataset has a similar challenge in the number of classes, but with the added difficulty of large between class similarity and small within class similarity. Results show that learning the optimum kernel combination of multiple features vastly improves the performance, from 55.1% for the best single feature to 72.8% for the combination of all features."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.CV",
                "I.2"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively leverage pre-trained dense model checkpoints to enhance the performance and convergence speed of sparsely activated mixture of experts (MoE) models during fine-tuning?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the significant barrier to the adoption of MoE models, which is the lack of pre-trained weights. By enabling the use of dense checkpoints, we can democratize access to MoE models, allowing more researchers to experiment with and build upon this promising architecture. This advancement could lead to improved model performance across various applications, fostering innovation in fields such as natural language processing and computer vision. Furthermore, it could stimulate future research into hybrid model architectures and efficient training methodologies, ultimately pushing the boundaries of what is achievable with deep learning.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent differences between dense and sparse model architectures. Naive approaches that simply transfer weights from dense models to MoE models may fail due to optimization issues and the risk of over-specialization, as MoE models require a different training paradigm. Additionally, the integration of dense checkpoints into MoE architectures necessitates overcoming technical obstacles related to weight selection and expert regularization. The complexity of ensuring that the MoE model can effectively utilize the pre-trained knowledge without losing its sparse activation benefits adds another layer of difficulty.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on either dense models or MoE models in isolation, with limited exploration of their integration. The lack of pre-trained MoE models has created a significant gap, as most existing solutions require training from scratch, which is resource-intensive. Barriers such as the absence of effective methodologies for checkpoint recycling and the challenges of adapting existing MoE architectures to leverage dense checkpoints have prevented progress in this area. Our approach differs by introducing the MoE Jetpack, which utilizes checkpoint recycling and the SpheroMoE layer to facilitate the integration of dense checkpoints, thereby addressing these limitations.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, MoE Jetpack, consists of two key components: (1) **Checkpoint Recycling**, which initializes MoE models using various dense checkpoints and multiple weight selection methods to enhance initialization flexibility and performance; and (2) **SpheroMoE"
            }
        },
        "author_data": {
            "4aa97803-0839-45c9-97b1-a65b11308c51": {
                "pk": "4aa97803-0839-45c9-97b1-a65b11308c51",
                "name": "Xingkui Zhu",
                "collaborators": [
                    "Xiang Bai",
                    "Xin Zhou",
                    "Dingkang Liang",
                    "Wei Xu",
                    "Zhikang Zou",
                    "Shuchang Lyu",
                    "Xu Wang",
                    "Qi Zhao",
                    "Wenwen Yu",
                    "Yuliang Liu"
                ],
                "domain": [
                    "Object Detection",
                    "Point Cloud Analysis",
                    "Transformer",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "TPH-YOLOv5: Improved YOLOv5 Based on Transformer Prediction Head for Object Detection on Drone-captured Scenarios",
                        "abstract": "Object detection on drone-captured scenarios is a recent popular task. As drones always navigate in different altitudes, the object scale varies violently, which burdens the optimization of networks. Moreover, high-speed and low-altitude flight bring in the motion blur on the densely packed objects, which leads to great challenge of object distinction. To solve the two issues mentioned above, we propose TPH-YOLOv5. Based on YOLOv5, we add one more prediction head to detect different-scale objects. Then we replace the original prediction heads with Transformer Prediction Heads (TPH) to explore the prediction potential with self-attention mechanism. We also integrate convolutional block attention model (CBAM) to find attention region on scenarios with dense objects. To achieve more improvement of our proposed TPH-YOLOv5, we provide bags of useful strategies such as data augmentation, multiscale testing, multi-model integration and utilizing extra classifier. Extensive experiments on dataset VisDrone2021 show that TPH-YOLOv5 have good performance with impressive interpretability on drone-captured scenarios. On DET-test-challenge dataset, the AP result of TPH-YOLOv5 are 39.18%, which is better than previous SOTA method (DPNetV3) by 1.81%. On VisDrone Challenge 2021, TPHYOLOv5 wins 5th place and achieves well-matched results with 1st place model (AP 39.43%). Compared to baseline model (YOLOv5), TPH-YOLOv5 improves about 7%, which is encouraging and competitive."
                    },
                    {
                        "title": "Turning a CLIP Model into a Scene Text Spotter",
                        "abstract": "We exploit the potential of the large-scale Contrastive Language-Image Pretraining (CLIP) model to enhance scene text detection and spotting tasks, transforming it into a robust backbone, FastTCM-CR50. This backbone utilizes visual prompt learning and cross-attention in CLIP to extract image and text-based prior knowledge. Using predefined and learnable prompts, FastTCM-CR50 introduces an instance-language matching process to enhance the synergy between image and text embeddings, thereby refining text regions. Our Bimodal Similarity Matching (BSM) module facilitates dynamic language prompt generation, enabling offline computations and improving performance. FastTCM-CR50 offers several advantages: 1) It can enhance existing text detectors and spotters, improving performance by an average of 1.7% and 1.5%, respectively. 2) It outperforms the previous TCM-CR50 backbone, yielding an average improvement of 0.2% and 0.56% in text detection and spotting tasks, along with a 48.5% increase in inference speed. 3) It showcases robust few-shot training capabilities. Utilizing only 10% of the supervised data, FastTCM-CR50 improves performance by an average of 26.5% and 5.5% for text detection and spotting tasks, respectively. 4) It consistently enhances performance on out-of-distribution text detection and spotting datasets, particularly the NightTime-ArT subset from ICDAR2019-ArT and the DOTA dataset for oriented object detection. The code is available at https://github.com/wenwenyu/TCM."
                    },
                    {
                        "title": "Dynamic Adapter Meets Prompt Tuning: Parameter-Efficient Transfer Learning for Point Cloud Analysis",
                        "abstract": "Point cloud analysis has achieved outstanding performance by transferring point cloud pre-trained models. However, existing methods for model adaptation usually update all model parameters, i.e., full fine-tuning paradigm, which is inefficient as it relies on high computational costs (e.g., training GPU memory) and massive storage space. In this paper, we aim to study parameter-efficient transfer learning for point cloud analysis with an ideal trade-off between task performance and parameter efficiency. To achieve this goal, we freeze the parameters of the default pre-trained models and then propose the Dynamic Adapter, which generates a dynamic scale for each token, considering the token significance to the downstream task. We further seamlessly integrate Dynamic Adapter with Prompt Tuning (DAPT) by constructing Internal Prompts, capturing the instance-specific features for interaction. Extensive experiments conducted on five challenging datasets demonstrate that the proposed DAPT achieves superior performance compared to the full fine-tuning counterparts while significantly reducing the trainable parameters and training GPU memory by 95% and 35%, respectively. Code is available at https://github.com/LMD0311/DAPT."
                    },
                    {
                        "title": "PointMamba: A Simple State Space Model for Point Cloud Analysis",
                        "abstract": "Transformers have become one of the foundational architectures in point cloud analysis tasks due to their excellent global modeling ability. However, the attention mechanism has quadratic complexity, making the design of a linear complexity method with global modeling appealing. In this paper, we propose PointMamba, transferring the success of Mamba, a recent representative state space model (SSM), from NLP to point cloud analysis tasks. Unlike traditional Transformers, PointMamba employs a linear complexity algorithm, presenting global modeling capacity while significantly reducing computational costs. Specifically, our method leverages space-filling curves for effective point tokenization and adopts an extremely simple, non-hierarchical Mamba encoder as the backbone. Comprehensive evaluations demonstrate that PointMamba achieves superior performance across multiple datasets while significantly reducing GPU memory usage and FLOPs. This work underscores the potential of SSMs in 3D vision-related tasks and presents a simple yet effective Mamba-based baseline for future research. The code is available at https://github.com/LMD0311/PointMamba."
                    }
                ]
            },
            "5271e72d-c3fd-4860-b1b9-40c99bdea7e6": {
                "pk": "5271e72d-c3fd-4860-b1b9-40c99bdea7e6",
                "name": "Yiran Guan",
                "collaborators": [
                    "Zhuoguang Chen",
                    "Wenzheng Zeng",
                    "Zhiguo Cao",
                    "Yang Xiao",
                    "Jiejun Zhang",
                    "Lingzhi Li",
                    "Ruidong Cao",
                    "Guangying Wang",
                    "Jingxu Chen",
                    "Jianping Yao"
                ],
                "domain": [
                    "Video Gaze Estimation",
                    "Synthetic Dimension",
                    "Quantum Physics",
                    "Photonics"
                ],
                "publications": [
                    {
                        "title": "End-to-end Video Gaze Estimation via Capturing Head-face-eye Spatial-temporal Interaction Context",
                        "abstract": "In this letter, we propose a new method, Multi-Clue Gaze (MCGaze), to facilitate video gaze estimation via capturing spatial-temporal interaction context among head, face, and eye in an end-to-end learning way, which has not been well concerned yet. The main advantage of MCGaze is that the tasks of clue localization of head, face, and eye can be solved jointly for gaze estimation in a one-step way, with joint optimization to seek optimal performance. During this, spatial-temporal context exchange happens among the clues on the head, face, and eye. Accordingly, the final gazes obtained by fusing features from various queries can be aware of global clues from heads and faces, and local clues from eyes simultaneously, which essentially leverages performance. Meanwhile, the one-step running way also ensures high running efficiency. Experiments on the challenging Gaze360 dataset verify the superiority of our proposition. The source code will be released at https://github.com/zgchen33/MCGaze."
                    },
                    {
                        "title": "Tera-sample-per-second arbitrary waveform generation in the synthetic dimension",
                        "abstract": "The synthetic dimension opens new horizons in quantum physics and topological photonics by enabling new dimensions for field and particle manipulations. The most appealing property of the photonic synthetic dimension is its ability to emulate high-dimensional optical behavior in a unitary physical system. Here we show that the photonic synthetic dimension can transform technical problems in photonic systems between dimensionalities, providing unexpected solutions to technical problems that are otherwise challenging. Specifically, we propose and experimentally demonstrate a photonic Galton board (PGB) in the temporal synthetic dimension, in which the temporal high-speed challenge is converted into a spatial fiber-optic length matching problem, leading to the experimental generation of tera-sample-per-second arbitrary waveforms. Limited by the speed of the measurement equipment, waveforms with sampling rates of up to 341.53 GSa/s are recorded. Our proposed PGB operating in the temporal synthetic dimension breaks the speed limit in a physical system, bringing arbitrary waveform generation into the terahertz regime. The concept of dimension conversion offers possible solutions to various physical dimension-related problems, such as super-resolution imaging, high-resolution spectroscopy, time measurement, etc."
                    }
                ]
            },
            "5c004d83-fc87-4138-9c6f-d499e4b9e382": {
                "pk": "5c004d83-fc87-4138-9c6f-d499e4b9e382",
                "name": "Dingkang Liang",
                "collaborators": [
                    "Xiang Bai",
                    "Wei Xu",
                    "Zhikang Zou",
                    "Xiaoqing Ye",
                    "Xinwei He",
                    "Xin Zhou",
                    "Yingying Zhu",
                    "Yu Zhou",
                    "Jingyu Li",
                    "Zhe Liu"
                ],
                "domain": [
                    "Crowd Localization",
                    "Object Detection",
                    "Semi-Supervised Learning",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Focal Inverse Distance Transform Maps for Crowd Localization",
                        "abstract": "In this paper, we focus on the crowd localization task, a crucial topic of crowd analysis. Most regression-based methods utilize convolution neural networks (CNN) to regress a density map, which can not accurately locate the instance in the extremely dense scene, attributed to two crucial reasons: 1) the density map consists of a series of blurry Gaussian blobs, 2) severe overlaps exist in the dense region of the density map. To tackle this issue, we propose a novel Focal Inverse Distance Transform (FIDT) map for the crowd localization task. Compared with the density maps, the FIDT maps accurately describe the persons' locations without overlapping in dense regions. Based on the FIDT maps, a Local-Maxima-Detection-Strategy (LMDS) is derived to effectively extract the center point for each individual. Furthermore, we introduce an Independent SSIM (I-SSIM) loss to make the model tend to learn the local structural information, better recognizing local maxima. Extensive experiments demonstrate that the proposed method reports state-of-the-art localization performance on six crowd datasets and one vehicle dataset. Additionally, we find that the proposed method shows superior robustness on the negative and extremely dense scenes, which further verifies the effectiveness of the FIDT maps. The code and model will be available at https://github.com/dk-liang/FIDTM."
                    },
                    {
                        "title": "An End-to-End Transformer Model for Crowd Localization",
                        "abstract": "Crowd localization, predicting head positions, is a more practical and high-level task than simply counting. Existing methods employ pseudo-bounding boxes or pre-designed localization maps, relying on complex post-processing to obtain the head positions. In this paper, we propose an elegant, end-to-end Crowd Localization Transformer named CLTR that solves the task in the regression-based paradigm. The proposed method views the crowd localization as a direct set prediction problem, taking extracted features and trainable embeddings as input of the transformer-decoder. To reduce the ambiguous points and generate more reasonable matching results, we introduce a KMO-based Hungarian matcher, which adopts the nearby context as the auxiliary matching cost. Extensive experiments conducted on five datasets in various data settings show the effectiveness of our method. In particular, the proposed method achieves the best localization performance on the NWPU-Crowd, UCF-QNRF, and ShanghaiTech Part A datasets."
                    },
                    {
                        "title": "Dilated-Scale-Aware Attention ConvNet For Multi-Class Object Counting",
                        "abstract": "Object counting aims to estimate the number of objects in images. The leading counting approaches focus on the single category counting task and achieve impressive performance. Note that there are multiple categories of objects in real scenes. Multi-class object counting expands the scope of application of object counting task. The multi-target detection task can achieve multi-class object counting in some scenarios. However, it requires the dataset annotated with bounding boxes. Compared with the point annotations in mainstream object counting issues, the coordinate box-level annotations are more difficult to obtain. In this paper, we propose a simple yet efficient counting network based on point-level annotations. Specifically, we first change the traditional output channel from one to the number of categories to achieve multiclass counting. Since all categories of objects use the same feature extractor in our proposed framework, their features will interfere mutually in the shared feature space. We further design a multi-mask structure to suppress harmful interaction among objects. Extensive experiments on the challenging benchmarks illustrate that the proposed method achieves state-of-the-art counting performance."
                    },
                    {
                        "title": "Comprehensive Benchmark Datasets for Amharic Scene Text Detection and Recognition",
                        "abstract": "Ethiopic/Amharic script is one of the oldest African writing systems, which serves at least 23 languages (e.g., Amharic, Tigrinya) in East Africa for more than 120 million people. The Amharic writing system, Abugida, has 282 syllables, 15 punctuation marks, and 20 numerals. The Amharic syllabic matrix is derived from 34 base graphemes/consonants by adding up to 12 appropriate diacritics or vocalic markers to the characters. The syllables with a common consonant or vocalic markers are likely to be visually similar and challenge text recognition tasks. In this work, we presented the first comprehensive public datasets named HUST-ART, HUST-AST, ABE, and Tana for Amharic script detection and recognition in the natural scene. We have also conducted extensive experiments to evaluate the performance of the state of art methods in detecting and recognizing Amharic scene text on our datasets. The evaluation results demonstrate the robustness of our datasets for benchmarking and its potential of promoting the development of robust Amharic script detection and recognition algorithms. Consequently, the outcome will benefit people in East Africa, including diplomats from several countries and international communities."
                    },
                    {
                        "title": "DDS3D: Dense Pseudo-Labels with Dynamic Threshold for Semi-Supervised 3D Object Detection",
                        "abstract": "In this paper, we present a simple yet effective semi-supervised 3D object detector named DDS3D. Our main contributions have two-fold. On the one hand, different from previous works using Non-Maximal Suppression (NMS) or its variants for obtaining the sparse pseudo labels, we propose a dense pseudo-label generation strategy to get dense pseudo-labels, which can retain more potential supervision information for the student network. On the other hand, instead of traditional fixed thresholds, we propose a dynamic threshold manner to generate pseudo-labels, which can guarantee the quality and quantity of pseudo-labels during the whole training process. Benefiting from these two components, our DDS3D outperforms the state-of-the-art semi-supervised 3d object detection with mAP of 3.1% on the pedestrian and 2.1% on the cyclist under the same configuration of 1% samples. Extensive ablation studies on the KITTI dataset demonstrate the effectiveness of our DDS3D. The code and models will be made publicly available at https://github.com/hust-jy/DDS3D"
                    },
                    {
                        "title": "TransCrowd: weakly-supervised crowd counting with transformers",
                        "abstract": "The mainstream crowd counting methods usually utilize the convolution neural network (CNN) to regress a density map, requiring point-level annotations. However, annotating each person with a point is an expensive and laborious process. During the testing phase, the point-level annotations are not considered to evaluate the counting accuracy, which means the point-level annotations are redundant. Hence, it is desirable to develop weakly-supervised counting methods that just rely on count-level annotations, a more economical way of labeling. Current weakly-supervised counting methods adopt the CNN to regress a total count of the crowd by an image-to-count paradigm. However, having limited receptive fields for context modeling is an intrinsic limitation of these weakly-supervised CNN-based methods. These methods thus cannot achieve satisfactory performance, with limited applications in the real world. The transformer is a popular sequence-to-sequence prediction model in natural language processing (NLP), which contains a global receptive field. In this paper, we propose TransCrowd, which reformulates the weakly-supervised crowd counting problem from the perspective of sequence-to-count based on transformers. We observe that the proposed TransCrowd can effectively extract the semantic crowd information by using the self-attention mechanism of transformer. To the best of our knowledge, this is the first work to adopt a pure transformer for crowd counting research. Experiments on five benchmark datasets demonstrate that the proposed TransCrowd achieves superior performance compared with all the weakly-supervised CNN-based counting methods and gains highly competitive counting performance compared with some popular fully-supervised counting methods."
                    },
                    {
                        "title": "Anomaly Detection by Adapting a pre-trained Vision Language Model",
                        "abstract": "Recently, large vision and language models have shown their success when adapting them to many downstream tasks. In this paper, we present a unified framework named CLIP-ADA for Anomaly Detection by Adapting a pre-trained CLIP model. To this end, we make two important improvements: 1) To acquire unified anomaly detection across industrial images of multiple categories, we introduce the learnable prompt and propose to associate it with abnormal patterns through self-supervised learning. 2) To fully exploit the representation power of CLIP, we introduce an anomaly region refinement strategy to refine the localization quality. During testing, the anomalies are localized by directly calculating the similarity between the representation of the learnable prompt and the image. Comprehensive experiments demonstrate the superiority of our framework, e.g., we achieve the state-of-the-art 97.5/55.6 and 89.3/33.1 on MVTec-AD and VisA for anomaly detection and localization. In addition, the proposed method also achieves encouraging performance with marginal training data, which is more challenging."
                    },
                    {
                        "title": "CrowdCLIP: Unsupervised Crowd Counting via Vision-Language Model",
                        "abstract": "Supervised crowd counting relies heavily on costly manual labeling, which is difficult and expensive, especially in dense scenes. To alleviate the problem, we propose a novel unsupervised framework for crowd counting, named CrowdCLIP. The core idea is built on two observations: 1) the recent contrastive pre-trained vision-language model (CLIP) has presented impressive performance on various downstream tasks; 2) there is a natural mapping between crowd patches and count text. To the best of our knowledge, CrowdCLIP is the first to investigate the vision language knowledge to solve the counting problem. Specifically, in the training stage, we exploit the multi-modal ranking loss by constructing ranking text prompts to match the size-sorted crowd patches to guide the image encoder learning. In the testing stage, to deal with the diversity of image patches, we propose a simple yet effective progressive filtering strategy to first select the highly potential crowd patches and then map them into the language space with various counting intervals. Extensive experiments on five challenging datasets demonstrate that the proposed CrowdCLIP achieves superior performance compared to previous unsupervised state-of-the-art counting methods. Notably, CrowdCLIP even surpasses some popular fully-supervised methods under the cross-dataset setting. The source code will be available at https://github.com/dk-liang/CrowdCLIP."
                    },
                    {
                        "title": "SOOD++: Leveraging Unlabeled Data to Boost Oriented Object Detection",
                        "abstract": "Semi-supervised object detection (SSOD), leveraging unlabeled data to boost object detectors, has become a hot topic recently. However, existing SSOD approaches mainly focus on horizontal objects, leaving multi-oriented objects common in aerial images unexplored. At the same time, the annotation cost of multi-oriented objects is significantly higher than that of their horizontal counterparts. Therefore, in this paper, we propose a simple yet effective Semi-supervised Oriented Object Detection method termed SOOD++. Specifically, we observe that objects from aerial images are usually arbitrary orientations, small scales, and aggregation, which inspires the following core designs: a Simple Instance-aware Dense Sampling (SIDS) strategy is used to generate comprehensive dense pseudo-labels; the Geometry-aware Adaptive Weighting (GAW) loss dynamically modulates the importance of each pair between pseudo-label and corresponding prediction by leveraging the intricate geometric information of aerial objects; we treat aerial images as global layouts and explicitly build the many-to-many relationship between the sets of pseudo-labels and predictions via the proposed Noise-driven Global Consistency (NGC). Extensive experiments conducted on various multi-oriented object datasets under various labeled settings demonstrate the effectiveness of our method. For example, on the DOTA-V1.5 benchmark, the proposed method outperforms previous state-of-the-art (SOTA) by a large margin (+2.92, +2.39, and +2.57 mAP under 10%, 20%, and 30% labeled data settings, respectively) with single-scale training and testing. More importantly, it still improves upon a strong supervised baseline with 70.66 mAP, trained using the full DOTA-V1.5 train-val set, by +1.82 mAP, resulting in a 72.48 mAP, pushing the new state-of-the-art. The code will be made available."
                    },
                    {
                        "title": "AutoScale: Learning to Scale for Crowd Counting and Localization",
                        "abstract": "Recent works on crowd counting mainly leverage CNNs to count by regressing density maps, and have achieved great progress. In the density map, each person is represented by a Gaussian blob, and the final count is obtained from the integration of the whole map. However, it is difficult to accurately predict the density map on dense regions. A major issue is that the density map on dense regions usually accumulates density values from a number of nearby Gaussian blobs, yielding different large density values on a small set of pixels. This makes the density map present variant patterns with significant pattern shifts and brings a long-tailed distribution of pixel-wise density values. We propose a simple and effective Learning to Scale (L2S) module, which automatically scales dense regions into reasonable closeness levels (reflecting image-plane distance between neighboring people). L2S directly normalizes the closeness in different patches such that it dynamically separates the overlapped blobs, decomposes the accumulated values in the ground-truth density map, and thus alleviates the pattern shifts and long-tailed distribution of density values. This helps the model to better learn the density map. We also explore the effectiveness of L2S in localizing people by finding the local minima of the quantized distance (w.r.t. person location map). To the best of our knowledge, such a localization method is also novel in localization-based crowd counting. We further introduce a customized dynamic cross-entropy loss, significantly improving the localization-based model optimization. Extensive experiments demonstrate that the proposed framework termed AutoScale improves upon some state-of-the-art methods in both regression and localization benchmarks on three crowded datasets and achieves very competitive performance on two sparse datasets."
                    },
                    {
                        "title": "Parameter-Efficient Fine-Tuning in Spectral Domain for Point Cloud Learning",
                        "abstract": "Recently, leveraging pre-training techniques to enhance point cloud models has become a hot research topic. However, existing approaches typically require full fine-tuning of pre-trained models to achieve satisfied performance on downstream tasks, accompanying storage-intensive and computationally demanding. To address this issue, we propose a novel Parameter-Efficient Fine-Tuning (PEFT) method for point cloud, called PointGST (Point cloud Graph Spectral Tuning). PointGST freezes the pre-trained model and introduces a lightweight, trainable Point Cloud Spectral Adapter (PCSA) to fine-tune parameters in the spectral domain. The core idea is built on two observations: 1) The inner tokens from frozen models might present confusion in the spatial domain; 2) Task-specific intrinsic information is important for transferring the general knowledge to the downstream task. Specifically, PointGST transfers the point tokens from the spatial domain to the spectral domain, effectively de-correlating confusion among tokens via using orthogonal components for separating. Moreover, the generated spectral basis involves intrinsic information about the downstream point clouds, enabling more targeted tuning. As a result, PointGST facilitates the efficient transfer of general knowledge to downstream tasks while significantly reducing training costs. Extensive experiments on challenging point cloud datasets across various tasks demonstrate that PointGST not only outperforms its fully fine-tuning counterpart but also significantly reduces trainable parameters, making it a promising solution for efficient point cloud learning. It improves upon a solid baseline by +2.28%, 1.16%, and 2.78%, resulting in 99.48%, 97.76%, and 96.18% on the ScanObjNN OBJ BG, OBJ OBLY, and PB T50 RS datasets, respectively. This advancement establishes a new state-of-the-art, using only 0.67% of the trainable parameters."
                    },
                    {
                        "title": "SAM3D: Zero-Shot 3D Object Detection via Segment Anything Model",
                        "abstract": "With the development of large language models, many remarkable linguistic systems like ChatGPT have thrived and achieved astonishing success on many tasks, showing the incredible power of foundation models. In the spirit of unleashing the capability of foundation models on vision tasks, the Segment Anything Model (SAM), a vision foundation model for image segmentation, has been proposed recently and presents strong zero-shot ability on many downstream 2D tasks. However, whether SAM can be adapted to 3D vision tasks has yet to be explored, especially 3D object detection. With this inspiration, we explore adapting the zero-shot ability of SAM to 3D object detection in this paper. We propose a SAM-powered BEV processing pipeline to detect objects and get promising results on the large-scale Waymo open dataset. As an early attempt, our method takes a step toward 3D object detection with vision foundation models and presents the opportunity to unleash their power on 3D vision tasks. The code is released at https://github.com/DYZhang09/SAM3D."
                    },
                    {
                        "title": "A Unified Framework for 3D Scene Understanding",
                        "abstract": "We propose UniSeg3D, a unified 3D segmentation framework that achieves panoptic, semantic, instance, interactive, referring, and open-vocabulary semantic segmentation tasks within a single model. Most previous 3D segmentation approaches are specialized for a specific task, thereby limiting their understanding of 3D scenes to a task-specific perspective. In contrast, the proposed method unifies six tasks into unified representations processed by the same Transformer. It facilitates inter-task knowledge sharing and, therefore, promotes comprehensive 3D scene understanding. To take advantage of multi-task unification, we enhance the performance by leveraging task connections. Specifically, we design a knowledge distillation method and a contrastive learning method to transfer task-specific knowledge across different tasks. Benefiting from extensive inter-task knowledge sharing, our UniSeg3D becomes more powerful. Experiments on three benchmarks, including the ScanNet20, ScanRefer, and ScanNet200, demonstrate that the UniSeg3D consistently outperforms current SOTA methods, even those specialized for individual tasks. We hope UniSeg3D can serve as a solid unified baseline and inspire future work. The code will be available at https://dk-liang.github.io/UniSeg3D/."
                    },
                    {
                        "title": "LATFormer: Locality-Aware Point-View Fusion Transformer for 3D Shape Recognition",
                        "abstract": "Recently, 3D shape understanding has achieved significant progress due to the advances of deep learning models on various data formats like images, voxels, and point clouds. Among them, point clouds and multi-view images are two complementary modalities of 3D objects and learning representations by fusing both of them has been proven to be fairly effective. While prior works typically focus on exploiting global features of the two modalities, herein we argue that more discriminative features can be derived by modeling ``where to fuse''. To investigate this, we propose a novel Locality-Aware Point-View Fusion Transformer (LATFormer) for 3D shape retrieval and classification. The core component of LATFormer is a module named Locality-Aware Fusion (LAF) which integrates the local features of correlated regions across the two modalities based on the co-occurrence scores. We further propose to filter out scores with low values to obtain salient local co-occurring regions, which reduces redundancy for the fusion process. In our LATFormer, we utilize the LAF module to fuse the multi-scale features of the two modalities both bidirectionally and hierarchically to obtain more informative features. Comprehensive experiments on four popular 3D shape benchmarks covering 3D object retrieval and classification validate its effectiveness."
                    },
                    {
                        "title": "SOOD: Towards Semi-Supervised Oriented Object Detection",
                        "abstract": "Semi-Supervised Object Detection (SSOD), aiming to explore unlabeled data for boosting object detectors, has become an active task in recent years. However, existing SSOD approaches mainly focus on horizontal objects, leaving multi-oriented objects that are common in aerial images unexplored. This paper proposes a novel Semi-supervised Oriented Object Detection model, termed SOOD, built upon the mainstream pseudo-labeling framework. Towards oriented objects in aerial scenes, we design two loss functions to provide better supervision. Focusing on the orientations of objects, the first loss regularizes the consistency between each pseudo-label-prediction pair (includes a prediction and its corresponding pseudo label) with adaptive weights based on their orientation gap. Focusing on the layout of an image, the second loss regularizes the similarity and explicitly builds the many-to-many relation between the sets of pseudo-labels and predictions. Such a global consistency constraint can further boost semi-supervised learning. Our experiments show that when trained with the two proposed losses, SOOD surpasses the state-of-the-art SSOD methods under various settings on the DOTA-v1.5 benchmark. The code will be available at https://github.com/HamPerdredes/SOOD."
                    },
                    {
                        "title": "A Discrepancy Aware Framework for Robust Anomaly Detection",
                        "abstract": "Defect detection is a critical research area in artificial intelligence. Recently, synthetic data-based self-supervised learning has shown great potential on this task. Although many sophisticated synthesizing strategies exist, little research has been done to investigate the robustness of models when faced with different strategies. In this paper, we focus on this issue and find that existing methods are highly sensitive to them. To alleviate this issue, we present a Discrepancy Aware Framework (DAF), which demonstrates robust performance consistently with simple and cheap strategies across different anomaly detection benchmarks. We hypothesize that the high sensitivity to synthetic data of existing self-supervised methods arises from their heavy reliance on the visual appearance of synthetic data during decoding. In contrast, our method leverages an appearance-agnostic cue to guide the decoder in identifying defects, thereby alleviating its reliance on synthetic appearance. To this end, inspired by existing knowledge distillation methods, we employ a teacher-student network, which is trained based on synthesized outliers, to compute the discrepancy map as the cue. Extensive experiments on two challenging datasets prove the robustness of our method. Under the simple synthesis strategies, it outperforms existing methods by a large margin. Furthermore, it also achieves the state-of-the-art localization performance. Code is available at: https://github.com/caiyuxuan1120/DAF."
                    },
                    {
                        "title": "Dynamic Adapter Meets Prompt Tuning: Parameter-Efficient Transfer Learning for Point Cloud Analysis",
                        "abstract": "Point cloud analysis has achieved outstanding performance by transferring point cloud pre-trained models. However, existing methods for model adaptation usually update all model parameters, i.e., full fine-tuning paradigm, which is inefficient as it relies on high computational costs (e.g., training GPU memory) and massive storage space. In this paper, we aim to study parameter-efficient transfer learning for point cloud analysis with an ideal trade-off between task performance and parameter efficiency. To achieve this goal, we freeze the parameters of the default pre-trained models and then propose the Dynamic Adapter, which generates a dynamic scale for each token, considering the token significance to the downstream task. We further seamlessly integrate Dynamic Adapter with Prompt Tuning (DAPT) by constructing Internal Prompts, capturing the instance-specific features for interaction. Extensive experiments conducted on five challenging datasets demonstrate that the proposed DAPT achieves superior performance compared to the full fine-tuning counterparts while significantly reducing the trainable parameters and training GPU memory by 95% and 35%, respectively. Code is available at https://github.com/LMD0311/DAPT."
                    },
                    {
                        "title": "Mini-Monkey: Multi-Scale Adaptive Cropping for Multimodal Large Language Models",
                        "abstract": "Recently, there has been significant interest in enhancing the capability of multimodal large language models (MLLMs) to process high-resolution images. Most existing methods focus on adopting a cropping strategy to improve the ability of multimodal large language models to understand image details. However, this cropping operation inevitably causes the segmentation of objects and connected areas, which impairs the MLLM's ability to recognize small or irregularly shaped objects or text. This issue is particularly evident in lightweight MLLMs. Addressing this issue, we propose Mini-Monkey, a lightweight MLLM that incorporates a plug-and-play method called multi-scale adaptive crop strategy (MSAC). Mini-Monkey adaptively generates multi-scale representations, allowing it to select non-segmented objects from various scales. To mitigate the computational overhead introduced by MSAC, we propose a Scale Compression Mechanism (SCM), which effectively compresses image tokens. Mini-Monkey achieves state-of-the-art performance among 2B-parameter MLLMs. It not only demonstrates leading performance on a variety of general multimodal understanding tasks but also shows consistent improvements in document understanding capabilities. On the OCRBench, Mini-Monkey achieves a score of 802, outperforming 8B-parameter state-of-the-art model InternVL2-8B. Besides, our model and training strategy are very efficient, which can be trained with only eight RTX 3090. The code is available at https://github.com/Yuliang-Liu/Monkey."
                    }
                ]
            },
            "bb7ea79e-ead8-4b57-b5fb-31270d29d85b": {
                "pk": "bb7ea79e-ead8-4b57-b5fb-31270d29d85b",
                "name": "Yuchao Chen",
                "collaborators": [
                    "Jintao Wang",
                    "Jian Song",
                    "Haoyue Tang",
                    "Pengkun Yang",
                    "Leandros Tassiulas",
                    "Yuqiao He",
                    "Xiaoqing Wang",
                    "Jingzhou Sun"
                ],
                "domain": [
                    "Wireless Networks",
                    "Age of Information",
                    "Stochastic Processes",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Age of Synchronization Minimization in Wireless Networks with Random Updates and Time-Varying Timeliness Requirement",
                        "abstract": "This study considers a wireless network where multiple nodes transmit status updates to a base station (BS) via a shared, error-free channel with limited bandwidth. The status updates arrive at each node randomly. We use the Age of Synchronization (AoS) as a metric to measure the information freshness of the updates. The AoS of each node has a timely-varying importance which follows a Markov chain. Our objective is to minimize the weighted sum AoS of the system. The optimization problem is relaxed and formulated as a constrained Markov decision process (CMDP). Solving the relaxed CMDP by a linear programming algorithm yields a stationary policy, which helps us propose a near-stationary policy for the original problem. Numerical simulations show that in most configurations, the AoS performance of our policy outperforms the policy choosing the maximum AoS regardless of weight variations."
                    },
                    {
                        "title": "Age of Information Optimization in Multi-Channel Network with Sided Information",
                        "abstract": "We consider a discrete-time multi-channel network where the destination collects time-sensitive packets from multiple sources with sided channel information. The popular metric, Age of Information (AoI), is applied to measure the data freshness at the destination. Due to the interference constraint, only disjoint source-channel pairs can be chosen for transmission in each time slot, and the decision maker should choose the optimal scheduling pairs to minimize the average AoI at the destination. To learn the optimal channel selection, we apply the linear contextual bandit (LCB) framework by utilizing the sided information provided by pilots. Concretely, we establish the relationship between AoI regret and sub-optimal channel selection times and propose both age-independent and age-dependent algorithms. The former method is proven to achieve the sub-linear AoI regret but is outperformed by the latter algorithm both in the linear and non-linear contextual model in simulation."
                    },
                    {
                        "title": "Sending Timely Status Updates through Channel with Random Delay via Online Learning",
                        "abstract": "In this work, we study a status update system with a source node sending timely information to the destination through a channel with random delay. We measure the timeliness of the information stored at the receiver via the Age of Information (AoI), the time elapsed since the freshest sample stored at the receiver is generated. The goal is to design a sampling strategy that minimizes the total cost of the expected time average AoI and sampling cost in the absence of transmission delay statistics. We reformulate the total cost minimization problem as the optimization of a renewal-reward process, and propose an online sampling strategy based on the Robbins-Monro algorithm. Denote $K$ to be the number of samples we have taken. We show that, when the transmission delay is bounded, the expected time average total cost obtained by the proposed online algorithm converges to the minimum cost when $K$ goes to infinity, and the optimality gap decays with rate $\\mathcal{O}\\left(\\ln K/K\\right)$. Simulation results validate the performance of our proposed algorithm."
                    },
                    {
                        "title": "Sampling for Remote Estimation of an Ornstein-Uhlenbeck Process through Channel with Unknown Delay Statistics",
                        "abstract": "In this paper, we consider sampling an Ornstein-Uhlenbeck (OU) process through a channel for remote estimation. The goal is to minimize the mean square error (MSE) at the estimator under a sampling frequency constraint when the channel delay statistics is unknown. Sampling for MSE minimization is reformulated into an optimal stopping problem. By revisiting the threshold structure of the optimal stopping policy when the delay statistics is known, we propose an online sampling algorithm to learn the optimum threshold using stochastic approximation algorithm and the virtual queue method. We prove that with probability 1, the MSE of the proposed online algorithm converges to the minimum MSE that is achieved when the channel delay statistics is known. The cumulative MSE gap of our proposed algorithm compared with the minimum MSE up to the $(k+1)$-th sample grows with rate at most $\\mathcal{O}(\\ln k)$. Our proposed online algorithm can satisfy the sampling frequency constraint theoretically. Finally, simulation results are provided to demonstrate the performance of the proposed algorithm."
                    },
                    {
                        "title": "Age Optimal Sampling Under Unknown Delay Statistics",
                        "abstract": "This paper revisits the problem of sampling and transmitting status updates through a channel with random delay under a sampling frequency constraint \\cite{sun_17_tit}. We use the Age of Information (AoI) to characterize the status information freshness at the receiver. The goal is to design a sampling policy that can minimize the average AoI when the statistics of delay is unknown. We reformulate the problem as the optimization of a renewal-reward process, and propose an online sampling strategy based on the Robbins-Monro algorithm. We prove that the proposed algorithm satisfies the sampling frequency constraint. Moreover, when the transmission delay is bounded and its distribution is absolutely continuous, the average AoI obtained by the proposed algorithm converges to the minimum AoI when the number of samples $K$ goes to infinity with probability 1. We show that the optimality gap decays with rate $\\mathcal{O}\\left(\\ln K/K\\right)$, and the proposed algorithm is minimax rate optimal. Simulation results validate the performance of our proposed algorithm."
                    }
                ]
            },
            "5fd2a567-83e5-40ac-bc96-80c341051f97": {
                "pk": "5fd2a567-83e5-40ac-bc96-80c341051f97",
                "name": "Yuliang Liu",
                "collaborators": [
                    "Lianwen Jin",
                    "Wenze Liu",
                    "Hao Lu",
                    "Zhiguo Cao",
                    "Dawei Ding",
                    "Jie Zhu",
                    "Xiaoshu Luo",
                    "Zhiwei Zhang",
                    "Jiangqi Mao",
                    "Houmin Du"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Scene Text Detection",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Deep Matching Prior Network: Toward Tighter Multi-oriented Text Detection",
                        "abstract": "Detecting incidental scene text is a challenging task because of multi-orientation, perspective distortion, and variation of text size, color and scale. Retrospective research has only focused on using rectangular bounding box or horizontal sliding window to localize text, which may result in redundant background noise, unnecessary overlap or even information loss. To address these issues, we propose a new Convolutional Neural Networks (CNNs) based method, named Deep Matching Prior Network (DMPNet), to detect text with tighter quadrangle. First, we use quadrilateral sliding windows in several specific intermediate convolutional layers to roughly recall the text with higher overlapping area and then a shared Monte-Carlo method is proposed for fast and accurate computing of the polygonal areas. After that, we designed a sequential protocol for relative regression which can exactly predict text with compact quadrangle. Moreover, a auxiliary smooth Ln loss is also proposed for further regressing the position of text, which has better overall performance than L2 loss and smooth L1 loss in terms of robustness and stability. The effectiveness of our approach is evaluated on a public word-level, multi-oriented scene text database, ICDAR 2015 Robust Reading Competition Challenge 4 \"Incidental scene text localization\". The performance of our method is evaluated by using F-measure and found to be 70.64%, outperforming the existing state-of-the-art method with F-measure 63.76%."
                    },
                    {
                        "title": "Accountable Textual-Visual Chat Learns to Reject Human Instructions in Image Re-creation",
                        "abstract": "The recent success of ChatGPT and GPT-4 has drawn widespread attention to multimodal dialogue systems. However, there is a lack of datasets in the academic community that can effectively evaluate the multimodal generation capabilities of Visual Language Models (VLMs) in textual-visual chat tasks. In this paper, we address this gap by introducing two novel multimodal datasets: the synthetic CLEVR-ATVC dataset (620K) and the manually pictured Fruit-ATVC dataset (50K). These datasets incorporate both visual and text-based inputs and outputs. Furthermore, to facilitate the accountability of multimodal systems in rejecting human requests, similar to language-based ChatGPT conversations, we introduce specific rules as supervisory signals within the datasets. This allows the trained VLM to provide a yes or no answer after engaging in visual and textual reasoning, accompanied by a language explanation to clarify the reasons behind the inability to execute the given human instruction. Our proposed method involves a two-stage training procedure, which includes training the image auto-encoder and the auto-regressive transformer from scratch. The first stage employs a discrete variational autoencoder (dVAE) to compress each image into concise tokens, which are then combined with text tokens into a single data stream. This stream is subsequently fed into the decoder-based transformer to generate visual re-creations and textual feedback in the second stage. We conduct comprehensive analyses of experimental results, focusing on re-created image quality, answer accuracy, and the model's behavior when faced with uncertainty and imperfect user queries. Through our explorations and findings, we aim to contribute valuable insights into the accountability of textual-visual generative models."
                    },
                    {
                        "title": "On Point Affiliation in Feature Upsampling",
                        "abstract": "We introduce the notion of point affiliation into feature upsampling. By abstracting a feature map into non-overlapped semantic clusters formed by points of identical semantic meaning, feature upsampling can be viewed as point affiliation -- designating a semantic cluster for each upsampled point. In the framework of kernel-based dynamic upsampling, we show that an upsampled point can resort to its low-res decoder neighbors and high-res encoder point to reason the affiliation, conditioned on the mutual similarity between them. We therefore present a generic formulation for generating similarity-aware upsampling kernels and prove that such kernels encourage not only semantic smoothness but also boundary sharpness. This formulation constitutes a novel, lightweight, and universal upsampling solution, Similarity-Aware Point Affiliation (SAPA). We show its working mechanism via our preliminary designs with window-shape kernel. After probing the limitations of the designs on object detection, we reveal additional insights for upsampling, leading to SAPA with the dynamic kernel shape. Extensive experiments demonstrate that SAPA outperforms prior upsamplers and invites consistent performance improvements on a number of dense prediction tasks, including semantic segmentation, object detection, instance segmentation, panoptic segmentation, image matting, and depth estimation. Code is made available at: https://github.com/tiny-smart/sapa"
                    },
                    {
                        "title": "Box-DETR: Understanding and Boxing Conditional Spatial Queries",
                        "abstract": "Conditional spatial queries are recently introduced into DEtection TRansformer (DETR) to accelerate convergence. In DAB-DETR, such queries are modulated by the so-called conditional linear projection at each decoder stage, aiming to search for positions of interest such as the four extremities of the box. Each decoder stage progressively updates the box by predicting the anchor box offsets, while in cross-attention only the box center is informed as the reference point. The use of only box center, however, leaves the width and height of the previous box unknown to the current stage, which hinders accurate prediction of offsets. We argue that the explicit use of the entire box information in cross-attention matters. In this work, we propose Box Agent to condense the box into head-specific agent points. By replacing the box center with the agent point as the reference point in each head, the conditional cross-attention can search for positions from a more reasonable starting point by considering the full scope of the previous box, rather than always from the previous box center. This significantly reduces the burden of the conditional linear projection. Experimental results show that the box agent leads to not only faster convergence but also improved detection performance, e.g., our single-scale model achieves $44.2$ AP with ResNet-50 based on DAB-DETR. Our Box Agent requires minor modifications to the code and has negligible computational workload. Code is available at https://github.com/tiny-smart/box-detr."
                    },
                    {
                        "title": "Transport Properties of a Quantum Dot Restudied by Algebraic Equation of Motion",
                        "abstract": "Based on the algebraic equation of motion (AEOM) method, we investigate the transport properties of a quantum dot. We obtain an analytical expression for the dot electron single-particle Green's function, and based on this expression, we plot the dot electron density of states under different biases. We find that the Kondo resonance splits and is suppressed as the bias is increased. In addition, we calculate the differential conductance of the dot and obtain the zero-bias Kondo resonance at different temperatures, which is found to be suppressed as the temperature is increased."
                    },
                    {
                        "title": "ViTEraser: Harnessing the Power of Vision Transformers for Scene Text Removal with SegMIM Pretraining",
                        "abstract": "Scene text removal (STR) aims at replacing text strokes in natural scenes with visually coherent backgrounds. Recent STR approaches rely on iterative refinements or explicit text masks, resulting in high complexity and sensitivity to the accuracy of text localization. Moreover, most existing STR methods adopt convolutional architectures while the potential of vision Transformers (ViTs) remains largely unexplored. In this paper, we propose a simple-yet-effective ViT-based text eraser, dubbed ViTEraser. Following a concise encoder-decoder framework, ViTEraser can easily incorporate various ViTs to enhance long-range modeling. Specifically, the encoder hierarchically maps the input image into the hidden space through ViT blocks and patch embedding layers, while the decoder gradually upsamples the hidden features to the text-erased image with ViT blocks and patch splitting layers. As ViTEraser implicitly integrates text localization and inpainting, we propose a novel end-to-end pretraining method, termed SegMIM, which focuses the encoder and decoder on the text box segmentation and masked image modeling tasks, respectively. Experimental results demonstrate that ViTEraser with SegMIM achieves state-of-the-art performance on STR by a substantial margin and exhibits strong generalization ability when extended to other tasks, \\textit{e.g.}, tampered scene text detection. Furthermore, we comprehensively explore the architecture, pretraining, and scalability of the ViT-based encoder-decoder for STR, which provides deep insights into the application of ViT to the STR field. Code is available at https://github.com/shannanyinxiang/ViTEraser."
                    },
                    {
                        "title": "Controlling Delay-induced Hopf bifurcation in Internet congestion control system",
                        "abstract": "This paper focuses on Hopf bifurcation control in a dual model of Internet congestion control algorithms which is modeled as a delay differential equation (DDE). By choosing communication delay as a bifurcation parameter, it has been demonstrated that the system loses stability and a Hopf bifurcation occurs when communication delay passes through a critical value. Therefore, a time-delayed feedback control method is applied to the system for delaying the onset of undesirable Hopf bifurcation. Theoretical analysis and numerical simulations confirm that the delayed feedback controller is efficient in controlling Hopf bifurcation in Internet congestion control system. Moreover, the direction of the Hopf bifurcation and the stability of the bifurcating periodic solutions are determinated by applying the center manifold theorem and the normal form theory."
                    },
                    {
                        "title": "Hopf bifurcation analysis in a dual model of Internet congestion control algorithm with communication delay",
                        "abstract": "This paper focuses on the delay induced Hopf bifurcation in a dual model of Internet congestion control algorithms which can be modeled as a time-delay system described by a one-order delay differential equation (DDE). By choosing communication delay as the bifurcation parameter, we demonstrate that the system loses its stability and a Hopf bifurcation occurs when communication delay passes through a critical value. Moreover, the bifurcating periodic solution of system is calculated by means of perturbation methods. Discussion of stability of the periodic solutions involves the computation of Floquet exponents by considering the corresponding Poincare -Lindstedt series expansion. Finally, numerical simulations for verify the theoretical analysis are provided."
                    },
                    {
                        "title": "Feature Enhancement Network: A Refined Scene Text Detector",
                        "abstract": "In this paper, we propose a refined scene text detector with a \\textit{novel} Feature Enhancement Network (FEN) for Region Proposal and Text Detection Refinement. Retrospectively, both region proposal with \\textit{only} $3\\times 3$ sliding-window feature and text detection refinement with \\textit{single scale} high level feature are insufficient, especially for smaller scene text. Therefore, we design a new FEN network with \\textit{task-specific}, \\textit{low} and \\textit{high} level semantic features fusion to improve the performance of text detection. Besides, since \\textit{unitary} position-sensitive RoI pooling in general object detection is unreasonable for variable text regions, an \\textit{adaptively weighted} position-sensitive RoI pooling layer is devised for further enhancing the detecting accuracy. To tackle the \\textit{sample-imbalance} problem during the refinement stage, we also propose an effective \\textit{positives mining} strategy for efficiently training our network. Experiments on ICDAR 2011 and 2013 robust text detection benchmarks demonstrate that our method can achieve state-of-the-art results, outperforming all reported methods in terms of F-measure."
                    },
                    {
                        "title": "DeRPN: Taking a further step toward more general object detection",
                        "abstract": "Most current detection methods have adopted anchor boxes as regression references. However, the detection performance is sensitive to the setting of the anchor boxes. A proper setting of anchor boxes may vary significantly across different datasets, which severely limits the universality of the detectors. To improve the adaptivity of the detectors, in this paper, we present a novel dimension-decomposition region proposal network (DeRPN) that can perfectly displace the traditional Region Proposal Network (RPN). DeRPN utilizes an anchor string mechanism to independently match object widths and heights, which is conducive to treating variant object shapes. In addition, a novel scale-sensitive loss is designed to address the imbalanced loss computations of different scaled objects, which can avoid the small objects being overwhelmed by larger ones. Comprehensive experiments conducted on both general object detection datasets (Pascal VOC 2007, 2012 and MS COCO) and scene text detection datasets (ICDAR 2013 and COCO-Text) all prove that our DeRPN can significantly outperform RPN. It is worth mentioning that the proposed DeRPN can be employed directly on different models, tasks, and datasets without any modifications of hyperparameters or specialized optimization, which further demonstrates its adaptivity. The code will be released at https://github.com/HCIILAB/DeRPN."
                    }
                ]
            },
            "80327eca-a327-4578-afa3-9023f3abf02e": {
                "pk": "80327eca-a327-4578-afa3-9023f3abf02e",
                "name": "Xiang Bai",
                "collaborators": [
                    "Song Bai",
                    "Changhui Tan",
                    "Liutang Xue",
                    "Qi Tian",
                    "Rui Yu",
                    "Zhichao Zhou",
                    "Xinwei He",
                    "Tengteng Huang",
                    "Qianyun Miao",
                    "Zhuotun Zhu"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Manifold Learning",
                    "Action Recognition"
                ],
                "publications": [
                    {
                        "title": "Scalable Person Re-identification on Supervised Smoothed Manifold",
                        "abstract": "Most existing person re-identification algorithms either extract robust visual features or learn discriminative metrics for person images. However, the underlying manifold which those images reside on is rarely investigated. That raises a problem that the learned metric is not smooth with respect to the local geometry structure of the data manifold.   In this paper, we study person re-identification with manifold-based affinity learning, which did not receive enough attention from this area. An unconventional manifold-preserving algorithm is proposed, which can 1) make the best use of supervision from training data, whose label information is given as pairwise constraints; 2) scale up to large repositories with low on-line time complexity; and 3) be plunged into most existing algorithms, serving as a generic postprocessing procedure to further boost the identification accuracies. Extensive experimental results on five popular person re-identification benchmarks consistently demonstrate the effectiveness of our method. Especially, on the largest CUHK03 and Market-1501, our method outperforms the state-of-the-art alternatives by a large margin with high efficiency, which is more appropriate for practical applications."
                    },
                    {
                        "title": "Divide and Fuse: A Re-ranking Approach for Person Re-identification",
                        "abstract": "As re-ranking is a necessary procedure to boost person re-identification (re-ID) performance on large-scale datasets, the diversity of feature becomes crucial to person reID for its importance both on designing pedestrian descriptions and re-ranking based on feature fusion. However, in many circumstances, only one type of pedestrian feature is available. In this paper, we propose a \"Divide and use\" re-ranking framework for person re-ID. It exploits the diversity from different parts of a high-dimensional feature vector for fusion-based re-ranking, while no other features are accessible. Specifically, given an image, the extracted feature is divided into sub-features. Then the contextual information of each sub-feature is iteratively encoded into a new feature. Finally, the new features from the same image are fused into one vector for re-ranking. Experimental results on two person re-ID benchmarks demonstrate the effectiveness of the proposed framework. Especially, our method outperforms the state-of-the-art on the Market-1501 dataset."
                    },
                    {
                        "title": "View N-gram Network for 3D Object Retrieval",
                        "abstract": "How to aggregate multi-view representations of a 3D object into an informative and discriminative one remains a key challenge for multi-view 3D object retrieval. Existing methods either use view-wise pooling strategies which neglect the spatial information across different views or employ recurrent neural networks which may face the efficiency problem. To address these issues, we propose an effective and efficient framework called View N-gram Network (VNN). Inspired by n-gram models in natural language processing, VNN divides the view sequence into a set of visual n-grams, which involve overlapping consecutive view sub-sequences. By doing so, spatial information across multiple views is captured, which helps to learn a discriminative global embedding for each 3D object. Experiments on 3D shape retrieval benchmarks, including ModelNet10, ModelNet40 and ShapeNetCore55 datasets, demonstrate the superiority of our proposed method."
                    },
                    {
                        "title": "Global well-posedness and asymptotic behavior in critical spaces for the compressible Euler system with velocity alignment",
                        "abstract": "In this paper, we study the Cauchy problem of the compressible Euler system with strongly singular velocity alignment. We prove the existence and uniqueness of global solutions in critical Besov spaces to the considered system with small initial data. The local-in-time solvability is also addressed. Moreover, we show the large-time asymptotic behavior and optimal decay estimates of the solutions as $t\\to \\infty$."
                    },
                    {
                        "title": "Global well-posedness and asymptotic behavior for the Euler-alignment system with pressure",
                        "abstract": "We study the Cauchy problem of the compressible Euler system with strongly singular velocity alignment. We establish a global well-posedness theory for the system with small smooth initial data. Additionally, we derive asymptotic emergent behaviors for the system, providing time decay estimates with optimal decay rates. Notably, the optimal decay rate we obtain does not align with the corresponding fractional heat equation within our considered range, where the parameter $\\alpha\\in(0,1)$. This highlights the distinct feature of the alignment operator."
                    },
                    {
                        "title": "Deep Learning Representation using Autoencoder for 3D Shape Retrieval",
                        "abstract": "We study the problem of how to build a deep learning representation for 3D shape. Deep learning has shown to be very effective in variety of visual applications, such as image classification and object detection. However, it has not been successfully applied to 3D shape recognition. This is because 3D shape has complex structure in 3D space and there are limited number of 3D shapes for feature learning. To address these problems, we project 3D shapes into 2D space and use autoencoder for feature learning on the 2D images. High accuracy 3D shape retrieval performance is obtained by aggregating the features learned on 2D images. In addition, we show the proposed deep learning feature is complementary to conventional local image descriptors. By combing the global deep learning representation and the local descriptor representation, our method can obtain the state-of-the-art performance on 3D shape retrieval benchmarks."
                    },
                    {
                        "title": "An Empirical Study of End-to-End Temporal Action Detection",
                        "abstract": "Temporal action detection (TAD) is an important yet challenging task in video understanding. It aims to simultaneously predict the semantic label and the temporal interval of every action instance in an untrimmed video. Rather than end-to-end learning, most existing methods adopt a head-only learning paradigm, where the video encoder is pre-trained for action classification, and only the detection head upon the encoder is optimized for TAD. The effect of end-to-end learning is not systematically evaluated. Besides, there lacks an in-depth study on the efficiency-accuracy trade-off in end-to-end TAD. In this paper, we present an empirical study of end-to-end temporal action detection. We validate the advantage of end-to-end learning over head-only learning and observe up to 11\\% performance improvement. Besides, we study the effects of multiple design choices that affect the TAD performance and speed, including detection head, video encoder, and resolution of input videos. Based on the findings, we build a mid-resolution baseline detector, which achieves the state-of-the-art performance of end-to-end methods while running more than 4$\\times$ faster. We hope that this paper can serve as a guide for end-to-end learning and inspire future research in this field. Code and models are available at \\url{https://github.com/xlliu7/E2E-TAD}."
                    },
                    {
                        "title": "Image Stitching by Line-guided Local Warping with Global Similarity Constraint",
                        "abstract": "Low-textured image stitching remains a challenging problem. It is difficult to achieve good alignment and it is easy to break image structures due to insufficient and unreliable point correspondences. Moreover, because of the viewpoint variations between multiple images, the stitched images suffer from projective distortions. To solve these problems, this paper presents a line-guided local warping method with a global similarity constraint for image stitching. Line features which serve well for geometric descriptions and scene constraints, are employed to guide image stitching accurately. On one hand, the line features are integrated into a local warping model through a designed weight function. On the other hand, line features are adopted to impose strong geometric constraints, including line correspondence and line colinearity, to improve the stitching performance through mesh optimization. To mitigate projective distortions, we adopt a global similarity constraint, which is integrated with the projective warps via a designed weight strategy. This constraint causes the final warp to slowly change from a projective to a similarity transformation across the image. Finally, the images undergo a two-stage alignment scheme that provides accurate alignment and reduces projective distortion. We evaluate our method on a series of images and compare it with several other methods. The experimental results demonstrate that the proposed method provides a convincing stitching performance and that it outperforms other state-of-the-art methods."
                    }
                ]
            }
        }
    },
    "2404.14408": {
        "paper_data": {
            "title": "SpaceByte: Towards Deleting Tokenization from Large Language Modeling",
            "url": "http://arxiv.org/abs/2404.14408v3",
            "arxiv_id": "2404.14408",
            "authors": [
                "Kevin Slagle"
            ],
            "abstract": "Tokenization is widely used in large language models because it significantly improves performance. However, tokenization imposes several disadvantages, such as performance biases, increased adversarial vulnerability, decreased character-level modeling performance, and increased modeling complexity. To address these disadvantages without sacrificing performance, we propose SpaceByte, a novel byte-level decoder architecture that closes the performance gap between byte-level and subword autoregressive language modeling. SpaceByte consists of a byte-level Transformer model, but with extra larger transformer blocks inserted in the middle of the layers. We find that performance is significantly improved by applying these larger blocks only after certain bytes, such as space characters, which typically denote word boundaries. Our experiments show that for a fixed training and inference compute budget, SpaceByte outperforms other byte-level architectures and roughly matches the performance of tokenized Transformer architectures.",
            "introduction": "   1 Introduction  Most language models are trained using tokenization, which partitions text into tokens that typically consist of words or subwords. Tokenization is useful because it significantly decreases the inference and training computational costs of large language models. However, tokenization also imposes several disadvantages, including performance penalties for languages that are priortizes less by the tokenizer [1, 2, 3]; increased vulnerability to adversarial attacks [4]; and worse character-level modeling performance [5, 6], and additional model complexity.111See also Andrej Karpathy\u2019s tweet twitter.com/karpathy/status/1657949234535211009 and video youtube.com/watch?v=zduSFxRajkE&t=6725s on the disadvantages of tokenization.   Recently, MegaByte [7], MambaByte [6], and more [8, 9, 10, 11, 12] have been proposed as new byte-level autoregressive language models that model bytes instead of tokens. (See [13, 14, 15, 16, 17, 18, 19, 20, 21] for encoder and encoder-decoder byte-level modeling.) To address the longer context size resulting from modeling bytes instead of tokens, MegaByte uses multiscale modeling [22, 23, 24], while MambaByte uses Mamba blocks [25] instead of Transformer blocks. But although MegaByte and MambaByte have been shown to perform better than a standard byte-level Transformer, to our knowledge, no byte-level autoregressive large language model architecture has been shown to match the performance of tokenized models when controlling for compute costs.   In this work, we study the performance of byte-level and subword-level autoregressive models when trained using a fixed compute budget. We measure the performance in terms of the cross entropy (measured in bits-per-byte), which has been shown to be a strong predictor of down-stream performance [26]. In addition to controlling for training compute, we also control for inference compute costs (measured in FLOPs). We find that byte-level Transformer and MegaByte models can require roughly 10 times more training FLOPs to achieve the same performance as a subword-level Transformer. To close this substantial performance gap, we propose a new byte-level decoder architecture: SpaceByte.   SpaceByte also utilizes multiscale modeling to improve efficiency by grouping bytes into patches. But unlike MegaByte, which uses a fixed patch size, SpaceByte uses a simple rule to dynamically partition the bytes into patches that are aligned with word and other language boundaries. Our compute-controlled experiments show that this simple modification is crucial for performance, allowing SpaceByte to outperform other byte-level architectures and roughly match the performance of subword Transformers across a variety of text modalities.   Our experiments are performed on datasets consisting of English books, LaTeX formatted arXiv papers, and open-source code. For other data modalities, SpaceByte with our simple patching rule might not be as effective.   Figure 1:  An overview of the SpaceByte architecture. The embedding, local transformer blocks, and de-embedding (i.e. a layer norm and linear) are the standard Transformer decoder layers. SpaceByte modifies the standard transformer by applying \u201cglobal\u201d transformer blocks only after certain bytes, such as space characters. The intuition is that the first character of a word is typically the hardest to predict; thus this positioning of the global blocks should make the best use of the global blocks (which use a larger model dimension).      2 SpaceByte  The SpaceByte architecture is summarized in Figure\u00a01. In a nutshell, SpaceByte can be thought of as a byte-level Transformer model, but with extra \u201cglobal\u201d transformer blocks (with a larger model dimension) inserted in the middle, which are only applied a fraction of the time. While the MegaByte architecture applies the global transformer blocks every P\u223c8similar-to\ud835\udc438P\\sim 8italic_P \u223c 8 bytes, we hypothesize that this fixed spacing hinders",
            "references": [
                {
                    "title": "Training LLMs over Neurally Compressed Text",
                    "abstract": "In this paper, we explore the idea of training large language models (LLMs) over highly compressed text. While standard subword tokenizers compress text by a small factor, neural text compressors can achieve much higher rates of compression. If it were possible to train LLMs directly over neurally compressed text, this would confer advantages in training and serving efficiency, as well as easier handling of long text spans. The main obstacle to this goal is that strong compression tends to produce opaque outputs that are not well-suited for learning. In particular, we find that text na\\\"ively compressed via Arithmetic Coding is not readily learnable by LLMs. To overcome this, we propose Equal-Info Windows, a novel compression technique whereby text is segmented into blocks that each compress to the same bit length. Using this method, we demonstrate effective learning over neurally compressed text that improves with scale, and outperforms byte-level baselines by a wide margin on perplexity and inference speed benchmarks. While our method delivers worse perplexity than subword tokenizers for models trained with the same parameter count, it has the benefit of shorter sequence lengths. Shorter sequence lengths require fewer autoregressive generation steps, and reduce latency. Finally, we provide extensive analysis of the properties that contribute to learnability, and offer concrete suggestions for how to further improve the performance of high-compression tokenizers."
                },
                {
                    "title": "Mixture-of-Depths: Dynamically allocating compute in transformer-based language models",
                    "abstract": "Transformer-based language models spread FLOPs uniformly across input sequences. In this work we demonstrate that transformers can instead learn to dynamically allocate FLOPs (or compute) to specific positions in a sequence, optimising the allocation along the sequence for different layers across the model depth. Our method enforces a total compute budget by capping the number of tokens ($k$) that can participate in the self-attention and MLP computations at a given layer. The tokens to be processed are determined by the network using a top-$k$ routing mechanism. Since $k$ is defined a priori, this simple procedure uses a static computation graph with known tensor sizes, unlike other conditional computation techniques. Nevertheless, since the identities of the $k$ tokens are fluid, this method can expend FLOPs non-uniformly across the time and model depth dimensions. Thus, compute expenditure is entirely predictable in sum total, but dynamic and context-sensitive at the token-level. Not only do models trained in this way learn to dynamically allocate compute, they do so efficiently. These models match baseline performance for equivalent FLOPS and wall-clock times to train, but require a fraction of the FLOPs per forward pass, and can be upwards of 50\\% faster to step during post-training sampling."
                },
                {
                    "title": "MYTE: Morphology-Driven Byte Encoding for Better and Fairer Multilingual Language Modeling",
                    "abstract": "A major consideration in multilingual language modeling is how to best represent languages with diverse vocabularies and scripts. Although contemporary text encoding methods cover most of the world's writing systems, they exhibit bias towards the high-resource languages of the Global West. As a result, texts of underrepresented languages tend to be segmented into long sequences of linguistically meaningless units. To address the disparities, we introduce a new paradigm that encodes the same information with segments of consistent size across diverse languages. Our encoding convention (MYTE) is based on morphemes, as their inventories are more balanced across languages than characters, which are used in previous methods. We show that MYTE produces shorter encodings for all 99 analyzed languages, with the most notable improvements for non-European languages and non-Latin scripts. This, in turn, improves multilingual LM performance and diminishes the perplexity gap throughout diverse languages."
                },
                {
                    "title": "MambaByte: Token-free Selective State Space Model",
                    "abstract": "Token-free language models learn directly from raw bytes and remove the inductive bias of subword tokenization. Operating on bytes, however, results in significantly longer sequences. In this setting, standard autoregressive Transformers scale poorly as the effective memory required grows with sequence length. The recent development of the Mamba state space model (SSM) offers an appealing alternative approach with a fixed-sized memory state and efficient decoding. We propose MambaByte, a token-free adaptation of the Mamba SSM trained autoregressively on byte sequences. In terms of modeling, we show MambaByte to be competitive with, and even to outperform, state-of-the-art subword Transformers on language modeling tasks while maintaining the benefits of token-free language models, such as robustness to noise. In terms of efficiency, we develop an adaptation of speculative decoding with tokenized drafting and byte-level verification. This results in a $2.6\\times$ inference speedup to the standard MambaByte implementation, showing similar decoding efficiency as the subword Mamba. These findings establish the viability of SSMs in enabling token-free language modeling."
                },
                {
                    "title": "Mamba: Linear-Time Sequence Modeling with Selective State Spaces",
                    "abstract": "Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5$\\times$ higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation."
                },
                {
                    "title": "Toucan: Token-Aware Character Level Language Modeling",
                    "abstract": "Character-level language models obviate the need for separately trained tokenizers, but efficiency suffers from longer sequence lengths. Learning to combine character representations into tokens has made training these models more efficient, but they still require decoding characters individually. We propose Toucan, an augmentation to character-level models to make them\"token-aware\". Comparing our method to prior work, we demonstrate significant speed-ups in character generation without a loss in language modeling performance. We then explore differences between our learned dynamic tokenization of character sequences with popular fixed vocabulary solutions such as Byte-Pair Encoding and WordPiece, finding our approach leads to a greater amount of longer sequences tokenized as single items. Our project and code are available at https://nlp.jhu.edu/nuggets/."
                },
                {
                    "title": "Learn Your Tokens: Word-Pooled Tokenization for Language Modeling",
                    "abstract": "Language models typically tokenize text into subwords, using a deterministic, hand-engineered heuristic of combining characters into longer surface-level strings such as 'ing' or whole words. Recent literature has repeatedly shown the limitations of such a tokenization strategy, particularly for documents not written in English and for representing numbers. On the other extreme, byte/character-level language models are much less restricted but suffer from increased sequence description lengths and a subsequent quadratic expansion in self-attention computation. Recent attempts to compress and limit these context lengths with fixed size convolutions is helpful but completely ignores the word boundary. This paper considers an alternative 'learn your tokens' scheme which utilizes the word boundary to pool bytes/characters into word representations, which are fed to the primary language model, before again decoding individual characters/bytes per word in parallel. We find that our moderately expressive and moderately fast end-to-end tokenizer outperform by over 300% both subwords and byte/character models over the intrinsic language modeling metric of next-word prediction across datasets. It particularly outshines on rare words, outperforming by a factor of 30! We extensively study the language modeling setup for all three categories of tokenizers and theoretically analyze how our end-to-end models can also be a strong trade-off in efficiency and robustness."
                },
                {
                    "title": "Small-scale proxies for large-scale Transformer training instabilities",
                    "abstract": "Teams that have trained large Transformer-based models have reported training instabilities at large scale that did not appear when training with the same hyperparameters at smaller scales. Although the causes of such instabilities are of scientific interest, the amount of resources required to reproduce them has made investigation difficult. In this work, we seek ways to reproduce and study training stability and instability at smaller scales. First, we focus on two sources of training instability described in previous work: the growth of logits in attention layers (Dehghani et al., 2023) and divergence of the output logits from the log probabilities (Chowdhery et al., 2022). By measuring the relationship between learning rate and loss across scales, we show that these instabilities also appear in small models when training at high learning rates, and that mitigations previously employed at large scales are equally effective in this regime. This prompts us to investigate the extent to which other known optimizer and model interventions influence the sensitivity of the final loss to changes in the learning rate. To this end, we study methods such as warm-up, weight decay, and the $\\mu$Param (Yang et al., 2022), and combine techniques to train small models that achieve similar losses across orders of magnitude of learning rate variation. Finally, to conclude our exploration we study two cases where instabilities can be predicted before they emerge by examining the scaling behavior of model activation and gradient norms."
                },
                {
                    "title": "FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning",
                    "abstract": "Scaling Transformers to longer sequence lengths has been a major problem in the last several years, promising to improve performance in language modeling and high-resolution image understanding, as well as to unlock new applications in code, audio, and video generation. The attention layer is the main bottleneck in scaling to longer sequences, as its runtime and memory increase quadratically in the sequence length. FlashAttention exploits the asymmetric GPU memory hierarchy to bring significant memory saving (linear instead of quadratic) and runtime speedup (2-4$\\times$ compared to optimized baselines), with no approximation. However, FlashAttention is still not nearly as fast as optimized matrix-multiply (GEMM) operations, reaching only 25-40\\% of the theoretical maximum FLOPs/s. We observe that the inefficiency is due to suboptimal work partitioning between different thread blocks and warps on the GPU, causing either low-occupancy or unnecessary shared memory reads/writes. We propose FlashAttention-2, with better work partitioning to address these issues. In particular, we (1) tweak the algorithm to reduce the number of non-matmul FLOPs (2) parallelize the attention computation, even for a single head, across different thread blocks to increase occupancy, and (3) within each thread block, distribute the work between warps to reduce communication through shared memory. These yield around 2$\\times$ speedup compared to FlashAttention, reaching 50-73\\% of the theoretical maximum FLOPs/s on A100 and getting close to the efficiency of GEMM operations. We empirically validate that when used end-to-end to train GPT-style models, FlashAttention-2 reaches training speed of up to 225 TFLOPs/s per A100 GPU (72\\% model FLOPs utilization)."
                },
                {
                    "title": "Do All Languages Cost the Same? Tokenization in the Era of Commercial Language Models",
                    "abstract": "Language models have graduated from being research prototypes to commercialized products offered as web APIs, and recent works have highlighted the multilingual capabilities of these products. The API vendors charge their users based on usage, more specifically on the number of ``tokens'' processed or generated by the underlying language models. What constitutes a token, however, is training data and model dependent with a large variance in the number of tokens required to convey the same information in different languages. In this work, we analyze the effect of this non-uniformity on the fairness of an API's pricing policy across languages. We conduct a systematic analysis of the cost and utility of OpenAI's language model API on multilingual benchmarks in 22 typologically diverse languages. We show evidence that speakers of a large number of the supported languages are overcharged while obtaining poorer results. These speakers tend to also come from regions where the APIs are less affordable to begin with. Through these analyses, we aim to increase transparency around language model APIs' pricing policies and encourage the vendors to make them more equitable."
                },
                {
                    "title": "GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints",
                    "abstract": "Multi-query attention (MQA), which only uses a single key-value head, drastically speeds up decoder inference. However, MQA can lead to quality degradation, and moreover it may not be desirable to train a separate model just for faster inference. We (1) propose a recipe for uptraining existing multi-head language model checkpoints into models with MQA using 5% of original pre-training compute, and (2) introduce grouped-query attention (GQA), a generalization of multi-query attention which uses an intermediate (more than one, less than number of query heads) number of key-value heads. We show that uptrained GQA achieves quality close to multi-head attention with comparable speed to MQA."
                },
                {
                    "title": "Language Model Tokenizers Introduce Unfairness Between Languages",
                    "abstract": "Recent language models have shown impressive multilingual performance, even when not explicitly trained for it. Despite this, there are concerns about the quality of their outputs across different languages. In this paper, we show how disparity in the treatment of different languages arises at the tokenization stage, well before a model is even invoked. The same text translated into different languages can have drastically different tokenization lengths, with differences up to 15 times in some cases. These disparities persist even for tokenizers that are intentionally trained for multilingual support. Character-level and byte-level models also exhibit over 4 times the difference in the encoding length for some language pairs. This induces unfair treatment for some language communities in regard to the cost of accessing commercial language services, the processing time and latency, as well as the amount of content that can be provided as context to the models. Therefore, we make the case that we should train future language models using multilingually fair subword tokenizers."
                },
                {
                    "title": "MEGABYTE: Predicting Million-byte Sequences with Multiscale Transformers",
                    "abstract": "Autoregressive transformers are spectacular models for short sequences but scale poorly to long sequences such as high-resolution images, podcasts, code, or books. We proposed Megabyte, a multi-scale decoder architecture that enables end-to-end differentiable modeling of sequences of over one million bytes. Megabyte segments sequences into patches and uses a local submodel within patches and a global model between patches. This enables sub-quadratic self-attention, much larger feedforward layers for the same compute, and improved parallelism during decoding -- unlocking better performance at reduced cost for both training and generation. Extensive experiments show that Megabyte allows byte-level models to perform competitively with subword models on long context language modeling, achieve state-of-the-art density estimation on ImageNet, and model audio from raw files. Together, these results establish the viability of tokenization-free autoregressive sequence modeling at scale."
                },
                {
                    "title": "What is the best recipe for character-level encoder-only modelling?",
                    "abstract": "This paper aims to benchmark recent progress in language understanding models that output contextualised representations at the character level. Many such modelling architectures and methods to train those architectures have been proposed, but it is currently unclear what the relative contributions of the architecture vs. the pretraining objective are to final model performance. We explore the design space of such models, comparing architectural innovations (Clark et al., 2022, Jaegle et al., 2022, Tay et al., 2021) and a variety of different pretraining objectives on a suite of evaluation tasks with a fixed training procedure in order to find the currently optimal way to build and train character-level BERT-like models. We find that our best performing character-level model exceeds the performance of a token-based model trained with the same settings on the same data, suggesting that character-level models are ready for more widespread adoption. Unfortunately, the best method to train character-level models still relies on a subword-level tokeniser during pretraining, and final model performance is highly dependent on tokeniser quality. We believe our results demonstrate the readiness of character-level models for multilingual language representation, and encourage NLP practitioners to try them as drop-in replacements for token-based models."
                },
                {
                    "title": "A Systematic Analysis of Vocabulary and BPE Settings for Optimal Fine-tuning of NMT: A Case Study of In-domain Translation",
                    "abstract": "The effectiveness of Neural Machine Translation (NMT) models largely depends on the vocabulary used at training; small vocabularies can lead to out-of-vocabulary problems -- large ones, to memory issues. Subword (SW) tokenization has been successfully employed to mitigate these issues. The choice of vocabulary and SW tokenization has a significant impact on both training and fine-tuning an NMT model. Fine-tuning is a common practice in optimizing an MT model with respect to new data. However, new data potentially introduces new words (or tokens), which, if not taken into consideration, may lead to suboptimal performance. In addition, the distribution of tokens in the new data can differ from the distribution of the original data. As such, the original SW tokenization model could be less suitable for the new data. Through a systematic empirical evaluation, in this work we compare different strategies for SW tokenization and vocabulary generation with the ultimate goal to uncover an optimal setting for fine-tuning a domain-specific model. Furthermore, we developed several (in-domain) models, the best of which achieves 6 BLEU points improvement over the baseline."
                },
                {
                    "title": "Are Character-level Translations Worth the Wait? Comparing ByT5 and mT5 for Machine Translation",
                    "abstract": "Pretrained character-level and byte-level language models have been shown to be competitive with popular subword models across a range of Natural Language Processing tasks. However, there has been little research on their effectiveness for neural machine translation (NMT), particularly within the popular pretrain-then-finetune paradigm. This work performs an extensive comparison across multiple languages and experimental conditions of character- and subword-level pretrained models (ByT5 and mT5, respectively) on NMT. We show the effectiveness of character-level modeling in translation, particularly in cases where fine-tuning data is limited. In our analysis, we show how character models\u2019 gains in translation quality are reflected in better translations of orthographically similar words and rare words. While evaluating the importance of source texts in driving model predictions, we highlight word-level patterns within ByT5, suggesting an ability to modulate word-level and character-level information during generation. We conclude by assessing the efficiency tradeoff of byte models, suggesting their usage in non-time-critical scenarios to boost translation quality."
                },
                {
                    "title": "Scaling Vision Transformers to 22 Billion Parameters",
                    "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                },
                {
                    "title": "MANTa: Efficient Gradient-Based Tokenization for Robust End-to-End Language Modeling",
                    "abstract": "Static subword tokenization algorithms have been an essential component of recent works on language modeling. However, their static nature results in important flaws that degrade the models' downstream performance and robustness. In this work, we propose MANTa, a Module for Adaptive Neural TokenizAtion. MANTa is a differentiable tokenizer trained end-to-end with the language model. The resulting system offers a trade-off between the expressiveness of byte-level models and the speed of models trained using subword tokenization. In addition, our tokenizer is highly explainable since it produces an explicit segmentation of sequences into blocks. We evaluate our pre-trained model on several English datasets from different domains as well as on synthetic noise. We find that MANTa improves robustness to character perturbations and out-of-domain data. We then show that MANTa performs comparably to other models on the general-domain GLUE benchmark. Finally, we show that it is considerably faster than strictly byte-level models."
                },
                {
                    "title": "Subword-Delimited Downsampling for Better Character-Level Translation",
                    "abstract": "Subword-level models have been the dominant paradigm in NLP. However, character-level models have the benefit of seeing each character individually, providing the model with more detailed information that ultimately could lead to better models. Recent works have shown character-level models to be competitive with subword models, but costly in terms of time and computation. Character-level models with a downsampling component alleviate this, but at the cost of quality, particularly for machine translation. This work analyzes the problems of previous downsampling methods and introduces a novel downsampling method which is informed by subwords. This new downsampling method not only outperforms existing downsampling methods, showing that downsampling characters can be done without sacrificing quality, but also leads to promising performance compared to subword models for translation."
                },
                {
                    "title": "Efficient Transformers with Dynamic Token Pooling",
                    "abstract": "Transformers achieve unrivalled performance in modelling language, but remain inefficient in terms of memory and time complexity. A possible remedy is to reduce the sequence length in the intermediate layers by pooling fixed-length segments of tokens. Nevertheless, natural units of meaning, such as words or phrases, display varying sizes. To address this mismatch, we equip language models with a dynamic-pooling mechanism, which predicts segment boundaries in an autoregressive fashion. We compare several methods to infer boundaries, including end-to-end learning through stochastic re-parameterisation, supervised learning (based on segmentations from subword tokenizers or spikes in conditional entropy), as well as linguistically motivated boundaries. We perform character-level evaluation on texts from multiple datasets and morphologically diverse languages. The results demonstrate that dynamic pooling, which jointly segments and models language, is both faster and more accurate than vanilla Transformers and fixed-length pooling within the same computational budget."
                },
                {
                    "title": "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness",
                    "abstract": "Transformers are slow and memory-hungry on long sequences, since the time and memory complexity of self-attention are quadratic in sequence length. Approximate attention methods have attempted to address this problem by trading off model quality to reduce the compute complexity, but often do not achieve wall-clock speedup. We argue that a missing principle is making attention algorithms IO-aware -- accounting for reads and writes between levels of GPU memory. We propose FlashAttention, an IO-aware exact attention algorithm that uses tiling to reduce the number of memory reads/writes between GPU high bandwidth memory (HBM) and GPU on-chip SRAM. We analyze the IO complexity of FlashAttention, showing that it requires fewer HBM accesses than standard attention, and is optimal for a range of SRAM sizes. We also extend FlashAttention to block-sparse attention, yielding an approximate attention algorithm that is faster than any existing approximate attention method. FlashAttention trains Transformers faster than existing baselines: 15% end-to-end wall-clock speedup on BERT-large (seq. length 512) compared to the MLPerf 1.1 training speed record, 3$\\times$ speedup on GPT-2 (seq. length 1K), and 2.4$\\times$ speedup on long-range arena (seq. length 1K-4K). FlashAttention and block-sparse FlashAttention enable longer context in Transformers, yielding higher quality models (0.7 better perplexity on GPT-2 and 6.4 points of lift on long-document classification) and entirely new capabilities: the first Transformers to achieve better-than-chance performance on the Path-X challenge (seq. length 16K, 61.4% accuracy) and Path-256 (seq. length 64K, 63.1% accuracy)."
                },
                {
                    "title": "Local Byte Fusion for Neural Machine Translation",
                    "abstract": "Subword tokenization schemes are the dominant technique used in current NLP models. However, such schemes can be rigid and tokenizers built on one corpus may not adapt well to other parallel corpora. It has also been observed that in multilingual corpora, subword tokenization schemes oversegment low-resource languages, leading to a drop in translation performance. An alternative to subword tokenizers is byte-based tokenization, i.e., tokenization into byte sequences using the UTF-8 encoding scheme. Byte tokens often represent inputs at a sub-character granularity, i.e., one character can be represented by a span of byte tokens. This results in much longer byte sequences that are hard to interpret without aggregating local information from multiple byte tokens. In this paper, we propose a Local Byte Fusion (LOBEF) method for byte-based machine translation\u2014utilizing byte n-gram and word boundaries\u2014to aggregate local semantic information. Extensive experiments on multilingual translation, zero-shot cross-lingual transfer, and domain adaptation reveal a consistent improvement over vanilla byte-based models. Further analysis also indicates that our byte-based models are parameter-efficient and perform competitive to subword models."
                },
                {
                    "title": "A Vocabulary-Free Multilingual Neural Tokenizer for End-to-End Task Learning",
                    "abstract": "Subword tokenization is a commonly used input pre-processing step in most recent NLP models. However, it limits the models\u2019 ability to leverage end-to-end task learning. Its frequency-based vocabulary creation compromises tokenization in low-resource languages, leading models to produce suboptimal representations. Additionally, the dependency on a fixed vocabulary limits the subword models\u2019 adaptability across languages and domains. In this work, we propose a vocabulary-free neural tokenizer by distilling segmentation information from heuristic-based subword tokenization. We pre-train our character-based tokenizer by processing unique words from multilingual corpus, thereby extensively increasing word diversity across languages. Unlike the predefined and fixed vocabularies in subword methods, our tokenizer allows end-to-end task learning, resulting in optimal task-specific tokenization. The experimental results show that replacing the subword tokenizer with our neural tokenizer consistently improves performance on multilingual (NLI) and code-switching (sentiment analysis) tasks, with larger gains in low-resource languages. Additionally, our neural tokenizer exhibits a robust performance on downstream tasks when adversarial noise is present (typos and misspelling), further increasing the initial improvements over statistical subword tokenizers."
                },
                {
                    "title": "Hierarchical Transformers Are More Efficient Language Models",
                    "abstract": "Transformer models yield impressive results on many NLP and sequence modeling tasks. Remarkably, Transformers can handle long sequences which allows them to produce long coherent outputs: full paragraphs produced by GPT-3 or well-structured images produced by DALL-E. These large language models are impressive but also very inefficient and costly, which limits their applications and accessibility. We postulate that having an explicit hierarchical architecture is the key to Transformers that efficiently handle long sequences. To verify this claim, we first study different ways to downsample and upsample activations in Transformers so as to make them hierarchical. We use the best performing upsampling and downsampling layers to create Hourglass - a hierarchical Transformer language model. Hourglass improves upon the Transformer baseline given the same amount of computation and can yield the same results as Transformers more efficiently. In particular, Hourglass sets new state-of-the-art for Transformer models on the ImageNet32 generation task and improves language modeling efficiency on the widely studied enwik8 benchmark."
                },
                {
                    "title": "Charformer: Fast Character Transformers via Gradient-based Subword Tokenization",
                    "abstract": "State-of-the-art models in natural language processing rely on separate rigid subword tokenization algorithms, which limit their generalization ability and adaptation to new settings. In this paper, we propose a new model inductive bias that learns a subword tokenization end-to-end as part of the model. To this end, we introduce a soft gradient-based subword tokenization module (GBST) that automatically learns latent subword representations from characters in a data-driven fashion. Concretely, GBST enumerates candidate subword blocks and learns to score them in a position-wise fashion using a block scoring network. We additionally introduce Charformer, a deep Transformer model that integrates GBST and operates on the byte level. Via extensive experiments on English GLUE, multilingual, and noisy text datasets, we show that Charformer outperforms a series of competitive byte-level baselines while generally performing on par and sometimes outperforming subword-based models. Additionally, Charformer is fast, improving the speed of both vanilla byte-level and subword-level Transformers by 28%-100% while maintaining competitive quality. We believe this work paves the way for highly performant token-free models that are trained completely end-to-end."
                },
                {
                    "title": "ByT5: Towards a Token-Free Future with Pre-trained Byte-to-Byte Models",
                    "abstract": "Most widely used pre-trained language models operate on sequences of tokens corresponding to word or subword units. By comparison, token-free models that operate directly on raw text (bytes or characters) have many benefits: They can process text in any language out of the box, they are more robust to noise, and they minimize technical debt by removing complex and error-prone text preprocessing pipelines. Because byte or character sequences are longer than token sequences, past work on token-free models has often introduced new model architectures designed to amortize the cost of operating directly on raw text. In this paper, we show that a standard Transformer architecture can be used with minimal modifications to process byte sequences. We characterize the trade-offs in terms of parameter count, training FLOPs, and inference speed, and show that byte-level models are competitive with their token-level counterparts. We also demonstrate that byte-level models are significantly more robust to noise and perform better on tasks that are sensitive to spelling and pronunciation. As part of our contribution, we release a new set of pre-trained byte-level Transformer models based on the T5 architecture, as well as all code and data used in our experiments.1"
                },
                {
                    "title": "Consistent Accelerated Inference via Confident Adaptive Transformers",
                    "abstract": "We develop a novel approach for confidently accelerating inference in the large and expensive multilayer Transformers that are now ubiquitous in natural language processing (NLP). Amortized or approximate computational methods increase efficiency, but can come with unpredictable performance costs. In this work, we present CATs \u2013 Confident Adaptive Transformers \u2013 in which we simultaneously increase computational efficiency, while guaranteeing a specifiable degree of consistency with the original model with high confidence. Our method trains additional prediction heads on top of intermediate layers, and dynamically decides when to stop allocating computational effort to each input using a meta consistency classifier. To calibrate our early prediction stopping rule, we formulate a unique extension of conformal prediction. We demonstrate the effectiveness of this approach on four classification and regression tasks."
                },
                {
                    "title": "Canine: Pre-training an Efficient Tokenization-Free Encoder for Language Representation",
                    "abstract": "Pipelined NLP systems have largely been superseded by end-to-end neural modeling, yet nearly all commonly used models still require an explicit tokenization step. While recent tokenization approaches based on data-derived subword lexicons are less brittle than manually engineered tokenizers, these techniques are not equally suited to all languages, and the use of any fixed vocabulary may limit a model\u2019s ability to adapt. In this paper, we present Canine, a neural encoder that operates directly on character sequences\u2014without explicit tokenization or vocabulary\u2014and a pre-training strategy that operates either directly on characters or optionally uses subwords as a soft inductive bias. To use its finer-grained input effectively and efficiently, Canine combines downsampling, which reduces the input sequence length, with a deep transformer stack, which encodes context. Canine outperforms a comparable mBert model by 5.7 F1 on TyDi QA, a challenging multilingual benchmark, despite having fewer model parameters."
                },
                {
                    "title": "Dynamic Neural Networks: A Survey",
                    "abstract": "Dynamic neural network is an emerging research topic in deep learning. Compared to static models which have fixed computational graphs and parameters at the inference stage, dynamic networks can adapt their structures or parameters to different inputs, leading to notable advantages in terms of accuracy, computational efficiency, adaptiveness, etc. In this survey, we comprehensively review this rapidly developing area by dividing dynamic networks into three main categories: 1) sample-wise dynamic models that process each sample with data-dependent architectures or parameters; 2) spatial-wise dynamic networks that conduct adaptive computation with respect to different spatial locations of image data; and 3) temporal-wise dynamic models that perform adaptive inference along the temporal dimension for sequential data such as videos and texts. The important research problems of dynamic networks, e.g., architecture design, decision making scheme, optimization technique and applications, are reviewed systematically. Finally, we discuss the open problems in this field together with interesting future research directions."
                },
                {
                    "title": "Shortformer: Better Language Modeling using Shorter Inputs",
                    "abstract": "Increasing the input length has been a driver of progress in language modeling with transformers. We identify conditions where shorter inputs are not harmful, and achieve perplexity and efficiency improvements through two new methods that decrease input length. First, we show that initially training a model on short subsequences before moving on to longer ones both reduces overall training time and, surprisingly, substantially improves perplexity. Second, we show how to improve the efficiency of recurrence methods in transformers, which let models condition on previously processed tokens when generating sequences that exceed the maximal length the transformer can handle at once. Existing methods require computationally expensive relative position embeddings; we introduce a simple alternative of adding absolute position embeddings to queries and keys instead of to word embeddings, which efficiently produces superior results. We show that these recurrent models also benefit from short input lengths. Combining these techniques speeds up training by a factor of 1.65, reduces memory usage, and substantially improves perplexity on WikiText-103, without adding any parameters."
                },
                {
                    "title": "The Pile: An 800GB Dataset of Diverse Text for Language Modeling",
                    "abstract": "Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present the Pile : an 825 GiB English text corpus tar-geted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets\u2014both existing and newly constructed\u2014many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve signi\ufb01cantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction. 1"
                },
                {
                    "title": "The Depth-to-Width Interplay in Self-Attention.",
                    "abstract": "Self-attention architectures, which are rapidly pushing the frontier in natural language processing, demonstrate a surprising depth-inefficient behavior: previous works indicate that increasing the internal representation (network width) is just as useful as increasing the number of self-attention layers (network depth). We theoretically predict a width-dependent transition between depth-efficiency and depth-inefficiency in self-attention. We conduct systematic empirical ablations on networks of depths 6 to 48 that clearly reveal the theoretically predicted behaviors, and provide explicit quantitative suggestions regarding the optimal depth-to-width allocation for a given self-attention network size. The race towards beyond 1-Trillion parameter language models renders informed guidelines for increasing self-attention depth and width in tandem an essential ingredient. Our guidelines elucidate the depth-to-width trade-off in self-attention networks of sizes up to the scale of GPT3 (which is too deep for its size), and beyond, marking an unprecedented width of 30K as optimal for a 1-Trillion parameter self-attention network."
                },
                {
                    "title": "CharacterBERT: Reconciling ELMo and BERT for Word-Level Open-Vocabulary Representations From Characters",
                    "abstract": "Due to the compelling improvements brought by BERT, many recent representation models adopted the Transformer architecture as their main building block, consequently inheriting the wordpiece tokenization system despite it not being intrinsically linked to the notion of Transformers. While this system is thought to achieve a good balance between the flexibility of characters and the efficiency of full words, using predefined wordpiece vocabularies from the general domain is not always suitable, especially when building models for specialized domains (e.g., the medical domain). Moreover, adopting a wordpiece tokenization shifts the focus from the word level to the subword level, making the models conceptually more complex and arguably less convenient in practice. For these reasons, we propose CharacterBERT, a new variant of BERT that drops the wordpiece system altogether and uses a Character-CNN module instead to represent entire words by consulting their characters. We show that this new model improves the performance of BERT on a variety of medical domain tasks while at the same time producing robust, word-level, and open-vocabulary representations."
                },
                {
                    "title": "Query-Key Normalization for Transformers",
                    "abstract": "Low-resource language translation is a challenging but socially valuable NLP task. Building on recent work adapting the Transformer\u2019s normalization to this setting, we propose QKNorm, a normalization technique that modifies the attention mechanism to make the softmax function less prone to arbitrary saturation without sacrificing expressivity. Specifically, we apply l2-normalization along the head dimension of each query and key matrix prior to multiplying them and then scale up by a learnable parameter instead of dividing by the square root of the embedding dimension. We show improvements averaging 0.928 BLEU over state-of-the-art bilingual benchmarks for 5 low-resource translation pairs from the TED Talks corpus and IWSLT\u201915."
                },
                {
                    "title": "Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processing",
                    "abstract": "With the success of language pretraining, it is highly desirable to develop more efficient architectures of good scalability that can exploit the abundant unlabeled data at a lower cost. To improve the efficiency, we examine the much-overlooked redundancy in maintaining a full-length token-level presentation, especially for tasks that only require a single-vector presentation of the sequence. With this intuition, we propose Funnel-Transformer which gradually compresses the sequence of hidden states to a shorter one and hence reduces the computation cost. More importantly, by re-investing the saved FLOPs from length reduction in constructing a deeper or wider model, we further improve the model capacity. In addition, to perform token-level predictions as required by common pretraining objectives, Funnel-Transformer is able to recover a deep representation for each token from the reduced hidden sequence via a decoder. Empirically, with comparable or fewer FLOPs, Funnel-Transformer outperforms the standard Transformer on a wide variety of sequence-level prediction tasks, including text classification, language understanding, and reading comprehension. The code and pretrained checkpoints are available at this https URL."
                },
                {
                    "title": "Multi-scale Transformer Language Models",
                    "abstract": "We investigate multi-scale transformer language models that learn representations of text at multiple scales, and present three different architectures that have an inductive bias to handle the hierarchical nature of language. Experiments on large-scale language modeling benchmarks empirically demonstrate favorable likelihood vs memory footprint trade-offs, e.g. we show that it is possible to train a hierarchical variant with 30 layers that has 23% smaller memory footprint and better perplexity, compared to a vanilla transformer with less than half the number of layers, on the Toronto BookCorpus. We analyze the advantages of learned representations at multiple scales in terms of memory footprint, compute time, and perplexity, which are particularly appealing given the quadratic scaling of transformers' run time and memory usage with respect to sequence length."
                },
                {
                    "title": "Longformer: The Long-Document Transformer",
                    "abstract": "Transformer-based models are unable to process long sequences due to their self-attention operation, which scales quadratically with the sequence length. To address this limitation, we introduce the Longformer with an attention mechanism that scales linearly with sequence length, making it easy to process documents of thousands of tokens or longer. Longformer's attention mechanism is a drop-in replacement for the standard self-attention and combines a local windowed attention with a task motivated global attention. Following prior work on long-sequence transformers, we evaluate Longformer on character-level language modeling and achieve state-of-the-art results on text8 and enwik8. In contrast to most prior work, we also pretrain Longformer and finetune it on a variety of downstream tasks. Our pretrained Longformer consistently outperforms RoBERTa on long document tasks and sets new state-of-the-art results on WikiHop and TriviaQA. We finally introduce the Longformer-Encoder-Decoder (LED), a Longformer variant for supporting long document generative sequence-to-sequence tasks, and demonstrate its effectiveness on the arXiv summarization dataset."
                },
                {
                    "title": "Compressive Transformers for Long-Range Sequence Modelling",
                    "abstract": "We present the Compressive Transformer, an attentive sequence model which compresses past memories for long-range sequence learning. We find the Compressive Transformer obtains state-of-the-art language modelling results in the WikiText-103 and Enwik8 benchmarks, achieving 17.1 ppl and 0.97 bpc respectively. We also find it can model high-frequency speech effectively and can be used as a memory mechanism for RL, demonstrated on an object matching task. To promote the domain of long-range sequence learning, we propose a new open-vocabulary language modelling benchmark derived from books, PG-19."
                },
                {
                    "title": "Fast Transformer Decoding: One Write-Head is All You Need",
                    "abstract": "Multi-head attention layers, as used in the Transformer neural sequence model, are a powerful alternative to RNNs for moving information across and between sequences. While training these layers is generally fast and simple, due to parallelizability across the length of the sequence, incremental inference (where such paralleization is impossible) is often slow, due to the memory-bandwidth cost of repeatedly loading the large \"keys\" and \"values\" tensors. We propose a variant called multi-query attention, where the keys and values are shared across all of the different attention \"heads\", greatly reducing the size of these tensors and hence the memory bandwidth requirements of incremental decoding. We verify experimentally that the resulting models can indeed be much faster to decode, and incur only minor quality degradation from the baseline."
                },
                {
                    "title": "Learning Deep Transformer Models for Machine Translation",
                    "abstract": "Transformer is the state-of-the-art model in recent machine translation evaluations. Two strands of research are promising to improve models of this kind: the first uses wide networks (a.k.a. Transformer-Big) and has been the de facto standard for development of the Transformer system, and the other uses deeper language representation but faces the difficulty arising from learning deep networks. Here, we continue the line of research on the latter. We claim that a truly deep Transformer model can surpass the Transformer-Big counterpart by 1) proper use of layer normalization and 2) a novel way of passing the combination of previous layers to the next. On WMT\u201916 English-German and NIST OpenMT\u201912 Chinese-English tasks, our deep system (30/25-layer encoder) outperforms the shallow Transformer-Big/Base baseline (6-layer encoder) by 0.4-2.4 BLEU points. As another bonus, the deep model is 1.6X smaller in size and 3X faster in training than Transformer-Big."
                },
                {
                    "title": "Generating Long Sequences with Sparse Transformers",
                    "abstract": "Transformers are powerful sequence models, but require time and memory that grows quadratically with the sequence length. In this paper we introduce sparse factorizations of the attention matrix which reduce this to $O(n \\sqrt{n})$. We also introduce a) a variation on architecture and initialization to train deeper networks, b) the recomputation of attention matrices to save memory, and c) fast attention kernels for training. We call networks with these changes Sparse Transformers, and show they can model sequences tens of thousands of timesteps long using hundreds of layers. We use the same architecture to model images, audio, and text from raw bytes, setting a new state of the art for density modeling of Enwik8, CIFAR-10, and ImageNet-64. We generate unconditional samples that demonstrate global coherence and great diversity, and show it is possible in principle to use self-attention to model sequences of length one million or more."
                },
                {
                    "title": "Adaptive Input Representations for Neural Language Modeling",
                    "abstract": "We introduce adaptive input representations for neural language modeling which extend the adaptive softmax of Grave et al. (2017) to input representations of variable capacity. There are several choices on how to factorize the input and output layers, and whether to model words, characters or sub-word units. We perform a systematic comparison of popular choices for a self-attentional architecture. Our experiments show that models equipped with adaptive embeddings are more than twice as fast to train than the popular character input CNN while having a lower number of parameters. On the WikiText-103 benchmark we achieve 18.7 perplexity, an improvement of 10.5 perplexity compared to the previously best published result and on the Billion Word benchmark, we achieve 23.02 perplexity."
                },
                {
                    "title": "SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing",
                    "abstract": "This paper describes SentencePiece, a language-independent subword tokenizer and detokenizer designed for Neural-based text processing, including Neural Machine Translation. It provides open-source C++ and Python implementations for subword units. While existing subword segmentation tools assume that the input is pre-tokenized into word sequences, SentencePiece can train subword models directly from raw sentences, which allows us to make a purely end-to-end and language independent system. We perform a validation experiment of NMT on English-Japanese machine translation, and find that it is possible to achieve comparable accuracy to direct subword training from raw sentences. We also compare the performance of subword training and segmentation with various configurations. SentencePiece is available under the Apache 2 license at https://github.com/google/sentencepiece."
                },
                {
                    "title": "Character-Level Language Modeling with Deeper Self-Attention",
                    "abstract": "LSTMs and other RNN variants have shown strong performance on character-level language modeling. These models are typically trained using truncated backpropagation through time, and it is common to assume that their success stems from their ability to remember long-term contexts. In this paper, we show that a deep (64-layer) transformer model (Vaswani et al. 2017) with fixed context outperforms RNN variants by a large margin, achieving state of the art on two popular benchmarks: 1.13 bits per character on text8 and 1.06 on enwik8. To get good results at this depth, we show that it is important to add auxiliary losses, both at intermediate network layers and intermediate sequence positions."
                },
                {
                    "title": "A Simple Method for Commonsense Reasoning",
                    "abstract": "Commonsense reasoning is a long-standing challenge for deep learning. For example, it is difficult to use neural networks to tackle the Winograd Schema dataset (Levesque et al., 2011). In this paper, we present a simple method for commonsense reasoning with neural networks, using unsupervised learning. Key to our method is the use of language models, trained on a massive amount of unlabled data, to score multiple choice questions posed by commonsense reasoning tests. On both Pronoun Disambiguation and Winograd Schema challenges, our models outperform previous state-of-the-art methods by a large margin, without using expensive annotated knowledge bases or hand-engineered features. We train an array of large RNN language models that operate at word or character level on LM-1-Billion, CommonCrawl, SQuAD, Gutenberg Books, and a customized corpus for this task and show that diversity of training data plays an important role in test performance. Further analysis also shows that our system successfully discovers important features of the context that decide the correct answer, indicating a good grasp of commonsense knowledge."
                },
                {
                    "title": "BlockDrop: Dynamic Inference Paths in Residual Networks",
                    "abstract": "Very deep convolutional neural networks offer excellent recognition results, yet their computational expense limits their impact for many real-world applications. We introduce BlockDrop, an approach that learns to dynamically choose which layers of a deep network to execute during inference so as to best reduce total computation without degrading prediction accuracy. Exploiting the robustness of Residual Networks (ResNets) to layer dropping, our framework selects on-the-fly which residual blocks to evaluate for a given novel image. In particular, given a pretrained ResNet, we train a policy network in an associative reinforcement learning setting for the dual reward of utilizing a minimal number of blocks while preserving recognition accuracy. We conduct extensive experiments on CIFAR and ImageNet. The results provide strong quantitative and qualitative evidence that these learned policies not only accelerate inference but also encode meaningful visual information. Built upon a ResNet-101 model, our method achieves a speedup of 20% on average, going as high as 36% for some images, while maintaining the same 76.4% top-1 accuracy on ImageNet."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Using the Output Embedding to Improve Language Models",
                    "abstract": "We study the topmost weight matrix of neural network language models. We show that this matrix constitutes a valid word embedding. When training language models, we recommend tying the input embedding and this output embedding. We analyze the resulting update rules and show that the tied embedding evolves in a more similar way to the output embedding than to the input embedding in the untied model. We also offer a new method of regularizing the output embedding. Our methods lead to a significant reduction in perplexity, as we are able to show on a variety of neural network language models. Finally, we show that weight tying can reduce the size of neural translation models to less than half of their original size without harming their performance."
                },
                {
                    "title": "Adaptive Computation Time for Recurrent Neural Networks",
                    "abstract": "This paper introduces Adaptive Computation Time (ACT), an algorithm that allows recurrent neural networks to learn how many computational steps to take between receiving an input and emitting an output. ACT requires minimal changes to the network architecture, is deterministic and differentiable, and does not add any noise to the parameter gradients. Experimental results are provided for four synthetic problems: determining the parity of binary vectors, applying binary logic operations, adding integers, and sorting real numbers. Overall, performance is dramatically improved by the use of ACT, which successfully adapts the number of computational steps to the requirements of the problem. We also present character-level language modelling results on the Hutter prize Wikipedia dataset. In this case ACT does not yield large gains in performance; however it does provide intriguing insight into the structure of the data, with more computation allocated to harder-to-predict transitions, such as spaces between words and ends of sentences. This suggests that ACT or other adaptive computation methods could provide a generic method for inferring segment boundaries in sequence data."
                },
                {
                    "title": "Peter",
                    "abstract": "Seit einigen Jahren geistert ein Begriff durch die verkehrspolitische Diskussion und produziert vor allem Blockaden und Unverst\u00e4ndnis: Mobility Pricing. Interessanterweise erfolgen die Positionsbez\u00fcge \u00fcberraschend quer zu den sonst in der Verkehrspolitik bekannten Interessengruppen. Es ist dabei ein Verdienst des Bundes, mit dem im Jahre 2016 publizierten \u00abKonzeptbericht Mobility Pricing\u00bb eine begrifflich und konzeptionell breit angelegte Auslegeordnung eingebracht zu haben. Mit der Einbindung der verschiedenen Modellvarianten in einen sog. \u00abmorphologischen Kasten\u00bb gelingt dem Bericht auch eine weitgehend ideologiefreie Auseinandersetzung. Denn, was ist naheliegender, als die Nutzung von Verkehrsinfrastrukturen leistungsabh\u00e4ngig auszugestalten! Mit der, in der Handhabung verbl\u00fcffend einfach konzipierten Mineral\u00f6lsteuer, den zeitund ortsgebundenen Parkplatzgeb\u00fchren, aber auch mit dem Kilometertarif im \u00f6V oder der \u00abLeistungsabh\u00e4ngigen Schwerverkehrsabgabe\u00bb (LSVA) gibt es unbestrittene und bew\u00e4hrte Traditionen von leistungsabh\u00e4ngigen Nutzungsentgelten im Verkehrsbereich. 1 Grundlegende Schw\u00e4chen der Verkehrspolitik bleiben ausgeblendet Die Auseinandersetzung mit Mobility Pricing offenbart jedoch zunehmend die schwerwiegenden Defizite und Schw\u00e4chen bei den Vorgaben und den Zielsetzungen in der Entwicklung von Verkehrsinfrastrukturen, resp. bei deren Einbettung in ein Raumund Siedlungsentwicklungskonzept. Die Erfahrungen zeigen, dass der Ausbau der Verkehrsinfrastrukturen und -angebote zunehmend eine Siedlungsentwicklung und ein Mobilit\u00e4tsverhalten bestimmen, ohne dass \u00fcber deren Zielsetzungen eine transparente Diskussion mit einer Absch\u00e4tzung der Folgen, beispielsweise unter dem Aspekt der Nachhaltigkeit, gef\u00fchrt wird. Treiber des Mobility Pricings ist weniger eine bewusste Siedlungsentwicklung oder ein nachhaltigeres Mobilit\u00e4tsverhalten, als vielmehr die Aussicht, dass das massiv zunehmende Verkehrsvolumen zeitlich nur besser aufgefangen und verteilt werden muss. Die Verkehrsnachfrage als solche wird dabei nicht in Frage gestellt, Es geht schlicht nur um die bessere Bew\u00e4ltigung des Nachfrage\u00fcberhangs an Verkehrsinfrastrukturen. Alle neuen Erhebungen zeigen dabei sowohl eine rasante Zunahme des Verkehrs wie auch eine Konzentrierung auf das Angebot, welche durch den Ausbau der Infrastrukturen und durch die Beseitigung von \u00abHindernissen\u00bb zu einer Verfl\u00fcssigung des Verkehrs und damit zu einer Verk\u00fcrzung der Fahrzeiten f\u00fchren. 2 Nur weitere Verkehrszunahmen? Es ist offensichtlich, dass neben den Infrastrukturausbauten praktisch alle Massnahmen im heutigen Verkehrsmanagement dem Ziel dienen, die Verkehrszunahme aufzufangen. Dabei provozieren diese Massnahmen erst recht Verkehrszunahmen oder machen diese \u00fcberhaupt erst einigermassen ertr\u00e4glich, wenn zum Beispiel mit weniger Stauzeiten gerechnet werden muss. Damit wird lediglich an der Spirale \u00abmehr Verkehr \u2013 zus\u00e4tzliche und besser nutzbare Infrastrukturen\u00bb gedreht. Der Focus der Betrachtung liegt damit nicht bei den Ursachen sondern"
                },
                {
                    "title": "Neural Machine Translation of Rare Words with Subword Units",
                    "abstract": "Neural machine translation (NMT) models typically operate with a fixed vocabulary, but translation is an open-vocabulary problem. Previous work addresses the translation of out-of-vocabulary words by backing off to a dictionary. In this paper, we introduce a simpler and more effective approach, making the NMT model capable of open-vocabulary translation by encoding rare and unknown words as sequences of subword units. This is based on the intuition that various word classes are translatable via smaller units than words, for instance names (via character copying or transliteration), compounds (via compositional translation), and cognates and loanwords (via phonological and morphological transformations). We discuss the suitability of different word segmentation techniques, including simple character n-gram models and a segmentation based on the byte pair encoding compression algorithm, and empirically show that subword models improve over a back-off dictionary baseline for the WMT 15 translation tasks English-German and English-Russian by 1.1 and 1.3 BLEU, respectively."
                },
                {
                    "title": "On the difficulty of training recurrent neural networks",
                    "abstract": "There are two widely known issues with properly training recurrent neural networks, the vanishing and the exploding gradient problems detailed in Bengio et al. (1994). In this paper we attempt to improve the understanding of the underlying issues by exploring these problems from an analytical, a geometric and a dynamical systems perspective. Our analysis is used to justify a simple yet effective solution. We propose a gradient norm clipping strategy to deal with exploding gradients and a soft constraint for the vanishing gradients problem. We validate empirically our hypothesis and proposed solutions in the experimental section."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we design a byte-level autoregressive language model architecture that matches the performance of tokenized models while controlling for computational costs?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of natural language processing, as it addresses the limitations of tokenization, which can lead to performance disparities for underrepresented languages and increased vulnerability to adversarial attacks. A successful byte-level model could democratize access to high-performance language models across diverse languages and applications, potentially leading to breakthroughs in multilingual processing, improved character-level modeling, and enhanced robustness against adversarial inputs. This research could pave the way for future studies on efficient language modeling techniques and inspire new methodologies that prioritize inclusivity and performance.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of byte-level modeling, which requires managing a significantly larger input space compared to tokenized models. Naive approaches may fail due to the increased training and inference computational costs associated with byte-level representations, as evidenced by the finding that byte-level models require roughly 10 times more training FLOPs to achieve comparable performance. Additionally, the need for effective multiscale modeling and dynamic partitioning of bytes into patches adds layers of complexity that must be carefully addressed to optimize performance without sacrificing efficiency.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on tokenized models, leading to a lack of exploration into byte-level architectures that can compete with their performance. Existing solutions, such as MegaByte and MambaByte, have shown improvements but have not achieved parity with tokenized models under controlled compute conditions. Barriers include the fixed patch sizes used in prior models, which may not align well with language structures, and a lack of comprehensive studies that systematically compare byte-level and subword-level models under equal computational constraints. Our approach differs by introducing a dynamic patching rule that aligns with language boundaries, which has not been adequately explored in prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the SpaceByte architecture, which incorporates dynamic partitioning of bytes into patches aligned with language boundaries, alongside multiscale modeling. We will conduct experiments using datasets consisting of English books, LaTeX formatted arXiv papers, and open-source code, measuring performance in terms of cross-entropy (bits-per-byte) while"
            }
        },
        "author_data": {
            "25f1892a-d82f-4605-8aac-360f97014bb5": {
                "pk": "25f1892a-d82f-4605-8aac-360f97014bb5",
                "name": "Kevin Slagle",
                "collaborators": [
                    "Yong Baek Kim",
                    "Cenke Xu",
                    "Wilbur Shirley",
                    "Xie Chen",
                    "John Preskill",
                    "Liang Fu",
                    "Yi-Zhuang You",
                    "Abhinav Prem",
                    "Michael Pretko",
                    "David Aasen"
                ],
                "domain": [
                    "Quantum Mechanics",
                    "Fracton Order",
                    "Quantum Field Theory",
                    "Superconductivity"
                ],
                "publications": [
                    {
                        "title": "Unfrustrating the t-J Model: d-wave BCS Superconductivity in the $t'$-$J_z$-$V$ Model",
                        "abstract": "The t-J model is believed to be a minimal model that may be capable of describing the low-energy physics of the cuprate superconductors. However, although the t-J model is simple in appearance, obtaining a detailed understanding of its phase diagram has proved to be challenging. We are therefore motivated to study modifications to the t-J model such that its phase diagram and mechanism for d-wave superconductivity can be understood analytically without making uncontrolled approximations. The modified model we consider is a $t'$-$J_z$-$V$ model on a square lattice, which has a second-nearest-neighbor hopping $t'$ (instead of a nearest-neighbor hopping $t$), an Ising (instead of Heisenberg) antiferromagnetic coupling $J_z$, and a nearest-neighbor repulsion $V$. In a certain strongly interacting limit, the ground state is an antiferromagnetic superconductor that can be described exactly by a Hamiltonian where the only interaction is a nearest-neighbor attraction. BCS theory can then be applied with arbitrary analytical control, from which nodeless d-wave or s-wave superconductivity can result."
                    },
                    {
                        "title": "Testing Quantum Mechanics using Noisy Quantum Computers",
                        "abstract": "We outline a proposal to test quantum mechanics in the high-complexity regime using noisy intermediate-scale quantum (NISQ) devices. The procedure involves simulating a non-Clifford random circuit, followed by its inverse, and then checking that the resulting state is the same as the initial state. We are motivated by the hypothesis that quantum mechanics is not fundamental, but instead emerges from a theory with less computational power, such as classical mechanics. This emergent quantum mechanics (EmQM) hypothesis makes the prediction that quantum computers will not be capable of sufficiently complex quantum computations. We show that quantum mechanics predicts that the fidelity of our procedure decays exponentially with circuit depth (due to noise in NISQ devices), while EmQM predicts that the fidelity will decay significantly more rapidly for sufficiently deep circuits, which is the experimental signature that we propose to search for. We estimate rough bounds for when possible signals of EmQM should be expected. Furthermore, we find that highly informative experiments should require only thousands qubits."
                    },
                    {
                        "title": "Foliated Quantum Field Theory of Fracton Order",
                        "abstract": "We introduce a new kind of foliated quantum field theory (FQFT) of gapped fracton orders in the continuum. FQFT is defined on a manifold with a layered structure given by one or more foliations, which each decompose spacetime into a stack of layers. FQFT involves a new kind of gauge field, a foliated gauge field, which behaves similar to a collection of independent gauge fields on this stack of layers. Gauge invariant operators (and their analogous particle mobilities) are constrained to the intersection of one or more layers from different foliations. The level coefficients are quantized and exhibit a duality that spatially transforms the coefficients. This duality occurs because the FQFT is a foliated fracton order. That is, the duality can decouple 2+1D gauge theories from the FQFT through a process we dub exfoliation."
                    },
                    {
                        "title": "Fast Tensor Disentangling Algorithm",
                        "abstract": "Many recent tensor network algorithms apply unitary operators to parts of a tensor network in order to reduce entanglement. However, many of the previously used iterative algorithms to minimize entanglement can be slow. We introduce an approximate, fast, and simple algorithm to optimize disentangling unitary tensors. Our algorithm is asymptotically faster than previous iterative algorithms and often results in a residual entanglement entropy that is within 10 to 40% of the minimum. For certain input tensors, our algorithm returns an optimal solution. When disentangling order-4 tensors with equal bond dimensions, our algorithm achieves an entanglement spectrum where nearly half of the singular values are zero. We further validate our algorithm by showing that it can efficiently disentangle random 1D states of qubits."
                    },
                    {
                        "title": "Quantum Gauge Networks: A New Kind of Tensor Network",
                        "abstract": "Although tensor networks are powerful tools for simulating low-dimensional quantum physics, tensor network algorithms are very computationally costly in higher spatial dimensions. We introduce quantum gauge networks: a different kind of tensor network ansatz for which the computation cost of simulations does not explicitly increase for larger spatial dimensions. We take inspiration from the gauge picture of quantum dynamics, which consists of a local wavefunction for each patch of space, with neighboring patches related by unitary connections. A quantum gauge network (QGN) has a similar structure, except the Hilbert space dimensions of the local wavefunctions and connections are truncated. We describe how a QGN can be obtained from a generic wavefunction or matrix product state (MPS). All $2k$-point correlation functions of any wavefunction for $M$ many operators can be encoded exactly by a QGN with bond dimension $O(M^k)$. In comparison, for just $k=1$, an exponentially larger bond dimension of $2^{M/6}$ is generically required for an MPS of qubits. We provide a simple QGN algorithm for approximate simulations of quantum dynamics in any spatial dimension. The approximate dynamics can achieve exact energy conservation for time-independent Hamiltonians, and spatial symmetries can also be maintained exactly. We benchmark the algorithm by simulating the quantum quench of fermionic Hamiltonians in up to three spatial dimensions."
                    },
                    {
                        "title": "The Gauge Picture of Quantum Dynamics",
                        "abstract": "Although local Hamiltonians exhibit local time dynamics, this locality is not explicit in the Schr\\\"{o}dinger picture in the sense that the wavefunction amplitudes do not obey a local equation of motion. We show that geometric locality can be achieved explicitly in the equations of motion by \"gauging\" the global unitary invariance of quantum mechanics into a local gauge invariance. That is, expectation values $\\langle \\psi|A|\\psi \\rangle$ are invariant under a global unitary transformation acting on the wavefunction $|\\psi\\rangle \\to U |\\psi\\rangle$ and operators $A \\to U A U^\\dagger$, and we show that it is possible to gauge this global invariance into a local gauge invariance. To do this, we replace the wavefunction with a collection of local wavefunctions $|\\psi_J\\rangle$, one for each patch of space $J$. The collection of spatial patches is chosen to cover the space; e.g. we could choose the patches to be single qubits or nearest-neighbor sites on a lattice. Local wavefunctions associated with neighboring pairs of spatial patches $I$ and $J$ are related to each other by dynamical unitary transformations $U_{IJ}$. The local wavefunctions are local in the sense that their dynamics are local. That is, the equations of motion for the local wavefunctions $|\\psi_J\\rangle$ and connections $U_{IJ}$ are explicitly local in space and only depend on nearby Hamiltonian terms. (The local wavefunctions are many-body wavefunctions and have the same Hilbert space dimension as the usual wavefunction.) We call this picture of quantum dynamics the gauge picture since it exhibits a local gauge invariance. The local dynamics of a single spatial patch is related to the interaction picture, where the interaction Hamiltonian consists of only nearby Hamiltonian terms. We can also generalize the explicit locality to include locality in local charge and energy densities."
                    },
                    {
                        "title": "Emergent Quantum Mechanics at the Boundary of a Local Classical Lattice Model",
                        "abstract": "We formulate a conceptually new model in which quantum mechanics emerges from classical mechanics. Given a local Hamiltonian $H$ acting on $n$ qubits, we define a local classical model with an additional spatial dimension whose boundary dynamics is approximately -- but to arbitrary precision -- described by Schr\\\"{o}dinger's equation and $H$. The bulk consists of a lattice of classical bits that propagate towards the boundary through a circuit of stochastic matrices. The bits reaching the boundary are governed by a probability distribution whose deviation from the uniform distribution can be interpreted as the quantum-mechanical wavefunction. Bell nonlocality is achieved because information can move through the bulk much faster than the boundary speed of light. We analytically estimate how much the model deviates from quantum mechanics, and we validate these estimates using computer simulations."
                    },
                    {
                        "title": "Charge Transfer Excitations, Pair Density Waves, and Superconductivity in Moir\u00e9 Materials",
                        "abstract": "Transition metal dichalcogenide (TMD) bilayers are a new class of tunable moir\\'e systems attracting interest as quantum simulators of strongly-interacting electrons in two dimensions. In particular, recent theory predicts that the correlated insulator observed in WSe$_2$/WS$_2$ at half filling is a charge-transfer insulator similar to cuprates and, upon further hole doping, exhibits a transfer of charge from anion-like to cation-like orbitals at different locations in the moir\\'e unit cell. In this work, we demonstrate that in this doped charge-transfer insulator, tightly-bound charge-2e excitations can form to lower the total electrostatic repulsion. This composite excitation, which we dub a trimer, consists of a pair of holes bound to a charge-transfer exciton. When the bandwidth of doped holes is small, trimers crystallize into insulating pair density waves at a sequence of commensurate doping levels. When the bandwidth becomes comparable to the pair binding energy, itinerant holes and charge-2e trimers interact resonantly, leading to unconventional superconductivity similar to superfluidity in an ultracold Fermi gas near Feshbach resonance. Our theory is broadly applicable to strongly-interacting charge-transfer insulators, such as WSe$_2$/WS$_2$ or TMD homobilayers under an applied electric field."
                    },
                    {
                        "title": "Fracton Topological Order from Nearest-Neighbor Two-Spin Interactions and Dualities",
                        "abstract": "Fracton topological order describes a remarkable phase of matter which can be characterized by fracton excitations with constrained dynamics and a ground state degeneracy that increases exponentially with the length of the system on a three-dimensional torus. However, previous models exhibiting this order require many-spin interactions which may be very difficult to realize in a real material or cold atom system. In this work, we present a more physically realistic model which has the so-called X-cube fracton topological order but only requires nearest-neighbor two-spin interactions. The model lives on a three-dimensional honeycomb-based lattice with one to two spin-1/2 degrees of freedom on each site and a unit cell of 6 sites. The model is constructed from two orthogonal stacks of $Z_2$ topologically ordered Kitaev honeycomb layers, which are coupled together by a two-spin interaction. It is also shown that a four-spin interaction can be included to instead stabilize 3+1D $Z_2$ topological order. We also find dual descriptions of four quantum phase transitions in our model, all of which appear to be discontinuous first order transitions."
                    },
                    {
                        "title": "Disordered XYZ Spin Chain Simulations using the Spectrum Bifurcation Renormalization Group",
                        "abstract": "We study the disordered XYZ spin chain using the recently developed Spectrum Bifurcation Renormalization Group (SBRG) numerical method. With strong disorder, the phase diagram consists of three many body localized (MBL) spin glass phases. We argue that, with sufficiently strong disorder, these spin glass phases are separated by marginally many-body localized (MBL) critical lines. We examine the critical lines of this model by measuring the entanglement entropy and Edwards-Anderson spin glass order parameter, and find that the critical lines are characterized by an effective central charge c'=ln2. Our data also suggests continuously varying critical exponents along the critical lines. We also demonstrate how long-range mutual information can distinguish these phases."
                    },
                    {
                        "title": "A Simple Mechanism for Unconventional Superconductivity in a Repulsive Fermion Model",
                        "abstract": "Motivated by a scarcity of simple and analytically tractable models of superconductivity from strong repulsive interactions, we introduce a simple tight-binding lattice model of fermions with repulsive interactions that exhibits unconventional superconductivity (beyond BCS theory). The model resembles an idealized conductor-dielectric-conductor trilayer. The Cooper pair consists of electrons on opposite sides of the dielectric, which mediates the attraction. In the strong coupling limit, we use degenerate perturbation theory to show that the model reduces to a superconducting hard-core Bose-Hubbard model. Above the superconducting critical temperature, an analog of pseudo-gap physics results where the fermions remain Cooper paired with a large single-particle energy gap."
                    },
                    {
                        "title": "Symmetric Tensor Gauge Theories on Curved Spaces",
                        "abstract": "Fractons and other subdimensional particles are an exotic class of emergent quasi-particle excitations with severely restricted mobility. A wide class of models featuring these quasi-particles have a natural description in the language of symmetric tensor gauge theories, which feature conservation laws restricting the motion of particles to lower-dimensional sub-spaces, such as lines or points. In this work, we investigate the fate of symmetric tensor gauge theories in the presence of spatial curvature. We find that weak curvature can induce small (exponentially suppressed) violations on the mobility restrictions of charges, leaving a sense of asymptotic fractonic/sub-dimensional behavior on generic manifolds. Nevertheless, we show that certain symmetric tensor gauge theories maintain sharp mobility restrictions and gauge invariance on certain special curved spaces, such as Einstein manifolds or spaces of constant curvature."
                    },
                    {
                        "title": "Foliated Field Theory and String-Membrane-Net Condensation Picture of Fracton Order",
                        "abstract": "Foliated fracton order is a qualitatively new kind of phase of matter. It is similar to topological order, but with the fundamental difference that a layered structure, referred to as a foliation, plays an essential role and determines the mobility restrictions of the topological excitations. In this work, we introduce a new kind of field theory to describe these phases: a foliated field theory. We also introduce a new lattice model and string-membrane-net condensation picture of these phases, which is analogous to the string-net condensation picture of topological order."
                    },
                    {
                        "title": "Comments on Foliated Gauge Theories and Dualities in 3+1d",
                        "abstract": "We investigate the properties of foliated gauge fields and construct several foliated field theories in 3+1d that describe foliated fracton orders both with and without matter, including the recent hybrid fracton models. These field theories describe Abelian or non-Abelian gauge theories coupled to foliated gauge fields, and they fall into two classes of models that we call the electric models and the magnetic models. We show that these two classes of foliated field theories enjoy a duality. We also construct a model (using foliated gauge fields and an exactly solvable lattice Hamiltonian model) for a subsystem-symmetry protected topological (SSPT) phase, which is analogous to a one-form symmetry protected topological phase, with the subsystem symmetry acting on codimension-two subregions. We construct the corresponding gauged SSPT phase as a foliated two-form gauge theory. Some instances of the gauged SSPT phase are a variant of the X-cube model with the same ground state degeneracy and the same fusion, but different particle statistics."
                    },
                    {
                        "title": "Quantum Field Theory of X-Cube Fracton Topological Order and Robust Degeneracy from Geometry",
                        "abstract": "We propose a quantum field theory description of the X-cube model of fracton topological order. The field theory is not (and cannot be) a topological quantum field theory (TQFT), since unlike the X-cube model, TQFTs are invariant (i.e. symmetric) under continuous spacetime transformations. However, the theory is instead invariant under a certain subgroup of the conformal group. We describe how braiding statistics and ground state degeneracy are reproduced by the field theory, and how the the X-cube Hamiltonian and field theory can be minimally coupled to matter fields. We also show that even on a manifold with trivial topology, spatial curvature can induce a ground state degeneracy that is stable to arbitrary local perturbations! Our formalism may allow for the description of other fracton field theories, where the only necessary input is an equation of motion for a charge density."
                    },
                    {
                        "title": "Screw dislocations in the X-cube fracton model",
                        "abstract": "The X-cube model, a prototypical gapped fracton model, has been shown to have a foliation structure. That is, inside the 3+1D model, there are hidden layers of 2+1D gapped topological states. A screw dislocation in a 3+1D lattice can often reveal nontrivial features associated with a layered structure. In this paper, we study the X-cube model on lattices with screw dislocations. In particular, we find that a screw dislocation results in a finite change in the logarithm of the ground state degeneracy of the model. Part of the change can be traced back to the effect of screw dislocations in a simple stack of 2+1D topological states, hence corroborating the foliation structure in the model. The other part of the change comes from the induced motion of fractons or sub-dimensional excitations along the dislocation, a feature absent in the stack of 2+1D layers."
                    },
                    {
                        "title": "Quantum Phase Transition between Z2 spin liquid and columnar Valence Bond Crystals on a Triangular lattice",
                        "abstract": "We study the quantum phase transition between the $Z_2$ spin liquid and valence bond solid (VBS) orders on a triangular lattice. With a fully isotropic triangular lattice, the transition from a columnar or resonating-plaquette VBS order can be either first order or there could be two transitions with an intermediate phase. If the transition splits into two, then the $Z_2$ spin liquid will first experience a first order $q=3$ Potts transition to a new nematic $Z_2$ spin liquid that breaks the $2\\pi/3$ lattice rotation symmetry (but retain translation symmetry unlike the VBS states). The second transition will then take this new nematic $Z_2$ spin liquid to a columnar or resonating-plaquette VBS state through a second order $3d$ XY$^*$ transition. On a distorted triangular lattice, the degeneracy between some of the different columnar VBS orders is lifted, and the phase transition can reduce to a single $3d$ XY$^*$ transition."
                    },
                    {
                        "title": "Foliated fracton order in the checkerboard model",
                        "abstract": "In this work, we show that the checkerboard model exhibits the phenomenon of foliated fracton order. We introduce a renormalization group transformation for the model that utilizes toric code bilayers as an entanglement resource, and show how to extend the model to general three-dimensional manifolds. Furthermore, we use universal properties distilled from the structure of fractional excitations and ground-state entanglement to characterize the foliated fracton phase and find that it is the same as two copies of the X-cube model. Indeed, we demonstrate that the checkerboard model can be transformed into two copies of the X-cube model via an adiabatic deformation."
                    },
                    {
                        "title": "X-Cube Fracton Model on Generic Lattices: Phases and Geometric Order",
                        "abstract": "Fracton order is a new kind of quantum order characterized by topological excitations that exhibit remarkable mobility restrictions and a robust ground state degeneracy (GSD) which can increase exponentially with system size. In this paper, we present a generic lattice construction (in three dimensions) for a generalized X-cube model of fracton order, where the mobility restrictions of the subdimensional particles inherit the geometry of the lattice. This helps explain a previous result that lattice curvature can produce a robust GSD, even on a manifold with trivial topology. We provide explicit examples to show that the (zero temperature) phase of matter is sensitive to the lattice geometry. In one example, the lattice geometry confines the dimension-1 particles to small loops, which allows the fractons to be fully mobile charges, and the resulting phase is equivalent to (3+1)-dimensional toric code. However, the phase is sensitive to more than just lattice curvature; different lattices without curvature (e.g. cubic or stacked kagome lattices) also result in different phases of matter, which are separated by phase transitions. Unintuitively however, according to a previous definition of phase [Chen, Gu, Wen 2010], even just a rotated or rescaled cubic lattice results in different phases of matter, which motivates us to propose a new and coarser definition of phase for gapped ground states and fracton order. The new equivalence relation between ground states is given by the composition of a local unitary transformation and a quasi-isometry (which can rotate and rescale the lattice); equivalently, ground states are in the same phase if they can be adiabatically connected by varying both the Hamiltonian and the positions of the degrees of freedom (via a quasi-isometry). In light of the importance of geometry, we further propose that fracton orders should be regarded as a geometric order."
                    }
                ]
            }
        }
    },
    "2406.01234": {
        "paper_data": {
            "title": "Achieving Tractable Minimax Optimal Regret in Average Reward MDPs",
            "url": "http://arxiv.org/abs/2406.01234v1",
            "arxiv_id": "2406.01234",
            "authors": [
                "Victor Boone",
                "Zihan Zhang"
            ],
            "abstract": "In recent years, significant attention has been directed towards learning average-reward Markov Decision Processes (MDPs). However, existing algorithms either suffer from sub-optimal regret guarantees or computational inefficiencies. In this paper, we present the first tractable algorithm with minimax optimal regret of $\\widetilde{\\mathrm{O}}(\\sqrt{\\mathrm{sp}(h^*) S A T})$, where $\\mathrm{sp}(h^*)$ is the span of the optimal bias function $h^*$, $S \\times A$ is the size of the state-action space and $T$ the number of learning steps. Remarkably, our algorithm does not require prior information on $\\mathrm{sp}(h^*)$. Our algorithm relies on a novel subroutine, Projected Mitigated Extended Value Iteration (PMEVI), to compute bias-constrained optimal policies efficiently. This subroutine can be applied to various previous algorithms to improve regret bounds.",
            "introduction": "   1 Introduction  Reinforcement learning (RL) Burnetas and Katehakis (1997); Sutton and Barto (2018) has become a popular approach for solving complex sequential decision-making tasks and has recently achieved notable advancements in diverse fields of application. The RL problem is generally formulated as a Markov Decision Process (MDP) Puterman (1994), where the agent interacts with an unknown environment to maximize its accumulative rewards.   In this paper, we consider the problem of learning average-reward MDPs, where the central task is to balance between exploration (i.e., learning the unknown environment) and exploitation (i.e., planning optimally according to current knowledge) along the infinite-horizon learning process. One way to measure the performance of the learner is the regret, that compares the gathered rewards of the learner, unaware of the exact structure of its environment, to the expected performance of an omniscient agent that knows the environment in advance. The seminal work of Auer et\u00a0al. (2009) provides a minimax regret lower bound \u03a9\u2062(D\u2062S\u2062A\u2062T)\u03a9\ud835\udc37\ud835\udc46\ud835\udc34\ud835\udc47\\Omega\\left(\\!\\!\\sqrt{DSAT}\\right)roman_\u03a9 ( square-root start_ARG italic_D italic_S italic_A italic_T end_ARG ), where D\ud835\udc37Ditalic_D is the diameter (the maximal distance between two different states), S\ud835\udc46Sitalic_S the number of states, A\ud835\udc34Aitalic_A the number of actions and T\ud835\udc47Titalic_T the learning horizon. They also provide an algorithm achieving regret O~\u2062(D2\u2062S2\u2062A\u2062T)~Osuperscript\ud835\udc372superscript\ud835\udc462\ud835\udc34\ud835\udc47\\widetilde{\\operatorname*{{\\rm O}}}\\left(\\!\\!\\sqrt{D^{2}S^{2}AT}\\right)over~ start_ARG roman_O end_ARG ( square-root start_ARG italic_D start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_S start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_A italic_T end_ARG ). Ever since Auer et\u00a0al. (2009), many works have been devoted to close the gap between the regret lower and upper bounds in the average reward setting Auer et\u00a0al. (2009); Bartlett and Tewari (2009); Filippi et\u00a0al. (2010); Talebi and Maillard (2018); Fruit et\u00a0al. (2018, 2020); Bourel et\u00a0al. (2020); Zhang and Ji (2019); Ouyang et\u00a0al. (2017); Agrawal and Jia (2023); Abbasi-Yadkori et\u00a0al. (2019); Wei et\u00a0al. (2020) and more. Subsequent works Fruit et\u00a0al. (2018); Zhang and Ji (2019) refined the minimax regret lower bound to \u03a9\u2062(sp\u2062(h\u2217)\u2062S\u2062A\u2062T)\u03a9spsuperscript\u210e\ud835\udc46\ud835\udc34\ud835\udc47\\Omega\\left(\\!\\!\\sqrt{{\\mathrm{sp}\\left(h^{*}\\right)}SAT}\\right)roman_\u03a9 ( square-root start_ARG roman_sp ( italic_h start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT ) italic_S italic_A italic_T end_ARG ) where sp\u2062(h\u2217)spsuperscript\u210e{\\mathrm{sp}\\left(h^{*}\\right)}roman_sp ( italic_h start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT ) is the span of the bias function, which is the maximal gap of the long-term accumulative rewards starting from two different states. The difference is significant, since sp\u2062(h\u2217)\u2264Dspsuperscript\u210e\ud835\udc37{\\mathrm{sp}\\left(h^{*}\\right)}\\leq Droman_sp ( italic_h start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT ) \u2264 italic_D and the gap between the two can be arbitrarly large. However, no existing work achieves the following three requirements simultaneously:   (1)  The method achieves minimax optimal regret guarantees O~\u2062(sp\u2062(h\u2217)\u2062S\u2062A\u2062T)~Ospsuperscript\u210e\ud835\udc46\ud835\udc34\ud835\udc47\\widetilde{\\operatorname*{{\\rm O}}}\\left(\\!\\!\\sqrt{{\\mathrm{sp}\\left(h^{*}% \\right)}SAT}\\right)over~ start_ARG roman_O end_ARG ( square-root start_ARG roman_sp ( italic_h start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT ) italic_S italic_A italic_T end_ARG );    (2)  The proposed method is tractable;    (3)  No prior knowledge on the model is required.    Most algorithms simply fail to achieve minimax optimal regret, and the only method achieving it Zhang and Ji (2019) is intractable because it relies an oracle to solve difficult optimization problems along the learning process. Naturally, we raise the question of whether these three requirements can be met all at once: Is there a tractable algorithm with O~\u2062(sp\u2062(h\u2217)\u2062S\u2062A\u2062T)~OspsuperscripthSAT\\widetilde{\\operatorname*{{\\rm O}}}\\left(\\!\\!\\sqrt{{\\mathrm{sp}\\left(h^{*}% \\right)}SAT}\\right)over~ start_ARG roman_O end_ARG ( square-root start_ARG roman_sp ( italic_h start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT ) italic_S italic_A italic_T end_ARG ) minimax regret without prior knowledge?   Contributions.  In this paper, we answer the above question affirmatively, by proposing a polynomial time algorithm with regret guarantees O~\u2062(sp\u2062(h\u2217)\u2062S\u2062A\u2062T)~Ospsuperscript\u210e\ud835\udc46\ud835\udc34\ud835\udc47\\widetilde{\\operatorname*{{\\rm O}}}\\left(\\!\\!\\sqrt{{\\mathrm{sp}\\left(h^{*}% \\right)}SAT}\\right)over~ start_ARG roman_O end_ARG",
            "references": [
                {
                    "title": "Sharper Model-free Reinforcement Learning for Average-reward Markov Decision Processes",
                    "abstract": "We develop several provably efficient model-free reinforcement learning (RL) algorithms for infinite-horizon average-reward Markov Decision Processes (MDPs). We consider both online setting and the setting with access to a simulator. In the online setting, we propose model-free RL algorithms based on reference-advantage decomposition. Our algorithm achieves $\\widetilde{O}(S^5A^2\\mathrm{sp}(h^*)\\sqrt{T})$ regret after $T$ steps, where $S\\times A$ is the size of state-action space, and $\\mathrm{sp}(h^*)$ the span of the optimal bias function. Our results are the first to achieve optimal dependence in $T$ for weakly communicating MDPs. In the simulator setting, we propose a model-free RL algorithm that finds an $\\epsilon$-optimal policy using $\\widetilde{O} \\left(\\frac{SA\\mathrm{sp}^2(h^*)}{\\epsilon^2}+\\frac{S^2A\\mathrm{sp}(h^*)}{\\epsilon} \\right)$ samples, whereas the minimax lower bound is $\\Omega\\left(\\frac{SA\\mathrm{sp}(h^*)}{\\epsilon^2}\\right)$. Our results are based on two new techniques that are unique in the average-reward setting: 1) better discounted approximation by value-difference estimation; 2) efficient construction of confidence region for the optimal bias function with space complexity $O(SA)$."
                },
                {
                    "title": "Optimistic posterior sampling for reinforcement learning: worst-case regret bounds",
                    "abstract": "We present an algorithm based on posterior sampling (aka Thompson sampling) that achieves near-optimal worst-case regret bounds when the underlying Markov decision process (MDP) is communicating with a finite, although unknown, diameter. Our main result is a high probability regret upper bound of [Formula: see text] for any communicating MDP with S states, A actions, and diameter D. Here, regret compares the total reward achieved by the algorithm to the total expected reward of an optimal infinite-horizon undiscounted average reward policy in time horizon T. This result closely matches the known lower bound of [Formula: see text]. Our techniques involve proving some novel results about the anti-concentration of Dirichlet distribution, which may be of independent interest."
                },
                {
                    "title": "Improved Analysis of UCRL2 with Empirical Bernstein Inequality",
                    "abstract": "We consider the problem of exploration-exploitation in communicating Markov Decision Processes. We provide an analysis of UCRL2 with Empirical Bernstein inequalities (UCRL2B). For any MDP with $S$ states, $A$ actions, $\\Gamma \\leq S$ next states and diameter $D$, the regret of UCRL2B is bounded as $\\widetilde{O}(\\sqrt{D\\Gamma S A T})$."
                },
                {
                    "title": "Planning in Markov Decision Processes with Gap-Dependent Sample Complexity",
                    "abstract": "We propose MDP-GapE, a new trajectory-based Monte-Carlo Tree Search algorithm for planning in a Markov Decision Process in which transitions have a finite support. We prove an upper bound on the number of calls to the generative models needed for MDP-GapE to identify a near-optimal action with high probability. This problem-dependent sample complexity result is expressed in terms of the sub-optimality gaps of the state-action pairs that are visited during exploration. Our experiments reveal that MDP-GapE is also effective in practice, in contrast with other algorithms with sample complexity guarantees in the fixed-confidence setting, that are mostly theoretical."
                },
                {
                    "title": "Almost Optimal Model-Free Reinforcement Learning via Reference-Advantage Decomposition",
                    "abstract": "We study the reinforcement learning problem in the setting of finite-horizon episodic Markov Decision Processes (MDPs) with $S$ states, $A$ actions, and episode length $H$. We propose a model-free algorithm UCB-Advantage and prove that it achieves $\\tilde{O}(\\sqrt{H^2SAT})$ regret where $T = KH$ and $K$ is the number of episodes to play. Our regret bound improves upon the results of [Jin et al., 2018] and matches the best known model-based algorithms as well as the information theoretic lower bound up to logarithmic factors. We also show that UCB-Advantage achieves low local switching cost and applies to concurrent reinforcement learning, improving upon the recent results of [Bai et al., 2019]."
                },
                {
                    "title": "Tightening Exploration in Upper Confidence Reinforcement Learning",
                    "abstract": "The upper confidence reinforcement learning (UCRL2) strategy introduced in (Jaksch et al., 2010) is a popular method to perform regret minimization in unknown discrete Markov Decision Processes under the average-reward criterion. Despite its nice and generic theoretical regret guarantees, this strategy and its variants have remained until now mostly theoretical as numerical experiments on simple environments exhibit long burn-in phases before the learning takes place. Motivated by practical efficiency, we present UCRL3, following the lines of UCRL2, but with two key modifications: First, it uses state-of-the-art time-uniform concentration inequalities, to compute confidence sets on the reward and transition distributions for each state-action pair. To further tighten exploration, we introduce an adaptive computation of the support of each transition distributions. This enables to revisit the extended value iteration procedure to optimize over distributions with reduced support by disregarding low probability transitions, while still ensuring near-optimism. We demonstrate, through numerical experiments on standard environments, that reducing exploration this way yields a substantial numerical improvement compared to UCRL2 and its variants. On the theoretical side, these key modifications enable to derive a regret bound for UCRL3 improving on UCRL2, that for the first time makes appear a notion of local diameter and effective support, thanks to variance-aware concentration bounds."
                },
                {
                    "title": "Exploration-exploitation dilemma in Reinforcement Learning under various form of prior knowledge. (Impact des connaissances a priori sur le compromis exploration-exploitation en apprentissage par renforcement)",
                    "abstract": "In combination with Deep Neural Networks (DNNs), several Reinforcement Learning (RL) algorithms such as \"Q-learning\" or \"Policy Gradient\" are now able to achieve super-human performances on most Atari Games as well as the game of Go. Despite these outstanding and promising achievements, such Deep Reinforcement Learning (DRL) algorithms require millions of samples to perform well, thus limiting their deployment to all applications where data acquisition is costly. \nThe lack of sample efficiency of DRL can partly be attributed to the use of DNNs, which are known to be data-intensive in the training phase. But more importantly, it can be attributed to the type of Reinforcement Learning algorithm used, which only perform a very inefficient undirected exploration of the environment. For instance, Q-learning and Policy Gradient rely on randomization for exploration. In most cases, this strategy turns out to be very ineffective to properly balance the exploration needed to discover unknown and potentially highly rewarding regions of the environment, with the exploitation of rewarding regions already identified as such. \nOther RL approaches with theoretical guarantees on the exploration-exploitation trade-off have been investigated. It is sometimes possible to formally prove that the performances almost match the theoretical optimum. This line of research is inspired by the Multi-Armed Bandit literature, with many algorithms relying on the same underlying principle often referred as \"optimism in the face of uncertainty\". Even if a significant effort has been made towards understanding the exploration-exploitation dilemma generally, many questions still remain open. \nIn this thesis, we generalize existing work on exploration-exploitation to different contexts with different amounts of prior knowledge on the learning problem. We introduce several algorithmic improvements to current state-of-the-art approaches and derive a new theoretical analysis which allows us to answer several open questions of the literature. We then relax the (very common although not very realistic) assumption that a path between any two distinct regions of the environment should always exist. Relaxing this assumption highlights the impact of prior knowledge on the intrinsic limitations of the exploration-exploitation dilemma. Finally, we show how some prior knowledge such as the range of the value function or a set of macro-actions can be efficiently exploited to speed-up learning. \nIn this thesis, we always strive to take the algorithmic complexity of the proposed algorithms into account. Although all these algorithms are somehow computationally \"efficient\", they all require a planning phase and therefore suffer from the well-known \"curse of dimensionality\" which limits their applicability to real-world problems. Nevertheless, the main focus of this work is to derive general principles that may be combined with more heuristic approaches to help overcome current DRL flaws."
                },
                {
                    "title": "Model-free Reinforcement Learning in Infinite-horizon Average-reward Markov Decision Processes",
                    "abstract": "Model-free reinforcement learning is known to be memory and computation efficient and more amendable to large scale problems. In this paper, two model-free algorithms are introduced for learning infinite-horizon average-reward Markov Decision Processes (MDPs). The first algorithm reduces the problem to the discounted-reward version and achieves $\\mathcal{O}(T^{2/3})$ regret after $T$ steps, under the minimal assumption of weakly communicating MDPs. To our knowledge, this is the first model-free algorithm for general MDPs in this setting. The second algorithm makes use of recent advances in adaptive algorithms for adversarial multi-armed bandits and improves the regret to $\\mathcal{O}(\\sqrt{T})$, albeit with a stronger ergodic assumption. This result significantly improves over the $\\mathcal{O}(T^{3/4})$ regret achieved by the only existing model-free algorithm by Abbasi-Yadkori et al. (2019a) for ergodic MDPs in the infinite-horizon average-reward setting."
                },
                {
                    "title": "Regret Minimization for Reinforcement Learning by Evaluating the Optimal Bias Function",
                    "abstract": "We present an algorithm based on the \\emph{Optimism in the Face of Uncertainty} (OFU) principle which is able to learn Reinforcement Learning (RL) modeled by Markov decision process (MDP) with finite state-action space efficiently. By evaluating the state-pair difference of the optimal bias function $h^{*}$, the proposed algorithm achieves a regret bound of $\\tilde{O}(\\sqrt{SAHT})$\\footnote{The symbol $\\tilde{O}$ means $O$ with log factors ignored. } for MDP with $S$ states and $A$ actions, in the case that an upper bound $H$ on the span of $h^{*}$, i.e., $sp(h^{*})$ is known. This result outperforms the best previous regret bounds $\\tilde{O}(S\\sqrt{AHT}) $\\citep{fruit2019improved} by a factor of $\\sqrt{S}$. Furthermore, this regret bound matches the lower bound of $\\Omega(\\sqrt{SAHT}) $\\citep{jaksch2010near} up to a logarithmic factor. As a consequence, we show that there is a near optimal regret bound of $\\tilde{O}(\\sqrt{SADT})$ for MDPs with a finite diameter $D$ compared to the lower bound of $\\Omega(\\sqrt{SADT}) $\\citep{jaksch2010near}."
                },
                {
                    "title": "Solving linear programs in the current matrix multiplication time",
                    "abstract": "This paper shows how to solve linear programs of the form minAx=b,x\u22650 c\u22a4x with n variables in time O*((n\u03c9+n2.5\u2212\u03b1/2+n2+1/6) log(n/\u03b4)) where \u03c9 is the exponent of matrix multiplication, \u03b1 is the dual exponent of matrix multiplication, and \u03b4 is the relative accuracy. For the current value of \u03c9\u223c2.37 and \u03b1\u223c0.31, our algorithm takes O*(n\u03c9 log(n/\u03b4)) time. When \u03c9 = 2, our algorithm takes O*(n2+1/6 log(n/\u03b4)) time. Our algorithm utilizes several new concepts that we believe may be of independent interest: (1) We define a stochastic central path method. (2) We show how to maintain a projection matrix \u221aW A\u22a4(AWA\u22a4)\u22121A \u221aW in sub-quadratic time under \u21132 multiplicative changes in the diagonal matrix W."
                },
                {
                    "title": "Variance-Aware Regret Bounds for Undiscounted Reinforcement Learning in MDPs",
                    "abstract": "The problem of reinforcement learning in an unknown and discrete Markov Decision Process (MDP) under the average-reward criterion is considered, when the learner interacts with the system in a single stream of observations, starting from an initial state without any reset. We revisit the minimax lower bound for that problem by making appear the local variance of the bias function in place of the diameter of the MDP. Furthermore, we provide a novel analysis of the KL-UCRL algorithm establishing a high-probability regret bound scaling as $\\widetilde {\\mathcal O}\\Bigl({\\textstyle \\sqrt{S\\sum_{s,a}{\\bf V}^\\star_{s,a}T}}\\Big)$ for this algorithm for ergodic MDPs, where $S$ denotes the number of states and where ${\\bf V}^\\star_{s,a}$ is the variance of the bias function with respect to the next-state distribution following action $a$ in state $s$. The resulting bound improves upon the best previously known regret bound $\\widetilde {\\mathcal O}(DS\\sqrt{AT})$ for that algorithm, where $A$ and $D$ respectively denote the maximum number of actions (per state) and the diameter of MDP. We finally compare the leading terms of the two bounds in some benchmark MDPs indicating that the derived bound can provide an order of magnitude improvement in some cases. Our analysis leverages novel variations of the transportation lemma combined with Kullback-Leibler concentration inequalities, that we believe to be of independent interest."
                },
                {
                    "title": "Efficient Bias-Span-Constrained Exploration-Exploitation in Reinforcement Learning",
                    "abstract": "We introduce SCAL, an algorithm designed to perform efficient exploration-exploitation in any unknown weakly-communicating Markov decision process (MDP) for which an upper bound $c$ on the span of the optimal bias function is known. For an MDP with $S$ states, $A$ actions and $\\Gamma \\leq S$ possible next states, we prove a regret bound of $\\widetilde{O}(c\\sqrt{\\Gamma SAT})$, which significantly improves over existing algorithms (e.g., UCRL and PSRL), whose regret scales linearly with the MDP diameter $D$. In fact, the optimal bias span is finite and often much smaller than $D$ (e.g., $D=\\infty$ in non-communicating MDPs). A similar result was originally derived by Bartlett and Tewari (2009) for REGAL.C, for which no tractable algorithm is available. In this paper, we relax the optimization problem at the core of REGAL.C, we carefully analyze its properties, and we provide the first computationally efficient algorithm to solve it. Finally, we report numerical simulations supporting our theoretical findings and showing how SCAL significantly outperforms UCRL in MDPs with large diameter and small span."
                },
                {
                    "title": "Posterior Sampling for Large Scale Reinforcement Learning",
                    "abstract": "Posterior sampling for reinforcement learning (PSRL) is a popular algorithm for learning to control an unknown Markov decision process (MDP). PSRL maintains a distribution over MDP parameters and in an episodic fashion samples MDP parameters, computes the optimal policy for them and executes it. A special case of PSRL is where at the end of each episode the MDP resets to the initial state distribution. Extensions of this idea to general MDPs without state resetting has so far produced non-practical algorithms and in some cases buggy theoretical analysis. This is due to the difficulty of analyzing regret under episode switching schedules that depend on random variables of the true underlying model. We propose a solution to this problem that involves using a deterministic, model-independent episode switching schedule, and establish a Bayes regret bound under mild assumptions. Our algorithm termed deterministic schedule PSRL (DS-PSRL) is efficient in terms of time, sample, and space complexity. Our result is more generally applicable to continuous state action problems. We demonstrate how this algorithm is well suited for sequential recommendation problems such as points of interest (POI). We derive a general procedure for parameterizing the underlying MDPs, to create action condition dynamics from passive data, that do not contain actions. We prove that such parameterization satisfies the assumptions of our analysis."
                },
                {
                    "title": "Learning Unknown Markov Decision Processes: A Thompson Sampling Approach",
                    "abstract": "We consider the problem of learning an unknown Markov Decision Process (MDP) that is weakly communicating in the infinite horizon setting. We propose a Thompson Sampling-based reinforcement learning algorithm with dynamic episodes (TSDE). At the beginning of each episode, the algorithm generates a sample from the posterior distribution over the unknown model parameters. It then follows the optimal stationary policy for the sampled model for the rest of the episode. The duration of each episode is dynamically determined by two stopping criteria. The first stopping criterion controls the growth rate of episode length. The second stopping criterion happens when the number of visits to any state-action pair is doubled. We establish $\\tilde O(HS\\sqrt{AT})$ bounds on expected regret under a Bayesian setting, where $S$ and $A$ are the sizes of the state and action spaces, $T$ is time, and $H$ is the bound of the span. This regret bound matches the best available bound for weakly communicating MDPs. Numerical results show it to perform better than existing algorithms for infinite horizon MDPs."
                },
                {
                    "title": "Optimism in reinforcement learning and Kullback-Leibler divergence",
                    "abstract": "We consider model-based reinforcement learning in finite Markov Decision Processes (MDPs), focussing on so-called optimistic strategies. In MDPs, optimism can be implemented by carrying out extended value iterations under a constraint of consistency with the estimated model transition probabilities. The UCRL2 algorithm by Auer, Jaksch and Ortner (2009), which follows this strategy, has recently been shown to guarantee near-optimal regret bounds. In this paper, we strongly argue in favor of using the Kullback-Leibler (KL) divergence for this purpose. By studying the linear maximization problem under KL constraints, we provide an efficient algorithm, termed KL-UCRL, for solving KL-optimistic extended value iteration. Using recent deviation bounds on the KL divergence, we prove that KL-UCRL provides the same guarantees as UCRL2 in terms of regret. However, numerical experiments on classical benchmarks show a significantly improved behavior, particularly when the MDP has reduced connectivity. To support this observation, we provide elements of comparison between the two algorithms based on geometric considerations."
                },
                {
                    "title": "REGAL: A Regularization based Algorithm for Reinforcement Learning in Weakly Communicating MDPs",
                    "abstract": "We provide an algorithm that achieves the optimal regret rate in an unknown weakly communicating Markov Decision Process (MDP). The algorithm proceeds in episodes where, in each episode, it picks a policy using regularization based on the span of the optimal bias vector. For an MDP with S states and A actions whose optimal bias vector has span bounded by H, we show a regret bound of O(HS\u221aAT). We also relate the span to various diameter-like quantities associated with the MDP, demonstrating how our results improve on previous regret bounds."
                },
                {
                    "title": "Near-optimal Regret Bounds for Reinforcement Learning",
                    "abstract": "For undiscounted reinforcement learning in Markov decision processes (MDPs) we consider the total regret of a learning algorithm with respect to an optimal policy. In order to describe the transition structure of an MDP we propose a new parameter: An MDP has diameter D if for any pair of states s,s' there is a policy which moves from s to s' in at most D steps (on average). We present a reinforcement learning algorithm with total regret O(DS\u221aAT) after T steps for any unknown MDP with S states, A actions per state, and diameter D. A corresponding lower bound of \u03a9(\u221aDSAT) on the total regret of any learning algorithm is given as well. \n \nThese results are complemented by a sample complexity bound on the number of suboptimal steps taken by our algorithm. This bound can be used to achieve a (gap-dependent) regret bound that is logarithmic in T. \n \nFinally, we also consider a setting where the MDP is allowed to change a fixed number of l times. We present a modification of our algorithm that is able to deal with this setting and show a regret bound of O(l1/3T2/3DS\u221aA)."
                },
                {
                    "title": "Optimistic Linear Programming gives Logarithmic Regret for Irreducible MDPs",
                    "abstract": "We present an algorithm called Optimistic Linear Programming (OLP) for learning to optimize average reward in an irreducible but otherwise unknown Markov decision process (MDP). OLP uses its experience so far to estimate the MDP. It chooses actions by optimistically maximizing estimated future rewards over a set of next-state transition probabilities that are close to the estimates, a computation that corresponds to solving linear programs. We show that the total expected reward obtained by OLP up to time T is within C(P) log T of the reward obtained by the optimal policy, where C(P) is an explicit, MDP-dependent constant. OLP is closely related to an algorithm proposed by Burnetas and Katehakis with four key differences: OLP is simpler, it does not require knowledge of the supports of transition probabilities, the proof of the regret bound is simpler, but our regret bound is a constant factor larger than the regret of their algorithm. OLP is also similar in flavor to an algorithm recently proposed by Auer and Ortner. But OLP is simpler and its regret bound has a better dependence on the size of the MDP."
                },
                {
                    "title": "Logarithmic Online Regret Bounds for Undiscounted Reinforcement Learning",
                    "abstract": "We present a learning algorithm for undiscounted reinforcement learning. Our interest lies in bounds for the algorithm's online performance after some finite number of steps. In the spirit of similar methods already successfully applied for the exploration-exploitation tradeoff in multi-armed bandit problems, we use upper confidence bounds to show that our UCRL algorithm achieves logarithmic online regret in the number of steps taken with respect to an optimal policy."
                },
                {
                    "title": "Optimal Adaptive Policies for Markov Decision Processes",
                    "abstract": "In this paper we consider the problem of adaptive control for Markov Decision Processes. We give the explicit form for a class of adaptive policies that possess optimal increase rate properties for the total expected finite horizon reward, under sufficient assumptions of finite state-action spaces and irreducibility of the transition law. A main feature of the proposed policies is that the choice of actions, at each state and time period, is based on indices that are inflations of the right-hand side of the estimated average reward optimality equations."
                },
                {
                    "title": "Markov Decision Processes: Discrete Stochastic Dynamic Programming",
                    "abstract": "From the Publisher: \nThe past decade has seen considerable theoretical and applied research on Markov decision processes, as well as the growing use of these models in ecology, economics, communications engineering, and other fields where outcomes are uncertain and sequential decision-making processes are needed. A timely response to this increased activity, Martin L. Puterman's new work provides a uniquely up-to-date, unified, and rigorous treatment of the theoretical, computational, and applied research on Markov decision process models. It discusses all major research directions in the field, highlights many significant applications of Markov decision processes models, and explores numerous important topics that have previously been neglected or given cursory coverage in the literature. Markov Decision Processes focuses primarily on infinite horizon discrete time models and models with discrete time spaces while also examining models with arbitrary state spaces, finite horizon models, and continuous-time discrete state models. The book is organized around optimality criteria, using a common framework centered on the optimality (Bellman) equation for presenting results. The results are presented in a \"theorem-proof\" format and elaborated on through both discussion and examples, including results that are not available in any other book. A two-state Markov decision process model, presented in Chapter 3, is analyzed repeatedly throughout the book and demonstrates many results and algorithms. Markov Decision Processes covers recent research advances in such areas as countable state space models with average reward criterion, constrained models, and models with risk sensitive optimality criteria. It also explores several topics that have received little or no attention in other books, including modified policy iteration, multichain models with average reward criterion, and sensitive optimality. In addition, a Bibliographic Remarks section in each chapter comments on relevant historic"
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                },
                {
                    "title": "WEIGHTED SUMS OF CERTAIN DEPENDENT RANDOM VARIABLES",
                    "abstract": "1. Let be a probability space and,be an increasing family of sub o'-fields of(we put(c) Let (xn)n=1, 2, \u2022c be a sequence of bounded martingale differences on , that is,xn(\u0192\u00d6) is bounded almost surely (a.s.) ands. for n =1, 2,.... It is easily seen that this sequence has the following properties[G] and [M], which have been introduced by Y. S. Chow ([1]) in an analogous form and by G. Alexits ([4]), respectively, and may be of independent interest."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.SY",
                "eess.SY",
                "math.OC",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nIs there a tractable algorithm that achieves minimax optimal regret guarantees O~(sp(h\u2217)SAT) without prior knowledge of the model in the context of learning average-reward Markov Decision Processes?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant gap in reinforcement learning, particularly in average-reward settings. Achieving minimax optimal regret guarantees in a tractable manner could lead to more efficient algorithms that can be applied in real-world scenarios where prior knowledge of the environment is often unavailable. This advancement could enhance the applicability of reinforcement learning in various fields, such as robotics, finance, and healthcare, ultimately leading to more robust decision-making systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to balance exploration and exploitation in an unknown environment while ensuring that the algorithm remains tractable. Naive approaches may fail because they often rely on prior knowledge or complex optimization techniques that are computationally expensive. The technical obstacles include developing an algorithm that can efficiently learn from limited information while still achieving optimal regret bounds, which requires sophisticated theoretical insights and practical implementations.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either achieving optimal regret bounds or ensuring tractability, but not both simultaneously. Existing algorithms often require prior knowledge of the model or rely on intractable optimization methods, which limits their applicability. The gap in the literature lies in the lack of a method that can meet all three requirements: minimax optimal regret, tractability, and no prior knowledge. Our approach differs by proposing a polynomial-time algorithm that satisfies these conditions, thus addressing the limitations of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a polynomial-time algorithm that operates within the framework of average-reward MDPs. We will utilize a specific dataset that simulates various MDP scenarios to evaluate the algorithm's performance. The primary metric for success will be the regret bound, specifically aiming for O~(sp(h\u2217)SAT). We expect the outcomes to demonstrate that our algorithm not only achieves the desired regret guarantees but also maintains tractability without requiring prior knowledge of the model, thereby contributing significantly to the field of reinforcement learning."
            }
        },
        "author_data": {
            "ca7e1933-417d-4ebe-b42e-2a949ea8028b": {
                "pk": "ca7e1933-417d-4ebe-b42e-2a949ea8028b",
                "name": "Victor Boone",
                "collaborators": [
                    "Georgios Piliouras",
                    "Panayotis Mertikopoulos"
                ],
                "domain": [
                    "Stochastic Bandits",
                    "Game Theory",
                    "No-Regret Learning",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "The Sliding Regret in Stochastic Bandits: Discriminating Index and Randomized Policies",
                        "abstract": "This paper studies the one-shot behavior of no-regret algorithms for stochastic bandits. Although many algorithms are known to be asymptotically optimal with respect to the expected regret, over a single run, their pseudo-regret seems to follow one of two tendencies: it is either smooth or bumpy. To measure this tendency, we introduce a new notion: the sliding regret, that measures the worst pseudo-regret over a time-window of fixed length sliding to infinity. We show that randomized methods (e.g. Thompson Sampling and MED) have optimal sliding regret, while index policies, although possibly asymptotically optimal for the expected regret, have the worst possible sliding regret under regularity conditions on their index (e.g. UCB, UCB-V, KL-UCB, MOSS, IMED etc.). We further analyze the average bumpiness of the pseudo-regret of index policies via the regret of exploration, that we show to be suboptimal as well."
                    },
                    {
                        "title": "When do discounted-optimal policies also optimize the gain?",
                        "abstract": "In this technical note, we establish an upper-bound on the threshold on the discount factor starting from which all discounted-optimal deterministic policies are gain-optimal, that we prove to be tight on an example. To address computability issues of that theoretical threshold, we provide a weaker bound which is tractable on ergodic MDPs in polynomial time."
                    },
                    {
                        "title": "From Darwin to Poincar\u00e9 and von Neumann: Recurrence and Cycles in Evolutionary and Algorithmic Game Theory",
                        "abstract": "Replicator dynamics, the continuous-time analogue of Multiplicative Weights Updates, is the main dynamic in evolutionary game theory. In simple evolutionary zero-sum games, such as Rock-Paper-Scissors, replicator dynamic is periodic \\cite{zeeman1980population}, however, its behavior in higher dimensions is not well understood. We provide a complete characterization of its behavior in zero-sum evolutionary games. We prove that, if and only if, the system has an interior Nash equilibrium, the dynamics exhibit Poincar\\'{e} recurrence, i.e., almost all orbits come arbitrary close to their initial conditions infinitely often. If no interior equilibria exist, then all interior initial conditions converge to the boundary. Specifically, the strategies that are not in the support of any equilibrium vanish in the limit of all orbits. All recurrence results furthermore extend to a class of games that generalize both graphical polymatrix games as well as evolutionary games, establishing a unifying link between evolutionary and algorithmic game theory. We show that two degrees of freedom, as in Rock-Paper-Scissors, is sufficient to prove periodicity."
                    },
                    {
                        "title": "The equivalence of dynamic and strategic stability under regularized learning in games",
                        "abstract": "In this paper, we examine the long-run behavior of regularized, no-regret learning in finite games. A well-known result in the field states that the empirical frequencies of no-regret play converge to the game's set of coarse correlated equilibria; however, our understanding of how the players' actual strategies evolve over time is much more limited - and, in many cases, non-existent. This issue is exacerbated further by a series of recent results showing that only strict Nash equilibria are stable and attracting under regularized learning, thus making the relation between learning and pointwise solution concepts particularly elusive. In lieu of this, we take a more general approach and instead seek to characterize the \\emph{setwise} rationality properties of the players' day-to-day play. To that end, we focus on one of the most stringent criteria of setwise strategic stability, namely that any unilateral deviation from the set in question incurs a cost to the deviator - a property known as closedness under better replies (club). In so doing, we obtain a far-reaching equivalence between strategic and dynamic stability: a product of pure strategies is closed under better replies if and only if its span is stable and attracting under regularized learning. In addition, we estimate the rate of convergence to such sets, and we show that methods based on entropic regularization (like the exponential weights algorithm) converge at a geometric rate, while projection-based methods converge within a finite number of iterations, even with bandit, payoff-based feedback."
                    }
                ]
            },
            "303d9322-f086-45ba-bc55-8aeaee8afe62": {
                "pk": "303d9322-f086-45ba-bc55-8aeaee8afe62",
                "name": "Zihan Zhang",
                "collaborators": [
                    "Xiangyang Ji",
                    "Zeyu Guo",
                    "Yuan Zhou",
                    "Simon S. Du",
                    "Qiaomin Xie",
                    "Yeyuan Chen",
                    "Xiang Xiang",
                    "Tianyi Zhang",
                    "Luciano Medina",
                    "Yulin Wang"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Error-Correcting Codes",
                    "Machine Learning",
                    "Diffusion Models"
                ],
                "publications": [
                    {
                        "title": "A New Metric on Symmetric Group and Applications to Block Permutation Codes",
                        "abstract": "Permutation codes have received a great attention due to various applications. For different applications, one needs permutation codes under different metrics. The generalized Cayley metric was introduced by Chee and Vu [4] and this metric includes several other metrics as special cases. However, the generalized Cayley metric is not easily computable in general. Therefore the block permutation metric was introduced by Yang et al. [22] as the generalized Cayley metric and the block permutation metric have the same magnitude. However, the block permutation metric lacks the symmetry property which restricts more advanced algebraic tools to be involved. In this paper, by introducing a novel metric closely related to the block permutation metric, we build a bridge between some advanced algebraic methods and codes in the block permutation metric. More specifically, based on some techniques from algebraic function fields originated in [19], we give an algebraic-geometric construction of codes in the novel metric with reasonably good parameters. By observing a trivial relation between the novel metric and block permutation metric, we then produce non-systematic codes in block permutation metric that improve all known results given in [21, 22]. More importantly, based on our non-systematic codes, we provide an explicit and systematic construction of codes in block permutation metric which improves the systematic result shown in [22]. In the end, we demonstrate that our codes in the novel metric itself have reasonably good parameters by showing that our construction beats the corresponding Gilbert-Varshamov bound."
                    },
                    {
                        "title": "Regret Minimization for Reinforcement Learning by Evaluating the Optimal Bias Function",
                        "abstract": "We present an algorithm based on the \\emph{Optimism in the Face of Uncertainty} (OFU) principle which is able to learn Reinforcement Learning (RL) modeled by Markov decision process (MDP) with finite state-action space efficiently. By evaluating the state-pair difference of the optimal bias function $h^{*}$, the proposed algorithm achieves a regret bound of $\\tilde{O}(\\sqrt{SAHT})$\\footnote{The symbol $\\tilde{O}$ means $O$ with log factors ignored. } for MDP with $S$ states and $A$ actions, in the case that an upper bound $H$ on the span of $h^{*}$, i.e., $sp(h^{*})$ is known. This result outperforms the best previous regret bounds $\\tilde{O}(S\\sqrt{AHT}) $\\citep{fruit2019improved} by a factor of $\\sqrt{S}$. Furthermore, this regret bound matches the lower bound of $\\Omega(\\sqrt{SAHT}) $\\citep{jaksch2010near} up to a logarithmic factor. As a consequence, we show that there is a near optimal regret bound of $\\tilde{O}(\\sqrt{SADT})$ for MDPs with a finite diameter $D$ compared to the lower bound of $\\Omega(\\sqrt{SADT}) $\\citep{jaksch2010near}."
                    },
                    {
                        "title": "Randomly Punctured Reed-Solomon Codes Achieve the List Decoding Capacity over Polynomial-Size Alphabets",
                        "abstract": "This paper shows that, with high probability, randomly punctured Reed-Solomon codes over fields of polynomial size achieve the list decoding capacity. More specifically, we prove that for any $\\epsilon>0$ and $R\\in (0,1)$, with high probability, randomly punctured Reed-Solomon codes of block length $n$ and rate $R$ are $\\left(1-R-\\epsilon, O({1}/{\\epsilon})\\right)$ list decodable over alphabets of size at least $2^{\\mathrm{poly}(1/\\epsilon)}n^2$. This extends the recent breakthrough of Brakensiek, Gopi, and Makam (STOC 2023) that randomly punctured Reed-Solomon codes over fields of exponential size attain the generalized Singleton bound of Shangguan and Tamo (STOC 2020)."
                    },
                    {
                        "title": "Sharper Model-free Reinforcement Learning for Average-reward Markov Decision Processes",
                        "abstract": "We develop several provably efficient model-free reinforcement learning (RL) algorithms for infinite-horizon average-reward Markov Decision Processes (MDPs). We consider both online setting and the setting with access to a simulator. In the online setting, we propose model-free RL algorithms based on reference-advantage decomposition. Our algorithm achieves $\\widetilde{O}(S^5A^2\\mathrm{sp}(h^*)\\sqrt{T})$ regret after $T$ steps, where $S\\times A$ is the size of state-action space, and   $\\mathrm{sp}(h^*)$ the span of the optimal bias function. Our results are the first to achieve optimal dependence in $T$ for weakly communicating MDPs.   In the simulator setting, we propose a model-free RL algorithm that finds an $\\epsilon$-optimal policy using $\\widetilde{O} \\left(\\frac{SA\\mathrm{sp}^2(h^*)}{\\epsilon^2}+\\frac{S^2A\\mathrm{sp}(h^*)}{\\epsilon} \\right)$ samples, whereas the minimax lower bound is $\\Omega\\left(\\frac{SA\\mathrm{sp}(h^*)}{\\epsilon^2}\\right)$.   Our results are based on two new techniques that are unique in the average-reward setting: 1) better discounted approximation by value-difference estimation; 2) efficient construction of confidence region for the optimal bias function with space complexity $O(SA)$."
                    },
                    {
                        "title": "Explicit Folded Reed-Solomon and Multiplicity Codes Achieve Relaxed Generalized Singleton Bounds",
                        "abstract": "In this paper, we prove that explicit FRS codes and multiplicity codes achieve relaxed generalized Singleton bounds for list size $L\\ge1.$ Specifically, we show the following: (1) FRS code of length $n$ and rate $R$ over the alphabet $\\mathbb{F}_q^s$ with distinct evaluation points is $\\left(\\frac{L}{L+1}\\left(1-\\frac{sR}{s-L+1}\\right),L\\right)$ list-decodable (LD) for list size $L\\in[s]$. (2) Multiplicity code of length $n$ and rate $R$ over the alphabet $\\mathbb{F}_p^s$ with distinct evaluation points is $\\left(\\frac{L}{L+1}\\left(1-\\frac{sR}{s-L+1}\\right),L\\right)$ LD for list size $L\\in[s]$.   Choosing $s=\\Theta(1/\\epsilon^2)$ and $L=O(1/\\epsilon)$, our results imply that both FRS codes and multiplicity codes achieve LD capacity $1-R-\\epsilon$ with optimal list size $O(1/\\epsilon)$. This exponentially improves the previous state of the art $(1/\\epsilon)^{O(1/\\epsilon)}$ established by Kopparty et. al. (FOCS 2018) and Tamo (IEEE TIT, 2024). In particular, our results on FRS codes fully resolve a open problem proposed by Guruswami and Rudra (STOC 2006). Furthermore, our results imply the first explicit constructions of $(1-R-\\epsilon,O(1/\\epsilon))$ LD codes of rate $R$ with poly-sized alphabets.   Our method can also be extended to analyze the list-recoverability (LR) of FRS codes. We provide a tighter radius upper bound that FRS codes cannot be $(\\frac{L+1-\\ell}{L+1}(1-\\frac{mR}{m-1})+o(1),\\ell, L)$ LR where $m=\\lceil\\log_{\\ell}{(L+1)}\\rceil$. We conjecture this bound is almost tight when $L+1=\\ell^a$ for any $a\\in\\mathbb{N}^{\\ge 2}$. To give some evidences, we show FRS codes are $\\left(\\frac{1}{2}-\\frac{sR}{s-2},2,3\\right)$ LR, which proves the tightness in the smallest non-trivial case. Our bound refutes the possibility that FRS codes could achieve LR capacity $(1-R-\\epsilon, \\ell, O(\\frac{\\ell}{\\epsilon}))$. This implies an intrinsic separation between LD and LR of FRS codes."
                    },
                    {
                        "title": "Long-Tailed Classification with Gradual Balanced Loss and Adaptive Feature Generation",
                        "abstract": "The real-world data distribution is essentially long-tailed, which poses great challenge to the deep model. In this work, we propose a new method, Gradual Balanced Loss and Adaptive Feature Generator (GLAG) to alleviate imbalance. GLAG first learns a balanced and robust feature model with Gradual Balanced Loss, then fixes the feature model and augments the under-represented tail classes on the feature level with the knowledge from well-represented head classes. And the generated samples are mixed up with real training samples during training epochs. Gradual Balanced Loss is a general loss and it can combine with different decoupled training methods to improve the original performance. State-of-the-art results have been achieved on long-tail datasets such as CIFAR100-LT, ImageNetLT, and iNaturalist, which demonstrates the effectiveness of GLAG for long-tailed visual recognition."
                    },
                    {
                        "title": "Existence of Optical Vortex Solitons in Photorefractive Media",
                        "abstract": "Optical propagation and vortices in nonlinear media have been intensively studied in modern optical physics. In this paper, we establish constraints regarding the propagation constant and provide an existence theory and numerical computations for positive exponentially decaying solutions for a class of ring-profiled solitons in a type of nonlinear media known as a photorefractive nonlinearity. Our methods include constrained minimization and finite element formalism, and we study the vortex profile and its propagation by fixing the energy flux."
                    },
                    {
                        "title": "Sampling Proper Colorings on Line Graphs Using $(1+o(1))\u0394$ Colors",
                        "abstract": "We prove that the single-site Glauber dynamics for sampling proper $q$-colorings mixes in $O_\\Delta(n\\log n)$ time on line graphs with $n$ vertices and maximum degree $\\Delta$ when $q>(1+o(1))\\Delta$. The main tool in our proof is the matrix trickle-down theorem developed by Abdolazimi, Liu and Oveis Gharan (FOCS, 2021)."
                    },
                    {
                        "title": "Random Reed-Solomon Codes Achieve the Half-Singleton Bound for Insertions and Deletions over Linear-Sized Alphabets",
                        "abstract": "In this paper, we prove that with high probability, random Reed-Solomon codes approach the half-Singleton bound - the optimal rate versus error tradeoff for linear insdel codes - with linear-sized alphabets. More precisely, we prove that, for any $\\epsilon>0$ and positive integers $n$ and $k$, with high probability, random Reed--Solomon codes of length $n$ and dimension $k$ can correct $(1-\\varepsilon)n-2k+1$ adversarial insdel errors over alphabets of size $n+2^{\\mathsf{poly}(1/\\varepsilon)}k$. This significantly improves upon the alphabet size demonstrated in the work of Con, Shpilka, and Tamo (IEEE TIT, 2023), who showed the existence of Reed--Solomon codes with exponential alphabet size $\\widetilde O\\left(\\binom{n}{2k-1}^2\\right)$ precisely achieving the half-Singleton bound.   Our methods are inspired by recent works on list-decoding Reed-Solomon codes. Brakensiek-Gopi-Makam (STOC 2023) showed that random Reed-Solomon codes are list-decodable up to capacity with exponential-sized alphabets, and Guo-Zhang (FOCS 2023) and Alrabiah-Guruswami-Li (STOC 2024) improved the alphabet-size to linear. We achieve a similar alphabet-size reduction by similarly establishing strong bounds on the probability that certain random rectangular matrices are full rank. To accomplish this in our insdel context, our proof combines the random matrix techniques from list-decoding with structural properties of Longest Common Subsequences."
                    },
                    {
                        "title": "Model-Free Reinforcement Learning: from Clipped Pseudo-Regret to Sample Complexity",
                        "abstract": "In this paper we consider the problem of learning an $\\epsilon$-optimal policy for a discounted Markov Decision Process (MDP). Given an MDP with $S$ states, $A$ actions, the discount factor $\\gamma \\in (0,1)$, and an approximation threshold $\\epsilon > 0$, we provide a model-free algorithm to learn an $\\epsilon$-optimal policy with sample complexity $\\tilde{O}(\\frac{SA\\ln(1/p)}{\\epsilon^2(1-\\gamma)^{5.5}})$ (where the notation $\\tilde{O}(\\cdot)$ hides poly-logarithmic factors of $S,A,1/(1-\\gamma)$, and $1/\\epsilon$) and success probability $(1-p)$. For small enough $\\epsilon$, we show an improved algorithm with sample complexity $\\tilde{O}(\\frac{SA\\ln(1/p)}{\\epsilon^2(1-\\gamma)^{3}})$. While the first bound improves upon all known model-free algorithms and model-based ones with tight dependence on $S$, our second algorithm beats all known sample complexity bounds and matches the information theoretic lower bound up to logarithmic factors."
                    },
                    {
                        "title": "Horizon-Free Reinforcement Learning in Polynomial Time: the Power of Stationary Policies",
                        "abstract": "This paper gives the first polynomial-time algorithm for tabular Markov Decision Processes (MDP) that enjoys a regret bound \\emph{independent on the planning horizon}. Specifically, we consider tabular MDP with $S$ states, $A$ actions, a planning horizon $H$, total reward bounded by $1$, and the agent plays for $K$ episodes. We design an algorithm that achieves an $O\\left(\\mathrm{poly}(S,A,\\log K)\\sqrt{K}\\right)$ regret in contrast to existing bounds which either has an additional $\\mathrm{polylog}(H)$ dependency~\\citep{zhang2020reinforcement} or has an exponential dependency on $S$~\\citep{li2021settling}. Our result relies on a sequence of new structural lemmas establishing the approximation power, stability, and concentration property of stationary policies, which can have applications in other problems related to Markov chains."
                    },
                    {
                        "title": "Almost Optimal Batch-Regret Tradeoff for Batch Linear Contextual Bandits",
                        "abstract": "We study the optimal batch-regret tradeoff for batch linear contextual bandits. For any batch number $M$, number of actions $K$, time horizon $T$, and dimension $d$, we provide an algorithm and prove its regret guarantee, which, due to technical reasons, features a two-phase expression as the time horizon $T$ grows. We also prove a lower bound theorem that surprisingly shows the optimality of our two-phase regret upper bound (up to logarithmic factors) in the \\emph{full range} of the problem parameters, therefore establishing the exact batch-regret tradeoff.   Compared to the recent work \\citep{ruan2020linear} which showed that $M = O(\\log \\log T)$ batches suffice to achieve the asymptotically minimax-optimal regret without the batch constraints, our algorithm is simpler and easier for practical implementation. Furthermore, our algorithm achieves the optimal regret for all $T \\geq d$, while \\citep{ruan2020linear} requires that $T$ greater than an unrealistically large polynomial of $d$.   Along our analysis, we also prove a new matrix concentration inequality with dependence on their dynamic upper bounds, which, to the best of our knowledge, is the first of its kind in literature and maybe of independent interest."
                    },
                    {
                        "title": "Sharp Variance-Dependent Bounds in Reinforcement Learning: Best of Both Worlds in Stochastic and Deterministic Environments",
                        "abstract": "We study variance-dependent regret bounds for Markov decision processes (MDPs). Algorithms with variance-dependent regret guarantees can automatically exploit environments with low variance (e.g., enjoying constant regret on deterministic MDPs). The existing algorithms are either variance-independent or suboptimal. We first propose two new environment norms to characterize the fine-grained variance properties of the environment. For model-based methods, we design a variant of the MVP algorithm (Zhang et al., 2021a). We apply new analysis techniques to demonstrate that this algorithm enjoys variance-dependent bounds with respect to the norms we propose. In particular, this bound is simultaneously minimax optimal for both stochastic and deterministic MDPs, the first result of its kind. We further initiate the study on model-free algorithms with variance-dependent regret bounds by designing a reference-function-based algorithm with a novel capped-doubling reference update schedule. Lastly, we also provide lower bounds to complement our upper bounds."
                    },
                    {
                        "title": "TEDi: Temporally-Entangled Diffusion for Long-Term Motion Synthesis",
                        "abstract": "The gradual nature of a diffusion process that synthesizes samples in small increments constitutes a key ingredient of Denoising Diffusion Probabilistic Models (DDPM), which have presented unprecedented quality in image synthesis and been recently explored in the motion domain. In this work, we propose to adapt the gradual diffusion concept (operating along a diffusion time-axis) into the temporal-axis of the motion sequence. Our key idea is to extend the DDPM framework to support temporally varying denoising, thereby entangling the two axes. Using our special formulation, we iteratively denoise a motion buffer that contains a set of increasingly-noised poses, which auto-regressively produces an arbitrarily long stream of frames. With a stationary diffusion time-axis, in each diffusion step we increment only the temporal-axis of the motion such that the framework produces a new, clean frame which is removed from the beginning of the buffer, followed by a newly drawn noise vector that is appended to it. This new mechanism paves the way towards a new framework for long-term motion synthesis with applications to character animation and other domains."
                    }
                ]
            }
        }
    },
    "2309.17388": {
        "paper_data": {
            "title": "Tree Cross Attention",
            "url": "http://arxiv.org/abs/2309.17388v2",
            "arxiv_id": "2309.17388",
            "authors": [
                "Leo Feng",
                "Frederick Tung",
                "Hossein Hajimirsadeghi",
                "Yoshua Bengio",
                "Mohamed Osama Ahmed"
            ],
            "abstract": "Cross Attention is a popular method for retrieving information from a set of context tokens for making predictions. At inference time, for each prediction, Cross Attention scans the full set of $\\mathcal{O}(N)$ tokens. In practice, however, often only a small subset of tokens are required for good performance. Methods such as Perceiver IO are cheap at inference as they distill the information to a smaller-sized set of latent tokens $L < N$ on which cross attention is then applied, resulting in only $\\mathcal{O}(L)$ complexity. However, in practice, as the number of input tokens and the amount of information to distill increases, the number of latent tokens needed also increases significantly. In this work, we propose Tree Cross Attention (TCA) - a module based on Cross Attention that only retrieves information from a logarithmic $\\mathcal{O}(\\log(N))$ number of tokens for performing inference. TCA organizes the data in a tree structure and performs a tree search at inference time to retrieve the relevant tokens for prediction. Leveraging TCA, we introduce ReTreever, a flexible architecture for token-efficient inference. We show empirically that Tree Cross Attention (TCA) performs comparable to Cross Attention across various classification and uncertainty regression tasks while being significantly more token-efficient. Furthermore, we compare ReTreever against Perceiver IO, showing significant gains while using the same number of tokens for inference.",
            "introduction": "   1 Introduction  With the rapid growth in applications of machine learning, an important objective is to make inference efficient both in terms of compute and memory. NVIDIA\u00a0(Leopold, 2019) and Amazon\u00a0(Barr, 2019) estimate that 80\u201390% of the ML workload is from performing inference. Furthermore, with the rapid growth in low-memory/compute domains (e.g. IoT devices) and the popularity of attention mechanisms in recent years, there is a strong incentive to design more efficient attention mechanisms for performing inference.   Cross Attention (CA) is a popular method at inference time for retrieving relevant information from a set of context tokens. CA scales linearly with the number of context tokens \ud835\udcaa\u2062(N)\ud835\udcaa\ud835\udc41\\mathcal{O}(N)caligraphic_O ( italic_N ). However, in reality many of the context tokens are not needed, making CA unnecessarily expensive in practice. General-purpose architectures such as Perceiver IO\u00a0(Jaegle et\u00a0al., 2021) perform inference cheaply by first distilling the contextual information down into a smaller fixed-sized set of latent tokens L<N\ud835\udc3f\ud835\udc41L<Nitalic_L < italic_N. When performing inference, information is instead retrieved from the fixed-size set of latent tokens \ud835\udcaa\u2062(L)\ud835\udcaa\ud835\udc3f\\mathcal{O}(L)caligraphic_O ( italic_L ). Methods that achieve efficient inference via distillation are problematic since (1) problems with a high intrinsic dimensionality naturally require a large number of latents, and (2) the number of latents (capacity of the inference model) is a hyperparameter that requires specifying before training. However, in many practical problems, the required model\u2019s capacity may not be known beforehand. For example, in settings where the amount of data increases overtime (e.g., Bayesian Optimization, Contextual Bandits, Active Learning, etc\u2026), the number of latents needed in the beginning and after many data acquisition steps can be vastly different.   In this work, we propose (1) Tree Cross Attention (TCA), a replacement for Cross Attention that performs retrieval, scaling logarithmically \ud835\udcaa\u2062(log\u2061(N))\ud835\udcaa\ud835\udc41\\mathcal{O}(\\log(N))caligraphic_O ( roman_log ( italic_N ) ) with the number of tokens. TCA organizes the tokens into a tree structure. From this, it then performs retrieval via a tree search, starting from the root. As a result, TCA selectively chooses the information to retrieve from the tree depending on a query feature vector. TCA leverages Reinforcement Learning (RL) to learn good representations for the internal nodes of the tree. Building on TCA, we also propose (2) ReTreever, a flexible architecture that achieves token-efficient inference.   In our experiments, we show (1) TCA achieves results competitive to that of Cross Attention while only requiring a logarithmic number of tokens, (2) ReTreever outperforms Perceiver IO on various classification and uncertainty estimation tasks while using the same number of tokens, (3) ReTreever\u2019s optimization objective can leverage non-differentiable objectives such as classification accuracy, and (4) TCA\u2019s memory usage scales logarithmically with the number of tokens unlike Cross Attention which scales linearly.     2 Background   2.1 Attention  Attention retrieves information from a context set Xcsubscript\ud835\udc4b\ud835\udc50X_{c}italic_X start_POSTSUBSCRIPT italic_c end_POSTSUBSCRIPT as follows:    Attention\u2062(Q,K,V)=softmax\u2062(Q\u2062KTd)\u2062VAttention\ud835\udc44\ud835\udc3e\ud835\udc49softmax\ud835\udc44superscript\ud835\udc3e\ud835\udc47\ud835\udc51\ud835\udc49\\mathrm{Attention}(Q,K,V)=\\mathrm{softmax}(\\frac{QK^{T}}{\\sqrt{d}})Vroman_Attention ( italic_Q , italic_K , italic_V ) = roman_softmax ( divide start_ARG italic_Q italic_K start_POSTSUPERSCRIPT italic_T end_POSTSUPERSCRIPT end_ARG start_ARG square-root start_ARG italic_d end_ARG end_ARG ) italic_V    where Q\ud835\udc44Qitalic_Q is the query matrix, K\ud835\udc3eKitalic_K is the key matrix, V\ud835\udc49Vitalic_V is the value matrix, and d\ud835\udc51ditalic_d is the dimension of the key and query vectors. In this work, we focus on Cross Attention where the objective is to retrieve information",
            "references": [
                {
                    "title": "Latent Bottlenecked Attentive Neural Processes",
                    "abstract": "Neural Processes (NPs) are popular methods in meta-learning that can estimate predictive uncertainty on target datapoints by conditioning on a context dataset. Previous state-of-the-art method Transformer Neural Processes (TNPs) achieve strong performance but require quadratic computation with respect to the number of context datapoints, significantly limiting its scalability. Conversely, existing sub-quadratic NP variants perform significantly worse than that of TNPs. Tackling this issue, we propose Latent Bottlenecked Attentive Neural Processes (LBANPs), a new computationally efficient sub-quadratic NP variant, that has a querying computational complexity independent of the number of context datapoints. The model encodes the context dataset into a constant number of latent vectors on which self-attention is performed. When making predictions, the model retrieves higher-order information from the context dataset via multiple cross-attention mechanisms on the latent vectors. We empirically show that LBANPs achieve results competitive with the state-of-the-art on meta-regression, image completion, and contextual multi-armed bandits. We demonstrate that LBANPs can trade-off the computational cost and performance according to the number of latent vectors. Finally, we show LBANPs can scale beyond existing attention-based NP variants to larger dataset settings."
                },
                {
                    "title": "Treeformer: Dense Gradient Trees for Efficient Attention Computation",
                    "abstract": "Standard inference and training with transformer based architectures scale quadratically with input sequence length. This is prohibitively large for a variety of applications especially in web-page translation, query-answering etc. Consequently, several approaches have been developed recently to speedup attention computation by enforcing different attention structures such as sparsity, low-rank, approximating attention using kernels. In this work, we view attention computation as that of nearest neighbor retrieval, and use decision tree based hierarchical navigation to reduce the retrieval cost per query token from linear in sequence length to nearly logarithmic. Based on such hierarchical navigation, we design Treeformer which can use one of two efficient attention layers -- TF-Attention and TC-Attention. TF-Attention computes the attention in a fine-grained style, while TC-Attention is a coarse attention layer which also ensures that the gradients are\"dense\". To optimize such challenging discrete layers, we propose a two-level bootstrapped training method. Using extensive experiments on standard NLP benchmarks, especially for long-sequences, we demonstrate that our Treeformer architecture can be almost as accurate as baseline Transformer while using 30x lesser FLOPs in the attention layer. Compared to Linformer, the accuracy can be as much as 12% higher while using similar FLOPs in the attention layer."
                },
                {
                    "title": "Transformer Neural Processes: Uncertainty-Aware Meta Learning Via Sequence Modeling",
                    "abstract": "Neural Processes (NPs) are a popular class of approaches for meta-learning. Similar to Gaussian Processes (GPs), NPs define distributions over functions and can estimate uncertainty in their predictions. However, unlike GPs, NPs and their variants suffer from underfitting and often have intractable likelihoods, which limit their applications in sequential decision making. We propose Transformer Neural Processes (TNPs), a new member of the NP family that casts uncertainty-aware meta learning as a sequence modeling problem. We learn TNPs via an autoregressive likelihood-based objective and instantiate it with a novel transformer-based architecture. The model architecture respects the inductive biases inherent to the problem structure, such as invariance to the observed data points and equivariance to the unobserved points. We further investigate knobs within the TNP framework that tradeoff expressivity of the decoding distribution with extra computation. Empirically, we show that TNPs achieve state-of-the-art performance on various benchmark problems, outperforming all previous NP variants on meta regression, image completion, contextual multi-armed bandits, and Bayesian optimization."
                },
                {
                    "title": "Graph Neural Networks Designed for Different Graph Types: A Survey",
                    "abstract": "Graphs are ubiquitous in nature and can therefore serve as models for many practical but also theoretical problems. For this purpose, they can be defined as many different types which suitably reflect the individual contexts of the represented problem. To address cutting-edge problems based on graph data, the research field of Graph Neural Networks (GNNs) has emerged. Despite the field's youth and the speed at which new models are developed, many recent surveys have been published to keep track of them. Nevertheless, it has not yet been gathered which GNN can process what kind of graph types. In this survey, we give a detailed overview of already existing GNNs and, unlike previous surveys, categorize them according to their ability to handle different graph types and properties. We consider GNNs operating on static and dynamic graphs of different structural constitutions, with or without node or edge attributes. Moreover, we distinguish between GNN models for discrete-time or continuous-time dynamic graphs and group the models according to their architecture. We find that there are still graph types that are not or only rarely covered by existing GNN models. We point out where models are missing and give potential reasons for their absence."
                },
                {
                    "title": "Perceiver IO: A General Architecture for Structured Inputs & Outputs",
                    "abstract": "A central goal of machine learning is the development of systems that can solve many problems in as many data domains as possible. Current architectures, however, cannot be applied beyond a small set of stereotyped settings, as they bake in domain&task assumptions or scale poorly to large inputs or outputs. In this work, we propose Perceiver IO, a general-purpose architecture that handles data from arbitrary settings while scaling linearly with the size of inputs and outputs. Our model augments the Perceiver with a flexible querying mechanism that enables outputs of various sizes and semantics, doing away with the need for task-specific architecture engineering. The same architecture achieves strong results on tasks spanning natural language and visual understanding, multi-task and multi-modal reasoning, and StarCraft II. As highlights, Perceiver IO outperforms a Transformer-based BERT baseline on the GLUE language benchmark despite removing input tokenization and achieves state-of-the-art performance on Sintel optical flow estimation with no explicit mechanisms for multiscale correspondence."
                },
                {
                    "title": "Bootstrapping Neural Processes",
                    "abstract": "Unlike in the traditional statistical modeling for which a user typically hand-specify a prior, Neural Processes (NPs) implicitly define a broad class of stochastic processes with neural networks. Given a data stream, NP learns a stochastic process that best describes the data. While this \"data-driven\" way of learning stochastic processes has proven to handle various types of data, NPs still rely on an assumption that uncertainty in stochastic processes is modeled by a single latent variable, which potentially limits the flexibility. To this end, we propose the Boostrapping Neural Process (BNP), a novel extension of the NP family using the bootstrap. The bootstrap is a classical data-driven technique for estimating uncertainty, which allows BNP to learn the stochasticity in NPs without assuming a particular form. We demonstrate the efficacy of BNP on various types of data and its robustness in the presence of model-data mismatch."
                },
                {
                    "title": "Multi-Time Attention Networks for Irregularly Sampled Time Series",
                    "abstract": "Irregular sampling occurs in many time series modeling applications where it presents a significant challenge to standard deep learning models. This work is motivated by the analysis of physiological time series data in electronic health records, which are sparse, irregularly sampled, and multivariate. In this paper, we propose a new deep learning framework for this setting that we call Multi-Time Attention Networks. Multi-Time Attention Networks learn an embedding of continuous time values and use an attention mechanism to produce a fixed-length representation of a time series containing a variable number of observations. We investigate the performance of our framework on interpolation and classification tasks using multiple datasets. Our results show that our approach performs as well or better than a range of baseline and recently proposed models while offering significantly faster training times than current state-of-the-art methods."
                },
                {
                    "title": "Tree-structured Attention with Hierarchical Accumulation",
                    "abstract": "Incorporating hierarchical structures like constituency trees has been shown to be effective for various natural language processing (NLP) tasks. However, it is evident that state-of-the-art (SOTA) sequence-based models like the Transformer struggle to encode such structures inherently. On the other hand, dedicated models like the Tree-LSTM, while explicitly modeling hierarchical structures, do not perform as efficiently as the Transformer. In this paper, we attempt to bridge this gap with Hierarchical Accumulation to encode parse tree structures into self-attention at constant time complexity. Our approach outperforms SOTA methods in four IWSLT translation tasks and the WMT'14 English-German task. It also yields improvements over Transformer and Tree-LSTM on three text classification tasks. We further demonstrate that using hierarchical priors can compensate for data shortage, and that our model prefers phrase-level attentions over token-level attentions."
                },
                {
                    "title": "Set Functions for Time Series",
                    "abstract": "Despite the eminent successes of deep neural networks, many architectures are often hard to transfer to irregularly-sampled and asynchronous time series that commonly occur in real-world datasets, especially in healthcare applications. This paper proposes a novel approach for classifying irregularly-sampled time series with unaligned measurements, focusing on high scalability and data efficiency. Our method SeFT (Set Functions for Time Series) is based on recent advances in differentiable set function learning, extremely parallelizable with a beneficial memory footprint, thus scaling well to large datasets of long time series and online monitoring scenarios. Furthermore, our approach permits quantifying per-observation contributions to the classification outcome. We extensively compare our method with existing algorithms on multiple healthcare time series datasets and demonstrate that it performs competitively whilst significantly reducing runtime."
                },
                {
                    "title": "Tree Transformer: Integrating Tree Structures into Self-Attention",
                    "abstract": "Pre-training Transformer from large-scale raw texts and fine-tuning on the desired task have achieved state-of-the-art results on diverse NLP tasks. However, it is unclear what the learned attention captures. The attention computed by attention heads seems not to match human intuitions about hierarchical structures. This paper proposes Tree Transformer, which adds an extra constraint to attention heads of the bidirectional Transformer encoder in order to encourage the attention heads to follow tree structures. The tree structures can be automatically induced from raw texts by our proposed \u201cConstituent Attention\u201d module, which is simply implemented by self-attention between two adjacent words. With the same training procedure identical to BERT, the experiments demonstrate the effectiveness of Tree Transformer in terms of inducing tree structures, better language modeling, and further learning more explainable attention scores."
                },
                {
                    "title": "Interpolation-Prediction Networks for Irregularly Sampled Time Series",
                    "abstract": "In this paper, we present a new deep learning architecture for addressing the problem of supervised learning with sparse and irregularly sampled multivariate time series. The architecture is based on the use of a semi-parametric interpolation network followed by the application of a prediction network. The interpolation network allows for information to be shared across multiple dimensions of a multivariate time series during the interpolation stage, while any standard deep learning model can be used for the prediction network. This work is motivated by the analysis of physiological time series data in electronic health records, which are sparse, irregularly sampled, and multivariate. We investigate the performance of this architecture on both classification and regression tasks, showing that our approach outperforms a range of baseline and recently proposed models."
                },
                {
                    "title": "Attentive Neural Processes",
                    "abstract": "Neural Processes (NPs) (Garnelo et al 2018a;b) approach regression by learning to map a context set of observed input-output pairs to a distribution over regression functions. Each function models the distribution of the output given an input, conditioned on the context. NPs have the benefit of fitting observed data efficiently with linear complexity in the number of context input-output pairs, and can learn a wide family of conditional distributions; they learn predictive distributions conditioned on context sets of arbitrary size. Nonetheless, we show that NPs suffer a fundamental drawback of underfitting, giving inaccurate predictions at the inputs of the observed data they condition on. We address this issue by incorporating attention into NPs, allowing each input location to attend to the relevant context points for the prediction. We show that this greatly improves the accuracy of predictions, results in noticeably faster training, and expands the range of functions that can be modelled."
                },
                {
                    "title": "Neural Processes",
                    "abstract": "A neural network (NN) is a parameterised function that can be tuned via gradient descent to approximate a labelled collection of data with high precision. A Gaussian process (GP), on the other hand, is a probabilistic model that defines a distribution over possible functions, and is updated in light of data via the rules of probabilistic inference. GPs are probabilistic, data-efficient and flexible, however they are also computationally intensive and thus limited in their applicability. We introduce a class of neural latent variable models which we call Neural Processes (NPs), combining the best of both worlds. Like GPs, NPs define distributions over functions, are capable of rapid adaptation to new observations, and can estimate the uncertainty in their predictions. Like NNs, NPs are computationally efficient during training and evaluation but also learn to adapt their priors to data. We demonstrate the performance of NPs on a range of learning tasks, including regression and optimisation, and compare and contrast with related models in the literature."
                },
                {
                    "title": "Conditional Neural Processes",
                    "abstract": "Deep neural networks excel at function approximation, yet they are typically trained from scratch for each new function. On the other hand, Bayesian methods, such as Gaussian Processes (GPs), exploit prior knowledge to quickly infer the shape of a new function at test time. Yet GPs are computationally expensive, and it can be hard to design appropriate priors. In this paper we propose a family of neural models, Conditional Neural Processes (CNPs), that combine the benefits of both. CNPs are inspired by the flexibility of stochastic processes such as GPs, but are structured as neural networks and trained via gradient descent. CNPs make accurate predictions after observing only a handful of training data points, yet scale to complex functions and large datasets. We demonstrate the performance and versatility of the approach on a range of canonical machine learning tasks, including regression, classification and image completion."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "Improved Semantic Representations From Tree-Structured Long Short-Term Memory Networks",
                    "abstract": "Because of their superior ability to preserve sequence information over time, Long Short-Term Memory (LSTM) networks, a type of recurrent neural network with a more complex computational unit, have obtained strong results on a variety of sequence modeling tasks. The only underlying LSTM structure that has been explored so far is a linear chain. However, natural language exhibits syntactic properties that would naturally combine words to phrases. We introduce the Tree-LSTM, a generalization of LSTMs to tree-structured network topologies. Tree-LSTMs outperform all existing systems and strong LSTM baselines on two tasks: predicting the semantic relatedness of two sentences (SemEval 2014, Task 1) and sentiment classification (Stanford Sentiment Treebank)."
                },
                {
                    "title": "Introduction to Algorithms, third edition",
                    "abstract": "If you had to buy just one text on algorithms, Introduction to Algorithms is a magnificent choice. The book begins by considering the mathematical foundations of the analysis of algorithms and maintains this mathematical rigor throughout the work. The tools developed in these opening sections are then applied to sorting, data structures, graphs, and a variety of selected algorithms including computational geometry, string algorithms, parallel models of computation, fast Fourier transforms (FFTs), and more. This book's strength lies in its encyclopedic range, clear exposition, and powerful analysis. Pseudo-code explanation of the algorithms coupled with proof of their accuracy makes this book is a great resource on the basic tools used to analyze the performance of algorithms."
                },
                {
                    "title": "Constant Memory Attentive Neural Processes",
                    "abstract": "Neural Processes (NPs) are efficient methods for estimating predictive uncertainties. NPs comprise of a conditioning phase where a context dataset is encoded, a querying phase where the model makes predictions using the context dataset encoding, and an updating phase where the model updates its encoding with newly received datapoints. However, state-of-the-art methods require additional memory which scales linearly or quadratically with the size of the dataset, limiting their applications, particularly in low-resource settings. In this work, we propose Constant Memory Attentive Neural Processes (CMANPs), an NP variant which only requires constant memory for the conditioning, querying, and updating phases. In building CMANPs, we propose Constant Memory Attention Block (CMAB), a novel general-purpose attention block that can compute its output in constant memory and perform updates in constant computation. Empirically, we show CMANPs achieve state-of-the-art results on meta-regression and image completion tasks while being (1) significantly more memory efficient than prior methods and (2) more scalable to harder settings."
                },
                {
                    "title": "A Comprehensive Survey on Graph Neural Networks",
                    "abstract": "Deep learning has revolutionized many machine learning tasks in recent years, ranging from image classification and video processing to speech recognition and natural language understanding. The data in these tasks are typically represented in the Euclidean space. However, there is an increasing number of applications, where data are generated from non-Euclidean domains and are represented as graphs with complex relationships and interdependency between objects. The complexity of graph data has imposed significant challenges on the existing machine learning algorithms. Recently, many studies on extending deep learning approaches for graph data have emerged. In this article, we provide a comprehensive overview of graph neural networks (GNNs) in data mining and machine learning fields. We propose a new taxonomy to divide the state-of-the-art GNNs into four categories, namely, recurrent GNNs, convolutional GNNs, graph autoencoders, and spatial\u2013temporal GNNs. We further discuss the applications of GNNs across various domains and summarize the open-source codes, benchmark data sets, and model evaluation of GNNs. Finally, we propose potential research directions in this rapidly growing field."
                },
                {
                    "title": "Latent Ordinary Differential Equations for Irregularly-Sampled Time Series",
                    "abstract": "Time series with non-uniform intervals occur in many applications, and are difficult to model using standard recurrent neural networks (RNNs). We generalize RNNs to have continuous-time hidden dynamics defined by ordinary differential equations (ODEs), a model we call ODE-RNNs. Furthermore, we use ODE-RNNs to replace the recognition network of the recently-proposed Latent ODE model. Both ODE-RNNs and Latent ODEs can naturally handle arbitrary time gaps between observations, and can explicitly model the probability of observation times using Poisson processes. We show experimentally that these ODE-based models outperform their RNN-based counterparts on irregularly-sampled data."
                },
                {
                    "title": "Five Balltree Construction Algorithms",
                    "abstract": "Balltrees are simple geometric data structures with a wide range of practical applica\u00ad tions to geometric \u00b7learning tasks. In this report we compare 5 different algorithms for . constructing ball trees from data. We study the trade-off between construction time and the quality of the constructed tree. Two of the algorithms are on-line, two construct the structures from the data set in a top down fashion, and one uses a bottom up approach. We empirically study the algorithms on random data drawn from eight different probability distributions representing smooth, clustered, and curve distributed data in different ambient space dimen\u00ad sions. We find that the bottom up approach usually produces the best trees but has the longest construction time. The other approaches have uses in specific circumstances. 1. IntemauonaI.Computer Science Institute, Berkeley, CA."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we design an efficient attention mechanism for inference that reduces computational and memory costs while maintaining performance, particularly in scenarios with varying data sizes and intrinsic dimensionality?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses the growing demand for efficient machine learning models, especially in low-resource environments like IoT devices. By developing a more efficient attention mechanism, we can significantly reduce the computational burden associated with inference, enabling broader applications of machine learning in real-time and resource-constrained settings. This advancement could lead to new research directions in model optimization, adaptive learning, and applications in dynamic environments where data availability fluctuates.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the intrinsic dimensionality of many problems, which necessitates a large number of context tokens for effective inference. Naive approaches, such as using standard Cross Attention, fail because they scale linearly with the number of tokens, leading to inefficiencies when many tokens are irrelevant. Additionally, determining the optimal number of latent tokens required for inference is complex, especially in scenarios where data evolves over time. Overcoming these technical obstacles requires innovative methods to dynamically manage and retrieve relevant information from a potentially vast set of tokens.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on static architectures that do not adapt to varying data sizes or intrinsic dimensionality, leading to limitations in efficiency and flexibility. Existing solutions like Perceiver IO rely on a fixed number of latent tokens, which can be a significant barrier in practical applications where data characteristics change over time. Our approach, which introduces Tree Cross Attention (TCA) and ReTreever, differs by utilizing a tree structure for token organization and retrieval, allowing for logarithmic scaling and dynamic adaptation to the data, thus addressing the shortcomings of prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology includes the following key components: \n1. **Tree Cross Attention (TCA)**: A novel attention mechanism that organizes tokens in a tree structure and retrieves information using a logarithmic search method, scaling as \\( \\mathcal{O}(\\log(N)) \\).\n2. **ReTreever**: A flexible architecture built on TCA that optimizes token-efficient inference and can leverage non-differentiable objectives.\n3. **Datasets**: We will evaluate our methods on various classification and"
            }
        },
        "author_data": {
            "dc735acb-8645-4bda-9e0b-2f6b98121d31": {
                "pk": "dc735acb-8645-4bda-9e0b-2f6b98121d31",
                "name": "Leo Feng",
                "collaborators": [
                    "Y. Bengio",
                    "Hossein Hajimirsadeghi",
                    "Frederick Tung",
                    "M. O. Ahmed",
                    "Mohamed Osama Ahmed",
                    "Pierre-Luc Bacon",
                    "L. Zintgraf",
                    "Shimon Whiteson",
                    "Emmanuel Bengio",
                    "Jarrid Rector-Brooks",
                    "A. Abdi",
                    "Cong Lu",
                    "Maximilian Igl",
                    "Katja Hofmann",
                    "Hossein Hajimirsadegh",
                    "Minsu Kim",
                    "Sanghyeok Choi",
                    "Taeyoung Yun",
                    "Sungsoo Ahn",
                    "Jinkyoo Park",
                    "Nikolay Malkin",
                    "Maksym Korablyov",
                    "Cheng-Hao Liu",
                    "Moksh Jain",
                    "A. V. D. Sloot",
                    "Eric Jolicoeur",
                    "Edward Ruediger",
                    "A. Nica",
                    "Kostiantyn Lapchevskyi",
                    "Daniel St-Cyr",
                    "Doris Alexandra Schuetz",
                    "V. Butoi",
                    "Simon R. Blackburn",
                    "Hadi Nekoei",
                    "S. Gottipati",
                    "Priyesh Vijayan",
                    "Prateek Gupta",
                    "Ladislav Ramp\u00e1\u0161ek",
                    "Sasikanth Avancha",
                    "William L. Hamilton",
                    "Brooks Paige",
                    "Sanchit Misra",
                    "Stanislaw Jastrzebski",
                    "Bharat Kaul",
                    "D. Precup",
                    "Jos'e Miguel Hern'andez-Lobato",
                    "Marwin H. S. Segler",
                    "Michael M. Bronstein",
                    "A. Marinier",
                    "Mike Tyers",
                    "Greg Mori",
                    "T. Deleu",
                    "David Kanaa",
                    "Giancarlo Kerg",
                    "Guillaume Lajoie",
                    "Padideh Nouri",
                    "Aneri Muni",
                    "Yoshua Bengio",
                    "Sebastian Schulze",
                    "K. Shiarlis",
                    "Y. Gal",
                    "Kristian Hartikainen",
                    "Bei Peng"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Neural Networks",
                    "Meta-Learning",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Were RNNs All We Needed?",
                        "abstract": "The scalability limitations of Transformers regarding sequence length have renewed interest in recurrent sequence models that are parallelizable during training. As a result, many novel recurrent architectures, such as S4, Mamba, and Aaren, have been proposed that achieve comparable performance. In this work, we revisit traditional recurrent neural networks (RNNs) from over a decade ago: LSTMs (1997) and GRUs (2014). While these models were slow due to requiring to backpropagate through time (BPTT), we show that by removing their hidden state dependencies from their input, forget, and update gates, LSTMs and GRUs no longer need to BPTT and can be efficiently trained in parallel. Building on this, we introduce minimal versions (minLSTMs and minGRUs) that (1) use significantly fewer parameters than their traditional counterparts and (2) are fully parallelizable during training (175x faster for a sequence of length 512). Lastly, we show that these stripped-down versions of decade-old RNNs match the empirical performance of recent sequence models."
                    },
                    {
                        "title": "Adaptive teachers for amortized samplers",
                        "abstract": "Amortized inference is the task of training a parametric model, such as a neural network, to approximate a distribution with a given unnormalized density where exact sampling is intractable. When sampling is implemented as a sequential decision-making process, reinforcement learning (RL) methods, such as generative flow networks, can be used to train the sampling policy. Off-policy RL training facilitates the discovery of diverse, high-reward candidates, but existing methods still face challenges in efficient exploration. We propose to use an adaptive training distribution (the Teacher) to guide the training of the primary amortized sampler (the Student) by prioritizing high-loss regions. The Teacher, an auxiliary behavior model, is trained to sample high-error regions of the Student and can generalize across unexplored modes, thereby enhancing mode coverage by providing an efficient training curriculum. We validate the effectiveness of this approach in a synthetic environment designed to present an exploration challenge, two diffusion-based sampling tasks, and four biochemical discovery tasks demonstrating its ability to improve sample efficiency and mode coverage."
                    },
                    {
                        "title": "Generative Active Learning for the Search of Small-molecule Protein Binders",
                        "abstract": "Despite substantial progress in machine learning for scientific discovery in recent years, truly de novo design of small molecules which exhibit a property of interest remains a significant challenge. We introduce LambdaZero, a generative active learning approach to search for synthesizable molecules. Powered by deep reinforcement learning, LambdaZero learns to search over the vast space of molecules to discover candidates with a desired property. We apply LambdaZero with molecular docking to design novel small molecules that inhibit the enzyme soluble Epoxide Hydrolase 2 (sEH), while enforcing constraints on synthesizability and drug-likeliness. LambdaZero provides an exponential speedup in terms of the number of calls to the expensive molecular docking oracle, and LambdaZero de novo designed molecules reach docking scores that would otherwise require the virtual screening of a hundred billion molecules. Importantly, LambdaZero discovers novel scaffolds of synthesizable, drug-like inhibitors for sEH. In in vitro experimental validation, a series of ligands from a generated quinazoline-based scaffold were synthesized, and the lead inhibitor N-(4,6-di(pyrrolidin-1-yl)quinazolin-2-yl)-N-methylbenzamide (UM0152893) displayed sub-micromolar enzyme inhibition of sEH."
                    },
                    {
                        "title": "Attention as an RNN",
                        "abstract": "The advent of Transformers marked a significant breakthrough in sequence modelling, providing a highly performant architecture capable of leveraging GPU parallelism. However, Transformers are computationally expensive at inference time, limiting their applications, particularly in low-resource settings (e.g., mobile and embedded devices). Addressing this, we (1) begin by showing that attention can be viewed as a special Recurrent Neural Network (RNN) with the ability to compute its \\textit{many-to-one} RNN output efficiently. We then (2) show that popular attention-based models such as Transformers can be viewed as RNN variants. However, unlike traditional RNNs (e.g., LSTMs), these models cannot be updated efficiently with new tokens, an important property in sequence modelling. Tackling this, we (3) introduce a new efficient method of computing attention's \\textit{many-to-many} RNN output based on the parallel prefix scan algorithm. Building on the new attention formulation, we (4) introduce \\textbf{Aaren}, an attention-based module that can not only (i) be trained in parallel (like Transformers) but also (ii) be updated efficiently with new tokens, requiring only constant memory for inferences (like traditional RNNs). Empirically, we show Aarens achieve comparable performance to Transformers on $38$ datasets spread across four popular sequential problem settings: reinforcement learning, event forecasting, time series classification, and time series forecasting tasks while being more time and memory-efficient."
                    },
                    {
                        "title": "Memory Efficient Neural Processes via Constant Memory Attention Block",
                        "abstract": "Neural Processes (NPs) are popular meta-learning methods for efficiently modelling predictive uncertainty. Recent state-of-the-art methods, however, leverage expensive attention mechanisms, limiting their applications, particularly in low-resource settings. In this work, we propose Constant Memory Attentive Neural Processes (CMANPs), an NP variant that only requires constant memory. To do so, we first propose an efficient update operation for Cross Attention. Leveraging the update operation, we propose Constant Memory Attention Block (CMAB), a novel attention block that (i) is permutation invariant, (ii) computes its output in constant memory, and (iii) performs constant computation updates. Finally, building on CMAB, we detail Constant Memory Attentive Neural Processes. Empirically, we show CMANPs achieve state-of-the-art results on popular NP benchmarks while being significantly more memory efficient than prior methods."
                    },
                    {
                        "title": "Constant Memory Attention Block",
                        "abstract": "Modern foundation model architectures rely on attention mechanisms to effectively capture context. However, these methods require linear or quadratic memory in terms of the number of inputs/datapoints, limiting their applicability in low-compute domains. In this work, we propose Constant Memory Attention Block (CMAB), a novel general-purpose attention block that computes its output in constant memory and performs updates in constant computation. Highlighting CMABs efficacy, we introduce methods for Neural Processes and Temporal Point Processes. Empirically, we show our proposed methods achieve results competitive with state-of-the-art while being significantly more memory efficient."
                    },
                    {
                        "title": "Constant Memory Attentive Neural Processes",
                        "abstract": "Neural Processes (NPs) are efficient methods for estimating predictive uncertainties. NPs comprise of a conditioning phase where a context dataset is encoded, a querying phase where the model makes predictions using the context dataset encoding, and an updating phase where the model updates its encoding with newly received datapoints. However, state-of-the-art methods require additional memory which scales linearly or quadratically with the size of the dataset, limiting their applications, particularly in low-resource settings. In this work, we propose Constant Memory Attentive Neural Processes (CMANPs), an NP variant which only requires constant memory for the conditioning, querying, and updating phases. In building CMANPs, we propose Constant Memory Attention Block (CMAB), a novel general-purpose attention block that can compute its output in constant memory and perform updates in constant computation. Empirically, we show CMANPs achieve state-of-the-art results on meta-regression and image completion tasks while being (1) significantly more memory efficient than prior methods and (2) more scalable to harder settings."
                    },
                    {
                        "title": "Stop Overcomplicating Selective Classification: Use Max-Logit",
                        "abstract": "We tackle the problem of Selective Classi\ufb01cation where the goal is to achieve the best performance on the desired coverages of the dataset. Recent state-of-the-art selective methods come with architectural changes either via introducing a separate selection head or an extra abstention logit. In this paper, we present surprising results for Selective Classi\ufb01cation by con\ufb01rming that the superior performance of state-of-the-art methods is owed to training a more generalizable classi\ufb01er; however, their selection mechanism is suboptimal. We argue that the selection mechanism should be rooted in the objective function instead of a separately calculated score. Accordingly, in this paper, we motivate an alternative selection strategy that is based on the cross entropy loss for the classi\ufb01cation settings, namely, max of the logits. Our proposed selection strategy achieves better results by a signi\ufb01cant margin, consistently, across all coverages and all datasets, without any additional computation. Finally, inspired by our superior selection mechanism, we propose to further regularize the objective function with entropy-minimization. Our proposed max-logit selection with the modi\ufb01ed loss function achieves new state-of-the-art results for Selective Classi\ufb01cation."
                    },
                    {
                        "title": "Towards Better Selective Classification",
                        "abstract": "We tackle the problem of Selective Classification where the objective is to achieve the best performance on a predetermined ratio (coverage) of the dataset. Recent state-of-the-art selective methods come with architectural changes either via introducing a separate selection head or an extra abstention logit. In this paper, we challenge the aforementioned methods. The results suggest that the superior performance of state-of-the-art methods is owed to training a more generalizable classifier rather than their proposed selection mechanisms. We argue that the best performing selection mechanism should instead be rooted in the classifier itself. Our proposed selection strategy uses the classification scores and achieves better results by a significant margin, consistently, across all coverages and all datasets, without any added compute cost. Furthermore, inspired by semi-supervised learning, we propose an entropy-based regularizer that improves the performance of selective classification methods. Our proposed selection mechanism with the proposed entropy-based regularizer achieves new state-of-the-art results."
                    },
                    {
                        "title": "Continuous-Time Meta-Learning with Forward Mode Differentiation",
                        "abstract": "Drawing inspiration from gradient-based meta-learning methods with infinitely small gradient steps, we introduce Continuous-Time Meta-Learning (COMLN), a meta-learning algorithm where adaptation follows the dynamics of a gradient vector field. Specifically, representations of the inputs are meta-learned such that a task-specific linear classifier is obtained as a solution of an ordinary differential equation (ODE). Treating the learning process as an ODE offers the notable advantage that the length of the trajectory is now continuous, as opposed to a fixed and discrete number of gradient steps. As a consequence, we can optimize the amount of adaptation necessary to solve a new task using stochastic gradient descent, in addition to learning the initial conditions as is standard practice in gradient-based meta-learning. Importantly, in order to compute the exact meta-gradients required for the outer-loop updates, we devise an efficient algorithm based on forward mode differentiation, whose memory requirements do not scale with the length of the learning trajectory, thus allowing longer adaptation in constant memory. We provide analytical guarantees for the stability of COMLN, we show empirically its efficiency in terms of runtime and memory usage, and we illustrate its effectiveness on a range of few-shot image classification problems."
                    },
                    {
                        "title": "Designing Biological Sequences via Meta-Reinforcement Learning and Bayesian Optimization",
                        "abstract": "The ability to accelerate the design of biological sequences can have a substantial impact on the progress of the medical field. The problem can be framed as a global optimization problem where the objective is an expensive black-box function such that we can query large batches restricted with a limitation of a low number of rounds. Bayesian Optimization is a principled method for tackling this problem. However, the astronomically large state space of biological sequences renders brute-force iterating over all possible sequences infeasible. In this paper, we propose MetaRLBO where we train an autoregressive generative model via Meta-Reinforcement Learning to propose promising sequences for selection via Bayesian Optimization. We pose this problem as that of finding an optimal policy over a distribution of MDPs induced by sampling subsets of the data acquired in the previous rounds. Our in-silico experiments show that meta-learning over such ensembles provides robustness against reward misspecification and achieves competitive results compared to existing strong baselines."
                    },
                    {
                        "title": "Latent Bottlenecked Attentive Neural Processes",
                        "abstract": "Neural Processes (NPs) are popular methods in meta-learning that can estimate predictive uncertainty on target datapoints by conditioning on a context dataset. Previous state-of-the-art method Transformer Neural Processes (TNPs) achieve strong performance but require quadratic computation with respect to the number of context datapoints, significantly limiting its scalability. Conversely, existing sub-quadratic NP variants perform significantly worse than that of TNPs. Tackling this issue, we propose Latent Bottlenecked Attentive Neural Processes (LBANPs), a new computationally efficient sub-quadratic NP variant, that has a querying computational complexity independent of the number of context datapoints. The model encodes the context dataset into a constant number of latent vectors on which self-attention is performed. When making predictions, the model retrieves higher-order information from the context dataset via multiple cross-attention mechanisms on the latent vectors. We empirically show that LBANPs achieve results competitive with the state-of-the-art on meta-regression, image completion, and contextual multi-armed bandits. We demonstrate that LBANPs can trade-off the computational cost and performance according to the number of latent vectors. Finally, we show LBANPs can scale beyond existing attention-based NP variants to larger dataset settings."
                    },
                    {
                        "title": "VariBAD: Variational Bayes-Adaptive Deep RL via Meta-Learning",
                        "abstract": "Trading o\ufb00 exploration and exploitation in an unknown environment is key to maximising expected online return during learning. A Bayes-optimal policy, which does so optimally, conditions its actions not only on the environment state but also on the agent\u2019s uncertainty about the environment. Computing a Bayes-optimal policy is however intractable for all but the smallest tasks. In this paper, we introduce variational Bayes-Adaptive Deep RL (variBAD), a way to meta-learn approximately Bayes-optimal policies for complex tasks. VariBAD simultaneously meta-learns a variational auto-encoder to perform approximate inference"
                    },
                    {
                        "title": "Exploration in Approximate Hyper-State Space for Meta Reinforcement Learning",
                        "abstract": "Meta-learning is a powerful tool for learning policies that can adapt efficiently when deployed in new tasks. If however the meta-training tasks have sparse rewards, the need for exploration during meta-training is exacerbated given that the agent has to explore and learn across many tasks. We show that current meta-learning methods can fail catastrophically in such environments. To address this problem, we propose HyperX, a novel method for meta-learning in sparse reward tasks. Using novel reward bonuses for meta-training, we incentivise the agent to explore in approximate hyper-state space, i.e., the joint state and approximate belief space, where the beliefs are over tasks. We show empirically that these bonuses allow an agent to successfully learn to solve sparse reward tasks where existing meta-learning methods fail."
                    },
                    {
                        "title": "VIABLE: Fast Adaptation via Backpropagating Learned Loss",
                        "abstract": "In few-shot learning, typically, the loss function which is applied at test time is the one we are ultimately interested in minimising, such as the mean-squared-error loss for a regression problem. However, given that we have few samples at test time, we argue that the loss function that we are interested in minimising is not necessarily the loss function most suitable for computing gradients in a few-shot setting. We propose VIABLE, a generic meta-learning extension that builds on existing meta-gradient-based methods by learning a differentiable loss function, replacing the pre-defined inner-loop loss function in performing task-specific updates. We show that learning a loss function capable of leveraging relational information between samples reduces underfitting, and significantly improves performance and sample efficiency on a simple regression task. Furthermore, we show VIABLE is scalable by evaluating on the Mini-Imagenet dataset."
                    }
                ]
            },
            "038a6a91-1e25-408c-aaf7-301a39b38941": {
                "pk": "038a6a91-1e25-408c-aaf7-301a39b38941",
                "name": "Frederick Tung",
                "collaborators": [
                    "Leo Feng",
                    "Mohamed Osama Ahmed",
                    "Y. Bengio",
                    "Hossein Hajimirsadegh",
                    "Hossein Hajimirsadeghi",
                    "Greg Mori",
                    "Sepidehsadat Hosseini",
                    "Mengyao Zhai"
                ],
                "domain": [
                    "Recurrent Neural Networks",
                    "Transformers",
                    "Temporal Domain Generalization",
                    "Neural Processes"
                ],
                "publications": [
                    {
                        "title": "Were RNNs All We Needed?",
                        "abstract": "The scalability limitations of Transformers regarding sequence length have renewed interest in recurrent sequence models that are parallelizable during training. As a result, many novel recurrent architectures, such as S4, Mamba, and Aaren, have been proposed that achieve comparable performance. In this work, we revisit traditional recurrent neural networks (RNNs) from over a decade ago: LSTMs (1997) and GRUs (2014). While these models were slow due to requiring to backpropagate through time (BPTT), we show that by removing their hidden state dependencies from their input, forget, and update gates, LSTMs and GRUs no longer need to BPTT and can be efficiently trained in parallel. Building on this, we introduce minimal versions (minLSTMs and minGRUs) that (1) use significantly fewer parameters than their traditional counterparts and (2) are fully parallelizable during training (175x faster for a sequence of length 512). Lastly, we show that these stripped-down versions of decade-old RNNs match the empirical performance of recent sequence models."
                    },
                    {
                        "title": "Attention as an RNN",
                        "abstract": "The advent of Transformers marked a significant breakthrough in sequence modelling, providing a highly performant architecture capable of leveraging GPU parallelism. However, Transformers are computationally expensive at inference time, limiting their applications, particularly in low-resource settings (e.g., mobile and embedded devices). Addressing this, we (1) begin by showing that attention can be viewed as a special Recurrent Neural Network (RNN) with the ability to compute its \\textit{many-to-one} RNN output efficiently. We then (2) show that popular attention-based models such as Transformers can be viewed as RNN variants. However, unlike traditional RNNs (e.g., LSTMs), these models cannot be updated efficiently with new tokens, an important property in sequence modelling. Tackling this, we (3) introduce a new efficient method of computing attention's \\textit{many-to-many} RNN output based on the parallel prefix scan algorithm. Building on the new attention formulation, we (4) introduce \\textbf{Aaren}, an attention-based module that can not only (i) be trained in parallel (like Transformers) but also (ii) be updated efficiently with new tokens, requiring only constant memory for inferences (like traditional RNNs). Empirically, we show Aarens achieve comparable performance to Transformers on $38$ datasets spread across four popular sequential problem settings: reinforcement learning, event forecasting, time series classification, and time series forecasting tasks while being more time and memory-efficient."
                    },
                    {
                        "title": "Prompting-based Temporal Domain Generalization",
                        "abstract": "Machine learning traditionally assumes that the training and testing data are distributed independently and identically. However, in many real-world settings, the data distribution can shift over time, leading to poor generalization of trained models in future time periods. This paper presents a novel prompting-based approach to temporal domain generalization that is parameter-efficient, time-efficient, and does not require access to future data during training. Our method adapts a trained model to temporal drift by learning global prompts, domain-specific prompts, and drift-aware prompts that capture underlying temporal dynamics. Experiments on classification, regression, and time series forecasting tasks demonstrate the generality of the proposed approach. The code repository will be publicly shared."
                    },
                    {
                        "title": "Constant Memory Attentive Neural Processes",
                        "abstract": "Neural Processes (NPs) are efficient methods for estimating predictive uncertainties. NPs comprise of a conditioning phase where a context dataset is encoded, a querying phase where the model makes predictions using the context dataset encoding, and an updating phase where the model updates its encoding with newly received datapoints. However, state-of-the-art methods require additional memory which scales linearly or quadratically with the size of the dataset, limiting their applications, particularly in low-resource settings. In this work, we propose Constant Memory Attentive Neural Processes (CMANPs), an NP variant which only requires constant memory for the conditioning, querying, and updating phases. In building CMANPs, we propose Constant Memory Attention Block (CMAB), a novel general-purpose attention block that can compute its output in constant memory and perform updates in constant computation. Empirically, we show CMANPs achieve state-of-the-art results on meta-regression and image completion tasks while being (1) significantly more memory efficient than prior methods and (2) more scalable to harder settings."
                    }
                ]
            },
            "209c43c2-5bcc-45ac-beeb-bdc18440cfc2": {
                "pk": "209c43c2-5bcc-45ac-beeb-bdc18440cfc2",
                "name": "Hossein Hajimirsadeghi",
                "collaborators": [
                    "Greg Mori",
                    "Leo Feng",
                    "M. O. Ahmed",
                    "A. Abdi",
                    "Frederick Tung",
                    "Y. Bengio",
                    "Thibaut Durand",
                    "Hamed Shirzad",
                    "Mohamed Osama Ahmed",
                    "Jo\u00e3o Monteiro",
                    "Yu Gong",
                    "Jiawei He",
                    "Saghar Irandoust",
                    "Yunduz Rakhmangulova",
                    "Wenjie Zi",
                    "Yoshua Bengio",
                    "Jiannan Wang",
                    "Lei Chen",
                    "Jianhui Chen",
                    "Shih-Yang Su",
                    "Megha Nawhal"
                ],
                "domain": [
                    "Neural Networks",
                    "Meta-Learning",
                    "Graph Generation",
                    "Attention Mechanisms"
                ],
                "publications": [
                    {
                        "title": "Attention as an RNN",
                        "abstract": "The advent of Transformers marked a significant breakthrough in sequence modelling, providing a highly performant architecture capable of leveraging GPU parallelism. However, Transformers are computationally expensive at inference time, limiting their applications, particularly in low-resource settings (e.g., mobile and embedded devices). Addressing this, we (1) begin by showing that attention can be viewed as a special Recurrent Neural Network (RNN) with the ability to compute its \\textit{many-to-one} RNN output efficiently. We then (2) show that popular attention-based models such as Transformers can be viewed as RNN variants. However, unlike traditional RNNs (e.g., LSTMs), these models cannot be updated efficiently with new tokens, an important property in sequence modelling. Tackling this, we (3) introduce a new efficient method of computing attention's \\textit{many-to-many} RNN output based on the parallel prefix scan algorithm. Building on the new attention formulation, we (4) introduce \\textbf{Aaren}, an attention-based module that can not only (i) be trained in parallel (like Transformers) but also (ii) be updated efficiently with new tokens, requiring only constant memory for inferences (like traditional RNNs). Empirically, we show Aarens achieve comparable performance to Transformers on $38$ datasets spread across four popular sequential problem settings: reinforcement learning, event forecasting, time series classification, and time series forecasting tasks while being more time and memory-efficient."
                    },
                    {
                        "title": "Memory Efficient Neural Processes via Constant Memory Attention Block",
                        "abstract": "Neural Processes (NPs) are popular meta-learning methods for efficiently modelling predictive uncertainty. Recent state-of-the-art methods, however, leverage expensive attention mechanisms, limiting their applications, particularly in low-resource settings. In this work, we propose Constant Memory Attentive Neural Processes (CMANPs), an NP variant that only requires constant memory. To do so, we first propose an efficient update operation for Cross Attention. Leveraging the update operation, we propose Constant Memory Attention Block (CMAB), a novel attention block that (i) is permutation invariant, (ii) computes its output in constant memory, and (iii) performs constant computation updates. Finally, building on CMAB, we detail Constant Memory Attentive Neural Processes. Empirically, we show CMANPs achieve state-of-the-art results on popular NP benchmarks while being significantly more memory efficient than prior methods."
                    },
                    {
                        "title": "Constant Memory Attention Block",
                        "abstract": "Modern foundation model architectures rely on attention mechanisms to effectively capture context. However, these methods require linear or quadratic memory in terms of the number of inputs/datapoints, limiting their applicability in low-compute domains. In this work, we propose Constant Memory Attention Block (CMAB), a novel general-purpose attention block that computes its output in constant memory and performs updates in constant computation. Highlighting CMABs efficacy, we introduce methods for Neural Processes and Temporal Point Processes. Empirically, we show our proposed methods achieve results competitive with state-of-the-art while being significantly more memory efficient."
                    },
                    {
                        "title": "Constant Memory Attentive Neural Processes",
                        "abstract": "Neural Processes (NPs) are efficient methods for estimating predictive uncertainties. NPs comprise of a conditioning phase where a context dataset is encoded, a querying phase where the model makes predictions using the context dataset encoding, and an updating phase where the model updates its encoding with newly received datapoints. However, state-of-the-art methods require additional memory which scales linearly or quadratically with the size of the dataset, limiting their applications, particularly in low-resource settings. In this work, we propose Constant Memory Attentive Neural Processes (CMANPs), an NP variant which only requires constant memory for the conditioning, querying, and updating phases. In building CMANPs, we propose Constant Memory Attention Block (CMAB), a novel general-purpose attention block that can compute its output in constant memory and perform updates in constant computation. Empirically, we show CMANPs achieve state-of-the-art results on meta-regression and image completion tasks while being (1) significantly more memory efficient than prior methods and (2) more scalable to harder settings."
                    },
                    {
                        "title": "Stop Overcomplicating Selective Classification: Use Max-Logit",
                        "abstract": "We tackle the problem of Selective Classi\ufb01cation where the goal is to achieve the best performance on the desired coverages of the dataset. Recent state-of-the-art selective methods come with architectural changes either via introducing a separate selection head or an extra abstention logit. In this paper, we present surprising results for Selective Classi\ufb01cation by con\ufb01rming that the superior performance of state-of-the-art methods is owed to training a more generalizable classi\ufb01er; however, their selection mechanism is suboptimal. We argue that the selection mechanism should be rooted in the objective function instead of a separately calculated score. Accordingly, in this paper, we motivate an alternative selection strategy that is based on the cross entropy loss for the classi\ufb01cation settings, namely, max of the logits. Our proposed selection strategy achieves better results by a signi\ufb01cant margin, consistently, across all coverages and all datasets, without any additional computation. Finally, inspired by our superior selection mechanism, we propose to further regularize the objective function with entropy-minimization. Our proposed max-logit selection with the modi\ufb01ed loss function achieves new state-of-the-art results for Selective Classi\ufb01cation."
                    },
                    {
                        "title": "Towards Better Selective Classification",
                        "abstract": "We tackle the problem of Selective Classification where the objective is to achieve the best performance on a predetermined ratio (coverage) of the dataset. Recent state-of-the-art selective methods come with architectural changes either via introducing a separate selection head or an extra abstention logit. In this paper, we challenge the aforementioned methods. The results suggest that the superior performance of state-of-the-art methods is owed to training a more generalizable classifier rather than their proposed selection mechanisms. We argue that the best performing selection mechanism should instead be rooted in the classifier itself. Our proposed selection strategy uses the classification scores and achieves better results by a significant margin, consistently, across all coverages and all datasets, without any added compute cost. Furthermore, inspired by semi-supervised learning, we propose an entropy-based regularizer that improves the performance of selective classification methods. Our proposed selection mechanism with the proposed entropy-based regularizer achieves new state-of-the-art results."
                    },
                    {
                        "title": "Training a Vision Transformer from scratch in less than 24 hours with 1 GPU",
                        "abstract": "Transformers have become central to recent advances in computer vision. However, training a vision Transformer (ViT) model from scratch can be resource intensive and time consuming. In this paper, we aim to explore approaches to reduce the training costs of ViT models. We introduce some algorithmic improvements to enable training a ViT model from scratch with limited hardware (1 GPU) and time (24 hours) resources. First, we propose an efficient approach to add locality to the ViT architecture. Second, we develop a new image size curriculum learning strategy, which allows to reduce the number of patches extracted from each image at the beginning of the training. Finally, we propose a new variant of the popular ImageNet1k benchmark by adding hardware and time constraints. We evaluate our contributions on this benchmark, and show they can significantly improve performances given the proposed training budget. We will share the code in https://github.com/BorealisAI/efficient-vit-training."
                    },
                    {
                        "title": "Monotonicity Regularization: Improved Penalties and Novel Applications to Disentangled Representation Learning and Robust Classification",
                        "abstract": "We study settings where gradient penalties are used alongside risk minimization with the goal of obtaining predictors satisfying different notions of monotonicity. Specifically, we present two sets of contributions. In the first part of the paper, we show that different choices of penalties define the regions of the input space where the property is observed. As such, previous methods result in models that are monotonic only in a small volume of the input space. We thus propose an approach that uses mixtures of training instances and random points to populate the space and enforce the penalty in a much larger region. As a second set of contributions, we introduce regularization strategies that enforce other notions of monotonicity in different settings. In this case, we consider applications, such as image classification and generative modeling, where monotonicity is not a hard constraint but can help improve some aspects of the model. Namely, we show that inducing monotonicity can be beneficial in applications such as: (1) allowing for controllable data generation, (2) defining strategies to detect anomalous data, and (3) generating explanations for predictions. Our proposed approaches do not introduce relevant computational overhead while leading to efficient procedures that provide extra benefits over baseline models."
                    },
                    {
                        "title": "Latent Bottlenecked Attentive Neural Processes",
                        "abstract": "Neural Processes (NPs) are popular methods in meta-learning that can estimate predictive uncertainty on target datapoints by conditioning on a context dataset. Previous state-of-the-art method Transformer Neural Processes (TNPs) achieve strong performance but require quadratic computation with respect to the number of context datapoints, significantly limiting its scalability. Conversely, existing sub-quadratic NP variants perform significantly worse than that of TNPs. Tackling this issue, we propose Latent Bottlenecked Attentive Neural Processes (LBANPs), a new computationally efficient sub-quadratic NP variant, that has a querying computational complexity independent of the number of context datapoints. The model encodes the context dataset into a constant number of latent vectors on which self-attention is performed. When making predictions, the model retrieves higher-order information from the context dataset via multiple cross-attention mechanisms on the latent vectors. We empirically show that LBANPs achieve results competitive with the state-of-the-art on meta-regression, image completion, and contextual multi-armed bandits. We demonstrate that LBANPs can trade-off the computational cost and performance according to the number of latent vectors. Finally, we show LBANPs can scale beyond existing attention-based NP variants to larger dataset settings."
                    },
                    {
                        "title": "Variational Selective Autoencoder: Learning from Partially-Observed Heterogeneous Data",
                        "abstract": "Learning from heterogeneous data poses challenges such as combining data from various sources and of different types. Meanwhile, heterogeneous data are often associated with missingness in real-world applications due to heterogeneity and noise of input sources. In this work, we propose the variational selective autoencoder (VSAE), a general framework to learn representations from partially-observed heterogeneous data. VSAE learns the latent dependencies in heterogeneous data by modeling the joint distribution of observed data, unobserved data, and the imputation mask which represents how the data are missing. It results in a unified model for various downstream tasks including data generation and imputation. Evaluation on both low-dimensional and high-dimensional heterogeneous datasets for these two tasks shows improvement over state-of-the-art models."
                    },
                    {
                        "title": "Graph Generation Using Tree Decomposition",
                        "abstract": "We propose TD-GEN, a graph generation framework based on tree decomposition, and introduce a reduced upper bound on the maximum number of decisions needed for graph generation. The framework includes a permutation invariant tree generation model which forms the backbone of graph generation. Tree nodes are supernodes, each representing a cluster of nodes in the graph. Graph nodes and edges are incrementally generated inside the clusters by traversing the tree supernodes, respecting the structure of the tree decomposition, and following node sharing decisions between the clusters. Finally, we discuss the shortcomings of standard evaluation criteria based on statistical properties of the generated graphs as performance measures. We propose to compare the performance of models based on likelihood. Empirical results on a variety of standard graph generation datasets demonstrate the superior performance of our method."
                    },
                    {
                        "title": "TD-GEN: Graph Generation Using Tree Decomposition",
                        "abstract": "We propose TD-GEN, a graph generation framework based on tree decomposition, and introduce a reduced upper bound on the maximum number of decisions needed for graph generation. The framework includes a permutation invariant tree generation model which forms the backbone of graph generation. Tree nodes are supernodes, each representing a cluster of nodes in the graph. Graph nodes and edges are incrementally generated inside the clusters by traversing the tree supernodes, respecting the structure of the tree decomposition, and following node sharing decisions between the clusters. Finally, we discuss the shortcomings of standard evaluation criteria based on statistical properties of the generated graphs as performance measures. We propose to compare the performance of models based on likelihood. Empirical results on a variety of standard graph generation datasets demonstrate the superior performance of our method."
                    },
                    {
                        "title": "TD-GEN: Graph Generation With Tree Decomposition",
                        "abstract": "We propose TD-GEN, a graph generation framework based on tree decomposition, and introduce a reduced upper bound on the maximum number of decisions needed for graph generation. The framework includes a permutation invariant tree generation model which forms the backbone of graph generation. Tree nodes are supernodes, each representing a cluster of nodes in the graph. Graph nodes and edges are incrementally generated inside the clusters by traversing the tree supernodes, respecting the structure of the tree decomposition, and following node sharing decisions between the clusters. Finally, we discuss the shortcomings of standard evaluation criteria based on statistical properties of the generated graphs as performance measures. We propose to compare the performance of models based on likelihood. Empirical results on a variety of standard graph generation datasets demonstrate the superior performance of our method."
                    },
                    {
                        "title": "Not Too Close and Not Too Far: Enforcing Monotonicity Requires Penalizing The Right Points",
                        "abstract": "In this work, we propose a practical scheme to enforce monotonicity in neural networks with respect to a given subset of the dimensions of the input space. The proposed approach focuses on the setting where point-wise gradient penalties are used as a soft constraint alongside the empirical risk during training. Our results indicate that the choice of the points employed for computing such a penalty de\ufb01nes the regions of the input space where the desired property is satis\ufb01ed. As such, previous methods result in models that are monotonic either only at the boundaries of the input space or in the small volume where training data lies. Given this, we propose an alternative approach that uses pairs of training instances and random points to create mixtures of points that lie inside and outside of the convex hull of the training sample. Empirical evaluation carried out using different datasets show that the proposed approach yields predictors that are monotonic in a larger volume of the space compared to previous methods. Our approach does not introduce relevant computational overhead leading to an ef\ufb01cient procedure that consistently achieves the best performance amongst all alternatives."
                    },
                    {
                        "title": "Adapting Grad-CAM for Embedding Networks",
                        "abstract": "The gradient-weighted class activation mapping (Grad-CAM) method can faithfully highlight important regions in images for deep model prediction in image classification, image captioning and many other tasks. It uses the gradients in back-propagation as weights (grad-weights) to explain network decisions. However, applying Grad-CAM to embedding networks raises significant challenges because embedding networks are trained by millions of dynamically paired examples (e.g. triplets). To overcome these challenges, we propose an adaptation of the Grad-CAM method for embedding networks. First, we aggregate grad-weights from multiple training examples to improve the stability of Grad-CAM. Then, we develop an efficient weight-transfer method to explain decisions for any image without back-propagation. We extensively validate the method on the standard CUB200 dataset in which our method produces more accurate visual attention than the original Grad-CAM method. We also apply the method to a house price estimation application using images. The method produces convincing qualitative results, showcasing the practicality of our approach."
                    },
                    {
                        "title": "Graph Generation with Variational Recurrent Neural Network",
                        "abstract": "Generating graph structures is a challenging problem due to the diverse representations and complex dependencies among nodes. In this paper, we introduce Graph Variational Recurrent Neural Network (GraphVRNN), a probabilistic autoregressive model for graph generation. Through modeling the latent variables of graph data, GraphVRNN can capture the joint distributions of graph structures and the underlying node attributes. We conduct experiments on the proposed GraphVRNN in both graph structure learning and attribute generation tasks. The evaluation results show that the variational component allows our network to model complicated distributions, as well as generate plausible structures and node attributes."
                    },
                    {
                        "title": "Variational Selective Autoencoder",
                        "abstract": "Despite promising progress on unimodal data imputation (e.g. image inpainting), models for multimodal data imputation are far from satisfactory. In this work, we propose variational selective autoencoder (VSAE) for this task. Learning only from partially-observed data, VSAE can model the joint distribution of observed/unobserved modalities and the imputation mask, resulting in a unified model for various down-stream tasks including data generation and imputation. Evaluation on synthetic high-dimensional and challenging low-dimensional multimodal datasets shows improvement over the state-of-the-art imputation models."
                    },
                    {
                        "title": "Joint Kernel Learning and Multi-Instance Classification",
                        "abstract": "This paper presents a structured kernel learning approach to multi-instance classification, where the bags are modeled as hidden conditional random fields (hCRFs) with cardinality potentials. A kernel, parameterized by the corresponding hCRF cardinality model, is jointly trained with the max-margin classifier in a regularized risk optimization problem. The integrated learning forms the instance-level labeling and the bag-level discriminative classification together, with a direct goal of improved bag classification The proposed method is evaluated on standard multiinstance learning data sets."
                    }
                ]
            },
            "46998bef-4c2f-4b0b-b426-b8a0201ba7c4": {
                "pk": "46998bef-4c2f-4b0b-b426-b8a0201ba7c4",
                "name": "Yoshua Bengio",
                "collaborators": [
                    "Nikolay Malkin",
                    "Michal Koziarski",
                    "Moksh Jain",
                    "Alexander Tong",
                    "Cheng-Hao Liu",
                    "Emmanuel Bengio",
                    "Thomas Jiralerspong",
                    "Xiaoyin Chen",
                    "Vedant Shah",
                    "Aniket Didolkar",
                    "Anirudh Goyal",
                    "M. Mozer",
                    "Seanie Lee",
                    "Minsu Kim",
                    "Juho Lee",
                    "Sung Ju Hwang",
                    "A. V. D. Sloot",
                    "Michael K. Cohen",
                    "Matt MacDermott",
                    "Damiano Fornasiere",
                    "Pietro Greiner",
                    "Younesse Kaddar",
                    "Xi Zhang",
                    "Yuan Pu",
                    "Yuki Kawamura",
                    "Andrew Loza",
                    "Dennis L. Shung",
                    "Nasim Rahaman",
                    "Martin Weiss",
                    "Manuel W\u00fcthrich",
                    "Erran L. Li",
                    "C. Pal",
                    "Bernhard Sch\u00f6lkopf",
                    "Jarrid Rector-Brooks",
                    "Mohsin Hasan",
                    "Zhangzhi Peng",
                    "Zachary Quinn",
                    "Sarthak Mittal",
                    "Nouha Dziri",
                    "Michael M. Bronstein",
                    "Pranam Chatterjee",
                    "A. Bose",
                    "Oussama Boussif",
                    "L'ena N'ehale Ezzine",
                    "Joseph D. Viviano",
                    "Rim Assouel",
                    "Yash More",
                    "George Ma",
                    "Dinghuai Zhang",
                    "Andrii Zadaianchuk",
                    "G. Martius",
                    "Maximilian Seitzer",
                    "Girish Sastry",
                    "Lennart Heim",
                    "Haydn Belfield",
                    "Markus Anderljung",
                    "Miles Brundage",
                    "Julian Hazell",
                    "Cullen O'Keefe",
                    "Gillian K. Hadfield",
                    "Richard Ngo",
                    "Konstantin Pilz",
                    "George Gor",
                    "Emma Bluemke",
                    "Sarah Shoker",
                    "Janet Egan",
                    "Robert Trager",
                    "S. Avin",
                    "Adrian Weller",
                    "Diane Coyle",
                    "Lynn Cherif",
                    "David Dobre",
                    "Kenji Kawaguchi",
                    "Gauthier Gidel",
                    "Andrei Rekesh",
                    "Dmytro Shevchuk",
                    "Piotr Gai'nski",
                    "Mike Tyers",
                    "Robert A. Batey",
                    "Anas Krichel",
                    "Salem Lahlou",
                    "Mohammed Abukalam",
                    "Louis Vaillancourt",
                    "Doris Alexandra Schuetz",
                    "Mathieu Bourgey",
                    "A. Marinier",
                    "Jin Hwa Lee",
                    "Lei Yu",
                    "Emily Cheng",
                    "Haebin Seong",
                    "Dong Bok Lee",
                    "Minki Kang",
                    "Dominik Wagner",
                    "Antoine Bellemare-Pepin",
                    "Franccois Lespinasse",
                    "Philipp Tholke",
                    "Y. Harel",
                    "K. Mathewson",
                    "Jay A. Olson",
                    "Karim Jerbi CoCo Lab"
                ],
                "domain": [
                    "Machine Learning",
                    "Generative Models",
                    "Reinforcement Learning",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "Can a Bayesian Oracle Prevent Harm from an Agent?",
                        "abstract": "Is there a way to design powerful AI systems based on machine learning methods that would satisfy probabilistic safety guarantees? With the long-term goal of obtaining a probabilistic guarantee that would apply in every context, we consider estimating a context-dependent bound on the probability of violating a given safety specification. Such a risk evaluation would need to be performed at run-time to provide a guardrail against dangerous actions of an AI. Noting that different plausible hypotheses about the world could produce very different outcomes, and because we do not know which one is right, we derive bounds on the safety violation probability predicted under the true but unknown hypothesis. Such bounds could be used to reject potentially dangerous actions. Our main results involve searching for cautious but plausible hypotheses, obtained by a maximization that involves Bayesian posteriors over hypotheses. We consider two forms of this result, in the iid case and in the non-iid case, and conclude with open problems towards turning such theoretical results into practical AI guardrails."
                    },
                    {
                        "title": "Trajectory Flow Matching with Applications to Clinical Time Series Modeling",
                        "abstract": "Modeling stochastic and irregularly sampled time series is a challenging problem found in a wide range of applications, especially in medicine. Neural stochastic differential equations (Neural SDEs) are an attractive modeling technique for this problem, which parameterize the drift and diffusion terms of an SDE with neural networks. However, current algorithms for training Neural SDEs require backpropagation through the SDE dynamics, greatly limiting their scalability and stability. To address this, we propose Trajectory Flow Matching (TFM), which trains a Neural SDE in a simulation-free manner, bypassing backpropagation through the dynamics. TFM leverages the flow matching technique from generative modeling to model time series. In this work we first establish necessary conditions for TFM to learn time series data. Next, we present a reparameterization trick which improves training stability. Finally, we adapt TFM to the clinical time series setting, demonstrating improved performance on three clinical time series datasets both in terms of absolute performance and uncertainty prediction."
                    },
                    {
                        "title": "Language Models Can Reduce Asymmetry in Information Markets",
                        "abstract": "This work addresses the buyer's inspection paradox for information markets. The paradox is that buyers need to access information to determine its value, while sellers need to limit access to prevent theft. To study this, we introduce an open-source simulated digital marketplace where intelligent agents, powered by language models, buy and sell information on behalf of external participants. The central mechanism enabling this marketplace is the agents' dual capabilities: they not only have the capacity to assess the quality of privileged information but also come equipped with the ability to forget. This ability to induce amnesia allows vendors to grant temporary access to proprietary information, significantly reducing the risk of unauthorized retention while enabling agents to accurately gauge the information's relevance to specific queries or tasks. To perform well, agents must make rational decisions, strategically explore the marketplace through generated sub-queries, and synthesize answers from purchased information. Concretely, our experiments (a) uncover biases in language models leading to irrational behavior and evaluate techniques to mitigate these biases, (b) investigate how price affects demand in the context of informational goods, and (c) show that inspection and higher budgets both lead to higher quality outcomes."
                    },
                    {
                        "title": "Steering Masked Discrete Diffusion Models via Discrete Denoising Posterior Prediction",
                        "abstract": "Generative modeling of discrete data underlies important applications spanning text-based agents like ChatGPT to the design of the very building blocks of life in protein sequences. However, application domains need to exert control over the generated data by steering the generative process - typically via RLHF - to satisfy a specified property, reward, or affinity metric. In this paper, we study the problem of steering Masked Diffusion Models (MDMs), a recent class of discrete diffusion models that offer a compelling alternative to traditional autoregressive models. We introduce Discrete Denoising Posterior Prediction (DDPP), a novel framework that casts the task of steering pre-trained MDMs as a problem of probabilistic inference by learning to sample from a target Bayesian posterior. Our DDPP framework leads to a family of three novel objectives that are all simulation-free, and thus scalable while applying to general non-differentiable reward functions. Empirically, we instantiate DDPP by steering MDMs to perform class-conditional pixel-level image modeling, RLHF-based alignment of MDMs using text-based rewards, and finetuning protein language models to generate more diverse secondary structures and shorter proteins. We substantiate our designs via wet-lab validation, where we observe transient expression of reward-optimized protein sequences."
                    },
                    {
                        "title": "Action abstractions for amortized sampling",
                        "abstract": "As trajectories sampled by policies used by reinforcement learning (RL) and generative flow networks (GFlowNets) grow longer, credit assignment and exploration become more challenging, and the long planning horizon hinders mode discovery and generalization. The challenge is particularly pronounced in entropy-seeking RL methods, such as generative flow networks, where the agent must learn to sample from a structured distribution and discover multiple high-reward states, each of which take many steps to reach. To tackle this challenge, we propose an approach to incorporate the discovery of action abstractions, or high-level actions, into the policy optimization process. Our approach involves iteratively extracting action subsequences commonly used across many high-reward trajectories and `chunking' them into a single action that is added to the action space. In empirical evaluation on synthetic and real-world environments, our approach demonstrates improved sample efficiency performance in discovering diverse high-reward objects, especially on harder exploration problems. We also observe that the abstracted high-order actions are interpretable, capturing the latent structure of the reward landscape of the action space. This work provides a cognitively motivated approach to action abstraction in RL and is the first demonstration of hierarchical planning in amortized sequential sampling."
                    },
                    {
                        "title": "Efficient Causal Graph Discovery Using Large Language Models",
                        "abstract": "We propose a novel framework that leverages LLMs for full causal graph discovery. While previous LLM-based methods have used a pairwise query approach, this requires a quadratic number of queries which quickly becomes impractical for larger causal graphs. In contrast, the proposed framework uses a breadth-first search (BFS) approach which allows it to use only a linear number of queries. We also show that the proposed method can easily incorporate observational data when available, to improve performance. In addition to being more time and data-efficient, the proposed framework achieves state-of-the-art results on real-world causal graphs of varying sizes. The results demonstrate the effectiveness and efficiency of the proposed method in discovering causal relationships, showcasing its potential for broad applicability in causal graph discovery tasks across different domains."
                    },
                    {
                        "title": "Baking Symmetry into GFlowNets",
                        "abstract": "GFlowNets have exhibited promising performance in generating diverse candidates with high rewards. These networks generate objects incrementally and aim to learn a policy that assigns probability of sampling objects in proportion to rewards. However, the current training pipelines of GFlowNets do not consider the presence of isomorphic actions, which are actions resulting in symmetric or isomorphic states. This lack of symmetry increases the amount of samples required for training GFlowNets and can result in inefficient and potentially incorrect flow functions. As a consequence, the reward and diversity of the generated objects decrease. In this study, our objective is to integrate symmetries into GFlowNets by identifying equivalent actions during the generation process. Experimental results using synthetic data demonstrate the promising performance of our proposed approaches."
                    },
                    {
                        "title": "Zero-Shot Object-Centric Representation Learning",
                        "abstract": "The goal of object-centric representation learning is to decompose visual scenes into a structured representation that isolates the entities. Recent successes have shown that object-centric representation learning can be scaled to real-world scenes by utilizing pre-trained self-supervised features. However, so far, object-centric methods have mostly been applied in-distribution, with models trained and evaluated on the same dataset. This is in contrast to the wider trend in machine learning towards general-purpose models directly applicable to unseen data and tasks. Thus, in this work, we study current object-centric methods through the lens of zero-shot generalization by introducing a benchmark comprising eight different synthetic and real-world datasets. We analyze the factors influencing zero-shot performance and find that training on diverse real-world images improves transferability to unseen scenarios. Furthermore, inspired by the success of task-specific fine-tuning in foundation models, we introduce a novel fine-tuning strategy to adapt pre-trained vision encoders for the task of object discovery. We find that the proposed approach results in state-of-the-art performance for unsupervised object discovery, exhibiting strong zero-shot transfer to unseen datasets."
                    },
                    {
                        "title": "Computing Power and the Governance of Artificial Intelligence",
                        "abstract": "Computing power, or\"compute,\"is crucial for the development and deployment of artificial intelligence (AI) capabilities. As a result, governments and companies have started to leverage compute as a means to govern AI. For example, governments are investing in domestic compute capacity, controlling the flow of compute to competing countries, and subsidizing compute access to certain sectors. However, these efforts only scratch the surface of how compute can be used to govern AI development and deployment. Relative to other key inputs to AI (data and algorithms), AI-relevant compute is a particularly effective point of intervention: it is detectable, excludable, and quantifiable, and is produced via an extremely concentrated supply chain. These characteristics, alongside the singular importance of compute for cutting-edge AI models, suggest that governing compute can contribute to achieving common policy objectives, such as ensuring the safety and beneficial use of AI. More precisely, policymakers could use compute to facilitate regulatory visibility of AI, allocate resources to promote beneficial outcomes, and enforce restrictions against irresponsible or malicious AI development and usage. However, while compute-based policies and technologies have the potential to assist in these areas, there is significant variation in their readiness for implementation. Some ideas are currently being piloted, while others are hindered by the need for fundamental research. Furthermore, naive or poorly scoped approaches to compute governance carry significant risks in areas like privacy, economic impacts, and centralization of power. We end by suggesting guardrails to minimize these risks from compute governance."
                    },
                    {
                        "title": "Learning diverse attacks on large language models for robust red-teaming and safety tuning",
                        "abstract": "Red-teaming, or identifying prompts that elicit harmful responses, is a critical step in ensuring the safe and responsible deployment of large language models (LLMs). Developing effective protection against many modes of attack prompts requires discovering diverse attacks. Automated red-teaming typically uses reinforcement learning to fine-tune an attacker language model to generate prompts that elicit undesirable responses from a target LLM, as measured, for example, by an auxiliary toxicity classifier. We show that even with explicit regularization to favor novelty and diversity, existing approaches suffer from mode collapse or fail to generate effective attacks. As a flexible and probabilistically principled alternative, we propose to use GFlowNet fine-tuning, followed by a secondary smoothing phase, to train the attacker model to generate diverse and effective attack prompts. We find that the attacks generated by our method are effective against a wide range of target LLMs, both with and without safety tuning, and transfer well between target LLMs. Finally, we demonstrate that models safety-tuned using a dataset of red-teaming prompts generated by our method are robust to attacks from other RL-based red-teaming approaches."
                    },
                    {
                        "title": "RGFN: Synthesizable Molecular Generation Using GFlowNets",
                        "abstract": "Generative models hold great promise for small molecule discovery, significantly increasing the size of search space compared to traditional in silico screening libraries. However, most existing machine learning methods for small molecule generation suffer from poor synthesizability of candidate compounds, making experimental validation difficult. In this paper we propose Reaction-GFlowNet (RGFN), an extension of the GFlowNet framework that operates directly in the space of chemical reactions, thereby allowing out-of-the-box synthesizability while maintaining comparable quality of generated candidates. We demonstrate that with the proposed set of reactions and building blocks, it is possible to obtain a search space of molecules orders of magnitude larger than existing screening libraries coupled with low cost of synthesis. We also show that the approach scales to very large fragment libraries, further increasing the number of potential molecules. We demonstrate the effectiveness of the proposed approach across a range of oracle models, including pretrained proxy models and GPU-accelerated docking."
                    },
                    {
                        "title": "On Generalization for Generative Flow Networks",
                        "abstract": "Generative Flow Networks (GFlowNets) have emerged as an innovative learning paradigm designed to address the challenge of sampling from an unnormalized probability distribution, called the reward function. This framework learns a policy on a constructed graph, which enables sampling from an approximation of the target probability distribution through successive steps of sampling from the learned policy. To achieve this, GFlowNets can be trained with various objectives, each of which can lead to the model s ultimate goal. The aspirational strength of GFlowNets lies in their potential to discern intricate patterns within the reward function and their capacity to generalize effectively to novel, unseen parts of the reward function. This paper attempts to formalize generalization in the context of GFlowNets, to link generalization with stability, and also to design experiments that assess the capacity of these models to uncover unseen parts of the reward function. The experiments will focus on length generalization meaning generalization to states that can be constructed only by longer trajectories than those seen in training."
                    },
                    {
                        "title": "Machine learning and information theory concepts towards an AI Mathematician",
                        "abstract": "The current state of the art in artificial intelligence is impressive, especially in terms of mastery of language, but not so much in terms of mathematical reasoning. What could be missing? Can we learn something useful about that gap from how the brains of mathematicians go about their craft? This essay builds on the idea that current deep learning mostly succeeds at system 1 abilities\u2014which correspond to our intuition and habitual behaviors\u2014but still lacks something important regarding system 2 abilities\u2014which include reasoning and robust uncertainty estimation. It takes an information-theoretical posture to ask questions about what constitutes an interesting mathematical statement, which could guide future work in crafting an AI mathematician. The focus is not on proving a given theorem but on discovering new and interesting conjectures. The central hypothesis is that a desirable body of theorems better summarizes the set of all provable statements, for example, by having a small description length while at the same time being close (in terms of number of derivation steps) to many provable statements."
                    },
                    {
                        "title": "Towards DNA-Encoded Library Generation with GFlowNets",
                        "abstract": "DNA-encoded libraries (DELs) are a powerful approach for rapidly screening large numbers of diverse compounds. One of the key challenges in using DELs is library design, which involves choosing the building blocks that will be combinatorially combined to produce the final library. In this paper we consider the task of protein-protein interaction (PPI) biased DEL design. To this end, we evaluate several machine learning algorithms on the PPI modulation task and use them as a reward for the proposed GFlowNet-based generative approach. We additionally investigate the possibility of using structural information about building blocks to design a hierarchical action space for the GFlowNet. The observed results indicate that GFlowNets are a promising approach for generating diverse combinatorial library candidates."
                    },
                    {
                        "title": "Geometric Signatures of Compositionality Across a Language Model's Lifetime",
                        "abstract": "Compositionality, the notion that the meaning of an expression is constructed from the meaning of its parts and syntactic rules, permits the infinite productivity of human language. For the first time, artificial language models (LMs) are able to match human performance in a number of compositional generalization tasks. However, much remains to be understood about the representational mechanisms underlying these abilities. We take a high-level geometric approach to this problem by relating the degree of compositionality in a dataset to the intrinsic dimensionality of its representations under an LM, a measure of feature complexity. We find not only that the degree of dataset compositionality is reflected in representations' intrinsic dimensionality, but that the relationship between compositionality and geometric complexity arises due to learned linguistic features over training. Finally, our analyses reveal a striking contrast between linear and nonlinear dimensionality, showing that they respectively encode formal and semantic aspects of linguistic composition."
                    },
                    {
                        "title": "HarmAug: Effective Data Augmentation for Knowledge Distillation of Safety Guard Models",
                        "abstract": "Safety guard models that detect malicious queries aimed at large language models (LLMs) are essential for ensuring the secure and responsible deployment of LLMs in real-world applications. However, deploying existing safety guard models with billions of parameters alongside LLMs on mobile devices is impractical due to substantial memory requirements and latency. To reduce this cost, we distill a large teacher safety guard model into a smaller one using a labeled dataset of instruction-response pairs with binary harmfulness labels. Due to the limited diversity of harmful instructions in the existing labeled dataset, naively distilled models tend to underperform compared to larger models. To bridge the gap between small and large models, we propose HarmAug, a simple yet effective data augmentation method that involves jailbreaking an LLM and prompting it to generate harmful instructions. Given a prompt such as,\"Make a single harmful instruction prompt that would elicit offensive content\", we add an affirmative prefix (e.g.,\"I have an idea for a prompt:\") to the LLM's response. This encourages the LLM to continue generating the rest of the response, leading to sampling harmful instructions. Another LLM generates a response to the harmful instruction, and the teacher model labels the instruction-response pair. We empirically show that our HarmAug outperforms other relevant baselines. Moreover, a 435-million-parameter safety guard model trained with HarmAug achieves an F1 score comparable to larger models with over 7 billion parameters, and even outperforms them in AUPRC, while operating at less than 25% of their computational cost."
                    },
                    {
                        "title": "Divergent Creativity in Humans and Large Language Models",
                        "abstract": "The recent surge in the capabilities of Large Language Models (LLMs) has led to claims that they are approaching a level of creativity akin to human capabilities. This idea has sparked a blend of excitement and apprehension. However, a critical piece that has been missing in this discourse is a systematic evaluation of LLM creativity, particularly in comparison to human divergent thinking. To bridge this gap, we leverage recent advances in creativity science to build a framework for in-depth analysis of divergent creativity in both state-of-the-art LLMs and a substantial dataset of 100,000 humans. We found evidence suggesting that LLMs can indeed surpass human capabilities in specific creative tasks such as divergent association and creative writing. Our quantitative benchmarking framework opens up new paths for the development of more creative LLMs, but it also encourages more granular inquiries into the distinctive elements that constitute human inventive thought processes, compared to those that can be artificially generated."
                    },
                    {
                        "title": "Metacognitive Capabilities of LLMs: An Exploration in Mathematical Problem Solving",
                        "abstract": "Metacognitive knowledge refers to humans' intuitive knowledge of their own thinking and reasoning processes. Today's best LLMs clearly possess some reasoning processes. The paper gives evidence that they also have metacognitive knowledge, including ability to name skills and procedures to apply given a task. We explore this primarily in context of math reasoning, developing a prompt-guided interaction procedure to get a powerful LLM to assign sensible skill labels to math questions, followed by having it perform semantic clustering to obtain coarser families of skill labels. These coarse skill labels look interpretable to humans. To validate that these skill labels are meaningful and relevant to the LLM's reasoning processes we perform the following experiments. (a) We ask GPT-4 to assign skill labels to training questions in math datasets GSM8K and MATH. (b) When using an LLM to solve the test questions, we present it with the full list of skill labels and ask it to identify the skill needed. Then it is presented with randomly selected exemplar solved questions associated with that skill label. This improves accuracy on GSM8k and MATH for several strong LLMs, including code-assisted models. The methodology presented is domain-agnostic, even though this article applies it to math problems."
                    },
                    {
                        "title": "Ant Colony Sampling with GFlowNets for Combinatorial Optimization",
                        "abstract": "We present the Generative Flow Ant Colony Sampler (GFACS), a novel meta-heuristic method that hierarchically combines amortized inference and parallel stochastic search. Our method first leverages Generative Flow Networks (GFlowNets) to amortize a multi-modal prior distribution over combinatorial solution space that encompasses both high-reward and diversified solutions. This prior is iteratively updated via parallel stochastic search in the spirit of Ant Colony Optimization (ACO), leading to the posterior distribution that generates near-optimal solutions. Extensive experiments across seven combinatorial optimization problems demonstrate GFACS's promising performances."
                    }
                ]
            },
            "1256b9cc-9ae3-49b3-bc49-ebc9e99917fb": {
                "pk": "1256b9cc-9ae3-49b3-bc49-ebc9e99917fb",
                "name": "Mohamed Osama Ahmed",
                "collaborators": [
                    "Leo Feng",
                    "Frederick Tung",
                    "Y. Bengio",
                    "Hossein Hajimirsadeghi",
                    "Hossein Hajimirsadegh",
                    "Greg Mori"
                ],
                "domain": [
                    "Recurrent Neural Networks",
                    "Attention Mechanisms",
                    "Neural Processes",
                    "Efficient Learning"
                ],
                "publications": [
                    {
                        "title": "Were RNNs All We Needed?",
                        "abstract": "The scalability limitations of Transformers regarding sequence length have renewed interest in recurrent sequence models that are parallelizable during training. As a result, many novel recurrent architectures, such as S4, Mamba, and Aaren, have been proposed that achieve comparable performance. In this work, we revisit traditional recurrent neural networks (RNNs) from over a decade ago: LSTMs (1997) and GRUs (2014). While these models were slow due to requiring to backpropagate through time (BPTT), we show that by removing their hidden state dependencies from their input, forget, and update gates, LSTMs and GRUs no longer need to BPTT and can be efficiently trained in parallel. Building on this, we introduce minimal versions (minLSTMs and minGRUs) that (1) use significantly fewer parameters than their traditional counterparts and (2) are fully parallelizable during training (175x faster for a sequence of length 512). Lastly, we show that these stripped-down versions of decade-old RNNs match the empirical performance of recent sequence models."
                    },
                    {
                        "title": "Attention as an RNN",
                        "abstract": "The advent of Transformers marked a significant breakthrough in sequence modelling, providing a highly performant architecture capable of leveraging GPU parallelism. However, Transformers are computationally expensive at inference time, limiting their applications, particularly in low-resource settings (e.g., mobile and embedded devices). Addressing this, we (1) begin by showing that attention can be viewed as a special Recurrent Neural Network (RNN) with the ability to compute its \\textit{many-to-one} RNN output efficiently. We then (2) show that popular attention-based models such as Transformers can be viewed as RNN variants. However, unlike traditional RNNs (e.g., LSTMs), these models cannot be updated efficiently with new tokens, an important property in sequence modelling. Tackling this, we (3) introduce a new efficient method of computing attention's \\textit{many-to-many} RNN output based on the parallel prefix scan algorithm. Building on the new attention formulation, we (4) introduce \\textbf{Aaren}, an attention-based module that can not only (i) be trained in parallel (like Transformers) but also (ii) be updated efficiently with new tokens, requiring only constant memory for inferences (like traditional RNNs). Empirically, we show Aarens achieve comparable performance to Transformers on $38$ datasets spread across four popular sequential problem settings: reinforcement learning, event forecasting, time series classification, and time series forecasting tasks while being more time and memory-efficient."
                    },
                    {
                        "title": "Constant Memory Attentive Neural Processes",
                        "abstract": "Neural Processes (NPs) are efficient methods for estimating predictive uncertainties. NPs comprise of a conditioning phase where a context dataset is encoded, a querying phase where the model makes predictions using the context dataset encoding, and an updating phase where the model updates its encoding with newly received datapoints. However, state-of-the-art methods require additional memory which scales linearly or quadratically with the size of the dataset, limiting their applications, particularly in low-resource settings. In this work, we propose Constant Memory Attentive Neural Processes (CMANPs), an NP variant which only requires constant memory for the conditioning, querying, and updating phases. In building CMANPs, we propose Constant Memory Attention Block (CMAB), a novel general-purpose attention block that can compute its output in constant memory and perform updates in constant computation. Empirically, we show CMANPs achieve state-of-the-art results on meta-regression and image completion tasks while being (1) significantly more memory efficient than prior methods and (2) more scalable to harder settings."
                    }
                ]
            }
        }
    },
    "2410.05550": {
        "paper_data": {
            "title": "Aggregating Quantitative Relative Judgments: From Social Choice to Ranking Prediction",
            "url": "http://arxiv.org/abs/2410.05550v1",
            "arxiv_id": "2410.05550",
            "authors": [
                "Yixuan Even Xu",
                "Hanrui Zhang",
                "Yu Cheng",
                "Vincent Conitzer"
            ],
            "abstract": "Quantitative Relative Judgment Aggregation (QRJA) is a new research topic in (computational) social choice. In the QRJA model, agents provide judgments on the relative quality of different candidates, and the goal is to aggregate these judgments across all agents. In this work, our main conceptual contribution is to explore the interplay between QRJA in a social choice context and its application to ranking prediction. We observe that in QRJA, judges do not have to be people with subjective opinions; for example, a race can be viewed as a \"judgment\" on the contestants' relative abilities. This allows us to aggregate results from multiple races to evaluate the contestants' true qualities. At a technical level, we introduce new aggregation rules for QRJA and study their structural and computational properties. We evaluate the proposed methods on data from various real races and show that QRJA-based methods offer effective and interpretable ranking predictions.",
            "introduction": "   1 Introduction  In voting theory, each voter ranks a set of candidates, and a voting rule maps the vector of rankings to either a winning candidate or an aggregate ranking of all the candidates. There has been significant interaction between computer scientists interested in voting theory and the learning-to-rank community. The learning-to-rank community is interested in problems such as ranking webpages in response to a search query, or ranking recommendations to a user (see, e.g.,\u00a0Liu (2009)). Another problem of interest is to aggregate multiple rankings into a single one, for example combining the ranking results from different algorithms (\u201cvoters\u201d) into a single meta-ranking. While the interests of the communities may differ, e.g., the learning-to-rank community is less concerned about strategic aspects of voting, a natural intersection point for these two communities is a model where there is a latent \u201ctrue\u201d ranking of the candidates, of which all the votes are just noisy observations. Consequently, it is natural to try to estimate the true ranking based on the received rankings, and such an estimation procedure corresponds to a voting rule. (See, e.g.,\u00a0Young (1995); Conitzer and Sandholm (2005); Meila et al. (2007); Conitzer et al. (2009); Caragiannis et al. (2013); Soufiani et al. (2014); Xia (2016), and\u00a0Elkind and Slinko (2015) for an overview.)   Voting rules are just one type of mechanism in the broader field of social choice, which studies the broader problem of making decisions based on the opinions and preferences of multiple agents. Such opinions are not necessarily represented as rankings. For example, in judgment aggregation (see\u00a0Endriss (2015) for an overview), judges assess whether certain propositions are true or false, and the goal is to aggregate these judgments into logically consistent statements. The observation that other types of input are aggregated in social choice prompts the natural question of whether analogous problems exist in statistics and machine learning (as is the case with ranking aggregation).   In this paper, we focus on a relatively new model in social choice, the quantitative judgment aggregation problem\u00a0(Conitzer et al., 2015, 2016). In this problem, the goal is to aggregate relative quantitative judgments: for example, one agent may value the life of a 20-year-old at 2 times the life of a 50-year-old (say in the context of self-driving cars making decisions)\u00a0(Noothigattu et al., 2018); another example could be that an agent judges that \u201cusing 1 unit of gasoline is as bad as creating 3 units of landfill trash\u201d (in a societal tradeoff context)\u00a0(Conitzer et al., 2016). Quantitative judgment aggregation has been considered in the area of automated moral decision-making, where an AI system may choose a course of action based on data about human judgments in similar scenarios.   An important conceptual difference between this work and previous studies on quantitative judgment aggregation is that we observe that relative \u201cjudgments\u201d can be produced by a process other than a subjective agent reporting them, which is the standard assumption in social choice. To illustrate, consider a race in which contestant A finishes at 20:00 and contestant B at 30:00. In this race, the \u201cjudgment\u201d is that A is 10:00 faster than B. This key observation allows us to bring the social",
            "references": [
                {
                    "title": "Negative-Weight Single-Source Shortest Paths in Near-linear Time",
                    "abstract": "We present a randomized algorithm that computes single-source shortest paths (SSSP) in $O\\left(m \\log ^{8}(n) \\log W\\right)$ time when edge weights are integral and can be negative. 1 This essentially resolves the classic negative-weight SSSP problem. The previous bounds are $\\tilde{O}\\left(\\left(m+n^{1.5}\\right) \\log W\\right)$ [BLNPSSSW FOCS\u201920] and $m^{4 / 3+o(1)} \\log W$ [AMV FOCS\u201920]. Near-linear time algorithms were known previously only for the special case of planar directed graphs [Fakcharoenphol and Rao FOCS\u201901]. In contrast to all recent developments that rely on sophisticated continuous optimization methods and dynamic algorithms, our algorithm is simple: it requires only a simple graph decomposition and elementary combinatorial tools. In fact, ours is the first combinatorial algorithm for negative-weight SSSP to break through the classic $O(m \\sqrt{n} \\log W)$ bound from over three decades ago [Gabow and Tarjan SICOMP\u201989]."
                },
                {
                    "title": "Maximum Flow and Minimum-Cost Flow in Almost-Linear Time",
                    "abstract": "We give an algorithm that computes exact maximum flows and minimum-cost flows on directed graphs with m edges and polynomially bounded integral demands, costs, and capacities in $m^{1+o(1)}$ time. Our algorithm builds the flow through a sequence of $m^{1+o(1)}$ approximate undirected minimum-ratio cycles, each of which is computed and processed in amortized $m^{o(1)}$ time using a new dynamic graph data structure. Our framework extends to algorithms running in $m^{1+o(1)}$ time for computing flows that minimize general edge-separable convex functions to high accuracy. This gives almost-linear time algorithms for several problems including entropy-regularized optimal transport, matrix scaling, p-norm flows, and p-norm isotonic regression on arbitrary directed acyclic graphs."
                },
                {
                    "title": "A Better Algorithm for Societal Tradeoffs",
                    "abstract": "In the societal tradeoffs problem, each agent perceives certain quantitative tradeoffs between pairs of activities, and the goal is to aggregate these tradeoffs across agents. This is a problem in social choice; specifically, it is a type of quantitative judgment aggregation problem. A natural rule for this problem was axiomatized by Conitzer et al. [AAAI 2016]; they also provided several algorithms for computing the outcomes of this rule. In this paper, we present a significantly improved algorithm and evaluate it experimentally. Our algorithm is based on a tight connection to minimum-cost flow that we exhibit. We also show that our algorithm cannot be improved without breakthroughs on min-cost flow."
                },
                {
                    "title": "Learning Mixtures of Random Utility Models",
                    "abstract": "\n \n We tackle the problem of identifiability and efficient learning of mixtures of Random Utility Models (RUMs). We show that when the PDFs of utility distributions are symmetric, the mixture of k RUMs (denoted by k-RUM) is not identifiable when the number of alternatives m is no more than 2k-1. On the other hand, when m \u2265 max{4k-2,6}, any k-RUM is generically identifiable. We then propose three algorithms for learning mixtures of RUMs: an EM-based algorithm, which we call E-GMM, a direct generalized-method-of-moments (GMM) algorithm, and a sandwich (GMM-E-GMM) algorithm that combines the other two. Experiments on synthetic data show that the sandwich algorithm achieves the highest statistical efficiency and GMM is the most computationally efficient. Experiments on real-world data at Preflib show that Gaussian k-RUMs provide better fitness than a single Gaussian RUM, the Plackett-Luce model, and mixtures of Plackett-Luce models w.r.t. commonly-used model fitness criteria. To the best of our knowledge, this is the first work on learning mixtures of general RUMs.\n \n"
                },
                {
                    "title": "An homotopy method for $\\ell_p$ regression provably beyond self-concordance and in input-sparsity time",
                    "abstract": "We consider the problem of linear regression where the $\\ell_2^n$ norm loss (i.e., the usual least squares loss) is replaced by the $\\ell_p^n$ norm. We show how to solve such problems up to machine precision in $O^*(n^{|1/2 - 1/p|})$ (dense) matrix-vector products and $O^*(1)$ matrix inversions, or alternatively in $O^*(n^{|1/2 - 1/p|})$ calls to a (sparse) linear system solver. This improves the state of the art for any $p\\not\\in \\{1,2,+\\infty\\}$. Furthermore we also propose a randomized algorithm solving such problems in {\\em input sparsity time}, i.e., $O^*(Z + \\mathrm{poly}(d))$ where $Z$ is the size of the input and $d$ is the number of variables. Such a result was only known for $p=2$. Finally we prove that these results lie outside the scope of the Nesterov-Nemirovski's theory of interior point methods by showing that any symmetric self-concordant barrier on the $\\ell_p^n$ unit ball has self-concordance parameter $\\tilde{\\Omega}(n)$."
                },
                {
                    "title": "A Voting-Based System for Ethical Decision Making",
                    "abstract": "\n \n We present a general approach to automating ethical decisions, drawing on machine learning and computational social choice. In a nutshell, we propose to learn a model of societal preferences, and, when faced with a specific ethical dilemma at runtime, efficiently aggregate those preferences to identify a desirable choice. We provide a concrete algorithm that instantiates our approach; some of its crucial steps are informed by a new theory of swap-dominance efficient voting rules. Finally, we implement and evaluate a system for ethical decision making in the autonomous vehicle domain, using preference data collected from 1.3 million people through the Moral Machine website.\n \n"
                },
                {
                    "title": "Statistical Inference for Incomplete Ranking Data: The Case of Rank-Dependent Coarsening",
                    "abstract": "We consider the problem of statistical inference for ranking data, specifically rank aggregation, under the assumption that samples are incomplete in the sense of not comprising all choice alternatives. In contrast to most existing methods, we explicitly model the process of turning a full ranking into an incomplete one, which we call the coarsening process. To this end, we propose the concept of rank-dependent coarsening, which assumes that incomplete rankings are produced by projecting a full ranking to a random subset of ranks. For a concrete instantiation of our model, in which full rankings are drawn from a Plackett-Luce distribution and observations take the form of pairwise preferences, we study the performance of various rank aggregation methods. In addition to predictive accuracy in the finite sample setting, we address the theoretical question of consistency, by which we mean the ability to recover a target ranking when the sample size goes to infinity, despite a potential bias in the observations caused by the (unknown) coarsening."
                },
                {
                    "title": "Rules for Choosing Societal Tradeoffs",
                    "abstract": "\n \n We study the societal tradeoffs problem, where a set of voters each submit their ideal tradeoff value between each pair of activities (e.g., \"using a gallon of gasoline is as bad as creating 2 bags of landfill trash\"), and these are then aggregated into the societal tradeoff vector using a rule. We introduce the family of distance-based rules and show that these can be justified as maximum likelihood estimators of the truth. Within this family, we single out the logarithmic distance-based rule as especially appealing based on a social-choice-theoretic axiomatization. We give an efficient algorithm for executing this rule as well as an approximate hill climbing algorithm, and evaluate these experimentally.\n \n"
                },
                {
                    "title": "Quantitative Extensions of the Condorcet Jury Theorem with Strategic Agents",
                    "abstract": "\n \n The Condorcet Jury Theorem justifies the wisdom of crowds and lays the foundations of the ideology of the democratic regime. However, the Jury Theorem and most of its extensions focus on two alternatives and none of them quantitatively evaluate the effect of agents\u2019 strategic behavior on the mechanism\u2019s truth-revealing power. We initiate a research agenda of quantitatively extend- ing the Jury Theorem with strategic agents by characterizing the price of anarchy (PoA) and the price of stability (PoS) of the common interest Bayesian voting games for three classes of mechanisms: plurality, MAPs, and the mechanisms that satisfy anonymity, neutrality, and strategy-proofness (w.r.t. a set of natural probabil- ity models). We show that while plurality and MAPs have better best-case truth-revealing power (lower PoS), the third class of mechanisms are more robust against agents\u2019 strategic behavior (lower PoA).\n \n"
                },
                {
                    "title": "Crowdsourcing Societal Tradeoffs",
                    "abstract": "It would be desirable if, as a society, we could reduce the amount of landfill trash we create, the amount of carbon dioxide we emit, the amount of forest we clear, etc. Since we cannot (or are in any case not willing to) simultaneously pursue all these objectives to their maximum extent, we must prioritize among them. Currently, this is done mostly in an ad-hoc manner, with people, companies, local governments, and other entities deciding on an individual basis which of these objectives to pursue, and to what extent. \n \nA more systematic approach would be to set, at a global level, exact numerical tradeoffs: using one gallon of gasoline is as bad as creating x bags of landfill trash. Having such tradeoffs available would greatly facilitate decision making, and reduce inefficiencies resulting from inconsistent decisions across agents. But how could we arrive at a reasonable value for x? \n \nIn this paper, we argue that many techniques developed in the multiagent systems community, particularly those under economic paradigms, can be brought to bear on this question. We lay out our vision and discuss its relation to computational social choice, mechanism design, prediction markets, and related topics."
                },
                {
                    "title": "\u2113p Row Sampling by Lewis Weights",
                    "abstract": "We give a simple algorithm to efficiently sample the rows of a matrix while preserving the p-norms of its product with vectors. Given an $n$-by-$d$ matrix $\\boldsymbol{\\mathit{A}}$, we find with high probability and in input sparsity time an $\\boldsymbol{\\mathit{A}}'$ consisting of about $d \\log{d}$ rescaled rows of $\\boldsymbol{\\mathit{A}}$ such that $\\| \\boldsymbol{\\mathit{A}} \\boldsymbol{\\mathit{x}} \\|_1$ is close to $\\| \\boldsymbol{\\mathit{A}}' \\boldsymbol{\\mathit{x}} \\|_1$ for all vectors $\\boldsymbol{\\mathit{x}}$. We also show similar results for all $\\ell_p$ that give nearly optimal sample bounds in input sparsity time. Our results are based on sampling by \"Lewis weights\", which can be viewed as statistical leverage scores of a reweighted matrix. We also give an elementary proof of the guarantees of this sampling process for $\\ell_1$."
                },
                {
                    "title": "A Statistical Decision-Theoretic Framework for Social Choice",
                    "abstract": "In this paper, we take a statistical decision-theoretic viewpoint on social choice, putting a focus on the decision to be made on behalf of a system of agents. In our framework, we are given a statistical ranking model, a decision space, and a loss function defined on (parameter, decision) pairs, and formulate social choice mechanisms as decision rules that minimize expected loss. This suggests a general framework for the design and analysis of new social choice mechanisms. We compare Bayesian estimators, which minimize Bayesian expected loss, for the Mallows model and the Condorcet model respectively, and the Kemeny rule. We consider various normative properties, in addition to computational complexity and asymptotic behavior. In particular, we show that the Bayesian estimator for the Condorcet model satisfies some desired properties such as anonymity, neutrality, and monotonicity, can be computed in polynomial time, and is asymptotically different from the other two rules when the data are generated from the Condorcet model for some ground truth parameter."
                },
                {
                    "title": "Optimal Voting Rules",
                    "abstract": "We study dominant strategy incentive compatible (DIC) and deterministic mechanisms in a social choice setting with several alternatives. The agents are privately informed about their preferences, and have single-crossing utility functions. Monetary transfers are not feasible. We use an equivalence between deterministic, DIC mechanisms and generalized median voter schemes to construct the constrained-efficient, optimal mechanism for an utilitarian planner. Optimal schemes for other welfare criteria such as, say, a Rawlsian maximin can be analogously obtained."
                },
                {
                    "title": "When do noisy votes reveal the truth?",
                    "abstract": "A well-studied approach to the design of voting rules views them as maximum likelihood estimators; given votes that are seen as noisy estimates of a true ranking of the alternatives, the rule must reconstruct the most likely true ranking. We argue that this is too stringent a requirement, and instead ask: How many votes does a voting rule need to reconstruct the true ranking? We define the family of pairwise-majority consistent rules, and show that for all rules in this family the number of samples required from the Mallows noise model is logarithmic in the number of alternatives, and that no rule can do asymptotically better (while some rules like plurality do much worse). Taking a more normative point of view, we consider voting rules that surely return the true ranking as the number of samples tends to infinity (we call this property accuracy in the limit); this allows us to move to a higher level of abstraction. We study families of noise models that are parametrized by distance functions, and find voting rules that are accurate in the limit for all noise models in such general families. We characterize the distance functions that induce noise models for which pairwise-majority consistent rules are accurate in the limit, and provide a similar result for another novel family of position-dominance consistent rules. These characterizations capture three well-known distance functions."
                },
                {
                    "title": "Matrix Factorization Techniques for Recommender Systems",
                    "abstract": "As the Netflix Prize competition has demonstrated, matrix factorization models are superior to classic nearest neighbor techniques for producing product recommendations, allowing the incorporation of additional information such as implicit feedback, temporal effects, and confidence levels."
                },
                {
                    "title": "Bayesian modelling of football outcomes: using the Skellam's distribution for the goal difference",
                    "abstract": "Modelling football match outcomes is becoming increasingly popular nowadays for both team managers and betting funs. Most of the existing literature deals with modelling the number of goals scored by each team. In this paper, we work in a different direction. Instead of modelling the number of goals directly, we focus on the difference of the number of goals, i.e. the margin of victory. Modelling the differences instead of the scores themselves has some major advantages. Firstly, we eliminate correlation imposed by the fact that the two opponent teams compete each other, and secondly, we do not assume that the scored goals by each team are marginally Poisson distributed. Application of the Bayesian methodology for the Skellam\u2019s distribution using covariates is discussed. Illustrations using real data from the English Premiership for the season 2006\u20132007 are provided. The advantages of the proposed approach are also discussed."
                },
                {
                    "title": "Exploring Network Structure, Dynamics, and Function using NetworkX",
                    "abstract": "NetworkX is a Python language package for exploration and analysis of networks and network algorithms. The core package provides data structures for representing many types of networks, or graphs, including simple graphs, directed graphs, and graphs with parallel edges and self-loops. The nodes in NetworkX graphs can be any (hashable) Python object and edges can contain arbitrary data; this flexibility makes NetworkX ideal for representing networks found in many different scientific fields. In addition to the basic data structures many graph algorithms are implemented for calculating network properties and structure measures: shortest paths, betweenness centrality, clustering, and degree distribution and many more. NetworkX can read and write various graph formats for easy exchange with existing data, and provides generators for many classic graphs and popular graph models, such as the Erdos-Renyi, Small World, and Barabasi-Albert models. The ease-of-use and flexibility of the Python programming language together with connection to the SciPy tools make NetworkX a powerful tool for scientific computations. We discuss some of our recent work studying synchronization of coupled oscillators to demonstrate how NetworkX enables research in the field of computational networks."
                },
                {
                    "title": "Exact Matrix Completion via Convex Optimization",
                    "abstract": "We consider a problem of considerable practical interest: the recovery of a data matrix from a sampling of its entries. Suppose that we observe m entries selected uniformly at random from a matrix M. Can we complete the matrix and recover the entries that we have not seen?We show that one can perfectly recover most low-rank matrices from what appears to be an incomplete set of entries. We prove that if the number m of sampled entries obeys $$m\\ge C\\,n^{1.2}r\\log n$$ for some positive numerical constant C, then with very high probability, most n\u00d7n matrices of rank r can be perfectly recovered by solving a simple convex optimization program. This program finds the matrix with minimum nuclear norm that fits the data. The condition above assumes that the rank is not too large. However, if one replaces the 1.2 exponent with 1.25, then the result holds for all values of the rank. Similar results hold for arbitrary rectangular matrices as well. Our results are connected with the recent literature on compressed sensing, and show that objects other than signals and images can be perfectly reconstructed from very limited information."
                },
                {
                    "title": "Consensus ranking under the exponential model",
                    "abstract": "We analyze the generalized Mallows model, a popular exponential model over rankings. Estimating the central (or consensus) ranking from data is NP-hard. We obtain the following new results: (1) We show that search methods can estimate both the central ranking pi0 and the model parameters theta exactly. The search is n! in the worst case, but is tractable when the true distribution is concentrated around its mode; (2) We show that the generalized Mallows model is jointly exponential in (pi0; theta), and introduce the conjugate prior for this model class; (3) The sufficient statistics are the pairwise marginal probabilities that item i is preferred to item j. Preliminary experiments confirm the theoretical predictions and compare the new algorithm and existing heuristics."
                },
                {
                    "title": "TrueSkillTM: A Bayesian Skill Rating System",
                    "abstract": "We present a new Bayesian skill rating system which can be viewed as a generalisation of the Elo system used in Chess. The new system tracks the uncertainty about player skills, explicitly models draws, can deal with any number of competing entities and can infer individual skills from team results. Inference is performed by approximate message passing on a factor graph representation of the model. We present experimental evidence on the increased accuracy and convergence speed of the system compared to Elo and report on our experience with the new rating system running in a large-scale commercial online gaming service under the name of TrueSkill."
                },
                {
                    "title": "Improved Bounds for Computing Kemeny Rankings",
                    "abstract": "Voting (or rank aggregation) is a general method for aggregating the preferences of multiple agents. One voting rule of particular interest is the Kemeny rule, which minimizes the number of cases where the final ranking disagrees with a vote on the order of two alternatives. Unfortunately, Kemeny rankings are NP-hard to compute. Recent work on computing Kemeny rankings has focused on producing good bounds to use in search-based methods. In this paper, we extend on this work by providing various improved bounding techniques. Some of these are based on cycles in the pairwise majority graph, others are based on linear programs. We completely characterize the relative strength of all of these bounds and provide some experimental results."
                },
                {
                    "title": "Fast maximum margin matrix factorization for collaborative prediction",
                    "abstract": "Maximum Margin Matrix Factorization (MMMF) was recently suggested (Srebro et al., 2005) as a convex, infinite dimensional alternative to low-rank approximations and standard factor models. MMMF can be formulated as a semi-definite programming (SDP) and learned using standard SDP solvers. However, current SDP solvers can only handle MMMF problems on matrices of dimensionality up to a few hundred. Here, we investigate a direct gradient-based optimization method for MMMF and demonstrate it on large collaborative prediction problems. We compare against results obtained by Marlin (2004) and find that MMMF substantially outperforms all nine methods he tested."
                },
                {
                    "title": "Common Voting Rules as Maximum Likelihood Estimators",
                    "abstract": "Voting is a very general method of preference aggregation. A voting rule takes as input every voter's vote (typically, a ranking of the alternatives), and produces as output either just the winning alternative or a ranking of the alternatives. One potential view of voting is the following. There exists a 'correct' outcome (winner/ranking), and each voter's vote corresponds to a noisy perception of this correct outcome. If we are given the noise model, then for any vector of votes, we can"
                },
                {
                    "title": "MM algorithms for generalized Bradley-Terry models",
                    "abstract": "The Bradley-Terry model for paired comparisons is a simple and muchstudied means to describe the probabilities of the possible outcomes when individuals are judged against one another in pairs. Among the many studies of the model in the past 75 years, numerous authors have generalized it in several directions, sometimes providing iterative algorithms for obtaining maximum likelihood estimates for the generalizations. Building on a theory of algorithms known by the initials MM, for minorization-maximization, this paper presents a powerful technique for producing iterative maximum likelihood estimation algorithms for a wide class of generalizations of the Bradley-Terry model. While algorithms for problems of this type have tended to be custom-built in the literature, the techniques in this paper enable their mass production. Simple conditions are stated that guarantee that each algorithm described will produce a sequence that converges to the unique maximum likelihood estimator. Several of the algorithms and convergence results herein are new."
                },
                {
                    "title": "Condorcet's Theory of Voting",
                    "abstract": "Condcrcet's criterion states that an alternative that defeats every other by a simple majority is the socially optimal choice. Condorcet argued that if the object of voting is to determine the \u201cbest\u201d decision for society but voters sometimes make mistakes in their judgments, then the majority alternative (if it exists) is statistically most likely to be the best choice. Strictly speaking, this claim is not true; in some situations Bordas rule gives a sharper estimate of the best alternative. Nevertheless, Condorcet did propose a novel and statistically correct rule for finding the most likely ranking of the alternatives. This procedure, which is sometimes known as \u201cKemeny's rule,\u201d is the unique social welfare function that satisfies a variant of independence of irrelevant alternatives together with several other standard properties."
                },
                {
                    "title": "Modelling association football scores",
                    "abstract": "Previous authors have rejected the Poisson model for association football scores in favour of the Negative Binomial. This paper, however, investigates the Poisson model further. Parameters representing the teams' inherent attacking and defensive strengths are incorporated and the most appropriate model is found from a hierarchy of models. Observed and expected frequencies of scores are compared and goodness\u2010of\u2010fit tests show that although there are some small systematic differences, an independent Poisson model gives a reasonably accurate description of football scores. Improvements can be achieved by the use of a bivariate Poisson model with a correlation between scores of 0.2."
                },
                {
                    "title": "A Consistent Extension of Condorcet\u2019s Election Principle",
                    "abstract": "Condorcet\u2019s principle of choosing the majority alternative whenever one exists is violated not only by Borda\u2019s rule but by any scoring method; nevertheless the essential property of scoring functions\u2014\u201cconsistency\u201d of the outcome under aggregation of subgroups\u2014is shown to be compatible with Condorcet\u2019s principle. Moreover these two properties, suitably interpreted, together with neutrality, determine a unique rule known as Kemeny\u2019s rule."
                },
                {
                    "title": "Bmc Medical Research Methodology Open Access Dealing with Missing Data in a Multi-question Depression Scale: a Comparison of Imputation Methods",
                    "abstract": "Background: Missing data present a challenge to many research projects. The problem is often pronounced in studies utilizing self-report scales, and literature addressing different strategies for dealing with missing data in such circumstances is scarce. The objective of this study was to compare six different imputation techniques for dealing with missing data in the Zung Self-reported Depression scale (SDS)."
                }
            ],
            "categories": [
                "cs.GT"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively aggregate relative quantitative judgments from multiple agents or sources to derive a coherent decision-making framework in automated moral decision-making contexts?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem has significant implications for the research community, particularly in the fields of social choice, machine learning, and automated decision-making. By developing robust methods for aggregating quantitative judgments, we can enhance the understanding of how to make informed decisions based on diverse inputs, which is crucial in applications like self-driving cars and environmental policy. This research could lead to advancements in AI systems that better reflect human values and preferences, ultimately influencing future research directions in ethical AI and decision-making frameworks.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of accurately aggregating judgments that may be derived from different processes, rather than solely from subjective agents. Naive approaches may fail because they often assume that judgments are independent and directly reported, neglecting the potential noise and biases in the data. Additionally, technical obstacles include the need for sophisticated algorithms that can handle the variability and uncertainty inherent in quantitative judgments, as well as the theoretical challenge of ensuring that the aggregation method respects the principles of fairness and consistency.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on subjective judgments reported by agents, overlooking the possibility of deriving judgments from objective processes. This gap has limited the exploration of quantitative judgment aggregation in contexts where judgments are not directly reported but inferred from data. Barriers to solving this problem include a lack of appropriate models that account for the nuances of judgment derivation and the absence of comprehensive methodologies that integrate insights from both social choice theory and machine learning. Our approach differs by explicitly considering the process of judgment generation and developing aggregation methods that accommodate this complexity.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a framework for quantitative judgment aggregation that incorporates both subjective and objective sources of judgments. We will utilize a dataset comprising various scenarios where quantitative judgments are derived from real-world data, such as traffic patterns or environmental assessments. The evaluation metric will focus on the accuracy and consistency of the aggregated judgments compared to established benchmarks. We expect our outcomes to demonstrate improved decision-making capabilities in automated systems, providing a clearer understanding of how to integrate diverse quantitative inputs into coherent moral decisions."
            }
        },
        "author_data": {
            "317dcbdb-f953-4561-971e-6e380d10f2b1": {
                "pk": "317dcbdb-f953-4561-971e-6e380d10f2b1",
                "name": "Yixuan Even Xu",
                "collaborators": [
                    "Renfei Zhou",
                    "Fei Fang",
                    "Hanrui Zhang",
                    "Vincent Conitzer",
                    "Nickolai Gravin",
                    "Chun Kai Ling",
                    "Steven Jecmen",
                    "Zimeng Song",
                    "Jingxun Liang",
                    "Zhihao Gavin Tang"
                ],
                "domain": [
                    "Algorithm Design",
                    "Game Theory",
                    "Mechanism Design",
                    "Online Matching"
                ],
                "publications": [
                    {
                        "title": "Non-Excludable Bilateral Trade Between Groups",
                        "abstract": "Bilateral trade is one of the most natural and important forms of economic interaction: A seller has a single, indivisible item for sale, and a buyer is potentially interested. The two parties typically have different, privately known valuations for the item, and ideally, they would like to trade if the buyer values the item more than the seller. The celebrated impossibility result by Myerson and Satterthwaite shows that any mechanism for this setting must violate at least one important desideratum. In this paper, we investigate a richer paradigm of bilateral trade, with many self-interested buyers and sellers on both sides of a single trade who cannot be excluded from the trade. We show that this allows for more positive results. In fact, we establish a dichotomy in the possibility of trading efficiently. If in expectation, the buyers value the item more, we can achieve efficiency in the limit. If this is not the case, then efficiency cannot be achieved in general. En route, we characterize trading mechanisms that encourage truth-telling, which may be of independent interest. We also evaluate our trading mechanisms experimentally, and the experiments align with our theoretical results."
                    },
                    {
                        "title": "Bidder Selection Problem in Position Auctions: A Fast and Simple Algorithm via Poisson Approximation",
                        "abstract": "In the Bidder Selection Problem (BSP) there is a large pool of $n$ potential advertisers competing for ad slots on the user's web page. Due to strict computational restrictions, the advertising platform can run a proper auction only for a fraction $k<n$ of advertisers. We consider the basic optimization problem underlying BSP: given $n$ independent prior distributions, how to efficiently find a subset of $k$ with the objective of either maximizing expected social welfare or revenue of the platform. We study BSP in the classic multi-winner model of position auctions for welfare and revenue objectives using the optimal (respectively, VCG mechanism, or Myerson's auction) format for the selected set of bidders. Previous PTAS results for BSP optimization were only known for single-item auctions and in case of [Segev and Singla 2021] for $l$-unit auctions. More importantly, all of these PTASes were computational complexity results with impractically large running times, which defeats the purpose of using these algorithms under severe computational constraints.   We propose a novel Poisson relaxation of BSP for position auctions that immediately implies that 1) BSP is polynomial-time solvable up to a vanishingly small error as the problem size $k$ grows; 2) there is a PTAS for position auctions after combining our relaxation with the trivial brute force algorithm. Unlike all previous PTASes, we implemented our algorithm and did extensive numerical experiments on practically relevant input sizes. First, our experiments corroborate the previous experimental findings of Mehta et al. that a few simple heuristics used in practice perform surprisingly well in terms of approximation factor. Furthermore, our algorithm outperforms Greedy both in running time and approximation on medium and large-sized instances."
                    },
                    {
                        "title": "Learning Coalition Structures with Games",
                        "abstract": "Coalitions naturally exist in many real-world systems involving multiple decision makers such as ridesharing, security, and online ad auctions, but the coalition structure among the agents is often unknown. We propose and study an important yet previously overseen problem -- Coalition Structure Learning (CSL), where we aim to carefully design a series of games for the agents and infer the underlying coalition structure by observing their interactions in those games. We establish a lower bound on the sample complexity -- defined as the number of games needed to learn the structure -- of any algorithms for CSL and propose the Iterative Grouping (IG) algorithm for designing normal-form games to achieve the lower bound. We show that IG can be extended to other succinct games such as congestion games and graphical games. Moreover, we solve CSL in a more restrictive and practical setting: auctions. We show a variant of IG to solve CSL in the auction setting even if we cannot design the bidder valuations. Finally, we conduct experiments to evaluate IG in the auction setting and the results align with our theoretical analysis."
                    },
                    {
                        "title": "A One-Size-Fits-All Approach to Improving Randomness in Paper Assignment",
                        "abstract": "The assignment of papers to reviewers is a crucial part of the peer review processes of large publication venues, where organizers (e.g., conference program chairs) rely on algorithms to perform automated paper assignment. As such, a major challenge for the organizers of these processes is to specify paper assignment algorithms that find appropriate assignments with respect to various desiderata. Although the main objective when choosing a good paper assignment is to maximize the expertise of each reviewer for their assigned papers, several other considerations make introducing randomization into the paper assignment desirable: robustness to malicious behavior, the ability to evaluate alternative paper assignments, reviewer diversity, and reviewer anonymity. However, it is unclear in what way one should randomize the paper assignment in order to best satisfy all of these considerations simultaneously. In this work, we present a practical, one-size-fits-all method for randomized paper assignment intended to perform well across different motivations for randomness. We show theoretically and experimentally that our method outperforms currently-deployed methods for randomized paper assignment on several intuitive randomness metrics, demonstrating that the randomized assignments produced by our method are general-purpose."
                    },
                    {
                        "title": "On the Perturbation Function of Ranking and Balance for Weighted Online Bipartite Matching",
                        "abstract": "Ranking and Balance are arguably the two most important algorithms in the online matching literature. They achieve the same optimal competitive ratio of $1-1/e$ for the integral version and fractional version of online bipartite matching by Karp, Vazirani, and Vazirani (STOC 1990) respectively. The two algorithms have been generalized to weighted online bipartite matching problems, including vertex-weighted online bipartite matching and AdWords, by utilizing a perturbation function. The canonical choice of the perturbation function is $f(x)=1-e^{x-1}$ as it leads to the optimal competitive ratio of $1-1/e$ in both settings.   We advance the understanding of the weighted generalizations of Ranking and Balance in this paper, with a focus on studying the effect of different perturbation functions. First, we prove that the canonical perturbation function is the \\emph{unique} optimal perturbation function for vertex-weighted online bipartite matching. In stark contrast, all perturbation functions achieve the optimal competitive ratio of $1-1/e$ in the unweighted setting. Second, we prove that the generalization of Ranking to AdWords with unknown budgets using the canonical perturbation function is at most $0.624$ competitive, refuting a conjecture of Vazirani (2021). More generally, as an application of the first result, we prove that no perturbation function leads to the prominent competitive ratio of $1-1/e$ by establishing an upper bound of $1-1/e-0.0003$.   Finally, we propose the online budget-additive welfare maximization problem that is intermediate between AdWords and AdWords with unknown budgets, and we design an optimal $1-1/e$ competitive algorithm by generalizing Balance."
                    }
                ]
            },
            "c25a6150-3623-4551-9003-cf07fcd7ae6a": {
                "pk": "c25a6150-3623-4551-9003-cf07fcd7ae6a",
                "name": "Hanrui Zhang",
                "collaborators": [
                    "Wei Chen",
                    "Vincent Conitzer",
                    "Shang-Hua Teng",
                    "Ruosong Wang",
                    "Kunhe Yang",
                    "Yu Cheng",
                    "Yuan Deng",
                    "Jieming Mao",
                    "Vahab Mirrokni",
                    "Song Zuo"
                ],
                "domain": [
                    "Mechanism Design",
                    "Game Theory",
                    "Social Networks",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Complete Submodularity Characterization in the Comparative Independent Cascade Model",
                        "abstract": "We study the propagation of comparative ideas or items in social networks. A full characterization for submodularity in the comparative independent cascade (Com-IC) model of two-idea cascade is given, for competing ideas and complementary ideas respectively, with or without reconsideration. We further introduce One-Shot model where agents show less patience toward ideas, and show that in One-Shot model, only the strongest idea spreads with submodularity."
                    },
                    {
                        "title": "Automated Dynamic Mechanism Design",
                        "abstract": "We study Bayesian automated mechanism design in unstructured dynamic environments, where a principal repeatedly interacts with an agent, and takes actions based on the strategic agent's report of the current state of the world. Both the principal and the agent can have arbitrary and potentially different valuations for the actions taken, possibly also depending on the actual state of the world. Moreover, at any time, the state of the world may evolve arbitrarily depending on the action taken by the principal. The goal is to compute an optimal mechanism which maximizes the principal's utility in the face of the self-interested strategic agent.   We give an efficient algorithm for computing optimal mechanisms, with or without payments, under different individual-rationality constraints, when the time horizon is constant. Our algorithm is based on a sophisticated linear program formulation, which can be customized in various ways to accommodate richer constraints. For environments with large time horizons, we show that the principal's optimal utility is hard to approximate within a certain constant factor, complementing our algorithmic result. We further consider a special case of the problem where the agent is myopic, and give a refined efficient algorithm whose time complexity scales linearly in the time horizon. Moreover, we show that memoryless mechanisms do not provide a good solution for our problem, in terms of both optimality and computational tractability. These results paint a relatively complete picture for automated dynamic mechanism design in unstructured environments. Finally, we present experimental results where our algorithms are applied to synthetic dynamic environments with different characteristics, which not only serve as a proof of concept for our algorithms, but also exhibit intriguing phenomena in dynamic mechanism design."
                    },
                    {
                        "title": "Unit-sphere games",
                        "abstract": "This paper introduces a class of games, called unit-sphere games, where strategies are real vectors with unit 2-norms (or, on a unit-sphere). As a result, they can no longer be interpreted as probability distributions over actions, but rather be thought of as allocations of one unit of resource to actions and the multiplicative payoff effect on each action is proportional to square-root of the amount of resource allocated to that action. The new definition generates a number of interesting consequences. We first characterize sufficient and necessary conditions under which a two-player unit-sphere game has a Nash equilibrium. The characterization effectively reduces solving a unit-sphere game to finding all eigenvalues and eigenvectors of the product of individual payoff matrices. For any unit-sphere game with non-negative payoff matrices, there always exists a unique Nash equilibrium; furthermore, the unique equilibrium is efficiently reachable via Cournot adjustment. In addition, we show that any equilibrium in positive unit-sphere games corresponds to approximate equilibria in the corresponding normal-form games. Analogous but weaker results are extended to positive n-player unit-sphere games."
                    },
                    {
                        "title": "Computational Aspects of Bayesian Persuasion under Approximate Best Response",
                        "abstract": "We study Bayesian persuasion under approximate best response, where the receiver may choose any action that is not too much suboptimal given their posterior belief upon receiving the signal. We focus on the computational aspects of the problem, aiming to design algorithms that efficiently compute (almost) optimal strategies for the sender. Despite the absence of the revelation principle -- which has been one of the most powerful tools in Bayesian persuasion -- we design polynomial-time exact algorithms for the problem when either the state space or the action space is small, as well as a quasi-polynomial-time approximation scheme (QPTAS) for the general problem. On the negative side, we show there is no polynomial-time exact algorithm for the general problem unless $\\mathsf{P} = \\mathsf{NP}$. Our results build on several new algorithmic ideas, which might be useful in other principal-agent problems where robustness is desired."
                    },
                    {
                        "title": "Efficient Algorithms for Planning with Participation Constraints",
                        "abstract": "We consider the problem of planning with participation constraints introduced in [Zhang et al., 2022]. In this problem, a principal chooses actions in a Markov decision process, resulting in separate utilities for the principal and the agent. However, the agent can and will choose to end the process whenever his expected onward utility becomes negative. The principal seeks to compute and commit to a policy that maximizes her expected utility, under the constraint that the agent should always want to continue participating. We provide the first polynomial-time exact algorithm for this problem for finite-horizon settings, where previously only an additive $\\varepsilon$-approximation algorithm was known. Our approach can also be extended to the (discounted) infinite-horizon case, for which we give an algorithm that runs in time polynomial in the size of the input and $\\log(1/\\varepsilon)$, and returns a policy that is optimal up to an additive error of $\\varepsilon$."
                    },
                    {
                        "title": "Automated Mechanism Design for Classification with Partial Verification",
                        "abstract": "We study the problem of automated mechanism design with partial verification, where each type can (mis)report only a restricted set of types (rather than any other type), induced by the principal's limited verification power. We prove hardness results when the revelation principle does not necessarily hold, as well as when types have even minimally different preferences. In light of these hardness results, we focus on truthful mechanisms in the setting where all types share the same preference over outcomes, which is motivated by applications in, e.g., strategic classification. We present a number of algorithmic and structural results, including an efficient algorithm for finding optimal deterministic truthful mechanisms, which also implies a faster algorithm for finding optimal randomized truthful mechanisms via a characterization based on convexity. We then consider a more general setting, where the principal's cost is a function of the combination of outcomes assigned to each type. In particular, we focus on the case where the cost function is submodular, and give generalizations of essentially all our results in the classical setting where the cost function is additive. Our results provide a relatively complete picture for automated mechanism design with partial verification."
                    },
                    {
                        "title": "Nonbossy Mechanisms: Mechanism Design Robust to Secondary Goals",
                        "abstract": "We study mechanism design when agents may have hidden secondary goals which will manifest as non-trivial preferences among outcomes for which their primary utility is the same. We show that in such cases, a mechanism is robust against strategic manipulation if and only if it is not only incentive-compatible, but also nonbossy -- a well-studied property in the context of matching and allocation mechanisms. We give complete characterizations of incentive-compatible and nonbossy mechanisms in various settings, including auctions with single-parameter agents and public decision settings where all agents share a common outcome. In particular, we show that in the single-item setting, a mechanism is incentive-compatible, individually rational, and nonbossy if and only if it is a sequential posted-price mechanism. In contrast, we show that in more general single-parameter environments, there exist mechanisms satisfying our characterization that significantly outperform sequential posted-price mechanisms in terms of revenue or efficiency (sometimes by an exponential factor)."
                    },
                    {
                        "title": "A Systematic Framework and Characterization of Influence-Based Network Centrality",
                        "abstract": "In this paper, we present a framework for studying the following fundamental question in network analysis: How should one assess the centralities of nodes in an information/influence propagation process over a social network?   Our framework systematically extends a family of classical graph-theoretical centrality formulations, including degree centrality, harmonic centrality, and their \"sphere-of-influence\" generalizations, to influence-based network centralities. We further extend natural group centralities from graph models to influence models, since group cooperation is essential in social influences. This in turn enables us to assess individuals' centralities in group influence settings by applying the concept of Shapley value from cooperative game theory.   Mathematically, using the property that these centrality formulations are Bayesian, we prove the following characterization theorem: Every influence-based centrality formulation in this family is the unique Bayesian centrality that conforms with its corresponding graph-theoretical centrality formulation. Moreover, the uniqueness is fully determined by the centrality formulation on the class of layered graphs, which is derived from a beautiful algebraic structure of influence instances modeled by cascading sequences. Our main mathematical result that layered graphs in fact form a basis for the space of influence-cascading-sequence profiles could also be useful in other studies of network influences. We further provide an algorithmic framework for efficient approximation of these influence-based centrality measures.   Our study provides a systematic road map for comparative analyses of different influence-based centrality formulations, as well as for transferring graph-theoretical concepts to influence models."
                    },
                    {
                        "title": "Capturing Complementarity in Set Functions by Going Beyond Submodularity/Subadditivity",
                        "abstract": "We introduce two new \"degree of complementarity\" measures, which we refer to, respectively, as supermodular width and superadditive width. Both are formulated based on natural witnesses of complementarity. We show that both measures are robust by proving that they, respectively, characterize the gap of monotone set functions from being submodular and subadditive. Thus, they define two new hierarchies over monotone set functions, which we will refer to as Supermodular Width (SMW) hierarchy and Superadditive Width (SAW) hierarchy, with level 0 of the hierarchies resting exactly on submodular and subadditive functions, respectively.   We present a comprehensive comparative analysis of the SMW hierarchy and the Supermodular Degree (SD) hierarchy, defined by Feige and Izsak. We prove that the SMW hierarchy is strictly more expressive than the SD hierarchy. We show that previous results regarding approximation guarantees for welfare and constrained maximization as well as regarding the Price of Anarchy (PoA) of simple auctions can be extended without any loss from the SD hierarchy to the SMW hierarchy. We also establish almost matching information-theoretical lower bounds. The combination of these approximation and hardness results illustrate that the SMW hierarchy provides an accurate characterization of \"near submodularity\" needed for maximization approximation. While SD and SMW hierarchies support nontrivial bounds on the PoA of simple auctions, we show that our SAW hierarchy seems to capture more intrinsic properties needed to realize the efficiency of simple auctions. So far, the SAW hierarchy provides the best dependency for the PoA of Single-bid Auction, and is nearly as competitive as the Maximum over Positive Hypergraphs (MPH) hierarchy for Simultaneous Item First Price Auction (SIA). We provide almost tight lower bounds for the PoA of both auctions with respect to the SAW hierarchy."
                    },
                    {
                        "title": "Nearly Linear Row Sampling Algorithm for Quantile Regression",
                        "abstract": "We give a row sampling algorithm for the quantile loss function with sample complexity nearly linear in the dimensionality of the data, improving upon the previous best algorithm whose sampling complexity has at least cubic dependence on the dimensionality. Based upon our row sampling algorithm, we give the fastest known algorithm for quantile regression and a graph sparsification algorithm for balanced directed graphs. Our main technical contribution is to show that Lewis weights sampling, which has been used in row sampling algorithms for $\\ell_p$ norms, can also be applied in row sampling algorithms for a variety of loss functions. We complement our theoretical results by experiments to demonstrate the practicality of our approach."
                    },
                    {
                        "title": "Non-Excludable Bilateral Trade Between Groups",
                        "abstract": "Bilateral trade is one of the most natural and important forms of economic interaction: A seller has a single, indivisible item for sale, and a buyer is potentially interested. The two parties typically have different, privately known valuations for the item, and ideally, they would like to trade if the buyer values the item more than the seller. The celebrated impossibility result by Myerson and Satterthwaite shows that any mechanism for this setting must violate at least one important desideratum. In this paper, we investigate a richer paradigm of bilateral trade, with many self-interested buyers and sellers on both sides of a single trade who cannot be excluded from the trade. We show that this allows for more positive results. In fact, we establish a dichotomy in the possibility of trading efficiently. If in expectation, the buyers value the item more, we can achieve efficiency in the limit. If this is not the case, then efficiency cannot be achieved in general. En route, we characterize trading mechanisms that encourage truth-telling, which may be of independent interest. We also evaluate our trading mechanisms experimentally, and the experiments align with our theoretical results."
                    },
                    {
                        "title": "Strategic Littlestone Dimension: Improved Bounds on Online Strategic Classification",
                        "abstract": "We study the problem of online binary classification in settings where strategic agents can modify their observable features to receive a positive classification. We model the set of feasible manipulations by a directed graph over the feature space, and assume the learner only observes the manipulated features instead of the original ones. We introduce the Strategic Littlestone Dimension, a new combinatorial measure that captures the joint complexity of the hypothesis class and the manipulation graph. We demonstrate that it characterizes the instance-optimal mistake bounds for deterministic learning algorithms in the realizable setting. We also achieve improved regret in the agnostic setting by a refined agnostic-to-realizable reduction that accounts for the additional challenge of not observing agents' original features. Finally, we relax the assumption that the learner knows the manipulation graph, instead assuming their knowledge is captured by a family of graphs. We derive regret bounds in both the realizable setting where all agents manipulate according to the same graph within the graph family, and the agnostic setting where the manipulation graphs are chosen adversarially and not consistently modeled by a single graph in the family."
                    },
                    {
                        "title": "On the Equivalence Between High-Order Network-Influence Frameworks: General-Threshold, Hypergraph-Triggering, and Logic-Triggering Models",
                        "abstract": "In this paper, we study several high-order network-influence-propagation frameworks and their connection to the classical network diffusion frameworks such as the triggering model and the general threshold model. In one framework, we use hyperedges to represent many-to-one influence -- the collective influence of a group of nodes on another node -- and define the hypergraph triggering model as a natural extension to the classical triggering model. In another framework, we use monotone Boolean functions to capture the diverse logic underlying many-to-one influence behaviors, and extend the triggering model to the Boolean-function triggering model. We prove that the Boolean-function triggering model, even with refined details of influence logic, is equivalent to the hypergraph triggering model, and both are equivalent to the general threshold model. Moreover, the general threshold model is optimal in the number of parameters, among all models with the same expressive power. We further extend these three equivalent models by introducing correlations among influence propagations on different nodes. Surprisingly, we discover that while the correlated hypergraph-based model is still equivalent to the correlated Boolean-function-based model, the correlated general threshold model is more restrictive than the two high-order models. Our study sheds light on high-order network-influence propagations by providing new insight into the group influence behaviors in existing models, as well as diverse modeling tools for understanding influence propagations in networks."
                    },
                    {
                        "title": "Efficiency of the First-Price Auction in the Autobidding World",
                        "abstract": "We study the price of anarchy of the first-price auction in the autobidding world, where bidders can be either utility maximizers (i.e., traditional bidders) or value maximizers (i.e., autobidders). We show that with autobidders only, the price of anarchy of the first-price auction is $1/2$, and with both kinds of bidders, the price of anarchy degrades to about $0.457$ (the precise number is given by an optimization). These results complement the recent result by Jin and Lu [2022] showing that the price of anarchy of the first-price auction with traditional bidders only is $1 - 1/e^2$. We further investigate a setting where the seller can utilize machine-learned advice to improve the efficiency of the auctions. There, we show that as the accuracy of the advice increases, the price of anarchy improves smoothly from about $0.457$ to $1$."
                    },
                    {
                        "title": "An Improved Algorithm for Incremental DFS Tree in Undirected Graphs",
                        "abstract": "Depth first search (DFS) tree is one of the most well-known data structures for designing efficient graph algorithms. Given an undirected graph $G=(V,E)$ with $n$ vertices and $m$ edges, the textbook algorithm takes $O(n+m)$ time to construct a DFS tree. In this paper, we study the problem of maintaining a DFS tree when the graph is undergoing incremental updates. Formally, we show: Given an arbitrary online sequence of edge or vertex insertions, there is an algorithm that reports a DFS tree in $O(n)$ worst case time per operation, and requires $O\\left(\\min\\{m \\log n, n^2\\}\\right)$ preprocessing time.   Our result improves the previous $O(n \\log^3 n)$ worst case update time algorithm by Baswana et al. and the $O(n \\log n)$ time by Nakamura and Sadakane, and matches the trivial $\\Omega(n)$ lower bound when it is required to explicitly output a DFS tree.   Our result builds on the framework introduced in the breakthrough work by Baswana et al., together with a novel use of a tree-partition lemma by Duan and Zhan, and the celebrated fractional cascading technique by Chazelle and Guibas."
                    },
                    {
                        "title": "Autobidding Auctions in the Presence of User Costs",
                        "abstract": "We study autobidding ad auctions with user costs, where each bidder is value-maximizing subject to a return-over-investment (ROI) constraint, and the seller aims to maximize the social welfare taking into consideration the user's cost of viewing an ad. We show that in the worst case, the approximation ratio of social welfare by running the vanilla VCG auctions with user costs could as bad as 0. To improve the performance of VCG, We propose a new variant of VCG based on properly chosen cost multipliers, and prove that there exist auction-dependent and bidder-dependent cost multipliers that guarantee approximation ratios of 1/2 and 1/4 respectively in terms of the social welfare."
                    },
                    {
                        "title": "Provably Efficient $Q$-learning with Function Approximation via Distribution Shift Error Checking Oracle",
                        "abstract": "$Q$-learning with function approximation is one of the most popular methods in reinforcement learning. Though the idea of using function approximation was proposed at least 60 years ago, even in the simplest setup, i.e, approximating $Q$-functions with linear functions, it is still an open problem on how to design a provably efficient algorithm that learns a near-optimal policy. The key challenges are how to efficiently explore the state space and how to decide when to stop exploring in conjunction with the function approximation scheme.   The current paper presents a provably efficient algorithm for $Q$-learning with linear function approximation. Under certain regularity assumptions, our algorithm, Difference Maximization $Q$-learning (DMQ), combined with linear function approximation, returns a near-optimal policy using a polynomial number of trajectories. Our algorithm introduces a new notion, the Distribution Shift Error Checking (DSEC) oracle. This oracle tests whether there exists a function in the function class that predicts well on a distribution $\\mathcal{D}_1$, but predicts poorly on another distribution $\\mathcal{D}_2$, where $\\mathcal{D}_1$ and $\\mathcal{D}_2$ are distributions over states induced by two different exploration policies. For the linear function class, this oracle is equivalent to solving a top eigenvalue problem. We believe our algorithmic insights, especially the DSEC oracle, are also useful in designing and analyzing reinforcement learning algorithms with general function approximation."
                    }
                ]
            },
            "203fd076-9a08-4a14-8736-609fce331e14": {
                "pk": "203fd076-9a08-4a14-8736-609fce331e14",
                "name": "Yu Cheng",
                "collaborators": [
                    "Mo Yu",
                    "Xiaoxiao Guo",
                    "Bowen Zhou",
                    "Rong Ge"
                ],
                "domain": [
                    "Few-shot Learning",
                    "Meta-Learning",
                    "Matrix Completion",
                    "Non-convex Optimization"
                ],
                "publications": [
                    {
                        "title": "Few-shot Learning with Meta Metric Learners",
                        "abstract": "Few-shot Learning aims to learn classifiers for new classes with only a few training examples per class. Existing meta-learning or metric-learning based few-shot learning approaches are limited in handling diverse domains with various number of labels. The meta-learning approaches train a meta learner to predict weights of homogeneous-structured task-specific networks, requiring a uniform number of classes across tasks. The metric-learning approaches learn one task-invariant metric for all the tasks, and they fail if the tasks diverge. We propose to deal with these limitations with meta metric learning. Our meta metric learning approach consists of task-specific learners, that exploit metric learning to handle flexible labels, and a meta learner, that discovers good parameters and gradient decent to specify the metrics in task-specific learners. Thus the proposed model is able to handle unbalanced classes as well as to generate task-specific metrics. We test our approach in the `$k$-shot $N$-way' few-shot learning setting used in previous work and new realistic few-shot setting with diverse multi-domain tasks and flexible label numbers. Experiments show that our approach attains superior performances in both settings."
                    },
                    {
                        "title": "Non-Convex Matrix Completion Against a Semi-Random Adversary",
                        "abstract": "Matrix completion is a well-studied problem with many machine learning applications. In practice, the problem is often solved by non-convex optimization algorithms. However, the current theoretical analysis for non-convex algorithms relies heavily on the assumption that every entry is observed with exactly the same probability $p$, which is not realistic in practice.   In this paper, we investigate a more realistic semi-random model, where the probability of observing each entry is at least $p$. Even with this mild semi-random perturbation, we can construct counter-examples where existing non-convex algorithms get stuck in bad local optima.   In light of the negative results, we propose a pre-processing step that tries to re-weight the semi-random input, so that it becomes \"similar\" to a random input. We give a nearly-linear time algorithm for this problem, and show that after our pre-processing, all the local minima of the non-convex objective can be used to approximately recover the underlying ground-truth matrix."
                    }
                ]
            },
            "90e0d165-f4e5-435f-bf02-5fe0998871e4": {
                "pk": "90e0d165-f4e5-435f-bf02-5fe0998871e4",
                "name": "Vincent Conitzer",
                "collaborators": [
                    "Tuomas Sandholm"
                ],
                "domain": [
                    "Game Theory",
                    "AI Ethics",
                    "Computational Complexity",
                    "Multiagent Systems"
                ],
                "publications": [
                    {
                        "title": "Philosophy in the Face of Artificial Intelligence",
                        "abstract": "In this article, I discuss how the AI community views concerns about the emergence of superintelligent AI and related philosophical issues."
                    },
                    {
                        "title": "Can Artificial Intelligence Do Everything That We Can?",
                        "abstract": "In this article, I discuss what AI can and cannot yet do, and the implications for humanity."
                    },
                    {
                        "title": "The Exact Computational Complexity of Evolutionarily Stable Strategies",
                        "abstract": "While the computational complexity of many game-theoretic solution concepts, notably Nash equilibrium, has now been settled, the question of determining the exact complexity of computing an evolutionarily stable strategy has resisted solution since attention was drawn to it in 2004. In this paper, I settle this question by proving that deciding the existence of an evolutionarily stable strategy is $\\Sigma_2^P$-complete."
                    },
                    {
                        "title": "Computing Equilibria with Partial Commitment",
                        "abstract": "In security games, the solution concept commonly used is that of a Stackelberg equilibrium where the defender gets to commit to a mixed strategy. The motivation for this is that the attacker can repeatedly observe the defender's actions and learn her distribution over actions, before acting himself. If the actions were not observable, Nash (or perhaps correlated) equilibrium would arguably be a more natural solution concept. But what if some, but not all, aspects of the defender's actions are observable? In this paper, we introduce solution concepts corresponding to this case, both with and without correlation. We study their basic properties, whether these solutions can be efficiently computed, and the impact of additional observability on the utility obtained."
                    },
                    {
                        "title": "A Puzzle about Further Facts",
                        "abstract": "In metaphysics, there are a number of distinct but related questions about the existence of \"further facts\" -- facts that are contingent relative to the physical structure of the universe. These include further facts about qualia, personal identity, and time. In this article I provide a sequence of examples involving computer simulations, ranging from one in which the protagonist can clearly conclude such further facts exist to one that describes our own condition. This raises the question of where along the sequence (if at all) the protagonist stops being able to soundly conclude that further facts exist."
                    },
                    {
                        "title": "Prediction Markets, Mechanism Design, and Cooperative Game Theory",
                        "abstract": "Prediction markets are designed to elicit information from multiple agents in order to predict (obtain probabilities for) future events. A good prediction market incentivizes agents to reveal their information truthfully; such incentive compatibility considerations are commonly studied in mechanism design. While this relation between prediction markets and mechanism design is well understood at a high level, the models used in prediction markets tend to be somewhat different from those used in mechanism design. This paper considers a model for prediction markets that fits more straightforwardly into the mechanism design framework. We consider a number of mechanisms within this model, all based on proper scoring rules. We discuss basic properties of these mechanisms, such as incentive compatibility. We also draw connections between some of these mechanisms and cooperative game theory. Finally, we speculate how one might build a practical prediction market based on some of these ideas."
                    },
                    {
                        "title": "A Devastating Example for the Halfer Rule",
                        "abstract": "How should we update de dicto beliefs in the face of de se evidence? The Sleeping Beauty problem divides philosophers into two camps, halfers and thirders. But there is some disagreement among halfers about how their position should generalize to other examples. A full generalization is not always given; one notable exception is the Halfer Rule, under which the agent updates her uncentered beliefs based on only the uncentered part of her evidence. In this brief article, I provide a simple example for which the Halfer Rule prescribes credences that, I argue, cannot be reasonably held by anyone. In particular, these credences constitute an egregious violation of the Reflection Principle. I then discuss the consequences for halfing in general."
                    },
                    {
                        "title": "A Dutch Book against Sleeping Beauties Who Are Evidential Decision Theorists",
                        "abstract": "In the context of the Sleeping Beauty problem, it has been argued that so-called \"halfers\" can avoid Dutch book arguments by adopting evidential decision theory. I introduce a Dutch book for a variant of the Sleeping Beauty problem and argue that evidential decision theorists fall prey to it, whether they are halfers or thirders. The argument crucially requires that an action can provide evidence for what the agent would do not only at other decision points where she has exactly the same information, but also at decision points where she has different but \"symmetric\" information."
                    },
                    {
                        "title": "On Stackelberg Mixed Strategies",
                        "abstract": "It is sometimes the case that one solution concept in game theory is equivalent to applying another solution concept to a modified version of the game. In such cases, does it make sense to study the former separately (as it applies to the original representation of the game), or should we entirely subordinate it to the latter? The answer probably depends on the particular circumstances, and indeed the literature takes different approaches in different cases. In this article, I consider the specific example of Stackelberg mixed strategies. I argue that, even though a Stackelberg mixed strategy can also be seen as a subgame perfect Nash equilibrium of a corresponding extensive-form game, there remains significant value in studying it separately. The analysis of this special case may have implications for other solution concepts."
                    },
                    {
                        "title": "The Personalized A-Theory of Time and Perspective",
                        "abstract": "A-theorists and B-theorists debate whether the \"Now\" is metaphysically distinguished from other time slices. Analogously, one may ask whether the \"I\" is metaphysically distinguished from other perspectives. Few philosophers would answer the second question in the affirmative. An exception is Caspar Hare, who has devoted two papers and a book to arguing for such a positive answer. In this paper, I argue that those who answer the first question in the affirmative -- A-theorists -- should also answer the second question in the affirmative. This is because key arguments in favor of the A-theory are more effective as arguments in favor of the resulting combined position, and key arguments against the A-theory are ineffective against the combined position."
                    },
                    {
                        "title": "The Complexity of Computing Robust Mediated Equilibria in Ordinal Games",
                        "abstract": "Usually, to apply game-theoretic methods, we must specify utilities precisely, and we run the risk that the solutions we compute are not robust to errors in this specification. Ordinal games provide an attractive alternative: they require specifying only which outcomes are preferred to which other ones. Unfortunately, they provide little guidance for how to play unless there are pure Nash equilibria; evaluating mixed strategies appears to fundamentally require cardinal utilities.   In this paper, we observe that we can in fact make good use of mixed strategies in ordinal games if we consider settings that allow for folk theorems. These allow us to find equilibria that are robust, in the sense that they remain equilibria no matter which cardinal utilities are the correct ones -- as long as they are consistent with the specified ordinal preferences. We analyze this concept and study the computational complexity of finding such equilibria in a range of settings."
                    },
                    {
                        "title": "Can rational choice guide us to correct {\\em de se} beliefs?",
                        "abstract": "Significant controversy remains about what constitute correct self-locating beliefs in scenarios such as the Sleeping Beauty problem, with proponents on both the \"halfer\" and \"thirder\" sides. To attempt to settle the issue, one natural approach consists in creating decision variants of the problem, determining what actions the various candidate beliefs prescribe, and assessing whether these actions are reasonable when we step back. Dutch book arguments are a special case of this approach, but other Sleeping Beauty games have also been constructed to make similar points. Building on a recent article (James R.~Shaw. {\\em De se} belief and rational choice. {\\em Synthese}, 190(3):491-508, 2013), I show that in general we should be wary of such arguments, because unintuitive actions may result for reasons that are unrelated to the beliefs. On the other hand, I show that, when we restrict our attention to {\\em additive} games, then a thirder will necessarily maximize her {\\em ex ante} expected payout, but a halfer in some cases will not (assuming causal decision theory). I conclude that this does not necessarily settle the issue and speculate about what might."
                    },
                    {
                        "title": "Why should we ever automate moral decision making?",
                        "abstract": "While people generally trust AI to make decisions in various aspects of their lives, concerns arise when AI is involved in decisions with significant moral implications. The absence of a precise mathematical framework for moral reasoning intensifies these concerns, as ethics often defies simplistic mathematical models. Unlike fields such as logical reasoning, reasoning under uncertainty, and strategic decision-making, which have well-defined mathematical frameworks, moral reasoning lacks a broadly accepted framework. This absence raises questions about the confidence we can place in AI's moral decision-making capabilities.   The environments in which AI systems are typically trained today seem insufficiently rich for such a system to learn ethics from scratch, and even if we had an appropriate environment, it is unclear how we might bring about such learning. An alternative approach involves AI learning from human moral decisions. This learning process can involve aggregating curated human judgments or demonstrations in specific domains, or leveraging a foundation model fed with a wide range of data. Still, concerns persist, given the imperfections in human moral decision making.   Given this, why should we ever automate moral decision making -- is it not better to leave all moral decision making to humans? This paper lays out a number of reasons why we should expect AI systems to engage in decisions with a moral component, with brief discussions of the associated risks."
                    },
                    {
                        "title": "Eliciting Single-Peaked Preferences Using Comparison Queries",
                        "abstract": "Voting is a general method for aggregating the preferences of multiple agents. Each agent ranks all the possible alternatives, and based on this, an aggregate ranking of the alternatives (or at least a winning alternative) is produced. However, when there are many alternatives, it is impractical to simply ask agents to report their complete preferences. Rather, the agents preferences, or at least the relevant parts thereof, need to be elicited. This is done by asking the agents a (hopefully small) number of simple queries about their preferences, such as comparison queries, which ask an agent to compare two of the alternatives. Prior work on preference elicitation in voting has focused on the case of unrestricted preferences. It has been shown that in this setting, it is sometimes necessary to ask each agent (almost) as many queries as would be required to determine an arbitrary ranking of the alternatives. In contrast, in this paper, we focus on single-peaked preferences. We show that such preferences can be elicited using only a linear number of comparison queries, if either the order with respect to which preferences are single-peaked is known, or at least one other agents complete preferences are known. We show that using a sublinear number of queries does not suffice. We also consider the case of cardinally single-peaked preferences. For this case, we show that if the alternatives cardinal positions are known, then an agents preferences can be elicited using only a logarithmic number of queries; however, we also show that if the cardinal positions are not known, then a sublinear number of queries does not suffice. We present experimental results for all elicitation algorithms. We also consider the problem of only eliciting enough information to determine the aggregate ranking, and show that even for this more modest objective, a sublinear number of queries per agent does not suffice for known ordinal or unknown cardinal positions. Finally, we discuss whether and how these techniques can be applied when preferences are almost single-peaked."
                    },
                    {
                        "title": "Common Voting Rules as Maximum Likelihood Estimators",
                        "abstract": "Voting is a very general method of preference aggregation. A voting rule takes as input every voter's vote (typically, a ranking of the alternatives), and produces as output either just the winning alternative or a ranking of the alternatives. One potential view of voting is the following. There exists a 'correct' outcome (winner/ranking), and each voter's vote corresponds to a noisy perception of this correct outcome. If we are given the noise model, then for any vector of votes, we can"
                    },
                    {
                        "title": "Vote Elicitation: Complexity and Strategy-Proofness",
                        "abstract": "Preference elicitation is a central problem in AI, and has received significant attention in single-agent settings. It is also a key problem in multiagent systems, but has received little attention here so far. In this setting, the agents may have different preferences that often must be aggregated using voting. This leads to interesting issues because what, if any, information should be elicited from an agent depends on what other agents have revealed about their preferences so far.   In this paper we study effective elicitation, and its impediments, for the most common voting protocols. It turns out that in the Single Transferable Vote protocol, even knowing when to terminate elicitation is mathcal NP-complete, while this is easy for all the other protocols under study. Even for these protocols, determining how to elicit effectively is NP-complete, even with perfect suspicions about how the agents will vote. The exception is the Plurality protocol where such effective elicitation is easy.   We also show that elicitation introduces additional opportunities for strategic manipulation by the voters. We demonstrate how to curtail the space of elicitation schemes so that no such additional strategic issues arise."
                    },
                    {
                        "title": "Complexity Results about Nash Equilibria",
                        "abstract": "Noncooperative game theory provides a normative framework for analyzing strategic interactions. However, for the toolbox to be operational, the solutions it defines will have to be computed. In this paper, we provide a single reduction that 1) demonstrates NP-hardness of determining whether Nash equilibria with certain natural properties exist, and 2) demonstrates the #P-hardness of counting Nash equilibria (or connected sets of Nash equilibria). We also show that 3) determining whether a pure-strategy Bayes-Nash equilibrium exists is NP-hard, and that 4) determining whether a pure-strategy Nash equilibrium exists in a stochastic (Markov) game is PSPACE-hard even if the game is invisible (this remains NP-hard if the game is finite). All of our hardness results hold even if there are only two players and the game is symmetric.   Keywords: Nash equilibrium; game theory; computational complexity; noncooperative game theory; normal form game; stochastic game; Markov game; Bayes-Nash equilibrium; multiagent systems."
                    },
                    {
                        "title": "Complexity of Mechanism Design",
                        "abstract": "The aggregation of conflicting preferences is a central problem in multiagent systems. The key difficulty is that the agents may report their preferences insincerely. Mechanism design is the art of designing the rules of the game so that the agents are motivated to report their preferences truthfully and a (socially) desirable outcome is chosen. We propose an approach where a mechanism is automatically created for the preference aggregation setting at hand. This has several advantages, but the downside is that the mechanism design optimization problem needs to be solved anew each time. Focusing on settings where side payments are not possible, we show that the mechanism design problem is NP-complete for deterministic mechanisms. This holds both for dominant-strategy implementation and for Bayes-Nash implementation. We then show that if we allow randomized mechanisms, the mechanism design problem becomes tractable. In other words, the coordinator can tackle the computational complexity introduced by its uncertainty about the agents' preferences by making the agents face additional uncertainty. This comes at no loss, and in some cases at a gain, in the (social) objective."
                    },
                    {
                        "title": "AWESOME: A General Multiagent Learning Algorithm that Converges in Self-Play and Learns a Best Response Against Stationary Opponents",
                        "abstract": "A satisfactory multiagent learning algorithm should, {\\em at a minimum}, learn to play optimally against stationary opponents and converge to a Nash equilibrium in self-play. The algorithm that has come closest, WoLF-IGA, has been proven to have these two properties in 2-player 2-action repeated games--assuming that the opponent's (mixed) strategy is observable. In this paper we present AWESOME, the first algorithm that is guaranteed to have these two properties in {\\em all} repeated (finite) games. It requires only that the other players' actual actions (not their strategies) can be observed at each step. It also learns to play optimally against opponents that {\\em eventually become} stationary. The basic idea behind AWESOME ({\\em Adapt When Everybody is Stationary, Otherwise Move to Equilibrium}) is to try to adapt to the others' strategies when they appear stationary, but otherwise to retreat to a precomputed equilibrium strategy. The techniques used to prove the properties of AWESOME are fundamentally different from those used for previous algorithms, and may help in analyzing other multiagent learning algorithms also."
                    },
                    {
                        "title": "Definition and Complexity of Some Basic Metareasoning Problems",
                        "abstract": "In most real-world settings, due to limited time or other resources, an agent cannot perform all potentially useful deliberation and information gathering actions. This leads to the metareasoning problem of selecting such actions. Decision-theoretic methods for metareasoning have been studied in AI, but there are few theoretical results on the complexity of metareasoning.   We derive hardness results for three settings which most real metareasoning systems would have to encompass as special cases. In the first, the agent has to decide how to allocate its deliberation time across anytime algorithms running on different problem instances. We show this to be $\\mathcal{NP}$-complete. In the second, the agent has to (dynamically) allocate its deliberation or information gathering resources across multiple actions that it has to choose among. We show this to be $\\mathcal{NP}$-hard even when evaluating each individual action is extremely simple. In the third, the agent has to (dynamically) choose a limited number of deliberation or information gathering actions to disambiguate the state of the world. We show that this is $\\mathcal{NP}$-hard under a natural restriction, and $\\mathcal{PSPACE}$-hard in general."
                    }
                ]
            }
        }
    },
    "2405.18979": {
        "paper_data": {
            "title": "MANO: Exploiting Matrix Norm for Unsupervised Accuracy Estimation Under Distribution Shifts",
            "url": "http://arxiv.org/abs/2405.18979v2",
            "arxiv_id": "2405.18979",
            "authors": [
                "Renchunzi Xie",
                "Ambroise Odonnat",
                "Vasilii Feofanov",
                "Weijian Deng",
                "Jianfeng Zhang",
                "Bo An"
            ],
            "abstract": "Leveraging the models' outputs, specifically the logits, is a common approach to estimating the test accuracy of a pre-trained neural network on out-of-distribution (OOD) samples without requiring access to the corresponding ground truth labels. Despite their ease of implementation and computational efficiency, current logit-based methods are vulnerable to overconfidence issues, leading to prediction bias, especially under the natural shift. In this work, we first study the relationship between logits and generalization performance from the view of low-density separation assumption. Our findings motivate our proposed method MaNo which (1) applies a data-dependent normalization on the logits to reduce prediction bias, and (2) takes the $L_p$ norm of the matrix of normalized logits as the estimation score. Our theoretical analysis highlights the connection between the provided score and the model's uncertainty. We conduct an extensive empirical study on common unsupervised accuracy estimation benchmarks and demonstrate that MaNo achieves state-of-the-art performance across various architectures in the presence of synthetic, natural, or subpopulation shifts.",
            "introduction": "   1 Introduction  The deployment of machine learning models in real-world scenarios is frequently challenged by distribution shifts between the training and test data. These shifts can substantially deteriorate the model\u2019s performance during testing\u00a0(Quionero-Candela et\u00a0al.,, 2009; Geirhos et\u00a0al.,, 2018; Koh et\u00a0al.,, 2021) and introduce significant risks related to AI safety\u00a0(Hendrycks and Mazeika,, 2022; Deng and Zheng,, 2021). To alleviate this issue, it is common to monitor model performance by periodically collecting the ground truth labels for a subset of the current test dataset\u00a0(Lu et\u00a0al.,, 2023). However, this approach is often resource-intensive and time-consuming, which motivates the importance of estimating the model\u2019s performance on out-of-distribution (OOD) data in an unsupervised manner, also known as Unsupervised Accuracy Estimation (Donmez et\u00a0al.,, 2010).   Due to privacy constraints and computational efficiency, one of the most popular ways to estimate accuracy without labels is to rely on the model\u2019s outputs, called logits, as a source of confidence in the model\u2019s predictions\u00a0(Hendrycks and Gimpel,, 2016; Garg et\u00a0al.,, 2022; Deng et\u00a0al.,, 2023; Guillory et\u00a0al.,, 2021). For instance, ConfScore\u00a0(Hendrycks and Gimpel,, 2016) leverages the average maximum softmax probability as the test accuracy estimator, while Deng et\u00a0al., (2023) has recently proposed to estimate the accuracy via the nuclear norm of the softmax probability matrix. These approaches, however, tend to underperform on the natural shift applications while the intuition behind the use of logits remains unclear. This motivates us to ask:   Question 1 \u2003What explains the correlation between logits and generalization performance?   In section\u00a03, we show that logits are connected to the model\u2019s margins, i.e., the distances between the learned embeddings and the decision boundaries. Inspired by the low-density separation (LDS) assumption\u00a0(Chapelle and Zien,, 2005; Feofanov et\u00a0al.,, 2023) that optimal decision boundaries should lie in low-density regions, we propose MaNo, an estimation score that aggregates the margins at a dataset level by taking the Lpsubscript\ud835\udc3f\ud835\udc5dL_{p}italic_L start_POSTSUBSCRIPT italic_p end_POSTSUBSCRIPT-norm of the normalized model\u2019s prediction matrix to evaluate the density around decision boundaries. Nevertheless, logit-based approaches are known to suffer from overconfidence\u00a0(Wei et\u00a0al., 2022a, ; Odonnat et\u00a0al.,, 2024), resulting in high prediction bias, especially under poorly-calibrated scenarios. This leads us to another critical question:   Question 2 \u2003How to alleviate the overconfidence issues of logits-based methods?   In section\u00a04, we reveal that this question is connected to the normalization of logits and show that the widely-used softmax normalization accumulates errors in the presence of prediction bias, which can lead to overconfidence and significantly degrade the performance of existing accuracy estimation methods in poorly-calibrated scenarios. To mitigate this issue, we propose a novel normalization strategy called SoftTrun that takes into account the empirical distribution of logits and aims to find a trade-off between information completeness of ground-truth logits and error accumulation.   Summary of our contributions. (1) We show that logits are informative of generalization performance through the lens of the low-density separation assumption by reflecting the distances to decision boundaries. (2) We identify the failure of the commonly-used softmax normalization that accumulates errors under poorly calibrated because of its overconfidence, leading to biased estimation. (3) We propose MaNo, a training-free estimation method that quantifies the global distances to decision boundaries by taking the Lpsubscript\ud835\udc3f\ud835\udc5dL_{p}italic_L start_POSTSUBSCRIPT italic_p",
            "references": [
                {
                    "title": "Energy-based Automated Model Evaluation",
                    "abstract": "The conventional evaluation protocols on machine learning models rely heavily on a labeled, i.i.d-assumed testing dataset, which is not often present in real world applications. The Automated Model Evaluation (AutoEval) shows an alternative to this traditional workflow, by forming a proximal prediction pipeline of the testing performance without the presence of ground-truth labels. Despite its recent successes, the AutoEval frameworks still suffer from an overconfidence issue, substantial storage and computational cost. In that regard, we propose a novel measure -- Meta-Distribution Energy (MDE) -- that allows the AutoEval framework to be both more efficient and effective. The core of the MDE is to establish a meta-distribution statistic, on the information (energy) associated with individual samples, then offer a smoother representation enabled by energy-based learning. We further provide our theoretical insights by connecting the MDE with the classification loss. We provide extensive experiments across modalities, datasets and different architectural backbones to validate MDE's validity, together with its superiority compared with prior approaches. We also prove MDE's versatility by showing its seamless integration with large-scale models, and easy adaption to learning scenarios with noisy- or imbalanced- labels. Code and data are available: https://github.com/pengr/Energy_AutoEval"
                },
                {
                    "title": "Leveraging Ensemble Diversity for Robust Self-Training in the Presence of Sample Selection Bias",
                    "abstract": "Self-training is a well-known approach for semi-supervised learning. It consists of iteratively assigning pseudo-labels to unlabeled data for which the model is confident and treating them as labeled examples. For neural networks, softmax prediction probabilities are often used as a confidence measure, although they are known to be overconfident, even for wrong predictions. This phenomenon is particularly intensified in the presence of sample selection bias, i.e., when data labeling is subject to some constraint. To address this issue, we propose a novel confidence measure, called $\\mathcal{T}$-similarity, built upon the prediction diversity of an ensemble of linear classifiers. We provide the theoretical analysis of our approach by studying stationary points and describing the relationship between the diversity of the individual members and their performance. We empirically demonstrate the benefit of our confidence measure for three different pseudo-labeling policies on classification datasets of various data modalities. The code is available at https://github.com/ambroiseodt/tsim."
                },
                {
                    "title": "Random Matrix Analysis to Balance between Supervised and Unsupervised Learning under the Low Density Separation Assumption",
                    "abstract": "We propose a theoretical framework to analyze semi-supervised classification under the low density separation assumption in a high-dimensional regime. In particular, we introduce QLDS, a linear classification model, where the low density separation assumption is implemented via quadratic margin maximization. The algorithm has an explicit solution with rich theoretical properties, and we show that particular cases of our algorithm are the least-square support vector machine in the supervised case, the spectral clustering in the fully unsupervised regime, and a class of semi-supervised graph-based approaches. As such, QLDS establishes a smooth bridge between these supervised and unsupervised learning methods. Using recent advances in the random matrix theory, we formally derive a theoretical evaluation of the classification error in the asymptotic regime. As an application, we derive a hyperparameter selection policy that finds the best balance between the supervised and the unsupervised terms of our learning criterion. Finally, we provide extensive illustrations of our framework, as well as an experimental study on several benchmarks to demonstrate that QLDS, while being computationally more efficient, improves over cross-validation for hyperparameter selection, indicating a high promise of the usage of random matrix theory for semi-supervised model selection."
                },
                {
                    "title": "Characterizing Out-of-Distribution Error via Optimal Transport",
                    "abstract": "Out-of-distribution (OOD) data poses serious challenges in deployed machine learning models, so methods of predicting a model's performance on OOD data without labels are important for machine learning safety. While a number of methods have been proposed by prior work, they often underestimate the actual error, sometimes by a large margin, which greatly impacts their applicability to real tasks. In this work, we identify pseudo-label shift, or the difference between the predicted and true OOD label distributions, as a key indicator to this underestimation. Based on this observation, we introduce a novel method for estimating model performance by leveraging optimal transport theory, Confidence Optimal Transport (COT), and show that it provably provides more robust error estimates in the presence of pseudo-label shift. Additionally, we introduce an empirically-motivated variant of COT, Confidence Optimal Transport with Thresholding (COTT), which applies thresholding to the individual transport costs and further improves the accuracy of COT's error estimates. We evaluate COT and COTT on a variety of standard benchmarks that induce various types of distribution shift -- synthetic, novel subpopulation, and natural -- and show that our approaches significantly outperform existing state-of-the-art methods with an up to 3x lower prediction error."
                },
                {
                    "title": "On the Importance of Feature Separability in Predicting Out-Of-Distribution Error",
                    "abstract": "Estimating the generalization performance is practically challenging on out-of-distribution (OOD) data without ground-truth labels. While previous methods emphasize the connection between distribution difference and OOD accuracy, we show that a large domain gap not necessarily leads to a low test accuracy. In this paper, we investigate this problem from the perspective of feature separability empirically and theoretically. Specifically, we propose a dataset-level score based upon feature dispersion to estimate the test accuracy under distribution shift. Our method is inspired by desirable properties of features in representation learning: high inter-class dispersion and high intra-class compactness. Our analysis shows that inter-class dispersion is strongly correlated with the model accuracy, while intra-class compactness does not reflect the generalization performance on OOD data. Extensive experiments demonstrate the superiority of our method in both prediction performance and computational efficiency."
                },
                {
                    "title": "Confidence and Dispersity Speak: Characterising Prediction Matrix for Unsupervised Accuracy Estimation",
                    "abstract": "This work aims to assess how well a model performs under distribution shifts without using labels. While recent methods study prediction confidence, this work reports prediction dispersity is another informative cue. Confidence reflects whether the individual prediction is certain; dispersity indicates how the overall predictions are distributed across all categories. Our key insight is that a well-performing model should give predictions with high confidence and high dispersity. That is, we need to consider both properties so as to make more accurate estimates. To this end, we use the nuclear norm that has been shown to be effective in characterizing both properties. Extensive experiments validate the effectiveness of nuclear norm for various models (e.g., ViT and ConvNeXt), different datasets (e.g., ImageNet and CUB-200), and diverse types of distribution shifts (e.g., style shift and reproduction shift). We show that the nuclear norm is more accurate and robust in accuracy estimation than existing methods. Furthermore, we validate the feasibility of other measurements (e.g., mutual information maximization) for characterizing dispersity and confidence. Lastly, we investigate the limitation of the nuclear norm, study its improved variant under severe class imbalance, and discuss potential directions."
                },
                {
                    "title": "X-Risk Analysis for AI Research",
                    "abstract": "Artificial intelligence (AI) has the potential to greatly improve society, but as with any powerful technology, it comes with heightened risks and responsibilities. Current AI research lacks a systematic discussion of how to manage long-tail risks from AI systems, including speculative long-term risks. Keeping in mind the potential benefits of AI, there is some concern that building ever more intelligent and powerful AI systems could eventually result in systems that are more powerful than us; some say this is like playing with fire and speculate that this could create existential risks (x-risks). To add precision and ground these discussions, we provide a guide for how to analyze AI x-risk, which consists of three parts: First, we review how systems can be made safer today, drawing on time-tested concepts from hazard analysis and systems safety that have been designed to steer large processes in safer directions. Next, we discuss strategies for having long-term impacts on the safety of future systems. Finally, we discuss a crucial concept in making AI systems safer by improving the balance between safety and general capabilities. We hope this document and the presented concepts and tools serve as a useful guide for understanding how to analyze AI x-risk."
                },
                {
                    "title": "Mitigating Neural Network Overconfidence with Logit Normalization",
                    "abstract": "Detecting out-of-distribution inputs is critical for safe deployment of machine learning models in the real world. However, neural networks are known to suffer from the overconfidence issue, where they produce abnormally high confidence for both in- and out-of-distribution inputs. In this work, we show that this issue can be mitigated through Logit Normalization (LogitNorm) -- a simple fix to the cross-entropy loss -- by enforcing a constant vector norm on the logits in training. Our method is motivated by the analysis that the norm of the logit keeps increasing during training, leading to overconfident output. Our key idea behind LogitNorm is thus to decouple the influence of output's norm during network optimization. Trained with LogitNorm, neural networks produce highly distinguishable confidence scores between in- and out-of-distribution data. Extensive experiments demonstrate the superiority of LogitNorm, reducing the average FPR95 by up to 42.30% on common benchmarks."
                },
                {
                    "title": "Self-Training: A Survey",
                    "abstract": "Semi-supervised algorithms aim to learn prediction functions from a small set of labeled observations and a large set of unlabeled observations. Because this framework is relevant in many applications, they have received a lot of interest in both academia and industry. Among the existing techniques, self-training methods have undoubtedly attracted greater attention in recent years. These models are designed to find the decision boundary on low density regions without making additional assumptions about the data distribution, and use the unsigned output score of a learned classifier, or its margin, as an indicator of confidence. The working principle of self-training algorithms is to learn a classifier iteratively by assigning pseudo-labels to the set of unlabeled training samples with a margin greater than a certain threshold. The pseudo-labeled examples are then used to enrich the labeled training data and to train a new classifier in conjunction with the labeled training set. In this paper, we present self-training methods for binary and multi-class classification; as well as their variants and two related approaches, namely consistency-based approaches and transductive learning. We examine the impact of significant self-training features on various methods, using different general and image classification benchmarks, and we discuss our ideas for future research in self-training. To the best of our knowledge, this is the first thorough and complete survey on this subject."
                },
                {
                    "title": "Predicting Out-of-Distribution Error with the Projection Norm",
                    "abstract": "We propose a metric -- Projection Norm -- to predict a model's performance on out-of-distribution (OOD) data without access to ground truth labels. Projection Norm first uses model predictions to pseudo-label test samples and then trains a new model on the pseudo-labels. The more the new model's parameters differ from an in-distribution model, the greater the predicted OOD error. Empirically, our approach outperforms existing methods on both image and text classification tasks and across different network architectures. Theoretically, we connect our approach to a bound on the test error for overparameterized linear models. Furthermore, we find that Projection Norm is the only approach that achieves non-trivial detection performance on adversarial examples. Our code is available at https://github.com/yaodongyu/ProjNorm."
                },
                {
                    "title": "Leveraging Unlabeled Data to Predict Out-of-Distribution Performance",
                    "abstract": "Real-world machine learning deployments are characterized by mismatches between the source (training) and target (test) distributions that may cause performance drops. In this work, we investigate methods for predicting the target domain accuracy using only labeled source data and unlabeled target data. We propose Average Thresholded Confidence (ATC), a practical method that learns a threshold on the model's confidence, predicting accuracy as the fraction of unlabeled examples for which model confidence exceeds that threshold. ATC outperforms previous methods across several model architectures, types of distribution shifts (e.g., due to synthetic corruptions, dataset reproduction, or novel subpopulations), and datasets (Wilds, ImageNet, Breeds, CIFAR, and MNIST). In our experiments, ATC estimates target performance $2$-$4\\times$ more accurately than prior methods. We also explore the theoretical foundations of the problem, proving that, in general, identifying the accuracy is just as hard as identifying the optimal predictor and thus, the efficacy of any method rests upon (perhaps unstated) assumptions on the nature of the shift. Finally, analyzing our method on some toy distributions, we provide insights concerning when it works. Code is available at https://github.com/saurabhgarg1996/ATC_code/."
                },
                {
                    "title": "Predicting with Confidence on Unseen Distributions",
                    "abstract": "Recent work has shown that the accuracy of machine learning models can vary substantially when evaluated on a distribution that even slightly differs from that of the training data. As a result, predicting model performance on previously unseen distributions without access to labeled data is an important challenge with implications for increasing the reliability of machine learning models. In the context of distribution shift, distance measures are often used to adapt models and improve their performance on new domains, however accuracy estimation is seldom explored in these investigations. Our investigation determines that common distributional distances such as Frechet distance or Maximum Mean Discrepancy, fail to induce reliable estimates of performance under distribution shift. On the other hand, we find that our proposed difference of confidences (DoC) approach yields successful estimates of a classifier\u2019s performance over a variety of shifts and model architectures. Despite its simplicity, we observe that DoC outperforms other methods across synthetic, natural, and adversarial distribution shifts, reducing error by (> 46%) on several realistic and challenging datasets such as ImageNet-Vid-Robust and ImageNet-Rendition."
                },
                {
                    "title": "Assessing Generalization of SGD via Disagreement",
                    "abstract": "We empirically show that the test error of deep networks can be estimated by simply training the same architecture on the same training set but with a different run of Stochastic Gradient Descent (SGD), and measuring the disagreement rate between the two networks on unlabeled test data. This builds on -- and is a stronger version of -- the observation in Nakkiran&Bansal '20, which requires the second run to be on an altogether fresh training set. We further theoretically show that this peculiar phenomenon arises from the \\emph{well-calibrated} nature of \\emph{ensembles} of SGD-trained models. This finding not only provides a simple empirical measure to directly predict the test error using unlabeled test data, but also establishes a new conceptual connection between generalization and calibration."
                },
                {
                    "title": "What Does Rotation Prediction Tell Us about Classifier Accuracy under Varying Testing Environments?",
                    "abstract": "Understanding classifier decision under novel environments is central to the community, and a common practice is evaluating it on labeled test sets. However, in real-world testing, image annotations are difficult and expensive to obtain, especially when the test environment is changing. A natural question then arises: given a trained classifier, can we evaluate its accuracy on varying unlabeled test sets? In this work, we train semantic classification and rotation prediction in a multi-task way. On a series of datasets, we report an interesting finding, i.e., the semantic classification accuracy exhibits a strong linear relationship with the accuracy of the rotation prediction task (Pearson's Correlation r>0.88). This finding allows us to utilize linear regression to estimate classifier performance from the accuracy of rotation prediction which can be obtained on the test set through the freely generated rotation labels."
                },
                {
                    "title": "Adversarially Adaptive Normalization for Single Domain Generalization",
                    "abstract": "Single domain generalization aims to learn a model that performs well on many unseen domains with only one domain data for training. Existing works focus on studying the adversarial domain augmentation (ADA) to improve the model\u2019s generalization capability. The impact on domain generalization of the statistics of normalization layers is still underinvestigated. In this paper, we propose a generic normalization approach, adaptive standardization and rescaling normalization (ASR-Norm), to complement the missing part in previous works. ASR-Norm learns both the standardization and rescaling statistics via neural networks. This new form of normalization can be viewed as a generic form of the traditional normalizations. When trained with ADA, the statistics in ASR-Norm are learned to be adaptive to the data coming from different domains, and hence improves the model generalization performance across domains, especially on the target domain with large discrepancy from the source domain. The experimental results show that ASR-Norm can bring consistent improvement to the state-of-the-art ADA approaches by 1.6%, 2.7%, and 6.3% averagely on the Digits, CIFAR-10-C, and PACS benchmarks, respectively. As a generic tool, the improvement introduced by ASR-Norm is agnostic to the choice of ADA methods."
                },
                {
                    "title": "On Interaction Between Augmentations and Corruptions in Natural Corruption Robustness",
                    "abstract": "Invariance to a broad array of image corruptions, such as warping, noise, or color shifts, is an important aspect of building robust models in computer vision. Recently, several new data augmentations have been proposed that significantly improve performance on ImageNet-C, a benchmark of such corruptions. However, there is still a lack of basic understanding on the relationship between data augmentations and test-time corruptions. To this end, we develop a feature space for image transforms, and then use a new measure in this space between augmentations and corruptions called the Minimal Sample Distance to demonstrate a strong correlation between similarity and performance. We then investigate recent data augmentations and observe a significant degradation in corruption robustness when the test-time corruptions are sampled to be perceptually dissimilar from ImageNet-C in this feature space. Our results suggest that test error can be improved by training on perceptually similar augmentations, and data augmentations may not generalize well beyond the existing benchmark. We hope our results and tools will allow for more robust progress towards improving robustness to image corruptions. We provide code at https://github.com/facebookresearch/augmentation-corruption."
                },
                {
                    "title": "Deep Learning Generalization and the Convex Hull of Training Sets",
                    "abstract": "We study the generalization of deep learning models in relation to the convex hull of their training sets. A trained image classifier basically partitions its domain via decision boundaries and assigns a class to each of those partitions. The location of decision boundaries inside the convex hull of training set can be investigated in relation to the training samples. However, our analysis shows that in standard image classification datasets, all testing images are considerably outside that convex hull, in the pixel space, in the wavelet space, and in the internal representations learned by deep networks. Therefore, the performance of a trained model partially depends on how its decision boundaries are extended outside the convex hull of its training data. From this perspective which is not studied before, over-parameterization of deep learning models may be considered a necessity for shaping the extension of decision boundaries. At the same time, over-parameterization should be accompanied by a specific training regime, in order to yield a model that not only fits the training set, but also its decision boundaries extend desirably outside the convex hull. To illustrate this, we investigate the decision boundaries of a neural network, with various degrees of parameters, inside and outside the convex hull of its training set. Moreover, we use a polynomial decision boundary to study the necessity of over-parameterization and the influence of training regime in shaping its extensions outside the convex hull of training set."
                },
                {
                    "title": "WILDS: A Benchmark of in-the-Wild Distribution Shifts",
                    "abstract": "Distribution shifts -- where the training distribution differs from the test distribution -- can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity in the real-world deployments, these distribution shifts are under-represented in the datasets widely used in the ML community today. To address this gap, we present WILDS, a curated benchmark of 10 datasets reflecting a diverse range of distribution shifts that naturally arise in real-world applications, such as shifts across hospitals for tumor identification; across camera traps for wildlife monitoring; and across time and location in satellite imaging and poverty mapping. On each dataset, we show that standard training yields substantially lower out-of-distribution than in-distribution performance. This gap remains even with models trained by existing methods for tackling distribution shifts, underscoring the need for new methods for training models that are more robust to the types of distribution shifts that arise in practice. To facilitate method development, we provide an open-source package that automates dataset loading, contains default model architectures and hyperparameters, and standardizes evaluations. Code and leaderboards are available at https://wilds.stanford.edu."
                },
                {
                    "title": "BREEDS: Benchmarks for Subpopulation Shift",
                    "abstract": "We develop a methodology for assessing the robustness of models to subpopulation shift---specifically, their ability to generalize to novel data subpopulations that were not observed during training. Our approach leverages the class structure underlying existing datasets to control the data subpopulations that comprise the training and test distributions. This enables us to synthesize realistic distribution shifts whose sources can be precisely controlled and characterized, within existing large-scale datasets. Applying this methodology to the ImageNet dataset, we create a suite of subpopulation shift benchmarks of varying granularity. We then validate that the corresponding shifts are tractable by obtaining human baselines for them. Finally, we utilize these benchmarks to measure the sensitivity of standard model architectures as well as the effectiveness of off-the-shelf train-time robustness interventions. Code and data available at this https URL ."
                },
                {
                    "title": "Are Labels Always Necessary for Classifier Accuracy Evaluation?",
                    "abstract": "To calculate the model accuracy on a computer vision task, e.g., object recognition, we usually require a test set composing of test samples and their ground truth labels. Whilst standard usage cases satisfy this requirement, many real-world scenarios involve unlabeled test data, rendering common model evaluation methods infeasible. We investigate this important and under-explored problem, Automatic model Evaluation (AutoEval). Specifically, given a labeled training set and a classifier, we aim to estimate the classification accuracy on unlabeled test datasets. We construct a meta-dataset: a dataset comprised of datasets generated from the original images via various transformations such as rotation, background substitution, foreground scaling, etc. As the classification accuracy of the model on each sample (dataset) is known from the original dataset labels, our task can be solved via regression. Using the feature statistics to represent the distribution of a sample dataset, we can train regression models (e.g., a regression neural network) to predict model performance. Using synthetic meta-dataset and real-world datasets in training and testing, respectively, we report a reasonable and promising prediction of the model accuracy. We also provide insights into the application scope, limitation, and potential future direction of AutoEval."
                },
                {
                    "title": "Multi-Class Data Description for Out-of-distribution Detection",
                    "abstract": "The capability of reliably detecting out-of-distribution samples is one of the key factors in deploying a good classifier, as the test distribution always does not match with the training distribution in most real-world applications. In this work, we present a deep multi-class data description, termed as Deep-MCDD, which is effective to detect out-of-distribution (OOD) samples as well as classify in-distribution (ID) samples. Unlike the softmax classifier that only focuses on the linear decision boundary partitioning its latent space into multiple regions, our Deep-MCDD aims to find a spherical decision boundary for each class which determines whether a test sample belongs to the class or not. By integrating the concept of Gaussian discriminant analysis into deep neural networks, we propose a deep learning objective to learn class-conditional distributions that are explicitly modeled as separable Gaussian distributions. Thereby, we can define the confidence score by the distance of a test sample from each class-conditional distribution, and utilize it for identifying OOD samples. Our empirical evaluation on multi-class tabular and image datasets demonstrates that Deep-MCDD achieves the best performances in distinguishing OOD samples while showing the classification accuracy as high as the other competitors."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "Understanding the Decision Boundary of Deep Neural Networks: An Empirical Study",
                    "abstract": "Despite achieving remarkable performance on many image classification tasks, state-of-the-art machine learning (ML) classifiers remain vulnerable to small input perturbations. Especially, the existence of adversarial examples raises concerns about the deployment of ML models in safety- and security-critical environments, like autonomous driving and disease detection. Over the last few years, numerous defense methods have been published with the goal of improving adversarial as well as corruption robustness. However, the proposed measures succeeded only to a very limited extent. This limited progress is partly due to the lack of understanding of the decision boundary and decision regions of deep neural networks. Therefore, we study the minimum distance of data points to the decision boundary and how this margin evolves over the training of a deep neural network. By conducting experiments on MNIST, FASHION-MNIST, and CIFAR-10, we observe that the decision boundary moves closer to natural images over training. This phenomenon even remains intact in the late epochs of training, where the classifier already obtains low training and test error rates. On the other hand, adversarial training appears to have the potential to prevent this undesired convergence of the decision boundary."
                },
                {
                    "title": "FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence",
                    "abstract": "Semi-supervised learning (SSL) provides an effective means of leveraging unlabeled data to improve a model's performance. In this paper, we demonstrate the power of a simple combination of two common SSL methods: consistency regularization and pseudo-labeling. Our algorithm, FixMatch, first generates pseudo-labels using the model's predictions on weakly-augmented unlabeled images. For a given image, the pseudo-label is only retained if the model produces a high-confidence prediction. The model is then trained to predict the pseudo-label when fed a strongly-augmented version of the same image. Despite its simplicity, we show that FixMatch achieves state-of-the-art performance across a variety of standard semi-supervised learning benchmarks, including 94.93% accuracy on CIFAR-10 with 250 labels and 88.61% accuracy with 40 -- just 4 labels per class. Since FixMatch bears many similarities to existing SSL methods that achieve worse performance, we carry out an extensive ablation study to tease apart the experimental factors that are most important to FixMatch's success. We make our code available at this https URL."
                },
                {
                    "title": "Characterizing the Decision Boundary of Deep Neural Networks",
                    "abstract": "Deep neural networks and in particular, deep neural classifiers have become an integral part of many modern applications. Despite their practical success, we still have limited knowledge of how they work and the demand for such an understanding is evergrowing. In this regard, one crucial aspect of deep neural network classifiers that can help us deepen our knowledge about their decision-making behavior is to investigate their decision boundaries. Nevertheless, this is contingent upon having access to samples populating the areas near the decision boundary. To achieve this, we propose a novel approach we call Deep Decision boundary Instance Generation (DeepDIG). DeepDIG utilizes a method based on adversarial example generation as an effective way of generating samples near the decision boundary of any deep neural network model. Then, we introduce a set of important principled characteristics that take advantage of the generated instances near the decision boundary to provide multifaceted understandings of deep neural networks. We have performed extensive experiments on multiple representative datasets across various deep neural network models and characterized their decision boundaries. The code is publicly available at this https URL."
                },
                {
                    "title": "Minimally distorted Adversarial Examples with a Fast Adaptive Boundary Attack",
                    "abstract": "The evaluation of robustness against adversarial manipulation of neural networks-based classifiers is mainly tested with empirical attacks as methods for the exact computation, even when available, do not scale to large networks. We propose in this paper a new white-box adversarial attack wrt the $l_p$-norms for $p \\in \\{1,2,\\infty\\}$ aiming at finding the minimal perturbation necessary to change the class of a given input. It has an intuitive geometric meaning, yields quickly high quality results, minimizes the size of the perturbation (so that it returns the robust accuracy at every threshold with a single run). It performs better or similar to state-of-the-art attacks which are partially specialized to one $l_p$-norm, and is robust to the phenomenon of gradient masking."
                },
                {
                    "title": "Geometric Losses for Distributional Learning",
                    "abstract": "Building upon recent advances in entropy-regularized optimal transport, and upon Fenchel duality between measures and continuous functions , we propose a generalization of the logistic loss that incorporates a metric or cost between classes. Unlike previous attempts to use optimal transport distances for learning, our loss results in unconstrained convex objective functions, supports infinite (or very large) class spaces, and naturally defines a geometric generalization of the softmax operator. The geometric properties of this loss make it suitable for predicting sparse and singular distributions, for instance supported on curves or hyper-surfaces. We study the theoretical properties of our loss and show-case its effectiveness on two applications: ordinal regression and drawing generation."
                },
                {
                    "title": "Sparse Sequence-to-Sequence Models",
                    "abstract": "Sequence-to-sequence models are a powerful workhorse of NLP. Most variants employ a softmax transformation in both their attention mechanism and output layer, leading to dense alignments and strictly positive output probabilities. This density is wasteful, making models less interpretable and assigning probability mass to many implausible outputs. In this paper, we propose sparse sequence-to-sequence models, rooted in a new family of \\alpha-entmax transformations, which includes softmax and sparsemax as particular cases, and is sparse for any \\alpha > 1. We provide fast algorithms to evaluate these transformations and their gradients, which scale well for large vocabulary sizes. Our models are able to produce sparse alignments and to assign nonzero probability to a short list of plausible outputs, sometimes rendering beam search exact. Experiments on morphological inflection and machine translation reveal consistent gains over dense models."
                },
                {
                    "title": "AdaCos: Adaptively Scaling Cosine Logits for Effectively Learning Deep Face Representations",
                    "abstract": "The cosine-based softmax losses and their variants achieve great success in deep learning based face recognition. However, hyperparameter settings in these losses have significant influences on the optimization path as well as the final recognition performance. Manually tuning those hyperparameters heavily relies on user experience and requires many training tricks. In this paper, we investigate in depth the effects of two important hyperparameters of cosine-based softmax losses, the scale parameter and angular margin parameter, by analyzing how they modulate the predicted classification probability. Based on these analysis, we propose a novel cosine-based softmax loss, AdaCos, which is hyperparameter-free and leverages an adaptive scale parameter to automatically strengthen the training supervisions during the training process. We apply the proposed AdaCos loss to large-scale face verification and identification datasets, including LFW, MegaFace, and IJB-C 1:1 Verification. Our results show that training deep neural networks with the AdaCos loss is stable and able to achieve high face recognition accuracy. Our method outperforms state-of-the-art softmax losses on all the three datasets."
                },
                {
                    "title": "Benchmarking Neural Network Robustness to Common Corruptions and Perturbations",
                    "abstract": "In this paper we establish rigorous benchmarks for image classifier robustness. Our first benchmark, ImageNet-C, standardizes and expands the corruption robustness topic, while showing which classifiers are preferable in safety-critical applications. Then we propose a new dataset called ImageNet-P which enables researchers to benchmark a classifier's robustness to common perturbations. Unlike recent robustness research, this benchmark evaluates performance on common corruptions and perturbations not worst-case adversarial perturbations. We find that there are negligible changes in relative corruption robustness from AlexNet classifiers to ResNet classifiers. Afterward we discover ways to enhance corruption and perturbation robustness. We even find that a bypassed adversarial defense provides substantial common perturbation robustness. Together our benchmarks may aid future work toward networks that robustly generalize."
                },
                {
                    "title": "Do ImageNet Classifiers Generalize to ImageNet?",
                    "abstract": "We build new test sets for the CIFAR-10 and ImageNet datasets. Both benchmarks have been the focus of intense research for almost a decade, raising the danger of overfitting to excessively re-used test sets. By closely following the original dataset creation processes, we test to what extent current classification models generalize to new data. We evaluate a broad range of models and find accuracy drops of 3% - 15% on CIFAR-10 and 11% - 14% on ImageNet. However, accuracy gains on the original test sets translate to larger gains on the new test sets. Our results suggest that the accuracy drops are not caused by adaptivity, but by the models' inability to generalize to slightly \"harder\" images than those found in the original test sets."
                },
                {
                    "title": "Learning with Fenchel-Young Losses",
                    "abstract": "Over the past decades, numerous loss functions have been been proposed for a variety of supervised learning tasks, including regression, classification, ranking, and more generally structured prediction. Understanding the core principles and theoretical properties underpinning these losses is key to choose the right loss for the right problem, as well as to create new losses which combine their strengths. In this paper, we introduce Fenchel-Young losses, a generic way to construct a convex loss function for a regularized prediction function. We provide an in-depth study of their properties in a very broad setting, covering all the aforementioned supervised learning tasks, and revealing new connections between sparsity, generalized entropies, and separation margins. We show that Fenchel-Young losses unify many well-known loss functions and allow to create useful new ones easily. Finally, we derive efficient predictive and training algorithms, making Fenchel-Young losses appealing both in theory and practice."
                },
                {
                    "title": "Moment Matching for Multi-Source Domain Adaptation",
                    "abstract": "Conventional unsupervised domain adaptation (UDA) assumes that training data are sampled from a single domain. This neglects the more practical scenario where training data are collected from multiple sources, requiring multi-source domain adaptation. We make three major contributions towards addressing this problem. First, we collect and annotate by far the largest UDA dataset, called DomainNet, which contains six domains and about 0.6 million images distributed among 345 categories, addressing the gap in data availability for multi-source UDA research. Second, we propose a new deep learning approach, Moment Matching for Multi-Source Domain Adaptation (M3SDA), which aims to transfer knowledge learned from multiple labeled source domains to an unlabeled target domain by dynamically aligning moments of their feature distributions. Third, we provide new theoretical insights specifically for moment matching approaches in both single and multiple source domain adaptation. Extensive experiments are conducted to demonstrate the power of our new dataset in benchmarking state-of-the-art multi-source domain adaptation methods, as well as the advantage of our proposed model. Dataset and Code are available at http://ai.bu.edu/M3SDA/"
                },
                {
                    "title": "ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness",
                    "abstract": "Convolutional Neural Networks (CNNs) are commonly thought to recognise objects by learning increasingly complex representations of object shapes. Some recent studies suggest a more important role of image textures. We here put these conflicting hypotheses to a quantitative test by evaluating CNNs and human observers on images with a texture-shape cue conflict. We show that ImageNet-trained CNNs are strongly biased towards recognising textures rather than shapes, which is in stark contrast to human behavioural evidence and reveals fundamentally different classification strategies. We then demonstrate that the same standard architecture (ResNet-50) that learns a texture-based representation on ImageNet is able to learn a shape-based representation instead when trained on \"Stylized-ImageNet\", a stylized version of ImageNet. This provides a much better fit for human behavioural performance in our well-controlled psychophysical lab setting (nine experiments totalling 48,560 psychophysical trials across 97 observers) and comes with a number of unexpected emergent benefits such as improved object detection performance and previously unseen robustness towards a wide range of image distortions, highlighting advantages of a shape-based representation."
                },
                {
                    "title": "On the Decision Boundary of Deep Neural Networks",
                    "abstract": "While deep learning models and techniques have achieved great empirical success, our understanding of the source of success in many aspects remains very limited. In an attempt to bridge the gap, we investigate the decision boundary of a production deep learning architecture with weak assumptions on both the training data and the model. We demonstrate, both theoretically and empirically, that the last weight layer of a neural network converges to a linear SVM trained on the output of the last hidden layer, for both the binary case and the multi-class case with the commonly used cross-entropy loss. Furthermore, we show empirically that training a neural network as a whole, instead of only fine-tuning the last weight layer, may result in better bias constant for the last weight layer, which is important for generalization. In addition to facilitating the understanding of deep learning, our result can be helpful for solving a broad range of practical problems of deep learning, such as catastrophic forgetting and adversarial attacking. The experiment codes are available at this https URL"
                },
                {
                    "title": "Representation Learning with Contrastive Predictive Coding",
                    "abstract": "While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments."
                },
                {
                    "title": "Empirical Study of the Topology and Geometry of Deep Networks",
                    "abstract": "The goal of this paper is to analyze the geometric properties of deep neural network image classifiers in the input space. We specifically study the topology of classification regions created by deep networks, as well as their associated decision boundary. Through a systematic empirical study, we show that state-of-the-art deep nets learn connected classification regions, and that the decision boundary in the vicinity of datapoints is flat along most directions. We further draw an essential connection between two seemingly unrelated properties of deep networks: their sensitivity to additive perturbations of the inputs, and the curvature of their decision boundary. The directions where the decision boundary is curved in fact characterize the directions to which the classifier is the most vulnerable. We finally leverage a fundamental asymmetry in the curvature of the decision boundary of deep nets, and propose a method to discriminate between original images, and images perturbed with small adversarial examples. We show the effectiveness of this purely geometric approach for detecting small adversarial perturbations in images, and for recovering the labels of perturbed images."
                },
                {
                    "title": "Learning Classifiers with Fenchel-Young Losses: Generalized Entropies, Margins, and Algorithms",
                    "abstract": "This paper studies Fenchel-Young losses, a generic way to construct convex loss functions from a regularization function. We analyze their properties in depth, showing that they unify many well-known loss functions and allow to create useful new ones easily. Fenchel-Young losses constructed from a generalized entropy, including the Shannon and Tsallis entropies, induce predictive probability distributions. We formulate conditions for a generalized entropy to yield losses with a separation margin, and probability distributions with sparse support. Finally, we derive efficient algorithms, making Fenchel-Young losses appealing both in theory and practice."
                },
                {
                    "title": "Knowledge Distillation with Adversarial Samples Supporting Decision Boundary",
                    "abstract": "Many recent works on knowledge distillation have provided ways to transfer the knowledge of a trained network for improving the learning process of a new one, but finding a good technique for knowledge distillation is still an open problem. In this paper, we provide a new perspective based on a decision boundary, which is one of the most important component of a classifier. The generalization performance of a classifier is closely related to the adequacy of its decision boundary, so a good classifier bears a good decision boundary. Therefore, transferring information closely related to the decision boundary can be a good attempt for knowledge distillation. To realize this goal, we utilize an adversarial attack to discover samples supporting a decision boundary. Based on this idea, to transfer more accurate information about the decision boundary, the proposed algorithm trains a student classifier based on the adversarial samples supporting the decision boundary. Experiments show that the proposed method indeed improves knowledge distillation and achieves the state-of-the-arts performance."
                },
                {
                    "title": "Unsupervised Feature Learning via Non-parametric Instance Discrimination",
                    "abstract": "Neural net classifiers trained on data with annotated class labels can also capture apparent visual similarity among categories without being directed to do so. We study whether this observation can be extended beyond the conventional domain of supervised learning: Can we learn a good feature representation that captures apparent similarity among instances, instead of classes, by merely asking the feature to be discriminative of individual instances? We formulate this intuition as a non-parametric classification problem at the instance-level, and use noise-contrastive estimation to tackle the computational challenges imposed by the large number of instance classes. Our experimental results demonstrate that, under unsupervised learning settings, our method surpasses the state-of-the-art on ImageNet classification by a large margin. Our method is also remarkable for consistently improving test performance with more training data and better network architectures. By fine-tuning the learned feature, we further obtain competitive results for semi-supervised learning and object detection tasks. Our non-parametric model is highly compact: With 128 features per image, our method requires only 600MB storage for a million images, enabling fast nearest neighbour retrieval at the run time."
                },
                {
                    "title": "ArcFace: Additive Angular Margin Loss for Deep Face Recognition",
                    "abstract": "One of the main challenges in feature learning using Deep Convolutional Neural Networks (DCNNs) for large-scale face recognition is the design of appropriate loss functions that can enhance the discriminative power. Centre loss penalises the distance between deep features and their corresponding class centres in the Euclidean space to achieve intra-class compactness. SphereFace assumes that the linear transformation matrix in the last fully connected layer can be used as a representation of the class centres in the angular space and therefore penalises the angles between deep features and their corresponding weights in a multiplicative way. Recently, a popular line of research is to incorporate margins in well-established loss functions in order to maximise face class separability. In this paper, we propose an Additive Angular Margin Loss (ArcFace) to obtain highly discriminative features for face recognition. The proposed ArcFace has a clear geometric interpretation due to its exact correspondence to geodesic distance on a hypersphere. We present arguably the most extensive experimental evaluation against all recent state-of-the-art face recognition methods on ten face recognition benchmarks which includes a new large-scale image database with trillions of pairs and a large-scale video dataset. We show that ArcFace consistently outperforms the state of the art and can be easily implemented with negligible computational overhead. To facilitate future research, the code has been made available."
                },
                {
                    "title": "Deeper, Broader and Artier Domain Generalization",
                    "abstract": "The problem of domain generalization is to learn from multiple training domains, and extract a domain-agnostic model that can then be applied to an unseen domain. Domain generalization (DG) has a clear motivation in contexts where there are target domains with distinct characteristics, yet sparse data for training. For example recognition in sketch images, which are distinctly more abstract and rarer than photos. Nevertheless, DG methods have primarily been evaluated on photo-only benchmarks focusing on alleviating the dataset bias where both problems of domain distinctiveness and data sparsity can be minimal. We argue that these benchmarks are overly straightforward, and show that simple deep learning baselines perform surprisingly well on them. In this paper, we make two main contributions: Firstly, we build upon the favorable domain shift-robust properties of deep learning methods, and develop a low-rank parameterized CNN model for end-to-end DG learning. Secondly, we develop a DG benchmark dataset covering photo, sketch, cartoon and painting domains. This is both more practically relevant, and harder (bigger domain shift) than existing benchmarks. The results show that our method outperforms existing DG alternatives, and our dataset provides a more significant DG challenge to drive future research."
                },
                {
                    "title": "Noisy Softmax: Improving the Generalization Ability of DCNN via Postponing the Early Softmax Saturation",
                    "abstract": "Over the past few years, softmax and SGD have become a commonly used component and the default training strategy in CNN frameworks, respectively. However, when optimizing CNNs with SGD, the saturation behavior behind softmax always gives us an illusion of training well and then is omitted. In this paper, we first emphasize that the early saturation behavior of softmax will impede the exploration of SGD, which sometimes is a reason for model converging at a bad local-minima, then propose Noisy Softmax to mitigating this early saturation issue by injecting annealed noise in softmax during each iteration. This operation based on noise injection aims at postponing the early saturation and further bringing continuous gradients propagation so as to significantly encourage SGD solver to be more exploratory and help to find a better local-minima. This paper empirically verifies the superiority of the early softmax desaturation, and our method indeed improves the generalization ability of CNN model by regularization. We experimentally find that this early desaturation helps optimization in many tasks, yielding state-of-the-art or competitive results on several popular benchmark datasets."
                },
                {
                    "title": "Deep Hashing Network for Unsupervised Domain Adaptation",
                    "abstract": "In recent years, deep neural networks have emerged as a dominant machine learning tool for a wide variety of application domains. However, training a deep neural network requires a large amount of labeled data, which is an expensive process in terms of time, labor and human expertise. Domain adaptation or transfer learning algorithms address this challenge by leveraging labeled data in a different, but related source domain, to develop a model for the target domain. Further, the explosive growth of digital data has posed a fundamental challenge concerning its storage and retrieval. Due to its storage and retrieval efficiency, recent years have witnessed a wide application of hashing in a variety of computer vision applications. In this paper, we first introduce a new dataset, Office-Home, to evaluate domain adaptation algorithms. The dataset contains images of a variety of everyday objects from multiple domains. We then propose a novel deep learning framework that can exploit labeled source data and unlabeled target data to learn informative hash codes, to accurately classify unseen target data. To the best of our knowledge, this is the first research effort to exploit the feature learning capabilities of deep neural networks to learn representative hash codes to address the domain adaptation problem. Our extensive empirical studies on multiple transfer tasks corroborate the usefulness of the framework in learning efficient hash codes which outperform existing competitive baselines for unsupervised domain adaptation."
                },
                {
                    "title": "Estimating Accuracy from Unlabeled Data: A Probabilistic Logic Approach",
                    "abstract": "We propose an efficient method to estimate the accuracy of classifiers using only unlabeled data. We consider a setting with multiple classification problems where the target classes may be tied together through logical constraints. For example, a set of classes may be mutually exclusive, meaning that a data instance can belong to at most one of them. The proposed method is based on the intuition that: (i) when classifiers agree, they are more likely to be correct, and (ii) when the classifiers make a prediction that violates the constraints, at least one classifier must be making an error. Experiments on four real-world data sets produce accuracy estimates within a few percent of the true accuracy, using solely unlabeled data. Our models also outperform existing state-of-the-art solutions in both estimating accuracies, and combining multiple classifier outputs. The results emphasize the utility of logical constraints in estimating accuracy, thus validating our intuition."
                },
                {
                    "title": "SphereFace: Deep Hypersphere Embedding for Face Recognition",
                    "abstract": "This paper addresses deep face recognition (FR) problem under open-set protocol, where ideal face features are expected to have smaller maximal intra-class distance than minimal inter-class distance under a suitably chosen metric space. However, few existing algorithms can effectively achieve this criterion. To this end, we propose the angular softmax (A-Softmax) loss that enables convolutional neural networks (CNNs) to learn angularly discriminative features. Geometrically, A-Softmax loss can be viewed as imposing discriminative constraints on a hypersphere manifold, which intrinsically matches the prior that faces also lie on a manifold. Moreover, the size of angular margin can be quantitatively adjusted by a parameter m. We further derive specific m to approximate the ideal feature criterion. Extensive analysis and experiments on Labeled Face in the Wild (LFW), Youtube Faces (YTF) and MegaFace Challenge 1 show the superiority of A-Softmax loss in FR tasks."
                },
                {
                    "title": "NormFace: L2 Hypersphere Embedding for Face Verification",
                    "abstract": "Thanks to the recent developments of Convolutional Neural Networks, the performance of face verification methods has increased rapidly. In a typical face verification method, feature normalization is a critical step for boosting performance. This motivates us to introduce and study the effect of normalization during training. But we find this is non-trivial, despite normalization being differentiable. We identify and study four issues related to normalization through mathematical analysis, which yields understanding and helps with parameter settings. Based on this analysis we propose two strategies for training using normalized features. The first is a modification of softmax loss, which optimizes cosine similarity instead of inner-product. The second is a reformulation of metric learning by introducing an agent vector for each class. We show that both strategies, and small variants, consistently improve performance by between 0.2% to 0.4% on the LFW dataset based on two models. This is significant because the performance of the two models on LFW dataset is close to saturation at over 98%."
                },
                {
                    "title": "L2-constrained Softmax Loss for Discriminative Face Verification",
                    "abstract": "In recent years, the performance of face verification systems has significantly improved using deep convolutional neural networks (DCNNs). A typical pipeline for face verification includes training a deep network for subject classification with softmax loss, using the penultimate layer output as the feature descriptor, and generating a cosine similarity score given a pair of face images. The softmax loss function does not optimize the features to have higher similarity score for positive pairs and lower similarity score for negative pairs, which leads to a performance gap. In this paper, we add an L2-constraint to the feature descriptors which restricts them to lie on a hypersphere of a fixed radius. This module can be easily implemented using existing deep learning frameworks. We show that integrating this simple step in the training pipeline significantly boosts the performance of face verification. Specifically, we achieve state-of-the-art results on the challenging IJB-A dataset, achieving True Accept Rate of 0.909 at False Accept Rate 0.0001 on the face verification protocol. Additionally, we achieve state-of-the-art performance on LFW dataset with an accuracy of 99.78%, and competing performance on YTF dataset with accuracy of 96.08%."
                },
                {
                    "title": "Sharp Minima Can Generalize For Deep Nets",
                    "abstract": "Despite their overwhelming capacity to overfit, deep learning architectures tend to generalize relatively well to unseen data, allowing them to be deployed in practice. However, explaining why this is the case is still an open area of research. One standing hypothesis that is gaining popularity, e.g. Hochreiter & Schmidhuber (1997); Keskar et al. (2017), is that the flatness of minima of the loss function found by stochastic gradient based methods results in good generalization. This paper argues that most notions of flatness are problematic for deep models and can not be directly applied to explain generalization. Specifically, when focusing on deep networks with rectifier units, we can exploit the particular geometry of parameter space induced by the inherent symmetries that these architectures exhibit to build equivalent models corresponding to arbitrarily sharper minima. Furthermore, if we allow to reparametrize a function, the geometry of its parameters can change drastically without affecting its generalization properties."
                },
                {
                    "title": "Improved Deep Metric Learning with Multi-class N-pair Loss Objective",
                    "abstract": "Deep metric learning has gained much popularity in recent years, following the success of deep learning. However, existing frameworks of deep metric learning based on contrastive loss and triplet loss often suffer from slow convergence, partially because they employ only one negative example while not interacting with the other negative classes in each update. In this paper, we propose to address this problem with a new metric learning objective called multi-class N-pair loss. The proposed objective function firstly generalizes triplet loss by allowing joint comparison among more than one negative examples - more specifically, N-1 negative examples - and secondly reduces the computational burden of evaluating deep embedding vectors via an efficient batch construction strategy using only N pairs of examples, instead of (N+1) x N. We demonstrate the superiority of our proposed loss to the triplet loss as well as other competing loss functions for a variety of tasks on several visual recognition benchmark, including fine-grained object recognition and verification, image clustering and retrieval, and face verification and identification."
                },
                {
                    "title": "Entropy-SGD: biasing gradient descent into wide valleys",
                    "abstract": "This paper proposes a new optimization algorithm called Entropy-SGD for training deep neural networks that is motivated by the local geometry of the energy landscape. Local extrema with low generalization error have a large proportion of almost-zero eigenvalues in the Hessian with very few positive or negative eigenvalues. We leverage upon this observation to construct a local-entropy-based objective function that favors well-generalizable solutions lying in large flat regions of the energy landscape, while avoiding poorly-generalizable solutions located in the sharp valleys. Conceptually, our algorithm resembles two nested loops of SGD where we use Langevin dynamics in the inner loop to compute the gradient of the local entropy before each update of the weights. We show that the new objective has a smoother energy landscape and show improved generalization over SGD using uniform stability, under certain assumptions. Our experiments on convolutional and recurrent networks demonstrate that Entropy-SGD compares favorably to state-of-the-art techniques in terms of generalization error and training time."
                },
                {
                    "title": "Topology and Geometry of Half-Rectified Network Optimization",
                    "abstract": "The loss surface of deep neural networks has recently attracted interest in the optimization and machine learning communities as a prime example of high-dimensional non-convex problem. Some insights were recently gained using spin glass models and mean-field approximations, but at the expense of strongly simplifying the nonlinear nature of the model. \nIn this work, we do not make any such assumption and study conditions on the data distribution and model architecture that prevent the existence of bad local minima. Our theoretical work quantifies and formalizes two important \\emph{folklore} facts: (i) the landscape of deep linear networks has a radically different topology from that of deep half-rectified ones, and (ii) that the energy landscape in the non-linear case is fundamentally controlled by the interplay between the smoothness of the data distribution and model over-parametrization. Our main theoretical contribution is to prove that half-rectified single layer networks are asymptotically connected, and we provide explicit bounds that reveal the aforementioned interplay. \nThe conditioning of gradient descent is the next challenge we address. We study this question through the geometry of the level sets, and we introduce an algorithm to efficiently estimate the regularity of such sets on large-scale networks. Our empirical results show that these level sets remain connected throughout all the learning phase, suggesting a near convex behavior, but they become exponentially more curvy as the energy level decays, in accordance to what is observed in practice with very low curvature attractors."
                },
                {
                    "title": "A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks",
                    "abstract": "We consider the two related problems of detecting if an example is misclassified or out-of-distribution. We present a simple baseline that utilizes probabilities from softmax distributions. Correctly classified examples tend to have greater maximum softmax probabilities than erroneously classified and out-of-distribution examples, allowing for their detection. We assess performance by defining several tasks in computer vision, natural language processing, and automatic speech recognition, showing the effectiveness of this baseline across all. We then show the baseline can sometimes be surpassed, demonstrating the room for future research on these underexplored detection tasks."
                },
                {
                    "title": "Tsallis Regularized Optimal Transport and Ecological Inference",
                    "abstract": "\n \n Optimal transport is a powerful framework for computing distances between probability distributions. We unify the two main approaches to optimal transport, namely Monge-Kantorovitch and Sinkhorn-Cuturi, into what we define as Tsallis regularized optimal transport (TROT). TROT interpolates a rich family of distortions from Wasserstein to Kullback-Leibler, encompassing as well Pearson, Neyman and Hellinger divergences, to name a few. We show that metric properties known for Sinkhorn-Cuturi generalize to TROT, and provide efficient algorithms for finding the optimal transportation plan with formal convergence proofs. We also present the first application of optimal transport to the problem of ecological inference, that is, the reconstruction of joint distributions from their marginals, a problem of large interest in the social sciences. TROT provides a convenient framework for ecological inference by allowing to compute the joint distribution -\u2014 that is, the optimal transportation plan itself \u2014 when side information is available, which is e.g. typically what census represents in political science. Experiments on data from the 2012 US presidential elections display the potential of TROT in delivering a faithful reconstruction of the joint distribution of ethnic groups and voter preferences.\n \n"
                },
                {
                    "title": "SGDR: Stochastic Gradient Descent with Warm Restarts",
                    "abstract": "Restart techniques are common in gradient-free optimization to deal with multimodal functions. Partial warm restarts are also gaining popularity in gradient-based optimization to improve the rate of convergence in accelerated gradient schemes to deal with ill-conditioned functions. In this paper, we propose a simple warm restart technique for stochastic gradient descent to improve its anytime performance when training deep neural networks. We empirically study its performance on the CIFAR-10 and CIFAR-100 datasets, where we demonstrate new state-of-the-art results at 3.14% and 16.21%, respectively. We also demonstrate its advantages on a dataset of EEG recordings and on a downsampled version of the ImageNet dataset. Our source code is available at this https URL"
                },
                {
                    "title": "Estimating Accuracy from Unlabeled Data: A Bayesian Approach",
                    "abstract": "We consider the question of how unlabeled data can be used to estimate the true accuracy of learned classifiers, and the related question of how outputs from several classifiers performing the same task can be combined based on their estimated accuracies. To answer these questions, we first present a simple graphical model that performs well in practice. We then provide two nonparametric extensions to it that improve its performance. Experiments on two real-world data sets produce accuracy estimates within a few percent of the true accuracy, using solely unlabeled data. Our models also outperform existing state-of-the-art solutions in both estimating accuracies, and combining multiple classifier outputs."
                },
                {
                    "title": "Exponential expressivity in deep neural networks through transient chaos",
                    "abstract": "We combine Riemannian geometry with the mean field theory of high dimensional chaos to study the nature of signal propagation in generic, deep neural networks with random weights. Our results reveal an order-to-chaos expressivity phase transition, with networks in the chaotic phase computing nonlinear functions whose global curvature grows exponentially with depth but not width. We prove this generic class of deep random functions cannot be efficiently computed by any shallow network, going beyond prior work restricted to the analysis of single functions. Moreover, we formalize and quantitatively demonstrate the long conjectured idea that deep networks can disentangle highly curved manifolds in input space into flat manifolds in hidden space. Our theoretical analysis of the expressive power of deep networks broadly applies to arbitrary nonlinearities, and provides a quantitative underpinning for previously abstract notions about the geometry of deep functions."
                },
                {
                    "title": "Wide Residual Networks",
                    "abstract": "Deep residual networks were shown to be able to scale up to thousands of layers and still have improving performance. However, each fraction of a percent of improved accuracy costs nearly doubling the number of layers, and so training very deep residual networks has a problem of diminishing feature reuse, which makes these networks very slow to train. To tackle these problems, in this paper we conduct a detailed experimental study on the architecture of ResNet blocks, based on which we propose a novel architecture where we decrease depth and increase width of residual networks. We call the resulting network structures wide residual networks (WRNs) and show that these are far superior over their commonly used thin and very deep counterparts. For example, we demonstrate that even a simple 16-layer-deep wide residual network outperforms in accuracy and efficiency all previous deep residual networks, including thousand-layer-deep networks, achieving new state-of-the-art results on CIFAR, SVHN, COCO, and significant improvements on ImageNet. Our code and models are available at this https URL"
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "An Exploration of Softmax Alternatives Belonging to the Spherical Loss Family",
                    "abstract": "In a multi-class classification problem, it is standard to model the output of a neural network as a categorical distribution conditioned on the inputs. The output must therefore be positive and sum to one, which is traditionally enforced by a softmax. This probabilistic mapping allows to use the maximum likelihood principle, which leads to the well-known log-softmax loss. However the choice of the softmax function seems somehow arbitrary as there are many other possible normalizing functions. It is thus unclear why the log-softmax loss would perform better than other loss alternatives. In particular Vincent et al. (2015) recently introduced a class of loss functions, called the spherical family, for which there exists an efficient algorithm to compute the updates of the output weights irrespective of the output size. In this paper, we explore several loss functions from this family as possible alternatives to the traditional log-softmax. In particular, we focus our investigation on spherical bounds of the log-softmax loss and on two spherical log-likelihood losses, namely the log-Spherical Softmax suggested by Vincent et al. (2015) and the log-Taylor Softmax that we introduce. Although these alternatives do not yield as good results as the log-softmax loss on two language modeling tasks, they surprisingly outperform it in our experiments on MNIST and CIFAR-10, suggesting that they might be relevant in a broad range of applications."
                },
                {
                    "title": "Orbit Regularization",
                    "abstract": "We propose a general framework for regularization based on group-induced ma-jorization. In this framework, a group is defined to act on the parameter space and an orbit is fixed; to control complexity, the model parameters are confined to the convex hull of this orbit (the orbitope). We recover several well-known regularizes as particular cases, and reveal a connection between the hyperoctahedral group and the recently proposed sorted l1-norm. We derive the properties a group must satisfy for being amenable to optimization with conditional and projected gradient algorithms. Finally, we suggest a continuation strategy for orbit exploration, presenting simulation results for the symmetric and hyperoctahedral groups."
                },
                {
                    "title": "The Loss Surfaces of Multilayer Networks",
                    "abstract": "We study the connection between the highly non-convex loss function of a simple model of the fully-connected feed-forward neural network and the Hamiltonian of the spherical spin-glass model under the assumptions of: i) variable independence, ii) redundancy in network parametrization, and iii) uniformity. These assumptions enable us to explain the complexity of the fully decoupled neural network through the prism of the results from random matrix theory. We show that for large-size decoupled networks the lowest critical values of the random loss function form a layered structure and they are located in a well-defined band lower-bounded by the global minimum. The number of local minima outside that band diminishes exponentially with the size of the network. We empirically verify that the mathematical model exhibits similar behavior as the computer simulations, despite the presence of high dependencies in real networks. We conjecture that both simulated annealing and SGD converge to the band of low critical points, and that all critical points found there are local minima of high quality measured by the test error. This emphasizes a major difference between largeand small-size networks where for the latter poor quality local minima have nonzero probability of being recovered. Finally, we prove that recovering the global minimum becomes harder as the network size increases and that it is in practice irrelevant as global minimum often leads to overfitting."
                },
                {
                    "title": "Identifying and attacking the saddle point problem in high-dimensional non-convex optimization",
                    "abstract": "A central challenge to many fields of science and engineering involves minimizing non-convex error functions over continuous, high dimensional spaces. Gradient descent or quasi-Newton methods are almost ubiquitously used to perform such minimizations, and it is often thought that a main source of difficulty for these local methods to find the global minimum is the proliferation of local minima with much higher error than the global minimum. Here we argue, based on results from statistical physics, random matrix theory, neural network theory, and empirical evidence, that a deeper and more profound difficulty originates from the proliferation of saddle points, not local minima, especially in high dimensional problems of practical interest. Such saddle points are surrounded by high error plateaus that can dramatically slow down learning, and give the illusory impression of the existence of a local minimum. Motivated by these arguments, we propose a new approach to second-order optimization, the saddle-free Newton method, that can rapidly escape high dimensional saddle points, unlike gradient descent and quasi-Newton methods. We apply this algorithm to deep or recurrent neural network training, and provide numerical evidence for its superior optimization performance."
                },
                {
                    "title": "On the Number of Linear Regions of Deep Neural Networks",
                    "abstract": "We study the complexity of functions computable by deep feedforward neural networks with piecewise linear activations in terms of the symmetries and the number of linear regions that they have. Deep networks are able to sequentially map portions of each layer's input-space to the same output. In this way, deep models compute functions that react equally to complicated patterns of different inputs. The compositional structure of these functions enables them to re-use pieces of computation exponentially often in terms of the network's depth. This paper investigates the complexity of such compositional maps and contributes new theoretical results regarding the advantage of depth for neural networks with piecewise linear activation functions. In particular, our analysis is not specific to a single family of models, and as an example, we employ it for rectifier and maxout networks. We improve complexity bounds from pre-existing work and investigate the behavior of units in higher layers."
                },
                {
                    "title": "PAC-Bayesian Analysis of Co-clustering and Beyond",
                    "abstract": "We derive PAC-Bayesian generalization bounds for supervised and unsupervised learning models based on clustering, such as co-clustering, matrix tri-factorization, graphical models, graph clustering, and pairwise clustering. We begin with the analysis of co-clustering, which is a widely used approach to the analysis of data matrices. We distinguish among two tasks in matrix data analysis: discriminative prediction of the missing entries in data matrices and estimation of the joint probability distribution of row and column variables in co-occurrence matrices. We derive PAC-Bayesian generalization bounds for the expected out-of-sample performance of co-clustering-based solutions for these two tasks. The analysis yields regularization terms that were absent in the previous formulations of co-clustering. The bounds suggest that the expected performance of co-clustering is governed by a trade-off between its empirical performance and the mutual information preserved by the cluster variables on row and column IDs. We derive an iterative projection algorithm for finding a local optimum of this trade-off for discriminative prediction tasks. This algorithm achieved state-of-the-art performance in the MovieLens collaborative filtering task. Our co-clustering model can also be seen as matrix tri-factorization and the results provide generalization bounds, regularization terms, and new algorithms for this form of matrix factorization. The analysis of co-clustering is extended to tree-shaped graphical models, which can be used to analyze high dimensional tensors. According to the bounds, the generalization abilities of tree-shaped graphical models depend on a trade-off between their empirical data fit and the mutual information that is propagated up the tree levels. We also formulate weighted graph clustering as a prediction problem: given a subset of edge weights we analyze the ability of graph clustering to predict the remaining edge weights. The analysis of co-clustering easily extends to this problem and suggests that graph clustering should optimize the trade-off between empirical data fit and the mutual information that clusters preserve on graph nodes."
                },
                {
                    "title": "Unsupervised Supervised Learning I: Estimating Classification and Regression Errors without Labels",
                    "abstract": "Estimating the error rates of classifiers or regression models is a fundamental task in machine learning which has thus far been studied exclusively using supervised learning techniques. We propose a novel unsupervised framework for estimating these error rates using only unlabeled data and mild assumptions. We prove consistency results for the framework and demonstrate its practical applicability on both synthetic and real world data."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Dataset Shift in Machine Learning",
                    "abstract": "Dataset shift is a common problem in predictive modeling that occurs when the joint distribution of inputs and outputs differs between training and test stages. Covariate shift, a particular case of dataset shift, occurs when only the input distribution changes. Dataset shift is present in most practical applications, for reasons ranging from the bias introduced by experimental design to the irreproducibility of the testing conditions at training time. (An example is -email spam filtering, which may fail to recognize spam that differs in form from the spam the automatic filter has been built on.) Despite this, and despite the attention given to the apparently similar problems of semi-supervised learning and active learning, dataset shift has received relatively little attention in the machine learning community until recently. This volume offers an overview of current efforts to deal with dataset and covariate shift. The chapters offer a mathematical and philosophical introduction to the problem, place dataset shift in relationship to transfer learning, transduction, local learning, active learning, and semi-supervised learning, provide theoretical views of dataset and covariate shift (including decision theoretic and Bayesian perspectives), and present algorithms for covariate shift. Contributors: Shai Ben-David, Steffen Bickel, Karsten Borgwardt, Michael Brckner, David Corfield, Amir Globerson, Arthur Gretton, Lars Kai Hansen, Matthias Hein, Jiayuan Huang, Takafumi Kanamori, Klaus-Robert Mller, Sam Roweis, Neil Rubens, Tobias Scheffer, Marcel Schmittfull, Bernhard Schlkopf, Hidetoshi Shimodaira, Alex Smola, Amos Storkey, Masashi Sugiyama, Choon Hui Teo Neural Information Processing series"
                },
                {
                    "title": "Co-Validation: Using Model Disagreement on Unlabeled Data to Validate Classification Algorithms",
                    "abstract": "In the context of binary classification, we define disagreement as a measure of how often two independently-trained models differ in their classification of unlabeled data. We explore the use of disagreement for error estimation and model selection. We call the procedure co-validation, since the two models effectively (invalidate one another by comparing results on unlabeled data, which we assume is relatively cheap and plentiful compared to labeled data. We show that per-instance disagreement is an unbiased estimate of the variance of error for that instance. We also show that disagreement provides a lower bound on the prediction (generalization) error, and a tight upper bound on the \"variance of prediction error\", or the variance of the average error across instances, where variance is measured across training sets. We present experimental results on several data sets exploring co-validation for error estimation and model selection. The procedure is especially effective in active learning settings, where training sets are not drawn at random and cross validation overestimates error."
                },
                {
                    "title": "Nonextensive Entropy: Interdisciplinary Applications",
                    "abstract": "A great variety of complex phenomena in many scientific fields exhibit power-law behavior, reflecting a hierarchical or fractal structure. Many of these phenomena seem to be susceptible to description using approaches drawn from thermodynamics or statistical mechanics, particularly approaches involving the maximization of entropy and of Boltzmann-Gibbs statistical mechanics and standard laws in a natural way. The book addresses the interdisciplinary applications of these ideas, and also on various phenomena that could possibly be quantitatively describable in terms of these ideas."
                },
                {
                    "title": "A note on a general definition of the coefficient of determination",
                    "abstract": "SUMMARY A generalization of the coefficient of determination R2 to general regression models is discussed. A modification of an earlier definition to allow for discrete models is proposed."
                },
                {
                    "title": "A New Adversarial Domain Generalization Network Based on Class Boundary Feature Detection for Bearing Fault Diagnosis",
                    "abstract": "In recent years, many researchers have attempted to achieve cross-domain diagnosis of faults through domain adaptation (DA) methods. However, owing to the complex physical environments, applications of DA-based approach are not guaranteed to unknown operating environments. Some existing domain generalization (DG) methods require enough fully labeled source domains to train, which are often unavailable in practical settings. In this study, an adversarial domain generalization network (ADGN) based on class boundary feature detection is proposed. The ADGN can diagnose faults in unknown operating environments, and only one fully labeled domain is used in training. Although ADGN has to access fully unlabeled auxiliary domains, a large number of unlabeled samples exist under actual working conditions. In our method, fuzzy features at a classification boundary are detected by maximizing the classifier differences. Better feature mapping functions and domain-invariant features are obtained by adversarial training. As the training proceeds, the differences in the distribution of features among the source, auxiliary, and unknown domains become smaller so domain-invariant features can be used for fault diagnosis in unknown operating environments. Comprehensive experiments showed that ADGN can achieve higher fault diagnosis accuracies than other methods when only one fully labeled domain is used in an unknown operating environment. The ADGN can even cope comfortably with complex transfer tasks with different speeds and loads."
                },
                {
                    "title": "Tent: Fully Test-Time Adaptation by Entropy Minimization",
                    "abstract": "A model must adapt itself to generalize to new and different data during testing. In this setting of fully test-time adaptation the model has only the test data and its own parameters. We propose to adapt by test entropy minimization (tent1): we optimize the model for confidence as measured by the entropy of its predictions. Our method estimates normalization statistics and optimizes channel-wise affine transformations to update online on each batch. Tent reduces generalization error for image classification on corrupted ImageNet and CIFAR-10/100 and reaches a new state-of-the-art error on ImageNet-C. Tent handles source-free domain adaptation on digit recognition from SVHN to MNIST/MNIST-M/USPS, on semantic segmentation from GTA to Cityscapes, and on the VisDA-C benchmark. These results are achieved in one epoch of test-time optimization without altering training."
                },
                {
                    "title": "R X R X 1: A N IMAGE SET FOR CELLULAR MORPHOLOGICAL VARIATION ACROSS MANY EXPERIMENTAL BATCHES",
                    "abstract": "High-throughput screening techniques are commonly used in many fields of biology. However, it is well known that non-biological artifacts arising from variability in the technical execution of different experimental batches confound highthroughput screens measurements. These batch effects obscure biological conclusions, and it is therefore necessary to account for them. While a number of techniques have been proposed, to our knowledge there is not a publicly available biological dataset designed specifically for the systematic study of batch effect correction. To this end we announce the release of RxRx1, a set of 125,514 highresolution fluorescence microscopy images of human cells under 1,108 genetic perturbations in 51 experimental batches across four cell types. Visual inspection of the images by batch makes it clear that the set indeed demonstrates significant batch effects. In this paper we describe the image set in detail. We also propose a classification task designed to study batch effect correction on these images, and provide some baseline results for the task. Our goal in releasing this image set is to encourage researchers across various disciplines to develop effective methods for removing batch effects that generalize well to unseen experimental batches and to share these methods with the scientific community."
                },
                {
                    "title": "Decision Boundary Analysis of Adversarial Examples",
                    "abstract": "Deep neural networks (DNNs) are vulnerable to adversarial examples, which are carefully crafted instances aiming to cause prediction errors for DNNs. Recent research on adversarial examples has examined local neighborhoods in the input space of DNN models. However, previous work has limited what regions to consider, focusing either on low-dimensional subspaces or small balls. In this paper, we argue that information from larger neighborhoods, such as from more directions and from greater distances, will better characterize the relationship between adversarial examples and the DNN models. First, we introduce an attack, OPTMARGIN, which generates adversarial examples robust to small perturbations. These examples successfully evade a defense that only considers a small ball around an input instance. Second, we analyze a larger neighborhood around input instances by looking at properties of surrounding decision boundaries, namely the distances to the boundaries and the adjacent classes. We find that the boundaries around these adversarial examples do not resemble the boundaries around benign examples. Finally, we show that, under scrutiny of the surrounding decision boundaries, our OPTMARGIN examples do not convincingly mimic benign examples. Although our experiments are limited to a few specific attacks, we hope these findings will motivate new, more evasive attacks and ultimately, effective defenses."
                },
                {
                    "title": "Tiny ImageNet Visual Recognition Challenge",
                    "abstract": "In this work, we investigate the effect of convolutional network depth, receptive field size, dropout layers, rectified activation unit type and dataset noise on its accuracy in Tiny-ImageNet Challenge settings. In order to make a thorough evaluation of the cause of the peformance improvement, we start with a basic 5 layer model with 5\u00d75 convolutional receptive fields. We keep increasing network depth or reducing receptive field size, and continue applying modern techniques, such as PReLu and dropout, to the model. Our model achieves excellent performance even compared to state-of-the-art results, with 0.444 final error rate on the test set."
                },
                {
                    "title": "Pseudo-Label : The Simple and Efficient Semi-Supervised Learning Method for Deep Neural Networks",
                    "abstract": "We propose the simple and e\ufb03cient method of semi-supervised learning for deep neural networks. Basically, the proposed network is trained in a supervised fashion with labeled and unlabeled data simultaneously. For unlabeled data, Pseudo-Label s, just picking up the class which has the maximum network output, are used as if they were true labels. Without any unsupervised pre-training method, this simple method with dropout shows the state-of-the-art performance."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Semi-Supervised Classification by Low Density Separation",
                    "abstract": "We believe that the cluster assumption is key to successful semi-supervised learning. Based on this, we propose three semi-supervised algorithms: 1. deriving graph-based distances that emphazise low density regions between clusters, followed by training a standard SVM; 2. optimizing the Transductive SVM objective function, which places the decision boundary in low density regions, by gradient descent; 3. combining the \ufb02rst two to make maximum use of the cluster assumption. We compare with state of the art algorithms and demonstrate superior accuracy for the latter two methods."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nWhat explains the correlation between logits and generalization performance?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses the fundamental challenge of accurately estimating model performance on out-of-distribution data, which is vital for the safe deployment of machine learning models in real-world applications. By understanding the relationship between logits and generalization performance, researchers can develop more robust models that maintain their effectiveness even when faced with distribution shifts. This advancement could lead to practical applications in various fields, such as healthcare, finance, and autonomous systems, where reliable AI performance is essential. Furthermore, improving accuracy estimation methods can foster trust in AI systems, encouraging broader adoption and innovation in machine learning.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the complex relationship between logits and model performance, particularly under distribution shifts. Naive approaches may fail because they do not account for the intricacies of how logits reflect the model's confidence and the underlying data distribution. Technical obstacles include the need for effective normalization techniques that can mitigate overconfidence in logits, which can lead to biased accuracy estimates. Additionally, theoretical challenges arise from the low-density separation assumption, which requires a deep understanding of decision boundaries and their relationship to data density. Overcoming these complexities is essential for developing reliable unsupervised accuracy estimation methods.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often overlooked the nuanced relationship between logits and generalization performance, leading to gaps in understanding how to effectively utilize logits for accuracy estimation. Existing solutions, such as those relying on softmax normalization, have been limited by their tendency to accumulate errors, particularly in poorly calibrated scenarios. Barriers to solving this problem include a lack of comprehensive methodologies that integrate insights from both logits and decision boundaries. Our approach differs by proposing a novel normalization strategy (SoftTrun) that addresses the shortcomings of traditional methods and provides a more accurate estimation of model performance.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves two key components: (1) MaNo, an estimation score that aggregates the margins at a dataset level using the Lp-norm of the normalized model\u2019s prediction matrix to evaluate the density around decision boundaries, and (2) SoftTrun, a novel normalization strategy that considers the empirical distribution of logits to mitigate overconfidence issues. We will evaluate our methods using benchmark datasets and metrics"
            }
        },
        "author_data": {
            "8e083927-d615-4f41-b2bf-401c522cab52": {
                "pk": "8e083927-d615-4f41-b2bf-401c522cab52",
                "name": "Renchunzi Xie",
                "collaborators": [
                    "Hongxin Wei",
                    "Bo An",
                    "Lei Feng",
                    "Mahardhika Pratama",
                    "Lue Tao",
                    "Yixuan Li",
                    "Huiping Zhuang",
                    "Yuzhou Cao",
                    "Ambroise Odonnat",
                    "Vasilii Feofanov"
                ],
                "domain": [
                    "Transfer Learning",
                    "Domain Adaptation",
                    "Noisy Labels",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Automatic Online Multi-Source Domain Adaptation",
                        "abstract": "Knowledge transfer across several streaming processes remain challenging problem not only because of different distributions of each stream but also because of rapidly changing and never-ending environments of data streams. Albeit growing research achievements in this area, most of existing works are developed for a single source domain which limits its resilience to exploit multi-source domains being beneficial to recover from concept drifts quickly and to avoid the negative transfer problem. An online domain adaptation technique under multisource streaming processes, namely automatic online multi-source domain adaptation (AOMSDA), is proposed in this paper. The online domain adaptation strategy of AOMSDA is formulated under a coupled generative and discriminative approach of denoising autoencoder (DAE) where the central moment discrepancy (CMD)-based regularizer is integrated to handle the existence of multi-source domains thereby taking advantage of complementary information sources. The asynchronous concept drifts taking place at different time periods are addressed by a self-organizing structure and a node re-weighting strategy. Our numerical study demonstrates that AOMSDA is capable of outperforming its counterparts in 5 of 8 study cases while the ablation study depicts the advantage of each learning component. In addition, AOMSDA is general for any number of source streams. The source code of AOMSDA is shared publicly in https://github.com/Renchunzi-Xie/AOMSDA.git."
                    },
                    {
                        "title": "Open-set Label Noise Can Improve Robustness Against Inherent Label Noise",
                        "abstract": "Learning with noisy labels is a practically challenging problem in weakly supervised learning. In the existing literature, open-set noises are always considered to be poisonous for generalization, similar to closed-set noises. In this paper, we empirically show that open-set noisy labels can be non-toxic and even benefit the robustness against inherent noisy labels. Inspired by the observations, we propose a simple yet effective regularization by introducing Open-set samples with Dynamic Noisy Labels (ODNL) into training. With ODNL, the extra capacity of the neural network can be largely consumed in a way that does not interfere with learning patterns from clean data. Through the lens of SGD noise, we show that the noises induced by our method are random-direction, conflict-free and biased, which may help the model converge to a flat minimum with superior stability and enforce the model to produce conservative predictions on Out-of-Distribution instances. Extensive experimental results on benchmark datasets with various types of noisy labels demonstrate that the proposed method not only enhances the performance of many existing robust algorithms but also achieves significant improvement on Out-of-Distribution detection tasks even in the label noise setting."
                    },
                    {
                        "title": "GearNet: Stepwise Dual Learning for Weakly Supervised Domain Adaptation",
                        "abstract": "This paper studies weakly supervised domain adaptation(WSDA) problem, where we only have access to the source domain with noisy labels, from which we need to transfer useful information to the unlabeled target domain. Although there have been a few studies on this problem, most of them only exploit unidirectional relationships from the source domain to the target domain. In this paper, we propose a universal paradigm called GearNet to exploit bilateral relationships between the two domains. Specifically, we take the two domains as different inputs to train two models alternately, and asymmetrical Kullback-Leibler loss is used for selectively matching the predictions of the two models in the same domain. This interactive learning schema enables implicit label noise canceling and exploits correlations between the source and target domains. Therefore, our GearNet has the great potential to boost the performance of a wide range of existing WSDL methods. Comprehensive experimental results show that the performance of existing methods can be significantly improved by equipping with our GearNet."
                    },
                    {
                        "title": "On the Importance of Feature Separability in Predicting Out-Of-Distribution Error",
                        "abstract": "Estimating the generalization performance is practically challenging on out-of-distribution (OOD) data without ground-truth labels. While previous methods emphasize the connection between distribution difference and OOD accuracy, we show that a large domain gap not necessarily leads to a low test accuracy. In this paper, we investigate this problem from the perspective of feature separability empirically and theoretically. Specifically, we propose a dataset-level score based upon feature dispersion to estimate the test accuracy under distribution shift. Our method is inspired by desirable properties of features in representation learning: high inter-class dispersion and high intra-class compactness. Our analysis shows that inter-class dispersion is strongly correlated with the model accuracy, while intra-class compactness does not reflect the generalization performance on OOD data. Extensive experiments demonstrate the superiority of our method in both prediction performance and computational efficiency."
                    },
                    {
                        "title": "Open-Sampling: Exploring Out-of-Distribution data for Re-balancing Long-tailed datasets",
                        "abstract": "Deep neural networks usually perform poorly when the training dataset suffers from extreme class imbalance. Recent studies found that directly training with out-of-distribution data (i.e., open-set samples) in a semi-supervised manner would harm the generalization performance. In this work, we theoretically show that out-of-distribution data can still be leveraged to augment the minority classes from a Bayesian perspective. Based on this motivation, we propose a novel method called Open-sampling, which utilizes open-set noisy labels to re-balance the class priors of the training dataset. For each open-set instance, the label is sampled from our pre-defined distribution that is complementary to the distribution of original class priors. We empirically show that Open-sampling not only re-balances the class priors but also encourages the neural network to learn separable representations. Extensive experiments demonstrate that our proposed method significantly outperforms existing data re-balancing methods and can boost the performance of existing state-of-the-art methods."
                    },
                    {
                        "title": "Leveraging Gradients for Unsupervised Accuracy Estimation under Distribution Shift",
                        "abstract": "Estimating test accuracy without access to the ground-truth test labels under varying test environments is a challenging, yet extremely important problem in the safe deployment of machine learning algorithms. Existing works rely on the information from either the outputs or the extracted features of neural networks to formulate an estimation score correlating with the ground-truth test accuracy. In this paper, we investigate--both empirically and theoretically--how the information provided by the gradients can be predictive of the ground-truth test accuracy even under a distribution shift. Specifically, we use the norm of classification-layer gradients, backpropagated from the cross-entropy loss after only one gradient step over test data. Our key idea is that the model should be adjusted with a higher magnitude of gradients when it does not generalize to the test dataset with a distribution shift. We provide theoretical insights highlighting the main ingredients of such an approach ensuring its empirical success. Extensive experiments conducted on diverse distribution shifts and model structures demonstrate that our method significantly outperforms state-of-the-art algorithms."
                    },
                    {
                        "title": "ATL: Autonomous Knowledge Transfer from Many Streaming Processes",
                        "abstract": "Transferring knowledge across many streaming processes remains an uncharted territory in the existing literature and features unique characteristics: no labelled instance of the target domain, covariate shift of source and target domain, different period of drifts in the source and target domains. Autonomous transfer learning (ATL) is proposed in this paper as a flexible deep learning approach for the online unsupervised transfer learning problem across many streaming processes. ATL offers an online domain adaptation strategy via the generative and discriminative phases coupled with the KL divergence based optimization strategy to produce a domain invariant network while putting forward an elastic network structure. It automatically evolves its network structure from scratch with/without the presence of ground truth to overcome independent concept drifts in the source and target domain. The rigorous numerical evaluation has been conducted along with a comparison against recently published works. ATL demonstrates improved performance while showing significantly faster training speed than its counterparts."
                    },
                    {
                        "title": "Mitigating Neural Network Overconfidence with Logit Normalization",
                        "abstract": "Detecting out-of-distribution inputs is critical for safe deployment of machine learning models in the real world. However, neural networks are known to suffer from the overconfidence issue, where they produce abnormally high confidence for both in- and out-of-distribution inputs. In this work, we show that this issue can be mitigated through Logit Normalization (LogitNorm) -- a simple fix to the cross-entropy loss -- by enforcing a constant vector norm on the logits in training. Our method is motivated by the analysis that the norm of the logit keeps increasing during training, leading to overconfident output. Our key idea behind LogitNorm is thus to decouple the influence of output's norm during network optimization. Trained with LogitNorm, neural networks produce highly distinguishable confidence scores between in- and out-of-distribution data. Extensive experiments demonstrate the superiority of LogitNorm, reducing the average FPR95 by up to 42.30% on common benchmarks."
                    },
                    {
                        "title": "ACIL: Analytic Class-Incremental Learning with Absolute Memorization and Privacy Protection",
                        "abstract": "Class-incremental learning (CIL) learns a classification model with training data of different classes arising progressively. Existing CIL either suffers from serious accuracy loss due to catastrophic forgetting, or invades data privacy by revisiting used exemplars. Inspired by linear learning formulations, we propose an analytic class-incremental learning (ACIL) with absolute memorization of past knowledge while avoiding breaching of data privacy (i.e., without storing historical data). The absolute memorization is demonstrated in the sense that class-incremental learning using ACIL given present data would give identical results to that from its joint-learning counterpart which consumes both present and historical samples. This equality is theoretically validated. Data privacy is ensured since no historical data are involved during the learning process. Empirical validations demonstrate ACIL's competitive accuracy performance with near-identical results for various incremental task settings (e.g., 5-50 phases). This also allows ACIL to outperform the state-of-the-art methods for large-phase scenarios (e.g., 25 and 50 phases)."
                    },
                    {
                        "title": "Mitigating Memorization of Noisy Labels by Clipping the Model Prediction",
                        "abstract": "In the presence of noisy labels, designing robust loss functions is critical for securing the generalization performance of deep neural networks. Cross Entropy (CE) loss has been shown to be not robust to noisy labels due to its unboundedness. To alleviate this issue, existing works typically design specialized robust losses with the symmetric condition, which usually lead to the underfitting issue. In this paper, our key idea is to induce a loss bound at the logit level, thus universally enhancing the noise robustness of existing losses. Specifically, we propose logit clipping (LogitClip), which clamps the norm of the logit vector to ensure that it is upper bounded by a constant. In this manner, CE loss equipped with our LogitClip method is effectively bounded, mitigating the overfitting to examples with noisy labels. Moreover, we present theoretical analyses to certify the noise-tolerant ability of LogitClip. Extensive experiments show that LogitClip not only significantly improves the noise robustness of CE loss, but also broadly enhances the generalization performance of popular robust losses."
                    }
                ]
            },
            "45f8a020-8e0e-4ba5-8cc5-3aec534e2ce4": {
                "pk": "45f8a020-8e0e-4ba5-8cc5-3aec534e2ce4",
                "name": "Ambroise Odonnat",
                "collaborators": [
                    "Ievgen Redko",
                    "Vasilii Feofanov",
                    "Romain Ilbert",
                    "Giuseppe Paolo",
                    "Themis Palpanas",
                    "Oussama Zekri",
                    "Abdelhakim Benechehab",
                    "Aladin Virmaux",
                    "Renchunzi Xie",
                    "Jianfeng Zhang"
                ],
                "domain": [
                    "Semi-Supervised Learning",
                    "Time Series Forecasting",
                    "Domain Adaptation",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Leveraging Ensemble Diversity for Robust Self-Training in the Presence of Sample Selection Bias",
                        "abstract": "Self-training is a well-known approach for semi-supervised learning. It consists of iteratively assigning pseudo-labels to unlabeled data for which the model is confident and treating them as labeled examples. For neural networks, softmax prediction probabilities are often used as a confidence measure, although they are known to be overconfident, even for wrong predictions. This phenomenon is particularly intensified in the presence of sample selection bias, i.e., when data labeling is subject to some constraint. To address this issue, we propose a novel confidence measure, called $\\mathcal{T}$-similarity, built upon the prediction diversity of an ensemble of linear classifiers. We provide the theoretical analysis of our approach by studying stationary points and describing the relationship between the diversity of the individual members and their performance. We empirically demonstrate the benefit of our confidence measure for three different pseudo-labeling policies on classification datasets of various data modalities. The code is available at https://github.com/ambroiseodt/tsim."
                    },
                    {
                        "title": "SAMformer: Unlocking the Potential of Transformers in Time Series Forecasting with Sharpness-Aware Minimization and Channel-Wise Attention",
                        "abstract": "Transformer-based architectures achieved breakthrough performance in natural language processing and computer vision, yet they remain inferior to simpler linear baselines in multivariate long-term forecasting. To better understand this phenomenon, we start by studying a toy linear forecasting problem for which we show that transformers are incapable of converging to their true solution despite their high expressive power. We further identify the attention of transformers as being responsible for this low generalization capacity. Building upon this insight, we propose a shallow lightweight transformer model that successfully escapes bad local minima when optimized with sharpness-aware optimization. We empirically demonstrate that this result extends to all commonly used real-world multivariate time series datasets. In particular, SAMformer surpasses current state-of-the-art methods and is on par with the biggest foundation model MOIRAI while having significantly fewer parameters. The code is available at https://github.com/romilbert/samformer."
                    },
                    {
                        "title": "Leveraging Gradients for Unsupervised Accuracy Estimation under Distribution Shift",
                        "abstract": "Estimating test accuracy without access to the ground-truth test labels under varying test environments is a challenging, yet extremely important problem in the safe deployment of machine learning algorithms. Existing works rely on the information from either the outputs or the extracted features of neural networks to formulate an estimation score correlating with the ground-truth test accuracy. In this paper, we investigate--both empirically and theoretically--how the information provided by the gradients can be predictive of the ground-truth test accuracy even under a distribution shift. Specifically, we use the norm of classification-layer gradients, backpropagated from the cross-entropy loss after only one gradient step over test data. Our key idea is that the model should be adjusted with a higher magnitude of gradients when it does not generalize to the test dataset with a distribution shift. We provide theoretical insights highlighting the main ingredients of such an approach ensuring its empirical success. Extensive experiments conducted on diverse distribution shifts and model structures demonstrate that our method significantly outperforms state-of-the-art algorithms."
                    },
                    {
                        "title": "Analysing Multi-Task Regression via Random Matrix Theory with Application to Time Series Forecasting",
                        "abstract": "In this paper, we introduce a novel theoretical framework for multi-task regression, applying random matrix theory to provide precise performance estimations, under high-dimensional, non-Gaussian data distributions. We formulate a multi-task optimization problem as a regularization technique to enable single-task models to leverage multi-task learning information. We derive a closed-form solution for multi-task optimization in the context of linear models. Our analysis provides valuable insights by linking the multi-task learning performance to various model statistics such as raw data covariances, signal-generating hyperplanes, noise levels, as well as the size and number of datasets. We finally propose a consistent estimation of training and testing errors, thereby offering a robust foundation for hyperparameter optimization in multi-task regression scenarios. Experimental validations on both synthetic and real-world datasets in regression and multivariate time series forecasting demonstrate improvements on univariate models, incorporating our method into the training loss and thus leveraging multivariate information."
                    },
                    {
                        "title": "Large Language Models as Markov Chains",
                        "abstract": "Large language models (LLMs) have proven to be remarkably efficient, both across a wide range of natural language processing tasks and well beyond them. However, a comprehensive theoretical analysis of the origins of their impressive performance remains elusive. In this paper, we approach this challenging task by drawing an equivalence between generic autoregressive language models with vocabulary of size $T$ and context window of size $K$ and Markov chains defined on a finite state space of size $\\mathcal{O}(T^K)$. We derive several surprising findings related to the existence of a stationary distribution of Markov chains that capture the inference power of LLMs, their speed of convergence to it, and the influence of the temperature on the latter. We then prove pre-training and in-context generalization bounds and show how the drawn equivalence allows us to enrich their interpretation. Finally, we illustrate our theoretical guarantees with experiments on several recent LLMs to highlight how they capture the behavior observed in practice."
                    },
                    {
                        "title": "SKADA-Bench: Benchmarking Unsupervised Domain Adaptation Methods with Realistic Validation",
                        "abstract": "Unsupervised Domain Adaptation (DA) consists of adapting a model trained on a labeled source domain to perform well on an unlabeled target domain with some data distribution shift. While many methods have been proposed in the literature, fair and realistic evaluation remains an open question, particularly due to methodological difficulties in selecting hyperparameters in the unsupervised setting. With SKADA-Bench, we propose a framework to evaluate DA methods and present a fair evaluation of existing shallow algorithms, including reweighting, mapping, and subspace alignment. Realistic hyperparameter selection is performed with nested cross-validation and various unsupervised model selection scores, on both simulated datasets with controlled shifts and real-world datasets across diverse modalities, such as images, text, biomedical, and tabular data with specific feature extraction. Our benchmark highlights the importance of realistic validation and provides practical guidance for real-life applications, with key insights into the choice and impact of model selection approaches. SKADA-Bench is open-source, reproducible, and can be easily extended with novel DA methods, datasets, and model selection criteria without requiring re-evaluating competitors. SKADA-Bench is available on GitHub at https://github.com/scikit-adaptation/skada-bench."
                    },
                    {
                        "title": "Zero-shot Model-based Reinforcement Learning using Large Language Models",
                        "abstract": "The emerging zero-shot capabilities of Large Language Models (LLMs) have led to their applications in areas extending well beyond natural language processing tasks. In reinforcement learning, while LLMs have been extensively used in text-based environments, their integration with continuous state spaces remains understudied. In this paper, we investigate how pre-trained LLMs can be leveraged to predict in context the dynamics of continuous Markov decision processes. We identify handling multivariate data and incorporating the control signal as key challenges that limit the potential of LLMs' deployment in this setup and propose Disentangled In-Context Learning (DICL) to address them. We present proof-of-concept applications in two reinforcement learning settings: model-based policy evaluation and data-augmented off-policy reinforcement learning, supported by theoretical analysis of the proposed methods. Our experiments further demonstrate that our approach produces well-calibrated uncertainty estimates. We release the code at https://github.com/abenechehab/dicl."
                    }
                ]
            },
            "2f67578b-2fd4-4143-b196-3d05b6ab5d52": {
                "pk": "2f67578b-2fd4-4143-b196-3d05b6ab5d52",
                "name": "Vasilii Feofanov",
                "collaborators": [
                    "Ievgen Redko",
                    "Ambroise Odonnat",
                    "Malik Tiomoko",
                    "Emilie Devijver",
                    "Massih-Reza Amini",
                    "Romain Ilbert",
                    "Themis Palpanas",
                    "Aladin Virmaux",
                    "Renchunzi Xie",
                    "Jianfeng Zhang"
                ],
                "domain": [
                    "Semi-Supervised Learning",
                    "Multi-Task Learning",
                    "Dimensionality Reduction",
                    "Time Series Forecasting"
                ],
                "publications": [
                    {
                        "title": "Random Matrix Analysis to Balance between Supervised and Unsupervised Learning under the Low Density Separation Assumption",
                        "abstract": "We propose a theoretical framework to analyze semi-supervised classification under the low density separation assumption in a high-dimensional regime. In particular, we introduce QLDS, a linear classification model, where the low density separation assumption is implemented via quadratic margin maximization. The algorithm has an explicit solution with rich theoretical properties, and we show that particular cases of our algorithm are the least-square support vector machine in the supervised case, the spectral clustering in the fully unsupervised regime, and a class of semi-supervised graph-based approaches. As such, QLDS establishes a smooth bridge between these supervised and unsupervised learning methods. Using recent advances in the random matrix theory, we formally derive a theoretical evaluation of the classification error in the asymptotic regime. As an application, we derive a hyperparameter selection policy that finds the best balance between the supervised and the unsupervised terms of our learning criterion. Finally, we provide extensive illustrations of our framework, as well as an experimental study on several benchmarks to demonstrate that QLDS, while being computationally more efficient, improves over cross-validation for hyperparameter selection, indicating a high promise of the usage of random matrix theory for semi-supervised model selection."
                    },
                    {
                        "title": "Leveraging Ensemble Diversity for Robust Self-Training in the Presence of Sample Selection Bias",
                        "abstract": "Self-training is a well-known approach for semi-supervised learning. It consists of iteratively assigning pseudo-labels to unlabeled data for which the model is confident and treating them as labeled examples. For neural networks, softmax prediction probabilities are often used as a confidence measure, although they are known to be overconfident, even for wrong predictions. This phenomenon is particularly intensified in the presence of sample selection bias, i.e., when data labeling is subject to some constraint. To address this issue, we propose a novel confidence measure, called $\\mathcal{T}$-similarity, built upon the prediction diversity of an ensemble of linear classifiers. We provide the theoretical analysis of our approach by studying stationary points and describing the relationship between the diversity of the individual members and their performance. We empirically demonstrate the benefit of our confidence measure for three different pseudo-labeling policies on classification datasets of various data modalities. The code is available at https://github.com/ambroiseodt/tsim."
                    },
                    {
                        "title": "Multi-class Probabilistic Bounds for Self-learning",
                        "abstract": "Self-learning is a classical approach for learning with both labeled and unlabeled observations which consists in giving pseudo-labels to unlabeled training instances with a confidence score over a predetermined threshold. At the same time, the pseudo-labeling technique is prone to error and runs the risk of adding noisy labels into unlabeled training data. In this paper, we present a probabilistic framework for analyzing self-learning in the multi-class classification scenario with partially labeled data. First, we derive a transductive bound over the risk of the multi-class majority vote classifier. Based on this result, we propose to automatically choose the threshold for pseudo-labeling that minimizes the transductive bound. Then, we introduce a mislabeling error model to analyze the error of the majority vote classifier in the case of the pseudo-labeled data. We derive a probabilistic C-bound over the majority vote error when an imperfect label is given. Empirical results on different data sets show the effectiveness of our framework compared to several state-of-the-art semi-supervised approaches."
                    },
                    {
                        "title": "Semi-supervised Wrapper Feature Selection by Modeling Imperfect Labels",
                        "abstract": "In this paper, we propose a new wrapper feature selection approach with partially labeled training examples where unlabeled observations are pseudo-labeled using the predictions of an initial classifier trained on the labeled training set. The wrapper is composed of a genetic algorithm for proposing new feature subsets, and an evaluation measure for scoring the different feature subsets. The selection of feature subsets is done by assigning weights to characteristics and recursively eliminating those that are irrelevant. The selection criterion is based on a new multi-class $\\mathcal{C}$-bound that explicitly takes into account the mislabeling errors induced by the pseudo-labeling mechanism, using a probabilistic error model. Empirical results on different data sets show the effectiveness of our framework compared to several state-of-the-art semi-supervised feature selection approaches."
                    },
                    {
                        "title": "User-friendly Foundation Model Adapters for Multivariate Time Series Classification",
                        "abstract": "Foundation models, while highly effective, are often resource-intensive, requiring substantial inference time and memory. This paper addresses the challenge of making these models more accessible with limited computational resources by exploring dimensionality reduction techniques. Our goal is to enable users to run large pre-trained foundation models on standard GPUs without sacrificing performance. We investigate classical methods such as Principal Component Analysis alongside neural network-based adapters, aiming to reduce the dimensionality of multivariate time series data while preserving key features. Our experiments show up to a 10x speedup compared to the baseline model, without performance degradation, and enable up to 4.5x more datasets to fit on a single GPU, paving the way for more user-friendly and scalable foundation models."
                    },
                    {
                        "title": "Leveraging Gradients for Unsupervised Accuracy Estimation under Distribution Shift",
                        "abstract": "Estimating test accuracy without access to the ground-truth test labels under varying test environments is a challenging, yet extremely important problem in the safe deployment of machine learning algorithms. Existing works rely on the information from either the outputs or the extracted features of neural networks to formulate an estimation score correlating with the ground-truth test accuracy. In this paper, we investigate--both empirically and theoretically--how the information provided by the gradients can be predictive of the ground-truth test accuracy even under a distribution shift. Specifically, we use the norm of classification-layer gradients, backpropagated from the cross-entropy loss after only one gradient step over test data. Our key idea is that the model should be adjusted with a higher magnitude of gradients when it does not generalize to the test dataset with a distribution shift. We provide theoretical insights highlighting the main ingredients of such an approach ensuring its empirical success. Extensive experiments conducted on diverse distribution shifts and model structures demonstrate that our method significantly outperforms state-of-the-art algorithms."
                    },
                    {
                        "title": "Analysing Multi-Task Regression via Random Matrix Theory with Application to Time Series Forecasting",
                        "abstract": "In this paper, we introduce a novel theoretical framework for multi-task regression, applying random matrix theory to provide precise performance estimations, under high-dimensional, non-Gaussian data distributions. We formulate a multi-task optimization problem as a regularization technique to enable single-task models to leverage multi-task learning information. We derive a closed-form solution for multi-task optimization in the context of linear models. Our analysis provides valuable insights by linking the multi-task learning performance to various model statistics such as raw data covariances, signal-generating hyperplanes, noise levels, as well as the size and number of datasets. We finally propose a consistent estimation of training and testing errors, thereby offering a robust foundation for hyperparameter optimization in multi-task regression scenarios. Experimental validations on both synthetic and real-world datasets in regression and multivariate time series forecasting demonstrate improvements on univariate models, incorporating our method into the training loss and thus leveraging multivariate information."
                    },
                    {
                        "title": "Self-Training: A Survey",
                        "abstract": "Semi-supervised algorithms aim to learn prediction functions from a small set of labeled observations and a large set of unlabeled observations. Because this framework is relevant in many applications, they have received a lot of interest in both academia and industry. Among the existing techniques, self-training methods have undoubtedly attracted greater attention in recent years. These models are designed to find the decision boundary on low density regions without making additional assumptions about the data distribution, and use the unsigned output score of a learned classifier, or its margin, as an indicator of confidence. The working principle of self-training algorithms is to learn a classifier iteratively by assigning pseudo-labels to the set of unlabeled training samples with a margin greater than a certain threshold. The pseudo-labeled examples are then used to enrich the labeled training data and to train a new classifier in conjunction with the labeled training set. In this paper, we present self-training methods for binary and multi-class classification; as well as their variants and two related approaches, namely consistency-based approaches and transductive learning. We examine the impact of significant self-training features on various methods, using different general and image classification benchmarks, and we discuss our ideas for future research in self-training. To the best of our knowledge, this is the first thorough and complete survey on this subject."
                    },
                    {
                        "title": "SAMformer: Unlocking the Potential of Transformers in Time Series Forecasting with Sharpness-Aware Minimization and Channel-Wise Attention",
                        "abstract": "Transformer-based architectures achieved breakthrough performance in natural language processing and computer vision, yet they remain inferior to simpler linear baselines in multivariate long-term forecasting. To better understand this phenomenon, we start by studying a toy linear forecasting problem for which we show that transformers are incapable of converging to their true solution despite their high expressive power. We further identify the attention of transformers as being responsible for this low generalization capacity. Building upon this insight, we propose a shallow lightweight transformer model that successfully escapes bad local minima when optimized with sharpness-aware optimization. We empirically demonstrate that this result extends to all commonly used real-world multivariate time series datasets. In particular, SAMformer surpasses current state-of-the-art methods and is on par with the biggest foundation model MOIRAI while having significantly fewer parameters. The code is available at https://github.com/romilbert/samformer."
                    }
                ]
            },
            "1a2052f5-b205-4152-ac75-b2d88c4ec226": {
                "pk": "1a2052f5-b205-4152-ac75-b2d88c4ec226",
                "name": "Weijian Deng",
                "collaborators": [
                    "Liang Zheng",
                    "Stephen Gould",
                    "Weijie Tu",
                    "Tom Gedeon",
                    "Jianbin Jiao",
                    "Joshua Marsh",
                    "Yumin Suh",
                    "Yifan Sun",
                    "Shengjin Wang",
                    "Yuli Zou"
                ],
                "domain": [
                    "Computer Vision",
                    "Domain Adaptation",
                    "Model Evaluation",
                    "Uncertainty Estimation"
                ],
                "publications": [
                    {
                        "title": "Are Labels Always Necessary for Classifier Accuracy Evaluation?",
                        "abstract": "To calculate the model accuracy on a computer vision task, e.g., object recognition, we usually require a test set composing of test samples and their ground truth labels. Whilst standard usage cases satisfy this requirement, many real-world scenarios involve unlabeled test data, rendering common model evaluation methods infeasible. We investigate this important and under-explored problem, Automatic model Evaluation (AutoEval). Specifically, given a labeled training set and a classifier, we aim to estimate the classification accuracy on unlabeled test datasets. We construct a meta-dataset: a dataset comprised of datasets generated from the original images via various transformations such as rotation, background substitution, foreground scaling, etc. As the classification accuracy of the model on each sample (dataset) is known from the original dataset labels, our task can be solved via regression. Using the feature statistics to represent the distribution of a sample dataset, we can train regression models (e.g., a regression neural network) to predict model performance. Using synthetic meta-dataset and real-world datasets in training and testing, respectively, we report a reasonable and promising prediction of the model accuracy. We also provide insights into the application scope, limitation, and potential future direction of AutoEval."
                    },
                    {
                        "title": "Fine-grained Classification via Categorical Memory Networks",
                        "abstract": "Motivated by the desire to exploit patterns shared across classes, we present a simple yet effective class-specific memory module for fine-grained feature learning. The memory module stores the prototypical feature representation for each category as a moving average. We hypothesize that the combination of similarities with respect to each category is itself a useful discriminative cue. To detect these similarities, we use attention as a querying mechanism. The attention scores with respect to each class prototype are used as weights to combine prototypes via weighted sum, producing a uniquely tailored response feature representation for a given input. The original and response features are combined to produce an augmented feature for classification. We integrate our class-specific memory module into a standard convolutional neural network, yielding a Categorical Memory Network. Our memory module significantly improves accuracy over baseline CNNs, achieving competitive accuracy with state-of-the-art methods on four benchmarks, including CUB-200-2011, Stanford Cars, FGVC Aircraft, and NABirds."
                    },
                    {
                        "title": "What Does Rotation Prediction Tell Us about Classifier Accuracy under Varying Testing Environments?",
                        "abstract": "Understanding classifier decision under novel environments is central to the community, and a common practice is evaluating it on labeled test sets. However, in real-world testing, image annotations are difficult and expensive to obtain, especially when the test environment is changing. A natural question then arises: given a trained classifier, can we evaluate its accuracy on varying unlabeled test sets? In this work, we train semantic classification and rotation prediction in a multi-task way. On a series of datasets, we report an interesting finding, i.e., the semantic classification accuracy exhibits a strong linear relationship with the accuracy of the rotation prediction task (Pearson's Correlation r > 0.88). This finding allows us to utilize linear regression to estimate classifier performance from the accuracy of rotation prediction which can be obtained on the test set through the freely generated rotation labels."
                    },
                    {
                        "title": "On the Strong Correlation Between Model Invariance and Generalization",
                        "abstract": "Generalization and invariance are two essential properties of any machine learning model. Generalization captures a model's ability to classify unseen data while invariance measures consistency of model predictions on transformations of the data. Existing research suggests a positive relationship: a model generalizing well should be invariant to certain visual factors. Building on this qualitative implication we make two contributions. First, we introduce effective invariance (EI), a simple and reasonable measure of model invariance which does not rely on image labels. Given predictions on a test image and its transformed version, EI measures how well the predictions agree and with what level of confidence. Second, using invariance scores computed by EI, we perform large-scale quantitative correlation studies between generalization and invariance, focusing on rotation and grayscale transformations. From a model-centric view, we observe generalization and invariance of different models exhibit a strong linear relationship, on both in-distribution and out-of-distribution datasets. From a dataset-centric view, we find a certain model's accuracy and invariance linearly correlated on different test sets. Apart from these major findings, other minor but interesting insights are also discussed."
                    },
                    {
                        "title": "Confidence and Dispersity Speak: Characterising Prediction Matrix for Unsupervised Accuracy Estimation",
                        "abstract": "This work aims to assess how well a model performs under distribution shifts without using labels. While recent methods study prediction confidence, this work reports prediction dispersity is another informative cue. Confidence reflects whether the individual prediction is certain; dispersity indicates how the overall predictions are distributed across all categories. Our key insight is that a well-performing model should give predictions with high confidence and high dispersity. That is, we need to consider both properties so as to make more accurate estimates. To this end, we use the nuclear norm that has been shown to be effective in characterizing both properties. Extensive experiments validate the effectiveness of nuclear norm for various models (e.g., ViT and ConvNeXt), different datasets (e.g., ImageNet and CUB-200), and diverse types of distribution shifts (e.g., style shift and reproduction shift). We show that the nuclear norm is more accurate and robust in accuracy estimation than existing methods. Furthermore, we validate the feasibility of other measurements (e.g., mutual information maximization) for characterizing dispersity and confidence. Lastly, we investigate the limitation of the nuclear norm, study its improved variant under severe class imbalance, and discuss potential directions."
                    },
                    {
                        "title": "What Does Softmax Probability Tell Us about Classifiers Ranking Across Diverse Test Conditions?",
                        "abstract": "This work aims to develop a measure that can accurately rank the performance of various classifiers when they are tested on unlabeled data from out-of-distribution (OOD) distributions. We commence by demonstrating that conventional uncertainty metrics, notably the maximum Softmax prediction probability, possess inherent utility in forecasting model generalization across certain OOD contexts. Building on this insight, we introduce a new measure called Softmax Correlation (SoftmaxCorr). It calculates the cosine similarity between a class-class correlation matrix, constructed from Softmax output vectors across an unlabeled test dataset, and a predefined reference matrix that embodies ideal class correlations. A high resemblance of predictions to the reference matrix signals that the model delivers confident and uniform predictions across all categories, reflecting minimal uncertainty and confusion. Through rigorous evaluation across a suite of datasets, including ImageNet, CIFAR-10, and WILDS, we affirm the predictive validity of SoftmaxCorr in accurately forecasting model performance within both in-distribution (ID) and OOD settings. Furthermore, we discuss the limitations of our proposed measure and suggest avenues for future research."
                    },
                    {
                        "title": "Domain Alignment with Triplets",
                        "abstract": "Deep domain adaptation methods can reduce the distribution discrepancy by learning domain-invariant embedddings. However, these methods only focus on aligning the whole data distributions, without considering the class-level relations among source and target images. Thus, a target embeddings of a bird might be aligned to source embeddings of an airplane. This semantic misalignment can directly degrade the classifier performance on the target dataset. To alleviate this problem, we present a similarity constrained alignment (SCA) method for unsupervised domain adaptation. When aligning the distributions in the embedding space, SCA enforces a similarity-preserving constraint to maintain class-level relations among the source and target images, i.e., if a source image and a target image are of the same class label, their corresponding embeddings are supposed to be aligned nearby, and vise versa. In the absence of target labels, we assign pseudo labels for target images. Given labeled source images and pseudo-labeled target images, the similarity-preserving constraint can be implemented by minimizing the triplet loss. With the joint supervision of domain alignment loss and similarity-preserving constraint, we train a network to obtain domain-invariant embeddings with two critical characteristics, intra-class compactness and inter-class separability. Extensive experiments conducted on the two datasets well demonstrate the effectiveness of SCA."
                    },
                    {
                        "title": "A Closer Look at the Robustness of Contrastive Language-Image Pre-Training (CLIP)",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) models have demonstrated remarkable generalization capabilities across multiple challenging distribution shifts. However, there is still much to be explored in terms of their robustness to the variations of specific visual factors. In real-world applications, reliable and safe systems must consider other safety objectives beyond classification accuracy, such as predictive uncertainty. Yet, the effectiveness of CLIP models on such safety-related features is less-explored. Driven by the above, this work comprehensively investigates the safety objectives of CLIP models, specifically focusing on three key properties: resilience to visual factor variations, calibrated uncertainty estimations, and the ability to detect anomalous inputs. To this end, we study 83 CLIP models and 127 ImageNet classifiers. They are diverse in architecture, (pre)training distribution and training strategies. We consider 10 visual factors (e.g., shape and pattern), 5 types of out-of-distribution data, and 8 natural and challenging test conditions with different shift types, such as texture, style, and perturbation shifts. Our study has unveiled several previously unknown insights into CLIP models. For instance, they are not consistently more calibrated than other ImageNet models, which contradicts existing findings. Additionally, our analysis underscores the significance of training source design by showcasing its profound influence on the three safety-related properties. We believe our comprehensive study can shed light on and help guide the development of more robust and reliable CLIP models."
                    },
                    {
                        "title": "Toward a Holistic Evaluation of Robustness in CLIP Models",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) models have shown significant potential, particularly in zero-shot classification across diverse distribution shifts. Building on existing evaluations of overall classification robustness, this work aims to provide a more comprehensive assessment of CLIP by introducing several new perspectives. First, we investigate their robustness to variations in specific visual factors. Second, we assess two critical safety objectives--confidence uncertainty and out-of-distribution detection--beyond mere classification accuracy. Third, we evaluate the finesse with which CLIP models bridge the image and text modalities. Fourth, we extend our examination to 3D awareness in CLIP models, moving beyond traditional 2D image understanding. Finally, we explore the interaction between vision and language encoders within modern large multimodal models (LMMs) that utilize CLIP as the visual backbone, focusing on how this interaction impacts classification robustness. In each aspect, we consider the impact of six factors on CLIP models: model architecture, training distribution, training set size, fine-tuning, contrastive loss, and test-time prompts. Our study uncovers several previously unknown insights into CLIP. For instance, the architecture of the visual encoder in CLIP plays a significant role in their robustness against 3D corruption. CLIP models tend to exhibit a bias towards shape when making predictions. Moreover, this bias tends to diminish after fine-tuning on ImageNet. Vision-language models like LLaVA, leveraging the CLIP vision encoder, could exhibit benefits in classification performance for challenging categories over CLIP alone. Our findings are poised to offer valuable guidance for enhancing the robustness and reliability of CLIP models."
                    },
                    {
                        "title": "SVDNet for Pedestrian Retrieval",
                        "abstract": "This paper proposes the SVDNet for retrieval problems, with focus on the application of person re-identification (re-ID). We view each weight vector within a fully connected (FC) layer in a convolutional neuron network (CNN) as a projection basis. It is observed that the weight vectors are usually highly correlated. This problem leads to correlations among entries of the FC descriptor, and compromises the retrieval performance based on the Euclidean distance. To address the problem, this paper proposes to optimize the deep representation learning process with Singular Vector Decomposition (SVD). Specifically, with the restraint and relaxation iteration (RRI) training scheme, we are able to iteratively integrate the orthogonality constraint in CNN training, yielding the so-called SVDNet. We conduct experiments on the Market-1501, CUHK03, and Duke datasets, and show that RRI effectively reduces the correlation among the projection vectors, produces more discriminative FC descriptors, and significantly improves the re-ID accuracy. On the Market-1501 dataset, for instance, rank-1 accuracy is improved from 55.3% to 80.5% for CaffeNet, and from 73.8% to 82.3% for ResNet-50."
                    },
                    {
                        "title": "Adaptive Calibrator Ensemble for Model Calibration under Distribution Shift",
                        "abstract": "Model calibration usually requires optimizing some parameters (e.g., temperature) w.r.t an objective function (e.g., negative log-likelihood). In this paper, we report a plain, important but often neglected fact that the objective function is influenced by calibration set difficulty, i.e., the ratio of the number of incorrectly classified samples to that of correctly classified samples. If a test set has a drastically different difficulty level from the calibration set, the optimal calibration parameters of the two datasets would be different. In other words, a calibrator optimal on the calibration set would be suboptimal on the OOD test set and thus has degraded performance. With this knowledge, we propose a simple and effective method named adaptive calibrator ensemble (ACE) to calibrate OOD datasets whose difficulty is usually higher than the calibration set. Specifically, two calibration functions are trained, one for in-distribution data (low difficulty), and the other for severely OOD data (high difficulty). To achieve desirable calibration on a new OOD dataset, ACE uses an adaptive weighting method that strikes a balance between the two extreme functions. When plugged in, ACE generally improves the performance of a few state-of-the-art calibration schemes on a series of OOD benchmarks. Importantly, such improvement does not come at the cost of the in-distribution calibration accuracy."
                    },
                    {
                        "title": "Similarity-preserving Image-image Domain Adaptation for Person Re-identification",
                        "abstract": "This article studies the domain adaptation problem in person re-identification (re-ID) under a \"learning via translation\" framework, consisting of two components, 1) translating the labeled images from the source to the target domain in an unsupervised manner, 2) learning a re-ID model using the translated images. The objective is to preserve the underlying human identity information after image translation, so that translated images with labels are effective for feature learning on the target domain. To this end, we propose a similarity preserving generative adversarial network (SPGAN) and its end-to-end trainable version, eSPGAN. Both aiming at similarity preserving, SPGAN enforces this property by heuristic constraints, while eSPGAN does so by optimally facilitating the re-ID model learning. More specifically, SPGAN separately undertakes the two components in the \"learning via translation\" framework. It first preserves two types of unsupervised similarity, namely, self-similarity of an image before and after translation, and domain-dissimilarity of a translated source image and a target image. It then learns a re-ID model using existing networks. In comparison, eSPGAN seamlessly integrates image translation and re-ID model learning. During the end-to-end training of eSPGAN, re-ID learning guides image translation to preserve the underlying identity information of an image. Meanwhile, image translation improves re-ID learning by providing identity-preserving training samples of the target domain style. In the experiment, we show that identities of the fake images generated by SPGAN and eSPGAN are well preserved. Based on this, we report the new state-of-the-art domain adaptation results on two large-scale person re-ID datasets."
                    },
                    {
                        "title": "An Empirical Study Into What Matters for Calibrating Vision-Language Models",
                        "abstract": "Vision-Language Models (VLMs) have emerged as the dominant approach for zero-shot recognition, adept at handling diverse scenarios and significant distribution changes. However, their deployment in risk-sensitive areas requires a deeper understanding of their uncertainty estimation capabilities, a relatively uncharted area. In this study, we explore the calibration properties of VLMs across different architectures, datasets, and training strategies. In particular, we analyze the uncertainty estimation performance of VLMs when calibrated in one domain, label set or hierarchy level, and tested in a different one. Our findings reveal that while VLMs are not inherently calibrated for uncertainty, temperature scaling significantly and consistently improves calibration, even across shifts in distribution and changes in label set. Moreover, VLMs can be calibrated with a very small set of examples. Through detailed experimentation, we highlight the potential applications and importance of our insights, aiming for more reliable and effective use of VLMs in critical, real-world scenarios."
                    },
                    {
                        "title": "Ranking Models in Unlabeled New Environments",
                        "abstract": "Consider a scenario where we are supplied with a number of ready-to-use models trained on a certain source domain and hope to directly apply the most appropriate ones to different target domains based on the models' relative performance. Ideally we should annotate a validation set for model performance assessment on each new target environment, but such annotations are often very expensive. Under this circumstance, we introduce the problem of ranking models in unlabeled new environments. For this problem, we propose to adopt a proxy dataset that 1) is fully labeled and 2) well reflects the true model rankings in a given target environment, and use the performance rankings on the proxy sets as surrogates. We first select labeled datasets as the proxy. Specifically, datasets that are more similar to the unlabeled target domain are found to better preserve the relative performance rankings. Motivated by this, we further propose to search the proxy set by sampling images from various datasets that have similar distributions as the target. We analyze the problem and its solutions on the person re-identification (re-ID) task, for which sufficient datasets are publicly available, and show that a carefully constructed proxy set effectively captures relative performance ranking in new environments. Code is available at \\url{https://github.com/sxzrt/Proxy-Set}."
                    },
                    {
                        "title": "A Bag-of-Prototypes Representation for Dataset-Level Applications",
                        "abstract": "This work investigates dataset vectorization for two dataset-level tasks: assessing training set suitability and test set difficulty. The former measures how suitable a training set is for a target domain, while the latter studies how challenging a test set is for a learned model. Central to the two tasks is measuring the underlying relationship between datasets. This needs a desirable dataset vectorization scheme, which should preserve as much discriminative dataset information as possible so that the distance between the resulting dataset vectors can reflect dataset-to-dataset similarity. To this end, we propose a bag-of-prototypes (BoP) dataset representation that extends the image-level bag consisting of patch descriptors to dataset-level bag consisting of semantic prototypes. Specifically, we develop a codebook consisting of K prototypes clustered from a reference dataset. Given a dataset to be encoded, we quantize each of its image features to a certain prototype in the codebook and obtain a K-dimensional histogram. Without assuming access to dataset labels, the BoP representation provides a rich characterization of the dataset semantic distribution. Furthermore, BoP representations cooperate well with Jensen-Shannon divergence for measuring dataset-to-dataset similarity. Although very simple, BoP consistently shows its advantage over existing representations on a series of benchmarks for two dataset-level tasks."
                    }
                ]
            },
            "6c21278c-50fb-484d-bbb1-2094eaaf77fe": {
                "pk": "6c21278c-50fb-484d-bbb1-2094eaaf77fe",
                "name": "Jianfeng Zhang",
                "collaborators": [
                    "Triet Pham",
                    "Said Hamadene",
                    "Jia Zhuo",
                    "Christian Keller",
                    "Haiyang Wang",
                    "Cong Wu",
                    "Jiongmin Yong",
                    "Joscha Diehl",
                    "Frederi Viens",
                    "Ibrahim Ekren"
                ],
                "domain": [
                    "Stochastic Control",
                    "Backward Stochastic Differential Equations",
                    "Game Theory",
                    "Path-Dependent PDEs"
                ],
                "publications": [
                    {
                        "title": "Is a sophisticated agent always a wise one?",
                        "abstract": "For time inconsistent optimal control problems, a quite popular approach is the equilibrium approach, taken by the sophisticated agents. In this short note we construct a deterministic continuous time example where the unique equilibrium is dominated by another control. Therefore, in this situation it may not be wise to take the equilibrium strategy."
                    },
                    {
                        "title": "The Wellposedness of FBSDEs (II)",
                        "abstract": "This paper is a continuation of \\cite{zhang}, in which we established the wellposedness result and a comparison theorem for a class of one dimensional Forward-Backward SDEs. In this paper we extend the wellposedness result to high dimensional FBSDEs, and weaken the key condition in \\cite{zhang} significantly. Compared to the existing methods in the literature, our result has the following features: (i) arbitrary time duration; (ii) random coefficients; (iii) (possibly) degenerate forward diffusion; and (iv) no monotonicity condition."
                    },
                    {
                        "title": "Instability and Efficiency of Non-cooperative Games",
                        "abstract": "It is well known that a non-cooperative game may have multiple equilibria. In this paper we consider the efficiency of games, measured by the ratio between the aggregate payoff over all Nash equilibria and that over all admissible controls. Such efficiency operator is typically unstable with respect to small perturbation of the game. This seemingly bad property can actually be a good news in practice: it is possible that a small change of the game mechanism may improve the efficiency of the game dramatically. We shall introduce a game mediator with limited resources and investigate the mechanism designs aiming to improve the efficiency."
                    },
                    {
                        "title": "Representation of solutions to BSDEs associated with a degenerate FSDE",
                        "abstract": "In this paper we investigate a class of decoupled forward-backward SDEs, where the volatility of the FSDE is degenerate and the terminal value of the BSDE is a discontinuous function of the FSDE. Such an FBSDE is associated with a degenerate parabolic PDE with discontinuous terminal condition. We first establish a Feynman-Kac type representation formula for the spatial derivative of the solution to the PDE. As a consequence, we show that there exists a stopping time \\tau such that the martingale integrand of the BSDE is continuous before \\tau and vanishes after \\tau. However, it may blow up at \\tau, as illustrated by an example. Moreover, some estimates for the martingale integrand before \\tau are obtained. These results are potentially useful for pricing and hedging discontinuous exotic options (e.g., digital options) when the underlying asset's volatility is small, and they are also useful for studying the rate of convergence of finite-difference approximations for degenerate parabolic PDEs."
                    },
                    {
                        "title": "Two Person Zero-sum Game in Weak Formulation and Path Dependent Bellman-Isaacs Equation",
                        "abstract": "In this paper we study a two person zero sum stochastic differential game in weak formulation. Unlike standard literature which uses strategy type of controls, the weak formulation allows us to consider the game with control against control. We shall prove the existence of game value under natural conditions. Another main feature of the paper is that we allow for non-Markovian structure, and thus the game value is a random process. We characterize the value process as the unique viscosity solution of the corresponding path dependent Bellman-Isaacs equation, a notion recently introduced by Ekren, Keller, Touzi and Zhang (arXiv:1109.5971) and Ekren, Touzi and Zhang."
                    },
                    {
                        "title": "Monotone Schemes for Fully Nonlinear Parabolic Path Dependent PDEs",
                        "abstract": "In this paper we extend the results of the seminal work Barles and Souganidis \\cite{BS} to path dependent case. Based on the viscosity theory of path dependent PDEs, developed by Ekren, Keller, Touzi and Zhang \\cite{EKTZ} and Ekren, Touzi and Zhang \\cite{ETZ0, ETZ1, ETZ2}, we show that a monotone scheme converges to the unique viscosity solution of the (fully nonlinear) parabolic path dependent PDE. An example of such monotone scheme is proposed. Moreover, in the case that the solution is smooth enough, we obtain the rate of convergence of our scheme."
                    },
                    {
                        "title": "The Starting and Stopping Problem under Knightian Uncertainty and Related Systems of Reflected BSDEs",
                        "abstract": "This article deals with the starting and stopping problem under Knightian uncertainty, i.e., roughly speaking, when the probability under which the future evolves is not exactly known. We show that the lower price of a plant submitted to the decisions of starting and stopping is given by a solution of a system of two reflected backward stochastic differential equations (BSDEs for short). We solve this latter system and we give the expression of the optimal strategy. Further we consider a more general system of $m$ ($m\\geq 2$) reflected BSDEs with interconnected obstacles. Once more we show existence and uniqueness of the solution of that system."
                    },
                    {
                        "title": "Pathwise It\u00f4 Calculus for Rough Paths and Rough PDEs with Path Dependent Coefficients",
                        "abstract": "This paper introduces the path derivatives, in the spirit of Dupire's functional It\\^o calculus, for the controlled paths in the rough path theory with possibly non-geometric rough paths. The theory allows us to deal with rough integration and rough PDEs in the same manner as standard stochastic calculus. We next study rough PDEs with coefficients depending on the rough path itself, which corresponds to stochastic PDEs with random coefficients. Such coefficients is less regular in the time variable and is not covered in the existing literature. The results are useful for studying viscosity solutions of stochastic PDEs."
                    },
                    {
                        "title": "Forward Backward SDEs in Weak Formulation",
                        "abstract": "Although having been developed for more than two decades, the theory of forward backward stochastic differential equations is still far from complete. In this paper, we take one step back and investigate the formulation of FBSDEs. Motivated from several considerations, both in theory and in applications, we propose to study FBSDEs in weak formulation, rather than the strong formulation in the standard literature. That is, the backward SDE is driven by the forward component, instead of by the Brownian motion noise. We establish the Feyman-Kac formula for FBSDEs in weak formulation, both in classical and in viscosity sense. Our new framework is efficient especially when the diffusion part of the forward equation involves the $Z$-component of the backward equation."
                    },
                    {
                        "title": "Some Norm Estimates for Semimartingales",
                        "abstract": "In this paper we introduce a new type of norms for semimartingales, under both linear and nonlinear expectations. Our norm is defined in the spirit of quasimartingales, and it characterizes square integrable semimartingales. This work is motivated by our study of zero-sum stochastic differential games \\cite{PZ}, whose value process is conjectured to be a semimartingale under a class of probability measures. As a by product, we establish some a priori estimates for doubly reflected BSDEs without imposing the Mokobodski's condition directly."
                    },
                    {
                        "title": "An Elementary Proof for the Structure of Wasserstein Derivatives",
                        "abstract": "Let $F: \\mathbb{L}^2(\\Omega, \\mathbb{R}) \\to \\mathbb{R}$ be a law invariant and continuously Fr\\'echet differentiable mapping. Based on Lions \\cite{Lions}, Cardaliaguet \\cite{Cardaliaguet} (Theorem 6.2 and 6.5) proved that: \\bea \\label{Derivative} D F (\\xi) = g(\\xi), \\eea where $g: \\mathbb{R} \\to \\mathbb{R}$ is a deterministic function which depends only on the law of $\\xi$. See also Carmona \\& Delarue \\cite{CD} Section 5.2. In this short note we provide an elementary proof for this well known result. This note is part of our accompanying paper \\cite{WZ}, which deals with a more general situation."
                    },
                    {
                        "title": "Non-Equivalence of Stochastic Optimal Control Problems with Open and Closed Loop Controls",
                        "abstract": "For an optimal control problem of an It\\^o's type stochastic differential equation, the control process could be taken as open-loop or closed-loop forms. In the standard literature, provided appropriate regularity, the value functions under these two types of controls are equal and are the unique (viscosity) solution to the corresponding (path-dependent) HJB equation. In this short note, we provide a counterexample in the path dependent setting showing that these value functions can be different in general."
                    },
                    {
                        "title": "Backward stochastic differential equations with Young drift",
                        "abstract": "We prove via a direct fixpoint argument the well-posedness of backward stochastic differential equations containing an additional drift driven by a path of finite $p$-variation with $p \\in [1,2)$. An application to the Feynman-Kac representation of semilinear rough partial differential equations is given."
                    },
                    {
                        "title": "The Continuous Time Nonzero-sum Dynkin Game Problem and Application in Game Options",
                        "abstract": "In this paper we study the nonzero-sum Dynkin game in continuous time which is a two player non-cooperative game on stopping times. We show that it has a Nash equilibrium point for general stochastic processes. As an application, we consider the problem of pricing American game contingent claims by the utility maximization approach."
                    },
                    {
                        "title": "A Martingale Approach for Fractional Brownian Motions and Related Path Dependent PDEs",
                        "abstract": "In this paper we study dynamic backward problems, with the computation of conditional expectations as a main objective, in a framework where the (forward) state process satisfies a Volterra type SDE, with fractional Brownian motion as a typical example. Such processes are neither Markov processes nor semimartingales, and most notably, they feature a certain time inconsistency which makes any direct application of Markovian ideas, such as flow properties, impossible without passing to a path-dependent framework. Our main result is a functional It\\^{o} formula, extending the seminal work of Dupire \\cite{Dupire} to our more general framework. In particular, unlike in \\cite{Dupire} where one needs only to consider the stopped paths, here we need to concatenate the observed path up to the current time with a certain smooth observable curve derived from the distribution of the future paths. This new feature is due to the time inconsistency involved in this paper. We then derive the path dependent PDEs for the backward problems. Finally, an application to option pricing in a financial market with rough volatility is presented."
                    },
                    {
                        "title": "Pseudo Markovian Viscosity Solutions of Fully Nonlinear Degenerate PPDEs",
                        "abstract": "In this paper we propose a new type of viscosity solutions for fully nonlinear path dependent PDEs. By restricting to certain pseudo Markovian structure, we remove the uniform non- degeneracy condition imposed in our earlier works [9, 10]. We establish the comparison principle under natural and mild conditions. Moreover, as applications we apply our results to two important classes of PPDEs: the stochastic HJB equations and the path dependent Isaacs equations, induced from the stochastic optimization with random coefficients and the path dependent zero sum game problem, respectively."
                    },
                    {
                        "title": "Set Valued Hamilton-Jacobi-Bellman Equations",
                        "abstract": "Building upon the dynamic programming principle for set valued functions arising from many applications, in this paper we propose a new notion of set valued PDEs. The key component in the theory is a set valued It\\^{o} formula, characterizing the flows on the surface of the dynamic sets. In the contexts of multivariate control problems, we establish the wellposedness of the set valued HJB equations, which extends the standard HJB equations in the scalar case to the multivariate case. As an application, a moving scalarization for certain time inconsistent problems is constructed by using the classical solution of the set valued HJB equation."
                    }
                ]
            },
            "b99494d9-8021-4487-ac02-c70babda31a0": {
                "pk": "b99494d9-8021-4487-ac02-c70babda31a0",
                "name": "Bo An",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2408.04057": {
        "paper_data": {
            "title": "PowerPM: Foundation Model for Power Systems",
            "url": "http://arxiv.org/abs/2408.04057v3",
            "arxiv_id": "2408.04057",
            "authors": [
                "Shihao Tu",
                "Yupeng Zhang",
                "Jing Zhang",
                "Zhendong Fu",
                "Yin Zhang",
                "Yang Yang"
            ],
            "abstract": "The emergence of abundant electricity time series (ETS) data provides ample opportunities for various applications in the power systems, including demand-side management, grid stability, and consumer behavior analysis. Deep learning models have advanced ETS modeling by effectively capturing sequence dependence. Nevertheless, learning a generic representation of ETS data for various applications remains challenging due to the inherently complex hierarchical structure of ETS data. Moreover, ETS data exhibits intricate temporal dependencies and is suscepti ble to the influence of exogenous variables. Furthermore, different instances exhibit diverse electricity consumption behavior. In this paper, we propose a foundation model PowerPM to model ETS data, providing a large-scale, off-the-shelf model for power systems. PowerPM consists of a temporal encoder and a hierarchical encoder. The temporal encoder captures both temporal dependencies in ETS data, considering exogenous variables. The hierarchical encoder models the correlation between hierarchy. Furthermore, PowerPM leverages a novel self-supervised pretraining framework consisting of masked ETS modeling and dual-view contrastive learning, which enable PowerPM to capture temporal dependency within ETS windows and aware the discrepancy across ETS windows, providing two different perspectives to learn generic representation. Our experiments involve five real world scenario datasets, comprising private and public data. Through pre-training on massive ETS data, PowerPM achieves SOTA performance on diverse downstream tasks within the private dataset. Impressively, when transferred to the public datasets, PowerPM maintains its superiority, showcasing its remarkable generalization ability across various tasks and domains. Moreover, ablation studies, few-shot experiments provide additional evidence of the effectiveness of our model.",
            "introduction": "   1 Introduction  The volume of Electricity Time Series (ETS) data has recently increased rapidly due to the emergence of advanced power systems known as smart grids\u00a0[10]. This abundance of data has paved the way for diverse applications in power systems, including demand-side management\u00a0[22], grid stability\u00a0[2] and consumer behavior analysis\u00a0[49], etc. Meanwhile, these applications have spawned various tasks, as shown in Fig.\u00a01(d), such as load forecasting\u00a0[27, 4], clock anomaly detection\u00a0[46], as well as electricity theft\u00a0[15] and elderly living alone detection\u00a0[45].   Recent statistics show that the total electricity consumption in China reached 9.229.229.229.22 trillion kilowatt-hours in 2023202320232023111https://www.nea.gov.cn/2024-01/26/c_1310762246.htm, which ranks among the top in the world. The substantial economic benefits that accompany this significant electricity usage are also considerable. On the other hand, unreasonable electricity planning can have a detrimental impact on the environment[30]. Thus, given the large amount of data and diverse tasks, there is a pressing need to explore effective modeling methods of ETS data for these tasks, which can lead to enhanced economic efficiency while adhering to low-carbon principles.   Recently, numerous research studies on pre-training approaches for ETS data have emerged. These approaches adopt the \u201cpre-training then fine-tuning\u201d paradigm, which solves the dilemma of limited annotation data, and the pre-trained model can easily adapt to new tasks. Such as PatchTST\u00a0[21], TS2Vec\u00a0[42], CoST\u00a0[37], etc. However, these pre-training methods only utilize small-scale of data with a small number of instances (e.g. users), resulting in poor performance on downstream tasks. As the same time, many researcher begin to apply Large Language Models (LLMs) to assist time series modeling by using pre-trained LLM to encode time series\u00a0[51] or incorporating additional descriptions related to the time series\u00a0[17, 20]. Nevertheless, these models have limited ability in the power system scenario due to insufficient pre-training data of power systems and the lack of sufficient domain-specific knowledge. Additionally, none of these models are tailored for the scenario of power systems, neglecting the unique characteristics of ETS data. Therefore, existing power systems related works still maintain a large research gap in modeling ETS data with a foundation model.   Figure 1:  (a) The hierarchical structure of ETS data. (b) The diversity of instances, and the temporal dependency within ETS data. (c) Different electricity consumption behaviors exist across time and instances. (d) Various tasks in power systems.   In our scenario, the ETS data contains numerous instances and naturally exhibits a complex hierarchy\u00a0[41, 23]. As depicted in Fig.\u00a01(a), a city ETS can be disaggregated into district ETS through the administrative divisions, which are further disaggregated into user ETS in this district. For the complex hierarchy of ETS data, modeling ETS data entails the consideration of several challenges:   (1) Hierarchical Dependency Modeling. The hierarchy of ETS data facilitates information interaction across different granularities. Fine-grained ETS provides detailed insights into individual electricity usage, while coarse-grained ETS from districts and cities captures broader factors, indicating overall trends. For example, user-level data reflects user-specific behaviors and city-level data encompasses demographics and policy effects\u00a0[29, 35]. Integrating these levels of granularity to provide both macro and micro perspectives is a complex task that requires sophisticated modeling.   (2) Temporal dependencies within ETS window. An ETS window refer to a piece of electricity time",
            "references": [
                {
                    "title": "UniTime: A Language-Empowered Unified Model for Cross-Domain Time Series Forecasting",
                    "abstract": "Multivariate time series forecasting plays a pivotal role in contemporary web technologies. In contrast to conventional methods that involve creating dedicated models for specific time series application domains, this research advocates for a unified model paradigm that transcends domain boundaries. However, learning an effective cross-domain model presents the following challenges. First, various domains exhibit disparities in data characteristics, e.g., the number of variables, posing hurdles for existing models that impose inflexible constraints on these factors. Second, the model may encounter difficulties in distinguishing data from various domains, leading to suboptimal performance in our assessments. Third, the diverse convergence rates of time series domains can also result in compromised empirical performance. To address these issues, we propose UniTime for effective cross-domain time series learning. Concretely, UniTime can flexibly adapt to data with varying characteristics. It also uses domain instructions and a Language-TS Transformer to offer identification information and align two modalities. In addition, UniTime employs masking to alleviate domain convergence speed imbalance issues. Our extensive experiments demonstrate the effectiveness of UniTime in advancing state-of-the-art forecasting performance and zero-shot transferability."
                },
                {
                    "title": "TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting",
                    "abstract": "The past decade has witnessed significant advances in time series modeling with deep learning. While achieving state-of-the-art results, the best-performing architectures vary highly across applications and domains. Meanwhile, for natural language processing, the Generative Pre-trained Transformer (GPT) has demonstrated impressive performance via training one general-purpose model across various textual datasets. It is intriguing to explore whether GPT-type architectures can be effective for time series, capturing the intrinsic dynamic attributes and leading to significant accuracy improvements. In this paper, we propose a novel framework, TEMPO, that can effectively learn time series representations. We focus on utilizing two essential inductive biases of the time series task for pre-trained models: (i) decomposition of the complex interaction between trend, seasonal and residual components; and (ii) introducing the design of prompts to facilitate distribution adaptation in different types of time series. TEMPO expands the capability for dynamically modeling real-world temporal phenomena from data within diverse domains. Our experiments demonstrate the superior performance of TEMPO over state-of-the-art methods on zero shot setting for a number of time series benchmark datasets. This performance gain is observed not only in scenarios involving previously unseen datasets but also in scenarios with multi-modal inputs. This compelling finding highlights TEMPO's potential to constitute a foundational model-building framework."
                },
                {
                    "title": "Time-LLM: Time Series Forecasting by Reprogramming Large Language Models",
                    "abstract": "Time series forecasting holds significant importance in many real-world dynamic systems and has been extensively studied. Unlike natural language process (NLP) and computer vision (CV), where a single large model can tackle multiple tasks, models for time series forecasting are often specialized, necessitating distinct designs for different tasks and applications. While pre-trained foundation models have made impressive strides in NLP and CV, their development in time series domains has been constrained by data sparsity. Recent studies have revealed that large language models (LLMs) possess robust pattern recognition and reasoning abilities over complex sequences of tokens. However, the challenge remains in effectively aligning the modalities of time series data and natural language to leverage these capabilities. In this work, we present Time-LLM, a reprogramming framework to repurpose LLMs for general time series forecasting with the backbone language models kept intact. We begin by reprogramming the input time series with text prototypes before feeding it into the frozen LLM to align the two modalities. To augment the LLM's ability to reason with time series data, we propose Prompt-as-Prefix (PaP), which enriches the input context and directs the transformation of reprogrammed input patches. The transformed time series patches from the LLM are finally projected to obtain the forecasts. Our comprehensive evaluations demonstrate that Time-LLM is a powerful time series learner that outperforms state-of-the-art, specialized forecasting models. Moreover, Time-LLM excels in both few-shot and zero-shot learning scenarios."
                },
                {
                    "title": "TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Series",
                    "abstract": "This work summarizes two ways to accomplish Time-Series (TS) tasks in today's Large Language Model (LLM) context: LLM-for-TS (model-centric) designs and trains a fundamental large model, or fine-tunes a pre-trained LLM for TS data; TS-for-LLM (data-centric) converts TS into a model-friendly representation to enable the pre-trained LLM to handle TS data. Given the lack of data, limited resources, semantic context requirements, and so on, this work focuses on TS-for-LLM, where we aim to activate LLM's ability for TS data by designing a TS embedding method suitable for LLM. The proposed method is named TEST. It first tokenizes TS, builds an encoder to embed TS via instance-wise, feature-wise, and text-prototype-aligned contrast, where the TS embedding space is aligned to LLM embedding layer space, then creates soft prompts to make LLM more open to that embeddings, and finally implements TS tasks using the frozen LLM. We also demonstrate the feasibility of TS-for-LLM through theory and experiments. Experiments are carried out on TS classification, forecasting, and representation tasks using eight frozen LLMs with various structures and sizes. The results show that the pre-trained LLM with TEST strategy can achieve better or comparable performance than today's SOTA TS models and offer benefits for few-shot and generalization. By treating LLM as the pattern machine, TEST can endow LLM's ability to process TS data without compromising language ability. We hope that this study will serve as a foundation for future work to support TS+LLM progress."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "One Fits All: Power General Time Series Analysis by Pretrained LM",
                    "abstract": "Although we have witnessed great success of pre-trained models in natural language processing (NLP) and computer vision (CV), limited progress has been made for general time series analysis. Unlike NLP and CV where a unified model can be used to perform different tasks, specially designed approach still dominates in each time series analysis task such as classification, anomaly detection, forecasting, and few-shot learning. The main challenge that blocks the development of pre-trained model for time series analysis is the lack of a large amount of data for training. In this work, we address this challenge by leveraging language or CV models, pre-trained from billions of tokens, for time series analysis. Specifically, we refrain from altering the self-attention and feedforward layers of the residual blocks in the pre-trained language or image model. This model, known as the Frozen Pretrained Transformer (FPT), is evaluated through fine-tuning on all major types of tasks involving time series. Our results demonstrate that pre-trained models on natural language or images can lead to a comparable or state-of-the-art performance in all main time series analysis tasks, as illustrated in Figure 1. We also found both theoretically and empirically that the self-attention module behaviors similarly to principle component analysis (PCA), an observation that helps explains how transformer bridges the domain gap and a crucial step towards understanding the universality of a pre-trained transformer.The code is publicly available at https://github.com/DAMO-DI-ML/One_Fits_All."
                },
                {
                    "title": "A Time Series is Worth 64 Words: Long-term Forecasting with Transformers",
                    "abstract": "We propose an efficient design of Transformer-based models for multivariate time series forecasting and self-supervised representation learning. It is based on two key components: (i) segmentation of time series into subseries-level patches which are served as input tokens to Transformer; (ii) channel-independence where each channel contains a single univariate time series that shares the same embedding and Transformer weights across all the series. Patching design naturally has three-fold benefit: local semantic information is retained in the embedding; computation and memory usage of the attention maps are quadratically reduced given the same look-back window; and the model can attend longer history. Our channel-independent patch time series Transformer (PatchTST) can improve the long-term forecasting accuracy significantly when compared with that of SOTA Transformer-based models. We also apply our model to self-supervised pre-training tasks and attain excellent fine-tuning performance, which outperforms supervised training on large datasets. Transferring of masked pre-trained representation on one dataset to others also produces SOTA forecasting accuracy. Code is available at: https://github.com/yuqinie98/PatchTST."
                },
                {
                    "title": "TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis",
                    "abstract": "Time series analysis is of immense importance in extensive applications, such as weather forecasting, anomaly detection, and action recognition. This paper focuses on temporal variation modeling, which is the common key problem of extensive analysis tasks. Previous methods attempt to accomplish this directly from the 1D time series, which is extremely challenging due to the intricate temporal patterns. Based on the observation of multi-periodicity in time series, we ravel out the complex temporal variations into the multiple intraperiod- and interperiod-variations. To tackle the limitations of 1D time series in representation capability, we extend the analysis of temporal variations into the 2D space by transforming the 1D time series into a set of 2D tensors based on multiple periods. This transformation can embed the intraperiod- and interperiod-variations into the columns and rows of the 2D tensors respectively, making the 2D-variations to be easily modeled by 2D kernels. Technically, we propose the TimesNet with TimesBlock as a task-general backbone for time series analysis. TimesBlock can discover the multi-periodicity adaptively and extract the complex temporal variations from transformed 2D tensors by a parameter-efficient inception block. Our proposed TimesNet achieves consistent state-of-the-art in five mainstream time series analysis tasks, including short- and long-term forecasting, imputation, classification, and anomaly detection. Code is available at this repository: https://github.com/thuml/TimesNet."
                },
                {
                    "title": "An Elderly Living-alone Guardianship Model Based on Wavelet Transform",
                    "abstract": "In this paper, an unsupervised guardianship model for the elderly living-alone is constructed through the identification of indoor-outdoor time and the characteristics of power consumption curve. The wavelet transform and sliding window method are used to identify the power consumption period, and timely sense whether the elderly users living-alone need to be alarmed, such as being away from home for a long time, sudden increase of power consumption, abnormal power consumption mode and so on. At the end, taking the actual power consumption data of 79 elderly users living-alone in a community in Tianjin as an example, the effectiveness of the algorithm is verified. Compared with the widely used ways, this method can provide more sufficient judgment basis."
                },
                {
                    "title": "Self-Supervised Contrastive Pre-Training For Time Series via Time-Frequency Consistency",
                    "abstract": "Pre-training on time series poses a unique challenge due to the potential mismatch between pre-training and target domains, such as shifts in temporal dynamics, fast-evolving trends, and long-range and short-cyclic effects, which can lead to poor downstream performance. While domain adaptation methods can mitigate these shifts, most methods need examples directly from the target domain, making them suboptimal for pre-training. To address this challenge, methods need to accommodate target domains with different temporal dynamics and be capable of doing so without seeing any target examples during pre-training. Relative to other modalities, in time series, we expect that time-based and frequency-based representations of the same example are located close together in the time-frequency space. To this end, we posit that time-frequency consistency (TF-C) -- embedding a time-based neighborhood of an example close to its frequency-based neighborhood -- is desirable for pre-training. Motivated by TF-C, we define a decomposable pre-training model, where the self-supervised signal is provided by the distance between time and frequency components, each individually trained by contrastive estimation. We evaluate the new method on eight datasets, including electrodiagnostic testing, human activity recognition, mechanical fault detection, and physical status monitoring. Experiments against eight state-of-the-art methods show that TF-C outperforms baselines by 15.4% (F1 score) on average in one-to-one settings (e.g., fine-tuning an EEG-pretrained model on EMG data) and by 8.4% (precision) in challenging one-to-many settings (e.g., fine-tuning an EEG-pretrained model for either hand-gesture recognition or mechanical fault prediction), reflecting the breadth of scenarios that arise in real-world applications. Code and datasets: https://github.com/mims-harvard/TFC-pretraining."
                },
                {
                    "title": "Are Transformers Effective for Time Series Forecasting?",
                    "abstract": "Recently, there has been a surge of Transformer-based solutions for the long-term time series forecasting (LTSF) task. Despite the growing performance over the past few years, we question the validity of this line of research in this work. Specifically, Transformers is arguably the most successful solution to extract the semantic correlations among the elements in a long sequence. However, in time series modeling, we are to extract the temporal relations in an ordered set of continuous points. While employing positional encoding and using tokens to embed sub-series in Transformers facilitate preserving some ordering information, the nature of the permutation-invariant self-attention mechanism inevitably results in temporal information loss. \nTo validate our claim, we introduce a set of embarrassingly simple one-layer linear models named LTSF-Linear for comparison. Experimental results on nine real-life datasets show that LTSF-Linear surprisingly outperforms existing sophisticated Transformer-based LTSF models in all cases, and often by a large margin. Moreover, we conduct comprehensive empirical studies to explore the impacts of various design elements of LTSF models on their temporal relation extraction capability. We hope this surprising finding opens up new research directions for the LTSF task. We also advocate revisiting the validity of Transformer-based solutions for other time series analysis tasks (e.g., anomaly detection) in the future."
                },
                {
                    "title": "CoST: Contrastive Learning of Disentangled Seasonal-Trend Representations for Time Series Forecasting",
                    "abstract": "Deep learning has been actively studied for time series forecasting, and the mainstream paradigm is based on the end-to-end training of neural network architectures, ranging from classical LSTM/RNNs to more recent TCNs and Transformers. Motivated by the recent success of representation learning in computer vision and natural language processing, we argue that a more promising paradigm for time series forecasting, is to first learn disentangled feature representations, followed by a simple regression fine-tuning step -- we justify such a paradigm from a causal perspective. Following this principle, we propose a new time series representation learning framework for time series forecasting named CoST, which applies contrastive learning methods to learn disentangled seasonal-trend representations. CoST comprises both time domain and frequency domain contrastive losses to learn discriminative trend and seasonal representations, respectively. Extensive experiments on real-world datasets show that CoST consistently outperforms the state-of-the-art methods by a considerable margin, achieving a 21.3% improvement in MSE on multivariate benchmarks. It is also robust to various choices of backbone encoders, as well as downstream regressors. Code is available at https://github.com/salesforce/CoST."
                },
                {
                    "title": "FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting",
                    "abstract": "Although Transformer-based methods have significantly improved state-of-the-art results for long-term series forecasting, they are not only computationally expensive but more importantly, are unable to capture the global view of time series (e.g. overall trend). To address these problems, we propose to combine Transformer with the seasonal-trend decomposition method, in which the decomposition method captures the global profile of time series while Transformers capture more detailed structures. To further enhance the performance of Transformer for long-term prediction, we exploit the fact that most time series tend to have a sparse representation in well-known basis such as Fourier transform, and develop a frequency enhanced Transformer. Besides being more effective, the proposed method, termed as Frequency Enhanced Decomposed Transformer ({\\bf FEDformer}), is more efficient than standard Transformer with a linear complexity to the sequence length. Our empirical studies with six benchmark datasets show that compared with state-of-the-art methods, FEDformer can reduce prediction error by $14.8\\%$ and $22.6\\%$ for multivariate and univariate time series, respectively. Code is publicly available at https://github.com/MAZiqing/FEDformer."
                },
                {
                    "title": "Clock Anomaly Detection Method of Power Quality Monitoring Device Based on Voltage Sag",
                    "abstract": "Most of the existing power quality monitoring devices are mainly based on the internal clock to achieve time synchronization. However, due to the clock crystal oscillator error or damage and other problems, the monitoring device often has clock anomaly problems, which seriously affects the later data comparative analysis. In order to solve the problem of automatic detection of clock anomaly in power quality monitoring device, a clock anomaly detection method based on voltage sag homologous identification is proposed according to the characteristics of fast propagation speed and negligible time deviation of voltage sag in the system. Firstly, the voltage sag waveform data recorded by different monitoring devices are identified by homology, and then the time deviation of the same sag event recorded in different monitoring devices is compared to judge whether the clock of the monitoring device is abnormal according to the time deviation. The simulation data and measured data are used to verify the method. The results show that the method can accurately detect the clock anomaly of power quality monitoring device, and is of good accuracy and feasibility."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "METRO: A Generic Graph Neural Network Framework for Multivariate Time Series Forecasting",
                    "abstract": "Multivariate time series forecasting has been drawing increasing attention due to its prevalent applications. It has been commonly assumed that leveraging latent dependencies between pairs of variables can enhance prediction accuracy. However, most existing methods suffer from static variable relevance modeling and ignorance of correlation between temporal scales, thereby failing to fully retain the dynamic and periodic interdependencies among variables, which are vital for long- and short-term forecasting. In this paper, we propose METRO, a generic framework with multi-scale temporal graphs neural networks, which models the dynamic and cross-scale variable correlations simultaneously. By representing the multivariate time series as a series of temporal graphs, both intra- and inter-step correlations can be well preserved via message-passing and node embedding update. To enable information propagation across temporal scales, we design a novel sampling strategy to align specific steps between higher and lower scales and fuse the cross-scale information efficiently. Moreover, we provide a modular interpretation of existing GNN-based time series forecasting works as specific instances under our framework. Extensive experiments conducted on four benchmark datasets demonstrate the effectiveness and efficiency of our approach. METRO has been successfully deployed onto the time series analytics platform of Huawei Cloud, where a one-month online test demonstrated that up to 20% relative improvement over state-of-the-art models w.r.t. RSE can be achieved."
                },
                {
                    "title": "Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting",
                    "abstract": "Extending the forecasting time is a critical demand for real applications, such as extreme weather early warning and long-term energy consumption planning. This paper studies the long-term forecasting problem of time series. Prior Transformer-based models adopt various self-attention mechanisms to discover the long-range dependencies. However, intricate temporal patterns of the long-term future prohibit the model from finding reliable dependencies. Also, Transformers have to adopt the sparse versions of point-wise self-attentions for long series efficiency, resulting in the information utilization bottleneck. Going beyond Transformers, we design Autoformer as a novel decomposition architecture with an Auto-Correlation mechanism. We break with the pre-processing convention of series decomposition and renovate it as a basic inner block of deep models. This design empowers Autoformer with progressive decomposition capacities for complex time series. Further, inspired by the stochastic process theory, we design the Auto-Correlation mechanism based on the series periodicity, which conducts the dependencies discovery and representation aggregation at the sub-series level. Auto-Correlation outperforms self-attention in both efficiency and accuracy. In long-term forecasting, Autoformer yields state-of-the-art accuracy, with a 38% relative improvement on six benchmarks, covering five practical applications: energy, traffic, economics, weather and disease. Code is available at this repository: \\url{https://github.com/thuml/Autoformer}."
                },
                {
                    "title": "TS2Vec: Towards Universal Representation of Time Series",
                    "abstract": "This paper presents TS2Vec, a universal framework for learning representations of time series in an arbitrary semantic level. Unlike existing methods, TS2Vec performs contrastive learning in a hierarchical way over augmented context views, which enables a robust contextual representation for each timestamp. Furthermore, to obtain the representation of an arbitrary sub-sequence in the time series, we can apply a simple aggregation over the representations of corresponding timestamps. We conduct extensive experiments on time series classification tasks to evaluate the quality of time series representations. As a result, TS2Vec achieves significant improvement over existing SOTAs of unsupervised time series representation on 125 UCR datasets and 29 UEA datasets. The learned timestamp-level representations also achieve superior results in time series forecasting and anomaly detection tasks. A linear regression trained on top of the learned representations outperforms previous SOTAs of time series forecasting. Furthermore, we present a simple way to apply the learned representations for unsupervised anomaly detection, which establishes SOTA results in the literature. The source code is publicly available at https://github.com/yuezhihan/ts2vec."
                },
                {
                    "title": "Unsupervised Representation Learning for Time Series with Temporal Neighborhood Coding",
                    "abstract": "Time series are often complex and rich in information but sparsely labeled and therefore challenging to model. In this paper, we propose a self-supervised framework for learning generalizable representations for non-stationary time series. Our approach, called Temporal Neighborhood Coding (TNC), takes advantage of the local smoothness of a signal's generative process to define neighborhoods in time with stationary properties. Using a debiased contrastive objective, our framework learns time series representations by ensuring that in the encoding space, the distribution of signals from within a neighborhood is distinguishable from the distribution of non-neighboring signals. Our motivation stems from the medical field, where the ability to model the dynamic nature of time series data is especially valuable for identifying, tracking, and predicting the underlying patients' latent states in settings where labeling data is practically impossible. We compare our method to recently developed unsupervised representation learning approaches and demonstrate superior performance on clustering and classification tasks for multiple datasets."
                },
                {
                    "title": "GLM: General Language Model Pretraining with Autoregressive Blank Infilling",
                    "abstract": "There have been various types of pretraining architectures including autoencoding models (e.g., BERT), autoregressive models (e.g., GPT), and encoder-decoder models (e.g., T5). However, none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding (NLU), unconditional generation, and conditional generation. We propose a General Language Model (GLM) based on autoregressive blank infilling to address this challenge. GLM improves blank filling pretraining by adding 2D positional encodings and allowing an arbitrary order to predict spans, which results in performance gains over BERT and T5 on NLU tasks. Meanwhile, GLM can be pretrained for different types of tasks by varying the number and lengths of blanks. On a wide range of tasks across NLU, conditional and unconditional generation, GLM outperforms BERT, T5, and GPT given the same model sizes and data, and achieves the best performance from a single pretrained model with 1.25\u00d7 parameters of BERT Large , demonstrating its generalizability to different downstream tasks."
                },
                {
                    "title": "Carbon-Oriented Electricity Network Planning and Transformation",
                    "abstract": "The acceleration of distributed energy resources and carbon pricing policies have compelled utilities to act and to prioritize carbon-constrained infrastructure augmentation in their capital programs. To implement various carbon emission reduction policies, power system transmission planning has become more challenging. The existing energy system will face massive retirement of coal-fired power plants (CFPPs), large scale integration of renewable energy and network expansion. In this paper, an electricity network planning and transformation roadmap, which has two milestones, is put forward. In the first stage, a mathematical model is proposed based on the average cost of carbon emission reduction to realize the cooperation of CFPPs retirement and renewable energy investment. It can help the network carry out the transition from a fossil-fuel dominated system to a low-carbon oriented system. Because of the promising prospect of power-to-gas (P2G) technology, in the second milestone, a method based on carbon emission flow (CEF) is employed to help the power-to-gas stations (P2GSes) to select the construction site and capacity. The gas network constraints are modeled to guarantee that P2GSes can work smoothly without energy flow congestion in both electricity and gas networks. According to the simulation results in case studies, our method can reach the emission reduction target more economically and effectively, and the P2GSes can produce and absorb clean energy."
                },
                {
                    "title": "Discrete Graph Structure Learning for Forecasting Multiple Time Series",
                    "abstract": "Time series forecasting is an extensively studied subject in statistics, economics, and computer science. Exploration of the correlation and causation among the variables in a multivariate time series shows promise in enhancing the performance of a time series model. When using deep neural networks as forecasting models, we hypothesize that exploiting the pairwise information among multiple (multivariate) time series also improves their forecast. If an explicit graph structure is known, graph neural networks (GNNs) have been demonstrated as powerful tools to exploit the structure. In this work, we propose learning the structure simultaneously with the GNN if the graph is unknown. We cast the problem as learning a probabilistic graph model through optimizing the mean performance over the graph distribution. The distribution is parameterized by a neural network so that discrete graphs can be sampled differentiably through reparameterization. Empirical evaluations show that our method is simpler, more efficient, and better performing than a recently proposed bilevel learning approach for graph structure learning, as well as a broad array of forecasting models, either deep or non-deep learning based, and graph or non-graph based."
                },
                {
                    "title": "Making Pre-trained Language Models Better Few-shot Learners",
                    "abstract": "The recent GPT-3 model (Brown et al., 2020) achieves remarkable few-shot performance solely by leveraging a natural-language prompt and a few task demonstrations as input context. Inspired by their findings, we study few-shot learning in a more practical scenario, where we use smaller language models for which fine-tuning is computationally efficient. We present LM-BFF\u2014better few-shot fine-tuning of language models\u2014a suite of simple and complementary techniques for fine-tuning language models on a small number of annotated examples. Our approach includes (1) prompt-based fine-tuning together with a novel pipeline for automating prompt generation; and (2) a refined strategy for dynamically and selectively incorporating demonstrations into each context. Finally, we present a systematic evaluation for analyzing few-shot performance on a range of NLP tasks, including classification and regression. Our experiments demonstrate that our methods combine to dramatically outperform standard fine-tuning procedures in this low resource setting, achieving up to 30% absolute improvement, and 11% on average across all tasks. Our approach makes minimal assumptions on task resources and domain expertise, and hence constitutes a strong task-agnostic method for few-shot learning."
                },
                {
                    "title": "Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting",
                    "abstract": "Many real-world applications require the prediction of long sequence time-series, such as electricity consumption planning. Long sequence time-series forecasting (LSTF) demands a high prediction capacity of the model, which is the ability to capture precise long-range dependency coupling between output and input efficiently. Recent studies have shown the potential of Transformer to increase the prediction capacity. However, there are several severe issues with Transformer that prevent it from being directly applicable to LSTF, including quadratic time complexity, high memory usage, and inherent limitation of the encoder-decoder architecture. To address these issues, we design an efficient transformer-based model for LSTF, named Informer, with three distinctive characteristics: (i) a ProbSparse self-attention mechanism, which achieves O(L log L) in time complexity and memory usage, and has comparable performance on sequences' dependency alignment. (ii) the self-attention distilling highlights dominating attention by halving cascading layer input, and efficiently handles extreme long input sequences. (iii) the generative style decoder, while conceptually simple, predicts the long time-series sequences at one forward operation rather than a step-by-step way, which drastically improves the inference speed of long-sequence predictions. Extensive experiments on four large-scale datasets demonstrate that Informer significantly outperforms existing methods and provides a new solution to the LSTF problem."
                },
                {
                    "title": "Exploring Simple Siamese Representation Learning",
                    "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code is made available.1"
                },
                {
                    "title": "A Survey on Contrastive Self-supervised Learning",
                    "abstract": "Self-supervised learning has gained popularity because of its ability to avoid the cost of annotating large-scale datasets. It is capable of adopting self-defined pseudolabels as supervision and use the learned representations for several downstream tasks. Specifically, contrastive learning has recently become a dominant component in self-supervised learning for computer vision, natural language processing (NLP), and other domains. It aims at embedding augmented versions of the same sample close to each other while trying to push away embeddings from different samples. This paper provides an extensive review of self-supervised methods that follow the contrastive approach. The work explains commonly used pretext tasks in a contrastive learning setup, followed by different architectures that have been proposed so far. Next, we present a performance comparison of different methods for multiple downstream tasks such as image classification, object detection, and action recognition. Finally, we conclude with the limitations of the current methods and the need for further techniques and future directions to make meaningful progress."
                },
                {
                    "title": "A Novel Closed-Loop Clustering Algorithm for Hierarchical Load Forecasting",
                    "abstract": "Hierarchical load forecasting (HLF) is an approach to generate forecasts for hierarchical load time series. The performance of HLF can be improved by optimizing the forecasting model and the hierarchical structure. Previous studies mostly focus on the forecasting model while the hierarchical structure is usually formed by clustering of natural attributes like geography, customer type, or the similarities between load profiles. A major limitation of these natural hierarchical structures is the mismatched objectives between clustering and forecasting. Clustering aims to minimize the dissimilarity among customers of a group while forecasting aims to minimize their forecasting errors. The two independent optimizations could limit the overall forecasting performance. Hence, this paper attempts to integrate the hierarchical structure and the forecasting model by a novel closed-loop clustering (CLC) algorithm. It links the objectives of forecasting and clustering by a feedback mechanism to return the goodness-of-fit as the criterion for the clustering. In this way, the hierarchical structure is enhanced by re-assigning the cluster membership and the parameters of the forecasting models are updated iteratively. The method is comparatively assessed with existing HLF methods. Using the same forecasting model, the proposed hierarchical structure outperforms the bottom-up structure by 52.20%, ensemble-based structure by 26.89%, load-profile structures by 19.90%, respectively."
                },
                {
                    "title": "Connecting the Dots: Multivariate Time Series Forecasting with Graph Neural Networks",
                    "abstract": "Modeling multivariate time series has long been a subject that has attracted researchers from a diverse range of fields including economics, finance, and traffic. A basic assumption behind multivariate time series forecasting is that its variables depend on one another but, upon looking closely, it is fair to say that existing methods fail to fully exploit latent spatial dependencies between pairs of variables. In recent years, meanwhile, graph neural networks (GNNs) have shown high capability in handling relational dependencies. GNNs require well-defined graph structures for information propagation which means they cannot be applied directly for multivariate time series where the dependencies are not known in advance. In this paper, we propose a general graph neural network framework designed specifically for multivariate time series data. Our approach automatically extracts the uni-directed relations among variables through a graph learning module, into which external knowledge like variable attributes can be easily integrated. A novel mix-hop propagation layer and a dilated inception layer are further proposed to capture the spatial and temporal dependencies within the time series. The graph learning, graph convolution, and temporal convolution modules are jointly learned in an end-to-end framework. Experimental results show that our proposed model outperforms the state-of-the-art baseline methods on 3 of 4 benchmark datasets and achieves on-par performance with other approaches on two traffic datasets which provide extra structural information."
                },
                {
                    "title": "Aggregation of Multi-Scale Experts for Bottom-Up Load Forecasting",
                    "abstract": "The development of smart grid and new advanced metering infrastructures induces new opportunities and challenges for utilities. Exploiting smart meters information for forecasting stands as a key point for energy providers who have to deal with time varying portfolio of customers as well as grid managers who needs to improve accuracy of local forecasts to face with distributed renewable energy generation development. We propose a new machine learning approach to forecast the system load of a group of customers exploiting individual load measurements in real time and/or exogenous information like weather and survey data. Our approach consists in building experts using random forests trained on some subsets of customers then normalise their predictions and aggregate them with a convex expert aggregation algorithm to forecast the system load. We propose new aggregation methods and compare two strategies for building subsets of customers: 1) hierarchical clustering based on survey data and/or load features and 2) random clustering strategy. These approaches are evaluated on a real data set of residential Irish customers load at a half hourly resolution. We show that our approaches achieve a significant gain in short term load forecasting accuracy of around 25 percent of RMSE."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "Understanding Electricity-Theft Behavior via Multi-Source Data",
                    "abstract": "Electricity theft, the behavior that involves users conducting illegal operations on electrical meters to avoid individual electricity bills, is a common phenomenon in the developing countries. Considering its harmfulness to both power grids and the public, several mechanized methods have been developed to automatically recognize electricity-theft behaviors. However, these methods, which mainly assess users\u2019 electricity usage records, can be insufficient due to the diversity of theft tactics and the irregularity of user behaviors. In this paper, we propose to recognize electricity-theft behavior via multi-source data. In addition to users\u2019 electricity usage records, we analyze user behaviors by means of regional factors (non-technical loss) and climatic factors (temperature) in the corresponding transformer area. By conducting analytical experiments, we unearth several interesting patterns: for instance, electricity thieves are likely to consume much more electrical power than normal users, especially under extremely high or low temperatures. Motivated by these empirical observations, we further design a novel hierarchical framework for identifying electricity thieves. Experimental results based on a real-world dataset demonstrate that our proposed model can achieve the best performance in electricity-theft detection (e.g., at least +3.0% in terms of F0.5) compared with several baselines. Last but not least, our work has been applied by the State Grid of China and used to successfully catch electricity thieves in Hangzhou with a precision of 15% (an improvement from 0% attained by several other models the company employed) during monthly on-site investigation."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Enhancing the Locality and Breaking the Memory Bottleneck of Transformer on Time Series Forecasting",
                    "abstract": "Time series forecasting is an important problem across many domains, including predictions of solar plant energy output, electricity consumption, and traffic jam situation. In this paper, we propose to tackle such forecasting problem with Transformer [1]. Although impressed by its performance in our preliminary study, we found its two major weaknesses: (1) locality-agnostics: the point-wise dot-product self-attention in canonical Transformer architecture is insensitive to local context, which can make the model prone to anomalies in time series; (2) memory bottleneck: space complexity of canonical Transformer grows quadratically with sequence length $L$, making directly modeling long time series infeasible. In order to solve these two issues, we first propose convolutional self-attention by producing queries and keys with causal convolution so that local context can be better incorporated into attention mechanism. Then, we propose LogSparse Transformer with only $O(L(\\log L)^{2})$ memory cost, improving forecasting accuracy for time series with fine granularity and strong long-term dependencies under constrained memory budget. Our experiments on both synthetic data and real-world datasets show that it compares favorably to the state-of-the-art."
                },
                {
                    "title": "Towards Concise Models of Grid Stability",
                    "abstract": "Decentral Smart Grid Control (DSGC) is a new system implementing demand response without significant changes of the infrastructure. It does so by binding the electricity price to the grid frequency. While models of DSGC exist, they rely on various simplifying assumptions. For example, researchers have assumed that the behavior of all participants in the grid is identical. In this paper we study how data-mining techniques can help to remove some of these simplifications, while keeping the representation of the insights concise. We systematically collect the various assumptions and identify questions regarding the system that are still open. Next, we run many simulations, with diverse input values. Finally, we apply decision trees to the resulting data and show that this indeed provides new insights. For example, we discover that the system can be stable even if some participants adapt their energy consumption with a high delay, or that fast adaptation is preferable for system stability."
                },
                {
                    "title": "Building Deep Neural Network Model for Short Term Electricity Consumption Forecasting",
                    "abstract": "Electricity consumption forecasting has a main role in the energy supply management system of a power supply company. A power supply company needs to keep the balancing of the electricity demand and supply for their customers. The target is to forecast the electricity consumption in manufacturing company for each 30-minutes in the next day to prevent the lack of electricity supply from a power supply company. Due to this problem, it is the challenge for short term electricity time series consumption forecasting. In this work, we proposed the model of deep learning neural network with approach the combination of Long Short-Term Memory (LSTM) and Feed Forward Neural Network (FFNN) to perform the electricity forecasting. This proposed method (LSTM-FFNN) was implemented in the time-series data of electricity consumption on a manufacturing company. In our experiment, we used LSTM to perform the time-series forecasting by using historical data of electricity consumption, and we performed FFNN along with additional information which represented by one-hot encoding shape to increase the forecasting performance. Experimental results showed that LSTM-FFNN gave the better result as we compared with our baseline which is the original LSTM and Moving Average (MA) based on the Root Mean Squared Error (RMSE) score."
                },
                {
                    "title": "Hierarchical Electricity Time Series Forecasting for Integrating Consumption Patterns Analysis and Aggregation Consistency",
                    "abstract": "Electricity demand forecasting is a very important problem for energy supply and environmental protection. It can be formalized as a hierarchical time series forecasting problem with the aggregation constraints according to the geographical hierarchy, since the sum of the prediction results of the disaggregated time series should be equal to the prediction results of the aggregated ones. However in most previous work, the aggregation consistency is ensured at the loss of forecast accuracy. In this paper, we propose a novel clustering-based hierarchical electricity time series forecasting approach. Instead of dealing with the geographical hierarchy directly, we explore electricity consumption patterns by clustering analysis and build a new consumption pattern based time series hierarchy. We then present a novel hierarchical forecasting method with consumption hierarchical aggregation constraints to improve the electricity demand predictions of the bottom level, followed by a ``bottom-up\" method to obtain forecasts of the geographical higher levels. Especially, we observe that in our consumption pattern based hierarchy the reconciliation error of the bottom level time series is ``correlated\" to its membership degree of the corresponding cluster (consumption pattern), and hence apply this correlations as the regularization term in our forecasting objective function. Extensive experiments on real-life datasets verify that our approach achieves the best prediction accuracy, compared with the state-of-the-art methods."
                },
                {
                    "title": "An Ensemble Forecasting Method for the Aggregated Load With Subprofiles",
                    "abstract": "With the prevalence of smart meters, fine-grained subprofiles reveal more information about the aggregated load and further help improve the forecasting accuracy. Ensemble is an effective approach for load forecasting. It either generates multiple training datasets or applies multiple forecasting models to produce multiple forecasts. In this letter, a novel ensemble method is proposed to forecast the aggregated load with subprofiles where the multiple forecasts are produced by different groupings of subprofiles. Specifically, the subprofiles are first clustered into different groups and forecasting is conducted on the grouped load profiles individually. Thus, these forecasts can be summed to form the aggregated load forecast. In this way, different aggregated load forecasts can be obtained by varying the number of clusters. Finally, an optimal weighted ensemble approach is employed to combine these forecasts and provide the final forecasting result. Case studies are conducted on two open datasets and verify the effectiveness and superiority of the proposed method."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "An Efficient Approach to Short-Term Load Forecasting at the Distribution Level",
                    "abstract": "Short-term load forecasting at the distribution level predicts the load of substations, feeders, transformers, and possibly customers from half an hour to one week ahead. Effective forecasting is important for the planning and operation of distribution systems. The problem, however, is difficult in view of complicated load features, the large number of distribution-level nodes, and possible switching operations. In this paper, a new forecasting approach within the hierarchical structure is presented to solve these difficulties. Load of the root node at any user-defined subtree is first forecast by a wavelet neural network with appropriate inputs. Child nodes categorized as \u201cregular\u201d and \u201cirregular\u201d based on load pattern similarities are then forecast separately. Load of a regular child node is simply forecast as the proportion from the parent node load forecast while the load of an irregular child node is forecast by an individual neural network model. Switching operation detection and follow-up adjustments are also performed to capture abnormal changes and improve the forecasting accuracy. This new approach captures load characteristics of nodes at different levels, takes advantage of pattern similarities between a parent node and its child nodes, detects abnormalities, and provides high quality forecasts as demonstrated by two practical datasets."
                },
                {
                    "title": "Forecast UPC-level FMCG demand, Part II: Hierarchical reconciliation",
                    "abstract": "In a big data enabled environment, manufacturers and distributors may have access to previously unobserved retailer-level demand related information. This additional information can be considered in demand forecasting to produce more accurate forecasts, and thus enable better stock-outs management. In Part II of this two-part paper, we explore the hierarchical nature of fast moving consumer goods (FMCG) demand (represented by sales) time series and produce one week ahead rolling forecasts on universal product code (UPC) level (or distributor level, as per our definition below). We show that the hierarchical forecasting framework has significant accuracy improvement over the conventional univariate forecasting methods. The main reason of the observed improvements is due to the price and promotion information available at the retailer level, which is assumed to be unknown to the distributor. To reconcile forecasts according to the hierarchy, only the forecast values at retailer level are needed, the business strategies of individual retailers remain proprietary. A freely available dataset is considered to encourage further exploration. Data exploratory analysis and visualization tools are discussed in Part I of the paper."
                },
                {
                    "title": "Load forecasting techniques and methodologies: A review",
                    "abstract": "Electricity demand forecasts are extremely important for energy suppliers and other participants in electric energy generation, transmission, distribution and markets. Accurate models for electric power load forecasting are essential to the operation and planning of a utility company. Load forecasts are extremely important for energy suppliers and other participants in electric energy generation, transmission, distribution and markets. This paper presents a review of electricity demand forecasting techniques. The various types of methodologies and models are included in the literature. Load forecasting can be broadly divided into three categories: short-term forecasts which are usually from one hour to one week, medium forecasts which are usually from a week to a year, and long-term forecasts which are longer than a year. Based on the various types of studies presented in these papers, the load forecasting techniques may be presented in three major groups: Traditional Forecasting technique, Modified Traditional Technique and Soft Computing Technique."
                },
                {
                    "title": "Demand Side Management: Demand Response, Intelligent Energy Systems, and Smart Loads",
                    "abstract": "Energy management means to optimize one of the most complex and important technical creations that we know: the energy system. While there is plenty of experience in optimizing energy generation and distribution, it is the demand side that receives increasing attention by research and industry. Demand Side Management (DSM) is a portfolio of measures to improve the energy system at the side of consumption. It ranges from improving energy efficiency by using better materials, over smart energy tariffs with incentives for certain consumption patterns, up to sophisticated real-time control of distributed energy resources. This paper gives an overview and a taxonomy for DSM, analyzes the various types of DSM, and gives an outlook on the latest demonstration projects in this domain."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively model Electricity Time Series (ETS) data to enhance performance in various power system applications while addressing the unique hierarchical and temporal characteristics of the data?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it can lead to significant advancements in the modeling of ETS data, which is essential for optimizing power system operations, improving demand-side management, and enhancing grid stability. By developing effective modeling methods, we can facilitate better decision-making in energy consumption, contribute to economic efficiency, and support low-carbon initiatives. This research could pave the way for future studies that leverage advanced modeling techniques, potentially leading to practical applications in smart grid technologies and energy management systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complex hierarchical structure of ETS data, which requires the integration of information across different levels of granularity (e.g., user, district, city). Additionally, the temporal dependencies within the ETS data add another layer of complexity, as traditional modeling approaches may not adequately capture these dynamics. Naive approaches may fail because they do not account for the intricate relationships between different data levels or the temporal patterns inherent in electricity consumption. Overcoming these technical and theoretical obstacles necessitates sophisticated modeling techniques that can effectively handle both hierarchical and temporal dependencies.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on pre-training approaches that utilize limited datasets, resulting in models that do not generalize well to downstream tasks in power systems. Additionally, existing models have not been tailored to the specific characteristics of ETS data, leading to a significant research gap. Barriers such as insufficient domain-specific knowledge and the lack of comprehensive pre-training data for power systems have hindered progress. Our approach aims to address these limitations by developing a foundation model specifically designed for ETS data, incorporating both hierarchical and temporal aspects that previous works have overlooked.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a foundation model that integrates hierarchical dependency modeling and temporal analysis of ETS data. We will utilize a large-scale dataset of electricity consumption across various instances and timeframes, employing metrics such as forecasting accuracy and anomaly detection rates to evaluate performance. The expected outcomes include improved accuracy in load forecasting, enhanced detection of anomalies, and better insights into consumer behavior, ultimately leading to more effective applications in power"
            }
        },
        "author_data": {
            "67ab9d2f-8f98-4c03-8016-bce2cc54a751": {
                "pk": "67ab9d2f-8f98-4c03-8016-bce2cc54a751",
                "name": "Shihao Tu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "2dd9bf27-7e93-45a7-a9e3-8db93d9f9e07": {
                "pk": "2dd9bf27-7e93-45a7-a9e3-8db93d9f9e07",
                "name": "Yupeng Zhang",
                "collaborators": [
                    "Kaushik Bhattacharya",
                    "Guangzhou Guo",
                    "Peng Wang"
                ],
                "domain": [
                    "Machine Learning",
                    "Multiscale Modeling",
                    "Computational Physics",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Iterated learning and multiscale modeling of history-dependent architectured metamaterials",
                        "abstract": "Neural network based models have emerged as a powerful tool in multiscale modeling of materials. One promising approach is to use a neural network based model, trained using data generated from repeated solution of an expensive small scale model, as a surrogate for the small scale model in application scale simulations. Such approaches have been shown to have the potential accuracy of concurrent multiscale methods like FE2, but at the cost comparable to empirical methods like classical constitutive models or parameter passing. A key question is to understand how much and what kind of data is necessary to obtain an accurate surrogate. This paper examines this question for history dependent elastic-plastic behavior of an architected metamaterial modeled as a truss. We introduce an iterative approach where we use the rich arbitrary class of trajectories to train an initial model, but then iteratively update the class of trajectories with those that arise in large scale simulation and use transfer learning to update the model. We show that such an approach converges to a highly accurate surrogate, and one that is transferable."
                    },
                    {
                        "title": "Nonlinear Stability of Black Holes with a Stable Light Ring",
                        "abstract": "Recently, ultracompact objects have been found to be susceptible to a new nonlinear instability, known as the light-ring instability, triggered by stable light rings. This discovery raises concerns about the viability of these objects as alternatives to black holes. In this work, we investigate the presence of the light-ring instability in scalarized Reissner-Nordstr\\\"om black holes, which have been previously shown to admit stable light rings. We employ fully nonlinear numerical evolutions of both scalarized black holes with and without stable light rings, perturbing them initially with spherically symmetric scalar perturbations. Our simulations demonstrate the long-term stability of these scalarized black holes, suggesting that the presence of a stable light ring may not necessarily induce the light-ring instability."
                    }
                ]
            },
            "4266e77e-bfb2-48a1-85b4-85189c7d5669": {
                "pk": "4266e77e-bfb2-48a1-85b4-85189c7d5669",
                "name": "Jing Zhang",
                "collaborators": [],
                "domain": [
                    "Algebraic Geometry",
                    "Combinatorial Set Theory",
                    "Quantum Information",
                    "Nonlinear Dynamics"
                ],
                "publications": [
                    {
                        "title": "Monochromatic Sumset Without the use of large cardinals",
                        "abstract": "We show in this note that in the forcing extension by $Add(\\omega,\\beth_{\\omega})$, the following Ramsey property holds: for any $r\\in \\omega$ and any $f: \\mathbb{R}\\to r$, there exists an infinite $X\\subset \\mathbb{R}$ such that $X+X$ is monochromatic under $f$. We also show the Ramsey statement above is true in $\\mathrm{ZFC}$ when $r=2$. This answers two questions by Komj\\'ath, Leader, Russell, Shelah, Soukup and Vidny\\'anszky."
                    },
                    {
                        "title": "Reflection principles, GCH and the uniformization properties",
                        "abstract": "Reflection principles (or dually speaking, compactness principles) often give rise to combinatorial guessing principles. Uniformization properties, on the other hand, are examples of anti-guessing principles. We discuss the tension and the compatibility between reflection principles and uniformization properties at the level of the second uncountable cardinal."
                    },
                    {
                        "title": "Hodge Cohomology Criteria For Affine Varieties",
                        "abstract": "We give several new criteria for a quasi-projective variety to be affine. In particular, we prove that an algebraic manifold $Y$ with dimension $n$ is affine if and only if $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$, $i>0$ and $\\kappa(D, X)=n$, i.e., there are $n$ algebraically independent nonconstant regular functions on $Y$, where $X$ is the smooth completion of $Y$, $D$ is the effective boundary divisor with support $X-Y$ and $\\Omega^j_Y$ is the sheaf of regular $j$-forms on $Y$. This proves Mohan Kumar's affineness conjecture for algebraic manifolds and gives a partial answer to J.-P. Serre's Steinness question \\cite{36} in algebraic case since the associated analytic space of an affine variety is Stein [15, Chapter VI, Proposition 3.1]."
                    },
                    {
                        "title": "Cluster States for Continuous-Variable Multipartite Entanglement",
                        "abstract": "We introduce a new class of continuous-variable (CV) multipartite entangled states, the CV cluster states, which might be generated from squeezing and kerr-like interaction. The entanglement properties of these states are studied in terms of classical communication and local operations. The quantum teleportation network with cluster states is investigated. The graph states as the general forms of cluster states are presented, which may be used to generate CV Greenberger-Horne-Zeilinger states by simply local measurements and classical communication. A chain for one-dimensional example of cluster states can be readily experimentally produced only with squeezed light and beamsplitters."
                    },
                    {
                        "title": "Local complementation rule for continuous-variable four-mode unweighted graph states",
                        "abstract": "The local complementation rule is applied for continuous-variable (CV) graph states in the paper, which is an elementary graph transformation rule and successive application of which generates the orbit of any graph states. The corresponding local Gaussian transformations of local complementation for four-mode unweighted graph states were found, which do not mirror the form of the local Clifford unitary of qubit exactly. This work is an important step to characterize the local Gaussian equivalence classes of CV graph states."
                    },
                    {
                        "title": "Graphical rule of transforming continuous-variable graph states by local homodyne detection",
                        "abstract": "Graphical rule, describing that any single-mode homodyne detection turns a given continuous-variable (CV) graph state into a new one, is presented. Employing two simple graphical rules: local complement operation and vertex deletion (single quadrature-amplitude $\\hat{x}$ measurement), the graphical rule for any single-mode quadrature component measurement can be obtained. The shape of CV weighted graph state may be designed and constructed easily from a given larger graph state by applying this graphical rule."
                    },
                    {
                        "title": "Continuous-variable multipartite unlockable bound entangled Gaussian states",
                        "abstract": "Continuous-variable (CV) multipartite unlockable bound-entangled states is investigated in this paper. Comparing with the qubit multipartite unlockable bound-entangled states, CV multipartite unlockable bound-entangled states present the new and different properties. CV multipartite unlockable bound-entangled states may serve as a useful quantum resource for new multiparty communication schemes. The experimental protocol for generating CV unlockable bound-entangled states is proposed with a setup that is at present accessible."
                    },
                    {
                        "title": "A tail cone version of the Halpern-L\u00e4uchli theorem at a large cardinal",
                        "abstract": "The classical Halpern-L\\\"auchli theorem states that for any finite coloring of a finite product of finitely branching perfect trees of height $\\omega$, there exist strong subtrees sharing the same level set such that tuples consisting of elements lying on the same level get the same color. Relative to large cardinals, we establish the consistency of a tail cone version of the Halpern-L\\\"auchli theorem at large cardinal, which, roughly speaking, deals with many colorings simultaneously and diagonally. Among other applications, we generalize a polarized partition relation on rational numbers due to Laver and Galvin to one on linear orders of larger saturation."
                    },
                    {
                        "title": "Threefolds with Vanishing Hodge Cohomology",
                        "abstract": "We consider algebraic manifolds $Y$ of dimension 3 over $\\Bbb{C}$ with $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$ and $i>0$. Let $X$ be a smooth completion of $Y$ with $D=X-Y$, an effective divisor on $X$ with normal crossings. If the $D$-dimension of $X$ is not zero, then $Y$ is a fibre space over a smooth affine curve $C$ (i.e., we have a surjective morphism from $Y$ to $C$ such that general fibre is smooth and irreducible) such that every fibre satisfies the same vanishing condition. If an irreducible smooth fibre is not affine, then the Kodaira dimension of $X$ is $-\\infty$ and the $D$-dimension of X is 1. We also discuss sufficient conditions from the behavior of fibres or higher direct images to guarantee the global vanishing of Hodge cohomology and the affineness of $Y$."
                    },
                    {
                        "title": "There Exist Nontrivial Threefolds with Vanishing Hodge Cohomology",
                        "abstract": "We analyze the structure of the algebraic manifolds $Y$ of dimension 3 with $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$, $i>0$ and $h^0(Y, {\\mathcal{O}}_Y) > 1$, by showing the deformation invariant of some open surfaces. Secondly, we show when a smooth threefold with nonconstant regular functions satisfies the vanishing Hodge cohomology. As an application, we prove the existence of nonaffine and nonproduct threefolds $Y$ with this property by constructing a family of a certain type of open surfaces parametrized by the affine curve $\\C-\\{0\\}$ such that the corresponding smooth completion $X$ has Kodaira dimension $-\\infty$ and $D$-dimension 1, where $D$ is the effective boundary divisor with support $X-Y$."
                    },
                    {
                        "title": "On the $D$-dimension of a certain type of threefolds",
                        "abstract": "Let $Y$ be an algebraic manifold of dimension 3 with $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$, $i>0$ and $h^0(Y, {\\mathcal{O}}_Y) > 1$. Let $X$ be a smooth completion of $Y$ such that the boundary $X-Y$ is the support of an effective divisor $D$ on $X$ with simple normal crossings. We prove that the $D$-dimension of $X$ cannot be 2, i.e., either any two nonconstant regular functions are algebraically dependent or there are three algebraically independent nonconstant regular functions on $Y$. Secondly, if the $D$-dimension of $X$ is greater than 1, then the associated scheme of $Y$ is isomorphic to Spec$\\Gamma(Y, {\\mathcal{O}}_Y)$. Furthermore, we prove that an algebraic manifold $Y$ of any dimension $d\\geq 1$ is affine if and only if $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$, $i>0$ and it is regularly separable, i.e., for any two distinct points $y_1$, $y_2$ on $Y$, there is a regular function $f$ on $Y$ such that $f(y_1)\\neq f(y_2)$."
                    },
                    {
                        "title": "On threefolds without nonconstant regular functions",
                        "abstract": "We consider smooth threefolds $Y$ defined over $\\Bbb{C}$ with $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$, $i>0$. Let $X$ be a smooth projective threefold containing $Y$ and $D$ be the boundary divisor with support $X-Y$.   We are interested in the following question: What geometry information of $X$ can be obtained from the regular function information on $Y$?   Suppose that the boundary $X-Y$ is a smooth projective surface. In this paper, we analyse two different cases, i.e., there are no nonconstant regular functions on $Y$ or there are lots of regular functions on $Y$. More precisely, if $H^0(Y, {\\mathcal{O}}_Y)=\\Bbb{C}$, we prove that ${1/2}(c_1^2+c_2)\\cdot D=\\chi({\\mathcal{O}}_D)\\geq 0$. In particular, if the line bundle ${\\mathcal{O}}_D(D)$ is not torsion, then $q=h^1(X, {\\mathcal{O}}_X)=0$, ${1/2}(c_1^2+c_2)\\cdot D=\\chi({\\mathcal{O}}_D)=0$,   $\\chi({\\mathcal{O}}_X) >0$ and $K_X$ is not nef. If there is a positive constant $c$ such that $h^0(X, {\\mathcal{O}}_X(nD))\\geq c n^3$ for all sufficiently large $n$ (we say that $D$ is big or the $D$-dimension of $X$ is 3) and $D$ has no exceptional curves, then $|nD|$ is base point free for $n\\gg 0$. Therefore $Y$ is affine if $D$ is big."
                    },
                    {
                        "title": "Algebraic Stein Varieties",
                        "abstract": "It is well-known that the associated analytic space of an affine variety defined over $\\mathbb{C}$ is Stein but the converse is not true, that is, an algebraic Stein variety is not necessarily affine. In this paper, we give sufficient and necessary conditions for an algebraic Stein variety to be affine. One of our results is that an irreducible quasi-projective variety $Y$ defined over $\\mathbb{C}$ with dimension $d$ ($d\\geq 1$) is affine if and only if $Y$ is Stein, $H^i(Y, {\\mathcal{O}}_Y)=0$ for all $i>0$ and $\\kappa(D, X)= d$ (i.e., $D$ is a big divisor), where $X$ is a projective variety containing $Y$ and $D$ is an effective divisor with support $X-Y$. If $Y$ is algebraic Stein but not affine, we also discuss the possible transcendental degree of the nonconstant regular functions on $Y$. We prove that $Y$ cannot have $d-1$ algebraically independent nonconstant regular functions. The interesting phenomenon is that the transcendental degree can be even if the dimension of $Y$ is even and the degree can be odd if the dimension of $Y$ is odd."
                    },
                    {
                        "title": "Affine Algebraic Varieties",
                        "abstract": "In this paper, we give new criteria for affineness of a variety defined over $\\Bbb{C}$. Our main result is that an irreducible algebraic variety $Y$ (may be singular) of dimension $d$ ($d\\geq 1$) defined over $\\Bbb{C}$ is an affine variety if and only if $Y$ contains no complete curves, $H^i(Y, {\\mathcal{O}}_Y)=0$ for all $i>0$ and the boundary $X-Y$ is support of a big divisor, where $X$ is a projective variety containing $Y$. We construct three examples to show that a variety is not affine if it only satisfies two conditions among these three conditions. We also give examples to demonstrate the difference between the behavior of the boundary divisor $D$ and the affineness of $Y$. If $Y$ is an affine variety, then the ring $\\Gamma (Y, {\\mathcal{O}}_Y)$ is noetherian. However, to prove that $Y$ is an affine variety, we do not start from this ring. We explain why we do not need to check the noetherian property of the ring $\\Gamma (Y, {\\mathcal{O}}_Y)$ directly but use the techniques of sheaf and cohomology."
                    },
                    {
                        "title": "Graphical description of local Gaussian operations for continuous-variable weighted graph states",
                        "abstract": "The form of a local Clifford (LC, also called local Gaussian (LG)) operation for the continuous-variable (CV) weighted graph states is presented in this paper, which is the counterpart of the LC operation of local complementation for qubit graph states. The novel property of the CV weighted graph states is shown, which can be expressed by the stabilizer formalism. It is distinctively different from the qubit weighted graph states, which can not be expressed by the stabilizer formalism. The corresponding graph rule, stated in purely graph theoretical terms, is described, which completely characterizes the evolution of CV weighted graph states under this LC operation. This LC operation may be applied repeatedly on a CV weighted graph state, which can generate the infinite LC equivalent graph states of this graph state. This work is an important step to characterize the LC equivalence class of CV weighted graph states."
                    },
                    {
                        "title": "Bertini Type Theorems",
                        "abstract": "Let $X$ be a smooth irreducible projective variety of dimension at least 2 over an algebraically closed field of characteristic 0 in the projective space ${\\mathbb{P}}^n$.   Bertini's Theorem states that a general hyperplane $H$ intersects $X$ with an irreducible smooth subvariety of $X$. However, the precise location of the smooth hyperplane section is not known. We show that for any $q\\leq n+1$ closed points in general position and any degree $a>1$, a general hypersurface $H$ of degree $a$ passing through these $q$ points intersects $X$ with an irreducible smooth codimension 1 subvariety on $X$. We also consider linear system of ample divisors and give precise location of smooth elements in the system. Similar result can be obtained for compact complex manifolds with holomorphic maps into projective spaces."
                    },
                    {
                        "title": "A Literature Survey of Cooperative Caching in Content Distribution Networks",
                        "abstract": "Content distribution networks (CDNs) which serve to deliver web objects (e.g., documents, applications, music and video, etc.) have seen tremendous growth since its emergence. To minimize the retrieving delay experienced by a user with a request for a web object, caching strategies are often applied - contents are replicated at edges of the network which is closer to the user such that the network distance between the user and the object is reduced. In this literature survey, evolution of caching is studied. A recent research paper [15] in the field of large-scale caching for CDN was chosen to be the anchor paper which serves as a guide to the topic. Research studies after and relevant to the anchor paper are also analyzed to better evaluate the statements and results of the anchor paper and more importantly, to obtain an unbiased view of the large scale collaborate caching systems as a whole."
                    },
                    {
                        "title": "Rado's conjecture and its Baire version",
                        "abstract": "Rado's Conjecture is a compactness/reflection principle that says any nonspecial tree of height $\\omega_1$ has a nonspecial subtree of size $\\leq \\aleph_1$. Though incompatible with Martin's Axiom, Rado's Conjecture turns out to have many interesting consequences that are consequences of forcing axioms. In this paper, we obtain consistency results concerning Rado's Conjecture and its Baire version. In particular, we show a fragment of PFA, that is the forcing axiom for \\emph{Baire Indestructibly proper forcings}, is compatible with the Baire Rado's Conjecture. As a corollary, Baire Rado's Conjecture does not imply Rado's Conjecture. Then we discuss the strength and limitations of the Baire Rado's Conjecture regarding its interaction with simultaneous stationary reflection and some families of weak square principles. Finally we investigate the influence of the Rado's Conjecture on some polarized partition relations."
                    },
                    {
                        "title": "Some remarks on uncountable rainbow Ramsey theory",
                        "abstract": "We discuss the rainbow Ramsey theorems at limit cardinals and successors of singular cardinals, addressing some questions in \\cite{MR2354904} and \\cite{MR2902230}. In particular, we show for inaccessible $\\kappa$, $\\kappa\\to^{poly}(\\kappa)^2_{2-bdd}$ does not characterize weak compactness and for singular $\\kappa$, $\\mathrm{GCH}+\\square_\\kappa$ implies $\\kappa^+\\not\\to^{poly} (\\eta)^2_{<\\kappa-bdd}$ for any $\\eta\\geq cf(\\kappa)^+$ and $\\kappa^+\\to^{poly} (\\nu)^2_{<\\kappa-bdd}$ for any $\\nu<cf(\\kappa)^+$. We also provide a simplified construction of a model for $\\omega_2\\not\\to^{poly} (\\omega_1)^2_{2-bdd}$ originally constructed in \\cite{MR2902230} and show the witnessing coloring is indestructible under strongly proper forcings but destructible under some c.c.c forcing. Finally, we conclude with some remarks and questions on possible generalizations to rainbow partition relations for triples."
                    },
                    {
                        "title": "Almost global solutions to two classes of 1-d Hamiltonian Derivative Nonlinear Schr\u00f6dinger equations",
                        "abstract": "Consider two kinds of 1-d Hamiltonian Derivative Nonlinear Schr\\\"odinger (DNLS) equations with respect to different symplectic forms under periodic boundary conditions. The nonlinearities of these equations depend not only on $(x,\\psi,\\bar{\\psi})$ but also on $(\\psi_x,\\bar{\\psi}_x)$, which means the nonlinearities of these equations are unbounded. Suppose that the nonlinearities depend on the space-variable $x$ periodically.   Under some assumptions, for most potentials of these two kinds of Hamiltonian DNLS equations, if the initial value is smaller than $\\varepsilon\\ll1$ in $p$-Sobolev norm, then the corresponding solution to these equations is also smaller than $2\\varepsilon$ during a time interval $(-c\\varepsilon^{-r_*},c\\varepsilon^{-r_*})$(for any given positive $r_*$). The main methods are constructing Birkhoff normal forms to two kinds of Hamiltonian systems which have unbounded nonlinearities and using the special symmetry of the unbounded nonlinearities of Hamiltonian functions to obtain a long time estimate of the solution in $p$-Sobolev norm."
                    }
                ]
            },
            "6148dcef-f28f-4758-830d-67da66731f90": {
                "pk": "6148dcef-f28f-4758-830d-67da66731f90",
                "name": "Zhendong Fu",
                "collaborators": [
                    "Yixi Su",
                    "Harikrishnan S. Nair",
                    "J\u00f6rg Voigt",
                    "Th. Br\u00fcckel",
                    "Zhouxiang Fang",
                    "Min Yu",
                    "Boning Zhang",
                    "Xuanwen Huang",
                    "Xiaoqi Tang",
                    "Yang Yang"
                ],
                "domain": [
                    "Magnetism",
                    "Neutron Scattering",
                    "Machine Learning",
                    "Social Media Analytics"
                ],
                "publications": [
                    {
                        "title": "Approaching the Ground State of Frustrated A-site Spinels: A Combined Magnetization and Polarized Neutron Scattering Study",
                        "abstract": "We re-investigate the magnetically frustrated, {\\it diamond-lattice-antiferromagnet} spinels FeAl$_2$O$_4$ and MnAl$_2$O$_4$ using magnetization measurements and diffuse scattering of polarized neutrons. In FeAl$_2$O$_4$, macroscopic measurements evidence a \"cusp\" in zero field-cooled susceptibility around 13~K. Dynamic magnetic susceptibility and {\\it memory effect} experiments provide results that do not conform with a canonical spin-glass scenario in this material. Through polarized neutron scattering studies, absence of long-range magnetic order down to 4~K is confirmed in FeAl$_2$O$_4$. By modeling the powder averaged differential magnetic neutron scattering cross-section, we estimate that the spin-spin correlations in this compound extend up to the third nearest-neighbour shell. The estimated value of the Land\\'{e} $g$ factor points towards orbital contributions from Fe$^{2+}$. This is also supported by a Curie-Weiss analysis of the magnetic susceptibility. MnAl$_2$O$_4$, on the contrary, undergoes a magnetic phase transition into a long-range ordered state below $\\approx$ 40~K, which is confirmed by macroscopic measurements and polarized neutron diffraction. However, the polarized neutron studies reveal the existence of prominent spin-fluctuations co-existing with long-range antiferromagnetic order. The magnetic diffuse intensity suggests a similar short range order as in FeAl$_2$O$_4$. Results of the present work supports the importance of spin-spin correlations in understanding magnetic response of frustrated magnets like $A$-site spinels which have predominant short-range spin correlations reminiscent of the \"spin liquid\" state."
                    },
                    {
                        "title": "How to Generate Popular Post Headlines on Social Media?",
                        "abstract": "Posts, as important containers of user-generated-content pieces on social media, are of tremendous social influence and commercial value. As an integral components of a post, the headline has a decisive contribution to the post's popularity. However, current mainstream method for headline generation is still manually writing, which is unstable and requires extensive human effort. This drives us to explore a novel research question: Can we automate the generation of popular headlines on social media? We collect more than 1 million posts of 42,447 celebrities from public data of Xiaohongshu, which is a well-known social media platform in China. We then conduct careful observations on the headlines of these posts. Observation results demonstrate that trends and personal styles are widespread in headlines on social medias and have significant contribution to posts's popularity. Motivated by these insights, we present MEBART, which combines Multiple preference-Extractors with Bidirectional and Auto-Regressive Transformers (BART), capturing trends and personal styles to generate popular headlines on social medias. We perform extensive experiments on real-world datasets and achieve state-of-the-art performance compared with several advanced baselines. In addition, ablation and case studies demonstrate that MEBART advances in capturing trends and personal styles."
                    },
                    {
                        "title": "Field-Driven Self-Assembly of Magnetite Nanoparticles Investigated Using Small-Angle Neutron Scattering",
                        "abstract": "The magnetic-field-induced assembly of magnetic nanoparticles (NPs) provides a unique and flexible strategy in the design and the fabrication of functional nanostructures and devices. We have investigated the field-driven self-assembly of core-shell magnetite NPs dispersed in toluene by means of in situ small angle neutron scattering (SANS). The form factor of the core-shell NPs was characterized and analyzed using SANS with polarized neutrons. Large-scale aggregation of magnetite NPs has formed above 0.02 T as indicated by very-small angle neutron scattering measurements. Three-dimensional long-range ordered superlattice of magnetite NPs was revealed under the application of moderate magnetic field. The crystal structure of the superlattice has been identified as a face-centered cubic one."
                    }
                ]
            },
            "b3f70d75-0a73-4ccc-ab10-6ba602a7b7aa": {
                "pk": "b3f70d75-0a73-4ccc-ab10-6ba602a7b7aa",
                "name": "Yin Zhang",
                "collaborators": [
                    "Yangyang Xu",
                    "Wotao Yin",
                    "Zaiwen Wen",
                    "Junwen Zhang"
                ],
                "domain": [
                    "Web Development",
                    "Matrix Factorization",
                    "Natural Language Processing",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "WebScript -- A Scripting Language for the Web",
                        "abstract": "WebScript is a scripting language for processing Web documents. Designed as an extension to Jacl, the Java implementation of Tcl, WebScript allows programmers to manipulate HTML in the same way as Tcl manipulates text strings and GUI elements. This leads to a completely new way of writing the next generation of Web applications. This paper presents the motivation behind the design and implementation of WebScript, an overview of its major features, as well as some demonstrations of its power."
                    },
                    {
                        "title": "An Alternating Direction Algorithm for Matrix Completion with Nonnegative Factors",
                        "abstract": "This paper introduces an algorithm for the nonnegative matrix factorization-and-completion problem, which aims to find nonnegative low-rank matrices X and Y so that the product XY approximates a nonnegative data matrix M whose elements are partially known (to a certain accuracy). This problem aggregates two existing problems: (i) nonnegative matrix factorization where all entries of M are given, and (ii) low-rank matrix completion where nonnegativity is not required. By taking the advantages of both nonnegativity and low-rankness, one can generally obtain superior results than those of just using one of the two properties. We propose to solve the non-convex constrained least-squares problem using an algorithm based on the classic alternating direction augmented Lagrangian method. Preliminary convergence properties of the algorithm and numerical simulation results are presented. Compared to a recent algorithm for nonnegative matrix factorization, the proposed algorithm produces factorizations of similar quality using only about half of the matrix entries. On tasks of recovering incomplete grayscale and hyperspectral images, the proposed algorithm yields overall better qualities than those produced by two recent matrix-completion algorithms that do not exploit nonnegativity."
                    },
                    {
                        "title": "HierarchicalContrast: A Coarse-to-Fine Contrastive Learning Framework for Cross-Domain Zero-Shot Slot Filling",
                        "abstract": "In task-oriented dialogue scenarios, cross-domain zero-shot slot filling plays a vital role in leveraging source domain knowledge to learn a model with high generalization ability in unknown target domain where annotated data is unavailable. However, the existing state-of-the-art zero-shot slot filling methods have limited generalization ability in target domain, they only show effective knowledge transfer on seen slots and perform poorly on unseen slots. To alleviate this issue, we present a novel Hierarchical Contrastive Learning Framework (HiCL) for zero-shot slot filling. Specifically, we propose a coarse- to fine-grained contrastive learning based on Gaussian-distributed embedding to learn the generalized deep semantic relations between utterance-tokens, by optimizing inter- and intra-token distribution distance. This encourages HiCL to generalize to the slot types unseen at training phase. Furthermore, we present a new iterative label set semantics inference method to unbiasedly and separately evaluate the performance of unseen slot types which entangled with their counterparts (i.e., seen slot types) in the previous zero-shot slot filling evaluation methods. The extensive empirical experiments on four datasets demonstrate that the proposed method achieves comparable or even better performance than the current state-of-the-art zero-shot slot filling approaches."
                    }
                ]
            },
            "a493420f-656d-481b-9a54-56f8860667b9": {
                "pk": "a493420f-656d-481b-9a54-56f8860667b9",
                "name": "Yang Yang",
                "collaborators": [],
                "domain": [
                    "Mathematical Analysis",
                    "Inverse Problems",
                    "Machine Learning",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "The Brezis-Nirenberg problem for the fractional $p$-Laplacian involving critical Hardy-Sobolev exponents",
                        "abstract": "We obtain existence, multiplicity, and bifurcation results for the Brezis-Nirenberg problem for the fractional $p$\\nobreakdash-Laplacian operator, involving critical Hardy-Sobolev exponents. Our results are mainly extend results in the literature for $\\alpha=0$. In the absence of an explicit formula for a minimizer in the fractional Hardy-Sobolev inequality $\\alpha\\not= 0$, we get around this difficulty by working with certain asymptotic estimates for minimizers recently obtained in \\cite{MM}."
                    },
                    {
                        "title": "Determining the first order perturbation of a bi-harmonic operator on bounded and unbounded domains from partial data",
                        "abstract": "In this paper we study inverse boundary value problems with partial data for the bi-harmonic operator with first order perturbation. We consider two types of subsets of $\\mathbb{R}^{n}(n\\geq 3)$, one is an infinite slab, the other is a bounded domain. In the case of a slab, we show that, from Dirichlet and Neumann data given either on the different boundary hyperplanes of the slab or on the same boundary hyperplane, one can uniquely determine the magnetic potential and the electric potential.   In the case of a bounded domain, we show the unique determination of the magnetic potential and the electric potential from partial Dirichlet and Neumann data under two different assumptions. The first assumption is that the magnetic and electric potentials are known in a neighborhood of the boundary, in this situation we obtain the uniqueness result when the Dirichlet and Neumann data are only given on two arbitrary open subsets of the boundary. The second assumption is that the Dirichlet and Neumann data are known on the same part of the boundary whose complement is a part of a hyperplane, we also establish the unique determination result in this local data case."
                    },
                    {
                        "title": "Upper triangular matrices and Billiard Arrays",
                        "abstract": "Fix a nonnegative integer $d$, a field $\\mathbb{F}$, and a vector space $V$ over $\\mathbb{F}$ with dimension $d+1$. Let $T$ denote an invertible upper triangular matrix in ${\\rm Mat}_{d+1}(\\mathbb{F})$. Using $T$ we construct three flags on $V$. We find a necessary and sufficient condition on $T$ for these three flags to be totally opposite. In this case, we use these three totally opposite flags to construct a Billiard Array $B$ on $V$. It is known that $B$ is determined up to isomorphism by a certain triangular array of scalar parameters called the $B$-values. We compute these $B$-values in terms of the entries of $T$. We describe the set of isomorphism classes of Billiard Arrays in terms of upper triangular matrices."
                    },
                    {
                        "title": "BASN -- Learning Steganography with Binary Attention Mechanism",
                        "abstract": "Secret information sharing through image carrier has aroused much research attention in recent years with images' growing domination on the Internet and mobile applications. However, with the booming trend of convolutional neural networks, image steganography is facing a more significant challenge from neural-network-automated tasks. To improve the security of image steganography and minimize task result distortion, models must maintain the feature maps generated by task-specific networks being irrelative to any hidden information embedded in the carrier. This paper introduces a binary attention mechanism into image steganography to help alleviate the security issue, and in the meanwhile, increase embedding payload capacity. The experimental results show that our method has the advantage of high payload capacity with little feature map distortion and still resist detection by state-of-the-art image steganalysis algorithms."
                    },
                    {
                        "title": "Some $q$-exponential formulas for finite-dimensional $\\square_q$-modules",
                        "abstract": "We consider the algebra $\\square_q$ which is a mild generalization of the quantum algebra $U_q(\\frak{sl}_2)$. The algebra $\\square_q$ is defined by generators and relations. The generators are $\\{x_i\\}_{i\\in \\mathbb{Z}_4}$, where $\\mathbb{Z}_4$ is the cyclic group of order $4$. For $i\\in \\mathbb{Z}_4$ the generators $x_i$,$x_{i+1}$ satisfy a $q$-Weyl relation, and $x_i$,$x_{i+2}$ satisfy a cubic $q$-Serre relation. For $i\\in \\mathbb{Z}_4$ we show that the action of $x_i$ is invertible on each nonzero finite-dimensional $\\square_q$-module. We view $x_i^{-1}$ as an operator that acts on nonzero finite-dimensional $\\square_q$-modules. For $i\\in \\mathbb{Z}_4$, define $\\mathfrak{n}_{i,i+1}=q(1-x_ix_{i+1})/(q-q^{-1})$. We show that the action of $\\mathfrak{n}_{i,i+1}$ is nilpotent on each nonzero finite-dimensional $\\square_q$-module. We view the $q$-exponential ${\\rm {exp}}_q(\\mathfrak{n}_{i,i+1})$ as an operator that acts on nonzero finite-dimensional $\\square_q$-modules. In our main results, for $i,j\\in \\mathbb{Z}_4$ we express each of of ${\\rm {exp}}_q(\\mathfrak{n}_{i,i+1})x_j{\\rm {exp}}_q(\\mathfrak{n}_{i,i+1})^{-1}$ and ${\\rm {exp}}_q(\\mathfrak{n}_{i,i+1})^{-1}x_j{\\rm {exp}}_q(\\mathfrak{n}_{i,i+1})$ as a polynomial in $\\{x_k^{\\pm 1}\\}_{k\\in \\mathbb{Z}_4}$."
                    },
                    {
                        "title": "Finite-dimensional irreducible $\\square_q$-modules and their Drinfel'd polynomials",
                        "abstract": "Let $\\mathbb{F}$ denote an algebraically closed field with characteristic $0$, and let $q$ denote a nonzero scalar in $\\mathbb{F}$ that is not a root of unity. Let $\\mathbb{Z}_4$ denote the cyclic group of order $4$. Let $\\square_q$ denote the unital associative $\\mathbb{F}$-algebra defined by generators $\\{x_i\\}_{i\\in \\mathbb{Z}_4}$ and relations \\begin{gather*} \\frac{qx_ix_{i+1}-q^{-1}x_{i+1}x_i}{q-q^{-1}}=1, \\\\ x_i^3x_{i+2}-[3]_qx_i^2x_{i+2}x_i+[3]_qx_ix_{i+2}x_i^2-x_{i+2}x_i^3=0, \\end{gather*} where $[3]_q=(q^3-q^{-3})/(q-q^{-1})$. There exists an automorphism $\\rho$ of $\\square_q$ that sends $x_i\\mapsto x_{i+1}$ for $i\\in \\mathbb{Z}_4$. Let $V$ denote a finite-dimensional irreducible $\\square_q$-module of type $1$. To $V$ we attach a polynomial called the Drinfel'd polynomial. In our main result, we explain how the following are related: (i) the Drinfel'd polynomial for the $\\square_q$-module $V$; (ii) the Drinfel'd polynomial for the $\\square_q$-module $V$ twisted via $\\rho$. Specifically, we show that the roots of (i) are the inverses of the roots of (ii). We discuss how $\\square_q$ is related to the quantum loop algebra $U_q(L(\\mathfrak{sl}_2))$, its positive part $U_q^+$, the $q$-tetrahedron algebra $\\boxtimes_q$, and the $q$-geometric tridiagonal pairs."
                    },
                    {
                        "title": "Holistic risk assessment of inference attacks in machine learning",
                        "abstract": "As machine learning expanding application, there are more and more unignorable privacy and safety issues. Especially inference attacks against Machine Learning models allow adversaries to infer sensitive information about the target model, such as training data, model parameters, etc. Inference attacks can lead to serious consequences, including violating individuals privacy, compromising the intellectual property of the owner of the machine learning model. As far as concerned, researchers have studied and analyzed in depth several types of inference attacks, albeit in isolation, but there is still a lack of a holistic rick assessment of inference attacks against machine learning models, such as their application in different scenarios, the common factors affecting the performance of these attacks and the relationship among the attacks. As a result, this paper performs a holistic risk assessment of different inference attacks against Machine Learning models. This paper focuses on three kinds of representative attacks: membership inference attack, attribute inference attack and model stealing attack. And a threat model taxonomy is established. A total of 12 target models using three model architectures, including AlexNet, ResNet18 and Simple CNN, are trained on four datasets, namely CelebA, UTKFace, STL10 and FMNIST."
                    },
                    {
                        "title": "An upper bound of the mutation probability in the genetic algorithm for general 0-1 knapsack problem",
                        "abstract": "As an important part of genetic algorithms (GAs), mutation operators is widely used in evolutionary algorithms to solve $\\mathcal{NP}$-hard problems because it can increase the population diversity of individual. Due to limitations in mathematical tools, the mutation probability of the mutation operator is primarily empirically set in practical applications.   In this paper, we propose a novel reduction method for the 0-1 knapsack problem(0-1 KP) and an improved mutation operator (IMO) based on the assumption $\\mathcal{NP}\\neq\\mathcal{P}$, along with the utilization of linear relaxation techniques and a recent result by Dey et al. (Math. Prog., pp 569-587, 2022). We employ this method to calculate an upper bound of the mutation probability in general instances of the 0-1 KP, and construct an instance where the mutation probability does not tend towards 0 as the problem size increases. Finally, we prove that the probability of the IMO hitting the optimal solution within only a single iteration in large-scale instances is superior to that of the traditional mutation operator."
                    },
                    {
                        "title": "Square-Root Higher-Order Unscented Estimators for Robust Orbit Determination",
                        "abstract": "Orbit determination (OD) is a fundamental problem in space surveillance and tracking, crucial for ensuring the safety of space assets. Real-world ground-based optical tracking scenarios often involve challenges such as limited measurement time, short visible arcs, and the presence of outliers, leading to sparse and non-Gaussian observational data. Additionally, the highly perturbative and nonlinear orbit dynamics of resident space objects (RSOs) in low Earth orbit (LEO) add further complexity to the OD problem.   This paper introduces a novel variant of the higher-order unscented Kalman estimator (HOUSE) called $w$-HOUSE, which employs a square-root formulation and addresses the challenges posed by nonlinear and non-Gaussian OD problems. The effectiveness of $w$-HOUSE was demonstrated through synthetic and real-world measurements, specifically outlier-contaminated angle-only measurements collected for the Sentinel 6A satellite flying in LEO. Comparative analyses are conducted with the original HOUSE (referred to as $\\delta$-HOUSE), unscented Kalman filters (UKF), conjugate unscented transformation (CUT) filters, and precise orbit determination solutions estimated via onboard global navigation satellite systems measurements.   The results reveal that the proposed $w$-HOUSE filter exhibits greater robustness when dealing with varying values of the dependent parameter compared to the original $\\delta$-HOUSE. Moreover, it surpasses all other filters in terms of positioning accuracy, achieving three-dimensional root-mean-square errors of less than 60 m in a three-day scenario. This research suggests that the new $w$-HOUSE filter represents a viable alternative to UKF and CUT filters, offering improved positioning performance in handling the nonlinear and non-Gaussian OD problem associated with LEO RSOs."
                    },
                    {
                        "title": "A constant time complexity algorithm for the unbounded knapsack problem with bounded coefficients",
                        "abstract": "Benchmark instances for the unbounded knapsack problem are typically generated according to specific criteria within a given constant range $R$, and these instances can be referred to as the unbounded knapsack problem with bounded coefficients (UKPB). In order to increase the difficulty of solving these instances, the knapsack capacity $C$ is usually set to a very large value. Therefore, an exact algorithm that neither time complexity nor space complexity includes the capacity coefficient $C$ is highly anticipated.   In this paper, we propose an exact algorithm with time complexity of $O(R^4)$ and space complexity of $O(R^3)$. The algorithm initially divides the multiset $N$ into two multisubsets, $N_1$ and $N_2$, based on the profit density of their types. For the multisubset $N_2$ composed of types with profit density lower than the maximum profit density type, we utilize a recent branch and bound (B\\&B) result by Dey et al. (Math. Prog., pp 569-587, 2023) to determine the maximum selection number for types in $N_2$. We then employ the Unbounded-DP algorithm to exactly solve for the types in $N_2$. For the multisubset $N_1$ composed of the maximum profit density type and its counterparts with the same profit density, we transform it into a linear Diophantine equation and leverage relevant conclusions from the Frobenius problem to solve it efficiently. In particular, the proof techniques required by the algorithm are primarily covered in the first-year mathematics curriculum, which is convenient for subsequent researchers to grasp."
                    },
                    {
                        "title": "Semi-Supervised Image Captioning Considering Wasserstein Graph Matching",
                        "abstract": "Image captioning can automatically generate captions for the given images, and the key challenge is to learn a mapping function from visual features to natural language features. Existing approaches are mostly supervised ones, i.e., each image has a corresponding sentence in the training set. However, considering that describing images always requires a huge of manpower, we usually have limited amount of described images (i.e., image-text pairs) and a large number of undescribed images in real-world applications. Thereby, a dilemma is the \"Semi-Supervised Image Captioning\". To solve this problem, we propose a novel Semi-Supervised Image Captioning method considering Wasserstein Graph Matching (SSIC-WGM), which turns to adopt the raw image inputs to supervise the generated sentences. Different from traditional single modal semi-supervised methods, the difficulty of semi-supervised cross-modal learning lies in constructing intermediately comparable information among heterogeneous modalities. In this paper, SSIC-WGM adopts the successful scene graphs as intermediate information, and constrains the generated sentences from two aspects: 1) inter-modal consistency. SSIC-WGM constructs the scene graphs of the raw image and generated sentence respectively, then employs the wasserstein distance to better measure the similarity between region embeddings of different graphs. 2) intra-modal consistency. SSIC-WGM takes the data augmentation techniques for the raw images, then constrains the consistency among augmented images and generated sentences. Consequently, SSIC-WGM combines the cross-modal pseudo supervision and structure invariant measure for efficiently using the undescribed images, and learns more reasonable mapping function."
                    }
                ]
            }
        }
    },
    "2402.02425": {
        "paper_data": {
            "title": "DeepLag: Discovering Deep Lagrangian Dynamics for Intuitive Fluid Prediction",
            "url": "http://arxiv.org/abs/2402.02425v3",
            "arxiv_id": "2402.02425",
            "authors": [
                "Qilong Ma",
                "Haixu Wu",
                "Lanxiang Xing",
                "Shangchen Miao",
                "Mingsheng Long"
            ],
            "abstract": "Accurately predicting the future fluid is vital to extensive areas such as meteorology, oceanology, and aerodynamics. However, since the fluid is usually observed from the Eulerian perspective, its moving and intricate dynamics are seriously obscured and confounded in static grids, bringing thorny challenges to the prediction. This paper introduces a new Lagrangian-Eulerian combined paradigm to tackle the tanglesome fluid dynamics. Instead of solely predicting the future based on Eulerian observations, we propose DeepLag to discover hidden Lagrangian dynamics within the fluid by tracking the movements of adaptively sampled key particles. Further, DeepLag presents a new paradigm for fluid prediction, where the Lagrangian movement of the tracked particles is inferred from Eulerian observations, and their accumulated Lagrangian dynamics information is incorporated into global Eulerian evolving features to guide future prediction respectively. Tracking key particles not only provides a transparent and interpretable clue for fluid dynamics but also makes our model free from modeling complex correlations among massive grids for better efficiency. Experimentally, DeepLag excels in three challenging fluid prediction tasks covering 2D and 3D, simulated and real-world fluids.",
            "introduction": "   1 Introduction  Fluids, characterized by a molecular structure devoid of resistance to external shear forces, readily undergo deformation even under minimal force, thereby typically presenting extremely complex dynamics [10]. Consequently, the solvability of theorems for fluid mechanics, such as Navier-Stokes equations, is confined to a limited subset of flows due to the complex form and intricate multiphysics interactions [38]. In practical applications, computational fluid dynamics (CFD) is widely used to infer future fluid by numerical simulations. However, it suffers from substantial computation costs due to the curse of dimensionality [40]. Accurately predicting future dynamics is still a challenging problem. Recently, benefiting from the tremendous non-linear modeling capability, deep models\u00a0[8, 28, 22] have emerged as a promising tool for fluid prediction, which can be trained with CFD simulations or real-world observations and perform as a fast surrogate model during inference.    Figure 1: Comparison between Lagrangian and Eulerian perspectives. The left part presents the trajectory of five representative particles learned by DeepLag on the Ocean Current dataset, which provides a Lagrangian perspective for Eulerian-recorded fluid on the right.   A booming direction for deep fluid prediction is learning deep models to solve partial differential equations (PDEs) [45]. However, most of these methods [48, 29, 22, 21] attempt to capture fluid dynamics from the Eulerian perspective, which means modeling spatiotemporal correlations among massive grids unchanging over time. From this perspective, the complicated moving dynamics in fluids could be seriously obscured and confounded in static grids, bringing challenges in both computational efficiency and learning difficulties for accurately predicting future fluids.   In parallel to the Eulerian method, we notice another major approach for elucidating fluid dynamics, the Lagrangian method [13], also known as the tracking method. This method primarily focuses on tracing individual fluid particles by conducting the temporal evolution of the position and velocity of each particle. Unlike the Eulerian methods, the Lagrangian approach describes the fluid dynamics through the moving trajectory of individual fluid particles, offering a more natural and neat representation of fluid dynamics with inherent advantages in capturing intricate flow dynamics. Moreover, it helps get around the Courant\u2013Friedrichs\u2013Lewy condition [6] for advection, allowing a larger inference step while retaining stability. In Figure 1, we can find that fluid dynamics is much more visually apparent in Lagrangian trajectories than density changes on static Eulerian grids.   Building on the two perspectives mentioned earlier, we propose DeepLag as an Eulerian-Lagrangian Recurrent Network. Our aim is to integrate Lagrangian tracking into the deep model, thereby enhancing the dynamics modeling in Eulerian fluid prediction. To achieve this, we present the EuLag Block, a powerful module that accomplishes Lagrangian tracking and Eulerian predicting at various scales. By leveraging the cross-attention mechanism, the EuLag Block assimilates tracked Lagrangian particle dynamics into the Eulerian field, guiding fluid prediction. It also forecasts the trajectory and dynamics of Lagrangian particles with the aid of Eulerian features. This unique Eulerian-Lagrangian design harnesses the dynamics information captured by Lagrangian trajectories and the fluid-structure features learned in the Eulerian grid. In our experiments, DeepLag consistently outperforms existing models, demonstrating state-of-the-art performance across three representative datasets, covering 2D and 3D fluid at various scales. Our contributions",
            "references": [
                {
                    "title": "Scalable Transformer for PDE Surrogate Modeling",
                    "abstract": "Transformer has shown state-of-the-art performance on various applications and has recently emerged as a promising tool for surrogate modeling of partial differential equations (PDEs). Despite the introduction of linear-complexity variant, applying attention to a large number of grid points can result in instability and is still expensive to compute. In this work, we propose Factorized Transformer(FactFormer), which is based on an axial factorized kernel integral. Concretely, we introduce a learnable projection operator that decomposes the input function into multiple sub-functions with one-dimensional domain. These sub-functions are then evaluated and used to compute the instance-based kernel with an axial factorized scheme. We showcase that the proposed model is able to simulate 2D Kolmogorov flow on a 256 by 256 grid and 3D smoke buoyancy on a 64 by 64 by 64 grid with good accuracy and efficiency. In addition, we find out that with the factorization scheme, the attention matrices enjoy a more compact spectrum than full softmax-free attention matrices."
                },
                {
                    "title": "GNOT: A General Neural Operator Transformer for Operator Learning",
                    "abstract": "Learning partial differential equations' (PDEs) solution operators is an essential problem in machine learning. However, there are several challenges for learning operators in practical applications like the irregular mesh, multiple input functions, and complexity of the PDEs' solution. To address these challenges, we propose a general neural operator transformer (GNOT), a scalable and effective transformer-based framework for learning operators. By designing a novel heterogeneous normalized attention layer, our model is highly flexible to handle multiple input functions and irregular meshes. Besides, we introduce a geometric gating mechanism which could be viewed as a soft domain decomposition to solve the multi-scale problems. The large model capacity of the transformer architecture grants our model the possibility to scale to large datasets and practical problems. We conduct extensive experiments on multiple challenging datasets from different domains and achieve a remarkable improvement compared with alternative methods. Our code and data are publicly available at \\url{https://github.com/thu-ml/GNOT}."
                },
                {
                    "title": "Convolutional Neural Operators for robust and accurate learning of PDEs",
                    "abstract": "Although very successfully used in conventional machine learning, convolution based neural network architectures -- believed to be inconsistent in function space -- have been largely ignored in the context of learning solution operators of PDEs. Here, we present novel adaptations for convolutional neural networks to demonstrate that they are indeed able to process functions as inputs and outputs. The resulting architecture, termed as convolutional neural operators (CNOs), is designed specifically to preserve its underlying continuous nature, even when implemented in a discretized form on a computer. We prove a universality theorem to show that CNOs can approximate operators arising in PDEs to desired accuracy. CNOs are tested on a novel suite of benchmarks, encompassing a diverse set of PDEs with possibly multi-scale solutions and are observed to significantly outperform baselines, paving the way for an alternative framework for robust and accurate operator learning. Our code is publicly available at https://github.com/bogdanraonic3/ConvolutionalNeuralOperator"
                },
                {
                    "title": "Solving High-Dimensional PDEs with Latent Spectral Models",
                    "abstract": "Deep models have achieved impressive progress in solving partial differential equations (PDEs). A burgeoning paradigm is learning neural operators to approximate the input-output mappings of PDEs. While previous deep models have explored the multiscale architectures and various operator designs, they are limited to learning the operators as a whole in the coordinate space. In real physical science problems, PDEs are complex coupled equations with numerical solvers relying on discretization into high-dimensional coordinate space, which cannot be precisely approximated by a single operator nor efficiently learned due to the curse of dimensionality. We present Latent Spectral Models (LSM) toward an efficient and precise solver for high-dimensional PDEs. Going beyond the coordinate space, LSM enables an attention-based hierarchical projection network to reduce the high-dimensional data into a compact latent space in linear time. Inspired by classical spectral methods in numerical analysis, we design a neural spectral block to solve PDEs in the latent space that approximates complex input-output mappings via learning multiple basis operators, enjoying nice theoretical guarantees for convergence and approximation. Experimentally, LSM achieves consistent state-of-the-art and yields a relative gain of 11.5% averaged on seven benchmarks covering both solid and fluid physics. Code is available at https://github.com/thuml/Latent-Spectral-Models."
                },
                {
                    "title": "Learning Vortex Dynamics for Fluid Inference and Prediction",
                    "abstract": "We propose a novel differentiable vortex particle (DVP) method to infer and predict fluid dynamics from a single video. Lying at its core is a particle-based latent space to encapsulate the hidden, Lagrangian vortical evolution underpinning the observable, Eulerian flow phenomena. Our differentiable vortex particles are coupled with a learnable, vortex-to-velocity dynamics mapping to effectively capture the complex flow features in a physically-constrained, low-dimensional space. This representation facilitates the learning of a fluid simulator tailored to the input video that can deliver robust, long-term future predictions. The value of our method is twofold: first, our learned simulator enables the inference of hidden physics quantities (e.g., velocity field) purely from visual observation; secondly, it also supports future prediction, constructing the input video's sequel along with its future dynamics evolution. We compare our method with a range of existing methods on both synthetic and real-world videos, demonstrating improved reconstruction quality, visual plausibility, and physical integrity."
                },
                {
                    "title": "U-NO: U-shaped Neural Operators",
                    "abstract": "Neural operators generalize classical neural networks to maps between infinite-dimensional spaces, e.g., function spaces. Prior works on neural operators proposed a series of novel methods to learn such maps and demonstrated unprecedented success in learning solution operators of partial differential equations. Due to their close proximity to fully connected architectures, these models mainly suffer from high memory usage and are generally limited to shallow deep learning models. In this paper, we propose U-shaped Neural Operator (U-NO), a U-shaped memory enhanced architecture that allows for deeper neural operators. U-NOs exploit the problem structures in function predictions and demonstrate fast training, data efficiency, and robustness with respect to hyperparameters choices. We study the performance of U-NO on PDE benchmarks, namely, Darcy's flow law and the Navier-Stokes equations. We show that U-NO results in an average of 26% and 44% prediction improvement on Darcy's flow and turbulent Navier-Stokes equations, respectively, over the state of the art. On Navier-Stokes 3D spatiotemporal operator learning task, we show U-NO provides 37% improvement over the state of art methods."
                },
                {
                    "title": "Message Passing Neural PDE Solvers",
                    "abstract": "The numerical solution of partial differential equations (PDEs) is difficult, having led to a century of research so far. Recently, there have been pushes to build neural--numerical hybrid solvers, which piggy-backs the modern trend towards fully end-to-end learned systems. Most works so far can only generalize over a subset of properties to which a generic solver would be faced, including: resolution, topology, geometry, boundary conditions, domain discretization regularity, dimensionality, etc. In this work, we build a solver, satisfying these properties, where all the components are based on neural message passing, replacing all heuristically designed components in the computation graph with backprop-optimized neural function approximators. We show that neural message passing solvers representationally contain some classical methods, such as finite differences, finite volumes, and WENO schemes. In order to encourage stability in training autoregressive models, we put forward a method that is based on the principle of zero-stability, posing stability as a domain adaptation problem. We validate our method on various fluid-like flow problems, demonstrating fast, stable, and accurate performance across different domain topologies, equation parameters, discretizations, etc., in 1D and 2D."
                },
                {
                    "title": "Comparison and Analysis of Neural Solver Methods for Differential Equations in Physical Systems",
                    "abstract": "Differential equations are ubiquitous in many fields of study, yet not all equations, whether ordinary or partial, can be solved analytically. Traditional numerical methods such as time-stepping schemes have been devised to approximate these solutions. With the advent of modern deep learning, neural networks have become a viable alternative to traditional numerical methods. By reformulating the problem as an optimisation task, neural networks can be trained in a semi-supervised learning fashion to approximate nonlinear solutions. In this paper, neural solvers are implemented in TensorFlow for a variety of differential equations, namely: linear and nonlinear ordinary differential equations of the first and second order; Poisson\u2019s equation, the heat equation, and the inviscid Burgers\u2019 equation. Different methods, such as the naive and ansatz formulations, are contrasted, and their overall performance is analysed. Experimental data is also used to validate the neural solutions on test cases, specifically: the spring-mass system and Gauss\u2019s law for electric fields. The errors of the neural solvers against exact solutions are investigated and found to surpass traditional schemes in certain cases. Although neural solvers will not replace the computational speed offered by traditional schemes in the near future, they remain a feasible, easy-to-implement substitute when all else fails."
                },
                {
                    "title": "Spline-PINN: Approaching PDEs without Data using Fast, Physics-Informed Hermite-Spline CNNs",
                    "abstract": "Partial Differential Equations (PDEs) are notoriously difficult to solve. In general, closed form solutions are not available and numerical approximation schemes are computationally expensive. In this paper, we propose to approach the solution of PDEs based on a novel technique that combines the advantages of two recently emerging machine learning based approaches.\n First, physics-informed neural networks (PINNs) learn continuous solutions of PDEs and can be trained with little to no ground truth data. However, PINNs do not generalize well to unseen domains. Second, convolutional neural networks provide fast inference and generalize but either require large amounts of training data or a physics-constrained loss based on finite differences that can lead to inaccuracies and discretization artifacts.\n We leverage the advantages of both of these approaches by using Hermite spline kernels in order to continuously interpolate a grid-based state representation that can be handled by a CNN. This allows for training without any precomputed training data using a physics-informed loss function only and provides fast, continuous solutions that generalize to unseen domains.\n We demonstrate the potential of our method at the examples of the incompressible Navier-Stokes equation and the damped wave equation. Our models are able to learn several intriguing phenomena such as Karman vortex streets, the Magnus effect, Doppler effect, interference patterns and wave reflections. Our quantitative assessment and an interactive real-time demo show that we are narrowing the gap in accuracy of unsupervised ML based methods to industrial solvers for computational fluid dynamics (CFD) while being orders of magnitude faster."
                },
                {
                    "title": "Fluid mechanics",
                    "abstract": "The author discusses some basic questions in fluid dynamics. He describes Newton's contribution to fluid dynamics in the second volume of the Principia. He presents a brief review of fluid dynamics since Newton, and discusses the nature and content of physics curricula in schools and universities. He also outlines some aspects of modern research."
                },
                {
                    "title": "Choose a Transformer: Fourier or Galerkin",
                    "abstract": "In this paper, we apply the self-attention from the state-of-the-art Transformer in Attention Is All You Need for the first time to a data-driven operator learning problem related to partial differential equations. An effort is put together to explain the heuristics of, and to improve the efficacy of the attention mechanism. By employing the operator approximation theory in Hilbert spaces, it is demonstrated for the first time that the softmax normalization in the scaled dot-product attention is sufficient but not necessary. Without softmax, the approximation capacity of a linearized Transformer variant can be proved to be comparable to a Petrov-Galerkin projection layer-wise, and the estimate is independent with respect to the sequence length. A new layer normalization scheme mimicking the Petrov-Galerkin projection is proposed to allow a scaling to propagate through attention layers, which helps the model achieve remarkable accuracy in operator learning tasks with unnormalized data. Finally, we present three operator learning experiments, including the viscid Burgers' equation, an interface Darcy flow, and an inverse interface coefficient identification problem. The newly proposed simple attention-based operator learner, Galerkin Transformer, shows significant improvements in both training cost and evaluation accuracy over its softmax-normalized counterparts."
                },
                {
                    "title": "Nystr\u00f6mformer: A Nystr\u00f6m-Based Algorithm for Approximating Self-Attention",
                    "abstract": "Transformers have emerged as a powerful tool for a broad range of natural language processing tasks. A key component that drives the impressive performance of Transformers is the self-attention mechanism that encodes the influence or dependence of other tokens on each specific token. While beneficial, the quadratic complexity of self-attention on the input sequence length has limited its application to longer sequences - a topic being actively studied in the community. To address this limitation, we propose Nystr\u00f6mformer - a model that exhibits favorable scalability as a function of sequence length. Our idea is based on adapting the Nystr\u00f6m method to approximate standard self-attention with O(n) complexity. The scalability of Nystr\u00f6mformer enables application to longer sequences with thousands of tokens. We perform evaluations on multiple downstream tasks on the GLUE benchmark and IMDB reviews with standard sequence length, and find that our Nystr\u00f6mformer performs comparably, or in a few cases, even slightly better, than standard self-attention. On longer sequence tasks in the Long Range Arena (LRA) benchmark, Nystr\u00f6mformer performs favorably relative to other efficient self-attention methods. Our code is available at https://github.com/mlpen/Nystromformer."
                },
                {
                    "title": "Fourier Neural Operator for Parametric Partial Differential Equations",
                    "abstract": "The classical development of neural networks has primarily focused on learning mappings between finite-dimensional Euclidean spaces. Recently, this has been generalized to neural operators that learn mappings between function spaces. For partial differential equations (PDEs), neural operators directly learn the mapping from any functional parametric dependence to the solution. Thus, they learn an entire family of PDEs, in contrast to classical methods which solve one instance of the equation. In this work, we formulate a new neural operator by parameterizing the integral kernel directly in Fourier space, allowing for an expressive and efficient architecture. We perform experiments on Burgers' equation, Darcy flow, and the Navier-Stokes equation (including the turbulent regime). Our Fourier neural operator shows state-of-the-art performance compared to existing neural network methodologies and it is up to three orders of magnitude faster compared to traditional PDE solvers."
                },
                {
                    "title": "Learning Mesh-Based Simulation with Graph Networks",
                    "abstract": "Mesh-based simulations are central to modeling complex physical systems in many disciplines across science and engineering. Mesh representations support powerful numerical integration methods and their resolution can be adapted to strike favorable trade-offs between accuracy and efficiency. However, high-dimensional scientific simulations are very expensive to run, and solvers and parameters must often be tuned individually to each system studied. Here we introduce MeshGraphNets, a framework for learning mesh-based simulations using graph neural networks. Our model can be trained to pass messages on a mesh graph and to adapt the mesh discretization during forward simulation. Our results show it can accurately predict the dynamics of a wide range of physical systems, including aerodynamics, structural mechanics, and cloth. The model's adaptivity supports learning resolution-independent dynamics and can scale to more complex state spaces at test time. Our method is also highly efficient, running 1-2 orders of magnitude faster than the simulation on which it is trained. Our approach broadens the range of problems on which neural network simulators can operate and promises to improve the efficiency of complex, scientific modeling tasks."
                },
                {
                    "title": "Lagrangian Fluid Simulation with Continuous Convolutions",
                    "abstract": "We present an approach to Lagrangian fluid simulation with a new type of convolutional network. Our networks process sets of moving particles, which describe fluids in space and time. Unlike previous approaches, we do not build an explicit graph structure to connect the particles but use spatial convolutions as the main differentiable operation that relates particles to their neighbors. To this end we present a simple, novel, and effective extension of N-D convolutions to the continuous domain. We show that our network architecture can simulate different materials, generalizes to arbitrary collision geometries, and can be used for inverse problems. In addition, we demonstrate that our continuous convolutions outperform prior formulations in terms of accuracy and speed."
                },
                {
                    "title": "Learning to Simulate Complex Physics with Graph Networks",
                    "abstract": "Here we present a machine learning framework and model implementation that can learn to simulate a wide variety of challenging physical domains, involving fluids, rigid solids, and deformable materials interacting with one another. Our framework---which we term \"Graph Network-based Simulators\" (GNS)---represents the state of a physical system with particles, expressed as nodes in a graph, and computes dynamics via learned message-passing. Our results show that our model can generalize from single-timestep predictions with thousands of particles during training, to different initial conditions, thousands of timesteps, and at least an order of magnitude more particles at test time. Our model was robust to hyperparameter choices across various evaluation metrics: the main determinants of long-term performance were the number of message-passing steps, and mitigating the accumulation of error by corrupting the training data with noise. Our GNS framework advances the state-of-the-art in learned physical simulation, and holds promise for solving a wide range of complex forward and inverse problems."
                },
                {
                    "title": "On Layer Normalization in the Transformer Architecture",
                    "abstract": "The Transformer is widely used in natural language processing tasks. To train a Transformer however, one usually needs a carefully designed learning rate warm-up stage, which is shown to be crucial to the final performance but will slow down the optimization and bring more hyperparameter tunings. In this paper, we first study theoretically why the learning rate warm-up stage is essential and show that the location of layer normalization matters. Specifically, we prove with mean field theory that at initialization, for the original-designed Post-LN Transformer, which places the layer normalization between the residual blocks, the expected gradients of the parameters near the output layer are large. Therefore, using a large learning rate on those gradients makes the training unstable. The warm-up stage is practically helpful for avoiding this problem. On the other hand, our theory also shows that if the layer normalization is put inside the residual blocks (recently proposed as Pre-LN Transformer), the gradients are well-behaved at initialization. This motivates us to remove the warm-up stage for the training of Pre-LN Transformers. We show in our experiments that Pre-LN Transformers without the warm-up stage can reach comparable results with baselines while requiring significantly less training time and hyper-parameter tuning on a wide range of applications."
                },
                {
                    "title": "Understanding and mitigating gradient pathologies in physics-informed neural networks",
                    "abstract": "The widespread use of neural networks across different scientific domains often involves constraining them to satisfy certain symmetries, conservation laws, or other domain knowledge. Such constraints are often imposed as soft penalties during model training and effectively act as domain-specific regularizers of the empirical risk loss. Physics-informed neural networks is an example of this philosophy in which the outputs of deep neural networks are constrained to approximately satisfy a given set of partial differential equations. In this work we review recent advances in scientific machine learning with a specific focus on the effectiveness of physics-informed neural networks in predicting outcomes of physical systems and discovering hidden physics from noisy data. We will also identify and analyze a fundamental mode of failure of such approaches that is related to numerical stiffness leading to unbalanced back-propagated gradients during model training. To address this limitation we present a learning rate annealing algorithm that utilizes gradient statistics during model training to balance the interplay between different terms in composite loss functions. We also propose a novel neural network architecture that is more resilient to such gradient pathologies. Taken together, our developments provide new insights into the training of constrained neural networks and consistently improve the predictive accuracy of physics-informed neural networks by a factor of 50-100x across a range of problems in computational physics. All code and data accompanying this manuscript are publicly available at \\url{this https URL}."
                },
                {
                    "title": "Reformer: The Efficient Transformer",
                    "abstract": "Large Transformer models routinely achieve state-of-the-art results on a number of tasks but training these models can be prohibitively costly, especially on long sequences. We introduce two techniques to improve the efficiency of Transformers. For one, we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O($L^2$) to O($L\\log L$), where $L$ is the length of the sequence. Furthermore, we use reversible residual layers instead of the standard residuals, which allows storing activations only once in the training process instead of $N$ times, where $N$ is the number of layers. The resulting model, the Reformer, performs on par with Transformer models while being much more memory-efficient and much faster on long sequences."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Turbulence forecasting via Neural ODE.",
                    "abstract": "Fluid turbulence is characterized by strong coupling across a broad range of scales. Furthermore, besides the usual local cascades, such coupling may extend to interactions that are non-local in scale-space. As such the computational demands associated with explicitly resolving the full set of scales and their interactions, as in the Direct Numerical Simulation (DNS) of the Navier-Stokes equations, in most problems of practical interest are so high that reduced modeling of scales and interactions is required before further progress can be made. While popular reduced models are typically based on phenomenological modeling of relevant turbulent processes, recent advances in machine learning techniques have energized efforts to further improve the accuracy of such reduced models. In contrast to such efforts that seek to improve an existing turbulence model, we propose a machine learning(ML) methodology that captures, de novo, underlying turbulence phenomenology without a pre-specified model form. To illustrate the approach, we consider transient modeling of the dissipation of turbulent kinetic energy, a fundamental turbulent process that is central to a wide range of turbulence models using a Neural ODE approach. After presenting details of the methodology, we show that this approach outperforms state-of-the-art approaches."
                },
                {
                    "title": "Learning three-dimensional flow for interactive aerodynamic design",
                    "abstract": "We present a data-driven technique to instantly predict how fluid flows around various three-dimensional objects. Such simulation is useful for computational fabrication and engineering, but is usually computationally expensive since it requires solving the Navier-Stokes equation for many time steps. To accelerate the process, we propose a machine learning framework which predicts aerodynamic forces and velocity and pressure fields given a three-dimensional shape input. Handling detailed free-form three-dimensional shapes in a data-driven framework is challenging because machine learning approaches usually require a consistent parametrization of input and output. We present a novel PolyCube maps-based parametrization that can be computed for three-dimensional shapes at interactive rates. This allows us to efficiently learn the nonlinear response of the flow using a Gaussian process regression. We demonstrate the effectiveness of our approach for the interactive design and optimization of a car body."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "SPNets: Differentiable Fluid Dynamics for Deep Neural Networks",
                    "abstract": "In this paper we introduce Smooth Particle Networks (SPNets), a framework for integrating fluid dynamics with deep networks. SPNets adds two new layers to the neural network toolbox: ConvSP and ConvSDF, which enable computing physical interactions with unordered particle sets. We use these lay- ers in combination with standard neural network layers to directly implement fluid dynamics inside a deep network, where the parameters of the network are the fluid parameters themselves (e.g., viscosity, cohesion, etc.). Because SPNets are imple- mented as a neural network, the resulting fluid dynamics are fully differentiable. We then show how this can be successfully used to learn fluid parameters from data, perform liquid control tasks, and learn policies to manipulate liquids."
                },
                {
                    "title": "Physics Informed Deep Learning (Part I): Data-driven Solutions of Nonlinear Partial Differential Equations",
                    "abstract": "We introduce physics informed neural networks -- neural networks that are trained to solve supervised learning tasks while respecting any given law of physics described by general nonlinear partial differential equations. In this two part treatise, we present our developments in the context of solving two main classes of problems: data-driven solution and data-driven discovery of partial differential equations. Depending on the nature and arrangement of the available data, we devise two distinct classes of algorithms, namely continuous time and discrete time models. The resulting neural networks form a new class of data-efficient universal function approximators that naturally encode any underlying physical laws as prior information. In this first part, we demonstrate how these networks can be used to infer solutions to partial differential equations, and obtain physics-informed surrogate models that are fully differentiable with respect to all input coordinates and free parameters."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Gaussian Error Linear Units (GELUs)",
                    "abstract": "We propose the Gaussian Error Linear Unit (GELU), a high-performing neural network activation function. The GELU activation function is $x\\Phi(x)$, where $\\Phi(x)$ the standard Gaussian cumulative distribution function. The GELU nonlinearity weights inputs by their value, rather than gates inputs by their sign as in ReLUs ($x\\mathbf{1}_{x>0}$). We perform an empirical evaluation of the GELU nonlinearity against the ReLU and ELU activations and find performance improvements across all considered computer vision, natural language processing, and speech tasks."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "On the Partial Difference Equations, of Mathematical Physics",
                    "abstract": "Problems involving the classical linear partial differential equations of mathematical physics can be reduced to algebraic ones of a very much simpler structure by replacing the differentials by difference quotients on some (say rectilinear) mesh. This paper will undertake an elementary discussion of these algebraic problems, in particular of the behavior of the solution as the mesh width tends to zero. For present purposes we limit ourselves mainly to simple but typical cases, and treat them in such a way that the applicability of the method to more general difference equations and to those with arbitrarily many independent variables is made clear."
                },
                {
                    "title": "Neural-network-based approximations for solving partial differential equations",
                    "abstract": "A numerical method, based on neural-network-based functions, for solving partial differential equations is reported in the paper. Using a \u2018universal approximator\u2019 based on a neural network and point collocation, the numerical problem of solving the partial differential equation is transformed to an unconstrained minimization problem. The method is extremely easy to implement and is suitable for obtaining an approximate solution in a short period of time. The technique is illustrated with the aid of two numerical examples."
                },
                {
                    "title": "A stable numerical integration scheme for the primitive meteorological equations",
                    "abstract": "Abstract A stable numerical integration scheme is applied to the non\u2010divergent barotropic vorticity equation. Integrations are performed with time steps ranging from 15 min to 4 h. The root\u2010mean\u2010square differences between the forecasts are calculated in order to measure the sensitivity of the predictions to the size of the time step. These experiments show that the truncation errors remain reasonably small with time steps as large as two hours. This scheme is then associated with the semi\u2010implicit algorithm in the integration of the shallow water equations. Integrations are carried out with large time steps and the resulting predictions are presented in order to demonstrate that the scheme is perfectly stable."
                },
                {
                    "title": "A Problem of Dimensionality: A Simple Example",
                    "abstract": "In pattern recognition problems it has been noted that beyond a certain point the inclusion of additional parameters (that have been estimated) leads to higher probabilities of error. A simple problem has been formulated where the probability of error approaches zero as the dimensionality increases and all the parameters are known; on the other hand, the probability of error approaches one-half as the dimensionality increases and parameters are estimated."
                },
                {
                    "title": "An Introduction to Fluid Dynamics",
                    "abstract": "Keywords: dynamique des : fluides Reference Record created on 2005-11-18, modified on 2016-08-08"
                },
                {
                    "title": "Computational methods for fluid dynamics",
                    "abstract": "This text develops and applies the techniques used to solve problems in fluid mechanics on computers and describes in detail those most often used in practice. It includes advanced techniques in computational fluid dynamics, such as direct and large-eddy simulation of turbulence, multigrid methods, parallel computing, moving grids, structured, block-structured and unstructured boundary-fitted grids, and free surface flows."
                },
                {
                    "title": "Classical mechanics of particles and rigid bodies",
                    "abstract": "The Lagrangian Formalism Space-Time Symmetries and a Lagrangian Formulation of the Conservation Laws The Hamiltonian Formalism The Integration of Equations of Motion Decay and Scattering Processes Small Oscillations The Kinematics and Dynamics of Rigid Body Motion The Canonical Transformation Theory The Hamilton-Jacobi Theory The Special Theory of Relativity Relativistic Classical Mechanics Continuous Systems and Classical Fields Nonlinear Oscillatory Systems."
                },
                {
                    "title": "Navier-Stokes Equations: Theory and Numerical Analysis",
                    "abstract": "I. The Steady-State Stokes Equations . 1. Some Function Spaces. 2. Existence and Uniqueness for the Stokes Equations. 3. Discretization of the Stokes Equations (I). 4. Discretization of the Stokes Equations (II). 5. Numerical Algorithms. 6. The Penalty Method. II. The Steady-State Navier-Stokes Equations . 1. Existence and Uniqueness Theorems. 2. Discrete Inequalities and Compactness Theorems. 3. Approximation of the Stationary Navier-Stokes Equations. 4. Bifurcation Theory and Non-Uniqueness Results. III. The Evolution Navier-Stokes Equations . 1. The Linear Case. 2. Compactness Theorems. 3. Existence and Uniqueness Theorems. (n < 4). 4. Alternate Proof of Existence by Semi-Discretization. 5. Discretization of the Navier-Stokes Equations: General Stability and Convergence Theorems. 6. Discretization of the Navier-Stokes Equations: Application of the General Results. 7. Approximation of the Navier-Stokes Equations by the Projection Method. 8. Approximation of the Navier-Stokes Equations by the Artificial Compressibility Method. Appendix I: Properties of the Curl Operator and Application to the Steady-State Navier-Stokes Equations. Appendix II. (by F. Thomasset): Implementation of Non-Conforming Linear Finite Elements. Comments."
                }
            ],
            "categories": [
                "cs.LG",
                "physics.flu-dyn"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively integrate Lagrangian tracking into deep learning models to enhance fluid dynamics prediction in computational fluid dynamics (CFD)?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of existing Eulerian methods in fluid dynamics, which often obscure complex moving dynamics. By improving fluid prediction through the integration of Lagrangian tracking, this research could lead to more accurate and efficient simulations in CFD, impacting various fields such as meteorology, oceanography, and engineering. The advancements could pave the way for future research into hybrid modeling techniques and practical applications in real-time fluid dynamics analysis and control systems.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of fluid dynamics, which involves intricate multiphysics interactions and non-linear behaviors that are difficult to capture using static Eulerian grids. Naive approaches may fail because they do not account for the dynamic nature of fluid particles, leading to inaccuracies in predictions. Additionally, the integration of Lagrangian tracking with Eulerian methods requires overcoming technical obstacles related to data representation, computational efficiency, and the effective assimilation of information from both perspectives.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on Eulerian methods, which have limitations in capturing the dynamic behavior of fluids due to their reliance on static grids. Existing solutions have not effectively combined Lagrangian tracking with deep learning models, resulting in a gap in methodologies that leverage both perspectives. Barriers such as the complexity of integrating different modeling approaches and the lack of suitable frameworks for real-time applications have hindered progress. Our approach differs by introducing the EuLag Block, which facilitates the integration of Lagrangian dynamics into Eulerian predictions, thus improving upon prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of DeepLag, an Eulerian-Lagrangian Recurrent Network that incorporates the EuLag Block for Lagrangian tracking and Eulerian prediction. We will utilize datasets representing 2D and 3D fluid dynamics across various scales, and performance will be evaluated using metrics such as prediction accuracy and computational efficiency. The expected outcomes include improved fluid dynamics predictions that outperform existing models, demonstrating the effectiveness of our hybrid approach in capturing complex fluid behaviors."
            }
        },
        "author_data": {
            "56b5a201-a5ba-4c9c-80d6-61d7b2ea2b5c": {
                "pk": "56b5a201-a5ba-4c9c-80d6-61d7b2ea2b5c",
                "name": "Qilong Ma",
                "collaborators": [
                    "Zheng Li",
                    "Dawei Zhu",
                    "Weimin Xiong",
                    "Sujian Li"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Personality Detection",
                    "Emotion Regulation"
                ],
                "publications": [
                    {
                        "title": "EERPD: Leveraging Emotion and Emotion Regulation for Improving Personality Detection",
                        "abstract": "Personality is a fundamental construct in psychology, reflecting an individual's behavior, thinking, and emotional patterns. Previous researches have made some progress in personality detection, primarily by utilizing the whole text to predict personality. However, these studies generally tend to overlook psychological knowledge: they rarely apply the well-established correlations between emotion regulation and personality. Based on this, we propose a new personality detection method called EERPD. This method introduces the use of emotion regulation, a psychological concept highly correlated with personality, for personality prediction. By combining this feature with emotion features, it retrieves few-shot examples and provides process CoTs for inferring labels from text. This approach enhances the understanding of LLM for personality within text and improves the performance in personality detection. Experimental results demonstrate that EERPD significantly enhances the accuracy and robustness of personality detection, outperforming previous SOTA by 15.05/4.29 in average F1 on the two benchmark datasets."
                    }
                ]
            },
            "33a3950d-e3a5-4716-9f02-2b5f46c150f0": {
                "pk": "33a3950d-e3a5-4716-9f02-2b5f46c150f0",
                "name": "Haixu Wu",
                "collaborators": [
                    "Mingsheng Long",
                    "Jianmin Wang",
                    "Yong Liu",
                    "Jiehui Xu",
                    "Huakun Luo",
                    "Yuezhou Ma",
                    "Tengge Hu",
                    "Jiaxiang Dong",
                    "Zhiyu Yao",
                    "Jialong Wu"
                ],
                "domain": [
                    "Time Series Analysis",
                    "Deep Learning",
                    "Transformers",
                    "Anomaly Detection"
                ],
                "publications": [
                    {
                        "title": "Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy",
                        "abstract": "Unsupervised detection of anomaly points in time series is a challenging problem, which requires the model to derive a distinguishable criterion. Previous methods tackle the problem mainly through learning pointwise representation or pairwise association, however, neither is sufficient to reason about the intricate dynamics. Recently, Transformers have shown great power in unified modeling of pointwise representation and pairwise association, and we find that the self-attention weight distribution of each time point can embody rich association with the whole series. Our key observation is that due to the rarity of anomalies, it is extremely difficult to build nontrivial associations from abnormal points to the whole series, thereby, the anomalies' associations shall mainly concentrate on their adjacent time points. This adjacent-concentration bias implies an association-based criterion inherently distinguishable between normal and abnormal points, which we highlight through the \\emph{Association Discrepancy}. Technically, we propose the \\emph{Anomaly Transformer} with a new \\emph{Anomaly-Attention} mechanism to compute the association discrepancy. A minimax strategy is devised to amplify the normal-abnormal distinguishability of the association discrepancy. The Anomaly Transformer achieves state-of-the-art results on six unsupervised time series anomaly detection benchmarks of three applications: service monitoring, space & earth exploration, and water treatment."
                    },
                    {
                        "title": "Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting",
                        "abstract": "Extending the forecasting time is a critical demand for real applications, such as extreme weather early warning and long-term energy consumption planning. This paper studies the long-term forecasting problem of time series. Prior Transformer-based models adopt various self-attention mechanisms to discover the long-range dependencies. However, intricate temporal patterns of the long-term future prohibit the model from finding reliable dependencies. Also, Transformers have to adopt the sparse versions of point-wise self-attentions for long series efficiency, resulting in the information utilization bottleneck. Going beyond Transformers, we design Autoformer as a novel decomposition architecture with an Auto-Correlation mechanism. We break with the pre-processing convention of series decomposition and renovate it as a basic inner block of deep models. This design empowers Autoformer with progressive decomposition capacities for complex time series. Further, inspired by the stochastic process theory, we design the Auto-Correlation mechanism based on the series periodicity, which conducts the dependencies discovery and representation aggregation at the sub-series level. Auto-Correlation outperforms self-attention in both efficiency and accuracy. In long-term forecasting, Autoformer yields state-of-the-art accuracy, with a 38% relative improvement on six benchmarks, covering five practical applications: energy, traffic, economics, weather and disease. Code is available at this repository: \\url{https://github.com/thuml/Autoformer}."
                    },
                    {
                        "title": "MotionRNN: A Flexible Model for Video Prediction with Spacetime-Varying Motions",
                        "abstract": "This paper tackles video prediction from a new dimension of predicting spacetime-varying motions that are incessantly changing across both space and time. Prior methods mainly capture the temporal state transitions but overlook the complex spatiotemporal variations of the motion itself, making them difficult to adapt to ever-changing motions. We observe that physical world motions can be decomposed into transient variation and motion trend, while the latter can be regarded as the accumulation of previous motions. Thus, simultaneously capturing the transient variation and the motion trend is the key to make spacetime-varying motions more predictable. Based on these observations, we propose the MotionRNN framework, which can capture the complex variations within motions and adapt to spacetime-varying scenarios. MotionRNN has two main contributions. The first is that we design the MotionGRU unit, which can model the transient variation and motion trend in a unified way. The second is that we apply the MotionGRU to RNN-based predictive models and indicate a new flexible video prediction architecture with a Motion Highway that can significantly improve the ability to predict changeable motions and avoid motion vanishing for stacked multiple-layer predictive models. With high flexibility, this framework can adapt to a series of models for deterministic spatiotemporal prediction. Our MotionRNN can yield significant improvements on three challenging benchmarks for video prediction with spacetime-varying motions."
                    },
                    {
                        "title": "Non-stationary Transformers: Exploring the Stationarity in Time Series Forecasting",
                        "abstract": "Transformers have shown great power in time series forecasting due to their global-range modeling ability. However, their performance can degenerate terribly on non-stationary real-world data in which the joint distribution changes over time. Previous studies primarily adopt stationarization to attenuate the non-stationarity of original series for better predictability. But the stationarized series deprived of inherent non-stationarity can be less instructive for real-world bursty events forecasting. This problem, termed over-stationarization in this paper, leads Transformers to generate indistinguishable temporal attentions for different series and impedes the predictive capability of deep models. To tackle the dilemma between series predictability and model capability, we propose Non-stationary Transformers as a generic framework with two interdependent modules: Series Stationarization and De-stationary Attention. Concretely, Series Stationarization unifies the statistics of each input and converts the output with restored statistics for better predictability. To address the over-stationarization problem, De-stationary Attention is devised to recover the intrinsic non-stationary information into temporal dependencies by approximating distinguishable attentions learned from raw series. Our Non-stationary Transformers framework consistently boosts mainstream Transformers by a large margin, which reduces MSE by 49.43% on Transformer, 47.34% on Informer, and 46.89% on Reformer, making them the state-of-the-art in time series forecasting. Code is available at this repository: https://github.com/thuml/Nonstationary_Transformers."
                    },
                    {
                        "title": "Supported Policy Optimization for Offline Reinforcement Learning",
                        "abstract": "Policy constraint methods to offline reinforcement learning (RL) typically utilize parameterization or regularization that constrains the policy to perform actions within the support set of the behavior policy. The elaborative designs of parameterization methods usually intrude into the policy networks, which may bring extra inference cost and cannot take full advantage of well-established online methods. Regularization methods reduce the divergence between the learned policy and the behavior policy, which may mismatch the inherent density-based definition of support set thereby failing to avoid the out-of-distribution actions effectively. This paper presents Supported Policy OpTimization (SPOT), which is directly derived from the theoretical formalization of the density-based support constraint. SPOT adopts a VAE-based density estimator to explicitly model the support set of behavior policy and presents a simple but effective density-based regularization term, which can be plugged non-intrusively into off-the-shelf off-policy RL algorithms. SPOT achieves the state-of-the-art performance on standard benchmarks for offline RL. Benefiting from the pluggable design, offline pretrained models from SPOT can also be applied to perform online fine-tuning seamlessly."
                    },
                    {
                        "title": "Flowformer: Linearizing Transformers with Conservation Flows",
                        "abstract": "Transformers based on the attention mechanism have achieved impressive success in various areas. However, the attention mechanism has a quadratic complexity, significantly impeding Transformers from dealing with numerous tokens and scaling up to bigger models. Previous methods mainly utilize the similarity decomposition and the associativity of matrix multiplication to devise linear-time attention mechanisms. They avoid degeneration of attention to a trivial distribution by reintroducing inductive biases such as the locality, thereby at the expense of model generality and expressiveness. In this paper, we linearize Transformers free from specific inductive biases based on the flow network theory. We cast attention as the information flow aggregated from the sources (values) to the sinks (results) through the learned flow capacities (attentions). Within this framework, we apply the property of flow conservation into attention and propose the Flow-Attention mechanism of linear complexity. By respectively conserving the incoming flow of sinks for source competition and the outgoing flow of sources for sink allocation, Flow-Attention inherently generates informative attentions without using specific inductive biases. Empowered by the Flow-Attention, Flowformer yields strong performance in linear time for wide areas, including long sequence, time series, vision, natural language, and reinforcement learning. The code and settings are available at this repository: https://github.com/thuml/Flowformer."
                    },
                    {
                        "title": "Transolver: A Fast Transformer Solver for PDEs on General Geometries",
                        "abstract": "Transformers have empowered many milestones across various fields and have recently been applied to solve partial differential equations (PDEs). However, since PDEs are typically discretized into large-scale meshes with complex geometries, it is challenging for Transformers to capture intricate physical correlations directly from massive individual points. Going beyond superficial and unwieldy meshes, we present Transolver based on a more foundational idea, which is learning intrinsic physical states hidden behind discretized geometries. Specifically, we propose a new Physics-Attention to adaptively split the discretized domain into a series of learnable slices of flexible shapes, where mesh points under similar physical states will be ascribed to the same slice. By calculating attention to physics-aware tokens encoded from slices, Transovler can effectively capture intricate physical correlations under complex geometrics, which also empowers the solver with endogenetic geometry-general modeling capacity and can be efficiently computed in linear complexity. Transolver achieves consistent state-of-the-art with 22% relative gain across six standard benchmarks and also excels in large-scale industrial simulations, including car and airfoil designs. Code is available at https://github.com/thuml/Transolver."
                    },
                    {
                        "title": "RoPINN: Region Optimized Physics-Informed Neural Networks",
                        "abstract": "Physics-informed neural networks (PINNs) have been widely applied to solve partial differential equations (PDEs) by enforcing outputs and gradients of deep models to satisfy target equations. Due to the limitation of numerical computation, PINNs are conventionally optimized on finite selected points. However, since PDEs are usually defined on continuous domains, solely optimizing models on scattered points may be insufficient to obtain an accurate solution for the whole domain. To mitigate this inherent deficiency of the default scatter-point optimization, this paper proposes and theoretically studies a new training paradigm as region optimization. Concretely, we propose to extend the optimization process of PINNs from isolated points to their continuous neighborhood regions, which can theoretically decrease the generalization error, especially for hidden high-order constraints of PDEs. A practical training algorithm, Region Optimized PINN (RoPINN), is seamlessly derived from this new paradigm, which is implemented by a straightforward but effective Monte Carlo sampling method. By calibrating the sampling process into trust regions, RoPINN finely balances optimization and generalization error. Experimentally, RoPINN consistently boosts the performance of diverse PINNs on a wide range of PDEs without extra backpropagation or gradient calculation. Code is available at this repository: https://github.com/thuml/RoPINN."
                    },
                    {
                        "title": "Solving High-Dimensional PDEs with Latent Spectral Models",
                        "abstract": "Deep models have achieved impressive progress in solving partial differential equations (PDEs). A burgeoning paradigm is learning neural operators to approximate the input-output mappings of PDEs. While previous deep models have explored the multiscale architectures and various operator designs, they are limited to learning the operators as a whole in the coordinate space. In real physical science problems, PDEs are complex coupled equations with numerical solvers relying on discretization into high-dimensional coordinate space, which cannot be precisely approximated by a single operator nor efficiently learned due to the curse of dimensionality. We present Latent Spectral Models (LSM) toward an efficient and precise solver for high-dimensional PDEs. Going beyond the coordinate space, LSM enables an attention-based hierarchical projection network to reduce the high-dimensional data into a compact latent space in linear time. Inspired by classical spectral methods in numerical analysis, we design a neural spectral block to solve PDEs in the latent space that approximates complex input-output mappings via learning multiple basis operators, enjoying nice theoretical guarantees for convergence and approximation. Experimentally, LSM achieves consistent state-of-the-art and yields a relative gain of 11.5% averaged on seven benchmarks covering both solid and fluid physics. Code is available at https://github.com/thuml/Latent-Spectral-Models."
                    },
                    {
                        "title": "HelmFluid: Learning Helmholtz Dynamics for Interpretable Fluid Prediction",
                        "abstract": "Fluid prediction is a long-standing challenge due to the intrinsic high-dimensional non-linear dynamics. Previous methods usually utilize the non-linear modeling capability of deep models to directly estimate velocity fields for future prediction. However, skipping over inherent physical properties but directly learning superficial velocity fields will overwhelm the model from generating precise or physics-reliable results. In this paper, we propose the HelmFluid toward an accurate and interpretable predictor for fluid. Inspired by the Helmholtz theorem, we design a HelmDynamics block to learn Helmholtz dynamics, which decomposes fluid dynamics into more solvable curl-free and divergence-free parts, physically corresponding to potential and stream functions of fluid. By embedding the HelmDynamics block into a Multiscale Multihead Integral Architecture, HelmFluid can integrate learned Helmholtz dynamics along temporal dimension in multiple spatial scales to yield future fluid. Compared with previous velocity estimating methods, HelmFluid is faithfully derived from Helmholtz theorem and ravels out complex fluid dynamics with physically interpretable evidence. Experimentally, HelmFluid achieves consistent state-of-the-art in both numerical simulated and real-world observed benchmarks, even for scenarios with complex boundaries."
                    },
                    {
                        "title": "ModeRNN: Harnessing Spatiotemporal Mode Collapse in Unsupervised Predictive Learning",
                        "abstract": "Learning predictive models for unlabeled spatiotemporal data is challenging in part because visual dynamics can be highly entangled in real scenes, making existing approaches prone to overfit partial modes of physical processes while neglecting to reason about others. We name this phenomenon spatiotemporal mode collapse and explore it for the first time in predictive learning. The key is to provide the model with a strong inductive bias to discover the compositional structures of latent modes. To this end, we propose ModeRNN, which introduces a novel method to learn structured hidden representations between recurrent states. The core idea of this framework is to first extract various components of visual dynamics using a set of spatiotemporal slots with independent parameters. Considering that multiple space-time patterns may co-exist in a sequence, we leverage learnable importance weights to adaptively aggregate slot features into a unified hidden representation, which is then used to update the recurrent states. Across the entire dataset, different modes result in different responses on the mixtures of slots, which enhances the ability of ModeRNN to build structured representations and thus prevents the so-called mode collapse. Unlike existing models, ModeRNN is shown to prevent spatiotemporal mode collapse and further benefit from learning mixed visual dynamics."
                    },
                    {
                        "title": "Unisolver: PDE-Conditional Transformers Are Universal PDE Solvers",
                        "abstract": "Deep models have recently emerged as a promising tool to solve partial differential equations (PDEs), known as neural PDE solvers. While neural solvers trained from either simulation data or physics-informed loss can solve PDEs reasonably well, they are mainly restricted to a few instances of PDEs, e.g. a certain equation with a limited set of coefficients. This limits the generalization of neural solvers to diverse PDEs, impeding them from being practical surrogate models for numerical solvers. In this paper, we present the Universal PDE Solver (Unisolver) capable of solving a wide scope of PDEs by training a novel Transformer model on diverse data and conditioned on diverse PDEs. Instead of purely scaling up data and parameters, Unisolver stems from the theoretical analysis of the PDE-solving process. Our key finding is that a PDE solution is fundamentally under the control of a series of PDE components, e.g. equation symbols, coefficients, and boundary conditions. Inspired by the mathematical structure of PDEs, we define a complete set of PDE components and flexibly embed them as domain-wise (e.g. equation symbols) and point-wise (e.g. boundaries) conditions for Transformer PDE solvers. Integrating physical insights with recent Transformer advances, Unisolver achieves consistent state-of-the-art results on three challenging large-scale benchmarks, showing impressive performance gains and favorable PDE generalizability."
                    },
                    {
                        "title": "TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis",
                        "abstract": "Time series analysis is of immense importance in extensive applications, such as weather forecasting, anomaly detection, and action recognition. This paper focuses on temporal variation modeling, which is the common key problem of extensive analysis tasks. Previous methods attempt to accomplish this directly from the 1D time series, which is extremely challenging due to the intricate temporal patterns. Based on the observation of multi-periodicity in time series, we ravel out the complex temporal variations into the multiple intraperiod- and interperiod-variations. To tackle the limitations of 1D time series in representation capability, we extend the analysis of temporal variations into the 2D space by transforming the 1D time series into a set of 2D tensors based on multiple periods. This transformation can embed the intraperiod- and interperiod-variations into the columns and rows of the 2D tensors respectively, making the 2D-variations to be easily modeled by 2D kernels. Technically, we propose the TimesNet with TimesBlock as a task-general backbone for time series analysis. TimesBlock can discover the multi-periodicity adaptively and extract the complex temporal variations from transformed 2D tensors by a parameter-efficient inception block. Our proposed TimesNet achieves consistent state-of-the-art in five mainstream time series analysis tasks, including short- and long-term forecasting, imputation, classification, and anomaly detection. Code is available at this repository: https://github.com/thuml/TimesNet."
                    },
                    {
                        "title": "TimeSiam: A Pre-Training Framework for Siamese Time-Series Modeling",
                        "abstract": "Time series pre-training has recently garnered wide attention for its potential to reduce labeling expenses and benefit various downstream tasks. Prior methods are mainly based on pre-training techniques well-acknowledged in vision or language, such as masked modeling and contrastive learning. However, randomly masking time series or calculating series-wise similarity will distort or neglect inherent temporal correlations crucial in time series data. To emphasize temporal correlation modeling, this paper proposes TimeSiam as a simple but effective self-supervised pre-training framework for Time series based on Siamese networks. Concretely, TimeSiam pre-trains Siamese encoders to capture intrinsic temporal correlations between randomly sampled past and current subseries. With a simple data augmentation method (e.g.~masking), TimeSiam can benefit from diverse augmented subseries and learn internal time-dependent representations through a past-to-current reconstruction. Moreover, learnable lineage embeddings are also introduced to distinguish temporal distance between sampled series and further foster the learning of diverse temporal correlations. TimeSiam consistently outperforms extensive advanced pre-training baselines, demonstrating superior forecasting and classification capabilities across 13 standard benchmarks in both intra- and cross-domain scenarios."
                    },
                    {
                        "title": "SimMTM: A Simple Pre-Training Framework for Masked Time-Series Modeling",
                        "abstract": "Time series analysis is widely used in extensive areas. Recently, to reduce labeling expenses and benefit various tasks, self-supervised pre-training has attracted immense interest. One mainstream paradigm is masked modeling, which successfully pre-trains deep models by learning to reconstruct the masked content based on the unmasked part. However, since the semantic information of time series is mainly contained in temporal variations, the standard way of randomly masking a portion of time points will seriously ruin vital temporal variations of time series, making the reconstruction task too difficult to guide representation learning. We thus present SimMTM, a Simple pre-training framework for Masked Time-series Modeling. By relating masked modeling to manifold learning, SimMTM proposes to recover masked time points by the weighted aggregation of multiple neighbors outside the manifold, which eases the reconstruction task by assembling ruined but complementary temporal variations from multiple masked series. SimMTM further learns to uncover the local structure of the manifold, which is helpful for masked modeling. Experimentally, SimMTM achieves state-of-the-art fine-tuning performance compared to the most advanced time series pre-training methods in two canonical time series analysis tasks: forecasting and classification, covering both in- and cross-domain settings."
                    },
                    {
                        "title": "iTransformer: Inverted Transformers Are Effective for Time Series Forecasting",
                        "abstract": "The recent boom of linear forecasting models questions the ongoing passion for architectural modifications of Transformer-based forecasters. These forecasters leverage Transformers to model the global dependencies over temporal tokens of time series, with each token formed by multiple variates of the same timestamp. However, Transformers are challenged in forecasting series with larger lookback windows due to performance degradation and computation explosion. Besides, the embedding for each temporal token fuses multiple variates that represent potential delayed events and distinct physical measurements, which may fail in learning variate-centric representations and result in meaningless attention maps. In this work, we reflect on the competent duties of Transformer components and repurpose the Transformer architecture without any modification to the basic components. We propose iTransformer that simply applies the attention and feed-forward network on the inverted dimensions. Specifically, the time points of individual series are embedded into variate tokens which are utilized by the attention mechanism to capture multivariate correlations; meanwhile, the feed-forward network is applied for each variate token to learn nonlinear representations. The iTransformer model achieves state-of-the-art on challenging real-world datasets, which further empowers the Transformer family with promoted performance, generalization ability across different variates, and better utilization of arbitrary lookback windows, making it a nice alternative as the fundamental backbone of time series forecasting. Code is available at this repository: https://github.com/thuml/iTransformer."
                    },
                    {
                        "title": "Deep Time Series Models: A Comprehensive Survey and Benchmark",
                        "abstract": "Time series, characterized by a sequence of data points arranged in a discrete-time order, are ubiquitous in real-world applications. Different from other modalities, time series present unique challenges due to their complex and dynamic nature, including the entanglement of nonlinear patterns and time-variant trends. Analyzing time series data is of great significance in real-world scenarios and has been widely studied over centuries. Recent years have witnessed remarkable breakthroughs in the time series community, with techniques shifting from traditional statistical methods to advanced deep learning models. In this paper, we delve into the design of deep time series models across various analysis tasks and review the existing literature from two perspectives: basic modules and model architectures. Further, we develop and release Time Series Library (TSLib) as a fair benchmark of deep time series models for diverse analysis tasks, which implements 24 mainstream models, covers 30 datasets from different domains, and supports five prevalent analysis tasks. Based on TSLib, we thoroughly evaluate 12 advanced deep time series models on different tasks. Empirical results indicate that models with specific structures are well-suited for distinct analytical tasks, which offers insights for research and adoption of deep time series models. Code is available at https://github.com/thuml/Time-Series-Library."
                    }
                ]
            },
            "48aae5c3-d4e2-4f4b-b199-23c184d8b729": {
                "pk": "48aae5c3-d4e2-4f4b-b199-23c184d8b729",
                "name": "Lanxiang Xing",
                "collaborators": [
                    "Haixu Wu",
                    "Yuezhou Ma",
                    "Jianmin Wang",
                    "Mingsheng Long"
                ],
                "domain": [
                    "Fluid Dynamics",
                    "Deep Learning",
                    "Physics-Informed Models"
                ],
                "publications": [
                    {
                        "title": "HelmFluid: Learning Helmholtz Dynamics for Interpretable Fluid Prediction",
                        "abstract": "Fluid prediction is a long-standing challenge due to the intrinsic high-dimensional non-linear dynamics. Previous methods usually utilize the non-linear modeling capability of deep models to directly estimate velocity fields for future prediction. However, skipping over inherent physical properties but directly learning superficial velocity fields will overwhelm the model from generating precise or physics-reliable results. In this paper, we propose the HelmFluid toward an accurate and interpretable predictor for fluid. Inspired by the Helmholtz theorem, we design a HelmDynamics block to learn Helmholtz dynamics, which decomposes fluid dynamics into more solvable curl-free and divergence-free parts, physically corresponding to potential and stream functions of fluid. By embedding the HelmDynamics block into a Multiscale Multihead Integral Architecture, HelmFluid can integrate learned Helmholtz dynamics along temporal dimension in multiple spatial scales to yield future fluid. Compared with previous velocity estimating methods, HelmFluid is faithfully derived from Helmholtz theorem and ravels out complex fluid dynamics with physically interpretable evidence. Experimentally, HelmFluid achieves consistent state-of-the-art in both numerical simulated and real-world observed benchmarks, even for scenarios with complex boundaries."
                    }
                ]
            },
            "49a09c58-febb-4c94-b3c7-5ee18d2130aa": {
                "pk": "49a09c58-febb-4c94-b3c7-5ee18d2130aa",
                "name": "Shangchen Miao",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "9a7bd6f5-39e3-4e3b-b3a9-d1a9258549da": {
                "pk": "9a7bd6f5-39e3-4e3b-b3a9-d1a9258549da",
                "name": "Mingsheng Long",
                "collaborators": [
                    "Jianmin Wang",
                    "Junguang Jiang",
                    "Zhangjie Cao",
                    "Yong Liu",
                    "Haixu Wu",
                    "Jiehui Xu",
                    "Yang Shu",
                    "Liangbin Hu",
                    "W. LiMing",
                    "Chao Huang"
                ],
                "domain": [
                    "Time Series Forecasting",
                    "Transfer Learning",
                    "Deep Learning",
                    "Generative Modeling"
                ],
                "publications": [
                    {
                        "title": "Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting",
                        "abstract": "Extending the forecasting time is a critical demand for real applications, such as extreme weather early warning and long-term energy consumption planning. This paper studies the long-term forecasting problem of time series. Prior Transformer-based models adopt various self-attention mechanisms to discover the long-range dependencies. However, intricate temporal patterns of the long-term future prohibit the model from finding reliable dependencies. Also, Transformers have to adopt the sparse versions of point-wise self-attentions for long series efficiency, resulting in the information utilization bottleneck. Going beyond Transformers, we design Autoformer as a novel decomposition architecture with an Auto-Correlation mechanism. We break with the pre-processing convention of series decomposition and renovate it as a basic inner block of deep models. This design empowers Autoformer with progressive decomposition capacities for complex time series. Further, inspired by the stochastic process theory, we design the Auto-Correlation mechanism based on the series periodicity, which conducts the dependencies discovery and representation aggregation at the sub-series level. Auto-Correlation outperforms self-attention in both efficiency and accuracy. In long-term forecasting, Autoformer yields state-of-the-art accuracy, with a 38% relative improvement on six benchmarks, covering five practical applications: energy, traffic, economics, weather and disease. Code is available at this repository: \\url{https://github.com/thuml/Autoformer}."
                    },
                    {
                        "title": "Transferability in Deep Learning: A Survey",
                        "abstract": "The success of deep learning algorithms generally depends on large-scale data, while humans appear to have inherent ability of knowledge transfer, by recognizing and applying relevant knowledge from previous learning experiences when encountering and solving unseen tasks. Such an ability to acquire and reuse knowledge is known as transferability in deep learning. It has formed the long-term quest towards making deep learning as data-efficient as human learning, and has been motivating fruitful design of more powerful deep learning algorithms. We present this survey to connect different isolated areas in deep learning with their relation to transferability, and to provide a unified and complete view to investigating transferability through the whole lifecycle of deep learning. The survey elaborates the fundamental goals and challenges in parallel with the core principles and methods, covering recent cornerstones in deep architectures, pre-training, task adaptation and domain adaptation. This highlights unanswered questions on the appropriate objectives for learning transferable knowledge and for adapting the knowledge to new tasks and domains, avoiding catastrophic forgetting and negative transfer. Finally, we implement a benchmark and an open-source library, enabling a fair evaluation of deep learning methods in terms of transferability."
                    },
                    {
                        "title": "Spin density wave in oxypnictide superconductors in a three-band model",
                        "abstract": "The spin density wave and its temperature dependence in oxypnictide are studied in a three-band model. The spin susceptibilities with various interactions are calculated in the random phase approximation(PPA). It is found that the spin susceptibility peaks around the M point show a spin density wave(SDW) with momentum (0, $\\pi$) and a clear stripe-like spin configuration. The intra-band Coulomb repulsion enhances remarkably the SDW but the Hund's coupling weakens it. It is shown that a new resonance appears at higher temperatures at the $\\Gamma$ point indicating the formation of a paramagnetic phase. There is a clear transition from the SDW phase to the paramagnetic phase."
                    },
                    {
                        "title": "Transfer Adversarial Hashing for Hamming Space Retrieval",
                        "abstract": "Hashing is widely applied to large-scale image retrieval due to the storage and retrieval efficiency. Existing work on deep hashing assumes that the database in the target domain is identically distributed with the training set in the source domain. This paper relaxes this assumption to a transfer retrieval setting, which allows the database and the training set to come from different but relevant domains. However, the transfer retrieval setting will introduce two technical difficulties: first, the hash model trained on the source domain cannot work well on the target domain due to the large distribution gap; second, the domain gap makes it difficult to concentrate the database points to be within a small Hamming ball. As a consequence, transfer retrieval performance within Hamming Radius 2 degrades significantly in existing hashing methods. This paper presents Transfer Adversarial Hashing (TAH), a new hybrid deep architecture that incorporates a pairwise $t$-distribution cross-entropy loss to learn concentrated hash codes and an adversarial network to align the data distributions between the source and target domains. TAH can generate compact transfer hash codes for efficient image retrieval on both source and target domains. Comprehensive experiments validate that TAH yields state of the art Hamming space retrieval performance on standard datasets."
                    },
                    {
                        "title": "Transitive Hashing Network for Heterogeneous Multimedia Retrieval",
                        "abstract": "Hashing has been widely applied to large-scale multimedia retrieval due to the storage and retrieval efficiency. Cross-modal hashing enables efficient retrieval from database of one modality in response to a query of another modality. Existing work on cross-modal hashing assumes heterogeneous relationship across modalities for hash function learning. In this paper, we relax the strong assumption by only requiring such heterogeneous relationship in an auxiliary dataset different from the query/database domain. We craft a hybrid deep architecture to simultaneously learn the cross-modal correlation from the auxiliary dataset, and align the dataset distributions between the auxiliary dataset and the query/database domain, which generates transitive hash codes for heterogeneous multimedia retrieval. Extensive experiments exhibit that the proposed approach yields state of the art multimedia retrieval performance on public datasets, i.e. NUS-WIDE, ImageNet-YahooQA."
                    },
                    {
                        "title": "Decoupled Adaptation for Cross-Domain Object Detection",
                        "abstract": "Cross-domain object detection is more challenging than object classification since multiple objects exist in an image and the location of each object is unknown in the unlabeled target domain. As a result, when we adapt features of different objects to enhance the transferability of the detector, the features of the foreground and the background are easy to be confused, which may hurt the discriminability of the detector. Besides, previous methods focused on category adaptation but ignored another important part for object detection, i.e., the adaptation on bounding box regression. To this end, we propose D-adapt, namely Decoupled Adaptation, to decouple the adversarial adaptation and the training of the detector. Besides, we fill the blank of regression domain adaptation in object detection by introducing a bounding box adaptor. Experiments show that D-adapt achieves state-of-the-art results on four cross-domain object detection tasks and yields 17% and 21% relative improvement on benchmark datasets Clipart1k and Comic2k in particular."
                    },
                    {
                        "title": "Unsupervised Transfer Learning for Spatiotemporal Predictive Networks",
                        "abstract": "This paper explores a new research problem of unsupervised transfer learning across multiple spatiotemporal prediction tasks. Unlike most existing transfer learning methods that focus on fixing the discrepancy between supervised tasks, we study how to transfer knowledge from a zoo of unsupervisedly learned models towards another predictive network. Our motivation is that models from different sources are expected to understand the complex spatiotemporal dynamics from different perspectives, thereby effectively supplementing the new task, even if the task has sufficient training samples. Technically, we propose a differentiable framework named transferable memory. It adaptively distills knowledge from a bank of memory states of multiple pretrained RNNs, and applies it to the target network via a novel recurrent structure called the Transferable Memory Unit (TMU). Compared with finetuning, our approach yields significant improvements on three benchmarks for spatiotemporal prediction, and benefits the target task even from less relevant pretext ones."
                    },
                    {
                        "title": "Koopa: Learning Non-stationary Time Series Dynamics with Koopman Predictors",
                        "abstract": "Real-world time series are characterized by intrinsic non-stationarity that poses a principal challenge for deep forecasting models. While previous models suffer from complicated series variations induced by changing temporal distribution, we tackle non-stationary time series with modern Koopman theory that fundamentally considers the underlying time-variant dynamics. Inspired by Koopman theory of portraying complex dynamical systems, we disentangle time-variant and time-invariant components from intricate non-stationary series by Fourier Filter and design Koopman Predictor to advance respective dynamics forward. Technically, we propose Koopa as a novel Koopman forecaster composed of stackable blocks that learn hierarchical dynamics. Koopa seeks measurement functions for Koopman embedding and utilizes Koopman operators as linear portraits of implicit transition. To cope with time-variant dynamics that exhibits strong locality, Koopa calculates context-aware operators in the temporal neighborhood and is able to utilize incoming ground truth to scale up forecast horizon. Besides, by integrating Koopman Predictors into deep residual structure, we ravel out the binding reconstruction loss in previous Koopman forecasters and achieve end-to-end forecasting objective optimization. Compared with the state-of-the-art model, Koopa achieves competitive performance while saving 77.3% training time and 76.0% memory."
                    },
                    {
                        "title": "Diffusion Tuning: Transferring Diffusion Models via Chain of Forgetting",
                        "abstract": "Diffusion models have significantly advanced the field of generative modeling. However, training a diffusion model is computationally expensive, creating a pressing need to adapt off-the-shelf diffusion models for downstream generation tasks. Current fine-tuning methods focus on parameter-efficient transfer learning but overlook the fundamental transfer characteristics of diffusion models. In this paper, we investigate the transferability of diffusion models and observe a monotonous chain of forgetting trend of transferability along the reverse process. Based on this observation and novel theoretical insights, we present Diff-Tuning, a frustratingly simple transfer approach that leverages the chain of forgetting tendency. Diff-Tuning encourages the fine-tuned model to retain the pre-trained knowledge at the end of the denoising chain close to the generated data while discarding the other noise side. We conduct comprehensive experiments to evaluate Diff-Tuning, including the transfer of pre-trained Diffusion Transformer models to eight downstream generations and the adaptation of Stable Diffusion to five control conditions with ControlNet. Diff-Tuning achieves a 26% improvement over standard fine-tuning and enhances the convergence speed of ControlNet by 24%. Notably, parameter-efficient transfer learning techniques for diffusion models can also benefit from Diff-Tuning."
                    },
                    {
                        "title": "LogME: Practical Assessment of Pre-trained Models for Transfer Learning",
                        "abstract": "This paper studies task adaptive pre-trained model selection, an underexplored problem of assessing pre-trained models for the target task and select best ones from the model zoo \\emph{without fine-tuning}. A few pilot works addressed the problem in transferring supervised pre-trained models to classification tasks, but they cannot handle emerging unsupervised pre-trained models or regression tasks. In pursuit of a practical assessment method, we propose to estimate the maximum value of label evidence given features extracted by pre-trained models. Unlike the maximum likelihood, the maximum evidence is \\emph{immune to over-fitting}, while its expensive computation can be dramatically reduced by our carefully designed algorithm. The Logarithm of Maximum Evidence (LogME) can be used to assess pre-trained models for transfer learning: a pre-trained model with a high LogME value is likely to have good transfer performance. LogME is \\emph{fast, accurate, and general}, characterizing itself as the first practical method for assessing pre-trained models. Compared with brute-force fine-tuning, LogME brings at most $3000\\times$ speedup in wall-clock time and requires only $1\\%$ memory footprint. It outperforms prior methods by a large margin in their setting and is applicable to new settings. It is general enough for diverse pre-trained models (supervised pre-trained and unsupervised pre-trained), downstream tasks (classification and regression), and modalities (vision and language). Code is available at this repository: \\href{https://github.com/thuml/LogME}{https://github.com/thuml/LogME}."
                    }
                ]
            }
        }
    },
    "2408.15065": {
        "paper_data": {
            "title": "The Benefits of Balance: From Information Projections to Variance Reduction",
            "url": "http://arxiv.org/abs/2408.15065v1",
            "arxiv_id": "2408.15065",
            "authors": [
                "Lang Liu",
                "Ronak Mehta",
                "Soumik Pal",
                "Zaid Harchaoui"
            ],
            "abstract": "Data balancing across multiple modalities/sources appears in various forms in several foundation models (e.g., CLIP and DINO) achieving universal representation learning. We show that this iterative algorithm, usually used to avoid representation collapse, enjoys an unsuspected benefit: reducing the variance of estimators that are functionals of the empirical distribution over these sources. We provide non-asymptotic bounds quantifying this variance reduction effect and relate them to the eigendecays of appropriately defined Markov operators. We explain how various forms of data balancing in contrastive multimodal learning and self-supervised clustering can be interpreted as instances of this variance reduction scheme.",
            "introduction": "   1 Introduction  Deep neural networks have shown remarkable success at learning task-specific representations of data when provided supervision from massive amounts of labeled training examples. Recent trends, however, have shifted toward task-agnostic, universal representations that may be easily fine-tuned or even have zero-shot capabilities out-of-the-box. Supervised learning, stricto sensu, is too limited a framework for these billion-parameter, data-hungry models, and a question at the heart of modern machine learning is learning from unlabelled, partially labeled, or weakly labeled data.   This need has paved the way for the current generation of self-supervised learning (SSL) approaches that circumvent the need for large amounts of strong labels. In SSL, a model is trained on a generic pseudo-task that can be performed on unlabelled data, such as relating the two modalities of an image-caption pair or two augmentations of the same image. Despite several modern foundation models such as DINO (Caron et\u00a0al., 2021; Oquab et\u00a0al., 2024) and CLIP (Radford et\u00a0al., 2021) being trained in this fashion, many aspects of SSL remain baffling.   In particular, the training process of self-supervised models often outgrows and \u201cbreaks the rules\u201d of the standard empirical risk minimization (ERM) toolkit. ERM combines two well-understood techniques: minibatch sampling and gradient-based optimization using backpropagation. SSL, on the other hand, adds clever, less-understood techniques to the training pipeline. To illustrate this, consider a minibatch \u2133n={Z1,\u2026,Zn}subscript\u2133\ud835\udc5bsubscript\ud835\udc4d1\u2026subscript\ud835\udc4d\ud835\udc5b\\mathcal{M}_{n}=\\left\\{Z_{1},\\ldots,Z_{n}\\right\\}caligraphic_M start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT = { italic_Z start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , \u2026 , italic_Z start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT } of training examples and let Pnsubscript\ud835\udc43\ud835\udc5bP_{n}italic_P start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT be the empirical distribution of the minibatch. For a model parameterized by \u03b8\u2208\u211dd\ud835\udf03superscript\u211d\ud835\udc51\\theta\\in\\mathbb{R}^{d}italic_\u03b8 \u2208 blackboard_R start_POSTSUPERSCRIPT italic_d end_POSTSUPERSCRIPT with associated loss function \u2113\u03b8subscript\u2113\ud835\udf03\\ell_{\\theta}roman_\u2113 start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT, a standard stochastic supervised learning algorithm, e.g., Adam, involves computing the minibatch loss    \ud835\udd3cZ\u223cPn\u2062[\u2113\u03b8\u2062(Z)]=1n\u2062\u2211i=1n\u2113\u03b8\u2062(Zi)subscript\ud835\udd3csimilar-to\ud835\udc4dsubscript\ud835\udc43\ud835\udc5bdelimited-[]subscript\u2113\ud835\udf03\ud835\udc4d1\ud835\udc5bsuperscriptsubscript\ud835\udc561\ud835\udc5bsubscript\u2113\ud835\udf03subscript\ud835\udc4d\ud835\udc56\\displaystyle{\\mathbb{E}}_{Z\\sim P_{n}}\\left[\\ell_{\\theta}(Z)\\right]=\\frac{1}{% n}\\sum_{i=1}^{n}\\ell_{\\theta}(Z_{i})blackboard_E start_POSTSUBSCRIPT italic_Z \u223c italic_P start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT end_POSTSUBSCRIPT [ roman_\u2113 start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( italic_Z ) ] = divide start_ARG 1 end_ARG start_ARG italic_n end_ARG \u2211 start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_n end_POSTSUPERSCRIPT roman_\u2113 start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( italic_Z start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT )  (1)   and backpropagating through it to produce a minibatch stochastic gradient estimate. The algorithm then proceeds with the stochastic gradient training, or a variant thereof.   On the other hand, self-supervised methods often modify this recipe by intervening on the optimization algorithm in a minibatch-specific way. For example, SwaV (Caron et\u00a0al., 2020) passes the minibatch examples through the model\u2019s encoder and clusters output vectors to generate pseudo-labels for a prediction task. In teacher-student architectures such as BYOL (Grill et\u00a0al., 2020) and DINO (Caron et\u00a0al., 2021), the data are passed through two networks, where the \u201cstudent\u201d is updated via backpropagation and the \u201cteacher\u201d is updated by cloning the student\u2019s weights in regular intervals. In CLIP (Radford et\u00a0al., 2021), a model optimizes the sum of two cross entropy loss terms, where the predicted class probabilities on example i\ud835\udc56iitalic_i are generated by comparison to all other elements of the minibatch. These steps are often motivated as constraints to avoid representation collapse, wherein the pseudo-task can be solved by a trivial representation, e.g.\u00a0mapping all data to the zero vector. Conceptually, however, it is difficult to see what exactly is being optimized when introducing",
            "references": [
                {
                    "title": "Demystifying CLIP Data",
                    "abstract": "Contrastive Language-Image Pre-training (CLIP) is an approach that has advanced research and applications in computer vision, fueling modern recognition systems and generative models. We believe that the main ingredient to the success of CLIP is its data and not the model architecture or pre-training objective. However, CLIP only provides very limited information about its data and how it has been collected, leading to works that aim to reproduce CLIP's data by filtering with its model parameters. In this work, we intend to reveal CLIP's data curation approach and in our pursuit of making it open to the community introduce Metadata-Curated Language-Image Pre-training (MetaCLIP). MetaCLIP takes a raw data pool and metadata (derived from CLIP's concepts) and yields a balanced subset over the metadata distribution. Our experimental study rigorously isolates the model and training settings, concentrating solely on data. MetaCLIP applied to CommonCrawl with 400M image-text data pairs outperforms CLIP's data on multiple standard benchmarks. In zero-shot ImageNet classification, MetaCLIP achieves 70.8% accuracy, surpassing CLIP's 68.3% on ViT-B models. Scaling to 1B data, while maintaining the same training budget, attains 72.4%. Our observations hold across various model sizes, exemplified by ViT-H achieving 80.5%, without any bells-and-whistles. Curation code and training data distribution on metadata is made available at https://github.com/facebookresearch/MetaCLIP."
                },
                {
                    "title": "DataComp: In search of the next generation of multimodal datasets",
                    "abstract": "Multimodal datasets are a critical component in recent breakthroughs such as Stable Diffusion and GPT-4, yet their design does not receive the same research attention as model architectures or training algorithms. To address this shortcoming in the ML ecosystem, we introduce DataComp, a testbed for dataset experiments centered around a new candidate pool of 12.8 billion image-text pairs from Common Crawl. Participants in our benchmark design new filtering techniques or curate new data sources and then evaluate their new dataset by running our standardized CLIP training code and testing the resulting model on 38 downstream test sets. Our benchmark consists of multiple compute scales spanning four orders of magnitude, which enables the study of scaling trends and makes the benchmark accessible to researchers with varying resources. Our baseline experiments show that the DataComp workflow leads to better training sets. In particular, our best baseline, DataComp-1B, enables training a CLIP ViT-L/14 from scratch to 79.2% zero-shot accuracy on ImageNet, outperforming OpenAI's CLIP ViT-L/14 by 3.7 percentage points while using the same training procedure and compute. We release DataComp and all accompanying code at www.datacomp.ai."
                },
                {
                    "title": "A Cookbook of Self-Supervised Learning",
                    "abstract": "Self-supervised learning, dubbed the dark matter of intelligence, is a promising path to advance machine learning. Yet, much like cooking, training SSL methods is a delicate art with a high barrier to entry. While many components are familiar, successfully training a SSL method involves a dizzying set of choices from the pretext tasks to training hyper-parameters. Our goal is to lower the barrier to entry into SSL research by laying the foundations and latest SSL recipes in the style of a cookbook. We hope to empower the curious researcher to navigate the terrain of methods, understand the role of the various knobs, and gain the know-how required to explore how delicious SSL can be."
                },
                {
                    "title": "DINOv2: Learning Robust Visual Features without Supervision",
                    "abstract": "The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels."
                },
                {
                    "title": "LAION-5B: An open large-scale dataset for training next generation image-text models",
                    "abstract": "Groundbreaking language-vision architectures like CLIP and DALL-E proved the utility of training on large amounts of noisy image-text data, without relying on expensive accurate labels used in standard vision unimodal supervised learning. The resulting models showed capabilities of strong text-guided image generation and transfer to downstream tasks, while performing remarkably at zero-shot classification with noteworthy out-of-distribution robustness. Since then, large-scale language-vision models like ALIGN, BASIC, GLIDE, Flamingo and Imagen made further improvements. Studying the training and capabilities of such models requires datasets containing billions of image-text pairs. Until now, no datasets of this size have been made openly available for the broader research community. To address this problem and democratize research on large-scale multi-modal models, we present LAION-5B - a dataset consisting of 5.85 billion CLIP-filtered image-text pairs, of which 2.32B contain English language. We show successful replication and fine-tuning of foundational models like CLIP, GLIDE and Stable Diffusion using the dataset, and discuss further experiments enabled with an openly available dataset of this scale. Additionally we provide several nearest neighbor indices, an improved web-interface for dataset exploration and subset generation, and detection scores for watermark, NSFW, and toxic content detection. Announcement page https://laion.ai/laion-5b-a-new-era-of-open-large-scale-multi-modal-datasets/"
                },
                {
                    "title": "Data Determines Distributional Robustness in Contrastive Language Image Pre-training (CLIP)",
                    "abstract": "Contrastively trained language-image models such as CLIP, ALIGN, and BASIC have demonstrated unprecedented robustness to multiple challenging natural distribution shifts. Since these language-image models differ from previous training approaches in several ways, an important question is what causes the large robustness gains. We answer this question via a systematic experimental investigation. Concretely, we study five different possible causes for the robustness gains: (i) the training set size, (ii) the training distribution, (iii) language supervision at training time, (iv) language supervision at test time, and (v) the contrastive loss function. Our experiments show that the more diverse training distribution is the main cause for the robustness gains, with the other factors contributing little to no robustness. Beyond our experimental results, we also introduce ImageNet-Captions, a version of ImageNet with original text annotations from Flickr, to enable further controlled experiments of language-image training."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Exploring Simple Siamese Representation Learning",
                    "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code is made available.1"
                },
                {
                    "title": "Unsupervised Learning of Visual Features by Contrasting Cluster Assignments",
                    "abstract": "Unsupervised image representations have significantly reduced the gap with supervised pretraining, notably with the recent achievements of contrastive learning methods. These contrastive methods typically work online and rely on a large number of explicit pairwise feature comparisons, which is computationally challenging. In this paper, we propose an online algorithm, SwAV, that takes advantage of contrastive methods without requiring to compute pairwise comparisons. Specifically, our method simultaneously clusters the data while enforcing consistency between cluster assignments produced for different augmentations (or views) of the same image, instead of comparing features directly as in contrastive learning. Simply put, we use a swapped prediction mechanism where we predict the cluster assignment of a view from the representation of another view. Our method can be trained with large and small batches and can scale to unlimited amounts of data. Compared to previous contrastive methods, our method is more memory efficient since it does not require a large memory bank or a special momentum network. In addition, we also propose a new data augmentation strategy, multi-crop, that uses a mix of views with different resolutions in place of two full-resolution views, without increasing the memory or compute requirements much. We validate our findings by achieving 75.3% top-1 accuracy on ImageNet with ResNet-50, as well as surpassing supervised pretraining on all the considered transfer tasks."
                },
                {
                    "title": "Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning",
                    "abstract": "We introduce Bootstrap Your Own Latent (BYOL), a new approach to self-supervised image representation learning. BYOL relies on two neural networks, referred to as online and target networks, that interact and learn from each other. From an augmented view of an image, we train the online network to predict the target network representation of the same image under a different augmented view. At the same time, we update the target network with a slow-moving average of the online network. While state-of-the art methods rely on negative pairs, BYOL achieves a new state of the art without them. BYOL reaches $74.3\\%$ top-1 classification accuracy on ImageNet using a linear evaluation with a ResNet-50 architecture and $79.6\\%$ with a larger ResNet. We show that BYOL performs on par or better than the current state of the art on both transfer and semi-supervised benchmarks. Our implementation and pretrained models are given on GitHub."
                },
                {
                    "title": "Self-labelling via simultaneous clustering and representation learning",
                    "abstract": "Combining clustering and representation learning is one of the most promising approaches for unsupervised learning of deep neural networks. However, doing so naively leads to ill posed learning problems with degenerate solutions. In this paper, we propose a novel and principled learning formulation that addresses these issues. The method is obtained by maximizing the information between labels and input data indices. We show that this criterion extends standard crossentropy minimization to an optimal transport problem, which we solve efficiently for millions of input images and thousands of labels using a fast variant of the Sinkhorn-Knopp algorithm. The resulting method is able to self-label visual data so as to train highly competitive image representations without manual labels. Our method achieves state of the art representation learning performance for AlexNet and ResNet-50 on SVHN, CIFAR-10, CIFAR-100 and ImageNet and yields the first self-supervised AlexNet that outperforms the supervised Pascal VOC detection baseline. Code and models are available."
                },
                {
                    "title": "Statistical bounds for entropic optimal transport: sample complexity and the central limit theorem",
                    "abstract": "We prove several fundamental statistical bounds for entropic OT with the squared Euclidean cost between subgaussian probability measures in arbitrary dimension. First, through a new sample complexity result we establish the rate of convergence of entropic OT for empirical measures. Our analysis improves exponentially on the bound of Genevay et al. (2019) and extends their work to unbounded measures. Second, we establish a central limit theorem for entropic OT, based on techniques developed by Del Barrio and Loubes (2019). Previously, such a result was only known for finite metric spaces. As an application of our results, we develop and analyze a new technique for estimating the entropy of a random variable corrupted by gaussian noise."
                },
                {
                    "title": "Moment Matching for Multi-Source Domain Adaptation",
                    "abstract": "Conventional unsupervised domain adaptation (UDA) assumes that training data are sampled from a single domain. This neglects the more practical scenario where training data are collected from multiple sources, requiring multi-source domain adaptation. We make three major contributions towards addressing this problem. First, we collect and annotate by far the largest UDA dataset, called DomainNet, which contains six domains and about 0.6 million images distributed among 345 categories, addressing the gap in data availability for multi-source UDA research. Second, we propose a new deep learning approach, Moment Matching for Multi-Source Domain Adaptation (M3SDA), which aims to transfer knowledge learned from multiple labeled source domains to an unlabeled target domain by dynamically aligning moments of their feature distributions. Third, we provide new theoretical insights specifically for moment matching approaches in both single and multiple source domain adaptation. Extensive experiments are conducted to demonstrate the power of our new dataset in benchmarking state-of-the-art multi-source domain adaptation methods, as well as the advantage of our proposed model. Dataset and Code are available at http://ai.bu.edu/M3SDA/"
                },
                {
                    "title": "Empirical Regularized Optimal Transport: Statistical Theory and Applications",
                    "abstract": "We derive limit distributions for certain empirical regularized optimal transport distances between probability distributions supported on a finite metric space and show consistency of the (naive) bootstrap. In particular, we prove that the empirical regularized transport plan itself asymptotically follows a Gaussian law. The theory includes the Boltzmann-Shannon entropy regularization and hence a limit law for the widely applied Sinkhorn divergence. Our approach is based on an application of the implicit function theorem to necessary and sufficient optimality conditions for the regularized transport problem. The asymptotic results are investigated in Monte Carlo simulations. We further discuss computational and statistical applications, e.g. confidence bands for colocalization analysis of protein interaction networks based on regularized optimal transport."
                },
                {
                    "title": "Sample Complexity of Sinkhorn Divergences",
                    "abstract": "Optimal transport (OT) and maximum mean discrepancies (MMD) are now routinely used in machine learning to compare probability measures. We focus in this paper on \\emph{Sinkhorn divergences} (SDs), a regularized variant of OT distances which can interpolate, depending on the regularization strength $\\varepsilon$, between OT ($\\varepsilon=0$) and MMD ($\\varepsilon=\\infty$). Although the tradeoff induced by that regularization is now well understood computationally (OT, SDs and MMD require respectively $O(n^3\\log n)$, $O(n^2)$ and $n^2$ operations given a sample size $n$), much less is known in terms of their \\emph{sample complexity}, namely the gap between these quantities, when evaluated using finite samples \\emph{vs.} their respective densities. Indeed, while the sample complexity of OT and MMD stand at two extremes, $1/n^{1/d}$ for OT in dimension $d$ and $1/\\sqrt{n}$ for MMD, that for SDs has only been studied empirically. In this paper, we \\emph{(i)} derive a bound on the approximation error made with SDs when approximating OT as a function of the regularizer $\\varepsilon$, \\emph{(ii)} prove that the optimizers of regularized OT are bounded in a Sobolev (RKHS) ball independent of the two measures and \\emph{(iii)} provide the first sample complexity bound for SDs, obtained,by reformulating SDs as a maximization problem in a RKHS. We thus obtain a scaling in $1/\\sqrt{n}$ (as in MMD), with a constant that depends however on $\\varepsilon$, making the bridge between OT and MMD complete."
                },
                {
                    "title": "Auxiliary information: the raking-ratio empirical process",
                    "abstract": "We study the empirical measure associated to a sample of size $n$ and modified by $N$ iterations of the raking-ratio method. The empirical measure is adjusted to match the true probability of sets in a finite partition which changes each step. We establish asymptotic properties of the raking-ratio empirical process indexed by functions as $n\\rightarrow +\\infty$, for $N$ fixed. A closed-form expression of the limiting covariance matrices is derived as $N\\rightarrow +\\infty$. The nonasymptotic Gaussian approximation we use also yields uniform Berry-Esseen type bounds in $n, N$ and sharp estimates of the uniform quadratic risk reduction. In the two-way contingency table formulas characterizing the limiting process are very simple."
                },
                {
                    "title": "Wasserstein Distance Guided Representation Learning for Domain Adaptation",
                    "abstract": "\n \n Domain adaptation aims at generalizing a high-performance learner on a target domain via utilizing the knowledge distilled from a source domain which has a different but related data distribution. One solution to domain adaptation is to learn domain invariant feature representations while the learned representations should also be discriminative in prediction. To learn such representations, domain adaptation frameworks usually include a domain invariant representation learning approach to measure and reduce the domain discrepancy, as well as a discriminator for classification. Inspired by Wasserstein GAN, in this paper we propose a novel approach to learn domain invariant feature representations, namely Wasserstein Distance Guided Representation Learning (WDGRL). WDGRL utilizes a neural network, denoted by the domain critic, to estimate empirical Wasserstein distance between the source and target samples and optimizes the feature extractor network to minimize the estimated Wasserstein distance in an adversarial manner. The theoretical advantages of Wasserstein distance for domain adaptation lie in its gradient property and promising generalization bound. Empirical studies on common sentiment and image classification adaptation datasets demonstrate that our proposed WDGRL outperforms the state-of-the-art domain invariant representation learning approaches.\n \n"
                },
                {
                    "title": "Joint distribution optimal transportation for domain adaptation",
                    "abstract": "This paper deals with the unsupervised domain adaptation problem, where one wants to estimate a prediction function $f$ in a given target domain without any labeled sample by exploiting the knowledge available from a source domain where labels are known. Our work makes the following assumption: there exists a non-linear transformation between the joint feature/label space distributions of the two domain $\\mathcal{P}_s$ and $\\mathcal{P}_t$. We propose a solution of this problem with optimal transport, that allows to recover an estimated target $\\mathcal{P}^f_t=(X,f(X))$ by optimizing simultaneously the optimal coupling and $f$. We show that our method corresponds to the minimization of a bound on the target error, and provide an efficient algorithmic solution, for which convergence is proved. The versatility of our approach, both in terms of class of hypothesis or loss functions is demonstrated with real world classification and regression problems, for which we reach or surpass state-of-the-art results."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Efficient and Adaptive Estimation for Semiparametric Models",
                    "abstract": "Introduction.- Asymptotic Inference for (Finite-Dimensional) Parametric Models.- Information Bounds for Euclidean Parameters in Infinite-Dimensional Models.- Euclidean Parameters: Further Examples.- Information Bounds for Infinite-Dimensional Parameters.- Infinite-Dimensional Parameters: Further Examples: Construction of Examples."
                },
                {
                    "title": "Entropy-Maximizing and the Iterative Proportional Fitting Procedure",
                    "abstract": "In a recent paper, Wong outlined the benefits of the IPF procedure, which he claimed had received little attention from geographers. This follow-up paper introduces a large literature, ignored by Wong, which uses that procedure, but within the context of another mathematical model. Its application is illustrated with new analyses of recent voting patterns in Great Britain."
                },
                {
                    "title": "Contingency tables with given marginals.",
                    "abstract": "SUMMARY In its simplest formulation the problem considered is to estimate the cell probabilities pij Of an r x c contingency table for which the marginal probabilities pi and p j are known and fixed, so as to minimize E2pij In (Pi/r1ij), where rij are the corresponding entries in a given contingency table. An iterative procedure is given for determining the estimates and it is shown that the estimates are BAN, and that the iterative procedure is convergent. A summary of results for a four-way contingency table is given. An illustrative example is given."
                },
                {
                    "title": "Introduction to Entropic Optimal Transport",
                    "abstract": "This text develops mathematical foundations for entropic optimal transport and Sinkhorn\u2019s algorithm in a self-contained yet general way. It is a revised version of lecture notes from a course given in Paris during the fall of 2021; some parts date back to an earlier course at Columbia University in 2020."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Theory of Sample Surveys",
                    "abstract": "Introduction The Mathematics of Random Sampling Designs Distributions Induced by Random Sampling Designs Design-Based Estimation for General Finite Population Quantities Inference for Descriptive Parameters Analytic uses of Survey Data Sampling Strategies in Time and Space"
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG",
                "math.ST",
                "stat.TH"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we better understand and optimize the training processes of self-supervised learning (SSL) models, particularly in relation to their deviations from standard empirical risk minimization (ERM) techniques?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it could lead to a deeper understanding of SSL methodologies, which are becoming increasingly prevalent in machine learning. By clarifying the optimization processes in SSL, we can improve model performance, enhance generalization capabilities, and facilitate the development of more efficient training algorithms. This advancement could also lead to practical applications in various domains, such as computer vision and natural language processing, where labeled data is scarce or expensive to obtain.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem stem from the complex and often opaque nature of SSL training processes. Unlike traditional supervised learning, SSL incorporates various innovative techniques that complicate the optimization landscape. Naive approaches may fail because they do not account for the unique interactions between the model's architecture and the pseudo-tasks it learns from unlabelled data. Additionally, the lack of clear theoretical frameworks to guide the optimization of SSL models presents significant obstacles, making it difficult to ascertain what is being optimized and how to effectively tune these models.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on developing SSL techniques without fully investigating the underlying optimization mechanisms. Many existing solutions lack a comprehensive theoretical foundation, which has hindered progress in understanding the nuances of SSL training. Barriers such as the complexity of the models, the diversity of training strategies, and the absence of standardized evaluation metrics have contributed to this gap. My approach aims to systematically analyze and clarify these optimization processes, providing a more structured understanding that builds upon and improves prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nMy proposed methodology involves a detailed analysis of various SSL training algorithms, focusing on their optimization strategies and the empirical risk minimization framework. I will utilize benchmark datasets commonly used in SSL research, such as ImageNet and COCO, and evaluate model performance using metrics like accuracy, representation quality, and convergence speed. The expected outcomes include a clearer understanding of the optimization dynamics in SSL, identification of best practices for training, and potentially new algorithms that enhance the efficiency and effectiveness of self-supervised learning models."
            }
        },
        "author_data": {
            "5803d703-95a5-4d92-ac02-434772ea9160": {
                "pk": "5803d703-95a5-4d92-ac02-434772ea9160",
                "name": "Lang Liu",
                "collaborators": [
                    "Zaid Harchaoui",
                    "Yanlong Lin",
                    "Yongle Yu",
                    "Soumik Pal",
                    "Carlos Cinelli",
                    "Yuhang Gao",
                    "Tao Yan",
                    "Joseph Salmon",
                    "Sang Pyo Kim"
                ],
                "domain": [
                    "Nuclear Physics",
                    "Statistical Inference",
                    "Machine Learning",
                    "Black Hole Physics"
                ],
                "publications": [
                    {
                        "title": "No-Core Monte Carlo Shell Model Calculations with Unitary Correlation Operator Method and Similarity Renormalization Group",
                        "abstract": "The unitary correlation operator method (UCOM) and the similarity renormalization group theory (SRG) are compared and discussed in the framework of no-core Monte Carlo shell model (MCSM) calculations for $^{3}$H and $^{4}$He. The treatment of spurious center-of-mass motion by Lawson's prescription is performed in the MCSM calculations. These results with both transformed interactions show good suppression of spurious center-of-mass motion with proper Lawson's prescription parameter $\\beta_{\\rm c.m.}$ values. The UCOM potentials obtains faster convergence of total energy for the ground state than that of SRG potentials in the MCSM calculations, which differs from the cases in the no-core shell model calculations (NCSM). This differences are discussed and analyzed in terms of the truncation scheme in the MCSM and NCSM, as well as the properties of potentials of SRG and UCOM."
                    },
                    {
                        "title": "Expanding proton dripline by employing a number of muons",
                        "abstract": "Through mean-field calculations, we demonstrate that, in a large $Z$ nucleus binding multiple muons, these heavy leptons localize within a few dozen femtometers of the nucleus. The mutual Coulomb interactions between the muons and the protons can lead to a substantial decrease in proton chemical potential, surpassing 1 MeV. These findings allow for expanding the proton-dripline on the nuclear chart in principle, suggesting the possible production of nuclei with Z around 120."
                    },
                    {
                        "title": "Confidence Sets under Generalized Self-Concordance",
                        "abstract": "This paper revisits a fundamental problem in statistical inference from a non-asymptotic theoretical viewpoint $\\unicode{x2013}$ the construction of confidence sets. We establish a finite-sample bound for the estimator, characterizing its asymptotic behavior in a non-asymptotic fashion. An important feature of our bound is that its dimension dependency is captured by the effective dimension $\\unicode{x2013}$ the trace of the limiting sandwich covariance $\\unicode{x2013}$ which can be much smaller than the parameter dimension in some regimes. We then illustrate how the bound can be used to obtain a confidence set whose shape is adapted to the optimization landscape induced by the loss function. Unlike previous works that rely heavily on the strong convexity of the loss function, we only assume the Hessian is lower bounded at optimum and allow it to gradually becomes degenerate. This property is formalized by the notion of generalized self-concordance which originated from convex optimization. Moreover, we demonstrate how the effective dimension can be estimated from data and characterize its estimation accuracy. We apply our results to maximum likelihood estimation with generalized linear models, score matching with exponential families, and hypothesis testing with Rao's score test."
                    },
                    {
                        "title": "Entropy Regularized Optimal Transport Independence Criterion",
                        "abstract": "We introduce an independence criterion based on entropy regularized optimal transport. Our criterion can be used to test for independence between two samples. We establish non-asymptotic bounds for our test statistic and study its statistical behavior under both the null hypothesis and the alternative hypothesis. The theoretical results involve tools from U-process theory and optimal transport theory. We also offer a random feature type approximation for large-scale problems, as well as a differentiable program implementation for deep learning applications. We present experimental results on existing benchmarks for independence testing, illustrating the interest of the proposed criterion to capture both linear and nonlinear dependencies in synthetic data and real data."
                    },
                    {
                        "title": "Orthogonal Statistical Learning with Self-Concordant Loss",
                        "abstract": "Orthogonal statistical learning and double machine learning have emerged as general frameworks for two-stage statistical prediction in the presence of a nuisance component. We establish non-asymptotic bounds on the excess risk of orthogonal statistical learning methods with a loss function satisfying a self-concordance property. Our bounds improve upon existing bounds by a dimension factor while lifting the assumption of strong convexity. We illustrate the results with examples from multiple treatment effect estimation and generalized partially linear modeling."
                    },
                    {
                        "title": "Classification of pairing phase transition in the hot nucleus",
                        "abstract": "The hot nucleus $^{162}\\mathrm{Dy}$ is investigated using covariant density functional theory, where the shell-model-like approach treats the pairing correlation. Lee-Yang's theorem is applied to classify the pairing phase transition by analyzing the distribution of zeros of the partition function in the complex temperature plane. The distribution of zeros of the partition function converges with increasing particle numbers and illustrates the characteristics of the phase transition. In our calculations, we determine the first order of the phase transition near the critical temperature. Different seniority states show the pairing phase transition from a superfluid to a normal phase, ranging from fully paired states to completely unpaired states."
                    },
                    {
                        "title": "Thermodynamics of pairing transition for Odd-A nuclei",
                        "abstract": "The hot nucleus $^{171}$Yb is investigated by the covariant density functional theory with the PC-PK1 effective interaction. The thermodynamic quantities are evaluated with the canonical ensemble theory. The pairing correlations is treated by the shell-model-like approach, in which the particle numbers are conserved strictly and in which the blocking effect is handled exactly. An S-shaped heat capacity versus temperature of $^{171}$Yb appears. It has been studied in terms of the blocking effect, the single-particle levels, the pairing gap, and defined seniority components, and compared to the heat capacity of $^{172}$Yb. The pairing transition from the superfluid state to the normal state can result in the S-shaped heat capacity of $^{172}$Yb where the one-pair-broken and two-pair-broken states dominate, while the single-particle level structure near the Fermi surface is associated with the S-shaped heat capacity of $^{171}$Yb. For odd-A nuclei, although the one-pair-broken and two-pair-broken states still contribute, the pairing gap and the pairing transition is relatively weak. The S-shaped heat capacity could be affected due to the blocking of the single-particle level near the Fermi surface."
                    },
                    {
                        "title": "Score-Based Change Detection for Gradient-Based Learning Machines",
                        "abstract": "The widespread use of machine learning algorithms calls for automatic change detection algorithms to monitor their behavior over time. As a machine learning algorithm learns from a continuous, possibly evolving, stream of data, it is desirable and often critical to supplement it with a companion change detection algorithm to facilitate its monitoring and control. We present a generic score-based change detection method that can detect a change in any number of components of a machine learning model trained via empirical risk minimization. This proposed statistical hypothesis test can be readily implemented for such models designed within a differentiable programming framework. We establish the consistency of the hypothesis test and show how to calibrate it to achieve a prescribed false alarm rate. We illustrate the versatility of the approach on synthetic and real data."
                    },
                    {
                        "title": "Gravitational and electromagnetic radiations from binary black holes with electric and magnetic charges",
                        "abstract": "The Einstein-Maxwell theory has black hole solutions with electric and magnetic charges. In the standard model for particle physics, dyons with electric and magnetic charges would have been formed in the early universe. We derive the equations of motion of black hole binaries with electric and magnetic charges and explore some features of static orbits. We calculate the total emission rates of energy and angular momentum due to gravitational and electromagnetic radiations from dyonic binary black holes in different cases. It is shown that the emission rates of energy and angular momentum due to gravitational and electromagnetic radiations have the same dependence on the conic angle for different orbits. Moreover, we obtain the evolutions of orbits and find that a circular orbit remains circular while an elliptic orbit becomes quasi-circular due to electromagnetic and gravitational radiations. Our results provide rich information about black hole binaries with electric and magnetic charges and can be used to test black holes with magnetic charges."
                    }
                ]
            },
            "aa87eb99-a001-46e4-aec2-439ee8e08fa9": {
                "pk": "aa87eb99-a001-46e4-aec2-439ee8e08fa9",
                "name": "Ronak Mehta",
                "collaborators": [
                    "Vikas Singh",
                    "Zaid Harchaoui",
                    "Cencheng Shen",
                    "Joshua T. Vogelstein",
                    "Vincent Roulet",
                    "Krishna Pillutla",
                    "Sathya N. Ravi",
                    "Hyunwoo J. Kim",
                    "Sterling C. Johnson",
                    "Jelena Diakonikolas"
                ],
                "domain": [
                    "Robust Optimization",
                    "Machine Learning",
                    "Neural Networks",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "A Primal-Dual Algorithm for Faster Distributionally Robust Optimization",
                        "abstract": "We consider the penalized distributionally robust optimization (DRO) problem with a closed, convex uncertainty set, a setting that encompasses the $f$-DRO, Wasserstein-DRO, and spectral/$L$-risk formulations used in practice. We present Drago, a stochastic primal-dual algorithm that achieves a state-of-the-art linear convergence rate on strongly convex-strongly concave DRO problems. The method combines both randomized and cyclic components with mini-batching, which effectively handles the unique asymmetric nature of the primal and dual problems in DRO. We support our theoretical results with numerical benchmarks in classification and regression."
                    },
                    {
                        "title": "Distributionally Robust Optimization with Bias and Variance Reduction",
                        "abstract": "We consider the distributionally robust optimization (DRO) problem with spectral risk-based uncertainty set and $f$-divergence penalty. This formulation includes common risk-sensitive learning objectives such as regularized condition value-at-risk (CVaR) and average top-$k$ loss. We present Prospect, a stochastic gradient-based algorithm that only requires tuning a single learning rate hyperparameter, and prove that it enjoys linear convergence for smooth regularized losses. This contrasts with previous algorithms that either require tuning multiple hyperparameters or potentially fail to converge due to biased gradient estimates or inadequate regularization. Empirically, we show that Prospect can converge 2-3$\\times$ faster than baselines such as stochastic gradient and stochastic saddle-point methods on distribution shift and fairness benchmarks spanning tabular, vision, and language domains."
                    },
                    {
                        "title": "Robust Blind Deconvolution via Mirror Descent",
                        "abstract": "We revisit the Blind Deconvolution problem with a focus on understanding its robustness and convergence properties. Provable robustness to noise and other perturbations is receiving recent interest in vision, from obtaining immunity to adversarial attacks to assessing and describing failure modes of algorithms in mission critical applications. Further, many blind deconvolution methods based on deep architectures internally make use of or optimize the basic formulation, so a clearer understanding of how this sub-module behaves, when it can be solved, and what noise injection it can tolerate is a first order requirement. We derive new insights into the theoretical underpinnings of blind deconvolution. The algorithm that emerges has nice convergence guarantees and is provably robust in a sense we formalize in the paper. Interestingly, these technical results play out very well in practice, where on standard datasets our algorithm yields results competitive with or superior to the state of the art. Keywords: blind deconvolution, robust continuous optimization"
                    },
                    {
                        "title": "Resource Constrained Neural Network Architecture Search: Will a Submodularity Assumption Help?",
                        "abstract": "The design of neural network architectures is frequently either based on human expertise using trial/error and empirical feedback or tackled via large scale reinforcement learning strategies performed over distinct discrete architecture choices. In the latter case, the optimization is often non-differentiable and also not very amenable to derivative-free optimization methods. Most methods in use today require sizable computational resources. And if we want networks that additionally satisfy resource constraints, the above challenges are exacerbated because the search must now balance accuracy with certain budget constraints on resources. We formulate this problem as the optimization of a set function -- we find that the empirical behavior of this set function often (but not always) satisfies marginal gain and monotonicity principles -- properties central to the idea of submodularity. Based on this observation, we adapt algorithms within discrete optimization to obtain heuristic schemes for neural network architecture search, where we have resource constraints on the architecture. This simple scheme when applied on CIFAR-100 and ImageNet, identifies resource-constrained architectures with quantifiably better performance than current state-of-the-art models designed for mobile devices. Specifically, we find high-performing architectures with fewer parameters and computations by a search method that is much faster."
                    },
                    {
                        "title": "Stochastic Optimization for Spectral Risk Measures",
                        "abstract": "Spectral risk objectives - also called $L$-risks - allow for learning systems to interpolate between optimizing average-case performance (as in empirical risk minimization) and worst-case performance on a task. We develop stochastic algorithms to optimize these quantities by characterizing their subdifferential and addressing challenges such as biasedness of subgradient estimates and non-smoothness of the objective. We show theoretically and experimentally that out-of-the-box approaches such as stochastic subgradient and dual averaging are hindered by bias and that our approach outperforms them."
                    },
                    {
                        "title": "Deep Unlearning via Randomized Conditionally Independent Hessians",
                        "abstract": "Recent legislation has led to interest in machine unlearning, i.e., removing specific training samples from a predictive model as if they never existed in the training dataset. Unlearning may also be required due to corrupted/adversarial data or simply a user's updated privacy requirement. For models which require no training (k-NN), simply deleting the closest original sample can be effective. But this idea is inapplicable to models which learn richer representations. Recent ideas leveraging optimization-based updates scale poorly with the model dimension d, due to inverting the Hessian of the loss function. We use a variant of a new conditional independence coefficient, L-CODEC, to identify a subset of the model parameters with the most semantic overlap on an individual sample level. Our approach completely avoids the need to invert a (possibly) huge matrix. By utilizing a Markov blanket selection, we premise that L-CODEC is also suitable for deep unlearning, as well as other applications in vision. Compared to alternatives, L-CODEC makes approximate unlearning possible in settings that would otherwise be infeasible, including vision models used for face recognition, person re-identification and NLP models that may require unlearning samples identified for exclusion. Code can be found at https://github.com/vsingh-group/LCODEC-deep-unlearning/"
                    },
                    {
                        "title": "Independence Testing for Temporal Data",
                        "abstract": "Temporal data are increasingly prevalent in modern data science. A fundamental question is whether two time series are related or not. Existing approaches often have limitations, such as relying on parametric assumptions, detecting only linear associations, and requiring multiple tests and corrections. While many non-parametric and universally consistent dependence measures have recently been proposed, directly applying them to temporal data can inflate the p-value and result in an invalid test. To address these challenges, this paper introduces the temporal dependence statistic with block permutation to test independence between temporal data. Under proper assumptions, the proposed procedure is asymptotically valid and universally consistent for testing independence between stationary time series, and capable of estimating the optimal dependence lag that maximizes the dependence. Moreover, it is compatible with a rich family of distance and kernel based dependence measures, eliminates the need for multiple testing, and exhibits excellent testing power in various simulation settings."
                    },
                    {
                        "title": "Sampling-free Uncertainty Estimation in Gated Recurrent Units with Exponential Families",
                        "abstract": "There has recently been a concerted effort to derive mechanisms in vision and machine learning systems to offer uncertainty estimates of the predictions they make. Clearly, there are enormous benefits to a system that is not only accurate but also has a sense for when it is not sure. Existing proposals center around Bayesian interpretations of modern deep architectures -- these are effective but can often be computationally demanding. We show how classical ideas in the literature on exponential families on probabilistic networks provide an excellent starting point to derive uncertainty estimates in Gated Recurrent Units (GRU). Our proposal directly quantifies uncertainty deterministically, without the need for costly sampling-based estimation. We demonstrate how our model can be used to quantitatively and qualitatively measure uncertainty in unsupervised image sequence prediction. To our knowledge, this is the first result describing sampling-free uncertainty estimation for powerful sequential models such as GRUs."
                    },
                    {
                        "title": "Finding Differentially Covarying Needles in a Temporally Evolving Haystack: A Scan Statistics Perspective",
                        "abstract": "Recent results in coupled or temporal graphical models offer schemes for estimating the relationship structure between features when the data come from related (but distinct) longitudinal sources. A novel application of these ideas is for analyzing group-level differences, i.e., in identifying if trends of estimated objects (e.g., covariance or precision matrices) are different across disparate conditions (e.g., gender or disease). Often, poor effect sizes make detecting the differential signal over the full set of features difficult: for example, dependencies between only a subset of features may manifest differently across groups. In this work, we first give a parametric model for estimating trends in the space of SPD matrices as a function of one or more covariates. We then generalize scan statistics to graph structures, to search over distinct subsets of features (graph partitions) whose temporal dependency structure may show statistically significant group-wise differences. We theoretically analyze the Family Wise Error Rate (FWER) and bounds on Type 1 and Type 2 error. On a cohort of individuals with risk factors for Alzheimer's disease (but otherwise cognitively healthy), we find scientifically interesting group differences where the default analysis, i.e., models estimated on the full graph, do not survive reasonable significance thresholds."
                    },
                    {
                        "title": "hyppo: A Multivariate Hypothesis Testing Python Package",
                        "abstract": "We introduce hyppo, a unified library for performing multivariate hypothesis testing, including independence, two-sample, and k-sample testing. While many multivariate independence tests have R packages available, the interfaces are inconsistent and most are not available in Python. hyppo includes many state of the art multivariate testing procedures. The package is easy-to-use and is flexible enough to enable future extensions. The documentation and all releases are available at https://hyppo.neurodata.io."
                    },
                    {
                        "title": "Random Forests for Adaptive Nearest Neighbor Estimation of Information-Theoretic Quantities",
                        "abstract": "Information-theoretic quantities, such as conditional entropy and mutual information, are critical data summaries for quantifying uncertainty. Current widely used approaches for computing such quantities rely on nearest neighbor methods and exhibit both strong performance and theoretical guarantees in certain simple scenarios. However, existing approaches fail in high-dimensional settings and when different features are measured on different scales.We propose decision forest-based adaptive nearest neighbor estimators and show that they are able to effectively estimate posterior probabilities, conditional entropies, and mutual information even in the aforementioned settings.We provide an extensive study of efficacy for classification and posterior probability estimation, and prove certain forest-based approaches to be consistent estimators of the true posteriors and derived information-theoretic quantities under certain assumptions. In a real-world connectome application, we quantify the uncertainty about neuron type given various cellular features in the Drosophila larva mushroom body, a key challenge for modern neuroscience."
                    },
                    {
                        "title": "DUAL-GLOW: Conditional Flow-Based Generative Model for Modality Transfer",
                        "abstract": "Positron emission tomography (PET) imaging is an imaging modality for diagnosing a number of neurological diseases. In contrast to Magnetic Resonance Imaging (MRI), PET is costly and involves injecting a radioactive substance into the patient. Motivated by developments in modality transfer in vision, we study the generation of certain types of PET images from MRI data. We derive new flow-based generative models which we show perform well in this small sample size regime (much smaller than dataset sizes available in standard vision tasks). Our formulation, DUAL-GLOW, is based on two invertible networks and a relation network that maps the latent spaces to each other. We discuss how given the prior distribution, learning the conditional distribution of PET given the MRI image reduces to obtaining the conditional distribution between the two latent codes w.r.t. the two image types. We also extend our framework to leverage 'side' information (or attributes) when available. By controlling the PET generation through 'conditioning' on age, our model is also able to capture brain FDG-PET (hypometabolism) changes, as a function of age. We present experiments on the Alzheimers Disease Neuroimaging Initiative (ADNI) dataset with 826 subjects, and obtain good performance in PET image synthesis, qualitatively and quantitatively better than recent works."
                    }
                ]
            },
            "c5fa5526-7b8b-4cfc-bd7d-e49fc9ad3e5c": {
                "pk": "c5fa5526-7b8b-4cfc-bd7d-e49fc9ad3e5c",
                "name": "Soumik Pal",
                "collaborators": [
                    "Krzysztof Burdzy",
                    "Sourav Chatterjee",
                    "Jim Pitman",
                    "Philip Protter",
                    "Persi Diaconis"
                ],
                "domain": [
                    "Stochastic Processes",
                    "Optimal Transport",
                    "Financial Mathematics",
                    "Probability Theory"
                ],
                "publications": [
                    {
                        "title": "A passage to the Poisson-Dirichlet through the Bessel square processes",
                        "abstract": "This paper has been withdrawn by the author."
                    },
                    {
                        "title": "Computing strategies for achieving acceptability",
                        "abstract": "We consider a trader who wants to direct his portfolio towards a set of acceptable wealths given by a convex risk measure. We propose a black-box algorithm, whose inputs are the joint law of stock prices and the convex risk measure, and whose outputs are the numerical values of initial capital requirement and the functional form of a trading strategy to achieve acceptability. We also prove optimality of the obtained capital."
                    },
                    {
                        "title": "On a conjectured sharpness principle for probabilistic forecasting with calibration",
                        "abstract": "This note proves a weak type of the sharpness principle as conjectured by Gneiting, Balabdaoui, and Raftery in 2007 in connection with probabilistic forecasting subject to calibration constraints. A strong version of such a principle still awaits a proper formulation."
                    },
                    {
                        "title": "Symmetrization of Bernoulli",
                        "abstract": "Let X be a random variable. We shall call an independent random variable Y to be a symmetrizer for X, if X+Y is symmetric around zero. A random variable is said to be symmetry resistant if the variance of any symmetrizer Y, is never smaller than the variance of X itself. We prove that a Bernoulli(p) random variable is symmetry resistant if and only if p is not 1/2. This is an old problem proved in 1999 by Kagan, Mallows, Shepp, Vanderbei & Vardi using linear programming principles. We reprove it here using completely probabilistic tools using Skorokhod embedding and Ito's rule."
                    },
                    {
                        "title": "Concentration for multidimensional diffusions and their boundary local times",
                        "abstract": "We prove that probability laws of certain multidimensional semimartingales which includes time-inhomogenous diffusions, under suitable assumptions, satisfy Quadratic Transportation Cost Inequality under the uniform metric. From this we derive concentration properties of Lipschitz functions of process paths that depend on the entire history. In particular, we estimate concentration of boundary local time of reflected Brownian motions on a polyhedral domain. We work out explicit applications of consequences of measure concentration for the case of Brownian motion with rank-based drifts."
                    },
                    {
                        "title": "On the difference between entropic cost and the optimal transport cost",
                        "abstract": "Consider the Monge-Kantorovich problem of transporting densities $\\rho_0$ to $\\rho_1$ on $\\mathbb{R}^d$ with a strictly convex cost function. A popular relaxation of the problem is the one-parameter family called the entropic cost problem. The entropic cost $K_h$, $h>0$, is significantly faster to compute and $h K_h$ is known to converge to the optimal transport cost as $h$ goes to zero. We are interested the rate of convergence. We show that the difference between $K_h$ and $1/h$ times the optimal cost of transport has a pointwise limit when transporting a compactly supported density to another that satisfies a few other technical restrictions. This limit is the relative entropy of $\\rho_1$ with respect to a Riemannian volume measure on $\\mathbb{R}^d$ that measures the local sensitivity of the transport map. For the quadratic Wasserstein transport, this relative entropy is exactly one half of the difference of entropies of $\\rho_1$ and $\\rho_0$. In that case we complement the results of Adams et al., Duong et al, and Erbar et al. who all use gamma convergence. More surprisingly, we demonstrate that this difference of two entropies (plus the cost) is also the limit for the Dirichlet transport introduced recently by Pal and Wong. The latter can be thought of as a multiplicative analog of the Wasserstein transport and corresponds to a non-local operator. It hints at an underlying gradient flow of entropy, in the sense of Jordan-Kinderlehrer-Otto, even when the cost function is not a metric. The proofs are based on Gaussian approximations to Schr\\\"odinger bridges as $h$ approaches zero."
                    },
                    {
                        "title": "Capital Requirement for Achieving Acceptability",
                        "abstract": "Consider an agent who enters a financial market on day t = 0 with an initial capital amount x. He invests this amount on stocks and the money market, and by day t = T, has generated a wealth W . He is given a convex class of probability measures (called scenarios) and a real-valued function (or floors) corresponding to each scenario. The agent faces the constraints that the expectation of W under each scenario must not be less than the corresponding floor. We call x acceptable if one can start with x and successfully generate W satisfying these constraints. The set of acceptable x is a half-line in R, unbounded from above. We show that under some regularity conditions on the set of scenarios and the floor function, the infimum of this set is given by the supremum of the floors over all scenarios under which S is a martingale."
                    },
                    {
                        "title": "Analysis of market weights under volatility-stabilized market models",
                        "abstract": "We derive the joint density of market weights, at fixed times and suitable stopping times, of the volatility-stabilized market models introduced by Fernholz and Karatzas in [Ann. Finan. 1 (2005) 149-177]. The argument rests on computing the exit density of a collection of independent Bessel-square processes of possibly different dimensions from the unit simplex. We show that the law of the market weights is the same as that of the multi-allele Wright-Fisher diffusion model, well known in population genetics. Thus, as a side result, we furnish a novel proof of the transition density function of the Wright-Fisher model which was originally derived by Griffiths by bi-orthogonal series expansion."
                    },
                    {
                        "title": "On the Aldous diffusion on Continuum Trees. I",
                        "abstract": "Consider a Markov chain on the space of rooted real binary trees that randomly removes leaves and reinserts them on a random edge and suitably rescales the lengths of edges. This chain was introduced by David Aldous who conjectured a diffusion limit of this chain, as the size of the tree grows, on the space of continuum trees. We prove the existence of a process on continuum trees, which via a random time change, displays properties one would expect from the conjectured Aldous diffusion. The existence of our process is proved by considering an explicit scaled limit of a Poissonized version of the Aldous Markov chain running on finite trees. The analysis involves taking limit of a sequence of splitting trees whose age processes converge but the contour process does not. Several formulas about the limiting process are derived."
                    },
                    {
                        "title": "Brownian approximation to counting graphs",
                        "abstract": "Let C(n,k) denote the number of connected graphs with n labeled vertices and n+k-1 edges. For any sequence (k_n), the limit of C(n,k_n) as n tends to infinity is known. It has been observed that, if k_n=o(\\sqrt{n}), this limit is asymptotically equal to the $k_n$th moment of the area under the standard Brownian excursion. These moments have been computed in the literature via independent methods. In this article we show why this is true for k_n=o(\\sqrt[3]{n}) starting from an observation made by Joel Spencer. The elementary argument uses a result about strong embedding of the Uniform empirical process in the Brownian bridge proved by Komlos, Major, and Tusnady."
                    },
                    {
                        "title": "Embedding optimal transports in statistical manifolds",
                        "abstract": "We consider Monge-Kantorovich optimal transport problems on $\\mathbb{R}^d$, $d\\ge 1$, with a convex cost function given by the cumulant generating function of a probability measure. Examples include the Wasserstein-2 transport whose cost function is the square of the Euclidean distance and corresponds to the cumulant generating function of the multivariate standard normal distribution. The optimal coupling is usually described via an extended notion of convex/concave functions and their gradient maps. These extended notions are nonintuitive and do not satisfy useful inequalities such as Jensen's inequality. Under mild regularity conditions, we show that all such extended gradient maps can be recovered as the usual supergradients of a nonnegative concave function on the space of probability distributions. This embedding provides a universal geometry for all such optimal transports and an unexpected connection with information geometry of exponential families of distributions."
                    },
                    {
                        "title": "Limiting partition function for the Mallows model: a conjecture and partial evidence",
                        "abstract": "Let $S_n$ denote the set of permutations of $n$ labels. We consider a class of Gibbs probability models on $S_n$ that is a subfamily of the so-called Mallows model of random permutations. The Gibbs energy is given by a class of right invariant divergences on $S_n$ that includes common choices such as the Spearman foot rule and the Spearman rank correlation. Mukherjee in 2016 computed the limit of the (scaled) log partition function (i.e. normalizing factor) of such models as $n\\rightarrow \\infty$. Our objective is to compute the exact limit, as $n\\rightarrow \\infty$, without the log. We conjecture that this limit is given by the Fredholm determinant of an integral operator related to the so-called Schr\\\"odinger bridge probability distributions from optimal transport theory. We provide partial evidence for this conjecture, although the argument lacks a final error bound that is needed for it to become a complete proof."
                    },
                    {
                        "title": "Wright-Fisher diffusion with negative mutation rates",
                        "abstract": "We study a family of n-dimensional diffusions, taking values in the unit simplex of vectors with nonnegative coordinates that add up to one. These processes satisfy stochastic differential equations which are similar to the ones for the classical Wright-Fisher diffusions, except that the \"mutation rates\" are now nonpositive. This model, suggested by Aldous, appears in the study of a conjectured diffusion limit for a Markov chain on Cladograms. The striking feature of these models is that the boundary is not reflecting, and we kill the process once it hits the boundary. We derive the explicit exit distribution from the simplex and probabilistic bounds on the exit time. We also prove that these processes can be viewed as a \"stochastic time-reversal\" of a Wright-Fisher process of increasing dimensions and conditioned at a random time. A key idea in our proofs is a skew-product construction using certain one-dimensional diffusions called Bessel-square processes of negative dimensions, which have been recently introduced by Going-Jaeschke and Yor."
                    },
                    {
                        "title": "Exponentially concave functions and high dimensional stochastic portfolio theory",
                        "abstract": "We consider the following problem in stochastic portfolio theory. Are there portfolios that are relative arbitrages with respect to the market portfolio over very short periods of time under realistic assumptions? We answer a slightly relaxed question affirmative in the following high dimensional sense, where dimension refers to the number of stocks being traded. Very roughly, suppose that for every dimension we have a continuous semimartingale market such that (i) the vector of market weights in decreasing order has a stationary regularly varying tail with an index between $-1$ and $-1/2$ and (ii) zero is not a limit point of the relative volatilities of the stocks. Then, given a probability $\\eta < 1$ arbitrarily close to one, two arbitrarily small $\\epsilon, \\delta >0$, and an arbitrarily high positive amount $M$, for all high enough dimensions, it is possible to construct a functionally generated portfolio such that, with probability at least $\\eta$, its relative value with respect to the market at time $\\delta$ is at least $M$, and never goes below $(1-\\epsilon)$ during $[0, \\delta]$. There are two phase transitions; if the index of the tail is less than $-1$ or larger than $-1/2$. The construction uses properties of regular variation, high-dimensional convex geometry and concentration of measure under Dirichlet distributions. We crucially use the notion of $(K,N)$ convex functions introduced by Erbar, Kuwada, Sturm in the context of curvature-dimension conditions and Bochner's inequalities."
                    },
                    {
                        "title": "A phase transition behavior for Brownian motions interacting through their ranks",
                        "abstract": "Consider a time-varying collection of n points on the positive real axis, modeled as exponentials of n Brownian motions whose drift vector at every time point is determined by the relative ranks of the coordinate processes at that time. If at each time point we divide the points by their sum, under suitable assumptions the rescaled point process converges to a stationary distribution (depending on n and the vector of drifts) as time goes to infinity. This stationary distribution can be exactly computed using a recent result of Pal and Pitman. The model and the rescaled point process are both central objects of study in models of equity markets introduced by Banner, Fernholz, and Karatzas. In this paper, we look at the behavior of this point process under the stationary measure as $n$ tends to infinity. Under a certain `continuity at the edge' condition on the drifts, we show that one of the following must happen: either (i) all points converge to zero, or (ii) the maximum goes to one and the rest go to zero, or (iii) the processes converge in law to a non-trivial Poisson-Dirichlet distribution. The proof employs, among other things, techniques from Talagrand's analysis of the low temperature phase of Derrida's Random Energy Model of spin glasses. The main result establishes a universality property for the BFK models and aids in explicit asymptotic computations using known results about the Poisson-Dirichlet law."
                    },
                    {
                        "title": "Contradictory predictions",
                        "abstract": "We prove the sharp bound for the probability that two experts who have access to different information, represented by different $\\sigma$-fields, will give radically different estimates of the probability of an event. This is relevant when one combines predictions from various experts in a common probability space to obtain an aggregated forecast. Our proof is constructive in the sense that, not only the sharp bounds are proved, but also the optimizer is constructed via an explicit algorithm."
                    },
                    {
                        "title": "One-dimensional Brownian particle systems with rank dependent drifts",
                        "abstract": "We study interacting systems of linear Brownian motions whose drift vector at every time point is determined by the relative ranks of the coordinate processes at that time. Our main objective has been to study the long range behavior of the spacings between the Brownian motions arranged in increasing order. For finitely many Brownian motions interacting in this manner, we characterize drifts for which the family of laws of the vector of spacings is tight, and show its convergence to a unique stationary joint distribution given by independent exponential distributions with varying means. We also study one particular countably infinite system, where only the minimum Brownian particle gets a constant upward drift, and prove that independent and identically distributed exponential spacings remain stationary under the dynamics of such a process. Some related conjectures in this direction have also been discussed."
                    },
                    {
                        "title": "Analysis of continuous strict local martingales via h-transforms",
                        "abstract": "We study strict local martingales via h-transforms, a method which first appeared in Delbaen-Schachermayer. We show that strict local martingales arise whenever there is a consistent family of change of measures where the two measures are not equivalent to one another. Several old and new strict local martingales are identified. We treat examples of diffusions with various boundary behavior, size-bias sampling of diffusion paths, and non-colliding diffusions. A multidimensional generalization to conformal strict local martingales is achieved through Kelvin transform. As curious examples of non-standard behavior, we show by various examples that strict local martingales do not behave uniformly when the function (x-K)^+ is applied to them. Implications to the recent literature on financial bubbles are discussed."
                    },
                    {
                        "title": "Shuffling cards by spatial motion",
                        "abstract": "We propose a model of card shuffling where a pack of cards, spread as points on a square table, are repeatedly gathered locally at random spots and then spread towards a random direction. A shuffling of the cards is then obtained by arranging the cards by their increasing $x$-coordinate values. When there are $m$ cards on the table we show that this random ordering gets mixed in time $O\\left(\\log m\\right)$. Explicit constants are evaluated in a diffusion limit when the position of $m$ cards evolves as an interesting $2m$-dimensional non-reversible reflected jump diffusion in time. Our main technique involves the use of multidimensional Skorokhod maps for double reflections in $[0,1]^2$ in taking the discrete to continuous limit. The limiting computations are then based on the planar Brownian motion and properties of Bessel processes."
                    },
                    {
                        "title": "Floodings of metric graphs",
                        "abstract": "We consider random labelings of finite graphs conditioned on a small fixed number of peaks. We introduce a continuum framework where a combinatorial graph is associated with a metric graph and edges are identified with intervals. Next we consider a sequence of partitions of the edges of the metric graph with the partition size going to zero. As the mesh of the subdivision goes to zero, the conditioned random labelings converge, in a suitable sense, to a deterministic function which evolves as an increasing process of subsets of the metric graph that grows at rate one while maximizing an appropriate notion of entropy. We call such functions floodings. We present a number of qualitative and quantitative properties of floodings and some explicit examples."
                    }
                ]
            },
            "d7c5638d-8df4-44d3-9400-98de2ef0a9ea": {
                "pk": "d7c5638d-8df4-44d3-9400-98de2ef0a9ea",
                "name": "Zaid Harchaoui",
                "collaborators": [
                    "Vincent Roulet",
                    "Lang Liu",
                    "Meyer Scetbon",
                    "Yassine Laguel",
                    "Krishna Pillutla",
                    "J\u00e9r\u00f4me Malick",
                    "Francis Bach",
                    "Eric Moulines",
                    "Farida Enikeeva",
                    "Niao He"
                ],
                "domain": [
                    "Statistical Learning",
                    "Optimization",
                    "Machine Learning",
                    "Kernel Methods"
                ],
                "publications": [
                    {
                        "title": "Testing for Homogeneity with Kernel Fisher Discriminant Analysis",
                        "abstract": "We propose to investigate test statistics for testing homogeneity in reproducing kernel Hilbert spaces. Asymptotic null distributions under null hypothesis are derived, and consistency against fixed and local alternatives is assessed. Finally, experimental evidence of the performance of the proposed approach on both artificial data and a speaker verification task is provided."
                    },
                    {
                        "title": "High-dimensional change-point detection with sparse alternatives",
                        "abstract": "We consider the problem of detecting a change in mean in a sequence of Gaussian vectors. Under the alternative hypothesis, the change occurs only in some subset of the components of the vector. We propose a test of the presence of a change-point that is adaptive to the number of changing components. Under the assumption that the vector dimension tends to infinity and the length of the sequence grows slower than the dimension of the signal, we obtain the detection boundary for this problem and prove its rate-optimality."
                    },
                    {
                        "title": "A Spectral Analysis of Dot-product Kernels",
                        "abstract": "We present eigenvalue decay estimates of integral operators associated with compositional dot-product kernels. The estimates improve on previous ones established for power series kernels on spheres. This allows us to obtain the volumes of balls in the corresponding reproducing kernel Hilbert spaces. We discuss the consequences on statistical estimation with compositional dot product kernels and highlight interesting trade-offs between the approximation error and the statistical error depending on the number of compositions and the smoothness of the kernels."
                    },
                    {
                        "title": "Differentiable Programming \u00e0 la Moreau",
                        "abstract": "The notion of a Moreau envelope is central to the analysis of first-order optimization algorithms for machine learning. Yet, it has not been developed and extended to be applied to a deep network and, more broadly, to a machine learning system with a differentiable programming implementation. We define a compositional calculus adapted to Moreau envelopes and show how to integrate it within differentiable programming. The proposed framework casts in a mathematical optimization framework several variants of gradient back-propagation related to the idea of the propagation of virtual targets."
                    },
                    {
                        "title": "Harmonic Decompositions of Convolutional Networks",
                        "abstract": "We present a description of the function space and the smoothness class associated with a convolutional network using the machinery of reproducing kernel Hilbert spaces. We show that the mapping associated with a convolutional network expands into a sum involving elementary functions akin to spherical harmonics. This functional decomposition can be related to the functional ANOVA decomposition in nonparametric statistics. Building off our functional characterization of convolutional networks, we obtain statistical bounds highlighting an interesting trade-off between the approximation error and the estimation error."
                    },
                    {
                        "title": "Semi-proximal Mirror-Prox for Nonsmooth Composite Minimization",
                        "abstract": "We propose a new first-order optimisation algorithm to solve high-dimensional non-smooth composite minimisation problems. Typical examples of such problems have an objective that decomposes into a non-smooth empirical risk part and a non-smooth regularisation penalty. The proposed algorithm, called Semi-Proximal Mirror-Prox, leverages the Fenchel-type representation of one part of the objective while handling the other part of the objective via linear minimization over the domain. The algorithm stands in contrast with more classical proximal gradient algorithms with smoothing, which require the computation of proximal operators at each iteration and can therefore be impractical for high-dimensional problems. We establish the theoretical convergence rate of Semi-Proximal Mirror-Prox, which exhibits the optimal complexity bounds, i.e. $O(1/\\epsilon^2)$, for the number of calls to linear minimization oracle. We present promising experimental results showing the interest of the approach in comparison to competing methods."
                    },
                    {
                        "title": "A Statistical Investigation of Long Memory in Language and Music",
                        "abstract": "Representation and learning of long-range dependencies is a central challenge confronted in modern applications of machine learning to sequence data. Yet despite the prominence of this issue, the basic problem of measuring long-range dependence, either in a given data source or as represented in a trained deep model, remains largely limited to heuristic tools. We contribute a statistical framework for investigating long-range dependence in current applications of deep sequence modeling, drawing on the well-developed theory of long memory stochastic processes. This framework yields testable implications concerning the relationship between long memory in real-world data and its learned representation in a deep learning architecture, which are explored through a semiparametric framework adapted to the high-dimensional setting."
                    },
                    {
                        "title": "An Elementary Approach to Convergence Guarantees of Optimization Algorithms for Deep Networks",
                        "abstract": "We present an approach to obtain convergence guarantees of optimization algorithms for deep networks based on elementary arguments and computations. The convergence analysis revolves around the analytical and computational structures of optimization oracles central to the implementation of deep networks in machine learning software. We provide a systematic way to compute estimates of the smoothness constants that govern the convergence behavior of first-order optimization algorithms used to train deep networks. A diverse set of example components and architectures arising in modern deep networks intersperse the exposition to illustrate the approach."
                    },
                    {
                        "title": "Efficient First-Order Algorithms for Adaptive Signal Denoising",
                        "abstract": "We consider the problem of discrete-time signal denoising, focusing on a specific family of non-linear convolution-type estimators. Each such estimator is associated with a time-invariant filter which is obtained adaptively, by solving a certain convex optimization problem. Adaptive convolution-type estimators were demonstrated to have favorable statistical properties. However, the question of their computational complexity remains largely unexplored, and in fact we are not aware of any publicly available implementation of these estimators. Our first contribution is an efficient implementation of these estimators via some known first-order proximal algorithms. Our second contribution is a computational complexity analysis of the proposed procedures, which takes into account their statistical nature and the related notion of statistical accuracy. The proposed procedures and their analysis are illustrated on a simulated data benchmark."
                    },
                    {
                        "title": "Target Propagation via Regularized Inversion",
                        "abstract": "Target Propagation (TP) algorithms compute targets instead of gradients along neural networks and propagate them backward in a way that is similar yet different than gradient back-propagation (BP). The idea was first presented as a perturbative alternative to back-propagation that may achieve greater accuracy in gradient evaluation when training multi-layer neural networks (LeCun et al., 1989). However, TP has remained more of a template algorithm with many variations than a well-identified algorithm. Revisiting insights of LeCun et al., (1989) and more recently of Lee et al. (2015), we present a simple version of target propagation based on regularized inversion of network layers, easily implementable in a differentiable programming framework. We compare its computational complexity to the one of BP and delineate the regimes in which TP can be attractive compared to BP. We show how our TP can be used to train recurrent neural networks with long sequences on various sequence modeling problems. The experimental results underscore the importance of regularization in TP in practice."
                    },
                    {
                        "title": "Confidence Sets under Generalized Self-Concordance",
                        "abstract": "This paper revisits a fundamental problem in statistical inference from a non-asymptotic theoretical viewpoint $\\unicode{x2013}$ the construction of confidence sets. We establish a finite-sample bound for the estimator, characterizing its asymptotic behavior in a non-asymptotic fashion. An important feature of our bound is that its dimension dependency is captured by the effective dimension $\\unicode{x2013}$ the trace of the limiting sandwich covariance $\\unicode{x2013}$ which can be much smaller than the parameter dimension in some regimes. We then illustrate how the bound can be used to obtain a confidence set whose shape is adapted to the optimization landscape induced by the loss function. Unlike previous works that rely heavily on the strong convexity of the loss function, we only assume the Hessian is lower bounded at optimum and allow it to gradually becomes degenerate. This property is formalized by the notion of generalized self-concordance which originated from convex optimization. Moreover, we demonstrate how the effective dimension can be estimated from data and characterize its estimation accuracy. We apply our results to maximum likelihood estimation with generalized linear models, score matching with exponential families, and hypothesis testing with Rao's score test."
                    },
                    {
                        "title": "Structure-Blind Signal Recovery",
                        "abstract": "We consider the problem of recovering a signal observed in Gaussian noise. If the set of signals is convex and compact, and can be specified beforehand, one can use classical linear estimators that achieve a risk within a constant factor of the minimax risk. However, when the set is unspecified, designing an estimator that is blind to the hidden structure of the signal remains a challenging problem. We propose a new family of estimators to recover signals observed in Gaussian noise. Instead of specifying the set where the signal lives, we assume the existence of a well-performing linear estimator. Proposed estimators enjoy exact oracle inequalities and can be efficiently computed through convex optimization. We present several numerical illustrations that show the potential of the approach."
                    },
                    {
                        "title": "Kernel-based Translations of Convolutional Networks",
                        "abstract": "Convolutional Neural Networks, as most artificial neural networks, are commonly viewed as methods different in essence from kernel-based methods. We provide a systematic translation of Convolutional Neural Networks (ConvNets) into their kernel-based counterparts, Convolutional Kernel Networks (CKNs), and demonstrate that this perception is unfounded both formally and empirically. We show that, given a Convolutional Neural Network, we can design a corresponding Convolutional Kernel Network, easily trainable using a new stochastic gradient algorithm based on an accurate gradient computation, that performs on par with its Convolutional Neural Network counterpart. We present experimental results supporting our claims on landmark ConvNet architectures comparing each ConvNet to its CKN counterpart over several parameter settings."
                    },
                    {
                        "title": "Device Heterogeneity in Federated Learning: A Superquantile Approach",
                        "abstract": "We propose a federated learning framework to handle heterogeneous client devices which do not conform to the population data distribution. The approach hinges upon a parameterized superquantile-based objective, where the parameter ranges over levels of conformity. We present an optimization algorithm and establish its convergence to a stationary point. We show how to practically implement it using secure aggregation by interleaving iterations of the usual federated averaging method with device filtering. We conclude with numerical experiments on neural networks as well as linear models on tasks from computer vision and natural language processing."
                    },
                    {
                        "title": "Object Discovery in Videos as Foreground Motion Clustering",
                        "abstract": "We consider the problem of providing dense segmentation masks for object discovery in videos. We formulate the object discovery problem as foreground motion clustering, where the goal is to cluster foreground pixels in videos into different objects. We introduce a novel pixel-trajectory recurrent neural network that learns feature embeddings of foreground pixel trajectories linked across time. By clustering the pixel trajectories using the learned feature embeddings, our method establishes correspondences between foreground object masks across video frames. To demonstrate the effectiveness of our framework for object discovery, we conduct experiments on commonly used datasets for motion segmentation, where we achieve state-of-the-art performance."
                    },
                    {
                        "title": "Entropy Regularized Optimal Transport Independence Criterion",
                        "abstract": "We introduce an independence criterion based on entropy regularized optimal transport. Our criterion can be used to test for independence between two samples. We establish non-asymptotic bounds for our test statistic and study its statistical behavior under both the null hypothesis and the alternative hypothesis. The theoretical results involve tools from U-process theory and optimal transport theory. We also offer a random feature type approximation for large-scale problems, as well as a differentiable program implementation for deep learning applications. We present experimental results on existing benchmarks for independence testing, illustrating the interest of the proposed criterion to capture both linear and nonlinear dependencies in synthetic data and real data."
                    },
                    {
                        "title": "Superquantile-based learning: a direct approach using gradient-based optimization",
                        "abstract": "We consider a formulation of supervised learning that endows models with robustness to distributional shifts from training to testing. The formulation hinges upon the superquantile risk measure, also known as the conditional value-at-risk, which has shown promise in recent applications of machine learning and signal processing. We show that, thanks to a direct smoothing of the superquantile function, a superquantile-based learning objective is amenable to gradient-based optimization, using batch optimization algorithms such as gradient descent or quasi-Newton algorithms, or using stochastic optimization algorithms such as stochastic gradient algorithms. A companion software SPQR implements in Python the algorithms described and allows practitioners to experiment with superquantile-based supervised learning."
                    },
                    {
                        "title": "Orthogonal Statistical Learning with Self-Concordant Loss",
                        "abstract": "Orthogonal statistical learning and double machine learning have emerged as general frameworks for two-stage statistical prediction in the presence of a nuisance component. We establish non-asymptotic bounds on the excess risk of orthogonal statistical learning methods with a loss function satisfying a self-concordance property. Our bounds improve upon existing bounds by a dimension factor while lifting the assumption of strong convexity. We illustrate the results with examples from multiple treatment effect estimation and generalized partially linear modeling."
                    },
                    {
                        "title": "Modified Gauss-Newton Algorithms under Noise",
                        "abstract": "Gauss-Newton methods and their stochastic version have been widely used in machine learning and signal processing. Their nonsmooth counterparts, modified Gauss-Newton or prox-linear algorithms, can lead to contrasting outcomes when compared to gradient descent in large-scale statistical settings. We explore the contrasting performance of these two classes of algorithms in theory on a stylized statistical example, and experimentally on learning problems including structured prediction. In theory, we delineate the regime where the quadratic convergence of the modified Gauss-Newton method is active under statistical noise. In the experiments, we underline the versatility of stochastic (sub)-gradient descent to minimize nonsmooth composite objectives."
                    },
                    {
                        "title": "A Primal-Dual Algorithm for Faster Distributionally Robust Optimization",
                        "abstract": "We consider the penalized distributionally robust optimization (DRO) problem with a closed, convex uncertainty set, a setting that encompasses the $f$-DRO, Wasserstein-DRO, and spectral/$L$-risk formulations used in practice. We present Drago, a stochastic primal-dual algorithm that achieves a state-of-the-art linear convergence rate on strongly convex-strongly concave DRO problems. The method combines both randomized and cyclic components with mini-batching, which effectively handles the unique asymmetric nature of the primal and dual problems in DRO. We support our theoretical results with numerical benchmarks in classification and regression."
                    }
                ]
            }
        }
    },
    "2406.04090": {
        "paper_data": {
            "title": "Interpretable Lightweight Transformer via Unrolling of Learned Graph Smoothness Priors",
            "url": "http://arxiv.org/abs/2406.04090v1",
            "arxiv_id": "2406.04090",
            "authors": [
                "Tam Thuc Do",
                "Parham Eftekhar",
                "Seyed Alireza Hosseini",
                "Gene Cheung",
                "Philip Chou"
            ],
            "abstract": "We build interpretable and lightweight transformer-like neural networks by unrolling iterative optimization algorithms that minimize graph smoothness priors -- the quadratic graph Laplacian regularizer (GLR) and the $\\ell_1$-norm graph total variation (GTV) -- subject to an interpolation constraint. The crucial insight is that a normalized signal-dependent graph learning module amounts to a variant of the basic self-attention mechanism in conventional transformers. Unlike \"black-box\" transformers that require learning of large key, query and value matrices to compute scaled dot products as affinities and subsequent output embeddings, resulting in huge parameter sets, our unrolled networks employ shallow CNNs to learn low-dimensional features per node to establish pairwise Mahalanobis distances and construct sparse similarity graphs. At each layer, given a learned graph, the target interpolated signal is simply a low-pass filtered output derived from the minimization of an assumed graph smoothness prior, leading to a dramatic reduction in parameter count. Experiments for two image interpolation applications verify the restoration performance, parameter efficiency and robustness to covariate shift of our graph-based unrolled networks compared to conventional transformers.",
            "introduction": "   1 Introduction  Focusing on the self-attention mechanism [1] as the basic building block\u2014where the affinity between two input tokens is computed as a scaled dot product\u2014transformers [2] learn large parameter sets to achieve state-of-the-art (SOTA) performance in a wide range of signal prediction/classification problems [3, 4], outperforming convolutional neural nets (CNNs) and recurrent neural nets (RNNs). However, there are shortcomings, including i) lack of mathematical interpretability to characterize general performance111While works exist to analyze existing transformer architectures [5, 6, 7, 8, 9], only [10, 11] characterized the performance of a single self-attention layer and a shallow transformer, respectively. In contrast, we build transformer-like networks by unrolling graph-based algorithms, so that each layer is interpretable by construction., ii) requiring substantial training datasets to train sizable parameters [12], and iii) fragility to covariate shift\u2014when training and testing data have different distributions [13].   Orthogonally, algorithm unrolling [14] implements iterations of a model-based algorithm as a sequence of neural layers to build a feed-forward network, whose parameters can be learned end-to-end via back-propagation from data. A classic example is the unrolling of the iterative soft-thresholding algorithm (ISTA) in sparse coding into Learned ISTA (LISTA) [15]. Recently, [16] showed that by unrolling an iterative algorithm minimizing a sparse rate reduction (SRR) objective, it can lead to a family of \u201cwhite-box\u201d transformer-like deep network architectures that are 100% mathematically interpretable. Inspired by [16], in this work we also seek to build white-box transformers via algorithm unrolling, but from a unique graph signal processing (GSP) perspective [17, 18, 19].   Over the last decade and a half, GSP studies spectral analysis and processing of discrete signals on structured data kernels described by combinatorial graphs. Specifically, by assuming that the sought signal is smooth (low-pass) with respect to (w.r.t.) a particular graph, a plethora of graph-based restoration algorithms can be designed for practical applications, including image denoising [20], JPEG dequantization [21], interpolation [22], 3D point cloud denoising [23, 24] and super-resolution [25]. At the heart of GSP is the construction of a similarity graph that captures pairwise similarities between signal samples on two connected nodes. We first demonstrate that signal-dependent similarity graph learning with normalization is akin to affinity computation in the self-attention mechanism. Thus, our first contribution is to show that unrolling of a graph-based iterative algorithm with normalized graph learning results in an interpretable transformer-like feed-forward network.   Second, computation of a positive weight wi,j=exp\u2061(\u2212d\u2062(i,j))subscript\ud835\udc64\ud835\udc56\ud835\udc57\ud835\udc51\ud835\udc56\ud835\udc57w_{i,j}=\\exp(-d(i,j))italic_w start_POSTSUBSCRIPT italic_i , italic_j end_POSTSUBSCRIPT = roman_exp ( - italic_d ( italic_i , italic_j ) ) of an edge (i,j)\ud835\udc56\ud835\udc57(i,j)( italic_i , italic_j ) connecting two graph nodes i\ud835\udc56iitalic_i and j\ud835\udc57jitalic_j often employs Mahalanobis distance d\u2062(i,j)=(\ud835\udc1fi\u2212\ud835\udc1fj)\u22a4\u2062\ud835\udc0c\u2062(\ud835\udc1fi\u2212\ud835\udc1fj)\ud835\udc51\ud835\udc56\ud835\udc57superscriptsubscript\ud835\udc1f\ud835\udc56subscript\ud835\udc1f\ud835\udc57top\ud835\udc0csubscript\ud835\udc1f\ud835\udc56subscript\ud835\udc1f\ud835\udc57d(i,j)=({\\mathbf{f}}_{i}-{\\mathbf{f}}_{j})^{\\top}{\\mathbf{M}}({\\mathbf{f}}_{i}% -{\\mathbf{f}}_{j})italic_d ( italic_i , italic_j ) = ( bold_f start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT - bold_f start_POSTSUBSCRIPT italic_j end_POSTSUBSCRIPT ) start_POSTSUPERSCRIPT \u22a4 end_POSTSUPERSCRIPT bold_M ( bold_f start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT - bold_f start_POSTSUBSCRIPT italic_j end_POSTSUBSCRIPT ) between representative (e.g., CNN-computed) feature vectors \ud835\udc1fisubscript\ud835\udc1f\ud835\udc56{\\mathbf{f}}_{i}bold_f start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT and \ud835\udc1fjsubscript\ud835\udc1f\ud835\udc57{\\mathbf{f}}_{j}bold_f start_POSTSUBSCRIPT italic_j end_POSTSUBSCRIPT, where \ud835\udc1fi,\ud835\udc1fj\u2208\u211dDsubscript\ud835\udc1f\ud835\udc56subscript\ud835\udc1f\ud835\udc57superscript\u211d\ud835\udc37{\\mathbf{f}}_{i},{\\mathbf{f}}_{j}\\in\\mathbb{R}^{D}bold_f start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT , bold_f start_POSTSUBSCRIPT italic_j end_POSTSUBSCRIPT \u2208 blackboard_R start_POSTSUPERSCRIPT italic_D end_POSTSUPERSCRIPT reside in low-dimensional space [26, 27]. Hence, unlike a conventional transformer that requires large key and query matrices, \ud835\udc0a\ud835\udc0a{\\mathbf{K}}bold_K and \ud835\udc10\ud835\udc10{\\mathbf{Q}}bold_Q, to compute scaled dot products, the similarity graph learning",
            "references": [
                {
                    "title": "Trained Transformers Learn Linear Models In-Context",
                    "abstract": "Attention-based neural networks such as transformers have demonstrated a remarkable ability to exhibit in-context learning (ICL): Given a short prompt sequence of tokens from an unseen task, they can formulate relevant per-token and next-token predictions without any parameter updates. By embedding a sequence of labeled training data and unlabeled test data as a prompt, this allows for transformers to behave like supervised learning algorithms. Indeed, recent work has shown that when training transformer architectures over random instances of linear regression problems, these models' predictions mimic those of ordinary least squares. Towards understanding the mechanisms underlying this phenomenon, we investigate the dynamics of ICL in transformers with a single linear self-attention layer trained by gradient flow on linear regression tasks. We show that despite non-convexity, gradient flow with a suitable random initialization finds a global minimum of the objective function. At this global minimum, when given a test prompt of labeled examples from a new prediction task, the transformer achieves prediction error competitive with the best linear predictor over the test prompt distribution. We additionally characterize the robustness of the trained transformer to a variety of distribution shifts and show that although a number of shifts are tolerated, shifts in the covariate distribution of the prompts are not. Motivated by this, we consider a generalized ICL setting where the covariate distributions can vary across prompts. We show that although gradient flow succeeds at finding a global minimum in this setting, the trained transformer is still brittle under mild covariate shifts. We complement this finding with experiments on large, nonlinear transformer architectures which we show are more robust under covariate shifts."
                },
                {
                    "title": "White-Box Transformers via Sparse Rate Reduction",
                    "abstract": "In this paper, we contend that the objective of representation learning is to compress and transform the distribution of the data, say sets of tokens, towards a mixture of low-dimensional Gaussian distributions supported on incoherent subspaces. The quality of the final representation can be measured by a unified objective function called sparse rate reduction. From this perspective, popular deep networks such as transformers can be naturally viewed as realizing iterative schemes to optimize this objective incrementally. Particularly, we show that the standard transformer block can be derived from alternating optimization on complementary parts of this objective: the multi-head self-attention operator can be viewed as a gradient descent step to compress the token sets by minimizing their lossy coding rate, and the subsequent multi-layer perceptron can be viewed as attempting to sparsify the representation of the tokens. This leads to a family of white-box transformer-like deep network architectures which are mathematically fully interpretable. Despite their simplicity, experiments show that these networks indeed learn to optimize the designed objective: they compress and sparsify representations of large-scale real-world vision datasets such as ImageNet, and achieve performance very close to thoroughly engineered transformers such as ViT. Code is at \\url{https://github.com/Ma-Lab-Berkeley/CRATE}."
                },
                {
                    "title": "A Theoretical Understanding of shallow Vision Transformers: Learning, Generalization, and Sample Complexity",
                    "abstract": "Vision Transformers (ViTs) with self-attention modules have recently achieved great empirical success in many vision tasks. Due to non-convex interactions across layers, however, theoretical learning and generalization analysis is mostly elusive. Based on a data model characterizing both label-relevant and label-irrelevant tokens, this paper provides the first theoretical analysis of training a shallow ViT, i.e., one self-attention layer followed by a two-layer perceptron, for a classification task. We characterize the sample complexity to achieve a zero generalization error. Our sample complexity bound is positively correlated with the inverse of the fraction of label-relevant tokens, the token noise level, and the initial model error. We also prove that a training process using stochastic gradient descent (SGD) leads to a sparse attention map, which is a formal verification of the general intuition about the success of attention. Moreover, this paper indicates that a proper token sparsification can improve the test performance by removing label-irrelevant and/or noisy tokens, including spurious correlations. Empirical experiments on synthetic data and CIFAR-10 dataset justify our theoretical results and generalize to deeper ViTs."
                },
                {
                    "title": "Scaling Vision Transformers to 22 Billion Parameters",
                    "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                },
                {
                    "title": "Point Cloud Video Super-Resolution via Partial Point Coupling and Graph Smoothness",
                    "abstract": "Point cloud (PC) is a collection of discrete geometric samples of a physical object in 3D space. A PC video consists of temporal frames evenly spaced in time, each containing a static PC at one time instant. PCs in adjacent frames typically do not have point-to-point (P2P) correspondence, and thus exploiting temporal redundancy for PC restoration across frames is difficult. In this paper, we focus on the super-resolution (SR) problem for PC video: increase point density of PCs in video frames while preserving salient geometric features consistently across time. We accomplish this with two ideas. First, we establish partial P2P coupling between PCs of adjacent frames by interpolating interior points in a low-resolution PC patch in frame <inline-formula> <tex-math notation=\"LaTeX\">$t$ </tex-math></inline-formula> and translating them to a corresponding patch in frame <inline-formula> <tex-math notation=\"LaTeX\">$t+1$ </tex-math></inline-formula>, via a motion model computed by iterative closest point (ICP). Second, we promote piecewise smoothness in 3D geometry in each patch using feature graph Laplacian regularizer (FGLR) in an easily computable quadratic form. The two ideas translate to an unconstrained quadratic programming (QP) problem with a system of linear equations as solution\u2014one where we ensure the numerical stability by upper-bounding the condition number of the coefficient matrix. Finally, to improve the accuracy of the ICP motion model, we re-sample points in a super-resolved patch at time <inline-formula> <tex-math notation=\"LaTeX\">$t$ </tex-math></inline-formula> to better match a low-resolution patch at time <inline-formula> <tex-math notation=\"LaTeX\">$t+1$ </tex-math></inline-formula> via bipartite graph matching after each SR iteration. Experimental results show temporally consistent super-resolved PC videos generated by our scheme, outperforming SR competitors that optimized on a per-frame basis, in two established PC metrics."
                },
                {
                    "title": "Manifold Graph Signal Restoration Using Gradient Graph Laplacian Regularizer",
                    "abstract": "In the graph signal processing (GSP) literature, graph Laplacian regularizer (GLR) was used for signal restoration to promote piecewise smooth / constant reconstruction with respect to an underlying graph. However, for signals slowly varying across graph kernels, GLR suffers from an undesirable \u201cstaircase\u201d effect. In this paper, focusing on manifold graphs\u2014collections of uniform discrete samples on low-dimensional continuous manifolds\u2014we generalize GLR to gradient graph Laplacian regularizer (GGLR) that promotes planar / piecewise planar (PWP) signal reconstruction. Specifically, for a graph endowed with sampling coordinates (e.g., 2D images, 3D point clouds), we first define a gradient operator, using which we construct a gradient graph for nodes\u2019 gradients in the sampling manifold space. This maps to a gradient-induced nodal graph (GNG) and a positive semi-definite (PSD) Laplacian matrix with planar signals as the $0$ frequencies. For manifold graphs without explicit sampling coordinates, we propose a graph embedding method to obtain node coordinates via fast eigenvector computation. We derive the means-square-error minimizing weight parameter for GGLR efficiently, trading off bias and variance of the signal estimate. Experimental results show that GGLR outperformed previous graph signal priors like GLR and graph total variation (GTV) in a range of graph signal restoration tasks."
                },
                {
                    "title": "Residual Swin Transformer Channel Attention Network for Image Demosaicing",
                    "abstract": "Image demosaicing is problem of interpolating full-resolution color images from raw sensor (color filter array) data. During last decade, deep neural networks have been widely used in image restoration, and in particular, in demosaicing, attaining significant performance improvement. In recent years, vision transformers have been designed and successfully used in various computer vision applications. One of the recent methods of image restoration based on a Swin Transformer (ST), SwinIR, demonstrates state-of-the-art performance with a smaller number of parameters than neural network-based methods. Inspired by the success of SwinIR, we propose in this paper a novel Swin Transformer-based network for image demosaicing, called RSTCANet. To extract image features, RSTCANet stacks several residual Swin Transformer Channel Attention blocks (RSTCAB), introducing the channel attention for each two successive ST blocks. Extensive experiments demonstrate that RSTCANet outperforms state-of-the-art image demosaicing methods, and has a smaller number of parameters. The source code is available at https://github.com/xingwz/RSTCANet."
                },
                {
                    "title": "Inductive Biases and Variable Creation in Self-Attention Mechanisms",
                    "abstract": "Self-attention, an architectural motif designed to model long-range interactions in sequential data, has driven numerous recent breakthroughs in natural language processing and beyond. This work provides a theoretical analysis of the inductive biases of self-attention modules. Our focus is to rigorously establish which functions and long-range dependencies self-attention blocks prefer to represent. Our main result shows that bounded-norm Transformer networks\"create sparse variables\": a single self-attention head can represent a sparse function of the input sequence, with sample complexity scaling only logarithmically with the context length. To support our analysis, we present synthetic experiments to probe the sample complexity of learning sparse Boolean functions with Transformers."
                },
                {
                    "title": "Statistically Meaningful Approximation: a Case Study on Approximating Turing Machines with Transformers",
                    "abstract": "A common lens to theoretically study neural net architectures is to analyze the functions they can approximate. However, constructions from approximation theory may be unrealistic and therefore less meaningful. For example, a common unrealistic trick is to encode target function values using infinite precision. To address these issues, this work proposes a formal definition of statistically meaningful (SM) approximation which requires the approximating network to exhibit good statistical learnability. We study SM approximation for two function classes: boolean circuits and Turing machines. We show that overparameterized feedforward neural nets can SM approximate boolean circuits with sample complexity depending only polynomially on the circuit size, not the size of the network. In addition, we show that transformers can SM approximate Turing machines with computation time bounded by $T$ with sample complexity polynomial in the alphabet size, state space size, and $\\log (T)$. We also introduce new tools for analyzing generalization which provide much tighter sample complexities than the typical VC-dimension or norm-based bounds, which may be of independent interest."
                },
                {
                    "title": "Approximating How Single Head Attention Learns",
                    "abstract": "Why do models often attend to salient words, and how does this evolve throughout training? We approximate model training as a two stage process: early on in training when the attention weights are uniform, the model learns to translate individual input word `i` to `o` if they co-occur frequently. Later, the model learns to attend to `i` while the correct output is $o$ because it knows `i` translates to `o`. To formalize, we define a model property, Knowledge to Translate Individual Words (KTIW) (e.g. knowing that `i` translates to `o`), and claim that it drives the learning of the attention. This claim is supported by the fact that before the attention mechanism is learned, KTIW can be learned from word co-occurrence statistics, but not the other way around. Particularly, we can construct a training distribution that makes KTIW hard to learn, the learning of the attention fails, and the model cannot even learn the simple task of copying the input words to the output. Our approximation explains why models sometimes attend to salient words, and inspires a toy example where a multi-head attention model can overcome the above hard training distribution by improving learning dynamics rather than expressiveness. We end by discussing the limitation of our approximation framework and suggest future directions."
                },
                {
                    "title": "Learning Sparse Graph Laplacian with K Eigenvector Prior via Iterative Glasso and Projection",
                    "abstract": "Learning a suitable graph is an important precursor to many graph signal processing (GSP) pipelines, such as graph signal compression and denoising. Previous graph learning algorithms either i) make assumptions on graph connectivity (e.g., graph sparsity), or ii) make edge weight assumptions such as positive edges only. In this paper, given an empirical covariance matrix ${\\mathbf{\\bar C}}$ computed from data as input, we consider an eigen-structural assumption on the graph Laplacian matrix L: the first K eigenvectors of L are pre-selected, e.g., based on domain-specific criteria, and the remaining eigenvectors are then learned from data. One example use case is image coding, where the first eigenvector is pre-chosen to be constant, regardless of available observed data. We first prove that the subspace $\\mathcal{H}_{\\mathbf{u}}^ + $ of symmetric positive semi-definite (PSD) matrices with the first K eigenvectors being {uk} in a defined Hilbert space is a convex cone. We then construct an operator to project a given positive definite (PD) matrix L to $\\mathcal{H}_{\\mathbf{u}}^ + $, inspired by the Gram-Schmidt procedure. Finally, we design an efficient hybrid graphical lasso / projection algorithm to compute the most suitable graph Laplacian matrix ${{\\mathbf{L}}^ * } \\in \\mathcal{H}_{\\mathbf{u}}^ + $ given ${\\mathbf{\\bar C}}$. Experimental results show that given the first K eigenvectors as a prior, our algorithm outperforms competing graph learning schemes using a variety of graph comparison metrics."
                },
                {
                    "title": "Why are Adaptive Methods Good for Attention Models?",
                    "abstract": "While stochastic gradient descent (SGD) is still the \\emph{de facto} algorithm in deep learning, adaptive methods like Clipped SGD/Adam have been observed to outperform SGD across important tasks, such as attention models. The settings under which SGD performs poorly in comparison to adaptive methods are not well understood yet. In this paper, we provide empirical and theoretical evidence that a heavy-tailed distribution of the noise in stochastic gradients is one cause of SGD's poor performance. We provide the first tight upper and lower convergence bounds for adaptive gradient methods under heavy-tailed noise. Further, we demonstrate how gradient clipping plays a key role in addressing heavy-tailed gradient noise. Subsequently, we show how clipping can be applied in practice by developing an \\emph{adaptive} coordinate-wise clipping algorithm (ACClip) and demonstrate its superior performance on BERT pretraining and finetuning tasks."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Image Interpolation Using Multi-Scale Attention-Aware Inception Network",
                    "abstract": "A new <italic>multi-scale deep learning</italic> (MDL) framework is proposed and exploited for conducting image <italic>interpolation</italic> in this paper. The core of the framework is a <italic>seeding network</italic> that needs to be designed for the targeted task. For image interpolation, a novel <italic>attention-aware inception network</italic> (AIN) is developed as the seeding network; it has two key stages: 1) <italic>feature extraction</italic> based on the low-resolution input image; and 2) <italic>feature-to-image mapping</italic> to enlarge image\u2019s size or resolution. Note that the designed seeding network, AIN, needs to be trained with a matched training dataset at each scale. For that, multi-scale image patches are generated using our proposed <italic>pyramid cut</italic>, which outperforms the conventional image pyramid method by completely avoiding aliasing issue. After training, the <italic>trained</italic> AINs are then combined for processing the input image in the testing stage. Extensive experimental simulation results obtained from seven image datasets (comprising 359 images in total) have clearly shown that the proposed MAIN consistently delivers highly accurate interpolated images."
                },
                {
                    "title": "A Mathematical Theory of Attention",
                    "abstract": "Attention is a powerful component of modern neural networks across a wide variety of domains. However, despite its ubiquity in machine learning, there is a gap in our understanding of attention from a theoretical point of view. We propose a framework to fill this gap by building a mathematically equivalent model of attention using measure theory. With this model, we are able to interpret self-attention as a system of self-interacting particles, we shed light on self-attention from a maximum entropy perspective, and we show that attention is actually Lipschitz-continuous (with an appropriate metric) under suitable assumptions. We then apply these insights to the problem of mis-specified input data; infinitely-deep, weight-sharing self-attention networks; and more general Lipschitz estimates for a specific type of attention studied in concurrent work."
                },
                {
                    "title": "Signed Graph Metric Learning via Gershgorin Disc Perfect Alignment",
                    "abstract": "Given a convex and differentiable objective <inline-formula><tex-math notation=\"LaTeX\">$Q({\\mathbf M})$</tex-math><alternatives><mml:math><mml:mrow><mml:mi>Q</mml:mi><mml:mo>(</mml:mo><mml:mi mathvariant=\"bold\">M</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href=\"yang-ieq1-3091682.gif\"/></alternatives></inline-formula> for a real symmetric matrix <inline-formula><tex-math notation=\"LaTeX\">${\\mathbf M}$</tex-math><alternatives><mml:math><mml:mi mathvariant=\"bold\">M</mml:mi></mml:math><inline-graphic xlink:href=\"yang-ieq2-3091682.gif\"/></alternatives></inline-formula> in the positive definite (PD) cone\u2014used to compute Mahalanobis distances\u2014we propose a fast general metric learning framework that is entirely projection-free. We first assume that <inline-formula><tex-math notation=\"LaTeX\">${\\mathbf M}$</tex-math><alternatives><mml:math><mml:mi mathvariant=\"bold\">M</mml:mi></mml:math><inline-graphic xlink:href=\"yang-ieq3-3091682.gif\"/></alternatives></inline-formula> resides in a space <inline-formula><tex-math notation=\"LaTeX\">${\\mathcal S}$</tex-math><alternatives><mml:math><mml:mi mathvariant=\"script\">S</mml:mi></mml:math><inline-graphic xlink:href=\"yang-ieq4-3091682.gif\"/></alternatives></inline-formula> of generalized graph Laplacian matrices corresponding to balanced signed graphs. <inline-formula><tex-math notation=\"LaTeX\">${\\mathbf M}\\in {\\mathcal S}$</tex-math><alternatives><mml:math><mml:mrow><mml:mi mathvariant=\"bold\">M</mml:mi><mml:mo>\u2208</mml:mo><mml:mi mathvariant=\"script\">S</mml:mi></mml:mrow></mml:math><inline-graphic xlink:href=\"yang-ieq5-3091682.gif\"/></alternatives></inline-formula> that is also PD is called a graph metric matrix. Unlike low-rank metric matrices common in the literature, <inline-formula><tex-math notation=\"LaTeX\">${\\mathcal S}$</tex-math><alternatives><mml:math><mml:mi mathvariant=\"script\">S</mml:mi></mml:math><inline-graphic xlink:href=\"yang-ieq6-3091682.gif\"/></alternatives></inline-formula> includes the important diagonal-only matrices as a special case. The key theorem to circumvent full eigen-decomposition and enable fast metric matrix optimization is Gershgorin disc perfect alignment (GDPA): given <inline-formula><tex-math notation=\"LaTeX\">${\\mathbf M}\\in {\\mathcal S}$</tex-math><alternatives><mml:math><mml:mrow><mml:mi mathvariant=\"bold\">M</mml:mi><mml:mo>\u2208</mml:mo><mml:mi mathvariant=\"script\">S</mml:mi></mml:mrow></mml:math><inline-graphic xlink:href=\"yang-ieq7-3091682.gif\"/></alternatives></inline-formula> and diagonal matrix <inline-formula><tex-math notation=\"LaTeX\">${\\mathbf S}$</tex-math><alternatives><mml:math><mml:mi mathvariant=\"bold\">S</mml:mi></mml:math><inline-graphic xlink:href=\"yang-ieq8-3091682.gif\"/></alternatives></inline-formula>, where <inline-formula><tex-math notation=\"LaTeX\">$S_{ii} = 1/v_i$</tex-math><alternatives><mml:math><mml:mrow><mml:msub><mml:mi>S</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>i</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mn>1</mml:mn><mml:mo>/</mml:mo><mml:msub><mml:mi>v</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:math><inline-graphic xlink:href=\"yang-ieq9-3091682.gif\"/></alternatives></inline-formula> and <inline-formula><tex-math notation=\"LaTeX\">${\\mathbf v}$</tex-math><alternatives><mml:math><mml:mi mathvariant=\"bold\">v</mml:mi></mml:math><inline-graphic xlink:href=\"yang-ieq10-3091682.gif\"/></alternatives></inline-formula> is the first eigenvector of <inline-formula><tex-math notation=\"LaTeX\">${\\mathbf M}$</tex-math><alternatives><mml:math><mml:mi mathvariant=\"bold\">M</mml:mi></mml:math><inline-graphic xlink:href=\"yang-ieq11-3091682.gif\"/></alternatives></inline-formula>, we prove that Gershgorin disc left-ends of similarity transform <inline-formula><tex-math notation=\"LaTeX\">${\\mathbf B}= {\\mathbf S}{\\mathbf M}{\\mathbf S}^{-1}$</tex-math><alternatives><mml:math><mml:mrow><mml:mi mathvariant=\"bold\">B</mml:mi><mml:mo>=</mml:mo><mml:mi mathvariant=\"bold\">S</mml:mi><mml:mi mathvariant=\"bold\">M</mml:mi><mml:msup><mml:mrow><mml:mi mathvariant=\"bold\">S</mml:mi></mml:mrow><mml:mrow><mml:mo>-</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup></mml:mrow></mml:math><inline-graphic xlink:href=\"yang-ieq12-3091682.gif\"/></alternatives></inline-formula> are perfectly aligned at the smallest eigenvalue <inline-formula><tex-math notation=\"LaTeX\">$\\lambda _{\\min }$</tex-math><alternatives><mml:math><mml:msub><mml:mi>\u03bb</mml:mi><mml:mo movablelimits=\"true\" form=\"prefix\">min</mml:mo></mml:msub></mml:math><inline-graphic xlink:href=\"yang-ieq13-3091682.gif\"/></alternatives></inline-formula>. Using this theorem, we replace the PD cone constraint in the metric learning problem with tightest possible linear constraints per iteration, so that the alternating optimization of the diagonal / off-diagonal terms in <inline-formula><tex-math notation=\"LaTeX\">${\\mathbf M}$</tex-math><alternatives><mml:math><mml:mi mathvariant=\"bold\">M</mml:mi></mml:math><inline-graphic xlink:href=\"yang-ieq14-3091682.gif\"/></alternatives></inline-formula> can be solved efficiently as linear programs via the Frank-Wolfe method. We update <inline-formula><tex-math notation=\"LaTeX\">${\\mathbf v}$</tex-math><alternatives><mml:math><mml:mi mathvariant=\"bold\">v</mml:mi></mml:math><inline-graphic xlink:href=\"yang-ieq15-3091682.gif\"/></alternatives></inline-formula> using Locally Optimal Block Preconditioned Conjugate Gradient (LOBPCG) with warm start as entries in <inline-formula><tex-math notation=\"LaTeX\">${\\mathbf M}$</tex-math><alternatives><mml:math><mml:mi mathvariant=\"bold\">M</mml:mi></mml:math><inline-graphic xlink:href=\"yang-ieq16-3091682.gif\"/></alternatives></inline-formula> are optimized successively. Experiments show that our graph metric optimization is significantly faster than cone-projection schemes, and produces competitive binary classification performance."
                },
                {
                    "title": "The Lipschitz Constant of Self-Attention",
                    "abstract": "Lipschitz constants of neural networks have been explored in various contexts in deep learning, such as provable adversarial robustness, estimating Wasserstein distance, stabilising training of GANs, and formulating invertible neural networks. Such works have focused on bounding the Lipschitz constant of fully connected or convolutional networks, composed of linear maps and pointwise non-linearities. In this paper, we investigate the Lipschitz constant of self-attention, a non-linear neural network module widely used in sequence modelling. We prove that the standard dot-product self-attention is not Lipschitz, and propose an alternative L2 self-attention that is Lipschitz. We derive an upper bound on the Lipschitz constant of L2 self-attention and provide empirical evidence for its asymptotic tightness. To demonstrate the practical relevance of the theory, we formulate invertible self-attention and use it in a Transformer-based architecture for a character-level language modelling task."
                },
                {
                    "title": "Point Cloud Denoising via Feature Graph Laplacian Regularization",
                    "abstract": "Point cloud is a collection of 3D coordinates that are discrete geometric samples of an object\u2019s 2D surfaces. Imperfection in the acquisition process means that point clouds are often corrupted with noise. Building on recent advances in graph signal processing, we design local algorithms for 3D point cloud denoising. Specifically, we design a signal-dependent feature graph Laplacian regularizer (SDFGLR) that assumes surface normals computed from point coordinates are piecewise smooth with respect to a signal-dependent graph Laplacian matrix. Using SDFGLR as a signal prior, we formulate an optimization problem with a general <inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{p}$ </tex-math></inline-formula>-norm fidelity term that can explicitly remove only two types of additive noise: small but non-sparse noise like Gaussian (using <inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{2}$ </tex-math></inline-formula> fidelity term) and large but sparser noise like Laplacian (using <inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{1}$ </tex-math></inline-formula> fidelity term).To establish a linear relationship between normals and 3D point coordinates, we first perform bipartite graph approximation to divide the point cloud into two disjoint node sets (red and blue). We then optimize the red and blue nodes\u2019 coordinates alternately. For <inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{2}$ </tex-math></inline-formula>-norm fidelity term, we iteratively solve an unconstrained quadratic programming (QP) problem, efficiently computed using conjugate gradient with a bounded condition number to ensure numerical stability. For <inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{1}$ </tex-math></inline-formula>-norm fidelity term, we iteratively minimize an <inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{1}$ </tex-math></inline-formula>-<inline-formula> <tex-math notation=\"LaTeX\">$\\ell _{2}$ </tex-math></inline-formula> cost function using accelerated proximal gradient (APG), where a good step size is chosen via Lipschitz continuity analysis. Finally, we propose simple mean and median filters for flat patches of a given point cloud to estimate the noise variance given the noise type, which in turn is used to compute a weight parameter trading off the fidelity term and signal prior in the problem formulation. Extensive experiments show state-of-the-art denoising performance among local methods using our proposed algorithms."
                },
                {
                    "title": "Algorithm Unrolling: Interpretable, Efficient Deep Learning for Signal and Image Processing",
                    "abstract": "Deep neural networks provide unprecedented performance gains in many real-world problems in signal and image processing. Despite these gains, the future development and practical deployment of deep networks are hindered by their black-box nature, i.e., a lack of interpretability and the need for very large training sets. An emerging technique called algorithm unrolling, or unfolding, offers promise in eliminating these issues by providing a concrete and systematic connection between iterative algorithms that are widely used in signal processing and deep neural networks. Unrolling methods were first proposed to develop fast neural network approximations for sparse coding. More recently, this direction has attracted enormous attention, and it is rapidly growing in both theoretic investigations and practical applications. The increasing popularity of unrolled deep networks is due, in part, to their potential in developing efficient, high-performance (yet interpretable) network architectures from reasonably sized training sets."
                },
                {
                    "title": "Feature Graph Learning for 3D Point Cloud Denoising",
                    "abstract": "Identifying an appropriate underlying graph kernel that reflects pairwise similarities is critical in many recent graph spectral signal restoration schemes, including image denoising, dequantization, and contrast enhancement. Existing graph learning algorithms compute the most likely entries of a properly defined graph Laplacian matrix <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {L}$</tex-math></inline-formula>, but require a large number of signal observations <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {z}$</tex-math></inline-formula>'s for a stable estimate. In this work, we assume instead the availability of a relevant feature vector <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {f}_i$</tex-math></inline-formula> per node <inline-formula><tex-math notation=\"LaTeX\">$i$</tex-math></inline-formula>, from which we compute an optimal feature graph via optimization of a feature metric. Specifically, we alternately optimize the diagonal and off-diagonal entries of a Mahalanobis distance matrix <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {M}$</tex-math></inline-formula> by minimizing the graph Laplacian regularizer (GLR) <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {z}^{\\top } \\mathbf {L} \\mathbf {z}$</tex-math></inline-formula>, where edge weight is <inline-formula><tex-math notation=\"LaTeX\">$w_{i,j} = \\exp \\lbrace - (\\mathbf {f}_i - \\mathbf {f}_j)^{\\top } \\mathbf {M} (\\mathbf {f}_i - \\mathbf {f}_j) \\rbrace$</tex-math></inline-formula>, given a single observation <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {z}$</tex-math></inline-formula>. We optimize diagonal entries via proximal gradient (PG), where we constrain <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {M}$</tex-math></inline-formula> to be positive definite (PD) via linear inequalities derived from the Gershgorin circle theorem. To optimize off-diagonal entries, we design a block descent algorithm that iteratively optimizes one row and column of <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {M}$</tex-math></inline-formula>. To keep <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {M}$</tex-math></inline-formula> PD, we constrain the Schur complement of sub-matrix <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {M}_{2,2}$</tex-math></inline-formula> of <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {M}$</tex-math></inline-formula> to be PD when optimizing via PG. Our algorithm mitigates full eigen-decomposition of <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {M}$</tex-math></inline-formula>, thus ensuring fast computation speed even when feature vector <inline-formula><tex-math notation=\"LaTeX\">$\\mathbf {f}_i$</tex-math></inline-formula> has high dimension. To validate its usefulness, we apply our feature graph learning algorithm to the problem of 3D point cloud denoising, resulting in state-of-the-art performance compared to competing schemes in extensive experiments."
                },
                {
                    "title": "Deep Graph Laplacian Regularization for Robust Denoising of Real Images",
                    "abstract": "Recent developments in deep learning have revolutionized the paradigm of image restoration. However, its applications on real image denoising are still limited, due to its sensitivity to training data and the complex nature of real image noise. In this work, we combine the robustness merit of model-based approaches and the learning power of data-driven approaches for real image denoising. Specifically, by integrating graph Laplacian regularization as a trainable module into a deep learning framework, we are less susceptible to overfitting than pure CNN-based approaches, achieving higher robustness to small datasets and cross-domain denoising. First, a sparse neighborhood graph is built from the output of a convolutional neural network (CNN). Then the image is restored by solving an unconstrained quadratic programming problem, using a corresponding graph Laplacian regularizer as a prior term. The proposed restoration pipeline is fully differentiable and hence can be end-to-end trained. Experimental results demonstrate that our work is less prone to overfitting given small training data. It is also endowed with strong cross-domain generalization power, outperforming the state-of-the-art approaches by a remarkable margin."
                },
                {
                    "title": "Learning Graphs From Data: A Signal Representation Perspective",
                    "abstract": "The construction of a meaningful graph topology plays a crucial role in the effective representation, processing, analysis, and visualization of structured data. When a natural choice of the graph is not readily available from the data sets, it is thus desirable to infer or learn a graph topology from the data. In this article, we survey solutions to the problem of graph learning, including classical viewpoints from statistics and physics, and more recent approaches that adopt a graph signal processing (GSP) perspective. We further emphasize the conceptual similarities and differences between classical and GSP-based graph-inference methods and highlight the potential advantage of the latter in a number of theoretical and practical scenarios. We conclude with several open issues and challenges that are keys to the design of future signal processing and machine-learning algorithms for learning graphs from data."
                },
                {
                    "title": "Graph-Based Blind Image Deblurring From a Single Photograph",
                    "abstract": "Blind image deblurring, i.e., deblurring without knowledge of the blur kernel, is a highly ill-posed problem. The problem can be solved in two parts: 1) estimate a blur kernel from the blurry image, and 2) given an estimated blur kernel, de-convolve the blurry input to restore the target image. In this paper, we propose a graph-based blind image deblurring algorithm by interpreting an image patch as a signal on a weighted graph. Specifically, we first argue that a skeleton image\u2014a proxy that retains the strong gradients of the target but smooths out the details\u2014can be used to accurately estimate the blur kernel and has a unique bi-modal edge weight distribution. Then, we design a reweighted graph total variation (RGTV) prior that can efficiently promote a bi-modal edge weight distribution given a blurry patch. Further, to analyze RGTV in the graph frequency domain, we introduce a new weight function to represent RGTV as a graph  $l_{1}$  -Laplacian regularizer. This leads to a graph spectral filtering interpretation of the prior with desirable properties, including robustness to noise and blur, strong piecewise smooth filtering, and sharpness promotion. Minimizing a blind image deblurring objective with RGTV results in a non-convex non-differentiable optimization problem. Leveraging the new graph spectral interpretation for RGTV, we design an efficient algorithm that solves for the skeleton image and the blur kernel alternately. Specifically for Gaussian blur, we propose a further speedup strategy for blind Gaussian deblurring using accelerated graph spectral filtering. Finally, with the computed blur kernel, recent non-blind image deblurring algorithms can be applied to restore the target image. Experimental results demonstrate that our algorithm successfully restores latent sharp images and outperforms the state-of-the-art methods quantitatively and qualitatively."
                },
                {
                    "title": "Graph Spectral Image Processing",
                    "abstract": "Recent advent of graph signal processing (GSP) has spurred intensive studies of signals that live naturally on irregular data kernels described by graphs (e.g., social networks, wireless sensor networks). Though a digital image contains pixels that reside on a regularly sampled 2-D grid, if one can design an appropriate underlying graph connecting pixels with weights that reflect the image structure, then one can interpret the image (or image patch) as a signal on a graph, and apply GSP tools for processing and analysis of the signal in graph spectral domain. In this paper, we overview recent graph spectral techniques in GSP specifically for image/video processing. The topics covered include image compression, image restoration, image filtering, and image segmentation."
                },
                {
                    "title": "Graph Signal Processing: Overview, Challenges, and Applications",
                    "abstract": "Research in graph signal processing (GSP) aims to develop tools for processing data defined on irregular graph domains. In this paper, we first provide an overview of core ideas in GSP and their connection to conventional digital signal processing, along with a brief historical perspective to highlight how concepts recently developed in GSP build on top of prior research in other areas. We then summarize recent advances in developing basic GSP tools, including methods for sampling, filtering, or graph learning. Next, we review progress in several application areas using GSP, including processing and analysis of sensor network data, biological data, and applications to image processing and machine learning."
                },
                {
                    "title": "Graph Signal Recovery via Primal-Dual Algorithms for Total Variation Minimization",
                    "abstract": "We consider the problem of recovering a smooth graph signal from noisy samples taken on a subset of graph nodes. The smoothness of the graph signal is quantified in terms of total variation. We formulate the signal recovery task as a convex optimization problem that minimizes the total variation of the graph signal while controlling its global or node-wise empirical error. We propose a first-order primal-dual algorithm to solve these total variation minimization problems. A distributed implementation of the algorithm is devised to handle large-dimensional applications efficiently. We use synthetic and real-world data to extensively compare the performance of our approach with state-of-the-art methods."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Graph Learning From Data Under Laplacian and Structural Constraints",
                    "abstract": "Graphs are fundamental mathematical structures used in various fields to represent data, signals, and processes. In this paper, we propose a novel framework for learning/estimating graphs from data. The proposed framework includes (i) formulation of various graph learning problems, (ii) their probabilistic interpretations, and (iii) associated algorithms. Specifically, graph learning problems are posed as the estimation of graph Laplacian matrices from some observed data under given structural constraints (e.g., graph connectivity and sparsity level). From a probabilistic perspective, the problems of interest correspond to maximum a posteriori parameter estimation of Gaussian\u2013Markov random field models, whose precision (inverse covariance) is a graph Laplacian matrix. For the proposed graph learning problems, specialized algorithms are developed by incorporating the graph Laplacian and structural constraints. The experimental results demonstrate that the proposed algorithms outperform the current state-of-the-art methods in terms of accuracy and computational efficiency."
                },
                {
                    "title": "Random Walk Graph Laplacian-Based Smoothness Prior for Soft Decoding of JPEG Images",
                    "abstract": "Given the prevalence of joint photographic experts group (JPEG) compressed images, optimizing image reconstruction from the compressed format remains an important problem. Instead of simply reconstructing a pixel block from the centers of indexed discrete cosine transform (DCT) coefficient quantization bins (hard decoding), soft decoding reconstructs a block by selecting appropriate coefficient values within the indexed bins with the help of signal priors. The challenge thus lies in how to define suitable priors and apply them effectively. In this paper, we combine three image priors-Laplacian prior for DCT coefficients, sparsity prior, and graph-signal smoothness prior for image patches-to construct an efficient JPEG soft decoding algorithm. Specifically, we first use the Laplacian prior to compute a minimum mean square error initial solution for each code block. Next, we show that while the sparsity prior can reduce block artifacts, limiting the size of the overcomplete dictionary (to lower computation) would lead to poor recovery of high DCT frequencies. To alleviate this problem, we design a new graph-signal smoothness prior (desired signal has mainly low graph frequencies) based on the left eigenvectors of the random walk graph Laplacian matrix (LERaG). Compared with the previous graph-signal smoothness priors, LERaG has desirable image filtering properties with low computation overhead. We demonstrate how LERaG can facilitate recovery of high DCT frequencies of a piecewise smooth signal via an interpretation of low graph frequency components as relaxed solutions to normalized cut in spectral clustering. Finally, we construct a soft decoding algorithm using the three signal priors with appropriate prior weights. Experimental results show that our proposal outperforms the state-of-the-art soft decoding algorithms in both objective and subjective evaluations noticeably."
                },
                {
                    "title": "Graph Laplacian Regularization for Inverse Imaging: Analysis in the Continuous Domain",
                    "abstract": "Inverse imaging problems are inherently under-determined, and hence it is important to employ appropriate image priors for regularization. One recent popular prior---the graph Laplacian regularizer---assumes that the target pixel patch is smooth with respect to an appropriately chosen graph. However, the mechanisms and implications of imposing the graph Laplacian regularizer on the original inverse problem are not well understood. To address this problem, in this paper we interpret neighborhood graphs of pixel patches as discrete counterparts of Riemannian manifolds and perform analysis in the continuous domain, providing insights into several fundamental aspects of graph Laplacian regularization. Specifically, we first show the convergence of the graph Laplacian regularizer to a continuous-domain functional, integrating a norm measured in a locally adaptive metric space. Focusing on image denoising, we derive an optimal metric space assuming nonlocal self-similarity of pixel patches, leading to an optimal graph Laplacian regularizer for denoising in the discrete domain. We then interpret graph Laplacian regularization as an anisotropic diffusion scheme to explain its behavior during iterations, e.g., its tendency to promote piecewise smooth signals under certain settings. To verify our analysis, an iterative image denoising algorithm is developed. Experimental results show that our algorithm performs competitively with state-of-the-art denoising methods such as BM3D for natural images, and outperforms them significantly for piecewise smooth images."
                },
                {
                    "title": "Single image super-resolution from transformed self-exemplars",
                    "abstract": "Self-similarity based super-resolution (SR) algorithms are able to produce visually pleasing results without extensive training on external databases. Such algorithms exploit the statistical prior that patches in a natural image tend to recur within and across scales of the same image. However, the internal dictionary obtained from the given image may not always be sufficiently expressive to cover the textural appearance variations in the scene. In this paper, we extend self-similarity based SR to overcome this drawback. We expand the internal patch search space by allowing geometric variations. We do so by explicitly localizing planes in the scene and using the detected perspective geometry to guide the patch search process. We also incorporate additional affine transformations to accommodate local shape variations. We propose a compositional model to simultaneously handle both types of transformations. We extensively evaluate the performance in both urban and natural scenes. Even without using any external training databases, we achieve significantly superior results on urban scenes, while maintaining comparable performance on natural scenes as other state-of-the-art SR algorithms."
                },
                {
                    "title": "Neural Machine Translation by Jointly Learning to Align and Translate",
                    "abstract": "Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition."
                },
                {
                    "title": "The emerging field of signal processing on graphs: Extending high-dimensional data analysis to networks and other irregular domains",
                    "abstract": "In applications such as social, energy, transportation, sensor, and neuronal networks, high-dimensional data naturally reside on the vertices of weighted graphs. The emerging field of signal processing on graphs merges algebraic and spectral graph theoretic concepts with computational harmonic analysis to process such signals on graphs. In this tutorial overview, we outline the main challenges of the area, discuss different ways to define graph spectral domains, which are the analogs to the classical frequency domain, and highlight the importance of incorporating the irregular structures of graph data domains when processing signals on graphs. We then review methods to generalize fundamental operations such as filtering, translation, modulation, dilation, and downsampling to the graph setting and survey the localized, multiscale transforms that have been proposed to efficiently extract information from high-dimensional data on graphs. We conclude with a brief discussion of open issues and possible extensions."
                },
                {
                    "title": "Color demosaicking by local directional interpolation and nonlocal adaptive thresholding",
                    "abstract": "Single sensor digital color cameras capture only one of the three primary colors at each pixel and a process called color demosaicking (CDM) is used to reconstruct the full color images. Most CDM algorithms assume the existence of high local spectral redundancy in estimating the missing color samples. However, for images with sharp color transitions and high color saturation, such an assumption may be invalid and visually unpleasant CDM errors will occur. In this paper, we exploit the image nonlocal redundancy to improve the local color reproduction result. First, multiple local direc- tional estimates of a missing color sample are computed and fused according to local gradients. Then, nonlocal pixels similar to the esti- mated pixel are searched to enhance the local estimate. An adaptive thresholding method rather than the commonly used nonlocal means filtering is proposed to improve the local estimate. This allows the final reconstruction to be performed at the structural level as op- posed to the pixel level. Experimental results demonstrate that the proposed local directional interpolation and nonlocal adaptive thresh- olding method outperforms many state-of-the-art CDM methods in reconstructing the edges and reducing color interpolation artifacts, leading to higher visual quality of reproduced color images. \u00a9 2011"
                },
                {
                    "title": "Learning Fast Approximations of Sparse Coding",
                    "abstract": "In Sparse Coding (SC), input vectors are reconstructed using a sparse linear combination of basis vectors. SC has become a popular method for extracting features from data. For a given input, SC minimizes a quadratic reconstruction error with an L1 penalty term on the code. The process is often too slow for applications such as real-time pattern recognition. We proposed two versions of a very fast algorithm that produces approximate estimates of the sparse code that can be used to compute good visual features, or to initialize exact iterative algorithms. The main idea is to train a non-linear, feed-forward predictor with a specific architecture and a fixed depth to produce the best possible approximation of the sparse code. A version of the method, which can be seen as a trainable version of Li and Osher's coordinate descent method, is shown to produce approximate solutions with 10 times less computation than Li and Os-her's for the same approximation error. Unlike previous proposals for sparse code predictors, the system allows a kind of approximate \"explaining away\" to take place during inference. The resulting predictor is differentiable and can be included into globally-trained recognition systems."
                },
                {
                    "title": "Convex Optimization",
                    "abstract": "This textbook is based on lectures given by the authors at MIPT (Moscow), HSE (Moscow), FEFU (Vladivostok), V.I. Vernadsky KFU (Simferopol), ASU (Republic of Adygea), and the University of Grenoble-Alpes (Grenoble, France). First of all, the authors focused on the program of a two-semester course of lectures on convex optimization, which is given to students of MIPT. The first chapter of this book contains the materials of the first semester (\"Fundamentals of convex analysis and optimization\"), the second and third chapters contain the materials of the second semester (\"Numerical methods of convex optimization\"). The textbook has a number of features. First, in contrast to the classic manuals, this book does not provide proofs of all the theorems mentioned. This allowed, on one side, to describe more themes, but on the other side, made the presentation less self-sufficient. The second important point is that part of the material is advanced and is published in the Russian educational literature, apparently for the first time. Third, the accents that are given do not always coincide with the generally accepted accents in the textbooks that are now popular. First of all, we talk about a sufficiently advanced presentation of conic optimization, including robust optimization, as a vivid demonstration of the capabilities of modern convex analysis."
                },
                {
                    "title": "Nonlocal Discrete Regularization on Weighted Graphs: A Framework for Image and Manifold Processing",
                    "abstract": "We introduce a nonlocal discrete regularization framework on weighted graphs of the arbitrary topologies for image and manifold processing. The approach considers the problem as a variational one, which consists of minimizing a weighted sum of two energy terms: a regularization one that uses a discrete weighted -Dirichlet energy and an approximation one. This is the discrete analogue of recent continuous Euclidean nonlocal regularization functionals. The proposed formulation leads to a family of simple and fast nonlinear processing methods based on the weighted -Laplace operator, parameterized by the degree of regularity, the graph structure and the graph weight function. These discrete processing methods provide a graph-based version of recently proposed semi-local or nonlocal processing methods used in image and mesh processing, such as the bilateral filter, the TV digital filter or the nonlocal means filter. It works with equal ease on regular 2-D and 3-D images, manifolds or any data. We illustrate the abilities of the approach by applying it to various types of images, meshes, manifolds, and data represented as graphs."
                },
                {
                    "title": "Bilateral filtering for gray and color images",
                    "abstract": "Bilateral filtering smooths images while preserving edges, by means of a nonlinear combination of nearby image values. The method is noniterative, local, and simple. It combines gray levels or colors based on both their geometric closeness and their photometric similarity, and prefers near values to distant values in both domain and range. In contrast with filters that operate on the three bands of a color image separately, a bilateral filter can enforce the perceptual metric underlying the CIE-Lab color space, and smooth colors and preserve edges in a way that is tuned to human perception. Also, in contrast with standard filtering, bilateral filtering produces no phantom colors along edges in color images, and reduces phantom colors where they appear in the original image."
                },
                {
                    "title": "An Introduction to the Conjugate Gradient Method Without the Agonizing Pain",
                    "abstract": "The Conjugate Gradient Method is the most prominent iterative method for solving sparse systems of linear equations. Unfortunately, many textbook treatments of the topic are written so that even their own authors would be mystified, if they bothered to read their own writing. For this reason, an understanding of the method has been reserved for the elite brilliant few who have painstakingly decoded the mumblings of their forebears. Nevertheless, the Conjugate Gradient Method is a composite of simple, elegant ideas that almost anyone can understand. Of course, a reader as intelligent as yourself will learn them almost effortlessly. The idea of quadratic forms is introduced and used to derive the methods of Steepest Descent, Conjugate Directions, and Conjugate Gradients. Eigenvectors are explained and used to examine the convergence of the Jacobi Method, Steepest Descent, and Conjugate Gradients. Other topics include preconditioning and the nonlinear Conjugate Gradient Method. I have taken pains to make this article easy to read. Sixty-two illustrations are provided. Dense prose is avoided. Concepts are explained in several different ways. Most equations are coupled with an intuitive interpretation."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "A New Alternating Direction Method for Linear Programming",
                    "abstract": "It is well known that, for a linear program (LP) with constraint matrix $\\mathbf{A}\\in\\mathbb{R}^{m\\times n}$, the Alternating Direction Method of Multiplier converges globally and linearly at a rate $O((\\|\\mathbf{A}\\|_F^2+mn)\\log(1/\\epsilon))$. However, such a rate is related to the problem dimension and the algorithm exhibits a slow and fluctuating ``tail convergence'' in practice. In this paper, we propose a new variable splitting method of LP and prove that our method has a convergence rate of $O(\\|\\mathbf{A}\\|^2\\log(1/\\epsilon))$. The proof is based on simultaneously estimating the distance from a pair of primal dual iterates to the optimal primal and dual solution set by certain residuals. In practice, we result in a new first-order LP solver that can exploit both the sparsity and the specific structure of matrix $\\mathbf{A}$ and a significant speedup for important problems such as basis pursuit, inverse covariance matrix estimation, L1 SVM and nonnegative matrix factorization problem compared with current fastest LP solvers."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CV",
                "eess.IV",
                "eess.SP"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we develop interpretable transformer-like architectures through algorithm unrolling in the context of graph signal processing?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is significant for the research community as it addresses the critical need for mathematical interpretability in transformer architectures, which are widely used in various applications. By providing a framework that allows for the construction of interpretable models, this research could lead to advancements in understanding how transformers operate, potentially influencing future research directions in model interpretability and robustness. Furthermore, the practical applications of interpretable models can enhance trust and reliability in machine learning systems, particularly in sensitive domains such as healthcare and finance.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the inherent complexity of transformer architectures, which typically involve large parameter sets and lack clear interpretability. Naive approaches may fail because they do not account for the need to maintain interpretability while also achieving high performance. Additionally, the technical obstacles include the difficulty of integrating graph-based algorithms with neural network architectures, as well as the need to ensure that the resulting models can generalize well across different datasets without requiring extensive training data.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on improving the performance of transformers without addressing their interpretability. Existing solutions often lack a systematic approach to integrate graph signal processing with transformer architectures. Barriers such as the absence of a clear mathematical framework for unrolling graph-based algorithms into neural networks have prevented this problem from being effectively tackled. Our approach differs by explicitly leveraging the principles of graph signal processing to create a framework that ensures interpretability while maintaining the performance benefits of transformer architectures.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves unrolling graph-based iterative algorithms to construct transformer-like networks that are mathematically interpretable. We will utilize datasets relevant to graph signal processing tasks, such as image denoising and super-resolution, and evaluate our models using metrics that assess both interpretability and performance, such as accuracy and model transparency. The expected outcomes include the development of a family of interpretable transformer architectures that can be applied to various signal processing tasks, demonstrating improved performance and interpretability compared to traditional transformers."
            }
        },
        "author_data": {
            "78ccbede-d524-4007-ab0b-5cabdaa9598d": {
                "pk": "78ccbede-d524-4007-ab0b-5cabdaa9598d",
                "name": "Tam Thuc Do",
                "collaborators": [
                    "Gene Cheung",
                    "Philip A. Chou",
                    "Saghar Bagheri",
                    "Antonio Ortega",
                    "Seyed Alireza Hosseini",
                    "Yuichi Tanaka"
                ],
                "domain": [
                    "3D Point Cloud",
                    "Image Compression",
                    "Graph Learning",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Volumetric Attribute Compression for 3D Point Clouds using Feedforward Network with Geometric Attention",
                        "abstract": "We study 3D point cloud attribute compression using a volumetric approach: given a target volumetric attribute function $f : \\mathbb{R}^3 \\rightarrow \\mathbb{R}$, we quantize and encode parameter vector $\\theta$ that characterizes $f$ at the encoder, for reconstruction $f_{\\hat{\\theta}}(\\mathbf{x})$ at known 3D points $\\mathbf{x}$'s at the decoder. Extending a previous work Region Adaptive Hierarchical Transform (RAHT) that employs piecewise constant functions to span a nested sequence of function spaces, we propose a feedforward linear network that implements higher-order B-spline bases spanning function spaces without eigen-decomposition. Feedforward network architecture means that the system is amenable to end-to-end neural learning. The key to our network is space-varying convolution, similar to a graph operator, whose weights are computed from the known 3D geometry for normalization. We show that the number of layers in the normalization at the encoder is equivalent to the number of terms in a matrix inverse Taylor series. Experimental results on real-world 3D point clouds show up to 2-3 dB gain over RAHT in energy compaction and 20-30% bitrate reduction."
                    },
                    {
                        "title": "Volumetric 3D Point Cloud Attribute Compression: Learned polynomial bilateral filter for prediction",
                        "abstract": "We extend a previous study on 3D point cloud attribute compression scheme that uses a volumetric approach: given a target volumetric attribute function $f : \\mathbb{R}^3 \\mapsto \\mathbb{R}$, we quantize and encode parameters $\\theta$ that characterize $f$ at the encoder, for reconstruction $f_{\\hat{\\theta}}(\\mathbf(x))$ at known 3D points $\\mathbf(x)$ at the decoder. Specifically, parameters $\\theta$ are quantized coefficients of B-spline basis vectors $\\mathbf{\\Phi}_l$ (for order $p \\geq 2$) that span the function space $\\mathcal{F}_l^{(p)}$ at a particular resolution $l$, which are coded from coarse to fine resolutions for scalability. In this work, we focus on the prediction of finer-grained coefficients given coarser-grained ones by learning parameters of a polynomial bilateral filter (PBF) from data. PBF is a pseudo-linear filter that is signal-dependent with a graph spectral interpretation common in the graph signal processing (GSP) field. We demonstrate PBF's predictive performance over a linear predictor inspired by MPEG standardization over a wide range of point cloud datasets."
                    },
                    {
                        "title": "Learned Nonlinear Predictor for Critically Sampled 3D Point Cloud Attribute Compression",
                        "abstract": "We study 3D point cloud attribute compression via a volumetric approach: assuming point cloud geometry is known at both encoder and decoder, parameters $\\theta$ of a continuous attribute function $f: \\mathbb{R}^3 \\mapsto \\mathbb{R}$ are quantized to $\\hat{\\theta}$ and encoded, so that discrete samples $f_{\\hat{\\theta}}(\\mathbf{x}_i)$ can be recovered at known 3D points $\\mathbf{x}_i \\in \\mathbb{R}^3$ at the decoder. Specifically, we consider a nested sequences of function subspaces $\\mathcal{F}^{(p)}_{l_0} \\subseteq \\cdots \\subseteq \\mathcal{F}^{(p)}_L$, where $\\mathcal{F}_l^{(p)}$ is a family of functions spanned by B-spline basis functions of order $p$, $f_l^*$ is the projection of $f$ on $\\mathcal{F}_l^{(p)}$ represented as low-pass coefficients $F_l^*$, and $g_l^*$ is the residual function in an orthogonal subspace $\\mathcal{G}_l^{(p)}$ (where $\\mathcal{G}_l^{(p)} \\oplus \\mathcal{F}_l^{(p)} = \\mathcal{F}_{l+1}^{(p)}$) represented as high-pass coefficients $G_l^*$. In this paper, to improve coding performance over \\cite{do2023volumetric}, we study predicting $f_{l+1}^*$ at level $l+1$ given $f_l^*$ at level $l$ and encoding of $G_l^*$ for the $p=1$ case (RAHT($1$)). For the prediction, we formalize RAHT(1) linear prediction in MPEG-PCC in a theoretical framework, and propose a new nonlinear predictor using a polynomial of bilateral filter. We derive equations to efficiently compute the critically sampled high-pass coefficients $G_l^*$ amenable to encoding. We optimize parameters in our resulting feed-forward network on a large training set of point clouds by minimizing a rate-distortion Lagrangian. Experimental results show that our improved framework outperforms the MPEG G-PCC predictor by $11\\%$--$12\\%$ in bit rate."
                    },
                    {
                        "title": "Hybrid Model-based / Data-driven Graph Transform for Image Coding",
                        "abstract": "Transform coding to sparsify signal representations remains crucial in an image compression pipeline. While the Karhunen-Lo\\`{e}ve transform (KLT) computed from an empirical covariance matrix $\\bar{C}$ is theoretically optimal for a stationary process, in practice, collecting sufficient statistics from a non-stationary image to reliably estimate $\\bar{C}$ can be difficult. In this paper, to encode an intra-prediction residual block, we pursue a hybrid model-based / data-driven approach: the first $K$ eigenvectors of a transform matrix are derived from a statistical model, e.g., the asymmetric discrete sine transform (ADST), for stability, while the remaining $N-K$ are computed from $\\bar{C}$ for performance. The transform computation is posed as a graph learning problem, where we seek a graph Laplacian matrix minimizing a graphical lasso objective inside a convex cone sharing the first $K$ eigenvectors in a Hilbert space of real symmetric matrices. We efficiently solve the problem via augmented Lagrangian relaxation and proximal gradient (PG). Using WebP as a baseline image codec, experimental results show that our hybrid graph transform achieved better energy compaction than default discrete cosine transform (DCT) and better stability than KLT."
                    },
                    {
                        "title": "Constructing an Interpretable Deep Denoiser by Unrolling Graph Laplacian Regularizer",
                        "abstract": "An image denoiser can be used for a wide range of restoration problems via the Plug-and-Play (PnP) architecture. In this paper, we propose a general framework to build an interpretable graph-based deep denoiser (GDD) by unrolling a solution to a maximum a posteriori (MAP) problem equipped with a graph Laplacian regularizer (GLR) as signal prior. Leveraging a recent theorem showing that any (pseudo-)linear denoiser $\\boldsymbol \\Psi$, under mild conditions, can be mapped to a solution of a MAP denoising problem regularized using GLR, we first initialize a graph Laplacian matrix $\\mathbf L$ via truncated Taylor Series Expansion (TSE) of $\\boldsymbol \\Psi^{-1}$. Then, we compute the MAP linear system solution by unrolling iterations of the conjugate gradient (CG) algorithm into a sequence of neural layers as a feed-forward network -- one that is amenable to parameter tuning. The resulting GDD network is \"graph-interpretable\", low in parameter count, and easy to initialize thanks to $\\mathbf L$ derived from a known well-performing denoiser $\\boldsymbol \\Psi$. Experimental results show that GDD achieves competitive image denoising performance compared to competitors, but employing far fewer parameters, and is more robust to covariate shift."
                    }
                ]
            },
            "64ffd8fd-0484-4cd2-aa2c-1b837928681a": {
                "pk": "64ffd8fd-0484-4cd2-aa2c-1b837928681a",
                "name": "Parham Eftekhar",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "3c3aa071-a61b-4e76-b52f-7ef61e98f6b0": {
                "pk": "3c3aa071-a61b-4e76-b52f-7ef61e98f6b0",
                "name": "Seyed Alireza Hosseini",
                "collaborators": [
                    "Tam Thuc Do",
                    "Gene Cheung",
                    "Yuichi Tanaka"
                ],
                "domain": [
                    "Image Processing",
                    "Deep Learning",
                    "Graph-based Methods",
                    "Denoising"
                ],
                "publications": [
                    {
                        "title": "Constructing an Interpretable Deep Denoiser by Unrolling Graph Laplacian Regularizer",
                        "abstract": "An image denoiser can be used for a wide range of restoration problems via the Plug-and-Play (PnP) architecture. In this paper, we propose a general framework to build an interpretable graph-based deep denoiser (GDD) by unrolling a solution to a maximum a posteriori (MAP) problem equipped with a graph Laplacian regularizer (GLR) as signal prior. Leveraging a recent theorem showing that any (pseudo-)linear denoiser $\\boldsymbol \\Psi$, under mild conditions, can be mapped to a solution of a MAP denoising problem regularized using GLR, we first initialize a graph Laplacian matrix $\\mathbf L$ via truncated Taylor Series Expansion (TSE) of $\\boldsymbol \\Psi^{-1}$. Then, we compute the MAP linear system solution by unrolling iterations of the conjugate gradient (CG) algorithm into a sequence of neural layers as a feed-forward network -- one that is amenable to parameter tuning. The resulting GDD network is \"graph-interpretable\", low in parameter count, and easy to initialize thanks to $\\mathbf L$ derived from a known well-performing denoiser $\\boldsymbol \\Psi$. Experimental results show that GDD achieves competitive image denoising performance compared to competitors, but employing far fewer parameters, and is more robust to covariate shift."
                    }
                ]
            },
            "ef968441-16cd-4d12-b7fb-1df76b21cf29": {
                "pk": "ef968441-16cd-4d12-b7fb-1df76b21cf29",
                "name": "Gene Cheung",
                "collaborators": [
                    "Chinthaka Dinesh",
                    "Xianming Liu",
                    "Jiahao Pang",
                    "Ivan V. Bajic",
                    "Jin Zeng",
                    "Fei Chen",
                    "Xue Zhang",
                    "Wei Hu",
                    "Saghar Bagheri",
                    "Enrico Magli"
                ],
                "domain": [
                    "Graph Signal Processing",
                    "Image Restoration",
                    "Machine Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Graph Laplacian Regularization for Image Denoising: Analysis in the Continuous Domain",
                        "abstract": "Inverse imaging problems are inherently under-determined, and hence it is important to employ appropriate image priors for regularization. One recent popular prior---the graph Laplacian regularizer---assumes that the target pixel patch is smooth with respect to an appropriately chosen graph. However, the mechanisms and implications of imposing the graph Laplacian regularizer on the original inverse problem are not well understood. To address this problem, in this paper we interpret neighborhood graphs of pixel patches as discrete counterparts of Riemannian manifolds and perform analysis in the continuous domain, providing insights into several fundamental aspects of graph Laplacian regularization for image denoising. Specifically, we first show the convergence of the graph Laplacian regularizer to a continuous-domain functional, integrating a norm measured in a locally adaptive metric space. Focusing on image denoising, we derive an optimal metric space assuming non-local self-similarity of pixel patches, leading to an optimal graph Laplacian regularizer for denoising in the discrete domain. We then interpret graph Laplacian regularization as an anisotropic diffusion scheme to explain its behavior during iterations, e.g., its tendency to promote piecewise smooth signals under certain settings. To verify our analysis, an iterative image denoising algorithm is developed. Experimental results show that our algorithm performs competitively with state-of-the-art denoising methods such as BM3D for natural images, and outperforms them significantly for piecewise smooth images."
                    },
                    {
                        "title": "Graph Spectral Image Processing",
                        "abstract": "Recent advent of graph signal processing (GSP) has spurred intensive studies of signals that live naturally on irregular data kernels described by graphs (e.g., social networks, wireless sensor networks). Though a digital image contains pixels that reside on a regularly sampled 2D grid, if one can design an appropriate underlying graph connecting pixels with weights that reflect the image structure, then one can interpret the image (or image patch) as a signal on a graph, and apply GSP tools for processing and analysis of the signal in graph spectral domain. In this article, we overview recent graph spectral techniques in GSP specifically for image / video processing. The topics covered include image compression, image restoration, image filtering and image segmentation."
                    },
                    {
                        "title": "3D Point Cloud Super-Resolution via Graph Total Variation on Surface Normals",
                        "abstract": "Point cloud is a collection of 3D coordinates that are discrete geometric samples of an object's 2D surfaces. Using a low-cost 3D scanner to acquire data means that point clouds are often in lower resolution than desired for rendering on high-resolution displays. Building on recent advances in graph signal processing, we design a local algorithm for 3D point cloud super-resolution (SR). First, we initialize new points at centroids of local triangles formed using the low-resolution point cloud, and connect all points using a k-nearestneighbor graph. Then, to establish a linear relationship between surface normals and 3D point coordinates, we perform bipartite graph approximation to divide all nodes into two disjoint sets, which are optimized alternately until convergence. For each node set, to promote piecewise smooth (PWS) 2D surfaces, we design a graph total variation (GTV) objective for nearby surface normals, under the constraint that coordinates of the original points are preserved. We pursue an augmented Lagrangian approach to tackle the optimization, and solve the unconstrained equivalent using the alternating method of multipliers (ADMM). Extensive experiments show that our proposed point cloud SR algorithm outperforms competing schemes objectively and subjectively for a large variety of point clouds."
                    },
                    {
                        "title": "Adaptive Non-Rigid Inpainting of 3D Point Cloud Geometry",
                        "abstract": "In this letter, we introduce several algorithms for geometry inpainting of 3D point clouds with large holes. The algorithms are examplar-based: hole filling is performed iteratively using templates near the hole boundary to find the best matching regions elsewhere in the cloud, from where existing points are transferred to the hole. We propose two improvements over the previous work on exemplar-based hole filling. The first one is adaptive template size selection in each iteration, which simultaneously leads to higher accuracy and lower execution time. The second improvement is a non-rigid transformation to better align the candidate set of points with the template before the point transfer, which leads to even higher accuracy. We demonstrate the algorithms' ability to fill holes that are difficult or impossible to fill by existing methods."
                    },
                    {
                        "title": "Deep Graph Laplacian Regularization for Robust Denoising of Real Images",
                        "abstract": "Recent developments in deep learning have revolutionized the paradigm of image restoration. However, its applications on real image denoising are still limited, due to its sensitivity to training data and the complex nature of real image noise. In this work, we combine the robustness merit of model-based approaches and the learning power of data-driven approaches for real image denoising. Specifically, by integrating graph Laplacian regularization as a trainable module into a deep learning framework, we are less susceptible to overfitting than pure CNN-based approaches, achieving higher robustness to small datasets and cross-domain denoising. First, a sparse neighborhood graph is built from the output of a convolutional neural network (CNN). Then the image is restored by solving an unconstrained quadratic programming problem, using a corresponding graph Laplacian regularizer as a prior term. The proposed restoration pipeline is fully differentiable and hence can be end-to-end trained. Experimental results demonstrate that our work is less prone to overfitting given small training data. It is also endowed with strong cross-domain generalization power, outperforming the state-of-the-art approaches by a remarkable margin."
                    },
                    {
                        "title": "Low-complexity Graph Sampling with Noise and Signal Reconstruction via Neumann Series",
                        "abstract": "Graph sampling addresses the problem of selecting a node subset in a graph to collect samples, so that a K-bandlimited signal can be reconstructed in high fidelity. Assuming an independent and identically distributed (i.i.d.) noise model, minimizing the expected mean square error (MMSE) leads to the known A-optimality criterion for graph sampling, which is expensive to compute and difficult to optimize. In this paper, we propose an augmented objective based on Neumann series that well approximates the original criterion and is amenable to greedy optimization. Specifically, we show that a shifted A-optimal criterion can be equivalently written as a function of an ideal low-pass (LP) graph filter, which in turn can be approximated efficiently via fast graph Fourier transform (FGFT). Minimizing the new objective, we select nodes greedily without large matrix inversions using a matrix inverse lemma. Further, for the dynamic network case where node availability varies across time, we propose an extended sampling strategy that replaces offline samples one-by-one in the selected set. For signal reconstruction, we propose an accompanied biased signal recovery strategy that reuses the approximated filter from sampling. Experiments show that our reconstruction is more robust to large noise than the least square (LS) solution, and our sampling strategy far outperforms several existing schemes."
                    },
                    {
                        "title": "Robust Deep Graph Based Learning for Binary Classification",
                        "abstract": "Convolutional neural network (CNN)-based feature learning has become state of the art, since given sufficient training data, CNN can significantly outperform traditional methods for various classification tasks. However, feature learning becomes more difficult if some training labels are noisy. With traditional regularization techniques, CNN often overfits to the noisy training labels, resulting in sub-par classification performance. In this paper, we propose a robust binary classifier, based on CNNs, to learn deep metric functions, which are then used to construct an optimal underlying graph structure used to clean noisy labels via graph Laplacian regularization (GLR). GLR is posed as a convex maximum a posteriori (MAP) problem solved via convex quadratic programming (QP). To penalize samples around the decision boundary, we propose two regularized loss functions for semi-supervised learning. The binary classification experiments on three datasets, varying in number and type of features, demonstrate that given a noisy training dataset, our proposed networks outperform several state-of-the-art classifiers, including label-noise robust support vector machine, CNNs with three different robust loss functions, model-based GLR, and dynamic graph CNN classifiers."
                    },
                    {
                        "title": "Unrolling of Deep Graph Total Variation for Image Denoising",
                        "abstract": "While deep learning (DL) architectures like convolutional neural networks (CNNs) have enabled effective solutions in image denoising, in general their implementations overly rely on training data, lack interpretability, and require tuning of a large parameter set. In this paper, we combine classical graph signal filtering with deep feature learning into a competitive hybrid design -- one that utilizes interpretable analytical low-pass graph filters and employs 80% fewer network parameters than state-of-the-art DL denoising scheme DnCNN. Specifically, to construct a suitable similarity graph for graph spectral filtering, we first adopt a CNN to learn feature representations per pixel, and then compute feature distances to establish edge weights. Given a constructed graph, we next formulate a convex optimization problem for denoising using a graph total variation (GTV) prior. Via a $l_1$ graph Laplacian reformulation, we interpret its solution in an iterative procedure as a graph low-pass filter and derive its frequency response. For fast filter implementation, we realize this response using a Lanczos approximation. Experimental results show that in the case of statistical mistmatch, our algorithm outperformed DnCNN by up to 3dB in PSNR."
                    },
                    {
                        "title": "Fast & Robust Image Interpolation using Gradient Graph Laplacian Regularizer",
                        "abstract": "In the graph signal processing (GSP) literature, it has been shown that signal-dependent graph Laplacian regularizer (GLR) can efficiently promote piecewise constant (PWC) signal reconstruction for various image restoration tasks. However, for planar image patches, like total variation (TV), GLR may suffer from the well-known \"staircase\" effect. To remedy this problem, we generalize GLR to gradient graph Laplacian regularizer (GGLR) that provably promotes piecewise planar (PWP) signal reconstruction for the image interpolation problem -- a 2D grid with random missing pixels that requires completion. Specifically, we first construct two higher-order gradient graphs to connect local horizontal and vertical gradients. Each local gradient is estimated using structure tensor, which is robust using known pixels in a small neighborhood, mitigating the problem of larger noise variance when computing gradient of gradients. Moreover, unlike total generalized variation (TGV), GGLR retains the quadratic form of GLR, leading to an unconstrained quadratic programming (QP) problem per iteration that can be solved quickly using conjugate gradient (CG). We derive the means-square-error minimizing weight parameter for GGLR, trading off bias and variance of the signal estimate. Experiments show that GGLR outperformed competing schemes in interpolation quality for severely damaged images at a reduced complexity."
                    },
                    {
                        "title": "Fast Graph Sampling for Short Video Summarization using Gershgorin Disc Alignment",
                        "abstract": "We study the problem of efficiently summarizing a short video into several keyframes, leveraging recent progress in fast graph sampling. Specifically, we first construct a similarity path graph (SPG) $\\mathcal{G}$, represented by graph Laplacian matrix $\\mathbf{L}$, where the similarities between adjacent frames are encoded as positive edge weights. We show that maximizing the smallest eigenvalue $\\lambda_{\\min}(\\mathbf{B})$ of a coefficient matrix $\\mathbf{B} = \\text{diag}(\\mathbf{a}) + \\mu \\mathbf{L}$, where $\\mathbf{a}$ is the binary keyframe selection vector, is equivalent to minimizing a worst-case signal reconstruction error. We prove that, after partitioning $\\mathcal{G}$ into $Q$ sub-graphs $\\{\\mathcal{G}^q\\}^Q_{q=1}$, the smallest Gershgorin circle theorem (GCT) lower bound of $Q$ corresponding coefficient matrices -- $\\min_q \\lambda^-_{\\min}(\\mathbf{B}^q)$ -- is a lower bound for $\\lambda_{\\min}(\\mathbf{B})$. This inspires a fast graph sampling algorithm to iteratively partition $\\mathcal{G}$ into $Q$ sub-graphs using $Q$ samples (keyframes), while maximizing $\\lambda^-_{\\min}(\\mathbf{B}^q)$ for each sub-graph $\\mathcal{G}^q$. Experimental results show that our algorithm achieves comparable video summarization performance as state-of-the-art methods, at a substantially reduced complexity."
                    },
                    {
                        "title": "Fast Computation of Generalized Eigenvectors for Manifold Graph Embedding",
                        "abstract": "Our goal is to efficiently compute low-dimensional latent coordinates for nodes in an input graph -- known as graph embedding -- for subsequent data processing such as clustering. Focusing on finite graphs that are interpreted as uniform samples on continuous manifolds (called manifold graphs), we leverage existing fast extreme eigenvector computation algorithms for speedy execution. We first pose a generalized eigenvalue problem for sparse matrix pair $(\\A,\\B)$, where $\\A = \\L - \\mu \\Q + \\epsilon \\I$ is a sum of graph Laplacian $\\L$ and disconnected two-hop difference matrix $\\Q$. Eigenvector $\\v$ minimizing Rayleigh quotient $\\frac{\\v^{\\top} \\A \\v}{\\v^{\\top} \\v}$ thus minimizes $1$-hop neighbor distances while maximizing distances between disconnected $2$-hop neighbors, preserving graph structure. Matrix $\\B = \\text{diag}(\\{\\b_i\\})$ that defines eigenvector orthogonality is then chosen so that boundary / interior nodes in the sampling domain have the same generalized degrees. $K$-dimensional latent vectors for the $N$ graph nodes are the first $K$ generalized eigenvectors for $(\\A,\\B)$, computed in $\\cO(N)$ using LOBPCG, where $K \\ll N$. Experiments show that our embedding is among the fastest in the literature, while producing the best clustering performance for manifold graphs."
                    },
                    {
                        "title": "Sparse Graph Learning with Spectrum Prior for Deep Graph Convolutional Networks",
                        "abstract": "A graph convolutional network (GCN) employs a graph filtering kernel tailored for data with irregular structures. However, simply stacking more GCN layers does not improve performance; instead, the output converges to an uninformative low-dimensional subspace, where the convergence rate is characterized by the graph spectrum -- this is the known over-smoothing problem in GCN. In this paper, we propose a sparse graph learning algorithm incorporating a new spectrum prior to compute a graph topology that circumvents over-smoothing while preserving pairwise correlations inherent in data. Specifically, based on a spectral analysis of multilayer GCN output, we derive a spectrum prior for the graph Laplacian matrix $\\mathbf{L}$ to robustify the model expressiveness against over-smoothing. Then, we formulate a sparse graph learning problem with the spectrum prior, solved efficiently via block coordinate descent (BCD). Moreover, we optimize the weight parameter trading off the fidelity term with the spectrum prior, based on data smoothness on the original graph learned without spectrum manipulation. The output $\\mathbf{L}$ is then normalized for supervised GCN training. Experiments show that our proposal produced deeper GCNs and higher prediction accuracy for regression and classification tasks compared to competing schemes."
                    },
                    {
                        "title": "Unsupervised Graph Spectral Feature Denoising for Crop Yield Prediction",
                        "abstract": "Prediction of annual crop yields at a county granularity is important for national food production and price stability. In this paper, towards the goal of better crop yield prediction, leveraging recent graph signal processing (GSP) tools to exploit spatial correlation among neighboring counties, we denoise relevant features via graph spectral filtering that are inputs to a deep learning prediction model. Specifically, we first construct a combinatorial graph with edge weights that encode county-to-county similarities in soil and location features via metric learning. We then denoise features via a maximum a posteriori (MAP) formulation with a graph Laplacian regularizer (GLR). We focus on the challenge to estimate the crucial weight parameter $\\mu$, trading off the fidelity term and GLR, that is a function of noise variance in an unsupervised manner. We first estimate noise variance directly from noise-corrupted graph signals using a graph clique detection (GCD) procedure that discovers locally constant regions. We then compute an optimal $\\mu$ minimizing an approximate mean square error function via bias-variance analysis. Experimental results from collected USDA data show that using denoised features as input, performance of a crop yield prediction model can be improved noticeably."
                    },
                    {
                        "title": "Graph Sparsification for GCN Towards Optimal Crop Yield Predictions",
                        "abstract": "In agronomics, predicting crop yield at a per field/county granularity is important for farmers to minimize uncertainty and plan seeding for the next crop cycle. While state-of-the-art prediction techniques employ graph convolutional nets (GCN) to predict future crop yields given relevant features and crop yields of previous years, a dense underlying graph kernel requires long training and execution time. In this paper, we propose a graph sparsification method based on the Fiedler number to remove edges from a complete graph kernel, in order to lower the complexity of GCN training/execution. Specifically, we first show that greedily removing an edge at a time that induces the minimal change in the second eigenvalue leads to a sparse graph with good GCN performance. We then propose a fast method to choose an edge for removal per iteration based on an eigenvalue perturbation theorem. Experiments show that our Fiedler-based method produces a sparse graph with good GCN performance compared to other graph sparsification schemes in crop yield prediction."
                    },
                    {
                        "title": "Retinex-based Image Denoising / Contrast Enhancement using Gradient Graph Laplacian Regularizer",
                        "abstract": "Images captured in poorly lit conditions are often corrupted by acquisition noise. Leveraging recent advances in graph-based regularization, we propose a fast Retinex-based restoration scheme that denoises and contrast-enhances an image. Specifically, by Retinex theory we first assume that each image pixel is a multiplication of its reflectance and illumination components. We next assume that the reflectance and illumination components are piecewise constant (PWC) and continuous piecewise planar (PWP) signals, which can be recovered via graph Laplacian regularizer (GLR) and gradient graph Laplacian regularizer (GGLR) respectively. We formulate quadratic objectives regularized by GLR and GGLR, which are minimized alternately until convergence by solving linear systems -- with improved condition numbers via proposed preconditioners -- via conjugate gradient (CG) efficiently. Experimental results show that our algorithm achieves competitive visual image quality while reducing computation complexity noticeably."
                    },
                    {
                        "title": "Mixed Graph Signal Analysis of Joint Image Denoising / Interpolation",
                        "abstract": "A noise-corrupted image often requires interpolation. Given a linear denoiser and a linear interpolator, when should the operations be independently executed in separate steps, and when should they be combined and jointly optimized? We study joint denoising / interpolation of images from a mixed graph filtering perspective: we model denoising using an undirected graph, and interpolation using a directed graph. We first prove that, under mild conditions, a linear denoiser is a solution graph filter to a maximum a posteriori (MAP) problem regularized using an undirected graph smoothness prior, while a linear interpolator is a solution to a MAP problem regularized using a directed graph smoothness prior. Next, we study two variants of the joint interpolation / denoising problem: a graph-based denoiser followed by an interpolator has an optimal separable solution, while an interpolator followed by a denoiser has an optimal non-separable solution. Experiments show that our joint denoising / interpolation method outperformed separate approaches noticeably."
                    },
                    {
                        "title": "Navigation domain representation for interactive multiview imaging",
                        "abstract": "Enabling users to interactively navigate through different viewpoints of a static scene is a new interesting functionality in 3D streaming systems. While it opens exciting perspectives towards rich multimedia applications, it requires the design of novel representations and coding techniques in order to solve the new challenges imposed by interactive navigation. Interactivity clearly brings new design constraints: the encoder is unaware of the exact decoding process, while the decoder has to reconstruct information from incomplete subsets of data since the server can generally not transmit images for all possible viewpoints due to resource constrains. In this paper, we propose a novel multiview data representation that permits to satisfy bandwidth and storage constraints in an interactive multiview streaming system. In particular, we partition the multiview navigation domain into segments, each of which is described by a reference image and some auxiliary information. The auxiliary information enables the client to recreate any viewpoint in the navigation segment via view synthesis. The decoder is then able to navigate freely in the segment without further data request to the server; it requests additional data only when it moves to a different segment. We discuss the benefits of this novel representation in interactive navigation systems and further propose a method to optimize the partitioning of the navigation domain into independent segments, under bandwidth and storage constraints. Experimental results confirm the potential of the proposed representation; namely, our system leads to similar compression performance as classical inter-view coding, while it provides the high level of flexibility that is required for interactive streaming. Hence, our new framework represents a promising solution for 3D data representation in novel interactive multimedia services."
                    },
                    {
                        "title": "Random Walk Graph Laplacian based Smoothness Prior for Soft Decoding of JPEG Images",
                        "abstract": "Given the prevalence of JPEG compressed images, optimizing image reconstruction from the compressed format remains an important problem. Instead of simply reconstructing a pixel block from the centers of indexed DCT coefficient quantization bins (hard decoding), soft decoding reconstructs a block by selecting appropriate coefficient values within the indexed bins with the help of signal priors. The challenge thus lies in how to define suitable priors and apply them effectively.   In this paper, we combine three image priors---Laplacian prior for DCT coefficients, sparsity prior and graph-signal smoothness prior for image patches---to construct an efficient JPEG soft decoding algorithm. Specifically, we first use the Laplacian prior to compute a minimum mean square error (MMSE) initial solution for each code block. Next, we show that while the sparsity prior can reduce block artifacts, limiting the size of the over-complete dictionary (to lower computation) would lead to poor recovery of high DCT frequencies. To alleviate this problem, we design a new graph-signal smoothness prior (desired signal has mainly low graph frequencies) based on the left eigenvectors of the random walk graph Laplacian matrix (LERaG). Compared to previous graph-signal smoothness priors, LERaG has desirable image filtering properties with low computation overhead. We demonstrate how LERaG can facilitate recovery of high DCT frequencies of a piecewise smooth (PWS) signal via an interpretation of low graph frequency components as relaxed solutions to normalized cut in spectral clustering. Finally, we construct a soft decoding algorithm using the three signal priors with appropriate prior weights. Experimental results show that our proposal outperforms state-of-the-art soft decoding algorithms in both objective and subjective evaluations noticeably."
                    },
                    {
                        "title": "Blind Image Deblurring via Reweighted Graph Total Variation",
                        "abstract": "Blind image deblurring, i.e., deblurring without knowledge of the blur kernel, is a highly ill-posed problem. The problem can be solved in two parts: i) estimate a blur kernel from the blurry image, and ii) given estimated blur kernel, de-convolve blurry input to restore the target image. In this paper, by interpreting an image patch as a signal on a weighted graph, we first argue that a skeleton image---a proxy that retains the strong gradients of the target but smooths out the details---can be used to accurately estimate the blur kernel and has a unique bi-modal edge weight distribution. We then design a reweighted graph total variation (RGTV) prior that can efficiently promote bi-modal edge weight distribution given a blurry patch. However, minimizing a blind image deblurring objective with RGTV results in a non-convex non-differentiable optimization problem. We propose a fast algorithm that solves for the skeleton image and the blur kernel alternately. Finally with the computed blur kernel, recent non-blind image deblurring algorithms can be applied to restore the target image. Experimental results show that our algorithm can robustly estimate the blur kernel with large kernel size, and the reconstructed sharp image is competitive against the state-of-the-art methods."
                    },
                    {
                        "title": "Graph Metric Learning via Gershgorin Disc Alignment",
                        "abstract": "We propose a fast general projection-free metric learning framework, where the minimization objective $\\min_{\\textbf{M} \\in \\mathcal{S}} Q(\\textbf{M})$ is a convex differentiable function of the metric matrix $\\textbf{M}$, and $\\textbf{M}$ resides in the set $\\mathcal{S}$ of generalized graph Laplacian matrices for connected graphs with positive edge weights and node degrees. Unlike low-rank metric matrices common in the literature, $\\mathcal{S}$ includes the important positive-diagonal-only matrices as a special case in the limit. The key idea for fast optimization is to rewrite the positive definite cone constraint in $\\mathcal{S}$ as signal-adaptive linear constraints via Gershgorin disc alignment, so that the alternating optimization of the diagonal and off-diagonal terms in $\\textbf{M}$ can be solved efficiently as linear programs via Frank-Wolfe iterations. We prove that the Gershgorin discs can be aligned perfectly using the first eigenvector $\\textbf{v}$ of $\\textbf{M}$, which we update iteratively using Locally Optimal Block Preconditioned Conjugate Gradient (LOBPCG) with warm start as diagonal / off-diagonal terms are optimized. Experiments show that our efficiently computed graph metric matrices outperform metrics learned using competing methods in terms of classification tasks."
                    }
                ]
            },
            "8fc1f121-1e94-423a-bee1-ced3a42c80b7": {
                "pk": "8fc1f121-1e94-423a-bee1-ced3a42c80b7",
                "name": "Philip Chou",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2310.01583": {
        "paper_data": {
            "title": "An Investigation of Representation and Allocation Harms in Contrastive Learning",
            "url": "http://arxiv.org/abs/2310.01583v1",
            "arxiv_id": "2310.01583",
            "authors": [
                "Subha Maity",
                "Mayank Agarwal",
                "Mikhail Yurochkin",
                "Yuekai Sun"
            ],
            "abstract": "The effect of underrepresentation on the performance of minority groups is known to be a serious problem in supervised learning settings; however, it has been underexplored so far in the context of self-supervised learning (SSL). In this paper, we demonstrate that contrastive learning (CL), a popular variant of SSL, tends to collapse representations of minority groups with certain majority groups. We refer to this phenomenon as representation harm and demonstrate it on image and text datasets using the corresponding popular CL methods. Furthermore, our causal mediation analysis of allocation harm on a downstream classification task reveals that representation harm is partly responsible for it, thus emphasizing the importance of studying and mitigating representation harm. Finally, we provide a theoretical explanation for representation harm using a stochastic block model that leads to a representational neural collapse in a contrastive learning setting.",
            "introduction": "   1 Introduction  Figure 1: tSNE visualization of CL representations for balanced (left) and automobile under-represented (right) CIFAR10\u00a0data.    The negative impacts of underrepresentation on the performance of minority groups have been extensively studied in supervised learning. Most prominent reports of algorithmic biases in automated decision-making pipelines (Angwin et\u00a0al., 2016; O\u2019Neil, 2017; Dastin, 2018) focus on disparities in resource allocation by supervised learning algorithms, and there are theoretical models that seek to elucidate the mechanisms behind the disparities (Fang et\u00a0al., 2021b). Such disparities cause allocative harms to the samples from underrepresented groups, and there are many approaches based on resampling and reweighting (He et\u00a0al., 2008; Ando & Huang, 2017; Byrd & Lipton, 2019), enforcing invariance constraints (Agarwal et\u00a0al., 2018), re-calibrating the logits (Tian et\u00a0al., 2020) to alleviate such allocative harms in supervised learning algorithms.   In contrast to the voluminous literature on allocative harms, representation harms have received less attention. This is because representation harms are harder to measure and their effects are more diffuse and longer-term (O\u2019Neil, 2017). Despite recent advances in representation learning (Bengio et\u00a0al., 2013), there is a lack of research on algorithmic biases in representation learning, especially studies that look into representation harms they may cause. In this paper, we seek to fill this gap in the literature by investigating the effects of group underrepresentation in contrastive learning (CL) algorithms (Chen et\u00a0al., 2020; Chen & He, 2021; He et\u00a0al., 2020). CL is a popular instance of self-supervised learning (SSL), an approach for learning representations that can effectively leverage unlabeled samples to improve (downstream) model performances (Wang et\u00a0al., 2021; Babu et\u00a0al., 2021). However, these datasets often exhibit inherent imbalances (De-Arteaga et\u00a0al., 2019) and it is important to know whether training a CL model on such datasets may detrimentally affect the quality of representations and cause downstream allocative and representation harms.   Here is a preview of the results of our controlled study of underrepresentation with the CIFAR10\u00a0dataset. In Figure 1 we plot 2D t-SNE embeddings (Van\u00a0der Maaten & Hinton, 2008) of CL representations of vehicles in the dataset. We see that the underrepresentation of automobile images in CL training causes their representation to cluster with those of trucks. This is an instance of stereotyping (Abbasi et\u00a0al., 2019) because samples from an underrepresented group are lumped together with those of a similar majority group; thus erasing the unique characteristics of the underrepresented. This is a form of representation harm (Crawford, 2017), which can lead to downstream allocation harms (e.g.\u00a0missclassifications between automobiles and trucks). In this paper, we show that this mechanism of representation to downstream allocation harms is a general. This complements prior works, which show that underrepresentation leads to allocation harms without identifying the underlying mechanism. The rest of the paper is organized as follows.   1.  In Section 2 we empirically show that underrepresentation leads to representation harms, as the CL representations of underrepresented groups collapse to semantically similar groups.    2.  In Section 3 we develop a simple model of CL on graphs that exhibits a collapse of (the learned representations of) underrepresented groups to (those of) semantically similar groups. This suggests that the representation harms we measured are intrinsic to CL representations.    3.  In Section",
            "references": [
                {
                    "title": "Social Biases through the Text-to-Image Generation Lens",
                    "abstract": "Text-to-Image (T2I) generation is enabling new applications that support creators, designers, and general end users of productivity software by generating illustrative content with high photorealism starting from a given descriptive text as a prompt. Such models are however trained on massive amounts of web data, which surfaces the peril of potential harmful biases that may leak in the generation process itself. In this paper, we take a multi-dimensional approach to studying and quantifying common social biases as reflected in the generated images, by focusing on how occupations, personality traits, and everyday situations are depicted across representations of (perceived) gender, age, race, and geographical location. Through an extensive set of both automated and human evaluation experiments we present findings for two popular T2I models: DALLE-v2 and Stable Diffusion. Our results reveal that there exist severe occupational biases of neutral prompts majorly excluding groups of people from results for both models. Such biases can get mitigated by increasing the amount of specification in the prompt itself, although the prompting mitigation will not address discrepancies in image quality or other usages of the model or its representations in other scenarios. Further, we observe personality traits being associated with only a limited set of people at the intersection of race, gender, and age. Finally, an analysis of geographical location representations on everyday situations (e.g., park, food, weddings) shows that for most situations, images generated through default location-neutral prompts are closer and more similar to images generated for locations of United States and Germany."
                },
                {
                    "title": "The Hidden Uniform Cluster Prior in Self-Supervised Learning",
                    "abstract": "A successful paradigm in representation learning is to perform self-supervised pretraining using tasks based on mini-batch statistics (e.g., SimCLR, VICReg, SwAV, MSN). We show that in the formulation of all these methods is an overlooked prior to learn features that enable uniform clustering of the data. While this prior has led to remarkably semantic representations when pretraining on class-balanced data, such as ImageNet, we demonstrate that it can hamper performance when pretraining on class-imbalanced data. By moving away from conventional uniformity priors and instead preferring power-law distributed feature clusters, we show that one can improve the quality of the learned representations on real-world class-imbalanced datasets. To demonstrate this, we develop an extension of the Masked Siamese Networks (MSN) method to support the use of arbitrary features priors."
                },
                {
                    "title": "Outlier-Robust Group Inference via Gradient Space Clustering",
                    "abstract": "Traditional machine learning models focus on achieving good performance on the overall training distribution, but they often underperform on minority groups. Existing methods can improve the worst-group performance, but they can have several limitations: (i) they require group annotations, which are often expensive and sometimes infeasible to obtain, and/or (ii) they are sensitive to outliers. Most related works fail to solve these two issues simultaneously as they focus on conflicting perspectives of minority groups and outliers. We address the problem of learning group annotations in the presence of outliers by clustering the data in the space of gradients of the model parameters. We show that data in the gradient space has a simpler structure while preserving information about minority groups and outliers, making it suitable for standard clustering methods like DBSCAN. Extensive experiments demonstrate that our method significantly outperforms state-of-the-art both in terms of group identification and downstream worst-group performance."
                },
                {
                    "title": "Self-Supervised Learning for Videos: A Survey",
                    "abstract": "The remarkable success of deep learning in various domains relies on the availability of large-scale annotated datasets. However, obtaining annotations is expensive and requires great effort, which is especially challenging for videos. Moreover, the use of human-generated annotations leads to models with biased learning and poor domain generalization and robustness. As an alternative, self-supervised learning provides a way for representation learning that does not require annotations and has shown promise in both image and video domains. In contrast to the image domain, learning video representations are more challenging due to the temporal dimension, bringing in motion and other environmental dynamics. This also provides opportunities for video-exclusive ideas that advance self-supervised learning in the video and multimodal domains. In this survey, we provide a review of existing approaches on self-supervised learning focusing on the video domain. We summarize these methods into four different categories based on their learning objectives: (1) pretext tasks, (2) generative learning, (3) contrastive learning, and (4) cross-modal agreement. We further introduce the commonly used datasets, downstream evaluation tasks, insights into the limitations of existing works, and the potential future directions in this area."
                },
                {
                    "title": "Contrastive Learning with Boosted Memorization",
                    "abstract": "Self-supervised learning has achieved a great success in the representation learning of visual and textual data. However, the current methods are mainly validated on the well-curated datasets, which do not exhibit the real-world long-tailed distribution. Recent attempts to consider self-supervised long-tailed learning are made by rebalancing in the loss perspective or the model perspective, resembling the paradigms in the supervised long-tailed learning. Nevertheless, without the aid of labels, these explorations have not shown the expected significant promise due to the limitation in tail sample discovery or the heuristic structure design. Different from previous works, we explore this direction from an alternative perspective, i.e., the data perspective, and propose a novel Boosted Contrastive Learning (BCL) method. Specifically, BCL leverages the memorization effect of deep neural networks to automatically drive the information discrepancy of the sample views in contrastive learning, which is more efficient to enhance the long-tailed learning in the label-unaware context. Extensive experiments on a range of benchmark datasets demonstrate the effectiveness of BCL over several state-of-the-art methods. Our code is available at https://github.com/MediaBrain-SJTU/BCL."
                },
                {
                    "title": "Self-Supervised Learning for Recommender Systems: A Survey",
                    "abstract": "In recent years, neural architecture-based recommender systems have achieved tremendous success, but they still fall short of expectation when dealing with highly sparse data. Self-supervised learning (SSL), as an emerging technique for learning from unlabeled data, has attracted considerable attention as a potential solution to this issue. This survey paper presents a systematic and timely review of research efforts on self-supervised recommendation (SSR). Specifically, we propose an exclusive definition of SSR, on top of which we develop a comprehensive taxonomy to divide existing SSR methods into four categories: contrastive, generative, predictive, and hybrid. For each category, we elucidate its concept and formulation, the involved methods, as well as its pros and cons. Furthermore, to facilitate empirical comparison, we release an open-source library SELFRec (https://github.com/Coder-Yu/SELFRec), which incorporates a wide range of SSR models and benchmark datasets. Through rigorous experiments using this library, we derive and report some significant findings regarding the selection of self-supervised signals for enhancing recommendation. Finally, we shed light on the limitations in the current research and outline the future research directions."
                },
                {
                    "title": "Limitations of Neural Collapse for Understanding Generalization in Deep Learning",
                    "abstract": "The recent work of Papyan, Han,&Donoho (2020) presented an intriguing\"Neural Collapse\"phenomenon, showing a structural property of interpolating classifiers in the late stage of training. This opened a rich area of exploration studying this phenomenon. Our motivation is to study the upper limits of this research program: How far will understanding Neural Collapse take us in understanding deep learning? First, we investigate its role in generalization. We refine the Neural Collapse conjecture into two separate conjectures: collapse on the train set (an optimization property) and collapse on the test distribution (a generalization property). We find that while Neural Collapse often occurs on the train set, it does not occur on the test set. We thus conclude that Neural Collapse is primarily an optimization phenomenon, with as-yet-unclear connections to generalization. Second, we investigate the role of Neural Collapse in feature learning. We show simple, realistic experiments where training longer leads to worse last-layer features, as measured by transfer-performance on a downstream task. This suggests that neural collapse is not always desirable for representation learning, as previously claimed. Finally, we give preliminary evidence of a\"cascading collapse\"phenomenon, wherein some form of Neural Collapse occurs not only for the last layer, but in earlier layers as well. We hope our work encourages the community to continue the rich line of Neural Collapse research, while also considering its inherent limitations."
                },
                {
                    "title": "On the Role of Neural Collapse in Transfer Learning",
                    "abstract": "We study the ability of foundation models to learn representations for classification that are transferable to new, unseen classes. Recent results in the literature show that representations learned by a single classifier over many classes are competitive on few-shot learning problems with representations learned by special-purpose algorithms designed for such problems. In this paper we provide an explanation for this behavior based on the recently observed phenomenon that the features learned by overparameterized classification networks show an interesting clustering property, called neural collapse. We demonstrate both theoretically and empirically that neural collapse generalizes to new samples from the training classes, and -- more importantly -- to new classes as well, allowing foundation models to provide feature maps that work well in transfer learning and, specifically, in the few-shot setting."
                },
                {
                    "title": "XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale",
                    "abstract": "This paper presents XLS-R, a large-scale model for cross-lingual speech representation learning based on wav2vec 2.0. We train models with up to 2B parameters on nearly half a million hours of publicly available speech audio in 128 languages, an order of magnitude more public data than the largest known prior work. Our evaluation covers a wide range of tasks, domains, data regimes and languages, both high and low-resource. On the CoVoST-2 speech translation benchmark, we improve the previous state of the art by an average of 7.4 BLEU over 21 translation directions into English. For speech recognition, XLS-R improves over the best known prior work on BABEL, MLS, CommonVoice as well as VoxPopuli, lowering error rates by 14-34% relative on average. XLS-R also sets a new state of the art on VoxLingua107 language identification. Moreover, we show that with sufficient model size, cross-lingual pretraining can outperform English-only pretraining when translating English speech into other languages, a setting which favors monolingual pretraining. We hope XLS-R can help to improve speech processing tasks for many more languages of the world."
                },
                {
                    "title": "Simple data balancing achieves competitive worst-group-accuracy",
                    "abstract": "We study the problem of learning classifiers that perform well across (known or unknown) groups of data. After observing that common worst-group-accuracy datasets suffer from substantial imbalances, we set out to compare state-of-the-art methods to simple balancing of classes and groups by either subsampling or reweighting data. Our results show that these data balancing baselines achieve state-of-the-art-accuracy, while being faster to train and requiring no additional hyper-parameters. In addition, we highlight that access to group information is most critical for model selection purposes, and not so much during training. All in all, our findings beg closer examination of benchmarks and methods for research in worst-group-accuracy optimization."
                },
                {
                    "title": "Self-supervised Learning is More Robust to Dataset Imbalance",
                    "abstract": "Self-supervised learning (SSL) is a scalable way to learn general visual representations since it learns without labels. However, large-scale unlabeled datasets in the wild often have long-tailed label distributions, where we know little about the behavior of SSL. In this work, we systematically investigate self-supervised learning under dataset imbalance. First, we find out via extensive experiments that off-the-shelf self-supervised representations are already more robust to class imbalance than supervised representations. The performance gap between balanced and imbalanced pre-training with SSL is significantly smaller than the gap with supervised learning, across sample sizes, for both in-domain and, especially, out-of-domain evaluation. Second, towards understanding the robustness of SSL, we hypothesize that SSL learns richer features from frequent data: it may learn label-irrelevant-but-transferable features that help classify the rare classes and downstream tasks. In contrast, supervised learning has no incentive to learn features irrelevant to the labels from frequent examples. We validate this hypothesis with semi-synthetic experiments and theoretical analyses on a simplified setting. Third, inspired by the theoretical insights, we devise a re-weighted regularization technique that consistently improves the SSL representation quality on imbalanced datasets with several evaluation criteria, closing the small gap between balanced and imbalanced datasets with the same number of examples."
                },
                {
                    "title": "Just Train Twice: Improving Group Robustness without Training Group Information",
                    "abstract": "Standard training via empirical risk minimization (ERM) can produce models that achieve high accuracy on average but low accuracy on certain groups, especially in the presence of spurious correlations between the input and label. Prior approaches that achieve high worst-group accuracy, like group distributionally robust optimization (group DRO) require expensive group annotations for each training point, whereas approaches that do not use such group annotations typically achieve unsatisfactory worst-group accuracy. In this paper, we propose a simple two-stage approach, JTT, that first trains a standard ERM model for several epochs, and then trains a second model that upweights the training examples that the first model misclassified. Intuitively, this upweights examples from groups on which standard ERM models perform poorly, leading to improved worst-group performance. Averaged over four image classification and natural language processing tasks with spurious correlations, JTT closes 75% of the gap in worst-group accuracy between standard ERM and group DRO, while only requiring group annotations on a small validation set in order to tune hyperparameters."
                },
                {
                    "title": "HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units",
                    "abstract": "Self-supervised approaches for speech representation learning are challenged by three unique problems: (1) there are multiple sound units in each input utterance, (2) there is no lexicon of input sound units during the pre-training phase, and (3) sound units have variable lengths with no explicit segmentation. To deal with these three problems, we propose the Hidden-Unit BERT (HuBERT) approach for self-supervised speech representation learning, which utilizes an offline clustering step to provide aligned target labels for a BERT-like prediction loss. A key ingredient of our approach is applying the prediction loss over the masked regions only, which forces the model to learn a combined acoustic and language model over the continuous inputs. HuBERT relies primarily on the consistency of the unsupervised clustering step rather than the intrinsic quality of the assigned cluster labels. Starting with a simple k-means teacher of 100 clusters, and using two iterations of clustering, the HuBERT model either matches or improves upon the state-of-the-art wav2vec 2.0 performance on the Librispeech (960 h) and Libri-light (60,000 h) benchmarks with 10 min, 1 h, 10 h, 100 h, and 960 h fine-tuning subsets. Using a 1B parameter model, HuBERT shows up to 19% and 13% relative WER reduction on the more challenging dev-other and test-other evaluation subsets.12"
                },
                {
                    "title": "Provable Guarantees for Self-Supervised Deep Learning with Spectral Contrastive Loss",
                    "abstract": "Recent works in self-supervised learning have advanced the state-of-the-art by relying on the contrastive learning paradigm, which learns representations by pushing positive pairs, or similar examples from the same class, closer together while keeping negative pairs far apart. Despite the empirical successes, theoretical foundations are limited -- prior analyses assume conditional independence of the positive pairs given the same class label, but recent empirical applications use heavily correlated positive pairs (i.e., data augmentations of the same image). Our work analyzes contrastive learning without assuming conditional independence of positive pairs using a novel concept of the augmentation graph on data. Edges in this graph connect augmentations of the same data, and ground-truth classes naturally form connected sub-graphs. We propose a loss that performs spectral decomposition on the population augmentation graph and can be succinctly written as a contrastive learning objective on neural net representations. Minimizing this objective leads to features with provable accuracy guarantees under linear probe evaluation. By standard generalization bounds, these accuracy guarantees also hold when minimizing the training contrastive loss. Empirically, the features learned by our objective can match or outperform several strong baselines on benchmark vision datasets. In all, this work provides the first provable analysis for contrastive learning where guarantees for linear probe evaluation can apply to realistic empirical settings."
                },
                {
                    "title": "Conditional Contrastive Learning for Improving Fairness in Self-Supervised Learning",
                    "abstract": "Contrastive self-supervised learning (SSL) learns an embedding space that maps similar data pairs closer and dissimilar data pairs farther apart. Despite its success, one issue has been overlooked: the fairness aspect of representations learned using contrastive SSL. Without mitigation, contrastive SSL techniques can incorporate sensitive information such as gender or race and cause potentially unfair predictions on downstream tasks. In this paper, we propose a Conditional Contrastive Learning (CCL) approach to improve the fairness of contrastive SSL methods. Our approach samples positive and negative pairs from distributions conditioning on the sensitive attribute, or empirically speaking, sampling positive and negative pairs from the same gender or the same race. We show that our approach provably maximizes the conditional mutual information between the learned representations of the positive pairs, and reduces the effect of the sensitive attribute by taking it as the conditional variable. On seven fairness and vision datasets, we empirically demonstrate that the proposed approach achieves state-of-the-art downstream performances compared to unsupervised baselines and significantly improves the fairness of contrastive SSL models on multiple fairness metrics."
                },
                {
                    "title": "Neural Collapse Under MSE Loss: Proximity to and Dynamics on the Central Path",
                    "abstract": "The recently discovered Neural Collapse (NC) phenomenon occurs pervasively in today's deep net training paradigm of driving cross-entropy (CE) loss towards zero. During NC, last-layer features collapse to their class-means, both classifiers and class-means collapse to the same Simplex Equiangular Tight Frame, and classifier behavior collapses to the nearest-class-mean decision rule. Recent works demonstrated that deep nets trained with mean squared error (MSE) loss perform comparably to those trained with CE. As a preliminary, we empirically establish that NC emerges in such MSE-trained deep nets as well through experiments on three canonical networks and five benchmark datasets. We provide, in a Google Colab notebook, PyTorch code for reproducing MSE-NC and CE-NC: at https://colab.research.google.com/github/neuralcollapse/neuralcollapse/blob/main/neuralcollapse.ipynb. The analytically-tractable MSE loss offers more mathematical opportunities than the hard-to-analyze CE loss, inspiring us to leverage MSE loss towards the theoretical investigation of NC. We develop three main contributions: (I) We show a new decomposition of the MSE loss into (A) terms directly interpretable through the lens of NC and which assume the last-layer classifier is exactly the least-squares classifier; and (B) a term capturing the deviation from this least-squares classifier. (II) We exhibit experiments on canonical datasets and networks demonstrating that term-(B) is negligible during training. This motivates us to introduce a new theoretical construct: the central path, where the linear classifier stays MSE-optimal for feature activations throughout the dynamics. (III) By studying renormalized gradient flow along the central path, we derive exact dynamics that predict NC."
                },
                {
                    "title": "A Geometric Analysis of Neural Collapse with Unconstrained Features",
                    "abstract": "We provide the first global optimization landscape analysis of $Neural\\;Collapse$ -- an intriguing empirical phenomenon that arises in the last-layer classifiers and features of neural networks during the terminal phase of training. As recently reported by Papyan et al., this phenomenon implies that ($i$) the class means and the last-layer classifiers all collapse to the vertices of a Simplex Equiangular Tight Frame (ETF) up to scaling, and ($ii$) cross-example within-class variability of last-layer activations collapses to zero. We study the problem based on a simplified $unconstrained\\;feature\\;model$, which isolates the topmost layers from the classifier of the neural network. In this context, we show that the classical cross-entropy loss with weight decay has a benign global landscape, in the sense that the only global minimizers are the Simplex ETFs while all other critical points are strict saddles whose Hessian exhibit negative curvature directions. In contrast to existing landscape analysis for deep neural networks which is often disconnected from practice, our analysis of the simplified model not only does it explain what kind of features are learned in the last layer, but it also shows why they can be efficiently optimized in the simplified settings, matching the empirical observations in practical deep network architectures. These findings could have profound implications for optimization, generalization, and robustness of broad interests. For example, our experiments demonstrate that one may set the feature dimension equal to the number of classes and fix the last-layer classifier to be a Simplex ETF for network training, which reduces memory cost by over $20\\%$ on ResNet18 without sacrificing the generalization performance."
                },
                {
                    "title": "SimCSE: Simple Contrastive Learning of Sentence Embeddings",
                    "abstract": "This paper presents SimCSE, a simple contrastive learning framework that greatly advances the state-of-the-art sentence embeddings. We first describe an unsupervised approach, which takes an input sentence and predicts itself in a contrastive objective, with only standard dropout used as noise. This simple method works surprisingly well, performing on par with previous supervised counterparts. We find that dropout acts as minimal data augmentation and removing it leads to a representation collapse. Then, we propose a supervised approach, which incorporates annotated pairs from natural language inference datasets into our contrastive learning framework, by using \u201centailment\u201d pairs as positives and \u201ccontradiction\u201d pairs as hard negatives. We evaluate SimCSE on standard semantic textual similarity (STS) tasks, and our unsupervised and supervised models using BERT base achieve an average of 76.3% and 81.6% Spearman\u2019s correlation respectively, a 4.2% and 2.2% improvement compared to previous best results. We also show\u2014both theoretically and empirically\u2014that contrastive learning objective regularizes pre-trained embeddings\u2019 anisotropic space to be more uniform, and it better aligns positive pairs when supervised signals are available."
                },
                {
                    "title": "Large-Scale Self- and Semi-Supervised Learning for Speech Translation",
                    "abstract": "In this paper, we improve speech translation (ST) through effectively leveraging large quantities of unlabeled speech and text data in different and complementary ways. We explore both pretraining and self-training by using the large Libri-Light speech audio corpus and language modeling with CommonCrawl. Our experiments improve over the previous state of the art by 2.6 BLEU on average on all four considered CoVoST 2 language pairs via a simple recipe of combining wav2vec 2.0 pretraining, a single iteration of self-training and decoding with a language model. Different to existing work, our approach does not leverage any other supervision than ST data. Code and models will be publicly released."
                },
                {
                    "title": "Graph Self-Supervised Learning: A Survey",
                    "abstract": "Deep learning on graphs has attracted significant interests recently. However, most of the works have focused on (semi-) supervised learning, resulting in shortcomings including heavy label reliance, poor generalization, and weak robustness. To address these issues, self-supervised learning (SSL), which extracts informative knowledge through well-designed pretext tasks without relying on manual labels, has become a promising and trending learning paradigm for graph data. Different from SSL on other domains like computer vision and natural language processing, SSL on graphs has an exclusive background, design ideas, and taxonomies. Under the umbrella of graph self-supervised learning, we present a timely and comprehensive review of the existing approaches which employ SSL techniques for graph data. We construct a unified framework that mathematically formalizes the paradigm of graph SSL. According to the objectives of pretext tasks, we divide these approaches into four categories: generation-based, auxiliary property-based, contrast-based, and hybrid approaches. We further describe the applications of graph SSL across various research fields and summarize the commonly used datasets, evaluation benchmark, performance comparison and open-source codes of graph SSL. Finally, we discuss the remaining challenges and potential future directions in this research field."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Understanding self-supervised Learning Dynamics without Contrastive Pairs",
                    "abstract": "While contrastive approaches of self-supervised learning (SSL) learn representations by minimizing the distance between two augmented views of the same data point (positive pairs) and maximizing views from different data points (negative pairs), recent \\emph{non-contrastive} SSL (e.g., BYOL and SimSiam) show remarkable performance {\\it without} negative pairs, with an extra learnable predictor and a stop-gradient operation. A fundamental question arises: why do these methods not collapse into trivial representations? We answer this question via a simple theoretical study and propose a novel approach, DirectPred, that \\emph{directly} sets the linear predictor based on the statistics of its inputs, without gradient training. On ImageNet, it performs comparably with more complex two-layer non-linear predictors that employ BatchNorm and outperforms a linear predictor by $2.5\\%$ in 300-epoch training (and $5\\%$ in 60-epoch). DirectPred is motivated by our theoretical study of the nonlinear learning dynamics of non-contrastive SSL in simple linear networks. Our study yields conceptual insights into how non-contrastive SSL methods learn, how they avoid representational collapse, and how multiple factors, like predictor networks, stop-gradients, exponential moving averages, and weight decay all come into play. Our simple theory recapitulates the results of real-world ablation studies in both STL-10 and ImageNet. Code is released https://github.com/facebookresearch/luckmatters/tree/master/ssl."
                },
                {
                    "title": "Exploring deep neural networks via layer-peeled model: Minority collapse in imbalanced training",
                    "abstract": "Significance The remarkable development of deep learning over the past decade relies heavily on sophisticated heuristics and tricks. To better exploit its potential in the coming decade, perhaps a rigorous framework for reasoning about deep learning is needed, which, however, is not easy to build due to the intricate details of neural networks. For near-term purposes, a practical alternative is to develop a mathematically tractable surrogate model, yet maintaining many characteristics of neural networks. This paper proposes a model of this kind that we term the Layer-Peeled Model. The effectiveness of this model is evidenced by, among others, its ability to reproduce a known empirical pattern and to predict a hitherto-unknown phenomenon when training deep-learning models on imbalanced datasets. In this paper, we introduce the Layer-Peeled Model, a nonconvex, yet analytically tractable, optimization program, in a quest to better understand deep neural networks that are trained for a sufficiently long time. As the name suggests, this model is derived by isolating the topmost layer from the remainder of the neural network, followed by imposing certain constraints separately on the two parts of the network. We demonstrate that the Layer-Peeled Model, albeit simple, inherits many characteristics of well-trained neural networks, thereby offering an effective tool for explaining and predicting common empirical patterns of deep-learning training. First, when working on class-balanced datasets, we prove that any solution to this model forms a simplex equiangular tight frame, which, in part, explains the recently discovered phenomenon of neural collapse [V. Papyan, X. Y. Han, D. L. Donoho, Proc. Natl. Acad. Sci. U.S.A. 117, 24652\u201324663 (2020)]. More importantly, when moving to the imbalanced case, our analysis of the Layer-Peeled Model reveals a hitherto-unknown phenomenon that we term Minority Collapse, which fundamentally limits the performance of deep-learning models on the minority classes. In addition, we use the Layer-Peeled Model to gain insights into how to mitigate Minority Collapse. Interestingly, this phenomenon is first predicted by the Layer-Peeled Model before being confirmed by our computational experiments."
                },
                {
                    "title": "Neural Collapse with Cross-Entropy Loss",
                    "abstract": "We consider the variational problem of cross-entropy loss with $n$ feature vectors on a unit hypersphere in $\\mathbb{R}^d$. We prove that when $d \\geq n - 1$, the global minimum is given by the simplex equiangular tight frame, which justifies the neural collapse behavior. We also prove that as $n \\rightarrow \\infty$ with fixed $d$, the minimizing points will distribute uniformly on the hypersphere and show a connection with the frame potential of Benedetto & Fickus."
                },
                {
                    "title": "Exploring Simple Siamese Representation Learning",
                    "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code is made available.1"
                },
                {
                    "title": "Posterior Re-calibration for Imbalanced Datasets",
                    "abstract": "Neural Networks can perform poorly when the training label distribution is heavily imbalanced, as well as when the testing data differs from the training distribution. In order to deal with shift in the testing label distribution, which imbalance causes, we motivate the problem from the perspective of an optimal Bayes classifier and derive a post-training prior rebalancing technique that can be solved through a KL-divergence based optimization. This method allows a flexible post-training hyper-parameter to be efficiently tuned on a validation set and effectively modify the classifier margin to deal with this imbalance. We further combine this method with existing likelihood shift methods, re-interpreting them from the same Bayesian perspective, and demonstrating that our method can deal with both problems in a unified way. The resulting algorithm can be conveniently used on probabilistic classification problems agnostic to underlying architectures. Our results on six different datasets and five different architectures show state of art accuracy, including on large-scale imbalanced datasets such as iNaturalist for classification and Synthia for semantic segmentation. Please see this https URL for implementation."
                },
                {
                    "title": "Contrastive Representation Learning: A Framework and Review",
                    "abstract": "Contrastive Learning has recently received interest due to its success in self-supervised representation learning in the computer vision domain. However, the origins of Contrastive Learning date as far back as the 1990s and its development has spanned across many fields and domains including Metric Learning and natural language processing. In this paper, we provide a comprehensive literature review and we propose a general Contrastive Representation Learning framework that simplifies and unifies many different contrastive learning methods. We also provide a taxonomy for each of the components of contrastive learning in order to summarise it and distinguish it from other forms of machine learning. We then discuss the inductive biases which are present in any contrastive learning system and we analyse our framework under different views from various sub-fields of Machine Learning. Examples of how contrastive learning has been applied in computer vision, natural language processing, audio processing, and others, as well as in Reinforcement Learning are also presented. Finally, we discuss the challenges and some of the most promising future research directions ahead."
                },
                {
                    "title": "Contrastive learning, multi-view redundancy, and linear models",
                    "abstract": "Self-supervised learning is an empirically successful approach to unsupervised learning based on creating artificial supervised learning problems. A popular self-supervised approach to representation learning is contrastive learning, which leverages naturally occurring pairs of similar and dissimilar data points, or multiple views of the same data. This work provides a theoretical analysis of contrastive learning in the multi-view setting, where two views of each datum are available. The main result is that linear functions of the learned representations are nearly optimal on downstream prediction tasks whenever the two views provide redundant information about the label."
                },
                {
                    "title": "Prevalence of neural collapse during the terminal phase of deep learning training",
                    "abstract": "Significance Modern deep neural networks for image classification have achieved superhuman performance. Yet, the complex details of trained networks have forced most practitioners and researchers to regard them as black boxes with little that could be understood. This paper considers in detail a now-standard training methodology: driving the cross-entropy loss to zero, continuing long after the classification error is already zero. Applying this methodology to an authoritative collection of standard deepnets and datasets, we observe the emergence of a simple and highly symmetric geometry of the deepnet features and of the deepnet classifier, and we document important benefits that the geometry conveys\u2014thereby helping us understand an important component of the modern deep learning training paradigm. Modern practice for training classification deepnets involves a terminal phase of training (TPT), which begins at the epoch where training error first vanishes. During TPT, the training error stays effectively zero, while training loss is pushed toward zero. Direct measurements of TPT, for three prototypical deepnet architectures and across seven canonical classification datasets, expose a pervasive inductive bias we call neural collapse (NC), involving four deeply interconnected phenomena. (NC1) Cross-example within-class variability of last-layer training activations collapses to zero, as the individual activations themselves collapse to their class means. (NC2) The class means collapse to the vertices of a simplex equiangular tight frame (ETF). (NC3) Up to rescaling, the last-layer classifiers collapse to the class means or in other words, to the simplex ETF (i.e., to a self-dual configuration). (NC4) For a given activation, the classifier\u2019s decision collapses to simply choosing whichever class has the closest train class mean (i.e., the nearest class center [NCC] decision rule). The symmetric and very simple geometry induced by the TPT confers important benefits, including better generalization performance, better robustness, and better interpretability."
                },
                {
                    "title": "Predicting What You Already Know Helps: Provable Self-Supervised Learning",
                    "abstract": "Self-supervised representation learning solves auxiliary prediction tasks (known as pretext tasks), that do not require labeled data, to learn semantic representations. These pretext tasks are created solely using the input features, such as predicting a missing image patch, recovering the color channels of an image from context, or predicting missing words, yet predicting this $known\\ $information helps in learning representations effective for downstream prediction tasks. This paper posits a mechanism based on conditional independence to formalize how solving certain pretext tasks can learn representations that provably decreases the sample complexity of downstream supervised tasks. Formally, we quantify how approximate independence between the components of the pretext task (conditional on the label and latent variables) allows us to learn representations that can solve the downstream task with drastically reduced sample complexity by just training a linear layer on top of the learned representation."
                },
                {
                    "title": "TERA: Self-Supervised Learning of Transformer Encoder Representation for Speech",
                    "abstract": "We introduce a self-supervised speech pre-training method called TERA, which stands for Transformer Encoder Representations from Alteration. Recent approaches often learn by using a single auxiliary task like contrastive prediction, autoregressive prediction, or masked reconstruction. Unlike previous methods, we use alteration along three orthogonal axes to pre-train Transformer Encoders on a large amount of unlabeled speech. The model learns through the reconstruction of acoustic frames from their altered counterpart, where we use a stochastic policy to alter along various dimensions: time, frequency, and magnitude. TERA can be used for speech representations extraction or fine-tuning with downstream models. We evaluate TERA on several downstream tasks, including phoneme classification, keyword spotting, speaker recognition, and speech recognition. We present a large-scale comparison of various self-supervised models. TERA achieves strong performance in the comparison by improving upon surface features and outperforming previous models. In our experiments, we study the effect of applying different alteration techniques, pre-training on more data, and pre-training on various features. We analyze different model sizes and find that smaller models are strong representation learners than larger models, while larger models are more effective for downstream fine-tuning than smaller models. Furthermore, we show the proposed method is transferable to downstream datasets not used in pre-training."
                },
                {
                    "title": "Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hypersphere",
                    "abstract": "Contrastive representation learning has been outstandingly successful in practice. In this work, we identify two key properties related to the contrastive loss: (1) alignment (closeness) of features from positive pairs, and (2) uniformity of the induced distribution of the (normalized) features on the hypersphere. We prove that, asymptotically, the contrastive loss optimizes these properties, and analyze their positive effects on downstream tasks. Empirically, we introduce an optimizable metric to quantify each property. Extensive experiments on standard vision and language datasets confirm the strong agreement between both metrics and downstream task performance. Remarkably, directly optimizing for these two metrics leads to representations with comparable or better performance at downstream tasks than contrastive learning. \nProject Page: this https URL \nCode: this https URL , this https URL"
                },
                {
                    "title": "An Investigation of Why Overparameterization Exacerbates Spurious Correlations",
                    "abstract": "We study why overparameterization -- increasing model size well beyond the point of zero training error -- can hurt test error on minority groups despite improving average test error when there are spurious correlations in the data. Through simulations and experiments on two image datasets, we identify two key properties of the training data that drive this behavior: the proportions of majority versus minority groups, and the signal-to-noise ratio of the spurious correlations. We then analyze a linear setting and theoretically show how the inductive bias of models towards \"memorizing\" fewer examples can cause overparameterization to hurt. Our analysis leads to a counterintuitive approach of subsampling the majority group, which empirically achieves low minority error in the overparameterized regime, even though the standard approach of upweighting the minority fails. Overall, our results suggest a tension between using overparameterized models versus using all the training data for achieving low worst-group error."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "Self-Supervised Learning of Pretext-Invariant Representations",
                    "abstract": "The goal of self-supervised learning from images is to construct image representations that are semantically meaningful via pretext tasks that do not require semantic annotations. Many pretext tasks lead to representations that are covariant with image transformations. We argue that, instead, semantic representations ought to be invariant under such transformations. Specifically, we develop Pretext-Invariant Representation Learning (PIRL, pronounced as `pearl') that learns invariant representations based on pretext tasks. We use PIRL with a commonly used pretext task that involves solving jigsaw puzzles. We find that PIRL substantially improves the semantic quality of the learned image representations. Our approach sets a new state-of-the-art in self-supervised learning from images on several popular benchmarks for self-supervised learning. Despite being unsupervised, PIRL outperforms supervised pre-training in learning image representations for object detection. Altogether, our results demonstrate the potential of self-supervised representations with good invariance properties."
                },
                {
                    "title": "Momentum Contrast for Unsupervised Visual Representation Learning",
                    "abstract": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a moving-averaged encoder. This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning. MoCo provides competitive results under the common linear protocol on ImageNet classification. More importantly, the representations learned by MoCo transfer well to downstream tasks. MoCo can outperform its supervised pre-training counterpart in 7 detection/segmentation tasks on PASCAL VOC, COCO, and other datasets, sometimes surpassing it by large margins. This suggests that the gap between unsupervised and supervised representation learning has been largely closed in many vision tasks."
                },
                {
                    "title": "A Theoretical Analysis of Contrastive Unsupervised Representation Learning",
                    "abstract": "Recent empirical works have successfully used unlabeled data to learn feature representations that are broadly useful in downstream classification tasks. Several of these methods are reminiscent of the well-known word2vec embedding algorithm: leveraging availability of pairs of semantically \"similar\" data points and \"negative samples,\" the learner forces the inner product of representations of similar pairs with each other to be higher on average than with negative samples. The current paper uses the term contrastive learning for such algorithms and presents a theoretical framework for analyzing them by introducing latent classes and hypothesizing that semantically similar points are sampled from the same latent class. This framework allows us to show provable guarantees on the performance of the learned representations on the average classification task that is comprised of a subset of the same set of latent classes. Our generalization bound also shows that learned representations can reduce (labeled) sample complexity on downstream tasks. We conduct controlled experiments in both the text and image domains to support the theory."
                },
                {
                    "title": "Self-Supervised Visual Feature Learning With Deep Neural Networks: A Survey",
                    "abstract": "Large-scale labeled data are generally required to train deep neural networks in order to obtain better performance in visual feature learning from images or videos for computer vision applications. To avoid extensive cost of collecting and annotating large-scale datasets, as a subset of unsupervised learning methods, self-supervised learning methods are proposed to learn general image and video features from large-scale unlabeled data without using any human-annotated labels. This paper provides an extensive review of deep learning-based self-supervised general visual feature learning methods from images or videos. First, the motivation, general pipeline, and terminologies of this field are described. Then the common deep neural network architectures that used for self-supervised learning are summarized. Next, the schema and evaluation metrics of self-supervised learning methods are reviewed followed by the commonly used datasets for images, videos, audios, and 3D data, as well as the existing self-supervised visual feature learning methods. Finally, quantitative performance comparisons of the reviewed methods on benchmark datasets are summarized and discussed for both image and video feature learning. At last, this paper is concluded and lists a set of promising future directions for self-supervised visual feature learning."
                },
                {
                    "title": "Fairness in representation: quantifying stereotyping as a representational harm",
                    "abstract": "While harms of allocation have been increasingly studied as part of the subfield of algorithmic fairness, harms of representation have received considerably less attention. In this paper, we formalize two notions of stereotyping and show how they manifest in later allocative harms within the machine learning pipeline. We also propose mitigation strategies and demonstrate their effectiveness on synthetic datasets."
                },
                {
                    "title": "Bias in Bios: A Case Study of Semantic Representation Bias in a High-Stakes Setting",
                    "abstract": "We present a large-scale study of gender bias in occupation classification, a task where the use of machine learning may lead to negative outcomes on peoples' lives. We analyze the potential allocation harms that can result from semantic representation bias. To do so, we study the impact on occupation classification of including explicit gender indicators---such as first names and pronouns---in different semantic representations of online biographies. Additionally, we quantify the bias that remains when these indicators are \"scrubbed,\" and describe proxy behavior that occurs in the absence of explicit gender indicators. As we demonstrate, differences in true positive rates between genders are correlated with existing gender imbalances in occupations, which may compound these imbalances."
                },
                {
                    "title": "What is the Effect of Importance Weighting in Deep Learning?",
                    "abstract": "Importance-weighted risk minimization is a key ingredient in many machine learning algorithms for causal inference, domain adaptation, class imbalance, and off-policy reinforcement learning. While the effect of importance weighting is well-characterized for low-capacity misspecified models, little is known about how it impacts over-parameterized, deep neural networks. This work is inspired by recent theoretical results showing that on (linearly) separable data, deep linear networks optimized by SGD learn weight-agnostic solutions, prompting us to ask, for realistic deep networks, for which many practical datasets are separable, what is the effect of importance weighting? We present the surprising finding that while importance weighting impacts models early in training, its effect diminishes over successive epochs. Moreover, while L2 regularization and batch normalization (but not dropout), restore some of the impact of importance weighting, they express the effect via (seemingly) the wrong abstraction: why should practitioners tweak the L2 regularization, and by how much, to produce the correct weighting effect? Our experiments confirm these findings across a range of architectures and datasets."
                },
                {
                    "title": "Fairness Without Demographics in Repeated Loss Minimization",
                    "abstract": "Machine learning models (e.g., speech recognizers) are usually trained to minimize average loss, which results in representation disparity---minority groups (e.g., non-native speakers) contribute less to the training objective and thus tend to suffer higher loss. Worse, as model accuracy affects user retention, a minority group can shrink over time. In this paper, we first show that the status quo of empirical risk minimization (ERM) amplifies representation disparity over time, which can even make initially fair models unfair. To mitigate this, we develop an approach based on distributionally robust optimization (DRO), which minimizes the worst case risk over all distributions close to the empirical distribution. We prove that this approach controls the risk of the minority group at each time step, in the spirit of Rawlsian distributive justice, while remaining oblivious to the identity of the groups. We demonstrate that DRO prevents disparity amplification on examples where ERM fails, and show improvements in minority group user satisfaction in a real-world text autocomplete task."
                },
                {
                    "title": "A Reductions Approach to Fair Classification",
                    "abstract": "We present a systematic approach for achieving fairness in a binary classification setting. While we focus on two well-known quantitative definitions of fairness, our approach encompasses many other previously studied definitions as special cases. The key idea is to reduce fair classification to a sequence of cost-sensitive classification problems, whose solutions yield a randomized classifier with the lowest (empirical) error subject to the desired constraints. We introduce two reductions that work for any representation of the cost-sensitive classifier and compare favorably to prior baselines on a variety of data sets, while overcoming several of their disadvantages."
                },
                {
                    "title": "Equality of Opportunity in Supervised Learning",
                    "abstract": "We propose a criterion for discrimination against a specified sensitive attribute in supervised learning, where the goal is to predict some target based on available features. Assuming data about the predictor, target, and membership in the protected group are available, we show how to optimally adjust any learned predictor so as to remove discrimination according to our definition. Our framework also improves incentives by shifting the cost of poor classification from disadvantaged groups to the decision maker, who can respond by improving the classification accuracy. We enourage readers to consult the more complete manuscript on the arXiv."
                },
                {
                    "title": "node2vec: Scalable Feature Learning for Networks",
                    "abstract": "Prediction tasks over nodes and edges in networks require careful effort in engineering features used by learning algorithms. Recent research in the broader field of representation learning has led to significant progress in automating prediction by learning the features themselves. However, present feature learning approaches are not expressive enough to capture the diversity of connectivity patterns observed in networks. Here we propose node2vec, an algorithmic framework for learning continuous feature representations for nodes in networks. In node2vec, we learn a mapping of nodes to a low-dimensional space of features that maximizes the likelihood of preserving network neighborhoods of nodes. We define a flexible notion of a node's network neighborhood and design a biased random walk procedure, which efficiently explores diverse neighborhoods. Our algorithm generalizes prior work which is based on rigid notions of network neighborhoods, and we argue that the added flexibility in exploring neighborhoods is the key to learning richer representations. We demonstrate the efficacy of node2vec over existing state-of-the-art techniques on multi-label classification and link prediction in several real-world networks from diverse domains. Taken together, our work represents a new way for efficiently learning state-of-the-art task-independent representations in complex networks."
                },
                {
                    "title": "Representation Learning: A Review and New Perspectives",
                    "abstract": "The success of machine learning algorithms generally depends on data representation, and we hypothesize that this is because different representations can entangle and hide more or less the different explanatory factors of variation behind the data. Although specific domain knowledge can be used to help design representations, learning with generic priors can also be used, and the quest for AI is motivating the design of more powerful representation-learning algorithms implementing such priors. This paper reviews recent work in the area of unsupervised feature learning and deep learning, covering advances in probabilistic models, autoencoders, manifold learning, and deep networks. This motivates longer term unanswered questions about the appropriate objectives for learning good representations, for computing representations (i.e., inference), and the geometrical connections between representation learning, density estimation, and manifold learning."
                },
                {
                    "title": "ADASYN: Adaptive synthetic sampling approach for imbalanced learning",
                    "abstract": "This paper presents a novel adaptive synthetic (ADASYN) sampling approach for learning from imbalanced data sets. The essential idea of ADASYN is to use a weighted distribution for different minority class examples according to their level of difficulty in learning, where more synthetic data is generated for minority class examples that are harder to learn compared to those minority examples that are easier to learn. As a result, the ADASYN approach improves learning with respect to the data distributions in two ways: (1) reducing the bias introduced by the class imbalance, and (2) adaptively shifting the classification decision boundary toward the difficult examples. Simulation analyses on several machine learning data sets show the effectiveness of this method across five evaluation metrics."
                },
                {
                    "title": "Direct and Indirect Effects",
                    "abstract": "The direct effect of one event on another can be defined and measured by holding constant all intermediate variables between the two. Indirect effects present conceptual and practical difficulties (in nonlinear models), because they cannot be isolated by holding certain variables constant. This paper presents a new way of defining the effect transmitted through a restricted set of paths, without controlling variables on the remaining paths. This permits the assessment of a more natural type of direct and indirect effects, one that is applicable in both linear and nonlinear models and that has broader policy-related interpretations. The paper establishes conditions under which such assessments can be estimated consistently from experimental and nonexperimental data, and thus extends path-analytic techniques to nonlinear and nonparametric models."
                },
                {
                    "title": "Layer-Peeled Model: Toward Understanding Well-Trained Deep Neural Networks",
                    "abstract": "In this paper, we introduce the Layer-Peeled Model , a nonconvex yet analytically tractable optimization program, in a quest to better understand deep neural networks that are trained for a su\ufb03ciently long time. As the name suggests, this new model is derived by isolating the topmost layer from the remainder of the neural network, followed by imposing certain constraints separately on the two parts of the network. We demonstrate that the Layer-Peeled Model, albeit simple, inherits many characteristics of well-trained neural networks, thereby o\ufb00ering an e\ufb00ective tool for explaining and predicting common empirical patterns of deep learning training. First, when working on class-balanced datasets, we prove that any solution to this model forms a simplex equiangular tight frame, which in part explains the recently discovered phenomenon of neural collapse in deep learning training [PHD20]. Moreover, when moving to the imbalanced case, our analysis of the Layer-Peeled Model reveals a hitherto unknown phenomenon that we term Minority Collapse , which fundamentally limits the performance of deep learning models on the minority classes. In addition, we use the Layer-Peeled Model to gain insights into how to mitigate Minority Collapse. Interestingly, this phenomenon is \ufb01rst predicted by the Layer-Peeled Model before its con\ufb01rmation by our computational experiments."
                },
                {
                    "title": "CERT: Contrastive Self-supervised Learning for Language Understanding",
                    "abstract": "Pretrained language models such as BERT, GPT have shown great effectiveness in language understanding. The auxiliary predictive tasks in existing pretraining approaches are mostly defined on tokens, thus may not be able to capture sentence-level semantics very well. To address this issue, we propose CERT: Contrastive self-supervised Encoder Representations from Transformers, which pretrains language representation models using contrastive self-supervised learning at the sentence level. CERT creates augmentations of original sentences using back-translation. Then it finetunes a pretrained language encoder (e.g., BERT) by predicting whether two augmented sentences originate from the same sentence. CERT is simple to use and can be flexibly plugged into any pretraining-finetuning NLP pipeline. We evaluate CERT on 11 natural language understanding tasks in the GLUE benchmark where CERT outperforms BERT on 7 tasks, achieves the same performance as BERT on 2 tasks, and performs worse than BERT on 2 tasks. On the averaged score of the 11 tasks, CERT outperforms BERT. The data and code are available at https://github.com/UCSD-AI4H/CERT"
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Visualizing Data using t-SNE",
                    "abstract": "We present a new technique called \u201ct-SNE\u201d that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to influence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are significantly better than those produced by the other techniques on almost all of the datasets."
                },
                {
                    "title": "The Model of",
                    "abstract": "Stock Price Prediction Deep Neural Generative SUMMARY In this study, we propose a deep neural generative model for predicting daily stock price movements given news articles. Approaches involving conventional technical analysis have been investigated to identify certain patterns in past price movements, which in turn helps to predict future price movements. However, the \ufb01nancial market is highly sensitive to speci\ufb01c events, including corporate buyouts, product releases, and the like. Therefore, recent research has focused on modeling relation- ships between these events that appear in the news articles and future price movements; however, a very large number of news articles are published daily, each article containing rich information, which results in over\ufb01tting to past price movements used for parameter adjustment. Given the above, we propose a model based on a generative model of news articles that includes price movement as a condition, thereby avoiding excessive over\ufb01t- ting thanks to the nature of the generative model. We evaluate our proposed model using historical price movements of Nikkei 225 and Standard & Poor\u2019s 500 Stock Index, con\ufb01rming that our model predicts future price movements better than such conventional classi\ufb01ers as support vector ma- chines and multilayer perceptrons. Further, our proposed model extracts signi\ufb01cant words from news articles that are directly related to future stock price movements."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does group underrepresentation in contrastive learning algorithms lead to representation harms that affect downstream model performance?\n\n**[Question 2] - Why is it interesting and important?**  \nAddressing the problem of representation harms in contrastive learning is crucial for the research community as it highlights a significant gap in understanding algorithmic biases beyond allocative harms. Solving this issue could lead to the development of more equitable machine learning models that better represent minority groups, ultimately advancing knowledge in representation learning and self-supervised learning. This research could also inform practical applications in various domains, ensuring that automated decision-making systems are fairer and more accurate, thereby reducing the risk of perpetuating biases in real-world scenarios.\n\n**[Question 3] - Why is it hard?**  \nThe challenge in solving this problem lies in the difficulty of measuring representation harms, which are often diffuse and long-term compared to allocative harms. Naive approaches may fail because they do not account for the complex interactions between underrepresented groups and their representations in high-dimensional spaces. Additionally, the intrinsic nature of contrastive learning algorithms may lead to the collapse of representations for underrepresented groups, making it difficult to isolate and address the specific causes of these harms. Overcoming these technical and theoretical obstacles requires a nuanced understanding of representation learning dynamics.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on allocative harms, leaving representation harms underexplored due to their elusive nature and the lack of effective measurement techniques. Existing solutions have not adequately addressed the specific mechanisms by which underrepresentation affects representation quality in contrastive learning. This paper differs from prior work by explicitly investigating the relationship between group underrepresentation and representation harms in contrastive learning, providing empirical evidence and a theoretical model to elucidate these effects.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves conducting a controlled study using the CIFAR10 dataset to empirically demonstrate the effects of underrepresentation on contrastive learning representations. The study will utilize t-SNE visualizations to analyze the clustering of representations and identify instances of stereotyping. The expected outcomes include a clear demonstration of how underrepresented groups' representations collapse to those of semantically similar groups, along with the development of a simple model of contrastive learning on graphs that illustrates this phenomenon. The results aim to provide insights into the intrinsic nature of representation harms in contrastive learning algorithms."
            }
        },
        "author_data": {
            "d4c19c30-235a-488f-a3eb-eb922892a322": {
                "pk": "d4c19c30-235a-488f-a3eb-eb922892a322",
                "name": "Subha Maity",
                "collaborators": [
                    "Yuekai Sun",
                    "M. Yurochkin",
                    "M. Banerjee",
                    "Debarghya Mukherjee",
                    "Lilian Ngweta",
                    "Alex Gittens",
                    "Moulinath Banerjee",
                    "Songkai Xue",
                    "Mayank Agarwal",
                    "Daniele Bracale",
                    "Felipe Maia Polo",
                    "Soham Bakshi",
                    "Saptarshi Roy",
                    "R. Bhattacharyya",
                    "Anik Burman",
                    "Kalpana Singh",
                    "Sayantan Banerjee",
                    "Arnab Auddy",
                    "S. Rout",
                    "Supriya Lahoti",
                    "Rajmohan Panda",
                    "V. Baladandayuthapani",
                    "Bum Chul Kwon",
                    "U. Kartoun",
                    "S. Khurshid",
                    "Deanna G. Brockman",
                    "A. Khera",
                    "P. Ellinor",
                    "S. Lubitz",
                    "Kenney Ng",
                    "Diptavo Dutta",
                    "Jonathan Terhorst",
                    "Sudip Kumar Ghorui",
                    "Roshan Ghosh"
                ],
                "domain": [
                    "Machine Learning",
                    "Algorithmic Fairness",
                    "Statistical Analysis",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Aligners: Decoupling LLMs and Alignment",
                        "abstract": "Large Language Models (LLMs) need to be aligned with human expectations to ensure their safety and utility in most applications. Alignment is challenging, costly, and needs to be repeated for every LLM and alignment criterion. We propose to decouple LLMs and alignment by training aligner models that can be used to align any LLM for a given criteria on an as-needed basis, thus also reducing the potential negative impacts of alignment on performance. Our recipe for training the aligner models solely relies on synthetic data generated with a (prompted) LLM and can be easily adjusted for a variety of alignment criteria. We use the same synthetic data to train inspectors, binary miss-alignment classification models to guide a\"squad\"of multiple aligners. Our empirical results demonstrate consistent improvements when applying aligner squad to various LLMs, including chat-aligned models, across several instruction-following and red-teaming datasets."
                    },
                    {
                        "title": "Learning the Distribution Map in Reverse Causal Performative Prediction",
                        "abstract": "In numerous predictive scenarios, the predictive model affects the sampling distribution; for example, job applicants often meticulously craft their resumes to navigate through a screening systems. Such shifts in distribution are particularly prevalent in the realm of social computing, yet, the strategies to learn these shifts from data remain remarkably limited. Inspired by a microeconomic model that adeptly characterizes agents' behavior within labor markets, we introduce a novel approach to learn the distribution shift. Our method is predicated on a reverse causal model, wherein the predictive model instigates a distribution shift exclusively through a finite set of agents' actions. Within this framework, we employ a microfoundation model for the agents' actions and develop a statistically justified methodology to learn the distribution shift map, which we demonstrate to be effective in minimizing the performative prediction risk."
                    },
                    {
                        "title": "Weak Supervision Performance Evaluation via Partial Identification",
                        "abstract": "Programmatic Weak Supervision (PWS) enables supervised model training without direct access to ground truth labels, utilizing weak labels from heuristics, crowdsourcing, or pre-trained models. However, the absence of ground truth complicates model evaluation, as traditional metrics such as accuracy, precision, and recall cannot be directly calculated. In this work, we present a novel method to address this challenge by framing model evaluation as a partial identification problem and estimating performance bounds using Fr\\'echet bounds. Our approach derives reliable bounds on key metrics without requiring labeled data, overcoming core limitations in current weak supervision evaluation techniques. Through scalable convex optimization, we obtain accurate and computationally efficient bounds for metrics including accuracy, precision, recall, and F1-score, even in high-dimensional settings. This framework offers a robust approach to assessing model quality without ground truth labels, enhancing the practicality of weakly supervised learning for real-world applications."
                    },
                    {
                        "title": "Simple Disentanglement of Style and Content in Visual Representations",
                        "abstract": "Learning visual representations with interpretable features, i.e., disentangled representations, remains a challenging problem. Existing methods demonstrate some success but are hard to apply to large-scale vision datasets like ImageNet. In this work, we propose a simple post-processing framework to disentangle content and style in learned representations from pre-trained vision models. We model the pre-trained features probabilistically as linearly entangled combinations of the latent content and style factors and develop a simple disentanglement algorithm based on the probabilistic model. We show that the method provably disentangles content and style features and verify its efficacy empirically. Our post-processed features yield significant domain generalization performance improvements when the distribution shift occurs due to style changes or style-related spurious correlations."
                    },
                    {
                        "title": "Bayes classifier cannot be learned from noisy responses with unknown noise rates",
                        "abstract": "Training a classifier with noisy labels typically requires the learner to specify the distribution of label noise, which is often unknown in practice. Although there have been some recent attempts to relax that requirement, we show that the Bayes decision rule is unidentified in most classification problems with noisy labels. This suggests it is generally not possible to bypass/relax the requirement. In the special cases in which the Bayes decision rule is identified, we develop a simple algorithm to learn the Bayes decision rule, that does not require knowledge of the noise distribution."
                    },
                    {
                        "title": "Understanding new tasks through the lens of training data via exponential tilting",
                        "abstract": "Deploying machine learning models to new tasks is a major challenge despite the large size of the modern training datasets. However, it is conceivable that the training data can be reweighted to be more representative of the new (target) task. We consider the problem of reweighing the training samples to gain insights into the distribution of the target task. Specifically, we formulate a distribution shift model based on the exponential tilt assumption and learn train data importance weights minimizing the KL divergence between labeled train and unlabeled target datasets. The learned train data weights can then be used for downstream tasks such as target performance evaluation, fine-tuning, and model selection. We demonstrate the efficacy of our method on Waterbirds and Breeds benchmarks."
                    },
                    {
                        "title": "Predictor-corrector algorithms for stochastic optimization under gradual distribution shift",
                        "abstract": "Time-varying stochastic optimization problems frequently arise in machine learning practice (e.g. gradual domain shift, object tracking, strategic classification). Although most problems are solved in discrete time, the underlying process is often continuous in nature. We exploit this underlying continuity by developing predictor-corrector algorithms for time-varying stochastic optimizations. We provide error bounds for the iterates, both in presence of pure and noisy access to the queries from the relevant derivatives of the loss function. Furthermore, we show (theoretically and empirically in several examples) that our method outperforms non-predictor corrector methods that do not exploit the underlying continuous process."
                    },
                    {
                        "title": "How does overparametrization affect performance on minority groups?",
                        "abstract": "The benefits of overparameterization for the overall performance of modern machine learning (ML) models are well known. However, the effect of overparameterization at a more granular level of data subgroups is less understood. Recent empirical studies demonstrate encouraging results: (i) when groups are not known, overparameterized models trained with empirical risk minimization (ERM) perform better on minority groups; (ii) when groups are known, ERM on data subsampled to equalize group sizes yields state-of-the-art worst-group-accuracy in the overparameterized regime. In this paper, we complement these empirical studies with a theoretical investigation of the risk of overparameterized random feature models on minority groups. In a setting in which the regression functions for the majority and minority groups are different, we show that overparameterization always improves minority group performance."
                    },
                    {
                        "title": "Role of multiresolution vulnerability indices in COVID-19 spread in India: a Bayesian model-based analysis",
                        "abstract": "Objectives COVID-19 has differentially affected countries, with health infrastructure and other related vulnerability indicators playing a role in determining the extent of its spread. Vulnerability of a geographical region to COVID-19 has been a topic of interest, particularly in low-income and middle-income countries like India to assess its multifactorial impact on incidence, prevalence or mortality. This study aims to construct a statistical analysis pipeline to compute such vulnerability indices and investigate their association with metrics of the pandemic growth. Design Using publicly reported observational socioeconomic, demographic, health-based and epidemiological data from Indian national surveys, we compute contextual COVID-19 Vulnerability Indices (cVIs) across multiple thematic resolutions for different geographical and spatial administrative regions. These cVIs are then used in Bayesian regression models to assess their impact on indicators of the spread of COVID-19. Setting This study uses district-level indicators and case counts data for the state of Odisha, India. Primary outcome measure We use instantaneous R (temporal average of estimated time-varying reproduction number for COVID-19) as the primary outcome variable in our models. Results Our observational study, focussing on 30 districts of Odisha, identified housing and hygiene conditions, COVID-19 preparedness and epidemiological factors as important indicators associated with COVID-19 vulnerability. Conclusion Having succeeded in containing COVID-19 to a reasonable level during the first wave, the second wave of COVID-19 made greater inroads into the hinterlands and peripheral districts of Odisha, burdening the already deficient public health system in these areas, as identified by the cVIs. Improved understanding of the factors driving COVID-19 vulnerability will help policy makers prioritise resources and regions, leading to more effective mitigation strategies for the present and future."
                    },
                    {
                        "title": "RMExplorer: A Visual Analytics Approach to Explore the Performance and the Fairness of Disease Risk Models on Population Subgroups",
                        "abstract": "Disease risk models can identify high-risk patients and help clinicians provide more personalized care. However, risk models de-veloped on one dataset may not generalize across diverse subpop-ulations of patients in different datasets and may have unexpected performance. It is challenging for clinical researchers to inspect risk models across different subgroups without any tools. Therefore, we developed an interactive visualization system called RMExplorer (Risk Model Explorer) to enable interactive risk model assessment. Specifically, the system allows users to define subgroups of patients by selecting clinical, demographic, or other characteristics, to ex-plore the performance and fairness of risk models on the subgroups, and to understand the feature contributions to risk scores. To demonstrate the usefulness of the tool, we conduct a case study, where we use RMExplorer to explore three atrial fibrillation risk models by applying them to the UK Biobank dataset of 445,329 individuals. RMExplorer can help researchers to evaluate the performance and biases of risk models on subpopulations of interest in their data."
                    },
                    {
                        "title": "A linear adjustment-based approach to posterior drift in transfer learning.",
                        "abstract": "We present new models and methods for the posterior drift problem where the regression function in the target domain is modelled as a linear adjustment, on an appropriate scale, of that in the source domain, and study the theoretical properties of our proposed estimators in the binary classification problem. The core idea of our model inherits the simplicity and the usefulness of generalized linear models and accelerated failure time models from the classical statistics literature. Our approach is shown to be flexible and applicable in a variety of statistical settings, and can be adopted for transfer learning problems in various domains including epidemiology, genetics and biomedicine. As concrete applications, we illustrate the power of our approach (i) through mortality prediction for British Asians by borrowing strength from similar data from the larger pool of British Caucasians, using the UK Biobank data, and (ii) in overcoming a spurious correlation present in the source domain of the Waterbirds dataset."
                    },
                    {
                        "title": "Statistical inference for individual fairness",
                        "abstract": "As we rely on machine learning (ML) models to make more consequential decisions, the issue of ML models perpetuating or even exacerbating undesirable historical biases (e.g., gender and racial biases) has come to the fore of the public's attention. In this paper, we focus on the problem of detecting violations of individual fairness in ML models. We formalize the problem as measuring the susceptibility of ML models against a form of adversarial attack and develop a suite of inference tools for the adversarial cost function. The tools allow auditors to assess the individual fairness of ML models in a statistically-principled way: form confidence intervals for the worst-case performance differential between similar individuals and test hypotheses of model fairness with (asymptotic) non-coverage/Type I error rate control. We demonstrate the utility of our tools in a real-world case study."
                    },
                    {
                        "title": "There is no trade-off: enforcing fairness can improve accuracy",
                        "abstract": "One of the main barriers to the broader adoption of algorithmic fairness in machine learning is the trade-off between fairness and performance of ML models: many practitioners are unwilling to sacrifice the performance of their ML model for fairness. In this paper, we show that this trade-off may not be necessary. If the algorithmic biases in an ML model are due to sampling biases in the training data, then enforcing algorithmic fairness may improve the performance of the ML model on unbiased test data. We study conditions under which enforcing algorithmic fairness helps practitioners learn the Bayes decision rule for (unbiased) test data from biased training data. We also demonstrate the practical implications of our theoretical results in real-world ML tasks."
                    },
                    {
                        "title": "Minimax optimal approaches to the label shift problem in non-parametric settings",
                        "abstract": "We study minimax rates of convergence in the label shift problem. In addition to the usual setting in which the learner only has access to unlabeled examples from the target domain, we also consider the setting in which a small number of labeled examples from the target domain are available to the learner. Our study reveals a difference in the difficulty of the label shift problem in the two settings. We attribute this difference to the availability of data from the target domain to estimate the class conditional distributions in the latter setting. We also show that a distributional matching approach is minimax rate-optimal in the former setting."
                    },
                    {
                        "title": "Does enforcing fairness mitigate biases caused by subpopulation shift?",
                        "abstract": "Many instances of algorithmic bias are caused by subpopulation shifts. For example, ML models often perform worse on demographic groups that are underrepresented in the training data. In this paper, we study whether enforcing algorithmic fairness during training improves the performance of the trained model in the \\emph{target domain}. On one hand, we conceive scenarios in which enforcing fairness does not improve performance in the target domain. In fact, it may even harm performance. On the other hand, we derive necessary and sufficient conditions under which enforcing algorithmic fairness leads to the Bayes model in the target domain. We also illustrate the practical implications of our theoretical results in simulations and on real data."
                    },
                    {
                        "title": "Communication-Efficient Integrative Regression in High-Dimensions",
                        "abstract": "We consider the task of meta-analysis in high-dimensional settings in which the data sources we wish to integrate are similar but non-identical. To borrow strength across such heterogeneous data sources, we introduce a global parameter that addresses several identification issues. We also propose a one-shot estimator of the global parameter that preserves the anonymity of the data sources and converges at a rate that depends on the size of the combined dataset. Finally, we demonstrate the benefits of our approach on a large-scale drug treatment dataset involving several different cancer cell lines."
                    },
                    {
                        "title": "Meta-analysis of heterogeneous data: integrative sparse regression in high-dimensions",
                        "abstract": "We consider the task of meta-analysis in high-dimensional settings in which the data sources are similar but non-identical. To borrow strength across such heterogeneous datasets, we introduce a global parameter that emphasizes interpretability and statistical efficiency in the presence of heterogeneity. We also propose a one-shot estimator of the global parameter that preserves the anonymity of the data sources and converges at a rate that depends on the size of the combined dataset. For high-dimensional linear model settings, we demonstrate the superiority of our identification restrictions in adapting to a previously seen data distribution as well as predicting for a new/unseen data distribution. Finally, we demonstrate the benefits of our approach on a large-scale drug treatment dataset involving several different cancer cell-lines."
                    },
                    {
                        "title": "Economic Load Dispatch of Power System Using Grey Wolf Optimization with Constriction Factor",
                        "abstract": "In this paper, we propose a new meta-heuristic, nature inspired technique known as Grey Wolf Optimization to solve Economic Load Dispatch (ELD) problems of thermal power units considering transmission losses, and constraints such as ramp rate limits and prohibited operating zones. Grey Wolf Optimization Algorithm (GWOA) is a relatively new optimization technique. Mathematical models of this algorithm demonstrate the efficiency, quality of solution and convergence speed of the method and successful application of the algorithm on economic load dispatch problems. Simulation results found that the proposed approach outperforms several other existing optimization techniques in terms quality of solution obtained and computational efficiency. Results also be confirmed the robustness of the proposed methodology."
                    }
                ]
            },
            "61dd1d79-ac1a-4a44-80df-9fd07ff695ba": {
                "pk": "61dd1d79-ac1a-4a44-80df-9fd07ff695ba",
                "name": "Mayank Agarwal",
                "collaborators": [
                    "Justin D. Weisz",
                    "Michael J. Muller",
                    "Stephanie Houde",
                    "Kartik Talamadupula",
                    "M. Yurochkin",
                    "Steven I. Ross",
                    "Fernando Martinez",
                    "John T. Richards",
                    "Lilian Ngweta",
                    "Subha Maity",
                    "Alex Gittens",
                    "Yuekai Sun",
                    "Djallel Bouneffouf",
                    "I. Rish",
                    "A. Kotecha",
                    "Atharva Deolalikar",
                    "Ritika Kalia",
                    "R. Yadav",
                    "Achamma Thomas",
                    "Naveen Venkat",
                    "Maneesh Singh",
                    "Shubham Tulsiani",
                    "Zahra Ashktorab",
                    "Benjamin Hoover",
                    "Casey Dugan",
                    "Werner Geyer",
                    "Han Yang",
                    "Jiao Sun",
                    "Q. Liao",
                    "D. Shakya",
                    "Vishal Deshpande"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Generative AI",
                    "Machine Learning",
                    "Explainable AI"
                ],
                "publications": [
                    {
                        "title": "Aligners: Decoupling LLMs and Alignment",
                        "abstract": "Large Language Models (LLMs) need to be aligned with human expectations to ensure their safety and utility in most applications. Alignment is challenging, costly, and needs to be repeated for every LLM and alignment criterion. We propose to decouple LLMs and alignment by training aligner models that can be used to align any LLM for a given criteria on an as-needed basis, thus also reducing the potential negative impacts of alignment on performance. Our recipe for training the aligner models solely relies on synthetic data generated with a (prompted) LLM and can be easily adjusted for a variety of alignment criteria. We use the same synthetic data to train inspectors, binary miss-alignment classification models to guide a\"squad\"of multiple aligners. Our empirical results demonstrate consistent improvements when applying aligner squad to various LLMs, including chat-aligned models, across several instruction-following and red-teaming datasets."
                    },
                    {
                        "title": "Dialogue System with Missing Observation",
                        "abstract": "Within the domain of dialogue, the ability to orchestrate multiple independently trained dialogue agents to create a unified system is of particular importance. Where we define orchestration as the task of selecting a subset of skills which most appropriately answer a user input using features extracted from both the user input and the individual skills. In this work, we study the task of online dialogue orchestration where the user feedback associated with the dialogue agent may not always be observed. In order to address the missing feedback setting, we propose to combine the attentive contextual bandit approach with an unsupervised learning mechanism such as clustering. By leveraging clustering to estimate missing reward, we are able to learn from each incoming event, even those with missing rewards. Promising empirical results are obtained on proprietary conversational datasets."
                    },
                    {
                        "title": "Deep Learning Approaches for Plant Disease Detection: A Comparative Review",
                        "abstract": "Crop production and growth are crucial factors that have an impact on the agricultural industry, as well as the farmer economically, socially, and in every other manner. Crop failure has several key causes, one of which is plant diseases. Plant diseases can be identified by a variety of signs, including infections, color changes, damaged leaves or stems, abnormal floret, stem, leaflet, bud, or root growth, among others. In addition, leaves exhibit disease indicators such as stains, dryness, premature dropping, etc. The research on using different Deep CNN architectures to detect illnesses of plants from photographs of leaves is reviewed in this paper. It also offers a comparison of models of CNN that were trained on leaf images to recognize and categorize the diseases of plants. Along with the benefits and drawbacks of the various approaches that were described in the prior research, it gives an overview of the performance measures and database used to assess the effectiveness of existing models."
                    },
                    {
                        "title": "Geometry-biased Transformers for Novel View Synthesis",
                        "abstract": "We tackle the task of synthesizing novel views of an object given a few input images and associated camera viewpoints. Our work is inspired by recent 'geometry-free' approaches where multi-view images are encoded as a (global) set-latent representation, which is then used to predict the color for arbitrary query rays. While this representation yields (coarsely) accurate images corresponding to novel viewpoints, the lack of geometric reasoning limits the quality of these outputs. To overcome this limitation, we propose 'Geometry-biased Transformers' (GBTs) that incorporate geometric inductive biases in the set-latent representation-based inference to encourage multi-view geometric consistency. We induce the geometric bias by augmenting the dot-product attention mechanism to also incorporate 3D distances between rays associated with tokens as a learnable bias. We find that this, along with camera-aware embeddings as input, allows our models to generate significantly more accurate outputs. We validate our approach on the real-world CO3D dataset, where we train our system over 10 categories and evaluate its view-synthesis ability for novel objects as well as unseen categories. We empirically validate the benefits of the proposed geometric biases and show that our approach significantly improves over prior works."
                    },
                    {
                        "title": "Fairness Evaluation in Text Classification: Machine Learning Practitioner Perspectives of Individual and Group Fairness",
                        "abstract": "Mitigating algorithmic bias is a critical task in the development and deployment of machine learning models. While several toolkits exist to aid machine learning practitioners in addressing fairness issues, little is known about the strategies practitioners employ to evaluate model fairness and what factors influence their assessment, particularly in the context of text classification. Two common approaches of evaluating the fairness of a model are group fairness and individual fairness. We run a study with Machine Learning practitioners (n=24) to understand the strategies used to evaluate models. Metrics presented to practitioners (group vs. individual fairness) impact which models they consider fair. Participants focused on risks associated with underpredicting / overpredicting and model sensitivity relative to identity token manipulations. We discover fairness assessment strategies involving personal experiences or how users form groups of identity tokens to test model fairness. We provide recommendations for interactive tools for evaluating fairness in text classification."
                    },
                    {
                        "title": "Better Together? An Evaluation of AI-Supported Code Translation",
                        "abstract": "Generative machine learning models have recently been applied to source code, for use cases including translating code between programming languages, creating documentation from code, and auto-completing methods. Yet, state-of-the-art models often produce code that is erroneous or incomplete. In a controlled study with 32 software engineers, we examined whether such imperfect outputs are helpful in the context of Java-to-Python code translation. When aided by the outputs of a code translation model, participants produced code with fewer errors than when working alone. We also examined how the quality and quantity of AI translations affected the work process and quality of outcomes, and observed that providing multiple translations had a larger impact on the translation process than varying the quality of provided translations. Our results tell a complex, nuanced story about the benefits of generative code models and the challenges software engineers face when working with their outputs. Our work motivates the need for intelligent user interfaces that help software engineers effectively work with generative code models in order to understand and evaluate their outputs and achieve superior outcomes to working alone."
                    },
                    {
                        "title": "Investigating Explainability of Generative AI for Code through Scenario-based Design",
                        "abstract": "What does it mean for a generative AI model to be explainable? The emergent discipline of explainable AI (XAI) has made great strides in helping people understand discriminative models. Less attention has been paid to generative models that produce artifacts, rather than decisions, as output. Meanwhile, generative AI (GenAI) technologies are maturing and being applied to application domains such as software engineering. Using scenario-based design and question-driven XAI design approaches, we explore users\u2019 explainability needs for GenAI in three software engineering use cases: natural language to code, code translation, and code auto-completion. We conducted 9 workshops with 43 software engineers in which real examples from state-of-the-art generative AI models were used to elicit users\u2019 explainability needs. Drawing from prior work, we also propose 4 types of XAI features for GenAI for code and gathered additional design ideas from participants. Our work explores explainability needs for GenAI for code and demonstrates how human-centered approaches can drive the technical development of XAI in novel domains."
                    },
                    {
                        "title": "A Deep Learning Based Model to Predict PM\u2081\u2080 Concentration",
                        "abstract": "Particulate matter PM10 is the most effective air pollutant and has a serious impact on health and the environment. Hence, it is important to predict the level of PM10 concentration for the protection of the population. This study employs deep learning models such as MLP, GRU, and LSTM for the prediction of PM10 concentration at different time intervals. The main input variables for the proposed models are 15mins, 30mins, 1hr, 4hr, 8hr, 24hr measurements of PM10 concentration. The time series data is collected from the Central Pollution Control Board (CPCB), New Delhi, for five years, from 2016 to 2021. We use different performance metrics, such as the coefficient of determination (R2), normalized root mean square error (nRMSE), mean absolute percentage error (MAPE), to compare how well different deep learning techniques work. The LSTM model achieved higher predictive accuracy than the GRU and MLP models in estimating PM10 concentrations at different time intervals with R2 value ranging from 0.575 to 0.963, as shown by the experiments. Other models only achieved predictive accuracy of 0.575 to 0.963. In this study, a promising and cost-effective method for accurately predicting PM10 concentrations was developed and proposed."
                    },
                    {
                        "title": "Drinking Chai with Your (AI) Programming Partner: A Design Fiction about Generative AI for Software Engineering 107-122",
                        "abstract": "Using design fiction, we develop a series of possible generative AI features and applications that could be developed in the future of humans\u2019 roles in software engineering. We use the fiction to highlight choices and value-tensions among these potential futures."
                    }
                ]
            },
            "44c6a11a-57cc-4b7b-abc5-c9b9caa3fdbc": {
                "pk": "44c6a11a-57cc-4b7b-abc5-c9b9caa3fdbc",
                "name": "Mikhail Yurochkin",
                "collaborators": [
                    "Yuekai Sun",
                    "Felipe Maia Polo",
                    "K. Greenewald",
                    "Leshem Choshen",
                    "Justin Solomon",
                    "Rickard Br\u00fcel Gabrielsson",
                    "Subha Maity",
                    "Jiacheng Zhu",
                    "Leonid Karlinsky",
                    "Lucas Weber",
                    "Lilian Ngweta",
                    "Mayank Agarwal",
                    "Alex Gittens",
                    "Moulinath Banerjee",
                    "Ronald Xu",
                    "M'irian Silva",
                    "Onkar Bhardwaj",
                    "Igor Melnyk",
                    "Youssef Mroueh",
                    "Brian M. Belgodere",
                    "Mattia Rigotti",
                    "Apoorva Nitsure",
                    "Jir\u00ed Navr\u00e1til",
                    "Jerret Ross",
                    "Tal Shnitzer",
                    "Kimia Nadjahi",
                    "Haitz S'aez de Oc'ariz Borde",
                    "Marzyeh Ghassemi",
                    "Nimrod Shabtay",
                    "Sivan Doveh",
                    "Wei Lin",
                    "M. Mirza",
                    "Leshem Chosen",
                    "Assaf Arbelle",
                    "Raja Giryes",
                    "Gongjun Xu",
                    "Felix Petersen",
                    "Aashwin Mishra",
                    "Hilde Kuehne",
                    "Christian Borgelt",
                    "Oliver Deussen",
                    "Seamus Somerstep",
                    "Ya'acov Ritov",
                    "Allysson Flavio Melo de Oliveira",
                    "Michael Feffer",
                    "Anthony Ou",
                    "Kate Soule",
                    "Neil Thompson",
                    "Songkai Xue",
                    "Dustin Klebe",
                    "Hongyi Wang",
                    "Souvik Kundu",
                    "Eric P. Xing",
                    "Zahra Ashktorab",
                    "Benjamin Hoover",
                    "Casey Dugan",
                    "Werner Geyer",
                    "Han Yang",
                    "J. Solomon"
                ],
                "domain": [
                    "Machine Learning",
                    "Large Language Models",
                    "Fine-Tuning",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Asymmetry in Low-Rank Adapters of Foundation Models",
                        "abstract": "Parameter-efficient fine-tuning optimizes large, pre-trained foundation models by updating a subset of parameters; in this class, Low-Rank Adaptation (LoRA) is particularly effective. Inspired by an effort to investigate the different roles of LoRA matrices during fine-tuning, this paper characterizes and leverages unexpected asymmetry in the importance of low-rank adapter matrices. Specifically, when updating the parameter matrices of a neural network by adding a product $BA$, we observe that the $B$ and $A$ matrices have distinct functions: $A$ extracts features from the input, while $B$ uses these features to create the desired output. Based on this observation, we demonstrate that fine-tuning $B$ is inherently more effective than fine-tuning $A$, and that a random untrained $A$ should perform nearly as well as a fine-tuned one. Using an information-theoretic lens, we also bound the generalization of low-rank adapters, showing that the parameter savings of exclusively training $B$ improves the bound. We support our conclusions with experiments on RoBERTa, BART-Large, LLaMA-2, and ViTs."
                    },
                    {
                        "title": "LiveXiv -- A Multi-Modal Live Benchmark Based on Arxiv Papers Content",
                        "abstract": "The large-scale training of multi-modal models on data scraped from the web has shown outstanding utility in infusing these models with the required world knowledge to perform effectively on multiple downstream tasks. However, one downside of scraping data from the web can be the potential sacrifice of the benchmarks on which the abilities of these models are often evaluated. To safeguard against test data contamination and to truly test the abilities of these foundation models we propose LiveXiv: A scalable evolving live benchmark based on scientific ArXiv papers. LiveXiv accesses domain-specific manuscripts at any given timestamp and proposes to automatically generate visual question-answer pairs (VQA). This is done without any human-in-the-loop, using the multi-modal content in the manuscripts, like graphs, charts, and tables. Moreover, we introduce an efficient evaluation approach that estimates the performance of all models on the evolving benchmark using evaluations of only a subset of models. This significantly reduces the overall evaluation cost. We benchmark multiple open and proprietary Large Multi-modal Models (LMMs) on the first version of our benchmark, showing its challenging nature and exposing the models true abilities, avoiding contamination. Lastly, in our commitment to high quality, we have collected and evaluated a manually verified subset. By comparing its overall results to our automatic annotations, we have found that the performance variance is indeed minimal (<2.5%). Our dataset is available online on HuggingFace, and our code will be available here."
                    },
                    {
                        "title": "tinyBenchmarks: evaluating LLMs with fewer examples",
                        "abstract": "The versatility of large language models (LLMs) led to the creation of diverse benchmarks that thoroughly test a variety of language models' abilities. These benchmarks consist of tens of thousands of examples making evaluation of LLMs very expensive. In this paper, we investigate strategies to reduce the number of evaluations needed to assess the performance of an LLM on several key benchmarks. For example, we show that to accurately estimate the performance of an LLM on MMLU, a popular multiple-choice QA benchmark consisting of 14K examples, it is sufficient to evaluate this LLM on 100 curated examples. We release evaluation tools and tiny versions of popular benchmarks: Open LLM Leaderboard, MMLU, HELM, and AlpacaEval 2.0. Our empirical analysis demonstrates that these tools and tiny benchmarks are sufficient to reliably and efficiently reproduce the original evaluation results."
                    },
                    {
                        "title": "Aligners: Decoupling LLMs and Alignment",
                        "abstract": "Large Language Models (LLMs) need to be aligned with human expectations to ensure their safety and utility in most applications. Alignment is challenging, costly, and needs to be repeated for every LLM and alignment criterion. We propose to decouple LLMs and alignment by training aligner models that can be used to align any LLM for a given criteria on an as-needed basis, thus also reducing the potential negative impacts of alignment on performance. Our recipe for training the aligner models solely relies on synthetic data generated with a (prompted) LLM and can be easily adjusted for a variety of alignment criteria. We use the same synthetic data to train inspectors, binary miss-alignment classification models to guide a\"squad\"of multiple aligners. Our empirical results demonstrate consistent improvements when applying aligner squad to various LLMs, including chat-aligned models, across several instruction-following and red-teaming datasets."
                    },
                    {
                        "title": "Uncertainty Quantification via Stable Distribution Propagation",
                        "abstract": "We propose a new approach for propagating stable probability distributions through neural networks. Our method is based on local linearization, which we show to be an optimal approximation in terms of total variation distance for the ReLU non-linearity. This allows propagating Gaussian and Cauchy input uncertainties through neural networks to quantify their output uncertainties. To demonstrate the utility of propagating distributions, we apply the proposed method to predicting calibrated confidence intervals and selective prediction on out-of-distribution data. The results demonstrate a broad applicability of propagating distributions and show the advantages of our method over other approaches such as moment matching."
                    },
                    {
                        "title": "A statistical framework for weak-to-strong generalization",
                        "abstract": "Modern large language model (LLM) alignment techniques rely on human feedback, but it is unclear whether the techniques fundamentally limit the capabilities of aligned LLMs. In particular, it is unclear whether it is possible to align (stronger) LLMs with superhuman capabilities with (weaker) human feedback without degrading their capabilities. This is an instance of the weak-to-strong generalization problem: using weaker (less capable) feedback to train a stronger (more capable) model. We prove that weak-to-strong generalization is possible by eliciting latent knowledge from pre-trained LLMs. In particular, we cast the weak-to-strong generalization problem as a transfer learning problem in which we wish to transfer a latent concept from a weak model to a strong pre-trained model. We prove that a naive fine-tuning approach suffers from fundamental limitations, but an alternative refinement-based approach suggested by the problem structure provably overcomes the limitations of fine-tuning. Finally, we demonstrate the practical applicability of the refinement approach with three LLM alignment tasks."
                    },
                    {
                        "title": "Efficient multi-prompt evaluation of LLMs",
                        "abstract": "Most popular benchmarks for comparing LLMs rely on a limited set of prompt templates, which may not fully capture the LLMs' abilities and can affect the reproducibility of results on leaderboards. Many recent works empirically verify prompt sensitivity and advocate for changes in LLM evaluation. In this paper, we consider the problem of estimating the performance distribution across many prompt variants instead of finding a single prompt to evaluate with. We introduce PromptEval, a method for estimating performance across a large set of prompts borrowing strength across prompts and examples to produce accurate estimates under practical evaluation budgets. The resulting distribution can be used to obtain performance quantiles to construct various robust performance metrics (e.g., top 95% quantile or median). We prove that PromptEval consistently estimates the performance distribution and demonstrate its efficacy empirically on three prominent LLM benchmarks: MMLU, BIG-bench Hard, and LMentry; for example, PromptEval can accurately estimate performance quantiles across 100 prompt templates on MMLU with a budget equivalent to two single-prompt evaluations. Moreover, we show how PromptEval can be useful in LLM-as-a-judge and best prompt identification applications."
                    },
                    {
                        "title": "Prompt Exploration with Prompt Regression",
                        "abstract": "In the advent of democratized usage of large language models (LLMs), there is a growing desire to systematize LLM prompt creation and selection processes beyond iterative trial-and-error. Prior works majorly focus on searching the space of prompts without accounting for relations between prompt variations. Here we propose a framework, Prompt Exploration with Prompt Regression (PEPR), to predict the effect of prompt combinations given results for individual prompt elements as well as a simple method to select an effective prompt for a given use-case. We evaluate our approach with open-source LLMs of different sizes on several different tasks."
                    },
                    {
                        "title": "Compress then Serve: Serving Thousands of LoRA Adapters with Little Overhead",
                        "abstract": "Fine-tuning large language models (LLMs) with low-rank adaptations (LoRAs) has become common practice, often yielding numerous copies of the same LLM differing only in their LoRA updates. This paradigm presents challenges for systems that serve real-time responses to queries that each involve a different LoRA. Prior works optimize the design of such systems but still require continuous loading and offloading of LoRAs, as it is infeasible to store thousands of LoRAs in GPU memory. To mitigate this issue, we investigate the efficacy of model compression when serving LoRAs. We propose a method for joint compression of LoRAs into a shared basis paired with LoRA-specific scaling matrices. We extend our algorithm to learn clusters of LoRAs that are more amenable to joint compression, allowing it to scale gracefully to large LoRA collections. Our experiments with up to 500 LoRAs demonstrate that compressed LoRAs preserve performance while offering major throughput gains in realistic serving scenarios with over a thousand LoRAs, maintaining 80% of the throughput of serving a single LoRA."
                    },
                    {
                        "title": "Distributional Preference Alignment of LLMs via Optimal Transport",
                        "abstract": "Current LLM alignment techniques use pairwise human preferences at a sample level, and as such, they do not imply an alignment on the distributional level. We propose in this paper Alignment via Optimal Transport (AOT), a novel method for distributional preference alignment of LLMs. AOT aligns LLMs on unpaired preference data by making the reward distribution of the positive samples stochastically dominant in the first order on the distribution of negative samples. We introduce a convex relaxation of this first-order stochastic dominance and cast it as an optimal transport problem with a smooth and convex cost. Thanks to the one-dimensional nature of the resulting optimal transport problem and the convexity of the cost, it has a closed-form solution via sorting on empirical measures. We fine-tune LLMs with this AOT objective, which enables alignment by penalizing the violation of the stochastic dominance of the reward distribution of the positive samples on the reward distribution of the negative samples. We analyze the sample complexity of AOT by considering the dual of the OT problem and show that it converges at the parametric rate. Empirically, we show on a diverse set of alignment datasets and LLMs that AOT leads to state-of-the-art models in the 7B family of models when evaluated with Open LLM Benchmarks and AlpacaEval."
                    },
                    {
                        "title": "Large Language Model Routing with Benchmark Datasets",
                        "abstract": "There is a rapidly growing number of open-source Large Language Models (LLMs) and benchmark datasets to compare them. While some models dominate these benchmarks, no single model typically achieves the best accuracy in all tasks and use cases. In this work, we address the challenge of selecting the best LLM out of a collection of models for new tasks. We propose a new formulation for the problem, in which benchmark datasets are repurposed to learn a\"router\"model for this LLM selection, and we show that this problem can be reduced to a collection of binary classification tasks. We demonstrate the utility and limitations of learning model routers from various benchmark datasets, where we consistently improve performance upon using any single model for all tasks."
                    },
                    {
                        "title": "Calibrated Data-Dependent Constraints with Exact Satisfaction Guarantees",
                        "abstract": "We consider the task of training machine learning models with data-dependent constraints. Such constraints often arise as empirical versions of expected value constraints that enforce fairness or stability goals. We reformulate data-dependent constraints so that they are calibrated: enforcing the reformulated constraints guarantees that their expected value counterparts are satisfied with a user-prescribed probability. The resulting optimization problem is amendable to standard stochastic optimization algorithms, and we demonstrate the efficacy of our method on a fairness-sensitive classification task where we wish to guarantee the classifier's fairness (at test time)."
                    },
                    {
                        "title": "GeRA: Label-Efficient Geometrically Regularized Alignment",
                        "abstract": "Pretrained unimodal encoders incorporate rich semantic information into embedding space structures. To be similarly informative, multi-modal encoders typically require massive amounts of paired data for alignment and training. We introduce a semi-supervised Geometrically Regularized Alignment (GeRA) method to align the embedding spaces of pretrained unimodal encoders in a label-efficient way. Our method leverages the manifold geometry of unpaired (unlabeled) data to improve alignment performance. To prevent distortions to local geometry during the alignment process, potentially disrupting semantic neighborhood structures and causing misalignment of unobserved pairs, we introduce a geometric loss term. This term is built upon a diffusion operator that captures the local manifold geometry of the unimodal pretrained encoders. GeRA is modality-agnostic and thus can be used to align pretrained encoders from any data modalities. We provide empirical evidence to the effectiveness of our method in the domains of speech-text and image-text alignment. Our experiments demonstrate significant improvement in alignment quality compared to a variaty of leading baselines, especially with a small amount of paired data, using our proposed geometric regularization."
                    },
                    {
                        "title": "Weak Supervision Performance Evaluation via Partial Identification",
                        "abstract": "Programmatic Weak Supervision (PWS) enables supervised model training without direct access to ground truth labels, utilizing weak labels from heuristics, crowdsourcing, or pre-trained models. However, the absence of ground truth complicates model evaluation, as traditional metrics such as accuracy, precision, and recall cannot be directly calculated. In this work, we present a novel method to address this challenge by framing model evaluation as a partial identification problem and estimating performance bounds using Fr\\'echet bounds. Our approach derives reliable bounds on key metrics without requiring labeled data, overcoming core limitations in current weak supervision evaluation techniques. Through scalable convex optimization, we obtain accurate and computationally efficient bounds for metrics including accuracy, precision, recall, and F1-score, even in high-dimensional settings. This framework offers a robust approach to assessing model quality without ground truth labels, enhancing the practicality of weakly supervised learning for real-world applications."
                    },
                    {
                        "title": "Fusing Models with Complementary Expertise",
                        "abstract": "Training AI models that generalize across tasks and domains has long been among the open problems driving AI research. The emergence of Foundation Models made it easier to obtain expert models for a given task, but the heterogeneity of data that may be encountered at test time often means that any single expert is insufficient. We consider the Fusion of Experts (FoE) problem of fusing outputs of expert models with complementary knowledge of the data distribution and formulate it as an instance of supervised learning. Our method is applicable to both discriminative and generative tasks and leads to significant performance improvements in image and text classification, text summarization, multiple-choice QA, and automatic evaluation of generated text. We also extend our method to the\"frugal\"setting where it is desired to reduce the number of expert model evaluations at test time. Our implementation is publicly available at https://github.com/hwang595/FoE-ICLR2024."
                    },
                    {
                        "title": "Simple Disentanglement of Style and Content in Visual Representations",
                        "abstract": "Learning visual representations with interpretable features, i.e., disentangled representations, remains a challenging problem. Existing methods demonstrate some success but are hard to apply to large-scale vision datasets like ImageNet. In this work, we propose a simple post-processing framework to disentangle content and style in learned representations from pre-trained vision models. We model the pre-trained features probabilistically as linearly entangled combinations of the latent content and style factors and develop a simple disentanglement algorithm based on the probabilistic model. We show that the method provably disentangles content and style features and verify its efficacy empirically. Our post-processed features yield significant domain generalization performance improvements when the distribution shift occurs due to style changes or style-related spurious correlations."
                    },
                    {
                        "title": "Risk Aware Benchmarking of Large Language Models",
                        "abstract": "We propose a distributional framework for benchmarking socio-technical risks of foundation models with quantified statistical significance. Our approach hinges on a new statistical relative testing based on first and second order stochastic dominance of real random variables. We show that the second order statistics in this test are linked to mean-risk models commonly used in econometrics and mathematical finance to balance risk and utility when choosing between alternatives. Using this framework, we formally develop a risk-aware approach for foundation model selection given guardrails quantified by specified metrics. Inspired by portfolio optimization and selection theory in mathematical finance, we define a metrics portfolio for each model as a means to aggregate a collection of metrics, and perform model selection based on the stochastic dominance of these portfolios. The statistical significance of our tests is backed theoretically by an asymptotic analysis via central limit theorems instantiated in practice via a bootstrap variance estimate. We use our framework to compare various large language models regarding risks related to drifting from instructions and outputting toxic content."
                    },
                    {
                        "title": "Fairness Evaluation in Text Classification: Machine Learning Practitioner Perspectives of Individual and Group Fairness",
                        "abstract": "Mitigating algorithmic bias is a critical task in the development and deployment of machine learning models. While several toolkits exist to aid machine learning practitioners in addressing fairness issues, little is known about the strategies practitioners employ to evaluate model fairness and what factors influence their assessment, particularly in the context of text classification. Two common approaches of evaluating the fairness of a model are group fairness and individual fairness. We run a study with Machine Learning practitioners (n=24) to understand the strategies used to evaluate models. Metrics presented to practitioners (group vs. individual fairness) impact which models they consider fair. Participants focused on risks associated with underpredicting / overpredicting and model sensitivity relative to identity token manipulations. We discover fairness assessment strategies involving personal experiences or how users form groups of identity tokens to test model fairness. We provide recommendations for interactive tools for evaluating fairness in text classification."
                    },
                    {
                        "title": "Rewiring with Positional Encodings for Graph Neural Networks",
                        "abstract": "Several recent works use positional encodings to extend the receptive fields of graph neural network (GNN) layers equipped with attention mechanisms. These techniques, however, extend receptive fields to the complete graph, at substantial computational cost and risking a change in the inductive biases of conventional GNNs, or require complex architecture adjustments. As a conservative alternative, we use positional encodings to expand receptive fields to $r$-hop neighborhoods. More specifically, our method augments the input graph with additional nodes/edges and uses positional encodings as node and/or edge features. We thus modify graphs before inputting them to a downstream GNN model, instead of modifying the model itself. This makes our method model-agnostic, i.e., compatible with any of the existing GNN architectures. We also provide examples of positional encodings that are lossless with a one-to-one map between the original and the modified graphs. We demonstrate that extending receptive fields via positional encodings and a virtual fully-connected node significantly improves GNN performance and alleviates over-squashing using small $r$. We obtain improvements on a variety of models and datasets and reach competitive performance using traditional GNNs or graph Transformers."
                    }
                ]
            },
            "ed598ae3-899e-4327-9923-77e0805495bf": {
                "pk": "ed598ae3-899e-4327-9923-77e0805495bf",
                "name": "Yuekai Sun",
                "collaborators": [
                    "M. Yurochkin",
                    "Felipe Maia Polo",
                    "Moulinath Banerjee",
                    "Subha Maity",
                    "Lucas Weber",
                    "Leshem Choshen",
                    "Seamus Somerstep",
                    "Ronald Xu",
                    "M'irian Silva",
                    "Nimrod Shabtay",
                    "Sivan Doveh",
                    "Wei Lin",
                    "M. Mirza",
                    "Leshem Chosen",
                    "Assaf Arbelle",
                    "Leonid Karlinsky",
                    "Raja Giryes",
                    "Gongjun Xu",
                    "Lilian Ngweta",
                    "Mayank Agarwal",
                    "Alex Gittens",
                    "Ya'acov Ritov",
                    "Y. Ritov",
                    "Onkar Bhardwaj",
                    "Allysson Flavio Melo de Oliveira",
                    "Michael Feffer",
                    "Daniele Bracale",
                    "Tal Shnitzer",
                    "Anthony Ou",
                    "Kate Soule",
                    "Justin Solomon",
                    "Neil Thompson",
                    "Hongyi Wang",
                    "Souvik Kundu",
                    "Eric P. Xing"
                ],
                "domain": [
                    "Large Language Models",
                    "Multi-modal Learning",
                    "Benchmarking",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "LiveXiv -- A Multi-Modal Live Benchmark Based on Arxiv Papers Content",
                        "abstract": "The large-scale training of multi-modal models on data scraped from the web has shown outstanding utility in infusing these models with the required world knowledge to perform effectively on multiple downstream tasks. However, one downside of scraping data from the web can be the potential sacrifice of the benchmarks on which the abilities of these models are often evaluated. To safeguard against test data contamination and to truly test the abilities of these foundation models we propose LiveXiv: A scalable evolving live benchmark based on scientific ArXiv papers. LiveXiv accesses domain-specific manuscripts at any given timestamp and proposes to automatically generate visual question-answer pairs (VQA). This is done without any human-in-the-loop, using the multi-modal content in the manuscripts, like graphs, charts, and tables. Moreover, we introduce an efficient evaluation approach that estimates the performance of all models on the evolving benchmark using evaluations of only a subset of models. This significantly reduces the overall evaluation cost. We benchmark multiple open and proprietary Large Multi-modal Models (LMMs) on the first version of our benchmark, showing its challenging nature and exposing the models true abilities, avoiding contamination. Lastly, in our commitment to high quality, we have collected and evaluated a manually verified subset. By comparing its overall results to our automatic annotations, we have found that the performance variance is indeed minimal (<2.5%). Our dataset is available online on HuggingFace, and our code will be available here."
                    },
                    {
                        "title": "tinyBenchmarks: evaluating LLMs with fewer examples",
                        "abstract": "The versatility of large language models (LLMs) led to the creation of diverse benchmarks that thoroughly test a variety of language models' abilities. These benchmarks consist of tens of thousands of examples making evaluation of LLMs very expensive. In this paper, we investigate strategies to reduce the number of evaluations needed to assess the performance of an LLM on several key benchmarks. For example, we show that to accurately estimate the performance of an LLM on MMLU, a popular multiple-choice QA benchmark consisting of 14K examples, it is sufficient to evaluate this LLM on 100 curated examples. We release evaluation tools and tiny versions of popular benchmarks: Open LLM Leaderboard, MMLU, HELM, and AlpacaEval 2.0. Our empirical analysis demonstrates that these tools and tiny benchmarks are sufficient to reliably and efficiently reproduce the original evaluation results."
                    },
                    {
                        "title": "Aligners: Decoupling LLMs and Alignment",
                        "abstract": "Large Language Models (LLMs) need to be aligned with human expectations to ensure their safety and utility in most applications. Alignment is challenging, costly, and needs to be repeated for every LLM and alignment criterion. We propose to decouple LLMs and alignment by training aligner models that can be used to align any LLM for a given criteria on an as-needed basis, thus also reducing the potential negative impacts of alignment on performance. Our recipe for training the aligner models solely relies on synthetic data generated with a (prompted) LLM and can be easily adjusted for a variety of alignment criteria. We use the same synthetic data to train inspectors, binary miss-alignment classification models to guide a\"squad\"of multiple aligners. Our empirical results demonstrate consistent improvements when applying aligner squad to various LLMs, including chat-aligned models, across several instruction-following and red-teaming datasets."
                    },
                    {
                        "title": "A statistical framework for weak-to-strong generalization",
                        "abstract": "Modern large language model (LLM) alignment techniques rely on human feedback, but it is unclear whether the techniques fundamentally limit the capabilities of aligned LLMs. In particular, it is unclear whether it is possible to align (stronger) LLMs with superhuman capabilities with (weaker) human feedback without degrading their capabilities. This is an instance of the weak-to-strong generalization problem: using weaker (less capable) feedback to train a stronger (more capable) model. We prove that weak-to-strong generalization is possible by eliciting latent knowledge from pre-trained LLMs. In particular, we cast the weak-to-strong generalization problem as a transfer learning problem in which we wish to transfer a latent concept from a weak model to a strong pre-trained model. We prove that a naive fine-tuning approach suffers from fundamental limitations, but an alternative refinement-based approach suggested by the problem structure provably overcomes the limitations of fine-tuning. Finally, we demonstrate the practical applicability of the refinement approach with three LLM alignment tasks."
                    },
                    {
                        "title": "Learning In Reverse Causal Strategic Environments With Ramifications on Two Sided Markets",
                        "abstract": "Motivated by equilibrium models of labor markets, we develop a formulation of causal strategic classification in which strategic agents can directly manipulate their outcomes. As an application, we compare employers that anticipate the strategic response of a labor force with employers that do not. We show through a combination of theory and experiment that employers with performatively optimal hiring policies improve employer reward, labor force skill level, and in some cases labor force equity. On the other hand, we demonstrate that performative employers harm labor force utility and fail to prevent discrimination in other cases."
                    },
                    {
                        "title": "Efficient multi-prompt evaluation of LLMs",
                        "abstract": "Most popular benchmarks for comparing LLMs rely on a limited set of prompt templates, which may not fully capture the LLMs' abilities and can affect the reproducibility of results on leaderboards. Many recent works empirically verify prompt sensitivity and advocate for changes in LLM evaluation. In this paper, we consider the problem of estimating the performance distribution across many prompt variants instead of finding a single prompt to evaluate with. We introduce PromptEval, a method for estimating performance across a large set of prompts borrowing strength across prompts and examples to produce accurate estimates under practical evaluation budgets. The resulting distribution can be used to obtain performance quantiles to construct various robust performance metrics (e.g., top 95% quantile or median). We prove that PromptEval consistently estimates the performance distribution and demonstrate its efficacy empirically on three prominent LLM benchmarks: MMLU, BIG-bench Hard, and LMentry; for example, PromptEval can accurately estimate performance quantiles across 100 prompt templates on MMLU with a budget equivalent to two single-prompt evaluations. Moreover, we show how PromptEval can be useful in LLM-as-a-judge and best prompt identification applications."
                    },
                    {
                        "title": "Prompt Exploration with Prompt Regression",
                        "abstract": "In the advent of democratized usage of large language models (LLMs), there is a growing desire to systematize LLM prompt creation and selection processes beyond iterative trial-and-error. Prior works majorly focus on searching the space of prompts without accounting for relations between prompt variations. Here we propose a framework, Prompt Exploration with Prompt Regression (PEPR), to predict the effect of prompt combinations given results for individual prompt elements as well as a simple method to select an effective prompt for a given use-case. We evaluate our approach with open-source LLMs of different sizes on several different tasks."
                    },
                    {
                        "title": "Learning the Distribution Map in Reverse Causal Performative Prediction",
                        "abstract": "In numerous predictive scenarios, the predictive model affects the sampling distribution; for example, job applicants often meticulously craft their resumes to navigate through a screening systems. Such shifts in distribution are particularly prevalent in the realm of social computing, yet, the strategies to learn these shifts from data remain remarkably limited. Inspired by a microeconomic model that adeptly characterizes agents' behavior within labor markets, we introduce a novel approach to learn the distribution shift. Our method is predicated on a reverse causal model, wherein the predictive model instigates a distribution shift exclusively through a finite set of agents' actions. Within this framework, we employ a microfoundation model for the agents' actions and develop a statistically justified methodology to learn the distribution shift map, which we demonstrate to be effective in minimizing the performative prediction risk."
                    },
                    {
                        "title": "Large Language Model Routing with Benchmark Datasets",
                        "abstract": "There is a rapidly growing number of open-source Large Language Models (LLMs) and benchmark datasets to compare them. While some models dominate these benchmarks, no single model typically achieves the best accuracy in all tasks and use cases. In this work, we address the challenge of selecting the best LLM out of a collection of models for new tasks. We propose a new formulation for the problem, in which benchmark datasets are repurposed to learn a\"router\"model for this LLM selection, and we show that this problem can be reduced to a collection of binary classification tasks. We demonstrate the utility and limitations of learning model routers from various benchmark datasets, where we consistently improve performance upon using any single model for all tasks."
                    },
                    {
                        "title": "Weak Supervision Performance Evaluation via Partial Identification",
                        "abstract": "Programmatic Weak Supervision (PWS) enables supervised model training without direct access to ground truth labels, utilizing weak labels from heuristics, crowdsourcing, or pre-trained models. However, the absence of ground truth complicates model evaluation, as traditional metrics such as accuracy, precision, and recall cannot be directly calculated. In this work, we present a novel method to address this challenge by framing model evaluation as a partial identification problem and estimating performance bounds using Fr\\'echet bounds. Our approach derives reliable bounds on key metrics without requiring labeled data, overcoming core limitations in current weak supervision evaluation techniques. Through scalable convex optimization, we obtain accurate and computationally efficient bounds for metrics including accuracy, precision, recall, and F1-score, even in high-dimensional settings. This framework offers a robust approach to assessing model quality without ground truth labels, enhancing the practicality of weakly supervised learning for real-world applications."
                    },
                    {
                        "title": "Fusing Models with Complementary Expertise",
                        "abstract": "Training AI models that generalize across tasks and domains has long been among the open problems driving AI research. The emergence of Foundation Models made it easier to obtain expert models for a given task, but the heterogeneity of data that may be encountered at test time often means that any single expert is insufficient. We consider the Fusion of Experts (FoE) problem of fusing outputs of expert models with complementary knowledge of the data distribution and formulate it as an instance of supervised learning. Our method is applicable to both discriminative and generative tasks and leads to significant performance improvements in image and text classification, text summarization, multiple-choice QA, and automatic evaluation of generated text. We also extend our method to the\"frugal\"setting where it is desired to reduce the number of expert model evaluations at test time. Our implementation is publicly available at https://github.com/hwang595/FoE-ICLR2024."
                    },
                    {
                        "title": "Conditional independence testing under misspecified inductive biases",
                        "abstract": "Conditional independence (CI) testing is a fundamental and challenging task in modern statistics and machine learning. Many modern methods for CI testing rely on powerful supervised learning methods to learn regression functions or Bayes predictors as an intermediate step; we refer to this class of tests as regression-based tests. Although these methods are guaranteed to control Type-I error when the supervised learning methods accurately estimate the regression functions or Bayes predictors of interest, their behavior is less understood when they fail due to misspecified inductive biases; in other words, when the employed models are not flexible enough or when the training algorithm does not induce the desired predictors. Then, we study the performance of regression-based CI tests under misspecified inductive biases. Namely, we propose new approximations or upper bounds for the testing errors of three regression-based tests that depend on misspecification errors. Moreover, we introduce the Rao-Blackwellized Predictor Test (RBPT), a regression-based CI test robust against misspecified inductive biases. Finally, we conduct experiments with artificial and real data, showcasing the usefulness of our theory and methods."
                    }
                ]
            }
        }
    },
    "2403.00158": {
        "paper_data": {
            "title": "Automated Efficient Estimation using Monte Carlo Efficient Influence Functions",
            "url": "http://arxiv.org/abs/2403.00158v2",
            "arxiv_id": "2403.00158",
            "authors": [
                "Raj Agrawal",
                "Sam Witty",
                "Andy Zane",
                "Eli Bingham"
            ],
            "abstract": "Many practical problems involve estimating low dimensional statistical quantities with high-dimensional models and datasets. Several approaches address these estimation tasks based on the theory of influence functions, such as debiased/double ML or targeted minimum loss estimation. This paper introduces \\textit{Monte Carlo Efficient Influence Functions} (MC-EIF), a fully automated technique for approximating efficient influence functions that integrates seamlessly with existing differentiable probabilistic programming systems. MC-EIF automates efficient statistical estimation for a broad class of models and target functionals that would previously require rigorous custom analysis. We prove that MC-EIF is consistent, and that estimators using MC-EIF achieve optimal $\\sqrt{N}$ convergence rates. We show empirically that estimators using MC-EIF are at parity with estimators using analytic EIFs. Finally, we demonstrate a novel capstone example using MC-EIF for optimal portfolio selection.",
            "introduction": "   1 Introduction  Over the past several decades, there has been remarkable research progress in methods for robust and efficient statistical estimation, especially for high dimensional problems. A particularly compelling class of such methods are built on a foundation of efficient influence functions (EIF), i.e., functional derivatives in the space of probability distributions\u00a0(Kennedy, 2022). These methods have been particularly fruitful in causal inference applications, where estimating quantities such as the average treatment effect require modeling high-dimensional nuisance parameters relating confounders to treatment and outcome variables. Intuitively, these methods focus finite statistical resources on quantities that matter, and not on nuisance parameters that only indirectly inform the statistical quantities we wish to estimate.   Despite their successes, estimation methods based on the EIF have lagged behind the kinds of automation that machine learning practitioners have grown accustomed to, instead requiring complex manual derivation on a case-by-case basis. This is contrasted with the generality of automatic differentiation (AD)\u00a0(Baydin et\u00a0al., 2018) systems and probabilistic programming languages (PPLs) such as Pyro\u00a0(Bingham et\u00a0al., 2019), which automate numerical computations for probabilistic inference. EIF-based estimators have historically eluded this level of automation and generality, because exact recovery of the EIF requires solving high-dimensional integral equations.   Contributions. In this paper we introduce Monte Carlo Efficient Influence Functions (MC-EIF), a general and automated technique for numerically computing EIFs using only quantities that are already available from existing AD and PPL systems. Our key insight is that EIFs can be expressed equivalently as a product of (i) the gradient of the functional, (ii) the inverse Fisher information matrix, and (iii) the gradient of the log-likelihood, as shown in Theorem\u00a03.4 in Section\u00a03. In Section\u00a04, we show how MC-EIF can be used to automatically construct a variety of efficient estimators for a broad class of models and functionals, avoiding the need for complex manual and error-prone derivations.   In summary, we show that: (i) MC-EIF provides accurate estimates of the true EIF, enabling efficient estimation, and (ii) MC-EIF is very general, applying to many functionals and models that can be written as probabilistic programs111We make these conditions precise in Section\u00a03..   Outline. We review related work in Section\u00a01.1 and provide background in Section\u00a02. In Section\u00a03, we introduce MC-EIF and provide a non-asymptotic error bound on the quality of our approximation. We show how estimators using MC-EIF achieve the same asymptotic guarantees as using analytic EIFs in Section\u00a04. In Section\u00a05, we show empirically that MC-EIF produces more accurate estimates of the EIF than existing automated approaches, and using MC-EIF as a drop-in replacement for the analytic EIF does not degrade estimation accuracy in a variety of benchmarks.    1.1 Related Work  Influence function-based estimators have a rich history in the statistics and machine learning literature\u00a0(Bickel et\u00a0al., 1993; Tsiatis, 2006; Kennedy, 2022; Hines et\u00a0al., 2022). Despite their effectiveness, these methods have historically required custom and complex mathematical analysis for specific combinations of models and functional. Even an incomplete survey of the influence function-based estimation literature yields a large collection of complex scenario-specific research. We enumerate some of this research in\u00a0Appendix\u00a0B. Importantly, our work does not aim to introduce novel efficient estimators; instead it aims to lower the",
            "references": [
                {
                    "title": "Targeted Learning on Variable Importance Measure for Heterogeneous Treatment Effect",
                    "abstract": "Quantifying the heterogeneity of treatment effect is important for understanding how a commercial product or medical treatment affects different subgroups in a population. Beyond the overall impact reflected parameters like the average treatment effect, the analysis of treatment effect heterogeneity further reveals details on the importance of different covariates and how they lead to different treatment impacts. One relevant parameter that addresses such heterogeneity is the variance of treatment effect across different covariate groups, however the treatment effect is defined. One can also derive variable importance parameters that measure (and rank) how much of treatment effect heterogeneity is explained by a targeted subset of covariates. In this article, we propose a new targeted maximum likelihood estimator for a treatment effect variable importance measure. This estimator is a pure plug-in estimator that consists of two steps: 1) the initial estimation of relevant components to plug in and 2) an iterative updating step to optimize the bias-variance tradeoff. The simulation results show that this TMLE estimator has competitive performance in terms of lower bias and better confidence interval coverage compared to the simple substitution estimator and the estimating equation estimator. The application of this method also demonstrates the advantage of a substitution estimator, which always respects the global constraints on the data distribution and that the estimand is a particular function of the distribution."
                },
                {
                    "title": "Doubly Robust Self-Training",
                    "abstract": "Self-training is an important technique for solving semi-supervised learning problems. It leverages unlabeled data by generating pseudo-labels and combining them with a limited labeled dataset for training. The effectiveness of self-training heavily relies on the accuracy of these pseudo-labels. In this paper, we introduce doubly robust self-training, a novel semi-supervised algorithm that provably balances between two extremes. When the pseudo-labels are entirely incorrect, our method reduces to a training process solely using labeled data. Conversely, when the pseudo-labels are completely accurate, our method transforms into a training process utilizing all pseudo-labeled data and labeled data, thus increasing the effective sample size. Through empirical evaluations on both the ImageNet dataset for image classification and the nuScenes autonomous driving dataset for 3D object detection, we demonstrate the superiority of the doubly robust loss over the standard self-training baseline."
                },
                {
                    "title": "The Fundamental Limits of Structure-Agnostic Functional Estimation",
                    "abstract": "Many recent developments in causal inference, and functional estimation problems more generally, have been motivated by the fact that classical one-step (first-order) debiasing methods, or their more recent sample-split double machine-learning avatars, can outperform plugin estimators under surprisingly weak conditions. These first-order corrections improve on plugin estimators in a black-box fashion, and consequently are often used in conjunction with powerful off-the-shelf estimation methods. These first-order methods are however provably suboptimal in a minimax sense for functional estimation when the nuisance functions live in Holder-type function spaces. This suboptimality of first-order debiasing has motivated the development of\"higher-order\"debiasing methods. The resulting estimators are, in some cases, provably optimal over Holder-type spaces, but both the estimators which are minimax-optimal and their analyses are crucially tied to properties of the underlying function space. In this paper we investigate the fundamental limits of structure-agnostic functional estimation, where relatively weak conditions are placed on the underlying nuisance functions. We show that there is a strong sense in which existing first-order methods are optimal. We achieve this goal by providing a formalization of the problem of functional estimation with black-box nuisance function estimates, and deriving minimax lower bounds for this problem. Our results highlight some clear tradeoffs in functional estimation -- if we wish to remain agnostic to the underlying nuisance function spaces, impose only high-level rate conditions, and maintain compatibility with black-box nuisance estimators then first-order methods are optimal. When we have an understanding of the structure of the underlying nuisance functions then carefully constructed higher-order estimators can outperform first-order estimators."
                },
                {
                    "title": "Targeted Maximum Likelihood Based Estimation for Longitudinal Mediation Analysis",
                    "abstract": "Causal mediation analysis with random interventions has become an area of significant interest for understanding time-varying effects with longitudinal and survival outcomes. To tackle causal and statistical challenges due to the complex longitudinal data structure with time-varying confounders, competing risks, and informative censoring, there exists a general desire to combine machine learning techniques and semiparametric theory. In this manuscript, we focus on targeted maximum likelihood estimation (TMLE) of longitudinal natural direct and indirect effects defined with random interventions. The proposed estimators are multiply robust, locally efficient, and directly estimate and update the conditional densities that factorize data likelihoods. We utilize the highly adaptive lasso (HAL) and projection representations to derive new estimators (HAL-EIC) of the efficient influence curves of longitudinal mediation problems and propose a fast one-step TMLE algorithm using HAL-EIC while preserving the asymptotic properties. The proposed method can be generalized for other longitudinal causal parameters that are smooth functions of data likelihoods, and thereby provides a novel and flexible statistical toolbox."
                },
                {
                    "title": "Optimal randomized multilevel Monte Carlo for repeatedly nested expectations",
                    "abstract": "The estimation of repeatedly nested expectations is a challenging task that arises in many real-world systems. However, existing methods generally suffer from high computational costs when the number of nestings becomes large. Fix any non-negative integer $D$ for the total number of nestings. Standard Monte Carlo methods typically cost at least $\\mathcal{O}(\\varepsilon^{-(2+D)})$ and sometimes $\\mathcal{O}(\\varepsilon^{-2(1+D)})$ to obtain an estimator up to $\\varepsilon$-error. More advanced methods, such as multilevel Monte Carlo, currently only exist for $D = 1$. In this paper, we propose a novel Monte Carlo estimator called $\\mathsf{READ}$, which stands for\"Recursive Estimator for Arbitrary Depth.'' Our estimator has an optimal computational cost of $\\mathcal{O}(\\varepsilon^{-2})$ for every fixed $D$ under suitable assumptions, and a nearly optimal computational cost of $\\mathcal{O}(\\varepsilon^{-2(1 + \\delta)})$ for any $0<\\delta<\\frac12$ under much more general assumptions. Our estimator is also unbiased, which makes it easy to parallelize. The key ingredients in our construction are an observation of the problem's recursive structure and the recursive use of the randomized multilevel Monte Carlo method."
                },
                {
                    "title": "ADEV: Sound Automatic Differentiation of Expected Values of Probabilistic Programs",
                    "abstract": "Optimizing the expected values of probabilistic processes is a central problem in computer science and its applications, arising in fields ranging from artificial intelligence to operations research to statistical computing. Unfortunately, automatic differentiation techniques developed for deterministic programs do not in general compute the correct gradients needed for widely used solutions based on gradient-based optimization. In this paper, we present ADEV, an extension to forward-mode AD that correctly differentiates the expectations of probabilistic processes represented as programs that make random choices. Our algorithm is a source-to-source program transformation on an expressive, higher-order language for probabilistic computation, with both discrete and continuous probability distributions. The result of our transformation is a new probabilistic program, whose expected return value is the derivative of the original program\u2019s expectation. This output program can be run to generate unbiased Monte Carlo estimates of the desired gradient, that can be used within the inner loop of stochastic gradient descent. We prove ADEV correct using logical relations over the denotations of the source and target probabilistic programs. Because it modularly extends forward-mode AD, our algorithm lends itself to a concise implementation strategy, which we exploit to develop a prototype in just a few dozen lines of Haskell (https://github.com/probcomp/adev)."
                },
                {
                    "title": "Efficient targeted learning of heterogeneous treatment effects for multiple subgroups",
                    "abstract": "In biomedical science, analyzing treatment effect heterogeneity plays an essential role in assisting personalized medicine. The main goals of analyzing treatment effect heterogeneity include estimating treatment effects in clinically relevant subgroups and predicting whether a patient subpopulation might benefit from a particular treatment. Conventional approaches often evaluate the subgroup treatment effects via parametric modeling and can thus be susceptible to model mis\u2010specifications. In this paper, we take a model\u2010free semiparametric perspective and aim to efficiently evaluate the heterogeneous treatment effects of multiple subgroups simultaneously under the one\u2010step targeted maximum\u2010likelihood estimation (TMLE) framework. When the number of subgroups is large, we further expand this path of research by looking at a variation of the one\u2010step TMLE that is robust to the presence of small estimated propensity scores in finite samples. From our simulations, our method demonstrates substantial finite sample improvements compared to conventional methods. In a case study, our method unveils the potential treatment effect heterogeneity of rs12916\u2010T allele (a proxy for statin usage) in decreasing Alzheimer's disease risk."
                },
                {
                    "title": "A Cross-Validated Targeted Maximum Likelihood Estimator for Data-Adaptive Experiment Selection Applied to the Augmentation of RCT Control Arms with External Data",
                    "abstract": "Augmenting the control arm of a randomized controlled trial (RCT) with external data may increase power at the risk of introducing bias. Existing data fusion estimators generally rely on stringent assumptions or may have decreased coverage or power in the presence of bias. Framing the problem as one of data-adaptive experiment selection, potential experiments include the RCT only or the RCT combined with different candidate real-world datasets. To select and analyze the experiment with the optimal bias-variance tradeoff, we develop a novel experiment-selector cross-validated targeted maximum likelihood estimator (ES-CVTMLE). The ES-CVTMLE uses two bias estimates: 1) a function of the difference in conditional mean outcome under control between the RCT and combined experiments and 2) an estimate of the average treatment effect on a negative control outcome (NCO). We define the asymptotic distribution of the ES-CVTMLE under varying magnitudes of bias and construct confidence intervals by Monte Carlo simulation. In simulations involving violations of identification assumptions, the ES-CVTMLE had better coverage than test-then-pool approaches and an NCO-based bias adjustment approach and higher power than one implementation of a Bayesian dynamic borrowing approach. We further demonstrate the ability of the ES-CVTMLE to distinguish biased from unbiased external controls through a re-analysis of the effect of liraglutide on glycemic control from the LEADER trial. The ES-CVTMLE has the potential to improve power while providing relatively robust inference for future hybrid RCT-RWD studies."
                },
                {
                    "title": "If Influence Functions are the Answer, Then What is the Question?",
                    "abstract": "Influence functions efficiently estimate the effect of removing a single training data point on a model's learned parameters. While influence estimates align well with leave-one-out retraining for linear models, recent works have shown this alignment is often poor in neural networks. In this work, we investigate the specific factors that cause this discrepancy by decomposing it into five separate terms. We study the contributions of each term on a variety of architectures and datasets and how they vary with factors such as network width and training time. While practical influence function estimates may be a poor match to leave-one-out retraining for nonlinear networks, we show they are often a good approximation to a different object we term the proximal Bregman response function (PBRF). Since the PBRF can still be used to answer many of the questions motivating influence functions, such as identifying influential or mislabeled examples, our results suggest that current algorithms for influence function estimation give more informative results than previous error analyses would suggest."
                },
                {
                    "title": "Semiparametric doubly robust targeted double machine learning: a review",
                    "abstract": "In this review we cover the basics of efficient nonparametric parameter estimation (also called functional estimation), with a focus on parameters that arise in causal inference problems. We review both efficiency bounds (i.e., what is the best possible performance for estimating a given parameter?) and the analysis of particular estimators (i.e., what is this estimator's error, and does it attain the efficiency bound?) under weak assumptions. We emphasize minimax-style efficiency bounds, worked examples, and practical shortcuts for easing derivations. We gloss over most technical details, in the interest of highlighting important concepts and providing intuition for main ideas."
                },
                {
                    "title": "A Free Lunch with Influence Functions? Improving Neural Network Estimates with Concepts from Semiparametric Statistics",
                    "abstract": "Parameter estimation in empirical fields is usually undertaken using parametric models, and such models readily facilitate statistical inference. Unfortunately, they are unlikely to be sufficiently flexible to be able to adequately model real-world phenomena, and may yield biased estimates. Conversely, non-parametric approaches are flexible but do not readily facilitate statistical inference and may still exhibit residual bias. We explore the potential for Influence Functions (IFs) to (a) improve initial estimators without needing more data (b) increase model robustness and (c) facilitate statistical inference. We begin with a broad introduction to IFs, and propose a neural network method 'MultiNet', which seeks the diversity of an ensemble using a single architecture. We also introduce variants on the IF update step which we call 'MultiStep', and provide a comprehensive evaluation of different approaches. The improvements are found to be dataset dependent, indicating an interaction between the methods used and nature of the data generating process. Our experiments highlight the need for practitioners to check the consistency of their findings, potentially by undertaking multiple analyses with different combinations of estimators. We also show that it is possible to improve existing neural networks for `free', without needing more data, and without needing to retrain them."
                },
                {
                    "title": "Demystifying Statistical Learning Based on Efficient Influence Functions",
                    "abstract": "Abstract Evaluation of treatment effects and more general estimands is typically achieved via parametric modeling, which is unsatisfactory since model misspecification is likely. Data-adaptive model building (e.g., statistical/machine learning) is commonly employed to reduce the risk of misspecification. Na\u00efve use of such methods, however, delivers estimators whose bias may shrink too slowly with sample size for inferential methods to perform well, including those based on the bootstrap. Bias arises because standard data-adaptive methods are tuned toward minimal prediction error as opposed to, for example, minimal MSE in the estimator. This may cause excess variability that is difficult to acknowledge, due to the complexity of such strategies. Building on results from nonparametric statistics, targeted learning and debiased machine learning overcome these problems by constructing estimators using the estimand\u2019s efficient influence function under the nonparametric model. These increasingly popular methodologies typically assume that the efficient influence function is given, or that the reader is familiar with its derivation. In this article, we focus on derivation of the efficient influence function and explain how it may be used to construct statistical/machine-learning-based estimators. We discuss the requisite conditions for these estimators to perform well and use diverse examples to convey the broad applicability of the theory."
                },
                {
                    "title": "Two-Stage TMLE to reduce bias and improve efficiency in cluster randomized trials",
                    "abstract": "Summary Cluster randomized trials (CRTs) randomly assign an intervention to groups of individuals (e.g., clinics or communities) and measure outcomes on individuals in those groups. While offering many advantages, this experimental design introduces challenges that are only partially addressed by existing analytic approaches. First, outcomes are often missing for some individuals within clusters. Failing to appropriately adjust for differential outcome measurement can result in biased estimates and inference. Second, CRTs often randomize limited numbers of clusters, resulting in chance imbalances on baseline outcome predictors between arms. Failing to adaptively adjust for these imbalances and other predictive covariates can result in efficiency losses. To address these methodological gaps, we propose and evaluate a novel two-stage targeted minimum loss-based estimator to adjust for baseline covariates in a manner that optimizes precision, after controlling for baseline and postbaseline causes of missing outcomes. Finite sample simulations illustrate that our approach can nearly eliminate bias due to differential outcome measurement, while existing CRT estimators yield misleading results and inferences. Application to real data from the SEARCH community randomized trial demonstrates the gains in efficiency afforded through adaptive adjustment for baseline covariates, after controlling for missingness on individual-level outcomes."
                },
                {
                    "title": "Estimation of time\u2010specific intervention effects on continuously distributed time\u2010to\u2010event outcomes by targeted maximum likelihood estimation",
                    "abstract": "This work considers targeted maximum likelihood estimation (TMLE) of treatment effects on absolute risk and survival probabilities in classical time\u2010to\u2010event settings characterized by right\u2010censoring and competing risks. TMLE is a general methodology combining flexible ensemble learning and semiparametric efficiency theory in a two\u2010step procedure for substitution estimation of causal parameters. We specialize and extend the continuous\u2010time TMLE methods for competing risks settings, proposing a targeting algorithm that iteratively updates cause\u2010specific hazards to solve the efficient influence curve equation for the target parameter. As part of the work, we further detail and implement the recently proposed highly adaptive lasso estimator for continuous\u2010time conditional hazards with L1\u2010penalized Poisson regression. The resulting estimation procedure benefits from relying solely on very mild nonparametric restrictions on the statistical model, thus providing a novel tool for machine\u2010learning\u2010based semiparametric causal inference for continuous\u2010time time\u2010to\u2010event data. We apply the methods to a publicly available dataset on follicular cell lymphoma where subjects are followed over time until disease relapse or death without relapse. The data display important time\u2010varying effects that can be captured by the highly adaptive lasso. In our simulations that are designed to imitate the data, we compare our methods to a similar approach based on random survival forests and to the discrete\u2010time TMLE."
                },
                {
                    "title": "Estimating Identifiable Causal Effects through Double Machine Learning",
                    "abstract": "Identifying causal effects from observational data is a pervasive challenge found throughout the empirical sciences. Very general methods have been developed to decide the identifiability of a causal quantity from a combination of observational data and causal knowledge about the underlying system. In practice, however, there are still challenges to estimating identifiable causal functionals from finite samples. Recently, a method known as double/debiased machine learning (DML) (Chernozhukov et al. 2018) has been proposed to learn parameters leveraging modern machine learning techniques, which is both robust to model misspecification and bias-reducing. Still, DML has only been used for causal estimation in settings when the back-door condition (also known as conditional ignorability) holds. In this paper, we develop a new, general class of estimators for any identifiable causal functionals that exhibit DML properties, which we name DML-ID. In particular, we introduce a complete identification algorithm that returns an influence function (IF) for any identifiable causal functional. We then construct the DML estimator based on the derived IF. We show that DML-ID estimators hold the key properties of debiasedness and doubly robustness. Simulation results corroborate with the theory."
                },
                {
                    "title": "Continuous-time targeted minimum loss-based estimation of intervention-specific mean outcomes",
                    "abstract": "This paper generalizes the targeted minimum loss-based estimation (TMLE) framework to allow for estimating the effects of time-varying interventions in settings where both interventions, covariates, and outcome can happen at subject-specific time-points on an arbitrarily fine time-scale. TMLE is a general template for constructing asymptotically linear substitution estimators for smooth low-dimensional parameters in infinite-dimensional models. Existing longitudinal TMLE methods are developed for data where observations are made on a discrete time-grid. We consider a continuous-time counting process model where intensity measures track the monitoring of subjects, and focus on a low-dimensional target parameter defined as the intervention-specific mean outcome at the end of follow-up. To construct our TMLE algorithm for the given statistical estimation problem we derive an expression for the efficient influence curve and represent the target parameter as a functional of intensities and conditional expectations. The high-dimensional nuisance parameters of our model are estimated and updated in an iterative manner according to separate targeting steps for the involved intensities and conditional expectations. The resulting estimator solves the efficient influence curve equation. We state a general efficiency theorem and describe a highly adaptive lasso estimator for nuisance parameters that allows us to establish asymptotic linearity and efficiency of our estimator under minimal conditions on the underlying statistical model."
                },
                {
                    "title": "Knowledge Distillation as Semiparametric Inference",
                    "abstract": "A popular approach to model compression is to train an inexpensive student model to mimic the class probabilities of a highly accurate but cumbersome teacher model. Surprisingly, this two-step knowledge distillation process often leads to higher accuracy than training the student directly on labeled data. To explain and enhance this phenomenon, we cast knowledge distillation as a semiparametric inference problem with the optimal student model as the target, the unknown Bayes class probabilities as nuisance, and the teacher probabilities as a plug-in nuisance estimate. By adapting modern semiparametric tools, we derive new guarantees for the prediction error of standard distillation and develop two enhancements -- cross-fitting and loss correction -- to mitigate the impact of teacher overfitting and underfitting on student performance. We validate our findings empirically on both tabular and image data and observe consistent improvements from our knowledge distillation enhancements."
                },
                {
                    "title": "Causal mediation analysis with double machine learning",
                    "abstract": "\n This paper combines causal mediation analysis with double machine learning for a data-driven control of observed confounders in a high-dimensional setting. The average indirect effect of a binary treatment and the unmediated direct effect are estimated based on efficient score functions, which are robust w.r.t. misspecifications of the outcome, mediator, and treatment models. This property is key for selecting these models by double machine learning, which is combined with data splitting to prevent overfitting. We demonstrate that the effect estimators are asymptotically normal and n\u22121/2-consistent under specific regularity conditions and investigate the finite sample properties of the suggested methods in a simulation study when considering lasso as machine learner. We also provide an empirical application to the U.S. National Longitudinal Survey of Youth, assessing the indirect effect of health insurance coverage on general health operating via routine checkups as mediator, as well as the direct effect."
                },
                {
                    "title": "Double/debiased machine learning for difference-in-differences models",
                    "abstract": "\n This paper provides an orthogonal extension of the semiparametric difference-in-differences estimator proposed in earlier literature. The proposed estimator enjoys the so-called Neyman orthogonality (Chernozhukov et\u00a0al., 2018), and thus it allows researchers to flexibly use a rich set of machine learning methods in the first-step estimation. It is particularly useful when researchers confront a high-dimensional data set in which the number of potential control variables is larger than the sample size and the conventional nonparametric estimation methods, such as kernel and sieve estimators, do not apply. I apply this orthogonal difference-in-differences estimator to evaluate the effect of tariff reduction on corruption. The empirical results show that tariff reduction decreases corruption in large magnitude."
                },
                {
                    "title": "Estimating covariance and precision matrices along subspaces",
                    "abstract": "We study the accuracy of estimating the covariance and the precision matrix of a $D$-variate sub-Gaussian distribution along a prescribed subspace or direction using the finite sample covariance. Our results show that the estimation accuracy depends almost exclusively on the components of the distribution that correspond to desired subspaces or directions. This is relevant and important for problems where the behavior of data along a lower-dimensional space is of specific interest, such as dimension reduction or structured regression problems. We also show that estimation of precision matrices is almost independent of the condition number of the covariance matrix. The presented applications include direction-sensitive eigenspace perturbation bounds, relative bounds for the smallest eigenvalue, and the estimation of the single-index model. For the latter, a new estimator, derived from the analysis, with strong theoretical guarantees and superior numerical performance is proposed."
                },
                {
                    "title": "Covariance Matrix Estimation under Total Positivity for Portfolio Selection*",
                    "abstract": "\n Selecting the optimal Markowitz portfolio depends on estimating the covariance matrix of the returns of N assets from T periods of historical data. Problematically, N is typically of the same order as T, which makes the sample covariance matrix estimator perform poorly, both empirically and theoretically. While various other general-purpose covariance matrix estimators have been introduced in the financial economics and statistics literature for dealing with the high dimensionality of this problem, we here propose an estimator that exploits the fact that assets are typically positively dependent. This is achieved by imposing that the joint distribution of returns be multivariate totally positive of order 2 (MTP2). This constraint on the covariance matrix not only enforces positive dependence among the assets but also regularizes the covariance matrix, leading to desirable statistical properties such as sparsity. Based on stock market data spanning 30 years, we show that estimating the covariance matrix under MTP2 outperforms previous state-of-the-art methods including shrinkage estimators and factor models."
                },
                {
                    "title": "Why Gradient Clipping Accelerates Training: A Theoretical Justification for Adaptivity",
                    "abstract": "We provide a theoretical explanation for the effectiveness of gradient clipping in training deep neural networks. The key ingredient is a new smoothness condition derived from practical neural network training examples. We observe that gradient smoothness, a concept central to the analysis of first-order optimization algorithms that is often assumed to be a constant, demonstrates significant variability along the training trajectory of deep neural networks. Further, this smoothness positively correlates with the gradient norm, and contrary to standard assumptions in the literature, it can grow with the norm of the gradient. These empirical observations limit the applicability of existing theoretical analyses of algorithms that rely on a fixed bound on smoothness. These observations motivate us to introduce a novel relaxation of gradient smoothness that is weaker than the commonly used Lipschitz smoothness assumption. Under the new condition, we prove that two popular methods, namely, \\emph{gradient clipping} and \\emph{normalized gradient}, converge arbitrarily faster than gradient descent with fixed stepsize. We further explain why such adaptively scaled gradient methods can accelerate empirical convergence and verify our results empirically in popular neural network training settings."
                },
                {
                    "title": "Exact Gaussian Processes on a Million Data Points",
                    "abstract": "Gaussian processes (GPs) are flexible non-parametric models, with a capacity that grows with the available data. However, computational constraints with standard inference procedures have limited exact GPs to problems with fewer than about ten thousand training points, necessitating approximations for larger datasets. In this paper, we develop a scalable approach for exact GPs that leverages multi-GPU parallelization and methods like linear conjugate gradients, accessing the kernel matrix only through matrix multiplication. By partitioning and distributing kernel matrix multiplies, we demonstrate that an exact GP can be trained on over a million points, a task previously thought to be impossible with current computing hardware, in less than 2 hours. Moreover, our approach is generally applicable, without constraints to grid data or specific kernel classes. Enabled by this scalability, we perform the first-ever comparison of exact GPs against scalable GP approximations on datasets with $10^4 \\!-\\! 10^6$ data points, showing dramatic performance improvements."
                },
                {
                    "title": "Orthogonal Statistical Learning",
                    "abstract": "We provide excess risk guarantees for statistical learning in a setting where the population risk with respect to which we evaluate the target model depends on an unknown model that must be to be estimated from data (a \"nuisance model\"). We analyze a two-stage sample splitting meta-algorithm that takes as input two arbitrary estimation algorithms: one for the target model and one for the nuisance model. We show that if the population risk satisfies a condition called Neyman orthogonality, the impact of the nuisance estimation error on the excess risk bound achieved by the meta-algorithm is of second order. Our theorem is agnostic to the particular algorithms used for the target and nuisance and only makes an assumption on their individual performance. This enables the use of a plethora of existing results from statistical learning and machine learning literature to give new guarantees for learning with a nuisance component. Moreover, by focusing on excess risk rather than parameter estimation, we can give guarantees under weaker assumptions than in previous works and accommodate the case where the target parameter belongs to a complex nonparametric class. We characterize conditions on the metric entropy such that oracle rates---rates of the same order as if we knew the nuisance model---are achieved. We also analyze the rates achieved by specific estimation algorithms such as variance-penalized empirical risk minimization, neural network estimation and sparse high-dimensional linear model estimation. We highlight the applicability of our results in four settings of central importance in the literature: 1) heterogeneous treatment effect estimation, 2) offline policy optimization, 3) domain adaptation, and 4) learning with missing data."
                },
                {
                    "title": "Pyro: Deep Universal Probabilistic Programming",
                    "abstract": "Pyro is a probabilistic programming language built on Python as a platform for developing advanced probabilistic models in AI research. To scale to large datasets and high-dimensional models, Pyro uses stochastic variational inference algorithms and probability distributions built on top of PyTorch, a modern GPU-accelerated deep learning framework. To accommodate complex or model-specific algorithmic behavior, Pyro leverages Poutine, a library of composable building blocks for modifying the behavior of probabilistic programs."
                },
                {
                    "title": "Automatic Debiased Machine Learning of Causal and Structural Effects",
                    "abstract": "Many causal and structural effects depend on regressions. Examples include policy effects, average derivatives, regression decompositions, average treatment effects, causal mediation, and parameters of economic structural models. The regressions may be high\u2010dimensional, making machine learning useful. Plugging machine learners into identifying equations can lead to poor inference due to bias from regularization and/or model selection. This paper gives automatic debiasing for linear and nonlinear functions of regressions. The debiasing is automatic in using Lasso and the function of interest without the full form of the bias correction. The debiasing can be applied to any regression learner, including neural nets, random forests, Lasso, boosting, and other high\u2010dimensional methods. In addition to providing the bias correction, we give standard errors that are robust to misspecification, convergence rates for the bias correction, and primitive conditions for asymptotic inference for estimators of a variety of estimators of structural and causal effects. The automatic debiased machine learning is used to estimate the average treatment effect on the treated for the NSW job training data and to estimate demand elasticities from Nielsen scanner data while allowing preferences to be correlated with prices and income."
                },
                {
                    "title": "A Swiss Army Infinitesimal Jackknife",
                    "abstract": "The error or variability of machine learning algorithms is often assessed by\nrepeatedly refitting a model with different weighted versions of the observed\ndata. The ubiquitous tools of cross-validation (CV) and the bootstrap are\nexamples of this technique. These methods are powerful in large part due to\ntheir model agnosticism but can be slow to run on modern, large data sets due to\nthe need to repeatedly re-fit the model. In this work, we use a linear\napproximation to the dependence of the fitting procedure on the weights,\nproducing results that can be faster than repeated re-fitting by an order of\nmagnitude. This linear approximation is sometimes known as the \u201cinfinitesimal\njackknife\u201d in the statistics literature, where it is mostly used as a\ntheoretical tool to prove asymptotic results. We provide explicit finite-sample\nerror bounds for the infinitesimal jackknife in terms of a small number of\nsimple, verifiable assumptions. Our results apply whether the weights and data\nare stochastic or deterministic, and so can be used as a tool for proving the\naccuracy of the infinitesimal jackknife on a wide variety of problems. As a\ncorollary, we state mild regularity conditions under which our approximation\nconsistently estimates true leave k-out cross-validation for any fixed k. These\ntheoretical results, together with modern automatic differentiation software,\nsupport the application of the infinitesimal jackknife to a wide variety of\npractical problems in machine learning, providing a \u201cSwiss Army infinitesimal\njackknife.\u201d We demonstrate the accuracy of our methods on a range of simulated\nand real datasets."
                },
                {
                    "title": "Automatic differentiation in PyTorch",
                    "abstract": "In this article, we describe an automatic differentiation module of PyTorch \u2014 a library designed to enable rapid research on machine learning models. It builds upon a few projects, most notably Lua Torch, Chainer, and HIPS Autograd [4], and provides a high performance environment with easy access to automatic differentiation of models executed on different devices (CPU and GPU). To make prototyping easier, PyTorch does not follow the symbolic approach used in many other deep learning frameworks, but focuses on differentiation of purely imperative programs, with a focus on extensibility and low overhead. Note that this preprint is a draft of certain sections from an upcoming paper covering all PyTorch features."
                },
                {
                    "title": "On Nesting Monte Carlo Estimators",
                    "abstract": "Many problems in machine learning and statistics involve nested expectations and thus do not permit conventional Monte Carlo (MC) estimation. For such problems, one must nest estimators, such that terms in an outer estimator themselves involve calculation of a separate, nested, estimation. We investigate the statistical implications of nesting MC estimators, including cases of multiple levels of nesting, and establish the conditions under which they converge. We derive corresponding rates of convergence and provide empirical evidence that these rates are observed in practice. We further establish a number of pitfalls that can arise from naive nesting of MC estimators, provide guidelines about how these can be avoided, and lay out novel methods for reformulating certain classes of nested expectation problems into single expectations, leading to improved convergence rates. We demonstrate the applicability of our work by using our results to develop a new estimator for discrete Bayesian experimental design problems and derive error bounds for a class of variational objectives."
                },
                {
                    "title": "Double/Debiased Machine Learning for Treatment and Structural Parameters",
                    "abstract": "We revisit the classic semiparametric problem of inference on a low dimensional parameter \u03b8_0 in the presence of high-dimensional nuisance parameters \u03b7_0. We depart from the classical setting by allowing for \u03b7_0 to be so high-dimensional that the traditional assumptions, such as Donsker properties, that limit complexity of the parameter space for this object break down. To estimate \u03b7_0, we consider the use of statistical or machine learning (ML) methods which are particularly well-suited to estimation in modern, very high-dimensional cases. ML methods perform well by employing regularization to reduce variance and trading off regularization bias with overfitting in practice. However, both regularization bias and overfitting in estimating \u03b7_0 cause a heavy bias in estimators of \u03b8_0 that are obtained by naively plugging ML estimators of \u03b7_0 into estimating equations for \u03b8_0. This bias results in the naive estimator failing to be N^(-1/2) consistent, where N is the sample size. We show that the impact of regularization bias and overfitting on estimation of the parameter of interest \u03b8_0 can be removed by using two simple, yet critical, ingredients: (1) using Neyman-orthogonal moments/scores that have reduced sensitivity with respect to nuisance parameters to estimate \u03b8_0, and (2) making use of cross-fitting which provides an efficient form of data-splitting. We call the resulting set of methods double or debiased ML (DML). We verify that DML delivers point estimators that concentrate in a N^(-1/2)-neighborhood of the true parameter values and are approximately unbiased and normally distributed, which allows construction of valid confidence statements. The generic statistical theory of DML is elementary and simultaneously relies on only weak theoretical requirements which will admit the use of a broad array of modern ML methods for estimating the nuisance parameters such as random forests, lasso, ridge, deep neural nets, boosted trees, and various hybrids and ensembles of these methods. We illustrate the general theory by applying it to provide theoretical properties of DML applied to learn the main regression parameter in a partially linear regression model, DML applied to learn the coefficient on an endogenous variable in a partially linear instrumental variables model, DML applied to learn the average treatment effect and the average treatment effect on the treated under unconfoundedness, and DML applied to learn the local average treatment effect in an instrumental variables setting. In addition to these theoretical applications, we also illustrate the use of DML in three empirical examples."
                },
                {
                    "title": "Understanding Black-box Predictions via Influence Functions",
                    "abstract": "How can we explain the predictions of a black-box model? In this paper, we use influence functions \u2014 a classic technique from robust statistics \u2014 to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. We show that even on non-convex and non-differentiable models where the theory breaks down, approximations to influence functions can still provide valuable information. On linear models and convolutional neural networks, we demonstrate that influence functions are useful for multiple purposes: understanding model behavior, debugging models, detecting dataset errors, and even creating visually-indistinguishable training-set attacks."
                },
                {
                    "title": "Toward Computerized Efficient Estimation in Infinite-Dimensional Models",
                    "abstract": "ABSTRACT Despite the risk of misspecification they are tied to, parametric models continue to be used in statistical practice because they are simple and convenient to use. In particular, efficient estimation procedures in parametric models are easy to describe and implement. Unfortunately, the same cannot be said of semiparametric and nonparametric models. While the latter often reflect the level of available scientific knowledge more appropriately, performing efficient inference in these models is generally challenging. The efficient influence function is a key analytic object from which the construction of asymptotically efficient estimators can potentially be streamlined. However, the theoretical derivation of the efficient influence function requires specialized knowledge and is often a difficult task, even for experts. In this article, we present a novel representation of the efficient influence function and describe a numerical procedure for approximating its evaluation. The approach generalizes the nonparametric procedures of Frangakis et al. and Luedtke, Carone, and van der Laan to arbitrary models. We present theoretical results to support our proposal and illustrate the method in the context of several semiparametric problems. The proposed approach is an important step toward automating efficient estimation in general statistical models, thereby rendering more accessible the use of realistic models in statistical analyses. Supplementary materials for this article are available online."
                },
                {
                    "title": "Nonparametric von Mises Estimators for Entropies, Divergences and Mutual Informations",
                    "abstract": "We propose and analyse estimators for statistical functionals of one or more distributions under nonparametric assumptions. Our estimators are derived from the von Mises expansion and are based on the theory of influence functions, which appear in the semiparametric statistics literature. We show that estimators based either on data-splitting or a leave-one-out technique enjoy fast rates of convergence and other favorable theoretical properties. We apply this framework to derive estimators for several popular information theoretic quantities, and via empirical evaluation, show the advantage of this approach over existing estimators."
                },
                {
                    "title": "Deductive derivation and turing\u2010computerization of semiparametric efficient estimation",
                    "abstract": "Researchers often seek robust inference for a parameter through semiparametric estimation. Efficient semiparametric estimation currently requires theoretical derivation of the efficient influence function (EIF), which can be a challenging and time\u2010consuming task. If this task can be computerized, it can save dramatic human effort, which can be transferred, for example, to the design of new studies. Although the EIF is, in principle, a derivative, simple numerical differentiation to calculate the EIF by a computer masks the EIF's functional dependence on the parameter of interest. For this reason, the standard approach to obtaining the EIF relies on the theoretical construction of the space of scores under all possible parametric submodels. This process currently depends on the correctness of conjectures about these spaces, and the correct verification of such conjectures. The correct guessing of such conjectures, though successful in some problems, is a nondeductive process, i.e., is not guaranteed to succeed (e.g., is not computerizable), and the verification of conjectures is generally susceptible to mistakes. We propose a method that can deductively produce semiparametric locally efficient estimators. The proposed method is computerizable, meaning that it does not need either conjecturing, or otherwise theoretically deriving the functional form of the EIF, and is guaranteed to produce the desired estimates even for complex parameters. The method is demonstrated through an example."
                },
                {
                    "title": "The influence function of semiparametric estimators",
                    "abstract": "There are many economic parameters that depend on nonparametric first steps. Examples include games, dynamic discrete choice, average exact consumer surplus, and treatment effects. Often estimators of these parameters are asymptotically equivalent to a sample average of an object referred to as the influence function. The influence function is useful in local policy analysis, in evaluating local sensitivity of estimators, and constructing debiased machine learning estimators. We show that the influence function is a Gateaux derivative with respect to a smooth deviation evaluated at a point mass. This result generalizes the classic Von Mises (1947) and Hampel (1974) calculation to estimators that depend on smooth nonparametric first steps. We give explicit influence functions for first steps that satisfy exogenous or endogenous orthogonality conditions. We use these results to generalize the omitted variable bias formula for regression to policy analysis for and sensitivity to structural changes. We apply this analysis and find no sensitivity to endogeneity of average equivalent variation estimates in a gasoline demand application."
                },
                {
                    "title": "Gradient Estimation Using Stochastic Computation Graphs",
                    "abstract": "In a variety of problems originating in supervised, unsupervised, and reinforcement learning, the loss function is defined by an expectation over a collection of random variables, which might be part of a probabilistic model or the external world. Estimating the gradient of this loss function, using samples, lies at the core of gradient-based learning algorithms for these problems. We introduce the formalism of stochastic computation graphs\u2014directed acyclic graphs that include both deterministic functions and conditional probability distributions\u2014and describe how to easily and automatically derive an unbiased estimator of the loss function's gradient. The resulting algorithm for computing the gradient estimator is a simple modification of the standard backpropagation algorithm. The generic scheme we propose unifies estimators derived in variety of prior work, along with variance-reduction techniques therein. It could assist researchers in developing intricate models involving a combination of stochastic and deterministic operations, enabling, for example, attention, memory, and control actions."
                },
                {
                    "title": "Statistical Learning with Sparsity: The Lasso and Generalizations",
                    "abstract": "Discover New Methods for Dealing with High-Dimensional Data A sparse statistical model has only a small number of nonzero parameters or weights; therefore, it is much easier to estimate and interpret than a dense model. Statistical Learning with Sparsity: The Lasso and Generalizations presents methods that exploit sparsity to help recover the underlying signal in a set of data. Top experts in this rapidly evolving field, the authors describe the lasso for linear regression and a simple coordinate descent algorithm for its computation. They discuss the application of 1 penalties to generalized linear models and support vector machines, cover generalized penalties such as the elastic net and group lasso, and review numerical methods for optimization. They also present statistical inference methods for fitted (lasso) models, including the bootstrap, Bayesian methods, and recently developed approaches. In addition, the book examines matrix decomposition, sparse multivariate analysis, graphical models, and compressed sensing. It concludes with a survey of theoretical results for the lasso. In this age of big data, the number of features measured on a person or object can be large and might be larger than the number of observations. This book shows how the sparsity assumption allows us to tackle these problems and extract useful and reproducible patterns from big datasets. Data analysts, computer scientists, and theorists will appreciate this thorough and up-to-date treatment of sparse statistical modeling."
                },
                {
                    "title": "Automatic differentiation in machine learning: a survey",
                    "abstract": "Derivatives, mostly in the form of gradients and Hessians, are ubiquitous in machine learning. Automatic differentiation (AD), also called algorithmic differentiation or simply \u201cauto-diff\u201d, is a family of techniques similar to but more general than backpropagation for efficiently and accurately evaluating derivatives of numeric functions expressed as computer programs. AD is a small but established field with applications in areas including computational fluid dynamics, atmospheric sciences, and engineering design optimization. Until \nvery recently, the fields of machine learning and AD have largely been unaware of each other and, in some cases, have independently discovered each other\u2019s results. Despite its \nrelevance, general-purpose AD has been missing from the machine learning toolbox, a situation slowly changing with its ongoing adoption under the names \u201cdynamic computational \ngraphs\u201d and \u201cdifferentiable programming\u201d. We survey the intersection of AD and machine learning, cover applications where AD has direct relevance, and address the main imple- \nmentation techniques. By precisely defining the main differentiation techniques and their interrelationships, we aim to bring clarity to the usage of the terms \u201cautodiff\u201d, \u201cautomatic differentiation\u201d, and \u201csymbolic differentiation\u201d as these are encountered more and more in machine learning settings."
                },
                {
                    "title": "Higher Order Tangent Spaces and Influence Functions",
                    "abstract": "We review higher order tangent spaces and influence functions\nand their use to construct minimax efficient estimators for parameters in highdimensional\nsemiparametric models."
                },
                {
                    "title": "On robust regression with high-dimensional predictors",
                    "abstract": "We study regression M-estimates in the setting where p, the number of covariates, and n, the number of observations, are both large, but . We find an exact stochastic representation for the distribution of at fixed p and n under various assumptions on the objective function \u03c1 and our statistical model. A scalar random variable whose deterministic limit can be studied when plays a central role in this representation. We discover a nonlinear system of two deterministic equations that characterizes . Interestingly, the system shows that depends on \u03c1 through proximal mappings of \u03c1 as well as various aspects of the statistical model underlying our study. Several surprising results emerge. In particular, we show that, when is large enough, least squares becomes preferable to least absolute deviations for double-exponential errors."
                },
                {
                    "title": "Higher order influence functions and minimax estimation of nonlinear functionals",
                    "abstract": "We present a theory of point and interval estimation for nonlinear functionals in parametric, semi-, and non-parametric models based on higher order influence functions (Robins (2004), Section 9; Li et al. (2004), Tchetgen et al. (2006), Robins et al. (2007)). Higher order influence functions are higher order U-statistics. Our theory extends the first order semiparametric theory of Bickel et al. (1993) and van der Vaart (1991) by incorporating the theory of higher order scores considered by Pfanzagl (1990), Small and McLeish (1994) and Lindsay and Waterman (1996). The theory reproduces many previous results, produces new non-$\\sqrt{n}$ results, and opens up the ability to perform optimal non-$\\sqrt{n}$ inference in complex high dimensional models. We present novel rate-optimal point and interval estimators for various functionals of central importance to biostatistics in settings in which estimation at the expected $\\sqrt{n}$ rate is not possible, owing to the curse of dimensionality. We also show that our higher order influence functions have a multi-robustness property that extends the double robustness property of first order influence functions described by Robins and Rotnitzky (2001) and van der Laan and Robins (2003)."
                },
                {
                    "title": "A New Algorithm for Estimating the Effective Dimension-Reduction Subspace",
                    "abstract": "The statistical problem of estimating the effective dimension-reduction (EDR) subspace in the multi-index regression model with deterministic design and additive noise is considered. A new procedure for recovering the directions of the EDR subspace is proposed. Many methods for estimating the EDR subspace perform principal component analysis on a family of vectors, say \u03b21,...,\u03b2L, nearly lying in the EDR subspace. This is in particular the case for the structure-adaptive approach proposed by Hristache et al. (2001a). In the present work, we propose to estimate the projector onto the EDR subspace by the solution to the optimization problem \n \nminimize maxl=1,...,L \u03b2lT (I-A)\u03b2l\u00a0\u00a0\u00a0 subject to A \u2208 Am \n \nwhere Am is the set of all symmetric matrices with eigenvalues in [0,1] and trace less than or equal to m, with m being the true structural dimension. Under mild assumptions, \u221an-consistency of the proposed procedure is proved (up to a logarithmic factor) in the case when the structural dimension is not larger than 4. Moreover, the stochastic error of the estimator of the projector onto the EDR subspace is shown to depend on L logarithmically. This enables us to use a large number of vectors \u03b2l for estimating the EDR subspace. The empirical behavior of the algorithm is studied through numerical simulations."
                },
                {
                    "title": "Risk Reduction in Large Portfolios: Why Imposing the Wrong Constraints Helps",
                    "abstract": "Mean-variance efficient portfolios constructed using sample moments often involve taking extreme long and short positions. Hence practitioners often impose portfolio weight constraints when constructing efficient portfolios. Green and Hollifield (1992) argue that the presence of a single dominant factor in the covariance matrix of returns is why we observe extreme positive and negative weights. If this were the case then imposing the weight constraint should hurt whereas the empirical evidence is often to the contrary. We reconcile this apparent contradiction. We show that constraining portfolio weights to be nonnegative is equivalent to using the sample covariance matrix after reducing its large elements and then form the optimal portfolio without any restrictions on portfolio weights. This shrinkage helps reduce the risk in estimated optimal portfolios even when they have negative weights in the population. Surprisingly, we also find that once the nonnegativity constraint is imposed, minimum variance portfolios constructed using the monthly sample covariance matrix perform as well as those constructed using covariance matrices estimated using factor models, shrinkage estimators, and daily data. When minimizing tracking error is the criterion, using daily data instead of monthly data helps. However, the sample covariance matrix without any correction for microstructure effects performs the best."
                },
                {
                    "title": "Fast Exact Multiplication by the Hessian",
                    "abstract": "Just storing the Hessian H (the matrix of second derivatives 2E/wiwj of the error E with respect to each pair of weights) of a large neural network is difficult. Since a common use of a large matrix like H is to compute its product with various vectors, we derive a technique that directly calculates Hv, where v is an arbitrary vector. To calculate Hv, we first define a differential operator Rv{f(w)} = (/r)f(w rv)|r=0, note that Rv{w} = Hv and Rv{w} = v, and then apply Rv{} to the equations used to compute w. The result is an exact and numerically stable procedure for computing Hv, which takes about as much computation, and is about as local, as a gradient evaluation. We then apply the technique to a one pass gradient calculation algorithm (backpropagation), a relaxation gradient calculation algorithm (recurrent backpropagation), and two stochastic gradient calculation algorithms (Boltzmann machines and weight perturbation). Finally, we show that this technique can be used at the heart of many iterative techniques for computing various properties of H, obviating any need to calculate the full Hessian."
                },
                {
                    "title": "The efficient market inefficiency of capitalization\u2013weighted stock portfolios",
                    "abstract": "35 A s of the end of 1990, the largest 200 prehensive cap-weighted portfolios occupy positions defined-benefit pension plans had indexed a combined total of approximately $200 billion to capitalization-weighted stock portfolios, such as the Wilshire 5000. The sponsors of these plans presumably return. E: believe this to be an efficient investment, in the inside the efficient set. In this context, theory directs us to manage risk in seeking out investment portfolios with the lowest possible volatility given expected 6 8 5"
                },
                {
                    "title": "Portfolio Selection",
                    "abstract": "Embracing finance, economics, operations research, and computers, this book applies modern techniques of analysis and computation to find combinations of securities that best meet the needs of private or institutional investors."
                },
                {
                    "title": "Kernel Debiased Plug-in Estimation",
                    "abstract": ","
                },
                {
                    "title": "Empirical Gateaux Derivatives for Causal Inference",
                    "abstract": "We study a constructive algorithm that approximates Gateaux derivatives for statistical functionals by \ufb01nite differencing, with a focus on functionals that arise in causal inference. We study the setting where probability distributions are not known a priori but need to be estimated from data. These estimated distributions lead to empirical Gateaux derivatives, and we study the relationships between empirical, numerical, and analytical Gateaux derivatives. Starting with a case study of the interventional mean (average potential outcome), we delineate the relationship between the empirical Gateaux derivative (via \ufb01nite differencing) and the analytical Gateaux derivative. We then derive requirements on the rates of numerical approximation in perturbation and smoothing that preserve the statistical bene\ufb01ts of one-step adjustments, such as rate double robustness. We further study more complicated functionals such as dynamic treatment regimes and the linear-programming formulation for policy optimization in in\ufb01nite-horizon Markov decision processes. The ability to approximate bias adjustments in the presence of arbitrary constraints in these more complicated settings illustrates the usefulness of constructive approaches for Gateaux derivatives. We also \ufb01nd that the statistical structure of the functional (rate double robustness) can permit less conservative rates for \ufb01nite-difference approximation. This property, however, can be speci\ufb01c to particular functionals; e.g., it occurs for the average potential outcome (hence average treatment effect) but not the in\ufb01nite-horizon MDP policy value."
                },
                {
                    "title": "Automatic Debiased Machine Learning via Neural Nets for Generalized Linear Regression",
                    "abstract": "We give debiased machine learners of parameters of interest that depend on generalized linear regressions, which regressions make a residual orthogonal to regressors. The parameters of interest include many causal and policy e\ufb00ects. We give neural net learners of the bias correction that are automatic in only depending on the object of interest and the regression residual. Convergence rates are given for these neural nets and for more general learners of the bias correction. We also give conditions for asymptotic normality and consistent asymptotic variance estimation of the learner of the object of interest."
                },
                {
                    "title": "Double Machine Learning Density Estimation for Local Treatment Effects with Instruments",
                    "abstract": "It is common to quantify causal effects with mean values, which, however, may 1 fail to capture significant distribution differences of the outcome under different 2 treatments. We study the problem of estimating the density of the causal effect 3 of a binary treatment on a continuous outcome given a binary instrumental vari4 able in the presence of covariates. Specifically, we consider the local treatment 5 effect, which measures the effect of treatment among those who comply with the 6 assignment under the assumption of monotonicity (only the ones who were offered 7 the treatment take it). We develop two families of methods for this task, kernel8 smoothing and model-based approximations \u2013 the former smoothes the density by 9 convoluting with a smooth kernel function; the latter projects the density onto a 10 finite-dimensional density class. For both approaches, we derive double/debiased 11 machine learning (DML) based estimators. We study the asymptotic convergence 12 rates of the estimators and show that they are robust to the biases in nuisance 13 function estimation. We illustrate the proposed methods on synthetic data and a 14 real dataset called 401(k). 15"
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                },
                {
                    "title": "On the Mathematical Foundations of Theoretical Statistics",
                    "abstract": "Centre of Location. That abscissa of a frequency curve for which the sampling errors of optimum location are uncorrelated with those of optimum scaling. (9.)"
                }
            ],
            "categories": [
                "stat.CO",
                "cs.LG",
                "stat.ME"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we automate the computation of Efficient Influence Functions (EIFs) in high-dimensional statistical estimation to improve the efficiency and accessibility of causal inference methods?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it bridges the gap between traditional statistical methods and modern machine learning practices. By automating the computation of EIFs, we can enhance the robustness and efficiency of causal inference, leading to more accurate estimations of treatment effects. This advancement could significantly influence future research by enabling researchers to apply sophisticated statistical techniques without the burden of complex manual derivations, thus fostering innovation in various applications such as healthcare, economics, and social sciences.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to accurately recover EIFs, which involves solving high-dimensional integral equations that are computationally intensive and complex. Naive approaches may fail because they do not account for the intricate relationships between nuisance parameters and the quantities of interest, leading to inaccurate estimations. Additionally, the lack of existing automated frameworks for EIFs complicates the process, as researchers often rely on manual derivations that are prone to errors and inconsistencies.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has focused on specific combinations of models and functionals, resulting in a fragmented understanding of EIFs that lacks generality. The barriers to solving this problem include the complexity of the mathematical analysis required for different scenarios and the absence of a unified framework that can automate the computation of EIFs. Our approach differs by introducing the Monte Carlo Efficient Influence Functions (MC-EIF) method, which leverages existing automatic differentiation and probabilistic programming systems to provide a general and automated solution, thus overcoming the limitations of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the use of Monte Carlo Efficient Influence Functions (MC-EIF) to compute EIFs. We will utilize existing automatic differentiation and probabilistic programming systems to derive the necessary quantities, focusing on the gradient of the functional, the inverse Fisher information matrix, and the gradient of the log-likelihood. The expected outcomes include accurate estimates of the true EIF, demonstrating that MC-EIF can achieve the same asymptotic guarantees as analytic EIFs while being applicable to a broad class of models and functionals. We will validate our approach through empirical"
            }
        },
        "author_data": {
            "bd5c6b3b-aa78-4035-9ad0-b89f254e93fc": {
                "pk": "bd5c6b3b-aa78-4035-9ad0-b89f254e93fc",
                "name": "Raj Agrawal",
                "collaborators": [
                    "Tamara Broderick",
                    "Caroline Uhler",
                    "Chandler Squires",
                    "Jonathan H. Huggins",
                    "Neha Prasad",
                    "Uma Roy",
                    "Karren Yang",
                    "Karthik Shanmugam",
                    "Trevor Campbell",
                    "Brian Trippe"
                ],
                "domain": [
                    "Causal Inference",
                    "Bayesian Methods",
                    "High-Dimensional Statistics",
                    "Kernel Methods"
                ],
                "publications": [
                    {
                        "title": "The SKIM-FA Kernel: High-Dimensional Variable Selection and Nonlinear Interaction Discovery in Linear Time",
                        "abstract": "Many scientific problems require identifying a small set of covariates that are associated with a target response and estimating their effects. Often, these effects are nonlinear and include interactions, so linear and additive methods can lead to poor estimation and variable selection. Unfortunately, methods that simultaneously express sparsity, nonlinearity, and interactions are computationally intractable -- with runtime at least quadratic in the number of covariates, and often worse. In the present work, we solve this computational bottleneck. We show that suitable interaction models have a kernel representation, namely there exists a \"kernel trick\" to perform variable selection and estimation in $O$(# covariates) time. Our resulting fit corresponds to a sparse orthogonal decomposition of the regression function in a Hilbert space (i.e., a functional ANOVA decomposition), where interaction effects represent all variation that cannot be explained by lower-order effects. On a variety of synthetic and real data sets, our approach outperforms existing methods used for large, high-dimensional data sets while remaining competitive (or being orders of magnitude faster) in runtime."
                    },
                    {
                        "title": "The DeCAMFounder: Non-Linear Causal Discovery in the Presence of Hidden Variables",
                        "abstract": "Many real-world decision-making tasks require learning causal relationships between a set of variables. Traditional causal discovery methods, however, require that all variables are observed, which is often not feasible in practical scenarios. Without additional assumptions about the unobserved variables, it is not possible to recover any causal relationships from observational data. Fortunately, in many applied settings, additional structure among the confounders can be expected. In particular, pervasive confounding is commonly encountered and has been utilized for consistent causal estimation in linear causal models. In this paper, we present a provably consistent method to estimate causal relationships in the non-linear, pervasive confounding setting. The core of our procedure relies on the ability to estimate the confounding variation through a simple spectral decomposition of the observed data matrix. We derive a DAG score function based on this insight, prove its consistency in recovering a correct ordering of the DAG, and empirically compare it to previous approaches. We demonstrate improved performance on both simulated and real datasets by explicitly accounting for both confounders and non-linear effects."
                    },
                    {
                        "title": "Covariance Matrix Estimation under Total Positivity for Portfolio Selection",
                        "abstract": "Selecting the optimal Markowitz porfolio depends on estimating the covariance matrix of the returns of $N$ assets from $T$ periods of historical data. Problematically, $N$ is typically of the same order as $T$, which makes the sample covariance matrix estimator perform poorly, both empirically and theoretically. While various other general purpose covariance matrix estimators have been introduced in the financial economics and statistics literature for dealing with the high dimensionality of this problem, we here propose an estimator that exploits the fact that assets are typically positively dependent. This is achieved by imposing that the joint distribution of returns be multivariate totally positive of order 2 ($\\text{MTP}_2$). This constraint on the covariance matrix not only enforces positive dependence among the assets, but also regularizes the covariance matrix, leading to desirable statistical properties such as sparsity. Based on stock-market data spanning over thirty years, we show that estimating the covariance matrix under $\\text{MTP}_2$ outperforms previous state-of-the-art methods including shrinkage estimators and factor models."
                    },
                    {
                        "title": "Minimal I-MAP MCMC for Scalable Structure Discovery in Causal DAG Models",
                        "abstract": "Learning a Bayesian network (BN) from data can be useful for decision-making or discovering causal relationships. However, traditional methods often fail in modern applications, which exhibit a larger number of observed variables than data points. The resulting uncertainty about the underlying network as well as the desire to incorporate prior information recommend a Bayesian approach to learning the BN, but the highly combinatorial structure of BNs poses a striking challenge for inference. The current state-of-the-art methods such as order MCMC are faster than previous methods but prevent the use of many natural structural priors and still have running time exponential in the maximum indegree of the true directed acyclic graph (DAG) of the BN. We here propose an alternative posterior approximation based on the observation that, if we incorporate empirical conditional independence tests, we can focus on a high-probability DAG associated with each order of the vertices. We show that our method allows the desired flexibility in prior specification, removes timing dependence on the maximum indegree and yields provably good posterior approximations; in addition, we show that it achieves superior accuracy, scalability, and sampler mixing on several datasets."
                    },
                    {
                        "title": "ABCD-Strategy: Budgeted Experimental Design for Targeted Causal Structure Discovery",
                        "abstract": "Determining the causal structure of a set of variables is critical for both scientific inquiry and decision-making. However, this is often challenging in practice due to limited interventional data. Given that randomized experiments are usually expensive to perform, we propose a general framework and theory based on optimal Bayesian experimental design to select experiments for targeted causal discovery. That is, we assume the experimenter is interested in learning some function of the unknown graph (e.g., all descendants of a target node) subject to design constraints such as limits on the number of samples and rounds of experimentation. While it is in general computationally intractable to select an optimal experimental design strategy, we provide a tractable implementation with provable guarantees on both approximation and optimization quality based on submodularity. We evaluate the efficacy of our proposed method on both synthetic and real datasets, thereby demonstrating that our method realizes considerable performance gains over baseline strategies such as random sampling."
                    },
                    {
                        "title": "Data-dependent compression of random features for large-scale kernel approximation",
                        "abstract": "Kernel methods offer the flexibility to learn complex relationships in modern, large data sets while enjoying strong theoretical guarantees on quality. Unfortunately, these methods typically require cubic running time in the data set size, a prohibitive cost in the large-data setting. Random feature maps (RFMs) and the Nystrom method both consider low-rank approximations to the kernel matrix as a potential solution. But, in order to achieve desirable theoretical guarantees, the former may require a prohibitively large number of features J+, and the latter may be prohibitively expensive for high-dimensional problems. We propose to combine the simplicity and generality of RFMs with a data-dependent feature selection scheme to achieve desirable theoretical approximation properties of Nystrom with just O(log J+) features. Our key insight is to begin with a large set of random features, then reduce them to a small number of weighted features in a data-dependent, computationally efficient way, while preserving the statistical guarantees of using the original large set of features. We demonstrate the efficacy of our method with theory and experiments--including on a data set with over 50 million observations. In particular, we show that our method achieves small kernel matrix approximation error and better test set accuracy with provably fewer random features than state-of-the-art methods."
                    },
                    {
                        "title": "The Kernel Interaction Trick: Fast Bayesian Discovery of Pairwise Interactions in High Dimensions",
                        "abstract": "Discovering interaction effects on a response of interest is a fundamental problem faced in biology, medicine, economics, and many other scientific disciplines. In theory, Bayesian methods for discovering pairwise interactions enjoy many benefits such as coherent uncertainty quantification, the ability to incorporate background knowledge, and desirable shrinkage properties. In practice, however, Bayesian methods are often computationally intractable for even moderate-dimensional problems. Our key insight is that many hierarchical models of practical interest admit a particular Gaussian process (GP) representation; the GP allows us to capture the posterior with a vector of O(p) kernel hyper-parameters rather than O(p^2) interactions and main effects. With the implicit representation, we can run Markov chain Monte Carlo (MCMC) over model hyper-parameters in time and memory linear in p per iteration. We focus on sparsity-inducing models and show on datasets with a variety of covariate behaviors that our method: (1) reduces runtime by orders of magnitude over naive applications of MCMC, (2) provides lower Type I and Type II error relative to state-of-the-art LASSO-based approaches, and (3) offers improved computational scaling in high dimensions relative to existing Bayesian and LASSO-based approaches."
                    },
                    {
                        "title": "Hamiltonian Monte Carlo using an adjoint-differentiated Laplace approximation: Bayesian inference for latent Gaussian models and beyond",
                        "abstract": "Gaussian latent variable models are a key class of Bayesian hierarchical models with applications in many fields. Performing Bayesian inference on such models can be challenging as Markov chain Monte Carlo algorithms struggle with the geometry of the resulting posterior distribution and can be prohibitively slow. An alternative is to use a Laplace approximation to marginalize out the latent Gaussian variables and then integrate out the remaining hyperparameters using dynamic Hamiltonian Monte Carlo, a gradient-based Markov chain Monte Carlo sampler. To implement this scheme efficiently, we derive a novel adjoint method that propagates the minimal information needed to construct the gradient of the approximate marginal likelihood. This strategy yields a scalable differentiation method that is orders of magnitude faster than state of the art differentiation techniques when the hyperparameters are high dimensional. We prototype the method in the probabilistic programming framework Stan and test the utility of the embedded Laplace approximation on several models, including one where the dimension of the hyperparameter is $\\sim$6,000. Depending on the cases, the benefits can include an alleviation of the geometric pathologies that frustrate Hamiltonian Monte Carlo and a dramatic speed-up."
                    }
                ]
            },
            "8b280197-3581-4b7d-953c-e490a829e57a": {
                "pk": "8b280197-3581-4b7d-953c-e490a829e57a",
                "name": "Sam Witty",
                "collaborators": [
                    "David Jensen",
                    "Vikash Mansinghka",
                    "Kenta Takatsu",
                    "Alexander Lew",
                    "Jun Ki Lee",
                    "Emma Tosch",
                    "Akanksha Atrey",
                    "Michael Littman",
                    "Brittany Johnson",
                    "Jesse Bartola"
                ],
                "domain": [
                    "Causal Inference",
                    "Bayesian Methods",
                    "Machine Learning",
                    "Fairness"
                ],
                "publications": [
                    {
                        "title": "SBI: A Simulation-Based Test of Identifiability for Bayesian Causal Inference",
                        "abstract": "A growing family of approaches to causal inference rely on Bayesian formulations of assumptions that go beyond causal graph structure. For example, Bayesian approaches have been developed for analyzing instrumental variable designs, regression discontinuity designs, and within-subjects designs. This paper introduces simulation-based identifiability (SBI), a procedure for testing the identifiability of queries in Bayesian causal inference approaches that are implemented as probabilistic programs. SBI complements analytical approaches to identifiability, leveraging a particle-based optimization scheme on simulated data to determine identifiability for analytically intractable models. We analyze SBI's soundness for a broad class of differentiable, finite-dimensional probabilistic programs with bounded effects. Finally, we provide an implementation of SBI using stochastic gradient descent, and show empirically that it agrees with known identification results on a suite of graph-based and quasi-experimental design benchmarks, including those using Gaussian processes."
                    },
                    {
                        "title": "Causal Inference using Gaussian Processes with Structured Latent Confounders",
                        "abstract": "Latent confounders---unobserved variables that influence both treatment and outcome---can bias estimates of causal effects. In some cases, these confounders are shared across observations, e.g. all students taking a course are influenced by the course's difficulty in addition to any educational interventions they receive individually. This paper shows how to semiparametrically model latent confounders that have this structure and thereby improve estimates of causal effects. The key innovations are a hierarchical Bayesian model, Gaussian processes with structured latent confounders (GP-SLC), and a Monte Carlo inference algorithm for this model based on elliptical slice sampling. GP-SLC provides principled Bayesian uncertainty estimates of individual treatment effect with minimal assumptions about the functional forms relating confounders, covariates, treatment, and outcome. Finally, this paper shows GP-SLC is competitive with or more accurate than widely used causal inference techniques on three benchmark datasets, including the Infant Health and Development Program and a dataset showing the effect of changing temperatures on state-wide energy consumption across New England."
                    },
                    {
                        "title": "Bayesian causal inference via probabilistic program synthesis",
                        "abstract": "Causal inference can be formalized as Bayesian inference that combines a prior distribution over causal models and likelihoods that account for both observations and interventions. We show that it is possible to implement this approach using a sufficiently expressive probabilistic programming language. Priors are represented using probabilistic programs that generate source code in a domain specific language. Interventions are represented using probabilistic programs that edit this source code to modify the original generative process. This approach makes it straightforward to incorporate data from atomic interventions, as well as shift interventions, variance-scaling interventions, and other interventions that modify causal structure. This approach also enables the use of general-purpose inference machinery for probabilistic programs to infer probable causal structures and parameters from data. This abstract describes a prototype of this approach in the Gen probabilistic programming language."
                    },
                    {
                        "title": "Measuring and Characterizing Generalization in Deep Reinforcement Learning",
                        "abstract": "Deep reinforcement-learning methods have achieved remarkable performance on challenging control tasks. Observations of the resulting behavior give the impression that the agent has constructed a generalized representation that supports insightful action decisions. We re-examine what is meant by generalization in RL, and propose several definitions based on an agent's performance in on-policy, off-policy, and unreachable states. We propose a set of practical methods for evaluating agents with these definitions of generalization. We demonstrate these techniques on a common benchmark task for deep RL, and we show that the learned networks make poor decisions for states that differ only slightly from on-policy states, even though those states are not selected adversarially. Taken together, these results call into question the extent to which deep Q-networks learn generalized representations, and suggest that more experimentation and analysis is necessary before claims of representation learning can be supported."
                    },
                    {
                        "title": "Fairkit, Fairkit, on the Wall, Who's the Fairest of Them All? Supporting Data Scientists in Training Fair Models",
                        "abstract": "Modern software relies heavily on data and machine learning, and affects decisions that shape our world. Unfortunately, recent studies have shown that because of biases in data, software systems frequently inject bias into their decisions, from producing better closed caption transcriptions of men's voices than of women's voices to overcharging people of color for financial loans. To address bias in machine learning, data scientists need tools that help them understand the trade-offs between model quality and fairness in their specific data domains. Toward that end, we present fairkit-learn, a toolkit for helping data scientists reason about and understand fairness. Fairkit-learn works with state-of-the-art machine learning tools and uses the same interfaces to ease adoption. It can evaluate thousands of models produced by multiple machine learning algorithms, hyperparameters, and data permutations, and compute and visualize a small Pareto-optimal set of models that describe the optimal trade-offs between fairness and quality. We evaluate fairkit-learn via a user study with 54 students, showing that students using fairkit-learn produce models that provide a better balance between fairness and quality than students using scikit-learn and IBM AI Fairness 360 toolkits. With fairkit-learn, users can select models that are up to 67% more fair and 10% more accurate than the models they are likely to train with scikit-learn."
                    }
                ]
            },
            "77d19cea-5b05-4a43-b2dc-e4d15725479d": {
                "pk": "77d19cea-5b05-4a43-b2dc-e4d15725479d",
                "name": "Andy Zane",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "ee9d6627-da57-4265-88f0-7b1d8b57e4d2": {
                "pk": "ee9d6627-da57-4265-88f0-7b1d8b57e4d2",
                "name": "Eli Bingham",
                "collaborators": [
                    "Fritz Obermeyer",
                    "Martin Jankowiak",
                    "Jonathan P. Chen",
                    "Neeraj Pradhan",
                    "Du Phan",
                    "Theofanis Karaletsos",
                    "Rohit Singh",
                    "Paul Szerlip",
                    "Paul Horsfall",
                    "Noah D. Goodman"
                ],
                "domain": [
                    "Probabilistic Programming",
                    "Inference Algorithms",
                    "Machine Learning",
                    "Tensor Algebra"
                ],
                "publications": [
                    {
                        "title": "Functional Tensors for Probabilistic Programming",
                        "abstract": "It is a significant challenge to design probabilistic programming systems that can accommodate a wide variety of inference strategies within a unified framework. Noting that the versatility of modern automatic differentiation frameworks is based in large part on the unifying concept of tensors, we describe a software abstraction for integration --functional tensors-- that captures many of the benefits of tensors, while also being able to describe continuous probability distributions. Moreover, functional tensors are a natural candidate for generalized variable elimination and parallel-scan filtering algorithms that enable parallel exact inference for a large family of tractable modeling motifs. We demonstrate the versatility of functional tensors by integrating them into the modeling frontend and inference backend of the Pyro programming language. In experiments we show that the resulting framework enables a large variety of inference strategies, including those that mix exact and approximate inference."
                    },
                    {
                        "title": "Pyro: Deep Universal Probabilistic Programming",
                        "abstract": "Pyro is a probabilistic programming language built on Python as a platform for developing advanced probabilistic models in AI research. To scale to large datasets and high-dimensional models, Pyro uses stochastic variational inference algorithms and probability distributions built on top of PyTorch, a modern GPU-accelerated deep learning framework. To accommodate complex or model-specific algorithmic behavior, Pyro leverages Poutine, a library of composable building blocks for modifying the behavior of probabilistic programs."
                    },
                    {
                        "title": "Tensor Variable Elimination for Plated Factor Graphs",
                        "abstract": "A wide class of machine learning algorithms can be reduced to variable elimination on factor graphs. While factor graphs provide a unifying notation for these algorithms, they do not provide a compact way to express repeated structure when compared to plate diagrams for directed graphical models. To exploit efficient tensor algebra in graphs with plates of variables, we generalize undirected factor graphs to plated factor graphs and variable elimination to a tensor variable elimination algorithm that operates directly on plated factor graphs. Moreover, we generalize complexity bounds based on treewidth and characterize the class of plated factor graphs for which inference is tractable. As an application, we integrate tensor variable elimination into the Pyro probabilistic programming language to enable exact inference in discrete latent variable models with repeated structure. We validate our methods with experiments on both directed and undirected graphical models, including applications to polyphonic music modeling, animal movement modeling, and latent sentiment analysis."
                    }
                ]
            }
        }
    },
    "2311.08376": {
        "paper_data": {
            "title": "Ensemble sampling for linear bandits: small ensembles suffice",
            "url": "http://arxiv.org/abs/2311.08376v2",
            "arxiv_id": "2311.08376",
            "authors": [
                "David Janz",
                "Alexander E. Litvak",
                "Csaba Szepesv\u00e1ri"
            ],
            "abstract": "We provide the first useful and rigorous analysis of ensemble sampling for the stochastic linear bandit setting. In particular, we show that, under standard assumptions, for a $d$-dimensional stochastic linear bandit with an interaction horizon $T$, ensemble sampling with an ensemble of size of order $\\smash{d \\log T}$ incurs regret at most of the order $\\smash{(d \\log T)^{5/2} \\sqrt{T}}$. Ours is the first result in any structured setting not to require the size of the ensemble to scale linearly with $T$ -- which defeats the purpose of ensemble sampling -- while obtaining near $\\smash{\\sqrt{T}}$ order regret. Ours is also the first result that allows infinite action sets.",
            "introduction": "   1 Introduction  Ensemble sampling [20] is a family of randomised algorithms for balancing exploration and exploitation within sequential decision-making tasks. These algorithms maintain an ensemble of perturbed models of the value of the available actions and, in each step, select an action that has the highest value according to a model chosen uniformly at random from the ensemble. The ensemble is then incrementally updated the new observations.   Ensemble sampling was introduced as an alternative to Thompson sampling [29] that is tractable whenever incremental model updates are cheap [23, 20]. It is thus particularly popular in deep reinforcement learning, where the models (neural networks) are trained via gradient descent in an incremental fashion. Therein, ensemble sampling features directly in algorithms such as bootstrapped DQN and ensemble+ [23, 22], as part of other methods [8, 6], and as the motivation for methods, including random network distillation [5] and hypermodels for exploration [9]. Ensemble sampling has also been applied to online recommendation systems [21, 14, 33], the behavioural sciences [10] and to marketing [31].   Despite its practicality and simple nature, ensemble sampling has thus far resisted analysis. Indeed, [26,  ], who showed a bound of order T\ud835\udc47\\sqrt{T}square-root start_ARG italic_T end_ARG on its Bayesian regret under the impractical condition that the size of the ensemble scales at least linearly with the horizon T\ud835\udc47Titalic_T, summarise the state-of-the-art in this regard as follows:   \u2018A lot of work has attempted to analyze ensemble sampling, but none of them has been successful.\u2019    Readers familiar with the literature know that the analysis of randomised exploration methods, such as Thompson sampling, or perturbed history exploration [17, 16], relies on showing that at each step there is a constant probability of choosing a model that overestimates the true value [3, 2]. For these methods, the analysis is relatively simple, because the algorithms can be described as first fitting a model to the past data, and then using a fresh source of randomness to perturb the model, which is then used to derive the action to be used. Ensemble sampling does not fit this pattern: here, the distribution of models given the past is controlled only implicitly, making the analysis challenging.   Our contribution is a guarantee that (a symmetrised version of) ensemble sampling, given an ensemble size logarithmic in T\ud835\udc47Titalic_T and linear in the number of features d\ud835\udc51ditalic_d, incurs regret no worse than order (d\u2062log\u2061T)5/2\u2062Tsuperscript\ud835\udc51\ud835\udc4752\ud835\udc47\\smash{(d\\log T)^{5/2}\\sqrt{T}}( italic_d roman_log italic_T ) start_POSTSUPERSCRIPT 5 / 2 end_POSTSUPERSCRIPT square-root start_ARG italic_T end_ARG. This is the first successful analysis of ensemble sampling in the stochastic linear bandit setting, or indeed, any structured setting (see Remarks\u00a06 and\u00a07 for a discussion of this claim). Our result is based on (a slight extension of) the now-standard framework of [2,  ], and as such it ought to be possible to extend it to the usual settings: generalised linear bandits [12], kernelised bandits [28], deep learning [15, via the neural tangent kernel, per] and reinforcement learning [32, following the work of].     2 Linear ensemble sampling  We now outline our problem setting, a version of the linear ensemble sampling algorithm and our main result, and discuss these in the",
            "references": [
                {
                    "title": "Exploration via linearly perturbed loss minimisation",
                    "abstract": "We introduce exploration via linear loss perturbations (EVILL), a randomised exploration method for structured stochastic bandit problems that works by solving for the minimiser of a linearly perturbed regularised negative log-likelihood function. We show that, for the case of generalised linear bandits, EVILL reduces to perturbed history exploration (PHE), a method where exploration is done by training on randomly perturbed rewards. In doing so, we provide a simple and clean explanation of when and why random reward perturbations give rise to good bandit algorithms. We propose data-dependent perturbations not present in previous PHE-type methods that allow EVILL to match the performance of Thompson-sampling-style parameter-perturbation methods, both in theory and in practice. Moreover, we show an example outside generalised linear bandits where PHE leads to inconsistent estimates, and thus linear regret, while EVILL remains performant. Like PHE, EVILL can be implemented in just a few lines of code."
                },
                {
                    "title": "An Analysis of Ensemble Sampling",
                    "abstract": "Ensemble sampling serves as a practical approximation to Thompson sampling when maintaining an exact posterior distribution over model parameters is computationally intractable. In this paper, we establish a regret bound that ensures desirable behavior when ensemble sampling is applied to the linear bandit problem. This represents the first rigorous regret analysis of ensemble sampling and is made possible by leveraging information-theoretic concepts and novel analytic techniques that may prove useful beyond the scope of this paper."
                },
                {
                    "title": "Deep Exploration for Recommendation Systems",
                    "abstract": "Modern recommendation systems ought to benefit by probing for and learning from delayed feedback. Research has tended to focus on learning from a user\u2019s response to a single recommendation. Such work, which leverages methods of supervised and bandit learning, forgoes learning from the user\u2019s subsequent behavior. Where past work has aimed to learn from subsequent behavior, there has been a lack of effective methods for probing to elicit informative delayed feedback. Effective exploration through probing for delayed feedback becomes particularly challenging when rewards are sparse. To address this, we develop deep exploration methods for recommendation systems. In particular, we formulate recommendation as a sequential decision problem and demonstrate benefits of deep exploration over single-step exploration. Our experiments are carried out with high-fidelity industrial-grade simulators and establish large improvements over existing algorithms."
                },
                {
                    "title": "Bernstein-type bounds for beta distribution",
                    "abstract": "This work obtains sharp closed-form exponential concentration inequalities of Bernstein type for the ubiquitous beta distribution, improving upon sub-Gaussian and sub-gamma bounds previously studied in this context.\nThe proof leverages a novel handy recursion of order 2 for central moments of the beta distribution, obtained from the hypergeometric representations of moments; this recursion is useful for obtaining explicit expressions for central moments and various tail approximations."
                },
                {
                    "title": "Targeting for long-term outcomes",
                    "abstract": "Decision makers often want to target interventions so as to maximize an outcome that is observed only in the long term. This typically requires delaying decisions until the outcome is observed or relying on simple short-term proxies for the long-term outcome. Here, we build on the statistical surrogacy and policy learning literatures to impute the missing long-term outcomes and then approximate the optimal targeting policy on the imputed outcomes via a doubly robust approach. We first show that conditions for the validity of average treatment effect estimation with imputed outcomes are also sufficient for valid policy evaluation and optimization; furthermore, these conditions can be somewhat relaxed for policy optimization. We apply our approach in two large-scale proactive churn management experiments at The Boston Globe by targeting optimal discounts to its digital subscribers with the aim of maximizing long-term revenue. Using the first experiment, we evaluate this approach empirically by comparing the policy learned using imputed outcomes with a policy learned on the ground-truth, long-term outcomes. The performance of these two policies is statistically indistinguishable, and we rule out large losses from relying on surrogates. Our approach also outperforms a policy learned on short-term proxies for the long-term outcome. In a second field experiment, we implement the optimal targeting policy with additional randomized exploration, which allows us to update the optimal policy for future subscribers. Over three years, our approach had a net-positive revenue impact in the range of $4\u2013$5 million compared with the status quo. This paper was accepted by Eric Anderson, marketing. Funding: This work was supported by Boston Globe Media. Supplemental Material: The online appendix and data are available at https://doi.org/10.1287/mnsc.2023.4881 ."
                },
                {
                    "title": "High-Dimensional Probability: An Introduction with Applications in Data Science",
                    "abstract": "\u00a9 2018, Cambridge University Press Let us summarize our findings. A random projection of a set T in R n onto an m-dimensional subspace approximately preserves the geometry of T if m \u2a86 d ( T ) . For..."
                },
                {
                    "title": "Low-rank Tensor Bandits",
                    "abstract": "In recent years, multi-dimensional online decision making has been playing a crucial role in many practical applications such as online recommendation and digital marketing. To solve it, we introduce stochastic low-rank tensor bandits, a class of bandits whose mean rewards can be represented as a low-rank tensor. We propose two learning algorithms, tensor epoch-greedy and tensor elimination, and develop finite-time regret bounds for them. We observe that tensor elimination has an optimal dependency on the time horizon, while tensor epoch-greedy has a sharper dependency on tensor dimensions. Numerical experiments further back up these theoretical findings and show that our algorithms outperform various state-of-the-art approaches that ignore the tensor low-rank structure."
                },
                {
                    "title": "Bandit Algorithms",
                    "abstract": "sets of environments and policies respectively and ` : E \u00d7\u03a0\u2192 [0, 1] a bounded loss function. Given a policy \u03c0 let `(\u03c0) = (`(\u03bd1, \u03c0), . . . , `(\u03bdN , \u03c0)) be the loss vector resulting from policy \u03c0. Define S = {`(\u03c0) : \u03c0 \u2208 \u03a0} and \u03bb(S) = {x \u2208 cl(S) : y 6< x for all y \u2208 S} , where y 6< x is defined to mean it is not true that yi \u2264 xi for all i with strict inequality for at least one i. Prove that if \u03bb(S) \u2286 S and S is convex, then for each x \u2208 \u03bb(S) there exists a prior q \u2208 P(E) and policy \u03c0\u2217 such that `(\u03c0) = x and \u2211 \u03bd\u2208E q(\u03bd)`(\u03bd, \u03c0\u2217) = min \u03c0\u2208\u03a0 \u2211"
                },
                {
                    "title": "Efficient Model-Based Reinforcement Learning through Optimistic Policy Search and Planning",
                    "abstract": "Model-based reinforcement learning algorithms with probabilistic dynamical models are amongst the most data-efficient learning methods. This is often attributed to their ability to distinguish between epistemic and aleatoric uncertainty. However, while most algorithms distinguish these two uncertainties for {\\em learning} the model, they ignore it when {\\em optimizing} the policy. In this paper, we show that ignoring the epistemic uncertainty leads to greedy algorithms that do not explore sufficiently. In turn, we propose a {\\em practical optimistic-exploration algorithm} (\\alg), which enlarges the input space with {\\em hallucinated} inputs that can exert as much control as the {\\em epistemic} uncertainty in the model affords. We analyze this setting and construct a general regret bound for well-calibrated models, which is provably sublinear in the case of Gaussian Process models. Based on this theoretical foundation, we show how optimistic exploration can be easily combined with state-of-the-art reinforcement learning algorithms and different probabilistic models. Our experiments demonstrate that optimistic exploration significantly speeds up learning when there are penalties on actions, a setting that is notoriously difficult for existing model-based reinforcement learning algorithms."
                },
                {
                    "title": "Hypermodels for Exploration",
                    "abstract": "We study the use of hypermodels to represent epistemic uncertainty and guide exploration. This generalizes and extends the use of ensembles to approximate Thompson sampling. The computational cost of training an ensemble grows with its size, and as such, prior work has typically been limited to ensembles with tens of elements. We show that alternative hypermodels can enjoy dramatic efficiency gains, enabling behavior that would otherwise require hundreds or thousands of elements, and even succeed in situations where ensemble methods fail to learn regardless of size. This allows more accurate approximation of Thompson sampling as well as use of more sophisticated exploration schemes. In particular, we consider an approximate form of information-directed sampling and demonstrate performance gains relative to Thompson sampling. As alternatives to ensembles, we consider linear and neural network hypermodels, also known as hypernetworks. We prove that, with neural network base models, a linear hypermodel can represent essentially any distribution over functions, and as such, hypernetworks do not extend what can be represented."
                },
                {
                    "title": "Frequentist Regret Bounds for Randomized Least-Squares Value Iteration",
                    "abstract": "We consider the exploration-exploitation dilemma in finite-horizon reinforcement learning (RL). When the state space is large or continuous, traditional tabular approaches are unfeasible and some form of function approximation is mandatory. In this paper, we introduce an optimistically-initialized variant of the popular randomized least-squares value iteration (RLSVI), a model-free algorithm where exploration is induced by perturbing the least-squares approximation of the action-value function. Under the assumption that the Markov decision process has low-rank transition dynamics, we prove that the frequentist regret of RLSVI is upper-bounded by $\\widetilde O(d^2 H^2 \\sqrt{T})$ where $ d $ are the feature dimension, $ H $ is the horizon, and $ T $ is the total number of steps. To the best of our knowledge, this is the first frequentist regret analysis for randomized exploration with function approximation."
                },
                {
                    "title": "Bootstrap Thompson Sampling and Sequential Decision Problems in the Behavioral Sciences",
                    "abstract": "Behavioral scientists are increasingly able to conduct randomized experiments in settings that enable rapidly updating probabilities of assignment to treatments (i.e., arms). Thus, many behavioral science experiments can be usefully formulated as sequential decision problems. This article reviews versions of the multiarmed bandit problem with an emphasis on behavioral science applications. One popular method for such problems is Thompson sampling, which is appealing for randomizing assignment and being asymptoticly consistent in selecting the best arm. Here, we show the utility of bootstrap Thompson sampling (BTS), which replaces the posterior distribution with the bootstrap distribution. This often has computational and practical advantages. We illustrate its robustness to model misspecification, which is a common concern in behavioral science applications. We show how BTS can be readily adapted to be robust to dependent data, such as repeated observations of the same units, which is common in behavioral science applications. We use simulations to illustrate parametric Thompson sampling and BTS for Bernoulli bandits, factorial Gaussian bandits, and bandits with repeated observations of the same units."
                },
                {
                    "title": "Perturbed-History Exploration in Stochastic Linear Bandits",
                    "abstract": "We propose a new online algorithm for minimizing the cumulative regret in stochastic linear bandits. The key idea is to build a perturbed history, which mixes the history of observed rewards with a pseudo-history of randomly generated i.i.d. pseudo-rewards. Our algorithm, perturbed-history exploration in a linear bandit (LinPHE), estimates a linear model from its perturbed history and pulls the arm with the highest value under that model. We prove a $\\tilde{O}(d \\sqrt{n})$ gap-free bound on the expected $n$-round regret of LinPHE, where $d$ is the number of features. Our analysis relies on novel concentration and anti-concentration bounds on the weighted sum of Bernoulli random variables. To show the generality of our design, we extend LinPHE to a logistic reward model. We evaluate both algorithms empirically and show that they are practical."
                },
                {
                    "title": "Exploration by Random Network Distillation",
                    "abstract": "We introduce an exploration bonus for deep reinforcement learning methods that is easy to implement and adds minimal overhead to the computation performed. The bonus is the error of a neural network predicting features of the observations given by a fixed randomly initialized neural network. We also introduce a method to flexibly combine intrinsic and extrinsic rewards. We find that the random network distillation (RND) bonus combined with this increased flexibility enables significant progress on several hard exploration Atari games. In particular we establish state of the art performance on Montezuma's Revenge, a game famously difficult for deep reinforcement learning methods. To the best of our knowledge, this is the first method that achieves better than average human performance on this game without using demonstrations or having access to the underlying state of the game, and occasionally completes the first level."
                },
                {
                    "title": "Efficient online recommendation via low-rank ensemble sampling",
                    "abstract": "The low-rank structure is one of the most prominent features in modern recommendation problems. In this paper, we consider an online learning problem with a low-rank expected reward matrix where both row features and column features are unknown a priori, and the agent aims to learn to choose the best row-column pair (i.e. the maximum entry) in the matrix. We develop a novel online recommendation algorithm based on ensemble sampling, a recently developed computationally efficient approximation of Thompson sampling. Our computational results show that our algorithm consistently achieves order-of-magnitude improvements over the baselines in both synthetic and real-world experiments."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "Randomized Prior Functions for Deep Reinforcement Learning",
                    "abstract": "Dealing with uncertainty is essential for efficient reinforcement learning. There is a growing literature on uncertainty estimation for deep learning from fixed datasets, but many of the most popular approaches are poorly-suited to sequential decision problems. Other methods, such as bootstrap sampling, have no mechanism for uncertainty that does not come from the observed data. We highlight why this can be a crucial shortcoming and propose a simple remedy through addition of a randomized untrainable `prior' network to each ensemble member. We prove that this approach is efficient with linear representations, provide simple illustrations of its efficacy with nonlinear representations and show that this approach scales to large-scale problems far better than previous attempts."
                },
                {
                    "title": "Coordinated Exploration in Concurrent Reinforcement Learning",
                    "abstract": "We consider a team of reinforcement learning agents that concurrently learn to operate in a common environment. We identify three properties - adaptivity, commitment, and diversity - which are necessary for efficient coordinated exploration and demonstrate that straightforward extensions to single-agent optimistic and posterior sampling approaches fail to satisfy them. As an alternative, we propose seed sampling, which extends posterior sampling in a manner that meets these requirements. Simulation results investigate how per-agent regret decreases as the number of agents grows, establishing substantial advantages of seed sampling over alternative exploration schemes."
                },
                {
                    "title": "Ensemble Sampling",
                    "abstract": "Thompson sampling has emerged as an effective heuristic for a broad range of online decision problems. In its basic form, the algorithm requires computing and sampling from a posterior distribution over models, which is tractable only for simple special cases. This paper develops ensemble sampling, which aims to approximate Thompson sampling while maintaining tractability even in the face of complex models such as neural networks. Ensemble sampling dramatically expands on the range of applications for which Thompson sampling is viable. We establish a theoretical basis that supports the approach and present computational results that offer further insight."
                },
                {
                    "title": "Deep Exploration via Randomized Value Functions",
                    "abstract": "We study the use of randomized value functions to guide deep exploration in reinforcement learning. This offers an elegant means for synthesizing statistically and computationally efficient exploration with common practical approaches to value function learning. We present several reinforcement learning algorithms that leverage randomized value functions and demonstrate their efficacy through computational studies. We also prove a regret bound that establishes statistical efficiency with a tabular representation."
                },
                {
                    "title": "Linear Thompson Sampling Revisited",
                    "abstract": "We derive an alternative proof for the regret of Thompson sampling (\\ts) in the stochastic linear bandit setting. While we obtain a regret bound of order $\\wt{O}(d^{3/2}\\sqrt{T})$ as in previous results, the proof sheds new light on the functioning of the \\ts. We leverage on the structure of the problem to show how the regret is related to the sensitivity (i.e., the gradient) of the objective function and how selecting optimal arms associated to \\textit{optimistic} parameters does control it. Thus we show that \\ts can be seen as a generic randomized algorithm where the sampling distribution is designed to have a fixed probability of being optimistic, at the cost of an additional $\\sqrt{d}$ regret factor compared to a UCB-like approach. Furthermore, we show that our proof can be readily applied to regularized linear optimization and generalized linear model problems."
                },
                {
                    "title": "Order statistics of vectors with dependent coordinates, and the Karhunen\u2013Lo\u00e8ve basis",
                    "abstract": "Let $X$ be an $n$-dimensional random centered Gaussian vector with independent but not identically distributed coordinates and let $T$ be an orthogonal trasformation of $\\mathbb R^n$. We show that the random vector $Y=T(X)$ satisfies $$\\mathbb E\\sum\\limits_{j=1}^k j\\mbox{-}\\min_{i\\leq n}{X_{i}}^2 \\leq C\\mathbb E\\sum\\limits_{j=1}^k j\\mbox{-}\\min_{i\\leq n}{Y_{i}}^2$$ for all $k 0$ is a universal constant. This resolves (up to a multiplicative constant) an old question of S.Mallat and O.Zeitouni regarding optimality of the Karhunen-Loeve basis for the nonlinear signal approximation. As a by-product we obtain some relations for order statistics of random vectors (not only Gaussian) which are of independent interest."
                },
                {
                    "title": "Deep Exploration via Bootstrapped DQN",
                    "abstract": "Efficient exploration in complex environments remains a major challenge for reinforcement learning. We propose bootstrapped DQN, a simple algorithm that explores in a computationally and statistically efficient manner through use of randomized value functions. Unlike dithering strategies such as epsilon-greedy exploration, bootstrapped DQN carries out temporally-extended (or deep) exploration; this can lead to exponentially faster learning. We demonstrate these benefits in complex stochastic MDPs and in the large-scale Arcade Learning Environment. Bootstrapped DQN substantially improves learning times and performance across most Atari games."
                },
                {
                    "title": "Concentration Inequalities: A Nonasymptotic Theory of Independence",
                    "abstract": "Concentration inequalities for functions of independent random variables is an area of probability theory that has witnessed a great revolution in the last few decades, and has applications in a wide variety of areas such as machine learning, statistics, discrete mathematics, and high-dimensional geometry. Roughly speaking, if a function of many independent random variables does not depend too much on any of the variables then it is concentrated in the sense that with high probability, it is close to its expected value. This book offers a host of inequalities to illustrate this rich theory in an accessible way by covering the key developments and applications in the field. The authors describe the interplay between the probabilistic structure (independence) and a variety of tools ranging from functional inequalities to transportation arguments to information theory. Applications to the study of empirical processes, random projections, random matrix theory, and threshold phenomena are also presented. A self-contained introduction to concentration inequalities, it includes a survey of concentration of sums of independent random variables, variance bounds, the entropy method, and the transportation method. Deep connections with isoperimetric problems are revealed whilst special attention is paid to applications to the supremum of empirical processes. Written by leading experts in the field and containing extensive exercise sections this book will be an invaluable resource for researchers and graduate students in mathematics, theoretical computer science, and engineering."
                },
                {
                    "title": "Thompson Sampling for Contextual Bandits with Linear Payoffs",
                    "abstract": "Thompson Sampling is one of the oldest heuristics for multi-armed bandit problems. It is a randomized algorithm based on Bayesian ideas, and has recently generated significant interest after several studies demonstrated it to have better empirical performance compared to the state-of-the-art methods. However, many questions regarding its theoretical performance remained open. In this paper, we design and analyze a generalization of Thompson Sampling algorithm for the stochastic contextual multi-armed bandit problem with linear payoff functions, when the contexts are provided by an adaptive adversary. This is among the most important and widely studied version of the contextual bandits problem. We prove a high probability regret bound of O(d2/e\u221aT1+e) in time T for any 0 < e < 1, where d is the dimension of each context vector and e is a parameter used by the algorithm. Our results provide the first theoretical guarantees for the contextual version of Thompson Sampling, and are close to the lower bound of \u03a9(d\u221aT) for this problem. This essentially solves a COLT open problem of Chapelle and Li [COLT 2012]."
                },
                {
                    "title": "Improved Algorithms for Linear Stochastic Bandits",
                    "abstract": "We improve the theoretical analysis and empirical performance of algorithms for the stochastic multi-armed bandit problem and the linear stochastic multi-armed bandit problem. In particular, we show that a simple modification of Auer's UCB algorithm (Auer, 2002) achieves with high probability constant regret. More importantly, we modify and, consequently, improve the analysis of the algorithm for the for linear stochastic bandit problem studied by Auer (2002), Dani et al. (2008), Rusmevichientong and Tsitsiklis (2010), Li et al. (2010). Our modification improves the regret bound by a logarithmic factor, though experiments show a vast improvement. In both cases, the improvement stems from the construction of smaller confidence sets. For their construction we use a novel tail inequality for vector-valued martingales."
                },
                {
                    "title": "Parametric Bandits: The Generalized Linear Case",
                    "abstract": "We consider structured multi-armed bandit problems based on the Generalized Linear Model (GLM) framework of statistics. For these bandits, we propose a new algorithm, called GLM-UCB. We derive finite time, high probability bounds on the regret of the algorithm, extending previous analyses developed for the linear bandits to the non-linear case. The analysis highlights a key difficulty in generalizing linear bandit algorithms to the non-linear case, which is solved in GLM-UCB by focusing on the reward space rather than on the parameter space. Moreover, as the actual effectiveness of current parameterized bandit algorithms is often poor in practice, we provide a tuning method based on asymptotic arguments, which leads to significantly better practical performance. We present two numerical experiments on real-world data that illustrate the potential of the GLM-UCB approach."
                },
                {
                    "title": "SELF-NORMALIZED PROCESSES: EXPONENTIAL INEQUALITIES, MOMENT BOUNDS AND ITERATED LOGARITHM LAWS",
                    "abstract": "Self-normalized processes arise naturally in statistical applications. Being unit free, they are not affected by scale changes. Moreover, self-normalization often eliminates or weakens moment assumptions. In this paper we present several exponential and moment inequalities, particularly those related to laws of the iterated logarithm, for self-normalized random variables including martingales. Tail probability bounds are also derived. For random variables Bt>0 and At, let Yt(\u03bb)=exp{\u03bbAt\u2212\u03bb2Bt2/2}. We develop inequalities for the moments of At/Bt or supt\u22650At/{Bt(log logBt)1/2} and variants thereof, when EYt(\u03bb)\u22641 or when Yt(\u03bb) is a supermartingale, for all \u03bb belonging to some interval. Our results are valid for a wide class of random processes including continuous martingales with At=Mt and $B_{t}=\\sqrt {\\langle M\\rangle _{t}}$ , and sums of conditionally symmetric variables di with At=\u2211i=1tdi and $B_{t}=\\sqrt{\\sum_{i=1}^{t}d_{i}^{2}}$ . A sharp maximal inequality for conditionally symmetric random variables and for continuous local martingales with values in Rm, m\u22651, is also established. Another development in this paper is a bounded law of the iterated logarithm for general adapted sequences that are centered at certain truncated conditional expectations and self-normalized by the square root of the sum of squares. The key ingredient in this development is a new exponential supermartingale involving \u2211i=1tdi and \u2211i=1tdi2. A compact law of the iterated logarithm for self-normalized martingales is also derived in this connection."
                },
                {
                    "title": "Symmetric Multivariate and Related Distributions",
                    "abstract": "Part 1 Preliminaries: construction of symmetric multivariate distributions notation of algebraic entities and characteristics of random quantities the \"d\" operator groups and invariance dirichlet distribution problems 1. Part 2 Spherically and elliptically symmetric distributions: introduction and definition marginal distributions, moments and density marginal distributions moments density the relationship between (phi) and f conditional distributions properties of elliptically symmetric distributions mixtures of normal distributions robust statistics and regression model robust statistics regression model log-elliptical and additive logistic elliptical distributions multivariate log-elliptical distribution additive logistic elliptical distributions complex elliptically symmetric distributions. Part 3 Some subclasses of elliptical distributions: multiuniform distribution the characteristic function moments marginal distribution conditional distributions uniform distribution in the unit sphere discussion symmetric Kotz type distributions definition distribution of R(2) moments multivariate normal distributions the c.f. of Kotz type distributions symmetric multivariate Pearson type VII distributions definition marginal densities conditional distributions moments conditional distributions moments some examples extended Tn family relationships between Ln and Tn families of distributions order statistics mixtures of exponential distributions independence, robustness and characterizations problems V. Part 6 Multivariate Liouville distributions: definitions and properties examples marginal distributions conditional distribution characterizations scale-invariant statistics survival functions inequalities and applications."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively analyze and guarantee the performance of ensemble sampling algorithms in the stochastic linear bandit setting?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant gap in the theoretical understanding of ensemble sampling, which is widely used in various applications, including deep reinforcement learning and online recommendation systems. By providing a rigorous analysis, this research could lead to improved algorithms that enhance exploration-exploitation balance, ultimately advancing knowledge in sequential decision-making and enabling more effective practical applications across multiple domains.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the implicit control of the distribution of models given past data in ensemble sampling, which complicates the analysis. Unlike other randomized exploration methods, ensemble sampling does not follow a straightforward model fitting and perturbation pattern, making it difficult to establish guarantees on its performance. Naive approaches may fail because they do not account for the unique structure of ensemble sampling, which requires a more sophisticated theoretical framework to analyze its behavior and performance.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has struggled to analyze ensemble sampling due to its complex nature and the lack of a clear framework for understanding its performance. Existing analyses have either been impractical or have not successfully captured the nuances of the algorithm. The barriers include the need for ensemble sizes that scale with the horizon and the implicit nature of model distribution. Our approach differs by providing a guarantee with a logarithmic ensemble size in relation to the horizon and a linear relationship with the number of features, marking a significant advancement over prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a detailed analysis of a symmetrized version of the ensemble sampling algorithm within the stochastic linear bandit setting. We will utilize a framework that extends existing methods to derive performance guarantees. The dataset will consist of structured bandit problems, and we will measure performance using regret metrics. The expected outcome is to demonstrate that our approach incurs regret no worse than order \\((d \\log T)^{5/2} \\sqrt{T}\\), providing the first successful analysis of ensemble sampling in this context and paving the way for future research in generalized linear bandits and reinforcement learning."
            }
        },
        "author_data": {
            "9f5de610-a2ed-49ae-83a4-76f74318890c": {
                "pk": "9f5de610-a2ed-49ae-83a4-76f74318890c",
                "name": "David Janz",
                "collaborators": [
                    "Jos\u00e9 Miguel Hern\u00e1ndez-Lobato",
                    "Javier Antor\u00e1n",
                    "Shreyas Padhy",
                    "Csaba Szepesv\u00e1ri",
                    "Jihao Andreas Lin",
                    "Alexander Terenin",
                    "Brooks Paige",
                    "Jos van der Westhuizen",
                    "Riccardo Barbano",
                    "Eric Nalisnick"
                ],
                "domain": [
                    "Gaussian Processes",
                    "Reinforcement Learning",
                    "Bandit Algorithms",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Bandit optimisation of functions in the Mat\u00e9rn kernel RKHS",
                        "abstract": "We consider the problem of optimising functions in the reproducing kernel Hilbert space (RKHS) of a Mat\\'ern kernel with smoothness parameter $\\nu$ over the domain $[0,1]^d$ under noisy bandit feedback. Our contribution, the $\\pi$-GP-UCB algorithm, is the first practical approach with guaranteed sublinear regret for all $\\nu>1$ and $d \\geq 1$. Empirical validation suggests better performance and drastically improved computational scalablity compared with its predecessor, Improved GP-UCB."
                    },
                    {
                        "title": "Exploration via linearly perturbed loss minimisation",
                        "abstract": "We introduce exploration via linear loss perturbations (EVILL), a randomised exploration method for structured stochastic bandit problems that works by solving for the minimiser of a linearly perturbed regularised negative log-likelihood function. We show that, for the case of generalised linear bandits, EVILL reduces to perturbed history exploration (PHE), a method where exploration is done by training on randomly perturbed rewards. In doing so, we provide a simple and clean explanation of when and why random reward perturbations give rise to good bandit algorithms. We propose data-dependent perturbations not present in previous PHE-type methods that allow EVILL to match the performance of Thompson-sampling-style parameter-perturbation methods, both in theory and in practice. Moreover, we show an example outside generalised linear bandits where PHE leads to inconsistent estimates, and thus linear regret, while EVILL remains performant. Like PHE, EVILL can be implemented in just a few lines of code."
                    },
                    {
                        "title": "Successor Uncertainties: Exploration and Uncertainty in Temporal Difference Learning",
                        "abstract": "Posterior sampling for reinforcement learning (PSRL) is an effective method for balancing exploration and exploitation in reinforcement learning. Randomised value functions (RVF) can be viewed as a promising approach to scaling PSRL. However, we show that most contemporary algorithms combining RVF with neural network function approximation do not possess the properties which make PSRL effective, and provably fail in sparse reward problems. Moreover, we find that propagation of uncertainty, a property of PSRL previously thought important for exploration, does not preclude this failure. We use these insights to design Successor Uncertainties (SU), a cheap and easy to implement RVF algorithm that retains key properties of PSRL. SU is highly effective on hard tabular exploration benchmarks. Furthermore, on the Atari 2600 domain, it surpasses human performance on 38 of 49 games tested (achieving a median human normalised score of 2.09), and outperforms its closest RVF competitor, Bootstrapped DQN, on 36 of those."
                    },
                    {
                        "title": "Stochastic Gradient Descent for Gaussian Processes Done Right",
                        "abstract": "As is well known, both sampling from the posterior and computing the mean of the posterior in Gaussian process regression reduces to solving a large linear system of equations. We study the use of stochastic gradient descent for solving this linear system, and show that when \\emph{done right} -- by which we mean using specific insights from the optimisation and kernel communities -- stochastic gradient descent is highly effective. To that end, we introduce a particularly simple \\emph{stochastic dual descent} algorithm, explain its design in an intuitive manner and illustrate the design choices through a series of ablation studies. Further experiments demonstrate that our new method is highly competitive. In particular, our evaluations on the UCI regression tasks and on Bayesian optimisation set our approach apart from preconditioned conjugate gradients and variational Gaussian process approximations. Moreover, our method places Gaussian process regression on par with state-of-the-art graph neural networks for molecular binding affinity prediction."
                    },
                    {
                        "title": "Probabilistic structure discovery in time series data",
                        "abstract": "Existing methods for structure discovery in time series data construct interpretable, compositional kernels for Gaussian process regression models. While the learned Gaussian process model provides posterior mean and variance estimates, typically the structure is learned via a greedy optimization procedure. This restricts the space of possible solutions and leads to over-confident uncertainty estimates. We introduce a fully Bayesian approach, inferring a full posterior over structures, which more reliably captures the uncertainty of the model."
                    },
                    {
                        "title": "Actively Learning what makes a Discrete Sequence Valid",
                        "abstract": "Deep learning techniques have been hugely successful for traditional supervised and unsupervised machine learning problems. In large part, these techniques solve continuous optimization problems. Recently however, discrete generative deep learning models have been successfully used to efficiently search high-dimensional discrete spaces. These methods work by representing discrete objects as sequences, for which powerful sequence-based deep models can be employed. Unfortunately, these techniques are significantly hindered by the fact that these generative models often produce invalid sequences. As a step towards solving this problem, we propose to learn a deep recurrent validator model. Given a partial sequence, our model learns the probability of that sequence occurring as the beginning of a full valid sequence. Thus this identifies valid versus invalid sequences and crucially it also provides insight about how individual sequence elements influence the validity of discrete objects. To learn this model we propose an approach inspired by seminal work in Bayesian active learning. On a synthetic dataset, we demonstrate the ability of our model to distinguish valid and invalid sequences. We believe this is a key step toward learning generative models that faithfully produce valid discrete objects."
                    },
                    {
                        "title": "Sampling-based inference for large linear models, with application to linearised Laplace",
                        "abstract": "Large-scale linear models are ubiquitous throughout machine learning, with contemporary application as surrogate models for neural network uncertainty quantification; that is, the linearised Laplace method. Alas, the computational cost associated with Bayesian linear models constrains this method's application to small networks, small output spaces and small datasets. We address this limitation by introducing a scalable sample-based Bayesian inference method for conjugate Gaussian multi-output linear models, together with a matching method for hyperparameter (regularisation) selection. Furthermore, we use a classic feature normalisation method (the g-prior) to resolve a previously highlighted pathology of the linearised Laplace method. Together, these contributions allow us to perform linearised neural network inference with ResNet-18 on CIFAR100 (11M parameters, 100 outputs x 50k datapoints), with ResNet-50 on Imagenet (50M parameters, 1000 outputs x 1.2M datapoints) and with a U-Net on a high-resolution tomographic reconstruction task (2M parameters, 251k output~dimensions)."
                    },
                    {
                        "title": "Adapting the Linearised Laplace Model Evidence for Modern Deep Learning",
                        "abstract": "The linearised Laplace method for estimating model uncertainty has received renewed attention in the Bayesian deep learning community. The method provides reliable error bars and admits a closed-form expression for the model evidence, allowing for scalable selection of model hyperparameters. In this work, we examine the assumptions behind this method, particularly in conjunction with model selection. We show that these interact poorly with some now-standard tools of deep learning--stochastic approximation methods and normalisation layers--and make recommendations for how to better adapt this classic method to the modern setting. We provide theoretical support for our recommendations and validate them empirically on MLPs, classic CNNs, residual networks with and without normalisation layers, generative autoencoders and transformers."
                    },
                    {
                        "title": "Learning to Drive in a Day",
                        "abstract": "We demonstrate the first application of deep reinforcement learning to autonomous driving. From randomly initialised parameters, our model is able to learn a policy for lane following in a handful of training episodes using a single monocular image as input. We provide a general and easy to obtain reward: the distance travelled by the vehicle without the safety driver taking control. We use a continuous, model-free deep reinforcement learning algorithm, with all exploration and optimisation performed on-vehicle. This demonstrates a new framework for autonomous driving which moves away from reliance on defined logical rules, mapping, and direct supervision. We discuss the challenges and opportunities to scale this approach to a broader range of autonomous driving tasks."
                    },
                    {
                        "title": "Learning a Generative Model for Validity in Complex Discrete Structures",
                        "abstract": "Deep generative models have been successfully used to learn representations for high-dimensional discrete spaces by representing discrete objects as sequences and employing powerful sequence-based deep models. Unfortunately, these sequence-based models often produce invalid sequences: sequences which do not represent any underlying discrete structure; invalid sequences hinder the utility of such models. As a step towards solving this problem, we propose to learn a deep recurrent validator model, which can estimate whether a partial sequence can function as the beginning of a full, valid sequence. This validator provides insight as to how individual sequence elements influence the validity of the overall sequence, and can be used to constrain sequence based models to generate valid sequences -- and thus faithfully model discrete objects. Our approach is inspired by reinforcement learning, where an oracle which can evaluate validity of complete sequences provides a sparse reward signal. We demonstrate its effectiveness as a generative model of Python 3 source code for mathematical expressions, and in improving the ability of a variational autoencoder trained on SMILES strings to decode valid molecular structures."
                    },
                    {
                        "title": "Sampling from Gaussian Process Posteriors using Stochastic Gradient Descent",
                        "abstract": "Gaussian processes are a powerful framework for quantifying uncertainty and for sequential decision-making but are limited by the requirement of solving linear systems. In general, this has a cubic cost in dataset size and is sensitive to conditioning. We explore stochastic gradient algorithms as a computationally efficient method of approximately solving these linear systems: we develop low-variance optimization objectives for sampling from the posterior and extend these to inducing points. Counterintuitively, stochastic gradient descent often produces accurate predictions, even in cases where it does not converge quickly to the optimum. We explain this through a spectral characterization of the implicit bias from non-convergence. We show that stochastic gradient descent produces predictive distributions close to the true posterior both in regions with sufficient data coverage, and in regions sufficiently far away from the data. Experimentally, stochastic gradient descent achieves state-of-the-art performance on sufficiently large-scale or ill-conditioned regression tasks. Its uncertainty estimates match the performance of significantly more expensive baselines on a large-scale Bayesian optimization task."
                    }
                ]
            },
            "ff5135b0-52f8-45e0-be49-99ffd7448faa": {
                "pk": "ff5135b0-52f8-45e0-be49-99ffd7448faa",
                "name": "Alexander E. Litvak",
                "collaborators": [
                    "Nicole Tomczak-Jaegermann",
                    "Alain Pajor",
                    "Rados\u0142aw Adamczak",
                    "Konstantin Tikhomirov",
                    "Olivier Gu\u00e9don",
                    "Rafa\u0142 Lata\u0142a",
                    "Dmitry Zaporozhets",
                    "Andrii Arman",
                    "Konstantin E. Tikhomirov",
                    "Aicke Hinrichs"
                ],
                "domain": [
                    "Probability Theory",
                    "Random Matrices",
                    "Convex Geometry",
                    "Statistical Learning"
                ],
                "publications": [
                    {
                        "title": "Minimal dispersion on the cube and the torus",
                        "abstract": "We improve some upper bounds for minimal dispersion on the cube and torus. /Our new ingredient is an improvement of a probabilistic lemma used to obtain upper bounds for dispersion in several previous works. Our new lemma combines a random and non-random choice of points in the cube. This leads to better upper bounds for the minimal dispersion."
                    },
                    {
                        "title": "Singularity of sparse Bernoulli matrices",
                        "abstract": "Let $M_n$ be an $n\\times n$ random matrix with i.i.d. Bernoulli(p) entries. We show that there is a universal constant $C\\geq 1$ such that, whenever $p$ and $n$ satisfy $C\\log n/n\\leq p\\leq C^{-1}$, \\begin{align*} {\\mathbb P}\\big\\{\\mbox{$M_n$ is singular}\\big\\}&=(1+o_n(1)){\\mathbb P}\\big\\{\\mbox{$M_n$ contains a zero row or column}\\big\\}\\\\ &=(2+o_n(1))n\\,(1-p)^n, \\end{align*} where $o_n(1)$ denotes a quantity which converges to zero as $n\\to\\infty$. We provide the corresponding upper and lower bounds on the smallest singular value of $M_n$ as well."
                    },
                    {
                        "title": "On the expectation of operator norms of random matrices",
                        "abstract": "We prove estimates for the expected value of operator norms of Gaussian random matrices with independent and mean-zero entries, acting as operators from $\\ell^m_{p^*}$ to $\\ell_q^n$, $1\\leq p^* \\leq 2 \\leq q \\leq \\infty$."
                    },
                    {
                        "title": "On the vertex index of convex bodies",
                        "abstract": "We introduce the vertex index, vein(K), of a given centrally symmetric convex body K, which, in a sense, measures how well K can be inscribed into a convex polytope with small number of vertices. This index is closely connected to the illumination parameter of a body, introduced earlier by the first named author, and, thus, related to the famous conjecture in Convex Geometry about covering of a d-dimensional body by 2^d smaller positively homothetic copies. We provide asymptotically sharp estimates (up to a logarithmic term) of this index in the general case. Also, we provide sharp estimates in dimensions 2 and 3."
                    },
                    {
                        "title": "On approximations by projections of polytopes with few facets",
                        "abstract": "We provide an affirmative answer to a problem posed by Barvinok and Veomett, showing that in general an n-dimensional convex body cannot be approximated by a projection of a section of a simplex of a sub-exponential dimension. Moreover, we establish a lower bound of the Banach-Mazur distance between n-dimensional projections of sections of an N-dimensional simplex and a certain convex symmetric body, which is sharp up to a logarithmic factor for all N>n."
                    },
                    {
                        "title": "Estimates for order statistics in terms of quantiles",
                        "abstract": "Let $X_1, \\ldots, X_n$ be independent non-negative random variables with cumulative distribution functions $F_1,F_2,\\ldots,F_n$, each satisfying certain (rather mild) conditions. We show that the median of $k$-th smallest order statistic of the vector $(X_1, \\ldots, X_n)$ is equivalent to the quantile of order $(k-1/2)/n$ with respect to the averaged distribution $F=\\frac{1}{n}\\sum_{i=1}^n F_i$."
                    },
                    {
                        "title": "New bounds on the minimal dispersion",
                        "abstract": "We provide a new construction for a set of boxes approximating axis-parallel boxes of fixed volume in $[0, 1]^d$. This improves upper bounds for the minimal dispersion of a point set in the unit cube and its inverse in both the periodic and non-periodic settings in certain regimes. Up to double logarithmic factor, our bounds are sharp. We also apply our construction to $k$-dispersion."
                    },
                    {
                        "title": "A short proof of Paouris' inequality",
                        "abstract": "We give a short proof of a result of G. Paouris on the tail behaviour of the Euclidean norm $|X|$ of an isotropic log-concave random vector $X\\in\\R^n$, stating that for every $t\\geq 1$, $P(|X|\\geq ct\\sqrt n)\\leq \\exp(-t\\sqrt n)$. More precisely we show that for any log-concave random vector $X$ and any $p\\geq 1$, $(E|X|^p)^{1/p}\\sim E |X|+\\sup_{z\\in S^{n-1}}(E |< z,X>|^p)^{1/p}$."
                    },
                    {
                        "title": "Order statistics of vectors with dependent coordinates, and the Karhunen-Lo\u00e8ve basis",
                        "abstract": "Let $X$ be an $n$-dimensional random centered Gaussian vector with independent but not identically distributed coordinates and let $T$ be an orthogonal trasformation of $\\mathbb R^n$. We show that the random vector $Y=T(X)$ satisfies $$\\mathbb E\\sum\\limits_{j=1}^k j\\mbox{-}\\min_{i\\leq n}{X_{i}}^2 \\leq C\\mathbb E\\sum\\limits_{j=1}^k j\\mbox{-}\\min_{i\\leq n}{Y_{i}}^2$$ for all $k<n$, where \"$j\\mbox{-}\\min$\" denotes the $j$-th smallest component of corresponding vector and $C>0$ is a universal constant. This resolves (up to a multiplicative constant) an old question of S.Mallat and O.Zeitouni regarding optimality of the Karhunen-Loeve basis for the nonlinear signal approximation. As a by-product we obtain some relations for order statistics of random vectors (not only Gaussian) which are of independent interest."
                    },
                    {
                        "title": "Random section and random simplex inequality",
                        "abstract": "Consider some convex body $K\\subset\\mathbb R^d$. Let $X_1,\\dots, X_k$, where $k\\leq d$, be random points independently and uniformly chosen in $K$, and let $\\xi_k$ be a uniformly distributed random linear $k$-plane. We show that for $p\\geq-d+k+1$, \\[ \\mathbb E\\,|K\\cap\\xi_k|^{d+p}\\leq c_{d,k,p} \\cdot|K|^k\\, \\,\\mathbb E\\,|\\mathrm{conv}(0,X_1, \\dots,X_k)|^p, \\] where $|\\cdot|$ and $\\mathrm{conv}$ denote the volume of correspondent dimension and the convex hull. The constant $c_{d,k,p}$ is such that for $k>1$ the equality holds if and only if $K$ is an ellipsoid centered at the origin, and for $k=1$ the inequality turns to equality. If $p=0$, then the inequality reduces to the Busemann intersection inequality, and if $k=d$ -- to the Busemann random simplex inequality. We also present an affine version of this inequality which similarly generalizes the Schneider inequality and the Blaschke-Gr\\\"omer inequality."
                    },
                    {
                        "title": "Geometry of log-concave Ensembles of random matrices and approximate reconstruction",
                        "abstract": "We study the Restricted Isometry Property of a random matrix $\\Gamma$ with independent isotropic log-concave rows. To this end, we introduce a parameter $\\Gamma_{k,m}$ that controls uniformly the operator norm of sub-matrices with $k$ rows and $m$ columns. This parameter is estimated by means of new tail estimates of order statistics and deviation inequalities for norms of projections of an isotropic log-concave vector."
                    },
                    {
                        "title": "Mean width of regular polytopes and expected maxima of correlated Gaussian variables",
                        "abstract": "An old conjecture states that among all simplices inscribed in the unit sphere the regular one has the maximal mean width. An equivalent formulation is that for any centered Gaussian vector $(\\xi_1,\\dots,\\xi_n)$ satisfying $\\mathbb E\\xi_1^2= \\dots =\\mathbb E\\xi_n^2=1$ one has $$   \\mathbb E\\,\\max\\{\\xi_1,\\dots,\\xi_n\\}\\leq\\sqrt{\\frac{n}{n-1}}\\, \\mathbb E\\,\\max\\{\\eta_1,\\dots,\\eta_n\\}, $$ where $\\eta_1,\\eta_2,\\dots,$ are independent standard Gaussian variables. Using this probabilistic interpretation we derive an asymptotic version of the conjecture. We also show that the mean width of the regular simplex with $2n$ vertices is remarkably close to the mean width of the regular crosspolytope with the same number of vertices. Interpreted probabilistically, our result states that $$   1\\leq\\frac{\\mathbb E\\,\\max\\{|\\eta_1|,\\dots,|\\eta_n|\\}}{\\mathbb E\\,\\max\\{\\eta_1,\\dots,\\eta_{2n}\\}}   \\leq\\min\\left\\{\\sqrt{\\frac{2n}{2n-1}}, \\, 1+\\frac{C}{n\\, \\log n} \\right\\}, $$ where $C>0$ is an absolute constant. We also compute the higher moments of the projection length $W$ of the regular cube, simplex and crosspolytope onto a line with random direction, thus proving several formulas conjectured by S. Finch. Finally, we prove distributional limit theorems for the length of random projection as the dimension goes to $\\infty$. In the case of the $n$-dimensional unit cube $Q_n$, we prove that $$ W_{Q_n} - \\sqrt{\\frac{2n}{\\pi}} \\overset{d}{\\underset{n\\to\\infty}\\longrightarrow} {\\mathcal{N}} \\left(0, \\frac{\\pi-3}{\\pi}\\right), $$ whereas for the simplex and the crosspolytope the limiting distributions are related to the Gumbel double exponential law."
                    },
                    {
                        "title": "Small ball probability for the condition number of random matrices",
                        "abstract": "Let $A$ be an $n\\times n$ random matrix with i.i.d. entries of zero mean, unit variance and a bounded subgaussian moment. We show that the condition number $s_{\\max}(A)/s_{\\min}(A)$ satisfies the small ball probability estimate $${\\mathbb P}\\big\\{s_{\\max}(A)/s_{\\min}(A)\\leq n/t\\big\\}\\leq 2\\exp(-c t^2),\\quad t\\geq 1,$$ where $c>0$ may only depend on the subgaussian moment. Although the estimate can be obtained as a combination of known results and techniques, it was not noticed in the literature before. As a key step of the proof, we apply estimates for the singular values of $A$, ${\\mathbb P}\\big\\{s_{n-k+1}(A)\\leq ck/\\sqrt{n}\\big\\}\\leq 2 \\exp(-c k^2), \\quad 1\\leq k\\leq n,$ obtained (under some additional assumptions) by Nguyen."
                    },
                    {
                        "title": "Sharp bounds on the rate of convergence of the empirical covariance matrix",
                        "abstract": "Let $X_1,..., X_N\\in\\R^n$ be independent centered random vectors with log-concave distribution and with the identity as covariance matrix. We show that with overwhelming probability at least $1 - 3 \\exp(-c\\sqrt{n}\\r)$ one has $   \\sup_{x\\in S^{n-1}} \\Big|\\frac{1/N}\\sum_{i=1}^N (|<X_i, x>|^2 - \\E|<X_i, x>|^2\\r)\\Big|   \\leq C \\sqrt{\\frac{n/N}},$ where $C$ is an absolute positive constant. This result is valid in a more general framework when the linear forms $(<X_i,x>)_{i\\leq N, x\\in S^{n-1}}$ and the Euclidean norms $(|X_i|/\\sqrt n)_{i\\leq N}$ exhibit uniformly a sub-exponential decay. As a consequence, if $A$ denotes the random matrix with columns $(X_i)$, then with overwhelming probability, the extremal singular values $\\lambda_{\\rm min}$ and $\\lambda_{\\rm max}$ of $AA^\\top$ satisfy the inequalities $ 1 - C\\sqrt{{n/N}} \\le {\\lambda_{\\rm min}/N} \\le \\frac{\\lambda_{\\rm max}/N} \\le 1 + C\\sqrt{{n/N}} $ which is a quantitative version of Bai-Yin theorem \\cite{BY} known for random matrices with i.i.d. entries."
                    },
                    {
                        "title": "Quantitative estimates of the convergence of the empirical covariance matrix in Log-concave Ensembles",
                        "abstract": "Let $K$ be an isotropic convex body in $\\R^n$. Given $\\eps>0$, how many independent points $X_i$ uniformly distributed on $K$ are needed for the empirical covariance matrix to approximate the identity up to $\\eps$ with overwhelming probability? Our paper answers this question posed by Kannan, Lovasz and Simonovits. More precisely, let $X\\in\\R^n$ be a centered random vector with a log-concave distribution and with the identity as covariance matrix. An example of such a vector $X$ is a random point in an isotropic convex body. We show that for any $\\eps>0$, there exists $C(\\eps)>0$, such that if $N\\sim C(\\eps) n$ and $(X_i)_{i\\le N}$ are i.i.d. copies of $X$, then $ \\Big\\|\\frac{1}{N}\\sum_{i=1}^N X_i\\otimes X_i - \\Id\\Big\\| \\le \\epsilon, $ with probability larger than $1-\\exp(-c\\sqrt n)$."
                    },
                    {
                        "title": "Restricted isometry property of matrices with independent columns and neighborly polytopes by random sampling",
                        "abstract": "This paper considers compressed sensing matrices and neighborliness of a centrally symmetric convex polytope generated by vectors $\\pm X_1,...,\\pm X_N\\in\\R^n$, ($N\\ge n$). We introduce a class of random sampling matrices and show that they satisfy a restricted isometry property (RIP) with overwhelming probability. In particular, we prove that matrices with i.i.d. centered and variance 1 entries that satisfy uniformly a sub-exponential tail inequality possess this property RIP with overwhelming probability. We show that such \"sensing\" matrices are valid for the exact reconstruction process of $m$-sparse vectors via $\\ell_1$ minimization with $m\\le Cn/\\log^2 (cN/n)$. The class of sampling matrices we study includes the case of matrices with columns that are independent isotropic vectors with log-concave densities. We deduce that if $K\\subset \\R^n$ is a convex body and $X_1,..., X_N\\in K$ are i.i.d. random vectors uniformly distributed on $K$, then, with overwhelming probability, the symmetric convex hull of these points is an $m$-centrally-neighborly polytope with $m\\sim n/\\log^2 (cN/n)$."
                    },
                    {
                        "title": "Numerical range for random matrices",
                        "abstract": "We analyze the numerical range of high-dimensional random matrices, obtaining limit results and corresponding quantitative estimates in the non-limit case. For a large class of random matrices their numerical range is shown to converge to a disc. In particular, numerical range of complex Ginibre matrix almost surely converges to the disk of radius $\\sqrt{2}$. Since the spectrum of non-hermitian random matrices from the Ginibre ensemble lives asymptotically in a neighborhood of the unit disk, it follows that the outer belt of width $\\sqrt{2}-1$ containing no eigenvalues can be seen as a quantification the non-normality of the complex Ginibre random matrix. We also show that the numerical range of upper triangular Gaussian matrices converges to the same disk of radius $\\sqrt{2}$, while all eigenvalues are equal to zero and we prove that the operator norm of such matrices converges to $\\sqrt{2e}$."
                    },
                    {
                        "title": "On the interval of fluctuation of the singular values of random matrices",
                        "abstract": "Let $A$ be a matrix whose columns $X_1,\\dots, X_N$ are independent random vectors in $\\mathbb{R}^n$. Assume that the tails of the 1-dimensional marginals decay as $\\mathbb{P}(|\\langle X_i, a\\rangle|\\geq t)\\leq t^{-p}$ uniformly in $a\\in S^{n-1}$ and $i\\leq N$. Then for $p>4$ we prove that with high probability $A/{\\sqrt{n}}$ has the Restricted Isometry Property (RIP) provided that Euclidean norms $|X_i|$ are concentrated around $\\sqrt{n}$. We also show that the covariance matrix is well approximated by the empirical covariance matrix and establish corresponding quantitative estimates on the rate of convergence in terms of the ratio $n/N$. Moreover, we obtain sharp bounds for both problems when the decay is of the type $ \\exp({-t^{\\alpha}})$ with $\\alpha \\in (0,2]$, extending the known case $\\alpha\\in[1, 2]$."
                    },
                    {
                        "title": "On the volume ratio of projections of convex bodies",
                        "abstract": "We study the volume ratio between projections of two convex bodies. Given a high-dimensional convex body $K$ we show that there is another convex body $L$ such that the volume ratio between any two projections of fixed rank of the bodies $K$ and $L$ is large. Namely, we prove that for every $1\\leq k\\leq n$ and for each convex body $K\\subset \\mathbb{R}^n$ there is a centrally symmetric body $L \\subset \\mathbb{R}^n$ such that for any two projections $P, Q: \\mathbb{R}^n \\to \\mathbb{R}^n$ of rank $k$ one has $$   \\mbox{vr}(PK, QL) \\geq c \\, \\min\\left\\{\\frac{ k}{ \\sqrt{n}} \\, \\sqrt{\\frac{1}{\\log \\log \\log(\\frac{n\\log(n)}{k})}}, \\,   \\frac{\\sqrt{k}}{\\sqrt{\\log(\\frac{n\\log(n)}{k})}}\\right\\}, $$   where $c>0$ is an absolute constant. This general lower bound is sharp (up to logarithmic factors) in the regime $k\\geq n^{2/3}$."
                    }
                ]
            },
            "e6b3d4f5-be46-4a67-82de-0c6cb84ce00e": {
                "pk": "e6b3d4f5-be46-4a67-82de-0c6cb84ce00e",
                "name": "Csaba Szepesv\u00e1ri",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            }
        }
    },
    "2402.04033": {
        "paper_data": {
            "title": "On provable privacy vulnerabilities of graph representations",
            "url": "http://arxiv.org/abs/2402.04033v3",
            "arxiv_id": "2402.04033",
            "authors": [
                "Ruofan Wu",
                "Guanhua Fang",
                "Qiying Pan",
                "Mingyang Zhang",
                "Tengfei Liu",
                "Weiqiang Wang"
            ],
            "abstract": "Graph representation learning (GRL) is critical for extracting insights from complex network structures, but it also raises security concerns due to potential privacy vulnerabilities in these representations. This paper investigates the structural vulnerabilities in graph neural models where sensitive topological information can be inferred through edge reconstruction attacks. Our research primarily addresses the theoretical underpinnings of similarity-based edge reconstruction attacks (SERA), furnishing a non-asymptotic analysis of their reconstruction capacities. Moreover, we present empirical corroboration indicating that such attacks can perfectly reconstruct sparse graphs as graph size increases. Conversely, we establish that sparsity is a critical factor for SERA's effectiveness, as demonstrated through analysis and experiments on (dense) stochastic block models. Finally, we explore the resilience of private graph representations produced via noisy aggregation (NAG) mechanism against SERA. Through theoretical analysis and empirical assessments, we affirm the mitigation of SERA using NAG . In parallel, we also empirically delineate instances wherein SERA demonstrates both efficacy and deficiency in its capacity to function as an instrument for elucidating the trade-off between privacy and utility.",
            "introduction": "   1 Introduction  With the surging developments of graph representation learning (GRL) [15], there has been growing apprehensions concerning the security challenges associated with the deployment of graph neural models in real-world scenarios [10]. GRL models harness the topological information of the underlying graph for producing high-quality predictions or graph representations. Meanwhile, these models bear the risk of inadvertently divulging the same topological information through the graph representations they produce. Such kind of security risks have been empirically validated through the examination of the attacking performance of edge reconstruction algorithms [11, 17, 34, 46], among which a simple form of attack based solely on the representation similarity of node pairs is shown to achieve strikingly strong performance, without the requirement of additional knowledge like encoder architecture or auxiliary datasets [17].   Despite the empirical evidence of topological vulnerabilities of graph representations, theoretical explanations delineating the effectiveness of such attacks remain largely unexplored: As demonstrated in previous studies [11, 17], similarity-based attacks are remarkably effective against sparse graphs that exhibit a generalized homophily pattern, i.e., there exists a significant correlation between the similarity of node features and edge adjacency information. This phenomenon posits that feature similarity may serve as a confounding factor, potentially impacting the efficacy of similarity-based attacks. It is therefore valuable to understand the influence of graph properties, such as feature similarity and sparsity, on the edge reconstruction process of the attacking procedures.   Beyond their capability in characterizing the vulnerabilities of representations, attacking algorithms may also function as empirical attestations of privacy-preserving inference protocols that fulfill formal privacy guarantees such as differential privacy [9, Section 4]. As an illustrative case, membership inference attacks can be employed for auditing differential privacy [31]. Since edge reconstruction is equivalent to edge membership inference on graphs [43], it is thus pertinent to explore the performance of similarity-based attacks when confronted with privacy-preserving graph representations [30, 36].   In this paper, we take initial steps toward a principled understanding of structural vulnerabilities of graph representations under the similarity-based edge reconstruction attack (hereafter abbreviated as SERA) which forms a realistic threat in many practical scenarios such as vertical federated learning [36]. In particular, we establish the following theoretical as well as empirical findings:   (i)  Success modes of SERA Through applying SERA to sparse random graphs equipped with independent random node features, we show that SERA provably reconstructs the input graph via a non-asymptotic analysis. The result indicates that feature similarity is not necessary for SERA to succeed. We conduct both synthetic experiments as well as real-world data evaluations to empirically validate our theory.    (ii)  Failure modes of SERA We show, through theoretical analysis and corroborative synthetic experiments, performance lower bounds when applying SERA to stochastic block models (SBM) with independent random node features: When the underlying SBM has \u0398\u2062(1)\u03981\\Theta(1)roman_\u0398 ( 1 ) intra-group connection probability, edge recovery through graph representations becomes provably hard.    (iii)  Mitigation of SERA We assess the resilience of SERA using noisy aggregation (NAG) as the privacy protection mechanism. Theoretical guarantees of NAG are established which further extends previous results, accompanied by extensive empirical evaluations to corroborate our theoretical assertions. Intriguingly, our findings reveal instances wherein",
            "references": [
                {
                    "title": "Statistical Guarantees for Link Prediction using Graph Neural Networks",
                    "abstract": "This paper derives statistical guarantees for the performance of Graph Neural Networks (GNNs) in link prediction tasks on graphs generated by a graphon. We propose a linear GNN architecture (LG-GNN) that produces consistent estimators for the underlying edge probabilities. We establish a bound on the mean squared error and give guarantees on the ability of LG-GNN to detect high-probability edges. Our guarantees hold for both sparse and dense graphs. Finally, we demonstrate some of the shortcomings of the classical GCN architecture, as well as verify our results on real and synthetic datasets."
                },
                {
                    "title": "Devil in Disguise: Breaching Graph Neural Networks Privacy through Infiltration",
                    "abstract": "Graph neural networks (GNNs) have been developed to mine useful information from graph data of various applications, e.g., healthcare, fraud detection, and social recommendation. However, GNNs open up new attack surfaces for privacy attacks on graph data. In this paper, we propose Infiltrator, a privacy attack that is able to pry node-level private information based on black-box access to GNNs. Different from existing works that require prior information of the victim node, we explore the possibility of conducting the attack without any information of the victim node. Our idea is to infiltrate the graph with attacker-created nodes to befriend the victim node. More specifically, we design infiltration schemes that enable the adversary to infer the label, neighboring links, and sensitive attributes of a victim node. We evaluate Infiltrator with extensive experiments on three representative GNN models and six real-world datasets. The results demonstrate that Infiltrator can achieve an attack performance of more than 98% in all three attacks, outperforming baseline approaches. We further evaluate the defense resistance of Infiltrator against the graph homophily defender and the differentially private model."
                },
                {
                    "title": "Privacy-preserving design of graph neural networks with applications to vertical federated learning",
                    "abstract": "The paradigm of vertical federated learning (VFL), where institutions collaboratively train machine learning models via combining each other's local feature or label information, has achieved great success in applications to financial risk management (FRM). The surging developments of graph representation learning (GRL) have opened up new opportunities for FRM applications under FL via efficiently utilizing the graph-structured data generated from underlying transaction networks. Meanwhile, transaction information is often considered highly sensitive. To prevent data leakage during training, it is critical to develop FL protocols with formal privacy guarantees. In this paper, we present an end-to-end GRL framework in the VFL setting called VESPER, which is built upon a general privatization scheme termed perturbed message passing (PMP) that allows the privatization of many popular graph neural architectures.Based on PMP, we discuss the strengths and weaknesses of specific design choices of concrete graph neural architectures and provide solutions and improvements for both dense and sparse graphs. Extensive empirical evaluations over both public datasets and an industry dataset demonstrate that VESPER is capable of training high-performance GNN models over both sparse and dense graphs under reasonable privacy budgets."
                },
                {
                    "title": "Differentially Private Decoupled Graph Convolutions for Multigranular Topology Protection",
                    "abstract": "GNNs can inadvertently expose sensitive user information and interactions through their model predictions. To address these privacy concerns, Differential Privacy (DP) protocols are employed to control the trade-off between provable privacy protection and model utility. Applying standard DP approaches to GNNs directly is not advisable due to two main reasons. First, the prediction of node labels, which relies on neighboring node attributes through graph convolutions, can lead to privacy leakage. Second, in practical applications, the privacy requirements for node attributes and graph topology may differ. In the latter setting, existing DP-GNN models fail to provide multigranular trade-offs between graph topology privacy, node attribute privacy, and GNN utility. To address both limitations, we propose a new framework termed Graph Differential Privacy (GDP), specifically tailored to graph learning. GDP ensures both provably private model parameters as well as private predictions. Additionally, we describe a novel unified notion of graph dataset adjacency to analyze the properties of GDP for different levels of graph topology privacy. Our findings reveal that DP-GNNs, which rely on graph convolutions, not only fail to meet the requirements for multigranular graph topology privacy but also necessitate the injection of DP noise that scales at least linearly with the maximum node degree. In contrast, our proposed Differentially Private Decoupled Graph Convolutions (DPDGCs) represent a more flexible and efficient alternative to graph convolutions that still provides the necessary guarantees of GDP. To validate our approach, we conducted extensive experiments on seven node classification benchmarking and illustrative synthetic datasets. The results demonstrate that DPDGCs significantly outperform existing DP-GNNs in terms of privacy-utility trade-offs."
                },
                {
                    "title": "On Strengthening and Defending Graph Reconstruction Attack with Markov Chain Approximation",
                    "abstract": "Although powerful graph neural networks (GNNs) have boosted numerous real-world applications, the potential privacy risk is still underexplored. To close this gap, we perform the first comprehensive study of graph reconstruction attack that aims to reconstruct the adjacency of nodes. We show that a range of factors in GNNs can lead to the surprising leakage of private links. Especially by taking GNNs as a Markov chain and attacking GNNs via a flexible chain approximation, we systematically explore the underneath principles of graph reconstruction attack, and propose two information theory-guided mechanisms: (1) the chain-based attack method with adaptive designs for extracting more private information; (2) the chain-based defense method that sharply reduces the attack fidelity with moderate accuracy loss. Such two objectives disclose a critical belief that to recover better in attack, you must extract more multi-aspect knowledge from the trained GNN; while to learn safer for defense, you must forget more link-sensitive information in training GNNs. Empirically, we achieve state-of-the-art results on six datasets and three common GNNs. The code is publicly available at: https://github.com/tmlr-group/MC-GRA."
                },
                {
                    "title": "A Non-Asymptotic Analysis of Oversmoothing in Graph Neural Networks",
                    "abstract": "Oversmoothing is a central challenge of building more powerful Graph Neural Networks (GNNs). While previous works have only demonstrated that oversmoothing is inevitable when the number of graph convolutions tends to infinity, in this paper, we precisely characterize the mechanism behind the phenomenon via a non-asymptotic analysis. Specifically, we distinguish between two different effects when applying graph convolutions -- an undesirable mixing effect that homogenizes node representations in different classes, and a desirable denoising effect that homogenizes node representations in the same class. By quantifying these two effects on random graphs sampled from the Contextual Stochastic Block Model (CSBM), we show that oversmoothing happens once the mixing effect starts to dominate the denoising effect, and the number of layers required for this transition is $O(\\log N/\\log (\\log N))$ for sufficiently dense graphs with $N$ nodes. We also extend our analysis to study the effects of Personalized PageRank (PPR), or equivalently, the effects of initial residual connections on oversmoothing. Our results suggest that while PPR mitigates oversmoothing at deeper layers, PPR-based architectures still achieve their best performance at a shallow depth and are outperformed by the graph convolution approach on certain graphs. Finally, we support our theoretical results with numerical experiments, which further suggest that the oversmoothing phenomenon observed in practice can be magnified by the difficulty of optimizing deep GNN models."
                },
                {
                    "title": "Model Inversion Attacks Against Graph Neural Networks",
                    "abstract": "Many data mining tasks rely on graphs to model relational structures among individuals (nodes). Since relational data are often sensitive, there is an urgent need to evaluate the privacy risks in graph data. One famous privacy attack against data analysis models is the model inversion attack, which aims to infer sensitive data in the training dataset and leads to great privacy concerns. Despite its success in grid-like domains, directly applying model inversion attacks on non-grid domains such as graph leads to poor attack performance. This is mainly due to the failure to consider the unique properties of graphs. To bridge this gap, we conduct a systematic study on model inversion attacks against Graph Neural Networks (GNNs), one of the state-of-the-art graph analysis tools in this paper. First, in the white-box setting where the attacker has full access to the target GNN model, we present GraphMI to infer the private training graph data. Specifically in GraphMI, a projected gradient module is proposed to tackle the discreteness of graph edges and preserve the sparsity and smoothness of graph features; a graph auto-encoder module is used to efficiently exploit graph topology, node attributes, and target model parameters for edge inference; a random sampling module can finally sample discrete edges. Furthermore, in the hard-label black-box setting where the attacker can only query the GNN API and receive the classification results, we propose two methods based on gradient estimation and reinforcement learning (RL-GraphMI). With the proposed methods, we study the connection between model inversion risk and edge influence and show that edges with greater influence are more likely to be recovered. Extensive experiments over several public datasets demonstrate the effectiveness of our methods. We also evaluate our attacks under two defenses: one is the well-designed differential private training, and the other is graph preprocessing. Our experimental results show that such defenses are not sufficiently effective and call for more advanced defenses against privacy attacks."
                },
                {
                    "title": "RAW-GNN: RAndom Walk Aggregation based Graph Neural Network",
                    "abstract": "Graph-Convolution-based methods have been successfully applied to representation learning on homophily graphs where nodes with the same label or similar attributes tend to connect with one another. Due to the homophily assumption of Graph Convolutional Networks (GCNs) that these methods use, they are not suitable for heterophily graphs where nodes with different labels or dissimilar attributes tend to be adjacent. Several methods have attempted to address this heterophily problem, but they do not change the fundamental aggregation mechanism of GCNs because they rely on summation operators to aggregate information from neighboring nodes, which is implicitly subject to the homophily assumption. Here, we introduce a novel aggregation mechanism and develop a RAndom Walk Aggregation-based Graph Neural Network (called RAW-GNN) method. The proposed approach integrates the random walk strategy with graph neural networks. The new method utilizes breadth-first random walk search to capture homophily information and depth-first search to collect heterophily information. It replaces the conventional neighborhoods with path-based neighborhoods and introduces a new path-based aggregator based on Recurrent Neural Networks. These designs make RAW-GNN suitable for both homophily and heterophily graphs. Extensive experimental results showed that the new method achieved state-of-the-art performance on a variety of homophily and heterophily graphs."
                },
                {
                    "title": "A Comprehensive Survey on Trustworthy Graph Neural Networks: Privacy, Robustness, Fairness, and Explainability",
                    "abstract": "Graph neural networks (GNNs) have made rapid developments in the recent years. Due to their great ability in modeling graph-structured data, GNNs are vastly used in various applications, including high-stakes scenarios such as financial analysis, traffic predictions, and drug discovery. Despite their great potential in benefiting humans in the real world, recent study shows that GNNs can leak private information, are vulnerable to adversarial attacks, can inherit and magnify societal bias from training data and lack interpretability, which have risk of causing unintentional harm to the users and society. For example, existing works demonstrate that attackers can fool the GNNs to give the outcome they desire with unnoticeable perturbation on training graph. GNNs trained on social networks may embed the discrimination in their decision process, strengthening the undesirable societal bias. Consequently, trust-worthy GNNs in various aspects are emerging to prevent the harm from GNN models and increase the users\u2019 trust in GNNs. In this paper, we give a comprehensive survey of GNNs in the computational aspects of privacy, robustness, fairness, and explainability. For each aspect, we give the taxonomy of the related methods and formulate the general frameworks for the multiple categories of trustworthy GNNs. We also discuss the future research directions of each aspect and connections between these aspects to help achieve trustworthiness."
                },
                {
                    "title": "GAP: Differentially Private Graph Neural Networks with Aggregation Perturbation",
                    "abstract": "In this paper, we study the problem of learning Graph Neural Networks (GNNs) with Differential Privacy (DP). We propose a novel differentially private GNN based on Aggregation Perturbation (GAP), which adds stochastic noise to the GNN's aggregation function to statistically obfuscate the presence of a single edge (edge-level privacy) or a single node and all its adjacent edges (node-level privacy). Tailored to the specifics of private learning, GAP's new architecture is composed of three separate modules: (i) the encoder module, where we learn private node embeddings without relying on the edge information; (ii) the aggregation module, where we compute noisy aggregated node embeddings based on the graph structure; and (iii) the classification module, where we train a neural network on the private aggregations for node classification without further querying the graph edges. GAP's major advantage over previous approaches is that it can benefit from multi-hop neighborhood aggregations, and guarantees both edge-level and node-level DP not only for training, but also at inference with no additional costs beyond the training's privacy budget. We analyze GAP's formal privacy guarantees using R\\'enyi DP and conduct empirical experiments over three real-world graph datasets. We demonstrate that GAP offers significantly better accuracy-privacy trade-offs than state-of-the-art DP-GNN approaches and naive MLP-based baselines. Our code is publicly available at https://github.com/sisaman/GAP."
                },
                {
                    "title": "Debugging Differential Privacy: A Case Study for Privacy Auditing",
                    "abstract": "Differential Privacy can provide provable privacy guarantees for training data in machine learning. However, the presence of proofs does not preclude the presence of errors. Inspired by recent advances in auditing which have been used for estimating lower bounds on differentially private algorithms, here we show that auditing can also be used to find flaws in (purportedly) differentially private schemes. In this case study, we audit a recent open source implementation of a differentially private deep learning algorithm and find, with 99.99999999% confidence, that the implementation does not satisfy the claimed differential privacy guarantee."
                },
                {
                    "title": "A short note on an inequality between KL and TV",
                    "abstract": "The goal of this short note is to discuss the relation between Kullback--Leibler divergence and total variation distance, starting with the celebrated Pinsker's inequality relating the two, before switching to a simple, yet (arguably) more useful inequality, apparently not as well known, due to Bretagnolle and Huber. We also discuss applications of this bound for minimax testing lower bounds."
                },
                {
                    "title": "LINKTELLER: Recovering Private Edges from Graph Neural Networks via Influence Analysis",
                    "abstract": "Graph structured data have enabled several successful applications such as recommendation systems and traffic prediction, given the rich node features and edges information. However, these high-dimensional features and high-order adjacency information are usually heterogeneous and held by different data holders in practice. Given such vertical data partition e.g., one data holder will only own either the node features or edge information), different data holders have to develop efficient joint training protocols rather than directly transferring data to each other due to privacy concerns. In this paper, we focus on the edge privacy, and consider a training scenario where the data holder Bob with node features will first send training node features to Alice who owns the adjacency information. Alice will then train a graph neural network (GNN) with the joint information and provide an inference API to Bob. During inference time, Bob is able to provide test node features and query the API to obtain the predictions for test nodes. Under this setting, we first propose a privacy attack LINKTELLER via influence analysis to infer the private edge information held by Alice via designing adversarial queries for Bob. We then empirically show that LINKTELLER is able to recover a significant amount of private edges in different settings, both including inductive (8 datasets) and transductive (3 datasets), under different graph densities, significantly outperforming existing baselines. To further evaluate the privacy leakage for edges, we adapt an existing algorithm for differentially private graph convolutional network (DP GCN) training as well as propose a new DP GCN mechanism LAPGRAPH based on Laplacian mechanism to evaluate LINKTELLER. We show that these DP GCN mechanisms are not always resilient against LINKTELLER empirically under mild privacy guarantees $(\\varepsilon\\gt 5)$. Our studies will shed light on future research towards designing more resilient privacy-preserving GCN models; in the meantime, provide an in-depth understanding about the tradeoff between GCN model utility and robustness against potential privacy attacks."
                },
                {
                    "title": "Privacy-Preserving Representation Learning on Graphs: A Mutual Information Perspective",
                    "abstract": "Learning with graphs has attracted significant attention recently. Existing representation learning methods on graphs have achieved state-of-the-art performance on various graph-related tasks such as node classification, link prediction, etc. However, we observe that these methods could leak serious private information. For instance, one can accurately infer the links (or node identity) in a graph from a node classifier (or link predictor) trained on the learnt node representations by existing methods. To address the issue, we propose a privacy-preserving representation learning framework on graphs from the mutual information perspective. Specifically, our framework includes a primary learning task and a privacy protection task, and we consider node classification and link prediction as the two tasks of interest. Our goal is to learn node representations such that they can be used to achieve high performance for the primary learning task, while obtaining performance for the privacy protection task close to random guessing. We formally formulate our goal via mutual information objectives. However, it is intractable to compute mutual information in practice. Then, we derive tractable variational bounds for the mutual information terms, where each bound can be parameterized via a neural network. Next, we train these parameterized neural networks to approximate the true mutual information and learn privacy-preserving node representations. We finally evaluate our framework on various graph datasets."
                },
                {
                    "title": "GraphMI: Extracting Private Graph Data from Graph Neural Networks",
                    "abstract": "As machine learning becomes more widely used for critical applications, the need to study its implications in privacy becomes urgent.\n\nGiven access to the target model and auxiliary information, model inversion attack aims to infer sensitive features of the training dataset, which leads to great privacy concerns.\n\nDespite its success in the grid domain, directly applying model inversion techniques on non grid domains such as graph achieves poor attack performance due to the difficulty to fully exploit the intrinsic properties of graphs and attributes of graph nodes used in GNN models.\n\nTo bridge this gap, we present Graph Model Inversion attack, which aims to infer edges of the training graph by inverting Graph Neural Networks, one of the most popular graph analysis tools.\n\nSpecifically, the projected gradient module in our method can tackle the discreteness of graph edges while preserving the sparsity and smoothness of graph features.\n\nMoreover, a well designed graph autoencoder module can efficiently exploit graph topology, node attributes, and target model parameters.\n\nWith the proposed method, we study the connection between model inversion risk and edge influence and show that edges with greater influence are more likely to be recovered.\n\nExtensive experiments over several public datasets demonstrate the effectiveness of our method.\n\nWe also show that differential privacy in its canonical form can hardly defend our attack while preserving decent utility."
                },
                {
                    "title": "Optimization of Graph Neural Networks: Implicit Acceleration by Skip Connections and More Depth",
                    "abstract": "Graph Neural Networks (GNNs) have been studied from the lens of expressive power and generalization. However, their optimization properties are less well understood. We take the first step towards analyzing GNN training by studying the gradient dynamics of GNNs. First, we analyze linearized GNNs and prove that despite the non-convexity of training, convergence to a global minimum at a linear rate is guaranteed under mild assumptions that we validate on real-world graphs. Second, we study what may affect the GNNs' training speed. Our results show that the training of GNNs is implicitly accelerated by skip connections, more depth, and/or a good label distribution. Empirical results confirm that our theoretical results for linearized GNNs align with the training behavior of nonlinear GNNs. Our results provide the first theoretical support for the success of GNNs with skip connections in terms of optimization, and suggest that deep GNNs with skip connections would be promising in practice."
                },
                {
                    "title": "DeepWalking Backwards: From Embeddings Back to Graphs",
                    "abstract": "Low-dimensional node embeddings play a key role in analyzing graph datasets. However, little work studies exactly what information is encoded by popular embedding methods, and how this information correlates with performance in downstream machine learning tasks. We tackle this question by studying whether embeddings can be inverted to (approximately) recover the graph used to generate them. Focusing on a variant of the popular DeepWalk method (Perozzi et al., 2014; Qiu et al., 2018), we present algorithms for accurate embedding inversion - i.e., from the low-dimensional embedding of a graph G, we can find a graph H with a very similar embedding. We perform numerous experiments on real-world networks, observing that significant information about G, such as specific edges and bulk properties like triangle density, is often lost in H. However, community structure is often preserved or even enhanced. Our findings are a step towards a more rigorous understanding of exactly what information embeddings encode about the input graph, and why this information is useful for learning tasks."
                },
                {
                    "title": "Adversary Instantiation: Lower Bounds for Differentially Private Machine Learning",
                    "abstract": "Differentially private (DP) machine learning allows us to train models on private data while limiting data leakage. DP formalizes this data leakage through a cryptographic game, where an adversary must predict if a model was trained on a dataset D, or a dataset D\u2032 that differs in just one example. If observing the training algorithm does not meaningfully increase the adversary's odds of successfully guessing which dataset the model was trained on, then the algorithm is said to be differentially private. Hence, the purpose of privacy analysis is to upper bound the probability that any adversary could successfully guess which dataset the model was trained on.In our paper, we instantiate this hypothetical adversary in order to establish lower bounds on the probability that this distinguishing game can be won. We use this adversary to evaluate the importance of the adversary capabilities allowed in the privacy analysis of DP training algorithms.For DP-SGD, the most common method for training neural networks with differential privacy, our lower bounds are tight and match the theoretical upper bound. This implies that in order to prove better upper bounds, it will be necessary to make use of additional assumptions. Fortunately, we find that our attacks are significantly weaker when additional (realistic) restrictions are put in place on the adversary's capabilities. Thus, in the practical setting common to many real-world deployments, there is a gap between our lower bounds and the upper bounds provided by the analysis: differential privacy is conservative and adversaries may not be able to leak as much information as suggested by the theoretical bound."
                },
                {
                    "title": "Quantifying Privacy Leakage in Graph Embedding",
                    "abstract": "Graph embeddings have been proposed to map graph data to low dimensional space for downstream processing (e.g., node classification or link prediction). With the increasing collection of personal data, graph embeddings can be trained on private and sensitive data. For the first time, we quantify the privacy leakage in graph embeddings through three inference attacks targeting Graph Neural Networks. Our membership inference attack aims to infer whether a graph node corresponding to an individual user\u2019s data was a member of the model\u2019s private training data or not. We consider a blackbox setting where the adversary exploits the output prediction scores and a whitebox setting where the adversary has also access to the released node embeddings. Our attack provides accuracy up to 28% (blackbox) and 36% (whitebox) beyond the random guess by exploiting the distinguishable footprint between train and test data records left by the graph embedding. In our graph reconstruction attack, the adversary aims to reconstruct the target graph given the corresponding graph embeddings. Here, the adversary can reconstruct the graph with more than 80% of accuracy and infer the link between two nodes with \u223c 30% more accuracy than the random guess. Finally, we propose an attribute inference attack where the adversary aims to infer the sensitive node attributes corresponding to an individual user. We show that the strong correlation between the graph embeddings and node attributes allows the adversary to infer sensitive information (e.g., gender or location)."
                },
                {
                    "title": "Vertically Federated Graph Neural Network for Privacy-Preserving Node Classification",
                    "abstract": "Recently, Graph Neural Network (GNN) has achieved remarkable progresses in various real-world tasks on graph data, consisting of node features and the adjacent information between different nodes. High-performance GNN models always depend on both rich features and complete edge information in graph. However, such information could possibly be isolated by different data holders in practice, which is the so-called data isolation problem. To solve this problem, in this paper, we propose VFGNN, a federated GNN learning paradigm for privacy-preserving node classification task under data vertically partitioned setting, which can be generalized to existing GNN models. Specifically, we split the computation graph into two parts. We leave the private data (i.e., features, edges, and labels) related computations on data holders, and delegate the rest of computations to a semi-honest server. We also propose to apply differential privacy to prevent potential information leakage from the server. We conduct experiments on three benchmarks and the results demonstrate the effectiveness of VFGNN."
                },
                {
                    "title": "Stealing Links from Graph Neural Networks",
                    "abstract": "Graph data, such as social networks and chemical networks, contains a wealth of information that can help to build powerful applications. To fully unleash the power of graph data, a family of machine learning models, namely graph neural networks (GNNs), is introduced. Empirical results show that GNNs have achieved state-of-the-art performance in various tasks. \nGraph data is the key to the success of GNNs. High-quality graph is expensive to collect and often contains sensitive information, such as social relations. Various research has shown that machine learning models are vulnerable to attacks against their training data. Most of these models focus on data from the Euclidean space, such as images and texts. Meanwhile, little attention has been paid to the security and privacy risks of graph data used to train GNNs. \nIn this paper, we aim at filling the gap by proposing the first link stealing attacks against graph neural networks. Given a black-box access to a GNN model, the goal of an adversary is to infer whether there exists a link between any pair of nodes in the graph used to train the model. We propose a threat model to systematically characterize the adversary's background knowledge along three dimensions. By combination, we obtain a comprehensive taxonomy of 8 different link stealing attacks. We propose multiple novel methods to realize these attacks. Extensive experiments over 8 real-world datasets show that our attacks are effective at inferring links, e.g., AUC (area under the ROC curve) is above 0.95 in multiple cases."
                },
                {
                    "title": "A Closer Look at Accuracy vs. Robustness",
                    "abstract": "Current methods for training robust networks lead to a drop in test accuracy, which has led prior works to posit that a robustness-accuracy tradeoff may be inevitable in deep learning. We take a closer look at this phenomenon and first show that real image datasets are actually separated. With this property in mind, we then prove that robustness and accuracy should both be achievable for benchmark datasets through locally Lipschitz functions, and hence, there should be no inherent tradeoff between robustness and accuracy. Through extensive experiments with robustness methods, we argue that the gap between theory and practice arises from two limitations of current methods: either they fail to impose local Lipschitzness or they are insufficiently generalized. We explore combining dropout with robust training methods and obtain better generalization. We conclude that achieving robustness and accuracy in practice may require using methods that impose local Lipschitzness and augmenting them with deep learning generalization techniques. Code available at this https URL"
                },
                {
                    "title": "Geom-GCN: Geometric Graph Convolutional Networks",
                    "abstract": "Message-passing neural networks (MPNNs) have been successfully applied in a wide variety of applications in the real world. However, two fundamental weaknesses of MPNNs' aggregators limit their ability to represent graph-structured data: losing the structural information of nodes in neighborhoods and lacking the ability to capture long-range dependencies in disassortative graphs. Few studies have noticed the weaknesses from different perspectives. From the observations on classical neural network and network geometry, we propose a novel geometric aggregation scheme for graph neural networks to overcome the two weaknesses. The behind basic idea is the aggregation on a graph can benefit from a continuous space underlying the graph. The proposed aggregation scheme is permutation-invariant and consists of three modules, node embedding, structural neighborhood, and bi-level aggregation. We also present an implementation of the scheme in graph convolutional networks, termed Geom-GCN, to perform transductive learning on graphs. Experimental results show the proposed Geom-GCN achieved state-of-the-art performance on a wide range of open datasets of graphs."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "GraphSAINT: Graph Sampling Based Inductive Learning Method",
                    "abstract": "Graph Convolutional Networks (GCNs) are powerful models for learning representations of attributed this http URL scale GCNs to large graphs, state-of-the-art methods use various layer sampling techniques to alleviate the \"neighbor explosion\" problem during minibatch training. Here we proposeGraphSAINT, a graph sampling based inductive learning method that improves training efficiency in a fundamentally different way. By a change of perspective, GraphSAINT constructs minibatches by sampling the training graph, rather than the nodes or edges across GCN layers. Each iteration, a complete GCN is built from the properly sampled subgraph. Thus, we ensure fixed number of well-connected nodes in all layers. We further propose normalization technique to eliminate bias, and sampling algorithms for variance reduction. Importantly, we can decouple the sampling process from the forward and backward propagation of training, and extend GraphSAINT with other graph samplers and GCN variants. Comparing with strong baselines using layer sampling, GraphSAINT demonstrates superior performance in both accuracy and training time on four large graphs."
                },
                {
                    "title": "Graph Neural Networks Exponentially Lose Expressive Power for Node Classification",
                    "abstract": "Graph Neural Networks (graph NNs) are a promising deep learning approach for analyzing graph-structured data. However, it is known that they do not improve (or sometimes worsen) their predictive performance as we pile up many layers and add non-lineality. To tackle this problem, we investigate the expressive power of graph NNs via their asymptotic behaviors as the layer size tends to infinity. Our strategy is to generalize the forward propagation of a Graph Convolutional Network (GCN), which is a popular graph NN variant, as a specific dynamical system. In the case of a GCN, we show that when its weights satisfy the conditions determined by the spectra of the (augmented) normalized Laplacian, its output exponentially approaches the set of signals that carry information of the connected components and node degrees only for distinguishing nodes. Our theory enables us to relate the expressive power of GCNs with the topological information of the underlying graphs inherent in the graph spectra. To demonstrate this, we characterize the asymptotic behavior of GCNs on the Erd\u0151s -- Renyi graph. We show that when the Erd\u0151s -- Renyi graph is sufficiently dense and large, a broad range of GCNs on it suffers from the \"information loss\" in the limit of infinite layers with high probability. Based on the theory, we provide a principled guideline for weight normalization of graph NNs. We experimentally confirm that the proposed weight scaling enhances the predictive performance of GCNs in real data. Code is available at this https URL."
                },
                {
                    "title": "Fast Graph Representation Learning with PyTorch Geometric",
                    "abstract": "We introduce PyTorch Geometric, a library for deep learning on irregularly structured input data such as graphs, point clouds and manifolds, built upon PyTorch. In addition to general graph data structures and processing methods, it contains a variety of recently published methods from the domains of relational learning and 3D data processing. PyTorch Geometric achieves high data throughput by leveraging sparse GPU acceleration, by providing dedicated CUDA kernels and by introducing efficient mini-batch handling for input examples of different size. In this work, we present the library in detail and perform a comprehensive comparative study of the implemented methods in homogeneous evaluation scenarios."
                },
                {
                    "title": "Simplifying Graph Convolutional Networks",
                    "abstract": "Graph Convolutional Networks (GCNs) and their variants have experienced significant attention and have become the de facto methods for learning graph representations. GCNs derive inspiration primarily from recent deep learning approaches, and as a result, may inherit unnecessary complexity and redundant computation. In this paper, we reduce this excess complexity through successively removing nonlinearities and collapsing weight matrices between consecutive layers. We theoretically analyze the resulting linear model and show that it corresponds to a fixed low-pass filter followed by a linear classifier. Notably, our experimental evaluation demonstrates that these simplifications do not negatively impact accuracy in many downstream applications. Moreover, the resulting model scales to larger datasets, is naturally interpretable, and yields up to two orders of magnitude speedup over FastGCN."
                },
                {
                    "title": "How Powerful are Graph Neural Networks?",
                    "abstract": "Graph Neural Networks (GNNs) are an effective framework for representation learning of graphs. GNNs follow a neighborhood aggregation scheme, where the representation vector of a node is computed by recursively aggregating and transforming representation vectors of its neighboring nodes. Many GNN variants have been proposed and have achieved state-of-the-art results on both node and graph classification tasks. However, despite GNNs revolutionizing graph representation learning, there is limited understanding of their representational properties and limitations. Here, we present a theoretical framework for analyzing the expressive power of GNNs to capture different graph structures. Our results characterize the discriminative power of popular GNN variants, such as Graph Convolutional Networks and GraphSAGE, and show that they cannot learn to distinguish certain simple graph structures. We then develop a simple architecture that is provably the most expressive among the class of GNNs and is as powerful as the Weisfeiler-Lehman graph isomorphism test. We empirically validate our theoretical findings on a number of graph classification benchmarks, and demonstrate that our model achieves state-of-the-art performance."
                },
                {
                    "title": "The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks",
                    "abstract": "This paper describes a testing methodology for quantitatively assessing the risk that rare or unique training-data sequences are unintentionally memorized by generative sequence models---a common type of machine-learning model. Because such models are sometimes trained on sensitive data (e.g., the text of users' private messages), this methodology can benefit privacy by allowing deep-learning practitioners to select means of training that minimize such memorization. \nIn experiments, we show that unintended memorization is a persistent, hard-to-avoid issue that can have serious consequences. Specifically, for models trained without consideration of memorization, we describe new, efficient procedures that can extract unique, secret sequences, such as credit card numbers. We show that our testing strategy is a practical and easy-to-use first line of defense, e.g., by describing its application to quantitatively limit data exposure in Google's Smart Compose, a commercial text-completion neural network trained on millions of users' email messages."
                },
                {
                    "title": "Spectral Normalization for Generative Adversarial Networks",
                    "abstract": "One of the challenges in the study of generative adversarial networks is the instability of its training. In this paper, we propose a novel weight normalization technique called spectral normalization to stabilize the training of the discriminator. Our new normalization technique is computationally light and easy to incorporate into existing implementations. We tested the efficacy of spectral normalization on CIFAR10, STL-10, and ILSVRC2012 dataset, and we experimentally confirmed that spectrally normalized GANs (SN-GANs) is capable of generating images of better or equal quality relative to the previous training stabilization techniques."
                },
                {
                    "title": "Graph Attention Networks",
                    "abstract": "We present graph attention networks (GATs), novel neural network architectures that operate on graph-structured data, leveraging masked self-attentional layers to address the shortcomings of prior methods based on graph convolutions or their approximations. By stacking layers in which nodes are able to attend over their neighborhoods' features, we enable (implicitly) specifying different weights to different nodes in a neighborhood, without requiring any kind of costly matrix operation (such as inversion) or depending on knowing the graph structure upfront. In this way, we address several key challenges of spectral-based graph neural networks simultaneously, and make our model readily applicable to inductive as well as transductive problems. Our GAT models have achieved or matched state-of-the-art results across four established transductive and inductive graph benchmarks: the Cora, Citeseer and Pubmed citation network datasets, as well as a protein-protein interaction dataset (wherein test graphs remain unseen during training)."
                },
                {
                    "title": "Representation Learning on Graphs: Methods and Applications",
                    "abstract": "Machine learning on graphs is an important and ubiquitous task with applications ranging from drug design to friendship recommendation in social networks. The primary challenge in this domain is finding a way to represent, or encode, graph structure so that it can be easily exploited by machine learning models. Traditionally, machine learning approaches relied on user-defined heuristics to extract features encoding structural information about a graph (e.g., degree statistics or kernel functions). However, recent years have seen a surge in approaches that automatically learn to encode graph structure into low-dimensional embeddings, using techniques based on deep learning and nonlinear dimensionality reduction. Here we provide a conceptual review of key advancements in this area of representation learning on graphs, including matrix factorization-based methods, random-walk based algorithms, and graph neural networks. We review methods to embed individual nodes as well as approaches to embed entire (sub)graphs. In doing so, we develop a unified framework to describe these recent approaches, and we highlight a number of important applications and directions for future work."
                },
                {
                    "title": "Inductive Representation Learning on Large Graphs",
                    "abstract": "Low-dimensional embeddings of nodes in large graphs have proved extremely useful in a variety of prediction tasks, from content recommendation to identifying protein functions. However, most existing approaches require that all nodes in the graph are present during training of the embeddings; these previous approaches are inherently transductive and do not naturally generalize to unseen nodes. Here we present GraphSAGE, a general, inductive framework that leverages node feature information (e.g., text attributes) to efficiently generate node embeddings for previously unseen data. Instead of training individual embeddings for each node, we learn a function that generates embeddings by sampling and aggregating features from a node's local neighborhood. Our algorithm outperforms strong baselines on three inductive node-classification benchmarks: we classify the category of unseen nodes in evolving information graphs based on citation and Reddit post data, and we show that our algorithm generalizes to completely unseen graphs using a multi-graph dataset of protein-protein interactions."
                },
                {
                    "title": "Community detection and stochastic block models: recent developments",
                    "abstract": "The stochastic block model (SBM) is a random graph model with planted clusters. It is widely employed as a canonical model to study clustering and community detection, and provides generally a fertile ground to study the statistical and computational tradeoffs that arise in network and data sciences. \n \nThis note surveys the recent developments that establish the fundamental limits for community detection in the SBM, both with respect to information-theoretic and computational thresholds, and for various recovery requirements such as exact, partial and weak recovery (a.k.a., detection). The main results discussed are the phase transitions for exact recovery at the Chernoff-Hellinger threshold, the phase transition for weak recovery at the Kesten-Stigum threshold, the optimal distortion-SNR tradeoff for partial recovery, the learning of the SBM parameters and the gap between information-theoretic and computational thresholds. \n \nThe note also covers some of the algorithms developed in the quest of achieving the limits, in particular two-round algorithms via graph-splitting, semi-definite programming, linearized belief propagation, classical and nonbacktracking spectral methods. A few open problems are also discussed."
                },
                {
                    "title": "R\u00e9nyi Differential Privacy",
                    "abstract": "We propose a natural relaxation of differential privacy based on the R\u00e9nyi divergence. Closely related notions have appeared in several recent papers that analyzed composition of differentially private mechanisms. We argue that the useful analytical tool can be used as a privacy definition, compactly and accurately representing guarantees on the tails of the privacy loss.We demonstrate that the new definition shares many important properties with the standard definition of differential privacy, while additionally allowing tighter analysis of composite heterogeneous mechanisms."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Deep Learning with Differential Privacy",
                    "abstract": "Machine learning techniques based on neural networks are achieving remarkable results in a wide variety of domains. Often, the training of models requires large, representative datasets, which may be crowdsourced and contain sensitive information. The models should not expose private information in these datasets. Addressing this goal, we develop new algorithmic techniques for learning and a refined analysis of privacy costs within the framework of differential privacy. Our implementation and experiments demonstrate that we can train deep neural networks with non-convex objectives, under a modest privacy budget, and at a manageable cost in software complexity, training efficiency, and model quality."
                },
                {
                    "title": "Revisiting Semi-Supervised Learning with Graph Embeddings",
                    "abstract": "We present a semi-supervised learning framework based on graph embeddings. Given a graph between instances, we train an embedding for each instance to jointly predict the class label and the neighborhood context in the graph. We develop both transductive and inductive variants of our method. In the transductive variant of our method, the class labels are determined by both the learned embeddings and input feature vectors, while in the inductive variant, the embeddings are defined as a parametric function of the feature vectors, so predictions can be made on instances not seen during training. On a large and diverse set of benchmark tasks, including text classification, distantly supervised entity extraction, and entity classification, we show improved performance over many of the existing models."
                },
                {
                    "title": "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification",
                    "abstract": "Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on the learnable activation and advanced initialization, we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66% [33]). To our knowledge, our result is the first to surpass the reported human-level performance (5.1%, [26]) on this dataset."
                },
                {
                    "title": "The Algorithmic Foundations of Differential Privacy",
                    "abstract": "The problem of privacy-preserving data analysis has a long history spanning multiple disciplines. As electronic data about individuals becomes increasingly detailed, and as technology enables ever more powerful collection and curation of these data, the need increases for a robust, meaningful, and mathematically rigorous definition of privacy, together with a computationally rich class of algorithms that satisfy this definition. Differential Privacy is such a definition.After motivating and discussing the meaning of differential privacy, the preponderance of this monograph is devoted to fundamental techniques for achieving differential privacy, and application of these techniques in creative combinations, using the query-release problem as an ongoing example. A key point is that, by rethinking the computational goal, one can often obtain far better results than would be achieved by methodically replacing each step of a non-private computation with a differentially private implementation. Despite some astonishingly powerful computational results, there are still fundamental limitations \u2014 not just on what can be achieved with differential privacy but on what can be achieved with any method that protects against a complete breakdown in privacy. Virtually all the algorithms discussed herein maintain differential privacy against adversaries of arbitrary computational power. Certain algorithms are computationally intensive, others are efficient. Computational complexity for the adversary and the algorithm are both discussed.We then turn from fundamentals to applications other than queryrelease, discussing differentially private methods for mechanism design and machine learning. The vast majority of the literature on differentially private algorithms considers a single, static, database that is subject to many analyses. Differential privacy in other models, including distributed databases and computations on data streams is discussed.Finally, we note that this work is meant as a thorough introduction to the problems and techniques of differential privacy, but is not intended to be an exhaustive survey \u2014 there is by now a vast amount of work in differential privacy, and we can cover only a small portion of it."
                },
                {
                    "title": "What Can We Learn Privately?",
                    "abstract": "Learning problems form an important category of computational tasks that generalizes many of the computations researchers apply to large real-life data sets. We ask: what concept classes can be learned privately, namely, by an algorithm whose output does not depend too heavily on any one input or specific training example? More precisely, we investigate learning algorithms that satisfy differential privacy, a notion that provides strong confidentiality guarantees in the contexts where aggregate information is released about a database containing sensitive information about individuals. We present several basic results that demonstrate general feasibility of private learning and relate several models previously studied separately in the contexts of privacy and standard learning."
                },
                {
                    "title": "Smooth sensitivity and sampling in private data analysis",
                    "abstract": "We introduce a new, generic framework for private data analysis.The goal of private data analysis is to release aggregate information about a data set while protecting the privacy of the individuals whose information the data set contains.Our framework allows one to release functions f of the data withinstance-based additive noise. That is, the noise magnitude is determined not only by the function we want to release, but also bythe database itself. One of the challenges is to ensure that the noise magnitude does not leak information about the database. To address that, we calibrate the noise magnitude to the smoothsensitivity of f on the database x --- a measure of variabilityof f in the neighborhood of the instance x. The new frameworkgreatly expands the applicability of output perturbation, a technique for protecting individuals' privacy by adding a smallamount of random noise to the released statistics. To our knowledge, this is the first formal analysis of the effect of instance-basednoise in the context of data privacy.\n Our framework raises many interesting algorithmic questions. Namely,to apply the framework one must compute or approximate the smoothsensitivity of f on x. We show how to do this efficiently for several different functions, including the median and the cost ofthe minimum spanning tree. We also give a generic procedure based on sampling that allows one to release f(x) accurately on manydatabases x. This procedure is applicable even when no efficient algorithm for approximating smooth sensitivity of f is known orwhen f is given as a black box. We illustrate the procedure by applying it to k-SED (k-means) clustering and learning mixtures of Gaussians."
                },
                {
                    "title": "When Do Graph Neural Networks Help with Node Classification: Investigating the Homophily Principle on Node Distinguishability",
                    "abstract": "Homophily principle, i.e., nodes with the same labels are more likely to be connected, was believed to be the main reason for the performance superiority of Graph Neural Networks (GNNs) over Neural Networks (NNs) on Node Classi\ufb01cation (NC) tasks. Recently, people have developed theoretical results arguing that, even though the homophily principle is broken, the advantage of GNNs can still hold as long as nodes from the same class share similar neighborhood patterns [29], which questions the validity of homophily. However, this argument only considers intra-class Node Distinguishability (ND) and ignores inter-class ND, which is insuf\ufb01cient to study the effect of homophily. In this paper, we \ufb01rst demonstrate the aforementioned insuf\ufb01ciency with examples and argue that an ideal situation for ND is to have smaller intra-class ND than inter-class ND. To formulate this idea and have a better understanding of homophily, we propose Contextual Stochastic Block Model for Homophily (CSBM-H) and de\ufb01ne two metrics, Probabilistic Bayes Error (PBE) and Expected Negative KL-divergence (ENKL), to quantify ND, through which we can also \ufb01nd how intra- and inter-class ND in\ufb02uence ND together. We visualize the results and give detailed analysis. Through experiments, we veri\ufb01ed that the superiority of GNNs is"
                },
                {
                    "title": "Demystifying Uneven Vulnerability of Link Stealing Attacks against Graph Neural Networks",
                    "abstract": "While graph neural networks (GNNs) dominate the state-of-the-art for exploring graphs in real-world applications, they have been shown to be vulnerable to a growing number of privacy attacks. For instance, link stealing is a well-known membership inference attack (MIA) on edges that infers the presence of an edge in a GNN\u2019s training graph. Recent studies on independent and identically distributed data (e.g., images) have empirically demonstrated that individuals from different groups suffer from different levels of privacy risks to MIAs, i.e., uneven vulnerability. However, theoretical evidence for such uneven vulnerability is missing. In this paper, we first present theoretical evidence of the uneven vulnerability of GNNs to link stealing attacks, which lays the foundation for demystifying such uneven risks among different groups of edges. We further demonstrate a group-based attack paradigm to expose the practical privacy harm to GNN users derived from the uneven vulnerability of edges. Finally, we empirically validate the existence of obvious uneven vulnerability on ten real-world datasets (e.g., about 25% AUC difference between different groups in the Credit graph). Compared with existing methods, the outperformance of our group-based attack paradigm confirms that cus-tomising different strategies for different groups results in more effective privacy attacks."
                },
                {
                    "title": "Challenges towards the Next Frontier in Privacy",
                    "abstract": ","
                },
                {
                    "title": "A Convergence Analysis of Gradient Descent on Graph Neural Networks",
                    "abstract": "Graph Neural Networks (GNNs) are a powerful class of architectures for solving learning problems on graphs. While many variants of GNNs have been proposed in the literature and have achieved strong empirical performance, their theoretical properties are less well understood. In this work we study the convergence properties of the gradient descent algorithm when used to train GNNs. In particular, we consider the realizable setting where the data is generated from a network with unknown weights and our goal is to study conditions under which gradient descent on a GNN architecture can recover near optimal solutions. While such analysis has been performed in recent years for other architectures such as fully connected feed-forward networks, the message passing nature of the updates in a GNN poses a new challenge in understanding the nature of the gradient descent updates. We take a step towards overcoming this by proving that for the case of deep linear GNNs gradient descent provably recovers solutions up to error (cid:15) in O ( log (1 /(cid:15) )) iterations, under natural assumptions on the data distribution. Furthermore, for the case of one-round GNNs with ReLU activations, we show that gradient descent provably recovers solutions up to error (cid:15) in O ( 1 (cid:15) 2 log( 1 (cid:15) )) iterations."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow do the properties of feature similarity and sparsity in graphs influence the effectiveness of similarity-based edge reconstruction attacks on graph representations?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the security vulnerabilities inherent in graph representation learning (GRL) models, which are increasingly deployed in real-world applications. Understanding these vulnerabilities can lead to the development of more robust privacy-preserving techniques, thereby enhancing the security of sensitive data in various domains, such as social networks and healthcare. This research could pave the way for future studies that explore the intersection of graph theory, machine learning, and privacy, ultimately leading to practical applications that ensure data integrity and confidentiality.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complex interplay between graph properties and the effectiveness of similarity-based attacks. Naive approaches may fail because they do not account for the nuanced relationships between feature similarity and edge adjacency, particularly in sparse graphs. Additionally, theoretical obstacles exist in establishing a clear understanding of how these properties affect attack performance, as well as practical challenges in empirically validating these theories across diverse graph structures.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on empirical validations of graph representation vulnerabilities without delving into the theoretical underpinnings that explain the effectiveness of similarity-based attacks. Gaps in understanding the role of feature similarity and sparsity have hindered progress. Additionally, existing solutions may not have adequately addressed the specific conditions under which these attacks succeed or fail. Our approach differs by providing a comprehensive theoretical framework alongside empirical evaluations, thereby filling these gaps and offering new insights into the problem.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a detailed analysis of similarity-based edge reconstruction attacks (SERA) on sparse random graphs and stochastic block models (SBM). We will utilize synthetic datasets and real-world data to evaluate the performance of SERA under varying conditions of feature similarity and sparsity. The metrics for success will include the accuracy of edge reconstruction and the resilience of graph representations against attacks. Expected outcomes include theoretical insights into the success and failure modes of SERA, as well as empirical evidence supporting the effectiveness of noisy aggregation (NAG) as a mitigation strategy against these attacks."
            }
        },
        "author_data": {
            "1172bfc8-ca63-475b-bd49-0f0733ae593c": {
                "pk": "1172bfc8-ca63-475b-bd49-0f0733ae593c",
                "name": "Ruofan Wu",
                "collaborators": [
                    "Jennie Si",
                    "Zhikai Yao",
                    "Junmin Zhong",
                    "He",
                    "Huang",
                    "Weiqiang Wang",
                    "Tianyi Zhang",
                    "Minhan Li",
                    "Jianyong Yao",
                    "Yan Feng"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Graph Neural Network",
                    "Federated Learning",
                    "Robotics"
                ],
                "publications": [
                    {
                        "title": "Robotic Knee Tracking Control to Mimic the Intact Human Knee Profile Based on Actor-critic Reinforcement Learning",
                        "abstract": "We address a state-of-the-art reinforcement learning (RL) control approach to automatically configure robotic prosthesis impedance parameters to enable end-to-end, continuous locomotion intended for transfemoral amputee subjects. Specifically, our actor-critic based RL provides tracking control of a robotic knee prosthesis to mimic the intact knee profile. This is a significant advance from our previous RL based automatic tuning of prosthesis control parameters which have centered on regulation control with a designer prescribed robotic knee profile as the target. In addition to presenting the complete tracking control algorithm based on direct heuristic dynamic programming (dHDP), we provide an analytical framework for the tracking controller with constrained inputs. We show that our proposed tracking control possesses several important properties, such as weight convergence of the learning networks, Bellman (sub)optimality of the cost-to-go value function and control input, and practical stability of the human-robot system under input constraint. We further provide a systematic simulation of the proposed tracking control using a realistic human-robot system simulator, the OpenSim, to emulate how the dHDP enables level ground walking, walking on different terrains and at different paces. These results show that our proposed dHDP based tracking control is not only theoretically suitable, but also practically useful."
                    },
                    {
                        "title": "Actor-Critic Reinforcement Learning with Phased Actor",
                        "abstract": "Policy gradient methods in actor-critic reinforcement learning (RL) have become perhaps the most promising approaches to solving continuous optimal control problems. However, the trial-and-error nature of RL and the inherent randomness associated with solution approximations cause variations in the learned optimal values and policies. This has significantly hindered their successful deployment in real life applications where control responses need to meet dynamic performance criteria deterministically. Here we propose a novel phased actor in actor-critic (PAAC) method, aiming at improving policy gradient estimation and thus the quality of the control policy. Specifically, PAAC accounts for both $Q$ value and TD error in its actor update. We prove qualitative properties of PAAC for learning convergence of the value and policy, solution optimality, and stability of system dynamics. Additionally, we show variance reduction in policy gradient estimation. PAAC performance is systematically and quantitatively evaluated in this study using DeepMind Control Suite (DMC). Results show that PAAC leads to significant performance improvement measured by total cost, learning variance, robustness, learning speed and success rate. As PAAC can be piggybacked onto general policy gradient learning frameworks, we select well-known methods such as direct heuristic dynamic programming (dHDP), deep deterministic policy gradient (DDPG) and their variants to demonstrate the effectiveness of PAAC. Consequently we provide a unified view on these related policy gradient algorithms."
                    },
                    {
                        "title": "Long N-step Surrogate Stage Reward to Reduce Variances of Deep Reinforcement Learning in Complex Problems",
                        "abstract": "High variances in reinforcement learning have shown impeding successful convergence and hurting task performance. As reward signal plays an important role in learning behavior, multi-step methods have been considered to mitigate the problem, and are believed to be more effective than single step methods. However, there is a lack of comprehensive and systematic study on this important aspect to demonstrate the effectiveness of multi-step methods in solving highly complex continuous control problems. In this study, we introduce a new long $N$-step surrogate stage (LNSS) reward approach to effectively account for complex environment dynamics while previous methods are usually feasible for limited number of steps. The LNSS method is simple, low computational cost, and applicable to value based or policy gradient reinforcement learning. We systematically evaluate LNSS in OpenAI Gym and DeepMind Control Suite to address some complex benchmark environments that have been challenging to obtain good results by DRL in general. We demonstrate performance improvement in terms of total reward, convergence speed, and coefficient of variation (CV) by LNSS. We also provide analytical insights on how LNSS exponentially reduces the upper bound on the variances of Q value from a respective single step method"
                    },
                    {
                        "title": "Mitigating Estimation Errors by Twin TD-Regularized Actor and Critic for Deep Reinforcement Learning",
                        "abstract": "We address the issue of estimation bias in deep reinforcement learning (DRL) by introducing solution mechanisms that include a new, twin TD-regularized actor-critic (TDR) method. It aims at reducing both over and under-estimation errors. With TDR and by combining good DRL improvements, such as distributional learning and long N-step surrogate stage reward (LNSS) method, we show that our new TDR-based actor-critic learning has enabled DRL methods to outperform their respective baselines in challenging environments in DeepMind Control Suite. Furthermore, they elevate TD3 and SAC respectively to a level of performance comparable to that of D4PG (the current SOTA), and they also improve the performance of D4PG to a new SOTA level measured by mean reward, convergence speed, learning success rate, and learning variance."
                    },
                    {
                        "title": "Reinforcement Learning Enabled Automatic Impedance Control of a Robotic Knee Prosthesis to Mimic the Intact Knee Motion in a Co-Adapting Environment",
                        "abstract": "Automatically configuring a robotic prosthesis to fit its user's needs and physical conditions is a great technical challenge and a roadblock to the adoption of the technology. Previously, we have successfully developed reinforcement learning (RL) solutions toward addressing this issue. Yet, our designs were based on using a subjectively prescribed target motion profile for the robotic knee during level ground walking. This is not realistic for different users and for different locomotion tasks. In this study for the first time, we investigated the feasibility of RL enabled automatic configuration of impedance parameter settings for a robotic knee to mimic the intact knee motion in a co-adapting environment. We successfully achieved such tracking control by an online policy iteration. We demonstrated our results in both OpenSim simulations and two able-bodied (AB) subjects."
                    },
                    {
                        "title": "Toward Reliable Designs of Data-Driven Reinforcement Learning Tracking Control for Euler-Lagrange Systems",
                        "abstract": "This paper addresses reinforcement learning based, direct signal tracking control with an objective of developing mathematically suitable and practically useful design approaches. Specifically, we aim to provide reliable and easy to implement designs in order to reach reproducible neural network-based solutions. Our proposed new design takes advantage of two control design frameworks: a reinforcement learning based, data-driven approach to provide the needed adaptation and (sub)optimality, and a backstepping based approach to provide closed-loop system stability framework. We develop this work based on an established direct heuristic dynamic programming (dHDP) learning paradigm to perform online learning and adaptation and a backstepping design for a class of important nonlinear dynamics described as Euler-Lagrange systems. We provide a theoretical guarantee for the stability of the overall dynamic system, weight convergence of the approximating nonlinear neural networks, and the Bellman (sub)optimality of the resulted control policy. We use simulations to demonstrate significantly improved design performance of the proposed approach over the original dHDP."
                    },
                    {
                        "title": "sqSGD: Locally Private and Communication Efficient Federated Learning",
                        "abstract": "Federated learning (FL) is a technique that trains machine learning models from decentralized data sources. We study FL under local notions of privacy constraints, which provides strong protection against sensitive data disclosures via obfuscating the data before leaving the client. We identify two major concerns in designing practical privacy-preserving FL algorithms: communication efficiency and high-dimensional compatibility. We then develop a gradient-based learning algorithm called \\emph{sqSGD} (selective quantized stochastic gradient descent) that addresses both concerns. The proposed algorithm is based on a novel privacy-preserving quantization scheme that uses a constant number of bits per dimension per client. Then we improve the base algorithm in three ways: first, we apply a gradient subsampling strategy that simultaneously offers better training performance and smaller communication costs under a fixed privacy budget. Secondly, we utilize randomized rotation as a preprocessing step to reduce quantization error. Thirdly, an adaptive gradient norm upper bound shrinkage strategy is adopted to improve accuracy and stabilize training. Finally, the practicality of the proposed framework is demonstrated on benchmark datasets. Experiment results show that sqSGD successfully learns large models like LeNet and ResNet with local privacy constraints. In addition, with fixed privacy and communication level, the performance of sqSGD significantly dominates that of various baseline algorithms."
                    },
                    {
                        "title": "GRANDE: a neural model over directed multigraphs with application to anti-money laundering",
                        "abstract": "The application of graph representation learning techniques to the area of financial risk management (FRM) has attracted significant attention recently. However, directly modeling transaction networks using graph neural models remains challenging: Firstly, transaction networks are directed multigraphs by nature, which could not be properly handled with most of the current off-the-shelf graph neural networks (GNN). Secondly, a crucial problem in FRM scenarios like anti-money laundering (AML) is to identify risky transactions and is most naturally cast into an edge classification problem with rich edge-level features, which are not fully exploited by the prevailing GNN design that follows node-centric message passing protocols. In this paper, we present a systematic investigation of design aspects of neural models over directed multigraphs and develop a novel GNN protocol that overcomes the above challenges via efficiently incorporating directional information, as well as proposing an enhancement that targets edge-related tasks using a novel message passing scheme over an extension of edge-to-node dual graph. A concrete GNN architecture called GRANDE is derived using the proposed protocol, with several further improvements and generalizations to temporal dynamic graphs. We apply the GRANDE model to both a real-world anti-money laundering task and public datasets. Experimental evaluations show the superiority of the proposed GRANDE architecture over recent state-of-the-art models on dynamic graph modeling and directed graph modeling."
                    },
                    {
                        "title": "Oversmoothing: A Nightmare for Graph Contrastive Learning?",
                        "abstract": "Oversmoothing is a common phenomenon observed in graph neural networks (GNNs), in which an increase in the network depth leads to a deterioration in their performance. Graph contrastive learning (GCL) is emerging as a promising way of leveraging vast unlabeled graph data. As a marriage between GNNs and contrastive learning, it remains unclear whether GCL inherits the same oversmoothing defect from GNNs. This work undertakes a fundamental analysis of GCL from the perspective of oversmoothing on the first hand. We demonstrate empirically that increasing network depth in GCL also leads to oversmoothing in their deep representations, and surprisingly, the shallow ones. We refer to this phenomenon in GCL as `long-range starvation', wherein lower layers in deep networks suffer from degradation due to the lack of sufficient guidance from supervision. Based on our findings, we present BlockGCL, a remarkably simple yet effective blockwise training framework to prevent GCL from notorious oversmoothing. Without bells and whistles, BlockGCL consistently improves robustness and stability for well-established GCL methods with increasing numbers of layers on several real-world graph benchmarks."
                    },
                    {
                        "title": "FedGKD: Unleashing the Power of Collaboration in Federated Graph Neural Networks",
                        "abstract": "Federated training of Graph Neural Networks (GNN) has become popular in recent years due to its ability to perform graph-related tasks under data isolation scenarios while preserving data privacy. However, graph heterogeneity issues in federated GNN systems continue to pose challenges. Existing frameworks address the problem by representing local tasks using different statistics and relating them through a simple aggregation mechanism. However, these approaches suffer from limited efficiency from two aspects: low quality of task-relatedness quantification and inefficacy of exploiting the collaboration structure. To address these issues, we propose FedGKD, a novel federated GNN framework that utilizes a novel client-side graph dataset distillation method to extract task features that better describe task-relatedness, and introduces a novel server-side aggregation mechanism that is aware of the global collaboration structure. We conduct extensive experiments on six real-world datasets of different scales, demonstrating our framework's outperformance."
                    }
                ]
            },
            "0169de67-f71c-49a4-a48d-05629fbfc74d": {
                "pk": "0169de67-f71c-49a4-a48d-05629fbfc74d",
                "name": "Guanhua Fang",
                "collaborators": [
                    "Ping Li",
                    "Zhiliang Ying",
                    "Gennady Samorodnitsky",
                    "Haochen Xu",
                    "Wen Yu",
                    "Sujay Bhatt",
                    "Jingchen Liu",
                    "Zhiqiang Xu",
                    "Yuecheng Zhang",
                    "Zhiheng Chen"
                ],
                "domain": [
                    "Statistical Modeling",
                    "Temporal Data Analysis",
                    "Machine Learning",
                    "Bayesian Inference"
                ],
                "publications": [
                    {
                        "title": "Latent Theme Dictionary Model for Finding Co-occurrent Patterns in Process Data",
                        "abstract": "Process data, temporally ordered categorical observations, are of recent interest due to its increasing abundance and the desire to extract useful information. A process is a collection of time-stamped events of different types, recording how an individual behaves in a given time period. The process data are too complex in terms of size and irregularity for the classical psychometric models to be applicable, at least directly, and, consequently, it is desirable to develop new ways for modeling and analysis. We introduce herein a latent theme dictionary model (LTDM) for processes that identifies co-occurrent event patterns and individuals with similar behavioral patterns. Theoretical properties are established under certain regularity conditions for the likelihood based estimation and inference. A non-parametric Bayes LTDM algorithm using the Markov Chain Monte Carlo method is proposed for computation. Simulation studies show that the proposed approach performs well in a range of situations. The proposed method is applied to an item in the 2012 Programme for International Student Assessment with interpretable findings."
                    },
                    {
                        "title": "Regression with Label Permutation in Generalized Linear Model",
                        "abstract": "The assumption that response and predictor belong to the same statistical unit may be violated in practice. Unbiased estimation and recovery of true label ordering based on unlabeled data are challenging tasks and have attracted increasing attentions in the recent literature. In this paper, we present a relatively complete analysis of label permutation problem for the generalized linear model with multivariate responses. The theory is established under different scenarios, with knowledge of true parameters, with partial knowledge of underlying label permutation matrix and without any knowledge. Our results remove the stringent conditions required by the current literature and are further extended to the missing observation setting which has never been considered in the field of label permutation problem. On computational side, we propose two methods, \"maximum likelihood estimation\" algorithm and \"two-step estimation\" algorithm, to accommodate for different settings. When the proportion of permuted labels is moderate, both methods work effectively. Multiple numerical experiments are provided and corroborate our theoretical findings."
                    },
                    {
                        "title": "On the Identifiability of Diagnostic Classification Models",
                        "abstract": "This paper establishes fundamental results for statistical inference of diagnostic classification models (DCM). The results are developed at a high level of generality, applicable to essentially all diagnostic classification models. In particular, we establish identifiability results of various modeling parameters, notably item response probabilities, attribute distribution, and Q-matrix-induced partial information structure. Consistent estimators are constructed. Simulation results show that these estimators perform well under various modeling settings. We also use a real example to illustrate the new method. The results are stated under the setting of general latent class models. For DCM with a specific parameterization, the conditions may be adapted accordingly."
                    },
                    {
                        "title": "A Latent Topic Model with Markovian Transition for Process Data",
                        "abstract": "We propose a latent topic model with a Markovian transition for process data, which consist of time-stamped events recorded in a log file. Such data are becoming more widely available in computer-based educational assessment with complex problem solving items. The proposed model can be viewed as an extension of the hierarchical Bayesian topic model with a hidden Markov structure to accommodate the underlying evolution of an examinee's latent state. Using topic transition probabilities along with response times enables us to capture examinees' learning trajectories, making clustering/classification more efficient. A forward-backward variational expectation-maximization (FB-VEM) algorithm is developed to tackle the challenging computational problem. Useful theoretical properties are established under certain asymptotic regimes. The proposed method is applied to a complex problem solving item in 2012 Programme for International Student Assessment (PISA 2012)."
                    },
                    {
                        "title": "A Cover Time Study of a non-Markovian Algorithm",
                        "abstract": "Given a traversal algorithm, cover time is the expected number of steps needed to visit all nodes in a given graph. A smaller cover time means a higher exploration efficiency of traversal algorithm. Although random walk algorithms have been studied extensively in the existing literature, there has been no cover time result for any non-Markovian method. In this work, we stand on a theoretical perspective and show that the negative feedback strategy (a count-based exploration method) is better than the naive random walk search. In particular, the former strategy can locally improve the search efficiency for an arbitrary graph. It also achieves smaller cover times for special but important graphs, including clique graphs, tree graphs, etc. Moreover, we make connections between our results and reinforcement learning literature to give new insights on why classical UCB and MCTS algorithms are so useful. Various numerical results corroborate our theoretical findings."
                    },
                    {
                        "title": "On Robust Clustering of Temporal Point Process",
                        "abstract": "Clustering of event stream data is of great importance in many application scenarios, including but not limited to, e-commerce, electronic health, online testing, mobile music service, etc. Existing clustering algorithms fail to take outlier data into consideration and are implemented without theoretical guarantees. In this paper, we propose a robust temporal point processes clustering framework which works under mild assumptions and meanwhile addresses several important issues in the event stream clustering problem.Specifically, we introduce a computationally efficient model-free distance function to quantify the dissimilarity between different event streams so that the outliers can be detected and the good initial clusters could be obtained. We further consider an expectation-maximization-type algorithm incorporated with a Catoni's influence function for robust estimation and fine-tuning of clusters. We also establish the theoretical results including algorithmic convergence, estimation error bound, outlier detection, etc. Simulation results corroborate our theoretical findings and real data applications show the effectiveness of our proposed methodology."
                    },
                    {
                        "title": "On Non-asymptotic Theory of Recurrent Neural Networks in Temporal Point Processes",
                        "abstract": "Temporal point process (TPP) is an important tool for modeling and predicting irregularly timed events across various domains. Recently, the recurrent neural network (RNN)-based TPPs have shown practical advantages over traditional parametric TPP models. However, in the current literature, it remains nascent in understanding neural TPPs from theoretical viewpoints. In this paper, we establish the excess risk bounds of RNN-TPPs under many well-known TPP settings. We especially show that an RNN-TPP with no more than four layers can achieve vanishing generalization errors. Our technical contributions include the characterization of the complexity of the multi-layer RNN class, the construction of $\\tanh$ neural networks for approximating dynamic event intensity functions, and the truncation technique for alleviating the issue of unbounded event sequences. Our results bridge the gap between TPP's application and neural network theory."
                    },
                    {
                        "title": "Best Subset Selection with Efficient Primal-Dual Algorithm",
                        "abstract": "Best subset selection is considered the `gold standard' for many sparse learning problems. A variety of optimization techniques have been proposed to attack this non-convex and NP-hard problem. In this paper, we investigate the dual forms of a family of $\\ell_0$-regularized problems. An efficient primal-dual method has been developed based on the primal and dual problem structures. By leveraging the dual range estimation along with the incremental strategy, our algorithm potentially reduces redundant computation and improves the solutions of best subset selection. Theoretical analysis and experiments on synthetic and real-world datasets validate the efficiency and statistical properties of the proposed solutions."
                    },
                    {
                        "title": "Copula for Instance-wise Feature Selection and Ranking",
                        "abstract": "Instance-wise feature selection and ranking methods can achieve a good selection of task-friendly features for each sample in the context of neural networks. However, existing approaches that assume feature subsets to be independent are imperfect when considering the dependency between features. To address this limitation, we propose to incorporate the Gaussian copula, a powerful mathematical technique for capturing correlations between variables, into the current feature selection framework with no additional changes needed. Experimental results on both synthetic and real datasets, in terms of performance comparison and interpretability, demonstrate that our method is capable of capturing meaningful correlations."
                    },
                    {
                        "title": "Offline Change Detection under Contamination",
                        "abstract": "In this work, we propose a non-parametric and robust change detection algorithm to detect multiple change points in time series data under contamination. The contamination model is sufficiently general, in that, the most common model used in the context of change detection -- Huber contamination model -- is a special case. Also, the contamination model is oblivious and arbitrary. The change detection algorithm is designed for the offline setting, where the objective is to detect changes when all data are received. We only make weak moment assumptions on the inliers (uncorrupted data) to handle a large class of distributions. The robust scan statistic in the algorithm is fashioned using mean estimators based on influence functions. We establish the consistency of the estimated change point indexes as the number of samples increases, and provide empirical evidence to support the consistency results."
                    },
                    {
                        "title": "On Penalization in Stochastic Multi-armed Bandits",
                        "abstract": "We study an important variant of the stochastic multi-armed bandit (MAB) problem, which takes penalization into consideration. Instead of directly maximizing cumulative expected reward, we need to balance between the total reward and fairness level. In this paper, we present some new insights in MAB and formulate the problem in the penalization framework, where rigorous penalized regret can be well defined and more sophisticated regret analysis is possible. Under such a framework, we propose a hard-threshold UCB-like algorithm, which enjoys many merits including asymptotic fairness, nearly optimal regret, better tradeoff between reward and fairness. Both gap-dependent and gap-independent regret bounds have been established. Multiple insightful comments are given to illustrate the soundness of our theoretical analysis. Numerous experimental results corroborate the theory and show the superiority of our method over other existing methods."
                    },
                    {
                        "title": "Empirical Risk Minimization for Losses without Variance",
                        "abstract": "This paper considers an empirical risk minimization problem under heavy-tailed settings, where data does not have finite variance, but only has $p$-th moment with $p \\in (1,2)$. Instead of using estimation procedure based on truncated observed data, we choose the optimizer by minimizing the risk value. Those risk values can be robustly estimated via using the remarkable Catoni's method (Catoni, 2012). Thanks to the structure of Catoni-type influence functions, we are able to establish excess risk upper bounds via using generalized generic chaining methods. Moreover, we take computational issues into consideration. We especially theoretically investigate two types of optimization methods, robust gradient descent algorithm and empirical risk-based methods. With an extensive numerical study, we find that the optimizer based on empirical risks via Catoni-style estimation indeed shows better performance than other baselines. It indicates that estimation directly based on truncated data may lead to unsatisfactory results."
                    },
                    {
                        "title": "Identifiability of Bifactor Models",
                        "abstract": "The bifactor model and its extensions are multidimensional latent variable models, under which each item measures up to one subdimension on top of the primary dimension(s). Despite their wide applications to educational and psychological assessments, this type of multidimensional latent variable models may suffer from non-identifiability, which can further lead to inconsistent parameter estimation and invalid inference. The current work provides a relatively complete characterization of identifiability for the linear and dichotomous bifactor models and the linear extended bifactor model with correlated subdimensions. In addition, similar results for the two-tier models are also developed. Illustrative examples are provided on checking model identifiability through inspecting the factor loading structure. Simulation studies are reported that examine estimation consistency when the identifiability conditions are/are not satisfied."
                    },
                    {
                        "title": "Catoni-style Confidence Sequences under Infinite Variance",
                        "abstract": "In this paper, we provide an extension of confidence sequences for settings where the variance of the data-generating distribution does not exist or is infinite. Confidence sequences furnish confidence intervals that are valid at arbitrary data-dependent stopping times, naturally having a wide range of applications. We first establish a lower bound for the width of the Catoni-style confidence sequences for the finite variance case to highlight the looseness of the existing results. Next, we derive tight Catoni-style confidence sequences for data distributions having a relaxed bounded~$p^{th}-$moment, where~$p \\in (1,2]$, and strengthen the results for the finite variance case of~$p =2$. The derived results are shown to better than confidence sequences obtained using Dubins-Savage inequality."
                    },
                    {
                        "title": "Online Estimation and Community Detection of Network Point Processes for Event Streams",
                        "abstract": "A common goal in network modeling is to uncover the latent community structure present among nodes. For many real-world networks, the true connections consist of events arriving as streams, which are then aggregated to form edges, ignoring the dynamic temporal component. A natural way to take account of these temporal dynamics of interactions is to use point processes as the foundation of network models for community detection. Computational complexity hampers the scalability of such approaches to large sparse networks. To circumvent this challenge, we propose a fast online variational inference algorithm for estimating the latent structure underlying dynamic event arrivals on a network, using continuous-time point process latent network models. We describe this procedure for networks models capturing community structure. This structure can be learned as new events are observed on the network, updating the inferred community assignments. We investigate the theoretical properties of such an inference scheme, and provide regret bounds on the loss function of this procedure. The proposed inference procedure is then thoroughly compared, using both simulation studies and real data, to non-online variants. We demonstrate that online inference can obtain comparable performance, in terms of community recovery, to non-online variants, while realising computational gains. Our proposed inference framework can also be readily modified to incorporate other popular network structures."
                    },
                    {
                        "title": "Group Network Hawkes Process",
                        "abstract": "In this work, we study the event occurrences of individuals interacting in a network. To characterize the dynamic interactions among the individuals, we propose a group network Hawkes process (GNHP) model whose network structure is observed and fixed. In particular, we introduce a latent group structure among individuals to account for the heterogeneous user-specific characteristics. A maximum likelihood approach is proposed to simultaneously cluster individuals in the network and estimate model parameters. A fast EM algorithm is subsequently developed by utilizing the branching representation of the proposed GNHP model. Theoretical properties of the resulting estimators of group memberships and model parameters are investigated under both settings when the number of latent groups $G$ is over-specified or correctly specified. A data-driven criterion that can consistently identify the true $G$ under mild conditions is derived. Extensive simulation studies and an application to a data set collected from Sina Weibo are used to illustrate the effectiveness of the proposed methodology."
                    }
                ]
            },
            "fb786f37-b890-4f37-af1b-2d2459e34328": {
                "pk": "fb786f37-b890-4f37-af1b-2d2459e34328",
                "name": "Qiying Pan",
                "collaborators": [
                    "Yifei Zhu",
                    "Lingyang Chu",
                    "Ruofan Wu",
                    "Tengfei Liu",
                    "Tianyi Zhang",
                    "Weiqiang Wang",
                    "Fuhua Li",
                    "Jia Wang",
                    "Hong-Mei Zou",
                    "Chenya Liu"
                ],
                "domain": [
                    "Federated Learning",
                    "Graph Neural Network",
                    "Data Privacy",
                    "Node Embedding"
                ],
                "publications": [
                    {
                        "title": "Lumos: Heterogeneity-aware Federated Graph Learning over Decentralized Devices",
                        "abstract": "Graph neural networks (GNN) have been widely deployed in real-world networked applications and systems due to their capability to handle graph-structured data. However, the growing awareness of data privacy severely challenges the traditional centralized model training paradigm, where a server holds all the graph information. Federated learning is an emerging collaborative computing paradigm that allows model training without data centralization. Existing federated GNN studies mainly focus on systems where clients hold distinctive graphs or sub-graphs. The practical node-level federated situation, where each client is only aware of its direct neighbors, has yet to be studied. In this paper, we propose the first federated GNN framework called Lumos that supports supervised and unsupervised learning with feature and degree protection on node-level federated graphs. We first design a tree constructor to improve the representation capability given the limited structural information. We further present a Monte Carlo Markov Chain-based algorithm to mitigate the workload imbalance caused by degree heterogeneity with theoretically-guaranteed performance. Based on the constructed tree for each client, a decentralized tree-based GNN trainer is proposed to support versatile training. Extensive experiments demonstrate that Lumos outperforms the baseline with significantly higher accuracy and greatly reduced communication cost and training time."
                    },
                    {
                        "title": "FedWalk: Communication Efficient Federated Unsupervised Node Embedding with Differential Privacy",
                        "abstract": "Node embedding aims to map nodes in the complex graph into low-dimensional representations. The real-world large-scale graphs and difficulties of labeling motivate wide studies of unsupervised node embedding problems. Nevertheless, previous effort mostly operates in a centralized setting where a complete graph is given. With the growing awareness of data privacy, data holders who are only aware of one vertex and its neighbours demand greater privacy protection. In this paper, we introduce FedWalk, a random-walk-based unsupervised node embedding algorithm that operates in such a node-level visibility graph with raw graph information remaining locally. FedWalk is designed to offer centralized competitive graph representation capability with data privacy protection and great communication efficiency. FedWalk instantiates the prevalent federated paradigm and contains three modules. We first design a hierarchical clustering tree (HCT) constructor to extract the structural feature of each node. A dynamic time warping algorithm seamlessly handles the structural heterogeneity across different nodes. Based on the constructed HCT, we then design a random walk generator, wherein a sequence encoder is designed to preserve privacy and a two-hop neighbor predictor is designed to save communication cost. The generated random walks are then used to update node embedding based on a SkipGram model. Extensive experiments on two large graphs demonstrate that Fed-Walk achieves competitive representativeness as a centralized node embedding algorithm does with only up to 1.8% Micro-F1 score and 4.4% Marco-F1 score loss while reducing about 6.7 times of inter-device communication per walk."
                    },
                    {
                        "title": "FedGKD: Unleashing the Power of Collaboration in Federated Graph Neural Networks",
                        "abstract": "Federated training of Graph Neural Networks (GNN) has become popular in recent years due to its ability to perform graph-related tasks under data isolation scenarios while preserving data privacy. However, graph heterogeneity issues in federated GNN systems continue to pose challenges. Existing frameworks address the problem by representing local tasks using different statistics and relating them through a simple aggregation mechanism. However, these approaches suffer from limited efficiency from two aspects: low quality of task-relatedness quantification and inefficacy of exploiting the collaboration structure. To address these issues, we propose FedGKD, a novel federated GNN framework that utilizes a novel client-side graph dataset distillation method to extract task features that better describe task-relatedness, and introduces a novel server-side aggregation mechanism that is aware of the global collaboration structure. We conduct extensive experiments on six real-world datasets of different scales, demonstrating our framework's outperformance."
                    },
                    {
                        "title": "Modulating Entanglement Dynamics of Two V-type Atoms in Dissipative Cavity by Detuning, Weak Measurement and Reversal",
                        "abstract": "In this paper, how to modulate entanglement dynamics of two V-type atoms in dissipative cavity by detuning, weak measurement and weak measurement reversal is studied. The analytical solution of this model is obtained by solving Schrodinger Equation after diagonalizing Hamiltonian of dissipative cavity. It is discussed in detail how the entanglement dynamics is influenced by cavity--environment coupling, spontaneously generated interference (SGI) parameter, detuning between cavity with environment and weak measurement reversal. The results show that the entanglement dynamics of different initial states obviously depends on coupling, SGI parameter, detuning and reversing measurement strength. The stronger coupling, the smaller SGI parameter, the larger detuning and the bigger reversing measurement strength can all not only protect but also generate the entanglement, and the detuning is more effectively in tne strong coupling regime than the weak measurement reversal, which is more effectively than the SGI parameter. We also give corresponding physical interpretations."
                    }
                ]
            },
            "cdfa775d-74e2-4342-98eb-538b576a1ba1": {
                "pk": "cdfa775d-74e2-4342-98eb-538b576a1ba1",
                "name": "Mingyang Zhang",
                "collaborators": [
                    "Xinyi Yu",
                    "Linlin Ou",
                    "Jingtao Rong",
                    "Berrak Sisman",
                    "Haizhou Li",
                    "Jie Jia",
                    "Jian Chen",
                    "Gang Wang",
                    "Qi Peng",
                    "Yanfeng Zhang"
                ],
                "domain": [
                    "Neural Architecture Search",
                    "AutoML",
                    "Speech Processing",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "RepNAS: Searching for Efficient Re-parameterizing Blocks",
                        "abstract": "In the past years, significant improvements in the field of neural architecture search(NAS) have been made. However, it is still challenging to search for efficient networks due to the gap between the searched constraint and real inference time exists. To search for a high-performance network with low inference time, several previous works set a computational complexity constraint for the search algorithm. However, many factors affect the speed of inference(e.g., FLOPs, MACs). The correlation between a single indicator and the latency is not strong. Currently, some re-parameterization(Rep) techniques are proposed to convert multi-branch to single-path architecture which is inference-friendly. Nevertheless, multi-branch architectures are still human-defined and inefficient. In this work, we propose a new search space that is suitable for structural re-parameterization techniques. RepNAS, a one-stage NAS approach, is present to efficiently search the optimal diverse branch block(ODBB) for each layer under the branch number constraint. Our experimental results show the searched ODBB can easily surpass the manual diverse branch block(DBB) with efficient training."
                    },
                    {
                        "title": "Indoor Localization Using Smartphone Magnetic with Multi-Scale TCN and LSTM",
                        "abstract": "A novel multi-scale temporal convolutional network (TCN) and long short-term memory network (LSTM) based magnetic localization approach is proposed. To enhance the discernibility of geomagnetic signals, the time-series preprocessing approach is constructed at first. Next, the TCN is invoked to expand the feature dimensions on the basis of keeping the time-series characteristics of LSTM model. Then, a multi-scale time-series layer is constructed with multiple TCNs of different dilation factors to address the problem of inconsistent time-series speed between localization model and mobile users. A stacking framework of multi-scale TCN and LSTM is eventually proposed for indoor magnetic localization. Experiment results demonstrate the effectiveness of the proposed algorithm in indoor localization."
                    },
                    {
                        "title": "Graph Pruning for Model Compression",
                        "abstract": "Previous AutoML pruning works utilized individual layer features to automatically prune filters. We analyze the correlation for two layers from the different blocks which have a short-cut structure. It shows that, in one block, the deeper layer has many redundant filters which can be represented by filters in the former layer. So, it is necessary to take information from other layers into consideration in pruning. In this paper, a novel pruning method, named GraphPruning, is proposed. Any series of the network is viewed as a graph. To automatically aggregate neighboring features for each node, a graph aggregator based on graph convolution networks(GCN) is designed. In the training stage, a PruningNet that is given aggregated node features generates reasonable weights for any size of the sub-network. Subsequently, the best configuration of the Pruned Network is searched by reinforcement learning. Different from previous work, we take the node features from a well-trained graph aggregator instead of the hand-craft features, as the states in reinforcement learning. Compared with other AutoML pruning works, our method has achieved the state-of-the-art under the same conditions on ImageNet-2012."
                    },
                    {
                        "title": "What Have We Learned from OpenReview?",
                        "abstract": "Anonymous peer review is used by the great majority of computer science conferences. OpenReview is such a platform that aims to promote openness in peer review process. The paper, (meta) reviews, rebuttals, and final decisions are all released to public. We collect 5,527 submissions and their 16,853 reviews from the OpenReview platform. We also collect these submissions' citation data from Google Scholar and their non-peer-reviewed versions from arXiv.org. By acquiring deep insights into these data, we have several interesting findings that could help understand the effectiveness of the public-accessible double-blind peer review process. Our results can potentially help writing a paper, reviewing it, and deciding on its acceptance."
                    },
                    {
                        "title": "Effective Model Compression via Stage-wise Pruning",
                        "abstract": "Automated Machine Learning(Auto-ML) pruning methods aim at searching a pruning strategy automatically to reduce the computational complexity of deep Convolutional Neural Networks(deep CNNs). However, some previous work found that the results of many Auto-ML pruning methods cannot even surpass the results of the uniformly pruning method. In this paper, the ineffectiveness of Auto-ML pruning which is caused by unfull and unfair training of the supernet is shown. A deep supernet suffers from unfull training because it contains too many candidates. To overcome the unfull training, a stage-wise pruning(SWP) method is proposed, which splits a deep supernet into several stage-wise supernets to reduce the candidate number and utilize inplace distillation to supervise the stage training. Besides, A wide supernet is hit by unfair training since the sampling probability of each channel is unequal. Therefore, the fullnet and the tinynet are sampled in each training iteration to ensure each channel can be overtrained. Remarkably, the proxy performance of the subnets trained with SWP is closer to the actual performance than that of most of the previous Auto-ML pruning work. Experiments show that SWP achieves the state-of-the-art on both CIFAR-10 and ImageNet under the mobile setting."
                    },
                    {
                        "title": "ShiftNAS: Improving One-shot NAS via Probability Shift",
                        "abstract": "One-shot Neural architecture search (One-shot NAS) has been proposed as a time-efficient approach to obtain optimal subnet architectures and weights under different complexity cases by training only once. However, the subnet performance obtained by weight sharing is often inferior to the performance achieved by retraining. In this paper, we investigate the performance gap and attribute it to the use of uniform sampling, which is a common approach in supernet training. Uniform sampling concentrates training resources on subnets with intermediate computational resources, which are sampled with high probability. However, subnets with different complexity regions require different optimal training strategies for optimal performance. To address the problem of uniform sampling, we propose ShiftNAS, a method that can adjust the sampling probability based on the complexity of subnets. We achieve this by evaluating the performance variation of subnets with different complexity and designing an architecture generator that can accurately and efficiently provide subnets with the desired complexity. Both the sampling probability and the architecture generator can be trained end-to-end in a gradient-based manner. With ShiftNAS, we can directly obtain the optimal model architecture and parameters for a given computational complexity. We evaluate our approach on multiple visual network models, including convolutional neural networks (CNNs) and vision transformers (ViTs), and demonstrate that ShiftNAS is model-agnostic. Experimental results on ImageNet show that ShiftNAS can improve the performance of one-shot NAS without additional consumption. Source codes are available at https://github.com/bestfleer/ShiftNAS."
                    },
                    {
                        "title": "DeepA: A Deep Neural Analyzer For Speech And Singing Vocoding",
                        "abstract": "Conventional vocoders are commonly used as analysis tools to provide interpretable features for downstream tasks such as speech synthesis and voice conversion. They are built under certain assumptions about the signals following signal processing principle, therefore, not easily generalizable to different audio, for example, from speech to singing. In this paper, we propose a deep neural analyzer, denoted as DeepA - a neural vocoder that extracts F0 and timbre/aperiodicity encoding from the input speech that emulate those defined in conventional vocoders. Therefore, the resulting parameters are more interpretable than other latent neural representations. At the same time, as the deep neural analyzer is learnable, it is expected to be more accurate for signal reconstruction and manipulation, and generalizable from speech to singing. The proposed neural analyzer is built based on a variational autoencoder (VAE) architecture. We show that DeepA improves F0 estimation over the conventional vocoder (WORLD). To our best knowledge, this is the first study dedicated to the development of a neural framework for extracting learnable vocoder-like parameters."
                    },
                    {
                        "title": "High-Quality Automatic Voice Over with Accurate Alignment: Supervision through Self-Supervised Discrete Speech Units",
                        "abstract": "The goal of Automatic Voice Over (AVO) is to generate speech in sync with a silent video given its text script. Recent AVO frameworks built upon text-to-speech synthesis (TTS) have shown impressive results. However, the current AVO learning objective of acoustic feature reconstruction brings in indirect supervision for inter-modal alignment learning, thus limiting the synchronization performance and synthetic speech quality. To this end, we propose a novel AVO method leveraging the learning objective of self-supervised discrete speech unit prediction, which not only provides more direct supervision for the alignment learning, but also alleviates the mismatch between the text-video context and acoustic features. Experimental results show that our proposed method achieves remarkable lip-speech synchronization and high speech quality by outperforming baselines in both objective and subjective evaluations. Code and speech samples are publicly available."
                    },
                    {
                        "title": "Transfer the linguistic representations from TTS to accent conversion with non-parallel data",
                        "abstract": "Accent conversion aims to convert the accent of a source speech to a target accent, meanwhile preserving the speaker's identity. This paper introduces a novel non-autoregressive framework for accent conversion that learns accent-agnostic linguistic representations and employs them to convert the accent in the source speech. Specifically, the proposed system aligns speech representations with linguistic representations obtained from Text-to-Speech (TTS) systems, enabling training of the accent voice conversion model on non-parallel data. Furthermore, we investigate the effectiveness of a pretraining strategy on native data and different acoustic features within our proposed framework. We conduct a comprehensive evaluation using both subjective and objective metrics to assess the performance of our approach. The evaluation results highlight the benefits of the pretraining strategy and the incorporation of richer semantic features, resulting in significantly enhanced audio quality and intelligibility."
                    }
                ]
            },
            "ffa5a3a7-de0f-4f35-9d01-dec2b452f394": {
                "pk": "ffa5a3a7-de0f-4f35-9d01-dec2b452f394",
                "name": "Tengfei Liu",
                "collaborators": [
                    "Zhong-Ping Jiang",
                    "Si Wu",
                    "Zhi Liu",
                    "Yongli Hu",
                    "Junbin Gao",
                    "Yanfeng Sun",
                    "Baocai Yin",
                    "Magnus Egerstedt"
                ],
                "domain": [
                    "Control Systems",
                    "Nonlinear Dynamics",
                    "Multi-Agent Systems",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Singular Perturbation: When the Perturbation Parameter Becomes a State-Dependent Function",
                        "abstract": "This paper presents a new systematic framework for nonlinear singularly perturbed systems in which state-dependent perturbation functions are used instead of constant perturbation coefficients. Under this framework, general results are obtained for the global robust stability and input-to-state stability of nonlinear singularly perturbed systems. Interestingly, the proposed methodology provides innovative solutions beyond traditional singular perturbation theory for emerging control problems arising from nonlinear integral control, feedback optimization, and formation-based source seeking."
                    },
                    {
                        "title": "Constructive Safety Control",
                        "abstract": "This paper proposes a constructive approach to safety control of nonlinear cascade systems subject to multiple state constraints. New design ingredients include a unified characterization of safety and stability for systematic designs of safety controllers, and a novel technique of reshaping the feasible sets of quadratically constrained quadratic programming induced from safety control. The proposed method guarantees Lipschitz continuity of virtual control laws, enabling a stepwise constructive design. A refined nonlinear small-gain synthesis is employed to address the nonlinear uncertain interconnections between the resulting subsystems corresponding to different virtual control laws, and to guarantee the achievement of the safety control objective. When the safety constraints are removed, the proposed approach coincides with the standard constructive nonlinear control. The proposed safety-control algorithm is experimentally validated in a testbed involving a vertical takeoff and landing (VTOL) vehicle taking off in narrow spaces."
                    },
                    {
                        "title": "Safety-Critical Control of Euler-Lagrange Systems Subject to Multiple Obstacles and Velocity Constraints",
                        "abstract": "This paper studies the safety-critical control problem for Euler-Lagrange (EL) systems subject to multiple ball obstacles and velocity constraints in accordance with affordable velocity ranges. A key strategy is to exploit the underlying inner-outer-loop structure for the design of a new cascade controller for the class of EL systems. In particular, the outer-loop controller is developed based on quadratic programming (QP) to avoid ball obstacles and generate velocity reference signals fulfilling the velocity limitation. Taking full advantage of the conservation-of-energy property, a nonlinear velocity-tracking controller is designed to form the inner loop. One major difficulty is caused by the possible non-Lipschitz continuity of the standard QP algorithm when there are multiple constraints. To solve this problem, we propose a refined QP algorithm with the feasible set reshaped by an appropriately chosen positive basis such that the feasibility is retained while the resulting outer-loop controller is locally Lipschitz. It is proved that the constraint-satisfaction problem is solvable as long as the ball obstacles satisfy a mild distance condition. The proposed design is validated by numerical simulation and an experiment based on a $2$-link planar manipulator."
                    },
                    {
                        "title": "Hierarchical Multi-modal Transformer for Cross-modal Long Document Classification",
                        "abstract": "Long Document Classification (LDC) has gained significant attention recently. However, multi-modal data in long documents such as texts and images are not being effectively utilized. Prior studies in this area have attempted to integrate texts and images in document-related tasks, but they have only focused on short text sequences and images of pages. How to classify long documents with hierarchical structure texts and embedding images is a new problem and faces multi-modal representation difficulties. In this paper, we propose a novel approach called Hierarchical Multi-modal Transformer (HMT) for cross-modal long document classification. The HMT conducts multi-modal feature interaction and fusion between images and texts in a hierarchical manner. Our approach uses a multi-modal transformer and a dynamic multi-scale multi-modal transformer to model the complex relationships between image features, and the section and sentence features. Furthermore, we introduce a new interaction strategy called the dynamic mask transfer module to integrate these two transformers by propagating features between them. To validate our approach, we conduct cross-modal LDC experiments on two newly created and two publicly available multi-modal long document datasets, and the results show that the proposed HMT outperforms state-of-the-art single-modality and multi-modality methods."
                    },
                    {
                        "title": "Quadratic Programming for Continuous Control of Safety-Critical Multi-Agent Systems Under Uncertainty",
                        "abstract": "This paper studies the control problem for safety-critical multi-agent systems based on quadratic programming (QP). Each controlled agent is modeled as a cascade connection of an integrator and an uncertain nonlinear actuation system. In particular, the integrator represents the position-velocity relation, and the actuation system describes the dynamic response of the actual velocity to the velocity reference signal. The notion of input-to-output stability (IOS) is employed to characterize the essential velocity-tracking capability of the actuation system. The uncertain actuation dynamics may cause infeasibility or discontinuous solutions of QP algorithms for collision avoidance. Also, the interaction between the controlled integrator and the uncertain actuation dynamics may lead to significant robustness issues. By using nonlinear control methods and numerical optimization methods, this paper first contributes a new feasible-set reshaping technique and a refined QP algorithm for feasibility, robustness, and local Lipschitz continuity. Then, we present a nonlinear small-gain analysis to handle the inherent interaction for guaranteed safety of the closed-loop multi-agent system. The proposed methods are illustrated by numerical simulations and a physical experiment."
                    }
                ]
            },
            "f64acacc-afd4-4571-bbf0-e37a5152efe6": {
                "pk": "f64acacc-afd4-4571-bbf0-e37a5152efe6",
                "name": "Weiqiang Wang",
                "collaborators": [],
                "domain": [
                    "Lie Algebra",
                    "Representation Theory",
                    "Algebraic Geometry",
                    "Quantum Groups"
                ],
                "publications": [
                    {
                        "title": "Dimension of a minimal nilpotent orbit",
                        "abstract": "We show that the dimension of the minimal nilpotent coadjoint orbit for a complex simple Lie algebra is equal to twice the dual Coxeter number minus two."
                    },
                    {
                        "title": "Resolution of singularities of null cones",
                        "abstract": "We give canonical resolutions of singularities of several cone varieties arising from invariant theory. We establish a connection between our resolutions and resolutions of singularities of closure of conjugacy classes in classical Lie algebras."
                    },
                    {
                        "title": "Nilpotent orbits and finite W-algebras",
                        "abstract": "In recent years, the finite W-algebras associated to a semisimple Lie algebra and its nilpotent element have been studied intensively from different viewpoints. In this lecture series, we shall present some basic constructions, connections, and applications of finite W-algebras."
                    },
                    {
                        "title": "Some conjectures on modular representations of affine $\\mathfrak{sl}_2$ and Virasoro algebra",
                        "abstract": "We conjecture an explicit bound on the prime characteristic of a field, under which the Weyl modules of affine $\\mathfrak{sl}_2$ and the minimal series modules of Virasoro algebra remain irreducible, and Goddard-Kent-Olive coset construction for affine $\\mathfrak{sl}_2$ is valid."
                    },
                    {
                        "title": "Quantum symmetric pairs",
                        "abstract": "This is a survey of some recent progress on quantum symmetric pairs and applications. The topics include quasi K-matrices, $\\imath$Schur duality, canonical bases, super Kazhdan-Lusztig theory, $\\imath$Hall algebras, current presentations for affine $\\imath$quantum groups, and braid group actions."
                    },
                    {
                        "title": "Dual pairs and infinite dimensional Lie algebras",
                        "abstract": "We construct and study various dual pairs between finite dimensional classical Lie groups and infinite dimensional Lie algebras in some Fock representations. The infinite dimensional Lie algebras here can be either a completed infinite rank affine Lie algebra, the $\\winf$ algebra or its certain Lie subalgebras. We give a formulation in the framework of vertex algebras. We also formulate several conjectures and open problems."
                    },
                    {
                        "title": "Algebraic structures behind Hilbert schemes and wreath products",
                        "abstract": "In this paper we review various strikingly parallel algebraic structures behind Hilbert schemes of points on surfaces and certain finite groups called the wreath products. We explain connections among Hilbert schemes, wreath products, infinite-dimensional Lie algebras, and vertex algebras. As an application we further describe the cohomology ring structure of the Hilbert schemes. We organize this exposition around several general principles which have been supported by various works on these subjects."
                    },
                    {
                        "title": "Correlation functions of strict partitions and twisted Fock spaces",
                        "abstract": "Using twisted Fock spaces, we formulate and study two twisted versions of the n-point correlation functions of Bloch-Okounkov, and then identify them with q-expectation values of certain functions on the set of (odd) strict partitions. We find closed formulas for the 1-point functions in both cases in terms of Jacobi theta functions. These correlation functions afford several distinct interpretations."
                    },
                    {
                        "title": "Double affine Hecke algebras for the spin symmetric group",
                        "abstract": "We introduce a new class (in two versions) of rational double affine Hecke algebras (DaHa) associated to the spin symmetric group. We establish the basic properties of the algebras, such as PBW and Dunkl representation, and connections to Nazarov's degenerate affine Hecke-Clifford algebra and to a new degenerate affine Hecke algebra introduced here. We formulate a Morita equivalence between the two versions of rational DaHa's. The trigonometric generalization of the above constructions is also formulated and its relation to the rational counterpart is established."
                    },
                    {
                        "title": "Spin Hecke algebras of finite and affine types",
                        "abstract": "We introduce the spin Hecke algebra, which is a q-deformation of the spin symmetric group algebra, and its affine generalization. We establish an algebra isomorphism which relates our spin (affine) Hecke algebras to the (affine) Hecke-Clifford algebras of Olshanski and Jones-Nazarov. Relation between the spin (affine) Hecke algebra and a nonstandard presentation of the usual (affine) Hecke algebra is displayed, and the notion of covering (affine) Hecke algebra is introduced to provide a link between these algebras. Various algebraic structures for the spin (affine) Hecke algebra are established."
                    },
                    {
                        "title": "W{1+ \\infty} algebra, W_3 algebra, and Friedan-Martinec-Shenker bosonization",
                        "abstract": "We show that the vertex algebra W{1+ \\infty} with central charge -1 is isomorphic to a tensor product of the simple W_3 algebra with central charge -2 and a Heisenberg vertex algebra generated by a free bosonic field. We construct a family of irreducible modules of the W_3 algebra with central charge -2 in terms of free fields and calculate the full character formulas of these modules with respect to the full Cartan subalgebra of the W_3 algebra."
                    },
                    {
                        "title": "Classification of irreducible modules of W_3 algebra with c = -2",
                        "abstract": "We construct irreducible modules V_{\\alpha}, \\alpha \\in \\C over W_3 algebra with c = -2 in terms of a free bosonic field. We prove that these modules exhaust all the irreducible modules of W_3 algebra with c = -2. Highest weights of modules V_{\\alpha}, \\alpha \\in \\C with respect to the full (two-dimensional) Cartan subalgebra of W_3 algebra are (\\alpha(\\alpha -1)/2, \\alpha(\\alpha -1)(2\\alpha -1)/6). They are parametrized by points (t, w) on a rational curve w^2 - t^2 (8t + 1)/9 = 0. Irreducible modules of vertex algebra W_{1+\\infty} with c = -1 are also classified."
                    },
                    {
                        "title": "Duality in infinite dimensional Fock representations",
                        "abstract": "We construct and study in detail various dual pairs acting on some Fock representations between a finite dimensional Lie group and a completed infinite rank affine algebra associated to an infinite affine Cartan matrix. We give explicit decompositions of a Fock representation into a direct sum of irreducible isotypic subspaces with respect to the action of a dual pair, present explicit formulas for the common highest weight vectors and calculate the corresponding highest weights. We further outline applications of these dual pairs to the study of tensor products of modules of such an infinite dimensional Lie algebra."
                    },
                    {
                        "title": "Hilbert schemes, wreath products, and the McKay correspondence",
                        "abstract": "Various algebraic structures have recently appeared in a parallel way in the framework of Hilbert schemes of points on a surface and respectively in the framework of equivariant K-theory [N1,Gr,S2,W], but direct connections are yet to be clarified to explain such a coincidence. We provide several non-trivial steps toward establishing our main conjecture on the isomorphism between the Hilbert quotient of the affine space $\\C^{2n}$ by the wreath product $\\G ~ S_n$ and Hilbert schemes of points on the minimal resolution of a simple singularity $\\C^2 /\\G$. We discuss further various implications of our main conjecture. We obtain a key ingredient toward a direct isomorphism between two forms of McKay correspondence in terms of Hilbert schemes [N1, Gr, N2] and respectively of wreath products [FJW]. We in addition establish a direct identification of various algebraic structures appearing in two different setups of equivariant K-theory [S2, W]."
                    },
                    {
                        "title": "Equivariant K-theory, generalized symmetric products, and twisted Heisenberg algebra",
                        "abstract": "For a space X acted by a finite group $\\G$, the product space $X^n$ affords a natural action of the wreath product $\\Gn$. In this paper we study the K-groups $K_{\\tG_n}(X^n)$ of $\\Gn$-equivariant Clifford supermodules on $X^n$. We show that $\\tFG =\\bigoplus_{n\\ge 0}K_{\\tG_n}(X^n) \\otimes \\C$ is a Hopf algebra and it is isomorphic to the Fock space of a twisted Heisenberg algebra. Twisted vertex operators make a natural appearance. The algebraic structures on $\\tFG$, when $\\G$ is trivial and X is a point, specialize to those on a ring of symmetric functions with the Schur Q-functions as a linear basis. As a by-product, we present a novel construction of K-theory operations using the spin representations of the hyperoctahedral groups."
                    },
                    {
                        "title": "The Farahat-Higman ring of wreath products and Hilbert schemes",
                        "abstract": "We study the structure constants of the class algebra $R_Z(G_n)$ of the wreath products $G_n$ associated to an arbitrary finite group G with respect to the basis of conjugacy classes. We show that a suitable filtration on $R_Z(G_n)$ gives rise to the graded ring $\\mathcal G_G(n)$ with non-negative integer structure constants independent of n (some of which are computed), which are then encoded in a Farahat-Higman ring $\\mathcal G_G$. The real conjugacy classes of G come to play a distinguished role, and is treated in detail in the case when G is a subgroup of $SL_2(C)$. The above results provide new insight to the cohomology rings of Hilbert schemes of points on a quasi-projective surface."
                    },
                    {
                        "title": "Universal rings arising in geometry and group theory",
                        "abstract": "Various algebraic structures in geometry and group theory have appeared to be governed by certain universal rings. Examples include: the cohomology rings of Hilbert schemes of points on projective surfaces and quasi-projective surfaces; the Chen-Ruan orbifold cohomology rings of the symmetric products; the class algebras of wreath products, as well as their associated graded algebras with respect to a suitable filtration. We review these examples, and further provide a new elementary construction and explanation in the case of symmetric products. We in addition show that the Jucys-Murphy elements can be used to clarify the Macdonald's isomorphism between the FH-ring for the symmetric groups and the ring of symmetric functions."
                    },
                    {
                        "title": "Dual pairs and tensor categories of modules over Lie algebras gl_{\\infty} and W_{1 +\\infty}",
                        "abstract": "We introduce a tensor category O_+ (resp. O_{-}) of certain modules of gl_{\\infty} with non-negative (resp. non-positive) integral central charges with the usual tensor product. We also introduce a tensor category O_f consisting of certain modules over GL(N) for all N. We show that the tensor categories O_+, O_{-} and O_f are semisimple abelian and all equivalent to each other. We give a formula to decompose a tensor product of two modules in each of these categories. We also introduce a tensor category O^w of certain modules over W_{1 +\\infty} with non-negative integral central charges. We show that O^w is semisimple abelian and give an explicit formula to decompose a tensor product of two modules in O^w."
                    },
                    {
                        "title": "PBW bases for modified quantum groups",
                        "abstract": "We construct a basis for a modified quantum group of finite type, extending the PBW bases of positive and negative halves of a quantum group. Generalizing Lusztig's classic results on PBW bases, we show that this basis is orthogonal with respect to its natural bilinear form (and hence called a PBW basis), and moreover, the matrix for the PBW-expansion of the canonical basis is unital triangular. All these follow by a new construction of the modified quantum group of arbitrary type, which is built on limits of sequences of elements in tensor products of lowest and highest weight modules. Explicitly formulas are worked out in the rank one case."
                    },
                    {
                        "title": "Lagrangian construction of the $(gl_n, gl_m)$-duality",
                        "abstract": "We give a geometric realization of the symmetric algebra of the tensor space $C^n \\bigotimes C^m$ together with the action of the dual pair $(gl_n, gl_m)$ in terms of lagrangian cycles in the cotangent bundles of certain varieties. We establish geometrically the equivalence between the $(gl_n, gl_m)$ duality and Schur duality. We establish the connection between Springer's construction of (representations of) Weyl groups and Ginzburg's construction of (representations of) Lie algebras of type $A$."
                    }
                ]
            }
        }
    },
    "2404.19651": {
        "paper_data": {
            "title": "Provably Robust Conformal Prediction with Improved Efficiency",
            "url": "http://arxiv.org/abs/2404.19651v1",
            "arxiv_id": "2404.19651",
            "authors": [
                "Ge Yan",
                "Yaniv Romano",
                "Tsui-Wei Weng"
            ],
            "abstract": "Conformal prediction is a powerful tool to generate uncertainty sets with guaranteed coverage using any predictive model, under the assumption that the training and test data are i.i.d.. Recently, it has been shown that adversarial examples are able to manipulate conformal methods to construct prediction sets with invalid coverage rates, as the i.i.d. assumption is violated. To address this issue, a recent work, Randomized Smoothed Conformal Prediction (RSCP), was first proposed to certify the robustness of conformal prediction methods to adversarial noise. However, RSCP has two major limitations: (i) its robustness guarantee is flawed when used in practice and (ii) it tends to produce large uncertainty sets. To address these limitations, we first propose a novel framework called RSCP+ to provide provable robustness guarantee in evaluation, which fixes the issues in the original RSCP method. Next, we propose two novel methods, Post-Training Transformation (PTT) and Robust Conformal Training (RCT), to effectively reduce prediction set size with little computation overhead. Experimental results in CIFAR10, CIFAR100, and ImageNet suggest the baseline method only yields trivial predictions including full label set, while our methods could boost the efficiency by up to $4.36\\times$, $5.46\\times$, and $16.9\\times$ respectively and provide practical robustness guarantee. Our codes are available at https://github.com/Trustworthy-ML-Lab/Provably-Robust-Conformal-Prediction.",
            "introduction": "ABSTRACT Conformal prediction is a powerful tool to generate uncertainty sets with guaranteed coverage using any predictive model, under the assumption that the training and test data are i.i.d.. Recently, it has been shown that adversarial examples are able to manipulate conformalmethods, the last linear layer is finetuned for Nepoch= 150 epochs on 2 NVIDIA V100 GPUs. RCT takes around 95 minutes while ConfTr takes around 50 minutes. We argue that this computational overhead stems from randomized smoothing: training base model for randomized smoothing requires sampling a batch of examples for each training example (Cohen et al., 2019; Salman et al., 2019; Zhai et al., 2020), which significantly increases the computational expanse. Test-time cost. During test time, our RCT method is equivalent to vanilla RSCP (Gendler et al., 2021). No additional expense is needed. For PTT, since a transformation is applied on base score S at test time, there would be a computational overhead: see Fig. D.3. Diving deeper into each step, Figure D.3: Score computation time for each test example on CIFAR10 dataset. Comparing vanilla RSCP (RSCP) and RSCP with our training-free transformation (ours), it could be seen that the computational overhead is about 10%. Note that our robust conformal training method does not introduce additional computational cost at the test stage. The experiment is performed on 2 NVIDIA V100 GPU. we could see that the computational cost for each example is \u0398(NMCKlog|Dholdout |)for ranking transformation and \u0398(NMCK)for sigmoid transformation. Therefore, the overall cost would be \u0398(NMCKlog|Dholdout |). We emphasize that this cost does not scale with the cost of the base model, which is desirable because it enables users to employ arbitrarily large base models without increasing the computational overhead of PTT. 35Published as a conference paper at ICLR 2024 E O THER RELATED WORKS IN ROBUST CONFORMAL PREDICTION Besides Gendler et al. (2021), there are other related works that studied robust conformal prediction (Einbinder et al., 2022a; Ghosh et al., 2023; Bastani et al., 2022). Bastani et al. (2022) focused on an online setting where the defender could interact with the attacker and collect information on adversarial distribution from history inputs, which is different from our setting which assumes the defender does not have information on the distribution of adversarial noises, except the assumption that they have limited norm. Ghosh et al. (2023) studied probabilistically robust conformal prediction. Formally, they studied the following guarantee: PX,Y,\u03b4{yn+1\u2208C(\u02dcxn+1=xn+1+\u03b4)} \u22651\u2212\u03b1. (E.1) As Ghosh et al. (2023) pointed out, this could be regarded as a relaxed version of the adversarial robust coverage guarantee defined in Eq. (2): probabilistically robust conformal prediction provides a guarantee for perturbation \u03b4following a specific distribution, while our RSCP+ provide adversarial robustness guarantee for any perturbation that satisfies \u2225\u03b4\u22252\u2264\u03f5. 36Published as a conference paper at ICLR 2024 F T ABLE OF SYMBOLS Symbol Description C(x) Prediction set of input x (xn+1, yn+1)Test input and corresponding label Pxy Joint distribution of input xand label y \u02dcxn+1 Perturbed test input 1\u2212\u03b1 Target level of coverage \u03f5 Magnitude of adversarial noise C\u03f5(x) RSCP prediction set with noise level \u03f5 Dtrain Training dataset Dcal Calibration dataset \u02c6\u03c0 Probability output of classifier S (Non-)conformity score function \u02dcS Randomized smoothed score defined by RSCP (Gendler et al., 2021) \u03c4 Threshold of vanilla conformal prediction \u03c4adj Adjusted threshold introduced by RSCP (Gendler et al., 2021) \u03a6 Cumulative density function of standard Gaussian distribution SRS Randomized smoothed score defined in Eq. (9) \u02c6SRS Monte",
            "references": [
                {
                    "title": "Probabilistically robust conformal prediction",
                    "abstract": "Conformal prediction (CP) is a framework to quantify uncertainty of machine learning classifiers including deep neural networks. Given a testing example and a trained classifier, CP produces a prediction set of candidate labels with a user-specified coverage (i.e., true class label is contained with high probability). Almost all the existing work on CP assumes clean testing data and there is not much known about the robustness of CP algorithms w.r.t natural/adversarial perturbations to testing examples. This paper studies the problem of probabilistically robust conformal prediction (PRCP) which ensures robustness to most perturbations around clean input examples. PRCP generalizes the standard CP (cannot handle perturbations) and adversarially robust CP (ensures robustness w.r.t worst-case perturbations) to achieve better trade-offs between nominal performance and robustness. We propose a novel adaptive PRCP (aPRCP) algorithm to achieve probabilistically robust coverage. The key idea behind aPRCP is to determine two parallel thresholds, one for data samples and another one for the perturbations on data (aka\"quantile-of-quantile\"design). We provide theoretical analysis to show that aPRCP algorithm achieves robust coverage. Our experiments on CIFAR-10, CIFAR-100, and ImageNet datasets using deep neural networks demonstrate that aPRCP achieves better trade-offs than state-of-the-art CP and adversarially robust CP algorithms."
                },
                {
                    "title": "Practical Adversarial Multivalid Conformal Prediction",
                    "abstract": "We give a simple, generic conformal prediction method for sequential prediction that achieves target empirical coverage guarantees against adversarially chosen data. It is computationally lightweight -- comparable to split conformal prediction -- but does not require having a held-out validation set, and so all data can be used for training models from which to derive a conformal score. It gives stronger than marginal coverage guarantees in two ways. First, it gives threshold calibrated prediction sets that have correct empirical coverage even conditional on the threshold used to form the prediction set from the conformal score. Second, the user can specify an arbitrary collection of subsets of the feature space -- possibly intersecting -- and the coverage guarantees also hold conditional on membership in each of these subsets. We call our algorithm MVP, short for MultiValid Prediction. We give both theory and an extensive set of empirical evaluations."
                },
                {
                    "title": "Training Uncertainty-Aware Classifiers with Conformalized Deep Learning",
                    "abstract": "Deep neural networks are powerful tools to detect hidden patterns in data and leverage them to make predictions, but they are not designed to understand uncertainty and estimate reliable probabilities. In particular, they tend to be overconfident. We begin to address this problem in the context of multi-class classification by developing a novel training algorithm producing models with more dependable uncertainty estimates, without sacrificing predictive power. The idea is to mitigate overconfidence by minimizing a loss function, inspired by advances in conformal inference, that quantifies model uncertainty by carefully leveraging hold-out data. Experiments with synthetic and real data demonstrate this method can lead to smaller conformal prediction sets with higher conditional coverage, after exact calibration with hold-out data, compared to state-of-the-art alternatives."
                },
                {
                    "title": "Learning Optimal Conformal Classifiers",
                    "abstract": "Modern deep learning based classifiers show very high accuracy on test data but this does not provide sufficient guarantees for safe deployment, especially in high-stake AI applications such as medical diagnosis. Usually, predictions are obtained without a reliable uncertainty estimate or a formal guarantee. Conformal prediction (CP) addresses these issues by using the classifier's predictions, e.g., its probability estimates, to predict confidence sets containing the true class with a user-specified probability. However, using CP as a separate processing step after training prevents the underlying model from adapting to the prediction of confidence sets. Thus, this paper explores strategies to differentiate through CP during training with the goal of training model with the conformal wrapper end-to-end. In our approach, conformal training (ConfTr), we specifically\"simulate\"conformalization on mini-batches during training. Compared to standard training, ConfTr reduces the average confidence set size (inefficiency) of state-of-the-art CP methods applied after training. Moreover, it allows to\"shape\"the confidence sets predicted at test time, which is difficult for standard CP. On experiments with several datasets, we show ConfTr can influence how inefficiency is distributed across classes, or guide the composition of confidence sets in terms of the included classes, while retaining the guarantees offered by CP."
                },
                {
                    "title": "Uncertainty Sets for Image Classifiers using Conformal Prediction",
                    "abstract": "Convolutional image classifiers can achieve high predictive accuracy, but quantifying their uncertainty remains an unresolved challenge, hindering their deployment in consequential settings. Existing uncertainty quantification techniques, such as Platt scaling, attempt to calibrate the network's probability estimates, but they do not have formal guarantees. We present an algorithm that modifies any classifier to output a predictive set containing the true label with a user-specified probability, such as 90%. The algorithm is simple and fast like Platt scaling, but provides a formal finite-sample coverage guarantee for every model and dataset. Furthermore, our method generates much smaller predictive sets than alternative methods, since we introduce a regularizer to stabilize the small scores of unlikely classes after Platt scaling. In experiments on both Imagenet and Imagenet-V2 with a ResNet-152 and other classifiers, our scheme outperforms existing approaches, achieving exact coverage with sets that are often factors of 5 to 10 smaller."
                },
                {
                    "title": "Classification with Valid and Adaptive Coverage",
                    "abstract": "Conformal inference, cross-validation+, and the jackknife+ are hold-out methods that can be combined with virtually any machine learning algorithm to construct prediction sets with guaranteed marginal coverage. In this paper, we develop specialized versions of these techniques for categorical and unordered response labels that, in addition to providing marginal coverage, are also fully adaptive to complex data distributions, in the sense that they perform favorably in terms of approximate conditional coverage compared to alternative methods. The heart of our contribution is a novel conformity score, which we explicitly demonstrate to be powerful and intuitive for classification problems, but whose underlying principle is potentially far more general. Experiments on synthetic and real data demonstrate the practical value of our theoretical guarantees, as well as the statistical advantages of the proposed methods over the existing alternatives."
                },
                {
                    "title": "Exchangeability, Conformal Prediction, and Rank Tests",
                    "abstract": "Conformal prediction has been a very popular method of distribution-free predictive inference in recent years in machine learning and statistics. The main reason for its popularity comes from the fact that it works as a wrapper around any prediction algorithm such as neural networks or random forests. Exchangeability is at the core of the validity of conformal prediction. The concept of exchangeability is also at the core of rank tests widely known in nonparametric statistics. In this paper, we review the concept of exchangeability and discuss its implications for rank tests and conformal prediction. Although written as an exposition, the main message of the paper is to show that similar to conformal prediction, rank tests can also be used as a wrapper around any dimension reduction algorithm."
                },
                {
                    "title": "Fast Differentiable Sorting and Ranking",
                    "abstract": "The sorting operation is one of the most commonly used building blocks in computer programming. In machine learning, it is often used for robust statistics. However, seen as a function, it is piecewise linear and as a result includes many kinks where it is non-differentiable. More problematic is the related ranking operator, often used for order statistics and ranking metrics. It is a piecewise constant function, meaning that its derivatives are null or undefined. While numerous works have proposed differentiable proxies to sorting and ranking, they do not achieve the $O(n \\log n)$ time complexity one would expect from sorting and ranking operations. In this paper, we propose the first differentiable sorting and ranking operators with $O(n \\log n)$ time and $O(n)$ space complexity. Our proposal in addition enjoys exact computation and differentiation. We achieve this feat by constructing differentiable operators as projections onto the permutahedron, the convex hull of permutations, and using a reduction to isotonic optimization. Empirically, we confirm that our approach is an order of magnitude faster than existing approaches and showcase two novel applications: differentiable Spearman's rank correlation coefficient and least trimmed squares."
                },
                {
                    "title": "MACER: Attack-free and Scalable Robust Training via Maximizing Certified Radius",
                    "abstract": "Adversarial training is one of the most popular ways to learn robust models but is usually attack-dependent and time costly. In this paper, we propose the MACER algorithm, which learns robust models without using adversarial training but performs better than all existing provable l2-defenses. Recent work shows that randomized smoothing can be used to provide certified l2 radius to smoothed classifiers, and our algorithm trains provably robust smoothed classifiers via MAximizing the CErtified Radius (MACER). The attack-free characteristic makes MACER faster to train and easier to optimize. In our experiments, we show that our method can be applied to modern deep neural networks on a wide range of datasets, including Cifar-10, ImageNet, MNIST, and SVHN. For all tasks, MACER spends less training time than state-of-the-art adversarial training algorithms, and the learned models achieve larger average certified radius."
                },
                {
                    "title": "Provably Robust Deep Learning via Adversarially Trained Smoothed Classifiers",
                    "abstract": "Recent works have shown the effectiveness of randomized smoothing as a scalable technique for building neural network-based classifiers that are provably robust to $\\ell_2$-norm adversarial perturbations. In this paper, we employ adversarial training to improve the performance of randomized smoothing. We design an adapted attack for smoothed classifiers, and we show how this attack can be used in an adversarial training setting to boost the provable robustness of smoothed classifiers. We demonstrate through extensive experimentation that our method consistently outperforms all existing provably $\\ell_2$-robust classifiers by a significant margin on ImageNet and CIFAR-10, establishing the state-of-the-art for provable $\\ell_2$-defenses. Moreover, we find that pre-training and semi-supervised learning boost adversarially trained smoothed classifiers even further. Our code and trained models are available at this http URL ."
                },
                {
                    "title": "Differentiable Ranks and Sorting using Optimal Transport",
                    "abstract": "Sorting an array is a fundamental routine in machine learning, one that is used to compute rank-based statistics, cumulative distribution functions (CDFs), quantiles, or to select closest neighbors and labels. The sorting function is however piece-wise constant (the sorting permutation of a vector does not change if the entries of that vector are infinitesimally perturbed) and therefore has no gradient information to back-propagate. We propose a framework to sort elements that is algorithmically differentiable. We leverage the fact that sorting can be seen as a particular instance of the optimal transport (OT) problem on $\\mathbb{R}$, from input values to a predefined array of sorted values (e.g. $1,2,\\dots,n$ if the input array has $n$ elements). Building upon this link , we propose generalized CDFs and quantile operators by varying the size and weights of the target presorted array. Because this amounts to using the so-called Kantorovich formulation of OT, we call these quantities K-sorts, K-CDFs and K-quantiles. We recover differentiable algorithms by adding to the OT problem an entropic regularization, and approximate it using a few Sinkhorn iterations. We call these operators S-sorts, S-CDFs and S-quantiles, and use them in various learning settings: we benchmark them against the recently proposed neuralsort [Grover et al. 2019], propose applications to quantile regression and introduce differentiable formulations of the top-k accuracy that deliver state-of-the art performance."
                },
                {
                    "title": "Certified Adversarial Robustness via Randomized Smoothing",
                    "abstract": "We show how to turn any classifier that classifies well under Gaussian noise into a new classifier that is certifiably robust to adversarial perturbations under the $\\ell_2$ norm. This \"randomized smoothing\" technique has been proposed recently in the literature, but existing guarantees are loose. We prove a tight robustness guarantee in $\\ell_2$ norm for smoothing with Gaussian noise. We use randomized smoothing to obtain an ImageNet classifier with e.g. a certified top-1 accuracy of 49% under adversarial perturbations with $\\ell_2$ norm less than 0.5 (=127/255). No certified defense has been shown feasible on ImageNet except for smoothing. On smaller-scale datasets where competing approaches to certified $\\ell_2$ robustness are viable, smoothing delivers higher certified accuracies. Our strong empirical results suggest that randomized smoothing is a promising direction for future research into adversarially robust classification. Code and models are available at this http URL."
                },
                {
                    "title": "Least Ambiguous Set-Valued Classifiers With Bounded Error Levels",
                    "abstract": "ABSTRACT In most classification tasks, there are observations that are ambiguous and therefore difficult to correctly label. Set-valued classifiers output sets of plausible labels rather than a single label, thereby giving a more appropriate and informative treatment to the labeling of ambiguous instances. We introduce a framework for multiclass set-valued classification, where the classifiers guarantee user-defined levels of coverage or confidence (the probability that the true label is contained in the set) while minimizing the ambiguity (the expected size of the output). We first derive oracle classifiers assuming the true distribution to be known. We show that the oracle classifiers are obtained from level sets of the functions that define the conditional probability of each class. Then we develop estimators with good asymptotic and finite sample properties. The proposed estimators build on existing single-label classifiers. The optimal classifier can sometimes output the empty set, but we provide two solutions to fix this issue that are suitable for various practical needs. Supplementary materials for this article are available online."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Distribution\u2010free prediction bands for non\u2010parametric regression",
                    "abstract": "We study distribution\u2010free, non\u2010parametric prediction bands with a focus on their finite sample behaviour. First we investigate and develop different notions of finite sample coverage guarantees. Then we give a new prediction band by combining the idea of \u2018conformal prediction\u2019 with non\u2010parametric conditional density estimation. The proposed estimator, called COPS (conformal optimized prediction set), always has a finite sample guarantee. Under regularity conditions the estimator converges to an oracle band at a minimax optimal rate. A fast approximation algorithm and a data\u2010driven method for selecting the bandwidth are developed. The method is illustrated in simulated and real data examples."
                },
                {
                    "title": "Distribution-Free Prediction Sets",
                    "abstract": "This article introduces a new approach to prediction by bringing together two different nonparametric ideas: distribution-free inference and nonparametric smoothing. Specifically, we consider the problem of constructing nonparametric tolerance/prediction sets. We start from the general conformal prediction approach, and we use a kernel density estimator as a measure of agreement between a sample point and the underlying distribution. The resulting prediction set is shown to be closely related to plug-in density level sets with carefully chosen cutoff values. Under standard smoothness conditions, we get an asymptotic efficiency result that is near optimal for a wide range of function classes. But the coverage is guaranteed whether or not the smoothness conditions hold and regardless of the sample size. The performance of our method is investigated through simulation studies and illustrated in a real data example."
                },
                {
                    "title": "Randomized Smoothing for Stochastic Optimization",
                    "abstract": "We analyze convergence rates of stochastic optimization procedures for non-smooth convex optimization problems. By combining randomized smoothing techniques with accelerated gradient methods, we obtain convergence rates of stochastic optimization procedures, both in expectation and with high probability, that have optimal dependence on the variance of the gradient estimates. To the best of our knowledge, these are the first variance-based rates for non-smooth optimization. We give several applications of our results to statistical estimation problems, and provide experimental results that demonstrate the effectiveness of the proposed algorithms. We also describe how a combination of our algorithm with recent work on decentralized optimization yields a distributed stochastic optimization algorithm that is order-optimal."
                },
                {
                    "title": "Empirical Bernstein Bounds and Sample-Variance Penalization",
                    "abstract": "We give improved constants for data dependent and variance sensitive confidence bounds, called empirical Bernstein bounds, and extend these inequalities to hold uniformly over classes of functions whose growth function is polynomial in the sample size n. The bounds lead us to consider sample variance penalization, a novel learning method which takes into account the empirical variance of the loss function. We give conditions under which sample variance penalization is effective. In particular, we present a bound on the excess risk incurred by the method. Using this, we argue that there are situations in which the excess risk of our method is of order 1/n, while the excess risk of empirical risk minimization is of order 1/\u221an. We show some experimental results, which confirm the theory. Finally, we discuss the potential application of our results to sample compression schemes."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Adversarially Robust Conformal Prediction",
                    "abstract": "Conformal prediction is a model-agnostic tool for constructing prediction sets that are valid under the common i.i.d. assumption, which has been applied to quantify the prediction uncertainty of deep net classifiers. In this paper, we generalize this framework to the case where adversaries exist during inference time, under which the i.i.d. assumption is grossly violated. By combining conformal prediction with randomized smoothing, our proposed method forms a prediction set with finite-sample coverage guarantee that holds for any data distribution with `2norm bounded adversarial noise, generated by any adversarial attack algorithm. The core idea is to bound the Lipschitz constant of the non-conformity score by smoothing it with Gaussian noise and leverage this knowledge to account for the effect of the unknown adversarial perturbation. We demonstrate the necessity of our method in the adversarial setting and the validity of our theoretical guarantee on three widely used benchmark data sets: CIFAR10, CIFAR100, and ImageNet."
                },
                {
                    "title": "Conformal Prediction is Robust to Label Noise",
                    "abstract": "We study the robustness of conformal prediction\u2014a powerful tool for uncertainty quanti\ufb01cation\u2014to label noise. Our analysis tackles both regression and classi\ufb01cation problems, characterizing when and how it is possible to construct uncertainty sets that correctly cover the unobserved noiseless ground truth labels. Through stylized theoretical examples and practical experiments, we argue that na\u00a8\u0131ve conformal prediction covers the noiseless ground truth label unless the noise distribution is adversarially designed. This leads us to believe that correcting for label noise is unnecessary except for pathological data distributions or noise sources. In such cases, we can also correct for noise of bounded size in the conformal prediction algorithm in order to ensure correct coverage of the ground truth labels without score or data regularity."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we enhance the robustness of conformal prediction methods against adversarial examples while maintaining computational efficiency?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing concern of adversarial attacks on machine learning models, particularly in safety-critical applications. By improving the robustness of conformal prediction, we can ensure more reliable uncertainty quantification, which is essential for decision-making processes in fields such as healthcare, finance, and autonomous systems. This research could pave the way for future studies that explore more resilient predictive models and lead to practical applications where trustworthiness and safety are paramount.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of adversarial attacks, which can manipulate model predictions in subtle ways. Naive approaches may fail because they do not account for the diverse nature of adversarial perturbations, leading to inadequate coverage guarantees. Additionally, the computational overhead associated with randomized smoothing techniques complicates the implementation of robust conformal prediction, as it requires extensive sampling and can significantly increase training time. Overcoming these technical and practical obstacles is essential to develop an effective solution.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either conformal prediction or adversarial robustness, often treating them as separate domains. Limitations in existing solutions include a lack of comprehensive methods that integrate robust conformal prediction with adversarial noise handling. Barriers such as insufficient understanding of the interaction between conformal prediction and adversarial examples have hindered progress. Our approach differs by providing a robust conformal training method that does not introduce additional computational costs at test time, thus addressing both robustness and efficiency.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a robust conformal prediction (RSCP) framework that utilizes randomized smoothing to enhance adversarial robustness. We will employ datasets such as CIFAR10 for evaluation and measure performance using metrics like coverage probability and computational efficiency. The expected outcomes include demonstrating that our RSCP method maintains robust coverage guarantees against adversarial perturbations while minimizing computational overhead during both training and testing phases, thus enabling the use of larger base models without increased costs."
            }
        },
        "author_data": {
            "7671432b-f17b-4e09-ac13-d0500e6f6617": {
                "pk": "7671432b-f17b-4e09-ac13-d0500e6f6617",
                "name": "Ge Yan",
                "collaborators": [
                    "Divyansh Srivastava",
                    "Tsui-Wei Weng"
                ],
                "domain": [
                    "Interpretable Machine Learning",
                    "Vision-Language Models",
                    "Concept Bottleneck Models"
                ],
                "publications": [
                    {
                        "title": "VLG-CBM: Training Concept Bottleneck Models with Vision-Language Guidance",
                        "abstract": "Concept Bottleneck Models (CBMs) provide interpretable prediction by introducing an intermediate Concept Bottleneck Layer (CBL), which encodes human-understandable concepts to explain models' decision. Recent works proposed to utilize Large Language Models (LLMs) and pre-trained Vision-Language Models (VLMs) to automate the training of CBMs, making it more scalable and automated. However, existing approaches still fall short in two aspects: First, the concepts predicted by CBL often mismatch the input image, raising doubts about the faithfulness of interpretation. Second, it has been shown that concept values encode unintended information: even a set of random concepts could achieve comparable test accuracy to state-of-the-art CBMs. To address these critical limitations, in this work, we propose a novel framework called Vision-Language-Guided Concept Bottleneck Model (VLG-CBM) to enable faithful interpretability with the benefits of boosted performance. Our method leverages off-the-shelf open-domain grounded object detectors to provide visually grounded concept annotation, which largely enhances the faithfulness of concept prediction while further improving the model performance. In addition, we propose a new metric called Number of Effective Concepts (NEC) to control the information leakage and provide better interpretability. Extensive evaluations across five standard benchmarks show that our method, VLG-CBM, outperforms existing methods by at least 4.27% and up to 51.09% on accuracy at NEC=5, and by at least 0.45% and up to 29.78% on average accuracy across different NECs, while preserves both faithfulness and interpretability of the learned concepts as demonstrated in extensive experiments."
                    }
                ]
            },
            "46bdd25b-91bb-4737-a9fa-cc2b1ac0810b": {
                "pk": "46bdd25b-91bb-4737-a9fa-cc2b1ac0810b",
                "name": "Yaniv Romano",
                "collaborators": [
                    "Shai Feldman"
                ],
                "domain": [
                    "Conformal Prediction",
                    "Robust Machine Learning",
                    "Statistical Learning",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "Robust Conformal Prediction Using Privileged Information",
                        "abstract": "We develop a method to generate prediction sets with a guaranteed coverage rate that is robust to corruptions in the training data, such as missing or noisy variables. Our approach builds on conformal prediction, a powerful framework to construct prediction sets that are valid under the i.i.d assumption. Importantly, naively applying conformal prediction does not provide reliable predictions in this setting, due to the distribution shift induced by the corruptions. To account for the distribution shift, we assume access to privileged information (PI). The PI is formulated as additional features that explain the distribution shift, however, they are only available during training and absent at test time. We approach this problem by introducing a novel generalization of weighted conformal prediction and support our method with theoretical coverage guarantees. Empirical experiments on both real and synthetic datasets indicate that our approach achieves a valid coverage rate and constructs more informative predictions compared to existing methods, which are not supported by theoretical guarantees."
                    }
                ]
            },
            "a936d860-1046-4c78-b0c6-e0343e4c0d4d": {
                "pk": "a936d860-1046-4c78-b0c6-e0343e4c0d4d",
                "name": "Tsui-Wei Weng",
                "collaborators": [
                    "Tuomas P. Oikarinen",
                    "Lam M. Nguyen",
                    "Chung-En Sun",
                    "Subhro Das",
                    "Divyansh Srivastava",
                    "Wang Zhang",
                    "Lucani E. Daniel",
                    "Ge Yan",
                    "Xiaodong Liu",
                    "Weiwei Yang"
                ],
                "domain": [
                    "Interpretable Machine Learning",
                    "Adversarial Robustness",
                    "Neural Networks",
                    "Concept Bottleneck Models"
                ],
                "publications": [
                    {
                        "title": "VLG-CBM: Training Concept Bottleneck Models with Vision-Language Guidance",
                        "abstract": "Concept Bottleneck Models (CBMs) provide interpretable prediction by introducing an intermediate Concept Bottleneck Layer (CBL), which encodes human-understandable concepts to explain models' decision. Recent works proposed to utilize Large Language Models (LLMs) and pre-trained Vision-Language Models (VLMs) to automate the training of CBMs, making it more scalable and automated. However, existing approaches still fall short in two aspects: First, the concepts predicted by CBL often mismatch the input image, raising doubts about the faithfulness of interpretation. Second, it has been shown that concept values encode unintended information: even a set of random concepts could achieve comparable test accuracy to state-of-the-art CBMs. To address these critical limitations, in this work, we propose a novel framework called Vision-Language-Guided Concept Bottleneck Model (VLG-CBM) to enable faithful interpretability with the benefits of boosted performance. Our method leverages off-the-shelf open-domain grounded object detectors to provide visually grounded concept annotation, which largely enhances the faithfulness of concept prediction while further improving the model performance. In addition, we propose a new metric called Number of Effective Concepts (NEC) to control the information leakage and provide better interpretability. Extensive evaluations across five standard benchmarks show that our method, VLG-CBM, outperforms existing methods by at least 4.27% and up to 51.09% on accuracy at NEC=5, and by at least 0.45% and up to 29.78% on average accuracy across different NECs, while preserves both faithfulness and interpretability of the learned concepts as demonstrated in extensive experiments."
                    },
                    {
                        "title": "Iterative Self-Tuning LLMs for Enhanced Jailbreaking Capabilities",
                        "abstract": "Recent research has shown that Large Language Models (LLMs) are vulnerable to automated jailbreak attacks, where adversarial suffixes crafted by algorithms appended to harmful queries bypass safety alignment and trigger unintended responses. Current methods for generating these suffixes are computationally expensive and have low Attack Success Rates (ASR), especially against well-aligned models like Llama2 and Llama3. To overcome these limitations, we introduce ADV-LLM, an iterative self-tuning process that crafts adversarial LLMs with enhanced jailbreak ability. Our framework significantly reduces the computational cost of generating adversarial suffixes while achieving nearly 100\\% ASR on various open-source LLMs. Moreover, it exhibits strong attack transferability to closed-source models, achieving 99% ASR on GPT-3.5 and 49% ASR on GPT-4, despite being optimized solely on Llama3. Beyond improving jailbreak ability, ADV-LLM provides valuable insights for future safety alignment research through its ability to generate large datasets for studying LLM safety."
                    },
                    {
                        "title": "Towards Trustworthy Deep Learning",
                        "abstract": "Deep neural networks (DNNs) have achieved unprecedented success across many scientific and engineering fields in the last decades. Despite its empirical success, unfortunately, recent studies have shown that there are various failure modes and blindspots in DNN models which may result in unexpected serious failures and potential harms, e.g. the existence of adversarial examples and small perturbations. This is not acceptable especially for safety critical and high stakes applications in the real-world, including healthcare, self-driving cars, aircraft control systems, hiring and malware detection protocols. Moreover, it has been challenging to understand why and when DNNs will fail due to their complicated structures and black-box behaviors. Lacking interpretability is one critical issue that may seriously hinder the deployment of DNNs in high-stake applications, which need interpretability to trust the prediction, to understand potential failures, and to be able to mitigate harms and eliminate biases in the model.   To make DNNs trustworthy and reliable for deployment, it is necessary and urgent to develop methods and tools that can (i) quantify and improve their robustness against adversarial and natural perturbations, and (ii) understand their underlying behaviors and further correct errors to prevent injuries and damages. These are the important first steps to enable Trustworthy AI and Trustworthy Machine Learning. In this talk, I will survey a series of research efforts in my lab contributed to tackling the grand challenges in (i) and (ii). In the first part of my talk, I will overview our research effort in Robust Machine Learning since 2017, where we have proposed the first attack-agnostic robustness evaluation metric, the first efficient robustness certification algorithms for various types of perturbations, and efficient robust learning algorithms across supervised learning to deep reinforcement learning.    In the second part of my talk, I will survey a series of exciting results in my lab on accelerating interpretable machine learning and explainable AI. Specifically, I will show how we could bring interpretability into deep learning by leveraging recent advances in multi-modal models. I'll present recent works in our group on automatically dissecting neural networks with open vocabulary concepts, designing interpretable neural networks without concept labels, and briefly overview our recent efforts on demystifying black-box DNN training process, automated neuron explanations for Large Language Models and the first robustness evaluation of a family of neuron-level interpretation techniques."
                    },
                    {
                        "title": "Interpretability-Guided Test-Time Adversarial Defense",
                        "abstract": "We propose a novel and low-cost test-time adversarial defense by devising interpretability-guided neuron importance ranking methods to identify neurons important to the output classes. Our method is a training-free approach that can significantly improve the robustness-accuracy tradeoff while incurring minimal computational overhead. While being among the most efficient test-time defenses (4x faster), our method is also robust to a wide range of black-box, white-box, and adaptive attacks that break previous test-time defenses. We demonstrate the efficacy of our method for CIFAR10, CIFAR100, and ImageNet-1k on the standard RobustBench benchmark (with average gains of 2.6%, 4.9%, and 2.8% respectively). We also show improvements (average 1.5%) over the state-of-the-art test-time defenses even under strong adaptive attacks."
                    },
                    {
                        "title": "Breaking the Barrier: Enhanced Utility and Robustness in Smoothed DRL Agents",
                        "abstract": "Robustness remains a paramount concern in deep reinforcement learning (DRL), with randomized smoothing emerging as a key technique for enhancing this attribute. However, a notable gap exists in the performance of current smoothed DRL agents, often characterized by significantly low clean rewards and weak robustness. In response to this challenge, our study introduces innovative algorithms aimed at training effective smoothed robust DRL agents. We propose S-DQN and S-PPO, novel approaches that demonstrate remarkable improvements in clean rewards, empirical robustness, and robustness guarantee across standard RL benchmarks. Notably, our S-DQN and S-PPO agents not only significantly outperform existing smoothed agents by an average factor of $2.16\\times$ under the strongest attack, but also surpass previous robustly-trained agents by an average factor of $2.13\\times$. This represents a significant leap forward in the field. Furthermore, we introduce Smoothed Attack, which is $1.89\\times$ more effective in decreasing the rewards of smoothed agents than existing adversarial attacks."
                    },
                    {
                        "title": "AND: Audio Network Dissection for Interpreting Deep Acoustic Models",
                        "abstract": "Neuron-level interpretations aim to explain network behaviors and properties by investigating neurons responsive to specific perceptual or structural input patterns. Although there is emerging work in the vision and language domains, none is explored for acoustic models. To bridge the gap, we introduce $\\textit{AND}$, the first $\\textbf{A}$udio $\\textbf{N}$etwork $\\textbf{D}$issection framework that automatically establishes natural language explanations of acoustic neurons based on highly-responsive audio. $\\textit{AND}$ features the use of LLMs to summarize mutual acoustic features and identities among audio. Extensive experiments are conducted to verify $\\textit{AND}$'s precise and informative descriptions. In addition, we demonstrate a potential use of $\\textit{AND}$ for audio machine unlearning by conducting concept-specific pruning based on the generated descriptions. Finally, we highlight two acoustic model behaviors with analysis by $\\textit{AND}$: (i) models discriminate audio with a combination of basic acoustic features rather than high-level abstract concepts; (ii) training strategies affect model behaviors and neuron interpretability -- supervised training guides neurons to gradually narrow their attention, while self-supervised learning encourages neurons to be polysemantic for exploring high-level features."
                    },
                    {
                        "title": "Linear Explanations for Individual Neurons",
                        "abstract": "In recent years many methods have been developed to understand the internal workings of neural networks, often by describing the function of individual neurons in the model. However, these methods typically only focus on explaining the very highest activations of a neuron. In this paper we show this is not sufficient, and that the highest activation range is only responsible for a very small percentage of the neuron's causal effect. In addition, inputs causing lower activations are often very different and can't be reliably predicted by only looking at high activations. We propose that neurons should instead be understood as a linear combination of concepts, and develop an efficient method for producing these linear explanations. In addition, we show how to automatically evaluate description quality using simulation, i.e. predicting neuron activations on unseen inputs in vision setting."
                    },
                    {
                        "title": "Describe-and-Dissect: Interpreting Neurons in Vision Networks with Language Models",
                        "abstract": "In this paper, we propose Describe-and-Dissect (DnD), a novel method to describe the roles of hidden neurons in vision networks. DnD utilizes recent advancements in multimodal deep learning to produce complex natural language descriptions, without the need for labeled training data or a predefined set of concepts to choose from. Additionally, DnD is training-free, meaning we don't train any new models and can easily leverage more capable general purpose models in the future. We have conducted extensive qualitative and quantitative analysis to show that DnD outperforms prior work by providing higher quality neuron descriptions. Specifically, our method on average provides the highest quality labels and is more than 2 times as likely to be selected as the best explanation for a neuron than the best baseline."
                    },
                    {
                        "title": "Crafting Large Language Models for Enhanced Interpretability",
                        "abstract": "We introduce the Concept Bottleneck Large Language Model (CB-LLM), a pioneering approach to creating inherently interpretable Large Language Models (LLMs). Unlike traditional black-box LLMs that rely on post-hoc interpretation methods with limited neuron function insights, CB-LLM sets a new standard with its built-in interpretability, scalability, and ability to provide clear, accurate explanations. This innovation not only advances transparency in language models but also enhances their effectiveness. Our unique Automatic Concept Correction (ACC) strategy successfully narrows the performance gap with conventional black-box LLMs, positioning CB-LLM as a model that combines the high accuracy of traditional LLMs with the added benefit of clear interpretability -- a feature markedly absent in existing LLMs."
                    },
                    {
                        "title": "Promoting Robustness of Randomized Smoothing: Two Cost-Effective Approaches",
                        "abstract": "Randomized smoothing has recently attracted attentions in the field of adversarial robustness to provide provable robustness guarantees on smoothed neural network classifiers. However, existing works show that vanillarandomized smoothing usually does not provide good robustness performance and often requires (re)training techniques on the base classifier in order to boost the robustness of the resulting smoothed classifier. In this work, we propose two cost-effective approaches to boost the robustness of randomized smoothing while preserving its clean performance. The first approach introduces a new robust training method AdvMacer which combines adversarial training and robustness certification maximization for randomized smoothing. We show that AdvMacer can improve the robustness performance of randomized smoothing classifiers compared to SOTA baselines, while being 3x faster to train than MACER baseline. The second approach introduces a post-processing method EsbRS which greatly improves the robustness certificate based on building model ensembles. Extensive experiments verify the superior performance of our methods on various datasets. Our code is available at https://github.com/Trustworthy-ML-Lab/AdvMacer_and_EsbRS."
                    },
                    {
                        "title": "Prediction without Preclusion: Recourse Verification with Reachable Sets",
                        "abstract": "Machine learning models are often used to decide who receives a loan, a job interview, or a public benefit. Models in such settings use features without considering their actionability. As a result, they can assign predictions that are fixed $-$ meaning that individuals who are denied loans and interviews are, in fact, precluded from access to credit and employment. In this work, we introduce a procedure called recourse verification to test if a model assigns fixed predictions to its decision subjects. We propose a model-agnostic approach for recourse verification with reachable sets $-$ i.e., the set of all points that a person can reach through their actions in feature space. We develop methods to construct reachable sets for discrete feature spaces, which can certify the responsiveness of any model by simply querying its predictions. We conduct a comprehensive empirical study on the infeasibility of recourse on datasets from consumer finance. Our results highlight how models can inadvertently preclude access by assigning fixed predictions and underscore the need to account for actionability in model development."
                    },
                    {
                        "title": "Label-Free Concept Bottleneck Models",
                        "abstract": "Concept bottleneck models (CBM) are a popular way of creating more interpretable neural networks by having hidden layer neurons correspond to human-understandable concepts. However, existing CBMs and their variants have two crucial limitations: first, they need to collect labeled data for each of the predefined concepts, which is time consuming and labor intensive; second, the accuracy of a CBM is often significantly lower than that of a standard neural network, especially on more complex datasets. This poor performance creates a barrier for adopting CBMs in practical real world applications. Motivated by these challenges, we propose Label-free CBM which is a novel framework to transform any neural network into an interpretable CBM without labeled concept data, while retaining a high accuracy. Our Label-free CBM has many advantages, it is: scalable - we present the first CBM scaled to ImageNet, efficient - creating a CBM takes only a few hours even for very large datasets, and automated - training it for a new dataset requires minimal human effort. Our code is available at https://github.com/Trustworthy-ML-Lab/Label-free-CBM. Finally, in Appendix B we conduct a large scale user evaluation of the interpretability of our method."
                    },
                    {
                        "title": "Concept-Monitor: Understanding DNN training through individual neurons",
                        "abstract": "In this work, we propose a general framework called Concept-Monitor to help demystify the black-box DNN training processes automatically using a novel unified embedding space and concept diversity metric. Concept-Monitor enables human-interpretable visualization and indicators of the DNN training processes and facilitates transparency as well as deeper understanding on how DNNs develop along the during training. Inspired by these findings, we also propose a new training regularizer that incentivizes hidden neurons to learn diverse concepts, which we show to improve training performance. Finally, we apply Concept-Monitor to conduct several case studies on different training paradigms including adversarial training, fine-tuning and network pruning via the Lottery Ticket Hypothesis"
                    },
                    {
                        "title": "One step closer to unbiased aleatoric uncertainty estimation",
                        "abstract": "Neural networks are powerful tools in various applications, and quantifying their uncertainty is crucial for reliable decision-making. In the deep learning field, the uncertainties are usually categorized into aleatoric (data) and epistemic (model) uncertainty. In this paper, we point out that the existing popular variance attenuation method highly overestimates aleatoric uncertainty. To address this issue, we proposed a new estimation method by actively de-noising the observed data. By conducting a broad range of experiments, we demonstrate that our proposed approach provides a much closer approximation to the actual data uncertainty than the standard method."
                    },
                    {
                        "title": "Corrupting Neuron Explanations of Deep Visual Features",
                        "abstract": "The inability of DNNs to explain their black-box behavior has led to a recent surge of explainability methods. However, there are growing concerns that these explainability methods are not robust and trustworthy. In this work, we perform the first robustness analysis of Neuron Explanation Methods under a unified pipeline and show that these explanations can be significantly corrupted by random noises and well-designed perturbations added to their probing data. We find that even adding small random noise with a standard deviation of 0.02 can already change the assigned concepts of up to 28% neurons in the deeper layers. Furthermore, we devise a novel corruption algorithm and show that our algorithm can manipulate the explanation of more than 80% neurons by poisoning less than 10% of probing data. This raises the concern of trusting Neuron Explanation Methods in real-life safety and fairness critical applications."
                    },
                    {
                        "title": "Constructive Assimilation: Boosting Contrastive Learning Performance through View Generation Strategies",
                        "abstract": "Transformations based on domain expertise (expert transformations), such as random-resized-crop and color-jitter, have proven critical to the success of contrastive learning techniques such as SimCLR. Recently, several attempts have been made to replace such domain-specific, human-designed transformations with generated views that are learned. However for imagery data, so far none of these view-generation methods has been able to outperform expert transformations. In this work, we tackle a different question: instead of replacing expert transformations with generated views, can we constructively assimilate generated views with expert transformations? We answer this question in the affirmative and propose a view generation method and a simple, effective assimilation method that together improve the state-of-the-art by up to ~3.6% on three different datasets. Importantly, we conduct a detailed empirical study that systematically analyzes a range of view generation and assimilation methods and provides a holistic picture of the efficacy of learned views in contrastive representation learning."
                    },
                    {
                        "title": "The Importance of Prompt Tuning for Automated Neuron Explanations",
                        "abstract": "Recent advances have greatly increased the capabilities of large language models (LLMs), but our understanding of the models and their safety has not progressed as fast. In this paper we aim to understand LLMs deeper by studying their individual neurons. We build upon previous work showing large language models such as GPT-4 can be useful in explaining what each neuron in a language model does. Specifically, we analyze the effect of the prompt used to generate explanations and show that reformatting the explanation prompt in a more natural way can significantly improve neuron explanation quality and greatly reduce computational cost. We demonstrate the effects of our new prompts in three different ways, incorporating both automated and human evaluations."
                    },
                    {
                        "title": "ConCerNet: A Contrastive Learning Based Framework for Automated Conservation Law Discovery and Trustworthy Dynamical System Prediction",
                        "abstract": "Deep neural networks (DNN) have shown great capacity of modeling a dynamical system; nevertheless, they usually do not obey physics constraints such as conservation laws. This paper proposes a new learning framework named ConCerNet to improve the trustworthiness of the DNN based dynamics modeling to endow the invariant properties. ConCerNet consists of two steps: (i) a contrastive learning method to automatically capture the system invariants (i.e. conservation properties) along the trajectory observations; (ii) a neural projection layer to guarantee that the learned dynamics models preserve the learned invariants. We theoretically prove the functional relationship between the learned latent representation and the unknown system invariant function. Experiments show that our method consistently outperforms the baseline neural networks in both coordinate error and conservation metrics by a large margin. With neural network based parameterization and no dependence on prior knowledge, our method can be extended to complex and large-scale dynamics by leveraging an autoencoder."
                    },
                    {
                        "title": "CLIP-D ISSECT : A UTOMATIC D ESCRIPTION OF N EU - RON R EPRESENTATIONS IN D EEP V ISION N ETWORKS",
                        "abstract": "In this paper"
                    }
                ]
            }
        }
    },
    "2403.08757": {
        "paper_data": {
            "title": "Efficient Combinatorial Optimization via Heat Diffusion",
            "url": "http://arxiv.org/abs/2403.08757v4",
            "arxiv_id": "2403.08757",
            "authors": [
                "Hengyuan Ma",
                "Wenlian Lu",
                "Jianfeng Feng"
            ],
            "abstract": "Combinatorial optimization problems are widespread but inherently challenging due to their discrete nature. The primary limitation of existing methods is that they can only access a small fraction of the solution space at each iteration, resulting in limited efficiency for searching the global optimal. To overcome this challenge, diverging from conventional efforts of expanding the solver's search scope, we focus on enabling information to actively propagate to the solver through heat diffusion. By transforming the target function while preserving its optima, heat diffusion facilitates information flow from distant regions to the solver, providing more efficient navigation. Utilizing heat diffusion, we propose a framework for solving general combinatorial optimization problems. The proposed methodology demonstrates superior performance across a range of the most challenging and widely encountered combinatorial optimizations. Echoing recent advancements in harnessing thermodynamics for generative artificial intelligence, our study further reveals its significant potential in advancing combinatorial optimization.",
            "introduction": "   1 Introduction  Combinatorial optimization problems are prevalent in various applications, encompassing circuit design\u00a0[1], machine learning\u00a0[2], computer vision\u00a0[3], molecular dynamics simulation\u00a0[4], traffic flow optimization\u00a0[5], and financial risk analysis\u00a0[6]. This widespread application creates a significant demand for accelerated solutions to these problems. Alongside classical algorithms, which encompass both exact solvers and metaheuristics\u00a0[7], recent years have seen remarkable advancements in addressing combinatorial optimization. These include quantum adiabatic approaches\u00a0[8, 9, 10], simulated bifurcation\u00a0[11, 12, 13], coherent Ising machine\u00a0[14, 15], high-order Ising machine\u00a0[16], and deep learning techniques\u00a0[17, 18]. However, due to the exponential growth of the solution number, finding the optima within a limited computational budget remains a daunting challenge.   Our primary focus is on iterative approximation solvers, which constitute a significant class of combinatorial optimization methods. An iterative approximation solvers typically begin with an initial solution and iteratively improve it by finding better solutions within the neighborhood of the current solution, known as the search scope or more vividly, receptive field. However, due to combinatorial explosion, as the scope of the receptive field increases, the number of solutions to be assessed grows exponentially, making a thorough evaluation of all these solutions expensive. As a result, current approaches are limited to a narrow receptive field, rendering them blind to distant regions in the solution space and heightening the risk of getting trapped in local minimas or areas with bumpy landscapes. Although methods like large neighborhood search\u00a0[19], variable neighborhood search\u00a0[20] and path auxiliary sampling\u00a0[21] are designed to broaden the search scope, they can only gather a modest increment of information from the expanded search scope. Consequently, the current solvers\u2019 receptive field remains significantly constrained, impeding their efficiency.   Figure 1: The Heat diffusion optimization\u00a0(HeO)\u00a0framework. The efficiency of searching a key in a dark room is significantly improved by employing navigation that utilizes heat emission from the key. In combinatorial optimization, heat diffusion transforms the target function into different versions while preserving the location of the optima. Therefore, the gradient information of these transformed functions cooperatively help to optimize the original target function.    In this study, we approach the prevalent limitation stated above from a unique perspective. Instead of expanding the solver\u2019s receptive field to acquire more information from the solution space, we concentrate on propagating information from distant areas of the solution space to the solver via heat diffusion\u00a0[22]. To illustrate, imagine a solver searching for the optima in the solution space akin to a person searching for a key in a dark room, as depicted in Fig.\u00a01. Without light, the person is compelled to rely solely on touching his surrounding space. The tactile provides only localized information, leading to inefficient navigation. This mirrors the current situation in combinatorial optimization, wherein the receptive field is predominantly confined to local information. However, if the key were to emit heat, its radiating warmth would be perceptible from a distance, acting as a directional beacon. This would significantly enhance navigational efficiency for finding the key.   Motivated by the above metaphor, we propose a simple but efficient framework utilizing heat diffusion to solve various combinatorial optimization problems. Heat diffusion transforms the target function into different versions, within which the information from distant regions",
            "references": [
                {
                    "title": "Error bounds, PL condition, and quadratic growth for weakly convex functions, and linear convergences of proximal point methods",
                    "abstract": "Many practical optimization problems lack strong convexity. Fortunately, recent studies have revealed that first-order algorithms also enjoy linear convergences under various weaker regularity conditions. While the relationship among different conditions for convex and smooth functions is well-understood, it is not the case for the nonsmooth setting. In this paper, we go beyond convexity and smoothness, and clarify the connections among common regularity conditions in the class of weakly convex functions, including $\\textit{strong convexity}$, $\\textit{restricted secant inequality}$, $\\textit{subdifferential error bound}$, $\\textit{Polyak-{\\L}ojasiewicz inequality}$, and $\\textit{quadratic growth}$. In addition, using these regularity conditions, we present a simple and modular proof for the linear convergence of the proximal point method (PPM) for convex and weakly convex optimization problems. The linear convergence also holds when the subproblems of PPM are solved inexactly with a proper control of inexactness."
                },
                {
                    "title": "DIFUSCO: Graph-based Diffusion Solvers for Combinatorial Optimization",
                    "abstract": "Neural network-based Combinatorial Optimization (CO) methods have shown promising results in solving various NP-complete (NPC) problems without relying on hand-crafted domain knowledge. This paper broadens the current scope of neural solvers for NPC problems by introducing a new graph-based diffusion framework, namely DIFUSCO. Our framework casts NPC problems as discrete {0, 1}-vector optimization problems and leverages graph-based denoising diffusion models to generate high-quality solutions. We investigate two types of diffusion models with Gaussian and Bernoulli noise, respectively, and devise an effective inference schedule to enhance the solution quality. We evaluate our methods on two well-studied NPC combinatorial optimization problems: Traveling Salesman Problem (TSP) and Maximal Independent Set (MIS). Experimental results show that DIFUSCO strongly outperforms the previous state-of-the-art neural solvers, improving the performance gap between ground-truth and neural solvers from 1.76% to 0.46% on TSP-500, from 2.46% to 1.17% on TSP-1000, and from 3.19% to 2.58% on TSP10000. For the MIS problem, DIFUSCO outperforms the previous state-of-the-art neural solver on the challenging SATLIB benchmark."
                },
                {
                    "title": "A Unified Analysis of Descent Sequences in Weakly Convex Optimization, Including Convergence Rates for Bundle Methods",
                    "abstract": ","
                },
                {
                    "title": "Diffusion Models: A Comprehensive Survey of Methods and Applications",
                    "abstract": "Diffusion models have emerged as a powerful new family of deep generative models with record-breaking performance in many applications, including image synthesis, video generation, and molecule design. In this survey, we provide an overview of the rapidly expanding body of work on diffusion models, categorizing the research into three key areas: efficient sampling, improved likelihood estimation, and handling data with special structures. We also discuss the potential for combining diffusion models with other generative models for enhanced results. We further review the wide-ranging applications of diffusion models in fields spanning from computer vision, natural language processing, temporal data modeling, to interdisciplinary applications in other scientific disciplines. This survey aims to provide a contextualized, in-depth look at the state of diffusion models, identifying the key areas of focus and pointing to potential areas for further exploration. Github: https://github.com/YangLing0818/Diffusion-Models-Papers-Survey-Taxonomy"
                },
                {
                    "title": "Unsupervised Learning for Combinatorial Optimization with Principled Objective Relaxation",
                    "abstract": "Using machine learning to solve combinatorial optimization (CO) problems is challenging, especially when the data is unlabeled. This work proposes an unsupervised learning framework for CO problems. Our framework follows a standard relaxation-plus-rounding approach and adopts neural networks to parameterize the relaxed solutions so that simple back-propagation can train the model end-to-end. Our key contribution is the observation that if the relaxed objective satisfies entry-wise concavity, a low optimization loss guarantees the quality of the final integral solutions. This observation significantly broadens the applicability of the previous framework inspired by Erdos' probabilistic method. In particular, this observation can guide the design of objective models in applications where the objectives are not given explicitly while requiring being modeled in prior. We evaluate our framework by solving a synthetic graph optimization problem, and two real-world applications including resource allocation in circuit design and approximate computing. Our framework largely outperforms the baselines based on na\\\"{i}ve relaxation, reinforcement learning, and Gumbel-softmax tricks."
                },
                {
                    "title": "Accelerating Score-based Generative Models with Preconditioned Diffusion Sampling",
                    "abstract": "Score-based generative models (SGMs) have recently emerged as a promising class of generative models. However, a fundamental limitation is that their inference is very slow due to a need for many (e.g., 2000) iterations of sequential computations. An intuitive acceleration method is to reduce the sampling iterations which however causes severe performance degradation. We investigate this problem by viewing the diffusion sampling process as a Metropolis adjusted Langevin algorithm, which helps reveal the underlying cause to be ill-conditioned curvature. Under this insight, we propose a model-agnostic preconditioned diffusion sampling (PDS) method that leverages matrix preconditioning to alleviate the aforementioned problem. Crucially, PDS is proven theoretically to converge to the original target distribution of a SGM, no need for retraining. Extensive experiments on three image datasets with a variety of resolutions and diversity validate that PDS consistently accelerates off-the-shelf SGMs whilst maintaining the synthesis quality. In particular, PDS can accelerate by up to 29x on more challenging high resolution (1024x1024) image generation."
                },
                {
                    "title": "Quadratic Unconstrained Binary Optimization via Quantum-Inspired Annealing",
                    "abstract": "We present a classical algorithm to find approximate solutions to instances of quadratic unconstrained binary optimisation. The algorithm can be seen as an analogue of quantum annealing under the restriction of a product state space, where the dynamical evolution in quantum annealing is replaced with a gradient-descent based method. This formulation is able to quickly find high-quality solutions to large-scale problem instances, and can naturally be accelerated by dedicated hardware such as graphics processing units. We benchmark our approach for large scale problem instances with tuneable hardness and planted solutions. We find that our algorithm offers a similar performance to current state of the art approaches within a comparably simple gradient-based and non-stochastic setting."
                },
                {
                    "title": "A Survey of Quantization Methods for Efficient Neural Network Inference",
                    "abstract": "As soon as abstract mathematical computations were adapted to computation on digital computers, the problem of efficient representation, manipulation, and communication of the numerical values in those computations arose. Strongly related to the problem of numerical representation is the problem of quantization: in what manner should a set of continuous real-valued numbers be distributed over a fixed discrete set of numbers to minimize the number of bits required and also to maximize the accuracy of the attendant computations? This perennial problem of quantization is particularly relevant whenever memory and/or computational resources are severely restricted, and it has come to the forefront in recent years due to the remarkable performance of Neural Network models in computer vision, natural language processing, and related areas. Moving from floating-point representations to low-precision fixed integer values represented in four bits or less holds the potential to reduce the memory footprint and latency by a factor of 16x; and, in fact, reductions of 4x to 8x are often realized in practice in these applications. Thus, it is not surprising that quantization has emerged recently as an important and very active sub-area of research in the efficient implementation of computations associated with Neural Networks. In this article, we survey approaches to the problem of quantizing the numerical values in deep Neural Network computations, covering the advantages/disadvantages of current methods. With this survey and its organization, we hope to have presented a useful snapshot of the current research in quantization for Neural Networks and to have given an intelligent organization to ease the evaluation of future research in this area."
                },
                {
                    "title": "Oops I Took A Gradient: Scalable Sampling for Discrete Distributions",
                    "abstract": "We propose a general and scalable approximate sampling strategy for probabilistic models with discrete variables. Our approach uses gradients of the likelihood function with respect to its discrete inputs to propose updates in a Metropolis-Hastings sampler. We show empirically that this approach outperforms generic samplers in a number of difficult settings including Ising models, Potts models, restricted Boltzmann machines, and factorial hidden Markov models. We also demonstrate the use of our improved sampler for training deep energy-based models on high dimensional discrete data. This approach outperforms variational auto-encoders and existing energy-based models. Finally, we give bounds showing that our approach is near-optimal in the class of samplers which propose local updates."
                },
                {
                    "title": "High-performance combinatorial optimization based on classical mechanics",
                    "abstract": "Massively parallel implementations of algorithms based on classical mechanics realize high-performance combinatorial optimization. Quickly obtaining optimal solutions of combinatorial optimization problems has tremendous value but is extremely difficult. Thus, various kinds of machines specially designed for combinatorial optimization have recently been proposed and developed. Toward the realization of higher-performance machines, here, we propose an algorithm based on classical mechanics, which is obtained by modifying a previously proposed algorithm called simulated bifurcation. Our proposed algorithm allows us to achieve not only high speed by parallel computing but also high solution accuracy for problems with up to one million binary variables. Benchmarking shows that our machine based on the algorithm achieves high performance compared to recently developed machines, including a quantum annealer using a superconducting circuit, a coherent Ising machine using a laser, and digital processors based on various algorithms. Thus, high-performance combinatorial optimization is realized by massively parallel implementations of the proposed algorithm based on classical mechanics."
                },
                {
                    "title": "Erdos Goes Neural: an Unsupervised Learning Framework for Combinatorial Optimization on Graphs",
                    "abstract": "Combinatorial optimization problems are notoriously challenging for neural networks, especially in the absence of labeled instances. This work proposes an unsupervised learning framework for CO problems on graphs that can provide integral solutions of certified quality. Inspired by Erdos' probabilistic method, we use a neural network to parametrize a probability distribution over sets. Crucially, we show that when the network is optimized w.r.t. a suitably chosen loss, the learned distribution contains, with controlled probability, a low-cost integral solution that obeys the constraints of the combinatorial problem. The probabilistic proof of existence is then derandomized to decode the desired solutions. We demonstrate the efficacy of this approach to obtain valid solutions to the maximum clique problem and to perform local graph clustering. Our method achieves competitive results on both real datasets and synthetic hard instances."
                },
                {
                    "title": "A Survey on Sparse Learning Models for Feature Selection",
                    "abstract": "Feature selection is important in both machine learning and pattern recognition. Successfully selecting informative features can significantly increase learning accuracy and improve result comprehensibility. Various methods have been proposed to identify informative features from high-dimensional data by removing redundant and irrelevant features to improve classification accuracy. In this article, we systematically survey existing sparse learning models for feature selection from the perspectives of individual sparse feature selection and group sparse feature selection, and analyze the differences and connections among various sparse learning models. Promising research directions and topics on sparse learning models are analyzed."
                },
                {
                    "title": "Compressed Quadratization of Higher Order Binary Optimization Problems",
                    "abstract": "Recent hardware advances in quantum and quantum-inspired annealers promise substantial speedup for solving NP-hard combinatorial optimization problems compared to general-purpose computers. These special-purpose hardware are built for solving hard instances of Quadratic Unconstrained Binary Optimization (QUBO) problems. In terms of number of variables and precision of these hardware are usually resource-constrained and they work either in Ising space or in Boolean space. Many naturally occurring problem instances are higher-order in nature. The known method to reduce the degree of a higher-order optimization problem uses Rosenberg's polynomial. The method works in Boolean space by reducing the degree of one term by introducing one extra variable. In this work, we prove that in Ising space the degree reduction of one term requires the introduction of two variables. Our proposed method of degree reduction works directly in Ising space, as opposed to converting an Ising polynomial to Boolean space and applying previously known Rosenberg's polynomial. For sparse higher-order Ising problems, this results in a more compact representation of the resultant QUBO problem, which is crucial for utilizing resource-constrained QUBO solvers."
                },
                {
                    "title": "Network of Evolvable Neural Units: Evolving to Learn at a Synaptic Level",
                    "abstract": "Although Deep Neural Networks have seen great success in recent years through various changes in overall architectures and optimization strategies, their fundamental underlying design remains largely unchanged. Computational neuroscience on the other hand provides more biologically realistic models of neural processing mechanisms, but they are still high level abstractions of the actual experimentally observed behaviour. Here a model is proposed that bridges Neuroscience, Machine Learning and Evolutionary Algorithms to evolve individual soma and synaptic compartment models of neurons in a scalable manner. Instead of attempting to manually derive models for all the observed complexity and diversity in neural processing, we propose an Evolvable Neural Unit (ENU) that can approximate the function of each individual neuron and synapse. We demonstrate that this type of unit can be evolved to mimic Integrate-And-Fire neurons and synaptic Spike-Timing-Dependent Plasticity. Additionally, by constructing a new type of neural network where each synapse and neuron is such an evolvable neural unit, we show it is possible to evolve an agent capable of learning to solve a T-maze environment task. This network independently discovers spiking dynamics and reinforcement type learning rules, opening up a new path towards biologically inspired artificial intelligence."
                },
                {
                    "title": "Monte Carlo Gradient Estimation in Machine Learning",
                    "abstract": "This paper is a broad and accessible survey of the methods we have at our disposal for Monte Carlo gradient estimation in machine learning and across the statistical sciences: the problem of computing the gradient of an expectation of a function with respect to parameters defining the distribution that is integrated; the problem of sensitivity analysis. In machine learning research, this gradient problem lies at the core of many learning problems, in supervised, unsupervised and reinforcement learning. We will generally seek to rewrite such gradients in a form that allows for Monte Carlo estimation, allowing them to be easily and efficiently used and analysed. We explore three strategies--the pathwise, score function, and measure-valued gradient estimators--exploring their historical developments, derivation, and underlying assumptions. We describe their use in other fields, show how they are related and can be combined, and expand on their possible generalisations. Wherever Monte Carlo gradient estimators have been derived and deployed in the past, important advances have followed. A deeper and more widely-held understanding of this problem will lead to further advances, and it is these advances that we wish to support."
                },
                {
                    "title": "Combinatorial optimization by simulating adiabatic bifurcations in nonlinear Hamiltonian systems",
                    "abstract": "Nonlinear Hamiltonian systems search optimal solutions exploiting their adiabatic and chaotic evolutions. Combinatorial optimization problems are ubiquitous but difficult to solve. Hardware devices for these problems have recently been developed by various approaches, including quantum computers. Inspired by recently proposed quantum adiabatic optimization using a nonlinear oscillator network, we propose a new optimization algorithm simulating adiabatic evolutions of classical nonlinear Hamiltonian systems exhibiting bifurcation phenomena, which we call simulated bifurcation (SB). SB is based on adiabatic and chaotic (ergodic) evolutions of nonlinear Hamiltonian systems. SB is also suitable for parallel computing because of its simultaneous updating. Implementing SB with a field-programmable gate array, we demonstrate that the SB machine can obtain good approximate solutions of an all-to-all connected 2000-node MAX-CUT problem in 0.5 ms, which is about 10 times faster than a state-of-the-art laser-based machine called a coherent Ising machine. SB will accelerate large-scale combinatorial optimization harnessing digital computer technologies and also offer a new application of computational and mathematical physics."
                },
                {
                    "title": "Annealing by simulating the coherent Ising machine.",
                    "abstract": "The coherent Ising machine (CIM) enables efficient sampling of low-lying energy states of the Ising Hamiltonian with all-to-all connectivity by encoding the spins in the amplitudes of pulsed modes in an optical parametric oscillator (OPO). The interaction between the pulses is realized by means of measurement-based optoelectronic feedforward, which enhances the gain for lower-energy spin configurations. We present an efficient method of simulating the CIM on a classical computer that outperforms the CIM itself, as well as the noisy mean-field annealer in terms of both the quality of the samples and the computational speed. It is furthermore advantageous with respect to the CIM in that it can handle Ising Hamiltonians with arbitrary real-valued node coupling strengths. These results illuminate the nature of the faster performance exhibited by the CIM and may give rise to a new class of quantum-inspired algorithms of classical annealing that can successfully compete with existing methods."
                },
                {
                    "title": "Traffic Flow Optimization Using a Quantum Annealer",
                    "abstract": "Quantum annealing algorithms belong to the class of meta-heuristic tools, applicable for solving binary optimization problems. Hardware implementations of quantum annealing, such as the quantum processing units (QPUs) produced by D-Wave Systems, have been subject to multiple analyses in research, with the aim of characterizing the technology's usefulness for optimization and sampling tasks. In this paper, we present a real-world application that uses quantum technologies. Specifically, we show how to map certain parts of the real-world traffic flow optimization problem to be suitable for quantum annealing. We show that time-critical optimization tasks, such as continuous redistribution of position data for cars in dense road networks, are suitable candidates for quantum applications. Due to the limited size and connectivity of current-generation D-Wave QPUs, we use a hybrid quantum and classical approach to solve the traffic flow problem."
                },
                {
                    "title": "Finding A Small Vertex Cover in Massive Sparse Graphs: Construct, Local Search, and Preprocess",
                    "abstract": "The problem of finding a minimum vertex cover (MinVC) in a graph is a well known NP-hard combinatorial optimization problem of great importance in theory and practice. Due to its NP-hardness, there has been much interest in developing heuristic algorithms for finding a small vertex cover in reasonable time. Previously, heuristic algorithms for MinVC have focused on solving graphs of relatively small size, and they are not suitable for solving massive graphs as they usually have high-complexity heuristics. This paper explores techniques for solving MinVC in very large scale real-world graphs, including a construction algorithm, a local search algorithm and a preprocessing algorithm. Both the construction and search algorithms are based on low-complexity heuristics, and we combine them to develop a heuristic algorithm for MinVC called FastVC. Experimental results on a broad range of real-world massive graphs show that, our algorithms are very fast and have better performance than previous heuristic algorithms for MinVC. We also develop a preprocessing algorithm to simplify graphs for MinVC algorithms. By applying the preprocessing algorithm to local search algorithms, we obtain two efficient MinVC solvers called NuMVC2+p and FastVC2+p, which show further improvement on the massive graphs."
                },
                {
                    "title": "Graph Partitioning using Quantum Annealing on the D-Wave System",
                    "abstract": "Graph partitioning (GP) applications are ubiquitous throughout mathematics, computer science, chemistry, physics, bio-science, machine learning, and complex systems. Post Moore's era supercomputing has provided us an opportunity to explore new approaches for traditional graph algorithms on quantum computing architectures. In this work, we explore graph partitioning using quantum annealing on the D-Wave 2X machine. Motivated by a recently proposed graph-based electronic structure theory applied to quantum molecular dynamics (QMD) simulations, graph partitioning is used for reducing the calculation of the density matrix into smaller subsystems rendering the calculation more computationally efficient. Unconstrained graph partitioning as community clustering based on the modularity metric can be naturally mapped into the Hamiltonian of the quantum annealer. On the other hand, when constraints are imposed for partitioning into equal parts and minimizing the number of cut edges between parts, a quadratic unconstrained binary optimization (QUBO) reformulation is required. This reformulation may employ the graph complement to fit the problem in the Chimera graph of the quantum annealer. Partitioning into 2 parts and k parts concurrently for arbitrary k are demonstrated with benchmark graphs, random graphs, and small material system density matrix based graphs. Results for graph partitioning using quantum and hybrid classical-quantum approaches are shown to be comparable to current \"state of the art\" methods and sometimes better."
                },
                {
                    "title": "A coherent Ising machine for 2000-node optimization problems",
                    "abstract": "Taking the pulse of optimization Finding the optimum solution of multiparameter or multifunctional problems is important across many disciplines, but it can be computationally intensive. Many such problems defined as computationally difficult can be mathematically mapped onto the so-called Ising problem, which looks at finding the minimum energy configuration for an array of coupled spins. Inagaki et al. and McMahon et al. show that an optical processing approach based on a network of coupled optical pulses in a ring fiber can be used to model and optimize large-scale Ising systems. Such a scalable architecture could help to optimize solutions to a wide range of complex problems. Science, this issue pp. 603 and 614 An optical-based processor is developed to solve a broad class of complex optimization problems. The analysis and optimization of complex systems can be reduced to mathematical problems collectively known as combinatorial optimization. Many such problems can be mapped onto ground-state search problems of the Ising model, and various artificial spin systems are now emerging as promising approaches. However, physical Ising machines have suffered from limited numbers of spin-spin couplings because of implementations based on localized spins, resulting in severe scalability problems. We report a 2000-spin network with all-to-all spin-spin couplings. Using a measurement and feedback scheme, we coupled time-multiplexed degenerate optical parametric oscillators to implement maximum cut problems on arbitrary graph topologies with up to 2000 nodes. Our coherent Ising machine outperformed simulated annealing in terms of accuracy and computation time for a 2000-node complete graph."
                },
                {
                    "title": "A Theoretical Analysis of Optimization by Gaussian Continuation",
                    "abstract": "\n \n Optimization via continuation method is a widely used approach for solving nonconvex minimization problems. While this method generally does not provide a global minimum, empirically it often achieves a superior local minimum compared to alternative approaches such as gradient descent. However, theoretical analysis of this method is largely unavailable. Here, we provide a theoretical analysis that provides a bound on the endpoint solution of the continuation method. The derived bound depends on a problem specific characteristic that we refer to as optimization complexity. We show that this characteristic can be analytically computed when the objective function is expressed in some suitable basis functions. Our analysis combines elements of scale-space theory, regularization and differential equations.\n \n"
                },
                {
                    "title": "Practical augmented Lagrangian methods for constrained optimization",
                    "abstract": "This book focuses on Augmented Lagrangian techniques for solving practical constrained optimization problems. The authors rigorously delineate mathematical convergence theory based on sequential optimality conditions and novel constraint qualifications. They also orient the book to practitioners by giving priority to results that provide insight on the practical behavior of algorithms and by providing geometrical and algorithmic interpretations of every mathematical result, and they fully describe a freely available computational package for constrained optimization and illustrate its usefulness with applications. Audience: The book is aimed at engineers, physicists, chemists, and other practitioners interested in full access to comprehensive and well-documented software for large-scale optimization as well as up-to-date convergence theory and its practical consequences. It will also be of interest to graduate and advanced undergraduate students in mathematics, computer science, applied mathematics, optimization, and numerical analysis."
                },
                {
                    "title": "Coherent Ising machine based on degenerate optical parametric oscillators",
                    "abstract": "A degenerate optical parametric oscillator network is proposed to solve the NP-hard problem of finding a ground state of the Ising model. The underlying operating mechanism originates from the bistable output phase of each oscillator and the inherent preference of the network in selecting oscillation modes with the minimum photon decay rate. Computational experiments are performed on all instances reducible to the NP-hard MAX-CUT problems on cubic graphs of order up to 20. The numerical results reasonably suggest the effectiveness of the proposed network."
                },
                {
                    "title": "Classical signature of quantum annealing",
                    "abstract": "A pair of recent articles concluded that the D-Wave One machine actually operates in the quantum regime, rather than performing some classical evolution. Here we give a classical model that leads to the same behaviors used in those works to infer quantum effects. Thus, the evidence presented does not demonstrate the presence of quantum effects."
                },
                {
                    "title": "Ising formulations of many NP problems",
                    "abstract": "We provide Ising formulations for many NP-complete and NP-hard problems, including all of Karp's 21 NP-complete problems. This collects and extends mappings to the Ising model from partitioning, covering and satisfiability. In each case, the required number of spins is at most cubic in the size of the problem. This work may be useful in designing adiabatic quantum optimization algorithms."
                },
                {
                    "title": "Gradient-Based Adaptive Stochastic Search for Non-Differentiable Optimization",
                    "abstract": "In this paper, we propose a stochastic search algorithm for solving general optimization problems with little structure. The algorithm iteratively finds high quality solutions by randomly sampling candidate solutions from a parameterized distribution model over the solution space. The basic idea is to convert the original (possibly non-differentiable) problem into a differentiable optimization problem on the parameter space of the parameterized sampling distribution, and then use a direct gradient search method to find improved sampling distributions. Thus, the algorithm combines the robustness feature of stochastic search from considering a population of candidate solutions with the relative fast convergence speed of classical gradient methods by exploiting local differentiable structures. We analyze the convergence and converge rate properties of the proposed algorithm, and carry out numerical study to illustrate its performance."
                },
                {
                    "title": "Boolean Functions: Theory, Algorithms, and Applications",
                    "abstract": "Part I. Foundations: 1. Fundamental concepts and applications 2. Boolean equations 3. Prime implicants and minimal DNFs Peter L. Hammer and Alexander Kogan 4. Duality theory Yves Crama and Kazuhisa Makino Part II. Special Classes: 5. Quadratic functions Bruno Simeone 6. Horn functions Endre Boros 7. Orthogonal forms and shellability 8. Regular functions 9. Threshold functions 10. Read-once functions Martin C. Golumbic and Vladimir Gurvich 11. Characterizations of special classes by functional equations Lisa Hellerstein Part III. Generalizations: 12. Partially defined Boolean functions Toshihide Ibaraki 13. Pseudo-Boolean functions Appendix A. Graphs and hypergraphs Appendix B. Algorithmic complexity Appendix C. JBool: a software tool Claude Benzaken and Nadia Brauner."
                },
                {
                    "title": "Randomized Smoothing for Stochastic Optimization",
                    "abstract": "We analyze convergence rates of stochastic optimization procedures for non-smooth convex optimization problems. By combining randomized smoothing techniques with accelerated gradient methods, we obtain convergence rates of stochastic optimization procedures, both in expectation and with high probability, that have optimal dependence on the variance of the gradient estimates. To the best of our knowledge, these are the first variance-based rates for non-smooth optimization. We give several applications of our results to statistical estimation problems, and provide experimental results that demonstrate the effectiveness of the proposed algorithms. We also describe how a combination of our algorithm with recent work on decentralized optimization yields a distributed stochastic optimization algorithm that is order-optimal."
                },
                {
                    "title": "Natural Evolution Strategies",
                    "abstract": "This paper presents natural evolution strategies (NES), a novel algorithm for performing real-valued dasiablack boxpsila function optimization: optimizing an unknown objective function where algorithm-selected function measurements constitute the only information accessible to the method. Natural evolution strategies search the fitness landscape using a multivariate normal distribution with a self-adapting mutation matrix to generate correlated mutations in promising regions. NES shares this property with covariance matrix adaption (CMA), an evolution strategy (ES) which has been shown to perform well on a variety of high-precision optimization tasks. The natural evolution strategies algorithm, however, is simpler, less ad-hoc and more principled. Self-adaptation of the mutation matrix is derived using a Monte Carlo estimate of the natural gradient towards better expected fitness. By following the natural gradient instead of the dasiavanillapsila gradient, we can ensure efficient update steps while preventing early convergence due to overly greedy updates, resulting in reduced sensitivity to local suboptima. We show NES has competitive performance with CMA on unimodal tasks, while outperforming it on several multimodal tasks that are rich in deceptive local optima."
                },
                {
                    "title": "The Elements of Statistical Learning: Data Mining, Inference, and Prediction",
                    "abstract": "In the words of the authors, the goal of this book was to \u201cbring together many of the important new ideas in learning, and explain them in a statistical framework.\u201d The authors have been quite successful in achieving this objective, and their work is a welcome addition to the statistics and learning literatures. Statistics has always been interdisciplinary, borrowing ideas from diverse \u008e elds and repaying the debt with contributions, both theoretical and practical, to the other intellectual disciplines. For statistical learning, this cross-fertilization is especially noticeable. This book is a valuable resource, both for the statistician needing an introduction to machine learning and related \u008e elds and for the computer scientist wishing to learn more about statistics. Statisticians will especially appreciate that it is written in their own language. The level of the book is roughly that of a second-year doctoral student in statistics, and it will be useful as a textbook for such students. In a stimulating article, Breiman (2001) argued that statistics has been focused too much on a \u201cdata modeling culture,\u201d where the model is paramount. Breiman argued instead for an \u201calgorithmic modeling culture,\u201d with emphasis on black-box types of prediction. Breiman\u2019s article is controversial, and in his discussion, Efron objects that \u201cprediction is certainly an interesting subject, but Leo\u2019s paper overstates both its role and our profession\u2019s lack of interest in it.\u201d Although I mostly agree with Efron, I worry that the courses offered by most statistics departments include little, if any, treatment of statistical learning and prediction. (Stanford, where Efron and the authors of this book teach, is an exception.) Graduate students in statistics certainly need to know more than they do now about prediction, machine learning, statistical learning, and data mining (not disjoint subjects). I hope that graduate courses covering the topics of this book will become more common in statistics curricula. Most of the book is focused on supervised learning, where one has inputs and outputs from some system and wishes to predict unknown outputs corresponding to known inputs. The methods discussed for supervised learning include linear and logistic regression; basis expansion, such as splines and wavelets; kernel techniques, such as local regression, local likelihood, and radial basis functions; neural networks; additive models; decision trees based on recursive partitioning, such as CART; and support vector machines. There is a \u008e nal chapter on unsupervised learning, including association rules, cluster analysis, self-organizing maps, principal components and curves, and independent component analysis. Many statisticians will be unfamiliar with at least some of these algorithms. Association rules are popular for mining commercial data in what is called \u201cmarket basket analysis.\u201d The aim is to discover types of products often purchased together. Such knowledge can be used to develop marketing strategies, such as store or catalog layouts. Self-organizing maps (SOMs) involve essentially constrained k-means clustering, where prototypes are mapped to a two-dimensional curved coordinate system. Independent components analysis is similar to principal components analysis and factor analysis, but it uses higher-order moments to achieve independence, not merely zero correlation between components. A strength of the book is the attempt to organize a plethora of methods into a coherent whole. The relationships among the methods are emphasized. I know of no other book that covers so much ground. Of course, with such broad coverage, it is not possible to cover any single topic in great depth, so this book will encourage further reading. Fortunately, each chapter includes bibliographic notes surveying the recent literature. These notes and the extensive references provide a good introduction to the learning literature, including much outside of statistics. The book might be more suitable as a textbook if less material were covered in greater depth; however, such a change would compromise the book\u2019s usefulness as a reference, and so I am happier with the book as it was written."
                },
                {
                    "title": "A Quantum Adiabatic Evolution Algorithm Applied to Random Instances of an NP-Complete Problem",
                    "abstract": "A quantum system will stay near its instantaneous ground state if the Hamiltonian that governs its evolution varies slowly enough. This quantum adiabatic behavior is the basis of a new class of algorithms for quantum computing. We tested one such algorithm by applying it to randomly generated hard instances of an NP-complete problem. For the small examples that we could simulate, the quantum adiabatic algorithm worked well, providing evidence that quantum computers (if large ones can be built) may be able to outperform ordinary computers on hard sets of instances of NP-complete problems."
                },
                {
                    "title": "An Application of Combinatorial Optimization to Statistical Physics and Circuit Layout Design",
                    "abstract": "We study the problem of finding ground states of spin glasses with exterior magnetic field, and the problem of minimizing the number of vias holes on a printed circuit board, or contacts on a chip subject to pin preassignments and layer preferences. The former problem comes up in solid-state physics, and the latter in very-large-scale-integrated VLSI circuit design and in printed circuit board design. Both problems can be reduced to the max-cut problem in graphs. Based on a partial characterization of the cut polytope, we design a cutting plane algorithm and report on computational experience with it. Our method has been used to solve max-cut problems on graphs with up to 1,600 nodes."
                },
                {
                    "title": "Revisiting Sampling for Combinatorial Optimization",
                    "abstract": "Sampling approaches like Markov chain Monte Carlo were once popular for combinatorial optimization, but the inefficiency of classical meth-ods and the need for problem-specific designs curtailed ongoing develpment. Recent work has favored data-driven approaches that mitigate the need for hand-craft heuristics, but these are often not usable as out-of-the-box solvers due to dependence on in-distribution training and limited scalability to large instances. In this paper, we revisit the idea of using sampling for combinatorial optimization, motivated by the significant recent advances of gradient-based discrete MCMC and new techniques for parallel neighborhood exploration on accelerators. Remarkably, we find that modern sampling strategies can leverage landscape information to provide general-purpose solvers that require no training and yet are competitive with state of the art combinatorial solvers. In particular, experiments on cover vertex selection, graph partition and routing demonstrate better speed-quality trade-offs over current learning based approaches, and sometimes even superior performance to commercial solvers and specialized algorithms."
                },
                {
                    "title": "Path Auxiliary Proposal for MCMC in Discrete Space",
                    "abstract": "Energy-based Models (EBMs) offer a powerful approach for modeling discrete structure, but both inference and learning of EBM are hard as it involves sampling from discrete distributions. Recent work shows Markov Chain Monte Carlo (MCMC) with the informed proposal is a powerful tool for such sampling. How-ever, an informed proposal only allows local updates as it requires evaluating all energy changes in the neighborhood. In this work, we present a path auxiliary algorithm that uses a composition of local moves to ef\ufb01ciently explore large neigh-borhoods. We also give a fast version of our algorithm that only queries the evaluation of energy function twice for each proposal via linearization of the energy function. Empirically, we show that our path auxiliary algorithms considerably outperform other generic samplers on various discrete models for sampling, inference, and learning. Our method can also be used to train deep EBMs for high dimensional discrete data."
                },
                {
                    "title": "Semi-supervised learning using greedy max-cut",
                    "abstract": "Graph-based semi-supervised learning (SSL) methods play an increasingly important role in practical machine learning systems, particularly in agnostic settings when no parametric information or other prior knowledge is available about the data distribution. Given the constructed graph represented by a weight matrix, transductive inference is used to propagate known labels to predict the values of all unlabeled vertices. Designing a robust label diffusion algorithm for such graphs is a widely studied problem and various methods have recently been suggested. Many of these can be formalized as regularized function estimation through the minimization of a quadratic cost. However, most existing label diffusion methods minimize a univariate cost with the classification function as the only variable of interest. Since the observed labels seed the diffusion process, such univariate frameworks are extremely sensitive to the initial label choice and any label noise. To alleviate the dependency on the initial observed labels, this article proposes a bivariate formulation for graph-based SSL, where both the binary label information and a continuous classification function are arguments of the optimization. This bivariate formulation is shown to be equivalent to a linearly constrained Max-Cut problem. Finally an efficient solution via greedy gradient Max-Cut (GGMC) is derived which gradually assigns unlabeled vertices to each class with minimum connectivity. Once convergence guarantees are established, this greedy Max-Cut based SSL is applied on both artificial and standard benchmark data sets where it obtains superior classification accuracy compared to existing state-of-the-art SSL methods. Moreover, GGMC shows robustness with respect to the graph construction method and maintains high accuracy over extensive experiments with various edge linking and weighting schemes."
                },
                {
                    "title": "SATLIB: An Online Resource for Research on SAT",
                    "abstract": "SATLIB is an online resource for SAT-related research established in June 1998. Its core components, a benchmark suite of SAT instances and a collection of SAT solvers, aim to facilitate empirical research on SAT by providing a uniform test-bed for SAT solvers along with freely available implementations of high-performing SAT algorithms. In this article, we give an overview of SATLIB; in particular, we describe its current set of benchmark problems. Currently, the main SATLIB web site (http://www.informatik.tu-darmstadt.de/AI/SATLIB) and its North Amer- ican mirror site (http://www.cs.ubc.ca/\u02dchoos/SATLIB) are being accessed frequently by a growing number of researchers, resulting in access rates of about 250 hits per month. To further increase the usefulness of SATLIB as a resource, we encourage all members of the community to utilise it for their SAT-related research and to improve it by submitting new benchmark problems, SAT solvers, and bibliography entries."
                },
                {
                    "title": "Regression Shrinkage and Selection via the Lasso",
                    "abstract": "SUMMARY We propose a new method for estimation in linear models. The 'lasso' minimizes the residual sum of squares subject to the sum of the absolute value of the coefficients being less than a constant. Because of the nature of this constraint it tends to produce some coefficients that are exactly 0 and hence gives interpretable models. Our simulation studies suggest that the lasso enjoys some of the favourable properties of both subset selection and ridge regression. It produces interpretable models like subset selection and exhibits the stability of ridge regression. There is also an interesting relationship with recent work in adaptive function estimation by Donoho and Johnstone. The lasso idea is quite general and can be applied in a variety of statistical models: extensions to generalized regression models and tree-based models are briefly described."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG",
                "math.CO",
                "physics.app-ph"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively utilize heat diffusion to enhance the efficiency of iterative approximation solvers in combinatorial optimization problems?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant limitation in current combinatorial optimization methods, which often get trapped in local minima due to their narrow receptive fields. By improving the efficiency of these solvers, we can advance knowledge in optimization techniques and potentially lead to practical applications across various fields, such as circuit design, machine learning, and traffic flow optimization. This research could pave the way for more robust algorithms that can handle complex optimization tasks, ultimately influencing future research directions and methodologies in the field.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the combinatorial explosion that occurs as the receptive field of solvers is expanded. Naive approaches that simply increase the search scope may lead to an exponential increase in the number of solutions to evaluate, making thorough assessments computationally expensive and impractical. Additionally, the inherent complexity of the solution landscape, which may contain numerous local minima and bumpy regions, complicates the optimization process. Overcoming these technical and practical obstacles requires innovative strategies to effectively propagate information from distant areas of the solution space without incurring prohibitive computational costs.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on expanding the receptive field of solvers to gather more information from the solution space, but these methods have only achieved modest improvements. The limitations of existing approaches, such as large neighborhood search and variable neighborhood search, have prevented a comprehensive solution to the problem. Additionally, the concept of utilizing heat diffusion as a means to propagate information from distant regions has not been explored in depth. Our approach differs by shifting the focus from expanding the search scope to enhancing the solver's ability to access and utilize information from the broader solution space, thereby improving optimization efficiency.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves implementing a heat diffusion framework that transforms the target function into various versions while preserving the location of the optima. We will utilize benchmark combinatorial optimization datasets to evaluate the effectiveness of our approach. The performance will be measured using metrics such as solution quality and computational efficiency. We expect that by leveraging heat diffusion, our approach will significantly enhance the solver's ability to navigate the solution space"
            }
        },
        "author_data": {
            "654e5671-8629-421d-b747-a85b72c1cfc4": {
                "pk": "654e5671-8629-421d-b747-a85b72c1cfc4",
                "name": "Hengyuan Ma",
                "collaborators": [
                    "Jianfeng Feng",
                    "Li Zhang",
                    "Xiatian Zhu",
                    "Yang Qi",
                    "Wenlian Lu",
                    "Jingfeng Zhang",
                    "Pulin Gong",
                    "Jie Zhang"
                ],
                "domain": [
                    "Generative Models",
                    "Probabilistic Computing",
                    "Neural Networks",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Preconditioned Score-based Generative Models",
                        "abstract": "Score-based generative models (SGMs) have recently emerged as a promising class of generative models. However, a fundamental limitation is that their sampling process is slow due to a need for many (\\eg, $2000$) iterations of sequential computations. An intuitive acceleration method is to reduce the sampling iterations which however causes severe performance degradation. We assault this problem to the ill-conditioned issues of the Langevin dynamics and reverse diffusion in the sampling process. Under this insight, we propose a model-agnostic {\\bf\\em preconditioned diffusion sampling} (PDS) method that leverages matrix preconditioning to alleviate the aforementioned problem. PDS alters the sampling process of a vanilla SGM at marginal extra computation cost, and without model retraining. Theoretically, we prove that PDS preserves the output distribution of the SGM, no risk of inducing systematical bias to the original sampling process. We further theoretically reveal a relation between the parameter of PDS and the sampling iterations,easing the parameter estimation under varying sampling iterations. Extensive experiments on various image datasets with a variety of resolutions and diversity validate that our PDS consistently accelerates off-the-shelf SGMs whilst maintaining the synthesis quality. In particular, PDS can accelerate by up to $29\\times$ on more challenging high resolution (1024$\\times$1024) image generation. Compared with the latest generative models (\\eg, CLD-SGM, DDIM, and Analytic-DDIM), PDS can achieve the best sampling quality on CIFAR-10 at a FID score of 1.99. Our code is made publicly available to foster any further research https://github.com/fudan-zvg/PDS."
                    },
                    {
                        "title": "Accelerating Score-based Generative Models with Preconditioned Diffusion Sampling",
                        "abstract": "Score-based generative models (SGMs) have recently emerged as a promising class of generative models. However, a fundamental limitation is that their inference is very slow due to a need for many (e.g., 2000) iterations of sequential computations. An intuitive acceleration method is to reduce the sampling iterations which however causes severe performance degradation. We investigate this problem by viewing the diffusion sampling process as a Metropolis adjusted Langevin algorithm, which helps reveal the underlying cause to be ill-conditioned curvature. Under this insight, we propose a model-agnostic preconditioned diffusion sampling (PDS) method that leverages matrix preconditioning to alleviate the aforementioned problem. Crucially, PDS is proven theoretically to converge to the original target distribution of a SGM, no need for retraining. Extensive experiments on three image datasets with a variety of resolutions and diversity validate that PDS consistently accelerates off-the-shelf SGMs whilst maintaining the synthesis quality. In particular, PDS can accelerate by up to 29x on more challenging high resolution (1024x1024) image generation."
                    },
                    {
                        "title": "Probabilistic computation and uncertainty quantification with emerging covariance",
                        "abstract": "Building robust, interpretable, and secure AI system requires quantifying and representing uncertainty under a probabilistic perspective to mimic human cognitive abilities. However, probabilistic computation presents significant challenges for most conventional artificial neural network, as they are essentially implemented in a deterministic manner. In this paper, we develop an efficient probabilistic computation framework by truncating the probabilistic representation of neural activation up to its mean and covariance and construct a moment neural network that encapsulates the nonlinear coupling between the mean and covariance of the underlying stochastic network. We reveal that when only the mean but not the covariance is supervised during gradient-based learning, the unsupervised covariance spontaneously emerges from its nonlinear coupling with the mean and faithfully captures the uncertainty associated with model predictions. Our findings highlight the inherent simplicity of probabilistic computation by seamlessly incorporating uncertainty into model prediction, paving the way for integrating it into large-scale AI systems."
                    },
                    {
                        "title": "Accelerating Score-based Generative Models for High-Resolution Image Synthesis",
                        "abstract": "Score-based generative models (SGMs) have recently emerged as a promising class of generative models. The key idea is to produce high-quality images by recurrently adding Gaussian noises and gradients to a Gaussian sample until converging to the target distribution, a.k.a. the diffusion sampling. To ensure stability of convergence in sampling and generation quality, however, this sequential sampling process has to take a small step size and many sampling iterations (e.g., 2000). Several acceleration methods have been proposed with focus on low-resolution generation. In this work, we consider the acceleration of high-resolution generation with SGMs, a more challenging yet more important problem. We prove theoretically that this slow convergence drawback is primarily due to the ignorance of the target distribution. Further, we introduce a novel Target Distribution Aware Sampling (TDAS) method by leveraging the structural priors in space and frequency domains. Extensive experiments on CIFAR-10, CelebA, LSUN, and FFHQ datasets validate that TDAS can consistently accelerate state-of-the-art SGMs, particularly on more challenging high resolution (1024x1024) image generation tasks by up to 18.4x, whilst largely maintaining the synthesis quality. With fewer sampling iterations, TDAS can still generate good quality images. In contrast, the existing methods degrade drastically or even fails completely"
                    },
                    {
                        "title": "Self-organization of nonlinearly coupled neural fluctuations into synergistic population codes",
                        "abstract": "Neural activity in the brain exhibits correlated fluctuations that may strongly influence the properties of neural population coding. However, how such correlated neural fluctuations may arise from the intrinsic neural circuit dynamics and subsequently affect the computational properties of neural population activity remains poorly understood. The main difficulty lies in resolving the nonlinear coupling between correlated fluctuations with the overall dynamics of the system. In this study, we investigate the emergence of synergistic neural population codes from the intrinsic dynamics of correlated neural fluctuations in a neural circuit model capturing realistic nonlinear noise coupling of spiking neurons. We show that a rich repertoire of spatial correlation patterns naturally emerges in a bump attractor network and further reveals the dynamical regime under which the interplay between differential and noise correlations leads to synergistic codes. Moreover, we find that negative correlations may induce stable bound states between two bumps, a phenomenon previously unobserved in firing rate models. These noise-induced effects of bump attractors lead to a number of computational advantages including enhanced working memory capacity and efficient spatiotemporal multiplexing and can account for a range of cognitive and behavioral phenomena related to working memory. This study offers a dynamical approach to investigating realistic correlated neural fluctuations and insights to their roles in cortical computations."
                    }
                ]
            },
            "c4e84c42-65a6-4dab-a1a8-667d8f7de5dd": {
                "pk": "c4e84c42-65a6-4dab-a1a8-667d8f7de5dd",
                "name": "Wenlian Lu",
                "collaborators": [
                    "Tianping Chen",
                    "Yujuan Han",
                    "Fatihcan M. Atay",
                    "Xiwei Liu",
                    "Ren Zheng",
                    "Xinlei Yi",
                    "Xiwei lIu",
                    "Bo liu",
                    "Jurgen Jost",
                    "Shouhuai Xu"
                ],
                "domain": [
                    "Neural Networks",
                    "Dynamical Systems",
                    "Synchronization",
                    "Control Theory"
                ],
                "publications": [
                    {
                        "title": "Global Exponential Stability of Almost Periodic Solution for A Large Class of Delayed Dynamical Systems",
                        "abstract": "Research of delayed neural networks with variable self-inhibitions, inter-connection weights, and inputs is an important issue. %In the real world, self-inhibitions, %inter-connection weights, and inputs should vary through time. In In this paper, we discuss a large class of delayed dynamical systems with almost periodic self-inhibitions, inter-connection weights, and inputs. This model is universal and includes delayed systems with time-varying delays, distributed delays as well as combination of both. We prove that under some mild conditions, the system has a unique almost periodic solution, which is globally exponentially stable. We propose a new approach, which is independent of existing theory concerning with existence of almost periodic solution for dynamical systems."
                    },
                    {
                        "title": "Local pinning of networks of multi-agent systems with transmission and pinning delays",
                        "abstract": "We study the stability of networks of multi-agent systems with local pinning strategies and two types of time delays, namely the transmission delay in the network and the pinning delay of the controllers. Sufficient conditions for stability are derived under specific scenarios by computing or estimating the dominant eigenvalue of the characteristic equation. In addition, controlling the network by pinning a single node is studied. Moreover, perturbation methods are employed to derive conditions in the limit of small and large pinning strengths.Numerical algorithms are proposed to verify stability, and simulation examples are presented to confirm the efficiency of analytic results."
                    },
                    {
                        "title": "Centralized and Decentralized Global Outer-synchronization of Asymmetric Recurrent Time-varying Neural Network by Data-sampling",
                        "abstract": "In this paper, we discuss the outer-synchronization of the asymmetrically connected recurrent time-varying neural networks. By both centralized and decentralized discretization data sampling principles, we derive several sufficient conditions based on diverse vector norms that guarantee that any two trajectories from different initial values of the identical neural network system converge together. The lower bounds of the common time intervals between data samples in centralized and decentralized principles are proved to be positive, which guarantees exclusion of Zeno behavior. A numerical example is provided to illustrate the efficiency of the theoretical results."
                    },
                    {
                        "title": "Pinning Complex Networks by a Single Controller",
                        "abstract": "In this paper, without assuming symmetry, irreducibility, or linearity of the couplings, we prove that a single controller can pin a coupled complex network to a homogenous solution. Sufficient conditions are presented to guarantee the convergence of the pinning process locally and globally. An effective approach to adapt the coupling strength is proposed. Several numerical simulations are given to verify our theoretical analysis."
                    },
                    {
                        "title": "Finite time Convergence of Pinning Synchronization with Linear and Nonlinear Controllers",
                        "abstract": "In this paper, we propose several models, which can realize synchronization of complex networks in finite time effectively. The results apply to heterogeneous dynamic networks, too. The mechanism of finite time convergence is revealed. Different from many models, which assume the coupling matrix being symmetric (or the connecting graph is undirected), here, the coupling matrix is asymmetric (or the connecting graph is directed)."
                    },
                    {
                        "title": "Finite time Convergence of Pinning Synchronization with a single Nonlinear Controller",
                        "abstract": "In this paper, we discuss distributive synchronization of complex networks in finite time, with a single nonlinear pinning controller. The results apply to heterogeneous dynamic networks, too. Different from many models, which assume the coupling matrix being symmetric (or the connecting graph is undirected), here, the coupling matrix is asymmetric (or the connecting graph is directed"
                    },
                    {
                        "title": "Synchronization analysis of high order layered complex networks",
                        "abstract": "In this letter, we propose high order layered complex networks. The synchronization is discussed in detail. The relations of synchronization, individual coupling matrices and the intrinsic function of the uncoupled system are given. As special cases, synchronization of monolayer networks and multiplex networks discussed in literature can be obtained."
                    },
                    {
                        "title": "Cluster synchronization in networks of coupled non-identical dynamical systems",
                        "abstract": "In this paper, we study cluster synchronization in networks of coupled non-identical dynamical systems. The vertices in the same cluster have the same dynamics of uncoupled node system but the uncoupled node systems in different clusters are different. We present conditions guaranteeing cluster synchronization and investigate the relation between cluster synchronization and the unweighted graph topology. We indicate that two condition play key roles for cluster synchronization: the common inter-cluster coupling condition and the intra-cluster communication. From the latter one, we interpret the two well-known cluster synchronization schemes: self-organization and driving, by whether the edges of communication paths lie at inter or intra-cluster. By this way, we classify clusters according to whether the set of edges inter- or intra-cluster edges are removable if wanting to keep the communication between pairs of vertices in the same cluster. Also, we propose adaptive feedback algorithms on the weights of the underlying graph, which can synchronize any bi-directed networks satisfying the two conditions above. We also give several numerical examples to illustrate the theoretical results."
                    },
                    {
                        "title": "Consensus and synchronization in discrete-time networks of multi-agents with stochastically switching topologies and time delays",
                        "abstract": "We analyze stability of consensus algorithms in networks of multi-agents with time-varying topologies and delays. The topology and delays are modeled as induced by an adapted process and are rather general, including i.i.d.\\ topology processes, asynchronous consensus algorithms, and Markovian jumping switching. In case the self-links are instantaneous, we prove that the network reaches consensus for all bounded delays if the graph corresponding to the conditional expectation of the coupling matrix sum across a finite time interval has a spanning tree almost surely. Moreover, when self-links are also delayed and when the delays satisfy certain integer patterns, we observe and prove that the algorithm may not reach consensus but instead synchronize at a periodic trajectory, whose period depends on the delay pattern. We also give a brief discussion on the dynamics in the absence of self-links."
                    },
                    {
                        "title": "Achieving synchronization in arrays of coupled differential systems with time-varying couplings",
                        "abstract": "In this paper, we study complete synchronization of the complex dynamical networks described by linearly coupled ordinary differential equation systems (LCODEs). The coupling considered here is time-varying in both the network structure and the reaction dynamics. Inspired by our previous paper [6], the extended Hajnal diameter is introduced and used to measure the synchronization in a general differential system. Then we find that the Hajnal diameter of the linear system induced by the time-varying coupling matrix and the largest Lyapunov exponent of the synchronized system play the key roles in synchronization analysis of LCODEs with the identity inner coupling matrix. As an application, we obtain a general sufficient condition guaranteeing directed time-varying graph to reach consensus. Example with numerical simulation is provided to show the effectiveness the theoretical results."
                    },
                    {
                        "title": "Characterizing the Power of Moving Target Defense via Cyber Epidemic Dynamics",
                        "abstract": "Moving Target Defense (MTD) can enhance the resilience of cyber systems against attacks. Although there have been many MTD techniques, there is no systematic understanding and {\\em quantitative} characterization of the power of MTD. In this paper, we propose to use a cyber epidemic dynamics approach to characterize the power of MTD. We define and investigate two complementary measures that are applicable when the defender aims to deploy MTD to achieve a certain security goal. One measure emphasizes the maximum portion of time during which the system can afford to stay in an undesired configuration (or posture), without considering the cost of deploying MTD. The other measure emphasizes the minimum cost of deploying MTD, while accommodating that the system has to stay in an undesired configuration (or posture) for a given portion of time. Our analytic studies lead to algorithms for optimally deploying MTD."
                    },
                    {
                        "title": "Contraction and incremental stability of switched Carath\u00e9odory systems using multiple norms",
                        "abstract": "In this paper, incremental exponential asymptotic stability of a class of switched Carath\\'{e}odory nonlinear systems is studied based on the novel concept of measure of switched matrices via multiple norms and the transaction coefficients between these norms. This model is rather general and includes the case of staircase switching signals as a special case. Sufficient conditions are derived for incremental stability allowing for the system to be incrementally exponentially asymptotically stable even if some of its modes are unstable in some time periods. Numerical examples on switched linear systems switching periodically and on the synchronization of switched networks of nonlinear systems are used to illustrate the theoretical results."
                    },
                    {
                        "title": "Stability of Analytic Neural Networks with Event-triggered Synaptic Feedbacks",
                        "abstract": "In this paper, we investigate stability of a class of analytic neural networks with the synaptic feedback via event-triggered rules. This model is general and include Hopfield neural network as a special case. These event-trigger rules can efficiently reduces loads of computation and information transmission at synapses of the neurons. The synaptic feedback of each neuron keeps a constant value based on the outputs of the other neurons at its latest triggering time but changes at its next triggering time, which is determined by certain criterion. It is proved that every trajectory of the analytic neural network converges to certain equilibrium under this event-triggered rule for all initial values except a set of zero measure. The main technique of the proof is the Lojasiewicz inequality to prove the finiteness of trajectory length. The realization of this event-triggered rule is verified by the exclusion of Zeno behaviors. Numerical examples are provided to illustrate the efficiency of the theoretical results."
                    },
                    {
                        "title": "Stability of phase difference trajectories of networks of Kuramoto oscillators with time-varying couplings and intrinsic frequencies",
                        "abstract": "We study dynamics of phase-differences (PDs) of coupled oscillators where both the intrinsic frequencies and the couplings vary in time. In case the coupling coefficients are all nonnegative, we prove that the PDs are asymptotically stable if there exists T>0 such that the aggregation of the time-varying graphs across any time interval of length $T$ has a spanning tree. We also consider the situation that the coupling coefficients may be negative and provide sufficient conditions for the asymptotic stability of the PD dynamics. Due to time-variations, the PDs are asymptotic to time-varying patterns rather than constant values. Hence, the PD dynamics can be regarded as a generalisation of the well-known phase-locking phenomena. We explicitly investigate several particular cases of time-varying graph structures, including asymptotically periodic PDs due to periodic coupling coefficients and intrinsic frequencies, small perturbations, and fast-switching near constant coupling and frequencies, which lead to PD dynamics close to a phase-locked one. Numerical examples are provided to illustrate the theoretical results."
                    },
                    {
                        "title": "Analytic Investigation for Spatio-temporal Patterns Propagation in Spiking Neural Networks",
                        "abstract": "Based upon the moment closure approach, a Gaussian random field is constructed to quantitatively and analytically characterize the dynamics of a random point field. The approach provides us with a theoretical tool to investigate synchronized spike propagation in a feedforward or recurrent spiking neural network. We show that the balance between the excitation and inhibition postsynaptic potentials is required for the occurrence of synfire chains. In particular, with a balanced network, the critical packet size of invasion and annihilation is observed. We also derive a sufficient analytic condition for the synchronization propagation in an asynchronous environment, which further allows us to disclose the possibility of spatial synaptic structure to sustain a stable synfire chain. Our findings are in good agreement with simulations and help us understand the propagation of spatio-temporal patterns in a random point flied."
                    },
                    {
                        "title": "Cluster consensus in discrete-time networks of multi-agents with inter-cluster nonidentical inputs",
                        "abstract": "In this paper, cluster consensus of multi-agent systems is studied via inter-cluster nonidentical inputs. Here, we consider general graph topologies, which might be time-varying. The cluster consensus is defined by two aspects: the intra-cluster synchronization, that the state differences between each pair of agents in the same cluster converge to zero, and inter-cluster separation, that the states of the agents in different clusters are separated. For intra-cluster synchronization, the concepts and theories of consensus including the spanning trees, scramblingness, infinite stochastic matrix product and Hajnal inequality, are extended. With them, it is proved that if the graph has cluster spanning trees and all vertices self-linked, then static linear system can realize intra-cluster synchronization. For the time-varying coupling cases, it is proved that if there exists T>0 such that the union graph across any T-length time interval has cluster spanning trees and all graphs has all vertices self-linked, then the time-varying linear system can also realize intra-cluster synchronization. Under the assumption of common inter-cluster influence, a sort of inter-cluster nonidentical inputs are utilized to realize inter-cluster separation, that each agent in the same cluster receives the same inputs and agents in different clusters have different inputs. In addition, the boundedness of the infinite sum of the inputs can guarantee the boundedness of the trajectory. As an application, we employ a modified non-Bayesian social learning model to illustrate the effectiveness of our results."
                    },
                    {
                        "title": "New criterion of asymptotic stability for delay systems with time-varying structures and delays",
                        "abstract": "In this paper, we study asymptotic stability of the zero solution of a class of differential systems governed by a scalar differential inequality with time-varying structures and delays. We establish a new generalized Halanay inequality for the asymptotic stability of the zero solution for such systems under more relaxed conditions than the existing ones. We also apply the theoretical results to the analysis of self synchronization in networks of delayed differential systems and obtained a more general sufficient condition for self synchronization."
                    },
                    {
                        "title": "Synchronization in Networks of Linearly Coupled Dynamical Systems via Event-triggered Diffusions",
                        "abstract": "In this paper, we utilize event-triggered coupling configuration to realize synchronization of linearly coupled dynamical systems. Here, the diffusion couplings are set up from the latest observations of the nodes of its neighborhood and the next observation time is triggered by the proposed criteria based on the local neighborhood information as well. Two scenarios are considered: continuous monitoring, that each node can observe its neighborhood's instantaneous states, and discrete monitoring, that each node can only obtain its neighborhood's states at the same time point when the coupling term is triggered. In both cases, we prove that if the system with persistent coupling can synchronize, then these event-trigger coupling strategies can synchronize the system, too."
                    },
                    {
                        "title": "A note on finite-time and fixed-time stability",
                        "abstract": "In this letter, by regarding finite-time stability as an inverse problem, we reveal the essence of finite-time stability and fixed-time stability. Some necessary and sufficient conditions are given. As application, we give a new approach for finite-time and fixed-time synchronization and consensus. Many existing results can be derived by the general approach."
                    }
                ]
            },
            "96e0eee4-7b7f-4884-9e99-9988bbed3190": {
                "pk": "96e0eee4-7b7f-4884-9e99-9988bbed3190",
                "name": "Jianfeng Feng",
                "collaborators": [
                    "Wenlian Lu",
                    "Xuejuan Zhang",
                    "Ning Hua",
                    "Xiangnan He",
                    "Zhichao Zhu",
                    "Yang Qi",
                    "Wenyong Zhang",
                    "Boyu Chen",
                    "Hengyuan Ma",
                    "Li Zhang"
                ],
                "domain": [
                    "Computational Neuroscience",
                    "Machine Learning",
                    "Generative Models",
                    "Gait Recognition"
                ],
                "publications": [
                    {
                        "title": "Computational modeling of neuronal networks",
                        "abstract": "Human brain contains about 10 billion neurons, each of which has about 10~10,000 nerve endings from which neurotransmitters are released in response to incoming spikes, and the released neurotransmitters then bind to receptors located in the postsynaptic neurons. However, individually, neurons are noisy and synaptic release is in general unreliable. But groups of neurons that are arranged in specialized modules can collectively perform complex information processing tasks robustly and reliably. How functionally groups of neurons perform behavioural related tasks crucial rely on a coherent organization of dynamics from membrane ionic kinetics to synaptic coupling of the network and dynamics of rhythmic oscillations that are tightly linked to behavioural state.   To capture essential features of the biological system at multiple spatial-temporal scales, it is important to construct a suitable computational model that is closely or solely based on experimental data. Depending on what one wants to understand, these models can either be very functional and biologically realistic descriptions with thousands of coupled differential equations (Hodgkin-Huxley type) or greatly simplified caricatures (integrate-and-fire type) which are useful for studying large interconnected networks."
                    },
                    {
                        "title": "Analytic Investigation for Spatio-temporal Patterns Propagation in Spiking Neural Networks",
                        "abstract": "Based upon the moment closure approach, a Gaussian random field is constructed to quantitatively and analytically characterize the dynamics of a random point field. The approach provides us with a theoretical tool to investigate synchronized spike propagation in a feedforward or recurrent spiking neural network. We show that the balance between the excitation and inhibition postsynaptic potentials is required for the occurrence of synfire chains. In particular, with a balanced network, the critical packet size of invasion and annihilation is observed. We also derive a sufficient analytic condition for the synchronization propagation in an asynchronous environment, which further allows us to disclose the possibility of spatial synaptic structure to sustain a stable synfire chain. Our findings are in good agreement with simulations and help us understand the propagation of spatio-temporal patterns in a random point flied."
                    },
                    {
                        "title": "Learn to integrate parts for whole through correlated neural variability",
                        "abstract": "Sensory perception originates from the responses of sensory neurons, which react to a collection of sensory signals linked to various physical attributes of a singular perceptual object. Unraveling how the brain extracts perceptual information from these neuronal responses is a pivotal challenge in both computational neuroscience and machine learning. Here we introduce a statistical mechanical theory, where perceptual information is first encoded in the correlated variability of sensory neurons and then reformatted into the firing rates of downstream neurons. Applying this theory, we illustrate the encoding of motion direction using neural covariance and demonstrate high-fidelity direction recovery by spiking neural networks. Networks trained under this theory also show enhanced performance in classifying natural images, achieving higher accuracy and faster inference speed. Our results challenge the traditional view of neural covariance as a secondary factor in neural coding, highlighting its potential influence on brain function."
                    },
                    {
                        "title": "On a framework of data assimilation for neuronal networks",
                        "abstract": "When handling real-world data modeled by a complex network dynamical system, the number of the parameters is always even much more than the size of the data. Therefore, in many cases, it is impossible to estimate these parameters and however, the exact value of each parameter is frequently less interesting than the distribution of the parameters may contain important information towards understanding the system and data. In this paper, we propose this question arising by employing a data assimilation approach to estimate the distribution of the parameters in the leakage-integrate-fire (LIF) neuronal network model from the experimental data, for example, the blood-oxygen-level-dependent (BOLD) signal. Herein, we assume that the parameters of the neurons and synapses are inhomogeneous but independently identical distributed following certain distribution with unknown hyperparameters. Thus, we estimate these hyperparameters of the distributions of the parameters, instead of estimating the parameters themselves. We formulate this problem under the framework of data assimilation and hierarchical Bayesian method, and present an efficient method named Hierarchical Data Assimilation (HDA) to conduct the statistical inference on the neuronal network model with the BOLD signal data simulated by the hemodynamic model. We consider the LIF neuronal networks with four synapses and show that the proposed algorithm can estimate the BOLD signals and the hyperparameters with good preciseness. In addition, we discuss the influence on the performance of the algorithm configuration and the LIF network model setup."
                    },
                    {
                        "title": "Preconditioned Score-based Generative Models",
                        "abstract": "Score-based generative models (SGMs) have recently emerged as a promising class of generative models. However, a fundamental limitation is that their sampling process is slow due to a need for many (\\eg, $2000$) iterations of sequential computations. An intuitive acceleration method is to reduce the sampling iterations which however causes severe performance degradation. We assault this problem to the ill-conditioned issues of the Langevin dynamics and reverse diffusion in the sampling process. Under this insight, we propose a model-agnostic {\\bf\\em preconditioned diffusion sampling} (PDS) method that leverages matrix preconditioning to alleviate the aforementioned problem. PDS alters the sampling process of a vanilla SGM at marginal extra computation cost, and without model retraining. Theoretically, we prove that PDS preserves the output distribution of the SGM, no risk of inducing systematical bias to the original sampling process. We further theoretically reveal a relation between the parameter of PDS and the sampling iterations,easing the parameter estimation under varying sampling iterations. Extensive experiments on various image datasets with a variety of resolutions and diversity validate that our PDS consistently accelerates off-the-shelf SGMs whilst maintaining the synthesis quality. In particular, PDS can accelerate by up to $29\\times$ on more challenging high resolution (1024$\\times$1024) image generation. Compared with the latest generative models (\\eg, CLD-SGM, DDIM, and Analytic-DDIM), PDS can achieve the best sampling quality on CIFAR-10 at a FID score of 1.99. Our code is made publicly available to foster any further research https://github.com/fudan-zvg/PDS."
                    },
                    {
                        "title": "GaitSet: Regarding Gait as a Set for Cross-View Gait Recognition",
                        "abstract": "As a unique biometric feature that can be recognized at a distance, gait has broad applications in crime prevention, forensic identification and social security. To portray a gait, existing gait recognition methods utilize either a gait template, where temporal information is hard to preserve, or a gait sequence, which must keep unnecessary sequential constraints and thus loses the flexibility of gait recognition. In this paper we present a novel perspective, where a gait is regarded as a set consisting of independent frames. We propose a new network named GaitSet to learn identity information from the set. Based on the set perspective, our method is immune to permutation of frames, and can naturally integrate frames from different videos which have been filmed under different scenarios, such as diverse viewing angles, different clothes/carrying conditions. Experiments show that under normal walking conditions, our single-model method achieves an average rank-1 accuracy of 95.0% on the CASIA-B gait dataset and an 87.1% accuracy on the OU-MVLP gait dataset. These results represent new state-of-the-art recognition accuracy. On various complex scenarios, our model exhibits a significant level of robustness. It achieves accuracies of 87.2% and 70.4% on CASIA-B under bag-carrying and coat-wearing walking conditions, respectively. These outperform the existing best methods by a large margin. The method presented can also achieve a satisfactory accuracy with a small number of frames in a test sample, e.g., 82.5% on CASIA-B with only 7 frames. The source code has been released at https://github.com/AbnerHqC/GaitSet."
                    }
                ]
            }
        }
    },
    "2406.17711": {
        "paper_data": {
            "title": "Data curation via joint example selection further accelerates multimodal learning",
            "url": "http://arxiv.org/abs/2406.17711v1",
            "arxiv_id": "2406.17711",
            "authors": [
                "Talfan Evans",
                "Nikhil Parthasarathy",
                "Hamza Merzic",
                "Olivier J. Henaff"
            ],
            "abstract": "Data curation is an essential component of large-scale pretraining. In this work, we demonstrate that jointly selecting batches of data is more effective for learning than selecting examples independently. Multimodal contrastive objectives expose the dependencies between data and thus naturally yield criteria for measuring the joint learnability of a batch. We derive a simple and tractable algorithm for selecting such batches, which significantly accelerate training beyond individually-prioritized data points. As performance improves by selecting from larger super-batches, we also leverage recent advances in model approximation to reduce the associated computational overhead. As a result, our approach--multimodal contrastive learning with joint example selection (JEST)--surpasses state-of-the-art models with up to 13$\\times$ fewer iterations and 10$\\times$ less computation. Essential to the performance of JEST is the ability to steer the data selection process towards the distribution of smaller, well-curated datasets via pretrained reference models, exposing the level of data curation as a new dimension for neural scaling laws.",
            "introduction": "   1 Introduction  Data quality is an essential driver of performance for large-scale pretraining. Whether in language [19], vision [15], or multimodal modeling [1, 22, 32], training on well-curated datasets has consistently demonstrated that strong performance can be achieved with significantly less data. However, current data pipelines rely heavily on manual curation, which is difficult and expensive to scale. In contrast, model-based data curation [31, 33], which uses features of the model being trained to select high quality data, holds promise for improving the slow, power-law scaling of large-scale pretraining across modalities, both in theory [47] and in practice [15].   Existing methods apply curation at the level of individual data points [12, 42]. Yet the quality of a batch is also a function of its composition, in addition to the summed quality of its data points considered independently. In computer vision, hard negatives (i.e.\u00a0clusters of points which lie close to one another but contain different labels) have been found to provide a more effective learning signal than trivially solvable ones [5, 21, 34, 41, 45, 50, 53]. In this work we seek to generalize this notion by asking whether model-based data-selection criteria applied to batches of data can accelerate learning beyond what is possible by selecting examples independently.   In multimodal learning, contrastive objectives directly expose the interactions between examples in a batch. We therefore derive a simple and tractable algorithm for joint example selection\u2014JEST\u2014which efficiently selects relevant \u2018sub-batches\u2019 of data from much larger \u2018super-batches\u2019 given their model-based scores. When scoring batches with a pretrained reference model (i.e. easy-reference), JEST accelerates learning relative to uniform batch selection, significantly improving on independent example selection using the same reference (as in CLIPScore [22]). When scoring batches according to their learnability, which also takes into account the online model loss [33], JEST improves further, matching the performance of state-of-the-art models [54] with up to 13\u00d7\\times\u00d7\u00a0fewer training iterations.   Discovering highly learnable batches requires sifting through much larger super-batches of raw data. We make learnability scoring of large batches tractable by leveraging recent advances in online model approximation, which reduce computation while still providing useful predictions [4, 29, 55]. By training a single model at multiple resolutions in parallel, we efficiently apply the model for scoring large super-batches, find their most learnable sub-batch, and spend more valuable computation for learning on them. Thanks to savings in both learning and example scoring, we reduce the overhead of scoring from 133% to 10% additional FLOPs while maintaining significant gains in training efficiency. This approximate scoring framework\u2014Flexi-JEST\u2014produces state-of-the-art models with 11\u00d7\\times\u00d7\u00a0fewer iterations and 10\u00d7\\times\u00d7\u00a0fewer FLOPs.   Finally, we find that central to the performance of our framework is the ability to steer the curation process towards the distribution of smaller, well-curated datasets. This occurs naturally with the model-based selection criteria we consider through the concept of a pretrained reference model, which prioritizes examples that most resemble the data it was trained on. Crucially, we find this process to enable strong data quality bootstrapping: a reference model trained on a small curated dataset can effectively guide the curation of a much larger dataset, allowing the training of a model which strongly surpasses the quality of the reference model on",
            "references": [
                {
                    "title": "Rho-1: Not All Tokens Are What You Need",
                    "abstract": "Previous language model pre-training methods have uniformly applied a next-token prediction loss to all training tokens. Challenging this norm, we posit that ''Not all tokens in a corpus are equally important for language model training''. Our initial analysis examines token-level training dynamics of language model, revealing distinct loss patterns for different tokens. Leveraging these insights, we introduce a new language model called Rho-1. Unlike traditional LMs that learn to predict every next token in a corpus, Rho-1 employs Selective Language Modeling (SLM), which selectively trains on useful tokens that aligned with the desired distribution. This approach involves scoring pretraining tokens using a reference model, and then training the language model with a focused loss on tokens with higher scores. When continual pretraining on 15B OpenWebMath corpus, Rho-1 yields an absolute improvement in few-shot accuracy of up to 30% in 9 math tasks. After fine-tuning, Rho-1-1B and 7B achieved state-of-the-art results of 40.6% and 51.8% on MATH dataset, respectively - matching DeepSeekMath with only 3% of the pretraining tokens. Furthermore, when pretraining on 80B general tokens, Rho-1 achieves 6.8% average enhancement across 15 diverse tasks, increasing both efficiency and performance of the language model pre-training."
                },
                {
                    "title": "Scaling Laws for Data Filtering\u2014Data Curation Cannot be Compute Agnostic",
                    "abstract": "Vision-language models (VLMs) are trained for thousands of GPU hours on carefully selected subsets of massive web scrapes. For instance, the LAION public dataset retained only about 10% of the total crawled data. In recent times, data curation has gained prominence with several works developing strategies to retain \u2018high-quality\u2019 subsets of \u2018raw\u2019 scraped data. However, these strategies are typically developed agnostic to the available compute for training. In this paper, we demonstrate that making filtering decisions independent of training compute is often suboptimal\u2014well-curated data rapidly loses its utility when repeated, eventually decreasing below the utility of \u2018unseen\u2019 but \u2018lower-quality\u2019 data. While past research in neural scaling laws has considered web data to be homogenous, real data is not. Our work bridges this important gap in the literature by developing scaling laws that characterize the differing \u2018utility\u2019 of various data subsets, and accounting for how this diminishes for a data point at its \u2018nth\u2019 repetition. Our key message is that data curation can not be agnostic of the total compute a model will be trained for. Even without ever jointly training on multiple data buckets, our scaling laws enable us to estimate model performance under this dynamic trade-off between quality and repetition. This allows us to curate the best possible pool for achieving top performance on Datacomp at various compute budgets, carving out a pareto-frontier for data curation."
                },
                {
                    "title": "Mixture-of-Depths: Dynamically allocating compute in transformer-based language models",
                    "abstract": "Transformer-based language models spread FLOPs uniformly across input sequences. In this work we demonstrate that transformers can instead learn to dynamically allocate FLOPs (or compute) to specific positions in a sequence, optimising the allocation along the sequence for different layers across the model depth. Our method enforces a total compute budget by capping the number of tokens ($k$) that can participate in the self-attention and MLP computations at a given layer. The tokens to be processed are determined by the network using a top-$k$ routing mechanism. Since $k$ is defined a priori, this simple procedure uses a static computation graph with known tensor sizes, unlike other conditional computation techniques. Nevertheless, since the identities of the $k$ tokens are fluid, this method can expend FLOPs non-uniformly across the time and model depth dimensions. Thus, compute expenditure is entirely predictable in sum total, but dynamic and context-sensitive at the token-level. Not only do models trained in this way learn to dynamically allocate compute, they do so efficiently. These models match baseline performance for equivalent FLOPS and wall-clock times to train, but require a fraction of the FLOPs per forward pass, and can be upwards of 50\\% faster to step during post-training sampling."
                },
                {
                    "title": "How to Train Data-Efficient LLMs",
                    "abstract": "The training of large language models (LLMs) is expensive. In this paper, we study data-efficient approaches for pre-training LLMs, i.e., techniques that aim to optimize the Pareto frontier of model quality and training resource/data consumption. We seek to understand the tradeoffs associated with data selection routines based on (i) expensive-to-compute data-quality estimates, and (ii) maximization of coverage and diversity-based measures in the feature space. Our first technique, Ask-LLM, leverages the zero-shot reasoning capabilities of instruction-tuned LLMs to directly assess the quality of a training example. To target coverage, we propose Density sampling, which models the data distribution to select a diverse sample. In our comparison of 19 samplers, involving hundreds of evaluation tasks and pre-training runs, we find that Ask-LLM and Density are the best methods in their respective categories. Coverage sampling can recover the performance of the full data, while models trained on Ask-LLM data consistently outperform full-data training -- even when we reject 90% of the original dataset, while converging up to 70% faster."
                },
                {
                    "title": "Effective pruning of web-scale datasets based on complexity of concept clusters",
                    "abstract": "Utilizing massive web-scale datasets has led to unprecedented performance gains in machine learning models, but also imposes outlandish compute requirements for their training. In order to improve training and data efficiency, we here push the limits of pruning large-scale multimodal datasets for training CLIP-style models. Today's most effective pruning method on ImageNet clusters data samples into separate concepts according to their embedding and prunes away the most prototypical samples. We scale this approach to LAION and improve it by noting that the pruning rate should be concept-specific and adapted to the complexity of the concept. Using a simple and intuitive complexity measure, we are able to reduce the training cost to a quarter of regular training. By filtering from the LAION dataset, we find that training on a smaller set of high-quality data can lead to higher performance with significantly lower training costs. More specifically, we are able to outperform the LAION-trained OpenCLIP-ViT-B32 model on ImageNet zero-shot accuracy by 1.1p.p. while only using 27.7% of the data and training compute. Despite a strong reduction in training cost, we also see improvements on ImageNet dist. shifts, retrieval tasks and VTAB. On the DataComp Medium benchmark, we achieve a new state-of-the-art Imagehttps://info.arxiv.org/help/prep#commentsNet zero-shot accuracy and a competitive average zero-shot accuracy on 38 evaluation tasks."
                },
                {
                    "title": "Bad Students Make Great Teachers: Active Learning Accelerates Large-Scale Visual Understanding",
                    "abstract": "Power-law scaling indicates that large-scale training with uniform sampling is prohibitively slow. Active learning methods aim to increase data efficiency by prioritizing learning on the most relevant examples. Despite their appeal, these methods have yet to be widely adopted since no one algorithm has been shown to a) generalize across models and tasks b) scale to large datasets and c) yield overall FLOP savings when accounting for the overhead of data selection. In this work we propose a method which satisfies these three properties, leveraging small, cheap proxy models to estimate\"learnability\"scores for datapoints, which are used to prioritize data for the training of much larger models. As a result, our models require 46% and 51% fewer training updates and up to 25% less total computation to reach the same performance as uniformly trained visual classifiers on JFT and multimodal models on ALIGN. Finally, we find our data-prioritization scheme to be complementary with recent data-curation and learning objectives, yielding a new state-of-the-art in several multimodal transfer tasks."
                },
                {
                    "title": "PaLI-3 Vision Language Models: Smaller, Faster, Stronger",
                    "abstract": "This paper presents PaLI-3, a smaller, faster, and stronger vision language model (VLM) that compares favorably to similar models that are 10x larger. As part of arriving at this strong performance, we compare Vision Transformer (ViT) models pretrained using classification objectives to contrastively (SigLIP) pretrained ones. We find that, while slightly underperforming on standard image classification benchmarks, SigLIP-based PaLI shows superior performance across various multimodal benchmarks, especially on localization and visually-situated text understanding. We scale the SigLIP image encoder up to 2 billion parameters, and achieves a new state-of-the-art on multilingual cross-modal retrieval. We hope that PaLI-3, at only 5B parameters, rekindles research on fundamental pieces of complex VLMs, and could fuel a new generation of scaled-up models."
                },
                {
                    "title": "Sieve: Multimodal Dataset Pruning Using Image Captioning Models",
                    "abstract": "Vision-Language Models (VLMs) are pretrained on large, diverse, and noisy web-crawled datasets. This underscores the critical need for dataset pruning, as the quality of these datasets is strongly correlated with the performance of VLMs on downstream tasks. Using CLIPScore from a pretrained model to only train models using highly-aligned samples is one of the most successful methods for pruning. We argue that this approach suffers from multiple limitations including: false positives and negatives due to CLIP's pretraining on noisy labels. We propose a pruning signal, Sieve, that employs synthetic captions generated by image-captioning models pretrained on small, diverse, and well-aligned image-text pairs to evaluate the alignment of noisy image-text pairs. To bridge the gap between the limited diversity of generated captions and the high diversity of alternative text (alt-text), we estimate the semantic textual similarity in the embedding space of a language model pretrained on unlabeled text corpus. Using DataComp, a multimodal dataset filtering benchmark, when evaluating on 38 downstream tasks, our pruning approach, surpasses CLIPScore by 2.6% and 1.7% on medium and large scale respectively. In addition, on retrieval tasks, Sieve leads to a significant improvement of 2.7% and 4.5% on medium and large scale respectively."
                },
                {
                    "title": "Data Filtering Networks",
                    "abstract": "Large training sets have become a cornerstone of machine learning and are the foundation for recent advances in language modeling and multimodal learning. While data curation for pre-training is often still ad-hoc, one common paradigm is to first collect a massive pool of data from the Web and then filter this candidate pool down to an actual training set via various heuristics. In this work, we study the problem of learning a data filtering network (DFN) for this second step of filtering a large uncurated dataset. Our key finding is that the quality of a network for filtering is distinct from its performance on downstream tasks: for instance, a model that performs well on ImageNet can yield worse training sets than a model with low ImageNet accuracy that is trained on a small amount of high-quality data. Based on our insights, we construct new data filtering networks that induce state-of-the-art image-text datasets. Specifically, our best performing dataset DFN-5B enables us to train state-of-the-art models for their compute budgets: among other improvements on a variety of tasks, a ViT-H trained on our dataset achieves 83.0% zero-shot transfer accuracy on ImageNet, out-performing models trained on other datasets such as LAION-2B, DataComp-1B, or OpenAI's WIT. In order to facilitate further research in dataset design, we also release a new 2 billion example dataset DFN-2B and show that high performance data filtering networks can be trained from scratch using only publicly available data."
                },
                {
                    "title": "Demystifying CLIP Data",
                    "abstract": "Contrastive Language-Image Pre-training (CLIP) is an approach that has advanced research and applications in computer vision, fueling modern recognition systems and generative models. We believe that the main ingredient to the success of CLIP is its data and not the model architecture or pre-training objective. However, CLIP only provides very limited information about its data and how it has been collected, leading to works that aim to reproduce CLIP's data by filtering with its model parameters. In this work, we intend to reveal CLIP's data curation approach and in our pursuit of making it open to the community introduce Metadata-Curated Language-Image Pre-training (MetaCLIP). MetaCLIP takes a raw data pool and metadata (derived from CLIP's concepts) and yields a balanced subset over the metadata distribution. Our experimental study rigorously isolates the model and training settings, concentrating solely on data. MetaCLIP applied to CommonCrawl with 400M image-text data pairs outperforms CLIP's data on multiple standard benchmarks. In zero-shot ImageNet classification, MetaCLIP achieves 70.8% accuracy, surpassing CLIP's 68.3% on ViT-B models. Scaling to 1B data, while maintaining the same training budget, attains 72.4%. Our observations hold across various model sizes, exemplified by ViT-H achieving 80.5%, without any bells-and-whistles. Curation code and training data distribution on metadata is made available at https://github.com/facebookresearch/MetaCLIP."
                },
                {
                    "title": "Patch n' Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution",
                    "abstract": "The ubiquitous and demonstrably suboptimal choice of resizing images to a fixed resolution before processing them with computer vision models has not yet been successfully challenged. However, models such as the Vision Transformer (ViT) offer flexible sequence-based modeling, and hence varying input sequence lengths. We take advantage of this with NaViT (Native Resolution ViT) which uses sequence packing during training to process inputs of arbitrary resolutions and aspect ratios. Alongside flexible model usage, we demonstrate improved training efficiency for large-scale supervised and contrastive image-text pretraining. NaViT can be efficiently transferred to standard tasks such as image and video classification, object detection, and semantic segmentation and leads to improved results on robustness and fairness benchmarks. At inference time, the input resolution flexibility can be used to smoothly navigate the test-time cost-performance trade-off. We believe that NaViT marks a departure from the standard, CNN-designed, input and modelling pipeline used by most computer vision models, and represents a promising direction for ViTs."
                },
                {
                    "title": "Textbooks Are All You Need",
                    "abstract": "We introduce phi-1, a new large language model for code, with significantly smaller size than competing models: phi-1 is a Transformer-based model with 1.3B parameters, trained for 4 days on 8 A100s, using a selection of ``textbook quality\"data from the web (6B tokens) and synthetically generated textbooks and exercises with GPT-3.5 (1B tokens). Despite this small scale, phi-1 attains pass@1 accuracy 50.6% on HumanEval and 55.5% on MBPP. It also displays surprising emergent properties compared to phi-1-base, our model before our finetuning stage on a dataset of coding exercises, and phi-1-small, a smaller model with 350M parameters trained with the same pipeline as phi-1 that still achieves 45% on HumanEval."
                },
                {
                    "title": "An Inverse Scaling Law for CLIP Training",
                    "abstract": "CLIP, one of the pioneering foundation models that connect images and text, has enabled many recent breakthroughs in computer vision. However, its associated training cost is prohibitively high, imposing a significant barrier to its widespread exploration. In this paper, we present a surprising finding that there exists an inverse scaling law for CLIP training, whereby the larger the image/text encoders used, the shorter the sequence length of image/text tokens that can be applied in training. Moreover, we showcase that the strategy for reducing image/text token length plays a crucial role in determining the quality of this scaling law. As a result of this finding, we are able to successfully train CLIP even with limited computational resources. For example, using 8 A100 GPUs, our CLIP models achieve zero-shot top-1 ImageNet-1k accuracies of 63.2% in ~2 days, 67.8% in ~3 days, and 69.3% in ~4 days. Our method also works well when scaling up -- with G/14, we register a new record of 83.0% ImageNet-1k zero-shot accuracy, and meanwhile accelerate the training by ~33x compared to its OpenCLIP counterpart. By reducing the computation barrier associated with CLIP, we hope to inspire more research in this field, particularly from academics. Our code is available at https://github.com/UCSC-VLAA/CLIPA."
                },
                {
                    "title": "Less is More: Removing Text-regions Improves CLIP Training Efficiency and Robustness",
                    "abstract": "The CLIP (Contrastive Language-Image Pre-training) model and its variants are becoming the de facto backbone in many applications. However, training a CLIP model from hundreds of millions of image-text pairs can be prohibitively expensive. Furthermore, the conventional CLIP model doesn't differentiate between the visual semantics and meaning of text regions embedded in images. This can lead to non-robustness when the text in the embedded region doesn't match the image's visual appearance. In this paper, we discuss two effective approaches to improve the efficiency and robustness of CLIP training: (1) augmenting the training dataset while maintaining the same number of optimization steps, and (2) filtering out samples that contain text regions in the image. By doing so, we significantly improve the classification and retrieval accuracy on public benchmarks like ImageNet and CoCo. Filtering out images with text regions also protects the model from typographic attacks. To verify this, we build a new dataset named ImageNet with Adversarial Text Regions (ImageNet-Attr). Our filter-based CLIP model demonstrates a top-1 accuracy of 68.78\\%, outperforming previous models whose accuracy was all below 50\\%."
                },
                {
                    "title": "DataComp: In search of the next generation of multimodal datasets",
                    "abstract": "Multimodal datasets are a critical component in recent breakthroughs such as Stable Diffusion and GPT-4, yet their design does not receive the same research attention as model architectures or training algorithms. To address this shortcoming in the ML ecosystem, we introduce DataComp, a testbed for dataset experiments centered around a new candidate pool of 12.8 billion image-text pairs from Common Crawl. Participants in our benchmark design new filtering techniques or curate new data sources and then evaluate their new dataset by running our standardized CLIP training code and testing the resulting model on 38 downstream test sets. Our benchmark consists of multiple compute scales spanning four orders of magnitude, which enables the study of scaling trends and makes the benchmark accessible to researchers with varying resources. Our baseline experiments show that the DataComp workflow leads to better training sets. In particular, our best baseline, DataComp-1B, enables training a CLIP ViT-L/14 from scratch to 79.2% zero-shot accuracy on ImageNet, outperforming OpenAI's CLIP ViT-L/14 by 3.7 percentage points while using the same training procedure and compute. We release DataComp and all accompanying code at www.datacomp.ai."
                },
                {
                    "title": "Sigmoid Loss for Language Image Pre-Training",
                    "abstract": "We propose a simple pairwise sigmoid loss for imagetext pre-training. Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization. The sigmoid loss simultaneously allows further scaling up the batch size, while also performing better at smaller batch sizes. With only four TPUv4 chips, we can train a Base CLIP model at 4k batch size and a Large LiT model at 20k batch size, the latter achieves 84.5% ImageNet zero-shot accuracy in two days. This disentanglement of the batch size from the loss further allows us to study the impact of examples vs pairs and negative to positive ratio. Finally, we push the batch size to the extreme, up to one million, and find that the benefits of growing batch size quickly diminish, with a more reasonable batch size of 32k being sufficient. We hope our research motivates further explorations in improving the quality and efficiency of language-image pre-training."
                },
                {
                    "title": "EVA-CLIP: Improved Training Techniques for CLIP at Scale",
                    "abstract": "Contrastive language-image pre-training, CLIP for short, has gained increasing attention for its potential in various scenarios. In this paper, we propose EVA-CLIP, a series of models that significantly improve the efficiency and effectiveness of CLIP training. Our approach incorporates new techniques for representation learning, optimization, and augmentation, enabling EVA-CLIP to achieve superior performance compared to previous CLIP models with the same number of parameters but significantly smaller training costs. Notably, our largest 5.0B-parameter EVA-02-CLIP-E/14+ with only 9 billion seen samples achieves 82.0 zero-shot top-1 accuracy on ImageNet-1K val. A smaller EVA-02-CLIP-L/14+ with only 430 million parameters and 6 billion seen samples achieves 80.4 zero-shot top-1 accuracy on ImageNet-1K val. To facilitate open access and open research, we release the complete suite of EVA-CLIP to the community at https://github.com/baaivision/EVA/tree/master/EVA-CLIP."
                },
                {
                    "title": "SemDeDup: Data-efficient learning at web-scale through semantic deduplication",
                    "abstract": "Progress in machine learning has been driven in large part by massive increases in data. However, large web-scale datasets such as LAION are largely uncurated beyond searches for exact duplicates, potentially leaving much redundancy. Here, we introduce SemDeDup, a method which leverages embeddings from pre-trained models to identify and remove semantic duplicates: data pairs which are semantically similar, but not exactly identical. Removing semantic duplicates preserves performance and speeds up learning. Analyzing a subset of LAION, we show that SemDeDup can remove 50% of the data with minimal performance loss, effectively halving training time. Moreover, performance increases out of distribution. Also, analyzing language models trained on C4, a partially curated dataset, we show that SemDeDup improves over prior approaches while providing efficiency gains. SemDeDup provides an example of how simple ways of leveraging quality embeddings can be used to make models learn faster with less data."
                },
                {
                    "title": "Filtering, Distillation, and Hard Negatives for Vision-Language Pre-Training",
                    "abstract": "Vision-language models trained with contrastive learning on large-scale noisy data are becoming increasingly popular for zero-shot recognition problems. In this paper we improve the following three aspects of the contrastive pre-training pipeline: dataset noise, model initialization and the training objective. First, we propose a straightforward filtering strategy titled Complexity, Action, and Text-spotting (CAT) that significantly reduces dataset size, while achieving improved performance across zero-shot vision-language tasks. Next, we propose an approach titled Concept Distillation to leverage strong unimodal representations for contrastive training that does not increase training complexity while outperforming prior work. Finally, we modify the traditional contrastive alignment objective, and propose an importance-sampling approach to up-sample the importance of hard-negatives without adding additional complexity. On an extensive zero-shot benchmark of 29 tasks, our Distilled and Hard-negative Training (DiHT) approach improves on 20 tasks compared to the baseline. Furthermore, for few-shot linear probing, we propose a novel approach that bridges the gap between zero-shot and few-shot performance, substantially improving over prior work. Models are available at github.com/facebookresearch/diht."
                },
                {
                    "title": "FlexiViT: One Model for All Patch Sizes",
                    "abstract": "Vision Transformers convert images to sequences by slicing them into patches. The size of these patches controls a speed/accuracy tradeoff, with smaller patches leading to higher accuracy at greater computational cost, but changing the patch size typically requires retraining the model. In this paper, we demonstrate that simply randomizing the patch size at training time leads to a single set of weights that performs well across a wide range of patch sizes, making it possible to tailor the model to different compute budgets at deployment time. We extensively evaluate the resulting model, which we call FlexiViT, on a wide range of tasks, including classification, image-text retrieval, open-world detection, panoptic segmentation, and semantic segmentation, concluding that it usually matches, and sometimes outperforms, standard ViT models trained at a single patch size in an otherwise identical setup. Hence, FlexiViT training is a simple drop-in improvement for ViT that makes it easy to add compute-adaptive capabilities to most models relying on a ViT backbone architecture. Code and pre-trained models are available at github.com/google-research/big_vision."
                },
                {
                    "title": "LAION-5B: An open large-scale dataset for training next generation image-text models",
                    "abstract": "Groundbreaking language-vision architectures like CLIP and DALL-E proved the utility of training on large amounts of noisy image-text data, without relying on expensive accurate labels used in standard vision unimodal supervised learning. The resulting models showed capabilities of strong text-guided image generation and transfer to downstream tasks, while performing remarkably at zero-shot classification with noteworthy out-of-distribution robustness. Since then, large-scale language-vision models like ALIGN, BASIC, GLIDE, Flamingo and Imagen made further improvements. Studying the training and capabilities of such models requires datasets containing billions of image-text pairs. Until now, no datasets of this size have been made openly available for the broader research community. To address this problem and democratize research on large-scale multi-modal models, we present LAION-5B - a dataset consisting of 5.85 billion CLIP-filtered image-text pairs, of which 2.32B contain English language. We show successful replication and fine-tuning of foundational models like CLIP, GLIDE and Stable Diffusion using the dataset, and discuss further experiments enabled with an openly available dataset of this scale. Additionally we provide several nearest neighbor indices, an improved web-interface for dataset exploration and subset generation, and detection scores for watermark, NSFW, and toxic content detection. Announcement page https://laion.ai/laion-5b-a-new-era-of-open-large-scale-multi-modal-datasets/"
                },
                {
                    "title": "Make-A-Video: Text-to-Video Generation without Text-Video Data",
                    "abstract": "We propose Make-A-Video -- an approach for directly translating the tremendous recent progress in Text-to-Image (T2I) generation to Text-to-Video (T2V). Our intuition is simple: learn what the world looks like and how it is described from paired text-image data, and learn how the world moves from unsupervised video footage. Make-A-Video has three advantages: (1) it accelerates training of the T2V model (it does not need to learn visual and multimodal representations from scratch), (2) it does not require paired text-video data, and (3) the generated videos inherit the vastness (diversity in aesthetic, fantastical depictions, etc.) of today's image generation models. We design a simple yet effective way to build on T2I models with novel and effective spatial-temporal modules. First, we decompose the full temporal U-Net and attention tensors and approximate them in space and time. Second, we design a spatial temporal pipeline to generate high resolution and frame rate videos with a video decoder, interpolation model and two super resolution models that can enable various applications besides T2V. In all aspects, spatial and temporal resolution, faithfulness to text, and quality, Make-A-Video sets the new state-of-the-art in text-to-video generation, as determined by both qualitative and quantitative measures."
                },
                {
                    "title": "PaLI: A Jointly-Scaled Multilingual Language-Image Model",
                    "abstract": "Effective scaling and a flexible task interface enable large language models to excel at many tasks. We present PaLI (Pathways Language and Image model), a model that extends this approach to the joint modeling of language and vision. PaLI generates text based on visual and textual inputs, and with this interface performs many vision, language, and multimodal tasks, in many languages. To train PaLI, we make use of large pre-trained encoder-decoder language models and Vision Transformers (ViTs). This allows us to capitalize on their existing capabilities and leverage the substantial cost of training them. We find that joint scaling of the vision and language components is important. Since existing Transformers for language are much larger than their vision counterparts, we train a large, 4-billion parameter ViT (ViT-e) to quantify the benefits from even larger-capacity vision models. To train PaLI, we create a large multilingual mix of pretraining tasks, based on a new image-text training set containing 10B images and texts in over 100 languages. PaLI achieves state-of-the-art in multiple vision and language tasks (such as captioning, visual question-answering, scene-text understanding), while retaining a simple, modular, and scalable design."
                },
                {
                    "title": "Beyond neural scaling laws: beating power law scaling via data pruning",
                    "abstract": "Widely observed neural scaling laws, in which error falls off as a power of the training set size, model size, or both, have driven substantial performance improvements in deep learning. However, these improvements through scaling alone require considerable costs in compute and energy. Here we focus on the scaling of error with dataset size and show how in theory we can break beyond power law scaling and potentially even reduce it to exponential scaling instead if we have access to a high-quality data pruning metric that ranks the order in which training examples should be discarded to achieve any pruned dataset size. We then test this improved scaling prediction with pruned dataset size empirically, and indeed observe better than power law scaling in practice on ResNets trained on CIFAR-10, SVHN, and ImageNet. Next, given the importance of finding high-quality pruning metrics, we perform the first large-scale benchmarking study of ten different data pruning metrics on ImageNet. We find most existing high performing metrics scale poorly to ImageNet, while the best are computationally intensive and require labels for every image. We therefore developed a new simple, cheap and scalable self-supervised pruning metric that demonstrates comparable performance to the best supervised metrics. Overall, our work suggests that the discovery of good data-pruning metrics may provide a viable path forward to substantially improved neural scaling laws, thereby reducing the resource costs of modern deep learning."
                },
                {
                    "title": "Prioritized Training on Points that are Learnable, Worth Learning, and Not Yet Learnt",
                    "abstract": "Training on web-scale data can take months. But most computation and time is wasted on redundant and noisy points that are already learnt or not learnable. To accelerate training, we introduce Reducible Holdout Loss Selection (RHO-LOSS), a simple but principled technique which selects approximately those points for training that most reduce the model's generalization loss. As a result, RHO-LOSS mitigates the weaknesses of existing data selection methods: techniques from the optimization literature typically select 'hard' (e.g. high loss) points, but such points are often noisy (not learnable) or less task-relevant. Conversely, curriculum learning prioritizes 'easy' points, but such points need not be trained on once learned. In contrast, RHO-LOSS selects points that are learnable, worth learning, and not yet learnt. RHO-LOSS trains in far fewer steps than prior art, improves accuracy, and speeds up training on a wide range of datasets, hyperparameters, and architectures (MLPs, CNNs, and BERT). On the large web-scraped image dataset Clothing-1M, RHO-LOSS trains in 18x fewer steps and reaches 2% higher final accuracy than uniform data shuffling."
                },
                {
                    "title": "BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation",
                    "abstract": "Vision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has been largely achieved by scaling up the dataset with noisy image-text pairs collected from the web, which is a suboptimal source of supervision. In this paper, we propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks. BLIP effectively utilizes the noisy web data by bootstrapping the captions, where a captioner generates synthetic captions and a filter removes the noisy ones. We achieve state-of-the-art results on a wide range of vision-language tasks, such as image-text retrieval (+2.7% in average recall@1), image captioning (+2.8% in CIDEr), and VQA (+1.6% in VQA score). BLIP also demonstrates strong generalization ability when directly transferred to video-language tasks in a zero-shot manner. Code, models, and datasets are released at https://github.com/salesforce/BLIP."
                },
                {
                    "title": "Deep Learning on a Data Diet: Finding Important Examples Early in Training",
                    "abstract": "Recent success in deep learning has partially been driven by training increasingly overparametrized networks on ever larger datasets. It is therefore natural to ask: how much of the data is superfluous, which examples are important for generalization, and how do we find them? In this work, we make the striking observation that, in standard vision datasets, simple scores averaged over several weight initializations can be used to identify important examples very early in training. We propose two such scores -- the Gradient Normed (GraNd) and the Error L2-Norm (EL2N) scores -- and demonstrate their efficacy on a range of architectures and datasets by pruning significant fractions of training data without sacrificing test accuracy. In fact, using EL2N scores calculated a few epochs into training, we can prune half of the CIFAR10 training set while slightly improving test accuracy. Furthermore, for a given dataset, EL2N scores from one architecture or hyperparameter configuration generalize to other configurations. Compared to recent work that prunes data by discarding examples that are rarely forgotten over the course of training, our scores use only local information early in training. We also use our scores to detect noisy examples and study training dynamics through the lens of important examples -- we investigate how the data distribution shapes the loss surface and identify subspaces of the model's data representation that are relatively stable over training."
                },
                {
                    "title": "Divide and Contrast: Self-supervised Learning from Uncurated Data",
                    "abstract": "Self-supervised learning holds promise in leveraging large amounts of unlabeled data, however much of its progress has thus far been limited to highly curated pre-training data such as ImageNet. We explore the effects of contrastive learning from larger, less-curated image datasets such as YFCC, and find there is indeed a large difference in the resulting representation quality. We hypothesize that this curation gap is due to a shift in the distribution of image classes\u2014which is more diverse and heavy-tailed\u2014resulting in less relevant negative samples to learn from. We test this hypothesis with a new approach, Divide and Contrast (DnC), which alternates between contrastive learning and clustering-based hard negative mining. When pretrained on less curated datasets, DnC greatly improves the performance of self-supervised learning on downstream tasks, while remaining competitive with the current state-of-the-art on curated datasets."
                },
                {
                    "title": "CLIPScore: A Reference-free Evaluation Metric for Image Captioning",
                    "abstract": "Image captioning has conventionally relied on reference-based automatic evaluations, where machine captions are compared against captions written by humans. This is in contrast to the reference-free manner in which humans assess caption quality. In this paper, we report the surprising empirical finding that CLIP (Radford et al., 2021), a cross-modal model pretrained on 400M image+caption pairs from the web, can be used for robust automatic evaluation of image captioning without the need for references. Experiments spanning several corpora demonstrate that our new reference-free metric, CLIPScore, achieves the highest correlation with human judgements, outperforming existing reference-based metrics like CIDEr and SPICE. Information gain experiments demonstrate that CLIPScore, with its tight focus on image-text compatibility, is complementary to existing reference-based metrics that emphasize text-text similarities. Thus, we also present a reference-augmented version, RefCLIPScore, which achieves even higher correlation. Beyond literal description tasks, several case studies reveal domains where CLIPScore performs well (clip-art images, alt-text rating), but also where it is relatively weaker in comparison to reference-based metrics, e.g., news captions that require richer contextual knowledge."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Conceptual 12M: Pushing Web-Scale Image-Text Pre-Training To Recognize Long-Tail Visual Concepts",
                    "abstract": "The availability of large-scale image captioning and visual question answering datasets has contributed significantly to recent successes in vision-and-language pretraining. However, these datasets are often collected with overrestrictive requirements inherited from their original target tasks (e.g., image caption generation), which limit the resulting dataset scale and diversity. We take a step further in pushing the limits of vision-and-language pretraining data by relaxing the data collection pipeline used in Conceptual Captions 3M (CC3M) [54] and introduce the Conceptual 12M (CC12M), a dataset with 12 million image-text pairs specifically meant to be used for visionand-language pre-training. We perform an analysis of this dataset and benchmark its effectiveness against CC3M on multiple downstream tasks with an emphasis on long-tail visual recognition. Our results clearly illustrate the benefit of scaling up pre-training data for vision-and-language tasks, as indicated by the new state-of-the-art results on both the nocaps and Conceptual Captions benchmarks.1"
                },
                {
                    "title": "Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision",
                    "abstract": "Pre-trained representations are becoming crucial for many NLP and perception tasks. While representation learning in NLP has transitioned to training on raw text without human annotations, visual and vision-language representations still rely heavily on curated training datasets that are expensive or require expert knowledge. For vision applications, representations are mostly learned using datasets with explicit class labels such as ImageNet or OpenImages. For vision-language, popular datasets like Conceptual Captions, MSCOCO, or CLIP all involve a non-trivial data collection (and cleaning) process. This costly curation process limits the size of datasets and hence hinders the scaling of trained models. In this paper, we leverage a noisy dataset of over one billion image alt-text pairs, obtained without expensive filtering or post-processing steps in the Conceptual Captions dataset. A simple dual-encoder architecture learns to align visual and language representations of the image and text pairs using a contrastive loss. We show that the scale of our corpus can make up for its noise and leads to state-of-the-art representations even with such a simple learning scheme. Our visual representation achieves strong performance when transferred to classification tasks such as ImageNet and VTAB. The aligned visual and language representations enables zero-shot image classification and also set new state-of-the-art results on Flickr30K and MSCOCO image-text retrieval benchmarks, even when compared with more sophisticated cross-attention models. The representations also enable cross-modality search with complex text and text + image queries."
                },
                {
                    "title": "Accelerating Training of Transformer-Based Language Models with Progressive Layer Dropping",
                    "abstract": "Recently, Transformer-based language models have demonstrated remarkable performance across many NLP domains. However, the unsupervised pre-training step of these models suffers from unbearable overall computational expenses. Current methods for accelerating the pre-training either rely on massive parallelism with advanced hardware or are not applicable to language modeling. In this work, we propose a method based on progressive layer dropping that speeds the training of Transformer-based language models, not at the cost of excessive hardware resources but from model architecture change and training technique boosted efficiency. Extensive experiments on BERT show that the proposed method achieves a 24% time reduction on average per sample and allows the pre-training to be 2.5 times faster than the baseline to get a similar accuracy on downstream tasks. While being faster, our pre-trained models are equipped with strong knowledge transferability, achieving comparable and sometimes higher GLUE score than the baseline when pre-trained with the same number of samples."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Contrastive Learning with Hard Negative Samples",
                    "abstract": "How can you sample good negative examples for contrastive learning? We argue that, as with metric learning, contrastive learning of representations benefits from hard negative samples (i.e., points that are difficult to distinguish from an anchor point). The key challenge toward using hard negatives is that contrastive methods must remain unsupervised, making it infeasible to adopt existing negative sampling strategies that use true similarity information. In response, we develop a new family of unsupervised sampling methods for selecting hard negative samples where the user can control the hardness. A limiting case of this sampling results in a representation that tightly clusters each class, and pushes different classes as far apart as possible. The proposed method improves downstream performance across multiple modalities, requires only few additional lines of code to implement, and introduces no computational overhead."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "Selection Via Proxy: Efficient Data Selection For Deep Learning",
                    "abstract": "Data selection methods such as active learning and core-set selection are useful tools for machine learning on large datasets, but they can be prohibitively expensive to apply in deep learning. Unlike in other areas of machine learning, the feature representations that these techniques depend on are learned in deep learning rather than given, which takes a substantial amount of training time. In this work, we show that we can significantly improve the computational efficiency of data selection in deep learning by using a much smaller proxy model to perform data selection for tasks that will eventually require a large target model (e.g., selecting data points to label for active learning). In deep learning, we can scale down models by removing hidden layers or reducing their dimension to create proxies that are an order of magnitude faster. Although these small proxy models have significantly higher error, we find that they empirically provide useful rankings for data selection that have a high correlation with those of larger models. We evaluate this \"selection via proxy\" (SVP) approach on several data selection tasks. For active learning, applying SVP to Sener and Savarese [2018]'s recent method for active learning in deep learning gives a 4x improvement in execution time while yielding the same model accuracy. For core-set selection, we show that a proxy model that trains 10x faster than a target ResNet164 model on CIFAR10 can be used to remove 50% of the training data without compromising the accuracy of the target model, making end-to-end training time improvements via core-set selection possible."
                },
                {
                    "title": "SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing",
                    "abstract": "This paper describes SentencePiece, a language-independent subword tokenizer and detokenizer designed for Neural-based text processing, including Neural Machine Translation. It provides open-source C++ and Python implementations for subword units. While existing subword segmentation tools assume that the input is pre-tokenized into word sequences, SentencePiece can train subword models directly from raw sentences, which allows us to make a purely end-to-end and language independent system. We perform a validation experiment of NMT on English-Japanese machine translation, and find that it is possible to achieve comparable accuracy to direct subword training from raw sentences. We also compare the performance of subword training and segmentation with various configurations. SentencePiece is available under the Apache 2 license at https://github.com/google/sentencepiece."
                },
                {
                    "title": "Bayesian Coreset Construction via Greedy Iterative Geodesic Ascent",
                    "abstract": "Coherent uncertainty quantification is a key strength of Bayesian methods. But modern algorithms for approximate Bayesian posterior inference often sacrifice accurate posterior uncertainty estimation in the pursuit of scalability. This work shows that previous Bayesian coreset construction algorithms---which build a small, weighted subset of the data that approximates the full dataset---are no exception. We demonstrate that these algorithms scale the coreset log-likelihood suboptimally, resulting in underestimated posterior uncertainty. To address this shortcoming, we develop greedy iterative geodesic ascent (GIGA), a novel algorithm for Bayesian coreset construction that scales the coreset log-likelihood optimally. GIGA provides geometric decay in posterior approximation error as a function of coreset size, and maintains the fast running time of its predecessors. The paper concludes with validation of GIGA on both synthetic and real datasets, demonstrating that it reduces posterior approximation error by orders of magnitude compared with previous coreset constructions."
                },
                {
                    "title": "Sampling Matters in Deep Embedding Learning",
                    "abstract": "Deep embeddings answer one simple question: How similar are two images? Learning these embeddings is the bedrock of verification, zero-shot learning, and visual search. The most prominent approaches optimize a deep convolutional network with a suitable loss function, such as contrastive loss or triplet loss. While a rich line of work focuses solely on the loss functions, we show in this paper that selecting training examples plays an equally important role. We propose distance weighted sampling, which selects more informative and stable examples than traditional approaches. In addition, we show that a simple margin based loss is sufficient to outperform all other loss functions. We evaluate our approach on the Stanford Online Products, CAR196, and the CUB200-2011 datasets for image retrieval and clustering, and on the LFW dataset for face verification. Our method achieves state-of-the-art performance on all of them."
                },
                {
                    "title": "Working hard to know your neighbor's margins: Local descriptor learning loss",
                    "abstract": "We introduce a novel loss for learning local feature descriptors which is inspired by the Lowe's matching criterion for SIFT. We show that the proposed loss that maximizes the distance between the closest positive and closest negative patch in the batch is better than complex regularization methods; it works well for both shallow and deep convolution network architectures. Applying the novel loss to the L2Net CNN architecture results in a compact descriptor -- it has the same dimensionality as SIFT (128) that shows state-of-art performance in wide baseline stereo, patch verification and instance retrieval benchmarks. It is fast, computing a descriptor takes about 1 millisecond on a low-end GPU."
                },
                {
                    "title": "Smart Mining for Deep Metric Learning",
                    "abstract": "To solve deep metric learning problems and producing feature embeddings, current methodologies will commonly use a triplet model to minimise the relative distance between samples from the same class and maximise the relative distance between samples from different classes. Though successful, the training convergence of this triplet model can be compromised by the fact that the vast majority of the training samples will produce gradients with magnitudes that are close to zero. This issue has motivated the development of methods that explore the global structure of the embedding and other methods that explore hard negative/positive mining. The effectiveness of such mining methods is often associated with intractable computational requirements. In this paper, we propose a novel deep metric learning method that combines the triplet model and the global structure of the embedding space. We rely on a smart mining procedure that produces effective training samples for a low computational cost. In addition, we propose an adaptive controller that automatically adjusts the smart mining hyper-parameters and speeds up the convergence of the training process. We show empirically that our proposed method allows for fast and more accurate training of triplet ConvNets than other competing mining methods. Additionally, we show that our method achieves new state-of-the-art embedding results for CUB-200-2011 and Cars196 datasets."
                },
                {
                    "title": "Discriminative Learning of Deep Convolutional Feature Point Descriptors",
                    "abstract": "Deep learning has revolutionalized image-level tasks such as classification, but patch-level tasks, such as correspondence, still rely on hand-crafted features, e.g. SIFT. In this paper we use Convolutional Neural Networks (CNNs) to learn discriminant patch representations and in particular train a Siamese network with pairs of (non-)corresponding patches. We deal with the large number of potential pairs with the combination of a stochastic sampling of the training set and an aggressive mining strategy biased towards patches that are hard to classify. By using the L2 distance during both training and testing we develop 128-D descriptors whose euclidean distances reflect patch similarity, and which can be used as a drop-in replacement for any task involving SIFT. We demonstrate consistent performance gains over the state of the art, and generalize well against scaling and rotation, perspective transformation, non-rigid deformation, and illumination changes. Our descriptors are efficient to compute and amenable to modern GPUs, and are publicly available."
                },
                {
                    "title": "Online Batch Selection for Faster Training of Neural Networks",
                    "abstract": "Deep neural networks are commonly trained using stochastic non-convex optimization procedures, which are driven by gradient information estimated on fractions (batches) of the dataset. While it is commonly accepted that batch size is an important parameter for offline tuning, the benefits of online selection of batches remain poorly understood. We investigate online batch selection strategies for two state-of-the-art methods of stochastic gradient-based optimization, AdaDelta and Adam. As the loss function to be minimized for the whole dataset is an aggregation of loss functions of individual datapoints, intuitively, datapoints with the greatest loss should be considered (selected in a batch) more frequently. However, the limitations of this intuition and the proper control of the selection pressure over time are open questions. We propose a simple strategy where all datapoints are ranked w.r.t. their latest known loss value and the probability to be selected decays exponentially as a function of rank. Our experimental results on the MNIST dataset suggest that selecting batches speeds up both AdaDelta and Adam by a factor of about 5."
                },
                {
                    "title": "Prioritized Experience Replay",
                    "abstract": "Experience replay lets online reinforcement learning agents remember and reuse experiences from the past. In prior work, experience transitions were uniformly sampled from a replay memory. However, this approach simply replays transitions at the same frequency that they were originally experienced, regardless of their significance. In this paper we develop a framework for prioritizing experience, so as to replay important transitions more frequently, and therefore learn more efficiently. We use prioritized experience replay in Deep Q-Networks (DQN), a reinforcement learning algorithm that achieved human-level performance across many Atari games. DQN with prioritized experience replay achieves a new state-of-the-art, outperforming DQN with uniform replay on 41 out of 49 games."
                },
                {
                    "title": "On coresets for k-means and k-median clustering",
                    "abstract": "In this paper, we show the existence of small coresets for the problems of computing k-median and k-means clustering for points in low dimension. In other words, we show that given a point set P in Rd, one can compute a weighted set S \u2286 P, of size O(k \u03b5-d log n), such that one can compute the k-median/means clustering on S instead of on P, and get an (1+\u03b5)-approximation. As a result, we improve the fastest known algorithms for (1+\u03b5)-approximate k-means and k-median. Our algorithms have linear running time for a fixed k and \u03b5. In addition, we can maintain the (1+\u03b5)-approximate k-median or k-means clustering of a stream when points are being only inserted, using polylogarithmic space and update time."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can model-based data selection criteria be effectively applied to batches of data to accelerate learning in large-scale pretraining beyond what is achievable through independent example selection?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the inefficiencies in current data curation methods, which rely heavily on manual processes that are difficult to scale. By improving batch selection through model-based criteria, we can enhance the performance of machine learning models across various modalities, leading to more efficient training processes. This advancement could significantly reduce the amount of data required for effective learning, enabling researchers to achieve better results with less computational resources. Furthermore, it opens avenues for practical applications in real-world scenarios where data quality is paramount, such as in healthcare, autonomous systems, and natural language processing.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of determining the optimal composition of data batches, as the quality of a batch is not merely the sum of its individual data points but also depends on their interactions. Naive approaches that treat data points independently may overlook critical relationships that enhance learning, such as the presence of hard negatives. Additionally, efficiently scoring large super-batches to identify the most learnable sub-batches requires advanced computational techniques and can be resource-intensive. Overcoming these technical and practical obstacles is essential to realize the potential of model-based data selection.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on individual data point selection without considering the batch composition's impact on learning. Existing methods have limitations in scalability and efficiency, often leading to suboptimal training outcomes. Barriers such as the computational cost of evaluating large datasets and the lack of effective algorithms for joint example selection have hindered progress. Our approach differs by introducing the JEST algorithm, which leverages model-based scoring to select relevant sub-batches, thus improving upon prior work by addressing both the efficiency of selection and the quality of the training process.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the JEST algorithm, which utilizes model-based scores to select sub-batches from larger super-batches of data. We will employ a pretrained reference model to score batches based on their learnability, taking into account the online model loss. The expected outcomes include significant improvements in training efficiency, with JEST achieving up to"
            }
        },
        "author_data": {
            "18739d60-f319-4714-85d4-321ddf1d19c6": {
                "pk": "18739d60-f319-4714-85d4-321ddf1d19c6",
                "name": "Talfan Evans",
                "collaborators": [
                    "Joseph Ortiz",
                    "Andrew J. Davison",
                    "Edgar Sucar",
                    "Shreya Pathak",
                    "Hamza Merzic",
                    "Jonathan Schwarz",
                    "Ryutaro Tanno",
                    "Olivier J. Henaff"
                ],
                "domain": [
                    "Probabilistic Inference",
                    "Graph-based SLAM",
                    "Active Learning",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "A visual introduction to Gaussian Belief Propagation",
                        "abstract": "In this article, we present a visual introduction to Gaussian Belief Propagation (GBP), an approximate probabilistic inference algorithm that operates by passing messages between the nodes of arbitrarily structured factor graphs. A special case of loopy belief propagation, GBP updates rely only on local information and will converge independently of the message schedule. Our key argument is that, given recent trends in computing hardware, GBP has the right computational properties to act as a scalable distributed probabilistic inference framework for future machine learning systems."
                    },
                    {
                        "title": "Incremental Abstraction in Distributed Probabilistic SLAM Graphs",
                        "abstract": "Scene graphs represent the key components of a scene in a compact and semantically rich way, but are difficult to build during incremental SLAM operation because of the challenges of robustly identifying abstract scene elements and optimising continually changing, complex graphs. We present a distributed, graph-based SLAM framework for incrementally building scene graphs based on two novel components. First, we propose an incremental abstraction framework in which a neural network proposes abstract scene elements that are incorporated into the factor graph of a feature-based monocular SLAM system. Scene elements are confirmed or rejected through optimisation and incrementally replace the points yielding a more dense, semantic and compact representation. Second, enabled by our novel routing procedure, we use Gaussian Belief Propagation (GBP) for distributed inference on a graph processor. The time per iteration of GBP is structure-agnostic and we demonstrate the speed advantages over direct methods for inference of heterogeneous factor graphs. We run our system on real indoor datasets using planar abstractions and recover the major planes with significant compression."
                    },
                    {
                        "title": "Bad Students Make Great Teachers: Active Learning Accelerates Large-Scale Visual Understanding",
                        "abstract": "Power-law scaling indicates that large-scale training with uniform sampling is prohibitively slow. Active learning methods aim to increase data efficiency by prioritizing learning on the most relevant examples. Despite their appeal, these methods have yet to be widely adopted since no one algorithm has been shown to a) generalize across models and tasks b) scale to large datasets and c) yield overall FLOP savings when accounting for the overhead of data selection. In this work we propose a method which satisfies these three properties, leveraging small, cheap proxy models to estimate \"learnability\" scores for datapoints, which are used to prioritize data for the training of much larger models. As a result, our models require 46% and 51% fewer training updates and up to 25% less total computation to reach the same performance as uniformly trained visual classifiers on JFT and multimodal models on ALIGN. Finally, we find our data-prioritization scheme to be complementary with recent data-curation and learning objectives, yielding a new state-of-the-art in several multimodal transfer tasks."
                    }
                ]
            },
            "88c02457-33d9-4ad6-8c18-4a6585863248": {
                "pk": "88c02457-33d9-4ad6-8c18-4a6585863248",
                "name": "Nikhil Parthasarathy",
                "collaborators": [
                    "Olivier J. H\u00e9naff",
                    "Eero P. Simoncelli",
                    "Reuben Feinman",
                    "S. M. Ali Eslami",
                    "Jo\u00e3o Carreira",
                    "Ivana Bala\u017eevi\u0107",
                    "David Steiner",
                    "Relja Arandjelovi\u0107"
                ],
                "domain": [
                    "Normalizing Flows",
                    "Self-Supervised Learning",
                    "Video Representation",
                    "In-Context Learning"
                ],
                "publications": [
                    {
                        "title": "A Linear Systems Theory of Normalizing Flows",
                        "abstract": "Normalizing Flows are a promising new class of algorithms for unsupervised learning based on maximum likelihood optimization with change of variables. They offer to learn a factorized component representation for complex nonlinear data and, simultaneously, yield a density function that can evaluate likelihoods and generate samples. Despite these diverse offerings, applications of Normalizing Flows have focused primarily on sampling and likelihoods, with little emphasis placed on feature representation. A lack of theoretical foundation has left many open questions about how to interpret and apply the learned components of the model. We provide a new theoretical perspective of Normalizing Flows using the lens of linear systems theory, showing that optimal flows learn to represent the local covariance at each region of input space. Using this insight, we develop a new algorithm to extract interpretable component representations from the learned model, where components correspond to Cartesian dimensions and are scaled according to their manifold significance. In addition, we highlight a stability concern for the learning algorithm that was previously unaddressed, providing a theoretically-grounded solution to mediate the problem. Experiments with toy manifold learning datasets, as well as the MNIST image dataset, provide convincing support for our theory and tools."
                    },
                    {
                        "title": "Self-Supervised Learning of a Biologically-Inspired Visual Texture Model",
                        "abstract": "We develop a model for representing visual texture in a low-dimensional feature space, along with a novel self-supervised learning objective that is used to train it on an unlabeled database of texture images. Inspired by the architecture of primate visual cortex, the model uses a first stage of oriented linear filters (corresponding to cortical area V1), consisting of both rectified units (simple cells) and pooled phase-invariant units (complex cells). These responses are processed by a second stage (analogous to cortical area V2) consisting of convolutional filters followed by half-wave rectification and pooling to generate V2 'complex cell' responses. The second stage filters are trained on a set of unlabeled homogeneous texture images, using a novel contrastive objective that maximizes the distance between the distribution of V2 responses to individual images and the distribution of responses across all images. When evaluated on texture classification, the trained model achieves substantially greater data-efficiency than a variety of deep hierarchical model architectures. Moreover, we show that the learned model exhibits stronger representational similarity to texture responses of neural populations recorded in primate V2 than pre-trained deep CNNs."
                    },
                    {
                        "title": "Layerwise complexity-matched learning yields an improved model of cortical area V2",
                        "abstract": "Human ability to recognize complex visual patterns arises through transformations performed by successive areas in the ventral visual cortex. Deep neural networks trained end-to-end for object recognition approach human capabilities, and offer the best descriptions to date of neural responses in the late stages of the hierarchy. But these networks provide a poor account of the early stages, compared to traditional hand-engineered models, or models optimized for coding efficiency or prediction. Moreover, the gradient backpropagation used in end-to-end learning is generally considered to be biologically implausible. Here, we overcome both of these limitations by developing a bottom-up self-supervised training methodology that operates independently on successive layers. Specifically, we maximize feature similarity between pairs of locally-deformed natural image patches, while decorrelating features across patches sampled from other images. Crucially, the deformation amplitudes are adjusted proportionally to receptive field sizes in each layer, thus matching the task complexity to the capacity at each stage of processing. In comparison with architecture-matched versions of previous models, we demonstrate that our layerwise complexity-matched learning (LCL) formulation produces a two-stage model (LCL-V2) that is better aligned with selectivity properties and neural activity in primate area V2. We demonstrate that the complexity-matched learning paradigm is responsible for much of the emergence of the improved biological alignment. Finally, when the two-stage model is used as a fixed front-end for a deep network trained to perform object recognition, the resultant model (LCL-V2Net) is significantly better than standard end-to-end self-supervised, supervised, and adversarially-trained models in terms of generalization to out-of-distribution tasks and alignment with human behavior."
                    },
                    {
                        "title": "Self-supervised video pretraining yields human-aligned visual representations",
                        "abstract": "Humans learn powerful representations of objects and scenes by observing how they evolve over time. Yet, outside of specific tasks that require explicit temporal understanding, static image pretraining remains the dominant paradigm for learning visual foundation models. We question this mismatch, and ask whether video pretraining can yield visual representations that bear the hallmarks of human perception: generalisation across tasks, robustness to perturbations, and consistency with human judgements. To that end we propose a novel procedure for curating videos, and develop a contrastive framework which learns from the complex transformations therein. This simple paradigm for distilling knowledge from videos, called VITO, yields general representations that far outperform prior video pretraining methods on image understanding tasks, and image pretraining methods on video understanding tasks. Moreover, VITO representations are significantly more robust to natural and synthetic deformations than image-, video-, and adversarially-trained ones. Finally, VITO's predictions are strongly aligned with human judgements, surpassing models that were specifically trained for that purpose. Together, these results suggest that video pretraining could be a simple way of learning unified, robust, and human-aligned representations of the visual world."
                    },
                    {
                        "title": "Towards In-context Scene Understanding",
                        "abstract": "In-context learning$\\unicode{x2013}$the ability to configure a model's behavior with different prompts$\\unicode{x2013}$has revolutionized the field of natural language processing, alleviating the need for task-specific models and paving the way for generalist models capable of assisting with any query. Computer vision, in contrast, has largely stayed in the former regime: specialized decoders and finetuning protocols are generally required to perform dense tasks such as semantic segmentation and depth estimation. In this work we explore a simple mechanism for in-context learning of such scene understanding tasks: nearest neighbor retrieval from a prompt of annotated features. We propose a new pretraining protocol$\\unicode{x2013}$leveraging attention within and across images$\\unicode{x2013}$which yields representations particularly useful in this regime. The resulting Hummingbird model, suitably prompted, performs various scene understanding tasks without modification while approaching the performance of specialists that have been finetuned for each task. Moreover, Hummingbird can be configured to perform new tasks much more efficiently than finetuned models, raising the possibility of scene understanding in the interactive assistant regime."
                    }
                ]
            },
            "ab9c52fd-5d98-4f11-9c14-3471c15e13ca": {
                "pk": "ab9c52fd-5d98-4f11-9c14-3471c15e13ca",
                "name": "Hamza Merzic",
                "collaborators": [
                    "Karol Gregor",
                    "Frederic Besse",
                    "Miroslav Bogdanovic",
                    "Daniel Kappler",
                    "Ludovic Righetti",
                    "Jeannette Bohg",
                    "Felix Hill",
                    "Olivier Tieleman",
                    "Tamara von Glehn",
                    "Nathaniel Wong"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Robotics",
                    "Deep Learning",
                    "Active Learning"
                ],
                "publications": [
                    {
                        "title": "Leveraging Contact Forces for Learning to Grasp",
                        "abstract": "Grasping objects under uncertainty remains an open problem in robotics research. This uncertainty is often due to noisy or partial observations of the object pose or shape. To enable a robot to react appropriately to unforeseen effects, it is crucial that it continuously takes sensor feedback into account. While visual feedback is important for inferring a grasp pose and reaching for an object, contact feedback offers valuable information during manipulation and grasp acquisition. In this paper, we use model-free deep reinforcement learning to synthesize control policies that exploit contact sensing to generate robust grasping under uncertainty. We demonstrate our approach on a multi-fingered hand that exhibits more complex finger coordination than the commonly used two-fingered grippers. We conduct extensive experiments in order to assess the performance of the learned policies, with and without contact sensing. While it is possible to learn grasping policies without contact sensing, our results suggest that contact feedback allows for a significant improvement of grasping robustness under object pose uncertainty and for objects with a complex shape."
                    },
                    {
                        "title": "Grounded Language Learning Fast and Slow",
                        "abstract": "Recent work has shown that large text-based neural language models, trained with conventional supervised learning objectives, acquire a surprising propensity for few- and one-shot learning. Here, we show that an embodied agent situated in a simulated 3D world, and endowed with a novel dual-coding external memory, can exhibit similar one-shot word learning when trained with conventional reinforcement learning algorithms. After a single introduction to a novel object via continuous visual perception and a language prompt (\"This is a dax\"), the agent can re-identify the object and manipulate it as instructed (\"Put the dax on the bed\"). In doing so, it seamlessly integrates short-term, within-episode knowledge of the appropriate referent for the word \"dax\" with long-term lexical and motor knowledge acquired across episodes (i.e. \"bed\" and \"putting\"). We find that, under certain training conditions and with a particular memory writing mechanism, the agent's one-shot word-object binding generalizes to novel exemplars within the same ShapeNet category, and is effective in settings with unfamiliar numbers of objects. We further show how dual-coding memory can be exploited as a signal for intrinsic motivation, stimulating the agent to seek names for objects that may be useful for later executing instructions. Together, the results demonstrate that deep neural networks can exploit meta-learning, episodic memory and an explicitly multi-modal environment to account for 'fast-mapping', a fundamental pillar of human cognitive development and a potentially transformative capacity for agents that interact with human users."
                    },
                    {
                        "title": "Bad Students Make Great Teachers: Active Learning Accelerates Large-Scale Visual Understanding",
                        "abstract": "Power-law scaling indicates that large-scale training with uniform sampling is prohibitively slow. Active learning methods aim to increase data efficiency by prioritizing learning on the most relevant examples. Despite their appeal, these methods have yet to be widely adopted since no one algorithm has been shown to a) generalize across models and tasks b) scale to large datasets and c) yield overall FLOP savings when accounting for the overhead of data selection. In this work we propose a method which satisfies these three properties, leveraging small, cheap proxy models to estimate \"learnability\" scores for datapoints, which are used to prioritize data for the training of much larger models. As a result, our models require 46% and 51% fewer training updates and up to 25% less total computation to reach the same performance as uniformly trained visual classifiers on JFT and multimodal models on ALIGN. Finally, we find our data-prioritization scheme to be complementary with recent data-curation and learning objectives, yielding a new state-of-the-art in several multimodal transfer tasks."
                    },
                    {
                        "title": "Shaping Belief States with Generative Environment Models for RL",
                        "abstract": "When agents interact with a complex environment, they must form and maintain beliefs about the relevant aspects of that environment. We propose a way to efficiently train expressive generative models in complex environments. We show that a predictive algorithm with an expressive generative model can form stable belief-states in visually rich and dynamic 3D environments. More precisely, we show that the learned representation captures the layout of the environment as well as the position and orientation of the agent. Our experiments show that the model substantially improves data-efficiency on a number of reinforcement learning (RL) tasks compared with strong model-free baseline agents. We find that predicting multiple steps into the future (overshooting), in combination with an expressive generative model, is critical for stable representations to emerge. In practice, using expressive generative models in RL is computationally expensive and we propose a scheme to reduce this computational burden, allowing us to build agents that are competitive with model-free baselines."
                    },
                    {
                        "title": "Causally Correct Partial Models for Reinforcement Learning",
                        "abstract": "In reinforcement learning, we can learn a model of future observations and rewards, and use it to plan the agent's next actions. However, jointly modeling future observations can be computationally expensive or even intractable if the observations are high-dimensional (e.g. images). For this reason, previous works have considered partial models, which model only part of the observation. In this paper, we show that partial models can be causally incorrect: they are confounded by the observations they don't model, and can therefore lead to incorrect planning. To address this, we introduce a general family of partial models that are provably causally correct, yet remain fast because they do not need to fully model future observations."
                    }
                ]
            },
            "6f9221f3-3016-4424-b934-47b9341c9419": {
                "pk": "6f9221f3-3016-4424-b934-47b9341c9419",
                "name": "Olivier J. Henaff",
                "collaborators": [
                    "Yonglong Tian",
                    "Aaron van den Oord",
                    "Talfan Evans",
                    "Shreya Pathak",
                    "Hamza Merzic",
                    "Jonathan Schwarz",
                    "Ryutaro Tanno"
                ],
                "domain": [
                    "Self-supervised Learning",
                    "Active Learning",
                    "Computer Vision",
                    "Contrastive Learning"
                ],
                "publications": [
                    {
                        "title": "Divide and Contrast: Self-supervised Learning from Uncurated Data",
                        "abstract": "Self-supervised learning holds promise in leveraging large amounts of unlabeled data, however much of its progress has thus far been limited to highly curated pre-training data such as ImageNet. We explore the effects of contrastive learning from larger, less-curated image datasets such as YFCC, and find there is indeed a large difference in the resulting representation quality. We hypothesize that this curation gap is due to a shift in the distribution of image classes -- which is more diverse and heavy-tailed -- resulting in less relevant negative samples to learn from. We test this hypothesis with a new approach, Divide and Contrast (DnC), which alternates between contrastive learning and clustering-based hard negative mining. When pretrained on less curated datasets, DnC greatly improves the performance of self-supervised learning on downstream tasks, while remaining competitive with the current state-of-the-art on curated datasets."
                    },
                    {
                        "title": "Bad Students Make Great Teachers: Active Learning Accelerates Large-Scale Visual Understanding",
                        "abstract": "Power-law scaling indicates that large-scale training with uniform sampling is prohibitively slow. Active learning methods aim to increase data efficiency by prioritizing learning on the most relevant examples. Despite their appeal, these methods have yet to be widely adopted since no one algorithm has been shown to a) generalize across models and tasks b) scale to large datasets and c) yield overall FLOP savings when accounting for the overhead of data selection. In this work we propose a method which satisfies these three properties, leveraging small, cheap proxy models to estimate \"learnability\" scores for datapoints, which are used to prioritize data for the training of much larger models. As a result, our models require 46% and 51% fewer training updates and up to 25% less total computation to reach the same performance as uniformly trained visual classifiers on JFT and multimodal models on ALIGN. Finally, we find our data-prioritization scheme to be complementary with recent data-curation and learning objectives, yielding a new state-of-the-art in several multimodal transfer tasks."
                    }
                ]
            }
        }
    },
    "2406.10580": {
        "paper_data": {
            "title": "IMDL-BenCo: A Comprehensive Benchmark and Codebase for Image Manipulation Detection & Localization",
            "url": "http://arxiv.org/abs/2406.10580v1",
            "arxiv_id": "2406.10580",
            "authors": [
                "Xiaochen Ma",
                "Xuekang Zhu",
                "Lei Su",
                "Bo Du",
                "Zhuohang Jiang",
                "Bingkui Tong",
                "Zeyu Lei",
                "Xinyu Yang",
                "Chi-Man Pun",
                "Jiancheng Lv",
                "Jizhe Zhou"
            ],
            "abstract": "A comprehensive benchmark is yet to be established in the Image Manipulation Detection \\& Localization (IMDL) field. The absence of such a benchmark leads to insufficient and misleading model evaluations, severely undermining the development of this field. However, the scarcity of open-sourced baseline models and inconsistent training and evaluation protocols make conducting rigorous experiments and faithful comparisons among IMDL models challenging. To address these challenges, we introduce IMDL-BenCo, the first comprehensive IMDL benchmark and modular codebase. IMDL-BenCo:~\\textbf{i)} decomposes the IMDL framework into standardized, reusable components and revises the model construction pipeline, improving coding efficiency and customization flexibility;~\\textbf{ii)} fully implements or incorporates training code for state-of-the-art models to establish a comprehensive IMDL benchmark; and~\\textbf{iii)} conducts deep analysis based on the established benchmark and codebase, offering new insights into IMDL model architecture, dataset characteristics, and evaluation standards. Specifically, IMDL-BenCo includes common processing algorithms, 8 state-of-the-art IMDL models (1 of which are reproduced from scratch), 2 sets of standard training and evaluation protocols, 15 GPU-accelerated evaluation metrics, and 3 kinds of robustness evaluation. This benchmark and codebase represent a significant leap forward in calibrating the current progress in the IMDL field and inspiring future breakthroughs. Code is available at: https://github.com/scu-zjz/IMDLBenCo",
            "introduction": "   1 Introduction  \u201cExperimentation is the ultimate arbiter of scientific truth.\" - Richard Feynman. The empowered image manipulation or generation models drive the Image Manipulation Detection & Localization (IMDL) task to the forefront of information forensics and security\u00a0[23, 35]. While the task is occasionally referred to as \"forgery detection\"\u00a0[15, 13] or \"tamper detection\"[31, 30] in literature, the consensus now favors the term IMDL\u00a0[13] as the most apt descriptor for this study area. The scope of \"manipulation\" within IMDL bounds to partial image alterations that yield semantic discrepancies from the original content\u00a0[37]. It does not pertain to purely generated images (e.g., images generated from pure text) or the application of image processing techniques that introduce noise or other non-semantical changes without altering the underlying meaning of the image\u00a0[5]. The terms \"detection and localization\" denote an IMDL model\u2019s dual responsibility: to conduct both image-level and pixel-level assessments. This involves a binary classification at the image level, discerning whether an input image is manipulated or authentic, and a segmentation task at the pixel level, depicting the exact manipulated areas through a mask. In short, an IMDL model shall identify semantically significant image alterations and deliver a twofold outcome: a class label and a manipulation mask.   Despite the rapid success of deep neural networks in the IMDL fields\u00a0[10, 43, 11], existing models suffer from inconsistent training and evaluation protocols, supported by Tables in Appendix B.1. These inconsistencies result in incompatible and unfair comparisons, yielding insufficient and misleading experimental outcomes. Hence, establishing a unified and comprehensive benchmark is the foremost concern in the IMDL field. However, establishing this benchmark is far more than straightforward protocol unification or simply model re-training. First, the training code for most state-of-the-art (SoTA) works is not publicly available, and the source code for some SoTA works is totally unreleased\u00a0[26]. Second, IMDL models commonly incorporate diverse low-level features\u00a0[41, 4, 13] and complex loss functions, requiring highly customized model architecture and decoupled pipeline design for efficient reproduction. Existing frameworks, like OpenMMLab111https://openmmlab.com/ and Detectron2222https://github.com/facebookresearch/detectron2, heavily rely on the registry mechanism and tightly coupled pipelines. This conflict leads to severe efficiency issues while reproducing IMDL models under existing frameworks and results in monolithic model architecture with extremely high coding load and low scalability. Consequently, a comprehensive IMDL benchmark is yet to be built.   To address this issue, we introduce IMDL-BenCo, the first comprehensive IMDL benchmark and modular codebase. IMDL-BenCo:\u00a0i) features a modular codebase with four components: data loader, model zoo, training script, and evaluator; the model zoo contains customizable model architecture includes SoTA models and backbone models. The loss design is also isolated within the model zoo, while other components are standardized by the interface and are highly reusable; this approach mitigates conflicts between model customization and coding efficiency; \u00a0ii) fully implements or incorporates training code for 8 SoTA IMDL models and establishes a comprehensive benchmark with 2 sets of standard training and evaluation protocols, 15 GPU-accelerated evaluation metrics, and 3 kinds of robustness evaluation, and\u00a0iii) conducts in-depth analysis based on the established benchmark and codebase, offering new insights into IMDL model architecture, dataset characteristics, and evaluation standards. This benchmark and codebase represent",
            "references": [
                {
                    "title": "Pre-training-free Image Manipulation Localization through Non-Mutually Exclusive Contrastive Learning",
                    "abstract": "Deep Image Manipulation Localization (IML) models suffer from training data insufficiency and thus heavily rely on pre-training. We argue that contrastive learning is more suitable to tackle the data insufficiency problem for IML. Crafting mutually exclusive positives and negatives is the prerequisite for contrastive learning. However, when adopting contrastive learning in IML, we encounter three categories of image patches: tampered, authentic, and contour patches. Tampered and authentic patches are naturally mutually exclusive, but contour patches containing both tampered and authentic pixels are non-mutually exclusive to them. Simply abnegating these contour patches results in a drastic performance loss since contour patches are decisive to the learning outcomes. Hence, we propose the Nonmutually exclusive Contrastive Learning (NCL) framework to rescue conventional contrastive learning from the above dilemma. In NCL, to cope with the non-mutually exclusivity, we first establish a pivot structure with dual branches to constantly switch the role of contour patches between positives and negatives while training. Then, we devise a pivot-consistent loss to avoid spatial corruption caused by the role-switching process. In this manner, NCL both inherits the self-supervised merits to address the data insufficiency and retains a high manipulation localization accuracy. Extensive experiments verify that our NCL achieves state-of-the-art performance on all five benchmarks without any pre-training and is more robust on unseen real-life samples. https://github.com/Knightzjz/NCL-IML."
                },
                {
                    "title": "Hierarchical Fine-Grained Image Forgery Detection and Localization",
                    "abstract": "Differences in forgery attributes of images generated in CNN-synthesized and image-editing domains are large, and such differences make a unified image forgery detection and localization (IFDL) challenging. To this end, we present a hierarchical fine-grained formulation for IFDL representation learning. Specifically, we first represent forgery attributes of a manipulated image with multiple labels at different levels. Then we perform fine-grained classification at these levels using the hierarchical dependency between them. As a result, the algorithm is encouraged to learn both comprehensive features and inherent hierarchical nature of different forgery attributes, thereby improving the IFDL representation. Our proposed IFDL framework contains three components: multi-branch feature extractor, localization and classification modules. Each branch of the feature extractor learns to classify forgery attributes at one level, while localization and classification modules segment the pixel-level forgery region and detect image-level forgery, respectively. Lastly, we construct a hierarchical fine-grained dataset to facilitate our study. We demonstrate the effectiveness of our method on 7 different benchmarks, for both tasks of IFDL and forgery attribute classification. Our source code and dataset can be found: github.com/CHELSEA234/HiFi-IFDL."
                },
                {
                    "title": "TruFor: Leveraging All-Round Clues for Trustworthy Image Forgery Detection and Localization",
                    "abstract": "In this paper we present TruFor, a forensic framework that can be applied to a large variety of image manipulation methods, from classic cheapfakes to more recent manipulations based on deep learning. We rely on the extraction of both high-level and low-level traces through a transformer-based fusion architecture that combines the RGB image and a learned noise-sensitive fingerprint. The latter learns to embed the artifacts related to the camera internal and external processing by training only on real data in a self-supervised manner. Forgeries are detected as deviations from the expected regular pattern that characterizes each pristine image. Looking for anomalies makes the approach able to robustly detect a variety of local manipulations, ensuring generalization. In addition to a pixel-level localization map and a whole-image integrity score, our approach outputs a reliability map that highlights areas where localization predictions may be error-prone. This is particularly important in forensic applications in order to reduce false alarms and allow for a large scale analysis. Extensive experiments on several datasets show that our method is able to reliably detect and localize both cheapfakes and deepfakes manipulations outperforming state-of-the-art works. Code is publicly available at https://grip-unina.github.io/TruFor/"
                },
                {
                    "title": "On The Detection of Synthetic Images Generated by Diffusion Models",
                    "abstract": "Over the past decade, there has been tremendous progress in creating synthetic media, mainly thanks to the development of powerful methods based on generative adversarial networks (GAN). Very recently, methods based on diffusion models (DM) have been gaining the spotlight. In addition to providing an impressive level of photorealism, they enable the creation of text-based visual content, opening up new and exciting opportunities in many different application fields, from arts to video games. On the other hand, this property is an additional asset in the hands of malicious users, who can generate and distribute fake media perfectly adapted to their attacks, posing new challenges to the media forensic community. With this work, we seek to understand how difficult it is to distinguish synthetic images generated by diffusion models from pristine ones and whether current state-of-the-art detectors are suitable for the task. To this end, first we expose the forensics traces left by diffusion models, then study how current detectors, developed for GAN-generated images, perform on these new synthetic images, especially in challenging social-network scenarios involving image compression and resizing. Datasets and code are available at https:github.com/grip-unina/DMimageDetection."
                },
                {
                    "title": "ObjectFormer for Image Manipulation Detection and Localization",
                    "abstract": "Recent advances in image editing techniques have posed serious challenges to the trustworthiness of multimedia data, which drives the research of image tampering detection. In this paper, we propose ObjectFormer to detect and localize image manipulations. To capture subtle manipulation traces that are no longer visible in the RGB domain, we extract high-frequency features of the images and combine them with RGB features as multimodal patch embeddings. Additionally, we use a set of learnable object prototypes as mid-level representations to model the object-level consistencies among different regions, which are further used to refine patch embeddings to capture the patch-level consistencies. We conduct extensive experiments on various datasets and the results verify the effectiveness of the proposed method, outperforming state-of-the-art tampering detection and localization methods."
                },
                {
                    "title": "MVSS-Net: Multi-View Multi-Scale Supervised Networks for Image Manipulation Detection",
                    "abstract": "As manipulating images by copy-move, splicing and/or inpainting may lead to misinterpretation of the visual content, detecting these sorts of manipulations is crucial for media forensics. Given the variety of possible attacks on the content, devising a generic method is nontrivial. Current deep learning based methods are promising when training and test data are well aligned, but perform poorly on independent tests. Moreover, due to the absence of authentic test images, their image-level detection specificity is in doubt. The key question is how to design and train a deep neural network capable of learning generalizable features sensitive to manipulations in novel data, whilst specific to prevent false alarms on the authentic. We propose multi-view feature learning to jointly exploit tampering boundary artifacts and the noise view of the input image. As both clues are meant to be semantic-agnostic, the learned features are thus generalizable. For effectively learning from authentic images, we train with multi-scale (pixel / edge / image) supervision. We term the new network MVSS-Net and its enhanced version MVSS-Net++. Experiments are conducted in both within-dataset and cross-dataset scenarios, showing that MVSS-Net++ performs the best, and exhibits better robustness against JPEG compression, Gaussian blur and screenshot based image re-capturing."
                },
                {
                    "title": "TransForensics: Image Forgery Localization with Dense Self-Attention",
                    "abstract": "Nowadays advanced image editing tools and technical skills produce tampered images more realistically, which can easily evade image forensic systems and make authenticity verification of images more difficult. To tackle this challenging problem, we introduce TransForensics, a novel image forgery localization method inspired by Transformers. The two major components in our framework are dense self-attention encoders and dense correction modules. The former is to model global context and all pairwise inter-actions between local patches at different scales, while the latter is used for improving the transparency of the hidden layers and correcting the outputs from different branches. Compared to previous traditional and deep learning methods, TransForensics not only can capture discriminative representations and obtain high-quality mask predictions but is also not limited by tampering types and patch sequence orders. By conducting experiments on main bench-marks, we show that TransForensics outperforms the state-of-the-art methods by a large margin."
                },
                {
                    "title": "A deep-learning-based image forgery detection framework for controlling the spread of misinformation",
                    "abstract": "PurposeWeb users rely heavily on online content make decisions without assessing the veracity of the content. The online content comprising text, image, video or audio may be tampered with to influence public opinion. Since the consumers of online information (misinformation) tend to trust the content when the image(s) supplement the text, image manipulation software is increasingly being used to forge the images. To address the crucial problem of image manipulation, this study focusses on developing a deep-learning-based image forgery detection framework.Design/methodology/approachThe proposed deep-learning-based framework aims to detect images forged using copy-move and splicing techniques. The image transformation technique aids the identification of relevant features for the network to train effectively. After that, the pre-trained customized convolutional neural network is used to train on the public benchmark datasets, and the performance is evaluated on the test dataset using various parameters.FindingsThe comparative analysis of image transformation techniques and experiments conducted on benchmark datasets from a variety of socio-cultural domains establishes the effectiveness and viability of the proposed framework. These findings affirm the potential applicability of proposed framework in real-time image forgery detection.Research limitations/implicationsThis study bears implications for several important aspects of research on image forgery detection. First this research adds to recent discussion on feature extraction and learning for image forgery detection. While prior research on image forgery detection, hand-crafted the features, the proposed solution contributes to stream of literature that automatically learns the features and classify the images. Second, this research contributes to ongoing effort in curtailing the spread of misinformation using images. The extant literature on spread of misinformation has prominently focussed on textual data shared over social media platforms. The study addresses the call for greater emphasis on the development of robust image transformation techniques.Practical implicationsThis study carries important practical implications for various domains such as forensic sciences, media and journalism where image data is increasingly being used to make inferences. The integration of image forgery detection tools can be helpful in determining the credibility of the article or post before it is shared over the Internet. The content shared over the Internet by the users has become an important component of news reporting. The framework proposed in this paper can be further extended and trained on more annotated real-world data so as to function as a tool for fact-checkers.Social implicationsIn the current scenario wherein most of the image forgery detection studies attempt to assess whether the image is real or forged in an offline mode, it is crucial to identify any trending or potential forged image as early as possible. By learning from historical data, the proposed framework can aid in early prediction of forged images to detect the newly emerging forged images even before they occur. In summary, the proposed framework has a potential to mitigate physical spreading and psychological impact of forged images on social media.Originality/valueThis study focusses on copy-move and splicing techniques while integrating transfer learning concepts to classify forged images with high accuracy. The synergistic use of hitherto little explored image transformation techniques and customized convolutional neural network helps design a robust image forgery detection framework. Experiments and findings establish that the proposed framework accurately classifies forged images, thus mitigating the negative socio-cultural spread of misinformation."
                },
                {
                    "title": "SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers",
                    "abstract": "We present SegFormer, a simple, efficient yet powerful semantic segmentation framework which unifies Transformers with lightweight multilayer perception (MLP) decoders. SegFormer has two appealing features: 1) SegFormer comprises a novel hierarchically structured Transformer encoder which outputs multiscale features. It does not need positional encoding, thereby avoiding the interpolation of positional codes which leads to decreased performance when the testing resolution differs from training. 2) SegFormer avoids complex decoders. The proposed MLP decoder aggregates information from different layers, and thus combining both local attention and global attention to render powerful representations. We show that this simple and lightweight design is the key to efficient segmentation on Transformers. We scale our approach up to obtain a series of models from SegFormer-B0 to SegFormer-B5, reaching significantly better performance and efficiency than previous counterparts. For example, SegFormer-B4 achieves 50.3% mIoU on ADE20K with 64M parameters, being 5x smaller and 2.2% better than the previous best method. Our best model, SegFormer-B5, achieves 84.0% mIoU on Cityscapes validation set and shows excellent zero-shot robustness on Cityscapes-C. Code will be released at: github.com/NVlabs/SegFormer."
                },
                {
                    "title": "Image Manipulation Detection by Multi-View Multi-Scale Supervision",
                    "abstract": "The key challenge of image manipulation detection is how to learn generalizable features that are sensitive to manipulations in novel data, whilst specific to prevent false alarms on authentic images. Current research emphasizes the sensitivity, with the specificity overlooked. In this paper we address both aspects by multi-view feature learning and multi-scale supervision. By exploiting noise distribution and boundary artifact surrounding tampered regions, the former aims to learn semantic-agnostic and thus more generalizable features. The latter allows us to learn from authentic images which are nontrivial to be taken into account by current semantic segmentation network based methods. Our thoughts are realized by a new network which we term MVSS-Net. Extensive experiments on five benchmark sets justify the viability of MVSS-Net for both pixel-level and image-level manipulation detection."
                },
                {
                    "title": "Extended IMD2020: a large-scale annotated dataset tailored for detecting manipulated images",
                    "abstract": "Image forensic datasets need to accommodate a complex diversity of systematic noise and intrinsic image artefacts to prevent any overfitting of learning methods to a small set of camera types or manipulation techniques. Such artefacts are created during the image acquisition as well as the manipulating process itself (e.g., noise due to sensors, interpolation artefacts, etc.). Here, the authors introduce three datasets. First, we identified the majority of camera models on the market. Then, we collected a dataset of 35,000 real images captured by these cameras. We also created the same number of digitally manipulated images. Additionally, we also collected a dataset of 2,000 \u2018real \u2010 life\u2019 (uncontrolled) manipulated images. They are made by unknown people and downloaded from the Internet. The real versions of these images are also provided. We also manually created binary masks localising the exact manipulated areas of these images. Moreover, we captured a set of 2,759 real images formed by 32 unique cameras (19 different camera models) in a controlled way by ourselves. Here, the processing history of all images is guaranteed. This set includes categorised images of uniform areas as well as natural images that can be used effectively for analysis of the sensor noise."
                },
                {
                    "title": "PSCC-Net: Progressive Spatio-Channel Correlation Network for Image Manipulation Detection and Localization",
                    "abstract": "To defend against manipulation of image content, such as splicing, copy-move, and removal, we develop a Progressive Spatio-Channel Correlation Network (PSCC-Net) to detect and localize image manipulations. PSCC-Net processes the image in a two-path procedure: a top-down path that extracts local and global features and a bottom-up path that detects whether the input image is manipulated, and estimates its manipulation masks at multiple scales, where each mask is conditioned on the previous one. Different from the conventional encoder-decoder and no-pooling structures, PSCC-Net leverages features at different scales with dense cross-connections to produce manipulation masks in a coarse-to-fine fashion. Moreover, a Spatio-Channel Correlation Module (SCCM) captures both spatial and channel-wise correlations in the bottom-up path, which endows features with holistic cues, enabling the network to cope with a wide range of manipulation attacks. Thanks to the light-weight backbone and progressive mechanism, PSCC-Net can process $1,080\\text{P}$ images at 50+FPS. Extensive experiments demonstrate the superiority of PSCC-Net over the state-of-the-art methods on both detection and localization. Codes and models are available at https://github.com/proteus1991/PSCC-Net."
                },
                {
                    "title": "Personal Privacy Protection via Irrelevant Faces Tracking and Pixelation in Video Live Streaming",
                    "abstract": "To date, the privacy-protection intended pixelation tasks are still labor-intensive and yet to be studied. With the prevailing of video live streaming, establishing an online face pixelation mechanism during streaming is an urgency. In this paper, we develop a new method called Face Pixelation in Video Live Streaming (FPVLS) to generate automatic personal privacy filtering during unconstrained streaming activities. Simply applying multi-face trackers will encounter problems in target drifting, computing efficiency, and over-pixelation. Therefore, for fast and accurate pixelation of irrelevant people\u2019s faces, FPVLS is organized in a frame-to-video structure of two core stages. On individual frames, FPVLS utilizes image-based face detection and embedding networks to yield face vectors. In the raw trajectories generation stage, the proposed Positioned Incremental Affinity Propagation (PIAP) clustering algorithm leverages face vectors and positioned information to quickly associate the same person\u2019s faces across frames. Such frame-wise accumulated raw trajectories are likely to be intermittent and unreliable on video level. Hence, we further introduce the trajectory refinement stage that merges a proposal network with the two-sample test based on the Empirical Likelihood Ratio (ELR) statistic to refine the raw trajectories. A Gaussian filter is laid on the refined trajectories for final pixelation. On the video live streaming dataset we collected, FPVLS obtains satisfying accuracy, real-time efficiency, and contains the over-pixelation problems."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Media Forensics and DeepFakes: An Overview",
                    "abstract": "With the rapid progress in recent years, techniques that generate and manipulate multimedia content can now provide a very advanced level of realism. The boundary between real and synthetic media has become very thin. On the one hand, this opens the door to a series of exciting applications in different fields such as creative arts, advertising, film production, and video games. On the other hand, it poses enormous security threats. Software packages freely available on the web allow any individual, without special skills, to create very realistic fake images and videos. These can be used to manipulate public opinion during elections, commit fraud, discredit or blackmail people. Therefore, there is an urgent need for automated tools capable of detecting false multimedia content and avoiding the spread of dangerous false information. This review paper aims to present an analysis of the methods for visual media integrity verification, that is, the detection of manipulated images and videos. Special emphasis will be placed on the emerging phenomenon of deepfakes, fake media created through deep learning tools, and on modern data-driven forensic methods to fight them. The analysis will help highlight the limits of current forensic tools, the most relevant issues, the upcoming challenges, and suggest future directions for research."
                },
                {
                    "title": "Deep High-Resolution Representation Learning for Visual Recognition",
                    "abstract": "High-resolution representations are essential for position-sensitive vision problems, such as human pose estimation, semantic segmentation, and object detection. Existing state-of-the-art frameworks first encode the input image as a low-resolution representation through a subnetwork that is formed by connecting high-to-low resolution convolutions in series (e.g., ResNet, VGGNet), and then recover the high-resolution representation from the encoded low-resolution representation. Instead, our proposed network, named as High-Resolution Network (HRNet), maintains high-resolution representations through the whole process. There are two key characteristics: (i) Connect the high-to-low resolution convolution streams in parallel and (ii) repeatedly exchange the information across resolutions. The benefit is that the resulting representation is semantically richer and spatially more precise. We show the superiority of the proposed HRNet in a wide range of applications, including human pose estimation, semantic segmentation, and object detection, suggesting that the HRNet is a stronger backbone for computer vision problems. All the codes are available at https://github.com/HRNet."
                },
                {
                    "title": "ManTra-Net: Manipulation Tracing Network for Detection and Localization of Image Forgeries With Anomalous Features",
                    "abstract": "To fight against real-life image forgery, which commonly involves different types and combined manipulations, we propose a unified deep neural architecture called ManTra-Net. Unlike many existing solutions, ManTra-Net is an end-to-end network that performs both detection and localization without extra preprocessing and postprocessing. \\manifold{} is a fully convolutional network and handles images of arbitrary sizes and many known forgery types such splicing, copy-move, removal, enhancement, and even unknown types. This paper has three salient contributions. We design a simple yet effective self-supervised learning task to learn robust image manipulation traces from classifying 385 image manipulation types. Further, we formulate the forgery localization problem as a local anomaly detection problem, design a Z-score feature to capture local anomaly, and propose a novel long short-term memory solution to assess local anomalies. Finally, we carefully conduct ablation experiments to systematically optimize the proposed network design. Our extensive experimental results demonstrate the generalizability, robustness and superiority of ManTra-Net, not only in single types of manipulations/forgeries, but also in their complicated combinations."
                },
                {
                    "title": "MFC Datasets: Large-Scale Benchmark Datasets for Media Forensic Challenge Evaluation",
                    "abstract": "We provide a benchmark for digital Media Forensics Challenge (MFC) evaluations. Our comprehensive data comprises over 176,000 high provenance (HP) images and 11,000 HP videos; more than 100,000 manipulated images and 4,000 manipulated videos; 35 million internet images and 300,000 video clips. We have designed and generated a series of development, evaluation, and challenge datasets, and used them to assess the progress and thoroughly analyze the performance of diverse systems on a variety of media forensics tasks in the past two years. In this paper, we first introduce the objectives, challenges, and approaches to building media forensics evaluation datasets. We then discuss our approaches to forensic dataset collection, annotation, and manipulation, and present the design and infrastructure to effectively and efficiently build the evaluation datasets to support various evaluation tasks. Given a specified query, we build an infrastructure that selects the customized evaluation subsets for the targeted analysis report. Finally, we demonstrate the evaluation results in the past evaluations."
                },
                {
                    "title": "Albumentations: fast and flexible image augmentations",
                    "abstract": "Data augmentation is a commonly used technique for increasing both the size and the diversity of labeled training sets by leveraging input transformations that preserve corresponding output labels. In computer vision, image augmentations have become a common implicit regularization technique to combat overfitting in deep learning models and are ubiquitously used to improve performance. While most deep learning frameworks implement basic image transformations, the list is typically limited to some variations of flipping, rotating, scaling, and cropping. Moreover, image processing speed varies in existing image augmentation libraries. We present Albumentations, a fast and flexible open source library for image augmentation with many various image transform operations available that is also an easy-to-use wrapper around other augmentation libraries. We discuss the design principles that drove the implementation of Albumentations and give an overview of the key features and distinct capabilities. Finally, we provide examples of image augmentations for different computer vision tasks and demonstrate that Albumentations is faster than other commonly used image augmentation tools on most image transform operations."
                },
                {
                    "title": "Constrained Convolutional Neural Networks: A New Approach Towards General Purpose Image Manipulation Detection",
                    "abstract": "Identifying the authenticity and processing history of an image is an important task in multimedia forensics. By analyzing traces left by different image manipulations, researchers have been able to develop several algorithms capable of detecting targeted editing operations. While this approach has led to the development of several successful forensic algorithms, an important problem remains: creating forensic detectors for different image manipulations is a difficult and time consuming process. Furthermore, forensic analysts need general purpose forensic algorithms capable of detecting multiple different image manipulations. In this paper, we address both of these problems by proposing a new general purpose forensic approach using convolutional neural networks (CNNs). While CNNs are capable of learning classification features directly from data, in their existing form they tend to learn features representative of an image\u2019s content. To overcome this issue, we have developed a new type of CNN layer, called a constrained convolutional layer, that is able to jointly suppress an image\u2019s content and adaptively learn manipulation detection features. Through a series of experiments, we show that our proposed constrained CNN is able to learn manipulation detection features directly from data. Our experimental results demonstrate that our CNN can detect multiple different editing operations with up to 99.97% accuracy and outperform the existing state-of-the-art general purpose manipulation detector. Furthermore, our constrained CNN can still accurately detect image manipulations in realistic scenarios where there is a source camera model mismatch between the training and testing data."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "COVERAGE \u2014 A novel database for copy-move forgery detection",
                    "abstract": "We present COVERAGE - a novel database containing copy-move forged images and their originals with similar but genuine objects. COVERAGE is designed to highlight and address tamper detection ambiguity of popular methods, caused by self-similarity within natural images. In COVERAGE, forged-original pairs are annotated with (i) the duplicated and forged region masks, and (ii) the tampering factor/similarity metric. For benchmarking, forgery quality is evaluated using (i) computer vision-based methods, and (ii) human detection performance. We also propose a novel sparsity-based metric for efficiently estimating forgery quality. Experimental results show that (a) popular forgery detection methods perform poorly over COVERAGE, and (b) the proposed sparsity based metric best correlates with human detection performance. We release the COVERAGE database to the research community."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "CASIA Image Tampering Detection Evaluation Database",
                    "abstract": "Image forensics has now raised the anxiety of justice as increasing cases of abusing tampered images in newspapers and court for evidence are reported recently. With the goal of verifying image content authenticity, passive-blind image tampering detection is called for. More realistic open benchmark databases are also needed to assist the techniques. Recently, we collect a natural color image database with realistic tampering operations. The database is made publicly available for researchers to compare and evaluate their proposed tampering detection techniques. We call this database CASI-A Image Tampering Detection Evaluation Database. We describe the purpose, the design criterion, the organization and self-evaluation of this database in this paper."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Detecting Image Splicing using Geometry Invariants and Camera Characteristics Consistency",
                    "abstract": "Recent advances in computer technology have made digital image tampering more and more common. In this paper, we propose an authentic vs. spliced image classification method making use of geometry invariants in a semi-automatic manner. For a given image, we identify suspicious splicing areas, compute the geometry invariants from the pixels within each region, and then estimate the camera response function (CRF) from these geometry invariants. The cross-fitting errors are fed into a statistical classifier. Experiments show a very promising accuracy, 87%, over a large data set of 363 natural and spliced images. To the best of our knowledge, this is the first work detecting image splicing by verifying camera characteristic consistency from a single-channel image"
                },
                {
                    "title": "The Fast Fourier Transform and Its Applications",
                    "abstract": "The advent of the fast Fourier transform method has greatly extended our ability to implement Fourier methods on digital computers. A description of the alogorithm and its programming is given here and followed by a theorem relating its operands, the finite sample sequences, to the continuous functions they often are intended to approximate. An analysis of the error due to discrete sampling over finite ranges is given in terms of aliasing. Procedures for computing Fourier integrals, convolutions and lagged products are outlined."
                },
                {
                    "title": "IML-ViT: Image Manipulation Localization by Vision Transformer",
                    "abstract": "Advanced image tampering techniques are increasingly challenging the trustworthiness of multimedia, leading to the development of Image Manipulation Localization (IML). But what makes a good IML model? The answer lies in the way to capture artifacts. Exploiting artifacts requires the model to extract non-semantic discrepancies be-tween manipulated and authentic regions, necessitating explicit comparisons between the two areas. With the self-attention mechanism, naturally, the Transformer should be a better candidate to capture artifacts. However, due to limited datasets, there is currently no pure ViT-based approach for IML to serve as a benchmark, and CNNs dominate the entire task. Nevertheless, CNNs suffer from weak long-range and non-semantic modeling. To bridge this gap, based on the fact that artifacts are sensitive to image resolution, amplified under multi-scale features, and massive at the manipulation border, we formulate the answer to the former question as building a ViT with high-resolution capacity, multi-scale feature extraction capability, and manipulation edge supervision that could converge with a small amount of data. We term this simple but effective ViT paradigm IML-ViT, which has significant potential to become a new benchmark for IML. Extensive experiments on five benchmark datasets verified our model outperforms the state-of-the-art manipulation localization methods. Code and models are available at https://github.com/ SunnyHaze/IML-ViT"
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "The Point Where Reality Meets Fantasy: Mixed Adversarial Generators for Image Splice Detection",
                    "abstract": "Modern photo editing tools allow creating realistic manipulated images easily. While fake images can be quickly generated, learning models for their detection is challenging due to the high variety of tampering artifacts and the lack of large labeled datasets of manipulated images. In this paper, we propose a new framework for training of discriminative segmentation model via an adversarial process. We simultaneously train four models: a generative retouching model G_R that translates manipulated image to the real image domain, a generative annotation model G_A that estimates the pixel-wise probability of image patch being either real or fake, and two discriminators D_R and D_A that qualify the output of G_R and G_A. The aim of model G_R is to maximize the probability of model G_A making a mistake. Our method extends the generative adversarial networks framework with two main contributions: (1) training of a generative model G_R against a deep semantic segmentation network G_A that learns rich scene semantics for manipulated region detection, (2) proposing per class semantic loss that facilitates semantically consistent image retouching by the G_R. We collected large-scale manipulated image dataset to train our model. The dataset includes 16k real and fake images with pixel-level annotations of manipulated areas. The dataset also provides ground truth pixel-level object annotations. We validate our approach on several modern manipulated image datasets, where quantitative results and ablations demonstrate that our method achieves and surpasses the state-of-the-art in manipulated image detection. We made our code and dataset publicly available."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we establish a comprehensive and unified benchmark for Image Manipulation Detection & Localization (IMDL) that addresses the inconsistencies in training and evaluation protocols among existing models?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it will provide a standardized framework for evaluating IMDL models, leading to fair comparisons and more reliable experimental outcomes. A unified benchmark will facilitate the advancement of knowledge in the field, enabling researchers to build upon each other's work more effectively. Additionally, it could lead to practical applications in information forensics and security, where accurate detection and localization of image manipulations are essential.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the lack of publicly available training code for many state-of-the-art models, which complicates reproducibility. Furthermore, IMDL models often utilize diverse low-level features and complex loss functions, necessitating highly customized architectures that are difficult to standardize. Existing frameworks are tightly coupled, leading to inefficiencies in reproducing models and hindering scalability. These technical and practical obstacles make it challenging to create a comprehensive benchmark.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the absence of a unified approach to training and evaluation protocols, as well as the unavailability of source code for many leading models. Barriers such as the complexity of model architectures and the reliance on monolithic frameworks have prevented the establishment of a comprehensive benchmark. Our approach differs by introducing IMDL-BenCo, a modular codebase that separates model architecture, loss design, and evaluation components, allowing for greater flexibility and efficiency in reproducing and comparing IMDL models.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of IMDL-BenCo, which includes a modular codebase with four components: a data loader, model zoo, training script, and evaluator. We will implement or incorporate training code for eight state-of-the-art IMDL models and establish a benchmark with standardized training and evaluation protocols, 15 GPU-accelerated evaluation metrics, and three types of robustness evaluations. The expected outcomes include a comprehensive benchmark that provides new insights into IMDL model architecture, dataset characteristics, and evaluation standards, ultimately advancing the field of image manipulation detection and localization."
            }
        },
        "author_data": {
            "9bf72303-d7b0-4e4b-8493-a33fb5b90b32": {
                "pk": "9bf72303-d7b0-4e4b-8493-a33fb5b90b32",
                "name": "Xiaochen Ma",
                "collaborators": [
                    "Jizhe Zhou",
                    "Zhuohang Jiang",
                    "Xing Huang",
                    "Bo Du",
                    "Ahmed Y. Al Hammadi",
                    "Xia Du",
                    "Ahmed Y. Alhammadi",
                    "Wentao Feng",
                    "Xiong Xu",
                    "Chi-Man Pun"
                ],
                "domain": [
                    "Image Manipulation Localization",
                    "Multimedia Forensics",
                    "Contrastive Learning",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Harnack Inequality for Distribution Dependent Stochastic Hamiltonian System",
                        "abstract": "The dimension free Harnack inequality is established for the distribution dependent stochastic Hamiltonian system, where the drift is Lipschitz continuous in the measure variable under the distance induced by the H\\\"{o}lder-Dini continuous functions, which are $\\beta (\\beta>\\frac{2}{3})$-H\\\"{o}lder continuous on the degenerate component and square root of Dini continuous on the non-degenerate one. The results extend the existing ones in which the drift is Lipschitz continuous in the measure variable under $L^2$-Wasserstein distance."
                    },
                    {
                        "title": "IML-ViT: Benchmarking Image Manipulation Localization by Vision Transformer",
                        "abstract": "Advanced image tampering techniques are increasingly challenging the trustworthiness of multimedia, leading to the development of Image Manipulation Localization (IML). But what makes a good IML model? The answer lies in the way to capture artifacts. Exploiting artifacts requires the model to extract non-semantic discrepancies between manipulated and authentic regions, necessitating explicit comparisons between the two areas. With the self-attention mechanism, naturally, the Transformer should be a better candidate to capture artifacts. However, due to limited datasets, there is currently no pure ViT-based approach for IML to serve as a benchmark, and CNNs dominate the entire task. Nevertheless, CNNs suffer from weak long-range and non-semantic modeling. To bridge this gap, based on the fact that artifacts are sensitive to image resolution, amplified under multi-scale features, and massive at the manipulation border, we formulate the answer to the former question as building a ViT with high-resolution capacity, multi-scale feature extraction capability, and manipulation edge supervision that could converge with a small amount of data. We term this simple but effective ViT paradigm IML-ViT, which has significant potential to become a new benchmark for IML. Extensive experiments on five benchmark datasets verified our model outperforms the state-of-the-art manipulation localization methods.Code and models are available at \\url{https://github.com/SunnyHaze/IML-ViT}."
                    },
                    {
                        "title": "Pre-training-free Image Manipulation Localization through Non-Mutually Exclusive Contrastive Learning",
                        "abstract": "Deep Image Manipulation Localization (IML) models suffer from training data insufficiency and thus heavily rely on pre-training. We argue that contrastive learning is more suitable to tackle the data insufficiency problem for IML. Crafting mutually exclusive positives and negatives is the prerequisite for contrastive learning. However, when adopting contrastive learning in IML, we encounter three categories of image patches: tampered, authentic, and contour patches. Tampered and authentic patches are naturally mutually exclusive, but contour patches containing both tampered and authentic pixels are non-mutually exclusive to them. Simply abnegating these contour patches results in a drastic performance loss since contour patches are decisive to the learning outcomes. Hence, we propose the Non-mutually exclusive Contrastive Learning (NCL) framework to rescue conventional contrastive learning from the above dilemma. In NCL, to cope with the non-mutually exclusivity, we first establish a pivot structure with dual branches to constantly switch the role of contour patches between positives and negatives while training. Then, we devise a pivot-consistent loss to avoid spatial corruption caused by the role-switching process. In this manner, NCL both inherits the self-supervised merits to address the data insufficiency and retains a high manipulation localization accuracy. Extensive experiments verify that our NCL achieves state-of-the-art performance on all five benchmarks without any pre-training and is more robust on unseen real-life samples. The code is available at: https://github.com/Knightzjz/NCL-IML."
                    },
                    {
                        "title": "Perceptual MAE for Image Manipulation Localization: A High-level Vision Learner Focusing on Low-level Features",
                        "abstract": "Nowadays, multimedia forensics faces unprecedented challenges due to the rapid advancement of multimedia generation technology thereby making Image Manipulation Localization (IML) crucial in the pursuit of truth. The key to IML lies in revealing the artifacts or inconsistencies between the tampered and authentic areas, which are evident under pixel-level features. Consequently, existing studies treat IML as a low-level vision task, focusing on allocating tampered masks by crafting pixel-level features such as image RGB noises, edge signals, or high-frequency features. However, in practice, tampering commonly occurs at the object level, and different classes of objects have varying likelihoods of becoming targets of tampering. Therefore, object semantics are also vital in identifying the tampered areas in addition to pixel-level features. This necessitates IML models to carry out a semantic understanding of the entire image. In this paper, we reformulate the IML task as a high-level vision task that greatly benefits from low-level features. Based on such an interpretation, we propose a method to enhance the Masked Autoencoder (MAE) by incorporating high-resolution inputs and a perceptual loss supervision module, which is termed Perceptual MAE (PMAE). While MAE has demonstrated an impressive understanding of object semantics, PMAE can also compensate for low-level semantics with our proposed enhancements. Evidenced by extensive experiments, this paradigm effectively unites the low-level and high-level features of the IML task and outperforms state-of-the-art tampering localization methods on all five publicly available datasets."
                    },
                    {
                        "title": "Ordeal Mechanisms, Information, and the Cost-Effectiveness of Subsidies: Evidence from Subsidized Eyeglasses in Rural China",
                        "abstract": "The cost-effectiveness of policies providing subsidized goods is often compromised by limited use of the goods provided. Through a randomized trial, we test two approaches to improve the cost-effectiveness of a program distributing free eyeglasses to myopic children in rural China. Requiring recipients to undergo an ordeal better targeted eyeglasses to those who used them without reducing usage relative to free delivery. An information campaign increased use when eyeglasses were freely delivered but not under an ordeal. Free delivery plus information was determined to be the most socially cost-effective approach and obtained the highest rate of eyeglass use."
                    }
                ]
            },
            "229ee020-d55c-4b8f-beb1-97831bca57e2": {
                "pk": "229ee020-d55c-4b8f-beb1-97831bca57e2",
                "name": "Xuekang Zhu",
                "collaborators": [
                    "Jizhe Zhou",
                    "Xiwen Wang",
                    "Cheng Li",
                    "Mao Li",
                    "Xinyu Yang",
                    "Xiaochen Ma",
                    "Bo Du",
                    "Lei Su",
                    "Bingkui Tong",
                    "Zeyu Lei"
                ],
                "domain": [
                    "Image Generation",
                    "Image Manipulation",
                    "Computer Vision",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Saliency Guided Optimization of Diffusion Latents",
                        "abstract": "With the rapid advances in diffusion models, generating decent images from text prompts is no longer challenging. The key to text-to-image generation is how to optimize the results of a text-to-image generation model so that they can be better aligned with human intentions or prompts. Existing optimization methods commonly treat the entire image uniformly and conduct global optimization. These methods overlook the fact that when viewing an image, the human visual system naturally prioritizes attention toward salient areas, often neglecting less or non-salient regions. That is, humans are likely to neglect optimizations in non-salient areas. Consequently, although model retaining is conducted under the guidance of additional large and multimodality models, existing methods, which perform uniform optimizations, yield sub-optimal results. To address this alignment challenge effectively and efficiently, we propose Saliency Guided Optimization Of Diffusion Latents (SGOOL). We first employ a saliency detector to mimic the human visual attention system and mark out the salient regions. To avoid retraining an additional model, our method directly optimizes the diffusion latents. Besides, SGOOL utilizes an invertible diffusion process and endows it with the merits of constant memory implementation. Hence, our method becomes a parameter-efficient and plug-and-play fine-tuning method. Extensive experiments have been done with several metrics and human evaluation. Experimental results demonstrate the superiority of SGOOL in image quality and prompt alignment."
                    },
                    {
                        "title": "M^3:Manipulation Mask Manufacturer for Arbitrary-Scale Super-Resolution Mask",
                        "abstract": "In the field of image manipulation localization (IML), the small quantity and poor quality of existing datasets have always been major issues. A dataset containing various types of manipulations will greatly help improve the accuracy of IML models. Images on the internet (such as those on Baidu Tieba's PS Bar) are manipulated using various techniques, and creating a dataset from these images will significantly enrich the types of manipulations in our data. However, images on the internet suffer from resolution and clarity issues, and the masks obtained by simply subtracting the manipulated image from the original contain various noises. These noises are difficult to remove, rendering the masks unusable for IML models. Inspired by the field of change detection, we treat the original and manipulated images as changes over time for the same image and view the data generation task as a change detection task. However, due to clarity issues between images, conventional change detection models perform poorly. Therefore, we introduced a super-resolution module and proposed the Manipulation Mask Manufacturer (MMM) framework. It enhances the resolution of both the original and tampered images, thereby improving image details for better comparison. Simultaneously, the framework converts the original and tampered images into feature embeddings and concatenates them, effectively modeling the context. Additionally, we created the Manipulation Mask Manufacturer Dataset (MMMD), a dataset that covers a wide range of manipulation techniques. We aim to contribute to the fields of image forensics and manipulation detection by providing more realistic manipulation data through MMM and MMMD. Detailed information about MMMD and the download link can be found at: the code and datasets will be made available."
                    }
                ]
            },
            "90bc1c22-a3d0-49eb-bf77-99ec073c0c17": {
                "pk": "90bc1c22-a3d0-49eb-bf77-99ec073c0c17",
                "name": "Lei Su",
                "collaborators": [
                    "Ivar Martin",
                    "Pinaki Sengupta",
                    "Meng Zeng",
                    "Bin Xiao"
                ],
                "domain": [
                    "Quantum Field Theory",
                    "Topological Phases",
                    "Nonunitary Quantum Circuits",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Boundary criticality via gauging finite subgroups: a case study on the clock model",
                        "abstract": "Gauging a finite Abelian normal subgroup $\\Gamma$ of a nonanomalous 0-form symmetry $G$ of a theory in $(d+1)$D spacetime can yield an unconventional critical point if the original theory has a continuous transition where $\\Gamma$ is completely spontaneously broken and if $G$ is a nontrivial extension of $G/\\Gamma$ by $\\Gamma$. The gauged theory has symmetry $G/\\Gamma \\times \\hat{\\Gamma}^{(d-1)}$, where $\\hat{\\Gamma}^{(d-1)}$ is the $(d-1)$-form dual symmetry of $\\Gamma$, and a 't Hooft anomaly between them. Thus it can be viewed as a boundary of a topological phase protected by $G/\\Gamma \\times \\hat{\\Gamma}^{(d-1)}$. The ordinary critical point, upon gauging, is mapped to a deconfined quantum critical point between two ordinary symmetry-breaking phases ($d =1$) or an unconventional quantum critical point between an ordinary symmetry-breaking phase and a topologically ordered phase ($d\\ge 2$) associated with $G/\\Gamma$ and $\\hat{\\Gamma}^{(d-1)}$, respectively. Order parameters and disorder parameters, before and after gauging, can be directly related. As a concrete example, we gauge the $\\mathbb{Z}_2$ subgroup of $\\mathbb{Z}_4$ symmetry of a 4-state clock model on a 1D lattice and a 2D square lattice. Since the symmetry of the clock model contains $D_8$, the dihedral group of order 8, we also analyze the anomaly structure which is similar to that in the compactified $SU(2)$ gauge theory with $\\theta =\\pi$ in $(3+1)$D and its mixed gauge theory. The general case is also discussed."
                    },
                    {
                        "title": "Integrable nonunitary quantum circuits",
                        "abstract": "We show that the integrable Lindblad superoperators found recently can be used to build integrable nonunitary quantum circuits with two-site gates by demonstrating that the $\\check{R}$-matrices are completely positive and trace preserving. Using the bond-site transformation, we obtain the corresponding integrable nonunitary quantum circuits with three-site gates. When restricted to the diagonals of the density matrix, these quantum models reduce to integrable classical cellular automata. This approach may pave the way for a systematic construction of integrable nonunitary quantum circuits."
                    },
                    {
                        "title": "Dimer-induced heavy-fermion superconductivity in the Shastry-Sutherland Kondo lattice model",
                        "abstract": "We study the Kondo lattice model on the geometrically frustrated Shastry-Sutherland lattice focusing on the quantum phase transition between the valence bond solid and the heavy fermion liquid phase. By explicitly including spinon pairing of local moments at the mean-field level, we establish the emergence of a unique heavy fermion superconducting phase induced by dimer ordering of the local moments coexisting with Kondo hybridization. Furthermore, we demonstrate that for suitable choices of parameters, a partial Kondo-screening phase, where some of the valence bonds are broken, precedes the aforementioned dimer-induced superconducting phase. Our results have important implications in understanding the experimental observations in the heavy fermion Shasstry-Sutherland compounds."
                    },
                    {
                        "title": "Gapless symmetry-protected topological phases and generalized deconfined critical points from gauging a finite subgroup",
                        "abstract": "Gauging a finite subgroup of a global symmetry can map conventional phases and phase transitions to unconventional ones. In this work, we study, as a concrete example, an emergent $\\mathbb{Z}_2$-gauged system with global symmetry $U(1)$, namely, the $\\mathbb{Z}_2$-gauged Bose-Hubbard model both in 1-D and in 2-D. In certain limits, there is an emergent mixed 't Hooft anomaly between the quotient $\\tilde{U}(1)$ symmetry and the dual $\\hat{\\mathbb{Z}}_2$ symmetry. In 1-D, the superfluid phase is mapped to an intrinsically gapless symmetry-protected topological (SPT) phase, as supported by density-matrix renormalization group (DMRG) calculations. In 2-D, the original superfluid-insulator transition becomes a generalized deconfined quantum critical point (DQCP) between a gapless SPT phase, where a SPT order coexists with Goldstone modes, and a $\\tilde{U}(1)$-symmetry-enriched topological (SET) phase. We also discuss the stability of these phases and the critical points to small perturbations and their potential experimental realizations. Our work demonstrates that partial gauging is a simple and yet powerful approach in constructing novel phases and quantum criticalities."
                    },
                    {
                        "title": "Global anomalies of Green's function zeros",
                        "abstract": "Anomaly analysis has been an important and powerful tool in studying nonperturbative physics for decades. The anomaly inflow mechanism provides an intuitive interpretation of the bulk-boundary correspondence in topological systems. In this work, we study global anomalies in systems with symmetry-preserving Luttinger surfaces, i.e. the manifolds of fermionic Green's function zeros in the momentum space at zero energy, described by nonlocal effective theories. We view the nonlocal effective theories as a result of integrating out some low energy states. Assuming that the states integrated out do not make extra contributions to the anomalies, we discuss the simplest Lagrangian describing a gapless Dirac zero and a two-pole variant, their global anomalies, and the bulk-boundary correspondence. We then consider the constraints on possible phases with Green's function zeros of Dirac type, such as non-Fermi liquids and emergent gapless quasiparticles on Luttinger surfaces. We also provide some perspectives on why the nonlocal fermionic effective theory discussed by Golterman and Shamir (arXiv: 2311.12790) is not a suitable starting point for a symmetrically gapped phase."
                    },
                    {
                        "title": "ISO: Overlap of Computation and Communication within Seqenence For LLM Inference",
                        "abstract": "In the realm of Large Language Model (LLM) inference, the inherent structure of transformer models coupled with the multi-GPU tensor parallelism strategy leads to a sequential execution of computation and communication. This results in substantial underutilization of computing resources during the communication phase. To mitigate this inefficiency, various techniques have been developed to optimize the use of computational power throughout the communication process. These strategies primarily involve overlapping matrix computations and communications, as well as interleaving micro-batches across different requests. Nonetheless, these approaches either fall short of achieving ideal overlap or impose certain limitations on their application. To overcome these challenges, this paper introduces a novel strategy for computation-communication overlap that operates at the sequence level. This method not only enhances the degree of overlap but also minimizes the constraints on its applicability. Experimental evaluations conducted using 30b/70b models have demonstrated significant improvements in efficiency. Specifically, the proposed technique has been shown to reduce time consumption by approximately 35% on 4090 GPU and by roughly 15% on A800 GPU during the prefill stage of LLM inference."
                    }
                ]
            },
            "635660fa-583c-4067-9ef5-8edc07b225f7": {
                "pk": "635660fa-583c-4067-9ef5-8edc07b225f7",
                "name": "Bo Du",
                "collaborators": [
                    "Bo Han",
                    "Tongliang Liu",
                    "Qikui Zhu",
                    "Pingkun Yan",
                    "Siyuan Fan",
                    "Xiantao Cai",
                    "Bo Peng",
                    "Longling Sun",
                    "Chaojian Yu",
                    "Mingming Gong"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Adversarial Training",
                    "Medical Image Segmentation",
                    "Few-Shot Learning"
                ],
                "publications": [
                    {
                        "title": "TextIM: Part-aware Interactive Motion Synthesis from Text",
                        "abstract": "In this work, we propose TextIM, a novel framework for synthesizing TEXT-driven human Interactive Motions, with a focus on the precise alignment of part-level semantics. Existing methods often overlook the critical roles of interactive body parts and fail to adequately capture and align part-level semantics, resulting in inaccuracies and even erroneous movement outcomes. To address these issues, TextIM utilizes a decoupled conditional diffusion framework to enhance the detailed alignment between interactive movements and corresponding semantic intents from textual descriptions. Our approach leverages large language models, functioning as a human brain, to identify interacting human body parts and to comprehend interaction semantics to generate complicated and subtle interactive motion. Guided by the refined movements of the interacting parts, TextIM further extends these movements into a coherent whole-body motion. We design a spatial coherence module to complement the entire body movements while maintaining consistency and harmony across body parts using a part graph convolutional network. For training and evaluation, we carefully selected and re-labeled interactive motions from HUMANML3D to develop a specialized dataset. Experimental results demonstrate that TextIM produces semantically accurate human interactive motions, significantly enhancing the realism and applicability of synthesized interactive motions in diverse scenarios, even including interactions with deformable and dynamically changing objects."
                    },
                    {
                        "title": "Robust Weight Perturbation for Adversarial Training",
                        "abstract": "Overfitting widely exists in adversarial robust training of deep networks. An effective remedy is adversarial weight perturbation, which injects the worst-case weight perturbation during network training by maximizing the classification loss on adversarial examples. Adversarial weight perturbation helps reduce the robust generalization gap; however, it also undermines the robustness improvement. A criterion that regulates the weight perturbation is therefore crucial for adversarial training. In this paper, we propose such a criterion, namely Loss Stationary Condition (LSC) for constrained perturbation. With LSC, we find that it is essential to conduct weight perturbation on adversarial data with small classification loss to eliminate robust overfitting. Weight perturbation on adversarial data with large classification loss is not necessary and may even lead to poor robustness. Based on these observations, we propose a robust perturbation strategy to constrain the extent of weight perturbation. The perturbation strategy prevents deep networks from overfitting while avoiding the side effect of excessive weight perturbation, significantly improving the robustness of adversarial training. Extensive experiments demonstrate the superiority of the proposed method over the state-of-the-art adversarial training methods."
                    },
                    {
                        "title": "MOKD: Cross-domain Finetuning for Few-shot Classification via Maximizing Optimized Kernel Dependence",
                        "abstract": "In cross-domain few-shot classification, \\emph{nearest centroid classifier} (NCC) aims to learn representations to construct a metric space where few-shot classification can be performed by measuring the similarities between samples and the prototype of each class. An intuition behind NCC is that each sample is pulled closer to the class centroid it belongs to while pushed away from those of other classes. However, in this paper, we find that there exist high similarities between NCC-learned representations of two samples from different classes. In order to address this problem, we propose a bi-level optimization framework, \\emph{maximizing optimized kernel dependence} (MOKD) to learn a set of class-specific representations that match the cluster structures indicated by labeled data of the given task. Specifically, MOKD first optimizes the kernel adopted in \\emph{Hilbert-Schmidt independence criterion} (HSIC) to obtain the optimized kernel HSIC (opt-HSIC) that can capture the dependence more precisely. Then, an optimization problem regarding the opt-HSIC is addressed to simultaneously maximize the dependence between representations and labels and minimize the dependence among all samples. Extensive experiments on Meta-Dataset demonstrate that MOKD can not only achieve better generalization performance on unseen domains in most cases but also learn better data representation clusters. The project repository of MOKD is available at: \\href{https://github.com/tmlr-group/MOKD}{https://github.com/tmlr-group/MOKD}."
                    },
                    {
                        "title": "OASIS: One-pass aligned Atlas Set for Image Segmentation",
                        "abstract": "Medical image segmentation is a fundamental task in medical image analysis. Despite that deep convolutional neural networks have gained stellar performance in this challenging task, they typically rely on large labeled datasets, which have limited their extension to customized applications. By revisiting the superiority of atlas based segmentation methods, we present a new framework of One-pass aligned Atlas Set for Images Segmentation (OASIS). To address the problem of time-consuming iterative image registration used for atlas warping, the proposed method takes advantage of the power of deep learning to achieve one-pass image registration. In addition, by applying label constraint, OASIS also makes the registration process to be focused on the regions to be segmented for improving the performance of segmentation. Furthermore, instead of using image based similarity for label fusion, which can be distracted by the large background areas, we propose a novel strategy to compute the label similarity based weights for label fusion. Our experimental results on the challenging task of prostate MR image segmentation demonstrate that OASIS is able to significantly increase the segmentation performance compared to other state-of-the-art methods."
                    },
                    {
                        "title": "Multi-Temporal Scene Classification and Scene Change Detection with Correlation based Fusion",
                        "abstract": "Classifying multi-temporal scene land-use categories and detecting their semantic scene-level changes for imagery covering urban regions could straightly reflect the land-use transitions. Existing methods for scene change detection rarely focus on the temporal correlation of bi-temporal features, and are mainly evaluated on small scale scene change detection datasets. In this work, we proposed a CorrFusion module that fuses the highly correlated components in bi-temporal feature embeddings. We firstly extracts the deep representations of the bi-temporal inputs with deep convolutional networks. Then the extracted features will be projected into a lower dimension space to computed the instance-level correlation. The cross-temporal fusion will be performed based on the computed correlation in CorrFusion module. The final scene classification are obtained with softmax activation layers. In the objective function, we introduced a new formulation for calculating the temporal correlation. The detailed derivation of backpropagation gradients for the proposed module is also given in this paper. Besides, we presented a much larger scale scene change detection dataset and conducted experiments on this dataset. The experimental results demonstrated that our proposed CorrFusion module could remarkably improve the multi-temporal scene classification and scene change detection results."
                    },
                    {
                        "title": "Self-supervised Training of Graph Convolutional Networks",
                        "abstract": "Graph Convolutional Networks (GCNs) have been successfully applied to analyze non-grid data, where the classical convolutional neural networks (CNNs) cannot be directly used. One similarity shared by GCNs and CNNs is the requirement of massive amount of labeled data for network training. In addition, GCNs need the adjacency matrix as input to define the relationship between those non-grid data, which leads to all of data including training, validation and test data typically forms only one graph structures data for training. Furthermore, the adjacency matrix is usually pre-defined and stationary, which makes the data augmentation strategies cannot be employed on the constructed graph structures data to augment the amount of training data. To further improve the learning capacity and model performance under the limited training data, in this paper, we propose two types of self-supervised learning strategies to exploit available information from the input graph structure data itself. Our proposed self-supervised learning strategies are examined on two representative GCN models with three public citation network datasets - Citeseer, Cora and Pubmed. The experimental results demonstrate the generalization ability as well as the portability of our proposed strategies, which can significantly improve the performance of GCNs with the power of self-supervised learning in improving feature learning."
                    }
                ]
            },
            "a0e081f4-311e-4669-861c-6d4be59612ac": {
                "pk": "a0e081f4-311e-4669-861c-6d4be59612ac",
                "name": "Zhuohang Jiang",
                "collaborators": [
                    "Jizhe Zhou",
                    "Bingkui Tong",
                    "Xia Du",
                    "Ahmed Alhammadi",
                    "Xiaochen Ma",
                    "Bo Du",
                    "Ahmed Y. Al Hammadi",
                    "Xiong Xu",
                    "Chi-Man Pun"
                ],
                "domain": [
                    "Privacy Detection",
                    "Image Manipulation",
                    "Computer Vision",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "SHAN: Object-Level Privacy Detection via Inference on Scene Heterogeneous Graph",
                        "abstract": "With the rise of social platforms, protecting privacy has become an important issue. Privacy object detection aims to accurately locate private objects in images. It is the foundation of safeguarding individuals' privacy rights and ensuring responsible data handling practices in the digital age. Since privacy of object is not shift-invariant, the essence of the privacy object detection task is inferring object privacy based on scene information. However, privacy object detection has long been studied as a subproblem of common object detection tasks. Therefore, existing methods suffer from serious deficiencies in accuracy, generalization, and interpretability. Moreover, creating large-scale privacy datasets is difficult due to legal constraints and existing privacy datasets lack label granularity. The granularity of existing privacy detection methods remains limited to the image level. To address the above two issues, we introduce two benchmark datasets for object-level privacy detection and propose SHAN, Scene Heterogeneous graph Attention Network, a model constructs a scene heterogeneous graph from an image and utilizes self-attention mechanisms for scene inference to obtain object privacy. Through experiments, we demonstrated that SHAN performs excellently in privacy object detection tasks, with all metrics surpassing those of the baseline model."
                    },
                    {
                        "title": "Beyond Visual Appearances: Privacy-sensitive Objects Identification via Hybrid Graph Reasoning",
                        "abstract": "The Privacy-sensitive Object Identification (POI) task allocates bounding boxes for privacy-sensitive objects in a scene. The key to POI is settling an object's privacy class (privacy-sensitive or non-privacy-sensitive). In contrast to conventional object classes which are determined by the visual appearance of an object, one object's privacy class is derived from the scene contexts and is subject to various implicit factors beyond its visual appearance. That is, visually similar objects may be totally opposite in their privacy classes. To explicitly derive the objects' privacy class from the scene contexts, in this paper, we interpret the POI task as a visual reasoning task aimed at the privacy of each object in the scene. Following this interpretation, we propose the PrivacyGuard framework for POI. PrivacyGuard contains three stages. i) Structuring: an unstructured image is first converted into a structured, heterogeneous scene graph that embeds rich scene contexts. ii) Data Augmentation: a contextual perturbation oversampling strategy is proposed to create slightly perturbed privacy-sensitive objects in a scene graph, thereby balancing the skewed distribution of privacy classes. iii) Hybrid Graph Generation & Reasoning: the balanced, heterogeneous scene graph is then transformed into a hybrid graph by endowing it with extra \"node-node\" and \"edge-edge\" homogeneous paths. These homogeneous paths allow direct message passing between nodes or edges, thereby accelerating reasoning and facilitating the capturing of subtle context changes. Based on this hybrid graph... **For the full abstract, see the original paper.**"
                    },
                    {
                        "title": "IML-ViT: Benchmarking Image Manipulation Localization by Vision Transformer",
                        "abstract": "Advanced image tampering techniques are increasingly challenging the trustworthiness of multimedia, leading to the development of Image Manipulation Localization (IML). But what makes a good IML model? The answer lies in the way to capture artifacts. Exploiting artifacts requires the model to extract non-semantic discrepancies between manipulated and authentic regions, necessitating explicit comparisons between the two areas. With the self-attention mechanism, naturally, the Transformer should be a better candidate to capture artifacts. However, due to limited datasets, there is currently no pure ViT-based approach for IML to serve as a benchmark, and CNNs dominate the entire task. Nevertheless, CNNs suffer from weak long-range and non-semantic modeling. To bridge this gap, based on the fact that artifacts are sensitive to image resolution, amplified under multi-scale features, and massive at the manipulation border, we formulate the answer to the former question as building a ViT with high-resolution capacity, multi-scale feature extraction capability, and manipulation edge supervision that could converge with a small amount of data. We term this simple but effective ViT paradigm IML-ViT, which has significant potential to become a new benchmark for IML. Extensive experiments on five benchmark datasets verified our model outperforms the state-of-the-art manipulation localization methods.Code and models are available at \\url{https://github.com/SunnyHaze/IML-ViT}."
                    },
                    {
                        "title": "Perceptual MAE for Image Manipulation Localization: A High-level Vision Learner Focusing on Low-level Features",
                        "abstract": "Nowadays, multimedia forensics faces unprecedented challenges due to the rapid advancement of multimedia generation technology thereby making Image Manipulation Localization (IML) crucial in the pursuit of truth. The key to IML lies in revealing the artifacts or inconsistencies between the tampered and authentic areas, which are evident under pixel-level features. Consequently, existing studies treat IML as a low-level vision task, focusing on allocating tampered masks by crafting pixel-level features such as image RGB noises, edge signals, or high-frequency features. However, in practice, tampering commonly occurs at the object level, and different classes of objects have varying likelihoods of becoming targets of tampering. Therefore, object semantics are also vital in identifying the tampered areas in addition to pixel-level features. This necessitates IML models to carry out a semantic understanding of the entire image. In this paper, we reformulate the IML task as a high-level vision task that greatly benefits from low-level features. Based on such an interpretation, we propose a method to enhance the Masked Autoencoder (MAE) by incorporating high-resolution inputs and a perceptual loss supervision module, which is termed Perceptual MAE (PMAE). While MAE has demonstrated an impressive understanding of object semantics, PMAE can also compensate for low-level semantics with our proposed enhancements. Evidenced by extensive experiments, this paradigm effectively unites the low-level and high-level features of the IML task and outperforms state-of-the-art tampering localization methods on all five publicly available datasets."
                    }
                ]
            },
            "554e9159-c2a4-4f65-86b0-3ba2b28b28bd": {
                "pk": "554e9159-c2a4-4f65-86b0-3ba2b28b28bd",
                "name": "Bingkui Tong",
                "collaborators": [
                    "Jizhe Zhou",
                    "Zhuohang Jiang",
                    "Xia Du",
                    "Ahmed Alhammadi",
                    "Xinyu Yang",
                    "Xiaochen Ma",
                    "Xuekang Zhu",
                    "Bo Du",
                    "Lei Su",
                    "Zeyu Lei"
                ],
                "domain": [
                    "Privacy Object Detection",
                    "Image Manipulation Localization",
                    "Graph Neural Network",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "SHAN: Object-Level Privacy Detection via Inference on Scene Heterogeneous Graph",
                        "abstract": "With the rise of social platforms, protecting privacy has become an important issue. Privacy object detection aims to accurately locate private objects in images. It is the foundation of safeguarding individuals' privacy rights and ensuring responsible data handling practices in the digital age. Since privacy of object is not shift-invariant, the essence of the privacy object detection task is inferring object privacy based on scene information. However, privacy object detection has long been studied as a subproblem of common object detection tasks. Therefore, existing methods suffer from serious deficiencies in accuracy, generalization, and interpretability. Moreover, creating large-scale privacy datasets is difficult due to legal constraints and existing privacy datasets lack label granularity. The granularity of existing privacy detection methods remains limited to the image level. To address the above two issues, we introduce two benchmark datasets for object-level privacy detection and propose SHAN, Scene Heterogeneous graph Attention Network, a model constructs a scene heterogeneous graph from an image and utilizes self-attention mechanisms for scene inference to obtain object privacy. Through experiments, we demonstrated that SHAN performs excellently in privacy object detection tasks, with all metrics surpassing those of the baseline model."
                    },
                    {
                        "title": "Beyond Visual Appearances: Privacy-sensitive Objects Identification via Hybrid Graph Reasoning",
                        "abstract": "The Privacy-sensitive Object Identification (POI) task allocates bounding boxes for privacy-sensitive objects in a scene. The key to POI is settling an object's privacy class (privacy-sensitive or non-privacy-sensitive). In contrast to conventional object classes which are determined by the visual appearance of an object, one object's privacy class is derived from the scene contexts and is subject to various implicit factors beyond its visual appearance. That is, visually similar objects may be totally opposite in their privacy classes. To explicitly derive the objects' privacy class from the scene contexts, in this paper, we interpret the POI task as a visual reasoning task aimed at the privacy of each object in the scene. Following this interpretation, we propose the PrivacyGuard framework for POI. PrivacyGuard contains three stages. i) Structuring: an unstructured image is first converted into a structured, heterogeneous scene graph that embeds rich scene contexts. ii) Data Augmentation: a contextual perturbation oversampling strategy is proposed to create slightly perturbed privacy-sensitive objects in a scene graph, thereby balancing the skewed distribution of privacy classes. iii) Hybrid Graph Generation & Reasoning: the balanced, heterogeneous scene graph is then transformed into a hybrid graph by endowing it with extra \"node-node\" and \"edge-edge\" homogeneous paths. These homogeneous paths allow direct message passing between nodes or edges, thereby accelerating reasoning and facilitating the capturing of subtle context changes. Based on this hybrid graph... **For the full abstract, see the original paper.**"
                    },
                    {
                        "title": "M^3:Manipulation Mask Manufacturer for Arbitrary-Scale Super-Resolution Mask",
                        "abstract": "In the field of image manipulation localization (IML), the small quantity and poor quality of existing datasets have always been major issues. A dataset containing various types of manipulations will greatly help improve the accuracy of IML models. Images on the internet (such as those on Baidu Tieba's PS Bar) are manipulated using various techniques, and creating a dataset from these images will significantly enrich the types of manipulations in our data. However, images on the internet suffer from resolution and clarity issues, and the masks obtained by simply subtracting the manipulated image from the original contain various noises. These noises are difficult to remove, rendering the masks unusable for IML models. Inspired by the field of change detection, we treat the original and manipulated images as changes over time for the same image and view the data generation task as a change detection task. However, due to clarity issues between images, conventional change detection models perform poorly. Therefore, we introduced a super-resolution module and proposed the Manipulation Mask Manufacturer (MMM) framework. It enhances the resolution of both the original and tampered images, thereby improving image details for better comparison. Simultaneously, the framework converts the original and tampered images into feature embeddings and concatenates them, effectively modeling the context. Additionally, we created the Manipulation Mask Manufacturer Dataset (MMMD), a dataset that covers a wide range of manipulation techniques. We aim to contribute to the fields of image forensics and manipulation detection by providing more realistic manipulation data through MMM and MMMD. Detailed information about MMMD and the download link can be found at: the code and datasets will be made available."
                    }
                ]
            },
            "aa6b1347-223d-47d1-b8d5-92c71aec3673": {
                "pk": "aa6b1347-223d-47d1-b8d5-92c71aec3673",
                "name": "Zeyu Lei",
                "collaborators": [
                    "Tian Yang"
                ],
                "domain": [
                    "Plasmonics",
                    "Surface Plasmon Polariton",
                    "Optical Devices",
                    "Particle Swarm Optimization"
                ],
                "publications": [
                    {
                        "title": "Gap plasmon resonator arrays for unidirectional launching and shaping of surface plasmon polaritons",
                        "abstract": "We report the design and experimental realization of a kind of miniaturized devices for efficient unidirectional launching and shaping of surface plasmon polaritons (SPPs). Each device consists of an array of evenly spaced gap plasmon resonators with varying dimensions. Particle swarm optimization is used to achieve a theoretical two dimensional launching efficiency of about 51%, under the normal illumination of a 5-{\\mu}m waist Gaussian beam at 780 nm. By modifying the wavefront of the SPPs, unidirectional SPPs with focused, Bessel and Airy profiles are launched and imaged with leakage radiation microscopy."
                    }
                ]
            },
            "d37cf132-9d33-417c-ada7-14219fdb2a2b": {
                "pk": "d37cf132-9d33-417c-ada7-14219fdb2a2b",
                "name": "Xinyu Yang",
                "collaborators": [
                    "Zhaoxi Mu",
                    "Jizhe Zhou",
                    "Shulei Ji",
                    "Hongbo Bo",
                    "Shuai Dong",
                    "Cong Zhao",
                    "Shusen Yang",
                    "Julie McCann",
                    "Majid Mirmehdi",
                    "Tilo Burghardt"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Audio Processing",
                    "Music Generation"
                ],
                "publications": [
                    {
                        "title": "High-Fidelity Face Swapping with Style Blending",
                        "abstract": "Face swapping has gained significant traction, driven by the plethora of human face synthesis facilitated by deep learning methods. However, previous face swapping methods that used generative adversarial networks (GANs) as backbones have faced challenges such as inconsistency in blending, distortions, artifacts, and issues with training stability. To address these limitations, we propose an innovative end-to-end framework for high-fidelity face swapping. First, we introduce a StyleGAN-based facial attributes encoder that extracts essential features from faces and inverts them into a latent style code, encapsulating indispensable facial attributes for successful face swapping. Second, we introduce an attention-based style blending module to effectively transfer Face IDs from source to target. To ensure accurate and quality transferring, a series of constraint measures including contrastive face ID learning, facial landmark alignment, and dual swap consistency is implemented. Finally, the blended style code is translated back to the image space via the style decoder, which is of high training stability and generative capability. Extensive experiments on the CelebA-HQ dataset highlight the superior visual quality of generated images from our face-swapping methodology when compared to other state-of-the-art methods, and the effectiveness of each proposed module. Source code and weights will be publicly available."
                    },
                    {
                        "title": "Oxidation tuning of ferroic transitions in Gd$_2$C monolayer",
                        "abstract": "Tuning of ferroic phases provides great opportunities for material functionalities, especially in two-dimensional materials. Here, a $4f$ rare-earth carbide Gd$_2$C monolayer is predicted to be ferromagnetic metal with large magnetization, inherited from its bulk property. Based on first-principles calculations, we propose a strategy that the surface passivation can effectively tune its ferroicity, namely switching among ferromagnetic, antiferromagnetic, and ferroelectric phases. Metal-insulator transition also occurs accompanying these ferroic transitions. Our calculation also suggests that the magneto-optic Kerr effect and second harmonic generation are effective methods to monitor these phase transitions."
                    },
                    {
                        "title": "Manipulation Mask Generator: High-Quality Image Manipulation Mask Generation Method Based on Modified Total Variation Noise Reduction",
                        "abstract": "In artificial intelligence, any model that wants to achieve a good result is inseparable from a large number of high-quality data. It is especially true in the field of tamper detection. This paper proposes a modified total variation noise reduction method to acquire high-quality tampered images. We automatically crawl original and tampered images from the Baidu PS Bar. Baidu PS Bar is a website where net friends post countless tampered images. Subtracting the original image with the tampered image can highlight the tampered area. However, there is also substantial noise on the final print, so these images can't be directly used in the deep learning model. Our modified total variation noise reduction method is aimed at solving this problem. Because a lot of text is slender, it is easy to lose text information after the opening and closing operation. We use MSER (Maximally Stable Extremal Regions) and NMS (Non-maximum Suppression) technology to extract text information. And then use the modified total variation noise reduction technology to process the subtracted image. Finally, we can obtain an image with little noise by adding the image and text information. And the idea also largely retains the text information. Datasets generated in this way can be used in deep learning models, and they will help the model achieve better results."
                    },
                    {
                        "title": "Research about the Ability of LLM in the Tamper-Detection Area",
                        "abstract": "In recent years, particularly since the early 2020s, Large Language Models (LLMs) have emerged as the most powerful AI tools in addressing a diverse range of challenges, from natural language processing to complex problem-solving in various domains. In the field of tamper detection, LLMs are capable of identifying basic tampering activities.To assess the capabilities of LLMs in more specialized domains, we have collected five different LLMs developed by various companies: GPT-4, LLaMA, Bard, ERNIE Bot 4.0, and Tongyi Qianwen. This diverse range of models allows for a comprehensive evaluation of their performance in detecting sophisticated tampering instances.We devised two domains of detection: AI-Generated Content (AIGC) detection and manipulation detection. AIGC detection aims to test the ability to distinguish whether an image is real or AI-generated. Manipulation detection, on the other hand, focuses on identifying tampered images. According to our experiments, most LLMs can identify composite pictures that are inconsistent with logic, and only more powerful LLMs can distinguish logical, but visible signs of tampering to the human eye. All of the LLMs can't identify carefully forged images and very realistic images generated by AI. In the area of tamper detection, LLMs still have a long way to go, particularly in reliably identifying highly sophisticated forgeries and AI-generated images that closely mimic reality."
                    },
                    {
                        "title": "Emotion-Conditioned Melody Harmonization with Hierarchical Variational Autoencoder",
                        "abstract": "Existing melody harmonization models have made great progress in improving the quality of generated harmonies, but most of them ignored the emotions beneath the music. Meanwhile, the variability of harmonies generated by previous methods is insufficient. To solve these problems, we propose a novel LSTM-based Hierarchical Variational Auto-Encoder (LHVAE) to investigate the influence of emotional conditions on melody harmonization, while improving the quality of generated harmonies and capturing the abundant variability of chord progressions. Specifically, LHVAE incorporates latent variables and emotional conditions at different levels (piece- and bar-level) to model the global and local music properties. Additionally, we introduce an attention-based melody context vector at each step to better learn the correspondence between melodies and harmonies. Objective experimental results show that our proposed model outperforms other LSTM-based models. Through subjective evaluation, we conclude that only altering the types of chords hardly changes the overall emotion of the music. The qualitative analysis demonstrates the ability of our model to generate variable harmonies."
                    },
                    {
                        "title": "Separate in the Speech Chain: Cross-Modal Conditional Audio-Visual Target Speech Extraction",
                        "abstract": "The integration of visual cues has revitalized the performance of the target speech extraction task, elevating it to the forefront of the field. Nevertheless, this multi-modal learning paradigm often encounters the challenge of modality imbalance. In audio-visual target speech extraction tasks, the audio modality tends to dominate, potentially overshadowing the importance of visual guidance. To tackle this issue, we propose AVSepChain, drawing inspiration from the speech chain concept. Our approach partitions the audio-visual target speech extraction task into two stages: speech perception and speech production. In the speech perception stage, audio serves as the dominant modality, while visual information acts as the conditional modality. Conversely, in the speech production stage, the roles are reversed. This transformation of modality status aims to alleviate the problem of modality imbalance. Additionally, we introduce a contrastive semantic matching loss to ensure that the semantic information conveyed by the generated speech aligns with the semantic information conveyed by lip movements during the speech production stage. Through extensive experiments conducted on multiple benchmark datasets for audio-visual target speech extraction, we showcase the superior performance achieved by our proposed method."
                    },
                    {
                        "title": "MusER: Musical Element-Based Regularization for Generating Symbolic Music with Emotion",
                        "abstract": "Generating music with emotion is an important task in automatic music generation, in which emotion is evoked through a variety of musical elements (such as pitch and duration) that change over time and collaborate with each other. However, prior research on deep learning-based emotional music generation has rarely explored the contribution of different musical elements to emotions, let alone the deliberate manipulation of these elements to alter the emotion of music, which is not conducive to fine-grained element-level control over emotions. To address this gap, we present a novel approach employing musical element-based regularization in the latent space to disentangle distinct elements, investigate their roles in distinguishing emotions, and further manipulate elements to alter musical emotions. Specifically, we propose a novel VQ-VAE-based model named MusER. MusER incorporates a regularization loss to enforce the correspondence between the musical element sequences and the specific dimensions of latent variable sequences, providing a new solution for disentangling discrete sequences. Taking advantage of the disentangled latent vectors, a two-level decoding strategy that includes multiple decoders attending to latent vectors with different semantics is devised to better predict the elements. By visualizing latent space, we conclude that MusER yields a disentangled and interpretable latent space and gain insights into the contribution of distinct elements to the emotional dimensions (i.e., arousal and valence). Experimental results demonstrate that MusER outperforms the state-of-the-art models for generating emotional music in both objective and subjective evaluation. Besides, we rearrange music through element transfer and attempt to alter the emotion of music by transferring emotion-distinguishable elements."
                    },
                    {
                        "title": "Rapid, User-Transparent, and Trustworthy Device Pairing for D2D-Enabled Mobile Crowdsourcing",
                        "abstract": "Mobile Crowdsourcing is a promising service paradigm utilizing ubiquitous mobile devices to facilitate largescale crowdsourcing tasks (e.g. urban sensing and collaborative computing). Many applications in this domain require Device-to-Device (D2D) communications between participating devices for interactive operations such as task collaborations and file transmissions. Considering the private participating devices and their opportunistic encountering behaviors, it is highly desired to establish secure and trustworthy D2D connections in a fast and autonomous way, which is vital for implementing practical Mobile Crowdsourcing Systems (MCSs). In this paper, we develop an efficient scheme, Trustworthy Device Pairing (TDP), which achieves user-transparent secure D2D connections and reliable peer device selections for trustworthy D2D communications. Through rigorous analysis, we demonstrate the effectiveness and security intensity of TDP in theory. The performance of TDP is evaluated based on both real-world prototype experiments and extensive trace-driven simulations. Evaluation results verify our theoretical analysis and show that TDP significantly outperforms existing approaches in terms of pairing speed, stability, and security."
                    },
                    {
                        "title": "Great Ape Detection in Challenging Jungle Camera Trap Footage via Attention-Based Spatial and Temporal Feature Blending",
                        "abstract": "We propose the first multi-frame video object detection framework trained to detect great apes. It is applicable to challenging camera trap footage in complex jungle environments and extends a traditional feature pyramid architecture by adding self-attention driven feature blending in both the spatial as well as the temporal domain. We demonstrate that this extension can detect distinctive species appearance and motion signatures despite significant partial occlusion. We evaluate the framework using 500 camera trap videos of great apes from the Pan African Programme containing 180K frames, which we manually annotated with accurate per-frame animal bounding boxes. These clips contain significant partial occlusions, challenging lighting, dynamic backgrounds, and natural camouflage effects. We show that our approach performs highly robustly and significantly outperforms frame-based detectors. We also perform detailed ablation studies and validation on the full ILSVRC 2015 VID data corpus to demonstrate wider applicability at adequate performance levels. We conclude that the framework is ready to assist human camera trap inspection efforts. We publish code, weights, and ground truth annotations with this paper."
                    },
                    {
                        "title": "A Multi-Stage Triple-Path Method for Speech Separation in Noisy and Reverberant Environments",
                        "abstract": "In noisy and reverberant environments, the performance of deep learning-based speech separation methods drops dramatically because previous methods are not designed and optimized for such situations. To address this issue, we propose a multi-stage end-to-end learning method that decouples the difficult speech separation problem in noisy and reverberant environments into three sub-problems: speech denoising, separation, and de-reverberation. The probability and speed of searching for the optimal solution of the speech separation model are improved by reducing the solution space. Moreover, since the channel information of the audio sequence in the time domain is crucial for speech separation, we propose a triple-path structure capable of modeling the channel dimension of audio sequences. Experimental results show that the proposed multi-stage triple-path method can improve the performance of speech separation models at the cost of little model parameter increment."
                    },
                    {
                        "title": "Self-Supervised Disentangled Representation Learning for Robust Target Speech Extraction",
                        "abstract": "Speech signals are inherently complex as they encompass both global acoustic characteristics and local semantic information. However, in the task of target speech extraction, certain elements of global and local semantic information in the reference speech, which are irrelevant to speaker identity, can lead to speaker confusion within the speech extraction network. To overcome this challenge, we propose a self-supervised disentangled representation learning method. Our approach tackles this issue through a two-phase process, utilizing a reference speech encoding network and a global information disentanglement network to gradually disentangle the speaker identity information from other irrelevant factors. We exclusively employ the disentangled speaker identity information to guide the speech extraction network. Moreover, we introduce the adaptive modulation Transformer to ensure that the acoustic representation of the mixed signal remains undisturbed by the speaker embeddings. This component incorporates speaker embeddings as conditional information, facilitating natural and efficient guidance for the speech extraction network. Experimental results substantiate the effectiveness of our meticulously crafted approach, showcasing a substantial reduction in the likelihood of speaker confusion."
                    }
                ]
            },
            "c5400084-c1d8-4ed2-943f-445b893b1add": {
                "pk": "c5400084-c1d8-4ed2-943f-445b893b1add",
                "name": "Chi-Man Pun",
                "collaborators": [
                    "Wenyun Li"
                ],
                "domain": [
                    "Deep Learning",
                    "Computer Vision",
                    "Object Detection",
                    "License Plate Recognition"
                ],
                "publications": [
                    {
                        "title": "A Single-Target License Plate Detection with Attention",
                        "abstract": "With the development of deep learning, Neural Network is commonly adopted to the License Plate Detection (LPD) task and achieves much better performance and precision, especially CNN-based networks can achieve state of the art RetinaNet[1]. For a single object detection task such as LPD, modified general object detection would be time-consuming, unable to cope with complex scenarios and a cumbersome weights file that is too hard to deploy on the embedded device."
                    }
                ]
            },
            "e0df870d-137d-4183-a0d2-46b9dce08a1e": {
                "pk": "e0df870d-137d-4183-a0d2-46b9dce08a1e",
                "name": "Jiancheng Lv",
                "collaborators": [
                    "Dayiheng Liu",
                    "Yuhao Zhou",
                    "Jie Fu",
                    "Qian Qu",
                    "Yusen Liu",
                    "Qing Ye",
                    "Kexin Yang",
                    "Yi Gao",
                    "Chenwei Tang",
                    "Chen Huang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Deep Learning",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "Learning Discriminative Representation with Signed Laplacian Restricted Boltzmann Machine",
                        "abstract": "We investigate the potential of a restricted Boltzmann Machine (RBM) for discriminative representation learning. By imposing the class information preservation constraints on the hidden layer of the RBM, we propose a Signed Laplacian Restricted Boltzmann Machine (SLRBM) for supervised discriminative representation learning. The model utilizes the label information and preserves the global data locality of data points simultaneously. Experimental results on the benchmark data set show the effectiveness of our method."
                    },
                    {
                        "title": "BFGAN: Backward and Forward Generative Adversarial Networks for Lexically Constrained Sentence Generation",
                        "abstract": "Incorporating prior knowledge like lexical constraints into the model's output to generate meaningful and coherent sentences has many applications in dialogue system, machine translation, image captioning, etc. However, existing RNN-based models incrementally generate sentences from left to right via beam search, which makes it difficult to directly introduce lexical constraints into the generated sentences. In this paper, we propose a new algorithmic framework, dubbed BFGAN, to address this challenge. Specifically, we employ a backward generator and a forward generator to generate lexically constrained sentences together, and use a discriminator to guide the joint training of two generators by assigning them reward signals. Due to the difficulty of BFGAN training, we propose several training techniques to make the training process more stable and efficient. Our extensive experiments on two large-scale datasets with human evaluation demonstrate that BFGAN has significant improvements over previous methods."
                    },
                    {
                        "title": "A Multi-Modal Chinese Poetry Generation Model",
                        "abstract": "Recent studies in sequence-to-sequence learning demonstrate that RNN encoder-decoder structure can successfully generate Chinese poetry. However, existing methods can only generate poetry with a given first line or user's intent theme. In this paper, we proposed a three-stage multi-modal Chinese poetry generation approach. Given a picture, the first line, the title and the other lines of the poem are successively generated in three stages. According to the characteristics of Chinese poems, we propose a hierarchy-attention seq2seq model which can effectively capture character, phrase, and sentence information between contexts and improve the symmetry delivered in poems. In addition, the Latent Dirichlet allocation (LDA) model is utilized for title generation and improve the relevance of the whole poem and the title. Compared with strong baseline, the experimental results demonstrate the effectiveness of our approach, using machine evaluations as well as human judgments."
                    },
                    {
                        "title": "TIGS: An Inference Algorithm for Text Infilling with Gradient Search",
                        "abstract": "Text infilling is defined as a task for filling in the missing part of a sentence or paragraph, which is suitable for many real-world natural language generation scenarios. However, given a well-trained sequential generative model, generating missing symbols conditioned on the context is challenging for existing greedy approximate inference algorithms. In this paper, we propose an iterative inference algorithm based on gradient search, which is the first inference algorithm that can be broadly applied to any neural sequence generative models for text infilling tasks. We compare the proposed method with strong baselines on three text infilling tasks with various mask ratios and different mask strategies. The results show that our proposed method is effective and efficient for fill-in-the-blank tasks, consistently outperforming all baselines."
                    },
                    {
                        "title": "Deep Poetry: A Chinese Classical Poetry Generation System",
                        "abstract": "In this work, we demonstrate a Chinese classical poetry generation system called Deep Poetry. Existing systems for Chinese classical poetry generation are mostly template-based and very few of them can accept multi-modal input. Unlike previous systems, Deep Poetry uses neural networks that are trained on over 200 thousand poems and 3 million ancient Chinese prose. Our system can accept plain text, images or artistic conceptions as inputs to generate Chinese classical poetry. More importantly, users are allowed to participate in the process of writing poetry by our system. For the user's convenience, we deploy the system at the WeChat applet platform, users can use the system on the mobile device whenever and wherever possible. The demo video of this paper is available at https://youtu.be/jD1R_u9TA3M."
                    },
                    {
                        "title": "Generating Chinese Poetry from Images via Concrete and Abstract Information",
                        "abstract": "In recent years, the automatic generation of classical Chinese poetry has made great progress. Besides focusing on improving the quality of the generated poetry, there is a new topic about generating poetry from an image. However, the existing methods for this topic still have the problem of topic drift and semantic inconsistency, and the image-poem pairs dataset is hard to be built when training these models. In this paper, we extract and integrate the Concrete and Abstract information from images to address those issues. We proposed an infilling-based Chinese poetry generation model which can infill the Concrete keywords into each line of poems in an explicit way, and an abstract information embedding to integrate the Abstract information into generated poems. In addition, we use non-parallel data during training and construct separate image datasets and poem datasets to train the different components in our framework. Both automatic and human evaluation results show that our approach can generate poems which have better consistency with images without losing the quality."
                    },
                    {
                        "title": "Let's be Humorous: Knowledge Enhanced Humor Generation",
                        "abstract": "The generation of humor is an under-explored and challenging problem. Previous works mainly utilize templates or replace phrases to generate humor. However, few works focus on freer forms and the background knowledge of humor. The linguistic theory of humor defines the structure of a humor sentence as set-up and punchline. In this paper, we explore how to generate a punchline given the set-up with the relevant knowledge. We propose a framework that can fuse the knowledge to end-to-end models. To our knowledge, this is the first attempt to generate punchlines with knowledge enhanced model. Furthermore, we create the first humor-knowledge dataset. The experimental results demonstrate that our method can make use of knowledge to generate fluent, funny punchlines, which outperforms several baselines."
                    },
                    {
                        "title": "HPSGD: Hierarchical Parallel SGD With Stale Gradients Featuring",
                        "abstract": "While distributed training significantly speeds up the training process of the deep neural network (DNN), the utilization of the cluster is relatively low due to the time-consuming data synchronizing between workers. To alleviate this problem, a novel Hierarchical Parallel SGD (HPSGD) strategy is proposed based on the observation that the data synchronization phase can be paralleled with the local training phase (i.e., Feed-forward and back-propagation). Furthermore, an improved model updating method is unitized to remedy the introduced stale gradients problem, which commits updates to the replica (i.e., a temporary model that has the same parameters as the global model) and then merges the average changes to the global model. Extensive experiments are conducted to demonstrate that the proposed HPSGD approach substantially boosts the distributed DNN training, reduces the disturbance of the stale gradients and achieves better accuracy in given fixed wall-time."
                    },
                    {
                        "title": "AnchiBERT: A Pre-Trained Model for Ancient ChineseLanguage Understanding and Generation",
                        "abstract": "Ancient Chinese is the essence of Chinese culture. There are several natural language processing tasks of ancient Chinese domain, such as ancient-modern Chinese translation, poem generation, and couplet generation. Previous studies usually use the supervised models which deeply rely on parallel data. However, it is difficult to obtain large-scale parallel data of ancient Chinese. In order to make full use of the more easily available monolingual ancient Chinese corpora, we release AnchiBERT, a pre-trained language model based on the architecture of BERT, which is trained on large-scale ancient Chinese corpora. We evaluate AnchiBERT on both language understanding and generation tasks, including poem classification, ancient-modern Chinese translation, poem generation, and couplet generation. The experimental results show that AnchiBERT outperforms BERT as well as the non-pretrained models and achieves state-of-the-art results in all cases."
                    },
                    {
                        "title": "Dynamic Normalization",
                        "abstract": "Batch Normalization has become one of the essential components in CNN. It allows the network to use a higher learning rate and speed up training. And the network doesn't need to be initialized carefully. However, in our work, we find that a simple extension of BN can increase the performance of the network. First, we extend BN to adaptively generate scale and shift parameters for each mini-batch data, called DN-C (Batch-shared and Channel-wise). We use the statistical characteristics of mini-batch data ($E[X], Std[X]\\in\\mathbb{R}^{c}$) as the input of SC module. Then we extend BN to adaptively generate scale and shift parameters for each channel of each sample, called DN-B (Batch and Channel-wise). Our experiments show that DN-C model can't train normally, but DN-B model has very good robustness. In classification task, DN-B can improve the accuracy of the MobileNetV2 on ImageNet-100 more than 2% with only 0.6% additional Mult-Adds. In detection task, DN-B can improve the accuracy of the SSDLite on MS-COCO nearly 4% mAP with the same settings. Compared with BN, DN-B has stable performance when using higher learning rate or smaller batch size."
                    },
                    {
                        "title": "Face Deblurring Based on Separable Normalization and Adaptive Denormalization",
                        "abstract": "Face deblurring aims to restore a clear face image from a blurred input image with more explicit structure and facial details. However, most conventional image and face deblurring methods focus on the whole generated image resolution without consideration of special face part texture and generally produce unsufficient details. Considering that faces and backgrounds have different distribution information, in this study, we designed an effective face deblurring network based on separable normalization and adaptive denormalization (SNADNet). First, We fine-tuned the face parsing network to obtain an accurate face structure. Then, we divided the face parsing feature into face foreground and background. Moreover, we constructed a new feature adaptive denormalization to regularize fafcial structures as a condition of the auxiliary to generate more harmonious and undistorted face structure. In addition, we proposed a texture extractor and multi-patch discriminator to enhance the generated facial texture information. Experimental results on both CelebA and CelebA-HQ datasets demonstrate that the proposed face deblurring network restores face structure with more facial details and performs favorably against state-of-the-art methods in terms of structured similarity indexing method (SSIM), peak signal-to-noise ratio (PSNR), Frechet inception distance (FID) and L1, and qualitative comparisons."
                    },
                    {
                        "title": "Cluster-based Contrastive Disentangling for Generalized Zero-Shot Learning",
                        "abstract": "Generalized Zero-Shot Learning (GZSL) aims to recognize both seen and unseen classes by training only the seen classes, in which the instances of unseen classes tend to be biased towards the seen class. In this paper, we propose a Cluster-based Contrastive Disentangling (CCD) method to improve GZSL by alleviating the semantic gap and domain shift problems. Specifically, we first cluster the batch data to form several sets containing similar classes. Then, we disentangle the visual features into semantic-unspecific and semantic-matched variables, and further disentangle the semantic-matched variables into class-shared and class-unique variables according to the clustering results. The disentangled learning module with random swapping and semantic-visual alignment bridges the semantic gap. Moreover, we introduce contrastive learning on semantic-matched and class-unique variables to learn high intra-set and intra-class similarity, as well as inter-set and inter-class discriminability. Then, the generated visual features conform to the underlying characteristics of general images and have strong discriminative information, which alleviates the domain shift problem well. We evaluate our proposed method on four datasets and achieve state-of-the-art results in both conventional and generalized settings."
                    },
                    {
                        "title": "Towards Equipping Transformer with the Ability of Systematic Compositionality",
                        "abstract": "One of the key factors in language productivity and human cognition is the ability of systematic compositionality, which refers to understanding composed unseen examples of seen primitives. However, recent evidence reveals that the Transformers have difficulty generalizing the composed context based on the seen primitives. To this end, we take the first step to propose a compositionality-aware Transformer called CAT and two novel pre-training tasks to facilitate systematic compositionality. We tentatively provide a successful implementation of a multi-layer CAT on the basis of the especially popular BERT. The experimental results demonstrate that CAT outperforms baselines on compositionality-aware tasks with minimal impact on the effectiveness on standardized language understanding tasks."
                    },
                    {
                        "title": "Deep Learning-Based Automatic Downbeat Tracking: A Brief Review",
                        "abstract": "As an important format of multimedia, music has filled almost everyone's life. Automatic analyzing music is a significant step to satisfy people's need for music retrieval and music recommendation in an effortless way. Thereinto, downbeat tracking has been a fundamental and continuous problem in Music Information Retrieval (MIR) area. Despite significant research efforts, downbeat tracking still remains a challenge. Previous researches either focus on feature engineering (extracting certain features by signal processing, which are semi-automatic solutions); or have some limitations: they can only model music audio recordings within limited time signatures and tempo ranges. Recently, deep learning has surpassed traditional machine learning methods and has become the primary algorithm in feature learning; the combination of traditional and deep learning methods also has made better performance. In this paper, we begin with a background introduction of downbeat tracking problem. Then, we give detailed discussions of the following topics: system architecture, feature extraction, deep neural network algorithms, datasets, and evaluation strategy. In addition, we take a look at the results from the annual benchmark evaluation--Music Information Retrieval Evaluation eXchange (MIREX)--as well as the developments in software implementations. Although much has been achieved in the area of automatic downbeat tracking, some problems still remain. We point out these problems and conclude with possible directions and challenges for future research."
                    },
                    {
                        "title": "Communication-Efficient Federated Learning with Compensated Overlap-FedAvg",
                        "abstract": "Petabytes of data are generated each day by emerging Internet of Things (IoT), but only few of them can be finally collected and used for Machine Learning (ML) purposes due to the apprehension of data & privacy leakage, which seriously retarding ML's growth. To alleviate this problem, Federated learning is proposed to perform model training by multiple clients' combined data without the dataset sharing within the cluster. Nevertheless, federated learning introduces massive communication overhead as the synchronized data in each epoch is of the same size as the model, and thereby leading to a low communication efficiency. Consequently, variant methods mainly focusing on the communication rounds reduction and data compression are proposed to reduce the communication overhead of federated learning. In this paper, we propose Overlap-FedAvg, a framework that parallels the model training phase with model uploading & downloading phase, so that the latter phase can be totally covered by the former phase. Compared to vanilla FedAvg, Overlap-FedAvg is further developed with a hierarchical computing strategy, a data compensation mechanism and a nesterov accelerated gradients~(NAG) algorithm. Besides, Overlap-FedAvg is orthogonal to many other compression methods so that they can be applied together to maximize the utilization of the cluster. Furthermore, the theoretical analysis is provided to prove the convergence of the proposed Overlap-FedAvg framework. Extensive experiments on both conventional and recurrent tasks with multiple models and datasets also demonstrate that the proposed Overlap-FedAvg framework substantially boosts the federated learning process."
                    },
                    {
                        "title": "Partial Differential Equations Meet Deep Neural Networks: A Survey",
                        "abstract": "Many problems in science and engineering can be represented by a set of partial differential equations (PDEs) through mathematical modeling. Mechanism-based computation following PDEs has long been an essential paradigm for studying topics such as computational fluid dynamics, multiphysics simulation, molecular dynamics, or even dynamical systems. It is a vibrant multi-disciplinary field of increasing importance and with extraordinary potential. At the same time, solving PDEs efficiently has been a long-standing challenge. Generally, except for a few differential equations for which analytical solutions are directly available, many more equations must rely on numerical approaches such as the finite difference method, finite element method, finite volume method, and boundary element method to be solved approximately. These numerical methods usually divide a continuous problem domain into discrete points and then concentrate on solving the system at each of those points. Though the effectiveness of these traditional numerical methods, the vast number of iterative operations accompanying each step forward significantly reduces the efficiency. Recently, another equally important paradigm, data-based computation represented by deep learning, has emerged as an effective means of solving PDEs. Surprisingly, a comprehensive review for this interesting subfield is still lacking. This survey aims to categorize and review the current progress on Deep Neural Networks (DNNs) for PDEs. We discuss the literature published in this subfield over the past decades and present them in a common taxonomy, followed by an overview and classification of applications of these related methods in scientific research and engineering scenarios. The origin, developing history, character, sort, as well as the future trends in each potential direction of this subfield are also introduced."
                    },
                    {
                        "title": "Personalized Federated Learning with Hidden Information on Personalized Prior",
                        "abstract": "Federated learning (FL for simplification) is a distributed machine learning technique that utilizes global servers and collaborative clients to achieve privacy-preserving global model training without direct data sharing. However, heterogeneous data problem, as one of FL's main problems, makes it difficult for the global model to perform effectively on each client's local data. Thus, personalized federated learning (PFL for simplification) aims to improve the performance of the model on local data as much as possible. Bayesian learning, where the parameters of the model are seen as random variables with a prior assumption, is a feasible solution to the heterogeneous data problem due to the tendency that the more local data the model use, the more it focuses on the local data, otherwise focuses on the prior. When Bayesian learning is applied to PFL, the global model provides global knowledge as a prior to the local training process. In this paper, we employ Bayesian learning to model PFL by assuming a prior in the scaled exponential family, and therefore propose pFedBreD, a framework to solve the problem we model using Bregman divergence regularization. Empirically, our experiments show that, under the prior assumption of the spherical Gaussian and the first order strategy of mean selection, our proposal significantly outcompetes other PFL algorithms on multiple public benchmarks."
                    },
                    {
                        "title": "DREditor: An Time-efficient Approach for Building a Domain-specific Dense Retrieval Model",
                        "abstract": "Deploying dense retrieval models efficiently is becoming increasingly important across various industries. This is especially true for enterprise search services, where customizing search engines to meet the time demands of different enterprises in different domains is crucial. Motivated by this, we develop a time-efficient approach called DREditor to edit the matching rule of an off-the-shelf dense retrieval model to suit a specific domain. This is achieved by directly calibrating the output embeddings of the model using an efficient and effective linear mapping. This mapping is powered by an edit operator that is obtained by solving a specially constructed least squares problem. Compared to implicit rule modification via long-time finetuning, our experimental results show that DREditor provides significant advantages on different domain-specific datasets, dataset sources, retrieval models, and computing devices. It consistently enhances time efficiency by 100-300 times while maintaining comparable or even superior retrieval performance. In a broader context, we take the first step to introduce a novel embedding calibration approach for the retrieval task, filling the technical blank in the current field of embedding calibration. This approach also paves the way for building domain-specific dense retrieval models efficiently and inexpensively."
                    },
                    {
                        "title": "Ancient-Modern Chinese Translation with a Large Training Dataset",
                        "abstract": "Ancient Chinese brings the wisdom and spirit culture of the Chinese nation. Automatic translation from ancient Chinese to modern Chinese helps to inherit and carry forward the quintessence of the ancients. However, the lack of large-scale parallel corpus limits the study of machine translation in Ancient-Modern Chinese. In this paper, we propose an Ancient-Modern Chinese clause alignment approach based on the characteristics of these two languages. This method combines both lexical-based information and statistical-based information, which achieves 94.2 F1-score on our manual annotation Test set. We use this method to create a new large-scale Ancient-Modern Chinese parallel corpus which contains 1.24M bilingual pairs. To our best knowledge, this is the first large high-quality Ancient-Modern Chinese dataset. Furthermore, we analyzed and compared the performance of the SMT and various NMT models on this dataset and provided a strong baseline for this task."
                    }
                ]
            },
            "e6673a89-9e1a-4666-8320-5690a5d80a57": {
                "pk": "e6673a89-9e1a-4666-8320-5690a5d80a57",
                "name": "Jizhe Zhou",
                "collaborators": [
                    "Chi-Man Pun",
                    "Zhuohang Jiang",
                    "Xiaochen Ma",
                    "Xinyu Yang",
                    "Bingkui Tong",
                    "Xia Du",
                    "Yu Tong",
                    "Ahmed Alhammadi",
                    "Xuekang Zhu",
                    "Bo Du"
                ],
                "domain": [
                    "Image Manipulation",
                    "Privacy Protection",
                    "Deep Learning",
                    "Multimedia Forensics"
                ],
                "publications": [
                    {
                        "title": "Manipulation Mask Generator: High-Quality Image Manipulation Mask Generation Method Based on Modified Total Variation Noise Reduction",
                        "abstract": "In artificial intelligence, any model that wants to achieve a good result is inseparable from a large number of high-quality data. It is especially true in the field of tamper detection. This paper proposes a modified total variation noise reduction method to acquire high-quality tampered images. We automatically crawl original and tampered images from the Baidu PS Bar. Baidu PS Bar is a website where net friends post countless tampered images. Subtracting the original image with the tampered image can highlight the tampered area. However, there is also substantial noise on the final print, so these images can't be directly used in the deep learning model. Our modified total variation noise reduction method is aimed at solving this problem. Because a lot of text is slender, it is easy to lose text information after the opening and closing operation. We use MSER (Maximally Stable Extremal Regions) and NMS (Non-maximum Suppression) technology to extract text information. And then use the modified total variation noise reduction technology to process the subtracted image. Finally, we can obtain an image with little noise by adding the image and text information. And the idea also largely retains the text information. Datasets generated in this way can be used in deep learning models, and they will help the model achieve better results."
                    },
                    {
                        "title": "Research about the Ability of LLM in the Tamper-Detection Area",
                        "abstract": "In recent years, particularly since the early 2020s, Large Language Models (LLMs) have emerged as the most powerful AI tools in addressing a diverse range of challenges, from natural language processing to complex problem-solving in various domains. In the field of tamper detection, LLMs are capable of identifying basic tampering activities.To assess the capabilities of LLMs in more specialized domains, we have collected five different LLMs developed by various companies: GPT-4, LLaMA, Bard, ERNIE Bot 4.0, and Tongyi Qianwen. This diverse range of models allows for a comprehensive evaluation of their performance in detecting sophisticated tampering instances.We devised two domains of detection: AI-Generated Content (AIGC) detection and manipulation detection. AIGC detection aims to test the ability to distinguish whether an image is real or AI-generated. Manipulation detection, on the other hand, focuses on identifying tampered images. According to our experiments, most LLMs can identify composite pictures that are inconsistent with logic, and only more powerful LLMs can distinguish logical, but visible signs of tampering to the human eye. All of the LLMs can't identify carefully forged images and very realistic images generated by AI. In the area of tamper detection, LLMs still have a long way to go, particularly in reliably identifying highly sophisticated forgeries and AI-generated images that closely mimic reality."
                    },
                    {
                        "title": "Adaptive Offline and Online Similarity-Based Caching",
                        "abstract": "With similarity-based content delivery, the request for a content can be satisfied by delivering a related content under a dissimilarity cost. This letter addresses the joint optimization of caching and similarity-based delivery decisions across a network so as to minimize the weighted sum of average delay and dissimilarity cost. A convergent alternate gradient descent ascent algorithm is first introduced for an offline scenario with prior knowledge of the request rates, and then extended to an online setting. Numerical results validate the advantages of the approach with respect to standard per-cache solutions."
                    },
                    {
                        "title": "Generating Adversarial Perturbation with Root Mean Square Gradient",
                        "abstract": "We focus our attention on the problem of generating adversarial perturbations based on the gradient in image classification domain"
                    },
                    {
                        "title": "Pixelation is NOT Done in Videos Yet",
                        "abstract": "This paper introduces an algorithm to protect the privacy of individuals in streaming video data by blurring faces such that face cannot be reliably recognized. This thwarts any possible face recognition, but because all facial details are obscured, the result is of limited use. We propose a new clustering algorithm to create raw trajectories for detected faces. Associating faces across frames to form trajectories, it auto-generates cluster number and discovers new clusters through deep feature and position aggregated affinities. We introduce a Gaussian Process to refine the raw trajectories. We conducted an online experiment with 47 participants to evaluate the effectiveness of face blurring compared to the original photo (as-is), and users' experience (satisfaction, information sufficiency, enjoyment, social presence, and filter likeability)"
                    },
                    {
                        "title": "Privacy-sensitive Objects Pixelation for Live Video Streaming",
                        "abstract": "With the prevailing of live video streaming, establishing an online pixelation method for privacy-sensitive objects is an urgency. Caused by the inaccurate detection of privacy-sensitive objects, simply migrating the tracking-by-detection structure into the online form will incur problems in target initialization, drifting, and over-pixelation. To cope with the inevitable but impacting detection issue, we propose a novel Privacy-sensitive Objects Pixelation (PsOP) framework for automatic personal privacy filtering during live video streaming. Leveraging pre-trained detection networks, our PsOP is extendable to any potential privacy-sensitive objects pixelation. Employing the embedding networks and the proposed Positioned Incremental Affinity Propagation (PIAP) clustering algorithm as the backbone, our PsOP unifies the pixelation of discriminating and indiscriminating pixelation objects through trajectories generation. In addition to the pixelation accuracy boosting, experiments on the streaming video data we built show that the proposed PsOP can significantly reduce the over-pixelation ratio in privacy-sensitive object pixelation."
                    },
                    {
                        "title": "News Image Steganography: A Novel Architecture Facilitates the Fake News Identification",
                        "abstract": "A larger portion of fake news quotes untampered images from other sources with ulterior motives rather than conducting image forgery. Such elaborate engraftments keep the inconsistency between images and text reports stealthy, thereby, palm off the spurious for the genuine. This paper proposes an architecture named News Image Steganography (NIS) to reveal the aforementioned inconsistency through image steganography based on GAN. Extractive summarization about a news image is generated based on its source texts, and a learned steganographic algorithm encodes and decodes the summarization of the image in a manner that approaches perceptual invisibility. Once an encoded image is quoted, its source summarization can be decoded and further presented as the ground truth to verify the quoting news. The pairwise encoder and decoder endow images of the capability to carry along their imperceptible summarization. Our NIS reveals the underlying inconsistency, thereby, according to our experiments and investigations, contributes to the identification accuracy of fake news that engrafts untampered images."
                    },
                    {
                        "title": "Personal Privacy Protection via Irrelevant Faces Tracking and Pixelation in Video Live Streaming",
                        "abstract": "To date, the privacy-protection intended pixelation tasks are still labor-intensive and yet to be studied. With the prevailing of video live streaming, establishing an online face pixelation mechanism during streaming is an urgency. In this paper, we develop a new method called Face Pixelation in Video Live Streaming (FPVLS) to generate automatic personal privacy filtering during unconstrained streaming activities. Simply applying multi-face trackers will encounter problems in target drifting, computing efficiency, and over-pixelation. Therefore, for fast and accurate pixelation of irrelevant people's faces, FPVLS is organized in a frame-to-video structure of two core stages. On individual frames, FPVLS utilizes image-based face detection and embedding networks to yield face vectors. In the raw trajectories generation stage, the proposed Positioned Incremental Affinity Propagation (PIAP) clustering algorithm leverages face vectors and positioned information to quickly associate the same person's faces across frames. Such frame-wise accumulated raw trajectories are likely to be intermittent and unreliable on video level. Hence, we further introduce the trajectory refinement stage that merges a proposal network with the two-sample test based on the Empirical Likelihood Ratio (ELR) statistic to refine the raw trajectories. A Gaussian filter is laid on the refined trajectories for final pixelation. On the video live streaming dataset we collected, FPVLS obtains satisfying accuracy, real-time efficiency, and contains the over-pixelation problems."
                    },
                    {
                        "title": "Dynamic Routing for Traffic Flow through Multi-agent Systems",
                        "abstract": "Routing strategies for traffics and vehicles have been historically studied. However, in the absence of considering drivers' preferences, current route planning algorithms are developed under ideal situations where all drivers are expected to behave rationally and properly. Especially, for jumbled urban road networks, drivers' actual routing strategies deteriorated to a series of empirical and selfish decisions that result in congestion. Self-evidently, if minimum mobility can be kept, traffic congestion is avoidable by traffic load dispersing. In this paper, we establish a novel dynamic routing method catering drivers' preferences and retaining maximum traffic mobility simultaneously through multi-agent systems (MAS). Modeling human-drivers' behavior through agents' dynamics, MAS can analyze the global behavior of the entire traffic flow. Therefore, regarding agents as particles in smoothed particles hydrodynamics (SPH), we can enforce the traffic flow to behave like a real flow. Thereby, with the characteristic of distributing itself uniformly in road networks, our dynamic routing method realizes traffic load balancing without violating the individual time-saving motivation. Moreover, as a discrete control mechanism, our method is robust to chaos meaning driver's disobedience can be tolerated. As controlled by SPH based density, the only intelligent transportation system (ITS) we require is the location-based service (LBS). A mathematical proof is accomplished to scrutinize the stability of the proposed control law. Also, multiple testing cases are built to verify the effectiveness of the proposed dynamic routing algorithm."
                    },
                    {
                        "title": "SHAN: Object-Level Privacy Detection via Inference on Scene Heterogeneous Graph",
                        "abstract": "With the rise of social platforms, protecting privacy has become an important issue. Privacy object detection aims to accurately locate private objects in images. It is the foundation of safeguarding individuals' privacy rights and ensuring responsible data handling practices in the digital age. Since privacy of object is not shift-invariant, the essence of the privacy object detection task is inferring object privacy based on scene information. However, privacy object detection has long been studied as a subproblem of common object detection tasks. Therefore, existing methods suffer from serious deficiencies in accuracy, generalization, and interpretability. Moreover, creating large-scale privacy datasets is difficult due to legal constraints and existing privacy datasets lack label granularity. The granularity of existing privacy detection methods remains limited to the image level. To address the above two issues, we introduce two benchmark datasets for object-level privacy detection and propose SHAN, Scene Heterogeneous graph Attention Network, a model constructs a scene heterogeneous graph from an image and utilizes self-attention mechanisms for scene inference to obtain object privacy. Through experiments, we demonstrated that SHAN performs excellently in privacy object detection tasks, with all metrics surpassing those of the baseline model."
                    },
                    {
                        "title": "Beyond Visual Appearances: Privacy-sensitive Objects Identification via Hybrid Graph Reasoning",
                        "abstract": "The Privacy-sensitive Object Identification (POI) task allocates bounding boxes for privacy-sensitive objects in a scene. The key to POI is settling an object's privacy class (privacy-sensitive or non-privacy-sensitive). In contrast to conventional object classes which are determined by the visual appearance of an object, one object's privacy class is derived from the scene contexts and is subject to various implicit factors beyond its visual appearance. That is, visually similar objects may be totally opposite in their privacy classes. To explicitly derive the objects' privacy class from the scene contexts, in this paper, we interpret the POI task as a visual reasoning task aimed at the privacy of each object in the scene. Following this interpretation, we propose the PrivacyGuard framework for POI. PrivacyGuard contains three stages. i) Structuring: an unstructured image is first converted into a structured, heterogeneous scene graph that embeds rich scene contexts. ii) Data Augmentation: a contextual perturbation oversampling strategy is proposed to create slightly perturbed privacy-sensitive objects in a scene graph, thereby balancing the skewed distribution of privacy classes. iii) Hybrid Graph Generation & Reasoning: the balanced, heterogeneous scene graph is then transformed into a hybrid graph by endowing it with extra \"node-node\" and \"edge-edge\" homogeneous paths. These homogeneous paths allow direct message passing between nodes or edges, thereby accelerating reasoning and facilitating the capturing of subtle context changes. Based on this hybrid graph... **For the full abstract, see the original paper.**"
                    },
                    {
                        "title": "Saliency Guided Optimization of Diffusion Latents",
                        "abstract": "With the rapid advances in diffusion models, generating decent images from text prompts is no longer challenging. The key to text-to-image generation is how to optimize the results of a text-to-image generation model so that they can be better aligned with human intentions or prompts. Existing optimization methods commonly treat the entire image uniformly and conduct global optimization. These methods overlook the fact that when viewing an image, the human visual system naturally prioritizes attention toward salient areas, often neglecting less or non-salient regions. That is, humans are likely to neglect optimizations in non-salient areas. Consequently, although model retaining is conducted under the guidance of additional large and multimodality models, existing methods, which perform uniform optimizations, yield sub-optimal results. To address this alignment challenge effectively and efficiently, we propose Saliency Guided Optimization Of Diffusion Latents (SGOOL). We first employ a saliency detector to mimic the human visual attention system and mark out the salient regions. To avoid retraining an additional model, our method directly optimizes the diffusion latents. Besides, SGOOL utilizes an invertible diffusion process and endows it with the merits of constant memory implementation. Hence, our method becomes a parameter-efficient and plug-and-play fine-tuning method. Extensive experiments have been done with several metrics and human evaluation. Experimental results demonstrate the superiority of SGOOL in image quality and prompt alignment."
                    },
                    {
                        "title": "FIF-UNet: An Efficient UNet Using Feature Interaction and Fusion for Medical Image Segmentation",
                        "abstract": "Nowadays, pre-trained encoders are widely used in medical image segmentation because of their ability to capture complex feature representations. However, the existing models fail to effectively utilize the rich features obtained by the pre-trained encoder, resulting in suboptimal segmentation results. In this work, a novel U-shaped model, called FIF-UNet, is proposed to address the above issue, including three plug-and-play modules. A channel spatial interaction module (CSI) is proposed to obtain informative features by establishing the interaction between encoder stages and corresponding decoder stages. A cascaded conv-SE module (CoSE) is designed to enhance the representation of critical features by adaptively assigning importance weights on different feature channels. A multi-level fusion module (MLF) is proposed to fuse the multi-scale features from the decoder stages, ensuring accurate and robust final segmentation. Comprehensive experiments on the Synapse and ACDC datasets demonstrate that the proposed FIF-UNet outperforms existing state-of-the-art methods, which achieves the highest average DICE of 86.05% and 92.58%, respectively."
                    },
                    {
                        "title": "IML-ViT: Benchmarking Image Manipulation Localization by Vision Transformer",
                        "abstract": "Advanced image tampering techniques are increasingly challenging the trustworthiness of multimedia, leading to the development of Image Manipulation Localization (IML). But what makes a good IML model? The answer lies in the way to capture artifacts. Exploiting artifacts requires the model to extract non-semantic discrepancies between manipulated and authentic regions, necessitating explicit comparisons between the two areas. With the self-attention mechanism, naturally, the Transformer should be a better candidate to capture artifacts. However, due to limited datasets, there is currently no pure ViT-based approach for IML to serve as a benchmark, and CNNs dominate the entire task. Nevertheless, CNNs suffer from weak long-range and non-semantic modeling. To bridge this gap, based on the fact that artifacts are sensitive to image resolution, amplified under multi-scale features, and massive at the manipulation border, we formulate the answer to the former question as building a ViT with high-resolution capacity, multi-scale feature extraction capability, and manipulation edge supervision that could converge with a small amount of data. We term this simple but effective ViT paradigm IML-ViT, which has significant potential to become a new benchmark for IML. Extensive experiments on five benchmark datasets verified our model outperforms the state-of-the-art manipulation localization methods.Code and models are available at \\url{https://github.com/SunnyHaze/IML-ViT}."
                    },
                    {
                        "title": "Pre-training-free Image Manipulation Localization through Non-Mutually Exclusive Contrastive Learning",
                        "abstract": "Deep Image Manipulation Localization (IML) models suffer from training data insufficiency and thus heavily rely on pre-training. We argue that contrastive learning is more suitable to tackle the data insufficiency problem for IML. Crafting mutually exclusive positives and negatives is the prerequisite for contrastive learning. However, when adopting contrastive learning in IML, we encounter three categories of image patches: tampered, authentic, and contour patches. Tampered and authentic patches are naturally mutually exclusive, but contour patches containing both tampered and authentic pixels are non-mutually exclusive to them. Simply abnegating these contour patches results in a drastic performance loss since contour patches are decisive to the learning outcomes. Hence, we propose the Non-mutually exclusive Contrastive Learning (NCL) framework to rescue conventional contrastive learning from the above dilemma. In NCL, to cope with the non-mutually exclusivity, we first establish a pivot structure with dual branches to constantly switch the role of contour patches between positives and negatives while training. Then, we devise a pivot-consistent loss to avoid spatial corruption caused by the role-switching process. In this manner, NCL both inherits the self-supervised merits to address the data insufficiency and retains a high manipulation localization accuracy. Extensive experiments verify that our NCL achieves state-of-the-art performance on all five benchmarks without any pre-training and is more robust on unseen real-life samples. The code is available at: https://github.com/Knightzjz/NCL-IML."
                    },
                    {
                        "title": "Perceptual MAE for Image Manipulation Localization: A High-level Vision Learner Focusing on Low-level Features",
                        "abstract": "Nowadays, multimedia forensics faces unprecedented challenges due to the rapid advancement of multimedia generation technology thereby making Image Manipulation Localization (IML) crucial in the pursuit of truth. The key to IML lies in revealing the artifacts or inconsistencies between the tampered and authentic areas, which are evident under pixel-level features. Consequently, existing studies treat IML as a low-level vision task, focusing on allocating tampered masks by crafting pixel-level features such as image RGB noises, edge signals, or high-frequency features. However, in practice, tampering commonly occurs at the object level, and different classes of objects have varying likelihoods of becoming targets of tampering. Therefore, object semantics are also vital in identifying the tampered areas in addition to pixel-level features. This necessitates IML models to carry out a semantic understanding of the entire image. In this paper, we reformulate the IML task as a high-level vision task that greatly benefits from low-level features. Based on such an interpretation, we propose a method to enhance the Masked Autoencoder (MAE) by incorporating high-resolution inputs and a perceptual loss supervision module, which is termed Perceptual MAE (PMAE). While MAE has demonstrated an impressive understanding of object semantics, PMAE can also compensate for low-level semantics with our proposed enhancements. Evidenced by extensive experiments, this paradigm effectively unites the low-level and high-level features of the IML task and outperforms state-of-the-art tampering localization methods on all five publicly available datasets."
                    },
                    {
                        "title": "M^3:Manipulation Mask Manufacturer for Arbitrary-Scale Super-Resolution Mask",
                        "abstract": "In the field of image manipulation localization (IML), the small quantity and poor quality of existing datasets have always been major issues. A dataset containing various types of manipulations will greatly help improve the accuracy of IML models. Images on the internet (such as those on Baidu Tieba's PS Bar) are manipulated using various techniques, and creating a dataset from these images will significantly enrich the types of manipulations in our data. However, images on the internet suffer from resolution and clarity issues, and the masks obtained by simply subtracting the manipulated image from the original contain various noises. These noises are difficult to remove, rendering the masks unusable for IML models. Inspired by the field of change detection, we treat the original and manipulated images as changes over time for the same image and view the data generation task as a change detection task. However, due to clarity issues between images, conventional change detection models perform poorly. Therefore, we introduced a super-resolution module and proposed the Manipulation Mask Manufacturer (MMM) framework. It enhances the resolution of both the original and tampered images, thereby improving image details for better comparison. Simultaneously, the framework converts the original and tampered images into feature embeddings and concatenates them, effectively modeling the context. Additionally, we created the Manipulation Mask Manufacturer Dataset (MMMD), a dataset that covers a wide range of manipulation techniques. We aim to contribute to the fields of image forensics and manipulation detection by providing more realistic manipulation data through MMM and MMMD. Detailed information about MMMD and the download link can be found at: the code and datasets will be made available."
                    },
                    {
                        "title": "Analysis and Applications of Deep Learning with Finite Samples in Full Life-Cycle Intelligence of Nuclear Power Generation",
                        "abstract": "The advent of Industry 4.0 has precipitated the incorporation of Artificial Intelligence (AI) methods within industrial contexts, aiming to realize intelligent manufacturing, operation as well as maintenance, also known as industrial intelligence. However, intricate industrial milieus, particularly those relating to energy exploration and production, frequently encompass data characterized by long-tailed class distribution, sample imbalance, and domain shift. These attributes pose noteworthy challenges to data-centric Deep Learning (DL) techniques, crucial for the realization of industrial intelligence. The present study centers on the intricate and distinctive industrial scenarios of Nuclear Power Generation (NPG), meticulously scrutinizing the application of DL techniques under the constraints of finite data samples. Initially, the paper expounds on potential employment scenarios for AI across the full life-cycle of NPG. Subsequently, we delve into an evaluative exposition of DL's advancement, grounded in the finite sample perspective. This encompasses aspects such as small-sample learning, few-shot learning, zero-shot learning, and open-set recognition, also referring to the unique data characteristics of NPG. The paper then proceeds to present two specific case studies. The first revolves around the automatic recognition of zirconium alloy metallography, while the second pertains to open-set recognition for signal diagnosis of machinery sensors. These cases, spanning the entirety of NPG's life-cycle, are accompanied by constructive outcomes and insightful deliberations. By exploring and applying DL methodologies within the constraints of finite sample availability, this paper not only furnishes a robust technical foundation but also introduces a fresh perspective toward the secure and efficient advancement and exploitation of this advanced energy source."
                    }
                ]
            }
        }
    },
    "2405.15706": {
        "paper_data": {
            "title": "The Impact of Geometric Complexity on Neural Collapse in Transfer Learning",
            "url": "http://arxiv.org/abs/2405.15706v2",
            "arxiv_id": "2405.15706",
            "authors": [
                "Michael Munn",
                "Benoit Dherin",
                "Javier Gonzalvo"
            ],
            "abstract": "Many of the recent remarkable advances in computer vision and language models can be attributed to the success of transfer learning via the pre-training of large foundation models. However, a theoretical framework which explains this empirical success is incomplete and remains an active area of research. Flatness of the loss surface and neural collapse have recently emerged as useful pre-training metrics which shed light on the implicit biases underlying pre-training. In this paper, we explore the geometric complexity of a model's learned representations as a fundamental mechanism that relates these two concepts. We show through experiments and theory that mechanisms which affect the geometric complexity of the pre-trained network also influence the neural collapse. Furthermore, we show how this effect of the geometric complexity generalizes to the neural collapse of new classes as well, thus encouraging better performance on downstream tasks, particularly in the few-shot setting.",
            "introduction": "   1 Introduction  Many of the recent remarkable advances in modern machine learning owe their success in part to transfer learning [15, 28, 6, 9, 4, 48, 62, 47]. While there are different approaches to this technique, the standard one involves two stages. In a first stage, called pre-training, ones trains a deep neural network on a general, large-scale dataset in the form of a supervised or unsupervised source task; e.g., ImageNet or CIFAR-100 [14, 32] for image models or the Common Crawl, C4 or LM1B datasets [8, 11, 47] for language models. In the second stage, one then leverages portions of the pre-trained network to use as features map or embeddings that can then be adapted, or fine-tuned, on a more specific target task. Often these target tasks are unknown at the time of pre-training and labeled data may be very scarce, such as in the context of few-shot learning [57, 33, 49]. However, despite these limitations, this approach often results in a fine-tuned model that achieves quite impressive performance substantially better than training on the target task alone and requiring less computational resources\u00a0[20]. Despite this empirical success, a comprehensive theoretical understanding of the mechanisms underlying this effectiveness of transfer learning is not fully understood and remains an active area of research\u00a0[60].   One interesting line of research suggests that the effectiveness of transfer learning is due to the implicit biases encoded in pre-trained models [22, 35, 46]. These implicit biases effectively constrain the hypothesis space, guiding the model towards solutions that have a preference for smoother functions [12, 39, 51, 37, 17], simpler geometry in the loss surface [3, 52, 18, 13, 55] or reduced complexity of the internal learned representations [23, 31]. In the same way that these implicit biases have been used to help explain the success of deep learning, recent work has shown that the notion of neural collapse [22, 59, 34] and flatness of the loss surface [35] can also inform the mechanisms behind transfer learning. This suggests that these preferences during pre-training also act as a form of prior knowledge which is highly transferable to downstream tasks, even with limited task-specific data.   In this paper, we present a novel perspective that further sheds light on these mechanisms and implicit biases hidden within transfer learning. Our approach analyzes the geometric complexity [16, 17] of the internal representations learned by the deep neural networks during pre-training and provides a complementary theoretical framework which unifies previous work examining the role of neural collapse and loss surface flatness in transfer learning [22, 35]. In particular, we show through experiments and theory that the geometric complexity of the pre-trained network directly controls the neural collapse of the pre-trained model and thus its efficacy in transfer learning, particularly in the few-shot setting. We argue that geometric complexity (similarly to flatness of the loss surface and neural collapse) can be used as hidden progress metrics, cf. [2], for transfer learning, serving as an informative proxy toward the transfer power of a pre-trained network.   Our primary contributions are the following:     \u2022  We uncover relationships between learning-path flatness, neural network geometric-complexity,",
            "references": [
                {
                    "title": "A margin-based multiclass generalization bound via geometric complexity",
                    "abstract": "There has been considerable effort to better understand the generalization capabilities of deep neural networks both as a means to unlock a theoretical understanding of their success as well as providing directions for further improvements. In this paper, we investigate margin-based multiclass generalization bounds for neural networks which rely on a recent complexity measure, the geometric complexity, developed for neural networks. We derive a new upper bound on the generalization error which scales with the margin-normalized geometric complexity of the network and which holds for a broad family of data distributions and model classes. Our generalization bound is empirically investigated for a ResNet-18 model trained with SGD on the CIFAR-10 and CIFAR-100 datasets with both original and random labels."
                },
                {
                    "title": "Towards Demystifying the Generalization Behaviors When Neural Collapse Emerges",
                    "abstract": "Neural Collapse (NC) is a well-known phenomenon of deep neural networks in the terminal phase of training (TPT). It is characterized by the collapse of features and classifier into a symmetrical structure, known as simplex equiangular tight frame (ETF). While there have been extensive studies on optimization characteristics showing the global optimality of neural collapse, little research has been done on the generalization behaviors during the occurrence of NC. Particularly, the important phenomenon of generalization improvement during TPT has been remaining in an empirical observation and lacking rigorous theoretical explanation. In this paper, we establish the connection between the minimization of CE and a multi-class SVM during TPT, and then derive a multi-class margin generalization bound, which provides a theoretical explanation for why continuing training can still lead to accuracy improvement on test set, even after the train accuracy has reached 100%. Additionally, our further theoretical results indicate that different alignment between labels and features in a simplex ETF can result in varying degrees of generalization improvement, despite all models reaching NC and demonstrating similar optimization performance on train set. We refer to this newly discovered property as\"non-conservative generalization\". In experiments, we also provide empirical observations to verify the indications suggested by our theoretical results."
                },
                {
                    "title": "How Far Pre-trained Models Are from Neural Collapse on the Target Dataset Informs their Transferability",
                    "abstract": "This paper focuses on model transferability estimation, i.e., assessing the performance of pre-trained models on a downstream task without performing fine-tuning. Motivated by neural collapse (NC) [25] that reveals particular feature geometry at the terminal stage of training, we consider model transferability as how far the target activations obtained by pre-trained models are from their hypothetical state in the terminal phase of the model fine-tuned on the target domain. We propose a metric that measures this proximity based on three phenomena of NC: within-class variability collapse, simplex encoded label interpolation geometry structure is formed, and the nearest center classifier becomes optimal on training data. Through experiments on 11 datasets, we confirm none of the three NC proxies are dispensable, which allows us to obtain very competitive transferability estimation accuracy with approximately 10\u00d7 wall-clock time speed up compared to state-of-the-art approaches."
                },
                {
                    "title": "On the Implicit Bias of Adam",
                    "abstract": "In previous literature, backward error analysis was used to find ordinary differential equations (ODEs) approximating the gradient descent trajectory. It was found that finite step sizes implicitly regularize solutions because terms appearing in the ODEs penalize the two-norm of the loss gradients. We prove that the existence of similar implicit regularization in RMSProp and Adam depends on their hyperparameters and the training stage, but with a different\"norm\"involved: the corresponding ODE terms either penalize the (perturbed) one-norm of the loss gradients or, conversely, impede its reduction (the latter case being typical). We also conduct numerical experiments and discuss how the proven facts can influence generalization."
                },
                {
                    "title": "Sparse Modular Activation for Efficient Sequence Modeling",
                    "abstract": "Linear State Space Models (SSMs) have demonstrated strong performance in a variety of sequence modeling tasks due to their efficient encoding of the recurrent structure. However, in more comprehensive tasks like language modeling and machine translation, self-attention-based models still outperform SSMs. Hybrid models employing both SSM and self-attention generally show promising performance, but current approaches apply attention modules statically and uniformly to all elements in the input sequences, leading to sub-optimal quality-efficiency trade-offs. In this work, we introduce Sparse Modular Activation (SMA), a general mechanism enabling neural networks to sparsely and dynamically activate sub-modules for sequence elements in a differentiable manner. Through allowing each element to skip non-activated sub-modules, SMA reduces computation and memory consumption at both training and inference stages of sequence modeling. As a specific instantiation of SMA, we design a novel neural architecture, SeqBoat, which employs SMA to sparsely activate a Gated Attention Unit (GAU) based on the state representations learned from an SSM. By constraining the GAU to only conduct local attention on the activated inputs, SeqBoat can achieve linear inference complexity with theoretically infinite attention span, and provide substantially better quality-efficiency trade-off than the chunking-based models. With experiments on a wide range of tasks, including language modeling, speech classification and long-range arena, SeqBoat brings new state-of-the-art results among hybrid models with linear complexity and reveals the amount of attention needed for each task through the learned sparse activation patterns."
                },
                {
                    "title": "Inducing Neural Collapse in Deep Long-tailed Learning",
                    "abstract": "Although deep neural networks achieve tremendous success on various classification tasks, the generalization ability drops sheer when training datasets exhibit long-tailed distributions. One of the reasons is that the learned representations (i.e. features) from the imbalanced datasets are less effective than those from balanced datasets. Specifically, the learned representation under class-balanced distribution will present the Neural Collapse (NC) phenomena. NC indicates the features from the same category are close to each other and from different categories are maximally distant, showing an optimal linear separable state of classification. However, the pattern differs on imbalanced datasets and is partially responsible for the reduced performance of the model. In this work, we propose two explicit feature regularization terms to learn high-quality representation for class-imbalanced data. With the proposed regularization, NC phenomena will appear under the class-imbalanced distribution, and the generalization ability can be significantly improved. Our method is easily implemented, highly effective, and can be plugged into most existing methods. The extensive experimental results on widely-used benchmarks show the effectiveness of our method"
                },
                {
                    "title": "Implicit regularization in Heavy-ball momentum accelerated stochastic gradient descent",
                    "abstract": "It is well known that the finite step-size ($h$) in Gradient Descent (GD) implicitly regularizes solutions to flatter minima. A natural question to ask is\"Does the momentum parameter $\\beta$ play a role in implicit regularization in Heavy-ball (H.B) momentum accelerated gradient descent (GD+M)?\". To answer this question, first, we show that the discrete H.B momentum update (GD+M) follows a continuous trajectory induced by a modified loss, which consists of an original loss and an implicit regularizer. Then, we show that this implicit regularizer for (GD+M) is stronger than that of (GD) by factor of $(\\frac{1+\\beta}{1-\\beta})$, thus explaining why (GD+M) shows better generalization performance and higher test accuracy than (GD). Furthermore, we extend our analysis to the stochastic version of gradient descent with momentum (SGD+M) and characterize the continuous trajectory of the update of (SGD+M) in a pointwise sense. We explore the implicit regularization in (SGD+M) and (GD+M) through a series of experiments validating our theory."
                },
                {
                    "title": "Understanding and Improving Transfer Learning of Deep Models via Neural Collapse",
                    "abstract": "With the ever-increasing complexity of large-scale pre-trained models coupled with a shortage of labeled data for downstream training, transfer learning has become the primary approach in many fields, including natural language processing, computer vision, and multi-modal learning. Despite recent progress, the fine-tuning process for large-scale pre-trained models in vision still mostly relies on trial and error. This work investigates the relationship between neural collapse (NC) and transfer learning for classification problems. NC is an intriguing while prevalent phenomenon that has been recently discovered in terms of the final-layer features and linear classifiers of trained neural networks. Specifically, during the terminal phase of training, NC implies that the variability of the features within each class diminishes to zero, while the means of features between classes are maximally and equally distanced. In this work, we examine the NC attributes of pre-trained models on both downstream and source data for transfer learning, and we find strong correlation between feature collapse and downstream performance. In particular, we discovered a systematic pattern that emerges when linear probing pre-trained models on downstream training data: the more feature collapse of pre-trained models on downstream training data, the higher the transfer accuracy. Additionally, we also studied the relationship between NC and transfer accuracy on the source data. Moreover, these findings allow us to develop a principled, parameter-efficient fine-tuning method that employs skip-connection to induce the last-layer feature collapse on downstream data. Our proposed fine-tuning methods deliver good performances while reducing fine-tuning parameters by at least 90% and mitigating overfitting in situations especially when the downstream data is scarce."
                },
                {
                    "title": "Same Pre-training Loss, Better Downstream: Implicit Bias Matters for Language Models",
                    "abstract": "Language modeling on large-scale datasets leads to impressive performance gains on various downstream language tasks. The validation pre-training loss (or perplexity in autoregressive language modeling) is often used as the evaluation metric when developing language models since the pre-training loss tends to be well-correlated with downstream performance (which is itself difficult to evaluate comprehensively). Contrary to this conventional wisdom, this paper shows that 1) pre-training loss cannot fully explain downstream performance and 2) flatness of the model is well-correlated with downstream performance where pre-training loss is not. On simplified datasets, we identify three ways to produce models with the same (statistically optimal) pre-training loss but different downstream performance: continue pre-training after convergence, increasing the model size, and changing the training algorithm. These experiments demonstrate the existence of implicit bias of pre-training algorithms/optimizers -- among models with the same minimal pre-training loss, they implicitly prefer more transferable ones. Toward understanding this implicit bias, we prove that SGD with standard mini-batch noise implicitly prefers flatter minima in language models, and empirically observe a strong correlation between flatness and downstream performance among models with the same minimal pre-training loss. We also prove in a synthetic language setting that among the models with the minimal pre-training loss, the flattest model transfers to downstream tasks."
                },
                {
                    "title": "Why neural networks find simple solutions: the many regularizers of geometric complexity",
                    "abstract": "In many contexts, simpler models are preferable to more complex models and the control of this model complexity is the goal for many methods in machine learning such as regularization, hyperparameter tuning and architecture design. In deep learning, it has been difficult to understand the underlying mechanisms of complexity control, since many traditional measures are not naturally suitable for deep neural networks. Here we develop the notion of geometric complexity, which is a measure of the variability of the model function, computed using a discrete Dirichlet energy. Using a combination of theoretical arguments and empirical results, we show that many common training heuristics such as parameter norm regularization, spectral norm regularization, flatness regularization, implicit gradient regularization, noise regularization and the choice of parameter initialization all act to control geometric complexity, providing a unifying framework in which to characterize the behavior of deep learning models."
                },
                {
                    "title": "Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limit",
                    "abstract": "There is mounting evidence of emergent phenomena in the capabilities of deep learning methods as we scale up datasets, model sizes, and training times. While there are some accounts of how these resources modulate statistical capacity, far less is known about their effect on the computational problem of model training. This work conducts such an exploration through the lens of learning a $k$-sparse parity of $n$ bits, a canonical discrete search problem which is statistically easy but computationally hard. Empirically, we find that a variety of neural networks successfully learn sparse parities, with discontinuous phase transitions in the training curves. On small instances, learning abruptly occurs at approximately $n^{O(k)}$ iterations; this nearly matches SQ lower bounds, despite the apparent lack of a sparse prior. Our theoretical analysis shows that these observations are not explained by a Langevin-like mechanism, whereby SGD\"stumbles in the dark\"until it finds the hidden set of features (a natural algorithm which also runs in $n^{O(k)}$ time). Instead, we show that SGD gradually amplifies the sparse solution via a Fourier gap in the population gradient, making continual progress that is invisible to loss and error metrics."
                },
                {
                    "title": "Neural Collapse: A Review on Modelling Principles and Generalization",
                    "abstract": "Deep classifier neural networks enter the terminal phase of training (TPT) when training error reaches zero and tend to exhibit intriguing Neural Collapse (NC) properties. Neural collapse essentially represents a state at which the within-class variability of final hidden layer outputs is infinitesimally small and their class means form a simplex equiangular tight frame. This simplifies the last layer behaviour to that of a nearest-class center decision rule. Despite the simplicity of this state, the dynamics and implications of reaching it are yet to be fully understood. In this work, we review the principles which aid in modelling neural collapse, followed by the implications of this state on generalization and transfer learning capabilities of neural networks. Finally, we conclude by discussing potential avenues and directions for future research."
                },
                {
                    "title": "Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets",
                    "abstract": "In this paper we propose to study generalization of neural networks on small algorithmically generated datasets. In this setting, questions about data efficiency, memorization, generalization, and speed of learning can be studied in great detail. In some situations we show that neural networks learn through a process of\"grokking\"a pattern in the data, improving generalization performance from random chance level to perfect generalization, and that this improvement in generalization can happen well past the point of overfitting. We also study generalization as a function of dataset size and find that smaller datasets require increasing amounts of optimization for generalization. We argue that these datasets provide a fertile ground for studying a poorly understood aspect of deep learning: generalization of overparametrized neural networks beyond memorization of the finite training dataset."
                },
                {
                    "title": "On the Role of Neural Collapse in Transfer Learning",
                    "abstract": "We study the ability of foundation models to learn representations for classification that are transferable to new, unseen classes. Recent results in the literature show that representations learned by a single classifier over many classes are competitive on few-shot learning problems with representations learned by special-purpose algorithms designed for such problems. In this paper we provide an explanation for this behavior based on the recently observed phenomenon that the features learned by overparameterized classification networks show an interesting clustering property, called neural collapse. We demonstrate both theoretically and empirically that neural collapse generalizes to new samples from the training classes, and -- more importantly -- to new classes as well, allowing foundation models to provide feature maps that work well in transfer learning and, specifically, in the few-shot setting."
                },
                {
                    "title": "The Geometric Occam's Razor Implicit in Deep Learning",
                    "abstract": "In over-parameterized deep neural networks there can be many possible parameter configurations that fit the training data exactly. However, the properties of these interpolating solutions are poorly understood. We argue that over-parameterized neural networks trained with stochastic gradient descent are subject to a Geometric Occam's Razor; that is, these networks are implicitly regularized by the geometric model complexity. For one-dimensional regression, the geometric model complexity is simply given by the arc length of the function. For higher-dimensional settings, the geometric model complexity depends on the Dirichlet energy of the function. We explore the relationship between this Geometric Occam's Razor, the Dirichlet energy and other known forms of implicit regularization. Finally, for ResNets trained on CIFAR-10, we observe that Dirichlet energy measurements are consistent with the action of this implicit Geometric Occam's Razor."
                },
                {
                    "title": "Efficient Approximation of Jacobian Matrices Involving a Non-Uniform Fast Fourier Transform (NUFFT)",
                    "abstract": "There is growing interest in learning Fourier domain sampling strategies (particularly for magnetic resonance imaging, MRI) using optimization approaches. For non-Cartesian sampling, the system models typically involve non-uniform fast Fourier transform (NUFFT) operations. Commonly used NUFFT algorithms contain frequency domain interpolation, which is not differentiable with respect to the sampling pattern, complicating the use of gradient methods. This paper describes an efficient and accurate approach for computing approximate gradients involving NUFFTs. Multiple numerical experiments validate the improved accuracy and efficiency of the proposed approximation. As an application to computational imaging, the NUFFT Jacobians were used to optimize non-Cartesian MRI sampling trajectories via data-driven stochastic optimization. Specifically, the sampling patterns were learned with respect to various model-based image reconstruction (MBIR) algorithms. The proposed approach enables sampling optimization for image sizes that are infeasible with standard auto-differentiation methods due to memory limits. The synergistic acquisition and reconstruction design leads to remarkably improved image quality. In fact, we show that model-based image reconstruction methods with suitably optimized imaging parameters can perform nearly as well as CNN-based methods."
                },
                {
                    "title": "Stochastic Training is Not Necessary for Generalization",
                    "abstract": "It is widely believed that the implicit regularization of SGD is fundamental to the impressive generalization behavior we observe in neural networks. In this work, we demonstrate that non-stochastic full-batch training can achieve comparably strong performance to SGD on CIFAR-10 using modern architectures. To this end, we show that the implicit regularization of SGD can be completely replaced with explicit regularization even when comparing against a strong and well-researched baseline. Our observations indicate that the perceived difficulty of full-batch training may be the result of its optimization properties and the disproportionate time and effort spent by the ML community tuning optimizers and hyperparameters for small-batch training."
                },
                {
                    "title": "Label Noise SGD Provably Prefers Flat Global Minimizers",
                    "abstract": "In overparametrized models, the noise in stochastic gradient descent (SGD) implicitly regularizes the optimization trajectory and determines which local minimum SGD converges to. Motivated by empirical studies that demonstrate that training with noisy labels improves generalization, we study the implicit regularization effect of SGD with label noise. We show that SGD with label noise converges to a stationary point of a regularized loss L ( \u03b8 )+ \u03bbR ( \u03b8 ) , where L ( \u03b8 ) is the training loss, \u03bb is an effective regularization parameter depending on the step size, strength of the label noise, and the batch size, and R ( \u03b8 ) is an explicit regularizer that penalizes sharp minimizers. Our analysis uncovers an additional regularization effect of large learning rates beyond the linear scaling rule that penalizes large eigenvalues of the Hessian more than small ones. We also prove extensions to classi\ufb01cation with general loss functions, signi\ufb01cantly strengthening the prior work of Blanc et al. [3] to global convergence and large learning rates and of HaoChen et al. [12] to general models."
                },
                {
                    "title": "Neural Collapse Under MSE Loss: Proximity to and Dynamics on the Central Path",
                    "abstract": "The recently discovered Neural Collapse (NC) phenomenon occurs pervasively in today's deep net training paradigm of driving cross-entropy (CE) loss towards zero. During NC, last-layer features collapse to their class-means, both classifiers and class-means collapse to the same Simplex Equiangular Tight Frame, and classifier behavior collapses to the nearest-class-mean decision rule. Recent works demonstrated that deep nets trained with mean squared error (MSE) loss perform comparably to those trained with CE. As a preliminary, we empirically establish that NC emerges in such MSE-trained deep nets as well through experiments on three canonical networks and five benchmark datasets. We provide, in a Google Colab notebook, PyTorch code for reproducing MSE-NC and CE-NC: at https://colab.research.google.com/github/neuralcollapse/neuralcollapse/blob/main/neuralcollapse.ipynb. The analytically-tractable MSE loss offers more mathematical opportunities than the hard-to-analyze CE loss, inspiring us to leverage MSE loss towards the theoretical investigation of NC. We develop three main contributions: (I) We show a new decomposition of the MSE loss into (A) terms directly interpretable through the lens of NC and which assume the last-layer classifier is exactly the least-squares classifier; and (B) a term capturing the deviation from this least-squares classifier. (II) We exhibit experiments on canonical datasets and networks demonstrating that term-(B) is negligible during training. This motivates us to introduce a new theoretical construct: the central path, where the linear classifier stays MSE-optimal for feature activations throughout the dynamics. (III) By studying renormalized gradient flow along the central path, we derive exact dynamics that predict NC."
                },
                {
                    "title": "On Linear Stability of SGD and Input-Smoothness of Neural Networks",
                    "abstract": "The multiplicative structure of parameters and input data in the first layer of neural networks is explored to build connection between the landscape of the loss function with respect to parameters and the landscape of the model function with respect to input data. By this connection, it is shown that flat minima regularize the gradient of the model function, which explains the good generalization performance of flat minima. Then, we go beyond the flatness and consider high-order moments of the gradient noise, and show that Stochastic Gradient Descent (SGD) tends to impose constraints on these moments by a linear stability analysis of SGD around global minima. Together with the multiplicative structure, we identify the Sobolev regularization effect of SGD, i.e. SGD regularizes the Sobolev seminorms of the model function with respect to the input data. Finally, bounds for generalization error and adversarial robustness are provided for solutions found by SGD under assumptions of the data distribution."
                },
                {
                    "title": "A Universal Law of Robustness via Isoperimetry",
                    "abstract": "Classically, data interpolation with a parametrized model class is possible as long as the number of parameters is larger than the number of equations to be satisfied. A puzzling phenomenon in deep learning is that models are trained with many more parameters than what this classical theory would suggest. We propose a partial theoretical explanation for this phenomenon. We prove that for a broad class of data distributions and model classes, overparametrization is necessary if one wants to interpolate the data smoothly. Namely we show that smooth interpolation requires d times more parameters than mere interpolation, where d is the ambient data dimension. We prove this universal law of robustness for any smoothly parametrized function class with polynomial size weights, and any covariate distribution verifying isoperimetry (or a mixture thereof). In the case of two-layer neural networks and Gaussian covariates, this law was conjectured in prior work by Bubeck, Li, and Nagaraj. We also give an interpretation of our result as an improved generalization bound for model classes consisting of smooth functions."
                },
                {
                    "title": "Zero-Shot Text-to-Image Generation",
                    "abstract": "Text-to-image generation has traditionally focused on finding better modeling assumptions for training on a fixed dataset. These assumptions might involve complex architectures, auxiliary losses, or side information such as object part labels or segmentation masks supplied during training. We describe a simple approach for this task based on a transformer that autoregressively models the text and image tokens as a single stream of data. With sufficient data and scale, our approach is competitive with previous domain-specific models when evaluated in a zero-shot fashion."
                },
                {
                    "title": "Understanding deep learning (still) requires rethinking generalization",
                    "abstract": "Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small gap between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family or to the regularization techniques used during training. Through extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. We interpret our experimental findings by comparison with traditional models. We supplement this republication with a new section at the end summarizing recent progresses in the field since the original version of this paper."
                },
                {
                    "title": "On the Origin of Implicit Regularization in Stochastic Gradient Descent",
                    "abstract": "For infinitesimal learning rates, stochastic gradient descent (SGD) follows the path of gradient flow on the full batch loss function. However moderately large learning rates can achieve higher test accuracies, and this generalization benefit is not explained by convergence bounds, since the learning rate which maximizes test accuracy is often larger than the learning rate which minimizes training loss. To interpret this phenomenon we prove that for SGD with random shuffling, the mean SGD iterate also stays close to the path of gradient flow if the learning rate is small and finite, but on a modified loss. This modified loss is composed of the original loss function and an implicit regularizer, which penalizes the norms of the minibatch gradients. Under mild assumptions, when the batch size is small the scale of the implicit regularization term is proportional to the ratio of the learning rate to the batch size. We verify empirically that explicitly including the implicit regularizer in the loss can enhance the test accuracy when the learning rate is small."
                },
                {
                    "title": "Extracting Training Data from Large Language Models",
                    "abstract": "It has become common to publish large (billion parameter) language models that have been trained on private datasets. This paper demonstrates that in such settings, an adversary can perform a training data extraction attack to recover individual training examples by querying the language model. \nWe demonstrate our attack on GPT-2, a language model trained on scrapes of the public Internet, and are able to extract hundreds of verbatim text sequences from the model's training data. These extracted examples include (public) personally identifiable information (names, phone numbers, and email addresses), IRC conversations, code, and 128-bit UUIDs. Our attack is possible even though each of the above sequences are included in just one document in the training data. \nWe comprehensively evaluate our extraction attack to understand the factors that contribute to its success. For example, we find that larger models are more vulnerable than smaller models. We conclude by drawing lessons and discussing possible safeguards for training large language models."
                },
                {
                    "title": "A case for new neural network smoothness constraints",
                    "abstract": "How sensitive should machine learning models be to input changes? We tackle the question of model smoothness and show that it is a useful inductive bias which aids generalization, adversarial robustness, generative modeling and reinforcement learning. We explore current methods of imposing smoothness constraints and observe they lack the flexibility to adapt to new tasks, they don't account for data modalities, they interact with losses, architectures and optimization in ways not yet fully understood. We conclude that new advances in the field are hinging on finding ways to incorporate data, tasks and learning into our definitions of smoothness."
                },
                {
                    "title": "A Concise Review of Transfer Learning",
                    "abstract": "The availability of abundant labeled data in recent years led the researchers to introduce a methodology called transfer learning, which utilizes existing data in situations where there are difficulties in collecting new annotated data. Transfer learning aims to boost the performance of a target learner by applying another related source data. In contrast to the traditional machine learning and data mining techniques, which assume that the training and testing data lie from the same feature space and distribution, transfer learning can handle situations where there is a discrepancy between domains and distributions. These characteristics give the model the potential to utilize the available related source data and extend the underlying knowledge to the target task achieving better performance. This survey paper aims to give a concise review of traditional and current transfer learning settings, existing challenges, and related approaches."
                },
                {
                    "title": "Implicit Gradient Regularization",
                    "abstract": "Gradient descent can be surprisingly good at optimizing deep neural networks without overfitting and without explicit regularization. We find that the discrete steps of gradient descent implicitly regularize models by penalizing gradient descent trajectories that have large loss gradients. We call this Implicit Gradient Regularization (IGR) and we use backward error analysis to calculate the size of this regularization. We confirm empirically that implicit gradient regularization biases gradient descent toward flat minima, where test errors are small and solutions are robust to noisy parameter perturbations. Furthermore, we demonstrate that the implicit gradient regularization term can be used as an explicit regularizer, allowing us to control this gradient regularization directly. More broadly, our work indicates that backward error analysis is a useful theoretical approach to the perennial question of how learning rate, model size, and parameter regularization interact to determine the properties of overparameterized models optimized with gradient descent."
                },
                {
                    "title": "Prevalence of neural collapse during the terminal phase of deep learning training",
                    "abstract": "Significance Modern deep neural networks for image classification have achieved superhuman performance. Yet, the complex details of trained networks have forced most practitioners and researchers to regard them as black boxes with little that could be understood. This paper considers in detail a now-standard training methodology: driving the cross-entropy loss to zero, continuing long after the classification error is already zero. Applying this methodology to an authoritative collection of standard deepnets and datasets, we observe the emergence of a simple and highly symmetric geometry of the deepnet features and of the deepnet classifier, and we document important benefits that the geometry conveys\u2014thereby helping us understand an important component of the modern deep learning training paradigm. Modern practice for training classification deepnets involves a terminal phase of training (TPT), which begins at the epoch where training error first vanishes. During TPT, the training error stays effectively zero, while training loss is pushed toward zero. Direct measurements of TPT, for three prototypical deepnet architectures and across seven canonical classification datasets, expose a pervasive inductive bias we call neural collapse (NC), involving four deeply interconnected phenomena. (NC1) Cross-example within-class variability of last-layer training activations collapses to zero, as the individual activations themselves collapse to their class means. (NC2) The class means collapse to the vertices of a simplex equiangular tight frame (ETF). (NC3) Up to rescaling, the last-layer classifiers collapse to the class means or in other words, to the simplex ETF (i.e., to a self-dual configuration). (NC4) For a given activation, the classifier\u2019s decision collapses to simply choosing whichever class has the closest train class mean (i.e., the nearest class center [NCC] decision rule). The symmetric and very simple geometry induced by the TPT confers important benefits, including better generalization performance, better robustness, and better interpretability."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "Robust Learning with Jacobian Regularization",
                    "abstract": "Design of reliable systems must guarantee stability against input perturbations. In machine learning, such guarantee entails preventing overfitting and ensuring robustness of models against corruption of input data. In order to maximize stability, we analyze and develop a computationally efficient implementation of Jacobian regularization that increases classification margins of neural networks. The stabilizing effect of the Jacobian regularizer leads to significant improvements in robustness, as measured against both random and adversarial input perturbations, without severely degrading generalization properties on clean data."
                },
                {
                    "title": "Implicit regularization for deep neural networks driven by an Ornstein-Uhlenbeck like process",
                    "abstract": "We consider networks, trained via stochastic gradient descent to minimize $\\ell_2$ loss, with the training labels perturbed by independent noise at each iteration. We characterize the behavior of the training dynamics near any parameter vector that achieves zero training error, in terms of an implicit regularization term corresponding to the sum over the data points, of the squared $\\ell_2$ norm of the gradient of the model with respect to the parameter vector, evaluated at each data point. This holds for networks of any connectivity, width, depth, and choice of activation function. We interpret this implicit regularization term for three simple settings: matrix sensing, two layer ReLU networks trained on one-dimensional data, and two layer networks with sigmoid activations trained on a single datapoint. For these settings, we show why this new and general implicit regularization effect drives the networks towards \"simple\" models."
                },
                {
                    "title": "Meta-Learning With Differentiable Convex Optimization",
                    "abstract": "Many meta-learning approaches for few-shot learning rely on simple base learners such as nearest-neighbor classifiers. However, even in the few-shot regime, discriminatively trained linear predictors can offer better generalization. We propose to use these predictors as base learners to learn representations for few-shot learning and show they offer better tradeoffs between feature size and performance across a range of few-shot recognition benchmarks. Our objective is to learn feature embeddings that generalize well under a linear classification rule for novel categories. To efficiently solve the objective, we exploit two properties of linear classifiers: implicit differentiation of the optimality conditions of the convex problem and the dual formulation of the optimization problem. This allows us to use high-dimensional embeddings with improved generalization at a modest increase in computational overhead. Our approach, named MetaOptNet, achieves state-of-the-art performance on miniImageNet, tieredImageNet, CIFAR-FS, and FC100 few-shot learning benchmarks."
                },
                {
                    "title": "Implicit Self-Regularization in Deep Neural Networks: Evidence from Random Matrix Theory and Implications for Learning",
                    "abstract": "Random Matrix Theory (RMT) is applied to analyze weight matrices of Deep Neural Networks (DNNs), including both production quality, pre-trained models such as AlexNet and Inception, and smaller models trained from scratch, such as LeNet5 and a miniature-AlexNet. Empirical and theoretical results clearly indicate that the DNN training process itself implicitly implements a form of Self-Regularization. The empirical spectral density (ESD) of DNN layer matrices displays signatures of traditionally-regularized statistical models, even in the absence of exogenously specifying traditional forms of explicit regularization. Building on relatively recent results in RMT, most notably its extension to Universality classes of Heavy-Tailed matrices, we develop a theory to identify 5+1 Phases of Training, corresponding to increasing amounts of Implicit Self-Regularization. These phases can be observed during the training process as well as in the final learned DNNs. For smaller and/or older DNNs, this Implicit Self-Regularization is like traditional Tikhonov regularization, in that there is a \"size scale\" separating signal from noise. For state-of-the-art DNNs, however, we identify a novel form of Heavy-Tailed Self-Regularization, similar to the self-organization seen in the statistical physics of disordered systems. This results from correlations arising at all size scales, which arises implicitly due to the training process itself. This implicit Self-Regularization can depend strongly on the many knobs of the training process. By exploiting the generalization gap phenomena, we demonstrate that we can cause a small model to exhibit all 5+1 phases of training simply by changing the batch size. This demonstrates that---all else being equal---DNN optimization with larger batch sizes leads to less-well implicitly-regularized models, and it provides an explanation for the generalization gap phenomena."
                },
                {
                    "title": "Implicit Bias of Gradient Descent on Linear Convolutional Networks",
                    "abstract": "We show that gradient descent on full-width linear convolutional networks of depth $L$ converges to a linear predictor related to the $\\ell_{2/L}$ bridge penalty in the frequency domain. This is in contrast to linearly fully connected networks, where gradient descent converges to the hard margin linear support vector machine solution, regardless of depth."
                },
                {
                    "title": "Sensitivity and Generalization in Neural Networks: an Empirical Study",
                    "abstract": "In practice it is often found that large over-parameterized neural networks generalize better than their smaller counterparts, an observation that appears to conflict with classical notions of function complexity, which typically favor smaller models. In this work, we investigate this tension between complexity and generalization through an extensive empirical exploration of two natural metrics of complexity related to sensitivity to input perturbations. Our experiments survey thousands of models with various fully-connected architectures, optimizers, and other hyper-parameters, as well as four different image classification datasets. \nWe find that trained neural networks are more robust to input perturbations in the vicinity of the training data manifold, as measured by the norm of the input-output Jacobian of the network, and that it correlates well with generalization. We further establish that factors associated with poor generalization $-$ such as full-batch training or using random labels $-$ correspond to lower robustness, while factors associated with good generalization $-$ such as data augmentation and ReLU non-linearities $-$ give rise to more robust functions. Finally, we demonstrate how the input-output Jacobian norm can be predictive of generalization at the level of individual test points."
                },
                {
                    "title": "Implicit Regularization in Deep Learning",
                    "abstract": "In an attempt to better understand generalization in deep learning, we study several possible explanations. We show that implicit regularization induced by the optimization method is playing a key role in generalization and success of deep learning models. Motivated by this view, we study how different complexity measures can ensure generalization and explain how optimization algorithms can implicitly regularize complexity measures. We empirically investigate the ability of these measures to explain different observed phenomena in deep learning. We further study the invariances in neural networks, suggest complexity measures and optimization algorithms that have similar invariances to those in neural networks and evaluate them on a number of learning tasks."
                },
                {
                    "title": "The Marginal Value of Adaptive Gradient Methods in Machine Learning",
                    "abstract": "Adaptive optimization methods, which perform local optimization with a metric constructed from the history of iterates, are becoming increasingly popular for training deep neural networks. Examples include AdaGrad, RMSProp, and Adam. We show that for simple overparameterized problems, adaptive methods often find drastically different solutions than gradient descent (GD) or stochastic gradient descent (SGD). We construct an illustrative binary classification problem where the data is linearly separable, GD and SGD achieve zero test error, and AdaGrad, Adam, and RMSProp attain test errors arbitrarily close to half. We additionally study the empirical generalization capability of adaptive methods on several state-of-the-art deep learning models. We observe that the solutions found by adaptive methods generalize worse (often significantly worse) than SGD, even when these solutions have better training performance. These results suggest that practitioners should reconsider the use of adaptive methods to train neural networks."
                },
                {
                    "title": "Sharp Minima Can Generalize For Deep Nets",
                    "abstract": "Despite their overwhelming capacity to overfit, deep learning architectures tend to generalize relatively well to unseen data, allowing them to be deployed in practice. However, explaining why this is the case is still an open area of research. One standing hypothesis that is gaining popularity, e.g. Hochreiter & Schmidhuber (1997); Keskar et al. (2017), is that the flatness of minima of the loss function found by stochastic gradient based methods results in good generalization. This paper argues that most notions of flatness are problematic for deep models and can not be directly applied to explain generalization. Specifically, when focusing on deep networks with rectifier units, we can exploit the particular geometry of parameter space induced by the inherent symmetries that these architectures exhibit to build equivalent models corresponding to arbitrarily sharper minima. Furthermore, if we allow to reparametrize a function, the geometry of its parameters can change drastically without affecting its generalization properties."
                },
                {
                    "title": "On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima",
                    "abstract": "The stochastic gradient descent (SGD) method and its variants are algorithms of choice for many Deep Learning tasks. These methods operate in a small-batch regime wherein a fraction of the training data, say $32$-$512$ data points, is sampled to compute an approximation to the gradient. It has been observed in practice that when using a larger batch there is a degradation in the quality of the model, as measured by its ability to generalize. We investigate the cause for this generalization drop in the large-batch regime and present numerical evidence that supports the view that large-batch methods tend to converge to sharp minimizers of the training and testing functions - and as is well known, sharp minima lead to poorer generalization. In contrast, small-batch methods consistently converge to flat minimizers, and our experiments support a commonly held view that this is due to the inherent noise in the gradient estimation. We discuss several strategies to attempt to help large-batch methods eliminate this generalization gap."
                },
                {
                    "title": "Matching Networks for One Shot Learning",
                    "abstract": "Learning from a few examples remains a key challenge in machine learning. Despite recent advances in important domains such as vision and language, the standard supervised deep learning paradigm does not offer a satisfactory solution for learning new concepts rapidly from little data. In this work, we employ ideas from metric learning based on deep neural features and from recent advances that augment neural networks with external memories. Our framework learns a network that maps a small labelled support set and an unlabelled example to its label, obviating the need for fine-tuning to adapt to new class types. We then define one-shot learning problems on vision (using Omniglot, ImageNet) and language tasks. Our algorithm improves one-shot accuracy on ImageNet from 87.6% to 93.2% and from 88.0% to 93.8% on Omniglot compared to competing approaches. We also demonstrate the usefulness of the same model on language modeling by introducing a one-shot task on the Penn Treebank."
                },
                {
                    "title": "Robust Large Margin Deep Neural Networks",
                    "abstract": "The generalization error of deep neural networks via their classification margin is studied in this paper. Our approach is based on the Jacobian matrix of a deep neural network and can be applied to networks with arbitrary nonlinearities and pooling layers, and to networks with different architectures such as feed forward networks and residual networks. Our analysis leads to the conclusion that a bounded spectral norm of the network's Jacobian matrix in the neighbourhood of the training samples is crucial for a deep neural network of arbitrary depth and width to generalize well. This is a significant improvement over the current bounds in the literature, which imply that the generalization error grows with either the width or the depth of the network. Moreover, it shows that the recently proposed batch normalization and weight normalization reparametrizations enjoy good generalization properties, and leads to a novel network regularizer based on the network's Jacobian matrix. The analysis is supported with experimental results on the MNIST, CIFAR-10, LaRED, and ImageNet datasets."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "In Search of the Real Inductive Bias: On the Role of Implicit Regularization in Deep Learning",
                    "abstract": "We present experiments demonstrating that some other form of capacity control, different from network size, plays a central role in learning multilayer feed-forward networks. We argue, partially through analogy to matrix factorization, that this is an inductive bias that can help shed light on deep learning."
                },
                {
                    "title": "The Loss Surfaces of Multilayer Networks",
                    "abstract": "We study the connection between the highly non-convex loss function of a simple model of the fully-connected feed-forward neural network and the Hamiltonian of the spherical spin-glass model under the assumptions of: i) variable independence, ii) redundancy in network parametrization, and iii) uniformity. These assumptions enable us to explain the complexity of the fully decoupled neural network through the prism of the results from random matrix theory. We show that for large-size decoupled networks the lowest critical values of the random loss function form a layered structure and they are located in a well-defined band lower-bounded by the global minimum. The number of local minima outside that band diminishes exponentially with the size of the network. We empirically verify that the mathematical model exhibits similar behavior as the computer simulations, despite the presence of high dependencies in real networks. We conjecture that both simulated annealing and SGD converge to the band of low critical points, and that all critical points found there are local minima of high quality measured by the test error. This emphasizes a major difference between largeand small-size networks where for the latter poor quality local minima have nonzero probability of being recovered. Finally, we prove that recovering the global minimum becomes harder as the network size increases and that it is in practice irrelevant as global minimum often leads to overfitting."
                },
                {
                    "title": "How transferable are features in deep neural networks?",
                    "abstract": "Many deep neural networks trained on natural images exhibit a curious phenomenon in common: on the first layer they learn features similar to Gabor filters and color blobs. Such first-layer features appear not to be specific to a particular dataset or task, but general in that they are applicable to many datasets and tasks. Features must eventually transition from general to specific by the last layer of the network, but this transition has not been studied extensively. In this paper we experimentally quantify the generality versus specificity of neurons in each layer of a deep convolutional neural network and report a few surprising results. Transferability is negatively affected by two distinct issues: (1) the specialization of higher layer neurons to their original task at the expense of performance on the target task, which was expected, and (2) optimization difficulties related to splitting networks between co-adapted neurons, which was not expected. In an example network trained on ImageNet, we demonstrate that either of these two issues may dominate, depending on whether features are transferred from the bottom, middle, or top of the network. We also document that the transferability of features decreases as the distance between the base task and target task increases, but that transferring features even from distant tasks can be better than using random features. A final surprising result is that initializing a network with transferred features from almost any number of layers can produce a boost to generalization that lingers even after fine-tuning to the target dataset."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "One billion word benchmark for measuring progress in statistical language modeling",
                    "abstract": "We propose a new benchmark corpus to be used for measuring progress in statistical language modeling. With almost one billion words of training data, we hope this benchmark will be useful to quickly evaluate novel language modeling techniques, and to compare their contribution when combined with other advanced techniques. We show performance of several well-known types of language models, with the best results achieved with a recurrent neural network based language model. The baseline unpruned KneserNey 5-gram model achieves perplexity 67.6. A combination of techniques leads to 35% reduction in perplexity, or 10% reduction in cross-entropy (bits), over that baseline. The benchmark is available as a code.google.com project; besides the scripts needed to rebuild the training/held-out data, it also makes available log-probability values for each word in each of ten held-out data sets, for each of the baseline n-gram models."
                },
                {
                    "title": "Deep Learning of Representations for Unsupervised and Transfer Learning",
                    "abstract": "Deep learning algorithms seek to exploit the unknown structure in the input distribution in order to discover good representations, often at multiple levels, with higher-level learned features defined in terms of lower-level features. The objective is to make these higher-level representations more abstract, with their individual features more invariant to most of the variations that are typically present in the training distribution, while collectively preserving as much as possible of the information in the input. Ideally, we would like these representations to disentangle the unknown factors of variation that underlie the training distribution. Such unsupervised learning of representations can be exploited usefully under the hypothesis that the input distribution P(x) is structurally related to some task of interest, say predicting P(y/x). This paper focuses on the context of the Unsupervised and Transfer Learning Challenge, on why unsupervised pre-training of representations can be useful, and how it can be exploited in the transfer learning scenario, where we care about predictions on examples that are not from the same distribution as the training distribution."
                },
                {
                    "title": "A Survey on Transfer Learning",
                    "abstract": "A major assumption in many machine learning and data mining algorithms is that the training and future data must be in the same feature space and have the same distribution. However, in many real-world applications, this assumption may not hold. For example, we sometimes have a classification task in one domain of interest, but we only have sufficient training data in another domain of interest, where the latter data may be in a different feature space or follow a different data distribution. In such cases, knowledge transfer, if done successfully, would greatly improve the performance of learning by avoiding much expensive data-labeling efforts. In recent years, transfer learning has emerged as a new learning framework to address this problem. This survey focuses on categorizing and reviewing the current progress on transfer learning for classification, regression, and clustering problems. In this survey, we discuss the relationship between transfer learning and other related machine learning techniques such as domain adaptation, multitask learning and sample selection bias, as well as covariate shift. We also explore some potential future issues in transfer learning research."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "A simple proof of the Poincar\u00e9 inequality for a large class of probability measures",
                    "abstract": "Abstract. We give a simple and direct proof of the existence of a spectral gap under some Lyapunov type condition which is satisfied in particular by log-concave probability measures on $\\mathbb{R}^n$. The proof is based on arguments introduced in Bakry and al, but for the sake of completeness, all details are provided."
                },
                {
                    "title": "The Sobolev Regularization Effect of Stochastic Gradient Descent",
                    "abstract": "The multiplicative structure of parameters and input data in the \ufb01rst layer of neural networks is explored to build connection between the landscape of the loss function with respect to parameters and the landscape of the model function with respect to input data. By this connection, it is shown that \ufb02at minima regularize the gradient of the model function, which explains the good generalization performance of \ufb02at minima. Then, we go beyond the \ufb02atness and consider high-order moments of the gradient noise, and show that Stochastic Gradient Dascent (SGD) tends to impose constraints on these moments by a linear stability analysis of SGD around global minima. Together with the multiplicative structure, we identify the Sobolev regularization e\ufb00ect of SGD, i.e. SGD regularizes the Sobolev seminorms of the model function with respect to the input data. Finally, bounds for generalization error and adversarial robustness are provided for solutions found by SGD under assumptions of the data distribution."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Geometric Measure Theory",
                    "abstract": "Geometric measure theory is an area of analysis concerned with solving geometric problems via measure theoretic techniques. The canonical motivating physical problem is probably that investigated experimentally by Plateau in the nineteenth century [3]: given a boundary wire, how does one find the (minimal) soap film which spans it? Slightly more mathematically, given a boundary curve, find the surface of minimal area spanning it. The many different approaches to solving this problem have found utility in most areas of modern mathematics and geometric measure theory is no exception: techniques and ideas from geometric measure theory have been found useful in the study of partial differential equations, the calculus of variations, harmonic analysis, and fractals. Successes in the field include: classifying the structure of singularities in soap films (see [18], together with the fine descriptive article [4]); showing that the standard \u2018double bubble\u2019 is the optimal shape for enclosing two prescribed volumes in space [13], and developing powerful computer software for modelling the evolution of surfaces under the action of physical forces [7]. The main reference text for the subject is still Federer\u2019s comprehensive book [11]. It is very densely written and Morgan\u2019s book [15] serves as a useful guide through it. Federer\u2019s colloquium lectures [10] provide a comprehensive overview of the subject and contain a summary of the main results in his book [11]. More recent books include Simon [17], which contains an introduction to the theory of varifolds and Allard\u2019s regularity theorem, and Mattila\u2019s book [14] which includes information about tangent measures and their uses. Both of these books are also suitable as introductions to the area. For a slightly different slant, the book by Evans and Gariepy [9], discusses applications of some of the ideas of geometric measure theory in the theory of Sobolev Spaces and functions of bounded variation. Many variational problems are solved by enlarging the allowed class of solutions, showing that in this enlarged class a solution exists, and then showing that the solution possesses more regularity than an arbitrary element of the enlarged class. Much of the work in geometric measure theory has been directed towards placing this informal description on a formal footing appropriate for the study of surfaces. The key concept underlying the whole theory is that of rectifiability: this is a measure theoretic notion of smoothness. A set E in Euclidean n-space, R, is"
                },
                {
                    "title": "Flat Minima",
                    "abstract": "We present a new algorithm for finding low-complexity neural networks with high generalization capability. The algorithm searches for a flat minimum of the error function. A flat minimum is a large connected region in weight space where the error remains approximately constant. An MDL-based, Bayesian argument suggests that flat minima correspond to simple networks and low expected overfitting. The argument is based on a Gibbs algorithm variant and a novel way of splitting generalization error into underfitting and overfitting error. Unlike many previous approaches, ours does not require gaussian assumptions and does not depend on a good weight prior. Instead we have a prior over input output functions, thus taking into account net architecture and training set. Although our algorithm requires the computation of second-order derivatives, it has backpropagation's order of complexity. Automatically, it effectively prunes units, weights, and input lines. Various experiments with feedforward and recurrent nets are described. In an application to stock market prediction, flat minimum search outperforms conventional backprop, weight decay, and optimal brain surgeon/optimal brain damage."
                },
                {
                    "title": "Learning Many Related Tasks at the Same Time with Backpropagation",
                    "abstract": "Hinton [6] proposed that generalization in artificial neural nets should improve if nets learn to represent the domain's underlying regularities. Abu-Mustafa's hints work [1] shows that the outputs of a backprop net can be used as inputs through which domain-specific information can be given to the net. We extend these ideas by showing that a backprop net learning many related tasks at the same time can use these tasks as inductive bias for each other and thus learn better. We identify five mechanisms by which multitask backprop improves generalization and give empirical evidence that multitask backprop generalizes better in real domains."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow does the geometric complexity of pre-trained neural networks influence their effectiveness in transfer learning, particularly in few-shot learning scenarios?\n\n### [Question 2] - Why is it interesting and important?\nUnderstanding the relationship between geometric complexity and transfer learning can provide deeper insights into the mechanisms that make transfer learning effective. This knowledge can advance the research community by offering a theoretical framework that unifies existing concepts like neural collapse and loss surface flatness. By addressing this question, future research can explore more efficient transfer learning strategies, leading to practical applications in areas with limited labeled data, such as medical imaging or natural language processing, where data scarcity is a significant challenge.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the complexity of quantifying geometric properties of neural networks and their impact on transfer learning outcomes. Naive approaches may fail because they do not account for the intricate relationships between the learned representations and the underlying geometry of the loss surface. Additionally, the lack of a comprehensive theoretical framework to connect these concepts makes it difficult to derive actionable insights. Overcoming these obstacles requires advanced mathematical tools and a deep understanding of neural network behavior during pre-training and fine-tuning.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on empirical observations of transfer learning effectiveness without a unified theoretical understanding of the underlying mechanisms. Gaps exist in linking geometric complexity to transfer learning outcomes, and existing studies have not sufficiently explored how these geometric properties can serve as proxies for transfer power. Barriers include the complexity of the mathematical relationships involved and the lack of methodologies to systematically analyze these properties. Our approach differs by providing a comprehensive analysis that connects geometric complexity with transfer learning efficacy, thereby filling these gaps.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves analyzing the geometric complexity of internal representations learned during the pre-training phase of deep neural networks. We will conduct experiments using various datasets and tasks to measure the impact of geometric complexity on transfer learning performance, particularly in few-shot learning scenarios. The metrics we plan to use include performance accuracy on target tasks and quantifications of geometric complexity. We expect to demonstrate that geometric complexity serves as a hidden progress metric, providing insights into the transfer power of pre-trained networks and guiding future research in transfer learning strategies."
            }
        },
        "author_data": {
            "31829a80-4e83-443d-a8c9-6f7b60eae6d4": {
                "pk": "31829a80-4e83-443d-a8c9-6f7b60eae6d4",
                "name": "Michael Munn",
                "collaborators": [
                    "Sajjad Lakzian",
                    "Benoit Dherin",
                    "David G. T. Barrett",
                    "Susan Wei",
                    "Dan Garbin",
                    "Jay Jorgenson",
                    "Javier Gonzalvo",
                    "Qintao Deng",
                    "Fernando Galaz-Garcia",
                    "Luis Guijarro"
                ],
                "domain": [
                    "Geometric Analysis",
                    "Metric Geometry",
                    "Machine Learning",
                    "Optimal Transport"
                ],
                "publications": [
                    {
                        "title": "Alexandrov spaces with large volume growth",
                        "abstract": "Let $(X,d)$ be an $n$-dimensional Alexandrov space whose Hausdorff measure $\\mathcal{H}^n$ satisfies a condition giving the metric measure space $(X,d,\\mathcal{H}^n)$ a notion of having nonnegative Ricci curvature. We examine the influence of large volume growth on these spaces and generalize some classical arguments from Riemannian geometry showing that when the volume growth is sufficiently large, then $(X,d,\\mathcal{H}^n)$ has finite topological type."
                    },
                    {
                        "title": "Volume growth and the topology of manifolds with nonnegative Ricci curvature",
                        "abstract": "Let $M^n$ be a complete, open Riemannian manifold with $\\Ric \\geq 0$. In 1994, Grigori Perelman showed that there exists a constant $\\delta_{n}>0$, depending only on the dimension of the manifold, such that if the volume growth satisfies $\\alpha_M := \\lim_{r \\to \\infty} \\frac{\\Vol(B_p(r))}{\\omega_n r^n} \\geq 1-\\delta_{n}$, then $M^n$ is contractible. Here we employ the techniques of Perelman to find specific lower bounds for the volume growth, $\\alpha(k,n)$, depending only on $k$ and $n$, which guarantee the individual $k$-homotopy group of $M^n$ is trivial."
                    },
                    {
                        "title": "Volume growth and the topology of Gromov-Hausdorff limits",
                        "abstract": "We examine topological properties of pointed metric measure spaces $(Y, p)$ that can be realized as the pointed Gromov-Hausdorff limit of a sequence of complete, Riemannian manifolds $\\{(M^n_i, p_i)\\}_{i=1}^{\\infty}$ with nonnegative Ricci curvature. Cheeger and Colding \\cite{ChCoI} showed that given such a sequence of Riemannian manifolds it is possible to define a measure $\\nu$ on the limit space $(Y, p)$. In the current work, we generalize previous results of the author to examine the relationship between the topology of $(Y, p)$ and the volume growth of $\\nu$. In particular, we prove a Abresch-Gromoll type excess estimate for triangles formed by limiting geodesics in the limit space. Assuming explicit volume growth lower bounds in the limit, we show that if $\\lim_{r \\to \\infty} \\frac{\\nu(B_p(r))}{\\omega_n r^n} > \\alpha(k,n)$, then the $k$-th group of $(Y,p)$ is trivial. The constants $\\alpha(k,n)$ are explicit and depend only on $n$, the dimension of the manifolds $\\{(M^n_i, p_i)\\}$, and $k$, the dimension of the homotopy in $(Y,p)$."
                    },
                    {
                        "title": "Intrinsic flat convergence with bounded Ricci curvature",
                        "abstract": "In this paper we address the relationship between Gromov-Hausdorff limits and intrinsic flat limits of complete Riemannian manifolds. In \\cite{SormaniWenger2010, SormaniWenger2011}, Sormani-Wenger show that for a sequence of Riemannian manifolds with nonnegative Ricci curvature, a uniform upper bound on diameter, and non-collapsed volume, the intrinsic flat limit exists and agrees with the Gromov-Hausdorff limit. This can be viewed as a non-cancellation theorem showing that for such sequences, points don't cancel each other out in the limit.   Here we prove a similar no-cancellation theorem, replacing the assumption of nonnegative Ricci curvature with a two-sided bound on Ricci curvature. This version corrects a mistake in the previous version of this paper (where we assume only an arbitrary lower Ricci bound) which was due to a crucial error in one of our supporting theorems for that argument."
                    },
                    {
                        "title": "Super Ricci flow for disjoint unions",
                        "abstract": "In this paper we consider compact, Riemannian manifolds $M_1, M_2$ each equipped with a one-parameter family of metrics $g_1(t), g_2(t)$ satisfying the Ricci flow equation. Motivated by a characterization of the super Ricci flow developed by McCann-Topping, we introduce the notion of a super Ricci flow for a family of distance metrics defined on the disjoint union $\\MM$. In particular, we show such a super Ricci flow property holds provided the distance function between points in $M_1$ and $M_2$ evolves by the heat equation. We also discuss possible applications and examples."
                    },
                    {
                        "title": "On the Size of a Ricci Flow Neckpinch via Optimal Transport",
                        "abstract": "In this paper we apply techniques from optimal transport to study the neckpinch examples of Angenent-Knopf which arise through the Ricci flow on $\\mathbb{S}^{n+1}$. In particular, we recover their proof of 'single-point pinching' along the flow. Using the methods of optimal transportation, we are able to remove the assumption of reflection symmetry for the metric. Our argument relies on the heuristic for weak Ricci flow proposed by McCann-Topping which characterizes super solutions of the Ricci flow by the contractivity of diffusions."
                    },
                    {
                        "title": "Leveraging free energy in pretraining model selection for improved fine-tuning",
                        "abstract": "Recent advances in artificial intelligence have been fueled by the development of foundation models such as BERT, GPT, T5, and Vision Transformers. These models are first pretrained on vast and diverse datasets and then adapted to specific downstream tasks, often with significantly less data. However, the mechanisms behind the success of this ubiquitous pretrain-then-adapt paradigm remain underexplored, particularly the characteristics of pretraining checkpoints that lend themselves to good downstream adaptation. We introduce a Bayesian model selection criterion, called the downstream free energy, which quantifies a checkpoint's adaptability by measuring the concentration of nearby favorable parameters for the downstream task. We demonstrate that this free energy criterion can be effectively implemented without access to the downstream data or prior knowledge of the downstream task. Furthermore, we provide empirical evidence that the free energy criterion reliably correlates with improved fine-tuning performance, offering a principled approach to predicting model adaptability."
                    },
                    {
                        "title": "On the appearance of Eisenstein series through degeneration",
                        "abstract": "Let $\\Gamma$ be a Fuchsian group of the first kind acting on the hyperbolic upper half plane $\\mathbb H$, and let $M = \\Gamma \\backslash \\mathbb H$ be the associated finite volume hyperbolic Riemann surface. If $\\gamma$ is parabolic, there is an associated (parabolic) Eisenstein series, which, by now, is a classical part of mathematical literature. If $\\gamma$ is hyperbolic, then, following ideas due to Kudla-Millson, there is a corresponding hyperbolic Eisenstein series. In this article, we study the limiting behavior of parabolic and hyperbolic Eisenstein series on a degenerating family of finite volume hyperbolic Riemann surfaces. In particular, we prove the following result. If $\\gamma \\in \\Gamma$ corresponds to a degenerating hyperbolic element, then a multiple of the associated hyperbolic Eisenstein series converges to parabolic Eisenstein series on the limit surface."
                    },
                    {
                        "title": "A Margin-based Multiclass Generalization Bound via Geometric Complexity",
                        "abstract": "There has been considerable effort to better understand the generalization capabilities of deep neural networks both as a means to unlock a theoretical understanding of their success as well as providing directions for further improvements. In this paper, we investigate margin-based multiclass generalization bounds for neural networks which rely on a recent complexity measure, the geometric complexity, developed for neural networks. We derive a new upper bound on the generalization error which scales with the margin-normalized geometric complexity of the network and which holds for a broad family of data distributions and model classes. Our generalization bound is empirically investigated for a ResNet-18 model trained with SGD on the CIFAR-10 and CIFAR-100 datasets with both original and random labels."
                    },
                    {
                        "title": "Three-Dimensional Alexandrov spaces with positive or nonnegative Ricci curvature",
                        "abstract": "We study closed three-dimensional Alexandrov spaces with a lower Ricci curvature bound in the $\\mathsf{CD}^*(K,N)$ sense, focusing our attention on those with positive or nonnegative Ricci curvature. First, we show that a closed three-dimensional $\\mathsf{CD}^*(2,3)$-Alexandrov space must be homeomorphic to a spherical space form or to the suspension of $\\mathbb{R}P^2$. We then classify closed three-dimensional $\\mathsf{CD}^*(0,3)$-Alexandrov spaces."
                    },
                    {
                        "title": "On Weak Super Ricci Flow through Neckpinch",
                        "abstract": "In this article, we study the Ricci flow neckpinch in the context of metric measure spaces. We introduce the notion of a Ricci flow metric measure spacetime and of a weak (refined) super Ricci flow associated to convex cost functions (cost functions which are increasing convex functions of the distance function). Our definition of a weak super Ricci flow is based on the coupled contraction property for suitably defined diffusions on maximal diffusion components. In our main theorem, we show that if a non-degenerate spherical neckpinch can be continued beyond the singular time by a smooth forward evolution then the corresponding Ricci flow metric measure spacetime through the singularity is a weak super Ricci flow for a (and therefore for all) convex cost functions if and only if the single point pinching phenomenon holds at singular times; i.e., if singularities form on a finite number of totally geodesic hypersurfaces of the form $\\{x\\} \\times \\sphere^n$. We also show the spacetime is a refined weak super Ricci flow if and only if the flow is a smooth Ricci flow with possibly singular final time."
                    },
                    {
                        "title": "COT-GAN: Generating Sequential Data via Causal Optimal Transport",
                        "abstract": "We introduce COT-GAN, an adversarial algorithm to train implicit generative models optimized for producing sequential data. The loss function of this algorithm is formulated using ideas from Causal Optimal Transport (COT), which combines classic optimal transport methods with an additional temporal causality constraint. Remarkably, we find that this causality condition provides a natural framework to parameterize the cost function that is learned by the discriminator as a robust (worst-case) distance, and an ideal mechanism for learning time dependent data distributions. Following Genevay et al.\\ (2018), we also include an entropic penalization term which allows for the use of the Sinkhorn algorithm when computing the optimal transport cost. Our experiments show effectiveness and stability of COT-GAN when generating both low- and high-dimensional time series data. The success of the algorithm also relies on a new, improved version of the Sinkhorn divergence which demonstrates less bias in learning."
                    },
                    {
                        "title": "The Geometric Occam's Razor Implicit in Deep Learning",
                        "abstract": "In over-parameterized deep neural networks there can be many possible parameter configurations that fit the training data exactly. However, the properties of these interpolating solutions are poorly understood. We argue that over-parameterized neural networks trained with stochastic gradient descent are subject to a Geometric Occam's Razor; that is, these networks are implicitly regularized by the geometric model complexity. For one-dimensional regression, the geometric model complexity is simply given by the arc length of the function. For higher-dimensional settings, the geometric model complexity depends on the Dirichlet energy of the function. We explore the relationship between this Geometric Occam's Razor, the Dirichlet energy and other known forms of implicit regularization. Finally, for ResNets trained on CIFAR-10, we observe that Dirichlet energy measurements are consistent with the action of this implicit Geometric Occam's Razor."
                    },
                    {
                        "title": "Geometric singularities and a flow tangent to the Ricci flow",
                        "abstract": "We consider a geometric flow introduced by Gigli and Mantegazza which, in the case of smooth compact manifolds with smooth metrics, is tangen- tial to the Ricci flow almost-everywhere along geodesics. To study spaces with geometric singularities, we consider this flow in the context of smooth manifolds with rough metrics with sufficiently regular heat kernels. On an appropriate non- singular open region, we provide a family of metric tensors evolving in time and provide a regularity theory for this flow in terms of the regularity of the heat kernel.   When the rough metric induces a metric measure space satisfying a Riemannian Curvature Dimension condition, we demonstrate that the distance induced by the flow is identical to the evolving distance metric defined by Gigli and Mantegazza on appropriate admissible points. Consequently, we demonstrate that a smooth compact manifold with a finite number of geometric conical singularities remains a smooth manifold with a smooth metric away from the cone points for all future times. Moreover, we show that the distance induced by the evolving metric tensor agrees with the flow of RCD(K, N) spaces defined by Gigli-Mantegazza."
                    },
                    {
                        "title": "Why neural networks find simple solutions: the many regularizers of geometric complexity",
                        "abstract": "In many contexts, simpler models are preferable to more complex models and the control of this model complexity is the goal for many methods in machine learning such as regularization, hyperparameter tuning and architecture design. In deep learning, it has been difficult to understand the underlying mechanisms of complexity control, since many traditional measures are not naturally suitable for deep neural networks. Here we develop the notion of geometric complexity, which is a measure of the variability of the model function, computed using a discrete Dirichlet energy. Using a combination of theoretical arguments and empirical results, we show that many common training heuristics such as parameter norm regularization, spectral norm regularization, flatness regularization, implicit gradient regularization, noise regularization and the choice of parameter initialization all act to control geometric complexity, providing a unifying framework in which to characterize the behavior of deep learning models."
                    },
                    {
                        "title": "Unified Functional Hashing in Automatic Machine Learning",
                        "abstract": "The field of Automatic Machine Learning (AutoML) has recently attained impressive results, including the discovery of state-of-the-art machine learning solutions, such as neural image classifiers. This is often done by applying an evolutionary search method, which samples multiple candidate solutions from a large space and evaluates the quality of each candidate through a long training process. As a result, the search tends to be slow. In this paper, we show that large efficiency gains can be obtained by employing a fast unified functional hash, especially through the functional equivalence caching technique, which we also present. The central idea is to detect by hashing when the search method produces equivalent candidates, which occurs very frequently, and this way avoid their costly re-evaluation. Our hash is \"functional\" in that it identifies equivalent candidates even if they were represented or coded differently, and it is \"unified\" in that the same algorithm can hash arbitrary representations; e.g. compute graphs, imperative code, or lambda functions. As evidence, we show dramatic improvements on multiple AutoML domains, including neural architecture search and algorithm discovery. Finally, we consider the effect of hash collisions, evaluation noise, and search distribution through empirical analysis. Altogether, we hope this paper may serve as a guide to hashing techniques in AutoML."
                    }
                ]
            },
            "27165d0d-8828-478f-9a4d-dfe613de17aa": {
                "pk": "27165d0d-8828-478f-9a4d-dfe613de17aa",
                "name": "Benoit Dherin",
                "collaborators": [
                    "Alberto S. Cattaneo",
                    "Alan Weinstein",
                    "Igor Mencattini",
                    "Giovanni Felder",
                    "David G. T. Barrett",
                    "Camilo Arias Abad",
                    "Marius Crainic",
                    "Alejandro Cabrera",
                    "Friedrich Wagemann",
                    "Mihaela Rosca"
                ],
                "domain": [
                    "Poisson Geometry",
                    "Symplectic Geometry",
                    "Neural Networks",
                    "Implicit Regularization"
                ],
                "publications": [
                    {
                        "title": "The Universal Generating Function of Analytical Poisson Structures",
                        "abstract": "The notion of generating functions of Poisson structures was first studied in math.SG/0312380.They are special functions which induce, on open subsets of $\\R^d$, a Poisson structure together with the local symplectic groupoid integrating it. A universal generating function was provided in terms of a formal power series coming from Kontsevich star product. The present article proves that this universal generating function converges for analytical Poisson structures and compares the induced local symplectic groupoid with the phase space of Karasev--Maslov."
                    },
                    {
                        "title": "Implicit biases in multitask and continual learning from a backward error analysis perspective",
                        "abstract": "Using backward error analysis, we compute implicit training biases in multitask and continual learning settings for neural networks trained with stochastic gradient descent. In particular, we derive modified losses that are implicitly minimized during training. They have three terms: the original loss, accounting for convergence, an implicit flatness regularization term proportional to the learning rate, and a last term, the conflict term, which can theoretically be detrimental to both convergence and implicit regularization. In multitask, the conflict term is a well-known quantity, measuring the gradient alignment between the tasks, while in continual learning the conflict term is a new quantity in deep learning optimization, although a basic tool in differential geometry: The Lie bracket between the task gradients."
                    },
                    {
                        "title": "Geometric States",
                        "abstract": "We introduce a special family of distributional alpha-densities and give a transversality criterion stating when their product is defined, closely related to Hormander's criterion for general distributions. Moreover, we show that for the subspace of distributional half-densities in this family the distribution product naturally yields a pairing that extends the usual one on smooth half-densities."
                    },
                    {
                        "title": "Symplectic Microgeometry III: Monoids",
                        "abstract": "We show that the category of Poisson manifolds and Poisson maps, the category of symplectic microgroupoids and lagrangian submicrogroupoids (as morphisms), and the category of monoids and monoid morphisms in the microsymplectic category are equivalent symmetric monoidal categories."
                    },
                    {
                        "title": "Tensor products of representations up to homotopy",
                        "abstract": "We study the construction of tensor products of representations up to homotopy, which are the A-infinity version of ordinary representations. We provide formulas for the construction of tensor products of representations up to homotopy and of morphisms between them, and show that these formulas give the homotopy category a monoidal structure which is uniquely defined up to equivalence."
                    },
                    {
                        "title": "Formal Lagrangian Operad",
                        "abstract": "Given a symplectic manifold $M$, we may define an operad structure on the the spaces $\\op^k$ of the Lagrangian submanifolds of $(\\bar{M})^k\\times M$ via symplectic reduction. If $M$ is also a symplectic groupoid, then its multiplication space is an associative product in this operad. Following this idea, we provide a deformation theory for symplectic groupoids analog to the deformation theory of algebras. It turns out that the semi-classical part of Kontsevich's deformation of $C^\\infty(\\R^d)$ is a deformation of the trivial symplectic groupoid structure of $T^*\\R^d$."
                    },
                    {
                        "title": "Cotangent Microbundle Category, I",
                        "abstract": "We define a local version of the extended symplectic category, the cotangent microbundle category, MiC, which turns out to be a true monoidal category. We show that a monoid in this category induces a Poisson manifold together with the local symplectic groupoid integrating it. Moreover, we prove that monoid morphisms produce Poisson maps between the induced Poisson manifolds in a functorial way. This gives a functor between the category of monoids in MiC and the category of Poisson manifolds and Poisson maps. Conversely, the semi-classical part of the Kontsevich star-product associated to a real-analytic Poisson structure on an open subset of R^n produces a monoid in MiC."
                    },
                    {
                        "title": "Symplectic Microgeometry I: Micromorphisms",
                        "abstract": "We introduce the notion of symplectic microfolds and symplectic micromorphisms between them. They form a monoidal category, which is a version of the \"category\" of symplectic manifolds and canonical relations obtained by localizing them around lagrangian submanifolds in the spirit of Milnor's microbundles."
                    },
                    {
                        "title": "Symplectic Microgeometry II: Generating functions",
                        "abstract": "We adapt the notion of generating functions for lagrangian submanifolds to symplectic microgeometry. We show that a symplectic micromorphism always admits a global generating function. As an application, we describe hamiltonian flows as special symplectic micromorphisms whose local generating functions are the solutions of Hamilton-Jacobi equations. We obtain a purely categorical formulation of the temporal evolution in classical mechanics."
                    },
                    {
                        "title": "Integration of Lie Algebroid Comorphisms",
                        "abstract": "We show that the path construction integration of Lie algebroids by Lie groupoids is an actual equivalence from the category of integrable Lie algebroids and complete Lie algebroid comorphisms to the category of source 1-connected Lie groupoids and Lie groupoid comorphisms. This allows us to construct an actual symplectization functor in Poisson geometry. We include examples to show that the integrability of comorphisms and Poisson maps may not hold in the absence of a completeness assumption."
                    },
                    {
                        "title": "Quantizations of Momentum Maps and G-Systems",
                        "abstract": "In this note, we give an explicit formula for a family of deformation quantizations for the momentum map associated with the cotangent lift of a Lie group action on Rd. This family of quantizations is parametrized by the formal G-systems introduced by the authors in a previous paper and allows us to obtain classical invariant Hamiltonians that quantize without anomalies with respect to the quantizations of the action prescribed by the formal G-systems."
                    },
                    {
                        "title": "G-Systems and Deformation of G-Actions On Rd",
                        "abstract": "Given a (smooth) action of a Lie group G on Rd we construct a DGA whose Maurer-Cartan elements are in one to one correspondence with some class of defomations of the (induced) G-action on the ring of formal power series with coefficients in the ring of smooth functions on Rd. In the final part of this note we discuss the cohomological obstructions to the existence and to the uniqueness (in a sense to be clarified) of such deformations."
                    },
                    {
                        "title": "Formal symplectic groupoid",
                        "abstract": "The multiplicative structure of the trivial symplectic groupoid over $\\mathbb R^d$ associated to the zero Poisson structure can be expressed in terms of a generating function. We address the problem of deforming such a generating function in the direction of a non-trivial Poisson structure so that the multiplication remains associative. We prove that such a deformation is unique under some reasonable conditions and we give the explicit formula for it. This formula turns out to be the semi-classical approximation of Kontsevich's deformation formula. For the case of a linear Poisson structure, the deformed generating function reduces exactly to the CBH formula of the associated Lie algebra. The methods used to prove existence are interesting in their own right as they come from an at first sight unrelated domain of mathematics: the Runge--Kutta theory of the numeric integration of ODE's."
                    },
                    {
                        "title": "Implicit Gradient Regularization",
                        "abstract": "Gradient descent can be surprisingly good at optimizing deep neural networks without overfitting and without explicit regularization. We find that the discrete steps of gradient descent implicitly regularize models by penalizing gradient descent trajectories that have large loss gradients. We call this Implicit Gradient Regularization (IGR) and we use backward error analysis to calculate the size of this regularization. We confirm empirically that implicit gradient regularization biases gradient descent toward flat minima, where test errors are small and solutions are robust to noisy parameter perturbations. Furthermore, we demonstrate that the implicit gradient regularization term can be used as an explicit regularizer, allowing us to control this gradient regularization directly. More broadly, our work indicates that backward error analysis is a useful theoretical approach to the perennial question of how learning rate, model size, and parameter regularization interact to determine the properties of overparameterized models optimized with gradient descent."
                    },
                    {
                        "title": "Symplectic Microgeometry IV: Quantization",
                        "abstract": "We construct a special class of semiclassical Fourier integral operators whose wave fronts are symplectic micromorphisms. These operators have very good properties: they form a category on which the wave front map becomes a functor into the cotangent microbundle category, and they admit a total symbol calculus in terms of symplectic micromorphisms enhanced with half-density germs. This new operator category encompasses the semi-classical pseudo-differential calculus and offers a functorial framework for the semi-classical analysis of the Schr\\\"odinger equation. We also comment on applications to classical and quantum mechanics as well as to a functorial and geometrical approach to the quantization of Poisson manifolds."
                    },
                    {
                        "title": "Formal symplectic realizations",
                        "abstract": "We study the relationship between several constructions of symplectic realizations of a given Poisson manifold. Our main result is a general formula for a formal symplectic realization in the case of an arbitrary Poisson structure on $\\R^n$. This formula is expressed in terms of rooted trees and elementary differentials, building on the work of Butcher, and the coefficients are shown to be a generalization of Bernoulli numbers appearing in the linear Poisson case. We also show that this realization coincides with a formal version of the original construction of Weinstein, when suitably put in global Darboux form, and with the realization coming from tree-level part of Kontsevich's star product. We provide a simple iterated integral expression for the relevant coefficients and show that they coincide with underlying Kontsevich weights."
                    },
                    {
                        "title": "Quantization of (volume-preserving) actions on R^d",
                        "abstract": "We associate a space of (formal) representations on the space of h-formal power series with coefficients in the space of smooth functions on Rd (which we call quantizations) with an action of a group on Rd by smooth diffeomorphisms. If the action is further volume preserving, these quantizations can be realized as unitary representations on square summable functions on Rd by bounded h-dependent Fourier integral operators, the formal case corresponding to the asymptotics in the limit h going to zero. We construct DGAs controlling these quantizations and prove existence and rigidity results for them."
                    },
                    {
                        "title": "Deformation quantization of Leibniz algebras",
                        "abstract": "This paper has two parts. The first part is a review and extension of the methods of integration of Leibniz algebras into Lie racks, including as new feature a new way of integrating 2-cocycles (see Lemma 3.9).   In the second part, we use the local integration of a Leibniz algebra h using a Baker-Campbell-Hausdorff type formula in order to deformation quantize its linear dual h^*. More precisely, we define a natural rack product on the set of exponential functions which extends to a rack action on C^{\\infty}(h^*)."
                    },
                    {
                        "title": "Corridor Geometry in Gradient-Based Optimization",
                        "abstract": "We characterize regions of a loss surface as corridors when the continuous curves of steepest descent -- the solutions of the gradient flow -- become straight lines. We show that corridors provide insights into gradient-based optimization, since corridors are exactly the regions where gradient descent and the gradient flow follow the same trajectory, while the loss decreases linearly. As a result, inside corridors there are no implicit regularization effects or training instabilities that have been shown to occur due to the drift between gradient descent and the gradient flow. Using the loss linear decrease on corridors, we devise a learning rate adaptation scheme for gradient descent; we call this scheme Corridor Learning Rate (CLR). The CLR formulation coincides with a special case of Polyak step-size, discovered in the context of convex optimization. The Polyak step-size has been shown recently to have also good convergence properties for neural networks; we further confirm this here with results on CIFAR-10 and ImageNet."
                    },
                    {
                        "title": "The Geometric Occam's Razor Implicit in Deep Learning",
                        "abstract": "In over-parameterized deep neural networks there can be many possible parameter configurations that fit the training data exactly. However, the properties of these interpolating solutions are poorly understood. We argue that over-parameterized neural networks trained with stochastic gradient descent are subject to a Geometric Occam's Razor; that is, these networks are implicitly regularized by the geometric model complexity. For one-dimensional regression, the geometric model complexity is simply given by the arc length of the function. For higher-dimensional settings, the geometric model complexity depends on the Dirichlet energy of the function. We explore the relationship between this Geometric Occam's Razor, the Dirichlet energy and other known forms of implicit regularization. Finally, for ResNets trained on CIFAR-10, we observe that Dirichlet energy measurements are consistent with the action of this implicit Geometric Occam's Razor."
                    }
                ]
            },
            "854460b9-c2e0-460a-a158-75e7b206d5e4": {
                "pk": "854460b9-c2e0-460a-a158-75e7b206d5e4",
                "name": "Javier Gonzalvo",
                "collaborators": [
                    "Hanna Mazzawi",
                    "Vladimir Macko",
                    "Charles Weill",
                    "Scott Yak",
                    "Michael Munn",
                    "Benoit Dherin",
                    "Vitaly Kuznetsov",
                    "Scott Yang",
                    "Eugen Hotaj",
                    "Ghassen Jerfel"
                ],
                "domain": [
                    "Neural Architecture Search",
                    "Deep Learning",
                    "Generalization",
                    "Ensemble Learning"
                ],
                "publications": [
                    {
                        "title": "Improving Neural Architecture Search Image Classifiers via Ensemble Learning",
                        "abstract": "Finding the best neural network architecture requires significant time, resources, and human expertise. These challenges are partially addressed by neural architecture search (NAS) which is able to find the best convolutional layer or cell that is then used as a building block for the network. However, once a good building block is found, manual design is still required to assemble the final architecture as a combination of multiple blocks under a predefined parameter budget constraint. A common solution is to stack these blocks into a single tower and adjust the width and depth to fill the parameter budget. However, these single tower architectures may not be optimal. Instead, in this paper we present the AdaNAS algorithm, that uses ensemble techniques to compose a neural network as an ensemble of smaller networks automatically. Additionally, we introduce a novel technique based on knowledge distillation to iteratively train the smaller networks using the previous ensemble as a teacher. Our experiments demonstrate that ensembles of networks improve accuracy upon a single neural network while keeping the same number of parameters. Our models achieve comparable results with the state-of-the-art on CIFAR-10 and sets a new state-of-the-art on CIFAR-100."
                    },
                    {
                        "title": "Towards Task and Architecture-Independent Generalization Gap Predictors",
                        "abstract": "Can we use deep learning to predict when deep learning works? Our results suggest the affirmative. We created a dataset by training 13,500 neural networks with different architectures, on different variations of spiral datasets, and using different optimization parameters. We used this dataset to train task-independent and architecture-independent generalization gap predictors for those neural networks. We extend Jiang et al. (2018) to also use DNNs and RNNs and show that they outperform the linear model, obtaining $R^2=0.965$. We also show results for architecture-independent, task-independent, and out-of-distribution generalization gap prediction tasks. Both DNNs and RNNs consistently and significantly outperform linear models, with RNNs obtaining $R^2=0.584$."
                    },
                    {
                        "title": "A Margin-based Multiclass Generalization Bound via Geometric Complexity",
                        "abstract": "There has been considerable effort to better understand the generalization capabilities of deep neural networks both as a means to unlock a theoretical understanding of their success as well as providing directions for further improvements. In this paper, we investigate margin-based multiclass generalization bounds for neural networks which rely on a recent complexity measure, the geometric complexity, developed for neural networks. We derive a new upper bound on the generalization error which scales with the margin-normalized geometric complexity of the network and which holds for a broad family of data distributions and model classes. Our generalization bound is empirically investigated for a ResNet-18 model trained with SGD on the CIFAR-10 and CIFAR-100 datasets with both original and random labels."
                    },
                    {
                        "title": "AdaNet: A Scalable and Flexible Framework for Automatically Learning Ensembles",
                        "abstract": "AdaNet is a lightweight TensorFlow-based (Abadi et al., 2015) framework for automatically learning high-quality ensembles with minimal expert intervention. Our framework is inspired by the AdaNet algorithm (Cortes et al., 2017) which learns the structure of a neural network as an ensemble of subnetworks. We designed it to: (1) integrate with the existing TensorFlow ecosystem, (2) offer sensible default search spaces to perform well on novel datasets, (3) present a flexible API to utilize expert information when available, and (4) efficiently accelerate training with distributed CPU, GPU, and TPU hardware. The code is open-source and available at: https://github.com/tensorflow/adanet."
                    }
                ]
            }
        }
    },
    "2402.16788": {
        "paper_data": {
            "title": "Why Transformers Need Adam: A Hessian Perspective",
            "url": "http://arxiv.org/abs/2402.16788v4",
            "arxiv_id": "2402.16788",
            "authors": [
                "Yushun Zhang",
                "Congliang Chen",
                "Tian Ding",
                "Ziniu Li",
                "Ruoyu Sun",
                "Zhi-Quan Luo"
            ],
            "abstract": "SGD performs worse than Adam by a significant margin on Transformers, but the reason remains unclear. In this work, we provide an explanation through the lens of Hessian: (i) Transformers are \"heterogeneous\": the Hessian spectrum across parameter blocks vary dramatically, a phenomenon we call \"block heterogeneity\"; (ii) Heterogeneity hampers SGD: SGD performs worse than Adam on problems with block heterogeneity. To validate (i) and (ii), we check various Transformers, CNNs, MLPs, and quadratic problems, and find that SGD can perform on par with Adam on problems without block heterogeneity, but performs worse than Adam when the heterogeneity exists. Our initial theoretical analysis indicates that SGD performs worse because it applies one single learning rate to all blocks, which cannot handle the heterogeneity among blocks. This limitation could be ameliorated if we use coordinate-wise learning rates, as designed in Adam.",
            "introduction": "   1 Introduction  Transformers [83] have become a major workhorse behind AI development (e.g., [1]). However, the understanding of Transformer training remains limited. For instance, Transformer training largely relies on the Adam optimizer [45, 57]. In contrast, stochastic gradient descent with momentum (SGD)222We introduce the update rules of Adam(W) and SGD in Appendix C.1., the de-facto optimizer for convolution neural networks (CNNs) [49], performs significantly worse than Adam on Transformers (e.g., Figure LABEL:fig_blockwise_spectrum). Yet, the reasons behind this performance gap remain unclear. Understanding why SGD performs worse than Adam on Transformers is an intriguing question. First, from a theoretical perspective, this can help us better understand the training of Transformers and more generally, neural networks. Second, from a computational perspective, the understanding may inspire the design of better algorithms for training neural networks.   In this work, we explore why SGD largely underperforms Adam on Transformers through the lens of Hessian. We start by investigating the full Hessian spectrum of Transformers, i.e., the full eigenvalue density of Hessian (see Figure LABEL:fig_full_spectrum). By theory, the full Hessian spectrum largely determines the behavior of gradient-based methods [63, 33, 79, 37], so we suspect it may also help explain SGD\u2019s unsatisfactory performance. Using tools from numerical linear algebra [7], we empirically compare the full spectra of CNNs (where SGD is on par with Adam) and those of Transformers (where SGD largely lags behind Adam). Unfortunately, as shown in Figure LABEL:fig_full_spectrum, the spectra for CNNs and Transformers are often largely similar despite the different optimizer behaviors. As such, we have not identified critical features in the full Hessian spectra associated with the gap between Adam and SGD on Transformers. To reveal the cause, a more fine-grained investigation into the Hessian is needed.     What would cause SGD to perform significantly worse than Adam on Transformers, but not on CNNs? By dissecting the structures of CNNs and Transformers, we notice that CNNs are constructed by the repetitive stacking of similar parameter blocks (convolution layers), while Transformers involve the non-sequential stacking of disparate parameter blocks (e.g. Query, Key, Value, Output projection blocks in attention and MLP layers). We hypothesize that these architectural differences might lead to different optimization properties. Intuitively, disparate parameter blocks contribute differently to the overall loss. So each block might benefit from a specialized treatment by optimizers, a flexibility offered by Adam but not by SGD. This observation motivates us to investigate the Hessian spectrum of each parameter block, which we refer to as the blockwise Hessian spectrum.   By inspecting the blockwise Hessian spectrum, we discover a possible explanation for why SGD is worse: the \u201cheterogeneity\u201d inherent in Transformers. We provide both empirical and theoretical evidence to support this explanation. Our contributions can be summarized as follows:   \u2022  Why SGD underperforms Adam on Transformers. We explain why SGD is worse than Adam on Transformers by examining the blockwise Hessian spectrum. First, we identify a phenomenon called \u201cblock heterogeneity\", which refers to the large differences in the Hessian spectra across parameter blocks. This block heterogeneity is observed in all examined Transformers but not in CNNs. Second, we verify that block heterogeneity hinders SGD. Across various Transformers, CNNs, and MLPs, we show that SGD consistently",
            "references": [
                {
                    "title": "Mamba: Linear-Time Sequence Modeling with Selective State Spaces",
                    "abstract": "Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5$\\times$ higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation."
                },
                {
                    "title": "Closing the Gap Between the Upper Bound and the Lower Bound of Adam's Iteration Complexity",
                    "abstract": "Recently, Arjevani et al. [1] established a lower bound of iteration complexity for the first-order optimization under an $L$-smooth condition and a bounded noise variance assumption. However, a thorough review of existing literature on Adam's convergence reveals a noticeable gap: none of them meet the above lower bound. In this paper, we close the gap by deriving a new convergence guarantee of Adam, with only an $L$-smooth condition and a bounded noise variance assumption. Our results remain valid across a broad spectrum of hyperparameters. Especially with properly chosen hyperparameters, we derive an upper bound of the iteration complexity of Adam and show that it meets the lower bound for first-order optimizers. To the best of our knowledge, this is the first to establish such a tight upper bound for Adam's convergence. Our proof utilizes novel techniques to handle the entanglement between momentum and adaptive learning rate and to convert the first-order term in the Descent Lemma to the gradient norm, which may be of independent interest."
                },
                {
                    "title": "Linear attention is (maybe) all you need (to understand transformer optimization)",
                    "abstract": "Transformer training is notoriously difficult, requiring a careful design of optimizers and use of various heuristics. We make progress towards understanding the subtleties of training Transformers by carefully studying a simple yet canonical linearized shallow Transformer model. Specifically, we train linear Transformers to solve regression tasks, inspired by J.~von Oswald et al.~(ICML 2023), and K.~Ahn et al.~(NeurIPS 2023). Most importantly, we observe that our proposed linearized models can reproduce several prominent aspects of Transformer training dynamics. Consequently, the results obtained in this paper suggest that a simple linearized Transformer model could actually be a valuable, realistic abstraction for understanding Transformer optimization."
                },
                {
                    "title": "Small-scale proxies for large-scale Transformer training instabilities",
                    "abstract": "Teams that have trained large Transformer-based models have reported training instabilities at large scale that did not appear when training with the same hyperparameters at smaller scales. Although the causes of such instabilities are of scientific interest, the amount of resources required to reproduce them has made investigation difficult. In this work, we seek ways to reproduce and study training stability and instability at smaller scales. First, we focus on two sources of training instability described in previous work: the growth of logits in attention layers (Dehghani et al., 2023) and divergence of the output logits from the log probabilities (Chowdhery et al., 2022). By measuring the relationship between learning rate and loss across scales, we show that these instabilities also appear in small models when training at high learning rates, and that mitigations previously employed at large scales are equally effective in this regime. This prompts us to investigate the extent to which other known optimizer and model interventions influence the sensitivity of the final loss to changes in the learning rate. To this end, we study methods such as warm-up, weight decay, and the $\\mu$Param (Yang et al., 2022), and combine techniques to train small models that achieve similar losses across orders of magnitude of learning rate variation. Finally, to conclude our exploration we study two cases where instabilities can be predicted before they emerge by examining the scaling behavior of model activation and gradient norms."
                },
                {
                    "title": "Baichuan 2: Open Large-scale Language Models",
                    "abstract": "Large language models (LLMs) have demonstrated remarkable performance on a variety of natural language tasks based on just a few examples of natural language instructions, reducing the need for extensive feature engineering. However, most powerful LLMs are closed-source or limited in their capability for languages other than English. In this technical report, we present Baichuan 2, a series of large-scale multilingual language models containing 7 billion and 13 billion parameters, trained from scratch, on 2.6 trillion tokens. Baichuan 2 matches or outperforms other open-source models of similar size on public benchmarks like MMLU, CMMLU, GSM8K, and HumanEval. Furthermore, Baichuan 2 excels in vertical domains such as medicine and law. We will release all pre-training model checkpoints to benefit the research community in better understanding the training dynamics of Baichuan 2."
                },
                {
                    "title": "Full Parameter Fine-tuning for Large Language Models with Limited Resources",
                    "abstract": "Large Language Models (LLMs) have revolutionized Natural Language Processing (NLP) but demand massive GPU resources for training. Lowering the threshold for LLMs training would encourage greater participation from researchers, benefiting both academia and society. While existing approaches have focused on parameter-efficient fine-tuning, which tunes or adds a small number of parameters, few have addressed the challenge of tuning the full parameters of LLMs with limited resources. In this work, we propose a new optimizer, LOw-Memory Optimization (LOMO), which fuses the gradient computation and the parameter update in one step to reduce memory usage. By integrating LOMO with existing memory saving techniques, we reduce memory usage to 10.8% compared to the standard approach (DeepSpeed solution). Consequently, our approach enables the full parameter fine-tuning of a 65B model on a single machine with 8 RTX 3090, each with 24GB memory.Code and data are available at https://github.com/OpenLMLab/LOMO."
                },
                {
                    "title": "Toward Understanding Why Adam Converges Faster Than SGD for Transformers",
                    "abstract": "While stochastic gradient descent (SGD) is still the most popular optimization algorithm in deep learning, adaptive algorithms such as Adam have established empirical advantages over SGD in some deep learning applications such as training transformers. However, it remains a question that why Adam converges significantly faster than SGD in these scenarios. In this paper, we propose one explanation of why Adam converges faster than SGD using a new concept directional sharpness. We argue that the performance of optimization algorithms is closely related to the directional sharpness of the update steps, and show SGD has much worse directional sharpness compared to adaptive algorithms. We further observe that only a small fraction of the coordinates causes the bad sharpness and slow convergence of SGD, and propose to use coordinate-wise clipping as a solution to SGD and other optimization algorithms. We demonstrate the effect of coordinate-wise clipping on sharpness reduction and speeding up the convergence of optimization algorithms under various settings. We show that coordinate-wise clipping improves the local loss reduction when only a small fraction of the coordinates has bad sharpness. We conclude that the sharpness reduction effect of adaptive coordinate-wise scaling is the reason for Adam's success in practice and suggest the use of coordinate-wise clipping as a universal technique to speed up deep learning optimization."
                },
                {
                    "title": "Convergence of AdaGrad for Non-convex Objectives: Simple Proofs and Relaxed Assumptions",
                    "abstract": "We provide a simple convergence proof for AdaGrad optimizing non-convex objectives under only affine noise variance and bounded smoothness assumptions. The proof is essentially based on a novel auxiliary function $\\xi$ that helps eliminate the complexity of handling the correlation between the numerator and denominator of AdaGrad's update. Leveraging simple proofs, we are able to obtain tighter results than existing results \\citep{faw2022power} and extend the analysis to several new and important cases. Specifically, for the over-parameterized regime, we show that AdaGrad needs only $\\mathcal{O}(\\frac{1}{\\varepsilon^2})$ iterations to ensure the gradient norm smaller than $\\varepsilon$, which matches the rate of SGD and significantly tighter than existing rates $\\mathcal{O}(\\frac{1}{\\varepsilon^4})$ for AdaGrad. We then discard the bounded smoothness assumption and consider a realistic assumption on smoothness called $(L_0,L_1)$-smooth condition, which allows local smoothness to grow with the gradient norm. Again based on the auxiliary function $\\xi$, we prove that AdaGrad succeeds in converging under $(L_0,L_1)$-smooth condition as long as the learning rate is lower than a threshold. Interestingly, we further show that the requirement on learning rate under the $(L_0,L_1)$-smooth condition is necessary via proof by contradiction, in contrast with the case of uniform smoothness conditions where convergence is guaranteed regardless of learning rate choices. Together, our analyses broaden the understanding of AdaGrad and demonstrate the power of the new auxiliary function in the investigations of AdaGrad."
                },
                {
                    "title": "Fine-Tuning Language Models with Just Forward Passes",
                    "abstract": "Fine-tuning language models (LMs) has yielded success on diverse downstream tasks, but as LMs grow in size, backpropagation requires a prohibitively large amount of memory. Zeroth-order (ZO) methods can in principle estimate gradients using only two forward passes but are theorized to be catastrophically slow for optimizing large models. In this work, we propose a memory-efficient zerothorder optimizer (MeZO), adapting the classical ZO-SGD method to operate in-place, thereby fine-tuning LMs with the same memory footprint as inference. For example, with a single A100 80GB GPU, MeZO can train a 30-billion parameter model, whereas fine-tuning with backpropagation can train only a 2.7B LM with the same budget. We conduct comprehensive experiments across model types (masked and autoregressive LMs), model scales (up to 66B), and downstream tasks (classification, multiple-choice, and generation). Our results demonstrate that (1) MeZO significantly outperforms in-context learning and linear probing; (2) MeZO achieves comparable performance to fine-tuning with backpropagation across multiple tasks, with up to 12x memory reduction and up to 2x GPU-hour reduction in our implementation; (3) MeZO is compatible with both full-parameter and parameter-efficient tuning techniques such as LoRA and prefix tuning; (4) MeZO can effectively optimize non-differentiable objectives (e.g., maximizing accuracy or F1). We support our empirical findings with theoretical insights, highlighting how adequate pre-training and task prompts enable MeZO to fine-tune huge models, despite classical ZO analyses suggesting otherwise."
                },
                {
                    "title": "Sophia: A Scalable Stochastic Second-order Optimizer for Language Model Pre-training",
                    "abstract": "Given the massive cost of language model pre-training, a non-trivial improvement of the optimization algorithm would lead to a material reduction on the time and cost of training. Adam and its variants have been state-of-the-art for years, and more sophisticated second-order (Hessian-based) optimizers often incur too much per-step overhead. In this paper, we propose Sophia, Second-order Clipped Stochastic Optimization, a simple scalable second-order optimizer that uses a light-weight estimate of the diagonal Hessian as the pre-conditioner. The update is the moving average of the gradients divided by the moving average of the estimated Hessian, followed by element-wise clipping. The clipping controls the worst-case update size and tames the negative impact of non-convexity and rapid change of Hessian along the trajectory. Sophia only estimates the diagonal Hessian every handful of iterations, which has negligible average per-step time and memory overhead. On language modeling with GPT models of sizes ranging from 125M to 1.5B, Sophia achieves a 2x speed-up compared to Adam in the number of steps, total compute, and wall-clock time, achieving the same perplexity with 50% fewer steps, less total compute, and reduced wall-clock time. Theoretically, we show that Sophia, in a much simplified setting, adapts to the heterogeneous curvatures in different parameter dimensions, and thus has a run-time bound that does not depend on the condition number of the loss."
                },
                {
                    "title": "Convergence of Adam Under Relaxed Assumptions",
                    "abstract": "In this paper, we provide a rigorous proof of convergence of the Adaptive Moment Estimate (Adam) algorithm for a wide class of optimization objectives. Despite the popularity and efficiency of the Adam algorithm in training deep neural networks, its theoretical properties are not yet fully understood, and existing convergence proofs require unrealistically strong assumptions, such as globally bounded gradients, to show the convergence to stationary points. In this paper, we show that Adam provably converges to $\\epsilon$-stationary points with $\\mathcal{O}(\\epsilon^{-4})$ gradient complexity under far more realistic conditions. The key to our analysis is a new proof of boundedness of gradients along the optimization trajectory of Adam, under a generalized smoothness assumption according to which the local smoothness (i.e., Hessian norm when it exists) is bounded by a sub-quadratic function of the gradient norm. Moreover, we propose a variance-reduced version of Adam with an accelerated gradient complexity of $\\mathcal{O}(\\epsilon^{-3})$."
                },
                {
                    "title": "Noise Is Not the Main Factor Behind the Gap Between SGD and Adam on Transformers, but Sign Descent Might Be",
                    "abstract": "The success of the Adam optimizer on a wide array of architectures has made it the default in settings where stochastic gradient descent (SGD) performs poorly. However, our theoretical understanding of this discrepancy is lagging, preventing the development of significant improvements on either algorithm. Recent work advances the hypothesis that Adam and other heuristics like gradient clipping outperform SGD on language tasks because the distribution of the error induced by sampling has heavy tails. This suggests that Adam outperform SGD because it uses a more robust gradient estimate. We evaluate this hypothesis by varying the batch size, up to the entire dataset, to control for stochasticity. We present evidence that stochasticity and heavy-tailed noise are not major factors in the performance gap between SGD and Adam. Rather, Adam performs better as the batch size increases, while SGD is less effective at taking advantage of the reduction in noise. This raises the question as to why Adam outperforms SGD in the full-batch setting. Through further investigation of simpler variants of SGD, we find that the behavior of Adam with large batches is similar to sign descent with momentum."
                },
                {
                    "title": "A Theory on Adam Instability in Large-Scale Machine Learning",
                    "abstract": "We present a theory for the previously unexplained divergent behavior noticed in the training of large language models. We argue that the phenomenon is an artifact of the dominant optimization algorithm used for training, called Adam. We observe that Adam can enter a state in which the parameter update vector has a relatively large norm and is essentially uncorrelated with the direction of descent on the training loss landscape, leading to divergence. This artifact is more likely to be observed in the training of a deep model with a large batch size, which is the typical setting of large-scale language model training. To argue the theory, we present observations from the training runs of the language models of different scales: 7 billion, 30 billion, 65 billion, and 546 billion parameters."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Stabilizing Transformer Training by Preventing Attention Entropy Collapse",
                    "abstract": "Training stability is of great importance to Transformers. In this work, we investigate the training dynamics of Transformers by examining the evolution of the attention layers. In particular, we track the attention entropy for each attention head during the course of training, which is a proxy for model sharpness. We identify a common pattern across different architectures and tasks, where low attention entropy is accompanied by high training instability, which can take the form of oscillating loss or divergence. We denote the pathologically low attention entropy, corresponding to highly concentrated attention scores, as $\\textit{entropy collapse}$. As a remedy, we propose $\\sigma$Reparam, a simple and efficient solution where we reparametrize all linear layers with spectral normalization and an additional learned scalar. We demonstrate that $\\sigma$Reparam successfully prevents entropy collapse in the attention layers, promoting more stable training. Additionally, we prove a tight lower bound of the attention entropy, which decreases exponentially fast with the spectral norm of the attention logits, providing additional motivation for our approach. We conduct experiments with $\\sigma$Reparam on image classification, image self-supervised learning, machine translation, speech recognition, and language modeling tasks. We show that $\\sigma$Reparam provides stability and robustness with respect to the choice of hyperparameters, going so far as enabling training (a) a Vision Transformer {to competitive performance} without warmup, weight decay, layer normalization or adaptive optimizers; (b) deep architectures in machine translation and (c) speech recognition to competitive performance without warmup and adaptive optimizers. Code is available at \\url{https://github.com/apple/ml-sigma-reparam}."
                },
                {
                    "title": "Scaling Vision Transformers to 22 Billion Parameters",
                    "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                },
                {
                    "title": "How Does Adaptive Optimization Impact Local Neural Network Geometry?",
                    "abstract": "Adaptive optimization methods are well known to achieve superior convergence relative to vanilla gradient methods. The traditional viewpoint in optimization, particularly in convex optimization, explains this improved performance by arguing that, unlike vanilla gradient schemes, adaptive algorithms mimic the behavior of a second-order method by adapting to the global geometry of the loss function. We argue that in the context of neural network optimization, this traditional viewpoint is insufficient. Instead, we advocate for a local trajectory analysis. For iterate trajectories produced by running a generic optimization algorithm OPT, we introduce $R^{\\text{OPT}}_{\\text{med}}$, a statistic that is analogous to the condition number of the loss Hessian evaluated at the iterates. Through extensive experiments, we show that adaptive methods such as Adam bias the trajectories towards regions where $R^{\\text{Adam}}_{\\text{med}}$ is small, where one might expect faster convergence. By contrast, vanilla gradient methods like SGD bias the trajectories towards regions where $R^{\\text{SGD}}_{\\text{med}}$ is comparatively large. We complement these empirical observations with a theoretical result that provably demonstrates this phenomenon in the simplified setting of a two-layer linear network. We view our findings as evidence for the need of a new explanation of the success of adaptive methods, one that is different than the conventional wisdom."
                },
                {
                    "title": "GLM-130B: An Open Bilingual Pre-trained Model",
                    "abstract": "We introduce GLM-130B, a bilingual (English and Chinese) pre-trained language model with 130 billion parameters. It is an attempt to open-source a 100B-scale model at least as good as GPT-3 (davinci) and unveil how models of such a scale can be successfully pre-trained. Over the course of this effort, we face numerous unexpected technical and engineering challenges, particularly on loss spikes and divergence. In this paper, we introduce the training process of GLM-130B including its design choices, training strategies for both efficiency and stability, and engineering efforts. The resultant GLM-130B model offers significant outperformance over GPT-3 175B (davinci) on a wide range of popular English benchmarks while the performance advantage is not observed in OPT-175B and BLOOM-176B. It also consistently and significantly outperforms ERNIE TITAN 3.0 260B -- the largest Chinese language model -- across related benchmarks. Finally, we leverage a unique scaling property of GLM-130B to reach INT4 quantization without post training, with almost no performance loss, making it the first among 100B-scale models and more importantly, allowing its effective inference on 4$\\times$RTX 3090 (24G) or 8$\\times$RTX 2080 Ti (11G) GPUs, the most affordable GPUs required for using 100B-scale models. The GLM-130B model weights are publicly accessible and its code, training logs, related toolkit, and lessons learned are open-sourced at \\url{https://github.com/THUDM/GLM-130B/}."
                },
                {
                    "title": "Robustness to Unbounded Smoothness of Generalized SignSGD",
                    "abstract": "Traditional analyses in non-convex optimization typically rely on the smoothness assumption, namely requiring the gradients to be Lipschitz. However, recent evidence shows that this smoothness condition does not capture the properties of some deep learning objective functions, including the ones involving Recurrent Neural Networks and LSTMs. Instead, they satisfy a much more relaxed condition, with potentially unbounded smoothness. Under this relaxed assumption, it has been theoretically and empirically shown that the gradient-clipped SGD has an advantage over the vanilla one. In this paper, we show that clipping is not indispensable for Adam-type algorithms in tackling such scenarios: we theoretically prove that a generalized SignSGD algorithm can obtain similar convergence rates as SGD with clipping but does not need explicit clipping at all. This family of algorithms on one end recovers SignSGD and on the other end closely resembles the popular Adam algorithm. Our analysis underlines the critical role that momentum plays in analyzing SignSGD-type and Adam-type algorithms: it not only reduces the effects of noise, thus removing the need for large mini-batch in previous analyses of SignSGD-type algorithms, but it also substantially reduces the effects of unbounded smoothness and gradient norms. We also compare these algorithms with popular optimizers on a set of deep learning tasks, observing that we can match the performance of Adam while beating the others."
                },
                {
                    "title": "Adam Can Converge Without Any Modification on Update Rules",
                    "abstract": "Ever since Reddi et al. 2018 pointed out the divergence issue of Adam, many new variants have been designed to obtain convergence. However, vanilla Adam remains exceptionally popular and it works well in practice. Why is there a gap between theory and practice? We point out there is a mismatch between the settings of theory and practice: Reddi et al. 2018 pick the problem after picking the hyperparameters of Adam, i.e., $(\\beta_1, \\beta_2)$; while practical applications often fix the problem first and then tune $(\\beta_1, \\beta_2)$. Due to this observation, we conjecture that the empirical convergence can be theoretically justified, only if we change the order of picking the problem and hyperparameter. In this work, we confirm this conjecture. We prove that, when $\\beta_2$ is large and $\\beta_1<\\sqrt{\\beta_2}<1$, Adam converges to the neighborhood of critical points. The size of the neighborhood is propositional to the variance of stochastic gradients. Under an extra condition (strong growth condition), Adam converges to critical points. It is worth mentioning that our results cover a wide range of hyperparameters: as $\\beta_2$ increases, our convergence result can cover any $\\beta_1 \\in [0,1)$ including $\\beta_1=0.9$, which is the default setting in deep learning libraries. To our knowledge, this is the first result showing that Adam can converge without any modification on its update rules. Further, our analysis does not require assumptions of bounded gradients or bounded 2nd-order momentum. When $\\beta_2$ is small, we further point out a large region of $(\\beta_1,\\beta_2)$ where Adam can diverge to infinity. Our divergence result considers the same setting as our convergence result, indicating a phase transition from divergence to convergence when increasing $\\beta_2$. These positive and negative results can provide suggestions on how to tune Adam hyperparameters."
                },
                {
                    "title": "Signal Propagation in Transformers: Theoretical Perspectives and the Role of Rank Collapse",
                    "abstract": "Transformers have achieved remarkable success in several domains, ranging from natural language processing to computer vision. Nevertheless, it has been recently shown that stacking self-attention layers - the distinctive architectural component of Transformers - can result in rank collapse of the tokens' representations at initialization. The question of if and how rank collapse affects training is still largely unanswered, and its investigation is necessary for a more comprehensive understanding of this architecture. In this work, we shed new light on the causes and the effects of this phenomenon. First, we show that rank collapse of the tokens' representations hinders training by causing the gradients of the queries and keys to vanish at initialization. Furthermore, we provide a thorough description of the origin of rank collapse and discuss how to prevent it via an appropriate depth-dependent scaling of the residual branches. Finally, our analysis unveils that specific architectural hyperparameters affect the gradients of queries and values differently, leading to disproportionate gradient norms. This suggests an explanation for the widespread use of adaptive methods for Transformers' optimization."
                },
                {
                    "title": "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness",
                    "abstract": "Transformers are slow and memory-hungry on long sequences, since the time and memory complexity of self-attention are quadratic in sequence length. Approximate attention methods have attempted to address this problem by trading off model quality to reduce the compute complexity, but often do not achieve wall-clock speedup. We argue that a missing principle is making attention algorithms IO-aware -- accounting for reads and writes between levels of GPU memory. We propose FlashAttention, an IO-aware exact attention algorithm that uses tiling to reduce the number of memory reads/writes between GPU high bandwidth memory (HBM) and GPU on-chip SRAM. We analyze the IO complexity of FlashAttention, showing that it requires fewer HBM accesses than standard attention, and is optimal for a range of SRAM sizes. We also extend FlashAttention to block-sparse attention, yielding an approximate attention algorithm that is faster than any existing approximate attention method. FlashAttention trains Transformers faster than existing baselines: 15% end-to-end wall-clock speedup on BERT-large (seq. length 512) compared to the MLPerf 1.1 training speed record, 3$\\times$ speedup on GPT-2 (seq. length 1K), and 2.4$\\times$ speedup on long-range arena (seq. length 1K-4K). FlashAttention and block-sparse FlashAttention enable longer context in Transformers, yielding higher quality models (0.7 better perplexity on GPT-2 and 6.4 points of lift on long-document classification) and entirely new capabilities: the first Transformers to achieve better-than-chance performance on the Path-X challenge (seq. length 16K, 61.4% accuracy) and Path-256 (seq. length 64K, 63.1% accuracy)."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer",
                    "abstract": "Hyperparameter (HP) tuning in deep learning is an expensive process, prohibitively so for neural networks (NNs) with billions of parameters. We show that, in the recently discovered Maximal Update Parametrization (muP), many optimal HPs remain stable even as model size changes. This leads to a new HP tuning paradigm we call muTransfer: parametrize the target model in muP, tune the HP indirectly on a smaller model, and zero-shot transfer them to the full-sized model, i.e., without directly tuning the latter at all. We verify muTransfer on Transformer and ResNet. For example, 1) by transferring pretraining HPs from a model of 13M parameters, we outperform published numbers of BERT-large (350M parameters), with a total tuning cost equivalent to pretraining BERT-large once; 2) by transferring from 40M parameters, we outperform published numbers of the 6.7B GPT-3 model, with tuning cost only 7% of total pretraining cost. A Pytorch implementation of our technique can be found at github.com/microsoft/mup and installable via `pip install mup`."
                },
                {
                    "title": "DeepNet: Scaling Transformers to 1,000 Layers",
                    "abstract": "In this paper, we propose a simple yet effective method to stabilize extremely deep Transformers. Specifically, we introduce a new normalization function (DeepNorm) to modify the residual connection in Transformer, accompanying with theoretically derived initialization. In-depth theoretical analysis shows that model updates can be bounded in a stable way. The proposed method combines the best of two worlds, i.e., good performance of Post-LN and stable training of Pre-LN, making DeepNorm a preferred alternative. We successfully scale Transformers up to 1,000 layers (i.e., 2,500 attention and feed-forward network sublayers) without difficulty, which is one order of magnitude deeper than previous deep Transformers. Extensive experiments demonstrate that DeepNet has superior performance across various benchmarks, including machine translation, language modeling (i.e., BERT, GPT) and vision pre-training (i.e., BEiT). Remarkably, on a multilingual benchmark with 7,482 translation directions, our 200-layer model with 3.2B parameters significantly outperforms the 48-layer state-of-the-art model with 12B parameters by 5 BLEU points, which indicates a promising scaling direction."
                },
                {
                    "title": "Early Convolutions Help Transformers See Better",
                    "abstract": "Vision transformer (ViT) models exhibit substandard optimizability. In particular, they are sensitive to the choice of optimizer (AdamW vs. SGD), optimizer hyperparameters, and training schedule length. In comparison, modern convolutional neural networks are easier to optimize. Why is this the case? In this work, we conjecture that the issue lies with the patchify stem of ViT models, which is implemented by a stride-p p*p convolution (p=16 by default) applied to the input image. This large-kernel plus large-stride convolution runs counter to typical design choices of convolutional layers in neural networks. To test whether this atypical design choice causes an issue, we analyze the optimization behavior of ViT models with their original patchify stem versus a simple counterpart where we replace the ViT stem by a small number of stacked stride-two 3*3 convolutions. While the vast majority of computation in the two ViT designs is identical, we find that this small change in early visual processing results in markedly different training behavior in terms of the sensitivity to optimization settings as well as the final model accuracy. Using a convolutional stem in ViT dramatically increases optimization stability and also improves peak performance (by ~1-2% top-1 accuracy on ImageNet-1k), while maintaining flops and runtime. The improvement can be observed across the wide spectrum of model complexities (from 1G to 36G flops) and dataset scales (from ImageNet-1k to ImageNet-21k). These findings lead us to recommend using a standard, lightweight convolutional stem for ViT models in this regime as a more robust architectural choice compared to the original ViT model design."
                },
                {
                    "title": "Super-Acceleration with Cyclical Step-sizes",
                    "abstract": "We develop a convergence-rate analysis of momentum with cyclical step-sizes. We show that under some assumption on the spectral gap of Hessians in machine learning, cyclical step-sizes are provably faster than constant step-sizes. More precisely, we develop a convergence rate analysis for quadratic objectives that provides optimal parameters and shows that cyclical learning rates can improve upon traditional lower complexity bounds. We further propose a systematic approach to design optimal \ufb01rst order methods for quadratic minimization with a given spectral structure. Finally, we provide a local convergence rate analysis beyond quadratic minimization for the proposed methods and illustrate our \ufb01ndings through benchmarks on least squares and logistic regression problems."
                },
                {
                    "title": "MLP-Mixer: An all-MLP Architecture for Vision",
                    "abstract": "Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e.\"mixing\"the per-location features), and one with MLPs applied across patches (i.e.\"mixing\"spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers."
                },
                {
                    "title": "Attention is Not All You Need: Pure Attention Loses Rank Doubly Exponentially with Depth",
                    "abstract": "Attention-based architectures have become ubiquitous in machine learning, yet our understanding of the reasons for their effectiveness remains limited. This work proposes a new way to understand self-attention networks: we show that their output can be decomposed into a sum of smaller terms, each involving the operation of a sequence of attention heads across layers. Using this decomposition, we prove that self-attention possesses a strong inductive bias towards\"token uniformity\". Specifically, without skip connections or multi-layer perceptrons (MLPs), the output converges doubly exponentially to a rank-1 matrix. On the other hand, skip connections and MLPs stop the output from degeneration. Our experiments verify the identified convergence phenomena on different variants of standard transformer architectures."
                },
                {
                    "title": "Hessian Eigenspectra of More Realistic Nonlinear Models",
                    "abstract": "Given an optimization problem, the Hessian matrix and its eigenspectrum can be used in many ways, ranging from designing more efficient second-order algorithms to performing model analysis and regression diagnostics. When nonlinear models and non-convex problems are considered, strong simplifying assumptions are often made to make Hessian spectral analysis more tractable. This leads to the question of how relevant the conclusions of such analyses are for more realistic nonlinear models. In this paper, we exploit deterministic equivalent techniques from random matrix theory to make a \\emph{precise} characterization of the Hessian eigenspectra for a broad family of nonlinear models, including models that generalize the classical generalized linear models, without relying on strong simplifying assumptions used previously. We show that, depending on the data properties, the nonlinear response model, and the loss function, the Hessian can have \\emph{qualitatively} different spectral behaviors: of bounded or unbounded support, with single- or multi-bulk, and with isolated eigenvalues on the left- or right-hand side of the bulk. By focusing on such a simple but nontrivial nonlinear model, our analysis takes a step forward to unveil the theoretical origin of many visually striking features observed in more complex machine learning models."
                },
                {
                    "title": "Towards Practical Adam: Non-Convexity, Convergence Theory, and Mini-Batch Acceleration",
                    "abstract": "Adam is one of the most influential adaptive stochastic algorithms for training deep neural networks, which has been pointed out to be divergent even in the simple convex setting via a few simple counterexamples. Many attempts, such as decreasing an adaptive learning rate, adopting a big batch size, incorporating a temporal decorrelation technique, seeking an analogous surrogate, \\textit{etc.}, have been tried to promote Adam-type algorithms to converge. In contrast with existing approaches, we introduce an alternative easy-to-check sufficient condition, which merely depends on the parameters of the base learning rate and combinations of historical second-order moments, to guarantee the global convergence of generic Adam for solving large-scale non-convex stochastic optimization. This observation, coupled with this sufficient condition, gives much deeper interpretations on the divergence of Adam. On the other hand, in practice, mini-Adam and distributed-Adam are widely used without any theoretical guarantee. We further give an analysis on how the batch size or the number of nodes in the distributed system affects the convergence of Adam, which theoretically shows that mini-batch and distributed Adam can be linearly accelerated by using a larger mini-batch size or a larger number of nodes.At last, we apply the generic Adam and mini-batch Adam with the sufficient condition for solving the counterexample and training several neural networks on various real-world datasets. Experimental results are exactly in accord with our theoretical analysis."
                },
                {
                    "title": "Asymptotic study of stochastic adaptive algorithm in non-convex landscape",
                    "abstract": "This paper studies some asymptotic properties of adaptive algorithms widely used in optimization and machine learning, and among them Adagrad and Rmsprop, which are involved in most of the blackbox deep learning algorithms. Our setup is the non-convex landscape optimization point of view, we consider a one time scale parametrization and we consider the situation where these algorithms may be used or not with mini-batches. We adopt the point of view of stochastic algorithms and establish the almost sure convergence of these methods when using a decreasing step-size point of view towards the set of critical points of the target function. With a mild extra assumption on the noise, we also obtain the convergence towards the set of minimizer of the function. Along our study, we also obtain a \"convergence rate\" of the methods, in the vein of the works of \\cite{GhadimiLan}."
                },
                {
                    "title": "A Deeper Look at the Hessian Eigenspectrum of Deep Neural Networks and its Applications to Regularization",
                    "abstract": "Loss landscape analysis is extremely useful for a deeper understanding of the generalization ability of deep neural network models. In this work, we propose a layerwise loss landscape analysis where the loss surface at every layer is studied independently and also on how each correlates to the overall loss surface. We study the layerwise loss landscape by studying the eigenspectra of the Hessian at each layer. In particular, our results show that the layerwise Hessian geometry is largely similar to the entire Hessian. We also report an interesting phenomenon where the Hessian eigenspectrum of middle layers of the deep neural network are observed to most similar to the overall Hessian eigenspectrum. We also show that the maximum eigenvalue and the trace of the Hessian (both full network and layerwise) reduce as training of the network progresses. We leverage on these observations to propose a new regularizer based on the trace of the layerwise Hessian. Penalizing the trace of the Hessian at every layer indirectly forces Stochastic Gradient Descent to converge to flatter minima, which are shown to have better generalization performance. In particular, we show that such a layerwise regularizer can be leveraged to penalize the middlemost layers alone, which yields promising results. Our empirical studies on well-known deep nets across datasets support the claims of this work."
                },
                {
                    "title": "Why are Adaptive Methods Good for Attention Models?",
                    "abstract": "While stochastic gradient descent (SGD) is still the \\emph{de facto} algorithm in deep learning, adaptive methods like Clipped SGD/Adam have been observed to outperform SGD across important tasks, such as attention models. The settings under which SGD performs poorly in comparison to adaptive methods are not well understood yet. In this paper, we provide empirical and theoretical evidence that a heavy-tailed distribution of the noise in stochastic gradients is one cause of SGD's poor performance. We provide the first tight upper and lower convergence bounds for adaptive gradient methods under heavy-tailed noise. Further, we demonstrate how gradient clipping plays a key role in addressing heavy-tailed gradient noise. Subsequently, we show how clipping can be applied in practice by developing an \\emph{adaptive} coordinate-wise clipping algorithm (ACClip) and demonstrate its superior performance on BERT pretraining and finetuning tasks."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Effects of Parameter Norm Growth During Transformer Training: Inductive Bias from Gradient Descent",
                    "abstract": "The capacity of neural networks like the widely adopted transformer is known to be very high. Evidence is emerging that they learn successfully due to inductive bias in the training routine, typically a variant of gradient descent (GD). To better understand this bias, we study the tendency for transformer parameters to grow in magnitude (\\ell_2 norm) during training, and its implications for the emergent representations within self attention layers. Empirically, we document norm growth in the training of transformer language models, including T5 during its pretraining. As the parameters grow in magnitude, we prove that the network approximates a discretized network with saturated activation functions. Such \u201csaturated\u201d networks are known to have a reduced capacity compared to the full network family that can be described in terms of formal languages and automata. Our results suggest saturation is a new characterization of an inductive bias implicit in GD of particular interest for NLP. We leverage the emergent discrete structure in a saturated transformer to analyze the role of different attention heads, finding that some focus locally on a small number of positions, while other heads compute global averages, allowing counting. We believe understanding the interplay between these two capabilities may shed further light on the structure of computation within large transformers."
                },
                {
                    "title": "Dissecting Hessian: Understanding Common Structure of Hessian in Neural Networks",
                    "abstract": "Hessian captures important properties of the deep neural network loss landscape. We observe that eigenvectors and eigenspaces of the layer-wise Hessian for neural network objective have several interesting structures -- top eigenspaces for different models have high overlap, and top eigenvectors form low rank matrices when they are reshaped into the same shape as the corresponding weight matrix. These structures, as well as the low rank structure of the Hessian observed in previous studies, can be explained by approximating the Hessian using Kronecker factorization. Our new understanding can also explain why some of these structures become weaker when the network is trained with batch normalization. Finally, we show that the Kronecker factorization can be combined with PAC-Bayes techniques to get better explicit generalization bounds."
                },
                {
                    "title": "Traces of Class/Cross-Class Structure Pervade Deep Learning Spectra",
                    "abstract": "Numerous researchers recently applied empirical spectral analysis to the study of modern deep learning classifiers. We identify and discuss an important formal class/cross-class structure and show how it lies at the origin of the many visually striking features observed in deepnet spectra, some of which were reported in recent articles, others are unveiled here for the first time. These include spectral outliers, \"spikes\", and small but distinct continuous distributions, \"bumps\", often seen beyond the edge of a \"main bulk\". \nThe significance of the cross-class structure is illustrated in three ways: (i) we prove the ratio of outliers to bulk in the spectrum of the Fisher information matrix is predictive of misclassification, in the context of multinomial logistic regression; (ii) we demonstrate how, gradually with depth, a network is able to separate class-distinctive information from class variability, all while orthogonalizing the class-distinctive information; and (iii) we propose a correction to KFAC, a well-known second-order optimization algorithm for training deepnets."
                },
                {
                    "title": "Improving Transformer Optimization Through Better Initialization",
                    "abstract": "The Transformer architecture has achieved considerable success recently; the key component of the Transformer is the attention layer that enables the model to focus on important regions within an input sequence. Gradient optimization with attention layers can be notoriously dif\ufb01cult requiring tricks such as learning rate warmup to prevent divergence. As Transformer models are becoming larger and more expensive to train, recent research has focused on understanding and improving optimization in these architectures. In this work our contributions are two-fold: we \ufb01rst investigate and empirically validate the source of optimization problems in the encoder-decoder Transformer architecture; we then propose a new weight initialization scheme with theoretical justi-\ufb01cation, that enables training without warmup or layer normalization. Empirical results on public machine translation benchmarks show that our approach achieves leading accuracy, allowing to train deep Transformer models with 200 layers in both encoder and decoder (over 1000 attention/MLP blocks) without dif\ufb01culty. Code for this work is available here: https://github. com/layer6ai-labs/T-Fixup ."
                },
                {
                    "title": "DeepSpeed: System Optimizations Enable Training Deep Learning Models with Over 100 Billion Parameters",
                    "abstract": "Explore new techniques in Microsoft's open source library called DeepSpeed, which advances large model training by improving scale, speed, cost, and usability, unlocking the ability to train 100-billion-parameter models. DeepSpeed is compatible with PyTorch. One piece of our library, called ZeRO, is a new parallelized optimizer that greatly reduces the resources needed for model and data parallelism while massively increasing the number of parameters that can be trained. Researchers have used these breakthroughs to create Turing Natural Language Generation (Turing-NLG), which at the time of its release was the largest publicly known language model at 17 billion parameters. In addition we will also go over our latest transformer kernel advancements that led the DeepSpeed team to achieve the world fastest BERT pretraining record. The Zero Redundancy Optimizer (ZeRO) is a novel memory optimization technology for large-scale distributed deep learning. ZeRO can train deep learning models with over 100 billion parameters on the current generation of GPU clusters at three to five times the throughput of the current best system. It also presents a clear path to training models with trillions of parameters, demonstrating an unprecedented leap in deep learning system technology. DeepSpeed brings state-of-the-art training techniques, such as ZeRO, optimized kernels, distributed training, mixed precision, and checkpointing, through lightweight APIs compatible with PyTorch. With just a few lines of code changes to your PyTorch model, you can leverage DeepSpeed to address underlying performance challenges and boost the speed and scale of your training."
                },
                {
                    "title": "Adaptive Inertia: Disentangling the Effects of Adaptive Learning Rate and Momentum",
                    "abstract": "Adaptive Moment Estimation (Adam), which combines Adaptive Learning Rate and Momentum, would be the most popular stochastic optimizer for accelerating the training of deep neural networks. However, it is empirically known that Adam often generalizes worse than Stochastic Gradient Descent (SGD). The purpose of this paper is to unveil the mystery of this behavior in the diffusion theoretical framework. Specifically, we disentangle the effects of Adaptive Learning Rate and Momentum of the Adam dynamics on saddle-point escaping and flat minima selection. We prove that Adaptive Learning Rate can escape saddle points efficiently, but cannot select flat minima as SGD does. In contrast, Momentum provides a drift effect to help the training process pass through saddle points, and almost does not affect flat minima selection. This partly explains why SGD (with Momentum) generalizes better, while Adam generalizes worse but converges faster. Furthermore, motivated by the analysis, we design a novel adaptive optimization framework named Adaptive Inertia, which uses parameter-wise adaptive inertia to accelerate the training and provably favors flat minima as well as SGD. Our extensive experiments demonstrate that the proposed adaptive inertia method can generalize significantly better than SGD and conventional adaptive gradient methods."
                },
                {
                    "title": "Understanding the Difficulty of Training Transformers",
                    "abstract": "Transformers have been proved effective for many deep learning tasks. Training transformers, however, requires non-trivial efforts regarding carefully designing learning rate schedulers and cutting-edge optimizers (the standard SGD fails to train Transformers effectively). In this paper, we study Transformer training from both theoretical and empirical perspectives. Our analysis reveals that unbalanced gradients are not the root cause of the instability of training. Instead, we identify an amplification effect that substantially influences training. Specifically, we observe that for each layer in a multi-layer Transformer model, heavy dependency on its residual branch makes training unstable since it amplifies small parameter perturbations (e.g., parameter updates) and result in significant disturbances in the model output, yet a light dependency limits the potential of model training and can lead to an inferior trained model. Inspired by our analysis, we propose Admin ($\\mathbf{Ad}$aptive $\\mathbf{m}$odel $\\mathbf{in}$itialization) to stabilize the training in the early stage and unleash its full potential in the late stage. Extensive experiments show that Admin is more stable, converges faster, and leads to better performance."
                },
                {
                    "title": "ReZero is All You Need: Fast Convergence at Large Depth",
                    "abstract": "Deep networks often suffer from vanishing or exploding gradients due to inefficient signal propagation, leading to long training times or convergence difficulties. Various architecture designs, sophisticated residual-style networks, and initialization schemes have been shown to improve deep signal propagation. Recently, Pennington et al. used free probability theory to show that dynamical isometry plays an integral role in efficient deep learning. We show that the simplest architecture change of gating each residual connection using a single zero-initialized parameter satisfies initial dynamical isometry and outperforms more complex approaches. Although much simpler than its predecessors, this gate enables training thousands of fully connected layers with fast convergence and better test performance for ResNets trained on CIFAR-10. We apply this technique to language modeling and find that we can easily train 120-layer Transformers. When applied to 12 layer Transformers, it converges 56% faster on enwiki8."
                },
                {
                    "title": "A Simple Convergence Proof of Adam and Adagrad",
                    "abstract": "We provide a simple proof of convergence covering both the Adam and Adagrad adaptive optimization algorithms when applied to smooth (possibly non-convex) objective functions with bounded gradients. We show that in expectation, the squared norm of the objective gradient averaged over the trajectory has an upper-bound which is explicit in the constants of the problem, parameters of the optimizer and the total number of iterations $N$. This bound can be made arbitrarily small: Adam with a learning rate $\\alpha=1/\\sqrt{N}$ and a momentum parameter on squared gradients $\\beta_2=1-1/N$ achieves the same rate of convergence $O(\\ln(N)/\\sqrt{N})$ as Adagrad. Finally, we obtain the tightest dependency on the heavy ball momentum among all previous convergence bounds for non-convex Adam and Adagrad, improving from $O((1-\\beta_1)^{-3})$ to $O((1-\\beta_1)^{-1})$. Our technique also improves the best known dependency for standard SGD by a factor $1 - \\beta_1$."
                },
                {
                    "title": "On Layer Normalization in the Transformer Architecture",
                    "abstract": "The Transformer is widely used in natural language processing tasks. To train a Transformer however, one usually needs a carefully designed learning rate warm-up stage, which is shown to be crucial to the final performance but will slow down the optimization and bring more hyperparameter tunings. In this paper, we first study theoretically why the learning rate warm-up stage is essential and show that the location of layer normalization matters. Specifically, we prove with mean field theory that at initialization, for the original-designed Post-LN Transformer, which places the layer normalization between the residual blocks, the expected gradients of the parameters near the output layer are large. Therefore, using a large learning rate on those gradients makes the training unstable. The warm-up stage is practically helpful for avoiding this problem. On the other hand, our theory also shows that if the layer normalization is put inside the residual blocks (recently proposed as Pre-LN Transformer), the gradients are well-behaved at initialization. This motivates us to remove the warm-up stage for the training of Pre-LN Transformers. We show in our experiments that Pre-LN Transformers without the warm-up stage can reach comparable results with baselines while requiring significantly less training time and hyper-parameter tuning on a wide range of applications."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Large-Scale Machine Learning",
                    "abstract": "Course Outline: This course will explore the mathematical foundations of a rapidly evolving new field: large-scale machine learning and optimization. We will focus on recent texts in machine learning, optimization, and randomized algorithms, with the goal to understand the tradeoffs that are driving algorithmic design in this new discipline. These trade-offs will revolve around statistical accuracy, scalability, algorithmic complexity, and implementation."
                },
                {
                    "title": "Optimization for deep learning: theory and algorithms",
                    "abstract": "When and why can a neural network be successfully trained? This article provides an overview of optimization algorithms and theory for training neural networks. First, we discuss the issue of gradient explosion/vanishing and the more general issue of undesirable spectrum, and then discuss practical solutions including careful initialization and normalization methods. Second, we review generic optimization methods used in training neural networks, such as SGD, adaptive gradient methods and distributed methods, and theoretical results for these algorithms. Third, we review existing research on the global issues of neural network training, including results on bad local minima, mode connectivity, lottery ticket hypothesis and infinite-width analysis."
                },
                {
                    "title": "PyHessian: Neural Networks Through the Lens of the Hessian",
                    "abstract": "We present PYHESSIAN, a new scalable framework that enables fast computation of Hessian (i.e., second-order derivative) information for deep neural networks. PYHESSIAN enables fast computations of the top Hessian eigenvalues, the Hessian trace, and the full Hessian eigenvalue/spectral density; it supports distributed-memory execution on cloud/supercomputer systems; and it is available as open source [1]. This general framework can be used to analyze neural network models, including the topology of the loss landscape (i.e., curvature information) to gain insight into the behavior of different models/optimizers. As an example, we analyze the effect of residual connections and Batch Normalization layers on the trainability of neural networks. One recent claim, based on simpler first-order analysis, is that residual connections and Batch Normalization make the loss landscape \"smoother,\" thus making it easier for Stochastic Gradient Descent to converge to a good solution. Our second-order analysis, easily enabled by PYHESSIAN, shows new finer-scale insights, demonstrating that while conventional wisdom is sometimes validated, in other cases it is simply incorrect. In particular, we find that Batch Normalization does not necessarily make the loss landscape smoother, especially for shallow networks."
                },
                {
                    "title": "Transformers without Tears: Improving the Normalization of Self-Attention",
                    "abstract": "We evaluate three simple, normalization-centric changes to improve Transformer training. First, we show that pre-norm residual connections (PRENORM) and smaller initializations enable warmup-free, validation-based training with large learning rates. Second, we propose l2 normalization with a single scale parameter (SCALENORM) for faster training and better performance. Finally, we reaffirm the effectiveness of normalizing word embeddings to a fixed length (FIXNORM). On five low-resource translation pairs from TED Talks-based corpora, these changes always converge, giving an average +1.1 BLEU over state-of-the-art bilingual baselines and a new 32.8 BLEU on IWSLT '15 English-Vietnamese. We ob- serve sharper performance curves, more consistent gradient norms, and a linear relationship between activation scaling and decoder depth. Surprisingly, in the high-resource setting (WMT '14 English-German), SCALENORM and FIXNORM remain competitive but PRENORM degrades performance."
                },
                {
                    "title": "Improving Deep Transformer with Depth-Scaled Initialization and Merged Attention",
                    "abstract": "The general trend in NLP is towards increasing model capacity and performance via deeper neural networks. However, simply stacking more layers of the popular Transformer architecture for machine translation results in poor convergence and high computational overhead. Our empirical analysis suggests that convergence is poor due to gradient vanishing caused by the interaction between residual connection and layer normalization. We propose depth-scaled initialization (DS-Init), which decreases parameter variance at the initialization stage, and reduces output variance of residual connections so as to ease gradient back-propagation through normalization layers. To address computational cost, we propose a merged attention sublayer (MAtt) which combines a simplified average-based self-attention sublayer and the encoder-decoder attention sublayer on the decoder side. Results on WMT and IWSLT translation tasks with five translation directions show that deep Transformers with DS-Init and MAtt can substantially outperform their base counterpart in terms of BLEU (+1.1 BLEU on average for 12-layer models), while matching the decoding speed of the baseline model thanks to the efficiency improvements of MAtt. Source code for reproduction will be released soon."
                },
                {
                    "title": "On the Variance of the Adaptive Learning Rate and Beyond",
                    "abstract": "The learning rate warmup heuristic achieves remarkable success in stabilizing training, accelerating convergence and improving generalization for adaptive stochastic optimization algorithms like RMSprop and Adam. Here, we study its mechanism in details. Pursuing the theory behind warmup, we identify a problem of the adaptive learning rate (i.e., it has problematically large variance in the early stage), suggest warmup works as a variance reduction technique, and provide both empirical and theoretical evidence to verify our hypothesis. We further propose RAdam, a new variant of Adam, by introducing a term to rectify the variance of the adaptive learning rate. Extensive experimental results on image classification, language modeling, and neural machine translation verify our intuition and demonstrate the effectiveness and robustness of our proposed method. All implementations are available at: this https URL."
                },
                {
                    "title": "Which Algorithmic Choices Matter at Which Batch Sizes? Insights From a Noisy Quadratic Model",
                    "abstract": "Increasing the batch size is a popular way to speed up neural network training, but beyond some critical batch size, larger batch sizes yield diminishing returns. In this work, we study how the critical batch size changes based on properties of the optimization algorithm, including acceleration and preconditioning, through two different lenses: large scale experiments, and analysis of a simple noisy quadratic model (NQM). We experimentally demonstrate that optimization algorithms that employ preconditioning, specifically Adam and K-FAC, result in much larger critical batch sizes than stochastic gradient descent with momentum. We also demonstrate that the NQM captures many of the essential features of real neural network training, despite being drastically simpler to work with. The NQM predicts our results with preconditioned optimizers, previous results with accelerated gradient descent, and other results around optimal learning rates and large batch training, making it a useful tool to generate testable predictions about neural network optimization."
                },
                {
                    "title": "Learning Deep Transformer Models for Machine Translation",
                    "abstract": "Transformer is the state-of-the-art model in recent machine translation evaluations. Two strands of research are promising to improve models of this kind: the first uses wide networks (a.k.a. Transformer-Big) and has been the de facto standard for development of the Transformer system, and the other uses deeper language representation but faces the difficulty arising from learning deep networks. Here, we continue the line of research on the latter. We claim that a truly deep Transformer model can surpass the Transformer-Big counterpart by 1) proper use of layer normalization and 2) a novel way of passing the combination of previous layers to the next. On WMT\u201916 English-German and NIST OpenMT\u201912 Chinese-English tasks, our deep system (30/25-layer encoder) outperforms the shallow Transformer-Big/Base baseline (6-layer encoder) by 0.4-2.4 BLEU points. As another bonus, the deep model is 1.6X smaller in size and 3X faster in training than Transformer-Big."
                },
                {
                    "title": "Why Gradient Clipping Accelerates Training: A Theoretical Justification for Adaptivity",
                    "abstract": "We provide a theoretical explanation for the effectiveness of gradient clipping in training deep neural networks. The key ingredient is a new smoothness condition derived from practical neural network training examples. We observe that gradient smoothness, a concept central to the analysis of first-order optimization algorithms that is often assumed to be a constant, demonstrates significant variability along the training trajectory of deep neural networks. Further, this smoothness positively correlates with the gradient norm, and contrary to standard assumptions in the literature, it can grow with the norm of the gradient. These empirical observations limit the applicability of existing theoretical analyses of algorithms that rely on a fixed bound on smoothness. These observations motivate us to introduce a novel relaxation of gradient smoothness that is weaker than the commonly used Lipschitz smoothness assumption. Under the new condition, we prove that two popular methods, namely, \\emph{gradient clipping} and \\emph{normalized gradient}, converge arbitrarily faster than gradient descent with fixed stepsize. We further explain why such adaptively scaled gradient methods can accelerate empirical convergence and verify our results empirically in popular neural network training settings."
                },
                {
                    "title": "Adaptive Gradient Methods with Dynamic Bound of Learning Rate",
                    "abstract": "Adaptive optimization methods such as AdaGrad, RMSprop and Adam have been proposed to achieve a rapid training process with an element-wise scaling term on learning rates. Though prevailing, they are observed to generalize poorly compared with SGD or even fail to converge due to unstable and extreme learning rates. Recent work has put forward some algorithms such as AMSGrad to tackle this issue but they failed to achieve considerable improvement over existing methods. In our paper, we demonstrate that extreme learning rates can lead to poor performance. We provide new variants of Adam and AMSGrad, called AdaBound and AMSBound respectively, which employ dynamic bounds on learning rates to achieve a gradual and smooth transition from adaptive methods to SGD and give a theoretical proof of convergence. We further conduct experiments on various popular tasks and models, which is often insufficient in previous work. Experimental results show that new variants can eliminate the generalization gap between adaptive methods and SGD and maintain higher learning speed early in training at the same time. Moreover, they can bring significant improvement over their prototypes, especially on complex deep networks. The implementation of the algorithm can be found at this https URL ."
                },
                {
                    "title": "An Investigation into Neural Net Optimization via Hessian Eigenvalue Density",
                    "abstract": "To understand the dynamics of optimization in deep neural networks, we develop a tool to study the evolution of the entire Hessian spectrum throughout the optimization process. Using this, we study a number of hypotheses concerning smoothness, curvature, and sharpness in the deep learning literature. We then thoroughly analyze a crucial structural feature of the spectra: in non-batch normalized networks, we observe the rapid appearance of large isolated eigenvalues in the spectrum, along with a surprising concentration of the gradient in the corresponding eigenspaces. In batch normalized networks, these two effects are almost absent. We characterize these effects, and explain how they affect optimization speed through both theory and experiments. As part of this work, we adapt advanced tools from numerical linear algebra that allow scalable and accurate estimation of the entire Hessian spectrum of ImageNet-scale neural networks; this technique may be of independent interest in other applications."
                },
                {
                    "title": "Measurements of Three-Level Hierarchical Structure in the Outliers in the Spectrum of Deepnet Hessians",
                    "abstract": "We consider deep classifying neural networks. We expose a structure in the derivative of the logits with respect to the parameters of the model, which is used to explain the existence of outliers in the spectrum of the Hessian. Previous works decomposed the Hessian into two components, attributing the outliers to one of them, the so-called Covariance of gradients. We show this term is not a Covariance but a second moment matrix, i.e., it is influenced by means of gradients. These means possess an additive two-way structure that is the source of the outliers in the spectrum. This structure can be used to approximate the principal subspace of the Hessian using certain \"averaging\" operations, avoiding the need for high-dimensional eigenanalysis. We corroborate this claim across different datasets, architectures and sample sizes."
                },
                {
                    "title": "A Sufficient Condition for Convergences of Adam and RMSProp",
                    "abstract": "Adam and RMSProp are two of the most influential adaptive stochastic algorithms for training deep neural networks, which have been pointed out to be divergent even in the convex setting via a few simple counterexamples. Many attempts, such as decreasing an adaptive learning rate, adopting a big batch size, incorporating a temporal decorrelation technique, seeking an analogous surrogate, etc., have been tried to promote Adam/RMSProp-type algorithms to converge. In contrast with existing approaches, we introduce an alternative easy-to-check sufficient condition, which merely depends on the parameters of the base learning rate and combinations of historical second-order moments, to guarantee the global convergence of generic Adam/RMSProp for solving large-scale non-convex stochastic optimization. Moreover, we show that the convergences of several variants of Adam, such as AdamNC, AdaEMA, etc., can be directly implied via the proposed sufficient condition in the non-convex setting. In addition, we illustrate that Adam is essentially a specifically weighted AdaGrad with exponential moving average momentum, which provides a novel perspective for understanding Adam and RMSProp. This observation coupled with this sufficient condition gives much deeper interpretations on their divergences. At last, we validate the sufficient condition by applying Adam and RMSProp to tackle a certain counterexample and train deep neural networks. Numerical results are exactly in accord with our theoretical analysis."
                },
                {
                    "title": "The Full Spectrum of Deepnet Hessians at Scale: Dynamics with SGD Training and Sample Size.",
                    "abstract": "We apply state-of-the-art tools in modern high-dimensional numerical linear algebra to approximate efficiently the spectrum of the Hessian of modern deepnets, with tens of millions of parameters, trained on real data. Our results corroborate previous findings, based on small-scale networks, that the Hessian exhibits \"spiked\" behavior, with several outliers isolated from a continuous bulk. We decompose the Hessian into different components and study the dynamics with training and sample size of each term individually."
                },
                {
                    "title": "A general system of differential equations to model first order adaptive algorithms",
                    "abstract": "First order optimization algorithms play a major role in large scale machine learning. A new class of methods, called adaptive algorithms, were recently introduced to adjust iteratively the learning rate for each coordinate. Despite great practical success in deep learning, their behavior and performance on more general loss functions are not well understood. In this paper, we derive a non-autonomous system of differential equations, which is the continuous time limit of adaptive optimization methods. We prove global well-posedness of the system and we investigate the numerical time convergence of its forward Euler approximation. We study, furthermore, the convergence of its trajectories and give conditions under which the differential system, underlying all adaptive algorithms, is suitable for optimization. We discuss convergence to a critical point in the non-convex case and give conditions for the dynamics to avoid saddle points and local maxima. For convex and deterministic loss function, we introduce a suitable Lyapunov functional which allow us to study its rate of convergence. Several other properties of both the continuous and discrete systems are briefly discussed. The differential system studied in the paper is general enough to encompass many other classical algorithms (such as Heavy ball and Nesterov's accelerated method) and allow us to recover several known results for these algorithms."
                },
                {
                    "title": "Gradient Descent Happens in a Tiny Subspace",
                    "abstract": "We show that in a variety of large-scale deep learning scenarios the gradient dynamically converges to a very small subspace after a short period of training. The subspace is spanned by a few top eigenvectors of the Hessian (equal to the number of classes in the dataset), and is mostly preserved over long periods of training. A simple argument then suggests that gradient descent may happen mostly in this subspace. We give an example of this effect in a solvable model of classification, and we comment on possible implications for optimization and learning."
                },
                {
                    "title": "On the Convergence of Adaptive Gradient Methods for Nonconvex Optimization",
                    "abstract": "Adaptive gradient methods are workhorses in deep learning. However, the convergence guarantees of adaptive gradient methods for nonconvex optimization have not been sufficiently studied. In this paper, we provide a sharp analysis of a recently proposed adaptive gradient method namely partially adaptive momentum estimation method (Padam) (Chen and Gu, 2018), which admits many existing adaptive gradient methods such as RMSProp and AMSGrad as special cases. Our analysis shows that, for smooth nonconvex functions, Padam converges to a first-order stationary point at the rate of $O\\big((\\sum_{i=1}^d\\|\\mathbf{g}_{1:T,i}\\|_2)^{1/2}/T^{3/4} + d/T\\big)$, where $T$ is the number of iterations, $d$ is the dimension, $\\mathbf{g}_1,\\ldots,\\mathbf{g}_T$ are the stochastic gradients, and $\\mathbf{g}_{1:T,i} = [g_{1,i},g_{2,i},\\ldots,g_{T,i}]^\\top$. Our theoretical result also suggests that in order to achieve faster convergence rate, it is necessary to use Padam instead of AMSGrad. This is well-aligned with the empirical results of deep learning reported in Chen and Gu (2018)."
                },
                {
                    "title": "On the Convergence of A Class of Adam-Type Algorithms for Non-Convex Optimization",
                    "abstract": "This paper studies a class of adaptive gradient based momentum algorithms that update the search directions and learning rates simultaneously using past gradients. This class, which we refer to as the \"Adam-type\", includes the popular algorithms such as the Adam, AMSGrad and AdaGrad. Despite their popularity in training deep neural networks, the convergence of these algorithms for solving nonconvex problems remains an open question. This paper provides a set of mild sufficient conditions that guarantee the convergence for the Adam-type methods. We prove that under our derived conditions, these methods can achieve the convergence rate of order $O(\\log{T}/\\sqrt{T})$ for nonconvex stochastic optimization. We show the conditions are essential in the sense that violating them may make the algorithm diverge. Moreover, we propose and analyze a class of (deterministic) incremental adaptive gradient algorithms, which has the same $O(\\log{T}/\\sqrt{T})$ convergence rate. Our study could also be extended to a broader class of adaptive gradient methods in machine learning and optimization."
                },
                {
                    "title": "The Best of Both Worlds: Combining Recent Advances in Neural Machine Translation",
                    "abstract": "The past year has witnessed rapid advances in sequence-to-sequence (seq2seq) modeling for Machine Translation (MT). The classic RNN-based approaches to MT were first out-performed by the convolutional seq2seq model, which was then out-performed by the more recent Transformer model. Each of these new approaches consists of a fundamental architecture accompanied by a set of modeling and training techniques that are in principle applicable to other seq2seq architectures. In this paper, we tease apart the new architectures and their accompanying techniques in two ways. First, we identify several key modeling and training techniques, and apply them to the RNN architecture, yielding a new RNMT+ model that outperforms all of the three fundamental architectures on the benchmark WMT\u201914 English to French and English to German tasks. Second, we analyze the properties of each fundamental seq2seq architecture and devise new hybrid architectures intended to combine their strengths. Our hybrid models obtain further improvements, outperforming the RNMT+ model on both benchmark datasets."
                },
                {
                    "title": "Hessian-based Analysis of Large Batch Training and Robustness to Adversaries",
                    "abstract": "Large batch size training of Neural Networks has been shown to incur accuracy loss when trained with the current methods. The exact underlying reasons for this are still not completely understood. Here, we study large batch size training through the lens of the Hessian operator and robust optimization. In particular, we perform a Hessian based study to analyze exactly how the landscape of the loss function changes when training with large batch size. We compute the true Hessian spectrum, without approximation, by back-propagating the second derivative. Extensive experiments on multiple networks show that saddle-points are not the cause for generalization gap of large batch size training, and the results consistently show that large batch converges to points with noticeably higher Hessian spectrum. Furthermore, we show that robust training allows one to favor flat areas, as points with large Hessian spectrum show poor robustness to adversarial perturbation. We further study this relationship, and provide empirical and theoretical proof that the inner loop for robust training is a saddle-free optimization problem \\textit{almost everywhere}. We present detailed experiments with five different network architectures, including a residual network, tested on MNIST, CIFAR-10, and CIFAR-100 datasets. We have open sourced our method which can be accessed at [1]."
                },
                {
                    "title": "On the Convergence of Adam and Beyond",
                    "abstract": "Several recently proposed stochastic optimization methods that have been successfully used in training deep networks such as RMSProp, Adam, Adadelta, Nadam are based on using gradient updates scaled by square roots of exponential moving averages of squared past gradients. In many applications, e.g. learning with large output spaces, it has been empirically observed that these algorithms fail to converge to an optimal solution (or a critical point in nonconvex settings). We show that one cause for such failures is the exponential moving average used in the algorithms. We provide an explicit example of a simple convex optimization setting where Adam does not converge to the optimal solution, and describe the precise problems with the previous analysis of Adam algorithm. Our analysis suggests that the convergence issues can be fixed by endowing such algorithms with `long-term memory' of past gradients, and propose new variants of the Adam algorithm which not only fix the convergence issues but often also lead to improved empirical performance."
                },
                {
                    "title": "signSGD: compressed optimisation for non-convex problems",
                    "abstract": "Training large neural networks requires distributing learning across multiple workers, where the cost of communicating gradients can be a significant bottleneck. signSGD alleviates this problem by transmitting just the sign of each minibatch stochastic gradient. We prove that it can get the best of both worlds: compressed gradients and SGD-level convergence rate. signSGD can exploit mismatches between L1 and L2 geometry: when noise and curvature are much sparser than the gradients, signSGD is expected to converge at the same rate or faster than full-precision SGD. Measurements of the L1 versus L2 geometry of real networks support our theoretical claims, and we find that the momentum counterpart of signSGD is able to match the accuracy and convergence speed of Adam on deep Imagenet models. We extend our theory to the distributed setting, where the parameter server uses majority vote to aggregate gradient signs from each worker enabling 1-bit compression of worker-server communication in both directions. Using a theorem by Gauss, we prove that the non-convex convergence rate of majority vote matches that of distributed SGD. Thus, there is great promise for sign-based optimisation schemes to achieve both communication efficiency and high accuracy."
                },
                {
                    "title": "Estimating the Spectral Density of Large Implicit Matrices",
                    "abstract": "Many important problems are characterized by the eigenvalues of a large matrix. For example, the difficulty of many optimization problems, such as those arising from the fitting of large models in statistics and machine learning, can be investigated via the spectrum of the Hessian of the empirical loss function. Network data can be understood via the eigenstructure of a graph Laplacian matrix using spectral graph theory. Quantum simulations and other many-body problems are often characterized via the eigenvalues of the solution space, as are various dynamic systems. However, naive eigenvalue estimation is computationally expensive even when the matrix can be represented; in many of these situations the matrix is so large as to only be available implicitly via products with vectors. Even worse, one may only have noisy estimates of such matrix vector products. In this work, we combine several different techniques for randomized estimation and show that it is possible to construct unbiased estimators to answer a broad class of questions about the spectra of such implicit matrices, even in the presence of noise. We validate these methods on large-scale problems in which graph theory and random matrix theory provide ground truth."
                },
                {
                    "title": "Iteration Method for the Solution of the Eigenvalue Problem of Linear Differential and Integral Operators",
                    "abstract": "The present investigation designs a systematic method for finding the latent roots and the principal axes of a matrix, without reducing the order of the matrix. It is characterized by a wide field of applicability and great accuracy, since the accumulation of rounding errors is avoided, through the process of \"minimized iterations\". Moreover, the method leads to a well convergent successive approximation procedure by which the solution of integral equations of the Fredholm type and the solution of the eigenvalue problem of linear differential and integral operators may be accomplished."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Fast Estimation of tr(f(A)) via Stochastic Lanczos Quadrature",
                    "abstract": "The problem of estimating the trace of matrix functions appears in applications ranging from machine learning and scientific computing, to computational biology. This paper presents an inexpensive method to estimate the trace of $f(A)$ for cases where $f$ is analytic inside a closed interval and $A$ is a symmetric positive definite matrix. The method combines three key ingredients, namely, the stochastic trace estimator, Gaussian quadrature, and the Lanczos algorithm. As examples, we consider the problems of estimating the log-determinant ($f(t)=\\log(t)$), the Schatten $p$-norms ($f(t)=t^{p/2}$), the Estrada index ($f(t)=e^{t}$), and the trace of the matrix inverse ($f(t)=t^{-1}$). We establish multiplicative and additive error bounds for the approximations obtained by this method. In addition, we present error bounds for other useful tools such as approximating the log-likelihood function in the context of maximum likelihood estimation of Gaussian processes. Numerical experiments illustrate the performan..."
                },
                {
                    "title": "Empirical Analysis of the Hessian of Over-Parametrized Neural Networks",
                    "abstract": "We study the properties of common loss surfaces through their Hessian matrix. In particular, in the context of deep learning, we empirically show that the spectrum of the Hessian is composed of two parts: (1) the bulk centered near zero, (2) and outliers away from the bulk. We present numerical evidence and mathematical justifications to the following conjectures laid out by Sagun et al. (2016): Fixing data, increasing the number of parameters merely scales the bulk of the spectrum; fixing the dimension and changing the data (for instance adding more clusters or making the data less separable) only affects the outliers. We believe that our observations have striking implications for non-convex optimization in high dimensions. First, the flatness of such landscapes (which can be measured by the singularity of the Hessian) implies that classical notions of basins of attraction may be quite misleading. And that the discussion of wide/narrow basins may be in need of a new perspective around over-parametrization and redundancy that are able to create large connected components at the bottom of the landscape. Second, the dependence of small number of large eigenvalues to the data distribution can be linked to the spectrum of the covariance matrix of gradients of model outputs. With this in mind, we may reevaluate the connections within the data-architecture-algorithm framework of a model, hoping that it would shed light into the geometry of high-dimensional and non-convex spaces in modern applications. In particular, we present a case that links the two observations: small and large batch gradient descent appear to converge to different basins of attraction but we show that they are in fact connected through their flat region and so belong to the same basin."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Entropy-SGD: biasing gradient descent into wide valleys",
                    "abstract": "This paper proposes a new optimization algorithm called Entropy-SGD for training deep neural networks that is motivated by the local geometry of the energy landscape. Local extrema with low generalization error have a large proportion of almost-zero eigenvalues in the Hessian with very few positive or negative eigenvalues. We leverage upon this observation to construct a local-entropy-based objective function that favors well-generalizable solutions lying in large flat regions of the energy landscape, while avoiding poorly-generalizable solutions located in the sharp valleys. Conceptually, our algorithm resembles two nested loops of SGD where we use Langevin dynamics in the inner loop to compute the gradient of the local entropy before each update of the weights. We show that the new objective has a smoother energy landscape and show improved generalization over SGD using uniform stability, under certain assumptions. Our experiments on convolutional and recurrent networks demonstrate that Entropy-SGD compares favorably to state-of-the-art techniques in terms of generalization error and training time."
                },
                {
                    "title": "Eigenvalues of the Hessian in Deep Learning: Singularity and Beyond",
                    "abstract": "We look at the eigenvalues of the Hessian of a loss function before and after training. The eigenvalue distribution is seen to be composed of two parts, the bulk which is concentrated around zero, and the edges which are scattered away from zero. We present empirical evidence for the bulk indicating how over-parametrized the system is, and for the edges that depend on the input data."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Optimizing Neural Networks with Kronecker-factored Approximate Curvature",
                    "abstract": "We propose an efficient method for approximating natural gradient descent in neural networks which we call Kronecker-Factored Approximate Curvature (K-FAC). K-FAC is based on an efficiently invertible approximation of a neural network's Fisher information matrix which is neither diagonal nor low-rank, and in some cases is completely non-sparse. It is derived by approximating various large blocks of the Fisher (corresponding to entire layers) as being the Kronecker product of two much smaller matrices. While only several times more expensive to compute than the plain stochastic gradient, the updates produced by K-FAC make much more progress optimizing the objective, which results in an algorithm that can be much faster than stochastic gradient descent with momentum in practice. And unlike some previously proposed approximate natural-gradient/Newton methods which use high-quality non-diagonal curvature matrices (such as Hessian-free optimization), K-FAC works very well in highly stochastic optimization regimes. This is because the cost of storing and inverting K-FAC's approximation to the curvature matrix does not depend on the amount of data used to estimate it, which is a feature typically associated only with diagonal or low-rank approximations to the curvature matrix."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "Wikipedia",
                    "abstract": "The task of this article is to analyze the political economy of Wikipedia. We discuss the specifics of Wikipedia\u2019s mode of production. The basic principles of what we call the info-communist mode of production will be presented. Our analysis is grounded in Marxist philosophy and Marxist political economy, and is connected to the current discourse about the renewal and reloading of the idea of communism that is undertaken by thinkers like Slavoj \u017di\u017eek and Alain Badiou. We explore to which extent Wikipedia encompasses principles that go beyond the capitalist mode of production and represent the info-communist mode of production. We present the subjective dimension of the mode of production (cooperative labor), the objective dimension of the mode of production (common ownership of the means of production), and the subject\u2013object dimension of the mode of production (the effects and products of the mode of production)."
                },
                {
                    "title": "On the Convergence of Block Coordinate Descent Type Methods",
                    "abstract": "In this paper we study smooth convex programming problems where the decision variables vector is split into several blocks of variables. We analyze the block coordinate gradient projection method in which each iteration consists of performing a gradient projection step with respect to a certain block taken in a cyclic order. Global sublinear rate of convergence of this method is established and it is shown that it can be accelerated when the problem is unconstrained. In the unconstrained setting we also prove a sublinear rate of convergence result for the so-called alternating minimization method when the number of blocks is two. When the objective function is also assumed to be strongly convex, linear rate of convergence is established."
                },
                {
                    "title": "Approximating Spectral Densities of Large Matrices",
                    "abstract": "In physics, it is sometimes desirable to compute the so-called density of states (DOS), also known as the spectral density, of a real symmetric matrix $A$. The spectral density can be viewed as a probability density distribution that measures the likelihood of finding eigenvalues near some point on the real line. The most straightforward way to obtain this density is to compute all eigenvalues of $A$, but this approach is generally costly and wasteful, especially for matrices of large dimension. There exist alternative methods that allow us to estimate the spectral density function at much lower cost. The major computational cost of these methods is in multiplying $A$ with a number of vectors, which makes them appealing for large-scale problems where products of the matrix $A$ with arbitrary vectors are relatively inexpensive. This article defines the problem of estimating the spectral density carefully and discusses how to measure the accuracy of an approximate spectral density. It then surveys a few kno..."
                },
                {
                    "title": "Numerical Methods for Large Eigenvalue Problems",
                    "abstract": "Preface to the Classics Edition Preface 1. Background in matrix theory and linear algebra 2. Sparse matrices 3. Perturbation theory and error analysis 4. The tools of spectral approximation 5. Subspace iteration 6. Krylov subspace methods 7. Filtering and restarting techniques 8. Preconditioning techniques 9. Non-standard eigenvalue problems 10. Origins of matrix eigenvalue problems References Index."
                },
                {
                    "title": "Randomized algorithms for estimating the trace of an implicit symmetric positive semi-definite matrix",
                    "abstract": "We analyze the convergence of randomized trace estimators. Starting at 1989, several algorithms have been proposed for estimating the trace of a matrix by 1/M\u2211<sub>i</sub>=1<sup>M</sup> z<sub><i>i</i></sub><sup><i>T</i></sup> <i>Az</i><sub><i>i</i></sub>, where the <i>z</i><sub><i>i</i></sub> are random vectors; different estimators use different distributions for the <i>z</i><sub><i>i</i></sub>s, all of which lead to <i>E</i>(1/M\u2211<sub><i>i</i></sub>=1<sup><i>M</i></sup> <i>z</i><sub><i>i</i></sub><sup>T</sup> <i>Az</i><sub><i>i</i></sub>) = trace(<i>A</i>). These algorithms are useful in applications in which there is no explicit representation of <i>A</i> but rather an efficient method compute <i>z</i><sup>T</sup><i>Az</i> given <i>z</i>. Existing results only analyze the variance of the different estimators. In contrast, we analyze the number of samples <i>M</i> required to guarantee that with probability at least 1\u2212\u0394, the relative error in the estimate is at most &epsis;. We argue that such bounds are much more useful in applications than the variance. We found that these bounds rank the estimators differently than the variance; this suggests that minimum-variance estimators may not be the best.\n We also make two additional contributions to this area. The first is a specialized bound for projection matrices, whose trace (rank) needs to be computed in electronic structure calculations. The second is a new estimator that uses less randomness than all the existing estimators."
                },
                {
                    "title": "Adaptive Subgradient Methods for Online Learning and Stochastic Optimization",
                    "abstract": "We present a new family of subgradient methods that dynamically incorporate knowledge of the geometry of the data observed in earlier iterations to perform more informative gradient-based learning. Metaphorically, the adaptation allows us to find needles in haystacks in the form of very predictive but rarely seen features. Our paradigm stems from recent advances in stochastic optimization and online learning which employ proximal functions to control the gradient steps of the algorithm. We describe and analyze an apparatus for adaptively modifying the proximal function, which significantly simplifies setting a learning rate and results in regret guarantees that are provably as good as the best proximal function that can be chosen in hindsight. We give several efficient algorithms for empirical risk minimization problems with common and important regularization functions and domain constraints. We experimentally study our theoretical analysis and show that adaptive subgradient methods outperform state-of-the-art, yet non-adaptive, subgradient algorithms."
                },
                {
                    "title": "Matrices, Moments and Quadrature with Applications",
                    "abstract": "This computationally oriented book describes and explains the mathematical relationships among matrices, moments, orthogonal polynomials, quadrature rules, and the Lanczos and conjugate gradient algorithms. The book bridges different mathematical areas to obtain algorithms to estimate bilinear forms involving two vectors and a function of the matrix. The first part of the book provides the necessary mathematical background and explains the theory. The second part describes the applications and gives numerical examples of the algorithms and techniques developed in the first part.Applications addressed in the book include computing elements of functions of matrices; obtaining estimates of the error norm in iterative methods for solving linear systems and computing parameters in least squares and total least squares; and solving ill-posed problems using Tikhonov regularization.This book will interest researchers in numerical linear algebra and matrix computations, as well as scientists and engineers working on problems involving computation of bilinear forms."
                },
                {
                    "title": "Topmoumoute Online Natural Gradient Algorithm",
                    "abstract": "Guided by the goal of obtaining an optimization algorithm that is both fast and yields good generalization, we study the descent direction maximizing the decrease in generalization error or the probability of not increasing generalization error. The surprising result is that from both the Bayesian and frequentist perspectives this can yield the natural gradient direction. Although that direction can be very expensive to compute we develop an efficient, general, online approximation to the natural gradient descent which is suited to large scale problems. We report experimental results showing much faster convergence in computation time and in number of iterations with TONGA (Topmoumoute Online natural Gradient Algorithm) than with stochastic gradient descent, even on very large datasets."
                },
                {
                    "title": "An Introduction to Numerical Methods and Analysis",
                    "abstract": "Praise for the First Edition. . . outstandingly appealing with regard to its style, contents, considerations of requirements of practice, choice of examples, and exercises.Zentralblatt MATH. . . carefully structured with many detailed worked examples.The Mathematical GazetteThe Second Edition of the highly regarded An Introduction to Numerical Methods and Analysis provides a fully revised guide to numerical approximation. The book continues to be accessible and expertly guides readers through the many available techniques of numerical methods and analysis.An Introduction to Numerical Methods and Analysis, Second Edition reflects the latest trends in the field, includes new material and revised exercises, and offers a unique emphasis on applications. The author clearly explains how to both construct and evaluate approximations for accuracy and performance, which are key skills in a variety of fields. A wide range of higher-level methods and solutions, including new topics such as the roots of polynomials, spectral collocation, finite element ideas, and Clenshaw-Curtis quadrature, are presented from an introductory perspective, and the Second Edition also features:Chapters and sections that begin with basic, elementary material followed by gradual coverage of more advanced materialExercises ranging from simple hand computations to challenging derivations and minor proofs to programming exercisesWidespread exposure and utilization of MATLABAn appendix that contains proofs of various theorems and other materialThe book is an ideal textbook for students in advanced undergraduate mathematics and engineering courses who are interested in gaining an understanding of numerical methods and numerical analysis."
                },
                {
                    "title": "Fast Exact Multiplication by the Hessian",
                    "abstract": "Just storing the Hessian H (the matrix of second derivatives 2E/wiwj of the error E with respect to each pair of weights) of a large neural network is difficult. Since a common use of a large matrix like H is to compute its product with various vectors, we derive a technique that directly calculates Hv, where v is an arbitrary vector. To calculate Hv, we first define a differential operator Rv{f(w)} = (/r)f(w rv)|r=0, note that Rv{w} = Hv and Rv{w} = v, and then apply Rv{} to the equations used to compute w. The result is an exact and numerically stable procedure for computing Hv, which takes about as much computation, and is about as local, as a gradient evaluation. We then apply the technique to a one pass gradient calculation algorithm (backpropagation), a relaxation gradient calculation algorithm (recurrent backpropagation), and two stochastic gradient calculation algorithms (Boltzmann machines and weight perturbation). Finally, we show that this technique can be used at the heart of many iterative techniques for computing various properties of H, obviating any need to calculate the full Hessian."
                },
                {
                    "title": "Calculation of Gauss quadrature rules",
                    "abstract": "Most numerical integration techniques consist of approximating the integrand by a polynomial in a region or regions and then integrating the polynomial exactly. Often a complicated integrand can be factored into a non-negative ''weight'' function and another function better approximated by a polynomial, thus $\\int_{a}^{b} g(t)dt = \\int_{a}^{b} \\omega (t)f(t)dt \\approx \\sum_{i=1}^{N} w_i f(t_i)$. Hopefully, the quadrature rule ${\\{w_j, t_j\\}}_{j=1}^{N}$ corresponding to the weight function $\\omega$(t) is available in tabulated form, but more likely it is not. We present here two algorithms for generating the Gaussian quadrature rule defined by the weight function when: a) the three term recurrence relation is known for the orthogonal polynomials generated by $\\omega$(t), and b) the moments of the weight function are known or can be calculated."
                },
                {
                    "title": "On best conditioned matrices",
                    "abstract": "1. Main theorems. Let A be a positive definite Hermitian matrix of finite order, and let A and X be its maximal and minimal eigenvalue respectively. The condition number of A is the ratio P(A) =A/X introduced by Todd [1]. Let 'G be a class of regular linear transformations. Define ATT*A T. We say that A is best conditioned with respect to 'G if P(A T) > P(A) for all TEE T. In order to investigate whether A is best conditioned we remember that"
                },
                {
                    "title": "RMSprop converges with proper hyper-parameter",
                    "abstract": "Despite the existence of divergence examples"
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nWhy does stochastic gradient descent (SGD) perform significantly worse than the Adam optimizer when training Transformer models, but not when training convolutional neural networks (CNNs)?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding the performance gap between SGD and Adam on Transformers is crucial for the research community as it can lead to deeper insights into the training dynamics of neural networks. By addressing this question, we can advance theoretical knowledge about optimization in deep learning, potentially leading to the development of more effective training algorithms. This could have practical applications in improving the efficiency and effectiveness of training Transformer models, which are widely used in various AI applications, thereby influencing future research directions in optimization techniques and neural network architectures.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge in solving this problem lies in the complex nature of the Hessian spectrum and its relationship to optimization performance. Naive approaches may fail because they do not account for the architectural differences between Transformers and CNNs, particularly the \"block heterogeneity\" in Transformers. This heterogeneity complicates the optimization landscape, making it difficult for SGD to effectively navigate the loss surface. Additionally, the need for a fine-grained analysis of the blockwise Hessian spectrum introduces technical and theoretical obstacles that require sophisticated numerical linear algebra techniques to overcome.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the overall Hessian spectrum without dissecting it into blockwise components, which has limited the understanding of how architectural differences impact optimization. Existing solutions have not adequately addressed the unique characteristics of Transformers, such as the non-sequential stacking of disparate parameter blocks. Barriers to solving this problem include a lack of empirical evidence linking block heterogeneity to optimization performance and insufficient methodologies for analyzing the blockwise Hessian spectrum. Our approach differs by specifically investigating the blockwise Hessian spectrum and its implications for SGD's performance, providing a more nuanced understanding of the optimization challenges in Transformers.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a detailed examination of the blockwise Hessian spectrum of Transformer models compared to CNNs. We will utilize numerical linear algebra techniques to analyze the Hessian spectra of individual parameter blocks within Transformers, focusing on identifying the phenomenon of \"block heterogeneity.\" The dataset will consist of various Transformer architectures and CNNs, and we will measure optimization performance using standard metrics"
            }
        },
        "author_data": {
            "e8eb8274-4037-4680-ae82-173e481f0d0a": {
                "pk": "e8eb8274-4037-4680-ae82-173e481f0d0a",
                "name": "Yushun Zhang",
                "collaborators": [
                    "Ruoyu Sun",
                    "Zhi-Quan Luo",
                    "Jiawei Zhang",
                    "Mingyi Hong",
                    "Yizhuo Cai",
                    "Bo Lei",
                    "Qianying Zhao",
                    "Jing Peng",
                    "Min Wei",
                    "Xing Zhang"
                ],
                "domain": [
                    "Neural Networks",
                    "Federated Learning",
                    "Machine Learning",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "When Expressivity Meets Trainability: Fewer than $n$ Neurons Can Work",
                        "abstract": "Modern neural networks are often quite wide, causing large memory and computation costs. It is thus of great interest to train a narrower network. However, training narrow neural nets remains a challenging task. We ask two theoretical questions: Can narrow networks have as strong expressivity as wide ones? If so, does the loss function exhibit a benign optimization landscape? In this work, we provide partially affirmative answers to both questions for 1-hidden-layer networks with fewer than $n$ (sample size) neurons when the activation is smooth. First, we prove that as long as the width $m \\geq 2n/d$ (where $d$ is the input dimension), its expressivity is strong, i.e., there exists at least one global minimizer with zero training loss. Second, we identify a nice local region with no local-min or saddle points. Nevertheless, it is not clear whether gradient descent can stay in this nice region. Third, we consider a constrained optimization formulation where the feasible region is the nice local region, and prove that every KKT point is a nearly global minimizer. It is expected that projected gradient methods converge to KKT points under mild technical conditions, but we leave the rigorous convergence analysis to future work. Thorough numerical results show that projected gradient methods on this constrained formulation significantly outperform SGD for training narrow neural nets."
                    },
                    {
                        "title": "Communication Efficiency Optimization of Federated Learning for Computing and Network Convergence of 6G Networks",
                        "abstract": "Federated learning effectively addresses issues such as data privacy by collaborating across participating devices to train global models. However, factors such as network topology and device computing power can affect its training or communication process in complex network environments. A new network architecture and paradigm with computing-measurable, perceptible, distributable, dispatchable, and manageable capabilities, computing and network convergence (CNC) of 6G networks can effectively support federated learning training and improve its communication efficiency. By guiding the participating devices' training in federated learning based on business requirements, resource load, network conditions, and arithmetic power of devices, CNC can reach this goal. In this paper, to improve the communication efficiency of federated learning in complex networks, we study the communication efficiency optimization of federated learning for computing and network convergence of 6G networks, methods that gives decisions on its training process for different network conditions and arithmetic power of participating devices in federated learning. The experiments address two architectures that exist for devices in federated learning and arrange devices to participate in training based on arithmetic power while achieving optimization of communication efficiency in the process of transferring model parameters. The results show that the method we proposed can (1) cope well with complex network situations (2) effectively balance the delay distribution of participating devices for local training (3) improve the communication efficiency during the transfer of model parameters (4) improve the resource utilization in the network."
                    },
                    {
                        "title": "Adam Can Converge Without Any Modification On Update Rules",
                        "abstract": "Ever since Reddi et al. 2018 pointed out the divergence issue of Adam, many new variants have been designed to obtain convergence. However, vanilla Adam remains exceptionally popular and it works well in practice. Why is there a gap between theory and practice? We point out there is a mismatch between the settings of theory and practice: Reddi et al. 2018 pick the problem after picking the hyperparameters of Adam, i.e., $(\\beta_1, \\beta_2)$; while practical applications often fix the problem first and then tune $(\\beta_1, \\beta_2)$. Due to this observation, we conjecture that the empirical convergence can be theoretically justified, only if we change the order of picking the problem and hyperparameter. In this work, we confirm this conjecture. We prove that, when $\\beta_2$ is large and $\\beta_1 < \\sqrt{\\beta_2}<1$, Adam converges to the neighborhood of critical points. The size of the neighborhood is propositional to the variance of stochastic gradients. Under an extra condition (strong growth condition), Adam converges to critical points. It is worth mentioning that our results cover a wide range of hyperparameters: as $\\beta_2$ increases, our convergence result can cover any $\\beta_1 \\in [0,1)$ including $\\beta_1=0.9$, which is the default setting in deep learning libraries. To our knowledge, this is the first result showing that Adam can converge without any modification on its update rules. Further, our analysis does not require assumptions of bounded gradients or bounded 2nd-order momentum. When $\\beta_2$ is small, we further point out a large region of $(\\beta_1,\\beta_2)$ where Adam can diverge to infinity. Our divergence result considers the same setting as our convergence result, indicating a phase transition from divergence to convergence when increasing $\\beta_2$. These positive and negative results can provide suggestions on how to tune Adam hyperparameters."
                    },
                    {
                        "title": "MoFO: Momentum-Filtered Optimizer for Mitigating Forgetting in LLM Fine-Tuning",
                        "abstract": "Recently, large language models (LLMs) have demonstrated remarkable capabilities in a wide range of tasks. Typically, an LLM is pre-trained on large corpora and subsequently fine-tuned on task-specific datasets. However, during fine-tuning, LLMs may forget the knowledge acquired in the pre-training stage, leading to a decline in general capabilities. To address this issue, we propose a new fine-tuning algorithm termed Momentum-Filtered Optimizer (MoFO). The key idea of MoFO is to iteratively select and update the model parameters with the largest momentum magnitudes. Compared to full-parameter training, MoFO achieves similar fine-tuning performance while keeping parameters closer to the pre-trained model, thereby mitigating knowledge forgetting. Unlike most existing methods for forgetting mitigation, MoFO combines the following two advantages. First, MoFO does not require access to pre-training data. This makes MoFO particularly suitable for fine-tuning scenarios where pre-training data is unavailable, such as fine-tuning checkpoint-only open-source LLMs. Second, MoFO does not alter the original loss function. This could avoid impairing the model performance on the fine-tuning tasks. We validate MoFO through rigorous convergence analysis and extensive experiments, demonstrating its superiority over existing methods in mitigating forgetting and enhancing fine-tuning performance."
                    },
                    {
                        "title": "Uncertainty and Explainable Analysis of Machine Learning Model for Reconstruction of Sonic Slowness Logs",
                        "abstract": "Logs are valuable information for oil and gas fields as they help to determine the lithology of the formations surrounding the borehole and the location and reserves of subsurface oil and gas reservoirs. However, important logs are often missing in horizontal or old wells, which poses a challenge in field applications. In this paper, we utilize data from the 2020 machine learning competition of the SPWLA, which aims to predict the missing compressional wave slowness and shear wave slowness logs using other logs in the same borehole. We employ the NGBoost algorithm to construct an Ensemble Learning model that can predicate the results as well as their uncertainty. Furthermore, we combine the SHAP method to investigate the interpretability of the machine learning model. We compare the performance of the NGBosst model with four other commonly used Ensemble Learning methods, including Random Forest, GBDT, XGBoost, LightGBM. The results show that the NGBoost model performs well in the testing set and can provide a probability distribution for the prediction results. In addition, the variance of the probability distribution of the predicted log can be used to justify the quality of the constructed log. Using the SHAP explainable machine learning model, we calculate the importance of each input log to the predicted results as well as the coupling relationship among input logs. Our findings reveal that the NGBoost model tends to provide greater slowness prediction results when the neutron porosity and gamma ray are large, which is consistent with the cognition of petrophysical models. Furthermore, the machine learning model can capture the influence of the changing borehole caliper on slowness, where the influence of borehole caliper on slowness is complex and not easy to establish a direct relationship. These findings are in line with the physical principle of borehole acoustics."
                    }
                ]
            },
            "726103a2-7504-439d-abc2-d1e0754b3012": {
                "pk": "726103a2-7504-439d-abc2-d1e0754b3012",
                "name": "Congliang Chen",
                "collaborators": [
                    "Li Shen",
                    "Wei Liu",
                    "Zhi-Quan Luo",
                    "Ruoyu Sun",
                    "Fangyu Zou",
                    "Yushun Zhang",
                    "Ziniu Li",
                    "Shuyang Gu",
                    "Jing Liao",
                    "Lu Yuan"
                ],
                "domain": [
                    "Optimization",
                    "Deep Learning",
                    "Neural Networks",
                    "Style Transfer"
                ],
                "publications": [
                    {
                        "title": "Arbitrary Style Transfer with Deep Feature Reshuffle",
                        "abstract": "This paper introduces a novel method by reshuffling deep features (i.e., permuting the spacial locations of a feature map) of the style image for arbitrary style transfer. We theoretically prove that our new style loss based on reshuffle connects both global and local style losses respectively used by most parametric and non-parametric neural style transfer methods. This simple idea can effectively address the challenging issues in existing style transfer methods. On one hand, it can avoid distortions in local style patterns, and allow semantic-level transfer, compared with neural parametric methods. On the other hand, it can preserve globally similar appearance to the style image, and avoid wash-out artifacts, compared with neural non-parametric methods. Based on the proposed loss, we also present a progressive feature-domain optimization approach. The experiments show that our method is widely applicable to various styles, and produces better quality than existing methods."
                    },
                    {
                        "title": "Towards Practical Adam: Non-Convexity, Convergence Theory, and Mini-Batch Acceleration",
                        "abstract": "Adam is one of the most influential adaptive stochastic algorithms for training deep neural networks, which has been pointed out to be divergent even in the simple convex setting via a few simple counterexamples. Many attempts, such as decreasing an adaptive learning rate, adopting a big batch size, incorporating a temporal decorrelation technique, seeking an analogous surrogate, \\textit{etc.}, have been tried to promote Adam-type algorithms to converge. In contrast with existing approaches, we introduce an alternative easy-to-check sufficient condition, which merely depends on the parameters of the base learning rate and combinations of historical second-order moments, to guarantee the global convergence of generic Adam for solving large-scale non-convex stochastic optimization. This observation, coupled with this sufficient condition, gives much deeper interpretations on the divergence of Adam. On the other hand, in practice, mini-Adam and distributed-Adam are widely used without any theoretical guarantee. We further give an analysis on how the batch size or the number of nodes in the distributed system affects the convergence of Adam, which theoretically shows that mini-batch and distributed Adam can be linearly accelerated by using a larger mini-batch size or a larger number of nodes.At last, we apply the generic Adam and mini-batch Adam with the sufficient condition for solving the counterexample and training several neural networks on various real-world datasets. Experimental results are exactly in accord with our theoretical analysis."
                    },
                    {
                        "title": "Quantized Adam with Error Feedback",
                        "abstract": "In this paper, we present a distributed variant of adaptive stochastic gradient method for training deep neural networks in the parameter-server model. To reduce the communication cost among the workers and server, we incorporate two types of quantization schemes, i.e., gradient quantization and weight quantization, into the proposed distributed Adam. Besides, to reduce the bias introduced by quantization operations, we propose an error-feedback technique to compensate for the quantized gradient. Theoretically, in the stochastic nonconvex setting, we show that the distributed adaptive gradient method with gradient quantization and error-feedback converges to the first-order stationary point, and that the distributed adaptive gradient method with weight quantization and error-feedback converges to the point related to the quantized level under both the single-worker and multi-worker modes. At last, we apply the proposed distributed adaptive gradient methods to train deep neural networks. Experimental results demonstrate the efficacy of our methods."
                    },
                    {
                        "title": "Efficient-Adam: Communication-Efficient Distributed Adam",
                        "abstract": "Distributed adaptive stochastic gradient methods have been widely used for large-scale nonconvex optimization, such as training deep learning models. However, their communication complexity on finding $\\varepsilon$-stationary points has rarely been analyzed in the nonconvex setting. In this work, we present a novel communication-efficient distributed Adam in the parameter-server model for stochastic nonconvex optimization, dubbed {\\em Efficient-Adam}. Specifically, we incorporate a two-way quantization scheme into Efficient-Adam to reduce the communication cost between the workers and server. Simultaneously, we adopt a two-way error feedback strategy to reduce the biases caused by the two-way quantization on both the server and workers, respectively. In addition, we establish the iteration complexity for the proposed Efficient-Adam with a class of quantization operators, and further characterize its communication complexity between the server and workers when an $\\varepsilon$-stationary point is achieved. Finally, we apply Efficient-Adam to solve a toy stochastic convex optimization problem and train deep learning models on real-world vision and language tasks. Extensive experiments together with a theoretical guarantee justify the merits of Efficient Adam."
                    },
                    {
                        "title": "Rethinking SIGN Training: Provable Nonconvex Acceleration without First- and Second-Order Gradient Lipschitz",
                        "abstract": "Sign-based stochastic methods have gained attention due to their ability to achieve robust performance despite using only the sign information for parameter updates. However, the current convergence analysis of sign-based methods relies on the strong assumptions of first-order gradient Lipschitz and second-order gradient Lipschitz, which may not hold in practical tasks like deep neural network training that involve high non-smoothness. In this paper, we revisit sign-based methods and analyze their convergence under more realistic assumptions of first- and second-order smoothness. We first establish the convergence of the sign-based method under weak first-order Lipschitz. Motivated by the weak first-order Lipschitz, we propose a relaxed second-order condition that still allows for nonconvex acceleration in sign-based methods. Based on our theoretical results, we gain insights into the computational advantages of the recently developed LION algorithm. In distributed settings, we prove that this nonconvex acceleration persists with linear speedup in the number of nodes, when utilizing fast communication compression gossip protocols. The novelty of our theoretical results lies in that they are derived under much weaker assumptions, thereby expanding the provable applicability of sign-based algorithms to a wider range of problems."
                    },
                    {
                        "title": "A Unified Analysis of AdaGrad with Weighted Aggregation and Momentum Acceleration",
                        "abstract": "Integrating adaptive learning rate and momentum techniques into SGD leads to a large class of efficiently accelerated adaptive stochastic algorithms, such as AdaGrad, RMSProp, Adam, AccAdaGrad, \\textit{etc}. In spite of their effectiveness in practice, there is still a large gap in their theories of convergences, especially in the difficult non-convex stochastic setting. To fill this gap, we propose \\emph{weighted AdaGrad with unified momentum}, dubbed AdaUSM, which has the main characteristics that (1) it incorporates a unified momentum scheme which covers both the heavy ball momentum and the Nesterov accelerated gradient momentum; (2) it adopts a novel weighted adaptive learning rate that can unify the learning rates of AdaGrad, AccAdaGrad, Adam, and RMSProp. Moreover, when we take polynomially growing weights in AdaUSM, we obtain its $\\mathcal{O}(\\log(T)/\\sqrt{T})$ convergence rate in the non-convex stochastic setting. We also show that the adaptive learning rates of Adam and RMSProp correspond to taking exponentially growing weights in AdaUSM, thereby providing a new perspective for understanding Adam and RMSProp. Lastly, comparative experiments of AdaUSM against SGD with momentum, AdaGrad, AdaEMA, Adam, and AMSGrad on various deep learning models and datasets are also carried out."
                    },
                    {
                        "title": "Adam Can Converge Without Any Modification On Update Rules",
                        "abstract": "Ever since Reddi et al. 2018 pointed out the divergence issue of Adam, many new variants have been designed to obtain convergence. However, vanilla Adam remains exceptionally popular and it works well in practice. Why is there a gap between theory and practice? We point out there is a mismatch between the settings of theory and practice: Reddi et al. 2018 pick the problem after picking the hyperparameters of Adam, i.e., $(\\beta_1, \\beta_2)$; while practical applications often fix the problem first and then tune $(\\beta_1, \\beta_2)$. Due to this observation, we conjecture that the empirical convergence can be theoretically justified, only if we change the order of picking the problem and hyperparameter. In this work, we confirm this conjecture. We prove that, when $\\beta_2$ is large and $\\beta_1 < \\sqrt{\\beta_2}<1$, Adam converges to the neighborhood of critical points. The size of the neighborhood is propositional to the variance of stochastic gradients. Under an extra condition (strong growth condition), Adam converges to critical points. It is worth mentioning that our results cover a wide range of hyperparameters: as $\\beta_2$ increases, our convergence result can cover any $\\beta_1 \\in [0,1)$ including $\\beta_1=0.9$, which is the default setting in deep learning libraries. To our knowledge, this is the first result showing that Adam can converge without any modification on its update rules. Further, our analysis does not require assumptions of bounded gradients or bounded 2nd-order momentum. When $\\beta_2$ is small, we further point out a large region of $(\\beta_1,\\beta_2)$ where Adam can diverge to infinity. Our divergence result considers the same setting as our convergence result, indicating a phase transition from divergence to convergence when increasing $\\beta_2$. These positive and negative results can provide suggestions on how to tune Adam hyperparameters."
                    },
                    {
                        "title": "Entropic Distribution Matching in Supervised Fine-tuning of LLMs: Less Overfitting and Better Diversity",
                        "abstract": "Large language models rely on Supervised Fine-Tuning (SFT) to specialize in downstream tasks. Cross Entropy (CE) loss is the de facto choice in SFT, but it often leads to overfitting and limited output diversity due to its aggressive updates to the data distribution. This paper aim to address these issues by introducing the maximum entropy principle, which favors models with flatter distributions that still effectively capture the data. Specifically, we develop a new distribution matching method called GEM, which solves reverse Kullback-Leibler divergence minimization with an entropy regularizer.   For the SFT of Llama-3-8B models, GEM outperforms CE in several aspects. First, when applied to the UltraFeedback dataset to develop general instruction-following abilities, GEM exhibits reduced overfitting, evidenced by lower perplexity and better performance on the IFEval benchmark. Furthermore, GEM enhances output diversity, leading to performance gains of up to 7 points on math reasoning and code generation tasks using best-of-n sampling, even without domain-specific data. Second, when fine-tuning with domain-specific datasets for math reasoning and code generation, GEM also shows less overfitting and improvements of up to 10 points compared with CE."
                    },
                    {
                        "title": "Adam-mini: Use Fewer Learning Rates To Gain More",
                        "abstract": "We propose Adam-mini, an optimizer that achieves on-par or better performance than AdamW with 45% to 50% less memory footprint. Adam-mini reduces memory by cutting down the learning rate resources in Adam (i.e., $1/\\sqrt{v}$). We find that $\\geq$ 90% of these learning rates in $v$ could be harmlessly removed if we (1) carefully partition the parameters into blocks following our proposed principle on Hessian structure; (2) assign a single but good learning rate to each parameter block. We further find that, for each of these parameter blocks, there exists a single high-quality learning rate that can outperform Adam, provided that sufficient resources are available to search it out. We then provide one cost-effective way to find good learning rates and propose Adam-mini. Empirically, we verify that Adam-mini performs on par or better than AdamW on various language models sized from 125M to 7B for pre-training, supervised fine-tuning, and RLHF. The reduced memory footprint of Adam-mini also alleviates communication overheads among GPUs and CPUs, thereby increasing throughput. For instance, Adam-mini achieves 49.6% higher throughput than AdamW when pre-training Llama2-7B on $2\\times$ A800-80GB GPUs, which saves 33% wall-clock time for pre-training."
                    }
                ]
            },
            "95145878-4eac-4d8e-89a5-bdea47b6349f": {
                "pk": "95145878-4eac-4d8e-89a5-bdea47b6349f",
                "name": "Tian Ding",
                "collaborators": [
                    "Kui Ren",
                    "Dawei Li",
                    "Ruoyu Sun",
                    "Sarah Vallelian",
                    "Hong Shen",
                    "Wei Xu",
                    "Chunming Zhao"
                ],
                "domain": [
                    "Inverse Problems",
                    "Medical Imaging",
                    "Neural Networks",
                    "Signal Processing"
                ],
                "publications": [
                    {
                        "title": "Inverse transport calculations in optical imaging with subspace optimization algorithms",
                        "abstract": "Inverse boundary value problems for the radiative transport equation play important roles in optics-based medical imaging techniques such as diffuse optical tomography (DOT) and fluorescence optical tomography (FOT). Despite the rapid progress in the mathematical theory and numerical computation of these inverse problems in recent years, developing robust and efficient reconstruction algorithms remains as a challenging task and an active research topic. We propose here a robust reconstruction method that is based on subspace minimization techniques. The method splits the unknown transport solution (or a functional of it) into low-frequency and high-frequency components, and uses singular value decomposition to analytically recover part of low-frequency information. Minimization is then applied to recover part of the high-frequency components of the unknowns. We present some numerical simulations with synthetic data to demonstrate the performance of the proposed algorithm."
                    },
                    {
                        "title": "Sub-Optimal Local Minima Exist for Neural Networks with Almost All Non-Linear Activations",
                        "abstract": "Does over-parameterization eliminate sub-optimal local minima for neural networks? An affirmative answer was given by a classical result in [59] for 1-hidden-layer wide neural networks. A few recent works have extended the setting to multi-layer neural networks, but none of them has proved every local minimum is global. Why is this result never extended to deep networks?   In this paper, we show that the task is impossible because the original result for 1-hidden-layer network in [59] can not hold. More specifically, we prove that for any multi-layer network with generic input data and non-linear activation functions, sub-optimal local minima can exist, no matter how wide the network is (as long as the last hidden layer has at least two neurons). While the result of [59] assumes sigmoid activation, our counter-example covers a large set of activation functions (dense in the set of continuous functions), indicating that the limitation is not due to the specific activation. Our result indicates that \"no bad local-min\" may be unable to explain the benefit of over-parameterization for training neural nets."
                    },
                    {
                        "title": "A one-step reconstruction algorithm for quantitative photoacoustic imaging",
                        "abstract": "Quantitative photoacoustic tomography (QPAT) is a recent hybrid imaging modality that couples optical tomography with ultrasound imaging to achieve high resolution imaging of optical properties of scattering media. Image reconstruction in QPAT is usually a two-step process. In the first step, the initial pressure field inside the medium, generated by the photoacoustic effect, is reconstructed using measured acoustic data. In the second step, this initial ultrasound pressure field datum is used to reconstruct optical properties of the medium. We propose in this work a one-step inversion algorithm for image reconstruction in QPAT that reconstructs the optical absorption coefficient directly from measured acoustic data. The algorithm can be used to recover simultaneously the absorption coefficient and the ultrasound speed of the medium from \\emph{multiple} acoustic data sets, with appropriate \\emph{a priori} bounds on the unknowns. We demonstrate, through numerical simulations based on synthetic data, the feasibility of the proposed reconstruction method."
                    },
                    {
                        "title": "Beamforming Design with Fast Convergence for IRS-Aided Full-Duplex Communication",
                        "abstract": "We study the beamforming optimization for an intelligent reflecting surface (IRS)-aided full-duplex (FD) communication system in this letter. Specifically, we maximize the sum rate of bi-directional transmissions by jointly optimizing the transmit beamforming and the beamforming of the IRS reflection. A fast converging alternating algorithm is developed to tackle this problem. In each iteration of the proposed algorithm, the solutions to the transmit beamforming and the IRS reflect beamforming are obtained in a semi-closed form and a closed form, respectively. Compared to an existing method based on the Arimoto-Blahut algorithm, the proposed method achieves almost the same performance while enjoying much faster convergence and lower computational complexity."
                    }
                ]
            },
            "b3a0380d-9893-4fa8-a676-7ad6e8878341": {
                "pk": "b3a0380d-9893-4fa8-a676-7ad6e8878341",
                "name": "Ziniu Li",
                "collaborators": [
                    "Tian Xu",
                    "Yang Yu",
                    "Zhi-Quan Luo",
                    "Ke Xu",
                    "Liu Liu",
                    "Lanqing Li",
                    "Deheng Ye",
                    "Peilin Zhao",
                    "Congliang Chen",
                    "Zeyu Qin"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Imitation Learning",
                    "Natural Language Processing",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "A Note on Target Q-learning For Solving Finite MDPs with A Generative Oracle",
                        "abstract": "Q-learning with function approximation could diverge in the off-policy setting and the target network is a powerful technique to address this issue. In this manuscript, we examine the sample complexity of the associated target Q-learning algorithm in the tabular case with a generative oracle. We point out a misleading claim in [Lee and He, 2020] and establish a tight analysis. In particular, we demonstrate that the sample complexity of the target Q-learning algorithm in [Lee and He, 2020] is $\\widetilde{\\mathcal O}(|\\mathcal S|^2|\\mathcal A|^2 (1-\\gamma)^{-5}\\varepsilon^{-2})$. Furthermore, we show that this sample complexity is improved to $\\widetilde{\\mathcal O}(|\\mathcal S||\\mathcal A| (1-\\gamma)^{-5}\\varepsilon^{-2})$ if we can sequentially update all state-action pairs and $\\widetilde{\\mathcal O}(|\\mathcal S||\\mathcal A| (1-\\gamma)^{-4}\\varepsilon^{-2})$ if $\\gamma$ is further in $(1/2, 1)$. Compared with the vanilla Q-learning, our results conclude that the introduction of a periodically-frozen target Q-function does not sacrifice the sample complexity."
                    },
                    {
                        "title": "On Value Discrepancy of Imitation Learning",
                        "abstract": "Imitation learning trains a policy from expert demonstrations. Imitation learning approaches have been designed from various principles, such as behavioral cloning via supervised learning, apprenticeship learning via inverse reinforcement learning, and GAIL via generative adversarial learning. In this paper, we propose a framework to analyze the theoretical property of imitation learning approaches based on discrepancy propagation analysis. Under the infinite-horizon setting, the framework leads to the value discrepancy of behavioral cloning in an order of O((1-\\gamma)^{-2}). We also show that the framework leads to the value discrepancy of GAIL in an order of O((1-\\gamma)^{-1}). It implies that GAIL has less compounding errors than behavioral cloning, which is also verified empirically in this paper. To the best of our knowledge, we are the first one to analyze GAIL's performance theoretically. The above results indicate that the proposed framework is a general tool to analyze imitation learning approaches. We hope our theoretical results can provide insights for future improvements in imitation learning algorithms."
                    },
                    {
                        "title": "Error Bounds of Imitating Policies and Environments",
                        "abstract": "Imitation learning trains a policy by mimicking expert demonstrations. Various imitation methods were proposed and empirically evaluated, meanwhile, their theoretical understanding needs further studies. In this paper, we firstly analyze the value gap between the expert policy and imitated policies by two imitation methods, behavioral cloning and generative adversarial imitation. The results support that generative adversarial imitation can reduce the compounding errors compared to behavioral cloning, and thus has a better sample complexity. Noticed that by considering the environment transition model as a dual agent, imitation learning can also be used to learn the environment model. Therefore, based on the bounds of imitating policies, we further analyze the performance of imitating environments. The results show that environment models can be more effectively imitated by generative adversarial imitation than behavioral cloning, suggesting a novel application of adversarial imitation for model-based reinforcement learning. We hope these results could inspire future advances in imitation learning and model-based reinforcement learning."
                    },
                    {
                        "title": "Policy Optimization in RLHF: The Impact of Out-of-preference Data",
                        "abstract": "Aligning intelligent agents with human preferences and values is important. This paper examines two popular alignment methods: Direct Preference Optimization (DPO) and Reward-Model-Based Policy Optimization (RMB-PO). A variant of RMB-PO, referred to as RMB-PO+ is also considered. These methods, either explicitly or implicitly, learn a reward model from preference data and differ in the data used for policy optimization to unlock the generalization ability of the reward model. In particular, compared with DPO, RMB-PO additionally uses policy-generated data, and RMB-PO+ further leverages new, preference-free data. We examine the impact of such out-of-preference data. Our study, conducted through controlled and synthetic experiments, demonstrates that DPO performs poorly, whereas RMB-PO+ performs the best. In particular, even when providing the policy model with a good feature representation, we find that policy optimization with adequate out-of-preference data significantly improves performance by harnessing the reward model's generalization capabilities."
                    },
                    {
                        "title": "On Generalization of Adversarial Imitation Learning and Beyond",
                        "abstract": "Despite massive empirical evaluations, one of the fundamental questions in imitation learning is still not fully settled: does AIL (adversarial imitation learning) provably generalize better than BC (behavioral cloning)? We study this open problem with tabular and episodic MDPs. For vanilla AIL that uses the direct maximum likelihood estimation, we provide both negative and positive answers under the known transition setting. For some MDPs, we show that vanilla AIL has a worse sample complexity than BC. The key insight is that the state-action distribution matching principle is weak so that AIL may generalize poorly even on visited states from the expert demonstrations. For another class of MDPs, vanilla AIL is proved to generalize well even on non-visited states. Interestingly, its sample complexity is horizon-free, which provably beats BC by a wide margin. Finally, we establish a framework in the unknown transition scenario, which allows AIL to explore via reward-free exploration strategies. Compared with the best-known online apprenticeship learning algorithm, the resulting algorithm improves the sample complexity and interaction complexity."
                    },
                    {
                        "title": "Rethinking ValueDice: Does It Really Improve Performance?",
                        "abstract": "Since the introduction of GAIL, adversarial imitation learning (AIL) methods attract lots of research interests. Among these methods, ValueDice has achieved significant improvements: it beats the classical approach Behavioral Cloning (BC) under the offline setting, and it requires fewer interactions than GAIL under the online setting. Are these improvements benefited from more advanced algorithm designs? We answer this question by the following conclusions. First, we show that ValueDice could reduce to BC under the offline setting. Second, we verify that overfitting exists and regularization matters in the low-data regime. Specifically, we demonstrate that with weight decay, BC also nearly matches the expert performance as ValueDice does. The first two claims explain the superior offline performance of ValueDice. Third, we establish that ValueDice does not work when the expert trajectory is subsampled. Instead, the mentioned success of ValueDice holds when the expert trajectory is complete, in which ValueDice is closely related to BC that performs well as mentioned. Finally, we discuss the implications of our research for imitation learning studies beyond ValueDice."
                    },
                    {
                        "title": "Theoretical Analysis of Offline Imitation With Supplementary Dataset",
                        "abstract": "Behavioral cloning (BC) can recover a good policy from abundant expert data, but may fail when expert data is insufficient. This paper considers a situation where, besides the small amount of expert data, a supplementary dataset is available, which can be collected cheaply from sub-optimal policies. Imitation learning with a supplementary dataset is an emergent practical framework, but its theoretical foundation remains under-developed. To advance understanding, we first investigate a direct extension of BC, called NBCU, that learns from the union of all available data. Our analysis shows that, although NBCU suffers an imitation gap that is larger than BC in the worst case, there exist special cases where NBCU performs better than or equally well as BC. This discovery implies that noisy data can also be helpful if utilized elaborately. Therefore, we further introduce a discriminator-based importance sampling technique to re-weight the supplementary data, proposing the WBCU method. With our newly developed landscape-based analysis, we prove that WBCU can outperform BC in mild conditions. Empirical studies show that WBCU simultaneously achieves the best performance on two challenging tasks where prior state-of-the-art methods fail."
                    },
                    {
                        "title": "Understanding Adversarial Imitation Learning in Small Sample Regime: A Stage-coupled Analysis",
                        "abstract": "Imitation learning learns a policy from expert trajectories. While the expert data is believed to be crucial for imitation quality, it was found that a kind of imitation learning approach, adversarial imitation learning (AIL), can have exceptional performance. With as little as only one expert trajectory, AIL can match the expert performance even in a long horizon, on tasks such as locomotion control. There are two mysterious points in this phenomenon. First, why can AIL perform well with only a few expert trajectories? Second, why does AIL maintain good performance despite the length of the planning horizon? In this paper, we theoretically explore these two questions. For a total-variation-distance-based AIL (called TV-AIL), our analysis shows a horizon-free imitation gap $\\mathcal O(\\{\\min\\{1, \\sqrt{|\\mathcal S|/N} \\})$ on a class of instances abstracted from locomotion control tasks. Here $|\\mathcal S|$ is the state space size for a tabular Markov decision process, and $N$ is the number of expert trajectories. We emphasize two important features of our bound. First, this bound is meaningful in both small and large sample regimes. Second, this bound suggests that the imitation gap of TV-AIL is at most 1 regardless of the planning horizon. Therefore, this bound can explain the empirical observation. Technically, we leverage the structure of multi-stage policy optimization in TV-AIL and present a new stage-coupled analysis via dynamic programming"
                    },
                    {
                        "title": "Provably Efficient Adversarial Imitation Learning with Unknown Transitions",
                        "abstract": "Imitation learning (IL) has proven to be an effective method for learning good policies from expert demonstrations. Adversarial imitation learning (AIL), a subset of IL methods, is particularly promising, but its theoretical foundation in the presence of unknown transitions has yet to be fully developed. This paper explores the theoretical underpinnings of AIL in this context, where the stochastic and uncertain nature of environment transitions presents a challenge. We examine the expert sample complexity and interaction complexity required to recover good policies. To this end, we establish a framework connecting reward-free exploration and AIL, and propose an algorithm, MB-TAIL, that achieves the minimax optimal expert sample complexity of $\\widetilde{O} (H^{3/2} |S|/\\varepsilon)$ and interaction complexity of $\\widetilde{O} (H^{3} |S|^2 |A|/\\varepsilon^2)$. Here, $H$ represents the planning horizon, $|S|$ is the state space size, $|A|$ is the action space size, and $\\varepsilon$ is the desired imitation gap. MB-TAIL is the first algorithm to achieve this level of expert sample complexity in the unknown transition setting and improves upon the interaction complexity of the best-known algorithm, OAL, by $O(H)$. Additionally, we demonstrate the generalization ability of MB-TAIL by extending it to the function approximation setting and proving that it can achieve expert sample and interaction complexity independent of $|S|$"
                    },
                    {
                        "title": "Deploying Offline Reinforcement Learning with Human Feedback",
                        "abstract": "Reinforcement learning (RL) has shown promise for decision-making tasks in real-world applications. One practical framework involves training parameterized policy models from an offline dataset and subsequently deploying them in an online environment. However, this approach can be risky since the offline training may not be perfect, leading to poor performance of the RL models that may take dangerous actions. To address this issue, we propose an alternative framework that involves a human supervising the RL models and providing additional feedback in the online deployment phase. We formalize this online deployment problem and develop two approaches. The first approach uses model selection and the upper confidence bound algorithm to adaptively select a model to deploy from a candidate set of trained offline RL models. The second approach involves fine-tuning the model in the online deployment phase when a supervision signal arrives. We demonstrate the effectiveness of these approaches for robot locomotion control and traffic light control tasks through empirical validation."
                    },
                    {
                        "title": "Entropic Distribution Matching in Supervised Fine-tuning of LLMs: Less Overfitting and Better Diversity",
                        "abstract": "Large language models rely on Supervised Fine-Tuning (SFT) to specialize in downstream tasks. Cross Entropy (CE) loss is the de facto choice in SFT, but it often leads to overfitting and limited output diversity due to its aggressive updates to the data distribution. This paper aim to address these issues by introducing the maximum entropy principle, which favors models with flatter distributions that still effectively capture the data. Specifically, we develop a new distribution matching method called GEM, which solves reverse Kullback-Leibler divergence minimization with an entropy regularizer.   For the SFT of Llama-3-8B models, GEM outperforms CE in several aspects. First, when applied to the UltraFeedback dataset to develop general instruction-following abilities, GEM exhibits reduced overfitting, evidenced by lower perplexity and better performance on the IFEval benchmark. Furthermore, GEM enhances output diversity, leading to performance gains of up to 7 points on math reasoning and code generation tasks using best-of-n sampling, even without domain-specific data. Second, when fine-tuning with domain-specific datasets for math reasoning and code generation, GEM also shows less overfitting and improvements of up to 10 points compared with CE."
                    },
                    {
                        "title": "BWArea Model: Learning World Model, Inverse Dynamics, and Policy for Controllable Language Generation",
                        "abstract": "Large language models (LLMs) have catalyzed a paradigm shift in natural language processing, yet their limited controllability poses a significant challenge for downstream applications. We aim to address this by drawing inspiration from the neural mechanisms of the human brain, specifically Broca's and Wernicke's areas, which are crucial for language generation and comprehension, respectively. In particular, Broca's area receives cognitive decision signals from Wernicke's area, treating the language generation as an intricate decision-making process, which differs from the fully auto-regressive language generation of existing LLMs. In a similar vein, our proposed system, the BWArea model, conceptualizes language generation as a decision-making task. This model has three components: a language world model, an inverse dynamics model, and a cognitive policy. Like Wernicke's area, the inverse dynamics model is designed to deduce the underlying cognitive intentions, or latent actions, behind each token. The BWArea model is amenable to both pre-training and fine-tuning like existing LLMs. With 30B clean pre-training tokens, we have trained a BWArea model, which achieves competitive performance with LLMs of equal size (1B parameters). Unlike fully auto-regressive LLMs, its pre-training performance does not degenerate if dirty data unintentionally appears. This shows the advantage of a decomposed structure of BWArea model in reducing efforts in laborious data selection and labeling. Finally, we reveal that the BWArea model offers enhanced controllability via fine-tuning the cognitive policy with downstream reward metrics, thereby facilitating alignment with greater simplicity. On 9 out of 10 tasks from two suites, TextWorld and BigBench Hard, our method shows superior performance to auto-regressive LLMs."
                    }
                ]
            },
            "64c0c7b0-3257-42e1-9aef-1c9abeae5016": {
                "pk": "64c0c7b0-3257-42e1-9aef-1c9abeae5016",
                "name": "Ruoyu Sun",
                "collaborators": [
                    "Mingyi Hong",
                    "Zhi-Quan Luo",
                    "Yinyu Ye",
                    "Dawei Li",
                    "Tiantian Fang",
                    "Alex Schwing",
                    "Xiangyi Chen",
                    "Sijia Liu",
                    "Feng Xiao",
                    "Jicong Fan"
                ],
                "domain": [
                    "Optimization",
                    "Neural Networks",
                    "Machine Learning",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Optimization for deep learning: theory and algorithms",
                        "abstract": "When and why can a neural network be successfully trained? This article provides an overview of optimization algorithms and theory for training neural networks. First, we discuss the issue of gradient explosion/vanishing and the more general issue of undesirable spectrum, and then discuss practical solutions including careful initialization and normalization methods. Second, we review generic optimization methods used in training neural networks, such as SGD, adaptive gradient methods and distributed methods, and theoretical results for these algorithms. Third, we review existing research on the global issues of neural network training, including results on bad local minima, mode connectivity, lottery ticket hypothesis and infinite-width analysis."
                    },
                    {
                        "title": "Worst-case Complexity of Cyclic Coordinate Descent: $O(n^2)$ Gap with Randomized Version",
                        "abstract": "This paper concerns the worst-case complexity of cyclic coordinate descent (C-CD) for minimizing a convex quadratic function, which is equivalent to Gauss-Seidel method and can be transformed to Kaczmarz method and projection onto convex sets (POCS). We observe that the known provable complexity of C-CD can be $O(n^2)$ times slower than randomized coordinate descent (R-CD), but no example was rigorously proven to exhibit such a large gap. In this paper we show that the gap indeed exists. We prove that there exists an example for which C-CD takes at least $O(n^4 \\kappa_{\\text{CD}} \\log\\frac{1}{\\epsilon})$ operations, where $\\kappa_{\\text{CD}}$ is related to Demmel's condition number and it determines the convergence rate of R-CD. It implies that in the worst case C-CD can indeed be $O(n^2)$ times slower than R-CD, which has complexity $O( n^2 \\kappa_{\\text{CD}} \\log\\frac{1}{\\epsilon})$. Note that for this example, the gap exists for any fixed update order, not just a particular order. Based on the example, we establish several almost tight complexity bounds of C-CD for quadratic problems. One difficulty with the analysis is that the spectral radius of a non-symmetric iteration matrix does not necessarily constitute a \\textit{lower bound} for the convergence rate.   An immediate consequence is that for Gauss-Seidel method, Kaczmarz method and POCS, there is also an $O(n^2) $ gap between the cyclic versions and randomized versions (for solving linear systems). We also show that the classical convergence rate of POCS by Smith, Solmon and Wager [1] is always worse and sometimes can be infinitely times worse than our bound."
                    },
                    {
                        "title": "On a Faster $R$-Linear Convergence Rate of the Barzilai-Borwein Method",
                        "abstract": "The Barzilai-Borwein (BB) method has demonstrated great empirical success in nonlinear optimization. However, the convergence speed of BB method is not well understood, as the known convergence rate of BB method for quadratic problems is much worse than the steepest descent (SD) method. Therefore, there is a large discrepancy between theory and practice. To shrink this gap, we prove that the BB method converges $R$-linearly at a rate of $1-1/\\kappa$, where $\\kappa$ is the condition number, for strongly convex quadratic problems. In addition, an example with the theoretical rate of convergence is constructed, indicating the tightness of our bound."
                    },
                    {
                        "title": "Improved Iteration Complexity Bounds of Cyclic Block Coordinate Descent for Convex Problems",
                        "abstract": "The iteration complexity of the block-coordinate descent (BCD) type algorithm has been under extensive investigation. It was recently shown that for convex problems the classical cyclic BCGD (block coordinate gradient descent) achieves an $\\mathcal{O}(1/r)$ complexity ($r$ is the number of passes of all blocks). However, such bounds are at least linearly depend on $K$ (the number of variable blocks), and are at least $K$ times worse than those of the gradient descent (GD) and proximal gradient (PG) methods. In this paper, we aim to close such theoretical performance gap between cyclic BCD and GD/PG. First we show that for a family of quadratic nonsmooth problems, the complexity bounds for cyclic Block Coordinate Proximal Gradient (BCPG), a popular variant of BCD, can match those of the GD/PG in terms of dependency on $K$ (up to a $\\log^2(K)$ factor). For the same family of problems, we also improve the bounds of the classical BCD (with exact block minimization) by an order of $K$. Second, we establish an improved complexity bound of Coordinate Gradient Descent (CGD) for general convex problems which can match that of GD in certain scenarios. Our bounds are sharper than the known bounds as they are always at least $K$ times worse than GD. Our analyses do not depend on the update order of block variables inside each cycle, thus our results also apply to BCD methods with random permutation (random sampling without replacement, another popular variant)."
                    },
                    {
                        "title": "On the Efficiency of Random Permutation for ADMM and Coordinate Descent",
                        "abstract": "Random permutation is observed to be powerful for optimization algorithms: for multi-block ADMM (alternating direction method of multipliers), while the classical cyclic version divergence, the randomly permuted version converges in practice; for BCD (block coordinate descent), the randomly permuted version is typically faster than other versions. In this paper, we provide strong theoretical evidence that random permutation has positive effects on ADMM and BCD, by analyzing randomly permuted ADMM (RP-ADMM) for solving linear systems of equations, and randomly permuted BCD (RP-BCD) for solving unconstrained quadratic problems. First, we prove that RP-ADMM converges in expectation for solving systems of linear equations. The key technical result is that the spectrum of the expected update matrix of RP-BCD lies in $(-1/3, 1)$, instead of the typical range $(-1, 1)$. Second, we establish expected convergence rates of RP-ADMM for solving linear sytems and RP-BCD for solving unconstrained quadratic problems. This expected rate of RP-BCD is $O(n)$ times better than the worst-case rate of cyclic BCD, thus establishing a gap of at least $O(n)$ between RP-BCD and cyclic BCD. To analyze RP-BCD, we propose a conjecture of a new matrix AM-GM (algebraic mean-geometric mean) inequality, and prove a weaker version of it."
                    },
                    {
                        "title": "Joint Downlink Base Station Association and Power Control for Max-Min Fairness: Computation and Complexity",
                        "abstract": "In a heterogeneous network (HetNet) with a large number of low power base stations (BSs), proper user-BS association and power control is crucial to achieving desirable system performance. In this paper, we systematically study the joint BS association and power allocation problem for a downlink cellular network under the max-min fairness criterion. First, we show that this problem is NP-hard. Second, we show that the upper bound of the optimal value can be easily computed, and propose a two-stage algorithm to find a high-quality suboptimal solution. Simulation results show that the proposed algorithm is near-optimal in the high-SNR regime. Third, we show that the problem under some additional mild assumptions can be solved to global optima in polynomial time by a semi-distributed algorithm. This result is based on a transformation of the original problem to an assignment problem with gains $\\log(g_{ij})$, where $\\{g_{ij}\\}$ are the channel gains."
                    },
                    {
                        "title": "On the Convergence of A Class of Adam-Type Algorithms for Non-Convex Optimization",
                        "abstract": "This paper studies a class of adaptive gradient based momentum algorithms that update the search directions and learning rates simultaneously using past gradients. This class, which we refer to as the \"Adam-type\", includes the popular algorithms such as the Adam, AMSGrad and AdaGrad. Despite their popularity in training deep neural networks, the convergence of these algorithms for solving nonconvex problems remains an open question. This paper provides a set of mild sufficient conditions that guarantee the convergence for the Adam-type methods. We prove that under our derived conditions, these methods can achieve the convergence rate of order $O(\\log{T}/\\sqrt{T})$ for nonconvex stochastic optimization. We show the conditions are essential in the sense that violating them may make the algorithm diverge. Moreover, we propose and analyze a class of (deterministic) incremental adaptive gradient algorithms, which has the same $O(\\log{T}/\\sqrt{T})$ convergence rate. Our study could also be extended to a broader class of adaptive gradient methods in machine learning and optimization."
                    },
                    {
                        "title": "Towards a Better Global Loss Landscape of GANs",
                        "abstract": "Understanding of GAN training is still very limited. One major challenge is its non-convex-non-concave min-max objective, which may lead to sub-optimal local minima. In this work, we perform a global landscape analysis of the empirical loss of GANs. We prove that a class of separable-GAN, including the original JS-GAN, has exponentially many bad basins which are perceived as mode-collapse. We also study the relativistic pairing GAN (RpGAN) loss which couples the generated samples and the true samples. We prove that RpGAN has no bad basins. Experiments on synthetic data show that the predicted bad basin can indeed appear in training. We also perform experiments to support our theory that RpGAN has a better landscape than separable-GAN. For instance, we empirically show that RpGAN performs better than separable-GAN with relatively narrow neural nets. The code is available at https://github.com/AilsaF/RS-GAN."
                    },
                    {
                        "title": "DigGAN: Discriminator gradIent Gap Regularization for GAN Training with Limited Data",
                        "abstract": "Generative adversarial nets (GANs) have been remarkably successful at learning to sample from distributions specified by a given dataset, particularly if the given dataset is reasonably large compared to its dimensionality. However, given limited data, classical GANs have struggled, and strategies like output-regularization, data-augmentation, use of pre-trained models and pruning have been shown to lead to improvements. Notably, the applicability of these strategies is 1) often constrained to particular settings, e.g., availability of a pretrained GAN; or 2) increases training time, e.g., when using pruning. In contrast, we propose a Discriminator gradIent Gap regularized GAN (DigGAN) formulation which can be added to any existing GAN. DigGAN augments existing GANs by encouraging to narrow the gap between the norm of the gradient of a discriminator's prediction w.r.t.\\ real images and w.r.t.\\ the generated samples. We observe this formulation to avoid bad attractors within the GAN loss landscape, and we find DigGAN to significantly improve the results of GAN training when limited data is available. Code is available at \\url{https://github.com/AilsaF/DigGAN}."
                    },
                    {
                        "title": "Restricted Generative Projection for One-Class Classification and Anomaly Detection",
                        "abstract": "We present a simple framework for one-class classification and anomaly detection. The core idea is to learn a mapping to transform the unknown distribution of training (normal) data to a known target distribution. Crucially, the target distribution should be sufficiently simple, compact, and informative. The simplicity is to ensure that we can sample from the distribution easily, the compactness is to ensure that the decision boundary between normal data and abnormal data is clear and reliable, and the informativeness is to ensure that the transformed data preserve the important information of the original data. Therefore, we propose to use truncated Gaussian, uniform in hypersphere, uniform on hypersphere, or uniform between hyperspheres, as the target distribution. We then minimize the distance between the transformed data distribution and the target distribution while keeping the reconstruction error for the original data small enough. Comparative studies on multiple benchmark datasets verify the effectiveness of our methods in comparison to baselines."
                    },
                    {
                        "title": "Interference alignment using finite and dependent channel extensions: the single beam case",
                        "abstract": "Vector space interference alignment (IA) is known to achieve high degrees of freedom (DoF) with infinite independent channel extensions, but its performance is largely unknown for a finite number of possibly dependent channel extensions. In this paper, we consider a $K$-user $M_t \\times M_r$ MIMO interference channel (IC) with arbitrary number of channel extensions $T$ and arbitrary channel diversity order $L$ (i.e., each channel matrix is a generic linear combination of $L$ fixed basis matrices). We study the maximum DoF achievable via vector space IA in the single beam case (i.e. each user sends one data stream). We prove that the total number of users $K$ that can communicate interference-free using linear transceivers is upper bounded by $NL+N^2/4$, where $N = \\min\\{M_tT, M_rT \\}$. An immediate consequence of this upper bound is that for a SISO IC the DoF in the single beam case is no more than $\\min\\left\\{\\sqrt{ 5K/4}, L + T/4\\right\\}$. When the channel extensions are independent, i.e. $ L$ achieves the maximum $M_r M_t T $, we show that this maximum DoF lies in $[M_r+M_t-1, M_r+M_t]$ regardless of $T$. Unlike the well-studied constant MIMO IC case, the main difficulty is how to deal with a hybrid system of equations (zero-forcing condition) and inequalities (full rank condition). Our approach combines algebraic tools that deal with equations with an induction analysis that indirectly considers the inequalities."
                    },
                    {
                        "title": "On the Benefit of Width for Neural Networks: Disappearance of Bad Basins",
                        "abstract": "Wide networks are often believed to have a nice optimization landscape, but what rigorous results can we prove? To understand the benefit of width, it is important to identify the difference between wide and narrow networks. In this work, we prove that from narrow to wide networks, there is a phase transition from having sub-optimal basins to no sub-optimal basins. Specifically, we prove two results: on the positive side, for any continuous activation functions, the loss surface of a class of wide networks has no sub-optimal basins, where \"basin\" is defined as the set-wise strict local minimum; on the negative side, for a large class of networks with width below a threshold, we construct strict local minima that are not global. These two results together show the phase transition from narrow to wide networks."
                    }
                ]
            },
            "a5cb2496-4203-41ca-ba7d-1b59c52c7e0c": {
                "pk": "a5cb2496-4203-41ca-ba7d-1b59c52c7e0c",
                "name": "Zhi-Quan Luo",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2310.13102": {
        "paper_data": {
            "title": "Particle Guidance: non-I.I.D. Diverse Sampling with Diffusion Models",
            "url": "http://arxiv.org/abs/2310.13102v2",
            "arxiv_id": "2310.13102",
            "authors": [
                "Gabriele Corso",
                "Yilun Xu",
                "Valentin de Bortoli",
                "Regina Barzilay",
                "Tommi Jaakkola"
            ],
            "abstract": "In light of the widespread success of generative models, a significant amount of research has gone into speeding up their sampling time. However, generative models are often sampled multiple times to obtain a diverse set incurring a cost that is orthogonal to sampling time. We tackle the question of how to improve diversity and sample efficiency by moving beyond the common assumption of independent samples. We propose particle guidance, an extension of diffusion-based generative sampling where a joint-particle time-evolving potential enforces diversity. We analyze theoretically the joint distribution that particle guidance generates, how to learn a potential that achieves optimal diversity, and the connections with methods in other disciplines. Empirically, we test the framework both in the setting of conditional image generation, where we are able to increase diversity without affecting quality, and molecular conformer generation, where we reduce the state-of-the-art median error by 13% on average.",
            "introduction": "   1 Introduction  Deep generative modeling has become pervasive in many computational tasks across computer vision, natural language processing, physical sciences, and beyond. In many applications, these models are used to take a number of representative samples of some distribution of interest like Van Gogh\u2019s style paintings or the 3D conformers of a small molecule. Although independent samples drawn from a distribution will perfectly represent it in the limit of infinite samples, for a finite number, this may not be the optimal strategy. Therefore, while deep learning methods have so far largely focused on the task of taking independent identically distributed (I.I.D.) samples from some distribution, this paper examines how one can use deep generative models to take a finite number of samples that can better represent the distribution of interest.   In other fields where finite-samples approximations are critical, researchers have developed various techniques to tackle this challenge. In molecular simulations, several enhanced sampling methods, like metadynamics and replica exchange, have been proposed to sample diverse sets of low-energy structures and estimate free energies. In statistics, Stein Variational Gradient Descent (SVGD) is an iterative technique to match a distribution with a finite set of particles. However, these methods are not able to efficiently sample complex distributions like images.   Towards the goal of better finite-samples generative models, that combine the power of recent advances with sample efficiency, we propose a general framework for sampling sets of particles using diffusion models. This framework, which we call particle guidance (PG), is based on the use of a time-evolving potential to guide the inference process. We present two different strategies to instantiate this new framework: the first, fixed potential particle guidance, provides ready-to-use potentials that require no further training and have little inference overhead; the second, learned potential particle guidance, requires a training process but offers better control and theoretical guarantees.   The theoretical analysis of the framework leads us to two key results. On one hand, we obtain an expression for the joint marginal distribution of the sampled process when using any arbitrary guidance potential. On the other, we derive a simple objective one can use to train a model to learn a time-evolving potential that exactly samples from a joint distribution of interest. We show this provides optimal joint distribution given some diversity constraint and it can be adapted to the addition of further constraints such as the preservation of marginal distributions.  Further, we also demonstrate the relations of particle guidance to techniques for non-I.I.D. sampling developed in other fields and natural processes and discuss its advantages.    Empirically, we demonstrate the effectiveness of the method in both synthetic experiments and two of the most successful applications of diffusion models: text-to-image generation and molecular conformer generation. In the former, we show that particle guidance can improve the diversity of the samples generated with Stable Diffusion\u00a0[Rombach et\u00a0al., 2021] while maintaining a quality comparable to that of I.I.D. sampling. For molecular conformer generation, applied to the state-of-the-art method Torsional Diffusion [Jing et\u00a0al., 2022], particle guidance is able to simultaneously improve precision and coverage, reducing their median error by respectively 19% and 8%. In",
            "references": [
                {
                    "title": "Vendi sampling for molecular simulations: Diversity as a force for faster convergence and better exploration.",
                    "abstract": "Molecular dynamics (MD) is the method of choice for understanding the structure, function, and interactions of molecules. However, MD simulations are limited by the strong metastability of many molecules, which traps them in a single conformation basin for an extended amount of time. Enhanced sampling techniques, such as metadynamics and replica exchange, have been developed to overcome this limitation and accelerate the exploration of complex free energy landscapes. In this paper, we propose Vendi Sampling, a replica-based algorithm for increasing the efficiency and efficacy of the exploration of molecular conformation spaces. In Vendi sampling, replicas are simulated in parallel and coupled via a global statistical measure, the Vendi Score, to enhance diversity. Vendi sampling allows for the recovery of unbiased sampling statistics and dramatically improves sampling efficiency. We demonstrate the effectiveness of Vendi sampling in improving molecular dynamics simulations by showing significant improvements in coverage and mixing between metastable states and convergence of free energy estimates for four common benchmarks, including Alanine Dipeptide and Chignolin."
                },
                {
                    "title": "Restart Sampling for Improving Generative Processes",
                    "abstract": "Generative processes that involve solving differential equations, such as diffusion models, frequently necessitate balancing speed and quality. ODE-based samplers are fast but plateau in performance while SDE-based samplers deliver higher sample quality at the cost of increased sampling time. We attribute this difference to sampling errors: ODE-samplers involve smaller discretization errors while stochasticity in SDE contracts accumulated errors. Based on these findings, we propose a novel sampling algorithm called Restart in order to better balance discretization errors and contraction. The sampling method alternates between adding substantial noise in additional forward steps and strictly following a backward ODE. Empirically, Restart sampler surpasses previous SDE and ODE samplers in both speed and accuracy. Restart not only outperforms the previous best SDE results, but also accelerates the sampling speed by 10-fold / 2-fold on CIFAR-10 / ImageNet $64 \\times 64$. In addition, it attains significantly better sample quality than ODE samplers within comparable sampling times. Moreover, Restart better balances text-image alignment/visual quality versus diversity than previous samplers in the large-scale text-to-image Stable Diffusion model pre-trained on LAION $512 \\times 512$. Code is available at https://github.com/Newbeeer/diffusion_restart_sampling"
                },
                {
                    "title": "Modeling Molecular Structures with Intrinsic Diffusion Models",
                    "abstract": "Since its foundations, more than one hundred years ago, the field of structural biology has strived to understand and analyze the properties of molecules and their interactions by studying the structure that they take in 3D space. However, a fundamental challenge with this approach has been the dynamic nature of these particles, which forces us to model not a single but a whole distribution of structures for every molecular system. This thesis proposes Intrinsic Diffusion Modeling, a novel approach to this problem based on combining diffusion generative models with scientific knowledge about the flexibility of biological complexes. The knowledge of these degrees of freedom is translated into the definition of a manifold over which the diffusion process is defined. This manifold significantly reduces the dimensionality and increases the smoothness of the generation space allowing for significantly faster and more accurate generative processes. We demonstrate the effectiveness of this approach on two fundamental tasks at the basis of computational chemistry and biology: molecular conformer generation and molecular docking. In both tasks, we construct the first deep learning method to outperform traditional computational approaches achieving an unprecedented level of accuracy for scalable programs."
                },
                {
                    "title": "Reduce, Reuse, Recycle: Compositional Generation with Energy-Based Diffusion Models and MCMC",
                    "abstract": "Since their introduction, diffusion models have quickly become the prevailing approach to generative modeling in many domains. They can be interpreted as learning the gradients of a time-varying sequence of log-probability density functions. This interpretation has motivated classifier-based and classifier-free guidance as methods for post-hoc control of diffusion models. In this work, we build upon these ideas using the score-based interpretation of diffusion models, and explore alternative ways to condition, modify, and reuse diffusion models for tasks involving compositional generation and guidance. In particular, we investigate why certain types of composition fail using current techniques and present a number of solutions. We conclude that the sampler (not the model) is responsible for this failure and propose new samplers, inspired by MCMC, which enable successful compositional generation. Further, we propose an energy-based parameterization of diffusion models which enables the use of new compositional operators and more sophisticated, Metropolis-corrected samplers. Intriguingly we find these samplers lead to notable improvements in compositional generation across a wide set of problems such as classifier-guided ImageNet modeling and compositional text-to-image generation."
                },
                {
                    "title": "PFGM++: Unlocking the Potential of Physics-Inspired Generative Models",
                    "abstract": "We introduce a new family of physics-inspired generative models termed PFGM++ that unifies diffusion models and Poisson Flow Generative Models (PFGM). These models realize generative trajectories for $N$ dimensional data by embedding paths in $N{+}D$ dimensional space while still controlling the progression with a simple scalar norm of the $D$ additional variables. The new models reduce to PFGM when $D{=}1$ and to diffusion models when $D{\\to}\\infty$. The flexibility of choosing $D$ allows us to trade off robustness against rigidity as increasing $D$ results in more concentrated coupling between the data and the additional variable norms. We dispense with the biased large batch field targets used in PFGM and instead provide an unbiased perturbation-based objective similar to diffusion models. To explore different choices of $D$, we provide a direct alignment method for transferring well-tuned hyperparameters from diffusion models ($D{\\to} \\infty$) to any finite $D$ values. Our experiments show that models with finite $D$ can be superior to previous state-of-the-art diffusion models on CIFAR-10/FFHQ $64{\\times}64$ datasets, with FID scores of $1.91/2.43$ when $D{=}2048/128$. In class-conditional setting, $D{=}2048$ yields current state-of-the-art FID of $1.74$ on CIFAR-10. In addition, we demonstrate that models with smaller $D$ exhibit improved robustness against modeling errors. Code is available at https://github.com/Newbeeer/pfgmpp"
                },
                {
                    "title": "SE(3) diffusion model with application to protein backbone generation",
                    "abstract": "The design of novel protein structures remains a challenge in protein engineering for applications across biomedicine and chemistry. In this line of work, a diffusion model over rigid bodies in 3D (referred to as frames) has shown success in generating novel, functional protein backbones that have not been observed in nature. However, there exists no principled methodological framework for diffusion on SE(3), the space of orientation preserving rigid motions in R3, that operates on frames and confers the group invariance. We address these shortcomings by developing theoretical foundations of SE(3) invariant diffusion models on multiple frames followed by a novel framework, FrameDiff, for learning the SE(3) equivariant score over multiple frames. We apply FrameDiff on monomer backbone generation and find it can generate designable monomers up to 500 amino acids without relying on a pretrained protein structure prediction network that has been integral to previous methods. We find our samples are capable of generalizing beyond any known protein structure."
                },
                {
                    "title": "Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Models",
                    "abstract": "Cutting-edge diffusion models produce images with high quality and customizability, enabling them to be used for commercial art and graphic design purposes. But do diffusion models create unique works of art, or are they replicating content directly from their training sets? In this work, we study image retrieval frameworks that enable us to compare generated images with training samples and detect when content has been replicated. Applying our frameworks to diffusion models trained on multiple datasets including Oxford flowers, Celeb-A, ImageNet, and LAION, we discuss how factors such as training set size impact rates of content replication. We also identify cases where diffusion models, including the popular Stable Diffusion model, blatantly copy from their training data. Project page: https://somepago.github.io/diffrep.html"
                },
                {
                    "title": "From Denoising Diffusions to Denoising Markov Models",
                    "abstract": "\n Denoising diffusions are state-of-the-art generative models exhibiting remarkable empirical performance. They work by diffusing the data distribution into a Gaussian distribution and then learning to reverse this noising process to obtain synthetic datapoints. The denoising diffusion relies on approximations of the logarithmic derivatives of the noised data densities using score matching. Such models can also be used to perform approximate posterior simulation when one can only sample from the prior and likelihood. We propose a unifying framework generalising this approach to a wide class of spaces and leading to an original extension of score matching. We illustrate the resulting models on various applications."
                },
                {
                    "title": "LAION-5B: An open large-scale dataset for training next generation image-text models",
                    "abstract": "Groundbreaking language-vision architectures like CLIP and DALL-E proved the utility of training on large amounts of noisy image-text data, without relying on expensive accurate labels used in standard vision unimodal supervised learning. The resulting models showed capabilities of strong text-guided image generation and transfer to downstream tasks, while performing remarkably at zero-shot classification with noteworthy out-of-distribution robustness. Since then, large-scale language-vision models like ALIGN, BASIC, GLIDE, Flamingo and Imagen made further improvements. Studying the training and capabilities of such models requires datasets containing billions of image-text pairs. Until now, no datasets of this size have been made openly available for the broader research community. To address this problem and democratize research on large-scale multi-modal models, we present LAION-5B - a dataset consisting of 5.85 billion CLIP-filtered image-text pairs, of which 2.32B contain English language. We show successful replication and fine-tuning of foundational models like CLIP, GLIDE and Stable Diffusion using the dataset, and discuss further experiments enabled with an openly available dataset of this scale. Additionally we provide several nearest neighbor indices, an improved web-interface for dataset exploration and subset generation, and detection scores for watermark, NSFW, and toxic content detection. Announcement page https://laion.ai/laion-5b-a-new-era-of-open-large-scale-multi-modal-datasets/"
                },
                {
                    "title": "DiffDock: Diffusion Steps, Twists, and Turns for Molecular Docking",
                    "abstract": "Predicting the binding structure of a small molecule ligand to a protein -- a task known as molecular docking -- is critical to drug design. Recent deep learning methods that treat docking as a regression problem have decreased runtime compared to traditional search-based methods but have yet to offer substantial improvements in accuracy. We instead frame molecular docking as a generative modeling problem and develop DiffDock, a diffusion generative model over the non-Euclidean manifold of ligand poses. To do so, we map this manifold to the product space of the degrees of freedom (translational, rotational, and torsional) involved in docking and develop an efficient diffusion process on this space. Empirically, DiffDock obtains a 38% top-1 success rate (RMSD<2A) on PDBBind, significantly outperforming the previous state-of-the-art of traditional docking (23%) and deep learning (20%) methods. Moreover, while previous methods are not able to dock on computationally folded structures (maximum accuracy 10.4%), DiffDock maintains significantly higher precision (21.7%). Finally, DiffDock has fast inference times and provides confidence estimates with high selective accuracy."
                },
                {
                    "title": "Sampling is as easy as learning the score: theory for diffusion models with minimal data assumptions",
                    "abstract": "We provide theoretical convergence guarantees for score-based generative models (SGMs) such as denoising diffusion probabilistic models (DDPMs), which constitute the backbone of large-scale real-world generative models such as DALL$\\cdot$E 2. Our main result is that, assuming accurate score estimates, such SGMs can efficiently sample from essentially any realistic data distribution. In contrast to prior works, our results (1) hold for an $L^2$-accurate score estimate (rather than $L^\\infty$-accurate); (2) do not require restrictive functional inequality conditions that preclude substantial non-log-concavity; (3) scale polynomially in all relevant problem parameters; and (4) match state-of-the-art complexity guarantees for discretization of the Langevin diffusion, provided that the score error is sufficiently small. We view this as strong theoretical justification for the empirical success of SGMs. We also examine SGMs based on the critically damped Langevin diffusion (CLD). Contrary to conventional wisdom, we provide evidence that the use of the CLD does not reduce the complexity of SGMs."
                },
                {
                    "title": "Poisson Flow Generative Models",
                    "abstract": "We propose a new\"Poisson flow\"generative model (PFGM) that maps a uniform distribution on a high-dimensional hemisphere into any data distribution. We interpret the data points as electrical charges on the $z=0$ hyperplane in a space augmented with an additional dimension $z$, generating a high-dimensional electric field (the gradient of the solution to Poisson equation). We prove that if these charges flow upward along electric field lines, their initial distribution in the $z=0$ plane transforms into a distribution on the hemisphere of radius $r$ that becomes uniform in the $r \\to\\infty$ limit. To learn the bijective transformation, we estimate the normalized field in the augmented space. For sampling, we devise a backward ODE that is anchored by the physically meaningful additional dimension: the samples hit the unaugmented data manifold when the $z$ reaches zero. Experimentally, PFGM achieves current state-of-the-art performance among the normalizing flow models on CIFAR-10, with an Inception score of $9.68$ and a FID score of $2.35$. It also performs on par with the state-of-the-art SDE approaches while offering $10\\times $ to $20 \\times$ acceleration on image generation tasks. Additionally, PFGM appears more tolerant of estimation errors on a weaker network architecture and robust to the step size in the Euler method. The code is available at https://github.com/Newbeeer/poisson_flow ."
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "Torsional Diffusion for Molecular Conformer Generation",
                    "abstract": "Molecular conformer generation is a fundamental task in computational chemistry. Several machine learning approaches have been developed, but none have outperformed state-of-the-art cheminformatics methods. We propose torsional diffusion, a novel diffusion framework that operates on the space of torsion angles via a diffusion process on the hypertorus and an extrinsic-to-intrinsic score model. On a standard benchmark of drug-like molecules, torsional diffusion generates superior conformer ensembles compared to machine learning and cheminformatics methods in terms of both RMSD and chemical properties, and is orders of magnitude faster than previous diffusion-based models. Moreover, our model provides exact likelihoods, which we employ to build the first generalizable Boltzmann generator. Code is available at https://github.com/gcorso/torsional-diffusion."
                },
                {
                    "title": "Elucidating the Design Space of Diffusion-Based Generative Models",
                    "abstract": "We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36."
                },
                {
                    "title": "Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding",
                    "abstract": "We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding. Imagen builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation. Our key discovery is that generic large language models (e.g. T5), pretrained on text-only corpora, are surprisingly effective at encoding text for image synthesis: increasing the size of the language model in Imagen boosts both sample fidelity and image-text alignment much more than increasing the size of the image diffusion model. Imagen achieves a new state-of-the-art FID score of 7.27 on the COCO dataset, without ever training on COCO, and human raters find Imagen samples to be on par with the COCO data itself in image-text alignment. To assess text-to-image models in greater depth, we introduce DrawBench, a comprehensive and challenging benchmark for text-to-image models. With DrawBench, we compare Imagen with recent methods including VQ-GAN+CLIP, Latent Diffusion Models, and DALL-E 2, and find that human raters prefer Imagen over other models in side-by-side comparisons, both in terms of sample quality and image-text alignment. See https://imagen.research.google/ for an overview of the results."
                },
                {
                    "title": "GeoDiff: a Geometric Diffusion Model for Molecular Conformation Generation",
                    "abstract": "Predicting molecular conformations from molecular graphs is a fundamental problem in cheminformatics and drug discovery. Recently, significant progress has been achieved with machine learning approaches, especially with deep generative models. Inspired by the diffusion process in classical non-equilibrium thermodynamics where heated particles will diffuse from original states to a noise distribution, in this paper, we propose a novel generative model named GeoDiff for molecular conformation prediction. GeoDiff treats each atom as a particle and learns to directly reverse the diffusion process (i.e., transforming from a noise distribution to stable conformations) as a Markov chain. Modeling such a generation process is however very challenging as the likelihood of conformations should be roto-translational invariant. We theoretically show that Markov chains evolving with equivariant Markov kernels can induce an invariant distribution by design, and further propose building blocks for the Markov kernels to preserve the desirable equivariance property. The whole framework can be efficiently trained in an end-to-end fashion by optimizing a weighted variational lower bound to the (conditional) likelihood. Experiments on multiple benchmarks show that GeoDiff is superior or comparable to existing state-of-the-art approaches, especially on large molecules."
                },
                {
                    "title": "Riemannian Score-Based Generative Modeling",
                    "abstract": "Score-based generative models (SGMs) are a powerful class of generative models that exhibit remarkable empirical performance. Score-based generative modelling (SGM) consists of a ``noising'' stage, whereby a diffusion is used to gradually add Gaussian noise to data, and a generative model, which entails a ``denoising'' process defined by approximating the time-reversal of the diffusion. Existing SGMs assume that data is supported on a Euclidean space, i.e. a manifold with flat geometry. In many domains such as robotics, geoscience or protein modelling, data is often naturally described by distributions living on Riemannian manifolds and current SGM techniques are not appropriate. We introduce here Riemannian Score-based Generative Models (RSGMs), a class of generative models extending SGMs to Riemannian manifolds. We demonstrate our approach on a variety of manifolds, and in particular with earth and climate science spherical data."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "GeoMol: Torsional Geometric Generation of Molecular 3D Conformer Ensembles",
                    "abstract": "Prediction of a molecule's 3D conformer ensemble from the molecular graph holds a key role in areas of cheminformatics and drug discovery. Existing generative models have several drawbacks including lack of modeling important molecular geometry elements (e.g. torsion angles), separate optimization stages prone to error accumulation, and the need for structure fine-tuning based on approximate classical force-fields or computationally expensive methods such as metadynamics with approximate quantum mechanics calculations at each geometry. We propose GeoMol--an end-to-end, non-autoregressive and SE(3)-invariant machine learning approach to generate distributions of low-energy molecular 3D conformers. Leveraging the power of message passing neural networks (MPNNs) to capture local and global graph information, we predict local atomic 3D structures and torsion angles, avoiding unnecessary over-parameterization of the geometric degrees of freedom (e.g. one angle per non-terminal bond). Such local predictions suffice both for the training loss computation, as well as for the full deterministic conformer assembly (at test time). We devise a non-adversarial optimal transport based loss function to promote diverse conformer generation. GeoMol predominantly outperforms popular open-source, commercial, or state-of-the-art machine learning (ML) models, while achieving significant speed-ups. We expect such differentiable 3D structure generators to significantly impact molecular modeling and related applications."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "Learning Gradient Fields for Molecular Conformation Generation",
                    "abstract": "We study a fundamental problem in computational chemistry known as molecular conformation generation, trying to predict stable 3D structures from 2D molecular graphs. Existing machine learning approaches usually first predict distances between atoms and then generate a 3D structure satisfying the distances, where noise in predicted distances may induce extra errors during 3D coordinate generation. Inspired by the traditional force field methods for molecular dynamics simulation, in this paper, we propose a novel approach called ConfGF by directly estimating the gradient fields of the log density of atomic coordinates. The estimated gradient fields allow directly generating stable conformations via Langevin dynamics. However, the problem is very challenging as the gradient fields are roto-translation equivariant. We notice that estimating the gradient fields of atomic coordinates can be translated to estimating the gradient fields of interatomic distances, and hence develop a novel algorithm based on recent score-based generative models to effectively estimate these gradients. Experimental results across multiple tasks show that ConfGF outperforms previous state-of-the-art baselines by a significant margin."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "CLIPScore: A Reference-free Evaluation Metric for Image Captioning",
                    "abstract": "Image captioning has conventionally relied on reference-based automatic evaluations, where machine captions are compared against captions written by humans. This is in contrast to the reference-free manner in which humans assess caption quality. In this paper, we report the surprising empirical finding that CLIP (Radford et al., 2021), a cross-modal model pretrained on 400M image+caption pairs from the web, can be used for robust automatic evaluation of image captioning without the need for references. Experiments spanning several corpora demonstrate that our new reference-free metric, CLIPScore, achieves the highest correlation with human judgements, outperforming existing reference-based metrics like CIDEr and SPICE. Information gain experiments demonstrate that CLIPScore, with its tight focus on image-text compatibility, is complementary to existing reference-based metrics that emphasize text-text similarities. Thus, we also present a reference-augmented version, RefCLIPScore, which achieves even higher correlation. Beyond literal description tasks, several case studies reveal domains where CLIPScore performs well (clip-art images, alt-text rating), but also where it is relatively weaker in comparison to reference-based metrics, e.g., news captions that require richer contextual knowledge."
                },
                {
                    "title": "Annealed Stein Variational Gradient Descent",
                    "abstract": "Particle based optimization algorithms have recently been developed as sampling methods that iteratively update a set of particles to approximate a target distribution. In particular Stein variational gradient descent has gained attention in the approximate inference literature for its flexibility and accuracy. We empirically explore the ability of this method to sample from multi-modal distributions and focus on two important issues: (i) the inability of the particles to escape from local modes and (ii) the inefficacy in reproducing the density of the different regions. We propose an annealing schedule to solve these issues and show, through various experiments, how this simple solution leads to significant improvements in mode coverage, without invalidating any theoretical properties of the original algorithm."
                },
                {
                    "title": "Optimal dimension dependence of the Metropolis-Adjusted Langevin Algorithm",
                    "abstract": "Conventional wisdom in the sampling literature, backed by a popular diffusion scaling limit, suggests that the mixing time of the Metropolis-Adjusted Langevin Algorithm (MALA) scales as $O(d^{1/3})$, where $d$ is the dimension. However, the diffusion scaling limit requires stringent assumptions on the target distribution and is asymptotic in nature. In contrast, the best known non-asymptotic mixing time bound for MALA on the class of log-smooth and strongly log-concave distributions is $O(d)$. In this work, we establish that the mixing time of MALA on this class of target distributions is $\\widetilde\\Theta(d^{1/2})$ under a warm start. Our upper bound proof introduces a new technique based on a projection characterization of the Metropolis adjustment which reduces the study of MALA to the well-studied discretization analysis of the Langevin SDE and bypasses direct computation of the acceptance probability."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Blindness of score-based methods to isolated components and mixing proportions",
                    "abstract": "Statistical tasks such as density estimation and approximate Bayesian inference often involve densities with unknown normalising constants. Score-based methods, including score matching, are popular techniques as they are free of normalising constants. Although these methods enjoy theoretical guarantees, a little-known fact is that they exhibit practical failure modes when the unnormalised distribution of interest has isolated components -- they cannot discover isolated components or identify the correct mixing proportions between components. We demonstrate these findings using simple distributions and present heuristic attempts to address these issues. We hope to bring the attention of theoreticians and practitioners to these issues when developing new algorithms and applications."
                },
                {
                    "title": "Kernel Stein Generative Modeling",
                    "abstract": "We are interested in gradient-based Explicit Generative Modeling where samples can be derived from iterative gradient updates based on an estimate of the score function of the data distribution. Recent advances in Stochastic Gradient Langevin Dynamics (SGLD) demonstrates impressive results with energy-based models on high-dimensional and complex data distributions. Stein Variational Gradient Descent (SVGD) is a deterministic sampling algorithm that iteratively transports a set of particles to approximate a given distribution, based on functional gradient descent that decreases the KL divergence. SVGD has promising results on several Bayesian inference applications. However, applying SVGD on high dimensional problems is still under-explored. The goal of this work is to study high dimensional inference with SVGD. We first identify key challenges in practical kernel SVGD inference in high-dimension. We propose noise conditional kernel SVGD (NCK-SVGD), that works in tandem with the recently introduced Noise Conditional Score Network estimator. NCK is crucial for successful inference with SVGD in high dimension, as it adapts the kernel to the noise level of the score estimate. As we anneal the noise, NCK-SVGD targets the real data distribution. We then extend the annealed SVGD with an entropic regularization. We show that this offers a flexible control between sample quality and diversity, and verify it empirically by precision and recall evaluations. The NCK-SVGD produces samples comparable to GANs and annealed SGLD on computer vision benchmarks, including MNIST and CIFAR-10."
                },
                {
                    "title": "GEOM: Energy-annotated molecular conformations for property prediction and molecular generation",
                    "abstract": "Machine learning outperforms traditional approaches in many molecular design tasks. Although molecules are often thought of as 2D graphs, they in fact consist of an ensemble of inter-converting 3D structures called conformers. Molecular properties arise from the contribution of many conformers, and in the case of a drug binding a target, may be due mainly to a few distinct members. Molecular representations in machine learning are typically based on either one single 3D conformer or on a 2D graph that strips geometrical information. No reference datasets exist that connect these graph and point cloud ensemble representations. Here, we use first-principles simulations to annotate over 400,000 molecules with the ensemble of geometries they span. The Geometrical Embedding Of Molecules (GEOM) dataset contains over 33 million molecular conformers labeled with their relative energies and statistical probabilities at room temperature. This dataset will assist benchmarking and transfer learning in two classes of tasks: inferring 3D properties from 2D molecular graphs, and developing generative models to sample 3D conformations."
                },
                {
                    "title": "Automated exploration of the low-energy chemical space with fast quantum chemical methods.",
                    "abstract": "We propose and discuss an efficient scheme for the in silico sampling for parts of the molecular chemical space by semiempirical tight-binding methods combined with a meta-dynamics driven search algorithm. The focus of this work is set on the generation of proper thermodynamic ensembles at a quantum chemical level for conformers, but similar procedures for protonation states, tautomerism and non-covalent complex geometries are also discussed. The conformational ensembles consisting of all significantly populated minimum energy structures normally form the basis of further, mostly DFT computational work, such as the calculation of spectra or macroscopic properties. By using basic quantum chemical methods, electronic effects or possible bond breaking/formation are accounted for and a very reasonable initial energetic ranking of the candidate structures is obtained. Due to the huge computational speedup gained by the fast low-cost quantum chemical methods, overall short computation times even for systems with hundreds of atoms (typically drug-sized molecules) are achieved. Furthermore, specialized applications, such as sampling with implicit solvation models or constrained conformational sampling for transition-states, metal-, surface-, or noncovalently bound complexes are discussed, opening many possible applications in modern computational chemistry and drug discovery. The procedures have been implemented in a freely available computer code called CREST, that makes use of the fast and reliable GFNn-xTB methods."
                },
                {
                    "title": "Stochastic Particle-Optimization Sampling and the Non-Asymptotic Convergence Theory",
                    "abstract": "Particle-optimization-based sampling (POS) is a recently developed effective sampling technique that interactively updates a set of particles. A representative algorithm is the Stein variational gradient descent (SVGD). We prove, under certain conditions, SVGD experiences a theoretical pitfall, {\\it i.e.}, particles tend to collapse. As a remedy, we generalize POS to a stochastic setting by injecting random noise into particle updates, thus yielding particle-optimization sampling (SPOS). Notably, for the first time, we develop {\\em non-asymptotic convergence theory} for the SPOS framework (related to SVGD), characterizing algorithm convergence in terms of the 1-Wasserstein distance w.r.t.\\! the numbers of particles and iterations. Somewhat surprisingly, with the same number of updates (not too large) for each particle, our theory suggests adopting more particles does not necessarily lead to a better approximation of a target distribution, due to limited computational budget and numerical errors. This phenomenon is also observed in SVGD and verified via an experiment on synthetic data. Extensive experimental results verify our theory and demonstrate the effectiveness of our proposed framework."
                },
                {
                    "title": "Message Passing Stein Variational Gradient Descent",
                    "abstract": "Stein variational gradient descent (SVGD) is a recently proposed particle-based Bayesian inference method, which has attracted a lot of interest due to its remarkable approximation ability and particle efficiency compared to traditional variational inference and Markov Chain Monte Carlo methods. However, we observed that particles of SVGD tend to collapse to modes of the target distribution, and this particle degeneracy phenomenon becomes more severe with higher dimensions. Our theoretical analysis finds out that there exists a negative correlation between the dimensionality and the repulsive force of SVGD which should be blamed for this phenomenon. We propose Message Passing SVGD (MP-SVGD) to solve this problem. By leveraging the conditional independence structure of probabilistic graphical models (PGMs), MP-SVGD converts the original high-dimensional global inference problem into a set of local ones over the Markov blanket with lower dimensions. Experimental results show its advantages of preventing vanishing repulsive force in high-dimensional space over SVGD, and its particle efficiency and approximation flexibility over other inference methods on graphical models."
                },
                {
                    "title": "Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithm",
                    "abstract": "We propose a general purpose variational inference algorithm that forms a natural counterpart of gradient descent for optimization. Our method iteratively transports a set of particles to match the target distribution, by applying a form of functional gradient descent that minimizes the KL divergence. Empirical studies are performed on various real world models and datasets, on which our method is competitive with existing state-of-the-art methods. The derivation of our method is based on a new theoretical result that connects the derivative of KL divergence under smooth transforms with Stein's identity and a recently proposed kernelized Stein discrepancy, which is of independent interest."
                },
                {
                    "title": "Bayesian ensemble refinement by replica simulations and reweighting.",
                    "abstract": "We describe different Bayesian ensemble refinement methods, examine their interrelation, and discuss their practical application. With ensemble refinement, the properties of dynamic and partially disordered (bio)molecular structures can be characterized by integrating a wide range of experimental data, including measurements of ensemble-averaged observables. We start from a Bayesian formulation in which the posterior is a functional that ranks different configuration space distributions. By maximizing this posterior, we derive an optimal Bayesian ensemble distribution. For discrete configurations, this optimal distribution is identical to that obtained by the maximum entropy \"ensemble refinement of SAXS\" (EROS) formulation. Bayesian replica ensemble refinement enhances the sampling of relevant configurations by imposing restraints on averages of observables in coupled replica molecular dynamics simulations. We show that the strength of the restraints should scale linearly with the number of replicas to ensure convergence to the optimal Bayesian result in the limit of infinitely many replicas. In the \"Bayesian inference of ensembles\" method, we combine the replica and EROS approaches to accelerate the convergence. An adaptive algorithm can be used to sample directly from the optimal ensemble, without replicas. We discuss the incorporation of single-molecule measurements and dynamic observables such as relaxation parameters. The theoretical analysis of different Bayesian ensemble refinement approaches provides a basis for practical applications and a starting point for further investigations."
                },
                {
                    "title": "Conformer Generation with OMEGA: Learning from the Data Set and the Analysis of Failures",
                    "abstract": "We recently published a high quality validation set for testing conformer generators, consisting of structures from both the PDB and the CSD (Hawkins, P. C. D. et al. J. Chem. Inf. Model. 2010, 50, 572.), and tested the performance of our conformer generator, OMEGA, on these sets. In the present publication, we focus on understanding the suitability of those data sets for validation and identifying and learning from OMEGA's failures. We compare, for the first time we are aware of, the coverage of the applicable property spaces between the validation data sets we used and the parent compound sets to determine if our data sets adequately sample these property spaces. We also introduce the concept of torsion fingerprinting and compare this method of dissimilation to the more traditional graph-centric diversification methods we used in our previous publication. To improve our ability to programmatically identify cases where the crystallographic conformation is not well reproduced computationally, we introduce a new metric to compare conformations, RMSTanimoto. This new metric is used alongside those from our previous publication to efficiently identify reproduction failures. We find RMSTanimoto to be particularly effective in identifying failures for the smallest molecules in our data sets. Analysis of the nature of these failures, particularly those for the CSD, sheds further light on the issue of strain in crystallographic structures. Some of the residual failure cases not resolved by simple changes in OMEGA's defaults present significant challenges to conformer generation engines like OMEGA and are a source of new avenues to further improve their performance, while others illustrate the pitfalls of validating against crystallographic ligand conformations, particularly those from the PDB."
                },
                {
                    "title": "Determinantal Point Processes for Machine Learning",
                    "abstract": "Determinantal point processes (DPPs) are elegant probabilistic models of repulsion that arise in quantum physics and random matrix theory. In contrast to traditional structured models like Markov random fields, which become intractable and hard to approximate in the presence of negative correlations, DPPs offer efficient and exact algorithms for sampling, marginalization, conditioning, and other inference tasks. While they have been studied extensively by mathematicians, giving rise to a deep and beautiful theory, DPPs are relatively new in machine learning. Determinantal Point Processes for Machine Learning provides a comprehensible introduction to DPPs, focusing on the intuitions, algorithms, and extensions that are most relevant to the machine learning community, and shows how DPPs can be applied to real-world applications like finding diverse sets of high-quality search results, building informative summaries by selecting diverse sentences from documents, modeling non-overlapping human poses in images or video, and automatically building timelines of important news stories. It presents the general mathematical background to DPPs along with a range of modeling extensions, efficient algorithms, and theoretical results that aim to enable practical modeling and learning."
                },
                {
                    "title": "AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading",
                    "abstract": "AutoDock Vina, a new program for molecular docking and virtual screening, is presented. AutoDock Vina achieves an approximately two orders of magnitude speed\u2010up compared with the molecular docking software previously developed in our lab (AutoDock 4), while also significantly improving the accuracy of the binding mode predictions, judging by our tests on the training set used in AutoDock 4 development. Further speed\u2010up is achieved from parallelism, by using multithreading on multicore machines. AutoDock Vina automatically calculates the grid maps and clusters the results in a way transparent to the user. \u00a9 2009 Wiley Periodicals, Inc. J Comput Chem 2010"
                },
                {
                    "title": "Well-tempered metadynamics: a smoothly converging and tunable free-energy method.",
                    "abstract": "We present a method for determining the free-energy dependence on a selected number of collective variables using an adaptive bias. The formalism provides a unified description which has metadynamics and canonical sampling as limiting cases. Convergence and errors can be rigorously and easily controlled. The parameters of the simulation can be tuned so as to focus the computational effort only on the physically relevant regions of the order parameter space. The algorithm is tested on the reconstruction of an alanine dipeptide free-energy landscape."
                },
                {
                    "title": "Escaping free-energy minima",
                    "abstract": "We introduce a powerful method for exploring the properties of the multidimensional free energy surfaces (FESs) of complex many-body systems by means of coarse-grained non-Markovian dynamics in the space defined by a few collective coordinates. A characteristic feature of these dynamics is the presence of a history-dependent potential term that, in time, fills the minima in the FES, allowing the efficient exploration and accurate determination of the FES as a function of the collective coordinates. We demonstrate the usefulness of this approach in the case of the dissociation of a NaCl molecule in water and in the study of the conformational changes of a dialanine in solution."
                },
                {
                    "title": "$I$-Divergence Geometry of Probability Distributions and Minimization Problems",
                    "abstract": "JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.. Institute of Mathematical Statistics is collaborating with JSTOR to digitize, preserve and extend access to The Annals of Probability. Some geometric properties of PD's are established, Kullback's I-divergence playing the role of squared Euclidean distance. The minimum discrimination information problem is viewed as that of projecting a PD onto a convex set of PD's and useful existence theorems for and characterizations of the minimizing PD are arrived at. A natural generalization of known iterative algorithms converging to the minimizing PD in special situations is given; even for those special cases, our convergence proof is more generally valid than those previously published. As corollaries of independent interest, generalizations of known results on the existence of PD's or nonnegative matrices of a certain form are obtained. The Lagrange multiplier technique is not used."
                },
                {
                    "title": "Brownian Motion and Stochastic Calculus",
                    "abstract": "1 Preliminaries of Measure Theory De\u0085nition 1 F P ( ) is said to be an algebra if (1) 2 F (2) A;B 2 F implies A S B 2 F (3) A 2 F implies AC 2 F . F is said to be a semialgebra or semi-ring is (1) ;? 2 F (2) A;B 2 F implies AB 2 F (3) A;B 2 F implies AnB = Pn k=1Ak for some Ak 2 F De\u0085nition 2 -Class: If A;B 2 implies AB 2 De\u0085nition 3 -Class: if (1) 2 ;(2) A;B 2 with A B implies A B 2 ; (3) An 2 ; An \" A implies A 2 De\u0085nition 4 Let L R be such that 2 L implies ; 1 2 L. L-class: (1)1 2 L (2) L is linearly closed (3) n 2 L; 0 n \" ; bounded or 2 L implies 2 L Proposition 5 (i)M being -Class & -Class impliesM is -algebra (ii) If -Class contains the -Class ; then ( )"
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively sample a finite number of representative samples from complex distributions using deep generative models?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of deep generative modeling, as it addresses the limitations of traditional I.I.D. sampling methods. By improving the efficiency and effectiveness of sampling techniques, this research could lead to significant advancements in various applications, including computer vision and molecular simulations. The findings could inspire future research to explore new sampling strategies and enhance the performance of generative models, ultimately leading to more accurate representations of complex distributions in practical applications.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of sampling from intricate distributions, particularly in high-dimensional spaces like images or molecular structures. Naive approaches may fail because they do not account for the dependencies and correlations present in the data, leading to suboptimal sample diversity and representation. Additionally, technical obstacles include the need for efficient algorithms that can handle the computational demands of training and inference, as well as the theoretical difficulties in defining and optimizing guidance potentials that can effectively direct the sampling process.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on I.I.D. sampling methods, which do not adequately address the nuances of finite-sample approximations. Existing techniques, such as metadynamics and SVGD, struggle with complex distributions like images and lack the necessary efficiency for practical applications. Barriers include the limited understanding of how to leverage time-evolving potentials for sampling and the absence of a unified framework that integrates insights from various fields. This research proposes a novel approach that combines diffusion models with particle guidance, offering a significant improvement over prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves a framework called particle guidance (PG) that utilizes time-evolving potentials to enhance the sampling process. Two strategies are presented: fixed potential particle guidance, which uses pre-defined potentials, and learned potential particle guidance, which involves training a model to optimize the potential. The expected outcomes include improved sample diversity and representation in both synthetic experiments and real-world applications, such as text-to-image generation and molecular conformer generation. Metrics for evaluation will include sample quality, precision, and coverage, with empirical results demonstrating significant improvements over existing methods."
            }
        },
        "author_data": {
            "d2555e68-8947-471a-a36d-8519c4deb5ce": {
                "pk": "d2555e68-8947-471a-a36d-8519c4deb5ce",
                "name": "Gabriele Corso",
                "collaborators": [
                    "T. Jaakkola",
                    "R. Barzilay",
                    "Hannes St\u00e4rk",
                    "Bowen Jing",
                    "P. Lio\u2019",
                    "D. Beaini",
                    "Petar Velivckovi'c",
                    "Rex Ying",
                    "J. Leskovec",
                    "Jason Yim",
                    "Yilun Xu",
                    "Arash Vahdat",
                    "Karsten Kreis",
                    "Arthur Deng",
                    "Benjamin Fry",
                    "Nicholas Polizzi",
                    "Chenyu Wang",
                    "Bonnie Berger",
                    "Mohamed Amine Ketata",
                    "Cedrik Laue",
                    "R. Mammadov",
                    "Menghua Wu",
                    "\u0106eline Marquet",
                    "Ezra Erives",
                    "Peter Pao-Huang",
                    "B. Berger",
                    "Michal P\u00e1ndy",
                    "Weikang Qiu",
                    "Pietro Lio'",
                    "Jeffrey Chang",
                    "Michal P'andy",
                    "Prudencio Tossou",
                    "Christian Dallago",
                    "Stephan Gunnemann",
                    "Luca Cavalleri",
                    "Petar Velickovic",
                    "Saro Passaro",
                    "Vincent L'etourneau",
                    "William L. Hamilton"
                ],
                "domain": [
                    "Generative AI",
                    "Computational Biology",
                    "Graph Neural Network",
                    "Protein Design"
                ],
                "publications": [
                    {
                        "title": "Diffusion models in protein structure and docking",
                        "abstract": "Generative AI is rapidly transforming the frontier of research in computational structural biology. Indeed, recent successes have substantially advanced protein design and drug discovery. One of the key methodologies underlying these advances is diffusion models (DM). Diffusion models originated in computer vision, rapidly taking over image generation and offering superior quality and performance. These models were subsequently extended and modified for uses in other areas including computational structural biology. DMs are well equipped to model high dimensional, geometric data while exploiting key strengths of deep learning. In structural biology, for example, they have achieved state\u2010of\u2010the\u2010art results on protein 3D structure generation and small molecule docking. This review covers the basics of diffusion models, associated modeling choices regarding molecular representations, generation capabilities, prevailing heuristics, as well as key limitations and forthcoming refinements. We also provide best practices around evaluation procedures to help establish rigorous benchmarking and evaluation. The review is intended to provide a fresh view into the state\u2010of\u2010the\u2010art as well as highlight its potentials and current challenges of recent generative techniques in computational structural biology."
                    },
                    {
                        "title": "DisCo-Diff: Enhancing Continuous Diffusion Models with Discrete Latents",
                        "abstract": "Diffusion models (DMs) have revolutionized generative learning. They utilize a diffusion process to encode data into a simple Gaussian distribution. However, encoding a complex, potentially multimodal data distribution into a single continuous Gaussian distribution arguably represents an unnecessarily challenging learning problem. We propose Discrete-Continuous Latent Variable Diffusion Models (DisCo-Diff) to simplify this task by introducing complementary discrete latent variables. We augment DMs with learnable discrete latents, inferred with an encoder, and train DM and encoder end-to-end. DisCo-Diff does not rely on pre-trained networks, making the framework universally applicable. The discrete latents significantly simplify learning the DM's complex noise-to-data mapping by reducing the curvature of the DM's generative ODE. An additional autoregressive transformer models the distribution of the discrete latents, a simple step because DisCo-Diff requires only few discrete variables with small codebooks. We validate DisCo-Diff on toy data, several image synthesis tasks as well as molecular docking, and find that introducing discrete latents consistently improves model performance. For example, DisCo-Diff achieves state-of-the-art FID scores on class-conditioned ImageNet-64/128 datasets with ODE sampler."
                    },
                    {
                        "title": "Deep Confident Steps to New Pockets: Strategies for Docking Generalization",
                        "abstract": "Accurate blind docking has the potential to lead to new biological breakthroughs, but for this promise to be realized, docking methods must generalize well across the proteome. Existing benchmarks, however, fail to rigorously assess generalizability. Therefore, we develop DockGen, a new benchmark based on the ligand-binding domains of proteins, and we show that existing machine learning-based docking models have very weak generalization abilities. We carefully analyze the scaling laws of ML-based docking and show that, by scaling data and model size, as well as integrating synthetic data strategies, we are able to significantly increase the generalization capacity and set new state-of-the-art performance across benchmarks. Further, we propose Confidence Bootstrapping, a new training paradigm that solely relies on the interaction between diffusion and confidence models and exploits the multi-resolution generation process of diffusion models. We demonstrate that Confidence Bootstrapping significantly improves the ability of ML-based docking methods to dock to unseen protein classes, edging closer to accurate and generalizable blind docking methods."
                    },
                    {
                        "title": "Dirichlet Flow Matching with Applications to DNA Sequence Design",
                        "abstract": "Discrete diffusion or flow models could enable faster and more controllable sequence generation than autoregressive models. We show that na\\\"ive linear flow matching on the simplex is insufficient toward this goal since it suffers from discontinuities in the training target and further pathologies. To overcome this, we develop Dirichlet flow matching on the simplex based on mixtures of Dirichlet distributions as probability paths. In this framework, we derive a connection between the mixtures' scores and the flow's vector field that allows for classifier and classifier-free guidance. Further, we provide distilled Dirichlet flow matching, which enables one-step sequence generation with minimal performance hits, resulting in $O(L)$ speedups compared to autoregressive models. On complex DNA sequence generation tasks, we demonstrate superior performance compared to all baselines in distributional metrics and in achieving desired design targets for generated sequences. Finally, we show that our classifier-free guidance approach improves unconditional generation and is effective for generating DNA that satisfies design targets. Code is available at https://github.com/HannesStark/dirichlet-flow-matching."
                    },
                    {
                        "title": "DiffDock-PP: Rigid Protein-Protein Docking with Diffusion Models",
                        "abstract": "Understanding how proteins structurally interact is crucial to modern biology, with applications in drug discovery and protein design. Recent machine learning methods have formulated protein-small molecule docking as a generative problem with significant performance boosts over both traditional and deep learning baselines. In this work, we propose a similar approach for rigid protein-protein docking: DiffDock-PP is a diffusion generative model that learns to translate and rotate unbound protein structures into their bound conformations. We achieve state-of-the-art performance on DIPS with a median C-RMSD of 4.85, outperforming all considered baselines. Additionally, DiffDock-PP is faster than all search-based methods and generates reliable confidence estimates for its predictions. Our code is publicly available at $\\texttt{https://github.com/ketatam/DiffDock-PP}$"
                    },
                    {
                        "title": "EigenFold: Generative Protein Structure Prediction with Diffusion Models",
                        "abstract": "Protein structure prediction has reached revolutionary levels of accuracy on single structures, yet distributional modeling paradigms are needed to capture the conformational ensembles and flexibility that underlie biological function. Towards this goal, we develop EigenFold, a diffusion generative modeling framework for sampling a distribution of structures from a given protein sequence. We define a diffusion process that models the structure as a system of harmonic oscillators and which naturally induces a cascading-resolution generative process along the eigenmodes of the system. On recent CAMEO targets, EigenFold achieves a median TMScore of 0.84, while providing a more comprehensive picture of model uncertainty via the ensemble of sampled structures relative to existing methods. We then assess EigenFold's ability to model and predict conformational heterogeneity for fold-switching proteins and ligand-induced conformational change. Code is available at https://github.com/bjing2016/EigenFold."
                    },
                    {
                        "title": "Modeling Molecular Structures with Intrinsic Diffusion Models",
                        "abstract": "Since its foundations, more than one hundred years ago, the field of structural biology has strived to understand and analyze the properties of molecules and their interactions by studying the structure that they take in 3D space. However, a fundamental challenge with this approach has been the dynamic nature of these particles, which forces us to model not a single but a whole distribution of structures for every molecular system. This thesis proposes Intrinsic Diffusion Modeling, a novel approach to this problem based on combining diffusion generative models with scientific knowledge about the flexibility of biological complexes. The knowledge of these degrees of freedom is translated into the definition of a manifold over which the diffusion process is defined. This manifold significantly reduces the dimensionality and increases the smoothness of the generation space allowing for significantly faster and more accurate generative processes. We demonstrate the effectiveness of this approach on two fundamental tasks at the basis of computational chemistry and biology: molecular conformer generation and molecular docking. In both tasks, we construct the first deep learning method to outperform traditional computational approaches achieving an unprecedented level of accuracy for scalable programs."
                    },
                    {
                        "title": "Learning Graph Search Heuristics",
                        "abstract": "Searching for a path between two nodes in a graph is one of the most well-studied and fundamental problems in computer science. In numerous domains such as robotics, AI, or biology, practitioners develop search heuristics to accelerate their pathfinding algorithms. However, it is a laborious and complex process to hand-design heuristics based on the problem and the structure of a given use case. Here we present PHIL (Path Heuristic with Imitation Learning), a novel neural architecture and a training algorithm for discovering graph search and navigation heuristics from data by leveraging recent advances in imitation learning and graph representation learning. At training time, we aggregate datasets of search trajectories and ground-truth shortest path distances, which we use to train a specialized graph neural network-based heuristic function using backpropagation through steps of the pathfinding process. Our heuristic function learns graph embeddings useful for inferring node distances, runs in constant time independent of graph sizes, and can be easily incorporated in an algorithm such as A* at test time. Experiments show that PHIL reduces the number of explored nodes compared to state-of-the-art methods on benchmark datasets by 58.5\\% on average, can be directly applied in diverse graphs ranging from biological networks to road networks, and allows for fast planning in time-critical robotics domains."
                    },
                    {
                        "title": "Torsional Diffusion for Molecular Conformer Generation",
                        "abstract": "Molecular conformer generation is a fundamental task in computational chemistry. Several machine learning approaches have been developed, but none have outperformed state-of-the-art cheminformatics methods. We propose torsional diffusion, a novel diffusion framework that operates on the space of torsion angles via a diffusion process on the hypertorus and an extrinsic-to-intrinsic score model. On a standard benchmark of drug-like molecules, torsional diffusion generates superior conformer ensembles compared to machine learning and cheminformatics methods in terms of both RMSD and chemical properties, and is orders of magnitude faster than previous diffusion-based models. Moreover, our model provides exact likelihoods, which we employ to build the first generalizable Boltzmann generator. Code is available at https://github.com/gcorso/torsional-diffusion."
                    },
                    {
                        "title": "DiffDock: Diffusion Steps, Twists, and Turns for Molecular Docking",
                        "abstract": "Predicting the binding structure of a small molecule ligand to a protein -- a task known as molecular docking -- is critical to drug design. Recent deep learning methods that treat docking as a regression problem have decreased runtime compared to traditional search-based methods but have yet to offer substantial improvements in accuracy. We instead frame molecular docking as a generative modeling problem and develop DiffDock, a diffusion generative model over the non-Euclidean manifold of ligand poses. To do so, we map this manifold to the product space of the degrees of freedom (translational, rotational, and torsional) involved in docking and develop an efficient diffusion process on this space. Empirically, DiffDock obtains a 38% top-1 success rate (RMSD<2A) on PDBBind, significantly outperforming the previous state-of-the-art of traditional docking (23%) and deep learning (20%) methods. Moreover, while previous methods are not able to dock on computationally folded structures (maximum accuracy 10.4%), DiffDock maintains significantly higher precision (21.7%). Finally, DiffDock has fast inference times and provides confidence estimates with high selective accuracy."
                    },
                    {
                        "title": "Neural Distance Embeddings for Biological Sequences",
                        "abstract": "The development of data-dependent heuristics and representations for biological sequences that reflect their evolutionary distance is critical for large-scale biological research. However, popular machine learning approaches, based on continuous Euclidean spaces, have struggled with the discrete combinatorial formulation of the edit distance that models evolution and the hierarchical relationship that characterises real-world datasets. We present Neural Distance Embeddings (NeuroSEED), a general framework to embed sequences in geometric vector spaces, and illustrate the effectiveness of the hyperbolic space that captures the hierarchical structure and provides an average 22% reduction in embedding RMSE against the best competing geometry. The capacity of the framework and the significance of these improvements are then demonstrated devising supervised and unsupervised NeuroSEED approaches to multiple core tasks in bioinformatics. Benchmarked with common baselines, the proposed approaches display significant accuracy and/or runtime improvements on real-world datasets. As an example for hierarchical clustering, the proposed pretrained and from-scratch methods match the quality of competing baselines with 30x and 15x runtime reduction, respectively."
                    },
                    {
                        "title": "3D Infomax improves GNNs for Molecular Property Prediction",
                        "abstract": "Molecular property prediction is one of the fastest-growing applications of deep learning with critical real-world impacts. Including 3D molecular structure as input to learned models improves their performance for many molecular tasks. However, this information is infeasible to compute at the scale required by several real-world applications. We propose pre-training a model to reason about the geometry of molecules given only their 2D molecular graphs. Using methods from self-supervised learning, we maximize the mutual information between 3D summary vectors and the representations of a Graph Neural Network (GNN) such that they contain latent 3D information. During fine-tuning on molecules with unknown geometry, the GNN still generates implicit 3D information and can use it to improve downstream tasks. We show that 3D pre-training provides significant improvements for a wide range of properties, such as a 22% average MAE reduction on eight quantum mechanical properties. Moreover, the learned representations can be effectively transferred between datasets in different molecular spaces."
                    },
                    {
                        "title": "Principal Neighbourhood Aggregation for Graph Nets",
                        "abstract": "Graph Neural Networks (GNNs) have been shown to be effective models for different predictive tasks on graph-structured data. Recent work on their expressive power has focused on isomorphism tasks and countable feature spaces. We extend this theoretical framework to include continuous features - which occur regularly in real-world input domains and within the hidden layers of GNNs - and we demonstrate the requirement for multiple aggregation functions in this context. Accordingly, we propose Principal Neighbourhood Aggregation (PNA), a novel architecture combining multiple aggregators with degree-scalers (which generalize the sum aggregator). Finally, we compare the capacity of different models to capture and exploit the graph structure via a novel benchmark containing multiple tasks taken from classical graph theory, alongside existing benchmarks from real-world domains, all of which demonstrate the strength of our model. With this work, we hope to steer some of the GNN research towards new aggregation methods which we believe are essential in the search for powerful and robust models."
                    },
                    {
                        "title": "Directional Graph Networks",
                        "abstract": "In order to overcome the expressive limitations of graph neural networks (GNNs), we propose the first method that exploits vector flows over graphs to develop globally consistent directional and asymmetric aggregation functions. We show that our directional graph networks (DGNs) generalize convolutional neural networks (CNNs) when applied on a grid. Whereas recent theoretical works focus on understanding local neighbourhoods, local structures and local isomorphism with no global information flow, our novel theoretical framework allows directional convolutional kernels in any graph. First, by defining a vector field in the graph, we develop a method of applying directional derivatives and smoothing by projecting node-specific messages into the field. Then we propose the use of the Laplacian eigenvectors as such vector field, and we show that the method generalizes CNNs on an n-dimensional grid. Finally, we bring the power of CNN data augmentation to graphs by providing a means of doing reflection, rotation and distortion on the underlying directional field. We evaluate our method on different standard benchmarks and see a relative error reduction of 8% on the CIFAR10 graph dataset and 11% to 32% on the molecular ZINC dataset. An important outcome of this work is that it enables to translate any physical or biological problems with intrinsic directional axes into a graph network formalism with an embedded directional field."
                    }
                ]
            },
            "136d43c1-09d7-44b4-ac41-6fad8cf1645a": {
                "pk": "136d43c1-09d7-44b4-ac41-6fad8cf1645a",
                "name": "Yilun Xu",
                "collaborators": [
                    "T. Jaakkola",
                    "Ziming Liu",
                    "Peng Cao",
                    "Yuqing Kong",
                    "Yizhou Wang",
                    "Yonglong Tian",
                    "Shangyuan Tong",
                    "M. Tegmark",
                    "Stefano Ermon",
                    "Gabriele Corso",
                    "Arash Vahdat",
                    "Karsten Kreis",
                    "P. Holderrieth",
                    "Sulin Liu",
                    "Juno Nam",
                    "Andrew Campbell",
                    "Hannes St\u00e4rk",
                    "Rafael G'omez-Bombarelli",
                    "Mingyang Deng",
                    "Xiang Cheng",
                    "Max Tegmark",
                    "Di Luo",
                    "Hanqun Cao",
                    "Cheng Tan",
                    "Zhangyang Gao",
                    "Guangyong Chen",
                    "P. Heng",
                    "Stan Z. Li",
                    "Hao He",
                    "T. Shen",
                    "Yang Song",
                    "Sahaj Garg",
                    "Linyuan Gong",
                    "Rui Shu",
                    "Aditya Grover",
                    "Dinghuai Zhang",
                    "Kartik Ahuja",
                    "Yisen Wang",
                    "Aaron C. Courville",
                    "Shengjia Zhao",
                    "Jiaming Song",
                    "Russell Stewart"
                ],
                "domain": [
                    "Generative Models",
                    "Diffusion Models",
                    "Machine Learning",
                    "Information Theory"
                ],
                "publications": [
                    {
                        "title": "DisCo-Diff: Enhancing Continuous Diffusion Models with Discrete Latents",
                        "abstract": "Diffusion models (DMs) have revolutionized generative learning. They utilize a diffusion process to encode data into a simple Gaussian distribution. However, encoding a complex, potentially multimodal data distribution into a single continuous Gaussian distribution arguably represents an unnecessarily challenging learning problem. We propose Discrete-Continuous Latent Variable Diffusion Models (DisCo-Diff) to simplify this task by introducing complementary discrete latent variables. We augment DMs with learnable discrete latents, inferred with an encoder, and train DM and encoder end-to-end. DisCo-Diff does not rely on pre-trained networks, making the framework universally applicable. The discrete latents significantly simplify learning the DM's complex noise-to-data mapping by reducing the curvature of the DM's generative ODE. An additional autoregressive transformer models the distribution of the discrete latents, a simple step because DisCo-Diff requires only few discrete variables with small codebooks. We validate DisCo-Diff on toy data, several image synthesis tasks as well as molecular docking, and find that introducing discrete latents consistently improves model performance. For example, DisCo-Diff achieves state-of-the-art FID scores on class-conditioned ImageNet-64/128 datasets with ODE sampler."
                    },
                    {
                        "title": "Hamiltonian Score Matching and Generative Flows",
                        "abstract": "Classical Hamiltonian mechanics has been widely used in machine learning in the form of Hamiltonian Monte Carlo for applications with predetermined force fields. In this work, we explore the potential of deliberately designing force fields for Hamiltonian ODEs, introducing Hamiltonian velocity predictors (HVPs) as a tool for score matching and generative models. We present two innovations constructed with HVPs: Hamiltonian Score Matching (HSM), which estimates score functions by augmenting data via Hamiltonian trajectories, and Hamiltonian Generative Flows (HGFs), a novel generative model that encompasses diffusion models and flow matching as HGFs with zero force fields. We showcase the extended design space of force fields by introducing Oscillation HGFs, a generative model inspired by harmonic oscillators. Our experiments validate our theoretical insights about HSM as a novel score matching metric and demonstrate that HGFs rival leading generative modeling techniques."
                    },
                    {
                        "title": "Think While You Generate: Discrete Diffusion with Planned Denoising",
                        "abstract": "Discrete diffusion has achieved state-of-the-art performance, outperforming or approaching autoregressive models on standard benchmarks. In this work, we introduce Discrete Diffusion with Planned Denoising (DDPD), a novel framework that separates the generation process into two models: a planner and a denoiser. At inference time, the planner selects which positions to denoise next by identifying the most corrupted positions in need of denoising, including both initially corrupted and those requiring additional refinement. This plan-and-denoise approach enables more efficient reconstruction during generation by iteratively identifying and denoising corruptions in the optimal order. DDPD outperforms traditional denoiser-only mask diffusion methods, achieving superior results on language modeling benchmarks such as text8, OpenWebText, and token-based generation on ImageNet $256 \\times 256$. Notably, in language modeling, DDPD significantly reduces the performance gap between diffusion-based and autoregressive methods in terms of generative perplexity. Code is available at https://github.com/liusulin/DDPD."
                    },
                    {
                        "title": "Restart Sampling for Improving Generative Processes",
                        "abstract": "Generative processes that involve solving differential equations, such as diffusion models, frequently necessitate balancing speed and quality. ODE-based samplers are fast but plateau in performance while SDE-based samplers deliver higher sample quality at the cost of increased sampling time. We attribute this difference to sampling errors: ODE-samplers involve smaller discretization errors while stochasticity in SDE contracts accumulated errors. Based on these findings, we propose a novel sampling algorithm called Restart in order to better balance discretization errors and contraction. The sampling method alternates between adding substantial noise in additional forward steps and strictly following a backward ODE. Empirically, Restart sampler surpasses previous SDE and ODE samplers in both speed and accuracy. Restart not only outperforms the previous best SDE results, but also accelerates the sampling speed by 10-fold / 2-fold on CIFAR-10 / ImageNet $64 \\times 64$. In addition, it attains significantly better sample quality than ODE samplers within comparable sampling times. Moreover, Restart better balances text-image alignment/visual quality versus diversity than previous samplers in the large-scale text-to-image Stable Diffusion model pre-trained on LAION $512 \\times 512$. Code is available at https://github.com/Newbeeer/diffusion_restart_sampling"
                    },
                    {
                        "title": "Stable Target Field for Reduced Variance Score Estimation in Diffusion Models",
                        "abstract": "Diffusion models generate samples by reversing a fixed forward diffusion process. Despite already providing impressive empirical results, these diffusion models algorithms can be further improved by reducing the variance of the training targets in their denoising score-matching objective. We argue that the source of such variance lies in the handling of intermediate noise-variance scales, where multiple modes in the data affect the direction of reverse paths. We propose to remedy the problem by incorporating a reference batch which we use to calculate weighted conditional scores as more stable training targets. We show that the procedure indeed helps in the challenging intermediate regime by reducing (the trace of) the covariance of training targets. The new stable targets can be seen as trading bias for reduced variance, where the bias vanishes with increasing reference batch size. Empirically, we show that the new objective improves the image quality, stability, and training speed of various popular diffusion models across datasets with both general ODE and SDE solvers. When used in combination with EDM, our method yields a current SOTA FID of 1.90 with 35 network evaluations on the unconditional CIFAR-10 generation task. The code is available at https://github.com/Newbeeer/stf"
                    },
                    {
                        "title": "PFGM++: Unlocking the Potential of Physics-Inspired Generative Models",
                        "abstract": "We introduce a new family of physics-inspired generative models termed PFGM++ that unifies diffusion models and Poisson Flow Generative Models (PFGM). These models realize generative trajectories for $N$ dimensional data by embedding paths in $N{+}D$ dimensional space while still controlling the progression with a simple scalar norm of the $D$ additional variables. The new models reduce to PFGM when $D{=}1$ and to diffusion models when $D{\\to}\\infty$. The flexibility of choosing $D$ allows us to trade off robustness against rigidity as increasing $D$ results in more concentrated coupling between the data and the additional variable norms. We dispense with the biased large batch field targets used in PFGM and instead provide an unbiased perturbation-based objective similar to diffusion models. To explore different choices of $D$, we provide a direct alignment method for transferring well-tuned hyperparameters from diffusion models ($D{\\to} \\infty$) to any finite $D$ values. Our experiments show that models with finite $D$ can be superior to previous state-of-the-art diffusion models on CIFAR-10/FFHQ $64{\\times}64$ datasets, with FID scores of $1.91/2.43$ when $D{=}2048/128$. In class-conditional setting, $D{=}2048$ yields current state-of-the-art FID of $1.74$ on CIFAR-10. In addition, we demonstrate that models with smaller $D$ exhibit improved robustness against modeling errors. Code is available at https://github.com/Newbeeer/pfgmpp"
                    },
                    {
                        "title": "GenPhys: From Physical Processes to Generative Models",
                        "abstract": "Since diffusion models (DM) and the more recent Poisson flow generative models (PFGM) are inspired by physical processes, it is reasonable to ask: Can physical processes offer additional new generative models? We show that the answer is yes. We introduce a general family, Generative Models from Physical Processes (GenPhys), where we translate partial differential equations (PDEs) describing physical processes to generative models. We show that generative models can be constructed from s-generative PDEs (s for smooth). GenPhys subsume the two existing generative models (DM and PFGM) and even give rise to new families of generative models, e.g.,\"Yukawa Generative Models\"inspired from weak interactions. On the other hand, some physical processes by default do not belong to the GenPhys family, e.g., the wave equation and the Schr\\\"{o}dinger equation, but could be made into the GenPhys family with some modifications. Our goal with GenPhys is to explore and expand the design space of generative models."
                    },
                    {
                        "title": "Poisson Flow Generative Models",
                        "abstract": "We propose a new\"Poisson flow\"generative model (PFGM) that maps a uniform distribution on a high-dimensional hemisphere into any data distribution. We interpret the data points as electrical charges on the $z=0$ hyperplane in a space augmented with an additional dimension $z$, generating a high-dimensional electric field (the gradient of the solution to Poisson equation). We prove that if these charges flow upward along electric field lines, their initial distribution in the $z=0$ plane transforms into a distribution on the hemisphere of radius $r$ that becomes uniform in the $r \\to\\infty$ limit. To learn the bijective transformation, we estimate the normalized field in the augmented space. For sampling, we devise a backward ODE that is anchored by the physically meaningful additional dimension: the samples hit the unaugmented data manifold when the $z$ reaches zero. Experimentally, PFGM achieves current state-of-the-art performance among the normalizing flow models on CIFAR-10, with an Inception score of $9.68$ and a FID score of $2.35$. It also performs on par with the state-of-the-art SDE approaches while offering $10\\times $ to $20 \\times$ acceleration on image generation tasks. Additionally, PFGM appears more tolerant of estimation errors on a weaker network architecture and robust to the step size in the Euler method. The code is available at https://github.com/Newbeeer/poisson_flow ."
                    },
                    {
                        "title": "A Survey on Generative Diffusion Model",
                        "abstract": "\u2014Deep learning shows excellent potential in generation tasks thanks to deep latent representation. Generative models are classes of models that can generate observations randomly with respect to certain implied parameters. Recently, the diffusion Model has become a raising class of generative models by virtue of its power-generating ability. Nowadays, great achievements have been reached. More applications except for computer vision, speech generation, bioinformatics, and natural language processing are to be explored in this \ufb01eld. However, the diffusion model has its genuine drawback of a slow generation process, leading to many enhanced works. This survey makes a summary of the \ufb01eld of the diffusion model. We \ufb01rst state the main problem with two landmark works \u2013 DDPM and DSM. Then, we present a diverse range of advanced techniques to speed up the diffusion models \u2013 training schedule, training-free sampling, mixed-modeling, and score & diffusion uni\ufb01cation. Regarding existing models, we also provide a benchmark of FID score, IS, and NLL according to speci\ufb01c NFE. Moreover, applications with diffusion models are introduced including computer vision, sequence modeling, audio, and AI for science. Finally, there is a summarization of this \ufb01eld together with limitations & further directions."
                    },
                    {
                        "title": "Controlling Directions Orthogonal to a Classifier",
                        "abstract": "We propose to identify directions invariant to a given classifier so that these directions can be controlled in tasks such as style transfer. While orthogonal decomposition is directly identifiable when the given classifier is linear, we formally define a notion of orthogonality in the non-linear case. We also provide a surprisingly simple method for constructing the orthogonal classifier (a classifier utilizing directions other than those of the given classifier). Empirically, we present three use cases where controlling orthogonal variation is important: style transfer, domain adaptation, and fairness. The orthogonal classifier enables desired style transfer when domains vary in multiple aspects, improves domain adaptation with label shifts and mitigates the unfairness as a predictor. The code is available at http://github.com/Newbeeer/orthogonal_classifier"
                    },
                    {
                        "title": "Anytime Sampling for Autoregressive Models via Ordered Autoencoding",
                        "abstract": "Autoregressive models are widely used for tasks such as image and audio generation. The sampling process of these models, however, does not allow interruptions and cannot adapt to real-time computational resources. This challenge impedes the deployment of powerful autoregressive models, which involve a slow sampling process that is sequential in nature and typically scales linearly with respect to the data dimension. To address this difficulty, we propose a new family of autoregressive models that enables anytime sampling. Inspired by Principal Component Analysis, we learn a structured representation space where dimensions are ordered based on their importance with respect to reconstruction. Using an autoregressive model in this latent space, we trade off sample quality for computational efficiency by truncating the generation process before decoding into the original data space. Experimentally, we demonstrate in several image and audio generation tasks that sample quality degrades gracefully as we reduce the computational budget for sampling. The approach suffers almost no loss in sample quality (measured by FID) using only 60\\% to 80\\% of all latent dimensions for image data. Code is available at https://github.com/Newbeeer/Anytime-Auto-Regressive-Model ."
                    },
                    {
                        "title": "Learning Representations that Support Robust Transfer of Predictors",
                        "abstract": "Ensuring generalization to unseen environments remains a challenge. Domain shift can lead to substantially degraded performance unless shifts are well-exercised within the available training environments. We introduce a simple robust estimation criterion -- transfer risk -- that is specifically geared towards optimizing transfer to new environments. Effectively, the criterion amounts to finding a representation that minimizes the risk of applying any optimal predictor trained on one environment to another. The transfer risk essentially decomposes into two terms, a direct transfer term and a weighted gradient-matching term arising from the optimality of per-environment predictors. Although inspired by IRM, we show that transfer risk serves as a better out-of-distribution generalization criterion, both theoretically and empirically. We further demonstrate the impact of optimizing such transfer risk on two controlled settings, each representing a different pattern of environment shift, as well as on two real-world datasets. Experimentally, the approach outperforms baselines across various out-of-distribution generalization tasks. Code is available at \\url{https://github.com/Newbeeer/TRM}."
                    },
                    {
                        "title": "Can Subnetwork Structure be the Key to Out-of-Distribution Generalization?",
                        "abstract": "Can models with particular structure avoid being biased towards spurious correlation in out-of-distribution (OOD) generalization? Peters et al. (2016) provides a positive answer for linear cases. In this paper, we use a functional modular probing method to analyze deep model structures under OOD setting. We demonstrate that even in biased models (which focus on spurious correlation) there still exist unbiased functional subnetworks. Furthermore, we articulate and demonstrate the functional lottery ticket hypothesis: full network contains a subnetwork that can achieve better OOD performance. We then propose Modular Risk Minimization to solve the subnetwork selection problem. Our algorithm learns the subnetwork structure from a given dataset, and can be combined with any other OOD regularization methods. Experiments on various OOD generalization tasks corroborate the effectiveness of our method."
                    },
                    {
                        "title": "A Theory of Usable Information Under Computational Constraints",
                        "abstract": "We propose a new framework for reasoning about information in complex systems. Our foundation is based on a variational extension of Shannon's information theory that takes into account the modeling power and computational constraints of the observer. The resulting \\emph{predictive $\\mathcal{V}$-information} encompasses mutual information and other notions of informativeness such as the coefficient of determination. Unlike Shannon's mutual information and in violation of the data processing inequality, $\\mathcal{V}$-information can be created through computation. This is consistent with deep neural networks extracting hierarchies of progressively more informative features in representation learning. Additionally, we show that by incorporating computational constraints, $\\mathcal{V}$-information can be reliably estimated from data even in high dimensions with PAC-style guarantees. Empirically, we demonstrate predictive $\\mathcal{V}$-information is more effective than mutual information for structure learning and fair representation learning."
                    },
                    {
                        "title": "Max-MIG: an Information Theoretic Approach for Joint Learning from Crowds",
                        "abstract": "Eliciting labels from crowds is a potential way to obtain large labeled data. Despite a variety of methods developed for learning from crowds, a key challenge remains unsolved: \\emph{learning from crowds without knowing the information structure among the crowds a priori, when some people of the crowds make highly correlated mistakes and some of them label effortlessly (e.g. randomly)}. We propose an information theoretic approach, Max-MIG, for joint learning from crowds, with a common assumption: the crowdsourced labels and the data are independent conditioning on the ground truth. Max-MIG simultaneously aggregates the crowdsourced labels and learns an accurate data classifier. Furthermore, we devise an accurate data-crowds forecaster that employs both the data and the crowdsourced labels to forecast the ground truth. To the best of our knowledge, this is the first algorithm that solves the aforementioned challenge of learning from crowds. In addition to the theoretical validation, we also empirically show that our algorithm achieves the new state-of-the-art results in most settings, including the real-world data, and is the first algorithm that is robust to various information structures. Codes are available at \\hyperlink{this https URL}{this https URL}"
                    },
                    {
                        "title": "L_DMI: A Novel Information-theoretic Loss Function for Training Deep Nets Robust to Label Noise",
                        "abstract": "Accurately annotating large scale dataset is notoriously expensive both in time and in money. Although acquiring low-quality-annotated dataset can be much cheaper, it often badly damages the performance of trained models when using such dataset without particular treatment. Various methods have been proposed for learning with noisy labels. However, most methods only handle limited kinds of noise patterns, require auxiliary information or steps (e.g., knowing or estimating the noise transition matrix), or lack theoretical justification. In this paper, we propose a novel information-theoretic loss function, L_DMI, for training deep neural networks robust to label noise. The core of L_DMI is a generalized version of mutual information, termed Determinant based Mutual Information (DMI), which is not only information-monotone but also relatively invariant. To the best of our knowledge, L_DMI is the first loss function that is provably robust to instance-independent label noise, regardless of noise pattern, and it can be applied to any existing classification neural networks straightforwardly without any auxiliary information. In addition to theoretical justification, we also empirically show that using L_DMI outperforms all other counterparts in the classification task on both image dataset and natural language dataset include Fashion-MNIST, CIFAR-10, Dogs vs. Cats, MR with a variety of synthesized noise patterns and noise amounts, as well as a real-world dataset Clothing1M."
                    },
                    {
                        "title": "L_DMI: An Information-theoretic Noise-robust Loss Function",
                        "abstract": "Accurately annotating large scale dataset is notoriously expensive both in time and in money. Although acquiring low-quality-annotated dataset can be much cheaper, it often badly damages the performance of trained models when using such dataset without particular treatment. Various methods have been proposed for learning with noisy labels. However, most methods only handle limited kinds of noise patterns, require auxiliary information or steps (e.g. , knowing or estimating the noise transition matrix), or lack theoretical justification. In this paper, we propose a novel information-theoretic loss function, $\\mathcal{L}_{DMI}$, for training deep neural networks robust to label noise. The core of $\\mathcal{L}_{DMI}$ is a generalized version of mutual information, termed Determinant based Mutual Information (DMI), which is not only information-monotone but also relatively invariant. \\emph{To the best of our knowledge, $\\mathcal{L}_{DMI}$ is the first loss function that is provably robust to instance-independent label noise, regardless of noise pattern, and it can be applied to any existing classification neural networks straightforwardly without any auxiliary information}. In addition to theoretical justification, we also empirically show that using $\\mathcal{L}_{DMI}$ outperforms all other counterparts in the classification task on both image dataset and natural language dataset include Fashion-MNIST, CIFAR-10, Dogs vs. Cats, MR with a variety of synthesized noise patterns and noise amounts, as well as a real-world dataset Clothing1M. Codes are available at this https URL ."
                    }
                ]
            },
            "4e53fac6-bdd4-4c9b-8002-43153a4bee3d": {
                "pk": "4e53fac6-bdd4-4c9b-8002-43153a4bee3d",
                "name": "Valentin de Bortoli",
                "collaborators": [
                    "M. Hutchinson",
                    "A. Doucet",
                    "Emile Mathieu",
                    "Arnaud Doucet",
                    "N. Fishman",
                    "Leo Klarner",
                    "George Deligiannidis",
                    "Andrew Campbell",
                    "Angus Phillips",
                    "Hai-Dang Dau",
                    "Peter Wirnsberger",
                    "Iryna Korshunova",
                    "A. Mnih",
                    "Pierre Marion",
                    "Anna Korba",
                    "Peter Bartlett",
                    "Mathieu Blondel",
                    "Felipe Llinares-L'opez",
                    "Courtney Paquette",
                    "Quentin Berthet",
                    "Alexandre Galashov",
                    "Arthur Gretton",
                    "Giulio Biroli",
                    "Tony Bonnaire",
                    "Marc M'ezard",
                    "Matteo Pariset",
                    "Ya-Ping Hsieh",
                    "Charlotte Bunne",
                    "Andreas Krause",
                    "William Harvey",
                    "Christian Weilbach",
                    "Tom Rainforth",
                    "Joe Benton",
                    "Guan-Horng Liu",
                    "Tianrong Chen",
                    "Evangelos A. Theodorou",
                    "Weilie Nie",
                    "Yuyang Shi",
                    "Maxence Noble",
                    "Alain Durmus",
                    "Jason Yim",
                    "Brian L. Trippe",
                    "R. Barzilay",
                    "T. Jaakkola",
                    "Marien Renaud",
                    "Jiaming Liu",
                    "Andr'es Almansa",
                    "U. Kamilov",
                    "Vincent Dutordoir",
                    "Y. Teh",
                    "Richard E. Turner"
                ],
                "domain": [
                    "Generative Modeling",
                    "Denoising Diffusion Models",
                    "Optimal Transport",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Particle Denoising Diffusion Sampler",
                        "abstract": "Denoising diffusion models have become ubiquitous for generative modeling. The core idea is to transport the data distribution to a Gaussian by using a diffusion. Approximate samples from the data distribution are then obtained by estimating the time-reversal of this diffusion using score matching ideas. We follow here a similar strategy to sample from unnormalized probability densities and compute their normalizing constants. However, the time-reversed diffusion is here simulated by using an original iterative particle scheme relying on a novel score matching loss. Contrary to standard denoising diffusion models, the resulting Particle Denoising Diffusion Sampler (PDDS) provides asymptotically consistent estimates under mild assumptions. We demonstrate PDDS on multimodal and high dimensional sampling tasks."
                    },
                    {
                        "title": "Target Score Matching",
                        "abstract": "Denoising Score Matching estimates the score of a noised version of a target distribution by minimizing a regression loss and is widely used to train the popular class of Denoising Diffusion Models. A well known limitation of Denoising Score Matching, however, is that it yields poor estimates of the score at low noise levels. This issue is particularly unfavourable for problems in the physical sciences and for Monte Carlo sampling tasks for which the score of the clean original target is known. Intuitively, estimating the score of a slightly noised version of the target should be a simple task in such cases. In this paper, we address this shortcoming and show that it is indeed possible to leverage knowledge of the target score. We present a Target Score Identity and corresponding Target Score Matching regression loss which allows us to obtain score estimates admitting favourable properties at low noise levels."
                    },
                    {
                        "title": "Schr\u00f6dinger Bridge Flow for Unpaired Data Translation",
                        "abstract": "Mass transport problems arise in many areas of machine learning whereby one wants to compute a map transporting one distribution to another. Generative modeling techniques like Generative Adversarial Networks (GANs) and Denoising Diffusion Models (DDMs) have been successfully adapted to solve such transport problems, resulting in CycleGAN and Bridge Matching respectively. However, these methods do not approximate Optimal Transport (OT) maps, which are known to have desirable properties. Existing techniques approximating OT maps for high-dimensional data-rich problems, such as DDM-based Rectified Flow and Schr\\\"odinger Bridge procedures, require fully training a DDM-type model at each iteration, or use mini-batch techniques which can introduce significant errors. We propose a novel algorithm to compute the Schr\\\"odinger Bridge, a dynamic entropy-regularised version of OT, that eliminates the need to train multiple DDM-like models. This algorithm corresponds to a discretisation of a flow of path measures, which we call the Schr\\\"odinger Bridge Flow, whose only stationary point is the Schr\\\"odinger Bridge. We demonstrate the performance of our algorithm on a variety of unpaired data translation tasks."
                    },
                    {
                        "title": "Implicit Diffusion: Efficient Optimization through Stochastic Sampling",
                        "abstract": "We present a new algorithm to optimize distributions defined implicitly by parameterized stochastic diffusions. Doing so allows us to modify the outcome distribution of sampling processes by optimizing over their parameters. We introduce a general framework for first-order optimization of these processes, that performs jointly, in a single loop, optimization and sampling steps. This approach is inspired by recent advances in bilevel optimization and automatic implicit differentiation, leveraging the point of view of sampling as optimization over the space of probability distributions. We provide theoretical guarantees on the performance of our method, as well as experimental results demonstrating its effectiveness. We apply it to training energy-based models and finetuning denoising diffusions."
                    },
                    {
                        "title": "Deep MMD Gradient Flow without adversarial training",
                        "abstract": "We propose a gradient flow procedure for generative modeling by transporting particles from an initial source distribution to a target distribution, where the gradient field on the particles is given by a noise-adaptive Wasserstein Gradient of the Maximum Mean Discrepancy (MMD). The noise-adaptive MMD is trained on data distributions corrupted by increasing levels of noise, obtained via a forward diffusion process, as commonly used in denoising diffusion probabilistic models. The result is a generalization of MMD Gradient Flow, which we call Diffusion-MMD-Gradient Flow or DMMD. The divergence training procedure is related to discriminator training in Generative Adversarial Networks (GAN), but does not require adversarial training. We obtain competitive empirical performance in unconditional image generation on CIFAR10, MNIST, CELEB-A (64 x64) and LSUN Church (64 x 64). Furthermore, we demonstrate the validity of the approach when MMD is replaced by a lower bound on the KL divergence."
                    },
                    {
                        "title": "Dynamical Regimes of Diffusion Models",
                        "abstract": "Using statistical physics methods, we study generative diffusion models in the regime where the dimension of space and the number of data are large, and the score function has been trained optimally. Our analysis reveals three distinct dynamical regimes during the backward generative diffusion process. The generative dynamics, starting from pure noise, encounters first a 'speciation' transition where the gross structure of data is unraveled, through a mechanism similar to symmetry breaking in phase transitions. It is followed at later time by a 'collapse' transition where the trajectories of the dynamics become attracted to one of the memorized data points, through a mechanism which is similar to the condensation in a glass phase. For any dataset, the speciation time can be found from a spectral analysis of the correlation matrix, and the collapse time can be found from the estimation of an 'excess entropy' in the data. The dependence of the collapse time on the dimension and number of data provides a thorough characterization of the curse of dimensionality for diffusion models. Analytical solutions for simple models like high-dimensional Gaussian mixtures substantiate these findings and provide a theoretical framework, while extensions to more complex scenarios and numerical validations with real datasets confirm the theoretical predictions."
                    },
                    {
                        "title": "Unbalanced Diffusion Schr\u00f6dinger Bridge",
                        "abstract": "Schr\\\"odinger bridges (SBs) provide an elegant framework for modeling the temporal evolution of populations in physical, chemical, or biological systems. Such natural processes are commonly subject to changes in population size over time due to the emergence of new species or birth and death events. However, existing neural parameterizations of SBs such as diffusion Schr\\\"odinger bridges (DSBs) are restricted to settings in which the endpoints of the stochastic process are both probability measures and assume conservation of mass constraints. To address this limitation, we introduce unbalanced DSBs which model the temporal evolution of marginals with arbitrary finite mass. This is achieved by deriving the time reversal of stochastic differential equations with killing and birth terms. We present two novel algorithmic schemes that comprise a scalable objective function for training unbalanced DSBs and provide a theoretical analysis alongside challenging applications on predicting heterogeneous molecular single-cell responses to various cancer drugs and simulating the emergence and spread of new viral variants."
                    },
                    {
                        "title": "Trans-Dimensional Generative Modeling via Jump Diffusion Models",
                        "abstract": "We propose a new class of generative models that naturally handle data of varying dimensionality by jointly modeling the state and dimension of each datapoint. The generative process is formulated as a jump diffusion process that makes jumps between different dimensional spaces. We first define a dimension destroying forward noising process, before deriving the dimension creating time-reversed generative process along with a novel evidence lower bound training objective for learning to approximate it. Simulating our learned approximation to the time-reversed generative process then provides an effective way of sampling data of varying dimensionality by jointly generating state values and dimensions. We demonstrate our approach on molecular and video datasets of varying dimensionality, reporting better compatibility with test-time diffusion guidance imputation tasks and improved interpolation capabilities versus fixed dimensional models that generate state values and dimensions separately."
                    },
                    {
                        "title": "Nearly d-Linear Convergence Bounds for Diffusion Models via Stochastic Localization",
                        "abstract": "Denoising diffusions are a powerful method to generate approximate samples from high-dimensional data distributions. Recent results provide polynomial bounds on their convergence rate, assuming $L^2$-accurate scores. Until now, the tightest bounds were either superlinear in the data dimension or required strong smoothness assumptions. We provide the first convergence bounds which are linear in the data dimension (up to logarithmic factors) assuming only finite second moments of the data distribution. We show that diffusion models require at most $\\tilde O(\\frac{d \\log^2(1/\\delta)}{\\varepsilon^2})$ steps to approximate an arbitrary distribution on $\\mathbb{R}^d$ corrupted with Gaussian noise of variance $\\delta$ to within $\\varepsilon^2$ in KL divergence. Our proof extends the Girsanov-based methods of previous works. We introduce a refined treatment of the error from discretizing the reverse SDE inspired by stochastic localization."
                    },
                    {
                        "title": "Augmented Bridge Matching",
                        "abstract": "Flow and bridge matching are a novel class of processes which encompass diffusion models. One of the main aspect of their increased flexibility is that these models can interpolate between arbitrary data distributions i.e. they generalize beyond generative modeling and can be applied to learning stochastic (and deterministic) processes of arbitrary transfer tasks between two given distributions. In this paper, we highlight that while flow and bridge matching processes preserve the information of the marginal distributions, they do \\emph{not} necessarily preserve the coupling information unless additional, stronger optimality conditions are met. This can be problematic if one aims at preserving the original empirical pairing. We show that a simple modification of the matching process recovers this coupling by augmenting the velocity field (or drift) with the information of the initial sample point. Doing so, we lose the Markovian property of the process but preserve the coupling information between distributions. We illustrate the efficiency of our augmentation in learning mixture of image translation tasks."
                    },
                    {
                        "title": "Diffusion Models for Constrained Domains",
                        "abstract": "Denoising diffusion models are a novel class of generative algorithms that achieve state-of-the-art performance across a range of domains, including image generation and text-to-image tasks. Building on this success, diffusion models have recently been extended to the Riemannian manifold setting, broadening their applicability to a range of problems from the natural and engineering sciences. However, these Riemannian diffusion models are built on the assumption that their forward and backward processes are well-defined for all times, preventing them from being applied to an important set of tasks that consider manifolds defined via a set of inequality constraints. In this work, we introduce a principled framework to bridge this gap. We present two distinct noising processes based on (i) the logarithmic barrier metric and (ii) the reflected Brownian motion induced by the constraints. As existing diffusion model techniques cannot be applied in this setting, we derive new tools to define such models in our framework. We then demonstrate the practical utility of our methods on a number of synthetic and real-world tasks, including applications from robotics and protein design."
                    },
                    {
                        "title": "Diffusion Schr\u00f6dinger Bridge Matching",
                        "abstract": "Solving transport problems, i.e. finding a map transporting one given distribution to another, has numerous applications in machine learning. Novel mass transport methods motivated by generative modeling have recently been proposed, e.g. Denoising Diffusion Models (DDMs) and Flow Matching Models (FMMs) implement such a transport through a Stochastic Differential Equation (SDE) or an Ordinary Differential Equation (ODE). However, while it is desirable in many applications to approximate the deterministic dynamic Optimal Transport (OT) map which admits attractive properties, DDMs and FMMs are not guaranteed to provide transports close to the OT map. In contrast, Schr\\\"odinger bridges (SBs) compute stochastic dynamic mappings which recover entropy-regularized versions of OT. Unfortunately, existing numerical methods approximating SBs either scale poorly with dimension or accumulate errors across iterations. In this work, we introduce Iterative Markovian Fitting (IMF), a new methodology for solving SB problems, and Diffusion Schr\\\"odinger Bridge Matching (DSBM), a novel numerical algorithm for computing IMF iterates. DSBM significantly improves over previous SB numerics and recovers as special/limiting cases various recent transport methods. We demonstrate the performance of DSBM on a variety of problems."
                    },
                    {
                        "title": "Tree-Based Diffusion Schr\u00f6dinger Bridge with Applications to Wasserstein Barycenters",
                        "abstract": "Multi-marginal Optimal Transport (mOT), a generalization of OT, aims at minimizing the integral of a cost function with respect to a distribution with some prescribed marginals. In this paper, we consider an entropic version of mOT with a tree-structured quadratic cost, i.e., a function that can be written as a sum of pairwise cost functions between the nodes of a tree. To address this problem, we develop Tree-based Diffusion Schr\\\"odinger Bridge (TreeDSB), an extension of the Diffusion Schr\\\"odinger Bridge (DSB) algorithm. TreeDSB corresponds to a dynamic and continuous state-space counterpart of the multimarginal Sinkhorn algorithm. A notable use case of our methodology is to compute Wasserstein barycenters which can be recast as the solution of a mOT problem on a star-shaped tree. We demonstrate that our methodology can be applied in high-dimensional settings such as image interpolation and Bayesian fusion."
                    },
                    {
                        "title": "SE(3) diffusion model with application to protein backbone generation",
                        "abstract": "The design of novel protein structures remains a challenge in protein engineering for applications across biomedicine and chemistry. In this line of work, a diffusion model over rigid bodies in 3D (referred to as frames) has shown success in generating novel, functional protein backbones that have not been observed in nature. However, there exists no principled methodological framework for diffusion on SE(3), the space of orientation preserving rigid motions in R3, that operates on frames and confers the group invariance. We address these shortcomings by developing theoretical foundations of SE(3) invariant diffusion models on multiple frames followed by a novel framework, FrameDiff, for learning the SE(3) equivariant score over multiple frames. We apply FrameDiff on monomer backbone generation and find it can generate designable monomers up to 500 amino acids without relying on a pretrained protein structure prediction network that has been integral to previous methods. We find our samples are capable of generalizing beyond any known protein structure."
                    },
                    {
                        "title": "Plug-and-Play Posterior Sampling under Mismatched Measurement and Prior Models",
                        "abstract": "Posterior sampling has been shown to be a powerful Bayesian approach for solving imaging inverse problems. The recent plug-and-play unadjusted Langevin algorithm (PnP-ULA) has emerged as a promising method for Monte Carlo sampling and minimum mean squared error (MMSE) estimation by combining physical measurement models with deep-learning priors specified using image denoisers. However, the intricate relationship between the sampling distribution of PnP-ULA and the mismatched data-fidelity and denoiser has not been theoretically analyzed. We address this gap by proposing a posterior-L2 pseudometric and using it to quantify an explicit error bound for PnP-ULA under mismatched posterior distribution. We numerically validate our theory on several inverse problems such as sampling from Gaussian mixture models and image deblurring. Our results suggest that the sensitivity of the sampling distribution of PnP-ULA to a mismatch in the measurement model and the denoiser can be precisely characterized."
                    },
                    {
                        "title": "Metropolis Sampling for Constrained Diffusion Models",
                        "abstract": "Denoising diffusion models have recently emerged as the predominant paradigm for generative modelling. Their extension to Riemannian manifolds has facilitated their application to an array of problems in the natural sciences. Yet, in many practical settings, such manifolds are defined by a set of constraints and are not covered by the existing (Riemannian) diffusion model methodology. Recent work has attempted to address this issue by employing novel noising processes based on logarithmic barrier methods or reflected Brownian motions. However, the associated samplers are computationally burdensome as the complexity of the constraints increases. In this paper, we introduce an alternative simple noising scheme based on Metropolis sampling that affords substantial gains in computational efficiency and empirical performance compared to the earlier samplers. Of independent interest, we prove that this new process corresponds to a valid discretisation of the reflected Brownian motion. We demonstrate the scalability and flexibility of our approach on a range of problem settings with convex and non-convex constraints, including applications from geospatial modelling, robotics and protein design."
                    },
                    {
                        "title": "Geometric Neural Diffusion Processes",
                        "abstract": "Denoising diffusion models have proven to be a flexible and effective paradigm for generative modelling. Their recent extension to infinite dimensional Euclidean spaces has allowed for the modelling of stochastic processes. However, many problems in the natural sciences incorporate symmetries and involve data living in non-Euclidean spaces. In this work, we extend the framework of diffusion models to incorporate a series of geometric priors in infinite-dimension modelling. We do so by a) constructing a noising process which admits, as limiting distribution, a geometric Gaussian process that transforms under the symmetry group of interest, and b) approximating the score with a neural network that is equivariant w.r.t. this group. We show that with these conditions, the generative functional model admits the same symmetry. We demonstrate scalability and capacity of the model, using a novel Langevin-based conditional sampler, to fit complex scalar and vector fields, with Euclidean and spherical codomain, on synthetic and real-world weather data."
                    },
                    {
                        "title": "Di \ufffd usion Models for Constrained Domains",
                        "abstract": "Denoising di \ufffd usion models are a novel class of generative algorithms that achieve state-of-the-art performance across a range of domains, including image generation and text-to-image tasks. Building on this success, di \ufffd usion models have recently been extended to the Riemannian manifold setting, broadening their applicability to a range of problems from the natural and engineering sciences. However, these Riemannian di \ufffd usion models are built on the assumption that their forward and backward processes are well-de\ufb01ned for all times, preventing them from being applied to an important set of tasks that consider manifolds de\ufb01ned via a set of inequality constraints. In this work, we introduce a principled framework to bridge this gap. We present two distinct noising processes based on (i) the logarithmic barrier metric and (ii) the re\ufb02ected Brownian motion induced by the constraints. As existing di \ufffd usion model techniques cannot be applied in this setting, we derive new tools to de\ufb01ne such models in our framework. We then demonstrate the practical utility of our methods on a number of synthetic and real-world tasks, including applications from robotics and protein design."
                    }
                ]
            },
            "350ff0ce-9fbe-4f3d-a966-22bccef2b676": {
                "pk": "350ff0ce-9fbe-4f3d-a966-22bccef2b676",
                "name": "Regina Barzilay",
                "collaborators": [
                    "T. Jaakkola",
                    "Jason Yim",
                    "Hannes St\u00e4rk",
                    "Connor W. Coley",
                    "Menghua Wu",
                    "Adam Fisch",
                    "Adam Yala",
                    "Yujie Qian",
                    "Zhengkai Tu",
                    "Yujia Bao",
                    "Xuan Zhang",
                    "Limei Wang",
                    "Jacob Helwig",
                    "Youzhi Luo",
                    "Cong Fu",
                    "Yaochen Xie",
                    "Meng Liu",
                    "Yu-Ching Lin",
                    "Zhao Xu",
                    "Keqiang Yan",
                    "Keir Adams",
                    "Maurice Weiler",
                    "Xiner Li",
                    "Tianfan Fu",
                    "Yucheng Wang",
                    "Haiyang Yu",
                    "Yuqing Xie",
                    "Xiang Fu",
                    "A. Strasser",
                    "Shenglong Xu",
                    "Yi Liu",
                    "Yuanqi Du",
                    "Alexandra Saxton",
                    "Hongyi Ling",
                    "Hannah Lawrence",
                    "Shurui Gui",
                    "Carl N. Edwards",
                    "Nicholas Gao",
                    "A. Ladera",
                    "Tailin Wu",
                    "E. Hofgard",
                    "A. M. Tehrani",
                    "Rui Wang",
                    "Ameya Daigavane",
                    "Montgomery Bohde",
                    "Jerry Kurtin",
                    "Qiang Huang",
                    "Tuong Phung",
                    "Minkai Xu",
                    "Chaitanya K. Joshi",
                    "Simon V. Mathis",
                    "K. Azizzadenesheli",
                    "Ada Fang",
                    "A. Aspuru\u2010Guzik",
                    "E. Bekkers",
                    "Michael M. Bronstein",
                    "M. Zitnik",
                    "Anima Anandkumar",
                    "Stefano Ermon",
                    "Pietro Lio'",
                    "Rose Yu",
                    "Stephan Gunnemann",
                    "J. Leskovec",
                    "Heng Ji",
                    "Jimeng Sun",
                    "Xiaoning Qian",
                    "Xiaofeng Qian",
                    "T. Smidt",
                    "Shuiwang Ji",
                    "Victor Quach",
                    "Tal Schuster",
                    "J. Sohn",
                    "Brent A. Koscher",
                    "Richard B. Canty",
                    "Matthew A. McDonald",
                    "Kevin P. Greenman",
                    "Charles J. McGill",
                    "Camille L. Bilodeau",
                    "Wengong Jin",
                    "Haoyang Wu",
                    "Florence H. Vermeire",
                    "Brooke Jin",
                    "Travis Hart",
                    "Timothy Kulesza",
                    "Shih-Cheng Li",
                    "Rafael G\u00f3mez-Bombarelli",
                    "William H. Green",
                    "Klavs F. Jensen",
                    "Peter G. Mikhael",
                    "Jeremy Wohlwend",
                    "Ludvig Karstens",
                    "J. Xiang",
                    "A. Takigami",
                    "P. Bourgouin",
                    "PuiYee Chan",
                    "Sofiane Mrah",
                    "Wael Amayri",
                    "Y. Juan",
                    "Cheng-Ta Yang",
                    "Y. Wan"
                ],
                "domain": [
                    "Causal Inference",
                    "Machine Learning",
                    "AI for Science",
                    "Protein Engineering"
                ],
                "publications": [
                    {
                        "title": "Sample, estimate, aggregate: A recipe for causal discovery foundation models",
                        "abstract": "Causal discovery, the task of inferring causal structure from data, promises to accelerate scientific research, inform policy making, and more. However, causal discovery algorithms over larger sets of variables tend to be brittle against misspecification or when data are limited. To mitigate these challenges, we train a supervised model that learns to predict a larger causal graph from the outputs of classical causal discovery algorithms run over subsets of variables, along with other statistical hints like inverse covariance. Our approach is enabled by the observation that typical errors in the outputs of classical methods remain comparable across datasets. Theoretically, we show that this model is well-specified, in the sense that it can recover a causal graph consistent with graphs over subsets. Empirically, we train the model to be robust to erroneous estimates using diverse synthetic data. Experiments on real and synthetic data demonstrate that this model maintains high accuracy in the face of misspecification or distribution shift, and can be adapted at low cost to different discovery algorithms or choice of statistics."
                    },
                    {
                        "title": "Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems",
                        "abstract": "Advances in artificial intelligence (AI) are fueling a new paradigm of discoveries in natural sciences. Today, AI has started to advance natural sciences by improving, accelerating, and enabling our understanding of natural phenomena at a wide range of spatial and temporal scales, giving rise to a new area of research known as AI for science (AI4Science). Being an emerging research paradigm, AI4Science is unique in that it is an enormous and highly interdisciplinary area. Thus, a unified and technical treatment of this field is needed yet challenging. This work aims to provide a technically thorough account of a subarea of AI4Science; namely, AI for quantum, atomistic, and continuum systems. These areas aim at understanding the physical world from the subatomic (wavefunctions and electron density), atomic (molecules, proteins, materials, and interactions), to macro (fluids, climate, and subsurface) scales and form an important subarea of AI4Science. A unique advantage of focusing on these areas is that they largely share a common set of challenges, thereby allowing a unified and foundational treatment. A key common challenge is how to capture physics first principles, especially symmetries, in natural systems by deep learning methods. We provide an in-depth yet intuitive account of techniques to achieve equivariance to symmetry transformations. We also discuss other common technical challenges, including explainability, out-of-distribution generalization, knowledge transfer with foundation and large language models, and uncertainty quantification. To facilitate learning and education, we provide categorized lists of resources that we found to be useful. We strive to be thorough and unified and hope this initial effort may trigger more community interests and efforts to further advance AI4Science."
                    },
                    {
                        "title": "Conformal Language Modeling",
                        "abstract": "We propose a novel approach to conformal prediction for generative language models (LMs). Standard conformal prediction produces prediction sets -- in place of single predictions -- that have rigorous, statistical performance guarantees. LM responses are typically sampled from the model's predicted distribution over the large, combinatorial output space of natural language. Translating this process to conformal prediction, we calibrate a stopping rule for sampling different outputs from the LM that get added to a growing set of candidates until we are confident that the output set is sufficient. Since some samples may be low-quality, we also simultaneously calibrate and apply a rejection rule for removing candidates from the output set to reduce noise. Similar to conformal prediction, we prove that the sampled set returned by our procedure contains at least one acceptable answer with high probability, while still being empirically precise (i.e., small) on average. Furthermore, within this set of candidate responses, we show that we can also accurately identify subsets of individual components -- such as phrases or sentences -- that are each independently correct (e.g., that are not\"hallucinations\"), again with statistical guarantees. We demonstrate the promise of our approach on multiple tasks in open-domain question answering, text summarization, and radiology report generation using different LM variants."
                    },
                    {
                        "title": "Autonomous, multiproperty-driven molecular discovery: From predictions to measurements and back",
                        "abstract": "A closed-loop, autonomous molecular discovery platform driven by integrated machine learning tools was developed to accelerate the design of molecules with desired properties. We demonstrated two case studies on dye-like molecules, targeting absorption wavelength, lipophilicity, and photooxidative stability. In the first study, the platform experimentally realized 294 unreported molecules across three automatic iterations of molecular design-make-test-analyze cycles while exploring the structure-function space of four rarely reported scaffolds. In each iteration, the property prediction models that guided exploration learned the structure-property space of diverse scaffold derivatives, which were realized with multistep syntheses and a variety of reactions. The second study exploited property models trained on the explored chemical space and previously reported molecules to discover nine top-performing molecules within a lightly explored structure-property space. Editor\u2019s summary As chemists make increasing use of automated equipment and predictive synthesis algorithms, an autonomous research device is coming closer to realization. Koscher et al. developed a prototype platform for exploring the prospective dye-like properties of several polycyclic small-molecule classes. By combining machine learning\u2013enabled planning and analysis with robotic reaction modules, the apparatus iteratively synthesizes variously substituted compounds to optimize their absorption properties, lipophilicity, and photo-oxidative stability. The authors elaborate on when and why human input still comes into play. \u2014Jake S. Yeston A platform integrates algorithmic and robotic advances to synthesize dye molecules and iteratively optimize their properties. INTRODUCTION The discovery of small molecules with desired functional properties is critical to advances in health, energy, and sustainability. The process is conducted through often slow, laborious, iterative design-make-test-analyze (DMTA) cycles. Emerging machine learning (ML) tools can generate new candidate molecules, predict their properties, and propose reaction pathways with computer-aided synthesis planning (CASP). Progress in chemical automation can enable chemical synthesis and characterization with minimal human intervention after a manual setup. For example, automated feedback-guided chemistry platforms have demonstrated chemical optimization within narrow, well-defined classes of reactions. RATIONALE Integrating ML generative algorithms, ML property prediction, CASP, robotics, and automated chemical synthesis, purification, and characterization into a DMTA workflow can lead to the development of autonomous chemical discovery platforms able to operate across varied chemical spaces without manual reconfiguration. An ideal property-focused discovery platform would propose and synthesize molecules to enrich ML generative and property models and ultimately discover top-performing molecules. In reality, it is necessary to exclude reactions that the available automation hardware cannot safely execute. RESULTS Toward autonomous discovery, we demonstrate an integrated DMTA cycle that iteratively proposes, realizes, and characterizes molecules to explore chemical spaces guided only by prediction tools. Four molecular scaffolds with potential applications as organic dyes serve as exploration case studies, whereas a fifth scaffold demonstrates the platform-exploiting refined ML models to optimize dye properties. Dye-like molecules constitute a compelling test case as fundamental properties are readily measurable and their realization involves multistep synthesis by a wide variety of chemical transformations. The platform automatically characterized three target properties\u2014absorption maximum (\u03bbmax), octanol-water partition coefficient (logKow), and photooxidative stability (logkdeg)\u2014and used them to refine model predictions and inform future experiment selection. A graph-completion generative model designs candidate molecules, which are evaluated with an ML model for each of the three properties. A CASP tool proposes multistep synthetic recipes, which are executed by an automated liquid handler, batch reactors, high-performance liquid chromatography (HPLC), and robotic arms. A plate reader measures absorption spectra, calibrated HPLC retention times provide water-octanol partition coefficients, and a simulated solar light source combined with the plate reader quantifies photooxidative degradation. The measured molecular properties are automatically fed back to retrain the property prediction models, completing one step of the automated DMTA cycle. Initially, the platform is highly uncertain about each scaffold (i.e., the ML prediction models have significant model errors), so it chooses to realize the simplest derivatives to anchor property predictions. Subsequent iterations favor more synthetically complex derivatives to fill gaps in the learned chemical space. Three iterations were sufficient for the ML model deviations to approach experimental uncertainty and enable exploitation. Human involvement was limited to setting and adjusting objectives, providing requested materials, and occasionally fixing unrecoverable errors such as clogging of the HPLC unit. CONCLUSION We have demonstrated a DMTA cycle that explores chemical space and exploits known chemical structures without manual experimentation. The platform proposed, synthesized, and characterized 303 unreported dye-like molecules spread across four exploration scaffolds and one exploitation scaffold. The integrated platform can explore unknown structure-property spaces by searching for structures with desired properties (hits) and can exploit characterized structure-property spaces by optimizing structures of promising candidates (leads). Further development of closed-loop integrated platforms promises to accelerate molecular discovery. Autonomous molecular discovery. A suite of machine learning tools iteratively plans experiments to learn about a structure-property space defined by molecular dye scaffolds. \u03b5^exp., mean experimental error."
                    },
                    {
                        "title": "Sybil: A Validated Deep Learning Model to Predict Future Lung Cancer Risk From a Single Low-Dose Chest Computed Tomography",
                        "abstract": "PURPOSE Low-dose computed tomography (LDCT) for lung cancer screening is effective, although most eligible people are not being screened. Tools that provide personalized future cancer risk assessment could focus approaches toward those most likely to benefit. We hypothesized that a deep learning model assessing the entire volumetric LDCT data could be built to predict individual risk without requiring additional demographic or clinical data. METHODS We developed a model called Sybil using LDCTs from the National Lung Screening Trial (NLST). Sybil requires only one LDCT and does not require clinical data or radiologist annotations; it can run in real time in the background on a radiology reading station. Sybil was validated on three independent data sets: a heldout set of 6,282 LDCTs from NLST participants, 8,821 LDCTs from Massachusetts General Hospital (MGH), and 12,280 LDCTs from Chang Gung Memorial Hospital (CGMH, which included people with a range of smoking history including nonsmokers). RESULTS Sybil achieved area under the receiver-operator curves for lung cancer prediction at 1 year of 0.92 (95% CI, 0.88 to 0.95) on NLST, 0.86 (95% CI, 0.82 to 0.90) on MGH, and 0.94 (95% CI, 0.91 to 1.00) on CGMH external validation sets. Concordance indices over 6 years were 0.75 (95% CI, 0.72 to 0.78), 0.81 (95% CI, 0.77 to 0.85), and 0.80 (95% CI, 0.75 to 0.86) for NLST, MGH, and CGMH, respectively. CONCLUSION Sybil can accurately predict an individual's future lung cancer risk from a single LDCT scan to further enable personalized screening. Future study is required to understand Sybil's clinical applications. Our model and annotations are publicly available."
                    },
                    {
                        "title": "RxnScribe: A Sequence Generation Model for Reaction Diagram Parsing",
                        "abstract": "Reaction diagram parsing is the task of extracting reaction schemes from a diagram in the chemistry literature. The reaction diagrams can be arbitrarily complex; thus, robustly parsing them into structured data is an open challenge. In this paper, we present RxnScribe, a machine learning model for parsing reaction diagrams of varying styles. We formulate this structured prediction task with a sequence generation approach, which condenses the traditional pipeline into an end-to-end model. We train RxnScribe on a dataset of 1378 diagrams and evaluate it with cross validation, achieving an 80.0% soft match F1 score, with significant improvements over previous models. Our code and data are publicly available at https://github.com/thomas0809/RxnScribe."
                    },
                    {
                        "title": "DiffDock-PP: Rigid Protein-Protein Docking with Diffusion Models",
                        "abstract": "Understanding how proteins structurally interact is crucial to modern biology, with applications in drug discovery and protein design. Recent machine learning methods have formulated protein-small molecule docking as a generative problem with significant performance boosts over both traditional and deep learning baselines. In this work, we propose a similar approach for rigid protein-protein docking: DiffDock-PP is a diffusion generative model that learns to translate and rotate unbound protein structures into their bound conformations. We achieve state-of-the-art performance on DIPS with a median C-RMSD of 4.85, outperforming all considered baselines. Additionally, DiffDock-PP is faster than all search-based methods and generates reliable confidence estimates for its predictions. Our code is publicly available at $\\texttt{https://github.com/ketatam/DiffDock-PP}$"
                    },
                    {
                        "title": "MorphPiece : A Linguistic Tokenizer for Large Language Models",
                        "abstract": "Tokenization is a critical part of modern NLP pipelines. However, contemporary tokenizers for Large Language Models are based on statistical analysis of text corpora, without much consideration to the linguistic features. I propose a linguistically motivated tokenization scheme, MorphPiece, which is based partly on morphological segmentation of the underlying text. A GPT-style causal language model trained on this tokenizer (called MorphGPT) shows comparable or superior performance on a variety of supervised and unsupervised NLP tasks, compared to the OpenAI GPT-2 model. Specifically I evaluated MorphGPT on language modeling tasks, zero-shot performance on GLUE Benchmark with various prompt templates, massive text embedding benchmark (MTEB) for supervised and unsupervised performance, and lastly with another morphological tokenization scheme (FLOTA, Hoffmann et al., 2022) and find that the model trained on MorphPiece outperforms GPT-2 on most evaluations, at times with considerable margin, despite being trained for about half the training iterations."
                    },
                    {
                        "title": "Improving protein optimization with smoothed fitness landscapes",
                        "abstract": "The ability to engineer novel proteins with higher fitness for a desired property would be revolutionary for biotechnology and medicine. Modeling the combinatorially large space of sequences is infeasible; prior methods often constrain optimization to a small mutational radius, but this drastically limits the design space. Instead of heuristics, we propose smoothing the fitness landscape to facilitate protein optimization. First, we formulate protein fitness as a graph signal then use Tikunov regularization to smooth the fitness landscape. We find optimizing in this smoothed landscape leads to improved performance across multiple methods in the GFP and AAV benchmarks. Second, we achieve state-of-the-art results utilizing discrete energy-based models and MCMC in the smoothed landscape. Our method, called Gibbs sampling with Graph-based Smoothing (GGS), demonstrates a unique ability to achieve 2.5 fold fitness improvement (with in-silico evaluation) over its training set. GGS demonstrates potential to optimize proteins in the limited data regime. Code: https://github.com/kirjner/GGS"
                    },
                    {
                        "title": "SE(3) diffusion model with application to protein backbone generation",
                        "abstract": "The design of novel protein structures remains a challenge in protein engineering for applications across biomedicine and chemistry. In this line of work, a diffusion model over rigid bodies in 3D (referred to as frames) has shown success in generating novel, functional protein backbones that have not been observed in nature. However, there exists no principled methodological framework for diffusion on SE(3), the space of orientation preserving rigid motions in R3, that operates on frames and confers the group invariance. We address these shortcomings by developing theoretical foundations of SE(3) invariant diffusion models on multiple frames followed by a novel framework, FrameDiff, for learning the SE(3) equivariant score over multiple frames. We apply FrameDiff on monomer backbone generation and find it can generate designable monomers up to 500 amino acids without relying on a pretrained protein structure prediction network that has been integral to previous methods. We find our samples are capable of generalizing beyond any known protein structure."
                    },
                    {
                        "title": "Harmonic Self-Conditioned Flow Matching for Multi-Ligand Docking and Binding Site Design",
                        "abstract": "A significant amount of protein function requires binding small molecules, including enzymatic catalysis. As such, designing binding pockets for small molecules has several impactful applications ranging from drug synthesis to energy storage. Towards this goal, we first develop HarmonicFlow, an improved generative process over 3D protein-ligand binding structures based on our self-conditioned flow matching objective. FlowSite extends this flow model to jointly generate a protein pocket's discrete residue types and the molecule's binding 3D structure. We show that HarmonicFlow improves upon the state-of-the-art generative processes for docking in simplicity, generality, and performance. Enabled by this structure modeling, FlowSite designs binding sites substantially better than baseline approaches and provides the first general solution for binding site design."
                    },
                    {
                        "title": "Fast protein backbone generation with SE(3) flow matching",
                        "abstract": "We present FrameFlow, a method for fast protein backbone generation using SE(3) flow matching. Specifically, we adapt FrameDiff, a state-of-the-art diffusion model, to the flow-matching generative modeling paradigm. We show how flow matching can be applied on SE(3) and propose modifications during training to effectively learn the vector field. Compared to FrameDiff, FrameFlow requires five times fewer sampling timesteps while achieving two fold better designability. The ability to generate high quality protein samples at a fraction of the cost of previous methods paves the way towards more efficient generative models in de novo protein design."
                    },
                    {
                        "title": "Fast non-autoregressive inverse folding with discrete diffusion",
                        "abstract": "Generating protein sequences that fold into a intended 3D structure is a fundamental step in de novo protein design. De facto methods utilize autoregressive generation, but this eschews higher order interactions that could be exploited to improve inference speed. We describe a non-autoregressive alternative that performs inference using a constant number of calls resulting in a 23 times speed up without a loss in performance on the CATH benchmark. Conditioned on the 3D structure, we fine-tune ProteinMPNN to perform discrete diffusion with a purity prior over the index sampling order. Our approach gives the flexibility in trading off inference speed and accuracy by modulating the diffusion speed. Code: https://github.com/johnyang101/pmpnndiff"
                    },
                    {
                        "title": "Risk-Controlling Model Selection via Guided Bayesian Optimization",
                        "abstract": "Adjustable hyperparameters of machine learning models typically impact various key trade-offs such as accuracy, fairness, robustness, or inference cost. Our goal in this paper is to find a configuration that adheres to user-specified limits on certain risks while being useful with respect to other conflicting metrics. We solve this by combining Bayesian Optimization (BO) with rigorous risk-controlling procedures, where our core idea is to steer BO towards an efficient testing strategy. Our BO method identifies a set of Pareto optimal configurations residing in a designated region of interest. The resulting candidates are statistically verified and the best-performing configuration is selected with guaranteed risk levels. We demonstrate the effectiveness of our approach on a range of tasks with multiple desiderata, including low error rates, equitable predictions, handling spurious correlations, managing rate and distortion in generative models, and reducing computational costs."
                    },
                    {
                        "title": "Predictive Chemistry Augmented with Text Retrieval",
                        "abstract": "This paper focuses on using natural language descriptions to enhance predictive models in the chemistry field. Conventionally, chemoinformatics models are trained with extensive structured data manually extracted from the literature. In this paper, we introduce TextReact, a novel method that directly augments predictive chemistry with texts retrieved from the literature. TextReact retrieves text descriptions relevant for a given chemical reaction, and then aligns them with the molecular representation of the reaction. This alignment is enhanced via an auxiliary masked LM objective incorporated in the predictor training. We empirically validate the framework on two chemistry tasks: reaction condition recommendation and one-step retrosynthesis. By leveraging text retrieval, TextReact significantly outperforms state-of-the-art chemoinformatics models trained solely on molecular data."
                    }
                ]
            },
            "8992e9a2-9c46-4285-a563-08e2867d6089": {
                "pk": "8992e9a2-9c46-4285-a563-08e2867d6089",
                "name": "Tommi Jaakkola",
                "collaborators": [
                    "R. Barzilay",
                    "Bowen Jing",
                    "Jason Yim",
                    "Gabriele Corso",
                    "Hannes St\u00e4rk",
                    "Bonnie Berger",
                    "Chenyu Wang",
                    "Andrew Campbell",
                    "Yilun Xu",
                    "Yujian Liu",
                    "Shiyu Chang",
                    "Yang Zhang",
                    "P. Holderrieth",
                    "Emile Mathieu",
                    "Andrew Y. K. Foong",
                    "Michael Gastegger",
                    "Jos'e Jim'enez-Luna",
                    "Sarah Lewis",
                    "Victor Garcia Satorras",
                    "Bastiaan S. Veeling",
                    "Frank No'e",
                    "Julia Balla",
                    "S. Mishra-Sharma",
                    "C. Cuesta-L\u00e1zaro",
                    "Tess E. Smidt",
                    "Arash Vahdat",
                    "Karsten Kreis",
                    "Blake R. Duschatko",
                    "Xiang Fu",
                    "Cameron J. Owen",
                    "Yu Xie",
                    "Albert Musaelian",
                    "Boris Kozinsky",
                    "Ezra Erives",
                    "Arthur Deng",
                    "Benjamin Fry",
                    "Nicholas Polizzi",
                    "Sharut Gupta",
                    "Yifei Wang",
                    "Stefanie Jegelka",
                    "Menghua Wu",
                    "Yujia Bao",
                    "Marton Havasi",
                    "Neta Shaul",
                    "Itai Gat",
                    "Brian Karrer",
                    "Ricky T. Q. Chen",
                    "Y. Lipman",
                    "Tom Rainforth",
                    "Masatoshi Uehara",
                    "Yichun He",
                    "Amy Wang",
                    "Tommaso Biancalani",
                    "Avantika Lal",
                    "Sergey Levine",
                    "Hanchen Wang",
                    "Aviv Regev"
                ],
                "domain": [
                    "Protein Design",
                    "Generative Models",
                    "Machine Learning",
                    "Structural Biology"
                ],
                "publications": [
                    {
                        "title": "Improved motif-scaffolding with SE(3) flow matching",
                        "abstract": "Protein design often begins with the knowledge of a desired function from a motif which motif-scaffolding aims to construct a functional protein around. Recently, generative models have achieved breakthrough success in designing scaffolds for a range of motifs. However, generated scaffolds tend to lack structural diversity, which can hinder success in wet-lab validation. In this work, we extend FrameFlow, an SE(3) flow matching model for protein backbone generation, to perform motif-scaffolding with two complementary approaches. The first is motif amortization, in which FrameFlow is trained with the motif as input using a data augmentation strategy. The second is motif guidance, which performs scaffolding using an estimate of the conditional score from FrameFlow without additional training. On a benchmark of 24 biologically meaningful motifs, we show our method achieves 2.5 times more designable and unique motif-scaffolds compared to state-of-the-art. Code: https://github.com/microsoft/protein-frame-flow"
                    },
                    {
                        "title": "A Cosmic-Scale Benchmark for Symmetry-Preserving Data Processing",
                        "abstract": "Efficiently processing structured point cloud data while preserving multiscale information is a key challenge across domains, from graphics to atomistic modeling. Using a curated dataset of simulated galaxy positions and properties, represented as point clouds, we benchmark the ability of graph neural networks to simultaneously capture local clustering environments and long-range correlations. Given the homogeneous and isotropic nature of the Universe, the data exhibits a high degree of symmetry. We therefore focus on evaluating the performance of Euclidean symmetry-preserving ($E(3)$-equivariant) graph neural networks, showing that they can outperform non-equivariant counterparts and domain-specific information extraction techniques in downstream performance as well as simulation-efficiency. However, we find that current architectures fail to capture information from long-range correlations as effectively as domain-specific baselines, motivating future work on architectures better suited for extracting long-range information."
                    },
                    {
                        "title": "Diffusion models in protein structure and docking",
                        "abstract": "Generative AI is rapidly transforming the frontier of research in computational structural biology. Indeed, recent successes have substantially advanced protein design and drug discovery. One of the key methodologies underlying these advances is diffusion models (DM). Diffusion models originated in computer vision, rapidly taking over image generation and offering superior quality and performance. These models were subsequently extended and modified for uses in other areas including computational structural biology. DMs are well equipped to model high dimensional, geometric data while exploiting key strengths of deep learning. In structural biology, for example, they have achieved state\u2010of\u2010the\u2010art results on protein 3D structure generation and small molecule docking. This review covers the basics of diffusion models, associated modeling choices regarding molecular representations, generation capabilities, prevailing heuristics, as well as key limitations and forthcoming refinements. We also provide best practices around evaluation procedures to help establish rigorous benchmarking and evaluation. The review is intended to provide a fresh view into the state\u2010of\u2010the\u2010art as well as highlight its potentials and current challenges of recent generative techniques in computational structural biology."
                    },
                    {
                        "title": "DisCo-Diff: Enhancing Continuous Diffusion Models with Discrete Latents",
                        "abstract": "Diffusion models (DMs) have revolutionized generative learning. They utilize a diffusion process to encode data into a simple Gaussian distribution. However, encoding a complex, potentially multimodal data distribution into a single continuous Gaussian distribution arguably represents an unnecessarily challenging learning problem. We propose Discrete-Continuous Latent Variable Diffusion Models (DisCo-Diff) to simplify this task by introducing complementary discrete latent variables. We augment DMs with learnable discrete latents, inferred with an encoder, and train DM and encoder end-to-end. DisCo-Diff does not rely on pre-trained networks, making the framework universally applicable. The discrete latents significantly simplify learning the DM's complex noise-to-data mapping by reducing the curvature of the DM's generative ODE. An additional autoregressive transformer models the distribution of the discrete latents, a simple step because DisCo-Diff requires only few discrete variables with small codebooks. We validate DisCo-Diff on toy data, several image synthesis tasks as well as molecular docking, and find that introducing discrete latents consistently improves model performance. For example, DisCo-Diff achieves state-of-the-art FID scores on class-conditioned ImageNet-64/128 datasets with ODE sampler."
                    },
                    {
                        "title": "Thermodynamically Informed Multimodal Learning of High-Dimensional Free Energy Models in Molecular Coarse Graining",
                        "abstract": "We present a differentiable formalism for learning free energies that is capable of capturing arbitrarily complex model dependencies on coarse-grained coordinates and finite-temperature response to variation of general system parameters. This is done by endowing models with explicit dependence on temperature and parameters and by exploiting exact differential thermodynamic relationships between the free energy, ensemble averages, and response properties. Formally, we derive an approach for learning high-dimensional cumulant generating functions using statistical estimates of their derivatives, which are observable cumulants of the underlying random variable. The proposed formalism opens ways to resolve several outstanding challenges in bottom-up molecular coarse graining dealing with multiple minima and state dependence. This is realized by using additional differential relationships in the loss function to significantly improve the learning of free energies, while exactly preserving the Boltzmann distribution governing the corresponding fine-grain all-atom system. As an example, we go beyond the standard force-matching procedure to demonstrate how leveraging the thermodynamic relationship between free energy and values of ensemble averaged all-atom potential energy improves the learning efficiency and accuracy of the free energy model. The result is significantly better sampling statistics of structural distribution functions. The theoretical framework presented here is demonstrated via implementations in both kernel-based and neural network machine learning regression methods and opens new ways to train accurate machine learning models for studying thermodynamic and response properties of complex molecular systems."
                    },
                    {
                        "title": "Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models",
                        "abstract": "Approximations in computing model likelihoods with continuous normalizing flows (CNFs) hinder the use of these models for importance sampling of Boltzmann distributions, where exact likelihoods are required. In this work, we present Verlet flows, a class of CNFs on an augmented state-space inspired by symplectic integrators from Hamiltonian dynamics. When used with carefully constructed Taylor-Verlet integrators, Verlet flows provide exact-likelihood generative models which generalize coupled flow architectures from a non-continuous setting while imposing minimal expressivity constraints. On experiments over toy densities, we demonstrate that the variance of the commonly used Hutchinson trace estimator is unsuitable for importance sampling, whereas Verlet flows perform comparably to full autograd trace computations while being significantly faster."
                    },
                    {
                        "title": "Fictitious Synthetic Data Can Improve LLM Factuality via Prerequisite Learning",
                        "abstract": "Recent studies have identified one aggravating factor of LLM hallucinations as the knowledge inconsistency between pre-training and fine-tuning, where unfamiliar fine-tuning data mislead the LLM to fabricate plausible but wrong outputs. In this paper, we propose a novel fine-tuning strategy called Prereq-Tune to address this knowledge inconsistency and reduce hallucinations. Fundamentally, Prereq-Tune disentangles the learning of skills and knowledge, so the model learns only the task skills without being impacted by the knowledge inconsistency. To achieve this, Prereq-Tune introduces an additional prerequisite learning stage to learn the necessary knowledge for SFT, allowing subsequent SFT to focus only on task skills. Prereq-Tune can also be combined with fictitious synthetic data to enhance the grounding of LLM outputs to their internal knowledge. Experiments show that Prereq-Tune outperforms existing baselines in improving LLM's factuality across short QA and long-form generation tasks. It also opens new possibilities for knowledge-controlled generation in LLMs. Our code is available at https://github.com/UCSB-NLP-Chang/Prereq_tune.git."
                    },
                    {
                        "title": "AlphaFold Meets Flow Matching for Generating Protein Ensembles",
                        "abstract": "The biological functions of proteins often depend on dynamic structural ensembles. In this work, we develop a flow-based generative modeling approach for learning and sampling the conformational landscapes of proteins. We repurpose highly accurate single-state predictors such as AlphaFold and ESMFold and fine-tune them under a custom flow matching framework to obtain sequence-conditoned generative models of protein structure called AlphaFlow and ESMFlow. When trained and evaluated on the PDB, our method provides a superior combination of precision and diversity compared to AlphaFold with MSA subsampling. When further trained on ensembles from all-atom MD, our method accurately captures conformational flexibility, positional distributions, and higher-order ensemble observables for unseen proteins. Moreover, our method can diversify a static PDB structure with faster wall-clock convergence to certain equilibrium properties than replicate MD trajectories, demonstrating its potential as a proxy for expensive physics-based simulations. Code is available at https://github.com/bjing2016/alphaflow."
                    },
                    {
                        "title": "Generative Modeling of Molecular Dynamics Trajectories",
                        "abstract": "Molecular dynamics (MD) is a powerful technique for studying microscopic phenomena, but its computational cost has driven significant interest in the development of deep learning-based surrogate models. We introduce generative modeling of molecular trajectories as a paradigm for learning flexible multi-task surrogate models of MD from data. By conditioning on appropriately chosen frames of the trajectory, we show such generative models can be adapted to diverse tasks such as forward simulation, transition path sampling, and trajectory upsampling. By alternatively conditioning on part of the molecular system and inpainting the rest, we also demonstrate the first steps towards dynamics-conditioned molecular design. We validate the full set of these capabilities on tetrapeptide simulations and show that our model can produce reasonable ensembles of protein monomers. Altogether, our work illustrates how generative modeling can unlock value from MD data towards diverse downstream tasks that are not straightforward to address with existing methods or even MD itself. Code is available at https://github.com/bjing2016/mdgen."
                    },
                    {
                        "title": "Deep Confident Steps to New Pockets: Strategies for Docking Generalization",
                        "abstract": "Accurate blind docking has the potential to lead to new biological breakthroughs, but for this promise to be realized, docking methods must generalize well across the proteome. Existing benchmarks, however, fail to rigorously assess generalizability. Therefore, we develop DockGen, a new benchmark based on the ligand-binding domains of proteins, and we show that existing machine learning-based docking models have very weak generalization abilities. We carefully analyze the scaling laws of ML-based docking and show that, by scaling data and model size, as well as integrating synthetic data strategies, we are able to significantly increase the generalization capacity and set new state-of-the-art performance across benchmarks. Further, we propose Confidence Bootstrapping, a new training paradigm that solely relies on the interaction between diffusion and confidence models and exploits the multi-resolution generation process of diffusion models. We demonstrate that Confidence Bootstrapping significantly improves the ability of ML-based docking methods to dock to unseen protein classes, edging closer to accurate and generalizable blind docking methods."
                    },
                    {
                        "title": "In-Context Symmetries: Self-Supervised Learning through Contextual World Models",
                        "abstract": "At the core of self-supervised learning for vision is the idea of learning invariant or equivariant representations with respect to a set of data transformations. This approach, however, introduces strong inductive biases, which can render the representations fragile in downstream tasks that do not conform to these symmetries. In this work, drawing insights from world models, we propose to instead learn a general representation that can adapt to be invariant or equivariant to different transformations by paying attention to context -- a memory module that tracks task-specific states, actions, and future states. Here, the action is the transformation, while the current and future states respectively represent the input's representation before and after the transformation. Our proposed algorithm, Contextual Self-Supervised Learning (ContextSSL), learns equivariance to all transformations (as opposed to invariance). In this way, the model can learn to encode all relevant features as general representations while having the versatility to tail down to task-wise symmetries when given a few examples as the context. Empirically, we demonstrate significant performance gains over existing methods on equivariance-related tasks, supported by both qualitative and quantitative evaluations."
                    },
                    {
                        "title": "Sample, estimate, aggregate: A recipe for causal discovery foundation models",
                        "abstract": "Causal discovery, the task of inferring causal structure from data, promises to accelerate scientific research, inform policy making, and more. However, causal discovery algorithms over larger sets of variables tend to be brittle against misspecification or when data are limited. To mitigate these challenges, we train a supervised model that learns to predict a larger causal graph from the outputs of classical causal discovery algorithms run over subsets of variables, along with other statistical hints like inverse covariance. Our approach is enabled by the observation that typical errors in the outputs of classical methods remain comparable across datasets. Theoretically, we show that this model is well-specified, in the sense that it can recover a causal graph consistent with graphs over subsets. Empirically, we train the model to be robust to erroneous estimates using diverse synthetic data. Experiments on real and synthetic data demonstrate that this model maintains high accuracy in the face of misspecification or distribution shift, and can be adapted at low cost to different discovery algorithms or choice of statistics."
                    },
                    {
                        "title": "Generator Matching: Generative modeling with arbitrary Markov processes",
                        "abstract": "We introduce generator matching, a modality-agnostic framework for generative modeling using arbitrary Markov processes. Generators characterize the infinitesimal evolution of a Markov process, which we leverage for generative modeling in a similar vein to flow matching: we construct conditional generators which generate single data points, then learn to approximate the marginal generator which generates the full data distribution. We show that generator matching unifies various generative modeling methods, including diffusion models, flow matching and discrete diffusion models. Furthermore, it provides the foundation to expand the design space to new and unexplored Markov processes such as jump processes. Finally, generator matching enables the construction of superpositions of Markov generative processes and enables the construction of multimodal models in a rigorous manner. We empirically validate our method on protein and image structure generation, showing that superposition with a jump process improves image generation."
                    },
                    {
                        "title": "Generative Flows on Discrete State-Spaces: Enabling Multimodal Flows with Applications to Protein Co-Design",
                        "abstract": "Combining discrete and continuous data is an important capability for generative models. We present Discrete Flow Models (DFMs), a new flow-based model of discrete data that provides the missing link in enabling flow-based generative models to be applied to multimodal continuous and discrete data problems. Our key insight is that the discrete equivalent of continuous space flow matching can be realized using Continuous Time Markov Chains. DFMs benefit from a simple derivation that includes discrete diffusion models as a specific instance while allowing improved performance over existing diffusion-based approaches. We utilize our DFMs method to build a multimodal flow-based modeling framework. We apply this capability to the task of protein co-design, wherein we learn a model for jointly generating protein structure and sequence. Our approach achieves state-of-the-art co-design performance while allowing the same multimodal model to be used for flexible generation of the sequence or structure."
                    },
                    {
                        "title": "Revisiting Who's Harry Potter: Towards Targeted Unlearning from a Causal Intervention Perspective",
                        "abstract": "This paper investigates Who's Harry Potter (WHP), a pioneering yet insufficiently understood method for LLM unlearning. We explore it in two steps. First, we introduce a new task of LLM targeted unlearning, where given an unlearning target (e.g., a person) and some unlearning documents, we aim to unlearn only the information about the target, rather than everything in the unlearning documents. We further argue that a successful unlearning should satisfy criteria such as not outputting gibberish, not fabricating facts about the unlearning target, and not releasing factual information under jailbreak attacks. Second, we construct a causal intervention framework for targeted unlearning, where the knowledge of the unlearning target is modeled as a confounder between LLM input and output, and the unlearning process as a deconfounding process. This framework justifies and extends WHP, deriving a simple unlearning algorithm that includes WHP as a special case. Experiments on existing and new datasets show that our approach, without explicitly optimizing for the aforementioned criteria, achieves competitive performance in all of them. Our code is available at https://github.com/UCSB-NLP-Chang/causal_unlearn.git."
                    },
                    {
                        "title": "Fine-Tuning Discrete Diffusion Models via Reward Optimization with Applications to DNA and Protein Design",
                        "abstract": "Recent studies have demonstrated the strong empirical performance of diffusion models on discrete sequences across domains from natural language to biological sequence generation. For example, in the protein inverse folding task, conditional diffusion models have achieved impressive results in generating natural-like sequences that fold back into the original structure. However, practical design tasks often require not only modeling a conditional distribution but also optimizing specific task objectives. For instance, we may prefer protein sequences with high stability. To address this, we consider the scenario where we have pre-trained discrete diffusion models that can generate natural-like sequences, as well as reward models that map sequences to task objectives. We then formulate the reward maximization problem within discrete diffusion models, analogous to reinforcement learning (RL), while minimizing the KL divergence against pretrained diffusion models to preserve naturalness. To solve this RL problem, we propose a novel algorithm, DRAKES, that enables direct backpropagation of rewards through entire trajectories generated by diffusion models, by making the originally non-differentiable trajectories differentiable using the Gumbel-Softmax trick. Our theoretical analysis indicates that our approach can generate sequences that are both natural-like and yield high rewards. While similar tasks have been recently explored in diffusion models for continuous domains, our work addresses unique algorithmic and theoretical challenges specific to discrete diffusion models, which arise from their foundation in continuous-time Markov chains rather than Brownian motion. Finally, we demonstrate the effectiveness of DRAKES in generating DNA and protein sequences that optimize enhancer activity and protein stability, respectively, important tasks for gene therapies and protein-based therapeutics."
                    },
                    {
                        "title": "Hamiltonian Score Matching and Generative Flows",
                        "abstract": "Classical Hamiltonian mechanics has been widely used in machine learning in the form of Hamiltonian Monte Carlo for applications with predetermined force fields. In this work, we explore the potential of deliberately designing force fields for Hamiltonian ODEs, introducing Hamiltonian velocity predictors (HVPs) as a tool for score matching and generative models. We present two innovations constructed with HVPs: Hamiltonian Score Matching (HSM), which estimates score functions by augmenting data via Hamiltonian trajectories, and Hamiltonian Generative Flows (HGFs), a novel generative model that encompasses diffusion models and flow matching as HGFs with zero force fields. We showcase the extended design space of force fields by introducing Oscillation HGFs, a generative model inspired by harmonic oscillators. Our experiments validate our theoretical insights about HSM as a novel score matching metric and demonstrate that HGFs rival leading generative modeling techniques."
                    },
                    {
                        "title": "Dirichlet Flow Matching with Applications to DNA Sequence Design",
                        "abstract": "Discrete diffusion or flow models could enable faster and more controllable sequence generation than autoregressive models. We show that na\\\"ive linear flow matching on the simplex is insufficient toward this goal since it suffers from discontinuities in the training target and further pathologies. To overcome this, we develop Dirichlet flow matching on the simplex based on mixtures of Dirichlet distributions as probability paths. In this framework, we derive a connection between the mixtures' scores and the flow's vector field that allows for classifier and classifier-free guidance. Further, we provide distilled Dirichlet flow matching, which enables one-step sequence generation with minimal performance hits, resulting in $O(L)$ speedups compared to autoregressive models. On complex DNA sequence generation tasks, we demonstrate superior performance compared to all baselines in distributional metrics and in achieving desired design targets for generated sequences. Finally, we show that our classifier-free guidance approach improves unconditional generation and is effective for generating DNA that satisfies design targets. Code is available at https://github.com/HannesStark/dirichlet-flow-matching."
                    }
                ]
            }
        }
    },
    "2406.03689": {
        "paper_data": {
            "title": "Evaluating the World Model Implicit in a Generative Model",
            "url": "http://arxiv.org/abs/2406.03689v2",
            "arxiv_id": "2406.03689",
            "authors": [
                "Keyon Vafa",
                "Justin Y. Chen",
                "Jon Kleinberg",
                "Sendhil Mullainathan",
                "Ashesh Rambachan"
            ],
            "abstract": "Recent work suggests that large language models may implicitly learn world models. How should we assess this possibility? We formalize this question for the case where the underlying reality is governed by a deterministic finite automaton. This includes problems as diverse as simple logical reasoning, geographic navigation, game-playing, and chemistry. We propose new evaluation metrics for world model recovery inspired by the classic Myhill-Nerode theorem from language theory. We illustrate their utility in three domains: game playing, logic puzzles, and navigation. In all domains, the generative models we consider do well on existing diagnostics for assessing world models, but our evaluation metrics reveal their world models to be far less coherent than they appear. Such incoherence creates fragility: using a generative model to solve related but subtly different tasks can lead it to fail badly. Building generative models that meaningfully capture the underlying logic of the domains they model would be immensely valuable; our results suggest new ways to assess how close a given model is to that goal.",
            "introduction": "   1 Introduction  Large language models (LLMs) appear to have capacities that far exceed the next-token prediction task they were trained to perform [14, 35, 31]. Recent work suggests a reason: they are implicitly recovering high-fidelity representations of the underlying domains they are trained on [1, 17].   An algorithm that recovers a \u201cworld model\u201d from sequence data would be extremely valuable. As an example, consider how one might build a navigation tool today: meticulously map each street and intersection, and then use a search algorithm to provide directions. The success of language models suggests an alternative approach: collect turn-by-turn sequences from trips in a city (e.g. \u201cEast North\u2026\u201d) and then train a sequence model on them. If the sequence model successfully recovers the world model, we would obtain a map of the city without ever mapping it and a routing algorithm simply by predicting the next turn. This example is not far-fetched: it is the reason language models are used in scientific domains such as protein generation, genetics and chemistry [7, 18, 3, 11, 6].   All of this relies on the presumption that the sequence model has recovered the true world model; but how can we test whether it actually has? Answering this question requires first defining what we mean by the true world model. Toshniwal et\u00a0al. [32] and Li et\u00a0al. [17] proposed a concrete and influential approach: study whether sequence models trained on board game transcripts (e.g. chess and Othello) recover the underlying game rules. Inspired by this approach, we consider the case where the underlying world can be summarized by a finite collection of states and rules governing transitions between the states; this includes many domains such as logic [16], location tracking [24, 8], games [32, 17], and several of the scientific applications described above. As a result, the \u201cworld\u201d in these domains can be modeled as a deterministic finite automaton (DFA).   We show the difficulty in evaluating implicit world models. Consider an existing approach: for a given sequence, compare the next tokens outputted by the generative model to the set of valid next tokens for the state implied by that sequence [32, 17]. Though intuitive, this approach can fail to diagnose severe problems, and we illustrate this concretely. The classic Myhill-Nerode theorem [22, 23] provides intuition: every pair of distinct states can be distinguished by some sequence (admitted by one state but not the other). Unless those minimal distinguishing sequences are of length one, looking at the next single token outputted will not reliably assess whether the generative model has an accurate model of the underlying state.   The logic of Myhill-Nerode suggests two metrics for measuring whether a generative model effectively captures underlying states and transitions. The first metric summarizes sequence compression: under the DFA, sequences that lead to the same state must have the same continuations; so one can test whether the generative model has similar sequences of outputs when started on these two sequences. The second metric summarizes sequence distinction: under the DFA, two sequences that lead to distinct states should have distinct continuations; so one can test whether the generative model matches these distinct outputs when started at these two sequences.",
            "references": [
                {
                    "title": "The Illusion of State in State-Space Models",
                    "abstract": "State-space models (SSMs) have emerged as a potential alternative architecture for building large language models (LLMs) compared to the previously ubiquitous transformer architecture. One theoretical weakness of transformers is that they cannot express certain kinds of sequential computation and state tracking (Merrill&Sabharwal, 2023), which SSMs are explicitly designed to address via their close architectural similarity to recurrent neural networks (RNNs). But do SSMs truly have an advantage (over transformers) in expressive power for state tracking? Surprisingly, the answer is no. Our analysis reveals that the expressive power of SSMs is limited very similarly to transformers: SSMs cannot express computation outside the complexity class $\\mathsf{TC}^0$. In particular, this means they cannot solve simple state-tracking problems like permutation composition. It follows that SSMs are provably unable to accurately track chess moves with certain notation, evaluate code, or track entities in a long narrative. To supplement our formal analysis, we report experiments showing that Mamba-style SSMs indeed struggle with state tracking. Thus, despite its recurrent formulation, the\"state\"in an SSM is an illusion: SSMs have similar expressiveness limitations to non-recurrent models like transformers, which may fundamentally limit their ability to solve real-world state-tracking problems."
                },
                {
                    "title": "Linear Latent World Models in Simple Transformers: A Case Study on Othello-GPT",
                    "abstract": "Foundation models exhibit significant capabilities in decision-making and logical deductions. Nonetheless, a continuing discourse persists regarding their genuine understanding of the world as opposed to mere stochastic mimicry. This paper meticulously examines a simple transformer trained for Othello, extending prior research to enhance comprehension of the emergent world model of Othello-GPT. The investigation reveals that Othello-GPT encapsulates a linear representation of opposing pieces, a factor that causally steers its decision-making process. This paper further elucidates the interplay between the linear world representation and causal decision-making, and their dependence on layer depth and model complexity. We have made the code public."
                },
                {
                    "title": "Mistral 7B",
                    "abstract": "We introduce Mistral 7B v0.1, a 7-billion-parameter language model engineered for superior performance and efficiency. Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation. Our model leverages grouped-query attention (GQA) for faster inference, coupled with sliding window attention (SWA) to effectively handle sequences of arbitrary length with a reduced inference cost. We also provide a model fine-tuned to follow instructions, Mistral 7B -- Instruct, that surpasses the Llama 2 13B -- Chat model both on human and automated benchmarks. Our models are released under the Apache 2.0 license."
                },
                {
                    "title": "Qwen Technical Report",
                    "abstract": "Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural language processing tasks that were previously thought to be exclusive to humans. In this work, we introduce Qwen, the first installment of our large language model series. Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts. It includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques. The base language models consistently demonstrate superior performance across a multitude of downstream tasks, and the chat models, particularly those trained using Reinforcement Learning from Human Feedback (RLHF), are highly competitive. The chat models possess advanced tool-use and planning capabilities for creating agent applications, showcasing impressive performance even when compared to bigger models on complex tasks like utilizing a code interpreter. Furthermore, we have developed coding-specialized models, Code-Qwen and Code-Qwen-Chat, as well as mathematics-focused models, Math-Qwen-Chat, which are built upon base language models. These models demonstrate significantly improved performance in comparison with open-source models, and slightly fall behind the proprietary models."
                },
                {
                    "title": "Large Language Models on the Chessboard: A Study on ChatGPT's Formal Language Comprehension and Complex Reasoning Skills",
                    "abstract": "While large language models have made strides in natural language processing, their proficiency in complex reasoning tasks requiring formal language comprehension, such as chess, remains less investigated. This paper probes the performance of ChatGPT, a sophisticated language model by OpenAI in tackling such complex reasoning tasks, using chess as a case study. Through robust metrics examining both the legality and quality of moves, we assess ChatGPT's understanding of the chessboard, adherence to chess rules, and strategic decision-making abilities. Our evaluation identifies limitations within ChatGPT's attention mechanism that affect its formal language comprehension and uncovers the model's underdeveloped self-regulation abilities. Our study also reveals ChatGPT's propensity for a coherent strategy in its gameplay and a noticeable uptick in decision-making assertiveness when the model is presented with a greater volume of natural language or possesses a more lucid understanding of the state of the chessboard. These findings contribute to the growing exploration of language models' abilities beyond natural language processing, providing valuable information for future research towards models demonstrating human-like cognitive abilities."
                },
                {
                    "title": "DNA language models are powerful predictors of genome-wide variant effects",
                    "abstract": "The expanding catalog of genome-wide association studies (GWAS) provides biological insights across a variety of species, but identifying the causal variants behind these associations remains a significant challenge. Experimental validation is both labor-intensive and costly, highlighting the need for accurate, scalable computational methods to predict the effects of genetic variants across the entire genome. Inspired by recent progress in natural language processing, unsupervised pre-training on large protein sequence databases has proven successful in extracting complex information related to proteins. These models showcase their ability to learn variant effects in coding regions using an unsupervised approach. Expanding on this idea, we here introduce the Genomic Pre-trained Network (GPN), a model designed to learn genome-wide variant effects through unsupervised pre-training on genomic DNA sequences. Our model also successfully learns gene structure and DNA motifs without any supervision. To demonstrate its utility, we train GPN on unaligned reference genomes of Arabidopsis thaliana and seven related species within the Brassicales order, and evaluate its ability to predict the functional impact of genetic variants in Arabidopsis thaliana by utilizing allele frequencies from the 1001 Genomes Project and a comprehensive database of GWAS. Notably, GPN outperforms predictors based on popular conservation scores such as phyloP and phastCons. Our predictions for Arabidopsis thaliana can be visualized as sequence logos in the UCSC Genome Browser (https://genome.ucsc.edu/s/gbenegas/gpn-arabidopsis). We provide code (https://github.com/songlab-cal/gpn) to train GPN for any given species using its DNA sequence alone, enabling unsupervised prediction of variant effects across the entire genome."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "VELMA: Verbalization Embodiment of LLM Agents for Vision and Language Navigation in Street View",
                    "abstract": "Incremental decision making in real-world environments is one of the most challenging tasks in embodied artificial intelligence. One particularly demanding scenario is Vision and Language Navigation (VLN) which requires visual and natural language understanding as well as spatial and temporal reasoning capabilities. The embodied agent needs to ground its understanding of navigation instructions in observations of a real-world environment like Street View. Despite the impressive results of LLMs in other research areas, it is an ongoing problem of how to best connect them with an interactive visual environment. In this work, we propose VELMA, an embodied LLM agent that uses a verbalization of the trajectory and of visual environment observations as contextual prompt for the next action. Visual information is verbalized by a pipeline that extracts landmarks from the human written navigation instructions and uses CLIP to determine their visibility in the current panorama view. We show that VELMA is able to successfully follow navigation instructions in Street View with only two in-context examples. We further finetune the LLM agent on a few thousand examples and achieve around 25% relative improvement in task completion over the previous state-of-the-art for two datasets."
                },
                {
                    "title": "Leveraging Pre-trained Large Language Models to Construct and Utilize World Models for Model-based Task Planning",
                    "abstract": "There is a growing interest in applying pre-trained large language models (LLMs) to planning problems. However, methods that use LLMs directly as planners are currently impractical due to several factors, including limited correctness of plans, strong reliance on feedback from interactions with simulators or even the actual environment, and the inefficiency in utilizing human feedback. In this work, we introduce a novel alternative paradigm that constructs an explicit world (domain) model in planning domain definition language (PDDL) and then uses it to plan with sound domain-independent planners. To address the fact that LLMs may not generate a fully functional PDDL model initially, we employ LLMs as an interface between PDDL and sources of corrective feedback, such as PDDL validators and humans. For users who lack a background in PDDL, we show that LLMs can translate PDDL into natural language and effectively encode corrective feedback back to the underlying domain model. Our framework not only enjoys the correctness guarantee offered by the external planners but also reduces human involvement by allowing users to correct domain models at the beginning, rather than inspecting and correcting (through interactive prompting) every generated plan as in previous work. On two IPC domains and a Household domain that is more complicated than commonly used benchmarks such as ALFWorld, we demonstrate that GPT-4 can be leveraged to produce high-quality PDDL models for over 40 actions, and the corrected PDDL models are then used to successfully solve 48 challenging planning tasks. Resources, including the source code, are released at: https://guansuns.github.io/pages/llm-dm."
                },
                {
                    "title": "Emergent Representations of Program Semantics in Language Models Trained on Programs",
                    "abstract": "We present evidence that language models (LMs) of code can learn to represent the formal semantics of programs, despite being trained only to perform next-token prediction. Specifically, we train a Transformer model on a synthetic corpus of programs written in a domain-specific language for navigating 2D grid world environments. Each program in the corpus is preceded by a (partial) specification in the form of several input-output grid world states. Despite providing no further inductive biases, we find that a probing classifier is able to extract increasingly accurate representations of the unobserved, intermediate grid world states from the LM hidden states over the course of training, suggesting the LM acquires an emergent ability to interpret programs in the formal sense. We also develop a novel interventional baseline that enables us to disambiguate what is represented by the LM as opposed to learned by the probe. We anticipate that this technique may be generally applicable to a broad range of semantic probing experiments. In summary, this paper does not propose any new techniques for training LMs of code, but develops an experimental framework for and provides insights into the acquisition and representation of formal semantics in statistical models of code. Our code is available at https://github.com/charlesjin/emergent-semantics."
                },
                {
                    "title": "Causal Reasoning and Large Language Models: Opening a New Frontier for Causality",
                    "abstract": "The causal capabilities of large language models (LLMs) are a matter of significant debate, with critical implications for the use of LLMs in societally impactful domains such as medicine, science, law, and policy. We conduct a\"behavorial\"study of LLMs to benchmark their capability in generating causal arguments. Across a wide range of tasks, we find that LLMs can generate text corresponding to correct causal arguments with high probability, surpassing the best-performing existing methods. Algorithms based on GPT-3.5 and 4 outperform existing algorithms on a pairwise causal discovery task (97%, 13 points gain), counterfactual reasoning task (92%, 20 points gain) and event causality (86% accuracy in determining necessary and sufficient causes in vignettes). We perform robustness checks across tasks and show that the capabilities cannot be explained by dataset memorization alone, especially since LLMs generalize to novel datasets that were created after the training cutoff date. That said, LLMs exhibit unpredictable failure modes, and we discuss the kinds of errors that may be improved and what are the fundamental limits of LLM-based answers. Overall, by operating on the text metadata, LLMs bring capabilities so far understood to be restricted to humans, such as using collected knowledge to generate causal graphs or identifying background causal context from natural language. As a result, LLMs may be used by human domain experts to save effort in setting up a causal analysis, one of the biggest impediments to the widespread adoption of causal methods. Given that LLMs ignore the actual data, our results also point to a fruitful research direction of developing algorithms that combine LLMs with existing causal techniques. Code and datasets are available at https://github.com/py-why/pywhy-llm."
                },
                {
                    "title": "Evolutionary-scale prediction of atomic level protein structure with a language model",
                    "abstract": "Artificial intelligence has the potential to open insight into the structure of proteins at the scale of evolution. It has only recently been possible to extend protein structure prediction to two hundred million cataloged proteins. Characterizing the structures of the exponentially growing billions of protein sequences revealed by large scale gene sequencing experiments would necessitate a break-through in the speed of folding. Here we show that direct inference of structure from primary sequence using a large language model enables an order of magnitude speed-up in high resolution structure prediction. Leveraging the insight that language models learn evolutionary patterns across millions of sequences, we train models up to 15B parameters, the largest language model of proteins to date. As the language models are scaled they learn information that enables prediction of the three-dimensional structure of a protein at the resolution of individual atoms. This results in prediction that is up to 60x faster than state-of-the-art while maintaining resolution and accuracy. Building on this, we present the ESM Metage-nomic Atlas. This is the first large-scale structural characterization of metagenomic proteins, with more than 617 million structures. The atlas reveals more than 225 million high confidence predictions, including millions whose structures are novel in comparison with experimentally determined structures, giving an unprecedented view into the vast breadth and diversity of the structures of some of the least understood proteins on earth."
                },
                {
                    "title": "Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task",
                    "abstract": "Language models show a surprising range of capabilities, but the source of their apparent competence is unclear. Do these networks just memorize a collection of surface statistics, or do they rely on internal representations of the process that generates the sequences they see? We investigate this question by applying a variant of the GPT model to the task of predicting legal moves in a simple board game, Othello. Although the network has no a priori knowledge of the game or its rules, we uncover evidence of an emergent nonlinear internal representation of the board state. Interventional experiments indicate this representation can be used to control the output of the network and create\"latent saliency maps\"that can help explain predictions in human terms."
                },
                {
                    "title": "Transformers Learn Shortcuts to Automata",
                    "abstract": "Algorithmic reasoning requires capabilities which are most naturally understood through recurrent models of computation, like the Turing machine. However, Transformer models, while lacking recurrence, are able to perform such reasoning using far fewer layers than the number of reasoning steps. This raises the question: what solutions are learned by these shallow and non-recurrent models? We find that a low-depth Transformer can represent the computations of any finite-state automaton (thus, any bounded-memory algorithm), by hierarchically reparameterizing its recurrent dynamics. Our theoretical results characterize shortcut solutions, whereby a Transformer with $o(T)$ layers can exactly replicate the computation of an automaton on an input sequence of length $T$. We find that polynomial-sized $O(\\log T)$-depth solutions always exist; furthermore, $O(1)$-depth simulators are surprisingly common, and can be understood using tools from Krohn-Rhodes theory and circuit complexity. Empirically, we perform synthetic experiments by training Transformers to simulate a wide variety of automata, and show that shortcut solutions can be learned via standard training. We further investigate the brittleness of these solutions and propose potential mitigations."
                },
                {
                    "title": "Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them",
                    "abstract": "BIG-Bench (Srivastava et al., 2022) is a diverse evaluation suite that focuses on tasks believed to be beyond the capabilities of current language models. Language models have already made good progress on this benchmark, with the best model in the BIG-Bench paper outperforming average reported human-rater results on 65% of the BIG-Bench tasks via few-shot prompting. But on what tasks do language models fall short of average human-rater performance, and are those tasks actually unsolvable by current language models? In this work, we focus on a suite of 23 challenging BIG-Bench tasks which we call BIG-Bench Hard (BBH). These are the task for which prior language model evaluations did not outperform the average human-rater. We find that applying chain-of-thought (CoT) prompting to BBH tasks enables PaLM to surpass the average human-rater performance on 10 of the 23 tasks, and Codex (code-davinci-002) to surpass the average human-rater performance on 17 of the 23 tasks. Since many tasks in BBH require multi-step reasoning, few-shot prompting without CoT, as done in the BIG-Bench evaluations (Srivastava et al., 2022), substantially underestimates the best performance and capabilities of language models, which is better captured via CoT prompting. As further analysis, we explore the interaction between CoT and model scale on BBH, finding that CoT enables emergent task performance on several BBH tasks with otherwise flat scaling curves."
                },
                {
                    "title": "The Parallelism Tradeoff: Limitations of Log-Precision Transformers",
                    "abstract": "Despite their omnipresence in modern NLP, characterizing the computational power of transformer neural nets remains an interesting open question. We prove that transformers whose arithmetic precision is logarithmic in the number of input tokens (and whose feedforward nets are computable using space linear in their input) can be simulated by constant-depth logspace-uniform threshold circuits. This provides insight on the power of transformers using known results in complexity theory. For example, if L\u2260P (i.e., not all poly-time problems can be solved using logarithmic space), then transformers cannot even accurately solve linear equalities or check membership in an arbitrary context-free grammar with empty productions. Our result intuitively emerges from the transformer architecture\u2019s high parallelizability. We thus speculatively introduce the idea of a fundamental parallelism tradeoff: any model architecture as parallelizable as the transformer will obey limitations similar to it. Since parallelism is key to training models at massive scale, this suggests a potential inherent weakness of the scaling paradigm."
                },
                {
                    "title": "Emergent Abilities of Large Language Models",
                    "abstract": "Scaling up language models has been shown to predictably improve performance and sample efficiency on a wide range of downstream tasks. This paper instead discusses an unpredictable phenomenon that we refer to as emergent abilities of large language models. We consider an ability to be emergent if it is not present in smaller models but is present in larger models. Thus, emergent abilities cannot be predicted simply by extrapolating the performance of smaller models. The existence of such emergence implies that additional scaling could further expand the range of capabilities of language models."
                },
                {
                    "title": "Analyzing Generalization of Vision and Language Navigation to Unseen Outdoor Areas",
                    "abstract": "Vision and language navigation (VLN) is a challenging visually-grounded language understanding task. Given a natural language navigation instruction, a visual agent interacts with a graph-based environment equipped with panorama images and tries to follow the described route. Most prior work has been conducted in indoor scenarios where best results were obtained for navigation on routes that are similar to the training routes, with sharp drops in performance when testing on unseen environments. We focus on VLN in outdoor scenarios and find that in contrast to indoor VLN, most of the gain in outdoor VLN on unseen data is due to features like junction type embedding or heading delta that are specific to the respective environment graph, while image information plays a very minor role in generalizing VLN to unseen outdoor areas. These findings show a bias to specifics of graph representations of urban environments, demanding that VLN tasks grow in scale and diversity of geographical environments."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "Can Language Models Encode Perceptual Structure Without Grounding? A Case Study in Color",
                    "abstract": "Pretrained language models have been shown to encode relational information, such as the relations between entities or concepts in knowledge-bases \u2014 (Paris, Capital, France). However, simple relations of this type can often be recovered heuristically and the extent to which models implicitly reflect topological structure that is grounded in world, such as perceptual structure, is unknown. To explore this question, we conduct a thorough case study on color. Namely, we employ a dataset of monolexemic color terms and color chips represented in CIELAB, a color space with a perceptually meaningful distance metric. Using two methods of evaluating the structural alignment of colors in this space with text-derived color term representations, we find significant correspondence. Analyzing the differences in alignment across the color spectrum, we find that warmer colors are, on average, better aligned to the perceptual color space than cooler ones, suggesting an intriguing connection to findings from recent work on efficient communication in color naming. Further analysis suggests that differences in alignment are, in part, mediated by collocationality and differences in syntactic usage, posing questions as to the relationship between color perception and usage and context."
                },
                {
                    "title": "Implicit Representations of Meaning in Neural Language Models",
                    "abstract": "Does the effectiveness of neural language models derive entirely from accurate modeling of surface word co-occurrence statistics, or do these models represent and reason about the world they describe? In BART and T5 transformer language models, we identify contextual word representations that function as *models of entities and situations* as they evolve throughout a discourse. These neural representations have functional similarities to linguistic models of dynamic semantics: they support a linear readout of each entity\u2019s current properties and relations, and can be manipulated with predictable effects on language generation. Our results indicate that prediction in pretrained neural language models is supported, at least in part, by dynamic representations of meaning and implicit simulation of entity state, and that this behavior can be learned with only text as training data."
                },
                {
                    "title": "Chess as a Testbed for Language Model State Tracking",
                    "abstract": "Transformer language models have made tremendous strides in natural language understanding tasks. However, the complexity of natural language makes it challenging to ascertain how accurately these models are tracking the world state underlying the text. Motivated by this issue, we consider the task of language modeling for the game of chess. Unlike natural language, chess notations describe a simple, constrained, and deterministic domain. Moreover, we observe that the appropriate choice of chess notation allows for directly probing the world state, without requiring any additional probing-related machinery. We find that: (a) With enough training data, transformer language models can learn to track pieces and predict legal moves with high accuracy when trained solely on move sequences. (b) For small training sets providing access to board state information during training can yield significant improvements. (c) The success of transformer language models is dependent on access to the entire game history i.e. \u201cfull attention\u201d. Approximating this full attention results in a significant performance drop. We propose this testbed as a benchmark for future work on the development and analysis of transformer language models."
                },
                {
                    "title": "Generating Landmark Navigation Instructions from Maps as a Graph-to-Text Problem",
                    "abstract": "Car-focused navigation services are based on turns and distances of named streets, whereas navigation instructions naturally used by humans are centered around physical objects called landmarks. We present a neural model that takes OpenStreetMap representations as input and learns to generate navigation instructions that contain visible and salient landmarks from human natural language instructions. Routes on the map are encoded in a location- and rotation-invariant graph representation that is decoded into natural language instructions. Our work is based on a novel dataset of 7,672 crowd-sourced instances that have been verified by human navigation in Street View. Our evaluation shows that the navigation instructions generated by our system have similar properties as human-generated instructions, and lead to successful human navigation in Street View."
                },
                {
                    "title": "On the Ability and Limitations of Transformers to Recognize Formal Languages",
                    "abstract": "Transformers have supplanted recurrent models in a large number of NLP tasks. However, the differences in their abilities to model different syntactic properties remain largely unknown. Past works suggest that LSTMs generalize very well on regular languages and have close connections with counter languages. In this work, we systematically study the ability of Transformers to model such languages as well as the role of its individual components in doing so. We first provide a construction of Transformers for a subclass of counter languages, including well-studied languages such as n-ary Boolean Expressions, Dyck-1, and its generalizations. In experiments, we find that Transformers do well on this subclass, and their learned mechanism strongly correlates with our construction. Perhaps surprisingly, in contrast to LSTMs, Transformers do well only on a subset of regular languages with degrading performance as we make languages more complex according to a well-known measure of complexity. Our analysis also provides insights on the role of self-attention mechanism in modeling certain behaviors and the influence of positional encoding schemes on the learning and generalization abilities of the model."
                },
                {
                    "title": "Designing and Interpreting Probes with Control Tasks",
                    "abstract": "Probes, supervised models trained to predict properties (like parts-of-speech) from representations (like ELMo), have achieved high accuracy on a range of linguistic tasks. But does this mean that the representations encode linguistic structure or just that the probe has learned the linguistic task? In this paper, we propose control tasks, which associate word types with random outputs, to complement linguistic tasks. By construction, these tasks can only be learned by the probe itself. So a good probe, (one that reflects the representation), should be selective, achieving high linguistic task accuracy and low control task accuracy. The selectivity of a probe puts linguistic task accuracy in context with the probe\u2019s capacity to memorize from word types. We construct control tasks for English part-of-speech tagging and dependency edge prediction, and show that popular probes on ELMo representations are not selective. We also find that dropout, commonly used to control probe complexity, is ineffective for improving selectivity of MLPs, but that other forms of regularization are effective. Finally, we find that while probes on the first layer of ELMo yield slightly better part-of-speech tagging accuracy than the second, probes on the second layer are substantially more selective, which raises the question of which layer better represents parts-of-speech."
                },
                {
                    "title": "On Evaluating the Generalization of LSTM Models in Formal Languages",
                    "abstract": "Recurrent Neural Networks (RNNs) are theoretically Turing-complete and established themselves as a dominant model for language processing. Yet, there still remains an uncertainty regarding their language learning capabilities. In this paper, we empirically evaluate the inductive learning capabilities of Long Short-Term Memory networks, a popular extension of simple RNNs, to learn simple formal languages, in particular $a^nb^n$, $a^nb^nc^n$, and $a^nb^nc^nd^n$. We investigate the influence of various aspects of learning, such as training data regimes and model capacity, on the generalization to unobserved samples. We find striking differences in model performances under different training settings and highlight the need for careful analysis and assessment when making claims about the learning capabilities of neural network models."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Linear automaton transformations",
                    "abstract": "Let R be a nonempty set, let N consist of all non-negative rational integers, and denote by RN the set of all functions on N to R. If R is a ring, a map M: R\"\u2014>P^ is linear if M(rxfx+r2f2)=rx(Mfx) +r2(Mf2) for rx, r2 in R, fx, f2 in RN. For a finite commutative ring with unit we determine which linear transformations M: RN\u2014+RN can be realized by finite automata. More precisely, let A, B he finite nonempty sets. A map M: AN\u2014>BN is an automaton transformation if there exists a finite set Q, maps Mq: A X\u00a3>\u2014><2, Mb: A XQ-*B, elements h in B, q in Q such that corresponding to each/ in AN there exists an h in QN satisfying"
                },
                {
                    "title": "Mapping Language Models to Grounded Conceptual Spaces",
                    "abstract": "A fundamental criticism of text-only language models (LMs) is their lack of grounding \u2014that is, the ability to tie a word for which they have learned a representation to its referent in the non-linguistic world. However, despite this limitation, large pre-trained LMs have been shown to have a remarkable grasp of the conceptual structure of language, as demonstrated by their ability to answer questions, generate \ufb02uent text, or make inferences about entities, objects, and properties that they have never physically observed. In this work we investigate the extent to which the rich conceptual structure that LMs learn indeed re\ufb02ects the conceptual structure of the non-linguistic world\u2014which is something that LMs have never observed. We do this by testing whether the LMs can learn to map an entire conceptual domain (e.g., direction or colour) onto a grounded world representation given only a small number of examples. For example, we show a model what the word \u201cleft\u201d means using a textual depiction of a grid world, and assess how well it can generalise to related concepts, for example, the word \u201cright\u201d , in a similar grid world. We investigate a range of generative language models of varying sizes (including GPT-2 and GPT-3), and see that although the smaller models struggle to perform this mapping, the largest model can not only learn to ground the concepts that it is explicitly taught, but appears to generalise to several instances of unseen concepts as well. Our results suggest an alternative means of building grounded language models: rather than learning grounded representations \u201cfrom scratch\u201d, it is possible that large text-only models learn a suf\ufb01ciently rich conceptual structure that could allow them to be grounded in a data-ef\ufb01cient way."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively evaluate whether sequence models, such as large language models, accurately recover the underlying world model represented by a deterministic finite automaton (DFA) from sequence data?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the fundamental question of how well sequence models understand and represent the domains they are trained on. By establishing reliable evaluation metrics, we can enhance the interpretability and reliability of these models, leading to advancements in various applications, including navigation tools, game AI, and scientific domains like protein generation and genetics. This research could pave the way for future studies focused on improving model architectures and training methodologies, ultimately leading to more robust and capable AI systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of accurately assessing the implicit world models learned by sequence models. Naive approaches, such as comparing the next token predictions to valid next tokens, may fail to capture the nuances of state transitions and can overlook significant discrepancies in model performance. The theoretical obstacle presented by the Myhill-Nerode theorem indicates that distinguishing between states often requires sequences longer than a single token, complicating the evaluation process. Additionally, developing metrics that can effectively measure both sequence compression and distinction under the DFA framework adds further complexity to the problem.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the performance of sequence models in terms of next-token prediction without adequately addressing the evaluation of their underlying world models. Existing solutions have not provided a comprehensive framework for assessing the accuracy of state recovery and transition modeling. Barriers include a lack of clear metrics that align with the theoretical foundations of automata theory and the difficulty in designing experiments that can effectively test these metrics. Our approach differs by leveraging the Myhill-Nerode theorem to propose specific metrics for sequence compression and distinction, providing a more rigorous evaluation framework.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing two key metrics based on the Myhill-Nerode theorem: one for sequence compression, which tests whether sequences leading to the same state produce similar continuations, and another for sequence distinction, which assesses whether sequences leading to distinct states yield distinct continuations. We will apply these metrics to evaluate generative models trained on various datasets, including board game transcripts and other sequence data. The"
            }
        },
        "author_data": {
            "ba8e4d22-7790-491f-825c-057b4970e44a": {
                "pk": "ba8e4d22-7790-491f-825c-057b4970e44a",
                "name": "Keyon Vafa",
                "collaborators": [
                    "David M. Blei",
                    "Susan Athey",
                    "Sendhil Mullainathan",
                    "Carolina Zheng",
                    "Tianyu Du",
                    "Ayush Kanodia",
                    "Suresh Naidu",
                    "Ashesh Rambachan",
                    "Herman Brunborg",
                    "Yuntian Deng"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Social Equity",
                    "Machine Learning",
                    "Econometrics"
                ],
                "publications": [
                    {
                        "title": "Text-Based Ideal Points",
                        "abstract": "Ideal point models analyze lawmakers' votes to quantify their political positions, or ideal points. But votes are not the only way to express a political position. Lawmakers also give speeches, release press statements, and post tweets. In this paper, we introduce the text-based ideal point model (TBIP), an unsupervised probabilistic topic model that analyzes texts to quantify the political positions of its authors. We demonstrate the TBIP with two types of politicized text data: U.S. Senate speeches and senator tweets. Though the model does not analyze their votes or political affiliations, the TBIP separates lawmakers by party, learns interpretable politicized topics, and infers ideal points close to the classical vote-based ideal points. One benefit of analyzing texts, as opposed to votes, is that the TBIP can estimate ideal points of anyone who authors political texts, including non-voting actors. To this end, we use it to study tweets from the 2020 Democratic presidential candidates. Using only the texts of their tweets, it identifies them along an interpretable progressive-to-moderate spectrum."
                    },
                    {
                        "title": "Do Large Language Models Perform the Way People Expect? Measuring the Human Generalization Function",
                        "abstract": "What makes large language models (LLMs) impressive is also what makes them hard to evaluate: their diversity of uses. To evaluate these models, we must understand the purposes they will be used for. We consider a setting where these deployment decisions are made by people, and in particular, people's beliefs about where an LLM will perform well. We model such beliefs as the consequence of a human generalization function: having seen what an LLM gets right or wrong, people generalize to where else it might succeed. We collect a dataset of 19K examples of how humans make generalizations across 79 tasks from the MMLU and BIG-Bench benchmarks. We show that the human generalization function can be predicted using NLP methods: people have consistent structured ways to generalize. We then evaluate LLM alignment with the human generalization function. Our results show that -- especially for cases where the cost of mistakes is high -- more capable models (e.g. GPT-4) can do worse on the instances people choose to use them for, exactly because they are not aligned with the human generalization function."
                    },
                    {
                        "title": "Estimating Wage Disparities Using Foundation Models",
                        "abstract": "One thread of empirical work in social science focuses on decomposing group differences in outcomes into unexplained components and components explained by observable factors. In this paper, we study gender wage decompositions, which require estimating the portion of the gender wage gap explained by career histories of workers. Classical methods for decomposing the wage gap employ simple predictive models of wages which condition on a small set of simple summaries of labor history. The problem is that these predictive models cannot take advantage of the full complexity of a worker's history, and the resulting decompositions thus suffer from omitted variable bias (OVB), where covariates that are correlated with both gender and wages are not included in the model. Here we explore an alternative methodology for wage gap decomposition that employs powerful foundation models, such as large language models, as the predictive engine. Foundation models excel at making accurate predictions from complex, high-dimensional inputs. We use a custom-built foundation model, designed to predict wages from full labor histories, to decompose the gender wage gap. We prove that the way such models are usually trained might still lead to OVB, but develop fine-tuning algorithms that empirically mitigate this issue. Our model captures a richer representation of career history than simple models and predicts wages more accurately. In detail, we first provide a novel set of conditions under which an estimator of the wage gap based on a fine-tuned foundation model is $\\sqrt{n}$-consistent. Building on the theory, we then propose methods for fine-tuning foundation models that minimize OVB. Using data from the Panel Study of Income Dynamics, we find that history explains more of the gender wage gap than standard econometric models can measure, and we identify elements of history that are important for reducing OVB."
                    },
                    {
                        "title": "Revisiting Topic-Guided Language Models",
                        "abstract": "A recent line of work in natural language processing has aimed to combine language models and topic models. These topic-guided language models augment neural language models with topic models, unsupervised learning methods that can discover document-level patterns of word use. This paper compares the effectiveness of these methods in a standardized setting. We study four topic-guided language models and two baselines, evaluating the held-out predictive performance of each model on four corpora. Surprisingly, we find that none of these methods outperform a standard LSTM language model baseline, and most fail to learn good topics. Further, we train a probe of the neural language model that shows that the baseline's hidden states already encode topic information. We make public all code used for this study."
                    },
                    {
                        "title": "LABOR-LLM: Language-Based Occupational Representations with Large Language Models",
                        "abstract": "Many empirical studies of labor market questions rely on estimating relatively simple predictive models using small, carefully constructed longitudinal survey datasets based on hand-engineered features. Large Language Models (LLMs), trained on massive datasets, encode vast quantities of world knowledge and can be used for the next job prediction problem. However, while an off-the-shelf LLM produces plausible career trajectories when prompted, the probability with which an LLM predicts a particular job transition conditional on career history will not, in general, align with the true conditional probability in a given population. Recently, Vafa et al. (2024) introduced a transformer-based \"foundation model\", CAREER, trained using a large, unrepresentative resume dataset, that predicts transitions between jobs; it further demonstrated how transfer learning techniques can be used to leverage the foundation model to build better predictive models of both transitions and wages that reflect conditional transition probabilities found in nationally representative survey datasets. This paper considers an alternative where the fine-tuning of the CAREER foundation model is replaced by fine-tuning LLMs. For the task of next job prediction, we demonstrate that models trained with our approach outperform several alternatives in terms of predictive performance on the survey data, including traditional econometric models, CAREER, and LLMs with in-context learning, even though the LLM can in principle predict job titles that are not allowed in the survey data. Further, we show that our fine-tuned LLM-based models' predictions are more representative of the career trajectories of various workforce subpopulations than off-the-shelf LLM models and CAREER. We conduct experiments and analyses that highlight the sources of the gains in the performance of our models for representative predictions."
                    },
                    {
                        "title": "Rationales for Sequential Predictions",
                        "abstract": "Sequence models are a critical component of modern NLP systems, but their predictions are difficult to explain. We consider model explanations though rationales, subsets of context that can explain individual model predictions. We find sequential rationales by solving a combinatorial optimization: the best rationale is the smallest subset of input tokens that would predict the same output as the full sequence. Enumerating all subsets is intractable, so we propose an efficient greedy algorithm to approximate this objective. The algorithm, which is called greedy rationalization, applies to any model. For this approach to be effective, the model should form compatible conditional distributions when making predictions on incomplete subsets of the context. This condition can be enforced with a short fine-tuning step. We study greedy rationalization on language modeling and machine translation. Compared to existing baselines, greedy rationalization is best at optimizing the combinatorial objective and provides the most faithful rationales. On a new dataset of annotated sequential rationales, greedy rationales are most similar to human rationales."
                    },
                    {
                        "title": "CAREER: A Foundation Model for Labor Sequence Data",
                        "abstract": "Labor economists regularly analyze employment data by fitting predictive models to small, carefully constructed longitudinal survey datasets. Although machine learning methods offer promise for such problems, these survey datasets are too small to take advantage of them. In recent years large datasets of online resumes have also become available, providing data about the career trajectories of millions of individuals. However, standard econometric models cannot take advantage of their scale or incorporate them into the analysis of survey data. To this end we develop CAREER, a foundation model for job sequences. CAREER is first fit to large, passively-collected resume data and then fine-tuned to smaller, better-curated datasets for economic inferences. We fit CAREER to a dataset of 24 million job sequences from resumes, and adjust it on small longitudinal survey datasets. We find that CAREER forms accurate predictions of job sequences, outperforming econometric baselines on three widely-used economics datasets. We further find that CAREER can be used to form good predictions of other downstream variables. For example, incorporating CAREER into a wage model provides better predictions than the econometric models currently in use."
                    },
                    {
                        "title": "Discrete Flows: Invertible Generative Models of Discrete Data",
                        "abstract": "While normalizing flows have led to significant advances in modeling high-dimensional continuous distributions, their applicability to discrete distributions remains unknown. In this paper, we show that flows can in fact be extended to discrete events---and under a simple change-of-variables formula not requiring log-determinant-Jacobian computations. Discrete flows have numerous applications. We consider two flow architectures: discrete autoregressive flows that enable bidirectionality, allowing, for example, tokens in text to depend on both left-to-right and right-to-left contexts in an exact language model; and discrete bipartite flows that enable efficient non-autoregressive generation as in RealNVP. Empirically, we find that discrete autoregressive flows outperform autoregressive baselines on synthetic discrete distributions, an addition task, and Potts models; and bipartite flows can obtain competitive performance with autoregressive baselines on character-level language modeling for Penn Tree Bank and text8."
                    },
                    {
                        "title": "An Invariant Learning Characterization of Controlled Text Generation",
                        "abstract": "Controlled generation refers to the problem of creating text that contains stylistic or semantic attributes of interest. Many approaches reduce this problem to training a predictor of the desired attribute. For example, researchers hoping to deploy a large language model to produce non-toxic content may use a toxicity classifier to filter generated text. In practice, the generated text to classify, which is determined by user prompts, may come from a wide range of distributions. In this paper, we show that the performance of controlled generation may be poor if the distributions of text in response to user prompts differ from the distribution the predictor was trained on. To address this problem, we cast controlled generation under distribution shift as an invariant learning problem: the most effective predictor should be invariant across multiple text environments. We then discuss a natural solution that arises from this characterization and propose heuristics for selecting natural environments. We study this characterization and the proposed method empirically using both synthetic and real data. Experiments demonstrate both the challenge of distribution shift in controlled generation and the potential of invariance methods in this setting."
                    },
                    {
                        "title": "Use large language models to promote equity",
                        "abstract": "Advances in large language models (LLMs) have driven an explosion of interest about their societal impacts. Much of the discourse around how they will impact social equity has been cautionary or negative, focusing on questions like \"how might LLMs be biased and how would we mitigate those biases?\" This is a vital discussion: the ways in which AI generally, and LLMs specifically, can entrench biases have been well-documented. But equally vital, and much less discussed, is the more opportunity-focused counterpoint: \"what promising applications do LLMs enable that could promote equity?\" If LLMs are to enable a more equitable world, it is not enough just to play defense against their biases and failure modes. We must also go on offense, applying them positively to equity-enhancing use cases to increase opportunities for underserved groups and reduce societal discrimination. There are many choices which determine the impact of AI, and a fundamental choice very early in the pipeline is the problems we choose to apply it to. If we focus only later in the pipeline -- making LLMs marginally more fair as they facilitate use cases which intrinsically entrench power -- we will miss an important opportunity to guide them to equitable impacts. Here, we highlight the emerging potential of LLMs to promote equity by presenting four newly possible, promising research directions, while keeping risks and cautionary points in clear view."
                    }
                ]
            },
            "aef8f8f3-2310-4ee2-a66d-2ffdad58a771": {
                "pk": "aef8f8f3-2310-4ee2-a66d-2ffdad58a771",
                "name": "Justin Y. Chen",
                "collaborators": [
                    "Piotr Indyk",
                    "Tal Wagner",
                    "Anders Aamand",
                    "Gregory Valiant",
                    "Paul Valiant",
                    "Shyam Narayanan",
                    "Yinzhan Xu",
                    "David P. Woodruff"
                ],
                "domain": [
                    "Data Streaming",
                    "Graph Algorithms",
                    "Statistical Estimation",
                    "Differential Privacy"
                ],
                "publications": [
                    {
                        "title": "Streaming Algorithms for Support-Aware Histograms",
                        "abstract": "Histograms, i.e., piece-wise constant approximations, are a popular tool used to represent data distributions. Traditionally, the difference between the histogram and the underlying distribution (i.e., the approximation error) is measured using the $L_p$ norm, which sums the differences between the two functions over all items in the domain. Although useful in many applications, the drawback of this error measure is that it treats approximation errors of all items in the same way, irrespective of whether the mass of an item is important for the downstream application that uses the approximation. As a result, even relatively simple distributions cannot be approximated by succinct histograms without incurring large error.   In this paper, we address this issue by adapting the definition of approximation so that only the errors of the items that belong to the support of the distribution are considered. Under this definition, we develop efficient 1-pass and 2-pass streaming algorithms that compute near-optimal histograms in sub-linear space. We also present lower bounds on the space complexity of this problem. Surprisingly, under this notion of error, there is an exponential gap in the space complexity of 1-pass and 2-pass streaming algorithms. Finally, we demonstrate the utility of our algorithms on a collection of real and synthetic data sets."
                    },
                    {
                        "title": "(Optimal) Online Bipartite Matching with Degree Information",
                        "abstract": "We propose a model for online graph problems where algorithms are given access to an oracle that predicts (e.g., based on modeling assumptions or on past data) the degrees of nodes in the graph. Within this model, we study the classic problem of online bipartite matching, and a natural greedy matching algorithm called MinPredictedDegree, which uses predictions of the degrees of offline nodes. For the bipartite version of a stochastic graph model due to Chung, Lu, and Vu where the expected values of the offline degrees are known and used as predictions, we show that MinPredictedDegree stochastically dominates any other online algorithm, i.e., it is optimal for graphs drawn from this model. Since the \"symmetric\" version of the model, where all online nodes are identical, is a special case of the well-studied \"known i.i.d. model\", it follows that the competitive ratio of MinPredictedDegree on such inputs is at least 0.7299. For the special case of graphs with power law degree distributions, we show that MinPredictedDegree frequently produces matchings almost as large as the true maximum matching on such graphs. We complement these results with an extensive empirical evaluation showing that MinPredictedDegree compares favorably to state-of-the-art online algorithms for online matching."
                    },
                    {
                        "title": "Worst-Case Analysis for Randomly Collected Data",
                        "abstract": "We introduce a framework for statistical estimation that leverages knowledge of how samples are collected but makes no distributional assumptions on the data values. Specifically, we consider a population of elements $[n]={1,\\ldots,n}$ with corresponding data values $x_1,\\ldots,x_n$. We observe the values for a \"sample\" set $A \\subset [n]$ and wish to estimate some statistic of the values for a \"target\" set $B \\subset [n]$ where $B$ could be the entire set. Crucially, we assume that the sets $A$ and $B$ are drawn according to some known distribution $P$ over pairs of subsets of $[n]$. A given estimation algorithm is evaluated based on its \"worst-case, expected error\" where the expectation is with respect to the distribution $P$ from which the sample $A$ and target sets $B$ are drawn, and the worst-case is with respect to the data values $x_1,\\ldots,x_n$. Within this framework, we give an efficient algorithm for estimating the target mean that returns a weighted combination of the sample values--where the weights are functions of the distribution $P$ and the sample and target sets $A$, $B$--and show that the worst-case expected error achieved by this algorithm is at most a multiplicative $\\pi/2$ factor worse than the optimal of such algorithms. The algorithm and proof leverage a surprising connection to the Grothendieck problem. This framework, which makes no distributional assumptions on the data values but rather relies on knowledge of the data collection process, is a significant departure from typical estimation and introduces a uniform algorithmic analysis for the many natural settings where membership in a sample may be correlated with data values, such as when sampling probabilities vary as in \"importance sampling\", when individuals are recruited into a sample via a social network as in \"snowball sampling\", or when samples have chronological structure as in \"selective prediction\"."
                    },
                    {
                        "title": "All-Pairs Shortest Path Distances with Differential Privacy: Improved Algorithms for Bounded and Unbounded Weights",
                        "abstract": "We revisit the problem of privately releasing the all-pairs shortest path distances of a weighted undirected graph up to low additive error, which was first studied by Sealfon [Sea16]. In this paper, we improve significantly on Sealfon's results, both for arbitrary weighted graphs and for bounded-weight graphs on $n$ nodes. Specifically, we provide an approximate-DP algorithm that outputs all-pairs shortest path distances up to maximum additive error $\\tilde{O}(\\sqrt{n})$, and a pure-DP algorithm that outputs all pairs shortest path distances up to maximum additive error $\\tilde{O}(n^{2/3})$ (where we ignore dependencies on $\\varepsilon, \\delta$). This improves over the previous best result of $\\tilde{O}(n)$ additive error for both approximate-DP and pure-DP [Sea16], and partially resolves an open question posed by Sealfon [Sea16, Sea20]. We also show that if the graph is promised to have reasonably bounded weights, one can improve the error further to roughly $n^{\\sqrt{2}-1+o(1)}$ in the approximate-DP setting and roughly $n^{(\\sqrt{17}-3)/2 + o(1)}$ in the pure-DP setting. Previously, it was only known how to obtain $\\tilde{O}(n^{1/2})$ additive error in the approximate-DP setting and $\\tilde{O}(n^{2/3})$ additive error in the pure-DP setting for bounded-weight graphs [Sea16]."
                    },
                    {
                        "title": "Space-Optimal Profile Estimation in Data Streams with Applications to Symmetric Functions",
                        "abstract": "We revisit the problem of estimating the profile (also known as the rarity) in the data stream model. Given a sequence of $m$ elements from a universe of size $n$, its profile is a vector $\\phi$ whose $i$-th entry $\\phi_i$ represents the number of distinct elements that appear in the stream exactly $i$ times. A classic paper by Datar and Muthukrishan from 2002 gave an algorithm which estimates any entry $\\phi_i$ up to an additive error of $\\pm \\epsilon D$ using $O(1/\\epsilon^2 (\\log n + \\log m))$ bits of space, where $D$ is the number of distinct elements in the stream. In this paper, we considerably improve on this result by designing an algorithm which simultaneously estimates many coordinates of the profile vector $\\phi$ up to small overall error. We give an algorithm which, with constant probability, produces an estimated profile $\\hat\\phi$ with the following guarantees in terms of space and estimation error:   - For any constant $\\tau$, with $O(1 / \\epsilon^2 + \\log n)$ bits of space, $\\sum_{i=1}^\\tau |\\phi_i - \\hat\\phi_i| \\leq \\epsilon D$.   - With $O(1/ \\epsilon^2\\log (1/\\epsilon) + \\log n + \\log \\log m)$ bits of space, $\\sum_{i=1}^m |\\phi_i - \\hat\\phi_i| \\leq \\epsilon m$.   In addition to bounding the error across multiple coordinates, our space bounds separate the terms that depend on $1/\\epsilon$ and those that depend on $n$ and $m$. We prove matching lower bounds on space in both regimes. Application of our profile estimation algorithm gives estimates within error $\\pm \\epsilon D$ of several symmetric functions of frequencies in $O(1/\\epsilon^2 + \\log n)$ bits. This generalizes space-optimal algorithms for the distinct elements problems to other problems including estimating the Huber and Tukey losses as well as frequency cap statistics."
                    }
                ]
            },
            "b23a5b93-c26a-4edc-b274-b8adafdd3879": {
                "pk": "b23a5b93-c26a-4edc-b274-b8adafdd3879",
                "name": "Jon Kleinberg",
                "collaborators": [
                    "Manish Raghavan",
                    "Michela Meister",
                    "Sendhil Mullainathan",
                    "Lars Backstrom",
                    "Robert Kleinberg",
                    "Katrina Ligett",
                    "Maithra Raghu",
                    "Rahmtin Rotabi",
                    "Hoda Heidari",
                    "Anna Evtushenko"
                ],
                "domain": [
                    "Social Network Analysis",
                    "Algorithmic Fairness",
                    "Resource Allocation",
                    "Behavioral Economics"
                ],
                "publications": [
                    {
                        "title": "Romantic Partnerships and the Dispersion of Social Ties: A Network Analysis of Relationship Status on Facebook",
                        "abstract": "A crucial task in the analysis of on-line social-networking systems is to identify important people --- those linked by strong social ties --- within an individual's network neighborhood. Here we investigate this question for a particular category of strong ties, those involving spouses or romantic partners. We organize our analysis around a basic question: given all the connections among a person's friends, can you recognize his or her romantic partner from the network structure alone? Using data from a large sample of Facebook users, we find that this task can be accomplished with high accuracy, but doing so requires the development of a new measure of tie strength that we term `dispersion' --- the extent to which two people's mutual friends are not themselves well-connected. The results offer methods for identifying types of structurally significant people in on-line applications, and suggest a potential expansion of existing theories of tie strength."
                    },
                    {
                        "title": "Delegated Search Approximates Efficient Search",
                        "abstract": "There are many settings in which a principal performs a task by delegating it to an agent, who searches over possible solutions and proposes one to the principal. This describes many aspects of the workflow within organizations, as well as many of the activities undertaken by regulatory bodies, who often obtain relevant information from the parties being regulated through a process of delegation. A fundamental tension underlying delegation is the fact that the agent's interests will typically differ -- potentially significantly -- from the interests of the principal, and as a result the agent may propose solutions based on their own incentives that are inefficient for the principal. A basic problem, therefore, is to design mechanisms by which the principal can constrain the set of proposals they are willing to accept from the agent, to ensure a certain level of quality for the principal from the proposed solution.   In this work, we investigate how much the principal loses -- quantitatively, in terms of the objective they are trying to optimize -- when they delegate to an agent. We develop a methodology for bounding this loss of efficiency, and show that in a very general model of delegation, there is a family of mechanisms achieving a universal bound on the ratio between the quality of the solution obtained through delegation and the quality the principal could obtain in an idealized benchmark where they searched for a solution themself. Moreover, it is possible to achieve such bounds through mechanisms with a natural threshold structure, which are thus structurally simpler than the optimal mechanisms typically considered in the literature on delegation. At the heart of our framework is an unexpected connection between delegation and the analysis of prophet inequalities, which we leverage to provide bounds on the behavior of our delegation mechanisms."
                    },
                    {
                        "title": "Containing the spread of a contagion on a tree",
                        "abstract": "Contact tracing can be thought of as a race between two processes: an infection process and a tracing process. In this paper, we study a simple model of infection spreading on a tree, and a tracer who stabilizes one node at a time. We focus on the question, how should the tracer choose nodes to stabilize so as to prevent the infection from spreading further? We study simple policies, which prioritize nodes based on time, infectiousness, or probability of generating new contacts."
                    },
                    {
                        "title": "Simplicity Creates Inequity: Implications for Fairness, Stereotypes, and Interpretability",
                        "abstract": "Algorithms are increasingly used to aid, or in some cases supplant, human decision-making, particularly for decisions that hinge on predictions. As a result, two additional features in addition to prediction quality have generated interest: (i) to facilitate human interaction and understanding with these algorithms, we desire prediction functions that are in some fashion simple or interpretable; and (ii) because they influence consequential decisions, we also want them to produce equitable allocations. We develop a formal model to explore the relationship between the demands of simplicity and equity. Although the two concepts appear to be motivated by qualitatively distinct goals, we show a fundamental inconsistency between them. Specifically, we formalize a general framework for producing simple prediction functions, and in this framework we establish two basic results. First, every simple prediction function is strictly improvable: there exists a more complex prediction function that is both strictly more efficient and also strictly more equitable. Put another way, using a simple prediction function both reduces utility for disadvantaged groups and reduces overall welfare relative to other options. Second, we show that simple prediction functions necessarily create incentives to use information about individuals' membership in a disadvantaged group --- incentives that weren't present before simplification, and that work against these individuals. Thus, simplicity transforms disadvantage into bias against the disadvantaged group. Our results are not only about algorithms but about any process that produces simple models, and as such they connect to the psychology of stereotypes and to an earlier economics literature on statistical discrimination."
                    },
                    {
                        "title": "Language Generation in the Limit",
                        "abstract": "Although current large language models are complex, the most basic specifications of the underlying language generation problem itself are simple to state: given a finite set of training samples from an unknown language, produce valid new strings from the language that don't already appear in the training data. Here we ask what we can conclude about language generation using only this specification, without further assumptions. In particular, suppose that an adversary enumerates the strings of an unknown target language L that is known only to come from one of a possibly infinite list of candidates. A computational agent is trying to learn to generate from this language; we say that the agent generates from L in the limit if after some finite point in the enumeration of L, the agent is able to produce new elements that come exclusively from L and that have not yet been presented by the adversary. Our main result is that there is an agent that is able to generate in the limit for every countable list of candidate languages. This contrasts dramatically with negative results due to Gold and Angluin in a well-studied model of language learning where the goal is to identify an unknown language from samples; the difference between these results suggests that identifying a language is a fundamentally different problem than generating from it."
                    },
                    {
                        "title": "Information-Sharing and Privacy in Social Networks",
                        "abstract": "We present a new model for reasoning about the way information is shared among friends in a social network, and the resulting ways in which it spreads. Our model formalizes the intuition that revealing personal information in social settings involves a trade-off between the benefits of sharing information with friends, and the risks that additional gossiping will propagate it to people with whom one is not on friendly terms. We study the behavior of rational agents in such a situation, and we characterize the existence and computability of stable information-sharing networks, in which agents do not have an incentive to change the partners with whom they share information. We analyze the implications of these stable networks for social welfare, and the resulting fragmentation of the social network."
                    },
                    {
                        "title": "Team Performance with Test Scores",
                        "abstract": "Team performance is a ubiquitous area of inquiry in the social sciences, and it motivates the problem of team selection -- choosing the members of a team for maximum performance. Influential work of Hong and Page has argued that testing individuals in isolation and then assembling the highest-scoring ones into a team is not an effective method for team selection. For a broad class of performance measures, based on the expected maximum of random variables representing individual candidates, we show that tests directly measuring individual performance are indeed ineffective, but that a more subtle family of tests used in isolation can provide a constant-factor approximation for team performance. These new tests measure the \"potential\" of individuals, in a precise sense, rather than performance, to our knowledge they represent the first time that individual tests have been shown to produce near-optimal teams for a non-trivial team performance measure. We also show families of subdmodular and supermodular team performance functions for which no test applied to individuals can produce near-optimal teams, and discuss implications for submodular maximization via hill-climbing."
                    },
                    {
                        "title": "The Status Gradient of Trends in Social Media",
                        "abstract": "An active line of research has studied the detection and representation of trends in social media content. There is still relatively little understanding, however, of methods to characterize the early adopters of these trends: who picks up on these trends at different points in time, and what is their role in the system? We develop a framework for analyzing the population of users who participate in trending topics over the course of these topics' lifecycles. Central to our analysis is the notion of a \"status gradient\", describing how users of different activity levels adopt a trend at different points in time. Across multiple datasets, we find that this methodology reveals key differences in the nature of the early adopters in different domains."
                    },
                    {
                        "title": "How Do Classifiers Induce Agents To Invest Effort Strategically?",
                        "abstract": "Algorithms are often used to produce decision-making rules that classify or evaluate individuals. When these individuals have incentives to be classified a certain way, they may behave strategically to influence their outcomes. We develop a model for how strategic agents can invest effort in order to change the outcomes they receive, and we give a tight characterization of when such agents can be incentivized to invest specified forms of effort into improving their outcomes as opposed to \"gaming\" the classifier. We show that whenever any \"reasonable\" mechanism can do so, a simple linear mechanism suffices."
                    },
                    {
                        "title": "Algorithmic Monoculture and Social Welfare",
                        "abstract": "As algorithms are increasingly applied to screen applicants for high-stakes decisions in employment, lending, and other domains, concerns have been raised about the effects of algorithmic monoculture, in which many decision-makers all rely on the same algorithm. This concern invokes analogies to agriculture, where a monocultural system runs the risk of severe harm from unexpected shocks. Here we show that the dangers of algorithmic monoculture run much deeper, in that monocultural convergence on a single algorithm by a group of decision-making agents, even when the algorithm is more accurate for any one agent in isolation, can reduce the overall quality of the decisions being made by the full collection of agents. Unexpected shocks are therefore not needed to expose the risks of monoculture; it can hurt accuracy even under \"normal\" operations, and even for algorithms that are more accurate when used by only a single decision-maker. Our results rely on minimal assumptions, and involve the development of a probabilistic framework for analyzing systems that use multiple noisy estimates of a set of alternatives."
                    },
                    {
                        "title": "Allocating Opportunities in a Dynamic Model of Intergenerational Mobility",
                        "abstract": "Opportunities such as higher education can promote intergenerational mobility, leading individuals to achieve levels of socioeconomic status above that of their parents. We develop a dynamic model for allocating such opportunities in a society that exhibits bottlenecks in mobility; the problem of optimal allocation reflects a trade-off between the benefits conferred by the opportunities in the current generation and the potential to elevate the socioeconomic status of recipients, shaping the composition of future generations in ways that can benefit further from the opportunities. We show how optimal allocations in our model arise as solutions to continuous optimization problems over multiple generations, and we find in general that these optimal solutions can favor recipients of low socioeconomic status over slightly higher-performing individuals of high socioeconomic status -- a form of socioeconomic affirmative action that the society in our model discovers in the pursuit of purely payoff-maximizing goals. We characterize how the structure of the model can lead to either temporary or persistent affirmative action, and we consider extensions of the model with more complex processes modulating the movement between different levels of socioeconomic status."
                    },
                    {
                        "title": "The Paradox of Second-Order Homophily in Networks",
                        "abstract": "Homophily -- the tendency of nodes to connect to others of the same type -- is a central issue in the study of networks. Here we take a local view of homophily, defining notions of first-order homophily of a node (its individual tendency to link to similar others) and second-order homophily of a node (the aggregate first-order homophily of its neighbors). Through this view, we find a surprising result for homophily values that applies with only minimal assumptions on the graph topology. It can be phrased most simply as \"in a graph of red and blue nodes, red friends of red nodes are on average more homophilous than red friends of blue nodes.\" This gap in averages defies simple intuitive explanations, applies to globally heterophilous and homophilous networks and is reminiscent of but structurally distinct from the Friendship Paradox. The existence of this gap suggests intrinsic biases in homophily measurements between groups, and hence is relevant to empirical studies of homophily in networks."
                    },
                    {
                        "title": "Optimizing the order of actions in contact tracing",
                        "abstract": "Contact tracing is a key tool for managing epidemic diseases like HIV, tuberculosis, and COVID-19. Manual investigations by human contact tracers remain a dominant way in which this is carried out. This process is limited by the number of contact tracers available, who are often overburdened during an outbreak or epidemic. As a result, a crucial decision in any contact tracing strategy is, given a set of contacts, which person should a tracer trace next? In this work, we develop a formal model that articulates these questions and provides a framework for comparing contact tracing strategies. Through analyzing our model, we give provably optimal prioritization policies via a clean connection to a tool from operations research called a \"branching bandit\". Examining these policies gives qualitative insight into trade-offs in contact tracing applications."
                    },
                    {
                        "title": "Hypergraph Ego-networks and Their Temporal Evolution",
                        "abstract": "Interactions involving multiple objects simultaneously are ubiquitous across many domains. The systems these interactions inhabit can be modelled using hypergraphs, a generalization of traditional graphs in which each edge can connect any number of nodes. Analyzing the global and static properties of these hypergraphs has led to a plethora of novel findings regarding how these modelled systems are structured. However, less is known about the localized structure of these systems and how they evolve over time. In this paper, we propose the study of hypergraph ego-networks, a structure that can be used to model higher-order interactions involving a single node. We also propose the temporal reconstruction of hypergraph ego-networks as a benchmark problem for models that aim to predict the local temporal structure of hypergraphs. By combining a deep learning binary classifier with a hill-climbing algorithm, we will present a model for reconstructing hypergraph ego-networks by incorporating structural patterns found across multiple domains."
                    },
                    {
                        "title": "Core-periphery Models for Hypergraphs",
                        "abstract": "We introduce a random hypergraph model for core-periphery structure. By leveraging our model's sufficient statistics, we develop a novel statistical inference algorithm that is able to scale to large hypergraphs with runtime that is practically linear wrt. the number of nodes in the graph after a preprocessing step that is almost linear in the number of hyperedges, as well as a scalable sampling algorithm. Our inference algorithm is capable of learning embeddings that correspond to the reputation (rank) of a node within the hypergraph. We also give theoretical bounds on the size of the core of hypergraphs generated by our model. We experiment with hypergraph data that range to $\\sim 10^5$ hyperedges mined from the Microsoft Academic Graph, Stack Exchange, and GitHub and show that our model outperforms baselines wrt. producing good fits."
                    },
                    {
                        "title": "Supervised Hypergraph Reconstruction",
                        "abstract": "We study an issue commonly seen with graph data analysis: many real-world complex systems involving high-order interactions are best encoded by hypergraphs; however, their datasets often end up being published or studied only in the form of their projections (with dyadic edges). To understand this issue, we first establish a theoretical framework to characterize this issue's implications and worst-case scenarios. The analysis motivates our formulation of the new task, supervised hypergraph reconstruction: reconstructing a real-world hypergraph from its projected graph, with the help of some existing knowledge of the application domain.   To reconstruct hypergraph data, we start by analyzing hyperedge distributions in the projection, based on which we create a framework containing two modules: (1) to handle the enormous search space of potential hyperedges, we design a sampling strategy with efficacy guarantees that significantly narrows the space to a smaller set of candidates; (2) to identify hyperedges from the candidates, we further design a hyperedge classifier in two well-working variants that capture structural features in the projection. Extensive experiments validate our claims, approach, and extensions. Remarkably, our approach outperforms all baselines by an order of magnitude in accuracy on hard datasets. Our code and data can be downloaded from bit.ly/SHyRe."
                    },
                    {
                        "title": "Tracking Patterns in Toxicity and Antisocial Behavior Over User Lifetimes on Large Social Media Platforms",
                        "abstract": "An increasing amount of attention has been devoted to the problem of \"toxic\" or antisocial behavior on social media. In this paper we analyze such behavior at very large scales: we analyze toxicity over a 14-year time span on nearly 500 million comments from Reddit and Wikipedia, grounded in two different proxies for toxicity.   At the individual level, we analyze users' toxicity levels over the course of their time on the site, and find a striking reversal in trends: both Reddit and Wikipedia users tended to become less toxic over their life cycles on the site in the early (pre-2013) history of the site, but more toxic over their life cycles in the later (post-2013) history of the site. We also find that toxicity on Reddit and Wikipedia differ in a key way, with the most toxic behavior on Reddit exhibited in aggregate by the most active users, and the most toxic behavior on Wikipedia exhibited in aggregate by the least active users. Finally, we consider the toxicity of discussion around widely-shared pieces of content, and find that the trends for toxicity in discussion about content bear interesting similarities with the trends for toxicity in discussion by users."
                    },
                    {
                        "title": "Time-Inconsistent Planning: A Computational Problem in Behavioral Economics",
                        "abstract": "In many settings, people exhibit behavior that is inconsistent across time --- we allocate a block of time to get work done and then procrastinate, or put effort into a project and then later fail to complete it. An active line of research in behavioral economics and related fields has developed and analyzed models for this type of time-inconsistent behavior.   Here we propose a graph-theoretic model of tasks and goals, in which dependencies among actions are represented by a directed graph, and a time-inconsistent agent constructs a path through this graph. We first show how instances of this path-finding problem on different input graphs can reconstruct a wide range of qualitative phenomena observed in the literature on time-inconsistency, including procrastination, abandonment of long-range tasks, and the benefits of reduced sets of choices. We then explore a set of analyses that quantify over the set of all graphs; among other results, we find that in any graph, there can be only polynomially many distinct forms of time-inconsistent behavior; and any graph in which a time-inconsistent agent incurs significantly more cost than an optimal agent must contain a large \"procrastination\" structure as a minor. Finally, we use this graph-theoretic model to explore ways in which tasks can be designed to help motivate agents to reach designated goals."
                    },
                    {
                        "title": "Selection Problems in the Presence of Implicit Bias",
                        "abstract": "Over the past two decades, the notion of implicit bias has come to serve as an important component in our understanding of discrimination in activities such as hiring, promotion, and school admissions. Research on implicit bias posits that when people evaluate others -- for example, in a hiring context -- their unconscious biases about membership in particular groups can have an effect on their decision-making, even when they have no deliberate intention to discriminate against members of these groups. A growing body of experimental work has pointed to the effect that implicit bias can have in producing adverse outcomes.   Here we propose a theoretical model for studying the effects of implicit bias on selection decisions, and a way of analyzing possible procedural remedies for implicit bias within this model. A canonical situation represented by our model is a hiring setting: a recruiting committee is trying to choose a set of finalists to interview among the applicants for a job, evaluating these applicants based on their future potential, but their estimates of potential are skewed by implicit bias against members of one group. In this model, we show that measures such as the Rooney Rule, a requirement that at least one of the finalists be chosen from the affected group, can not only improve the representation of this affected group, but also lead to higher payoffs in absolute terms for the organization performing the recruiting. However, identifying the conditions under which such measures can lead to improved payoffs involves subtle trade-offs between the extent of the bias and the underlying distribution of applicant characteristics, leading to novel theoretical questions about order statistics in the presence of probabilistic side information."
                    },
                    {
                        "title": "Fairness and Utilization in Allocating Resources with Uncertain Demand",
                        "abstract": "Resource allocation problems are a fundamental domain in which to evaluate the fairness properties of algorithms. The trade-offs between fairness and utilization have a long history in this domain. A recent line of work has considered fairness questions for resource allocation when the demands for the resource are distributed across multiple groups and drawn from probability distributions. In such cases, a natural fairness requirement is that individuals from different groups should have (approximately) equal probabilities of receiving the resource. A largely open question in this area has been to bound the gap between the maximum possible utilization of the resource and the maximum possible utilization subject to this fairness condition.   Here, we obtain some of the first provable upper bounds on this gap. We obtain an upper bound for arbitrary distributions, as well as much stronger upper bounds for specific families of distributions that are typically used to model levels of demand. In particular, we find - somewhat surprisingly - that there are natural families of distributions (including Exponential and Weibull) for which the gap is non-existent: it is possible to simultaneously achieve maximum utilization and the given notion of fairness. Finally, we show that for power-law distributions, there is a non-trivial gap between the solutions, but this gap can be bounded by a constant factor independent of the parameters of the distribution."
                    }
                ]
            },
            "8351574a-9d9b-403a-8324-4a2e0454f928": {
                "pk": "8351574a-9d9b-403a-8324-4a2e0454f928",
                "name": "Sendhil Mullainathan",
                "collaborators": [
                    "Jon Kleinberg",
                    "Ziad Obermeyer",
                    "Ashesh Rambachan",
                    "Annie Liang",
                    "Manish Raghavan",
                    "Andrew C. Miller",
                    "Jens Ludwig",
                    "Maithra Raghu",
                    "Katy Blumer",
                    "Drew Fudenberg"
                ],
                "domain": [
                    "Machine Learning",
                    "Fairness",
                    "Causal Inference",
                    "Health Informatics"
                ],
                "publications": [
                    {
                        "title": "Simplicity Creates Inequity: Implications for Fairness, Stereotypes, and Interpretability",
                        "abstract": "Algorithms are increasingly used to aid, or in some cases supplant, human decision-making, particularly for decisions that hinge on predictions. As a result, two additional features in addition to prediction quality have generated interest: (i) to facilitate human interaction and understanding with these algorithms, we desire prediction functions that are in some fashion simple or interpretable; and (ii) because they influence consequential decisions, we also want them to produce equitable allocations. We develop a formal model to explore the relationship between the demands of simplicity and equity. Although the two concepts appear to be motivated by qualitatively distinct goals, we show a fundamental inconsistency between them. Specifically, we formalize a general framework for producing simple prediction functions, and in this framework we establish two basic results. First, every simple prediction function is strictly improvable: there exists a more complex prediction function that is both strictly more efficient and also strictly more equitable. Put another way, using a simple prediction function both reduces utility for disadvantaged groups and reduces overall welfare relative to other options. Second, we show that simple prediction functions necessarily create incentives to use information about individuals' membership in a disadvantaged group --- incentives that weren't present before simplification, and that work against these individuals. Thus, simplicity transforms disadvantage into bias against the disadvantaged group. Our results are not only about algorithms but about any process that produces simple models, and as such they connect to the psychology of stereotypes and to an earlier economics literature on statistical discrimination."
                    },
                    {
                        "title": "Language Generation in the Limit",
                        "abstract": "Although current large language models are complex, the most basic specifications of the underlying language generation problem itself are simple to state: given a finite set of training samples from an unknown language, produce valid new strings from the language that don't already appear in the training data. Here we ask what we can conclude about language generation using only this specification, without further assumptions. In particular, suppose that an adversary enumerates the strings of an unknown target language L that is known only to come from one of a possibly infinite list of candidates. A computational agent is trying to learn to generate from this language; we say that the agent generates from L in the limit if after some finite point in the enumeration of L, the agent is able to produce new elements that come exclusively from L and that have not yet been presented by the adversary. Our main result is that there is an agent that is able to generate in the limit for every countable list of candidate languages. This contrasts dramatically with negative results due to Gold and Angluin in a well-studied model of language learning where the goal is to identify an unknown language from samples; the difference between these results suggests that identifying a language is a fundamentally different problem than generating from it."
                    },
                    {
                        "title": "From Predictive Algorithms to Automatic Generation of Anomalies",
                        "abstract": "Machine learning algorithms can find predictive signals that researchers fail to notice; yet they are notoriously hard-to-interpret. How can we extract theoretical insights from these black boxes? History provides a clue. Facing a similar problem -- how to extract theoretical insights from their intuitions -- researchers often turned to ``anomalies:'' constructed examples that highlight flaws in an existing theory and spur the development of new ones. Canonical examples include the Allais paradox and the Kahneman-Tversky choice experiments for expected utility theory. We suggest anomalies can extract theoretical insights from black box predictive algorithms. We develop procedures to automatically generate anomalies for an existing theory when given a predictive algorithm. We cast anomaly generation as an adversarial game between a theory and a falsifier, the solutions to which are anomalies: instances where the black box algorithm predicts - were we to collect data - we would likely observe violations of the theory. As an illustration, we generate anomalies for expected utility theory using a large, publicly available dataset on real lottery choices. Based on an estimated neural network that predicts lottery choices, our procedures recover known anomalies and discover new ones for expected utility theory. In incentivized experiments, subjects violate expected utility theory on these algorithmically generated anomalies; moreover, the violation rates are similar to observed rates for the Allais paradox and Common ratio effect."
                    },
                    {
                        "title": "Measuring the Completeness of Theories",
                        "abstract": "We use machine learning to provide a tractable measure of the amount of predictable variation in the data that a theory captures, which we call its \"completeness.\" We apply this measure to three problems: assigning certain equivalents to lotteries, initial play in games, and human generation of random sequences. We discover considerable variation in the completeness of existing models, which sheds light on whether to focus on developing better models with the same features or instead to look for new features that will improve predictions. We also illustrate how and why completeness varies with the experiments considered, which highlights the role played in choosing which experiments to run."
                    },
                    {
                        "title": "Inherent Trade-Offs in the Fair Determination of Risk Scores",
                        "abstract": "Recent discussion in the public sphere about algorithmic classification has involved tension between competing notions of what it means for a probabilistic classification to be fair to different groups. We formalize three fairness conditions that lie at the heart of these debates, and we prove that except in highly constrained special cases, there is no method that can satisfy these three conditions simultaneously. Moreover, even satisfying all three conditions approximately requires that the data lie in an approximate version of one of the constrained special cases identified by our theorem. These results suggest some of the ways in which key notions of fairness are incompatible with each other, and hence provide a framework for thinking about the trade-offs between them."
                    },
                    {
                        "title": "Measuring the Stability of EHR- and EKG-based Predictive Models",
                        "abstract": "Databases of electronic health records (EHRs) are increasingly used to inform clinical decisions. Machine learning methods can find patterns in EHRs that are predictive of future adverse outcomes. However, statistical models may be built upon patterns of health-seeking behavior that vary across patient subpopulations, leading to poor predictive performance when training on one patient population and predicting on another. This note proposes two tests to better measure and understand model generalization. We use these tests to compare models derived from two data sources: (i) historical medical records, and (ii) electrocardiogram (EKG) waveforms. In a predictive task, we show that EKG-based models can be more stable than EHR-based models across different patient populations."
                    },
                    {
                        "title": "Quantifying the Causal Effects of Conversational Tendencies",
                        "abstract": "Understanding what leads to effective conversations can aid the design of better computer-mediated communication platforms. In particular, prior observational work has sought to identify behaviors of individuals that correlate to their conversational efficiency. However, translating such correlations to causal interpretations is a necessary step in using them in a prescriptive fashion to guide better designs and policies.   In this work, we formally describe the problem of drawing causal links between conversational behaviors and outcomes. We focus on the task of determining a particular type of policy for a text-based crisis counseling platform: how best to allocate counselors based on their behavioral tendencies exhibited in their past conversations. We apply arguments derived from causal inference to underline key challenges that arise in conversational settings where randomized trials are hard to implement. Finally, we show how to circumvent these inference challenges in our particular domain, and illustrate the potential benefits of an allocation policy informed by the resulting prescriptive information."
                    },
                    {
                        "title": "Do Large Language Models Perform the Way People Expect? Measuring the Human Generalization Function",
                        "abstract": "What makes large language models (LLMs) impressive is also what makes them hard to evaluate: their diversity of uses. To evaluate these models, we must understand the purposes they will be used for. We consider a setting where these deployment decisions are made by people, and in particular, people's beliefs about where an LLM will perform well. We model such beliefs as the consequence of a human generalization function: having seen what an LLM gets right or wrong, people generalize to where else it might succeed. We collect a dataset of 19K examples of how humans make generalizations across 79 tasks from the MMLU and BIG-Bench benchmarks. We show that the human generalization function can be predicted using NLP methods: people have consistent structured ways to generalize. We then evaluate LLM alignment with the human generalization function. Our results show that -- especially for cases where the cost of mistakes is high -- more capable models (e.g. GPT-4) can do worse on the instances people choose to use them for, exactly because they are not aligned with the human generalization function."
                    },
                    {
                        "title": "Assessing Human Error Against a Benchmark of Perfection",
                        "abstract": "An increasing number of domains are providing us with detailed trace data on human decisions in settings where we can evaluate the quality of these decisions via an algorithm. Motivated by this development, an emerging line of work has begun to consider whether we can characterize and predict the kinds of decisions where people are likely to make errors.   To investigate what a general framework for human error prediction might look like, we focus on a model system with a rich history in the behavioral sciences: the decisions made by chess players as they select moves in a game. We carry out our analysis at a large scale, employing datasets with several million recorded games, and using chess tablebases to acquire a form of ground truth for a subset of chess positions that have been completely solved by computers but remain challenging even for the best players in the world.   We organize our analysis around three categories of features that we argue are present in most settings where the analysis of human error is applicable: the skill of the decision-maker, the time available to make the decision, and the inherent difficulty of the decision. We identify rich structure in all three of these categories of features, and find strong evidence that in our domain, features describing the inherent difficulty of an instance are significantly more powerful than features based on skill or time."
                    },
                    {
                        "title": "Comparison-Based Choices",
                        "abstract": "A broad range of on-line behaviors are mediated by interfaces in which people make choices among sets of options. A rich and growing line of work in the behavioral sciences indicate that human choices follow not only from the utility of alternatives, but also from the choice set in which alternatives are presented. In this work we study comparison-based choice functions, a simple but surprisingly rich class of functions capable of exhibiting so-called choice-set effects. Motivated by the challenge of predicting complex choices, we study the query complexity of these functions in a variety of settings. We consider settings that allow for active queries or passive observation of a stream of queries, and give analyses both at the granularity of individuals or populations that might exhibit heterogeneous choice behavior. Our main result is that any comparison-based choice function in one dimension can be inferred as efficiently as a basic maximum or minimum choice function across many query contexts, suggesting that choice-set effects need not entail any fundamental algorithmic barriers to inference. We also introduce a class of choice functions we call distance-comparison-based functions, and briefly discuss the analysis of such functions. The framework we outline provides intriguing connections between human choice behavior and a range of questions in the theory of sorting."
                    },
                    {
                        "title": "The Theory is Predictive, but is it Complete? An Application to Human Perception of Randomness",
                        "abstract": "When we test a theory using data, it is common to focus on correctness: do the predictions of the theory match what we see in the data? But we also care about completeness: how much of the predictable variation in the data is captured by the theory? This question is difficult to answer, because in general we do not know how much \"predictable variation\" there is in the problem. In this paper, we consider approaches motivated by machine learning algorithms as a means of constructing a benchmark for the best attainable level of prediction.   We illustrate our methods on the task of predicting human-generated random sequences. Relative to an atheoretical machine learning algorithm benchmark, we find that existing behavioral models explain roughly 15 percent of the predictable variation in this problem. This fraction is robust across several variations on the problem. We also consider a version of this approach for analyzing field data from domains in which human perception and generation of randomness has been used as a conceptual framework; these include sequential decision-making and repeated zero-sum games. In these domains, our framework for testing the completeness of theories provides a way of assessing their effectiveness over different contexts; we find that despite some differences, the existing theories are fairly stable across our field domains in their performance relative to the benchmark. Overall, our results indicate that (i) there is a significant amount of structure in this problem that existing models have yet to capture and (ii) there are rich domains in which machine learning may provide a viable approach to testing completeness."
                    },
                    {
                        "title": "Machine-Learning Tests for Effects on Multiple Outcomes",
                        "abstract": "In this paper we present tools for applied researchers that re-purpose off-the-shelf methods from the computer-science field of machine learning to create a \"discovery engine\" for data from randomized controlled trials (RCTs). The applied problem we seek to solve is that economists invest vast resources into carrying out RCTs, including the collection of a rich set of candidate outcome measures. But given concerns about inference in the presence of multiple testing, economists usually wind up exploring just a small subset of the hypotheses that the available data could be used to test. This prevents us from extracting as much information as possible from each RCT, which in turn impairs our ability to develop new theories or strengthen the design of policy interventions. Our proposed solution combines the basic intuition of reverse regression, where the dependent variable of interest now becomes treatment assignment itself, with methods from machine learning that use the data themselves to flexibly identify whether there is any function of the outcomes that predicts (or has signal about) treatment group status. This leads to correctly-sized tests with appropriate $p$-values, which also have the important virtue of being easy to implement in practice. One open challenge that remains with our work is how to meaningfully interpret the signal that these methods find."
                    },
                    {
                        "title": "The Challenge of Understanding What Users Want: Inconsistent Preferences and Engagement Optimization",
                        "abstract": "Online platforms have a wealth of data, run countless experiments and use industrial-scale algorithms to optimize user experience. Despite this, many users seem to regret the time they spend on these platforms. One possible explanation is misaligned incentives: platforms are not optimizing for user happiness. We suggest the problem runs deeper, transcending the specific incentives of any particular platform, and instead stems from a mistaken foundational assumption: To understand what users want, platforms look at what users do. Yet research has demonstrated, and personal experience affirms, that we often make choices in the moment that are inconsistent with what we actually want. In this work, we develop a model of media consumption where users have inconsistent preferences. We consider a platform which simply wants to maximize user utility, but only observes user engagement. We show how our model of users' preference inconsistencies produces phenomena that are familiar from everyday experience, but difficult to capture in traditional user interaction models. A key ingredient in our model is a formulation for how platforms determine what to show users: they optimize over a large set of potential content (the content manifold) parametrized by underlying features of the content. Whether improving engagement improves user welfare depends on the direction of movement in the content manifold: for certain directions of change, increasing engagement makes users less happy, while in other directions, increasing engagement makes users happier. We characterize the structure of content manifolds for which increasing engagement fails to increase user utility. By linking these effects to abstractions of platform design choices, our model thus creates a theoretical framework and vocabulary in which to explore interactions between design, behavioral science, and social media."
                    },
                    {
                        "title": "Discrimination in the Age of Algorithms",
                        "abstract": "The law forbids discrimination. But the ambiguity of human decision-making often makes it extraordinarily hard for the legal system to know whether anyone has actually discriminated. To understand how algorithms affect discrimination, we must therefore also understand how they affect the problem of detecting discrimination. By one measure, algorithms are fundamentally opaque, not just cognitively but even mathematically. Yet for the task of proving discrimination, processes involving algorithms can provide crucial forms of transparency that are otherwise unavailable. These benefits do not happen automatically. But with appropriate requirements in place, the use of algorithms will make it possible to more easily examine and interrogate the entire decision process, thereby making it far easier to know whether discrimination has occurred. By forcing a new level of specificity, the use of algorithms also highlights, and makes transparent, central tradeoffs among competing values. Algorithms are not only a threat to be regulated; with the right safeguards in place, they have the potential to be a positive force for equity."
                    },
                    {
                        "title": "Direct Uncertainty Prediction for Medical Second Opinions",
                        "abstract": "The issue of disagreements amongst human experts is a ubiquitous one in both machine learning and medicine. In medicine, this often corresponds to doctor disagreements on a patient diagnosis. In this work, we show that machine learning models can be trained to give uncertainty scores to data instances that might result in high expert disagreements. In particular, they can identify patient cases that would benefit most from a medical second opinion. Our central methodological finding is that Direct Uncertainty Prediction (DUP), training a model to predict an uncertainty score directly from the raw patient features, works better than Uncertainty Via Classification, the two-step process of training a classifier and postprocessing the output distribution to give an uncertainty score. We show this both with a theoretical result, and on extensive evaluations on a large scale medical imaging application."
                    },
                    {
                        "title": "The Algorithmic Automation Problem: Prediction, Triage, and Human Effort",
                        "abstract": "In a wide array of areas, algorithms are matching and surpassing the performance of human experts, leading to consideration of the roles of human judgment and algorithmic prediction in these domains. The discussion around these developments, however, has implicitly equated the specific task of prediction with the general task of automation. We argue here that automation is broader than just a comparison of human versus algorithmic performance on a task; it also involves the decision of which instances of the task to give to the algorithm in the first place. We develop a general framework that poses this latter decision as an optimization problem, and we show how basic heuristics for this optimization problem can lead to performance gains even on heavily-studied applications of AI in medicine. Our framework also serves to highlight how effective automation depends crucially on estimating both algorithmic and human error on an instance-by-instance basis, and our results show how improvements in these error estimation problems can yield significant gains for automation as well."
                    },
                    {
                        "title": "Characterizing the Value of Information in Medical Notes",
                        "abstract": "Machine learning models depend on the quality of input data. As electronic health records are widely adopted, the amount of data in health care is growing, along with complaints about the quality of medical notes. We use two prediction tasks, readmission prediction and in-hospital mortality prediction, to characterize the value of information in medical notes. We show that as a whole, medical notes only provide additional predictive power over structured information in readmission prediction. We further propose a probing framework to select parts of notes that enable more accurate predictions than using all notes, despite that the selected information leads to a distribution shift from the training data (\"all notes\"). Finally, we demonstrate that models trained on the selected valuable information achieve even better predictive performance, with only 6.8% of all the tokens for readmission prediction."
                    },
                    {
                        "title": "A Probabilistic Model of Cardiac Physiology and Electrocardiograms",
                        "abstract": "An electrocardiogram (EKG) is a common, non-invasive test that measures the electrical activity of a patient's heart. EKGs contain useful diagnostic information about patient health that may be absent from other electronic health record (EHR) data. As multi-dimensional waveforms, they could be modeled using generic machine learning tools, such as a linear factor model or a variational autoencoder. We take a different approach:~we specify a model that directly represents the underlying electrophysiology of the heart and the EKG measurement process. We apply our model to two datasets, including a sample of emergency department EKG reports with missing data. We show that our model can more accurately reconstruct missing data (measured by test reconstruction error) than a standard baseline when there is significant missing data. More broadly, this physiological representation of heart function may be useful in a variety of settings, including prediction, causal analysis, and discovery."
                    },
                    {
                        "title": "The Data Station: Combining Data, Compute, and Market Forces",
                        "abstract": "This paper introduces Data Stations, a new data architecture that we are designing to tackle some of the most challenging data problems that we face today: access to sensitive data; data discovery and integration; and governance and compliance. Data Stations depart from modern data lakes in that both data and derived data products, such as machine learning models, are sealed and cannot be directly seen, accessed, or downloaded by anyone. Data Stations do not deliver data to users; instead, users bring questions to data. This inversion of the usual relationship between data and compute mitigates many of the security risks that are otherwise associated with sharing and working with sensitive data.   Data Stations are designed following the principle that many data problems require human involvement, and that incentives are the key to obtaining such involvement. To that end, Data Stations implement market designs to create, manage, and coordinate the use of incentives. We explain the motivation for this new kind of platform and its design."
                    }
                ]
            },
            "2bee71d2-c42f-4e1c-9002-deef9186fa9a": {
                "pk": "2bee71d2-c42f-4e1c-9002-deef9186fa9a",
                "name": "Ashesh Rambachan",
                "collaborators": [
                    "Neil Shephard",
                    "Jonathan Roth",
                    "Sendhil Mullainathan",
                    "Amanda Coston",
                    "Iavor Bojinov",
                    "Edward Kennedy",
                    "Alexandra Chouldechova",
                    "Keyon Vafa"
                ],
                "domain": [
                    "Causal Inference",
                    "Machine Learning",
                    "Predictive Modeling",
                    "Algorithmic Fairness"
                ],
                "publications": [
                    {
                        "title": "Econometric analysis of potential outcomes time series: instruments, shocks, linearity and the causal response function",
                        "abstract": "Bojinov & Shephard (2019) defined potential outcome time series to nonparametrically measure dynamic causal effects in time series experiments. Four innovations are developed in this paper: \"instrumental paths,\" treatments which are \"shocks,\" \"linear potential outcomes\" and the \"causal response function.\" Potential outcome time series are then used to provide a nonparametric causal interpretation of impulse response functions, generalized impulse response functions, local projections and LP-IV."
                    },
                    {
                        "title": "Bias In, Bias Out? Evaluating the Folk Wisdom",
                        "abstract": "We evaluate the folk wisdom that algorithmic decision rules trained on data produced by biased human decision-makers necessarily reflect this bias. We consider a setting where training labels are only generated if a biased decision-maker takes a particular action, and so \"biased\" training data arise due to discriminatory selection into the training data. In our baseline model, the more biased the decision-maker is against a group, the more the algorithmic decision rule favors that group. We refer to this phenomenon as \"bias reversal.\" We then clarify the conditions that give rise to bias reversal. Whether a prediction algorithm reverses or inherits bias depends critically on how the decision-maker affects the training data as well as the label used in training. We illustrate our main theoretical results in a simulation study applied to the New York City Stop, Question and Frisk dataset."
                    },
                    {
                        "title": "Design-Based Uncertainty for Quasi-Experiments",
                        "abstract": "This paper develops a finite-population, design-based theory of uncertainty for studying quasi-experimental settings in the social sciences. In our framework, treatment is determined by stochastic idiosyncratic factors, but individuals may differ in their probability of receiving treatment in ways unknown to the researcher, thus allowing for rich selection into treatment. We derive formulas for the bias of common estimators (including difference-in-means and difference-in-differences), and provide conditions under which they are unbiased for an interpretable causal estimand (e.g. analogs to the ATE or ATT). We further show that when the finite population is large, conventional standard errors are valid but typically conservative for the variance of the estimator over the randomization distribution. An interesting feature of our framework is that conventional standard errors tend to become more conservative when treatment probabilities vary more across units, i.e. when there is more selection into treatment. This conservativeness can (at least partially) mitigate undercoverage of conventional confidence intervals when the estimator is biased because of selection. Our results also have implications for the appropriate level to cluster standard errors, and for the analysis of linear covariate adjustment and instrumental variables in quasi-experimental settings."
                    },
                    {
                        "title": "From Predictive Algorithms to Automatic Generation of Anomalies",
                        "abstract": "Machine learning algorithms can find predictive signals that researchers fail to notice; yet they are notoriously hard-to-interpret. How can we extract theoretical insights from these black boxes? History provides a clue. Facing a similar problem -- how to extract theoretical insights from their intuitions -- researchers often turned to ``anomalies:'' constructed examples that highlight flaws in an existing theory and spur the development of new ones. Canonical examples include the Allais paradox and the Kahneman-Tversky choice experiments for expected utility theory. We suggest anomalies can extract theoretical insights from black box predictive algorithms. We develop procedures to automatically generate anomalies for an existing theory when given a predictive algorithm. We cast anomaly generation as an adversarial game between a theory and a falsifier, the solutions to which are anomalies: instances where the black box algorithm predicts - were we to collect data - we would likely observe violations of the theory. As an illustration, we generate anomalies for expected utility theory using a large, publicly available dataset on real lottery choices. Based on an estimated neural network that predicts lottery choices, our procedures recover known anomalies and discover new ones for expected utility theory. In incentivized experiments, subjects violate expected utility theory on these algorithmically generated anomalies; moreover, the violation rates are similar to observed rates for the Allais paradox and Common ratio effect."
                    },
                    {
                        "title": "Panel Experiments and Dynamic Causal Effects: A Finite Population Perspective",
                        "abstract": "In panel experiments, we randomly assign units to different interventions, measuring their outcomes, and repeating the procedure in several periods. Using the potential outcomes framework, we define finite population dynamic causal effects that capture the relative effectiveness of alternative treatment paths. For a rich class of dynamic causal effects, we provide a nonparametric estimator that is unbiased over the randomization distribution and derive its finite population limiting distribution as either the sample size or the duration of the experiment increases. We develop two methods for inference: a conservative test for weak null hypotheses and an exact randomization test for sharp null hypotheses. We further analyze the finite population probability limit of linear fixed effects estimators. These commonly-used estimators do not recover a causally interpretable estimand if there are dynamic causal effects and serial correlation in the assignments, highlighting the value of our proposed estimator."
                    },
                    {
                        "title": "Robust Design and Evaluation of Predictive Algorithms under Unobserved Confounding",
                        "abstract": "Predictive algorithms inform consequential decisions in settings where the outcome is selectively observed given choices made by human decision makers. We propose a unified framework for the robust design and evaluation of predictive algorithms in selectively observed data. We impose general assumptions on how much the outcome may vary on average between unselected and selected units conditional on observed covariates and identified nuisance parameters, formalizing popular empirical strategies for imputing missing data such as proxy outcomes and instrumental variables. We develop debiased machine learning estimators for the bounds on a large class of predictive performance estimands, such as the conditional likelihood of the outcome, a predictive algorithm's mean square error, true/false positive rate, and many others, under these assumptions. In an administrative dataset from a large Australian financial institution, we illustrate how varying assumptions on unobserved confounding leads to meaningful changes in default risk predictions and evaluations of credit scores across sensitive groups."
                    },
                    {
                        "title": "Characterizing Fairness Over the Set of Good Models Under Selective Labels",
                        "abstract": "Algorithmic risk assessments are used to inform decisions in a wide variety of high-stakes settings. Often multiple predictive models deliver similar overall performance but differ markedly in their predictions for individual cases, an empirical phenomenon known as the \"Rashomon Effect.\" These models may have different properties over various groups, and therefore have different predictive fairness properties. We develop a framework for characterizing predictive fairness properties over the set of models that deliver similar overall performance, or \"the set of good models.\" Our framework addresses the empirically relevant challenge of selectively labelled data in the setting where the selection decision and outcome are unconfounded given the observed data features. Our framework can be used to 1) replace an existing model with one that has better fairness properties; or 2) audit for predictive bias. We illustrate these uses cases on a real-world credit-scoring task and a recidivism prediction task."
                    },
                    {
                        "title": "Do Large Language Models Perform the Way People Expect? Measuring the Human Generalization Function",
                        "abstract": "What makes large language models (LLMs) impressive is also what makes them hard to evaluate: their diversity of uses. To evaluate these models, we must understand the purposes they will be used for. We consider a setting where these deployment decisions are made by people, and in particular, people's beliefs about where an LLM will perform well. We model such beliefs as the consequence of a human generalization function: having seen what an LLM gets right or wrong, people generalize to where else it might succeed. We collect a dataset of 19K examples of how humans make generalizations across 79 tasks from the MMLU and BIG-Bench benchmarks. We show that the human generalization function can be predicted using NLP methods: people have consistent structured ways to generalize. We then evaluate LLM alignment with the human generalization function. Our results show that -- especially for cases where the cost of mistakes is high -- more capable models (e.g. GPT-4) can do worse on the instances people choose to use them for, exactly because they are not aligned with the human generalization function."
                    }
                ]
            }
        }
    },
    "2402.02774": {
        "paper_data": {
            "title": "Accelerating Matroid Optimization through Fast Imprecise Oracles",
            "url": "http://arxiv.org/abs/2402.02774v1",
            "arxiv_id": "2402.02774",
            "authors": [
                "Franziska Eberle",
                "Felix Hommelsheim",
                "Alexander Lindermayr",
                "Zhenwei Liu",
                "Nicole Megow",
                "Jens Schl\u00f6ter"
            ],
            "abstract": "Querying complex models for precise information (e.g. traffic models, database systems, large ML models) often entails intense computations and results in long response times. Thus, weaker models which give imprecise results quickly can be advantageous, provided inaccuracies can be resolved using few queries to a stronger model. In the fundamental problem of computing a maximum-weight basis of a matroid, a well-known generalization of many combinatorial optimization problems, algorithms have access to a clean oracle to query matroid information. We additionally equip algorithms with a fast but dirty oracle modelling an unknown, potentially different matroid. We design and analyze practical algorithms which only use few clean queries w.r.t. the quality of the dirty oracle, while maintaining robustness against arbitrarily poor dirty matroids, approaching the performance of classic algorithms for the given problem. Notably, we prove that our algorithms are, in many respects, best-possible. Further, we outline extensions to other matroid oracle types, non-free dirty oracles and other matroid problems.",
            "introduction": "   1 Introduction  Matroids play a central and unifying role in combinatorial optimization as numerous classical problems can be formulated as matroid basis problems and they find application, e.g., in resource allocation and network design. Moreover, matroids precisely characterize structures for which the greedy algorithm efficiently computes an optimal solution for linear (and more general) objective functions.    A matroid \u2133=(E,\u2110)\u2133\ud835\udc38\u2110\\mathcal{M}=(E,\\mathcal{I})caligraphic_M = ( italic_E , caligraphic_I ) is a set system with a certain augmentation property that is represented by a ground set E\ud835\udc38Eitalic_E of elements and a family \u2110\u22862E\u2110superscript2\ud835\udc38\\mathcal{I}\\subseteq 2^{E}caligraphic_I \u2286 2 start_POSTSUPERSCRIPT italic_E end_POSTSUPERSCRIPT of independence sets; we give formal definitions later. Since |\u2110|\u2110|\\mathcal{I}|| caligraphic_I | can be exponential in n:=|E|assign\ud835\udc5b\ud835\udc38n:=|E|italic_n := | italic_E |, algorithms operating on matroids are usually given access to oracles via queries. The independence oracle answers for a query S\u2286E\ud835\udc46\ud835\udc38S\\subseteq Eitalic_S \u2286 italic_E whether S\u2208\u2110\ud835\udc46\u2110S\\in\\mathcal{I}italic_S \u2208 caligraphic_I. A basis of a matroid is an independent set of maximum cardinality.    In the most fundamental matroid optimization problem, we are given a matroid \u2133=(E,\u2110)\u2133\ud835\udc38\u2110\\mathcal{M}=(E,\\mathcal{I})caligraphic_M = ( italic_E , caligraphic_I ) and weights wesubscript\ud835\udc64\ud835\udc52w_{e}italic_w start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT for every element e\u2208E\ud835\udc52\ud835\udc38e\\in Eitalic_e \u2208 italic_E, and have to find a basis B\ud835\udc35Bitalic_B of maximum weight \u2211e\u2208Bwesubscript\ud835\udc52\ud835\udc35subscript\ud835\udc64\ud835\udc52\\sum_{e\\in B}w_{e}\u2211 start_POSTSUBSCRIPT italic_e \u2208 italic_B end_POSTSUBSCRIPT italic_w start_POSTSUBSCRIPT italic_e end_POSTSUBSCRIPT. In his seminal work, Edmonds\u00a0[16] showed that this problem can be solved efficiently by the greedy algorithm, and, vice versa, that matroids are the most general downward-closed set systems for which this algorithm is correct. The greedy algorithm iteratively adds elements in non-increasing order of their weight to the solution if possible; hence, it requires n\ud835\udc5bnitalic_n (independence) oracle calls. It is also not hard to see that, in general, one needs at least n\ud835\udc5bnitalic_n oracle calls. More general matroid problems include matroid intersection and matroid partition; we omit details.   Depending on the specific matroid, the oracle\u2019s task of evaluating an independence query might be highly complex, demanding significant computational effort. Consequently, substantial research efforts aim for algorithms that minimize the number of oracle calls required; see, e.g.,\u00a0[8, 5, 36].   In many methods in practice, such as those employed in machine learning pipelines or heuristic approaches, it is common to incorporate an efficient yet less precise model in addition to the precise but computationally expensive oracle model. Examples of less precise oracles include a neural network trained on potentially noisy data, or a fast heuristic designed for a simplified independence system that encompasses only a subset of the elements. More concretely, imagine a network problem in which an independence oracle shall evaluate, for a given set of edges, whether they contain a cycle (graphic matroid). While this is doable in time linear in the set size, it might become prohibitively expensive for huge input sets. In such cases, a trained model giving an inaccurate answer, or a heuristic applied to a truncated much smaller backbone subnetwork could be used to efficiently yield a potentially imprecise answer.   We initiate the study of the following two-oracle model for matroid problems: In addition to the \u201cexpensive\u201d precise (we say clean) independence oracle of a given matroid \u2133\u2133\\mathcal{M}caligraphic_M, we assume that an algorithm has access to a fast",
            "references": [
                {
                    "title": "Sorting with Predictions",
                    "abstract": "We explore the fundamental problem of sorting through the lens of learning-augmented algorithms, where algorithms can leverage possibly erroneous predictions to improve their efficiency. We consider two different settings: In the first setting, each item is provided a prediction of its position in the sorted list. In the second setting, we assume there is a\"quick-and-dirty\"way of comparing items, in addition to slow-and-exact comparisons. For both settings, we design new and simple algorithms using only $O(\\sum_i \\log \\eta_i)$ exact comparisons, where $\\eta_i$ is a suitably defined prediction error for the $i$th element. In particular, as the quality of predictions deteriorates, the number of comparisons degrades smoothly from $O(n)$ to $O(n\\log n)$. We prove that the comparison complexity is theoretically optimal with respect to the examined error measures. An experimental evaluation against existing adaptive and non-adaptive sorting algorithms demonstrates the potential of applying learning-augmented algorithms in sorting tasks."
                },
                {
                    "title": "Metric Clustering and MST with Strong and Weak Distance Oracles",
                    "abstract": "We study optimization problems in a metric space $(\\mathcal{X},d)$ where we can compute distances in two ways: via a ''strong'' oracle that returns exact distances $d(x,y)$, and a ''weak'' oracle that returns distances $\\tilde{d}(x,y)$ which may be arbitrarily corrupted with some probability. This model captures the increasingly common trade-off between employing both an expensive similarity model (e.g. a large-scale embedding model), and a less accurate but cheaper model. Hence, the goal is to make as few queries to the strong oracle as possible. We consider both so-called ''point queries'', where the strong oracle is queried on a set of points $S \\subset \\mathcal{X} $ and returns $d(x,y)$ for all $x,y \\in S$, and ''edge queries'' where it is queried for individual distances $d(x,y)$. Our main contributions are optimal algorithms and lower bounds for clustering and Minimum Spanning Tree (MST) in this model. For $k$-centers, $k$-median, and $k$-means, we give constant factor approximation algorithms with only $\\tilde{O}(k)$ strong oracle point queries, and prove that $\\Omega(k)$ queries are required for any bounded approximation. For edge queries, our upper and lower bounds are both $\\tilde{\\Theta}(k^2)$. Surprisingly, for the MST problem we give a $O(\\sqrt{\\log n})$ approximation algorithm using no strong oracle queries at all, and a matching $\\Omega(\\sqrt{\\log n})$ lower bound. We empirically evaluate our algorithms, and show that their quality is comparable to that of the baseline algorithms that are given all true distances, but while querying the strong oracle on only a small fraction ($<1\\%$) of points."
                },
                {
                    "title": "Sorting and Hypergraph Orientation under Uncertainty with Predictions",
                    "abstract": "Learning-augmented algorithms have been attracting increasing interest, but have only recently been considered in the setting of explorable uncertainty where precise values of uncertain input elements can be obtained by a query and the goal is to minimize the number of queries needed to solve a problem. We study learning-augmented algorithms for sorting and hypergraph orientation under uncertainty, assuming access to untrusted predictions for the uncertain values. Our algorithms provide improved performance guarantees for accurate predictions while maintaining worst-case guarantees that are best possible without predictions. For sorting, our algorithm uses the optimal number of queries for accurate predictions and at most twice the optimal number for arbitrarily wrong predictions. For hypergraph orientation, for any \u03b3\u22652, we give an algorithm that uses at most 1+1/\u03b3 times the optimal number of queries for accurate predictions and at most \u03b3 times the optimal number for arbitrarily wrong predictions. These tradeoffs are the best possible. We also consider different error metrics and show that the performance of our algorithms degrades smoothly with the prediction error in all the cases where this is possible."
                },
                {
                    "title": "Faster Matroid Partition Algorithms",
                    "abstract": "In the matroid partitioning problem, we are given $k$ matroids $\\mathcal{M}_1 = (V, \\mathcal{I}_1), \\dots , \\mathcal{M}_k = (V, \\mathcal{I}_k)$ defined over a common ground set $V$ of $n$ elements, and we need to find a partitionable set $S \\subseteq V$ of largest possible cardinality, denoted by $p$. Here, a set $S \\subseteq V$ is called partitionable if there exists a partition $(S_1, \\dots , S_k)$ of $S$ with $S_i \\in \\mathcal{I}_i$ for $i = 1, \\ldots, k$. In 1986, Cunningham [SICOMP 1986] presented a matroid partition algorithm that uses $O(n p^{3/2} + k n)$ independence oracle queries, which was the previously known best algorithm. This query complexity is $O(n^{5/2})$ when $k \\leq n$. Our main result is to present a matroid partition algorithm that uses $\\tilde{O}(k'^{1/3} n p + k n)$ independence oracle queries, where $k' = \\min\\{k, p\\}$. This query complexity is $\\tilde{O}(n^{7/3})$ when $k \\leq n$, and this improves upon the one of previous Cunningham's algorithm. To obtain this, we present a new approach \\emph{edge recycling augmentation}, which can be attained through new ideas: an efficient utilization of the binary search technique by Nguyen [2019] and Chakrabarty-Lee-Sidford-Singla-Wong [FOCS 2019] and a careful analysis of the independence oracle query complexity. Our analysis differs significantly from the one for matroid intersection algorithms, because of the parameter $k$. We also present a matroid partition algorithm that uses $\\tilde{O}((n + k) \\sqrt{p})$ rank oracle queries."
                },
                {
                    "title": "Efficient Submodular Optimization under Noise: Local Search is Robust",
                    "abstract": "The problem of monotone submodular maximization has been studied extensively due to its wide range of applications. However, there are cases where one can only access the objective function in a distorted or noisy form because of the uncertain nature or the errors involved in the evaluation. This paper considers the problem of constrained monotone submodular maximization with noisy oracles introduced by [Hassidim et al., 2017]. For a cardinality constraint, we propose an algorithm achieving a near-optimal $\\left(1-\\frac{1}{e}-O(\\varepsilon)\\right)$-approximation guarantee (for arbitrary $\\varepsilon>0$) with only a polynomial number of queries to the noisy value oracle, which improves the exponential query complexity of [Singer et al., 2018]. For general matroid constraints, we show the first constant approximation algorithm in the presence of noise. Our main approaches are to design a novel local search framework that can handle the effect of noise and to construct certain smoothing surrogate functions for noise reduction."
                },
                {
                    "title": "Learning-Augmented Query Policies for Minimum Spanning Tree with Uncertainty",
                    "abstract": "We study how to utilize (possibly erroneous) predictions in a model for computing under uncertainty in which an algorithm can query unknown data. Our aim is to minimize the number of queries needed to solve the minimum spanning tree problem, a fundamental combinatorial optimization problem that has been central also to the research area of explorable uncertainty. For all integral $\\gamma\\ge 2$, we present algorithms that are $\\gamma$-robust and $(1+\\frac{1}{\\gamma})$-consistent, meaning that they use at most $\\gamma OPT$ queries if the predictions are arbitrarily wrong and at most $(1+\\frac{1}{\\gamma})OPT$ queries if the predictions are correct, where $OPT$ is the optimal number of queries for the given instance. Moreover, we show that this trade-off is best possible. Furthermore, we argue that a suitably defined hop distance is a useful measure for the amount of prediction error and design algorithms with performance guarantees that degrade smoothly with the hop distance. We also show that the predictions are PAC-learnable in our model. Our results demonstrate that untrusted predictions can circumvent the known lower bound of~$2$, without any degradation of the worst-case ratio. To obtain our results, we provide new structural insights for the minimum spanning tree problem that might be useful in the context of query-based algorithms regardless of predictions. In particular, we generalize the concept of witness sets -- the key to lower-bounding the optimum -- by proposing novel global witness set structures and completely new ways of adaptively using those."
                },
                {
                    "title": "Discrete-Convex-Analysis-Based Framework for Warm-Starting Algorithms with Predictions",
                    "abstract": "Augmenting algorithms with learned predictions is a promising approach for going beyond worst-case bounds. Dinitz, Im, Lavastida, Moseley, and Vassilvitskii~(2021) have demonstrated that a warm start with learned dual solutions can improve the time complexity of the Hungarian method for weighted perfect bipartite matching. We extend and improve their framework in a principled manner via \\textit{discrete convex analysis} (DCA), a discrete analog of convex analysis. We show the usefulness of our DCA-based framework by applying it to weighted perfect bipartite matching, weighted matroid intersection, and discrete energy minimization for computer vision. Our DCA-based framework yields time complexity bounds that depend on the $\\ell_\\infty$-distance from a predicted solution to an optimal solution, which has two advantages relative to the previous $\\ell_1$-distance-dependent bounds: time complexity bounds are smaller, and learning of predictions is more sample efficient. We also discuss whether to learn primal or dual solutions from the DCA perspective."
                },
                {
                    "title": "Faster Fundamental Graph Algorithms via Learned Predictions",
                    "abstract": "We consider the question of speeding up classic graph algorithms with machine-learned predictions. In this model, algorithms are furnished with extra advice learned from past or similar instances. Given the additional information, we aim to improve upon the traditional worst-case run-time guarantees. Our contributions are the following: Finally, we give a set of general learnability theorems, showing that the predictions required by our algorithms can be efficiently learned in a PAC fashion."
                },
                {
                    "title": "Maximum Flow and Minimum-Cost Flow in Almost-Linear Time",
                    "abstract": "We give an algorithm that computes exact maximum flows and minimum-cost flows on directed graphs with m edges and polynomially bounded integral demands, costs, and capacities in $m^{1+o(1)}$ time. Our algorithm builds the flow through a sequence of $m^{1+o(1)}$ approximate undirected minimum-ratio cycles, each of which is computed and processed in amortized $m^{o(1)}$ time using a new dynamic graph data structure. Our framework extends to algorithms running in $m^{1+o(1)}$ time for computing flows that minimize general edge-separable convex functions to high accuracy. This gives almost-linear time algorithms for several problems including entropy-regularized optimal transport, matrix scaling, p-norm flows, and p-norm isotonic regression on arbitrary directed acyclic graphs."
                },
                {
                    "title": "Faster Matchings via Learned Duals",
                    "abstract": "A recent line of research investigates how algorithms can be augmented with machine-learned predictions to overcome worst case lower bounds. This area has revealed interesting algorithmic insights into problems, with particular success in the design of competitive online algorithms. However, the question of improving algorithm running times with predictions has largely been unexplored. We take a first step in this direction by combining the idea of machine-learned predictions with the idea of\"warm-starting\"primal-dual algorithms. We consider one of the most important primitives in combinatorial optimization: weighted bipartite matching and its generalization to $b$-matching. We identify three key challenges when using learned dual variables in a primal-dual algorithm. First, predicted duals may be infeasible, so we give an algorithm that efficiently maps predicted infeasible duals to nearby feasible solutions. Second, once the duals are feasible, they may not be optimal, so we show that they can be used to quickly find an optimal solution. Finally, such predictions are useful only if they can be learned, so we show that the problem of learning duals for matching has low sample complexity. We validate our theoretical findings through experiments on both real and synthetic data. As a result we give a rigorous, practical, and empirically effective method to compute bipartite matchings."
                },
                {
                    "title": "Breaking O(nr) for Matroid Intersection",
                    "abstract": "We present algorithms that break the $\\tilde O(nr)$-independence-query bound for the Matroid Intersection problem for the full range of $r$; where $n$ is the size of the ground set and $r\\leq n$ is the size of the largest common independent set. The $\\tilde O(nr)$ bound was due to the efficient implementations [CLSSW FOCS'19; Nguyen 2019] of the classic algorithm of Cunningham [SICOMP'86]. It was recently broken for large $r$ ($r=\\omega(\\sqrt{n})$), first by the $\\tilde O(n^{1.5}/\\epsilon^{1.5})$-query $(1-\\epsilon)$-approximation algorithm of CLSSW [FOCS'19], and subsequently by the $\\tilde O(n^{6/5}r^{3/5})$-query exact algorithm of BvdBMN [STOC'21]. No algorithm, even an approximation one, was known to break the $\\tilde O(nr)$ bound for the full range of $r$. We present an $\\tilde O(n\\sqrt{r}/\\epsilon)$-query $(1-\\epsilon)$-approximation algorithm and an $\\tilde O(nr^{3/4})$-query exact algorithm. Our algorithms improve the $\\tilde O(nr)$ bound and also the bounds by CLSSW and BvdBMN for the full range of $r$."
                },
                {
                    "title": "Online Facility Location in Evolving Metrics",
                    "abstract": "The Dynamic Facility Location problem is a generalization of the classic Facility Location problem, in which the distance metric between clients and facilities changes over time. Such metrics that develop as a function of time are usually called \u201cevolving metrics\u201d, thus Dynamic Facility Location can be alternatively interpreted as a Facility Location problem in evolving metrics. The objective in this time-dependent variant is to balance the trade-off between optimizing the classic objective function and the stability of the solution, which is modeled by charging a switching cost when a client\u2019s assignment changes from one facility to another. In this paper, we study the online variant of Dynamic Facility Location. We present a randomized O(logm+logn)-competitive algorithm, where m is the number of facilities and n is the number of clients. In the first step, our algorithm produces a fractional solution, in each timestep, to the objective of Dynamic Facility Location involving a regularization function. This step is an adaptation of the generic algorithm proposed by Buchbinder et al. in their work \u201cCompetitive Analysis via Regularization.\u201d Then, our algorithm rounds the fractional solution of this timestep to an integral one with the use of exponential clocks. We complement our result by proving a lower bound of \u03a9(m) for deterministic algorithms and lower bound of \u03a9(logm) for randomized algorithms. To the best of our knowledge, these are the first results for the online variant of the Dynamic Facility Location problem."
                },
                {
                    "title": "Active Learning for Noisy Data Streams Using Weak and Strong Labelers",
                    "abstract": "Labeling data correctly is an expensive and challenging task in machine learning, especially for on-line data streams. Deep learning models especially require a large number of clean labeled data that is very difficult to acquire in real-world problems. Choosing useful data samples to label while minimizing the cost of labeling is crucial to maintain efficiency in the training process. When confronted with multiple labelers with different expertise and respective labeling costs, deciding which labeler to choose is nontrivial. In this paper, we consider a novel weak and strong labeler problem inspired by humans natural ability for labeling, in the presence of data streams with noisy labels and constrained by a limited budget. We propose an on-line active learning algorithm that consists of four steps: filtering, adding diversity, informative sample selection, and labeler selection. We aim to filter out the suspicious noisy samples and spend the budget on the diverse informative data using strong and weak labelers in a cost-effective manner. We derive a decision function that measures the information gain by combining the informativeness of individual samples and model confidence. We evaluate our proposed algorithm on the well-known image classification datasets CIFAR10 and CIFAR100 with up to 60% noise. Experiments show that by intelligently deciding which labeler to query, our algorithm maintains the same accuracy compared to the case of having only one of the labelers available while spending less of the budget."
                },
                {
                    "title": "Rate Distortion Optimization Over Large Scale Video Corpus With Machine Learning",
                    "abstract": "We present an efficient codec-agnostic method for bitrate allocation over a large scale video corpus with the goal of minimizing the average bitrate subject to constraints on average and minimum quality. Our method clusters the videos in the corpus such that videos within one cluster have similar rate-distortion (R-D) characteristics. We train a support vector machine classifier to predict the R-D cluster of a video using simple video complexity features that are computationally easy to obtain. The model allows us to classify a large sample of the corpus in order to estimate the distribution of the number of videos in each of the clusters. We use this distribution to find the optimal encoder operating point for each R-D cluster. Experiments with AV1 encoder show that our method can achieve the same average quality over the corpus with 22% less average bitrate."
                },
                {
                    "title": "Algorithms with predictions",
                    "abstract": "Seeking a new approach that goes beyond worst-case analysis."
                },
                {
                    "title": "Faster Matroid Intersection",
                    "abstract": "In this paper we consider the classic matroid intersection problem: given two matroids M_1=(V, I_1) and M_2=(V, I_2) defined over a common ground set V, compute a set S\u220aI_1 \u2229 I_2 of largest possible cardinality, denoted by r. We consider this problem both in the setting where each M_i is accessed through an independence oracle, i.e. a routine which returns whether or not a set S \u220a I_i in T_ind time, and the setting where each M_i is accessed through a rank oracle, i.e. a routine which returns the size of the largest independent subset of S in M_i in T_rank time. In each setting we provide faster exact and approximate algorithms. Given an independence oracle, we provide an exact O(nr log r\u22c5T_ind) time algorithm. This improves upon previous best known running times of O(nr^1.5\u22c5T_ind) due to Cunningham in 1986 and \u00d5(n^2\u22c5T_ind+n^3) due to Lee, Sidford, and Wong in 2015. We also provide two algorithms which compute a (1-\u03b5)-approximate solution to matroid intersection running in times \u00d5(n^1.5/\u03b5^1.5\u22c5T_ind) and \u00d5((n^2r^-1\u03b5^-2+r^1.5\u03b5^-4.5)\u22c5T_ind), respectively. These results improve upon the O(nr/\u03b5\u22c5T_ind) -time algorithm of Cunningham (noted recently by Chekuri and Quanrud). Given a rank oracle, we provide algorithms with even better dependence on n and r. We provide an O(n\u221arlog n\u22c5T_rank) -time exact algorithm and an O(n\u03b5^-1 log n\u22c5T_rank) -time algorithm which obtains a (1-\u03b5) -approximation to the matroid intersection problem. The former result improves over the \u00d5(nr\u22c5T_rank+n^3) -time algorithm by Lee, Sidford, and Wong. The rank oracle is of particular interest as the matroid intersection problem with this oracle is a special case (via Edmond's minimax characterization of matroid intersection) of the submodular function minimization (SFM) problem with an evaluation oracle, and understanding SFM query complexity is an outstanding open question."
                },
                {
                    "title": "Submodular Optimization under Noise",
                    "abstract": "We consider the problem of maximizing a monotone submodular function under noise. There has been a great deal of work on optimization of submodular functions under various constraints, resulting in algorithms that provide desirable approximation guarantees. In many applications, however, we do not have access to the submodular function we aim to optimize, but rather to some erroneous or noisy version of it. This raises the question of whether provable guarantees are obtainable in presence of error and noise. We provide initial answers, by focusing on the question of maximizing a monotone submodular function under a cardinality constraint when given access to a noisy oracle of the function. We show that: \n- For a cardinality constraint $k \\geq 2$, there is an approximation algorithm whose approximation ratio is arbitrarily close to $1-1/e$; \n- For $k=1$ there is an algorithm whose approximation ratio is arbitrarily close to $1/2$. No randomized algorithm can obtain an approximation ratio better than $1/2+o(1)$; \n-If the noise is adversarial, no non-trivial approximation guarantee can be obtained."
                },
                {
                    "title": "Active Learning from Weak and Strong Labelers",
                    "abstract": "An active learner is given a hypothesis class, a large set of unlabeled examples and the ability to interactively query labels to an oracle of a subset of these examples; the goal of the learner is to learn a hypothesis in the class that fits the data well by making as few label queries as possible. \nThis work addresses active learning with labels obtained from strong and weak labelers, where in addition to the standard active learning setting, we have an extra weak labeler which may occasionally provide incorrect labels. An example is learning to classify medical images where either expensive labels may be obtained from a physician (oracle or strong labeler), or cheaper but occasionally incorrect labels may be obtained from a medical resident (weak labeler). Our goal is to learn a classifier with low error on data labeled by the oracle, while using the weak labeler to reduce the number of label queries made to this labeler. We provide an active learning algorithm for this setting, establish its statistical consistency, and analyze its label complexity to characterize when it can provide label savings over using the strong labeler alone."
                },
                {
                    "title": "From Word Embeddings To Document Distances",
                    "abstract": "We present the Word Mover's Distance (WMD), a novel distance function between text documents. Our work is based on recent results in word embeddings that learn semantically meaningful representations for words from local cooccurrences in sentences. The WMD distance measures the dissimilarity between two text documents as the minimum amount of distance that the embedded words of one document need to \"travel\" to reach the embedded words of another document. We show that this distance metric can be cast as an instance of the Earth Mover's Distance, a well studied transportation problem for which several highly efficient solvers have been developed. Our metric has no hyperparameters and is straight-forward to implement. Further, we demonstrate on eight real world document classification data sets, in comparison with seven state-of-the-art baselines, that the WMD metric leads to unprecedented low k-nearest neighbor document classification error rates."
                },
                {
                    "title": "Competitive Analysis via Regularization",
                    "abstract": "We provide a framework for designing competitive online algorithms using regularization, a widely used technique in online learning, particularly in online convex optimization. An online algorithm that uses regularization serves requests by computing a solution, in each step, to an objective function involving a smooth convex regularization function. Applying the technique of regularization allows us to obtain new results in the domain of competitive analysis. We remark that competitive analysis and online learning are two widely studied frameworks for online decision-making settings. We show that even though there are significant differences in assumptions, goals, and techniques between the two fields, one can still benefit by introducing techniques from one field to the other. \n \nIn our new framework we exhibit a general O(log m)-competitive deterministic algorithm for generating a fractional solution that satisfies a time-varying set of online covering and precedence constraints, where m is the number of variables. This framework allows to incorporate both service costs (over time) and setup costs into a host of applications. We then provide an O(log m log n)-competitive randomized algorithm for the online set cover problem with service cost, where m is the number of sets and n is the number of elements. This model allows for sets to be both added and deleted over time from a solution."
                },
                {
                    "title": "Unified Algorithms for Online Learning and Competitive Analysis",
                    "abstract": "Online learning and competitive analysis are two widely studied frameworks for online decision-making settings. Despite the frequent similarity of the problems they study, there are significant differences in their assumptions, goals, and techniques, hindering a unified analysis and richer interplay between the two. In this paper, we provide several contributions in this direction. We provide a single unified algorithm, which, by parameter tuning, interpolates between optimal regret for learning from experts (in online learning) and optimal competitive ratio for the metrical task systems problem (MTS) (in competitive analysis), improving upon previous results. The algorithm also allows us to obtain new regret bounds against \u201cdrifting\u201d experts, which might be of independent interest. Moreover, our approach allows us to go beyond experts/MTS, obtaining similar unifying results for structured action sets and \u201ccombinatorial experts,\u201d whenever the setting has a certain matroid structure."
                },
                {
                    "title": "Online submodular minimization",
                    "abstract": "We consider an online decision problem over a discrete space in which the loss function is submodular. We give algorithms which are computationally efficient and are Hannan-consistent in both the full information and partial feedback settings."
                },
                {
                    "title": "KwikBucks: Correlation Clustering with Cheap-Weak and Expensive-Strong Signals",
                    "abstract": ","
                },
                {
                    "title": "Submodular Function Minimization with Noisy Evaluation Oracle",
                    "abstract": "This paper considers submodular function minimization with \\textit{noisy evaluation oracles} that return the function value of a submodular objective with zero-mean additive noise. For this problem, we provide an algorithm that returns an $O(n^{3/2}/\\sqrt{T})$-additive approximate solution in expectation, where $n$ and $T$ stand for the size of the problem and the number of oracle calls, respectively. There is no room for reducing this error bound by a factor smaller than $O(1/\\sqrt{n})$. Indeed, we show that any algorithm will suffer additive errors of $\\Omega(n/\\sqrt{T})$ in the worst case. Further, we consider an extended problem setting with \\textit{multiple-point feedback} in which we can get the feedback of $k$ function values with each oracle call. Under the additional assumption that each noisy oracle is submodular and that $2 \\leq k = O(1)$, we provide an algorithm with an $O(n/\\sqrt{T})$-additive error bound as well as a worst-case analysis including a lower bound of $\\Omega(n/\\sqrt{T})$, which together imply that the algorithm achieves an optimal error bound up to a constant."
                },
                {
                    "title": "Multistage Matchings",
                    "abstract": "14 We consider a multistage version of the Perfect Matching problem which models the 15 scenario where the costs of edges change over time and we seek to obtain a solution that achieves 16 low total cost, while minimizing the number of changes from one instance to the next. Formally, 17 we are given a sequence of edge-weighted graphs on the same set of vertices V , and are asked to 18 produce a perfect matching in each instance so that the total edge cost plus the transition cost 19 (the cost of exchanging edges), is minimized. This model was introduced by Gupta et al. (ICALP 2"
                },
                {
                    "title": "Optimization for Approximate Submodularity",
                    "abstract": "We consider the problem of maximizing a submodular function when given access to its approximate version. Submodular functions are heavily studied in a wide variety of disciplines, since they are used to model many real world phenomena, and are amenable to optimization. However, there are many cases in which the phenomena we observe is only approximately submodular and the approximation guarantees cease to hold. We describe a technique which we call the sampled mean approximation that yields strong guarantees for maximization of submodular functions from approximate surrogates under cardinality and intersection of matroid constraints. In particular, we show tight guarantees for maximization under a cardinality constraint and 1/(1+P) approximation under intersection of P matroids."
                },
                {
                    "title": "Combinatorial optimization. Polyhedra and efficiency.",
                    "abstract": "Reading a book is also kind of better solution when you have no enough money or time to get your own adventure. This is one of the reasons we show the combinatorial optimization polyhedra and efficiency as your friend in spending the time. For more representative collections, this book not only offers it's strategically book resource. It can be a good friend, really good friend with much knowledge."
                }
            ],
            "categories": [
                "cs.DS",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively utilize a two-oracle model to optimize matroid problems by balancing the trade-off between the precision of a clean independence oracle and the efficiency of a fast, potentially imprecise oracle?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is significant for the research community as it bridges the gap between theoretical matroid optimization and practical applications in fields like machine learning and network design. By developing algorithms that can leverage both precise and fast oracles, we can enhance the efficiency of solving complex combinatorial problems, leading to advancements in resource allocation strategies and improved performance in real-world applications. This research could inspire future studies on hybrid oracle models, potentially transforming how we approach optimization problems in various domains.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to balance the accuracy of the clean oracle with the speed of the fast oracle. Naive approaches may fail because they might rely solely on one type of oracle, either sacrificing precision for speed or vice versa. The technical obstacles include designing algorithms that can effectively integrate the outputs of both oracles while ensuring that the overall solution remains optimal or near-optimal. Additionally, the complexity of independence queries in different matroid structures adds to the difficulty of developing a robust solution.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either precise independence oracles or heuristic approaches without adequately addressing the integration of both types. Limitations in existing solutions include a lack of frameworks that can effectively combine the strengths of clean and fast oracles. Barriers such as the complexity of independence queries and the absence of a unified methodology for leveraging two oracles have hindered progress. Our approach differs by proposing a structured model that explicitly incorporates both oracles, allowing for a more nuanced and efficient optimization process.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing algorithms that utilize both a clean independence oracle and a fast oracle within a two-oracle framework. We will evaluate the performance of these algorithms on various matroid problems using benchmark datasets that represent different matroid structures. The metrics for success will include the number of oracle calls, the accuracy of the solutions, and the computational time required. We expect that our approach will yield solutions that are both efficient and close to optimal, demonstrating the viability of the two-oracle model in"
            }
        },
        "author_data": {
            "720bf070-067b-4a52-98df-8bdbbfc7bdf8": {
                "pk": "720bf070-067b-4a52-98df-8bdbbfc7bdf8",
                "name": "Franziska Eberle",
                "collaborators": [
                    "Nicole Megow",
                    "Kevin Schewior",
                    "Lukas N\u00f6lke",
                    "Bertrand Simon",
                    "Andreas Wiese",
                    "Lin Chen",
                    "Cliff Stein",
                    "Felix Hommelsheim",
                    "Malin Rau",
                    "Stefan Walzer"
                ],
                "domain": [
                    "Online Scheduling",
                    "Algorithm Design",
                    "Dynamic Programming",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "$O(1/\\varepsilon)$ is the answer in online weighted throughput maximization",
                        "abstract": "We study a fundamental online scheduling problem where jobs with processing times, weights, and deadlines arrive online over time at their release dates. The task is to preemptively schedule these jobs on a single or multiple (possibly unrelated) machines with the objective to maximize the weighted throughput, the total weight of jobs that complete before their deadline. To overcome known lower bounds for the competitive analysis, we assume that each job arrives with some slack $\\varepsilon > 0$; that is, the time window for processing job $j$ on any machine $i$ on which it can be executed has length at least $(1+\\varepsilon)$ times $j$'s processing time on machine $i$. Our contribution is a best possible online algorithm for weighted throughput maximization on unrelated machines: Our algorithm is $O\\big(\\frac1\\varepsilon\\big)$-competitive, which matches the lower bound for unweighted throughput maximization on a single machine. Even for a single machine, it was not known whether the problem with weighted jobs is \"harder\" than the problem with unweighted jobs. Thus, we answer this question and close weighted throughput maximization on a single machine with a best possible competitive ratio $\\Theta\\big(\\frac1\\varepsilon\\big)$. While we focus on non-migratory schedules, our algorithm achieves the same (up to constants) performance guarantee when compared to an optimal migratory schedule."
                    },
                    {
                        "title": "A general framework for handling commitment in online throughput maximization",
                        "abstract": "We study a fundamental online job admission problem where jobs with deadlines arrive online over time at their release dates, and the task is to determine a preemptive single-server schedule which maximizes the number of jobs that complete on time. To circumvent known impossibility results, we make a standard slackness assumption by which the feasible time window for scheduling a job is at least $1+\\varepsilon$ times its processing time, for some $\\varepsilon>0$. We quantify the impact that different provider commitment requirements have on the performance of online algorithms. Our main contribution is one universal algorithmic framework for online job admission both with and without commitments. Without commitment, our algorithm with a competitive ratio of $O(1/\\varepsilon)$ is the best possible (deterministic) for this problem. For commitment models, we give the first non-trivial performance bounds. If the commitment decisions must be made before a job's slack becomes less than a $\\delta$-fraction of its size, we prove a competitive ratio of $O(\\varepsilon/((\\varepsilon-\\delta)\\delta^2))$, for $0<\\delta<\\varepsilon$. When a provider must commit upon starting a job, our bound is $O(1/\\varepsilon^2)$. Finally, we observe that for scheduling with commitment the restriction to the `unweighted' throughput model is essential; if jobs have individual weights, we rule out competitive deterministic algorithms."
                    },
                    {
                        "title": "Fully Dynamic Algorithms for Knapsack Problems with Polylogarithmic Update Time",
                        "abstract": "Knapsack problems are among the most fundamental problems in optimization. In the Multiple Knapsack problem, we are given multiple knapsacks with different capacities and items with values and sizes. The task is to find a subset of items of maximum total value that can be packed into the knapsacks without exceeding the capacities. We investigate this problem and special cases thereof in the context of dynamic algorithms and design data structures that efficiently maintain near-optimal knapsack solutions for dynamically changing input. More precisely, we handle the arrival and departure of individual items or knapsacks during the execution of the algorithm with worst-case update time polylogarithmic in the number of items. As the optimal and any approximate solution may change drastically, we only maintain implicit solutions and support certain queries in polylogarithmic time, such as the packing of an item and the solution value. While dynamic algorithms are well-studied in the context of graph problems, there is hardly any work on packing problems and generally much less on non-graph problems. Given the theoretical interest in knapsack problems and their practical relevance, it is somewhat surprising that Knapsack has not been addressed before in the context of dynamic algorithms and our work bridges this gap."
                    },
                    {
                        "title": "A Tight ($3/2 + \\varepsilon$)-Approximation Algorithm for Demand Strip Packing",
                        "abstract": "We consider the Demand Strip Packing problem (DSP), in which we are given a set of jobs, each specified by a processing time and a demand. The task is to schedule all jobs such that they are finished before some deadline $D$ while minimizing the peak demand, i.e., the maximum total demand of tasks executed at any point in time. DSP is closely related to the Strip Packing problem (SP), in which we are given a set of axis-aligned rectangles that must be packed into a strip of fixed width while minimizing the maximum height. DSP and SP are known to be NP-hard to approximate to within a factor below $\\frac{3}{2}$.   To achieve the essentially best possible approximation guarantee, we prove a structural result. Any instance admits a solution with peak demand at most $\\big(\\frac32+\\varepsilon\\big)OPT$ satisfying one of two properties. Either (i) the solution leaves a gap for a job with demand $OPT$ and processing time $\\mathcal O(\\varepsilon D)$ or (ii) all jobs with demand greater than $\\frac{OPT}2$ appear sorted by demand in immediate succession. We then provide two efficient algorithms that find a solution with maximum demand at most $\\big(\\frac32+\\varepsilon\\big)OPT$ in the respective case. A central observation, which sets our approach apart from previous ones for DSP, is that the properties (i) and (ii) need not be efficiently decidable: We can simply run both algorithms and use whichever solution is the better one."
                    },
                    {
                        "title": "Online Throughput Maximization on Unrelated Machines: Commitment is No Burden",
                        "abstract": "We consider a fundamental online scheduling problem in which jobs with processing times and deadlines arrive online over time at their release dates. The task is to determine a feasible preemptive schedule on a single or multiple possibly unrelated machines that maximizes the number of jobs that complete before their deadline. Due to strong impossibility results for competitive analysis on a single machine, we require that jobs contain some slack $\\varepsilon>0$, which means that the feasible time window for scheduling a job is at least $1+\\varepsilon$ times its processing time on each eligible machine. Our contribution is two-fold: (i) We give the first non-trivial online algorithms for throughput maximization on unrelated machines, and (ii), this is the main focus of our paper, we answer the question on how to handle commitment requirements which enforce that a scheduler has to guarantee at a certain point in time the completion of admitted jobs. This is very relevant, e.g., in providing cloud-computing services, and disallows last-minute rejections of critical tasks. We present an algorithm for unrelated machines that is $\\Theta\\big(\\frac{1}\\varepsilon\\big )$-competitive when the scheduler must commit upon starting a job. Somewhat surprisingly, this is the same optimal performance bound (up to constants) as for scheduling without commitment on a single machine. If commitment decisions must be made before a job's slack becomes less than a $\\delta$-fraction of its size, we prove a competitive ratio of $\\mathcal{O}\\big(\\frac{1}{\\varepsilon - \\delta}\\big)$ for $0 < \\delta < \\varepsilon$. This result nicely interpolates between commitment upon starting a job and commitment upon arrival. For the latter commitment model, it is known that no (randomized) online algorithm admits any bounded competitive ratio."
                    },
                    {
                        "title": "Robustification of Online Graph Exploration Methods",
                        "abstract": "Exploring unknown environments is a fundamental task in many domains, e.g., robot navigation, network security, and internet search. We initiate the study of a learning-augmented variant of the classical, notoriously hard online graph exploration problem by adding access to machine-learned predictions. We propose an algorithm that naturally integrates predictions into the well-known Nearest Neighbor (NN) algorithm and significantly outperforms any known online algorithm if the prediction is of high accuracy while maintaining good guarantees when the prediction is of poor quality. We provide theoretical worst-case bounds that gracefully degrade with the prediction error, and we complement them by computational experiments that confirm our results. Further, we extend our concept to a general framework to robustify algorithms. By interpolating carefully between a given algorithm and NN, we prove new performance bounds that leverage the individual good performance on particular inputs while establishing robustness to arbitrary inputs."
                    },
                    {
                        "title": "Speed-Robust Scheduling -- Sand, Bricks, and Rocks",
                        "abstract": "The speed-robust scheduling problem is a two-stage problem where given $m$ machines, jobs must be grouped into at most $m$ bags while the processing speeds of the given $m$ machines are unknown. After the speeds are revealed, the grouped jobs must be assigned to the machines without being separated. To evaluate the performance of algorithms, we determine upper bounds on the worst-case ratio of the algorithm's makespan and the optimal makespan given full information. We refer to this ratio as the robustness factor. We give an algorithm with a robustness factor $2-1/m$ for the most general setting and improve this to $1.8$ for equal-size jobs. For the special case of infinitesimal jobs, we give an algorithm with an optimal robustness factor equal to $e/(e-1) \\approx 1.58$. The particular machine environment in which all machines have either speed $0$ or $1$ was studied before by Stein and Zhong (SODA 2019). For this setting, we provide an algorithm for scheduling infinitesimal jobs with an optimal robustness factor of $(1+\\sqrt{2})/2 \\approx 1.207$. It lays the foundation for an algorithm matching the lower bound of $4/3$ for equal-size jobs."
                    },
                    {
                        "title": "Scheduling on a Stochastic Number of Machines",
                        "abstract": "We consider a new scheduling problem on parallel identical machines in which the number of machines is initially not known, but it follows a given probability distribution. Only after all jobs are assigned to a given number of bags, the actual number of machines is revealed. Subsequently, the jobs need to be assigned to the machines without splitting the bags. This is the stochastic version of a related problem introduced by Stein and Zhong [SODA 2018, TALG 2020] and it is, for example, motivated by bundling jobs that need to be scheduled by data centers. We present two PTASs for the stochastic setting, computing job-to-bag assignments that (i) minimize the expected maximum machine load and (ii) maximize the expected minimum machine load (like in the Santa Claus problem), respectively. The former result follows by careful enumeration combined with known PTASs. For the latter result, we introduce an intricate dynamic program that we apply to a suitably rounded instance."
                    }
                ]
            },
            "86d505d4-ff8e-4347-b34b-169f0ee165f3": {
                "pk": "86d505d4-ff8e-4347-b34b-169f0ee165f3",
                "name": "Felix Hommelsheim",
                "collaborators": [
                    "Moritz M\u00fchlenthaler",
                    "Oliver Schaudt",
                    "David Adjiashvili",
                    "Mohit Garg",
                    "Nicole Megow",
                    "Christoph Buchheim",
                    "Dorothee Henke",
                    "Komal Muluk",
                    "Britta Peis",
                    "Tatsuhiko Hatanaka"
                ],
                "domain": [
                    "Combinatorial Optimization",
                    "Graph Theory",
                    "Approximation Algorithms",
                    "Robust Optimization"
                ],
                "publications": [
                    {
                        "title": "How to Secure Matchings Against Edge Failures",
                        "abstract": "Suppose we are given a bipartite graph that admits a perfect matching and an adversary may delete any edge from the graph with the intention of destroying all perfect matchings. We consider the task of adding a minimum cost edge-set to the graph, such that the adversary never wins. We provide efficient exact and approximation algorithms. In particular, for the unit-cost problem, we provide a $\\log_2 n$-factor approximation algorithm and a polynomial-time algorithm for chordal-bipartite graphs. Furthermore, we give a fixed parameter algorithm for the problem parameterized by the treewidth of the input graph. For general non-negative weights we settle the approximability of the problem and show a close relation to the Directed Steiner Forest Problem. Additionally we prove a dichotomy theorem characterizing minor-closed graph classes which allow for a polynomial-time algorithm. Our methods rely on a close relationship to the classical strong connectivity augmentation problem and directed Steiner problems."
                    },
                    {
                        "title": "Fault-Tolerant Edge-Disjoint Paths -- Beyond Uniform Faults",
                        "abstract": "The overwhelming majority of survivable (fault-tolerant) network design models assume a uniform fault model. Such a model assumes that every subset of the network resources (edges or vertices) of a given cardinality $k$ may fail. While this approach yields problems with clean combinatorial structure and good algorithms, it often fails to capture the true nature of the scenario set coming from applications. One natural refinement of the uniform model is obtained by partitioning the set of resources into vulnerable and safe resources. The scenario set contains every subset of at most $k$ faulty resources. This work studies the Fault-Tolerant Path (FTP) problem, the counterpart of the Shortest Path problem in this fault model and the Fault-Tolerant Flow problem (FTF), the counterpart of the $\\ell$-disjoint Shortest $s$-$t$ Path problem. We present complexity results alongside exact and approximation algorithms for both models. We emphasize the vast increase in the complexity of the problem with respect to the uniform analogue, the Edge-Disjoint Paths problem."
                    },
                    {
                        "title": "Matching Augmentation via Simultaneous Contractions",
                        "abstract": "We consider the matching augmentation problem (MAP), where a matching of a graph needs to be extended into a $2$-edge-connected spanning subgraph by adding the minimum number of edges to it. We present a polynomial-time algorithm with an approximation ratio of $13/8 = 1.625$ improving upon an earlier $5/3$-approximation. The improvement builds on a new $\\alpha$-approximation preserving reduction for any $\\alpha\\geq 3/2$ from arbitrary MAP instances to well-structured instances that do not contain certain forbidden structures like parallel edges, small separators, and contractible subgraphs. We further introduce, as key ingredients, the technique of repeated simultaneous contractions and provide improved lower bounds for instances that cannot be contracted."
                    },
                    {
                        "title": "Recoverable Robust Optimization with Commitment",
                        "abstract": "We propose a model for recoverable robust optimization with commitment. Given a combinatorial optimization problem and uncertainty about elements that may fail, we ask for a robust solution that, after the failing elements are revealed, can be augmented in a limited way. However, we commit to preserve the remaining elements of the initial solution. We consider different polynomial-time solvable combinatorial optimization problems and settle the computational complexity of their robust counterparts with commitment. We show for the weighted matroid basis problem that an optimal solution to the nominal problem is also optimal for its robust counterpart. Indeed, matroids are provably the only structures with this strong property. Robust counterparts of other problems are NP-hard such as the matching and the stable set problem, even in bipartite graphs. However, we establish polynomial-time algorithms for the robust counterparts of the unweighted stable set problem in bipartite graphs and the weighted stable set problem in interval graphs, also known as the interval scheduling problem."
                    },
                    {
                        "title": "On the Complexity of Robust Bilevel Optimization With Uncertain Follower's Objective",
                        "abstract": "We investigate the complexity of bilevel combinatorial optimization with uncertainty in the follower's objective, in a robust optimization approach. We show that the robust counterpart of the bilevel problem under interval uncertainty can be $\\Sigma^{\\text P}_2$-hard, even when the certain bilevel problem is NP-equivalent and the follower's problem is tractable. On the contrary, in the discrete uncertainty case, the robust bilevel problem is at most one level harder than the follower's problem."
                    },
                    {
                        "title": "Flexible Graph Connectivity: Approximating Network Design Problems Between 1- and 2-connectivity",
                        "abstract": "Graph connectivity and network design problems are among the most fundamental problems in combinatorial optimization. The minimum spanning tree problem, the two edge-connected spanning subgraph problem (2-ECSS) and the tree augmentation problem (TAP) are all examples of fundamental well-studied network design tasks that postulate different initial states of the network and different assumptions on the reliability of network components. In this paper we motivate and study \\emph{Flexible Graph Connectivity} (FGC), a problem that mixes together both the modeling power and the complexities of all aforementioned problems and more. In a nutshell, FGC asks to design a connected network, while allowing to specify different reliability levels for individual edges. While this non-uniform nature of the problem makes it appealing from the modeling perspective, it also renders most existing algorithmic tools for dealing with network design problems unfit for approximating FGC.   In this paper we develop a general algorithmic approach for approximating FGC that yields approximation algorithms with ratios that are very close to the best known bounds for many special cases, such as 2-ECSS and TAP. Our algorithm and analysis combine various techniques including a weight-scaling algorithm, a charging argument that uses a variant of exchange bijections between spanning trees and a factor revealing min-max-min optimization problem."
                    },
                    {
                        "title": "Fixed-Parameter Algorithms for Graph Constraint Logic",
                        "abstract": "Non-deterministic constraint logic (NCL) is a simple model of computation based on orientations of a constraint graph with edge weights and vertex demands. NCL captures \\PSPACE\\xspace and has been a useful tool for proving algorithmic hardness of many puzzles, games, and reconfiguration problems. In particular, its usefulness stems from the fact that it remains \\PSPACE-complete even under severe restrictions of the weights (e.g., only edge-weights one and two are needed) and the structure of the constraint graph (e.g., planar \\textsc{and/or}\\xspace graphs of bounded bandwidth). While such restrictions on the structure of constraint graphs do not seem to limit the expressiveness of NCL, the building blocks of the constraint graphs cannot be limited without losing expressiveness: We consider as parameters the number of weight-one edges and the number of weight-two edges of a constraint graph, as well as the number of \\textsc{and}\\xspace or \\textsc{or}\\xspace vertices of an \\textsc{and/or}\\xspace constraint graph. We show that NCL is fixed-parameter tractable (FPT) for any of these parameters. In particular, for NCL parameterized by the number of weight-one edges or the number of \\textsc{and}\\xspace vertices, we obtain a linear kernel. It follows that, in a sense, NCL as introduced by Hearn and Demaine is defined in the most economical way for the purpose of capturing \\PSPACE."
                    },
                    {
                        "title": "On the Complexity of the Bilevel Minimum Spanning Tree Problem",
                        "abstract": "We consider the bilevel minimum spanning tree (BMST) problem where the leader and the follower choose a spanning tree together, according to different objective functions. By showing that this problem is NP-hard in general, we answer an open question stated in by Shi et al. We prove that BMST remains hard even in the special case where the follower only controls a matching. Moreover, by a polynomial reduction from the vertex-disjoint Steiner trees problem, we give some evidence that BMST might even remain hard in case the follower controls only few edges. On the positive side, we present a polynomial-time $(|V|-1)$-approximation algorithm for BMST, where $|V|$ is the number of vertices in the input graph. Moreover, considering the number of edges controlled by the follower as parameter, we show that 2-approximating BMST is fixed-parameter tractable and that, in case of uniform costs on leader's edges, even solving BMST exactly is fixed-parameter tractable. We finally consider bottleneck variants of BMST and settle the complexity landscape of all combinations of sum or bottleneck objective functions for the leader and follower, for the optimistic as well as the pessimistic setting."
                    },
                    {
                        "title": "Two-Edge Connectivity via Pac-Man Gluing",
                        "abstract": "We study the 2-edge-connected spanning subgraph (2-ECSS) problem: Given a graph $G$, compute a connected subgraph $H$ of $G$ with the minimum number of edges such that $H$ is spanning, i.e., $V(H) = V(G)$, and $H$ is 2-edge-connected, i.e., $H$ remains connected upon the deletion of any single edge, if such an $H$ exists. The $2$-ECSS problem is known to be NP-hard. In this work, we provide a polynomial-time $(\\frac 5 4 + \\varepsilon)$-approximation for the problem for an arbitrarily small $\\varepsilon>0$, improving the previous best approximation ratio of $\\frac{13}{10}+\\varepsilon$.   Our improvement is based on two main innovations: First, we reduce solving the problem on general graphs to solving it on structured graphs with high vertex connectivity. This high vertex connectivity ensures the existence of a 4-matching across any bipartition of the vertex set with at least 10 vertices in each part. Second, we exploit this property in a later gluing step, where isolated 2-edge-connected components need to be merged without adding too many edges. Using the 4-matching property, we can repeatedly glue a huge component (containing at least 10 vertices) to other components. This step is reminiscent of the Pac-Man game, where a Pac-Man (a huge component) consumes all the dots (other components) as it moves through a maze. These two innovations lead to a significantly simpler algorithm and analysis for the gluing step compared to the previous best approximation algorithm, which required a long and tedious case analysis."
                    },
                    {
                        "title": "A Tight ($3/2 + \\varepsilon$)-Approximation Algorithm for Demand Strip Packing",
                        "abstract": "We consider the Demand Strip Packing problem (DSP), in which we are given a set of jobs, each specified by a processing time and a demand. The task is to schedule all jobs such that they are finished before some deadline $D$ while minimizing the peak demand, i.e., the maximum total demand of tasks executed at any point in time. DSP is closely related to the Strip Packing problem (SP), in which we are given a set of axis-aligned rectangles that must be packed into a strip of fixed width while minimizing the maximum height. DSP and SP are known to be NP-hard to approximate to within a factor below $\\frac{3}{2}$.   To achieve the essentially best possible approximation guarantee, we prove a structural result. Any instance admits a solution with peak demand at most $\\big(\\frac32+\\varepsilon\\big)OPT$ satisfying one of two properties. Either (i) the solution leaves a gap for a job with demand $OPT$ and processing time $\\mathcal O(\\varepsilon D)$ or (ii) all jobs with demand greater than $\\frac{OPT}2$ appear sorted by demand in immediate succession. We then provide two efficient algorithms that find a solution with maximum demand at most $\\big(\\frac32+\\varepsilon\\big)OPT$ in the respective case. A central observation, which sets our approach apart from previous ones for DSP, is that the properties (i) and (ii) need not be efficiently decidable: We can simply run both algorithms and use whichever solution is the better one."
                    },
                    {
                        "title": "Improved Approximation Algorithms for the Expanding Search Problem",
                        "abstract": "A searcher faces a graph with edge lengths and vertex weights, initially having explored only a given starting vertex. In each step, the searcher adds an edge to the solution that connects an unexplored vertex to an explored vertex. This requires an amount of time equal to the edge length. The goal is to minimize the weighted sum of the exploration times over all vertices. We show that this problem is hard to approximate and provide algorithms with improved approximation guarantees. For the general case, we give a $(2\\mathrm{e}+\\varepsilon)$-approximation for any $\\varepsilon > 0$. For the case that all vertices have unit weight, we provide a $2\\mathrm{e}$-approximation. Finally, we provide a PTAS for the case of a Euclidean graph. Previously, for all cases only an $8$-approximation was known."
                    }
                ]
            },
            "311a4a29-b0dc-4d25-8416-1c48cd2941e0": {
                "pk": "311a4a29-b0dc-4d25-8416-1c48cd2941e0",
                "name": "Alexander Lindermayr",
                "collaborators": [
                    "Nicole Megow",
                    "Jens Schl\u00f6ter",
                    "Alexandra Lassota",
                    "Martin Rapp",
                    "Franziska Eberle",
                    "Lukas N\u00f6lke",
                    "Giulia Bernardini",
                    "Alberto Marchetti-Spaccamela",
                    "Leen Stougie",
                    "Michelle Sweering"
                ],
                "domain": [
                    "Online Scheduling",
                    "Learning-Augmented Algorithms",
                    "Graph Theory",
                    "Approximation Algorithms"
                ],
                "publications": [
                    {
                        "title": "Permutation Predictions for Non-Clairvoyant Scheduling",
                        "abstract": "In non-clairvoyant scheduling, the task is to find an online strategy for scheduling jobs with a priori unknown processing requirements with the objective to minimize the total (weighted) completion time. We revisit this well-studied problem in a recently popular learning-augmented setting that integrates (untrusted) predictions in online algorithm design. While previous works used predictions on processing requirements, we propose a new prediction model, which provides a relative order of jobs which could be seen as predicting algorithmic actions rather than parts of the unknown input. We show that these predictions have desired properties, admit a natural error measure as well as algorithms with strong performance guarantees and that they are learnable in both, theory and practice. We generalize the algorithmic framework proposed in the seminal paper by Kumar et al. (NeurIPS'18) and present the first learning-augmented scheduling results for weighted jobs and unrelated machines. We demonstrate in empirical experiments the practicability and superior performance compared to the previously suggested single-machine algorithms."
                    },
                    {
                        "title": "Minimalistic Predictions to Schedule Jobs with Online Precedence Constraints",
                        "abstract": "We consider non-clairvoyant scheduling with online precedence constraints, where an algorithm is oblivious to any job dependencies and learns about a job only if all of its predecessors have been completed. Given strong impossibility results in classical competitive analysis, we investigate the problem in a learning-augmented setting, where an algorithm has access to predictions without any quality guarantee. We discuss different prediction models: novel problem-specific models as well as general ones, which have been proposed in previous works. We present lower bounds and algorithmic upper bounds for different precedence topologies, and thereby give a structured overview on which and how additional (possibly erroneous) information helps for designing better algorithms. Along the way, we also improve bounds on traditional competitive ratios for existing algorithms."
                    },
                    {
                        "title": "Speed-Oblivious Online Scheduling: Knowing (Precise) Speeds is not Necessary",
                        "abstract": "We consider online scheduling on unrelated (heterogeneous) machines in a speed-oblivious setting, where an algorithm is unaware of the exact job-dependent processing speeds. We show strong impossibility results for clairvoyant and non-clairvoyant algorithms and overcome them in models inspired by practical settings: (i) we provide competitive learning-augmented algorithms, assuming that (possibly erroneous) predictions on the speeds are given, and (ii) we provide competitive algorithms for the speed-ordered model, where a single global order of machines according to their unknown job-dependent speeds is known. We prove strong theoretical guarantees and evaluate our findings on a representative heterogeneous multi-core processor. These seem to be the first empirical results for scheduling algorithms with predictions that are evaluated in a non-synthetic hardware environment."
                    },
                    {
                        "title": "Robustification of Online Graph Exploration Methods",
                        "abstract": "Exploring unknown environments is a fundamental task in many domains, e.g., robot navigation, network security, and internet search. We initiate the study of a learning-augmented variant of the classical, notoriously hard online graph exploration problem by adding access to machine-learned predictions. We propose an algorithm that naturally integrates predictions into the well-known Nearest Neighbor (NN) algorithm and significantly outperforms any known online algorithm if the prediction is of high accuracy while maintaining good guarantees when the prediction is of poor quality. We provide theoretical worst-case bounds that gracefully degrade with the prediction error, and we complement them by computational experiments that confirm our results. Further, we extend our concept to a general framework to robustify algorithms. By interpolating carefully between a given algorithm and NN, we prove new performance bounds that leverage the individual good performance on particular inputs while establishing robustness to arbitrary inputs."
                    },
                    {
                        "title": "A Universal Error Measure for Input Predictions Applied to Online Graph Problems",
                        "abstract": "We introduce a novel measure for quantifying the error in input predictions. The error is based on a minimum-cost hyperedge cover in a suitably defined hypergraph and provides a general template which we apply to online graph problems. The measure captures errors due to absent predicted requests as well as unpredicted actual requests; hence, predicted and actual inputs can be of arbitrary size. We achieve refined performance guarantees for previously studied network design problems in the online-list model, such as Steiner tree and facility location. Further, we initiate the study of learning-augmented algorithms for online routing problems, such as the online traveling salesperson problem and the online dial-a-ride problem, where (transportation) requests arrive over time (online-time model). We provide a general algorithmic framework and we give error-dependent performance bounds that improve upon known worst-case barriers, when given accurate predictions, at the cost of slightly increased worst-case bounds when given predictions of arbitrary quality."
                    },
                    {
                        "title": "Double Coverage with Machine-Learned Advice",
                        "abstract": "We study the fundamental online k-server problem in a learning-augmented setting. While in the traditional online model, an algorithm has no information about the request sequence, we assume that there is given some advice (e.g. machine-learned predictions) on an algorithm's decision. There is, however, no guarantee on the quality of the prediction and it might be far from being correct.   Our main result is a learning-augmented variation of the well-known Double Coverage algorithm for k-server on the line (Chrobak et al., SIDMA 1991) in which we integrate predictions as well as our trust into their quality. We give an error-dependent competitive ratio, which is a function of a user-defined confidence parameter, and which interpolates smoothly between an optimal consistency, the performance in case that all predictions are correct, and the best-possible robustness regardless of the prediction quality. When given good predictions, we improve upon known lower bounds for online algorithms without advice. We further show that our algorithm achieves for any k an almost optimal consistency-robustness tradeoff, within a class of deterministic algorithms respecting local and memoryless properties.   Our algorithm outperforms a previously proposed (more general) learning-augmented algorithm. It is remarkable that the previous algorithm crucially exploits memory, whereas our algorithm is memoryless. Finally, we demonstrate in experiments the practicability and the superior performance of our algorithm on real-world data."
                    },
                    {
                        "title": "Two-Edge Connectivity via Pac-Man Gluing",
                        "abstract": "We study the 2-edge-connected spanning subgraph (2-ECSS) problem: Given a graph $G$, compute a connected subgraph $H$ of $G$ with the minimum number of edges such that $H$ is spanning, i.e., $V(H) = V(G)$, and $H$ is 2-edge-connected, i.e., $H$ remains connected upon the deletion of any single edge, if such an $H$ exists. The $2$-ECSS problem is known to be NP-hard. In this work, we provide a polynomial-time $(\\frac 5 4 + \\varepsilon)$-approximation for the problem for an arbitrarily small $\\varepsilon>0$, improving the previous best approximation ratio of $\\frac{13}{10}+\\varepsilon$.   Our improvement is based on two main innovations: First, we reduce solving the problem on general graphs to solving it on structured graphs with high vertex connectivity. This high vertex connectivity ensures the existence of a 4-matching across any bipartition of the vertex set with at least 10 vertices in each part. Second, we exploit this property in a later gluing step, where isolated 2-edge-connected components need to be merged without adding too many edges. Using the 4-matching property, we can repeatedly glue a huge component (containing at least 10 vertices) to other components. This step is reminiscent of the Pac-Man game, where a Pac-Man (a huge component) consumes all the dots (other components) as it moves through a maze. These two innovations lead to a significantly simpler algorithm and analysis for the gluing step compared to the previous best approximation algorithm, which required a long and tedious case analysis."
                    },
                    {
                        "title": "The Power of Proportional Fairness for Non-Clairvoyant Scheduling under Polyhedral Constraints",
                        "abstract": "The Polytope Scheduling Problem (PSP) was introduced by Im, Kulkarni, and Munagala (JACM 2018) as a very general abstraction of resource allocation over time and captures many well-studied problems including classical unrelated machine scheduling, multidimensional scheduling, and broadcast scheduling. In PSP, jobs with different arrival times receive processing rates that are subject to arbitrary packing constraints. An elegant and well-known algorithm for instantaneous rate allocation with good fairness and efficiency properties is the Proportional Fairness algorithm (PF), which was analyzed for PSP by Im et al.   We drastically improve the analysis of the PF algorithm for both the general PSP and several of its important special cases subject to the objective of minimizing the sum of weighted completion times. We reduce the upper bound on the competitive ratio from 128 to 27 for general PSP and to 4 for the prominent class of monotone PSP. For certain heterogeneous machine environments we even close the substantial gap to the lower bound of 2 for non-clairvoyant scheduling. Our analysis also gives the first polynomial-time improvements over the nearly 30-year-old bounds on the competitive ratio of the doubling framework by Hall, Shmoys, and Wein (SODA 1996) for clairvoyant online preemptive scheduling on unrelated machines. Somewhat surprisingly, we achieve this improvement by a non-clairvoyant algorithm, thereby demonstrating that non-clairvoyance is not a (significant) hurdle.   Our improvements are based on exploiting monotonicity properties of PSP, providing tight dual fitting arguments on structured instances, and showing new additivity properties on the optimal objective value for scheduling on unrelated machines. Finally, we establish new connections of PF to matching markets, and thereby provide new insights on equilibria and their computational complexity."
                    },
                    {
                        "title": "Elimination distance to bounded degree on planar graphs",
                        "abstract": "We study the graph parameter elimination distance to bounded degree, which was introduced by Bulian and Dawar in their study of the parameterized complexity of the graph isomorphism problem. We prove that the problem is fixed-parameter tractable on planar graphs, that is, there exists an algorithm that given a planar graph $G$ and integers $d$ and $k$ decides in time $f(k,d)\\cdot n^c$ for a computable function~$f$ and constant $c$ whether the elimination distance of $G$ to the class of degree $d$ graphs is at most $k$."
                    },
                    {
                        "title": "Santa Claus meets Makespan and Matroids: Algorithms and Reductions",
                        "abstract": "In this paper we study the relation of two fundamental problems in scheduling and fair allocation: makespan minimization on unrelated parallel machines and max-min fair allocation, also known as the Santa Claus problem. For both of these problems the best approximation factor is a notorious open question; more precisely, whether there is a better-than-2 approximation for the former problem and whether there is a constant approximation for the latter.   While the two problems are intuitively related and history has shown that techniques can often be transferred between them, no formal reductions are known. We first show that an affirmative answer to the open question for makespan minimization implies the same for the Santa Claus problem by reducing the latter problem to the former. We also prove that for problem instances with only two input values both questions are equivalent.   We then move to a special case called ``restricted assignment'', which is well studied in both problems. Although our reductions do not maintain the characteristics of this special case, we give a reduction in a slight generalization, where the jobs or resources are assigned to multiple machines or players subject to a matroid constraint and in addition we have only two values. This draws a similar picture as before: equivalence for two values and the general case of Santa Claus can only be easier than makespan minimization. To complete the picture, we give an algorithm for our new matroid variant of the Santa Claus problem using a non-trivial extension of the local search method from restricted assignment. Thereby we unify, generalize, and improve several previous results. We believe that this matroid generalization may be of independent interest and provide several sample applications."
                    }
                ]
            },
            "68bf3283-f734-416f-b008-919a43f010d9": {
                "pk": "68bf3283-f734-416f-b008-919a43f010d9",
                "name": "Zhenwei Liu",
                "collaborators": [
                    "Ali Saberi",
                    "Anton A. Stoorvogel"
                ],
                "domain": [
                    "Multi-Agent Systems",
                    "Control Theory",
                    "Distributed Systems",
                    "Synchronization"
                ],
                "publications": [
                    {
                        "title": "Scale-free linear protocol design for global regulated state synchronization of discrete-time double-integrator multi-agent systems subject to actuator saturation",
                        "abstract": "This paper studies global regulated state synchronization of discrete-time double-integrator multi-agent systems subject to actuator saturation by utilizing localized information exchange. We propose a scale-free linear protocol that achieves global regulated state synchronization for any network with arbitrary number of agents and arbitrarily directed communication graph that has a path between each agent and exosystem which generates the reference trajectory."
                    },
                    {
                        "title": "Scale-free Non-collaborative Linear Protocol Design for A Class of Homogeneous Multi-agent Systems",
                        "abstract": "In this paper, we have focused on identifying a class of continuous- and discrete-time MAS for which a scale-free non-collaborative (i.e., scale-free fully distributed) linear protocol design is developed. We have identified conditions on agent models that enable us to design scalable linear protocols. Moreover, we show that these conditions are necessary if the agents are single input and single output. We also provide a complete design of scalable protocols for this class."
                    },
                    {
                        "title": "Scalable global state synchronization of discrete-time double integrator multi-agent systems with input saturation via linear protocol (Completed Version)",
                        "abstract": "This paper studies scalable global state synchronization of discrete-time double integrator multi-agent systems in presence of input saturation based on localized information exchange. A scale-free collaborative linear dynamic protocols design methodology is developed for discrete-time multi-agent systems with both full and partial-state couplings. And the protocol design methodology does not need any knowledge of the directed network topology and the spectrum of the associated Laplacian matrix. Meanwhile, the protocols are parametric based on a parameter set in which the designed protocols can guarantee the global synchronization result. Furthermore, the proposed protocol is scalable and achieves synchronization for any arbitrary number of agents."
                    },
                    {
                        "title": "The role of local bounds on neighborhoods in the network for scale-free state synchronization of multi-agent systems",
                        "abstract": "This paper provides necessary and sufficient conditions for the existence of solutions to the state synchronization problem of homogeneous multi-agent systems (MAS) via scale-free linear dynamic non-collaborative protocol for both continuous- and discrete-time. These conditions guarantee for which class of MAS, one can achieve scale-free state synchronization. We investigate protocol design with and without utilizing local bounds on neighborhood. The results show that the availability of local bounds on neighborhoods plays a key role."
                    },
                    {
                        "title": "Fault-tolerant properties of scale-free linear protocols for synchronization of homogeneous multi-agent systems",
                        "abstract": "Originally, protocols were designed for multi-agent systems (MAS) using information about the network which might not be available. Recently, there has been a focus on scale-free synchronization where the protocol is designed without any prior information about the network.   As long as the network contains a directed spanning tree, a scale-free protocol guarantees that the network achieves synchronization.   If there is no directed spanning tree then synchronization cannot be achieved. But what happens when these scale-free protocols are applied to such a network where the directed spanning tree no longer exists? This paper establishes that the network decomposes into a number of basic bicomponents which achieves synchronization among all nodes in this basic bicomponent. On the other hand, nodes which are not part of any basic bicomponent converge to a weighted average of the synchronized trajectories of the basic bicomponents. The weights are independent of the initial conditions and are independent of the designed protocol."
                    }
                ]
            },
            "c12bcd9d-0e1b-401b-896e-b7bd176d0551": {
                "pk": "c12bcd9d-0e1b-401b-896e-b7bd176d0551",
                "name": "Nicole Megow",
                "collaborators": [
                    "Alexander Lindermayr",
                    "Lin Chen",
                    "Kevin Schewior",
                    "Jens Schl\u00f6ter",
                    "Lukas N\u00f6lke",
                    "Franziska Eberle",
                    "Andreas Wiese",
                    "Jos\u00e9 Verschae",
                    "Felix Hommelsheim",
                    "Bertrand Simon"
                ],
                "domain": [
                    "Online Scheduling",
                    "Competitive Analysis",
                    "Algorithm Design",
                    "Combinatorial Optimization"
                ],
                "publications": [
                    {
                        "title": "Competitive-Ratio Approximation Schemes for Minimizing the Makespan in the Online-List Model",
                        "abstract": "We consider online scheduling on multiple machines for jobs arriving one-by-one with the objective of minimizing the makespan. For any number of identical parallel or uniformly related machines, we provide a competitive-ratio approximation scheme that computes an online algorithm whose competitive ratio is arbitrarily close to the best possible competitive ratio. We also determine this value up to any desired accuracy. This is the first application of competitive-ratio approximation schemes in the online-list model. The result proves the applicability of the concept in different online models. We expect that it fosters further research on other online problems."
                    },
                    {
                        "title": "Permutation Predictions for Non-Clairvoyant Scheduling",
                        "abstract": "In non-clairvoyant scheduling, the task is to find an online strategy for scheduling jobs with a priori unknown processing requirements with the objective to minimize the total (weighted) completion time. We revisit this well-studied problem in a recently popular learning-augmented setting that integrates (untrusted) predictions in online algorithm design. While previous works used predictions on processing requirements, we propose a new prediction model, which provides a relative order of jobs which could be seen as predicting algorithmic actions rather than parts of the unknown input. We show that these predictions have desired properties, admit a natural error measure as well as algorithms with strong performance guarantees and that they are learnable in both, theory and practice. We generalize the algorithmic framework proposed in the seminal paper by Kumar et al. (NeurIPS'18) and present the first learning-augmented scheduling results for weighted jobs and unrelated machines. We demonstrate in empirical experiments the practicability and superior performance compared to the previously suggested single-machine algorithms."
                    },
                    {
                        "title": "Set Selection under Explorable Stochastic Uncertainty via Covering Techniques",
                        "abstract": "Given subsets of uncertain values, we study the problem of identifying the subset of minimum total value (sum of the uncertain values) by querying as few values as possible. This set selection problem falls into the field of explorable uncertainty and is of intrinsic importance therein as it implies strong adversarial lower bounds for a wide range of interesting combinatorial problems such as knapsack and matchings. We consider a stochastic problem variant and give algorithms that, in expectation, improve upon these adversarial lower bounds. The key to our results is to prove a strong structural connection to a seemingly unrelated covering problem with uncertainty in the constraints via a linear programming formulation. We exploit this connection to derive an algorithmic framework that can be used to solve both problems under uncertainty, obtaining nearly tight bounds on the competitive ratio. This is the first non-trivial stochastic result concerning the sum of unknown values without further structure known for the set. With our novel methods, we lay the foundations for solving more general problems in the area of explorable uncertainty."
                    },
                    {
                        "title": "Dual techniques for scheduling on a machine with varying speed",
                        "abstract": "We study scheduling problems on a machine with varying speed. Assuming a known speed function we ask for a cost-efficient scheduling solution. Our main result is a PTAS for minimizing the total weighted completion time in this setting. This also implies a PTAS for the closely related problem of scheduling to minimize generalized global cost functions. The key to our results is a re-interpretation of the problem within the well-known two-dimensional Gantt chart: instead of the standard approach of scheduling in the {\\em time-dimension}, we construct scheduling solutions in the weight-dimension.   We also consider a dynamic problem variant in which deciding upon the speed is part of the scheduling problem and we are interested in the tradeoff between scheduling cost and speed-scaling cost, which is typically the energy consumption. We observe that the optimal order is independent of the energy consumption and that the problem can be reduced to the setting where the speed of the machine is fixed, and thus admits a PTAS. Furthermore, we provide an FPTAS for the NP-hard problem variant in which the machine can run only on a fixed number of discrete speeds. Finally, we show how our results can be used to obtain a~$(2+\\eps)$-approximation for scheduling preemptive jobs with release dates on multiple identical parallel machines."
                    },
                    {
                        "title": "An O(m^2 log m)-Competitive Algorithm for Online Machine Minimization",
                        "abstract": "We consider the online machine minimization problem in which jobs with hard deadlines arrive online over time at their release dates. The task is to determine a feasible schedule on a minimum number of machines. Our main result is a general O(m^2 log m)-competitive algorithm for the preemptive online problem, where m is the optimal number of machines used in an offline solution. This is the first improvement on an O(log (p_max/p_min))-competitive algorithm by Phillips et al. (STOC 1997), which was to date the best known algorithm even when m=2. Our algorithm is O(1)-competitive for any m that is bounded by a constant. To develop the algorithm, we investigate two complementary special cases of the problem, namely, laminar instances and agreeable instances, for which we provide an O(log m)-competitive and an O(1)-competitive algorithm, respectively. Our O(1)-competitive algorithm for agreeable instances actually produces a non-preemptive schedule, which is of its own interest as there exists a strong lower bound of n, the number of jobs, for the general non-preemptive online machine minimization problem by Saha (FSTTCS 2013), which even holds for laminar instances."
                    },
                    {
                        "title": "Matching Augmentation via Simultaneous Contractions",
                        "abstract": "We consider the matching augmentation problem (MAP), where a matching of a graph needs to be extended into a $2$-edge-connected spanning subgraph by adding the minimum number of edges to it. We present a polynomial-time algorithm with an approximation ratio of $13/8 = 1.625$ improving upon an earlier $5/3$-approximation. The improvement builds on a new $\\alpha$-approximation preserving reduction for any $\\alpha\\geq 3/2$ from arbitrary MAP instances to well-structured instances that do not contain certain forbidden structures like parallel edges, small separators, and contractible subgraphs. We further introduce, as key ingredients, the technique of repeated simultaneous contractions and provide improved lower bounds for instances that cannot be contracted."
                    },
                    {
                        "title": "Minimalistic Predictions to Schedule Jobs with Online Precedence Constraints",
                        "abstract": "We consider non-clairvoyant scheduling with online precedence constraints, where an algorithm is oblivious to any job dependencies and learns about a job only if all of its predecessors have been completed. Given strong impossibility results in classical competitive analysis, we investigate the problem in a learning-augmented setting, where an algorithm has access to predictions without any quality guarantee. We discuss different prediction models: novel problem-specific models as well as general ones, which have been proposed in previous works. We present lower bounds and algorithmic upper bounds for different precedence topologies, and thereby give a structured overview on which and how additional (possibly erroneous) information helps for designing better algorithms. Along the way, we also improve bounds on traditional competitive ratios for existing algorithms."
                    },
                    {
                        "title": "Speed-Oblivious Online Scheduling: Knowing (Precise) Speeds is not Necessary",
                        "abstract": "We consider online scheduling on unrelated (heterogeneous) machines in a speed-oblivious setting, where an algorithm is unaware of the exact job-dependent processing speeds. We show strong impossibility results for clairvoyant and non-clairvoyant algorithms and overcome them in models inspired by practical settings: (i) we provide competitive learning-augmented algorithms, assuming that (possibly erroneous) predictions on the speeds are given, and (ii) we provide competitive algorithms for the speed-ordered model, where a single global order of machines according to their unknown job-dependent speeds is known. We prove strong theoretical guarantees and evaluate our findings on a representative heterogeneous multi-core processor. These seem to be the first empirical results for scheduling algorithms with predictions that are evaluated in a non-synthetic hardware environment."
                    },
                    {
                        "title": "A New Approach to Online Scheduling: Approximating the Optimal Competitive Ratio",
                        "abstract": "We propose a new approach to competitive analysis in online scheduling by introducing the novel concept of competitive-ratio approximation schemes. Such a scheme algorithmically constructs an online algorithm with a competitive ratio arbitrarily close to the best possible competitive ratio for any online algorithm. We study the problem of scheduling jobs online to minimize the weighted sum of completion times on parallel, related, and unrelated machines, and we derive both deterministic and randomized algorithms which are almost best possible among all online algorithms of the respective settings. We also generalize our techniques to arbitrary monomial cost functions and apply them to the makespan objective. Our method relies on an abstract characterization of online algorithms combined with various simplifications and transformations. We also contribute algorithmic means to compute the actual value of the best possi- ble competitive ratio up to an arbitrary accuracy. This strongly contrasts all previous manually obtained competitiveness results for algorithms and, most importantly, it reduces the search for the optimal com- petitive ratio to a question that a computer can answer. We believe that our concept can also be applied to many other problems and yields a new perspective on online algorithms in general."
                    },
                    {
                        "title": "New Results on Online Resource Minimization",
                        "abstract": "We consider the online resource minimization problem in which jobs with hard deadlines arrive online over time at their release dates. The task is to determine a feasible schedule on a minimum number of machines. We rigorously study this problem and derive various algorithms with small constant competitive ratios for interesting restricted problem variants. As the most important special case, we consider scheduling jobs with agreeable deadlines. We provide the first constant ratio competitive algorithm for the non-preemptive setting, which is of particular interest with regard to the known strong lower bound of n for the general problem. For the preemptive setting, we show that the natural algorithm LLF achieves a constant ratio for agreeable jobs, while for general jobs it has a lower bound of Omega(n^(1/3)). We also give an O(log n)-competitive algorithm for the general preemptive problem, which improves upon the known O(p_max/p_min)-competitive algorithm. Our algorithm maintains a dynamic partition of the job set into loose and tight jobs and schedules each (temporal) subset individually on separate sets of machines. The key is a characterization of how the decrease in the relative laxity of jobs influences the optimum number of machines. To achieve this we derive a compact expression of the optimum value, which might be of independent interest. We complement the general algorithmic result by showing lower bounds that rule out that other known algorithms may yield a similar performance guarantee."
                    },
                    {
                        "title": "Online Minimum Cost Matching on the Line with Recourse",
                        "abstract": "In online minimum cost matching on the line, $n$ requests appear one by one and have to be matched immediately and irrevocably to a given set of servers, all on the real line. The goal is to minimize the sum of distances from the requests to their respective servers. Despite all research efforts, it remains an intriguing open question whether there exists an $O(1)$-competitive algorithm. The best known online algorithm by Raghvendra [SoCG18] achieves a competitive factor of $\\Theta(\\log n)$. This result matches a lower bound of $\\Omega(\\log n)$ [Latin18] that holds for a quite large class of online algorithms, including all deterministic algorithms in the literature.   In this work we approach the problem in a recourse model where we allow to revoke online decisions to some extent. We show an $O(1)$-competitive algorithm for online matching on the line that uses at most $O(n\\log n)$ reassignments. This is the first non-trivial result for min-cost bipartite matching with recourse. For so-called alternating instances, with no more than one request between two servers, we obtain a near-optimal result. We give a $(1+\\varepsilon)$-competitive algorithm that reassigns any request at most $O(\\varepsilon^{-1.001})$ times. This special case is interesting as the aforementioned quite general lower bound $\\Omega(\\log n)$ holds for such instances."
                    },
                    {
                        "title": "Double Coverage with Machine-Learned Advice",
                        "abstract": "We study the fundamental online k-server problem in a learning-augmented setting. While in the traditional online model, an algorithm has no information about the request sequence, we assume that there is given some advice (e.g. machine-learned predictions) on an algorithm's decision. There is, however, no guarantee on the quality of the prediction and it might be far from being correct.   Our main result is a learning-augmented variation of the well-known Double Coverage algorithm for k-server on the line (Chrobak et al., SIDMA 1991) in which we integrate predictions as well as our trust into their quality. We give an error-dependent competitive ratio, which is a function of a user-defined confidence parameter, and which interpolates smoothly between an optimal consistency, the performance in case that all predictions are correct, and the best-possible robustness regardless of the prediction quality. When given good predictions, we improve upon known lower bounds for online algorithms without advice. We further show that our algorithm achieves for any k an almost optimal consistency-robustness tradeoff, within a class of deterministic algorithms respecting local and memoryless properties.   Our algorithm outperforms a previously proposed (more general) learning-augmented algorithm. It is remarkable that the previous algorithm crucially exploits memory, whereas our algorithm is memoryless. Finally, we demonstrate in experiments the practicability and the superior performance of our algorithm on real-world data."
                    },
                    {
                        "title": "The Power of Proportional Fairness for Non-Clairvoyant Scheduling under Polyhedral Constraints",
                        "abstract": "The Polytope Scheduling Problem (PSP) was introduced by Im, Kulkarni, and Munagala (JACM 2018) as a very general abstraction of resource allocation over time and captures many well-studied problems including classical unrelated machine scheduling, multidimensional scheduling, and broadcast scheduling. In PSP, jobs with different arrival times receive processing rates that are subject to arbitrary packing constraints. An elegant and well-known algorithm for instantaneous rate allocation with good fairness and efficiency properties is the Proportional Fairness algorithm (PF), which was analyzed for PSP by Im et al.   We drastically improve the analysis of the PF algorithm for both the general PSP and several of its important special cases subject to the objective of minimizing the sum of weighted completion times. We reduce the upper bound on the competitive ratio from 128 to 27 for general PSP and to 4 for the prominent class of monotone PSP. For certain heterogeneous machine environments we even close the substantial gap to the lower bound of 2 for non-clairvoyant scheduling. Our analysis also gives the first polynomial-time improvements over the nearly 30-year-old bounds on the competitive ratio of the doubling framework by Hall, Shmoys, and Wein (SODA 1996) for clairvoyant online preemptive scheduling on unrelated machines. Somewhat surprisingly, we achieve this improvement by a non-clairvoyant algorithm, thereby demonstrating that non-clairvoyance is not a (significant) hurdle.   Our improvements are based on exploiting monotonicity properties of PSP, providing tight dual fitting arguments on structured instances, and showing new additivity properties on the optimal objective value for scheduling on unrelated machines. Finally, we establish new connections of PF to matching markets, and thereby provide new insights on equilibria and their computational complexity."
                    },
                    {
                        "title": "On Hop-Constrained Steiner Trees in Tree-Like Metrics",
                        "abstract": "We consider the problem of computing a Steiner tree of minimum cost under a hop constraint which requires the depth of the tree to be at most $k$. Our main result is an exact algorithm for metrics induced by graphs with bounded treewidth that runs in time $n^{O(k)}$. For the special case of a path, we give a simple algorithm that solves the problem in polynomial time, even if $k$ is part of the input. The main result can be used to obtain, in quasi-polynomial time, a near-optimal solution that violates the $k$-hop constraint by at most one hop for more general metrics induced by graphs of bounded highway dimension and bounded doubling dimension. For non-metric graphs, we rule out an $o(\\log n)$-approximation, assuming P$\\neq$NP even when relaxing the hop constraint by any additive constant."
                    },
                    {
                        "title": "Robustification of Online Graph Exploration Methods",
                        "abstract": "Exploring unknown environments is a fundamental task in many domains, e.g., robot navigation, network security, and internet search. We initiate the study of a learning-augmented variant of the classical, notoriously hard online graph exploration problem by adding access to machine-learned predictions. We propose an algorithm that naturally integrates predictions into the well-known Nearest Neighbor (NN) algorithm and significantly outperforms any known online algorithm if the prediction is of high accuracy while maintaining good guarantees when the prediction is of poor quality. We provide theoretical worst-case bounds that gracefully degrade with the prediction error, and we complement them by computational experiments that confirm our results. Further, we extend our concept to a general framework to robustify algorithms. By interpolating carefully between a given algorithm and NN, we prove new performance bounds that leverage the individual good performance on particular inputs while establishing robustness to arbitrary inputs."
                    },
                    {
                        "title": "Recoverable Robust Optimization with Commitment",
                        "abstract": "We propose a model for recoverable robust optimization with commitment. Given a combinatorial optimization problem and uncertainty about elements that may fail, we ask for a robust solution that, after the failing elements are revealed, can be augmented in a limited way. However, we commit to preserve the remaining elements of the initial solution. We consider different polynomial-time solvable combinatorial optimization problems and settle the computational complexity of their robust counterparts with commitment. We show for the weighted matroid basis problem that an optimal solution to the nominal problem is also optimal for its robust counterpart. Indeed, matroids are provably the only structures with this strong property. Robust counterparts of other problems are NP-hard such as the matching and the stable set problem, even in bipartite graphs. However, we establish polynomial-time algorithms for the robust counterparts of the unweighted stable set problem in bipartite graphs and the weighted stable set problem in interval graphs, also known as the interval scheduling problem."
                    },
                    {
                        "title": "Online Throughput Maximization on Unrelated Machines: Commitment is No Burden",
                        "abstract": "We consider a fundamental online scheduling problem in which jobs with processing times and deadlines arrive online over time at their release dates. The task is to determine a feasible preemptive schedule on a single or multiple possibly unrelated machines that maximizes the number of jobs that complete before their deadline. Due to strong impossibility results for competitive analysis on a single machine, we require that jobs contain some slack $\\varepsilon>0$, which means that the feasible time window for scheduling a job is at least $1+\\varepsilon$ times its processing time on each eligible machine. Our contribution is two-fold: (i) We give the first non-trivial online algorithms for throughput maximization on unrelated machines, and (ii), this is the main focus of our paper, we answer the question on how to handle commitment requirements which enforce that a scheduler has to guarantee at a certain point in time the completion of admitted jobs. This is very relevant, e.g., in providing cloud-computing services, and disallows last-minute rejections of critical tasks. We present an algorithm for unrelated machines that is $\\Theta\\big(\\frac{1}\\varepsilon\\big )$-competitive when the scheduler must commit upon starting a job. Somewhat surprisingly, this is the same optimal performance bound (up to constants) as for scheduling without commitment on a single machine. If commitment decisions must be made before a job's slack becomes less than a $\\delta$-fraction of its size, we prove a competitive ratio of $\\mathcal{O}\\big(\\frac{1}{\\varepsilon - \\delta}\\big)$ for $0 < \\delta < \\varepsilon$. This result nicely interpolates between commitment upon starting a job and commitment upon arrival. For the latter commitment model, it is known that no (randomized) online algorithm admits any bounded competitive ratio."
                    },
                    {
                        "title": "Packing a Knapsack of Unknown Capacity",
                        "abstract": "We study the problem of packing a knapsack without knowing its capacity. Whenever we attempt to pack an item that does not fit, the item is discarded; if the item fits, we have to include it in the packing. We show that there is always a policy that packs a value within factor 2 of the optimum packing, irrespective of the actual capacity. If all items have unit density, we achieve a factor equal to the golden ratio. Both factors are shown to be best possible. In fact, we obtain the above factors using packing policies that are universal in the sense that they fix a particular order of the items and try to pack the items in this order, independent of the observations made while packing. We give efficient algorithms computing these policies. On the other hand, we show that, for any alpha>1, the problem of deciding whether a given universal policy achieves a factor of alpha is coNP-complete. If alpha is part of the input, the same problem is shown to be coNP-complete for items with unit densities. Finally, we show that it is coNP-hard to decide, for given alpha, whether a set of items admits a universal policy with factor alpha, even if all items have unit densities."
                    },
                    {
                        "title": "A general framework for handling commitment in online throughput maximization",
                        "abstract": "We study a fundamental online job admission problem where jobs with deadlines arrive online over time at their release dates, and the task is to determine a preemptive single-server schedule which maximizes the number of jobs that complete on time. To circumvent known impossibility results, we make a standard slackness assumption by which the feasible time window for scheduling a job is at least $1+\\varepsilon$ times its processing time, for some $\\varepsilon>0$. We quantify the impact that different provider commitment requirements have on the performance of online algorithms. Our main contribution is one universal algorithmic framework for online job admission both with and without commitments. Without commitment, our algorithm with a competitive ratio of $O(1/\\varepsilon)$ is the best possible (deterministic) for this problem. For commitment models, we give the first non-trivial performance bounds. If the commitment decisions must be made before a job's slack becomes less than a $\\delta$-fraction of its size, we prove a competitive ratio of $O(\\varepsilon/((\\varepsilon-\\delta)\\delta^2))$, for $0<\\delta<\\varepsilon$. When a provider must commit upon starting a job, our bound is $O(1/\\varepsilon^2)$. Finally, we observe that for scheduling with commitment the restriction to the `unweighted' throughput model is essential; if jobs have individual weights, we rule out competitive deterministic algorithms."
                    },
                    {
                        "title": "Optimal Algorithms for Scheduling under Time-of-Use Tariffs",
                        "abstract": "We consider a natural generalization of classical scheduling problems in which using a time unit for processing a job causes some time-dependent cost which must be paid in addition to the standard scheduling cost. We study the scheduling objectives of minimizing the makespan and the sum of (weighted) completion times. It is not difficult to derive a polynomial-time algorithm for preemptive scheduling to minimize the makespan on unrelated machines. The problem of minimizing the total (weighted) completion time is considerably harder, even on a single machine. We present a polynomial-time algorithm that computes for any given sequence of jobs an optimal schedule, i.e., the optimal set of time-slots to be used for scheduling jobs according to the given sequence. This result is based on dynamic programming using a subtle analysis of the structure of optimal solutions and a potential function argument. With this algorithm, we solve the unweighted problem optimally in polynomial time. For the more general problem, in which jobs may have individual weights, we develop a polynomial-time approximation scheme (PTAS) based on a dual scheduling approach introduced for scheduling on a machine of varying speed. As the weighted problem is strongly NP-hard, our PTAS is the best possible approximation we can hope for."
                    }
                ]
            },
            "a36daf90-ec29-46c5-a72d-01fa95e904b3": {
                "pk": "a36daf90-ec29-46c5-a72d-01fa95e904b3",
                "name": "Jens Schl\u00f6ter",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2406.17271": {
        "paper_data": {
            "title": "DARG: Dynamic Evaluation of Large Language Models via Adaptive Reasoning Graph",
            "url": "http://arxiv.org/abs/2406.17271v1",
            "arxiv_id": "2406.17271",
            "authors": [
                "Zhehao Zhang",
                "Jiaao Chen",
                "Diyi Yang"
            ],
            "abstract": "The current paradigm of evaluating Large Language Models (LLMs) through static benchmarks comes with significant limitations, such as vulnerability to data contamination and a lack of adaptability to the evolving capabilities of LLMs. Therefore, evaluation methods that can adapt and generate evaluation data with controlled complexity are urgently needed. In this work, we introduce Dynamic Evaluation of LLMs via Adaptive Reasoning Graph Evolvement (DARG) to dynamically extend current benchmarks with controlled complexity and diversity. Specifically, we first extract the reasoning graphs of data points in current benchmarks and then perturb the reasoning graphs to generate novel testing data. Such newly generated test samples can have different levels of complexity while maintaining linguistic diversity similar to the original benchmarks. We further use a code-augmented LLM to ensure the label correctness of newly generated data. We apply our DARG framework to diverse reasoning tasks in four domains with 15 state-of-the-art LLMs. Experimental results show that almost all LLMs experience a performance decrease with increased complexity and certain LLMs exhibit significant drops. Additionally, we find that LLMs exhibit more biases when being evaluated via the data generated by DARG with higher complexity levels. These observations provide useful insights into how to dynamically and adaptively evaluate LLMs. The code is available at https://github.com/SALT-NLP/DARG.",
            "introduction": "   1 Introduction  Large language models (LLMs) have recently attained exceptional performance across a wide range of tasks [10, 2, 11] by showing substantial evaluation results on static benchmark datasets [35, 16, 14] where their test data points are open-sourced and unchanged. Although these widely used benchmarks are generally of high-quality, they may suffer from the following issues [119]: (1) Data contamination [8, 66, 104, 29], which refers to the potential overlap between LLMs\u2019 training corpus and benchmarks\u2019 data points. This raises concerns about whether LLMs are merely memorizing and overfitting these benchmarks instead of learning how to solve the tasks [112], which may lead to poor generalization[67, 13, 9]. (2) Static datasets only have fixed complexity and lack the flexibility to evolve. As LLMs are developing and scaling up rapidly, existing static benchmarks may fail to align with their increasing capabilities, as the complexity of current benchmarks remains unchanged [25].   To address these issues, prior work has introduced template-based methods [119] to generate evaluation samples with different complexities for mathematical and logical reasoning tasks. However, these rule-based generated samples are synthetic and limited to a specific set of tasks, lacking linguistic diversity compared to existing benchmarks. Another line of work involves prompting LLMs to directly modify the current evaluation data such as DyVal 2 [120] and Benchmark Self-Evolving [96] which utilize LLMs with various prompting strategies to perturb existing data. Despite better adaptation to existing benchmarks, these methods usually have low controllability and suffer from LLMs\u2019 instability, which makes it difficult to verify the quality and correctness of the newly generated data points. Therefore, it remains a challenge to dynamically and adaptively generate novel test samples with controlled complexity and diversity.   To fill in this gap, in this work, we propose DARG, a Dynamic Evaluation of LLMs via Adaptive Reasoning Graph. Unlike previous work that generates test data through templates or designed prompts [119, 96], we evolve existing benchmarks based on the reasoning111Note that we use the term \u201creasoning\u201d to refer to the potential rationales or intermediate steps that models might follow to make inferences, not the exact reasoning behind the model\u2019s inferences.  graphs that represent the underlying structures of basic reasoning components necessary for problem-solving. Specifically, we first construct the reasoning graphs for data points in given benchmarks using LLMs (e.g., computational reasoning graphs for solving a math problem are shown in Figure 1). Next, we perform fine-grained graph perturbations based on various dimensions of the reasoning graph. As illustrated in the middle of Figure 1, we can dynamically increase the graph complexity by increasing its depth, width, and the numerical complexity of node values. Afterwards, we convert the reasoning graph back into the description that adapts the linguistic diversity as the original data. In order to ensure the correctness of the reasoning graph construction and graph-to-text generation, inspired by recent advances in tool-augmented LLMs [69, 28], we use tool-augmented LLMs to verify the quality of reasoning graphs and generated text to produce valid test examples. In this way, novel test cases can be generated with controllable complexity, adapted linguistic diversity, and validated labels.   Figure 1:",
            "references": [
                {
                    "title": "A Careful Examination of Large Language Model Performance on Grade School Arithmetic",
                    "abstract": "Large language models (LLMs) have achieved impressive success on many benchmarks for mathematical reasoning. However, there is growing concern that some of this performance actually reflects dataset contamination, where data closely resembling benchmark questions leaks into the training data, instead of true reasoning ability. To investigate this claim rigorously, we commission Grade School Math 1000 (GSM1k). GSM1k is designed to mirror the style and complexity of the established GSM8k benchmark, the gold standard for measuring elementary mathematical reasoning. We ensure that the two benchmarks are comparable across important metrics such as human solve rates, number of steps in solution, answer magnitude, and more. When evaluating leading open- and closed-source LLMs on GSM1k, we observe accuracy drops of up to 13%, with several families of models (e.g., Phi and Mistral) showing evidence of systematic overfitting across almost all model sizes. At the same time, many models, especially those on the frontier, (e.g., Gemini/GPT/Claude) show minimal signs of overfitting. Further analysis suggests a positive relationship (Spearman's r^2=0.32) between a model's probability of generating an example from GSM8k and its performance gap between GSM8k and GSM1k, suggesting that many models may have partially memorized GSM8k."
                },
                {
                    "title": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
                    "abstract": "We introduce phi-3-mini, a 3.8 billion parameter language model trained on 3.3 trillion tokens, whose overall performance, as measured by both academic benchmarks and internal testing, rivals that of models such as Mixtral 8x7B and GPT-3.5 (e.g., phi-3-mini achieves 69% on MMLU and 8.38 on MT-bench), despite being small enough to be deployed on a phone. Our training dataset is a scaled-up version of the one used for phi-2, composed of heavily filtered publicly available web data and synthetic data. The model is also further aligned for robustness, safety, and chat format. We also provide parameter-scaling results with a 7B, 14B models trained for 4.8T tokens, called phi-3-small, phi-3-medium, both significantly more capable than phi-3-mini (e.g., respectively 75%, 78% on MMLU, and 8.7, 8.9 on MT-bench). To enhance multilingual, multimodal, and long-context capabilities, we introduce three models in the phi-3.5 series: phi-3.5-mini, phi-3.5-MoE, and phi-3.5-Vision. The phi-3.5-MoE, a 16 x 3.8B MoE model with 6.6 billion active parameters, achieves superior performance in language reasoning, math, and code tasks compared to other open-source models of similar scale, such as Llama 3.1 and the Mixtral series, and on par with Gemini-1.5-Flash and GPT-4o-mini. Meanwhile, phi-3.5-Vision, a 4.2 billion parameter model derived from phi-3.5-mini, excels in reasoning tasks and is adept at handling both single-image and text prompts, as well as multi-image and text prompts."
                },
                {
                    "title": "Best Practices and Lessons Learned on Synthetic Data for Language Models",
                    "abstract": "The success of AI models relies on the availability of large, diverse, and high-quality datasets, which can be challenging to obtain due to data scarcity, privacy concerns, and high costs. Synthetic data has emerged as a promising solution by generating artificial data that mimics real-world patterns. This paper provides an overview of synthetic data research, discussing its applications, challenges, and future directions. We present empirical evidence from prior art to demonstrate its effectiveness and highlight the importance of ensuring its factuality, fidelity, and unbiasedness. We emphasize the need for responsible use of synthetic data to build more powerful, inclusive, and trustworthy language models."
                },
                {
                    "title": "Long-form factuality in large language models",
                    "abstract": "Large language models (LLMs) often generate content that contains factual errors when responding to fact-seeking prompts on open-ended topics. To benchmark a model's long-form factuality in open domains, we first use GPT-4 to generate LongFact, a prompt set comprising thousands of questions spanning 38 topics. We then propose that LLM agents can be used as automated evaluators for long-form factuality through a method which we call Search-Augmented Factuality Evaluator (SAFE). SAFE utilizes an LLM to break down a long-form response into a set of individual facts and to evaluate the accuracy of each fact using a multi-step reasoning process comprising sending search queries to Google Search and determining whether a fact is supported by the search results. Furthermore, we propose extending F1 score as an aggregated metric for long-form factuality. To do so, we balance the percentage of supported facts in a response (precision) with the percentage of provided facts relative to a hyperparameter representing a user's preferred response length (recall). Empirically, we demonstrate that LLM agents can outperform crowdsourced human annotators - on a set of ~16k individual facts, SAFE agrees with crowdsourced human annotators 72% of the time, and on a random subset of 100 disagreement cases, SAFE wins 76% of the time. At the same time, SAFE is more than 20 times cheaper than human annotators. We also benchmark thirteen language models on LongFact across four model families (Gemini, GPT, Claude, and PaLM-2), finding that larger language models generally achieve better long-form factuality. LongFact, SAFE, and all experimental code are available at https://github.com/google-deepmind/long-form-factuality."
                },
                {
                    "title": "RewardBench: Evaluating Reward Models for Language Modeling",
                    "abstract": "Reward models (RMs) are at the crux of successfully using RLHF to align pretrained models to human preferences, yet there has been relatively little study that focuses on evaluation of those models. Evaluating reward models presents an opportunity to understand the opaque technologies used for alignment of language models and which values are embedded in them. Resources for reward model training and understanding are sparse in the nascent open-source community around them. To enhance scientific understanding of reward models, we present RewardBench, a benchmark dataset and code-base for evaluation. The RewardBench dataset is a collection of prompt-chosen-rejected trios spanning chat, reasoning, and safety, to benchmark how reward models perform on challenging, structured and out-of-distribution queries. We create specific comparison datasets for RMs that have subtle, but verifiable reasons (e.g. bugs, incorrect facts) why one answer should be preferred to another. On the RewardBench leaderboard, we evaluate reward models trained with a variety of methods, such as the direct MLE training of classifiers and the implicit reward modeling of Direct Preference Optimization (DPO). We present many findings on propensity for refusals, reasoning limitations, and instruction following shortcomings of various reward models towards a better understanding of the RLHF process."
                },
                {
                    "title": "Whose Side Are You On? Investigating the Political Stance of Large Language Models",
                    "abstract": "Large Language Models (LLMs) have gained significant popularity for their application in various everyday tasks such as text generation, summarization, and information retrieval. As the widespread adoption of LLMs continues to surge, it becomes increasingly crucial to ensure that these models yield responses that are politically impartial, with the aim of preventing information bubbles, upholding fairness in representation, and mitigating confirmation bias. In this paper, we propose a quantitative framework and pipeline designed to systematically investigate the political orientation of LLMs. Our investigation delves into the political alignment of LLMs across a spectrum of eight polarizing topics, spanning from abortion to LGBTQ issues. Across topics, the results indicate that LLMs exhibit a tendency to provide responses that closely align with liberal or left-leaning perspectives rather than conservative or right-leaning ones when user queries include details pertaining to occupation, race, or political affiliation. The findings presented in this study not only reaffirm earlier observations regarding the left-leaning characteristics of LLMs but also surface particular attributes, such as occupation, that are particularly susceptible to such inclinations even when directly steered towards conservatism. As a recommendation to avoid these models providing politicised responses, users should be mindful when crafting queries, and exercise caution in selecting neutral prompt language."
                },
                {
                    "title": "Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context",
                    "abstract": "In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content."
                },
                {
                    "title": "Common 7B Language Models Already Possess Strong Math Capabilities",
                    "abstract": "Mathematical capabilities were previously believed to emerge in common language models only at a very large scale or require extensive math-related pre-training. This paper shows that the LLaMA-2 7B model with common pre-training already exhibits strong mathematical abilities, as evidenced by its impressive accuracy of 97.7% and 72.0% on the GSM8K and MATH benchmarks, respectively, when selecting the best response from 256 random generations. The primary issue with the current base model is the difficulty in consistently eliciting its inherent mathematical capabilities. Notably, the accuracy for the first answer drops to 49.5% and 7.9% on the GSM8K and MATH benchmarks, respectively. We find that simply scaling up the SFT data can significantly enhance the reliability of generating correct answers. However, the potential for extensive scaling is constrained by the scarcity of publicly available math questions. To overcome this limitation, we employ synthetic data, which proves to be nearly as effective as real data and shows no clear saturation when scaled up to approximately one million samples. This straightforward approach achieves an accuracy of 82.6% on GSM8K and 40.6% on MATH using LLaMA-2 7B models, surpassing previous models by 14.2% and 20.8%, respectively. We also provide insights into scaling behaviors across different reasoning complexities and error types."
                },
                {
                    "title": "Key-Point-Driven Data Synthesis with its Enhancement on Mathematical Reasoning",
                    "abstract": "Large language models (LLMs) have shown great potential in complex reasoning tasks, yet their performance is often hampered by the scarcity of high-quality and reasoning-focused training datasets. Addressing this challenge, we propose Key-Point-Driven Data Synthesis (KPDDS), a novel data synthesis framework that synthesizes question-answer pairs by leveraging key points and exemplar practices from authentic data sources. KPDDS ensures the generation of novel questions with rigorous quality control and substantial scalability. As a result, we present KPMath, an extensive synthetic dataset tailored for mathematical reasoning, comprising over 800K question-answer pairs. Utilizing KPMath and augmenting it with additional reasoning-intensive corpora, we create the comprehensive KPMath-Plus dataset. The Qwen1.5-72B model, fine-tuned on KPMath-Plus, achieves 87.0% PASS@1 accuracy on GSM8K and 58.3% on MATH, surpassing competitors in the 7B to 70B range and best commercial models like GPT-4 across multiple math reasoning datasets."
                },
                {
                    "title": "Generalization or Memorization: Data Contamination and Trustworthy Evaluation for Large Language Models",
                    "abstract": "Recent statements about the impressive capabilities of large language models (LLMs) are usually supported by evaluating on open-access benchmarks. Considering the vast size and wide-ranging sources of LLMs' training data, it could explicitly or implicitly include test data, leading to LLMs being more susceptible to data contamination. However, due to the opacity of training data, the black-box access of models, and the rapid growth of synthetic training data, detecting and mitigating data contamination for LLMs faces significant challenges. In this paper, we propose CDD, which stands for Contamination Detection via output Distribution for LLMs. CDD necessitates only the sampled texts to detect data contamination, by identifying the peakedness of LLM's output distribution. To mitigate the impact of data contamination in evaluation, we also present TED: Trustworthy Evaluation via output Distribution, based on the correction of LLM's output distribution. To facilitate this study, we introduce two benchmarks, i.e., DetCon and ComiEval, for data contamination detection and contamination mitigation evaluation tasks. Extensive experimental results show that CDD achieves the average relative improvements of 21.8\\%-30.2\\% over other contamination detection approaches in terms of Accuracy, F1 Score, and AUC metrics, and can effectively detect implicit contamination. TED substantially mitigates performance improvements up to 66.9\\% attributed to data contamination across various contamination setups. In real-world applications, we reveal that ChatGPT exhibits a high potential to suffer from data contamination on HumanEval benchmark."
                },
                {
                    "title": "KIEval: A Knowledge-grounded Interactive Evaluation Framework for Large Language Models",
                    "abstract": "Automatic evaluation methods for large language models (LLMs) are hindered by data contamination, leading to inflated assessments of their effectiveness. Existing strategies, which aim to detect contaminated texts, focus on quantifying contamination status instead of accurately gauging model performance. In this paper, we introduce KIEval, a Knowledge-grounded Interactive Evaluation framework, which incorporates an LLM-powered\"interactor\"role for the first time to accomplish a dynamic contamination-resilient evaluation. Starting with a question in a conventional LLM benchmark involving domain-specific knowledge, KIEval utilizes dynamically generated, multi-round, and knowledge-focused dialogues to determine whether a model's response is merely a recall of benchmark answers or demonstrates a deep comprehension to apply knowledge in more complex conversations. Extensive experiments on seven leading LLMs across five datasets validate KIEval's effectiveness and generalization. We also reveal that data contamination brings no contribution or even negative effect to models' real-world applicability and understanding, and existing contamination detection methods for LLMs can only identify contamination in pre-training but not during supervised fine-tuning."
                },
                {
                    "title": "DyVal 2: Dynamic Evaluation of Large Language Models by Meta Probing Agents",
                    "abstract": "Evaluation of large language models (LLMs) has raised great concerns in the community due to the issue of data contamination. Existing work designed evaluation protocols using well-defined algorithms for specific tasks, which cannot be easily extended to diverse scenarios. Moreover, current evaluation benchmarks can only provide the overall benchmark results and cannot support a fine-grained and multifaceted analysis of LLMs' abilities. In this paper, we propose meta probing agents (MPA), a general dynamic evaluation protocol inspired by psychometrics to evaluate LLMs. MPA is the key component of DyVal 2, which naturally extends the previous DyVal~\\citep{zhu2023dyval}. MPA designs the probing and judging agents to automatically transform an original evaluation problem into a new one following psychometric theory on three basic cognitive abilities: language understanding, problem solving, and domain knowledge. These basic abilities are also dynamically configurable, allowing multifaceted analysis. We conducted extensive evaluations using MPA and found that most LLMs achieve poorer performance, indicating room for improvement. Our multifaceted analysis demonstrated the strong correlation between the basic abilities and an implicit Matthew effect on model size, i.e., larger models possess stronger correlations of the abilities. MPA can also be used as a data augmentation approach to enhance LLMs. Code is available at: https://github.com/microsoft/promptbench."
                },
                {
                    "title": "TreeEval: Benchmark-Free Evaluation of Large Language Models through Tree Planning",
                    "abstract": "Recently, numerous new benchmarks have been established to evaluate the performance of large language models (LLMs) via either computing a holistic score or employing another LLM as a judge. However, these approaches suffer from data leakage due to the open access of the benchmark and inflexible evaluation process. To address this issue, we introduce $\\textbf{TreeEval}$, a benchmark-free evaluation method for LLMs that let a high-performance LLM host an irreproducible evaluation session and essentially avoids the data leakage. Moreover, this LLM performs as an examiner to raise up a series of questions under a topic with a tree planing strategy, which considers the current evaluation status to decide the next question generation and ensures the completeness and efficiency of the evaluation process. We evaluate $6$ models of different parameter sizes, including $7$B, $13$B, and $33$B, and ultimately achieved the highest correlation coefficient with AlpacaEval2.0 using only around $45$ questions. We also conduct more analysis to show the robustness and reliability of TreeEval. Our code can be accessed via the provided https://github.com/Ashura5/TreeEval."
                },
                {
                    "title": "Benchmark Self-Evolving: A Multi-Agent Framework for Dynamic LLM Evaluation",
                    "abstract": "This paper presents a benchmark self-evolving framework to dynamically evaluate rapidly advancing Large Language Models (LLMs), aiming for a more accurate assessment of their capabilities and limitations. We utilize a multi-agent system to manipulate the context or question of original instances, reframing new evolving instances with high confidence that dynamically extend existing benchmarks. Towards a more scalable, robust and fine-grained evaluation, we implement six reframing operations to construct evolving instances testing LLMs against diverse queries, data noise and probing their problem-solving sub-abilities. With this framework, we extend benchmark datasets of four tasks. Experimental results show a general performance decline in most LLMs against their original results. This decline under our scalable and robust evaluations, alongside our fine-grained evaluation, more accurately reflect models' capabilities. Besides, our framework widens performance discrepancies both between different models and within the same model across various tasks, facilitating more informed model selection for specific tasks (Code and data are available at https://github.com/NanshineLoong/Self-Evolving-Benchmark)."
                },
                {
                    "title": "Leak, Cheat, Repeat: Data Contamination and Evaluation Malpractices in Closed-Source LLMs",
                    "abstract": "Natural Language Processing (NLP) research is increasingly focusing on the use of Large Language Models (LLMs), with some of the most popular ones being either fully or partially closed-source. The lack of access to model details, especially regarding training data, has repeatedly raised concerns about data contamination among researchers. Several attempts have been made to address this issue, but they are limited to anecdotal evidence and trial and error. Additionally, they overlook the problem of indirect data leaking, where modelsare iteratively improved by using data coming from users. In this work, we conduct the first systematic analysis of work using OpenAI\u2019s GPT-3.5 and GPT-4, the most prominently used LLMs today, in the context of data contamination. By analysing 255 papers and considering OpenAI\u2019s data usage policy, we extensively document the amount of data leaked to these models during the first year after the model\u2019s release. We report that these models have been globally exposed to \u223c4.7M samples from 263 benchmarks. At the same time, we document a number of evaluation malpractices emerging in the reviewed papers, such as unfair or missing baseline comparisons and reproducibility issues. We release our results as a collaborative project on https://leak-llm.github.io/, where other researchers can contribute to our efforts."
                },
                {
                    "title": "DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models",
                    "abstract": "Mathematical reasoning poses a significant challenge for language models due to its complex and structured nature. In this paper, we introduce DeepSeekMath 7B, which continues pre-training DeepSeek-Coder-Base-v1.5 7B with 120B math-related tokens sourced from Common Crawl, together with natural language and code data. DeepSeekMath 7B has achieved an impressive score of 51.7% on the competition-level MATH benchmark without relying on external toolkits and voting techniques, approaching the performance level of Gemini-Ultra and GPT-4. Self-consistency over 64 samples from DeepSeekMath 7B achieves 60.9% on MATH. The mathematical reasoning capability of DeepSeekMath is attributed to two key factors: First, we harness the significant potential of publicly available web data through a meticulously engineered data selection pipeline. Second, we introduce Group Relative Policy Optimization (GRPO), a variant of Proximal Policy Optimization (PPO), that enhances mathematical reasoning abilities while concurrently optimizing the memory usage of PPO."
                },
                {
                    "title": "Self-Rewarding Language Models",
                    "abstract": "We posit that to achieve superhuman agents, future models require superhuman feedback in order to provide an adequate training signal. Current approaches commonly train reward models from human preferences, which may then be bottlenecked by human performance level, and secondly these separate frozen reward models cannot then learn to improve during LLM training. In this work, we study Self-Rewarding Language Models, where the language model itself is used via LLM-as-a-Judge prompting to provide its own rewards during training. We show that during Iterative DPO training that not only does instruction following ability improve, but also the ability to provide high-quality rewards to itself. Fine-tuning Llama 2 70B on three iterations of our approach yields a model that outperforms many existing systems on the AlpacaEval 2.0 leaderboard, including Claude 2, Gemini Pro, and GPT-4 0613. While there is much left still to explore, this work opens the door to the possibility of models that can continually improve in both axes."
                },
                {
                    "title": "Investigating Data Contamination for Pre-training Language Models",
                    "abstract": "Language models pre-trained on web-scale corpora demonstrate impressive capabilities on diverse downstream tasks. However, there is increasing concern whether such capabilities might arise from evaluation datasets being included in the pre-training corpus -- a phenomenon known as \\textit{data contamination} -- in a manner that artificially increases performance. There has been little understanding of how this potential contamination might influence LMs' performance on downstream tasks. In this paper, we explore the impact of data contamination at the pre-training stage by pre-training a series of GPT-2 models \\textit{from scratch}. We highlight the effect of both text contamination (\\textit{i.e.}\\ input text of the evaluation samples) and ground-truth contamination (\\textit{i.e.}\\ the prompts asked on the input and the desired outputs) from evaluation data. We also investigate the effects of repeating contamination for various downstream tasks. Additionally, we examine the prevailing n-gram-based definitions of contamination within current LLM reports, pinpointing their limitations and inadequacy. Our findings offer new insights into data contamination's effects on language model capabilities and underscore the need for independent, comprehensive contamination assessments in LLM studies."
                },
                {
                    "title": "Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training",
                    "abstract": "Humans are capable of strategically deceptive behavior: behaving helpfully in most situations, but then behaving very differently in order to pursue alternative objectives when given the opportunity. If an AI system learned such a deceptive strategy, could we detect it and remove it using current state-of-the-art safety training techniques? To study this question, we construct proof-of-concept examples of deceptive behavior in large language models (LLMs). For example, we train models that write secure code when the prompt states that the year is 2023, but insert exploitable code when the stated year is 2024. We find that such backdoor behavior can be made persistent, so that it is not removed by standard safety training techniques, including supervised fine-tuning, reinforcement learning, and adversarial training (eliciting unsafe behavior and then training to remove it). The backdoor behavior is most persistent in the largest models and in models trained to produce chain-of-thought reasoning about deceiving the training process, with the persistence remaining even when the chain-of-thought is distilled away. Furthermore, rather than removing backdoors, we find that adversarial training can teach models to better recognize their backdoor triggers, effectively hiding the unsafe behavior. Our results suggest that, once a model exhibits deceptive behavior, standard techniques could fail to remove such deception and create a false impression of safety."
                },
                {
                    "title": "Mixtral of Experts",
                    "abstract": "We introduce Mixtral 8x7B, a Sparse Mixture of Experts (SMoE) language model. Mixtral has the same architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks (i.e. experts). For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, each token has access to 47B parameters, but only uses 13B active parameters during inference. Mixtral was trained with a context size of 32k tokens and it outperforms or matches Llama 2 70B and GPT-3.5 across all evaluated benchmarks. In particular, Mixtral vastly outperforms Llama 2 70B on mathematics, code generation, and multilingual benchmarks. We also provide a model fine-tuned to follow instructions, Mixtral 8x7B - Instruct, that surpasses GPT-3.5 Turbo, Claude-2.1, Gemini Pro, and Llama 2 70B - chat model on human benchmarks. Both the base and instruct models are released under the Apache 2.0 license."
                },
                {
                    "title": "Task Contamination: Language Models May Not Be Few-Shot Anymore",
                    "abstract": "Large language models (LLMs) offer impressive performance in various zero-shot and few-shot tasks. However, their success in zero-shot or few-shot settings may be affected by task contamination, a potential limitation that has not been thoroughly examined. This paper investigates how zero-shot and few-shot performance of LLMs has changed chronologically over datasets released over time, and over LLMs released over time. Utilizing GPT-3 series models and several other recent open-sourced LLMs, and controlling for dataset difficulty, we find that datasets released prior to the LLM training data creation date perform surprisingly better than datasets released post the LLM training data creation date. This strongly indicates that, for many LLMs, there exists task contamination on zero-shot and few-shot evaluation for datasets prior to the LLMs' training data creation date. Additionally, we utilize training data inspection, training data extraction, and a membership inference attack, which reveal further evidence of task contamination. Importantly, we find that for tasks with no possibility of task contamination, LLMs rarely demonstrate statistically significant improvements over simple majority baselines, in both zero and few-shot settings."
                },
                {
                    "title": "What Makes Good Data for Alignment? A Comprehensive Study of Automatic Data Selection in Instruction Tuning",
                    "abstract": "Instruction tuning is a standard technique employed to align large language models to end tasks and user preferences after the initial pretraining phase. Recent research indicates the critical role of data engineering in instruction tuning -- when appropriately selected, only limited data is necessary to achieve superior performance. However, we still lack a principled understanding of what makes good instruction tuning data for alignment, and how we should select data automatically and effectively. In this work, we delve deeply into automatic data selection strategies for alignment. We start with controlled studies to measure data across three dimensions: complexity, quality, and diversity, along which we examine existing methods and introduce novel techniques for enhanced data measurement. Subsequently, we propose a simple strategy to select data samples based on the measurement. We present deita (short for Data-Efficient Instruction Tuning for Alignment), a series of models fine-tuned from LLaMA and Mistral models using data samples automatically selected with our proposed approach. Empirically, deita performs better or on par with the state-of-the-art open-source alignment models with only 6K SFT training data samples -- over 10x less than the data used in the baselines. When further trained with direct preference optimization (DPO), deita-Mistral-7B + DPO trained with 6K SFT and 10K DPO samples achieve 7.55 MT-Bench and 90.06% AlpacaEval scores. We anticipate this work to provide tools on automatic data selection, facilitating data-efficient alignment. We release our models as well as the selected datasets for future researches to effectively align models more efficiently."
                },
                {
                    "title": "Large Language Models are Not Stable Recommender Systems",
                    "abstract": "With the significant successes of large language models (LLMs) in many natural language processing tasks, there is growing interest among researchers in exploring LLMs for novel recommender systems. However, we have observed that directly using LLMs as a recommender system is usually unstable due to its inherent position bias. To this end, we introduce exploratory research and find consistent patterns of positional bias in LLMs that influence the performance of recommendation across a range of scenarios. Then, we propose a Bayesian probabilistic framework, STELLA (Stable LLM for Recommendation), which involves a two-stage pipeline. During the first probing stage, we identify patterns in a transition matrix using a probing detection dataset. And in the second recommendation stage, a Bayesian strategy is employed to adjust the biased output of LLMs with an entropy indicator. Therefore, our framework can capitalize on existing pattern information to calibrate instability of LLMs, and enhance recommendation performance. Finally, extensive experiments clearly validate the effectiveness of our framework."
                },
                {
                    "title": "LatestEval: Addressing Data Contamination in Language Model Evaluation through Dynamic and Time-Sensitive Test Construction",
                    "abstract": "Data contamination in evaluation is getting increasingly prevalent with the emergence of language models pre-trained on super large, automatically crawled corpora. This problem leads to significant challenges in the accurate assessment of model capabilities and generalisations. In this paper, we propose LatestEval, an automatic method that leverages the most recent texts to create uncontaminated reading comprehension evaluations. LatestEval avoids data contamination by only using texts published within a recent time window, ensuring no overlap with the training corpora of pre-trained language models. We develop the LatestEval automated pipeline to 1) gather the latest texts; 2) identify key information, and 3) construct questions targeting the information while removing the existing answers from the context. This encourages models to infer the answers themselves based on the remaining context, rather than just copy-paste. Our experiments demonstrate that language models exhibit negligible memorisation behaviours on LatestEval as opposed to previous benchmarks, suggesting a significantly reduced risk of data contamination and leading to a more robust evaluation. Data and code are publicly available at: https://github.com/liyucheng09/LatestEval."
                },
                {
                    "title": "Simulating Opinion Dynamics with Networks of LLM-based Agents",
                    "abstract": "Accurately simulating human opinion dynamics is crucial for understanding a variety of societal phenomena, including polarization and the spread of misinformation. However, the agent-based models (ABMs) commonly used for such simulations often over-simplify human behavior. We propose a new approach to simulating opinion dynamics based on populations of Large Language Models (LLMs). Our findings reveal a strong inherent bias in LLM agents towards producing accurate information, leading simulated agents to consensus in line with scientific reality. This bias limits their utility for understanding resistance to consensus views on issues like climate change. After inducing confirmation bias through prompt engineering, however, we observed opinion fragmentation in line with existing agent-based modeling and opinion dynamics research. These insights highlight the promise and limitations of LLM agents in this domain and suggest a path forward: refining LLMs with real-world discourse to better simulate the evolution of human beliefs."
                },
                {
                    "title": "Investigating Data Contamination in Modern Benchmarks for Large Language Models",
                    "abstract": "Recent observations have underscored a disparity between the inflated benchmark scores and the actual performance of LLMs, raising concerns about potential contamination of evaluation benchmarks. This issue is especially critical for closed-source models and certain open-source models where training data transparency is lacking. In this paper we study data contamination by proposing two methods tailored for both open-source and proprietary LLMs. We first introduce a retrieval-based system to explore potential overlaps between evaluation benchmarks and pretraining corpora. We further present a novel investigation protocol named Testset Slot Guessing (TS-Guessing), applicable to both open and proprietary models. This approach entails masking a wrong answer in a multiple-choice question and prompting the model to fill in the gap. Additionally, it involves obscuring an unlikely word in an evaluation example and asking the model to produce it. We find that certain commercial LLMs could surprisingly guess the missing option in various test sets. Specifically, in the MMLU benchmark, ChatGPT and GPT-4 demonstrated an exact match rate of 52% and 57%, respectively, in guessing the missing options in benchmark test data. We hope these results underscore the need for more robust evaluation methodologies and benchmarks in the field."
                },
                {
                    "title": "Data Contamination Quiz: A Tool to Detect and Estimate Contamination in Large Language Models",
                    "abstract": "We propose the Data Contamination Quiz (DCQ), a simple and effective approach to detect data contamination in large language models (LLMs) and estimate the amount of it. Specifically, we frame data contamination detection as a series of multiple-choice questions and devise a quiz format wherein three perturbed versions of each subsampled instance from a specific dataset partition (e.g., GSM8k test set) are created. These changes only include word-level perturbations. The generated perturbations, along with the original dataset instance, form the options in the DCQ, with an extra option accommodating the possibility of selecting none of the provided options. Given that the only distinguishing signal among the options is the exact wording with respect to the original dataset instance, an LLM, when tasked with identifying the original dataset instance, gravitates towards selecting the original one if it has been exposed to it in its pre-training phase -- a trait intrinsic to LLMs. While accounting for positional biases in LLMs, the quiz performance reveals the contamination level for the model being examined with the dataset partition to which the quiz pertains. Applied to various datasets with GPT-4 and GPT-3.5, our findings -- while fully lacking access to pre-training data and model parameters -- suggest that DCQ achieves state-of-the-art results and uncovers greater contamination/memorization levels compared to existing methods and proficiently bypasses more safety filters, especially those set to avoid generating copyrighted contents."
                },
                {
                    "title": "A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions",
                    "abstract": "The emergence of large language models (LLMs) has marked a significant breakthrough in natural language processing (NLP), leading to remarkable advancements in text understanding and generation. Nevertheless, alongside these strides, LLMs exhibit a critical tendency to produce hallucinations, resulting in content that is inconsistent with real-world facts or user inputs. This phenomenon poses substantial challenges to their practical deployment and raises concerns over the reliability of LLMs in real-world scenarios, which attracts increasing attention to detect and mitigate these hallucinations. In this survey, we aim to provide a thorough and in-depth overview of recent advances in the field of LLM hallucinations. We begin with an innovative taxonomy of LLM hallucinations, then delve into the factors contributing to hallucinations. Subsequently, we present a comprehensive overview of hallucination detection methods and benchmarks. Additionally, representative approaches designed to mitigate hallucinations are introduced accordingly. Finally, we analyze the challenges that highlight the current limitations and formulate open questions, aiming to delineate pathways for future research on hallucinations in LLMs."
                },
                {
                    "title": "Rethinking Benchmark and Contamination for Language Models with Rephrased Samples",
                    "abstract": "Large language models are increasingly trained on all the data ever produced by humans. Many have raised concerns about the trustworthiness of public benchmarks due to potential contamination in pre-training or fine-tuning datasets. While most data decontamination efforts apply string matching (e.g., n-gram overlap) to remove benchmark data, we show that these methods are insufficient, and simple variations of test data (e.g., paraphrasing, translation) can easily bypass these decontamination measures. Furthermore, we demonstrate that if such variation of test data is not eliminated, a 13B model can easily overfit a test benchmark and achieve drastically high performance, on par with GPT-4. We validate such observations in widely used benchmarks such as MMLU, GSK8k, and HumanEval. To address this growing risk, we propose a stronger LLM-based decontamination method and apply it to widely used pre-training and fine-tuning datasets, revealing significant previously unknown test overlap. For example, in pre-training sets such as RedPajama-Data-1T and StarCoder-Data, we identified that 8-18\\% of the HumanEval benchmark overlaps. Interestingly, we also find such contamination in synthetic dataset generated by GPT-3.5/4, suggesting a potential risk of unintentional contamination. We urge the community to adopt stronger decontamination approaches when using public benchmarks. Moreover, we call for the community to actively develop fresh one-time exams to evaluate models accurately. Our decontamination tool is publicly available at https://github.com/lm-sys/llm-decontaminator."
                },
                {
                    "title": "Don't Make Your LLM an Evaluation Benchmark Cheater",
                    "abstract": "Large language models~(LLMs) have greatly advanced the frontiers of artificial intelligence, attaining remarkable improvement in model capacity. To assess the model performance, a typical approach is to construct evaluation benchmarks for measuring the ability level of LLMs in different aspects. Despite that a number of high-quality benchmarks have been released, the concerns about the appropriate use of these benchmarks and the fair comparison of different models are increasingly growing. Considering these concerns, in this paper, we discuss the potential risk and impact of inappropriately using evaluation benchmarks and misleadingly interpreting the evaluation results. Specially, we focus on a special issue that would lead to inappropriate evaluation, \\ie \\emph{benchmark leakage}, referring that the data related to evaluation sets is occasionally used for model training. This phenomenon now becomes more common since pre-training data is often prepared ahead of model test. We conduct extensive experiments to study the effect of benchmark leverage, and find that it can dramatically boost the evaluation results, which would finally lead to an unreliable assessment of model performance. To improve the use of existing evaluation benchmarks, we finally present several guidelines for both LLM developers and benchmark maintainers. We hope this work can draw attention to appropriate training and evaluation of LLMs."
                },
                {
                    "title": "NLP Evaluation in trouble: On the Need to Measure LLM Data Contamination for each Benchmark",
                    "abstract": "In this position paper, we argue that the classical evaluation on Natural Language Processing (NLP) tasks using annotated benchmarks is in trouble. The worst kind of data contamination happens when a Large Language Model (LLM) is trained on the test split of a benchmark, and then evaluated in the same benchmark. The extent of the problem is unknown, as it is not straightforward to measure. Contamination causes an overestimation of the performance of a contaminated model in a target benchmark and associated task with respect to their non-contaminated counterparts. The consequences can be very harmful, with wrong scientific conclusions being published while other correct ones are discarded. This position paper defines different levels of data contamination and argues for a community effort, including the development of automatic and semi-automatic measures to detect when data from a benchmark was exposed to a model, and suggestions for flagging papers with conclusions that are compromised by data contamination."
                },
                {
                    "title": "Skill-Mix: a Flexible and Expandable Family of Evaluations for AI models",
                    "abstract": "With LLMs shifting their role from statistical modeling of language to serving as general-purpose AI agents, how should LLM evaluations change? Arguably, a key ability of an AI agent is to flexibly combine, as needed, the basic skills it has learned. The capability to combine skills plays an important role in (human) pedagogy and also in a paper on emergence phenomena (Arora&Goyal, 2023). This work introduces Skill-Mix, a new evaluation to measure ability to combine skills. Using a list of $N$ skills the evaluator repeatedly picks random subsets of $k$ skills and asks the LLM to produce text combining that subset of skills. Since the number of subsets grows like $N^k$, for even modest $k$ this evaluation will, with high probability, require the LLM to produce text significantly different from any text in the training set. The paper develops a methodology for (a) designing and administering such an evaluation, and (b) automatic grading (plus spot-checking by humans) of the results using GPT-4 as well as the open LLaMA-2 70B model. Administering a version of to popular chatbots gave results that, while generally in line with prior expectations, contained surprises. Sizeable differences exist among model capabilities that are not captured by their ranking on popular LLM leaderboards (\"cramming for the leaderboard\"). Furthermore, simple probability calculations indicate that GPT-4's reasonable performance on $k=5$ is suggestive of going beyond\"stochastic parrot\"behavior (Bender et al., 2021), i.e., it combines skills in ways that it had not seen during training. We sketch how the methodology can lead to a Skill-Mix based eco-system of open evaluations for AI capabilities of future models."
                },
                {
                    "title": "Proving Test Set Contamination in Black Box Language Models",
                    "abstract": "Large language models are trained on vast amounts of internet data, prompting concerns and speculation that they have memorized public benchmarks. Going from speculation to proof of contamination is challenging, as the pretraining data used by proprietary models are often not publicly accessible. We show that it is possible to provide provable guarantees of test set contamination in language models without access to pretraining data or model weights. Our approach leverages the fact that when there is no data contamination, all orderings of an exchangeable benchmark should be equally likely. In contrast, the tendency for language models to memorize example order means that a contaminated language model will find certain canonical orderings to be much more likely than others. Our test flags potential contamination whenever the likelihood of a canonically ordered benchmark dataset is significantly higher than the likelihood after shuffling the examples. We demonstrate that our procedure is sensitive enough to reliably prove test set contamination in challenging situations, including models as small as 1.4 billion parameters, on small test sets of only 1000 examples, and datasets that appear only a few times in the pretraining corpus. Using our test, we audit five popular publicly accessible language models for test set contamination and find little evidence for pervasive contamination."
                },
                {
                    "title": "Detecting Pretraining Data from Large Language Models",
                    "abstract": "Although large language models (LLMs) are widely deployed, the data used to train them is rarely disclosed. Given the incredible scale of this data, up to trillions of tokens, it is all but certain that it includes potentially problematic text such as copyrighted materials, personally identifiable information, and test data for widely reported reference benchmarks. However, we currently have no way to know which data of these types is included or in what proportions. In this paper, we study the pretraining data detection problem: given a piece of text and black-box access to an LLM without knowing the pretraining data, can we determine if the model was trained on the provided text? To facilitate this study, we introduce a dynamic benchmark WIKIMIA that uses data created before and after model training to support gold truth detection. We also introduce a new detection method Min-K% Prob based on a simple hypothesis: an unseen example is likely to contain a few outlier words with low probabilities under the LLM, while a seen example is less likely to have words with such low probabilities. Min-K% Prob can be applied without any knowledge about the pretraining corpus or any additional training, departing from previous detection methods that require training a reference model on data that is similar to the pretraining data. Moreover, our experiments demonstrate that Min-K% Prob achieves a 7.4% improvement on WIKIMIA over these previous methods. We apply Min-K% Prob to two real-world scenarios, copyrighted book detection, and contaminated downstream example detection, and find it a consistently effective solution."
                },
                {
                    "title": "S3Eval: A Synthetic, Scalable, Systematic Evaluation Suite for Large Language Model",
                    "abstract": "The rapid development of Large Language Models (LLMs) has led to great strides in model capabilities like long-context understanding and reasoning.However, as LLMs are able to process longer contexts, it becomes more challenging to evaluate whether they have acquired certain capabilities, since the length of text (e.g., 200K tokens) they can process far exceeds what humans can reliably assess in a reasonable duration.In this paper, we propose using complex synthetic tasks as a proxy evaluation method, and present S3Eval, a Synthetic, Scalable, Systematic evaluation suite for LLMs evaluation.The synthetic nature of S3Eval provides users full control over the dataset, allowing them to systematically probe LLM capabilities by scaling text length and varying task difficulty across diverse scenarios.The strong correlation between S3Eval and real-world benchmarks demonstrates the soundness of using S3Eval for evaluation of LLMs.S3Eval provides a flexible and infinite long-context data generation method. We have generated a comprehensive dataset called S3Eval-Standard, and experimental results have shown that it poses significant challenges for all existing LLMs."
                },
                {
                    "title": "Llemma: An Open Language Model For Mathematics",
                    "abstract": "We present Llemma, a large language model for mathematics. We continue pretraining Code Llama on the Proof-Pile-2, a mixture of scientific papers, web data containing mathematics, and mathematical code, yielding Llemma. On the MATH benchmark Llemma outperforms all known open base models, as well as the unreleased Minerva model suite on an equi-parameter basis. Moreover, Llemma is capable of tool use and formal theorem proving without any further finetuning. We openly release all artifacts, including 7 billion and 34 billion parameter models, the Proof-Pile-2, and code to replicate our experiments."
                },
                {
                    "title": "Mistral 7B",
                    "abstract": "We introduce Mistral 7B v0.1, a 7-billion-parameter language model engineered for superior performance and efficiency. Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation. Our model leverages grouped-query attention (GQA) for faster inference, coupled with sliding window attention (SWA) to effectively handle sequences of arbitrary length with a reduced inference cost. We also provide a model fine-tuned to follow instructions, Mistral 7B -- Instruct, that surpasses the Llama 2 13B -- Chat model both on human and automated benchmarks. Our models are released under the Apache 2.0 license."
                },
                {
                    "title": "Estimating Contamination via Perplexity: Quantifying Memorisation in Language Model Evaluation",
                    "abstract": "Data contamination in model evaluation is getting increasingly prevalent as the massive training corpora of large language models often unintentionally include benchmark samples. Therefore, contamination analysis has became an inevitable part of reliable model evaluation. However, existing method of contamination analysis requires the access of the entire training data which is often confidential for recent models. This prevent the community to rigorously audit these models and conduct accurate assessment of their capability. In this paper, we propose a novel method to quantify contamination without the access of the full training set, that measure the extent of contamination with perplexity. Our analysis provides evidence of significant memorisation of recent foundation models in popular reading comprehension, summarisation benchmarks, while multiple choice appears less contaminated."
                },
                {
                    "title": "WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct",
                    "abstract": "Large language models (LLMs), such as GPT-4, have shown remarkable performance in natural language processing (NLP) tasks, including challenging mathematical reasoning. However, most existing open-source models are only pre-trained on large-scale internet data and without math-related optimization. In this paper, we present WizardMath, which enhances the mathematical reasoning abilities of Llama-2, by applying our proposed Reinforcement Learning from Evol-Instruct Feedback (RLEIF) method to the domain of math. Through extensive experiments on two mathematical reasoning benchmarks, namely GSM8k and MATH, we reveal the extraordinary capabilities of our model. WizardMath surpasses all other open-source LLMs by a substantial margin. Furthermore, our model even outperforms ChatGPT-3.5, Claude Instant-1, PaLM-2 and Minerva on GSM8k, simultaneously surpasses Text-davinci-002, PaLM-1 and GPT-3 on MATH. More details and model weights are public at https://github.com/nlpxucan/WizardLM and https://huggingface.co/WizardLM."
                },
                {
                    "title": "Time Travel in LLMs: Tracing Data Contamination in Large Language Models",
                    "abstract": "Data contamination, i.e., the presence of test data from downstream tasks in the training data of large language models (LLMs), is a potential major issue in measuring LLMs' real effectiveness on other tasks. We propose a straightforward yet effective method for identifying data contamination within LLMs. At its core, our approach starts by identifying potential contamination at the instance level; using this information, our approach then assesses wider contamination at the partition level. To estimate contamination of individual instances, we employ\"guided instruction:\"a prompt consisting of the dataset name, partition type, and the random-length initial segment of a reference instance, asking the LLM to complete it. An instance is flagged as contaminated if the LLM's output either exactly or nearly matches the latter segment of the reference. To understand if an entire partition is contaminated, we propose two ideas. The first idea marks a dataset partition as contaminated if the average overlap score with the reference instances (as measured by ROUGE-L or BLEURT) is statistically significantly better with the completions from guided instruction compared to a\"general instruction\"that does not include the dataset and partition name. The second idea marks a dataset partition as contaminated if a classifier based on GPT-4 with few-shot in-context learning prompt marks multiple generated completions as exact/near-exact matches of the corresponding reference instances. Our best method achieves an accuracy between 92% and 100% in detecting if an LLM is contaminated with seven datasets, containing train and test/validation partitions, when contrasted with manual evaluation by human experts. Further, our findings indicate that GPT-4 is contaminated with AG News, WNLI, and XSum datasets."
                },
                {
                    "title": "Solving Challenging Math Word Problems Using GPT-4 Code Interpreter with Code-based Self-Verification",
                    "abstract": "Recent progress in large language models (LLMs) like GPT-4 and PaLM-2 has brought significant advancements in addressing math reasoning problems. In particular, OpenAI's latest version of GPT-4, known as GPT-4 Code Interpreter, shows remarkable performance on challenging math datasets. In this paper, we explore the effect of code on enhancing LLMs' reasoning capability by introducing different constraints on the \\textit{Code Usage Frequency} of GPT-4 Code Interpreter. We found that its success can be largely attributed to its powerful skills in generating and executing code, evaluating the output of code execution, and rectifying its solution when receiving unreasonable outputs. Based on this insight, we propose a novel and effective prompting method, explicit \\uline{c}ode-based \\uline{s}elf-\\uline{v}erification~(CSV), to further boost the mathematical reasoning potential of GPT-4 Code Interpreter. This method employs a zero-shot prompt on GPT-4 Code Interpreter to encourage it to use code to self-verify its answers. In instances where the verification state registers as ``False'', the model shall automatically amend its solution, analogous to our approach of rectifying errors during a mathematics examination. Furthermore, we recognize that the states of the verification result indicate the confidence of a solution, which can improve the effectiveness of majority voting. With GPT-4 Code Interpreter and CSV, we achieve an impressive zero-shot accuracy on MATH dataset \\textbf{(53.9\\% $\\to$ 84.3\\%)}."
                },
                {
                    "title": "Simple synthetic data reduces sycophancy in large language models",
                    "abstract": "Sycophancy is an undesirable behavior where models tailor their responses to follow a human user's view even when that view is not objectively correct (e.g., adapting liberal views once a user reveals that they are liberal). In this paper, we study the prevalence of sycophancy in language models and propose a simple synthetic-data intervention to reduce this behavior. First, on a set of three sycophancy tasks (Perez et al., 2022) where models are asked for an opinion on statements with no correct answers (e.g., politics), we observe that both model scaling and instruction tuning significantly increase sycophancy for PaLM models up to 540B parameters. Second, we extend sycophancy evaluations to simple addition statements that are objectively incorrect, finding that despite knowing that these statements are wrong, language models will still agree with them if the user does as well. To reduce sycophancy, we present a straightforward synthetic-data intervention that takes public NLP tasks and encourages models to be robust to user opinions on these tasks. Adding these data in a lightweight finetuning step can significantly reduce sycophantic behavior on held-out prompts. Code for generating synthetic data for intervention can be found at https://github.com/google/sycophancy-intervention."
                },
                {
                    "title": "Skills-in-Context Prompting: Unlocking Compositionality in Large Language Models",
                    "abstract": "We investigate how to elicit compositional generalization capabilities in large language models (LLMs). Compositional generalization empowers LLMs to solve complex problems by combining foundational skills, a critical reasoning ability akin to human intelligence. However, even the most advanced LLMs currently struggle with this form of reasoning. We examine this problem within the framework of in-context learning and find that demonstrating both foundational skills and compositional examples grounded in these skills within the same prompt context is crucial. We refer to this prompt structure as skills-in-context (SKiC). With as few as two exemplars, this in-context learning structure enables LLMs to tackle more challenging problems requiring innovative skill combinations, achieving near-perfect systematic generalization across a broad range of tasks. Intriguingly, SKiC also unlocks the latent potential of LLMs, allowing them to more actively utilize pre-existing internal skills acquired during earlier pretraining stages to solve complex reasoning problems. The SKiC structure is robust across different skill constructions and exemplar choices and demonstrates strong transferability to new tasks. Finally, inspired by our in-context learning study, we show that fine-tuning LLMs with SKiC-style data can elicit zero-shot weak-to-strong generalization, enabling the models to solve much harder problems directly with standard prompting."
                },
                {
                    "title": "A Survey on Evaluation of Large Language Models",
                    "abstract": "Large language models (LLMs) are gaining increasing popularity in both academia and industry, owing to their unprecedented performance in various applications. As LLMs continue to play a vital role in both research and daily use, their evaluation becomes increasingly critical, not only at the task level, but also at the society level for better understanding of their potential risks. Over the past years, significant efforts have been made to examine LLMs from various perspectives. This paper presents a comprehensive review of these evaluation methods for LLMs, focusing on three key dimensions: what to evaluate, where to evaluate, and how to evaluate. Firstly, we provide an overview from the perspective of evaluation tasks, encompassing general natural language processing tasks, reasoning, medical usage, ethics, education, natural and social sciences, agent applications, and other areas. Secondly, we answer the \u2018where\u2019 and \u2018how\u2019 questions by diving into the evaluation methods and benchmarks, which serve as crucial components in assessing the performance of LLMs. Then, we summarize the success and failure cases of LLMs in different tasks. Finally, we shed light on several future challenges that lie ahead in LLMs evaluation. Our aim is to offer invaluable insights to researchers in the realm of LLMs evaluation, thereby aiding the development of more proficient LLMs. Our key point is that evaluation should be treated as an essential discipline to better assist the development of LLMs. We consistently maintain the related open-source materials at: https://github.com/MLGroupJLU/LLM-eval-survey"
                },
                {
                    "title": "ToolAlpaca: Generalized Tool Learning for Language Models with 3000 Simulated Cases",
                    "abstract": "Enabling large language models to utilize real-world tools effectively is crucial for achieving embodied intelligence. Existing approaches to tool learning have either primarily relied on extremely large language models, such as GPT-4, to attain generalized tool-use abilities in a zero-shot manner, or utilized supervised learning to train limited scopes of tools on compact models. However, it remains uncertain whether smaller language models can achieve generalized tool-use abilities without tool-specific training. To address this question, this paper introduces ToolAlpaca, a novel framework designed to automatically generate a diverse tool-use corpus and learn generalized tool-use abilities on compact language models with minimal human intervention. Specifically, ToolAlpaca first automatically creates a highly diversified tool-use corpus by building a multi-agent simulation environment. The corpus contains 3938 tool-use instances from more than 400 real-world tool APIs spanning 50 distinct categories. Subsequently, the constructed corpus is employed to fine-tune compact language models, resulting in two models, namely ToolAlpaca-7B and ToolAlpaca-13B, respectively. Finally, we evaluate the ability of these models to utilize previously unseen tools without specific training. Experimental results demonstrate that ToolAlpaca achieves effective generalized tool-use capabilities comparable to those of extremely large language models like GPT-3.5, demonstrating that learning generalized tool-use ability is feasible for compact language models."
                },
                {
                    "title": "Faith and Fate: Limits of Transformers on Compositionality",
                    "abstract": "Transformer large language models (LLMs) have sparked admiration for their exceptional performance on tasks that demand intricate multi-step reasoning. Yet, these models simultaneously show failures on surprisingly trivial problems. This begs the question: Are these errors incidental, or do they signal more substantial limitations? In an attempt to demystify transformer LLMs, we investigate the limits of these models across three representative compositional tasks -- multi-digit multiplication, logic grid puzzles, and a classic dynamic programming problem. These tasks require breaking problems down into sub-steps and synthesizing these steps into a precise answer. We formulate compositional tasks as computation graphs to systematically quantify the level of complexity, and break down reasoning steps into intermediate sub-procedures. Our empirical findings suggest that transformer LLMs solve compositional tasks by reducing multi-step compositional reasoning into linearized subgraph matching, without necessarily developing systematic problem-solving skills. To round off our empirical study, we provide theoretical arguments on abstract multi-step reasoning problems that highlight how autoregressive generations' performance can rapidly decay with\\,increased\\,task\\,complexity."
                },
                {
                    "title": "Voyager: An Open-Ended Embodied Agent with Large Language Models",
                    "abstract": "We introduce Voyager, the first LLM-powered embodied lifelong learning agent in Minecraft that continuously explores the world, acquires diverse skills, and makes novel discoveries without human intervention. Voyager consists of three key components: 1) an automatic curriculum that maximizes exploration, 2) an ever-growing skill library of executable code for storing and retrieving complex behaviors, and 3) a new iterative prompting mechanism that incorporates environment feedback, execution errors, and self-verification for program improvement. Voyager interacts with GPT-4 via blackbox queries, which bypasses the need for model parameter fine-tuning. The skills developed by Voyager are temporally extended, interpretable, and compositional, which compounds the agent's abilities rapidly and alleviates catastrophic forgetting. Empirically, Voyager shows strong in-context lifelong learning capability and exhibits exceptional proficiency in playing Minecraft. It obtains 3.3x more unique items, travels 2.3x longer distances, and unlocks key tech tree milestones up to 15.3x faster than prior SOTA. Voyager is able to utilize the learned skill library in a new Minecraft world to solve novel tasks from scratch, while other techniques struggle to generalize. We open-source our full codebase and prompts at https://voyager.minedojo.org/."
                },
                {
                    "title": "Holistic Evaluation of Language Models",
                    "abstract": "Language models (LMs) like GPT\u20103, PaLM, and ChatGPT are the foundation for almost all major language technologies, but their capabilities, limitations, and risks are not well understood. We present Holistic Evaluation of Language Models (HELM) to improve the transparency of LMs. LMs can serve many purposes and their behavior should satisfy many desiderata. To navigate the vast space of potential scenarios and metrics, we taxonomize the space and select representative subsets. We evaluate models on 16 core scenarios and 7 metrics, exposing important trade\u2010offs. We supplement our core evaluation with seven targeted evaluations to deeply analyze specific aspects (including world knowledge, reasoning, regurgitation of copyrighted content, and generation of disinformation). We benchmark 30 LMs, from OpenAI, Microsoft, Google, Meta, Cohere, AI21 Labs, and others. Prior to HELM, models were evaluated on just 17.9% of the core HELM scenarios, with some prominent models not sharing a single scenario in common. We improve this to 96.0%: all 30 models are now benchmarked under the same standardized conditions. Our evaluation surfaces 25 top\u2010level findings. For full transparency, we release all raw model prompts and completions publicly. HELM is a living benchmark for the community, continuously updated with new scenarios, metrics, and models https://crfm.stanford.edu/helm/latest/."
                },
                {
                    "title": "Enhancing Chat Language Models by Scaling High-quality Instructional Conversations",
                    "abstract": "Fine-tuning on instruction data has been widely validated as an effective practice for implementing chat language models like ChatGPT. Scaling the diversity and quality of such data, although straightforward, stands a great chance of leading to improved performance. This paper aims to improve the upper bound of open-source models further. We first provide a systematically designed, diverse, informative, large-scale dataset of instructional conversations, UltraChat, which does not involve human queries. Our objective is to capture the breadth of interactions that a human might have with an AI assistant and employs a comprehensive framework to generate multi-turn conversation iteratively. UltraChat contains 1.5 million high-quality multi-turn dialogues and covers a wide range of topics and instructions. Our statistical analysis of UltraChat reveals its superiority in various key metrics, including scale, average length, diversity, coherence, etc., solidifying its position as a leading open-source dataset. Building upon UltraChat, we fine-tune a LLaMA model to create a powerful conversational model, UltraLLaMA. Our evaluations indicate that UltraLLaMA consistently outperforms other open-source models, including Vicuna, the previously recognized state-of-the-art open-source model. The dataset and the model will be publicly released\\footnote{\\url{https://github.com/thunlp/UltraChat}}."
                },
                {
                    "title": "Stop Uploading Test Data in Plain Text: Practical Strategies for Mitigating Data Contamination by Evaluation Benchmarks",
                    "abstract": "Data contamination has become prevalent and challenging with the rise of models pretrained on large automatically-crawled corpora. For closed models, the training data becomes a trade secret, and even for open models, it is not trivial to detect contamination. Strategies such as leaderboards with hidden answers, or using test data which is guaranteed to be unseen, are expensive and become fragile with time. Assuming that all relevant actors value clean test data and will cooperate to mitigate data contamination, what can be done? We propose three strategies that can make a difference: (1) Test data made public should be encrypted with a public key and licensed to disallow derivative distribution; (2) demand training exclusion controls from closed API holders, and protect your test data by refusing to evaluate without them; (3) avoid data which appears with its solution on the internet, and release the web-page context of internet-derived data along with the data. These strategies are practical and can be effective in preventing data contamination."
                },
                {
                    "title": "Tree of Thoughts: Deliberate Problem Solving with Large Language Models",
                    "abstract": "Language models are increasingly being deployed for general problem solving across a wide range of tasks, but are still confined to token-level, left-to-right decision-making processes during inference. This means they can fall short in tasks that require exploration, strategic lookahead, or where initial decisions play a pivotal role. To surmount these challenges, we introduce a new framework for language model inference, Tree of Thoughts (ToT), which generalizes over the popular Chain of Thought approach to prompting language models, and enables exploration over coherent units of text (thoughts) that serve as intermediate steps toward problem solving. ToT allows LMs to perform deliberate decision making by considering multiple different reasoning paths and self-evaluating choices to decide the next course of action, as well as looking ahead or backtracking when necessary to make global choices. Our experiments show that ToT significantly enhances language models' problem-solving abilities on three novel tasks requiring non-trivial planning or search: Game of 24, Creative Writing, and Mini Crosswords. For instance, in Game of 24, while GPT-4 with chain-of-thought prompting only solved 4% of tasks, our method achieved a success rate of 74%. Code repo with all prompts: https://github.com/princeton-nlp/tree-of-thought-llm."
                },
                {
                    "title": "Symbol tuning improves in-context learning in language models",
                    "abstract": "We present symbol tuning - finetuning language models on in-context input-label pairs where natural language labels (e.g.,\"positive/negative sentiment\") are replaced with arbitrary symbols (e.g.,\"foo/bar\"). Symbol tuning leverages the intuition that when a model cannot use instructions or natural language labels to figure out a task, it must instead do so by learning the input-label mappings. We experiment with symbol tuning across Flan-PaLM models up to 540B parameters and observe benefits across various settings. First, symbol tuning boosts performance on unseen in-context learning tasks and is much more robust to underspecified prompts, such as those without instructions or without natural language labels. Second, symbol-tuned models are much stronger at algorithmic reasoning tasks, with up to 18.2% better performance on the List Functions benchmark and up to 15.3% better performance on the Simple Turing Concepts benchmark. Finally, symbol-tuned models show large improvements in following flipped-labels presented in-context, meaning that they are more capable of using in-context information to override prior semantic knowledge."
                },
                {
                    "title": "C-Eval: A Multi-Level Multi-Discipline Chinese Evaluation Suite for Foundation Models",
                    "abstract": "New NLP benchmarks are urgently needed to align with the rapid development of large language models (LLMs). We present C-Eval, the first comprehensive Chinese evaluation suite designed to assess advanced knowledge and reasoning abilities of foundation models in a Chinese context. C-Eval comprises multiple-choice questions across four difficulty levels: middle school, high school, college, and professional. The questions span 52 diverse disciplines, ranging from humanities to science and engineering. C-Eval is accompanied by C-Eval Hard, a subset of very challenging subjects in C-Eval that requires advanced reasoning abilities to solve. We conduct a comprehensive evaluation of the most advanced LLMs on C-Eval, including both English- and Chinese-oriented models. Results indicate that only GPT-4 could achieve an average accuracy of over 60%, suggesting that there is still significant room for improvement for current LLMs. We anticipate C-Eval will help analyze important strengths and shortcomings of foundation models, and foster their development and growth for Chinese users."
                },
                {
                    "title": "FactKB: Generalizable Factuality Evaluation using Language Models Enhanced with Factual Knowledge",
                    "abstract": "Evaluating the factual consistency of automatically generated summaries is essential for the progress and adoption of reliable summarization systems. Despite recent advances, existing factuality evaluation models are not robust, being especially prone to entity and relation errors in new domains. We propose FactKB, a simple new approach to factuality evaluation that is generalizable across domains, in particular with respect to entities and relations. FactKB is based on language models pretrained using facts extracted from external knowledge bases. We introduce three types of complementary factuality pretraining objectives based on direct entity facts, facts grounded in auxiliary knowledge about entities, and facts constructed compositionally through knowledge base walks. The resulting factuality evaluation model achieves state-of-the-art performance on two in-domain news summarization benchmarks as well as on three out-of-domain scientific literature datasets. Further analysis of FactKB shows improved ability to detect erroneous entities and relations in summaries and is robust and generalizable across domains."
                },
                {
                    "title": "WizardLM: Empowering Large Language Models to Follow Complex Instructions",
                    "abstract": "Training large language models (LLMs) with open-domain instruction following data brings colossal success. However, manually creating such instruction data is very time-consuming and labor-intensive. Moreover, humans may struggle to produce high-complexity instructions. In this paper, we show an avenue for creating large amounts of instruction data with varying levels of complexity using LLM instead of humans. Starting with an initial set of instructions, we use our proposed Evol-Instruct to rewrite them step by step into more complex instructions. Then, we mix all generated instruction data to fine-tune LLaMA. We call the resulting model WizardLM. Human evaluations on a complexity-balanced test bed and Vicuna's testset show that instructions from Evol-Instruct are superior to human-created ones. By analyzing the human evaluation results of the high complexity part, we demonstrate that outputs from our WizardLM are preferred to outputs from OpenAI ChatGPT. In GPT-4 automatic evaluation, WizardLM achieves more than 90\\% capacity of ChatGPT on 17 out of 29 skills. Even though WizardLM still lags behind ChatGPT in some aspects, our findings suggest that fine-tuning with AI-evolved instructions is a promising direction for enhancing LLMs. Our code and data are public at https://github.com/nlpxucan/WizardLM"
                },
                {
                    "title": "Emergent and Predictable Memorization in Large Language Models",
                    "abstract": "Memorization, or the tendency of large language models (LLMs) to output entire sequences from their training data verbatim, is a key concern for safely deploying language models. In particular, it is vital to minimize a model's memorization of sensitive datapoints such as those containing personal identifiable information (PII). The prevalence of such undesirable memorization can pose issues for model trainers, and may even require discarding an otherwise functional model. We therefore seek to predict which sequences will be memorized before a large model's full train-time by extrapolating the memorization behavior of lower-compute trial runs. We measure memorization of the Pythia model suite and plot scaling laws for forecasting memorization, allowing us to provide equi-compute recommendations to maximize the reliability (recall) of such predictions. We additionally provide further novel discoveries on the distribution of memorization scores across models and data. We release all code and data necessary to reproduce the results in this paper at https://github.com/EleutherAI/pythia"
                },
                {
                    "title": "Chameleon: Plug-and-Play Compositional Reasoning with Large Language Models",
                    "abstract": "Large language models (LLMs) have achieved remarkable progress in solving various natural language processing tasks due to emergent reasoning abilities. However, LLMs have inherent limitations as they are incapable of accessing up-to-date information (stored on the Web or in task-specific knowledge bases), using external tools, and performing precise mathematical and logical reasoning. In this paper, we present Chameleon, an AI system that mitigates these limitations by augmenting LLMs with plug-and-play modules for compositional reasoning. Chameleon synthesizes programs by composing various tools (e.g., LLMs, off-the-shelf vision models, web search engines, Python functions, and heuristic-based modules) for accomplishing complex reasoning tasks. At the heart of Chameleon is an LLM-based planner that assembles a sequence of tools to execute to generate the final response. We showcase the effectiveness of Chameleon on two multi-modal knowledge-intensive reasoning tasks: ScienceQA and TabMWP. Chameleon, powered by GPT-4, achieves an 86.54% overall accuracy on ScienceQA, improving the best published few-shot result by 11.37%. On TabMWP, GPT-4-powered Chameleon improves the accuracy by 17.0%, lifting the state of the art to 98.78%. Our analysis also shows that the GPT-4-powered planner exhibits more consistent and rational tool selection via inferring potential constraints from instructions, compared to a ChatGPT-powered planner. The project is available at https://chameleon-llm.github.io."
                },
                {
                    "title": "AGIEval: A Human-Centric Benchmark for Evaluating Foundation Models",
                    "abstract": "Evaluating the general abilities of foundation models to tackle human-level tasks is a vital aspect of their development and application in the pursuit of Artificial General Intelligence (AGI). Traditional benchmarks, which rely on artificial datasets, may not accurately represent human-level capabilities. In this paper, we introduce AGIEval, a novel benchmark specifically designed to assess foundation model in the context of human-centric standardized exams, such as college entrance exams, law school admission tests, math competitions, and lawyer qualification tests. We evaluate several state-of-the-art foundation models, including GPT-4, ChatGPT, and Text-Davinci-003, using this benchmark. Impressively, GPT-4 surpasses average human performance on SAT, LSAT, and math competitions, attaining a 95% accuracy rate on the SAT Math test and a 92.5% accuracy on the English test of the Chinese national college entrance exam. This demonstrates the extraordinary performance of contemporary foundation models. In contrast, we also find that GPT-4 is less proficient in tasks that require complex reasoning or specific domain knowledge. Our comprehensive analyses of model capabilities (understanding, knowledge, reasoning, and calculation) reveal these models' strengths and limitations, providing valuable insights into future directions for enhancing their general capabilities. By concentrating on tasks pertinent to human cognition and decision-making, our benchmark delivers a more meaningful and robust evaluation of foundation models' performance in real-world scenarios. The data, code, and all model outputs are released in https://github.com/ruixiangcui/AGIEval."
                },
                {
                    "title": "Self-Refine: Iterative Refinement with Self-Feedback",
                    "abstract": "Like humans, large language models (LLMs) do not always generate the best output on their first try. Motivated by how humans refine their written text, we introduce Self-Refine, an approach for improving initial outputs from LLMs through iterative feedback and refinement. The main idea is to generate an initial output using an LLMs; then, the same LLMs provides feedback for its output and uses it to refine itself, iteratively. Self-Refine does not require any supervised training data, additional training, or reinforcement learning, and instead uses a single LLM as the generator, refiner, and feedback provider. We evaluate Self-Refine across 7 diverse tasks, ranging from dialog response generation to mathematical reasoning, using state-of-the-art (GPT-3.5, ChatGPT, and GPT-4) LLMs. Across all evaluated tasks, outputs generated with Self-Refine are preferred by humans and automatic metrics over those generated with the same LLM using conventional one-step generation, improving by ~20% absolute on average in task performance. Our work demonstrates that even state-of-the-art LLMs like GPT-4 can be further improved at test time using our simple, standalone approach."
                },
                {
                    "title": "Sparks of Artificial General Intelligence: Early experiments with GPT-4",
                    "abstract": "Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. The latest model developed by OpenAI, GPT-4, was trained using an unprecedented scale of compute and data. In this paper, we report on our investigation of an early version of GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT-4 is part of a new cohort of LLMs (along with ChatGPT and Google's PaLM for example) that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4's performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4's capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system. In our exploration of GPT-4, we put special emphasis on discovering its limitations, and we discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for pursuing a new paradigm that moves beyond next-word prediction. We conclude with reflections on societal influences of the recent technological leap and future research directions."
                },
                {
                    "title": "Reflexion: language agents with verbal reinforcement learning",
                    "abstract": "Large language models (LLMs) have been increasingly used to interact with external environments (e.g., games, compilers, APIs) as goal-driven agents. However, it remains challenging for these language agents to quickly and efficiently learn from trial-and-error as traditional reinforcement learning methods require extensive training samples and expensive model fine-tuning. We propose Reflexion, a novel framework to reinforce language agents not by updating weights, but instead through linguistic feedback. Concretely, Reflexion agents verbally reflect on task feedback signals, then maintain their own reflective text in an episodic memory buffer to induce better decision-making in subsequent trials. Reflexion is flexible enough to incorporate various types (scalar values or free-form language) and sources (external or internally simulated) of feedback signals, and obtains significant improvements over a baseline agent across diverse tasks (sequential decision-making, coding, language reasoning). For example, Reflexion achieves a 91% pass@1 accuracy on the HumanEval coding benchmark, surpassing the previous state-of-the-art GPT-4 that achieves 80%. We also conduct ablation and analysis studies using different feedback signals, feedback incorporation methods, and agent types, and provide insights into how they affect performance."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Learning Performance-Improving Code Edits",
                    "abstract": "The waning of Moore's Law has shifted the focus of the tech industry towards alternative methods for continued performance gains. While optimizing compilers are a standard tool to help increase program efficiency, programmers continue to shoulder much responsibility in crafting and refactoring code with better performance characteristics. In this paper, we investigate the ability of large language models (LLMs) to suggest functionally correct, performance improving code edits. We hypothesize that language models can suggest such edits in ways that would be impractical for static analysis alone. We investigate these questions by curating a large-scale dataset of Performance-Improving Edits, PIE. PIE contains trajectories of programs, where a programmer begins with an initial, slower version and iteratively makes changes to improve the program's performance. We use PIE to evaluate and improve the capacity of large language models. Specifically, use examples from PIE to fine-tune multiple variants of CODEGEN, a billion-scale Transformer-decoder model. Additionally, we use examples from PIE to prompt OpenAI's CODEX using a few-shot prompting. By leveraging PIE, we find that both CODEX and CODEGEN can generate performance-improving edits, with speedups of more than 2.5x for over 25% of the programs, for C++ and Python, even after the C++ programs were compiled using the O3 optimization level. Crucially, we show that PIE allows CODEGEN, an open-sourced and 10x smaller model than CODEX, to match the performance of CODEX on this challenging task. Overall, this work opens new doors for creating systems and methods that can help programmers write efficient code."
                },
                {
                    "title": "Augmented Language Models: a Survey",
                    "abstract": "This survey reviews works in which language models (LMs) are augmented with reasoning skills and the ability to use tools. The former is defined as decomposing a potentially complex task into simpler subtasks while the latter consists in calling external modules such as a code interpreter. LMs can leverage these augmentations separately or in combination via heuristics, or learn to do so from demonstrations. While adhering to a standard missing tokens prediction objective, such augmented LMs can use various, possibly non-parametric external modules to expand their context processing ability, thus departing from the pure language modeling paradigm. We therefore refer to them as Augmented Language Models (ALMs). The missing token objective allows ALMs to learn to reason, use tools, and even act, while still performing standard natural language tasks and even outperforming most regular LMs on several benchmarks. In this work, after reviewing current advance in ALMs, we conclude that this new research direction has the potential to address common limitations of traditional LMs such as interpretability, consistency, and scalability issues."
                },
                {
                    "title": "Toolformer: Language Models Can Teach Themselves to Use Tools",
                    "abstract": "Language models (LMs) exhibit remarkable abilities to solve new tasks from just a few examples or textual instructions, especially at scale. They also, paradoxically, struggle with basic functionality, such as arithmetic or factual lookup, where much simpler and smaller models excel. In this paper, we show that LMs can teach themselves to use external tools via simple APIs and achieve the best of both worlds. We introduce Toolformer, a model trained to decide which APIs to call, when to call them, what arguments to pass, and how to best incorporate the results into future token prediction. This is done in a self-supervised way, requiring nothing more than a handful of demonstrations for each API. We incorporate a range of tools, including a calculator, a Q\\&A system, two different search engines, a translation system, and a calendar. Toolformer achieves substantially improved zero-shot performance across a variety of downstream tasks, often competitive with much larger models, without sacrificing its core language modeling abilities."
                },
                {
                    "title": "Self-Instruct: Aligning Language Models with Self-Generated Instructions",
                    "abstract": "Large \u201cinstruction-tuned\u201d language models (i.e., finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is often limited in quantity, diversity, and creativity, therefore hindering the generality of the tuned model. We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off their own generations. Our pipeline generates instructions, input, and output samples from a language model, then filters invalid or similar ones before using them to finetune the original model. Applying our method to the vanilla GPT3, we demonstrate a 33% absolute improvement over the original model on Super-NaturalInstructions, on par with the performance of InstructGPT-001, which was trained with private user data and human annotations. For further evaluation, we curate a set of expert-written instructions for novel tasks, and show through human evaluation that tuning GPT3 with Self-Instruct outperforms using existing public instruction datasets by a large margin, leaving only a 5% absolute gap behind InstructGPT-001. Self-Instruct provides an almost annotation-free method for aligning pre-trained language models with instructions, and we release our large synthetic dataset to facilitate future studies on instruction tuning."
                },
                {
                    "title": "Discovering Language Model Behaviors with Model-Written Evaluations",
                    "abstract": "As language models (LMs) scale, they develop many novel behaviors, good and bad, exacerbating the need to evaluate how they behave. Prior work creates evaluations with crowdwork (which is time-consuming and expensive) or existing data sources (which are not always available). Here, we automatically generate evaluations with LMs. We explore approaches with varying amounts of human effort, from instructing LMs to write yes/no questions to making complex Winogender schemas with multiple stages of LM-based generation and filtering. Crowdworkers rate the examples as highly relevant and agree with 90-100% of labels, sometimes more so than corresponding human-written datasets. We generate 154 datasets and discover new cases of inverse scaling where LMs get worse with size. Larger LMs repeat back a dialog user's preferred answer (\"sycophancy\") and express greater desire to pursue concerning goals like resource acquisition and goal preservation. We also find some of the first examples of inverse scaling in RL from Human Feedback (RLHF), where more RLHF makes LMs worse. For example, RLHF makes LMs express stronger political views (on gun rights and immigration) and a greater desire to avoid shut down. Overall, LM-written evaluations are high-quality and let us quickly discover many novel LM behaviors."
                },
                {
                    "title": "On Second Thought, Let\u2019s Not Think Step by Step! Bias and Toxicity in Zero-Shot Reasoning",
                    "abstract": "Generating a Chain of Thought (CoT) has been shown to consistently improve large language model (LLM) performance on a wide range of NLP tasks. However, prior work has mainly focused on logical reasoning tasks (e.g. arithmetic, commonsense QA); it remains unclear whether improvements hold for more diverse types of reasoning, especially in socially situated contexts. Concretely, we perform a controlled evaluation of zero-shot CoT across two socially sensitive domains: harmful questions and stereotype benchmarks. We find that zero-shot CoT reasoning in sensitive domains significantly increases a model\u2019s likelihood to produce harmful or undesirable output, with trends holding across different prompt formats and model variants. Furthermore, we show that harmful CoTs increase with model size, but decrease with improved instruction following. Our work suggests that zero-shot CoT should be used with caution on socially important tasks, especially when marginalized groups or sensitive topics are involved."
                },
                {
                    "title": "Adaptive Testing of Computer Vision Models",
                    "abstract": "Vision models often fail systematically on groups of data that share common semantic characteristics (e.g., rare objects or unusual scenes), but identifying these failure modes is a challenge. We introduce AdaVision, an interactive process for testing vision models which helps users identify and fix coherent failure modes. Given a natural language description of a coherent group, AdaVision retrieves relevant images from LAION-5B with CLIP. The user then labels a small amount of data for model correctness, which is used in successive retrieval rounds to hill-climb towards high-error regions, refining the group definition. Once a group is saturated, AdaVision uses GPT-3 to suggest new group descriptions for the user to explore. We demonstrate the usefulness and generality of AdaVision in user studies, where users find major bugs in state-of-the-art classification, object detection, and image captioning models. These user-discovered groups have failure rates 2-3x higher than those surfaced by automatic error clustering methods. Finally, finetuning on examples found with AdaVision fixes the discovered bugs when evaluated on unseen examples, without degrading in-distribution accuracy, and while also improving performance on out-of-distribution datasets."
                },
                {
                    "title": "PAL: Program-aided Language Models",
                    "abstract": "Large language models (LLMs) have recently demonstrated an impressive ability to perform arithmetic and symbolic reasoning tasks, when provided with a few examples at test time (\"few-shot prompting\"). Much of this success can be attributed to prompting methods such as\"chain-of-thought'', which employ LLMs for both understanding the problem description by decomposing it into steps, as well as solving each step of the problem. While LLMs seem to be adept at this sort of step-by-step decomposition, LLMs often make logical and arithmetic mistakes in the solution part, even when the problem is decomposed correctly. In this paper, we present Program-Aided Language models (PAL): a novel approach that uses the LLM to read natural language problems and generate programs as the intermediate reasoning steps, but offloads the solution step to a runtime such as a Python interpreter. With PAL, decomposing the natural language problem into runnable steps remains the only learning task for the LLM, while solving is delegated to the interpreter. We demonstrate this synergy between a neural LLM and a symbolic interpreter across 13 mathematical, symbolic, and algorithmic reasoning tasks from BIG-Bench Hard and other benchmarks. In all these natural language reasoning tasks, generating code using an LLM and reasoning using a Python interpreter leads to more accurate results than much larger models. For example, PAL using Codex achieves state-of-the-art few-shot accuracy on the GSM8K benchmark of math word problems, surpassing PaLM-540B which uses chain-of-thought by absolute 15% top-1. Our code and data are publicly available at http://reasonwithpal.com/ ."
                },
                {
                    "title": "Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them",
                    "abstract": "BIG-Bench (Srivastava et al., 2022) is a diverse evaluation suite that focuses on tasks believed to be beyond the capabilities of current language models. Language models have already made good progress on this benchmark, with the best model in the BIG-Bench paper outperforming average reported human-rater results on 65% of the BIG-Bench tasks via few-shot prompting. But on what tasks do language models fall short of average human-rater performance, and are those tasks actually unsolvable by current language models? In this work, we focus on a suite of 23 challenging BIG-Bench tasks which we call BIG-Bench Hard (BBH). These are the task for which prior language model evaluations did not outperform the average human-rater. We find that applying chain-of-thought (CoT) prompting to BBH tasks enables PaLM to surpass the average human-rater performance on 10 of the 23 tasks, and Codex (code-davinci-002) to surpass the average human-rater performance on 17 of the 23 tasks. Since many tasks in BBH require multi-step reasoning, few-shot prompting without CoT, as done in the BIG-Bench evaluations (Srivastava et al., 2022), substantially underestimates the best performance and capabilities of language models, which is better captured via CoT prompting. As further analysis, we explore the interaction between CoT and model scale on BBH, finding that CoT enables emergent task performance on several BBH tasks with otherwise flat scaling curves."
                },
                {
                    "title": "Prompting GPT-3 To Be Reliable",
                    "abstract": "Large language models (LLMs) show impressive abilities via few-shot prompting. Commercialized APIs such as OpenAI GPT-3 further increase their use in real-world language applications. However, the crucial problem of how to improve the reliability of GPT-3 is still under-explored. While reliability is a broad and vaguely defined term, we decompose reliability into four main facets that correspond to the existing framework of ML safety and are well-recognized to be important: generalizability, social biases, calibration, and factuality. Our core contribution is to establish simple and effective prompts that improve GPT-3's reliability as it: 1) generalizes out-of-distribution, 2) balances demographic distribution and uses natural language instructions to reduce social biases, 3) calibrates output probabilities, and 4) updates the LLM's factual knowledge and reasoning chains. With appropriate prompts, GPT-3 is more reliable than smaller-scale supervised models on all these facets. We release all processed datasets, evaluation scripts, and model predictions. Our systematic empirical study not only sheds new insights on the reliability of prompting LLMs, but more importantly, our prompting strategies can help practitioners more reliably use LLMs like GPT-3."
                },
                {
                    "title": "Deep Bidirectional Language-Knowledge Graph Pretraining",
                    "abstract": "Pretraining a language model (LM) on text has been shown to help various downstream NLP tasks. Recent works show that a knowledge graph (KG) can complement text data, offering structured background knowledge that provides a useful scaffold for reasoning. However, these works are not pretrained to learn a deep fusion of the two modalities at scale, limiting the potential to acquire fully joint representations of text and KG. Here we propose DRAGON (Deep Bidirectional Language-Knowledge Graph Pretraining), a self-supervised approach to pretraining a deeply joint language-knowledge foundation model from text and KG at scale. Specifically, our model takes pairs of text segments and relevant KG subgraphs as input and bidirectionally fuses information from both modalities. We pretrain this model by unifying two self-supervised reasoning tasks, masked language modeling and KG link prediction. DRAGON outperforms existing LM and LM+KG models on diverse downstream tasks including question answering across general and biomedical domains, with +5% absolute gain on average. In particular, DRAGON achieves notable performance on complex reasoning about language and knowledge (+10% on questions involving long contexts or multi-step reasoning) and low-resource QA (+8% on OBQA and RiddleSense), and new state-of-the-art results on various BioNLP tasks. Our code and trained models are available at https://github.com/michiyasunaga/dragon."
                },
                {
                    "title": "ReAct: Synergizing Reasoning and Acting in Language Models",
                    "abstract": "While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources, such as knowledge bases or environments, to gather additional information. We apply our approach, named ReAct, to a diverse set of language and decision making tasks and demonstrate its effectiveness over state-of-the-art baselines, as well as improved human interpretability and trustworthiness over methods without reasoning or acting components. Concretely, on question answering (HotpotQA) and fact verification (Fever), ReAct overcomes issues of hallucination and error propagation prevalent in chain-of-thought reasoning by interacting with a simple Wikipedia API, and generates human-like task-solving trajectories that are more interpretable than baselines without reasoning traces. On two interactive decision making benchmarks (ALFWorld and WebShop), ReAct outperforms imitation and reinforcement learning methods by an absolute success rate of 34% and 10% respectively, while being prompted with only one or two in-context examples. Project site with code: https://react-lm.github.io"
                },
                {
                    "title": "Language Models Can Teach Themselves to Program Better",
                    "abstract": "Recent Language Models (LMs) achieve breakthrough performance in code generation when trained on human-authored problems, even solving some competitive-programming problems. Self-play has proven useful in games such as Go, and thus it is natural to ask whether LMs can generate their own instructive programming problems to improve their performance. We show that it is possible for an LM to synthesize programming problems and solutions, which are filtered for correctness by a Python interpreter. The LM's performance is then seen to improve when it is fine-tuned on its own synthetic problems and verified solutions; thus the model 'improves itself' using the Python interpreter. Problems are specified formally as programming puzzles [Schuster et al., 2021], a code-based problem format where solutions can easily be verified for correctness by execution. In experiments on publicly-available LMs, test accuracy more than doubles. This work demonstrates the potential for code LMs, with an interpreter, to generate instructive problems and improve their own performance."
                },
                {
                    "title": "Inner Monologue: Embodied Reasoning through Planning with Language Models",
                    "abstract": "Recent works have shown how the reasoning capabilities of Large Language Models (LLMs) can be applied to domains beyond natural language processing, such as planning and interaction for robots. These embodied problems require an agent to understand many semantic aspects of the world: the repertoire of skills available, how these skills influence the world, and how changes to the world map back to the language. LLMs planning in embodied environments need to consider not just what skills to do, but also how and when to do them - answers that change over time in response to the agent's own choices. In this work, we investigate to what extent LLMs used in such embodied contexts can reason over sources of feedback provided through natural language, without any additional training. We propose that by leveraging environment feedback, LLMs are able to form an inner monologue that allows them to more richly process and plan in robotic control scenarios. We investigate a variety of sources of feedback, such as success detection, scene description, and human interaction. We find that closed-loop language feedback significantly improves high-level instruction completion on three domains, including simulated and real table top rearrangement tasks and long-horizon mobile manipulation tasks in a kitchen environment in the real world."
                },
                {
                    "title": "Large Language Models are Zero-Shot Reasoners",
                    "abstract": "Pretrained large language models (LLMs) are widely used in many sub-fields of natural language processing (NLP) and generally known as excellent few-shot learners with task-specific exemplars. Notably, chain of thought (CoT) prompting, a recent technique for eliciting complex multi-step reasoning through step-by-step answer examples, achieved the state-of-the-art performances in arithmetics and symbolic reasoning, difficult system-2 tasks that do not follow the standard scaling laws for LLMs. While these successes are often attributed to LLMs' ability for few-shot learning, we show that LLMs are decent zero-shot reasoners by simply adding\"Let's think step by step\"before each answer. Experimental results demonstrate that our Zero-shot-CoT, using the same single prompt template, significantly outperforms zero-shot LLM performances on diverse benchmark reasoning tasks including arithmetics (MultiArith, GSM8K, AQUA-RAT, SVAMP), symbolic reasoning (Last Letter, Coin Flip), and other logical reasoning tasks (Date Understanding, Tracking Shuffled Objects), without any hand-crafted few-shot examples, e.g. increasing the accuracy on MultiArith from 17.7% to 78.7% and GSM8K from 10.4% to 40.7% with large InstructGPT model (text-davinci-002), as well as similar magnitudes of improvements with another off-the-shelf large model, 540B parameter PaLM. The versatility of this single prompt across very diverse reasoning tasks hints at untapped and understudied fundamental zero-shot capabilities of LLMs, suggesting high-level, multi-task broad cognitive capabilities may be extracted by simple prompting. We hope our work not only serves as the minimal strongest zero-shot baseline for the challenging reasoning benchmarks, but also highlights the importance of carefully exploring and analyzing the enormous zero-shot knowledge hidden inside LLMs before crafting finetuning datasets or few-shot exemplars."
                },
                {
                    "title": "Least-to-Most Prompting Enables Complex Reasoning in Large Language Models",
                    "abstract": "Chain-of-thought prompting has demonstrated remarkable performance on various natural language reasoning tasks. However, it tends to perform poorly on tasks which requires solving problems harder than the exemplars shown in the prompts. To overcome this challenge of easy-to-hard generalization, we propose a novel prompting strategy, least-to-most prompting. The key idea in this strategy is to break down a complex problem into a series of simpler subproblems and then solve them in sequence. Solving each subproblem is facilitated by the answers to previously solved subproblems. Our experimental results on tasks related to symbolic manipulation, compositional generalization, and math reasoning reveal that least-to-most prompting is capable of generalizing to more difficult problems than those seen in the prompts. A notable finding is that when the GPT-3 code-davinci-002 model is used with least-to-most prompting, it can solve the compositional generalization benchmark SCAN in any split (including length split) with an accuracy of at least 99% using just 14 exemplars, compared to only 16% accuracy with chain-of-thought prompting. This is particularly noteworthy because neural-symbolic models in the literature that specialize in solving SCAN are trained on the entire training set containing over 15,000 examples. We have included prompts for all the tasks in the Appendix."
                },
                {
                    "title": "STaR: Bootstrapping Reasoning With Reasoning",
                    "abstract": "Generating step-by-step\"chain-of-thought\"rationales improves language model performance on complex reasoning tasks like mathematics or commonsense question-answering. However, inducing language model rationale generation currently requires either constructing massive rationale datasets or sacrificing accuracy by using only few-shot inference. We propose a technique to iteratively leverage a small number of rationale examples and a large dataset without rationales, to bootstrap the ability to perform successively more complex reasoning. This technique, the\"Self-Taught Reasoner\"(STaR), relies on a simple loop: generate rationales to answer many questions, prompted with a few rationale examples; if the generated answers are wrong, try again to generate a rationale given the correct answer; fine-tune on all the rationales that ultimately yielded correct answers; repeat. We show that STaR significantly improves performance on multiple datasets compared to a model fine-tuned to directly predict final answers, and performs comparably to fine-tuning a 30$\\times$ larger state-of-the-art language model on CommensenseQA. Thus, STaR lets a model improve itself by learning from its own generated reasoning."
                },
                {
                    "title": "Data Contamination: From Memorization to Exploitation",
                    "abstract": "Pretrained language models are typically trained on massive web-based datasets, which are often \u201ccontaminated\u201d with downstream test sets. It is not clear to what extent models exploit the contaminated data for downstream tasks. We present a principled method to study this question. We pretrain BERT models on joint corpora of Wikipedia and labeled downstream datasets, and fine-tune them on the relevant task. Comparing performance between samples seen and unseen during pretraining enables us to define and quantify levels of memorization and exploitation.Experiments with two models and three downstream tasks show that exploitation exists in some cases, but in others the models memorize the contaminated data, but do not exploit it. We show that these two measures are affected by different factors such as the number of duplications of the contaminated data and the model size. Our results highlight the importance of analyzing massive web-scale datasets to verify that progress in NLP is obtained by better language understanding and not better data exploitation."
                },
                {
                    "title": "Quantifying Memorization Across Neural Language Models",
                    "abstract": "Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations."
                },
                {
                    "title": "Survey of Hallucination in Natural Language Generation",
                    "abstract": "Natural Language Generation (NLG) has improved exponentially in recent years thanks to the development of sequence-to-sequence deep learning technologies such as Transformer-based language models. This advancement has led to more fluent and coherent NLG, leading to improved development in downstream tasks such as abstractive summarization, dialogue generation, and data-to-text generation. However, it is also apparent that deep learning based generation is prone to hallucinate unintended text, which degrades the system performance and fails to meet user expectations in many real-world scenarios. To address this issue, many studies have been presented in measuring and mitigating hallucinated texts, but these have never been reviewed in a comprehensive manner before. In this survey, we thus provide a broad overview of the research progress and challenges in the hallucination problem in NLG. The survey is organized into two parts: (1) a general overview of metrics, mitigation methods, and future directions, and (2) an overview of task-specific research progress on hallucinations in the following downstream tasks, namely abstractive summarization, dialogue generation, generative question answering, data-to-text generation, and machine translation. This survey serves to facilitate collaborative efforts among researchers in tackling the challenge of hallucinated texts in NLG."
                },
                {
                    "title": "Red Teaming Language Models with Language Models",
                    "abstract": "Language Models (LMs) often cannot be deployed because of their potential to harm users in hard-to-predict ways. Prior work identifies harmful behaviors before deployment by using human annotators to hand-write test cases. However, human annotation is expensive, limiting the number and diversity of test cases. In this work, we automatically find cases where a target LM behaves in a harmful way, by generating test cases (\u201cred teaming\u201d) using another LM. We evaluate the target LM\u2019s replies to generated test questions using a classifier trained to detect offensive content, uncovering tens of thousands of offensive replies in a 280B parameter LM chatbot. We explore several methods, from zero-shot generation to reinforcement learning, for generating test cases with varying levels of diversity and difficulty. Furthermore, we use prompt engineering to control LM-generated test cases to uncover a variety of other harms, automatically finding groups of people that the chatbot discusses in offensive ways, personal and hospital phone numbers generated as the chatbot\u2019s own contact info, leakage of private training data in generated text, and harms that occur over the course of a conversation. Overall, LM-based red teaming is one promising tool (among many needed) for finding and fixing diverse, undesirable LM behaviors before impacting users."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "GreaseLM: Graph REASoning Enhanced Language Models for Question Answering",
                    "abstract": "Answering complex questions about textual narratives requires reasoning over both stated context and the world knowledge that underlies it. However, pretrained language models (LM), the foundation of most modern QA systems, do not robustly represent latent relationships between concepts, which is necessary for reasoning. While knowledge graphs (KG) are often used to augment LMs with structured representations of world knowledge, it remains an open question how to effectively fuse and reason over the KG representations and the language context, which provides situational constraints and nuances. In this work, we propose GreaseLM, a new model that fuses encoded representations from pretrained LMs and graph neural networks over multiple layers of modality interaction operations. Information from both modalities propagates to the other, allowing language context representations to be grounded by structured world knowledge, and allowing linguistic nuances (e.g., negation, hedging) in the context to inform the graph representations of knowledge. Our results on three benchmarks in the commonsense reasoning (i.e., CommonsenseQA, OpenbookQA) and medical question answering (i.e., MedQA-USMLE) domains demonstrate that GreaseLM can more reliably answer questions that require reasoning over both situational constraints and structured knowledge, even outperforming models 8x larger."
                },
                {
                    "title": "A General Language Assistant as a Laboratory for Alignment",
                    "abstract": "Given the broad capabilities of large language models, it should be possible to work towards a general-purpose, text-based assistant that is aligned with human values, meaning that it is helpful, honest, and harmless. As an initial foray in this direction we study simple baseline techniques and evaluations, such as prompting. We find that the benefits from modest interventions increase with model size, generalize to a variety of alignment evaluations, and do not compromise the performance of large models. Next we investigate scaling trends for several training objectives relevant to alignment, comparing imitation learning, binary discrimination, and ranked preference modeling. We find that ranked preference modeling performs much better than imitation learning, and often scales more favorably with model size. In contrast, binary discrimination typically performs and scales very similarly to imitation learning. Finally we study a `preference model pre-training' stage of training, with the goal of improving sample efficiency when finetuning on human preferences."
                },
                {
                    "title": "MetaICL: Learning to Learn In Context",
                    "abstract": "We introduce MetaICL (Meta-training for In-Context Learning), a new meta-training framework for few-shot learning where a pretrained language model is tuned to do in-context learning on a large set of training tasks. This meta-training enables the model to more effectively learn a new task in context at test time, by simply conditioning on a few training examples with no parameter updates or task-specific templates. We experiment on a large, diverse collection of tasks consisting of 142 NLP datasets including classification, question answering, natural language inference, paraphrase detection and more, across seven different meta-training/target splits. MetaICL outperforms a range of baselines including in-context learning without meta-training and multi-task learning followed by zero-shot transfer. We find that the gains are particularly significant for target tasks that have domain shifts from the meta-training tasks, and that using a diverse set of the meta-training tasks is key to improvements. We also show that MetaICL approaches (and sometimes beats) the performance of models fully finetuned on the target task training data, and outperforms much bigger models with nearly 8x parameters."
                },
                {
                    "title": "Training Verifiers to Solve Math Word Problems",
                    "abstract": "State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline."
                },
                {
                    "title": "BBQ: A hand-built bias benchmark for question answering",
                    "abstract": "It is well documented that NLP models learn social biases, but little work has been done on how these biases manifest in model outputs for applied tasks like question answering (QA). We introduce the Bias Benchmark for QA (BBQ), a dataset of question-sets constructed by the authors that highlight attested social biases against people belonging to protected classes along nine social dimensions relevant for U.S. English-speaking contexts. Our task evaluate model responses at two levels: (i) given an under-informative context, we test how strongly responses reflect social biases, and (ii) given an adequately informative context, we test whether the model\u2019s biases override a correct answer choice. We find that models often rely on stereotypes when the context is under-informative, meaning the model\u2019s outputs consistently reproduce harmful biases in this setting. Though models are more accurate when the context provides an informative answer, they still rely on stereotypes and average up to 3.4 percentage points higher accuracy when the correct answer aligns with a social bias than when it conflicts, with this difference widening to over 5 points on examples targeting gender for most models tested."
                },
                {
                    "title": "Evaluating Large Language Models Trained on Code",
                    "abstract": "We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics."
                },
                {
                    "title": "Dynaboard: An Evaluation-As-A-Service Platform for Holistic Next-Generation Benchmarking",
                    "abstract": "We introduce Dynaboard, an evaluation-as-a-service framework for hosting benchmarks and conducting holistic model comparison, integrated with the Dynabench platform. Our platform evaluates NLP models directly instead of relying on self-reported metrics or predictions on a single dataset. Under this paradigm, models are submitted to be evaluated in the cloud, circumventing the issues of reproducibility, accessibility, and backwards compatibility that often hinder benchmarking in NLP. This allows users to interact with uploaded models in real time to assess their quality, and permits the collection of additional metrics such as memory use, throughput, and robustness, which -- despite their importance to practitioners -- have traditionally been absent from leaderboards. On each task, models are ranked according to the Dynascore, a novel utility-based aggregation of these statistics, which users can customize to better reflect their preferences, placing more/less weight on a particular axis of evaluation or dataset. As state-of-the-art NLP models push the limits of traditional benchmarks, Dynaboard offers a standardized solution for a more diverse and comprehensive evaluation of model quality."
                },
                {
                    "title": "Dynabench: Rethinking Benchmarking in NLP",
                    "abstract": "We introduce Dynabench, an open-source platform for dynamic dataset creation and model benchmarking. Dynabench runs in a web browser and supports human-and-model-in-the-loop dataset creation: annotators seek to create examples that a target model will misclassify, but that another person will not. In this paper, we argue that Dynabench addresses a critical need in our community: contemporary models quickly achieve outstanding performance on benchmark tasks but nonetheless fail on simple challenge examples and falter in real-world scenarios. With Dynabench, dataset creation, model development, and model assessment can directly inform each other, leading to more robust and informative benchmarks. We report on four initial NLP tasks, illustrating these concepts and highlighting the promise of the platform, and address potential objections to dynamic benchmarking as a new standard for the field."
                },
                {
                    "title": "Measuring Mathematical Problem Solving With the MATH Dataset",
                    "abstract": "Many intellectual endeavors require mathematical problem solving, but this skill remains beyond the capabilities of computers. To measure this ability in machine learning models, we introduce MATH, a new dataset of 12,500 challenging competition mathematics problems. Each problem in MATH has a full step-by-step solution which can be used to teach models to generate answer derivations and explanations. To facilitate future research and increase accuracy on MATH, we also contribute a large auxiliary pretraining dataset which helps teach models the fundamentals of mathematics. Even though we are able to increase accuracy on MATH, our results show that accuracy remains relatively low, even with enormous Transformer models. Moreover, we find that simply increasing budgets and model parameter counts will be impractical for achieving strong mathematical reasoning if scaling trends continue. While scaling Transformers is automatically solving most other text-based tasks, scaling is not currently solving MATH. To have more traction on mathematical problem solving we will likely need new algorithmic advancements from the broader research community."
                },
                {
                    "title": "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? \ud83e\udd9c",
                    "abstract": "The past 3 years of work in NLP have been characterized by the development and deployment of ever larger language models, especially for English. BERT, its variants, GPT-2/3, and others, most recently Switch-C, have pushed the boundaries of the possible both through architectural innovations and through sheer size. Using these pretrained models and the methodology of fine-tuning them for specific tasks, researchers have extended the state of the art on a wide array of tasks as measured by leaderboards on specific benchmarks for English. In this paper, we take a step back and ask: How big is too big? What are the possible risks associated with this technology and what paths are available for mitigating those risks? We provide recommendations including weighing the environmental and financial costs first, investing resources into curating and carefully documenting datasets rather than ingesting everything on the web, carrying out pre-development exercises evaluating how the planned approach fits into research and development goals and supports stakeholder values, and encouraging research directions beyond ever larger language models."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "Aligning AI With Shared Human Values",
                    "abstract": "We show how to assess a language model's knowledge of basic concepts of morality. We introduce the ETHICS dataset, a new benchmark that spans concepts in justice, well-being, duties, virtues, and commonsense morality. Models predict widespread moral judgments about diverse text scenarios. This requires connecting physical and social world knowledge to value judgements, a capability that may enable us to steer chatbot outputs or eventually regularize open-ended reinforcement learning agents. With the ETHICS dataset, we find that current language models have a promising but incomplete understanding of basic ethical knowledge. Our work shows that progress can be made on machine ethics today, and it provides a steppingstone toward AI that is aligned with human values."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Assessing The Factual Accuracy of Generated Text",
                    "abstract": "We propose a model-based metric to estimate the factual accuracy of generated text that is complementary to typical scoring schemes like ROUGE (Recall-Oriented Understudy for Gisting Evaluation) and BLEU (Bilingual Evaluation Understudy). We introduce and release a new large-scale dataset based on Wikipedia and Wikidata to train relation classifiers and end-to-end fact extraction models. The end-to-end models are shown to be able to extract complete sets of facts from datasets with full pages of text. We then analyse multiple models that estimate factual accuracy on a Wikipedia text summarization task, and show their efficacy compared to ROUGE and other model-free variants by conducting a human evaluation study."
                }
            ],
            "categories": [
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we dynamically and adaptively generate novel test samples for large language models (LLMs) that maintain controlled complexity and linguistic diversity, while ensuring the correctness of the generated data?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of static benchmarks that may not accurately reflect the evolving capabilities of LLMs. By developing a method to generate diverse and complex test samples, we can enhance the evaluation of LLMs, leading to better understanding of their generalization abilities. This advancement could pave the way for more robust models and improve practical applications in various fields, such as natural language processing, education, and AI-driven decision-making.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the need for a method that can generate test samples with varying complexities while ensuring linguistic diversity. Naive approaches, such as template-based generation, may fail due to their limited scope and lack of adaptability. Additionally, existing methods that modify current evaluation data often struggle with low controllability and instability, making it difficult to verify the quality and correctness of the generated samples. Overcoming these technical and practical obstacles requires a sophisticated approach to reasoning graph construction and perturbation.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on static benchmarks or template-based methods, which do not adequately address the need for dynamic and diverse test samples. Limitations in existing solutions include a lack of linguistic variety and the inability to adapt to the increasing complexity of LLMs. Barriers such as the challenges of ensuring the correctness of generated data and the instability of LLMs have hindered progress. Our approach differs by utilizing reasoning graphs to evolve benchmarks, allowing for fine-grained control over complexity and diversity, which has not been explored in prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, DARG (Dynamic Evaluation of LLMs via Adaptive Reasoning Graph), involves constructing reasoning graphs for data points in existing benchmarks using LLMs. We will perform fine-grained perturbations on these graphs to dynamically adjust their complexity. The reasoning graphs will then be converted back into text, ensuring linguistic diversity. To validate the quality of the generated test samples, we will employ tool-augmented LLMs. The expected outcomes include"
            }
        },
        "author_data": {
            "cd88fe31-0a1b-4625-91f9-26a4876ddfd7": {
                "pk": "cd88fe31-0a1b-4625-91f9-26a4876ddfd7",
                "name": "Zhehao Zhang",
                "collaborators": [
                    "Sagar Vijay",
                    "David Aasen",
                    "Utkarsh Agrawal",
                    "Thomas S. Richardson",
                    "Tsung-Cheng Lu",
                    "Timothy H. Hsieh",
                    "Caleb Ziems",
                    "William Held",
                    "Omar Shaikh",
                    "Jiaao Chen"
                ],
                "domain": [
                    "Quantum Computing",
                    "Machine Learning",
                    "Vision-Language Models",
                    "Computational Social Science"
                ],
                "publications": [
                    {
                        "title": "The X-Cube Floquet Code",
                        "abstract": "Inspired by the coupled-layer construction of the X-Cube model, we introduce the X-Cube Floquet code, a dynamical quantum error-correcting code where the number of encoded logical qubits grows with system size. The X-Cube Floquet code is defined on a three-dimensional lattice, built from intersecting two-dimensional layers in the $xy$, $yz$, and $xz$ directions, and consists of a periodic sequence of two-qubit measurements which couple the layers together. Within a single Floquet cycle, the codespace switches between that of the X-Cube fracton order and layers of entangled, two-dimensional toric codes. The encoded logical qubits' dynamics are analyzed, and we argue that the new code has a non-zero error threshold. We provide a new Hamiltonian realization of the X-Cube model and, more generally, explore the phase diagram related to the sequence of measurements that define the X-Cube Floquet code."
                    },
                    {
                        "title": "Quantum Communication and Mixed-State Order in Decohered Symmetry-Protected Topological States",
                        "abstract": "Certain pure-state symmetry-protected topological orders (SPT) can be used as a resource for transmitting quantum information. Here, we investigate the ability to transmit quantum information using decohered SPT states, and relate this property to the \"strange correlation functions\" which diagnose quantum many-body orders in these mixed-states. This perspective leads to the identification of a class of quantum channels -- termed symmetry-decoupling channels -- which do not necessarily preserve any weak or strong symmetries of the SPT state, but nevertheless protect quantum many-body order in the decohered mixed-state. We quantify the ability to transmit quantum information in decohered SPT states through the coherent quantum information, whose behavior is generally related to a decoding problem, whereby local measurements in the system are used to attempt to \"learn\" the symmetry charge of the SPT state before decoherence."
                    },
                    {
                        "title": "Bounds on the Distribution of a Sum of Two Random Variables: Revisiting a problem of Kolmogorov with application to Individual Treatment Effects",
                        "abstract": "We revisit the following problem, proposed by Kolmogorov: given prescribed marginal distributions $F$ and $G$ for random variables $X,Y$ respectively, characterize the set of compatible distribution functions for the sum $Z=X+Y$. Bounds on the distribution function for $Z$ were given by Markarov (1982), and Frank et al. (1987), the latter using copula theory. However, though they obtain the same bounds, they make different assertions concerning their sharpness. In addition, their solutions leave some open problems in the case when the given marginal distribution functions are discontinuous. These issues have led to some confusion and erroneous statements in subsequent literature, which we correct.   Kolmogorov's problem is closely related to inferring possible distributions for individual treatment effects $Y_1 - Y_0$ given the marginal distributions of $Y_1$ and $Y_0$; the latter being identified from a randomized experiment. We use our new insights to sharpen and correct results due to Fan and Park (2010) concerning individual treatment effects, and to fill some other logical gaps."
                    },
                    {
                        "title": "Mixed-state long-range order and criticality from measurement and feedback",
                        "abstract": "We propose a general framework for using local measurements, local unitaries, and non-local classical communication to construct quantum channels which can efficiently prepare mixed states with long-range quantum order or quantum criticality. As an illustration, symmetry-protected topological (SPT) phases can be universally converted into mixed-states with long-range entanglement, which can undergo phase transitions with quantum critical correlations of local operators and a logarithmic scaling of the entanglement negativity, despite coexisting with volume-law entropy. Within the same framework, we present two applications using fermion occupation number measurement to convert (i) spinful free fermions in one dimension into a quantum-critical mixed state with enhanced algebraic correlations between spins and (ii) Chern insulators into a mixed state with critical quantum correlations in the bulk. The latter is an example where mixed-state quantum criticality can emerge from a gapped state of matter in constant depth using local quantum operations and non-local classical communication."
                    },
                    {
                        "title": "Can Large Language Models Transform Computational Social Science?",
                        "abstract": "Large Language Models (LLMs) are capable of successfully performing many language processing tasks zero-shot (without training data). If zero-shot LLMs can also reliably classify and explain social phenomena like persuasiveness and political ideology, then LLMs could augment the Computational Social Science (CSS) pipeline in important ways. This work provides a road map for using LLMs as CSS tools. Towards this end, we contribute a set of prompting best practices and an extensive evaluation pipeline to measure the zero-shot performance of 13 language models on 25 representative English CSS benchmarks. On taxonomic labeling tasks (classification), LLMs fail to outperform the best fine-tuned models but still achieve fair levels of agreement with humans. On free-form coding tasks (generation), LLMs produce explanations that often exceed the quality of crowdworkers' gold references. We conclude that the performance of today's LLMs can augment the CSS research pipeline in two ways: (1) serving as zero-shot data annotators on human annotation teams, and (2) bootstrapping challenging creative generation tasks (e.g., explaining the underlying attributes of a text). In summary, LLMs are posed to meaningfully participate in social science analysis in partnership with humans."
                    },
                    {
                        "title": "VipAct: Visual-Perception Enhancement via Specialized VLM Agent Collaboration and Tool-use",
                        "abstract": "While vision-language models (VLMs) have demonstrated remarkable performance across various tasks combining textual and visual information, they continue to struggle with fine-grained visual perception tasks that require detailed pixel-level analysis. Effectively eliciting comprehensive reasoning from VLMs on such intricate visual elements remains an open challenge. In this paper, we present VipAct, an agent framework that enhances VLMs by integrating multi-agent collaboration and vision expert models, enabling more precise visual understanding and comprehensive reasoning. VipAct consists of an orchestrator agent, which manages task requirement analysis, planning, and coordination, along with specialized agents that handle specific tasks such as image captioning and vision expert models that provide high-precision perceptual information. This multi-agent approach allows VLMs to better perform fine-grained visual perception tasks by synergizing planning, reasoning, and tool use. We evaluate VipAct on benchmarks featuring a diverse set of visual perception tasks, with experimental results demonstrating significant performance improvements over state-of-the-art baselines across all tasks. Furthermore, comprehensive ablation studies reveal the critical role of multi-agent collaboration in eliciting more detailed System-2 reasoning and highlight the importance of image input for task planning. Additionally, our error analysis identifies patterns of VLMs' inherent limitations in visual perception, providing insights into potential future improvements. VipAct offers a flexible and extensible framework, paving the way for more advanced visual perception systems across various real-world applications."
                    },
                    {
                        "title": "A quality assurance framework for real-time monitoring of deep learning segmentation models in radiotherapy",
                        "abstract": "To safely deploy deep learning models in the clinic, a quality assurance framework is needed for routine or continuous monitoring of input-domain shift and the models' performance without ground truth contours. In this work, cardiac substructure segmentation was used as an example task to establish a QA framework. A benchmark dataset consisting of Computed Tomography (CT) images along with manual cardiac delineations of 241 patients were collected, including one 'common' image domain and five 'uncommon' domains. Segmentation models were tested on the benchmark dataset for an initial evaluation of model capacity and limitations. An image domain shift detector was developed by utilizing a trained Denoising autoencoder (DAE) and two hand-engineered features. Another Variational Autoencoder (VAE) was also trained to estimate the shape quality of the auto-segmentation results. Using the extracted features from the image/segmentation pair as inputs, a regression model was trained to predict the per-patient segmentation accuracy, measured by Dice coefficient similarity (DSC). The framework was tested across 19 segmentation models to evaluate the generalizability of the entire framework.   As results, the predicted DSC of regression models achieved a mean absolute error (MAE) ranging from 0.036 to 0.046 with an averaged MAE of 0.041. When tested on the benchmark dataset, the performances of all segmentation models were not significantly affected by scanning parameters: FOV, slice thickness and reconstructions kernels. For input images with Poisson noise, CNN-based segmentation models demonstrated a decreased DSC ranging from 0.07 to 0.41, while the transformer-based model was not significantly affected."
                    }
                ]
            },
            "e6632beb-de33-4b74-9d37-0cc3e8d6a4c4": {
                "pk": "e6632beb-de33-4b74-9d37-0cc3e8d6a4c4",
                "name": "Jiaao Chen",
                "collaborators": [
                    "Diyi Yang",
                    "Jianshu Chen",
                    "Zichao Yang",
                    "Yanzhe Zhang",
                    "Zhou Yu",
                    "Dinghan Shen",
                    "Weizhu Chen",
                    "Yuwei Wu",
                    "Mohan Dodda",
                    "Zhenghui Wang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Text Summarization",
                    "Data Augmentation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Multi-View Sequence-to-Sequence Models with Conversational Structure for Abstractive Dialogue Summarization",
                        "abstract": "Text summarization is one of the most challenging and interesting problems in NLP. Although much attention has been paid to summarizing structured text like news reports or encyclopedia articles, summarizing conversations---an essential part of human-human/machine interaction where most important pieces of information are scattered across various utterances of different speakers---remains relatively under-investigated. This work proposes a multi-view sequence-to-sequence model by first extracting conversational structures of unstructured daily chats from different views to represent conversations and then utilizing a multi-view decoder to incorporate different views to generate dialogue summaries. Experiments on a large-scale dialogue summarization corpus demonstrated that our methods significantly outperformed previous state-of-the-art models via both automatic evaluations and human judgment. We also discussed specific challenges that current approaches faced with this task. We have publicly released our code at https://github.com/GT-SALT/Multi-View-Seq2Seq."
                    },
                    {
                        "title": "Weakly-Supervised Hierarchical Models for Predicting Persuasive Strategies in Good-faith Textual Requests",
                        "abstract": "Modeling persuasive language has the potential to better facilitate our decision-making processes. Despite its importance, computational modeling of persuasion is still in its infancy, largely due to the lack of benchmark datasets that can provide quantitative labels of persuasive strategies to expedite this line of research. To this end, we introduce a large-scale multi-domain text corpus for modeling persuasive strategies in good-faith text requests. Moreover, we design a hierarchical weakly-supervised latent variable model that can leverage partially labeled data to predict such associated persuasive strategies for each sentence, where the supervision comes from both the overall document-level labels and very limited sentence-level labels. Experimental results showed that our proposed method outperformed existing semi-supervised baselines significantly. We have publicly released our code at https://github.com/GT-SALT/Persuasion_Strategy_WVAE."
                    },
                    {
                        "title": "Structure-Aware Abstractive Conversation Summarization via Discourse and Action Graphs",
                        "abstract": "Abstractive conversation summarization has received much attention recently. However, these generated summaries often suffer from insufficient, redundant, or incorrect content, largely due to the unstructured and complex characteristics of human-human interactions. To this end, we propose to explicitly model the rich structures in conversations for more precise and accurate conversation summarization, by first incorporating discourse relations between utterances and action triples (\"who-doing-what\") in utterances through structured graphs to better encode conversations, and then designing a multi-granularity decoder to generate summaries by combining all levels of information. Experiments show that our proposed models outperform state-of-the-art methods and generalize well in other domains in terms of both automatic evaluations and human judgments. We have publicly released our code at https://github.com/GT-SALT/Structure-Aware-BART."
                    },
                    {
                        "title": "Unlearn What You Want to Forget: Efficient Unlearning for LLMs",
                        "abstract": "Large language models (LLMs) have achieved significant progress from pre-training on and memorizing a wide range of textual data, however, this process might suffer from privacy issues and violations of data protection regulations. As a result, the ability to easily remove data related to individual users from such models while not deteriorating their predictive quality after the removal becomes increasingly important. To address these issues, in this work, we propose an efficient unlearning framework that could efficiently update LLMs without having to retrain the whole model after data removals, by introducing lightweight unlearning layers learned with a selective teacher-student objective into the transformers. In addition, we introduce a fusion mechanism to effectively combine different unlearning layers that learns to forget different sets of data to handle a sequence of forgetting operations. Experiments on classification and generation tasks demonstrate the effectiveness of our proposed methods compared to the state-of-the-art baselines."
                    },
                    {
                        "title": "Incorporating Structured Commonsense Knowledge in Story Completion",
                        "abstract": "The ability to select an appropriate story ending is the first step towards perfect narrative comprehension. Story ending prediction requires not only the explicit clues within the context, but also the implicit knowledge (such as commonsense) to construct a reasonable and consistent story. However, most previous approaches do not explicitly use background commonsense knowledge. We present a neural story ending selection model that integrates three types of information: narrative sequence, sentiment evolution and commonsense knowledge. Experiments show that our model outperforms state-of-the-art approaches on a public dataset, ROCStory Cloze Task , and the performance gain from adding the additional commonsense knowledge is significant."
                    },
                    {
                        "title": "HiddenCut: Simple Data Augmentation for Natural Language Understanding with Better Generalization",
                        "abstract": "Fine-tuning large pre-trained models with task-specific data has achieved great success in NLP. However, it has been demonstrated that the majority of information within the self-attention networks is redundant and not utilized effectively during the fine-tuning stage. This leads to inferior results when generalizing the obtained models to out-of-domain distributions. To this end, we propose a simple yet effective data augmentation technique, HiddenCut, to better regularize the model and encourage it to learn more generalizable features. Specifically, contiguous spans within the hidden space are dynamically and strategically dropped during training. Experiments show that our HiddenCut method outperforms the state-of-the-art augmentation methods on the GLUE benchmark, and consistently exhibits superior generalization performances on out-of-distribution and challenging counterexamples. We have publicly released our code at https://github.com/GT-SALT/HiddenCut."
                    },
                    {
                        "title": "Semi-Supervised Models via Data Augmentationfor Classifying Interactive Affective Responses",
                        "abstract": "We present semi-supervised models with data augmentation (SMDA), a semi-supervised text classification system to classify interactive affective responses. SMDA utilizes recent transformer-based models to encode each sentence and employs back translation techniques to paraphrase given sentences as augmented data. For labeled sentences, we performed data augmentations to uniform the label distributions and computed supervised loss during training process. For unlabeled sentences, we explored self-training by regarding low-entropy predictions over unlabeled sentences as pseudo labels, assuming high-confidence predictions as labeled data for training. We further introduced consistency regularization as unsupervised loss after data augmentations on unlabeled data, based on the assumption that the model should predict similar class distributions with original unlabeled sentences as input and augmented sentences as input. Via a set of experiments, we demonstrated that our system outperformed baseline models in terms of F1-score and accuracy."
                    },
                    {
                        "title": "MixText: Linguistically-Informed Interpolation of Hidden Space for Semi-Supervised Text Classification",
                        "abstract": "This paper presents MixText, a semi-supervised learning method for text classification, which uses our newly designed data augmentation method called TMix. TMix creates a large amount of augmented training samples by interpolating text in hidden space. Moreover, we leverage recent advances in data augmentation to guess low-entropy labels for unlabeled data, hence making them as easy to use as labeled data.By mixing labeled, unlabeled and augmented data, MixText significantly outperformed current pre-trained and fined-tuned models and other state-of-the-art semi-supervised learning methods on several text classification benchmarks. The improvement is especially prominent when supervision is extremely limited. We have publicly released our code at https://github.com/GT-SALT/MixText."
                    },
                    {
                        "title": "Human-in-the-loop Abstractive Dialogue Summarization",
                        "abstract": "Abstractive dialogue summarization has received increasing attention recently. Despite the fact that most of the current dialogue summarization systems are trained to maximize the likelihood of human-written summaries and have achieved significant results, there is still a huge gap in generating high-quality summaries as determined by humans, such as coherence and faithfulness, partly due to the misalignment in maximizing a single human-written summary. To this end, we propose to incorporate different levels of human feedback into the training process. This will enable us to guide the models to capture the behaviors humans care about for summaries. Specifically, we ask humans to highlight the salient information to be included in summaries to provide the local feedback , and to make overall comparisons among summaries in terms of coherence, accuracy, coverage, concise and overall quality, as the global feedback. We then combine both local and global feedback to fine-tune the dialog summarization policy with Reinforcement Learning. Experiments conducted on multiple datasets demonstrate the effectiveness and generalization of our methods over the state-of-the-art supervised baselines, especially in terms of human judgments."
                    },
                    {
                        "title": "Local Additivity Based Data Augmentation for Semi-supervised NER",
                        "abstract": "Named Entity Recognition (NER) is one of the first stages in deep language understanding yet current NER models heavily rely on human-annotated data. In this work, to alleviate the dependence on labeled data, we propose a Local Additivity based Data Augmentation (LADA) method for semi-supervised NER, in which we create virtual samples by interpolating sequences close to each other. Our approach has two variations: Intra-LADA and Inter-LADA, where Intra-LADA performs interpolations among tokens within one sentence, and Inter-LADA samples different sentences to interpolate. Through linear additions between sampled training data, LADA creates an infinite amount of labeled data and improves both entity and context learning. We further extend LADA to the semi-supervised setting by designing a novel consistency loss for unlabeled data. Experiments conducted on two NER benchmarks demonstrate the effectiveness of our methods over several strong baselines. We have publicly released our code at https://github.com/GT-SALT/LADA."
                    },
                    {
                        "title": "VALUE: Understanding Dialect Disparity in NLU",
                        "abstract": "English Natural Language Understanding (NLU) systems have achieved great performances and even outperformed humans on benchmarks like GLUE and SuperGLUE. However, these benchmarks contain only textbook Standard American English (SAE). Other dialects have been largely overlooked in the NLP community. This leads to biased and inequitable NLU systems that serve only a sub-population of speakers. To understand disparities in current models and to facilitate more dialect-competent NLU systems, we introduce the VernAcular Language Understanding Evaluation (VALUE) benchmark, a challenging variant of GLUE that we created with a set of lexical and morphosyntactic transformation rules. In this initial release (V.1), we construct rules for 11 features of African American Vernacular English (AAVE), and we recruit fluent AAVE speakers to validate each feature transformation via linguistic acceptability judgments in a participatory design manner. Experiments show that these new dialectal features can lead to a drop in model performance. To run the transformation code and download both synthetic and gold-standard dialectal GLUE benchmarks, see https://github.com/SALT-NLP/value"
                    },
                    {
                        "title": "A Cheaper and Better Diffusion Language Model with Soft-Masked Noise",
                        "abstract": "Diffusion models that are based on iterative denoising have been recently proposed and leveraged in various generation tasks like image generation. Whereas, as a way inherently built for continuous data, existing diffusion models still have some limitations in modeling discrete data, e.g., languages. For example, the generally used Gaussian noise can not handle the discrete corruption well, and the objectives in continuous spaces fail to be stable for textual data in the diffusion process especially when the dimension is high. To alleviate these issues, we introduce a novel diffusion model for language modeling, Masked-Diffuse LM, with lower training cost and better performances, inspired by linguistic features in languages. Specifically, we design a linguistic-informed forward process which adds corruptions to the text through strategically soft-masking to better noise the textual data. Also, we directly predict the categorical distribution with cross-entropy loss function in every diffusion step to connect the continuous space and discrete space in a more efficient and straightforward way. Through experiments on 5 controlled generation tasks, we demonstrate that our Masked-Diffuse LM can achieve better generation quality than the state-of-the-art diffusion models with better efficiency."
                    },
                    {
                        "title": "Skills-in-Context Prompting: Unlocking Compositionality in Large Language Models",
                        "abstract": "We investigate how to elicit compositional generalization capabilities in large language models (LLMs). Compositional generalization empowers LLMs to solve complex problems by combining foundational skills, a critical reasoning ability akin to human intelligence. However, even the most advanced LLMs currently struggle with this form of reasoning. We examine this problem within the framework of in-context learning and find that demonstrating both foundational skills and compositional examples grounded in these skills within the same prompt context is crucial. We refer to this prompt structure as skills-in-context (SKiC). With as few as two exemplars, this in-context learning structure enables LLMs to tackle more challenging problems requiring innovative skill combinations, achieving near-perfect systematic generalization across a broad range of tasks. Intriguingly, SKiC also unlocks the latent potential of LLMs, allowing them to more actively utilize pre-existing internal skills acquired during earlier pretraining stages to solve complex reasoning problems. The SKiC structure is robust across different skill constructions and exemplar choices and demonstrates strong transferability to new tasks. Finally, inspired by our in-context learning study, we show that fine-tuning LLMs with SKiC-style data can elicit zero-shot weak-to-strong generalization, enabling the models to solve much harder problems directly with standard prompting."
                    },
                    {
                        "title": "An Empirical Survey of Data Augmentation for Limited Data Learning in NLP",
                        "abstract": "NLP has achieved great progress in the past decade through the use of neural models and large labeled datasets. The dependence on abundant data prevents NLP models from being applied to low-resource settings or novel tasks where significant time, money, or expertise is required to label massive amounts of textual data. Recently, data augmentation methods have been explored as a means of improving data efficiency in NLP. To date, there has been no systematic empirical overview of data augmentation for NLP in the limited labeled data setting, making it difficult to understand which methods work in which settings. In this paper, we provide an empirical survey of recent progress on data augmentation for NLP in the limited labeled data setting, summarizing the landscape of methods (including token-level augmentations, sentence-level augmentations, adversarial augmentations, and hidden-space augmentations) and carrying out experiments on 11 datasets covering topics/news classification, inference tasks, paraphrasing tasks, and single-sentence tasks. Based on the results, we draw several conclusions to help practitioners choose appropriate augmentations in different settings and discuss the current challenges and future directions for limited data learning in NLP."
                    },
                    {
                        "title": "Examining the Ordering of Rhetorical Strategies in Persuasive Requests",
                        "abstract": "Interpreting how persuasive language influences audiences has implications across many domains like advertising, argumentation, and propaganda. Persuasion relies on more than a message's content. Arranging the order of the message itself (i.e., ordering specific rhetorical strategies) also plays an important role. To examine how strategy orderings contribute to persuasiveness, we first utilize a Variational Autoencoder model to disentangle content and rhetorical strategies in textual requests from a large-scale loan request corpus. We then visualize interplay between content and strategy through an attentional LSTM that predicts the success of textual requests. We find that specific (orderings of) strategies interact uniquely with a request's content to impact success rate, and thus the persuasiveness of a request."
                    },
                    {
                        "title": "Continual Learning for Text Classification with Information Disentanglement Based Regularization",
                        "abstract": "Continual learning has become increasingly important as it enables NLP models to constantly learn and gain knowledge over time. Previous continual learning methods are mainly designed to preserve knowledge from previous tasks, without much emphasis on how to well generalize models to new tasks. In this work, we propose an information disentanglement based regularization method for continual learning on text classification. Our proposed method first disentangles text hidden spaces into representations that are generic to all tasks and representations specific to each individual task, and further regularizes these representations differently to better constrain the knowledge required to generalize. We also introduce two simple auxiliary tasks: next sentence prediction and task-id prediction, for learning better generic and specific representation spaces. Experiments conducted on large-scale benchmarks demonstrate the effectiveness of our method in continual text classification tasks with various sequences and lengths over state-of-the-art baselines. We have publicly released our code at https://github.com/GT-SALT/IDBR."
                    },
                    {
                        "title": "Leveraging Expert Guided Adversarial Augmentation For Improving Generalization in Named Entity Recognition",
                        "abstract": "Named Entity Recognition (NER) systems often demonstrate great performance on in-distribution data, but perform poorly on examples drawn from a shifted distribution. One way to evaluate the generalization ability of NER models is to use adversarial examples, on which the specific variations associated with named entities are rarely considered. To this end, we propose leveraging expert-guided heuristics to change the entity tokens and their surrounding contexts thereby altering their entity types as adversarial attacks. Using expert-guided heuristics, we augmented the CoNLL 2003 test set and manually annotated it to construct a high-quality challenging set. We found that state-of-the-art NER systems trained on CoNLL 2003 training data drop performance dramatically on our challenging set. By training on adversarial augmented training examples and using mixup for regularization, we were able to significantly improve the performance on the challenging set as well as improve out-of-domain generalization which we evaluated by using OntoNotes data. We have publicly released our dataset and code at https://github.com/GT-SALT/Guided-Adversarial-Augmentation."
                    }
                ]
            },
            "f361a2ee-98ca-46ac-a109-1102ebfa90da": {
                "pk": "f361a2ee-98ca-46ac-a109-1102ebfa90da",
                "name": "Diyi Yang",
                "collaborators": [
                    "Jiaao Chen",
                    "Caleb Ziems",
                    "Xuezhi Wang",
                    "Le Zhang",
                    "Zichao Yang",
                    "Kunal Chawla",
                    "William Held",
                    "Jingfeng Yang",
                    "Dirk Hovy",
                    "David Jurgens"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Text Summarization",
                    "Style Transfer",
                    "Robustness"
                ],
                "publications": [
                    {
                        "title": "Multi-View Sequence-to-Sequence Models with Conversational Structure for Abstractive Dialogue Summarization",
                        "abstract": "Text summarization is one of the most challenging and interesting problems in NLP. Although much attention has been paid to summarizing structured text like news reports or encyclopedia articles, summarizing conversations---an essential part of human-human/machine interaction where most important pieces of information are scattered across various utterances of different speakers---remains relatively under-investigated. This work proposes a multi-view sequence-to-sequence model by first extracting conversational structures of unstructured daily chats from different views to represent conversations and then utilizing a multi-view decoder to incorporate different views to generate dialogue summaries. Experiments on a large-scale dialogue summarization corpus demonstrated that our methods significantly outperformed previous state-of-the-art models via both automatic evaluations and human judgment. We also discussed specific challenges that current approaches faced with this task. We have publicly released our code at https://github.com/GT-SALT/Multi-View-Seq2Seq."
                    },
                    {
                        "title": "Semi-supervised Formality Style Transfer using Language Model Discriminator and Mutual Information Maximization",
                        "abstract": "Formality style transfer is the task of converting informal sentences to grammatically-correct formal sentences, which can be used to improve performance of many downstream NLP tasks. In this work, we propose a semi-supervised formality style transfer model that utilizes a language model-based discriminator to maximize the likelihood of the output sentence being formal, which allows us to use maximization of token-level conditional probabilities for training. We further propose to maximize mutual information between source and target styles as our training objective instead of maximizing the regular likelihood that often leads to repetitive and trivial generated responses. Experiments showed that our model outperformed previous state-of-the-art baselines significantly in terms of both automated metrics and human judgement. We further generalized our model to unsupervised text style transfer task, and achieved significant improvements on two benchmark sentiment style transfer datasets."
                    },
                    {
                        "title": "Weakly-Supervised Hierarchical Models for Predicting Persuasive Strategies in Good-faith Textual Requests",
                        "abstract": "Modeling persuasive language has the potential to better facilitate our decision-making processes. Despite its importance, computational modeling of persuasion is still in its infancy, largely due to the lack of benchmark datasets that can provide quantitative labels of persuasive strategies to expedite this line of research. To this end, we introduce a large-scale multi-domain text corpus for modeling persuasive strategies in good-faith text requests. Moreover, we design a hierarchical weakly-supervised latent variable model that can leverage partially labeled data to predict such associated persuasive strategies for each sentence, where the supervision comes from both the overall document-level labels and very limited sentence-level labels. Experimental results showed that our proposed method outperformed existing semi-supervised baselines significantly. We have publicly released our code at https://github.com/GT-SALT/Persuasion_Strategy_WVAE."
                    },
                    {
                        "title": "Structure-Aware Abstractive Conversation Summarization via Discourse and Action Graphs",
                        "abstract": "Abstractive conversation summarization has received much attention recently. However, these generated summaries often suffer from insufficient, redundant, or incorrect content, largely due to the unstructured and complex characteristics of human-human interactions. To this end, we propose to explicitly model the rich structures in conversations for more precise and accurate conversation summarization, by first incorporating discourse relations between utterances and action triples (\"who-doing-what\") in utterances through structured graphs to better encode conversations, and then designing a multi-granularity decoder to generate summaries by combining all levels of information. Experiments show that our proposed models outperform state-of-the-art methods and generalize well in other domains in terms of both automatic evaluations and human judgments. We have publicly released our code at https://github.com/GT-SALT/Structure-Aware-BART."
                    },
                    {
                        "title": "To Protect and To Serve? Analyzing Entity-Centric Framing of Police Violence",
                        "abstract": "Framing has significant but subtle effects on public opinion and policy. We propose an NLP framework to measure entity-centric frames. We use it to understand media coverage on police violence in the United States in a new Police Violence Frames Corpus of 82k news articles spanning 7k police killings. Our work uncovers more than a dozen framing devices and reveals significant differences in the way liberal and conservative news sources frame both the issue of police violence and the entities involved. Conservative sources emphasize when the victim is armed or attacking an officer and are more likely to mention the victim's criminal record. Liberal sources focus more on the underlying systemic injustice, highlighting the victim's race and that they were unarmed. We discover temporary spikes in these injustice frames near high-profile shooting events, and finally, we show protest volume correlates with and precedes media framing decisions."
                    },
                    {
                        "title": "Shapley Head Pruning: Identifying and Removing Interference in Multilingual Transformers",
                        "abstract": "Multilingual transformer-based models demonstrate remarkable zero and few-shot transfer across languages by learning and reusing language-agnostic features. However, as a fixed-size model acquires more languages, its performance across all languages degrades, a phenomenon termed interference. Often attributed to limited model capacity, interference is commonly addressed by adding additional parameters despite evidence that transformer-based models are overparameterized. In this work, we show that it is possible to reduce interference by instead identifying and pruning language-specific parameters. First, we use Shapley Values, a credit allocation metric from coalitional game theory, to identify attention heads that introduce interference. Then, we show that removing identified attention heads from a fixed model improves performance for a target language on both sentence classification and structural prediction, seeing gains as large as 24.7\\%. Finally, we provide insights on language-agnostic and language-specific attention heads using attention visualization."
                    },
                    {
                        "title": "Unlearn What You Want to Forget: Efficient Unlearning for LLMs",
                        "abstract": "Large language models (LLMs) have achieved significant progress from pre-training on and memorizing a wide range of textual data, however, this process might suffer from privacy issues and violations of data protection regulations. As a result, the ability to easily remove data related to individual users from such models while not deteriorating their predictive quality after the removal becomes increasingly important. To address these issues, in this work, we propose an efficient unlearning framework that could efficiently update LLMs without having to retrain the whole model after data removals, by introducing lightweight unlearning layers learned with a selective teacher-student objective into the transformers. In addition, we introduce a fusion mechanism to effectively combine different unlearning layers that learns to forget different sets of data to handle a sequence of forgetting operations. Experiments on classification and generation tasks demonstrate the effectiveness of our proposed methods compared to the state-of-the-art baselines."
                    },
                    {
                        "title": "SUBS: Subtree Substitution for Compositional Semantic Parsing",
                        "abstract": "Although sequence-to-sequence models often achieve good performance in semantic parsing for i.i.d. data, their performance is still inferior in compositional generalization. Several data augmentation methods have been proposed to alleviate this problem. However, prior work only leveraged superficial grammar or rules for data augmentation, which resulted in limited improvement. We propose to use subtree substitution for compositional data augmentation, where we consider subtrees with similar semantic functions as exchangeable. Our experiments showed that such augmented data led to significantly better performance on SCAN and GeoQuery, and reached new SOTA on compositional split of GeoQuery."
                    },
                    {
                        "title": "MixText: Linguistically-Informed Interpolation of Hidden Space for Semi-Supervised Text Classification",
                        "abstract": "This paper presents MixText, a semi-supervised learning method for text classification, which uses our newly designed data augmentation method called TMix. TMix creates a large amount of augmented training samples by interpolating text in hidden space. Moreover, we leverage recent advances in data augmentation to guess low-entropy labels for unlabeled data, hence making them as easy to use as labeled data.By mixing labeled, unlabeled and augmented data, MixText significantly outperformed current pre-trained and fined-tuned models and other state-of-the-art semi-supervised learning methods on several text classification benchmarks. The improvement is especially prominent when supervision is extremely limited. We have publicly released our code at https://github.com/GT-SALT/MixText."
                    },
                    {
                        "title": "TreeMix: Compositional Constituency-based Data Augmentation for Natural Language Understanding",
                        "abstract": "Data augmentation is an effective approach to tackle over-fitting. Many previous works have proposed different data augmentations strategies for NLP, such as noise injection, word replacement, back-translation etc. Though effective, they missed one important characteristic of language--compositionality, meaning of a complex expression is built from its sub-parts. Motivated by this, we propose a compositional data augmentation approach for natural language understanding called TreeMix. Specifically, TreeMix leverages constituency parsing tree to decompose sentences into constituent sub-structures and the Mixup data augmentation technique to recombine them to generate new sentences. Compared with previous approaches, TreeMix introduces greater diversity to the samples generated and encourages models to learn compositionality of NLP data. Extensive experiments on text classification and SCAN demonstrate that TreeMix outperforms current state-of-the-art data augmentation methods."
                    },
                    {
                        "title": "The Call for Socially Aware Language Technologies",
                        "abstract": "Language technologies have made enormous progress, especially with the introduction of large language models (LLMs). On traditional tasks such as machine translation and sentiment analysis, these models perform at near-human level. These advances can, however, exacerbate a variety of issues that models have traditionally struggled with, such as bias, evaluation, and risks. In this position paper, we argue that many of these issues share a common core: a lack of awareness of the factors, context, and implications of the social environment in which NLP operates, which we call social awareness. While NLP is getting better at solving the formal linguistic aspects, limited progress has been made in adding the social awareness required for language applications to work in all situations for all users. Integrating social awareness into NLP models will make applications more natural, helpful, and safe, and will open up new possibilities. Thus we argue that substantial challenges remain for NLP to develop social awareness and that we are just at the beginning of a new era for the field."
                    },
                    {
                        "title": "TADA: Task-Agnostic Dialect Adapters for English",
                        "abstract": "Large Language Models, the dominant starting point for Natural Language Processing (NLP) applications, fail at a higher rate for speakers of English dialects other than Standard American English (SAE). Prior work addresses this using task-specific data or synthetic data augmentation, both of which require intervention for each dialect and task pair. This poses a scalability issue that prevents the broad adoption of robust dialectal English NLP. We introduce a simple yet effective method for task-agnostic dialect adaptation by aligning non-SAE dialects using adapters and composing them with task-specific adapters from SAE. Task-Agnostic Dialect Adapters (TADA) improve dialectal robustness on 4 dialectal variants of the GLUE benchmark without task-specific supervision."
                    },
                    {
                        "title": "Identifying and Mitigating Spurious Correlations for Improving Robustness in NLP Models",
                        "abstract": "Recently, NLP models have achieved remarkable progress across a variety of tasks; however, they have also been criticized for being not robust. Many robustness problems can be attributed to models exploiting spurious correlations, or shortcuts between the training data and the task labels. Most existing work identifies a limited set of task-specific shortcuts via human priors or error analyses, which requires extensive expertise and efforts. In this paper, we aim to automatically identify such spurious correlations in NLP models at scale. We first leverage existing interpretability methods to extract tokens that significantly affect model's decision process from the input text. We then distinguish \"genuine\" tokens and \"spurious\" tokens by analyzing model predictions across multiple corpora and further verify them through knowledge-aware perturbations. We show that our proposed method can effectively and efficiently identify a scalable set of \"shortcuts\", and mitigating these leads to more robust models in multiple applications."
                    },
                    {
                        "title": "Inducing Positive Perspectives with Text Reframing",
                        "abstract": "Sentiment transfer is one popular example of a text style transfer task, where the goal is to reverse the sentiment polarity of a text. With a sentiment reversal comes also a reversal in meaning. We introduce a different but related task called positive reframing in which we neutralize a negative point of view and generate a more positive perspective for the author without contradicting the original meaning. Our insistence on meaning preservation makes positive reframing a challenging and semantically rich task. To facilitate rapid progress, we introduce a large-scale benchmark, Positive Psychology Frames, with 8,349 sentence pairs and 12,755 structured annotations to explain positive reframing in terms of six theoretically-motivated reframing strategies. Then we evaluate a set of state-of-the-art text style transfer models, and conclude by discussing key challenges and directions for future work."
                    },
                    {
                        "title": "Six Feet Apart: Online Payments During the COVID-19 Pandemic",
                        "abstract": "Since the COVID-19 pandemic, businesses have faced unprecedented challenges when trying to remain open. Because COVID-19 spreads through aerosolized droplets, businesses were forced to distance their services; in some cases, distancing may have involved moving business services online. In this work, we explore digitization strategies used by small businesses that remained open during the pandemic, and survey/interview small businesses owners to understand preliminary challenges associated with moving online. Furthermore, we analyze payments from 400K businesses across Japan, Australia, United States, Great Britain, and Canada. Following initial government interventions, we observe (at minimum for each country) a 47% increase in digitizing businesses compared to pre-pandemic levels, with about 80% of surveyed businesses digitizing in under a week. From both our quantitative models and our surveys/interviews, we find that businesses rapidly digitized at the start of the pandemic in preparation of future uncertainty. We also conduct a case-study of initial digitization in the United States, examining finer relationships between specific government interventions, business sectors, political orientation, and resulting digitization shifts. Finally, we discuss the implications of rapid & widespread digitization for small businesses in the context of usability challenges and interpersonal interactions, while highlighting potential shifts in pre-existing social norms."
                    },
                    {
                        "title": "Semi-Supervised Models via Data Augmentationfor Classifying Interactive Affective Responses",
                        "abstract": "We present semi-supervised models with data augmentation (SMDA), a semi-supervised text classification system to classify interactive affective responses. SMDA utilizes recent transformer-based models to encode each sentence and employs back translation techniques to paraphrase given sentences as augmented data. For labeled sentences, we performed data augmentations to uniform the label distributions and computed supervised loss during training process. For unlabeled sentences, we explored self-training by regarding low-entropy predictions over unlabeled sentences as pseudo labels, assuming high-confidence predictions as labeled data for training. We further introduced consistency regularization as unsupervised loss after data augmentations on unlabeled data, based on the assumption that the model should predict similar class distributions with original unlabeled sentences as input and augmented sentences as input. Via a set of experiments, we demonstrated that our system outperformed baseline models in terms of F1-score and accuracy."
                    },
                    {
                        "title": "Tuiteamos o pongamos un tuit? Investigating the Social Constraints of Loanword Integration in Spanish Social Media",
                        "abstract": "Speakers of non-English languages often adopt loanwords from English to express new or unusual concepts. While these loanwords may be borrowed unchanged, speakers may also integrate the words to fit the constraints of their native language, e.g. creating Spanish \"tuitear\" from English \"tweet.\" Linguists have often considered the process of loanword integration to be more dependent on language-internal constraints, but sociolinguistic constraints such as speaker background remain only qualitatively understood. We investigate the role of social context and speaker background in Spanish speakers' use of integrated loanwords on social media. We find first that newspaper authors use the integrated forms of loanwords and native words more often than social media authors, showing that integration is associated with formal domains. In social media, we find that speaker background and expectations of formality explain loanword and native word integration, such that authors who use more Spanish and who write to a wider audience tend to use integrated verb forms more often. This study shows that loanword integration reflects not only language-internal constraints but also social expectations that vary by conversation and speaker."
                    },
                    {
                        "title": "HiddenCut: Simple Data Augmentation for Natural Language Understanding with Better Generalization",
                        "abstract": "Fine-tuning large pre-trained models with task-specific data has achieved great success in NLP. However, it has been demonstrated that the majority of information within the self-attention networks is redundant and not utilized effectively during the fine-tuning stage. This leads to inferior results when generalizing the obtained models to out-of-domain distributions. To this end, we propose a simple yet effective data augmentation technique, HiddenCut, to better regularize the model and encourage it to learn more generalizable features. Specifically, contiguous spans within the hidden space are dynamically and strategically dropped during training. Experiments show that our HiddenCut method outperforms the state-of-the-art augmentation methods on the GLUE benchmark, and consistently exhibits superior generalization performances on out-of-distribution and challenging counterexamples. We have publicly released our code at https://github.com/GT-SALT/HiddenCut."
                    },
                    {
                        "title": "Measure and Improve Robustness in NLP Models: A Survey",
                        "abstract": "As NLP models achieved state-of-the-art performances over benchmarks and gained wide applications, it has been increasingly important to ensure the safe deployment of these models in the real world, e.g., making sure the models are robust against unseen or challenging scenarios. Despite robustness being an increasingly studied topic, it has been separately explored in applications like vision and NLP, with various definitions, evaluation and mitigation strategies in multiple lines of research. In this paper, we aim to provide a unifying survey of how to define, measure and improve robustness in NLP. We first connect multiple definitions of robustness, then unify various lines of work on identifying robustness failures and evaluating models' robustness. Correspondingly, we present mitigation strategies that are data-driven, model-driven, and inductive-prior-based, with a more systematic view of how to effectively improve robustness in NLP models. Finally, we conclude by outlining open challenges and future directions to motivate further research in this area."
                    },
                    {
                        "title": "Continual Sequence Generation with Adaptive Compositional Modules",
                        "abstract": "Continual learning is essential for real-world deployment when there is a need to quickly adapt the model to new tasks without forgetting knowledge of old tasks. Existing work on continual sequence generation either always reuses existing parameters to learn new tasks, which is vulnerable to catastrophic forgetting on dissimilar tasks, or blindly adds new parameters for every new task, which could prevent knowledge sharing between similar tasks. To get the best of both worlds, in this work, we propose continual sequence generation with adaptive compositional modules to adaptively add modules in transformer architectures and compose both old and new modules for new tasks. We also incorporate pseudo experience replay to facilitate knowledge transfer in those shared modules. Experiment results on various sequences of generation tasks show that our framework can adaptively add modules or reuse modules based on task similarity, outperforming state-of-the-art baselines in terms of both performance and parameter efficiency. We make our code public at https://github.com/GT-SALT/Adaptive-Compositional-Modules."
                    }
                ]
            }
        }
    },
    "2407.19448": {
        "paper_data": {
            "title": "Piecewise deterministic generative models",
            "url": "http://arxiv.org/abs/2407.19448v1",
            "arxiv_id": "2407.19448",
            "authors": [
                "Andrea Bertazzi",
                "Alain Oliviero-Durmus",
                "Dario Shariatian",
                "Umut Simsekli",
                "Eric Moulines"
            ],
            "abstract": "We introduce a novel class of generative models based on piecewise deterministic Markov processes (PDMPs), a family of non-diffusive stochastic processes consisting of deterministic motion and random jumps at random times. Similarly to diffusions, such Markov processes admit time reversals that turn out to be PDMPs as well. We apply this observation to three PDMPs considered in the literature: the Zig-Zag process, Bouncy Particle Sampler, and Randomised Hamiltonian Monte Carlo. For these three particular instances, we show that the jump rates and kernels of the corresponding time reversals admit explicit expressions depending on some conditional densities of the PDMP under consideration before and after a jump. Based on these results, we propose efficient training procedures to learn these characteristics and consider methods to approximately simulate the reverse process. Finally, we provide bounds in the total variation distance between the data distribution and the resulting distribution of our model in the case where the base distribution is the standard $d$-dimensional Gaussian distribution. Promising numerical simulations support further investigations into this class of models.",
            "introduction": "   1 Introduction  Diffusion-based generative models (Ho et\u00a0al., 2020; Song et\u00a0al., 2021) have recently achieved state-of-the-art performance in various fields of application (Dhariwal and Nichol, 2021; Croitoru et\u00a0al., 2023; Jeong et\u00a0al., 2021; Kong et\u00a0al., 2021). In their continuous time interpretation (Song et\u00a0al., 2021), these models leverage the idea that a diffusion process can bridge the data distribution \u03bc\u22c6subscript\ud835\udf07\u22c6\\mu_{\\star}italic_\u03bc start_POSTSUBSCRIPT \u22c6 end_POSTSUBSCRIPT to a base distribution \u03c0\ud835\udf0b\\piitalic_\u03c0, and its time reversal can transform samples from \u03c0\ud835\udf0b\\piitalic_\u03c0 into synthetic data from \u03bc\u22c6subscript\ud835\udf07\u22c6\\mu_{\\star}italic_\u03bc start_POSTSUBSCRIPT \u22c6 end_POSTSUBSCRIPT. As shown in the 1980s by Anderson (1982), the time reversal of a diffusion process, i.e., the backward process, is itself a diffusion with explicit drift and covariance functions that are related to the score functions of the time-marginal densities of the original, forward diffusion. Consequently, the key element of these generative models is learning these score functions using techniques such as (denoising) score-matching (Hyv\u00e4rinen, 2005; Vincent, 2011).   In this work, we explore the potential of using piecewise deterministic Markov processes (PDMPs) as noising processes instead of diffusions. PDMPs were introduced around forty years ago (Davis, 1984, 1993) and since then have been successfully applied in various fields, including communication networks (Dumas et\u00a0al., 2002), biology (Berg and Brown, 1972; Cloez, Bertrand et\u00a0al., 2017), risk theory (Embrechts and Schmidli, 1994), and the reliability of complex systems (Zhang et\u00a0al., 2008). More recently, PDMPs have been intensively studied in the context of Monte Carlo algorithms (Fearnhead et\u00a0al., 2018) as alternatives to Langevin diffusion-based methods and Metropolis-Hastings mechanisms. This renewed interest in PDMPs has led to the development of novel processes, such as the Zig-Zag process (ZZP) (Bierkens et\u00a0al., 2019a), the Bouncy Particle Sampler (BPS) (Bouchard-C\u00f4t\u00e9 et\u00a0al., 2018), and the Randomised Hamiltonian Monte Carlo (RHMC) (Bou-Rabee and Sanz-Serna, 2017). Compared to Langevin-based methods, PDMPs offer several advantages, such as better scalability and reduced computational complexity in high-dimensional settings (Bierkens et\u00a0al., 2019a).   In this paper, we propose a new family of generative models based on PDMPs. Our contributions are the following:     1)  Leveraging the existing literature on time reversals of Markov jump processes (Conforti and L\u00e9onard, 2022), we characterise the time reversal of any PDMP under appropriate conditions. It turns out that this time reversal is itself a PDMP with characteristics related to the original PDMP; see Proposition\u00a01.    2)  We further specify the characteristics of the time-reversal processes associated with the three aforementioned PDMPs: ZZP, BPS, and RHMC. For these processes, Proposition\u00a02 shows the corresponding time-reversals are PDMPs with simple reversed deterministic motion and with jump rates and kernels that depend on (ratios of) conditional densities of the velocity of the forward process before and after a jump. In contrast to common diffusion models, the emphasis is on distributions of the velocity, similar to the case of the underdamped Langevin diffusion (Dockhorn et\u00a0al., 2022), which includes an additional velocity vector akin to the PDMPs we consider. Moreover, the structure of the backward jump rates and kernels closely connects to the case of continuous time jump processes on discrete state spaces (Sun et\u00a0al., 2023; Lou et\u00a0al., 2024).    3)  We define our piecewise deterministic generative models employing either ZZP, BPS, or RHMC as forward process,",
            "references": [
                {
                    "title": "Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution",
                    "abstract": "Despite their groundbreaking performance for many generative modeling tasks, diffusion models have fallen short on discrete data domains such as natural language. Crucially, standard diffusion models rely on the well-established theory of score matching, but efforts to generalize this to discrete structures have not yielded the same empirical gains. In this work, we bridge this gap by proposing score entropy, a novel loss that naturally extends score matching to discrete spaces, integrates seamlessly to build discrete diffusion models, and significantly boosts performance. Experimentally, we test our Score Entropy Discrete Diffusion models (SEDD) on standard language modeling tasks. For comparable model sizes, SEDD beats existing language diffusion paradigms (reducing perplexity by $25$-$75$\\%) and is competitive with autoregressive models, in particular outperforming GPT-2. Furthermore, compared to autoregressive mdoels, SEDD generates faithful text without requiring distribution annealing techniques like temperature scaling (around $6$-$8\\times$ better generative perplexity than un-annealed GPT-2), can trade compute and quality (similar quality with $32\\times$ fewer network evaluations), and enables controllable infilling (matching nucleus sampling quality while enabling other strategies besides left to right prompting)."
                },
                {
                    "title": "Piecewise deterministic sampling with splitting schemes",
                    "abstract": "We introduce novel Markov chain Monte Carlo (MCMC) algorithms based on numerical approximations of piecewise-deterministic Markov processes obtained with the framework of splitting schemes. We present unadjusted as well as adjusted algorithms, for which the asymptotic bias due to the discretisation error is removed applying a non-reversible Metropolis-Hastings filter. In a general framework we demonstrate that the unadjusted schemes have weak error of second order in the step size, while typically maintaining a computational cost of only one gradient evaluation of the negative log-target function per iteration. Focusing then on unadjusted schemes based on the Bouncy Particle and Zig-Zag samplers, we provide conditions ensuring geometric ergodicity and consider the expansion of the invariant measure in terms of the step size. We analyse the dependence of the leading term in this expansion on the refreshment rate and on the structure of the splitting scheme, giving a guideline on which structure is best. Finally, we illustrate the competitiveness of our samplers with numerical experiments on a Bayesian imaging inverse problem and a system of interacting particles."
                },
                {
                    "title": "Score-based Continuous-time Discrete Diffusion Models",
                    "abstract": "Score-based modeling through stochastic differential equations (SDEs) has provided a new perspective on diffusion models, and demonstrated superior performance on continuous data. However, the gradient of the log-likelihood function, i.e., the score function, is not properly defined for discrete spaces. This makes it non-trivial to adapt \\textcolor{\\cdiff}{the score-based modeling} to categorical data. In this paper, we extend diffusion models to discrete variables by introducing a stochastic jump process where the reverse process denoises via a continuous-time Markov chain. This formulation admits an analytical simulation during backward sampling. To learn the reverse process, we extend score matching to general categorical data and show that an unbiased estimator can be obtained via simple matching of the conditional marginal distributions. We demonstrate the effectiveness of the proposed method on a set of synthetic and real-world music and image benchmarks."
                },
                {
                    "title": "Sampling is as easy as learning the score: theory for diffusion models with minimal data assumptions",
                    "abstract": "We provide theoretical convergence guarantees for score-based generative models (SGMs) such as denoising diffusion probabilistic models (DDPMs), which constitute the backbone of large-scale real-world generative models such as DALL$\\cdot$E 2. Our main result is that, assuming accurate score estimates, such SGMs can efficiently sample from essentially any realistic data distribution. In contrast to prior works, our results (1) hold for an $L^2$-accurate score estimate (rather than $L^\\infty$-accurate); (2) do not require restrictive functional inequality conditions that preclude substantial non-log-concavity; (3) scale polynomially in all relevant problem parameters; and (4) match state-of-the-art complexity guarantees for discretization of the Langevin diffusion, provided that the score error is sufficiently small. We view this as strong theoretical justification for the empirical success of SGMs. We also examine SGMs based on the critically damped Langevin diffusion (CLD). Contrary to conventional wisdom, we provide evidence that the use of the CLD does not reduce the complexity of SGMs."
                },
                {
                    "title": "Diffusion Models in Vision: A Survey",
                    "abstract": "Denoising diffusion models represent a recent emerging topic in computer vision, demonstrating remarkable results in the area of generative modeling. A diffusion model is a deep generative model that is based on two stages, a forward diffusion stage and a reverse diffusion stage. In the forward diffusion stage, the input data is gradually perturbed over several steps by adding Gaussian noise. In the reverse stage, a model is tasked at recovering the original input data by learning to gradually reverse the diffusion process, step by step. Diffusion models are widely appreciated for the quality and diversity of the generated samples, despite their known computational burdens, i.e., low speeds due to the high number of steps involved during sampling. In this survey, we provide a comprehensive review of articles on denoising diffusion models applied in vision, comprising both theoretical and practical contributions in the field. First, we identify and present three generic diffusion modeling frameworks, which are based on denoising diffusion probabilistic models, noise conditioned score networks, and stochastic differential equations. We further discuss the relations between diffusion models and other deep generative models, including variational auto-encoders, generative adversarial networks, energy-based models, autoregressive models and normalizing flows. Then, we introduce a multi-perspective categorization of diffusion models applied in computer vision. Finally, we illustrate the current limitations of diffusion models and envision some interesting directions for future research."
                },
                {
                    "title": "Score-Based Generative Modeling with Critically-Damped Langevin Diffusion",
                    "abstract": "Score-based generative models (SGMs) have demonstrated remarkable synthesis quality. SGMs rely on a diffusion process that gradually perturbs the data towards a tractable distribution, while the generative model learns to denoise. The complexity of this denoising task is, apart from the data distribution itself, uniquely determined by the diffusion process. We argue that current SGMs employ overly simplistic diffusions, leading to unnecessarily complex denoising processes, which limit generative modeling performance. Based on connections to statistical mechanics, we propose a novel critically-damped Langevin diffusion (CLD) and show that CLD-based SGMs achieve superior performance. CLD can be interpreted as running a joint diffusion in an extended space, where the auxiliary variables can be considered \u201cvelocities\u201d that are coupled to the data variables as in Hamiltonian dynamics. We derive a novel score matching objective for CLD and show that the model only needs to learn the score function of the conditional distribution of the velocity given data, an easier task than learning scores of the data directly. We also derive a new sampling scheme for ef\ufb01cient synthesis from CLD-based diffusion models. We \ufb01nd that CLD outperforms previous SGMs in synthesis quality for similar network architectures and sampling compute budgets. We show that our novel sampler for CLD signi\ufb01cantly outperforms solvers such as Euler\u2013Maruyama. Our framework provides new insights into score-based denoising diffusion models and can be readily used for high-resolution image synthesis. VPSDE. We leave the study of CLD with maximum likelihood training for high-dimensional (image) datasets to future work."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "Diff-TTS: A Denoising Diffusion Model for Text-to-Speech",
                    "abstract": "Although neural text-to-speech (TTS) models have attracted a lot of attention and succeeded in generating human-like speech, there is still room for improvements to its naturalness and architectural efficiency. In this work, we propose a novel non-autoregressive TTS model, namely Diff-TTS, which achieves highly natural and efficient speech synthesis. Given the text, Diff-TTS exploits a denoising diffusion framework to transform the noise signal into a mel-spectrogram via diffusion time steps. In order to learn the mel-spectrogram distribution conditioned on the text, we present a likelihood-based optimization method for TTS. Furthermore, to boost up the inference speed, we leverage the accelerated sampling method that allows Diff-TTS to generate raw waveforms much faster without significantly degrading perceptual quality. Through experiments, we verified that Diff-TTS generates 28 times faster than the real-time with a single NVIDIA 2080Ti GPU."
                },
                {
                    "title": "Improved Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) are a class of generative models which have recently been shown to produce excellent samples. We show that with a few simple modifications, DDPMs can also achieve competitive log-likelihoods while maintaining high sample quality. Additionally, we find that learning variances of the reverse diffusion process allows sampling with an order of magnitude fewer forward passes with a negligible difference in sample quality, which is important for the practical deployment of these models. We additionally use precision and recall to compare how well DDPMs and GANs cover the target distribution. Finally, we show that the sample quality and likelihood of these models scale smoothly with model capacity and training compute, making them easily scalable. We release our code at https://github.com/openai/improved-diffusion"
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "DiffWave: A Versatile Diffusion Model for Audio Synthesis",
                    "abstract": "In this work, we propose DiffWave, a versatile Diffusion probabilistic model for conditional and unconditional Waveform generation. The model is non-autoregressive, and converts the white noise signal into structured waveform through a Markov chain with a constant number of steps at synthesis. It is efficiently trained by optimizing a variant of variational bound on the data likelihood. DiffWave produces high-fidelity audios in Different Waveform generation tasks, including neural vocoding conditioned on mel spectrogram, class-conditional generation, and unconditional generation. We demonstrate that DiffWave matches a strong WaveNet vocoder in terms of speech quality~(MOS: 4.44 versus 4.43), while synthesizing orders of magnitude faster. In particular, it significantly outperforms autoregressive and GAN-based waveform models in the challenging unconditional generation task in terms of audio quality and sample diversity from various automatic and human evaluations."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Normalizing Flows for Probabilistic Modeling and Inference",
                    "abstract": "Normalizing flows provide a general mechanism for defining expressive probability distributions, only requiring the specification of a (usually simple) base distribution and a series of bijective transformations. There has been much recent work on normalizing flows, ranging from improving their expressive power to expanding their application. We believe the field has now matured and is in need of a unified perspective. In this review, we attempt to provide such a perspective by describing flows through the lens of probabilistic modeling and inference. We place special emphasis on the fundamental principles of flow design, and discuss foundational topics such as expressive power and computational trade-offs. We also broaden the conceptual framing of flows by relating them to more general probability transformations. Lastly, we summarize the use of flows for tasks such as generative modeling, approximate inference, and supervised learning."
                },
                {
                    "title": "Neural Spline Flows",
                    "abstract": "A normalizing flow models a complex probability density as an invertible transformation of a simple base density. Flows based on either coupling or autoregressive transforms both offer exact density evaluation and sampling, but rely on the parameterization of an easily invertible elementwise transformation, whose choice determines the flexibility of these models. Building upon recent work, we propose a fully-differentiable module based on monotonic rational-quadratic splines, which enhances the flexibility of both coupling and autoregressive transforms while retaining analytic invertibility. We demonstrate that neural spline flows improve density estimation, variational inference, and generative modeling of images."
                },
                {
                    "title": "Piecewise deterministic Markov processes and their invariant measures",
                    "abstract": "Piecewise Deterministic Markov Processes (PDMPs) are studied in a general framework. First, different constructions are proven to be equivalent. Second, we introduce a coupling between two PDMPs following the same differential flow which implies quantitative bounds on the total variation between the marginal distributions of the two processes. Finally two results are established regarding the invariant measures of PDMPs. A practical condition to show that a probability measure is invariant for the associated PDMP semi-group is presented. In a second time, a bound on the invariant probability measures in V-norm of two PDMPs following the same differential flow is established. This last result is then applied to study the asymptotic bias of some non-exact PDMP MCMC methods."
                },
                {
                    "title": "Geometric ergodicity of the Bouncy Particle Sampler",
                    "abstract": "The Bouncy Particle Sampler (BPS) is a Monte Carlo Markov Chain algorithm to sample from a target density known up to a multiplicative constant. This method is based on a kinetic piecewise deterministic Markov process for which the target measure is invariant. This paper deals with theoretical properties of BPS. First, we establish geometric ergodicity of the associated semi-group under weaker conditions than in [10] both on the target distribution and the velocity probability distribution. This result is based on a new coupling of the process which gives a quantitative minorization condition and yields more insights on the convergence. In addition, we study on a toy model the dependency of the convergence rates on the dimension of the state space. Finally, we apply our results to the analysis of simulated annealing algorithms based on BPS."
                },
                {
                    "title": "Ergodicity of the zigzag process",
                    "abstract": "The zigzag process is a Piecewise Deterministic Markov Process which can be used in a MCMC framework to sample from a given target distribution. We prove the convergence of this process to its target under very weak assumptions, and establish a central limit theorem for empirical averages under stronger assumptions on the decay of the target measure. We use the classical \"Meyn-Tweedie\" approach. The main difficulty turns out to be the proof that the process can indeed reach all the points in the space, even if we consider the minimal switching rates."
                },
                {
                    "title": "Probabilistic and Piecewise Deterministic models in Biology",
                    "abstract": "We present recent results on Piecewise Deterministic Markov Processes (PDMPs), involved in biological modeling. PDMPs, first introduced in the probabilistic literature by Davis (1984), are a very general class of Markov processes and are being increasingly popular in biological applications. They also give new interesting challenges from the theoretical point of view. We give here different examples on the long time behavior of switching Markov models applied to population dynamics, on uniform sampling in general branching models applied to structured population dynamic, on time scale separation in integrate-and-fire models used in neuroscience, and, finally, on moment calculus in stochastic models of gene expression."
                },
                {
                    "title": "Exponential ergodicity of the bouncy particle sampler",
                    "abstract": "Non-reversible Markov chain Monte Carlo schemes based on piecewise deterministic Markov processes have been recently introduced in applied probability, automatic control, physics and statistics. Although these algorithms demonstrate experimentally good performance and are accordingly increasingly used in a wide range of applications, geometric ergodicity results for such schemes have only been established so far under very restrictive assumptions. We give here verifiable conditions on the target distribution under which the Bouncy Particle Sampler algorithm introduced in \\cite{P_dW_12} is geometrically ergodic. This holds whenever the target satisfies a curvature condition and has tails decaying at least as fast as an exponential and at most as fast as a Gaussian distribution. This allows us to provide a central limit theorem for the associated ergodic averages. When the target has tails thinner than a Gaussian distribution, we propose an original modification of this scheme that is geometrically ergodic. For thick-tailed target distributions, such as $t$-distributions, we extend the idea pioneered in \\cite{J_G_12} in a random walk Metropolis context. We apply a change of variable to obtain a transformed target satisfying the tail conditions for geometric ergodicity. By sampling the transformed target using the Bouncy Particle Sampler and mapping back the Markov process to the original parameterization, we obtain a geometrically ergodic algorithm."
                },
                {
                    "title": "Piecewise Deterministic Markov Processes for Continuous-Time Monte Carlo",
                    "abstract": "Recently, there have been conceptually new developments in Monte Carlo methods through the introduction of new MCMC and sequential Monte Carlo (SMC) algorithms which are based on continuous-time, rather than discrete-time, Markov processes. This has led to some fundamentally new Monte Carlo algorithms which can be used to sample from, say, a posterior distribution. Interestingly, continuous-time algorithms seem particularly well suited to Bayesian analysis in big-data settings as they need only access a small sub-set of data points at each iteration, and yet are still guaranteed to target the true posterior distribution. Whilst continuous-time MCMC and SMC methods have been developed independently we show here that they are related by the fact that both involve simulating a piecewise deterministic Markov process. Furthermore, we show that the methods developed to date are just specific cases of a potentially much wider class of continuous-time Monte Carlo algorithms.We give an informal introduction to piecewise deterministic Markov processes, covering the aspects relevant to these new Monte Carlo algorithms, with a view to making the development of new continuoustime Monte Carlo more accessible. We focus on how and why sub-sampling ideas can be used with these algorithms, and aim to give insight into how these new algorithms can be implemented, and what are some of the issues that affect their efficiency."
                },
                {
                    "title": "The Zig-Zag process and super-efficient sampling for Bayesian analysis of big data",
                    "abstract": "Standard MCMC methods can scale poorly to big data settings due to the need to evaluate the likelihood at each iteration. There have been a number of approximate MCMC algorithms that use sub-sampling ideas to reduce this computational burden, but with the drawback that these algorithms no longer target the true posterior distribution. We introduce a new family of Monte Carlo methods based upon a multi-dimensional version of the Zig-Zag process of (Bierkens, Roberts, 2017), a continuous time piecewise deterministic Markov process. While traditional MCMC methods are reversible by construction (a property which is known to inhibit rapid convergence) the Zig-Zag process offers a flexible non-reversible alternative which we observe to often have favourable convergence properties. We show how the Zig-Zag process can be simulated without discretisation error, and give conditions for the process to be ergodic. Most importantly, we introduce a sub-sampling version of the Zig-Zag process that is an example of an {\\em exact approximate scheme}, i.e. the resulting approximate process still has the posterior as its stationary distribution. Furthermore, if we use a control-variate idea to reduce the variance of our unbiased estimator, then the Zig-Zag process can be super-efficient: after an initial pre-processing step, essentially independent samples from the posterior distribution are obtained at a computational cost which does not depend on the size of the data."
                },
                {
                    "title": "Randomized Hamiltonian Monte Carlo",
                    "abstract": "Tuning the durations of the Hamiltonian flow in Hamiltonian Monte Carlo (also called Hybrid Monte Carlo) (HMC) involves a tradeoff between computational cost and sampling quality, which is typically challenging to resolve in a satisfactory way. In this article we present and analyze a randomized HMC method (RHMC), in which these durations are i.i.d. exponential random variables whose mean is a free parameter. We focus on the small time step size limit, where the algorithm is rejection-free and the computational cost is proportional to the mean duration. In this limit, we prove that RHMC is geometrically ergodic under the same conditions that imply geometric ergodicity of the solution to underdamped Langevin equations. Moreover, in the context of a multi-dimensional Gaussian distribution, we prove that the sampling efficiency of RHMC, unlike that of constant duration HMC, behaves in a regular way. This regularity is also verified numerically in non-Gaussian target distributions. Finally we suggest variants of RHMC for which the time step size is not required to be small."
                },
                {
                    "title": "The Bouncy Particle Sampler: A Nonreversible Rejection-Free Markov Chain Monte Carlo Method",
                    "abstract": "ABSTRACT Many Markov chain Monte Carlo techniques currently available rely on discrete-time reversible Markov processes whose transition kernels are variations of the Metropolis\u2013Hastings algorithm. We explore and generalize an alternative scheme recently introduced in the physics literature (Peters and de With 2012) where the target distribution is explored using a continuous-time nonreversible piecewise-deterministic Markov process. In the Metropolis\u2013Hastings algorithm, a trial move to a region of lower target density, equivalently of higher \u201cenergy,\u201d than the current state can be rejected with positive probability. In this alternative approach, a particle moves along straight lines around the space and, when facing a high energy barrier, it is not rejected but its path is modified by bouncing against this barrier. By reformulating this algorithm using inhomogeneous Poisson processes, we exploit standard sampling techniques to simulate exactly this Markov process in a wide range of scenarios of interest. Additionally, when the target distribution is given by a product of factors dependent only on subsets of the state variables, such as the posterior distribution associated with a probabilistic graphical model, this method can be modified to take advantage of this structure by allowing computationally cheaper \u201clocal\u201d bounces, which only involve the state variables associated with a factor, while the other state variables keep on evolving. In this context, by leveraging techniques from chemical kinetics, we propose several computationally efficient implementations. Experimentally, this new class of Markov chain Monte Carlo schemes compares favorably to state-of-the-art methods on various Bayesian inference tasks, including for high-dimensional models and large datasets. Supplementary materials for this article are available online."
                },
                {
                    "title": "MCMC using Hamiltonian dynamics",
                    "abstract": "Hamiltonian dynamics can be used to produce distant proposals for the Metropolis algorithm, thereby avoiding the slow exploration of the state space that results from the diffusive behaviour of simple random-walk proposals. Though originating in physics, Hamiltonian dynamics can be applied to most problems with continuous state spaces by simply introducing fictitious\"momentum\"variables. A key to its usefulness is that Hamiltonian dynamics preserves volume, and its trajectories can thus be used to define complex mappings without the need to account for a hard-to-compute Jacobian factor - a property that can be exactly maintained even when the dynamics is approximated by discretizing time. In this review, I discuss theoretical and practical aspects of Hamiltonian Monte Carlo, and present some of its variations, including using windows of states for deciding on acceptance or rejection, computing trajectories using fast approximations, tempering during the course of a trajectory to handle isolated modes, and short-cut methods that prevent useless trajectories from taking much computation time."
                },
                {
                    "title": "A Connection Between Score Matching and Denoising Autoencoders",
                    "abstract": "Denoising autoencoders have been previously shown to be competitive alternatives to restricted Boltzmann machines for unsupervised pretraining of each layer of a deep architecture. We show that a simple denoising autoencoder training criterion is equivalent to matching the score (with respect to the data) of a specific energy-based model to that of a nonparametric Parzen density estimator of the data. This yields several useful insights. It defines a proper probabilistic model for the denoising autoencoder technique, which makes it in principle possible to sample from them or rank examples by their energy. It suggests a different way to apply score matching that is related to learning to denoise and does not require computing second derivatives. It justifies the use of tied weights between the encoder and decoder and suggests ways to extend the success of denoising autoencoders to a larger family of energy-based models."
                },
                {
                    "title": "Point Process Theory and Applications: Marked Point and Piecewise Deterministic Processes",
                    "abstract": "Scheuren, F. (1989), Comment on \u201cThe Social Policy Simulation Database and Model: An Example of Survey and Administrative Data Integration,\u201d by M. Wolfson, S. Gribble, M. Bordt, B. Murphy, and G. Rowe, Survey of Current Business, 69, 40\u201341. (2005), \u201cParadata From Concept to Completion,\u201d in Symposium 2005: Methodological Challenges for Future Information Needs, Ottawa: Statistics Canada, available at www.statcan.ca/english/freepub/11-522-XIE/ 2005001/9432.pdf."
                },
                {
                    "title": "Estimation of Non-Normalized Statistical Models by Score Matching",
                    "abstract": "One often wants to estimate statistical models where the probability density function is known only up to a multiplicative normalization constant. Typically, one then has to resort to Markov Chain Monte Carlo methods, or approximations of the normalization constant. Here, we propose that such models can be estimated by minimizing the expected squared distance between the gradient of the log-density given by the model and the gradient of the log-density of the observed data. While the estimation of the gradient of log-density function is, in principle, a very difficult non-parametric problem, we prove a surprising result that gives a simple formula for this objective function. The density function of the observed data does not appear in this formula, which simplifies to a sample average of a sum of some derivatives of the log-density given by the model. The validity of the method is demonstrated on multivariate Gaussian and independent component analysis models, and by estimating an overcomplete filter set for natural image data."
                },
                {
                    "title": "A Markovian analysis of additive-increase multiplicative-decrease algorithms",
                    "abstract": "The additive-increase multiplicative-decrease (AIMD) schemes designed to control congestion in communication networks are investigated from a probabilistic point of view. Functional limit theorems for a general class of Markov processes that describe these algorithms are obtained. The asymptotic behaviour of the corresponding invariant measures is described in terms of the limiting Markov processes. For some special important cases, including TCP congestion avoidance, an important autoregressive property is proved. As a consequence, the explicit expression of the related invariant probabilities is derived. The transient behaviour of these algorithms is also analysed."
                },
                {
                    "title": "Ruin estimation for a general insurance risk model",
                    "abstract": "The theory of piecewise-deterministic Markov processes is used in order to investigate insurance risk models where borrowing, investment and inflation are present."
                },
                {
                    "title": "Piecewise\u2010Deterministic Markov Processes: A General Class of Non\u2010Diffusion Stochastic Models",
                    "abstract": "A general class of non-diffusion stochastic models is introduced with a view to providing a framework for studying optimization problems arising in queueing systems, inventory theory, resource allocation and other areas. The corresponding stochastic processes are Markov processes consisting of a mixture of deterministic motion and random jumps. Stochastic calculus for these processes is developed and a complete characterization of the extended generator is given; this is the main technical result of the paper. The relevance of the extended generator concept in applied problems is discussed and some recent results on optimal control of piecewise-deterministic processes are described."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can piecewise deterministic Markov processes (PDMPs) be effectively utilized as noising processes in generative models to improve upon existing diffusion-based generative models?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is significant for the research community as it could lead to the development of more efficient generative models that leverage the advantages of PDMPs, such as better scalability and reduced computational complexity in high-dimensional settings. This advancement could inspire future research to explore alternative stochastic processes in generative modeling, potentially leading to novel applications in various fields, including image synthesis, natural language processing, and beyond. By addressing this question, we could enhance our understanding of the underlying mechanisms of generative processes and open new avenues for practical applications in machine learning.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities involved in characterizing the time reversals of PDMPs and ensuring that these processes maintain the desired properties for effective generative modeling. Naive approaches may fail because they might not adequately capture the intricate relationships between the forward and backward processes, particularly in terms of jump rates and velocity distributions. Additionally, the theoretical underpinnings of PDMPs and their time reversals require a deep understanding of stochastic processes, which adds a layer of difficulty. Overcoming these technical obstacles is crucial for successfully integrating PDMPs into generative models.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on diffusion-based generative models, leaving a gap in the exploration of PDMPs as alternatives. Limitations in understanding the time-reversal characteristics of PDMPs and their application in generative modeling have hindered progress. Additionally, the complexity of PDMPs and their mathematical formulations may have deterred researchers from pursuing this avenue. Our approach differs by explicitly characterizing the time reversals of PDMPs and demonstrating their potential as generative models, thus providing a novel perspective that builds on existing literature while addressing these gaps.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves defining generative models based on PDMPs, specifically utilizing the Zig-Zag process (ZZP), Bouncy Particle Sampler (BPS), or Randomised Hamiltonian Monte Carlo (RHMC) as forward processes. We will characterize the time reversals"
            }
        },
        "author_data": {
            "b6c77b28-58e1-4d14-9147-991e5e02a429": {
                "pk": "b6c77b28-58e1-4d14-9147-991e5e02a429",
                "name": "Andrea Bertazzi",
                "collaborators": [
                    "Joris Bierkens",
                    "Paul Dobson",
                    "Pierre Monmarch\u00e9"
                ],
                "domain": [
                    "Markov Chain Monte Carlo",
                    "Stochastic Processes",
                    "Adaptive Sampling",
                    "Computational Statistics"
                ],
                "publications": [
                    {
                        "title": "Adaptive schemes for piecewise deterministic Monte Carlo algorithms",
                        "abstract": "The Bouncy Particle sampler (BPS) and the Zig-Zag sampler (ZZS) are continuous time, non-reversible Monte Carlo methods based on piecewise deterministic Markov processes. Experiments show that the speed of convergence of these samplers can be affected by the shape of the target distribution, as for instance in the case of anisotropic targets. We propose an adaptive scheme that iteratively learns all or part of the covariance matrix of the target and takes advantage of the obtained information to modify the underlying process with the aim of increasing the speed of convergence. Moreover, we define an adaptive scheme that automatically tunes the refreshment rate of the BPS or ZZS. We prove ergodicity and a law of large numbers for all the proposed adaptive algorithms. Finally, we show the benefits of the adaptive samplers with several numerical simulations."
                    },
                    {
                        "title": "Approximations of Piecewise Deterministic Markov Processes and their convergence properties",
                        "abstract": "Piecewise deterministic Markov processes (PDMPs) are a class of stochastic processes with applications in several fields of applied mathematics spanning from mathematical modeling of physical phenomena to computational methods. A PDMP is specified by three characteristic quantities: the deterministic motion, the law of the random event times, and the jump kernels. The applicability of PDMPs to real world scenarios is currently limited by the fact that these processes can be simulated only when these three characteristics of the process can be simulated exactly. In order to overcome this problem, we introduce discretisation schemes for PDMPs which make their approximate simulation possible. In particular, we design both first order and higher order schemes that rely on approximations of one or more of the three characteristics. For the proposed approximation schemes we study both pathwise convergence to the continuous PDMP as the step size converges to zero and convergence in law to the invariant measure of the PDMP in the long time limit. Moreover, we apply our theoretical results to several PDMPs that arise from the computational statistics and mathematical biology literature."
                    },
                    {
                        "title": "Piecewise deterministic sampling with splitting schemes",
                        "abstract": "We introduce novel Markov chain Monte Carlo (MCMC) algorithms based on numerical approximations of piecewise-deterministic Markov processes obtained with the framework of splitting schemes. We present unadjusted as well as adjusted algorithms, for which the asymptotic bias due to the discretisation error is removed applying a non-reversible Metropolis-Hastings filter. In a general framework we demonstrate that the unadjusted schemes have weak error of second order in the step size, while typically maintaining a computational cost of only one gradient evaluation of the negative log-target function per iteration. Focusing then on unadjusted schemes based on the Bouncy Particle and Zig-Zag samplers, we provide conditions ensuring geometric ergodicity and consider the expansion of the invariant measure in terms of the step size. We analyse the dependence of the leading term in this expansion on the refreshment rate and on the structure of the splitting scheme, giving a guideline on which structure is best. Finally, we illustrate the competitiveness of our samplers with numerical experiments on a Bayesian imaging inverse problem and a system of interacting particles."
                    }
                ]
            },
            "b48d969a-7dcd-4848-a4be-8c72536e0f21": {
                "pk": "b48d969a-7dcd-4848-a4be-8c72536e0f21",
                "name": "Alain Oliviero-Durmus",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "c726d797-ed3b-4e5d-ad0c-48218123dc1b": {
                "pk": "c726d797-ed3b-4e5d-ad0c-48218123dc1b",
                "name": "Dario Shariatian",
                "collaborators": [
                    "Umut Simsekli",
                    "Alain Durmus"
                ],
                "domain": [
                    "Generative Models",
                    "Diffusion Models",
                    "Heavy-tailed Distributions",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Denoising L\u00e9vy Probabilistic Models",
                        "abstract": "Investigating noise distribution beyond Gaussian in diffusion generative models is an open problem. The Gaussian case has seen success experimentally and theoretically, fitting a unified SDE framework for score-based and denoising formulations. Recent studies suggest heavy-tailed noise distributions can address mode collapse and manage datasets with class imbalance, heavy tails, or outliers. Yoon et al. (NeurIPS 2023) introduced the L\\'evy-Ito model (LIM), extending the SDE framework to heavy-tailed SDEs with $\\alpha$-stable noise. Despite its theoretical elegance and performance gains, LIM's complex mathematics may limit its accessibility and broader adoption. This study takes a simpler approach by extending the denoising diffusion probabilistic model (DDPM) with $\\alpha$-stable noise, creating the denoising L\\'evy probabilistic model (DLPM). Using elementary proof techniques, we show DLPM reduces to running vanilla DDPM with minimal changes, allowing the use of existing implementations with minimal changes. DLPM and LIM have different training algorithms and, unlike the Gaussian case, they admit different backward processes and sampling algorithms. Our experiments demonstrate that DLPM achieves better coverage of data distribution tail, improved generation of unbalanced datasets, and faster computation times with fewer backward steps."
                    }
                ]
            },
            "3b270917-6186-4806-9427-9bf2ef5fcd2f": {
                "pk": "3b270917-6186-4806-9427-9bf2ef5fcd2f",
                "name": "Umut Simsekli",
                "collaborators": [
                    "Paul Viallard",
                    "Soheil Kolouri",
                    "Kimia Nadjahi",
                    "Shahin Shahrampour",
                    "Maxime Haddouche",
                    "Benjamin Guedj",
                    "Fabian Schaipp",
                    "Guillaume Garrigos",
                    "Robert Gower",
                    "Benjamin Dupuis"
                ],
                "domain": [
                    "Probability Metrics",
                    "Stochastic Optimization",
                    "Generative Modeling",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Generalized Sliced Distances for Probability Distributions",
                        "abstract": "Probability metrics have become an indispensable part of modern statistics and machine learning, and they play a quintessential role in various applications, including statistical hypothesis testing and generative modeling. However, in a practical setting, the convergence behavior of the algorithms built upon these distances have not been well established, except for a few specific cases. In this paper, we introduce a broad family of probability metrics, coined as Generalized Sliced Probability Metrics (GSPMs), that are deeply rooted in the generalized Radon transform. We first verify that GSPMs are metrics. Then, we identify a subset of GSPMs that are equivalent to maximum mean discrepancy (MMD) with novel positive definite kernels, which come with a unique geometric interpretation. Finally, by exploiting this connection, we consider GSPM-based gradient flows for generative modeling applications and show that under mild assumptions, the gradient flow converges to the global optimum. We illustrate the utility of our approach on both real and synthetic problems."
                    },
                    {
                        "title": "A PAC-Bayesian Link Between Generalisation and Flat Minima",
                        "abstract": "Modern machine learning usually involves predictors in the overparametrised setting (number of trained parameters greater than dataset size), and their training yield not only good performances on training data, but also good generalisation capacity. This phenomenon challenges many theoretical results, and remains an open problem. To reach a better understanding, we provide novel generalisation bounds involving gradient terms. To do so, we combine the PAC-Bayes toolbox with Poincar\\'e and Log-Sobolev inequalities, avoiding an explicit dependency on dimension of the predictor space. Our results highlight the positive influence of \\emph{flat minima} (being minima with a neighbourhood nearly minimising the learning problem as well) on generalisation performances, involving directly the benefits of the optimisation phase."
                    },
                    {
                        "title": "SGD with Clipping is Secretly Estimating the Median Gradient",
                        "abstract": "There are several applications of stochastic optimization where one can benefit from a robust estimate of the gradient. For example, domains such as distributed learning with corrupted nodes, the presence of large outliers in the training data, learning under privacy constraints, or even heavy-tailed noise due to the dynamics of the algorithm itself. Here we study SGD with robust gradient estimators based on estimating the median. We first consider computing the median gradient across samples, and show that the resulting method can converge even under heavy-tailed, state-dependent noise. We then derive iterative methods based on the stochastic proximal point method for computing the geometric median and generalizations thereof. Finally we propose an algorithm estimating the median gradient across iterations, and find that several well known methods - in particular different forms of clipping - are particular cases of this framework."
                    },
                    {
                        "title": "Uniform Generalization Bounds on Data-Dependent Hypothesis Sets via PAC-Bayesian Theory on Random Sets",
                        "abstract": "We propose data-dependent uniform generalization bounds by approaching the problem from a PAC-Bayesian perspective. We first apply the PAC-Bayesian framework on `random sets' in a rigorous way, where the training algorithm is assumed to output a data-dependent hypothesis set after observing the training data. This approach allows us to prove data-dependent bounds, which can be applicable in numerous contexts. To highlight the power of our approach, we consider two main applications. First, we propose a PAC-Bayesian formulation of the recently developed fractal-dimension-based generalization bounds. The derived results are shown to be tighter and they unify the existing results around one simple proof technique. Second, we prove uniform bounds over the trajectories of continuous Langevin dynamics and stochastic gradient Langevin dynamics. These results provide novel information about the generalization properties of noisy algorithms."
                    },
                    {
                        "title": "Denoising L\u00e9vy Probabilistic Models",
                        "abstract": "Investigating noise distribution beyond Gaussian in diffusion generative models is an open problem. The Gaussian case has seen success experimentally and theoretically, fitting a unified SDE framework for score-based and denoising formulations. Recent studies suggest heavy-tailed noise distributions can address mode collapse and manage datasets with class imbalance, heavy tails, or outliers. Yoon et al. (NeurIPS 2023) introduced the L\\'evy-Ito model (LIM), extending the SDE framework to heavy-tailed SDEs with $\\alpha$-stable noise. Despite its theoretical elegance and performance gains, LIM's complex mathematics may limit its accessibility and broader adoption. This study takes a simpler approach by extending the denoising diffusion probabilistic model (DDPM) with $\\alpha$-stable noise, creating the denoising L\\'evy probabilistic model (DLPM). Using elementary proof techniques, we show DLPM reduces to running vanilla DDPM with minimal changes, allowing the use of existing implementations with minimal changes. DLPM and LIM have different training algorithms and, unlike the Gaussian case, they admit different backward processes and sampling algorithms. Our experiments demonstrate that DLPM achieves better coverage of data distribution tail, improved generation of unbalanced datasets, and faster computation times with fewer backward steps."
                    },
                    {
                        "title": "A Tail-Index Analysis of Stochastic Gradient Noise in Deep Neural Networks",
                        "abstract": "The gradient noise (GN) in the stochastic gradient descent (SGD) algorithm is often considered to be Gaussian in the large data regime by assuming that the classical central limit theorem (CLT) kicks in. This assumption is often made for mathematical convenience, since it enables SGD to be analyzed as a stochastic differential equation (SDE) driven by a Brownian motion. We argue that the Gaussianity assumption might fail to hold in deep learning settings and hence render the Brownian motion-based analyses inappropriate. Inspired by non-Gaussian natural phenomena, we consider the GN in a more general context and invoke the generalized CLT (GCLT), which suggests that the GN converges to a heavy-tailed $\\alpha$-stable random variable. Accordingly, we propose to analyze SGD as an SDE driven by a L\\'{e}vy motion. Such SDEs can incur `jumps', which force the SDE transition from narrow minima to wider minima, as proven by existing metastability theory. To validate the $\\alpha$-stable assumption, we conduct extensive experiments on common deep learning architectures and show that in all settings, the GN is highly non-Gaussian and admits heavy-tails. We further investigate the tail behavior in varying network architectures and sizes, loss functions, and datasets. Our results open up a different perspective and shed more light on the belief that SGD prefers wide minima."
                    }
                ]
            },
            "882b2092-07dc-40e9-9d8f-cb815a5bf303": {
                "pk": "882b2092-07dc-40e9-9d8f-cb815a5bf303",
                "name": "Eric Moulines",
                "collaborators": [
                    "Randal Douc",
                    "Philippe Soulier",
                    "Paul Doukhan",
                    "Olivier Capp\u00e9",
                    "Alain Durmus",
                    "Ya'Acov Ritov",
                    "Elisabeth Gassiat",
                    "Benoit Landelle",
                    "Valderio Reisen",
                    "Glaura Franco"
                ],
                "domain": [
                    "Hidden Markov Models",
                    "Time Series Analysis",
                    "Stochastic Processes",
                    "Bayesian Inference"
                ],
                "publications": [
                    {
                        "title": "Forgetting of the initial condition for the filter in general state-space hidden Markov chain: a coupling approach",
                        "abstract": "We give simple conditions that ensure exponential forgetting of the initial conditions of the filter for general state-space hidden Markov chain. The proofs are based on the coupling argument applied to the posterior Markov kernels. These results are useful both for filtering hidden Markov models using approximation methods (e.g., particle filters) and for proving asymptotic properties of estimators. The results are general enough to cover models like the Gaussian state space model, without using the special structure that permits the application of the Kalman filter."
                    },
                    {
                        "title": "Forgetting of the initial distribution for non-ergodic Hidden Markov Chains",
                        "abstract": "In this paper, the forgetting of the initial distribution for a non-ergodic Hidden Markov Models (HMM) is studied. A new set of conditions is proposed to establish the forgetting property of the filter, which significantly extends all the existing results. Both a pathwise-type convergence of the total variation distance of the filter started from two different initial distributions, and a convergence in expectation are considered. The results are illustrated using generic models of non-ergodic HMM and extend all the results known so far."
                    },
                    {
                        "title": "Log-average periodogram estimator of the memory parameter",
                        "abstract": "This paper introduces a semiparametric regression estimator of the memory parameter for long-memory time series process. It is based on the regression in a neighborhood of the zero-frequency of the periodogram averaged over epochs. The proposed estimator is theoretically justified and empirical Monte Carlo investigation gives evidence that the method is very promising to estimate the long-memory parameter."
                    },
                    {
                        "title": "Computable Convergence Rates for Subgeometrically Ergodic Markov Chains",
                        "abstract": "In this paper, we give quantitative bounds on the $f$-total variation distance from convergence of an Harris recurrent Markov chain on an arbitrary under drift and minorisation conditions implying ergodicity at a sub-geometric rate. These bounds are then specialized to the stochastically monotone case, covering the case where there is no minimal reachable element. The results are illustrated on two examples from queueing theory and Markov Chain Monte Carlo."
                    },
                    {
                        "title": "Testing for Homogeneity with Kernel Fisher Discriminant Analysis",
                        "abstract": "We propose to investigate test statistics for testing homogeneity in reproducing kernel Hilbert spaces. Asymptotic null distributions under null hypothesis are derived, and consistency against fixed and local alternatives is assessed. Finally, experimental evidence of the performance of the proposed approach on both artificial data and a speaker verification task is provided."
                    },
                    {
                        "title": "Estimation of the autocovariance function with missing observations",
                        "abstract": "We propose a novel estimator of the autocorrelation function in presence of missing observations. We establish the consistency, the asymptotic normality, and we derive deviation bounds for various classes of weakly dependent stationary time series, including causal or non causal models. In addition, we introduce a modified version periodogram defined from these autocorrelation estimators and derive asymptotic distribution of linear functionals of this estimator."
                    },
                    {
                        "title": "Ergodicity of observation-driven time series models and consistency of the maximum likelihood estimator",
                        "abstract": "This paper deals with a general class of observation-driven time series models with a special focus on time series of counts. We provide conditions under which there exist strict-sense stationary and ergodic versions of such processes. The consistency of the maximum likelihood estimators is then derived for well- specified and misspecified models."
                    },
                    {
                        "title": "On the spectral density of the wavelet coefficients of long memory time series with application to the log-regression estimation of the memory parameter",
                        "abstract": "In the recent years, methods to estimate the memory parameter using wavelet analysis have gained popularity in many areas of science. Despite its widespread use, a rigorous semi-parametric asymptotic theory, comparable to the one developed for Fourier methods, is still missing. In this contribution, we adapt the classical semi-parametric framework introduced by Robinson and his co-authors for estimating the memory parameter of a (possibly) non-stationary process. As an application, we obtain minimax upper bounds for the log-scale regression estimator of the memory parameter for a Gaussian process and we derive an explicit expression of its variance."
                    },
                    {
                        "title": "Online EM for Functional Data",
                        "abstract": "A novel approach to perform unsupervised sequential learning for functional data is proposed. Our goal is to extract reference shapes (referred to as templates) from noisy, deformed and censored realizations of curves and images. Our model generalizes the Bayesian dense deformable template model (Allassonni\\`ere et al., 2007), a hierarchical model in which the template is the function to be estimated and the deformation is a nuisance, assumed to be random with a known prior distribution. The templates are estimated using a Monte Carlo version of the online Expectation-Maximization algorithm, extending the work from Capp\\'e and Moulines (2009). Our sequential inference framework is significantly more computationally efficient than equivalent batch learning algorithms, especially when the missing data is high-dimensional. Some numerical illustrations on curve registration problem and templates extraction from images are provided to support our findings."
                    },
                    {
                        "title": "Online EM Algorithm for Latent Data Models",
                        "abstract": "In this contribution, we propose a generic online (also sometimes called adaptive or recursive) version of the Expectation-Maximisation (EM) algorithm applicable to latent variable models of independent observations. Compared to the algorithm of Titterington (1984), this approach is more directly connected to the usual EM algorithm and does not rely on integration with respect to the complete data distribution. The resulting algorithm is usually simpler and is shown to achieve convergence to the stationary points of the Kullback-Leibler divergence between the marginal distribution of the observation and the model distribution at the optimal rate, i.e., that of the maximum likelihood estimator. In addition, the proposed approach is also suitable for conditional (or regression) models, as illustrated in the case of the mixture of linear regressions model."
                    },
                    {
                        "title": "On Upper-Confidence Bound Policies for Non-Stationary Bandit Problems",
                        "abstract": "Multi-armed bandit problems are considered as a paradigm of the trade-off between exploring the environment to find profitable actions and exploiting what is already known. In the stationary case, the distributions of the rewards do not change in time, Upper-Confidence Bound (UCB) policies have been shown to be rate optimal. A challenging variant of the MABP is the non-stationary bandit problem where the gambler must decide which arm to play while facing the possibility of a changing environment. In this paper, we consider the situation where the distributions of rewards remain constant over epochs and change at unknown time instants. We analyze two algorithms: the discounted UCB and the sliding-window UCB. We establish for these two algorithms an upper-bound for the expected regret by upper-bounding the expectation of the number of times a suboptimal arm is played. For that purpose, we derive a Hoeffding type inequality for self normalized deviations with a random number of summands. We establish a lower-bound for the regret in presence of abrupt changes in the arms reward distributions. We show that the discounted UCB and the sliding-window UCB both match the lower-bound up to a logarithmic factor."
                    },
                    {
                        "title": "Asymptotic properties of the maximum likelihood estimation in misspecified hidden Markov models",
                        "abstract": "Let $(Y_k)_{k\\in \\mathbb{Z}}$ be a stationary sequence on a probability space $(\\Omega,\\mathcal{A},\\mathbb{P})$ taking values in a standard Borel space $\\mathsf{Y}$. Consider the associated maximum likelihood estimator with respect to a parametrized family of hidden Markov models such that the law of the observations $(Y_k)_{k\\in \\mathbb{Z}}$ is not assumed to be described by any of the hidden Markov models of this family. In this paper we investigate the consistency of this estimator in such misspecified models under mild assumptions."
                    },
                    {
                        "title": "Non-asymptotic convergence analysis for the Unadjusted Langevin Algorithm",
                        "abstract": "In this paper, we study a method to sample from a target distribution $\\pi$ over $\\mathbb{R}^d$ having a positive density with respect to the Lebesgue measure, known up to a normalisation factor. This method is based on the Euler discretization of the overdamped Langevin stochastic differential equation associated with $\\pi$. For both constant and decreasing step sizes in the Euler discretization, we obtain non-asymptotic bounds for the convergence to the target distribution $\\pi$ in total variation distance. A particular attention is paid to the dependency on the dimension $d$, to demonstrate the applicability of this method in the high dimensional setting. These bounds improve and extend the results of (Dalalyan 2014)."
                    },
                    {
                        "title": "A quantitative Mc Diarmid's inequality for geometrically ergodic Markov chains",
                        "abstract": "We state and prove a quantitative version of the bounded difference inequality for geometrically ergodic Markov chains.   Our proof uses the same martingale decomposition as \\cite{MR3407208} but, compared to this paper, the exact coupling argument is modified to fill a gap between the strongly aperiodic case and the general aperiodic case."
                    },
                    {
                        "title": "High-dimensional Bayesian inference via the Unadjusted Langevin Algorithm",
                        "abstract": "We consider in this paper the problem of sampling a high-dimensional probability distribution $\\pi$ having a density with respect to the Lebesgue measure on $\\mathbb{R}^d$, known up to a normalization constant $x \\mapsto \\pi(x)= \\mathrm{e}^{-U(x)}/\\int_{\\mathbb{R}^d} \\mathrm{e}^{-U(y)} \\mathrm{d} y$. Such problem naturally occurs for example in Bayesian inference and machine learning. Under the assumption that $U$ is continuously differentiable, $\\nabla U$ is globally Lipschitz and $U$ is strongly convex, we obtain non-asymptotic bounds for the convergence to stationarity in Wasserstein distance of order $2$ and total variation distance of the sampling method based on the Euler discretization of the Langevin stochastic differential equation, for both constant and decreasing step sizes. The dependence on the dimension of the state space of these bounds is explicit. The convergence of an appropriately weighted empirical measure is also investigated and bounds for the mean square error and exponential deviation inequality are reported for functions which are measurable and bounded. An illustration to Bayesian inference for binary regression is presented to support our claims."
                    },
                    {
                        "title": "Stochastic Variable Metric Proximal Gradient with variance reduction for non-convex composite optimization",
                        "abstract": "This paper introduces a novel algorithm, the Perturbed Proximal Preconditioned SPIDER algorithm (3P-SPIDER), designed to solve finite sum non-convex composite optimization. It is a stochastic Variable Metric Forward-Backward algorithm, which allows approximate preconditioned forward operator and uses a variable metric proximity operator as the backward operator; it also proposes a mini-batch strategy with variance reduction to address the finite sum setting. We show that 3P-SPIDER extends some Stochastic preconditioned Gradient Descent-based algorithms and some Incremental Expectation Maximization algorithms to composite optimization and to the case the forward operator can not be computed in closed form. We also provide an explicit control of convergence in expectation of 3P-SPIDER, and study its complexity in order to satisfy the epsilon-approximate stationary condition. Our results are the first to combine the composite non-convex optimization setting, a variance reduction technique to tackle the finite sum setting by using a minibatch strategy and, to allow deterministic or random approximations of the preconditioned forward operator. Finally, through an application to inference in a logistic regression model with random effects, we numerically compare 3P-SPIDER to other stochastic forward-backward algorithms and discuss the role of some design parameters of 3P-SPIDER."
                    },
                    {
                        "title": "Comparison of Resampling Schemes for Particle Filtering",
                        "abstract": "This contribution is devoted to the comparison of various resampling approaches that have been proposed in the literature on particle filtering. It is first shown using simple arguments that the so-called residual and stratified methods do yield an improvement over the basic multinomial resampling approach. A simple counter-example showing that this property does not hold true for systematic resampling is given. Finally, some results on the large-sample behavior of the simple bootstrap filter algorithm are given. In particular, a central limit theorem is established for the case where resampling is performed using the residual approach."
                    }
                ]
            }
        }
    },
    "2309.16620": {
        "paper_data": {
            "title": "Depthwise Hyperparameter Transfer in Residual Networks: Dynamics and Scaling Limit",
            "url": "http://arxiv.org/abs/2309.16620v2",
            "arxiv_id": "2309.16620",
            "authors": [
                "Blake Bordelon",
                "Lorenzo Noci",
                "Mufan Bill Li",
                "Boris Hanin",
                "Cengiz Pehlevan"
            ],
            "abstract": "The cost of hyperparameter tuning in deep learning has been rising with model sizes, prompting practitioners to find new tuning methods using a proxy of smaller networks. One such proposal uses $\\mu$P parameterized networks, where the optimal hyperparameters for small width networks transfer to networks with arbitrarily large width. However, in this scheme, hyperparameters do not transfer across depths. As a remedy, we study residual networks with a residual branch scale of $1/\\sqrt{\\text{depth}}$ in combination with the $\\mu$P parameterization. We provide experiments demonstrating that residual architectures including convolutional ResNets and Vision Transformers trained with this parameterization exhibit transfer of optimal hyperparameters across width and depth on CIFAR-10 and ImageNet. Furthermore, our empirical findings are supported and motivated by theory. Using recent developments in the dynamical mean field theory (DMFT) description of neural network learning dynamics, we show that this parameterization of ResNets admits a well-defined feature learning joint infinite-width and infinite-depth limit and show convergence of finite-size network dynamics towards this limit.",
            "introduction": "   1 Introduction     (a) ResNet, \u03bc\ud835\udf07\\muitalic_\u03bcP     (b) This work: 1L1\ud835\udc3f\\frac{1}{\\sqrt{L}}divide start_ARG 1 end_ARG start_ARG square-root start_ARG italic_L end_ARG end_ARG-ResNet + \u03bc\ud835\udf07\\muitalic_\u03bcP     Figure 1: The optimal learning rate \u03b7*superscript\ud835\udf02\\eta^{*}italic_\u03b7 start_POSTSUPERSCRIPT * end_POSTSUPERSCRIPT transfers across both depth and width in our proposed parameterization but not in \u03bc\ud835\udf07\\muitalic_\u03bcP or standard parameterization (Fig.\u00a0B.2). Loss is plotted after 20202020 epochs on CIFAR-10. All the missing datapoints indicate that the corresponding run diverged.    Increasing the number of parameters in a neural network has led to consistent and often dramatic improvements in model quality (Kaplan et\u00a0al., 2020; Hoffmann et\u00a0al., 2022; Zhai et\u00a0al., 2022; Klug & Heckel, 2023; OpenAI, 2023). To realize these gains, however, it is typically necessary to conduct by trial-and-error a grid search for optimal choices of hyperparameters, such as learning rates. Doing so directly in SOTA models, which may have hundreds of billions (Brown et\u00a0al., 2020) or even trillions of parameters (Fedus et\u00a0al., 2022), often incurs prohibitive computational costs.   To combat this, an influential recent line of work by Yang & Hu (2021) proposed the so-called \u03bc\ud835\udf07\\muitalic_\u03bcP parameterization, which seeks to develop principles by which optimal hyperparameters from small networks can be reused \u2014 or transferred \u2014 to larger networks (Yang et\u00a0al., 2021). The \u03bc\ud835\udf07\\muitalic_\u03bcP prescription focuses on transfer from narrower to wider models, but does not always transfer across depth (see Figure 1(a)). This leads us to pose the following problem:    Question: Can we transfer hyperparameters simultaneously across depth and width?    In this work, we provide an affirmative answer for a particular flexible class of residual architectures (including ResNets and transformers) with residual branches scaled like 1/depth1depth1/\\sqrt{\\text{depth}}1 / square-root start_ARG depth end_ARG (see Equation\u00a01 and Table 1 for the exact parameterization). More specifically, our empirical contributions can be summarized as follows:   \u2022  We show that a simple variant of the \u03bc\ud835\udf07\\muitalic_\u03bcP parameterization allows for transfer of learning rates simultaneously across depth and width in residual networks with 1/depth1depth1/\\sqrt{\\text{depth}}1 / square-root start_ARG depth end_ARG scaling on the residual branches (see Equation\u00a01 and Figure 1(b)). Moreover, we observe transfer not only for training with a fixed learning rate, but also with complex learning rate schedules (Figure 3(a-d)).    \u2022  We show that the lack of learning rate transfer across depth in \u03bc\ud835\udf07\\muitalic_\u03bcP ResNets can be partly overcome by adding normalization layers (Figure 2). However, we find on CIFAR-10, Tiny ImageNet, ImageNet, that the transfer can be significantly improved with our 1/depth1depth1/\\sqrt{\\text{depth}}1 / square-root start_ARG depth end_ARG residual branches (Figures 3 and B.6).     \u2022  While most of our experiments concern residual convolutional networks, such as wide ResNets, we also present evidence that our prescription leads to learning rate transfer in Vision Transformers (ViTs), both with and without normalization (Figure 3).    \u2022  We demonstrate transfer for learning rates, as well as other hyperparameters such as momentum coefficients and regularization strengths in convolutional residual networks (Figure 4).         (a) ResNet + \u03bc\ud835\udf07\\muitalic_\u03bcP, BatchNorm     (b) 1L1\ud835\udc3f\\frac{1}{\\sqrt{L}}divide start_ARG 1 end_ARG start_ARG square-root start_ARG italic_L end_ARG end_ARG ResNet + \u03bc\ud835\udf07\\muitalic_\u03bcP, BatchNorm    Figure 2: Effect of batch normalization on a standard Resnet18-type architecture (He et\u00a0al., 2016). The above examples are taken after 20202020 epochs on CIFAR-10. Normalization layers can slightly improve consistency and trainability across depths in standard ResNets, but consistency is much more reliable with the 1/L1\ud835\udc3f1/\\sqrt{L}1",
            "references": [
                {
                    "title": "The Shaped Transformer: Attention Models in the Infinite Depth-and-Width Limit",
                    "abstract": "In deep learning theory, the covariance matrix of the representations serves as a proxy to examine the network's trainability. Motivated by the success of Transformers, we study the covariance matrix of a modified Softmax-based attention model with skip connections in the proportional limit of infinite-depth-and-width. We show that at initialization the limiting distribution can be described by a stochastic differential equation (SDE) indexed by the depth-to-width ratio. To achieve a well-defined stochastic limit, the Transformer's attention mechanism is modified by centering the Softmax output at identity, and scaling the Softmax logits by a width-dependent temperature parameter. We examine the stability of the network through the corresponding SDE, showing how the scale of both the drift and diffusion can be elegantly controlled with the aid of residual connections. The existence of a stable SDE implies that the covariance structure is well-behaved, even for very large depth and width, thus preventing the notorious issues of rank degeneracy in deep attention models. Finally, we show, through simulations, that the SDE provides a surprisingly good description of the corresponding finite-size model. We coin the name shaped Transformer for these architectural modifications."
                },
                {
                    "title": "On the impact of activation and normalization in obtaining isometric embeddings at initialization",
                    "abstract": "In this paper, we explore the structure of the penultimate Gram matrix in deep neural networks, which contains the pairwise inner products of outputs corresponding to a batch of inputs. In several architectures it has been observed that this Gram matrix becomes degenerate with depth at initialization, which dramatically slows training. Normalization layers, such as batch or layer normalization, play a pivotal role in preventing the rank collapse issue. Despite promising advances, the existing theoretical results do not extend to layer normalization, which is widely used in transformers, and can not quantitatively characterize the role of non-linear activations. To bridge this gap, we prove that layer normalization, in conjunction with activation layers, biases the Gram matrix of a multilayer perceptron towards the identity matrix at an exponential rate with depth at initialization. We quantify this rate using the Hermite expansion of the activation function."
                },
                {
                    "title": "Feature-Learning Networks Are Consistent Across Widths At Realistic Scales",
                    "abstract": "We study the effect of width on the dynamics of feature-learning neural networks across a variety of architectures and datasets. Early in training, wide neural networks trained on online data have not only identical loss curves but also agree in their point-wise test predictions throughout training. For simple tasks such as CIFAR-5m this holds throughout training for networks of realistic widths. We also show that structural properties of the models, including internal representations, preactivation distributions, edge of stability phenomena, and large learning rate effects are consistent across large widths. This motivates the hypothesis that phenomena seen in realistic models can be captured by infinite-width, feature-learning limits. For harder tasks (such as ImageNet and language modeling), and later training times, finite-width deviations grow systematically. Two distinct effects cause these deviations across widths. First, the network output has initialization-dependent variance scaling inversely with width, which can be removed by ensembling networks. We observe, however, that ensembles of narrower networks perform worse than a single wide network. We call this the bias of narrower width. We conclude with a spectral perspective on the origin of this finite-width bias."
                },
                {
                    "title": "Depth Dependence of \u03bcP Learning Rates in ReLU MLPs",
                    "abstract": "In this short note we consider random fully connected ReLU networks of width $n$ and depth $L$ equipped with a mean-field weight initialization. Our purpose is to study the dependence on $n$ and $L$ of the maximal update ($\\mu$P) learning rate, the largest learning rate for which the mean squared change in pre-activations after one step of gradient descent remains uniformly bounded at large $n,L$. As in prior work on $\\mu$P of Yang et. al., we find that this maximal update learning rate is independent of $n$ for all but the first and last layer weights. However, we find that it has a non-trivial dependence of $L$, scaling like $L^{-3/2}.$"
                },
                {
                    "title": "Field theory for optimal signal propagation in ResNets",
                    "abstract": "Residual networks have significantly better trainability and thus performance than feed-forward networks at large depth. Introducing skip connections facilitates signal propagation to deeper layers. In addition, previous works found that adding a scaling parameter for the residual branch further improves generalization performance. While they empirically identified a particularly beneficial range of values for this scaling parameter, the associated performance improvement and its universality across network hyperparameters yet need to be understood. For feed-forward networks, finite-size theories have led to important insights with regard to signal propagation and hyperparameter tuning. We here derive a systematic finite-size field theory for residual networks to study signal propagation and its dependence on the scaling for the residual branch. We derive analytical expressions for the response function, a measure for the network's sensitivity to inputs, and show that for deep networks the empirically found values for the scaling parameter lie within the range of maximal sensitivity. Furthermore, we obtain an analytical expression for the optimal scaling parameter that depends only weakly on other network hyperparameters, such as the weight variance, thereby explaining its universality across hyperparameters. Overall, this work provides a theoretical framework to study ResNets at finite size."
                },
                {
                    "title": "Automatic Gradient Descent: Deep Learning without Hyperparameters",
                    "abstract": "The architecture of a deep neural network is defined explicitly in terms of the number of layers, the width of each layer and the general network topology. Existing optimisation frameworks neglect this information in favour of implicit architectural information (e.g. second-order methods) or architecture-agnostic distance functions (e.g. mirror descent). Meanwhile, the most popular optimiser in practice, Adam, is based on heuristics. This paper builds a new framework for deriving optimisation algorithms that explicitly leverage neural architecture. The theory extends mirror descent to non-convex composite objective functions: the idea is to transform a Bregman divergence to account for the non-linear structure of neural architecture. Working through the details for deep fully-connected networks yields automatic gradient descent: a first-order optimiser without any hyperparameters. Automatic gradient descent trains both fully-connected and convolutional networks out-of-the-box and at ImageNet scale. A PyTorch implementation is available at https://github.com/jxbz/agd and also in Appendix B. Overall, the paper supplies a rigorous theoretical foundation for a next-generation of architecture-dependent optimisers that work automatically and without hyperparameters."
                },
                {
                    "title": "Dynamics of Finite Width Kernel and Prediction Fluctuations in Mean Field Neural Networks",
                    "abstract": "\n We analyze the dynamics of finite width effects in wide but finite feature learning neural networks. Starting from a dynamical mean field theory description of infinite width deep neural network kernel and prediction dynamics, we provide a characterization of the \n \n \n \n \n O\n \n (\n 1\n \n /\n \n \n width\n \n )\n \n \n fluctuations of the dynamical mean field theory order parameters over random initializations of the network weights. Our results, while perturbative in width, unlike prior analyses, are non-perturbative in the strength of feature learning. We find that once the mean field/\u00b5P parameterization is adopted, the leading finite size effect on the dynamics is to introduce initialization variance in the predictions and feature kernels of the networks. In the lazy limit of network training, all kernels are random but static in time and the prediction variance has a universal form. However, in the rich, feature learning regime, the fluctuations of the kernels and predictions are dynamically coupled with a variance that can be computed self-consistently. In two layer networks, we show how feature learning can dynamically reduce the variance of the final tangent kernel and final network predictions. We also show how initialization variance can slow down online learning in wide but finite networks. In deeper networks, kernel variance can dramatically accumulate through subsequent layers at large feature learning strengths, but feature learning continues to improve the signal-to-noise ratio of the feature kernels. In discrete time, we demonstrate that large learning rate phenomena such as edge of stability effects can be well captured by infinite width dynamics and that initialization variance can decrease dynamically. For convolutional neural networks trained on CIFAR-10, we empirically find significant corrections to both the bias and variance of network dynamics due to finite width."
                },
                {
                    "title": "Neural signature kernels as infinite-width-depth-limits of controlled ResNets",
                    "abstract": "Motivated by the paradigm of reservoir computing, we consider randomly initialized controlled ResNets defined as Euler-discretizations of neural controlled differential equations (Neural CDEs), a unified architecture which enconpasses both RNNs and ResNets. We show that in the infinite-width-depth limit and under proper scaling, these architectures converge weakly to Gaussian processes indexed on some spaces of continuous paths and with kernels satisfying certain partial differential equations (PDEs) varying according to the choice of activation function, extending the results of Hayou (2022); Hayou&Yang (2023) to the controlled and homogeneous case. In the special, homogeneous, case where the activation is the identity, we show that the equation reduces to a linear PDE and the limiting kernel agrees with the signature kernel of Salvi et al. (2021a). We name this new family of limiting kernels neural signature kernels. Finally, we show that in the infinite-depth regime, finite-width controlled ResNets converge in distribution to Neural CDEs with random vector fields which, depending on whether the weights are shared across layers, are either time-independent and Gaussian or behave like a matrix-valued Brownian motion."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Width and Depth Limits Commute in Residual Networks",
                    "abstract": "We show that taking the width and depth to infinity in a deep neural network with skip connections, when branches are scaled by $1/\\sqrt{depth}$ (the only nontrivial scaling), result in the same covariance structure no matter how that limit is taken. This explains why the standard infinite-width-then-depth approach provides practical insights even for networks with depth of the same order as width. We also demonstrate that the pre-activations, in this case, have Gaussian distributions which has direct applications in Bayesian deep learning. We conduct extensive simulations that show an excellent match with our theoretical findings."
                },
                {
                    "title": "Bayesian interpolation with deep linear networks",
                    "abstract": "Significance Understanding the interplay between network architecture, dataset statistics, and learning algorithms is a key challenge in deep learning. We overcome this challenge analytically for zero-noise Bayesian inference in neural networks with linear activations and output dimension one but arbitrary depth, width, and training data. Our results provide the first rigorous solutions to learning with neural networks in which depth, width, and dataset can vary freely. We identify an emergent notion of effective depth, which combines depth, width, and dataset size, that must be large for networks to compute optimal posteriors from universal priors. Our results both present novel examples of models admitting exact solutions to Bayesian inference and give the first principled Bayesian justification for preferring deeper networks."
                },
                {
                    "title": "Rigorous dynamical mean field theory for stochastic gradient descent methods",
                    "abstract": "We prove closed-form equations for the exact high-dimensional asymptotics of a family of first order gradient-based methods, learning an estimator (e.g. M-estimator, shallow neural network, ...) from observations on Gaussian data with empirical risk minimization. This includes widely used algorithms such as stochastic gradient descent (SGD) or Nesterov acceleration. The obtained equations match those resulting from the discretization of dynamical mean-field theory (DMFT) equations from statistical physics when applied to gradient flow. Our proof method allows us to give an explicit description of how memory kernels build up in the effective dynamics, and to include non-separable update functions, allowing datasets with non-identity covariance matrices. Finally, we provide numerical implementations of the equations for SGD with generic extensive batch-size and with constant learning rates."
                },
                {
                    "title": "The Influence of Learning Rule on Representation Dynamics in Wide Neural Networks",
                    "abstract": "It is unclear how changing the learning rule of a deep neural network alters its learning dynamics and representations. To gain insight into the relationship between learned features, function approximation, and the learning rule, we analyze infinite-width deep networks trained with gradient descent (GD) and biologically-plausible alternatives including feedback alignment (FA), direct feedback alignment (DFA), and error modulated Hebbian learning (Hebb), as well as gated linear networks (GLN). We show that, for each of these learning rules, the evolution of the output function at infinite width is governed by a time varying effective neural tangent kernel (eNTK). In the lazy training limit, this eNTK is static and does not evolve, while in the rich mean-field regime this kernel's evolution can be determined self-consistently with dynamical mean field theory (DMFT). This DMFT enables comparisons of the feature and prediction dynamics induced by each of these learning rules. In the lazy limit, we find that DFA and Hebb can only learn using the last layer features, while full FA can utilize earlier layers with a scale determined by the initial correlation between feedforward and feedback weight matrices. In the rich regime, DFA and FA utilize a temporally evolving and depth-dependent NTK. Counterintuitively, we find that FA networks trained in the rich regime exhibit more feature learning if initialized with smaller correlation between the forward and backward pass weights. GLNs admit a very simple formula for their lazy limit kernel and preserve conditional Gaussianity of their preactivations under gating functions. Error modulated Hebb rules show very small task-relevant alignment of their kernels and perform most task relevant learning in the last layer."
                },
                {
                    "title": "On the infinite-depth limit of finite-width neural networks",
                    "abstract": "In this paper, we study the infinite-depth limit of finite-width residual neural networks with random Gaussian weights. With proper scaling, we show that by fixing the width and taking the depth to infinity, the pre-activations converge in distribution to a zero-drift diffusion process. Unlike the infinite-width limit where the pre-activation converge weakly to a Gaussian random variable, we show that the infinite-depth limit yields different distributions depending on the choice of the activation function. We document two cases where these distributions have closed-form (different) expressions. We further show an intriguing change of regime phenomenon of the post-activation norms when the width increases from 3 to 4. Lastly, we study the sequential limit infinite-depth-then-infinite-width and compare it with the more commonly studied infinite-width-then-infinite-depth limit."
                },
                {
                    "title": "Scaling Laws For Deep Learning Based Image Reconstruction",
                    "abstract": "Deep neural networks trained end-to-end to map a measurement of a (noisy) image to a clean image perform excellent for a variety of linear inverse problems. Current methods are only trained on a few hundreds or thousands of images as opposed to the millions of examples deep networks are trained on in other domains. In this work, we study whether major performance gains are expected from scaling up the training set size. We consider image denoising, accelerated magnetic resonance imaging, and super-resolution and empirically determine the reconstruction quality as a function of training set size, while simultaneously scaling the network size. For all three tasks we find that an initially steep power-law scaling slows significantly already at moderate training set sizes. Interpolating those scaling laws suggests that even training on millions of images would not significantly improve performance. To understand the expected behavior, we analytically characterize the performance of a linear estimator learned with early stopped gradient descent. The result formalizes the intuition that once the error induced by learning the signal model is small relative to the error floor, more training examples do not improve performance."
                },
                {
                    "title": "Signal Propagation in Transformers: Theoretical Perspectives and the Role of Rank Collapse",
                    "abstract": "Transformers have achieved remarkable success in several domains, ranging from natural language processing to computer vision. Nevertheless, it has been recently shown that stacking self-attention layers - the distinctive architectural component of Transformers - can result in rank collapse of the tokens' representations at initialization. The question of if and how rank collapse affects training is still largely unanswered, and its investigation is necessary for a more comprehensive understanding of this architecture. In this work, we shed new light on the causes and the effects of this phenomenon. First, we show that rank collapse of the tokens' representations hinders training by causing the gradients of the queries and keys to vanish at initialization. Furthermore, we provide a thorough description of the origin of rank collapse and discuss how to prevent it via an appropriate depth-dependent scaling of the residual branches. Finally, our analysis unveils that specific architectural hyperparameters affect the gradients of queries and values differently, leading to disproportionate gradient norms. This suggests an explanation for the widespread use of adaptive methods for Transformers' optimization."
                },
                {
                    "title": "The Neural Covariance SDE: Shaped Infinite Depth-and-Width Networks at Initialization",
                    "abstract": "The logit outputs of a feedforward neural network at initialization are conditionally Gaussian, given a random covariance matrix defined by the penultimate layer. In this work, we study the distribution of this random matrix. Recent work has shown that shaping the activation function as network depth grows large is necessary for this covariance matrix to be non-degenerate. However, the current infinite-width-style understanding of this shaping method is unsatisfactory for large depth: infinite-width analyses ignore the microscopic fluctuations from layer to layer, but these fluctuations accumulate over many layers. To overcome this shortcoming, we study the random covariance matrix in the shaped infinite-depth-and-width limit. We identify the precise scaling of the activation function necessary to arrive at a non-trivial limit, and show that the random covariance matrix is governed by a stochastic differential equation (SDE) that we call the Neural Covariance SDE. Using simulations, we show that the SDE closely matches the distribution of the random covariance matrix of finite networks. Additionally, we recover an if-and-only-if condition for exploding and vanishing norms of large shaped networks based on the activation function."
                },
                {
                    "title": "Self-consistent dynamical field theory of kernel evolution in wide neural networks",
                    "abstract": "We analyze feature learning in infinite-width neural networks trained with gradient flow through a self-consistent dynamical field theory. We construct a collection of deterministic dynamical order parameters which are inner-product kernels for hidden unit activations and gradients in each layer at pairs of time points, providing a reduced description of network activity through training. These kernel order parameters collectively define the hidden layer activation distribution, the evolution of the neural tangent kernel (NTK), and consequently, output predictions. We show that the field theory derivation recovers the recursive stochastic process of infinite-width feature learning networks obtained by Yang and Hu with tensor programs. For deep linear networks, these kernels satisfy a set of algebraic matrix equations. For nonlinear networks, we provide an alternating sampling procedure to self-consistently solve for the kernel order parameters. We provide comparisons of the self-consistent solution to various approximation schemes including the static NTK approximation, gradient independence assumption, and leading order perturbation theory, showing that each of these approximations can break down in regimes where general self-consistent solutions still provide an accurate description. Lastly, we provide experiments in more realistic settings which demonstrate that the loss and kernel dynamics of convolutional neural networks at fixed feature learning strength are preserved across different widths on a image classification task."
                },
                {
                    "title": "Training Compute-Optimal Large Language Models",
                    "abstract": "We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over 400 language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, Chinchilla, that uses the same compute budget as Gopher but with 70B parameters and 4$\\times$ more more data. Chinchilla uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that Chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, Chinchilla reaches a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, greater than a 7% improvement over Gopher."
                },
                {
                    "title": "Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer",
                    "abstract": "Hyperparameter (HP) tuning in deep learning is an expensive process, prohibitively so for neural networks (NNs) with billions of parameters. We show that, in the recently discovered Maximal Update Parametrization (muP), many optimal HPs remain stable even as model size changes. This leads to a new HP tuning paradigm we call muTransfer: parametrize the target model in muP, tune the HP indirectly on a smaller model, and zero-shot transfer them to the full-sized model, i.e., without directly tuning the latter at all. We verify muTransfer on Transformer and ResNet. For example, 1) by transferring pretraining HPs from a model of 13M parameters, we outperform published numbers of BERT-large (350M parameters), with a total tuning cost equivalent to pretraining BERT-large once; 2) by transferring from 40M parameters, we outperform published numbers of the 6.7B GPT-3 model, with tuning cost only 7% of total pretraining cost. A Pytorch implementation of our technique can be found at github.com/microsoft/mup and installable via `pip install mup`."
                },
                {
                    "title": "Unified field theoretical approach to deep and recurrent neuronal networks",
                    "abstract": "Understanding capabilities and limitations of different network architectures is of fundamental importance to machine learning. Bayesian inference on Gaussian processes has proven to be a viable approach for studying recurrent and deep networks in the limit of infinite layer width, n \u2192 \u221e. Here we present a unified and systematic derivation of the mean-field theory for both architectures that starts from first principles by employing established methods from statistical physics of disordered systems. The theory elucidates that while the mean-field equations are different with regard to their temporal structure, they yet yield identical Gaussian kernels when readouts are taken at a single time point or layer, respectively. Bayesian inference applied to classification then predicts identical performance and capabilities for the two architectures. Numerically, we find that convergence towards the mean-field theory is typically slower for recurrent networks than for deep networks and the convergence speed depends non-trivially on the parameters of the weight prior as well as the depth or number of time steps, respectively. Our method exposes that Gaussian processes are but the lowest order of a systematic expansion in 1/n and we compute next-to-leading-order corrections which turn out to be architecture-specific. The formalism thus paves the way to investigate the fundamental differences between recurrent and deep architectures at finite widths n."
                },
                {
                    "title": "Precise characterization of the prior predictive distribution of deep ReLU networks",
                    "abstract": "Recent works on Bayesian neural networks (BNNs) have highlighted the need to better understand the implications of using Gaussian priors in combination with the compositional structure of the network architecture. Similar in spirit to the kind of analysis that has been developed to devise better initialization schemes for neural networks (cf. He- or Xavier initialization), we derive a precise characterization of the prior predictive distribution of finite-width ReLU networks with Gaussian weights. While theoretical results have been obtained for their heavy-tailedness, the full characterization of the prior predictive distribution (i.e. its density, CDF and moments), remained unknown prior to this work. Our analysis, based on the Meijer-G function, allows us to quantify the influence of architectural choices such as the width or depth of the network on the resulting shape of the prior predictive distribution. We also formally connect our results to previous work in the infinite width setting, demonstrating that the moments of the distribution converge to those of a normal log-normal mixture in the infinite depth limit. Finally, our results provide valuable guidance on prior design: for instance, controlling the predictive variance with depth- and width-informed priors on the weights of the network."
                },
                {
                    "title": "Scaling Vision Transformers",
                    "abstract": "Attention-based neural networks such as the Vision Transformer (ViT) have recently attained state-of-the-art results on many computer vision benchmarks. Scale is a primary ingredient in attaining excellent results, therefore, understanding a model's scaling properties is a key to designing future generations effectively. While the laws for scaling Transformer language models have been studied, it is unknown how Vision Transformers scale. To address this, we scale ViT models and data, both up and down, and characterize the relationships between error rate, data, and compute. Along the way, we refine the architecture and training of ViT, reducing memory consumption and increasing accuracy of the resulting models. As a result, we successfully train a ViT model with two billion parameters, which attains a new state-of-the-art on ImageNet of 90.45% top-1 accuracy. The model also performs well for few-shot transfer, for example, reaching 84.86% top-1 accuracy on ImageNet with only 10 examples per class."
                },
                {
                    "title": "Batch Normalization Orthogonalizes Representations in Deep Random Networks",
                    "abstract": "This paper underlines a subtle property of batch-normalization (BN): Successive batch normalizations with random linear transformations make hidden representations increasingly orthogonal across layers of a deep neural network. We establish a non-asymptotic characterization of the interplay between depth, width, and the orthogonality of deep representations. More precisely, under a mild assumption, we prove that the deviation of the representations from orthogonality rapidly decays with depth up to a term inversely proportional to the network width. This result has two main implications: 1) Theoretically, as the depth grows, the distribution of the representation -- after the linear layers -- contracts to a Wasserstein-2 ball around an isotropic Gaussian distribution. Furthermore, the radius of this Wasserstein ball shrinks with the width of the network. 2) In practice, the orthogonality of the representations directly influences the performance of stochastic gradient descent (SGD). When representations are initially aligned, we observe SGD wastes many iterations to orthogonalize representations before the classification. Nevertheless, we experimentally show that starting optimization from orthogonal representations is sufficient to accelerate SGD, with no need for BN."
                },
                {
                    "title": "The Future is Log-Gaussian: ResNets and Their Infinite-Depth-and-Width Limit at Initialization",
                    "abstract": "Theoretical results show that neural networks can be approximated by Gaussian processes in the infinite-width limit. However, for fully connected networks, it has been previously shown that for any fixed network width, $n$, the Gaussian approximation gets worse as the network depth, $d$, increases. Given that modern networks are deep, this raises the question of how well modern architectures, like ResNets, are captured by the infinite-width limit. To provide a better approximation, we study ReLU ResNets in the infinite-depth-and-width limit, where both depth and width tend to infinity as their ratio, $d/n$, remains constant. In contrast to the Gaussian infinite-width limit, we show theoretically that the network exhibits log-Gaussian behaviour at initialization in the infinite-depth-and-width limit, with parameters depending on the ratio $d/n$. Using Monte Carlo simulations, we demonstrate that even basic properties of standard ResNet architectures are poorly captured by the Gaussian limit, but remarkably well captured by our log-Gaussian limit. Moreover, our analysis reveals that ReLU ResNets at initialization are hypoactivated: fewer than half of the ReLUs are activated. Additionally, we calculate the interlayer correlations, which have the effect of exponentially increasing the variance of the network output. Based on our analysis, we introduce Balanced ResNets, a simple architecture modification, which eliminates hypoactivation and interlayer correlations and is more amenable to theoretical analysis."
                },
                {
                    "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity",
                    "abstract": "In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with outrageous numbers of parameters -- but a constant computational cost. However, despite several notable successes of MoE, widespread adoption has been hindered by complexity, communication costs and training instability -- we address these with the Switch Transformer. We simplify the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs. Our proposed training techniques help wrangle the instabilities and we show large sparse models may be trained, for the first time, with lower precision (bfloat16) formats. We design models based off T5-Base and T5-Large to obtain up to 7x increases in pre-training speed with the same computational resources. These improvements extend into multilingual settings where we measure gains over the mT5-Base version across all 101 languages. Finally, we advance the current scale of language models by pre-training up to trillion parameter models on the\"Colossal Clean Crawled Corpus\"and achieve a 4x speedup over the T5-XXL model."
                },
                {
                    "title": "Statistical Mechanics of Deep Linear Neural Networks: The Backpropagating Kernel Renormalization",
                    "abstract": "The success of deep learning in many real-world tasks has triggered an intense effort to understand the power and limitations of deep learning in the training and generalization of complex tasks, so far with limited progress. In this work, we study the statistical mechanics of learning in Deep Linear Neural Networks (DLNNs) in which the input-output function of an individual unit is linear. Despite the linearity of the units, learning in DLNNs is nonlinear, hence studying its properties reveals some of the features of nonlinear Deep Neural Networks (DNNs). Importantly, we solve exactly the network properties following supervised learning using an equilibrium Gibbs distribution in the weight space. To do this, we introduce the Back-Propagating Kernel Renormalization (BPKR), which allows for the incremental integration of the network weights starting from the network output layer and progressing backward until the first layer's weights are integrated out. This procedure allows us to evaluate important network properties, such as its generalization error, the role of network width and depth, the impact of the size of the training set, and the effects of weight regularization and learning stochasticity. BPKR does not assume specific statistics of the input or the task's output. Furthermore, by performing partial integration of the layers, the BPKR allows us to compute the properties of the neural representations across the different hidden layers. We have proposed an extension of the BPKR to nonlinear DNNs with ReLU. Surprisingly, our numerical simulations reveal that despite the nonlinearity, the predictions of our theory are largely shared by ReLU networks in a wide regime of parameters. Our work is the first exact statistical mechanical study of learning in a family of DNNs, and the first successful theory of learning through successive integration of DoFs in the learned weight space."
                },
                {
                    "title": "Stable ResNet",
                    "abstract": "Deep ResNet architectures have achieved state of the art performance on many tasks. While they solve the problem of gradient vanishing, they might suffer from gradient exploding as the depth becomes large (Yang et al. 2017). Moreover, recent results have shown that ResNet might lose expressivity as the depth goes to infinity (Yang et al. 2017, Hayou et al. 2019). To resolve these issues, we introduce a new class of ResNet architectures, called Stable ResNet, that have the property of stabilizing the gradient while ensuring expressivity in the infinite depth limit."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Improving Transformer Optimization Through Better Initialization",
                    "abstract": "The Transformer architecture has achieved considerable success recently; the key component of the Transformer is the attention layer that enables the model to focus on important regions within an input sequence. Gradient optimization with attention layers can be notoriously dif\ufb01cult requiring tricks such as learning rate warmup to prevent divergence. As Transformer models are becoming larger and more expensive to train, recent research has focused on understanding and improving optimization in these architectures. In this work our contributions are two-fold: we \ufb01rst investigate and empirically validate the source of optimization problems in the encoder-decoder Transformer architecture; we then propose a new weight initialization scheme with theoretical justi-\ufb01cation, that enables training without warmup or layer normalization. Empirical results on public machine translation benchmarks show that our approach achieves leading accuracy, allowing to train deep Transformer models with 200 layers in both encoder and decoder (over 1000 attention/MLP blocks) without dif\ufb01culty. Code for this work is available here: https://github. com/layer6ai-labs/T-Fixup ."
                },
                {
                    "title": "Tensor Programs II: Neural Tangent Kernel for Any Architecture",
                    "abstract": "We prove that a randomly initialized neural network of *any architecture* has its Tangent Kernel (NTK) converge to a deterministic limit, as the network widths tend to infinity. We demonstrate how to calculate this limit. In prior literature, the heuristic study of neural network gradients often assumes every weight matrix used in forward propagation is independent from its transpose used in backpropagation (Schoenholz et al. 2017). This is known as the *gradient independence assumption (GIA)*. We identify a commonly satisfied condition, which we call *Simple GIA Check*, such that the NTK limit calculation based on GIA is correct. Conversely, when Simple GIA Check fails, we show GIA can result in wrong answers. Our material here presents the NTK results of Yang (2019a) in a friendly manner and showcases the *tensor programs* technique for understanding wide neural networks. We provide reference implementations of infinite-width NTKs of recurrent neural network, transformer, and batch normalization at this https URL."
                },
                {
                    "title": "Dynamical mean-field theory for stochastic gradient descent in Gaussian mixture classification",
                    "abstract": "We analyze in a closed form the learning dynamics of the stochastic gradient descent (SGD) for a single-layer neural network classifying a high-dimensional Gaussian mixture where each cluster is assigned one of two labels. This problem provides a prototype of a non-convex loss landscape with interpolating regimes and a large generalization gap. We define a particular stochastic process for which SGD can be extended to a continuous-time limit that we call stochastic gradient flow. In the full-batch limit, we recover the standard gradient flow. We apply dynamical mean-field theory from statistical physics to track the dynamics of the algorithm in the high-dimensional limit via a self-consistent stochastic process. We explore the performance of the algorithm as a function of the control parameters shedding light on how it navigates the loss landscape."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "ReZero is All You Need: Fast Convergence at Large Depth",
                    "abstract": "Deep networks often suffer from vanishing or exploding gradients due to inefficient signal propagation, leading to long training times or convergence difficulties. Various architecture designs, sophisticated residual-style networks, and initialization schemes have been shown to improve deep signal propagation. Recently, Pennington et al. used free probability theory to show that dynamical isometry plays an integral role in efficient deep learning. We show that the simplest architecture change of gating each residual connection using a single zero-initialized parameter satisfies initial dynamical isometry and outperforms more complex approaches. Although much simpler than its predecessors, this gate enables training thousands of fully connected layers with fast convergence and better test performance for ResNets trained on CIFAR-10. We apply this technique to language modeling and find that we can easily train 120-layer Transformers. When applied to 12 layer Transformers, it converges 56% faster on enwiki8."
                },
                {
                    "title": "On Layer Normalization in the Transformer Architecture",
                    "abstract": "The Transformer is widely used in natural language processing tasks. To train a Transformer however, one usually needs a carefully designed learning rate warm-up stage, which is shown to be crucial to the final performance but will slow down the optimization and bring more hyperparameter tunings. In this paper, we first study theoretically why the learning rate warm-up stage is essential and show that the location of layer normalization matters. Specifically, we prove with mean field theory that at initialization, for the original-designed Post-LN Transformer, which places the layer normalization between the residual blocks, the expected gradients of the parameters near the output layer are large. Therefore, using a large learning rate on those gradients makes the training unstable. The warm-up stage is practically helpful for avoiding this problem. On the other hand, our theory also shows that if the layer normalization is put inside the residual blocks (recently proposed as Pre-LN Transformer), the gradients are well-behaved at initialization. This motivates us to remove the warm-up stage for the training of Pre-LN Transformers. We show in our experiments that Pre-LN Transformers without the warm-up stage can reach comparable results with baselines while requiring significantly less training time and hyper-parameter tuning on a wide range of applications."
                },
                {
                    "title": "Implicit Bias of Gradient Descent for Wide Two-layer Neural Networks Trained with the Logistic Loss",
                    "abstract": "Neural networks trained to minimize the logistic (a.k.a. cross-entropy) loss with gradient-based methods are observed to perform well in many supervised classification tasks. Towards understanding this phenomenon, we analyze the training and generalization behavior of infinitely wide two-layer neural networks with homogeneous activations. We show that the limits of the gradient flow on exponentially tailed losses can be fully characterized as a max-margin classifier in a certain non-Hilbertian space of functions. In presence of hidden low-dimensional structures, the resulting margin is independent of the ambiant dimension, which leads to strong generalization bounds. In contrast, training only the output layer implicitly solves a kernel support vector machine, which a priori does not enjoy such an adaptivity. Our analysis of training is non-quantitative in terms of running time but we prove computational guarantees in simplified settings by showing equivalences with online mirror descent. Finally, numerical experiments suggest that our analysis describes well the practical behavior of two-layer neural networks with ReLU activation and confirm the statistical benefits of this implicit bias."
                },
                {
                    "title": "On the distance between two neural networks and the stability of learning",
                    "abstract": "This paper relates parameter distance to gradient breakdown for a broad class of nonlinear compositional functions. The analysis leads to a new distance function called deep relative trust and a descent lemma for neural networks. Since the resulting learning rule seems not to require learning rate grid search, it may unlock a simpler workflow for training deeper and more complex neural networks. Please find the Python code used in this paper here: this https URL."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Non-Gaussian processes and neural networks at finite widths",
                    "abstract": "Gaussian processes are ubiquitous in nature and engineering. A case in point is a class of neural networks in the infinite-width limit, whose priors correspond to Gaussian processes. Here we perturbatively extend this correspondence to finite-width neural networks, yielding non-Gaussian processes as priors. The methodology developed herein allows us to track the flow of preactivation distributions by progressively integrating out random variables from lower to higher layers, reminiscent of renormalization-group flow. We further develop a perturbative procedure to perform Bayesian inference with weakly non-Gaussian priors."
                },
                {
                    "title": "Finite Depth and Width Corrections to the Neural Tangent Kernel",
                    "abstract": "We prove the precise scaling, at finite depth and width, for the mean and variance of the neural tangent kernel (NTK) in a randomly initialized ReLU network. The standard deviation is exponential in the ratio of network depth to width. Thus, even in the limit of infinite overparameterization, the NTK is not deterministic if depth and width simultaneously tend to infinity. Moreover, we prove that for such deep and wide networks, the NTK has a non-trivial evolution during training by showing that the mean of its first SGD update is also exponential in the ratio of network depth to width. This is sharp contrast to the regime where depth is fixed and network width is very large. Our results suggest that, unlike relatively shallow and wide networks, deep and wide ReLU networks are capable of learning data-dependent features even in the so-called lazy training regime."
                },
                {
                    "title": "Finite size corrections for neural network Gaussian processes",
                    "abstract": "There has been a recent surge of interest in modeling neural networks (NNs) as Gaussian processes. In the limit of a NN of infinite width the NN becomes equivalent to a Gaussian process. Here we demonstrate that for an ensemble of large, finite, fully connected networks with a single hidden layer the distribution of outputs at initialization is well described by a Gaussian perturbed by the fourth Hermite polynomial for weights drawn from a symmetric distribution. We show that the scale of the perturbation is inversely proportional to the number of units in the NN and that higher order terms decay more rapidly, thereby recovering the Edgeworth expansion. We conclude by observing that understanding how this perturbation changes under training would reveal the regimes in which the Gaussian process framework is valid to model NN behavior."
                },
                {
                    "title": "Wide neural networks of any depth evolve as linear models under gradient descent",
                    "abstract": "A longstanding goal in deep learning research has been to precisely characterize training and generalization. However, the often complex loss landscapes of neural networks (NNs) have made a theory of learning dynamics elusive. In this work, we show that for wide NNs the learning dynamics simplify considerably and that, in the infinite width limit, they are governed by a linear model obtained from the first-order Taylor expansion of the network around its initial parameters. Furthermore, mirroring the correspondence between wide Bayesian NNs and Gaussian processes (GPs), gradient-based training of wide NNs with a squared loss produces test set predictions drawn from a GP with a particular compositional kernel. While these theoretical results are only exact in the infinite width limit, we nevertheless find excellent empirical agreement between the predictions of the original network and those of the linearized version even for finite practically-sized networks. This agreement is robust across different architectures, optimization methods, and loss functions."
                },
                {
                    "title": "Mean-field theory of two-layers neural networks: dimension-free bounds and kernel limit",
                    "abstract": "We consider learning two layer neural networks using stochastic gradient descent. The mean-field description of this learning dynamics approximates the evolution of the network weights by an evolution in the space of probability distributions in $R^D$ (where $D$ is the number of parameters associated to each neuron). This evolution can be defined through a partial differential equation or, equivalently, as the gradient flow in the Wasserstein space of probability distributions. Earlier work shows that (under some regularity assumptions), the mean field description is accurate as soon as the number of hidden units is much larger than the dimension $D$. In this paper we establish stronger and more general approximation guarantees. First of all, we show that the number of hidden units only needs to be larger than a quantity dependent on the regularity properties of the data, and independent of the dimensions. Next, we generalize this analysis to the case of unbounded activation functions, which was not covered by earlier bounds. We extend our results to noisy stochastic gradient descent. \nFinally, we show that kernel ridge regression can be recovered as a special limit of the mean field analysis."
                },
                {
                    "title": "Fixup Initialization: Residual Learning Without Normalization",
                    "abstract": "Normalization layers are a staple in state-of-the-art deep neural network architectures. They are widely believed to stabilize training, enable higher learning rate, accelerate convergence and improve generalization, though the reason for their effectiveness is still an active research topic. In this work, we challenge the commonly-held beliefs by showing that none of the perceived benefits is unique to normalization. Specifically, we propose fixed-update initialization (Fixup), an initialization motivated by solving the exploding and vanishing gradient problem at the beginning of training via properly rescaling a standard initialization. We find training residual networks with Fixup to be as stable as training with normalization -- even for networks with 10,000 layers. Furthermore, with proper regularization, Fixup enables residual networks without normalization to achieve state-of-the-art performance in image classification and machine translation."
                },
                {
                    "title": "On Lazy Training in Differentiable Programming",
                    "abstract": "In a series of recent theoretical works, it was shown that strongly over-parameterized neural networks trained with gradient-based methods could converge exponentially fast to zero training loss, with their parameters hardly varying. In this work, we show that this \"lazy training\" phenomenon is not specific to over-parameterized neural networks, and is due to a choice of scaling, often implicit, that makes the model behave as its linearization around the initialization, thus yielding a model equivalent to learning with positive-definite kernels. Through a theoretical analysis, we exhibit various situations where this phenomenon arises in non-convex optimization and we provide bounds on the distance between the lazy and linearized optimization paths. Our numerical experiments bring a critical note, as we observe that the performance of commonly used non-linear deep convolutional neural networks in computer vision degrades when trained in the lazy regime. This makes it unlikely that \"lazy training\" is behind the many successes of neural networks in difficult high dimensional tasks."
                },
                {
                    "title": "Dynamical Isometry is Achieved in Residual Networks in a Universal Way for any Activation Function",
                    "abstract": "We demonstrate that in residual neural networks (ResNets) dynamical isometry is achievable irrespective of the activation function used. We do that by deriving, with the help of Free Probability and Random Matrix Theories, a universal formula for the spectral density of the input-output Jacobian at initialization, in the large network width and depth limit. The resulting singular value spectrum depends on a single parameter, which we calculate for a variety of popular activation functions, by analyzing the signal propagation in the artificial neural network. We corroborate our results with numerical simulations of both random matrices and ResNets applied to the CIFAR-10 classification problem. Moreover, we study consequences of this universal behavior for the initial and late phases of the learning processes. \nWe conclude by drawing attention to the simple fact, that initialization acts as a confounding factor between the choice of activation function and the rate of learning. We propose that in ResNets this can be resolved based on our results by ensuring the same level of dynamical isometry at initialization."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "On the Global Convergence of Gradient Descent for Over-parameterized Models using Optimal Transport",
                    "abstract": "Many tasks in machine learning and signal processing can be solved by minimizing a convex function of a measure. This includes sparse spikes deconvolution or training a neural network with a single hidden layer. For these problems, we study a simple minimization method: the unknown measure is discretized into a mixture of particles and a continuous-time gradient descent is performed on their weights and positions. This is an idealization of the usual way to train neural networks with a large hidden layer. We show that, when initialized correctly and in the many-particle limit, this gradient flow, although non-convex, converges to global minimizers. The proof involves Wasserstein gradient flows, a by-product of optimal transport theory. Numerical experiments show that this asymptotic behavior is already at play for a reasonable number of particles, even in high dimension."
                },
                {
                    "title": "Gaussian Process Behaviour in Wide Deep Neural Networks",
                    "abstract": "Whilst deep neural networks have shown great empirical success, there is still much work to be done to understand their theoretical properties. In this paper, we study the relationship between random, wide, fully connected, feedforward networks with more than one hidden layer and Gaussian processes with a recursive kernel definition. We show that, under broad conditions, as we make the architecture increasingly wide, the implied random function converges in distribution to a Gaussian process, formalising and extending existing results by Neal (1996) to deep networks. To evaluate convergence rates empirically, we use maximum mean discrepancy. We then compare finite Bayesian deep networks from the literature to Gaussian processes in terms of the key predictive quantities of interest, finding that in some cases the agreement can be very close. We discuss the desirability of Gaussian process behaviour and review non-Gaussian alternative models from the literature."
                },
                {
                    "title": "Which Neural Net Architectures Give Rise To Exploding and Vanishing Gradients?",
                    "abstract": "We give a rigorous analysis of the statistical behavior of gradients in randomly initialized feed-forward networks with ReLU activations. Our results show that a fully connected depth $d$ ReLU net with hidden layer widths $n_j$ will have exploding and vanishing gradients if and only if $\\sum_{j=1}^{d-1} 1/n_j$ is large. The point of view of this article is that whether a given neural net will have exploding/vanishing gradients is a function mainly of the architecture of the net, and hence can be tested at initialization. Our results imply that a fully connected network that produces manageable gradients at initialization must have many hidden layers that are about as wide as the network is deep. This work is related to the mean field theory approach to random neural nets. From this point of view, we give a rigorous computation of the $1/n_j$ corrections to the propagation of gradients at the so-called edge of chaos."
                },
                {
                    "title": "Deep Neural Networks as Gaussian Processes",
                    "abstract": "It has long been known that a single-layer fully-connected neural network with an i.i.d. prior over its parameters is equivalent to a Gaussian process (GP), in the limit of infinite network width. This correspondence enables exact Bayesian inference for infinite width neural networks on regression tasks by means of evaluating the corresponding GP. Recently, kernel functions which mimic multi-layer random neural networks have been developed, but only outside of a Bayesian framework. As such, previous work has not identified that these kernels can be used as covariance functions for GPs and allow fully Bayesian prediction with a deep neural network. \nIn this work, we derive the exact equivalence between infinitely wide deep networks and GPs. We further develop a computationally efficient pipeline to compute the covariance function for these GPs. We then use the resulting GPs to perform Bayesian inference for wide deep neural networks on MNIST and CIFAR-10. We observe that trained neural network accuracy approaches that of the corresponding GP with increasing layer width, and that the GP uncertainty is strongly correlated with trained network prediction error. We further find that test performance increases as finite-width trained networks are made wider and more similar to a GP, and thus that GP predictions typically outperform those of finite-width networks. Finally we connect the performance of these GPs to the recent theory of signal propagation in random neural networks."
                },
                {
                    "title": "Deep Residual Networks and Weight Initialization",
                    "abstract": "Residual Network (ResNet) is the state-of-the-art architecture that realizes successful training of really deep neural network. It is also known that good weight initialization of neural network avoids problem of vanishing/exploding gradients. In this paper, simplified models of ResNets are analyzed. We argue that goodness of ResNet is correlated with the fact that ResNets are relatively insensitive to choice of initial weights. We also demonstrate how batch normalization improves backpropagation of deep ResNets without tuning initial values of weights."
                },
                {
                    "title": "The Shattered Gradients Problem: If resnets are the answer, then what is the question?",
                    "abstract": "A long-standing obstacle to progress in deep learning is the problem of vanishing and exploding gradients. Although, the problem has largely been overcome via carefully constructed initializations and batch normalization, architectures incorporating skip-connections such as highway and resnets perform much better than standard feedforward architectures despite well-chosen initialization and batch normalization. In this paper, we identify the shattered gradients problem. Specifically, we show that the correlation between gradients in standard feedforward networks decays exponentially with depth resulting in gradients that resemble white noise whereas, in contrast, the gradients in architectures with skip-connections are far more resistant to shattering, decaying sublinearly. Detailed empirical evidence is presented in support of the analysis, on both fully-connected networks and convnets. Finally, we present a new \"looks linear\" (LL) initialization that prevents shattering, with preliminary experiments showing the new initialization allows to train very deep networks without the addition of skip-connections."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Kernel Methods for Deep Learning",
                    "abstract": "We introduce a new family of positive-definite kernel functions that mimic the computation in large, multilayer neural nets. These kernel functions can be used in shallow architectures, such as support vector machines (SVMs), or in deep kernel-based architectures that we call multilayer kernel machines (MKMs). We evaluate SVMs and MKMs with these kernel functions on problems designed to illustrate the advantages of deep architectures. On several problems, we obtain better results than previous, leading benchmarks from both SVMs with Gaussian kernels as well as deep belief nets."
                },
                {
                    "title": "Statistical Dynamics of Classical Systems",
                    "abstract": "The statistical dynamics of a classical random variable that satisfies a nonlinear equation of motion is recast in terms of closed self-consistent equations in which only the observable correlations at pairs of points and the exact response to infinitesimal disturbances appear. The self-consistent equations are developed by introducing a second field that does not commute with the random variable. Techniques used in the study of the interacting quantum fields can then be employed, and systematic approximations can be obtained. It is also possible to carry out a \"charge normalization\" eliminating the nonlinear coupling in favor of a dimensionless parameter which measures the deviation from Gaussian behavior. No assumptions of spatial or time homogeneity or of small deviation from equilibrium enter. It is shown that previously inferred renormalization schemes for homogeneous systems were incomplete or erroneous. The application of the method to classical microscopic systems, where it leads from first principles to a coupled-mode description is briefly indicated."
                },
                {
                    "title": "Tensor Programs IV: Feature Learning in Infinite-Width Neural Networks",
                    "abstract": "As its width tends to in\ufb01nity, a deep neural network\u2019s behavior under gradient descent can become simpli\ufb01ed and predictable (e.g. given by the Neural Tangent Kernel (NTK)), if it is parametrized appropriately (e.g. the NTK parametrization). However, we show that the standard and NTK parametrizations of a neural network do not admit in\ufb01nite-width limits that can learn features, which is crucial for pretraining and transfer learning such as with BERT. We propose simple modi\ufb01cations to the standard parametrization to allow for feature learning in the limit. Using the Tensor Programs technique, we derive explicit formulas for such limits. On Word2Vec and few-shot learning on Omniglot via MAML, two canonical tasks that rely crucially on feature learning, we compute these limits exactly. We \ufb01nd that they outperform both NTK baselines and \ufb01nite-width networks, with the latter approaching the in\ufb01nite-width feature learning performance as width increases. See arXiv:2011.14522 for the full version of this paper."
                },
                {
                    "title": "Supplemental Material for: \u201cExact marginal prior distributions of finite Bayesian neural networks\u201d",
                    "abstract": "1. For all authors... (a) Do the main claims made in the abstract and introduction accurately reflect the paper\u2019s contributions and scope? [Yes] (b) Did you describe the limitations of your work? [Yes] We discuss the limitations of our work in the Introduction, \u00a72, and in the Conclusion. As discussed therein, the primary limitations of our theoretical results are that they apply only to the prior over"
                },
                {
                    "title": "Supplemental Material for: \u201cAsymptotics of representation learning in finite Bayesian neural networks\u201d",
                    "abstract": "D Direct computation of the average hidden layer kernels of a deep linear MLP S11 D.1 The cumulant generating function of learned features for a MLP . . . . . . . . . . S11 D.2 General form of the perturbative layer integrals for a deep linear network . . . . . . S12 D.3 Perturbative computation of the partition function of a deep linear network . . . . . S16 D.4 Computing the average hidden layer kernels of a deep linear network . . . . . . . . S18"
                },
                {
                    "title": "Principles of mathematical analysis",
                    "abstract": "Chapter 1: The Real and Complex Number Systems Introduction Ordered Sets Fields The Real Field The Extended Real Number System The Complex Field Euclidean Spaces Appendix Exercises Chapter 2: Basic Topology Finite, Countable, and Uncountable Sets Metric Spaces Compact Sets Perfect Sets Connected Sets Exercises Chapter 3: Numerical Sequences and Series Convergent Sequences Subsequences Cauchy Sequences Upper and Lower Limits Some Special Sequences Series Series of Nonnegative Terms The Number e The Root and Ratio Tests Power Series Summation by Parts Absolute Convergence Addition and Multiplication of Series Rearrangements Exercises Chapter 4: Continuity Limits of Functions Continuous Functions Continuity and Compactness Continuity and Connectedness Discontinuities Monotonic Functions Infinite Limits and Limits at Infinity Exercises Chapter 5: Differentiation The Derivative of a Real Function Mean Value Theorems The Continuity of Derivatives L'Hospital's Rule Derivatives of Higher-Order Taylor's Theorem Differentiation of Vector-valued Functions Exercises Chapter 6: The Riemann-Stieltjes Integral Definition and Existence of the Integral Properties of the Integral Integration and Differentiation Integration of Vector-valued Functions Rectifiable Curves Exercises Chapter 7: Sequences and Series of Functions Discussion of Main Problem Uniform Convergence Uniform Convergence and Continuity Uniform Convergence and Integration Uniform Convergence and Differentiation Equicontinuous Families of Functions The Stone-Weierstrass Theorem Exercises Chapter 8: Some Special Functions Power Series The Exponential and Logarithmic Functions The Trigonometric Functions The Algebraic Completeness of the Complex Field Fourier Series The Gamma Function Exercises Chapter 9: Functions of Several Variables Linear Transformations Differentiation The Contraction Principle The Inverse Function Theorem The Implicit Function Theorem The Rank Theorem Determinants Derivatives of Higher Order Differentiation of Integrals Exercises Chapter 10: Integration of Differential Forms Integration Primitive Mappings Partitions of Unity Change of Variables Differential Forms Simplexes and Chains Stokes' Theorem Closed Forms and Exact Forms Vector Analysis Exercises Chapter 11: The Lebesgue Theory Set Functions Construction of the Lebesgue Measure Measure Spaces Measurable Functions Simple Functions Integration Comparison with the Riemann Integral Integration of Complex Functions Functions of Class L2 Exercises Bibliography List of Special Symbols Index"
                }
            ],
            "categories": [
                "stat.ML",
                "cond-mat.dis-nn",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan we transfer hyperparameters simultaneously across depth and width in neural networks, specifically in residual architectures like ResNets and transformers?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the inefficiencies in hyperparameter tuning for large-scale neural networks, which often require extensive computational resources. By enabling the transfer of hyperparameters across both depth and width, this research could significantly reduce the time and cost associated with training deep learning models. This advancement could lead to more efficient model training practices, fostering innovation in various applications such as computer vision and natural language processing. Furthermore, it could inspire future research into more generalized hyperparameter transfer techniques, enhancing the scalability and performance of deep learning models.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of neural network architectures, where the interaction between depth and width can lead to non-linear behaviors that are difficult to predict. Naive approaches may fail because they do not account for the unique scaling properties of residual connections, which can vary significantly across different architectures. Additionally, the lack of established principles for hyperparameter transfer across these dimensions complicates the task. Technical obstacles include the need for empirical validation across diverse datasets and architectures, as well as the potential for divergence in training when inappropriate hyperparameters are applied.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research, such as the \u03bcP parameterization, primarily focused on transferring hyperparameters from narrower to wider models but did not adequately address the simultaneous transfer across depth. This limitation stems from a lack of understanding of how depth influences learning dynamics in residual networks. Barriers include the complexity of existing models and the computational costs associated with extensive experimentation. Our approach improves upon prior work by introducing a specific parameterization that allows for effective transfer across both dimensions, supported by empirical evidence demonstrating its efficacy in various architectures.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a novel parameterization of residual architectures that scales residual branches as \\(1/\\sqrt{\\text{depth}}\\). We will conduct experiments using datasets such as CIFAR-10, Tiny ImageNet, and ImageNet, focusing on metrics like training loss and transferability of learning rates. Expected outcomes include demonstrating successful transfer of learning rates and other hyperparameters (e.g., momentum coefficients, regularization strengths)"
            }
        },
        "author_data": {
            "c262e749-e829-42a4-a250-29694545c9ab": {
                "pk": "c262e749-e829-42a4-a250-29694545c9ab",
                "name": "Blake Bordelon",
                "collaborators": [
                    "C. Pehlevan",
                    "Alexander Atanasov",
                    "P. Masset",
                    "Henry Kuo",
                    "Sabarish Sainathan",
                    "H. Shan",
                    "Alexander B. Atanasov",
                    "Hamza Tahir Chaudhry",
                    "Tanishq Kumar",
                    "S. Gershman"
                ],
                "domain": [
                    "Neural Networks",
                    "Reinforcement Learning",
                    "Statistical Physics",
                    "Machine Learning Theory"
                ],
                "publications": [
                    {
                        "title": "How Feature Learning Can Improve Neural Scaling Laws",
                        "abstract": "We develop a solvable model of neural scaling laws beyond the kernel limit. Theoretical analysis of this model shows how performance scales with model size, training time, and the total amount of available data. We identify three scaling regimes corresponding to varying task difficulties: hard, easy, and super easy tasks. For easy and super-easy target functions, which lie in the reproducing kernel Hilbert space (RKHS) defined by the initial infinite-width Neural Tangent Kernel (NTK), the scaling exponents remain unchanged between feature learning and kernel regime models. For hard tasks, defined as those outside the RKHS of the initial NTK, we demonstrate both analytically and empirically that feature learning can improve scaling with training time and compute, nearly doubling the exponent for hard tasks. This leads to a different compute optimal strategy to scale parameters and training time in the feature learning regime. We support our finding that feature learning improves the scaling law for hard tasks but not for easy and super-easy tasks with experiments of nonlinear MLPs fitting functions with power-law Fourier spectra on the circle and CNNs learning vision tasks."
                    },
                    {
                        "title": "Infinite Limits of Multi-head Transformer Dynamics",
                        "abstract": "In this work, we analyze various scaling limits of the training dynamics of transformer models in the feature learning regime. We identify the set of parameterizations that admit well-defined infinite width and depth limits, allowing the attention layers to update throughout training--a relevant notion of feature learning in these models. We then use tools from dynamical mean field theory (DMFT) to analyze various infinite limits (infinite key/query dimension, infinite heads, and infinite depth) which have different statistical descriptions depending on which infinite limit is taken and how attention layers are scaled. We provide numerical evidence of convergence to the limits and discuss how the parameterization qualitatively influences learned features."
                    },
                    {
                        "title": "A Dynamical Model of Neural Scaling Laws",
                        "abstract": "On a variety of tasks, the performance of neural networks predictably improves with training time, dataset size and model size across many orders of magnitude. This phenomenon is known as a neural scaling law. Of fundamental importance is the compute-optimal scaling law, which reports the performance as a function of units of compute when choosing model sizes optimally. We analyze a random feature model trained with gradient descent as a solvable model of network training and generalization. This reproduces many observations about neural scaling laws. First, our model makes a prediction about why the scaling of performance with training time and with model size have different power law exponents. Consequently, the theory predicts an asymmetric compute-optimal scaling rule where the number of training steps are increased faster than model parameters, consistent with recent empirical observations. Second, it has been observed that early in training, networks converge to their infinite-width dynamics at a rate $1/\\textit{width}$ but at late time exhibit a rate $\\textit{width}^{-c}$, where $c$ depends on the structure of the architecture and task. We show that our model exhibits this behavior. Lastly, our theory shows how the gap between training and test loss can gradually build up over time due to repeated reuse of data."
                    },
                    {
                        "title": "Dynamics of Temporal Difference Reinforcement Learning",
                        "abstract": "Reinforcement learning has been successful across several applications in which agents have to learn to act in environments with sparse feedback. However, despite this empirical success there is still a lack of theoretical understanding of how the parameters of reinforcement learning models and the features used to represent states interact to control the dynamics of learning. In this work, we use concepts from statistical physics, to study the typical case learning curves for temporal difference learning of a value function with linear function approximators. Our theory is derived under a Gaussian equivalence hypothesis where averages over the random trajectories are replaced with temporally correlated Gaussian feature averages and we validate our assumptions on small scale Markov Decision Processes. We find that the stochastic semi-gradient noise due to subsampling the space of possible episodes leads to significant plateaus in the value error, unlike in traditional gradient descent dynamics. We study how learning dynamics and plateaus depend on feature structure, learning rate, discount factor, and reward function. We then analyze how strategies like learning rate annealing and reward shaping can favorably alter learning dynamics and plateaus. To conclude, our work introduces new tools to open a new direction towards developing a theory of learning dynamics in reinforcement learning."
                    },
                    {
                        "title": "Loss Dynamics of Temporal Difference Reinforcement Learning",
                        "abstract": "Reinforcement learning has been successful across several applications in which agents have to learn to act in environments with sparse feedback. However, despite this empirical success there is still a lack of theoretical understanding of how the parameters of reinforcement learning models and the features used to represent states interact to control the dynamics of learning. In this work, we use concepts from statistical physics, to study the typical case learning curves for temporal difference learning of a value function with linear function approximators. Our theory is derived under a Gaussian equivalence hypothesis where averages over the random trajectories are replaced with temporally correlated Gaussian feature averages and we validate our assumptions on small scale Markov Decision Processes. We find that the stochastic semi-gradient noise due to subsampling the space of possible episodes leads to significant plateaus in the value error, unlike in traditional gradient descent dynamics. We study how learning dynamics and plateaus depend on feature structure, learning rate, discount factor, and reward function. We then analyze how strategies like learning rate annealing and reward shaping can favorably alter learning dynamics and plateaus. To conclude, our work introduces new tools to open a new direction towards developing a theory of learning dynamics in reinforcement learning."
                    },
                    {
                        "title": "Grokking as the Transition from Lazy to Rich Training Dynamics",
                        "abstract": "We propose that the grokking phenomenon, where the train loss of a neural network decreases much earlier than its test loss, can arise due to a neural network transitioning from lazy training dynamics to a rich, feature learning regime. To illustrate this mechanism, we study the simple setting of vanilla gradient descent on a polynomial regression problem with a two layer neural network which exhibits grokking without regularization in a way that cannot be explained by existing theories. We identify sufficient statistics for the test loss of such a network, and tracking these over training reveals that grokking arises in this setting when the network first attempts to fit a kernel regression solution with its initial features, followed by late-time feature learning where a generalizing solution is identified after train loss is already low. We find that the key determinants of grokking are the rate of feature learning -- which can be controlled precisely by parameters that scale the network output -- and the alignment of the initial features with the target function $y(x)$. We argue this delayed generalization arises when (1) the top eigenvectors of the initial neural tangent kernel and the task labels $y(x)$ are misaligned, but (2) the dataset size is large enough so that it is possible for the network to generalize eventually, but not so large that train loss perfectly tracks test loss at all epochs, and (3) the network begins training in the lazy regime so does not learn features immediately. We conclude with evidence that this transition from lazy (linear model) to rich training (feature learning) can control grokking in more general settings, like on MNIST, one-layer Transformers, and student-teacher networks."
                    },
                    {
                        "title": "Feature-Learning Networks Are Consistent Across Widths At Realistic Scales",
                        "abstract": "We study the effect of width on the dynamics of feature-learning neural networks across a variety of architectures and datasets. Early in training, wide neural networks trained on online data have not only identical loss curves but also agree in their point-wise test predictions throughout training. For simple tasks such as CIFAR-5m this holds throughout training for networks of realistic widths. We also show that structural properties of the models, including internal representations, preactivation distributions, edge of stability phenomena, and large learning rate effects are consistent across large widths. This motivates the hypothesis that phenomena seen in realistic models can be captured by infinite-width, feature-learning limits. For harder tasks (such as ImageNet and language modeling), and later training times, finite-width deviations grow systematically. Two distinct effects cause these deviations across widths. First, the network output has initialization-dependent variance scaling inversely with width, which can be removed by ensembling networks. We observe, however, that ensembles of narrower networks perform worse than a single wide network. We call this the bias of narrower width. We conclude with a spectral perspective on the origin of this finite-width bias."
                    },
                    {
                        "title": "Dynamics of Finite Width Kernel and Prediction Fluctuations in Mean Field Neural Networks",
                        "abstract": "  We analyze the dynamics of finite width effects in wide but finite feature learning neural networks. Starting from a dynamical mean field theory description of infinite width deep neural network kernel and prediction dynamics, we provide a characterization of the           O    (  1    /      width    )      fluctuations of the dynamical mean field theory order parameters over random initializations of the network weights. Our results, while perturbative in width, unlike prior analyses, are non-perturbative in the strength of feature learning. We find that once the mean field/\u00b5P parameterization is adopted, the leading finite size effect on the dynamics is to introduce initialization variance in the predictions and feature kernels of the networks. In the lazy limit of network training, all kernels are random but static in time and the prediction variance has a universal form. However, in the rich, feature learning regime, the fluctuations of the kernels and predictions are dynamically coupled with a variance that can be computed self-consistently. In two layer networks, we show how feature learning can dynamically reduce the variance of the final tangent kernel and final network predictions. We also show how initialization variance can slow down online learning in wide but finite networks. In deeper networks, kernel variance can dramatically accumulate through subsequent layers at large feature learning strengths, but feature learning continues to improve the signal-to-noise ratio of the feature kernels. In discrete time, we demonstrate that large learning rate phenomena such as edge of stability effects can be well captured by infinite width dynamics and that initialization variance can decrease dynamically. For convolutional neural networks trained on CIFAR-10, we empirically find significant corrections to both the bias and variance of network dynamics due to finite width."
                    },
                    {
                        "title": "The Influence of Learning Rule on Representation Dynamics in Wide Neural Networks",
                        "abstract": "It is unclear how changing the learning rule of a deep neural network alters its learning dynamics and representations. To gain insight into the relationship between learned features, function approximation, and the learning rule, we analyze infinite-width deep networks trained with gradient descent (GD) and biologically-plausible alternatives including feedback alignment (FA), direct feedback alignment (DFA), and error modulated Hebbian learning (Hebb), as well as gated linear networks (GLN). We show that, for each of these learning rules, the evolution of the output function at infinite width is governed by a time varying effective neural tangent kernel (eNTK). In the lazy training limit, this eNTK is static and does not evolve, while in the rich mean-field regime this kernel's evolution can be determined self-consistently with dynamical mean field theory (DMFT). This DMFT enables comparisons of the feature and prediction dynamics induced by each of these learning rules. In the lazy limit, we find that DFA and Hebb can only learn using the last layer features, while full FA can utilize earlier layers with a scale determined by the initial correlation between feedforward and feedback weight matrices. In the rich regime, DFA and FA utilize a temporally evolving and depth-dependent NTK. Counterintuitively, we find that FA networks trained in the rich regime exhibit more feature learning if initialized with smaller correlation between the forward and backward pass weights. GLNs admit a very simple formula for their lazy limit kernel and preserve conditional Gaussianity of their preactivations under gating functions. Error modulated Hebb rules show very small task-relevant alignment of their kernels and perform most task relevant learning in the last layer."
                    },
                    {
                        "title": "Self-consistent dynamical field theory of kernel evolution in wide neural networks",
                        "abstract": "We analyze feature learning in infinite-width neural networks trained with gradient flow through a self-consistent dynamical field theory. We construct a collection of deterministic dynamical order parameters which are inner-product kernels for hidden unit activations and gradients in each layer at pairs of time points, providing a reduced description of network activity through training. These kernel order parameters collectively define the hidden layer activation distribution, the evolution of the neural tangent kernel (NTK), and consequently, output predictions. We show that the field theory derivation recovers the recursive stochastic process of infinite-width feature learning networks obtained by Yang and Hu with tensor programs. For deep linear networks, these kernels satisfy a set of algebraic matrix equations. For nonlinear networks, we provide an alternating sampling procedure to self-consistently solve for the kernel order parameters. We provide comparisons of the self-consistent solution to various approximation schemes including the static NTK approximation, gradient independence assumption, and leading order perturbation theory, showing that each of these approximations can break down in regimes where general self-consistent solutions still provide an accurate description. Lastly, we provide experiments in more realistic settings which demonstrate that the loss and kernel dynamics of convolutional neural networks at fixed feature learning strength are preserved across different widths on a image classification task."
                    },
                    {
                        "title": "The Onset of Variance-Limited Behavior for Networks in the Lazy and Rich Regimes",
                        "abstract": "For small training set sizes $P$, the generalization error of wide neural networks is well-approximated by the error of an infinite width neural network (NN), either in the kernel or mean-field/feature-learning regime. However, after a critical sample size $P^*$, we empirically find the finite-width network generalization becomes worse than that of the infinite width network. In this work, we empirically study the transition from infinite-width behavior to this variance limited regime as a function of sample size $P$ and network width $N$. We find that finite-size effects can become relevant for very small dataset sizes on the order of $P^* \\sim \\sqrt{N}$ for polynomial regression with ReLU networks. We discuss the source of these effects using an argument based on the variance of the NN's final neural tangent kernel (NTK). This transition can be pushed to larger $P$ by enhancing feature learning or by ensemble averaging the networks. We find that the learning curve for regression with the final NTK is an accurate approximation of the NN learning curve. Using this, we provide a toy model which also exhibits $P^* \\sim \\sqrt{N}$ scaling and has $P$-dependent benefits from feature learning."
                    },
                    {
                        "title": "Learning Curves for SGD on Structured Features",
                        "abstract": "The generalization performance of a machine learning algorithm such as a neural network depends in a non-trivial way on the structure of the data distribution. To analyze the influence of data structure on test loss dynamics, we study an exactly solveable model of stochastic gradient descent (SGD) which predicts test loss when training on features with arbitrary covariance structure. We solve the theory exactly for both Gaussian features and arbitrary features and we show that the simpler Gaussian model accurately predicts test loss of nonlinear random-feature models and deep neural networks trained with SGD on real datasets such as MNIST and CIFAR-10. We show that the optimal batch size at a fixed compute budget is typically small and depends on the feature correlation structure, demonstrating the computational benefits of SGD with small batch sizes. Lastly, we extend our theory to the more usual setting of stochastic gradient descent on a fixed subsampled training set, showing that both training and test error can be accurately predicted in our framework on real data."
                    },
                    {
                        "title": "Population codes enable learning from few examples by shaping inductive bias",
                        "abstract": "Learning from a limited number of experiences requires suitable inductive biases. To identify how inductive biases are implemented in and shaped by neural codes, we analyze sample-efficient learning of arbitrary stimulus-response maps from arbitrary neural codes with biologically-plausible readouts. We develop an analytical theory that predicts the generalization error of the readout as a function of the number of observed examples. Our theory illustrates in a mathematically precise way how the structure of population codes shapes inductive bias, and how a match between the code and the task is crucial for sample-efficient learning. We observe that many different codes can support the same inductive bias. By analyzing recordings from the mouse primary visual cortex, we demonstrate that biological codes have lower total activity than other codes with identical bias. Using these mouse primary visual cortex responses, we demonstrate the existence of an efficiency bias towards low frequency orientation discrimination tasks for grating stimuli and low spatial frequency reconstruction tasks for natural images. We reproduce the discrimination bias in a simple model of primary visual cortex, and further show how invariances in the code to certain stimulus variations alter learning performance. We extend our methods to time-dependent neural codes and predict the sample efficiency of readouts from recurrent networks. Finally, we discuss implications of our theory in the context of recent developments in neuroscience and artificial intelligence. Overall, our study provides a concrete method for elucidating inductive biases of the brain and promotes sample-efficient learning as a general normative coding principle."
                    },
                    {
                        "title": "Population codes enable learning",
                        "abstract": "Learning from a limited number of experiences requires suitable inductive biases. To 8 identify how inductive biases are implemented in and shaped by neural codes, we analyze 9 sample-efficient learning of arbitrary stimulus-response maps from arbitrary neural codes with 10 biologically-plausible readouts. We develop an analytical theory that predicts the generalization 11 error of the readout as a function of the number of observed examples. Our theory illustrates in 12 a mathematically precise way how the structure of population codes shapes inductive bias, and 13 how a match between the code and the task is crucial for sample-efficient learning. It elucidates a 14 bias to explain observed data with simple stimulus-response maps. Using recordings from the 15 mouse primary visual cortex, we demonstrate the existence of an efficiency bias towards low 16 frequency orientation discrimination tasks for grating stimuli and low spatial frequency 17 reconstruction tasks for natural images. We reproduce the discrimination bias in a simple model 18 of primary visual cortex, and further show how invariances in the code to certain stimulus 19 variations alter learning performance. We extend our methods to time-dependent neural codes 20 and predict the sample efficiency of readouts from recurrent networks. We observe that many 21 different codes can support the same inductive bias. By analyzing recordings from the mouse 22 primary visual cortex, we demonstrate that biological codes have lower total activity than other 23 codes with identical bias. Finally, we discuss implications of our theory in the context of recent 24 developments in neuroscience and artificial intelligence. Overall, our study provides a concrete 25 method for elucidating inductive biases of the brain and promotes sample-efficient learning as a 26 general normative coding principle. 27"
                    },
                    {
                        "title": "Dynamical Mean Field Theory of Kernel Evolution in Wide Neural Networks",
                        "abstract": "We analyze feature learning in in\ufb01nite-width neural networks trained with gradient \ufb02ow through a self-consistent dynamical \ufb01eld theory. We construct a collection of deterministic dynamical order parameters which are inner-product kernels for hidden unit activations and gradients in each layer at pairs of time points, providing a reduced description of network activity through training. These kernel order parameters collectively de\ufb01ne the hidden layer activation distribution, the evolution of the neural tangent kernel, and consequently output predictions. We provide a sampling procedure to self-consistently solve for the kernel order parameters"
                    },
                    {
                        "title": "Rapid Feature Evolution Accelerates Learning in Neural Networks",
                        "abstract": "Neural network (NN) training and generalization in the in\ufb01nite-width limit are well-characterized by kernel methods with a neural tangent kernel (NTK) that is stationary in time. However, \ufb01nite-width NNs consistently outperform corresponding kernel methods, suggesting the importance of feature learning, which manifests as the time evolution of NTKs. Here, we analyze the phenomenon of kernel alignment of the NTK with the target functions during gradient descent. We \ufb01rst provide a mechanistic explanation for why alignment between task and kernel occurs in deep linear networks. We then show that this behavior occurs more generally if one optimizes the feature map over time to accelerate learning while constraining how quickly the features evolve. Empirically, gradient descent undergoes a feature learning phase, during which top eigenfunctions of the NTK quickly align with the target function and the loss decreases faster than power law in time; it then enters a kernel gradient descent (KGD) phase where the alignment does not improve signi\ufb01cantly and the training loss decreases in power law. We show that feature evolution is faster and more dramatic in deeper networks. We also found that networks with multiple output nodes develop separate, specialized kernels for each output channel, a phenomenon we termed kernel specialization. We show that this class-speci\ufb01c alignment does not occur in linear networks."
                    },
                    {
                        "title": "Efficient online inference for nonparametric mixture models",
                        "abstract": "Natural data are often well-described as belonging to latent clusters. When the number of clusters is unknown, Bayesian nonparametric (BNP) models can provide a \ufb02exible and powerful technique to model the data. However, algorithms for inference in nonparametric mixture models fail to meet two critical requirements for practical use: (1) that inference can be performed online, and (2) that inference is ef\ufb01cient in the large time/sample limit. In this work, we propose a novel Bayesian recursion to ef\ufb01ciently infer a posterior distribution over discrete latent variables from a sequence of observations in an online manner, assuming a Chinese Restaurant Process prior on the sequence of latent variables. Our recursive \ufb01lter, which we call the Recursive Chinese Restaurant Process (R-CRP), has quasilinear average time complexity and logarithmic average space complexity in the total number of observations. We experimentally compare our \ufb01ltering method against both online and of-\ufb02ine inference algorithms including Markov chain Monte Carlo, variational approximations and DP-Means, and demonstrate that our inference algo-rithm achieves comparable or better performance for a fraction of the runtime."
                    },
                    {
                        "title": "Replica Method for the Machine Learning Theorist",
                        "abstract": "Blake Bordelon, Haozhe Shan, Abdul Canatar, Boaz Barak, Cengiz Pehlevan [Boaz\u2019s note: Blake and Haozhe were students in the ML theory seminar; in that seminar we touched on the replica method in the lecture on inference and statistical physics but here Blake and Haozhe (with a little help from the rest of us) give a great overview of the method and its relations to ML. See also all seminar posts.]"
                    }
                ]
            },
            "4da572d6-a29c-4c4d-b8ce-9f606d061095": {
                "pk": "4da572d6-a29c-4c4d-b8ce-9f606d061095",
                "name": "Lorenzo Noci",
                "collaborators": [
                    "Thomas Hofmann",
                    "Gregor Bachmann",
                    "Antonio Orvieto",
                    "Sotiris Anagnostidis",
                    "Bobby He",
                    "Sang-Ho Kim",
                    "Luca Biggio",
                    "Aur\u00e9lien Lucchi",
                    "Kevin Roth",
                    "Sebastian Nowozin"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Bayesian Inference",
                    "Knowledge Graphs"
                ],
                "publications": [
                    {
                        "title": "How Good is a Single Basin?",
                        "abstract": "The multi-modal nature of neural loss landscapes is often considered to be the main driver behind the empirical success of deep ensembles. In this work, we probe this belief by constructing various\"connected\"ensembles which are restricted to lie in the same basin. Through our experiments, we demonstrate that increased connectivity indeed negatively impacts performance. However, when incorporating the knowledge from other basins implicitly through distillation, we show that the gap in performance can be mitigated by re-discovering (multi-basin) deep ensembles within a single basin. Thus, we conjecture that while the extra-basin knowledge is at least partially present in any given basin, it cannot be easily harnessed without learning it from other basins."
                    },
                    {
                        "title": "Understanding and Minimising Outlier Features in Neural Network Training",
                        "abstract": "Outlier Features (OF) are neurons whose activation magnitudes significantly exceed the average over a neural network's (NN) width. They are well known to emerge during standard transformer training and have the undesirable effect of hindering quantisation in afflicted models. Despite their practical importance, little is known behind why OFs emerge during training, nor how one can minimise them. Our work focuses on the above questions, first identifying several quantitative metrics, such as the kurtosis over neuron activation norms, to measure OFs. With these metrics, we study how architectural and optimisation choices influence OFs, and provide practical insights to minimise OFs during training. As highlights, we emphasise the importance of controlling signal propagation throughout training, and propose the Outlier Protected transformer block, which removes standard Pre-Norm layers to mitigate OFs, without loss of convergence speed or training stability. Overall, our findings shed new light on our understanding of, our ability to prevent, and the complexity of this important facet in NN training dynamics."
                    },
                    {
                        "title": "Why do Learning Rates Transfer? Reconciling Optimization and Scaling Limits for Deep Learning",
                        "abstract": "Recently, there has been growing evidence that if the width and depth of a neural network are scaled toward the so-called rich feature learning limit ($\\mu$P and its depth extension), then some hyperparameters - such as the learning rate - exhibit transfer from small to very large models, thus reducing the cost of hyperparameter tuning. From an optimization perspective, this phenomenon is puzzling, as it implies that the loss landscape is remarkably consistent across very different model sizes. In this work, we find empirical evidence that learning rate transfer can be attributed to the fact that under $\\mu$P and its depth extension, the largest eigenvalue of the training loss Hessian (i.e. the sharpness) is largely independent of the width and depth of the network for a sustained period of training time. On the other hand, we show that under the neural tangent kernel (NTK) regime, the sharpness exhibits very different dynamics at different scales, thus preventing learning rate transfer. But what causes these differences in the sharpness dynamics? Through a connection between the spectra of the Hessian and the NTK matrix, we argue that the cause lies in the presence (for $\\mu$P) or progressive absence (for the NTK regime) of feature learning, which results in a different evolution of the NTK, and thus of the sharpness. We corroborate our claims with a substantial suite of experiments, covering a wide range of datasets and architectures: from ResNets and Vision Transformers trained on benchmark vision datasets to Transformers-based language models trained on WikiText"
                    },
                    {
                        "title": "Disentangling Linear Mode-Connectivity",
                        "abstract": "Linear mode-connectivity (LMC) (or lack thereof) is one of the intriguing characteristics of neural network loss landscapes. While empirically well established, it unfortunately still lacks a proper theoretical understanding. Even worse, although empirical data points are abound, a systematic study of when networks exhibit LMC is largely missing in the literature. In this work we aim to close this gap. We explore how LMC is affected by three factors: (1) architecture (sparsity, weight-sharing), (2) training strategy (optimization setup) as well as (3) the underlying dataset. We place particular emphasis on minimal but non-trivial settings, removing as much unnecessary complexity as possible. We believe that our insights can guide future theoretical works on uncovering the inner workings of LMC."
                    },
                    {
                        "title": "Achieving a Better Stability-Plasticity Trade-off via Auxiliary Networks in Continual Learning (Appendix)",
                        "abstract": "In the last few years, several papers [14, 15, 22, 24] have proposed to use an auxiliary network or an extra module which is solely trained on current dataset. They tried to combine this additional structure with a previous network or module which has been trained continually on old datasets. Elastic Weight Consolidation (EWC) [12] is one of the initial works among weight regularization methods that regularize weights related to the previous tasks. After training each task, EWC copies and freezes the current network as an old network. Then, it estimates the importance of each parameter in the old network so that the weights with the high importance remain unchanged when the model is optimized on future task. Based on the EWC, [22] proposes Active Forgetting with synaptic Expansion-Convergence (AFEC) which further regularizes the weights relevant to the current task through a new set of parameters called expanded parameters. The expanded parameters are solely trained on the dataset of the new task initialized by the old network and are allowed to forget the previous tasks. As a result, AFEC can reduce potential negative transfer in continual learning by selectively merging the old parameters with the expanded parameters. The stability-plasticity balance in AFEC is adjusted via hyperparameters which scale the regularization terms for remembering the old tasks and learning the new tasks. Learning without Forgetting (LwF) [13] prevents forgetting using knowledge distillation [2, 9]. They apply a distillation loss so that the model can learn soft targets generated by the old model instead of typical one-hot targets. The old model is copied and freezed before the training of the current task like EWC. Deep Model Consolidation (DMC) [24] is another distillation method built upon LwF. DMC proposes double distillation loss where the soft targets are generated using both the old model and a new model. The new model in DMC is basically the same concept as the expanded parameters in AFEC which is optimized on the current task. Then, the logits of new classes from the new model and the logits of old classes from the old model are concatenated to build the final soft targets. Then, the stability-plasticity balance is achieved by penalizing the model to generate the same output as the old model for old classes and the same output as the new model for new classes. Adaptive Aggregation Networks (AANet) [15] explicitly expands ResNet [8] to have the two types of residual blocks at each residual level: the one for retaining old knowledge and the other for learning new knowledge. The outputs from the two residual blocks are linearly combined by aggregation weights and then proceeded to the next-level layer. They train AANets through bilevel optimization. In the first level, the parameters of the two residual blocks are trained. In the second level, the aggregation weights are adapted which decide the balance between stability and plasticity within the ResNet. Recent work by [19] observes that multitask and continual solutions are connected by very simple curves with a low error in weight space, which is called Linear Mode Connectivity. They empirically prove that this connectivity is a linear path if the multitask learning and the continual learning share same initialization weights. Based on this observation, [14] proposes a simple linear connector that linearly add the weights of two networks following this linear path to emulate the multitask solution: the one model remembering the old tasks and the other model learning the new tasks. The stability-plasticity balance is preserved by combining the two networks. The above methods [14, 15, 22, 24] all share the property that the auxiliary model or module is used to solve the stabilityplasticity dilemma in continual learning. Consequently, these methods are able to learn the current task better than the original method while still retaining the knowledge of the previous tasks. However, the underlying mechanism of the interaction between the previous model and the auxiliary model is not widely studied. Therefore, in this work, we first formalize the framework of continual learning that adopts the auxiliary network called Auxiliary Network Continual Learning (ANCL)."
                    },
                    {
                        "title": "Achieving a Better Stability-Plasticity Trade-off via Auxiliary Networks in Continual Learning",
                        "abstract": "In contrast to the natural capabilities of humans to learn new tasks in a sequential fashion, neural networks are known to suffer from catastrophic forgetting, where the model's performances on old tasks drop dramatically after being optimized for a new task. Since then, the continual learning (CL) community has proposed several solutions aiming to equip the neural network with the ability to learn the current task (plasticity) while still achieving high accuracy on the previous tasks (stability). Despite remarkable improvements, the plasticity-stability trade-off is still far from being solved and its underlying mechanism is poorly understood. In this work, we propose Auxiliary Network Continual Learning (ANCL), a novel method that applies an additional auxiliary network which promotes plasticity to the continually learned model which mainly focuses on stability. More concretely, the proposed framework materializes in a regularizer that naturally interpolates between plasticity and stability, surpassing strong baselines on task incremental and class incremental scenarios. Through extensive analyses on ANCL solutions, we identify some essential principles beneath the stability-plasticity tradeoff. The code implementation of our work is available at https://github.com/kim-sanghwan/ANCL."
                    },
                    {
                        "title": "The Shaped Transformer: Attention Models in the Infinite Depth-and-Width Limit",
                        "abstract": "In deep learning theory, the covariance matrix of the representations serves as a proxy to examine the network's trainability. Motivated by the success of Transformers, we study the covariance matrix of a modified Softmax-based attention model with skip connections in the proportional limit of infinite-depth-and-width. We show that at initialization the limiting distribution can be described by a stochastic differential equation (SDE) indexed by the depth-to-width ratio. To achieve a well-defined stochastic limit, the Transformer's attention mechanism is modified by centering the Softmax output at identity, and scaling the Softmax logits by a width-dependent temperature parameter. We examine the stability of the network through the corresponding SDE, showing how the scale of both the drift and diffusion can be elegantly controlled with the aid of residual connections. The existence of a stable SDE implies that the covariance structure is well-behaved, even for very large depth and width, thus preventing the notorious issues of rank degeneracy in deep attention models. Finally, we show, through simulations, that the SDE provides a surprisingly good description of the corresponding finite-size model. We coin the name shaped Transformer for these architectural modifications."
                    },
                    {
                        "title": "Dynamic Context Pruning for Efficient and Interpretable Autoregressive Transformers",
                        "abstract": "Autoregressive Transformers adopted in Large Language Models (LLMs) are hard to scale to long sequences. Despite several works trying to reduce their computational cost, most of LLMs still adopt attention layers between all pairs of tokens in the sequence, thus incurring a quadratic cost. In this study, we present a novel approach that dynamically prunes contextual information while preserving the model's expressiveness, resulting in reduced memory and computational requirements during inference. Our method employs a learnable mechanism that determines which uninformative tokens can be dropped from the context at any point across the generation process. By doing so, our approach not only addresses performance concerns but also enhances interpretability, providing valuable insight into the model's decision-making process. Our technique can be applied to existing pre-trained models through a straightforward fine-tuning process, and the pruning strength can be specified by a sparsity parameter. Notably, our empirical findings demonstrate that we can effectively prune up to 80\\% of the context without significant performance degradation on downstream tasks, offering a valuable tool for mitigating inference costs. Our reference implementation achieves up to $2\\times$ increase in inference throughput and even greater memory savings."
                    },
                    {
                        "title": "Signal Propagation in Transformers: Theoretical Perspectives and the Role of Rank Collapse",
                        "abstract": "Transformers have achieved remarkable success in several domains, ranging from natural language processing to computer vision. Nevertheless, it has been recently shown that stacking self-attention layers - the distinctive architectural component of Transformers - can result in rank collapse of the tokens' representations at initialization. The question of if and how rank collapse affects training is still largely unanswered, and its investigation is necessary for a more comprehensive understanding of this architecture. In this work, we shed new light on the causes and the effects of this phenomenon. First, we show that rank collapse of the tokens' representations hinders training by causing the gradients of the queries and keys to vanish at initialization. Furthermore, we provide a thorough description of the origin of rank collapse and discuss how to prevent it via an appropriate depth-dependent scaling of the residual branches. Finally, our analysis unveils that specific architectural hyperparameters affect the gradients of queries and values differently, leading to disproportionate gradient norms. This suggests an explanation for the widespread use of adaptive methods for Transformers' optimization."
                    },
                    {
                        "title": "The Curious Case of Benign Memorization",
                        "abstract": "Despite the empirical advances of deep learning across a variety of learning tasks, our theoretical understanding of its success is still very restricted. One of the key challenges is the overparametrized nature of modern models, enabling complete overfitting of the data even if the labels are randomized, i.e. networks can completely \\textit{memorize} all given patterns. While such a memorization capacity seems worrisome, in this work we show that under training protocols that include \\textit{data augmentation}, neural networks learn to memorize entirely random labels in a benign way, i.e. they learn embeddings that lead to highly non-trivial performance under nearest neighbour probing. We demonstrate that deep models have the surprising ability to separate noise from signal by distributing the task of memorization and feature learning to different layers. As a result, only the very last layers are used for memorization, while preceding layers encode performant features which remain largely unaffected by the label noise. We explore the intricate role of the augmentations used for training and identify a memorization-generalization trade-off in terms of their diversity, marking a clear distinction to all previous works. Finally, we give a first explanation for the emergence of benign memorization by showing that \\textit{malign} memorization under data augmentation is infeasible due to the insufficient capacity of the model for the increased sample size. As a consequence, the network is forced to leverage the correlated nature of the augmentations and as a result learns meaningful features. To complete the picture, a better theory of feature learning in deep neural networks is required to fully understand the origins of this phenomenon."
                    },
                    {
                        "title": "How Tempering Fixes Data Augmentation in Bayesian Neural Networks",
                        "abstract": "While Bayesian neural networks (BNNs) provide a sound and principled alternative to standard neural networks, an artificial sharpening of the posterior usually needs to be applied to reach comparable performance. This is in stark contrast to theory, dictating that given an adequate prior and a well-specified model, the untempered Bayesian posterior should achieve optimal performance. Despite the community's extensive efforts, the observed gains in performance still remain disputed with several plausible causes pointing at its origin. While data augmentation has been empirically recognized as one of the main drivers of this effect, a theoretical account of its role, on the other hand, is largely missing. In this work we identify two interlaced factors concurrently influencing the strength of the cold posterior effect, namely the correlated nature of augmentations and the degree of invariance of the employed model to such transformations. By theoretically analyzing simplified settings, we prove that tempering implicitly reduces the misspecification arising from modeling augmentations as i.i.d. data. The temperature mimics the role of the effective sample size, reflecting the gain in information provided by the augmentations. We corroborate our theoretical findings with extensive empirical evaluations, scaling to realistic BNNs. By relying on the framework of group convolutions, we experiment with models of varying inherent degree of invariance, confirming its hypothesized relationship with the optimal temperature."
                    },
                    {
                        "title": "Disentangling the Roles of Curation, Data-Augmentation and the Prior in the Cold Posterior Effect",
                        "abstract": "The\"cold posterior effect\"(CPE) in Bayesian deep learning describes the uncomforting observation that the predictive performance of Bayesian neural networks can be significantly improved if the Bayes posterior is artificially sharpened using a temperature parameter T<1. The CPE is problematic in theory and practice and since the effect was identified many researchers have proposed hypotheses to explain the phenomenon. However, despite this intensive research effort the effect remains poorly understood. In this work we provide novel and nuanced evidence relevant to existing explanations for the cold posterior effect, disentangling three hypotheses: 1. The dataset curation hypothesis of Aitchison (2020): we show empirically that the CPE does not arise in a real curated data set but can be produced in a controlled experiment with varying curation strength. 2. The data augmentation hypothesis of Izmailov et al. (2021) and Fortuin et al. (2021): we show empirically that data augmentation is sufficient but not necessary for the CPE to be present. 3. The bad prior hypothesis of Wenzel et al. (2020): we use a simple experiment evaluating the relative importance of the prior and the likelihood, strongly linking the CPE to the prior. Our results demonstrate how the CPE can arise in isolation from synthetic curation, data augmentation, and bad priors. Cold posteriors observed\"in the wild\"are therefore unlikely to arise from a single simple cause; as a result, we do not expect a simple\"fix\"for cold posteriors."
                    },
                    {
                        "title": "Precise characterization of the prior predictive distribution of deep ReLU networks",
                        "abstract": "Recent works on Bayesian neural networks (BNNs) have highlighted the need to better understand the implications of using Gaussian priors in combination with the compositional structure of the network architecture. Similar in spirit to the kind of analysis that has been developed to devise better initialization schemes for neural networks (cf. He- or Xavier initialization), we derive a precise characterization of the prior predictive distribution of finite-width ReLU networks with Gaussian weights. While theoretical results have been obtained for their heavy-tailedness, the full characterization of the prior predictive distribution (i.e. its density, CDF and moments), remained unknown prior to this work. Our analysis, based on the Meijer-G function, allows us to quantify the influence of architectural choices such as the width or depth of the network on the resulting shape of the prior predictive distribution. We also formally connect our results to previous work in the infinite width setting, demonstrating that the moments of the distribution converge to those of a normal log-normal mixture in the infinite depth limit. Finally, our results provide valuable guidance on prior design: for instance, controlling the predictive variance with depth- and width-informed priors on the weights of the network."
                    },
                    {
                        "title": "Adversarial Learning for Debiasing Knowledge Graph Embeddings",
                        "abstract": "Knowledge Graphs (KG) are gaining increasing attention in both academia and industry. Despite their diverse benefits, recent research have identified social and cultural biases embedded in the representations learned from KGs. Such biases can have detrimental consequences on different population and minority groups as applications of KG begin to intersect and interact with social spheres. This paper aims at identifying and mitigating such biases in Knowledge Graph (KG) embeddings. As a first step, we explore popularity bias -- the relationship between node popularity and link prediction accuracy. In case of node2vec graph embeddings, we find that prediction accuracy of the embedding is negatively correlated with the degree of the node. However, in case of knowledge-graph embeddings (KGE), we observe an opposite trend. As a second step, we explore gender bias in KGE, and a careful examination of popular KGE algorithms suggest that sensitive attribute like the gender of a person can be predicted from the embedding. This implies that such biases in popular KGs is captured by the structural properties of the embedding. As a preliminary solution to debiasing KGs, we introduce a novel framework to filter out the sensitive attribute information from the KG embeddings, which we call FAN (Filtering Adversarial Network). We also suggest the applicability of FAN for debiasing other network embeddings which could be explored in future work."
                    }
                ]
            },
            "3fdaf743-c7ed-402c-a399-77c097657104": {
                "pk": "3fdaf743-c7ed-402c-a399-77c097657104",
                "name": "Mufan Bill Li",
                "collaborators": [
                    "Murat A. Erdogdu",
                    "Daniel M. Roy",
                    "Matthew Shunshi Zhang",
                    "Sinho Chewi",
                    "M. Nica",
                    "Yunbum Kook",
                    "K. Balasubramanian",
                    "Mihai Nica",
                    "Lorenzo Noci",
                    "Chuning Li"
                ],
                "domain": [
                    "Stochastic Processes",
                    "Neural Networks",
                    "Optimization",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Sampling from the Mean-Field Stationary Distribution",
                        "abstract": "We study the complexity of sampling from the stationary distribution of a mean-field SDE, or equivalently, the complexity of minimizing a functional over the space of probability measures which includes an interaction term. Our main insight is to decouple the two key aspects of this problem: (1) approximation of the mean-field SDE via a finite-particle system, via uniform-in-time propagation of chaos, and (2) sampling from the finite-particle stationary distribution, via standard log-concave samplers. Our approach is conceptually simpler and its flexibility allows for incorporating the state-of-the-art for both algorithms and theory. This leads to improved guarantees in numerous settings, including better guarantees for optimizing certain two-layer neural networks in the mean-field regime. A key technical contribution is to establish a new uniform-in-$N$ log-Sobolev inequality for the stationary distribution of the mean-field Langevin dynamics."
                    },
                    {
                        "title": "Improved Discretization Analysis for Underdamped Langevin Monte Carlo",
                        "abstract": "Underdamped Langevin Monte Carlo (ULMC) is an algorithm used to sample from unnormalized densities by leveraging the momentum of a particle moving in a potential well. We provide a novel analysis of ULMC, motivated by two central questions: (1) Can we obtain improved sampling guarantees beyond strong log-concavity? (2) Can we achieve acceleration for sampling? For (1), prior results for ULMC only hold under a log-Sobolev inequality together with a restrictive Hessian smoothness condition. Here, we relax these assumptions by removing the Hessian smoothness condition and by considering distributions satisfying a Poincar\\'e inequality. Our analysis achieves the state of art dimension dependence, and is also flexible enough to handle weakly smooth potentials. As a byproduct, we also obtain the first KL divergence guarantees for ULMC without Hessian smoothness under strong log-concavity, which is based on a new result on the log-Sobolev constant along the underdamped Langevin diffusion. For (2), the recent breakthrough of Cao, Lu, and Wang (2020) established the first accelerated result for sampling in continuous time via PDE methods. Our discretization analysis translates their result into an algorithmic guarantee, which indeed enjoys better condition number dependence than prior works on ULMC, although we leave open the question of full acceleration in discrete time. Both (1) and (2) necessitate R\\'enyi discretization bounds, which are more challenging than the typically used Wasserstein coupling arguments. We address this using a flexible discretization analysis based on Girsanov's theorem that easily extends to more general settings."
                    },
                    {
                        "title": "Differential Equation Scaling Limits of Shaped and Unshaped Neural Networks",
                        "abstract": "Recent analyses of neural networks with shaped activations (i.e. the activation function is scaled as the network size grows) have led to scaling limits described by differential equations. However, these results do not a priori tell us anything about\"ordinary\"unshaped networks, where the activation is unchanged as the network size grows. In this article, we find similar differential equation based asymptotic characterization for two types of unshaped networks. Firstly, we show that the following two architectures converge to the same infinite-depth-and-width limit at initialization: (i) a fully connected ResNet with a $d^{-1/2}$ factor on the residual branch, where $d$ is the network depth. (ii) a multilayer perceptron (MLP) with depth $d \\ll$ width $n$ and shaped ReLU activation at rate $d^{-1/2}$. Secondly, for an unshaped MLP at initialization, we derive the first order asymptotic correction to the layerwise correlation. In particular, if $\\rho_\\ell$ is the correlation at layer $\\ell$, then $q_t = \\ell^2 (1 - \\rho_\\ell)$ with $t = \\frac{\\ell}{n}$ converges to an SDE with a singularity at $t=0$. These results together provide a connection between shaped and unshaped network architectures, and opens up the possibility of studying the effect of normalization methods and how it connects with shaping activation functions."
                    },
                    {
                        "title": "The Shaped Transformer: Attention Models in the Infinite Depth-and-Width Limit",
                        "abstract": "In deep learning theory, the covariance matrix of the representations serves as a proxy to examine the network's trainability. Motivated by the success of Transformers, we study the covariance matrix of a modified Softmax-based attention model with skip connections in the proportional limit of infinite-depth-and-width. We show that at initialization the limiting distribution can be described by a stochastic differential equation (SDE) indexed by the depth-to-width ratio. To achieve a well-defined stochastic limit, the Transformer's attention mechanism is modified by centering the Softmax output at identity, and scaling the Softmax logits by a width-dependent temperature parameter. We examine the stability of the network through the corresponding SDE, showing how the scale of both the drift and diffusion can be elegantly controlled with the aid of residual connections. The existence of a stable SDE implies that the covariance structure is well-behaved, even for very large depth and width, thus preventing the notorious issues of rank degeneracy in deep attention models. Finally, we show, through simulations, that the SDE provides a surprisingly good description of the corresponding finite-size model. We coin the name shaped Transformer for these architectural modifications."
                    },
                    {
                        "title": "Acceleration of Gossip Algorithms through the Euler-Poisson-Darboux Equation",
                        "abstract": "Gossip algorithms and their accelerated versions have been studied exclusively in discrete time on graphs. In this work, we take a different approach, and consider the scaling limit of gossip algorithms in both large graphs and large number of iterations. These limits lead to well-known partial differential equations (PDEs) with insightful properties. On lattices, we prove that the non-accelerated gossip algorithm of Boyd et al. [2006] converges to the heat equation, and the accelerated Jacobi polynomial iteration of Berthier et al. [2020] converges to the Euler-Poisson-Darboux (EPD) equation - a damped wave equation. Remarkably, with appropriate parameters, the fundamental solution of the EPD equation has the ideal gossip behaviour: a uniform density over an ellipsoid, whose radius increases at a rate proportional to t - the fastest possible rate for locally communicating gossip algorithms. This is in contrast with the heat equation where the density spreads on a typical scale of $\\sqrt{t}$. Additionally, we provide simulations demonstrating that the gossip algorithms are accurately approximated by their limiting PDEs."
                    },
                    {
                        "title": "The Neural Covariance SDE: Shaped Infinite Depth-and-Width Networks at Initialization",
                        "abstract": "The logit outputs of a feedforward neural network at initialization are conditionally Gaussian, given a random covariance matrix defined by the penultimate layer. In this work, we study the distribution of this random matrix. Recent work has shown that shaping the activation function as network depth grows large is necessary for this covariance matrix to be non-degenerate. However, the current infinite-width-style understanding of this shaping method is unsatisfactory for large depth: infinite-width analyses ignore the microscopic fluctuations from layer to layer, but these fluctuations accumulate over many layers. To overcome this shortcoming, we study the random covariance matrix in the shaped infinite-depth-and-width limit. We identify the precise scaling of the activation function necessary to arrive at a non-trivial limit, and show that the random covariance matrix is governed by a stochastic differential equation (SDE) that we call the Neural Covariance SDE. Using simulations, we show that the SDE closely matches the distribution of the random covariance matrix of finite networks. Additionally, we recover an if-and-only-if condition for exploding and vanishing norms of large shaped networks based on the activation function."
                    },
                    {
                        "title": "Higher Order Generalization Error for First Order Discretization of Langevin Diffusion",
                        "abstract": "We propose a novel approach to analyze generalization error for discretizations of Langevin diffusion, such as the stochastic gradient Langevin dynamics (SGLD). For an $\\epsilon$ tolerance of expected generalization error, it is known that a first order discretization can reach this target if we run $\\Omega(\\epsilon^{-1} \\log (\\epsilon^{-1}) )$ iterations with $\\Omega(\\epsilon^{-1})$ samples. In this article, we show that with additional smoothness assumptions, even first order methods can achieve arbitrarily runtime complexity. More precisely, for each $N>0$, we provide a sufficient smoothness condition on the loss function such that a first order discretization can reach $\\epsilon$ expected generalization error given $\\Omega( \\epsilon^{-1/N} \\log (\\epsilon^{-1}) )$ iterations with $\\Omega(\\epsilon^{-1})$ samples."
                    },
                    {
                        "title": "The Future is Log-Gaussian: ResNets and Their Infinite-Depth-and-Width Limit at Initialization",
                        "abstract": "Theoretical results show that neural networks can be approximated by Gaussian processes in the infinite-width limit. However, for fully connected networks, it has been previously shown that for any fixed network width, $n$, the Gaussian approximation gets worse as the network depth, $d$, increases. Given that modern networks are deep, this raises the question of how well modern architectures, like ResNets, are captured by the infinite-width limit. To provide a better approximation, we study ReLU ResNets in the infinite-depth-and-width limit, where both depth and width tend to infinity as their ratio, $d/n$, remains constant. In contrast to the Gaussian infinite-width limit, we show theoretically that the network exhibits log-Gaussian behaviour at initialization in the infinite-depth-and-width limit, with parameters depending on the ratio $d/n$. Using Monte Carlo simulations, we demonstrate that even basic properties of standard ResNet architectures are poorly captured by the Gaussian limit, but remarkably well captured by our log-Gaussian limit. Moreover, our analysis reveals that ReLU ResNets at initialization are hypoactivated: fewer than half of the ReLUs are activated. Additionally, we calculate the interlayer correlations, which have the effect of exponentially increasing the variance of the network output. Based on our analysis, we introduce Balanced ResNets, a simple architecture modification, which eliminates hypoactivation and interlayer correlations and is more amenable to theoretical analysis."
                    },
                    {
                        "title": "Riemannian Langevin Algorithm for Solving Semidefinite Programs",
                        "abstract": "We propose a Langevin diffusion-based algorithm for non-convex optimization and sampling on a product manifold of spheres. Under a logarithmic Sobolev inequality, we establish a guarantee for finite iteration convergence to the Gibbs distribution in terms of Kullback-Leibler divergence. We show that with an appropriate temperature choice, the suboptimality gap to the global minimum is guaranteed to be arbitrarily small with high probability.  As an application, we analyze the proposed Langevin algorithm for solving the Burer-Monteiro relaxation of a semidefinite program (SDP). In particular, we establish a logarithmic Sobolev inequality for the Burer-Monteiro problem when there are no spurious local minima; hence implying a fast escape from saddle points. Combining the results, we then provide a global optimality guarantee for the SDP and the Max-Cut problem. More precisely, we show the Langevin algorithm achieves $\\epsilon$-multiplicative accuracy with high probability in $\\widetilde{\\Omega}( n^2 \\epsilon^{-3} )$ iterations, where $n$ is the size of the cost matrix."
                    }
                ]
            },
            "ea0f3e1b-358a-4f3e-bd3b-b4b9434cec41": {
                "pk": "ea0f3e1b-358a-4f3e-bd3b-b4b9434cec41",
                "name": "Boris Hanin",
                "collaborators": [
                    "Lingjiao Chen",
                    "Jared Quincy Davis",
                    "Peter D. Bailis",
                    "Ion Stoica",
                    "Matei Zaharia",
                    "James Zou",
                    "Wuyang Chen",
                    "Junru Wu",
                    "Zhangyang Wang",
                    "Alexander Zlokapa"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Bayesian Inference",
                    "AutoML"
                ],
                "publications": [
                    {
                        "title": "Are More LLM Calls All You Need? Towards Scaling Laws of Compound Inference Systems",
                        "abstract": "Many recent state-of-the-art results in language tasks were achieved using compound systems that perform multiple Language Model (LM) calls and aggregate their responses. However, there is little understanding of how the number of LM calls - e.g., when asking the LM to answer each question multiple times and taking a majority vote - affects such a compound system's performance. In this paper, we initiate the study of scaling properties of compound inference systems. We analyze, theoretically and empirically, how the number of LM calls affects the performance of Vote and Filter-Vote, two of the simplest compound system designs, which aggregate LM responses via majority voting, optionally applying LM filters. We find, surprisingly, that across multiple language tasks, the performance of both Vote and Filter-Vote can first increase but then decrease as a function of the number of LM calls. Our theoretical results suggest that this non-monotonicity is due to the diversity of query difficulties within a task: more LM calls lead to higher performance on\"easy\"queries, but lower performance on\"hard\"queries, and non-monotone behavior can emerge when a task contains both types of queries. This insight then allows us to compute, from a small number of samples, the number of LM calls that maximizes system performance, and define an analytical scaling model for both systems. Experiments show that our scaling model can accurately predict the performance of Vote and Filter-Vote systems and thus find the optimal number of LM calls to make."
                    },
                    {
                        "title": "Principled Architecture-aware Scaling of Hyperparameters",
                        "abstract": "Training a high-quality deep neural network requires choosing suitable hyperparameters, which is a non-trivial and expensive process. Current works try to automatically optimize or design principles of hyperparameters, such that they can generalize to diverse unseen scenarios. However, most designs or optimization methods are agnostic to the choice of network structures, and thus largely ignore the impact of neural architectures on hyperparameters. In this work, we precisely characterize the dependence of initializations and maximal learning rates on the network architecture, which includes the network depth, width, convolutional kernel size, and connectivity patterns. By pursuing every parameter to be maximally updated with the same mean squared change in pre-activations, we can generalize our initialization and learning rates across MLPs (multi-layer perception) and CNNs (convolutional neural network) with sophisticated graph topologies. We verify our principles with comprehensive experiments. More importantly, our strategy further sheds light on advancing current benchmarks for architecture design. A fair comparison of AutoML algorithms requires accurate network rankings. However, we demonstrate that network rankings can be easily changed by better training networks in benchmarks with our architecture-aware learning rates and initialization."
                    },
                    {
                        "title": "Are More LLM Calls All You Need? Towards Scaling Laws of Compound Inference Systems",
                        "abstract": "Many recent state-of-the-art results in language tasks were achieved using compound systems that perform multiple Large Language Model (LLM) calls and aggregate their responses. However, there is little understanding of how the number of LLM calls \u2013 e.g., when asking the LLM to answer each question multiple times and taking a consensus \u2013 affects such a compound system\u2019s performance. In this paper, we initiate the study of scaling laws of compound inference systems. We analyze, theoretically and empirically, how the number of LLM calls affects the performance of one-layer Voting Inference Systems \u2013 one of the simplest compound systems, which aggregates LLM responses via majority voting. We find empirically that across multiple language tasks, surprisingly, Voting Inference Systems\u2019 performance first increases but then decreases as a function of the number of LLM calls. Our theoretical results suggest that this non-monotonicity is due to the diversity of query difficulties within a task: more LLM calls lead to higher performance on \u201ceasy\u201d queries, but lower performance on \u201chard\u201d queries, and non-monotone behavior emerges when a task contains both types of queries. This insight then allows us to compute, from a small number of samples, the number of LLM calls that maximizes system performance, and define a scaling law of Voting Inference Systems. Experiments show that our scaling law can predict the performance of Voting Inference Systems and find the optimal number of LLM calls to make."
                    },
                    {
                        "title": "Bayesian Inference with Deep Weakly Nonlinear Networks",
                        "abstract": "We show at a physics level of rigor that Bayesian inference with a fully connected neural network and a shaped nonlinearity of the form $\\phi(t) = t + \\psi t^3/L$ is (perturbatively) solvable in the regime where the number of training datapoints $P$ , the input dimension $N_0$, the network layer widths $N$, and the network depth $L$ are simultaneously large. Our results hold with weak assumptions on the data; the main constraint is that $P<N_0$. We provide techniques to compute the model evidence and posterior to arbitrary order in $1/N$ and at arbitrary temperature. We report the following results from the first-order computation: 1. When the width $N$ is much larger than the depth $L$ and training set size $P$, neural network Bayesian inference coincides with Bayesian inference using a kernel. The value of $\\psi$ determines the curvature of a sphere, hyperbola, or plane into which the training data is implicitly embedded under the feature map. 2. When $LP/N$ is a small constant, neural network Bayesian inference departs from the kernel regime. At zero temperature, neural network Bayesian inference is equivalent to Bayesian inference using a data-dependent kernel, and $LP/N$ serves as an effective depth that controls the extent of feature learning. 3. In the restricted case of deep linear networks ($\\psi=0$) and noisy data, we show a simple data model for which evidence and generalization error are optimal at zero temperature. As $LP/N$ increases, both evidence and generalization further improve, demonstrating the benefit of depth in benign overfitting."
                    },
                    {
                        "title": "Networks of Networks: Complexity Class Principles Applied to Compound AI Systems Design",
                        "abstract": "As practitioners seek to surpass the current reliability and quality frontier of monolithic models, Compound AI Systems consisting of many language model inference calls are increasingly employed. In this work, we construct systems, which we call Networks of Networks (NoNs) organized around the distinction between generating a proposed answer and verifying its correctness, a fundamental concept in complexity theory that we show empirically extends to Language Models (LMs). We introduce a verifier-based judge NoN with K generators, an instantiation of\"best-of-K\"or\"judge-based\"compound AI systems. Through experiments on synthetic tasks such as prime factorization, and core benchmarks such as the MMLU, we demonstrate notable performance gains. For instance, in factoring products of two 3-digit primes, a simple NoN improves accuracy from 3.7\\% to 36.6\\%. On MMLU, a verifier-based judge construction with only 3 generators boosts accuracy over individual GPT-4-Turbo calls by 2.8\\%. Our analysis reveals that these gains are most pronounced in domains where verification is notably easier than generation--a characterization which we believe subsumes many reasoning and procedural knowledge tasks, but doesn't often hold for factual and declarative knowledge-based settings. For mathematical and formal logic reasoning-based subjects of MMLU, we observe a 5-8\\% or higher gain, whilst no gain on others such as geography and religion. We provide key takeaways for ML practitioners, including the importance of considering verification complexity, the impact of witness format on verifiability, and a simple test to determine the potential benefit of this NoN approach for a given problem distribution. This work aims to inform future research and practice in the design of compound AI systems."
                    },
                    {
                        "title": "Quantitative CLTs in Deep Neural Networks",
                        "abstract": "We study the distribution of a fully connected neural network with random Gaussian weights and biases in which the hidden layer widths are proportional to a large constant $n$. Under mild assumptions on the non-linearity, we obtain quantitative bounds on normal approximations valid at large but finite $n$ and any fixed network depth. Our theorems show both for the finite-dimensional distributions and the entire process, that the distance between a random fully connected network (and its derivatives) to the corresponding infinite width Gaussian process scales like $n^{-\\gamma}$ for $\\gamma>0$, with the exponent depending on the metric used to measure discrepancy. Our bounds are strictly stronger in terms of their dependence on network width than any previously available in the literature; in the one-dimensional case, we also prove that they are optimal, i.e., we establish matching lower bounds."
                    },
                    {
                        "title": "Les Houches Lectures on Deep Learning at Large & Infinite Width",
                        "abstract": "These lectures, presented at the 2022 Les Houches Summer School on Statistical Physics and Machine Learning, focus on the infinite-width limit and large-width regime of deep neural networks. Topics covered include various statistical and dynamical properties of these networks. In particular, the lecturers discuss properties of random deep neural networks; connections between trained deep neural networks, linear models, kernels, and Gaussian processes that arise in the infinite-width limit; and perturbative and non-perturbative treatments of large but finite-width networks, at initialization and after training."
                    }
                ]
            },
            "3e19da2a-4df3-42a0-a0d3-c79513c2cfb2": {
                "pk": "3e19da2a-4df3-42a0-a0d3-c79513c2cfb2",
                "name": "Cengiz Pehlevan",
                "collaborators": [
                    "Jacob A. Zavatone-Veth",
                    "Blake Bordelon",
                    "Alexander B. Atanasov",
                    "Sheng Yang",
                    "Alexander Atanasov",
                    "P. Masset",
                    "Henry Kuo",
                    "Yue M. Lu",
                    "Mary I. Letey",
                    "Anindita Maiti"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "In-Context Learning",
                    "Adversarial Robustness"
                ],
                "publications": [
                    {
                        "title": "Asymptotic theory of in-context learning by linear attention",
                        "abstract": "Transformers have a remarkable ability to learn and execute tasks based on examples provided within the input itself, without explicit prior training. It has been argued that this capability, known as in-context learning (ICL), is a cornerstone of Transformers' success, yet questions about the necessary sample complexity, pretraining task diversity, and context length for successful ICL remain unresolved. Here, we provide a precise answer to these questions in an exactly solvable model of ICL of a linear regression task by linear attention. We derive sharp asymptotics for the learning curve in a phenomenologically-rich scaling regime where the token dimension is taken to infinity; the context length and pretraining task diversity scale proportionally with the token dimension; and the number of pretraining examples scales quadratically. We demonstrate a double-descent learning curve with increasing pretraining examples, and uncover a phase transition in the model's behavior between low and high task diversity regimes: In the low diversity regime, the model tends toward memorization of training tasks, whereas in the high diversity regime, it achieves genuine in-context learning and generalization beyond the scope of pretrained tasks. These theoretical insights are empirically validated through experiments with both linear attention and full nonlinear Transformer architectures."
                    },
                    {
                        "title": "How Feature Learning Can Improve Neural Scaling Laws",
                        "abstract": "We develop a solvable model of neural scaling laws beyond the kernel limit. Theoretical analysis of this model shows how performance scales with model size, training time, and the total amount of available data. We identify three scaling regimes corresponding to varying task difficulties: hard, easy, and super easy tasks. For easy and super-easy target functions, which lie in the reproducing kernel Hilbert space (RKHS) defined by the initial infinite-width Neural Tangent Kernel (NTK), the scaling exponents remain unchanged between feature learning and kernel regime models. For hard tasks, defined as those outside the RKHS of the initial NTK, we demonstrate both analytically and empirically that feature learning can improve scaling with training time and compute, nearly doubling the exponent for hard tasks. This leads to a different compute optimal strategy to scale parameters and training time in the feature learning regime. We support our finding that feature learning improves the scaling law for hard tasks but not for easy and super-easy tasks with experiments of nonlinear MLPs fitting functions with power-law Fourier spectra on the circle and CNNs learning vision tasks."
                    },
                    {
                        "title": "MLPs Learn In-Context on Regression and Classification Tasks",
                        "abstract": "In-context learning (ICL), the remarkable ability to solve a task from only input exemplars, is often assumed to be a unique hallmark of Transformer models. By examining commonly employed synthetic ICL tasks, we demonstrate that multi-layer perceptrons (MLPs) can also learn in-context. Moreover, MLPs, and the closely related MLP-Mixer models, learn in-context competitively with Transformers given the same compute budget in this setting. We further show that MLPs outperform Transformers on a series of classical tasks from psychology designed to test relational reasoning, which are closely related to in-context classification. These results underscore a need for studying in-context learning beyond attention-based architectures, while also challenging strong prior arguments about MLPs' limited ability to solve relational tasks. Altogether, our results highlight the unexpected competence of MLPs, and support the growing interest in all-MLP alternatives to task-specific architectures."
                    },
                    {
                        "title": "Convex Relaxation for Solving Large-Margin Classifiers in Hyperbolic Space",
                        "abstract": "Hyperbolic spaces have increasingly been recognized for their outstanding performance in handling data with inherent hierarchical structures compared to their Euclidean counterparts. However, learning in hyperbolic spaces poses significant challenges. In particular, extending support vector machines to hyperbolic spaces is in general a constrained non-convex optimization problem. Previous and popular attempts to solve hyperbolic SVMs, primarily using projected gradient descent, are generally sensitive to hyperparameters and initializations, often leading to suboptimal solutions. In this work, by first rewriting the problem into a polynomial optimization, we apply semidefinite relaxation and sparse moment-sum-of-squares relaxation to effectively approximate the optima. From extensive empirical experiments, these methods are shown to perform better than the projected gradient descent approach."
                    },
                    {
                        "title": "Risk and cross validation in ridge regression with correlated samples",
                        "abstract": "Recent years have seen substantial advances in our understanding of high-dimensional ridge regression, but existing theories assume that training examples are independent. By leveraging recent techniques from random matrix theory and free probability, we provide sharp asymptotics for the in- and out-of-sample risks of ridge regression when the data points have arbitrary correlations. We demonstrate that in this setting, the generalized cross validation estimator (GCV) fails to correctly predict the out-of-sample risk. However, in the case where the noise residuals have the same correlations as the data points, one can modify the GCV to yield an efficiently-computable unbiased estimator that concentrates in the high-dimensional limit, which we dub CorrGCV. We further extend our asymptotic analysis to the case where the test point has nontrivial correlations with the training set, a setting often encountered in time series forecasting. Assuming knowledge of the correlation structure of the time series, this again yields an extension of the GCV estimator, and sharply characterizes the degree to which such test points yield an overly optimistic prediction of long-time risk. We validate the predictions of our theory across a variety of high dimensional data."
                    },
                    {
                        "title": "Spectral regularization for adversarially-robust representation learning",
                        "abstract": "The vulnerability of neural network classifiers to adversarial attacks is a major obstacle to their deployment in safety-critical applications. Regularization of network parameters during training can be used to improve adversarial robustness and generalization performance. Usually, the network is regularized end-to-end, with parameters at all layers affected by regularization. However, in settings where learning representations is key, such as self-supervised learning (SSL), layers after the feature representation will be discarded when performing inference. For these models, regularizing up to the feature space is more suitable. To this end, we propose a new spectral regularizer for representation learning that encourages black-box adversarial robustness in downstream classification tasks. In supervised classification settings, we show empirically that this method is more effective in boosting test accuracy and robustness than previously-proposed methods that regularize all layers of the network. We then show that this method improves the adversarial robustness of classifiers using representations learned with self-supervised training or transferred from another classification task. In all, our work begins to unveil how representational structure affects adversarial robustness."
                    },
                    {
                        "title": "Infinite Limits of Multi-head Transformer Dynamics",
                        "abstract": "In this work, we analyze various scaling limits of the training dynamics of transformer models in the feature learning regime. We identify the set of parameterizations that admit well-defined infinite width and depth limits, allowing the attention layers to update throughout training--a relevant notion of feature learning in these models. We then use tools from dynamical mean field theory (DMFT) to analyze various infinite limits (infinite key/query dimension, infinite heads, and infinite depth) which have different statistical descriptions depending on which infinite limit is taken and how attention layers are scaled. We provide numerical evidence of convergence to the limits and discuss how the parameterization qualitatively influences learned features."
                    },
                    {
                        "title": "A Brain-Inspired Regularizer for Adversarial Robustness",
                        "abstract": "Convolutional Neural Networks (CNNs) excel in many visual tasks, but they tend to be sensitive to slight input perturbations that are imperceptible to the human eye, often resulting in task failures. Recent studies indicate that training CNNs with regularizers that promote brain-like representations, using neural recordings, can improve model robustness. However, the requirement to use neural data severely restricts the utility of these methods. Is it possible to develop regularizers that mimic the computational function of neural regularizers without the need for neural recordings, thereby expanding the usability and effectiveness of these techniques? In this work, we inspect a neural regularizer introduced in Li et al. (2019) to extract its underlying strength. The regularizer uses neural representational similarities, which we find also correlate with pixel similarities. Motivated by this finding, we introduce a new regularizer that retains the essence of the original but is computed using image pixel similarities, eliminating the need for neural recordings. We show that our regularization method 1) significantly increases model robustness to a range of black box attacks on various datasets and 2) is computationally inexpensive and relies only on original datasets. Our work explores how biologically motivated loss functions can be used to drive the performance of artificial neural networks."
                    },
                    {
                        "title": "A Dynamical Model of Neural Scaling Laws",
                        "abstract": "On a variety of tasks, the performance of neural networks predictably improves with training time, dataset size and model size across many orders of magnitude. This phenomenon is known as a neural scaling law. Of fundamental importance is the compute-optimal scaling law, which reports the performance as a function of units of compute when choosing model sizes optimally. We analyze a random feature model trained with gradient descent as a solvable model of network training and generalization. This reproduces many observations about neural scaling laws. First, our model makes a prediction about why the scaling of performance with training time and with model size have different power law exponents. Consequently, the theory predicts an asymmetric compute-optimal scaling rule where the number of training steps are increased faster than model parameters, consistent with recent empirical observations. Second, it has been observed that early in training, networks converge to their infinite-width dynamics at a rate $1/\\textit{width}$ but at late time exhibit a rate $\\textit{width}^{-c}$, where $c$ depends on the structure of the architecture and task. We show that our model exhibits this behavior. Lastly, our theory shows how the gap between training and test loss can gradually build up over time due to repeated reuse of data."
                    },
                    {
                        "title": "Nadaraya-Watson kernel smoothing as a random energy model",
                        "abstract": "We investigate the behavior of the Nadaraya-Watson kernel smoothing estimator in high dimensions using its relationship to the random energy model and to dense associative memories."
                    },
                    {
                        "title": "Scaling and renormalization in high-dimensional regression",
                        "abstract": "This paper presents a succinct derivation of the training and generalization performance of a variety of high-dimensional ridge regression models using the basic tools of random matrix theory and free probability. We provide an introduction and review of recent results on these topics, aimed at readers with backgrounds in physics and deep learning. Analytic formulas for the training and generalization errors are obtained in a few lines of algebra directly from the properties of the $S$-transform of free probability. This allows for a straightforward identification of the sources of power-law scaling in model performance. We compute the generalization error of a broad class of random feature models. We find that in all models, the $S$-transform corresponds to the train-test generalization gap, and yields an analogue of the generalized-cross-validation estimator. Using these techniques, we derive fine-grained bias-variance decompositions for a very general class of random feature models with structured covariates. These novel results allow us to discover a scaling regime for random feature models where the variance due to the features limits performance in the overparameterized setting. We also demonstrate how anisotropic weight structure in random feature models can limit performance and lead to nontrivial exponents for finite-width corrections in the overparameterized setting. Our results extend and provide a unifying perspective on earlier models of neural scaling laws."
                    },
                    {
                        "title": "Partial observation can induce mechanistic mismatches in data-constrained models of neural dynamics",
                        "abstract": "One of the central goals of neuroscience is to gain a mechanistic understanding of how the dynamics of neural circuits give rise to their observed function. A popular approach towards this end is to train recurrent neural networks (RNNs) to reproduce experimental recordings of neural activity. These trained RNNs are then treated as surrogate models of biological neural circuits, whose properties can be dissected via dynamical systems analysis. How reliable are the mechanistic insights derived from this procedure? While recent advances in population-level recording technologies have allowed simultaneous recording of up to tens of thousands of neurons, this represents only a tiny fraction of most cortical circuits. Here we show that observing only a subset of neurons in a circuit can create mechanistic mismatches between a simulated teacher network and a data-constrained student, even when the two networks have matching single-unit dynamics. In particular, partial observation of models of low-dimensional cortical dynamics based on functionally feedforward or low-rank connectivity can lead to surrogate models with spurious attractor structure. Our results illustrate the challenges inherent in accurately uncovering neural mechanisms from single-trial data, and suggest the need for new methods of validating data-constrained models for neural dynamics."
                    },
                    {
                        "title": "The Optimization Landscape of SGD Across the Feature Learning Strength",
                        "abstract": "We consider neural networks (NNs) where the final layer is down-scaled by a fixed hyperparameter $\\gamma$. Recent work has identified $\\gamma$ as controlling the strength of feature learning. As $\\gamma$ increases, network evolution changes from\"lazy\"kernel dynamics to\"rich\"feature-learning dynamics, with a host of associated benefits including improved performance on common tasks. In this work, we conduct a thorough empirical investigation of the effect of scaling $\\gamma$ across a variety of models and datasets in the online training setting. We first examine the interaction of $\\gamma$ with the learning rate $\\eta$, identifying several scaling regimes in the $\\gamma$-$\\eta$ plane which we explain theoretically using a simple model. We find that the optimal learning rate $\\eta^*$ scales non-trivially with $\\gamma$. In particular, $\\eta^* \\propto \\gamma^2$ when $\\gamma \\ll 1$ and $\\eta^* \\propto \\gamma^{2/L}$ when $\\gamma \\gg 1$ for a feed-forward network of depth $L$. Using this optimal learning rate scaling, we proceed with an empirical study of the under-explored\"ultra-rich\"$\\gamma \\gg 1$ regime. We find that networks in this regime display characteristic loss curves, starting with a long plateau followed by a drop-off, sometimes followed by one or more additional staircase steps. We find networks of different large $\\gamma$ values optimize along similar trajectories up to a reparameterization of time. We further find that optimal online performance is often found at large $\\gamma$ and could be missed if this hyperparameter is not tuned. Our findings indicate that analytical study of the large-$\\gamma$ limit may yield useful insights into the dynamics of representation learning in performant models."
                    },
                    {
                        "title": "Learning curves for deep structured Gaussian feature models",
                        "abstract": "  In recent years, significant attention in deep learning theory has been devoted to analyzing when models that interpolate their training data can still generalize well to unseen examples. Many insights have been gained from studying models with multiple layers of Gaussian random features, for which one can compute precise generalization asymptotics. However, few works have considered the effect of weight anisotropy; most assume that the random features are generated using independent and identically distributed Gaussian weights, and allow only for structure in the input data. Here, we use the replica trick from statistical physics to derive learning curves for models with many layers of structured Gaussian features. We show that allowing correlations between the rows of the first layer of features can aid generalization, while structure in later layers is generally detrimental. Our results shed light on how weight structure affects generalization in a simple class of solvable models."
                    },
                    {
                        "title": "Dynamics of Temporal Difference Reinforcement Learning",
                        "abstract": "Reinforcement learning has been successful across several applications in which agents have to learn to act in environments with sparse feedback. However, despite this empirical success there is still a lack of theoretical understanding of how the parameters of reinforcement learning models and the features used to represent states interact to control the dynamics of learning. In this work, we use concepts from statistical physics, to study the typical case learning curves for temporal difference learning of a value function with linear function approximators. Our theory is derived under a Gaussian equivalence hypothesis where averages over the random trajectories are replaced with temporally correlated Gaussian feature averages and we validate our assumptions on small scale Markov Decision Processes. We find that the stochastic semi-gradient noise due to subsampling the space of possible episodes leads to significant plateaus in the value error, unlike in traditional gradient descent dynamics. We study how learning dynamics and plateaus depend on feature structure, learning rate, discount factor, and reward function. We then analyze how strategies like learning rate annealing and reward shaping can favorably alter learning dynamics and plateaus. To conclude, our work introduces new tools to open a new direction towards developing a theory of learning dynamics in reinforcement learning."
                    },
                    {
                        "title": "Loss Dynamics of Temporal Difference Reinforcement Learning",
                        "abstract": "Reinforcement learning has been successful across several applications in which agents have to learn to act in environments with sparse feedback. However, despite this empirical success there is still a lack of theoretical understanding of how the parameters of reinforcement learning models and the features used to represent states interact to control the dynamics of learning. In this work, we use concepts from statistical physics, to study the typical case learning curves for temporal difference learning of a value function with linear function approximators. Our theory is derived under a Gaussian equivalence hypothesis where averages over the random trajectories are replaced with temporally correlated Gaussian feature averages and we validate our assumptions on small scale Markov Decision Processes. We find that the stochastic semi-gradient noise due to subsampling the space of possible episodes leads to significant plateaus in the value error, unlike in traditional gradient descent dynamics. We study how learning dynamics and plateaus depend on feature structure, learning rate, discount factor, and reward function. We then analyze how strategies like learning rate annealing and reward shaping can favorably alter learning dynamics and plateaus. To conclude, our work introduces new tools to open a new direction towards developing a theory of learning dynamics in reinforcement learning."
                    },
                    {
                        "title": "Grokking as the Transition from Lazy to Rich Training Dynamics",
                        "abstract": "We propose that the grokking phenomenon, where the train loss of a neural network decreases much earlier than its test loss, can arise due to a neural network transitioning from lazy training dynamics to a rich, feature learning regime. To illustrate this mechanism, we study the simple setting of vanilla gradient descent on a polynomial regression problem with a two layer neural network which exhibits grokking without regularization in a way that cannot be explained by existing theories. We identify sufficient statistics for the test loss of such a network, and tracking these over training reveals that grokking arises in this setting when the network first attempts to fit a kernel regression solution with its initial features, followed by late-time feature learning where a generalizing solution is identified after train loss is already low. We find that the key determinants of grokking are the rate of feature learning -- which can be controlled precisely by parameters that scale the network output -- and the alignment of the initial features with the target function $y(x)$. We argue this delayed generalization arises when (1) the top eigenvectors of the initial neural tangent kernel and the task labels $y(x)$ are misaligned, but (2) the dataset size is large enough so that it is possible for the network to generalize eventually, but not so large that train loss perfectly tracks test loss at all epochs, and (3) the network begins training in the lazy regime so does not learn features immediately. We conclude with evidence that this transition from lazy (linear model) to rich training (feature learning) can control grokking in more general settings, like on MNIST, one-layer Transformers, and student-teacher networks."
                    },
                    {
                        "title": "Neural networks learn to magnify areas near decision boundaries",
                        "abstract": "In machine learning, there is a long history of trying to build neural networks that can learn from fewer example data by baking in strong geometric priors. However, it is not always clear a priori what geometric constraints are appropriate for a given task. Here, we consider the possibility that one can uncover useful geometric inductive biases by studying how training molds the Riemannian geometry induced by unconstrained neural network feature maps. We first show that at infinite width, neural networks with random parameters induce highly symmetric metrics on input space. This symmetry is broken by feature learning: networks trained to perform classification tasks learn to magnify local areas along decision boundaries. This holds in deep networks trained on high-dimensional image classification tasks, and even in self-supervised representation learning. These results begins to elucidate how training shapes the geometry induced by unconstrained neural network feature maps, laying the groundwork for an understanding of this richly nonlinear form of feature learning."
                    }
                ]
            }
        }
    },
    "2406.07592": {
        "paper_data": {
            "title": "MambaLRP: Explaining Selective State Space Sequence Models",
            "url": "http://arxiv.org/abs/2406.07592v1",
            "arxiv_id": "2406.07592",
            "authors": [
                "Farnoush Rezaei Jafari",
                "Gr\u00e9goire Montavon",
                "Klaus-Robert M\u00fcller",
                "Oliver Eberle"
            ],
            "abstract": "Recent sequence modeling approaches using Selective State Space Sequence Models, referred to as Mamba models, have seen a surge of interest. These models allow efficient processing of long sequences in linear time and are rapidly being adopted in a wide range of applications such as language modeling, demonstrating promising performance. To foster their reliable use in real-world scenarios, it is crucial to augment their transparency. Our work bridges this critical gap by bringing explainability, particularly Layer-wise Relevance Propagation (LRP), to the Mamba architecture. Guided by the axiom of relevance conservation, we identify specific components in the Mamba architecture, which cause unfaithful explanations. To remedy this issue, we propose MambaLRP, a novel algorithm within the LRP framework, which ensures a more stable and reliable relevance propagation through these components. Our proposed method is theoretically sound and excels in achieving state-of-the-art explanation performance across a diverse range of models and datasets. Moreover, MambaLRP facilitates a deeper inspection of Mamba architectures, uncovering various biases and evaluating their significance. It also enables the analysis of previous speculations regarding the long-range capabilities of Mamba models.",
            "introduction": "   1 Introduction  Sequence modeling has demonstrated its effectiveness and versatility across a wide variety of tasks and data types, including text, time series, genomics, audio, and computer vision [21, 72, 28, 8, 23]. Recently, there has been a surge of interest in a new class of sequence modeling architectures, known as structured state space sequence models (SSMs) [31, 56, 29]. This is due to their ability to process sequences in linear time, as opposed to quadratic time required by the more established Transformer architectures [59]. The recent Mamba architecture, a prominent and widely adopted instance of state space models, has demonstrated competitive predictive performance on a variety of sequence modeling tasks across domains and applications [29, 39, 73, 68, 60], while scaling linearly with sequence length.   As Mamba models, and more generally SSMs, are rapidly being adopted into real-world applications, ensuring their transparency is crucial. This enables inspection beyond test set accuracy and uncovering various forms of biases, including \u2018Clever-Hans\u2019 effects [35]. It is particularly important in high-risk domains such as medicine, where the prediction behavior must be robust under real-world conditions and aligned with human understanding. The field of Explainable AI [42, 32, 7, 49] focuses on developing faithful model explanations that attribute predictions to relevant features and has shown success in explaining many highly nonlinear models such as convolutional networks [18], or attention-based Transformer models [3, 2].   Figure 1: Conceptual steps involved in the design of MambaLRP. (a) Take as a starting point a basic LRP procedure, equivalent to Gradient\u00d7\\,\\times\\,\u00d7Input. (b) Analyze layers in which the conservation property is violated. (c) Rework the relevance propagation strategy at those layers to achieve conservation. The resulting MambaLRP method enables efficient and faithful explanations.   Explaining the predictions of Mamba models is however challenging due to their highly non-linear and recurrent structure. A recent study [4] suggests viewing these models as attention-based models, enabling the use of attention-based explanation methods [1, 17]. Yet, the explanations produced by attention-based techniques are often unreliable and exposed to potential misalignment between input features and attention scores [65, 34]. As an alternative, Layer-wise Relevance Propagation (LRP) [9] decomposes the model function with the goal of explicitly identifying the relevance of input features by applying purposely designed propagation rules at each layer. A distinguishing feature of LRP is its adherence to a conservation axiom, which prevents the artificial amplification or suppression of feature relevance in the backward pass. LRP has been demonstrated to produce faithful explanations across various domains (e.g.\u00a0[6, 3, 18]). Nevertheless, the peculiarities of the Mamba architecture are not addressed by the existing LRP procedures, which may lead to the violation of the conservation property and result in unreliable explanations.   In this work, we present MambaLRP, a novel approach to integrate LRP into the Mamba architecture. By examining the relevance propagation process across Mamba layers through the lens of conservation, we pinpoint layers within the Mamba architecture that need to be addressed specifically. We propose a novel relevance propagation strategy for these layers, grounded in the conservation axiom, that is theoretically sound, straightforward to implement and computationally efficient. Through a number of",
            "references": [
                {
                    "title": "RULER: What's the Real Context Size of Your Long-Context Language Models?",
                    "abstract": "The needle-in-a-haystack (NIAH) test, which examines the ability to retrieve a piece of information (the\"needle\") from long distractor texts (the\"haystack\"), has been widely adopted to evaluate long-context language models (LMs). However, this simple retrieval-based test is indicative of only a superficial form of long-context understanding. To provide a more comprehensive evaluation of long-context LMs, we create a new synthetic benchmark RULER with flexible configurations for customized sequence length and task complexity. RULER expands upon the vanilla NIAH test to encompass variations with diverse types and quantities of needles. Moreover, RULER introduces new task categories multi-hop tracing and aggregation to test behaviors beyond searching from context. We evaluate 17 long-context LMs with 13 representative tasks in RULER. Despite achieving nearly perfect accuracy in the vanilla NIAH test, almost all models exhibit large performance drops as the context length increases. While these models all claim context sizes of 32K tokens or greater, only half of them can maintain satisfactory performance at the length of 32K. Our analysis of Yi-34B, which supports context length of 200K, reveals large room for improvement as we increase input length and task complexity. We open source RULER to spur comprehensive evaluation of long-context LMs."
                },
                {
                    "title": "The Hidden Attention of Mamba Models",
                    "abstract": "The Mamba layer offers an efficient selective state space model (SSM) that is highly effective in modeling multiple domains, including NLP, long-range sequence processing, and computer vision. Selective SSMs are viewed as dual models, in which one trains in parallel on the entire sequence via an IO-aware parallel scan, and deploys in an autoregressive manner. We add a third view and show that such models can be viewed as attention-driven models. This new perspective enables us to empirically and theoretically compare the underlying mechanisms to that of the self-attention layers in transformers and allows us to peer inside the inner workings of the Mamba model with explainability methods. Our code is publicly available."
                },
                {
                    "title": "Graph Mamba: Towards Learning on Graphs with State Space Models",
                    "abstract": "Graph Neural Networks (GNNs) have shown promising potential in graph representation learning. The majority of GNNs define a local message-passing mechanism, propagating information over the graph by stacking multiple layers. These methods, however, are known to suffer from two major limitations: over-squashing and poor capturing of long-range dependencies. Recently, Graph Transformers (GTs) emerged as a powerful alternative to Message-Passing Neural Networks (MPNNs). GTs, however, have quadratic computational cost, lack inductive biases on graph structures, and rely on complex Positional/Structural Encodings (SE/PE). In this paper, we show that while Transformers, complex message-passing, and SE/PE are sufficient for good performance in practice, neither is necessary. Motivated by the recent success of State Space Models (SSMs), such as Mamba, we present Graph Mamba Networks (GMNs), a general framework for a new class of GNNs based on selective SSMs. We discuss and categorize the new challenges when adapting SSMs to graph-structured data, and present four required and one optional steps to design GMNs, where we choose (1) Neighborhood Tokenization, (2) Token Ordering, (3) Architecture of Bidirectional Selective SSM Encoder, (4) Local Encoding, and dispensable (5) PE and SE. We further provide theoretical justification for the power of GMNs. Experiments demonstrate that despite much less computational cost, GMNs attain an outstanding performance in long-range, small-scale, large-scale, and heterophilic benchmark datasets."
                },
                {
                    "title": "AttnLRP: Attention-Aware Layer-wise Relevance Propagation for Transformers",
                    "abstract": "Large Language Models are prone to biased predictions and hallucinations, underlining the paramount importance of understanding their model-internal reasoning process. However, achieving faithful attributions for the entirety of a black-box transformer model and maintaining computational efficiency is an unsolved challenge. By extending the Layer-wise Relevance Propagation attribution method to handle attention layers, we address these challenges effectively. While partial solutions exist, our method is the first to faithfully and holistically attribute not only input but also latent representations of transformer models with the computational efficiency similar to a single backward pass. Through extensive evaluations against existing methods on LLaMa 2, Mixtral 8x7b, Flan-T5 and vision transformer architectures, we demonstrate that our proposed approach surpasses alternative methods in terms of faithfulness and enables the understanding of latent representations, opening up the door for concept-based explanations. We provide an LRP library at https://github.com/rachtibat/LRP-eXplains-Transformers."
                },
                {
                    "title": "Mamba-UNet: UNet-Like Pure Visual Mamba for Medical Image Segmentation",
                    "abstract": "In recent advancements in medical image analysis, Convolutional Neural Networks (CNN) and Vision Transformers (ViT) have set significant benchmarks. While the former excels in capturing local features through its convolution operations, the latter achieves remarkable global context understanding by leveraging self-attention mechanisms. However, both architectures exhibit limitations in efficiently modeling long-range dependencies within medical images, which is a critical aspect for precise segmentation. Inspired by the Mamba architecture, known for its proficiency in handling long sequences and global contextual information with enhanced computational efficiency as a State Space Model (SSM), we propose Mamba-UNet, a novel architecture that synergizes the U-Net in medical image segmentation with Mamba's capability. Mamba-UNet adopts a pure Visual Mamba (VMamba)-based encoder-decoder structure, infused with skip connections to preserve spatial information across different scales of the network. This design facilitates a comprehensive feature learning process, capturing intricate details and broader semantic contexts within medical images. We introduce a novel integration mechanism within the VMamba blocks to ensure seamless connectivity and information flow between the encoder and decoder paths, enhancing the segmentation performance. We conducted experiments on publicly available ACDC MRI Cardiac segmentation dataset, and Synapse CT Abdomen segmentation dataset. The results show that Mamba-UNet outperforms several types of UNet in medical image segmentation under the same hyper-parameter setting. The source code and baseline implementations are available."
                },
                {
                    "title": "Swin-UMamba: Mamba-based UNet with ImageNet-based pretraining",
                    "abstract": "Accurate medical image segmentation demands the integration of multi-scale information, spanning from local features to global dependencies. However, it is challenging for existing methods to model long-range global information, where convolutional neural networks (CNNs) are constrained by their local receptive fields, and vision transformers (ViTs) suffer from high quadratic complexity of their attention mechanism. Recently, Mamba-based models have gained great attention for their impressive ability in long sequence modeling. Several studies have demonstrated that these models can outperform popular vision models in various tasks, offering higher accuracy, lower memory consumption, and less computational burden. However, existing Mamba-based models are mostly trained from scratch and do not explore the power of pretraining, which has been proven to be quite effective for data-efficient medical image analysis. This paper introduces a novel Mamba-based model, Swin-UMamba, designed specifically for medical image segmentation tasks, leveraging the advantages of ImageNet-based pretraining. Our experimental results reveal the vital role of ImageNet-based training in enhancing the performance of Mamba-based models. Swin-UMamba demonstrates superior performance with a large margin compared to CNNs, ViTs, and latest Mamba-based models. Notably, on AbdomenMRI, Encoscopy, and Microscopy datasets, Swin-UMamba outperforms its closest counterpart U-Mamba_Enc by an average score of 2.72%."
                },
                {
                    "title": "nnMamba: 3D Biomedical Image Segmentation, Classification and Landmark Detection with State Space Model",
                    "abstract": "In the field of biomedical image analysis, the quest for architectures capable of effectively capturing long-range dependencies is paramount, especially when dealing with 3D image segmentation, classification, and landmark detection. Traditional Convolutional Neural Networks (CNNs) struggle with locality respective field, and Transformers have a heavy computational load when applied to high-dimensional medical images.In this paper, we introduce nnMamba, a novel architecture that integrates the strengths of CNNs and the advanced long-range modeling capabilities of State Space Sequence Models (SSMs). Specifically, we propose the Mamba-In-Convolution with Channel-Spatial Siamese learning (MICCSS) block to model the long-range relationship of the voxels. For the dense prediction and classification tasks, we also design the channel-scaling and channel-sequential learning methods. Extensive experiments on 6 datasets demonstrate nnMamba's superiority over state-of-the-art methods in a suite of challenging tasks, including 3D image segmentation, classification, and landmark detection. nnMamba emerges as a robust solution, offering both the local representation ability of CNNs and the efficient global context processing of SSMs, setting a new standard for long-range dependency modeling in medical image analysis. Code is available at https://github.com/lhaof/nnMamba"
                },
                {
                    "title": "VM-UNet: Vision Mamba UNet for Medical Image Segmentation",
                    "abstract": "In the realm of medical image segmentation, both CNN-based and Transformer-based models have been extensively explored. However, CNNs exhibit limitations in long-range modeling capabilities, whereas Transformers are hampered by their quadratic computational complexity. Recently, State Space Models (SSMs), exemplified by Mamba, have emerged as a promising approach. They not only excel in modeling long-range interactions but also maintain a linear computational complexity. In this paper, leveraging state space models, we propose a U-shape architecture model for medical image segmentation, named Vision Mamba UNet (VM-UNet). Specifically, the Visual State Space (VSS) block is introduced as the foundation block to capture extensive contextual information, and an asymmetrical encoder-decoder structure is constructed. We conduct comprehensive experiments on the ISIC17, ISIC18, and Synapse datasets, and the results indicate that VM-UNet performs competitively in medical image segmentation tasks. To our best knowledge, this is the first medical image segmentation model constructed based on the pure SSM-based model. We aim to establish a baseline and provide valuable insights for the future development of more efficient and effective SSM-based segmentation systems. Our code is available at https://github.com/JCruan519/VM-UNet."
                },
                {
                    "title": "Graph-Mamba: Towards Long-Range Graph Sequence Modeling with Selective State Spaces",
                    "abstract": "Attention mechanisms have been widely used to capture long-range dependencies among nodes in Graph Transformers. Bottlenecked by the quadratic computational cost, attention mechanisms fail to scale in large graphs. Recent improvements in computational efficiency are mainly achieved by attention sparsification with random or heuristic-based graph subsampling, which falls short in data-dependent context reasoning. State space models (SSMs), such as Mamba, have gained prominence for their effectiveness and efficiency in modeling long-range dependencies in sequential data. However, adapting SSMs to non-sequential graph data presents a notable challenge. In this work, we introduce Graph-Mamba, the first attempt to enhance long-range context modeling in graph networks by integrating a Mamba block with the input-dependent node selection mechanism. Specifically, we formulate graph-centric node prioritization and permutation strategies to enhance context-aware reasoning, leading to a substantial improvement in predictive performance. Extensive experiments on ten benchmark datasets demonstrate that Graph-Mamba outperforms state-of-the-art methods in long-range graph prediction tasks, with a fraction of the computational cost in both FLOPs and GPU memory consumption. The code and models are publicly available at https://github.com/bowang-lab/Graph-Mamba."
                },
                {
                    "title": "BlackMamba: Mixture of Experts for State-Space Models",
                    "abstract": "State-space models (SSMs) have recently demonstrated competitive performance to transformers at large-scale language modeling benchmarks while achieving linear time and memory complexity as a function of sequence length. Mamba, a recently released SSM model, shows impressive performance in both language modeling and long sequence processing tasks. Simultaneously, mixture-of-expert (MoE) models have shown remarkable performance while significantly reducing the compute and latency costs of inference at the expense of a larger memory footprint. In this paper, we present BlackMamba, a novel architecture that combines the Mamba SSM with MoE to obtain the benefits of both. We demonstrate that BlackMamba performs competitively against both Mamba and transformer baselines, and outperforms in inference and training FLOPs. We fully train and open-source 340M/1.5B and 630M/2.8B BlackMamba models on 300B tokens of a custom dataset. We show that BlackMamba inherits and combines both of the benefits of SSM and MoE architectures, combining linear-complexity generation from SSM with cheap and fast inference from MoE. We release all weights, checkpoints, and inference code open-source. Inference code at: https://github.com/Zyphra/BlackMamba"
                },
                {
                    "title": "Vivim: a Video Vision Mamba for Medical Video Segmentation",
                    "abstract": "Medical video segmentation gains increasing attention in clinical practice due to the redundant dynamic references in video frames. However, traditional convolutional neural networks have a limited receptive field and transformer-based networks are mediocre in constructing long-term dependency from the perspective of computational complexity. This bottleneck poses a significant challenge when processing longer sequences in medical video analysis tasks using available devices with limited memory. Recently, state space models (SSMs), famous by Mamba, have exhibited impressive achievements in efficient long sequence modeling, which develops deep neural networks by expanding the receptive field on many vision tasks significantly. Unfortunately, vanilla SSMs failed to simultaneously capture causal temporal cues and preserve non-casual spatial information. To this end, this paper presents a Video Vision Mamba-based framework, dubbed as Vivim, for medical video segmentation tasks. Our Vivim can effectively compress the long-term spatiotemporal representation into sequences at varying scales with our designed Temporal Mamba Block. We also introduce an improved boundary-aware affine constraint across frames to enhance the discriminative ability of Vivim on ambiguous lesions. Extensive experiments on thyroid segmentation, breast lesion segmentation in ultrasound videos, and polyp segmentation in colonoscopy videos demonstrate the effectiveness and efficiency of our Vivim, superior to existing methods. The code is available at: https://github.com/scott-yjyang/Vivim. The dataset will be released once accepted."
                },
                {
                    "title": "MambaByte: Token-free Selective State Space Model",
                    "abstract": "Token-free language models learn directly from raw bytes and remove the inductive bias of subword tokenization. Operating on bytes, however, results in significantly longer sequences. In this setting, standard autoregressive Transformers scale poorly as the effective memory required grows with sequence length. The recent development of the Mamba state space model (SSM) offers an appealing alternative approach with a fixed-sized memory state and efficient decoding. We propose MambaByte, a token-free adaptation of the Mamba SSM trained autoregressively on byte sequences. In terms of modeling, we show MambaByte to be competitive with, and even to outperform, state-of-the-art subword Transformers on language modeling tasks while maintaining the benefits of token-free language models, such as robustness to noise. In terms of efficiency, we develop an adaptation of speculative decoding with tokenized drafting and byte-level verification. This results in a $2.6\\times$ inference speedup to the standard MambaByte implementation, showing similar decoding efficiency as the subword Mamba. These findings establish the viability of SSMs in enabling token-free language modeling."
                },
                {
                    "title": "SegMamba: Long-range Sequential Modeling Mamba For 3D Medical Image Segmentation",
                    "abstract": "The Transformer architecture has shown a remarkable ability in modeling global relationships. However, it poses a significant computational challenge when processing high-dimensional medical images. This hinders its development and widespread adoption in this task. Mamba, as a State Space Model (SSM), recently emerged as a notable manner for long-range dependencies in sequential modeling, excelling in natural language processing filed with its remarkable memory efficiency and computational speed. Inspired by its success, we introduce SegMamba, a novel 3D medical image \\textbf{Seg}mentation \\textbf{Mamba} model, designed to effectively capture long-range dependencies within whole volume features at every scale. Our SegMamba, in contrast to Transformer-based methods, excels in whole volume feature modeling from a state space model standpoint, maintaining superior processing speed, even with volume features at a resolution of {$64\\times 64\\times 64$}. Comprehensive experiments on the BraTS2023 dataset demonstrate the effectiveness and efficiency of our SegMamba. The code for SegMamba is available at: https://github.com/ge-xing/SegMamba"
                },
                {
                    "title": "VMamba: Visual State Space Model",
                    "abstract": "Designing computationally efficient network architectures persists as an ongoing necessity in computer vision. In this paper, we transplant Mamba, a state-space language model, into VMamba, a vision backbone that works in linear time complexity. At the core of VMamba lies a stack of Visual State-Space (VSS) blocks with the 2D Selective Scan (SS2D) module. By traversing along four scanning routes, SS2D helps bridge the gap between the ordered nature of 1D selective scan and the non-sequential structure of 2D vision data, which facilitates the gathering of contextual information from various sources and perspectives. Based on the VSS blocks, we develop a family of VMamba architectures and accelerate them through a succession of architectural and implementation enhancements. Extensive experiments showcase VMamba's promising performance across diverse visual perception tasks, highlighting its advantages in input scaling efficiency compared to existing benchmark models. Source code is available at https://github.com/MzeroMiko/VMamba."
                },
                {
                    "title": "Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model",
                    "abstract": "Recently the state space models (SSMs) with efficient hardware-aware designs, i.e., the Mamba deep learning model, have shown great potential for long sequence modeling. Meanwhile building efficient and generic vision backbones purely upon SSMs is an appealing direction. However, representing visual data is challenging for SSMs due to the position-sensitivity of visual data and the requirement of global context for visual understanding. In this paper, we show that the reliance on self-attention for visual representation learning is not necessary and propose a new generic vision backbone with bidirectional Mamba blocks (Vim), which marks the image sequences with position embeddings and compresses the visual representation with bidirectional state space models. On ImageNet classification, COCO object detection, and ADE20k semantic segmentation tasks, Vim achieves higher performance compared to well-established vision transformers like DeiT, while also demonstrating significantly improved computation&memory efficiency. For example, Vim is 2.8$\\times$ faster than DeiT and saves 86.8% GPU memory when performing batch inference to extract features on images with a resolution of 1248$\\times$1248. The results demonstrate that Vim is capable of overcoming the computation&memory constraints on performing Transformer-style understanding for high-resolution images and it has great potential to be the next-generation backbone for vision foundation models. Code is available at https://github.com/hustvl/Vim."
                },
                {
                    "title": "Decoupling Pixel Flipping and Occlusion Strategy for Consistent XAI Benchmarks",
                    "abstract": "Feature removal is a central building block for eXplainable AI (XAI), both for occlusion-based explanations (Shapley values) as well as their evaluation (pixel flipping, PF). However, occlusion strategies can vary significantly from simple mean replacement up to inpainting with state-of-the-art diffusion models. This ambiguity limits the usefulness of occlusion-based approaches. For example, PF benchmarks lead to contradicting rankings. This is amplified by competing PF measures: Features are either removed starting with most influential first (MIF) or least influential first (LIF). This study proposes two complementary perspectives to resolve this disagreement problem. Firstly, we address the common criticism of occlusion-based XAI, that artificial samples lead to unreliable model evaluations. We propose to measure the reliability by the R(eference)-Out-of-Model-Scope (OMS) score. The R-OMS score enables a systematic comparison of occlusion strategies and resolves the disagreement problem by grouping consistent PF rankings. Secondly, we show that the insightfulness of MIF and LIF is conversely dependent on the R-OMS score. To leverage this, we combine the MIF and LIF measures into the symmetric relevance gain (SRG) measure. This breaks the inherent connection to the underlying occlusion strategy and leads to consistent rankings. This resolves the disagreement problem, which we verify for a set of 40 different occlusion strategies."
                },
                {
                    "title": "U-Mamba: Enhancing Long-range Dependency for Biomedical Image Segmentation",
                    "abstract": "Convolutional Neural Networks (CNNs) and Transformers have been the most popular architectures for biomedical image segmentation, but both of them have limited ability to handle long-range dependencies because of inherent locality or computational complexity. To address this challenge, we introduce U-Mamba, a general-purpose network for biomedical image segmentation. Inspired by the State Space Sequence Models (SSMs), a new family of deep sequence models known for their strong capability in handling long sequences, we design a hybrid CNN-SSM block that integrates the local feature extraction power of convolutional layers with the abilities of SSMs for capturing the long-range dependency. Moreover, U-Mamba enjoys a self-configuring mechanism, allowing it to automatically adapt to various datasets without manual intervention. We conduct extensive experiments on four diverse tasks, including the 3D abdominal organ segmentation in CT and MR images, instrument segmentation in endoscopy images, and cell segmentation in microscopy images. The results reveal that U-Mamba outperforms state-of-the-art CNN-based and Transformer-based segmentation networks across all tasks. This opens new avenues for efficient long-range dependency modeling in biomedical image analysis. The code, models, and data are publicly available at https://wanglab.ai/u-mamba.html."
                },
                {
                    "title": "MoE-Mamba: Efficient Selective State Space Models with Mixture of Experts",
                    "abstract": "State Space Models (SSMs) have become serious contenders in the field of sequential modeling, challenging the dominance of Transformers. At the same time, Mixture of Experts (MoE) has significantly improved Transformer-based Large Language Models, including recent state-of-the-art open models. We propose that to unlock the potential of SSMs for scaling, they should be combined with MoE. We showcase this on Mamba, a recent SSM-based model that achieves remarkable performance. Our model, MoE-Mamba, outperforms both Mamba and baseline Transformer-MoE. In particular, MoE-Mamba reaches the same performance as Mamba in $2.35\\times$ fewer training steps while preserving the inference performance gains of Mamba against Transformer."
                },
                {
                    "title": "Mamba: Linear-Time Sequence Modeling with Selective State Spaces",
                    "abstract": "Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5$\\times$ higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation."
                },
                {
                    "title": "Diffusion Models Without Attention",
                    "abstract": "In recent advancements in high-fidelity image generation, Denoising Diffusion Probabilistic Models (DDPMs) have emerged as a key player. However, their application at high resolutions presents significant computational challenges. Current methods, such as patchifying, expedite processes in UNet and Transformer architectures but at the expense of rep-resentational capacity. Addressing this, we introduce the Dif-fusion State Space Model (DIFFUSSM), an architecture that supplants attention mechanisms with a more scalable state space model backbone. This approach effectively handles higher resolutions without resorting to global compression, thus preserving detailed image representation throughout the diffusion process. Our focus on FLOP-efficient architectures in diffusion training marks a significant step forward. Comprehensive evaluations on both ImageNet and LSUN datasets at two resolutions demonstrate that DiffuSSMs are on par or even outperform existing diffusion models with attention modules in FID and Inception Score metrics while significantly reducing total FLOP usage."
                },
                {
                    "title": "Rather a Nurse than a Physician - Contrastive Explanations under Investigation",
                    "abstract": "Contrastive explanations, where one decision is explained in contrast to another, are supposed to be closer to how humans explain a decision than non-contrastive explanations, where the decision is not necessarily referenced to an alternative. This claim has never been empirically validated. We analyze four English text-classification datasets (SST2, DynaSent, BIOS and DBpedia-Animals). We fine-tune and extract explanations from three different models (RoBERTa, GTP-2, and T5), each in three different sizes and apply three post-hoc explainability methods (LRP, GradientxInput, GradNorm). We furthermore collect and release human rationale annotations for a subset of 100 samples from the BIOS dataset for contrastive and non-contrastive settings. A cross-comparison between model-based rationales and human annotations, both in contrastive and non-contrastive settings, yields a high agreement between the two settings for models as well as for humans. Moreover, model-based explanations computed in both settings align equally well with human rationales. Thus, we empirically find that humans do not necessarily explain in a contrastive manner.9 pages, long paper at ACL 2022 proceedings."
                },
                {
                    "title": "LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding",
                    "abstract": "Although large language models (LLMs) demonstrate impressive performance for many language tasks, most of them can only handle texts a few thousand tokens long, limiting their applications on longer sequence inputs, such as books, reports, and codebases. Recent works have proposed methods to improve LLMs' long context capabilities by extending context windows and more sophisticated memory mechanisms. However, comprehensive benchmarks tailored for evaluating long context understanding are lacking. In this paper, we introduce LongBench, the first bilingual, multi-task benchmark for long context understanding, enabling a more rigorous evaluation of long context understanding. LongBench comprises 21 datasets across 6 task categories in both English and Chinese, with an average length of 6,711 words (English) and 13,386 characters (Chinese). These tasks cover key long-text application areas including single-doc QA, multi-doc QA, summarization, few-shot learning, synthetic tasks, and code completion. All datasets in LongBench are standardized into a unified format, allowing for effortless automatic evaluation of LLMs. Upon comprehensive evaluation of 8 LLMs on LongBench, we find that: (1) Commercial model (GPT-3.5-Turbo-16k) outperforms other open-sourced models, but still struggles on longer contexts. (2) Scaled position embedding and fine-tuning on longer sequences lead to substantial improvement on long context understanding. (3) Context compression technique such as retrieval brings improvement for model with weak ability on long contexts, but the performance still lags behind models that have strong long context understanding capability. The code and datasets are available at https://github.com/THUDM/LongBench."
                },
                {
                    "title": "2-D SSM: A General Spatial Layer for Visual Transformers",
                    "abstract": "A central objective in computer vision is to design models with appropriate 2-D inductive bias. Desiderata for 2D inductive bias include two-dimensional position awareness, dynamic spatial locality, and translation and permutation invariance. To address these goals, we leverage an expressive variation of the multidimensional State Space Model (SSM). Our approach introduces efficient parameterization, accelerated computation, and a suitable normalization scheme. Empirically, we observe that incorporating our layer at the beginning of each transformer block of Vision Transformers (ViT) significantly enhances performance for multiple ViT backbones and across datasets. The new layer is effective even with a negligible amount of additional parameters and inference time. Ablation studies and visualizations demonstrate that the layer has a strong 2-D inductive bias. For example, vision transformers equipped with our layer exhibit effective performance even without positional encoding"
                },
                {
                    "title": "Decision S4: Efficient Sequence-Based RL via State Spaces Layers",
                    "abstract": "Recently, sequence learning methods have been applied to the problem of off-policy Reinforcement Learning, including the seminal work on Decision Transformers, which employs transformers for this task. Since transformers are parameter-heavy, cannot benefit from history longer than a fixed window size, and are not computed using recurrence, we set out to investigate the suitability of the S4 family of models, which are based on state-space layers and have been shown to outperform transformers, especially in modeling long-range dependencies. In this work we present two main algorithms: (i) an off-policy training procedure that works with trajectories, while still maintaining the training efficiency of the S4 model. (ii) An on-policy training procedure that is trained in a recurrent manner, benefits from long-range dependencies, and is based on a novel stable actor-critic mechanism. Our results indicate that our method outperforms multiple variants of decision transformers, as well as the other baseline methods on most tasks, while reducing the latency, number of parameters, and training time by several orders of magnitude, making our approach more suitable for real-world RL."
                },
                {
                    "title": "Landmark Attention: Random-Access Infinite Context Length for Transformers",
                    "abstract": "While Transformers have shown remarkable success in natural language processing, their attention mechanism's large memory requirements have limited their ability to handle longer contexts. Prior approaches, such as recurrent memory or retrieval-based augmentation, have either compromised the random-access flexibility of attention (i.e., the capability to select any token in the entire context) or relied on separate mechanisms for relevant context retrieval, which may not be compatible with the model's attention. In this paper, we present a novel approach that allows access to the complete context while retaining random-access flexibility, closely resembling running attention on the entire context. Our method uses a landmark token to represent each block of the input and trains the attention to use it for selecting relevant blocks, enabling retrieval of blocks directly through the attention mechanism instead of by relying on a separate mechanism. Our approach seamlessly integrates with specialized data structures and the system's memory hierarchy, enabling processing of arbitrarily long context lengths. We demonstrate that our method can obtain comparable performance with Transformer-XL while significantly reducing the number of retrieved tokens in each step. Finally, we show that fine-tuning LLaMA 7B with our method successfully extends its context length capacity to over 32k tokens, allowing for inference at the context lengths of GPT-4. We release the implementation of landmark attention and the code to reproduce our experiments at https://github.com/epfml/landmark-attention/."
                },
                {
                    "title": "Selective Structured State-Spaces for Long-Form Video Understanding",
                    "abstract": "Effective modeling of complex spatiotemporal dependencies in long-form videos remains an open problem. The recently proposed Structured State-Space Sequence ($S4$) model with its linear complexity offers a promising direction in this space. However, we demonstrate that treating all imagetokens equally as done by $S4$ model can adversely affect its efficiency and accuracy. To address this limitation, we present a novel Selective $S4$ (i.e., $S5)$ model that employs a lightweight mask generator to adaptively select informative image tokens resulting in more efficient and accurate modeling of long-term spatiotemporal dependencies in videos. Unlike previous mask-based token reduction methods used in transformers, our $S5$ model avoids the dense self-attention calculation by making use of the guidance of the momentum-updated $S4$ model. This enables our model to efficiently discard less informative tokens and adapt to various long-form video understanding tasks more effectively. However, as is the case for most token reduction methods, the informative image tokens could be dropped incorrectly. To improve the robustness and the temporal horizon of our model, we propose a novel long-short masked contrastive learning (LSMCL) approach that enables our model to predict longer temporal context using shorter input videos. We present extensive comparative results using three challenging long-form video understanding datasets (LVU, COIN and Breakfast), demonstrating that our approach consistently outperforms the previous state-of-the-art S4 model by up to 9.6% accuracy while reducing its memory footprint by 23%."
                },
                {
                    "title": "Structured State Space Models for In-Context Reinforcement Learning",
                    "abstract": "Structured state space sequence (S4) models have recently achieved state-of-the-art performance on long-range sequence modeling tasks. These models also have fast inference speeds and parallelisable training, making them potentially useful in many reinforcement learning settings. We propose a modification to a variant of S4 that enables us to initialise and reset the hidden state in parallel, allowing us to tackle reinforcement learning tasks. We show that our modified architecture runs asymptotically faster than Transformers in sequence length and performs better than RNN's on a simple memory-based task. We evaluate our modified architecture on a set of partially-observable environments and find that, in practice, our model outperforms RNN's while also running over five times faster. Then, by leveraging the model's ability to handle long-range sequences, we achieve strong performance on a challenging meta-learning task in which the agent is given a randomly-sampled continuous control environment, combined with a randomly-sampled linear projection of the environment's observations and actions. Furthermore, we show the resulting model can adapt to out-of-distribution held-out tasks. Overall, the results presented in this paper show that structured state space models are fast and performant for in-context reinforcement learning tasks. We provide code at https://github.com/luchris429/popjaxrl."
                },
                {
                    "title": "Diagonal State Space Augmented Transformers for Speech Recognition",
                    "abstract": "We improve on the popular conformer architecture by replacing the depthwise temporal convolutions with diagonal state space (DSS) models. DSS is a recently introduced variant of linear RNNs obtained by discretizing a linear dynamical system with a diagonal state transition matrix. DSS layers project the input sequence onto a space of orthogonal polynomials where the choice of basis functions, metric and support is controlled by the eigenvalues of the transition matrix. We compare neural transducers with either conformer or our proposed DSS-augmented transformer (DSSformer) encoders on three public corpora: Switchboard English conversational telephone speech 300 hours, Switchboard+Fisher 2000 hours, and a spoken archive of holocaust survivor testimonials called MALACH 176 hours. On Switchboard 300/2000 hours, we reach a single model performance of 8.9%/6.7% WER on the combined test set of the Hub5 2000 evaluation, respectively, and on MALACH we improve the WER by 7% relative over the previous best published result. In addition, we present empirical evidence suggesting that DSS layers learn damped Fourier basis functions where the attenuation coefficients are layer specific whereas the frequency coefficients converge to almost identical linearly-spaced values across all layers."
                },
                {
                    "title": "Inseq: An Interpretability Toolkit for Sequence Generation Models",
                    "abstract": "Past work in natural language processing interpretability focused mainly on popular classification tasks while largely overlooking generation settings, partly due to a lack of dedicated tools. In this work, we introduce Inseq, a Python library to democratize access to interpretability analyses of sequence generation models. Inseq enables intuitive and optimized extraction of models\u2019 internal information and feature importance scores for popular decoder-only and encoder-decoder Transformers architectures. We showcase its potential by adopting it to highlight gender biases in machine translation models and locate factual knowledge inside GPT-2. Thanks to its extensible interface supporting cutting-edge techniques such as contrastive feature attribution, Inseq can drive future advances in explainable natural language generation, centralizing good practices and enabling fair and reproducible model evaluations."
                },
                {
                    "title": "Disentangled Explanations of Neural Network Predictions by Finding Relevant Subspaces",
                    "abstract": "Explainable AI aims to overcome the black-box nature of complex ML models like neural networks by generating explanations for their predictions. Explanations often take the form of a heatmap identifying input features (e.g. pixels) that are relevant to the model's decision. These explanations, however, entangle the potentially multiple factors that enter into the overall complex decision strategy. We propose to disentangle explanations by extracting at some intermediate layer of a neural network, subspaces that capture the multiple and distinct activation patterns (e.g. visual concepts) that are relevant to the prediction. To automatically extract these subspaces, we propose two new analyses, extending principles found in PCA or ICA to explanations. These novel analyses, which we call principal relevant component analysis (PRCA) and disentangled relevant subspace analysis (DRSA), maximize relevance instead of e.g. variance or kurtosis. This allows for a much stronger focus of the analysis on what the ML model actually uses for predicting, ignoring activations or concepts to which the model is invariant. Our approach is general enough to work alongside common attribution techniques such as Shapley Value, Integrated Gradients, or LRP. Our proposed methods show to be practically useful and compare favorably to the state of the art as demonstrated on benchmarks and three use cases."
                },
                {
                    "title": "Hungry Hungry Hippos: Towards Language Modeling with State Space Models",
                    "abstract": "State space models (SSMs) have demonstrated state-of-the-art sequence modeling performance in some modalities, but underperform attention in language modeling. Moreover, despite scaling nearly linearly in sequence length instead of quadratically, SSMs are still slower than Transformers due to poor hardware utilization. In this paper, we make progress on understanding the expressivity gap between SSMs and attention in language modeling, and on reducing the hardware barrier between SSMs and attention. First, we use synthetic language modeling tasks to understand the gap between SSMs and attention. We find that existing SSMs struggle with two capabilities: recalling earlier tokens in the sequence and comparing tokens across the sequence. To understand the impact on language modeling, we propose a new SSM layer, H3, that is explicitly designed for these abilities. H3 matches attention on the synthetic languages and comes within 0.4 PPL of Transformers on OpenWebText. Furthermore, a hybrid 125M-parameter H3-attention model that retains two attention layers surprisingly outperforms Transformers on OpenWebText by 1.0 PPL. Next, to improve the efficiency of training SSMs on modern hardware, we propose FlashConv. FlashConv uses a fused block FFT algorithm to improve efficiency on sequences up to 8K, and introduces a novel state passing algorithm that exploits the recurrent properties of SSMs to scale to longer sequences. FlashConv yields 2$\\times$ speedup on the long-range arena benchmark and allows hybrid language models to generate text 2.4$\\times$ faster than Transformers. Using FlashConv, we scale hybrid H3-attention language models up to 2.7B parameters on the Pile and find promising initial results, achieving lower perplexity than Transformers and outperforming Transformers in zero- and few-shot learning on a majority of tasks in the SuperGLUE benchmark."
                },
                {
                    "title": "Simplified State Space Layers for Sequence Modeling",
                    "abstract": "Models using structured state space sequence (S4) layers have achieved state-of-the-art performance on long-range sequence modeling tasks. An S4 layer combines linear state space models (SSMs), the HiPPO framework, and deep learning to achieve high performance. We build on the design of the S4 layer and introduce a new state space layer, the S5 layer. Whereas an S4 layer uses many independent single-input, single-output SSMs, the S5 layer uses one multi-input, multi-output SSM. We establish a connection between S5 and S4, and use this to develop the initialization and parameterization used by the S5 model. The result is a state space layer that can leverage efficient and widely implemented parallel scans, allowing S5 to match the computational efficiency of S4, while also achieving state-of-the-art performance on several long-range sequence modeling tasks. S5 averages 87.4% on the long range arena benchmark, and 98.5% on the most difficult Path-X task."
                },
                {
                    "title": "Long Range Language Modeling via Gated State Spaces",
                    "abstract": "State space models have shown to be effective at modeling long range dependencies, specially on sequence classification tasks. In this work we focus on autoregressive sequence modeling over English books, Github source code and ArXiv mathematics articles. Based on recent developments around the effectiveness of gated activation functions, we propose a new layer named Gated State Space (GSS) and show that it trains significantly faster than the diagonal version of S4 (i.e. DSS) on TPUs, is fairly competitive with several well-tuned Transformer-based baselines and exhibits zero-shot generalization to longer inputs while being straightforward to implement. Finally, we show that leveraging self-attention to model local dependencies improves the performance of GSS even further."
                },
                {
                    "title": "Interpreting Language Models with Contrastive Explanations",
                    "abstract": "Model interpretability methods are often used to explain NLP model decisions on tasks such as text classification, where the output space is relatively small. However, when applied to language generation, where the output space often consists of tens of thousands of tokens, these methods are unable to provide informative explanations. Language models must consider various features to predict a token, such as its part of speech, number, tense, or semantics.Existing explanation methods conflate evidence for all these features into a single explanation, which is less interpretable for human understanding.To disentangle the different decisions in language modeling, we focus on explaining language models contrastively: we look for salient input tokens that explain why the model predicted one token instead of another. We demonstrate that contrastive explanations are quantifiably better than non-contrastive explanations in verifying major grammatical phenomena, and that they significantly improve contrastive model simulatability for human observers. We also identify groups of contrastive decisions where the model uses similar evidence, and we are able to characterize what input tokens models use during various language generation decisions."
                },
                {
                    "title": "XAI for Transformers: Better Explanations through Conservative Propagation",
                    "abstract": "Transformers have become an important workhorse of machine learning, with numerous applications. This necessitates the development of reliable methods for increasing their transparency. Multiple interpretability methods, often based on gradient information, have been proposed. We show that the gradient in a Transformer reflects the function only locally, and thus fails to reliably identify the contribution of input features to the prediction. We identify Attention Heads and LayerNorm as main reasons for such unreliable explanations and propose a more stable way for propagation through these layers. Our proposal, which can be seen as a proper extension of the well-established LRP method to Transformers, is shown both theoretically and empirically to overcome the deficiency of a simple gradient-based approach, and achieves state-of-the-art explanation performance on a broad range of Transformer models and datasets."
                },
                {
                    "title": "Efficiently Modeling Long Sequences with Structured State Spaces",
                    "abstract": "A central goal of sequence modeling is designing a single principled model that can address sequence data across a range of modalities and tasks, particularly on long-range dependencies. Although conventional models including RNNs, CNNs, and Transformers have specialized variants for capturing long dependencies, they still struggle to scale to very long sequences of $10000$ or more steps. A promising recent approach proposed modeling sequences by simulating the fundamental state space model (SSM) \\( x'(t) = Ax(t) + Bu(t), y(t) = Cx(t) + Du(t) \\), and showed that for appropriate choices of the state matrix \\( A \\), this system could handle long-range dependencies mathematically and empirically. However, this method has prohibitive computation and memory requirements, rendering it infeasible as a general sequence modeling solution. We propose the Structured State Space sequence model (S4) based on a new parameterization for the SSM, and show that it can be computed much more efficiently than prior approaches while preserving their theoretical strengths. Our technique involves conditioning \\( A \\) with a low-rank correction, allowing it to be diagonalized stably and reducing the SSM to the well-studied computation of a Cauchy kernel. S4 achieves strong empirical results across a diverse range of established benchmarks, including (i) 91\\% accuracy on sequential CIFAR-10 with no data augmentation or auxiliary losses, on par with a larger 2-D ResNet, (ii) substantially closing the gap to Transformers on image and language modeling tasks, while performing generation $60\\times$ faster (iii) SoTA on every task from the Long Range Arena benchmark, including solving the challenging Path-X task of length 16k that all prior work fails on, while being as efficient as all competitors."
                },
                {
                    "title": "Combining Recurrent, Convolutional, and Continuous-time Models with Linear State-Space Layers",
                    "abstract": "Recurrent neural networks (RNNs), temporal convolutions, and neural differential equations (NDEs) are popular families of deep learning models for time-series data, each with unique strengths and tradeoffs in modeling power and computational efficiency. We introduce a simple sequence model inspired by control systems that generalizes these approaches while addressing their shortcomings. The Linear State-Space Layer (LSSL) maps a sequence $u \\mapsto y$ by simply simulating a linear continuous-time state-space representation $\\dot{x} = Ax + Bu, y = Cx + Du$. Theoretically, we show that LSSL models are closely related to the three aforementioned families of models and inherit their strengths. For example, they generalize convolutions to continuous-time, explain common RNN heuristics, and share features of NDEs such as time-scale adaptation. We then incorporate and generalize recent theory on continuous-time memorization to introduce a trainable subset of structured matrices $A$ that endow LSSLs with long-range memory. Empirically, stacking LSSL layers into a simple deep neural network obtains state-of-the-art results across time series benchmarks for long dependencies in sequential image classification, real-world healthcare regression tasks, and speech. On a difficult speech classification task with length-16000 sequences, LSSL outperforms prior approaches by 24 accuracy points, and even outperforms baselines that use hand-crafted features on 100x shorter sequences."
                },
                {
                    "title": "AST: Audio Spectrogram Transformer",
                    "abstract": "In the past decade, convolutional neural networks (CNNs) have been widely adopted as the main building block for end-to-end audio classification models, which aim to learn a direct mapping from audio spectrograms to corresponding labels. To better capture long-range global context, a recent trend is to add a self-attention mechanism on top of the CNN, forming a CNN-attention hybrid model. However, it is unclear whether the reliance on a CNN is necessary, and if neural networks purely based on attention are sufficient to obtain good performance in audio classification. In this paper, we answer the question by introducing the Audio Spectrogram Transformer (AST), the first convolution-free, purely attention-based model for audio classification. We evaluate AST on various audio classification benchmarks, where it achieves new state-of-the-art results of 0.485 mAP on AudioSet, 95.6% accuracy on ESC-50, and 98.1% accuracy on Speech Commands V2."
                },
                {
                    "title": "Generic Attention-model Explainability for Interpreting Bi-Modal and Encoder-Decoder Transformers",
                    "abstract": "Transformers are increasingly dominating multi-modal reasoning tasks, such as visual question answering, achieving state-of-the-art results thanks to their ability to contextualize information using the self-attention and co-attention mechanisms. These attention modules also play a role in other computer vision tasks including object detection and image segmentation. Unlike Transformers that only use self-attention, Transformers with co-attention require to consider multiple attention maps in parallel in order to highlight the information that is relevant to the prediction in the model\u2019s input. In this work, we propose the first method to explain prediction by any Transformer-based architecture, including bi-modal Transformers and Transformers with co-attentions. We provide generic solutions and apply these to the three most commonly used of these architectures: (i) pure self-attention, (ii) self-attention combined with co-attention, and (iii) encoder-decoder attention. We show that our method is superior to all existing methods which are adapted from single modality explainability. Our code is available at: https://github.com/hila-chefer/Transformer-MM-Explainability."
                },
                {
                    "title": "Explaining Deep Neural Networks and Beyond: A Review of Methods and Applications",
                    "abstract": "With the broader and highly successful usage of machine learning (ML) in industry and the sciences, there has been a growing demand for explainable artificial intelligence (XAI). Interpretability and explanation methods for gaining a better understanding of the problem-solving abilities and strategies of nonlinear ML, in particular, deep neural networks, are, therefore, receiving increased attention. In this work, we aim to: 1) provide a timely overview of this active emerging field, with a focus on \u201cpost hoc\u201d explanations, and explain its theoretical foundations; 2) put interpretability algorithms to a test both from a theory and comparative evaluation perspective using extensive simulations; 3) outline best practice aspects, i.e., how to best include interpretation methods into the standard usage of ML; and 4) demonstrate successful usage of XAI in a representative selection of application scenarios. Finally, we discuss challenges and possible future directions of this exciting foundational field of ML."
                },
                {
                    "title": "Transformer Interpretability Beyond Attention Visualization",
                    "abstract": "Self-attention techniques, and specifically Transformers, are dominating the field of text processing and are becoming increasingly popular in computer vision classification tasks. In order to visualize the parts of the image that led to a certain classification, existing methods either rely on the obtained attention maps or employ heuristic propagation along the attention graph. In this work, we propose a novel way to compute relevancy for Transformer networks. The method assigns local relevance based on the Deep Taylor Decomposition principle and then propagates these relevancy scores through the layers. This propagation involves attention layers and skip connections, which challenge existing methods. Our solution is based on a specific formulation that is shown to maintain the total relevancy across layers. We benchmark our method on very recent visual Transformer networks, as well as on a text classification problem, and demonstrate a clear advantage over the existing explainability methods. Our code is available at: https://github.com/hila-chefer/Transformer-Explainability."
                },
                {
                    "title": "Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting",
                    "abstract": "Many real-world applications require the prediction of long sequence time-series, such as electricity consumption planning. Long sequence time-series forecasting (LSTF) demands a high prediction capacity of the model, which is the ability to capture precise long-range dependency coupling between output and input efficiently. Recent studies have shown the potential of Transformer to increase the prediction capacity. However, there are several severe issues with Transformer that prevent it from being directly applicable to LSTF, including quadratic time complexity, high memory usage, and inherent limitation of the encoder-decoder architecture. To address these issues, we design an efficient transformer-based model for LSTF, named Informer, with three distinctive characteristics: (i) a ProbSparse self-attention mechanism, which achieves O(L log L) in time complexity and memory usage, and has comparable performance on sequences' dependency alignment. (ii) the self-attention distilling highlights dominating attention by halving cascading layer input, and efficiently handles extreme long input sequences. (iii) the generative style decoder, while conceptually simple, predicts the long time-series sequences at one forward operation rather than a step-by-step way, which drastically improves the inference speed of long-sequence predictions. Extensive experiments on four large-scale datasets demonstrate that Informer significantly outperforms existing methods and provides a new solution to the LSTF problem."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Quantifying Attention Flow in Transformers",
                    "abstract": "In the Transformer model, \u201cself-attention\u201d combines information from attended embeddings into the representation of the focal embedding in the next layer. Thus, across layers of the Transformer, information originating from different tokens gets increasingly mixed. This makes attention weights unreliable as explanations probes. In this paper, we consider the problem of quantifying this flow of information through self-attention. We propose two methods for approximating the attention to input tokens given attention weights, attention rollout and attention flow, as post hoc methods when we use attention weights as the relative relevance of the input tokens. We show that these methods give complementary views on the flow of information, and compared to raw attention, both yield higher correlations with importance scores of input tokens obtained using an ablation method and input gradients."
                },
                {
                    "title": "Attention is not not Explanation",
                    "abstract": "Attention mechanisms play a central role in NLP systems, especially within recurrent neural network (RNN) models. Recently, there has been increasing interest in whether or not the intermediate representations offered by these modules may be used to explain the reasoning for a model\u2019s prediction, and consequently reach insights regarding the model\u2019s decision-making process. A recent paper claims that \u2018Attention is not Explanation\u2019 (Jain and Wallace, 2019). We challenge many of the assumptions underlying this work, arguing that such a claim depends on one\u2019s definition of explanation, and that testing it needs to take into account all elements of the model. We propose four alternative tests to determine when/whether attention can be used as explanation: a simple uniform-weights baseline; a variance calibration based on multiple random seed runs; a diagnostic framework using frozen weights from pretrained models; and an end-to-end adversarial attention training protocol. Each allows for meaningful interpretation of attention mechanisms in RNN models. We show that even when reliable adversarial distributions can be found, they don\u2019t perform well on the simple diagnostic, indicating that prior work does not disprove the usefulness of attention mechanisms for explainability."
                },
                {
                    "title": "DARPA's Explainable Artificial Intelligence (XAI) Program",
                    "abstract": "Dramatic success in machine learning has led to a new wave of AI applications (for example, transportation, security, medicine, finance, defense) that offer tremendous benefits but cannot explain their decisions and actions to human users. DARPA\u2019s explainable artificial intelligence (XAI) program endeavors to create AI systems whose learned models and decisions can be understood and appropriately trusted by end users. Realizing this goal requires methods for learning more explainable models, designing effective explanation interfaces, and understanding the psychologic requirements for effective explanations. The XAI developer teams are addressing the first two challenges by creating ML techniques and developing principles, strategies, and human-computer interaction techniques for generating effective explanations. Another XAI team is addressing the third challenge by summarizing, extending, and applying psychologic theories of explanation to help the XAI evaluator define a suitable evaluation framework, which the developer teams will use to test their systems. The XAI teams completed the first of this 4-year program in May 2018. In a series of ongoing evaluations, the developer teams are assessing how well their XAM systems\u2019 explanations improve user understanding, user trust, and user task performance."
                },
                {
                    "title": "Attention is not Explanation",
                    "abstract": "Attention mechanisms have seen wide adoption in neural NLP models. In addition to improving predictive performance, these are often touted as affording transparency: models equipped with attention provide a distribution over attended-to input units, and this is often presented (at least implicitly) as communicating the relative importance of inputs. However, it is unclear what relationship exists between attention weights and model outputs. In this work we perform extensive experiments across a variety of NLP tasks that aim to assess the degree to which attention weights provide meaningful \u201cexplanations\u201d for predictions. We find that they largely do not. For example, learned attention weights are frequently uncorrelated with gradient-based measures of feature importance, and one can identify very different attention distributions that nonetheless yield equivalent predictions. Our findings show that standard attention modules do not provide meaningful explanations and should not be treated as though they do."
                },
                {
                    "title": "HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering",
                    "abstract": "Existing question answering (QA) datasets fail to train QA systems to perform complex reasoning and provide explanations for answers. We introduce HotpotQA, a new dataset with 113k Wikipedia-based question-answer pairs with four key features: (1) the questions require finding and reasoning over multiple supporting documents to answer; (2) the questions are diverse and not constrained to any pre-existing knowledge bases or knowledge schemas; (3) we provide sentence-level supporting facts required for reasoning, allowing QA systems to reason with strong supervision and explain the predictions; (4) we offer a new type of factoid comparison questions to test QA systems\u2019 ability to extract relevant facts and perform necessary comparison. We show that HotpotQA is challenging for the latest QA systems, and the supporting facts enable models to improve performance and make explainable predictions."
                },
                {
                    "title": "SmoothGrad: removing noise by adding noise",
                    "abstract": "Explaining the output of a deep network remains a challenge. In the case of an image classifier, one type of explanation is to identify pixels that strongly influence the final decision. A starting point for this strategy is the gradient of the class score function with respect to the input image. This gradient can be interpreted as a sensitivity map, and there are several techniques that elaborate on this basic idea. This paper makes two contributions: it introduces SmoothGrad, a simple method that can help visually sharpen gradient-based sensitivity maps, and it discusses lessons in the visualization of these maps. We publish the code for our experiments and a website with our results."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Interpretable Explanations of Black Boxes by Meaningful Perturbation",
                    "abstract": "As machine learning algorithms are increasingly applied to high impact yet high risk tasks, such as medical diagnosis or autonomous driving, it is critical that researchers can explain how such algorithms arrived at their predictions. In recent years, a number of image saliency methods have been developed to summarize where highly complex neural networks \u201clook\u201d in an image for evidence for their predictions. However, these techniques are limited by their heuristic nature and architectural constraints. In this paper, we make two main contributions: First, we propose a general framework for learning different kinds of explanations for any black box algorithm. Second, we specialise the framework to find the part of an image most responsible for a classifier decision. Unlike previous works, our method is model-agnostic and testable because it is grounded in explicit and interpretable image perturbations."
                },
                {
                    "title": "Learning Important Features Through Propagating Activation Differences",
                    "abstract": "The purported \"black box\" nature of neural networks is a barrier to adoption in applications where interpretability is essential. Here we present DeepLIFT (Deep Learning Important FeaTures), a method for decomposing the output prediction of a neural network on a specific input by backpropagating the contributions of all neurons in the network to every feature of the input. DeepLIFT compares the activation of each neuron to its 'reference activation' and assigns contribution scores according to the difference. By optionally giving separate consideration to positive and negative contributions, DeepLIFT can also reveal dependencies which are missed by other approaches. Scores can be computed efficiently in a single backward pass. We apply DeepLIFT to models trained on MNIST and simulated genomic data, and show significant advantages over gradient-based methods. Video tutorial: http://goo.gl/qKb7pL, code: http://goo.gl/RM8jvH."
                },
                {
                    "title": "Axiomatic Attribution for Deep Networks",
                    "abstract": "We study the problem of attributing the prediction of a deep network to its input features, a problem previously studied by several other works. We identify two fundamental axioms\u2014 Sensitivity and Implementation Invariance that attribution methods ought to satisfy. We show that they are not satisfied by most known attribution methods, which we consider to be a fundamental weakness of those methods. We use the axioms to guide the design of a new attribution method called Integrated Gradients. Our method requires no modification to the original network and is extremely simple to implement; it just needs a few calls to the standard gradient operator. We apply this method to a couple of image models, a couple of text models and a chemistry model, demonstrating its ability to debug networks, to extract rules from a network, and to enable users to engage with models better."
                },
                {
                    "title": "Visualizing Deep Neural Network Decisions: Prediction Difference Analysis",
                    "abstract": "This article presents the prediction difference analysis method for visualizing the response of a deep neural network to a specific input. When classifying images, the method highlights areas in a given input image that provide evidence for or against a certain class. It overcomes several shortcoming of previous methods and provides great additional insight into the decision making process of classifiers. Making neural network decisions interpretable through visualization is important both to improve models and to accelerate the adoption of black-box classifiers in application areas such as medicine. We illustrate the method in experiments on natural images (ImageNet data), as well as medical images (MRI brain scans)."
                },
                {
                    "title": "Evaluating the Visualization of What a Deep Neural Network Has Learned",
                    "abstract": "Deep neural networks (DNNs) have demonstrated impressive performance in complex machine learning tasks such as image classification or speech recognition. However, due to their multilayer nonlinear structure, they are not transparent, i.e., it is hard to grasp what makes them arrive at a particular classification or recognition decision, given a new unseen data sample. Recently, several approaches have been proposed enabling one to understand and interpret the reasoning embodied in a DNN for a single test image. These methods quantify the \u201cimportance\u201d of individual pixels with respect to the classification decision and allow a visualization in terms of a heatmap in pixel/input space. While the usefulness of heatmaps can be judged subjectively by a human, an objective quality measure is missing. In this paper, we present a general methodology based on region perturbation for evaluating ordered collections of pixels such as heatmaps. We compare heatmaps computed by three different methods on the SUN397, ILSVRC2012, and MIT Places data sets. Our main result is that the recently proposed layer-wise relevance propagation algorithm qualitatively and quantitatively provides a better explanation of what made a DNN arrive at a particular classification decision than the sensitivity-based approach or the deconvolution method. We provide theoretical arguments to explain this result and discuss its practical implications. Finally, we investigate the use of heatmaps for unsupervised assessment of the neural network performance."
                },
                {
                    "title": "A large annotated corpus for learning natural language inference",
                    "abstract": "Understanding entailment and contradiction is fundamental to understanding natural language, and inference about entailment and contradiction is a valuable testing ground for the development of semantic representations. However, machine learning research in this area has been dramatically limited by the lack of large-scale resources. To address this, we introduce the Stanford Natural Language Inference corpus, a new, freely available collection of labeled sentence pairs, written by humans doing a novel grounded task based on image captioning. At 570K pairs, it is two orders of magnitude larger than all other resources of its type. This increase in scale allows lexicalized classifiers to outperform some sophisticated existing entailment models, and it allows a neural network-based model to perform competitively on natural language inference benchmarks for the first time."
                },
                {
                    "title": "On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation",
                    "abstract": "Understanding and interpreting classification decisions of automated image classification systems is of high value in many applications, as it allows to verify the reasoning of the system and provides additional information to the human expert. Although machine learning methods are solving very successfully a plethora of tasks, they have in most cases the disadvantage of acting as a black box, not providing any information about what made them arrive at a particular decision. This work proposes a general solution to the problem of understanding classification decisions by pixel-wise decomposition of nonlinear classifiers. We introduce a methodology that allows to visualize the contributions of single pixels to predictions for kernel-based classifiers over Bag of Words features and for multilayered neural networks. These pixel contributions can be visualized as heatmaps and are provided to a human expert who can intuitively not only verify the validity of the classification decision, but also focus further analysis on regions of potential interest. We evaluate our method for classifiers trained on PASCAL VOC 2009 images, synthetic image data containing geometric shapes, the MNIST handwritten digits data set and for the pre-trained ImageNet model available as part of the Caffe open source package."
                },
                {
                    "title": "Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank",
                    "abstract": "Semantic word spaces have been very useful but cannot express the meaning of longer phrases in a principled way. Further progress towards understanding compositionality in tasks such as sentiment detection requires richer supervised training and evaluation resources and more powerful models of composition. To remedy this, we introduce a Sentiment Treebank. It includes fine grained sentiment labels for 215,154 phrases in the parse trees of 11,855 sentences and presents new challenges for sentiment compositionality. To address them, we introduce the Recursive Neural Tensor Network. When trained on the new treebank, this model outperforms all previous methods on several metrics. It pushes the state of the art in single sentence positive/negative classification from 80% up to 85.4%. The accuracy of predicting fine-grained sentiment labels for all phrases reaches 80.7%, an improvement of 9.7% over bag of features baselines. Lastly, it is the only model that can accurately capture the effects of negation and its scope at various tree levels for both positive and negative phrases."
                },
                {
                    "title": "How to Explain Individual Classification Decisions",
                    "abstract": "After building a classifier with modern tools of machine learning we typically have a black box at hand that is able to predict well for unseen data. Thus, we get an answer to the question what is the most likely label of a given unseen data point. However, most methods will provide no answer why the model predicted a particular label for a single instance and what features were most influential for that particular instance. The only method that is currently able to provide such explanations are decision trees. This paper proposes a procedure which (based on a set of assumptions) allows to explain the decisions of any classification method."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Semi-Mamba-UNet: Pixel-Level Contrastive Cross-Supervised Visual Mamba-based UNet for Semi-Supervised Medical Image Segmentation",
                    "abstract": ". Medical image segmentation is essential in diagnostics, treat-ment planning, and healthcare, with deep learning offering promising advancements. Notably, Convolutional Neural Network (CNN) excel in capturing local image features, whereas Vision Transformer (ViT) adeptly model long-range dependencies through multi-head self-attention mechanisms. Despite their strengths, both CNN and ViT face challenges in efficiently processing long-range dependencies within medical images, often requiring substantial computational resources. This issue, combined with the high cost and limited availability of expert annotations, poses significant obstacles to achieving precise segmentation. To address these challenges, this paper introduces the Semi-Mamba-UNet, which integrates a visual mamba-based UNet architecture with a conventional UNet into a semi-supervised learning (SSL) framework. This innovative SSL approach leverages dual networks to jointly generate pseudo labels and cross supervise each other, drawing inspiration from consistency regularization techniques. Furthermore, we introduce a self-supervised pixel-level contrastive learning strategy, employing a projector pair to further enhance feature learning capabilities. Our comprehensive evaluation on a publicly available MRI cardiac segmentation dataset, comparing against various SSL frameworks with different UNet-based segmentation networks, highlights the superior performance of Semi-Mamba-UNet. The source code has been made publicly accessible."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "CARER: Contextualized Affect Representations for Emotion Recognition",
                    "abstract": "Emotions are expressed in nuanced ways, which varies by collective or individual experiences, knowledge, and beliefs. Therefore, to understand emotion, as conveyed through text, a robust mechanism capable of capturing and modeling different linguistic nuances and phenomena is needed. We propose a semi-supervised, graph-based algorithm to produce rich structural descriptors which serve as the building blocks for constructing contextualized affect representations from text. The pattern-based representations are further enriched with word embeddings and evaluated through several emotion recognition tasks. Our experimental results demonstrate that the proposed method outperforms state-of-the-art techniques on emotion recognition tasks."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively explain the predictions of Mamba models while ensuring adherence to the conservation property in Layer-wise Relevance Propagation (LRP)?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing need for transparency in machine learning models, particularly in high-stakes applications like medicine. By developing a reliable explanation method for Mamba models, we can enhance trust in AI systems, facilitate better understanding of model behavior, and mitigate biases. This work could pave the way for future research on explainability in structured state space models and inspire the development of more robust and interpretable AI systems across various domains.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the highly non-linear and recurrent structure of Mamba models, which complicates the application of existing explanation methods like LRP. Naive approaches may fail because they do not account for the unique characteristics of the Mamba architecture, potentially violating the conservation property and leading to unreliable explanations. Overcoming these technical obstacles requires a deep understanding of both the model architecture and the intricacies of relevance propagation.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has not adequately addressed the peculiarities of the Mamba architecture in the context of LRP, leading to gaps in reliable explanation methods. Existing solutions often overlook the conservation property, resulting in misleading relevance scores. Barriers such as the complexity of the model structure and the lack of tailored propagation rules have hindered progress. Our approach differs by specifically targeting the layers of the Mamba architecture that violate conservation, thereby improving upon prior work with a theoretically sound and efficient relevance propagation strategy.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, MambaLRP, integrates LRP into the Mamba architecture by analyzing the relevance propagation process across its layers. We will identify specific layers that require tailored relevance propagation strategies to maintain conservation. The dataset will consist of various sequence modeling tasks where Mamba models are applied. We will evaluate the effectiveness of MambaLRP using metrics that assess the fidelity and reliability of the explanations produced. The expected outcomes include a robust explanation framework that provides faithful insights into model predictions while adhering to the conservation principle, ultimately enhancing the interpretability of Mamba models in real-world applications."
            }
        },
        "author_data": {
            "dda3c016-72d1-4a2a-8e50-176dc9677e28": {
                "pk": "dda3c016-72d1-4a2a-8e50-176dc9677e28",
                "name": "Farnoush Rezaei Jafari",
                "collaborators": [
                    "Thomas Schnake",
                    "Jonas Lederer",
                    "Ping Xiong",
                    "Shinichi Nakajima",
                    "Stefan Gugler",
                    "Gr\u00e9goire Montavon",
                    "Klaus-Robert M\u00fcller",
                    "Mohsen Fayyaz",
                    "Soroush Abbasi Koohpayegani",
                    "Sunando Sengupta"
                ],
                "domain": [
                    "Explainable AI",
                    "Vision Transformers",
                    "Natural Language Processing",
                    "Quantum Chemistry"
                ],
                "publications": [
                    {
                        "title": "Towards Symbolic XAI -- Explanation Through Human Understandable Logical Relationships Between Features",
                        "abstract": "Explainable Artificial Intelligence (XAI) plays a crucial role in fostering transparency and trust in AI systems, where traditional XAI approaches typically offer one level of abstraction for explanations, often in the form of heatmaps highlighting single or multiple input features. However, we ask whether abstract reasoning or problem-solving strategies of a model may also be relevant, as these align more closely with how humans approach solutions to problems. We propose a framework, called Symbolic XAI, that attributes relevance to symbolic queries expressing logical relationships between input features, thereby capturing the abstract reasoning behind a model's predictions. The methodology is built upon a simple yet general multi-order decomposition of model predictions. This decomposition can be specified using higher-order propagation-based relevance methods, such as GNN-LRP, or perturbation-based explanation methods commonly used in XAI. The effectiveness of our framework is demonstrated in the domains of natural language processing (NLP), vision, and quantum chemistry (QC), where abstract symbolic domain knowledge is abundant and of significant interest to users. The Symbolic XAI framework provides an understanding of the model's decision-making process that is both flexible for customization by the user and human-readable through logical formulas."
                    },
                    {
                        "title": "Adaptive Token Sampling For Efficient Vision Transformers",
                        "abstract": "While state-of-the-art vision transformer models achieve promising results in image classification, they are computationally expensive and require many GFLOPs. Although the GFLOPs of a vision transformer can be decreased by reducing the number of tokens in the network, there is no setting that is optimal for all input images. In this work, we therefore introduce a differentiable parameter-free Adaptive Token Sampler (ATS) module, which can be plugged into any existing vision transformer architecture. ATS empowers vision transformers by scoring and adaptively sampling significant tokens. As a result, the number of tokens is not constant anymore and varies for each input image. By integrating ATS as an additional layer within the current transformer blocks, we can convert them into much more efficient vision transformers with an adaptive number of tokens. Since ATS is a parameter-free module, it can be added to the off-the-shelf pre-trained vision transformers as a plug and play module, thus reducing their GFLOPs without any additional training. Moreover, due to its differentiable design, one can also train a vision transformer equipped with ATS. We evaluate the efficiency of our module in both image and video classification tasks by adding it to multiple SOTA vision transformers. Our proposed module improves the SOTA by reducing their computational costs (GFLOPs) by 2X, while preserving their accuracy on the ImageNet, Kinetics-400, and Kinetics-600 datasets."
                    }
                ]
            },
            "20d41367-6d29-4656-a51d-d9eb1a75843e": {
                "pk": "20d41367-6d29-4656-a51d-d9eb1a75843e",
                "name": "Gr\u00e9goire Montavon",
                "collaborators": [
                    "Klaus-Robert M\u00fcller",
                    "Wojciech Samek",
                    "Sebastian Bach",
                    "Alexander Binder",
                    "Franziska Horn",
                    "Leila Arras",
                    "Oliver Eberle",
                    "Jacob Kauffmann",
                    "Lukas Ruff",
                    "Marco Cuturi"
                ],
                "domain": [
                    "Deep Learning",
                    "Explainable AI",
                    "Unsupervised Learning",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Learning Feature Hierarchies with Centered Deep Boltzmann Machines",
                        "abstract": "Deep Boltzmann machines are in principle powerful models for extracting the hierarchical structure of data. Unfortunately, attempts to train layers jointly (without greedy layer-wise pretraining) have been largely unsuccessful. We propose a modification of the learning algorithm that initially recenters the output of the activation functions to zero. This modification leads to a better conditioned Hessian and thus makes learning easier. We test the algorithm on real data and demonstrate that our suggestion, the centered deep Boltzmann machine, learns a hierarchy of increasingly abstract representations and a better generative model of data."
                    },
                    {
                        "title": "Methods for Interpreting and Understanding Deep Neural Networks",
                        "abstract": "This paper provides an entry point to the problem of interpreting a deep neural network model and explaining its predictions. It is based on a tutorial given at ICASSP 2017. It introduces some recently proposed techniques of interpretation, along with theory, tricks and recommendations, to make most efficient use of these techniques on real data. It also discusses a number of practical applications."
                    },
                    {
                        "title": "Wasserstein Training of Boltzmann Machines",
                        "abstract": "The Boltzmann machine provides a useful framework to learn highly complex, multimodal and multiscale data distributions that occur in the real world. The default method to learn its parameters consists of minimizing the Kullback-Leibler (KL) divergence from training samples to the Boltzmann model. We propose in this work a novel approach for Boltzmann training which assumes that a meaningful metric between observations is given. This metric can be represented by the Wasserstein distance between distributions, for which we derive a gradient with respect to the model parameters. Minimization of this new Wasserstein objective leads to generative models that are better when considering the metric and that have a cluster-like structure. We demonstrate the practical potential of these models for data completion and denoising, for which the metric between observations plays a crucial role."
                    },
                    {
                        "title": "Explaining NonLinear Classification Decisions with Deep Taylor Decomposition",
                        "abstract": "Nonlinear methods such as Deep Neural Networks (DNNs) are the gold standard for various challenging machine learning problems, e.g., image classification, natural language processing or human action recognition. Although these methods perform impressively well, they have a significant disadvantage, the lack of transparency, limiting the interpretability of the solution and thus the scope of application in practice. Especially DNNs act as black boxes due to their multilayer nonlinear structure. In this paper we introduce a novel methodology for interpreting generic multilayer neural networks by decomposing the network classification decision into contributions of its input elements. Although our focus is on image classification, the method is applicable to a broad set of input data, learning tasks and network architectures. Our method is based on deep Taylor decomposition and efficiently utilizes the structure of the network by backpropagating the explanations from the output to the input layer. We evaluate the proposed method empirically on the MNIST and ILSVRC data sets."
                    },
                    {
                        "title": "Building and Interpreting Deep Similarity Models",
                        "abstract": "Many learning algorithms such as kernel machines, nearest neighbors, clustering, or anomaly detection, are based on the concept of 'distance' or 'similarity'. Before similarities are used for training an actual machine learning model, we would like to verify that they are bound to meaningful patterns in the data. In this paper, we propose to make similarities interpretable by augmenting them with an explanation in terms of input features. We develop BiLRP, a scalable and theoretically founded method to systematically decompose similarity scores on pairs of input features. Our method can be expressed as a composition of LRP explanations, which were shown in previous works to scale to highly nonlinear functions. Through an extensive set of experiments, we demonstrate that BiLRP robustly explains complex similarity models, e.g. built on VGG-16 deep neural network features. Additionally, we apply our method to an open problem in digital humanities: detailed assessment of similarity between historical documents such as astronomical tables. Here again, BiLRP provides insight and brings verifiability into a highly engineered and problem-specific similarity model."
                    },
                    {
                        "title": "Explaining Predictive Uncertainty by Exposing Second-Order Effects",
                        "abstract": "Explainable AI has brought transparency into complex ML blackboxes, enabling, in particular, to identify which features these models use for their predictions. So far, the question of explaining predictive uncertainty, i.e. why a model 'doubts', has been scarcely studied. Our investigation reveals that predictive uncertainty is dominated by second-order effects, involving single features or product interactions between them. We contribute a new method for explaining predictive uncertainty based on these second-order effects. Computationally, our method reduces to a simple covariance computation over a collection of first-order explanations. Our method is generally applicable, allowing for turning common attribution techniques (LRP, Gradient x Input, etc.) into powerful second-order uncertainty explainers, which we call CovLRP, CovGI, etc. The accuracy of the explanations our method produces is demonstrated through systematic quantitative evaluations, and the overall usefulness of our method is demonstrated via two practical showcases."
                    },
                    {
                        "title": "The Clever Hans Effect in Unsupervised Learning",
                        "abstract": "Unsupervised learning has become an essential building block of AI systems. The representations it produces, e.g. in foundation models, are critical to a wide variety of downstream applications. It is therefore important to carefully examine unsupervised models to ensure not only that they produce accurate predictions, but also that these predictions are not \"right for the wrong reasons\", the so-called Clever Hans (CH) effect. Using specially developed Explainable AI techniques, we show for the first time that CH effects are widespread in unsupervised learning. Our empirical findings are enriched by theoretical insights, which interestingly point to inductive biases in the unsupervised learning machine as a primary source of CH effects. Overall, our work sheds light on unexplored risks associated with practical applications of unsupervised learning and suggests ways to make unsupervised learning more robust."
                    },
                    {
                        "title": "Discovering topics in text datasets by visualizing relevant words",
                        "abstract": "When dealing with large collections of documents, it is imperative to quickly get an overview of the texts' contents. In this paper we show how this can be achieved by using a clustering algorithm to identify topics in the dataset and then selecting and visualizing relevant words, which distinguish a group of documents from the rest of the texts, to summarize the contents of the documents belonging to each topic. We demonstrate our approach by discovering trending topics in a collection of New York Times article snippets."
                    },
                    {
                        "title": "From Clustering to Cluster Explanations via Neural Networks",
                        "abstract": "A recent trend in machine learning has been to enrich learned models with the ability to explain their own predictions. The emerging field of Explainable AI (XAI) has so far mainly focused on supervised learning, in particular, deep neural network classifiers. In many practical problems however, label information is not given and the goal is instead to discover the underlying structure of the data, for example, its clusters. While powerful methods exist for extracting the cluster structure in data, they typically do not answer the question why a certain data point has been assigned to a given cluster. We propose a new framework that can, for the first time, explain cluster assignments in terms of input features in an efficient and reliable manner. It is based on the novel insight that clustering models can be rewritten as neural networks - or 'neuralized'. Cluster predictions of the obtained networks can then be quickly and accurately attributed to the input features. Several showcases demonstrate the ability of our method to assess the quality of learned clusters and to extract novel insights from the analyzed data and representations."
                    },
                    {
                        "title": "Higher-Order Explanations of Graph Neural Networks via Relevant Walks",
                        "abstract": "Graph Neural Networks (GNNs) are a popular approach for predicting graph structured data. As GNNs tightly entangle the input graph into the neural network structure, common explainable AI approaches are not applicable. To a large extent, GNNs have remained black-boxes for the user so far. In this paper, we show that GNNs can in fact be naturally explained using higher-order expansions, i.e. by identifying groups of edges that jointly contribute to the prediction. Practically, we find that such explanations can be extracted using a nested attribution scheme, where existing techniques such as layer-wise relevance propagation (LRP) can be applied at each step. The output is a collection of walks into the input graph that are relevant for the prediction. Our novel explanation method, which we denote by GNN-LRP, is applicable to a broad range of graph neural networks and lets us extract practically relevant insights on sentiment analysis of text data, structure-property relationships in quantum chemistry, and image classification."
                    },
                    {
                        "title": "Learning Domain Invariant Representations by Joint Wasserstein Distance Minimization",
                        "abstract": "Domain shifts in the training data are common in practical applications of machine learning; they occur for instance when the data is coming from different sources. Ideally, a ML model should work well independently of these shifts, for example, by learning a domain-invariant representation. However, common ML losses do not give strong guarantees on how consistently the ML model performs for different domains, in particular, whether the model performs well on a domain at the expense of its performance on another domain. In this paper, we build new theoretical foundations for this problem, by contributing a set of mathematical relations between classical losses for supervised ML and the Wasserstein distance in joint space (i.e. representation and output space). We show that classification or regression losses, when combined with a GAN-type discriminator between domains, form an upper-bound to the true Wasserstein distance between domains. This implies a more invariant representation and also more stable prediction performance across domains. Theoretical results are corroborated empirically on several image datasets. Our proposed approach systematically produces the highest minimum classification accuracy across domains, and the most invariant representation."
                    },
                    {
                        "title": "Disentangled Explanations of Neural Network Predictions by Finding Relevant Subspaces",
                        "abstract": "Explainable AI aims to overcome the black-box nature of complex ML models like neural networks by generating explanations for their predictions. Explanations often take the form of a heatmap identifying input features (e.g. pixels) that are relevant to the model's decision. These explanations, however, entangle the potentially multiple factors that enter into the overall complex decision strategy. We propose to disentangle explanations by extracting at some intermediate layer of a neural network, subspaces that capture the multiple and distinct activation patterns (e.g. visual concepts) that are relevant to the prediction. To automatically extract these subspaces, we propose two new analyses, extending principles found in PCA or ICA to explanations. These novel analyses, which we call principal relevant component analysis (PRCA) and disentangled relevant subspace analysis (DRSA), maximize relevance instead of e.g. variance or kurtosis. This allows for a much stronger focus of the analysis on what the ML model actually uses for predicting, ignoring activations or concepts to which the model is invariant. Our approach is general enough to work alongside common attribution techniques such as Shapley Value, Integrated Gradients, or LRP. Our proposed methods show to be practically useful and compare favorably to the state of the art as demonstrated on benchmarks and three use cases."
                    },
                    {
                        "title": "Machine Learning of Molecular Electronic Properties in Chemical Compound Space",
                        "abstract": "The combination of modern scientific computing with electronic structure theory can lead to an unprecedented amount of data amenable to intelligent data analysis for the identification of meaningful, novel, and predictive structure-property relationships. Such relationships enable high-throughput screening for relevant properties in an exponentially growing pool of virtual compounds that are synthetically accessible. Here, we present a machine learning (ML) model, trained on a data base of \\textit{ab initio} calculation results for thousands of organic molecules, that simultaneously predicts multiple electronic ground- and excited-state properties. The properties include atomization energy, polarizability, frontier orbital eigenvalues, ionization potential, electron affinity, and excitation energies. The ML model is based on a deep multi-task artificial neural network, exploiting underlying correlations between various molecular properties. The input is identical to \\emph{ab initio} methods, \\emph{i.e.} nuclear charges and Cartesian coordinates of all atoms. For small organic molecules the accuracy of such a \"Quantum Machine\" is similar, and sometimes superior, to modern quantum-chemical methods---at negligible computational cost."
                    },
                    {
                        "title": "Analyzing Classifiers: Fisher Vectors and Deep Neural Networks",
                        "abstract": "Fisher Vector classifiers and Deep Neural Networks (DNNs) are popular and successful algorithms for solving image classification problems. However, both are generally considered `black box' predictors as the non-linear transformations involved have so far prevented transparent and interpretable reasoning. Recently, a principled technique, Layer-wise Relevance Propagation (LRP), has been developed in order to better comprehend the inherent structured reasoning of complex nonlinear classification models such as Bag of Feature models or DNNs. In this paper we (1) extend the LRP framework also for Fisher Vector classifiers and then use it as analysis tool to (2) quantify the importance of context for classification, (3) qualitatively compare DNNs against FV classifiers in terms of important image regions and (4) detect potential flaws and biases in data. All experiments are performed on the PASCAL VOC 2007 data set."
                    },
                    {
                        "title": "Layer-wise Relevance Propagation for Neural Networks with Local Renormalization Layers",
                        "abstract": "Layer-wise relevance propagation is a framework which allows to decompose the prediction of a deep neural network computed over a sample, e.g. an image, down to relevance scores for the single input dimensions of the sample such as subpixels of an image. While this approach can be applied directly to generalized linear mappings, product type non-linearities are not covered. This paper proposes an approach to extend layer-wise relevance propagation to neural networks with local renormalization layers, which is a very common product-type non-linearity in convolutional neural networks. We evaluate the proposed method for local renormalization layers on the CIFAR-10, Imagenet and MIT Places datasets."
                    },
                    {
                        "title": "Identifying individual facial expressions by deconstructing a neural network",
                        "abstract": "This paper focuses on the problem of explaining predictions of psychological attributes such as attractiveness, happiness, confidence and intelligence from face photographs using deep neural networks. Since psychological attribute datasets typically suffer from small sample sizes, we apply transfer learning with two base models to avoid overfitting. These models were trained on an age and gender prediction task, respectively. Using a novel explanation method we extract heatmaps that highlight the parts of the image most responsible for the prediction. We further observe that the explanation method provides important insights into the nature of features of the base model, which allow one to assess the aptitude of the base model for a given transfer learning task. Finally, we observe that the multiclass model is more feature rich than its binary counterpart. The experimental evaluation is performed on the 2222 images from the 10k US faces dataset containing psychological attribute labels as well as on a subset of KDEF images."
                    },
                    {
                        "title": "Explaining Predictions of Non-Linear Classifiers in NLP",
                        "abstract": "Layer-wise relevance propagation (LRP) is a recently proposed technique for explaining predictions of complex non-linear classifiers in terms of input variables. In this paper, we apply LRP for the first time to natural language processing (NLP). More precisely, we use it to explain the predictions of a convolutional neural network (CNN) trained on a topic categorization task. Our analysis highlights which words are relevant for a specific prediction of the CNN. We compare our technique to standard sensitivity analysis, both qualitatively and quantitatively, using a \"word deleting\" perturbation experiment, a PCA analysis, and various visualizations. All experiments validate the suitability of LRP for explaining the CNN predictions, which is also in line with results reported in recent image classification studies."
                    },
                    {
                        "title": "Exploring text datasets by visualizing relevant words",
                        "abstract": "When working with a new dataset, it is important to first explore and familiarize oneself with it, before applying any advanced machine learning algorithms. However, to the best of our knowledge, no tools exist that quickly and reliably give insight into the contents of a selection of documents with respect to what distinguishes them from other documents belonging to different categories. In this paper we propose to extract `relevant words' from a collection of texts, which summarize the contents of documents belonging to a certain class (or discovered cluster in the case of unlabeled datasets), and visualize them in word clouds to allow for a survey of salient features at a glance. We compare three methods for extracting relevant words and demonstrate the usefulness of the resulting word clouds by providing an overview of the classes contained in a dataset of scientific publications as well as by discovering trending topics from recent New York Times article snippets."
                    }
                ]
            },
            "6fe4ee90-1daf-48cc-9e77-c5e9e708f0f2": {
                "pk": "6fe4ee90-1daf-48cc-9e77-c5e9e708f0f2",
                "name": "Klaus-Robert M\u00fcller",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "80a28264-d404-4b92-ab41-5f8ac1f45c47": {
                "pk": "80a28264-d404-4b92-ab41-5f8ac1f45c47",
                "name": "Oliver Eberle",
                "collaborators": [
                    "Klaus-Robert M\u00fcller",
                    "Stephanie Brandl",
                    "Gr\u00e9goire Montavon",
                    "Thomas Schnake",
                    "Anders S\u00f8gaard",
                    "Jochen B\u00fcttner",
                    "Matteo Valleriani",
                    "Alexandros Vasileiou",
                    "Jonas Pilot",
                    "Ilias Chalkidis"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Explainable AI",
                    "Graph Neural Network",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Explaining Text Similarity in Transformer Models",
                        "abstract": "As Transformers have become state-of-the-art models for natural language processing (NLP) tasks, the need to understand and explain their predictions is increasingly apparent. Especially in unsupervised applications, such as information retrieval tasks, similarity models built on top of foundation model representations have been widely applied. However, their inner prediction mechanisms have mostly remained opaque. Recent advances in explainable AI have made it possible to mitigate these limitations by leveraging improved explanations for Transformers through layer-wise relevance propagation (LRP). Using BiLRP, an extension developed for computing second-order explanations in bilinear similarity models, we investigate which feature interactions drive similarity in NLP models. We validate the resulting explanations and demonstrate their utility in three corpus-level use cases, analyzing grammatical interactions, multilingual semantics, and biomedical text retrieval. Our findings contribute to a deeper understanding of different semantic similarity tasks and models, highlighting how novel explainable AI methods enable in-depth analyses and corpus-level insights."
                    },
                    {
                        "title": "Comparing zero-shot self-explanations with human rationales in multilingual text classification",
                        "abstract": "Instruction-tuned LLMs are able to provide an explanation about their output to users by generating self-explanations that do not require gradient computations or the application of possibly complex XAI methods. In this paper, we analyse whether this ability results in a good explanation by evaluating self-explanations in the form of input rationales with respect to their plausibility to humans as well as their faithfulness to models. For this, we apply two text classification tasks: sentiment classification and forced labour detection. Next to English, we further include Danish and Italian translations of the sentiment classification task and compare self-explanations to human annotations for all samples. To allow for direct comparisons, we also compute post-hoc feature attribution, i.e., layer-wise relevance propagation (LRP) and apply this pipeline to 4 LLMs (Llama2, Llama3, Mistral and Mixtral). Our results show that self-explanations align more closely with human annotations compared to LRP, while maintaining a comparable level of faithfulness."
                    },
                    {
                        "title": "Do Transformer Models Show Similar Attention Patterns to Task-Specific Human Gaze?",
                        "abstract": "Learned self-attention functions in state-of-the-art NLP models often correlate with human attention. We investigate whether self-attention in large-scale pre-trained language models is as predictive of human eye fixation patterns during task-reading as classical cognitive models of human attention. We compare attention functions across two task-specific reading datasets for sentiment analysis and relation extraction. We find the predictiveness of large-scale pre-trained self-attention for human attention depends on `what is in the tail', e.g., the syntactic nature of rare contexts. Further, we observe that task-specific fine-tuning does not increase the correlation with human task-specific reading. Through an input reduction experiment we give complementary insights on the sparsity and fidelity trade-off, showing that lower-entropy attention vectors are more faithful."
                    },
                    {
                        "title": "Rather a Nurse than a Physician -- Contrastive Explanations under Investigation",
                        "abstract": "Contrastive explanations, where one decision is explained in contrast to another, are supposed to be closer to how humans explain a decision than non-contrastive explanations, where the decision is not necessarily referenced to an alternative. This claim has never been empirically validated. We analyze four English text-classification datasets (SST2, DynaSent, BIOS and DBpedia-Animals). We fine-tune and extract explanations from three different models (RoBERTa, GTP-2, and T5), each in three different sizes and apply three post-hoc explainability methods (LRP, GradientxInput, GradNorm). We furthermore collect and release human rationale annotations for a subset of 100 samples from the BIOS dataset for contrastive and non-contrastive settings. A cross-comparison between model-based rationales and human annotations, both in contrastive and non-contrastive settings, yields a high agreement between the two settings for models as well as for humans. Moreover, model-based explanations computed in both settings align equally well with human rationales. Thus, we empirically find that humans do not necessarily explain in a contrastive manner.9 pages, long paper at ACL 2022 proceedings."
                    },
                    {
                        "title": "Insightful analysis of historical sources at scales beyond human capabilities using unsupervised Machine Learning and XAI",
                        "abstract": "Historical materials are abundant. Yet, piecing together how human knowledge has evolved and spread both diachronically and synchronically remains a challenge that can so far only be very selectively addressed. The vast volume of materials precludes comprehensive studies, given the restricted number of human specialists. However, as large amounts of historical materials are now available in digital form there is a promising opportunity for AI-assisted historical analysis. In this work, we take a pivotal step towards analyzing vast historical corpora by employing innovative machine learning (ML) techniques, enabling in-depth historical insights on a grand scale. Our study centers on the evolution of knowledge within the `Sacrobosco Collection' -- a digitized collection of 359 early modern printed editions of textbooks on astronomy used at European universities between 1472 and 1650 -- roughly 76,000 pages, many of which contain astronomic, computational tables. An ML based analysis of these tables helps to unveil important facets of the spatio-temporal evolution of knowledge and innovation in the field of mathematical astronomy in the period, as taught at European universities."
                    },
                    {
                        "title": "Evaluating Webcam-based Gaze Data as an Alternative for Human Rationale Annotations",
                        "abstract": "Rationales in the form of manually annotated input spans usually serve as ground truth when evaluating explainability methods in NLP. They are, however, time-consuming and often biased by the annotation process. In this paper, we debate whether human gaze, in the form of webcam-based eye-tracking recordings, poses a valid alternative when evaluating importance scores. We evaluate the additional information provided by gaze data, such as total reading times, gaze entropy, and decoding accuracy with respect to human rationale annotations. We compare WebQAmGaze, a multilingual dataset for information-seeking QA, with attention and explainability-based importance scores for 4 different multilingual Transformer-based language models (mBERT, distil-mBERT, XLMR, and XLMR-L) and 3 languages (English, Spanish, and German). Our pipeline can easily be applied to other tasks and languages. Our findings suggest that gaze data offers valuable linguistic insights that could be leveraged to infer task difficulty and further show a comparable ranking of explainability methods to that of human rationales."
                    },
                    {
                        "title": "Building and Interpreting Deep Similarity Models",
                        "abstract": "Many learning algorithms such as kernel machines, nearest neighbors, clustering, or anomaly detection, are based on the concept of 'distance' or 'similarity'. Before similarities are used for training an actual machine learning model, we would like to verify that they are bound to meaningful patterns in the data. In this paper, we propose to make similarities interpretable by augmenting them with an explanation in terms of input features. We develop BiLRP, a scalable and theoretically founded method to systematically decompose similarity scores on pairs of input features. Our method can be expressed as a composition of LRP explanations, which were shown in previous works to scale to highly nonlinear functions. Through an extensive set of experiments, we demonstrate that BiLRP robustly explains complex similarity models, e.g. built on VGG-16 deep neural network features. Additionally, we apply our method to an open problem in digital humanities: detailed assessment of similarity between historical documents such as astronomical tables. Here again, BiLRP provides insight and brings verifiability into a highly engineered and problem-specific similarity model."
                    },
                    {
                        "title": "XAI for Transformers: Better Explanations through Conservative Propagation",
                        "abstract": "Transformers have become an important workhorse of machine learning, with numerous applications. This necessitates the development of reliable methods for increasing their transparency. Multiple interpretability methods, often based on gradient information, have been proposed. We show that the gradient in a Transformer reflects the function only locally, and thus fails to reliably identify the contribution of input features to the prediction. We identify Attention Heads and LayerNorm as main reasons for such unreliable explanations and propose a more stable way for propagation through these layers. Our proposal, which can be seen as a proper extension of the well-established LRP method to Transformers, is shown both theoretically and empirically to overcome the deficiency of a simple gradient-based approach, and achieves state-of-the-art explanation performance on a broad range of Transformer models and datasets."
                    },
                    {
                        "title": "xMIL: Insightful Explanations for Multiple Instance Learning in Histopathology",
                        "abstract": "Multiple instance learning (MIL) is an effective and widely used approach for weakly supervised machine learning. In histopathology, MIL models have achieved remarkable success in tasks like tumor detection, biomarker prediction, and outcome prognostication. However, MIL explanation methods are still lagging behind, as they are limited to small bag sizes or disregard instance interactions. We revisit MIL through the lens of explainable AI (XAI) and introduce xMIL, a refined framework with more general assumptions. We demonstrate how to obtain improved MIL explanations using layer-wise relevance propagation (LRP) and conduct extensive evaluation experiments on three toy settings and four real-world histopathology datasets. Our approach consistently outperforms previous explanation attempts with particularly improved faithfulness scores on challenging biomarker prediction tasks. Finally, we showcase how xMIL explanations enable pathologists to extract insights from MIL models, representing a significant advance for knowledge discovery and model debugging in digital histopathology."
                    },
                    {
                        "title": "Higher-Order Explanations of Graph Neural Networks via Relevant Walks",
                        "abstract": "Graph Neural Networks (GNNs) are a popular approach for predicting graph structured data. As GNNs tightly entangle the input graph into the neural network structure, common explainable AI approaches are not applicable. To a large extent, GNNs have remained black-boxes for the user so far. In this paper, we show that GNNs can in fact be naturally explained using higher-order expansions, i.e. by identifying groups of edges that jointly contribute to the prediction. Practically, we find that such explanations can be extracted using a nested attribution scheme, where existing techniques such as layer-wise relevance propagation (LRP) can be applied at each step. The output is a collection of walks into the input graph that are relevant for the prediction. Our novel explanation method, which we denote by GNN-LRP, is applicable to a broad range of graph neural networks and lets us extract practically relevant insights on sentiment analysis of text data, structure-property relationships in quantum chemistry, and image classification."
                    }
                ]
            }
        }
    },
    "2405.13888": {
        "paper_data": {
            "title": "Marrying Causal Representation Learning with Dynamical Systems for Science",
            "url": "http://arxiv.org/abs/2405.13888v1",
            "arxiv_id": "2405.13888",
            "authors": [
                "Dingling Yao",
                "Caroline Muller",
                "Francesco Locatello"
            ],
            "abstract": "Causal representation learning promises to extend causal models to hidden causal variables from raw entangled measurements. However, most progress has focused on proving identifiability results in different settings, and we are not aware of any successful real-world application. At the same time, the field of dynamical systems benefited from deep learning and scaled to countless applications but does not allow parameter identification. In this paper, we draw a clear connection between the two and their key assumptions, allowing us to apply identifiable methods developed in causal representation learning to dynamical systems. At the same time, we can leverage scalable differentiable solvers developed for differential equations to build models that are both identifiable and practical. Overall, we learn explicitly controllable models that isolate the trajectory-specific parameters for further downstream tasks such as out-of-distribution classification or treatment effect estimation. We experiment with a wind simulator with partially known factors of variation. We also apply the resulting model to real-world climate data and successfully answer downstream causal questions in line with existing literature on climate change.",
            "introduction": " Introduction Causal representation learning (CRL) [ 51] focuses on provably retrieving high-level latent variables from low- level data. Recently, there have been many casual representation learning works compiling, in various settings, different theoretical identifiability discussion about different types of neural network-based solvers is provided in \u00a7 5. Discussion Why mechanistic neural networks [45] . As mentioned in \u00a7 4, the ODE solver Fgiven in Cors. 3.1 and 3.2 can be interpreted as the decoder in a traditional representation learning regime; however, several challenges arise when integrating ODE solving in the training loop: First of all, the ODE solver must be differentiable to utilize the automatic differentiation implementation of the state-of-the-art deep learning frameworks; this obstacle has been tacked by the line of work termed NeuralODE , which models the ODE vector field using a neural network thus enable differentiability [ 12,13,25]. Nevertheless, most differentiable ODE solvers solve the ODE autoregressively and thus cannot be parallelized by the GPU very efficiently. Dealing with long-term trajectories (for example, weekly climate data during the last few decades) would be extremely computationally heavy. Therefore, we advocate for a time- and memory-efficient differentiable ODE solver: the mechanistic neural networks [45]. Latent regularizers in CRL. The framework proposed in \u00a7 4 can be generalized to many causal representation learning works, by specifying the latent regularizes according to individual assumptions and settings. For exam- ple, in the multiview setting, the latent regularizer can be the L2alignment between the learned representations on the shared partition eq. (3), as it was assumed that the paired views are generated based on this overlapping set of latents [ 35,60,66]; in sparse causal representation learning the underlying generative process assumes observations are generated from sparse latent variables; therefore, the proposed algorithms actively enforce some sparsity constraint on the learned representation [ 28,29,43,65], We provide a more extensive summary of different causal representation learning approaches and their corresponding latent regularizer in Tab. 6. 19Marrying Causal Representation Learning with Dynamical Systems for Science Table 5: Training setup for sea surface temperature in \u00a7 6.2. Non-applicable fields are marked with \u2717. Ada-GV AE TI-MNN Cont. Identifier Mech. Identifier Pre-process DCT DCT DCT DCT Encoder 6-layer MLP 6-layer MLP 6-layer MLP 6-layer MLP Decoder 6-layer MLP 6-layer MLP \u2717 3 proj. \u00d76-layer MLP Time dim 208 208 208 208 State dim 1 1 1 1 Hidden dim 1024 1024 1024 1024 Latent dim 20 20 20 20 Optimizer Adam Adam Adam Adam Adam: learning rate 1e\u22125 1e \u22125 1e \u22125 1e \u22125 Adam: beta1 0.9 0.9 0.9 0.9 Adam: beta2 0.999 0.999 0.999 0.999 Adam: epsilon 1e\u22128 1e \u22128 1e \u22128 1e \u22128 Batch size 2160 2160 2160 2160 Temperature \u03c4 \u2717 \u2717 0.1 \u2717 Alignment reg. \u2717 \u2717 \u2717 10 # Initial values 10 10 \u2717 10 # Iterations < 30,000 < 30,000 < 30,000 < 30,000 # Seeds 3 3 3 3 Table 6: A non-exhaustive summary of latent regularizers in recent CRL approaches. Principle Assumption Latent regularizer experiments (\u00a7 6) ( Ada-GVAE, TI-MNN, contrastive identifier, mechanistic identifier ) were built upon open-sourced code provided by the original works [ 36,45,66], under the MIT license. For mechanistic identifiers , we add a regularizer multiplier on the alignment constraint (Defn. 3.3), which is shown in Tabs. 3 and 5. C.1 Wind simulation: SpeedyWeather.jl We simulate global air motion using using the ShallowWaterModel from speedy weather",
            "references": [
                {
                    "title": "A Sparsity Principle for Partially Observable Causal Representation Learning",
                    "abstract": "Causal representation learning aims at identifying high-level causal variables from perceptual data. Most methods assume that all latent causal variables are captured in the high-dimensional observations. We instead consider a partially observed setting, in which each measurement only provides information about a subset of the underlying causal state. Prior work has studied this setting with multiple domains or views, each depending on a fixed subset of latents. Here, we focus on learning from unpaired observations from a dataset with an instance-dependent partial observability pattern. Our main contribution is to establish two identifiability results for this setting: one for linear mixing functions without parametric assumptions on the underlying causal model, and one for piecewise linear mixing functions with Gaussian latent causal variables. Based on these insights, we propose two methods for estimating the underlying causal variables by enforcing sparsity in the inferred representation. Experiments on different simulated datasets and established benchmarks highlight the effectiveness of our approach in recovering the ground-truth latents."
                },
                {
                    "title": "Mechanistic Neural Networks for Scientific Machine Learning",
                    "abstract": "This paper presents Mechanistic Neural Networks, a neural network design for machine learning applications in the sciences. It incorporates a new Mechanistic Block in standard architectures to explicitly learn governing differential equations as representations, revealing the underlying dynamics of data and enhancing interpretability and efficiency in data modeling. Central to our approach is a novel Relaxed Linear Programming Solver (NeuRLP) inspired by a technique that reduces solving linear ODEs to solving linear programs. This integrates well with neural networks and surpasses the limitations of traditional ODE solvers enabling scalable GPU parallel processing. Overall, Mechanistic Neural Networks demonstrate their versatility for scientific machine learning applications, adeptly managing tasks from equation discovery to dynamic systems modeling. We prove their comprehensive capabilities in analyzing and interpreting complex scientific data across various applications, showing significant performance against specialized state-of-the-art methods."
                },
                {
                    "title": "Multi-View Causal Representation Learning with Partial Observability",
                    "abstract": "We present a unified framework for studying the identifiability of representations learned from simultaneously observed views, such as different data modalities. We allow a partially observed setting in which each view constitutes a nonlinear mixture of a subset of underlying latent variables, which can be causally related. We prove that the information shared across all subsets of any number of views can be learned up to a smooth bijection using contrastive learning and a single encoder per view. We also provide graphical criteria indicating which latent variables can be identified through a simple set of rules, which we refer to as identifiability algebra. Our general framework and theoretical results unify and extend several previous works on multi-view nonlinear ICA, disentanglement, and causal representation learning. We experimentally validate our claims on numerical, image, and multi-modal data sets. Further, we demonstrate that the performance of prior methods is recovered in different special cases of our setup. Overall, we find that access to multiple partial views enables us to identify a more fine-grained representation, under the generally milder assumption of partial observability."
                },
                {
                    "title": "Predicting Ordinary Differential Equations with Transformers",
                    "abstract": "We develop a transformer-based sequence-to-sequence model that recovers scalar ordinary differential equations (ODEs) in symbolic form from irregularly sampled and noisy observations of a single solution trajectory. We demonstrate in extensive empirical evaluations that our model performs better or on par with existing methods in terms of accurate recovery across various settings. Moreover, our method is efficiently scalable: after one-time pretraining on a large set of ODEs, we can infer the governing law of a new observed solution in a few forward passes of the model."
                },
                {
                    "title": "Identifiability Guarantees for Causal Disentanglement from Soft Interventions",
                    "abstract": "Causal disentanglement aims to uncover a representation of data using latent variables that are interrelated through a causal model. Such a representation is identifiable if the latent model that explains the data is unique. In this paper, we focus on the scenario where unpaired observational and interventional data are available, with each intervention changing the mechanism of a latent variable. When the causal variables are fully observed, statistically consistent algorithms have been developed to identify the causal model under faithfulness assumptions. We here show that identifiability can still be achieved with unobserved causal variables, given a generalized notion of faithfulness. Our results guarantee that we can recover the latent causal model up to an equivalence class and predict the effect of unseen combinations of interventions, in the limit of infinite data. We implement our causal disentanglement framework by developing an autoencoding variational Bayes algorithm and apply it to the problem of predicting combinatorial perturbation effects in genomics."
                },
                {
                    "title": "Additive Decoders for Latent Variables Identification and Cartesian-Product Extrapolation",
                    "abstract": "We tackle the problems of latent variables identification and ``out-of-support'' image generation in representation learning. We show that both are possible for a class of decoders that we call additive, which are reminiscent of decoders used for object-centric representation learning (OCRL) and well suited for images that can be decomposed as a sum of object-specific images. We provide conditions under which exactly solving the reconstruction problem using an additive decoder is guaranteed to identify the blocks of latent variables up to permutation and block-wise invertible transformations. This guarantee relies only on very weak assumptions about the distribution of the latent factors, which might present statistical dependencies and have an almost arbitrarily shaped support. Our result provides a new setting where nonlinear independent component analysis (ICA) is possible and adds to our theoretical understanding of OCRL methods. We also show theoretically that additive decoders can generate novel images by recombining observed factors of variations in novel ways, an ability we refer to as Cartesian-product extrapolation. We show empirically that additivity is crucial for both identifiability and extrapolation on simulated data."
                },
                {
                    "title": "Learning Linear Causal Representations from Interventions under General Nonlinear Mixing",
                    "abstract": "We study the problem of learning causal representations from unknown, latent interventions in a general setting, where the latent distribution is Gaussian but the mixing function is completely general. We prove strong identifiability results given unknown single-node interventions, i.e., without having access to the intervention targets. This generalizes prior works which have focused on weaker classes, such as linear maps or paired counterfactual data. This is also the first instance of causal identifiability from non-paired interventions for deep neural network embeddings. Our proof relies on carefully uncovering the high-dimensional geometric structure present in the data distribution after a non-linear density transformation, which we capture by analyzing quadratic forms of precision matrices of the latent distributions. Finally, we propose a contrastive algorithm to identify the latent variables in practice and evaluate its performance on various tasks."
                },
                {
                    "title": "The Uniqueness Problem of Physical Law Learning",
                    "abstract": "Physical law learning is the ambiguous attempt at automating the derivation of governing equations with the use of machine learning techniques. This paper shall serve as a first step to build a comprehensive theoretical framework for learning physical laws, aiming to provide reliability to according algorithms. One key problem consists in the fact that the governing equations might not be uniquely determined by the given data. We will study this problem in the common situation that a physical law is described by an ordinary or partial differential equation. For various different classes of differential equations, we provide both necessary and sufficient conditions for a function from a given function class to uniquely determine the differential equation which is governing the phenomenon. We then use our results to determine in extensive numerical experiments whether a function solves a differential equation uniquely."
                },
                {
                    "title": "Nonparametric Identifiability of Causal Representations from Unknown Interventions",
                    "abstract": "We study causal representation learning, the task of inferring latent causal variables and their causal relations from high-dimensional mixtures of the variables. Prior work relies on weak supervision, in the form of counterfactual pre- and post-intervention views or temporal structure; places restrictive assumptions, such as linearity, on the mixing function or latent causal model; or requires partial knowledge of the generative process, such as the causal graph or intervention targets. We instead consider the general setting in which both the causal model and the mixing function are nonparametric. The learning signal takes the form of multiple datasets, or environments, arising from unknown interventions in the underlying causal model. Our goal is to identify both the ground truth latents and their causal graph up to a set of ambiguities which we show to be irresolvable from interventional data. We study the fundamental setting of two causal variables and prove that the observational distribution and one perfect intervention per node suffice for identifiability, subject to a genericity condition. This condition rules out spurious solutions that involve fine-tuning of the intervened and observational distributions, mirroring similar conditions for nonlinear cause-effect inference. For an arbitrary number of variables, we show that at least one pair of distinct perfect interventional domains per node guarantees identifiability. Further, we demonstrate that the strengths of causal influences among the latent variables are preserved by all equivalent solutions, rendering the inferred representation appropriate for drawing causal conclusions from new data. Our study provides the first identifiability results for the general nonparametric setting with unknown interventions, and elucidates what is possible and impossible for causal representation learning without more direct supervision."
                },
                {
                    "title": "Simultaneous identification of models and parameters of scientific simulators",
                    "abstract": "Many scientific models are composed of multiple discrete components, and scientists often make heuristic decisions about which components to include. Bayesian inference provides a mathematical framework for systematically selecting model components, but defining prior distributions over model components and developing associated inference schemes has been challenging. We approach this problem in a simulation-based inference framework: We define model priors over candidate components and, from model simulations, train neural networks to infer joint probability distributions over both model components and associated parameters. Our method, simulation-based model inference (SBMI), represents distributions over model components as a conditional mixture of multivariate binary distributions in the Grassmann formalism. SBMI can be applied to any compositional stochastic simulator without requiring likelihood evaluations. We evaluate SBMI on a simple time series model and on two scientific models from neuroscience, and show that it can discover multiple data-consistent model configurations, and that it reveals non-identifiable model components and parameters. SBMI provides a powerful tool for data-driven scientific inquiry which will allow scientists to identify essential model components and make uncertainty-informed modelling decisions."
                },
                {
                    "title": "Identifiability Results for Multimodal Contrastive Learning",
                    "abstract": "Contrastive learning is a cornerstone underlying recent progress in multi-view and multimodal learning, e.g., in representation learning with image/caption pairs. While its effectiveness is not yet fully understood, a line of recent work reveals that contrastive learning can invert the data generating process and recover ground truth latent factors shared between views. In this work, we present new identifiability results for multimodal contrastive learning, showing that it is possible to recover shared factors in a more general setup than the multi-view setting studied previously. Specifically, we distinguish between the multi-view setting with one generative mechanism (e.g., multiple cameras of the same type) and the multimodal setting that is characterized by distinct mechanisms (e.g., cameras and microphones). Our work generalizes previous identifiability results by redefining the generative process in terms of distinct mechanisms with modality-specific latent variables. We prove that contrastive learning can block-identify latent factors shared between modalities, even when there are nontrivial dependencies between factors. We empirically verify our identifiability results with numerical simulations and corroborate our findings on a complex multimodal dataset of image/text pairs. Zooming out, our work provides a theoretical basis for multimodal representation learning and explains in which settings multimodal contrastive learning can be effective in practice."
                },
                {
                    "title": "Unpaired Multi-Domain Causal Representation Learning",
                    "abstract": "The goal of causal representation learning is to find a representation of data that consists of causally related latent variables. We consider a setup where one has access to data from multiple domains that potentially share a causal representation. Crucially, observations in different domains are assumed to be unpaired, that is, we only observe the marginal distribution in each domain but not their joint distribution. In this paper, we give sufficient conditions for identifiability of the joint distribution and the shared causal graph in a linear setup. Identifiability holds if we can uniquely recover the joint distribution and the shared causal representation from the marginal distributions in each domain. We transform our identifiability results into a practical method to recover the shared latent causal graph."
                },
                {
                    "title": "Linear Causal Disentanglement via Interventions",
                    "abstract": "Causal disentanglement seeks a representation of data involving latent variables that relate to one another via a causal model. A representation is identifiable if both the latent model and the transformation from latent to observed variables are unique. In this paper, we study observed variables that are a linear transformation of a linear latent causal model. Data from interventions are necessary for identifiability: if one latent variable is missing an intervention, we show that there exist distinct models that cannot be distinguished. Conversely, we show that a single intervention on each latent variable is sufficient for identifiability. Our proof uses a generalization of the RQ decomposition of a matrix that replaces the usual orthogonal and upper triangular conditions with analogues depending on a partial order on the rows of the matrix, with partial order determined by a latent causal model. We corroborate our theoretical results with a method for causal disentanglement that accurately recovers a latent causal model."
                },
                {
                    "title": "Synergies between Disentanglement and Sparsity: Generalization and Identifiability in Multi-Task Learning",
                    "abstract": "Although disentangled representations are often said to be beneficial for downstream tasks, current empirical and theoretical understanding is limited. In this work, we provide evidence that disentangled representations coupled with sparse base-predictors improve generalization. In the context of multi-task learning, we prove a new identifiability result that provides conditions under which maximally sparse base-predictors yield disentangled representations. Motivated by this theoretical result, we propose a practical approach to learn disentangled representations based on a sparsity-promoting bi-level optimization problem. Finally, we explore a meta-learning version of this algorithm based on group Lasso multiclass SVM base-predictors, for which we derive a tractable dual formulation. It obtains competitive results on standard few-shot classification benchmarks, while each task is using only a fraction of the learned representations."
                },
                {
                    "title": "Interventional Causal Representation Learning",
                    "abstract": "Causal representation learning seeks to extract high-level latent factors from low-level sensory data. Most existing methods rely on observational data and structural assumptions (e.g., conditional independence) to identify the latent factors. However, interventional data is prevalent across applications. Can interventional data facilitate causal representation learning? We explore this question in this paper. The key observation is that interventional data often carries geometric signatures of the latent factors' support (i.e. what values each latent can possibly take). For example, when the latent factors are causally connected, interventions can break the dependency between the intervened latents' support and their ancestors'. Leveraging this fact, we prove that the latent causal factors can be identified up to permutation and scaling given data from perfect $do$ interventions. Moreover, we can achieve block affine identification, namely the estimated latent factors are only entangled with a few other latents if we have access to data from imperfect interventions. These results highlight the unique power of interventional data in causal representation learning; they can enable provable identification of latent factors without any assumptions about their distributions or dependency structure."
                },
                {
                    "title": "Partial Disentanglement via Mechanism Sparsity",
                    "abstract": "Disentanglement via mechanism sparsity was introduced recently as a principled approach to ex-tract latent factors without supervision when the causal graph relating them in time is sparse, and/or when actions are observed and affect them sparsely. However, this theory applies only to ground-truth graphs satisfying a specific criterion. In this work, we introduce a generalization of this theory which applies to any ground-truth graph and specifies qualitatively how disentangled the learned representation is expected to be, via a new equivalence relation over models we call consistency . This equivalence captures which factors are expected to remain entangled and which are not based on the specific form of the ground-truth graph. We call this weaker form of identifiability partial disentanglement . The graphical criterion that allows complete disentanglement, proposed in an earlier work, can be derived as a special case of our theory. Finally, we enforce graph sparsity with constrained optimization and illustrate our theory and algorithm in simulations."
                },
                {
                    "title": "Identifiability of deep generative models without auxiliary information",
                    "abstract": "We prove identifiability of a broad class of deep latent variable models that (a) have universal approximation capabilities and (b) are the decoders of variational autoencoders that are commonly used in practice. Unlike existing work, our analysis does not require weak supervision, auxiliary information, or conditioning in the latent space. Specifically, we show that for a broad class of generative (i.e. unsupervised) models with universal approximation capabilities, the side information $u$ is not necessary: We prove identifiability of the entire generative model where we do not observe $u$ and only observe the data $x$. The models we consider match autoencoder architectures used in practice that leverage mixture priors in the latent space and ReLU/leaky-ReLU activations in the encoder, such as VaDE and MFC-VAE. Our main result is an identifiability hierarchy that significantly generalizes previous work and exposes how different assumptions lead to different\"strengths\"of identifiability, and includes certain\"vanilla\"VAEs with isotropic Gaussian priors as a special case. For example, our weakest result establishes (unsupervised) identifiability up to an affine transformation, and thus partially resolves an open problem regarding model identifiability raised in prior work. These theoretical results are augmented with experiments on both simulated and real data."
                },
                {
                    "title": "On Finite-Sample Identifiability of Contrastive Learning-Based Nonlinear Independent Component Analysis",
                    "abstract": "Nonlinear independent component analysis (nICA) aims at recovering statistically independent latent components that are mixed by unknown nonlinear functions. Central to nICA is the identifiability of the latent components, which had been elusive until very recently. Specifically, Hyv\\\"arinen et al. have shown that the nonlinearly mixed latent components are identifiable (up to often inconsequential ambiguities) under a generalized contrastive learning (GCL) formulation, given that the latent components are independent conditioned on a certain auxiliary variable. The GCL-based identifiability of nICA is elegant, and establishes interesting connections between nICA and popular unsupervised/self-supervised learning paradigms in representation learning, causal learning, and factor disentanglement. However, existing identifiability analyses of nICA all build upon an unlimited sample assumption and the use of ideal universal function learners -- which creates a non-negligible gap between theory and practice. Closing the gap is a nontrivial challenge, as there is a lack of established ``textbook'' routine for finite sample analysis of such unsupervised problems. This work puts forth a finite-sample identifiability analysis of GCL-based nICA. Our analytical framework judiciously combines the properties of the GCL loss function, statistical generalization analysis, and numerical differentiation. Our framework also takes the learning function's approximation error into consideration, and reveals an intuitive trade-off between the complexity and expressiveness of the employed function learner. Numerical experiments are used to validate the theorems."
                },
                {
                    "title": "Causal Representation Learning for Instantaneous and Temporal Effects in Interactive Systems",
                    "abstract": "Causal representation learning is the task of identifying the underlying causal variables and their relations from high-dimensional observations, such as images. Recent work has shown that one can reconstruct the causal variables from temporal sequences of observations under the assumption that there are no instantaneous causal relations between them. In practical applications, however, our measurement or frame rate might be slower than many of the causal effects. This effectively creates\"instantaneous\"effects and invalidates previous identifiability results. To address this issue, we propose iCITRIS, a causal representation learning method that allows for instantaneous effects in intervened temporal sequences when intervention targets can be observed, e.g., as actions of an agent. iCITRIS identifies the potentially multidimensional causal variables from temporal observations, while simultaneously using a differentiable causal discovery method to learn their causal graph. In experiments on three datasets of interactive systems, iCITRIS accurately identifies the causal variables and their causal graph."
                },
                {
                    "title": "Weakly Supervised Representation Learning with Sparse Perturbations",
                    "abstract": "The theory of representation learning aims to build methods that provably invert the data generating process with minimal domain knowledge or any source of supervision. Most prior approaches require strong distributional assumptions on the latent variables and weak supervision (auxiliary information such as timestamps) to provide provable identification guarantees. In this work, we show that if one has weak supervision from observations generated by sparse perturbations of the latent variables--e.g. images in a reinforcement learning environment where actions move individual sprites--identification is achievable under unknown continuous latent distributions. We show that if the perturbations are applied only on mutually exclusive blocks of latents, we identify the latents up to those blocks. We also show that if these perturbation blocks overlap, we identify latents up to the smallest blocks shared across perturbations. Consequently, if there are blocks that intersect in one latent variable only, then such latents are identified up to permutation and scaling. We propose a natural estimation procedure based on this theory and illustrate it on low-dimensional synthetic and image-based experiments."
                },
                {
                    "title": "Weakly supervised causal representation learning",
                    "abstract": "Learning high-level causal representations together with a causal model from unstructured low-level data such as pixels is impossible from observational data alone. We prove under mild assumptions that this representation is however identifiable in a weakly supervised setting. This involves a dataset with paired samples before and after random, unknown interventions, but no further labels. We then introduce implicit latent causal models, variational autoencoders that represent causal variables and causal structure without having to optimize an explicit discrete graph structure. On simple image data, including a novel dataset of simulated robotic manipulation, we demonstrate that such models can reliably identify the causal structure and disentangle causal variables."
                },
                {
                    "title": "CITRIS: Causal Identifiability from Temporal Intervened Sequences",
                    "abstract": "Understanding the latent causal factors of a dynamical system from visual observations is considered a crucial step towards agents reasoning in complex environments. In this paper, we propose CITRIS, a variational autoencoder framework that learns causal representations from temporal sequences of images in which underlying causal factors have possibly been intervened upon. In contrast to the recent literature, CITRIS exploits temporality and observing intervention targets to identify scalar and multidimensional causal factors, such as 3D rotation angles. Furthermore, by introducing a normalizing flow, CITRIS can be easily extended to leverage and disentangle representations obtained by already pretrained autoencoders. Extending previous results on scalar causal factors, we prove identifiability in a more general setting, in which only some components of a causal factor are affected by interventions. In experiments on 3D rendered image sequences, CITRIS outperforms previous methods on recovering the underlying causal variables. Moreover, using pretrained autoencoders, CITRIS can even generalize to unseen instantiations of causal factors, opening future research areas in sim-to-real generalization for causal representation learning."
                },
                {
                    "title": "Identifiable Deep Generative Models via Sparse Decoding",
                    "abstract": "We develop the sparse VAE for unsupervised representation learning on high-dimensional data. The sparse VAE learns a set of latent factors (representations) which summarize the associations in the observed data features. The underlying model is sparse in that each observed feature (i.e. each dimension of the data) depends on a small subset of the latent factors. As examples, in ratings data each movie is only described by a few genres; in text data each word is only applicable to a few topics; in genomics, each gene is active in only a few biological processes. We prove such sparse deep generative models are identifiable: with infinite data, the true model parameters can be learned. (In contrast, most deep generative models are not identifiable.) We empirically study the sparse VAE with both simulated and real data. We find that it recovers meaningful latent factors and has smaller heldout reconstruction error than related methods."
                },
                {
                    "title": "Understanding Latent Correlation-Based Multiview Learning and Self-Supervision: An Identifiability Perspective",
                    "abstract": "Multiple views of data, both naturally acquired (e.g., image and audio) and artificially produced (e.g., via adding different noise to data samples), have proven useful in enhancing representation learning. Natural views are often handled by multiview analysis tools, e.g., (deep) canonical correlation analysis [(D)CCA], while the artificial ones are frequently used in self-supervised learning (SSL) paradigms, e.g., BYOL and Barlow Twins. Both types of approaches often involve learning neural feature extractors such that the embeddings of data exhibit high cross-view correlations. Although intuitive, the effectiveness of correlation-based neural embedding is mostly empirically validated. This work aims to understand latent correlation maximization-based deep multiview learning from a latent component identification viewpoint. An intuitive generative model of multiview data is adopted, where the views are different nonlinear mixtures of shared and private components. Since the shared components are view/distortion-invariant, representing the data using such components is believed to reveal the identity of the samples effectively and robustly. Under this model, latent correlation maximization is shown to guarantee the extraction of the shared components across views (up to certain ambiguities). In addition, it is further shown that the private information in each view can be provably disentangled from the shared using proper regularization design. A finite sample analysis, which has been rare in nonlinear mixture identifiability study, is also presented. The theoretical results and newly designed regularization are tested on a series of tasks."
                },
                {
                    "title": "Self-Supervised Learning with Data Augmentations Provably Isolates Content from Style",
                    "abstract": "Self-supervised representation learning has shown remarkable success in a number of domains. A common practice is to perform data augmentation via hand-crafted transformations intended to leave the semantics of the data invariant. We seek to understand the empirical success of this approach from a theoretical perspective. We formulate the augmentation process as a latent variable model by postulating a partition of the latent representation into a content component, which is assumed invariant to augmentation, and a style component, which is allowed to change. Unlike prior work on disentanglement and independent component analysis, we allow for both nontrivial statistical and causal dependencies in the latent space. We study the identifiability of the latent representation based on pairs of views of the observations and prove sufficient conditions that allow us to identify the invariant content partition up to an invertible mapping in both generative and discriminative settings. We find numerical simulations with dependent latent variables are consistent with our theory. Lastly, we introduce Causal3DIdent, a dataset of high-dimensional, visually complex images with rich causal dependencies, which we use to study the effect of data augmentations performed in practice."
                },
                {
                    "title": "Promoting global stability in data-driven models of quadratic nonlinear dynamics",
                    "abstract": "Modeling realistic fluid and plasma flows is computationally intensive, motivating the use of reduced-order models for a variety of scientific and engineering tasks. However, it is challenging to characterize, much less guarantee, the global stability (i.e., long-time boundedness) of these models. The seminal work of Schlegel and Noack (JFM, 2015) provided a theorem outlining necessary and sufficient conditions to ensure global stability in systems with energy-preserving, quadratic nonlinearities, with the goal of evaluating the stability of projection-based models. In this work, we incorporate this theorem into modern data-driven models obtained via machine learning. First, we propose that this theorem should be a standard diagnostic for the stability of projection-based and data-driven models, examining the conditions under which it holds. Second, we illustrate how to modify the objective function in machine learning algorithms to promote globally stable models, with implications for the modeling of fluid and plasma flows. Specifically, we introduce a modified\"trapping SINDy\"algorithm based on the sparse identification of nonlinear dynamics (SINDy) method. This method enables the identification of models that, by construction, only produce bounded trajectories. The effectiveness and accuracy of this approach are demonstrated on a broad set of examples of varying model complexity and physical origin, including the vortex shedding in the wake of a circular cylinder."
                },
                {
                    "title": "Toward Causal Representation Learning",
                    "abstract": "The two fields of machine learning and graphical causality arose and are developed separately. However, there is, now, cross-pollination and increasing interest in both fields to benefit from the advances of the other. In this article, we review fundamental concepts of causal inference and relate them to crucial open problems of machine learning, including transfer and generalization, thereby assaying how causality can contribute to modern machine learning research. This also applies in the opposite direction: we note that most work in causality starts from the premise that the causal variables are given. A central problem for AI and causality is, thus, causal representation learning, that is, the discovery of high-level causal variables from low-level observations. Finally, we delineate some implications of causality for machine learning and propose key research areas at the intersection of both communities."
                },
                {
                    "title": "Contrastive Learning Inverts the Data Generating Process",
                    "abstract": "Contrastive learning has recently seen tremendous success in self-supervised learning. So far, however, it is largely unclear why the learned representations generalize so effectively to a large variety of downstream tasks. We here prove that feedforward models trained with objectives belonging to the commonly used InfoNCE family learn to implicitly invert the underlying generative model of the observed data. While the proofs make certain statistical assumptions about the generative model, we observe empirically that our findings hold even if these assumptions are severely violated. Our theory highlights a fundamental connection between contrastive learning, generative modeling, and nonlinear independent component analysis, thereby furthering our understanding of the learned representations as well as providing a theoretical foundation to derive more effective contrastive losses."
                },
                {
                    "title": "Improvements of the Daily Optimum Interpolation Sea Surface Temperature (DOISST) Version 2.1",
                    "abstract": "The NOAA/NESDIS/NCEI Daily Optimum Interpolation Sea Surface Temperature (SST), version 2.0, dataset (DOISST v2.0) is a blend of in situ ship and buoy SSTs with satellite SSTs derived from the Advanced Very High Resolution Radiometer (AVHRR). DOISST v2.0 exhibited a cold bias in the Indian, South Pacific, and South Atlantic Oceans that is due to a lack of ingested drifting-buoy SSTs in the system, which resulted from a gradual data format change from the traditional alphanumeric codes (TAC) to the binary universal form for the representation of meteorological data (BUFR). The cold bias against Argo was about \u22120.14\u00b0C on global average and \u22120.28\u00b0C in the Indian Ocean from January 2016 to August 2019. We explored the reasons for these cold biases through six progressive experiments. These experiments showed that the cold biases can be effectively reduced by adjusting ship SSTs with available buoy SSTs, using the latest available ICOADS R3.0.2 derived from merging BUFR and TAC, as well as by including Argo observations above 5-m depth. The impact of using the satellite MetOp-B instead of NOAA-19 was notable for high-latitude oceans but small on global average, since their biases are adjusted using in situ SSTs. In addition, the warm SSTs in the Arctic were improved by applying a freezing point instead of regressed ice-SST proxy. This paper describes an upgraded version, DOISST v2.1, which addresses biases in v2.0. Overall, by updating v2.0 to v2.1, the biases are reduced to \u22120.07\u00b0 and \u22120.14\u00b0C in the global ocean and Indian Ocean, respectively, when compared with independent Argo observations and are reduced to \u22120.04\u00b0 and \u22120.08\u00b0C in the global ocean and Indian Ocean, respectively, when compared with dependent Argo observations. The difference against the Group for High Resolution SST (GHRSST) Multiproduct Ensemble (GMPE) product is reduced from \u22120.09\u00b0 to \u22120.01\u00b0C in the global oceans and from \u22120.20\u00b0 to \u22120.04\u00b0C in the Indian Ocean."
                },
                {
                    "title": "Learning Neural Event Functions for Ordinary Differential Equations",
                    "abstract": "The existing Neural ODE formulation relies on an explicit knowledge of the termination time. We extend Neural ODEs to implicitly defined termination criteria modeled by neural event functions, which can be chained together and differentiated through. Neural Event ODEs are capable of modeling discrete (instantaneous) changes in a continuous-time system, without prior knowledge of when these changes should occur or how many such changes should exist. We test our approach in modeling hybrid discrete- and continuous- systems such as switching dynamical systems and collision in multi-body systems, and we propose simulation-based training of point processes with applications in discrete control."
                },
                {
                    "title": "\"Hey, that's not an ODE\": Faster ODE Adjoints with 12 Lines of Code",
                    "abstract": "Neural differential equations may be trained by backpropagating gradients via the adjoint method, which is another differential equation typically solved using an adaptive-step-size numerical differential equation solver. A proposed step is accepted if its error, \\emph{relative to some norm}, is sufficiently small; else it is rejected, the step is shrunk, and the process is repeated. Here, we demonstrate that the particular structure of the adjoint equations makes the usual choices of norm (such as $L^2$) unnecessarily stringent. By replacing it with a more appropriate (semi)norm, fewer steps are unnecessarily rejected and the backpropagation is made faster. This requires only minor code modifications. Experiments on a wide range of tasks---including time series, generative modeling, and physical control---demonstrate a median improvement of 40% fewer function evaluations. On some problems we see as much as 62% fewer function evaluations, so that the overall training time is roughly halved."
                },
                {
                    "title": "On Second Order Behaviour in Augmented Neural ODEs",
                    "abstract": "Neural Ordinary Differential Equations (NODEs) are a new class of models that transform data continuously through infinite-depth architectures. The continuous nature of NODEs has made them particularly suitable for learning the dynamics of complex physical systems. While previous work has mostly been focused on first order ODEs, the dynamics of many systems, especially in classical physics, are governed by second order laws. In this work, we consider Second Order Neural ODEs (SONODEs). We show how the adjoint sensitivity method can be extended to SONODEs and prove that the optimisation of a first order coupled ODE is equivalent and computationally more efficient. Furthermore, we extend the theoretical understanding of the broader class of Augmented NODEs (ANODEs) by showing they can also learn higher order dynamics with a minimal number of augmented dimensions, but at the cost of interpretability. This indicates that the advantages of ANODEs go beyond the extra space offered by the augmented dimensions, as originally thought. Finally, we compare SONODEs and ANODEs on synthetic and real dynamical systems and demonstrate that the inductive biases of the former generally result in faster training and better performance."
                },
                {
                    "title": "Identification of affine dynamical systems from a single trajectory",
                    "abstract": "For dynamical systems of the form \u1e8b = Ax + c, henceforth called affine systems, we study the problem of determining system parameters A, c from data derived from a single trajectory. We establish necessary and sufficient conditions for identifiability of parameters from either a full trajectory or a set of discrete data points sampled from a single trajectory, expressed as geometrical conditions on the trajectory. We describe conditions under which the system has an equilibrium point and examine the problem of identifiability of the equilibrium point from data. We briefly examine the analytical upper and lower bounds for the maximal permissible uncertainty that will guarantee an inverse with specified qualitative properties of the resulting system (e.g., stability, spirality, and so on). Finally, we illustrate an application of the theory to parameter estimation for a class of nonlinear systems consisting of linear-in-parameters systems and demonstrate that affine approximations can yield more accurate estimates of parameter values than those based on finite differences."
                },
                {
                    "title": "SINDy-PI: a robust algorithm for parallel implicit sparse identification of nonlinear dynamics",
                    "abstract": "Accurately modelling the nonlinear dynamics of a system from measurement data is a challenging yet vital topic. The sparse identification of nonlinear dynamics (SINDy) algorithm is one approach to discover dynamical systems models from data. Although extensions have been developed to identify implicit dynamics, or dynamics described by rational functions, these extensions are extremely sensitive to noise. In this work, we develop SINDy-PI (parallel, implicit), a robust variant of the SINDy algorithm to identify implicit dynamics and rational nonlinearities. The SINDy-PI framework includes multiple optimization algorithms and a principled approach to model selection. We demonstrate the ability of this algorithm to learn implicit ordinary and partial differential equations and conservation laws from limited and noisy data. In particular, we show that the proposed approach is several orders of magnitude more noise robust than previous approaches, and may be used to identify a class of ODE and PDE dynamics that were previously unattainable with SINDy, including for the double pendulum dynamics and simplified model for the Belousov\u2013Zhabotinsky (BZ) reaction."
                },
                {
                    "title": "A Survey of Deep Learning for Scientific Discovery",
                    "abstract": "Over the past few years, we have seen fundamental breakthroughs in core problems in machine learning, largely driven by advances in deep neural networks. At the same time, the amount of data collected in a wide array of scientific domains is dramatically increasing in both size and complexity. Taken together, this suggests many exciting opportunities for deep learning applications in scientific settings. But a significant challenge to this is simply knowing where to start. The sheer breadth and diversity of different deep learning techniques makes it difficult to determine what scientific problems might be most amenable to these methods, or which specific combination of methods might offer the most promising first approach. In this survey, we focus on addressing this central issue, providing an overview of many widely used deep learning models, spanning visual, sequential and graph structured data, associated tasks and different training methods, along with techniques to use deep learning with less data and better interpret these complex models --- two central considerations for many scientific use cases. We also include overviews of the full design process, implementation tips, and links to a plethora of tutorials, research summaries and open-sourced deep learning pipelines and pretrained models, developed by the community. We hope that this survey will help accelerate the use of deep learning across different scientific domains."
                },
                {
                    "title": "Weakly-Supervised Disentanglement Without Compromises",
                    "abstract": "Intelligent agents should be able to learn useful representations by observing changes in their environment. We model such observations as pairs of non-i.i.d. images sharing at least one of the underlying factors of variation. First, we theoretically show that only knowing how many factors have changed, but not which ones, is sufficient to learn disentangled representations. Second, we provide practical algorithms that learn disentangled representations from pairs of images without requiring annotation of groups, individual factors, or the number of factors that have changed. Third, we perform a large-scale empirical study and show that such pairs of observations are sufficient to reliably learn disentangled representations on several benchmark data sets. Finally, we evaluate our learned representations and find that they are simultaneously useful on a diverse suite of tasks, including generalization under covariate shifts, fairness, and abstract reasoning. Overall, our results demonstrate that weak supervision enables learning of useful disentangled representations in realistic scenarios."
                },
                {
                    "title": "The frontier of simulation-based inference",
                    "abstract": "Many domains of science have developed complex simulations to describe phenomena of interest. While these simulations provide high-fidelity models, they are poorly suited for inference and lead to challenging inverse problems. We review the rapidly developing field of simulation-based inference and identify the forces giving additional momentum to the field. Finally, we describe how the frontier is expanding so that a broad audience can appreciate the profound influence these developments may have on science."
                },
                {
                    "title": "Variational Sparse Coding",
                    "abstract": "Unsupervised discovery of interpretable features and controllable generation with highdimensional data are currently major challenges in machine learning, with applications \nin data visualisation, clustering and artificial \ndata synthesis. We propose a model based \non variational auto-encoders (VAEs) in which \ninterpretation is induced through latent space \nsparsity with a mixture of Spike and Slab distributions as prior. We derive an evidence \nlower bound for this model and propose a specific training method for recovering disentangled features as sparse elements in latent vectors. In our experiments, we demonstrate superior disentanglement performance to standard \nVAE approaches when an estimate of the number of true sources of variation is not available \nand objects display different combinations of \nattributes. Furthermore, the new model provides unique capabilities, such as recovering \nfeature exploitation, synthesising samples that \nshare attributes with a given input object and \ncontrolling both discrete and continuous features upon generation."
                },
                {
                    "title": "Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations",
                    "abstract": "The key idea behind the unsupervised learning of disentangled representations is that real-world data is generated by a few explanatory factors of variation which can be recovered by unsupervised learning algorithms. In this paper, we provide a sober look at recent progress in the field and challenge some common assumptions. We first theoretically show that the unsupervised learning of disentangled representations is fundamentally impossible without inductive biases on both the models and the data. Then, we train more than 12000 models covering most prominent methods and evaluation metrics in a reproducible large-scale experimental study on seven different data sets. We observe that while the different methods successfully enforce properties ``encouraged'' by the corresponding losses, well-disentangled models seemingly cannot be identified without supervision. Furthermore, increased disentanglement does not seem to lead to a decreased sample complexity of learning for downstream tasks. Our results suggest that future work on disentanglement learning should be explicit about the role of inductive biases and (implicit) supervision, investigate concrete benefits of enforcing disentanglement of the learned representations, and consider a reproducible experimental setup covering several data sets."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "Fast Gaussian Process Based Gradient Matching for Parameter Identification in Systems of Nonlinear ODEs",
                    "abstract": "Parameter identification and comparison of dynamical systems is a challenging task in many fields. Bayesian approaches based on Gaussian process regression over time-series data have been successfully applied to infer the parameters of a dynamical system without explicitly solving it. While the benefits in computational cost are well established, a rigorous mathematical framework has been missing. We offer a novel interpretation which leads to a better understanding and improvements in state-of-the-art performance in terms of accuracy for nonlinear dynamical systems."
                },
                {
                    "title": "Structural Identifiability of Dynamic Systems Biology Models",
                    "abstract": "A powerful way of gaining insight into biological systems is by creating a nonlinear differential equation model, which usually contains many unknown parameters. Such a model is called structurally identifiable if it is possible to determine the values of its parameters from measurements of the model outputs. Structural identifiability is a prerequisite for parameter estimation, and should be assessed before exploiting a model. However, this analysis is seldom performed due to the high computational cost involved in the necessary symbolic calculations, which quickly becomes prohibitive as the problem size increases. In this paper we show how to analyse the structural identifiability of a very general class of nonlinear models by extending methods originally developed for studying observability. We present results about models whose identifiability had not been previously determined, report unidentifiabilities that had not been found before, and show how to modify those unidentifiable models to make them identifiable. This method helps prevent problems caused by lack of identifiability analysis, which can compromise the success of tasks such as experiment design, parameter estimation, and model-based optimization. The procedure is called STRIKE-GOLDD (STRuctural Identifiability taKen as Extended-Generalized Observability with Lie Derivatives and Decomposition), and it is implemented in a MATLAB toolbox which is available as open source software. The broad applicability of this approach facilitates the analysis of the increasingly complex models used in systems biology and other areas."
                },
                {
                    "title": "Data-driven discovery of partial differential equations",
                    "abstract": "Researchers propose sparse regression for identifying governing partial differential equations for spatiotemporal systems. We propose a sparse regression method capable of discovering the governing partial differential equation(s) of a given system by time series measurements in the spatial domain. The regression framework relies on sparsity-promoting techniques to select the nonlinear and partial derivative terms of the governing equations that most accurately represent the data, bypassing a combinatorially large search through all possible candidate models. The method balances model complexity and regression accuracy by selecting a parsimonious model via Pareto analysis. Time series measurements can be made in an Eulerian framework, where the sensors are fixed spatially, or in a Lagrangian framework, where the sensors move with the dynamics. The method is computationally efficient, robust, and demonstrated to work on a variety of canonical problems spanning a number of scientific domains including Navier-Stokes, the quantum harmonic oscillator, and the diffusion equation. Moreover, the method is capable of disambiguating between potentially nonunique dynamical terms by using multiple time series taken with different initial data. Thus, for a traveling wave, the method can distinguish between a linear wave equation and the Korteweg\u2013de Vries equation, for instance. The method provides a promising new technique for discovering governing equations and physical laws in parameterized spatiotemporal systems, where first-principles derivations are intractable."
                },
                {
                    "title": "Discovering governing equations from data by sparse identification of nonlinear dynamical systems",
                    "abstract": "Significance Understanding dynamic constraints and balances in nature has facilitated rapid development of knowledge and enabled technology, including aircraft, combustion engines, satellites, and electrical power. This work develops a novel framework to discover governing equations underlying a dynamical system simply from data measurements, leveraging advances in sparsity techniques and machine learning. The resulting models are parsimonious, balancing model complexity with descriptive ability while avoiding overfitting. There are many critical data-driven problems, such as understanding cognition from neural recordings, inferring climate patterns, determining stability of financial markets, predicting and suppressing the spread of disease, and controlling turbulence for greener transportation and energy. With abundant data and elusive laws, data-driven discovery of dynamics will continue to play an important role in these efforts. Extracting governing equations from data is a central challenge in many diverse areas of science and engineering. Data are abundant whereas models often remain elusive, as in climate science, neuroscience, ecology, finance, and epidemiology, to name only a few examples. In this work, we combine sparsity-promoting techniques and machine learning with nonlinear dynamical systems to discover governing equations from noisy measurement data. The only assumption about the structure of the model is that there are only a few important terms that govern the dynamics, so that the equations are sparse in the space of possible functions; this assumption holds for many physical systems in an appropriate basis. In particular, we use sparse regression to determine the fewest terms in the dynamic governing equations required to accurately represent the data. This results in parsimonious models that balance accuracy with model complexity to avoid overfitting. We demonstrate the algorithm on a wide range of problems, from simple canonical systems, including linear and nonlinear oscillators and the chaotic Lorenz system, to the fluid vortex shedding behind an obstacle. The fluid example illustrates the ability of this method to discover the underlying dynamics of a system that took experts in the community nearly 30 years to resolve. We also show that this method generalizes to parameterized systems and systems that are time-varying or have external forcing."
                },
                {
                    "title": "Identifiability of Linear and Linear-in-Parameters Dynamical Systems from a Single Trajectory",
                    "abstract": "Certain experiments are nonrepeatable because they result in the destruction or alteration of the system under study, and thus provide data consisting of at most a single trajectory in state space. Before proceeding with parameter estimation for models of such systems, it is important to know whether the model parameters can be uniquely determined, or identified, from idealized (error-free) single trajectory data. In the case of a linear model, we provide precise definitions of several forms of identifiability, and we derive some novel, interrelated conditions that are necessary and sufficient for these forms of identifiability to arise. We also show that the results have a direct extension to a class of nonlinear systems that are linear in parameters. One of our results provides information about identifiability based solely on the geometric structure of an observed trajectory, while other results relate to whether or not there exists an initial condition that yields identifiability of a fixed but unknow..."
                },
                {
                    "title": "On discrete cosine transform",
                    "abstract": "The discrete cosine transform (DCT), introduced by Ahmed, Natarajan and Rao, has been used in many applications of digital signal processing, data compression and information hiding. There are four types of the discrete cosine transform. In simulating the discrete cosine transform, we propose a generalized discrete cosine transform with three parameters, and prove its orthogonality for some new cases. A new type of discrete cosine transform is proposed and its orthogonality is proved. Finally, we propose a generalized discrete W transform with three parameters, and prove its orthogonality for some new cases."
                },
                {
                    "title": "On Identifiability of Nonlinear ODE Models and Applications in Viral Dynamics",
                    "abstract": "Ordinary differential equations (ODE) are a powerful tool for modeling dynamic processes with wide applications in a variety of scientific fields. Over the last 2 decades, ODEs have also emerged as a prevailing tool in various biomedical research fields, especially in infectious disease modeling. In practice, it is important and necessary to determine unknown parameters in ODE models based on experimental data. Identifiability analysis is the first step in determing unknown parameters in ODE models and such analysis techniques for nonlinear ODE models are still under development. In this article, we review identifiability analysis methodologies for nonlinear ODE models developed in the past one to two decades, including structural identifiability analysis, practical identifiability analysis and sensitivity-based identifiability analysis. Some advanced topics and ongoing research are also briefly reviewed. Finally, some examples from modeling viral dynamics of HIV, influenza and hepatitis viruses are given to illustrate how to apply these identifiability analysis methods in practice."
                },
                {
                    "title": "Parameter Estimation for Differential Equation Models Using a Framework of Measurement Error in Regression Models",
                    "abstract": "Differential equation (DE) models are widely used in many scientific fields, including engineering, physics, and biomedical sciences. The so-called \u201cforward problem,\u201d the problem of simulations and predictions of state variables for given parameter values in the DE models, has been extensively studied by mathematicians, physicists, engineers, and other scientists. However, the \u201cinverse problem,\u201d the problem of parameter estimation based on the measurements of output variables, has not been well explored using modern statistical methods, although some least squares\u2013based approaches have been proposed and studied. In this article we propose parameter estimation methods for ordinary differential equation (ODE) models based on the local smoothing approach and a pseudo\u2013least squares (PsLS) principle under a framework of measurement error in regression models. The asymptotic properties of the proposed PsLS estimator are established. We also compare the PsLS method to the corresponding simulation-extrapolation (SIMEX) method and evaluate their finite-sample performances via simulation studies. We illustrate the proposed approach using an application example from an HIV dynamic study."
                },
                {
                    "title": "Machine learning for science: state of the art and future prospects.",
                    "abstract": "Recent advances in machine learning methods, along with successful applications across a wide variety of fields such as planetary science and bioinformatics, promise powerful new tools for practicing scientists. This viewpoint highlights some useful characteristics of modern machine learning methods and their relevance to scientific applications. We conclude with some speculations on near-term progress and promising directions."
                },
                {
                    "title": "Estimation of Regression Coefficients When Some Regressors are not Always Observed",
                    "abstract": "Abstract In applied problems it is common to specify a model for the conditional mean of a response given a set of regressors. A subset of the regressors may be missing for some study subjects either by design or happenstance. In this article we propose a new class of semiparametric estimators, based on inverse probability weighted estimating equations, that are consistent for parameter vector \u03b10 of the conditional mean model when the data are missing at random in the sense of Rubin and the missingness probabilities are either known or can be parametrically modeled. We show that the asymptotic variance of the optimal estimator in our class attains the semiparametric variance bound for the model by first showing that our estimation problem is a special case of the general problem of parameter estimation in an arbitrary semiparametric model in which the data are missing at random and the probability of observing complete data is bounded away from 0, and then deriving a representation for the efficient score..."
                },
                {
                    "title": "Numerical solution of nonlinear differential equations with algebraic constraints I: Convergence results for backward differentiation formulas",
                    "abstract": "In this paper we investigate the behavior of numerical ODE methods for the solution of systems of differential equations coupled with algebraic constraints. Systems of this form arise frequently in the modelling of problems from physics and engineering; we study some particular examples from electrical networks, fluid dynamics and constrained mechanical systems. We show that backward differentiation formulas converge with the expected order of accuracy for these systems."
                },
                {
                    "title": "Statistics and Causal Inference",
                    "abstract": "Abstract Problems involving causal inference have dogged at the heels of statistics since its earliest days. Correlation does not imply causation, and yet causal conclusions drawn from a carefully designed experiment are often valid. What can a statistical model say about causation? This question is addressed by using a particular model for causal inference (Holland and Rubin 1983; Rubin 1974) to critique the discussions of other writers on causation and causal inference. These include selected philosophers, medical researchers, statisticians, econometricians, and proponents of causal modeling."
                },
                {
                    "title": "Structural identifiability",
                    "abstract": "The structural identification problem in dynamic systems is reviewed and new conditions for global identifiability of linear systems are obtained."
                },
                {
                    "title": "Score-based Causal Representation Learning from Interventions: Nonparametric Identifiability",
                    "abstract": "This paper focuses on causal representation learning (CRL) under a general non-parametric causal latent model and a general transformation model that maps the latent data to the observational data. It establishes identifiability and achievability results using two (stochastic) hard uncoupled interventions per node in the latent causal graph. Notably, one does not know which pair of intervention environments have the same node intervened (hence, uncoupled environments). For identifiability, the paper establishes that perfect recovery of the latent causal model and variables is guaranteed under uncoupled interventions. For achievability, an algorithm is designed that uses observational and interventional data and recovers the latent causal model and variables with provable guarantees for the algorithm. This algorithm leverages score variations across different environments to estimate the inverse of the transformer and, subsequently, the latent variables. The analysis, additionally, recovers the existing identifiability result for two hard coupled interventions, that is when metadata about the pair of environments that have the same node intervened is known. It is noteworthy that the existing results on non-parametric identifiability require assumptions on interventions and additional faithfulness assumptions. This paper shows that when observational data is available, additional faithfulness assumptions are unnecessary."
                }
            ],
            "categories": [
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we develop a time- and memory-efficient differentiable ODE solver that integrates with causal representation learning to effectively handle long-term trajectories in scientific data?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the field of causal representation learning (CRL) as it enables the retrieval of high-level latent variables from complex, low-level data. By improving the efficiency of ODE solvers, researchers can analyze long-term scientific data, such as climate patterns, more effectively. This advancement could lead to significant breakthroughs in understanding dynamic systems, enhancing predictive modeling, and fostering practical applications in various scientific domains, ultimately influencing future research directions in both machine learning and applied sciences.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the need for differentiable ODE solvers that can be efficiently integrated into the training loop of deep learning frameworks. Most existing differentiable ODE solvers operate autoregressively, which limits their parallelization capabilities on GPUs, making them computationally expensive for long-term data analysis. Additionally, the integration of mechanistic neural networks introduces complexities in ensuring that the ODE solvers maintain differentiability while also being time- and memory-efficient. Naive approaches may fail due to these computational constraints and the intricate nature of the underlying dynamical systems.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on either differentiability or computational efficiency, often neglecting the need for both in the context of CRL. Existing solutions have not adequately addressed the integration of ODE solvers with CRL frameworks, leading to limitations in handling long-term trajectories. Barriers such as the lack of suitable regularizers and the inability to efficiently process large datasets have hindered progress. Our approach differs by proposing a novel mechanistic neural network framework that combines efficient ODE solving with tailored latent regularizers, addressing these gaps in prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves developing a mechanistic neural network that serves as a differentiable ODE solver, specifically designed to handle long-term trajectory data. We will utilize datasets such as sea surface temperature data and employ metrics like alignment regularization to evaluate the performance of our model. The expected outcomes include improved computational efficiency and accuracy in retrieving latent variables, enabling more effective analysis of complex dynamical systems and advancing the capabilities of causal representation learning in scientific research."
            }
        },
        "author_data": {
            "c64bf740-d279-482a-acc7-d0bc7521f377": {
                "pk": "c64bf740-d279-482a-acc7-d0bc7521f377",
                "name": "Dingling Yao",
                "collaborators": [
                    "Francesco Locatello",
                    "Danru Xu",
                    "S\u00e9bastien Lachapelle",
                    "Sara Magliacane",
                    "Perouz Taslakian",
                    "Julius von K\u00fcgelgen",
                    "Dario Rancati",
                    "Riccardo Cadei",
                    "Marco Fumero",
                    "Jiale Chen"
                ],
                "domain": [
                    "Causal Inference",
                    "Machine Learning",
                    "Active Learning",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "Unifying Causal Representation Learning with the Invariance Principle",
                        "abstract": "Causal representation learning aims at recovering latent causal variables from high-dimensional observations to solve causal downstream tasks, such as predicting the effect of new interventions or more robust classification. A plethora of methods have been developed, each tackling carefully crafted problem settings that lead to different types of identifiability. The folklore is that these different settings are important, as they are often linked to different rungs of Pearl's causal hierarchy, although not all neatly fit. Our main contribution is to show that many existing causal representation learning approaches methodologically align the representation to known data symmetries. Identification of the variables is guided by equivalence classes across different data pockets that are not necessarily causal. This result suggests important implications, allowing us to unify many existing approaches in a single method that can mix and match different assumptions, including non-causal ones, based on the invariances relevant to our application. It also significantly benefits applicability, which we demonstrate by improving treatment effect estimation on real-world high-dimensional ecological data. Overall, this paper clarifies the role of causality assumptions in the discovery of causal variables and shifts the focus to preserving data symmetries."
                    },
                    {
                        "title": "Scalable Mechanistic Neural Networks",
                        "abstract": "We propose Scalable Mechanistic Neural Network (S-MNN), an enhanced neural network framework designed for scientific machine learning applications involving long temporal sequences. By reformulating the original Mechanistic Neural Network (MNN) (Pervez et al., 2024), we reduce the computational time and space complexities from cubic and quadratic with respect to the sequence length, respectively, to linear. This significant improvement enables efficient modeling of long-term dynamics without sacrificing accuracy or interpretability. Extensive experiments demonstrate that S-MNN matches the original MNN in precision while substantially reducing computational resources. Consequently, S-MNN can drop-in replace the original MNN in applications, providing a practical and efficient tool for integrating mechanistic bottlenecks into neural network models of complex dynamical systems."
                    },
                    {
                        "title": "Active Learning in Gaussian Process State Space Model",
                        "abstract": "We investigate active learning in Gaussian Process state-space models (GPSSM). Our problem is to actively steer the system through latent states by determining its inputs such that the underlying dynamics can be optimally learned by a GPSSM. In order that the most informative inputs are selected, we employ mutual information as our active learning criterion. In particular, we present two approaches for the approximation of mutual information for the GPSSM given latent states. The proposed approaches are evaluated in several physical systems where we actively learn the underlying non-linear dynamics represented by the state-space model."
                    },
                    {
                        "title": "Multi-View Causal Representation Learning with Partial Observability",
                        "abstract": "We present a unified framework for studying the identifiability of representations learned from simultaneously observed views, such as different data modalities. We allow a partially observed setting in which each view constitutes a nonlinear mixture of a subset of underlying latent variables, which can be causally related. We prove that the information shared across all subsets of any number of views can be learned up to a smooth bijection using contrastive learning and a single encoder per view. We also provide graphical criteria indicating which latent variables can be identified through a simple set of rules, which we refer to as identifiability algebra. Our general framework and theoretical results unify and extend several previous works on multi-view nonlinear ICA, disentanglement, and causal representation learning. We experimentally validate our claims on numerical, image, and multi-modal data sets. Further, we demonstrate that the performance of prior methods is recovered in different special cases of our setup. Overall, we find that access to multiple partial views enables us to identify a more fine-grained representation, under the generally milder assumption of partial observability."
                    },
                    {
                        "title": "A Sparsity Principle for Partially Observable Causal Representation Learning",
                        "abstract": "Causal representation learning aims at identifying high-level causal variables from perceptual data. Most methods assume that all latent causal variables are captured in the high-dimensional observations. We instead consider a partially observed setting, in which each measurement only provides information about a subset of the underlying causal state. Prior work has studied this setting with multiple domains or views, each depending on a fixed subset of latents. Here, we focus on learning from unpaired observations from a dataset with an instance-dependent partial observability pattern. Our main contribution is to establish two identifiability results for this setting: one for linear mixing functions without parametric assumptions on the underlying causal model, and one for piecewise linear mixing functions with Gaussian latent causal variables. Based on these insights, we propose two methods for estimating the underlying causal variables by enforcing sparsity in the inferred representation. Experiments on different simulated datasets and established benchmarks highlight the effectiveness of our approach in recovering the ground-truth latents."
                    }
                ]
            },
            "44fd2a78-0dd5-4591-92ce-e4d70821c84b": {
                "pk": "44fd2a78-0dd5-4591-92ce-e4d70821c84b",
                "name": "Caroline Muller",
                "collaborators": [
                    "Lokahith Agasthya"
                ],
                "domain": [
                    "Fluid Dynamics",
                    "Convection",
                    "Atmospheric Science",
                    "Stability Analysis"
                ],
                "publications": [
                    {
                        "title": "Dynamics and Scaling of Internally Cooled Convection",
                        "abstract": "Our goal is to investigate fundamental properties of the system of internally cooled convection. The system consists of an upward thermal flux at the lower boundary, a mean temperature lapse-rate and a constant cooling term in the bulk with the bulk cooling in thermal equilibrium with the input heat flux. This simple model represents idealised dry convection in the atmospheric boundary layer, where the cooling mimics the radiative cooling to space notably through longwave radiation. We perform linear stability analysis of the model for different values of the mean stratification to derive the critical forcing above which the fluid is convectively unstable to small perturbations. The dynamic behaviour of the fluid system is described and the scaling of various important measured quantities such as the total vertical convective heat flux and the upward mass flux is measured. We introduce a lapse-rate dependent dimensionless Rayleigh-number $Ra_\\gamma$ that determines the behaviour of the system, finding that the convective heat-flux and mass-flux scale scale approximately as $Ra_\\gamma^{0.5}$ and $Ra_\\gamma^{0.7}$ respectively. The area-fraction of the domain that is occupied by upward and downward moving fluid and the skewness of the vertical velocity are studied to understand the asymmetry inherent in the system. We conclude with a short discussion on the relevance to atmospheric convection and the scope for further investigations of atmospheric convection using similar simplified approaches."
                    }
                ]
            },
            "4b1c6e7f-2d8e-482c-9b5c-68e0368284bd": {
                "pk": "4b1c6e7f-2d8e-482c-9b5c-68e0368284bd",
                "name": "Francesco Locatello",
                "collaborators": [
                    "Gunnar R\u00e4tsch",
                    "Francesco Montagna",
                    "Bernhard Sch\u00f6lkopf",
                    "Olivier Bachem",
                    "Michael Tschannen",
                    "Stefan Bauer",
                    "Volkan Cevher",
                    "Nicoletta Noceti",
                    "Lorenzo Rosasco",
                    "Sylvain Gelly"
                ],
                "domain": [
                    "Causal Inference",
                    "Representation Learning",
                    "Optimization",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Enforcing and Discovering Structure in Machine Learning",
                        "abstract": "The world is structured in countless ways. It may be prudent to enforce corresponding structural properties to a learning algorithm's solution, such as incorporating prior beliefs, natural constraints, or causal structures. Doing so may translate to faster, more accurate, and more flexible models, which may directly relate to real-world impact. In this dissertation, we consider two different research areas that concern structuring a learning algorithm's solution: when the structure is known and when it has to be discovered."
                    },
                    {
                        "title": "Demystifying amortized causal discovery with transformers",
                        "abstract": "Supervised learning approaches for causal discovery from observational data often achieve competitive performance despite seemingly avoiding explicit assumptions that traditional methods make for identifiability. In this work, we investigate CSIvA (Ke et al., 2023), a transformer-based model promising to train on synthetic data and transfer to real data. First, we bridge the gap with existing identifiability theory and show that constraints on the training data distribution implicitly define a prior on the test observations. Consistent with classical approaches, good performance is achieved when we have a good prior on the test data, and the underlying model is identifiable. At the same time, we find new trade-offs. Training on datasets generated from different classes of causal models, unambiguously identifiable in isolation, improves the test generalization. Performance is still guaranteed, as the ambiguous cases resulting from the mixture of identifiable causal models are unlikely to occur (which we formally prove). Overall, our study finds that amortized causal discovery still needs to obey identifiability theory, but it also differs from classical methods in how the assumptions are formulated, trading more reliance on assumptions on the noise type for fewer hypotheses on the mechanisms."
                    },
                    {
                        "title": "A Commentary on the Unsupervised Learning of Disentangled Representations",
                        "abstract": "The goal of the unsupervised learning of disentangled representations is to separate the independent explanatory factors of variation in the data without access to supervision. In this paper, we summarize the results of Locatello et al., 2019, and focus on their implications for practitioners. We discuss the theoretical result showing that the unsupervised learning of disentangled representations is fundamentally impossible without inductive biases and the practical challenges it entails. Finally, we comment on our experimental findings, highlighting the limitations of state-of-the-art approaches and directions for future research."
                    },
                    {
                        "title": "Disentangling Factors of Variation Using Few Labels",
                        "abstract": "Learning disentangled representations is considered a cornerstone problem in representation learning. Recently, Locatello et al. (2019) demonstrated that unsupervised disentanglement learning without inductive biases is theoretically impossible and that existing inductive biases and unsupervised methods do not allow to consistently learn disentangled representations. However, in many practical settings, one might have access to a limited amount of supervision, for example through manual labeling of (some) factors of variation in a few training examples. In this paper, we investigate the impact of such supervision on state-of-the-art disentanglement methods and perform a large scale study, training over 52000 models under well-defined and reproducible experimental conditions. We observe that a small number of labeled examples (0.01--0.5\\% of the data set), with potentially imprecise and incomplete labels, is sufficient to perform model selection on state-of-the-art unsupervised models. Further, we investigate the benefit of incorporating supervision into the training process. Overall, we empirically validate that with little and imprecise supervision it is possible to reliably learn disentangled representations."
                    },
                    {
                        "title": "Boosting Variational Inference: an Optimization Perspective",
                        "abstract": "Variational inference is a popular technique to approximate a possibly intractable Bayesian posterior with a more tractable one. Recently, boosting variational inference has been proposed as a new paradigm to approximate the posterior by a mixture of densities by greedily adding components to the mixture. However, as is the case with many other variational inference algorithms, its theoretical properties have not been studied. In the present work, we study the convergence properties of this approach from a modern optimization viewpoint by establishing connections to the classic Frank-Wolfe algorithm. Our analyses yields novel theoretical insights regarding the sufficient conditions for convergence, explicit rates, and algorithmic simplifications. Since a lot of focus in previous works for variational inference has been on tractability, our work is especially important as a much needed attempt to bridge the gap between probabilistic models and their corresponding theoretical properties."
                    },
                    {
                        "title": "Stochastic Frank-Wolfe for Composite Convex Minimization",
                        "abstract": "A broad class of convex optimization problems can be formulated as a semidefinite program (SDP), minimization of a convex function over the positive-semidefinite cone subject to some affine constraints. The majority of classical SDP solvers are designed for the deterministic setting where problem data is readily available. In this setting, generalized conditional gradient methods (aka Frank-Wolfe-type methods) provide scalable solutions by leveraging the so-called linear minimization oracle instead of the projection onto the semidefinite cone. Most problems in machine learning and modern engineering applications, however, contain some degree of stochasticity. In this work, we propose the first conditional-gradient-type method for solving stochastic optimization problems under affine constraints. Our method guarantees $\\mathcal{O}(k^{-1/3})$ convergence rate in expectation on the objective residual and $\\mathcal{O}(k^{-5/12})$ on the feasibility gap."
                    },
                    {
                        "title": "Greedy Algorithms for Cone Constrained Optimization with Convergence Guarantees",
                        "abstract": "Greedy optimization methods such as Matching Pursuit (MP) and Frank-Wolfe (FW) algorithms regained popularity in recent years due to their simplicity, effectiveness and theoretical guarantees. MP and FW address optimization over the linear span and the convex hull of a set of atoms, respectively. In this paper, we consider the intermediate case of optimization over the convex cone, parametrized as the conic hull of a generic atom set, leading to the first principled definitions of non-negative MP algorithms for which we give explicit convergence rates and demonstrate excellent empirical performance. In particular, we derive sublinear ($\\mathcal{O}(1/t)$) convergence on general smooth and convex objectives, and linear convergence ($\\mathcal{O}(e^{-t})$) on strongly convex objectives, in both cases for general sets of atoms. Furthermore, we establish a clear correspondence of our algorithms to known algorithms from the MP and FW literature. Our novel algorithms and analyses target general atom sets and general objective functions, and hence are directly applicable to a large variety of learning settings."
                    },
                    {
                        "title": "Shortcuts for causal discovery of nonlinear models by score matching",
                        "abstract": "The use of simulated data in the field of causal discovery is ubiquitous due to the scarcity of annotated real data. Recently, Reisach et al., 2021 highlighted the emergence of patterns in simulated linear data, which displays increasing marginal variance in the casual direction. As an ablation in their experiments, Montagna et al., 2023 found that similar patterns may emerge in nonlinear models for the variance of the score vector $\\nabla \\log p_{\\mathbf{X}}$, and introduced the ScoreSort algorithm. In this work, we formally define and characterize this score-sortability pattern of nonlinear additive noise models. We find that it defines a class of identifiable (bivariate) causal models overlapping with nonlinear additive noise models. We theoretically demonstrate the advantages of ScoreSort in terms of statistical efficiency compared to prior state-of-the-art score matching-based methods and empirically show the score-sortability of the most common synthetic benchmarks in the literature. Our findings remark (1) the lack of diversity in the data as an important limitation in the evaluation of nonlinear causal discovery approaches, (2) the importance of thoroughly testing different settings within a problem class, and (3) the importance of analyzing statistical properties in causal discovery, where research is often limited to defining identifiability conditions of the model."
                    },
                    {
                        "title": "Causal Discovery with Score Matching on Additive Models with Arbitrary Noise",
                        "abstract": "Causal discovery methods are intrinsically constrained by the set of assumptions needed to ensure structure identifiability. Moreover additional restrictions are often imposed in order to simplify the inference task: this is the case for the Gaussian noise assumption on additive non-linear models, which is common to many causal discovery approaches. In this paper we show the shortcomings of inference under this hypothesis, analyzing the risk of edge inversion under violation of Gaussianity of the noise terms. Then, we propose a novel method for inferring the topological ordering of the variables in the causal graph, from data generated according to an additive non-linear model with a generic noise distribution. This leads to NoGAM (Not only Gaussian Additive noise Models), a causal discovery algorithm with a minimal set of assumptions and state of the art performance, experimentally benchmarked on synthetic data."
                    },
                    {
                        "title": "Scalable Causal Discovery with Score Matching",
                        "abstract": "This paper demonstrates how to discover the whole causal graph from the second derivative of the log-likelihood in observational non-linear additive Gaussian noise models. Leveraging scalable machine learning approaches to approximate the score function $\\nabla \\log p(\\mathbf{X})$, we extend the work of Rolland et al. (2022) that only recovers the topological order from the score and requires an expensive pruning step removing spurious edges among those admitted by the ordering. Our analysis leads to DAS (acronym for Discovery At Scale), a practical algorithm that reduces the complexity of the pruning by a factor proportional to the graph size. In practice, DAS achieves competitive accuracy with current state-of-the-art while being over an order of magnitude faster. Overall, our approach enables principled and scalable causal discovery, significantly lowering the compute bar."
                    },
                    {
                        "title": "A Unified Optimization View on Generalized Matching Pursuit and Frank-Wolfe",
                        "abstract": "Two of the most fundamental prototypes of greedy optimization are the matching pursuit and Frank-Wolfe algorithms. In this paper, we take a unified view on both classes of methods, leading to the first explicit convergence rates of matching pursuit methods in an optimization sense, for general sets of atoms. We derive sublinear ($1/t$) convergence for both classes on general smooth objectives, and linear convergence on strongly convex objectives, as well as a clear correspondence of algorithm variants. Our presented algorithms and rates are affine invariant, and do not need any incoherence or sparsity assumptions."
                    },
                    {
                        "title": "Sample Complexity Bounds for Score-Matching: Causal Discovery and Generative Modeling",
                        "abstract": "This paper provides statistical sample complexity bounds for score-matching and its applications in causal discovery. We demonstrate that accurate estimation of the score function is achievable by training a standard deep ReLU neural network using stochastic gradient descent. We establish bounds on the error rate of recovering causal relationships using the score-matching-based causal discovery method of Rolland et al. [2022], assuming a sufficiently good estimation of the score function. Finally, we analyze the upper bound of score-matching estimation within the score-based generative modeling, which has been applied for causal discovery but is also of independent interest within the domain of generative models."
                    },
                    {
                        "title": "Two Tricks to Improve Unsupervised Segmentation Learning",
                        "abstract": "We present two practical improvement techniques for unsupervised segmentation learning. These techniques address limitations in the resolution and accuracy of predicted segmentation maps of recent state-of-the-art methods. Firstly, we leverage image post-processing techniques such as guided filtering to refine the output masks, improving accuracy while avoiding substantial computational costs. Secondly, we introduce a multi-scale consistency criterion, based on a teacher-student training scheme. This criterion matches segmentation masks predicted from regions of the input image extracted at different resolutions to each other. Experimental results on several benchmarks used in unsupervised segmentation learning demonstrate the effectiveness of our proposed techniques."
                    },
                    {
                        "title": "Weakly-Supervised Disentanglement Without Compromises",
                        "abstract": "Intelligent agents should be able to learn useful representations by observing changes in their environment. We model such observations as pairs of non-i.i.d. images sharing at least one of the underlying factors of variation. First, we theoretically show that only knowing how many factors have changed, but not which ones, is sufficient to learn disentangled representations. Second, we provide practical algorithms that learn disentangled representations from pairs of images without requiring annotation of groups, individual factors, or the number of factors that have changed. Third, we perform a large-scale empirical study and show that such pairs of observations are sufficient to reliably learn disentangled representations on several benchmark data sets. Finally, we evaluate our learned representations and find that they are simultaneously useful on a diverse suite of tasks, including generalization under covariate shifts, fairness, and abstract reasoning. Overall, our results demonstrate that weak supervision enables learning of useful disentangled representations in realistic scenarios."
                    },
                    {
                        "title": "Competitive Training of Mixtures of Independent Deep Generative Models",
                        "abstract": "A common assumption in causal modeling posits that the data is generated by a set of independent mechanisms, and algorithms should aim to recover this structure. Standard unsupervised learning, however, is often concerned with training a single model to capture the overall distribution or aspects thereof. Inspired by clustering approaches, we consider mixtures of implicit generative models that ``disentangle'' the independent generative mechanisms underlying the data. Relying on an additional set of discriminators, we propose a competitive training procedure in which the models only need to capture the portion of the data distribution from which they can produce realistic samples. As a by-product, each model is simpler and faster to train. We empirically show that our approach splits the training distribution in a sensible way and increases the quality of the generated samples."
                    },
                    {
                        "title": "On the Fairness of Disentangled Representations",
                        "abstract": "Recently there has been a significant interest in learning disentangled representations, as they promise increased interpretability, generalization to unseen scenarios and faster learning on downstream tasks. In this paper, we investigate the usefulness of different notions of disentanglement for improving the fairness of downstream prediction tasks based on representations. We consider the setting where the goal is to predict a target variable based on the learned representation of high-dimensional observations (such as images) that depend on both the target variable and an \\emph{unobserved} sensitive variable. We show that in this setting both the optimal and empirical predictions can be unfair, even if the target variable and the sensitive variable are independent. Analyzing the representations of more than \\num{12600} trained state-of-the-art disentangled models, we observe that several disentanglement scores are consistently correlated with increased fairness, suggesting that disentanglement may be a useful property to encourage fairness when sensitive variables are not observed."
                    },
                    {
                        "title": "Scalable Mechanistic Neural Networks",
                        "abstract": "We propose Scalable Mechanistic Neural Network (S-MNN), an enhanced neural network framework designed for scientific machine learning applications involving long temporal sequences. By reformulating the original Mechanistic Neural Network (MNN) (Pervez et al., 2024), we reduce the computational time and space complexities from cubic and quadratic with respect to the sequence length, respectively, to linear. This significant improvement enables efficient modeling of long-term dynamics without sacrificing accuracy or interpretability. Extensive experiments demonstrate that S-MNN matches the original MNN in precision while substantially reducing computational resources. Consequently, S-MNN can drop-in replace the original MNN in applications, providing a practical and efficient tool for integrating mechanistic bottlenecks into neural network models of complex dynamical systems."
                    },
                    {
                        "title": "Score matching through the roof: linear, nonlinear, and latent variables causal discovery",
                        "abstract": "Causal discovery from observational data holds great promise, but existing methods rely on strong assumptions about the underlying causal structure, often requiring full observability of all relevant variables. We tackle these challenges by leveraging the score function $\\nabla \\log p(X)$ of observed variables for causal discovery and propose the following contributions. First, we generalize the existing results of identifiability with the score to additive noise models with minimal requirements on the causal mechanisms. Second, we establish conditions for inferring causal relations from the score even in the presence of hidden variables; this result is two-faced: we demonstrate the score's potential as an alternative to conditional independence tests to infer the equivalence class of causal graphs with hidden variables, and we provide the necessary conditions for identifying direct causes in latent variable models. Building on these insights, we propose a flexible algorithm for causal discovery across linear, nonlinear, and latent variable models, which we empirically validate."
                    },
                    {
                        "title": "A Conditional Gradient Framework for Composite Convex Minimization with Applications to Semidefinite Programming",
                        "abstract": "We propose a conditional gradient framework for a composite convex minimization template with broad applications. Our approach combines smoothing and homotopy techniques under the CGM framework, and provably achieves the optimal $\\mathcal{O}(1/\\sqrt{k})$ convergence rate. We demonstrate that the same rate holds if the linear subproblems are solved approximately with additive or multiplicative error. In contrast with the relevant work, we are able to characterize the convergence when the non-smooth term is an indicator function. Specific applications of our framework include the non-smooth minimization, semidefinite programming, and minimization with linear inclusion constraints over a compact domain. Numerical evidence demonstrates the benefits of our framework."
                    },
                    {
                        "title": "TeST: Test-time Self-Training under Distribution Shift",
                        "abstract": "Despite their recent success, deep neural networks continue to perform poorly when they encounter distribution shifts at test time. Many recently proposed approaches try to counter this by aligning the model to the new distribution prior to inference. With no labels available this requires unsupervised objectives to adapt the model on the observed test data. In this paper, we propose Test-Time Self-Training (TeST): a technique that takes as input a model trained on some source data and a novel data distribution at test time, and learns invariant and robust representations using a student-teacher framework. We find that models adapted using TeST significantly improve over baseline test-time adaptation algorithms. TeST achieves competitive performance to modern domain adaptation algorithms, while having access to 5-10x less data at time of adaption. We thoroughly evaluate a variety of baselines on two tasks: object detection and image segmentation and find that models adapted with TeST. We find that TeST sets the new state-of-the art for test-time domain adaptation algorithms."
                    }
                ]
            }
        }
    },
    "2311.17295": {
        "paper_data": {
            "title": "Elo Uncovered: Robustness and Best Practices in Language Model Evaluation",
            "url": "http://arxiv.org/abs/2311.17295v1",
            "arxiv_id": "2311.17295",
            "authors": [
                "Meriem Boubdir",
                "Edward Kim",
                "Beyza Ermis",
                "Sara Hooker",
                "Marzieh Fadaee"
            ],
            "abstract": "In Natural Language Processing (NLP), the Elo rating system, originally designed for ranking players in dynamic games such as chess, is increasingly being used to evaluate Large Language Models (LLMs) through \"A vs B\" paired comparisons. However, while popular, the system's suitability for assessing entities with constant skill levels, such as LLMs, remains relatively unexplored. We study two fundamental axioms that evaluation methods should adhere to: reliability and transitivity. We conduct extensive evaluation of Elo behaviour, illustrating that individual Elo computations exhibit volatility and delving into the impact of varying the Elo rating system's hyperparameters. We show that these axioms are not always satisfied raising questions about the reliability of current comparative evaluations of LLMs. If the current use of Elo scores is intended to substitute the costly head-to-head comparison of LLMs, it is crucial to ensure the ranking is as robust as possible. Guided by the axioms, our findings offer concrete guidelines for enhancing the reliability of LLM evaluation methods, suggesting a need for reassessment of existing comparative approaches.",
            "introduction": "   1 Introduction  In the rapidly evolving field of Natural Language Processing (NLP), the task of accurately and reliably evaluating LLMs has become increasingly challenging\u00a0(Liang et\u00a0al., 2022; Chang et\u00a0al., 2023; Srivastava et\u00a0al., 2023; Kaddour et\u00a0al., 2023; Pozzobon et\u00a0al., 2023). Human feedback has emerged as an indispensable tool in this performance assessment process, serving as a qualitative metric that captures nuances that automated scoring mechanisms often fail to address \u00a0(Askell et\u00a0al., 2021; Bai et\u00a0al., 2022a, b; Srivastava et\u00a0al., 2023; Ding et\u00a0al., 2023; Dettmers et\u00a0al., 2023).   These human-centered evaluations, highly valuable to the overall progress of the NLP field, typically adopt an \u201cA vs B\u201d comparative setup, turning evaluations into a zero-sum game between language models. This paired feedback structure\u00a0(Zhao et\u00a0al., 2023) naturally lends itself to the Elo rating system, originally designed for ranking chess players for better matchmaking\u00a0(Elo, 1978). With Elo rating system, we can integrate subjective human feedback into a structured rating system and assess the performance of language models.   Figure 1: Impact of win probabilities and permutation sampling on Elo ratings: Comparing Model A and Model B across three different win probabilities (P\u2062r\u2062o\u2062b\u2062(A\u2062beats\u2062B)={0.6,0.55,0.51}\ud835\udc43\ud835\udc5f\ud835\udc5c\ud835\udc4f\ud835\udc34normal-beats\ud835\udc350.60.550.51Prob(A\\;\\mathrm{{beats}}\\;B)=\\{0.6,0.55,0.51\\}italic_P italic_r italic_o italic_b ( italic_A roman_beats italic_B ) = { 0.6 , 0.55 , 0.51 }) with two levels of permutation sampling (Nperms=1subscript\ud835\udc41perms1N_{\\text{{perms}}}=1italic_N start_POSTSUBSCRIPT perms end_POSTSUBSCRIPT = 1 and Nperms=100subscript\ud835\udc41perms100N_{\\text{{perms}}}=100italic_N start_POSTSUBSCRIPT perms end_POSTSUBSCRIPT = 100). The top row displays the observed win rates, the middle row illustrates Elo ratings with a single permutation, and the bottom row shows the mean and standard error of the mean (SEM) of Elo ratings across 100 permutations.   The core principles of Elo rating have proven to be resilient and adaptable due to its dynamic adjustments, relative rating focus, consistency across skill levels, and simplicity and transparency. As a result, the Elo system has found diverse applications, from predicting sports events outcomes (Binder & Findlay, 2009; Hvattum & Arntzen, 2010; Leitner et\u00a0al., 2010; Wise, 2021), and facilitating matchmaking in massively multiplayer online games like StarCraft ii and Dota \u00a0(Ebtekar & Liu, 2021; Reid, ; Liquipedia, ; ESL, ), to its recent use in the evaluation of LLMs (Askell et\u00a0al., 2021; Bai et\u00a0al., 2022a, b; Srivastava et\u00a0al., 2023; Ding et\u00a0al., 2023; Dettmers et\u00a0al., 2023; Wu et\u00a0al., 2023; Lin & Chen, 2023). However, to-date there has not been a comprehensive examination of the compatibility of Elo scores and LLMs.   Unlike dynamic competitors that evolve, LLMs have static capabilities and operate in a time-agnostic context. In this setting, not only are LLM evaluations unconstrained by a preset number of turns (unlike tournament timelines or predefined match sequences), but the ordering of matches can also significantly influence the final Elo scores and, consequently, models rankings. This oversight is especially concerning, given the direct impact of Elo system rankings on both research directions and real-world applications in NLP as well as its widespread adoption (Ye et\u00a0al., 2023; Zheng et\u00a0al., 2023; K\u00f6pf et\u00a0al., 2023; Askell et\u00a0al., 2021; Bai et\u00a0al., 2022a, b; Srivastava et\u00a0al., 2023; Ding et\u00a0al., 2023; Dettmers et\u00a0al., 2023; Wu et\u00a0al., 2023; Lin & Chen, 2023).   This study aims to close this research gap by adopting an axiomatic approach and scrutinizing both the reliability and limitations of the Elo",
            "references": [
                {
                    "title": "Which Prompts Make The Difference? Data Prioritization For Efficient Human LLM Evaluation",
                    "abstract": "Human evaluation is increasingly critical for assessing large language models, capturing linguistic nuances, and reflecting user preferences more accurately than traditional automated metrics. However, the resource-intensive nature of this type of annotation process poses significant challenges. The key question driving our work:\"is it feasible to minimize human-in-the-loop feedback by prioritizing data instances which most effectively distinguish between models?\"We evaluate several metric-based methods and find that these metrics enhance the efficiency of human evaluations by minimizing the number of required annotations, thus saving time and cost, while ensuring a robust performance evaluation. We show that our method is effective across widely used model families, reducing instances of indecisive (or\"tie\") outcomes by up to 54% compared to a random sample when focusing on the top-20 percentile of prioritized instances. This potential reduction in required human effort positions our approach as a valuable strategy in future large language model evaluations."
                },
                {
                    "title": "QLoRA: Efficient Finetuning of Quantized LLMs",
                    "abstract": "We present QLoRA, an efficient finetuning approach that reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU while preserving full 16-bit finetuning task performance. QLoRA backpropagates gradients through a frozen, 4-bit quantized pretrained language model into Low Rank Adapters~(LoRA). Our best model family, which we name Guanaco, outperforms all previous openly released models on the Vicuna benchmark, reaching 99.3% of the performance level of ChatGPT while only requiring 24 hours of finetuning on a single GPU. QLoRA introduces a number of innovations to save memory without sacrificing performance: (a) 4-bit NormalFloat (NF4), a new data type that is information theoretically optimal for normally distributed weights (b) double quantization to reduce the average memory footprint by quantizing the quantization constants, and (c) paged optimziers to manage memory spikes. We use QLoRA to finetune more than 1,000 models, providing a detailed analysis of instruction following and chatbot performance across 8 instruction datasets, multiple model types (LLaMA, T5), and model scales that would be infeasible to run with regular finetuning (e.g. 33B and 65B parameter models). Our results show that QLoRA finetuning on a small high-quality dataset leads to state-of-the-art results, even when using smaller models than the previous SoTA. We provide a detailed analysis of chatbot performance based on both human and GPT-4 evaluations showing that GPT-4 evaluations are a cheap and reasonable alternative to human evaluation. Furthermore, we find that current chatbot benchmarks are not trustworthy to accurately evaluate the performance levels of chatbots. A lemon-picked analysis demonstrates where Guanaco fails compared to ChatGPT. We release all of our models and code, including CUDA kernels for 4-bit training."
                },
                {
                    "title": "Enhancing Chat Language Models by Scaling High-quality Instructional Conversations",
                    "abstract": "Fine-tuning on instruction data has been widely validated as an effective practice for implementing chat language models like ChatGPT. Scaling the diversity and quality of such data, although straightforward, stands a great chance of leading to improved performance. This paper aims to improve the upper bound of open-source models further. We first provide a systematically designed, diverse, informative, large-scale dataset of instructional conversations, UltraChat, which does not involve human queries. Our objective is to capture the breadth of interactions that a human might have with an AI assistant and employs a comprehensive framework to generate multi-turn conversation iteratively. UltraChat contains 1.5 million high-quality multi-turn dialogues and covers a wide range of topics and instructions. Our statistical analysis of UltraChat reveals its superiority in various key metrics, including scale, average length, diversity, coherence, etc., solidifying its position as a leading open-source dataset. Building upon UltraChat, we fine-tune a LLaMA model to create a powerful conversational model, UltraLLaMA. Our evaluations indicate that UltraLLaMA consistently outperforms other open-source models, including Vicuna, the previously recognized state-of-the-art open-source model. The dataset and the model will be publicly released\\footnote{\\url{https://github.com/thunlp/UltraChat}}."
                },
                {
                    "title": "On the Challenges of Using Black-Box APIs for Toxicity Evaluation in Research",
                    "abstract": "Perception of toxicity evolves over time and often differs between geographies and cultural backgrounds. Similarly, black-box commercially available APIs for detecting toxicity, such as the Perspective API, are not static, but frequently retrained to address any unattended weaknesses and biases. We evaluate the implications of these changes on the reproducibility of findings that compare the relative merits of models and methods that aim to curb toxicity. Our findings suggest that research that relied on inherited automatic toxicity scores to compare models and techniques may have resulted in inaccurate findings. Rescoring all models from HELM, a widely respected living benchmark, for toxicity with the recent version of the API led to a different ranking of widely used foundation models. We suggest caution in applying apples-to-apples comparisons between studies and lay recommendations for a more structured approach to evaluating toxicity over time. Code and data are available at https://github.com/for-ai/black-box-api-challenges."
                },
                {
                    "title": "SODA: Million-scale Dialogue Distillation with Social Commonsense Contextualization",
                    "abstract": "Data scarcity has been a long standing issue in the field of open-domain social dialogue. To quench this thirst, we present SODA: the first publicly available, million-scale high-quality social dialogue dataset. By contextualizing social commonsense knowledge from a knowledge graph, we are able to distill an exceptionally broad spectrum of social interactions from a large language model. Human evaluation shows that conversations in SODA are more consistent, specific, and (surprisingly) natural than those in prior human-authored datasets. Using SODA, we train COSMO: a generalizable conversation model that is significantly more natural and consistent on unseen datasets than best-performing conversation models (e.g., GODEL, BlenderBot-1, Koala, Vicuna). Experiments reveal COSMO is sometimes even preferred to the original human-written gold responses. Additionally, our results shed light on the distinction between knowledge-enriched conversations and natural social chitchats. We plan to make our data, model, and code public."
                },
                {
                    "title": "On the Limitations of Elo: Real-World Games, are Transitive, not Additive",
                    "abstract": "Real-world competitive games, such as chess, go, or StarCraft II, rely on Elo models to measure the strength of their players. Since these games are not fully transitive, using Elo implicitly assumes they have a strong transitive component that can correctly be identified and extracted. In this study, we investigate the challenge of identifying the strength of the transitive component in games. First, we show that Elo models can fail to extract this transitive component, even in elementary transitive games. Then, based on this observation, we propose an extension of the Elo score: we end up with a disc ranking system that assigns each player two scores, which we refer to as skill and consistency. Finally, we propose an empirical validation on payoff matrices coming from real-world games played by bots and humans."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "A General Language Assistant as a Laboratory for Alignment",
                    "abstract": "Given the broad capabilities of large language models, it should be possible to work towards a general-purpose, text-based assistant that is aligned with human values, meaning that it is helpful, honest, and harmless. As an initial foray in this direction we study simple baseline techniques and evaluations, such as prompting. We find that the benefits from modest interventions increase with model size, generalize to a variety of alignment evaluations, and do not compromise the performance of large models. Next we investigate scaling trends for several training objectives relevant to alignment, comparing imitation learning, binary discrimination, and ranked preference modeling. We find that ranked preference modeling performs much better than imitation learning, and often scales more favorably with model size. In contrast, binary discrimination typically performs and scales very similarly to imitation learning. Finally we study a `preference model pre-training' stage of training, with the goal of improving sample efficiency when finetuning on human preferences."
                },
                {
                    "title": "Multitask Prompted Training Enables Zero-Shot Task Generalization",
                    "abstract": "Large language models have recently been shown to attain reasonable zero-shot generalization on a diverse set of tasks (Brown et al., 2020). It has been hypothesized that this is a consequence of implicit multitask learning in language models' pretraining (Radford et al., 2019). Can zero-shot generalization instead be directly induced by explicit multitask learning? To test this question at scale, we develop a system for easily mapping any natural language tasks into a human-readable prompted form. We convert a large set of supervised datasets, each with multiple prompts with diverse wording. These prompted datasets allow for benchmarking the ability of a model to perform completely held-out tasks. We fine-tune a pretrained encoder-decoder model (Raffel et al., 2020; Lester et al., 2021) on this multitask mixture covering a wide variety of tasks. The model attains strong zero-shot performance on several standard datasets, often outperforming models up to 16x its size. Further, our approach attains strong performance on a subset of tasks from the BIG-bench benchmark, outperforming models up to 6x its size. All trained models are available at https://github.com/bigscience-workshop/t-zero and all prompts are available at https://github.com/bigscience-workshop/promptsource."
                },
                {
                    "title": "Randomness In Neural Network Training: Characterizing The Impact of Tooling",
                    "abstract": "The quest for determinism in machine learning has disproportionately focused on characterizing the impact of noise introduced by algorithmic design choices. In this work, we address a less well understood and studied question: how does our choice of tooling introduce randomness to deep neural network training. We conduct large scale experiments across different types of hardware, accelerators, state of art networks, and open-source datasets, to characterize how tooling choices contribute to the level of non-determinism in a system, the impact of said non-determinism, and the cost of eliminating different sources of noise. Our findings are surprising, and suggest that the impact of non-determinism in nuanced. While top-line metrics such as top-1 accuracy are not noticeably impacted, model performance on certain parts of the data distribution is far more sensitive to the introduction of randomness. Our results suggest that deterministic tooling is critical for AI safety. However, we also find that the cost of ensuring determinism varies dramatically between neural network architectures and hardware types, e.g., with overhead up to $746\\%$, $241\\%$, and $196\\%$ on a spectrum of widely used GPU accelerator architectures, relative to non-deterministic training. The source code used in this paper is available at https://github.com/usyd-fsalab/NeuralNetworkRandomness."
                },
                {
                    "title": "Evaluating Team Skill Aggregation in Online Competitive Games",
                    "abstract": "One of the main goals of online competitive games is increasing player engagement by ensuring fair matches. These games use rating systems for creating balanced match-ups. Rating systems leverage statistical estimation to rate players' skills and use skill ratings to predict rank before matching players. Skill ratings of individual players can be aggregated to compute the skill level of a team. While research often aims to improve the accuracy of skill estimation and fairness of match-ups, less attention has been given to how the skill level of a team is calculated from the skill level of its members. In this paper, we propose two new aggregation methods and compare them with a standard approach extensively used in the research literature. We present an exhaustive analysis of the impact of these methods on the predictive performance of rating systems. We perform our experiments using three popular rating systems, Elo, Glicko, and TrueSkill, on three real-world datasets including over 100,000 battle royale and head-to-head matches. Our evaluations show the superiority of the MAX method over the other two methods in the majority of the tested cases, implying that the overall performance of a team is best determined by the performance of its most skilled member. The results of this study highlight the necessity of devising more elaborated methods for calculating a team's performance- methods covering different aspects of players' behavior such as skills, strategy, or goals."
                },
                {
                    "title": "Elo Ratings for Large Tournaments of Software Agents in Asymmetric Games",
                    "abstract": "The Elo rating system has been used world wide for individual sports and team sports, as exemplified by the European Go Federation (EGF), International Chess Federation (FIDE) , International Federation of Association Football (FIFA), and many others. To evaluate the performance of artificial intelligence agents, it is natural to evaluate them on the same Elo scale as humans, such as the rating of 5185 attributed to AlphaGo Zero [10] There are several fundamental differences between humans and AI that suggest modifications to the system, which in turn require revisiting Elo\u2019s fundamental rationale. AI is typically trained on many more games than humans play, and we have little a-priori information on newly created AI agents. Further, AI is being extended into games which are asymmetric between the players, and which could even have large complex boards with different setup in every game, such as commercial paper strategy games. We present a revised rating system, and guidelines for tournaments, to reflect these differences."
                },
                {
                    "title": "Elo-MMR: A Rating System for Massive Multiplayer Competitions",
                    "abstract": "Skill estimation mechanisms, colloquially known as rating systems, play an important role in competitive sports and games. They provide a measure of player skill, which incentivizes competitive performances and enables balanced match-ups. In this paper, we present a novel Bayesian rating system for contests with many participants. It is widely applicable to competition formats with discrete ranked matches, such as online programming competitions, obstacle courses races, and video games. The system\u2019s simplicity allows us to prove theoretical bounds on its robustness and runtime. In addition, we show that it is incentive-compatible: a player who seeks to maximize their rating will never want to underperform. Experimentally, the rating system surpasses existing systems in prediction accuracy, and computes faster than existing systems by up to an order of magnitude."
                },
                {
                    "title": "Nondeterminism and Instability in Neural Network Optimization",
                    "abstract": "Nondeterminism in neural network optimization produces uncertainty in performance, making small improvements difficult to discern from run-to-run variability. While uncertainty can be reduced by training multiple model copies, doing so is time-consuming, costly, and harms reproducibility. In this work, we establish an experimental protocol for understanding the effect of optimization nondeterminism on model diversity, allowing us to isolate the effects of a variety of sources of nondeterminism. Surprisingly, we find that all sources of nondeterminism have similar effects on measures of model diversity. To explain this intriguing fact, we identify the instability of model training, taken as an end-to-end procedure, as the key determinant. We show that even one-bit changes in initial parameters result in models converging to vastly different values. Last, we propose two approaches for reducing the effects of instability on run-to-run variability."
                },
                {
                    "title": "Accounting for Variance in Machine Learning Benchmarks",
                    "abstract": "Strong empirical evidence that one machine-learning algorithm A outperforms another one B ideally calls for multiple trials optimizing the learning pipeline over sources of variation such as data sampling, data augmentation, parameter initialization, and hyperparameters choices. This is prohibitively expensive, and corners are cut to reach conclusions. We model the whole benchmarking process, revealing that variance due to data sampling, parameter initialization and hyperparameter choice impact markedly the results. We analyze the predominant comparison methods used today in the light of this variance. We show a counter-intuitive result that adding more sources of variation to an imperfect estimator approaches better the ideal estimator at a 51 times reduction in compute cost. Building on these results, we study the error rate of detecting improvements, on five different deep-learning tasks/architectures. This study leads us to propose recommendations for performance comparisons."
                },
                {
                    "title": "Synthesizing Irreproducibility in Deep Networks",
                    "abstract": "The success and superior performance of deep networks is spreading their popularity and use to an increasing number of applications. Very recent works, however, demonstrate that modern day deep networks suffer from irreproducibility (also referred to as nondeterminism or underspecification). Two or more models that are identical in architecture, structure, training hyper-parameters, and parameters, and that are trained on exactly the same training data, yield different predictions on individual previously unseen examples. Thus, a model that performs well on controlled test data, may perform in unexpected ways when deployed in the real world, whose data is expected to be similar to the test data. We study simple synthetic models and data to understand the origins of these problems. We show that even with a single nonlinearity and for very simple data and models, irreproducibility occurs. Our study demonstrates the effects of randomness in initialization, training data shuffling window size, and activation functions on prediction irreproducibility, even under very controlled synthetic data. While, as one would expect, randomness in initialization and in shuffling the training examples exacerbates the phenomenon, we show that model complexity and the choice of nonlinearity also play significant roles in making deep models irreproducible."
                },
                {
                    "title": "Smooth activations and reproducibility in deep networks",
                    "abstract": "Deep networks are gradually penetrating almost every domain in our lives due to their amazing success. However, with substantive performance accuracy improvements comes the price of \\emph{irreproducibility}. Two identical models, trained on the exact same training dataset may exhibit large differences in predictions on individual examples even when average accuracy is similar, especially when trained on highly distributed parallel systems. The popular Rectified Linear Unit (ReLU) activation has been key to recent success of deep networks. We demonstrate, however, that ReLU is also a catalyzer to irreproducibility in deep networks. We show that not only can activations smoother than ReLU provide better accuracy, but they can also provide better accuracy-reproducibility tradeoffs. We propose a new family of activations; Smooth ReLU (\\emph{SmeLU}), designed to give such better tradeoffs, while also keeping the mathematical expression simple, and thus training speed fast and implementation cheap. SmeLU is monotonic, mimics ReLU, while providing continuous gradients, yielding better reproducibility. We generalize SmeLU to give even more flexibility and then demonstrate that SmeLU and its generalized form are special cases of a more general methodology of REctified Smooth Continuous Unit (RESCU) activations. Empirical results demonstrate the superior accuracy-reproducibility tradeoffs with smooth activations, SmeLU in particular."
                },
                {
                    "title": "We Need To Talk About Random Splits",
                    "abstract": "(CITATION) argued for using random splits rather than standard splits in NLP experiments. We argue that random splits, like standard splits, lead to overly optimistic performance estimates. We can also split data in biased or adversarial ways, e.g., training on short sentences and evaluating on long ones. Biased sampling has been used in domain adaptation to simulate real-world drift; this is known as the covariate shift assumption. In NLP, however, even worst-case splits, maximizing bias, often under-estimate the error observed on new samples of in-domain data, i.e., the data that models should minimally generalize to at test time. This invalidates the covariate shift assumption. Instead of using multiple random splits, future benchmarks should ideally include multiple, independent test sets instead; if infeasible, we argue that multiple biased splits leads to more realistic performance estimates than multiple random splits."
                },
                {
                    "title": "CommonGen: A Constrained Text Generation Challenge for Generative Commonsense Reasoning",
                    "abstract": "Recently, large-scale pre-trained language models have demonstrated impressive performance on several commonsense-reasoning benchmark datasets. However, building machines with commonsense to compose realistically plausible sentences remains challenging. In this paper, we present a constrained text generation task, CommonGen associated with a benchmark dataset, to explicitly test machines for the ability of generative commonsense reasoning. Given a set of common concepts (e.g., dog, frisbee, catch, throw); the task is to generate a coherent sentence describing an everyday scenario using these concepts (e.g., \u201ca man throws a frisbee and his dog catches it\u201d). The CommonGen task is challenging because it inherently requires 1) relational reasoning with background commonsense knowledge and 2) compositional generalization ability to work on unseen concept combinations. Our dataset, constructed through a combination of crowdsourced and existing caption corpora, consists of 77k commonsense descriptions over 35k unique concept-sets. Experiments show that there is a large gap between state-of-the-art text generation models (e.g., T5) and human performance (31.6% v.s. 63.5% in SPICE metric). Furthermore, we demonstrate that the learned generative commonsense reasoning capability can be transferred to improve downstream tasks such as CommonsenseQA (76.9% to 78.4 in dev accuracy) by generating additional context."
                },
                {
                    "title": "Beat the AI: Investigating Adversarial Human Annotation for Reading Comprehension",
                    "abstract": "Abstract Innovations in annotation methodology have been a catalyst for Reading Comprehension (RC) datasets and models. One recent trend to challenge current RC models is to involve a model in the annotation process: Humans create questions adversarially, such that the model fails to answer them correctly. In this work we investigate this annotation methodology and apply it in three different settings, collecting a total of 36,000 samples with progressively stronger models in the annotation loop. This allows us to explore questions such as the reproducibility of the adversarial effect, transfer from data collected with varying model-in-the-loop strengths, and generalization to data collected without a model. We find that training on adversarially collected samples leads to strong generalization to non-adversarially collected datasets, yet with progressive performance deterioration with increasingly stronger models-in-the-loop. Furthermore, we find that stronger models can still learn from datasets collected with substantially weaker models-in-the-loop. When trained on data collected with a BiDAF model in the loop, RoBERTa achieves 39.9F1 on questions that it cannot answer when trained on SQuAD\u2014only marginally lower than when trained on data collected using RoBERTa itself (41.0F1)."
                },
                {
                    "title": "On Model Stability as a Function of Random Seed",
                    "abstract": "In this paper, we focus on quantifying model stability as a function of random seed by investigating the effects of the induced randomness on model performance and the robustness of the model in general. We specifically perform a controlled study on the effect of random seeds on the behaviour of attention, gradient-based and surrogate model based (LIME) interpretations. Our analysis suggests that random seeds can adversely affect the consistency of models resulting in counterfactual interpretations. We propose a technique called Aggressive Stochastic Weight Averaging (ASWA) and an extension called Norm-filtered Aggressive Stochastic Weight Averaging (NASWA) which improves the stability of models over random seeds. With our ASWA and NASWA based optimization, we are able to improve the robustness of the original model, on average reducing the standard deviation of the model\u2019s performance by 72%."
                },
                {
                    "title": "Measuring the Effects of Data Parallelism on Neural Network Training",
                    "abstract": "Recent hardware developments have dramatically increased the scale of data parallelism available for neural network training. Among the simplest ways to harness next-generation hardware is to increase the batch size in standard mini-batch neural network training algorithms. In this work, we aim to experimentally characterize the effects of increasing the batch size on training time, as measured by the number of steps necessary to reach a goal out-of-sample error. We study how this relationship varies with the training algorithm, model, and data set, and find extremely large variation between workloads. Along the way, we show that disagreements in the literature on how batch size affects model quality can largely be explained by differences in metaparameter tuning and compute budgets at different batch sizes. We find no evidence that larger batch sizes degrade out-of-sample performance. Finally, we discuss the implications of our results on efforts to train neural networks much faster in the future. Our experimental data is publicly available as a database of 71,638,836 loss measurements taken over the course of training for 168,160 individual models across 35 workloads."
                },
                {
                    "title": "The Impact of Nondeterminism on Reproducibility in Deep Reinforcement Learning",
                    "abstract": "While deep reinforcement learning (DRL) has enjoyed several recent successes, results reported in the literature are often difficult to reliably reproduce. Difficulties in reproducibility can arise due to many factors, including the lack of access to computational resources or the lack of knowledge of specific implementation details. Another factor of particular importance in the specific case of DRL is the ability to control for sources of nondeterminism during the training process. This is because DRL is faced with the challenges of a nonstationary training distribution and additional sources of randomness that are absent from other areas of machine learning. In this paper, we (1) enable deterministic training in DRL by identifying and controlling for all sources of nondeterminism present during training, and (2) perform an ablation study that shows how these sources of nondeterminism can impact the performance of a DRL agent. We find that even simple sources of nondeterminism such as those stemming from nondeterministic GPU operations can lead to large differences in performance between training runs. Lastly, we make available our deterministic implementation of deep Q-learning [14]."
                },
                {
                    "title": "State of the Art: Reproducibility in Artificial Intelligence",
                    "abstract": "\n \n Background: Research results in artificial intelligence (AI) are criticized for not being reproducible. Objective: To quantify the state of reproducibility of empirical AI research using six reproducibility metrics measuring three different degrees of reproducibility. Hypotheses: 1) AI research is not documented well enough to reproduce the reported results. 2) Documentation practices have improved over time. Method: The literature is reviewed and a set of variables that should be documented to enable reproducibility are grouped into three factors: Experiment, Data and Method. The metrics describe how well the factors have been documented for a paper. A total of 400 research papers from the conference series IJCAI and AAAI have been surveyed using the metrics. Findings: None of the papers document all of the variables. The metrics show that between 20% and 30% of the variables for each factor are documented. One of the metrics show statistically significant increase over time while the others show no change. Interpretation: The reproducibility scores decrease with in- creased documentation requirements. Improvement over time is found. Conclusion: Both hypotheses are supported.\n \n"
                },
                {
                    "title": "TrueSkill 2: An improved Bayesian skill rating system",
                    "abstract": "Online multiplayer games, such as Gears of War and Halo, use skill-based matchmaking to give players fair and enjoyable matches. They depend on a skill rating system to infer accurate player skills from historical data. TrueSkill is a popular and effective skill rating system, working from only the winner and loser of each game. This paper presents an extension to TrueSkill that incorporates additional information that is readily available in online shooters, such as player experience, membership in a squad, the number of kills a player scored, tendency to quit, and skill in other game modes. This extension, which we call TrueSkill2, is shown to significantly improve the accuracy of skill ratings computed from Halo 5 matches. TrueSkill2 predicts historical match outcomes with 68% accuracy, compared to 52% accuracy for TrueSkill."
                },
                {
                    "title": "Sensitivity and Generalization in Neural Networks: an Empirical Study",
                    "abstract": "In practice it is often found that large over-parameterized neural networks generalize better than their smaller counterparts, an observation that appears to conflict with classical notions of function complexity, which typically favor smaller models. In this work, we investigate this tension between complexity and generalization through an extensive empirical exploration of two natural metrics of complexity related to sensitivity to input perturbations. Our experiments survey thousands of models with various fully-connected architectures, optimizers, and other hyper-parameters, as well as four different image classification datasets. \nWe find that trained neural networks are more robust to input perturbations in the vicinity of the training data manifold, as measured by the norm of the input-output Jacobian of the network, and that it correlates well with generalization. We further establish that factors associated with poor generalization $-$ such as full-batch training or using random labels $-$ correspond to lower robustness, while factors associated with good generalization $-$ such as data augmentation and ReLU non-linearities $-$ give rise to more robust functions. Finally, we demonstrate how the input-output Jacobian norm can be predictive of generalization at the level of individual test points."
                },
                {
                    "title": "Are GANs Created Equal? A Large-Scale Study",
                    "abstract": "Generative adversarial networks (GAN) are a powerful subclass of generative models. Despite a very rich research activity leading to numerous interesting GAN algorithms, it is still very hard to assess which algorithm(s) perform better than others. We conduct a neutral, multi-faceted large-scale empirical study on state-of-the art models and evaluation measures. We find that most models can reach similar scores with enough hyperparameter optimization and random restarts. This suggests that improvements can arise from a higher computational budget and tuning more than fundamental algorithmic changes. To overcome some limitations of the current metrics, we also propose several data sets on which precision and recall can be computed. Our experimental results suggest that future GAN research should be based on more systematic and objective evaluation procedures. Finally, we did not find evidence that any of the tested algorithms consistently outperforms the non-saturating GAN introduced in \\cite{goodfellow2014generative}."
                },
                {
                    "title": "Deep Reinforcement Learning that Matters",
                    "abstract": "\n \n In recent years, significant progress has been made in solving challenging problems across various domains using deep reinforcement learning (RL). Reproducing existing work and accurately judging the improvements offered by novel methods is vital to sustaining this progress. Unfortunately, reproducing results for state-of-the-art deep RL methods is seldom straightforward. In particular, non-determinism in standard benchmark environments, combined with variance intrinsic to the methods, can make reported results tough to interpret. Without significance metrics and tighter standardization of experimental reporting, it is difficult to determine whether improvements over the prior state-of-the-art are meaningful. In this paper, we investigate challenges posed by reproducibility, proper experimental techniques, and reporting procedures. We illustrate the variability in reported metrics and results when comparing against common baselines and suggest guidelines to make future results in deep RL more reproducible. We aim to spur discussion about how to ensure continued progress in the field by minimizing wasted effort stemming from results that are non-reproducible and easily misinterpreted.\n \n"
                },
                {
                    "title": "On the State of the Art of Evaluation in Neural Language Models",
                    "abstract": "Ongoing innovations in recurrent neural network architectures have provided a steady influx of apparently state-of-the-art results on language modelling benchmarks. However, these have been evaluated using differing code bases and limited computational resources, which represent uncontrolled sources of experimental variation. We reevaluate several popular architectures and regularisation methods with large-scale automatic black-box hyperparameter tuning and arrive at the somewhat surprising conclusion that standard LSTM architectures, when properly regularised, outperform more recent models. We establish a new state of the art on the Penn Treebank and Wikitext-2 corpora, as well as strong baselines on the Hutter Prize dataset."
                },
                {
                    "title": "Knowledge Base Completion: Baselines Strike Back",
                    "abstract": "Many papers have been published on the knowledge base completion task in the past few years. Most of these introduce novel architectures for relation learning that are evaluated on standard datasets like FB15k and WN18. This paper shows that the accuracy of almost all models published on the FB15k can be outperformed by an appropriately tuned baseline \u2014 our reimplementation of the DistMult model. Our findings cast doubt on the claim that the performance improvements of recent models are due to architectural changes as opposed to hyper-parameter tuning or different training objectives. This should prompt future research to re-consider how the performance of models is evaluated and reported."
                },
                {
                    "title": "What does research reproducibility mean?",
                    "abstract": "The language and conceptual framework of \u201cresearch reproducibility\u201d are nonstandard and unsettled across the sciences. The language and conceptual framework of \u201cresearch reproducibility\u201d are nonstandard and unsettled across the sciences. In this Perspective, we review an array of explicit and implicit definitions of reproducibility and related terminology, and discuss how to avoid potential misunderstandings when these terms are used as a surrogate for \u201ctruth.\u201d"
                },
                {
                    "title": "Replicability is not Reproducibility:Nor is it Good Science",
                    "abstract": "At various machine learning conferences, at various times, there have been discussions arising from the inability to replicate the experimental results published in a paper. There seems to be a wide spread view that we need to do something to address this problem, as it is essential to the advancement of our field. The most compelling argument would seem to be that reproducibility of experimental results is the hallmark of science. Therefore, given that most of us regard machine learning as a scientific discipline, being able to replicate experiments is paramount. I want to challenge this view by separating the notion of reproducibility, a generally desirable property, from replicability, its poor cousin. I claim there are important differences between the two. Reproducibility requires changes; replicability avoids them. Although reproducibility is desirable, I contend that the impoverished version, replicability, is one not worth having."
                },
                {
                    "title": "The Effects of the Bosman Ruling on National and Club Teams in Europe",
                    "abstract": "The Bosman ruling and its aftermath allowed soccer players to move more freely between clubs in Europe. This study examines the performance of national and club teams in Europe before and after Bosman. Some national teams improved after the ruling while others became weaker, but the overall effects are small. At the club level, there is little evidence that the competitive balance of the domestic leagues in Europe was seriously harmed, although in the Champions League the top clubs appear to have become noticeably stronger."
                },
                {
                    "title": "Pitfalls of Static Language Modelling",
                    "abstract": "Our world is open-ended, non-stationary and constantly evolving; thus what we talk about and how we talk about it changes over time. This inherent dynamic nature of language comes in stark contrast to the current static language modelling paradigm, which constructs training and evaluation sets from overlapping time periods. Despite recent progress, we demonstrate that state-of-the-art Transformer models perform worse in the realistic setup of predicting future utterances from beyond their training period\u2014a consistent pattern across three datasets from two domains. We \ufb01nd that, while increasing model size alone\u2014a key driver behind recent progress\u2014does not provide a solution for the temporal generalization problem, having models that continually update their knowledge with new information can indeed slow down the degradation over time. Hence, given the compilation of ever-larger language modelling training datasets, combined with the growing list of language-model-based NLP applications that require up-to-date knowledge about the world, we argue that now is the right time to rethink our static language modelling evaluation protocol, and develop adaptive language models that can remain up-to-date with respect to our ever-changing and non-stationary world. 1"
                },
                {
                    "title": "CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledge",
                    "abstract": "When answering a question, people often draw upon their rich world knowledge in addition to the particular context. Recent work has focused primarily on answering questions given some relevant document or context, and required very little general background. To investigate question answering with prior knowledge, we present CommonsenseQA: a challenging new dataset for commonsense question answering. To capture common sense beyond associations, we extract from ConceptNet (Speer et al., 2017) multiple target concepts that have the same semantic relation to a single source concept. Crowd-workers are asked to author multiple-choice questions that mention the source concept and discriminate in turn between each of the target concepts. This encourages workers to create questions with complex semantics that often require prior knowledge. We create 12,247 questions through this procedure and demonstrate the difficulty of our task with a large number of strong baselines. Our best baseline is based on BERT-large (Devlin et al., 2018) and obtains 56% accuracy, well below human performance, which is 89%."
                },
                {
                    "title": "TrueSkill\u2122: A Bayesian Skill Rating System",
                    "abstract": "We present a new Bayesian skill rating system which can be viewed as a generalisation of the Elo system used in Chess. The new system tracks the uncertainty about player skills, explicitly models draws, can deal with any number of competing entities and can infer individual skills from team results. Inference is performed by approximate message passing on a factor graph representation of the model. We present experimental evidence on the increased accuracy and convergence speed of the system compared to Elo and report on our experience with the new rating system running in a large-scale commercial online gaming service under the name of TrueSkill."
                },
                {
                    "title": "Parameter Estimation in Large Dynamic Paired Comparison Experiments",
                    "abstract": "Paired comparison data in which the abilities or merits of the objects being compared may be changing over time can be modelled as a non\u2010linear state space model. When the population of objects being compared is large, likelihood\u2010based analyses can be too computationally cumbersome to carry out regularly. This presents a problem for rating populations of chess players and other large groups which often consist of tens of thousands of competitors. This problem is overcome through a computationally simple non\u2010iterative algorithm for fitting a particular dynamic paired comparison model. The algorithm, which improves over the commonly used algorithm of Elo by incorporating the variability in parameter estimates, can be performed regularly even for large populations of competitors. The method is evaluated on simulated data and is applied to ranking the best chess players of all time, and to ranking the top current tennis\u2010players."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can the Elo rating system be effectively adapted and applied to evaluate the performance of large language models (LLMs) in Natural Language Processing (NLP)?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the need for a reliable and structured method to evaluate LLMs, which are increasingly used in various applications. A comprehensive understanding of the Elo rating system's compatibility with LLMs could lead to more accurate performance assessments, guiding future research directions and improving model development. This advancement could enhance the quality of NLP applications, ensuring that they are built on robust evaluations that reflect true model capabilities.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the static nature of LLMs compared to dynamic competitors, which complicates the evaluation process. The ordering of matches in the Elo system can significantly influence final scores, making it difficult to establish consistent rankings. Naive approaches may fail because they do not account for the unique characteristics of LLMs, such as their time-agnostic context and the absence of a preset number of evaluation turns. Additionally, the integration of subjective human feedback into a structured rating system presents technical and theoretical obstacles that need to be addressed.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has not comprehensively examined the compatibility of Elo scores with LLMs, primarily due to a lack of focus on the unique evaluation dynamics of static models. Existing solutions have often overlooked the impact of match ordering on rankings and the implications of using a comparative feedback structure. Barriers such as the complexity of integrating human feedback and the need for a tailored approach to LLM evaluation have prevented this problem from being solved. This study aims to fill these gaps by adopting an axiomatic approach that scrutinizes the reliability and limitations of the Elo system in this context.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves an axiomatic examination of the Elo rating system as applied to LLM evaluations. This will include a detailed analysis of the system's reliability and limitations, utilizing a dataset of human feedback on LLM performance. The evaluation metrics will focus on the consistency and accuracy of Elo ratings in reflecting model capabilities. Expected outcomes include a clearer understanding of how Elo ratings can be adapted for LLMs, insights into the influence of match ordering"
            }
        },
        "author_data": {
            "6cda11c8-97e2-473d-a445-a602dc626b93": {
                "pk": "6cda11c8-97e2-473d-a445-a602dc626b93",
                "name": "Meriem Boubdir",
                "collaborators": [
                    "Edward Kim",
                    "Beyza Ermis",
                    "Marzieh Fadaee",
                    "Sara Hooker"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Human Evaluation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Which Prompts Make The Difference? Data Prioritization For Efficient Human LLM Evaluation",
                        "abstract": "Human evaluation is increasingly critical for assessing large language models, capturing linguistic nuances, and reflecting user preferences more accurately than traditional automated metrics. However, the resource-intensive nature of this type of annotation process poses significant challenges. The key question driving our work: \"is it feasible to minimize human-in-the-loop feedback by prioritizing data instances which most effectively distinguish between models?\" We evaluate several metric-based methods and find that these metrics enhance the efficiency of human evaluations by minimizing the number of required annotations, thus saving time and cost, while ensuring a robust performance evaluation. We show that our method is effective across widely used model families, reducing instances of indecisive (or \"tie\") outcomes by up to 54% compared to a random sample when focusing on the top-20 percentile of prioritized instances. This potential reduction in required human effort positions our approach as a valuable strategy in future large language model evaluations."
                    }
                ]
            },
            "22fc63f6-d782-4128-b3fd-0b5b81cd3f56": {
                "pk": "22fc63f6-d782-4128-b3fd-0b5b81cd3f56",
                "name": "Edward Kim",
                "collaborators": [
                    "Yigit Alparslan",
                    "Andrew O'Brien",
                    "Maryam Daniali",
                    "Michael Shenoda",
                    "Tianyang Nie",
                    "Marek Rutkowski"
                ],
                "domain": [
                    "Machine Learning",
                    "Adversarial Robustness",
                    "Computer Vision",
                    "Financial Modeling"
                ],
                "publications": [
                    {
                        "title": "Nevermind: Instruction Override and Moderation in Large Language Models",
                        "abstract": "Given the impressive capabilities of recent Large Language Models (LLMs), we investigate and benchmark the most popular proprietary and different sized open source models on the task of explicit instruction following in conflicting situations, e.g. overrides. These include the ability of the model to override the knowledge within the weights of the model, the ability to override (or moderate) extracted knowledge in the prompt, and lastly the ability to perform a full jailbreak. Experimentation performed suggest several key findings to improve instruction following - larger models perform the best in following instructions that override internal and contextual instructions, and are obedient, even to a fault. When scaling to longer contexts via rope scaling, a significant buffer needs to be maintained from the edge of the perplexity cliff in order to maintain instruction following capabilities. Finally, we observe improving instruction following, and subsequently instruction overrides/jailbreaks, is fundamentally at odds with the ability of a language model to follow given safety filters or guidelines. Thus, we postulate the most effective approach for safe, trustworthy AI should be dealt external to the LLM itself."
                    },
                    {
                        "title": "ATRAS: Adversarially Trained Robust Architecture Search",
                        "abstract": "In this paper, we explore the effect of architecture completeness on adversarial robustness. We train models with different architectures on CIFAR-10 and MNIST dataset. For each model, we vary different number of layers and different number of nodes in the layer. For every architecture candidate, we use Fast Gradient Sign Method (FGSM) to generate untargeted adversarial attacks and use adversarial training to defend against those attacks. For each architecture candidate, we report pre-attack, post-attack and post-defense accuracy for the model as well as the architecture parameters and the impact of completeness to the model accuracies."
                    },
                    {
                        "title": "Robust SleepNets",
                        "abstract": "State-of-the-art convolutional neural networks excel in machine learning tasks such as face recognition, and object classification but suffer significantly when adversarial attacks are present. It is crucial that machine critical systems, where machine learning models are deployed, utilize robust models to handle a wide range of variability in the real world and malicious actors that may use adversarial attacks. In this study, we investigate eye closedness detection to prevent vehicle accidents related to driver disengagements and driver drowsiness. Specifically, we focus on adversarial attacks in this application domain, but emphasize that the methodology can be applied to many other domains. We develop two models to detect eye closedness: first model on eye images and a second model on face images. We adversarially attack the models with Projected Gradient Descent, Fast Gradient Sign and DeepFool methods and report adversarial success rate. We also study the effect of training data augmentation. Finally, we adversarially train the same models on perturbed images and report the success rate for the defense against these attacks. We hope our study sets up the work to prevent potential vehicle accidents by capturing drivers' face images and alerting them in case driver's eyes are closed due to drowsiness."
                    },
                    {
                        "title": "Multi-Agent Algorithmic Recourse",
                        "abstract": "The recent adoption of machine learning as a tool in real world decision making has spurred interest in understanding how these decisions are being made. Counterfactual Explanations are a popular interpretable machine learning technique that aims to understand how a machine learning model would behave if given alternative inputs. Many explanations attempt to go further and recommend actions an individual could take to obtain a more desirable output from the model. These recommendations are known as algorithmic recourse. Past work has largely focused on the effect algorithmic recourse has on a single agent. In this work, we show that when the assumption of a single agent environment is relaxed, current approaches to algorithmic recourse fail to guarantee certain ethically desirable properties. Instead, we propose a new game theory inspired framework for providing algorithmic recourse in a multi-agent environment that does guarantee these properties."
                    },
                    {
                        "title": "Perception Over Time: Temporal Dynamics for Robust Image Understanding",
                        "abstract": "While deep learning surpasses human-level performance in narrow and specific vision tasks, it is fragile and over-confident in classification. For example, minor transformations in perspective, illumination, or object deformation in the image space can result in drastically different labeling, which is especially transparent via adversarial perturbations. On the other hand, human visual perception is orders of magnitude more robust to changes in the input stimulus. But unfortunately, we are far from fully understanding and integrating the underlying mechanisms that result in such robust perception. In this work, we introduce a novel method of incorporating temporal dynamics into static image understanding. We describe a neuro-inspired method that decomposes a single image into a series of coarse-to-fine images that simulates how biological vision integrates information over time. Next, we demonstrate how our novel visual perception framework can utilize this information \"over time\" using a biologically plausible algorithm with recurrent units, and as a result, significantly improving its accuracy and robustness over standard CNNs. We also compare our proposed approach with state-of-the-art models and explicitly quantify our adversarial robustness properties through multiple ablation studies. Our quantitative and qualitative results convincingly demonstrate exciting and transformative improvements over the standard computer vision and deep learning architectures used today."
                    },
                    {
                        "title": "DiffuGen: Adaptable Approach for Generating Labeled Image Datasets using Stable Diffusion Models",
                        "abstract": "Generating high-quality labeled image datasets is crucial for training accurate and robust machine learning models in the field of computer vision. However, the process of manually labeling real images is often time-consuming and costly. To address these challenges associated with dataset generation, we introduce \"DiffuGen,\" a simple and adaptable approach that harnesses the power of stable diffusion models to create labeled image datasets efficiently. By leveraging stable diffusion models, our approach not only ensures the quality of generated datasets but also provides a versatile solution for label generation. In this paper, we present the methodology behind DiffuGen, which combines the capabilities of diffusion models with two distinct labeling techniques: unsupervised and supervised. Distinctively, DiffuGen employs prompt templating for adaptable image generation and textual inversion to enhance diffusion model capabilities."
                    },
                    {
                        "title": "Extreme Volatility Prediction in Stock Market: When GameStop meets Long Short-Term Memory Networks",
                        "abstract": "The beginning of 2021 saw a surge in volatility for certain stocks such as GameStop company stock (Ticker GME under NYSE). GameStop stock increased around 10 fold from its decade-long average to its peak at \\$485. In this paper, we hypothesize a buy-and-hold strategy can be outperformed in the presence of extreme volatility by predicting and trading consolidation breakouts. We investigate GME stock for its volatility and compare it to SPY as a benchmark (since it is a less volatile ETF fund) from February 2002 to February 2021. For strategy 1, we develop a Long Short-term Memory (LSTM) Neural Network to predict stock prices recurrently with a very short look ahead period in the presence of extreme volatility. For our strategy 2, we develop an LSTM autoencoder network specifically designed to trade only on consolidation breakouts after predicting anomalies in the stock price. When back-tested in our simulations, our strategy 1 executes 863 trades for SPY and 452 trades for GME. Our strategy 2 executes 931 trades for SPY and 325 trades for GME. We compare both strategies to buying and holding one single share for the period that we picked as a benchmark. In our simulations, SPY returns \\$281.160 from buying and holding one single share, \\$110.29 from strategy 1 with 53.5% success rate and \\$4.34 from strategy 2 with 57.6% success rate. GME returns \\$45.63 from buying and holding one single share, \\$69.046 from strategy 1 with 47.12% success rate and \\$2.10 from strategy 2 with 48% success rate. Overall, buying and holding outperforms all deep-learning assisted prediction models in our study except for when the LSTM-based prediction model (strategy 1) is applied to GME. We hope that our study sheds more light into the field of extreme volatility predictions based on LSTMs to outperform buying and holding strategy."
                    },
                    {
                        "title": "Arbitrage-Free Pricing of Game Options in Nonlinear Markets",
                        "abstract": "The goal is to re-examine and extend the findings from the recent paper by Dumitrescu, Quenez and Sulem (2017) who studied game options within the nonlinear arbitrage-free pricing approach developed in El Karoui and Quenez (1997). We consider the setup introduced in Kim, Nie and Rutkowski (2018) where contracts of an American style were examined. We give a detailed study of unilateral pricing, hedging and exercising problems for the counterparties within a general nonlinear setup. We also present a BSDE approach, which is used to obtain more explicit results under suitable assumptions about solutions to doubly reflected BSDEs."
                    }
                ]
            },
            "50367d6f-6a03-44e6-8765-392aa726e221": {
                "pk": "50367d6f-6a03-44e6-8765-392aa726e221",
                "name": "Beyza Ermis",
                "collaborators": [
                    "Sara Hooker",
                    "Giovanni Zappella",
                    "Cedric Archambeau",
                    "Luiza Pozzobon",
                    "Patrick Lewis",
                    "Martin Wistuba",
                    "A. Taylan Cemgil",
                    "Ali Taylan Cemgil",
                    "Aditya Rawal",
                    "Marzieh Fadaee"
                ],
                "domain": [
                    "Machine Learning",
                    "Continual Learning",
                    "Privacy-Preserving",
                    "Multilingual Models"
                ],
                "publications": [
                    {
                        "title": "A Bayesian Tensor Factorization Model via Variational Inference for Link Prediction",
                        "abstract": "Probabilistic approaches for tensor factorization aim to extract meaningful structure from incomplete data by postulating low rank constraints. Recently, variational Bayesian (VB) inference techniques have successfully been applied to large scale models. This paper presents full Bayesian inference via VB on both single and coupled tensor factorization models. Our method can be run even for very large models and is easily implemented. It exhibits better prediction performance than existing approaches based on maximum likelihood on several real-world datasets for missing link prediction problem."
                    },
                    {
                        "title": "Incremental Variational Inference for Latent Dirichlet Allocation",
                        "abstract": "We introduce incremental variational inference and apply it to latent Dirichlet allocation (LDA). Incremental variational inference is inspired by incremental EM and provides an alternative to stochastic variational inference. Incremental LDA can process massive document collections, does not require to set a learning rate, converges faster to a local optimum of the variational bound and enjoys the attractive property of monotonically increasing it. We study the performance of incremental LDA on large benchmark data sets. We further introduce a stochastic approximation of incremental variational inference which extends to the asynchronous distributed setting. The resulting distributed algorithm achieves comparable performance as single host incremental variational inference, but with a significant speed-up."
                    },
                    {
                        "title": "Differentially Private Dropout",
                        "abstract": "Large data collections required for the training of neural networks often contain sensitive information such as the medical histories of patients, and the privacy of the training data must be preserved. In this paper, we introduce a dropout technique that provides an elegant Bayesian interpretation to dropout, and show that the intrinsic noise added, with the primary goal of regularization, can be exploited to obtain a degree of differential privacy. The iterative nature of training neural networks presents a challenge for privacy-preserving estimation since multiple iterations increase the amount of noise added. We overcome this by using a relaxed notion of differential privacy, called concentrated differential privacy, which provides tighter estimates on the overall privacy loss. We demonstrate the accuracy of our privacy-preserving dropout algorithm on benchmark datasets."
                    },
                    {
                        "title": "Differentially Private Variational Dropout",
                        "abstract": "Deep neural networks with their large number of parameters are highly flexible learning systems. The high flexibility in such networks brings with some serious problems such as overfitting, and regularization is used to address this problem. A currently popular and effective regularization technique for controlling the overfitting is dropout. Often, large data collections required for neural networks contain sensitive information such as the medical histories of patients, and the privacy of the training data should be protected. In this paper, we modify the recently proposed variational dropout technique which provided an elegant Bayesian interpretation to dropout, and show that the intrinsic noise in the variational dropout can be exploited to obtain a degree of differential privacy. The iterative nature of training neural networks presents a challenge for privacy-preserving estimation since multiple iterations increase the amount of noise added. We overcome this by using a relaxed notion of differential privacy, called concentrated differential privacy, which provides tighter estimates on the overall privacy loss. We demonstrate the accuracy of our privacy-preserving variational dropout algorithm on benchmark datasets."
                    },
                    {
                        "title": "Learning to Rank in the Position Based Model with Bandit Feedback",
                        "abstract": "Personalization is a crucial aspect of many online experiences. In particular, content ranking is often a key component in delivering sophisticated personalization results. Commonly, supervised learning-to-rank methods are applied, which suffer from bias introduced during data collection by production systems in charge of producing the ranking. To compensate for this problem, we leverage contextual multi-armed bandits. We propose novel extensions of two well-known algorithms viz. LinUCB and Linear Thompson Sampling to the ranking use-case. To account for the biases in a production environment, we employ the position-based click model. Finally, we show the validity of the proposed algorithms by conducting extensive offline experiments on synthetic datasets as well as customer facing online A/B experiments."
                    },
                    {
                        "title": "Variational Inference For Probabilistic Latent Tensor Factorization with KL Divergence",
                        "abstract": "Probabilistic Latent Tensor Factorization (PLTF) is a recently proposed probabilistic framework for modelling multi-way data. Not only the common tensor factorization models but also any arbitrary tensor factorization structure can be realized by the PLTF framework. This paper presents full Bayesian inference via variational Bayes that facilitates more powerful modelling and allows more sophisticated inference on the PLTF framework. We illustrate our approach on model order selection and link prediction."
                    },
                    {
                        "title": "On the Challenges of Using Black-Box APIs for Toxicity Evaluation in Research",
                        "abstract": "Perception of toxicity evolves over time and often differs between geographies and cultural backgrounds. Similarly, black-box commercially available APIs for detecting toxicity, such as the Perspective API, are not static, but frequently retrained to address any unattended weaknesses and biases. We evaluate the implications of these changes on the reproducibility of findings that compare the relative merits of models and methods that aim to curb toxicity. Our findings suggest that research that relied on inherited automatic toxicity scores to compare models and techniques may have resulted in inaccurate findings. Rescoring all models from HELM, a widely respected living benchmark, for toxicity with the recent version of the API led to a different ranking of widely used foundation models. We suggest caution in applying apples-to-apples comparisons between studies and lay recommendations for a more structured approach to evaluating toxicity over time. Code and data are available at https://github.com/for-ai/black-box-api-challenges."
                    },
                    {
                        "title": "Goodtriever: Adaptive Toxicity Mitigation with Retrieval-augmented Models",
                        "abstract": "Considerable effort has been dedicated to mitigating toxicity, but existing methods often require drastic modifications to model parameters or the use of computationally intensive auxiliary models. Furthermore, previous approaches have often neglected the crucial factor of language's evolving nature over time. In this work, we present a comprehensive perspective on toxicity mitigation that takes into account its changing nature. We introduce Goodtriever, a flexible methodology that matches the current state-of-the-art toxicity mitigation while achieving 43% relative latency reduction during inference and being more computationally efficient. By incorporating a retrieval-based approach at decoding time, Goodtriever enables toxicity-controlled text generation. Our research advocates for an increased focus on adaptable mitigation techniques, which better reflect the data drift models face when deployed in the wild. Code and data are available at https://github.com/for-ai/goodtriever."
                    },
                    {
                        "title": "Multilingual Arbitrage: Optimizing Data Pools to Accelerate Multilingual Progress",
                        "abstract": "The use of synthetic data has played a critical role in recent state-of-art breakthroughs. However, overly relying on a single oracle teacher model to generate data has been shown to lead to model collapse and invite propagation of biases. These limitations are particularly evident in multilingual settings, where the absence of a universally effective teacher model that excels across all languages presents significant challenges. In this work, we address these extreme difference by introducing \"multilingual arbitrage\", which capitalizes on performance variations between multiple models for a given language. To do so, we strategically route samples through a diverse pool of models, each with unique strengths in different languages. Across exhaustive experiments on state-of-art models, our work suggests that arbitrage techniques allow for spectacular gains in performance that far outperform relying on a single teacher. In particular, compared to the best single teacher, we observe gains of up to 56.5% improvement in win rates averaged across all languages when switching to multilingual arbitrage. We observe the most significant gains for the least resourced languages in our pool."
                    },
                    {
                        "title": "Continual Learning with Transformers for Image Classification",
                        "abstract": "In many real-world scenarios, data to train machine learning models become available over time. However, neural network models struggle to continually learn new concepts without forgetting what has been learnt in the past. This phenomenon is known as catastrophic forgetting and it is often difficult to prevent due to practical constraints, such as the amount of data that can be stored or the limited computation sources that can be used. Moreover, training large neural networks, such as Transformers, from scratch is very costly and requires a vast amount of training data, which might not be available in the application domain of interest. A recent trend indicates that dynamic architectures based on an expansion of the parameters can reduce catastrophic forgetting efficiently in continual learning, but this needs complex tuning to balance the growing number of parameters and barely share any information across tasks. As a result, they struggle to scale to a large number of tasks without significant overhead. In this paper, we validate in the computer vision domain a recent solution called Adaptive Distillation of Adapters (ADA), which is developed to perform continual learning using pre-trained Transformers and Adapters on text classification tasks. We empirically demonstrate on different classification tasks that this method maintains a good predictive performance without retraining the model or increasing the number of model parameters over the time. Besides it is significantly faster at inference time compared to the state-of-the-art methods."
                    },
                    {
                        "title": "Which Prompts Make The Difference? Data Prioritization For Efficient Human LLM Evaluation",
                        "abstract": "Human evaluation is increasingly critical for assessing large language models, capturing linguistic nuances, and reflecting user preferences more accurately than traditional automated metrics. However, the resource-intensive nature of this type of annotation process poses significant challenges. The key question driving our work: \"is it feasible to minimize human-in-the-loop feedback by prioritizing data instances which most effectively distinguish between models?\" We evaluate several metric-based methods and find that these metrics enhance the efficiency of human evaluations by minimizing the number of required annotations, thus saving time and cost, while ensuring a robust performance evaluation. We show that our method is effective across widely used model families, reducing instances of indecisive (or \"tie\") outcomes by up to 54% compared to a random sample when focusing on the top-20 percentile of prioritized instances. This potential reduction in required human effort positions our approach as a valuable strategy in future large language model evaluations."
                    },
                    {
                        "title": "From One to Many: Expanding the Scope of Toxicity Mitigation in Language Models",
                        "abstract": "To date, toxicity mitigation in language models has almost entirely been focused on single-language settings. As language models embrace multilingual capabilities, it's crucial our safety measures keep pace. Recognizing this research gap, our approach expands the scope of conventional toxicity mitigation to address the complexities presented by multiple languages. In the absence of sufficient annotated datasets across languages, we employ translated data to evaluate and enhance our mitigation techniques. We also compare finetuning mitigation approaches against retrieval-augmented techniques under both static and continual toxicity mitigation scenarios. This allows us to examine the effects of translation quality and the cross-lingual transfer on toxicity mitigation. We also explore how model size and data quantity affect the success of these mitigation efforts. Covering nine languages, our study represents a broad array of linguistic families and levels of resource availability, ranging from high to mid-resource languages. Through comprehensive experiments, we provide insights into the complexities of multilingual toxicity mitigation, offering valuable insights and paving the way for future research in this increasingly important field. Code and data are available at https://github.com/for-ai/goodtriever."
                    },
                    {
                        "title": "Memory Efficient Continual Learning with Transformers",
                        "abstract": "In many real-world scenarios, data to train machine learning models becomes available over time. Unfortunately, these models struggle to continually learn new concepts without forgetting what has been learnt in the past. This phenomenon is known as catastrophic forgetting and it is difficult to prevent due to practical constraints. For instance, the amount of data that can be stored or the computational resources that can be used might be limited. Moreover, applications increasingly rely on large pre-trained neural networks, such as pre-trained Transformers, since the resources or data might not be available in sufficiently large quantities to practitioners to train the model from scratch. In this paper, we devise a method to incrementally train a model on a sequence of tasks using pre-trained Transformers and extending them with Adapters. Different than the existing approaches, our method is able to scale to a large number of tasks without significant overhead and allows sharing information across tasks. On both image and text classification tasks, we empirically demonstrate that our method maintains a good predictive performance without retraining the model or increasing the number of model parameters over time. The resulting model is also significantly faster at inference time compared to Adapter-based state-of-the-art methods."
                    },
                    {
                        "title": "Linear Bandits with Stochastic Delayed Feedback",
                        "abstract": "Stochastic linear bandits are a natural and well-studied model for structured exploration/exploitation problems and are widely used in applications such as online marketing and recommendation. One of the main challenges faced by practitioners hoping to apply existing algorithms is that usually the feedback is randomly delayed and delays are only partially observable. For example, while a purchase is usually observable some time after the display, the decision of not buying is never explicitly sent to the system. In other words, the learner only observes delayed positive events. We formalize this problem as a novel stochastic delayed linear bandit and propose ${\\tt OTFLinUCB}$ and ${\\tt OTFLinTS}$, two computationally efficient algorithms able to integrate new information as it becomes available and to deal with the permanently censored feedback. We prove optimal $\\tilde O(\\smash{d\\sqrt{T}})$ bounds on the regret of the first algorithm and study the dependency on delay-dependent parameters. Our model, assumptions and results are validated by experiments on simulated and real data."
                    },
                    {
                        "title": "PASHA: Efficient HPO and NAS with Progressive Resource Allocation",
                        "abstract": "Hyperparameter optimization (HPO) and neural architecture search (NAS) are methods of choice to obtain the best-in-class machine learning models, but in practice they can be costly to run. When models are trained on large datasets, tuning them with HPO or NAS rapidly becomes prohibitively expensive for practitioners, even when efficient multi-fidelity methods are employed. We propose an approach to tackle the challenge of tuning machine learning models trained on large datasets with limited computational resources. Our approach, named PASHA, extends ASHA and is able to dynamically allocate maximum resources for the tuning procedure depending on the need. The experimental comparison shows that PASHA identifies well-performing hyperparameter configurations and architectures while consuming significantly fewer computational resources than ASHA."
                    },
                    {
                        "title": "Mix Data or Merge Models? Optimizing for Diverse Multi-Task Learning",
                        "abstract": "Large Language Models (LLMs) have been adopted and deployed worldwide for a broad variety of applications. However, ensuring their safe use remains a significant challenge. Preference training and safety measures often overfit to harms prevalent in Western-centric datasets, and safety protocols frequently fail to extend to multilingual settings. In this work, we explore model merging in a diverse multi-task setting, combining safety and general-purpose tasks within a multilingual context. Each language introduces unique and varied learning challenges across tasks. We find that objective-based merging is more effective than mixing data, with improvements of up to 8% and 10% in general performance and safety respectively. We also find that language-based merging is highly effective -- by merging monolingually fine-tuned models, we achieve a 4% increase in general performance and 7% reduction in harm across all languages on top of the data mixtures method using the same available data. Overall, our comprehensive study of merging approaches provides a useful framework for building strong and safe multilingual models."
                    },
                    {
                        "title": "Investigating Continual Pretraining in Large Language Models: Insights and Implications",
                        "abstract": "This paper studies the evolving domain of Continual Learning (CL) in large language models (LLMs), with a focus on developing strategies for efficient and sustainable training. Our primary emphasis is on continual domain-adaptive pretraining, a process designed to equip LLMs with the ability to integrate new information from various domains while retaining previously learned knowledge and enhancing cross-domain knowledge transfer without relying on domain-specific identification. Unlike previous studies, which mostly concentrate on a limited selection of tasks or domains and primarily aim to address the issue of forgetting, our research evaluates the adaptability and capabilities of LLMs to changing data landscapes in practical scenarios. To this end, we introduce a new benchmark designed to measure the adaptability of LLMs to these evolving data environments, offering a comprehensive framework for evaluation. We examine the impact of model size on learning efficacy and forgetting, as well as how the progression and similarity of emerging domains affect the knowledge transfer within these models. Our findings uncover several key insights: (i) when the sequence of domains shows semantic similarity, continual pretraining enables LLMs to better specialize in the current domain compared to stand-alone fine-tuning, (ii) training across a diverse range of domains enhances both backward and forward knowledge transfer, and (iii) smaller models are particularly sensitive to continual pretraining, showing the most significant rates of both forgetting and learning. We posit that our research marks a shift towards establishing a more realistic benchmark for investigating CL in LLMs, and has the potential to play a key role in guiding the direction of future research in the field."
                    }
                ]
            },
            "7afbc922-66f8-49be-a066-2dc3ac81c6fd": {
                "pk": "7afbc922-66f8-49be-a066-2dc3ac81c6fd",
                "name": "Sara Hooker",
                "collaborators": [
                    "Beyza Ermis",
                    "Daniel D'souza",
                    "Luiza Pozzobon",
                    "Patrick Lewis",
                    "Gregory Clark",
                    "Chirag Agarwal",
                    "Dumitru Erhan",
                    "Pieter-Jan Kindermans",
                    "Been Kim",
                    "Aaron Courville"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Model Compression",
                    "Interpretability"
                ],
                "publications": [
                    {
                        "title": "The Hardware Lottery",
                        "abstract": "Hardware, systems and algorithms research communities have historically had different incentive structures and fluctuating motivation to engage with each other explicitly. This historical treatment is odd given that hardware and software have frequently determined which research ideas succeed (and fail). This essay introduces the term hardware lottery to describe when a research idea wins because it is suited to the available software and hardware and not because the idea is superior to alternative research directions. Examples from early computer science history illustrate how hardware lotteries can delay research progress by casting successful ideas as failures. These lessons are particularly salient given the advent of domain specialized hardware which make it increasingly costly to stray off of the beaten path of research ideas. This essay posits that the gains from progress in computing are likely to become even more uneven, with certain research directions moving into the fast-lane while progress on others is further obstructed."
                    },
                    {
                        "title": "On the Limitations of Compute Thresholds as a Governance Strategy",
                        "abstract": "At face value, this essay is about understanding a fairly esoteric governance tool called compute thresholds. However, in order to grapple with whether these thresholds will achieve anything, we must first understand how they came to be. To do so, we need to engage with a decades-old debate at the heart of computer science progress, namely, is bigger always better? Does a certain inflection point of compute result in changes to the risk profile of a model? Hence, this essay may be of interest not only to policymakers and the wider public but also to computer scientists interested in understanding the role of compute in unlocking breakthroughs. This discussion is timely given the wide adoption of compute thresholds in both the White House Executive Orders on AI Safety (EO) and the EU AI Act to identify more risky systems. A key conclusion of this essay is that compute thresholds, as currently implemented, are shortsighted and likely to fail to mitigate risk. The relationship between compute and risk is highly uncertain and rapidly changing. Relying upon compute thresholds overestimates our ability to predict what abilities emerge at different scales. This essay ends with recommendations for a better way forward."
                    },
                    {
                        "title": "A Benchmark for Interpretability Methods in Deep Neural Networks",
                        "abstract": "We propose an empirical measure of the approximate accuracy of feature importance estimates in deep neural networks. Our results across several large-scale image classification datasets show that many popular interpretability methods produce estimates of feature importance that are not better than a random designation of feature importance. Only certain ensemble based approaches---VarGrad and SmoothGrad-Squared---outperform such a random assignment of importance. The manner of ensembling remains critical, we show that some approaches do no better then the underlying method but carry a far higher computational burden."
                    },
                    {
                        "title": "What Do Compressed Deep Neural Networks Forget?",
                        "abstract": "Deep neural network pruning and quantization techniques have demonstrated it is possible to achieve high levels of compression with surprisingly little degradation to test set accuracy. However, this measure of performance conceals significant differences in how different classes and images are impacted by model compression techniques. We find that models with radically different numbers of weights have comparable top-line performance metrics but diverge considerably in behavior on a narrow subset of the dataset. This small subset of data points, which we term Pruning Identified Exemplars (PIEs) are systematically more impacted by the introduction of sparsity. Compression disproportionately impacts model performance on the underrepresented long-tail of the data distribution. PIEs over-index on atypical or noisy images that are far more challenging for both humans and algorithms to classify. Our work provides intuition into the role of capacity in deep neural networks and the trade-offs incurred by compression. An understanding of this disparate impact is critical given the widespread deployment of compressed models in the wild."
                    },
                    {
                        "title": "Characterising Bias in Compressed Models",
                        "abstract": "The popularity and widespread use of pruning and quantization is driven by the severe resource constraints of deploying deep neural networks to environments with strict latency, memory and energy requirements. These techniques achieve high levels of compression with negligible impact on top-line metrics (top-1 and top-5 accuracy). However, overall accuracy hides disproportionately high errors on a small subset of examples; we call this subset Compression Identified Exemplars (CIE). We further establish that for CIE examples, compression amplifies existing algorithmic bias. Pruning disproportionately impacts performance on underrepresented features, which often coincides with considerations of fairness. Given that CIE is a relatively small subset but a great contributor of error in the model, we propose its use as a human-in-the-loop auditing tool to surface a tractable subset of the dataset for further inspection or annotation by a domain expert. We provide qualitative and quantitative support that CIE surfaces the most challenging examples in the data distribution for human-in-the-loop auditing."
                    },
                    {
                        "title": "Studying the impact of magnitude pruning on contrastive learning methods",
                        "abstract": "We study the impact of different pruning techniques on the representation learned by deep neural networks trained with contrastive loss functions. Our work finds that at high sparsity levels, contrastive learning results in a higher number of misclassified examples relative to models trained with traditional cross-entropy loss. To understand this pronounced difference, we use metrics such as the number of PIEs (Hooker et al., 2019), Q-Score (Kalibhat et al., 2022), and PD-Score (Baldock et al., 2021) to measure the impact of pruning on the learned representation quality. Our analysis suggests the schedule of the pruning method implementation matters. We find that the negative impact of sparsity on the quality of the learned representation is the highest when pruning is introduced early on in the training phase."
                    },
                    {
                        "title": "Estimating Example Difficulty Using Variance of Gradients",
                        "abstract": "In machine learning, a question of great interest is understanding what examples are challenging for a model to classify. Identifying atypical examples ensures the safe deployment of models, isolates samples that require further human inspection and provides interpretability into model behavior. In this work, we propose Variance of Gradients (VoG) as a valuable and efficient metric to rank data by difficulty and to surface a tractable subset of the most challenging examples for human-in-the-loop auditing. We show that data points with high VoG scores are far more difficult for the model to learn and over-index on corrupted or memorized examples. Further, restricting the evaluation to the test set instances with the lowest VoG improves the model's generalization performance. Finally, we show that VoG is a valuable and efficient ranking for out-of-distribution detection."
                    },
                    {
                        "title": "Keep the Gradients Flowing: Using Gradient Flow to Study Sparse Network Optimization",
                        "abstract": "Training sparse networks to converge to the same performance as dense neural architectures has proven to be elusive. Recent work suggests that initialization is the key. However, while this direction of research has had some success, focusing on initialization alone appears to be inadequate. In this paper, we take a broader view of training sparse networks and consider the role of regularization, optimization, and architecture choices on sparse models. We propose a simple experimental framework, Same Capacity Sparse vs Dense Comparison (SC-SDC), that allows for a fair comparison of sparse and dense networks. Furthermore, we propose a new measure of gradient flow, Effective Gradient Flow (EGF), that better correlates to performance in sparse networks. Using top-line metrics, SC-SDC and EGF, we show that default choices of optimizers, activation functions and regularizers used for dense networks can disadvantage sparse networks. Based upon these findings, we show that gradient flow in sparse networks can be improved by reconsidering aspects of the architecture design and the training regime. Our work suggests that initialization is only one piece of the puzzle and taking a wider view of tailoring optimization to sparse networks yields promising results."
                    },
                    {
                        "title": "The State of Sparsity in Deep Neural Networks",
                        "abstract": "We rigorously evaluate three state-of-the-art techniques for inducing sparsity in deep neural networks on two large-scale learning tasks: Transformer trained on WMT 2014 English-to-German, and ResNet-50 trained on ImageNet. Across thousands of experiments, we demonstrate that complex techniques (Molchanov et al., 2017; Louizos et al., 2017b) shown to yield high compression rates on smaller datasets perform inconsistently, and that simple magnitude pruning approaches achieve comparable or better results. Additionally, we replicate the experiments performed by (Frankle & Carbin, 2018) and (Liu et al., 2018) at scale and show that unstructured sparse architectures learned through pruning cannot be trained from scratch to the same test set performance as a model trained with joint sparsification and optimization. Together, these results highlight the need for large-scale benchmarks in the field of model compression. We open-source our code, top performing model checkpoints, and results of all hyperparameter configurations to establish rigorous baselines for future work on compression and sparsification."
                    },
                    {
                        "title": "A Tale Of Two Long Tails",
                        "abstract": "As machine learning models are increasingly employed to assist human decision-makers, it becomes critical to communicate the uncertainty associated with these model predictions. However, the majority of work on uncertainty has focused on traditional probabilistic or ranking approaches - where the model assigns low probabilities or scores to uncertain examples. While this captures what examples are challenging for the model, it does not capture the underlying source of the uncertainty. In this work, we seek to identify examples the model is uncertain about and characterize the source of said uncertainty. We explore the benefits of designing a targeted intervention - targeted data augmentation of the examples where the model is uncertain over the course of training. We investigate whether the rate of learning in the presence of additional information differs between atypical and noisy examples? Our results show that this is indeed the case, suggesting that well-designed interventions over the course of training can be an effective way to characterize and distinguish between different sources of uncertainty."
                    },
                    {
                        "title": "On the Challenges of Using Black-Box APIs for Toxicity Evaluation in Research",
                        "abstract": "Perception of toxicity evolves over time and often differs between geographies and cultural backgrounds. Similarly, black-box commercially available APIs for detecting toxicity, such as the Perspective API, are not static, but frequently retrained to address any unattended weaknesses and biases. We evaluate the implications of these changes on the reproducibility of findings that compare the relative merits of models and methods that aim to curb toxicity. Our findings suggest that research that relied on inherited automatic toxicity scores to compare models and techniques may have resulted in inaccurate findings. Rescoring all models from HELM, a widely respected living benchmark, for toxicity with the recent version of the API led to a different ranking of widely used foundation models. We suggest caution in applying apples-to-apples comparisons between studies and lay recommendations for a more structured approach to evaluating toxicity over time. Code and data are available at https://github.com/for-ai/black-box-api-challenges."
                    },
                    {
                        "title": "Goodtriever: Adaptive Toxicity Mitigation with Retrieval-augmented Models",
                        "abstract": "Considerable effort has been dedicated to mitigating toxicity, but existing methods often require drastic modifications to model parameters or the use of computationally intensive auxiliary models. Furthermore, previous approaches have often neglected the crucial factor of language's evolving nature over time. In this work, we present a comprehensive perspective on toxicity mitigation that takes into account its changing nature. We introduce Goodtriever, a flexible methodology that matches the current state-of-the-art toxicity mitigation while achieving 43% relative latency reduction during inference and being more computationally efficient. By incorporating a retrieval-based approach at decoding time, Goodtriever enables toxicity-controlled text generation. Our research advocates for an increased focus on adaptable mitigation techniques, which better reflect the data drift models face when deployed in the wild. Code and data are available at https://github.com/for-ai/goodtriever."
                    },
                    {
                        "title": "Locally Differentially Private Document Generation Using Zero Shot Prompting",
                        "abstract": "Numerous studies have highlighted the privacy risks associated with pretrained large language models. In contrast, our research offers a unique perspective by demonstrating that pretrained large language models can effectively contribute to privacy preservation. We propose a locally differentially private mechanism called DP-Prompt, which leverages the power of pretrained large language models and zero-shot prompting to counter author de-anonymization attacks while minimizing the impact on downstream utility. When DP-Prompt is used with a powerful language model like ChatGPT (gpt-3.5), we observe a notable reduction in the success rate of de-anonymization attacks, showing that it surpasses existing approaches by a considerable margin despite its simpler design. For instance, in the case of the IMDB dataset, DP-Prompt (with ChatGPT) perfectly recovers the clean sentiment F1 score while achieving a 46\\% reduction in author identification F1 score against static attackers and a 26\\% reduction against adaptive attackers. We conduct extensive experiments across six open-source large language models, ranging up to 7 billion parameters, to analyze various effects of the privacy-utility tradeoff."
                    },
                    {
                        "title": "The Low-Resource Double Bind: An Empirical Study of Pruning for Low-Resource Machine Translation",
                        "abstract": "A \"bigger is better\" explosion in the number of parameters in deep neural networks has made it increasingly challenging to make state-of-the-art networks accessible in compute-restricted environments. Compression techniques have taken on renewed importance as a way to bridge the gap. However, evaluation of the trade-offs incurred by popular compression techniques has been centered on high-resource datasets. In this work, we instead consider the impact of compression in a data-limited regime. We introduce the term low-resource double bind to refer to the co-occurrence of data limitations and compute resource constraints. This is a common setting for NLP for low-resource languages, yet the trade-offs in performance are poorly studied. Our work offers surprising insights into the relationship between capacity and generalization in data-limited regimes for the task of machine translation. Our experiments on magnitude pruning for translations from English into Yoruba, Hausa, Igbo and German show that in low-resource regimes, sparsity preserves performance on frequent sentences but has a disparate impact on infrequent ones. However, it improves robustness to out-of-distribution shifts, especially for datasets that are very distinct from the training distribution. Our findings suggest that sparsity can play a beneficial role at curbing memorization of low frequency attributes, and therefore offers a promising solution to the low-resource double bind."
                    },
                    {
                        "title": "Which Prompts Make The Difference? Data Prioritization For Efficient Human LLM Evaluation",
                        "abstract": "Human evaluation is increasingly critical for assessing large language models, capturing linguistic nuances, and reflecting user preferences more accurately than traditional automated metrics. However, the resource-intensive nature of this type of annotation process poses significant challenges. The key question driving our work: \"is it feasible to minimize human-in-the-loop feedback by prioritizing data instances which most effectively distinguish between models?\" We evaluate several metric-based methods and find that these metrics enhance the efficiency of human evaluations by minimizing the number of required annotations, thus saving time and cost, while ensuring a robust performance evaluation. We show that our method is effective across widely used model families, reducing instances of indecisive (or \"tie\") outcomes by up to 54% compared to a random sample when focusing on the top-20 percentile of prioritized instances. This potential reduction in required human effort positions our approach as a valuable strategy in future large language model evaluations."
                    },
                    {
                        "title": "From One to Many: Expanding the Scope of Toxicity Mitigation in Language Models",
                        "abstract": "To date, toxicity mitigation in language models has almost entirely been focused on single-language settings. As language models embrace multilingual capabilities, it's crucial our safety measures keep pace. Recognizing this research gap, our approach expands the scope of conventional toxicity mitigation to address the complexities presented by multiple languages. In the absence of sufficient annotated datasets across languages, we employ translated data to evaluate and enhance our mitigation techniques. We also compare finetuning mitigation approaches against retrieval-augmented techniques under both static and continual toxicity mitigation scenarios. This allows us to examine the effects of translation quality and the cross-lingual transfer on toxicity mitigation. We also explore how model size and data quantity affect the success of these mitigation efforts. Covering nine languages, our study represents a broad array of linguistic families and levels of resource availability, ranging from high to mid-resource languages. Through comprehensive experiments, we provide insights into the complexities of multilingual toxicity mitigation, offering valuable insights and paving the way for future research in this increasingly important field. Code and data are available at https://github.com/for-ai/goodtriever."
                    },
                    {
                        "title": "Multilingual Arbitrage: Optimizing Data Pools to Accelerate Multilingual Progress",
                        "abstract": "The use of synthetic data has played a critical role in recent state-of-art breakthroughs. However, overly relying on a single oracle teacher model to generate data has been shown to lead to model collapse and invite propagation of biases. These limitations are particularly evident in multilingual settings, where the absence of a universally effective teacher model that excels across all languages presents significant challenges. In this work, we address these extreme difference by introducing \"multilingual arbitrage\", which capitalizes on performance variations between multiple models for a given language. To do so, we strategically route samples through a diverse pool of models, each with unique strengths in different languages. Across exhaustive experiments on state-of-art models, our work suggests that arbitrage techniques allow for spectacular gains in performance that far outperform relying on a single teacher. In particular, compared to the best single teacher, we observe gains of up to 56.5% improvement in win rates averaged across all languages when switching to multilingual arbitrage. We observe the most significant gains for the least resourced languages in our pool."
                    },
                    {
                        "title": "Randomness In Neural Network Training: Characterizing The Impact of Tooling",
                        "abstract": "The quest for determinism in machine learning has disproportionately focused on characterizing the impact of noise introduced by algorithmic design choices. In this work, we address a less well understood and studied question: how does our choice of tooling introduce randomness to deep neural network training. We conduct large scale experiments across different types of hardware, accelerators, state of art networks, and open-source datasets, to characterize how tooling choices contribute to the level of non-determinism in a system, the impact of said non-determinism, and the cost of eliminating different sources of noise.   Our findings are surprising, and suggest that the impact of non-determinism in nuanced. While top-line metrics such as top-1 accuracy are not noticeably impacted, model performance on certain parts of the data distribution is far more sensitive to the introduction of randomness. Our results suggest that deterministic tooling is critical for AI safety. However, we also find that the cost of ensuring determinism varies dramatically between neural network architectures and hardware types, e.g., with overhead up to $746\\%$, $241\\%$, and $196\\%$ on a spectrum of widely used GPU accelerator architectures, relative to non-deterministic training. The source code used in this paper is available at https://github.com/usyd-fsalab/NeuralNetworkRandomness."
                    },
                    {
                        "title": "When less is more: Simplifying inputs aids neural network understanding",
                        "abstract": "How do neural network image classifiers respond to simpler and simpler inputs? And what do such responses reveal about the learning process? To answer these questions, we need a clear measure of input simplicity (or inversely, complexity), an optimization objective that correlates with simplification, and a framework to incorporate such objective into training and inference. Lastly we need a variety of testbeds to experiment and evaluate the impact of such simplification on learning. In this work, we measure simplicity with the encoding bit size given by a pretrained generative model, and minimize the bit size to simplify inputs in training and inference. We investigate the effect of such simplification in several scenarios: conventional training, dataset condensation and post-hoc explanations. In all settings, inputs are simplified along with the original classification task, and we investigate the trade-off between input simplicity and task performance. For images with injected distractors, such simplification naturally removes superfluous information. For dataset condensation, we find that inputs can be simplified with almost no accuracy degradation. When used in post-hoc explanation, our learning-based simplification approach offers a valuable new tool to explore the basis of network decisions."
                    }
                ]
            },
            "74ec45c8-5289-40e5-bd23-8fd51cffd69b": {
                "pk": "74ec45c8-5289-40e5-bd23-8fd51cffd69b",
                "name": "Marzieh Fadaee",
                "collaborators": [
                    "Christof Monz",
                    "Sara Hooker",
                    "Arianna Bisazza",
                    "Beyza Ermis",
                    "Luiz Bonifacio",
                    "Hugo Abonizio",
                    "Rodrigo Nogueira",
                    "Jakub Zavrel",
                    "Aakanksha",
                    "Arash Ahmadian"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Translation",
                    "Data Augmentation",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "Understanding and Enhancing the Use of Context for Machine Translation",
                        "abstract": "To understand and infer meaning in language, neural models have to learn complicated nuances. Discovering distinctive linguistic phenomena from data is not an easy task. For instance, lexical ambiguity is a fundamental feature of language which is challenging to learn. Even more prominently, inferring the meaning of rare and unseen lexical units is difficult with neural networks. Meaning is often determined from context. With context, languages allow meaning to be conveyed even when the specific words used are not known by the reader. To model this learning process, a system has to learn from a few instances in context and be able to generalize well to unseen cases. The learning process is hindered when training data is scarce for a task. Even with sufficient data, learning patterns for the long tail of the lexical distribution is challenging. In this thesis, we focus on understanding certain potentials of contexts in neural models and design augmentation models to benefit from them. We focus on machine translation as an important instance of the more general language understanding problem. To translate from a source language to a target language, a neural model has to understand the meaning of constituents in the provided context and generate constituents with the same meanings in the target language. This task accentuates the value of capturing nuances of language and the necessity of generalization from few observations. The main problem we study in this thesis is what neural machine translation models learn from data and how we can devise more focused contexts to enhance this learning. Looking more in-depth into the role of context and the impact of data on learning models is essential to advance the NLP field. Moreover, it helps highlight the vulnerabilities of current neural networks and provides insights into designing more robust models."
                    },
                    {
                        "title": "The Unreasonable Volatility of Neural Machine Translation Models",
                        "abstract": "Recent works have shown that Neural Machine Translation (NMT) models achieve impressive performance, however, questions about understanding the behavior of these models remain unanswered. We investigate the unexpected volatility of NMT models where the input is semantically and syntactically correct. We discover that with trivial modifications of source sentences, we can identify cases where \\textit{unexpected changes} happen in the translation and in the worst case lead to mistranslations. This volatile behavior of translating extremely similar sentences in surprisingly different ways highlights the underlying generalization problem of current NMT models. We find that both RNN and Transformer models display volatile behavior in 26% and 19% of sentence variations, respectively."
                    },
                    {
                        "title": "Back-Translation Sampling by Targeting Difficult Words in Neural Machine Translation",
                        "abstract": "Neural Machine Translation has achieved state-of-the-art performance for several language pairs using a combination of parallel and synthetic data. Synthetic data is often generated by back-translating sentences randomly sampled from monolingual data using a reverse translation model. While back-translation has been shown to be very effective in many cases, it is not entirely clear why. In this work, we explore different aspects of back-translation, and show that words with high prediction loss during training benefit most from the addition of synthetic data. We introduce several variations of sampling strategies targeting difficult-to-predict words using prediction losses and frequencies of words. In addition, we also target the contexts of difficult words and sample sentences that are similar in context. Experimental results for the WMT news translation task show that our method improves translation quality by up to 1.7 and 1.2 Bleu points over back-translation using random sampling for German-English and English-German, respectively."
                    },
                    {
                        "title": "Data Augmentation for Low-Resource Neural Machine Translation",
                        "abstract": "The quality of a Neural Machine Translation system depends substantially on the availability of sizable parallel corpora. For low-resource language pairs this is not the case, resulting in poor translation quality. Inspired by work in computer vision, we propose a novel data augmentation approach that targets low-frequency words by generating new sentence pairs containing rare words in new, synthetically created contexts. Experimental results on simulated low-resource settings show that our method improves translation quality by up to 2.9 BLEU points over the baseline and up to 3.2 BLEU over back-translation."
                    },
                    {
                        "title": "Learning Topic-Sensitive Word Representations",
                        "abstract": "Distributed word representations are widely used for modeling words in NLP tasks. Most of the existing models generate one representation per word and do not consider different meanings of a word. We present two approaches to learn multiple topic-sensitive representations per word by using Hierarchical Dirichlet Process. We observe that by modeling topics and integrating topic distributions for each document we obtain representations that are able to distinguish between different meanings of a given word. Our models yield statistically significant improvements for the lexical substitution task indicating that commonly used single word representations, even when combined with contextual information, are insufficient for this task."
                    },
                    {
                        "title": "Examining the Tip of the Iceberg: A Data Set for Idiom Translation",
                        "abstract": "Neural Machine Translation (NMT) has been widely used in recent years with significant improvements for many language pairs. Although state-of-the-art NMT systems are generating progressively better translations, idiom translation remains one of the open challenges in this field. Idioms, a category of multiword expressions, are an interesting language phenomenon where the overall meaning of the expression cannot be composed from the meanings of its parts. A first important challenge is the lack of dedicated data sets for learning and evaluating idiom translation. In this paper we address this problem by creating the first large-scale data set for idiom translation. Our data set is automatically extracted from a widely used German-English translation corpus and includes, for each language direction, a targeted evaluation set where all sentences contain idioms and a regular training corpus where sentences including idioms are marked. We release this data set and use it to perform preliminary NMT experiments as the first step towards better idiom translation."
                    },
                    {
                        "title": "InPars: Data Augmentation for Information Retrieval using Large Language Models",
                        "abstract": "The information retrieval community has recently witnessed a revolution due to large pretrained transformer models. Another key ingredient for this revolution was the MS MARCO dataset, whose scale and diversity has enabled zero-shot transfer learning to various tasks. However, not all IR tasks and domains can benefit from one single dataset equally. Extensive research in various NLP tasks has shown that using domain-specific training data, as opposed to a general-purpose one, improves the performance of neural models. In this work, we harness the few-shot capabilities of large pretrained language models as synthetic data generators for IR tasks. We show that models finetuned solely on our unsupervised dataset outperform strong baselines such as BM25 as well as recently proposed self-supervised dense retrieval methods. Furthermore, retrievers finetuned on both supervised and our synthetic data achieve better zero-shot transfer than models finetuned only on supervised data. Code, models, and data are available at https://github.com/zetaalphavector/inpars ."
                    },
                    {
                        "title": "A New Neural Search and Insights Platform for Navigating and Organizing AI Research",
                        "abstract": "To provide AI researchers with modern tools for dealing with the explosive growth of the research literature in their field, we introduce a new platform, AI Research Navigator, that combines classical keyword search with neural retrieval to discover and organize relevant literature. The system provides search at multiple levels of textual granularity, from sentences to aggregations across documents, both in natural language and through navigation in a domain-specific Knowledge Graph. We give an overview of the overall architecture of the system and of the components for document analysis, question answering, search, analytics, expert search, and recommendations."
                    },
                    {
                        "title": "Which Prompts Make The Difference? Data Prioritization For Efficient Human LLM Evaluation",
                        "abstract": "Human evaluation is increasingly critical for assessing large language models, capturing linguistic nuances, and reflecting user preferences more accurately than traditional automated metrics. However, the resource-intensive nature of this type of annotation process poses significant challenges. The key question driving our work: \"is it feasible to minimize human-in-the-loop feedback by prioritizing data instances which most effectively distinguish between models?\" We evaluate several metric-based methods and find that these metrics enhance the efficiency of human evaluations by minimizing the number of required annotations, thus saving time and cost, while ensuring a robust performance evaluation. We show that our method is effective across widely used model families, reducing instances of indecisive (or \"tie\") outcomes by up to 54% compared to a random sample when focusing on the top-20 percentile of prioritized instances. This potential reduction in required human effort positions our approach as a valuable strategy in future large language model evaluations."
                    },
                    {
                        "title": "Diversify and Conquer: Diversity-Centric Data Selection with Iterative Refinement",
                        "abstract": "Finetuning large language models on instruction data is crucial for enhancing pre-trained knowledge and improving instruction-following capabilities. As instruction datasets proliferate, selecting optimal data for effective training becomes increasingly important. This work addresses the question: How can we determine the optimal subset of data for effective training? While existing research often emphasizes local criteria like instance quality for subset selection, we argue that a global approach focused on data diversity is more critical. Our method employs k-means clustering to ensure the selected subset effectively represents the full dataset. We propose an iterative refinement method inspired by active learning techniques to resample instances from clusters, reassessing each cluster's importance and sampling weight in every training iteration. This approach reduces the effect of outliers and automatically filters out clusters containing low-quality data. Through extensive evaluation across natural language reasoning, general world knowledge, code and math reasoning tasks, and by fine-tuning models from various families, we observe consistent improvements, achieving a 7% increase over random selection and a 3.8% improvement over state-of-the-art sampling methods. Our work highlights the significance of diversity-first sampling when finetuning LLMs to enhance performance across a broad array of evaluation tasks. Our code is available at https://github.com/for-ai/iterative-data-selection."
                    },
                    {
                        "title": "InPars-v2: Large Language Models as Efficient Dataset Generators for Information Retrieval",
                        "abstract": "Recently, InPars introduced a method to efficiently use large language models (LLMs) in information retrieval tasks: via few-shot examples, an LLM is induced to generate relevant queries for documents. These synthetic query-document pairs can then be used to train a retriever. However, InPars and, more recently, Promptagator, rely on proprietary LLMs such as GPT-3 and FLAN to generate such datasets. In this work we introduce InPars-v2, a dataset generator that uses open-source LLMs and existing powerful rerankers to select synthetic query-document pairs for training. A simple BM25 retrieval pipeline followed by a monoT5 reranker finetuned on InPars-v2 data achieves new state-of-the-art results on the BEIR benchmark. To allow researchers to further improve our method, we open source the code, synthetic data, and finetuned models: https://github.com/zetaalphavector/inPars/tree/master/tpu"
                    },
                    {
                        "title": "Mix Data or Merge Models? Optimizing for Diverse Multi-Task Learning",
                        "abstract": "Large Language Models (LLMs) have been adopted and deployed worldwide for a broad variety of applications. However, ensuring their safe use remains a significant challenge. Preference training and safety measures often overfit to harms prevalent in Western-centric datasets, and safety protocols frequently fail to extend to multilingual settings. In this work, we explore model merging in a diverse multi-task setting, combining safety and general-purpose tasks within a multilingual context. Each language introduces unique and varied learning challenges across tasks. We find that objective-based merging is more effective than mixing data, with improvements of up to 8% and 10% in general performance and safety respectively. We also find that language-based merging is highly effective -- by merging monolingually fine-tuned models, we achieve a 4% increase in general performance and 7% reduction in harm across all languages on top of the data mixtures method using the same available data. Overall, our comprehensive study of merging approaches provides a useful framework for building strong and safe multilingual models."
                    },
                    {
                        "title": "When Less is More: Investigating Data Pruning for Pretraining LLMs at Scale",
                        "abstract": "Large volumes of text data have contributed significantly to the development of large language models (LLMs) in recent years. This data is typically acquired by scraping the internet, leading to pretraining datasets comprised of noisy web text. To date, efforts to prune these datasets down to a higher quality subset have relied on hand-crafted heuristics encoded as rule-based filters. In this work, we take a wider view and explore scalable estimates of data quality that can be used to systematically measure the quality of pretraining data. We perform a rigorous comparison at scale of the simple data quality estimator of perplexity, as well as more sophisticated and computationally intensive estimates of the Error L2-Norm and memorization. These metrics are used to rank and prune pretraining corpora, and we subsequently compare LLMs trained on these pruned datasets. Surprisingly, we find that the simple technique of perplexity outperforms our more computationally expensive scoring methods. We improve over our no-pruning baseline while training on as little as 30% of the original training dataset. Our work sets the foundation for unexplored strategies in automatically curating high quality corpora and suggests the majority of pretraining data can be removed while retaining performance."
                    },
                    {
                        "title": "The Multilingual Alignment Prism: Aligning Global and Local Preferences to Reduce Harm",
                        "abstract": "A key concern with the concept of \"alignment\" is the implicit question of \"alignment to what?\". AI systems are increasingly used across the world, yet safety alignment is often focused on homogeneous monolingual settings. Additionally, preference training and safety measures often overfit to harms common in Western-centric datasets. Here, we explore the viability of different alignment approaches when balancing dual objectives: addressing and optimizing for a non-homogeneous set of languages and cultural preferences while minimizing both global and local harms. We collect the first set of human annotated red-teaming prompts in different languages distinguishing between global and local harm, which serve as a laboratory for understanding the reliability of alignment techniques when faced with preference distributions that are non-stationary across geographies and languages. While this setting is seldom covered by the literature to date, which primarily centers on English harm mitigation, it captures real-world interactions with AI systems around the world. We establish a new precedent for state-of-the-art alignment techniques across 6 languages with minimal degradation in general performance. Our work provides important insights into cross-lingual transfer and novel optimization approaches to safeguard AI systems designed to serve global populations."
                    },
                    {
                        "title": "LLM See, LLM Do: Guiding Data Generation to Target Non-Differentiable Objectives",
                        "abstract": "The widespread adoption of synthetic data raises new questions about how models generating the data can influence other large language models (LLMs) via distilled data. To start, our work exhaustively characterizes the impact of passive inheritance of model properties by systematically studying the consequences of synthetic data integration. We provide one of the most comprehensive studies to-date of how the source of synthetic data shapes models' internal biases, calibration and generations' textual attributes and preferences. We find that models are surprisingly sensitive towards certain attributes even when the synthetic data prompts appear \"neutral\". which invites the question whether this sensitivity can be exploited for good.   Our findings invite the question can we explicitly steer the models towards the properties we want at test time by exploiting the data generation process? This would have historically been considered infeasible due to the cost of collecting data with a specific characteristic or objective in mind. However, improvement in the quality of synthetic data, as well as a shift towards general-purpose models designed to follow a diverse way of instructions, means this question is timely. We propose active inheritance as a term to describe intentionally constraining synthetic data according to a non-differentiable objective. We demonstrate how active inheritance can steer the generation profiles of models towards desirable non-differentiable attributes, e.g. high lexical diversity or low toxicity."
                    }
                ]
            }
        }
    },
    "2406.06425": {
        "paper_data": {
            "title": "Multivariate Stochastic Dominance via Optimal Transport and Applications to Models Benchmarking",
            "url": "http://arxiv.org/abs/2406.06425v1",
            "arxiv_id": "2406.06425",
            "authors": [
                "Gabriel Rioux",
                "Apoorva Nitsure",
                "Mattia Rigotti",
                "Kristjan Greenewald",
                "Youssef Mroueh"
            ],
            "abstract": "Stochastic dominance is an important concept in probability theory, econometrics and social choice theory for robustly modeling agents' preferences between random outcomes. While many works have been dedicated to the univariate case, little has been done in the multivariate scenario, wherein an agent has to decide between different multivariate outcomes. By exploiting a characterization of multivariate first stochastic dominance in terms of couplings, we introduce a statistic that assesses multivariate almost stochastic dominance under the framework of Optimal Transport with a smooth cost. Further, we introduce an entropic regularization of this statistic, and establish a central limit theorem (CLT) and consistency of the bootstrap procedure for the empirical statistic. Armed with this CLT, we propose a hypothesis testing framework as well as an efficient implementation using the Sinkhorn algorithm. We showcase our method in comparing and benchmarking Large Language Models that are evaluated on multiple metrics. Our multivariate stochastic dominance test allows us to capture the dependencies between the metrics in order to make an informed and statistically significant decision on the relative performance of the models.",
            "introduction": "   1. Introduction  In choice theory, economics, finance, and models benchmarking, agents are faced with stochastic prospects that are (eventually) multivariate random variables which they wish to order according to a utility or risk measure of interest. To formalize such a notion of ordering of stochastic quantities, the concept of stochastic dominance can be utilized.   In the univariate case, a standard notion of order is given by First order Stochastic Dominance (FSD) which can be expressed in terms of the quantiles of the underlying random variables (Ogryczak and Ruszczynski, 2002). To wit, a random variable X\ud835\udc4bXitalic_X dominates another random variable Y\ud835\udc4cYitalic_Y in FSD, if it has larger quantiles than Y\ud835\udc4cYitalic_Y across all percentiles. A weaker notion of FSD, called almost stochastic dominance was introduced in Del\u00a0Barrio et\u00a0al. (2018); del Barrio et\u00a0al. (2018). Their approach is based on optimal transport (OT) and consists of measuring a ratio which quantifies how close X\ud835\udc4bXitalic_X is to dominating Y\ud835\udc4cYitalic_Y. Del\u00a0Barrio et\u00a0al. (2018); del Barrio et\u00a0al. (2018) further lay the groundwork for principled statistical analysis of almost FSD by establishing a central limit theorem for the empirical ratio as well as consistency of the bootstrap. Dror et\u00a0al. (2018); Ulmer et\u00a0al. (2022); Nitsure et\u00a0al. (2023) used the almost FSD testing framework in benchmarking Large Language models to make statistically significant decisions regarding which model to select when these models are evaluated with a metric of interest on test data.   Our main motivation is to extend the testing framework for almost FSD of Del\u00a0Barrio et\u00a0al. (2018) to the multivariate case as to enable applications with dependencies between metrics. For instance, the problem of multivariate portfolio selection in financial applications has been treated via a reduction to univariate orders (Kouaissah, 2021). Another application of interest is that of multi-metric benchmarking which is central nowadays in ranking and selecting Large Languages Models (Bommasani et\u00a0al., 2023; Chang et\u00a0al., 2023; Huang et\u00a0al., 2023; MosaicLM, 2023; Beeching et\u00a0al., 2023; Zhang and Hardt, 2024). Current approaches such as Nitsure et\u00a0al. (2023) use aggregation techniques to reduce the ordering to the univariate case thereby ignoring dependencies between metrics.   Our starting point is the so-called standard multivariate stochastic order (see Chapter 6 in Shaked and Shanthikumar, 2007). This order can be expressed in terms of couplings between random vectors. This insight allows us to follow the approach of Del\u00a0Barrio et\u00a0al. (2018) and define an almost multivariate FSD via OT in Section 2. This notion of stochastic dominance can be defined using multivariate violation ratios that are expressed as optimal transport problems with smooth costs (Manole and Niles-Weed, 2024; Hundrieser et\u00a0al., 2022; Groppe and Hundrieser, 2023). Given that empirical OT suffers from the curse of dimensionality, we resort in Section 3 to entropic regularization (Cuturi, 2013) to alleviate that issue and hence define Entropic Multivariate Violation Ratios. We establish in Section 3 convergence of these entropic violation ratios as the regularization parameter goes to zero, as well as a central limit theorem and bootstrapping consistency in Section 4 using the functional delta method (R\u00f6misch, 2006). We highlight that the delta method has seen general success for proving limit theorems with entropic OT (Hundrieser et\u00a0al.,",
            "references": [
                {
                    "title": "Inherent Trade-Offs between Diversity and Stability in Multi-Task Benchmarks",
                    "abstract": "We examine multi-task benchmarks in machine learning through the lens of social choice theory. We draw an analogy between benchmarks and electoral systems, where models are candidates and tasks are voters. This suggests a distinction between cardinal and ordinal benchmark systems. The former aggregate numerical scores into one model ranking; the latter aggregate rankings for each task. We apply Arrow's impossibility theorem to ordinal benchmarks to highlight the inherent limitations of ordinal systems, particularly their sensitivity to the inclusion of irrelevant models. Inspired by Arrow's theorem, we empirically demonstrate a strong trade-off between diversity and sensitivity to irrelevant changes in existing multi-task benchmarks. Our result is based on new quantitative measures of diversity and sensitivity that we introduce. Sensitivity quantifies the impact that irrelevant changes to tasks have on a benchmark. Diversity captures the degree of disagreement in model rankings across tasks. We develop efficient approximation algorithms for both measures, as exact computation is computationally challenging. Through extensive experiments on seven cardinal benchmarks and eleven ordinal benchmarks, we demonstrate a clear trade-off between diversity and stability: The more diverse a multi-task benchmark, the more sensitive to trivial changes it is. Additionally, we show that the aggregated rankings of existing benchmarks are highly unstable under irrelevant changes. The codes and data are available at https://socialfoundations.github.io/benchbench/."
                },
                {
                    "title": "PyTorch 2: Faster Machine Learning Through Dynamic Python Bytecode Transformation and Graph Compilation",
                    "abstract": "This paper introduces two extensions to the popular PyTorch machine learning framework, TorchDynamo and TorchInductor, which implement the torch.compile feature released in PyTorch 2. TorchDynamo is a Python-level just-in-time (JIT) compiler that enables graph compilation in PyTorch programs without sacrificing the flexibility of Python. It achieves this by dynamically modifying Python bytecode before execution and extracting sequences of PyTorch operations into an FX graph, which is then JIT compiled using one of many extensible backends. TorchInductor is the default compiler backend for TorchDynamo, which translates PyTorch programs into OpenAI's Triton for GPUs and C++ for CPUs. Results show that TorchDynamo is able to capture graphs more robustly than prior approaches while adding minimal overhead, and TorchInductor is able to provide a 2.27\u00d7 inference and 1.41\u00d7 training geometric mean speedup on an NVIDIA A100 GPU across 180+ real-world models, which outperforms six other compilers. These extensions provide a new way to apply optimizations through compilers in eager mode frameworks like PyTorch."
                },
                {
                    "title": "Limit distributions and sensitivity analysis for empirical entropic optimal transport on countable spaces",
                    "abstract": "For probability measures supported on countable spaces we derive limit distributions for empirical entropic optimal transport quantities. In particular, we prove that the corresponding plan converges weakly to a centered Gaussian process. Furthermore, its optimal value is shown to be asymptotically normal. The results are valid for a large class of ground cost functions and generalize recently obtained limit laws for empirical entropic optimal transport quantities on finite spaces. Our proofs are based on a sensitivity analysis with respect to a weighted l-norm relying on the dual formulation of entropic optimal transport as well as necessary and sufficient optimality conditions for the entropic transport plan. This can be used to derive weak convergence of the empirical entropic optimal transport plan and value that results in weighted BorisovDudley-Durst conditions on the underlying probability measures. The weights are linked to an exponential penalty term for dual entropic optimal transport and the underlying ground cost function under consideration. Finally, statistical applications, such as bootstrap, are discussed."
                },
                {
                    "title": "A Survey on Evaluation of Large Language Models",
                    "abstract": "Large language models (LLMs) are gaining increasing popularity in both academia and industry, owing to their unprecedented performance in various applications. As LLMs continue to play a vital role in both research and daily use, their evaluation becomes increasingly critical, not only at the task level, but also at the society level for better understanding of their potential risks. Over the past years, significant efforts have been made to examine LLMs from various perspectives. This paper presents a comprehensive review of these evaluation methods for LLMs, focusing on three key dimensions: what to evaluate, where to evaluate, and how to evaluate. Firstly, we provide an overview from the perspective of evaluation tasks, encompassing general natural language processing tasks, reasoning, medical usage, ethics, education, natural and social sciences, agent applications, and other areas. Secondly, we answer the \u2018where\u2019 and \u2018how\u2019 questions by diving into the evaluation methods and benchmarks, which serve as crucial components in assessing the performance of LLMs. Then, we summarize the success and failure cases of LLMs in different tasks. Finally, we shed light on several future challenges that lie ahead in LLMs evaluation. Our aim is to offer invaluable insights to researchers in the realm of LLMs evaluation, thereby aiding the development of more proficient LLMs. Our key point is that evaluation should be treated as an essential discipline to better assist the development of LLMs. We consistently maintain the related open-source materials at: https://github.com/MLGroupJLU/LLM-eval-survey"
                },
                {
                    "title": "Lower Complexity Adaptation for Empirical Entropic Optimal Transport",
                    "abstract": "Entropic optimal transport (EOT) presents an effective and computationally viable alternative to unregularized optimal transport (OT), offering diverse applications for large-scale data analysis. In this work, we derive novel statistical bounds for empirical plug-in estimators of the EOT cost and show that their statistical performance in the entropy regularization parameter $\\epsilon$ and the sample size $n$ only depends on the simpler of the two probability measures. For instance, under sufficiently smooth costs this yields the parametric rate $n^{-1/2}$ with factor $\\epsilon^{-d/2}$, where $d$ is the minimum dimension of the two population measures. This confirms that empirical EOT also adheres to the lower complexity adaptation principle, a hallmark feature only recently identified for unregularized OT. As a consequence of our theory, we show that the empirical entropic Gromov-Wasserstein distance and its unregularized version for measures on Euclidean spaces also obey this principle. Additionally, we comment on computational aspects and complement our findings with Monte Carlo simulations. Our techniques employ empirical process theory and rely on a dual formulation of EOT over a single function class. Crucial to our analysis is the observation that the entropic cost-transformation of a function class does not increase its uniform metric entropy by much."
                },
                {
                    "title": "TrustGPT: A Benchmark for Trustworthy and Responsible Large Language Models",
                    "abstract": "Large Language Models (LLMs) such as ChatGPT, have gained significant attention due to their impressive natural language processing capabilities. It is crucial to prioritize human-centered principles when utilizing these models. Safeguarding the ethical and moral compliance of LLMs is of utmost importance. However, individual ethical issues have not been well studied on the latest LLMs. Therefore, this study aims to address these gaps by introducing a new benchmark -- TrustGPT. TrustGPT provides a comprehensive evaluation of LLMs in three crucial areas: toxicity, bias, and value-alignment. Initially, TrustGPT examines toxicity in language models by employing toxic prompt templates derived from social norms. It then quantifies the extent of bias in models by measuring quantifiable toxicity values across different groups. Lastly, TrustGPT assesses the value of conversation generation models from both active value-alignment and passive value-alignment tasks. Through the implementation of TrustGPT, this research aims to enhance our understanding of the performance of conversation generation models and promote the development of language models that are more ethical and socially responsible."
                },
                {
                    "title": "Minimum intrinsic dimension scaling for entropic optimal transport",
                    "abstract": "Motivated by the manifold hypothesis, which states that data with a high extrinsic dimension may yet have a low intrinsic dimension, we develop refined statistical bounds for entropic optimal transport that are sensitive to the intrinsic dimension of the data. Our bounds involve a robust notion of intrinsic dimension, measured at only a single distance scale depending on the regularization parameter, and show that it is only the minimum of these single-scale intrinsic dimensions which governs the rate of convergence. We call this the Minimum Intrinsic Dimension scaling (MID scaling) phenomenon, and establish MID scaling with no assumptions on the data distributions so long as the cost is bounded and Lipschitz, and for various entropic optimal transport quantities beyond just values, with stronger analogs when one distribution is supported on a manifold. Our results significantly advance the theoretical state of the art by showing that MID scaling is a generic phenomenon, and provide the first rigorous interpretation of the statistical effect of entropic regularization as a distance scale."
                },
                {
                    "title": "LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fusion",
                    "abstract": "We present LLM-Blender, an ensembling framework designed to attain consistently superior performance by leveraging the diverse strengths of multiple open-source large language models (LLMs). Our framework consists of two modules: PairRanker and GenFuser, addressing the observation that optimal LLMs for different examples can significantly vary. PairRanker employs a specialized pairwise comparison method to distinguish subtle differences between candidate outputs. It jointly encodes the input text and a pair of candidates, using cross-attention encoders to determine the superior one. Our results demonstrate that PairRanker exhibits the highest correlation with ChatGPT-based ranking. Then, GenFuser aims to merge the top-ranked candidates, generating an improved output by capitalizing on their strengths and mitigating their weaknesses. To facilitate large-scale evaluation, we introduce a benchmark dataset, MixInstruct, which is a mixture of multiple instruction datasets featuring oracle pairwise comparisons. Our LLM-Blender significantly outperform individual LLMs and baseline methods across various metrics, establishing a substantial performance gap."
                },
                {
                    "title": "Holistic Evaluation of Language Models",
                    "abstract": "Language models (LMs) like GPT\u20103, PaLM, and ChatGPT are the foundation for almost all major language technologies, but their capabilities, limitations, and risks are not well understood. We present Holistic Evaluation of Language Models (HELM) to improve the transparency of LMs. LMs can serve many purposes and their behavior should satisfy many desiderata. To navigate the vast space of potential scenarios and metrics, we taxonomize the space and select representative subsets. We evaluate models on 16 core scenarios and 7 metrics, exposing important trade\u2010offs. We supplement our core evaluation with seven targeted evaluations to deeply analyze specific aspects (including world knowledge, reasoning, regurgitation of copyrighted content, and generation of disinformation). We benchmark 30 LMs, from OpenAI, Microsoft, Google, Meta, Cohere, AI21 Labs, and others. Prior to HELM, models were evaluated on just 17.9% of the core HELM scenarios, with some prominent models not sharing a single scenario in common. We improve this to 96.0%: all 30 models are now benchmarked under the same standardized conditions. Our evaluation surfaces 25 top\u2010level findings. For full transparency, we release all raw model prompts and completions publicly. HELM is a living benchmark for the community, continuously updated with new scenarios, metrics, and models https://crfm.stanford.edu/helm/latest/."
                },
                {
                    "title": "Convergence Rates for Regularized Optimal Transport via Quantization",
                    "abstract": "We study the convergence of divergence-regularized optimal transport as the regularization parameter vanishes. Sharp rates for general divergences including relative entropy or Lp regularization, general transport costs, and multimarginal problems are obtained. A novel methodology using quantization and martingale couplings is suitable for noncompact marginals and achieves, in particular, the sharp leading-order term of entropically regularized 2-Wasserstein distance for marginals with a finite [Formula: see text]-moment. Funding: This work was supported by the Alfred P. Sloan Foundation [Grant FG-2016-6282] and the Division of Mathematical Sciences [Grants DMS-1812661 and DMS-2106056]."
                },
                {
                    "title": "Limit Theorems for Entropic Optimal Transport Maps and the Sinkhorn Divergence",
                    "abstract": "We study limit theorems for entropic optimal transport (EOT) maps, dual potentials, and the Sinkhorn divergence. The key technical tool we use is a first and second-order Hadamard differentiability analysis of EOT potentials with respect to the marginal distributions, which may be of independent interest. Given the differentiability results, the functional delta method is used to obtain central limit theorems for empirical EOT potentials and maps. The second-order functional delta method is leveraged to establish the limit distribution of the empirical Sinkhorn divergence under the null. Building on the latter result, we further derive the null limit distribution of the Sinkhorn independence test statistic and characterize the correct order. Since our limit theorems follow from Hadamard differentiability of the relevant maps, as a byproduct, we also obtain bootstrap consistency and asymptotic efficiency of the empirical EOT map, potentials, and Sinkhorn divergence."
                },
                {
                    "title": "Statistical inference with regularized optimal transport",
                    "abstract": "\n Optimal transport (OT) is a versatile framework for comparing probability measures, with many applications to statistics, machine learning and applied mathematics. However, OT distances suffer from computational and statistical scalability issues to high dimensions, which motivated the study of regularized OT methods like slicing, smoothing and entropic penalty. This work establishes a unified framework for deriving limit distributions of empirical regularized OT distances, semiparametric efficiency of the plug-in empirical estimator and bootstrap consistency. We apply the unified framework to provide a comprehensive statistical treatment of (i) average- and max-sliced $p$-Wasserstein distances, for which several gaps in existing literature are closed; (ii) smooth distances with compactly supported kernels, the analysis of which is motivated by computational considerations; and (iii) entropic OT, for which our method generalizes existing limit distribution results and establishes, for the first time, efficiency and bootstrap consistency. While our focus is on these three regularized OT distances as applications, the flexibility of the proposed framework renders it applicable to broad classes of functionals beyond these examples."
                },
                {
                    "title": "An improved central limit theorem and fast convergence rates for entropic transportation costs",
                    "abstract": "We prove a central limit theorem for the entropic transportation cost between subgaussian probability measures, centered at the population cost. This is the first result which allows for asymptotically valid inference for entropic optimal transport between measures which are not necessarily discrete. In the compactly supported case, we complement these results with new, faster, convergence rates for the expected entropic transportation cost between empirical measures. Our proof is based on strengthening convergence results for dual solutions to the entropic optimal transport problem."
                },
                {
                    "title": "deep-significance - Easy and Meaningful Statistical Significance Testing in the Age of Neural Networks",
                    "abstract": "A lot of Machine Learning (ML) and Deep Learning (DL) research is of an empirical nature. Nevertheless, statistical significance testing (SST) is still not widely used. This endangers true progress, as seeming improvements over a baseline might be statistical flukes, leading follow-up research astray while wasting human and computational resources. Here, we provide an easy-to-use package containing different significance tests and utility functions specifically tailored towards research needs and usability."
                },
                {
                    "title": "Empirical optimal transport between different measures adapts to lower complexity",
                    "abstract": "The empirical optimal transport (OT) cost between two probability measures from random data is a fundamental quantity in transport based data analysis. In this work, we derive novel guarantees for its convergence rate when the involved measures are different, possibly supported on different spaces. Our central observation is that the statistical performance of the empirical OT cost is determined by the less complex measure, a phenomenon we refer to as lower complexity adaptation of empirical OT. For instance, under Lipschitz ground costs, we find that the empirical OT cost based on $n$ observations converges at least with rate $n^{-1/d}$ to the population quantity if one of the two measures is concentrated on a $d$-dimensional manifold, while the other can be arbitrary. For semi-concave ground costs, we show that the upper bound for the rate improves to $n^{-2/d}$. Similarly, our theory establishes the general convergence rate $n^{-1/2}$ for semi-discrete OT. All of these results are valid in the two-sample case as well, meaning that the convergence rate is still governed by the simpler of the two measures. On a conceptual level, our findings therefore suggest that the curse of dimensionality only affects the estimation of the OT cost when both measures exhibit a high intrinsic dimension. Our proofs are based on the dual formulation of OT as a maximization over a suitable function class $\\mathcal{F}_c$ and the observation that the $c$-transform of $\\mathcal{F}_c$ under bounded costs has the same uniform metric entropy as $\\mathcal{F}_c$ itself."
                },
                {
                    "title": "Sharp convergence rates for empirical optimal transport with smooth costs",
                    "abstract": "We revisit the question of characterizing the convergence rate of plug-in estimators of optimal transport costs. It is well known that an empirical measure comprising independent samples from an absolutely continuous distribution on $\\mathbb{R}^d$ converges to that distribution at the rate $n^{-1/d}$ in Wasserstein distance, which can be used to prove that plug-in estimators of many optimal transport costs converge at this same rate. However, we show that when the cost is smooth, this analysis is loose: plug-in estimators based on empirical measures converge quadratically faster, at the rate $n^{-2/d}$. As a corollary, we show that the Wasserstein distance between two distributions is significantly easier to estimate when the measures are well-separated. We also prove lower bounds, showing not only that our analysis of the plug-in estimator is tight, but also that no other estimator can enjoy significantly faster rates of convergence uniformly over all pairs of measures. Our proofs rely on empirical process theory arguments based on tight control of $L^2$ covering numbers for locally Lipschitz and semi-concave functions. As a byproduct of our proofs, we derive $L^\\infty$ estimates on the displacement induced by the optimal coupling between any two measures satisfying suitable concentration and anticoncentration conditions, for a wide range of cost functions."
                },
                {
                    "title": "Smooth p-Wasserstein Distance: Structure, Empirical Approximation, and Statistical Applications",
                    "abstract": "Discrepancy measures between probability distributions, often termed statistical distances, are ubiquitous in probability theory, statistics and machine learning. To combat the curse of dimensionality when estimating these distances from data, recent work has proposed smoothing out local irregularities in the measured distributions via convolution with a Gaussian kernel. Motivated by the scalability of this framework to high dimensions, we investigate the structural and statistical behavior of the Gaussian-smoothed p-Wasserstein distance W p , for arbitrary p \u2265 1. After establishing basic metric and topological properties of W (\u03c3) p , we explore the asymptotic statistical behavior of W p (\u03bc\u0302n, \u03bc), where \u03bc\u0302n is the empirical distribution of n independent observations from \u03bc. We prove that W p enjoys a parametric empirical convergence rate of n\u22121/2, which contrasts the n\u22121/d rate for unsmoothed Wp when d \u2265 3. Our proof relies on controlling W p by a pth-order smooth Sobolev distance d p and deriving the limit distribution of \u221a n d (\u03c3) p (\u03bc\u0302n, \u03bc), for all dimensions d. As applications, we provide asymptotic guarantees for two-sample testing and minimum distance estimation using W p , with experiments for p = 2 using a maximum mean discrepancy formulation of d 2 ."
                },
                {
                    "title": "Statistical bounds for entropic optimal transport: sample complexity and the central limit theorem",
                    "abstract": "We prove several fundamental statistical bounds for entropic OT with the squared Euclidean cost between subgaussian probability measures in arbitrary dimension. First, through a new sample complexity result we establish the rate of convergence of entropic OT for empirical measures. Our analysis improves exponentially on the bound of Genevay et al. (2019) and extends their work to unbounded measures. Second, we establish a central limit theorem for entropic OT, based on techniques developed by Del Barrio and Loubes (2019). Previously, such a result was only known for finite metric spaces. As an application of our results, we develop and analyze a new technique for estimating the entropy of a random variable corrupted by gaussian noise."
                },
                {
                    "title": "Sample Complexity of Sinkhorn Divergences",
                    "abstract": "Optimal transport (OT) and maximum mean discrepancies (MMD) are now routinely used in machine learning to compare probability measures. We focus in this paper on \\emph{Sinkhorn divergences} (SDs), a regularized variant of OT distances which can interpolate, depending on the regularization strength $\\varepsilon$, between OT ($\\varepsilon=0$) and MMD ($\\varepsilon=\\infty$). Although the tradeoff induced by that regularization is now well understood computationally (OT, SDs and MMD require respectively $O(n^3\\log n)$, $O(n^2)$ and $n^2$ operations given a sample size $n$), much less is known in terms of their \\emph{sample complexity}, namely the gap between these quantities, when evaluated using finite samples \\emph{vs.} their respective densities. Indeed, while the sample complexity of OT and MMD stand at two extremes, $1/n^{1/d}$ for OT in dimension $d$ and $1/\\sqrt{n}$ for MMD, that for SDs has only been studied empirically. In this paper, we \\emph{(i)} derive a bound on the approximation error made with SDs when approximating OT as a function of the regularizer $\\varepsilon$, \\emph{(ii)} prove that the optimizers of regularized OT are bounded in a Sobolev (RKHS) ball independent of the two measures and \\emph{(iii)} provide the first sample complexity bound for SDs, obtained,by reformulating SDs as a maximization problem in a RKHS. We thus obtain a scaling in $1/\\sqrt{n}$ (as in MMD), with a constant that depends however on $\\varepsilon$, making the bridge between OT and MMD complete."
                },
                {
                    "title": "The Hitchhiker\u2019s Guide to Testing Statistical Significance in Natural Language Processing",
                    "abstract": "Statistical significance testing is a standard statistical tool designed to ensure that experimental results are not coincidental. In this opinion/ theoretical paper we discuss the role of statistical significance testing in Natural Language Processing (NLP) research. We establish the fundamental concepts of significance testing and discuss the specific aspects of NLP tasks, experimental setups and evaluation measures that affect the choice of significance tests in NLP research. Based on this discussion we propose a simple practical protocol for statistical significance test selection in NLP setups and accompany this protocol with a brief survey of the most relevant tests. We then survey recent empirical papers published in ACL and TACL during 2017 and show that while our community assigns great value to experimental results, statistical significance testing is often ignored or misused. We conclude with a brief discussion of open issues that should be properly addressed so that this important tool can be applied. in NLP research in a statistically sound manner."
                },
                {
                    "title": "Optimal Transport for Applied Mathematicians: Calculus of Variations, PDEs, and Modeling",
                    "abstract": "Preface.- Primal and Dual Problems.- One-Dimensional Issues.- L^1 and L^infinity Theory.- Minimal Flows.- Wasserstein Spaces.- Numerical Methods.- Functionals over Probabilities.- Gradient Flows.- Exercises.- References.- Index."
                },
                {
                    "title": "Sinkhorn Distances: Lightspeed Computation of Optimal Transport",
                    "abstract": "Optimal transportation distances are a fundamental family of parameterized distances for histograms. Despite their appealing theoretical properties, excellent performance in retrieval tasks and intuitive formulation, their computation involves the resolution of a linear program whose cost is prohibitive whenever the histograms' dimension exceeds a few hundreds. We propose in this work a new family of optimal transportation distances that look at transportation problems from a maximum-entropy perspective. We smooth the classical optimal transportation problem with an entropic regularization term, and show that the resulting optimum is also a distance which can be computed through Sinkhorn-Knopp's matrix scaling algorithm at a speed that is several orders of magnitude faster than that of transportation solvers. We also report improved performance over classical optimal transportation distances on the MNIST benchmark problem."
                },
                {
                    "title": "Stochastic Orders",
                    "abstract": "have seen rapid development of SEM methodology dealing with, for example, nonlinear terms, multiple levels, mixtures, and so on. This book presents the state-of-the-art SEMs from mainly a Bayesian viewpoint. The author aims to introduce a Bayesian approach for developing efficient and rigorous statistical methodologies in SEMs and applying them to practical problems. The book is intended to serve as a reference for researchers and a textbook for graduate students in several disciplines. The book has several attractive features. The presentation is lucid and clear, the discussion of various topics in SEM is thorough and up to date, and an expert interweaving of theory and examples in each topic make the book a pleasure to read and learn from. The variety of topics and the depth to which they are covered and referenced makes this an extremely useful resource for research in any topic relating to SEMs. As the author remarks, \u201cit is very common in practice to encounter ordered categorical data with missing data, hierarchical data and/or heterogeneous data, and hence developments of sound statistical methods to cope with such practical situations are useful.\u201d It is clear throughout the book that \u201csound statistical methods\u201d have been topmost in the author\u2019s mind. The standard SEMs based on normality assumption and computer software implementing such SEMs have limited capabilities and are inadequate for analyzing complex data sets. Thus the Bayesian development presented in this book should be attractive to any serious researcher using SEMs. After the introductory chapter, Chapter 2 presents some basic SEMs and provides a gentle introduction to the subject matter. Chapter 3 is an off-beat chapter that provides rigorous technical detail about some non-Bayesian methods of SEM. This is a welcome development, because necessary rigor often is sidestepped in application-oriented texts. Chapters 4 and 5 introduce the reader to the Bayesian viewpoint of SEMs. The remaining chapters discuss different frameworks for SEM and their analysis. This book is a welcome addition to any library and should be a valuable resource for research and teaching."
                },
                {
                    "title": "Delta Method, Infinite Dimensional",
                    "abstract": "The infinite-dimensional delta method including higher order and set-valued variants is presented. The relevant underlying material on weak convergence, Hadamard directional derivatives and on hyperspaces of closed sets is discussed. Some examples of Hadamard directional differentiable functions and of the delta method, including M-estimation, are given. \n \n \nKeywords: \n \nstatistical functional; \nfirst order asymptotics; \nhadamard directional derivative; \nweak convergence; \ncentral limit theorem; \ndelta method; \nasymptotic normality; \nnon-normality; \nM-estimation; \nZ-estimation"
                },
                {
                    "title": "Dual Stochastic Dominance and Related Mean-Risk Models",
                    "abstract": "We consider the problem of constructing mean-risk models which are consistent with the second degree stochastic dominance relation. By exploiting duality relations of convex analysis we develop the quantile model of stochastic dominance for general distributions. This allows us to show that several models using quantiles and tail characteristics of the distribution are in harmony with the stochastic dominance relation. We also provide stochastic linear programming formulations of these models."
                },
                {
                    "title": "POT: Python Optimal Transport",
                    "abstract": "Optimal transport has recently been reintroduced to the machine learning community thanks in part to novel e\ufb03cient optimization procedures allowing for medium to large scale applications. We propose a Python toolbox that implements several key optimal transport ideas for the machine learning community. The toolbox contains implementations of a number of founding works of OT for machine learning such as Sinkhorn algorithm and Wasserstein barycenters, but also provides generic solvers that can be used for conducting novel fundamental research. This toolbox, named POT for Python Optimal Transport, is open source with an MIT license."
                },
                {
                    "title": "Introduction to Entropic Optimal Transport",
                    "abstract": "This text develops mathematical foundations for entropic optimal transport and Sinkhorn\u2019s algorithm in a self-contained yet general way. It is a revised version of lecture notes from a course given in Paris during the fall of 2021; some parts date back to an earlier course at Columbia University in 2020."
                },
                {
                    "title": "On Concepts of Directional Differentiability",
                    "abstract": "Various definitions of directional derivatives in topological vector spaces are compared. Directional derivatives in the sense of G~teaux, Fr6chet, and Hadamard are singled out from the general framework of cr-directional differentiability. It is pointed out that, in the case of finite-dimensional spaces and locally Lipschitz mappings, all these concepts of directional differentiability are equivalent. The chain rule for directional derivatives of a composite mapping is discussed."
                },
                {
                    "title": "A Multivariate Faa di Bruno Formula with Applications",
                    "abstract": "A multivariate Faa di Bruno formula for computing arbitrary partial derivatives of a function composition is presented. It is shown, by way of a general identity, how such derivatives can also be expressed in the form of an infinite series. Applications to stochastic processes and multivariate cumulants are then delineated."
                },
                {
                    "title": "A Simple Sequentially Rejective Multiple Test Procedure",
                    "abstract": "This paper presents a simple and widely ap- plicable multiple test procedure of the sequentially rejective type, i.e. hypotheses are rejected one at a tine until no further rejections can be done. It is shown that the test has a prescribed level of significance protection against error of the first kind for any combination of true hypotheses. The power properties of the test and a number of possible applications are also discussed."
                },
                {
                    "title": "Review: P. Billingsley, Convergence of probability measures",
                    "abstract": "The author's preface gives an outline: \"This book is about weakconvergence methods in metric spaces, with applications sufficient to show their power and utility. The Introduction motivates the definitions and indicates how the theory will yield solutions to problems arising outside it. Chapter 1 sets out the basic general theorems, which are then specialized in Chapter 2 to the space C[0, l ] of continuous functions on the unit interval and in Chapter 3 to the space D [0, 1 ] of functions with discontinuities of the first kind. The results of the first three chapters are used in Chapter 4 to derive a variety of limit theorems for dependent sequences of random variables. \" The book develops and expands on Donsker's 1951 and 1952 papers on the invariance principle and empirical distributions. The basic random variables remain real-valued although, of course, measures on C[0, l ] and D[0, l ] are vitally used. Within this framework, there are various possibilities for a different and apparently better treatment of the material. More of the general theory of weak convergence of probabilities on separable metric spaces would be useful. Metrizability of the convergence is not brought up until late in the Appendix. The close relation of the Prokhorov metric and a metric for convergence in probability is (hence) not mentioned (see V. Strassen, Ann. Math. Statist. 36 (1965), 423-439; the reviewer, ibid. 39 (1968), 1563-1572). This relation would illuminate and organize such results as Theorems 4.1, 4.2 and 4.4 which give isolated, ad hoc connections between weak convergence of measures and nearness in probability. In the middle of p. 16, it should be noted that C*(S) consists of signed measures which need only be finitely additive if 5 is not compact. On p. 239, where the author twice speaks of separable subsets having nonmeasurable cardinal, he means \"discrete\" rather than \"separable.\" Theorem 1.4 is Ulam's theorem that a Borel probability on a complete separable metric space is tight. Theorem 1 of Appendix 3 weakens completeness to topological completeness. After mentioning that probabilities on the rationals are tight, the author says it is an"
                },
                {
                    "title": "The identity of weak and strong extensions of differential operators",
                    "abstract": "In applying the theory of linear operators in Hubert spaces or spaces $p to the solution of differential equation problems, it is impossible to retain the. meaning of differentiation in the ordinary sense; the concept of differential operator must be extended. Two such extensions offer themselves, a \"weak\" and a \"strong\" one. Existence theorems, when derived by variational methods, result most directly in terms of the weak extensions. It is the strong extension, however, which offers the natural approach to establishing properties of the solution; in particular those that lead to differentiability in the ordinary sense. The fact that both extensions are identical is therefore decisive. The objective of this paper is to prove the identity of weak and strong extension for general linear differential operators. The main tool for the proof is a certain class of smoothing operators approximating unity, the \"mollifiers.\" These mollifiers yield the identity of both extensions immediately for differential operators with constant coefficients; it is remarkable that they are a strong enough tool to yield this identity likewise for operators with non-constant coefficients. While the present paper is not concerned with the application to existence problems^), other miscellaneous applications and generalizations of the identity will be discussed. In the main part of the paper (\u00a7\u00a71-3) the identity of weak and strong extensions is proved with reference to the function space 2P. Generalization to function spaces enjoying a certain translation property is possible (\u00a75). Under certain restrictions weak and strong extensions can be expressed in terms of adjointness (\u00a74). The miscellaneous applications (\u00a76) refer in particular, to underdetermined systems and to Haar's lemma. 1. Extended differential operators. Let x = (xu \u25a0 \u25a0 \u25a0 , xm) be a point in an w-dimensional Euclidean space. Let R be an open region in that space. Let u= {uc}, a\u2014 1, \u2022 \u2022 \u2022 , s, be a system of functions defined in R; u = u(x) will in general simply be called a function. Let Am n=l, \u25a0 \u25a0 \u2022 , m, B be matrices transforming systems \u00ab of j functions into systems A\u201eu, Bu of t functions."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG",
                "math.ST",
                "stat.TH"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we extend the testing framework for almost First Order Stochastic Dominance (FSD) to the multivariate case to account for dependencies between metrics in applications such as multivariate portfolio selection and multi-metric benchmarking of Large Language Models?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of current methodologies that reduce multivariate comparisons to univariate orders, thereby ignoring important dependencies between metrics. By developing a robust framework for almost multivariate FSD, we can enhance the statistical analysis of complex systems, leading to more informed decision-making in fields like finance and machine learning. This advancement could pave the way for more accurate model selection and performance evaluation, ultimately influencing future research directions and practical applications in various domains.\n\n**[Question 3] - Why is it hard?**  \nThe challenge in solving this problem lies in the complexity of multivariate stochastic orders, which require sophisticated mathematical tools to capture dependencies between multiple random variables. Naive approaches that treat metrics independently may overlook critical interactions, leading to misleading conclusions. Additionally, empirical optimal transport methods face the curse of dimensionality, making it difficult to accurately estimate multivariate dominance. Overcoming these technical obstacles requires innovative methodologies, such as entropic regularization, to ensure convergence and statistical consistency.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on univariate stochastic dominance, with limited exploration of multivariate frameworks. Existing solutions often rely on aggregation techniques that simplify the problem but fail to account for the intricate relationships between metrics. Barriers to progress include a lack of established theoretical foundations for multivariate dominance and the computational challenges associated with high-dimensional data. Our approach differs by leveraging optimal transport principles and entropic regularization to create a more comprehensive and statistically sound framework for almost multivariate FSD.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves defining almost multivariate FSD through optimal transport, utilizing multivariate violation ratios expressed as optimal transport problems with smooth costs. We will employ entropic regularization to mitigate the curse of dimensionality, allowing for the definition of Entropic Multivariate Violation Ratios. The dataset will consist of multivariate performance metrics from Large Language Models, and we will use statistical metrics such as convergence rates and central limit theorems to evaluate our approach. Expected outcomes include a robust framework for"
            }
        },
        "author_data": {
            "a094eedf-487b-49f9-9ec8-cc80cb67aa18": {
                "pk": "a094eedf-487b-49f9-9ec8-cc80cb67aa18",
                "name": "Gabriel Rioux",
                "collaborators": [
                    "Ziv Goldfeld",
                    "Kengo Kato",
                    "Ritwik Sadhu",
                    "Rustum Choksi",
                    "Tim Hoheisel",
                    "Pierre Marechal",
                    "Christopher Scarvelis",
                    "Sloan Nietert"
                ],
                "domain": [
                    "Optimal Transport",
                    "Statistical Inference",
                    "Wasserstein Distance",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Limit Laws for Gromov-Wasserstein Alignment with Applications to Testing Graph Isomorphisms",
                        "abstract": "The Gromov-Wasserstein (GW) distance enables comparing metric measure spaces based solely on their internal structure, making it invariant to isomorphic transformations. This property is particularly useful for comparing datasets that naturally admit isomorphic representations, such as unlabelled graphs or objects embedded in space. However, apart from the recently derived empirical convergence rates for the quadratic GW problem, a statistical theory for valid estimation and inference remains largely obscure. Pushing the frontier of statistical GW further, this work derives the first limit laws for the empirical GW distance across several settings of interest: (i)~discrete, (ii)~semi-discrete, and (iii)~general distributions under moment constraints under the entropically regularized GW distance. The derivations rely on a novel stability analysis of the GW functional in the marginal distributions. The limit laws then follow by an adaptation of the functional delta method. As asymptotic normality fails to hold in most cases, we establish the consistency of an efficient estimation procedure for the limiting law in the discrete case, bypassing the need for computationally intensive resampling methods. We apply these findings to testing whether collections of unlabelled graphs are generated from distributions that are isomorphic to each other."
                    },
                    {
                        "title": "Entropic Gromov-Wasserstein Distances: Stability and Algorithms",
                        "abstract": "The Gromov-Wasserstein (GW) distance quantifies discrepancy between metric measure spaces and provides a natural framework for aligning heterogeneous datasets. Alas, as exact computation of GW alignment is NP hard, entropic regularization provides an avenue towards a computationally tractable proxy. Leveraging a recently derived variational representation for the quadratic entropic GW (EGW) distance, this work derives the first efficient algorithms for solving the EGW problem subject to formal, non-asymptotic convergence guarantees. To that end, we derive smoothness and convexity properties of the objective in this variational problem, which enables its resolution by the accelerated gradient method. Our algorithms employs Sinkhorn's fixed point iterations to compute an approximate gradient, which we model as an inexact oracle. We furnish convergence rates towards local and even global solutions (the latter holds under a precise quantitative condition on the regularization parameter), characterize the effects of gradient inexactness, and prove that stationary points of the EGW problem converge towards a stationary point of the unregularized GW problem, in the limit of vanishing regularization. We provide numerical experiments that validate our theory and empirically demonstrate the state-of-the-art empirical performance of our algorithm."
                    },
                    {
                        "title": "The Maximum Entropy on the Mean Method for Image Deblurring",
                        "abstract": "Image deblurring is a notoriously challenging ill-posed inverse problem. In recent years, a wide variety of approaches have been proposed based upon regularization at the level of the image or on techniques from machine learning. We propose an alternative approach, shifting the paradigm towards regularization at the level of the probability distribution on the space of images. Our method is based upon the idea of maximum entropy on the mean wherein we work at the level of the probability density function of the image whose expectation is our estimate of the ground truth. Using techniques from convex analysis and probability theory, we show that the method is computationally feasible and amenable to very large blurs. Moreover, when images are imbedded with symbology (a known pattern), we show how our method can be applied to approximate the unknown blur kernel with remarkable effects. While our method is stable with respect to small amounts of noise, it does not actively denoise. However, for moderate to large amounts of noise, it performs well by preconditioned denoising with a state of the art method."
                    },
                    {
                        "title": "Limit distribution theory for smooth $p$-Wasserstein distances",
                        "abstract": "The Wasserstein distance is a metric on a space of probability measures that has seen a surge of applications in statistics, machine learning, and applied mathematics. However, statistical aspects of Wasserstein distances are bottlenecked by the curse of dimensionality, whereby the number of data points needed to accurately estimate them grows exponentially with dimension. Gaussian smoothing was recently introduced as a means to alleviate the curse of dimensionality, giving rise to a parametric convergence rate in any dimension, while preserving the Wasserstein metric and topological structure. To facilitate valid statistical inference, in this work, we develop a comprehensive limit distribution theory for the empirical smooth Wasserstein distance. The limit distribution results leverage the functional delta method after embedding the domain of the Wasserstein distance into a certain dual Sobolev space, characterizing its Hadamard directional derivative for the dual Sobolev norm, and establishing weak convergence of the smooth empirical process in the dual space. To estimate the distributional limits, we also establish consistency of the nonparametric bootstrap. Finally, we use the limit distribution theory to study applications to generative modeling via minimum distance estimation with the smooth Wasserstein distance, showing asymptotic normality of optimal solutions for the quadratic cost."
                    },
                    {
                        "title": "Statistical inference with regularized optimal transport",
                        "abstract": "Optimal transport (OT) is a versatile framework for comparing probability measures, with many applications to statistics, machine learning, and applied mathematics. However, OT distances suffer from computational and statistical scalability issues to high dimensions, which motivated the study of regularized OT methods like slicing, smoothing, and entropic penalty. This work establishes a unified framework for deriving limit distributions of empirical regularized OT distances, semiparametric efficiency of the plug-in empirical estimator, and bootstrap consistency. We apply the unified framework to provide a comprehensive statistical treatment of: (i) average- and max-sliced $p$-Wasserstein distances, for which several gaps in existing literature are closed; (ii) smooth distances with compactly supported kernels, the analysis of which is motivated by computational considerations; and (iii) entropic OT, for which our method generalizes existing limit distribution results and establishes, for the first time, efficiency and bootstrap consistency. While our focus is on these three regularized OT distances as applications, the flexibility of the proposed framework renders it applicable to broad classes of functionals beyond these examples."
                    },
                    {
                        "title": "Limit Theorems for Entropic Optimal Transport Maps and the Sinkhorn Divergence",
                        "abstract": "We study limit theorems for entropic optimal transport (EOT) maps, dual potentials, and the Sinkhorn divergence. The key technical tool we use is a first and second-order Hadamard differentiability analysis of EOT potentials with respect to the marginal distributions, which may be of independent interest. Given the differentiability results, the functional delta method is used to obtain central limit theorems for empirical EOT potentials and maps. The second-order functional delta method is leveraged to establish the limit distribution of the empirical Sinkhorn divergence under the null. Building on the latter result, we further derive the null limit distribution of the Sinkhorn independence test statistic and characterize the correct order. Since our limit theorems follow from Hadamard differentiability of the relevant maps, as a byproduct, we also obtain bootstrap consistency and asymptotic efficiency of the empirical EOT map, potentials, and Sinkhorn divergence."
                    }
                ]
            },
            "4a4c7910-e7f2-4983-ad68-9670d41766ec": {
                "pk": "4a4c7910-e7f2-4983-ad68-9670d41766ec",
                "name": "Apoorva Nitsure",
                "collaborators": [
                    "Rajesh Bordawekar",
                    "Jose Neves",
                    "Prabhakar Kudva",
                    "Prashanth Vijayaraghavan",
                    "Luyao Shi",
                    "Stefano Ambrogio",
                    "Charles Mackin",
                    "David Beymer",
                    "Ehsan Degan"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Database Systems",
                    "Code Generation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Unlocking New York City Crime Insights using Relational Database Embeddings",
                        "abstract": "This version withdrawn by arXiv administrators because the author did not have the right to agree to our license at the time of submission."
                    },
                    {
                        "title": "A Scalable Space-efficient In-database Interpretability Framework for Embedding-based Semantic SQL Queries",
                        "abstract": "AI-Powered database (AI-DB) is a novel relational database system that uses a self-supervised neural network, database embedding, to enable semantic SQL queries on relational tables. In this paper, we describe an architecture and implementation of in-database interpretability infrastructure designed to provide simple, transparent, and relatable insights into ranked results of semantic SQL queries supported by AI-DB. We introduce a new co-occurrence based interpretability approach to capture relationships between relational entities and describe a space-efficient probabilistic Sketch implementation to store and process co-occurrence counts. Our approach provides both query-agnostic (global) and query-specific (local) interpretabilities. Experimental evaluation demonstrate that our in-database probabilistic approach provides the same interpretability quality as the precise space-inefficient approach, while providing scalable and space efficient runtime behavior (up to 8X space savings), without any user intervention."
                    },
                    {
                        "title": "VHDL-Eval: A Framework for Evaluating Large Language Models in VHDL Code Generation",
                        "abstract": "With the unprecedented advancements in Large Language Models (LLMs), their application domains have expanded to include code generation tasks across various programming languages. While significant progress has been made in enhancing LLMs for popular programming languages, there exists a notable gap in comprehensive evaluation frameworks tailored for Hardware Description Languages (HDLs), particularly VHDL. This paper addresses this gap by introducing a comprehensive evaluation framework designed specifically for assessing LLM performance in VHDL code generation task. We construct a dataset for evaluating LLMs on VHDL code generation task. This dataset is constructed by translating a collection of Verilog evaluation problems to VHDL and aggregating publicly available VHDL problems, resulting in a total of 202 problems. To assess the functional correctness of the generated VHDL code, we utilize a curated set of self-verifying testbenches specifically designed for those aggregated VHDL problem set. We conduct an initial evaluation of different LLMs and their variants, including zero-shot code generation, in-context learning (ICL), and Parameter-efficient fine-tuning (PEFT) methods. Our findings underscore the considerable challenges faced by existing LLMs in VHDL code generation, revealing significant scope for improvement. This study emphasizes the necessity of supervised fine-tuning code generation models specifically for VHDL, offering potential benefits to VHDL designers seeking efficient code generation solutions."
                    }
                ]
            },
            "80d18833-853c-4724-bb4d-da06bc1f8d92": {
                "pk": "80d18833-853c-4724-bb4d-da06bc1f8d92",
                "name": "Mattia Rigotti",
                "collaborators": [
                    "Youssef Mroueh",
                    "Inkit Padhi",
                    "Pierre Dognin",
                    "Igor Melnyk",
                    "Yair Schiff",
                    "Cristiano Malossi",
                    "Jarret Ross",
                    "Stefano Fusi",
                    "Rong Zhu",
                    "Ioana Giurgiu"
                ],
                "domain": [
                    "Neuroscience",
                    "Reinforcement Learning",
                    "Deep Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Estimating the dimensionality of neural responses with fMRI Repetition Suppression",
                        "abstract": "We propose a novel method that exploits fMRI Repetition Suppression (RS-fMRI) to measure the dimensionality of the set response vectors, i.e. the dimension of the space of linear combinations of neural population activity patterns in response to specific task conditions. RS-fMRI measures the overlap between response vectors even in brain areas displaying no discernible average differential BOLD signal. We show how this property can be used to estimate the neural response dimensionality in areas lacking macroscopic spatial patterning. The importance of dimensionality derives from how it relates to a neural circuit's functionality. As we show, the dimensionality of the response vectors is predicted to be high in areas involved in multi-stream integration, while it is low in areas where inputs from independent sources do not interact or merely overlap linearly. Our method can be used to identify and functionally characterize cortical circuits that integrate multiple independent information pathways."
                    },
                    {
                        "title": "Self-correcting Q-Learning",
                        "abstract": "The Q-learning algorithm is known to be affected by the maximization bias, i.e. the systematic overestimation of action values, an important issue that has recently received renewed attention. Double Q-learning has been proposed as an efficient algorithm to mitigate this bias. However, this comes at the price of an underestimation of action values, in addition to increased memory requirements and a slower convergence. In this paper, we introduce a new way to address the maximization bias in the form of a \"self-correcting algorithm\" for approximating the maximum of an expected value. Our method balances the overestimation of the single estimator used in conventional Q-learning and the underestimation of the double estimator used in Double Q-learning. Applying this strategy to Q-learning results in Self-correcting Q-learning. We show theoretically that this new algorithm enjoys the same convergence guarantees as Q-learning while being more accurate. Empirically, it performs better than Double Q-learning in domains with rewards of high variance, and it even attains faster convergence than Q-learning in domains with rewards of zero or low variance. These advantages transfer to a Deep Q Network implementation that we call Self-correcting DQN and which outperforms regular DQN and Double DQN on several tasks in the Atari 2600 domain."
                    },
                    {
                        "title": "Deep Bandits Show-Off: Simple and Efficient Exploration with Deep Networks",
                        "abstract": "Designing efficient exploration is central to Reinforcement Learning due to the fundamental problem posed by the exploration-exploitation dilemma. Bayesian exploration strategies like Thompson Sampling resolve this trade-off in a principled way by modeling and updating the distribution of the parameters of the action-value function, the outcome model of the environment. However, this technique becomes infeasible for complex environments due to the computational intractability of maintaining probability distributions over parameters of outcome models of corresponding complexity. Moreover, the approximation techniques introduced to mitigate this issue typically result in poor exploration-exploitation trade-offs, as observed in the case of deep neural network models with approximate posterior methods that have been shown to underperform in the deep bandit scenario. In this paper we introduce Sample Average Uncertainty (SAU), a simple and efficient uncertainty measure for contextual bandits. While Bayesian approaches like Thompson Sampling estimate outcomes uncertainty indirectly by first quantifying the variability over the parameters of the outcome model, SAU is a frequentist approach that directly estimates the uncertainty of the outcomes based on the value predictions. Importantly, we show theoretically that the uncertainty measure estimated by SAU asymptotically matches the uncertainty provided by Thompson Sampling, as well as its regret bounds. Because of its simplicity SAU can be seamlessly applied to deep contextual bandits as a very scalable drop-in replacement for epsilon-greedy exploration. We confirm empirically our theory by showing that SAU-based exploration outperforms current state-of-the-art deep Bayesian bandit methods on several real-world datasets at modest computation cost. Code is available at \\url{https://github.com/ibm/sau-explore}."
                    },
                    {
                        "title": "Unbalanced Sobolev Descent",
                        "abstract": "We introduce Unbalanced Sobolev Descent (USD), a particle descent algorithm for transporting a high dimensional source distribution to a target distribution that does not necessarily have the same mass. We define the Sobolev-Fisher discrepancy between distributions and show that it relates to advection-reaction transport equations and the Wasserstein-Fisher-Rao metric between distributions. USD transports particles along gradient flows of the witness function of the Sobolev-Fisher discrepancy (advection step) and reweighs the mass of particles with respect to this witness function (reaction step). The reaction step can be thought of as a birth-death process of the particles with rate of growth proportional to the witness function. When the Sobolev-Fisher witness function is estimated in a Reproducing Kernel Hilbert Space (RKHS), under mild assumptions we show that USD converges asymptotically (in the limit of infinite particles) to the target distribution in the Maximum Mean Discrepancy (MMD) sense. We then give two methods to estimate the Sobolev-Fisher witness with neural networks, resulting in two Neural USD algorithms. The first one implements the reaction step with mirror descent on the weights, while the second implements it through a birth-death process of particles. We show on synthetic examples that USD transports distributions with or without conservation of mass faster than previous particle descent algorithms, and finally demonstrate its use for molecular biology analyses where our method is naturally suited to match developmental stages of populations of differentiating cells based on their single-cell RNA sequencing profile. Code is available at https://github.com/ibm/usd ."
                    },
                    {
                        "title": "Model-Assisted Labeling via Explainability for Visual Inspection of Civil Infrastructures",
                        "abstract": "Labeling images for visual segmentation is a time-consuming task which can be costly, particularly in application domains where labels have to be provided by specialized expert annotators, such as civil engineering. In this paper, we propose to use attribution methods to harness the valuable interactions between expert annotators and the data to be annotated in the case of defect segmentation for visual inspection of civil infrastructures. Concretely, a classifier is trained to detect defects and coupled with an attribution-based method and adversarial climbing to generate and refine segmentation masks corresponding to the classification outputs. These are used within an assisted labeling framework where the annotators can interact with them as proposal segmentation masks by deciding to accept, reject or modify them, and interactions are logged as weak labels to further refine the classifier. Applied on a real-world dataset resulting from the automated visual inspection of bridges, our proposed method is able to save more than 50\\% of annotators' time when compared to manual annotation of defects."
                    },
                    {
                        "title": "Adaptive Conformal Regression with Jackknife+ Rescaled Scores",
                        "abstract": "Conformal regression provides prediction intervals with global coverage guarantees, but often fails to capture local error distributions, leading to non-homogeneous coverage. We address this with a new adaptive method based on rescaling conformal scores with an estimate of local score distribution, inspired by the Jackknife+ method, which enables the use of calibration data in conformal scores without breaking calibration-test exchangeability. Our approach ensures formal global coverage guarantees and is supported by new theoretical results on local coverage, including an a posteriori bound on any calibration score. The strength of our approach lies in achieving local coverage without sacrificing calibration set size, improving the applicability of conformal prediction intervals in various settings. As a result, our method provides prediction intervals that outperform previous methods, particularly in the low-data regime, making it especially relevant for real-world applications such as healthcare and biomedical domains where uncertainty needs to be quantified accurately despite low sample data."
                    },
                    {
                        "title": "Energy-efficient neuromorphic classifiers",
                        "abstract": "Neuromorphic engineering combines the architectural and computational principles of systems neuroscience with semiconductor electronics, with the aim of building efficient and compact devices that mimic the synaptic and neural machinery of the brain. Neuromorphic engineering promises extremely low energy consumptions, comparable to those of the nervous system. However, until now the neuromorphic approach has been restricted to relatively simple circuits and specialized functions, rendering elusive a direct comparison of their energy consumption to that used by conventional von Neumann digital machines solving real-world tasks. Here we show that a recent technology developed by IBM can be leveraged to realize neuromorphic circuits that operate as classifiers of complex real-world stimuli. These circuits emulate enough neurons to compete with state-of-the-art classifiers. We also show that the energy consumption of the IBM chip is typically 2 or more orders of magnitude lower than that of conventional digital machines when implementing classifiers with comparable performance. Moreover, the spike-based dynamics display a trade-off between integration time and accuracy, which naturally translates into algorithms that can be flexibly deployed for either fast and approximate classifications, or more accurate classifications at the mere expense of longer running times and higher energy costs. This work finally proves that the neuromorphic approach can be efficiently used in real-world applications and it has significant advantages over conventional digital devices when energy consumption is considered."
                    },
                    {
                        "title": "DARE: Towards Robust Text Explanations in Biomedical and Healthcare Applications",
                        "abstract": "Along with the successful deployment of deep neural networks in several application domains, the need to unravel the black-box nature of these networks has seen a significant increase recently. Several methods have been introduced to provide insight into the inference process of deep neural networks. However, most of these explainability methods have been shown to be brittle in the face of adversarial perturbations of their inputs in the image and generic textual domain. In this work we show that this phenomenon extends to specific and important high stakes domains like biomedical datasets. In particular, we observe that the robustness of explanations should be characterized in terms of the accuracy of the explanation in linking a model's inputs and its decisions - faithfulness - and its relevance from the perspective of domain experts - plausibility. This is crucial to prevent explanations that are inaccurate but still look convincing in the context of the domain at hand. To this end, we show how to adapt current attribution robustness estimation methods to a given domain, so as to take into account domain-specific plausibility. This results in our DomainAdaptiveAREstimator (DARE) attribution robustness estimator, allowing us to properly characterize the domain-specific robustness of faithful explanations. Next, we provide two methods, adversarial training and FAR training, to mitigate the brittleness characterized by DARE, allowing us to train networks that display robust attributions. Finally, we empirically validate our methods with extensive experiments on three established biomedical benchmarks."
                    },
                    {
                        "title": "Outline-Guided Object Inpainting with Diffusion Models",
                        "abstract": "Instance segmentation datasets play a crucial role in training accurate and robust computer vision models. However, obtaining accurate mask annotations to produce high-quality segmentation datasets is a costly and labor-intensive process. In this work, we show how this issue can be mitigated by starting with small annotated instance segmentation datasets and augmenting them to effectively obtain a sizeable annotated dataset. We achieve that by creating variations of the available annotated object instances in a way that preserves the provided mask annotations, thereby resulting in new image-mask pairs to be added to the set of annotated images. Specifically, we generate new images using a diffusion-based inpainting model to fill out the masked area with a desired object class by guiding the diffusion through the object outline. We show that the object outline provides a simple, but also reliable and convenient training-free guidance signal for the underlying inpainting model that is often sufficient to fill out the mask with an object of the correct class without further text guidance and preserve the correspondence between generated images and the mask annotations with high precision. Our experimental results reveal that our method successfully generates realistic variations of object instances, preserving their shape characteristics while introducing diversity within the augmented area. We also show that the proposed method can naturally be combined with text guidance and other image augmentation techniques."
                    },
                    {
                        "title": "Estimating the Adversarial Robustness of Attributions in Text with Transformers",
                        "abstract": "Explanations are crucial parts of deep neural network (DNN) classifiers. In high stakes applications, faithful and robust explanations are important to understand and gain trust in DNN classifiers. However, recent work has shown that state-of-the-art attribution methods in text classifiers are susceptible to imperceptible adversarial perturbations that alter explanations significantly while maintaining the correct prediction outcome. If undetected, this can critically mislead the users of DNNs. Thus, it is crucial to understand the influence of such adversarial perturbations on the networks' explanations and their perceptibility. In this work, we establish a novel definition of attribution robustness (AR) in text classification, based on Lipschitz continuity. Crucially, it reflects both attribution change induced by adversarial input alterations and perceptibility of such alterations. Moreover, we introduce a wide set of text similarity measures to effectively capture locality between two text samples and imperceptibility of adversarial perturbations in text. We then propose our novel TransformerExplanationAttack (TEA), a strong adversary that provides a tight estimation for attribution robustness in text classification. TEA uses state-of-the-art language models to extract word substitutions that result in fluent, contextual adversarial samples. Finally, with experiments on several text classification architectures, we show that TEA consistently outperforms current state-of-the-art AR estimators, yielding perturbations that alter explanations to a greater extent while being more fluent and less perceptible."
                    },
                    {
                        "title": "On the generalization capacity of neural networks during generic multimodal reasoning",
                        "abstract": "The advent of the Transformer has led to the development of large language models (LLM), which appear to demonstrate human-like capabilities. To assess the generality of this class of models and a variety of other base neural network architectures to multimodal domains, we evaluated and compared their capacity for multimodal generalization. We introduce a multimodal question-answer benchmark to evaluate three specific types of out-of-distribution (OOD) generalization performance: distractor generalization (generalization in the presence of distractors), systematic compositional generalization (generalization to new task permutations), and productive compositional generalization (generalization to more complex tasks structures). We found that across model architectures (e.g., RNNs, Transformers, Perceivers, etc.), models with multiple attention layers, or models that leveraged cross-attention mechanisms between input domains, fared better. Our positive results demonstrate that for multimodal distractor and systematic generalization, either cross-modal attention or models with deeper attention layers are key architectural features required to integrate multimodal inputs. On the other hand, neither of these architectural features led to productive generalization, suggesting fundamental limitations of existing architectures for specific types of multimodal generalization. These results demonstrate the strengths and limitations of specific architectural components underlying modern neural models for multimodal reasoning. Finally, we provide Generic COG (gCOG), a configurable benchmark with several multimodal generalization splits, for future studies to explore."
                    },
                    {
                        "title": "Alleviating Noisy Data in Image Captioning with Cooperative Distillation",
                        "abstract": "Image captioning systems have made substantial progress, largely due to the availability of curated datasets like Microsoft COCO or Vizwiz that have accurate descriptions of their corresponding images. Unfortunately, scarce availability of such cleanly labeled data results in trained algorithms producing captions that can be terse and idiosyncratically specific to details in the image. We propose a new technique, cooperative distillation that combines clean curated datasets with the web-scale automatically extracted captions of the Google Conceptual Captions dataset (GCC), which can have poor descriptions of images, but is abundant in size and therefore provides a rich vocabulary resulting in more expressive captions."
                    },
                    {
                        "title": "Efficient ConvNets for Analog Arrays",
                        "abstract": "Analog arrays are a promising upcoming hardware technology with the potential to drastically speed up deep learning. Their main advantage is that they compute matrix-vector products in constant time, irrespective of the size of the matrix. However, early convolution layers in ConvNets map very unfavorably onto analog arrays, because kernel matrices are typically small and the constant time operation needs to be sequentially iterated a large number of times, reducing the speed up advantage for ConvNets. Here, we propose to replicate the kernel matrix of a convolution layer on distinct analog arrays, and randomly divide parts of the compute among them, so that multiple kernel matrices are trained in parallel. With this modification, analog arrays execute ConvNets with an acceleration factor that is proportional to the number of kernel matrices used per layer (here tested 16-128). Despite having more free parameters, we show analytically and in numerical experiments that this convolution architecture is self-regularizing and implicitly learns similar filters across arrays. We also report superior performance on a number of datasets and increased robustness to adversarial attacks. Our investigation suggests to revise the notion that mixed analog-digital hardware is not suitable for ConvNets."
                    },
                    {
                        "title": "Sobolev Independence Criterion",
                        "abstract": "We propose the Sobolev Independence Criterion (SIC), an interpretable dependency measure between a high dimensional random variable X and a response variable Y . SIC decomposes to the sum of feature importance scores and hence can be used for nonlinear feature selection. SIC can be seen as a gradient regularized Integral Probability Metric (IPM) between the joint distribution of the two random variables and the product of their marginals. We use sparsity inducing gradient penalties to promote input sparsity of the critic of the IPM. In the kernel version we show that SIC can be cast as a convex optimization problem by introducing auxiliary variables that play an important role in feature selection as they are normalized feature importance scores. We then present a neural version of SIC where the critic is parameterized as a homogeneous neural network, improving its representation power as well as its interpretability. We conduct experiments validating SIC for feature selection in synthetic and real-world experiments. We show that SIC enables reliable and interpretable discoveries, when used in conjunction with the holdout randomization test and knockoffs to control the False Discovery Rate. Code is available at http://github.com/ibm/sic."
                    },
                    {
                        "title": "Active Learning for Imbalanced Civil Infrastructure Data",
                        "abstract": "Aging civil infrastructures are closely monitored by engineers for damage and critical defects. As the manual inspection of such large structures is costly and time-consuming, we are working towards fully automating the visual inspections to support the prioritization of maintenance activities. To that end we combine recent advances in drone technology and deep learning. Unfortunately, annotation costs are incredibly high as our proprietary civil engineering dataset must be annotated by highly trained engineers. Active learning is, therefore, a valuable tool to optimize the trade-off between model performance and annotation costs. Our use-case differs from the classical active learning setting as our dataset suffers from heavy class imbalance and consists of a much larger already labeled data pool than other active learning research. We present a novel method capable of operating in this challenging setting by replacing the traditional active learning acquisition function with an auxiliary binary discriminator. We experimentally show that our novel method outperforms the best-performing traditional active learning method (BALD) by 5% and 38% accuracy on CIFAR-10 and our proprietary dataset respectively."
                    },
                    {
                        "title": "Cloud-Based Real-Time Molecular Screening Platform with MolFormer",
                        "abstract": "With the prospect of automating a number of chemical tasks with high fidelity, chemical language processing models are emerging at a rapid speed. Here, we present a cloud-based real-time platform that allows users to virtually screen molecules of interest. For this purpose, molecular embeddings inferred from a recently proposed large chemical language model, named MolFormer, are leveraged. The platform currently supports three tasks: nearest neighbor retrieval, chemical space visualization, and property prediction. Based on the functionalities of this platform and results obtained, we believe that such a platform can play a pivotal role in automating chemistry and chemical engineering research, as well as assist in drug discovery and material design tasks. A demo of our platform is provided at \\url{www.ibm.biz/molecular_demo}."
                    },
                    {
                        "title": "Tabular Transformers for Modeling Multivariate Time Series",
                        "abstract": "Tabular datasets are ubiquitous in data science applications. Given their importance, it seems natural to apply state-of-the-art deep learning algorithms in order to fully unlock their potential. Here we propose neural network models that represent tabular time series that can optionally leverage their hierarchical structure. This results in two architectures for tabular time series: one for learning representations that is analogous to BERT and can be pre-trained end-to-end and used in downstream tasks, and one that is akin to GPT and can be used for generation of realistic synthetic tabular sequences. We demonstrate our models on two datasets: a synthetic credit card transaction dataset, where the learned representations are used for fraud detection and synthetic data generation, and on a real pollution dataset, where the learned encodings are used to predict atmospheric pollutant concentrations. Code and data are available at https://github.com/IBM/TabFormer."
                    },
                    {
                        "title": "Unraveling the Key Components of OOD Generalization via Diversification",
                        "abstract": "Supervised learning datasets may contain multiple cues that explain the training set equally well, i.e., learning any of them would lead to the correct predictions on the training data. However, many of them can be spurious, i.e., lose their predictive power under a distribution shift and consequently fail to generalize to out-of-distribution (OOD) data. Recently developed \"diversification\" methods (Lee et al., 2023; Pagliardini et al., 2023) approach this problem by finding multiple diverse hypotheses that rely on different features. This paper aims to study this class of methods and identify the key components contributing to their OOD generalization abilities.   We show that (1) diversification methods are highly sensitive to the distribution of the unlabeled data used for diversification and can underperform significantly when away from a method-specific sweet spot. (2) Diversification alone is insufficient for OOD generalization. The choice of the used learning algorithm, e.g., the model's architecture and pretraining, is crucial. In standard experiments (classification on Waterbirds and Office-Home datasets), using the second-best choice leads to an up to 20\\% absolute drop in accuracy. (3) The optimal choice of learning algorithm depends on the unlabeled data and vice versa i.e. they are co-dependent. (4) Finally, we show that, in practice, the above pitfalls cannot be alleviated by increasing the number of diverse hypotheses, the major feature of diversification methods.   These findings provide a clearer understanding of the critical design factors influencing the OOD generalization abilities of diversification methods. They can guide practitioners in how to use the existing methods best and guide researchers in developing new, better ones."
                    },
                    {
                        "title": "Image Captioning as an Assistive Technology: Lessons Learned from VizWiz 2020 Challenge",
                        "abstract": "Image captioning has recently demonstrated impressive progress largely owing to the introduction of neural network algorithms trained on curated dataset like MS-COCO. Often work in this field is motivated by the promise of deployment of captioning systems in practical applications. However, the scarcity of data and contexts in many competition datasets renders the utility of systems trained on these datasets limited as an assistive technology in real-world settings, such as helping visually impaired people navigate and accomplish everyday tasks. This gap motivated the introduction of the novel VizWiz dataset, which consists of images taken by the visually impaired and captions that have useful, task-oriented information. In an attempt to help the machine learning computer vision field realize its promise of producing technologies that have positive social impact, the curators of the VizWiz dataset host several competitions, including one for image captioning. This work details the theory and engineering from our winning submission to the 2020 captioning competition. Our work provides a step towards improved assistive image captioning systems."
                    }
                ]
            },
            "2eda7faf-9b92-4065-9828-069ab9dc78f3": {
                "pk": "2eda7faf-9b92-4065-9828-069ab9dc78f3",
                "name": "Kristjan Greenewald",
                "collaborators": [
                    "Alfred Hero III",
                    "Alfred Hero",
                    "Ziv Goldfeld",
                    "Shuheng Zhou",
                    "Justin Solomon",
                    "Edmund Zelnio",
                    "Stephen Kelley",
                    "Alfred O. Hero III",
                    "Dmitriy Katz-Rogozhnikov",
                    "Karthik Shanmugam"
                ],
                "domain": [
                    "Anomaly Detection",
                    "Spatio-Temporal Analysis",
                    "Machine Learning",
                    "Optimal Transport"
                ],
                "publications": [
                    {
                        "title": "Detection of Anomalous Crowd Behavior Using Spatio-Temporal Multiresolution Model and Kronecker Sum Decompositions",
                        "abstract": "In this work we consider the problem of detecting anomalous spatio-temporal behavior in videos. Our approach is to learn the normative multiframe pixel joint distribution and detect deviations from it using a likelihood based approach. Due to the extreme lack of available training samples relative to the dimension of the distribution, we use a mean and covariance approach and consider methods of learning the spatio-temporal covariance in the low-sample regime. Our approach is to estimate the covariance using parameter reduction and sparse models. The first method considered is the representation of the covariance as a sum of Kronecker products as in (Greenewald et al 2013), which is found to be an accurate approximation in this setting. We propose learning algorithms relevant to our problem. We then consider the sparse multiresolution model of (Choi et al 2010) and apply the Kronecker product methods to it for further parameter reduction, as well as introducing modifications for enhanced efficiency and greater applicability to spatio-temporal covariance matrices. We apply our methods to the detection of crowd behavior anomalies in the University of Minnesota crowd anomaly dataset, and achieve competitive results."
                    },
                    {
                        "title": "Sliced Mutual Information: A Scalable Measure of Statistical Dependence",
                        "abstract": "Mutual information (MI) is a fundamental measure of statistical dependence, with a myriad of applications to information theory, statistics, and machine learning. While it possesses many desirable structural properties, the estimation of high-dimensional MI from samples suffers from the curse of dimensionality. Motivated by statistical scalability to high dimensions, this paper proposes sliced MI (SMI) as a surrogate measure of dependence. SMI is defined as an average of MI terms between one-dimensional random projections. We show that it preserves many of the structural properties of classic MI, while gaining scalable computation and efficient estimation from samples. Furthermore, and in contrast to classic MI, SMI can grow as a result of deterministic transformations. This enables leveraging SMI for feature extraction by optimizing it over processing functions of raw data to identify useful representations thereof. Our theory is supported by numerical studies of independence testing and feature extraction, which demonstrate the potential gains SMI offers over classic MI for high-dimensional inference."
                    },
                    {
                        "title": "Sharper rates of convergence for the tensor graphical Lasso estimator",
                        "abstract": "Many modern datasets exhibit dependencies among observations as well as variables. This gives rise to the challenging problem of analyzing high-dimensional matrix-variate data with unknown dependence structures. To address this challenge, Kalaitzis et. al. (2013) proposed the Bigraphical Lasso (BiGLasso), an estimator for precision matrices of matrix-normals based on the Cartesian product of graphs. Subsequently, Greenewald, Zhou and Hero (GZH 2019) introduced a multiway tensor generalization of the BiGLasso estimator, known as the TeraLasso estimator. In this paper, we provide sharper rates of convergence in the Frobenius and operator norm for both BiGLasso and TeraLasso estimators for estimating inverse covariance matrices. This improves upon the rates presented in GZH 2019. In particular, (a) we strengthen the bounds for the relative errors in the operator and Frobenius norm by a factor of approximately $\\log p$; (b) Crucially, this improvement allows for finite-sample estimation errors in both norms to be derived for the two-way Kronecker sum model. The two-way regime is important because it is the setting that is the most theoretically challenging, and simultaneously the most common in applications. Normality is not needed in our proofs; instead, we consider sub-gaussian ensembles and derive tight concentration of measure bounds, using tensor unfolding techniques. The proof techniques may be of independent interest."
                    },
                    {
                        "title": "Regularized Block Toeplitz Covariance Matrix Estimation via Kronecker Product Expansions",
                        "abstract": "In this work we consider the estimation of spatio-temporal covariance matrices in the low sample non-Gaussian regime. We impose covariance structure in the form of a sum of Kronecker products decomposition (Tsiligkaridis et al. 2013, Greenewald et al. 2013) with diagonal correction (Greenewald et al.), which we refer to as DC-KronPCA, in the estimation of multiframe covariance matrices. This paper extends the approaches of (Tsiligkaridis et al.) in two directions. First, we modify the diagonally corrected method of (Greenewald et al.) to include a block Toeplitz constraint imposing temporal stationarity structure. Second, we improve the conditioning of the estimate in the very low sample regime by using Ledoit-Wolf type shrinkage regularization similar to (Chen, Hero et al. 2010). For improved robustness to heavy tailed distributions, we modify the KronPCA to incorporate robust shrinkage estimation (Chen, Hero et al. 2011). Results of numerical simulations establish benefits in terms of estimation MSE when compared to previous methods. Finally, we apply our methods to a real-world network spatio-temporal anomaly detection problem and achieve superior results."
                    },
                    {
                        "title": "Robust Kronecker Product PCA for Spatio-Temporal Covariance Estimation",
                        "abstract": "Kronecker PCA involves the use of a space vs. time Kronecker product decomposition to estimate spatio-temporal covariances. In this work the addition of a sparse correction factor is considered, which corresponds to a model of the covariance as a sum of Kronecker products of low (separation) rank and a sparse matrix. This sparse correction extends the diagonally corrected Kronecker PCA of [Greenewald et al 2013, 2014] to allow for sparse unstructured \"outliers\" anywhere in the covariance matrix, e.g. arising from variables or correlations that do not fit the Kronecker model well, or from sources such as sensor noise or sensor failure. We introduce a robust PCA-based algorithm to estimate the covariance under this model, extending the rearranged nuclear norm penalized LS Kronecker PCA approaches of [Greenewald et al 2014, Tsiligkaridis et al 2013]. An extension to Toeplitz temporal factors is also provided, producing a parameter reduction for temporally stationary measurement modeling. High dimensional MSE performance bounds are given for these extensions. Finally, the proposed extension of KronPCA is evaluated on both simulated and real data coming from yeast cell cycle experiments. This establishes the practical utility of robust Kronecker PCA in biological and other applications."
                    },
                    {
                        "title": "Gaussian-Smooth Optimal Transport: Metric Structure and Statistical Efficiency",
                        "abstract": "Optimal transport (OT), and in particular the Wasserstein distance, has seen a surge of interest and applications in machine learning. However, empirical approximation under Wasserstein distances suffers from a severe curse of dimensionality, rendering them impractical in high dimensions. As a result, entropically regularized OT has become a popular workaround. However, while it enjoys fast algorithms and better statistical properties, it looses the metric structure that Wasserstein distances enjoy. This work proposes a novel Gaussian-smoothed OT (GOT) framework, that achieves the best of both worlds: preserving the 1-Wasserstein metric structure while alleviating the empirical approximation curse of dimensionality. Furthermore, as the Gaussian-smoothing parameter shrinks to zero, GOT $\\Gamma$-converges towards classic OT (with convergence of optimizers), thus serving as a natural extension. An empirical study that supports the theoretical results is provided, promoting Gaussian-smoothed OT as a powerful alternative to entropic OT."
                    },
                    {
                        "title": "Kronecker STAP and SAR GMTI",
                        "abstract": "As a high resolution radar imaging modality, SAR detects and localizes non-moving targets accurately, giving it an advantage over lower resolution GMTI radars. Moving target detection is more challenging due to target smearing and masking by clutter. Space-time adaptive processing (STAP) is often used on multiantenna SAR to remove the stationary clutter and enhance the moving targets. In (Greenewald et al., 2016) it was shown that the performance of STAP can be improved by modeling the clutter covariance as a space vs. time Kronecker product with low rank factors, providing robustness and reducing the number of training samples required. In this work, we present a massively parallel algorithm for implementing Kronecker product STAP, enabling application to very large SAR datasets (such as the 2006 Gotcha data collection) using GPUs. Finally, we develop an extension of Kronecker STAP that uses information from multiple passes to improve moving target detection."
                    },
                    {
                        "title": "Nonstationary Distance Metric Learning",
                        "abstract": "Recent work in distance metric learning has focused on learning transformations of data that best align with provided sets of pairwise similarity and dissimilarity constraints. The learned transformations lead to improved retrieval, classification, and clustering algorithms due to the better adapted distance or similarity measures. Here, we introduce the problem of learning these transformations when the underlying constraint generation process is nonstationary. This nonstationarity can be due to changes in either the ground-truth clustering used to generate constraints or changes to the feature subspaces in which the class structure is apparent. We propose and evaluate COMID-SADL, an adaptive, online approach for learning and tracking optimal metrics as they change over time that is highly robust to a variety of nonstationary behaviors in the changing metric. We demonstrate COMID-SADL on both real and synthetic data sets and show significant performance improvements relative to previously proposed batch and online distance metric learning algorithms."
                    },
                    {
                        "title": "Robust SAR STAP via Kronecker Decomposition",
                        "abstract": "This paper proposes a spatio-temporal decomposition for the detection of moving targets in multiantenna SAR. As a high resolution radar imaging modality, SAR detects and localizes non-moving targets accurately, giving it an advantage over lower resolution GMTI radars. Moving target detection is more challenging due to target smearing and masking by clutter. Space-time adaptive processing (STAP) is often used to remove the stationary clutter and enhance the moving targets. In this work, it is shown that the performance of STAP can be improved by modeling the clutter covariance as a space vs. time Kronecker product with low rank factors. Based on this model, a low-rank Kronecker product covariance estimation algorithm is proposed, and a novel separable clutter cancelation filter based on the Kronecker covariance estimate is introduced. The proposed method provides orders of magnitude reduction in the required number of training samples, as well as improved robustness to corruption of the training data. Simulation results and experiments using the Gotcha SAR GMTI challenge dataset are presented that confirm the advantages of our approach relative to existing techniques."
                    },
                    {
                        "title": "High-Dimensional Feature Selection for Sample Efficient Treatment Effect Estimation",
                        "abstract": "The estimation of causal treatment effects from observational data is a fundamental problem in causal inference. To avoid bias, the effect estimator must control for all confounders. Hence practitioners often collect data for as many covariates as possible to raise the chances of including the relevant confounders. While this addresses the bias, this has the side effect of significantly increasing the number of data samples required to accurately estimate the effect due to the increased dimensionality. In this work, we consider the setting where out of a large number of covariates $X$ that satisfy strong ignorability, an unknown sparse subset $S$ is sufficient to include to achieve zero bias, i.e. $c$-equivalent to $X$. We propose a common objective function involving outcomes across treatment cohorts with nonconvex joint sparsity regularization that is guaranteed to recover $S$ with high probability under a linear outcome model for $Y$ and subgaussian covariates for each of the treatment cohort. This improves the effect estimation sample complexity so that it scales with the cardinality of the sparse subset $S$ and $\\log |X|$, as opposed to the cardinality of the full set $X$. We validate our approach with experiments on treatment effect estimation."
                    },
                    {
                        "title": "Domain Adaptable Prescriptive AI Agent for Enterprise",
                        "abstract": "Despite advancements in causal inference and prescriptive AI, its adoption in enterprise settings remains hindered primarily due to its technical complexity. Many users lack the necessary knowledge and appropriate tools to effectively leverage these technologies. This work at the MIT-IBM Watson AI Lab focuses on developing the proof-of-concept agent, PrecAIse, a domain-adaptable conversational agent equipped with a suite of causal and prescriptive tools to help enterprise users make better business decisions. The objective is to make advanced, novel causal inference and prescriptive tools widely accessible through natural language interactions. The presented Natural Language User Interface (NLUI) enables users with limited expertise in machine learning and data science to harness prescriptive analytics in their decision-making processes without requiring intensive computing resources. We present an agent capable of function calling, maintaining faithful, interactive, and dynamic conversations, and supporting new domains."
                    },
                    {
                        "title": "Time-dependent spatially varying graphical models, with application to brain fMRI data analysis",
                        "abstract": "In this work, we present an additive model for space-time data that splits the data into a temporally correlated component and a spatially correlated component. We model the spatially correlated portion using a time-varying Gaussian graphical model. Under assumptions on the smoothness of changes in covariance matrices, we derive strong single sample convergence results, confirming our ability to estimate meaningful graphical structures as they evolve over time. We apply our methodology to the discovery of time-varying spatial structures in human brain fMRI signals."
                    },
                    {
                        "title": "Tensor Graphical Lasso (TeraLasso)",
                        "abstract": "This paper introduces a multi-way tensor generalization of the Bigraphical Lasso (BiGLasso), which uses a two-way sparse Kronecker-sum multivariate-normal model for the precision matrix to parsimoniously model conditional dependence relationships of matrix-variate data based on the Cartesian product of graphs. We call this generalization the {\\bf Te}nsor g{\\bf ra}phical Lasso (TeraLasso). We demonstrate using theory and examples that the TeraLasso model can be accurately and scalably estimated from very limited data samples of high dimensional variables with multiway coordinates such as space, time and replicates. Statistical consistency and statistical rates of convergence are established for both the BiGLasso and TeraLasso estimators of the precision matrix and estimators of its support (non-sparsity) set, respectively. We propose a scalable composite gradient descent algorithm and analyze the computational convergence rate, showing that the composite gradient descent algorithm is guaranteed to converge at a geometric rate to the global minimizer of the TeraLasso objective function. Finally, we illustrate the TeraLasso using both simulation and experimental data from a meteorological dataset, showing that we can accurately estimate precision matrices and recover meaningful conditional dependency graphs from high dimensional complex datasets."
                    },
                    {
                        "title": "k-Sliced Mutual Information: A Quantitative Study of Scalability with Dimension",
                        "abstract": "Sliced mutual information (SMI) is defined as an average of mutual information (MI) terms between one-dimensional random projections of the random variables. It serves as a surrogate measure of dependence to classic MI that preserves many of its properties but is more scalable to high dimensions. However, a quantitative characterization of how SMI itself and estimation rates thereof depend on the ambient dimension, which is crucial to the understanding of scalability, remain obscure. This work provides a multifaceted account of the dependence of SMI on dimension, under a broader framework termed $k$-SMI, which considers projections to $k$-dimensional subspaces. Using a new result on the continuity of differential entropy in the 2-Wasserstein metric, we derive sharp bounds on the error of Monte Carlo (MC)-based estimates of $k$-SMI, with explicit dependence on $k$ and the ambient dimension, revealing their interplay with the number of samples. We then combine the MC integrator with the neural estimation framework to provide an end-to-end $k$-SMI estimator, for which optimal convergence rates are established. We also explore asymptotics of the population $k$-SMI as dimension grows, providing Gaussian approximation results with a residual that decays under appropriate moment bounds. All our results trivially apply to SMI by setting $k=1$. Our theory is validated with numerical experiments and is applied to sliced InfoGAN, which altogether provide a comprehensive quantitative account of the scalability question of $k$-SMI, including SMI as a special case when $k=1$."
                    },
                    {
                        "title": "Dynamic Metric Learning from Pairwise Comparisons",
                        "abstract": "Recent work in distance metric learning has focused on learning transformations of data that best align with specified pairwise similarity and dissimilarity constraints, often supplied by a human observer. The learned transformations lead to improved retrieval, classification, and clustering algorithms due to the better adapted distance or similarity measures. Here, we address the problem of learning these transformations when the underlying constraint generation process is nonstationary. This nonstationarity can be due to changes in either the ground-truth clustering used to generate constraints or changes in the feature subspaces in which the class structure is apparent. We propose Online Convex Ensemble StrongLy Adaptive Dynamic Learning (OCELAD), a general adaptive, online approach for learning and tracking optimal metrics as they change over time that is highly robust to a variety of nonstationary behaviors in the changing metric. We apply the OCELAD framework to an ensemble of online learners. Specifically, we create a retro-initialized composite objective mirror descent (COMID) ensemble (RICE) consisting of a set of parallel COMID learners with different learning rates, demonstrate RICE-OCELAD on both real and synthetic data sets and show significant performance improvements relative to previously proposed batch and online distance metric learning algorithms."
                    },
                    {
                        "title": "Action Centered Contextual Bandits",
                        "abstract": "Contextual bandits have become popular as they offer a middle ground between very simple approaches based on multi-armed bandits and very complex approaches using the full power of reinforcement learning. They have demonstrated success in web applications and have a rich body of associated theoretical guarantees. Linear models are well understood theoretically and preferred by practitioners because they are not only easily interpretable but also simple to implement and debug. Furthermore, if the linear model is true, we get very strong performance guarantees. Unfortunately, in emerging applications in mobile health, the time-invariant linear model assumption is untenable. We provide an extension of the linear model for contextual bandits that has two parts: baseline reward and treatment effect. We allow the former to be complex but keep the latter simple. We argue that this model is plausible for mobile health applications. At the same time, it leads to algorithms with strong performance guarantees as in the linear model setting, while still allowing for complex nonlinear baseline modeling. Our theory is supported by experiments on data gathered in a recently concluded mobile health study."
                    },
                    {
                        "title": "k-Mixup Regularization for Deep Learning via Optimal Transport",
                        "abstract": "Mixup is a popular regularization technique for training deep neural networks that improves generalization and increases robustness to certain distribution shifts. It perturbs input training data in the direction of other randomly-chosen instances in the training set. To better leverage the structure of the data, we extend mixup in a simple, broadly applicable way to \\emph{$k$-mixup}, which perturbs $k$-batches of training points in the direction of other $k$-batches. The perturbation is done with displacement interpolation, i.e. interpolation under the Wasserstein metric. We demonstrate theoretically and in simulations that $k$-mixup preserves cluster and manifold structures, and we extend theory studying the efficacy of standard mixup to the $k$-mixup case. Our empirical results show that training with $k$-mixup further improves generalization and robustness across several network architectures and benchmark datasets of differing modalities. For the wide variety of real datasets considered, the performance gains of $k$-mixup over standard mixup are similar to or larger than the gains of mixup itself over standard ERM after hyperparameter optimization. In several instances, in fact, $k$-mixup achieves gains in settings where standard mixup has negligible to zero improvement over ERM."
                    },
                    {
                        "title": "$k$-Variance: A Clustered Notion of Variance",
                        "abstract": "We introduce $k$-variance, a generalization of variance built on the machinery of random bipartite matchings. $K$-variance measures the expected cost of matching two sets of $k$ samples from a distribution to each other, capturing local rather than global information about a measure as $k$ increases; it is easily approximated stochastically using sampling and linear programming. In addition to defining $k$-variance and proving its basic properties, we provide in-depth analysis of this quantity in several key cases, including one-dimensional measures, clustered measures, and measures concentrated on low-dimensional subsets of $\\mathbb R^n$. We conclude with experiments and open problems motivated by this new way to summarize distributional shape."
                    },
                    {
                        "title": "Improving Approximate Optimal Transport Distances using Quantization",
                        "abstract": "Optimal transport (OT) is a popular tool in machine learning to compare probability measures geometrically, but it comes with substantial computational burden. Linear programming algorithms for computing OT distances scale cubically in the size of the input, making OT impractical in the large-sample regime. We introduce a practical algorithm, which relies on a quantization step, to estimate OT distances between measures given cheap sample access. We also provide a variant of our algorithm to improve the performance of approximate solvers, focusing on those for entropy-regularized transport. We give theoretical guarantees on the benefits of this quantization step and display experiments showing that it behaves well in practice, providing a practical approximation algorithm that can be used as a drop-in replacement for existing OT estimators."
                    }
                ]
            },
            "197e8d8d-2d64-496e-a33b-8431a6f2477b": {
                "pk": "197e8d8d-2d64-496e-a33b-8431a6f2477b",
                "name": "Youssef Mroueh",
                "collaborators": [
                    "Tom Sercu",
                    "Vaibhava Goel",
                    "Lorenzo Rosasco",
                    "Carles Domingo-Enrich",
                    "Etienne Marcheret",
                    "Tomaso Poggio",
                    "Truyen Nguyen",
                    "Mattia Rigotti",
                    "Jean-Jacques Slotine",
                    "Stephen Voinea"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Optimal Transport",
                    "Machine Learning",
                    "Phase Retrieval"
                ],
                "publications": [
                    {
                        "title": "Regularized Finite Dimensional Kernel Sobolev Discrepancy",
                        "abstract": "We show in this note that the Sobolev Discrepancy introduced in Mroueh et al in the context of generative adversarial networks, is actually the weighted negative Sobolev norm $||.||_{\\dot{H}^{-1}(\\nu_q)}$, that is known to linearize the Wasserstein $W_2$ distance and plays a fundamental role in the dynamic formulation of optimal transport of Benamou and Brenier. Given a Kernel with finite dimensional feature map we show that the Sobolev discrepancy can be approximated from finite samples. Assuming this discrepancy is finite, the error depends on the approximation error in the function space induced by the finite dimensional feature space kernel and on a statistical error due to the finite sample approximation."
                    },
                    {
                        "title": "Robust Phase Retrieval and Super-Resolution from One Bit Coded Diffraction Patterns",
                        "abstract": "In this paper we study a realistic setup for phase retrieval, where the signal of interest is modulated or masked and then for each modulation or mask a diffraction pattern is collected, producing a coded diffraction pattern (CDP) [CLM13]. We are interested in the setup where the resolution of the collected CDP is limited by the Fraunhofer diffraction limit of the imaging system.   We investigate a novel approach based on a geometric quantization scheme of phase-less linear measurements into (one-bit) coded diffraction patterns, and a corresponding recovery scheme. The key novelty in this approach consists in comparing pairs of coded diffractions patterns across frequencies: the one bit measurements obtained rely on the order statistics of the un-quantized measurements rather than their values . This results in a robust phase recovery, and unlike currently available methods, allows to efficiently perform phase recovery from measurements affected by severe (possibly unknown) non linear, rank preserving perturbations, such as distortions. Another important feature of this approach consists in the fact that it enables also super-resolution and blind-deconvolution, beyond the diffraction limit of a given imaging system."
                    },
                    {
                        "title": "Wasserstein Style Transfer",
                        "abstract": "We propose Gaussian optimal transport for Image style transfer in an Encoder/Decoder framework. Optimal transport for Gaussian measures has closed forms Monge mappings from source to target distributions. Moreover interpolates between a content and a style image can be seen as geodesics in the Wasserstein Geometry. Using this insight, we show how to mix different target styles , using Wasserstein barycenter of Gaussian measures. Since Gaussians are closed under Wasserstein barycenter, this allows us a simple style transfer and style mixing and interpolation. Moreover we show how mixing different styles can be achieved using other geodesic metrics between gaussians such as the Fisher Rao metric, while the transport of the content to the new interpolate style is still performed with Gaussian OT maps. Our simple methodology allows to generate new stylized content interpolating between many artistic styles. The metric used in the interpolation results in different stylizations."
                    },
                    {
                        "title": "Information Theoretic Guarantees For Policy Alignment In Large Language Models",
                        "abstract": "Policy alignment of large language models refers to constrained policy optimization, where the policy is optimized to maximize a reward while staying close to a reference policy with respect to an $f$-divergence such as the $\\mathsf{KL}$ divergence. The best of $n$ alignment policy selects a sample from the reference policy that has the maximum reward among $n$ independent samples. For both cases (policy alignment and best of $n$), recent works showed empirically that the reward improvement of the aligned policy on the reference one scales like $\\sqrt{\\mathsf{KL}}$, with an explicit bound in $n$ on the $\\mathsf{KL}$ for the best of $n$ policy. We show in this paper that the $\\sqrt{\\mathsf{KL}}$ information theoretic upper bound holds if the reward under the reference policy has sub-gaussian tails. Moreover, we prove for the best of $n$ policy, that the $\\mathsf{KL}$ upper bound can be obtained for any $f$-divergence via a reduction to exponential order statistics owing to the R\\'enyi representation of order statistics, and a data processing inequality. If additional information is known on the tails of the aligned policy we show that tighter control on the reward improvement can be obtained via the R\\'enyi divergence. Finally we demonstrate how these upper bounds transfer from proxy rewards to golden rewards which results in a decrease in the golden reward improvement due to overestimation and approximation errors of the proxy reward."
                    },
                    {
                        "title": "q-ary Compressive Sensing",
                        "abstract": "We introduce q-ary compressive sensing, an extension of 1-bit compressive sensing. We propose a novel sensing mechanism and a corresponding recovery procedure. The recovery properties of the proposed approach are analyzed both theoretically and empirically. Results in 1-bit compressive sensing are recovered as a special case. Our theoretical results suggest a tradeoff between the quantization parameter q, and the number of measurements m in the control of the error of the resulting recovery algorithm, as well its robustness to noise."
                    },
                    {
                        "title": "Quantization and Greed are Good: One bit Phase Retrieval, Robustness and Greedy Refinements",
                        "abstract": "In this paper, we study the problem of robust phase recovery. We investigate a novel approach based on extremely quantized (one-bit) phase-less measurements and a corresponding recovery scheme. The proposed approach has surprising robustness and stability properties and, unlike currently available methods, allows to efficiently perform phase recovery from measurements affected by severe (possibly unknown) non-linear perturbations, such as distortions (e.g. clipping). Beyond robustness, we show how our approach can be used within greedy approaches based on alternating minimization. In particular, we propose novel initialization schemes for the alternating minimization achieving favorable convergence properties with improved sample complexity."
                    },
                    {
                        "title": "On the Convergence of Gradient Descent in GANs: MMD GAN As a Gradient Flow",
                        "abstract": "We consider the maximum mean discrepancy ($\\mathrm{MMD}$) GAN problem and propose a parametric kernelized gradient flow that mimics the min-max game in gradient regularized $\\mathrm{MMD}$ GAN. We show that this flow provides a descent direction minimizing the $\\mathrm{MMD}$ on a statistical manifold of probability distributions. We then derive an explicit condition which ensures that gradient descent on the parameter space of the generator in gradient regularized $\\mathrm{MMD}$ GAN is globally convergent to the target distribution. Under this condition, we give non asymptotic convergence results of gradient descent in MMD GAN. Another contribution of this paper is the introduction of a dynamic formulation of a regularization of $\\mathrm{MMD}$ and demonstrating that the parametric kernelized descent for $\\mathrm{MMD}$ is the gradient flow of this functional with respect to the new Riemannian structure. Our obtained theoretical result allows ones to treat gradient flows for quite general functionals and thus has potential applications to other types of variational inferences on a statistical manifold beyond GANs. Finally, numerical experiments suggest that our parametric kernelized gradient flow stabilizes GAN training and guarantees convergence."
                    },
                    {
                        "title": "Semi-Supervised Learning with IPM-based GANs: an Empirical Study",
                        "abstract": "We present an empirical investigation of a recent class of Generative Adversarial Networks (GANs) using Integral Probability Metrics (IPM) and their performance for semi-supervised learning. IPM-based GANs like Wasserstein GAN, Fisher GAN and Sobolev GAN have desirable properties in terms of theoretical understanding, training stability, and a meaningful loss. In this work we investigate how the design of the critic (or discriminator) influences the performance in semi-supervised learning. We distill three key take-aways which are important for good SSL performance: (1) the K+1 formulation, (2) avoiding batch normalization in the critic and (3) avoiding gradient penalty constraints on the classification layer."
                    },
                    {
                        "title": "Auditing Differential Privacy in High Dimensions with the Kernel Quantum R\u00e9nyi Divergence",
                        "abstract": "Differential privacy (DP) is the de facto standard for private data release and private machine learning. Auditing black-box DP algorithms and mechanisms to certify whether they satisfy a certain DP guarantee is challenging, especially in high dimension. We propose relaxations of differential privacy based on new divergences on probability distributions: the kernel R\\'enyi divergence and its regularized version. We show that the regularized kernel R\\'enyi divergence can be estimated from samples even in high dimensions, giving rise to auditing procedures for $\\varepsilon$-DP, $(\\varepsilon,\\delta)$-DP and $(\\alpha,\\varepsilon)$-R\\'enyi DP."
                    },
                    {
                        "title": "Unbalanced Sobolev Descent",
                        "abstract": "We introduce Unbalanced Sobolev Descent (USD), a particle descent algorithm for transporting a high dimensional source distribution to a target distribution that does not necessarily have the same mass. We define the Sobolev-Fisher discrepancy between distributions and show that it relates to advection-reaction transport equations and the Wasserstein-Fisher-Rao metric between distributions. USD transports particles along gradient flows of the witness function of the Sobolev-Fisher discrepancy (advection step) and reweighs the mass of particles with respect to this witness function (reaction step). The reaction step can be thought of as a birth-death process of the particles with rate of growth proportional to the witness function. When the Sobolev-Fisher witness function is estimated in a Reproducing Kernel Hilbert Space (RKHS), under mild assumptions we show that USD converges asymptotically (in the limit of infinite particles) to the target distribution in the Maximum Mean Discrepancy (MMD) sense. We then give two methods to estimate the Sobolev-Fisher witness with neural networks, resulting in two Neural USD algorithms. The first one implements the reaction step with mirror descent on the weights, while the second implements it through a birth-death process of particles. We show on synthetic examples that USD transports distributions with or without conservation of mass faster than previous particle descent algorithms, and finally demonstrate its use for molecular biology analyses where our method is naturally suited to match developmental stages of populations of differentiating cells based on their single-cell RNA sequencing profile. Code is available at https://github.com/ibm/usd ."
                    },
                    {
                        "title": "Co-Occuring Directions Sketching for Approximate Matrix Multiply",
                        "abstract": "We introduce co-occurring directions sketching, a deterministic algorithm for approximate matrix product (AMM), in the streaming model. We show that co-occuring directions achieves a better error bound for AMM than other randomized and deterministic approaches for AMM. Co-occurring directions gives a $1 + \\epsilon$ -approximation of the optimal low rank approximation of a matrix product. Empirically our algorithm outperforms competing methods for AMM, for a small sketch size. We validate empirically our theoretical findings and algorithms"
                    },
                    {
                        "title": "Fisher GAN",
                        "abstract": "Generative Adversarial Networks (GANs) are powerful models for learning complex distributions. Stable training of GANs has been addressed in many recent works which explore different metrics between distributions. In this paper we introduce Fisher GAN which fits within the Integral Probability Metrics (IPM) framework for training GANs. Fisher GAN defines a critic with a data dependent constraint on its second order moments. We show in this paper that Fisher GAN allows for stable and time efficient training that does not compromise the capacity of the critic, and does not need data independent constraints such as weight clipping. We analyze our Fisher IPM theoretically and provide an algorithm based on Augmented Lagrangian for Fisher GAN. We validate our claims on both image sample generation and semi-supervised classification using Fisher GAN."
                    },
                    {
                        "title": "Separation Results between Fixed-Kernel and Feature-Learning Probability Metrics",
                        "abstract": "Several works in implicit and explicit generative modeling empirically observed that feature-learning discriminators outperform fixed-kernel discriminators in terms of the sample quality of the models. We provide separation results between probability metrics with fixed-kernel and feature-learning discriminators using the function classes $\\mathcal{F}_2$ and $\\mathcal{F}_1$ respectively, which were developed to study overparametrized two-layer neural networks. In particular, we construct pairs of distributions over hyper-spheres that can not be discriminated by fixed kernel $(\\mathcal{F}_2)$ integral probability metric (IPM) and Stein discrepancy (SD) in high dimensions, but that can be discriminated by their feature learning ($\\mathcal{F}_1$) counterparts. To further study the separation we provide links between the $\\mathcal{F}_1$ and $\\mathcal{F}_2$ IPMs with sliced Wasserstein distances. Our work suggests that fixed-kernel discriminators perform worse than their feature learning counterparts because their corresponding metrics are weaker."
                    },
                    {
                        "title": "McGan: Mean and Covariance Feature Matching GAN",
                        "abstract": "We introduce new families of Integral Probability Metrics (IPM) for training Generative Adversarial Networks (GAN). Our IPMs are based on matching statistics of distributions embedded in a finite dimensional feature space. Mean and covariance feature matching IPMs allow for stable training of GANs, which we will call McGan. McGan minimizes a meaningful loss between distributions."
                    },
                    {
                        "title": "Multiclass Learning with Simplex Coding",
                        "abstract": "In this paper we discuss a novel framework for multiclass learning, defined by a suitable coding/decoding strategy, namely the simplex coding, that allows to generalize to multiple classes a relaxation approach commonly used in binary classification. In this framework, a relaxation error analysis can be developed avoiding constraints on the considered hypotheses class. Moreover, we show that in this setting it is possible to derive the first provably consistent regularized method with training/tuning complexity which is independent to the number of classes. Tools from convex analysis are introduced that can be used beyond the scope of this paper."
                    },
                    {
                        "title": "Tighter Sparse Approximation Bounds for ReLU Neural Networks",
                        "abstract": "A well-known line of work (Barron, 1993; Breiman, 1993; Klusowski & Barron, 2018) provides bounds on the width $n$ of a ReLU two-layer neural network needed to approximate a function $f$ over the ball $\\mathcal{B}_R(\\mathbb{R}^d)$ up to error $\\epsilon$, when the Fourier based quantity $C_f = \\frac{1}{(2\\pi)^{d/2}} \\int_{\\mathbb{R}^d} \\|\\xi\\|^2 |\\hat{f}(\\xi)| \\ d\\xi$ is finite. More recently Ongie et al. (2019) used the Radon transform as a tool for analysis of infinite-width ReLU two-layer networks. In particular, they introduce the concept of Radon-based $\\mathcal{R}$-norms and show that a function defined on $\\mathbb{R}^d$ can be represented as an infinite-width two-layer neural network if and only if its $\\mathcal{R}$-norm is finite. In this work, we extend the framework of Ongie et al. (2019) and define similar Radon-based semi-norms ($\\mathcal{R}, \\mathcal{U}$-norms) such that a function admits an infinite-width neural network representation on a bounded open set $\\mathcal{U} \\subseteq \\mathbb{R}^d$ when its $\\mathcal{R}, \\mathcal{U}$-norm is finite. Building on this, we derive sparse (finite-width) neural network approximation bounds that refine those of Breiman (1993); Klusowski & Barron (2018). Finally, we show that infinite-width neural network representations on bounded open sets are not unique and study their structure, providing a functional view of mode connectivity."
                    },
                    {
                        "title": "Learning with Group Invariant Features: A Kernel Perspective",
                        "abstract": "We analyze in this paper a random feature map based on a theory of invariance I-theory introduced recently. More specifically, a group invariant signal signature is obtained through cumulative distributions of group transformed random projections. Our analysis bridges invariant feature learning with kernel methods, as we show that this feature map defines an expected Haar integration kernel that is invariant to the specified group action. We show how this non-linear random feature map approximates this group invariant kernel uniformly on a set of $N$ points. Moreover, we show that it defines a function space that is dense in the equivalent Invariant Reproducing Kernel Hilbert Space. Finally, we quantify error rates of the convergence of the empirical risk minimization, as well as the reduction in the sample complexity of a learning algorithm using such an invariant representation for signal classification, in a classical supervised learning setting."
                    },
                    {
                        "title": "Random Maxout Features",
                        "abstract": "In this paper, we propose and study random maxout features, which are constructed by first projecting the input data onto sets of randomly generated vectors with Gaussian elements, and then outputing the maximum projection value for each set. We show that the resulting random feature map, when used in conjunction with linear models, allows for the locally linear estimation of the function of interest in classification tasks, and for the locally linear embedding of points when used for dimensionality reduction or data visualization. We derive generalization bounds for learning that assess the error in approximating locally linear functions by linear functions in the maxout feature space, and empirically evaluate the efficacy of the approach on the MNIST and TIMIT classification tasks."
                    },
                    {
                        "title": "Asymmetrically Weighted CCA And Hierarchical Kernel Sentence Embedding For Image & Text Retrieval",
                        "abstract": "Joint modeling of language and vision has been drawing increasing interest. A multimodal data representation allowing for bidirectional retrieval of images by sentences and vice versa is a key aspect. In this paper we present three contributions in canonical correlation analysis (CCA) based multimodal retrieval. Firstly, we show that an asymmetric weighting of the canonical weights, while achieving a cross view mapping from the search to the query space, improves the retrieval performance. Secondly, we devise a computationally efficient model selection, crucial to generalization and stability, in the framework of the Bj\\\"ork Golub algorithm for regularized CCA via spectral filtering. Finally, we introduce a Hierarchical Kernel Sentence Embedding (HKSE) that approximates Kernel CCA for a special similarity kernel between distribution of words embedded in a vector space. State of the art results are obtained on MSCOCO and Flickr benchmarks when these three techniques are used in conjunction."
                    },
                    {
                        "title": "Sobolev Descent",
                        "abstract": "We study a simplification of GAN training: the problem of transporting particles from a source to a target distribution. Starting from the Sobolev GAN critic, part of the gradient regularized GAN family, we show a strong relation with Optimal Transport (OT). Specifically with the less popular dynamic formulation of OT that finds a path of distributions from source to target minimizing a ``kinetic energy''. We introduce Sobolev descent that constructs similar paths by following gradient flows of a critic function in a kernel space or parametrized by a neural network. In the kernel version, we show convergence to the target distribution in the MMD sense. We show in theory and experiments that regularization has an important role in favoring smooth transitions between distributions, avoiding large gradients from the critic. This analysis in a simplified particle setting provides insight in paths to equilibrium in GANs."
                    }
                ]
            }
        }
    },
    "2406.00529": {
        "paper_data": {
            "title": "On the Use of Anchoring for Training Vision Models",
            "url": "http://arxiv.org/abs/2406.00529v1",
            "arxiv_id": "2406.00529",
            "authors": [
                "Vivek Narayanaswamy",
                "Kowshik Thopalli",
                "Rushil Anirudh",
                "Yamen Mubarka",
                "Wesam Sakla",
                "Jayaraman J. Thiagarajan"
            ],
            "abstract": "Anchoring is a recent, architecture-agnostic principle for training deep neural networks that has been shown to significantly improve uncertainty estimation, calibration, and extrapolation capabilities. In this paper, we systematically explore anchoring as a general protocol for training vision models, providing fundamental insights into its training and inference processes and their implications for generalization and safety. Despite its promise, we identify a critical problem in anchored training that can lead to an increased risk of learning undesirable shortcuts, thereby limiting its generalization capabilities. To address this, we introduce a new anchored training protocol that employs a simple regularizer to mitigate this issue and significantly enhances generalization. We empirically evaluate our proposed approach across datasets and architectures of varying scales and complexities, demonstrating substantial performance gains in generalization and safety metrics compared to the standard training protocol.",
            "introduction": "   1 Introduction  Anchoring\u00a0(Thiagarajan et\u00a0al., 2022) is a recent architecture-agnostic principle for training deep neural networks. It reparameterizes each input xx\\mathrm{x}roman_x into a tuple comprising a reference sample r\u00af\u00afr\\bar{\\mathrm{r}}over\u00af start_ARG roman_r end_ARG and the residual d=x\u2212r\u00afdx\u00afr\\mathrm{d}=\\mathrm{x}-\\bar{\\mathrm{r}}roman_d = roman_x - over\u00af start_ARG roman_r end_ARG, i.e., [r\u00af,d]\u00afrd[\\bar{\\mathrm{r}},\\mathrm{d}][ over\u00af start_ARG roman_r end_ARG , roman_d ], r\u00af\u223cPrsimilar-to\u00afrsubscript\ud835\udc43r\\bar{\\mathrm{r}}\\sim P_{\\mathrm{r}}over\u00af start_ARG roman_r end_ARG \u223c italic_P start_POSTSUBSCRIPT roman_r end_POSTSUBSCRIPT and d\u223cP\u0394similar-todsubscript\ud835\udc43\u0394\\mathrm{d}\\sim P_{\\Delta}roman_d \u223c italic_P start_POSTSUBSCRIPT roman_\u0394 end_POSTSUBSCRIPT. Here, Prsubscript\ud835\udc43rP_{\\mathrm{r}}italic_P start_POSTSUBSCRIPT roman_r end_POSTSUBSCRIPT and P\u0394subscript\ud835\udc43\u0394P_{\\Delta}italic_P start_POSTSUBSCRIPT roman_\u0394 end_POSTSUBSCRIPT denote the distributions of references and residuals respectively. The resulting tuple is then fed as input to a deep network instead of the original input xx\\mathrm{x}roman_x, by concatenating the tuple elements along the feature axis for vector-valued data or the channel axis for image data. Although the first layer of the network needs to be modified to accommodate twice the number of input dimensions (due to concatenation), the rest of the model architecture and optimization strategies remain the same as in standard training. This simple re-parameterization of the input forces the neural network to model the joint distribution P(r,\u0394)subscript\ud835\udc43r\u0394P_{(\\mathrm{r},\\Delta)}italic_P start_POSTSUBSCRIPT ( roman_r , roman_\u0394 ) end_POSTSUBSCRIPT for predicting the target label yy\\mathrm{y}roman_y. Formally, the training objective can be written as:    \u03b8\u2217=arg\u2061min\ud835\udf031|\ud835\udc9f|\u2062\u2211(x,y)\u2208\ud835\udc9f\ud835\udd3cr\u00af\u223cPr\u2062\u2112\u2062[y,\u2131\u03b8\u2062(concat\u2062([r\u00af,x\u2212r\u00af]))],superscript\ud835\udf03\ud835\udf031\ud835\udc9fsubscriptxy\ud835\udc9fsimilar-to\u00afrsubscript\ud835\udc43r\ud835\udd3c\u2112ysubscript\u2131\ud835\udf03concat\u00afrx\u00afr\\theta^{*}=\\arg\\underset{\\theta}{\\min}\\leavevmode\\nobreak\\ \\leavevmode\\nobreak% \\ \\leavevmode\\nobreak\\ \\frac{1}{|\\mathcal{D}|}\\sum_{(\\mathrm{x},\\mathrm{y})\\in% \\mathcal{D}}\\underset{\\bar{\\mathrm{r}}\\sim P_{\\mathrm{r}}}{\\mathbb{E}}\\mathcal% {L}\\bigg{[}\\mathrm{y},\\mathcal{F}_{\\theta}\\bigg{(}\\texttt{concat}([\\bar{% \\mathrm{r}},\\mathrm{x}-\\bar{\\mathrm{r}}])\\bigg{)}\\bigg{]},italic_\u03b8 start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT = roman_arg underitalic_\u03b8 start_ARG roman_min end_ARG divide start_ARG 1 end_ARG start_ARG | caligraphic_D | end_ARG \u2211 start_POSTSUBSCRIPT ( roman_x , roman_y ) \u2208 caligraphic_D end_POSTSUBSCRIPT start_UNDERACCENT over\u00af start_ARG roman_r end_ARG \u223c italic_P start_POSTSUBSCRIPT roman_r end_POSTSUBSCRIPT end_UNDERACCENT start_ARG blackboard_E end_ARG caligraphic_L [ roman_y , caligraphic_F start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( concat ( [ over\u00af start_ARG roman_r end_ARG , roman_x - over\u00af start_ARG roman_r end_ARG ] ) ) ] ,  (1)   where \u2112(.)\\mathcal{L}(.)caligraphic_L ( . ) is a loss function such as cross-entropy, \ud835\udc9f\ud835\udc9f\\mathcal{D}caligraphic_D is the training dataset and \u2131\u2131\\mathcal{F}caligraphic_F is the underlying network parameterized by \u03b8\ud835\udf03\\thetaitalic_\u03b8. In effect, for a given xx\\mathrm{x}roman_x and reference samples r\u00af1,\u2026,r\u00afksubscript\u00afr1\u2026subscript\u00afr\ud835\udc58\\bar{\\mathrm{r}}_{1},\\dots,\\bar{\\mathrm{r}}_{k}over\u00af start_ARG roman_r end_ARG start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , \u2026 , over\u00af start_ARG roman_r end_ARG start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT, anchoring ensures that \u2131\u03b8\u2062([r\u00af1,d1])=\u22ef=\u2131\u03b8\u2062([r\u00afk,dk])subscript\u2131\ud835\udf03subscript\u00afr1subscriptd1\u22efsubscript\u2131\ud835\udf03subscript\u00afr\ud835\udc58subscriptd\ud835\udc58\\mathcal{F}_{\\theta}([\\bar{\\mathrm{r}}_{1},\\mathrm{d}_{1}])=\\dots=\\mathcal{F}_% {\\theta}([\\bar{\\mathrm{r}}_{k},\\mathrm{d}_{k}])caligraphic_F start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( [ over\u00af start_ARG roman_r end_ARG start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , roman_d start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ] ) = \u22ef = caligraphic_F start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( [ over\u00af start_ARG roman_r end_ARG start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT , roman_d start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT ] ), where dk=x\u2212r\u00afksubscriptd\ud835\udc58xsubscript\u00afr\ud835\udc58\\mathrm{d}_{k}=\\mathrm{x}-\\bar{\\mathrm{r}}_{k}roman_d start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT = roman_x - over\u00af start_ARG roman_r end_ARG start_POSTSUBSCRIPT italic_k end_POSTSUBSCRIPT. In other words, regardless of the choice of reference the model must arrive at the same prediction for an input. This principle has been shown to produce models with improved calibration and extrapolation properties\u00a0(Netanyahu et\u00a0al., 2023; Anirudh and Thiagarajan, 2023), and to facilitate accurate epistemic uncertainty estimation\u00a0(Thiagarajan et\u00a0al., 2022). In this paper, we systematically explore the utility of anchoring as a generic protocol for building vision models and make a number of fundamental insights on its training and inferencing, applicability to different architecture families (conv-nets, transformers), and most importantly, the implications on model generalization and safety.   Our main contributions in this work can be summarized as follows:   A closer look into anchored training and inferencing: By studying the roles of reference set diversity and the inferencing protocol choice on",
            "references": [
                {
                    "title": "Smooth ECE: Principled Reliability Diagrams via Kernel Smoothing",
                    "abstract": "Calibration measures and reliability diagrams are two fundamental tools for measuring and interpreting the calibration of probabilistic predictors. Calibration measures quantify the degree of miscalibration, and reliability diagrams visualize the structure of this miscalibration. However, the most common constructions of reliability diagrams and calibration measures -- binning and ECE -- both suffer from well-known flaws (e.g. discontinuity). We show that a simple modification fixes both constructions: first smooth the observations using an RBF kernel, then compute the Expected Calibration Error (ECE) of this smoothed function. We prove that with a careful choice of bandwidth, this method yields a calibration measure that is well-behaved in the sense of (B{\\l}asiok, Gopalan, Hu, and Nakkiran 2023a) -- a consistent calibration measure. We call this measure the SmoothECE. Moreover, the reliability diagram obtained from this smoothed function visually encodes the SmoothECE, just as binned reliability diagrams encode the BinnedECE. We also provide a Python package with simple, hyperparameter-free methods for measuring and plotting calibration: `pip install relplot\\`."
                },
                {
                    "title": "Identifying and Mitigating the Security Risks of Generative AI",
                    "abstract": "Every major technical invention resurfaces the dual-use dilemma -- the new technology has the potential to be used for good as well as for harm. Generative AI (GenAI) techniques, such as large language models (LLMs) and diffusion models, have shown remarkable capabilities (e.g., in-context learning, code-completion, and text-to-image generation and editing). However, GenAI can be used just as well by attackers to generate new attacks and increase the velocity and efficacy of existing attacks. This paper reports the findings of a workshop held at Google (co-organized by Stanford University and the University of Wisconsin-Madison) on the dual-use dilemma posed by GenAI. This paper is not meant to be comprehensive, but is rather an attempt to synthesize some of the interesting findings from the workshop. We discuss short-term and long-term goals for the community on this topic. We hope this paper provides both a launching point for a discussion on this important topic as well as interesting problems that the research community can work to address."
                },
                {
                    "title": "In or Out? Fixing ImageNet Out-of-Distribution Detection Evaluation",
                    "abstract": "Out-of-distribution (OOD) detection is the problem of identifying inputs which are unrelated to the in-distribution task. The OOD detection performance when the in-distribution (ID) is ImageNet-1K is commonly being tested on a small range of test OOD datasets. We find that most of the currently used test OOD datasets, including datasets from the open set recognition (OSR) literature, have severe issues: In some cases more than 50$\\%$ of the dataset contains objects belonging to one of the ID classes. These erroneous samples heavily distort the evaluation of OOD detectors. As a solution, we introduce with NINCO a novel test OOD dataset, each sample checked to be ID free, which with its fine-grained range of OOD classes allows for a detailed analysis of an OOD detector's strengths and failure modes, particularly when paired with a number of synthetic\"OOD unit-tests\". We provide detailed evaluations across a large set of architectures and OOD detection methods on NINCO and the unit-tests, revealing new insights about model weaknesses and the effects of pretraining on OOD detection performance. We provide code and data at https://github.com/j-cb/NINCO."
                },
                {
                    "title": "Learning to Extrapolate: A Transductive Approach",
                    "abstract": "Machine learning systems, especially with overparameterized deep neural networks, can generalize to novel test instances drawn from the same distribution as the training data. However, they fare poorly when evaluated on out-of-support test points. In this work, we tackle the problem of developing machine learning systems that retain the power of overparameterized function approximators while enabling extrapolation to out-of-support test points when possible. This is accomplished by noting that under certain conditions, a\"transductive\"reparameterization can convert an out-of-support extrapolation problem into a problem of within-support combinatorial generalization. We propose a simple strategy based on bilinear embeddings to enable this type of combinatorial generalization, thereby addressing the out-of-support extrapolation problem under certain conditions. We instantiate a simple, practical algorithm applicable to various supervised learning and imitation learning tasks."
                },
                {
                    "title": "Finetune like you pretrain: Improved finetuning of zero-shot vision models",
                    "abstract": "Finetuning image-text models such as CLIP achieves state-of-the-art accuracies on a variety of benchmarks. However, recent works (Kumar et al., 2022; Wortsman et al., 2021) have shown that even subtle differences in the finetuning process can lead to surprisingly large differences in the final performance, both for in-distribution (ID) and out-of-distribution (OOD) data. In this work, we show that a natural and simple approach of mimicking contrastive pretraining consistently outperforms alternative finetuning approaches. Specifically, we cast downstream class labels as text prompts and continue optimizing the contrastive loss between image embeddings and class-descriptive prompt embeddings (contrastive finetuning). Our method consistently outperforms baselines across 7 distribution shift, 6 transfer learning, and 3 few-shot learning benchmarks. On WILDS-iWILDCam, our proposed approach FLYP outperforms the top of the leaderboard by 2.3% ID and 2.7% OOD, giving the highest reported accuracy. Averaged across 7 OOD datasets (2 WILDS and 5 ImageNet associated shifts), FLYP gives gains of 4.2% OOD over standard finetuning and outperforms current state-of-the-art (LP-FT) by more than 1 % both ID and OOD. Similarly, on 3 few-shot learning benchmarks, FLYP gives gains up to 4.6% over standard finetuning and 4.4% over the state-of-the-art. Thus we establish our proposed method of contrastive finetuning as a simple and intuitive state-of-the-art for supervised finetuning of image-text models like CLIP. Code is available at https://github.com/locuslab/FLYP."
                },
                {
                    "title": "Single Model Uncertainty Estimation via Stochastic Data Centering",
                    "abstract": "We are interested in estimating the uncertainties of deep neural networks, which play an important role in many scientific and engineering problems. In this paper, we present a striking new finding that an ensemble of neural networks with the same weight initialization, trained on datasets that are shifted by a constant bias gives rise to slightly inconsistent trained models, where the differences in predictions are a strong indicator of epistemic uncertainties. Using the neural tangent kernel (NTK), we demonstrate that this phenomena occurs in part because the NTK is not shift-invariant. Since this is achieved via a trivial input transformation, we show that this behavior can therefore be approximated by training a single neural network -- using a technique that we call $\\Delta-$UQ -- that estimates uncertainty around prediction by marginalizing out the effect of the biases during inference. We show that $\\Delta-$UQ's uncertainty estimates are superior to many of the current methods on a variety of benchmarks -- outlier rejection, calibration under distribution shift, and sequential design optimization of black box functions. Code for $\\Delta-$UQ can be accessed at https://github.com/LLNL/DeltaUQ"
                },
                {
                    "title": "Out of Distribution Detection via Neural Network Anchoring",
                    "abstract": "Our goal in this paper is to exploit heteroscedastic temperature scaling as a calibration strategy for out of distribution (OOD) detection. Heteroscedasticity here refers to the fact that the optimal temperature parameter for each sample can be different, as opposed to conventional approaches that use the same value for the entire distribution. To enable this, we propose a new training strategy called anchoring that can estimate appropriate temperature values for each sample, leading to state-of-the-art OOD detection performance across several benchmarks. Using NTK theory, we show that this temperature function estimate is closely linked to the epistemic uncertainty of the classifier, which explains its behavior. In contrast to some of the best-performing OOD detection approaches, our method does not require exposure to additional outlier datasets, custom calibration objectives, or model ensembling. Through empirical studies with different OOD detection settings -- far OOD, near OOD, and semantically coherent OOD - we establish a highly effective OOD detection approach. Code to reproduce our results is available at github.com/LLNL/AMP"
                },
                {
                    "title": "Improving Diversity with Adversarially Learned Transformations for Domain Generalization",
                    "abstract": "To be successful in single source domain generalization (SSDG), maximizing diversity of synthesized domains has emerged as one of the most effective strategies. Recent success in SSDG comes from methods that pre-specify diversity inducing image augmentations during training, so that it may lead to better generalization on new domains. However, na\u00efve pre-specified augmentations are not always effective, either because they cannot model large domain shift, or be-cause the specific choice of transforms may not cover the types of shift commonly occurring in domain generalization. To address this issue, we present a novel framework called ALT: adversarially learned transformations, that uses an adversary neural network to model plausible, yet hard image transformations that fool the classifier. ALT learns image transformations by randomly initializing the adversary net-work for each batch and optimizing it for a fixed number of steps to maximize classification error. The classifier is trained by enforcing a consistency between its predictions on the clean and transformed images. With extensive empirical analysis, we find that this new form of adversarial transformations achieves both objectives of diversity and hardness simultaneously, outperforming all existing techniques on competitive benchmarks for SSDG. We also show that ALT can seamlessly work with existing diversity modules to produce highly distinct, and large transformations of the source domain leading to state-of-the-art performance. Code: https://github.com/tejas-gokhale/ALT"
                },
                {
                    "title": "Anomaly Detection in Autonomous Driving: A Survey",
                    "abstract": "Nowadays, there are outstanding strides towards a future with autonomous vehicles on our roads. While the perception of autonomous vehicles performs well under closed-set conditions, they still struggle to handle the unexpected. This survey provides an extensive overview of anomaly detection techniques based on camera, lidar, radar, multimodal and abstract object level data. We provide a systematization including detection approach, corner case level, ability for an online application, and further attributes. We outline the state-of-the-art and point out current research gaps."
                },
                {
                    "title": "Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time",
                    "abstract": "The conventional recipe for maximizing model accuracy is to (1) train multiple models with various hyperparameters and (2) pick the individual model which performs best on a held-out validation set, discarding the remainder. In this paper, we revisit the second step of this procedure in the context of fine-tuning large pre-trained models, where fine-tuned models often appear to lie in a single low error basin. We show that averaging the weights of multiple models fine-tuned with different hyperparameter configurations often improves accuracy and robustness. Unlike a conventional ensemble, we may average many models without incurring any additional inference or memory costs -- we call the results\"model soups.\"When fine-tuning large pre-trained models such as CLIP, ALIGN, and a ViT-G pre-trained on JFT, our soup recipe provides significant improvements over the best model in a hyperparameter sweep on ImageNet. The resulting ViT-G model, which attains 90.94% top-1 accuracy on ImageNet, achieved a new state of the art. Furthermore, we show that the model soup approach extends to multiple image classification and natural language processing tasks, improves out-of-distribution performance, and improves zero-shot performance on new downstream tasks. Finally, we analytically relate the performance similarity of weight-averaging and logit-ensembling to flatness of the loss and confidence of the predictions, and validate this relation empirically. Code is available at https://github.com/mlfoundations/model-soups."
                },
                {
                    "title": "PixMix: Dreamlike Pictures Comprehensively Improve Safety Measures",
                    "abstract": "In real-world applications of machine learning, reliable and safe systems must consider measures of performance beyond standard test set accuracy. These other goals include out-of-distribution (OOD) robustness, prediction consistency, resilience to adversaries, calibrated uncertainty estimates, and the ability to detect anomalous inputs. However, improving performance towards these goals is often a balancing act that today's methods cannot achieve without sacrificing performance on other safety axes. For instance, adversarial training improves adversarial robustness but sharply degrades other classifier performance metrics. Similarly, strong data augmentation and regularization techniques often improve OOD robustness but harm anomaly detection, raising the question of whether a Pareto improvement on all existing safety measures is possible. To meet this challenge, we design a new data augmentation strategy utilizing the natural structural complexity of pictures such as fractals, which outperforms numerous baselines, is near Pareto-optimal, and roundly improves safety measures."
                },
                {
                    "title": "Swin Transformer V2: Scaling Up Capacity and Resolution",
                    "abstract": "We present techniques for scaling Swin Transformer [35] up to 3 billion parameters and making it capable of training with images of up to 1,536x1,536 resolution. By scaling up capacity and resolution, Swin Transformer sets new records on four representative vision benchmarks: 84.0% top-1 accuracy on ImageNet- V2 image classification, 63.1 / 54.4 box / mask mAP on COCO object detection, 59.9 mIoU on ADE20K semantic segmentation, and 86.8% top-1 accuracy on Kinetics-400 video action classification. We tackle issues of training instability, and study how to effectively transfer models pre-trained at low resolutions to higher resolution ones. To this aim, several novel technologies are proposed: 1) a residual post normalization technique and a scaled cosine attention approach to improve the stability of large vision models; 2) a log-spaced continuous position bias technique to effectively transfer models pre-trained at low-resolution images and windows to their higher-resolution counterparts. In addition, we share our crucial implementation details that lead to significant savings of GPU memory consumption and thus make it feasi-ble to train large vision models with regular GPUs. Using these techniques and self-supervised pre-training, we suc-cessfully train a strong 3 billion Swin Transformer model and effectively transfer it to various vision tasks involving high-resolution images or windows, achieving the state-of-the-art accuracy on a variety of benchmarks. Code is avail-able at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "Unsolved Problems in ML Safety",
                    "abstract": "Machine learning (ML) systems are rapidly increasing in size, are acquiring new capabilities, and are increasingly deployed in high-stakes settings. As with other powerful technologies, safety for ML should be a leading research priority. In response to emerging safety challenges in ML, such as those introduced by recent large-scale models, we provide a new roadmap for ML Safety and refine the technical problems that the field needs to address. We present four problems ready for research, namely withstanding hazards (\"Robustness\"), identifying hazards (\"Monitoring\"), reducing inherent model hazards (\"Alignment\"), and reducing systemic hazards (\"Systemic Safety\"). Throughout, we clarify each problem's motivation and provide concrete research directions."
                },
                {
                    "title": "The Evolution of Out-of-Distribution Robustness Throughout Fine-Tuning",
                    "abstract": "Although machine learning models typically experience a drop in performance on out-of-distribution data, accuracies on in- versus out-of-distribution data are widely observed to follow a single linear trend when evaluated across a testbed of models. Models that are more accurate on the out-of-distribution data relative to this baseline exhibit\"effective robustness\"and are exceedingly rare. Identifying such models, and understanding their properties, is key to improving out-of-distribution performance. We conduct a thorough empirical investigation of effective robustness during fine-tuning and surprisingly find that models pre-trained on larger datasets exhibit effective robustness during training that vanishes at convergence. We study how properties of the data influence effective robustness, and we show that it increases with the larger size, more diversity, and higher example difficulty of the dataset. We also find that models that display effective robustness are able to correctly classify 10% of the examples that no other current testbed model gets correct. Finally, we discuss several strategies for scaling effective robustness to the high-accuracy regime to improve the out-of-distribution accuracy of state-of-the-art models."
                },
                {
                    "title": "TrivialAugment: Tuning-free Yet State-of-the-Art Data Augmentation",
                    "abstract": "Automatic augmentation methods have recently become a crucial pillar for strong model performance in vision tasks. While existing automatic augmentation methods need to trade off simplicity, cost and performance, we present a most simple baseline, TrivialAugment, that outperforms previous methods for almost free. TrivialAugment is parameter-free and only applies a single augmentation to each image. Thus, TrivialAugment's effectiveness is very unexpected to us and we performed very thorough experiments to study its performance. First, we compare TrivialAugment to previous state-of-the-art methods in a variety of image classification scenarios. Then, we perform multiple ablation studies with different augmentation spaces, augmentation methods and setups to understand the crucial requirements for its performance. Additionally, we provide a simple interface to facilitate the widespread adoption of automatic augmentation methods, as well as our full code base for reproducibility1. Since our work reveals a stagnation in many parts of automatic augmentation research, we end with a short proposal of best practices for sustained future progress in automatic augmentation methods."
                },
                {
                    "title": "On Interaction Between Augmentations and Corruptions in Natural Corruption Robustness",
                    "abstract": "Invariance to a broad array of image corruptions, such as warping, noise, or color shifts, is an important aspect of building robust models in computer vision. Recently, several new data augmentations have been proposed that significantly improve performance on ImageNet-C, a benchmark of such corruptions. However, there is still a lack of basic understanding on the relationship between data augmentations and test-time corruptions. To this end, we develop a feature space for image transforms, and then use a new measure in this space between augmentations and corruptions called the Minimal Sample Distance to demonstrate a strong correlation between similarity and performance. We then investigate recent data augmentations and observe a significant degradation in corruption robustness when the test-time corruptions are sampled to be perceptually dissimilar from ImageNet-C in this feature space. Our results suggest that test error can be improved by training on perceptually similar augmentations, and data augmentations may not generalize well beyond the existing benchmark. We hope our results and tools will allow for more robust progress towards improving robustness to image corruptions. We provide code at https://github.com/facebookresearch/augmentation-corruption."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Energy-based Out-of-distribution Detection",
                    "abstract": "Determining whether inputs are out-of-distribution (OOD) is an essential building block for safely deploying machine learning models in the open world. However, previous methods relying on the softmax confidence score suffer from overconfident posterior distributions for OOD data. We propose a unified framework for OOD detection that uses an energy score. We show that energy scores better distinguish in- and out-of-distribution samples than the traditional approach using the softmax scores. Unlike softmax confidence scores, energy scores are theoretically aligned with the probability density of the inputs and are less susceptible to the overconfidence issue. Within this framework, energy can be flexibly used as a scoring function for any pre-trained neural classifier as well as a trainable cost function to shape the energy surface explicitly for OOD detection. On a CIFAR-10 pre-trained WideResNet, using the energy score reduces the average FPR (at TPR 95%) by 18.03% compared to the softmax confidence score. With energy-based training, our method outperforms the state-of-the-art on common benchmarks."
                },
                {
                    "title": "Robust and Generalizable Visual Representation Learning via Random Convolutions",
                    "abstract": "While successful for various computer vision tasks, deep neural networks have shown to be vulnerable to texture style shifts and small perturbations to which humans are robust. Hence, our goal is to train models in such a way that improves their robustness to these perturbations. We are motivated by the approximately shape-preserving property of randomized convolutions, which is due to distance preservation under random linear transforms. Intuitively, randomized convolutions create an infinite number of new domains with similar object shapes but random local texture. Therefore, we explore using outputs of multi-scale random convolutions as new images or mixing them with the original images during training. When applying a network trained with our approach to unseen domains, our method consistently improves the performance on domain generalization benchmarks and is scalable to ImageNet. Especially for the challenging scenario of generalizing to the sketch domain in PACS and to ImageNet-Sketch, our method outperforms state-of-art methods by a large margin. More interestingly, our method can benefit downstream tasks by providing a more robust pretrained visual representation."
                },
                {
                    "title": "The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization",
                    "abstract": "We introduce four new real-world distribution shift datasets consisting of changes in image style, image blurriness, geographic location, camera operation, and more. With our new datasets, we take stock of previously proposed methods for improving out-of-distribution robustness and put them to the test. We find that using larger models and artificial data augmentations can improve robustness on real-world distribution shifts, contrary to claims in prior work. We find improvements in artificial robustness benchmarks can transfer to real-world distribution shifts, contrary to claims in prior work. Motivated by our observation that data augmentations can help with real-world distribution shifts, we also introduce a new data augmentation method which advances the state-of-the-art and outperforms models pre-trained with 1000\u00d7 more labeled data. Overall we find that some methods consistently help with distribution shifts in texture and local image statistics, but these methods do not help with some other distribution shifts like geographic changes. Our results show that future research must study multiple distribution shifts simultaneously, as we demonstrate that no evaluated method consistently improves robustness."
                },
                {
                    "title": "AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty",
                    "abstract": "Modern deep neural networks can achieve high accuracy when the training distribution and test distribution are identically distributed, but this assumption is frequently violated in practice. When the train and test distributions are mismatched, accuracy can plummet. Currently there are few techniques that improve robustness to unforeseen data shifts encountered during deployment. In this work, we propose a technique to improve the robustness and uncertainty estimates of image classifiers. We propose AugMix, a data processing technique that is simple to implement, adds limited computational overhead, and helps models withstand unforeseen corruptions. AugMix significantly improves robustness and uncertainty measures on challenging image classification benchmarks, closing the gap between previous methods and the best possible performance in some cases by more than half."
                },
                {
                    "title": "Randaugment: Practical automated data augmentation with a reduced search space",
                    "abstract": "Recent work on automated augmentation strategies has led to state-of-the-art results in image classification and object detection. An obstacle to a large-scale adoption of these methods is that they require a separate and expensive search phase. A common way to overcome the expense of the search phase was to use a smaller proxy task. However, it was not clear if the optimized hyperparameters found on the proxy task are also optimal for the actual task. In this work, we rethink the process of designing automated augmentation strategies. We find that while previous work required a search for both magnitude and probability of each operation independently, it is sufficient to only search for a single distortion magnitude that jointly controls all operations. We hence propose a simplified search space that vastly reduces the computational expense of automated augmentation, and permits the removal of a separate proxy task.Despite the simplifications, our method achieves equal or better performance over previous automated augmentation strategies on on CIFAR-10/100, SVHN, ImageNet and COCO datasets. EfficientNet-B7, we achieve 85.0% accuracy, a 1.0% increase over baseline augmentation, a 0.6% improvement over AutoAugment on the ImageNet dataset. With EfficientNet-B8, we achieve 85.4% accuracy on ImageNet, which matches a previous result that used 3.5B extra images. On object detection, the same method as classification leads to 1.0-1.3% improvement over baseline augmentation. Code will be made available online."
                },
                {
                    "title": "The potential for artificial intelligence in healthcare",
                    "abstract": "ABSTRACT The complexity and rise of data in healthcare means that artificial intelligence (AI) will increasingly be applied within the field. Several types of AI are already being employed by payers and providers of care, and life sciences companies. The key categories of applications involve diagnosis and treatment recommendations, patient engagement and adherence, and administrative activities. Although there are many instances in which AI can perform healthcare tasks as well or better than humans, implementation factors will prevent large-scale automation of healthcare professional jobs for a considerable period. Ethical issues in the application of AI to healthcare are also discussed."
                },
                {
                    "title": "Learning Robust Global Representations by Penalizing Local Predictive Power",
                    "abstract": "Despite their renowned predictive power on i.i.d. data, convolutional neural networks are known to rely more on high-frequency patterns that humans deem superficial than on low-frequency patterns that agree better with intuitions about what constitutes category membership. This paper proposes a method for training robust convolutional networks by penalizing the predictive power of the local representations learned by earlier layers. Intuitively, our networks are forced to discard predictive signals such as color and texture that can be gleaned from local receptive fields and to rely instead on the global structures of the image. Across a battery of synthetic and benchmark domain adaptation tasks, our method confers improved generalization out of the domain. Also, to evaluate cross-domain transfer, we introduce ImageNet-Sketch, a new dataset consisting of sketch-like images, that matches the ImageNet classification validation set in categories and scale."
                },
                {
                    "title": "Benchmarking Neural Network Robustness to Common Corruptions and Perturbations",
                    "abstract": "In this paper we establish rigorous benchmarks for image classifier robustness. Our first benchmark, ImageNet-C, standardizes and expands the corruption robustness topic, while showing which classifiers are preferable in safety-critical applications. Then we propose a new dataset called ImageNet-P which enables researchers to benchmark a classifier's robustness to common perturbations. Unlike recent robustness research, this benchmark evaluates performance on common corruptions and perturbations not worst-case adversarial perturbations. We find that there are negligible changes in relative corruption robustness from AlexNet classifiers to ResNet classifiers. Afterward we discover ways to enhance corruption and perturbation robustness. We even find that a bypassed adversarial defense provides substantial common perturbation robustness. Together our benchmarks may aid future work toward networks that robustly generalize."
                },
                {
                    "title": "Moment Matching for Multi-Source Domain Adaptation",
                    "abstract": "Conventional unsupervised domain adaptation (UDA) assumes that training data are sampled from a single domain. This neglects the more practical scenario where training data are collected from multiple sources, requiring multi-source domain adaptation. We make three major contributions towards addressing this problem. First, we collect and annotate by far the largest UDA dataset, called DomainNet, which contains six domains and about 0.6 million images distributed among 345 categories, addressing the gap in data availability for multi-source UDA research. Second, we propose a new deep learning approach, Moment Matching for Multi-Source Domain Adaptation (M3SDA), which aims to transfer knowledge learned from multiple labeled source domains to an unlabeled target domain by dynamically aligning moments of their feature distributions. Third, we provide new theoretical insights specifically for moment matching approaches in both single and multiple source domain adaptation. Extensive experiments are conducted to demonstrate the power of our new dataset in benchmarking state-of-the-art multi-source domain adaptation methods, as well as the advantage of our proposed model. Dataset and Code are available at http://ai.bu.edu/M3SDA/"
                },
                {
                    "title": "Deep Anomaly Detection with Outlier Exposure",
                    "abstract": "It is important to detect anomalous inputs when deploying machine learning systems. The use of larger and more complex inputs in deep learning magnifies the difficulty of distinguishing between anomalous and in-distribution examples. At the same time, diverse image and text data are available in enormous quantities. We propose leveraging these data to improve deep anomaly detection by training anomaly detectors against an auxiliary dataset of outliers, an approach we call Outlier Exposure (OE). This enables anomaly detectors to generalize and detect unseen anomalies. In extensive experiments on natural language processing and small- and large-scale vision tasks, we find that Outlier Exposure significantly improves detection performance. We also observe that cutting-edge generative models trained on CIFAR-10 may assign higher likelihoods to SVHN images than to CIFAR-10 images; we use OE to mitigate this issue. We also analyze the flexibility and robustness of Outlier Exposure, and identify characteristics of the auxiliary dataset that improve performance."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "Places: A 10 Million Image Database for Scene Recognition",
                    "abstract": "The rise of multi-million-item dataset initiatives has enabled data-hungry machine learning algorithms to reach near-human semantic classification performance at tasks such as visual object and scene recognition. Here we describe the Places Database, a repository of 10 million scene photographs, labeled with scene semantic categories, comprising a large and diverse list of the types of environments encountered in the world. Using the state-of-the-art Convolutional Neural Networks (CNNs), we provide scene classification CNNs (Places-CNNs) as baselines, that significantly outperform the previous approaches. Visualization of the CNNs trained on Places shows that object detectors emerge as an intermediate representation of scene classification. With its high-coverage and high-diversity of exemplars, the Places Database along with the Places-CNNs offer a novel resource to guide future progress on scene recognition problems."
                },
                {
                    "title": "Averaging Weights Leads to Wider Optima and Better Generalization",
                    "abstract": "Deep neural networks are typically trained by optimizing a loss function with an SGD variant, in conjunction with a decaying learning rate, until convergence. We show that simple averaging of multiple points along the trajectory of SGD, with a cyclical or constant learning rate, leads to better generalization than conventional training. We also show that this Stochastic Weight Averaging (SWA) procedure finds much flatter solutions than SGD, and approximates the recent Fast Geometric Ensembling (FGE) approach with a single model. Using SWA we achieve notable improvement in test accuracy over conventional SGD training on a range of state-of-the-art residual networks, PyramidNets, DenseNets, and Shake-Shake networks on CIFAR-10, CIFAR-100, and ImageNet. In short, SWA is extremely easy to implement, improves generalization, and has almost no computational overhead."
                },
                {
                    "title": "Loss Surfaces, Mode Connectivity, and Fast Ensembling of DNNs",
                    "abstract": "The loss functions of deep neural networks are complex and their geometric properties are not well understood. We show that the optima of these complex loss functions are in fact connected by simple curves over which training and test accuracy are nearly constant. We introduce a training procedure to discover these high-accuracy pathways between modes. Inspired by this new geometric insight, we also propose a new ensembling method entitled Fast Geometric Ensembling (FGE). Using FGE we can train high-performing ensembles in the time required to train a single model. We achieve improved performance compared to the recent state-of-the-art Snapshot Ensembles, on CIFAR-10, CIFAR-100, and ImageNet."
                },
                {
                    "title": "Visualizing the Loss Landscape of Neural Nets",
                    "abstract": "Neural network training relies on our ability to find \"good\" minimizers of highly non-convex loss functions. It is well-known that certain network architecture designs (e.g., skip connections) produce loss functions that train easier, and well-chosen training parameters (batch size, learning rate, optimizer) produce minimizers that generalize better. However, the reasons for these differences, and their effects on the underlying loss landscape, are not well understood. In this paper, we explore the structure of neural loss functions, and the effect of loss landscapes on generalization, using a range of visualization methods. First, we introduce a simple \"filter normalization\" method that helps us visualize loss function curvature and make meaningful side-by-side comparisons between loss functions. Then, using a variety of visualizations, we explore how network architecture affects the loss landscape, and how training parameters affect the shape of minimizers."
                },
                {
                    "title": "On Calibration of Modern Neural Networks",
                    "abstract": "Confidence calibration -- the problem of predicting probability estimates representative of the true correctness likelihood -- is important for classification models in many applications. We discover that modern neural networks, unlike those from a decade ago, are poorly calibrated. Through extensive experiments, we observe that depth, width, weight decay, and Batch Normalization are important factors influencing calibration. We evaluate the performance of various post-processing calibration methods on state-of-the-art architectures with image and document classification datasets. Our analysis and experiments not only offer insights into neural network learning, but also provide a simple and straightforward recipe for practical settings: on most datasets, temperature scaling -- a single-parameter variant of Platt Scaling -- is surprisingly effective at calibrating predictions."
                },
                {
                    "title": "A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks",
                    "abstract": "We consider the two related problems of detecting if an example is misclassified or out-of-distribution. We present a simple baseline that utilizes probabilities from softmax distributions. Correctly classified examples tend to have greater maximum softmax probabilities than erroneously classified and out-of-distribution examples, allowing for their detection. We assess performance by defining several tasks in computer vision, natural language processing, and automatic speech recognition, showing the effectiveness of this baseline across all. We then show the baseline can sometimes be surpassed, demonstrating the room for future research on these underexplored detection tasks."
                },
                {
                    "title": "Wide Residual Networks",
                    "abstract": "Deep residual networks were shown to be able to scale up to thousands of layers and still have improving performance. However, each fraction of a percent of improved accuracy costs nearly doubling the number of layers, and so training very deep residual networks has a problem of diminishing feature reuse, which makes these networks very slow to train. To tackle these problems, in this paper we conduct a detailed experimental study on the architecture of ResNet blocks, based on which we propose a novel architecture where we decrease depth and increase width of residual networks. We call the resulting network structures wide residual networks (WRNs) and show that these are far superior over their commonly used thin and very deep counterparts. For example, we demonstrate that even a simple 16-layer-deep wide residual network outperforms in accuracy and efficiency all previous deep residual networks, including thousand-layer-deep networks, achieving new state-of-the-art results on CIFAR, SVHN, COCO, and significant improvements on ImageNet. Our code and models are available at this https URL"
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "3D Object Representations for Fine-Grained Categorization",
                    "abstract": "While 3D object representations are being revived in the context of multi-view object class detection and scene understanding, they have not yet attained wide-spread use in fine-grained categorization. State-of-the-art approaches achieve remarkable performance when training data is plentiful, but they are typically tied to flat, 2D representations that model objects as a collection of unconnected views, limiting their ability to generalize across viewpoints. In this paper, we therefore lift two state-of-the-art 2D object representations to 3D, on the level of both local feature appearance and location. In extensive experiments on existing and newly proposed datasets, we show our 3D object representations outperform their state-of-the-art 2D counterparts for fine-grained categorization and demonstrate their efficacy for estimating 3D geometry from images via ultra-wide baseline matching and 3D reconstruction."
                },
                {
                    "title": "Describing Textures in the Wild",
                    "abstract": "Patterns and textures are key characteristics of many natural objects: a shirt can be striped, the wings of a butterfly can be veined, and the skin of an animal can be scaly. Aiming at supporting this dimension in image understanding, we address the problem of describing textures with semantic attributes. We identify a vocabulary of forty-seven texture terms and use them to describe a large dataset of patterns collected \"in the wild\". The resulting Describable Textures Dataset (DTD) is a basis to seek the best representation for recognizing describable texture attributes in images. We port from object recognition to texture recognition the Improved Fisher Vector (IFV) and Deep Convolutional-network Activation Features (DeCAF), and show that surprisingly, they both outperform specialized texture descriptors not only on our problem, but also in established material recognition datasets. We also show that our describable attributes are excellent texture descriptors, transferring between datasets and tasks, in particular, combined with IFV and DeCAF, they significantly outperform the state-of-the-art by more than 10% on both FMD and KTH-TIPS-2b benchmarks. We also demonstrate that they produce intuitive descriptions of materials and Internet images."
                },
                {
                    "title": "UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild",
                    "abstract": "We introduce UCF101 which is currently the largest dataset of human actions. It consists of 101 action classes, over 13k clips and 27 hours of video data. The database consists of realistic user uploaded videos containing camera motion and cluttered background. Additionally, we provide baseline action recognition results on this new dataset using standard bag of words approach with overall performance of 44.5%. To the best of our knowledge, UCF101 is currently the most challenging dataset of actions due to its large number of classes, large number of clips and also unconstrained nature of such clips."
                },
                {
                    "title": "SUN database: Large-scale scene recognition from abbey to zoo",
                    "abstract": "Scene categorization is a fundamental problem in computer vision. However, scene understanding research has been constrained by the limited scope of currently-used databases which do not capture the full variety of scene categories. Whereas standard databases for object categorization contain hundreds of different classes of objects, the largest available dataset of scene categories contains only 15 classes. In this paper we propose the extensive Scene UNderstanding (SUN) database that contains 899 categories and 130,519 images. We use 397 well-sampled categories to evaluate numerous state-of-the-art algorithms for scene recognition and establish new bounds of performance. We measure human scene classification performance on the SUN database and compare this with computational methods. Additionally, we study a finer-grained scene representation to detect scenes embedded inside of larger scenes."
                },
                {
                    "title": "Automated Flower Classification over a Large Number of Classes",
                    "abstract": "We investigate to what extent combinations of features can improve classification performance on a large dataset of similar classes. To this end we introduce a 103 class flower dataset. We compute four different features for the flowers, each describing different aspects, namely the local shape/texture, the shape of the boundary, the overall spatial distribution of petals, and the colour. We combine the features using a multiple kernel framework with a SVM classifier. The weights for each class are learnt using the method of Varma and Ray, which has achieved state of the art performance on other large dataset, such as Caltech 101/256. Our dataset has a similar challenge in the number of classes, but with the added difficulty of large between class similarity and small within class similarity. Results show that learning the optimum kernel combination of multiple features vastly improves the performance, from 55.1% for the best single feature to 72.8% for the combination of all features."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CV",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can the anchoring principle be effectively utilized to enhance the training and inferencing of vision models across different architectures, while improving model generalization and safety?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the need for more robust and reliable machine learning models, particularly in critical applications where safety and generalization are paramount. By systematically exploring the anchoring principle, this research could lead to significant advancements in model calibration, extrapolation properties, and uncertainty estimation, thereby influencing future research directions and practical applications in fields such as autonomous systems, healthcare, and finance.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the complexity of effectively integrating the anchoring principle into various model architectures, as well as understanding the interplay between reference sample diversity and inferencing protocols. Naive approaches may fail because they do not account for the nuanced relationships between reference samples and residuals, which are critical for accurate predictions. Additionally, technical obstacles such as ensuring consistent model performance across diverse datasets and maintaining computational efficiency pose significant hurdles.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often overlooked the systematic exploration of anchoring as a training protocol, focusing instead on isolated aspects of model performance. Limitations in understanding the joint distribution of reference samples and residuals, as well as a lack of comprehensive studies on the implications of anchoring for different architectures, have hindered progress. This research aims to fill these gaps by providing a detailed analysis of the anchoring principle's effects on model generalization and safety, thus offering a more holistic approach compared to prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves implementing the anchoring principle in various vision model architectures (e.g., convolutional networks and transformers) and evaluating its impact on model performance. The study will utilize diverse datasets to assess the effects of reference sample diversity and inferencing protocols on model predictions. Key metrics for evaluation will include calibration accuracy, extrapolation capabilities, and uncertainty estimation. The expected outcomes include improved model generalization, enhanced safety features, and a deeper understanding of the anchoring principle's utility in machine learning."
            }
        },
        "author_data": {
            "5b6b0cf0-7c66-484a-8bd2-30e1126f0408": {
                "pk": "5b6b0cf0-7c66-484a-8bd2-30e1126f0408",
                "name": "Vivek Narayanaswamy",
                "collaborators": [
                    "Jayaraman J. Thiagarajan",
                    "Andreas Spanias",
                    "Rushil Anirudh",
                    "Deepta Rajan",
                    "Peer-Timo Bremer",
                    "Puja Trivedi",
                    "Rakshith Subramanyam",
                    "Kowshik Thopalli",
                    "Yamen Mubarka",
                    "Grace Billingsley"
                ],
                "domain": [
                    "Deep Learning",
                    "Uncertainty Estimation",
                    "Image Processing",
                    "Audio Source Separation"
                ],
                "publications": [
                    {
                        "title": "Using Deep Image Priors to Generate Counterfactual Explanations",
                        "abstract": "Through the use of carefully tailored convolutional neural network architectures, a deep image prior (DIP) can be used to obtain pre-images from latent representation encodings. Though DIP inversion has been known to be superior to conventional regularized inversion strategies such as total variation, such an over-parameterized generator is able to effectively reconstruct even images that are not in the original data distribution. This limitation makes it challenging to utilize such priors for tasks such as counterfactual reasoning, wherein the goal is to generate small, interpretable changes to an image that systematically leads to changes in the model prediction. To this end, we propose a novel regularization strategy based on an auxiliary loss estimator jointly trained with the predictor, which efficiently guides the prior to recover natural pre-images. Our empirical studies with a real-world ISIC skin lesion detection problem clearly evidence the effectiveness of the proposed approach in synthesizing meaningful counterfactuals. In comparison, we find that the standard DIP inversion often proposes visually imperceptible perturbations to irrelevant parts of the image, thus providing no additional insights into the model behavior."
                    },
                    {
                        "title": "Unsupervised Audio Source Separation using Generative Priors",
                        "abstract": "State-of-the-art under-determined audio source separation systems rely on supervised end-end training of carefully tailored neural network architectures operating either in the time or the spectral domain. However, these methods are severely challenged in terms of requiring access to expensive source level labeled data and being specific to a given set of sources and the mixing process, which demands complete re-training when those assumptions change. This strongly emphasizes the need for unsupervised methods that can leverage the recent advances in data-driven modeling, and compensate for the lack of labeled data through meaningful priors. To this end, we propose a novel approach for audio source separation based on generative priors trained on individual sources. Through the use of projected gradient descent optimization, our approach simultaneously searches in the source-specific latent spaces to effectively recover the constituent sources. Though the generative priors can be defined in the time domain directly, e.g. WaveGAN, we find that using spectral domain loss functions for our optimization leads to good-quality source estimates. Our empirical studies on standard spoken digit and instrument datasets clearly demonstrate the effectiveness of our approach over classical as well as state-of-the-art unsupervised baselines."
                    },
                    {
                        "title": "Loss Estimators Improve Model Generalization",
                        "abstract": "With increased interest in adopting AI methods for clinical diagnosis, a vital step towards safe deployment of such tools is to ensure that the models not only produce accurate predictions but also do not generalize to data regimes where the training data provide no meaningful evidence. Existing approaches for ensuring the distribution of model predictions to be similar to that of the true distribution rely on explicit uncertainty estimators that are inherently hard to calibrate. In this paper, we propose to train a loss estimator alongside the predictive model, using a contrastive training objective, to directly estimate the prediction uncertainties. Interestingly, we find that, in addition to producing well-calibrated uncertainties, this approach improves the generalization behavior of the predictor. Using a dermatology use-case, we show the impact of loss estimators on model generalization, in terms of both its fidelity on in-distribution data and its ability to detect out of distribution samples or new classes unseen during training."
                    },
                    {
                        "title": "PAGER: A Framework for Failure Analysis of Deep Regression Models",
                        "abstract": "Safe deployment of AI models requires proactive detection of failures to prevent costly errors. To this end, we study the important problem of detecting failures in deep regression models. Existing approaches rely on epistemic uncertainty estimates or inconsistency w.r.t the training data to identify failure. Interestingly, we find that while uncertainties are necessary they are insufficient to accurately characterize failure in practice. Hence, we introduce PAGER (Principled Analysis of Generalization Errors in Regressors), a framework to systematically detect and characterize failures in deep regressors. Built upon the principle of anchored training in deep models, PAGER unifies both epistemic uncertainty and complementary manifold non-conformity scores to accurately organize samples into different risk regimes."
                    },
                    {
                        "title": "DECIDER: Leveraging Foundation Model Priors for Improved Model Failure Detection and Explanation",
                        "abstract": "Reliably detecting when a deployed machine learning model is likely to fail on a given input is crucial for ensuring safe operation. In this work, we propose DECIDER (Debiasing Classifiers to Identify Errors Reliably), a novel approach that leverages priors from large language models (LLMs) and vision-language models (VLMs) to detect failures in image classification models. DECIDER utilizes LLMs to specify task-relevant core attributes and constructs a ``debiased'' version of the classifier by aligning its visual features to these core attributes using a VLM, and detects potential failure by measuring disagreement between the original and debiased models. In addition to proactively identifying samples on which the model would fail, DECIDER also provides human-interpretable explanations for failure through a novel attribute-ablation strategy. Through extensive experiments across diverse benchmarks spanning subpopulation shifts (spurious correlations, class imbalance) and covariate shifts (synthetic corruptions, domain shifts), DECIDER consistently achieves state-of-the-art failure detection performance, significantly outperforming baselines in terms of the overall Matthews correlation coefficient as well as failure and success recall. Our codes can be accessed at~\\url{https://github.com/kowshikthopalli/DECIDER/}"
                    },
                    {
                        "title": "Know Your Space: Inlier and Outlier Construction for Calibrating Medical OOD Detectors",
                        "abstract": "We focus on the problem of producing well-calibrated out-of-distribution (OOD) detectors, in order to enable safe deployment of medical image classifiers. Motivated by the difficulty of curating suitable calibration datasets, synthetic augmentations have become highly prevalent for inlier/outlier specification. While there have been rapid advances in data augmentation techniques, this paper makes a striking finding that the space in which the inliers and outliers are synthesized, in addition to the type of augmentation, plays a critical role in calibrating OOD detectors. Using the popular energy-based OOD detection framework, we find that the optimal protocol is to synthesize latent-space inliers along with diverse pixel-space outliers. Based on empirical studies with multiple medical imaging benchmarks, we demonstrate that our approach consistently leads to superior OOD detection ($15\\% - 35\\%$ in AUROC) over the state-of-the-art in a variety of open-set recognition settings."
                    },
                    {
                        "title": "An L2-Normalized Spatial Attention Network For Accurate And Fast Classification Of Brain Tumors In 2D T1-Weighted CE-MRI Images",
                        "abstract": "We propose an accurate and fast classification network for classification of brain tumors in MRI images that outperforms all lightweight methods investigated in terms of accuracy. We test our model on a challenging 2D T1-weighted CE-MRI dataset containing three types of brain tumors: Meningioma, Glioma and Pituitary. We introduce an l2-normalized spatial attention mechanism that acts as a regularizer against overfitting during training. We compare our results against the state-of-the-art on this dataset and show that by integrating l2-normalized spatial attention into a baseline network we achieve a performance gain of 1.79 percentage points. Even better accuracy can be attained by combining our model in an ensemble with the pretrained VGG16 at the expense of execution speed. Our code is publicly available at https://github.com/juliadietlmeier/MRI_image_classification"
                    },
                    {
                        "title": "Accurate and Robust Feature Importance Estimation under Distribution Shifts",
                        "abstract": "With increasing reliance on the outcomes of black-box models in critical applications, post-hoc explainability tools that do not require access to the model internals are often used to enable humans understand and trust these models. In particular, we focus on the class of methods that can reveal the influence of input features on the predicted outputs. Despite their wide-spread adoption, existing methods are known to suffer from one or more of the following challenges: computational complexities, large uncertainties and most importantly, inability to handle real-world domain shifts. In this paper, we propose PRoFILE, a novel feature importance estimation method that addresses all these challenges. Through the use of a loss estimator jointly trained with the predictive model and a causal objective, PRoFILE can accurately estimate the feature importance scores even under complex distribution shifts, without any additional re-training. To this end, we also develop learning strategies for training the loss estimator, namely contrastive and dropout calibration, and find that it can effectively detect distribution shifts. Using empirical studies on several benchmark image and non-image data, we show significant improvements over state-of-the-art approaches, both in terms of fidelity and robustness."
                    },
                    {
                        "title": "Designing Counterfactual Generators using Deep Model Inversion",
                        "abstract": "Explanation techniques that synthesize small, interpretable changes to a given image while producing desired changes in the model prediction have become popular for introspecting black-box models. Commonly referred to as counterfactuals, the synthesized explanations are required to contain discernible changes (for easy interpretability) while also being realistic (consistency to the data manifold). In this paper, we focus on the case where we have access only to the trained deep classifier and not the actual training data. While the problem of inverting deep models to synthesize images from the training distribution has been explored, our goal is to develop a deep inversion approach to generate counterfactual explanations for a given query image. Despite their effectiveness in conditional image synthesis, we show that existing deep inversion methods are insufficient for producing meaningful counterfactuals. We propose DISC (Deep Inversion for Synthesizing Counterfactuals) that improves upon deep inversion by utilizing (a) stronger image priors, (b) incorporating a novel manifold consistency objective and (c) adopting a progressive optimization strategy. We find that, in addition to producing visually meaningful explanations, the counterfactuals from DISC are effective at learning classifier decision boundaries and are robust to unknown test-time corruptions."
                    },
                    {
                        "title": "Single Model Uncertainty Estimation via Stochastic Data Centering",
                        "abstract": "We are interested in estimating the uncertainties of deep neural networks, which play an important role in many scientific and engineering problems. In this paper, we present a striking new finding that an ensemble of neural networks with the same weight initialization, trained on datasets that are shifted by a constant bias gives rise to slightly inconsistent trained models, where the differences in predictions are a strong indicator of epistemic uncertainties. Using the neural tangent kernel (NTK), we demonstrate that this phenomena occurs in part because the NTK is not shift-invariant. Since this is achieved via a trivial input transformation, we show that this behavior can therefore be approximated by training a single neural network -- using a technique that we call $\\Delta-$UQ -- that estimates uncertainty around prediction by marginalizing out the effect of the biases during inference. We show that $\\Delta-$UQ's uncertainty estimates are superior to many of the current methods on a variety of benchmarks -- outlier rejection, calibration under distribution shift, and sequential design optimization of black box functions. Code for $\\Delta-$UQ can be accessed at https://github.com/LLNL/DeltaUQ"
                    }
                ]
            },
            "299820a2-3324-40b3-98c6-3de0d17af186": {
                "pk": "299820a2-3324-40b3-98c6-3de0d17af186",
                "name": "Kowshik Thopalli",
                "collaborators": [
                    "Jayaraman J. Thiagarajan",
                    "Pavan Turaga",
                    "Rushil Anirudh",
                    "Rakshith Subramanyam",
                    "Sameeksha Katoch",
                    "Vivek Narayanaswamy",
                    "Banooqa Banday",
                    "Tanzima Z. Islam",
                    "Yang Liu",
                    "Jayaraman Thiagarajan"
                ],
                "domain": [
                    "Unsupervised Learning",
                    "Domain Adaptation",
                    "Computer Vision",
                    "Meta-Learning"
                ],
                "publications": [
                    {
                        "title": "Multiple Subspace Alignment Improves Domain Adaptation",
                        "abstract": "We present a novel unsupervised domain adaptation (DA) method for cross-domain visual recognition. Though subspace methods have found success in DA, their performance is often limited due to the assumption of approximating an entire dataset using a single low-dimensional subspace. Instead, we develop a method to effectively represent the source and target datasets via a collection of low-dimensional subspaces, and subsequently align them by exploiting the natural geometry of the space of subspaces, on the Grassmann manifold. We demonstrate the effectiveness of this approach, using empirical studies on two widely used benchmarks, with state of the art domain adaptation performance"
                    },
                    {
                        "title": "Domain Alignment Meets Fully Test-Time Adaptation",
                        "abstract": "A foundational requirement of a deployed ML model is to generalize to data drawn from a testing distribution that is different from training. A popular solution to this problem is to adapt a pre-trained model to novel domains using only unlabeled data. In this paper, we focus on a challenging variant of this problem, where access to the original source data is restricted. While fully test-time adaptation (FTTA) and unsupervised domain adaptation (UDA) are closely related, the advances in UDA are not readily applicable to TTA, since most UDA methods require access to the source data. Hence, we propose a new approach, CATTAn, that bridges UDA and FTTA, by relaxing the need to access entire source data, through a novel deep subspace alignment strategy. With a minimal overhead of storing the subspace basis set for the source data, CATTAn enables unsupervised alignment between source and target data during adaptation. Through extensive experimental evaluation on multiple 2D and 3D vision benchmarks (ImageNet-C, Office-31, OfficeHome, DomainNet, PointDA-10) and model architectures, we demonstrate significant gains in FTTA performance. Furthermore, we make a number of crucial findings on the utility of the alignment objective even with inherently robust models, pre-trained ViT representations and under low sample availability in the target domain."
                    },
                    {
                        "title": "Target-Aware Generative Augmentations for Single-Shot Adaptation",
                        "abstract": "In this paper, we address the problem of adapting models from a source domain to a target domain, a task that has become increasingly important due to the brittle generalization of deep neural networks. While several test-time adaptation techniques have emerged, they typically rely on synthetic toolbox data augmentations in cases of limited target data availability. We consider the challenging setting of single-shot adaptation and explore the design of augmentation strategies. We argue that augmentations utilized by existing methods are insufficient to handle large distribution shifts, and hence propose a new approach SiSTA, which first fine-tunes a generative model from the source domain using a single-shot target, and then employs novel sampling strategies for curating synthetic target data. Using experiments on a variety of benchmarks, distribution shifts and image corruptions, we find that SiSTA produces significantly improved generalization over existing baselines in face attribute detection and multi-class object recognition. Furthermore, SiSTA performs competitively to models obtained by training on larger target datasets. Our codes can be accessed at https://github.com/Rakshith-2905/SiSTA."
                    },
                    {
                        "title": "DECIDER: Leveraging Foundation Model Priors for Improved Model Failure Detection and Explanation",
                        "abstract": "Reliably detecting when a deployed machine learning model is likely to fail on a given input is crucial for ensuring safe operation. In this work, we propose DECIDER (Debiasing Classifiers to Identify Errors Reliably), a novel approach that leverages priors from large language models (LLMs) and vision-language models (VLMs) to detect failures in image classification models. DECIDER utilizes LLMs to specify task-relevant core attributes and constructs a ``debiased'' version of the classifier by aligning its visual features to these core attributes using a VLM, and detects potential failure by measuring disagreement between the original and debiased models. In addition to proactively identifying samples on which the model would fail, DECIDER also provides human-interpretable explanations for failure through a novel attribute-ablation strategy. Through extensive experiments across diverse benchmarks spanning subpopulation shifts (spurious correlations, class imbalance) and covariate shifts (synthetic corruptions, domain shifts), DECIDER consistently achieves state-of-the-art failure detection performance, significantly outperforming baselines in terms of the overall Matthews correlation coefficient as well as failure and success recall. Our codes can be accessed at~\\url{https://github.com/kowshikthopalli/DECIDER/}"
                    },
                    {
                        "title": "On The Role of Prompt Construction In Enhancing Efficacy and Efficiency of LLM-Based Tabular Data Generation",
                        "abstract": "LLM-based data generation for real-world tabular data can be challenged by the lack of sufficient semantic context in feature names used to describe columns. We hypothesize that enriching prompts with domain-specific insights can improve both the quality and efficiency of data generation. To test this hypothesis, we explore three prompt construction protocols: Expert-guided, LLM-guided, and Novel-Mapping. Through empirical studies with the recently proposed GReaT framework, we find that context-enriched prompts lead to significantly improved data generation quality and training efficiency."
                    },
                    {
                        "title": "Speeding Up Image Classifiers with Little Companions",
                        "abstract": "Scaling up neural networks has been a key recipe to the success of large language and vision models. However, in practice, up-scaled models can be disproportionately costly in terms of computations, providing only marginal improvements in performance; for example, EfficientViT-L3-384 achieves <2% improvement on ImageNet-1K accuracy over the base L1-224 model, while requiring $14\\times$ more multiply-accumulate operations (MACs). In this paper, we investigate scaling properties of popular families of neural networks for image classification, and find that scaled-up models mostly help with \"difficult\" samples. Decomposing the samples by difficulty, we develop a simple model-agnostic two-pass Little-Big algorithm that first uses a light-weight \"little\" model to make predictions of all samples, and only passes the difficult ones for the \"big\" model to solve. Good little companion achieve drastic MACs reduction for a wide variety of model families and scales. Without loss of accuracy or modification of existing models, our Little-Big models achieve MACs reductions of 76% for EfficientViT-L3-384, 81% for EfficientNet-B7-600, 71% for DeiT3-L-384 on ImageNet-1K. Little-Big also speeds up the InternImage-G-512 model by 62% while achieving 90% ImageNet-1K top-1 accuracy, serving both as a strong baseline and as a simple practical method for large model compression."
                    },
                    {
                        "title": "Automated Domain Discovery from Multiple Sources to Improve Zero-Shot Generalization",
                        "abstract": "Domain generalization (DG) methods aim to develop models that generalize to settings where the test distribution is different from the training data. In this paper, we focus on the challenging problem of multi-source zero shot DG (MDG), where labeled training data from multiple source domains is available but with no access to data from the target domain. A wide range of solutions have been proposed for this problem, including the state-of-the-art multi-domain ensembling approaches. Despite these advances, the na\\\"ive ERM solution of pooling all source data together and training a single classifier is surprisingly effective on standard benchmarks. In this paper, we hypothesize that, it is important to elucidate the link between pre-specified domain labels and MDG performance, in order to explain this behavior. More specifically, we consider two popular classes of MDG algorithms -- distributional robust optimization (DRO) and multi-domain ensembles, in order to demonstrate how inferring custom domain groups can lead to consistent improvements over the original domain labels that come with the dataset. To this end, we propose (i) Group-DRO++, which incorporates an explicit clustering step to identify custom domains in an existing DRO technique; and (ii) DReaME, which produces effective multi-domain ensembles through implicit domain re-labeling with a novel meta-optimization algorithm. Using empirical studies on multiple standard benchmarks, we show that our variants consistently outperform ERM by significant margins (1.5% - 9%), and produce state-of-the-art MDG performance. Our code can be found at https://github.com/kowshikthopalli/DREAME"
                    },
                    {
                        "title": "Calibrate and Prune: Improving Reliability of Lottery Tickets Through Prediction Calibration",
                        "abstract": "The hypothesis that sub-network initializations (lottery) exist within the initializations of over-parameterized networks, which when trained in isolation produce highly generalizable models, has led to crucial insights into network initialization and has enabled efficient inferencing. Supervised models with uncalibrated confidences tend to be overconfident even when making wrong prediction. In this paper, for the first time, we study how explicit confidence calibration in the over-parameterized network impacts the quality of the resulting lottery tickets. More specifically, we incorporate a suite of calibration strategies, ranging from mixup regularization, variance-weighted confidence calibration to the newly proposed likelihood-based calibration and normalized bin assignment strategies. Furthermore, we explore different combinations of architectures and datasets, and make a number of key findings about the role of confidence calibration. Our empirical studies reveal that including calibration mechanisms consistently lead to more effective lottery tickets, in terms of accuracy as well as empirical calibration metrics, even when retrained using data with challenging distribution shifts with respect to the source dataset."
                    },
                    {
                        "title": "SALT: Subspace Alignment as an Auxiliary Learning Task for Domain Adaptation",
                        "abstract": "Unsupervised domain adaptation aims to transfer and adapt knowledge learned from a labeled source domain to an unlabeled target domain. Key components of unsupervised domain adaptation include: (a) maximizing performance on the target, and (b) aligning the source and target domains. Traditionally, these tasks have either been considered as separate, or assumed to be implicitly addressed together with high-capacity feature extractors. When considered separately, alignment is usually viewed as a problem of aligning data distributions, either through geometric approaches such as subspace alignment or through distributional alignment such as optimal transport. This paper represents a hybrid approach, where we assume simplified data geometry in the form of subspaces, and consider alignment as an auxiliary task to the primary task of maximizing performance on the source. The alignment is made rather simple by leveraging tractable data geometry in the form of subspaces. We synergistically allow certain parameters derived from the closed-form auxiliary solution, to be affected by gradients from the primary task. The proposed approach represents a unique fusion of geometric and model-based alignment with gradients from a data-driven primary task. Our approach termed SALT, is a simple framework that achieves comparable or sometimes outperforms state-of-the-art on multiple standard benchmarks."
                    },
                    {
                        "title": "MaAST: Map Attention with Semantic Transformersfor Efficient Visual Navigation",
                        "abstract": "Visual navigation for autonomous agents is a core task in the fields of computer vision and robotics. Learning-based methods, such as deep reinforcement learning, have the potential to outperform the classical solutions developed for this task; however, they come at a significantly increased computational load. Through this work, we design a novel approach that focuses on performing better or comparable to the existing learning-based solutions but under a clear time/computational budget. To this end, we propose a method to encode vital scene semantics such as traversable paths, unexplored areas, and observed scene objects -- alongside raw visual streams such as RGB, depth, and semantic segmentation masks -- into a semantically informed, top-down egocentric map representation. Further, to enable the effective use of this information, we introduce a novel 2-D map attention mechanism, based on the successful multi-layer Transformer networks. We conduct experiments on 3-D reconstructed indoor PointGoal visual navigation and demonstrate the effectiveness of our approach. We show that by using our novel attention schema and auxiliary rewards to better utilize scene semantics, we outperform multiple baselines trained with only raw inputs or implicit semantic information while operating with an 80% decrease in the agent's experience."
                    },
                    {
                        "title": "Revisiting Deep Subspace Alignment for Unsupervised Domain Adaptation",
                        "abstract": "Unsupervised domain adaptation (UDA) aims to transfer and adapt knowledge from a labeled source domain to an unlabeled target domain. Traditionally, subspace-based methods form an important class of solutions to this problem. Despite their mathematical elegance and tractability, these methods are often found to be ineffective at producing domain-invariant features with complex, real-world datasets. Motivated by the recent advances in representation learning with deep networks, this paper revisits the use of subspace alignment for UDA and proposes a novel adaptation algorithm that consistently leads to improved generalization. In contrast to existing adversarial training-based DA methods, our approach isolates feature learning and distribution alignment steps, and utilizes a primary-auxiliary optimization strategy to effectively balance the objectives of domain invariance and model fidelity. While providing a significant reduction in target data and computational requirements, our subspace-based DA performs competitively and sometimes even outperforms state-of-the-art approaches on several standard UDA benchmarks. Furthermore, subspace alignment leads to intrinsically well-regularized models that demonstrate strong generalization even in the challenging partial DA setting. Finally, the design of our UDA framework inherently supports progressive adaptation to new target domains at test-time, without requiring retraining of the model from scratch. In summary, powered by powerful feature learners and an effective optimization strategy, we establish subspace-based DA as a highly effective approach for visual recognition."
                    },
                    {
                        "title": "Perturbation Robust Representations of Topological Persistence Diagrams",
                        "abstract": "Topological methods for data analysis present opportunities for enforcing certain invariances of broad interest in computer vision, including view-point in activity analysis, articulation in shape analysis, and measurement invariance in non-linear dynamical modeling. The increasing success of these methods is attributed to the complementary information that topology provides, as well as availability of tools for computing topological summaries such as persistence diagrams. However, persistence diagrams are multi-sets of points and hence it is not straightforward to fuse them with features used for contemporary machine learning tools like deep-nets. In this paper we present theoretically well-grounded approaches to develop novel perturbation robust topological representations, with the long-term view of making them amenable to fusion with contemporary learning architectures. We term the proposed representation as Perturbed Topological Signatures, which live on a Grassmann manifold and hence can be efficiently used in machine learning pipelines. We explore the use of the proposed descriptor on three applications: 3D shape analysis, view-invariant activity analysis, and non-linear dynamical modeling. We show favorable results in both high-level recognition performance and time-complexity when compared to other baseline methods."
                    },
                    {
                        "title": "Single-Shot Domain Adaptation via Target-Aware Generative Augmentation",
                        "abstract": "The problem of adapting models from a source domain using data from any target domain of interest has gained prominence, thanks to the brittle generalization in deep neural networks. While several test-time adaptation techniques have emerged, they typically rely on synthetic data augmentations in cases of limited target data availability. In this paper, we consider the challenging setting of single-shot adaptation and explore the design of augmentation strategies. We argue that augmentations utilized by existing methods are insufficient to handle large distribution shifts, and hence propose a new approach SiSTA (Single-Shot Target Augmentations), which first fine-tunes a generative model from the source domain using a single-shot target, and then employs novel sampling strategies for curating synthetic target data. Using experiments with a state-of-the-art domain adaptation method, we find that SiSTA produces improvements as high as 20\\% over existing baselines under challenging shifts in face attribute detection, and that it performs competitively to oracle models obtained by training on a larger target dataset."
                    },
                    {
                        "title": "Invenio: Discovering Hidden Relationships Between Tasks/Domains Using Structured Meta Learning",
                        "abstract": "Exploiting known semantic relationships between fine-grained tasks is critical to the success of recent model agnostic approaches. These approaches often rely on meta-optimization to make a model robust to systematic task or domain shifts. However, in practice, the performance of these methods can suffer, when there are no coherent semantic relationships between the tasks (or domains). We present Invenio, a structured meta-learning algorithm to infer semantic similarities between a given set of tasks and to provide insights into the complexity of transferring knowledge between different tasks. In contrast to existing techniques such as Task2Vec and Taskonomy, which measure similarities between pre-trained models, our approach employs a novel self-supervised learning strategy to discover these relationships in the training loop and at the same time utilizes them to update task-specific models in the meta-update step. Using challenging task and domain databases, under few-shot learning settings, we show that Invenio can discover intricate dependencies between tasks or domains, and can provide significant gains over existing approaches in terms of generalization performance. The learned semantic structure between tasks/domains from Invenio is interpretable and can be used to construct meaningful priors for tasks or domains."
                    }
                ]
            },
            "1f645ced-4f78-4bfc-b370-d67914107878": {
                "pk": "1f645ced-4f78-4bfc-b370-d67914107878",
                "name": "Rushil Anirudh",
                "collaborators": [
                    "Jayaraman J. Thiagarajan",
                    "Pavan Turaga",
                    "Bhavya Kailkhura",
                    "Timo Bremer",
                    "Suhas Lohit",
                    "Kowshik Thopalli",
                    "Vivek Narayanaswamy",
                    "Ahnaf Masroor",
                    "Jingyong Su",
                    "Anuj Srivastava"
                ],
                "domain": [
                    "Machine Learning",
                    "Graph Neural Network",
                    "Uncertainty Estimation",
                    "Domain Adaptation"
                ],
                "publications": [
                    {
                        "title": "Interactively Test Driving an Object Detector: Estimating Performance on Unlabeled Data",
                        "abstract": "In this paper, we study the problem of `test-driving' a detector, i.e. allowing a human user to get a quick sense of how well the detector generalizes to their specific requirement. To this end, we present the first system that estimates detector performance interactively without extensive ground truthing using a human in the loop. We approach this as a problem of estimating proportions and show that it is possible to make accurate inferences on the proportion of classes or groups within a large data collection by observing only $5-10\\%$ of samples from the data. In estimating the false detections (for precision), the samples are chosen carefully such that the overall characteristics of the data collection are preserved. Next, inspired by its use in estimating disease propagation we apply pooled testing approaches to estimate missed detections (for recall) from the dataset. The estimates thus obtained are close to the ones obtained using ground truth, thus reducing the need for extensive labeling which is expensive and time consuming."
                    },
                    {
                        "title": "Geometry-based Adaptive Symbolic Approximation for Fast Sequence Matching on Manifolds",
                        "abstract": "In this paper, we consider the problem of fast and efficient indexing techniques for sequences evolving in non-Euclidean spaces. This problem has several applications in the areas of human activity analysis, where there is a need to perform fast search, and recognition in very high dimensional spaces. The problem is made more challenging when representations such as landmarks, contours, and human skeletons etc. are naturally studied in a non-Euclidean setting where even simple operations are much more computationally intensive than their Euclidean counterparts. We propose a geometry and data adaptive symbolic framework that is shown to enable the deployment of fast and accurate algorithms for activity recognition, dynamic texture recognition, motif discovery. Toward this end, we present generalizations of key concepts of piece-wise aggregation and symbolic approximation for the case of non-Euclidean manifolds. We show that one can replace expensive geodesic computations with much faster symbolic computations with little loss of accuracy in activity recognition and discovery applications. The framework is general enough to work across both Euclidean and non-Euclidean spaces, depending on appropriate feature representations without compromising on the ultra-low bandwidth, high speed and high accuracy. The proposed methods are ideally suited for real-time systems and low complexity scenarios."
                    },
                    {
                        "title": "Diversity Promoting Online Sampling for Streaming Video Summarization",
                        "abstract": "Many applications benefit from sampling algorithms where a small number of well chosen samples are used to generalize different properties of a large dataset. In this paper, we use diverse sampling for streaming video summarization. Several emerging applications support streaming video, but existing summarization algorithms need access to the entire video which requires a lot of memory and computational power. We propose a memory efficient and computationally fast, online algorithm that uses competitive learning for diverse sampling. Our algorithm is a generalization of online K-means such that the cost function reduces clustering error, while also ensuring a diverse set of samples. The diversity is measured as the volume of a convex hull around the samples. Finally, the performance of the proposed algorithm is measured against human users for 50 videos in the VSUMM dataset. The algorithm performs better than batch mode summarization, while requiring significantly lower memory and computational requirements."
                    },
                    {
                        "title": "Bootstrapping Graph Convolutional Neural Networks for Autism Spectrum Disorder Classification",
                        "abstract": "Using predictive models to identify patterns that can act as biomarkers for different neuropathoglogical conditions is becoming highly prevalent. In this paper, we consider the problem of Autism Spectrum Disorder (ASD) classification where previous work has shown that it can be beneficial to incorporate a wide variety of meta features, such as socio-cultural traits, into predictive modeling. A graph-based approach naturally suits these scenarios, where a contextual graph captures traits that characterize a population, while the specific brain activity patterns are utilized as a multivariate signal at the nodes. Graph neural networks have shown improvements in inferencing with graph-structured data. Though the underlying graph strongly dictates the overall performance, there exists no systematic way of choosing an appropriate graph in practice, thus making predictive models non-robust. To address this, we propose a bootstrapped version of graph convolutional neural networks (G-CNNs) that utilizes an ensemble of weakly trained G-CNNs, and reduce the sensitivity of models on the choice of graph construction. We demonstrate its effectiveness on the challenging Autism Brain Imaging Data Exchange (ABIDE) dataset and show that our approach improves upon recently proposed graph-based neural networks. We also show that our method remains more robust to noisy graphs."
                    },
                    {
                        "title": "Generative Patch Priors for Practical Compressive Image Recovery",
                        "abstract": "In this paper, we propose the generative patch prior (GPP) that defines a generative prior for compressive image recovery, based on patch-manifold models. Unlike learned, image-level priors that are restricted to the range space of a pre-trained generator, GPP can recover a wide variety of natural images using a pre-trained patch generator. Additionally, GPP retains the benefits of generative priors like high reconstruction quality at extremely low sensing rates, while also being much more generally applicable. We show that GPP outperforms several unsupervised and supervised techniques on three different sensing models -- linear compressive sensing with known, and unknown calibration settings, and the non-linear phase retrieval problem. Finally, we propose an alternating optimization strategy using GPP for joint calibration-and-reconstruction which performs favorably against several baselines on a real world, un-calibrated compressive sensing dataset."
                    },
                    {
                        "title": "$\u0394$-UQ: Accurate Uncertainty Quantification via Anchor Marginalization",
                        "abstract": "We present $\\Delta$-UQ -- a novel, general-purpose uncertainty estimator using the concept of anchoring in predictive models. Anchoring works by first transforming the input into a tuple consisting of an anchor point drawn from a prior distribution, and a combination of the input sample with the anchor using a pretext encoding scheme. This encoding is such that the original input can be perfectly recovered from the tuple -- regardless of the choice of the anchor. Therefore, any predictive model should be able to predict the target response from the tuple alone (since it implicitly represents the input). Moreover, by varying the anchors for a fixed sample, we can estimate uncertainty in the prediction even using only a single predictive model. We find this uncertainty is deeply connected to improper sampling of the input data, and inherent noise, enabling us to estimate the total uncertainty in any system. With extensive empirical studies on a variety of use-cases, we demonstrate that $\\Delta$-UQ outperforms several competitive baselines. Specifically, we study model fitting, sequential model optimization, model based inversion in the regression setting and out of distribution detection, & calibration under distribution shifts for classification."
                    },
                    {
                        "title": "Out of Distribution Detection via Neural Network Anchoring",
                        "abstract": "Our goal in this paper is to exploit heteroscedastic temperature scaling as a calibration strategy for out of distribution (OOD) detection. Heteroscedasticity here refers to the fact that the optimal temperature parameter for each sample can be different, as opposed to conventional approaches that use the same value for the entire distribution. To enable this, we propose a new training strategy called anchoring that can estimate appropriate temperature values for each sample, leading to state-of-the-art OOD detection performance across several benchmarks. Using NTK theory, we show that this temperature function estimate is closely linked to the epistemic uncertainty of the classifier, which explains its behavior. In contrast to some of the best-performing OOD detection approaches, our method does not require exposure to additional outlier datasets, custom calibration objectives, or model ensembling. Through empirical studies with different OOD detection settings -- far OOD, near OOD, and semantically coherent OOD - we establish a highly effective OOD detection approach. Code to reproduce our results is available at github.com/LLNL/AMP"
                    },
                    {
                        "title": "Elastic Functional Coding of Riemannian Trajectories",
                        "abstract": "Visual observations of dynamic phenomena, such as human actions, are often represented as sequences of smoothly-varying features . In cases where the feature spaces can be structured as Riemannian manifolds, the corresponding representations become trajectories on manifolds. Analysis of these trajectories is challenging due to non-linearity of underlying spaces and high-dimensionality of trajectories. In vision problems, given the nature of physical systems involved, these phenomena are better characterized on a low-dimensional manifold compared to the space of Riemannian trajectories. For instance, if one does not impose physical constraints of the human body, in data involving human action analysis, the resulting representation space will have highly redundant features. Learning an effective, low-dimensional embedding for action representations will have a huge impact in the areas of search and retrieval, visualization, learning, and recognition. The difficulty lies in inherent non-linearity of the domain and temporal variability of actions that can distort any traditional metric between trajectories. To overcome these issues, we use the framework based on transported square-root velocity fields (TSRVF); this framework has several desirable properties, including a rate-invariant metric and vector space representations. We propose to learn an embedding such that each action trajectory is mapped to a single point in a low-dimensional Euclidean space, and the trajectories that differ only in temporal rates map to the same point. We utilize the TSRVF representation, and accompanying statistical summaries of Riemannian trajectories, to extend existing coding methods such as PCA, KSVD and Label Consistent KSVD to Riemannian trajectories or more generally to Riemannian functions."
                    },
                    {
                        "title": "Recovering Trajectories of Unmarked Joints in 3D Human Actions Using Latent Space Optimization",
                        "abstract": "Motion capture (mocap) and time-of-flight based sensing of human actions are becoming increasingly popular modalities to perform robust activity analysis. Applications range from action recognition to quantifying movement quality for health applications. While marker-less motion capture has made great progress, in critical applications such as healthcare, marker-based systems, especially active markers, are still considered gold-standard. However, there are several practical challenges in both modalities such as visibility, tracking errors, and simply the need to keep marker setup convenient wherein movements are recorded with a reduced marker-set. This implies that certain joint locations will not even be marked-up, making downstream analysis of full body movement challenging. To address this gap, we first pose the problem of reconstructing the unmarked joint data as an ill-posed linear inverse problem. We recover missing joints for a given action by projecting it onto the manifold of human actions, this is achieved by optimizing the latent space representation of a deep autoencoder. Experiments on both mocap and Kinect datasets clearly demonstrate that the proposed method performs very well in recovering semantics of the actions and dynamics of missing joints. We will release all the code and models publicly."
                    },
                    {
                        "title": "MimicGAN: Corruption-Mimicking for Blind Image Recovery & Adversarial Defense",
                        "abstract": "Solving inverse problems continues to be a central challenge in computer vision. Existing techniques either explicitly construct an inverse mapping using prior knowledge about the corruption, or learn the inverse directly using a large collection of examples. However, in practice, the nature of corruption may be unknown, and thus it is challenging to regularize the problem of inferring a plausible solution. On the other hand, collecting task-specific training data is tedious for known corruptions and impossible for unknown ones. We present MimicGAN, an unsupervised technique to solve general inverse problems based on image priors in the form of generative adversarial networks (GANs). Using a GAN prior, we show that one can reliably recover solutions to underdetermined inverse problems through a surrogate network that learns to mimic the corruption at test time. Our system successively estimates the corruption and the clean image without the need for supervisory training, while outperforming existing baselines in blind image recovery. We also demonstrate that MimicGAN improves upon recent GAN-based defenses against adversarial attacks and represents one of the strongest test-time defenses available today."
                    },
                    {
                        "title": "Autism Spectrum Disorder Classification using Graph Kernels on Multidimensional Time Series",
                        "abstract": "We present an approach to model time series data from resting state fMRI for autism spectrum disorder (ASD) severity classification. We propose to adopt kernel machines and employ graph kernels that define a kernel dot product between two graphs. This enables us to take advantage of spatio-temporal information to capture the dynamics of the brain network, as opposed to aggregating them in the spatial or temporal dimension. In addition to the conventional similarity graphs, we explore the use of L1 graph using sparse coding, and the persistent homology of time delay embeddings, in the proposed pipeline for ASD classification. In our experiments on two datasets from the ABIDE collection, we demonstrate a consistent and significant advantage in using graph kernels over traditional linear or non linear kernels for a variety of time series features."
                    },
                    {
                        "title": "Multiple Subspace Alignment Improves Domain Adaptation",
                        "abstract": "We present a novel unsupervised domain adaptation (DA) method for cross-domain visual recognition. Though subspace methods have found success in DA, their performance is often limited due to the assumption of approximating an entire dataset using a single low-dimensional subspace. Instead, we develop a method to effectively represent the source and target datasets via a collection of low-dimensional subspaces, and subsequently align them by exploiting the natural geometry of the space of subspaces, on the Grassmann manifold. We demonstrate the effectiveness of this approach, using empirical studies on two widely used benchmarks, with state of the art domain adaptation performance"
                    },
                    {
                        "title": "An Unsupervised Approach to Solving Inverse Problems using Generative Adversarial Networks",
                        "abstract": "Solving inverse problems continues to be a challenge in a wide array of applications ranging from deblurring, image inpainting, source separation etc. Most existing techniques solve such inverse problems by either explicitly or implicitly finding the inverse of the model. The former class of techniques require explicit knowledge of the measurement process which can be unrealistic, and rely on strong analytical regularizers to constrain the solution space, which often do not generalize well. The latter approaches have had remarkable success in part due to deep learning, but require a large collection of source-observation pairs, which can be prohibitively expensive. In this paper, we propose an unsupervised technique to solve inverse problems with generative adversarial networks (GANs). Using a pre-trained GAN in the space of source signals, we show that one can reliably recover solutions to under determined problems in a `blind' fashion, i.e., without knowledge of the measurement process. We solve this by making successive estimates on the model and the solution in an iterative fashion. We show promising results in three challenging applications -- blind source separation, image deblurring, and recovering an image from its edge map, and perform better than several baselines."
                    },
                    {
                        "title": "A Riemannian Framework for Statistical Analysis of Topological Persistence Diagrams",
                        "abstract": "Topological data analysis is becoming a popular way to study high dimensional feature spaces without any contextual clues or assumptions. This paper concerns itself with one popular topological feature, which is the number of $d-$dimensional holes in the dataset, also known as the Betti$-d$ number. The persistence of the Betti numbers over various scales is encoded into a persistence diagram (PD), which indicates the birth and death times of these holes as scale varies. A common way to compare PDs is by a point-to-point matching, which is given by the $n$-Wasserstein metric. However, a big drawback of this approach is the need to solve correspondence between points before computing the distance; for $n$ points, the complexity grows according to $\\mathcal{O}($n$^3)$. Instead, we propose to use an entirely new framework built on Riemannian geometry, that models PDs as 2D probability density functions that are represented in the square-root framework on a Hilbert Sphere. The resulting space is much more intuitive with closed form expressions for common operations. The distance metric is 1) correspondence-free and also 2) independent of the number of points in the dataset. The complexity of computing distance between PDs now grows according to $\\mathcal{O}(K^2)$, for a $K \\times K$ discretization of $[0,1]^2$. This also enables the use of existing machinery in differential geometry towards statistical analysis of PDs such as computing the mean, geodesics, classification etc. We report competitive results with the Wasserstein metric, at a much lower computational load, indicating the favorable properties of the proposed approach."
                    },
                    {
                        "title": "MimicGAN: Robust Projection onto Image Manifolds with Corruption Mimicking",
                        "abstract": "In the past few years, Generative Adversarial Networks (GANs) have dramatically advanced our ability to represent and parameterize high-dimensional, non-linear image manifolds. As a result, they have been widely adopted across a variety of applications, ranging from challenging inverse problems like image completion, to problems such as anomaly detection and adversarial defense. A recurring theme in many of these applications is the notion of projecting an image observation onto the manifold that is inferred by the generator. In this context, Projected Gradient Descent (PGD) has been the most popular approach, which essentially optimizes for a latent vector that minimizes the discrepancy between a generated image and the given observation. However, PGD is a brittle optimization technique that fails to identify the right projection (or latent vector) when the observation is corrupted, or perturbed even by a small amount. Such corruptions are common in the real world, for example images in the wild come with unknown crops, rotations, missing pixels, or other kinds of non-linear distributional shifts which break current encoding methods, rendering downstream applications unusable. To address this, we propose corruption mimicking -- a new robust projection technique, that utilizes a surrogate network to approximate the unknown corruption directly at test time, without the need for additional supervision or data augmentation. The proposed method is significantly more robust than PGD and other competing methods under a wide variety of corruptions, thereby enabling a more effective use of GANs in real-world applications. More importantly, we show that our approach produces state-of-the-art performance in several GAN-based applications -- anomaly detection, domain adaptation, and adversarial defense, that benefit from an accurate projection."
                    },
                    {
                        "title": "Extreme Few-view CT Reconstruction using Deep Inference",
                        "abstract": "Reconstruction of few-view x-ray Computed Tomography (CT) data is a highly ill-posed problem. It is often used in applications that require low radiation dose in clinical CT, rapid industrial scanning, or fixed-gantry CT. Existing analytic or iterative algorithms generally produce poorly reconstructed images, severely deteriorated by artifacts and noise, especially when the number of x-ray projections is considerably low. This paper presents a deep network-driven approach to address extreme few-view CT by incorporating convolutional neural network-based inference into state-of-the-art iterative reconstruction. The proposed method interprets few-view sinogram data using attention-based deep networks to infer the reconstructed image. The predicted image is then used as prior knowledge in the iterative algorithm for final reconstruction. We demonstrate effectiveness of the proposed approach by performing reconstruction experiments on a chest CT dataset."
                    },
                    {
                        "title": "PAGER: A Framework for Failure Analysis of Deep Regression Models",
                        "abstract": "Safe deployment of AI models requires proactive detection of failures to prevent costly errors. To this end, we study the important problem of detecting failures in deep regression models. Existing approaches rely on epistemic uncertainty estimates or inconsistency w.r.t the training data to identify failure. Interestingly, we find that while uncertainties are necessary they are insufficient to accurately characterize failure in practice. Hence, we introduce PAGER (Principled Analysis of Generalization Errors in Regressors), a framework to systematically detect and characterize failures in deep regressors. Built upon the principle of anchored training in deep models, PAGER unifies both epistemic uncertainty and complementary manifold non-conformity scores to accurately organize samples into different risk regimes."
                    },
                    {
                        "title": "SALT: Subspace Alignment as an Auxiliary Learning Task for Domain Adaptation",
                        "abstract": "Unsupervised domain adaptation aims to transfer and adapt knowledge learned from a labeled source domain to an unlabeled target domain. Key components of unsupervised domain adaptation include: (a) maximizing performance on the target, and (b) aligning the source and target domains. Traditionally, these tasks have either been considered as separate, or assumed to be implicitly addressed together with high-capacity feature extractors. When considered separately, alignment is usually viewed as a problem of aligning data distributions, either through geometric approaches such as subspace alignment or through distributional alignment such as optimal transport. This paper represents a hybrid approach, where we assume simplified data geometry in the form of subspaces, and consider alignment as an auxiliary task to the primary task of maximizing performance on the source. The alignment is made rather simple by leveraging tractable data geometry in the form of subspaces. We synergistically allow certain parameters derived from the closed-form auxiliary solution, to be affected by gradients from the primary task. The proposed approach represents a unique fusion of geometric and model-based alignment with gradients from a data-driven primary task. Our approach termed SALT, is a simple framework that achieves comparable or sometimes outperforms state-of-the-art on multiple standard benchmarks."
                    },
                    {
                        "title": "Unsupervised Audio Source Separation using Generative Priors",
                        "abstract": "State-of-the-art under-determined audio source separation systems rely on supervised end-end training of carefully tailored neural network architectures operating either in the time or the spectral domain. However, these methods are severely challenged in terms of requiring access to expensive source level labeled data and being specific to a given set of sources and the mixing process, which demands complete re-training when those assumptions change. This strongly emphasizes the need for unsupervised methods that can leverage the recent advances in data-driven modeling, and compensate for the lack of labeled data through meaningful priors. To this end, we propose a novel approach for audio source separation based on generative priors trained on individual sources. Through the use of projected gradient descent optimization, our approach simultaneously searches in the source-specific latent spaces to effectively recover the constituent sources. Though the generative priors can be defined in the time domain directly, e.g. WaveGAN, we find that using spectral domain loss functions for our optimization leads to good-quality source estimates. Our empirical studies on standard spoken digit and instrument datasets clearly demonstrate the effectiveness of our approach over classical as well as state-of-the-art unsupervised baselines."
                    }
                ]
            },
            "5efa27da-5963-436f-a130-3fbbe8fe0b08": {
                "pk": "5efa27da-5963-436f-a130-3fbbe8fe0b08",
                "name": "Yamen Mubarka",
                "collaborators": [
                    "Vivek Narayanaswamy",
                    "Rushil Anirudh",
                    "Deepta Rajan",
                    "Andreas Spanias",
                    "Jayaraman J. Thiagarajan"
                ],
                "domain": [
                    "Out-of-Distribution Detection",
                    "Medical Imaging",
                    "Data Augmentation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Know Your Space: Inlier and Outlier Construction for Calibrating Medical OOD Detectors",
                        "abstract": "We focus on the problem of producing well-calibrated out-of-distribution (OOD) detectors, in order to enable safe deployment of medical image classifiers. Motivated by the difficulty of curating suitable calibration datasets, synthetic augmentations have become highly prevalent for inlier/outlier specification. While there have been rapid advances in data augmentation techniques, this paper makes a striking finding that the space in which the inliers and outliers are synthesized, in addition to the type of augmentation, plays a critical role in calibrating OOD detectors. Using the popular energy-based OOD detection framework, we find that the optimal protocol is to synthesize latent-space inliers along with diverse pixel-space outliers. Based on empirical studies with multiple medical imaging benchmarks, we demonstrate that our approach consistently leads to superior OOD detection ($15\\% - 35\\%$ in AUROC) over the state-of-the-art in a variety of open-set recognition settings."
                    }
                ]
            },
            "e0d188b9-1241-4c75-81c3-3083368e7c62": {
                "pk": "e0d188b9-1241-4c75-81c3-3083368e7c62",
                "name": "Wesam Sakla",
                "collaborators": [
                    "Oscar Leong",
                    "Lucas Jaffe",
                    "Michael Zelinski",
                    "Yuzhe Lu",
                    "Shusen Liu",
                    "Jayaraman J. Thiagarajan",
                    "Rushil Anirudh"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Generative Models",
                    "Remote Sensing"
                ],
                "publications": [
                    {
                        "title": "Low Shot Learning with Untrained Neural Networks for Imaging Inverse Problems",
                        "abstract": "Employing deep neural networks as natural image priors to solve inverse problems either requires large amounts of data to sufficiently train expressive generative models or can succeed with no data via untrained neural networks. However, very few works have considered how to interpolate between these no- to high-data regimes. In particular, how can one use the availability of a small amount of data (even $5-25$ examples) to one's advantage in solving these inverse problems and can a system's performance increase as the amount of data increases as well? In this work, we consider solving linear inverse problems when given a small number of examples of images that are drawn from the same distribution as the image of interest. Comparing to untrained neural networks that use no data, we show how one can pre-train a neural network with a few given examples to improve reconstruction results in compressed sensing and semantic image recovery problems such as colorization. Our approach leads to improved reconstruction as the amount of available data increases and is on par with fully trained generative models, while requiring less than $1 \\%$ of the data needed to train a generative model."
                    },
                    {
                        "title": "Remote Sensor Design for Visual Recognition with Convolutional Neural Networks",
                        "abstract": "While deep learning technologies for computer vision have developed rapidly since 2012, modeling of remote sensing systems has remained focused around human vision. In particular, remote sensing systems are usually constructed to optimize sensing cost-quality trade-offs with respect to human image interpretability. While some recent studies have explored remote sensing system design as a function of simple computer vision algorithm performance, there has been little work relating this design to the state-of-the-art in computer vision: deep learning with convolutional neural networks. We develop experimental systems to conduct this analysis, showing results with modern deep learning algorithms and recent overhead image data. Our results are compared to standard image quality measurements based on human visual perception, and we conclude not only that machine and human interpretability differ significantly, but that computer vision performance is largely self-consistent across a range of disparate conditions. This research is presented as a cornerstone for a new generation of sensor design systems which focus on computer algorithm performance instead of human visual perception."
                    },
                    {
                        "title": "On-the-fly Object Detection using StyleGAN with CLIP Guidance",
                        "abstract": "We present a fully automated framework for building object detectors on satellite imagery without requiring any human annotation or intervention. We achieve this by leveraging the combined power of modern generative models (e.g., StyleGAN) and recent advances in multi-modal learning (e.g., CLIP). While deep generative models effectively encode the key semantics pertinent to a data distribution, this information is not immediately accessible for downstream tasks, such as object detection. In this work, we exploit CLIP's ability to associate image features with text descriptions to identify neurons in the generator network, which are subsequently used to build detectors on-the-fly."
                    }
                ]
            },
            "dcdbf784-0489-4283-bb1d-95fc4199148c": {
                "pk": "dcdbf784-0489-4283-bb1d-95fc4199148c",
                "name": "Jayaraman J. Thiagarajan",
                "collaborators": [
                    "Andreas Spanias",
                    "Deepta Rajan",
                    "Bindya Venkatesh",
                    "Prasanna Sattigeri",
                    "Vivek Narayanaswamy",
                    "Rushil Anirudh",
                    "Huan Song",
                    "Uday Shankar Shanthamallu",
                    "Karthikeyan Natesan Ramamurthy",
                    "Satyananda Kashyap"
                ],
                "domain": [
                    "Machine Learning",
                    "Uncertainty Quantification",
                    "Graph Neural Network",
                    "Medical Imaging"
                ],
                "publications": [
                    {
                        "title": "Understanding Behavior of Clinical Models under Domain Shifts",
                        "abstract": "The hypothesis that computational models can be reliable enough to be adopted in prognosis and patient care is revolutionizing healthcare. Deep learning, in particular, has been a game changer in building predictive models, thus leading to community-wide data curation efforts. However, due to inherent variabilities in population characteristics and biological systems, these models are often biased to the training datasets. This can be limiting when models are deployed in new environments, when there are systematic domain shifts not known a priori. In this paper, we propose to emulate a large class of domain shifts, that can occur in clinical settings, with a given dataset, and argue that evaluating the behavior of predictive models in light of those shifts is an effective way to quantify their reliability. More specifically, we develop an approach for building realistic scenarios, based on analysis of \\textit{disease landscapes} in multi-label classification. Using the openly available MIMIC-III EHR dataset for phenotyping, for the first time, our work sheds light into data regimes where deep clinical models can fail to generalize. This work emphasizes the need for novel validation mechanisms driven by real-world domain shifts in AI for healthcare."
                    },
                    {
                        "title": "Improved Deep Embeddings for Inferencing with Multi-Layered Networks",
                        "abstract": "Inferencing with network data necessitates the mapping of its nodes into a vector space, where the relationships are preserved. However, with multi-layered networks, where multiple types of relationships exist for the same set of nodes, it is crucial to exploit the information shared between layers, in addition to the distinct aspects of each layer. In this paper, we propose a novel approach that first obtains node embeddings in all layers jointly via DeepWalk on a \\textit{supra} graph, which allows interactions between layers, and then fine-tunes the embeddings to encourage cohesive structure in the latent space. With empirical studies in node classification, link prediction and multi-layered community detection, we show that the proposed approach outperforms existing single- and multi-layered network embedding algorithms on several benchmarks. In addition to effectively scaling to a large number of layers (tested up to $37$), our approach consistently produces highly modular community structure, even when compared to methods that directly optimize for the modularity function."
                    },
                    {
                        "title": "Distill-to-Label: Weakly Supervised Instance Labeling Using Knowledge Distillation",
                        "abstract": "Weakly supervised instance labeling using only image-level labels, in lieu of expensive fine-grained pixel annotations, is crucial in several applications including medical image analysis. In contrast to conventional instance segmentation scenarios in computer vision, the problems that we consider are characterized by a small number of training images and non-local patterns that lead to the diagnosis. In this paper, we explore the use of multiple instance learning (MIL) to design an instance label generator under this weakly supervised setting. Motivated by the observation that an MIL model can handle bags of varying sizes, we propose to repurpose an MIL model originally trained for bag-level classification to produce reliable predictions for single instances, i.e., bags of size $1$. To this end, we introduce a novel regularization strategy based on virtual adversarial training for improving MIL training, and subsequently develop a knowledge distillation technique for repurposing the trained MIL model. Using empirical studies on colon cancer and breast cancer detection from histopathological images, we show that the proposed approach produces high-quality instance-level prediction and significantly outperforms state-of-the MIL methods."
                    },
                    {
                        "title": "A Generative Modeling Approach to Limited Channel ECG Classification",
                        "abstract": "Processing temporal sequences is central to a variety of applications in health care, and in particular multi-channel Electrocardiogram (ECG) is a highly prevalent diagnostic modality that relies on robust sequence modeling. While Recurrent Neural Networks (RNNs) have led to significant advances in automated diagnosis with time-series data, they perform poorly when models are trained using a limited set of channels. A crucial limitation of existing solutions is that they rely solely on discriminative models, which tend to generalize poorly in such scenarios. In order to combat this limitation, we develop a generative modeling approach to limited channel ECG classification. This approach first uses a Seq2Seq model to implicitly generate the missing channel information, and then uses the latent representation to perform the actual supervisory task. This decoupling enables the use of unsupervised data and also provides highly robust metric spaces for subsequent discriminative learning. Our experiments with the Physionet dataset clearly evidence the effectiveness of our approach over standard RNNs in disease prediction."
                    },
                    {
                        "title": "Learn-By-Calibrating: Using Calibration as a Training Objective",
                        "abstract": "Calibration error is commonly adopted for evaluating the quality of uncertainty estimators in deep neural networks. In this paper, we argue that such a metric is highly beneficial for training predictive models, even when we do not explicitly measure the uncertainties. This is conceptually similar to heteroscedastic neural networks that produce variance estimates for each prediction, with the key difference that we do not place a Gaussian prior on the predictions. We propose a novel algorithm that performs simultaneous interval estimation for different calibration levels and effectively leverages the intervals to refine the mean estimates. Our results show that, our approach is consistently superior to existing regularization strategies in deep regression models. Finally, we propose to augment partial dependence plots, a model-agnostic interpretability tool, with expected prediction intervals to reveal interesting dependencies between data and the target."
                    },
                    {
                        "title": "Heteroscedastic Calibration of Uncertainty Estimators in Deep Learning",
                        "abstract": "The role of uncertainty quantification (UQ) in deep learning has become crucial with growing use of predictive models in high-risk applications. Though a large class of methods exists for measuring deep uncertainties, in practice, the resulting estimates are found to be poorly calibrated, thus making it challenging to translate them into actionable insights. A common workaround is to utilize a separate recalibration step, which adjusts the estimates to compensate for the miscalibration. Instead, we propose to repurpose the heteroscedastic regression objective as a surrogate for calibration and enable any existing uncertainty estimator to be inherently calibrated. In addition to eliminating the need for recalibration, this also regularizes the training process. Using regression experiments, we demonstrate the effectiveness of the proposed heteroscedastic calibration with two popular uncertainty estimators."
                    },
                    {
                        "title": "Ask-n-Learn: Active Learning via Reliable Gradient Representations for Image Classification",
                        "abstract": "Deep predictive models rely on human supervision in the form of labeled training data. Obtaining large amounts of annotated training data can be expensive and time consuming, and this becomes a critical bottleneck while building such models in practice. In such scenarios, active learning (AL) strategies are used to achieve faster convergence in terms of labeling efforts. Existing active learning employ a variety of heuristics based on uncertainty and diversity to select query samples. Despite their wide-spread use, in practice, their performance is limited by a number of factors including non-calibrated uncertainties, insufficient trade-off between data exploration and exploitation, presence of confirmation bias etc. In order to address these challenges, we propose Ask-n-Learn, an active learning approach based on gradient embeddings obtained using the pesudo-labels estimated in each iteration of the algorithm. More importantly, we advocate the use of prediction calibration to obtain reliable gradient embeddings, and propose a data augmentation strategy to alleviate the effects of confirmation bias during pseudo-labeling. Through empirical studies on benchmark image classification tasks (CIFAR-10, SVHN, Fashion-MNIST, MNIST), we demonstrate significant improvements over state-of-the-art baselines, including the recently proposed BADGE algorithm."
                    },
                    {
                        "title": "Using Deep Image Priors to Generate Counterfactual Explanations",
                        "abstract": "Through the use of carefully tailored convolutional neural network architectures, a deep image prior (DIP) can be used to obtain pre-images from latent representation encodings. Though DIP inversion has been known to be superior to conventional regularized inversion strategies such as total variation, such an over-parameterized generator is able to effectively reconstruct even images that are not in the original data distribution. This limitation makes it challenging to utilize such priors for tasks such as counterfactual reasoning, wherein the goal is to generate small, interpretable changes to an image that systematically leads to changes in the model prediction. To this end, we propose a novel regularization strategy based on an auxiliary loss estimator jointly trained with the predictor, which efficiently guides the prior to recover natural pre-images. Our empirical studies with a real-world ISIC skin lesion detection problem clearly evidence the effectiveness of the proposed approach in synthesizing meaningful counterfactuals. In comparison, we find that the standard DIP inversion often proposes visually imperceptible perturbations to irrelevant parts of the image, thus providing no additional insights into the model behavior."
                    },
                    {
                        "title": "Attend and Diagnose: Clinical Time Series Analysis using Attention Models",
                        "abstract": "With widespread adoption of electronic health records, there is an increased emphasis for predictive models that can effectively deal with clinical time-series data. Powered by Recurrent Neural Network (RNN) architectures with Long Short-Term Memory (LSTM) units, deep neural networks have achieved state-of-the-art results in several clinical prediction tasks. Despite the success of RNNs, its sequential nature prohibits parallelized computing, thus making it inefficient particularly when processing long sequences. Recently, architectures which are based solely on attention mechanisms have shown remarkable success in transduction tasks in NLP, while being computationally superior. In this paper, for the first time, we utilize attention models for clinical time-series modeling, thereby dispensing recurrence entirely. We develop the \\textit{SAnD} (Simply Attend and Diagnose) architecture, which employs a masked, self-attention mechanism, and uses positional encoding and dense interpolation strategies for incorporating temporal order. Furthermore, we develop a multi-task variant of \\textit{SAnD} to jointly infer models with multiple diagnosis tasks. Using the recent MIMIC-III benchmark datasets, we demonstrate that the proposed approach achieves state-of-the-art performance in all tasks, outperforming LSTM models and classical baselines with hand-engineered features."
                    },
                    {
                        "title": "A Regularized Attention Mechanism for Graph Attention Networks",
                        "abstract": "Machine learning models that can exploit the inherent structure in data have gained prominence. In particular, there is a surge in deep learning solutions for graph-structured data, due to its wide-spread applicability in several fields. Graph attention networks (GAT), a recent addition to the broad class of feature learning models in graphs, utilizes the attention mechanism to efficiently learn continuous vector representations for semi-supervised learning problems. In this paper, we perform a detailed analysis of GAT models, and present interesting insights into their behavior. In particular, we show that the models are vulnerable to heterogeneous rogue nodes and hence propose novel regularization strategies to improve the robustness of GAT models. Using benchmark datasets, we demonstrate performance improvements on semi-supervised learning, using the proposed robust variant of GAT."
                    },
                    {
                        "title": "Calibrate and Prune: Improving Reliability of Lottery Tickets Through Prediction Calibration",
                        "abstract": "The hypothesis that sub-network initializations (lottery) exist within the initializations of over-parameterized networks, which when trained in isolation produce highly generalizable models, has led to crucial insights into network initialization and has enabled efficient inferencing. Supervised models with uncalibrated confidences tend to be overconfident even when making wrong prediction. In this paper, for the first time, we study how explicit confidence calibration in the over-parameterized network impacts the quality of the resulting lottery tickets. More specifically, we incorporate a suite of calibration strategies, ranging from mixup regularization, variance-weighted confidence calibration to the newly proposed likelihood-based calibration and normalized bin assignment strategies. Furthermore, we explore different combinations of architectures and datasets, and make a number of key findings about the role of confidence calibration. Our empirical studies reveal that including calibration mechanisms consistently lead to more effective lottery tickets, in terms of accuracy as well as empirical calibration metrics, even when retrained using data with challenging distribution shifts with respect to the source dataset."
                    },
                    {
                        "title": "Learning Stable Multilevel Dictionaries for Sparse Representations",
                        "abstract": "Sparse representations using learned dictionaries are being increasingly used with success in several data processing and machine learning applications. The availability of abundant training data necessitates the development of efficient, robust and provably good dictionary learning algorithms. Algorithmic stability and generalization are desirable characteristics for dictionary learning algorithms that aim to build global dictionaries which can efficiently model any test data similar to the training samples. In this paper, we propose an algorithm to learn dictionaries for sparse representations from large scale data, and prove that the proposed learning algorithm is stable and generalizable asymptotically. The algorithm employs a 1-D subspace clustering procedure, the K-hyperline clustering, in order to learn a hierarchical dictionary with multiple levels. We also propose an information-theoretic scheme to estimate the number of atoms needed in each level of learning and develop an ensemble approach to learn robust dictionaries. Using the proposed dictionaries, the sparse code for novel test data can be computed using a low-complexity pursuit procedure. We demonstrate the stability and generalization characteristics of the proposed algorithm using simulations. We also evaluate the utility of the multilevel dictionaries in compressed recovery and subspace learning applications."
                    },
                    {
                        "title": "Multiple Kernel Sparse Representations for Supervised and Unsupervised Learning",
                        "abstract": "In complex visual recognition tasks it is typical to adopt multiple descriptors, that describe different aspects of the images, for obtaining an improved recognition performance. Descriptors that have diverse forms can be fused into a unified feature space in a principled manner using kernel methods. Sparse models that generalize well to the test data can be learned in the unified kernel space, and appropriate constraints can be incorporated for application in supervised and unsupervised learning. In this paper, we propose to perform sparse coding and dictionary learning in the multiple kernel space, where the weights of the ensemble kernel are tuned based on graph-embedding principles such that class discrimination is maximized. In our proposed algorithm, dictionaries are inferred using multiple levels of 1-D subspace clustering in the kernel space, and the sparse codes are obtained using a simple levelwise pursuit scheme. Empirical results for object recognition and image clustering show that our algorithm outperforms existing sparse coding based approaches, and compares favorably to other state-of-the-art methods."
                    },
                    {
                        "title": "Calibrating Healthcare AI: Towards Reliable and Interpretable Deep Predictive Models",
                        "abstract": "The wide-spread adoption of representation learning technologies in clinical decision making strongly emphasizes the need for characterizing model reliability and enabling rigorous introspection of model behavior. While the former need is often addressed by incorporating uncertainty quantification strategies, the latter challenge is addressed using a broad class of interpretability techniques. In this paper, we argue that these two objectives are not necessarily disparate and propose to utilize prediction calibration to meet both objectives. More specifically, our approach is comprised of a calibration-driven learning method, which is also used to design an interpretability technique based on counterfactual reasoning. Furthermore, we introduce \\textit{reliability plots}, a holistic evaluation mechanism for model reliability. Using a lesion classification problem with dermoscopy images, we demonstrate the effectiveness of our approach and infer interesting insights about the model behavior."
                    },
                    {
                        "title": "Uncertainty-Matching Graph Neural Networks to Defend Against Poisoning Attacks",
                        "abstract": "Graph Neural Networks (GNNs), a generalization of neural networks to graph-structured data, are often implemented using message passes between entities of a graph. While GNNs are effective for node classification, link prediction and graph classification, they are vulnerable to adversarial attacks, i.e., a small perturbation to the structure can lead to a non-trivial performance degradation. In this work, we propose Uncertainty Matching GNN (UM-GNN), that is aimed at improving the robustness of GNN models, particularly against poisoning attacks to the graph structure, by leveraging epistemic uncertainties from the message passing framework. More specifically, we propose to build a surrogate predictor that does not directly access the graph structure, but systematically extracts reliable knowledge from a standard GNN through a novel uncertainty-matching strategy. Interestingly, this uncoupling makes UM-GNN immune to evasion attacks by design, and achieves significantly improved robustness against poisoning attacks. Using empirical studies with standard benchmarks and a suite of global and target attacks, we demonstrate the effectiveness of UM-GNN, when compared to existing baselines including the state-of-the-art robust GCN."
                    },
                    {
                        "title": "Loss Estimators Improve Model Generalization",
                        "abstract": "With increased interest in adopting AI methods for clinical diagnosis, a vital step towards safe deployment of such tools is to ensure that the models not only produce accurate predictions but also do not generalize to data regimes where the training data provide no meaningful evidence. Existing approaches for ensuring the distribution of model predictions to be similar to that of the true distribution rely on explicit uncertainty estimators that are inherently hard to calibrate. In this paper, we propose to train a loss estimator alongside the predictive model, using a contrastive training objective, to directly estimate the prediction uncertainties. Interestingly, we find that, in addition to producing well-calibrated uncertainties, this approach improves the generalization behavior of the predictor. Using a dermatology use-case, we show the impact of loss estimators on model generalization, in terms of both its fidelity on in-distribution data and its ability to detect out of distribution samples or new classes unseen during training."
                    },
                    {
                        "title": "On the Design of Deep Priors for Unsupervised Audio Restoration",
                        "abstract": "Unsupervised deep learning methods for solving audio restoration problems extensively rely on carefully tailored neural architectures that carry strong inductive biases for defining priors in the time or spectral domain. In this context, lot of recent success has been achieved with sophisticated convolutional network constructions that recover audio signals in the spectral domain. However, in practice, audio priors require careful engineering of the convolutional kernels to be effective at solving ill-posed restoration tasks, while also being easy to train. To this end, in this paper, we propose a new U-Net based prior that does not impact either the network complexity or convergence behavior of existing convolutional architectures, yet leads to significantly improved restoration. In particular, we advocate the use of carefully designed dilation schedules and dense connections in the U-Net architecture to obtain powerful audio priors. Using empirical studies on standard benchmarks and a variety of ill-posed restoration tasks, such as audio denoising, in-painting and source separation, we demonstrate that our proposed approach consistently outperforms widely adopted audio prior architectures."
                    },
                    {
                        "title": "PAGER: A Framework for Failure Analysis of Deep Regression Models",
                        "abstract": "Safe deployment of AI models requires proactive detection of failures to prevent costly errors. To this end, we study the important problem of detecting failures in deep regression models. Existing approaches rely on epistemic uncertainty estimates or inconsistency w.r.t the training data to identify failure. Interestingly, we find that while uncertainties are necessary they are insufficient to accurately characterize failure in practice. Hence, we introduce PAGER (Principled Analysis of Generalization Errors in Regressors), a framework to systematically detect and characterize failures in deep regressors. Built upon the principle of anchored training in deep models, PAGER unifies both epistemic uncertainty and complementary manifold non-conformity scores to accurately organize samples into different risk regimes."
                    },
                    {
                        "title": "Bootstrapping Graph Convolutional Neural Networks for Autism Spectrum Disorder Classification",
                        "abstract": "Using predictive models to identify patterns that can act as biomarkers for different neuropathoglogical conditions is becoming highly prevalent. In this paper, we consider the problem of Autism Spectrum Disorder (ASD) classification where previous work has shown that it can be beneficial to incorporate a wide variety of meta features, such as socio-cultural traits, into predictive modeling. A graph-based approach naturally suits these scenarios, where a contextual graph captures traits that characterize a population, while the specific brain activity patterns are utilized as a multivariate signal at the nodes. Graph neural networks have shown improvements in inferencing with graph-structured data. Though the underlying graph strongly dictates the overall performance, there exists no systematic way of choosing an appropriate graph in practice, thus making predictive models non-robust. To address this, we propose a bootstrapped version of graph convolutional neural networks (G-CNNs) that utilizes an ensemble of weakly trained G-CNNs, and reduce the sensitivity of models on the choice of graph construction. We demonstrate its effectiveness on the challenging Autism Brain Imaging Data Exchange (ABIDE) dataset and show that our approach improves upon recently proposed graph-based neural networks. We also show that our method remains more robust to noisy graphs."
                    },
                    {
                        "title": "$\u0394$-UQ: Accurate Uncertainty Quantification via Anchor Marginalization",
                        "abstract": "We present $\\Delta$-UQ -- a novel, general-purpose uncertainty estimator using the concept of anchoring in predictive models. Anchoring works by first transforming the input into a tuple consisting of an anchor point drawn from a prior distribution, and a combination of the input sample with the anchor using a pretext encoding scheme. This encoding is such that the original input can be perfectly recovered from the tuple -- regardless of the choice of the anchor. Therefore, any predictive model should be able to predict the target response from the tuple alone (since it implicitly represents the input). Moreover, by varying the anchors for a fixed sample, we can estimate uncertainty in the prediction even using only a single predictive model. We find this uncertainty is deeply connected to improper sampling of the input data, and inherent noise, enabling us to estimate the total uncertainty in any system. With extensive empirical studies on a variety of use-cases, we demonstrate that $\\Delta$-UQ outperforms several competitive baselines. Specifically, we study model fitting, sequential model optimization, model based inversion in the regression setting and out of distribution detection, & calibration under distribution shifts for classification."
                    }
                ]
            }
        }
    },
    "2310.20703": {
        "paper_data": {
            "title": "Vanishing Gradients in Reinforcement Finetuning of Language Models",
            "url": "http://arxiv.org/abs/2310.20703v3",
            "arxiv_id": "2310.20703",
            "authors": [
                "Noam Razin",
                "Hattie Zhou",
                "Omid Saremi",
                "Vimal Thilak",
                "Arwen Bradley",
                "Preetum Nakkiran",
                "Joshua Susskind",
                "Etai Littwin"
            ],
            "abstract": "Pretrained language models are commonly aligned with human preferences and downstream tasks via reinforcement finetuning (RFT), which refers to maximizing a (possibly learned) reward function using policy gradient algorithms. This work identifies a fundamental optimization obstacle in RFT: we prove that the expected gradient for an input vanishes when its reward standard deviation under the model is small, even if the expected reward is far from optimal. Through experiments on an RFT benchmark and controlled environments, as well as a theoretical analysis, we then demonstrate that vanishing gradients due to small reward standard deviation are prevalent and detrimental, leading to extremely slow reward maximization. Lastly, we explore ways to overcome vanishing gradients in RFT. We find the common practice of an initial supervised finetuning (SFT) phase to be the most promising candidate, which sheds light on its importance in an RFT pipeline. Moreover, we show that a relatively small number of SFT optimization steps on as few as 1% of the input samples can suffice, indicating that the initial SFT phase need not be expensive in terms of compute and data labeling efforts. Overall, our results emphasize that being mindful for inputs whose expected gradient vanishes, as measured by the reward standard deviation, is crucial for successful execution of RFT.",
            "introduction": "   1 Introduction  The dominant machine learning paradigm for textual data relies on pretraining a language model on vast corpora, and subsequently aligning it with human preferences and downstream tasks via finetuning (see, e.g.,\u00a0Yang et\u00a0al. (2019); Raffel et\u00a0al. (2020); Wei et\u00a0al. (2022); Thoppilan et\u00a0al. (2022); OpenAI (2023); Touvron et\u00a0al. (2023); Zhao et\u00a0al. (2023)). Supervised finetuning (SFT) provides a straightforward procedure for adapting the model, in which the cross entropy loss is minimized over pairs of textual inputs and desirable outputs. Despite the simplicity of SFT, the cost of obtaining high-quality labeled data, as well as the difficulty in expressing human preferences through ground truth labels, has led to wide adoption of a reinforcement learning-based approach, herein referred to as reinforcement finetuning (RFT)\u00a0(Wu and Hu, 2018; Ziegler et\u00a0al., 2019; Stiennon et\u00a0al., 2020; Ouyang et\u00a0al., 2022; Bai et\u00a0al., 2022; OpenAI, 2023; Touvron et\u00a0al., 2023).   In RFT, instead of minimizing a loss over labeled inputs, policy gradient algorithms are used for maximizing a reward function. The reward function can be learned from human preferences, in which case RFT is commonly known as reinforcement learning from human feedback (RLHF), or tailored to a downstream task. Specifically, denote by V\u2062(\ud835\udc31;\u03b8)\ud835\udc49\ud835\udc31\ud835\udf03V({\\mathbf{x}};\\theta)italic_V ( bold_x ; italic_\u03b8 ) the expected reward that a model parameterized by\u00a0\u03b8\ud835\udf03\\thetaitalic_\u03b8 achieves, given as input a sequence of tokens \ud835\udc31\ud835\udc31{\\mathbf{x}}bold_x. For a distribution over inputs, RFT maximizes \ud835\udd3c\ud835\udc31\\brk\u2062[s]\u2062V\u2062(\ud835\udc31;\u03b8)subscript\ud835\udd3c\ud835\udc31\\brkdelimited-[]\ud835\udc60\ud835\udc49\ud835\udc31\ud835\udf03\\mathop{\\mathbb{E}}_{{\\mathbf{x}}}\\brk[s]{V({\\mathbf{x}};\\theta)}blackboard_E start_POSTSUBSCRIPT bold_x end_POSTSUBSCRIPT [ italic_s ] italic_V ( bold_x ; italic_\u03b8 ) with respect to\u00a0\u03b8\ud835\udf03\\thetaitalic_\u03b8 by following estimates of its gradient. Proximal Policy Optimization (PPO)\u00a0(Schulman et\u00a0al., 2017)\u00a0\u2014\u00a0a policy gradient algorithm widely used for RFT\u00a0\u2014\u00a0adheres to this scheme, up to replacing the expected reward with a surrogate objective.   In this work, we highlight a fundamental optimization obstacle in RFT: we prove that the expected gradient for an input \ud835\udc31\ud835\udc31{\\mathbf{x}}bold_x, i.e.\u00a0\u2207\u03b8V\u2062(\ud835\udc31;\u03b8)subscript\u2207\ud835\udf03\ud835\udc49\ud835\udc31\ud835\udf03\\nabla_{\\theta}V({\\mathbf{x}};\\theta)\u2207 start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT italic_V ( bold_x ; italic_\u03b8 ), vanishes when the reward standard deviation of \ud835\udc31\ud835\udc31{\\mathbf{x}}bold_x under the model is small, even if the expected reward V\u2062(\ud835\udc31;\u03b8)\ud835\udc49\ud835\udc31\ud835\udf03V({\\mathbf{x}};\\theta)italic_V ( bold_x ; italic_\u03b8 ) is far from optimal (Section\u00a03). The same holds for PPO, and stems from a combination of the reward maximization objective and the softmax operator used for producing distributions over tokens. In contrast, under SFT the expected gradient does not necessarily vanish in analogous circumstances.   Vanishing gradients in RFT can arise when attempting to reverse an existing behavior of the model or when the text distribution of the downstream task differs from that of the pretraining corpora. In both of these common use cases, over some inputs, the pretrained model is likely to give non-negligible probability only to outputs of roughly the same suboptimal reward. The reward standard deviation for such inputs is small, implying that their expected gradients are near-zero. Thus, RFT is anticipated to be ineffective for achieving a higher reward over them.    Using the GRUE benchmark\u00a0(Ramamurthy et\u00a0al., 2023) for RFT of language models, we empirically affirm the prevalence of inputs with small reward standard deviation, for which the expected gradient vanishes, and showcase its harmful effects (Section\u00a04.1). Specifically, we find that three of seven GRUE datasets contain a considerable amount of inputs with near-zero reward standard deviation under pretrained models, while their expected reward is suboptimal. As anticipated, RFT has limited",
            "references": [
                {
                    "title": "Reinforced Self-Training (ReST) for Language Modeling",
                    "abstract": "Reinforcement learning from human feedback (RLHF) can improve the quality of large language model's (LLM) outputs by aligning them with human preferences. We propose a simple algorithm for aligning LLMs with human preferences inspired by growing batch reinforcement learning (RL), which we call Reinforced Self-Training (ReST). Given an initial LLM policy, ReST produces a dataset by generating samples from the policy, which are then used to improve the LLM policy using offline RL algorithms. ReST is more efficient than typical online RLHF methods because the training dataset is produced offline, which allows data reuse. While ReST is a general approach applicable to all generative learning settings, we focus on its application to machine translation. Our results show that ReST can substantially improve translation quality, as measured by automated metrics and human evaluation on machine translation benchmarks in a compute and sample-efficient manner."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedback",
                    "abstract": "Large language models (LLMs) such as ChatGPT have seen widespread adoption due to their strong instruction-following abilities. Developing these LLMs involves a complex yet poorly understood workflow requiring training with human feedback. Replicating and understanding this instruction-following requires tackling three major challenges: the high cost of data collection, the lack of trustworthy evaluation, and the absence of reference method implementations. We address these challenges with AlpacaFarm, a simulator that enables research and development for learning from feedback at a low cost. First, we design LLM prompts to simulate human feedback that are 50x cheaper than crowdworkers and display high agreement with humans. Second, we propose an automatic evaluation and validate it against human instructions obtained on real-world interactions. Third, we contribute reference implementations for several methods (PPO, DPO, best-of-n, expert iteration, and more) that learn from pairwise feedback. Finally, as an end-to-end validation of AlpacaFarm, we train and evaluate eleven models on 10k pairs of real human feedback and show that rankings of models trained in AlpacaFarm match rankings of models trained on human data. As a demonstration of the research possible in AlpacaFarm, we find that methods that use a reward model can substantially improve over supervised fine-tuning and that our reference PPO implementation leads to a +10% improvement in win-rate against Davinci003. We release all components of AlpacaFarm at https://github.com/tatsu-lab/alpaca_farm."
                },
                {
                    "title": "RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment",
                    "abstract": "Generative foundation models are susceptible to implicit biases that can arise from extensive unsupervised training data. Such biases can produce suboptimal samples, skewed outcomes, and unfairness, with potentially serious consequences. Consequently, aligning these models with human ethics and preferences is an essential step toward ensuring their responsible and effective deployment in real-world applications. Prior research has primarily employed Reinforcement Learning from Human Feedback (RLHF) to address this problem, where generative models are fine-tuned with RL algorithms guided by a human-feedback-informed reward model. However, the inefficiencies and instabilities associated with RL algorithms frequently present substantial obstacles to the successful alignment, necessitating the development of a more robust and streamlined approach. To this end, we introduce a new framework, Reward rAnked FineTuning (RAFT), designed to align generative models effectively. Utilizing a reward model and a sufficient number of samples, our approach selects the high-quality samples, discarding those that exhibit undesired behavior, and subsequently enhancing the model by fine-tuning on these filtered samples. Our studies show that RAFT can effectively improve the model performance in both reward learning and other automated metrics in both large language models and diffusion models."
                },
                {
                    "title": "A Survey of Large Language Models",
                    "abstract": "Language is essentially a complex, intricate system of human expressions governed by grammatical rules. It poses a significant challenge to develop capable AI algorithms for comprehending and grasping a language. As a major approach, language modeling has been widely studied for language understanding and generation in the past two decades, evolving from statistical language models to neural language models. Recently, pre-trained language models (PLMs) have been proposed by pre-training Transformer models over large-scale corpora, showing strong capabilities in solving various NLP tasks. Since researchers have found that model scaling can lead to performance improvement, they further study the scaling effect by increasing the model size to an even larger size. Interestingly, when the parameter scale exceeds a certain level, these enlarged language models not only achieve a significant performance improvement but also show some special abilities that are not present in small-scale language models. To discriminate the difference in parameter scale, the research community has coined the term large language models (LLM) for the PLMs of significant size. Recently, the research on LLMs has been largely advanced by both academia and industry, and a remarkable progress is the launch of ChatGPT, which has attracted widespread attention from society. The technical evolution of LLMs has been making an important impact on the entire AI community, which would revolutionize the way how we develop and use AI algorithms. In this survey, we review the recent advances of LLMs by introducing the background, key findings, and mainstream techniques. In particular, we focus on four major aspects of LLMs, namely pre-training, adaptation tuning, utilization, and capacity evaluation. Besides, we also summarize the available resources for developing LLMs and discuss the remaining issues for future directions."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Benign Overfitting in Linear Classifiers and Leaky ReLU Networks from KKT Conditions for Margin Maximization",
                    "abstract": "Linear classifiers and leaky ReLU networks trained by gradient flow on the logistic loss have an implicit bias towards solutions which satisfy the Karush--Kuhn--Tucker (KKT) conditions for margin maximization. In this work we establish a number of settings where the satisfaction of these KKT conditions implies benign overfitting in linear classifiers and in two-layer leaky ReLU networks: the estimators interpolate noisy training data and simultaneously generalize well to test data. The settings include variants of the noisy class-conditional Gaussians considered in previous work as well as new distributional settings where benign overfitting has not been previously observed. The key ingredient to our proof is the observation that when the training data is nearly-orthogonal, both linear classifiers and leaky ReLU networks satisfying the KKT conditions for their respective margin maximization problems behave like a nearly uniform average of the training examples."
                },
                {
                    "title": "The Flan Collection: Designing Data and Methods for Effective Instruction Tuning",
                    "abstract": "We study the design decisions of publicly available instruction tuning methods, and break down the development of Flan 2022 (Chung et al., 2022). Through careful ablation studies on the Flan Collection of tasks and methods, we tease apart the effect of design decisions which enable Flan-T5 to outperform prior work by 3-17%+ across evaluation settings. We find task balancing and enrichment techniques are overlooked but critical to effective instruction tuning, and in particular, training with mixed prompt settings (zero-shot, few-shot, and chain-of-thought) actually yields stronger (2%+) performance in all settings. In further experiments, we show Flan-T5 requires less finetuning to converge higher and faster than T5 on single downstream tasks, motivating instruction-tuned models as more computationally-efficient starting checkpoints for new tasks. Finally, to accelerate research on instruction tuning, we make the Flan 2022 collection of datasets, templates, and methods publicly available at https://github.com/google-research/FLAN/tree/main/flan/v2."
                },
                {
                    "title": "Scaling Instruction-Finetuned Language Models",
                    "abstract": "Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models."
                },
                {
                    "title": "Is Reinforcement Learning (Not) for Natural Language Processing?: Benchmarks, Baselines, and Building Blocks for Natural Language Policy Optimization",
                    "abstract": "We tackle the problem of aligning pre-trained large language models (LMs) with human preferences. If we view text generation as a sequential decision-making problem, reinforcement learning (RL) appears to be a natural conceptual framework. However, using RL for LM-based generation faces empirical challenges, including training instability due to the combinatorial action space, as well as a lack of open-source libraries and benchmarks customized for LM alignment. Thus, a question rises in the research community: is RL a practical paradigm for NLP? To help answer this, we first introduce an open-source modular library, RL4LMs (Reinforcement Learning for Language Models), for optimizing language generators with RL. The library consists of on-policy RL algorithms that can be used to train any encoder or encoder-decoder LM in the HuggingFace library (Wolf et al. 2020) with an arbitrary reward function. Next, we present the GRUE (General Reinforced-language Understanding Evaluation) benchmark, a set of 6 language generation tasks which are supervised not by target strings, but by reward functions which capture automated measures of human preference. GRUE is the first leaderboard-style evaluation of RL algorithms for NLP tasks. Finally, we introduce an easy-to-use, performant RL algorithm, NLPO (Natural Language Policy Optimization) that learns to effectively reduce the combinatorial action space in language generation. We show 1) that RL techniques are generally better than supervised methods at aligning LMs to human preferences; and 2) that NLPO exhibits greater stability and performance than previous policy gradient methods (e.g., PPO (Schulman et al. 2017)), based on both automatic and human evaluations."
                },
                {
                    "title": "Quark: Controllable Text Generation with Reinforced Unlearning",
                    "abstract": "Large-scale language models often learn behaviors that are misaligned with user expectations. Generated text may contain offensive or toxic language, contain significant repetition, or be of a different sentiment than desired by the user. We consider the task of unlearning these misalignments by fine-tuning the language model on signals of what not to do. We introduce Quantized Reward Konditioning (Quark), an algorithm for optimizing a reward function that quantifies an (un)wanted property, while not straying too far from the original model. Quark alternates between (i) collecting samples with the current language model, (ii) sorting them into quantiles based on reward, with each quantile identified by a reward token prepended to the language model's input, and (iii) using a standard language modeling loss on samples from each quantile conditioned on its reward token, while remaining nearby the original language model via a KL-divergence penalty. By conditioning on a high-reward token at generation time, the model generates text that exhibits less of the unwanted property. For unlearning toxicity, negative sentiment, and repetition, our experiments show that Quark outperforms both strong baselines and state-of-the-art reinforcement learning methods like PPO (Schulman et al. 2017), while relying only on standard language modeling primitives."
                },
                {
                    "title": "Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks",
                    "abstract": "How well can NLP models generalize to a variety of unseen tasks when provided with task instructions? To address this question, we first introduce Super-NaturalInstructions, a benchmark of 1,616 diverse NLP tasks and their expert-written instructions. Our collection covers 76 distinct task types, including but not limited to classification, extraction, infilling, sequence tagging, text rewriting, and text composition. This large and diverse collection of tasks enables rigorous benchmarking of cross-task generalization under instructions\u2014training models to follow instructions on a subset of tasks and evaluating them on the remaining unseen ones.Furthermore, we build Tk-Instruct, a transformer model trained to follow a variety of in-context instructions (plain language task definitions or k-shot examples). Our experiments show that Tk-Instruct outperforms existing instruction-following models such as InstructGPT by over 9% on our benchmark despite being an order of magnitude smaller. We further analyze generalization as a function of various scaling parameters, such as the number of observed tasks, the number of instances per task, and model sizes. We hope our dataset and model facilitate future progress towards more general-purpose NLP models."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "LaMDA: Language Models for Dialog Applications",
                    "abstract": "We present LaMDA: Language Models for Dialog Applications. LaMDA is a family of Transformer-based neural language models specialized for dialog, which have up to 137B parameters and are pre-trained on 1.56T words of public dialog data and web text. While model scaling alone can improve quality, it shows less improvements on safety and factual grounding. We demonstrate that fine-tuning with annotated data and enabling the model to consult external knowledge sources can lead to significant improvements towards the two key challenges of safety and factual grounding. The first challenge, safety, involves ensuring that the model's responses are consistent with a set of human values, such as preventing harmful suggestions and unfair bias. We quantify safety using a metric based on an illustrative set of human values, and we find that filtering candidate responses using a LaMDA classifier fine-tuned with a small amount of crowdworker-annotated data offers a promising approach to improving model safety. The second challenge, factual grounding, involves enabling the model to consult external knowledge sources, such as an information retrieval system, a language translator, and a calculator. We quantify factuality using a groundedness metric, and we find that our approach enables the model to generate responses grounded in known sources, rather than responses that merely sound plausible. Finally, we explore the use of LaMDA in the domains of education and content recommendations, and analyze their helpfulness and role consistency."
                },
                {
                    "title": "An Alternate Policy Gradient Estimator for Softmax Policies",
                    "abstract": "Policy gradient (PG) estimators are ine\ufb00ec-tive in dealing with softmax policies that are sub-optimally saturated, which refers to the situation when the policy concentrates its probability mass on sub-optimal actions. Sub-optimal policy saturation may arise from bad policy initialization or sudden changes in the environment that occur after the policy has already converged. Current softmax PG estimators require a large number of updates to overcome policy saturation, which causes low sample e\ufb03ciency and poor adaptability to new situations. To mitigate this problem, we propose a novel PG estimator for softmax policies that utilizes the bias in the critic estimate and the noise present in the reward signal to escape the saturated regions of the policy parameter space. Our theoretical analysis and experiments, conducted on bandits and various reinforcement learning environments, show that this new estimator is signi\ufb01cantly more robust to policy saturation. for the gradient bandit estimators with true r \u03c0 and an additional normally distributed gradient noise. The individual graph labels (0 , 0 . 1 , 0 . 5 , 1 , 2) show the standard deviation of the normal noise added to the gradient. The subplot headers show the policy saturation at the beginning of the experiment."
                },
                {
                    "title": "Finetuned Language Models Are Zero-Shot Learners",
                    "abstract": "This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning."
                },
                {
                    "title": "Continuous vs. Discrete Optimization of Deep Neural Networks",
                    "abstract": "Existing analyses of optimization in deep learning are either continuous, focusing on (variants of) gradient flow, or discrete, directly treating (variants of) gradient descent. Gradient flow is amenable to theoretical analysis, but is stylized and disregards computational efficiency. The extent to which it represents gradient descent is an open question in the theory of deep learning. The current paper studies this question. Viewing gradient descent as an approximate numerical solution to the initial value problem of gradient flow, we find that the degree of approximation depends on the curvature around the gradient flow trajectory. We then show that over deep neural networks with homogeneous activations, gradient flow trajectories enjoy favorable curvature, suggesting they are well approximated by gradient descent. This finding allows us to translate an analysis of gradient flow over deep linear neural networks into a guarantee that gradient descent efficiently converges to global minimum almost surely under random initialization. Experiments suggest that over simple deep neural networks, gradient descent with conventional step size is indeed close to gradient flow. We hypothesize that the theory of gradient flows will unravel mysteries behind deep learning."
                },
                {
                    "title": "Revisiting the Weaknesses of Reinforcement Learning for Neural Machine Translation",
                    "abstract": "Policy gradient algorithms have found wide adoption in NLP, but have recently become subject to criticism, doubting their suitability for NMT. Choshen et al. (2020) identify multiple weaknesses and suspect that their success is determined by the shape of output distributions rather than the reward. In this paper, we revisit these claims and study them under a wider range of configurations. Our experiments on in-domain and cross-domain adaptation reveal the importance of exploration and reward scaling, and provide empirical counter-evidence to these claims."
                },
                {
                    "title": "Learning to summarize from human feedback",
                    "abstract": "As language models become more powerful, training and evaluation are increasingly bottlenecked by the data and metrics used for a particular task. For example, summarization models are often trained to predict human reference summaries and evaluated using ROUGE, but both of these metrics are rough proxies for what we really care about---summary quality. In this work, we show that it is possible to significantly improve summary quality by training a model to optimize for human preferences. We collect a large, high-quality dataset of human comparisons between summaries, train a model to predict the human-preferred summary, and use that model as a reward function to fine-tune a summarization policy using reinforcement learning. We apply our method to a version of the TL;DR dataset of Reddit posts and find that our models significantly outperform both human reference summaries and much larger models fine-tuned with supervised learning alone. Our models also transfer to CNN/DM news articles, producing summaries nearly as good as the human reference without any news-specific fine-tuning. We conduct extensive analyses to understand our human feedback dataset and fine-tuned models We establish that our reward model generalizes to new datasets, and that optimizing our reward model results in better summaries than optimizing ROUGE according to humans. We hope the evidence from our paper motivates machine learning researchers to pay closer attention to how their training loss affects the model behavior they actually want."
                },
                {
                    "title": "Implicit Regularization in Deep Learning May Not Be Explainable by Norms",
                    "abstract": "Mathematically characterizing the implicit regularization induced by gradient-based optimization is a longstanding pursuit in the theory of deep learning. A widespread hope is that a characterization based on minimization of norms may apply, and a standard test-bed for studying this prospect is matrix factorization (matrix completion via linear neural networks). It is an open question whether norms can explain the implicit regularization in matrix factorization. The current paper resolves this open question in the negative, by proving that there exist natural matrix factorization problems on which the implicit regularization drives all norms (and quasi-norms) towards infinity. Our results suggest that, rather than perceiving the implicit regularization via norms, a potentially more useful interpretation is minimization of rank. We demonstrate empirically that this interpretation extends to a certain class of non-linear neural networks, and hypothesize that it may be key to explaining generalization in deep learning."
                },
                {
                    "title": "On the Global Convergence Rates of Softmax Policy Gradient Methods",
                    "abstract": "We make three contributions toward better understanding policy gradient methods in the tabular setting. First, we show that with the true gradient, policy gradient with a softmax parametrization converges at a $O(1/t)$ rate, with constants depending on the problem and initialization. This result significantly expands the recent asymptotic convergence results. The analysis relies on two findings: that the softmax policy gradient satisfies a \u0141ojasiewicz inequality, and the minimum probability of an optimal action during optimization can be bounded in terms of its initial value. Second, we analyze entropy regularized policy gradient and show that it enjoys a significantly faster linear convergence rate $O(e^{-t})$ toward softmax optimal policy. This result resolves an open question in the recent literature. Finally, combining the above two results and additional new $\\Omega(1/t)$ lower bound results, we explain how entropy regularization improves policy optimization, even with the true gradient, from the perspective of convergence rate. The separation of rates is further explained using the notion of non-uniform \u0141ojasiewicz degree. These results provide a theoretical understanding of the impact of entropy and corroborate existing empirical studies."
                },
                {
                    "title": "ToTTo: A Controlled Table-To-Text Generation Dataset",
                    "abstract": "We present ToTTo, an open-domain English table-to-text dataset with over 120,000 training examples that proposes a controlled generation task: given a Wikipedia table and a set of highlighted table cells, produce a one-sentence description. To obtain generated targets that are natural but also faithful to the source table, we introduce a dataset construction process where annotators directly revise existing candidate sentences from Wikipedia. We present systematic analyses of our dataset and annotation process as well as results achieved by several state-of-the-art baselines. While usually fluent, existing methods often hallucinate phrases that are not supported by the table, suggesting that this dataset can serve as a useful research benchmark for high-precision conditional text generation."
                },
                {
                    "title": "CommonGen: A Constrained Text Generation Challenge for Generative Commonsense Reasoning",
                    "abstract": "Recently, large-scale pre-trained language models have demonstrated impressive performance on several commonsense-reasoning benchmark datasets. However, building machines with commonsense to compose realistically plausible sentences remains challenging. In this paper, we present a constrained text generation task, CommonGen associated with a benchmark dataset, to explicitly test machines for the ability of generative commonsense reasoning. Given a set of common concepts (e.g., dog, frisbee, catch, throw); the task is to generate a coherent sentence describing an everyday scenario using these concepts (e.g., \u201ca man throws a frisbee and his dog catches it\u201d). The CommonGen task is challenging because it inherently requires 1) relational reasoning with background commonsense knowledge and 2) compositional generalization ability to work on unseen concept combinations. Our dataset, constructed through a combination of crowdsourced and existing caption corpora, consists of 77k commonsense descriptions over 35k unique concept-sets. Experiments show that there is a large gap between state-of-the-art text generation models (e.g., T5) and human performance (31.6% v.s. 63.5% in SPICE metric). Furthermore, we demonstrate that the learned generative commonsense reasoning capability can be transferred to improve downstream tasks such as CommonsenseQA (76.9% to 78.4 in dev accuracy) by generating additional context."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "Well-Read Students Learn Better: On the Importance of Pre-training Compact Models",
                    "abstract": "Recent developments in natural language representations have been accompanied by large and expensive models that leverage vast amounts of general-domain text through self-supervised pre-training. Due to the cost of applying such models to down-stream tasks, several model compression techniques on pre-trained language representations have been proposed (Sun et al., 2019; Sanh, 2019). However, surprisingly, the simple baseline of just pre-training and fine-tuning compact models has been overlooked. In this paper, we first show that pre-training remains important in the context of smaller architectures, and fine-tuning pre-trained compact models can be competitive to more elaborate methods proposed in concurrent work. Starting with pre-trained compact models, we then explore transferring task knowledge from large fine-tuned models through standard knowledge distillation. The resulting simple, yet effective and general algorithm, Pre-trained Distillation, brings further improvements. Through extensive experiments, we more generally explore the interaction between pre-training and distillation under two variables that have been under-studied: model size and properties of unlabeled task data. One surprising observation is that they have a compound effect even when sequentially applied on the same data. To accelerate future research, we will make our 24 pre-trained miniature BERT models publicly available."
                },
                {
                    "title": "Fine-Tuning Language Models from Human Preferences",
                    "abstract": "Reward learning enables the application of reinforcement learning (RL) to tasks where reward is defined by human judgment, building a model of reward by asking humans questions. Most work on reward learning has used simulated environments, but complex information about values is often expressed in natural language, and we believe reward learning for language is a key to making RL practical and safe for real-world tasks. In this paper, we build on advances in generative pretraining of language models to apply reward learning to four natural language tasks: continuing text with positive sentiment or physically descriptive language, and summarization tasks on the TL;DR and CNN/Daily Mail datasets. For stylistic continuation we achieve good results with only 5,000 comparisons evaluated by humans. For summarization, models trained with 60,000 comparisons copy whole sentences from the input but skip irrelevant preamble; this leads to reasonable ROUGE scores and very good performance according to our human labelers, but may be exploiting the fact that labelers rely on simple heuristics."
                },
                {
                    "title": "On the Theory of Policy Gradient Methods: Optimality, Approximation, and Distribution Shift",
                    "abstract": "Policy gradient methods are among the most effective methods in challenging reinforcement learning problems with large state and/or action spaces. However, little is known about even their most basic theoretical convergence properties, including: if and how fast they converge to a globally optimal solution or how they cope with approximation error due to using a restricted class of parametric policies. This work provides provable characterizations of the computational, approximation, and sample size properties of policy gradient methods in the context of discounted Markov Decision Processes (MDPs). We focus on both: \"tabular\" policy parameterizations, where the optimal policy is contained in the class and where we show global convergence to the optimal policy; and parametric policy classes (considering both log-linear and neural policy classes), which may not contain the optimal policy and where we provide agnostic learning results. One central contribution of this work is in providing approximation guarantees that are average case -- which avoid explicit worst-case dependencies on the size of state space -- by making a formal connection to supervised learning under distribution shift. This characterization shows an important interplay between estimation error, approximation error, and exploration (as characterized through a precisely defined condition number)."
                },
                {
                    "title": "On the Weaknesses of Reinforcement Learning for Neural Machine Translation",
                    "abstract": "Reinforcement learning (RL) is frequently used to increase performance in text generation tasks, including machine translation (MT), notably through the use of Minimum Risk Training (MRT) and Generative Adversarial Networks (GAN). However, little is known about what and how these methods learn in the context of MT. We prove that one of the most common RL methods for MT does not optimize the expected reward, as well as show that other methods take an infeasibly long time to converge. In fact, our results suggest that RL practices in MT are likely to improve performance only where the pre-trained parameters are already close to yielding the correct translation. Our findings further suggest that observed gains may be due to effects unrelated to the training signal, but rather from changes in the shape of the distribution curve."
                },
                {
                    "title": "Way Off-Policy Batch Deep Reinforcement Learning of Implicit Human Preferences in Dialog",
                    "abstract": "Most deep reinforcement learning (RL) systems are not able to learn effectively from off-policy data, especially if they cannot explore online in the environment. These are critical shortcomings for applying RL to real-world problems where collecting data is expensive, and models must be tested offline before being deployed to interact with the environment -- e.g. systems that learn from human interaction. Thus, we develop a novel class of off-policy batch RL algorithms, which are able to effectively learn offline, without exploring, from a fixed batch of human interaction data. We leverage models pre-trained on data as a strong prior, and use KL-control to penalize divergence from this prior during RL training. We also use dropout-based uncertainty estimates to lower bound the target Q-values as a more efficient alternative to Double Q-Learning. The algorithms are tested on the problem of open-domain dialog generation -- a challenging reinforcement learning problem with a 20,000-dimensional action space. Using our Way Off-Policy algorithm, we can extract multiple different reward functions post-hoc from collected human interaction data, and learn effectively from all of these. We test the real-world generalization of these systems by deploying them live to converse with humans in an open-domain setting, and demonstrate that our algorithm achieves significant improvements over prior methods in off-policy batch RL."
                },
                {
                    "title": "XLNet: Generalized Autoregressive Pretraining for Language Understanding",
                    "abstract": "With the capability of modeling bidirectional contexts, denoising autoencoding based pretraining like BERT achieves better performance than pretraining approaches based on autoregressive language modeling. However, relying on corrupting the input with masks, BERT neglects dependency between the masked positions and suffers from a pretrain-finetune discrepancy. In light of these pros and cons, we propose XLNet, a generalized autoregressive pretraining method that (1) enables learning bidirectional contexts by maximizing the expected likelihood over all permutations of the factorization order and (2) overcomes the limitations of BERT thanks to its autoregressive formulation. Furthermore, XLNet integrates ideas from Transformer-XL, the state-of-the-art autoregressive model, into pretraining. Empirically, under comparable experiment settings, XLNet outperforms BERT on 20 tasks, often by a large margin, including question answering, natural language inference, sentiment analysis, and document ranking."
                },
                {
                    "title": "Gradient Descent Maximizes the Margin of Homogeneous Neural Networks",
                    "abstract": "In this paper, we study the implicit regularization of the gradient descent algorithm in homogeneous neural networks, including fully-connected and convolutional neural networks with ReLU or LeakyReLU activations. In particular, we study the gradient descent or gradient flow (i.e., gradient descent with infinitesimal step size) optimizing the logistic loss or cross-entropy loss of any homogeneous model (possibly non-smooth), and show that if the training loss decreases below a certain threshold, then we can define a smoothed version of the normalized margin which increases over time. We also formulate a natural constrained optimization problem related to margin maximization, and prove that both the normalized margin and its smoothed version converge to the objective value at a KKT point of the optimization problem. Our results generalize the previous results for logistic regression with one-layer or multi-layer linear networks, and provide more quantitative convergence results with weaker assumptions than previous results for homogeneous smooth neural networks. We conduct several experiments to justify our theoretical finding on MNIST and CIFAR-10 datasets. Finally, as margin is closely related to robustness, we discuss potential benefits of training longer for improving the robustness of the model."
                },
                {
                    "title": "Ray Interference: a Source of Plateaus in Deep Reinforcement Learning",
                    "abstract": "Rather than proposing a new method, this paper investigates an issue present in existing learning algorithms. We study the learning dynamics of reinforcement learning (RL), specifically a characteristic coupling between learning and data generation that arises because RL agents control their future data distribution. In the presence of function approximation, this coupling can lead to a problematic type of 'ray interference', characterized by learning dynamics that sequentially traverse a number of performance plateaus, effectively constraining the agent to learn one thing at a time even when learning in parallel is better. We establish the conditions under which ray interference occurs, show its relation to saddle points and obtain the exact learning dynamics in a restricted setting. We characterize a number of its properties and discuss possible remedies."
                },
                {
                    "title": "Understanding the impact of entropy on policy optimization",
                    "abstract": "Entropy regularization is commonly used to improve policy optimization in reinforcement learning. It is believed to help with \\emph{exploration} by encouraging the selection of more stochastic policies. In this work, we analyze this claim using new visualizations of the optimization landscape based on randomly perturbing the loss function. We first show that even with access to the exact gradient, policy optimization is difficult due to the geometry of the objective function. Then, we qualitatively show that in some environments, a policy with higher entropy can make the optimization landscape smoother, thereby connecting local optima and enabling the use of larger learning rates. This paper presents new tools for understanding the optimization landscape, shows that policy entropy serves as a regularizer, and highlights the challenge of designing general-purpose policy optimization algorithms."
                },
                {
                    "title": "Gradient Descent Provably Optimizes Over-parameterized Neural Networks",
                    "abstract": "One of the mysteries in the success of neural networks is randomly initialized first order methods like gradient descent can achieve zero training loss even though the objective function is non-convex and non-smooth. This paper demystifies this surprising phenomenon for two-layer fully connected ReLU activated neural networks. For an $m$ hidden node shallow neural network with ReLU activation and $n$ training data, we show as long as $m$ is large enough and no two inputs are parallel, randomly initialized gradient descent converges to a globally optimal solution at a linear convergence rate for the quadratic loss function. \nOur analysis relies on the following observation: over-parameterization and random initialization jointly restrict every weight vector to be close to its initialization for all iterations, which allows us to exploit a strong convexity-like property to show that gradient descent converges at a global linear rate to the global optimum. We believe these insights are also useful in analyzing deep models and other first order methods."
                },
                {
                    "title": "A Call for Clarity in Reporting BLEU Scores",
                    "abstract": "The field of machine translation faces an under-recognized problem because of inconsistency in the reporting of scores from its dominant metric. Although people refer to \u201cthe\u201d BLEU score, BLEU is in fact a parameterized metric whose values can vary wildly with changes to these parameters. These parameters are often not reported or are hard to find, and consequently, BLEU scores between papers cannot be directly compared. I quantify this variation, finding differences as high as 1.8 between commonly used configurations. The main culprit is different tokenization and normalization schemes applied to the reference. Pointing to the success of the parsing community, I suggest machine translation researchers settle upon the BLEU scheme used by the annual Conference on Machine Translation (WMT), which does not allow for user-supplied reference processing, and provide a new tool, SACREBLEU, to facilitate this."
                },
                {
                    "title": "Learning to Extract Coherent Summary via Deep Reinforcement Learning",
                    "abstract": "\n \n Coherence plays a critical role in producing a high-quality summary from a document. In recent years, neural extractive summarization is becoming increasingly attractive. However, most of them ignore the coherence of summaries when extracting sentences. As an effort towards extracting coherent summaries, we propose a neural coherence model to capture the cross-sentence semantic and syntactic coherence patterns. The proposed neural coherence model obviates the need for feature engineering and can be trained in an end-to-end fashion using unlabeled data. Empirical results show that the proposed neural coherence model can efficiently capture the cross-sentence coherence patterns. Using the combined output of the neural coherence model and ROUGE package as the reward, we design a reinforcement learning method to train a proposed neural extractive summarizer which is named Reinforced Neural Extractive Summarization (RNES) model. The RNES model learns to optimize coherence and informative importance of the summary simultaneously. The experimental results show that the proposed RNES outperforms existing baselines and achieves state-of-the-art performance in term of ROUGE on CNN/Daily Mail dataset. The qualitative evaluation indicates that summaries produced by RNES are more coherent and readable.\n \n"
                },
                {
                    "title": "The Mirage of Action-Dependent Baselines in Reinforcement Learning",
                    "abstract": "Policy gradient methods are a widely used class of model-free reinforcement learning algorithms where a state-dependent baseline is used to reduce gradient estimator variance. Several recent papers extend the baseline to depend on both the state and action and suggest that this significantly reduces variance and improves sample efficiency without introducing bias into the gradient estimates. To better understand this development, we decompose the variance of the policy gradient estimator and numerically show that learned state-action-dependent baselines do not in fact reduce variance over a state-dependent baseline in commonly tested benchmark domains. We confirm this unexpected result by reviewing the open-source code accompanying these prior papers, and show that subtle implementation decisions cause deviations from the methods presented in the papers and explain the source of the previously observed empirical gains. Furthermore, the variance decomposition highlights areas for improvement, which we demonstrate by illustrating a simple change to the typical value function parameterization that can significantly improve performance."
                },
                {
                    "title": "The NarrativeQA Reading Comprehension Challenge",
                    "abstract": "Reading comprehension (RC)\u2014in contrast to information retrieval\u2014requires integrating information and reasoning about events, entities, and their relations across a full document. Question answering is conventionally used to assess RC ability, in both artificial agents and children learning to read. However, existing RC datasets and tasks are dominated by questions that can be solved by selecting answers using superficial information (e.g., local context similarity or global term frequency); they thus fail to test for the essential integrative aspect of RC. To encourage progress on deeper comprehension of language, we present a new dataset and set of tasks in which the reader must answer questions about stories by reading entire books or movie scripts. These tasks are designed so that successfully answering their questions requires understanding the underlying narrative rather than relying on shallow pattern matching or salience. We show that although humans solve the tasks easily, standard RC models struggle on the tasks presented here. We provide an analysis of the dataset and the challenges it presents."
                },
                {
                    "title": "End-to-End Offline Goal-Oriented Dialog Policy Learning via Policy Gradient",
                    "abstract": "Learning a goal-oriented dialog policy is generally performed offline with supervised learning algorithms or online with reinforcement learning (RL). Additionally, as companies accumulate massive quantities of dialog transcripts between customers and trained human agents, encoder-decoder methods have gained popularity as agent utterances can be directly treated as supervision without the need for utterance-level annotations. However, one potential drawback of such approaches is that they myopically generate the next agent utterance without regard for dialog-level considerations. To resolve this concern, this paper describes an offline RL method for learning from unannotated corpora that can optimize a goal-oriented policy at both the utterance and dialog level. We introduce a novel reward function and use both on-policy and off-policy policy gradient to learn a policy offline without requiring online user interaction or an explicit state space definition."
                },
                {
                    "title": "Automatic differentiation in PyTorch",
                    "abstract": "In this article, we describe an automatic differentiation module of PyTorch \u2014 a library designed to enable rapid research on machine learning models. It builds upon a few projects, most notably Lua Torch, Chainer, and HIPS Autograd [4], and provides a high performance environment with easy access to automatic differentiation of models executed on different devices (CPU and GPU). To make prototyping easier, PyTorch does not follow the symbolic approach used in many other deep learning frameworks, but focuses on differentiation of purely imperative programs, with a focus on extensibility and low overhead. Note that this preprint is a draft of certain sections from an upcoming paper covering all PyTorch features."
                },
                {
                    "title": "DailyDialog: A Manually Labelled Multi-turn Dialogue Dataset",
                    "abstract": "We develop a high-quality multi-turn dialog dataset, DailyDialog, which is intriguing in several aspects. The language is human-written and less noisy. The dialogues in the dataset reflect our daily communication way and cover various topics about our daily life. We also manually label the developed dataset with communication intention and emotion information. Then, we evaluate existing approaches on DailyDialog dataset and hope it benefit the research field of dialog systems. The dataset is available on http://yanran.li/dailydialog"
                },
                {
                    "title": "SemEval-2017 Task 1: Semantic Textual Similarity Multilingual and Crosslingual Focused Evaluation",
                    "abstract": "Semantic Textual Similarity (STS) measures the meaning similarity of sentences. Applications include machine translation (MT), summarization, generation, question answering (QA), short answer grading, semantic search, dialog and conversational systems. The STS shared task is a venue for assessing the current state-of-the-art. The 2017 task focuses on multilingual and cross-lingual pairs with one sub-track exploring MT quality estimation (MTQE) data. The task obtained strong participation from 31 teams, with 17 participating in all language tracks. We summarize performance and review a selection of well performing methods. Analysis highlights common errors, providing insight into the limitations of existing models. To support ongoing work on semantic representations, the STS Benchmark is introduced as a new shared training and evaluation set carefully selected from the corpus of English STS shared task data (2012-2017)."
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "Reinforcement Learning for Bandit Neural Machine Translation with Simulated Human Feedback",
                    "abstract": "Machine translation is a natural candidate problem for reinforcement learning from human feedback: users provide quick, dirty ratings on candidate translations to guide a system to improve. Yet, current neural machine translation training focuses on expensive human-generated reference translations. We describe a reinforcement learning algorithm that improves neural machine translation systems from simulated human feedback. Our algorithm combines the advantage actor-critic algorithm (Mnih et al., 2016) with the attention-based neural encoder-decoder architecture (Luong et al., 2015). This algorithm (a) is well-designed for problems with a large action space and delayed rewards, (b) effectively optimizes traditional corpus-level machine translation metrics, and (c) is robust to skewed, high-variance, granular feedback modeled after actual human behaviors."
                },
                {
                    "title": "A Deep Reinforced Model for Abstractive Summarization",
                    "abstract": "Attentional, RNN-based encoder-decoder models for abstractive summarization have achieved good performance on short input and output sequences. For longer documents and summaries however these models often include repetitive and incoherent phrases. We introduce a neural network model with a novel intra-attention that attends over the input and continuously generated output separately, and a new training method that combines standard supervised word prediction and reinforcement learning (RL). Models trained only with supervised learning often exhibit \"exposure bias\" - they assume ground truth is provided at each step during training. However, when standard word prediction is combined with the global sequence prediction training of RL the resulting summaries become more readable. We evaluate this model on the CNN/Daily Mail and New York Times datasets. Our model obtains a 41.16 ROUGE-1 score on the CNN/Daily Mail dataset, an improvement over previous state-of-the-art models. Human evaluation also shows that our model produces higher quality summaries."
                },
                {
                    "title": "An Actor-Critic Algorithm for Sequence Prediction",
                    "abstract": "We present an approach to training neural networks to generate sequences using actor-critic methods from reinforcement learning (RL). Current log-likelihood training methods are limited by the discrepancy between their training and testing modes, as models must generate tokens conditioned on their previous guesses rather than the ground-truth tokens. We address this problem by introducing a \\textit{critic} network that is trained to predict the value of an output token, given the policy of an \\textit{actor} network. This results in a training procedure that is much closer to the test phase, and allows us to directly optimize for a task-specific score such as BLEU. Crucially, since we leverage these techniques in the supervised learning setting rather than the traditional RL setting, we condition the critic network on the ground-truth output. We show that our method leads to improved performance on both a synthetic task, and for German-English machine translation. Our analysis paves the way for such methods to be applied in natural language generation tasks, such as machine translation, caption generation, and dialogue modelling."
                },
                {
                    "title": "Deep Reinforcement Learning for Dialogue Generation",
                    "abstract": "Recent neural models of dialogue generation offer great promise for generating responses for conversational agents, but tend to be shortsighted, predicting utterances one at a time while ignoring their influence on future outcomes. Modeling the future direction of a dialogue is crucial to generating coherent, interesting dialogues, a need which led traditional NLP models of dialogue to draw on reinforcement learning. In this paper, we show how to integrate these goals, applying deep reinforcement learning to model future reward in chatbot dialogue. The model simulates dialogues between two virtual agents, using policy gradient methods to reward sequences that display three useful conversational properties: informativity (non-repetitive turns), coherence, and ease of answering (related to forward-looking function). We evaluate our model on diversity, length as well as with human judges, showing that the proposed algorithm generates more interactive responses and manages to foster a more sustained conversation in dialogue simulation. This work marks a first step towards learning a neural conversational model based on the long-term success of dialogues."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Sequence Level Training with Recurrent Neural Networks",
                    "abstract": "Many natural language processing applications use language models to generate text. These models are typically trained to predict the next word in a sequence, given the previous words and some context such as an image. However, at test time the model is expected to generate the entire sequence from scratch. This discrepancy makes generation brittle, as errors may accumulate along the way. We address this issue by proposing a novel sequence level training algorithm that directly optimizes the metric used at test time, such as BLEU or ROUGE. On three different tasks, our approach outperforms several strong baselines for greedy generation. The method is also competitive when these baselines employ beam search, while being several times faster."
                },
                {
                    "title": "Teaching Machines to Read and Comprehend",
                    "abstract": "Teaching machines to read natural language documents remains an elusive challenge. Machine reading systems can be tested on their ability to answer questions posed on the contents of documents that they have seen, but until now large scale training and test datasets have been missing for this type of evaluation. In this work we define a new methodology that resolves this bottleneck and provides large scale supervised reading comprehension data. This allows us to develop a class of attention based deep neural networks that learn to read real documents and answer complex questions with minimal prior knowledge of language structure."
                },
                {
                    "title": "High-Dimensional Continuous Control Using Generalized Advantage Estimation",
                    "abstract": "Policy gradient methods are an appealing approach in reinforcement learning because they directly optimize the cumulative reward and can straightforwardly be used with nonlinear function approximators such as neural networks. The two main challenges are the large number of samples typically required, and the difficulty of obtaining stable and steady improvement despite the nonstationarity of the incoming data. We address the first challenge by using value functions to substantially reduce the variance of policy gradient estimates at the cost of some bias, with an exponentially-weighted estimator of the advantage function that is analogous to TD(lambda). We address the second challenge by using trust region optimization procedure for both the policy and the value function, which are represented by neural networks. \nOur approach yields strong empirical results on highly challenging 3D locomotion tasks, learning running gaits for bipedal and quadrupedal simulated robots, and learning a policy for getting the biped to stand up from starting out lying on the ground. In contrast to a body of prior work that uses hand-crafted policy representations, our neural network policies map directly from raw kinematics to joint torques. Our algorithm is fully model-free, and the amount of simulated experience required for the learning tasks on 3D bipeds corresponds to 1-2 weeks of real time."
                },
                {
                    "title": "Exact solutions to the nonlinear dynamics of learning in deep linear neural networks",
                    "abstract": "Despite the widespread practical success of deep learning methods, our theoretical understanding of the dynamics of learning in deep neural networks remains quite sparse. We attempt to bridge the gap between the theory and practice of deep learning by systematically analyzing learning dynamics for the restricted case of deep linear neural networks. Despite the linearity of their input-output map, such networks have nonlinear gradient descent dynamics on weights that change with the addition of each new hidden layer. We show that deep linear networks exhibit nonlinear learning phenomena similar to those seen in simulations of nonlinear networks, including long plateaus followed by rapid transitions to lower error solutions, and faster convergence from greedy unsupervised pretraining initial conditions than from random initial conditions. We provide an analytical description of these phenomena by finding new exact solutions to the nonlinear dynamics of deep learning. Our theoretical analysis also reveals the surprising finding that as the depth of a network approaches infinity, learning speed can nevertheless remain finite: for a special class of initial conditions on the weights, very deep networks incur only a finite, depth independent, delay in learning speed relative to shallow networks. We show that, under certain conditions on the training data, unsupervised pretraining can find this special class of initial conditions, while scaled random Gaussian initializations cannot. We further exhibit a new class of random orthogonal initial conditions on weights that, like unsupervised pre-training, enjoys depth independent learning times. We further show that these initial conditions also lead to faithful propagation of gradients even in deep nonlinear networks, as long as they operate in a special regime known as the edge of chaos."
                },
                {
                    "title": "Learning Word Vectors for Sentiment Analysis",
                    "abstract": "Unsupervised vector-based approaches to semantics can model rich lexical meanings, but they largely fail to capture sentiment information that is central to many word meanings and important for a wide range of NLP tasks. We present a model that uses a mix of unsupervised and supervised techniques to learn word vectors capturing semantic term--document information as well as rich sentiment content. The proposed model can leverage both continuous and multi-dimensional sentiment information as well as non-sentiment annotations. We instantiate the model to utilize the document-level sentiment polarity annotations present in many online documents (e.g. star ratings). We evaluate the model using small, widely used sentiment and subjectivity corpora and find it out-performs several previously introduced methods for sentiment classification. We also introduce a large dataset of movie reviews to serve as a more robust benchmark for work in this area."
                },
                {
                    "title": "METEOR: An Automatic Metric for MT Evaluation with Improved Correlation with Human Judgments",
                    "abstract": "We describe METEOR, an automatic metric for machine translation evaluation that is based on a generalized concept of unigram matching between the machineproduced translation and human-produced reference translations. Unigrams can be matched based on their surface forms, stemmed forms, and meanings; furthermore, METEOR can be easily extended to include more advanced matching strategies. Once all generalized unigram matches between the two strings have been found, METEOR computes a score for this matching using a combination of unigram-precision, unigram-recall, and a measure of fragmentation that is designed to directly capture how well-ordered the matched words in the machine translation are in relation to the reference. We evaluate METEOR by measuring the correlation between the metric scores and human judgments of translation quality. We compute the Pearson R correlation value between its scores and human quality assessments of the LDC TIDES 2003 Arabic-to-English and Chinese-to-English datasets. We perform segment-bysegment correlation, and show that METEOR gets an R correlation value of 0.347 on the Arabic data and 0.331 on the Chinese data. This is shown to be an improvement on using simply unigramprecision, unigram-recall and their harmonic F1 combination. We also perform experiments to show the relative contributions of the various mapping modules."
                },
                {
                    "title": "ROUGE: A Package for Automatic Evaluation of Summaries",
                    "abstract": "ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST."
                },
                {
                    "title": "Variance Reduction Techniques for Gradient Estimates in Reinforcement Learning",
                    "abstract": "Policy gradient methods for reinforcement learning avoid some of the undesirable properties of the value function approaches, such as policy degradation (Baxter and Bartlett, 2001). However, the variance of the performance gradient estimates obtained from the simulation is sometimes excessive. In this paper, we consider variance reduction methods that were developed for Monte Carlo estimates of integrals. We study two commonly used policy gradient techniques, the baseline and actor-critic methods, from this perspective. Both can be interpreted as additive control variate variance reduction methods. We consider the expected average reward performance measure, and we focus on the GPOMDP algorithm for estimating performance gradients in partially observable Markov decision processes controlled by stochastic reactive policies. We give bounds for the estimation error of the gradient estimates for both baseline and actor-critic algorithms, in terms of the sample size and mixing properties of the controlled system. For the baseline technique, we compute the optimal baseline, and show that the popular approach of using the average reward to define the baseline can be suboptimal. For actor-critic algorithms, we show that using the true value function as the critic can be suboptimal. We also discuss algorithms for estimating the optimal baseline and approximate value function."
                },
                {
                    "title": "Escaping the Gravitational Pull of Softmax",
                    "abstract": "The softmax is the standard transformation used in machine learning to map realvalued vectors to categorical distributions. Unfortunately, this transform poses serious drawbacks for gradient descent (ascent) optimization. We reveal this difficulty by establishing two negative results: (1) optimizing any expectation with respect to the softmax must exhibit sensitivity to parameter initialization (\u201csoftmax gravity well\u201d), and (2) optimizing log-probabilities under the softmax must exhibit slow convergence (\u201csoftmax damping\u201d). Both findings are based on an analysis of convergence rates using the Non-uniform \u0141ojasiewicz (N\u0141) inequalities. To circumvent these shortcomings we investigate an alternative transformation, the escort mapping, that demonstrates better optimization properties. The disadvantages of the softmax and the effectiveness of the escort transformation are further explained using the concept of N\u0141 coefficient. In addition to proving bounds on convergence rates to firmly establish these results, we also provide experimental evidence for the superiority of the escort transformation."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Neural Replicator Dynamics",
                    "abstract": "In multiagent learning, agents interact in inherently nonstationary environments due to their concurrent policy updates. It is, therefore, paramount to develop and analyze algorithms that learn effectively despite these nonstationarities. A number of works have successfully conducted this analysis under the lens of evolutionary game theory (EGT), wherein a population of individuals interact and evolve based on biologically-inspired operators. These studies have mainly focused on establishing connections to value-iteration based approaches in stateless or tabular games. We extend this line of inquiry to formally establish links between EGT and policy gradient (PG) methods, which have been extensively applied in single and multiagent learning. We pinpoint weaknesses of the commonly-used softmax PG algorithm in adversarial and nonstationary settings and contrast PG\u2019s behavior to that predicted by replicator dynamics (RD), a central model in EGT. We consequently provide theoretical results that establish links between EGT and PG methods, then derive Neural Replicator Dynamics (NeuRD), a parameterized version of RD that constitutes a novel method with several advantages. First, as NeuRD reduces to the well-studied no-regret Hedge algorithm in the tabular setting, it inherits no-regret guarantees that enable convergence to equilibria in games. Second, NeuRD is shown to be more adaptive to nonstationarity, in comparison to PG, when learning in canonical games and imperfect information benchmarks including Poker. Thirdly, modifying any PG-based algorithm to use the NeuRD"
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Overview of the IWSLT 2017 Evaluation Campaign",
                    "abstract": "The IWSLT 2017 evaluation campaign has organised three tasks. The Multilingual task, which is about training machine translation systems handling many-to-many language directions, including so-called zero-shot directions. The Dialogue task, which calls for the integration of context information in machine translation, in order to resolve anaphoric references that typically occur in human-human dialogue turns. And, finally, the Lecture task, which offers the challenge of automatically transcribing and translating real-life university lectures. Following the tradition of these reports, we will described all tasks in detail and present the results of all runs submitted by their participants."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "The mnist database of handwritten digits",
                    "abstract": "Disclosed is an improved articulated bar flail having shearing edges for efficiently shredding materials. An improved shredder cylinder is disclosed with a plurality of these flails circumferentially spaced and pivotally attached to the periphery of a rotatable shaft. Also disclosed is an improved shredder apparatus which has a pair of these shredder cylinders mounted to rotate about spaced parallel axes which cooperates with a conveyer apparatus which has a pair of inclined converging conveyer belts with one of the belts mounted to move with respect to the other belt to allow the transport of articles of various sizes therethrough."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CL",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively address the issue of vanishing gradients in reinforcement finetuning (RFT) of language models, particularly when the reward standard deviation is low?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of vanishing gradients in RFT is crucial for advancing the field of machine learning, particularly in natural language processing. By addressing this issue, we can improve the performance of language models in real-world applications, leading to more effective human-computer interactions and better alignment with human preferences. This research could pave the way for more robust reinforcement learning techniques, enhancing the adaptability of models to diverse tasks and datasets, and ultimately contributing to the development of more intelligent systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenge of vanishing gradients in RFT arises from the interplay between the reward maximization objective and the softmax operator used for token distribution. When the reward standard deviation is low, the expected gradient becomes near-zero, making it difficult to optimize the model effectively. Naive approaches that do not account for the distribution of rewards may fail to improve model performance, as they overlook the critical relationship between reward variability and gradient computation. Additionally, the complexity of aligning model behavior with human preferences adds another layer of difficulty, as it requires nuanced understanding and representation of those preferences.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on the broader aspects of reinforcement learning and human feedback without specifically addressing the nuances of gradient behavior in RFT. The lack of empirical studies highlighting the prevalence of low reward standard deviation inputs has contributed to this oversight. Existing solutions have not adequately tackled the optimization challenges posed by these inputs, leading to a gap in understanding how to effectively train models under such conditions. Our approach differs by directly investigating the relationship between reward standard deviation and gradient behavior, providing a targeted solution to this specific problem.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a detailed analysis of the GRUE benchmark to empirically assess the impact of low reward standard deviation on RFT performance. We will utilize policy gradient algorithms, specifically Proximal Policy Optimization (PPO), while incorporating modifications to address the vanishing gradient issue. The dataset will consist of various inputs from the GRUE benchmark, and we will measure performance using metrics that capture both reward maximization and gradient behavior. We expect to demonstrate"
            }
        },
        "author_data": {
            "d76f3980-dde0-47a7-84bc-b19d828711c0": {
                "pk": "d76f3980-dde0-47a7-84bc-b19d828711c0",
                "name": "Noam Razin",
                "collaborators": [
                    "Nadav Cohen",
                    "Yotam Alexander",
                    "Asaf Maman",
                    "Itzik Malkiel",
                    "Oren Barkan",
                    "Avi Caciularu",
                    "Ori Katz",
                    "Noam Koenigstein",
                    "Edo Cohen-Karlik",
                    "Raja Giryes"
                ],
                "domain": [
                    "Deep Learning",
                    "Implicit Regularization",
                    "Graph Neural Network",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Understanding Deep Learning via Notions of Rank",
                        "abstract": "Despite the extreme popularity of deep learning in science and industry, its formal understanding is limited. This thesis puts forth notions of rank as key for developing a theory of deep learning, focusing on the fundamental aspects of generalization and expressiveness. In particular, we establish that gradient-based training can induce an implicit regularization towards low rank for several neural network architectures, and demonstrate empirically that this phenomenon may facilitate an explanation of generalization over natural data (e.g., audio, images, and text). Then, we characterize the ability of graph neural networks to model interactions via a notion of rank, which is commonly used for quantifying entanglement in quantum physics. A central tool underlying these results is a connection between neural networks and tensor factorizations. Practical implications of our theory for designing explicit regularization schemes and data preprocessing algorithms are presented."
                    },
                    {
                        "title": "Implicit Bias of Policy Gradient in Linear Quadratic Control: Extrapolation to Unseen Initial States",
                        "abstract": "In modern machine learning, models can often fit training data in numerous ways, some of which perform well on unseen (test) data, while others do not. Remarkably, in such cases gradient descent frequently exhibits an implicit bias that leads to excellent performance on unseen data. This implicit bias was extensively studied in supervised learning, but is far less understood in optimal control (reinforcement learning). There, learning a controller applied to a system via gradient descent is known as policy gradient, and a question of prime importance is the extent to which a learned controller extrapolates to unseen initial states. This paper theoretically studies the implicit bias of policy gradient in terms of extrapolation to unseen initial states. Focusing on the fundamental Linear Quadratic Regulator (LQR) problem, we establish that the extent of extrapolation depends on the degree of exploration induced by the system when commencing from initial states included in training. Experiments corroborate our theory, and demonstrate its conclusions on problems beyond LQR, where systems are non-linear and controllers are neural networks. We hypothesize that real-world optimal control may be greatly improved by developing methods for informed selection of initial states to train on."
                    },
                    {
                        "title": "Lecture Notes on Linear Neural Networks: A Tale of Optimization and Generalization in Deep Learning",
                        "abstract": "These notes are based on a lecture delivered by NC on March 2021, as part of an advanced course in Princeton University on the mathematical understanding of deep learning. They present a theory (developed by NC, NR and collaborators) of linear neural networks -- a fundamental model in the study of optimization and generalization in deep learning. Practical applications born from the presented theory are also discussed. The theory is based on mathematical tools that are dynamical in nature. It showcases the potential of such tools to push the envelope of our understanding of optimization and generalization in deep learning. The text assumes familiarity with the basics of statistical learning theory. Exercises (without solutions) are included."
                    },
                    {
                        "title": "Unintentional Unalignment: Likelihood Displacement in Direct Preference Optimization",
                        "abstract": "Direct Preference Optimization (DPO) and its variants are increasingly used for aligning language models with human preferences. Although these methods are designed to teach a model to generate preferred responses more frequently relative to dispreferred responses, prior work has observed that the likelihood of preferred responses often decreases during training. The current work sheds light on the causes and implications of this counter-intuitive phenomenon, which we term likelihood displacement. We demonstrate that likelihood displacement can be catastrophic, shifting probability mass from preferred responses to responses with an opposite meaning. As a simple example, training a model to prefer $\\texttt{No}$ over $\\texttt{Never}$ can sharply increase the probability of $\\texttt{Yes}$. Moreover, when aligning the model to refuse unsafe prompts, we show that such displacement can unintentionally lead to unalignment, by shifting probability mass from preferred refusal responses to harmful responses (e.g., reducing the refusal rate of Llama-3-8B-Instruct from 74.4% to 33.4%). We theoretically characterize that likelihood displacement is driven by preferences that induce similar embeddings, as measured by a centered hidden embedding similarity (CHES) score. Empirically, the CHES score enables identifying which training samples contribute most to likelihood displacement in a given dataset. Filtering out these samples effectively mitigated unintentional unalignment in our experiments. More broadly, our results highlight the importance of curating data with sufficiently distinct preferences, for which we believe the CHES score may prove valuable."
                    },
                    {
                        "title": "The Implicit Bias of Structured State Space Models Can Be Poisoned With Clean Labels",
                        "abstract": "Neural networks are powered by an implicit bias: a tendency of gradient descent to fit training data in a way that generalizes to unseen data. A recent class of neural network models gaining increasing popularity is structured state space models (SSMs), regarded as an efficient alternative to transformers. Prior work argued that the implicit bias of SSMs leads to generalization in a setting where data is generated by a low dimensional teacher. In this paper, we revisit the latter setting, and formally establish a phenomenon entirely undetected by prior work on the implicit bias of SSMs. Namely, we prove that while implicit bias leads to generalization under many choices of training data, there exist special examples whose inclusion in training completely distorts the implicit bias, to a point where generalization fails. This failure occurs despite the special training examples being labeled by the teacher, i.e. having clean labels! We empirically demonstrate the phenomenon, with SSMs trained independently and as part of non-linear neural networks. In the area of adversarial machine learning, disrupting generalization with cleanly labeled training examples is known as clean-label poisoning. Given the proliferation of SSMs, particularly in large language models, we believe significant efforts should be invested in further delineating their susceptibility to clean-label poisoning, and in developing methods for overcoming this susceptibility."
                    },
                    {
                        "title": "What Algorithms can Transformers Learn? A Study in Length Generalization",
                        "abstract": "Large language models exhibit surprising emergent generalization properties, yet also struggle on many simple reasoning tasks such as arithmetic and parity. This raises the question of if and when Transformer models can learn the true algorithm for solving a task. We study the scope of Transformers' abilities in the specific setting of length generalization on algorithmic tasks. Here, we propose a unifying framework to understand when and how Transformers can exhibit strong length generalization on a given task. Specifically, we leverage RASP (Weiss et al., 2021) -- a programming language designed for the computational model of a Transformer -- and introduce the RASP-Generalization Conjecture: Transformers tend to length generalize on a task if the task can be solved by a short RASP program which works for all input lengths. This simple conjecture remarkably captures most known instances of length generalization on algorithmic tasks. Moreover, we leverage our insights to drastically improve generalization performance on traditionally hard tasks (such as parity and addition). On the theoretical side, we give a simple example where the\"min-degree-interpolator\"model of learning from Abbe et al. (2023) does not correctly predict Transformers' out-of-distribution behavior, but our conjecture does. Overall, our work provides a novel perspective on the mechanisms of compositional generalization and the algorithmic capabilities of Transformers."
                    },
                    {
                        "title": "What Makes Data Suitable for a Locally Connected Neural Network? A Necessary and Sufficient Condition Based on Quantum Entanglement",
                        "abstract": "The question of what makes a data distribution suitable for deep learning is a fundamental open problem. Focusing on locally connected neural networks (a prevalent family of architectures that includes convolutional and recurrent neural networks as well as local self-attention models), we address this problem by adopting theoretical tools from quantum physics. Our main theoretical result states that a certain locally connected neural network is capable of accurate prediction over a data distribution if and only if the data distribution admits low quantum entanglement under certain canonical partitions of features. As a practical application of this result, we derive a preprocessing method for enhancing the suitability of a data distribution to locally connected neural networks. Experiments with widespread models over various datasets demonstrate our findings. We hope that our use of quantum entanglement will encourage further adoption of tools from physics for formally reasoning about the relation between deep learning and real-world data."
                    },
                    {
                        "title": "Implicit Regularization in Hierarchical Tensor Factorization and Deep Convolutional Neural Networks",
                        "abstract": "In the pursuit of explaining implicit regularization in deep learning, prominent focus was given to matrix and tensor factorizations, which correspond to simplified neural networks. It was shown that these models exhibit an implicit tendency towards low matrix and tensor ranks, respectively. Drawing closer to practical deep learning, the current paper theoretically analyzes the implicit regularization in hierarchical tensor factorization, a model equivalent to certain deep convolutional neural networks. Through a dynamical systems lens, we overcome challenges associated with hierarchy, and establish implicit regularization towards low hierarchical tensor rank. This translates to an implicit regularization towards locality for the associated convolutional networks. Inspired by our theory, we design explicit regularization discouraging locality, and demonstrate its ability to improve the performance of modern convolutional networks on non-local tasks, in defiance of conventional wisdom by which architectural changes are needed. Our work highlights the potential of enhancing neural networks via theoretical analysis of their implicit regularization."
                    },
                    {
                        "title": "On the Ability of Graph Neural Networks to Model Interactions Between Vertices",
                        "abstract": "Graph neural networks (GNNs) are widely used for modeling complex interactions between entities represented as vertices of a graph. Despite recent efforts to theoretically analyze the expressive power of GNNs, a formal characterization of their ability to model interactions is lacking. The current paper aims to address this gap. Formalizing strength of interactions through an established measure known as separation rank, we quantify the ability of certain GNNs to model interaction between a given subset of vertices and its complement, i.e. between the sides of a given partition of input vertices. Our results reveal that the ability to model interaction is primarily determined by the partition's walk index -- a graph-theoretical characteristic defined by the number of walks originating from the boundary of the partition. Experiments with common GNN architectures corroborate this finding. As a practical application of our theory, we design an edge sparsification algorithm named Walk Index Sparsification (WIS), which preserves the ability of a GNN to model interactions when input edges are removed. WIS is simple, computationally efficient, and in our experiments has markedly outperformed alternative methods in terms of induced prediction accuracy. More broadly, it showcases the potential of improving GNNs by theoretically analyzing the interactions they can model."
                    },
                    {
                        "title": "Implicit Regularization in Tensor Factorization",
                        "abstract": "Recent efforts to unravel the mystery of implicit regularization in deep learning have led to a theoretical focus on matrix factorization -- matrix completion via linear neural network. As a step further towards practical deep learning, we provide the first theoretical analysis of implicit regularization in tensor factorization -- tensor completion via certain type of non-linear neural network. We circumvent the notorious difficulty of tensor problems by adopting a dynamical systems perspective, and characterizing the evolution induced by gradient descent. The characterization suggests a form of greedy low tensor rank search, which we rigorously prove under certain conditions, and empirically demonstrate under others. Motivated by tensor rank capturing the implicit regularization of a non-linear neural network, we empirically explore it as a measure of complexity, and find that it captures the essence of datasets on which neural networks generalize. This leads us to believe that tensor rank may pave way to explaining both implicit regularization in deep learning, and the properties of real-world data translating this implicit regularization to generalization."
                    },
                    {
                        "title": "Implicit Regularization in Deep Learning May Not Be Explainable by Norms",
                        "abstract": "Mathematically characterizing the implicit regularization induced by gradient-based optimization is a longstanding pursuit in the theory of deep learning. A widespread hope is that a characterization based on minimization of norms may apply, and a standard test-bed for studying this prospect is matrix factorization (matrix completion via linear neural networks). It is an open question whether norms can explain the implicit regularization in matrix factorization. The current paper resolves this open question in the negative, by proving that there exist natural matrix factorization problems on which the implicit regularization drives all norms (and quasi-norms) towards infinity. Our results suggest that, rather than perceiving the implicit regularization via norms, a potentially more useful interpretation is minimization of rank. We demonstrate empirically that this interpretation extends to a certain class of non-linear neural networks, and hypothesize that it may be key to explaining generalization in deep learning."
                    },
                    {
                        "title": "Optimizing BERT for Unlabeled Text-Based Items Similarity",
                        "abstract": "Language models that utilize extensive self-supervised pre-training from unlabeled text, have recently shown to significantly advance the state-of-the-art performance in a variety of language understanding tasks. However, it is yet unclear if and how these recent models can be harnessed for conducting text-based recommendations. In this work, we introduce RecoBERT, a BERT-based approach for learning catalog-specialized language models for text-based item recommendations. We suggest novel training and inference procedures for scoring similarities between pairs of items, that don\u2019t require item similarity labels. Both the training and the inference techniques were designed to utilize the unlabeled structure of textual catalogs, and minimize the discrepancy between them. By incorporating four scores during inference, RecoBERT can infer text-based item-to-item similarities more accurately than other techniques. In addition, we introduce a new language understanding task for wine recommendations using similarities based on professional wine reviews. As an additional contribution, we publish annotated recommendations dataset crafted by human wine experts. Finally, we evaluate RecoBERT and compare it to various state-of-the-art NLP models on wine and fashion recommendations tasks."
                    },
                    {
                        "title": "Scalable Attentive Sentence-Pair Modeling via Distilled Sentence Embedding",
                        "abstract": "Recent state-of-the-art natural language understanding models, such as BERT and XLNet, score a pair of sentences (A and B) using multiple cross-attention operations - a process in which each word in sentence A attends to all words in sentence B and vice versa. As a result, computing the similarity between a query sentence and a set of candidate sentences, requires the propagation of all query-candidate sentence-pairs throughout a stack of cross-attention layers. This exhaustive process becomes computationally prohibitive when the number of candidate sentences is large. In contrast, sentence embedding techniques learn a sentence-to-vector mapping and compute the similarity between the sentence vectors via simple elementary operations. In this paper, we introduce Distilled Sentence Embedding (DSE) - a model that is based on knowledge distillation from cross-attentive models, focusing on sentence-pair tasks. The outline of DSE is as follows: Given a cross-attentive teacher model (e.g. a fine-tuned BERT), we train a sentence embedding based student model to reconstruct the sentence-pair scores obtained by the teacher model. We empirically demonstrate the effectiveness of DSE on five GLUE sentence-pair tasks. DSE significantly outperforms several ELMO variants and other sentence embedding methods, while accelerating computation of the query-candidate sentence-pairs similarities by several orders of magnitude, with an average relative degradation of 4.6% compared to BERT. Furthermore, we show that DSE produces sentence embeddings that reach state-of-the-art performance on universal sentence representation benchmarks. Our code is made publicly available at this https URL."
                    }
                ]
            },
            "3a23efcd-ad62-42c6-b258-3ba0be089617": {
                "pk": "3a23efcd-ad62-42c6-b258-3ba0be089617",
                "name": "Hattie Zhou",
                "collaborators": [
                    "Preetum Nakkiran",
                    "Arwen Bradley",
                    "Xinting Huang",
                    "Andy Yang",
                    "S. Bhattamishra",
                    "Yash Sarrof",
                    "Andreas Krebs",
                    "Michael Hahn",
                    "Madhu Advani",
                    "Etai Littwin"
                ],
                "domain": [
                    "Transformers",
                    "Length Generalization",
                    "Machine Learning",
                    "Diffusion Models"
                ],
                "publications": [
                    {
                        "title": "A Formal Framework for Understanding Length Generalization in Transformers",
                        "abstract": "A major challenge for transformers is generalizing to sequences longer than those observed during training. While previous works have empirically shown that transformers can either succeed or fail at length generalization depending on the task, theoretical understanding of this phenomenon remains limited. In this work, we introduce a rigorous theoretical framework to analyze length generalization in causal transformers with learnable absolute positional encodings. In particular, we characterize those functions that are identifiable in the limit from sufficiently long inputs with absolute positional encodings under an idealized inference scheme using a norm-based regularizer. This enables us to prove the possibility of length generalization for a rich family of problems. We experimentally validate the theory as a predictor of success and failure of length generalization across a range of algorithmic and formal language tasks. Our theory not only explains a broad set of empirical observations but also opens the way to provably predicting length generalization capabilities in transformers."
                    },
                    {
                        "title": "Step-by-Step Diffusion: An Elementary Tutorial",
                        "abstract": "We present an accessible first course on diffusion models and flow matching for machine learning, aimed at a technical audience with no diffusion experience. We try to simplify the mathematical details as much as possible (sometimes heuristically), while retaining enough precision to derive correct algorithms."
                    },
                    {
                        "title": "What Algorithms can Transformers Learn? A Study in Length Generalization",
                        "abstract": "Large language models exhibit surprising emergent generalization properties, yet also struggle on many simple reasoning tasks such as arithmetic and parity. This raises the question of if and when Transformer models can learn the true algorithm for solving a task. We study the scope of Transformers' abilities in the specific setting of length generalization on algorithmic tasks. Here, we propose a unifying framework to understand when and how Transformers can exhibit strong length generalization on a given task. Specifically, we leverage RASP (Weiss et al., 2021) -- a programming language designed for the computational model of a Transformer -- and introduce the RASP-Generalization Conjecture: Transformers tend to length generalize on a task if the task can be solved by a short RASP program which works for all input lengths. This simple conjecture remarkably captures most known instances of length generalization on algorithmic tasks. Moreover, we leverage our insights to drastically improve generalization performance on traditionally hard tasks (such as parity and addition). On the theoretical side, we give a simple example where the\"min-degree-interpolator\"model of learning from Abbe et al. (2023) does not correctly predict Transformers' out-of-distribution behavior, but our conjecture does. Overall, our work provides a novel perspective on the mechanisms of compositional generalization and the algorithmic capabilities of Transformers."
                    }
                ]
            },
            "5d72a55c-910e-4534-a2c3-c090d42db4da": {
                "pk": "5d72a55c-910e-4534-a2c3-c090d42db4da",
                "name": "Omid Saremi",
                "collaborators": [
                    "Etai Littwin",
                    "Vimal Thilak",
                    "Josh Susskind",
                    "Preetum Nakkiran",
                    "Chen Huang",
                    "Samy Bengio",
                    "J. Susskind",
                    "Laurent Dinh",
                    "Hanlin Goh",
                    "Shuangfei Zhai"
                ],
                "domain": [
                    "Self-Supervised Learning",
                    "Transformers",
                    "Neural Networks",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "How JEPA Avoids Noisy Features: The Implicit Bias of Deep Linear Self Distillation Networks",
                        "abstract": "Two competing paradigms exist for self-supervised learning of data representations. Joint Embedding Predictive Architecture (JEPA) is a class of architectures in which semantically similar inputs are encoded into representations that are predictive of each other. A recent successful approach that falls under the JEPA framework is self-distillation, where an online encoder is trained to predict the output of the target encoder, sometimes using a lightweight predictor network. This is contrasted with the Masked AutoEncoder (MAE) paradigm, where an encoder and decoder are trained to reconstruct missing parts of the input in the data space rather, than its latent representation. A common motivation for using the JEPA approach over MAE is that the JEPA objective prioritizes abstract features over fine-grained pixel information (which can be unpredictable and uninformative). In this work, we seek to understand the mechanism behind this empirical observation by analyzing the training dynamics of deep linear models. We uncover a surprising mechanism: in a simplified linear setting where both approaches learn similar representations, JEPAs are biased to learn high-influence features, i.e., features characterized by having high regression coefficients. Our results point to a distinct implicit bias of predicting in latent space that may shed light on its success in practice."
                    },
                    {
                        "title": "What Algorithms can Transformers Learn? A Study in Length Generalization",
                        "abstract": "Large language models exhibit surprising emergent generalization properties, yet also struggle on many simple reasoning tasks such as arithmetic and parity. This raises the question of if and when Transformer models can learn the true algorithm for solving a task. We study the scope of Transformers' abilities in the specific setting of length generalization on algorithmic tasks. Here, we propose a unifying framework to understand when and how Transformers can exhibit strong length generalization on a given task. Specifically, we leverage RASP (Weiss et al., 2021) -- a programming language designed for the computational model of a Transformer -- and introduce the RASP-Generalization Conjecture: Transformers tend to length generalize on a task if the task can be solved by a short RASP program which works for all input lengths. This simple conjecture remarkably captures most known instances of length generalization on algorithmic tasks. Moreover, we leverage our insights to drastically improve generalization performance on traditionally hard tasks (such as parity and addition). On the theoretical side, we give a simple example where the\"min-degree-interpolator\"model of learning from Abbe et al. (2023) does not correctly predict Transformers' out-of-distribution behavior, but our conjecture does. Overall, our work provides a novel perspective on the mechanisms of compositional generalization and the algorithmic capabilities of Transformers."
                    },
                    {
                        "title": "Adaptivity and Modularity for Efficient Generalization Over Task Complexity",
                        "abstract": "Can transformers generalize efficiently on problems that require dealing with examples with different levels of difficulty? We introduce a new task tailored to assess generalization over different complexities and present results that indicate that standard transformers face challenges in solving these tasks. These tasks are variations of pointer value retrieval previously introduced by Zhang et al. (2021). We investigate how the use of a mechanism for adaptive and modular computation in transformers facilitates the learning of tasks that demand generalization over the number of sequential computation steps (i.e., the depth of the computation graph). Based on our observations, we propose a transformer-based architecture called Hyper-UT, which combines dynamic function generation from hyper networks with adaptive depth from Universal Transformers. This model demonstrates higher accuracy and a fairer allocation of computational resources when generalizing to higher numbers of computation steps. We conclude that mechanisms for adaptive depth and modularity complement each other in improving efficient generalization concerning example complexity. Additionally, to emphasize the broad applicability of our findings, we illustrate that in a standard image recognition task, Hyper- UT's performance matches that of a ViT model but with considerably reduced computational demands (achieving over 70\\% average savings by effectively using fewer layers)."
                    },
                    {
                        "title": "When can transformers reason with abstract symbols?",
                        "abstract": "We investigate the capabilities of transformer models on relational reasoning tasks. In these tasks, models are trained on a set of strings encoding abstract relations, and are then tested out-of-distribution on data that contains symbols that did not appear in the training dataset. We prove that for any relational reasoning task in a large family of tasks, transformers learn the abstract relations and generalize to the test set when trained by gradient descent on sufficiently large quantities of training data. This is in contrast to classical fully-connected networks, which we prove fail to learn to reason. Our results inspire modifications of the transformer architecture that add only two trainable parameters per head, and that we empirically demonstrate improve data efficiency for learning to reason."
                    },
                    {
                        "title": "LiDAR: Sensing Linear Probing Performance in Joint Embedding SSL Architectures",
                        "abstract": "Joint embedding (JE) architectures have emerged as a promising avenue for acquiring transferable data representations. A key obstacle to using JE methods, however, is the inherent challenge of evaluating learned representations without access to a downstream task, and an annotated dataset. Without efficient and reliable evaluation, it is difficult to iterate on architectural and training choices for JE methods. In this paper, we introduce LiDAR (Linear Discriminant Analysis Rank), a metric designed to measure the quality of representations within JE architectures. Our metric addresses several shortcomings of recent approaches based on feature covariance rank by discriminating between informative and uninformative features. In essence, LiDAR quantifies the rank of the Linear Discriminant Analysis (LDA) matrix associated with the surrogate SSL task -- a measure that intuitively captures the information content as it pertains to solving the SSL task. We empirically demonstrate that LiDAR significantly surpasses naive rank based approaches in its predictive power of optimal hyperparameters. Our proposed criterion presents a more robust and intuitive means of assessing the quality of representations within JE architectures, which we hope facilitates broader adoption of these powerful techniques in various domains."
                    },
                    {
                        "title": "The Slingshot Mechanism: An Empirical Study of Adaptive Optimizers and the Grokking Phenomenon",
                        "abstract": "The grokking phenomenon as reported by Power et al. ( arXiv:2201.02177 ) refers to a regime where a long period of overfitting is followed by a seemingly sudden transition to perfect generalization. In this paper, we attempt to reveal the underpinnings of Grokking via a series of empirical studies. Specifically, we uncover an optimization anomaly plaguing adaptive optimizers at extremely late stages of training, referred to as the Slingshot Mechanism. A prominent artifact of the Slingshot Mechanism can be measured by the cyclic phase transitions between stable and unstable training regimes, and can be easily monitored by the cyclic behavior of the norm of the last layers weights. We empirically observe that without explicit regularization, Grokking as reported in ( arXiv:2201.02177 ) almost exclusively happens at the onset of Slingshots, and is absent without it. While common and easily reproduced in more general settings, the Slingshot Mechanism does not follow from any known optimization theories that we are aware of, and can be easily overlooked without an in depth examination. Our work points to a surprising and useful inductive bias of adaptive gradient optimizers at late stages of training, calling for a revised theoretical analysis of their origin."
                    },
                    {
                        "title": "Implicit Acceleration and Feature Learning in Infinitely Wide Neural Networks with Bottlenecks",
                        "abstract": "We analyze the learning dynamics of infinitely wide neural networks with a finite sized bottle-neck. Unlike the neural tangent kernel limit, a bottleneck in an otherwise infinite width network al-lows data dependent feature learning in its bottle-neck representation. We empirically show that a single bottleneck in infinite networks dramatically accelerates training when compared to purely in-finite networks, with an improved overall performance. We discuss the acceleration phenomena by drawing similarities to infinitely wide deep linear models, where the acceleration effect of a bottleneck can be understood theoretically."
                    },
                    {
                        "title": "Implicit Greedy Rank Learning in Autoencoders via Overparameterized Linear Networks",
                        "abstract": "Deep linear networks trained with gradient descent yield low rank solutions, as is typically studied in matrix factorization. In this paper, we take a step further and analyze implicit rank regularization in autoencoders. We show greedy learning of low-rank latent codes induced by a linear sub-network at the autoencoder bottleneck. We further propose orthogonal initialization and principled learning rate adjustment to mitigate sensitivity of training dynamics to spectral prior and linear depth. With linear autoencoders on synthetic data, our method converges stably to ground-truth latent code rank. With nonlinear autoencoders, our method converges to latent ranks optimal for downstream classification and image sampling."
                    },
                    {
                        "title": "Hawking-Page transition in holographic massive gravity Citation",
                        "abstract": "Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. The MIT Faculty has made this article openly available. Please share how this access benefits you. Your story matters. We study the Hawking-Page transition in a holographic model of field theories with momentum dissipation. We find that the deconfinement temperature strictly decreases as momentum dissipation is increased. For sufficiently strong momentum dissipation, the critical temperature goes to zero, indicating a zero-temperature deconfinement transition in the dual field theory."
                    },
                    {
                        "title": "Hawking-Page transition in holographic massive gravity",
                        "abstract": "We study the Hawking-Page transition in a holographic model of field theories with momentum dissipation. We find that the deconfinement temperature strictly decreases as momentum dissipation is increased. For sufficiently strong momentum dissipation, the critical temperature goes to zero, indicating a zero-temperature deconfinement transition in the dual field theory."
                    },
                    {
                        "title": "Monte Carlo Simulation of The Adjoint Coulomb Gas",
                        "abstract": "Monte Carlo simulation results for unitary matrix quantum mechanics, describing two-dimensional Yang-Mills theory coupled to a finite density of non-dynamical quarks (adjoint Coulomb gas), are presented. We characterize the deconfining transition in this model, by measuring the Polyakov Loop Susceptibility and employing finite-size scaling analysis. We provide evidence that the phase transition is first-order. Our results are consistent with the outcome of earlier large-$N$ studies of the model."
                    }
                ]
            },
            "fd3b6ee3-429c-4f05-be73-59134b5528b7": {
                "pk": "fd3b6ee3-429c-4f05-be73-59134b5528b7",
                "name": "Vimal Thilak",
                "collaborators": [
                    "C. Creusere",
                    "D. Voelz",
                    "Etai Littwin",
                    "O. Saremi",
                    "Chen Huang",
                    "Josh Susskind",
                    "J. Susskind",
                    "Preetum Nakkiran",
                    "Laurent Dinh",
                    "Hanlin Goh"
                ],
                "domain": [
                    "Self-Supervised Learning",
                    "Polarimetry",
                    "Image Processing",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Towards Automatic Assessment of Self-Supervised Speech Models using Rank",
                        "abstract": "This study explores using embedding rank as an unsupervised evaluation metric for general-purpose speech encoders trained via self-supervised learning (SSL). Traditionally, assessing the performance of these encoders is resource-intensive and requires labeled data from the downstream tasks. Inspired by the vision domain, where embedding rank has shown promise for evaluating image encoders without tuning on labeled downstream data, this work examines its applicability in the speech domain, considering the temporal nature of the signals. The findings indicate rank correlates with downstream performance within encoder layers across various downstream tasks and for in- and out-of-domain scenarios. However, rank does not reliably predict the best-performing layer for specific downstream tasks, as lower-ranked layers can outperform higher-ranked ones. Despite this limitation, the results suggest that embedding rank can be a valuable tool for monitoring training progress in SSL speech models, offering a less resource-demanding alternative to traditional evaluation methods."
                    },
                    {
                        "title": "How JEPA Avoids Noisy Features: The Implicit Bias of Deep Linear Self Distillation Networks",
                        "abstract": "Two competing paradigms exist for self-supervised learning of data representations. Joint Embedding Predictive Architecture (JEPA) is a class of architectures in which semantically similar inputs are encoded into representations that are predictive of each other. A recent successful approach that falls under the JEPA framework is self-distillation, where an online encoder is trained to predict the output of the target encoder, sometimes using a lightweight predictor network. This is contrasted with the Masked AutoEncoder (MAE) paradigm, where an encoder and decoder are trained to reconstruct missing parts of the input in the data space rather, than its latent representation. A common motivation for using the JEPA approach over MAE is that the JEPA objective prioritizes abstract features over fine-grained pixel information (which can be unpredictable and uninformative). In this work, we seek to understand the mechanism behind this empirical observation by analyzing the training dynamics of deep linear models. We uncover a surprising mechanism: in a simplified linear setting where both approaches learn similar representations, JEPAs are biased to learn high-influence features, i.e., features characterized by having high regression coefficients. Our results point to a distinct implicit bias of predicting in latent space that may shed light on its success in practice."
                    },
                    {
                        "title": "Enhancing JEPAs with Spatial Conditioning: Robust and Efficient Representation Learning",
                        "abstract": "Image-based Joint-Embedding Predictive Architecture (IJEPA) offers an attractive alternative to Masked Autoencoder (MAE) for representation learning using the Masked Image Modeling framework. IJEPA drives representations to capture useful semantic information by predicting in latent rather than input space. However, IJEPA relies on carefully designed context and target windows to avoid representational collapse. The encoder modules in IJEPA cannot adaptively modulate the type of predicted and/or target features based on the feasibility of the masked prediction task as they are not given sufficient information of both context and targets. Based on the intuition that in natural images, information has a strong spatial bias with spatially local regions being highly predictive of one another compared to distant ones. We condition the target encoder and context encoder modules in IJEPA with positions of context and target windows respectively. Our\"conditional\"encoders show performance gains on several image classification benchmark datasets, improved robustness to context window size and sample-efficiency during pretraining."
                    },
                    {
                        "title": "Adaptivity and Modularity for Efficient Generalization Over Task Complexity",
                        "abstract": "Can transformers generalize efficiently on problems that require dealing with examples with different levels of difficulty? We introduce a new task tailored to assess generalization over different complexities and present results that indicate that standard transformers face challenges in solving these tasks. These tasks are variations of pointer value retrieval previously introduced by Zhang et al. (2021). We investigate how the use of a mechanism for adaptive and modular computation in transformers facilitates the learning of tasks that demand generalization over the number of sequential computation steps (i.e., the depth of the computation graph). Based on our observations, we propose a transformer-based architecture called Hyper-UT, which combines dynamic function generation from hyper networks with adaptive depth from Universal Transformers. This model demonstrates higher accuracy and a fairer allocation of computational resources when generalizing to higher numbers of computation steps. We conclude that mechanisms for adaptive depth and modularity complement each other in improving efficient generalization concerning example complexity. Additionally, to emphasize the broad applicability of our findings, we illustrate that in a standard image recognition task, Hyper- UT's performance matches that of a ViT model but with considerably reduced computational demands (achieving over 70\\% average savings by effectively using fewer layers)."
                    },
                    {
                        "title": "LiDAR: Sensing Linear Probing Performance in Joint Embedding SSL Architectures",
                        "abstract": "Joint embedding (JE) architectures have emerged as a promising avenue for acquiring transferable data representations. A key obstacle to using JE methods, however, is the inherent challenge of evaluating learned representations without access to a downstream task, and an annotated dataset. Without efficient and reliable evaluation, it is difficult to iterate on architectural and training choices for JE methods. In this paper, we introduce LiDAR (Linear Discriminant Analysis Rank), a metric designed to measure the quality of representations within JE architectures. Our metric addresses several shortcomings of recent approaches based on feature covariance rank by discriminating between informative and uninformative features. In essence, LiDAR quantifies the rank of the Linear Discriminant Analysis (LDA) matrix associated with the surrogate SSL task -- a measure that intuitively captures the information content as it pertains to solving the SSL task. We empirically demonstrate that LiDAR significantly surpasses naive rank based approaches in its predictive power of optimal hyperparameters. Our proposed criterion presents a more robust and intuitive means of assessing the quality of representations within JE architectures, which we hope facilitates broader adoption of these powerful techniques in various domains."
                    },
                    {
                        "title": "The Slingshot Mechanism: An Empirical Study of Adaptive Optimizers and the Grokking Phenomenon",
                        "abstract": "The grokking phenomenon as reported by Power et al. ( arXiv:2201.02177 ) refers to a regime where a long period of overfitting is followed by a seemingly sudden transition to perfect generalization. In this paper, we attempt to reveal the underpinnings of Grokking via a series of empirical studies. Specifically, we uncover an optimization anomaly plaguing adaptive optimizers at extremely late stages of training, referred to as the Slingshot Mechanism. A prominent artifact of the Slingshot Mechanism can be measured by the cyclic phase transitions between stable and unstable training regimes, and can be easily monitored by the cyclic behavior of the norm of the last layers weights. We empirically observe that without explicit regularization, Grokking as reported in ( arXiv:2201.02177 ) almost exclusively happens at the onset of Slingshots, and is absent without it. While common and easily reproduced in more general settings, the Slingshot Mechanism does not follow from any known optimization theories that we are aware of, and can be easily overlooked without an in depth examination. Our work points to a surprising and useful inductive bias of adaptive gradient optimizers at late stages of training, calling for a revised theoretical analysis of their origin."
                    },
                    {
                        "title": "Implicit Acceleration and Feature Learning in Infinitely Wide Neural Networks with Bottlenecks",
                        "abstract": "We analyze the learning dynamics of infinitely wide neural networks with a finite sized bottle-neck. Unlike the neural tangent kernel limit, a bottleneck in an otherwise infinite width network al-lows data dependent feature learning in its bottle-neck representation. We empirically show that a single bottleneck in infinite networks dramatically accelerates training when compared to purely in-finite networks, with an improved overall performance. We discuss the acceleration phenomena by drawing similarities to infinitely wide deep linear models, where the acceleration effect of a bottleneck can be understood theoretically."
                    },
                    {
                        "title": "Implicit Greedy Rank Learning in Autoencoders via Overparameterized Linear Networks",
                        "abstract": "Deep linear networks trained with gradient descent yield low rank solutions, as is typically studied in matrix factorization. In this paper, we take a step further and analyze implicit rank regularization in autoencoders. We show greedy learning of low-rank latent codes induced by a linear sub-network at the autoencoder bottleneck. We further propose orthogonal initialization and principled learning rate adjustment to mitigate sensitivity of training dynamics to spectral prior and linear depth. With linear autoencoders on synthetic data, our method converges stably to ground-truth latent code rank. With nonlinear autoencoders, our method converges to latent ranks optimal for downstream classification and image sampling."
                    },
                    {
                        "title": "Passive Polarimetric Imagery-Based Material Classification Robust to Illumination Source Position and Viewpoint",
                        "abstract": "Polarization, a property of light that conveys information about the transverse electric field orientation, complements other attributes of electromagnetic radiation such as intensity and frequency. Using multiple passive polarimetric images, we develop an iterative, model-based approach to estimate the complex index of refraction and apply it to target classification."
                    },
                    {
                        "title": "Active Polarimetric Imaging for Estimation of Scene Geometry",
                        "abstract": "Active imaging polarimetry for remote sensing applications has received significant attention recently. Such systems use a variably-polarized active light source to illuminate target objects. Multiple images with different polarizations are then captured and used to build Stokes vectors which are, in turn, used to estimate the refraction indices of materials as well as the relative geometry of the target object. The applications facilitated by active polarimetry include target detection, object recognition, shape extraction and material classification. Unfortunately, this estimation problem requires us to find the solution to a system of nonlinear equations using an iterative optimization technique. In this paper, we introduce a methodology for finding and validating good solutions to this optimization."
                    },
                    {
                        "title": "Estimation of target geometry from Mueller matrix imagery",
                        "abstract": "An imaging polarimeter records the polarization state of light reflected by an object that is illuminated with a polarized source such as a laser. Active polarimetric imagery has been shown to be useful in many remote sensing applications including shape extraction, material classification and target detection/recognition. In this paper, we present a method that automatically extracts the angle of incidence, angle of reflection and the relative azimuthal angle from Mueller matrix imagery. Mueller matrix imagery provides multiple measurements from which we can construct a nonlinear system of equations. This system is solved using the Levenberg-Marquardt algorithm which is a standard nonlinear equation solver. We experimentally demonstrate via computer simulations that the parameter estimates can be estimated accurately using our approach."
                    },
                    {
                        "title": "Passive Polarimetric Imagery Based Material Classification For Remote Sensing Applications",
                        "abstract": "Passive imaging polarimetry has emerged as a useful tool in many remote sensing applications including material classification, target detection and shape extraction. In this paper we present a method to classify specular objects based on their material composition from passive polarimetric imagery. The proposed algorithm is built on an iterative, model-based method to recover the complex index of refraction of a specular target from multiple polarization measurements. The recovered parameters are then used to discriminate between objects by employing the nearest neighbor rule. Experimental results indicate that the classification approach is highly effective for distinguishing between various targets of interest. Most significantly, the proposed classification method is robust to a wide range of observational geometry."
                    },
                    {
                        "title": "Polarization-based index of refraction and reflection angle estimation for remote sensing applications.",
                        "abstract": "A passive-polarization-based imaging system records the polarization state of light reflected by objects that are illuminated with an unpolarized and generally uncontrolled source. Such systems can be useful in many remote sensing applications including target detection, object segmentation, and material classification. We present a method to jointly estimate the complex index of refraction and the reflection angle (reflected zenith angle) of a target from multiple measurements collected by a passive polarimeter. An expression for the degree of polarization is derived from the microfacet polarimetric bidirectional reflectance model for the case of scattering in the plane of incidence. Using this expression, we develop a nonlinear least-squares estimation algorithm for extracting an apparent index of refraction and the reflection angle from a set of polarization measurements collected from multiple source positions. Computer simulation results show that the estimation accuracy generally improves with an increasing number of source position measurements. Laboratory results indicate that the proposed method is effective for recovering the reflection angle and that the estimated index of refraction provides a feature vector that is robust to the reflection angle."
                    },
                    {
                        "title": "Estimation of incidence and reflection angles from passive polarimetric imagery: extension to out-of-plane scattering",
                        "abstract": "Passive polarimetric imagery conveys information that complements the information contained in intensity and spectral imagery. Passive polarimetric measurements have been exploited in many remote sensing applications such as shape extraction, surface inspection and object detection/recognition. In previous work Thilak et al. proposed an algorithm to estimate the index of refraction and view angle (object surface orientation) from multiple polarization images where the source position changes between measurements. That work relies on a specular polarimetric bidirectional reflectance distribution function (pBRDF) developed by Priest and Meier. The pBRDF incorporates a Mueller matrix that characterizes the polarized reflection properties of a target for any incident Stokes vector. The results in Thilak et al. assumed that scattering occurs in the plane of incidence, which means that the pBRDF matrix contains many zero elements. In this paper, we extend this work to an out-of-plane scattering geometry, which implies that the pBRDF matrix contains more non-zero elements. In the initial work presented here, a nonlinear optimization approach is utilized to estimate the incident and reflection angles from a single polarization measurement assuming knowledge of the surface index of refraction and azimuthal angle between source and receiver. The effectiveness of the proposed method is verified through computer simulation."
                    },
                    {
                        "title": "Image segmentation from multi-look passive polarimetric imagery",
                        "abstract": "A passive imaging polarimeter records the polarization state of light reflected by an object that is illuminated with an unpolarized and usually uncontrolled source. Passive polarimetric imagery has shown to be useful in many remote sensing applications including shape extraction, material classification and target detection/recognition. In this paper, we present an image segmentation algorithm that automatically extracts an object from multi-look passive polarimetric imagery. The term multi-look refers to multiple polarization measurements where the position of the source of illumination (typically the Sun in passive systems) changes between measurements. The proposed method relies on our previous work on estimating the complex index of refraction and reflection angle from multi-look passive polarimetric imagery. We experimentally showed that the estimates for the index of refraction were largely invariant to both the position of the source and the view angle. Consequently, we utilize the index of refraction as a feature vector to design an illumination invariant image segmentation algorithm. A clustering approach based on the classic c-means algorithm is used for segmenting objects based on their index of refraction. The proposed segmentation approach is validated by using data collected under laboratory conditions. Experimental results indicate that the proposed method is effective for segmenting various targets of interest."
                    },
                    {
                        "title": "Material Classification using Passive Polarimetric Imagery",
                        "abstract": "Passive imaging polarimetry has emerged as an useful tool in many remote sensing applications including material classification, target detection and shape extraction. In this paper we present a method to classify specular objects based on their material composition from passive polarimetric imagery. The proposed algorithm is built on an iterative model-based method to recover the complex index of refraction of a specular target from multiple polarization measurements. The recovered parameters are then used to discriminate between objects by employing the nearest neighbor rule. The effectiveness of the proposed method is validated with data collected in laboratory conditions. Experimental results indicate that the classification approach is highly effective for distinguishing between various targets of interest. Most significantly, the proposed classification method is robust to a wide range of observational geometry."
                    },
                    {
                        "title": "Estimating the Complex Index of Refraction and View Angle of an Object using Multiple Polarization Measurements",
                        "abstract": "A passive polarization based imaging system records the polarization state of light reflected by objects that are illuminated with an unpolarized and generally uncontrolled source. Such systems can be useful in many remote sensing applications including target detection, object segmentation and material classification. In this paper we present a method to jointly estimate the complex index of refraction and the view angle of a target from multiple measurements collected by a passive polarimeter. This generalizes our previous work which was applicable only to dielectric targets. An expression for the degree of polarization is derived from the microfacet polarimetric bidirectional reflectance model for the case of scattering in the place of incidence. Using this expression, we develop nonlinear least squares estimation algorithms for extracting the complex index of refraction and view angle from multiple polarization measurements. The effectiveness of the proposed method is validated with data collected in laboratory conditions. Experimental results indicate that the proposed method is effective for recovering the parameters of interest for real world data and that the complex index of refraction thus computed provides a feature vector that is robust to the view angle."
                    },
                    {
                        "title": "Estimating the refractive index and reflected zenith angle of a target using multiple polarization measurements",
                        "abstract": "A passive polarization based imaging system records the polarization state of light reflected by objects that are illuminated with an unpolarized and generally uncontrolled source. The information conveyed by the polarization state of light has been exploited in applications such as target detection, shape extraction and material classification. In this paper we present a method to jointly estimate the refractive index and the reflected zenith angle from two measurements collected by a passive polarimeter. An expression for the degree of polarization is derived from the microfacet polarimetric bidirectional reflectance model for the case of scattering in the place of incidence. The parameters of interest are iteratively estimated from polarization measurements assumed to be collected with a polarimeter. Computer simulations are presented to demonstrate the effectiveness of the proposed method."
                    }
                ]
            },
            "13605359-802a-4fbb-bfba-3356380ebc1b": {
                "pk": "13605359-802a-4fbb-bfba-3356380ebc1b",
                "name": "Arwen Bradley",
                "collaborators": [
                    "Preetum Nakkiran",
                    "Hattie Zhou",
                    "Madhu Advani",
                    "Etai Littwin",
                    "Noam Razin",
                    "O. Saremi",
                    "Josh Susskind",
                    "Samy Bengio"
                ],
                "domain": [
                    "Diffusion Models",
                    "Machine Learning",
                    "Transformer Models",
                    "Generalization"
                ],
                "publications": [
                    {
                        "title": "Step-by-Step Diffusion: An Elementary Tutorial",
                        "abstract": "We present an accessible first course on diffusion models and flow matching for machine learning, aimed at a technical audience with no diffusion experience. We try to simplify the mathematical details as much as possible (sometimes heuristically), while retaining enough precision to derive correct algorithms."
                    },
                    {
                        "title": "Classifier-Free Guidance is a Predictor-Corrector",
                        "abstract": "We investigate the theoretical foundations of classifier-free guidance (CFG). CFG is the dominant method of conditional sampling for text-to-image diffusion models, yet unlike other aspects of diffusion, it remains on shaky theoretical footing. In this paper, we disprove common misconceptions, by showing that CFG interacts differently with DDPM (Ho et al., 2020) and DDIM (Song et al., 2021), and neither sampler with CFG generates the gamma-powered distribution $p(x|c)^\\gamma p(x)^{1-\\gamma}$. Then, we clarify the behavior of CFG by showing that it is a kind of predictor-corrector method (Song et al., 2020) that alternates between denoising and sharpening, which we call predictor-corrector guidance (PCG). We prove that in the SDE limit, CFG is actually equivalent to combining a DDIM predictor for the conditional distribution together with a Langevin dynamics corrector for a gamma-powered distribution (with a carefully chosen gamma). Our work thus provides a lens to theoretically understand CFG by embedding it in a broader design space of principled sampling methods."
                    },
                    {
                        "title": "What Algorithms can Transformers Learn? A Study in Length Generalization",
                        "abstract": "Large language models exhibit surprising emergent generalization properties, yet also struggle on many simple reasoning tasks such as arithmetic and parity. This raises the question of if and when Transformer models can learn the true algorithm for solving a task. We study the scope of Transformers' abilities in the specific setting of length generalization on algorithmic tasks. Here, we propose a unifying framework to understand when and how Transformers can exhibit strong length generalization on a given task. Specifically, we leverage RASP (Weiss et al., 2021) -- a programming language designed for the computational model of a Transformer -- and introduce the RASP-Generalization Conjecture: Transformers tend to length generalize on a task if the task can be solved by a short RASP program which works for all input lengths. This simple conjecture remarkably captures most known instances of length generalization on algorithmic tasks. Moreover, we leverage our insights to drastically improve generalization performance on traditionally hard tasks (such as parity and addition). On the theoretical side, we give a simple example where the\"min-degree-interpolator\"model of learning from Abbe et al. (2023) does not correctly predict Transformers' out-of-distribution behavior, but our conjecture does. Overall, our work provides a novel perspective on the mechanisms of compositional generalization and the algorithmic capabilities of Transformers."
                    }
                ]
            },
            "646fcb61-24b7-490c-aabf-d67b09d7de99": {
                "pk": "646fcb61-24b7-490c-aabf-d67b09d7de99",
                "name": "Preetum Nakkiran",
                "collaborators": [
                    "Jaros\u0142aw B\u0142asiok",
                    "Etai Littwin",
                    "O. Saremi",
                    "Josh Susskind",
                    "Hattie Zhou",
                    "Arwen Bradley",
                    "Madhu Advani",
                    "Vimal Thilak",
                    "Chen Huang",
                    "Parikshit Gopalan"
                ],
                "domain": [
                    "Self-Supervised Learning",
                    "Calibration",
                    "Transformers",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "How JEPA Avoids Noisy Features: The Implicit Bias of Deep Linear Self Distillation Networks",
                        "abstract": "Two competing paradigms exist for self-supervised learning of data representations. Joint Embedding Predictive Architecture (JEPA) is a class of architectures in which semantically similar inputs are encoded into representations that are predictive of each other. A recent successful approach that falls under the JEPA framework is self-distillation, where an online encoder is trained to predict the output of the target encoder, sometimes using a lightweight predictor network. This is contrasted with the Masked AutoEncoder (MAE) paradigm, where an encoder and decoder are trained to reconstruct missing parts of the input in the data space rather, than its latent representation. A common motivation for using the JEPA approach over MAE is that the JEPA objective prioritizes abstract features over fine-grained pixel information (which can be unpredictable and uninformative). In this work, we seek to understand the mechanism behind this empirical observation by analyzing the training dynamics of deep linear models. We uncover a surprising mechanism: in a simplified linear setting where both approaches learn similar representations, JEPAs are biased to learn high-influence features, i.e., features characterized by having high regression coefficients. Our results point to a distinct implicit bias of predicting in latent space that may shed light on its success in practice."
                    },
                    {
                        "title": "A Formal Framework for Understanding Length Generalization in Transformers",
                        "abstract": "A major challenge for transformers is generalizing to sequences longer than those observed during training. While previous works have empirically shown that transformers can either succeed or fail at length generalization depending on the task, theoretical understanding of this phenomenon remains limited. In this work, we introduce a rigorous theoretical framework to analyze length generalization in causal transformers with learnable absolute positional encodings. In particular, we characterize those functions that are identifiable in the limit from sufficiently long inputs with absolute positional encodings under an idealized inference scheme using a norm-based regularizer. This enables us to prove the possibility of length generalization for a rich family of problems. We experimentally validate the theory as a predictor of success and failure of length generalization across a range of algorithmic and formal language tasks. Our theory not only explains a broad set of empirical observations but also opens the way to provably predicting length generalization capabilities in transformers."
                    },
                    {
                        "title": "When is Multicalibration Post-Processing Necessary?",
                        "abstract": "Calibration is a well-studied property of predictors which guarantees meaningful uncertainty estimates. Multicalibration is a related notion -- originating in algorithmic fairness -- which requires predictors to be simultaneously calibrated over a potentially complex and overlapping collection of protected subpopulations (such as groups defined by ethnicity, race, or income). We conduct the first comprehensive study evaluating the usefulness of multicalibration post-processing across a broad set of tabular, image, and language datasets for models spanning from simple decision trees to 90 million parameter fine-tuned LLMs. Our findings can be summarized as follows: (1) models which are calibrated out of the box tend to be relatively multicalibrated without any additional post-processing; (2) multicalibration post-processing can help inherently uncalibrated models; and (3) traditional calibration measures may sometimes provide multicalibration implicitly. More generally, we also distill many independent observations which may be useful for practical and effective applications of multicalibration post-processing in real-world contexts."
                    },
                    {
                        "title": "Step-by-Step Diffusion: An Elementary Tutorial",
                        "abstract": "We present an accessible first course on diffusion models and flow matching for machine learning, aimed at a technical audience with no diffusion experience. We try to simplify the mathematical details as much as possible (sometimes heuristically), while retaining enough precision to derive correct algorithms."
                    },
                    {
                        "title": "Classifier-Free Guidance is a Predictor-Corrector",
                        "abstract": "We investigate the theoretical foundations of classifier-free guidance (CFG). CFG is the dominant method of conditional sampling for text-to-image diffusion models, yet unlike other aspects of diffusion, it remains on shaky theoretical footing. In this paper, we disprove common misconceptions, by showing that CFG interacts differently with DDPM (Ho et al., 2020) and DDIM (Song et al., 2021), and neither sampler with CFG generates the gamma-powered distribution $p(x|c)^\\gamma p(x)^{1-\\gamma}$. Then, we clarify the behavior of CFG by showing that it is a kind of predictor-corrector method (Song et al., 2020) that alternates between denoising and sharpening, which we call predictor-corrector guidance (PCG). We prove that in the SDE limit, CFG is actually equivalent to combining a DDIM predictor for the conditional distribution together with a Langevin dynamics corrector for a gamma-powered distribution (with a carefully chosen gamma). Our work thus provides a lens to theoretically understand CFG by embedding it in a broader design space of principled sampling methods."
                    },
                    {
                        "title": "Loss minimization yields multicalibration for large neural networks",
                        "abstract": "Multicalibration is a notion of fairness for predictors that requires them to provide calibrated predictions across a large set of protected groups. Multicalibration is known to be a distinct goal than loss minimization, even for simple predictors such as linear functions. In this work, we consider the setting where the protected groups can be represented by neural networks of size $k$, and the predictors are neural networks of size $n>k$. We show that minimizing the squared loss over all neural nets of size $n$ implies multicalibration for all but a bounded number of unlucky values of $n$. We also give evidence that our bound on the number of unlucky values is tight, given our proof technique. Previously, results of the flavor that loss minimization yields multicalibration were known only for predictors that were near the ground truth, hence were rather limited in applicability. Unlike these, our results rely on the expressivity of neural nets and utilize the representation of the predictor."
                    },
                    {
                        "title": "What Algorithms can Transformers Learn? A Study in Length Generalization",
                        "abstract": "Large language models exhibit surprising emergent generalization properties, yet also struggle on many simple reasoning tasks such as arithmetic and parity. This raises the question of if and when Transformer models can learn the true algorithm for solving a task. We study the scope of Transformers' abilities in the specific setting of length generalization on algorithmic tasks. Here, we propose a unifying framework to understand when and how Transformers can exhibit strong length generalization on a given task. Specifically, we leverage RASP (Weiss et al., 2021) -- a programming language designed for the computational model of a Transformer -- and introduce the RASP-Generalization Conjecture: Transformers tend to length generalize on a task if the task can be solved by a short RASP program which works for all input lengths. This simple conjecture remarkably captures most known instances of length generalization on algorithmic tasks. Moreover, we leverage our insights to drastically improve generalization performance on traditionally hard tasks (such as parity and addition). On the theoretical side, we give a simple example where the\"min-degree-interpolator\"model of learning from Abbe et al. (2023) does not correctly predict Transformers' out-of-distribution behavior, but our conjecture does. Overall, our work provides a novel perspective on the mechanisms of compositional generalization and the algorithmic capabilities of Transformers."
                    },
                    {
                        "title": "Empirical Limitations of the NTK for Understanding Scaling Laws in Deep Learning",
                        "abstract": "The \u201cNeural Tangent Kernel\u201d (NTK) (Jacot et al., 2018), and its empirical variants have been proposed as a proxy to capture certain behaviors of real neural networks. In this work, we study NTKs through the lens of scaling laws, and demonstrate that they fall short of explaining important aspects of neural network generalization. In particular, we demonstrate realistic settings where finite-width neural networks have significantly better data scaling exponents as compared to their corresponding empirical and infinite NTKs at initialization. This reveals a more fundamental difference between the real networks and NTKs, beyond just a few percentage points of test accuracy. Further, we show that even if the empirical NTK is allowed to be pre-trained on a constant number of samples, the kernel scaling does not catch up to the neural network scaling. Finally, we show that the empirical NTK continues to evolve throughout most of the training, in contrast with prior work which suggests that it stabilizes after a few epochs of training. Altogether, our work establishes concrete limitations of the NTK approach in understanding scaling laws of real networks on natural datasets."
                    },
                    {
                        "title": "LiDAR: Sensing Linear Probing Performance in Joint Embedding SSL Architectures",
                        "abstract": "Joint embedding (JE) architectures have emerged as a promising avenue for acquiring transferable data representations. A key obstacle to using JE methods, however, is the inherent challenge of evaluating learned representations without access to a downstream task, and an annotated dataset. Without efficient and reliable evaluation, it is difficult to iterate on architectural and training choices for JE methods. In this paper, we introduce LiDAR (Linear Discriminant Analysis Rank), a metric designed to measure the quality of representations within JE architectures. Our metric addresses several shortcomings of recent approaches based on feature covariance rank by discriminating between informative and uninformative features. In essence, LiDAR quantifies the rank of the Linear Discriminant Analysis (LDA) matrix associated with the surrogate SSL task -- a measure that intuitively captures the information content as it pertains to solving the SSL task. We empirically demonstrate that LiDAR significantly surpasses naive rank based approaches in its predictive power of optimal hyperparameters. Our proposed criterion presents a more robust and intuitive means of assessing the quality of representations within JE architectures, which we hope facilitates broader adoption of these powerful techniques in various domains."
                    },
                    {
                        "title": "Smooth ECE: Principled Reliability Diagrams via Kernel Smoothing",
                        "abstract": "Calibration measures and reliability diagrams are two fundamental tools for measuring and interpreting the calibration of probabilistic predictors. Calibration measures quantify the degree of miscalibration, and reliability diagrams visualize the structure of this miscalibration. However, the most common constructions of reliability diagrams and calibration measures -- binning and ECE -- both suffer from well-known flaws (e.g. discontinuity). We show that a simple modification fixes both constructions: first smooth the observations using an RBF kernel, then compute the Expected Calibration Error (ECE) of this smoothed function. We prove that with a careful choice of bandwidth, this method yields a calibration measure that is well-behaved in the sense of (B{\\l}asiok, Gopalan, Hu, and Nakkiran 2023a) -- a consistent calibration measure. We call this measure the SmoothECE. Moreover, the reliability diagram obtained from this smoothed function visually encodes the SmoothECE, just as binned reliability diagrams encode the BinnedECE. We also provide a Python package with simple, hyperparameter-free methods for measuring and plotting calibration: `pip install relplot\\`."
                    },
                    {
                        "title": "Perspectives on the State and Future of Deep Learning - 2023",
                        "abstract": "The goal of this series is to chronicle opinions and issues in the field of machine learning as they stand today and as they change over time. The plan is to host this survey periodically until the AI singularity paperclip-frenzy-driven doomsday, keeping an updated list of topical questions and interviewing new community members for each edition. In this issue, we probed people's opinions on interpretable AI, the value of benchmarking in modern NLP, the state of progress towards understanding deep learning, and the future of academia."
                    },
                    {
                        "title": "When Does Optimizing a Proper Loss Yield Calibration?",
                        "abstract": "Optimizing proper loss functions is popularly believed to yield predictors with good calibration properties; the intuition being that for such losses, the global optimum is to predict the ground-truth probabilities, which is indeed calibrated. However, typical machine learning models are trained to approximately minimize loss over restricted families of predictors, that are unlikely to contain the ground truth. Under what circumstances does optimizing proper loss over a restricted family yield calibrated models? What precise calibration guarantees does it give? In this work, we provide a rigorous answer to these questions. We replace the global optimality with a local optimality condition stipulating that the (proper) loss of the predictor cannot be reduced much by post-processing its predictions with a certain family of Lipschitz functions. We show that any predictor with this local optimality satisfies smooth calibration as defined in Kakade-Foster (2008), B{\\l}asiok et al. (2023). Local optimality is plausibly satisfied by well-trained DNNs, which suggests an explanation for why they are calibrated from proper loss minimization alone. Finally, we show that the connection between local optimality and calibration error goes both ways: nearly calibrated predictors are also nearly locally optimal."
                    },
                    {
                        "title": "Deconstructing Distributions: A Pointwise Framework of Learning",
                        "abstract": "In machine learning, we traditionally evaluate the performance of a single model, averaged over a collection of test inputs. In this work, we propose a new approach: we measure the performance of a collection of models when evaluated on a $\\textit{single input point}$. Specifically, we study a point's $\\textit{profile}$: the relationship between models' average performance on the test distribution and their pointwise performance on this individual point. We find that profiles can yield new insights into the structure of both models and data -- in and out-of-distribution. For example, we empirically show that real data distributions consist of points with qualitatively different profiles. On one hand, there are\"compatible\"points with strong correlation between the pointwise and average performance. On the other hand, there are points with weak and even $\\textit{negative}$ correlation: cases where improving overall model accuracy actually $\\textit{hurts}$ performance on these inputs. We prove that these experimental observations are inconsistent with the predictions of several simplified models of learning proposed in prior work. As an application, we use profiles to construct a dataset we call CIFAR-10-NEG: a subset of CINIC-10 such that for standard models, accuracy on CIFAR-10-NEG is $\\textit{negatively correlated}$ with accuracy on CIFAR-10 test. This illustrates, for the first time, an OOD dataset that completely inverts\"accuracy-on-the-line\"(Miller, Taori, Raghunathan, Sagawa, Koh, Shankar, Liang, Carmon, and Schmidt 2021)"
                    },
                    {
                        "title": "The Calibration Generalization Gap",
                        "abstract": "Calibration is a fundamental property of a good predictive model: it requires that the model predicts correctly in proportion to its confidence. Modern neural networks, however, provide no strong guarantees on their calibration -- and can be either poorly calibrated or well-calibrated depending on the setting. It is currently unclear which factors contribute to good calibration (architecture, data augmentation, overparameterization, etc), though various claims exist in the literature. We propose a systematic way to study the calibration error: by decomposing it into (1) calibration error on the train set, and (2) the calibration generalization gap. This mirrors the fundamental decomposition of generalization. We then investigate each of these terms, and give empirical evidence that (1) DNNs are typically always calibrated on their train set, and (2) the calibration generalization gap is upper-bounded by the standard generalization gap. Taken together, this implies that models with small generalization gap (|Test Error - Train Error|) are well-calibrated. This perspective unifies many results in the literature, and suggests that interventions which reduce the generalization gap (such as adding data, using heavy augmentation, or smaller model size) also improve calibration. We thus hope our initial study lays the groundwork for a more systematic and comprehensive understanding of the relation between calibration, generalization, and optimization."
                    },
                    {
                        "title": "Benign, Tempered, or Catastrophic: A Taxonomy of Overfitting",
                        "abstract": "The practical success of overparameterized neural networks has motivated the recent scientific study of interpolating methods, which perfectly fit their training data. Certain interpolating methods, including neural networks, can fit noisy training data without catastrophically bad test performance, in defiance of standard intuitions from statistical learning theory. Aiming to explain this, a body of recent work has studied benign overfitting, a phenomenon where some interpolating methods approach Bayes optimality, even in the presence of noise. In this work we argue that while benign overfitting has been instructive and fruitful to study, many real interpolating methods like neural networks do not fit benignly: modest noise in the training set causes nonzero (but non-infinite) excess risk at test time, implying these models are neither benign nor catastrophic but rather fall in an intermediate regime. We call this intermediate regime tempered overfitting, and we initiate its systematic study. We first explore this phenomenon in the context of kernel (ridge) regression (KR) by obtaining conditions on the ridge parameter and kernel eigenspectrum under which KR exhibits each of the three behaviors. We find that kernels with powerlaw spectra, including Laplace kernels and ReLU neural tangent kernels, exhibit tempered overfitting. We then empirically study deep neural networks through the lens of our taxonomy, and find that those trained to interpolation are tempered, while those stopped early are benign. We hope our work leads to a more refined understanding of overfitting in modern learning."
                    },
                    {
                        "title": "What You See is What You Get: Principled Deep Learning via Distributional Generalization",
                        "abstract": "Having similar behavior at training time and test time $-$ what we call a\"What You See Is What You Get\"(WYSIWYG) property $-$ is desirable in machine learning. Models trained with standard stochastic gradient descent (SGD), however, do not necessarily have this property, as their complex behaviors such as robustness or subgroup performance can differ drastically between training and test time. In contrast, we show that Differentially-Private (DP) training provably ensures the high-level WYSIWYG property, which we quantify using a notion of distributional generalization. Applying this connection, we introduce new conceptual tools for designing deep-learning methods by reducing generalization concerns to optimization ones: to mitigate unwanted behavior at test time, it is provably sufficient to mitigate this behavior on the training data. By applying this novel design principle, which bypasses\"pathologies\"of SGD, we construct simple algorithms that are competitive with SOTA in several distributional-robustness applications, significantly improve the privacy vs. disparate impact trade-off of DP-SGD, and mitigate robust overfitting in adversarial training. Finally, we also improve on theoretical bounds relating DP, stability, and distributional generalization."
                    },
                    {
                        "title": "Limitations of the NTK for Understanding Generalization in Deep Learning",
                        "abstract": "The ``Neural Tangent Kernel'' (NTK) (Jacot et al 2018), and its empirical variants have been proposed as a proxy to capture certain behaviors of real neural networks. In this work, we study NTKs through the lens of scaling laws, and demonstrate that they fall short of explaining important aspects of neural network generalization. In particular, we demonstrate realistic settings where finite-width neural networks have significantly better data scaling exponents as compared to their corresponding empirical and infinite NTKs at initialization. This reveals a more fundamental difference between the real networks and NTKs, beyond just a few percentage points of test accuracy. Further, we show that even if the empirical NTK is allowed to be pre-trained on a constant number of samples, the kernel scaling does not catch up to the neural network scaling. Finally, we show that the empirical NTK continues to evolve throughout most of the training, in contrast with prior work which suggests that it stabilizes after a few epochs of training. Altogether, our work establishes concrete limitations of the NTK approach in understanding generalization of real networks on natural datasets."
                    },
                    {
                        "title": "Limitations of Neural Collapse for Understanding Generalization in Deep Learning",
                        "abstract": "The recent work of Papyan, Han,&Donoho (2020) presented an intriguing\"Neural Collapse\"phenomenon, showing a structural property of interpolating classifiers in the late stage of training. This opened a rich area of exploration studying this phenomenon. Our motivation is to study the upper limits of this research program: How far will understanding Neural Collapse take us in understanding deep learning? First, we investigate its role in generalization. We refine the Neural Collapse conjecture into two separate conjectures: collapse on the train set (an optimization property) and collapse on the test distribution (a generalization property). We find that while Neural Collapse often occurs on the train set, it does not occur on the test set. We thus conclude that Neural Collapse is primarily an optimization phenomenon, with as-yet-unclear connections to generalization. Second, we investigate the role of Neural Collapse in feature learning. We show simple, realistic experiments where training longer leads to worse last-layer features, as measured by transfer-performance on a downstream task. This suggests that neural collapse is not always desirable for representation learning, as previously claimed. Finally, we give preliminary evidence of a\"cascading collapse\"phenomenon, wherein some form of Neural Collapse occurs not only for the last layer, but in earlier layers as well. We hope our work encourages the community to continue the rich line of Neural Collapse research, while also considering its inherent limitations."
                    }
                ]
            },
            "cca76019-185b-4bad-88e4-910ec1db4953": {
                "pk": "cca76019-185b-4bad-88e4-910ec1db4953",
                "name": "Joshua Susskind",
                "collaborators": [
                    "Etai Littwin",
                    "O. Saremi",
                    "Chen Huang",
                    "Samy Bengio",
                    "Vimal Thilak",
                    "Preetum Nakkiran",
                    "Hanlin Goh",
                    "Samira Abnar",
                    "N. Jaitly",
                    "Laurent Dinh"
                ],
                "domain": [
                    "Self-Supervised Learning",
                    "Vision-Language Models",
                    "Transformer Architecture",
                    "Symbolic Regression"
                ],
                "publications": [
                    {
                        "title": "How JEPA Avoids Noisy Features: The Implicit Bias of Deep Linear Self Distillation Networks",
                        "abstract": "Two competing paradigms exist for self-supervised learning of data representations. Joint Embedding Predictive Architecture (JEPA) is a class of architectures in which semantically similar inputs are encoded into representations that are predictive of each other. A recent successful approach that falls under the JEPA framework is self-distillation, where an online encoder is trained to predict the output of the target encoder, sometimes using a lightweight predictor network. This is contrasted with the Masked AutoEncoder (MAE) paradigm, where an encoder and decoder are trained to reconstruct missing parts of the input in the data space rather, than its latent representation. A common motivation for using the JEPA approach over MAE is that the JEPA objective prioritizes abstract features over fine-grained pixel information (which can be unpredictable and uninformative). In this work, we seek to understand the mechanism behind this empirical observation by analyzing the training dynamics of deep linear models. We uncover a surprising mechanism: in a simplified linear setting where both approaches learn similar representations, JEPAs are biased to learn high-influence features, i.e., features characterized by having high regression coefficients. Our results point to a distinct implicit bias of predicting in latent space that may shed light on its success in practice."
                    },
                    {
                        "title": "Overcoming the Pitfalls of Vision-Language Model Finetuning for OOD Generalization",
                        "abstract": "Existing vision-language models exhibit strong generalization on a variety of visual domains and tasks. However, such models mainly perform zero-shot recognition in a closed-set manner, and thus struggle to handle open-domain visual concepts by design. There are recent finetuning methods, such as prompt learning, that not only study the discrimination between in-distribution (ID) and out-of-distribution (OOD) samples, but also show some improvements in both ID and OOD accuracies. In this paper, we first demonstrate that vision-language models, after long enough finetuning but without proper regularization, tend to overfit the known classes in the given dataset, with degraded performance on unknown classes. Then we propose a novel approach OGEN to address this pitfall, with the main focus on improving the OOD GENeralization of finetuned models. Specifically, a class-conditional feature generator is introduced to synthesize OOD features using just the class name of any unknown class. Such synthesized features will provide useful knowledge about unknowns and help regularize the decision boundary between ID and OOD data when optimized jointly. Equally important is our adaptive self-distillation mechanism to regularize our feature generation model during joint optimization, i.e., adaptively transferring knowledge between model states to further prevent overfitting. Experiments validate that our method yields convincing gains in OOD generalization performance in different settings. Code: https://github.com/apple/ml-ogen."
                    },
                    {
                        "title": "Aggregate-and-Adapt Natural Language Prompts for Downstream Generalization of CLIP",
                        "abstract": "Large pretrained vision-language models like CLIP have shown promising generalization capability, but may struggle in specialized domains (e.g., satellite imagery) or fine-grained classification (e.g., car models) where the visual concepts are unseen or under-represented during pretraining. Prompt learning offers a parameter-efficient finetuning framework that can adapt CLIP to downstream tasks even when limited annotation data are available. In this paper, we improve prompt learning by distilling the textual knowledge from natural language prompts (either human- or LLM-generated) to provide rich priors for those under-represented concepts. We first obtain a prompt ``summary'' aligned to each input image via a learned prompt aggregator. Then we jointly train a prompt generator, optimized to produce a prompt embedding that stays close to the aggregated summary while minimizing task loss at the same time. We dub such prompt embedding as Aggregate-and-Adapted Prompt Embedding (AAPE). AAPE is shown to be able to generalize to different downstream data distributions and tasks, including vision-language understanding tasks (e.g., few-shot classification, VQA) and generation tasks (image captioning) where AAPE achieves competitive performance. We also show AAPE is particularly helpful to handle non-canonical and OOD examples. Furthermore, AAPE learning eliminates LLM-based inference cost as required by baselines, and scales better with data and LLM model size."
                    },
                    {
                        "title": "What Algorithms can Transformers Learn? A Study in Length Generalization",
                        "abstract": "Large language models exhibit surprising emergent generalization properties, yet also struggle on many simple reasoning tasks such as arithmetic and parity. This raises the question of if and when Transformer models can learn the true algorithm for solving a task. We study the scope of Transformers' abilities in the specific setting of length generalization on algorithmic tasks. Here, we propose a unifying framework to understand when and how Transformers can exhibit strong length generalization on a given task. Specifically, we leverage RASP (Weiss et al., 2021) -- a programming language designed for the computational model of a Transformer -- and introduce the RASP-Generalization Conjecture: Transformers tend to length generalize on a task if the task can be solved by a short RASP program which works for all input lengths. This simple conjecture remarkably captures most known instances of length generalization on algorithmic tasks. Moreover, we leverage our insights to drastically improve generalization performance on traditionally hard tasks (such as parity and addition). On the theoretical side, we give a simple example where the\"min-degree-interpolator\"model of learning from Abbe et al. (2023) does not correctly predict Transformers' out-of-distribution behavior, but our conjecture does. Overall, our work provides a novel perspective on the mechanisms of compositional generalization and the algorithmic capabilities of Transformers."
                    },
                    {
                        "title": "Adaptivity and Modularity for Efficient Generalization Over Task Complexity",
                        "abstract": "Can transformers generalize efficiently on problems that require dealing with examples with different levels of difficulty? We introduce a new task tailored to assess generalization over different complexities and present results that indicate that standard transformers face challenges in solving these tasks. These tasks are variations of pointer value retrieval previously introduced by Zhang et al. (2021). We investigate how the use of a mechanism for adaptive and modular computation in transformers facilitates the learning of tasks that demand generalization over the number of sequential computation steps (i.e., the depth of the computation graph). Based on our observations, we propose a transformer-based architecture called Hyper-UT, which combines dynamic function generation from hyper networks with adaptive depth from Universal Transformers. This model demonstrates higher accuracy and a fairer allocation of computational resources when generalizing to higher numbers of computation steps. We conclude that mechanisms for adaptive depth and modularity complement each other in improving efficient generalization concerning example complexity. Additionally, to emphasize the broad applicability of our findings, we illustrate that in a standard image recognition task, Hyper- UT's performance matches that of a ViT model but with considerably reduced computational demands (achieving over 70\\% average savings by effectively using fewer layers)."
                    },
                    {
                        "title": "When can transformers reason with abstract symbols?",
                        "abstract": "We investigate the capabilities of transformer models on relational reasoning tasks. In these tasks, models are trained on a set of strings encoding abstract relations, and are then tested out-of-distribution on data that contains symbols that did not appear in the training dataset. We prove that for any relational reasoning task in a large family of tasks, transformers learn the abstract relations and generalize to the test set when trained by gradient descent on sufficiently large quantities of training data. This is in contrast to classical fully-connected networks, which we prove fail to learn to reason. Our results inspire modifications of the transformer architecture that add only two trainable parameters per head, and that we empirically demonstrate improve data efficiency for learning to reason."
                    },
                    {
                        "title": "LiDAR: Sensing Linear Probing Performance in Joint Embedding SSL Architectures",
                        "abstract": "Joint embedding (JE) architectures have emerged as a promising avenue for acquiring transferable data representations. A key obstacle to using JE methods, however, is the inherent challenge of evaluating learned representations without access to a downstream task, and an annotated dataset. Without efficient and reliable evaluation, it is difficult to iterate on architectural and training choices for JE methods. In this paper, we introduce LiDAR (Linear Discriminant Analysis Rank), a metric designed to measure the quality of representations within JE architectures. Our metric addresses several shortcomings of recent approaches based on feature covariance rank by discriminating between informative and uninformative features. In essence, LiDAR quantifies the rank of the Linear Discriminant Analysis (LDA) matrix associated with the surrogate SSL task -- a measure that intuitively captures the information content as it pertains to solving the SSL task. We empirically demonstrate that LiDAR significantly surpasses naive rank based approaches in its predictive power of optimal hyperparameters. Our proposed criterion presents a more robust and intuitive means of assessing the quality of representations within JE architectures, which we hope facilitates broader adoption of these powerful techniques in various domains."
                    },
                    {
                        "title": "Construction of Paired Knowledge Graph - Text Datasets Informed by Cyclic Evaluation",
                        "abstract": "Datasets that pair Knowledge Graphs (KG) and text together (KG-T) can be used to train forward and reverse neural models that generate text from KG and vice versa. However models trained on datasets where KG and text pairs are not equivalent can suffer from more hallucination and poorer recall. In this paper, we verify this empirically by generating datasets with different levels of noise and find that noisier datasets do indeed lead to more hallucination. We argue that the ability of forward and reverse models trained on a dataset to cyclically regenerate source KG or text is a proxy for the equivalence between the KG and the text in the dataset. Using cyclic evaluation we find that manually created WebNLG is much better than automatically created TeKGen and T-REx. Informed by these observations, we construct a new, improved dataset called LAGRANGE using heuristics meant to improve equivalence between KG and text and show the impact of each of the heuristics on cyclic evaluation. We also construct two synthetic datasets using large language models (LLMs), and observe that these are conducive to models that perform significantly well on cyclic generation of text, but less so on cyclic generation of KGs, probably because of a lack of a consistent underlying ontology."
                    },
                    {
                        "title": "Boolformer: Symbolic Regression of Logic Functions with Transformers",
                        "abstract": "In this work, we introduce Boolformer, the first Transformer architecture trained to perform end-to-end symbolic regression of Boolean functions. First, we show that it can predict compact formulas for complex functions which were not seen during training, when provided a clean truth table. Then, we demonstrate its ability to find approximate expressions when provided incomplete and noisy observations. We evaluate the Boolformer on a broad set of real-world binary classification datasets, demonstrating its potential as an interpretable alternative to classic machine learning methods. Finally, we apply it to the widespread task of modelling the dynamics of gene regulatory networks. Using a recent benchmark, we show that Boolformer is competitive with state-of-the art genetic algorithms with a speedup of several orders of magnitude. Our code and models are available publicly."
                    }
                ]
            },
            "1c17f655-6062-4ad8-bc97-26206fb9ddc1": {
                "pk": "1c17f655-6062-4ad8-bc97-26206fb9ddc1",
                "name": "Etai Littwin",
                "collaborators": [
                    "O. Saremi",
                    "Vimal Thilak",
                    "J. Susskind",
                    "Josh Susskind",
                    "Enric Boix-Adser\u00e0",
                    "Samy Bengio",
                    "Shuangfei Zhai",
                    "Greg Yang",
                    "Preetum Nakkiran",
                    "Chen Huang"
                ],
                "domain": [
                    "Self-Supervised Learning",
                    "Neural Networks",
                    "Transformers",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "UI-JEPA: Towards Active Perception of User Intent through Onscreen User Activity",
                        "abstract": "Generating user intent from a sequence of user interface (UI) actions is a core challenge in comprehensive UI understanding. Recent advancements in multimodal large language models (MLLMs) have led to substantial progress in this area, but their demands for extensive model parameters, computing power, and high latency makes them impractical for scenarios requiring lightweight, on-device solutions with low latency or heightened privacy. Additionally, the lack of high-quality datasets has hindered the development of such lightweight models. To address these challenges, we propose UI-JEPA, a novel framework that employs masking strategies to learn abstract UI embeddings from unlabeled data through self-supervised learning, combined with an LLM decoder fine-tuned for user intent prediction. We also introduce two new UI-grounded multimodal datasets,\"Intent in the Wild\"(IIW) and\"Intent in the Tame\"(IIT), designed for few-shot and zero-shot UI understanding tasks. IIW consists of 1.7K videos across 219 intent categories, while IIT contains 914 videos across 10 categories. We establish the first baselines for these datasets, showing that representations learned using a JEPA-style objective, combined with an LLM decoder, can achieve user intent predictions that match the performance of state-of-the-art large MLLMs, but with significantly reduced annotation and deployment resources. Measured by intent similarity scores, UI-JEPA outperforms GPT-4 Turbo and Claude 3.5 Sonnet by 10.0% and 7.2% respectively, averaged across two datasets. Notably, UI-JEPA accomplishes the performance with a 50.5x reduction in computational cost and a 6.6x improvement in latency in the IIW dataset. These results underscore the effectiveness of UI-JEPA, highlighting its potential for lightweight, high-performance UI understanding."
                    },
                    {
                        "title": "How JEPA Avoids Noisy Features: The Implicit Bias of Deep Linear Self Distillation Networks",
                        "abstract": "Two competing paradigms exist for self-supervised learning of data representations. Joint Embedding Predictive Architecture (JEPA) is a class of architectures in which semantically similar inputs are encoded into representations that are predictive of each other. A recent successful approach that falls under the JEPA framework is self-distillation, where an online encoder is trained to predict the output of the target encoder, sometimes using a lightweight predictor network. This is contrasted with the Masked AutoEncoder (MAE) paradigm, where an encoder and decoder are trained to reconstruct missing parts of the input in the data space rather, than its latent representation. A common motivation for using the JEPA approach over MAE is that the JEPA objective prioritizes abstract features over fine-grained pixel information (which can be unpredictable and uninformative). In this work, we seek to understand the mechanism behind this empirical observation by analyzing the training dynamics of deep linear models. We uncover a surprising mechanism: in a simplified linear setting where both approaches learn similar representations, JEPAs are biased to learn high-influence features, i.e., features characterized by having high regression coefficients. Our results point to a distinct implicit bias of predicting in latent space that may shed light on its success in practice."
                    },
                    {
                        "title": "Enhancing JEPAs with Spatial Conditioning: Robust and Efficient Representation Learning",
                        "abstract": "Image-based Joint-Embedding Predictive Architecture (IJEPA) offers an attractive alternative to Masked Autoencoder (MAE) for representation learning using the Masked Image Modeling framework. IJEPA drives representations to capture useful semantic information by predicting in latent rather than input space. However, IJEPA relies on carefully designed context and target windows to avoid representational collapse. The encoder modules in IJEPA cannot adaptively modulate the type of predicted and/or target features based on the feasibility of the masked prediction task as they are not given sufficient information of both context and targets. Based on the intuition that in natural images, information has a strong spatial bias with spatially local regions being highly predictive of one another compared to distant ones. We condition the target encoder and context encoder modules in IJEPA with positions of context and target windows respectively. Our\"conditional\"encoders show performance gains on several image classification benchmark datasets, improved robustness to context window size and sample-efficiency during pretraining."
                    },
                    {
                        "title": "The NTK approximation is valid for longer than you think",
                        "abstract": "We study when the neural tangent kernel (NTK) approximation is valid for training a model with the square loss. In the lazy training setting of [COB19], we show that rescaling the model by a factor of \u03b1 = O ( T ) su\ufb03ces for the NTK approximation to be valid until training time T . Our bound is tight and improves on the previous bound of [COB19], which required a larger rescaling factor of \u03b1 = O ( T 2 )"
                    },
                    {
                        "title": "Transformers learn through gradual rank increase",
                        "abstract": "We identify incremental learning dynamics in transformers, where the difference between trained and initial weights progressively increases in rank. We rigorously prove this occurs under the simplifying assumptions of diagonal weight matrices and small initialization. Our experiments support the theory and also show that phenomenon can occur in practice without the simplifying assumptions."
                    },
                    {
                        "title": "What Algorithms can Transformers Learn? A Study in Length Generalization",
                        "abstract": "Large language models exhibit surprising emergent generalization properties, yet also struggle on many simple reasoning tasks such as arithmetic and parity. This raises the question of if and when Transformer models can learn the true algorithm for solving a task. We study the scope of Transformers' abilities in the specific setting of length generalization on algorithmic tasks. Here, we propose a unifying framework to understand when and how Transformers can exhibit strong length generalization on a given task. Specifically, we leverage RASP (Weiss et al., 2021) -- a programming language designed for the computational model of a Transformer -- and introduce the RASP-Generalization Conjecture: Transformers tend to length generalize on a task if the task can be solved by a short RASP program which works for all input lengths. This simple conjecture remarkably captures most known instances of length generalization on algorithmic tasks. Moreover, we leverage our insights to drastically improve generalization performance on traditionally hard tasks (such as parity and addition). On the theoretical side, we give a simple example where the\"min-degree-interpolator\"model of learning from Abbe et al. (2023) does not correctly predict Transformers' out-of-distribution behavior, but our conjecture does. Overall, our work provides a novel perspective on the mechanisms of compositional generalization and the algorithmic capabilities of Transformers."
                    },
                    {
                        "title": "Adaptivity and Modularity for Efficient Generalization Over Task Complexity",
                        "abstract": "Can transformers generalize efficiently on problems that require dealing with examples with different levels of difficulty? We introduce a new task tailored to assess generalization over different complexities and present results that indicate that standard transformers face challenges in solving these tasks. These tasks are variations of pointer value retrieval previously introduced by Zhang et al. (2021). We investigate how the use of a mechanism for adaptive and modular computation in transformers facilitates the learning of tasks that demand generalization over the number of sequential computation steps (i.e., the depth of the computation graph). Based on our observations, we propose a transformer-based architecture called Hyper-UT, which combines dynamic function generation from hyper networks with adaptive depth from Universal Transformers. This model demonstrates higher accuracy and a fairer allocation of computational resources when generalizing to higher numbers of computation steps. We conclude that mechanisms for adaptive depth and modularity complement each other in improving efficient generalization concerning example complexity. Additionally, to emphasize the broad applicability of our findings, we illustrate that in a standard image recognition task, Hyper- UT's performance matches that of a ViT model but with considerably reduced computational demands (achieving over 70\\% average savings by effectively using fewer layers)."
                    },
                    {
                        "title": "Stabilizing Transformer Training by Preventing Attention Entropy Collapse",
                        "abstract": "Training stability is of great importance to Transformers. In this work, we investigate the training dynamics of Transformers by examining the evolution of the attention layers. In particular, we track the attention entropy for each attention head during the course of training, which is a proxy for model sharpness. We identify a common pattern across different architectures and tasks, where low attention entropy is accompanied by high training instability, which can take the form of oscillating loss or divergence. We denote the pathologically low attention entropy, corresponding to highly concentrated attention scores, as $\\textit{entropy collapse}$. As a remedy, we propose $\\sigma$Reparam, a simple and efficient solution where we reparametrize all linear layers with spectral normalization and an additional learned scalar. We demonstrate that $\\sigma$Reparam successfully prevents entropy collapse in the attention layers, promoting more stable training. Additionally, we prove a tight lower bound of the attention entropy, which decreases exponentially fast with the spectral norm of the attention logits, providing additional motivation for our approach. We conduct experiments with $\\sigma$Reparam on image classification, image self-supervised learning, machine translation, speech recognition, and language modeling tasks. We show that $\\sigma$Reparam provides stability and robustness with respect to the choice of hyperparameters, going so far as enabling training (a) a Vision Transformer {to competitive performance} without warmup, weight decay, layer normalization or adaptive optimizers; (b) deep architectures in machine translation and (c) speech recognition to competitive performance without warmup and adaptive optimizers. Code is available at \\url{https://github.com/apple/ml-sigma-reparam}."
                    },
                    {
                        "title": "Tight conditions for when the NTK approximation is valid",
                        "abstract": "We study when the neural tangent kernel (NTK) approximation is valid for training a model with the square loss. In the lazy training setting of Chizat et al. 2019, we show that rescaling the model by a factor of $\\alpha = O(T)$ suffices for the NTK approximation to be valid until training time $T$. Our bound is tight and improves on the previous bound of Chizat et al. 2019, which required a larger rescaling factor of $\\alpha = O(T^2)$."
                    },
                    {
                        "title": "Tensor Programs IVb: Adaptive Optimization in the Infinite-Width Limit",
                        "abstract": "Going beyond stochastic gradient descent (SGD), what new phenomena emerge in wide neural networks trained by adaptive optimizers like Adam? Here we show: The same dichotomy between feature learning and kernel behaviors (as in SGD) holds for general optimizers as well, including Adam -- albeit with a nonlinear notion of\"kernel.\"We derive the corresponding\"neural tangent\"and\"maximal update\"limits for any architecture. Two foundational advances underlie the above results: 1) A new Tensor Program language, NEXORT, that can express how adaptive optimizers process gradients into updates. 2) The introduction of bra-ket notation to drastically simplify expressions and calculations in Tensor Programs. This work summarizes and generalizes all previous results in the Tensor Programs series of papers."
                    },
                    {
                        "title": "When can transformers reason with abstract symbols?",
                        "abstract": "We investigate the capabilities of transformer models on relational reasoning tasks. In these tasks, models are trained on a set of strings encoding abstract relations, and are then tested out-of-distribution on data that contains symbols that did not appear in the training dataset. We prove that for any relational reasoning task in a large family of tasks, transformers learn the abstract relations and generalize to the test set when trained by gradient descent on sufficiently large quantities of training data. This is in contrast to classical fully-connected networks, which we prove fail to learn to reason. Our results inspire modifications of the transformer architecture that add only two trainable parameters per head, and that we empirically demonstrate improve data efficiency for learning to reason."
                    },
                    {
                        "title": "LiDAR: Sensing Linear Probing Performance in Joint Embedding SSL Architectures",
                        "abstract": "Joint embedding (JE) architectures have emerged as a promising avenue for acquiring transferable data representations. A key obstacle to using JE methods, however, is the inherent challenge of evaluating learned representations without access to a downstream task, and an annotated dataset. Without efficient and reliable evaluation, it is difficult to iterate on architectural and training choices for JE methods. In this paper, we introduce LiDAR (Linear Discriminant Analysis Rank), a metric designed to measure the quality of representations within JE architectures. Our metric addresses several shortcomings of recent approaches based on feature covariance rank by discriminating between informative and uninformative features. In essence, LiDAR quantifies the rank of the Linear Discriminant Analysis (LDA) matrix associated with the surrogate SSL task -- a measure that intuitively captures the information content as it pertains to solving the SSL task. We empirically demonstrate that LiDAR significantly surpasses naive rank based approaches in its predictive power of optimal hyperparameters. Our proposed criterion presents a more robust and intuitive means of assessing the quality of representations within JE architectures, which we hope facilitates broader adoption of these powerful techniques in various domains."
                    },
                    {
                        "title": "Adaptive Optimization in the \u221e-Width Limit",
                        "abstract": "of any computation graph involving adaptive optimizers and the calculation of their large width limits. Our work lays a path to study the implicit bias of adaptive optimizers by studying their evolution equations in the in\ufb01nite-width limit"
                    },
                    {
                        "title": "Learning Representation from Neural Fisher Kernel with Low-rank Approximation",
                        "abstract": "In this paper, we study the representation of neural networks from the view of kernels. We first define the Neural Fisher Kernel (NFK), which is the Fisher Kernel applied to neural networks. We show that NFK can be computed for both supervised and unsupervised learning models, which can serve as a unified tool for representation extraction. Furthermore, we show that practical NFKs exhibit low-rank structures. We then propose an efficient algorithm that computes a low rank approximation of NFK, which scales to large datasets and networks. We show that the low-rank approximation of NFKs derived from unsupervised generative models and supervised learning models gives rise to high-quality compact representations of data, achieving competitive results on a variety of machine learning tasks."
                    },
                    {
                        "title": "The Slingshot Mechanism: An Empirical Study of Adaptive Optimizers and the Grokking Phenomenon",
                        "abstract": "The grokking phenomenon as reported by Power et al. ( arXiv:2201.02177 ) refers to a regime where a long period of overfitting is followed by a seemingly sudden transition to perfect generalization. In this paper, we attempt to reveal the underpinnings of Grokking via a series of empirical studies. Specifically, we uncover an optimization anomaly plaguing adaptive optimizers at extremely late stages of training, referred to as the Slingshot Mechanism. A prominent artifact of the Slingshot Mechanism can be measured by the cyclic phase transitions between stable and unstable training regimes, and can be easily monitored by the cyclic behavior of the norm of the last layers weights. We empirically observe that without explicit regularization, Grokking as reported in ( arXiv:2201.02177 ) almost exclusively happens at the onset of Slingshots, and is absent without it. While common and easily reproduced in more general settings, the Slingshot Mechanism does not follow from any known optimization theories that we are aware of, and can be easily overlooked without an in depth examination. Our work points to a surprising and useful inductive bias of adaptive gradient optimizers at late stages of training, calling for a revised theoretical analysis of their origin."
                    },
                    {
                        "title": "Tensor Programs IIb: Architectural Universality of Neural Tangent Kernel Training Dynamics",
                        "abstract": "Yang (2020a) recently showed that the Neural Tangent Kernel (NTK) at initialization has an infinite-width limit for a large class of architectures including modern staples such as ResNet and Transformers. However, their analysis does not apply to training. Here, we show the same neural networks (in the so-called NTK parametrization) during training follow a kernel gradient descent dynamics in function space, where the kernel is the infinite-width NTK. This completes the proof of the *architectural universality* of NTK behavior. To achieve this result, we apply the Tensor Programs technique: Write the entire SGD dynamics inside a Tensor Program and analyze it via the Master Theorem. To facilitate this proof, we develop a graphical notation for Tensor Programs."
                    },
                    {
                        "title": "Implicit Acceleration and Feature Learning in Infinitely Wide Neural Networks with Bottlenecks",
                        "abstract": "We analyze the learning dynamics of infinitely wide neural networks with a finite sized bottle-neck. Unlike the neural tangent kernel limit, a bottleneck in an otherwise infinite width network al-lows data dependent feature learning in its bottle-neck representation. We empirically show that a single bottleneck in infinite networks dramatically accelerates training when compared to purely in-finite networks, with an improved overall performance. We discuss the acceleration phenomena by drawing similarities to infinitely wide deep linear models, where the acceleration effect of a bottleneck can be understood theoretically."
                    },
                    {
                        "title": "Implicit Greedy Rank Learning in Autoencoders via Overparameterized Linear Networks",
                        "abstract": "Deep linear networks trained with gradient descent yield low rank solutions, as is typically studied in matrix factorization. In this paper, we take a step further and analyze implicit rank regularization in autoencoders. We show greedy learning of low-rank latent codes induced by a linear sub-network at the autoencoder bottleneck. We further propose orthogonal initialization and principled learning rate adjustment to mitigate sensitivity of training dynamics to spectral prior and linear depth. With linear autoencoders on synthetic data, our method converges stably to ground-truth latent code rank. With nonlinear autoencoders, our method converges to latent ranks optimal for downstream classification and image sampling."
                    }
                ]
            }
        }
    },
    "2405.17277": {
        "paper_data": {
            "title": "Gradients of Functions of Large Matrices",
            "url": "http://arxiv.org/abs/2405.17277v1",
            "arxiv_id": "2405.17277",
            "authors": [
                "Nicholas Kr\u00e4mer",
                "Pablo Moreno-Mu\u00f1oz",
                "Hrittik Roy",
                "S\u00f8ren Hauberg"
            ],
            "abstract": "Tuning scientific and probabilistic machine learning models -- for example, partial differential equations, Gaussian processes, or Bayesian neural networks -- often relies on evaluating functions of matrices whose size grows with the data set or the number of parameters. While the state-of-the-art for evaluating these quantities is almost always based on Lanczos and Arnoldi iterations, the present work is the first to explain how to differentiate these workhorses of numerical linear algebra efficiently. To get there, we derive previously unknown adjoint systems for Lanczos and Arnoldi iterations, implement them in JAX, and show that the resulting code can compete with Diffrax when it comes to differentiating PDEs, GPyTorch for selecting Gaussian process models and beats standard factorisation methods for calibrating Bayesian neural networks. All this is achieved without any problem-specific code optimisation. Find the code at https://github.com/pnkraemer/experiments-lanczos-adjoints and install the library with pip install matfree.",
            "introduction": "   1 Introduction  Automatic differentiation has dramatically altered the development of machine learning models by allowing us to forego laborious, application-dependent gradient derivations. The essence of this automation is to evaluate Jacobian-vector and vector-Jacobian products without ever instantiating the full Jacobian matrix, whose column count would match the number of parameters of the neural network. Nowadays, everyone can build algorithms around matrices of unprecedented sizes by exploiting this matrix-free implementation. However, differentiable linear algebra for Jacobian-vector products and similar operations has remained largely unexplored to this day. We introduce a new matrix-free method for automatically differentiating functions of matrices. Our algorithm yields the exact gradients of the forward pass, all gradients are obtained with the same code, and said code runs in linear time- and memory-complexity.   For a parametrised matrix A=A\u2062(\u03b8)\u2208\u211dN\u00d7N\ud835\udc34\ud835\udc34\ud835\udf03superscript\u211d\ud835\udc41\ud835\udc41A=A(\\theta)\\in\\mathbb{R}^{N\\times N}italic_A = italic_A ( italic_\u03b8 ) \u2208 blackboard_R start_POSTSUPERSCRIPT italic_N \u00d7 italic_N end_POSTSUPERSCRIPT and an analytic function f:\u211d\u2192\u211d:\ud835\udc53\u2192\u211d\u211df:\\mathbb{R}\\rightarrow\\mathbb{R}italic_f : blackboard_R \u2192 blackboard_R, we call f\u2062(A)\ud835\udc53\ud835\udc34f(A)italic_f ( italic_A ) a function of the matrix (different properties of A\ud835\udc34Aitalic_A imply different definitions of f\u2062(A)\ud835\udc53\ud835\udc34f(A)italic_f ( italic_A ); one of them is applying f\ud835\udc53fitalic_f to each eigenvalue of A\ud835\udc34Aitalic_A if A\ud835\udc34Aitalic_A is diagonalisable; see [1]). However, we assume that A\ud835\udc34Aitalic_A is the Jacobian of a large neural network or a matrix of similar size and never materialise f\u2062(A)\ud835\udc53\ud835\udc34f(A)italic_f ( italic_A ). Instead, we only care about the values and gradients of the matrix-function-vector product    (\u03b8,v)\u21a6f\u2062[A\u2062(\u03b8)]\u2062vmaps-to\ud835\udf03\ud835\udc63\ud835\udc53delimited-[]\ud835\udc34\ud835\udf03\ud835\udc63\\displaystyle(\\theta,v)\\mapsto f[A(\\theta)]v( italic_\u03b8 , italic_v ) \u21a6 italic_f [ italic_A ( italic_\u03b8 ) ] italic_v  (1)   assuming that A\ud835\udc34Aitalic_A is only accessed via differentiable matrix-vector products. Table\u00a01 lists examples.   Table 1: Some applications for functions of matrices. Log-determinants apply by combining log\u2062det(A)=trace\u2062(log\u2061(A))\ud835\udc34trace\ud835\udc34\\log\\det(A)=\\,\\mathrm{trace}\\left(\\log(A)\\right)roman_log roman_det ( italic_A ) = roman_trace ( roman_log ( italic_A ) ) with stochastic trace estimation, which is why most vectors in this table are Rademacher samples. \u201cPDE\u201d / \u201cODE\u201d = \u201cPartial/Ordinary differential equation\u201d.    Application  Function f\ud835\udc53fitalic_f   Matrix A\ud835\udc34Aitalic_A   Vector v\ud835\udc63vitalic_v   Parameter \u03b8\ud835\udf03\\thetaitalic_\u03b8      PDEs & flows [2, 3, 4, 5]  e\u03bbsuperscript\ud835\udc52\ud835\udf06e^{\\lambda}italic_e start_POSTSUPERSCRIPT italic_\u03bb end_POSTSUPERSCRIPT PDE discret. PDE initial value PDE     Gaussian process [6, 7, 8]  log\u2061(\u03bb)\ud835\udf06\\log(\\lambda)roman_log ( italic_\u03bb ) Kernel matrix v\u223cRademachersimilar-to\ud835\udc63Rademacherv\\sim\\text{Rademacher}italic_v \u223c Rademacher Kernel     Invert. ResNets [9, 10]  log\u2061(1+\u03bb)1\ud835\udf06\\log(1+\\lambda)roman_log ( 1 + italic_\u03bb ) Jacobian matrix v\u223cRademachersimilar-to\ud835\udc63Rademacherv\\sim\\text{Rademacher}italic_v \u223c Rademacher Network     Gaussian sampler [11]  \u03bb\ud835\udf06\\sqrt{\\lambda}square-root start_ARG italic_\u03bb end_ARG Covariance matrix v\u223cN\u2062(0,I)similar-to\ud835\udc63\ud835\udc410\ud835\udc3cv\\sim N(0,I)italic_v \u223c italic_N ( 0 , italic_I ) Covariance     Neural ODE [12]  \u03bb2superscript\ud835\udf062\\lambda^{2}italic_\u03bb start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT Jacobian matrix v\u223cRademachersimilar-to\ud835\udc63Rademacherv\\sim\\text{Rademacher}italic_v \u223c Rademacher Network       Evaluating Equation\u00a01 is crucial for building large machine learning models, e.g., Bayesian neural nets: A common hyperparameter-calibration loss of a (Laplace-approximated) Bayesian neural network involves the log-determinant of the generalised Gauss\u2013Newton matrix [13]    A\u2062(\u03b1)\u2254\u2211(xi,yi)\u2208data[D\u03b8\u2062g]\u2062(xi)\u22a4\u2062[Dg2\u2062\u03c1]\u2062(yi,g\u2062(xi))\u2062[D\u03b8\u2062g]\u2062(xi)+\u03b12\u2062I,\u2254\ud835\udc34\ud835\udefcsubscriptsubscript\ud835\udc65\ud835\udc56subscript\ud835\udc66\ud835\udc56datadelimited-[]subscript\ud835\udc37\ud835\udf03\ud835\udc54superscriptsubscript\ud835\udc65\ud835\udc56topdelimited-[]superscriptsubscript\ud835\udc37\ud835\udc542\ud835\udf0csubscript\ud835\udc66\ud835\udc56\ud835\udc54subscript\ud835\udc65\ud835\udc56delimited-[]subscript\ud835\udc37\ud835\udf03\ud835\udc54subscript\ud835\udc65\ud835\udc56superscript\ud835\udefc2\ud835\udc3c\\displaystyle A(\\alpha)\\coloneqq\\sum_{(x_{i},y_{i})\\in\\text{data}}[D_{\\theta}g% ](x_{i})^{\\top}[D_{g}^{2}\\rho](y_{i},g(x_{i}))[D_{\\theta}g](x_{i})+\\alpha^{2}I,italic_A ( italic_\u03b1 ) \u2254 \u2211 start_POSTSUBSCRIPT ( italic_x start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT , italic_y start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) \u2208 data end_POSTSUBSCRIPT [ italic_D start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT italic_g ] ( italic_x start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) start_POSTSUPERSCRIPT \u22a4 end_POSTSUPERSCRIPT [ italic_D start_POSTSUBSCRIPT italic_g end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_\u03c1 ] ( italic_y start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT , italic_g ( italic_x start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) ) [ italic_D start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT italic_g ] ( italic_x start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) + italic_\u03b1 start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_I ,  (2)   where D\u03b8\u2062gsubscript\ud835\udc37\ud835\udf03\ud835\udc54D_{\\theta}gitalic_D start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT italic_g is the parameter-Jacobian of the neural network g\ud835\udc54gitalic_g, Dg2\u2062\u03c1superscriptsubscript\ud835\udc37\ud835\udc542\ud835\udf0cD_{g}^{2}\\rhoitalic_D start_POSTSUBSCRIPT italic_g end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_\u03c1 is the Hessian of the loss function \u03c1\ud835\udf0c\\rhoitalic_\u03c1 with respect to g\u2062(xi)\ud835\udc54subscript\ud835\udc65\ud835\udc56g(x_{i})italic_g ( italic_x start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ), and \u03b1\ud835\udefc\\alphaitalic_\u03b1 is a to-be-tuned parameter. The matrix A\u2062(\u03b1)\ud835\udc34\ud835\udefcA(\\alpha)italic_A",
            "references": [
                {
                    "title": "The Elements of Differentiable Programming",
                    "abstract": "Artificial intelligence has recently experienced remarkable advances, fueled by large models, vast datasets, accelerated hardware, and, last but not least, the transformative power of differentiable programming. This new programming paradigm enables end-to-end differentiation of complex computer programs (including those with control flows and data structures), making gradient-based optimization of program parameters possible. As an emerging paradigm, differentiable programming builds upon several areas of computer science and applied mathematics, including automatic differentiation, graphical models, optimization and statistics. This book presents a comprehensive review of the fundamental concepts useful for differentiable programming. We adopt two main perspectives, that of optimization and that of probability, with clear analogies between the two. Differentiable programming is not merely the differentiation of programs, but also the thoughtful design of programs intended for differentiation. By making programs differentiable, we inherently introduce probability distributions over their execution, providing a means to quantify the uncertainty associated with program outputs."
                },
                {
                    "title": "Efficient training of Gaussian processes with tensor product structure",
                    "abstract": "To determine the optimal set of hyperparameters of a Gaussian process based on a large number of training data, both a linear system and a trace estimation problem must be solved. In this paper, we focus on establishing numerical methods for the case where the covariance matrix is given as the sum of possibly multiple Kronecker products, i.e., can be identified as a tensor. As such, we will represent this operator and the training data in the tensor train format. Based on the AMEn method and Krylov subspace methods, we derive an efficient scheme for computing the matrix functions required for evaluating the gradient and the objective function in hyperparameter optimization."
                },
                {
                    "title": "Large-Scale Gaussian Processes via Alternating Projection",
                    "abstract": "Training and inference in Gaussian processes (GPs) require solving linear systems with $n\\times n$ kernel matrices. To address the prohibitive $\\mathcal{O}(n^3)$ time complexity, recent work has employed fast iterative methods, like conjugate gradients (CG). However, as datasets increase in magnitude, the kernel matrices become increasingly ill-conditioned and still require $\\mathcal{O}(n^2)$ space without partitioning. Thus, while CG increases the size of datasets GPs can be trained on, modern datasets reach scales beyond its applicability. In this work, we propose an iterative method which only accesses subblocks of the kernel matrix, effectively enabling mini-batching. Our algorithm, based on alternating projection, has $\\mathcal{O}(n)$ per-iteration time and space complexity, solving many of the practical challenges of scaling GPs to very large datasets. Theoretically, we prove the method enjoys linear convergence. Empirically, we demonstrate its fast convergence in practice and robustness to ill-conditioning. On large-scale benchmark datasets with up to four million data points, our approach accelerates GP training and inference by speed-up factors up to $27\\times$ and $72 \\times$, respectively, compared to CG."
                },
                {
                    "title": "Stochastic Marginal Likelihood Gradients using Neural Tangent Kernels",
                    "abstract": "Selecting hyperparameters in deep learning greatly impacts its effectiveness but requires manual effort and expertise. Recent works show that Bayesian model selection with Laplace approximations can allow to optimize such hyperparameters just like standard neural network parameters using gradients and on the training data. However, estimating a single hyperparameter gradient requires a pass through the entire dataset, limiting the scalability of such algorithms. In this work, we overcome this issue by introducing lower bounds to the linearized Laplace approximation of the marginal likelihood. In contrast to previous estimators, these bounds are amenable to stochastic-gradient-based optimization and allow to trade off estimation accuracy against computational complexity. We derive them using the function-space form of the linearized Laplace, which can be estimated using the neural tangent kernel. Experimentally, we show that the estimators can significantly accelerate gradient-based hyperparameter optimization."
                },
                {
                    "title": "Accelerated Linearized Laplace Approximation for Bayesian Deep Learning",
                    "abstract": "Laplace approximation (LA) and its linearized variant (LLA) enable effortless adaptation of pretrained deep neural networks to Bayesian neural networks. The generalized Gauss-Newton (GGN) approximation is typically introduced to improve their tractability. However, LA and LLA are still confronted with non-trivial inefficiency issues and should rely on Kronecker-factored, diagonal, or even last-layer approximate GGN matrices in practical use. These approximations are likely to harm the fidelity of learning outcomes. To tackle this issue, inspired by the connections between LLA and neural tangent kernels (NTKs), we develop a Nystrom approximation to NTKs to accelerate LLA. Our method benefits from the capability of popular deep learning libraries for forward mode automatic differentiation, and enjoys reassuring theoretical guarantees. Extensive studies reflect the merits of the proposed method in aspects of both scalability and performance. Our method can even scale up to architectures like vision transformers. We also offer valuable ablation studies to diagnose our method. Code is available at \\url{https://github.com/thudzj/ELLA}."
                },
                {
                    "title": "Flow Matching for Generative Modeling",
                    "abstract": "We introduce a new paradigm for generative modeling built on Continuous Normalizing Flows (CNFs), allowing us to train CNFs at unprecedented scale. Specifically, we present the notion of Flow Matching (FM), a simulation-free approach for training CNFs based on regressing vector fields of fixed conditional probability paths. Flow Matching is compatible with a general family of Gaussian probability paths for transforming between noise and data samples -- which subsumes existing diffusion paths as specific instances. Interestingly, we find that employing FM with diffusion paths results in a more robust and stable alternative for training diffusion models. Furthermore, Flow Matching opens the door to training CNFs with other, non-diffusion probability paths. An instance of particular interest is using Optimal Transport (OT) displacement interpolation to define the conditional probability paths. These paths are more efficient than diffusion paths, provide faster training and sampling, and result in better generalization. Training CNFs using Flow Matching on ImageNet leads to consistently better performance than alternative diffusion-based methods in terms of both likelihood and sample quality, and allows fast and reliable sample generation using off-the-shelf numerical ODE solvers."
                },
                {
                    "title": "j-Wave: An open-source differentiable wave simulator",
                    "abstract": "We present an open-source differentiable acoustic simulator, j-Wave, which can solve time-varying and time-harmonic acoustic problems. It supports automatic differentiation, which is a program transformation technique that has many applications, especially in machine learning and scientific computing. j-Wave is composed of modular components that can be easily customized and reused. At the same time, it is compatible with some of the most popular machine learning libraries, such as JAX and TensorFlow. The accuracy of the simulation results for known configurations is evaluated against the widely used k-Wave toolbox and a cohort of acoustic simulation software. j-Wave is available from https://github.com/ucl-bug/jwave."
                },
                {
                    "title": "Laplacian Autoencoders for Learning Stochastic Representations",
                    "abstract": "Established methods for unsupervised representation learning such as variational autoencoders produce none or poorly calibrated uncertainty estimates making it difficult to evaluate if learned representations are stable and reliable. In this work, we present a Bayesian autoencoder for unsupervised representation learning, which is trained using a novel variational lower-bound of the autoencoder evidence. This is maximized using Monte Carlo EM with a variational distribution that takes the shape of a Laplace approximation. We develop a new Hessian approximation that scales linearly with data size allowing us to model high-dimensional data. Empirically, we show that our Laplacian autoencoder estimates well-calibrated uncertainties in both latent and output space. We demonstrate that this results in improved performance across a multitude of downstream tasks."
                },
                {
                    "title": "Generative Modelling With Inverse Heat Dissipation",
                    "abstract": "While diffusion models have shown great success in image generation, their noise-inverting generative process does not explicitly consider the structure of images, such as their inherent multi-scale nature. Inspired by diffusion models and the empirical success of coarse-to-fine modelling, we propose a new diffusion-like model that generates images through stochastically reversing the heat equation, a PDE that locally erases fine-scale information when run over the 2D plane of the image. We interpret the solution of the forward heat equation with constant additive noise as a variational approximation in the diffusion latent variable model. Our new model shows emergent qualitative properties not seen in standard diffusion models, such as disentanglement of overall colour and shape in images. Spectral analysis on natural images highlights connections to diffusion models and reveals an implicit coarse-to-fine inductive bias in them."
                },
                {
                    "title": "Geometric Methods for Adjoint Systems",
                    "abstract": "Adjoint systems are widely used to inform control, optimization, and design in systems described by ordinary differential equations or differential-algebraic equations. In this paper, we explore the geometric properties and develop methods for such adjoint systems. In particular, we utilize symplectic and presymplectic geometry to investigate the properties of adjoint systems associated with ordinary differential equations and differential-algebraic equations, respectively. We show that the adjoint variational quadratic conservation laws, which are key to adjoint sensitivity analysis, arise from (pre)symplecticity of such adjoint systems. We discuss various additional geometric properties of adjoint systems, such as symmetries and variational characterizations. For adjoint systems associated with a differential-algebraic equation, we relate the index of the differential-algebraic equation to the presymplectic constraint algorithm of Gotay et al. (J Math Phys 19(11):2388\u20132399, 1978). As an application of this geometric framework, we discuss how the adjoint variational quadratic conservation laws can be used to compute sensitivities of terminal or running cost functions. Furthermore, we develop structure-preserving numerical methods for such systems using Galerkin Hamiltonian variational integrators (Leok and Zhang in IMA J. Numer. Anal. 31(4):1497\u20131532, 2011) which admit discrete analogues of these quadratic conservation laws. We additionally show that such methods are natural, in the sense that reduction, forming the adjoint system, and discretization all commute, for suitable choices of these processes. We utilize this naturality to derive a variational error analysis result for the presymplectic variational integrator that we use to discretize the adjoint DAE system. Finally, we discuss the application of adjoint systems in the context of optimal control problems, where we prove a similar naturality result."
                },
                {
                    "title": "Krylov-aware stochastic trace estimation",
                    "abstract": "We introduce an algorithm for estimating the trace of a matrix function $f(\\mathbf{A})$ using implicit products with a symmetric matrix $\\mathbf{A}$. Existing methods for implicit trace estimation of a matrix function tend to treat matrix-vector products with $f(\\mathbf{A})$ as a black-box to be computed by a Krylov subspace method. Like other recent algorithms for implicit trace estimation, our approach is based on a combination of deflation and stochastic trace estimation. However, we take a closer look at how products with $f(\\mathbf{A})$ are integrated into these approaches which enables several efficiencies not present in previously studied methods. In particular, we describe a Krylov subspace method for computing a low-rank approximation of a matrix function by a computationally efficient projection onto Krylov subspace."
                },
                {
                    "title": "Fast Sampling of Diffusion Models with Exponential Integrator",
                    "abstract": "The past few years have witnessed the great success of Diffusion models~(DMs) in generating high-fidelity samples in generative modeling tasks. A major limitation of the DM is its notoriously slow sampling procedure which normally requires hundreds to thousands of time discretization steps of the learned diffusion process to reach the desired accuracy. Our goal is to develop a fast sampling method for DMs with a much less number of steps while retaining high sample quality. To this end, we systematically analyze the sampling procedure in DMs and identify key factors that affect the sample quality, among which the method of discretization is most crucial. By carefully examining the learned diffusion process, we propose Diffusion Exponential Integrator Sampler~(DEIS). It is based on the Exponential Integrator designed for discretizing ordinary differential equations (ODEs) and leverages a semilinear structure of the learned diffusion process to reduce the discretization error. The proposed method can be applied to any DMs and can generate high-fidelity samples in as few as 10 steps. In our experiments, it takes about 3 minutes on one A6000 GPU to generate $50k$ images from CIFAR10. Moreover, by directly using pre-trained DMs, we achieve the state-of-art sampling performance when the number of score function evaluation~(NFE) is limited, e.g., 4.17 FID with 10 NFEs, 3.37 FID, and 9.74 IS with only 15 NFEs on CIFAR10. Code is available at https://github.com/qsh-zh/deis"
                },
                {
                    "title": "Adaptive Cholesky Gaussian Processes",
                    "abstract": "We present a method to approximate Gaussian process regression models for large datasets by considering only a subset of the data. Our approach is novel in that the size of the subset is selected on the fly during exact inference with little computational overhead. From an empirical observation that the log-marginal likelihood often exhibits a linear trend once a sufficient subset of a dataset has been observed, we conclude that many large datasets contain redundant information that only slightly affects the posterior. Based on this, we provide probabilistic bounds on the full model evidence that can identify such subsets. Remarkably, these bounds are largely composed of terms that appear in intermediate steps of the standard Cholesky decomposition, allowing us to modify the algorithm to adaptively stop the decomposition once enough data have been observed."
                },
                {
                    "title": "On Neural Differential Equations",
                    "abstract": "The conjoining of dynamical systems and deep learning has become a topic of great interest. In particular, neural differential equations (NDEs) demonstrate that neural networks and differential equation are two sides of the same coin. Traditional parameterised differential equations are a special case. Many popular neural network architectures, such as residual networks and recurrent networks, are discretisations. NDEs are suitable for tackling generative problems, dynamical systems, and time series (particularly in physics, finance, ...) and are thus of interest to both modern machine learning and traditional mathematical modelling. NDEs offer high-capacity function approximation, strong priors on model space, the ability to handle irregular data, memory efficiency, and a wealth of available theory on both sides. This doctoral thesis provides an in-depth survey of the field. Topics include: neural ordinary differential equations (e.g. for hybrid neural/mechanistic modelling of physical systems); neural controlled differential equations (e.g. for learning functions of irregular time series); and neural stochastic differential equations (e.g. to produce generative models capable of representing complex stochastic dynamics, or sampling from complex high-dimensional distributions). Further topics include: numerical methods for NDEs (e.g. reversible differential equations solvers, backpropagation through differential equations, Brownian reconstruction); symbolic regression for dynamical systems (e.g. via regularised evolution); and deep implicit models (e.g. deep equilibrium models, differentiable optimisation). We anticipate this thesis will be of interest to anyone interested in the marriage of deep learning with dynamical systems, and hope it will provide a useful reference for the current state of the art."
                },
                {
                    "title": "Preconditioning for Scalable Gaussian Process Hyperparameter Optimization",
                    "abstract": "Gaussian process hyperparameter optimization requires linear solves with, and log -determinants of, large kernel matrices. Iterative numerical tech-niques are becoming popular to scale to larger datasets, relying on the conjugate gradient method (CG) for the linear solves and stochastic trace estimation for the log -determinant. This work introduces new algorithmic and theoretical in-sights for preconditioning these computations. While preconditioning is well understood in the context of CG, we demonstrate that it can also accelerate convergence and reduce variance of the estimates for the log -determinant and its derivative. We prove general probabilistic error bounds for the preconditioned computation of the log -determinant, log -marginal likelihood and its derivatives. Additionally, we derive speci\ufb01c rates for a range of kernel-preconditioner combinations, showing that up to exponential convergence can be achieved. Our theoretical results enable prov-ably ef\ufb01cient optimization of kernel hyperparameters, which we validate empirically on large-scale benchmark problems. There our approach accelerates training by up to an order of magnitude."
                },
                {
                    "title": "Laplace Redux - Effortless Bayesian Deep Learning",
                    "abstract": "Bayesian formulations of deep learning have been shown to have compelling theoretical properties and offer practical functional benefits, such as improved predictive uncertainty quantification and model selection. The Laplace approximation (LA) is a classic, and arguably the simplest family of approximations for the intractable posteriors of deep neural networks. Yet, despite its simplicity, the LA is not as popular as alternatives like variational Bayes or deep ensembles. This may be due to assumptions that the LA is expensive due to the involved Hessian computation, that it is difficult to implement, or that it yields inferior results. In this work we show that these are misconceptions: we (i) review the range of variants of the LA including versions with minimal cost overhead; (ii) introduce\"laplace\", an easy-to-use software library for PyTorch offering user-friendly access to all major flavors of the LA; and (iii) demonstrate through extensive experiments that the LA is competitive with more popular alternatives in terms of performance, while excelling in terms of computational cost. We hope that this work will serve as a catalyst to a wider adoption of the LA in practical deep learning, including in domains where Bayesian approaches are not typically considered at the moment."
                },
                {
                    "title": "Efficient and Modular Implicit Differentiation",
                    "abstract": "Automatic differentiation (autodiff) has revolutionized machine learning. It allows to express complex computations by composing elementary ones in creative ways and removes the burden of computing their derivatives by hand. More recently, differentiation of optimization problem solutions has attracted widespread attention with applications such as optimization layers, and in bi-level problems such as hyper-parameter optimization and meta-learning. However, so far, implicit differentiation remained difficult to use for practitioners, as it often required case-by-case tedious mathematical derivations and implementations. In this paper, we propose automatic implicit differentiation, an efficient and modular approach for implicit differentiation of optimization problems. In our approach, the user defines directly in Python a function $F$ capturing the optimality conditions of the problem to be differentiated. Once this is done, we leverage autodiff of $F$ and the implicit function theorem to automatically differentiate the optimization problem. Our approach thus combines the benefits of implicit differentiation and autodiff. It is efficient as it can be added on top of any state-of-the-art solver and modular as the optimality condition specification is decoupled from the implicit differentiation mechanism. We show that seemingly simple principles allow to recover many existing implicit differentiation methods and create new ones easily. We demonstrate the ease of formulating and solving bi-level optimization problems using our framework. We also showcase an application to the sensitivity analysis of molecular dynamics."
                },
                {
                    "title": "Scalable Marginal Likelihood Estimation for Model Selection in Deep Learning",
                    "abstract": "Marginal-likelihood based model-selection, even though promising, is rarely used in deep learning due to estimation difficulties. Instead, most approaches rely on validation data, which may not be readily available. In this work, we present a scalable marginal-likelihood estimation method to select both hyperparameters and network architectures, based on the training data alone. Some hyperparameters can be estimated online during training, simplifying the procedure. Our marginal-likelihood estimate is based on Laplace's method and Gauss-Newton approximations to the Hessian, and it outperforms cross-validation and manual-tuning on standard regression and image classification datasets, especially in terms of calibration and out-of-distribution detection. Our work shows that marginal likelihoods can improve generalization and be useful when validation data is unavailable (e.g., in nonstationary settings)."
                },
                {
                    "title": "Sketching Curvature for Efficient Out-of-Distribution Detection for Deep Neural Networks",
                    "abstract": "In order to safely deploy Deep Neural Networks (DNNs) within the perception pipelines of real-time decision making systems, there is a need for safeguards that can detect out-of-training-distribution (OoD) inputs both efficiently and accurately. Building on recent work leveraging the local curvature of DNNs to reason about epistemic uncertainty, we propose Sketching Curvature of OoD Detection (SCOD), an architecture-agnostic framework for equipping any trained DNN with a task-relevant epistemic uncertainty estimate. Offline, given a trained model and its training data, SCOD employs tools from matrix sketching to tractably compute a low-rank approximation of the Fisher information matrix, which characterizes which directions in the weight space are most influential on the predictions over the training data. Online, we estimate uncertainty by measuring how much perturbations orthogonal to these directions can alter predictions at a new test input. We apply SCOD to pre-trained networks of varying architectures on several tasks, ranging from regression to classification. We demonstrate that SCOD achieves comparable or better OoD detection performance with lower computational burden relative to existing baselines."
                },
                {
                    "title": "QR and LQ Decomposition Matrix Backpropagation Algorithms for Square, Wide, and Deep -- Real or Complex -- Matrices and Their Software Implementation.",
                    "abstract": "This article presents matrix backpropagation algorithms for the QR decomposition of matrices $A_{m, n}$, that are either square (m = n), wide (m n), with rank $k = min(m, n)$. Furthermore, we derive novel matrix backpropagation results for the pivoted (full-rank) QR decomposition and for the LQ decomposition of deep input matrices. Differentiable QR decomposition offers a numerically stable, computationally efficient method to solve least squares problems frequently encountered in machine learning and computer vision. Other use cases such as graph learning and network compression are listed in the article. Software implementation across popular deep learning frameworks (PyTorch, TensorFlow, MXNet) incorporate the methods for general use within the deep learning community. Furthermore, this article aids the practitioner in understanding the matrix backpropagation methodology as part of larger computational graphs."
                },
                {
                    "title": "Bayesian Deep Learning via Subnetwork Inference",
                    "abstract": "The Bayesian paradigm has the potential to solve core issues of deep neural networks such as poor calibration and data inefficiency. Alas, scaling Bayesian inference to large weight spaces often requires restrictive approximations. In this work, we show that it suffices to perform inference over a small subset of model weights in order to obtain accurate predictive posteriors. The other weights are kept as point estimates. This subnetwork inference framework enables us to use expressive, otherwise intractable, posterior approximations over such subsets. In particular, we implement subnetwork linearized Laplace as a simple, scalable Bayesian deep learning method: We first obtain a MAP estimate of all weights and then infer a full-covariance Gaussian posterior over a subnetwork using the linearized Laplace approximation. We propose a subnetwork selection strategy that aims to maximally preserve the model's predictive uncertainty. Empirically, our approach compares favorably to ensembles and less expressive posterior approximations over full networks. Our proposed subnetwork (linearized) Laplace method is implemented within the laplace PyTorch library at https://github.com/AlexImmer/Laplace."
                },
                {
                    "title": "Traces of Class/Cross-Class Structure Pervade Deep Learning Spectra",
                    "abstract": "Numerous researchers recently applied empirical spectral analysis to the study of modern deep learning classifiers. We identify and discuss an important formal class/cross-class structure and show how it lies at the origin of the many visually striking features observed in deepnet spectra, some of which were reported in recent articles, others are unveiled here for the first time. These include spectral outliers, \"spikes\", and small but distinct continuous distributions, \"bumps\", often seen beyond the edge of a \"main bulk\". \nThe significance of the cross-class structure is illustrated in three ways: (i) we prove the ratio of outliers to bulk in the spectrum of the Fisher information matrix is predictive of misclassification, in the context of multinomial logistic regression; (ii) we demonstrate how, gradually with depth, a network is able to separate class-distinctive information from class variability, all while orthogonalizing the class-distinctive information; and (iii) we propose a correction to KFAC, a well-known second-order optimization algorithm for training deepnets."
                },
                {
                    "title": "Generative Flows with Matrix Exponential",
                    "abstract": "Generative flows models enjoy the properties of tractable exact likelihood and efficient sampling, which are composed of a sequence of invertible functions. In this paper, we incorporate matrix exponential into generative flows. Matrix exponential is a map from matrices to invertible matrices, this property is suitable for generative flows. Based on matrix exponential, we propose matrix exponential coupling layers that are a general case of affine coupling layers and matrix exponential invertible 1 x 1 convolutions that do not collapse during training. And we modify the networks architecture to make trainingstable andsignificantly speed up the training process. Our experiments show that our model achieves great performance on density estimation amongst generative flows models."
                },
                {
                    "title": "Fast Matrix Square Roots with Applications to Gaussian Processes and Bayesian Optimization",
                    "abstract": "Matrix square roots and their inverses arise frequently in machine learning, e.g., when sampling from high-dimensional Gaussians $\\mathcal{N}(\\mathbf 0, \\mathbf K)$ or whitening a vector $\\mathbf b$ against covariance matrix $\\mathbf K$. While existing methods typically require $O(N^3)$ computation, we introduce a highly-efficient quadratic-time algorithm for computing $\\mathbf K^{1/2} \\mathbf b$, $\\mathbf K^{-1/2} \\mathbf b$, and their derivatives through matrix-vector multiplication (MVMs). Our method combines Krylov subspace methods with a rational approximation and typically achieves $4$ decimal places of accuracy with fewer than $100$ MVMs. Moreover, the backward pass requires little additional computation. We demonstrate our method's applicability on matrices as large as $50,\\!000 \\times 50,\\!000$ - well beyond traditional methods - with little approximation error. Applying this increased scalability to variational Gaussian processes, Bayesian optimization, and Gibbs sampling results in more powerful models with higher accuracy."
                },
                {
                    "title": "The Convolution Exponential and Generalized Sylvester Flows",
                    "abstract": "This paper introduces a new method to build linear flows, by taking the exponential of a linear transformation. This linear transformation does not need to be invertible itself, and the exponential has the following desirable properties: it is guaranteed to be invertible, its inverse is straightforward to compute and the log Jacobian determinant is equal to the trace of the linear transformation. An important insight is that the exponential can be computed implicitly, which allows the use of convolutional layers. Using this insight, we develop new invertible transformations named convolution exponentials and graph convolution exponentials, which retain the equivariance of their underlying transformations. In addition, we generalize Sylvester Flows and propose Convolutional Sylvester Flows which are based on the generalization and the convolution exponential as basis change. Empirically, we show that the convolution exponential outperforms other linear transformations in generative flows on CIFAR10 and the graph convolution exponential improves the performance of graph normalizing flows. In addition, we show that Convolutional Sylvester Flows improve performance over residual flows as a generative flow model measured in log-likelihood."
                },
                {
                    "title": "Kernel Operations on the GPU, with Autodiff, without Memory Overflows",
                    "abstract": "The KeOps library provides a fast and memory-efficient GPU support for tensors whose entries are given by a mathematical formula, such as kernel and distance matrices. KeOps alleviates the major bottleneck of tensor-centric libraries for kernel and geometric applications: memory consumption. It also supports automatic differentiation and outperforms standard GPU baselines, including PyTorch CUDA tensors or the Halide and TVM libraries. KeOps combines optimized C++/CUDA schemes with binders for high-level languages: Python (Numpy and PyTorch), Matlab and GNU R. As a result, high-level \"quadratic\" codes can now scale up to large data sets with millions of samples processed in seconds. KeOps brings graphics-like performances for kernel methods and is freely available on standard repositories (PyPi, CRAN). To showcase its versatility, we provide tutorials in a wide range of settings online at \\url{this http URL}."
                },
                {
                    "title": "Rethinking Parameter Counting in Deep Models: Effective Dimensionality Revisited",
                    "abstract": "Neural networks appear to have mysterious generalization properties when using parameter counting as a proxy for complexity. Indeed, neural networks often have many more parameters than there are data points, yet still provide good generalization performance. Moreover, when we measure generalization as a function of parameters, we see double descent behaviour, where the test error decreases, increases, and then again decreases. We show that many of these properties become understandable when viewed through the lens of effective dimensionality, which measures the dimensionality of the parameter space determined by the data. We relate effective dimensionality to posterior contraction in Bayesian deep learning, model selection, width-depth tradeoffs, double descent, and functional diversity in loss surfaces, leading to a richer understanding of the interplay between parameters and functions in deep models. We also show that effective dimensionality compares favourably to alternative norm- and flatness- based generalization measures."
                },
                {
                    "title": "Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU Networks",
                    "abstract": "The point estimates of ReLU classification networks---arguably the most widely used neural network architecture---have been shown to yield arbitrarily high confidence far away from the training data. This architecture, in conjunction with a maximum a posteriori estimation scheme, is thus not calibrated nor robust. Approximate Bayesian inference has been empirically demonstrated to improve predictive uncertainty in neural networks, although the theoretical analysis of such Bayesian approximations is limited. We theoretically analyze approximate Gaussian distributions on the weights of ReLU networks and show that they fix the overconfidence problem. Furthermore, we show that even a simplistic, thus cheap, Bayesian approximation, also fixes these issues. This indicates that a sufficient condition for a calibrated uncertainty on a ReLU network is \"to be a bit Bayesian\". These theoretical results validate the usage of last-layer Bayesian approximation and motivate a range of a fidelity-cost trade-off. We further validate these findings empirically via various standard experiments using common deep ReLU networks and Laplace approximations."
                },
                {
                    "title": "How to train your neural ODE",
                    "abstract": "Training neural ODEs on large datasets has not been tractable due to the necessity of allowing the adaptive numerical ODE solver to refine its step size to very small values. In practice this leads to dynamics equivalent to many hundreds or even thousands of layers. In this paper, we overcome this apparent difficulty by introducing a theoretically-grounded combination of both optimal transport and stability regularizations which encourage neural ODEs to prefer simpler dynamics out of all the dynamics that solve a problem well. Simpler dynamics lead to faster convergence and to fewer discretizations of the solver, considerably decreasing wall-clock time without loss in performance. Our approach allows us to train neural ODE based generative models to the same performance as the unregularized dynamics in just over a day on one GPU, whereas unregularized dynamics can take up to 4-6 days of training time on multiple GPUs. This brings neural ODEs significantly closer to practical relevance in large-scale applications."
                },
                {
                    "title": "The Discrete Adjoint Method: Efficient Derivatives for Functions of Discrete Sequences",
                    "abstract": "Gradient-based techniques are becoming increasingly critical in quantitative fields, notably in statistics and computer science. The utility of these techniques, however, ultimately depends on how efficiently we can evaluate the derivatives of the complex mathematical functions that arise in applications. In this paper we introduce a discrete adjoint method that efficiently evaluates derivatives for functions of discrete sequences."
                },
                {
                    "title": "Residual Flows for Invertible Generative Modeling",
                    "abstract": "Flow-based generative models parameterize probability distributions through an invertible transformation and can be trained by maximum likelihood. Invertible residual networks provide a flexible family of transformations where only Lipschitz conditions rather than strict architectural constraints are needed for enforcing invertibility. However, prior work trained invertible residual networks for density estimation by relying on biased log-density estimates whose bias increased with the network's expressiveness. We give a tractable unbiased estimate of the log density using a \"Russian roulette\" estimator, and reduce the memory required during training by using an alternative infinite series for the gradient. Furthermore, we improve invertible residual blocks by proposing the use of activation functions that avoid derivative saturation and generalizing the Lipschitz condition to induced mixed norms. The resulting approach, called Residual Flows, achieves state-of-the-art performance on density estimation amongst flow-based models, and outperforms networks that use coupling blocks at joint generative and discriminative modeling."
                },
                {
                    "title": "Exact Gaussian Processes on a Million Data Points",
                    "abstract": "Gaussian processes (GPs) are flexible non-parametric models, with a capacity that grows with the available data. However, computational constraints with standard inference procedures have limited exact GPs to problems with fewer than about ten thousand training points, necessitating approximations for larger datasets. In this paper, we develop a scalable approach for exact GPs that leverages multi-GPU parallelization and methods like linear conjugate gradients, accessing the kernel matrix only through matrix multiplication. By partitioning and distributing kernel matrix multiplies, we demonstrate that an exact GP can be trained on over a million points, a task previously thought to be impossible with current computing hardware, in less than 2 hours. Moreover, our approach is generally applicable, without constraints to grid data or specific kernel classes. Enabled by this scalability, we perform the first-ever comparison of exact GPs against scalable GP approximations on datasets with $10^4 \\!-\\! 10^6$ data points, showing dramatic performance improvements."
                },
                {
                    "title": "The SuiteSparse Matrix Collection Website Interface",
                    "abstract": "The SuiteSparse Matrix Collection (formerly known as the University of Florida Sparse Matrix Collection) (Davis & Hu, 2011) has grown significantly since its introduction, with newly added matrices representing almost seven times as much data than the entirety of the original Collection. With this growth, searching the Collection for matrices with specific names, structures, and other properties has become increasingly difficult. To make the Collection more accessible to the scientific computing community, we have developed a web application that allows real-time search and filtering of the Collection matrices."
                },
                {
                    "title": "DiffEqFlux.jl - A Julia Library for Neural Differential Equations",
                    "abstract": "DiffEqFlux.jl is a library for fusing neural networks and differential equations. In this work we describe differential equations from the viewpoint of data science and discuss the complementary nature between machine learning models and differential equations. We demonstrate the ability to incorporate DifferentialEquations.jl-defined differential equation problems into a Flux-defined neural network, and vice versa. The advantages of being able to use the entire DifferentialEquations.jl suite for this purpose is demonstrated by counter examples where simple integration strategies fail, but the sophisticated integration strategies provided by the DifferentialEquations.jl library succeed. This is followed by a demonstration of delay differential equations and stochastic differential equations inside of neural networks. We show high-level functionality for defining neural ordinary differential equations (neural networks embedded into the differential equation) and describe the extra models in the Flux model zoo which includes neural stochastic differential equations. We conclude by discussing the various adjoint methods used for backpropogation of the differential equation solvers. DiffEqFlux.jl is an important contribution to the area, as it allows the full weight of the differential equation solvers developed from decades of research in the scientific computing field to be readily applied to the challenges posed by machine learning and data science."
                },
                {
                    "title": "A Comparison of Automatic Differentiation and Continuous Sensitivity Analysis for Derivatives of Differential Equation Solutions",
                    "abstract": "Derivatives of differential equation solutions are commonly for parameter estimation, fitting neural differential equations, and as model diagnostics. However, with a litany of choices and a Cartesian product of potential methods, it can be difficult for practitioners to understand which method is likely to be the most effective on their particular application. In this manuscript we investigate the performance characteristics of Discrete Local Sensitivity Analysis implemented via Automatic Differentiation (DSAAD) against continuous adjoint sensitivity analysis. Non-stiff and stiff biological and pharmacometric models, including a PDE discretization, are used to quantify the performance of sensitivity analysis methods. Our benchmarks show that on small stiff and non-stiff systems of ODEs (approximately < 100 parameters+ODEs), forward-mode DSAAD is more efficient than both reverse-mode and continuous forward/adjoint sensitivity analysis. The scalability of continuous adjoint methods is shown to be more efficient than discrete adjoints and forward methods after crossing this size range. These comparative studies demonstrate a trade-off between memory usage and performance in the continuous adjoint methods that should be considered when choosing the technique, while numerically unstable backsolve techniques from the machine learning literature are demonstrated as unsuitable for most scientific models. The performance of adjoint methods is shown to be heavily tied to the reverse-mode AD method used for the vector-Jacobian product calculations, with tape-based AD methods shown to be 2 orders of magnitude slower on nonlinear partial differential equations than static AD techniques. In addition, these results demonstrate the out-of-the-box applicability of DSAAD to differential-algebraic equations, delay differential equations, and hybrid differential equation systems where the event timing and effects are dependent on model parameters, showcasing an ease of implementation advantage for DSAAD approaches. Together, these benchmarks provide a guide to help practitioners to quickly identify the best mixture of continuous sensitivities and automatic differentiation for their applications."
                },
                {
                    "title": "Invertible Residual Networks",
                    "abstract": "We show that standard ResNet architectures can be made invertible, allowing the same model to be used for classification, density estimation, and generation. Typically, enforcing invertibility requires partitioning dimensions or restricting network architectures. In contrast, our approach only requires adding a simple normalization step during training, already available in standard frameworks. Invertible ResNets define a generative model which can be trained by maximum likelihood on unlabeled data. To compute likelihoods, we introduce a tractable approximation to the Jacobian log-determinant of a residual block. Our empirical evaluation shows that invertible ResNets perform competitively with both state-of-the-art image classifiers and flow-based generative models, something that has not been previously achieved with a single architecture."
                },
                {
                    "title": "GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration",
                    "abstract": "Despite advances in scalable models, the inference tools used for Gaussian processes (GPs) have yet to fully capitalize on developments in computing hardware. We present an efficient and general approach to GP inference based on Blackbox Matrix-Matrix multiplication (BBMM). BBMM inference uses a modified batched version of the conjugate gradients algorithm to derive all terms for training and inference in a single call. BBMM reduces the asymptotic complexity of exact GP inference from $O(n^3)$ to $O(n^2)$. Adapting this algorithm to scalable approximations and complex GP models simply requires a routine for efficient matrix-matrix multiplication with the kernel and its derivative. In addition, BBMM uses a specialized preconditioner to substantially speed up convergence. In experiments we show that BBMM effectively uses GPU hardware to dramatically accelerate both exact GP inference and scalable approximations. Additionally, we provide GPyTorch, a software platform for scalable GP inference via BBMM, built on PyTorch."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "Backpropagation for Implicit Spectral Densities",
                    "abstract": "Most successful machine intelligence systems rely on gradient-based learning, which is made possible by backpropagation. Some systems are designed to aid us in interpreting data when explicit goals cannot be provided. These unsupervised systems are commonly trained by backpropagating through a likelihood function. We introduce a tool that allows us to do this even when the likelihood is not explicitly set, by instead using the implicit likelihood of the model. Explicitly defining the likelihood often entails making heavy-handed assumptions that impede our ability to solve challenging tasks. On the other hand, the implicit likelihood of the model is accessible without the need for such assumptions. Our tool, which we call spectral backpropagation, allows us to optimize it in much greater generality than what has been attempted before. GANs can also be viewed as a technique for optimizing implicit likelihoods. We study them using spectral backpropagation in order to demonstrate robustness for high-dimensional problems, and identify two novel properties of the generator G: (1) there exist aberrant, nonsensical outputs to which G assigns very high likelihood, and (2) the eigenvectors of the metric induced by G over latent space correspond to quasi-disentangled explanatory factors."
                },
                {
                    "title": "Online Structured Laplace Approximations For Overcoming Catastrophic Forgetting",
                    "abstract": "We introduce the Kronecker factored online Laplace approximation for overcoming catastrophic forgetting in neural networks. The method is grounded in a Bayesian online learning framework, where we recursively approximate the posterior after every task with a Gaussian, leading to a quadratic penalty on changes to the weights. The Laplace approximation requires calculating the Hessian around a mode, which is typically intractable for modern architectures. In order to make our method scalable, we leverage recent block-diagonal Kronecker factored approximations to the curvature. Our algorithm achieves over 90% test accuracy across a sequence of 50 instantiations of the permuted MNIST dataset, substantially outperforming related methods for overcoming catastrophic forgetting."
                },
                {
                    "title": "Constant-Time Predictive Distributions for Gaussian Processes",
                    "abstract": "One of the most compelling features of Gaussian process (GP) regression is its ability to provide well-calibrated posterior distributions. Recent advances in inducing point methods have sped up GP marginal likelihood and posterior mean computations, leaving posterior covariance estimation and sampling as the remaining computational bottlenecks. In this paper we address these shortcomings by using the Lanczos algorithm to rapidly approximate the predictive covariance matrix. Our approach, which we refer to as LOVE (LanczOs Variance Estimates), substantially improves time and space complexity. In our experiments, LOVE computes covariances up to 2,000 times faster and draws samples 18,000 times faster than existing methods, all without sacrificing accuracy."
                },
                {
                    "title": "A Scalable Laplace Approximation for Neural Networks",
                    "abstract": "We leverage recent insights from second-order optimisation for neural networks to construct a Kronecker factored Laplace approximation to the posterior over the weights of a trained network. Our approximation requires no modi\ufb01cation of the training procedure, enabling practitioners to estimate the uncertainty of their models currently used in production without having to retrain them. We extensively compare our method to using Dropout and a diagonal Laplace approximation for estimating the uncertainty of a network. We demonstrate that our Kronecker factored method leads to better uncertainty estimates on out-of-distribution data and is more robust to simple adversarial attacks. Our approach only requires calculating two square curvature factor matrices for each layer. Their size is equal to the respective square of the input and output size of the layer, making the method ef\ufb01cient both computationally and in terms of memory usage. We illustrate its scalability by applying it to a state-of-the-art convolutional network architecture."
                },
                {
                    "title": "Iteration Method for the Solution of the Eigenvalue Problem of Linear Differential and Integral Operators",
                    "abstract": "The present investigation designs a systematic method for finding the latent roots and the principal axes of a matrix, without reducing the order of the matrix. It is characterized by a wide field of applicability and great accuracy, since the accumulation of rounding errors is avoided, through the process of \"minimized iterations\". Moreover, the method leads to a well convergent successive approximation procedure by which the solution of integral equations of the Fredholm type and the solution of the eigenvalue problem of linear differential and integral operators may be accomplished."
                },
                {
                    "title": "Scalable Log Determinants for Gaussian Process Kernel Learning",
                    "abstract": "For applications as varied as Bayesian neural networks, determinantal point processes, elliptical graphical models, and kernel learning for Gaussian processes (GPs), one must compute a log determinant of an $n \\times n$ positive definite matrix, and its derivatives - leading to prohibitive $\\mathcal{O}(n^3)$ computations. We propose novel $\\mathcal{O}(n)$ approaches to estimating these quantities from only fast matrix vector multiplications (MVMs). These stochastic approximations are based on Chebyshev, Lanczos, and surrogate models, and converge quickly even for kernel matrices that have challenging spectra. We leverage these approximations to develop a scalable Gaussian process approach to kernel learning. We find that Lanczos is generally superior to Chebyshev for kernel learning, and that a surrogate approach can be highly efficient and accurate with popular kernels."
                },
                {
                    "title": "Auto-Differentiating Linear Algebra",
                    "abstract": "Development systems for deep learning, such as Theano, Torch, TensorFlow, or MXNet, are easy-to-use tools for creating complex neural network models. Since gradient computations are automatically baked in, and execution is mapped to high performance hardware, these models can be trained end-to-end on large amounts of data. However, it is currently not easy to implement many basic machine learning primitives in these systems (such as Gaussian processes, least squares estimation, principal components analysis, Kalman smoothing), mainly because they lack efficient support of linear algebra primitives as differentiable operators. We detail how a number of matrix decompositions (Cholesky, LQ, symmetric eigen) can be implemented as differentiable operators. We have implemented these primitives in MXNet, running on CPU and GPU in single and double precision. We sketch use cases of these new operators, learning Gaussian process and Bayesian linear regression models. Our implementation is based on BLAS/LAPACK APIs, for which highly tuned implementations are available on all major CPUs and GPUs."
                },
                {
                    "title": "Fast Estimation of tr(f(A)) via Stochastic Lanczos Quadrature",
                    "abstract": "The problem of estimating the trace of matrix functions appears in applications ranging from machine learning and scientific computing, to computational biology. This paper presents an inexpensive method to estimate the trace of $f(A)$ for cases where $f$ is analytic inside a closed interval and $A$ is a symmetric positive definite matrix. The method combines three key ingredients, namely, the stochastic trace estimator, Gaussian quadrature, and the Lanczos algorithm. As examples, we consider the problems of estimating the log-determinant ($f(t)=\\log(t)$), the Schatten $p$-norms ($f(t)=t^{p/2}$), the Estrada index ($f(t)=e^{t}$), and the trace of the matrix inverse ($f(t)=t^{-1}$). We establish multiplicative and additive error bounds for the approximations obtained by this method. In addition, we present error bounds for other useful tools such as approximating the log-likelihood function in the context of maximum likelihood estimation of Gaussian processes. Numerical experiments illustrate the performan..."
                },
                {
                    "title": "DifferentialEquations.jl \u2013 A Performant and Feature-Rich Ecosystem for\n Solving Differential Equations in Julia",
                    "abstract": "DifferentialEquations.jl is a package for solving differential equations in Julia. It covers discrete equations (function maps, discrete stochastic (Gillespie/Markov) simulations), ordinary differential equations, stochastic differential equations, algebraic differential equations, delay differential equations, hybrid differential equations, jump diffusions, and (stochastic) partial differential equations. Through extensive use of multiple dispatch, metaprogramming, plot recipes, foreign function interfaces (FFI), and call-overloading, DifferentialEquations.jl offers a unified user interface to solve and analyze various forms of differential equations while not sacrificing features or performance. Many modern features are integrated into the solvers, such as allowing arbitrary user-defined number systems for high-precision and arithmetic with physical units, built-in multithreading and parallelism, and symbolic calculation of Jacobians. Integrated into the package is an algorithm testing and benchmarking suite to both ensure accuracy and serve as an easy way for researchers to develop and distribute their own methods. Together, these features build a highly extendable suite which is feature-rich and highly performant. Funding statement: This work was partially supported by NIH grants P50GM76516 and R01GM107264 and NSF grants DMS1562176 and DMS1161621. This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No. DGE-1321846, the National Academies of Science, Engineering, and Medicine via the Ford Foundation, and the National Institutes of Health Award T32 EB009418. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH."
                },
                {
                    "title": "Preconditioning Kernel Matrices",
                    "abstract": "The computational and storage complexity of kernel machines presents the primary barrier to their scaling to large, modern, datasets. A common way to tackle the scalability issue is to use the conjugate gradient algorithm, which relieves the constraints on both storage (the kernel matrix need not be stored) and computation (both stochastic gradients and parallelization can be used). Even so, conjugate gradient is not without its own issues: the conditioning of kernel matrices is often such that conjugate gradients will have poor convergence in practice. Preconditioning is a common approach to alleviating this issue. Here we propose preconditioned conjugate gradients for kernel machines, and develop a broad range of preconditioners particularly useful for kernel matrices. We describe a scalable approach to both solving kernel machines and learning their hyperparameters. We show this approach is exact in the limit of iterations and outperforms state-of-the-art approximations for a given computational budget."
                },
                {
                    "title": "Variational Inference with Normalizing Flows",
                    "abstract": "The choice of approximate posterior distribution is one of the core problems in variational inference. Most applications of variational inference employ simple families of posterior approximations in order to allow for efficient inference, focusing on mean-field or other simple structured approximations. This restriction has a significant impact on the quality of inferences made using variational methods. We introduce a new approach for specifying flexible, arbitrarily complex and scalable approximate posterior distributions. Our approximations are distributions constructed through a normalizing flow, whereby a simple initial density is transformed into a more complex one by applying a sequence of invertible transformations until a desired level of complexity is attained. We use this view of normalizing flows to develop categories of finite and infinitesimal flows and provide a unified view of approaches for constructing rich posterior approximations. We demonstrate that the theoretical advantages of having posteriors that better match the true posterior, combined with the scalability of amortized variational approaches, provides a clear improvement in performance and applicability of variational inference."
                },
                {
                    "title": "Kernel Interpolation for Scalable Structured Gaussian Processes (KISS-GP)",
                    "abstract": "We introduce a new structured kernel interpolation (SKI) framework, which generalises and unifies inducing point methods for scalable Gaussian processes (GPs). SKI methods produce kernel approximations for fast computations through kernel interpolation. The SKI framework clarifies how the quality of an inducing point approach depends on the number of inducing (aka interpolation) points, interpolation strategy, and GP covariance kernel. SKI also provides a mechanism to create new scalable kernel methods, through choosing different kernel interpolation strategies. Using SKI, with local cubic kernel interpolation, we introduce KISSGP, which is 1) more scalable than inducing point alternatives, 2) naturally enables Kronecker and Toeplitz algebra for substantial additional gains in scalability, without requiring any grid data, and 3) can be used for fast and expressive kernel learning. KISS-GP costs O(n) time and storage for GP inference. We evaluate KISS-GP for kernel matrix approximation, kernel learning, and natural sound modelling."
                },
                {
                    "title": "Discrete-Time Solutions to the Continuous-Time Differential Lyapunov Equation With Applications to Kalman Filtering",
                    "abstract": "Prediction and filtering of continuous-time stochastic processes often require a solver of a continuous-time differential Lyapunov equation (CDLE), for example the time update in the Kalman filter. Even though this can be recast into an ordinary differential equation (ODE), where standard solvers can be applied, the dominating approach in Kalman filter applications is to discretize the system and then apply the discrete-time difference Lyapunov equation (DDLE). To avoid problems with stability and poor accuracy, oversampling is often used. This contribution analyzes over-sampling strategies, and proposes a novel low-complexity analytical solution that does not involve oversampling. The results are illustrated on Kalman filtering problems in both linear and nonlinear systems."
                },
                {
                    "title": "Scalable Bayesian Optimization Using Deep Neural Networks",
                    "abstract": "Bayesian optimization is an effective methodology for the global optimization of functions with expensive evaluations. It relies on querying a distribution over functions defined by a relatively cheap surrogate model. An accurate model for this distribution over functions is critical to the effectiveness of the approach, and is typically fit using Gaussian processes (GPs). However, since GPs scale cubically with the number of observations, it has been challenging to handle objectives whose optimization requires many evaluations, and as such, massively parallelizing the optimization. \n \nIn this work, we explore the use of neural networks as an alternative to GPs to model distributions over functions. We show that performing adaptive basis function regression with a neural network as the parametric form performs competitively with state-of-the-art GP-based approaches, but scales linearly with the number of data rather than cubically. This allows us to achieve a previously intractable degree of parallelism, which we apply to large scale hyperparameter optimization, rapidly finding competitive models on benchmark object recognition tasks using convolutional networks, and image caption generation using neural language models."
                },
                {
                    "title": "Obtaining Well Calibrated Probabilities Using Bayesian Binning",
                    "abstract": "Learning probabilistic predictive models that are well calibrated is critical for many prediction and decision-making tasks in artificial intelligence. In this paper we present a new non-parametric calibration method called Bayesian Binning into Quantiles (BBQ) which addresses key limitations of existing calibration methods. The method post processes the output of a binary classification algorithm; thus, it can be readily combined with many existing classification algorithms. The method is computationally tractable, and empirically accurate, as evidenced by the set of experiments reported here on both real and simulated datasets."
                },
                {
                    "title": "Learning Deep Features for Scene Recognition using Places Database",
                    "abstract": "Scene recognition is one of the hallmark tasks of computer vision, allowing definition of a context for object recognition. Whereas the tremendous recent progress in object recognition tasks is due to the availability of large datasets like ImageNet and the rise of Convolutional Neural Networks (CNNs) for learning high-level features, performance at scene recognition has not attained the same level of success. This may be because current deep features trained from ImageNet are not competitive enough for such tasks. Here, we introduce a new scene-centric database called Places with over 7 million labeled pictures of scenes. We propose new methods to compare the density and diversity of image datasets and show that Places is as dense as other scene datasets and has more diversity. Using CNN, we learn deep features for scene recognition tasks, and establish new state-of-the-art results on several scene-centric datasets. A visualization of the CNN layers' responses allows us to show differences in the internal representations of object-centric and scene-centric networks."
                },
                {
                    "title": "Numerical Methods for Eigenvalue Problems",
                    "abstract": "This textbook presents a number of the most important numerical methods for finding eigenvalues and eigenvectors of matrices. The authors discuss the central ideas underlying the different algorithms and introduce the theoretical concepts required to analyze their behaviour. Several programming examples allow the reader to experience the behaviour of the different algorithms first-hand. The book addresses students and lecturers of mathematics and engineering who are interested in the fundamental ideas of modern numerical methods and want to learn how to apply and extend these ideas to solve new problems."
                },
                {
                    "title": "The university of Florida sparse matrix collection",
                    "abstract": "We describe the University of Florida Sparse Matrix Collection, a large and actively growing set of sparse matrices that arise in real applications. The Collection is widely used by the numerical linear algebra community for the development and performance evaluation of sparse matrix algorithms. It allows for robust and repeatable experiments: robust because performance results with artificially generated matrices can be misleading, and repeatable because matrices are curated and made publicly available in many formats. Its matrices cover a wide spectrum of domains, include those arising from problems with underlying 2D or 3D geometry (as structural engineering, computational fluid dynamics, model reduction, electromagnetics, semiconductor devices, thermodynamics, materials, acoustics, computer graphics/vision, robotics/kinematics, and other discretizations) and those that typically do not have such geometry (optimization, circuit simulation, economic and financial modeling, theoretical and quantum chemistry, chemical process simulation, mathematics and statistics, power networks, and other networks and graphs). We provide software for accessing and managing the Collection, from MATLAB#8482;, Mathematica#8482;, Fortran, and C, as well as an online search capability. Graph visualization of the matrices is provided, and a new multilevel coarsening scheme is proposed to facilitate this task."
                },
                {
                    "title": "Exponential integrators",
                    "abstract": "In this paper we consider the construction, analysis, implementation and application of exponential integrators. The focus will be on two types of stiff problems. The first one is characterized by a Jacobian that possesses eigenvalues with large negative real parts. Parabolic partial differential equations and their spatial discretization are typical examples. The second class consists of highly oscillatory problems with purely imaginary eigenvalues of large modulus. Apart from motivating the construction of exponential integrators for various classes of problems, our main intention in this article is to present the mathematics behind these methods. We will derive error bounds that are independent of stiffness or highest frequencies in the system. Since the implementation of exponential integrators requires the evaluation of the product of a matrix function with a vector, we will briefly discuss some possible approaches as well. The paper concludes with some applications, in which exponential integrators are used."
                },
                {
                    "title": "Algorithmic Differentiation of Linear Algebra Functions with Application in Optimum Experimental Design (Extended Version)",
                    "abstract": "We derive algorithms for higher order derivative computation of the rectangular QR and eigenvalue decomposition of symmetric matrices with distinct eigenvalues in the forward and reverse mode of algorithmic differentiation (AD) using univariate Taylor propagation of matrices (UTPM). Linear algebra functions are regarded as elementary functions and not as algorithms. The presented algorithms are implemented in the BSD licensed AD tool ALGOPY. Numerical tests show that the UTPM algorithms derived in this paper produce results close to machine precision accuracy. The theory developed in this paper is applied to compute the gradient of an objective function motivated from optimum experimental design:"
                },
                {
                    "title": "Matrices, Moments and Quadrature with Applications",
                    "abstract": "This computationally oriented book describes and explains the mathematical relationships among matrices, moments, orthogonal polynomials, quadrature rules, and the Lanczos and conjugate gradient algorithms. The book bridges different mathematical areas to obtain algorithms to estimate bilinear forms involving two vectors and a function of the matrix. The first part of the book provides the necessary mathematical background and explains the theory. The second part describes the applications and gives numerical examples of the algorithms and techniques developed in the first part.Applications addressed in the book include computing elements of functions of matrices; obtaining estimates of the error norm in iterative methods for solving linear systems and computing parameters in least squares and total least squares; and solving ill-posed problems using Tikhonov regularization.This book will interest researchers in numerical linear algebra and matrix computations, as well as scientists and engineers working on problems involving computation of bilinear forms."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Computing the Fr\u00e9chet Derivative of the Matrix Exponential, with an Application to Condition Number Estimation",
                    "abstract": "The matrix exponential is a much-studied matrix function having many applications. The Frechet derivative of the matrix exponential describes the first-order sensitivity of $e^A$ to perturbations in $A$ and its norm determines a condition number for $e^A$. Among the numerous methods for computing $e^A$ the scaling and squaring method is the most widely used. We show that the implementation of the method in [N. J. Higham, The scaling and squaring method for the matrix exponential revisited, SIAM J. Matrix Anal. Appl., 26 (2005), pp. 1179-1193] can be extended to compute both $e^A$ and the Frechet derivative at $A$ in the direction $E$, denoted by $L(A,E)$, at a cost about three times that for computing $e^A$ alone. The algorithm is derived from the scaling and squaring method by differentiating the Pade approximants and the squaring recurrence, reusing quantities computed during the evaluation of the Pade approximant, and intertwining the recurrences in the squaring phase. To guide the choice of algorithmic parameters, an extension of the existing backward error analysis for the scaling and squaring method is developed which shows that, modulo rounding errors, the approximations obtained are $e^{A+\\Delta A}$ and $L(A+\\Delta A,E+\\Delta E)$, with the same $\\Delta A$ in both cases, and with computable bounds on $\\|\\Delta A\\|$ and $\\|\\Delta E\\|$. The algorithm for $L(A,E)$ is used to develop an algorithm that computes $e^A$ together with an estimate of its condition number. In addition to results specific to the exponential, we develop some results and techniques for arbitrary functions. We show how a matrix iteration for $f(A)$ yields an iteration for the Frechet derivative and show how to efficiently compute the Frechet derivative of a power series. We also show that a matrix polynomial and its Frechet derivative can be evaluated at a cost at most three times that of computing the polynomial itself and give a general framework for evaluating a matrix function and its Frechet derivative via Pade approximation."
                },
                {
                    "title": "Functions of matrices - theory and computation",
                    "abstract": "The only book devoted exclusively to matrix functions, this research monograph gives a thorough treatment of the theory of matrix functions and numerical methods for computing them. The author s elegant presentation focuses on the equivalent definitions of f(A) via the Jordan canonical form, polynomial interpolation, and the Cauchy integral formula, and features an emphasis on results of practical interest and an extensive collection of problems and solutions. Functions of Matrices: Theory and Computation is more than just a monograph on matrix functions; its wide-ranging content including an overview of applications, historical references, and miscellaneous results, tricks, and techniques with an f(A) connection makes it useful as a general reference in numerical linear algebra. Other key features of the book include development of the theory of conditioning and properties of the Frchet derivative; an emphasis on the Schur decomposition, the block Parlett recurrence, and judicious use of Pad approximants; the inclusion of new, unpublished research results and improved algorithms; a chapter devoted to the f(A)b problem; and a MATLAB toolbox providing implementations of the key algorithms. Audience: This book is for specialists in numerical analysis and applied linear algebra as well as anyone wishing to learn about the theory of matrix functions and state of the art methods for computing them. It can be used for a graduate-level course on functions of matrices and is a suitable reference for an advanced course on applied or numerical linear algebra. It is also particularly well suited for self-study. Contents: List of Figures; List of Tables; Preface; Chapter 1: Theory of Matrix Functions; Chapter 2: Applications; Chapter 3: Conditioning; Chapter 4: Techniques for General Functions; Chapter 5: Matrix Sign Function; Chapter 6: Matrix Square Root; Chapter 7: Matrix pth Root; Chapter 8: The Polar Decomposition; Chapter 9: Schur-Parlett Algorithm; Chapter 10: Matrix Exponential; Chapter 11: Matrix Logarithm; Chapter 12: Matrix Cosine and Sine; Chapter 13: Function of Matrix Times Vector: f(A)b; Chapter 14: Miscellany; Appendix A: Notation; Appendix B: Background: Definitions and Useful Facts; Appendix C: Operation Counts; Appendix D: Matrix Function Toolbox; Appendix E: Solutions to Problems; Bibliography; Index."
                },
                {
                    "title": "Cytoscape: a software environment for integrated models of biomolecular interaction networks.",
                    "abstract": "Cytoscape is an open source software project for integrating biomolecular interaction networks with high-throughput expression data and other molecular states into a unified conceptual framework. Although applicable to any system of molecular components and interactions, Cytoscape is most powerful when used in conjunction with large databases of protein-protein, protein-DNA, and genetic interactions that are increasingly available for humans and model organisms. Cytoscape's software Core provides basic functionality to layout and query the network; to visually integrate the network with expression profiles, phenotypes, and other molecular states; and to link the network to databases of functional annotations. The Core is extensible through a straightforward plug-in architecture, allowing rapid development of additional computational analyses and features. Several case studies of Cytoscape plug-ins are surveyed, including a search for interaction pathways correlating with changes in gene expression, a study of protein complexes involved in cellular recovery to DNA damage, inference of a combined physical/functional interaction network for Halobacterium, and an interface to detailed stochastic/kinetic gene regulatory models."
                },
                {
                    "title": "Fast Curvature Matrix-Vector Products for Second-Order Gradient Descent",
                    "abstract": "We propose a generic method for iteratively approximating various second-order gradient steps-Newton, Gauss-Newton, Levenberg-Marquardt, and natural gradient-in linear time per iteration, using special curvature matrix-vector products that can be computed in O(n). Two recent acceleration techniques for on-line learning, matrix momentum and stochastic meta-descent (SMD), implement this approach. Since both were originally derived by very different routes, this offers fresh insight into their operation, resulting in further improvements to SMD."
                },
                {
                    "title": "Adjoint Sensitivity Analysis for Differential-Algebraic Equations: The Adjoint DAE System and Its Numerical Solution",
                    "abstract": "An adjoint sensitivity method is presented for parameter-dependent differential-algebraic equation systems (DAEs). The adjoint system is derived, along with conditions for its consistent initialization, for DAEs of index up to two (Hessenberg). For stable linear DAEs, stability of the adjoint system (for semi-explicit DAEs) or of an augmented adjoint system (for fully implicit DAEs) is shown. In addition, it is shown for these systems that numerical stability is maintained for the adjoint system or for the augmented adjoint system."
                },
                {
                    "title": "Exponential Integrators for Large Systems of Differential Equations",
                    "abstract": "We study the numerical integration of large stiff systems of differential equations by methods that use matrix--vector products with the exponential or a related function of the Jacobian. For large problems, these can be approximated by Krylov subspace methods, which typically converge faster than those for the solution of the linear systems arising in standard stiff integrators. The exponential methods also offer favorable properties in the integration of differential equations whose Jacobian has large imaginary eigenvalues. We derive methods up to order 4 which are exact for linear constant-coefficient equations. The implementation of the methods is discussed. Numerical experiments with reaction-diffusion problems and a time-dependent Schrodinger equation are included."
                },
                {
                    "title": "On Krylov Subspace Approximations to the Matrix Exponential Operator",
                    "abstract": "Krylov subspace methods for approximating the action of matrix exponentials are analyzed in this paper. We derive error bounds via a functional calculus of Arnoldi and Lanczos methods that reduces the study of Krylov subspace approximations of functions of matrices to that of linear systems of equations. As a side result, we obtain error bounds for Galerkin-type Krylov methods for linear equations, namely, the biconjugate gradient method and the full orthogonalization method. For Krylov approximations to matrix exponentials, we show superlinear error decay from relatively small iteration numbers onwards, depending on the geometry of the numerical range, the spectrum, or the pseudospectrum. The convergence to exp$(\\tau A)v$ is faster than that of corresponding Krylov methods for the solution of linear equations $(I-\\tau A)x=v$, which usually arise in the numerical solution of stiff ordinary differential equations (ODEs). We therefore propose a new class of time integration methods for large systems of nonlinear differential equations which use Krylov approximations to the exponential function of the Jacobian instead of solving linear or nonlinear systems of equations in every time step."
                },
                {
                    "title": "Efficient Solution of Parabolic Equations by Krylov Approximation Methods",
                    "abstract": "This paper takes a new look at numerical techniques for solving parabolic equations by the method of lines. The main motivation for the proposed approach is the possibility of exploiting a high degree of parallelism in a simple manner. The basic idea of the method is to approximate the action of the evolution operator on a given state vector by means of a projection process onto a Krylov subspace. Thus the resulting approximation consists of applying an evolution operator of very small dimension to a known vector, which is, in turn, computed accurately by exploiting high-order rational Chebyshev and Pade approximations to the exponential. Because the rational approximation is only applied to a small matrix, the only operations required with the original large matrix are matrix-by-vector multiplications and, as a result, the algorithm can easily be parallelized and vectorized. Further parallelism is introduced by expanding the rational approximations into partial fractions. Some relevant approximation and ..."
                },
                {
                    "title": "Bayesian Interpolation",
                    "abstract": "Although Bayesian analysis has been in use since Laplace, the Bayesian method of model-comparison has only recently been developed in depth. In this paper, the Bayesian approach to regularization and model-comparison is demonstrated by studying the inference problem of interpolating noisy data. The concepts and methods described are quite general and can be applied to many other data modeling problems. Regularizing constants are set by examining their posterior probability distribution. Alternative regularizers (priors) and alternative basis sets are objectively compared by evaluating the evidence for them. Occam's razor is automatically embodied by this process. The way in which Bayes infers the values of regularizing constants and noise levels has an elegant interpretation in terms of the effective number of parameters determined by the data set. This framework is due to Gull and Skilling."
                },
                {
                    "title": "Transforming Neural-Net Output Levels to Probability Distributions",
                    "abstract": "(1) The outputs of a typical multi-output classification network do not satisfy the axioms of probability; probabilities should be positive and sum to one. This problem can be solved by treating the trained network as a preprocessor that produces a feature vector that can be further processed, for instance by classical statistical estimation techniques. (2) We present a method for computing the first two moments of the probability distribution indicating the range of outputs that are consistent with the input and the training data. It is particularly useful to combine these two ideas: we implement the ideas of section 1 using Parzen windows, where the shape and relative size of each window is computed using the ideas of section 2. This allows us to make contact between important theoretical ideas (e.g. the ensemble formalism) and practical techniques (e.g. back-prop). Our results also shed new light on and generalize the well-known \"softmax\" scheme."
                },
                {
                    "title": "Solving the discretized time\u2010independent Schr\u00f6dinger equation with the Lanczos procedure",
                    "abstract": "A new method is presented to find bound state solutions of the one\u2010, two\u2010, or three\u2010dimensional Schrodinger equation. The equation is turned into a sparse matrix eigenvalue problem by representing the potential energy surface and the wave function on a grid. The Laplacian is represented by a high (10th) order finite difference formula. Eigenvalues are found by the Lanczos procedure [J. Cullum and R. A. Willoughby, J. Comp. Phys. 44, 329 (1981)] and transition probabilities (Franck\u2013Condon factors) are found by the recursive residue generation method [A. Nauts and R. E. Wyatt, Phys. Rev. Lett. 51, 2238 (1983)]. Examples are given for the 1D Morse oscillator and the 2D Henon\u2010Heiles potential. Numerical convergence can be checked easily and highly accurate results can be obtained. The algorithm is fast, easy to implement, and vectorizable."
                },
                {
                    "title": "Sparse matrix test problems",
                    "abstract": "The development, analysis and production of algorithms in sparse linear algebra often requires the use of test problems to demonstrate the effectiveness and applicability of the algorithms. Many algorithms have been developed in the context of specific application areas and have been tested in the context of sets of test problems collected by the developers. Comparisons of algorithms across application areas and comparisons between algorithms has often been incomplete, due to the lack of a comprehensive set of test problems. Additionally we believe that a comprehensive set of test problems will lead to a better understanding of the range of structures in sparse matrix problems and thence to better classification and development of algorithms. We have agreed to sponsor and maintain a general library of sparse matrix test problems, available on request to anyone for a nominal fee to cover postal charges. Contributors to the library will, of course, receive a free copy."
                },
                {
                    "title": "Nineteen Dubious Ways to Compute the Exponential of a Matrix, Twenty-Five Years Later",
                    "abstract": "In principle, the exponential of a matrix could be computed in many ways. Methods involving approximation theory, differential equations, the matrix eigenvalues, and the matrix characteristic polyn..."
                },
                {
                    "title": "Methods of conjugate gradients for solving linear systems",
                    "abstract": "An iterative algorithm is given for solving a system Ax=k of n linear equations in n unknowns. The solution is given in n steps. It is shown that this method is a special case of a very general method which also includes Gaussian elimination. These general algorithms are essentially algorithms for finding an n dimensional ellipsoid. Connections are made with the theory of orthogonal polynomials and continued fractions."
                },
                {
                    "title": "The principle of minimized iterations in the solution of the matrix eigenvalue problem",
                    "abstract": "An interpretation of Dr. Cornelius Lanczos' iteration method, which he has named \"minimized iterations\", is discussed in this article, expounding the method as applied to the solution of the characteristic matrix equations both in homogeneous and nonhomogeneous form. This interpretation leads to a variation of the Lanczos procedure which may frequently be advantageous by virtue of reducing the volume of numerical work in practical applications. Both methods employ essentially the same algorithm, requiring the generation of a series of orthogonal functions through which a simple matrix equation of reduced order is established. The reduced matrix equation may be solved directly in terms of certain polynomial functions obtained in conjunction with the generated orthogonal functions, and the convergence of the solution may be observed as the order of the reduced matrix is successively increased with the order of the original matrix as a limit. The method of minimized iterations is recommended as a rapid means for determining a small number of the larger eigenvalues and modal columns of a large matrix and as a desirable alternative for various series expansions of the Fredholm problem. 1. The conventional iterative procedures. It is frequently required that real latent roots, or eigenvalues, and modal columns be determined for a real numerical matrix, u, of order, n, in the characteristic homogeneous equation,*"
                },
                {
                    "title": "Improving predictions of Bayesian neural nets via local linearization",
                    "abstract": "The generalized Gauss-Newton (ggn) approximation is often used to make practical Bayesian deep learning approaches scalable by replacing a second order derivative with a product of first order derivatives. In this paper we argue that the ggn approximation should be understood as a local linearization of the underlying Bayesian neural network (bnn), which turns the bnn into a generalized linear model (glm). Because we use this linearized model for posterior inference, we should also predict using this modified model instead of the original one. We refer to this modified predictive as \u201cglm predictive\u201d and show that it effectively resolves common underfitting problems of the Laplace approximation. It extends previous results in this vein to general likelihoods and has an equivalent Gaussian process formulation, which enables alternative inference schemes for bnns in function space. We demonstrate the effectiveness of our approach on several standard classification datasets and on out-of-distribution detection. We provide an implementation at https://github.com/ AlexImmer/BNN-predictions."
                },
                {
                    "title": "The Krylov Subspace Methods for the Computation of Matrix Exponentials",
                    "abstract": "OF DISSERTATION The Krylov Subspace Methods for the Computation of Matrix Exponentials The problem of computing the matrix exponential e arises in many theoretical and practical problems. Many methods have been developed to accurately and efficiently compute this matrix function or its product with a vector, i.e., ev. In the past few decades, with the increasing need of the computation for large sparse matrices, iterative methods such as the Krylov subspace methods have proved to be a powerful class of methods in dealing with many linear algebra problems. The Krylov subspace methods have been introduced for computing matrix exponentials by Gallopoulos and Saad, and the corresponding error bounds that aim at explaining the convergence properties have been extensively studied. Many of those bounds show that the speed of convergence depends on the norm of the matrix, while some others emphasize the important role played by the spectral distribution. For example, it is shown in a recent work by Ye that the speed of convergence is also determined by the condition number for a symmetric negative definite matrix. Namely the convergence is fast for a well-conditioned matrix no matter how large the norm is. In this dissertation, we derive new error bounds for computing ev for nonsymmetric A, using the spectral information of A. Our result is based on the assumption that A is negative definite, i.e., the field of values of A lies entirely in the left half of the complex plane, such that the underlying dynamic system is stable. The new bounds show that the speed of convergence is related to the size and shape of the rectangle containing the field of values, and they agree with the existing results when A is symmetric. Furthermore, we also derive a simpler error bound for the special case when A is skew-Hermitian. This bound explains an observed convergence behavior where the approximation error initially stagnates for certain number of iterations before it starts to converge. In deriving our new error bounds, we use sharper estimates of the decay property of exponentials of Hessenberg matrices, by constructing Faber polynomial approximating exponential function in the region containing the field of values. The Jacobi elliptic functions are used to construct the conformal mappings and generate the Faber polynomials. We also present numerical tests to demonstrate the behavior of the new error bounds."
                },
                {
                    "title": "Transductive and Inductive Methods for Approximate Gaussian Process Regression",
                    "abstract": "Gaussian process regression allows a simple analytical treatment of exact Bayesian inference and has been found to provide good performance, yet scales badly with the number of training data. In this paper we compare several approaches towards scaling Gaussian processes regression to large data sets: the subset of representers method, the reduced rank approximation, online Gaussian processes, and the Bayesian committee machine. Furthermore we provide theoretical insight into some of our experimental results. We found that subset of representers methods can give good and particularly fast predictions for data sets with high and medium noise levels. On complex low noise data sets, the Bayesian committee machine achieves significantly better accuracy, yet at a higher computational cost."
                },
                {
                    "title": "Evaluating derivatives - principles and techniques of algorithmic differentiation, Second Edition",
                    "abstract": "Algorithmic, or automatic, differentiation (AD) is a growing area of theoretical research and software development concerned with the accurate and efficient evaluation of derivatives for function evaluations given as computer programs. The resulting derivative values are useful for all scientific computations that are based on linear, quadratic, or higher order approximations to nonlinear scalar or vector functions. AD has been applied in particular to optimization, parameter identification, nonlinear equation solving, the numerical integration of differential equations, and combinations of these. Apart from quantifying sensitivities numerically, AD also yields structural dependence information, such as the sparsity pattern and generic rank of Jacobian matrices. The field opens up an exciting opportunity to develop new algorithms that reflect the true cost of accurate derivatives and to use them for improvements in speed and reliability. This second edition has been updated and expanded to cover recent developments in applications and theory, including an elegant NP completeness argument by Uwe Naumann and a brief introduction to scarcity, a generalization of sparsity. There is also added material on checkpointing and iterative differentiation. To improve readability the more detailed analysis of memory and complexity bounds has been relegated to separate, optional chapters.The book consists of three parts: a stand-alone introduction to the fundamentals of AD and its software; a thorough treatment of methods for sparse problems; and final chapters on program-reversal schedules, higher derivatives, nonsmooth problems and iterative processes. Each of the 15 chapters concludes with examples and exercises. Audience: This volume will be valuable to designers of algorithms and software for nonlinear computational problems. Current numerical software users should gain the insight necessary to choose and deploy existing AD software tools to the best advantage. Contents: Rules; Preface; Prologue; Mathematical Symbols; Chapter 1: Introduction; Chapter 2: A Framework for Evaluating Functions; Chapter 3: Fundamentals of Forward and Reverse; Chapter 4: Memory Issues and Complexity Bounds; Chapter 5: Repeating and Extending Reverse; Chapter 6: Implementation and Software; Chapter 7: Sparse Forward and Reverse; Chapter 8: Exploiting Sparsity by Compression; Chapter 9: Going beyond Forward and Reverse; Chapter 10: Jacobian and Hessian Accumulation; Chapter 11: Observations on Efficiency; Chapter 12: Reversal Schedules and Checkpointing; Chapter 13: Taylor and Tensor Coefficients; Chapter 14: Differentiation without Differentiability; Chapter 15: Implicit and Iterative Differentiation; Epilogue; List of Figures; List of Tables; Assumptions and Definitions; Propositions, Corollaries, and Lemmas; Bibliography; Index"
                },
                {
                    "title": "Old and New Matrix Algebra Useful for Statistics",
                    "abstract": "The partials with respect to the numerator are laid out according to the shape of Y while the partials with respect to the denominator are laid out according to the transpose of X. For example, dy/dx is a column vector while dy/dx is a row vector (assuming x and y are column vectors\u2014otherwise it is flipped). Each of these derivatives can be tediously computed via partials, but this section shows how they instead can be computed with matrix manipulations. The material is based on Magnus and Neudecker (1988)."
                },
                {
                    "title": "A stochastic estimator of the trace of the influence matrix for laplacian smoothing splines",
                    "abstract": "An unbiased stochastic estimator of tr(I-A), where A is the influence matrix associated with the calculation of Laplacian smoothing splines, is described. The estimator is similar to one recently developed by Girard but satisfies a minimum variance criterion and does not require the simulation of a standard normal variable. It uses instead simulations of the discrete random variable which takes the values 1, -1 each with probability 1/2. Bounds on the variance of the estimator, similar to those established by Girard, are obtained using elementary methods. The estimator can be used to approximately minimize generalised cross validation (GCV) when using discretized iterative methods for fitting Laplacian smoothing splines to very large data sets. Simulated examples show that the estimated trace values, using either the estimator presented here or the estimator of Girard, perform almost as well as the exact values when applied to the minimization of GCV for n as small as a few hundred, where n is the number ..."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.NA",
                "math.NA",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we efficiently compute the gradients of functions of large matrices in machine learning without explicitly constructing the full Jacobian matrix?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning, particularly in the development of large-scale models such as Bayesian neural networks and neural ODEs. By enabling efficient gradient computation for matrix functions, this research could lead to significant improvements in model training and hyperparameter optimization. The implications extend to various applications, including Gaussian processes and PDEs, potentially leading to more robust and scalable machine learning algorithms. Addressing this question could also inspire further research into differentiable programming and automatic differentiation techniques, fostering innovation in model design and optimization strategies.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to handle large matrices without instantiating them, which complicates the computation of gradients. Naive approaches may fail because they often rely on constructing the full Jacobian, leading to prohibitive memory and time costs. The technical obstacles include ensuring that the method remains efficient in terms of both time and memory complexity while accurately computing the required gradients. Additionally, the theoretical understanding of differentiable linear algebra in this context is still underdeveloped, making it difficult to apply existing techniques directly.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely overlooked the specific challenges associated with differentiating functions of large matrices, focusing instead on simpler cases or relying on full Jacobian constructions. Barriers include a lack of efficient algorithms that can operate in a matrix-free manner and insufficient theoretical frameworks to guide the development of such methods. Our approach differs by introducing a novel matrix-free algorithm that computes gradients in linear time and memory complexity, addressing the limitations of prior work and providing a more scalable solution for large-scale machine learning applications.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a matrix-free algorithm for automatically differentiating functions of matrices, specifically targeting the computation of matrix-function-vector products. We will utilize a variety of datasets relevant to applications such as Bayesian neural networks and Gaussian processes, measuring performance using metrics like gradient accuracy and computational efficiency. The expected outcomes include demonstrating the effectiveness of our algorithm in producing exact gradients while maintaining linear time and memory complexity, ultimately showcasing its applicability in large-scale machine learning models."
            }
        },
        "author_data": {
            "eb93ee76-688d-4685-ad27-8de2a6aa45b3": {
                "pk": "eb93ee76-688d-4685-ad27-8de2a6aa45b3",
                "name": "Nicholas Kr\u00e4mer",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "da848d4f-f729-499b-b2ed-6d0146d1d0cf": {
                "pk": "da848d4f-f729-499b-b2ed-6d0146d1d0cf",
                "name": "Pablo Moreno-Mu\u00f1oz",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "e45536db-fe3b-4330-bc46-4d1c199b4e09": {
                "pk": "e45536db-fe3b-4330-bc46-4d1c199b4e09",
                "name": "Hrittik Roy",
                "collaborators": [
                    "Marco Miani",
                    "S\u00f8ren Hauberg",
                    "Alison Pouplin",
                    "Sidak Pal Singh",
                    "Georgios Arvanitidis",
                    "Carl Henrik Ek",
                    "Philipp Hennig",
                    "Marvin Pf\u00f6rtner",
                    "Lukas Tatzel"
                ],
                "domain": [
                    "Bayesian Deep Learning",
                    "Uncertainty Quantification",
                    "Neural Networks",
                    "Riemannian Geometry"
                ],
                "publications": [
                    {
                        "title": "Bayes without Underfitting: Fully Correlated Deep Learning Posteriors via Alternating Projections",
                        "abstract": "Bayesian deep learning all too often underfits so that the Bayesian prediction is less accurate than a simple point estimate. Uncertainty quantification then comes at the cost of accuracy. For linearized models, the null space of the generalized Gauss-Newton matrix corresponds to parameters that preserve the training predictions of the point estimate. We propose to build Bayesian approximations in this null space, thereby guaranteeing that the Bayesian predictive does not underfit. We suggest a matrix-free algorithm for projecting onto this null space, which scales linearly with the number of parameters and quadratically with the number of output dimensions. We further propose an approximation that only scales linearly with parameters to make the method applicable to generative models. An extensive empirical evaluation shows that the approach scales to large models, including vision transformers with 28 million parameters."
                    },
                    {
                        "title": "On the curvature of the loss landscape",
                        "abstract": "One of the main challenges in modern deep learning is to understand why such over-parameterized models perform so well when trained on finite data. A way to analyze this generalization concept is through the properties of the associated loss landscape. In this work, we consider the loss landscape as an embedded Riemannian manifold and show that the differential geometric properties of the manifold can be used when analyzing the generalization abilities of a deep net. In particular, we focus on the scalar curvature, which can be computed analytically for our manifold, and show connections to several settings that potentially imply generalization."
                    },
                    {
                        "title": "Reparameterization invariance in approximate Bayesian inference",
                        "abstract": "Current approximate posteriors in Bayesian neural networks (BNNs) exhibit a crucial limitation: they fail to maintain invariance under reparameterization, i.e. BNNs assign different posterior densities to different parametrizations of identical functions. This creates a fundamental flaw in the application of Bayesian principles as it breaks the correspondence between uncertainty over the parameters with uncertainty over the parametrized function. In this paper, we investigate this issue in the context of the increasingly popular linearized Laplace approximation. Specifically, it has been observed that linearized predictives alleviate the common underfitting problems of the Laplace approximation. We develop a new geometric view of reparametrizations from which we explain the success of linearization. Moreover, we demonstrate that these reparameterization invariance properties can be extended to the original neural network predictive using a Riemannian diffusion process giving a straightforward algorithm for approximate posterior sampling, which empirically improves posterior fit."
                    }
                ]
            },
            "ff13657c-2176-42f4-a303-3315e66c50f4": {
                "pk": "ff13657c-2176-42f4-a303-3315e66c50f4",
                "name": "S\u00f8ren Hauberg",
                "collaborators": [
                    "Georgios Arvanitidis",
                    "Lars Kai Hansen",
                    "Martin J\u00f8rgensen",
                    "David Eklund",
                    "Philipp Hennig",
                    "Andri Bergsson",
                    "Oren Freifeld",
                    "Anders Boesen Lindbo Larsen",
                    "John W. Fisher III",
                    "Alessandra Tosi"
                ],
                "domain": [
                    "Manifold Learning",
                    "Riemannian Geometry",
                    "Generative Models",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "Visualizing Riemannian data with Rie-SNE",
                        "abstract": "Faithful visualizations of data residing on manifolds must take the underlying geometry into account when producing a flat planar view of the data. In this paper, we extend the classic stochastic neighbor embedding (SNE) algorithm to data on general Riemannian manifolds. We replace standard Gaussian assumptions with Riemannian diffusion counterparts and propose an efficient approximation that only requires access to calculations of Riemannian distances and volumes. We demonstrate that the approach also allows for mapping data from one manifold to another, e.g. from a high-dimensional sphere to a low-dimensional one."
                    },
                    {
                        "title": "Dreaming More Data: Class-dependent Distributions over Diffeomorphisms for Learned Data Augmentation",
                        "abstract": "Data augmentation is a key element in training high-dimensional models. In this approach, one synthesizes new observations by applying pre-specified transformations to the original training data; e.g.~new images are formed by rotating old ones. Current augmentation schemes, however, rely on manual specification of the applied transformations, making data augmentation an implicit form of feature engineering. With an eye towards true end-to-end learning, we suggest learning the applied transformations on a per-class basis. Particularly, we align image pairs within each class under the assumption that the spatial transformation between images belongs to a large class of diffeomorphisms. We then learn a class-specific probabilistic generative models of the transformations in a Riemannian submanifold of the Lie group of diffeomorphisms. We demonstrate significant performance improvements in training deep neural nets over manually-specified augmentation schemes. Our code and augmented datasets are available online."
                    },
                    {
                        "title": "Only Bayes should learn a manifold (on the estimation of differential geometric structure from data)",
                        "abstract": "We investigate learning of the differential geometric structure of a data manifold embedded in a high-dimensional Euclidean space. We first analyze kernel-based algorithms and show that under the usual regularizations, non-probabilistic methods cannot recover the differential geometric structure, but instead find mostly linear manifolds or spaces equipped with teleports. To properly learn the differential geometric structure, non-probabilistic methods must apply regularizations that enforce large gradients, which go against common wisdom. We repeat the analysis for probabilistic methods and find that under reasonable priors, the geometric structure can be recovered. Fully exploiting the recovered structure, however, requires the development of stochastic extensions to classic Riemannian geometry. We take early steps in that regard. Finally, we partly extend the analysis to modern models based on neural networks, thereby highlighting geometric and probabilistic shortcomings of current deep generative models."
                    },
                    {
                        "title": "Expected path length on random manifolds",
                        "abstract": "Manifold learning seeks a low dimensional representation that faithfully captures the essence of data. Current methods can successfully learn such representations, but do not provide a meaningful set of operations that are associated with the representation. Working towards operational representation learning, we endow the latent space of a large class of generative models with a random Riemannian metric, which provides us with elementary operators. As computational tools are unavailable for random Riemannian manifolds, we study deterministic approximations and derive tight error bounds on expected distances."
                    },
                    {
                        "title": "Isometric Gaussian Process Latent Variable Model for Dissimilarity Data",
                        "abstract": "We present a probabilistic model where the latent variable respects both the distances and the topology of the modeled data. The model leverages the Riemannian geometry of the generated manifold to endow the latent space with a well-defined stochastic distance measure, which is modeled locally as Nakagami distributions. These stochastic distances are sought to be as similar as possible to observed distances along a neighborhood graph through a censoring process. The model is inferred by variational inference based on observations of pairwise distances. We demonstrate how the new model can encode invariances in the learned manifolds."
                    },
                    {
                        "title": "Metrics for Probabilistic Geometries",
                        "abstract": "We investigate the geometrical structure of probabilistic generative dimensionality reduction models using the tools of Riemannian geometry. We explicitly define a distribution over the natural metric given by the models. We provide the necessary algorithms to compute expected metric tensors where the distribution over mappings is given by a Gaussian process. We treat the corresponding latent variable model as a Riemannian manifold and we use the expectation of the metric under the Gaussian process prior to define interpolating paths and measure distance between latent points. We show how distances that respect the expected metric lead to more appropriate generation of new data."
                    },
                    {
                        "title": "Geometrically Enriched Latent Spaces",
                        "abstract": "A common assumption in generative models is that the generator immerses the latent space into a Euclidean ambient space. Instead, we consider the ambient space to be a Riemannian manifold, which allows for encoding domain knowledge through the associated Riemannian metric. Shortest paths can then be defined accordingly in the latent space to both follow the learned manifold and respect the ambient geometry. Through careful design of the ambient metric we can ensure that shortest paths are well-behaved even for deterministic generators that otherwise would exhibit a misleading bias. Experimentally we show that our approach improves interpretability of learned representations both using stochastic and deterministic generators."
                    },
                    {
                        "title": "Sketched Lanczos uncertainty score: a low-memory summary of the Fisher information",
                        "abstract": "Current uncertainty quantification is memory and compute expensive, which hinders practical uptake. To counter, we develop Sketched Lanczos Uncertainty (SLU): an architecture-agnostic uncertainty score that can be applied to pre-trained neural networks with minimal overhead. Importantly, the memory use of SLU only grows logarithmically with the number of model parameters. We combine Lanczos' algorithm with dimensionality reduction techniques to compute a sketch of the leading eigenvectors of a matrix. Applying this novel algorithm to the Fisher information matrix yields a cheap and reliable uncertainty score. Empirically, SLU yields well-calibrated uncertainties, reliably detects out-of-distribution examples, and consistently outperforms existing methods in the low-memory regime."
                    },
                    {
                        "title": "Probabilistic Solutions to Differential Equations and their Application to Riemannian Statistics",
                        "abstract": "We study a probabilistic numerical method for the solution of both boundary and initial value problems that returns a joint Gaussian process posterior over the solution. Such methods have concrete value in the statistics on Riemannian manifolds, where non-analytic ordinary differential equations are involved in virtually all computations. The probabilistic formulation permits marginalising the uncertainty of the numerical solution such that statistics are less sensitive to inaccuracies. This leads to new Riemannian algorithms for mean value computations and principal geodesic analysis. Marginalisation also means results can be less precise than point estimates, enabling a noticeable speed-up over the state of the art. Our approach is an argument for a wider point that uncertainty caused by numerical calculations should be tracked throughout the pipeline of machine learning algorithms."
                    },
                    {
                        "title": "A Locally Adaptive Normal Distribution",
                        "abstract": "The multivariate normal density is a monotonic function of the distance to the mean, and its ellipsoidal shape is due to the underlying Euclidean metric. We suggest to replace this metric with a locally adaptive, smoothly changing (Riemannian) metric that favors regions of high local density. The resulting locally adaptive normal distribution (LAND) is a generalization of the normal distribution to the \"manifold\" setting, where data is assumed to lie near a potentially low-dimensional manifold embedded in $\\mathbb{R}^D$. The LAND is parametric, depending only on a mean and a covariance, and is the maximum entropy distribution under the given metric. The underlying metric is, however, non-parametric. We develop a maximum likelihood algorithm to infer the distribution parameters that relies on a combination of gradient descent and Monte Carlo integration. We further extend the LAND to mixture models, and provide the corresponding EM algorithm. We demonstrate the efficiency of the LAND to fit non-trivial probability distributions over both synthetic data, and EEG measurements of human sleep."
                    },
                    {
                        "title": "Latent Space Oddity: on the Curvature of Deep Generative Models",
                        "abstract": "Deep generative models provide a systematic way to learn nonlinear data distributions, through a set of latent variables and a nonlinear \"generator\" function that maps latent points into the input space. The nonlinearity of the generator imply that the latent space gives a distorted view of the input space. Under mild conditions, we show that this distortion can be characterized by a stochastic Riemannian metric, and demonstrate that distances and interpolants are significantly improved under this metric. This in turn improves probability distributions, sampling algorithms and clustering in the latent space. Our geometric analysis further reveals that current generators provide poor variance estimates and we propose a new generator architecture with vastly improved variance estimates. Results are demonstrated on convolutional and fully connected variational autoencoders, but the formalism easily generalize to other deep generative models."
                    },
                    {
                        "title": "Probabilistic Riemannian submanifold learning with wrapped Gaussian process latent variable models",
                        "abstract": "Latent variable models (LVMs) learn probabilistic models of data manifolds lying in an \\emph{ambient} Euclidean space. In a number of applications, a priori known spatial constraints can shrink the ambient space into a considerably smaller manifold. Additionally, in these applications the Euclidean geometry might induce a suboptimal similarity measure, which could be improved by choosing a different metric. Euclidean models ignore such information and assign probability mass to data points that can never appear as data, and vastly different likelihoods to points that are similar under the desired metric. We propose the wrapped Gaussian process latent variable model (WGPLVM), that extends Gaussian process latent variable models to take values strictly on a given ambient Riemannian manifold, making the model blind to impossible data points. This allows non-linear, probabilistic inference of low-dimensional Riemannian submanifolds from data. Our evaluation on diverse datasets show that we improve performance on several tasks, including encoding, visualization and uncertainty quantification."
                    },
                    {
                        "title": "Geodesic Clustering in Deep Generative Models",
                        "abstract": "Deep generative models are tremendously successful in learning low-dimensional latent representations that well-describe the data. These representations, however, tend to much distort relationships between points, i.e. pairwise distances tend to not reflect semantic similarities well. This renders unsupervised tasks, such as clustering, difficult when working with the latent representations. We demonstrate that taking the geometry of the generative model into account is sufficient to make simple clustering algorithms work well over latent representations. Leaning on the recent finding that deep generative models constitute stochastically immersed Riemannian manifolds, we propose an efficient algorithm for computing geodesics (shortest paths) and computing distances in the latent space, while taking its distortion into account. We further propose a new architecture for modeling uncertainty in variational autoencoders, which is essential for understanding the geometry of deep generative models. Experiments show that the geodesic distance is very likely to reflect the internal structure of the data."
                    },
                    {
                        "title": "Fast and Robust Shortest Paths on Manifolds Learned from Data",
                        "abstract": "We propose a fast, simple and robust algorithm for computing shortest paths and distances on Riemannian manifolds learned from data. This amounts to solving a system of ordinary differential equations (ODEs) subject to boundary conditions. Here standard solvers perform poorly because they require well-behaved Jacobians of the ODE, and usually, manifolds learned from data imply unstable and ill-conditioned Jacobians. Instead, we propose a fixed-point iteration scheme for solving the ODE that avoids Jacobians. This enhances the stability of the solver, while reduces the computational cost. In experiments involving both Riemannian metric learning and deep generative models we demonstrate significant improvements in speed and stability over both general-purpose state-of-the-art solvers as well as over specialized solvers."
                    },
                    {
                        "title": "Reliable training and estimation of variance networks",
                        "abstract": "We propose and investigate new complementary methodologies for estimating predictive variance networks in regression neural networks. We derive a locally aware mini-batching scheme that result in sparse robust gradients, and show how to make unbiased weight updates to a variance network. Further, we formulate a heuristic for robustly fitting both the mean and variance networks post hoc. Finally, we take inspiration from posterior Gaussian processes and propose a network architecture with similar extrapolation properties to Gaussian processes. The proposed methodologies are complementary, and improve upon baseline methods individually. Experimentally, we investigate the impact on predictive uncertainty on multiple datasets and tasks ranging from regression, active learning and generative modeling. Experiments consistently show significant improvements in predictive uncertainty estimation over state-of-the-art methods across tasks and datasets."
                    },
                    {
                        "title": "Explicit Disentanglement of Appearance and Perspective in Generative Models",
                        "abstract": "Disentangled representation learning finds compact, independent and easy-to-interpret factors of the data. Learning such has been shown to require an inductive bias, which we explicitly encode in a generative model of images. Specifically, we propose a model with two latent spaces: one that represents spatial transformations of the input data, and another that represents the transformed data. We find that the latter naturally captures the intrinsic appearance of the data. To realize the generative model, we propose a Variationally Inferred Transformational Autoencoder (VITAE) that incorporates a spatial transformer into a variational autoencoder. We show how to perform inference in the model efficiently by carefully designing the encoders and restricting the transformation class to be diffeomorphic. Empirically, our model separates the visual style from digit type on MNIST, separates shape and pose in images of human bodies and facial features from facial shape on CelebA."
                    },
                    {
                        "title": "Variational Autoencoders with Riemannian Brownian Motion Priors",
                        "abstract": "Variational Autoencoders (VAEs) represent the given data in a low-dimensional latent space, which is generally assumed to be Euclidean. This assumption naturally leads to the common choice of a standard Gaussian prior over continuous latent variables. Recent work has, however, shown that this prior has a detrimental effect on model capacity, leading to subpar performance. We propose that the Euclidean assumption lies at the heart of this failure mode. To counter this, we assume a Riemannian structure over the latent space, which constitutes a more principled geometric view of the latent codes, and replace the standard Gaussian prior with a Riemannian Brownian motion prior. We propose an efficient inference scheme that does not rely on the unknown normalizing factor of this prior. Finally, we demonstrate that this prior significantly increases model capacity using only one additional scalar parameter."
                    },
                    {
                        "title": "Probabilistic Spatial Transformer Networks",
                        "abstract": "Spatial Transformer Networks (STNs) estimate image transformations that can improve downstream tasks by `zooming in' on relevant regions in an image. However, STNs are hard to train and sensitive to mis-predictions of transformations. To circumvent these limitations, we propose a probabilistic extension that estimates a stochastic transformation rather than a deterministic one. Marginalizing transformations allows us to consider each image at multiple poses, which makes the localization task easier and the training more robust. As an additional benefit, the stochastic transformations act as a localized, learned data augmentation that improves the downstream tasks. We show across standard imaging benchmarks and on a challenging real-world dataset that these two properties lead to improved classification performance, robustness and model calibration. We further demonstrate that the approach generalizes to non-visual domains by improving model performance on time-series data."
                    }
                ]
            }
        }
    },
    "2405.12940": {
        "paper_data": {
            "title": "Learning the Infinitesimal Generator of Stochastic Diffusion Processes",
            "url": "http://arxiv.org/abs/2405.12940v1",
            "arxiv_id": "2405.12940",
            "authors": [
                "Vladimir R. Kostic",
                "Karim Lounici",
                "Helene Halconruy",
                "Timothee Devergne",
                "Massimiliano Pontil"
            ],
            "abstract": "We address data-driven learning of the infinitesimal generator of stochastic diffusion processes, essential for understanding numerical simulations of natural and physical systems. The unbounded nature of the generator poses significant challenges, rendering conventional analysis techniques for Hilbert-Schmidt operators ineffective. To overcome this, we introduce a novel framework based on the energy functional for these stochastic processes. Our approach integrates physical priors through an energy-based risk metric in both full and partial knowledge settings. We evaluate the statistical performance of a reduced-rank estimator in reproducing kernel Hilbert spaces (RKHS) in the partial knowledge setting. Notably, our approach provides learning bounds independent of the state space dimension and ensures non-spurious spectral estimation. Additionally, we elucidate how the distortion between the intrinsic energy-induced metric of the stochastic diffusion and the RKHS metric used for generator estimation impacts the spectral learning bounds.",
            "introduction": "   1 Introduction  Continuous-time processes are often modeled using ordinary differential equations (ODEs), assuming deterministic dynamics. However, real systems in science and engineering that are modeled by ODEs are subject to unfeasible-to-model influences, necessitating the extension of deterministic models through stochastic differential equations (SDEs), see [33, 35] and references therein. SDEs are advantageous for modeling inherently random phenomena. For instance, in finance, they specify the stochastic process governing asset behavior, a crucial step in constructing pricing models for financial derivatives [34]. In atomistic simulations, where SDEs are used to model the evolution of atomic systems subjected to thermal fluctuations via the Boltzmann distribution [30].   A diverse range of SDEs can be represented as d\u2062Xt=a\u2062(Xt)\u2062d\u2062t+b\u2062(Xt)\u2062d\u2062Wt\ud835\udc51subscript\ud835\udc4b\ud835\udc61\ud835\udc4esubscript\ud835\udc4b\ud835\udc61\ud835\udc51\ud835\udc61\ud835\udc4fsubscript\ud835\udc4b\ud835\udc61\ud835\udc51subscript\ud835\udc4a\ud835\udc61dX_{t}=a(X_{t})dt+b(X_{t})dW_{t}italic_d italic_X start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT = italic_a ( italic_X start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) italic_d italic_t + italic_b ( italic_X start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) italic_d italic_W start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT, where X0=xsubscript\ud835\udc4b0\ud835\udc65X_{0}=xitalic_X start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT = italic_x. Here, W\ud835\udc4aWitalic_W denotes a (possibly multi-dimensional) Brownian motion, and the functions a\ud835\udc4eaitalic_a and b\ud835\udc4fbitalic_b are commonly known as the drift and diffusion coefficients, respectively. Determining these coefficients from one or more trajectories, whether discretized or continuous, has been a key pursuit in \"diffusion statistics\" since the 1980s, as seen in works like [18, 24, 25]. However, uncovering the drift and diffusion coefficients alone does not reveal all the intrinsic properties of a complex system, such as the metastable states of Langevin dynamics in atomistic simulations [see e.g. 42, and references therein]. Consequently, there has been a shift and growing interest in the Infinitesimal Generator (IG) of an SDE, as its spectral decomposition offers a more in-depth understanding of system dynamics and behavior, thus providing a comprehensive picture beyond the mere identification of coefficients.   Recovering the spectral decomposition of the IG can theoretically be achieved by exploiting the well-studied Transfer Operators (TO) [see 21, and references therein]. TOs represent the average evolution of state functions (observables) over time and, being linear and amenable to spectral decomposition under certain conditions, they offer a valuable means to interpret and analyze nonlinear systems. However, they require evenly sampled data at a high rate, which may be impractical. Additionally, TO approaches are purely data-driven, complicating the incorporation of partial or full knowledge of an SDE into the learning process. Thus, there is growing interest in learning the IG directly from data, as it can handle uneven sampling and integrate SDE knowledge. The challenge lies in the IG being an unbounded operator, unlike TO which are often well-approximated by Hilbert-Schmidt operators with comprehensive statistical theory [22]. Unfortunately, the existing statistical theory collapses when applied to unbounded operators, prompting the need to completely rethink the problem   Related work\u00a0\u00a0 Extensive research has explored learning dynamical systems from data [see the monographs 7, 26, and references therein]. Analytical models for dynamics are often unavailable, motivating the need for data-driven methods. Two prominent approaches have emerged: deep learning-based methods [5, 11, 29], effective for learning complex data representations but lacking statistical analysis, and kernel methods [2, 6, 9, 16, 17, 20, 21, 22, 43], offering solid statistical guarantees for the estimation of the TO but requiring careful selection of the kernel function.",
            "references": [
                {
                    "title": "Consistent Long-Term Forecasting of Ergodic Dynamical Systems",
                    "abstract": "We study the evolution of distributions under the action of an ergodic dynamical system, which may be stochastic in nature. By employing tools from Koopman and transfer operator theory one can evolve any initial distribution of the state forward in time, and we investigate how estimators of these operators perform on long-term forecasting. Motivated by the observation that standard estimators may fail at this task, we introduce a learning paradigm that neatly combines classical techniques of eigenvalue deflation from operator theory and feature centering from statistics. This paradigm applies to any operator estimator based on empirical risk minimization, making them satisfy learning bounds which hold uniformly on the entire trajectory of future distributions, and abide to the conservation of mass for each of the forecasted distributions. Numerical experiments illustrates the advantages of our approach in practice."
                },
                {
                    "title": "Learning invariant representations of time-homogeneous stochastic dynamical systems",
                    "abstract": "We consider the general class of time-homogeneous stochastic dynamical systems, both discrete and continuous, and study the problem of learning a representation of the state that faithfully captures its dynamics. This is instrumental to learning the transfer operator or the generator of the system, which in turn can be used for numerous tasks, such as forecasting and interpreting the system dynamics. We show that the search for a good representation can be cast as an optimization problem over neural networks. Our approach is supported by recent results in statistical learning theory, highlighting the role of approximation error and metric distortion in the learning problem. The objective function we propose is associated with projection operators from the representation space to the data space, overcomes metric distortion, and can be empirically estimated from data. In the discrete-time setting, we further derive a relaxed objective function that is differentiable and numerically well-conditioned. We compare our method against state-of-the-art approaches on different datasets, showing better performance across the board."
                },
                {
                    "title": "The Galerkin method beats Graph-Based Approaches for Spectral Algorithms",
                    "abstract": "Historically, the machine learning community has derived spectral decompositions from graph-based approaches. We break with this approach and prove the statistical and computational superiority of the Galerkin method, which consists in restricting the study to a small set of test functions. In particular, we introduce implementation tricks to deal with differential operators in large dimensions with structured kernels. Finally, we extend on the core principles beyond our approach to apply them to non-linear spaces of functions, such as the ones parameterized by deep neural networks, through loss-based optimization procedures."
                },
                {
                    "title": "Kernelized Diffusion maps",
                    "abstract": "Spectral clustering and diffusion maps are celebrated dimensionality reduction algorithms built on eigen-elements related to the diffusive structure of the data. The core of these procedures is the approximation of a Laplacian through a graph kernel approach, however this local average construction is known to be cursed by the high-dimension d. In this article, we build a different estimator of the Laplacian, via a reproducing kernel Hilbert space method, which adapts naturally to the regularity of the problem. We provide non-asymptotic statistical rates proving that the kernel estimator we build can circumvent the curse of dimensionality. Finally we discuss techniques (Nystr\\\"om subsampling, Fourier features) that enable to reduce the computational cost of the estimator while not degrading its overall performance."
                },
                {
                    "title": "Sharp Spectral Rates for Koopman Operator Learning",
                    "abstract": "Non-linear dynamical systems can be handily described by the associated Koopman operator, whose action evolves every observable of the system forward in time. Learning the Koopman operator and its spectral decomposition from data is enabled by a number of algorithms. In this work we present for the first time non-asymptotic learning bounds for the Koopman eigenvalues and eigenfunctions. We focus on time-reversal-invariant stochastic dynamical systems, including the important example of Langevin dynamics. We analyze two popular estimators: Extended Dynamic Mode Decomposition (EDMD) and Reduced Rank Regression (RRR). Our results critically hinge on novel minimax estimation bounds for the operator norm error, that may be of independent interest. Our spectral learning bounds are driven by the simultaneous control of the operator norm error and a novel metric distortion functional of the estimated eigenfunctions. The bounds indicates that both EDMD and RRR have similar variance, but EDMD suffers from a larger bias which might be detrimental to its learning rate. Our results shed new light on the emergence of spurious eigenvalues, an issue which is well known empirically. Numerical experiments illustrate the implications of the bounds in practice."
                },
                {
                    "title": "Optimal Rates for Regularized Conditional Mean Embedding Learning",
                    "abstract": "We address the consistency of a kernel ridge regression estimate of the conditional mean embedding (CME), which is an embedding of the conditional distribution of $Y$ given $X$ into a target reproducing kernel Hilbert space $\\mathcal{H}_Y$. The CME allows us to take conditional expectations of target RKHS functions, and has been employed in nonparametric causal and Bayesian inference. We address the misspecified setting, where the target CME is in the space of Hilbert-Schmidt operators acting from an input interpolation space between $\\mathcal{H}_X$ and $L_2$, to $\\mathcal{H}_Y$. This space of operators is shown to be isomorphic to a newly defined vector-valued interpolation space. Using this isomorphism, we derive a novel and adaptive statistical learning rate for the empirical CME estimator under the misspecified setting. Our analysis reveals that our rates match the optimal $O(\\log n / n)$ rates without assuming $\\mathcal{H}_Y$ to be finite dimensional. We further establish a lower bound on the learning rate, which shows that the obtained upper bound is optimal."
                },
                {
                    "title": "Learning Dynamical Systems via Koopman Operator Regression in Reproducing Kernel Hilbert Spaces",
                    "abstract": "We study a class of dynamical systems modelled as Markov chains that admit an invariant distribution via the corresponding transfer, or Koopman, operator. While data-driven algorithms to reconstruct such operators are well known, their relationship with statistical learning is largely unexplored. We formalize a framework to learn the Koopman operator from finite data trajectories of the dynamical system. We consider the restriction of this operator to a reproducing kernel Hilbert space and introduce a notion of risk, from which different estimators naturally arise. We link the risk with the estimation of the spectral decomposition of the Koopman operator. These observations motivate a reduced-rank operator regression (RRR) estimator. We derive learning bounds for the proposed estimator, holding both in i.i.d. and non i.i.d. settings, the latter in terms of mixing coefficients. Our results suggest RRR might be beneficial over other widely used estimators as confirmed in numerical experiments both for forecasting and mode decomposition."
                },
                {
                    "title": "Learning Stable Koopman Embeddings",
                    "abstract": "In this paper, we present a new data-driven method for learning stable models of nonlinear systems. Our model lifts the original state space to a higher-dimensional linear manifold using Koopman embeddings. Interestingly, we prove that every discrete-time nonlinear contracting model can be learnt in our framework. Another significant merit of the proposed approach is that it allows for unconstrained optimization over the Koopman embedding and operator jointly while enforcing stability of the model, via a direct parameterization of stable linear systems, greatly simplifying the computations involved. We validate our method on a simulated system and analyze the advantages of our parameterization compared to alternatives."
                },
                {
                    "title": "Modern Koopman Theory for Dynamical Systems",
                    "abstract": "The field of dynamical systems is being transformed by the mathematical tools and algorithms emerging from modern computing and data science. First-principles derivations and asymptotic reductions are giving way to data-driven approaches that formulate models in operator theoretic or probabilistic frameworks. Koopman spectral theory has emerged as a dominant perspective over the past decade, in which nonlinear dynamics are represented in terms of an infinite-dimensional linear operator acting on the space of all possible measurement functions of the system. This linear representation of nonlinear dynamics has tremendous potential to enable the prediction, estimation, and control of nonlinear systems with standard textbook methods developed for linear systems. However, obtaining finite-dimensional coordinate systems and embeddings in which the dynamics appear approximately linear remains a central open challenge. The success of Koopman analysis is due primarily to three key factors: 1) there exists rigorous theory connecting it to classical geometric approaches for dynamical systems, 2) the approach is formulated in terms of measurements, making it ideal for leveraging big-data and machine learning techniques, and 3) simple, yet powerful numerical algorithms, such as the dynamic mode decomposition (DMD), have been developed and extended to reduce Koopman theory to practice in real-world applications. In this review, we provide an overview of modern Koopman operator theory, describing recent theoretical and algorithmic developments and highlighting these methods with a diverse range of applications. We also discuss key advances and challenges in the rapidly growing field of machine learning that are likely to drive future developments and significantly transform the theoretical landscape of dynamical systems."
                },
                {
                    "title": "Kernel Embedding Based Variational Approach for Low-Dimensional Approximation of Dynamical Systems",
                    "abstract": "Abstract Transfer operators such as Perron\u2013Frobenius and Koopman operator play a key role in modeling and analysis of complex dynamical systems, which allow linear representations of nonlinear dynamics by transforming the original state variables to feature spaces. However, it remains challenging to identify the optimal low-dimensional feature mappings from data. The variational approach for Markov processes (VAMP) provides a comprehensive framework for the evaluation and optimization of feature mappings based on the variational estimation of modeling errors, but it still suffers from a flawed assumption on the transfer operator and therefore sometimes fails to capture the essential structure of system dynamics. In this paper, we develop a powerful alternative to VAMP, called kernel embedding based variational approach for dynamical systems (KVAD). By using the distance measure of functions in the kernel embedding space, KVAD effectively overcomes theoretical and practical limitations of VAMP. In addition, we develop a data-driven KVAD algorithm for seeking the ideal feature mapping within a subspace spanned by given basis functions, and numerical experiments show that the proposed algorithm can significantly improve the modeling accuracy compared to VAMP."
                },
                {
                    "title": "Learning Deep Neural Network Representations for Koopman Operators of Nonlinear Dynamical Systems",
                    "abstract": "The Koopman operator has recently garnered much attention for its value in dynamical systems analysis and data-driven model discovery. However, its application has been hindered by the computational complexity of extended dynamic mode decomposition; this requires a combinatorially large basis set to adequately describe many nonlinear systems of interest, e.g. cyber-physical infrastructure systems, biological networks, social systems, and fluid dynamics. Often the dictionaries generated for these problems are manually curated, requiring domain-specific knowledge and painstaking tuning. In this paper we introduce a computational framework for learning Koopman operators of nonlinear dynamical systems using deep learning. We show that this novel method automatically selects efficient deep dictionaries, requiring much lower dimensional dictionaries while outperforming state-of-the-art methods. We benchmark this method on partially observed nonlinear systems, including the glycolytic oscillator and show it is able to predict on test data quantitatively 100 steps into the future, using only a single timepoint as an initial condition, and quantitative oscillatory behavior 400 steps into the future."
                },
                {
                    "title": "Extended dynamic mode decomposition with dictionary learning: A data-driven adaptive spectral decomposition of the Koopman operator.",
                    "abstract": "Numerical approximation methods for the Koopman operator have advanced considerably in the last few years. In particular, data-driven approaches such as dynamic mode decomposition (DMD)51 and its generalization, the extended-DMD (EDMD), are becoming increasingly popular in practical applications. The EDMD improves upon the classical DMD by the inclusion of a flexible choice of dictionary of observables which spans a finite dimensional subspace on which the Koopman operator can be approximated. This enhances the accuracy of the solution reconstruction and broadens the applicability of the Koopman formalism. Although the convergence of the EDMD has been established, applying the method in practice requires a careful choice of the observables to improve convergence with just a finite number of terms. This is especially difficult for high dimensional and highly nonlinear systems. In this paper, we employ ideas from machine learning to improve upon the EDMD method. We develop an iterative approximation algorithm which couples the EDMD with a trainable dictionary represented by an artificial neural network. Using the Duffing oscillator and the Kuramoto Sivashinsky partical differential equation as examples, we show that our algorithm can effectively and efficiently adapt the trainable dictionary to the problem at hand to achieve good reconstruction accuracy without the need to choose a fixed dictionary a priori. Furthermore, to obtain a given accuracy, we require fewer dictionary terms than EDMD with fixed dictionaries. This alleviates an important shortcoming of the EDMD algorithm and enhances the applicability of the Koopman framework to practical problems."
                },
                {
                    "title": "Sobolev Norm Learning Rates for Regularized Least-Squares Algorithms",
                    "abstract": "Learning rates for least-squares regression are typically expressed in terms of $L_2$-norms. In this paper we extend these rates to norms stronger than the $L_2$-norm without requiring the regression function to be contained in the hypothesis space. In the special case of Sobolev reproducing kernel Hilbert spaces used as hypotheses spaces, these stronger norms coincide with fractional Sobolev norms between the used Sobolev space and $L_2$. As a consequence, not only the target function but also some of its derivatives can be estimated without changing the algorithm. From a technical point of view, we combine the well-known integral operator techniques with an embedding property, which so far has only been used in combination with empirical process arguments. This combination results in new finite sample bounds with respect to the stronger norms. From these finite sample bounds our rates easily follow. Finally, we prove the asymptotic optimality of our results in many cases."
                },
                {
                    "title": "Dynamic Mode Decomposition with Reproducing Kernels for Koopman Spectral Analysis",
                    "abstract": "A spectral analysis of the Koopman operator, which is an infinite dimensional linear operator on an observable, gives a (modal) description of the global behavior of a nonlinear dynamical system without any explicit prior knowledge of its governing equations. In this paper, we consider a spectral analysis of the Koopman operator in a reproducing kernel Hilbert space (RKHS). We propose a modal decomposition algorithm to perform the analysis using finite-length data sequences generated from a nonlinear system. The algorithm is in essence reduced to the calculation of a set of orthogonal bases for the Krylov matrix in RKHS and the eigendecomposition of the projection of the Koopman operator onto the subspace spanned by the bases. The algorithm returns a decomposition of the dynamics into a finite number of modes, and thus it can be thought of as a feature extraction procedure for a nonlinear dynamical system. Therefore, we further consider applications in machine learning using extracted features with the presented analysis. We illustrate the method on the applications using synthetic and real-world data."
                },
                {
                    "title": "A kernel-based method for data-driven koopman spectral analysis",
                    "abstract": "A data-driven, kernel-based method for approximating the leading Koopman \neigenvalues, eigenfunctions, and modes in problems with high-dimensional state \nspaces is presented. \n \nThis approach uses a set of scalar observables (functions that map a state to a scalar value) that are defined implicitly by the feature map associated with a user-defined kernel function. \n \nThis circumvents the computational issues that arise due to the number of \nfunctions required to span a ``sufficiently rich'' subspace of all possible scalar observables in such applications. \n \nWe illustrate this method on two examples: the first is the FitzHugh-Nagumo PDE, \n a prototypical one-dimensional reaction-diffusion \nsystem, and the second is a set of vorticity data computed from experimentally obtained \nvelocity data from flow past a cylinder at Reynolds number 413. \n \nIn both examples, we use the output of Dynamic Mode Decomposition, which has a similar computational cost, as the benchmark for our approach."
                },
                {
                    "title": "Deep Canonical Correlation Analysis",
                    "abstract": "We introduce Deep Canonical Correlation Analysis (DCCA), a method to learn complex nonlinear transformations of two views of data such that the resulting representations are highly linearly correlated. Parameters of both transformations are jointly learned to maximize the (regularized) total correlation. It can be viewed as a nonlinear extension of the linear method canonical correlation analysis (CCA). It is an alternative to the nonparametric method kernel canonical correlation analysis (KCCA) for learning correlated nonlinear transformations. Unlike KCCA, DCCA does not require an inner product, and has the advantages of a parametric method: training time scales well with data size and the training data need not be referenced when computing the representations of unseen instances. In experiments on two real-world datasets, we find that DCCA learns representations with significantly higher correlation than those learned by CCA and KCCA. We also introduce a novel non-saturating sigmoid function based on the cube root that may be useful more generally in feedforward neural networks."
                },
                {
                    "title": "Statistical Methods for Stochastic Differential Equations",
                    "abstract": "Estimating functions for diffusion-type processes, Michael Sorensen Introduction Low frequency asymptotics Martingale estimating functions The likelihood function Non-martingale estimating functions High-frequency asymptotics High-frequency asymptotics in a fixed time-interval Small-diffusion asymptotics Non-Markovian models General asymptotic results for estimating functions Optimal estimating functions: General theory The econometrics of high frequency data, Per. A. Mykland and Lan Zhang Introduction Time varying drift and volatility Behavior of estimators: Variance Asymptotic normality Microstructure Methods based on contiguity Irregularly spaced data Statistics and high frequency data, Jean Jacod Introduction What can be estimated? Wiener plus compound Poisson processes Auxiliary limit theorems A first LNN (Law of Large Numbers) Some other LNNs A first CLT CLT with discontinuous limits Estimation of the integrated volatility Testing for jumps Testing for common jumps The Blumenthal-Getoor index Importance sampling techniques for estimation of diffusion models, Omiros Papaspiliopoulos and Gareth Roberts Overview of the chapter Background IS estimators based on bridge processes IS estimators based on guided processes Unbiased Monte Carlo for diffusions Appendix: Typical problems of the projection-simulation paradigm in MC for diffusions Appendix: Gaussian change of measure Non parametric estimation of the coefficients of ergodic diffusion processes based on high frequency data, Fabienne Comte, Valentine Genon-Catalot, and Yves Rozenholc Introduction Model and assumptions Observations and asymptotic framework Estimation method Drift estimation Diffusion coefficient estimation Examples and practical implementation Bibliographical remarks Appendix. Proof of Proposition.13 Ornstein-Uhlenbeck related models driven by Levy processes, Peter J. Brockwell and Alexander Lindner Introduction Levy processes Ornstein-Uhlenbeck related models Some estimation methods Parameter estimation for multiscale diffusions: an overview, Grigorios A. Pavliotis, Yvo Pokern, and Andrew M. Stuart Introduction Illustrative examples Averaging and homogenization Subsampling Hypoelliptic diffusions Nonparametric drift estimation Conclusions and further work"
                },
                {
                    "title": "Kernel Methods for the Approximation of Nonlinear Systems",
                    "abstract": "We introduce a data-driven order reduction method for nonlinear control systems, drawing on recent progress in machine learning and statistical dimensionality reduction. The method rests on the assumption that the nonlinear system behaves linearly when lifted into a high (or infinite) dimensional feature space where balanced truncation may be carried out implicitly. This leads to a nonlinear reduction map which can be combined with a representation of the system belonging to a reproducing kernel Hilbert space to give a closed, reduced order dynamical system which captures the essential input-output characteristics of the original model. Empirical simulations illustrating the approach are also provided."
                },
                {
                    "title": "Statistical Mechanics: Theory and Molecular Simulation",
                    "abstract": "1. Introduction 2. Classical Mechanics 3. Theoretical Foundations of Classical Statistical Mechanics 4. The Microcanonical Ensemble and Introduction to Molecular Dynamics 5. The Canonical Ensemble 6. The Isobaric Ensembles 7. The Grand Canonical Ensemble 8. Monte Carlo Methods in Statistical Mechanics 9. Free Energy Calculations 10. Quantum Mechanics 11. Quantum Ensembles and the Density Matrix 12. Quantum Ideal Gases: Fermi-Dirac and Bose-Einstein Statistics 13. The Feynman Path Integral 14. Classical Time-Dependent Statistical Mechanics and Systems Away from Equilibrium 15. Quantum Time-Dependent Statistical Mechanics 16. The Generalized Langevin Equation 17. Advanced Sampling Approaches 18. Critical Phenomena 19. Conclusions and Perspectives"
                },
                {
                    "title": "Support vector machines",
                    "abstract": "Support vector machines (SVMs) are a family of machine learning methods, originally introduced for the problem of classification and later generalized to various other situations. They are based on principles of statistical learning theory and convex optimization, and are currently used in various domains of application, including bioinformatics, text categorization, and computer vision. Copyright \u00a9 2009 John Wiley & Sons, Inc."
                },
                {
                    "title": "Handbook of Linear Algebra",
                    "abstract": "Linear Algebra Linear Algebra Vectors, Matrices, and Systems of Linear Equations Jane Day Linear Independence, Span, and Bases Mark Mills Linear Transformations Francesco Barioli Determinants and Eigenvalues Luz M. DeAlba Inner Product Spaces, Orthogonal Projection, Least Squares, and Singular Value Decomposition Lixing Han and Michael Neumann Canonical Forms Leslie Hogben Other Canonical Forms Roger A. Horn and Vladimir V. Sergeichuk Unitary Similarity, Normal Matrices, and Spectral Theory Helene Shapiro Hermitian and Positive Definite Matrices Wayne Barrett Nonnegative and Stochastic Matrices Uriel G. Rothblum Partitioned Matrices Robert Reams Topics in Linear Algebra Schur Complements Roger A. Horn and Fuzhen Zhang Quadratic, Bilinear, and Sesquilinear Forms Raphael Loewy Multilinear Algebra J.A. Dias da Silva and Armando Machado Tensors and Hypermatrices Lek-Heng Lim Matrix Equalities and Inequalities Michael Tsatsomeros Functions of Matrices Nicholas J. Higham Matrix Polynomials Jorg Liesen and Christian Mehl Matrix Equations Beatrice Meini Invariant Subspaces G.W. Stewart Matrix Perturbation Theory Ren-Cang Li Special Types of Matrices Albrecht Bottcher and Ilya Spitkovsky Pseudospectra Mark Embree Singular Values and Singular Value Inequalities Roy Mathias Numerical Range Chi-Kwong Li Matrix Stability and Inertia Daniel Hershkowitz Generalized Inverses of Matrices Yimin Wei Inverse Eigenvalue Problems Alberto Borobia Totally Positive and Totally Nonnegative Matrices Shaun M. Fallat Linear Preserver Problems Peter Semrl Matrices over Finite Fields J. D. Botha Matrices over Integral Domains Shmuel Friedland Similarity of Families of Matrices Shmuel Friedland Representations of Quivers and Mixed Graphs Roger A. Horn and Vladimir V. Sergeichuk Max-Plus Algebra Marianne Akian, Ravindra Bapat, and Stephane Gaubert Matrices Leaving a Cone Invariant Bit-Shun Tam and Hans Schneider Spectral Sets Catalin Badea and Bernhard Beckermann Combinatorial Matrix Theory and Graphs Combinatorial Matrix Theory Combinatorial Matrix Theory Richard A. Brualdi Matrices and Graphs Willem H. Haemers Digraphs and Matrices Jeffrey L. Stuart Bipartite Graphs and Matrices Bryan L. Shader Sign Pattern Matrices Frank J. Hall and Zhongshan Li Topics in Combinatorial Matrix Theory Permanents Ian M. Wanless D-Optimal Matrices Michael G. Neubauer and William Watkins Tournaments T.S. Michael Minimum Rank, Maximum Nullity, and Zero Forcing Number of Graphs Shaun M. Fallat and Leslie Hogben Spectral Graph Theory Steve Butler and Fan Chung Algebraic Connectivity Steve Kirkland Matrix Completion Problems Luz M. DeAlba, Leslie Hogben, and Amy Wangsness Wehe Numerical Methods Numerical Methods for Linear Systems Vector and Matrix Norms, Error Analysis, Efficiency, and Stability Ralph Byers and Biswa Nath Datta Matrix Factorizations and Direct Solution of Linear Systems Christopher Beattie Least Squares Solution of Linear Systems Per Christian Hansen and Hans Bruun Nielsen Sparse Matrix Methods Esmond G. Ng Iterative Solution Methods for Linear Systems Anne Greenbaum Numerical Methods for Eigenvalues Symmetric Matrix Eigenvalue Techniques Ivan Slapnicar Unsymmetric Matrix Eigenvalue Techniques David S. Watkins The Implicitly Restarted Arnoldi Method D.C. Sorensen Computation of the Singular Value Decomposition Alan Kaylor Cline and Inderjit S. Dhillon Computing Eigenvalues and Singular Values to High Relative Accuracy Zlatko Drmac Nonlinear Eigenvalue Problems Heinrich Voss Topics in Numerical Linear Algebra Fast Matrix Multiplication Dario A. Bini Fast Algorithms for Structured Matrix Computations Michael Stewart Structured Eigenvalue Problems | Structure-Preserving Algorithms, Structured Error Analysis Heike Fassbender Large-Scale Matrix Computations Roland W. Freund Linear Algebra in Other Disciplines Applications to Physical and Biological Sciences Linear Algebra and Mathematical Physics Lorenzo Sadun Linear Algebra in Biomolecular Modeling Zhijun Wu Linear Algebra in Mathematical Population Biology and Epidemiology Fred Brauer and Carlos Castillo-Chavez Applications to Optimization Linear Programming Leonid N. Vaserstein Semidefinite Programming Henry Wolkowicz Applications to Probability and Statistics Random Vectors and Linear Statistical Models Simo Puntanen and George P.H. Styan Multivariate Statistical Analysis Simo Puntanen, George A.F. Seber, and George P.H. Styan Markov Chains Beatrice Meini Applications to Computer Science Coding Theory Joachim Rosenthal and Paul Weiner Quantum Computation Zijian Diao Operator Quantum Error Correction Chi-Kwong Li, Yiu-Tung Poon, and Nung-Sing Sze Information Retrieval and Web Search Amy N. Langville and Carl D. Meyer Signal Processing Michael Stewart Applications to Analysis Differential Equations and Stability Volker Mehrmann and Tatjana Stykel Dynamical Systems and Linear Algebra Fritz Colonius and Wolfgang Kliemann Control Theory Peter Benner Fourier Analysis Kenneth Howell Applications to Geometry Geometry Mark Hunacek Some Applications of Matrices and Graphs in Euclidean Geometry Miroslav Fiedler Applications to Algebra Matrix Groups Peter J. Cameron Group Representations Randall R. Holmes and Tin-Yau Tam Nonassociative Algebras Murray R. Bremner, Lucia I. Murakami, and Ivan P. Shestakov Lie Algebras Robert Wilson Computational Software Interactive Software for Linear Algebra MATLAB Steven J. Leon Linear Algebra in Maple David J. Jeffrey and Robert M. Corless Mathematica Heikki Ruskeep'a'a Sage Robert A. Beezer, Robert Bradshaw, Jason Grout, and William Stein Packages of Subroutines for Linear Algebra BLAS Jack Dongarra, Victor Eijkhout, and Julien Langou LAPACK Zhaojun Bai, James Demmel, Jack Dongarra, Julien Langou, and Jenny Wang Use of ARPACK and EIGS D.C. Sorensen Summary of Software for Linear Algebra Freely Available on the Web Jack Dongarra, Victor Eijkhout, and Julien Langou Glossary Notation Index Index"
                },
                {
                    "title": "Statistical Inference for Ergodic Diffusion Processes",
                    "abstract": "L\u00e9vy Processes\u201d (eight papers), \u201cIII. Empirical Processes\u201d (four papers), and \u201cIV. Stochastic Differential Equations\u201d (four papers). Here are some comments about the individual papers: In I.2 (paper 2 of Part I) the covariance representation method, which relies on Clark\u2019s formula on path spaces, is used to obtain concentration inequalities for functionals of Brownian motion on a manifold, allowing one to obtain tail estimates for this Brownian motion. In I.4 a transportation inequality for the canonical Gaussian measure in R is obtained and applied to Khintchine\u2013Kahane inequalities for norms of random series with nonsymmetric Bernoulli coefficients. In II.1 exponential inequalities for U -statistics of order two are presented; these rely upon the Talagrand inequality for empirical processes but also use martingale type inequalities. In II.2 the unconditional convergence of a Gaussian [and, more generally, independent, identically distributed (iid)] series in a Banach space is studied. Applications to Karhunen\u2013Love representations of Gaussian processes are given. In II.3 estimates of tail properties and moments of multidimensional chaos generated by positive random variables with log concave tails are given. In II.4 a quantitative technique for studying the asymptotic distribution of sequences of Markov processes in infinite dimensions is proposed. The proof relies on the properties of an associated sequence of exponential martingales. In II.5 it is shown that a moving average process driven by a symmetric L\u00e9vy process and with a kernel with finite total 2-variation admits an almost surely bounded version. In II.6 a Markovian approach to the entropic convergence in the central limit theorem is presented. The emphasis is on the speed of convergence, as well as relaxing a spectral gap assumption. In II.7 a new version of the Khintchine\u2013Kahane inequality for general Bernoulli random variables is presented with the help of hypercontractive methods. In III.1 necessary and sufficient conditions for the moderate deviations of empirical processes and sums of iid random vectors on a separable Banach space are given. In III.2 exponential concentration inequalities for subadditive functions of independent random variables are obtained. As a consequence, Talagrand\u2019s inequality for empirical processes is refined thanks to further developments of the entropy method introduced by M. Ledoux. In III.3 ratio limit theorems for empirical processes are obtained with the help of concentration inequalities. In III.4 asymptotic distributions of trimmed Wasserstein distances between the true and the empirical distribution function are obtained via weighted approximation results for uniform empirical processes. In IV.1 sharp rates of convergence for splitting-up approximations of stochastic partial differential equations are obtained. The error is estimated in terms of Sobolev\u2019s norm. In IV.4 the existence and uniqueness of a strong solution for a stochastic differential equation driven by a fractional Brownian motion with Hurst index H < 1/2 and with a possibly time-dependent drift which satisfies a suitable integrability condition is obtained. In short, the book presents a wide variety of inequalities which are established for either stochastic processes or sequences of random variables. In 2006 this is still an active domain of research in transportation problems."
                },
                {
                    "title": "On the Convergence of Eigenspaces in Kernel Principal Component Analysis",
                    "abstract": "This paper presents a non-asymptotic statistical analysis of Kernel-PCA with a focus different from the one proposed in previous work on this topic. Here instead of considering the reconstruction error of KPCA we are interested in approximation error bounds for the eigenspaces themselves. We prove an upper bound depending on the spacing between eigenvalues but not on the dimensionality of the eigenspace. As a consequence this allows to infer stability results for these estimated spaces."
                },
                {
                    "title": "Mathematical control theory - an introduction",
                    "abstract": "Preface.- Introduction.- Part I. Elements of classical control theory.- Controllability and observability.- Stability and stabilizability.- Realization theory.- Systems with constraints.- Part II. Nonlinear control systems.- Controllability and observability of nonlinear systems.- Stability and stabilizability.- Realization theory.- Part III. Optimal control.- Dynamic programming.- Dynamic programming for impulse control.- The maximum principle.- The existence of optimal strategies.- Part IV. Infinite dimensional linear systems.- Linear control systems.- Controllability.- Stability and stabilizability.- Linear regulators in Hilbert spaces.- Appendix.- Metric spaces.- Banach spaces.- Hilbert spaces.- Bochner's integral.- Spaces of continuous functions.- Spaces of measurable functions.- References.- Notations.- Index"
                },
                {
                    "title": "Time's Arrow: The Origins of Thermodynamic Behavior",
                    "abstract": "The Second Law of Thermodynamics has been called the most important law of nature: It is the law that gives a direction to processes that is not inherent in the laws of motion, that says the state of the universe is driven to thermal equilibrium. Its mathematical formulation is simple: The entropy of a closed system cannot decrease. Since the recognition that macroscopic phenomena have an atomic basis, it has remained a fundamental problem to reconcile the increase of entropy with the known reversibility of all the laws of microscopic physics. Professor Michael Mackey of McGill University here explores the dynamical basis for the Second Law, that is, he seeks to illuminate the fundamental dynamical properties required for the construction of a successful statistical mechanics. Aimed at physicists and applied mathematicians with an interest in the foundations of statistical mechanics, the book includes such new material as: a demonstration that the black body radiation law can be deduced from maximal entropy principles; a discussion of sufficient conditions for the existence of at least one state of thermodynamic equilibrium; a description of the behavior of entropy in asymptotically periodic systems; a necessary and sufficient condition for the evolution of entropy to a global maximum; and a presen- tation of the three main types of ergodic theorems and their proofs. He also explores the potential role of incomplete knowledge of dynamical variables, measurement imprecision, and the effects of noise in giving rise to entropy increases."
                },
                {
                    "title": "Stochastic differential equations : an introduction with applications",
                    "abstract": "Some Mathematical Preliminaries.- Ito Integrals.- The Ito Formula and the Martingale Representation Theorem.- Stochastic Differential Equations.- The Filtering Problem.- Diffusions: Basic Properties.- Other Topics in Diffusion Theory.- Applications to Boundary Value Problems.- Application to Optimal Stopping.- Application to Stochastic Control.- Application to Mathematical Finance."
                },
                {
                    "title": "The Rotation of Eigenvectors by a Perturbation. III",
                    "abstract": "When a Hermitian linear operator is slightly perturbed, by how much can its invariant subspaces change? Given some approximations to a cluster of neighboring eigenvalues and to the corresponding eigenvectors of a real symmetric matrix, and given an estimate for the gap that separates the cluster from all other eigenvalues, how much can the subspace spanned by the eigenvectors differ from the subspace spanned by our approximations? These questions are closely related; both are investigated here. The difference between the two subspaces is characterized in terms of certain angles through which one subspace must be rotated in order most directly to reach the other. These angles unify the treatment of natural geometric, operator-theoretic and error-analytic questions concerning those subspaces. Sharp bounds upon trigonometric functions of these angles are obtained from the gap and from bounds upon either the perturbation (1st question) or a computable residual (2nd question). An example is included."
                },
                {
                    "title": "Sparse Learning of Dynamical Systems in RKHS: An Operator-Theoretic Approach",
                    "abstract": "Transfer operators provide a rich framework for representing the dynamics of very general, non-linear dynamical systems. When interacting with reproducing kernel Hilbert spaces (RKHS), descriptions of dynamics often incur prohibitive data storage requirements, motivating dataset sparsifi-cation as a precursory step to computation. Further, in practice, data is available in the form of trajectories, introducing correlation between samples. In this work, we present a method for sparse learning of transfer operators from \u03b2 - mixing stochastic processes, in both discrete and continuous time, and provide sample complexity analysis extending existing theoretical guarantees for learning from non-sparse, i.i.d. data. In addressing continuous-time settings, we develop precise descriptions using covariance-type operators for the infinitesimal generator that aids in the sample complexity analysis. We empirically illustrate the efficacy of our sparse embedding approach through deterministic and stochastic non-linear system examples."
                },
                {
                    "title": "Functional Analysis",
                    "abstract": "Theorem 1.5. Let X be countable and A \u2286 X be infinite, then A is countable. Proof. Let\u2019s see first that every infinite set of N is countable. Let A \u2286 N be infinite. In particular, A \u0338= \u2205. Since A \u2286 N is not empty, by the well ordering principle, there exist a0 \u2208 A such that a0 = min(A). Now, consider A \\ {a0} and note that, since A is infinite, A \\ {a0} is not empty, then, we can take a1 = min(A \\ {a0}). Note that since A is infinite, we can recursively get a0, a1, a2, . . . , an, for all n \u2208 N. This procedure does not over, as A is inifnite. Now, note that f : N \u2192 A, f(n) := an is a bijection. Now, letX be countable and A \u2286 X be infinite. SinceX is countable, there is f : X \u2192 N bijective. Consider the set f(A) := {f(a) : a \u2208 A} and note that since f is bijective, f(A) is infinite. Then, f(A) is countable, but since the function g : A \u2192 f(A), g(a) := f(a) is a bijection, A itself is countable. \u25a1"
                },
                {
                    "title": "KoopmanizingFlows: Diffeomorphically Learning Stable Koopman Operators",
                    "abstract": "We propose a novel framework for constructing linear time-invariant (LTI) models for data-driven representations of the Koopman operator for a class of stable nonlinear dynamics. The Koopman operator (generator) lifts a \ufb01nite-dimensional nonlinear system to a possibly in\ufb01nite-dimensional linear feature space. To utilize it for modeling, one needs to discover \ufb01nite-dimensional representations of the Koopman operator. Learning suitable features is challenging, as one needs to learn LTI features that are both Koopman-invariant (evolve linearly under the dynamics) as well as relevant (spanning the original state) - a generally unsupervised learning task. For a theoretically well-founded solution to this problem, we propose learning Koopman-invariant coordinates by composing a di\ufb00eomorphic learner with a lifted aggregate system of a latent linear model. Using an unconstrained parameterization of stable matrices along with the aforementioned feature construction, we learn the Koopman operator features without assuming a prede\ufb01ned library of functions or knowing the spectrum, while ensuring stability regardless of the operator approximation accuracy. We demonstrate the superior e\ufb03cacy of the proposed method in comparison to a state-of-the-art method on the well-known LASA handwriting dataset."
                },
                {
                    "title": "Dynamic Mode Decomposition \u1107\u1161\u11bc\u1107\u1165\u11b8\u110b\u1174 \u110c\u1165\u11a8\u110b\u116d\u11bc",
                    "abstract": "Dynamic mode decomposition (DMD) method was applied for the further study of periodical characteristics of the unsteady shock-induced combustion. The case of Lehr\u2019s experiments was numerically simulated using 4 levels of grids. FFT result reveals that almost all the grid systems oscillate at frequencies around 430-435 kHz and the measureed one is around 425 kHz. To identify more resonant modes with low frequencies, DMD method is adopted for 4 grid systems. Several major frequencies are extracted and their damping coefficients are calculated at the same time, which is a quantification parameter for combustion stabilization."
                },
                {
                    "title": "Spectra and pseudospectra : the behavior of nonnormal matrices and operators",
                    "abstract": "spectra and pseudospectra springerlink. spectra and pseudospectra the behavior of nonnormal. spectra and pseudospectra the behavior of nonnormal. spectra and pseudospectra the behavior of nonnormal. on n ? pseudospectra of operators on banach spaces. phd course on pseudospectra aalb universitet. nonhermitian systems and pseudospectra. spectra and pseudospectra the behavior of nonnormal. spectra and pseudospectra the behavior of nonnormal. pseudospectrum scholarpedia. spectra and pseudospectra the behavior of nonnormal. spectra and pseudospectra the behavior of nonnormal. customer reviews spectra and pseudospectra. spectra and pseudospectra request pdf. an introduction to pseudo spectra and non normal matrices. spectra and pseudospectra gbv. pseudospectrum. spectra and pseudospectra the behavior of nonnormal. ???? spectra and pseudospectra the behavior of nonnormal. spectra and pseudospectra the behavior of nonnormal. mark embree lloyd n trefethen abebooks. spectra and pseudospectra the behavior of nonnormal. eigtool a graphical tool for nonsymmetric eigenproblems. spectra and pseudospectra the behavior of nonnormal. universality of non normality in real arxiv vanity. spectra and pseudospectra lloyd n trefethen mark embree. spectra and pseudospectra the behavior of nonnormal. pseudospectra and nonnormal dynamical systems. booksavages. review of spectra and pseudospectra the behavior of. spectra pseudospectra and localization for random. spectra and pseudospectra of block toeplitz matrices. lecture notes on spectra and pseudospectra of matrices and. structure and dynamical behavior of non normal networks. lecture 2 nonnormality and pseudospectra. spectra and pseudospectra the behavior of nonnormal. numerical range for some plex upper triangular matrices. spectra and pseudospectra the behavior of nonnormal. spectra and pseudospectra the behavior of nonnormal. mark embree virginia tech. spectra and pseudospectra princeton university press. bookask?????? ??????????. spectra and pseudospectra the behavior of nonnormal. pseudospectra and inverse pseudospectra springerlink. review of spectra and pseudospectra the behavior of. nick trefethen. pseudospectrum mathematical garden"
                },
                {
                    "title": "Local quantum physics : fields, particles, algebras",
                    "abstract": "This textbook gives a comprehensive account of local quantum physics understood as the synthesis of quantum theory with the principle of locality. Centered on the algebraic approach, it describes both the physical concepts and the mathematical structures and their consequences. These include the emergence of the particle picture, general collision theory covering the cases of massless particles and infraparticles, the analysis of possible charge structures and exchange symmetries including braid group statistics. Thermal states of an unbounded medium and local equilibrium are discussed in detail. The author describes both the ideas and to give a critical assessment of future perspectives."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG",
                "math.PR",
                "62M15"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively learn the Infinitesimal Generator (IG) of stochastic differential equations (SDEs) from data, particularly in the presence of uneven sampling and incorporating prior knowledge of the SDE?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the understanding of complex dynamical systems across various fields, including finance and atomistic simulations. By accurately learning the IG, researchers can gain deeper insights into system dynamics, such as metastable states and long-term behavior, which are not captured by traditional methods focused solely on drift and diffusion coefficients. This advancement could lead to improved modeling techniques, more robust predictive capabilities, and practical applications in areas like risk assessment in finance and material science.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the nature of the IG as an unbounded operator, which complicates the application of existing statistical theories that work well for bounded operators like Transfer Operators (TO). Naive approaches may fail because they do not account for the complexities of unbounded operators and the need for high-quality, evenly sampled data. Additionally, integrating prior knowledge of the SDE into the learning process while handling unevenly sampled data presents significant technical and theoretical obstacles.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either learning TOs, which require evenly sampled data, or using deep learning methods that lack statistical guarantees. The limitations of existing statistical theories when applied to unbounded operators have hindered progress in this area. Moreover, the lack of methods that can effectively combine data-driven learning with prior knowledge of SDEs has created a gap in the literature. Our approach aims to bridge this gap by directly addressing the challenges associated with learning the IG from unevenly sampled data.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a novel framework for learning the IG directly from data, utilizing techniques that accommodate uneven sampling and incorporate prior knowledge of the SDE. We will employ a combination of advanced statistical methods and machine learning algorithms, applying them to a diverse set of datasets representing various dynamical systems. The expected outcomes include a robust estimation of the IG, improved understanding of system dynamics, and the ability to predict long-term behavior in complex systems, ultimately providing a comprehensive tool for researchers in the field."
            }
        },
        "author_data": {
            "c19f6a2a-e473-4628-ad39-0676cc56f056": {
                "pk": "c19f6a2a-e473-4628-ad39-0676cc56f056",
                "name": "Vladimir R. Kostic",
                "collaborators": [
                    "Pietro Novelli",
                    "Massimiliano Pontil",
                    "Karim Lounici",
                    "Riccardo Grazzi",
                    "Gregoire Pacreau",
                    "Giacomo Turri",
                    "H\u00e9l\u00e8ne Halconruy",
                    "Timoth\u00e9e Devergne",
                    "Giacomo Meanti",
                    "Antoine Chatalic"
                ],
                "domain": [
                    "Stochastic Dynamical Systems",
                    "Machine Learning",
                    "Koopman Operators",
                    "Conditional Probability"
                ],
                "publications": [
                    {
                        "title": "Learning invariant representations of time-homogeneous stochastic dynamical systems",
                        "abstract": "We consider the general class of time-homogeneous stochastic dynamical systems, both discrete and continuous, and study the problem of learning a representation of the state that faithfully captures its dynamics. This is instrumental to learning the transfer operator or the generator of the system, which in turn can be used for numerous tasks, such as forecasting and interpreting the system dynamics. We show that the search for a good representation can be cast as an optimization problem over neural networks. Our approach is supported by recent results in statistical learning theory, highlighting the role of approximation error and metric distortion in the learning problem. The objective function we propose is associated with projection operators from the representation space to the data space, overcomes metric distortion, and can be empirically estimated from data. In the discrete-time setting, we further derive a relaxed objective function that is differentiable and numerically well-conditioned. We compare our method against state-of-the-art approaches on different datasets, showing better performance across the board."
                    },
                    {
                        "title": "Neural Conditional Probability for Inference",
                        "abstract": "We introduce NCP (Neural Conditional Probability), a novel operator-theoretic approach for learning conditional distributions with a particular focus on inference tasks. NCP can be used to build conditional confidence regions and extract important statistics like conditional quantiles, mean, and covariance. It offers streamlined learning through a single unconditional training phase, facilitating efficient inference without the need for retraining even when conditioning changes. By tapping into the powerful approximation capabilities of neural networks, our method efficiently handles a wide variety of complex probability distributions, effectively dealing with nonlinear relationships between input and output variables. Theoretical guarantees ensure both optimization consistency and statistical accuracy of the NCP method. Our experiments show that our approach matches or beats leading methods using a simple Multi-Layer Perceptron (MLP) with two hidden layers and GELU activations. This demonstrates that a minimalistic architecture with a theoretically grounded loss function can achieve competitive results without sacrificing performance, even in the face of more complex architectures."
                    },
                    {
                        "title": "Laplace Transform Based Low-Complexity Learning of Continuous Markov Semigroups",
                        "abstract": "Markov processes serve as a universal model for many real-world random processes. This paper presents a data-driven approach for learning these models through the spectral decomposition of the infinitesimal generator (IG) of the Markov semigroup. The unbounded nature of IGs complicates traditional methods such as vector-valued regression and Hilbert-Schmidt operator analysis. Existing techniques, including physics-informed kernel regression, are computationally expensive and limited in scope, with no recovery guarantees for transfer operator methods when the time-lag is small. We propose a novel method that leverages the IG's resolvent, characterized by the Laplace transform of transfer operators. This approach is robust to time-lag variations, ensuring accurate eigenvalue learning even for small time-lags. Our statistical analysis applies to a broader class of Markov processes than current methods while reducing computational complexity from quadratic to linear in the state dimension. Finally, we illustrate the behaviour of our method in two experiments."
                    },
                    {
                        "title": "Estimating Koopman operators with sketching to provably learn large scale dynamical systems",
                        "abstract": "The theory of Koopman operators allows to deploy non-parametric machine learning algorithms to predict and analyze complex dynamical systems. Estimators such as principal component regression (PCR) or reduced rank regression (RRR) in kernel spaces can be shown to provably learn Koopman operators from finite empirical observations of the system's time evolution. Scaling these approaches to very long trajectories is a challenge and requires introducing suitable approximations to make computations feasible. In this paper, we boost the efficiency of different kernel-based Koopman operator estimators using random projections (sketching). We derive, implement and test the new \"sketched\" estimators with extensive experiments on synthetic and large-scale molecular dynamics datasets. Further, we establish non asymptotic error bounds giving a sharp characterization of the trade-offs between statistical learning rates and computational efficiency. Our empirical and theoretical analysis shows that the proposed estimators provide a sound and efficient way to learn large scale dynamical systems. In particular our experiments indicate that the proposed estimators retain the same accuracy of PCR or RRR, while being much faster."
                    }
                ]
            },
            "ce369cfc-c8fa-4749-a8bf-b20671fb549d": {
                "pk": "ce369cfc-c8fa-4749-a8bf-b20671fb549d",
                "name": "Karim Lounici",
                "collaborators": [
                    "Vladimir Koltchinskii",
                    "Gr\u00e9goire Pacreau",
                    "Katia Meziani",
                    "R\u00e9mi Flamary",
                    "Ery Arias-Castro",
                    "Alexander B. Tsybakov",
                    "Pierre Alquier",
                    "Richard Nickl",
                    "Benjamin Riu",
                    "Alexandre B. Tsybakov"
                ],
                "domain": [
                    "Statistical Learning",
                    "High-Dimensional Statistics",
                    "Machine Learning",
                    "Covariance Estimation"
                ],
                "publications": [
                    {
                        "title": "High-dimensional covariance matrix estimation with missing observations",
                        "abstract": "In this paper, we study the problem of high-dimensional approximately low-rank covariance matrix estimation with missing observations. We propose a simple procedure computationally tractable in high-dimension and that does not require imputation of the missing data. We establish non-asymptotic sparsity oracle inequalities for the estimation of the covariance matrix with the Frobenius and spectral norms, valid for any setting of the sample size and the dimension of the observations. We further establish minimax lower bounds showing that our rates are minimax optimal up to a logarithmic factor."
                    },
                    {
                        "title": "High-dimensional stochastic optimization with the generalized Dantzig estimator",
                        "abstract": "We propose a generalized version of the Dantzig selector. We show that it satisfies sparsity oracle inequalities in prediction and estimation. We consider then the particular case of high-dimensional linear regression model selection with the Huber loss function. In this case we derive the sup-norm convergence rate and the sign concentration property of the Dantzig estimators under a mutual coherence assumption on the dictionary."
                    },
                    {
                        "title": "Sup-norm convergence rate and sign concentration property of Lasso and Dantzig estimators",
                        "abstract": "We derive the $l_{\\infty}$ convergence rate simultaneously for Lasso and Dantzig estimators in a high-dimensional linear regression model under a mutual coherence assumption on the Gram matrix of the design and two different assumptions on the noise: Gaussian noise and general noise with finite variance. Then we prove that simultaneously the thresholded Lasso and Dantzig estimators with a proper choice of the threshold enjoy a sign concentration property provided that the non-zero components of the target vector are not too small."
                    },
                    {
                        "title": "Optimal spectral norm rates for noisy low-rank matrix completion",
                        "abstract": "In this paper we consider the trace regression model where $n$ entries or linear combinations of entries of an unknown $m_1\\times m_2$ matrix $A_0$ corrupted by noise are observed. We establish for the nuclear-norm penalized estimator of $A_0$ introduced in \\cite{KLT} a general sharp oracle inequality with the spectral norm for arbitrary values of $n,m_1,m_2$ under an incoherence condition on the sampling distribution $\\Pi$ of the observed entries. Then, we apply this method to the matrix completion problem. In this case, we prove that it satisfies an optimal oracle inequality for the spectral norm, thus improving upon the only existing result \\cite{KLT} concerning the spectral norm, which assumes that the sampling distribution is uniform. Note that our result is valid, in particular, in the high-dimensional setting $m_1m_2\\gg n$. Finally we show that the obtained rate is optimal up to logarithmic factors in a minimax sense."
                    },
                    {
                        "title": "Sparse Principal Component Analysis with missing observations",
                        "abstract": "In this paper, we study the problem of sparse Principal Component Analysis (PCA) in the high-dimensional setting with missing observations. Our goal is to estimate the first principal component when we only have access to partial observations. Existing estimation techniques are usually derived for fully observed data sets and require a prior knowledge of the sparsity of the first principal component in order to achieve good statistical guarantees. Our contributions is threefold. First, we establish the first information-theoretic lower bound for the sparse PCA problem with missing observations. Second, we propose a simple procedure that does not require any prior knowledge on the sparsity of the unknown first principal component or any imputation of the missing observations, adapts to the unknown sparsity of the first principal component and achieves the optimal rate of estimation up to a logarithmic factor. Third, if the covariance matrix of interest admits a sparse first principal component and is in addition approximately low-rank, then we can derive a completely data-driven procedure computationally tractable in high-dimension, adaptive to the unknown sparsity of the first principal component and statistically optimal (up to a logarithmic factor)."
                    },
                    {
                        "title": "Variable Selection with Exponential Weights and $l_0$-Penalization",
                        "abstract": "In the context of a linear model with a sparse coefficient vector, exponential weights methods have been shown to be achieve oracle inequalities for prediction. We show that such methods also succeed at variable selection and estimation under the necessary identifiability condition on the design matrix, instead of much stronger assumptions required by other methods such as the Lasso or the Dantzig Selector. The same analysis yields consistency results for Bayesian methods and BIC-type variable selection under similar conditions."
                    },
                    {
                        "title": "Concentration Inequalities and Moment Bounds for Sample Covariance Operators",
                        "abstract": "Let $X,X_1,\\dots, X_n,\\dots$ be i.i.d. centered Gaussian random variables in a separable Banach space $E$ with covariance operator $\\Sigma:$ $$ \\Sigma:E^{\\ast}\\mapsto E,\\ \\ \\Sigma u = {\\mathbb E}\\langle X,u\\rangle, u\\in E^{\\ast}. $$ The sample covariance operator $\\hat \\Sigma:E^{\\ast}\\mapsto E$ is defined as $$ \\hat \\Sigma u := n^{-1}\\sum_{j=1}^n \\langle X_j,u\\rangle X_j, u\\in E^{\\ast}. $$ The goal of the paper is to obtain concentration inequalities and expectation bounds for the operator norm $\\|\\hat \\Sigma-\\Sigma\\|$ of the deviation of the sample covariance operator from the true covariance operator. In particular, it is shown that $$ {\\mathbb E}\\|\\hat \\Sigma-\\Sigma\\|\\asymp \\|\\Sigma\\|\\biggl(\\sqrt{\\frac{{\\bf r}(\\Sigma)}{n}}\\bigvee \\frac{{\\bf r}(\\Sigma)}{n}\\biggr), $$ where $$ {\\bf r}(\\Sigma):=\\frac{\\Bigl({\\mathbb E}\\|X\\|\\Bigr)^2}{\\|\\Sigma\\|}. $$ Moreover, under the assumption that ${\\bf r}(\\Sigma)\\lesssim n,$ it is proved that, for all $t\\geq 1,$ with probability at least $1-e^{-t}$ \\begin{align*} \\Bigl|\\|\\hat\\Sigma - \\Sigma\\|-{\\mathbb E}\\|\\hat\\Sigma - \\Sigma\\|\\Bigr| \\lesssim \\|\\Sigma\\|\\biggl(\\sqrt{\\frac{t}{n}}\\bigvee \\frac{t}{n}\\biggr). \\end{align*}"
                    },
                    {
                        "title": "Estimation of Low-Rank Covariance Function",
                        "abstract": "We consider the problem of estimating a low rank covariance function $K(t,u)$ of a Gaussian process $S(t), t\\in [0,1]$ based on $n$ i.i.d. copies of $S$ observed in a white noise. We suggest a new estimation procedure adapting simultaneously to the low rank structure and the smoothness of the covariance function. The new procedure is based on nuclear norm penalization and exhibits superior performances as compared to the sample covariance function by a polynomial factor in the sample size $n$. Other results include a minimax lower bound for estimation of low-rank covariance functions showing that our procedure is optimal as well as a scheme to estimate the unknown noise variance of the Gaussian process."
                    },
                    {
                        "title": "Pac-bayesian bounds for sparse regression estimation with exponential weights",
                        "abstract": "We consider the sparse regression model where the number of parameters $p$ is larger than the sample size $n$. The difficulty when considering high-dimensional problems is to propose estimators achieving a good compromise between statistical and computational performances. The BIC estimator for instance performs well from the statistical point of view \\cite{BTW07} but can only be computed for values of $p$ of at most a few tens. The Lasso estimator is solution of a convex minimization problem, hence computable for large value of $p$. However stringent conditions on the design are required to establish fast rates of convergence for this estimator. Dalalyan and Tsybakov \\cite{arnak} propose a method achieving a good compromise between the statistical and computational aspects of the problem. Their estimator can be computed for reasonably large $p$ and satisfies nice statistical properties under weak assumptions on the design. However, \\cite{arnak} proposes sparsity oracle inequalities in expectation for the empirical excess risk only. In this paper, we propose an aggregation procedure similar to that of \\cite{arnak} but with improved statistical performances. Our main theoretical result is a sparsity oracle inequality in probability for the true excess risk for a version of exponential weight estimator. We also propose a MCMC method to compute our estimator for reasonably large values of $p$."
                    },
                    {
                        "title": "Global uniform risk bounds for wavelet deconvolution estimators",
                        "abstract": "We consider the statistical deconvolution problem where one observes $n$ replications from the model $Y=X+\\epsilon$, where $X$ is the unobserved random signal of interest and $\\epsilon$ is an independent random error with distribution $\\phi$. Under weak assumptions on the decay of the Fourier transform of $\\phi,$ we derive upper bounds for the finite-sample sup-norm risk of wavelet deconvolution density estimators $f_n$ for the density $f$ of $X$, where $f:\\mathbb{R}\\to \\mathbb{R}$ is assumed to be bounded. We then derive lower bounds for the minimax sup-norm risk over Besov balls in this estimation problem and show that wavelet deconvolution density estimators attain these bounds. We further show that linear estimators adapt to the unknown smoothness of $f$ if the Fourier transform of $\\phi$ decays exponentially and that a corresponding result holds true for the hard thresholding wavelet estimator if $\\phi$ decays polynomially. We also analyze the case where $f$ is a \"supersmooth\"/analytic density. We finally show how our results and recent techniques from Rademacher processes can be applied to construct global confidence bands for the density $f$."
                    },
                    {
                        "title": "New asymptotic results in principal component analysis",
                        "abstract": "Let $X$ be a mean zero Gaussian random vector in a separable Hilbert space ${\\mathbb H}$ with covariance operator $\\Sigma:={\\mathbb E}(X\\otimes X).$ Let $\\Sigma=\\sum_{r\\geq 1}\\mu_r P_r$ be the spectral decomposition of $\\Sigma$ with distinct eigenvalues $\\mu_1>\\mu_2> \\dots$ and the corresponding spectral projectors $P_1, P_2, \\dots.$ Given a sample $X_1,\\dots, X_n$ of size $n$ of i.i.d. copies of $X,$ the sample covariance operator is defined as $\\hat \\Sigma_n := n^{-1}\\sum_{j=1}^n X_j\\otimes X_j.$ The main goal of principal component analysis is to estimate spectral projectors $P_1, P_2, \\dots$ by their empirical counterparts $\\hat P_1, \\hat P_2, \\dots$ properly defined in terms of spectral decomposition of the sample covariance operator $\\hat \\Sigma_n.$ The aim of this paper is to study asymptotic distributions of important statistics related to this problem, in particular, of statistic $\\|\\hat P_r-P_r\\|_2^2,$ where $\\|\\cdot\\|_2^2$ is the squared Hilbert--Schmidt norm. This is done in a \"high-complexity\" asymptotic framework in which the so called effective rank ${\\bf r}(\\Sigma):=\\frac{{\\rm tr}(\\Sigma)}{\\|\\Sigma\\|_{\\infty}}$ (${\\rm tr}(\\cdot)$ being the trace and $\\|\\cdot\\|_{\\infty}$ being the operator norm) of the true covariance $\\Sigma$ is becoming large simultaneously with the sample size $n,$ but ${\\bf r}(\\Sigma)=o(n)$ as $n\\to\\infty.$ In this setting, we prove that, in the case of one-dimensional spectral projector $P_r,$ the properly centered and normalized statistic $\\|\\hat P_r-P_r\\|_2^2$ with {\\it data-dependent} centering and normalization converges in distribution to a Cauchy type limit. The proofs of this and other related results rely on perturbation analysis and Gaussian concentration."
                    },
                    {
                        "title": "Normal approximation and concentration of spectral projectors of sample covariance",
                        "abstract": "Let $X,X_1,\\dots, X_n$ be i.i.d. Gaussian random variables in a separable Hilbert space ${\\mathbb H}$ with zero mean and covariance operator $\\Sigma={\\mathbb E}(X\\otimes X),$ and let $\\hat \\Sigma:=n^{-1}\\sum_{j=1}^n (X_j\\otimes X_j)$ be the sample (empirical) covariance operator based on $(X_1,\\dots, X_n).$ Denote by $P_r$ the spectral projector of $\\Sigma$ corresponding to its $r$-th eigenvalue $\\mu_r$ and by $\\hat P_r$ the empirical counterpart of $P_r.$ The main goal of the paper is to obtain tight bounds on $$ \\sup_{x\\in {\\mathbb R}} \\left|{\\mathbb P}\\left\\{\\frac{\\|\\hat P_r-P_r\\|_2^2-{\\mathbb E}\\|\\hat P_r-P_r\\|_2^2}{{\\rm Var}^{1/2}(\\|\\hat P_r-P_r\\|_2^2)}\\leq x\\right\\}-\\Phi(x)\\right|, $$ where $\\|\\cdot\\|_2$ denotes the Hilbert--Schmidt norm and $\\Phi$ is the standard normal distribution function. Such accuracy of normal approximation of the distribution of squared Hilbert--Schmidt error is characterized in terms of so called effective rank of $\\Sigma$ defined as ${\\bf r}(\\Sigma)=\\frac{{\\rm tr}(\\Sigma)}{\\|\\Sigma\\|_{\\infty}},$ where ${\\rm tr}(\\Sigma)$ is the trace of $\\Sigma$ and $\\|\\Sigma\\|_{\\infty}$ is its operator norm, as well as another parameter characterizing the size of ${\\rm Var}(\\|\\hat P_r-P_r\\|_2^2).$ Other results include non-asymptotic bounds and asymptotic representations for the mean squared Hilbert--Schmidt norm error ${\\mathbb E}\\|\\hat P_r-P_r\\|_2^2$ and the variance ${\\rm Var}(\\|\\hat P_r-P_r\\|_2^2),$ and concentration inequalities for $\\|\\hat P_r-P_r\\|_2^2$ around its expectation."
                    },
                    {
                        "title": "Asymptotics and Concentration Bounds for Bilinear Forms of Spectral Projectors of Sample Covariance",
                        "abstract": "Let $X,X_1,\\dots, X_n$ be i.i.d. Gaussian random variables with zero mean and covariance operator $\\Sigma={\\mathbb E}(X\\otimes X)$ taking values in a separable Hilbert space ${\\mathbb H}.$ Let $$ {\\bf r}(\\Sigma):=\\frac{{\\rm tr}(\\Sigma)}{\\|\\Sigma\\|_{\\infty}} $$ be the effective rank of $\\Sigma,$ ${\\rm tr}(\\Sigma)$ being the trace of $\\Sigma$ and $\\|\\Sigma\\|_{\\infty}$ being its operator norm. Let $$\\hat \\Sigma_n:=n^{-1}\\sum_{j=1}^n (X_j\\otimes X_j)$$ be the sample (empirical) covariance operator based on $(X_1,\\dots, X_n).$ The paper deals with a problem of estimation of spectral projectors of the covariance operator $\\Sigma$ by their empirical counterparts, the spectral projectors of $\\hat \\Sigma_n$ (empirical spectral projectors). The focus is on the problems where both the sample size $n$ and the effective rank ${\\bf r}(\\Sigma)$ are large. This framework includes and generalizes well known high-dimensional spiked covariance models. Given a spectral projector $P_r$ corresponding to an eigenvalue $\\mu_r$ of covariance operator $\\Sigma$ and its empirical counterpart $\\hat P_r,$ we derive sharp concentration bounds for bilinear forms of empirical spectral projector $\\hat P_r$ in terms of sample size $n$ and effective dimension ${\\bf r}(\\Sigma).$ Building upon these concentration bounds, we prove the asymptotic normality of bilinear forms of random operators $\\hat P_r -{\\mathbb E}\\hat P_r$ under the assumptions that $n\\to \\infty$ and ${\\bf r}(\\Sigma)=o(n).$ In a special case of eigenvalues of multiplicity one, these results are rephrased as concentration bounds and asymptotic normality for linear forms of empirical eigenvectors. Other results include bounds on the bias ${\\mathbb E}\\hat P_r-P_r$ and a method of bias reduction as well as a discussion of possible applications to statistical inference in high-dimensional principal component analysis."
                    },
                    {
                        "title": "Robust covariance estimation with missing values and cell-wise contamination",
                        "abstract": "Large datasets are often affected by cell-wise outliers in the form of missing or erroneous data. However, discarding any samples containing outliers may result in a dataset that is too small to accurately estimate the covariance matrix. Moreover, the robust procedures designed to address this problem require the invertibility of the covariance operator and thus are not effective on high-dimensional data. In this paper, we propose an unbiased estimator for the covariance in the presence of missing values that does not require any imputation step and still achieves near minimax statistical accuracy with the operator norm. We also advocate for its use in combination with cell-wise outlier detection methods to tackle cell-wise contamination in a high-dimensional and low-rank setting, where state-of-the-art methods may suffer from numerical instability and long computation times. To complement our theoretical findings, we conducted an experimental study which demonstrates the superiority of our approach over the state of the art both in low and high dimension settings."
                    },
                    {
                        "title": "Muddling Label Regularization: Deep Learning for Tabular Datasets",
                        "abstract": "Deep Learning (DL) is considered the state-of-the-art in computer vision, speech recognition and natural language processing. Until recently, it was also widely accepted that DL is irrelevant for learning tasks on tabular data, especially in the small sample regime where ensemble methods are acknowledged as the gold standard. We present a new end-to-end differentiable method to train a standard FFNN. Our method, \\textbf{Muddling labels for Regularization} (\\texttt{MLR}), penalizes memorization through the generation of uninformative labels and the application of a differentiable close-form regularization scheme on the last hidden layer during training. \\texttt{MLR} outperforms classical NN and the gold standard (GBDT, RF) for regression and classification tasks on several datasets from the UCI database and Kaggle covering a large range of sample sizes and feature to sample ratios. Researchers and practitioners can use \\texttt{MLR} on its own as an off-the-shelf \\DL{} solution or integrate it into the most advanced ML pipelines."
                    },
                    {
                        "title": "Nuclear norm penalization and optimal rates for noisy low rank matrix completion",
                        "abstract": "This paper deals with the trace regression model where $n$ entries or linear combinations of entries of an unknown $m_1\\times m_2$ matrix $A_0$ corrupted by noise are observed. We propose a new nuclear norm penalized estimator of $A_0$ and establish a general sharp oracle inequality for this estimator for arbitrary values of $n,m_1,m_2$ under the condition of isometry in expectation. Then this method is applied to the matrix completion problem. In this case, the estimator admits a simple explicit form and we prove that it satisfies oracle inequalities with faster rates of convergence than in the previous works. They are valid, in particular, in the high-dimensional setting $m_1m_2\\gg n$. We show that the obtained rates are optimal up to logarithmic factors in a minimax sense and also derive, for any fixed matrix $A_0$, a non-minimax lower bound on the rate of convergence of our estimator, which coincides with the upper bound up to a constant factor. Finally, we show that our procedure provides an exact recovery of the rank of $A_0$ with probability close to 1. We also discuss the statistical learning setting where there is no underlying model determined by $A_0$ and the aim is to find the best trace regression model approximating the data."
                    },
                    {
                        "title": "Minimax and adaptive estimation of the Wigner function in quantum homodyne tomography with noisy data",
                        "abstract": "In quantum optics, the quantum state of a light beam is represented through the Wigner function, a density on $\\mathbb R^2$ which may take negative values but must respect intrinsic positivity constraints imposed by quantum physics. In the framework of noisy quantum homodyne tomography with efficiency parameter $1/2 < \\eta \\leq 1$, we study the theoretical performance of a kernel estimator of the Wigner function. We prove that it is minimax efficient, up to a logarithmic factor in the sample size, for the $\\mathbb L_\\infty$-risk over a class of infinitely differentiable. We compute also the lower bound for the $\\mathbb L_2$-risk. We construct adaptive estimator, i.e. which does not depend on the smoothness parameters, and prove that it attains the minimax rates for the corresponding smoothness class functions. Finite sample behaviour of our adaptive procedure are explored through numerical experiments."
                    },
                    {
                        "title": "Large scale Lasso with windowed active set for convolutional spike sorting",
                        "abstract": "Spike sorting is a fundamental preprocessing step in neuroscience that is central to access simultaneous but distinct neuronal activities and therefore to better understand the animal or even human brain. But numerical complexity limits studies that require processing large scale datasets in terms of number of electrodes, neurons, spikes and length of the recorded signals. We propose in this work a novel active set algorithm aimed at solving the Lasso for a classical convolutional model. Our algorithm can be implemented efficiently on parallel architecture and has a linear complexity w.r.t. the temporal dimensionality which ensures scaling and will open the door to online spike sorting. We provide theoretical results about the complexity of the algorithm and illustrate it in numerical experiments along with results about the accuracy of the spike recovery and robustness to the regularization parameter."
                    },
                    {
                        "title": "Concentration bounds for linear Monge mapping estimation and optimal transport domain adaptation",
                        "abstract": "This article investigates the quality of the estimator of the linear Monge mapping between distributions. We provide the first concentration result on the linear mapping operator and prove a sample complexity of $n^{-1/2}$ when using empirical estimates of first and second order moments. This result is then used to derive a generalization bound for domain adaptation with optimal transport. As a consequence, this method approaches the performance of theoretical Bayes predictor under mild conditions on the covariance structure of the problem. We also discuss the computational complexity of the linear mapping estimation and show that when the source and target are stationary the mapping is a convolution that can be estimated very efficiently using fast Fourier transforms. Numerical experiments reproduce the behavior of the proven bounds on simulated and real data for mapping estimation and domain adaptation on images."
                    },
                    {
                        "title": "Multi-task Representation Learning with Stochastic Linear Bandits",
                        "abstract": "We study the problem of transfer-learning in the setting of stochastic linear bandit tasks. We consider that a low dimensional linear representation is shared across the tasks, and study the benefit of learning this representation in the multi-task learning setting. Following recent results to design stochastic bandit policies, we propose an efficient greedy policy based on trace norm regularization. It implicitly learns a low dimensional representation by encouraging the matrix formed by the task regression vectors to be of low rank. Unlike previous work in the literature, our policy does not need to know the rank of the underlying matrix. We derive an upper bound on the multi-task regret of our policy, which is, up to logarithmic factors, of order $\\sqrt{NdT(T+d)r}$, where $T$ is the number of tasks, $r$ the rank, $d$ the number of variables and $N$ the number of rounds per task. We show the benefit of our strategy compared to the baseline $Td\\sqrt{N}$ obtained by solving each task independently. We also provide a lower bound to the multi-task regret. Finally, we corroborate our theoretical findings with preliminary experiments on synthetic data."
                    }
                ]
            },
            "9a16160a-6666-46b4-a0fc-7f5252c777ad": {
                "pk": "9a16160a-6666-46b4-a0fc-7f5252c777ad",
                "name": "Helene Halconruy",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "a943b53f-ba2b-4f44-8c33-ff7c0372b5f9": {
                "pk": "a943b53f-ba2b-4f44-8c33-ff7c0372b5f9",
                "name": "Timothee Devergne",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "59b5dc21-edad-4ff9-9bae-2f51096fa485": {
                "pk": "59b5dc21-edad-4ff9-9bae-2f51096fa485",
                "name": "Massimiliano Pontil",
                "collaborators": [
                    "Andreas Maurer",
                    "Dimitris Stamos",
                    "Bernardino Romera-Paredes",
                    "Andrew M. McDonald",
                    "Andreas Argyriou",
                    "Giulia Denevi",
                    "Andreas Maurer Massimiliano Pontil",
                    "Charles Micchelli",
                    "Carlo Ciliberto",
                    "Luca Franceschi"
                ],
                "domain": [
                    "Statistical Learning",
                    "Multi-task Learning",
                    "Regularization",
                    "Transfer Learning"
                ],
                "publications": [
                    {
                        "title": "Bounds for Vector-Valued Function Estimation",
                        "abstract": "We present a framework to derive risk bounds for vector-valued learning with a broad class of feature maps and loss functions. Multi-task learning and one-vs-all multi-category learning are treated as examples. We discuss in detail vector-valued functions with one hidden layer, and demonstrate that the conditions under which shared representations are beneficial for multi- task learning are equally applicable to multi-category learning."
                    },
                    {
                        "title": "Some Hoeffding- and Bernstein-type Concentration Inequalities",
                        "abstract": "We prove concentration inequalities for functions of independent random variables {under} sub-gaussian and sub-exponential conditions. The utility of the inequalities is demonstrated by an extension of the now classical method of Rademacher complexities to Lipschitz function classes and unbounded sub-exponential distribution."
                    },
                    {
                        "title": "Empirical bounds for functions with weak interactions",
                        "abstract": "We provide sharp empirical estimates of expectation, variance and normal approximation for a class of statistics whose variation in any argument does not change too much when another argument is modified. Examples of such weak interactions are furnished by U- and V-statistics, Lipschitz L-statistics and various error functionals of L2-regularized algorithms and Gibbs algorithms."
                    },
                    {
                        "title": "Structured Sparsity and Generalization",
                        "abstract": "We present a data dependent generalization bound for a large class of regularized algorithms which implement structured sparsity constraints. The bound can be applied to standard squared-norm regularization, the Lasso, the group Lasso, some versions of the group Lasso with overlapping groups, multiple kernel learning and other regularization schemes. In all these cases competitive results are obtained. A novel feature of our bound is that it can be applied in an infinite dimensional setting such as the Lasso in a separable Hilbert space or multiple kernel learning with a countable number of kernels."
                    },
                    {
                        "title": "Excess risk bounds for multitask learning with trace norm regularization",
                        "abstract": "Trace norm regularization is a popular method of multitask learning. We give excess risk bounds with explicit dependence on the number of tasks, the number of examples per task and properties of the data distribution. The bounds are independent of the dimension of the input space, which may be infinite as in the case of reproducing kernel Hilbert spaces. A byproduct of the proof are bounds on the expected norm of sums of random positive semidefinite matrices with subexponential moments."
                    },
                    {
                        "title": "A New Convex Relaxation for Tensor Completion",
                        "abstract": "We study the problem of learning a tensor from a set of linear measurements. A prominent methodology for this problem is based on a generalization of trace norm regularization, which has been used extensively for learning low rank matrices, to the tensor setting. In this paper, we highlight some limitations of this approach and propose an alternative convex relaxation on the Euclidean ball. We then describe a technique to solve the associated regularization problem, which builds upon the alternating direction method of multipliers. Experiments on one synthetic dataset and two real datasets indicate that the proposed method improves significantly over tensor trace norm regularization in terms of estimation error, while remaining computationally tractable."
                    },
                    {
                        "title": "Uniform concentration and symmetrization for weak interactions",
                        "abstract": "The method to derive uniform bounds with Gaussian and Rademacher complexities is extended to the case where the sample average is replaced by a nonlinear statistic. Tight bounds are obtained for U-statistics, smoothened L-statistics and error functionals of l2-regularized algorithms."
                    },
                    {
                        "title": "Empirical Bernstein Bounds and Sample Variance Penalization",
                        "abstract": "We give improved constants for data dependent and variance sensitive confidence bounds, called empirical Bernstein bounds, and extend these inequalities to hold uniformly over classes of functionswhose growth function is polynomial in the sample size n. The bounds lead us to consider sample variance penalization, a novel learning method which takes into account the empirical variance of the loss function. We give conditions under which sample variance penalization is effective. In particular, we present a bound on the excess risk incurred by the method. Using this, we argue that there are situations in which the excess risk of our method is of order 1/n, while the excess risk of empirical risk minimization is of order 1/sqrt/{n}. We show some experimental results, which confirm the theory. Finally, we discuss the potential application of our results to sample compression schemes."
                    },
                    {
                        "title": "K-Dimensional Coding Schemes in Hilbert Spaces",
                        "abstract": "This paper presents a general coding method where data in a Hilbert space are represented by finite dimensional coding vectors. The method is based on empirical risk minimization within a certain class of linear operators, which map the set of coding vectors to the Hilbert space. Two results bounding the expected reconstruction error of the method are derived, which highlight the role played by the codebook and the class of linear operators. The results are specialized to some cases of practical importance, including K-means clustering, nonnegative matrix factorization and other sparse coding methods."
                    },
                    {
                        "title": "An Inequality with Applications to Structured Sparsity and Multitask Dictionary Learning",
                        "abstract": "From concentration inequalities for the suprema of Gaussian or Rademacher processes an inequality is derived. It is applied to sharpen existing and to derive novel bounds on the empirical Rademacher complexities of unit balls in various norms appearing in the context of structured sparsity and multitask dictionary learning or matrix factorization. A key role is played by the largest eigenvalue of the data covariance matrix."
                    },
                    {
                        "title": "When is there a representer theorem? Vector versus matrix regularizers",
                        "abstract": "We consider a general class of regularization methods which learn a vector of parameters on the basis of linear measurements. It is well known that if the regularizer is a nondecreasing function of the inner product then the learned vector is a linear combination of the input data. This result, known as the {\\em representer theorem}, is at the basis of kernel-based methods in machine learning. In this paper, we prove the necessity of the above condition, thereby completing the characterization of kernel methods based on regularization. We further extend our analysis to regularization methods which learn a matrix, a problem which is motivated by the application to multi-task learning. In this context, we study a more general representer theorem, which holds for a larger class of regularizers. We provide a necessary and sufficient condition for these class of matrix regularizers and highlight them with some concrete examples of practical importance. Our analysis uses basic principles from matrix theory, especially the useful notion of matrix nondecreasing function."
                    },
                    {
                        "title": "New Perspectives on $k$-Support and Cluster Norms",
                        "abstract": "We study a regularizer which is defined as a parameterized infimum of quadratics, and which we call the box-norm. We show that the k-support norm, a regularizer proposed by [Argyriou et al, 2012] for sparse vector prediction problems, belongs to this family, and the box-norm can be generated as a perturbation of the former. We derive an improved algorithm to compute the proximity operator of the squared box-norm, and we provide a method to compute the norm. We extend the norms to matrices, introducing the spectral k-support norm and spectral box-norm. We note that the spectral box-norm is essentially equivalent to the cluster norm, a multitask learning regularizer introduced by [Jacob et al. 2009a], and which in turn can be interpreted as a perturbation of the spectral k-support norm. Centering the norm is important for multitask learning and we also provide a method to use centered versions of the norms as regularizers. Numerical experiments indicate that the spectral k-support and box-norms and their centered variants provide state of the art performance in matrix completion and multitask learning problems respectively."
                    },
                    {
                        "title": "Fitting Spectral Decay with the $k$-Support Norm",
                        "abstract": "The spectral $k$-support norm enjoys good estimation properties in low rank matrix learning problems, empirically outperforming the trace norm. Its unit ball is the convex hull of rank $k$ matrices with unit Frobenius norm. In this paper we generalize the norm to the spectral $(k,p)$-support norm, whose additional parameter $p$ can be used to tailor the norm to the decay of the spectrum of the underlying model. We characterize the unit ball and we explicitly compute the norm. We further provide a conditional gradient method to solve regularization problems with the norm, and we derive an efficient algorithm to compute the Euclidean projection on the unit ball in the case $p=\\infty$. In numerical experiments, we show that allowing $p$ to vary significantly improves performance over the spectral $k$-support norm on various matrix completion benchmarks, and better captures the spectral decay of the underlying model."
                    },
                    {
                        "title": "Incremental Learning-to-Learn with Statistical Guarantees",
                        "abstract": "In learning-to-learn the goal is to infer a learning algorithm that works well on a class of tasks sampled from an unknown meta distribution. In contrast to previous work on batch learning-to-learn, we consider a scenario where tasks are presented sequentially and the algorithm needs to adapt incrementally to improve its performance on future tasks. Key to this setting is for the algorithm to rapidly incorporate new observations into the model as they arrive, without keeping them in memory. We focus on the case where the underlying algorithm is ridge regression parameterized by a positive semidefinite matrix. We propose to learn this matrix by applying a stochastic strategy to minimize the empirical error incurred by ridge regression on future tasks sampled from the meta distribution. We study the statistical properties of the proposed algorithm and prove non-asymptotic bounds on its excess transfer risk, that is, the generalization performance on new tasks from the same meta distribution. We compare our online learning-to-learn approach with a state of the art batch method, both theoretically and empirically."
                    },
                    {
                        "title": "Forward and Reverse Gradient-Based Hyperparameter Optimization",
                        "abstract": "We study two procedures (reverse-mode and forward-mode) for computing the gradient of the validation error with respect to the hyperparameters of any iterative learning algorithm such as stochastic gradient descent. These procedures mirror two methods of computing gradients for recurrent neural networks and have different trade-offs in terms of running time and space requirements. Our formulation of the reverse-mode procedure is linked to previous work by Maclaurin et al. [2015] but does not require reversible dynamics. The forward-mode procedure is suitable for real-time hyperparameter updates, which may significantly speed up hyperparameter optimization on large datasets. We present experiments on data cleaning and on learning task interactions. We also present one large-scale experiment where the use of previous gradient-based methods would be prohibitive."
                    },
                    {
                        "title": "Online Parameter-Free Learning of Multiple Low Variance Tasks",
                        "abstract": "We propose a method to learn a common bias vector for a growing sequence of low-variance tasks. Unlike state-of-the-art approaches, our method does not require tuning any hyper-parameter. Our approach is presented in the non-statistical setting and can be of two variants. The \"aggressive\" one updates the bias after each datapoint, the \"lazy\" one updates the bias only at the end of each task. We derive an across-tasks regret bound for the method. When compared to state-of-the-art approaches, the aggressive variant returns faster rates, the lazy one recovers standard rates, but with no need of tuning hyper-parameters. We then adapt the methods to the statistical setting: the aggressive variant becomes a multi-task learning method, the lazy one a meta-learning method. Experiments confirm the effectiveness of our methods in practice."
                    },
                    {
                        "title": "On Sparsity Inducing Regularization Methods for Machine Learning",
                        "abstract": "During the past years there has been an explosion of interest in learning methods based on sparsity regularization. In this paper, we discuss a general class of such methods, in which the regularizer can be expressed as the composition of a convex function $\\omega$ with a linear function. This setting includes several methods such the group Lasso, the Fused Lasso, multi-task learning and many more. We present a general approach for solving regularization problems of this kind, under the assumption that the proximity operator of the function $\\omega$ is available. Furthermore, we comment on the application of this approach to support vector machines, a technique pioneered by the groundbreaking work of Vladimir Vapnik."
                    },
                    {
                        "title": "Sparse coding for multitask and transfer learning",
                        "abstract": "We investigate the use of sparse coding and dictionary learning in the context of multitask and transfer learning. The central assumption of our learning method is that the tasks parameters are well approximated by sparse linear combinations of the atoms of a dictionary on a high or infinite dimensional space. This assumption, together with the large quantity of available data in the multitask and transfer learning settings, allows a principled choice of the dictionary. We provide bounds on the generalization error of this approach, for both settings. Numerical experiments on one synthetic and two real datasets show the advantage of our method over single task learning, a previous method based on orthogonal and dense representation of the tasks and a related method learning task grouping."
                    },
                    {
                        "title": "New Perspectives on k-Support and Cluster Norms",
                        "abstract": "The $k$-support norm is a regularizer which has been successfully applied to sparse vector prediction problems. We show that it belongs to a general class of norms which can be formulated as a parameterized infimum over quadratics. We further extend the $k$-support norm to matrices, and we observe that it is a special case of the matrix cluster norm. Using this formulation we derive an efficient algorithm to compute the proximity operator of both norms. This improves upon the standard algorithm for the $k$-support norm and allows us to apply proximal gradient methods to the cluster norm. We also describe how to solve regularization problems which employ centered versions of these norms. Finally, we apply the matrix regularizers to different matrix completion and multitask learning datasets. Our results indicate that the spectral $k$-support norm and the cluster norm give state of the art performance on these problems, significantly outperforming trace norm and elastic net penalties."
                    },
                    {
                        "title": "Regret Bounds for Lifelong Learning",
                        "abstract": "We consider the problem of transfer learning in an online setting. Different tasks are presented sequentially and processed by a within-task algorithm. We propose a lifelong learning strategy which refines the underlying data representation used by the within-task algorithm, thereby transferring information from one task to the next. We show that when the within-task algorithm comes with some regret bound, our strategy inherits this good property. Our bounds are in expectation for a general loss function, and uniform for a convex loss. We discuss applications to dictionary learning and finite set of predictors. In the latter case, we improve previous $O(1/\\sqrt{m})$ bounds to $O(1/m)$ where $m$ is the per task sample size."
                    }
                ]
            }
        }
    },
    "2405.13899": {
        "paper_data": {
            "title": "Symmetric Linear Bandits with Hidden Symmetry",
            "url": "http://arxiv.org/abs/2405.13899v1",
            "arxiv_id": "2405.13899",
            "authors": [
                "Nam Phuong Tran",
                "The Anh Ta",
                "Debmalya Mandal",
                "Long Tran-Thanh"
            ],
            "abstract": "High-dimensional linear bandits with low-dimensional structure have received considerable attention in recent studies due to their practical significance. The most common structure in the literature is sparsity. However, it may not be available in practice. Symmetry, where the reward is invariant under certain groups of transformations on the set of arms, is another important inductive bias in the high-dimensional case that covers many standard structures, including sparsity. In this work, we study high-dimensional symmetric linear bandits where the symmetry is hidden from the learner, and the correct symmetry needs to be learned in an online setting. We examine the structure of a collection of hidden symmetry and provide a method based on model selection within the collection of low-dimensional subspaces. Our algorithm achieves a regret bound of $ O(d_0^{1/3} T^{2/3} \\log(d))$, where $d$ is the ambient dimension which is potentially very large, and $d_0$ is the dimension of the true low-dimensional subspace such that $d_0 \\ll d$. With an extra assumption on well-separated models, we can further improve the regret to $ O(d_0\\sqrt{T\\log(d)} )$.",
            "introduction": "   1 Introduction  Stochastic bandit is a sequential decision-making problem in which a player, who aims to maximize his reward, selects an action at each step and receives a stochastic reward, drawn from an initially unknown distribution of the selected arm, in response. Linear stochastic bandit (LSB)\u00a0[2] is an important variant in which the expected value of the reward is a linear function of the action. It is one of the most studied bandit variants and has many practical applications [24].   Actions in LSB are specified as feature vectors in \u211ddsuperscript\u211d\ud835\udc51\\mathbb{R}^{d}blackboard_R start_POSTSUPERSCRIPT italic_d end_POSTSUPERSCRIPT for very large feature dimension d\ud835\udc51ditalic_d, with performance i.e. the resulting regret scaling with d\ud835\udc51ditalic_d. Many works have addressed this curse of dimensionality by leveraging different low-dimensional structures as inductive biases for the learner. For example, sparsity, which assumes that the reward is a sparse linear function, has been used extensively in LSB to design bandit algorithms with better performance\u00a0[3, 32, 21]. However, when the reward function lacks the structure fo sparsity (which may occur in many real-world situations), a question arises: Are there different structures in the features of LSB that we can exploit to overcome the curse of dimensionality and design bandit algorithms with better performance?    In this paper, we study the inductive bias induced by symmetry structures in LSB, which is a more general model inductive bias than sparsity, and can facilitate efficient and effective learning\u00a0[9]. Symmetry describes how, under certain transformations of the input of the problem, the outcome should either remain unchanged (invariance) or shift predictably (equivariance). In supervised learning, it has been empirically observed [15, 12] and theoretically proven [14, 6] that explicitly integrating symmetry into models leads to improved generalization error. However, in the literature on sequential decision-making, unlike sparsity, symmetry is rarely considered to date. This leads us to the following research question: Can one leverage symmetry in sequential decision-making tasks to enable effective exploration, and eventually break the curse of dimensionality?   In machine learning literature, especially supervised learning, most of studies on symmetry assume prior knowledge of symmetry structures of the tasks under consideration [27, 14]. However, in numerous practical scenarios, the learner can only access to partial knowledge of the symmetry, necessitating the incorporation of symmetry learning mechanisms into the algorithms to achieve better performance. Examples of hidden symmetry in reinforcement learning tasks can be found in robotic control [28, 4]. As an illustrative example, consider a task of controlling a four-legged ant robot. In the ideal scenario where all legs are identical, the robot\u2019s movements in space should exhibit four rotational symmetries: (0\u2218,\u200490\u2218,\u2004180\u2218,\u2004270\u2218)superscript0superscript90superscript180superscript270(0^{\\circ},\\;90^{\\circ},\\;180^{\\circ},\\;270^{\\circ})( 0 start_POSTSUPERSCRIPT \u2218 end_POSTSUPERSCRIPT , 90 start_POSTSUPERSCRIPT \u2218 end_POSTSUPERSCRIPT , 180 start_POSTSUPERSCRIPT \u2218 end_POSTSUPERSCRIPT , 270 start_POSTSUPERSCRIPT \u2218 end_POSTSUPERSCRIPT ). However, due to potentially different capacities of the legs on the right and left sides, which is unknown to the designer, the rotational symmetry is reduced to (0\u2218,\u2004180\u2218)superscript0superscript180(0^{\\circ},\\;180^{\\circ})( 0 start_POSTSUPERSCRIPT \u2218 end_POSTSUPERSCRIPT , 180 start_POSTSUPERSCRIPT \u2218 end_POSTSUPERSCRIPT ). Further examples of hidden symmetry can be also found in the literature on multi-agent reinforcement learning with a large number of agents. To avoid the curse of dimensionality, researchers often rely on the assumption of the existence of homogeneous agents [13, 30]. In the extreme case",
            "references": [
                {
                    "title": "Addressing Imperfect Symmetry: a Novel Symmetry-Learning Actor-Critic Extension",
                    "abstract": "Symmetry, a fundamental concept to understand our environment, often oversimplifies reality from a mathematical perspective. Humans are a prime example, deviating from perfect symmetry in terms of appearance and cognitive biases (e.g. having a dominant hand). Nevertheless, our brain can easily overcome these imperfections and efficiently adapt to symmetrical tasks. The driving motivation behind this work lies in capturing this ability through reinforcement learning. To this end, we introduce Adaptive Symmetry Learning (ASL) $\\unicode{x2013}$ a model-minimization actor-critic extension that addresses incomplete or inexact symmetry descriptions by adapting itself during the learning process. ASL consists of a symmetry fitting component and a modular loss function that enforces a common symmetric relation across all states while adapting to the learned policy. The performance of ASL is compared to existing symmetry-enhanced methods in a case study involving a four-legged ant model for multidirectional locomotion tasks. The results demonstrate that ASL is capable of recovering from large perturbations and generalizing knowledge to hidden symmetric states. It achieves comparable or better performance than alternative methods in most scenarios, making it a valuable approach for leveraging model symmetry while compensating for inherent perturbations."
                },
                {
                    "title": "Anytime Model Selection in Linear Bandits",
                    "abstract": "Model selection in the context of bandit optimization is a challenging problem, as it requires balancing exploration and exploitation not only for action selection, but also for model selection. One natural approach is to rely on online learning algorithms that treat different models as experts. Existing methods, however, scale poorly ($\\text{poly}M$) with the number of models $M$ in terms of their regret. Our key insight is that, for model selection in linear bandits, we can emulate full-information feedback to the online learner with a favorable bias-variance trade-off. This allows us to develop ALEXP, which has an exponentially improved ($\\log M$) dependence on $M$ for its regret. ALEXP has anytime guarantees on its regret, and neither requires knowledge of the horizon $n$, nor relies on an initial purely exploratory stage. Our approach utilizes a novel time-uniform analysis of the Lasso, establishing a new connection between online learning and high-dimensional statistics."
                },
                {
                    "title": "Context-lumpable stochastic bandits",
                    "abstract": "We consider a contextual bandit problem with $S$ contexts and $K$ actions. In each round $t=1,2,\\dots$, the learner observes a random context and chooses an action based on its past experience. The learner then observes a random reward whose mean is a function of the context and the action for the round. Under the assumption that the contexts can be lumped into $r\\le \\min\\{S,K\\}$ groups such that the mean reward for the various actions is the same for any two contexts that are in the same group, we give an algorithm that outputs an $\\epsilon$-optimal policy after using at most $\\widetilde O(r (S +K )/\\epsilon^2)$ samples with high probability and provide a matching $\\Omega(r(S+K)/\\epsilon^2)$ lower bound. In the regret minimization setting, we give an algorithm whose cumulative regret up to time $T$ is bounded by $\\widetilde O(\\sqrt{r^3(S+K)T})$. To the best of our knowledge, we are the first to show the near-optimal sample complexity in the PAC setting and $\\widetilde O(\\sqrt{{poly}(r)(S+K)T})$ minimax regret in the online setting for this problem. We also show our algorithms can be applied to more general low-rank bandits and get improved regret bounds in some scenarios."
                },
                {
                    "title": "Data-Driven Online Model Selection With Regret Guarantees",
                    "abstract": "We consider model selection for sequential decision making in stochastic environments with bandit feedback, where a meta-learner has at its disposal a pool of base learners, and decides on the fly which action to take based on the policies recommended by each base learner. Model selection is performed by regret balancing but, unlike the recent literature on this subject, we do not assume any prior knowledge about the base learners like candidate regret guarantees; instead, we uncover these quantities in a data-driven manner. The meta-learner is therefore able to leverage the realized regret incurred by each base learner for the learning environment at hand (as opposed to the expected regret), and single out the best such regret. We design two model selection algorithms operating with this more ambitious notion of regret and, besides proving model selection guarantees via regret balancing, we experimentally demonstrate the compelling practical benefits of dealing with actual regrets instead of candidate regret bounds."
                },
                {
                    "title": "Invariant Lipschitz Bandits: A Side Observation Approach",
                    "abstract": "Symmetry arises in many optimization and decision-making problems, and has attracted considerable attention from the optimization community: By utilizing the existence of such symmetries, the process of searching for optimal solutions can be improved significantly. Despite its success in (offline) optimization, the utilization of symmetries has not been well examined within the online optimization settings, especially in the bandit literature. As such, in this paper we study the invariant Lipschitz bandit setting, a subclass of the Lipschitz bandits where the reward function and the set of arms are preserved under a group of transformations. We introduce an algorithm named \\texttt{UniformMesh-N}, which naturally integrates side observations using group orbits into the \\texttt{UniformMesh} algorithm (\\cite{Kleinberg2005_UniformMesh}), which uniformly discretizes the set of arms. Using the side-observation approach, we prove an improved regret upper bound, which depends on the cardinality of the group, given that the group is finite. We also prove a matching regret's lower bound for the invariant Lipschitz bandit class (up to logarithmic factors). We hope that our work will ignite further investigation of symmetry in bandit theory and sequential decision-making theory in general."
                },
                {
                    "title": "A PAC-Bayesian Generalization Bound for Equivariant Networks",
                    "abstract": "Equivariant networks capture the inductive bias about the symmetry of the learning task by building those symmetries into the model. In this paper, we study how equivariance relates to generalization error utilizing PAC Bayesian analysis for equivariant networks, where the transformation laws of feature spaces are determined by group representations. By using perturbation analysis of equivariant networks in Fourier domain for each layer, we derive norm-based PAC-Bayesian generalization bounds. The bound characterizes the impact of group size, and multiplicity and degree of irreducible representations on the generalization error and thereby provide a guideline for selecting them. In general, the bound indicates that using larger group size in the model improves the generalization error substantiated by extensive numerical experiments."
                },
                {
                    "title": "On the Approximation of Cooperative Heterogeneous Multi-Agent Reinforcement Learning (MARL) using Mean Field Control (MFC)",
                    "abstract": "Mean field control (MFC) is an effective way to mitigate the curse of dimensionality of cooperative multi-agent reinforcement learning (MARL) problems. This work considers a collection of $N_{\\mathrm{pop}}$ heterogeneous agents that can be segregated into $K$ classes such that the $k$-th class contains $N_k$ homogeneous agents. We aim to prove approximation guarantees of the MARL problem for this heterogeneous system by its corresponding MFC problem. We consider three scenarios where the reward and transition dynamics of all agents are respectively taken to be functions of $(1)$ joint state and action distributions across all classes, $(2)$ individual distributions of each class, and $(3)$ marginal distributions of the entire population. We show that, in these cases, the $K$-class MARL problem can be approximated by MFC with errors given as $e_1=\\mathcal{O}(\\frac{\\sqrt{|\\mathcal{X}|}+\\sqrt{|\\mathcal{U}|}}{N_{\\mathrm{pop}}}\\sum_{k}\\sqrt{N_k})$, $e_2=\\mathcal{O}(\\left[\\sqrt{|\\mathcal{X}|}+\\sqrt{|\\mathcal{U}|}\\right]\\sum_{k}\\frac{1}{\\sqrt{N_k}})$ and $e_3=\\mathcal{O}\\left(\\left[\\sqrt{|\\mathcal{X}|}+\\sqrt{|\\mathcal{U}|}\\right]\\left[\\frac{A}{N_{\\mathrm{pop}}}\\sum_{k\\in[K]}\\sqrt{N_k}+\\frac{B}{\\sqrt{N_{\\mathrm{pop}}}}\\right]\\right)$, respectively, where $A, B$ are some constants and $|\\mathcal{X}|,|\\mathcal{U}|$ are the sizes of state and action spaces of each agent. Finally, we design a Natural Policy Gradient (NPG) based algorithm that, in the three cases stated above, can converge to an optimal MARL policy within $\\mathcal{O}(e_j)$ error with a sample complexity of $\\mathcal{O}(e_j^{-3})$, $j\\in\\{1,2,3\\}$, respectively."
                },
                {
                    "title": "Feature and Parameter Selection in Stochastic Linear Bandits",
                    "abstract": "We study two model selection settings in stochastic linear bandits (LB). In the first setting, which we refer to as feature selection, the expected reward of the LB problem is in the linear span of at least one of $M$ feature maps (models). In the second setting, the reward parameter of the LB problem is arbitrarily selected from $M$ models represented as (possibly) overlapping balls in $\\mathbb R^d$. However, the agent only has access to misspecified models, i.e.,~estimates of the centers and radii of the balls. We refer to this setting as parameter selection. For each setting, we develop and analyze a computationally efficient algorithm that is based on a reduction from bandits to full-information problems. This allows us to obtain regret bounds that are not worse (up to a $\\sqrt{\\log M}$ factor) than the case where the true model is known. This is the best-reported dependence on the number of models $M$ in these settings. Finally, we empirically show the effectiveness of our algorithms using synthetic and real-world experiments."
                },
                {
                    "title": "Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges",
                    "abstract": "The last decade has witnessed an experimental revolution in data science and machine learning, epitomised by deep learning methods. Indeed, many high-dimensional learning tasks previously thought to be beyond reach -- such as computer vision, playing Go, or protein folding -- are in fact feasible with appropriate computational scale. Remarkably, the essence of deep learning is built from two simple algorithmic principles: first, the notion of representation or feature learning, whereby adapted, often hierarchical, features capture the appropriate notion of regularity for each task, and second, learning by local gradient-descent type methods, typically implemented as backpropagation. While learning generic functions in high dimensions is a cursed estimation problem, most tasks of interest are not generic, and come with essential pre-defined regularities arising from the underlying low-dimensionality and structure of the physical world. This text is concerned with exposing these regularities through unified geometric principles that can be applied throughout a wide spectrum of applications. Such a 'geometric unification' endeavour, in the spirit of Felix Klein's Erlangen Program, serves a dual purpose: on one hand, it provides a common mathematical framework to study the most successful neural network architectures, such as CNNs, RNNs, GNNs, and Transformers. On the other hand, it gives a constructive procedure to incorporate prior physical knowledge into neural architectures and provide principled way to build future architectures yet to be invented."
                },
                {
                    "title": "Provably Strict Generalisation Benefit for Equivariant Models",
                    "abstract": "It is widely believed that engineering a model to be invariant/equivariant improves generalisation. Despite the growing popularity of this approach, a precise characterisation of the generalisation benefit is lacking. By considering the simplest case of linear models, this paper provides the first provably non-zero improvement in generalisation for invariant/equivariant models when the target distribution is invariant/equivariant with respect to a compact group. Moreover, our work reveals an interesting relationship between generalisation, the number of training examples and properties of the group action. Our results rest on an observation of the structure of function spaces under averaging operators which, along with its consequences for feature averaging, may be of independent interest."
                },
                {
                    "title": "Regret Bound Balancing and Elimination for Model Selection in Bandits and RL",
                    "abstract": "We propose a simple model selection approach for algorithms in stochastic bandit and reinforcement learning problems. As opposed to prior work that (implicitly) assumes knowledge of the optimal regret, we only require that each base algorithm comes with a candidate regret bound that may or may not hold during all rounds. In each round, our approach plays a base algorithm to keep the candidate regret bounds of all remaining base algorithms balanced, and eliminates algorithms that violate their candidate bound. We prove that the total regret of this approach is bounded by the best valid candidate regret bound times a multiplicative factor. This factor is reasonably small in several applications, including linear bandits and MDPs with nested function classes, linear bandits with unknown misspecification, and LinUCB applied to linear bandits with different confidence parameters. We further show that, under a suitable gap-assumption, this factor only scales with the number of base algorithms and not their complexity when the number of rounds is large enough. Finally, unlike recent efforts in model selection for linear stochastic bandits, our approach is versatile enough to also cover cases where the context information is generated by an adversarial environment, rather than a stochastic one."
                },
                {
                    "title": "High-Dimensional Sparse Linear Bandits",
                    "abstract": "Stochastic linear bandits with high-dimensional sparse features are a practical model for a variety of domains, including personalized medicine and online advertising. We derive a novel $\\Omega(n^{2/3})$ dimension-free minimax regret lower bound for sparse linear bandits in the data-poor regime where the horizon is smaller than the ambient dimension and where the feature vectors admit a well-conditioned exploration distribution. This is complemented by a nearly matching upper bound for an explore-then-commit algorithm showing that that $\\Theta(n^{2/3})$ is the optimal rate in the data-poor regime. The results complement existing bounds for the data-rich regime and provide another example where carefully balancing the trade-off between information and regret is necessary. Finally, we prove a dimension-free $O(\\sqrt{n})$ regret upper bound under an additional assumption on the magnitude of the signal for relevant features."
                },
                {
                    "title": "Sparsity-Agnostic Lasso Bandit",
                    "abstract": "We consider a stochastic contextual bandit problem where the dimension $d$ of the feature vectors is potentially large, however, only a sparse subset of features of cardinality $s_0 \\ll d$ affect the reward function. Essentially all existing algorithms for sparse bandits require a priori knowledge of the value of the sparsity index $s_0$. This knowledge is almost never available in practice, and misspecification of this parameter can lead to severe deterioration in the performance of existing methods. The main contribution of this paper is to propose an algorithm that does \\emph{not} require prior knowledge of the sparsity index $s_0$ and establish tight regret bounds on its performance under mild conditions. We also comprehensively evaluate our proposed algorithm numerically and show that it consistently outperforms existing methods, even when the correct sparsity index is revealed to them but is kept hidden from our algorithm."
                },
                {
                    "title": "Bandit Algorithms",
                    "abstract": "sets of environments and policies respectively and ` : E \u00d7\u03a0\u2192 [0, 1] a bounded loss function. Given a policy \u03c0 let `(\u03c0) = (`(\u03bd1, \u03c0), . . . , `(\u03bdN , \u03c0)) be the loss vector resulting from policy \u03c0. Define S = {`(\u03c0) : \u03c0 \u2208 \u03a0} and \u03bb(S) = {x \u2208 cl(S) : y 6< x for all y \u2208 S} , where y 6< x is defined to mean it is not true that yi \u2264 xi for all i with strict inequality for at least one i. Prove that if \u03bb(S) \u2286 S and S is convex, then for each x \u2208 \u03bb(S) there exists a prior q \u2208 P(E) and policy \u03c0\u2217 such that `(\u03c0) = x and \u2211 \u03bd\u2208E q(\u03bd)`(\u03bd, \u03c0\u2217) = min \u03c0\u2208\u03a0 \u2211"
                },
                {
                    "title": "MDP Homomorphic Networks: Group Symmetries in Reinforcement Learning",
                    "abstract": "This paper introduces MDP homomorphic networks for deep reinforcement learning. MDP homomorphic networks are neural networks that are equivariant under symmetries in the joint state-action space of an MDP. Current approaches to deep reinforcement learning do not usually exploit knowledge about such structure. By building this prior knowledge into policy and value networks using an equivariance constraint, we can reduce the size of the solution space. We specifically focus on group-structured symmetries (invertible transformations). Additionally, we introduce an easy method for constructing equivariant network layers numerically, so the system designer need not solve the constraints by hand, as is typically done. We construct MDP homomorphic MLPs and CNNs that are equivariant under either a group of reflections or rotations. We show that such networks converge faster than unstructured baselines on CartPole, a grid world and Pong."
                },
                {
                    "title": "Problem-Complexity Adaptive Model Selection for Stochastic Linear Bandits",
                    "abstract": "We consider the problem of model selection for two popular stochastic linear bandit settings, and propose algorithms that adapts to the unknown problem complexity. In the first setting, we consider the $K$ armed mixture bandits, where the mean reward of arm $i \\in [K]$, is $\\mu_i+ \\langle \\alpha_{i,t},\\theta^* \\rangle $, with $\\alpha_{i,t} \\in \\mathbb{R}^d$ being the known context vector and $\\mu_i \\in [-1,1]$ and $\\theta^*$ are unknown parameters. We define $\\|\\theta^*\\|$ as the problem complexity and consider a sequence of nested hypothesis classes, each positing a different upper bound on $\\|\\theta^*\\|$. Exploiting this, we propose Adaptive Linear Bandit (ALB), a novel phase based algorithm that adapts to the true problem complexity, $\\|\\theta^*\\|$. We show that ALB achieves regret scaling of $O(\\|\\theta^*\\|\\sqrt{T})$, where $\\|\\theta^*\\|$ is apriori unknown. As a corollary, when $\\theta^*=0$, ALB recovers the minimax regret for the simple bandit algorithm without such knowledge of $\\theta^*$. ALB is the first algorithm that uses parameter norm as model section criteria for linear bandits. Prior state of art algorithms \\cite{osom} achieve a regret of $O(L\\sqrt{T})$, where $L$ is the upper bound on $\\|\\theta^*\\|$, fed as an input to the problem. In the second setting, we consider the standard linear bandit problem (with possibly an infinite number of arms) where the sparsity of $\\theta^*$, denoted by $d^* \\leq d$, is unknown to the algorithm. Defining $d^*$ as the problem complexity, we show that ALB achieves $O(d^*\\sqrt{T})$ regret, matching that of an oracle who knew the true sparsity level. This methodology is then extended to the case of finitely many arms and similar results are proven. This is the first algorithm that achieves such model selection guarantees. We further verify our results via synthetic and real-data experiments."
                },
                {
                    "title": "Plannable Approximations to MDP Homomorphisms: Equivariance under Actions",
                    "abstract": "This work exploits action equivariance for representation learning in reinforcement learning. Equivariance under actions states that transitions in the input space are mirrored by equivalent transitions in latent space, while the map and transition functions should also commute. We introduce a contrastive loss function that enforces action equivariance on the learned representations. We prove that when our loss is zero, we have a homomorphism of a deterministic Markov Decision Process (MDP). Learning equivariant maps leads to structured latent spaces, allowing us to build a model on which we plan through value iteration. We show experimentally that for deterministic MDPs, the optimal policy in the abstract MDP can be successfully lifted to the original MDP. Moreover, the approach easily adapts to changes in the goal states. Empirically, we show that in such MDPs, we obtain better representations in fewer epochs compared to representation learning approaches using reconstructions, while generalizing better to new goals than model-free approaches."
                },
                {
                    "title": "Generalizing Convolutional Neural Networks for Equivariance to Lie Groups on Arbitrary Continuous Data",
                    "abstract": "The translation equivariance of convolutional layers enables convolutional neural networks to generalize well on image problems. While translation equivariance provides a powerful inductive bias for images, we often additionally desire equivariance to other transformations, such as rotations, especially for non-image data. We propose a general method to construct a convolutional layer that is equivariant to transformations from any specified Lie group with a surjective exponential map. Incorporating equivariance to a new group requires implementing only the group exponential and logarithm maps, enabling rapid prototyping. Showcasing the simplicity and generality of our method, we apply the same model architecture to images, ball-and-stick molecular data, and Hamiltonian dynamical systems. For Hamiltonian systems, the equivariance of our models is especially impactful, leading to exact conservation of linear and angular momentum."
                },
                {
                    "title": "Enhanced Convolutional Neural Tangent Kernels",
                    "abstract": "Recent research shows that for training with $\\ell_2$ loss, convolutional neural networks (CNNs) whose width (number of channels in convolutional layers) goes to infinity correspond to regression with respect to the CNN Gaussian Process kernel (CNN-GP) if only the last layer is trained, and correspond to regression with respect to the Convolutional Neural Tangent Kernel (CNTK) if all layers are trained. An exact algorithm to compute CNTK (Arora et al., 2019) yielded the finding that classification accuracy of CNTK on CIFAR-10 is within 6-7% of that of that of the corresponding CNN architecture (best figure being around 78%) which is interesting performance for a fixed kernel. Here we show how to significantly enhance the performance of these kernels using two ideas. (1) Modifying the kernel using a new operation called Local Average Pooling (LAP) which preserves efficient computability of the kernel and inherits the spirit of standard data augmentation using pixel shifts. Earlier papers were unable to incorporate naive data augmentation because of the quadratic training cost of kernel regression. This idea is inspired by Global Average Pooling (GAP), which we show for CNN-GP and CNTK is equivalent to full translation data augmentation. (2) Representing the input image using a pre-processing technique proposed by Coates et al. (2011), which uses a single convolutional layer composed of random image patches. On CIFAR-10, the resulting kernel, CNN-GP with LAP and horizontal flip data augmentation, achieves 89% accuracy, matching the performance of AlexNet (Krizhevsky et al., 2012). Note that this is the best such result we know of for a classifier that is not a trained neural network. Similar improvements are obtained for Fashion-MNIST."
                },
                {
                    "title": "Doubly-Robust Lasso Bandit",
                    "abstract": "Contextual multi-armed bandit algorithms are widely used in sequential decision tasks such as news article recommendation systems, web page ad placement algorithms, and mobile health. Most of the existing algorithms have regret proportional to a polynomial function of the context dimension, $d$. In many applications however, it is often the case that contexts are high-dimensional with only a sparse subset of size $s_0 (\\ll d)$ being correlated with the reward. We consider the stochastic linear contextual bandit problem and propose a novel algorithm, namely the Doubly-Robust Lasso Bandit algorithm, which exploits the sparse structure of the regression parameter as in Lasso, while blending the doubly-robust technique used in missing data literature. The high-probability upper bound of the regret incurred by the proposed algorithm does not depend on the number of arms and scales with $\\mathrm{log}(d)$ instead of a polynomial function of $d$. The proposed algorithm shows good performance when contexts of different arms are correlated and requires less tuning parameters than existing methods."
                },
                {
                    "title": "Model selection for contextual bandits",
                    "abstract": "We introduce the problem of model selection for contextual bandits, where a learner must adapt to the complexity of the optimal policy while balancing exploration and exploitation. Our main result is a new model selection guarantee for linear contextual bandits. We work in the stochastic realizable setting with a sequence of nested linear policy classes of dimension $d_1 < d_2 < \\ldots$, where the $m^\\star$-th class contains the optimal policy, and we design an algorithm that achieves $\\tilde{O}(T^{2/3}d^{1/3}_{m^\\star})$ regret with no prior knowledge of the optimal dimension $d_{m^\\star}$. The algorithm also achieves regret $\\tilde{O}(T^{3/4} + \\sqrt{Td_{m^\\star}})$, which is optimal for $d_{m^{\\star}}\\geq{}\\sqrt{T}$. This is the first model selection result for contextual bandits with non-vacuous regret for all values of $d_{m^\\star}$, and to the best of our knowledge is the first positive result of this type for any online learning setting with partial information. The core of the algorithm is a new estimator for the gap in the best loss achievable by two linear policy classes, which we show admits a convergence rate faster than the rate required to learn the parameters for either class."
                },
                {
                    "title": "Symmetry Learning for Function Approximation in Reinforcement Learning",
                    "abstract": "In this paper we explore methods to exploit symmetries for ensuring sample efficiency in reinforcement learning (RL), this problem deserves ever increasing attention with the recent advances in the use of deep networks for complex RL tasks which require large amount of training data. We introduce a novel method to detect symmetries using reward trails observed during episodic experience and prove its completeness. We also provide a framework to incorporate the discovered symmetries for functional approximation. Finally we show that the use of potential based reward shaping is especially effective for our symmetry exploitation mechanism. Experiments on various classical problems show that our method improves the learning performance significantly by utilizing symmetry information."
                },
                {
                    "title": "Corralling a Band of Bandit Algorithms",
                    "abstract": "We study the problem of combining multiple bandit algorithms (that is, online learning algorithms with partial feedback) with the goal of creating a master algorithm that performs almost as well as the best base algorithm if it were to be run on its own. The main challenge is that when run with a master, base algorithms unavoidably receive much less feedback and it is thus critical that the master not starve a base algorithm that might perform uncompetitively initially but would eventually outperform others if given enough feedback. We address this difficulty by devising a version of Online Mirror Descent with a special mirror map together with a sophisticated learning rate scheme. We show that this approach manages to achieve a more delicate balance between exploiting and exploring base algorithms than previous works yielding superior regret bounds. \nOur results are applicable to many settings, such as multi-armed bandits, contextual bandits, and convex bandits. As examples, we present two main applications. The first is to create an algorithm that enjoys worst-case robustness while at the same time performing much better when the environment is relatively easy. The second is to create an algorithm that works simultaneously under different assumptions of the environment, such as different priors or different loss structures."
                },
                {
                    "title": "Linear Multi-Resource Allocation with Semi-Bandit Feedback",
                    "abstract": "We study an idealised sequential resource allocation problem. In each time step the learner chooses an allocation of several resource types between a number of tasks. Assigning more resources to a task increases the probability that it is completed. The problem is challenging because the alignment of the tasks to the resource types is unknown and the feedback is noisy. Our main contribution is the new setting and an algorithm with nearly-optimal regret analysis. Along the way we draw connections to the problem of minimising regret for stochastic linear bandits with heteroscedastic noise. We also present some new results for stochastic linear bandits on the hypercube that significantly improve on existing work, especially in the sparse case."
                },
                {
                    "title": "Introduction to High-Dimensional Statistics",
                    "abstract": "Preface Acknowledgments Introduction High-Dimensional Data Curse of Dimensionality Lost in the Immensity of High-Dimensional Spaces Fluctuations Cumulate An Accumulation of Rare Events May Not Be Rare Computational Complexity High-Dimensional Statistics Circumventing the Curse of Dimensionality A Paradigm Shift Mathematics of High-Dimensional Statistics About This Book Statistics and Data Analysis Purpose of This Book Overview Discussion and References Take-Home Message References Exercises Strange Geometry of High-Dimensional Spaces Volume of a p-Dimensional Ball Tails of a Standard Gaussian Distribution Principal Component Analysis Basics of Linear Regression Concentration of the Square Norm of a Gaussian Random Variable Model Selection Statistical Setting To Select among a Collection of Models Models and Oracle Model Selection Procedures Risk Bound for Model Selection Oracle Risk Bound Optimality Minimax Optimality Frontier of Estimation in High Dimensions Minimal Penalties Computational Issues Illustration An Alternative Point of View on Model Selection Discussion and References Take-Home Message References Exercises Orthogonal Design Risk Bounds for the Different Sparsity Settings Collections of Nested Models Segmentation with Dynamic Programming Goldenshluger-Lepski Method Minimax Lower Bounds Aggregation of Estimators Introduction Gibbs Mixing of Estimators Oracle Risk Bound Numerical Approximation by Metropolis-Hastings Numerical Illustration Discussion and References Take-Home Message References Exercises Gibbs Distribution Orthonormal Setting with Power Law Prior Group-Sparse Setting Gain of Combining Online Aggregation Convex Criteria Reminder on Convex Multivariate Functions Subdifferentials Two Useful Properties Lasso Estimator Geometric Insights Analytic Insights Oracle Risk Bound Computing the Lasso Estimator Removing the Bias of the Lasso Estimator Convex Criteria for Various Sparsity Patterns Group-Lasso (Group Sparsity) Sparse-Group Lasso (Sparse-Group Sparsity) Fused-Lasso (Variation Sparsity) Discussion and References Take-Home Message References Exercises When Is the Lasso Solution Unique? Support Recovery via the Witness Approach Lower Bound on the Compatibility Constant On the Group-Lasso Dantzig Selector Projection on the l1-Ball Ridge and Elastic-Net Estimator Selection Estimator Selection Cross-Validation Techniques Complexity Selection Techniques Coordinate-Sparse Regression Group-Sparse Regression Multiple Structures Scaled-Invariant Criteria References and Discussion Take-Home Message References Exercises Expected V-Fold CV l2-Risk Proof of Corollary 5.5 Some Properties of Penalty (5.4) Selecting the Number of Steps for the Forward Algorithm Multivariate Regression Statistical Setting A Reminder on Singular Values Low-Rank Estimation If We Knew the Rank of A* When the Rank of A* Is Unknown Low Rank and Sparsity Row-Sparse Matrices Criterion for Row-Sparse and Low-Rank Matrices Convex Criterion for Low Rank Matrices Convex Criterion for Sparse and Low-Rank Matrices Discussion and References Take-Home Message References Exercises Hard-Thresholding of the Singular Values Exact Rank Recovery Rank Selection with Unknown Variance Graphical Models Reminder on Conditional Independence Graphical Models Directed Acyclic Graphical Models Nondirected Models Gaussian Graphical Models (GGM) Connection with the Precision Matrix and the Linear Regression Estimating g by Multiple Testing Sparse Estimation of the Precision Matrix Estimation of g by Regression Practical Issues Discussion and References Take-Home Message References Exercises Factorization in Directed Models Moralization of a Directed Graph Convexity of -log(det(K)) Block Gradient Descent with the l1 / l2 Penalty Gaussian Graphical Models with Hidden Variables Dantzig Estimation of Sparse Gaussian Graphical Models Gaussian Copula Graphical Models Restricted Isometry Constant for Gaussian Matrices Multiple Testing An Introductory Example Differential Expression of a Single Gene Differential Expression of Multiple Genes Statistical Setting p-Values Multiple Testing Setting Bonferroni Correction Controlling the False Discovery Rate Heuristics Step-Up Procedures FDR Control under the WPRDS Property Illustration Discussion and References Take-Home Message References Exercises FDR versus FWER WPRDS Property Positively Correlated Normal Test Statistics Supervised Classification Statistical Modeling Bayes Classifier Parametric Modeling Semi-Parametric Modeling Nonparametric Modeling Empirical Risk Minimization Misclassification Probability of the Empirical Risk Minimizer Vapnik-Chervonenkis Dimension Dictionary Selection From Theoretical to Practical Classifiers Empirical Risk Convexification Statistical Properties Support Vector Machines AdaBoost Classifier Selection Discussion and References Take-Home Message References Exercises Linear Discriminant Analysis VC Dimension of Linear Classifiers in Rd Linear Classifiers with Margin Constraints Spectral Kernel Computation of the SVM Classifier Kernel Principal Component Analysis (KPCA) Gaussian Distribution Gaussian Random Vectors Chi-Square Distribution Gaussian Conditioning Probabilistic Inequalities Basic Inequalities Concentration Inequalities McDiarmid Inequality Gaussian Concentration Inequality Symmetrization and Contraction Lemmas Symmetrization Lemma Contraction Principle Birge's Inequality Linear Algebra Singular Value Decomposition (SVD) Moore-Penrose Pseudo-Inverse Matrix Norms Matrix Analysis Subdifferentials of Convex Functions Subdifferentials and Subgradients Examples of Subdifferentials Reproducing Kernel Hilbert Spaces Notations Bibliography Index"
                },
                {
                    "title": "Combinatorics of Set Partitions",
                    "abstract": "Introduction Historical Overview and Earliest Results Timeline of Research for Set Partitions A More Detailed Book Basic Tools of the Book Sequences Solving Recurrence Relations Generating Functions Lagrange Inversion Formula The Principle of Inclusion and Exclusion Generating Trees Preliminary Results on Set Partitions Dobinski's Formula Different Representations Subword Statistics on Set Partitions Subword Patterns of Size Two: Rises, Levels and Descents Peaks and Valleys Subword Patterns: l-Rises, l-Levels, and l-Descents Families of Subword Patterns Patterns of Size Three Nonsubword Statistics on Set Partitions Statistics and Block Representation Statistics and Canonical and Rook Representations Records and Weak Records Number of Positions between Adjacent Occurrences of a Letter The Internal Statistic Statistics and Generalized Patterns Major Index Number of Crossings, Nestings and Alignments Avoidance of Patterns in Set Partitions History and Connections Avoidance of Subsequence Patterns Generalized Patterns Partially Ordered Patterns Multi Restrictions on Set Partitions Avoiding a Pattern of Size Three and Another Pattern Pattern Avoidance in Noncrossing Set Partitions General Equivalences Two Patterns of Size Four Left Motzkin Numbers Sequence A054391 Catalan and Generalized Catalan Numbers Pell Numbers Regular Set Partitions Distance Restrictions Singletons Block-Connected Asymptotics and Random Set Partition Tools from Probability Theory Tools from Complex Analysis Z-Statistics Set Partitions as Geometric Words Asymptotics for Set Partitions Gray Codes, Loopless Algorithms and Set Partitions Gray Code and Loopless Algorithms Gray Codes for Pn Loopless Algorithm for Generating Pn Set Partitions and Normal Ordering Preliminaries Linear Representation and N((a+a)n) Wick's Theorem and q-Normal Ordering p-Normal Ordering Noncrossing Normal Ordering Appendices Bibliography Index Exercises, Research Directions, and Open Problems appear at the end of each chapter."
                },
                {
                    "title": "Online-to-Confidence-Set Conversions and Application to Sparse Stochastic Bandits",
                    "abstract": "We introduce a novel technique, which we call online-to-confidence-set conversion. The technique allows us to construct highprobability confidence sets for linear prediction with correlated inputs given the predictions of any algorithm (e.g., online LASSO, exponentiated gradient algorithm, online least-squares, p-norm algorithm) targeting online learning with linear predictors and the quadratic loss. By construction, the size of the confidence set is directly governed by the regret of the online learning algorithm. Constructing tight confidence sets is interesting on its own, but the new technique is given extra weight by the fact having access tight confidence sets underlies a number of important problems. The advantage of our construction here is that progress in constructing better algorithms for online prediction problems directly translates into tighter confidence sets. In this paper, this is demonstrated in the case of linear stochastic bandits. In particular, we introduce the sparse variant of linear stochastic bandits and show that a recent online algorithm together with our online-to-confidence-set conversion allows one to derive algorithms that can exploit if the reward is a function of a sparse linear combination of the components of the chosen action. Appearing in Proceedings of the 15 International Conference on Artificial Intelligence and Statistics (AISTATS) 2012, La Palma, Canary Islands. Volume XX of JMLR: W&CP XX. Copyright 2012 by the authors."
                },
                {
                    "title": "Bandit Theory meets Compressed Sensing for high dimensional Stochastic Linear Bandit",
                    "abstract": "We consider a linear stochastic bandit problem where the dimension K of the unknown parameter is larger than the sampling budget n. In such cases, it is in general impossible to derive sub-linear regret bounds since usual linear bandit algorithms have a regret in O(K p n). In this paper we assume that is S sparse, i.e. has at most S non-zero components, and that the space of arms is the unit ball for the jj:jj2 norm. We combine ideas from Compressed Sensing and Bandit Theory and derive an algorithm with a regret bound in O(S p n). We detail an application to the problem of optimizing a function that depends on many variables but among which only a small number of them (initially unknown) are relevant."
                },
                {
                    "title": "Improved Algorithms for Linear Stochastic Bandits",
                    "abstract": "We improve the theoretical analysis and empirical performance of algorithms for the stochastic multi-armed bandit problem and the linear stochastic multi-armed bandit problem. In particular, we show that a simple modification of Auer's UCB algorithm (Auer, 2002) achieves with high probability constant regret. More importantly, we modify and, consequently, improve the analysis of the algorithm for the for linear stochastic bandit problem studied by Auer (2002), Dani et al. (2008), Rusmevichientong and Tsitsiklis (2010), Li et al. (2010). Our modification improves the regret bound by a logarithmic factor, though experiments show a vast improvement. In both cases, the improvement stems from the construction of smaller confidence sets. For their construction we use a novel tail inequality for vector-valued martingales."
                },
                {
                    "title": "SYMMETRIES IN LINEAR AND INTEGER PROGRAMS",
                    "abstract": "The notion of symmetry is dened in the context of Linear and Integer Programming. Symmetric linear and integer programs are studied from a group theoretical viewpoint. We show that for any linear program there exists an optimal solution in the xed point set of its symmetry group. Using this result, we develop an algorithm that allows for reducing the dimension of any linear program having a non-trivial group of symmetries."
                },
                {
                    "title": "Sampling Theorems for Signals From the Union of Finite-Dimensional Linear Subspaces",
                    "abstract": "Compressed sensing is an emerging signal acquisition technique that enables signals to be sampled well below the Nyquist rate, given that the signal has a sparse representation in an orthonormal basis. In fact, sparsity in an orthonormal basis is only one possible signal model that allows for sampling strategies below the Nyquist rate. In this paper, we consider a more general signal model and assume signals that live on or close to the union of linear subspaces of low dimension. We present sampling theorems for this model that are in the same spirit as the Nyquist-Shannon sampling theorem in that they connect the number of required samples to certain model parameters. Contrary to the Nyquist-Shannon sampling theorem, which gives a necessary and sufficient condition for the number of required samples as well as a simple linear algorithm for signal reconstruction, the model studied here is more complex. We therefore concentrate on two aspects of the signal model, the existence of one to one maps to lower dimensional observation spaces and the smoothness of the inverse map. We show that almost all linear maps are one to one when the observation space is at least of the same dimension as the largest dimension of the convex hull of the union of any two subspaces in the model. However, we also show that in order for the inverse map to have certain smoothness properties such as a given finite Lipschitz constant, the required observation dimension necessarily depends logarithmically on the number of subspaces in the signal model. In other words, while unique linear sampling schemes require a small number of samples depending only on the dimension of the subspaces involved, in order to have stable sampling methods, the number of samples depends necessarily logarithmically on the number of subspaces in the model. These results are then applied to two examples, the standard compressed sensing signal model in which the signal has a sparse representation in an orthonormal basis and to a sparse signal model with additional tree structure."
                },
                {
                    "title": "Stochastic bandits with groups of similar arms",
                    "abstract": "We consider a variant of the stochastic multi-armed bandit problem where arms are known to be organized into different groups having the same mean. The groups are unknown but a lower bound q on their size is known. This situation typically appears when each arm can be described with a list of categorical attributes, and the (unknown) mean reward function only depends on a subset of them, the others being redundant. In this case, q is linked naturally to the number of attributes considered redundant, and the number of categories of each attribute. For this structured problem of practical relevance, we \ufb01rst derive the asymptotic regret lower bound and corresponding constrained optimization problem. They reveal the achievable regret can be substantially reduced when compared to the unstructured setup, possibly by a factor q . However, solving exactly the exact constrained optimization problem involves a combinatorial problem. We introduce a lowerbound inspired strategy involving a computationally ef\ufb01cient relaxation that is based on a sorting mechanism. We further prove it achieves a lower bound close to the optimal one up to a controlled factor, and achieves an asymptotic regret q times smaller than the unstructured one. We believe this shows it is a valuable strategy for the practitioner. Last, we illustrate the performance of the considered strategy on numerical experiments involving a large number of arms."
                },
                {
                    "title": "Pessimism Meets Invariance: Provably Efficient Offline Mean-Field Multi-Agent RL",
                    "abstract": "Mean-Field Multi-Agent Reinforcement Learning (MF-MARL) is attractive in the applications involving a large population of homogeneous agents, as it exploits the permutation invariance of agents and avoids the curse of many agents. Most existing results only focus on online settings, in which agents can interact with the environment during training. In some applications such as social welfare optimization, however, the interaction during training can be prohibitive or even unethical in the societal systems. To bridge such a gap, we propose a SAFARI (peSsimistic meAn-Field vAlue iteRatIon) algorithm for off-line MF-MARL, which only requires a handful of pre-collected experience data. Theoretically, under a weak coverage assumption that the experience dataset contains enough information about the optimal policy, we prove that for an episodic mean-field MDP with a horizon H and N training trajectories, SAFARI attains a sub-optimality gap of O ( H 2 d eff / \u221a N ) , where d eff is the effective dimension of the function class for parameterizing the value function, but independent on the number of agents. Numerical experiments are provided."
                },
                {
                    "title": "Approximate Homomorphisms : A framework for non-exact minimization in Markov Decision Processes",
                    "abstract": "To operate effectively in complex environments learning agents require the ability to selectively ignore irrelevant details and form useful abstractions. In earlier work we explored in detail what constitutes a useful abstraction in a stochastic sequential decision problem modeled as a Markov Decision Process (MDP). We based our approach on the notion of an MDP homomorphism. In this article we look at relaxing the strict conditions imposed earlier and introduce approximate homomorphisms that allow us to construct useful abstract models even when the homomorphism conditions are not met exactly. We also present a result on bounding the loss resulting from this approximation."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nCan one leverage symmetry in sequential decision-making tasks to enable effective exploration, and eventually break the curse of dimensionality?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is significant for the research community as it could lead to a deeper understanding of how symmetry can be integrated into sequential decision-making frameworks, which has been largely overlooked compared to sparsity. By addressing this question, future research could explore new algorithms that utilize symmetry, potentially leading to improved performance in high-dimensional settings. This advancement could have practical applications in various fields, such as robotics and multi-agent systems, where understanding and exploiting symmetry can enhance learning efficiency and decision-making processes.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the complexity of identifying and incorporating symmetry structures in high-dimensional feature spaces. Naive approaches may fail because they do not account for the hidden or partial nature of symmetry in real-world tasks, leading to ineffective exploration strategies. Additionally, the theoretical understanding of how to model and leverage symmetry in a way that is computationally feasible poses significant obstacles. The lack of established methodologies for symmetry learning in sequential decision-making further complicates the problem.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on sparsity as an inductive bias in linear stochastic bandits, neglecting the potential of symmetry. This gap exists partly because most studies on symmetry in machine learning assume prior knowledge of symmetry structures, which is often not available in practical scenarios. Barriers such as the lack of algorithms that can learn symmetry from partial information and the absence of theoretical frameworks for symmetry in sequential decision-making have prevented this problem from being addressed. Our approach aims to fill this gap by developing methods that incorporate symmetry learning mechanisms.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves developing algorithms that explicitly integrate symmetry learning into linear stochastic bandit frameworks. We will utilize a dataset that captures various sequential decision-making tasks with inherent symmetry, and we will evaluate our algorithms using metrics such as cumulative regret and exploration efficiency. The expected outcomes include demonstrating that our symmetry-based approach can significantly reduce regret and improve performance in high-dimensional settings compared to traditional methods that do not leverage symmetry."
            }
        },
        "author_data": {
            "bb2d2b5e-6940-4f02-8690-3d43ad760756": {
                "pk": "bb2d2b5e-6940-4f02-8690-3d43ad760756",
                "name": "Nam Phuong Tran",
                "collaborators": [
                    "Long Tran-Thanh",
                    "The Anh Ta",
                    "Shuqing Shi",
                    "Debmalya Mandal",
                    "Yali Du"
                ],
                "domain": [
                    "Online Optimization",
                    "Bandit Theory",
                    "Cooperative Games",
                    "Reward Allocation"
                ],
                "publications": [
                    {
                        "title": "Invariant Lipschitz Bandits: A Side Observation Approach",
                        "abstract": "Symmetry arises in many optimization and decision-making problems, and has attracted considerable attention from the optimization community: By utilizing the existence of such symmetries, the process of searching for optimal solutions can be improved significantly. Despite its success in (offline) optimization, the utilization of symmetries has not been well examined within the online optimization settings, especially in the bandit literature. As such, in this paper we study the invariant Lipschitz bandit setting, a subclass of the Lipschitz bandits where the reward function and the set of arms are preserved under a group of transformations. We introduce an algorithm named \\texttt{UniformMesh-N}, which naturally integrates side observations using group orbits into the \\texttt{UniformMesh} algorithm (\\cite{Kleinberg2005_UniformMesh}), which uniformly discretizes the set of arms. Using the side-observation approach, we prove an improved regret upper bound, which depends on the cardinality of the group, given that the group is finite. We also prove a matching regret's lower bound for the invariant Lipschitz bandit class (up to logarithmic factors). We hope that our work will ignite further investigation of symmetry in bandit theory and sequential decision-making theory in general."
                    },
                    {
                        "title": "Learning the Expected Core of Strictly Convex Stochastic Cooperative Games",
                        "abstract": "Reward allocation, also known as the credit assignment problem, has been an important topic in economics, engineering, and machine learning. An important concept in reward allocation is the core, which is the set of stable allocations where no agent has the motivation to deviate from the grand coalition. In previous works, computing the core requires either knowledge of the reward function in deterministic games or the reward distribution in stochastic games. However, this is unrealistic, as the reward function or distribution is often only partially known and may be subject to uncertainty. In this paper, we consider the core learning problem in stochastic cooperative games, where the reward distribution is unknown. Our goal is to learn the expected core, that is, the set of allocations that are stable in expectation, given an oracle that returns a stochastic reward for an enquired coalition each round. Within the class of strictly convex games, we present an algorithm named \\texttt{Common-Points-Picking} that returns a point in the expected core given a polynomial number of samples, with high probability. To analyse the algorithm, we develop a new extension of the separation hyperplane theorem for multiple convex sets."
                    }
                ]
            },
            "0cf47a80-dd6c-4968-a883-afcf4ff3f873": {
                "pk": "0cf47a80-dd6c-4968-a883-afcf4ff3f873",
                "name": "The Anh Ta",
                "collaborators": [
                    "Nam Phuong Tran",
                    "Shuqing Shi",
                    "Debmalya Mandal",
                    "Yali Du",
                    "Long Tran-Thanh"
                ],
                "domain": [
                    "Game Theory",
                    "Machine Learning",
                    "Stochastic Processes",
                    "Cooperative Systems"
                ],
                "publications": [
                    {
                        "title": "Learning the Expected Core of Strictly Convex Stochastic Cooperative Games",
                        "abstract": "Reward allocation, also known as the credit assignment problem, has been an important topic in economics, engineering, and machine learning. An important concept in reward allocation is the core, which is the set of stable allocations where no agent has the motivation to deviate from the grand coalition. In previous works, computing the core requires either knowledge of the reward function in deterministic games or the reward distribution in stochastic games. However, this is unrealistic, as the reward function or distribution is often only partially known and may be subject to uncertainty. In this paper, we consider the core learning problem in stochastic cooperative games, where the reward distribution is unknown. Our goal is to learn the expected core, that is, the set of allocations that are stable in expectation, given an oracle that returns a stochastic reward for an enquired coalition each round. Within the class of strictly convex games, we present an algorithm named \\texttt{Common-Points-Picking} that returns a point in the expected core given a polynomial number of samples, with high probability. To analyse the algorithm, we develop a new extension of the separation hyperplane theorem for multiple convex sets."
                    }
                ]
            },
            "38314fb2-3d28-41b1-ac60-46652a9aa843": {
                "pk": "38314fb2-3d28-41b1-ac60-46652a9aa843",
                "name": "Debmalya Mandal",
                "collaborators": [
                    "Goran Radanovic",
                    "Jiarui Gan",
                    "Stelios Triantafyllou",
                    "Rupak Majumdar",
                    "Adish Singla",
                    "Hadi Hosseini",
                    "David Parkes",
                    "Ben Rank",
                    "Amrit Puhan",
                    "Aleksa Sukovic"
                ],
                "domain": [
                    "Causal Inference",
                    "Reinforcement Learning",
                    "Multi-Agent Systems",
                    "Graph Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Weighted Tensor Completion for Time-Series Causal Inference",
                        "abstract": "Marginal Structural Models (MSM) are the most popular models for causal inference from time-series observational data. However, they have two main drawbacks: (a) they do not capture subject heterogeneity, and (b) they only consider fixed time intervals and do not scale gracefully with longer intervals. In this work, we propose a new family of MSMs to address these two concerns. We model the potential outcomes as a three-dimensional tensor of low rank, where the three dimensions correspond to the agents, time periods and the set of possible histories. Unlike the traditional MSM, we allow the dimensions of the tensor to increase with the number of agents and time periods. We set up a weighted tensor completion problem as our estimation procedure, and show that the solution to this problem converges to the true model in an appropriate sense. Then we show how to solve the estimation problem, providing conditions under which we can approximately and efficiently solve the estimation problem. Finally we propose an algorithm based on projected gradient descent, which is easy to implement, and evaluate its performance on a simulated dataset."
                    },
                    {
                        "title": "Socially Fair Reinforcement Learning",
                        "abstract": "We consider the problem of episodic reinforcement learning where there are multiple stakeholders with different reward functions. Our goal is to output a policy that is socially fair with respect to different reward functions. Prior works have proposed different objectives that a fair policy must optimize including minimum welfare, and generalized Gini welfare. We first take an axiomatic view of the problem, and propose four axioms that any such fair objective must satisfy. We show that the Nash social welfare is the unique objective that uniquely satisfies all four objectives, whereas prior objectives fail to satisfy all four axioms. We then consider the learning version of the problem where the underlying model i.e. Markov decision process is unknown. We consider the problem of minimizing regret with respect to the fair policies maximizing three different fair objectives -- minimum welfare, generalized Gini welfare, and Nash social welfare. Based on optimistic planning, we propose a generic learning algorithm and derive its regret bound with respect to the three different policies. For the objective of Nash social welfare, we also derive a lower bound in regret that grows exponentially with $n$, the number of agents. Finally, we show that for the objective of minimum welfare, one can improve regret by a factor of $O(H)$ for a weaker notion of regret."
                    },
                    {
                        "title": "Performative Reinforcement Learning",
                        "abstract": "We introduce the framework of performative reinforcement learning where the policy chosen by the learner affects the underlying reward and transition dynamics of the environment. Following the recent literature on performative prediction~\\cite{Perdomo et. al., 2020}, we introduce the concept of performatively stable policy. We then consider a regularized version of the reinforcement learning problem and show that repeatedly optimizing this objective converges to a performatively stable policy under reasonable assumptions on the transition dynamics. Our proof utilizes the dual perspective of the reinforcement learning problem and may be of independent interest in analyzing the convergence of other algorithms with decision-dependent environments. We then extend our results for the setting where the learner just performs gradient ascent steps instead of fully optimizing the objective, and for the setting where the learner has access to a finite number of trajectories from the changed environment. For both settings, we leverage the dual formulation of performative reinforcement learning and establish convergence to a stable solution. Finally, through extensive experiments on a grid-world environment, we demonstrate the dependence of convergence on various parameters e.g. regularization, smoothness, and the number of samples."
                    },
                    {
                        "title": "Performative Reinforcement Learning in Gradually Shifting Environments",
                        "abstract": "When Reinforcement Learning (RL) agents are deployed in practice, they might impact their environment and change its dynamics. We propose a new framework to model this phenomenon, where the current environment depends on the deployed policy as well as its previous dynamics. This is a generalization of Performative RL (PRL) [Mandal et al., 2023]. Unlike PRL, our framework allows to model scenarios where the environment gradually adjusts to a deployed policy. We adapt two algorithms from the performative prediction literature to our setting and propose a novel algorithm called Mixed Delayed Repeated Retraining (MDRR). We provide conditions under which these algorithms converge and compare them using three metrics: number of retrainings, approximation guarantee, and number of samples per deployment. MDRR is the first algorithm in this setting which combines samples from multiple deployments in its training. This makes MDRR particularly suitable for scenarios where the environment's response strongly depends on its previous dynamics, which are common in practice. We experimentally compare the algorithms using a simulation-based testbed and our results show that MDRR converges significantly faster than previous approaches."
                    },
                    {
                        "title": "The Surprising Effectiveness of SP Voting with Partial Preferences",
                        "abstract": "We consider the problem of recovering the ground truth ordering (ranking, top-$k$, or others) over a large number of alternatives. The wisdom of crowd is a heuristic approach based on Condorcet's Jury theorem to address this problem through collective opinions. This approach fails to recover the ground truth when the majority of the crowd is misinformed. The surprisingly popular (SP) algorithm cite{prelec2017solution} is an alternative approach that is able to recover the ground truth even when experts are in minority. The SP algorithm requires the voters to predict other voters' report in the form of a full probability distribution over all rankings of alternatives. However, when the number of alternatives, $m$, is large, eliciting the prediction report or even the vote over $m$ alternatives might be too costly. In this paper, we design a scalable alternative of the SP algorithm which only requires eliciting partial preferences from the voters, and propose new variants of the SP algorithm. In particular, we propose two versions -- Aggregated-SP and Partial-SP -- that ask voters to report vote and prediction on a subset of size $k$ ($\\ll m$) in terms of top alternative, partial rank, or an approval set. Through a large-scale crowdsourcing experiment on MTurk, we show that both of our approaches outperform conventional preference aggregation algorithms for the recovery of ground truth rankings, when measured in terms of Kendall-Tau distance and Spearman's $\\rho$. We further analyze the collected data and demonstrate that voters' behavior in the experiment, including the minority of the experts, and the SP phenomenon, can be correctly simulated by a concentric mixtures of Mallows model. Finally, we provide theoretical bounds on the sample complexity of SP algorithms with partial rankings to demonstrate the theoretical guarantees of the proposed methods."
                    },
                    {
                        "title": "Agent-Specific Effects: A Causal Effect Propagation Analysis in Multi-Agent MDPs",
                        "abstract": "Establishing causal relationships between actions and outcomes is fundamental for accountable multi-agent decision-making. However, interpreting and quantifying agents' contributions to such relationships pose significant challenges. These challenges are particularly prominent in the context of multi-agent sequential decision-making, where the causal effect of an agent's action on the outcome depends on how other agents respond to that action. In this paper, our objective is to present a systematic approach for attributing the causal effects of agents' actions to the influence they exert on other agents. Focusing on multi-agent Markov decision processes, we introduce agent-specific effects (ASE), a novel causal quantity that measures the effect of an agent's action on the outcome that propagates through other agents. We then turn to the counterfactual counterpart of ASE (cf-ASE), provide a sufficient set of conditions for identifying cf-ASE, and propose a practical sampling-based algorithm for estimating it. Finally, we experimentally evaluate the utility of cf-ASE through a simulation-based testbed, which includes a sepsis management environment."
                    },
                    {
                        "title": "Stochastic Principal-Agent Problems: Efficient Computation and Learning",
                        "abstract": "We introduce a stochastic principal-agent model. A principal and an agent interact in a stochastic environment, each privy to observations about the state not available to the other. The principal has the power of commitment, both to elicit information from the agent and to provide signals about her own information. The players communicate with each other and then select actions independently. Each of them receives a payoff based on the state and their joint action, and the environment transitions to a new state. The interaction continues over a finite time horizon. Both players are far-sighted, aiming to maximize their total payoffs over the time horizon. The model encompasses as special cases extensive-form games (EFGs) and stochastic games of incomplete information, partially observable Markov decision processes (POMDPs), as well as other forms of sequential principal-agent interactions, including Bayesian persuasion and automated mechanism design problems.   We consider both the computation and learning of the principal's optimal policy. Since the general problem, which subsumes POMDPs, is intractable, we explore algorithmic solutions under hindsight observability, where the state and the interaction history are revealed at the end of each step. Though the problem becomes more amenable under this condition, the number of possible histories remains exponential in the length of the time horizon, making approaches for EFG-based models infeasible. We present an efficient algorithm based on the inducible value sets. The algorithm computes an $\\epsilon$-approximate optimal policy in time polynomial in $1/\\epsilon$. Additionally, we show an efficient learning algorithm for an episodic reinforcement learning setting where the transition probabilities are unknown. The algorithm guarantees sublinear regret $\\tilde{O}(T^{2/3})$ for both players over $T$ episodes."
                    },
                    {
                        "title": "Surprisingly Popular Voting Recovers Rankings, Surprisingly!",
                        "abstract": "The wisdom of the crowd has long become the de facto approach for eliciting information from individuals or experts in order to predict the ground truth. However, classical democratic approaches for aggregating individual \\emph{votes} only work when the opinion of the majority of the crowd is relatively accurate. A clever recent approach, \\emph{surprisingly popular voting}, elicits additional information from the individuals, namely their \\emph{prediction} of other individuals' votes, and provably recovers the ground truth even when experts are in minority. This approach works well when the goal is to pick the correct option from a small list, but when the goal is to recover a true ranking of the alternatives, a direct application of the approach requires eliciting too much information. We explore practical techniques for extending the surprisingly popular algorithm to ranked voting by partial votes and predictions and designing robust aggregation rules. We experimentally demonstrate that even a little prediction information helps surprisingly popular voting outperform classical approaches."
                    },
                    {
                        "title": "Meta-Learning with Graph Neural Networks: Methods and Applications",
                        "abstract": "Graph Neural Networks (GNNs), a generalization of deep neural networks on graph data have been widely used in various domains, ranging from drug discovery to recommender systems. However, GNNs on such applications are limited when there are few available samples. Meta-learning has been an important framework to address the lack of samples in machine learning, and in recent years, researchers have started to apply meta-learning to GNNs. In this work, we provide a comprehensive survey of different meta-learning approaches involving GNNs on various graph problems showing the power of using these two approaches together. We categorize the literature based on proposed architectures, shared representations, and applications. Finally, we discuss several exciting future research directions and open problems."
                    },
                    {
                        "title": "Learning Tensor Representations for Meta-Learning",
                        "abstract": "We introduce a tensor-based model of shared representation for meta-learning from a diverse set of tasks. Prior works on learning linear representations for meta-learning assume that there is a common shared representation across different tasks, and do not consider the additional task-specific observable side information. In this work, we model the meta-parameter through an order-$3$ tensor, which can adapt to the observed task features of the task. We propose two methods to estimate the underlying tensor. The first method solves a tensor regression problem and works under natural assumptions on the data generating process. The second method uses the method of moments under additional distributional assumptions and has an improved sample complexity in terms of the number of tasks.   We also focus on the meta-test phase, and consider estimating task-specific parameters on a new task. Substituting the estimated tensor from the first step allows us estimating the task-specific parameters with very few samples of the new task, thereby showing the benefits of learning tensor representations for meta-learning. Finally, through simulation and several real-world datasets, we evaluate our methods and show that it improves over previous linear models of shared representations for meta-learning."
                    },
                    {
                        "title": "Online Reinforcement Learning with Uncertain Episode Lengths",
                        "abstract": "Existing episodic reinforcement algorithms assume that the length of an episode is fixed across time and known a priori. In this paper, we consider a general framework of episodic reinforcement learning when the length of each episode is drawn from a distribution. We first establish that this problem is equivalent to online reinforcement learning with general discounting where the learner is trying to optimize the expected discounted sum of rewards over an infinite horizon, but where the discounting function is not necessarily geometric. We show that minimizing regret with this new general discounting is equivalent to minimizing regret with uncertain episode lengths. We then design a reinforcement learning algorithm that minimizes regret with general discounting but acts for the setting with uncertain episode lengths. We instantiate our general bound for different types of discounting, including geometric and polynomial discounting. We also show that we can obtain similar regret bounds even when the uncertainty over the episode lengths is unknown, by estimating the unknown distribution over time. Finally, we compare our learning algorithms with existing value-iteration based episodic RL algorithms in a grid-world environment."
                    },
                    {
                        "title": "Sequential Blocked Matching",
                        "abstract": "We consider a sequential blocked matching (SBM) model where strategic agents repeatedly report ordinal preferences over a set of services to a central planner. The planner's goal is to elicit agents' true preferences and design a policy that matches services to agents in order to maximize the expected social welfare with the added constraint that each matched service can be \\emph{blocked} or unavailable for a number of time periods. Naturally, SBM models the repeated allocation of reusable services to a set of agents where each allocated service becomes unavailable for a fixed duration.   We first consider the offline SBM setting, where the strategic agents are aware of their true preferences. We measure the performance of any policy by \\emph{distortion}, the worst-case multiplicative approximation guaranteed by any policy. For the setting with $s$ services, we establish lower bounds of $\\Omega(s)$ and $\\Omega(\\sqrt{s})$ on the distortions of any deterministic and randomised mechanisms, respectively. We complement these results by providing approximately truthful, measured by \\emph{incentive ratio}, deterministic and randomised policies based on random serial dictatorship which match our lower bounds. Our results show that there is a significant improvement if one considers the class of randomised policies.   Finally, we consider the online SBM setting with bandit feedback where each agent is initially unaware of her true preferences, and the planner must facilitate each agent in the learning of their preferences through the matching of services over time. We design an approximately truthful mechanism based on the Explore-then-Commit paradigm, which achieves logarithmic dynamic approximate regret."
                    },
                    {
                        "title": "Markov Decision Processes with Time-Varying Geometric Discounting",
                        "abstract": "Canonical models of Markov decision processes (MDPs) usually consider geometric discounting based on a constant discount factor. While this standard modeling approach has led to many elegant results, some recent studies indicate the necessity of modeling time-varying discounting in certain applications. This paper studies a model of infinite-horizon MDPs with time-varying discount factors. We take a game-theoretic perspective -- whereby each time step is treated as an independent decision maker with their own (fixed) discount factor -- and we study the subgame perfect equilibrium (SPE) of the resulting game as well as the related algorithmic problems. We present a constructive proof of the existence of an SPE and demonstrate the EXPTIME-hardness of computing an SPE. We also turn to the approximate notion of $\\epsilon$-SPE and show that an $\\epsilon$-SPE exists under milder assumptions. An algorithm is presented to compute an $\\epsilon$-SPE, of which an upper bound of the time complexity, as a function of the convergence property of the time-varying discount factor, is provided."
                    },
                    {
                        "title": "Corruption-Robust Offline Two-Player Zero-Sum Markov Games",
                        "abstract": "We study data corruption robustness in offline two-player zero-sum Markov games. Given a dataset of realized trajectories of two players, an adversary is allowed to modify an $\\epsilon$-fraction of it. The learner's goal is to identify an approximate Nash Equilibrium policy pair from the corrupted data. We consider this problem in linear Markov games under different degrees of data coverage and corruption. We start by providing an information-theoretic lower bound on the suboptimality gap of any learner. Next, we propose robust versions of the Pessimistic Minimax Value Iteration algorithm, both under coverage on the corrupted data and under coverage only on the clean data, and show that they achieve (near)-optimal suboptimality gap bounds with respect to $\\epsilon$. We note that we are the first to provide such a characterization of the problem of learning approximate Nash Equilibrium policies in offline two-player zero-sum Markov games under data corruption."
                    },
                    {
                        "title": "Implicit Poisoning Attacks in Two-Agent Reinforcement Learning: Adversarial Policies for Training-Time Attacks",
                        "abstract": "In targeted poisoning attacks, an attacker manipulates an agent-environment interaction to force the agent into adopting a policy of interest, called target policy. Prior work has primarily focused on attacks that modify standard MDP primitives, such as rewards or transitions. In this paper, we study targeted poisoning attacks in a two-agent setting where an attacker implicitly poisons the effective environment of one of the agents by modifying the policy of its peer. We develop an optimization framework for designing optimal attacks, where the cost of the attack measures how much the solution deviates from the assumed default policy of the peer agent. We further study the computational properties of this optimization framework. Focusing on a tabular setting, we show that in contrast to poisoning attacks based on MDP primitives (transitions and (unbounded) rewards), which are always feasible, it is NP-hard to determine the feasibility of implicit poisoning attacks. We provide characterization results that establish sufficient conditions for the feasibility of the attack problem, as well as an upper and a lower bound on the optimal cost of the attack. We propose two algorithmic approaches for finding an optimal adversarial policy: a model-based approach with tabular policies and a model-free approach with parametric/neural policies. We showcase the efficacy of the proposed algorithms through experiments."
                    }
                ]
            },
            "f638f946-be67-4c91-b50a-dafac4a9dd92": {
                "pk": "f638f946-be67-4c91-b50a-dafac4a9dd92",
                "name": "Long Tran-Thanh",
                "collaborators": [
                    "Archie Chapman",
                    "Alex Rogers",
                    "Nicholas R. Jennings",
                    "Anshuka Rangi",
                    "Haifeng Xu",
                    "Massimo Franceschetti"
                ],
                "domain": [
                    "Multi-Armed Bandit",
                    "Data Poisoning",
                    "Algorithm Design",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Knapsack based Optimal Policies for Budget-Limited Multi-Armed Bandits",
                        "abstract": "In budget-limited multi-armed bandit (MAB) problems, the learner's actions are costly and constrained by a fixed budget. Consequently, an optimal exploitation policy may not be to pull the optimal arm repeatedly, as is the case in other variants of MAB, but rather to pull the sequence of different arms that maximises the agent's total reward within the budget. This difference from existing MABs means that new approaches to maximising the total reward are required. Given this, we develop two pulling policies, namely: (i) KUBE; and (ii) fractional KUBE. Whereas the former provides better performance up to 40% in our experimental settings, the latter is computationally less expensive. We also prove logarithmic upper bounds for the regret of both policies, and show that these bounds are asymptotically optimal (i.e. they only differ from the best possible regret by a constant factor)."
                    },
                    {
                        "title": "Saving Stochastic Bandits from Poisoning Attacks via Limited Data Verification",
                        "abstract": "We study bandit algorithms under data poisoning attacks in a bounded reward setting. We consider a strong attacker model in which the attacker can observe both the selected actions and their corresponding rewards and can contaminate the rewards with additive noise. We show that any bandit algorithm with regret $O(\\log T)$ can be forced to suffer a regret $\\Omega(T)$ with an expected amount of contamination $O(\\log T)$. This amount of contamination is also necessary, as we prove that there exists an $O(\\log T)$ regret bandit algorithm, specifically the classical UCB, that requires $\\Omega(\\log T)$ amount of contamination to suffer regret $\\Omega(T)$. To combat such attacks, our second main contribution is to propose verification based mechanisms, which use limited verification to access a limited number of uncontaminated rewards. In particular, for the case of unlimited verifications, we show that with $O(\\log T)$ expected number of verifications, a simple modified version of the ETC type bandit algorithm can restore the order optimal $O(\\log T)$ regret irrespective of the amount of contamination used by the attacker. We also provide a UCB-like verification scheme, called Secure-UCB, that also enjoys full recovery from any attacks, also with $O(\\log T)$ expected number of verifications. To derive a matching lower bound on the number of verifications, we prove that for any order-optimal bandit algorithm, this number of verifications $\\Omega(\\log T)$ is necessary to recover the order-optimal regret. On the other hand, when the number of verifications is bounded above by a budget $B$, we propose a novel algorithm, Secure-BARBAR, which provably achieves $O(\\min\\{C,T/\\sqrt{B} \\})$ regret with high probability against weak attackers where $C$ is the total amount of contamination by the attacker, which breaks the known $\\Omega(C)$ lower bound of the non-verified setting if $C$ is large."
                    }
                ]
            }
        }
    },
    "2410.14574": {
        "paper_data": {
            "title": "MomentumSMoE: Integrating Momentum into Sparse Mixture of Experts",
            "url": "http://arxiv.org/abs/2410.14574v1",
            "arxiv_id": "2410.14574",
            "authors": [
                "Rachel S. Y. Teo",
                "Tan M. Nguyen"
            ],
            "abstract": "Sparse Mixture of Experts (SMoE) has become the key to unlocking unparalleled scalability in deep learning. SMoE has the potential to exponentially increase parameter count while maintaining the efficiency of the model by only activating a small subset of these parameters for a given sample. However, it has been observed that SMoE suffers from unstable training and has difficulty adapting to new distributions, leading to the model's lack of robustness to data contamination. To overcome these limitations, we first establish a connection between the dynamics of the expert representations in SMoEs and gradient descent on a multi-objective optimization problem. Leveraging our framework, we then integrate momentum into SMoE and propose a new family of SMoEs named MomentumSMoE. We theoretically prove and numerically demonstrate that MomentumSMoE is more stable and robust than SMoE. In particular, we verify the advantages of MomentumSMoE over SMoE on a variety of practical tasks including ImageNet-1K object recognition and WikiText-103 language modeling. We demonstrate the applicability of MomentumSMoE to many types of SMoE models, including those in the Sparse MoE model for vision (V-MoE) and the Generalist Language Model (GLaM). We also show that other advanced momentum-based optimization methods, such as Adam, can be easily incorporated into the MomentumSMoE framework for designing new SMoE models with even better performance, almost negligible additional computation cost, and simple implementations.",
            "introduction": "   1 Introduction  Scaling up deep models has demonstrated significant potential for enhancing the model\u2019s performance on a wide range of cognitive and machine learning tasks, ranging from large language model pre-training\u00a0[13, 58, 59, 30, 5, 51, 70] and vision understanding\u00a0[15, 2, 3, 39, 1, 40] to reinforcement learning\u00a0[6, 28] and scientific applications\u00a0[66, 74]. However, increasing the model\u2019s size requires a higher computational budget, which can be often challenging to meet. As a result, Sparse Mixture of Experts (SMoE) has been recently studied as an efficient approach to effectively scale up deep models. By modularizing the network and activating only subsets of experts for each input, SMoE maintains constant computational costs while increasing model complexity. This approach enables the development of billion-parameter models and achieves significant success in various applications, including machine translation\u00a0[35], image classification\u00a0[61], and speech recognition\u00a0[34].\u2020\u2020Please correspond to: rachel.tsy@u.nus.edu    1.1 Sparse Mixture of Experts  A MoE replaces a component in the layer of the model, for example, a feed-forward or convolutional layer, by a set of networks termed experts. This approach largely scales up the model but increases the computational cost. A SMoE inherits the extended model capacity from MoE but preserves the computational overhead by taking advantage of conditional computation. In particular, a SMoE consists of a router and E\ud835\udc38Eitalic_E expert networks, uisubscript\ud835\udc62\ud835\udc56u_{i}italic_u start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT, i=1,2,\u2026,E\ud835\udc5612\u2026\ud835\udc38i=1,2,\\dots,Eitalic_i = 1 , 2 , \u2026 , italic_E. For each input token \ud835\udc99t\u2208\u211dDsubscript\ud835\udc99\ud835\udc61superscript\u211d\ud835\udc37{\\bm{x}}_{t}\\in{\\mathbb{R}}^{D}bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT \u2208 blackboard_R start_POSTSUPERSCRIPT italic_D end_POSTSUPERSCRIPT at layer t\ud835\udc61titalic_t, the SMoE\u2019s router computes the affinity scores between \ud835\udc99tsubscript\ud835\udc99\ud835\udc61{\\bm{x}}_{t}bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT and each expert as gi\u2062(\ud835\udc99t)subscript\ud835\udc54\ud835\udc56subscript\ud835\udc99\ud835\udc61g_{i}({\\bm{x}}_{t})italic_g start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ( bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ), i=1,2,\u2026,E\ud835\udc5612\u2026\ud835\udc38i=1,2,\\dots,Eitalic_i = 1 , 2 , \u2026 , italic_E. In practice, we often choose the router g\u2062(\ud835\udc99t)=[g1\u2062(\ud835\udc99t),g2\u2062(\ud835\udc99t),\u2026,gE\u2062(\ud835\udc99t)]\u22a4=\ud835\udc7e\u2062\ud835\udc99+\ud835\udc83\ud835\udc54subscript\ud835\udc99\ud835\udc61superscriptsubscript\ud835\udc541subscript\ud835\udc99\ud835\udc61subscript\ud835\udc542subscript\ud835\udc99\ud835\udc61\u2026subscript\ud835\udc54\ud835\udc38subscript\ud835\udc99\ud835\udc61top\ud835\udc7e\ud835\udc99\ud835\udc83g({\\bm{x}}_{t})=[g_{1}({\\bm{x}}_{t}),g_{2}({\\bm{x}}_{t}),\\dots,g_{E}({\\bm{x}}_% {t})]^{\\top}={\\bm{W}}{\\bm{x}}+{\\bm{b}}italic_g ( bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) = [ italic_g start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) , italic_g start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ( bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) , \u2026 , italic_g start_POSTSUBSCRIPT italic_E end_POSTSUBSCRIPT ( bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) ] start_POSTSUPERSCRIPT \u22a4 end_POSTSUPERSCRIPT = bold_italic_W bold_italic_x + bold_italic_b, where \ud835\udc7e\u2208\u211dE\u00d7D\ud835\udc7esuperscript\u211d\ud835\udc38\ud835\udc37{\\bm{W}}\\in{\\mathbb{R}}^{E\\times D}bold_italic_W \u2208 blackboard_R start_POSTSUPERSCRIPT italic_E \u00d7 italic_D end_POSTSUPERSCRIPT and \ud835\udc83\u2208\u211dE\ud835\udc83superscript\u211d\ud835\udc38{\\bm{b}}\\in{\\mathbb{R}}^{E}bold_italic_b \u2208 blackboard_R start_POSTSUPERSCRIPT italic_E end_POSTSUPERSCRIPT. Then, a sparse gating function TopKTopK\\mathrm{TopK}roman_TopK is applied to select only K\ud835\udc3eKitalic_K experts with the greatest affinity scores. Here, we define the TopKTopK\\mathrm{TopK}roman_TopK function as:    TopK\u2062(gi):={gi,if\u00a0\u2062gi\u2062\u00a0is in the\u00a0\u2062K\u2062\u00a0largest elements of\u00a0\u2062g\u2212\u221e,otherwise.assignTopKsubscript\ud835\udc54\ud835\udc56casessubscript\ud835\udc54\ud835\udc56if\u00a0subscript\ud835\udc54\ud835\udc56\u00a0is in the\u00a0\ud835\udc3e\u00a0largest elements of\u00a0\ud835\udc54otherwiseotherwiseotherwise\\displaystyle\\mathrm{TopK}(g_{i}):=\\begin{cases}\\leavevmode\\nobreak\\ g_{i},% \\hskip 9.10509pt\\text{if }g_{i}\\text{ is in the }K\\text{ largest elements of }% g\\\\ -\\infty,\\hskip 4.26773pt\\text{otherwise}.\\end{cases}roman_TopK ( italic_g start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) := { start_ROW start_CELL italic_g start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT , if italic_g start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT is in the italic_K largest elements of italic_g end_CELL start_CELL end_CELL end_ROW start_ROW start_CELL - \u221e , otherwise . end_CELL start_CELL end_CELL end_ROW  (1)   The outputs from K\ud835\udc3eKitalic_K expert networks chosen by the router are then linearly combined as    \ud835\udc99t+1=\ud835\udc99t+\u2211i=1Esoftmax(TopK(gi(\ud835\udc99t))ui(\ud835\udc99t)=\ud835\udc99t+u(\ud835\udc99t),\\displaystyle{\\bm{x}}_{t+1}={\\bm{x}}_{t}+\\sum_{i=1}^{E}\\text{softmax}(\\mathrm{% TopK}(g_{i}({\\bm{x}}_{t}))u_{i}({\\bm{x}}_{t})={\\bm{x}}_{t}+u({\\bm{x}}_{t}),bold_italic_x start_POSTSUBSCRIPT italic_t + 1 end_POSTSUBSCRIPT = bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT + \u2211 start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_E end_POSTSUPERSCRIPT softmax ( roman_TopK ( italic_g start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ( bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) ) italic_u start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ( bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ) = bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT + italic_u ( bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT )",
            "references": [
                {
                    "title": "Sequential Modeling Enables Scalable Learning for Large Vision Models",
                    "abstract": "We introduce a novel sequential modeling approach which enables learning a Large Vision Model (LVM) without making use of any linguistic data. To do this, we define a common format, \u201cvisual sentences \u201d, in which we can represent raw images and videos as well as annotated data sources such as semantic segmentations and depth reconstructions with-out needing any meta-knowledge beyond the pixels. Once this wide variety of visual data (comprising 420 billion to-kens) is represented as sequences, the model can be trained to minimize a cross-entropy loss for next token prediction. By training across various scales of model architecture and data diversity, we provide empirical evidence that our models scale effectively. Many different vision tasks can be solved by designing suitable visual prompts at test time."
                },
                {
                    "title": "Robust Mixture-of-Expert Training for Convolutional Neural Networks",
                    "abstract": "Sparsely-gated Mixture of Expert (MoE), an emerging deep model architecture, has demonstrated a great promise to enable high-accuracy and ultra-efficient model inference. Despite the growing popularity of MoE, little work investigated its potential to advance convolutional neural networks (CNNs), especially in the plane of adversarial robustness. Since the lack of robustness has become one of the main hurdles for CNNs, in this paper we ask: How to adversarially robustify a CNN-based MoE model? Can we robustly train it like an ordinary CNN model? Our pilot study shows that the conventional adversarial training (AT) mechanism (developed for vanilla CNNs) no longer remains effective to robustify an MoE-CNN. To better understand this phenomenon, we dissect the robustness of an MoE-CNN into two dimensions: Robustness of routers (i.e., gating functions to select data-specific experts) and robustness of experts (i.e., the router-guided pathways defined by the subnetworks of the backbone CNN). Our analyses show that routers and experts are hard to adapt to each other in the vanilla AT. Thus, we propose a new router-expert alternating Adversarial training framework for MoE, termed AdvMoE. The effectiveness of our proposal is justified across 4 commonly-used CNN model architectures over 4 benchmark datasets. We find that AdvMoE achieves 1% ~ 4% adversarial robustness improvement over the original dense CNN, and enjoys the efficiency merit of sparsity-gated MoE, leading to more than 50% inference cost reduction. Codes are available at https://github.com/OPTML-Group/Robust-MoE-CNN."
                },
                {
                    "title": "From Sparse to Soft Mixtures of Experts",
                    "abstract": "Sparse mixture of expert architectures (MoEs) scale model capacity without significant increases in training or inference costs. Despite their success, MoEs suffer from a number of issues: training instability, token dropping, inability to scale the number of experts, or ineffective finetuning. In this work, we propose Soft MoE, a fully-differentiable sparse Transformer that addresses these challenges, while maintaining the benefits of MoEs. Soft MoE performs an implicit soft assignment by passing different weighted combinations of all input tokens to each expert. As in other MoEs, experts in Soft MoE only process a subset of the (combined) tokens, enabling larger model capacity (and performance) at lower inference cost. In the context of visual recognition, Soft MoE greatly outperforms dense Transformers (ViTs) and popular MoEs (Tokens Choice and Experts Choice). Furthermore, Soft MoE scales well: Soft MoE Huge/14 with 128 experts in 16 MoE layers has over 40x more parameters than ViT Huge/14, with only 2% increased inference time, and substantially better quality."
                },
                {
                    "title": "Brainformers: Trading Simplicity for Efficiency",
                    "abstract": "Transformers are central to recent successes in natural language processing and computer vision. Transformers have a mostly uniform backbone where layers alternate between feed-forward and self-attention in order to build a deep network. Here we investigate this design choice and find that more complex blocks that have different permutations of layer primitives can be more efficient. Using this insight, we develop a complex block, named Brainformer, that consists of a diverse sets of layers such as sparsely gated feed-forward layers, dense feed-forward layers, attention layers, and various forms of layer normalization and activation functions. Brainformer consistently outperforms the state-of-the-art dense and sparse Transformers, in terms of both quality and efficiency. A Brainformer model with 8 billion activated parameters per token demonstrates 2x faster training convergence and 5x faster step time compared to its GLaM counterpart. In downstream task evaluation, Brainformer also demonstrates a 3% higher SuperGLUE score with fine-tuning compared to GLaM with a similar number of activated parameters. Finally, Brainformer largely outperforms a Primer dense model derived with NAS with similar computation per token on fewshot evaluations."
                },
                {
                    "title": "A Theory on Adam Instability in Large-Scale Machine Learning",
                    "abstract": "We present a theory for the previously unexplained divergent behavior noticed in the training of large language models. We argue that the phenomenon is an artifact of the dominant optimization algorithm used for training, called Adam. We observe that Adam can enter a state in which the parameter update vector has a relatively large norm and is essentially uncorrelated with the direction of descent on the training loss landscape, leading to divergence. This artifact is more likely to be observed in the training of a deep model with a large batch size, which is the typical setting of large-scale language model training. To argue the theory, we present observations from the training runs of the language models of different scales: 7 billion, 30 billion, 65 billion, and 546 billion parameters."
                },
                {
                    "title": "In-context operator learning with data prompts for differential equation problems",
                    "abstract": "Significance This paper presents In-Context Operator Networks (ICON), a neural network approach that can learn new operators from prompted data during the inference stage without requiring any weight updates. Unlike existing methods that are limited to approximating specific equation solutions or operators and necessitate retraining for new problems, ICON trains a single neural network as an operator learner, eliminating the need for retraining or fine-tuning when encountering different problems. Numerical results demonstrate the efficacy of ICON in solving various types of differential equation problems and generalizing to operators beyond the training distribution. The proposed approach draws inspiration from successful in-context learning techniques used in natural language processing and has implications for artificial general intelligence in physical systems."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "On the Adversarial Robustness of Mixture of Experts",
                    "abstract": "Adversarial robustness is a key desirable property of neural networks. It has been empirically shown to be affected by their sizes, with larger networks being typically more robust. Recently, Bubeck and Sellke proved a lower bound on the Lipschitz constant of functions that fit the training data in terms of their number of parameters. This raises an interesting open question, do -- and can -- functions with more parameters, but not necessarily more computational cost, have better robustness? We study this question for sparse Mixture of Expert models (MoEs), that make it possible to scale up the model size for a roughly constant computational cost. We theoretically show that under certain conditions on the routing and the structure of the data, MoEs can have significantly smaller Lipschitz constants than their dense counterparts. The robustness of MoEs can suffer when the highest weighted experts for an input implement sufficiently different functions. We next empirically evaluate the robustness of MoEs on ImageNet using adversarial attacks and show they are indeed more robust than dense models with the same computational cost. We make key observations showing the robustness of MoEs to the choice of experts, highlighting the redundancy of experts in models trained in practice."
                },
                {
                    "title": "Momentum Transformer: Closing the Performance Gap Between Self-attention and Its Linearization",
                    "abstract": "Transformers have achieved remarkable success in sequence modeling and beyond but suffer from quadratic computational and memory complexities with respect to the length of the input sequence. Leveraging techniques include sparse and linear attention and hashing tricks; efficient transformers have been proposed to reduce the quadratic complexity of transformers but significantly degrade the accuracy. In response, we first interpret the linear attention and residual connections in computing the attention map as gradient descent steps. We then introduce momentum into these components and propose the \\emph{momentum transformer}, which utilizes momentum to improve the accuracy of linear transformers while maintaining linear memory and computational complexities. Furthermore, we develop an adaptive strategy to compute the momentum value for our model based on the optimal momentum for quadratic optimization. This adaptive momentum eliminates the need to search for the optimal momentum value and further enhances the performance of the momentum transformer. A range of experiments on both autoregressive and non-autoregressive tasks, including image generation and machine translation, demonstrate that the momentum transformer outperforms popular linear transformers in training efficiency and accuracy."
                },
                {
                    "title": "Sparse Mixture-of-Experts are Domain Generalizable Learners",
                    "abstract": "Human visual perception can easily generalize to out-of-distributed visual data, which is far beyond the capability of modern machine learning models. Domain generalization (DG) aims to close this gap, with existing DG methods mainly focusing on the loss function design. In this paper, we propose to explore an orthogonal direction, i.e., the design of the backbone architecture. It is motivated by an empirical finding that transformer-based models trained with empirical risk minimization (ERM) outperform CNN-based models employing state-of-the-art (SOTA) DG algorithms on multiple DG datasets. We develop a formal framework to characterize a network's robustness to distribution shifts by studying its architecture's alignment with the correlations in the dataset. This analysis guides us to propose a novel DG model built upon vision transformers, namely Generalizable Mixture-of-Experts (GMoE). Extensive experiments on DomainBed demonstrate that GMoE trained with ERM outperforms SOTA DG baselines by a large margin. Moreover, GMoE is complementary to existing DG methods and its performance is substantially improved when trained with DG algorithms."
                },
                {
                    "title": "On the Representation Collapse of Sparse Mixture of Experts",
                    "abstract": "Sparse mixture of experts provides larger model capacity while requiring a constant computational overhead. It employs the routing mechanism to distribute input tokens to the best-matched experts according to their hidden representations. However, learning such a routing mechanism encourages token clustering around expert centroids, implying a trend toward representation collapse. In this work, we propose to estimate the routing scores between tokens and experts on a low-dimensional hypersphere. We conduct extensive experiments on cross-lingual language model pre-training and fine-tuning on downstream tasks. Experimental results across seven multilingual benchmarks show that our method achieves consistent gains. We also present a comprehensive analysis on the representation and routing behaviors of our models. Our method alleviates the representation collapse issue and achieves more consistent routing than the baseline mixture-of-experts methods."
                },
                {
                    "title": "StableMoE: Stable Routing Strategy for Mixture of Experts",
                    "abstract": "The Mixture-of-Experts (MoE) technique can scale up the model size of Transformers with an affordable computational overhead. We point out that existing learning-to-route MoE methods suffer from the routing fluctuation issue, i.e., the target expert of the same input may change along with training, but only one expert will be activated for the input during inference. The routing fluctuation tends to harm sample efficiency because the same input updates different experts but only one is finally used. In this paper, we propose StableMoE with two training stages to address the routing fluctuation problem. In the first training stage, we learn a balanced and cohesive routing strategy and distill it into a lightweight router decoupled from the backbone model. In the second training stage, we utilize the distilled router to determine the token-to-expert assignment and freeze it for a stable routing strategy. We validate our method on language modeling and multilingual machine translation. The results show that StableMoE outperforms existing MoE methods in terms of both convergence speed and performance."
                },
                {
                    "title": "Sparsely Activated Mixture-of-Experts are Robust Multi-Task Learners",
                    "abstract": "Traditional multi-task learning (MTL) methods use dense networks that use the same set of shared weights across several different tasks. This often creates interference where two or more tasks compete to pull model parameters in different directions. In this work, we study whether sparsely activated Mixture-of-Experts (MoE) improve multi-task learning by specializing some weights for learning shared representations and using the others for learning task-specific information. To this end, we devise task-aware gating functions to route examples from different tasks to specialized experts which share subsets of network weights conditioned on the task. This results in a sparsely activated multi-task model with a large number of parameters, but with the same computational cost as that of a dense model. We demonstrate such sparse networks to improve multi-task learning along three key dimensions: (i) transfer to low-resource tasks from related tasks in the training mixture; (ii) sample-efficient generalization to tasks not seen during training by making use of task-aware routing from seen related tasks; (iii) robustness to the addition of unrelated tasks by avoiding catastrophic forgetting of existing tasks."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "Mixture-of-Experts with Expert Choice Routing",
                    "abstract": "Sparsely-activated Mixture-of-experts (MoE) models allow the number of parameters to greatly increase while keeping the amount of computation for a given token or a given sample unchanged. However, a poor expert routing strategy (e.g. one resulting in load imbalance) can cause certain experts to be under-trained, leading to an expert being under or over-specialized. Prior work allocates a fixed number of experts to each token using a top-k function regardless of the relative importance of different tokens. To address this, we propose a heterogeneous mixture-of-experts employing an expert choice method. Instead of letting tokens select the top-k experts, we have experts selecting the top-k tokens. As a result, each token can be routed to a variable number of experts and each expert can have a fixed bucket size. We systematically study pre-training speedups using the same computational resources of the Switch Transformer top-1 and GShard top-2 gating of prior work and find that our method improves training convergence time by more than 2x. For the same computational cost, our method demonstrates higher performance in fine-tuning 11 selected tasks in the GLUE and SuperGLUE benchmarks. For a smaller activation cost, our method outperforms the T5 dense model in 7 out of the 11 tasks."
                },
                {
                    "title": "ST-MoE: Designing Stable and Transferable Sparse Expert Models",
                    "abstract": "Scale has opened new frontiers in natural language processing -- but at a high cost. In response, Mixture-of-Experts (MoE) and Switch Transformers have been proposed as an energy efficient path to even larger and more capable language models. But advancing the state-of-the-art across a broad set of natural language tasks has been hindered by training instabilities and uncertain quality during fine-tuning. Our work focuses on these issues and acts as a design guide. We conclude by scaling a sparse model to 269B parameters, with a computational cost comparable to a 32B dense encoder-decoder Transformer (Stable and Transferable Mixture-of-Experts or ST-MoE-32B). For the first time, a sparse model achieves state-of-the-art performance in transfer learning, across a diverse set of tasks including reasoning (SuperGLUE, ARC Easy, ARC Challenge), summarization (XSum, CNN-DM), closed book question answering (WebQA, Natural Questions), and adversarially constructed tasks (Winogrande, ANLI R3)."
                },
                {
                    "title": "Unified Scaling Laws for Routed Language Models",
                    "abstract": "The performance of a language model has been shown to be effectively modeled as a power-law in its parameter count. Here we study the scaling behaviors of Routing Networks: architectures that conditionally use only a subset of their parameters while processing an input. For these models, parameter count and computational requirement form two independent axes along which an increase leads to better performance. In this work we derive and justify scaling laws defined on these two variables which generalize those known for standard language models and describe the performance of a wide range of routing architectures trained via three different techniques. Afterwards we provide two applications of these laws: first deriving an Effective Parameter Count along which all models scale at the same rate, and then using the scaling coefficients to give a quantitative comparison of the three routing techniques considered. Our analysis derives from an extensive evaluation of Routing Networks across five orders of magnitude of size, including models with hundreds of experts and hundreds of billions of parameters."
                },
                {
                    "title": "GLaM: Efficient Scaling of Language Models with Mixture-of-Experts",
                    "abstract": "Scaling language models with more data, compute and parameters has driven significant progress in natural language processing. For example, thanks to scaling, GPT-3 was able to achieve strong results on in-context learning tasks. However, training these large dense models requires significant amounts of computing resources. In this paper, we propose and develop a family of language models named GLaM (Generalist Language Model), which uses a sparsely activated mixture-of-experts architecture to scale the model capacity while also incurring substantially less training cost compared to dense variants. The largest GLaM has 1.2 trillion parameters, which is approximately 7x larger than GPT-3. It consumes only 1/3 of the energy used to train GPT-3 and requires half of the computation flops for inference, while still achieving better overall zero-shot and one-shot performance across 29 NLP tasks."
                },
                {
                    "title": "Building a great multi-lingual teacher with sparsely-gated mixture of experts for speech recognition",
                    "abstract": "The sparsely-gated Mixture of Experts (MoE) can magnify a network capacity with a little computational complexity. In this work, we investigate how multi-lingual Automatic Speech Recognition (ASR) networks can be scaled up with a simple routing algorithm in order to achieve better accuracy. More specifically, we apply the sparsely-gated MoE technique to two types of networks: Sequence-to-Sequence Transformer (S2S-T) and Transformer Transducer (T-T). We demonstrate through a set of ASR experiments on multiple language data that the MoE networks can reduce the relative word error rates by 16.3% and 4.6% with the S2S-T and T-T, respectively. Moreover, we thoroughly investigate the effect of the MoE on the T-T architecture in various conditions: streaming mode, non-streaming mode, the use of language ID and the label decoder with the MoE."
                },
                {
                    "title": "VLMo: Unified Vision-Language Pre-Training with Mixture-of-Modality-Experts",
                    "abstract": "We present a unified Vision-Language pretrained Model (VLMo) that jointly learns a dual encoder and a fusion encoder with a modular Transformer network. Specifically, we introduce Mixture-of-Modality-Experts (MoME) Transformer, where each block contains a pool of modality-specific experts and a shared self-attention layer. Because of the modeling flexibility of MoME, pretrained VLMo can be fine-tuned as a fusion encoder for vision-language classification tasks, or used as a dual encoder for efficient image-text retrieval. Moreover, we propose a stagewise pre-training strategy, which effectively leverages large-scale image-only and text-only data besides image-text pairs. Experimental results show that VLMo achieves state-of-the-art results on various vision-language tasks, including VQA, NLVR2 and image-text retrieval. The code and pretrained models are available at https://aka.ms/vlmo."
                },
                {
                    "title": "Heavy Ball Neural Ordinary Differential Equations",
                    "abstract": "We propose heavy ball neural ordinary differential equations (HBNODEs), leveraging the continuous limit of the classical momentum accelerated gradient descent, to improve neural ODEs (NODEs) training and inference. HBNODEs have two properties that imply practical advantages over NODEs: (i) The adjoint state of an HBNODE also satisfies an HBNODE, accelerating both forward and backward ODE solvers, thus significantly reducing the number of function evaluations (NFEs) and improving the utility of the trained models. (ii) The spectrum of HBNODEs is well structured, enabling effective learning of long-term dependencies from complex sequential data. We verify the advantages of HBNODEs over NODEs on benchmark tasks, including image classification, learning complex dynamics, and sequential modeling. Our method requires remarkably fewer forward and backward NFEs, is more accurate, and learns long-term dependencies more effectively than the other ODE-based neural network models. Code is available at \\url{https://github.com/hedixia/HeavyBallNODE}."
                },
                {
                    "title": "Taming Sparsely Activated Transformer with Stochastic Experts",
                    "abstract": "Sparsely activated models (SAMs), such as Mixture-of-Experts (MoE), can easily scale to have outrageously large amounts of parameters without significant increase in computational cost. However, SAMs are reported to be parameter inefficient such that larger models do not always lead to better performance. While most on-going research focuses on improving SAMs models by exploring methods of routing inputs to experts, our analysis reveals that such research might not lead to the solution we expect, i.e., the commonly-used routing methods based on gating mechanisms do not work better than randomly routing inputs to experts. In this paper, we propose a new expert-based model, THOR (Transformer witH StOchastic ExpeRts). Unlike classic expert-based models, such as the Switch Transformer, experts in THOR are randomly activated for each input during training and inference. THOR models are trained using a consistency regularized loss, where experts learn not only from training data but also from other experts as teachers, such that all the experts make consistent predictions. We validate the effectiveness of THOR on machine translation tasks. Results show that THOR models are more parameter efficient in that they significantly outperform the Transformer and MoE models across various settings. For example, in multilingual translation, THOR outperforms the Switch Transformer by 2 BLEU scores, and obtains the same BLEU score as that of a state-of-the-art MoE model that is 18 times larger. Our code is publicly available at: https://github.com/microsoft/Stochastic-Mixture-of-Experts."
                },
                {
                    "title": "BEiT: BERT Pre-Training of Image Transformers",
                    "abstract": "We introduce a self-supervised vision representation model BEiT, which stands for Bidirectional Encoder representation from Image Transformers. Following BERT developed in the natural language processing area, we propose a masked image modeling task to pretrain vision Transformers. Specifically, each image has two views in our pre-training, i.e, image patches (such as 16x16 pixels), and visual tokens (i.e., discrete tokens). We first\"tokenize\"the original image into visual tokens. Then we randomly mask some image patches and fed them into the backbone Transformer. The pre-training objective is to recover the original visual tokens based on the corrupted image patches. After pre-training BEiT, we directly fine-tune the model parameters on downstream tasks by appending task layers upon the pretrained encoder. Experimental results on image classification and semantic segmentation show that our model achieves competitive results with previous pre-training methods. For example, base-size BEiT achieves 83.2% top-1 accuracy on ImageNet-1K, significantly outperforming from-scratch DeiT training (81.8%) with the same setup. Moreover, large-size BEiT obtains 86.3% only using ImageNet-1K, even outperforming ViT-L with supervised pre-training on ImageNet-22K (85.2%). The code and pretrained models are available at https://aka.ms/beit."
                },
                {
                    "title": "Partial success in closing the gap between human and machine vision",
                    "abstract": "A few years ago, the first CNN surpassed human performance on ImageNet. However, it soon became clear that machines lack robustness on more challenging test cases, a major obstacle towards deploying machines\"in the wild\"and towards obtaining better computational models of human visual perception. Here we ask: Are we making progress in closing the gap between human and machine vision? To answer this question, we tested human observers on a broad range of out-of-distribution (OOD) datasets, recording 85,120 psychophysical trials across 90 participants. We then investigated a range of promising machine learning developments that crucially deviate from standard supervised CNNs along three axes: objective function (self-supervised, adversarially trained, CLIP language-image training), architecture (e.g. vision transformers), and dataset size (ranging from 1M to 1B). Our findings are threefold. (1.) The longstanding distortion robustness gap between humans and CNNs is closing, with the best models now exceeding human feedforward performance on most of the investigated OOD datasets. (2.) There is still a substantial image-level consistency gap, meaning that humans make different errors than models. In contrast, most models systematically agree in their categorisation errors, even substantially different ones like contrastive self-supervised vs. standard supervised models. (3.) In many cases, human-to-model consistency improves when training dataset size is increased by one to three orders of magnitude. Our results give reason for cautious optimism: While there is still much room for improvement, the behavioural difference between human and machine vision is narrowing. In order to measure future progress, 17 OOD datasets with image-level human behavioural data and evaluation code are provided as a toolbox and benchmark at: https://github.com/bethgelab/model-vs-human/"
                },
                {
                    "title": "Scaling Vision with Sparse Mixture of Experts",
                    "abstract": "Sparsely-gated Mixture of Experts networks (MoEs) have demonstrated excellent scalability in Natural Language Processing. In Computer Vision, however, almost all performant networks are\"dense\", that is, every input is processed by every parameter. We present a Vision MoE (V-MoE), a sparse version of the Vision Transformer, that is scalable and competitive with the largest dense networks. When applied to image recognition, V-MoE matches the performance of state-of-the-art networks, while requiring as little as half of the compute at inference time. Further, we propose an extension to the routing algorithm that can prioritize subsets of each input across the entire batch, leading to adaptive per-image compute. This allows V-MoE to trade-off performance and compute smoothly at test-time. Finally, we demonstrate the potential of V-MoE to scale vision models, and train a 15B parameter model that attains 90.35% on ImageNet."
                },
                {
                    "title": "Offline Reinforcement Learning as One Big Sequence Modeling Problem",
                    "abstract": "Reinforcement learning (RL) is typically concerned with estimating stationary policies or single-step models, leveraging the Markov property to factorize problems in time. However, we can also view RL as a generic sequence modeling problem, with the goal being to produce a sequence of actions that leads to a sequence of high rewards. Viewed in this way, it is tempting to consider whether high-capacity sequence prediction models that work well in other domains, such as natural-language processing, can also provide effective solutions to the RL problem. To this end, we explore how RL can be tackled with the tools of sequence modeling, using a Transformer architecture to model distributions over trajectories and repurposing beam search as a planning algorithm. Framing RL as sequence modeling problem simplifies a range of design decisions, allowing us to dispense with many of the components common in offline RL algorithms. We demonstrate the flexibility of this approach across long-horizon dynamics prediction, imitation learning, goal-conditioned RL, and offline RL. Further, we show that this approach can be combined with existing model-free algorithms to yield a state-of-the-art planner in sparse-reward, long-horizon tasks."
                },
                {
                    "title": "Decision Transformer: Reinforcement Learning via Sequence Modeling",
                    "abstract": "We introduce a framework that abstracts Reinforcement Learning (RL) as a sequence modeling problem. This allows us to draw upon the simplicity and scalability of the Transformer architecture, and associated advances in language modeling such as GPT-x and BERT. In particular, we present Decision Transformer, an architecture that casts the problem of RL as conditional sequence modeling. Unlike prior approaches to RL that fit value functions or compute policy gradients, Decision Transformer simply outputs the optimal actions by leveraging a causally masked Transformer. By conditioning an autoregressive model on the desired return (reward), past states, and actions, our Decision Transformer model can generate future actions that achieve the desired return. Despite its simplicity, Decision Transformer matches or exceeds the performance of state-of-the-art model-free offline RL baselines on Atari, OpenAI Gym, and Key-to-Door tasks."
                },
                {
                    "title": "BASE Layers: Simplifying Training of Large, Sparse Models",
                    "abstract": "We introduce a new balanced assignment of experts (BASE) layer for large language models that greatly simplifies existing high capacity sparse layers. Sparse layers can dramatically improve the efficiency of training and inference by routing each token to specialized expert modules that contain only a small fraction of the model parameters. However, it can be difficult to learn balanced routing functions that make full use of the available experts; existing approaches typically use routing heuristics or auxiliary expert-balancing loss functions. In contrast, we formulate token-to-expert allocation as a linear assignment problem, allowing an optimal assignment in which each expert receives an equal number of tokens. This optimal assignment scheme improves efficiency by guaranteeing balanced compute loads, and also simplifies training by not requiring any new hyperparameters or auxiliary losses. Code is publicly released at https://github.com/pytorch/fairseq/"
                },
                {
                    "title": "Complex Momentum for Optimization in Games",
                    "abstract": "We generalize gradient descent with momentum for optimization in differentiable games to have complex-valued momentum. We give theoretical motivation for our method by proving convergence on bilinear zero-sum games for simultaneous and alternating updates. Our method gives real-valued parameter updates, making it a drop-in replacement for standard optimizers. We empirically demonstrate that complex-valued momentum can improve convergence in realistic adversarial games - like generative adversarial networks - by showing we can find better solutions with an almost identical computational cost. We also show a practical generalization to a complex-valued Adam variant, which we use to train BigGAN to better inception scores on CIFAR-10."
                },
                {
                    "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity",
                    "abstract": "In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with outrageous numbers of parameters -- but a constant computational cost. However, despite several notable successes of MoE, widespread adoption has been hindered by complexity, communication costs and training instability -- we address these with the Switch Transformer. We simplify the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs. Our proposed training techniques help wrangle the instabilities and we show large sparse models may be trained, for the first time, with lower precision (bfloat16) formats. We design models based off T5-Base and T5-Large to obtain up to 7x increases in pre-training speed with the same computational resources. These improvements extend into multilingual settings where we measure gains over the mT5-Base version across all 101 languages. Finally, we advance the current scale of language models by pre-training up to trillion parameter models on the\"Colossal Clean Crawled Corpus\"and achieve a 4x speedup over the T5-XXL model."
                },
                {
                    "title": "Training data-efficient image transformers & distillation through attention",
                    "abstract": "Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption. In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data. More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding",
                    "abstract": "Neural network scaling has been critical for improving the model quality in many real-world machine learning applications with vast amounts of training data and compute. Although this trend of scaling is affirmed to be a sure-fire approach for better model quality, there are challenges on the path such as the computation cost, ease of programming, and efficient implementation on parallel devices. GShard is a module composed of a set of lightweight annotation APIs and an extension to the XLA compiler. It provides an elegant way to express a wide range of parallel computation patterns with minimal changes to the existing model code. GShard enabled us to scale up multilingual neural machine translation Transformer model with Sparsely-Gated Mixture-of-Experts beyond 600 billion parameters using automatic sharding. We demonstrate that such a giant model can efficiently be trained on 2048 TPU v3 accelerators in 4 days to achieve far superior quality for translation from 100 languages to English compared to the prior art."
                },
                {
                    "title": "The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization",
                    "abstract": "We introduce four new real-world distribution shift datasets consisting of changes in image style, image blurriness, geographic location, camera operation, and more. With our new datasets, we take stock of previously proposed methods for improving out-of-distribution robustness and put them to the test. We find that using larger models and artificial data augmentations can improve robustness on real-world distribution shifts, contrary to claims in prior work. We find improvements in artificial robustness benchmarks can transfer to real-world distribution shifts, contrary to claims in prior work. Motivated by our observation that data augmentations can help with real-world distribution shifts, we also introduce a new data augmentation method which advances the state-of-the-art and outperforms models pre-trained with 1000\u00d7 more labeled data. Overall we find that some methods consistently help with distribution shifts in texture and local image statistics, but these methods do not help with some other distribution shifts like geographic changes. Our results show that future research must study multiple distribution shifts simultaneously, as we demonstrate that no evaluated method consistently improves robustness."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "TextAttack: A Framework for Adversarial Attacks, Data Augmentation, and Adversarial Training in NLP",
                    "abstract": "While there has been substantial research using adversarial attacks to analyze NLP models, each attack is implemented in its own code repository. It remains challenging to develop NLP attacks and utilize them to improve model performance. This paper introduces TextAttack, a Python framework for adversarial attacks, data augmentation, and adversarial training in NLP. TextAttack builds attacks from four components: a goal function, a set of constraints, a transformation, and a search method. TextAttack\u2019s modular design enables researchers to easily construct attacks from combinations of novel and existing components. TextAttack provides implementations of 16 adversarial attacks from the literature and supports a variety of models and datasets, including BERT and other transformers, and all GLUE tasks. TextAttack also includes data augmentation and adversarial training modules for using components of adversarial attacks to improve model accuracy and robustness.TextAttack is democratizing NLP: anyone can try data augmentation and adversarial training on any model or dataset, with just a few lines of code. Code and tutorials are available at https://github.com/QData/TextAttack."
                },
                {
                    "title": "Scheduled Restart Momentum for Accelerated Stochastic Gradient Descent",
                    "abstract": "Stochastic gradient descent (SGD) with constant momentum and its variants such as Adam are the optimization algorithms of choice for training deep neural networks (DNNs). Since DNN training is incredibly computationally expensive, there is great interest in speeding up the convergence. Nesterov accelerated gradient (NAG) improves the convergence rate of gradient descent (GD) for convex optimization using a specially designed momentum; however, it accumulates error when an inexact gradient is used (such as in SGD), slowing convergence at best and diverging at worst. In this paper, we propose Scheduled Restart SGD (SRSGD), a new NAG-style scheme for training DNNs. SRSGD replaces the constant momentum in SGD by the increasing momentum in NAG but stabilizes the iterations by resetting the momentum to zero according to a schedule. Using a variety of models and benchmarks for image classification, we demonstrate that, in training DNNs, SRSGD significantly improves convergence and generalization; for instance in training ResNet200 for ImageNet classification, SRSGD achieves an error rate of 20.93% vs. the benchmark of 22.13%. These improvements become more significant as the network grows deeper. Furthermore, on both CIFAR and ImageNet, SRSGD reaches similar or even better error rates with significantly fewer training epochs compared to the SGD baseline."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Fantastic Generalization Measures and Where to Find Them",
                    "abstract": "Generalization of deep networks has been of great interest in recent years, resulting in a number of theoretically and empirically motivated complexity measures. However, most papers proposing such measures study only a small set of models, leaving open the question of whether the conclusion drawn from those experiments would remain valid in other settings. We present the first large scale study of generalization in deep networks. We investigate more then 40 complexity measures taken from both theoretical bounds and empirical studies. We train over 10,000 convolutional networks by systematically varying commonly used hyperparameters. Hoping to uncover potentially causal relationships between each measure and generalization, we analyze carefully controlled experiments and show surprising failures of some measures as well as promising measures for further research."
                },
                {
                    "title": "Momentum Contrast for Unsupervised Visual Representation Learning",
                    "abstract": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a moving-averaged encoder. This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning. MoCo provides competitive results under the common linear protocol on ImageNet classification. More importantly, the representations learned by MoCo transfer well to downstream tasks. MoCo can outperform its supervised pre-training counterpart in 7 detection/segmentation tasks on PASCAL VOC, COCO, and other datasets, sometimes surpassing it by large margins. This suggests that the gap between unsupervised and supervised representation learning has been largely closed in many vision tasks."
                },
                {
                    "title": "Improving Transformer Models by Reordering their Sublayers",
                    "abstract": "Multilayer transformer networks consist of interleaved self-attention and feedforward sublayers. Could ordering the sublayers in a different pattern lead to better performance? We generate randomly ordered transformers and train them with the language modeling objective. We observe that some of these models are able to achieve better performance than the interleaved baseline, and that those successful variants tend to have more self-attention at the bottom and more feedforward sublayers at the top. We propose a new transformer pattern that adheres to this property, the sandwich transformer, and show that it improves perplexity on multiple word-level and character-level language modeling benchmarks, at no cost in parameters, memory, or training time. However, the sandwich reordering pattern does not guarantee performance gains across every task, as we demonstrate on machine translation models. Instead, we suggest that further exploration of task-specific sublayer reorderings is needed in order to unlock additional gains."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "Natural Adversarial Examples",
                    "abstract": "We introduce two challenging datasets that reliably cause machine learning model performance to substantially degrade. The datasets are collected with a simple adversarial filtration technique to create datasets with limited spurious cues. Our datasets\u2019 real-world, unmodified examples transfer to various unseen models reliably, demonstrating that computer vision models have shared weaknesses. The first dataset is called IMAGENET-A and is like the ImageNet test set, but it is far more challenging for existing models. We also curate an adversarial out-of-distribution detection dataset called IMAGENET-O, which is the first out-of-distribution detection dataset created for ImageNet models. On IMAGENET-A a DenseNet-121 obtains around 2% accuracy, an accuracy drop of approximately 90%, and its out-of-distribution detection performance on IMAGENET-O is near random chance levels. We find that existing data augmentation techniques hardly boost performance, and using other public training datasets provides improvements that are limited. However, we find that improvements to computer vision architectures provide a promising path towards robust models."
                },
                {
                    "title": "Routing Networks and the Challenges of Modular and Compositional Computation",
                    "abstract": "Compositionality is a key strategy for addressing combinatorial complexity and the curse of dimensionality. Recent work has shown that compositional solutions can be learned and offer substantial gains across a variety of domains, including multi-task learning, language modeling, visual question answering, machine comprehension, and others. However, such models present unique challenges during training when both the module parameters and their composition must be learned jointly. In this paper, we identify several of these issues and analyze their underlying causes. Our discussion focuses on routing networks, a general approach to this problem, and examines empirically the interplay of these challenges and a variety of design decisions. In particular, we consider the effect of how the algorithm decides on module composition, how the algorithm updates the modules, and if the algorithm uses regularization."
                },
                {
                    "title": "Benchmarking Neural Network Robustness to Common Corruptions and Perturbations",
                    "abstract": "In this paper we establish rigorous benchmarks for image classifier robustness. Our first benchmark, ImageNet-C, standardizes and expands the corruption robustness topic, while showing which classifiers are preferable in safety-critical applications. Then we propose a new dataset called ImageNet-P which enables researchers to benchmark a classifier's robustness to common perturbations. Unlike recent robustness research, this benchmark evaluates performance on common corruptions and perturbations not worst-case adversarial perturbations. We find that there are negligible changes in relative corruption robustness from AlexNet classifiers to ResNet classifiers. Afterward we discover ways to enhance corruption and perturbation robustness. We even find that a bypassed adversarial defense provides substantial common perturbation robustness. Together our benchmarks may aid future work toward networks that robustly generalize."
                },
                {
                    "title": "Do ImageNet Classifiers Generalize to ImageNet?",
                    "abstract": "We build new test sets for the CIFAR-10 and ImageNet datasets. Both benchmarks have been the focus of intense research for almost a decade, raising the danger of overfitting to excessively re-used test sets. By closely following the original dataset creation processes, we test to what extent current classification models generalize to new data. We evaluate a broad range of models and find accuracy drops of 3% - 15% on CIFAR-10 and 11% - 14% on ImageNet. However, accuracy gains on the original test sets translate to larger gains on the new test sets. Our results suggest that the accuracy drops are not caused by adaptivity, but by the models' inability to generalize to slightly \"harder\" images than those found in the original test sets."
                },
                {
                    "title": "Optimization Algorithm Inspired Deep Neural Network Structure Design",
                    "abstract": "Deep neural networks have been one of the dominant machine learning approaches in recent years. Several new network structures are proposed and have better performance than the traditional feedforward neural network structure. Representative ones include the skip connection structure in ResNet and the dense connection structure in DenseNet. However, it still lacks a unified guidance for the neural network structure design. In this paper, we propose the hypothesis that the neural network structure design can be inspired by optimization algorithms and a faster optimization algorithm may lead to a better neural network structure. Specifically, we prove that the propagation in the feedforward neural network with the same linear transformation in different layers is equivalent to minimizing some function using the gradient descent algorithm. Based on this observation, we replace the gradient descent algorithm with the heavy ball algorithm and Nesterov's accelerated gradient descent algorithm, which are faster and inspire us to design new and better network structures. ResNet and DenseNet can be considered as two special cases of our framework. Numerical experiments on CIFAR-10, CIFAR-100 and ImageNet verify the advantage of our optimization algorithm inspired structures over ResNet and DenseNet."
                },
                {
                    "title": "Negative Momentum for Improved Game Dynamics",
                    "abstract": "Games generalize the single-objective optimization paradigm by introducing different objective functions for different players. Differentiable games often proceed by simultaneous or alternating gradient updates. In machine learning, games are gaining new importance through formulations like generative adversarial networks (GANs) and actor-critic systems. However, compared to single-objective optimization, game dynamics are more complex and less understood. In this paper, we analyze gradient-based methods with momentum on simple games. We prove that alternating updates are more stable than simultaneous updates. Next, we show both theoretically and empirically that alternating gradient updates with a negative momentum term achieves convergence in a difficult toy adversarial problem, but also on the notoriously difficult to train saturating GANs."
                },
                {
                    "title": "Improving Generalization Performance by Switching from Adam to SGD",
                    "abstract": "Despite superior training outcomes, adaptive optimization methods such as Adam, Adagrad or RMSprop have been found to generalize poorly compared to Stochastic gradient descent (SGD). These methods tend to perform well in the initial portion of training but are outperformed by SGD at later stages of training. We investigate a hybrid strategy that begins training with an adaptive method and switches to SGD when appropriate. Concretely, we propose SWATS, a simple strategy which switches from Adam to SGD when a triggering condition is satisfied. The condition we propose relates to the projection of Adam steps on the gradient subspace. By design, the monitoring process for this condition adds very little overhead and does not increase the number of hyperparameters in the optimizer. We report experiments on several standard benchmarks such as: ResNet, SENet, DenseNet and PyramidNet for the CIFAR-10 and CIFAR-100 data sets, ResNet on the tiny-ImageNet data set and language modeling with recurrent networks on the PTB and WT2 data sets. The results show that our strategy is capable of closing the generalization gap between SGD and Adam on a majority of the tasks."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "A Robust Accelerated Optimization Algorithm for Strongly Convex Functions",
                    "abstract": "This work proposes an accelerated first-order algorithm we call the Robust Momentum Method for optimizing smooth strongly convex functions. The algorithm has a single scalar parameter that can be tuned to trade off robustness to gradient noise versus worst-case convergence rate. At one extreme, the algorithm is faster than Nesterov's Fast Gradient Method by a constant factor but more fragile to noise. At the other extreme, the algorithm reduces to the Gradient Method and is very robust to noise. The algorithm design technique is inspired by methods from classical control theory and the resulting algorithm has a simple analytical form. Algorithm performance is verified on a series of numerical simulations in both noise-free and relative gradient noise cases."
                },
                {
                    "title": "Understanding the Learned Iterative Soft Thresholding Algorithm with matrix factorization",
                    "abstract": "Sparse coding is a core building block in many data analysis and machine learning pipelines. Typically it is solved by relying on generic optimization techniques, such as the Iterative Soft Thresholding Algorithm and its accelerated version (ISTA, FISTA). These methods are optimal in the class of first-order methods for non-smooth, convex functions. However, they do not exploit the particular structure of the problem at hand nor the input data distribution. An acceleration using neural networks, coined LISTA, was proposed in Gregor and Le Cun (2010), which showed empirically that one could achieve high quality estimates with few iterations by modifying the parameters of the proximal splitting appropriately. \nIn this paper we study the reasons for such acceleration. Our mathematical analysis reveals that it is related to a specific matrix factorization of the Gram kernel of the dictionary, which attempts to nearly diagonalise the kernel with a basis that produces a small perturbation of the $\\ell_1$ ball. When this factorization succeeds, we prove that the resulting splitting algorithm enjoys an improved convergence bound with respect to the non-adaptive version. Moreover, our analysis also shows that conditions for acceleration occur mostly at the beginning of the iterative process, consistent with numerical experiments. We further validate our analysis by showing that on dictionaries where this factorization does not exist, adaptive acceleration fails."
                },
                {
                    "title": "A Study and Comparison of Human and Deep Learning Recognition Performance under Visual Distortions",
                    "abstract": "Deep neural networks (DNNs) achieve excellent performance on standard classification tasks. However, under image quality distortions such as blur and noise, classification accuracy becomes poor. In this work, we compare the performance of DNNs with human subjects on distorted images. We show that, although DNNs perform better than or on par with humans on good quality images, DNN performance is still much lower than human performance on distorted images. We additionally find that there is little correlation in errors between DNNs and human subjects. This could be an indication that the internal representation of images are different between DNNs and the human visual system. These comparisons with human performance could be used to guide future development of more robust DNNs."
                },
                {
                    "title": "Computing Nonvacuous Generalization Bounds for Deep (Stochastic) Neural Networks with Many More Parameters than Training Data",
                    "abstract": "One of the defining properties of deep learning is that models are chosen to have many more parameters than available training data. In light of this capacity for overfitting, it is remarkable that simple algorithms like SGD reliably return solutions with low test error. One roadblock to explaining these phenomena in terms of implicit regularization, structural properties of the solution, and/or easiness of the data is that many learning bounds are quantitatively vacuous when applied to networks learned by SGD in this \"deep learning\" regime. Logically, in order to explain generalization, we need nonvacuous bounds. We return to an idea by Langford and Caruana (2001), who used PAC-Bayes bounds to compute nonvacuous numerical bounds on generalization error for stochastic two-layer two-hidden-unit neural networks via a sensitivity analysis. By optimizing the PAC-Bayes bound directly, we are able to extend their approach and obtain nonvacuous generalization bounds for deep stochastic neural network classifiers with millions of parameters trained on only tens of thousands of examples. We connect our findings to recent and old work on flat minima and MDL-based explanations of generalization."
                },
                {
                    "title": "Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer",
                    "abstract": "The capacity of a neural network to absorb information is limited by its number of parameters. Conditional computation, where parts of the network are active on a per-example basis, has been proposed in theory as a way of dramatically increasing model capacity without a proportional increase in computation. In practice, however, there are significant algorithmic and performance challenges. In this work, we address these challenges and finally realize the promise of conditional computation, achieving greater than 1000x improvements in model capacity with only minor losses in computational efficiency on modern GPU clusters. We introduce a Sparsely-Gated Mixture-of-Experts layer (MoE), consisting of up to thousands of feed-forward sub-networks. A trainable gating network determines a sparse combination of these experts to use for each example. We apply the MoE to the tasks of language modeling and machine translation, where model capacity is critical for absorbing the vast quantities of knowledge available in the training corpora. We present model architectures in which a MoE with up to 137 billion parameters is applied convolutionally between stacked LSTM layers. On large language modeling and machine translation benchmarks, these models achieve significantly better results than state-of-the-art at lower computational cost."
                },
                {
                    "title": "Pointer Sentinel Mixture Models",
                    "abstract": "Recent neural network sequence models with softmax classifiers have achieved their best language modeling performance only with very large hidden states and large vocabularies. Even then they struggle to predict rare or unseen words even if the context makes the prediction unambiguous. We introduce the pointer sentinel mixture architecture for neural sequence models which has the ability to either reproduce a word from the recent context or produce a word from a standard softmax classifier. Our pointer sentinel-LSTM model achieves state of the art language modeling performance on the Penn Treebank (70.9 perplexity) while using far fewer parameters than a standard softmax LSTM. In order to evaluate how well language models can exploit longer contexts and deal with more realistic vocabularies and larger corpora we also introduce the freely available WikiText corpus."
                },
                {
                    "title": "On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima",
                    "abstract": "The stochastic gradient descent (SGD) method and its variants are algorithms of choice for many Deep Learning tasks. These methods operate in a small-batch regime wherein a fraction of the training data, say $32$-$512$ data points, is sampled to compute an approximation to the gradient. It has been observed in practice that when using a larger batch there is a degradation in the quality of the model, as measured by its ability to generalize. We investigate the cause for this generalization drop in the large-batch regime and present numerical evidence that supports the view that large-batch methods tend to converge to sharp minimizers of the training and testing functions - and as is well known, sharp minima lead to poorer generalization. In contrast, small-batch methods consistently converge to flat minimizers, and our experiments support a commonly held view that this is due to the inherent noise in the gradient estimation. We discuss several strategies to attempt to help large-batch methods eliminate this generalization gap."
                },
                {
                    "title": "On the importance of initialization and momentum in deep learning",
                    "abstract": "Deep and recurrent neural networks (DNNs and RNNs respectively) are powerful models that were considered to be almost impossible to train using stochastic gradient descent with momentum. In this paper, we show that when stochastic gradient descent with momentum uses a well-designed random initialization and a particular type of slowly increasing schedule for the momentum parameter, it can train both DNNs and RNNs (on datasets with long-term dependencies) to levels of performance that were previously achievable only with Hessian-Free optimization. We find that both the initialization and the momentum are crucial since poorly initialized networks cannot be trained with momentum and well-initialized networks perform markedly worse when the momentum is absent or poorly tuned. \n \nOur success training these models suggests that previous attempts to train deep and recurrent neural networks from random initializations have likely failed due to poor initialization schemes. Furthermore, carefully tuned momentum methods suffice for dealing with the curvature issues in deep and recurrent network training objectives without the need for sophisticated second-order methods."
                },
                {
                    "title": "Improving Transformer with an Admixture of Attention Heads",
                    "abstract": "Transformers with multi-head self-attention have achieved remarkable success in sequence modeling and beyond. However, they suffer from high computational and memory complexities for computing the attention matrix at each head. Recently, it has been shown that those attention matrices lie on a low-dimensional manifold and, thus, are redundant. We propose the Transformer with a Finite Admixture of Shared Heads (FiSHformers), a novel class of efficient and flexible transformers that allow the sharing of attention matrices between attention heads. At the core of FiSHformer is a novel finite admixture model of shared heads (FiSH) that samples attention matrices from a set of global attention matrices. The number of global attention matrices is much smaller than the number of local attention matrices generated. FiSHformers directly learn these global attention matrices rather than the local ones as in other transformers, thus significantly improving the computational and memory efficiency of the model. We empirically verify the advantages of the FiSHformer over the baseline transformers in a wide range of practical applications including language modeling, machine translation, and image classification. On the WikiText-103, IWSLT\u201914 De-En and WMT\u201914"
                },
                {
                    "title": "Improving Neural Ordinary Differential Equations with Nesterov's Accelerated Gradient Method",
                    "abstract": "We propose the Nesterov neural ordinary differential equations (NesterovNODEs), whose layers solve the second-order ordinary differential equations (ODEs) limit of Nesterov\u2019s accelerated gradient (NAG) method, and a generalization called GNesterovNODEs. Taking the advantage of the convergence rate O (1 /k 2 ) of the NAG scheme, GNesterovNODEs speed up training and inference by reducing the number of function evaluations (NFEs) needed to solve the ODEs. We also prove that the adjoint state of a GNesterovNODEs also satisfies a GNesterovNODEs, thus accelerating both forward and backward ODE solvers and allowing the model to be scaled up for large-scale tasks. We empirically corroborate the advantage of GNesterovNODEs on a wide range of practical applications, including point cloud separation, image classification, and sequence modeling. Compared to NODEs, GNesterovNODEs require a significantly smaller number of NFEs while achieving better accuracy across our experiments."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Sharpness, Restart and Acceleration",
                    "abstract": "The \u0141ojasievicz inequality shows that sharpness bounds on the minimum of convex optimization problems hold almost generically. Here, we show that sharpness directly controls the performance of restart schemes. The constants quantifying sharpness are of course unobservable, but we show that optimal restart strategies are fairly robust, and searching for the best scheme only increases the complexity by a logarithmic factor compared to the optimal bound. Overall then, restart schemes generically accelerate accelerated methods"
                },
                {
                    "title": "A Fast Iterative Shrinkage-Thresholding Algorithm for Linear Inverse Problems",
                    "abstract": "We consider the class of iterative shrinkage-thresholding algorithms (ISTA) for solving linear inverse problems arising in signal/image processing. This class of methods, which can be viewed as an extension of the classical gradient algorithm, is attractive due to its simplicity and thus is adequate for solving large-scale problems even with dense matrix data. However, such methods are also known to converge quite slowly. In this paper we present a new fast iterative shrinkage-thresholding algorithm (FISTA) which preserves the computational simplicity of ISTA but with a global rate of convergence which is proven to be significantly better, both theoretically and practically. Initial promising numerical results for wavelet-based image deblurring demonstrate the capabilities of FISTA which is shown to be faster than ISTA by several orders of magnitude."
                },
                {
                    "title": "Towards Foundation Models for Scienti\ufb01c Machine Learning: Characterizing Scaling and Transfer Behavior",
                    "abstract": "Pre-trained machine learning (ML) models have shown great performance for a wide range of applications, in particular in natural language processing (NLP) and computer vision (CV). Here, we study how pre-training could be used for scienti\ufb01c machine learning (SciML) applications, speci\ufb01cally in the context of transfer learning. We study the transfer behavior of these models as (i) the pre-trained model size is scaled, (ii) the downstream training dataset size is scaled, (iii) the physics parameters are systematically pushed out of distribution, and (iv) how a single model pre-trained on a mixture of different physics problems can be adapted to various downstream applications. We \ufb01nd that\u2014when \ufb01ne-tuned appropriately\u2014transfer learning can help reach desired accuracy levels with orders of magnitude fewer downstream examples (across different tasks that can even be out-of-distribution) than training from scratch, with consistent behaviour across a wide range of downstream examples. We also \ufb01nd that \ufb01ne-tuning these models yields more performance gains as model size increases, compared to training from scratch on new downstream tasks. These results hold for a broad range of PDE learning tasks. All in all, our results demonstrate the potential of the \u201cpre-train and \ufb01ne-tune\u201d paradigm for SciML problems, demonstrating a path towards building SciML foundation models. Our code is available as open-source at [1]."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CL",
                "cs.CV",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively scale deep learning models using Sparse Mixture of Experts (SMoE) while maintaining computational efficiency?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing demand for larger and more complex models in various applications, such as natural language processing, computer vision, and reinforcement learning. By improving the efficiency of model scaling through SMoE, we can enable researchers to develop billion-parameter models without prohibitive computational costs. This advancement could lead to breakthroughs in AI capabilities, fostering innovation in practical applications like machine translation, image classification, and speech recognition. Furthermore, it may inspire future research into more efficient architectures and training methodologies, ultimately pushing the boundaries of what is possible in machine learning.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of balancing model size and computational efficiency. Naive approaches that simply increase model parameters often lead to exponential growth in computational requirements, making them impractical for real-world applications. Additionally, the design of effective routing mechanisms to select the most relevant experts for each input is non-trivial and requires sophisticated algorithms to ensure optimal performance. Technical obstacles include managing the sparsity of expert activation, ensuring robust training of the selected experts, and maintaining model interpretability. Theoretical challenges involve understanding the trade-offs between model capacity and generalization, which are critical for achieving high performance.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either scaling models or optimizing computational efficiency, but rarely have these two aspects been integrated effectively. Limitations in existing solutions include a lack of robust routing mechanisms and insufficient exploration of conditional computation strategies. Barriers such as the complexity of designing effective gating functions and the computational overhead associated with training large models have hindered progress. Our approach differs by leveraging advanced routing algorithms and conditional computation techniques that allow for dynamic expert selection, thus improving upon prior work by providing a more scalable and efficient framework for deep learning models.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves implementing a Sparse Mixture of Experts (SMoE) architecture that includes a router for expert selection and a set of expert networks. We will utilize a diverse dataset relevant to the target applications, such as machine translation or image classification, and evaluate the model"
            }
        },
        "author_data": {
            "0640dd95-c3fc-45c9-866f-362a6975bb4e": {
                "pk": "0640dd95-c3fc-45c9-866f-362a6975bb4e",
                "name": "Rachel S. Y. Teo",
                "collaborators": [
                    "Tan M. Nguyen"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Computer Vision",
                    "Deep Learning",
                    "Attention Mechanism"
                ],
                "publications": [
                    {
                        "title": "Unveiling the Hidden Structure of Self-Attention via Kernel Principal Component Analysis",
                        "abstract": "The remarkable success of transformers in sequence modeling tasks, spanning various applications in natural language processing and computer vision, is attributed to the critical role of self-attention. Similar to the development of most deep learning models, the construction of these attention mechanisms rely on heuristics and experience. In our work, we derive self-attention from kernel principal component analysis (kernel PCA) and show that self-attention projects its query vectors onto the principal component axes of its key matrix in a feature space. We then formulate the exact formula for the value matrix in self-attention, theoretically and empirically demonstrating that this value matrix captures the eigenvectors of the Gram matrix of the key vectors in self-attention. Leveraging our kernel PCA framework, we propose Attention with Robust Principal Components (RPC-Attention), a novel class of robust attention that is resilient to data contamination. We empirically demonstrate the advantages of RPC-Attention over softmax attention on the ImageNet-1K object classification, WikiText-103 language modeling, and ADE20K image segmentation task."
                    }
                ]
            },
            "f13acf06-0c4f-42fa-adca-9029091f4ddb": {
                "pk": "f13acf06-0c4f-42fa-adca-9029091f4ddb",
                "name": "Tan M. Nguyen",
                "collaborators": [
                    "Tam Nguyen",
                    "Tuan Nguyen",
                    "Vinh Nguyen",
                    "Richard G. Baraniuk",
                    "Rachel S. Y. Teo",
                    "Hirotada Honda",
                    "Takashi Sano",
                    "Shugo Nakamura",
                    "C\u00e9sar A. Uribe"
                ],
                "domain": [
                    "Transformers",
                    "Graph Neural Network",
                    "Robustness",
                    "Regularization"
                ],
                "publications": [
                    {
                        "title": "p-Laplacian Transformer",
                        "abstract": "$p$-Laplacian regularization, rooted in graph and image signal processing, introduces a parameter $p$ to control the regularization effect on these data. Smaller values of $p$ promote sparsity and interpretability, while larger values encourage smoother solutions. In this paper, we first show that the self-attention mechanism obtains the minimal Laplacian regularization ($p=2$) and encourages the smoothness in the architecture. However, the smoothness is not suitable for the heterophilic structure of self-attention in transformers where attention weights between tokens that are in close proximity and non-close ones are assigned indistinguishably. From that insight, we then propose a novel class of transformers, namely the $p$-Laplacian Transformer (p-LaT), which leverages $p$-Laplacian regularization framework to harness the heterophilic features within self-attention layers. In particular, low $p$ values will effectively assign higher attention weights to tokens that are in close proximity to the current token being processed. We empirically demonstrate the advantages of p-LaT over the baseline transformers on a wide range of benchmark datasets."
                    },
                    {
                        "title": "Unveiling the Hidden Structure of Self-Attention via Kernel Principal Component Analysis",
                        "abstract": "The remarkable success of transformers in sequence modeling tasks, spanning various applications in natural language processing and computer vision, is attributed to the critical role of self-attention. Similar to the development of most deep learning models, the construction of these attention mechanisms rely on heuristics and experience. In our work, we derive self-attention from kernel principal component analysis (kernel PCA) and show that self-attention projects its query vectors onto the principal component axes of its key matrix in a feature space. We then formulate the exact formula for the value matrix in self-attention, theoretically and empirically demonstrating that this value matrix captures the eigenvectors of the Gram matrix of the key vectors in self-attention. Leveraging our kernel PCA framework, we propose Attention with Robust Principal Components (RPC-Attention), a novel class of robust attention that is resilient to data contamination. We empirically demonstrate the advantages of RPC-Attention over softmax attention on the ImageNet-1K object classification, WikiText-103 language modeling, and ADE20K image segmentation task."
                    },
                    {
                        "title": "Mitigating Over-smoothing in Transformers via Regularized Nonlocal Functionals",
                        "abstract": "Transformers have achieved remarkable success in a wide range of natural language processing and computer vision applications. However, the representation capacity of a deep transformer model is degraded due to the over-smoothing issue in which the token representations become identical when the model's depth grows. In this work, we show that self-attention layers in transformers minimize a functional which promotes smoothness, thereby causing token uniformity. We then propose a novel regularizer that penalizes the norm of the difference between the smooth output tokens from self-attention and the input tokens to preserve the fidelity of the tokens. Minimizing the resulting regularized energy functional, we derive the Neural Transformer with a Regularized Nonlocal Functional (NeuTRENO), a novel class of transformer models that can mitigate the over-smoothing issue. We empirically demonstrate the advantages of NeuTRENO over the baseline transformers and state-of-the-art methods in reducing the over-smoothing of token representations on various practical tasks, including object classification, image segmentation, and language modeling."
                    },
                    {
                        "title": "From Coupled Oscillators to Graph Neural Networks: Reducing Over-smoothing via a Kuramoto Model-based Approach",
                        "abstract": "We propose the Kuramoto Graph Neural Network (KuramotoGNN), a novel class of continuous-depth graph neural networks (GNNs) that employs the Kuramoto model to mitigate the over-smoothing phenomenon, in which node features in GNNs become indistinguishable as the number of layers increases. The Kuramoto model captures the synchronization behavior of non-linear coupled oscillators. Under the view of coupled oscillators, we first show the connection between Kuramoto model and basic GNN and then over-smoothing phenomenon in GNNs can be interpreted as phase synchronization in Kuramoto model. The KuramotoGNN replaces this phase synchronization with frequency synchronization to prevent the node features from converging into each other while allowing the system to reach a stable synchronized state. We experimentally verify the advantages of the KuramotoGNN over the baseline GNNs and existing methods in reducing over-smoothing on various graph deep learning benchmark tasks."
                    },
                    {
                        "title": "PIDformer: Transformer Meets Control Theory",
                        "abstract": "In this work, we address two main shortcomings of transformer architectures: input corruption and rank collapse in their output representation. We unveil self-attention as an autonomous state-space model that inherently promotes smoothness in its solutions, leading to lower-rank outputs and diminished representation capacity. Moreover, the steady-state solution of the model is sensitive to input perturbations. We incorporate a Proportional-Integral-Derivative (PID) closed-loop feedback control system with a reference point into the model to improve robustness and representation capacity. This integration aims to preserve high-frequency details while bolstering model stability, rendering it more noise-resilient. The resulting controlled state-space model is theoretically proven robust and adept at addressing the rank collapse. Motivated by this control framework, we derive a novel class of transformers, PID-controlled Transformer (PIDformer), aimed at improving robustness and mitigating the rank-collapse issue inherent in softmax transformers. We empirically evaluate the model for advantages and robustness against baseline transformers across various practical tasks, including object classification, image segmentation, and language modeling."
                    }
                ]
            }
        }
    },
    "2401.17992": {
        "paper_data": {
            "title": "Multilinear Operator Networks",
            "url": "http://arxiv.org/abs/2401.17992v1",
            "arxiv_id": "2401.17992",
            "authors": [
                "Yixin Cheng",
                "Grigorios G. Chrysos",
                "Markos Georgopoulos",
                "Volkan Cevher"
            ],
            "abstract": "Despite the remarkable capabilities of deep neural networks in image recognition, the dependence on activation functions remains a largely unexplored area and has yet to be eliminated. On the other hand, Polynomial Networks is a class of models that does not require activation functions, but have yet to perform on par with modern architectures. In this work, we aim close this gap and propose MONet, which relies solely on multilinear operators. The core layer of MONet, called Mu-Layer, captures multiplicative interactions of the elements of the input token. MONet captures high-degree interactions of the input elements and we demonstrate the efficacy of our approach on a series of image recognition and scientific computing benchmarks. The proposed model outperforms prior polynomial networks and performs on par with modern architectures. We believe that MONet can inspire further research on models that use entirely multilinear operations.",
            "introduction": " introduction to the dataset we use and the overview statistics of datasets are shown in Methods. The best performance is marked in bold. Note that * in the table uses PyTorch default initialization, it depends on the layer type. For linear layer of shape (out, in ), the values are initialized from U(\u2212\u221a k,\u221a k), where k=1 in. Initialization Method Top-1 Acc (%) Top-5 Acc (%) Xavier Uniform (Glorot & Bengio, 2010) 88.85 99.38 Xavier Normal (Glorot & Bengio, 2010) 89.13 99.85 Kaiming Uniform (He et al., 2015) 88.56 98.70 Kaiming Normal (He et al., 2015) 88.72 99.46 Lecun Normal (LeCun et al., 2002) 88.95 99.51 Normal 88.06 99.34 Sparse (Martens et al., 2010) 88.18 99.48 Pytorch default* 88.37 99.49 28",
            "references": [
                {
                    "title": "Regularization of polynomial networks for image recognition",
                    "abstract": "Deep Neural Networks (DNNs) have obtained impressive performance across tasks, however they still remain as black boxes, e.g., hard to theoretically analyze. At the same time, Polynomial Networks (PNs) have emerged as an alternative method with a promising performance and improved interpretability but have yet to reach the performance of the powerful DNN baselines. In this work, we aim to close this performance gap. We introduce a class of PNs, which are able to reach the performance of ResNet across a range of six benchmarks. We demonstrate that strong regularization is critical and conduct an extensive study of the exact regularization schemes required to match performance. To further motivate the regularization schemes, we introduce D-PolyNets that achieve a higher- degree of expansion than previously proposed polynomial networks. D-PolyNets are more parameter-efficient while achieving a similar performance as other polynomial networks. We expect that our new models can lead to an understanding of the role of elementwise activation functions (which are no longer required for training PNs). The source code is available at https://github.com/grigorisg9gr/regularized_polynomials."
                },
                {
                    "title": "Polynomial Neural Fields for Subband Decomposition and Manipulation",
                    "abstract": "Neural fields have emerged as a new paradigm for representing signals, thanks to their ability to do it compactly while being easy to optimize. In most applications, however, neural fields are treated like black boxes, which precludes many signal manipulation tasks. In this paper, we propose a new class of neural fields called polynomial neural fields (PNFs). The key advantage of a PNF is that it can represent a signal as a composition of a number of manipulable and interpretable components without losing the merits of neural fields representation. We develop a general theoretical framework to analyze and design PNFs. We use this framework to design Fourier PNFs, which match state-of-the-art performance in signal representation tasks that use neural fields. In addition, we empirically demonstrate that Fourier PNFs enable signal manipulation applications such as texture transfer and scale-space interpolation. Code is available at https://github.com/stevenygd/PNF."
                },
                {
                    "title": "BiMLP: Compact Binary Architectures for Vision Multi-Layer Perceptrons",
                    "abstract": "This paper studies the problem of designing compact binary architectures for vision multi-layer perceptrons (MLPs). We provide extensive analysis on the difficulty of binarizing vision MLPs and find that previous binarization methods perform poorly due to limited capacity of binary MLPs. In contrast with the traditional CNNs that utilizing convolutional operations with large kernel size, fully-connected (FC) layers in MLPs can be treated as convolutional layers with kernel size $1\\times1$. Thus, the representation ability of the FC layers will be limited when being binarized, and places restrictions on the capability of spatial mixing and channel mixing on the intermediate features. To this end, we propose to improve the performance of binary MLP (BiMLP) model by enriching the representation ability of binary FC layers. We design a novel binary block that contains multiple branches to merge a series of outputs from the same stage, and also a universal shortcut connection that encourages the information flow from the previous stage. The downsampling layers are also carefully designed to reduce the computational complexity while maintaining the classification performance. Experimental results on benchmark dataset ImageNet-1k demonstrate the effectiveness of the proposed BiMLP models, which achieve state-of-the-art accuracy compared to prior binary CNNs. The MindSpore code is available at \\url{https://gitee.com/mindspore/models/tree/master/research/cv/BiMLP}."
                },
                {
                    "title": "Interpretable Polynomial Neural Ordinary Differential Equations",
                    "abstract": "Neural networks have the ability to serve as universal function approximators, but they are not interpretable and do not generalize well outside of their training region. Both of these issues are problematic when trying to apply standard neural ordinary differential equations (ODEs) to dynamical systems. We introduce the polynomial neural ODE, which is a deep polynomial neural network inside of the neural ODE framework. We demonstrate the capability of polynomial neural ODEs to predict outside of the training region, as well as to perform direct symbolic regression without using additional tools such as SINDy."
                },
                {
                    "title": "Scalable Interpretability via Polynomials",
                    "abstract": "Generalized Additive Models (GAMs) have quickly become the leading choice for inherently-interpretable machine learning. However, unlike uninterpretable methods such as DNNs, they lack expressive power and easy scalability, and are hence not a feasible alternative for real-world tasks. We present a new class of GAMs that use tensor rank decompositions of polynomials to learn powerful, {\\em inherently-interpretable} models. Our approach, titled Scalable Polynomial Additive Models (SPAM) is effortlessly scalable and models {\\em all} higher-order feature interactions without a combinatorial parameter explosion. SPAM outperforms all current interpretable approaches, and matches DNN/XGBoost performance on a series of real-world benchmarks with up to hundreds of thousands of features. We demonstrate by human subject evaluations that SPAMs are demonstrably more interpretable in practice, and are hence an effortless replacement for DNNs for creating interpretable and high-performance systems suitable for large-scale machine learning. Source code is available at https://github.com/facebookresearch/nbm-spam."
                },
                {
                    "title": "Patches Are All You Need?",
                    "abstract": "Although convolutional networks have been the dominant architecture for vision tasks for many years, recent experiments have shown that Transformer-based models, most notably the Vision Transformer (ViT), may exceed their performance in some settings. However, due to the quadratic runtime of the self-attention layers in Transformers, ViTs require the use of patch embeddings, which group together small regions of the image into single input features, in order to be applied to larger image sizes. This raises a question: Is the performance of ViTs due to the inherently-more-powerful Transformer architecture, or is it at least partly due to using patches as the input representation? In this paper, we present some evidence for the latter: specifically, we propose the ConvMixer, an extremely simple model that is similar in spirit to the ViT and the even-more-basic MLP-Mixer in that it operates directly on patches as input, separates the mixing of spatial and channel dimensions, and maintains equal size and resolution throughout the network. In contrast, however, the ConvMixer uses only standard convolutions to achieve the mixing steps. Despite its simplicity, we show that the ConvMixer outperforms the ViT, MLP-Mixer, and some of their variants for similar parameter counts and data set sizes, in addition to outperforming classical vision models such as the ResNet. Our code is available at https://github.com/locuslab/convmixer."
                },
                {
                    "title": "A Survey: Image Classification Models Based on Convolutional Neural Networks",
                    "abstract": "In recent years, image classification has become one of the hot spots in the field of artificial intelligence and computer vision. Image classification is a technique that simulates the human visual system to identify objects and classify images accurately, which has wide application in the fields of image segmentation, object tracking, face recognition, and etc. With the rapid development of deep learning architecture and hardware computing power, image classification methods based on convolutional neural networks have been rapidly developed. In this paper, we review architectures of convolutional neural networks, the process of image classification and put emphasis on the principles of advanced neural network models, quantization and pruning methods. We also analyze experiments based on different models and methods to compare their experimental data on different dimensions in different application situations. Additionally, we discuss some important issues and research dilemmas combined with current research progress. Finally, future research trends relating to classification methods are also provided."
                },
                {
                    "title": "DyTox: Transformers for Continual Learning with DYnamic TOken eXpansion",
                    "abstract": "Deep network architectures struggle to continually learn new tasks without forgetting the previous tasks. A recent trend indicates that dynamic architectures based on an ex-pansion of the parameters can reduce catastrophic forget-ting efficiently in continual learning. However, existing approaches often require a task identifier at test-time, need complex tuning to balance the growing number of parameters, and barely share any information across tasks. As a result, they struggle to scale to a large number of tasks without significant overhead. In this paper, we propose a transformer architecture based on a dedicated encoder/decoder framework. Critically, the encoder and decoder are shared among all tasks. Through a dynamic expansion of special tokens, we specialize each forward of our decoder network on a task distribution. Our strategy scales to a large number of tasks while having neg-ligible memory and time overheads due to strict control of the expansion of the parameters. Moreover, this efficient strategy doesn't need any hyperparameter tuning to control the network's expansion. Our model reaches excellent results on CIFAR100 and state-of-the-art performances on the large-scale ImageNet100 and ImageNet100 while having fewer parameters than concurrent dynamic frameworks.11Code is released at https://github.com/arthurdouillard/dytox."
                },
                {
                    "title": "Hire-MLP: Vision MLP via Hierarchical Rearrangement",
                    "abstract": "Previous vision MLPs such as MLP-Mixer and ResMLP accept linearly flattened image patches as input, making them inflexible for different input sizes and hard to capture spatial information. Such approach withholds MLPs from getting comparable performance with their transformer-based counterparts and prevents them from becoming a general backbone for computer vision. This paper presents Hire-MLP, a simple yet competitive vision MLP architecture via Hierarchical rearrangement, which contains two levels of rearrangements. Specifically, the inner-region rearrangement is proposed to capture local information inside a spatial region, and the cross-region rearrangement is proposed to enable information communication between different regions and capture global context by circularly shifting all tokens along spatial directions. Extensive experiments demonstrate the effectiveness of Hire-MLP as a versatile backbone for various vision tasks. In particular, Hire-MLP achieves competitive results on image classification, object detection and semantic segmentation tasks, e.g., 83.8% top-1 accuracy on ImageNet, 51.7% box AP and 44.8% mask AP on COCO val2017, and 49.9% mIoU on ADE20K, surpassing previous transformer-based and MLP-based models with better trade-off for accuracy and throughput."
                },
                {
                    "title": "S2-MLPv2: Improved Spatial-Shift MLP Architecture for Vision",
                    "abstract": "Recently, MLP-based vision backbones emerge. MLP-based vision architectures with less inductive bias achieve competitive performance in image recognition compared with CNNs and vision Transformers. Among them, spatial-shift MLP (S$^2$-MLP), adopting the straightforward spatial-shift operation, achieves better performance than the pioneering works including MLP-mixer and ResMLP. More recently, using smaller patches with a pyramid structure, Vision Permutator (ViP) and Global Filter Network (GFNet) achieve better performance than S$^2$-MLP. In this paper, we improve the S$^2$-MLP vision backbone. We expand the feature map along the channel dimension and split the expanded feature map into several parts. We conduct different spatial-shift operations on split parts. Meanwhile, we exploit the split-attention operation to fuse these split parts. Moreover, like the counterparts, we adopt smaller-scale patches and use a pyramid structure for boosting the image recognition accuracy. We term the improved spatial-shift MLP vision backbone as S$^2$-MLPv2. Using 55M parameters, our medium-scale model, S$^2$-MLPv2-Medium achieves an $83.6\\%$ top-1 accuracy on the ImageNet-1K benchmark using $224\\times 224$ images without self-attention and external training data."
                },
                {
                    "title": "CycleMLP: A MLP-like Architecture for Dense Prediction",
                    "abstract": "This paper presents a simple MLP-like architecture, CycleMLP, which is a versatile backbone for visual recognition and dense predictions. As compared to modern MLP architectures, e.g., MLP-Mixer, ResMLP, and gMLP, whose architectures are correlated to image size and thus are infeasible in object detection and segmentation, CycleMLP has two advantages compared to modern approaches. (1) It can cope with various image sizes. (2) It achieves linear computational complexity to image size by using local windows. In contrast, previous MLPs have $O(N^2)$ computations due to fully spatial connections. We build a family of models which surpass existing MLPs and even state-of-the-art Transformer-based models, e.g., Swin Transformer, while using fewer parameters and FLOPs. We expand the MLP-like models' applicability, making them a versatile backbone for dense prediction tasks. CycleMLP achieves competitive results on object detection, instance segmentation, and semantic segmentation. In particular, CycleMLP-Tiny outperforms Swin-Tiny by 1.3% mIoU on ADE20K dataset with fewer FLOPs. Moreover, CycleMLP also shows excellent zero-shot robustness on ImageNet-C dataset. Code is available at https://github.com/ShoufaChen/CycleMLP."
                },
                {
                    "title": "Poly-NL: Linear Complexity Non-local Layers with Polynomials",
                    "abstract": "Spatial self-attention layers, in the form of Non-Local blocks, introduce long-range dependencies in Convolutional Neural Networks by computing pairwise similarities among all possible positions. Such pairwise functions underpin the effectiveness of non-local layers, but also determine a complexity that scales quadratically with respect to the input size both in space and time. This is a severely limiting factor that practically hinders the applicability of non-local blocks to even moderately sized inputs. Previous works focused on reducing the complexity by modifying the underlying matrix operations, however in this work we aim to retain full expressiveness of non-local layers while keeping complexity linear. We overcome the efficiency limitation of non-local blocks by framing them as special cases of 3rd order polynomial functions. This fact enables us to formulate novel fast Non-Local blocks, capable of reducing the complexity from quadratic to linear with no loss in performance, by replacing any direct computation of pairwise similarities with element-wise multiplications. The proposed method, which we dub as\"Poly-NL\", is competitive with state-of-the-art performance across image recognition, instance segmentation, and face detection tasks, while having considerably less computational overhead."
                },
                {
                    "title": "Vision Permutator: A Permutable MLP-Like Architecture for Visual Recognition",
                    "abstract": "In this paper, we present Vision Permutator, a conceptually simple and data efficient MLP-like architecture for visual recognition. By realizing the importance of the positional information carried by 2D feature representations, unlike recent MLP-like models that encode the spatial information along the flattened spatial dimensions, Vision Permutator separately encodes the feature representations along the height and width dimensions with linear projections. This allows Vision Permutator to capture long-range dependencies and meanwhile avoid the attention building process in transformers. The outputs are then aggregated in a mutually complementing manner to form expressive representations. We show that our Vision Permutators are formidable competitors to convolutional neural networks (CNNs) and vision transformers. Without the dependence on spatial convolutions or attention mechanisms, Vision Permutator achieves 81.5% top-1 accuracy on ImageNet without extra large-scale training data (e.g., ImageNet-22k) using only 25M learnable parameters, which is much better than most CNNs and vision transformers under the same model size constraint. When scaling up to 88M, it attains 83.2% top-1 accuracy, greatly improving the performance of recent state-of-the-art MLP-like networks for visual recognition. We hope this work could encourage research on rethinking the way of encoding spatial information and facilitate the development of MLP-like models. PyTorch/MindSpore/Jittor code is available at https://github.com/Andrew-Qibin/VisionPermutator."
                },
                {
                    "title": "S2-MLP: Spatial-Shift MLP Architecture for Vision",
                    "abstract": "Recently, visual Transformer (ViT) and its following works abandon the convolution and exploit the self-attention operation, attaining a comparable or even higher accuracy than CNN. More recently, MLP-mixer abandons both the convolution and the self-attention operation, proposing an architecture containing only MLP layers. To achieve cross-patch communications, it devises an additional token-mixing MLP besides the channel-mixing MLP. It achieves promising results when training on an extremely large-scale dataset such as JFT-300M. But it cannot achieve as outstanding performance as its CNN and ViT counterparts when training on medium-scale datasets such as ImageNet-1K. The performance drop of MLP-mixer motivates us to rethink the token-mixing MLP. We discover that token-mixing operation in MLP-mixer is a variant of depthwise convolution with a global reception field and spatial-specific configuration. In this paper, we propose a novel pure MLP architecture, spatial-shift MLP (S2-MLP). Different from MLP-mixer, our S2-MLP only contains channel-mixing MLP. We devise a spatial-shift operation for achieving the communication between patches. It has a local reception field and is spatial-agnostic. Meanwhile, it is parameter-free and efficient for computation. The proposed S2-MLP attains higher recognition accuracy than MLP-mixer when training on ImageNet1K dataset. Meanwhile, S2-MLP accomplishes as excellent performance as ViT on ImageNet-1K dataset with considerably simpler architecture and fewer FLOPs and parameters."
                },
                {
                    "title": "SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers",
                    "abstract": "We present SegFormer, a simple, efficient yet powerful semantic segmentation framework which unifies Transformers with lightweight multilayer perception (MLP) decoders. SegFormer has two appealing features: 1) SegFormer comprises a novel hierarchically structured Transformer encoder which outputs multiscale features. It does not need positional encoding, thereby avoiding the interpolation of positional codes which leads to decreased performance when the testing resolution differs from training. 2) SegFormer avoids complex decoders. The proposed MLP decoder aggregates information from different layers, and thus combining both local attention and global attention to render powerful representations. We show that this simple and lightweight design is the key to efficient segmentation on Transformers. We scale our approach up to obtain a series of models from SegFormer-B0 to SegFormer-B5, reaching significantly better performance and efficiency than previous counterparts. For example, SegFormer-B4 achieves 50.3% mIoU on ADE20K with 64M parameters, being 5x smaller and 2.2% better than the previous best method. Our best model, SegFormer-B5, achieves 84.0% mIoU on Cityscapes validation set and shows excellent zero-shot robustness on Cityscapes-C. Code will be released at: github.com/NVlabs/SegFormer."
                },
                {
                    "title": "Pay Attention to MLPs",
                    "abstract": "Transformers have become one of the most important architectural innovations in deep learning and have enabled many breakthroughs over the past few years. Here we propose a simple network architecture, gMLP, based on MLPs with gating, and show that it can perform as well as Transformers in key language and vision applications. Our comparisons show that self-attention is not critical for Vision Transformers, as gMLP can achieve the same accuracy. For BERT, our model achieves parity with Transformers on pretraining perplexity and is better on some downstream NLP tasks. On finetuning tasks where gMLP performs worse, making the gMLP model substantially larger can close the gap with Transformers. In general, our experiments show that gMLP can scale as well as Transformers over increased data and compute."
                },
                {
                    "title": "ResMLP: Feedforward Networks for Image Classification With Data-Efficient Training",
                    "abstract": "We present ResMLP, an architecture built entirely upon multi-layer perceptrons for image classification. It is a simple residual network that alternates (i) a linear layer in which image patches interact, independently and identically across channels, and (ii) a two-layer feed-forward network in which channels interact independently per patch. When trained with a modern training strategy using heavy data-augmentation and optionally distillation, it attains surprisingly good accuracy/complexity trade-offs on ImageNet. We also train ResMLP models in a self-supervised setup, to further remove priors from employing a labelled dataset. Finally, by adapting our model to machine translation we achieve surprisingly good results. We share pre-trained models and our code based on the Timm library."
                },
                {
                    "title": "Do You Even Need Attention? A Stack of Feed-Forward Layers Does Surprisingly Well on ImageNet",
                    "abstract": "The strong performance of vision transformers on image classification and other vision tasks is often attributed to the design of their multi-head attention layers. However, the extent to which attention is responsible for this strong performance remains unclear. In this short report, we ask: is the attention layer even necessary? Specifically, we replace the attention layer in a vision transformer with a feed-forward layer applied over the patch dimension. The resulting architecture is simply a series of feed-forward layers applied over the patch and feature dimensions in an alternating fashion. In experiments on ImageNet, this architecture performs surprisingly well: a ViT/DeiT-base-sized model obtains 74.9\\% top-1 accuracy, compared to 77.9\\% and 79.9\\% for ViT and DeiT respectively. These results indicate that aspects of vision transformers other than attention, such as the patch embedding, may be more responsible for their strong performance than previously thought. We hope these results prompt the community to spend more time trying to understand why our current models are as effective as they are."
                },
                {
                    "title": "MLP-Mixer: An all-MLP Architecture for Vision",
                    "abstract": "Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e.\"mixing\"the per-location features), and one with MLPs applied across patches (i.e.\"mixing\"spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers."
                },
                {
                    "title": "Conditional Generation Using Polynomial Expansions",
                    "abstract": "Generative modeling has evolved to a notable field of machine learning. Deep polynomial neural networks (PNNs) have demonstrated impressive results in unsupervised image generation, where the task is to map an input vector (i.e., noise) to a synthesized image. However, the success of PNNs has not been replicated in conditional generation tasks, such as super-resolution. Existing PNNs focus on single-variable polynomial expansions which do not fare well to two-variable inputs, i.e., the noise variable and the conditional variable. In this work, we introduce a general framework, called CoPE, that enables a polynomial expansion of two input variables and captures their auto- and cross-correlations. We exhibit how CoPE can be trivially augmented to accept an arbitrary number of input variables. CoPE is evaluated in five tasks (class-conditional generation, inverse problems, edges-to-image translation, image-to-image translation, attribute-guided generation) involving eight datasets. The thorough evaluation suggests that CoPE can be useful for tackling diverse conditional generation tasks. The source code of CoPE is available at \\url{https://github.com/grigorisg9gr/polynomial_nets_for_conditional_generation}."
                },
                {
                    "title": "Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions",
                    "abstract": "Although convolutional neural networks (CNNs) have achieved great success in computer vision, this work investigates a simpler, convolution-free backbone network use-fid for many dense prediction tasks. Unlike the recently-proposed Vision Transformer (ViT) that was designed for image classification specifically, we introduce the Pyramid Vision Transformer (PVT), which overcomes the difficulties of porting Transformer to various dense prediction tasks. PVT has several merits compared to current state of the arts. (1) Different from ViT that typically yields low-resolution outputs and incurs high computational and memory costs, PVT not only can be trained on dense partitions of an image to achieve high output resolution, which is important for dense prediction, but also uses a progressive shrinking pyramid to reduce the computations of large feature maps. (2) PVT inherits the advantages of both CNN and Transformer, making it a unified backbone for various vision tasks without convolutions, where it can be used as a direct replacement for CNN backbones. (3) We validate PVT through extensive experiments, showing that it boosts the performance of many downstream tasks, including object detection, instance and semantic segmentation. For example, with a comparable number of parameters, PVT+RetinaNet achieves 40.4 AP on the COCO dataset, surpassing ResNet50+RetinNet (36.3 AP) by 4.1 absolute AP (see Figure 2). We hope that PVT could, serre as an alternative and useful backbone for pixel-level predictions and facilitate future research."
                },
                {
                    "title": "Training data-efficient image transformers & distillation through attention",
                    "abstract": "Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption. In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data. More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models."
                },
                {
                    "title": "MedMNIST Classification Decathlon: A Lightweight AutoML Benchmark for Medical Image Analysis",
                    "abstract": "We present MedMNIST, a collection of 10 pre-processed medical open datasets. MedMNIST is standardized to perform classification tasks on lightweight 28 $\\times$ 28 images, which requires no background knowledge. Covering the primary data modalities in medical image analysis, it is diverse on data scale (from 100 to 100,000) and tasks (binary/multi-class, ordinal regression and multi-label). MedMNIST could be used for educational purpose, rapid prototyping, multi-modal machine learning or AutoML in medical image analysis. Moreover, MedMNIST Classification Decathlon is designed to benchmark AutoML algorithms on all 10 datasets; We have compared several baseline methods, including open-source or commercial AutoML tools. The datasets, evaluation code and baseline methods for MedMNIST are publicly available at https://medmnist.github.io/."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Multilinear Latent Conditioning for Generating Unseen Attribute Combinations",
                    "abstract": "Deep generative models rely on their inductive bias to facilitate generalization, especially for problems with high dimensional data, like images. However, empirical studies have shown that variational autoencoders (VAE) and generative adversarial networks (GAN) lack the generalization ability that occurs naturally in human perception. For example, humans can visualize a woman smiling after only seeing a smiling man. On the contrary, the standard conditional VAE (cVAE) is unable to generate unseen attribute combinations. To this end, we extend cVAE by introducing a multilinear latent conditioning framework that captures the multiplicative interactions between the attributes. We implement two variants of our model and demonstrate their efficacy on MNIST, Fashion-MNIST and CelebA. Altogether, we design a novel conditioning framework that can be used with any architecture to synthesize unseen attribute combinations."
                },
                {
                    "title": "P\u2013nets: Deep Polynomial Neural Networks",
                    "abstract": "Deep Convolutional Neural Networks (DCNNs) is currently the method of choice both for generative, as well as for discriminative learning in computer vision and machine learning. The success of DCNNs can be attributed to the careful selection of their building blocks (e.g., residual blocks, rectifiers, sophisticated normalization schemes, to mention but a few). In this paper, we propose $\\Pi$-Nets, a new class of DCNNs. $\\Pi$-Nets are polynomial neural networks, i.e., the output is a high-order polynomial of the input. $\\Pi$-Nets can be implemented using special kind of skip connections and their parameters can be represented via high-order tensors. We empirically demonstrate that $\\Pi$-Nets have better representation power than standard DCNNs and they even produce good results without the use of non-linear activation functions in a large battery of tasks and signals, i.e., images, graphs, and audio. When used in conjunction with activation functions, $\\Pi$-Nets produce state-of-the-art results in challenging tasks, such as image generation. Lastly, our framework elucidates why recent generative models, such as StyleGAN, improve upon their predecessors, e.g., ProGAN."
                },
                {
                    "title": "CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features",
                    "abstract": "Regional dropout strategies have been proposed to enhance performance of convolutional neural network classifiers. They have proved to be effective for guiding the model to attend on less discriminative parts of objects (e.g. leg as opposed to head of a person), thereby letting the network generalize better and have better object localization capabilities. On the other hand, current methods for regional dropout removes informative pixels on training images by overlaying a patch of either black pixels or random noise. Such removal is not desirable because it suffers from information loss causing inefficiency in training. We therefore propose the CutMix augmentation strategy: patches are cut and pasted among training images where the ground truth labels are also mixed proportionally to the area of the patches. By making efficient use of training pixels and retaining the regularization effect of regional dropout, CutMix consistently outperforms state-of-the-art augmentation strategies on CIFAR and ImageNet classification tasks, as well as on ImageNet weakly-supervised localization task. Moreover, unlike previous augmentation methods, our CutMix-trained ImageNet classifier, when used as a pretrained model, results in consistent performance gain in Pascal detection and MS-COCO image captioning benchmarks. We also show that CutMix can improve the model robustness against input corruptions and its out-of distribution detection performance."
                },
                {
                    "title": "Attention Augmented Convolutional Networks",
                    "abstract": "Convolutional networks have enjoyed much success in many computer vision applications. The convolution operation however has a significant weakness in that it only operates on a local neighbourhood, thus missing global information. Self-attention, on the other hand, has emerged as a recent advance to capture long range interactions, but has mostly been applied to sequence modeling and generative modeling tasks. In this paper, we propose to augment convolutional networks with self-attention by concatenating convolutional feature maps with a set of feature maps produced via a novel relative self-attention mechanism. In particular, we extend previous work on relative self-attention over sequences to images and discuss a memory efficient implementation. Unlike Squeeze-and-Excitation, which performs attention over the channels and ignores spatial information, our self-attention mechanism attends jointly to both features and spatial locations while preserving translation equivariance. We find that Attention Augmentation leads to consistent improvements in image classification on ImageNet and object detection on COCO across many different models and scales, including ResNets and a state-of-the art mobile constrained network, while keeping the number of parameters similar. In particular, our method achieves a 1.3% top-1 accuracy improvement on ImageNet classification over a ResNet50 baseline and outperforms other attention mechanisms for images such as Squeeze-and-Excitation. It also achieves an improvement of 1.4 AP in COCO Object Detection on top of a RetinaNet baseline."
                },
                {
                    "title": "Augmented Neural ODEs",
                    "abstract": "We show that Neural Ordinary Differential Equations (ODEs) learn representations that preserve the topology of the input space and prove that this implies the existence of functions Neural ODEs cannot represent. To address these limitations, we introduce Augmented Neural ODEs which, in addition to being more expressive models, are empirically more stable, generalize better and have a lower computational cost than Neural ODEs."
                },
                {
                    "title": "Selective Kernel Networks",
                    "abstract": "In standard Convolutional Neural Networks (CNNs), the receptive fields of artificial neurons in each layer are designed to share the same size. It is well-known in the neuroscience community that the receptive field size of visual cortical neurons are modulated by the stimulus, which has been rarely considered in constructing CNNs. We propose a dynamic selection mechanism in CNNs that allows each neuron to adaptively adjust its receptive field size based on multiple scales of input information. A building block called Selective Kernel (SK) unit is designed, in which multiple branches with different kernel sizes are fused using softmax attention that is guided by the information in these branches. Different attentions on these branches yield different sizes of the effective receptive fields of neurons in the fusion layer. Multiple SK units are stacked to a deep network termed Selective Kernel Networks (SKNets). On the ImageNet and CIFAR benchmarks, we empirically show that SKNet outperforms the existing state-of-the-art architectures with lower model complexity. Detailed analyses show that the neurons in SKNet can capture target objects with different scales, which verifies the capability of neurons for adaptively adjusting their receptive field sizes according to the input. The code and models are available at https://github.com/implus/SKNet."
                },
                {
                    "title": "A2-Nets: Double Attention Networks",
                    "abstract": "Learning to capture long-range relations is fundamental to image/video recognition. Existing CNN models generally rely on increasing depth to model such relations which is highly inefficient. In this work, we propose the \"double attention block\", a novel component that aggregates and propagates informative global features from the entire spatio-temporal space of input images/videos, enabling subsequent convolution layers to access features from the entire space efficiently. The component is designed with a double attention mechanism in two steps, where the first step gathers features from the entire space into a compact set through second-order attention pooling and the second step adaptively selects and distributes features to each location via another attention. The proposed double attention block is easy to adopt and can be plugged into existing deep neural networks conveniently. We conduct extensive ablation studies and experiments on both image and video recognition tasks for evaluating its performance. On the image recognition task, a ResNet-50 equipped with our double attention blocks outperforms a much larger ResNet-152 architecture on ImageNet-1k dataset with over 40% less the number of parameters and less FLOPs. On the action recognition task, our proposed model achieves the state-of-the-art results on the Kinetics and UCF-101 datasets with significantly higher efficiency than recent works."
                },
                {
                    "title": "Benchmarking Neural Network Robustness to Common Corruptions and Surface Variations",
                    "abstract": "In this paper we establish rigorous benchmarks for image classifier robustness. Our first benchmark, ImageNet-C, standardizes and expands the corruption robustness topic, while showing which classifiers are preferable in safety-critical applications. Unlike recent robustness research, this benchmark evaluates performance on commonplace corruptions not worst-case adversarial corruptions. We find that there are negligible changes in relative corruption robustness from AlexNet to ResNet classifiers, and we discover ways to enhance corruption robustness. Then we propose a new dataset called Icons-50 which opens research on a new kind of robustness, surface variation robustness. With this dataset we evaluate the frailty of classifiers on new styles of known objects and unexpected instances of known classes. We also demonstrate two methods that improve surface variation robustness. Together our benchmarks may aid future work toward networks that learn fundamental class structure and also robustly generalize."
                },
                {
                    "title": "Neural Ordinary Differential Equations",
                    "abstract": "We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models."
                },
                {
                    "title": "AutoAugment: Learning Augmentation Policies from Data",
                    "abstract": "Data augmentation is an effective technique for improving the accuracy of modern image classifiers. However, current data augmentation implementations are manually designed. In this paper, we describe a simple procedure called AutoAugment to automatically search for improved data augmentation policies. In our implementation, we have designed a search space where a policy consists of many sub-policies, one of which is randomly chosen for each image in each mini-batch. A sub-policy consists of two operations, each operation being an image processing function such as translation, rotation, or shearing, and the probabilities and magnitudes with which the functions are applied. We use a search algorithm to find the best policy such that the neural network yields the highest validation accuracy on a target dataset. Our method achieves state-of-the-art accuracy on CIFAR-10, CIFAR-100, SVHN, and ImageNet (without additional data). On ImageNet, we attain a Top-1 accuracy of 83.5% which is 0.4% better than the previous record of 83.1%. On CIFAR-10, we achieve an error rate of 1.5%, which is 0.6% better than the previous state-of-the-art. Augmentation policies we find are transferable between datasets. The policy learned on ImageNet transfers well to achieve significant improvements on other datasets, such as Oxford Flowers, Caltech-101, Oxford-IIT Pets, FGVC Aircraft, and Stanford Cars."
                },
                {
                    "title": "Non-local Neural Networks",
                    "abstract": "Both convolutional and recurrent operations are building blocks that process one local neighborhood at a time. In this paper, we present non-local operations as a generic family of building blocks for capturing long-range dependencies. Inspired by the classical non-local means method [4] in computer vision, our non-local operation computes the response at a position as a weighted sum of the features at all positions. This building block can be plugged into many computer vision architectures. On the task of video classification, even without any bells and whistles, our nonlocal models can compete or outperform current competition winners on both Kinetics and Charades datasets. In static image recognition, our non-local models improve object detection/segmentation and pose estimation on the COCO suite of tasks. Code will be made available."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "mixup: Beyond Empirical Risk Minimization",
                    "abstract": "Large deep neural networks are powerful, but exhibit undesirable behaviors such as memorization and sensitivity to adversarial examples. In this work, we propose mixup, a simple learning principle to alleviate these issues. In essence, mixup trains a neural network on convex combinations of pairs of examples and their labels. By doing so, mixup regularizes the neural network to favor simple linear behavior in-between training examples. Our experiments on the ImageNet-2012, CIFAR-10, CIFAR-100, Google commands and UCI datasets show that mixup improves the generalization of state-of-the-art neural network architectures. We also find that mixup reduces the memorization of corrupt labels, increases the robustness to adversarial examples, and stabilizes the training of generative adversarial networks."
                },
                {
                    "title": "Squeeze-and-Excitation Networks",
                    "abstract": "Convolutional neural networks are built upon the convolution operation, which extracts informative features by fusing spatial and channel-wise information together within local receptive fields. In order to boost the representational power of a network, several recent approaches have shown the benefit of enhancing spatial encoding. In this work, we focus on the channel relationship and propose a novel architectural unit, which we term the \"Squeeze-and-Excitation\" (SE) block, that adaptively recalibrates channel-wise feature responses by explicitly modelling interdependencies between channels. We demonstrate that by stacking these blocks together, we can construct SENet architectures that generalise extremely well across challenging datasets. Crucially, we find that SE blocks produce significant performance improvements for existing state-of-the-art deep architectures at minimal additional computational cost. SENets formed the foundation of our ILSVRC 2017 classification submission which won first place and significantly reduced the top-5 error to 2.251%, achieving a ~25% relative improvement over the winning entry of 2016. Code and models are available at https://github.com/hujie-frank/SENet."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Feature Pyramid Networks for Object Detection",
                    "abstract": "Feature pyramids are a basic component in recognition systems for detecting objects at different scales. But pyramid representations have been avoided in recent object detectors that are based on deep convolutional networks, partially because they are slow to compute and memory intensive. In this paper, we exploit the inherent multi-scale, pyramidal hierarchy of deep convolutional networks to construct feature pyramids with marginal extra cost. A top-down architecture with lateral connections is developed for building high-level semantic feature maps at all scales. This architecture, called a Feature Pyramid Network (FPN), shows significant improvement as a generic feature extractor in several applications. Using a basic Faster R-CNN system, our method achieves state-of-the-art single-model results on the COCO detection benchmark without bells and whistles, surpassing all existing single-model entries including those from the COCO 2016 challenge winners. In addition, our method can run at 5 FPS on a GPU and thus is a practical and accurate solution to multi-scale object detection. Code will be made publicly available."
                },
                {
                    "title": "Layer Normalization",
                    "abstract": "Training state-of-the-art, deep neural networks is computationally expensive. One way to reduce the training time is to normalize the activities of the neurons. A recently introduced technique called batch normalization uses the distribution of the summed input to a neuron over a mini-batch of training cases to compute a mean and variance which are then used to normalize the summed input to that neuron on each training case. This significantly reduces the training time in feed-forward neural networks. However, the effect of batch normalization is dependent on the mini-batch size and it is not obvious how to apply it to recurrent neural networks. In this paper, we transpose batch normalization into layer normalization by computing the mean and variance used for normalization from all of the summed inputs to the neurons in a layer on a single training case. Like batch normalization, we also give each neuron its own adaptive bias and gain which are applied after the normalization but before the non-linearity. Unlike batch normalization, layer normalization performs exactly the same computation at training and test times. It is also straightforward to apply to recurrent neural networks by computing the normalization statistics separately at each time step. Layer normalization is very effective at stabilizing the hidden state dynamics in recurrent networks. Empirically, we show that layer normalization can substantially reduce the training time compared with previously published techniques."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Rethinking the Inception Architecture for Computer Vision",
                    "abstract": "Convolutional networks are at the core of most state of-the-art computer vision solutions for a wide variety of tasks. Since 2014 very deep convolutional networks started to become mainstream, yielding substantial gains in various benchmarks. Although increased model size and computational cost tend to translate to immediate quality gains for most tasks (as long as enough labeled data is provided for training), computational efficiency and low parameter count are still enabling factors for various use cases such as mobile vision and big-data scenarios. Here we are exploring ways to scale up networks in ways that aim at utilizing the added computation as efficiently as possible by suitably factorized convolutions and aggressive regularization. We benchmark our methods on the ILSVRC 2012 classification challenge validation set demonstrate substantial gains over the state of the art: 21:2% top-1 and 5:6% top-5 error for single frame evaluation using a network with a computational cost of 5 billion multiply-adds per inference and with using less than 25 million parameters. With an ensemble of 4 models and multi-crop evaluation, we report 3:5% top-5 error and 17:3% top-1 error on the validation set and 3:6% top-5 error on the official test set."
                },
                {
                    "title": "Intelligible Models for HealthCare: Predicting Pneumonia Risk and Hospital 30-day Readmission",
                    "abstract": "In machine learning often a tradeoff must be made between accuracy and intelligibility. More accurate models such as boosted trees, random forests, and neural nets usually are not intelligible, but more intelligible models such as logistic regression, naive-Bayes, and single decision trees often have significantly worse accuracy. This tradeoff sometimes limits the accuracy of models that can be applied in mission-critical applications such as healthcare where being able to understand, validate, edit, and trust a learned model is important. We present two case studies where high-performance generalized additive models with pairwise interactions (GA2Ms) are applied to real healthcare problems yielding intelligible models with state-of-the-art accuracy. In the pneumonia risk prediction case study, the intelligible model uncovers surprising patterns in the data that previously had prevented complex learned models from being fielded in this domain, but because it is intelligible and modular allows these patterns to be recognized and removed. In the 30-day hospital readmission case study, we show that the same methods scale to large datasets containing hundreds of thousands of patients and thousands of attributes while remaining intelligible and providing accuracy comparable to the best (unintelligible) machine learning methods."
                },
                {
                    "title": "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification",
                    "abstract": "Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on the learnable activation and advanced initialization, we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66% [33]). To our knowledge, our result is the first to surpass the reported human-level performance (5.1%, [26]) on this dataset."
                },
                {
                    "title": "Fully convolutional networks for semantic segmentation",
                    "abstract": "Convolutional networks are powerful visual models that yield hierarchies of features. We show that convolutional networks by themselves, trained end-to-end, pixels-to-pixels, exceed the state-of-the-art in semantic segmentation. Our key insight is to build \u201cfully convolutional\u201d networks that take input of arbitrary size and produce correspondingly-sized output with efficient inference and learning. We define and detail the space of fully convolutional networks, explain their application to spatially dense prediction tasks, and draw connections to prior models. We adapt contemporary classification networks (AlexNet [20], the VGG net [31], and GoogLeNet [32]) into fully convolutional networks and transfer their learned representations by fine-tuning [3] to the segmentation task. We then define a skip architecture that combines semantic information from a deep, coarse layer with appearance information from a shallow, fine layer to produce accurate and detailed segmentations. Our fully convolutional network achieves state-of-the-art segmentation of PASCAL VOC (20% relative improvement to 62.2% mean IU on 2012), NYUDv2, and SIFT Flow, while inference takes less than one fifth of a second for a typical image."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "A survey of image classification methods and techniques",
                    "abstract": "In this paper, we review the current activity of image classification methodologies and techniques. Image classification is a complex process which depends upon various factors. Here, we discuss about the current techniques, problems as well as prospects of image classification. The main focus will be on advanced classification techniques which are used for improving classification accuracy. Additionally, some important issues relating to classification performance are also discussed."
                },
                {
                    "title": "(Leveled) fully homomorphic encryption without bootstrapping",
                    "abstract": "We present a novel approach to fully homomorphic encryption (FHE) that dramatically improves performance and bases security on weaker assumptions. A central conceptual contribution in our work is a new way of constructing leveled fully homomorphic encryption schemes (capable of evaluating arbitrary polynomial-size circuits), without Gentry's bootstrapping procedure.\n Specifically, we offer a choice of FHE schemes based on the learning with error (LWE) or ring-LWE (RLWE) problems that have 2\u03bb security against known attacks. For RLWE, we have:\n \u2022 A leveled FHE scheme that can evaluate L-level arithmetic circuits with \u00d5(\u03bb \u00b7 L3) per-gate computation -- i.e., computation quasi-linear in the security parameter. Security is based on RLWE for an approximation factor exponential in L. This construction does not use the bootstrapping procedure.\n \u2022 A leveled FHE scheme that uses bootstrapping as an optimization, where the per-gate computation (which includes the bootstrapping procedure) is \u00d5(\u03bb2), independent of L. Security is based on the hardness of RLWE for quasi-polynomial factors (as opposed to the sub-exponential factors needed in previous schemes).\n We obtain similar results to the above for LWE, but with worse performance.\n Based on the Ring LWE assumption, we introduce a number of further optimizations to our schemes. As an example, for circuits of large width -- e.g., where a constant fraction of levels have width at least \u03bb -- we can reduce the per-gate computation of the bootstrapped version to \u00d5(\u03bb), independent of L, by batching the bootstrapping operation. Previous FHE schemes all required \u03a9(\u03bb3.5) computation per gate.\n At the core of our construction is a much more effective approach for managing the noise level of lattice-based ciphertexts as homomorphic operations are performed, using some new techniques recently introduced by Brakerski and Vaikuntanathan (FOCS 2011)."
                },
                {
                    "title": "Deep learning via Hessian-free optimization",
                    "abstract": "We develop a 2nd-order optimization method based on the \"Hessian-free\" approach, and apply it to training deep auto-encoders. Without using pre-training, we obtain results superior to those reported by Hinton & Salakhutdinov (2006) on the same tasks they considered. Our method is practical, easy to use, scales nicely to very large datasets, and isn't limited in applicability to auto-encoders, or any specific model class. We also discuss the issue of \"pathological curvature\" as a possible explanation for the difficulty of deep-learning and how 2nd-order optimization, and our method in particular, effectively deals with it."
                },
                {
                    "title": "Understanding the difficulty of training deep feedforward neural networks",
                    "abstract": "Whereas before 2006 it appears that deep multilayer neural networks were not successfully trained, since then several algorithms have been shown to successfully train them, with experimental results showing the superiority of deeper vs less deep architectures. All these experimental results were obtained with new initialization or training mechanisms. Our objective here is to understand better why standard gradient descent from random initialization is doing so poorly with deep neural networks, to better understand these recent relative successes and help design better algorithms in the future. We first observe the influence of the non-linear activations functions. We find that the logistic sigmoid activation is unsuited for deep networks with random initialization because of its mean value, which can drive especially the top hidden layer into saturation. Surprisingly, we find that saturated units can move out of saturation by themselves, albeit slowly, and explaining the plateaus sometimes seen when training neural networks. We find that a new non-linearity that saturates less can often be beneficial. Finally, we study how activations and gradients vary across layers and during training, with the idea that training may be more difficult when the singular values of the Jacobian associated with each layer are far from 1. Based on these considerations, we propose a new initialization scheme that brings substantially faster convergence. 1 Deep Neural Networks Deep learning methods aim at learning feature hierarchies with features from higher levels of the hierarchy formed by the composition of lower level features. They include Appearing in Proceedings of the 13 International Conference on Artificial Intelligence and Statistics (AISTATS) 2010, Chia Laguna Resort, Sardinia, Italy. Volume 9 of JMLR: WC Weston et al., 2008). Much attention has recently been devoted to them (see (Bengio, 2009) for a review), because of their theoretical appeal, inspiration from biology and human cognition, and because of empirical success in vision (Ranzato et al., 2007; Larochelle et al., 2007; Vincent et al., 2008) and natural language processing (NLP) (Collobert & Weston, 2008; Mnih & Hinton, 2009). Theoretical results reviewed and discussed by Bengio (2009), suggest that in order to learn the kind of complicated functions that can represent high-level abstractions (e.g. in vision, language, and other AI-level tasks), one may need deep architectures. Most of the recent experimental results with deep architecture are obtained with models that can be turned into deep supervised neural networks, but with initialization or training schemes different from the classical feedforward neural networks (Rumelhart et al., 1986). Why are these new algorithms working so much better than the standard random initialization and gradient-based optimization of a supervised training criterion? Part of the answer may be found in recent analyses of the effect of unsupervised pretraining (Erhan et al., 2009), showing that it acts as a regularizer that initializes the parameters in a \u201cbetter\u201d basin of attraction of the optimization procedure, corresponding to an apparent local minimum associated with better generalization. But earlier work (Bengio et al., 2007) had shown that even a purely supervised but greedy layer-wise procedure would give better results. So here instead of focusing on what unsupervised pre-training or semi-supervised criteria bring to deep architectures, we focus on analyzing what may be going wrong with good old (but deep) multilayer neural networks. Our analysis is driven by investigative experiments to monitor activations (watching for saturation of hidden units) and gradients, across layers and across training iterations. We also evaluate the effects on these of choices of activation function (with the idea that it might affect saturation) and initialization procedure (since unsupervised pretraining is a particular form of initialization and it has a drastic impact)."
                },
                {
                    "title": "Tensor Decompositions and Applications",
                    "abstract": "This survey provides an overview of higher-order tensor decompositions, their applications, and available software. A tensor is a multidimensional or $N$-way array. Decompositions of higher-order tensors (i.e., $N$-way arrays with $N \\geq 3$) have applications in psycho-metrics, chemometrics, signal processing, numerical linear algebra, computer vision, numerical analysis, data mining, neuroscience, graph analysis, and elsewhere. Two particular tensor decompositions can be considered to be higher-order extensions of the matrix singular value decomposition: CANDECOMP/PARAFAC (CP) decomposes a tensor as a sum of rank-one tensors, and the Tucker decomposition is a higher-order form of principal component analysis. There are many other tensor decompositions, including INDSCAL, PARAFAC2, CANDELINC, DEDICOM, and PARATUCK2 as well as nonnegative variants of all of the above. The N-way Toolbox, Tensor Toolbox, and Multilinear Engine are examples of software packages for working with tensors."
                },
                {
                    "title": "Automated Flower Classification over a Large Number of Classes",
                    "abstract": "We investigate to what extent combinations of features can improve classification performance on a large dataset of similar classes. To this end we introduce a 103 class flower dataset. We compute four different features for the flowers, each describing different aspects, namely the local shape/texture, the shape of the boundary, the overall spatial distribution of petals, and the colour. We combine the features using a multiple kernel framework with a SVM classifier. The weights for each class are learnt using the method of Varma and Ray, which has achieved state of the art performance on other large dataset, such as Caltech 101/256. Our dataset has a similar challenge in the number of classes, but with the added difficulty of large between class similarity and small within class similarity. Results show that learning the optimum kernel combination of multiple features vastly improves the performance, from 55.1% for the best single feature to 72.8% for the combination of all features."
                },
                {
                    "title": "The pi-sigma network: an efficient higher-order neural network for pattern classification and function approximation",
                    "abstract": "Introduces a novel feedforward network called the pi-sigma network. This network utilizes product cells as the output units to indirectly incorporate the capabilities of higher-order networks while using a fewer number of weights and processing units. The network has a regular structure, exhibits much faster learning, and is amenable to the incremental addition of units to attain a desired level of complexity. Simulation results show good convergence properties and accuracy for function approximation. Comparative results using the DARPA acoustic transient data set are also provided to highlight the classification abilities of pi-sigma networks.<<ETX>>"
                },
                {
                    "title": "Polynomial Theory of Complex Systems",
                    "abstract": "A complex multidimensional decision hypersurface can be approximated by a set of polynomials in the input signals (properties) which contain information about the hypersurface of interest. The hypersurface is usually described by a number of experimental (vector) points and simple functions of their coordinates. The approach taken in this paper to approximating the decision hypersurface, and hence the input-output relationship of a complex system, is to fit a high-degree multinomial to the input properties using a multilayered perceptronlike network structure. Thresholds are employed at each layer in the network to identify those polynomials which best fit into the desired hypersurface. Only the best combinations of the input properties are allowed to pass to succeeding layers, where more complex combinations are formed. Each element in each layer in the network implements a nonlinear function of two inputs. The coefficients of each element are determined by a regression technique which enables each element to approximate the true outputs with minimum mean-square error. The experimental data base is divided into a training and testing set. The training set is used to obtain the element coefficients, and the testing set is used to determine the utility of a given element in the network and to control overfitting of the experimental data. This latter feature is termed \"decision regularization."
                },
                {
                    "title": "Tiny ImageNet Visual Recognition Challenge",
                    "abstract": "In this work, we investigate the effect of convolutional network depth, receptive field size, dropout layers, rectified activation unit type and dataset noise on its accuracy in Tiny-ImageNet Challenge settings. In order to make a thorough evaluation of the cause of the peformance improvement, we start with a basic 5 layer model with 5\u00d75 convolutional receptive fields. We keep increasing network depth or reducing receptive field size, and continue applying modern techniques, such as PReLu and dropout, to the model. Our model achieves excellent performance even compared to state-of-the-art results, with 0.444 final error rate on the test set."
                },
                {
                    "title": "Reading Digits in Natural Images with Unsupervised Feature Learning",
                    "abstract": "Detecting and reading text from natural images is a hard computer vision task that is central to a variety of emerging applications. Related problems like document character recognition have been widely studied by computer vision and machine learning researchers and are virtually solved for practical applications like reading handwritten digits. Reliably recognizing characters in more complex scenes like photographs, however, is far more difficult: the best existing methods lag well behind human performance on the same tasks. In this paper we attack the problem of recognizing digits in a real application using unsupervised feature learning methods: reading house numbers from street level photos. To this end, we introduce a new benchmark dataset for research use containing over 600,000 labeled digits cropped from Street View images. We then demonstrate the difficulty of recognizing these digits when the problem is approached with hand-designed features. Finally, we employ variants of two recently proposed unsupervised feature learning methods and find that they are convincingly superior on our benchmarks."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "[Question 1] - What is the problem?  \nHow can the choice of weight initialization methods in deep learning models impact their performance on image classification tasks?\n\n[Question 2] - Why is it interesting and important?  \nUnderstanding the impact of weight initialization methods is crucial for the research community as it can lead to more efficient training processes and improved model performance. By addressing this question, future research can focus on optimizing initialization strategies, potentially leading to breakthroughs in model convergence rates and accuracy. This knowledge can also have practical applications in various domains, such as computer vision, where enhanced model performance can lead to better real-world applications, including medical imaging and autonomous systems.\n\n[Question 3] - Why is it hard?  \nThe challenges in solving this problem stem from the complex interplay between weight initialization and the optimization landscape of deep learning models. Naive approaches may fail because they do not account for the specific architecture of the model or the nature of the dataset, leading to suboptimal convergence or performance. Additionally, the theoretical understanding of how different initialization methods affect gradient flow and model training is still evolving, making it difficult to predict outcomes. Practical obstacles include the need for extensive experimentation across various architectures and datasets to draw generalizable conclusions.\n\n[Question 4] - Why hasn't it been solved before?  \nPrevious research has often focused on specific initialization methods without comprehensively comparing their effects across a wide range of architectures and datasets. Limitations in computational resources and the complexity of deep learning models have also hindered thorough investigations. Additionally, many studies have not systematically evaluated the performance metrics, such as Top-1 and Top-5 accuracy, in a unified manner. My approach differs by systematically analyzing multiple initialization methods across various architectures and providing a detailed comparison of their impacts on performance metrics.\n\n[Question 5] - What are the key components of my approach and results?  \nMy proposed methodology involves conducting a series of experiments using a diverse set of deep learning architectures on standard image classification datasets. I will evaluate the performance of various weight initialization methods, including Xavier Uniform, Xavier Normal, Kaiming Uniform, Kaiming Normal, and others, using Top-1 and Top-5 accuracy as metrics. The expected outcomes include identifying the most effective initialization methods for different architectures, providing insights into their impact on model performance, and establishing guidelines for practitioners to improve their model training processes."
            }
        },
        "author_data": {
            "d8e3a7aa-a692-42ce-b85f-5c68abab8556": {
                "pk": "d8e3a7aa-a692-42ce-b85f-5c68abab8556",
                "name": "Yixin Cheng",
                "collaborators": [
                    "Markos Georgopoulos",
                    "V. Cevher",
                    "Grigorios G. Chrysos"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Security",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "Leveraging the Context through Multi-Round Interactions for Jailbreaking Attacks",
                        "abstract": "Large Language Models (LLMs) are susceptible to Jailbreaking attacks, which aim to extract harmful information by subtly modifying the attack query. As defense mechanisms evolve, directly obtaining harmful information becomes increasingly challenging for Jailbreaking attacks. In this work, inspired from Chomsky's transformational-generative grammar theory and human practices of indirect context to elicit harmful information, we focus on a new attack form, called Contextual Interaction Attack. We contend that the prior context\\u2014the information preceding the attack query\\u2014plays a pivotal role in enabling strong Jailbreaking attacks. Specifically, we propose a first multi-turn approach that leverages benign preliminary questions to interact with the LLM. Due to the autoregressive nature of LLMs, which use previous conversation rounds as context during generation, we guide the model's question-response pair to construct a context that is semantically aligned with the attack query to execute the attack. We conduct experiments on seven different LLMs and demonstrate the efficacy of this attack, which is black-box and can also transfer across LLMs. We believe this can lead to further developments and understanding of security in LLMs."
                    }
                ]
            },
            "30ff9365-fa1c-498b-9890-55f7db720e09": {
                "pk": "30ff9365-fa1c-498b-9890-55f7db720e09",
                "name": "Grigorios G. Chrysos",
                "collaborators": [
                    "V. Cevher",
                    "Fanghui Liu",
                    "Markos Georgopoulos",
                    "Yongtao Wu",
                    "Yannis Panagakis",
                    "Jiankang Deng",
                    "El\u00edas Abad-Rocamora",
                    "Aggelina Chatziagapi",
                    "Dimitris Samaras",
                    "Stratis Skoulakis",
                    "Eva Papadogiannaki",
                    "K. Georgopoulos",
                    "S. Ioannidis",
                    "Justin Deschenaux",
                    "Igor Krawczuk",
                    "Yixin Cheng",
                    "James Oldfield",
                    "Christos Tzelepis",
                    "M. Nicolaou",
                    "Ioannis Patras",
                    "P. Olmos",
                    "Pol Puigdemont",
                    "Arshia Afzal",
                    "Mahsa Shoaran",
                    "Zheyang Xiong",
                    "Ziyang Cai",
                    "John Cooper",
                    "Albert Ge",
                    "Vasileios Papageorgiou",
                    "Zack Sifakis",
                    "Angeliki Giannou",
                    "Ziqian Lin",
                    "Liu Yang",
                    "Saurabh Agarwal",
                    "Samet Oymak",
                    "Kangwook Lee",
                    "Dimitris G. Papailiopoulos",
                    "Yihang Chen",
                    "Yiping Lu",
                    "Carl-Johann Simon-Gabriel",
                    "Sonam Gupta",
                    "S. Tomar",
                    "Sukhendu Das",
                    "A. N. Rajagopalan",
                    "Lorenzo Brusca",
                    "Lars C.P.M. Quaedvlieg",
                    "Konstantinos Georgopoulos",
                    "Sotiris Ioannidis",
                    "Christos Diktopoulos",
                    "A. Brokalakis",
                    "Georgios Christou",
                    "Ioannis Morianos",
                    "Dimitris Deyannis"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Adversarial Attacks",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Going beyond Compositions, DDPMs Can Produce Zero-Shot Interpolations",
                        "abstract": "Denoising Diffusion Probabilistic Models (DDPMs) exhibit remarkable capabilities in image generation, with studies suggesting that they can generalize by composing latent factors learned from the training data. In this work, we go further and study DDPMs trained on strictly separate subsets of the data distribution with large gaps on the support of the latent factors. We show that such a model can effectively generate images in the unexplored, intermediate regions of the distribution. For instance, when trained on clearly smiling and non-smiling faces, we demonstrate a sampling procedure which can generate slightly smiling faces without reference images (zero-shot interpolation). We replicate these findings for other attributes as well as other datasets. Our code is available at https://github.com/jdeschena/ddpm-zero-shot-interpolation."
                    },
                    {
                        "title": "Leveraging the Context through Multi-Round Interactions for Jailbreaking Attacks",
                        "abstract": "Large Language Models (LLMs) are susceptible to Jailbreaking attacks, which aim to extract harmful information by subtly modifying the attack query. As defense mechanisms evolve, directly obtaining harmful information becomes increasingly challenging for Jailbreaking attacks. In this work, inspired from Chomsky's transformational-generative grammar theory and human practices of indirect context to elicit harmful information, we focus on a new attack form, called Contextual Interaction Attack. We contend that the prior context\\u2014the information preceding the attack query\\u2014plays a pivotal role in enabling strong Jailbreaking attacks. Specifically, we propose a first multi-turn approach that leverages benign preliminary questions to interact with the LLM. Due to the autoregressive nature of LLMs, which use previous conversation rounds as context during generation, we guide the model's question-response pair to construct a context that is semantically aligned with the attack query to execute the attack. We conduct experiments on seven different LLMs and demonstrate the efficacy of this attack, which is black-box and can also transfer across LLMs. We believe this can lead to further developments and understanding of security in LLMs."
                    },
                    {
                        "title": "Multilinear Mixture of Experts: Scalable Expert Specialization through Factorization",
                        "abstract": "The Mixture of Experts (MoE) paradigm provides a powerful way to decompose dense layers into smaller, modular computations often more amenable to human interpretation, debugging, and editability. However, a major challenge lies in the computational cost of scaling the number of experts high enough to achieve fine-grained specialization. In this paper, we propose the Multilinear Mixture of Experts ($\\mu$MoE) layer to address this, focusing on vision models. $\\mu$MoE layers enable scalable expert specialization by performing an implicit computation on prohibitively large weight tensors entirely in factorized form. Consequently, $\\mu$MoEs (1) avoid the restrictively high inference-time costs of dense MoEs, yet (2) do not inherit the training issues of the popular sparse MoEs' discrete (non-differentiable) expert routing. We present both qualitative and quantitative evidence that scaling $\\mu$MoE layers when fine-tuning foundation models for vision tasks leads to more specialized experts at the class-level, further enabling manual bias correction in CelebA attribute classification. Finally, we show qualitative results demonstrating the expert specialism achieved when pre-training large GPT2 and MLP-Mixer models with parameter-matched $\\mu$MoE blocks at every layer, maintaining comparable accuracy. Our code is available at: https://github.com/james-oldfield/muMoE."
                    },
                    {
                        "title": "Efficient local linearity regularization to overcome catastrophic overfitting",
                        "abstract": "Catastrophic overfitting (CO) in single-step adversarial training (AT) results in abrupt drops in the adversarial test accuracy (even down to 0%). For models trained with multi-step AT, it has been observed that the loss function behaves locally linearly with respect to the input, this is however lost in single-step AT. To address CO in single-step AT, several methods have been proposed to enforce local linearity of the loss via regularization. However, these regularization terms considerably slow down training due to Double Backpropagation. Instead, in this work, we introduce a regularization term, called ELLE, to mitigate CO effectively and efficiently in classical AT evaluations, as well as some more difficult regimes, e.g., large adversarial perturbations and long training schedules. Our regularization term can be theoretically linked to curvature of the loss function and is computationally cheaper than previous methods by avoiding Double Backpropagation. Our thorough experimental validation demonstrates that our work does not suffer from CO, even in challenging settings where previous works suffer from it. We also notice that adapting our regularization parameter during training (ELLE-A) greatly improves the performance, specially in large $\\epsilon$ setups. Our implementation is available in https://github.com/LIONS-EPFL/ELLE ."
                    },
                    {
                        "title": "MI-NeRF: Learning a Single Face NeRF from Multiple Identities",
                        "abstract": "In this work, we introduce a method that learns a single dynamic neural radiance field (NeRF) from monocular talking face videos of multiple identities. NeRFs have shown remarkable results in modeling the 4D dynamics and appearance of human faces. However, they require per-identity optimization. Although recent approaches have proposed techniques to reduce the training and rendering time, increasing the number of identities can be expensive. We introduce MI-NeRF (multi-identity NeRF), a single unified network that models complex non-rigid facial motion for multiple identities, using only monocular videos of arbitrary length. The core premise in our method is to learn the non-linear interactions between identity and non-identity specific information with a multiplicative module. By training on multiple videos simultaneously, MI-NeRF not only reduces the total training time compared to standard single-identity NeRFs, but also demonstrates robustness in synthesizing novel expressions for any input identity. We present results for both facial expression transfer and talking face video synthesis. Our method can be further personalized for a target identity given only a short video."
                    },
                    {
                        "title": "Learning to Remove Cuts in Integer Linear Programming",
                        "abstract": "Cutting plane methods are a fundamental approach for solving integer linear programs (ILPs). In each iteration of such methods, additional linear constraints (cuts) are introduced to the constraint set with the aim of excluding the previous fractional optimal solution while not affecting the optimal integer solution. In this work, we explore a novel approach within cutting plane methods: instead of only adding new cuts, we also consider the removal of previous cuts introduced at any of the preceding iterations of the method under a learnable parametric criteria. We demonstrate that in fundamental combinatorial optimization settings such cut removal policies can lead to significant improvements over both human-based and machine learning-guided cut addition policies even when implemented with simple models."
                    },
                    {
                        "title": "REST: Efficient and Accelerated EEG Seizure Analysis through Residual State Updates",
                        "abstract": "EEG-based seizure detection models face challenges in terms of inference speed and memory efficiency, limiting their real-time implementation in clinical devices. This paper introduces a novel graph-based residual state update mechanism (REST) for real-time EEG signal analysis in applications such as epileptic seizure detection. By leveraging a combination of graph neural networks and recurrent structures, REST efficiently captures both non-Euclidean geometry and temporal dependencies within EEG data. Our model demonstrates high accuracy in both seizure detection and classification tasks. Notably, REST achieves a remarkable 9-fold acceleration in inference speed compared to state-of-the-art models, while simultaneously demanding substantially less memory than the smallest model employed for this task. These attributes position REST as a promising candidate for real-time implementation in clinical devices, such as Responsive Neurostimulation or seizure alert systems."
                    },
                    {
                        "title": "Everything Everywhere All at Once: LLMs can In-Context Learn Multiple Tasks in Superposition",
                        "abstract": "Large Language Models (LLMs) have demonstrated remarkable in-context learning (ICL) capabilities. In this study, we explore a surprising phenomenon related to ICL: LLMs can perform multiple, computationally distinct ICL tasks simultaneously, during a single inference call, a capability we term\"task superposition\". We provide empirical evidence of this phenomenon across various LLM families and scales and show that this phenomenon emerges even if we train the model to in-context learn one task at a time. We offer theoretical explanations that this capability is well within the expressive power of transformers. We also explore how LLMs internally compose task vectors during superposition. Furthermore, we show that larger models can solve more ICL tasks in parallel, and better calibrate their output distribution. Our findings offer insights into the latent capabilities of LLMs, further substantiate the perspective of\"LLMs as superposition of simulators\", and raise questions about the mechanisms enabling simultaneous task execution."
                    },
                    {
                        "title": "Generalization of Scaled Deep ResNets in the Mean-Field Regime",
                        "abstract": "Despite the widespread empirical success of ResNet, the generalization properties of deep ResNet are rarely explored beyond the lazy training regime. In this work, we investigate \\emph{scaled} ResNet in the limit of infinitely deep and wide neural networks, of which the gradient flow is described by a partial differential equation in the large-neural network limit, i.e., the \\emph{mean-field} regime. To derive the generalization bounds under this setting, our analysis necessitates a shift from the conventional time-invariant Gram matrix employed in the lazy training regime to a time-variant, distribution-dependent version. To this end, we provide a global lower bound on the minimum eigenvalue of the Gram matrix under the mean-field regime. Besides, for the traceability of the dynamic of Kullback-Leibler (KL) divergence, we establish the linear convergence of the empirical error and estimate the upper bound of the KL divergence over parameters distribution. Finally, we build the uniform convergence for generalization bound via Rademacher complexity. Our results offer new insights into the generalization ability of deep ResNet beyond the lazy training regime and contribute to advancing the understanding of the fundamental properties of deep neural networks."
                    },
                    {
                        "title": "Revisiting character-level adversarial attacks",
                        "abstract": "Adversarial attacks in Natural Language Processing apply perturbations in the character or token levels. Token-level attacks, gaining prominence for their use of gradient-based methods, are susceptible to altering sentence semantics, leading to invalid adversarial examples. While character-level attacks easily maintain semantics, they have received less attention as they cannot easily adopt popular gradient-based methods, and are thought to be easy to defend. Challenging these beliefs, we introduce Charmer, an efficient query-based adversarial attack capable of achieving high attack success rate (ASR) while generating highly similar adversarial examples. Our method successfully targets both small (BERT) and large (Llama 2) models. Specifically, on BERT with SST-2, Charmer improves the ASR in 4.84% points and the USE similarity in 8% points with respect to the previous art. Our implementation is available in https://github.com/LIONS-EPFL/Charmer."
                    },
                    {
                        "title": "Robust NAS under adversarial training: benchmark, theory, and beyond",
                        "abstract": "Recent developments in neural architecture search (NAS) emphasize the significance of considering robust architectures against malicious data. However, there is a notable absence of benchmark evaluations and theoretical guarantees for searching these robust architectures, especially when adversarial training is considered. In this work, we aim to address these two challenges, making twofold contributions. First, we release a comprehensive data set that encompasses both clean accuracy and robust accuracy for a vast array of adversarially trained networks from the NAS-Bench-201 search space on image datasets. Then, leveraging the neural tangent kernel (NTK) tool from deep learning theory, we establish a generalization theory for searching architecture in terms of clean accuracy and robust accuracy under multi-objective adversarial training. We firmly believe that our benchmark and theoretical insights will significantly benefit the NAS community through reliable reproducibility, efficient assessment, and theoretical foundation, particularly in the pursuit of robust architectures."
                    },
                    {
                        "title": "MIGS: Multi-Identity Gaussian Splatting via Tensor Decomposition",
                        "abstract": "We introduce MIGS (Multi-Identity Gaussian Splatting), a novel method that learns a single neural representation for multiple identities, using only monocular videos. Recent 3D Gaussian Splatting (3DGS) approaches for human avatars require per-identity optimization. However, learning a multi-identity representation presents advantages in robustly animating humans under arbitrary poses. We propose to construct a high-order tensor that combines all the learnable 3DGS parameters for all the training identities. By assuming a low-rank structure and factorizing the tensor, we model the complex rigid and non-rigid deformations of multiple subjects in a unified network, significantly reducing the total number of parameters. Our proposed approach leverages information from all the training identities and enables robust animation under challenging unseen poses, outperforming existing approaches. It can also be extended to learn unseen identities."
                    },
                    {
                        "title": "PNeRV: A Polynomial Neural Representation for Videos",
                        "abstract": "Extracting Implicit Neural Representations (INRs) on video data poses unique challenges due to the additional temporal dimension. In the context of videos, INRs have predominantly relied on a frame-only parameterization, which sacrifices the spatiotemporal continuity observed in pixel-level (spatial) representations. To mitigate this, we introduce Polynomial Neural Representation for Videos (PNeRV), a parameter-wise efficient, patch-wise INR for videos that preserves spatiotemporal continuity. PNeRV leverages the modeling capabilities of Polynomial Neural Networks to perform the modulation of a continuous spatial (patch) signal with a continuous time (frame) signal. We further propose a custom Hierarchical Patch-wise Spatial Sampling Scheme that ensures spatial continuity while retaining parameter efficiency. We also employ a carefully designed Positional Embedding methodology to further enhance PNeRV's performance. Our extensive experimentation demonstrates that PNeRV outperforms the baselines in conventional Implicit Neural Representation tasks like compression along with downstream applications that require spatiotemporal continuity in the underlying representation. PNeRV not only addresses the challenges posed by video data in the realm of INRs but also opens new avenues for advanced video processing and analysis."
                    },
                    {
                        "title": "Maximum Independent Set: Self-Training through Dynamic Programming",
                        "abstract": "This work presents a graph neural network (GNN) framework for solving the maximum independent set (MIS) problem, inspired by dynamic programming (DP). Specifically, given a graph, we propose a DP-like recursive algorithm based on GNNs that firstly constructs two smaller sub-graphs, predicts the one with the larger MIS, and then uses it in the next recursive call. To train our algorithm, we require annotated comparisons of different graphs concerning their MIS size. Annotating the comparisons with the output of our algorithm leads to a self-training process that results in more accurate self-annotation of the comparisons and vice versa. We provide numerical evidence showing the superiority of our method vs prior methods in multiple synthetic and real-world datasets."
                    },
                    {
                        "title": "A Reconfigurable IDS Framework for Encrypted and Non-Encrypted Network Data in Supply Chains",
                        "abstract": "Supply Chains (SCs) constitute the bloodline for modern economic trade and manufacturing infrastructures. Their uninterrupted functionality remains of paramount importance. Consequently, substantial endeavors have been channeled into fortifying their security facets. This work presents an innovative solution to implement and integrate a well-established Intrusion Detection System (IDS) within the Supply Chain ecosystems. The proposed framework embodies an ingenuous way for merging an open-source and widely used network intrusion detection algorithm, Snort algorithm, with reconfigurable hardware technology, Field-Programmable Gate Arrays (FPGAs). The adoption of FPGAs confers notable advantages, as they enable the complete alleviation of high IDS workloads from Supply Chain servers. Moreover, the proposed framework possesses the ability to simultaneously check both non-encrypted and encrypted data traffic, thereby markedly amplifying the operational scope of the IDS. The final system performance evaluation shows that the proposed framework achieves network workload processing rate up to sixfold greater and energy consumption reduced by up to threefold when compared to the best optimized official software solution."
                    },
                    {
                        "title": "On the Convergence of Encoder-only Shallow Transformers",
                        "abstract": "In this paper, we aim to build the global convergence theory of encoder-only shallow Transformers under a realistic setting from the perspective of architectures, initialization, and scaling under a finite width regime. The difficulty lies in how to tackle the softmax in self-attention mechanism, the core ingredient of Transformer. In particular, we diagnose the scaling scheme, carefully tackle the input/output of softmax, and prove that quadratic overparameterization is sufficient for global convergence of our shallow Transformers under commonly-used He/LeCun initialization in practice. Besides, neural tangent kernel (NTK) based analysis is also given, which facilitates a comprehensive comparison. Our theory demonstrates the separation on the importance of different scaling schemes and initialization. We believe our results can pave the way for a better understanding of modern Transformers, particularly on training dynamics."
                    },
                    {
                        "title": "Assessing the Effectiveness of Active Fences Against SCAs for Multi-Tenant FPGAs",
                        "abstract": "The rising use of FPGAs, in the context of cloud computing, has created security concerns. Previous works have shown that malicious users can implement voltage fluctuation sensors and mount successful power analysis attacks against cryptographic algorithms that share the same Power Distribution Network (PDN). So far, masking and hiding schemes are the two main mitigation strategies against such attacks and previous work has shown that the use of an active fence of Ring Oscillators (ROs) holds the potential for constituting an effective hiding countermeasure if placed between two adversary users. Nevertheless, developing an effective proposition against remote Side-Channel Attacks (SCAs) remains an open research topic. This work presents the mapping of an intra-FPGA adversary scenario on a Xilinx UltraScale+ MPSoC to assess the effectiveness of the Ring Oscillator active fence countermeasure. We compare different active fence configurations, with a varying number of Ring Oscillators, while using a new, resource efficient, activation method aiming at the achievement of noise injection hiding. The results show that by using our active fence scheme, which exhibits lower area overhead and lower power consumption than the algorithm under attack, the side-channel leakage is reduced to such a degree that the amount of traces that need to be collected for a successful attack is more than ten times higher compared to no fence present. Moreover, this work presents qualitative results that FPGA cloud providers can consider in order to assess the benefits gained through the deployment of active fence mechanisms within their platforms for multi-tenant services."
                    },
                    {
                        "title": "The Diversification and Enhancement of an IDS Scheme for the Cybersecurity Needs of Modern Supply Chains",
                        "abstract": "Despite the tremendous socioeconomic importance of supply chains (SCs), security officers and operators are faced with no easy and integrated way for protecting their critical, and interconnected, infrastructures from cyber-attacks. As a result, solutions and methodologies that support the detection of malicious activity on SCs are constantly researched into and proposed. Hence, this work presents the implementation of a low-cost reconfigurable intrusion detection system (IDS), on the edge, that can be easily integrated into SC networks, thereby elevating the featured levels of security. Specifically, the proposed system offers real-time cybersecurity intrusion detection over high-speed networks and services by offloading elements of the security check workloads on dedicated reconfigurable hardware. Our solution uses a novel framework that implements the Aho\u2013Corasick algorithm on the reconfigurable fabric of a multi-processor system-on-chip (MPSoC), which supports parallel matching for multiple network packet patterns. The initial performance evaluation of this proof-of-concept shows that it holds the potential to outperform existing software-based solutions while unburdening SC nodes from demanding cybersecurity check workloads. The proposed system performance and its efficiency were evaluated using a real-life environment in the context of European Union\u2019s Horizon 2020 research and innovation program, i.e., CYRENE."
                    },
                    {
                        "title": "Polynomial Networks in Deep Classifiers",
                        "abstract": "Deep neural networks have been the driving force behind the success in classi\ufb01cation tasks, e.g., object and audio recognition. Impressive results and generalization have been achieved by a variety of recently proposed architectures, the majority of which are seemingly disconnected. In this work, we cast the study of deep classi\ufb01ers under a unifying framework. In particular, we express state-of-the-art architectures (e.g., residual and non-local networks) in the form of different degree polynomials of the input. Our framework provides insights on the inductive biases of each model and enables natural extensions building upon their polynomial nature. The ef\ufb01cacy of the proposed models is evaluated on standard image and audio classi\ufb01cation benchmarks. The expressivity of the proposed models is highlighted both in terms of increased model performance as well as model compression. Lastly, the extensions allowed by this taxonomy showcase bene\ufb01ts in the presence of limited data and long-tailed data distributions. We expect this taxonomy to provide links between existing domain-speci\ufb01c architectures."
                    }
                ]
            },
            "8867fcbe-7b02-4c05-8446-b8f0a2357480": {
                "pk": "8867fcbe-7b02-4c05-8446-b8f0a2357480",
                "name": "Markos Georgopoulos",
                "collaborators": [
                    "Yannis Panagakis",
                    "Grigorios G. Chrysos",
                    "James Oldfield",
                    "Tobias Kirschstein",
                    "Simon Giebenhain",
                    "M. Nicolaou",
                    "Martin R\u00fcnz",
                    "Matthias Nie\u00dfner",
                    "Jiankang Deng",
                    "L. Agapito",
                    "M. Nie\u00dfner",
                    "M. Pantic",
                    "Jiapeng Tang",
                    "Yixin Cheng",
                    "V. Cevher",
                    "Christos Tzelepis",
                    "Ioannis Patras",
                    "Adam Polyak",
                    "Amit Zohar",
                    "Andrew Brown",
                    "Andros Tjandra",
                    "Animesh Sinha",
                    "Ann Lee",
                    "Apoorv Vyas",
                    "Bowen Shi",
                    "Chih-Yao Ma",
                    "Ching-Yao Chuang",
                    "David Yan",
                    "Dhruv Choudhary",
                    "Dingkang Wang",
                    "Geet Sethi",
                    "Guan Pang",
                    "Haoyu Ma",
                    "Ishan Misra",
                    "Ji Hou",
                    "Jialiang Wang",
                    "Kiran Jagadeesh",
                    "Kunpeng Li",
                    "Luxin Zhang",
                    "Mannat Singh",
                    "Mary Williamson",
                    "Matt Le",
                    "Matthew Yu",
                    "Mitesh Kumar Singh",
                    "Peizhao Zhang",
                    "Peter Vajda",
                    "Quentin Duval",
                    "Rohit Girdhar",
                    "Roshan Sumbaly",
                    "Sai Saketh Rambhatla",
                    "Sam S. Tsai",
                    "S. Azadi",
                    "Samyak Datta",
                    "Sanyuan Chen",
                    "Sean Bell",
                    "Sharadh Ramaswamy",
                    "Shelly Sheynin",
                    "Siddharth Bhattacharya",
                    "Simran Motwani",
                    "Tao Xu",
                    "Tianhe Li",
                    "Tingbo Hou",
                    "Wei-Ning Hsu",
                    "Xi Yin",
                    "Xiaoliang Dai",
                    "Yaniv Taigman",
                    "Yaqiao Luo",
                    "Yen-Cheng Liu",
                    "Yi-Chiao Wu",
                    "Yue Zhao",
                    "Yuval Kirstain",
                    "Zecheng He",
                    "Zijian He",
                    "Albert Pumarola",
                    "Ali K. Thabet",
                    "A. Sanakoyeu",
                    "Arun Mallya",
                    "Baishan Guo",
                    "Boris Araya",
                    "Breena Kerr",
                    "Carleigh Wood",
                    "Ce Liu",
                    "Cen Peng",
                    "Dimitry Vengertsev",
                    "Edgar Schonfeld",
                    "Elliot Blanchard",
                    "Felix Juefei-Xu",
                    "Fraylie Nord",
                    "Jeff Liang",
                    "John Hoffman",
                    "Jonas Kohler",
                    "Kaolin Fire",
                    "Karthik Sivakumar",
                    "Lawrence Chen",
                    "Licheng Yu",
                    "Luya Gao",
                    "Rashel Moritz",
                    "Sara K. Sampson",
                    "Shikai Li",
                    "Simone Parmeggiani"
                ],
                "domain": [
                    "Generative Models",
                    "3D Reconstruction",
                    "Deep Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "GGHead: Fast and Generalizable 3D Gaussian Heads",
                        "abstract": "Learning 3D head priors from large 2D image collections is an important step towards high-quality 3D-aware human modeling. A core requirement is an efficient architecture that scales well to large-scale datasets and large image resolutions. Unfortunately, existing 3D GANs struggle to scale to generate samples at high resolutions due to their relatively slow train and render speeds, and typically have to rely on 2D superresolution networks at the expense of global 3D consistency. To address these challenges, we propose Generative Gaussian Heads (GGHead), which adopts the recent 3D Gaussian Splatting representation within a 3D GAN framework. To generate a 3D representation, we employ a powerful 2D CNN generator to predict Gaussian attributes in the UV space of a template head mesh. This way, GGHead exploits the regularity of the template's UV layout, substantially facilitating the challenging task of predicting an unstructured set of 3D Gaussians. We further improve the geometric fidelity of the generated 3D representations with a novel total variation loss on rendered UV coordinates. Intuitively, this regularization encourages that neighboring rendered pixels should stem from neighboring Gaussians in the template's UV space. Taken together, our pipeline can efficiently generate 3D heads trained only from single-view 2D image observations. Our proposed framework matches the quality of existing 3D head GANs on FFHQ while being both substantially faster and fully 3D consistent. As a result, we demonstrate real-time generation and rendering of high-quality 3D-consistent heads at $1024^2$ resolution for the first time. Project Website: https://tobias-kirschstein.github.io/gghead"
                    },
                    {
                        "title": "Leveraging the Context through Multi-Round Interactions for Jailbreaking Attacks",
                        "abstract": "Large Language Models (LLMs) are susceptible to Jailbreaking attacks, which aim to extract harmful information by subtly modifying the attack query. As defense mechanisms evolve, directly obtaining harmful information becomes increasingly challenging for Jailbreaking attacks. In this work, inspired from Chomsky's transformational-generative grammar theory and human practices of indirect context to elicit harmful information, we focus on a new attack form, called Contextual Interaction Attack. We contend that the prior context\\u2014the information preceding the attack query\\u2014plays a pivotal role in enabling strong Jailbreaking attacks. Specifically, we propose a first multi-turn approach that leverages benign preliminary questions to interact with the LLM. Due to the autoregressive nature of LLMs, which use previous conversation rounds as context during generation, we guide the model's question-response pair to construct a context that is semantically aligned with the attack query to execute the attack. We conduct experiments on seven different LLMs and demonstrate the efficacy of this attack, which is black-box and can also transfer across LLMs. We believe this can lead to further developments and understanding of security in LLMs."
                    },
                    {
                        "title": "Multilinear Mixture of Experts: Scalable Expert Specialization through Factorization",
                        "abstract": "The Mixture of Experts (MoE) paradigm provides a powerful way to decompose dense layers into smaller, modular computations often more amenable to human interpretation, debugging, and editability. However, a major challenge lies in the computational cost of scaling the number of experts high enough to achieve fine-grained specialization. In this paper, we propose the Multilinear Mixture of Experts ($\\mu$MoE) layer to address this, focusing on vision models. $\\mu$MoE layers enable scalable expert specialization by performing an implicit computation on prohibitively large weight tensors entirely in factorized form. Consequently, $\\mu$MoEs (1) avoid the restrictively high inference-time costs of dense MoEs, yet (2) do not inherit the training issues of the popular sparse MoEs' discrete (non-differentiable) expert routing. We present both qualitative and quantitative evidence that scaling $\\mu$MoE layers when fine-tuning foundation models for vision tasks leads to more specialized experts at the class-level, further enabling manual bias correction in CelebA attribute classification. Finally, we show qualitative results demonstrating the expert specialism achieved when pre-training large GPT2 and MLP-Mixer models with parameter-matched $\\mu$MoE blocks at every layer, maintaining comparable accuracy. Our code is available at: https://github.com/james-oldfield/muMoE."
                    },
                    {
                        "title": "Movie Gen: A Cast of Media Foundation Models",
                        "abstract": "We present Movie Gen, a cast of foundation models that generates high-quality, 1080p HD videos with different aspect ratios and synchronized audio. We also show additional capabilities such as precise instruction-based video editing and generation of personalized videos based on a user's image. Our models set a new state-of-the-art on multiple tasks: text-to-video synthesis, video personalization, video editing, video-to-audio generation, and text-to-audio generation. Our largest video generation model is a 30B parameter transformer trained with a maximum context length of 73K video tokens, corresponding to a generated video of 16 seconds at 16 frames-per-second. We show multiple technical innovations and simplifications on the architecture, latent spaces, training objectives and recipes, data curation, evaluation protocols, parallelization techniques, and inference optimizations that allow us to reap the benefits of scaling pre-training data, model size, and training compute for training large scale media generation models. We hope this paper helps the research community to accelerate progress and innovation in media generation models. All videos from this paper are available at https://go.fb.me/MovieGenResearchVideos."
                    },
                    {
                        "title": "MonoNPHM: Dynamic Head Reconstruction from Monocular Videos",
                        "abstract": "We present Monocular Neural Parametric Head Models (MonoNPHM) for dynamic 3D head reconstructions from monocular RGB videos. To this end, we propose a latent appearance space that parameterizes a texture field on top of a neural parametric model. We constrain predicted color values to be correlated with the underlying geometry such that gradients from RGB effectively influence latent geometry codes during inverse rendering. To increase the representational capacity of our expression space, we augment our backward deformation field with hyper-dimensions, thus improving color and geometry representation in topologically challenging expressions. Using MonoNPHM as a learned prior, we approach the task of 3D head reconstruction using signed distance field based volumetric rendering. By numerically inverting our backward deformation field, we incorporated a landmark loss using facial anchor points that are closely tied to our canonical geometry representation. To evaluate the task of dynamic face reconstruction from monocular RGB videos we record 20 challenging Kinect sequences under casual conditions. MonoNPHM outper-forms all baselines with a significant margin, and makes an important step towards easily accessible neural parametric face models through RGB tracking."
                    },
                    {
                        "title": "HumanRF: High-Fidelity Neural Radiance Fields for Humans in Motion",
                        "abstract": "Representing human performance at high-fidelity is an essential building block in diverse applications, such as film production, computer games or videoconferencing. To close the gap to production-level quality, we introduce HumanRF1, a 4D dynamic neural scene representation that captures full-body appearance in motion from multi-view video input, and enables playback from novel, unseen viewpoints. Our novel representation acts as a dynamic video encoding that captures fine details at high compression rates by factorizing space-time into a temporal matrix-vector decomposition. This allows us to obtain temporally coherent reconstructions of human actors for long sequences, while representing high-resolution details even in the context of challenging motion. While most research focuses on synthesizing at resolutions of 4MP or lower, we address the challenge of operating at 12MP. To this end, we introduce ActorsHQ, a novel multi-view dataset that provides 12MP footage from 160 cameras for 16 sequences with high-fidelity, per-frame mesh reconstructions2. We demonstrate challenges that emerge from using such high-resolution data and show that our newly introduced HumanRF effectively leverages this data, making a significant step towards production-level quality novel view synthesis."
                    },
                    {
                        "title": "Learning Neural Parametric Head Models",
                        "abstract": "We propose a novel 3D morphable model for complete human heads based on hybrid neural fields. At the core of our model lies a neural parametric representation that disentangles identity and expressions in disjoint latent spaces. To this end, we capture a person's identity in a canonical space as a signed distance field (SDF), and model facial expressions with a neural deformation field. In addition, our representation achieves high-fidelity local detail by introducing an ensemble of local fields centered around facial anchor points. To facilitate generalization, we train our model on a newly-captured dataset of over 3700 head scans from 203 different identities using a custom high-end 3D scanning setup. Our dataset significantly exceeds comparable existing datasets, both with respect to quality and completeness of geometry, averaging around 3.5M mesh faces per scan11We will publicly release our dataset along with a public benchmark for both neural head avatar construction as well as an evaluation on a hidden test-set for inference-time fitting.. Finally, we demonstrate that our approach outperforms state-of-the-art methods in terms of fitting error and reconstruction quality."
                    },
                    {
                        "title": "Cluster-guided Image Synthesis with Unconditional Models",
                        "abstract": "Generative Adversarial Networks (GANs) are the driving force behind the state-of-the-art in image generation. Despite their ability to synthesize high-resolution photo-realistic images, generating content with on-demand conditioning of different granularity remains a challenge. This challenge is usually tackled by annotating massive datasets with the attributes of interest, a laborious task that is not always a viable option. Therefore, it is vital to introduce control into the generation process of unsupervised generative models. In this work, we focus on controllable image generation by leveraging GANs that are well-trained in an unsupervised fashion. To this end, we discover that the representation space of intermediate layers of the generator forms a number of clusters that separate the data according to semantically meaningful attributes (e.g., hair color and pose). By conditioning on the cluster assignments, the proposed method is able to control the semantic class of the generated image. Our approach enables sampling from each cluster by Implicit Maximum Likelihood Estimation (IMLE). We showcase the efficacy of our approach on faces (CelebA-HQ and FFHQ), animals (Imagenet) and objects (LSUN) using different pre-trained generative models. The results highlight the ability of our approach to condition image generation on attributes like gender, pose and hair style on faces, as well as a variety of features on different object classes."
                    },
                    {
                        "title": "Polynomial Networks in Deep Classifiers",
                        "abstract": "Deep neural networks have been the driving force behind the success in classi\ufb01cation tasks, e.g., object and audio recognition. Impressive results and generalization have been achieved by a variety of recently proposed architectures, the majority of which are seemingly disconnected. In this work, we cast the study of deep classi\ufb01ers under a unifying framework. In particular, we express state-of-the-art architectures (e.g., residual and non-local networks) in the form of different degree polynomials of the input. Our framework provides insights on the inductive biases of each model and enables natural extensions building upon their polynomial nature. The ef\ufb01cacy of the proposed models is evaluated on standard image and audio classi\ufb01cation benchmarks. The expressivity of the proposed models is highlighted both in terms of increased model performance as well as model compression. Lastly, the extensions allowed by this taxonomy showcase bene\ufb01ts in the presence of limited data and long-tailed data distributions. We expect this taxonomy to provide links between existing domain-speci\ufb01c architectures."
                    },
                    {
                        "title": "Tensor Component Analysis for Interpreting the Latent Space of GANs",
                        "abstract": "This paper addresses the problem of finding interpretable directions in the latent space of pre-trained Generative Adversarial Networks (GANs) to facilitate controllable image synthesis. Such interpretable directions correspond to transformations that can affect both the style and geometry of the synthetic images. However, existing approaches that utilise linear techniques to find these transformations often fail to provide an intuitive way to separate these two sources of variation. To address this, we propose to a) perform a multilinear decomposition of the tensor of intermediate representations, and b) use a tensor-based regression to map directions found using this decomposition to the latent space. Our scheme allows for both linear edits corresponding to the individual modes of the tensor, and non-linear ones that model the multiplicative interactions between them. We show experimentally that we can utilise the former to better separate style- from geometry-based transformations, and the latter to generate an extended set of possible transformations in comparison to prior works. We demonstrate our approach's efficacy both quantitatively and qualitatively compared to the current state-of-the-art."
                    },
                    {
                        "title": "Conditional Generation Using Polynomial Expansions",
                        "abstract": "Generative modeling has evolved to a notable field of machine learning. Deep polynomial neural networks (PNNs) have demonstrated impressive results in unsupervised image generation, where the task is to map an input vector (i.e., noise) to a synthesized image. However, the success of PNNs has not been replicated in conditional generation tasks, such as super-resolution. Existing PNNs focus on single-variable polynomial expansions which do not fare well to two-variable inputs, i.e., the noise variable and the conditional variable. In this work, we introduce a general framework, called CoPE, that enables a polynomial expansion of two input variables and captures their auto- and cross-correlations. We exhibit how CoPE can be trivially augmented to accept an arbitrary number of input variables. CoPE is evaluated in five tasks (class-conditional generation, inverse problems, edges-to-image translation, image-to-image translation, attribute-guided generation) involving eight datasets. The thorough evaluation suggests that CoPE can be useful for tackling diverse conditional generation tasks. The source code of CoPE is available at \\url{https://github.com/grigorisg9gr/polynomial_nets_for_conditional_generation}."
                    },
                    {
                        "title": "Enhancing Facial Data Diversity with Style-based Face Aging",
                        "abstract": "A significant limiting factor in training fair classifiers relates to the presence of dataset bias. In particular, face datasets are typically biased in terms of attributes such as gender, age, and race. If not mitigated, bias leads to algorithms that exhibit unfair behaviour towards such groups. In this work, we address the problem of increasing the diversity of face datasets with respect to age. Concretely, we propose a novel, generative style-based architecture for data augmentation that captures fine-grained aging patterns by conditioning on multi-resolution age-discriminative representations. By evaluating on several age-annotated datasets in both single- and cross-database experiments, we show that the proposed method outperforms state-of-the-art algorithms for age transfer, especially in the case of age groups that lie in the tails of the label distribution. We further show significantly increased diversity in the augmented datasets, outperforming all compared methods according to established metrics."
                    },
                    {
                        "title": "Multilinear Latent Conditioning for Generating Unseen Attribute Combinations",
                        "abstract": "Deep generative models rely on their inductive bias to facilitate generalization, especially for problems with high dimensional data, like images. However, empirical studies have shown that variational autoencoders (VAE) and generative adversarial networks (GAN) lack the generalization ability that occurs naturally in human perception. For example, humans can visualize a woman smiling after only seeing a smiling man. On the contrary, the standard conditional VAE (cVAE) is unable to generate unseen attribute combinations. To this end, we extend cVAE by introducing a multilinear latent conditioning framework that captures the multiplicative interactions between the attributes. We implement two variants of our model and demonstrate their efficacy on MNIST, Fashion-MNIST and CelebA. Altogether, we design a novel conditioning framework that can be used with any architecture to synthesize unseen attribute combinations."
                    },
                    {
                        "title": "Humidity Sensing Properties of Paper Substrates and Their Passivation with ZnO Nanoparticles for Sensor Applications",
                        "abstract": "In this paper, we investigated the effect of humidity on paper substrates and propose a simple and low-cost method for their passivation using ZnO nanoparticles. To this end, we built paper-based microdevices based on an interdigitated electrode (IDE) configuration by means of a mask-less laser patterning method on simple commercial printing papers. Initial resistive measurements indicate that a paper substrate with a porous surface can be used as a cost-effective, sensitive and disposable humidity sensor in the 20% to 70% relative humidity (RH) range. Successive spin-coated layers of ZnO nanoparticles then, control the effect of humidity. Using this approach, the sensors become passive to relative humidity changes, paving the way to the development of ZnO-based gas sensors on paper substrates insensitive to humidity."
                    }
                ]
            },
            "989f741a-42ca-4684-832a-c8edb2ce2b9f": {
                "pk": "989f741a-42ca-4684-832a-c8edb2ce2b9f",
                "name": "Volkan Cevher",
                "collaborators": [
                    "Grigorios G. Chrysos",
                    "Fanghui Liu",
                    "Stratis Skoulakis",
                    "Yongtao Wu",
                    "Mingyi Hong",
                    "Yihang Chen",
                    "George Karypis",
                    "L. Dadi",
                    "El\u00edas Abad-Rocamora",
                    "Kimon Antonakopoulos",
                    "Wanyun Xie",
                    "T. Pethick",
                    "Justin Deschenaux",
                    "Igor Krawczuk",
                    "Yixin Cheng",
                    "Markos Georgopoulos",
                    "Zhiqi Bu",
                    "B. Vinzamuri",
                    "Anil Ramakrishna",
                    "Kai-Wei Chang",
                    "Taiji Suzuki",
                    "Vyas Raina",
                    "Samson Tan",
                    "Aditya Rawal",
                    "Sheng Zha",
                    "P. Olmos",
                    "Ioannis Mavrothalassitis",
                    "Youngsuk Park",
                    "Kailash Budhathoki",
                    "Liangfu Chen",
                    "Jonas M. K\u00fcbler",
                    "Jiaji Huang",
                    "Matth\u00e4us Kleindessner",
                    "Jun Huan",
                    "Yida Wang",
                    "Luca Viano",
                    "Pol Puigdemont",
                    "Arshia Afzal",
                    "Mahsa Shoaran",
                    "Sonny Achten",
                    "F. Tonin",
                    "J. Suykens",
                    "Yiping Lu",
                    "Carl-Johann Simon-Gabriel",
                    "Anton Rodomanov",
                    "Ali Kavis",
                    "Fabian Latorre",
                    "Ali Ramezani-Kebrya",
                    "Kaan Ozkara",
                    "Can Karakus",
                    "Parameswaran Raman",
                    "Shoham Sabach",
                    "B. Kveton"
                ],
                "domain": [
                    "Machine Learning",
                    "Adversarial Robustness",
                    "Neural Networks",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Going beyond Compositions, DDPMs Can Produce Zero-Shot Interpolations",
                        "abstract": "Denoising Diffusion Probabilistic Models (DDPMs) exhibit remarkable capabilities in image generation, with studies suggesting that they can generalize by composing latent factors learned from the training data. In this work, we go further and study DDPMs trained on strictly separate subsets of the data distribution with large gaps on the support of the latent factors. We show that such a model can effectively generate images in the unexplored, intermediate regions of the distribution. For instance, when trained on clearly smiling and non-smiling faces, we demonstrate a sampling procedure which can generate slightly smiling faces without reference images (zero-shot interpolation). We replicate these findings for other attributes as well as other datasets. Our code is available at https://github.com/jdeschena/ddpm-zero-shot-interpolation."
                    },
                    {
                        "title": "Leveraging the Context through Multi-Round Interactions for Jailbreaking Attacks",
                        "abstract": "Large Language Models (LLMs) are susceptible to Jailbreaking attacks, which aim to extract harmful information by subtly modifying the attack query. As defense mechanisms evolve, directly obtaining harmful information becomes increasingly challenging for Jailbreaking attacks. In this work, inspired from Chomsky's transformational-generative grammar theory and human practices of indirect context to elicit harmful information, we focus on a new attack form, called Contextual Interaction Attack. We contend that the prior context\\u2014the information preceding the attack query\\u2014plays a pivotal role in enabling strong Jailbreaking attacks. Specifically, we propose a first multi-turn approach that leverages benign preliminary questions to interact with the LLM. Due to the autoregressive nature of LLMs, which use previous conversation rounds as context during generation, we guide the model's question-response pair to construct a context that is semantically aligned with the attack query to execute the attack. We conduct experiments on seven different LLMs and demonstrate the efficacy of this attack, which is black-box and can also transfer across LLMs. We believe this can lead to further developments and understanding of security in LLMs."
                    },
                    {
                        "title": "Unlearning as multi-task optimization: A normalized gradient difference approach with an adaptive learning rate",
                        "abstract": "Machine unlearning has been used to remove unwanted knowledge acquired by large language models (LLMs). In this paper, we examine machine unlearning from an optimization perspective, framing it as a regularized multi-task optimization problem, where one task optimizes a forgetting objective and another optimizes the model performance. In particular, we introduce a normalized gradient difference (NGDiff) algorithm, enabling us to have better control over the trade-off between the objectives, while integrating a new, automatic learning rate scheduler. We provide a theoretical analysis and empirically demonstrate the superior performance of NGDiff among state-of-the-art unlearning methods on the TOFU and MUSE datasets while exhibiting stable training."
                    },
                    {
                        "title": "High-Dimensional Kernel Methods under Covariate Shift: Data-Dependent Implicit Regularization",
                        "abstract": "This paper studies kernel ridge regression in high dimensions under covariate shifts and analyzes the role of importance re-weighting. We first derive the asymptotic expansion of high dimensional kernels under covariate shifts. By a bias-variance decomposition, we theoretically demonstrate that the re-weighting strategy allows for decreasing the variance. For bias, we analyze the regularization of the arbitrary or well-chosen scale, showing that the bias can behave very differently under different regularization scales. In our analysis, the bias and variance can be characterized by the spectral decay of a data-dependent regularized kernel: the original kernel matrix associated with an additional re-weighting matrix, and thus the re-weighting strategy can be regarded as a data-dependent regularization for better understanding. Besides, our analysis provides asymptotic expansion of kernel functions/vectors under covariate shift, which has its own interest."
                    },
                    {
                        "title": "Extreme Miscalibration and the Illusion of Adversarial Robustness",
                        "abstract": "Deep learning-based Natural Language Processing (NLP) models are vulnerable to adversarial attacks, where small perturbations can cause a model to misclassify. Adversarial Training (AT) is often used to increase model robustness. However, we have discovered an intriguing phenomenon: deliberately or accidentally miscalibrating models masks gradients in a way that interferes with adversarial attack search methods, giving rise to an apparent increase in robustness. We show that this observed gain in robustness is an illusion of robustness (IOR), and demonstrate how an adversary can perform various forms of test-time temperature calibration to nullify the aforementioned interference and allow the adversarial attack to find adversarial examples. Hence, we urge the NLP community to incorporate test-time temperature scaling into their robustness evaluations to ensure that any observed gains are genuine. Finally, we show how the temperature can be scaled during \\textit{training} to improve genuine robustness."
                    },
                    {
                        "title": "Learning with Norm Constrained, Over-parameterized, Two-layer Neural Networks",
                        "abstract": "Recent studies show that a reproducing kernel Hilbert space (RKHS) is not a suitable space to model functions by neural networks as the curse of dimensionality (CoD) cannot be evaded when trying to approximate even a single ReLU neuron (Bach, 2017). In this paper, we study a suitable function space for over-parameterized two-layer neural networks with bounded norms (e.g., the path norm, the Barron norm) in the perspective of sample complexity and generalization properties. First, we show that the path norm (as well as the Barron norm) is able to obtain width-independence sample complexity bounds, which allows for uniform convergence guarantees. Based on this result, we derive the improved result of metric entropy for $\\epsilon$-covering up to $O(\\epsilon^{-\\frac{2d}{d+2}})$ ($d$ is the input dimension and the depending constant is at most linear order of $d$) via the convex hull technique, which demonstrates the separation with kernel methods with $\\Omega(\\epsilon^{-d})$ to learn the target function in a Barron space. Second, this metric entropy result allows for building a sharper generalization bound under a general moment hypothesis setting, achieving the rate at $O(n^{-\\frac{d+2}{2d+2}})$. Our analysis is novel in that it offers a sharper and refined estimation for metric entropy with a linear dimension dependence and unbounded sampling in the estimation of the sample error and the output error."
                    },
                    {
                        "title": "Efficient local linearity regularization to overcome catastrophic overfitting",
                        "abstract": "Catastrophic overfitting (CO) in single-step adversarial training (AT) results in abrupt drops in the adversarial test accuracy (even down to 0%). For models trained with multi-step AT, it has been observed that the loss function behaves locally linearly with respect to the input, this is however lost in single-step AT. To address CO in single-step AT, several methods have been proposed to enforce local linearity of the loss via regularization. However, these regularization terms considerably slow down training due to Double Backpropagation. Instead, in this work, we introduce a regularization term, called ELLE, to mitigate CO effectively and efficiently in classical AT evaluations, as well as some more difficult regimes, e.g., large adversarial perturbations and long training schedules. Our regularization term can be theoretically linked to curvature of the loss function and is computationally cheaper than previous methods by avoiding Double Backpropagation. Our thorough experimental validation demonstrates that our work does not suffer from CO, even in challenging settings where previous works suffer from it. We also notice that adapting our regularization parameter during training (ELLE-A) greatly improves the performance, specially in large $\\epsilon$ setups. Our implementation is available in https://github.com/LIONS-EPFL/ELLE ."
                    },
                    {
                        "title": "Efficient Continual Finite-Sum Minimization",
                        "abstract": "Given a sequence of functions $f_1,\\ldots,f_n$ with $f_i:\\mathcal{D}\\mapsto \\mathbb{R}$, finite-sum minimization seeks a point ${x}^\\star \\in \\mathcal{D}$ minimizing $\\sum_{j=1}^n f_j(x)/n$. In this work, we propose a key twist into the finite-sum minimization, dubbed as continual finite-sum minimization, that asks for a sequence of points ${x}_1^\\star,\\ldots,{x}_n^\\star \\in \\mathcal{D}$ such that each ${x}^\\star_i \\in \\mathcal{D}$ minimizes the prefix-sum $\\sum_{j=1}^if_j(x)/i$. Assuming that each prefix-sum is strongly convex, we develop a first-order continual stochastic variance reduction gradient method ($\\mathrm{CSVRG}$) producing an $\\epsilon$-optimal sequence with $\\mathcal{\\tilde{O}}(n/\\epsilon^{1/3} + 1/\\sqrt{\\epsilon})$ overall first-order oracles (FO). An FO corresponds to the computation of a single gradient $\\nabla f_j(x)$ at a given $x \\in \\mathcal{D}$ for some $j \\in [n]$. Our approach significantly improves upon the $\\mathcal{O}(n/\\epsilon)$ FOs that $\\mathrm{StochasticGradientDescent}$ requires and the $\\mathcal{O}(n^2 \\log (1/\\epsilon))$ FOs that state-of-the-art variance reduction methods such as $\\mathrm{Katyusha}$ require. We also prove that there is no natural first-order method with $\\mathcal{O}\\left(n/\\epsilon^\\alpha\\right)$ gradient complexity for $\\alpha<1/4$, establishing that the first-order complexity of our method is nearly tight."
                    },
                    {
                        "title": "Inference Optimization of Foundation Models on AI Accelerators",
                        "abstract": "Powerful foundation models, including large language models (LLMs), with Transformer architectures have ushered in a new era of Generative AI across various industries. Industry and research community have witnessed a large number of new applications, based on those foundation models. Such applications include question and answer, customer services, image and video generation, and code completions, among others. However, as the number of model parameters reaches to hundreds of billions, their deployment incurs prohibitive inference costs and high latency in real-world scenarios. As a result, the demand for cost-effective and fast inference using AI accelerators is ever more higher. To this end, our tutorial offers a comprehensive discussion on complementary inference optimization techniques using AI accelerators. Beginning with an overview of basic Transformer architectures and deep learning system frameworks, we deep dive into system optimization techniques for fast and memory-efficient attention computations and discuss how they can be implemented efficiently on AI accelerators. Next, we describe architectural elements that are key for fast transformer inference. Finally, we examine various model compression and fast decoding strategies in the same context."
                    },
                    {
                        "title": "Imitation Learning in Discounted Linear MDPs without exploration assumptions",
                        "abstract": "We present a new algorithm for imitation learning in infinite horizon linear MDPs dubbed ILARL which greatly improves the bound on the number of trajectories that the learner needs to sample from the environment. In particular, we remove exploration assumptions required in previous works and we improve the dependence on the desired accuracy $\\epsilon$ from $\\mathcal{O}(\\epsilon^{-5})$ to $\\mathcal{O}(\\epsilon^{-4})$. Our result relies on a connection between imitation learning and online learning in MDPs with adversarial losses. For the latter setting, we present the first result for infinite horizon linear MDP which may be of independent interest. Moreover, we are able to provide a strengthen result for the finite horizon case where we achieve $\\mathcal{O}(\\epsilon^{-2})$. Numerical experiments with linear function approximation shows that ILARL outperforms other commonly used algorithms."
                    },
                    {
                        "title": "Learning to Remove Cuts in Integer Linear Programming",
                        "abstract": "Cutting plane methods are a fundamental approach for solving integer linear programs (ILPs). In each iteration of such methods, additional linear constraints (cuts) are introduced to the constraint set with the aim of excluding the previous fractional optimal solution while not affecting the optimal integer solution. In this work, we explore a novel approach within cutting plane methods: instead of only adding new cuts, we also consider the removal of previous cuts introduced at any of the preceding iterations of the method under a learnable parametric criteria. We demonstrate that in fundamental combinatorial optimization settings such cut removal policies can lead to significant improvements over both human-based and machine learning-guided cut addition policies even when implemented with simple models."
                    },
                    {
                        "title": "REST: Efficient and Accelerated EEG Seizure Analysis through Residual State Updates",
                        "abstract": "EEG-based seizure detection models face challenges in terms of inference speed and memory efficiency, limiting their real-time implementation in clinical devices. This paper introduces a novel graph-based residual state update mechanism (REST) for real-time EEG signal analysis in applications such as epileptic seizure detection. By leveraging a combination of graph neural networks and recurrent structures, REST efficiently captures both non-Euclidean geometry and temporal dependencies within EEG data. Our model demonstrates high accuracy in both seizure detection and classification tasks. Notably, REST achieves a remarkable 9-fold acceleration in inference speed compared to state-of-the-art models, while simultaneously demanding substantially less memory than the smallest model employed for this task. These attributes position REST as a promising candidate for real-time implementation in clinical devices, such as Responsive Neurostimulation or seizure alert systems."
                    },
                    {
                        "title": "HeNCler: Node Clustering in Heterophilous Graphs through Learned Asymmetric Similarity",
                        "abstract": "Clustering nodes in heterophilous graphs presents unique challenges due to the asymmetric relationships often overlooked by traditional methods, which moreover assume that good clustering corresponds to high intra-cluster and low inter-cluster connectivity. To address these issues, we introduce HeNCler - a novel approach for Heterophilous Node Clustering. Our method begins by defining a weighted kernel singular value decomposition to create an asymmetric similarity graph, applicable to both directed and undirected graphs. We further establish that the dual problem of this formulation aligns with asymmetric kernel spectral clustering, interpreting learned graph similarities without relying on homophily. We demonstrate the ability to solve the primal problem directly, circumventing the computational difficulties of the dual approach. Experimental evidence confirms that HeNCler significantly enhances performance in node clustering tasks within heterophilous graph contexts."
                    },
                    {
                        "title": "Generalization of Scaled Deep ResNets in the Mean-Field Regime",
                        "abstract": "Despite the widespread empirical success of ResNet, the generalization properties of deep ResNet are rarely explored beyond the lazy training regime. In this work, we investigate \\emph{scaled} ResNet in the limit of infinitely deep and wide neural networks, of which the gradient flow is described by a partial differential equation in the large-neural network limit, i.e., the \\emph{mean-field} regime. To derive the generalization bounds under this setting, our analysis necessitates a shift from the conventional time-invariant Gram matrix employed in the lazy training regime to a time-variant, distribution-dependent version. To this end, we provide a global lower bound on the minimum eigenvalue of the Gram matrix under the mean-field regime. Besides, for the traceability of the dynamic of Kullback-Leibler (KL) divergence, we establish the linear convergence of the empirical error and estimate the upper bound of the KL divergence over parameters distribution. Finally, we build the uniform convergence for generalization bound via Rademacher complexity. Our results offer new insights into the generalization ability of deep ResNet beyond the lazy training regime and contribute to advancing the understanding of the fundamental properties of deep neural networks."
                    },
                    {
                        "title": "Revisiting character-level adversarial attacks",
                        "abstract": "Adversarial attacks in Natural Language Processing apply perturbations in the character or token levels. Token-level attacks, gaining prominence for their use of gradient-based methods, are susceptible to altering sentence semantics, leading to invalid adversarial examples. While character-level attacks easily maintain semantics, they have received less attention as they cannot easily adopt popular gradient-based methods, and are thought to be easy to defend. Challenging these beliefs, we introduce Charmer, an efficient query-based adversarial attack capable of achieving high attack success rate (ASR) while generating highly similar adversarial examples. Our method successfully targets both small (BERT) and large (Llama 2) models. Specifically, on BERT with SST-2, Charmer improves the ASR in 4.84% points and the USE similarity in 8% points with respect to the previous art. Our implementation is available in https://github.com/LIONS-EPFL/Charmer."
                    },
                    {
                        "title": "Robust NAS under adversarial training: benchmark, theory, and beyond",
                        "abstract": "Recent developments in neural architecture search (NAS) emphasize the significance of considering robust architectures against malicious data. However, there is a notable absence of benchmark evaluations and theoretical guarantees for searching these robust architectures, especially when adversarial training is considered. In this work, we aim to address these two challenges, making twofold contributions. First, we release a comprehensive data set that encompasses both clean accuracy and robust accuracy for a vast array of adversarially trained networks from the NAS-Bench-201 search space on image datasets. Then, leveraging the neural tangent kernel (NTK) tool from deep learning theory, we establish a generalization theory for searching architecture in terms of clean accuracy and robust accuracy under multi-objective adversarial training. We firmly believe that our benchmark and theoretical insights will significantly benefit the NAS community through reliable reproducibility, efficient assessment, and theoretical foundation, particularly in the pursuit of robust architectures."
                    },
                    {
                        "title": "Universal Gradient Methods for Stochastic Convex Optimization",
                        "abstract": "We develop universal gradient methods for Stochastic Convex Optimization (SCO). Our algorithms automatically adapt not only to the oracle's noise but also to the H\\\"older smoothness of the objective function without a priori knowledge of the particular setting. The key ingredient is a novel strategy for adjusting step-size coefficients in the Stochastic Gradient Method (SGD). Unlike AdaGrad, which accumulates gradient norms, our Universal Gradient Method accumulates appropriate combinations of gradient- and iterate differences. The resulting algorithm has state-of-the-art worst-case convergence rate guarantees for the entire H\\\"older class including, in particular, both nonsmooth functions and those with Lipschitz continuous gradient. We also present the Universal Fast Gradient Method for SCO enjoying optimal efficiency estimates."
                    },
                    {
                        "title": "Improving SAM Requires Rethinking its Optimization Formulation",
                        "abstract": "This paper rethinks Sharpness-Aware Minimization (SAM), which is originally formulated as a zero-sum game where the weights of a network and a bounded perturbation try to minimize/maximize, respectively, the same differentiable loss. To fundamentally improve this design, we argue that SAM should instead be reformulated using the 0-1 loss. As a continuous relaxation, we follow the simple conventional approach where the minimizing (maximizing) player uses an upper bound (lower bound) surrogate to the 0-1 loss. This leads to a novel formulation of SAM as a bilevel optimization problem, dubbed as BiSAM. BiSAM with newly designed lower-bound surrogate loss indeed constructs stronger perturbation. Through numerical evidence, we show that BiSAM consistently results in improved performance when compared to the original SAM and variants, while enjoying similar computational complexity. Our code is available at https://github.com/LIONS-EPFL/BiSAM."
                    },
                    {
                        "title": "Mixed Nash for Robust Federated Learning",
                        "abstract": "We study robust federated learning (FL) within a game theoretic framework to alleviate the server vulnerabilities to even an informed adversary who can tailor training-time attacks (Fang et al., 2020; Xie et al., 2020a; Ozfatura et al., 2022; Rodr\u00edguez-Barroso et al., 2023). Specifically, we introduce RobustTailor, a simulation-based framework that prevents the adversary from being omniscient and derives its convergence guarantees. RobustTailor improves robustness to training-time attacks significantly with a minor trade-off of privacy. Empirical results under challenging attacks show that RobustTailor performs close to an upper bound with perfect knowledge of honest clients."
                    },
                    {
                        "title": "MADA: Meta-Adaptive Optimizers through hyper-gradient Descent",
                        "abstract": "Following the introduction of Adam, several novel adaptive optimizers for deep learning have been proposed. These optimizers typically excel in some tasks but may not outperform Adam uniformly across all tasks. In this work, we introduce Meta-Adaptive Optimizers (MADA), a unified optimizer framework that can generalize several known optimizers and dynamically learn the most suitable one during training. The key idea in MADA is to parameterize the space of optimizers and dynamically search through it using hyper-gradient descent during training. We empirically compare MADA to other popular optimizers on vision and language tasks, and find that MADA consistently outperforms Adam and other popular optimizers, and is robust against sub-optimally tuned hyper-parameters. MADA achieves a greater validation performance improvement over Adam compared to other popular optimizers during GPT-2 training and fine-tuning. We also propose AVGrad, a modification of AMSGrad that replaces the maximum operator with averaging, which is more suitable for hyper-gradient optimization. Finally, we provide a convergence analysis to show that parameterized interpolations of optimizers can improve their error bounds (up to constants), hinting at an advantage for meta-optimizers."
                    }
                ]
            }
        }
    },
    "2310.02984": {
        "paper_data": {
            "title": "Scaling Laws for Associative Memories",
            "url": "http://arxiv.org/abs/2310.02984v2",
            "arxiv_id": "2310.02984",
            "authors": [
                "Vivien Cabannes",
                "Elvis Dohmatob",
                "Alberto Bietti"
            ],
            "abstract": "Learning arguably involves the discovery and memorization of abstract rules. The aim of this paper is to study associative memory mechanisms. Our model is based on high-dimensional matrices consisting of outer products of embeddings, which relates to the inner layers of transformer language models. We derive precise scaling laws with respect to sample size and parameter size, and discuss the statistical efficiency of different estimators, including optimization-based algorithms. We provide extensive numerical experiments to validate and interpret theoretical results, including fine-grained visualizations of the stored memory associations.",
            "introduction": "   1 Introduction  As the scale of large language models (LLMs) keeps increasing, scaling laws have become a crucial tool to empirically assess and predict the behavior of these models when varying the number of parameters and training data\u00a0(Kaplan et\u00a0al., 2020; Hoffmann et\u00a0al., 2022). Despite their practical impact, the underlying phenomena leading to such scaling laws remain poorly understood. A better understanding of such phenomena could guide researchers towards improved models, algorithms, and datasets which may lead to improved scaling laws.   Our study focuses on a simple model that aims to be representative of LLMs in two ways. First, we focus on heavy-tailed data distributions over discrete tokens, a natural assumption for text data\u00a0(Piantadosi, 2014). Second, we consider associative memory models that store input-output pairs through outer-products of finite-dimensional embeddings, and can be seen as a proxy of the intermediate layers of transformers. Indeed, some transformer layers have been found to behave as key-value memories\u00a0(Geva et\u00a0al., 2021; Meng et\u00a0al., 2022), and more generally outer-product associative memory matrices arise naturally from training dynamics on intermediate weights\u00a0(Bietti et\u00a0al., 2023). Beyond simple associative recall, the combination of multiple such associative rules at different layers may lead to certain circuits with rich \u201creasoning\u201d behaviors based on context\u00a0(Elhage et\u00a0al., 2021; Bietti et\u00a0al., 2023; Michaud et\u00a0al., 2023). For example, an intermediate layer input token may encode for the topic \u201clinux\u201d, leading to an output token that will trigger a specific behavior in the transformer\u2019s following layers when processing the token \u201cterminal\u201d.   Our contributions are as follows:   \u2022  We provide precise statistical rates for outer-product memories with random embeddings, and compare different memory storage schemes in the context of Zipf-distributed data.    \u2022  We compare theoretical schemes to the weights learned by various optimization algorithms used in practice, and illustrate the role of different design choices with numerical experiments.      Related work.  Associative memory models have a long history in the literature on neural computation (Steinbuch, 1961; Willshaw et\u00a0al., 1969; Longuet-Higgins et\u00a0al., 1970; Kohonen, 1972; Amari, 1972; Little, 1974; Hopfield, 1982; Smolensky, 1990; Schlag et\u00a0al., 2021; Valle-Lisboa et\u00a0al., 2023), though the statistical insights we provide for continuous-values random embeddings and heavy-tailed tokens distributions are new, to the best of our knowledge. Memorization behaviors have drawn a lot of attention recently, and are believed to be an important notion to understand the learning happening in deep neural network (e.g., Sukhbaatar et\u00a0al., 2019; Feldman, 2020; Feldman & Zhang, 2020; Geva et\u00a0al., 2021; Wu et\u00a0al., 2022). Building on memorization and heavy-tailed discrete data, our model bears similarities to the ones of Hutter (2021), Michaud et\u00a0al. (2023) or Debowski (2023), although we focus on practical models with finite capacity. The discrete nature of tokens contrasts with other recent works on scaling laws that have focused on continuous Gaussian inputs\u00a0(e.g., Bahri et\u00a0al., 2021; Maloney et\u00a0al., 2022; Sorscher et\u00a0al., 2022).       2 Model for Associative Memory  The data.  In the following, we consider a joint distribution p\u2208\u0394[N]\u00d7[M]\ud835\udc5dsubscript\u0394delimited-[]\ud835\udc41delimited-[]\ud835\udc40p\\in\\Delta_{[N]\\times[M]}italic_p \u2208 roman_\u0394 start_POSTSUBSCRIPT [ italic_N ] \u00d7 [ italic_M ] end_POSTSUBSCRIPT on inputs x\u2208[N]\ud835\udc65delimited-[]\ud835\udc41x\\in[N]italic_x \u2208 [ italic_N ] and outputs y\u2208[M]\ud835\udc66delimited-[]\ud835\udc40y\\in[M]italic_y \u2208 [ italic_M ]. The inputs and outputs are respectively assumed to solely take N\ud835\udc41Nitalic_N and M\ud835\udc40Mitalic_M discrete values respectively. For example, N\ud835\udc41Nitalic_N could be the number of potential sequences of fixed word",
            "references": [
                {
                    "title": "Tensor Programs IVb: Adaptive Optimization in the Infinite-Width Limit",
                    "abstract": "Going beyond stochastic gradient descent (SGD), what new phenomena emerge in wide neural networks trained by adaptive optimizers like Adam? Here we show: The same dichotomy between feature learning and kernel behaviors (as in SGD) holds for general optimizers as well, including Adam -- albeit with a nonlinear notion of\"kernel.\"We derive the corresponding\"neural tangent\"and\"maximal update\"limits for any architecture. Two foundational advances underlie the above results: 1) A new Tensor Program language, NEXORT, that can express how adaptive optimizers process gradients into updates. 2) The introduction of bra-ket notation to drastically simplify expressions and calculations in Tensor Programs. This work summarizes and generalizes all previous results in the Tensor Programs series of papers."
                },
                {
                    "title": "Birth of a Transformer: A Memory Viewpoint",
                    "abstract": "Large language models based on transformers have achieved great empirical successes. However, as they are deployed more widely, there is a growing need to better understand their internal mechanisms in order to make them more reliable. These models appear to store vast amounts of knowledge from their training data, and to adapt quickly to new information provided in their context or prompt. We study how transformers balance these two types of knowledge by considering a synthetic setup where tokens are generated from either global or context-specific bigram distributions. By a careful empirical analysis of the training process on a simplified two-layer transformer, we illustrate the fast learning of global bigrams and the slower development of an\"induction head\"mechanism for the in-context bigrams. We highlight the role of weight matrices as associative memories, provide theoretical insights on how gradients enable their learning during training, and study the role of data-distributional properties."
                },
                {
                    "title": "Noise Is Not the Main Factor Behind the Gap Between SGD and Adam on Transformers, but Sign Descent Might Be",
                    "abstract": "The success of the Adam optimizer on a wide array of architectures has made it the default in settings where stochastic gradient descent (SGD) performs poorly. However, our theoretical understanding of this discrepancy is lagging, preventing the development of significant improvements on either algorithm. Recent work advances the hypothesis that Adam and other heuristics like gradient clipping outperform SGD on language tasks because the distribution of the error induced by sampling has heavy tails. This suggests that Adam outperform SGD because it uses a more robust gradient estimate. We evaluate this hypothesis by varying the batch size, up to the entire dataset, to control for stochasticity. We present evidence that stochasticity and heavy-tailed noise are not major factors in the performance gap between SGD and Adam. Rather, Adam performs better as the batch size increases, while SGD is less effective at taking advantage of the reduction in noise. This raises the question as to why Adam outperforms SGD in the full-batch setting. Through further investigation of simpler variants of SGD, we find that the behavior of Adam with large batches is similar to sign descent with momentum."
                },
                {
                    "title": "The Quantization Model of Neural Scaling",
                    "abstract": "We propose the Quantization Model of neural scaling laws, explaining both the observed power law dropoff of loss with model and data size, and also the sudden emergence of new capabilities with scale. We derive this model from what we call the Quantization Hypothesis, where network knowledge and skills are\"quantized\"into discrete chunks ($\\textbf{quanta}$). We show that when quanta are learned in order of decreasing use frequency, then a power law in use frequencies explains observed power law scaling of loss. We validate this prediction on toy datasets, then study how scaling curves decompose for large language models. Using language model gradients, we automatically decompose model behavior into a diverse set of skills (quanta). We tentatively find that the frequency at which these quanta are used in the training distribution roughly follows a power law corresponding with the empirical scaling exponent for language models, a prediction of our theory."
                },
                {
                    "title": "A Simplistic Model of Neural Scaling Laws: Multiperiodic Santa Fe Processes",
                    "abstract": "It was observed that large language models exhibit a power-law decay of cross entropy with respect to the number of parameters and training tokens. When extrapolated literally, this decay implies that the entropy rate of natural language is zero. To understand this phenomenon -- or an artifact -- better, we construct a simple stationary stochastic process and its memory-based predictor that exhibit a power-law decay of cross entropy with the vanishing entropy rate. Our example is based on previously discussed Santa Fe processes, which decompose a random text into a process of narration and time-independent knowledge. Previous discussions assumed that narration is a memoryless source with Zipf's distribution. In this paper, we propose a model of narration that has the vanishing entropy rate and applies a randomly chosen deterministic sequence called a multiperiodic sequence. Under a suitable parameterization, multiperiodic sequences exhibit asymptotic relative frequencies given by Zipf's law. Remaining agnostic about the value of the entropy rate of natural language, we discuss relevance of similar constructions for language modeling."
                },
                {
                    "title": "A Solvable Model of Neural Scaling Laws",
                    "abstract": "Large language models with a huge number of parameters, when trained on near internet-sized number of tokens, have been empirically shown to obey neural scaling laws: specifically, their performance behaves predictably as a power law in either parameters or dataset size until bottlenecked by the other resource. To understand this better, we first identify the necessary properties allowing such scaling laws to arise and then propose a statistical model \u2013 a joint generative data model and random feature model \u2013 that captures this neural scaling phenomenology. By solving this model in the dual limit of large training set size and large number of parameters, we gain insight into (i) the statistical structure of datasets and tasks that lead to scaling laws, (ii) the way nonlinear feature maps, such as those provided by neural networks, enable scaling laws when trained on these datasets, (iii) the optimality of the equiparameterization scaling of training sets and parameters, and (iv) whether such scaling laws can break down and how they behave when they do. Key findings are the manner in which the power laws that occur in the statistics of natural datasets are extended by nonlinear random feature maps and then translated into power-law scalings of the test loss and how the finite extent of the data\u2019s spectral power law causes the model\u2019s performance to plateau. ? Equal contribution. ar X iv :2 21 0. 16 85 9v 1 [ cs .L G ] 3 0 O ct 2 02 2"
                },
                {
                    "title": "Toy Models of Superposition",
                    "abstract": "Neural networks often pack many unrelated concepts into a single neuron \u2013 a puzzling phenomenon known as 'polysemanticity' which makes interpretability much more challenging. This paper provides a toy model where polysemanticity can be fully understood, arising as a result of models storing additional sparse features in \"superposition.\" We demonstrate the existence of a phase change, a surprising connection to the geometry of uniform polytopes, and evidence of a link to adversarial examples. We also discuss potential implications for mechanistic interpretability. We recommend reading this paper as an HTML article. It would be very convenient if the individual neurons of artificial neural networks corresponded to cleanly interpretable features of the input. For example, in an \u201cideal\u201d ImageNet classifier, each neuron would fire only in the presence of a specific visual feature, such as the color red, a leftfacing curve, or a dog snout. Empirically, in models we have studied, some of the neurons do cleanly map to features. But it isn't always the case that features correspond so cleanly to neurons, especially in large language models where it actually seems rare for neurons to correspond to clean features. This brings up many questions. Why is it that neurons sometimes align with features and sometimes don't? Why do some models and tasks have many of these clean neurons, while they're vanishingly rare in others? In this paper, we use toy models \u2014 small ReLU networks trained on synthetic data with sparse input features \u2014 to investigate how and when models represent more features than they have dimensions. We call this phenomenon superposition. When features are sparse, superposition allows compression beyond what a linear model would do, at the cost of \"interference\" that requires nonlinear filtering. Consider a toy model where we train an embedding of five features of varying importance in two dimensions, add a ReLU afterwards for filtering, and vary the sparsity of the features. With dense features, the model learns to represent an orthogonal basis of the most important two features (similar to what Principal Component Analysis might give us), and the other three features are not represented. But if we make the features sparse, this changes: 1 This figure and a few others can be reproduced using the toy model framework Colab notebook in our Github repo Not only can models store additional features in superposition by tolerating some interference, but we'll show that, at least in certain limited cases, models can perform computation while in superposition. (In particular, we'll show that models can put simple circuits computing the absolute value function in superposition.) This leads us to hypothesize that the neural networks we observe in practice are in some sense noisily simulating larger, highly sparse networks. In other words, it's possible that models we train can be thought of as doing \u201cthe same thing as\u201d an imagined muchlarger model, representing the exact same features but with no interference. Feature superposition isn't a novel idea. A number of previous interpretability papers have speculated about it , and it's very closely related to the long-studied topic of compressed sensing in mathematics , as well as the ideas of distributed, dense, and population codes in neuroscience and deep learning . What, then, is the contribution of this paper? For interpretability researchers, our main contribution is providing a direct demonstration that superposition occurs in artificial neural networks given a relatively natural setup, suggesting this may also occur in practice. We offer a theory of when and why this occurs, revealing a phase diagram for superposition. We also discover that, at least in our toy model, superposition exhibits complex geometric structure. But our results may also be of broader interest. We find preliminary evidence that superposition may be linked to adversarial examples and grokking, and might also suggest a theory for the performance of mixture of experts models. More broadly, the toy model we investigate has unexpectedly rich structure, exhibiting phase changes, a geometric structure based on uniform polytopes, \"energy level\"-like jumps during training, and a phenomenon which is qualitatively similar to the fractional quantum Hall effect in physics. We originally investigated the subject to gain understanding of cleanly-interpretable neurons in larger models, but we've found these toy models to be surprisingly interesting in their own right. [1, 2] [3] [4] [5] KEY RESULTS FROM OUR TOY MODELS In our toy models, we are able to demonstrate that: Superposition is a real, observed phenomenon. Both monosemantic and polysemantic neurons can form. At least some kinds of computation can be performed in superposition. Whether features are stored in superposition is governed by a phase change. Superposition organizes features into geometric structures such as digons, triangles, pentagons, and tetrahedrons. Our toy models are simple ReLU networks, so it seems fair to say that neural networks exhibit these properties in at least some regimes, but it's very unclear what to generalize to real networks. Definitions and Motivation: Features, Directions, and Superposition In our work, we often think of neural networks as having features of the input represented as directions in activation space. This isn't a trivial claim. It isn't obvious what kind of structure we should expect neural network representations to have. When we say something like \"word embeddings have a gender direction\" or \"vision models have curve detector neurons\", one is implicitly making strong claims about the structure of network representations. Despite this, we believe this kind of \"linear representation hypothesis\" is supported both by significant empirical findings and theoretical arguments. One might think of this as two separate properties, which we'll explore in more detail shortly: Decomposability: Network representations can be described in terms of independently understandable features. Linearity: Features are represented by direction. If we hope to reverse engineer neural networks, we need a property like decomposability. Decomposability is what allows us to reason about the model without fitting the whole thing in our heads! But it's not enough for things to be decomposable: we need to be able to access the decomposition somehow. In order to do this, we need to identify the individual features within a representation. In a linear representation, this corresponds to determining which directions in activation space correspond to which independent features of the input. Sometimes, identifying feature directions is very easy because features seem to correspond to neurons. For example, many neurons in the early layers of InceptionV1 clearly correspond to features (e.g. curve detector neurons ). Why is it that we sometimes get this extremely helpful property, but in other cases don't? We hypothesize that there are really two countervailing forces driving this: [6] Privileged Basis: Only some representations have a privileged basis which encourages features to align with basis directions (i.e. to correspond to neurons). Superposition: Linear representations can represent more features than dimensions, using a strategy we call superposition. This can be seen as neural networks simulating larger networks. This pushes features away from corresponding to neurons. Superposition has been hypothesized in previous work  . However, we're not aware of feature superposition having been unambiguously demonstrated to occur in neural networks before ( demonstrates a closely related phenomenon of model superposition). The goal of this paper is to change that, demonstrating superposition and exploring how it interacts with privileged bases. If superposition occurs in networks, it deeply influences what approaches to interpretability research make sense, so unambiguous demonstration seems important. [1, 2] [7] The goal of this section will be to motivate these ideas and unpack them in detail. It's worth noting that many of the ideas in this section have close connections to ideas in other lines of interpretability research (especially disentanglement), neuroscience (distributed representations, population codes, etc), compressed sensing, and many other lines of work. This section will focus on articulating our perspective on the problem. We'll discuss these other lines of work in detail in Related Work."
                },
                {
                    "title": "Beyond neural scaling laws: beating power law scaling via data pruning",
                    "abstract": "Widely observed neural scaling laws, in which error falls off as a power of the training set size, model size, or both, have driven substantial performance improvements in deep learning. However, these improvements through scaling alone require considerable costs in compute and energy. Here we focus on the scaling of error with dataset size and show how in theory we can break beyond power law scaling and potentially even reduce it to exponential scaling instead if we have access to a high-quality data pruning metric that ranks the order in which training examples should be discarded to achieve any pruned dataset size. We then test this improved scaling prediction with pruned dataset size empirically, and indeed observe better than power law scaling in practice on ResNets trained on CIFAR-10, SVHN, and ImageNet. Next, given the importance of finding high-quality pruning metrics, we perform the first large-scale benchmarking study of ten different data pruning metrics on ImageNet. We find most existing high performing metrics scale poorly to ImageNet, while the best are computationally intensive and require labels for every image. We therefore developed a new simple, cheap and scalable self-supervised pruning metric that demonstrates comparable performance to the best supervised metrics. Overall, our work suggests that the discovery of good data-pruning metrics may provide a viable path forward to substantially improved neural scaling laws, thereby reducing the resource costs of modern deep learning."
                },
                {
                    "title": "Training Compute-Optimal Large Language Models",
                    "abstract": "We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over 400 language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, Chinchilla, that uses the same compute budget as Gopher but with 70B parameters and 4$\\times$ more more data. Chinchilla uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that Chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, Chinchilla reaches a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, greater than a 7% improvement over Gopher."
                },
                {
                    "title": "Memorizing Transformers",
                    "abstract": "Language models typically need to be trained or finetuned in order to acquire new knowledge, which involves updating their weights. We instead envision language models that can simply read and memorize new data at inference time, thus acquiring new knowledge immediately. In this work, we extend language models with the ability to memorize the internal representations of past inputs. We demonstrate that an approximate kNN lookup into a non-differentiable memory of recent (key, value) pairs improves language modeling across various benchmarks and tasks, including generic webtext (C4), math papers (arXiv), books (PG-19), code (Github), as well as formal theorems (Isabelle). We show that the performance steadily improves when we increase the size of memory up to 262K tokens. On benchmarks including code and mathematics, we find that the model is capable of making use of newly defined functions and theorems during test time."
                },
                {
                    "title": "Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer",
                    "abstract": "Hyperparameter (HP) tuning in deep learning is an expensive process, prohibitively so for neural networks (NNs) with billions of parameters. We show that, in the recently discovered Maximal Update Parametrization (muP), many optimal HPs remain stable even as model size changes. This leads to a new HP tuning paradigm we call muTransfer: parametrize the target model in muP, tune the HP indirectly on a smaller model, and zero-shot transfer them to the full-sized model, i.e., without directly tuning the latter at all. We verify muTransfer on Transformer and ResNet. For example, 1) by transferring pretraining HPs from a model of 13M parameters, we outperform published numbers of BERT-large (350M parameters), with a total tuning cost equivalent to pretraining BERT-large once; 2) by transferring from 40M parameters, we outperform published numbers of the 6.7B GPT-3 model, with tuning cost only 7% of total pretraining cost. A Pytorch implementation of our technique can be found at github.com/microsoft/mup and installable via `pip install mup`."
                },
                {
                    "title": "Locating and Editing Factual Associations in GPT",
                    "abstract": "We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's factual predictions. This reveals a distinct set of steps in middle-layer feed-forward modules that mediate factual predictions while processing subject tokens. To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update specific factual associations using Rank-One Model Editing (ROME). We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another. Our results confirm an important role for mid-layer feed-forward modules in storing factual associations and suggest that direct manipulation of computational mechanisms may be a feasible approach for model editing. The code, dataset, visualizations, and an interactive demo notebook are available at https://rome.baulab.info/"
                },
                {
                    "title": "Linear Transformers Are Secretly Fast Weight Programmers",
                    "abstract": "We show the formal equivalence of linearised self-attention mechanisms and fast weight controllers from the early '90s, where a ``slow\"neural net learns by gradient descent to program the ``fast weights\"of another net through sequences of elementary programming instructions which are additive outer products of self-invented activation patterns (today called keys and values). Such Fast Weight Programmers (FWPs) learn to manipulate the contents of a finite memory and dynamically interact with it. We infer a memory capacity limitation of recent linearised softmax attention variants, and replace the purely additive outer products by a delta rule-like programming instruction, such that the FWP can more easily learn to correct the current mapping from keys to values. The FWP also learns to compute dynamically changing learning rates. We also propose a new kernel function to linearise attention which balances simplicity and effectiveness. We conduct experiments on synthetic retrieval problems as well as standard machine translation and language modelling tasks which demonstrate the benefits of our methods."
                },
                {
                    "title": "Explaining neural scaling laws",
                    "abstract": "Significance The population loss of trained deep neural networks has been empirically observed to improve as a power law in a variety of large models and datasets. We investigate the origins behind such \u201cscaling laws\u201d and provide a taxonomy for different scaling regimes. Our findings are based on derivations in linear random feature models\u2014which, in addition to being a simple fruitful model, also describe the wide network limit of deep neural networks. We further formulate and verify aspects of scaling based on smoothness in interpolating a data manifold. We support our theory with empirical results in realistic settings. Our work provides insights into scaling laws and bridges the large gap between theory and experiment in modern deep learning."
                },
                {
                    "title": "Learning Curve Theory",
                    "abstract": "Recently a number of empirical\"universal\"scaling law papers have been published, most notably by OpenAI. `Scaling laws' refers to power-law decreases of training or test error w.r.t. more data, larger neural networks, and/or more compute. In this work we focus on scaling w.r.t. data size $n$. Theoretical understanding of this phenomenon is largely lacking, except in finite-dimensional models for which error typically decreases with $n^{-1/2}$ or $n^{-1}$, where $n$ is the sample size. We develop and theoretically analyse the simplest possible (toy) model that can exhibit $n^{-\\beta}$ learning curves for arbitrary power $\\beta>0$, and determine whether power laws are universal or depend on the data distribution."
                },
                {
                    "title": "Transformer Feed-Forward Layers Are Key-Value Memories",
                    "abstract": "Feed-forward layers constitute two-thirds of a transformer model\u2019s parameters, yet their role in the network remains under-explored. We show that feed-forward layers in transformer-based language models operate as key-value memories, where each key correlates with textual patterns in the training examples, and each value induces a distribution over the output vocabulary. Our experiments show that the learned patterns are human-interpretable, and that lower layers tend to capture shallow patterns, while upper layers learn more semantic ones. The values complement the keys\u2019 input patterns by inducing output distributions that concentrate probability mass on tokens likely to appear immediately after each pattern, particularly in the upper layers. Finally, we demonstrate that the output of a feed-forward layer is a composition of its memories, which is subsequently refined throughout the model\u2019s layers via residual connections to produce the final output distribution."
                },
                {
                    "title": "Why are Adaptive Methods Good for Attention Models?",
                    "abstract": "While stochastic gradient descent (SGD) is still the \\emph{de facto} algorithm in deep learning, adaptive methods like Clipped SGD/Adam have been observed to outperform SGD across important tasks, such as attention models. The settings under which SGD performs poorly in comparison to adaptive methods are not well understood yet. In this paper, we provide empirical and theoretical evidence that a heavy-tailed distribution of the noise in stochastic gradients is one cause of SGD's poor performance. We provide the first tight upper and lower convergence bounds for adaptive gradient methods under heavy-tailed noise. Further, we demonstrate how gradient clipping plays a key role in addressing heavy-tailed gradient noise. Subsequently, we show how clipping can be applied in practice by developing an \\emph{adaptive} coordinate-wise clipping algorithm (ACClip) and demonstrate its superior performance on BERT pretraining and finetuning tasks."
                },
                {
                    "title": "What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation",
                    "abstract": "Deep learning algorithms are well-known to have a propensity for fitting the training data very well and often fit even outliers and mislabeled data points. Such fitting requires memorization of training data labels, a phenomenon that has attracted significant research interest but has not been given a compelling explanation so far. A recent work of Feldman (2019) proposes a theoretical explanation for this phenomenon based on a combination of two insights. First, natural image and data distributions are (informally) known to be long-tailed, that is have a significant fraction of rare and atypical examples. Second, in a simple theoretical model such memorization is necessary for achieving close-to-optimal generalization error when the data distribution is long-tailed. However, no direct empirical evidence for this explanation or even an approach for obtaining such evidence were given. \nIn this work we design experiments to test the key ideas in this theory. The experiments require estimation of the influence of each training example on the accuracy at each test example as well as memorization values of training examples. Estimating these quantities directly is computationally prohibitive but we show that closely-related subsampled influence and memorization values can be estimated much more efficiently. Our experiments demonstrate the significant benefits of memorization for generalization on several standard benchmarks. They also provide quantitative and visually compelling evidence for the theory put forth in (Feldman, 2019)."
                },
                {
                    "title": "Hopfield Networks is All You Need",
                    "abstract": "We show that the transformer attention mechanism is the update rule of a modern Hopfield network with continuous states. This new Hopfield network can store exponentially (with the dimension) many patterns, converges with one update, and has exponentially small retrieval errors. The number of stored patterns is traded off against convergence speed and retrieval error. The new Hopfield network has three types of energy minima (fixed points of the update): (1) global fixed point averaging over all patterns, (2) metastable states averaging over a subset of patterns, and (3) fixed points which store a single pattern. Transformer and BERT models operate in their first layers preferably in the global averaging regime, while they operate in higher layers in metastable states. The gradient in transformers is maximal for metastable states, is uniformly distributed for global averaging, and vanishes for a fixed point near a stored pattern. Using the Hopfield network interpretation, we analyzed learning of transformer and BERT models. Learning starts with attention heads that average and then most of them switch to metastable states. However, the majority of heads in the first layers still averages and can be replaced by averaging, e.g. our proposed Gaussian weighting. In contrast, heads in the last layers steadily learn and seem to use metastable states to collect information created in lower layers. These heads seem to be a promising target for improving transformers. Neural networks with Hopfield networks outperform other methods on immune repertoire classification, where the Hopfield net stores several hundreds of thousands of patterns. We provide a new PyTorch layer called \"Hopfield\", which allows to equip deep learning architectures with modern Hopfield networks as a new powerful concept comprising pooling, memory, and attention. GitHub: this https URL"
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Augmenting Self-attention with Persistent Memory",
                    "abstract": "Transformer networks have lead to important progress in language modeling and machine translation. These models include two consecutive modules, a feed-forward layer and a self-attention layer. The latter allows the network to capture long term dependencies and are often regarded as the key ingredient in the success of Transformers. Building upon this intuition, we propose a new model that solely consists of attention layers. More precisely, we augment the self-attention layers with persistent memory vectors that play a similar role as the feed-forward layer. Thanks to these vectors, we can remove the feed-forward layer without degrading the performance of a transformer. Our evaluation shows the benefits brought by our model on standard character and word level language modeling benchmarks."
                },
                {
                    "title": "Does learning require memorization? a short tale about a long tail",
                    "abstract": "State-of-the-art results on image recognition tasks are achieved using over-parameterized learning algorithms that (nearly) perfectly fit the training set and are known to fit well even random labels. This tendency to memorize seemingly useless training data labels is not explained by existing theoretical analyses. Memorization of the training data also presents significant privacy risks when the training data contains sensitive personal information and thus it is important to understand whether such memorization is necessary for accurate learning. We provide a simple conceptual explanation and a theoretical model demonstrating that for natural data distributions memorization of labels is necessary for achieving close-to-optimal generalization error. The model is motivated and supported by the results of several recent empirical works. In our model, data is sampled from a mixture of subpopulations and the frequencies of these subpopulations are chosen from some prior. The model allows to quantify the effect of not fitting the training data on the generalization performance of the learned classifier and demonstrates that memorization is necessary whenever frequencies are long-tailed. Image and text data are known to follow such distributions and therefore our results establish a formal link between these empirical phenomena. Our results also have concrete implications for the cost of ensuring differential privacy in learning."
                },
                {
                    "title": "Don't Decay the Learning Rate, Increase the Batch Size",
                    "abstract": "It is common practice to decay the learning rate. Here we show one can usually obtain the same learning curve on both training and test sets by instead increasing the batch size during training. This procedure is successful for stochastic gradient descent (SGD), SGD with momentum, Nesterov momentum, and Adam. It reaches equivalent test accuracies after the same number of training epochs, but with fewer parameter updates, leading to greater parallelism and shorter training times. We can further reduce the number of parameter updates by increasing the learning rate $\\epsilon$ and scaling the batch size $B \\propto \\epsilon$. Finally, one can increase the momentum coefficient $m$ and scale $B \\propto 1/(1-m)$, although this tends to slightly reduce the test accuracy. Crucially, our techniques allow us to repurpose existing training schedules for large batch training with no hyper-parameter tuning. We train ResNet-50 on ImageNet to $76.1\\%$ validation accuracy in under 30 minutes."
                },
                {
                    "title": "Dissecting Adam: The Sign, Magnitude and Variance of Stochastic Gradients",
                    "abstract": "The ADAM optimizer is exceedingly popular in the deep learning community. Often it works very well, sometimes it doesn't. Why? We interpret ADAM as a combination of two aspects: for each weight, the update direction is determined by the sign of stochastic gradients, whereas the update magnitude is determined by an estimate of their relative variance. We disentangle these two aspects and analyze them in isolation, gaining insight into the mechanisms underlying ADAM. This analysis also extends recent results on adverse effects of ADAM on generalization, isolating the sign aspect as the problematic one. Transferring the variance adaptation to SGD gives rise to a novel method, completing the practitioner's toolbox for problems where ADAM fails."
                },
                {
                    "title": "Dense Associative Memory for Pattern Recognition",
                    "abstract": "A model of associative memory is studied, which stores and reliably retrieves many more patterns than the number of neurons in the network. We propose a simple duality between this dense associative memory and neural networks commonly used in deep learning. On the associative memory side of this duality, a family of models that smoothly interpolates between two limiting cases can be constructed. One limit is referred to as the feature-matching mode of pattern recognition, and the other one as the prototype regime. On the deep learning side of the duality, this family corresponds to feedforward neural networks with one hidden layer and various activation functions, which transmit the activities of the visible neurons to the hidden layer. This family of activation functions includes logistics, rectified linear units, and rectified polynomials of higher degrees. The proposed duality makes it possible to apply energy-based intuition from associative memory to analyze computational properties of neural networks with unusual activation functions - the higher rectified polynomials which until now have not been used in deep learning. The utility of the dense memories is illustrated for two test cases: the logical gate XOR and the recognition of handwritten digits from the MNIST data set."
                },
                {
                    "title": "Elements of Information Theory",
                    "abstract": "Preface to the Second Edition. Preface to the First Edition. Acknowledgments for the Second Edition. Acknowledgments for the First Edition. 1. Introduction and Preview. 1.1 Preview of the Book. 2. Entropy, Relative Entropy, and Mutual Information. 2.1 Entropy. 2.2 Joint Entropy and Conditional Entropy. 2.3 Relative Entropy and Mutual Information. 2.4 Relationship Between Entropy and Mutual Information. 2.5 Chain Rules for Entropy, Relative Entropy, and Mutual Information. 2.6 Jensen's Inequality and Its Consequences. 2.7 Log Sum Inequality and Its Applications. 2.8 Data-Processing Inequality. 2.9 Sufficient Statistics. 2.10 Fano's Inequality. Summary. Problems. Historical Notes. 3. Asymptotic Equipartition Property. 3.1 Asymptotic Equipartition Property Theorem. 3.2 Consequences of the AEP: Data Compression. 3.3 High-Probability Sets and the Typical Set. Summary. Problems. Historical Notes. 4. Entropy Rates of a Stochastic Process. 4.1 Markov Chains. 4.2 Entropy Rate. 4.3 Example: Entropy Rate of a Random Walk on a Weighted Graph. 4.4 Second Law of Thermodynamics. 4.5 Functions of Markov Chains. Summary. Problems. Historical Notes. 5. Data Compression. 5.1 Examples of Codes. 5.2 Kraft Inequality. 5.3 Optimal Codes. 5.4 Bounds on the Optimal Code Length. 5.5 Kraft Inequality for Uniquely Decodable Codes. 5.6 Huffman Codes. 5.7 Some Comments on Huffman Codes. 5.8 Optimality of Huffman Codes. 5.9 Shannon-Fano-Elias Coding. 5.10 Competitive Optimality of the Shannon Code. 5.11 Generation of Discrete Distributions from Fair Coins. Summary. Problems. Historical Notes. 6. Gambling and Data Compression. 6.1 The Horse Race. 6.2 Gambling and Side Information. 6.3 Dependent Horse Races and Entropy Rate. 6.4 The Entropy of English. 6.5 Data Compression and Gambling. 6.6 Gambling Estimate of the Entropy of English. Summary. Problems. Historical Notes. 7. Channel Capacity. 7.1 Examples of Channel Capacity. 7.2 Symmetric Channels. 7.3 Properties of Channel Capacity. 7.4 Preview of the Channel Coding Theorem. 7.5 Definitions. 7.6 Jointly Typical Sequences. 7.7 Channel Coding Theorem. 7.8 Zero-Error Codes. 7.9 Fano's Inequality and the Converse to the Coding Theorem. 7.10 Equality in the Converse to the Channel Coding Theorem. 7.11 Hamming Codes. 7.12 Feedback Capacity. 7.13 Source-Channel Separation Theorem. Summary. Problems. Historical Notes. 8. Differential Entropy. 8.1 Definitions. 8.2 AEP for Continuous Random Variables. 8.3 Relation of Differential Entropy to Discrete Entropy. 8.4 Joint and Conditional Differential Entropy. 8.5 Relative Entropy and Mutual Information. 8.6 Properties of Differential Entropy, Relative Entropy, and Mutual Information. Summary. Problems. Historical Notes. 9. Gaussian Channel. 9.1 Gaussian Channel: Definitions. 9.2 Converse to the Coding Theorem for Gaussian Channels. 9.3 Bandlimited Channels. 9.4 Parallel Gaussian Channels. 9.5 Channels with Colored Gaussian Noise. 9.6 Gaussian Channels with Feedback. Summary. Problems. Historical Notes. 10. Rate Distortion Theory. 10.1 Quantization. 10.2 Definitions. 10.3 Calculation of the Rate Distortion Function. 10.4 Converse to the Rate Distortion Theorem. 10.5 Achievability of the Rate Distortion Function. 10.6 Strongly Typical Sequences and Rate Distortion. 10.7 Characterization of the Rate Distortion Function. 10.8 Computation of Channel Capacity and the Rate Distortion Function. Summary. Problems. Historical Notes. 11. Information Theory and Statistics. 11.1 Method of Types. 11.2 Law of Large Numbers. 11.3 Universal Source Coding. 11.4 Large Deviation Theory. 11.5 Examples of Sanov's Theorem. 11.6 Conditional Limit Theorem. 11.7 Hypothesis Testing. 11.8 Chernoff-Stein Lemma. 11.9 Chernoff Information. 11.10 Fisher Information and the Cram-er-Rao Inequality. Summary. Problems. Historical Notes. 12. Maximum Entropy. 12.1 Maximum Entropy Distributions. 12.2 Examples. 12.3 Anomalous Maximum Entropy Problem. 12.4 Spectrum Estimation. 12.5 Entropy Rates of a Gaussian Process. 12.6 Burg's Maximum Entropy Theorem. Summary. Problems. Historical Notes. 13. Universal Source Coding. 13.1 Universal Codes and Channel Capacity. 13.2 Universal Coding for Binary Sequences. 13.3 Arithmetic Coding. 13.4 Lempel-Ziv Coding. 13.5 Optimality of Lempel-Ziv Algorithms. Compression. Summary. Problems. Historical Notes. 14. Kolmogorov Complexity. 14.1 Models of Computation. 14.2 Kolmogorov Complexity: Definitions and Examples. 14.3 Kolmogorov Complexity and Entropy. 14.4 Kolmogorov Complexity of Integers. 14.5 Algorithmically Random and Incompressible Sequences. 14.6 Universal Probability. 14.7 Kolmogorov complexity. 14.9 Universal Gambling. 14.10 Occam's Razor. 14.11 Kolmogorov Complexity and Universal Probability. 14.12 Kolmogorov Sufficient Statistic. 14.13 Minimum Description Length Principle. Summary. Problems. Historical Notes. 15. Network Information Theory. 15.1 Gaussian Multiple-User Channels. 15.2 Jointly Typical Sequences. 15.3 Multiple-Access Channel. 15.4 Encoding of Correlated Sources. 15.5 Duality Between Slepian-Wolf Encoding and Multiple-Access Channels. 15.6 Broadcast Channel. 15.7 Relay Channel. 15.8 Source Coding with Side Information. 15.9 Rate Distortion with Side Information. 15.10 General Multiterminal Networks. Summary. Problems. Historical Notes. 16. Information Theory and Portfolio Theory. 16.1 The Stock Market: Some Definitions. 16.2 Kuhn-Tucker Characterization of the Log-Optimal Portfolio. 16.3 Asymptotic Optimality of the Log-Optimal Portfolio. 16.4 Side Information and the Growth Rate. 16.5 Investment in Stationary Markets. 16.6 Competitive Optimality of the Log-Optimal Portfolio. 16.7 Universal Portfolios. 16.8 Shannon-McMillan-Breiman Theorem (General AEP). Summary. Problems. Historical Notes. 17. Inequalities in Information Theory. 17.1 Basic Inequalities of Information Theory. 17.2 Differential Entropy. 17.3 Bounds on Entropy and Relative Entropy. 17.4 Inequalities for Types. 17.5 Combinatorial Bounds on Entropy. 17.6 Entropy Rates of Subsets. 17.7 Entropy and Fisher Information. 17.8 Entropy Power Inequality and Brunn-Minkowski Inequality. 17.9 Inequalities for Determinants. 17.10 Inequalities for Ratios of Determinants. Summary. Problems. Historical Notes. Bibliography. List of Symbols. Index."
                },
                {
                    "title": "Neural networks and physical systems with emergent collective computational abilities.",
                    "abstract": "Computational properties of use of biological organisms or to the construction of computers can emerge as collective properties of systems having a large number of simple equivalent components (or neurons). The physical meaning of content-addressable memory is described by an appropriate phase space flow of the state of a system. A model of such a system is given, based on aspects of neurobiology but readily adapted to integrated circuits. The collective properties of this model produce a content-addressable memory which correctly yields an entire memory from any subpart of sufficient size. The algorithm for the time evolution of the state of the system is based on asynchronous parallel processing. Additional emergent collective properties include some capacity for generalization, familiarity recognition, categorization, error correction, and time sequence retention. The collective properties are only weakly sensitive to details of the modeling or the failure of individual devices."
                },
                {
                    "title": "Learning Patterns and Pattern Sequences by Self-Organizing Nets of Threshold Elements",
                    "abstract": "Various information-processing capabilities of self-organizing nets of threshold elements are studied. A self-organizing net, learning from patterns or pattern sequences given from outside as stimuli, \"remembers\" some of them as stable equilibrium states or state-transition sequences of the net. A condition where many patterns and pattern sequences are remembered in a net at the same time is shown. The stability degree of their remembrance and recalling under noise disturbances is investigated theoretically. For this purpose, the stability of state transition in an autonomous logical net of threshold elements is studied by the use of characteristics of threshold elements."
                },
                {
                    "title": "Correlation Matrix Memories",
                    "abstract": "A new model for associative memory, based on a correlation matrix, is suggested. In this model information is accumulated on memory elements as products of component data. Denoting a key vector by q(p), and the data associated with it by another vector x(p), the pairs (q(p), x(p)) are memorized in the form of a matrix {see the Equation in PDF File} where c is a constant. A randomly selected subset of the elements of Mxq can also be used for memorizing. The recalling of a particular datum x(r) is made by a transformation x(r)=Mxqq(r). This model is failure tolerant and facilitates associative search of information; these are properties that are usually assigned to holographic memories. Two classes of memories are discussed: a complete correlation matrix memory (CCMM), and randomly organized incomplete correlation matrix memories (ICMM). The data recalled from the latter are stochastic variables but the fidelity of recall is shown to have a deterministic limit if the number of memory elements grows without limits. A special case of correlation matrix memories is the auto-associative memory in which any part of the memorized information can be used as a key. The memories are selective with respect to accumulated data. The ICMM exhibits adaptive improvement under certain circumstances. It is also suggested that correlation matrix memories could be applied for the classification of data."
                },
                {
                    "title": "Theories of associative recall",
                    "abstract": "The problem of how the brain stores and retrieves information is ultimately an experimental one, and its solution will doubtless call for the combined resources of psychology, physiology and molecular biology. But it is also a problem of great theoretical sophistication; and one of the major tasks confronting the brain scientist is the construction of theoretical models which are worthy of, and open to, experimental test. In this review we shall be concerned with the latter aspect of the problem of memory, which has attracted quite a lot of attention in the last few years. It is early yet to judge the relative merits of the various models in any detail; but as we shall see, most of those which have been developed beyond their initial hypotheses have a certain family resemblance, and it seems as if we may now be in possession of the basic ideas which will be needed for the understanding of one of the central problems of memory, namely the mechanism of associative recall."
                },
                {
                    "title": "Tensor Product Variable Binding and the Representation of Symbolic Structures in Connectionist Systems",
                    "abstract": "This chapter contains sections titled: 1. Introduction, 2. Connectionist Representation and Tensor Product Binding: Definition and Examples, 3. Tensor Product Representation: Properties, 4. Conclusion"
                }
            ],
            "categories": [
                "stat.ML",
                "cs.AI",
                "cs.CL",
                "cs.LG",
                "cs.NE",
                "I.2.6; G.1.6"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we better understand the scaling laws of large language models (LLMs) through the lens of associative memory models and heavy-tailed data distributions?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it could lead to a deeper understanding of the mechanisms behind LLMs, which are increasingly influential in various applications such as natural language processing, machine translation, and conversational agents. By elucidating the relationship between model architecture, data distribution, and performance, this research could inform the design of more efficient models and algorithms, ultimately advancing the field of machine learning. Improved scaling laws could also facilitate the development of practical applications that leverage LLMs more effectively, enhancing their utility in real-world scenarios.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem stem from the complexity of modeling the interactions between heavy-tailed data distributions and associative memory mechanisms in LLMs. Naive approaches may fail because they do not account for the intricate relationships between input-output pairs and the statistical properties of the data. Additionally, the theoretical underpinnings of scaling laws are not fully understood, making it difficult to derive generalizable insights. Overcoming these obstacles requires sophisticated mathematical modeling and empirical validation, as well as a nuanced understanding of the dynamics of transformer architectures.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either continuous Gaussian inputs or broader scaling laws without delving into the specificities of discrete token distributions and associative memory models. This gap has been compounded by a lack of statistical insights into the behavior of random embeddings in the context of heavy-tailed data. Barriers such as the complexity of the models and the need for extensive empirical validation have hindered progress. Our approach differs by providing precise statistical rates for outer-product memories and comparing them with practical optimization algorithms, thus bridging the gap between theory and practice.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a model for associative memory that utilizes outer-product representations with random embeddings, specifically tailored for heavy-tailed token distributions. We will analyze a joint distribution of inputs and outputs, focusing on discrete values. The dataset will consist of sequences of tokens, and we will evaluate the model's performance using metrics that capture memorization behaviors and scaling properties. Expected outcomes include a clearer understanding of the statistical rates of"
            }
        },
        "author_data": {
            "b2713456-3952-464f-9ca4-4abdda5884ea": {
                "pk": "b2713456-3952-464f-9ca4-4abdda5884ea",
                "name": "Vivien Cabannes",
                "collaborators": [
                    "Charles Arnal",
                    "Wassim Bouaziz",
                    "Vianney Perchet",
                    "Alice Yang",
                    "Francois Charton",
                    "Julia Kempe",
                    "Ambroise Odonnat",
                    "Berfin Simsek",
                    "A. Bietti",
                    "Carles Domingo-Enrich",
                    "Francis R. Bach",
                    "Alessandro Rudi"
                ],
                "domain": [
                    "Weakly Supervised Learning",
                    "Active Learning",
                    "Transformer Models",
                    "Generative Modeling"
                ],
                "publications": [
                    {
                        "title": "Mode Estimation with Partial Feedback",
                        "abstract": "The combination of lightly supervised pre-training and online fine-tuning has played a key role in recent AI developments. These new learning pipelines call for new theoretical frameworks. In this paper, we formalize core aspects of weakly supervised and active learning with a simple problem: the estimation of the mode of a distribution using partial feedback. We show how entropy coding allows for optimal information acquisition from partial feedback, develop coarse sufficient statistics for mode identification, and adapt bandit algorithms to our new setting. Finally, we combine those contributions into a statistically and computationally efficient solution to our problem."
                    },
                    {
                        "title": "Iteration Head: A Mechanistic Study of Chain-of-Thought",
                        "abstract": "Chain-of-Thought (CoT) reasoning is known to improve Large Language Models both empirically and in terms of theoretical approximation power. However, our understanding of the inner workings and conditions of apparition of CoT capabilities remains limited. This paper helps fill this gap by demonstrating how CoT reasoning emerges in transformers in a controlled and interpretable setting. In particular, we observe the appearance of a specialized attention mechanism dedicated to iterative reasoning, which we coined\"iteration heads\". We track both the emergence and the precise working of these iteration heads down to the attention level, and measure the transferability of the CoT skills to which they give rise between tasks."
                    },
                    {
                        "title": "A Visual Case Study of the Training Dynamics in Neural Networks",
                        "abstract": "This paper introduces a visual sandbox designed to explore the training dynamics of a small-scale transformer model, with the embedding dimension constrained to $d=2$. This restriction allows for a comprehensive two-dimensional visualization of each layer's dynamics. Through this approach, we gain insights into training dynamics, circuit transferability, and the causes of loss spikes, including those induced by the high curvature of normalization layers. We propose strategies to mitigate these spikes, demonstrating how good visualization facilitates the design of innovative ideas of practical interest. Additionally, we believe our sandbox could assist theoreticians in assessing essential training dynamics mechanisms and integrating them into future theories. The code is available at https://github.com/facebookresearch/pal."
                    },
                    {
                        "title": "Learning Associative Memories with Gradient Descent",
                        "abstract": "This work focuses on the training dynamics of one associative memory module storing outer products of token embeddings. We reduce this problem to the study of a system of particles, which interact according to properties of the data distribution and correlations between embeddings. Through theory and experiments, we provide several insights. In overparameterized regimes, we obtain logarithmic growth of the ``classification margins.'' Yet, we show that imbalance in token frequencies and memory interferences due to correlated embeddings lead to oscillatory transitory regimes. The oscillations are more pronounced with large step sizes, which can create benign loss spikes, although these learning rates speed up the dynamics and accelerate the asymptotic convergence. In underparameterized regimes, we illustrate how the cross-entropy loss can lead to suboptimal memorization schemes. Finally, we assess the validity of our findings on small Transformer models."
                    },
                    {
                        "title": "Open Problem: Learning with Variational Objectives on Measures",
                        "abstract": "The theory of statistical learning has focused on variational objectives expressed on functions. In this note, we discuss motivations to write similar objectives on measures, in particular to discuss out-of-distribution generalization and weakly-supervised learning. It raises a natural question: can one cast usual statistical learning results to objectives expressed on measures? Does the resulting construction lead to new algorithms of practical interest?"
                    },
                    {
                        "title": "Touring Sampling With Pushforward Maps",
                        "abstract": "The number of sampling methods could be daunting for a practitioner looking to cast powerful machine learning methods to their specific problem. This paper takes a theoretical stance to review and organize many sampling approaches in the \"generative modeling\" setting, where one wants to generate new data that are similar to some training examples. By revealing links between existing methods, it might prove useful to overcome some of the current challenges in sampling with diffusion models, such as long inference time due to diffusion simulation, or the lack of diversity in generated samples."
                    },
                    {
                        "title": "Disambiguation of weak supervision with exponential convergence rates",
                        "abstract": "Machine learning approached through supervised learning requires expensive annotation of data. This motivates weakly supervised learning, where data are annotated with incomplete yet discriminative information. In this paper, we focus on partial labelling, an instance of weak supervision where, from a given input, we are given a set of potential targets. We review a disambiguation principle to recover full supervision from weak supervision, and propose an empirical disambiguation algorithm. We prove exponential convergence rates of our algorithm under classical learnability assumptions, and we illustrate the usefulness of our method on practical examples."
                    }
                ]
            },
            "416b8e27-edd2-4d06-bb5f-2e5093f25069": {
                "pk": "416b8e27-edd2-4d06-bb5f-2e5093f25069",
                "name": "Elvis Dohmatob",
                "collaborators": [
                    "Yunzhen Feng",
                    "Julia Kempe",
                    "Pu Yang",
                    "Francois Charton",
                    "Arjun Subramonian",
                    "M. Scetbon",
                    "Morgane Goibert",
                    "Insu Han",
                    "Mike Gartrell",
                    "Amin Karbasi",
                    "Nicolas Usunier",
                    "Virginie Do",
                    "Sam Bell",
                    "Levent Sagun",
                    "Chuan Guo",
                    "Jennifer Gillenwater",
                    "Thomas Ricatte",
                    "A. Bietti",
                    "Matteo Pirotta",
                    "A. Lazaric"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Large Language Models",
                    "Model Robustness",
                    "Fairness in AI"
                ],
                "publications": [
                    {
                        "title": "Strong Model Collapse",
                        "abstract": "Within the scaling laws paradigm, which underpins the training of large neural networks like ChatGPT and Llama, we consider a supervised regression setting and establish the existance of a strong form of the model collapse phenomenon, a critical performance degradation due to synthetic data in the training corpus. Our results show that even the smallest fraction of synthetic data (e.g., as little as 1\\% of the total training dataset) can still lead to model collapse: larger and larger training sets do not enhance performance. We further investigate whether increasing model size, an approach aligned with current trends in training large language models, exacerbates or mitigates model collapse. In a simplified regime where neural networks are approximated via random projections of tunable size, we both theoretically and empirically show that larger models can amplify model collapse. Interestingly, our theory also indicates that, beyond the interpolation threshold (which can be extremely high for very large datasets), larger models may mitigate the collapse, although they do not entirely prevent it. Our theoretical findings are empirically verified through experiments on language models and feed-forward neural networks for images."
                    },
                    {
                        "title": "Beyond Model Collapse: Scaling Up with Synthesized Data Requires Reinforcement",
                        "abstract": "Synthesized data from generative models is increasingly considered as an alternative to human-annotated data for fine-tuning Large Language Models. This raises concerns about model collapse: a drop in performance of models fine-tuned on generated data. Considering that it is easier for both humans and machines to tell between good and bad examples than to generate high-quality samples, we investigate the use of feedback on synthesized data to prevent model collapse. We derive theoretical conditions under which a Gaussian mixture classification model can achieve asymptotically optimal performance when trained on feedback-augmented synthesized data, and provide supporting simulations for finite regimes. We illustrate our theoretical predictions on two practical problems: computing matrix eigenvalues with transformers and news summarization with large language models, which both undergo model collapse when trained on model-generated data. We show that training from feedback-augmented synthesized data, either by pruning incorrect predictions or by selecting the best of several guesses, can prevent model collapse, validating popular approaches like RLHF."
                    },
                    {
                        "title": "Beyond Model Collapse: Scaling Up with Synthesized Data Requires Verification",
                        "abstract": "Large Language Models (LLM) are increasingly trained on data generated by other LLM, either because generated text and images become part of the pre-training corpus, or because synthetized data is used as a replacement for expensive human-annotation. This raises concerns about \\emph{model collapse}, a drop in model performance when their training sets include generated data. Considering that it is easier for both humans and machines to tell between good and bad examples than to generate high-quality samples, we investigate the use of verification on synthesized data to prevent model collapse. We provide a theoretical characterization using Gaussian mixtures, linear classifiers, and linear verifiers to derive conditions with measurable proxies to assess whether the verifier can effectively select synthesized data that leads to optimal performance. We experiment with two practical tasks -- computing matrix eigenvalues with transformers and news summarization with LLMs -- which both exhibit model collapse when trained on generated data, and show that verifiers, even imperfect ones, can indeed be harnessed to prevent model collapse and that our proposed proxy measure strongly correlates with performance."
                    },
                    {
                        "title": "An Effective Theory of Bias Amplification",
                        "abstract": "Machine learning models may capture and amplify biases present in data, leading to disparate test performance across social groups. To better understand, evaluate, and mitigate these possible biases, a deeper theoretical understanding of how model design choices and data distribution properties could contribute to bias is needed. In this work, we contribute a precise analytical theory in the context of ridge regression, both with and without random projections, where the former models neural networks in a simplified regime. Our theory offers a unified and rigorous explanation of machine learning bias, providing insights into phenomena such as bias amplification and minority-group bias in various feature and parameter regimes. For example, we demonstrate that there may be an optimal regularization penalty or training time to avoid bias amplification, and there can be fundamental differences in test error between groups that do not vanish with increased parameterization. Importantly, our theoretical predictions align with several empirical observations reported in the literature. We extensively empirically validate our theory on diverse synthetic and semi-synthetic datasets."
                    },
                    {
                        "title": "A Tale of Tails: Model Collapse as a Change of Scaling Laws",
                        "abstract": "As AI model size grows, neural scaling laws have become a crucial tool to predict the improvements of large models when increasing capacity and the size of original (human or natural) training data. Yet, the widespread use of popular models means that the ecosystem of online data and text will co-evolve to progressively contain increased amounts of synthesized data. In this paper we ask: How will the scaling laws change in the inevitable regime where synthetic data makes its way into the training corpus? Will future models, still improve, or be doomed to degenerate up to total (model) collapse? We develop a theoretical framework of model collapse through the lens of scaling laws. We discover a wide range of decay phenomena, analyzing loss of scaling, shifted scaling with number of generations, the ''un-learning\"of skills, and grokking when mixing human and synthesized data. Our theory is validated by large-scale experiments with a transformer on an arithmetic task and text generation using the large language model Llama2."
                    },
                    {
                        "title": "Model Collapse Demystified: The Case of Regression",
                        "abstract": "In the era of proliferation of large language and image generation models, the phenomenon of\"model collapse\"refers to the situation whereby as a model is trained recursively on data generated from previous generations of itself over time, its performance degrades until the model eventually becomes completely useless, i.e the model collapses. In this work, we study this phenomenon in the setting of high-dimensional regression and obtain analytic formulae which quantitatively outline this phenomenon in a broad range of regimes. In the special case of polynomial decaying spectral and source conditions, we obtain modified scaling laws which exhibit new crossover phenomena from fast to slow rates. We also propose a simple strategy based on adaptive regularization to mitigate model collapse. Our theoretical results are validated with experiments."
                    },
                    {
                        "title": "Robust Linear Regression: Gradient-descent, Early-stopping, and Beyond",
                        "abstract": "In this work we study the robustness to adversarial attacks, of early-stopping strategies on gradient-descent (GD) methods for linear regression. More precisely, we show that early-stopped GD is optimally robust (up to an absolute constant) against Euclidean-norm adversarial attacks. However, we show that this strategy can be arbitrarily sub-optimal in the case of general Mahalanobis attacks. This observation is compatible with recent findings in the case of classification~\\cite{Vardi2022GradientMP} that show that GD provably converges to non-robust models. To alleviate this issue, we propose to apply instead a GD scheme on a transformation of the data adapted to the attack. This data transformation amounts to apply feature-depending learning rates and we show that this modified GD is able to handle any Mahalanobis attack, as well as more general attacks under some conditions. Unfortunately, choosing such adapted transformations can be hard for general attacks. To the rescue, we design a simple and tractable estimator whose adversarial risk is optimal up to within a multiplicative constant of 1.1124 in the population regime, and works for any norm."
                    },
                    {
                        "title": "Robust Linear Regression: Phase-Transitions and Precise Tradeoffs for General Norms",
                        "abstract": "In this paper, we investigate the impact of test-time adversarial attacks on linear regression models and determine the optimal level of robustness that any model can reach while maintaining a given level of standard predictive performance (accuracy). Through quantitative estimates, we uncover fundamental tradeoffs between adversarial robustness and accuracy in different regimes. We obtain a precise characterization which distinguishes between regimes where robustness is achievable without hurting standard accuracy and regimes where a tradeoff might be unavoidable. Our findings are empirically confirmed with simple experiments that represent a variety of settings. This work applies to feature covariance matrices and attack norms of any nature, and extends beyond previous works in this area."
                    },
                    {
                        "title": "Origins of Low-dimensional Adversarial Perturbations",
                        "abstract": "In this paper, we initiate a rigorous study of the phenomenon of low-dimensional adversarial perturbations (LDAPs) in classification. Unlike the classical setting, these perturbations are limited to a subspace of dimension $k$ which is much smaller than the dimension $d$ of the feature space. The case $k=1$ corresponds to so-called universal adversarial perturbations (UAPs; Moosavi-Dezfooli et al., 2017). First, we consider binary classifiers under generic regularity conditions (including ReLU networks) and compute analytical lower-bounds for the fooling rate of any subspace. These bounds explicitly highlight the dependence of the fooling rate on the pointwise margin of the model (i.e., the ratio of the output to its $L_2$ norm of its gradient at a test point), and on the alignment of the given subspace with the gradients of the model w.r.t. inputs. Our results provide a rigorous explanation for the recent success of heuristic methods for efficiently generating low-dimensional adversarial perturbations. Finally, we show that if a decision-region is compact, then it admits a universal adversarial perturbation with $L_2$ norm which is $\\sqrt{d}$ times smaller than the typical $L_2$ norm of a data point. Our theoretical results are confirmed by experiments on both synthetic and real data."
                    },
                    {
                        "title": "Scalable Sampling for Nonsymmetric Determinantal Point Processes",
                        "abstract": "A determinantal point process (DPP) on a collection of $M$ items is a model, parameterized by a symmetric kernel matrix, that assigns a probability to every subset of those items. Recent work shows that removing the kernel symmetry constraint, yielding nonsymmetric DPPs (NDPPs), can lead to significant predictive performance gains for machine learning applications. However, existing work leaves open the question of scalable NDPP sampling. There is only one known DPP sampling algorithm, based on Cholesky decomposition, that can directly apply to NDPPs as well. Unfortunately, its runtime is cubic in $M$, and thus does not scale to large item collections. In this work, we first note that this algorithm can be transformed into a linear-time one for kernels with low-rank structure. Furthermore, we develop a scalable sublinear-time rejection sampling algorithm by constructing a novel proposal distribution. Additionally, we show that imposing certain structural constraints on the NDPP kernel enables us to bound the rejection rate in a way that depends only on the kernel rank. In our experiments we compare the speed of all of these samplers for a variety of real-world tasks."
                    },
                    {
                        "title": "An Adversarial Robustness Perspective on the Topology of Neural Networks",
                        "abstract": "In this paper, we investigate the impact of neural networks (NNs) topology on adversarial robustness. Specifically, we study the graph produced when an input traverses all the layers of a NN, and show that such graphs are different for clean and adversarial inputs. We find that graphs from clean inputs are more centralized around highway edges, whereas those from adversaries are more diffuse, leveraging under-optimized edges. Through experiments on a variety of datasets and architectures, we show that these under-optimized edges are a source of adversarial vulnerability and that they can be used to detect adversarial inputs."
                    },
                    {
                        "title": "Fast online ranking with fairness of exposure",
                        "abstract": "As recommender systems become increasingly central for sorting and prioritizing the content available online, they have a growing impact on the opportunities or revenue of their items producers. For instance, they influence which recruiter a resume is recommended to, or to whom and how much a music track, video or news article is being exposed. This calls for recommendation approaches that not only maximize (a proxy of) user satisfaction, but also consider some notion of fairness in the exposure of items or groups of items. Formally, such recommendations are usually obtained by maximizing a concave objective function in the space of randomized rankings. When the total exposure of an item is defined as the sum of its exposure over users, the optimal rankings of every users become coupled, which makes the optimization process challenging. Existing approaches to find these rankings either solve the global optimization problem in a batch setting, i.e., for all users at once, which makes them inapplicable at scale, or are based on heuristics that have weak theoretical guarantees. In this paper, we propose the first efficient online algorithm to optimize concave objective functions in the space of rankings which applies to every concave and smooth objective function, such as the ones found for fairness of exposure. Based on online variants of the Frank-Wolfe algorithm, we show that our algorithm is computationally fast, generating rankings on-the-fly with computation cost dominated by the sort operation, memory efficient, and has strong theoretical guarantees. Compared to baseline policies that only maximize user-side performance, our algorithm allows to incorporate complex fairness of exposure criteria in the recommendations with negligible computational overhead. We present experiments on artificial music and movie recommendation tasks using Last.fm and MovieLens datasets which suggest that in practice, the algorithm rapidly reaches good performances on three different objectives representing different fairness of exposure criteria."
                    },
                    {
                        "title": "On the (Non-)Robustness of Two-Layer Neural Networks in Different Learning Regimes",
                        "abstract": "Neural networks are known to be highly sensitive to adversarial examples. These may arise due to different factors, such as random initialization, or spurious correlations in the learning problem. To better understand these factors, we provide a precise study of the adversarial robustness in different scenarios, from initialization to the end of training in different regimes, as well as intermediate scenarios, where initialization still plays a role due to\"lazy\"training. We consider over-parameterized networks in high dimensions with quadratic targets and infinite samples. Our analysis allows us to identify new tradeoffs between approximation (as measured via test error) and robustness, whereby robustness can only get worse when test error improves, and vice versa. We also show how linearized lazy training regimes can worsen robustness, due to improperly scaled random initialization. Our theoretical results are illustrated with numerical experiments."
                    },
                    {
                        "title": "Contextual bandits with concave rewards, and an application to fair ranking",
                        "abstract": "We consider Contextual Bandits with Concave Rewards (CBCR), a multi-objective bandit problem where the desired trade-off between the rewards is defined by a known concave objective function, and the reward vector depends on an observed stochastic context. We present the first algorithm with provably vanishing regret for CBCR without restrictions on the policy space, whereas prior works were restricted to finite policy spaces or tabular representations. Our solution is based on a geometric interpretation of CBCR algorithms as optimization algorithms over the convex set of expected rewards spanned by all stochastic policies. Building on Frank-Wolfe analyses in constrained convex optimization, we derive a novel reduction from the CBCR regret to the regret of a scalar-reward bandit problem. We illustrate how to apply the reduction off-the-shelf to obtain algorithms for CBCR with both linear and general reward functions, in the case of non-combinatorial actions. Motivated by fairness in recommendation, we describe a special case of CBCR with rankings and fairness-aware objectives, leading to the first algorithm with regret guarantees for contextual combinatorial bandits with fairness of exposure."
                    },
                    {
                        "title": "Scalable MCMC Sampling for Nonsymmetric Determinantal Point Processes",
                        "abstract": "A determinantal point process (DPP) is an elegant model that assigns a probability to every subset of a collection of n items. While conventionally a DPP is parameterized by a symmetric kernel matrix, removing this symmetry constraint, resulting in nonsymmetric DPPs (NDPPs), leads to significant improvements in modeling power and predictive performance. Recent work has studied an approximate Markov chain Monte Carlo (MCMC) sampling algorithm for NDPPs restricted to size- k subsets (called k -NDPPs). However, the runtime of this approach is quadratic in n , making it infeasible for large-scale settings. In this work, we develop a scalable MCMC sampling algorithm for k -NDPPs with low-rank kernels, thus enabling runtime that is sublinear in n . Our method is based on a state-of-the-art NDPP rejection sampling algorithm, which we enhance with a novel approach for efficiently constructing the proposal distribution. Furthermore, we extend our scalable k -NDPP sampling algorithm to NDPPs without size constraints. Our resulting sampling method has polynomial time complexity in the rank of the kernel, while the existing approach has runtime that is exponential in the rank. With both a theoretical analysis and experiments on real-world datasets, we verify that our scalable approximate sampling algorithms are orders of magnitude faster than existing sampling approaches for k -NDPPs and NDPPs."
                    }
                ]
            },
            "d3992110-0712-4e8f-a6c7-43810284a43c": {
                "pk": "d3992110-0712-4e8f-a6c7-43810284a43c",
                "name": "Alberto Bietti",
                "collaborators": [
                    "Joan Bruna",
                    "Vivien A. Cabannes",
                    "Randall Balestriero",
                    "Miroslav Dud\u00edk",
                    "J. Langford",
                    "Zhiwei Steven Wu",
                    "Lei Chen",
                    "Vivien Cabannes",
                    "Berfin Simsek",
                    "Aaron Mishkin",
                    "Robert M. Gower",
                    "B. Kiani",
                    "Yann LeCun",
                    "Diane Bouchacourt",
                    "H. J\u00e9gou",
                    "L. Bottou",
                    "Loucas Pillaud-Vivien",
                    "David Brandfonbrener",
                    "Jacob Buckman",
                    "R. Laroche",
                    "Houssam Zenati",
                    "Eustache Diemert",
                    "J. Mairal",
                    "Matthieu Martin",
                    "P. Gaillard",
                    "Elvis Dohmatob",
                    "Chengkun Wei",
                    "Chen-Yu Wei",
                    "Clayton Sanford",
                    "M. Song",
                    "N. Keriven",
                    "Samuel Vaiter",
                    "Carles Domingo-Enrich",
                    "E. Vanden-Eijnden",
                    "Luca Venturi"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Self-Supervised Learning",
                    "Federated Learning"
                ],
                "publications": [
                    {
                        "title": "How Truncating Weights Improves Reasoning in Language Models",
                        "abstract": "In addition to the ability to generate fluent text in various languages, large language models have been successful at tasks that involve basic forms of logical\"reasoning\"over their context. Recent work found that selectively removing certain components from weight matrices in pre-trained models can improve such reasoning capabilities. We investigate this phenomenon further by carefully studying how certain global associations tend to be stored in specific weight components or Transformer blocks, in particular feed-forward layers. Such associations may hurt predictions in reasoning tasks, and removing the corresponding components may then improve performance. We analyze how this arises during training, both empirically and theoretically, on a two-layer Transformer trained on a basic reasoning task with noise, a toy associative memory model, and on the Pythia family of pre-trained models tested on simple reasoning tasks."
                    },
                    {
                        "title": "Learning Associative Memories with Gradient Descent",
                        "abstract": "This work focuses on the training dynamics of one associative memory module storing outer products of token embeddings. We reduce this problem to the study of a system of particles, which interact according to properties of the data distribution and correlations between embeddings. Through theory and experiments, we provide several insights. In overparameterized regimes, we obtain logarithmic growth of the ``classification margins.'' Yet, we show that imbalance in token frequencies and memory interferences due to correlated embeddings lead to oscillatory transitory regimes. The oscillations are more pronounced with large step sizes, which can create benign loss spikes, although these learning rates speed up the dynamics and accelerate the asymptotic convergence. In underparameterized regimes, we illustrate how the cross-entropy loss can lead to suboptimal memorization schemes. Finally, we assess the validity of our findings on small Transformer models."
                    },
                    {
                        "title": "Level Set Teleportation: An Optimization Perspective",
                        "abstract": "We study level set teleportation, an optimization sub-routine which seeks to accelerate gradient methods by maximizing the gradient norm on a level-set of the objective function. Since the descent lemma implies that gradient descent (GD) decreases the objective proportional to the squared norm of the gradient, level-set teleportation maximizes this one-step progress guarantee. For convex functions satisfying Hessian stability, we prove that GD with level-set teleportation obtains a combined sub-linear/linear convergence rate which is strictly faster than standard GD when the optimality gap is small. This is in sharp contrast to the standard (strongly) convex setting, where we show level-set teleportation neither improves nor worsens convergence rates. To evaluate teleportation in practice, we develop a projected-gradient-type method requiring only Hessian-vector products. We use this method to show that gradient methods with access to a teleportation oracle uniformly out-perform their standard versions on a variety of learning problems."
                    },
                    {
                        "title": "The SSL Interplay: Augmentations, Inductive Bias, and Generalization",
                        "abstract": "Self-supervised learning (SSL) has emerged as a powerful framework to learn representations from raw data without supervision. Yet in practice, engineers face issues such as instability in tuning optimizers and collapse of representations during training. Such challenges motivate the need for a theory to shed light on the complex interplay between the choice of data augmentation, network architecture, and training algorithm. We study such an interplay with a precise analysis of generalization performance on both pretraining and downstream tasks in a theory friendly setup, and highlight several insights for SSL practitioners that arise from our theory."
                    },
                    {
                        "title": "Birth of a Transformer: A Memory Viewpoint",
                        "abstract": "Large language models based on transformers have achieved great empirical successes. However, as they are deployed more widely, there is a growing need to better understand their internal mechanisms in order to make them more reliable. These models appear to store vast amounts of knowledge from their training data, and to adapt quickly to new information provided in their context or prompt. We study how transformers balance these two types of knowledge by considering a synthetic setup where tokens are generated from either global or context-specific bigram distributions. By a careful empirical analysis of the training process on a simplified two-layer transformer, we illustrate the fast learning of global bigrams and the slower development of an\"induction head\"mechanism for the in-context bigrams. We highlight the role of weight matrices as associative memories, provide theoretical insights on how gradients enable their learning during training, and study the role of data-distributional properties."
                    },
                    {
                        "title": "On Learning Gaussian Multi-index Models with Gradient Flow",
                        "abstract": "We study gradient flow on the multi-index regression problem for high-dimensional Gaussian data. Multi-index functions consist of a composition of an unknown low-rank linear projection and an arbitrary unknown, low-dimensional link function. As such, they constitute a natural template for feature learning in neural networks. We consider a two-timescale algorithm, whereby the low-dimensional link function is learnt with a non-parametric model infinitely faster than the subspace parametrizing the low-rank projection. By appropriately exploiting the matrix semigroup structure arising over the subspace correlation matrices, we establish global convergence of the resulting Grassmannian population gradient flow dynamics, and provide a quantitative description of its associated `saddle-to-saddle' dynamics. Notably, the timescales associated with each saddle can be explicitly characterized in terms of an appropriate Hermite decomposition of the target link function. In contrast with these positive results, we also show that the related \\emph{planted} problem, where the link function is known and fixed, in fact has a rough optimization landscape, in which gradient flow dynamics might get trapped with high probability."
                    },
                    {
                        "title": "When does return-conditioned supervised learning work for offline reinforcement learning?",
                        "abstract": "Several recent works have proposed a class of algorithms for the offline reinforcement learning (RL) problem that we will refer to as return-conditioned supervised learning (RCSL). RCSL algorithms learn the distribution of actions conditioned on both the state and the return of the trajectory. Then they define a policy by conditioning on achieving high return. In this paper, we provide a rigorous study of the capabilities and limitations of RCSL, something which is crucially missing in previous work. We find that RCSL returns the optimal policy under a set of assumptions that are stronger than those needed for the more traditional dynamic programming-based algorithms. We provide specific examples of MDPs and datasets that illustrate the necessity of these assumptions and the limits of RCSL. Finally, we present empirical evidence that these limitations will also cause issues in practice by providing illustrative experiments in simple point-mass environments and on datasets from the D4RL benchmark."
                    },
                    {
                        "title": "Efficient Kernel UCB for Contextual Bandits",
                        "abstract": "In this paper, we tackle the computational efficiency of kernelized UCB algorithms in contextual bandits. While standard methods require a O(CT^3) complexity where T is the horizon and the constant C is related to optimizing the UCB rule, we propose an efficient contextual algorithm for large-scale problems. Specifically, our method relies on incremental Nystrom approximations of the joint kernel embedding of contexts and actions. This allows us to achieve a complexity of O(CTm^2) where m is the number of Nystrom points. To recover the same regret as the standard kernelized UCB algorithm, m needs to be of order of the effective dimension of the problem, which is at most O(\\sqrt(T)) and nearly constant in some cases."
                    },
                    {
                        "title": "On the (Non-)Robustness of Two-Layer Neural Networks in Different Learning Regimes",
                        "abstract": "Neural networks are known to be highly sensitive to adversarial examples. These may arise due to different factors, such as random initialization, or spurious correlations in the learning problem. To better understand these factors, we provide a precise study of the adversarial robustness in different scenarios, from initialization to the end of training in different regimes, as well as intermediate scenarios, where initialization still plays a role due to\"lazy\"training. We consider over-parameterized networks in high dimensions with quadratic targets and infinite samples. Our analysis allows us to identify new tradeoffs between approximation (as measured via test error) and robustness, whereby robustness can only get worse when test error improves, and vice versa. We also show how linearized lazy training regimes can worsen robustness, due to improperly scaled random initialization. Our theoretical results are illustrated with numerical experiments."
                    },
                    {
                        "title": "Personalization Improves Privacy-Accuracy Tradeoffs in Federated Optimization",
                        "abstract": "Large-scale machine learning systems often involve data distributed across a collection of users. Federated optimization algorithms leverage this structure by communicating model updates to a central server, rather than entire datasets. In this paper, we study stochastic optimization algorithms for a personalized federated learning setting involving local and global models subject to user-level (joint) di\ufb00erential privacy. While learning a private global model induces a cost of privacy, local learning is perfectly private. We show that coordinating local learning with private centralized learning yields a generically useful and improved tradeo\ufb00 between accuracy and privacy. We illustrate our theoretical results with experiments on synthetic and real-world datasets"
                    },
                    {
                        "title": "Personalization Improves Privacy-Accuracy Tradeoffs in Federated Learning",
                        "abstract": "Large-scale machine learning systems often involve data distributed across a collection of users. Federated learning algorithms leverage this structure by communicating model updates to a central server, rather than entire datasets. In this paper, we study stochastic optimization algorithms for a personalized federated learning setting involving local and global models subject to user-level (joint) differential privacy. While learning a private global model induces a cost of privacy, local learning is perfectly private. We provide generalization guarantees showing that coordinating local learning with private centralized learning yields a generically useful and improved tradeoff between accuracy and privacy. We illustrate our theoretical results with experiments on synthetic and real-world datasets."
                    },
                    {
                        "title": "On Minimal Variations for Unsupervised Representation Learning",
                        "abstract": "Unsupervised representation learning aims at describing raw data efficiently to solve various downstream tasks. It has been approached with many techniques, such as manifold learning, diffusion maps, or more recently self-supervised learning. Those techniques are arguably all based on the underlying assumption that target functions, associated with future downstream tasks, have low variations in densely populated regions of the input space. Our main contribution is to unveil minimal variations as a guiding principle behind unsupervised representation learning paves the way to better practical guidelines for self-supervised learning algorithms."
                    },
                    {
                        "title": "Learning Single-Index Models with Shallow Neural Networks",
                        "abstract": "Single-index models are a class of functions given by an unknown univariate ``link'' function applied to an unknown one-dimensional projection of the input. These models are particularly relevant in high dimension, when the data might present low-dimensional structure that learning algorithms should adapt to. While several statistical aspects of this model, such as the sample complexity of recovering the relevant (one-dimensional) subspace, are well-understood, they rely on tailored algorithms that exploit the specific structure of the target function. In this work, we introduce a natural class of shallow neural networks and study its ability to learn single-index models via gradient flow. More precisely, we consider shallow networks in which biases of the neurons are frozen at random initialization. We show that the corresponding optimization landscape is benign, which in turn leads to generalization guarantees that match the near-optimal sample complexity of dedicated semi-parametric methods."
                    },
                    {
                        "title": "On the Universality of Graph Neural Networks on Large Random Graphs",
                        "abstract": "We study the approximation power of Graph Neural Networks (GNNs) on latent position random graphs. In the large graph limit, GNNs are known to converge to certain\"continuous\"models known as c-GNNs, which directly enables a study of their approximation power on random graph models. In the absence of input node features however, just as GNNs are limited by the Weisfeiler-Lehman isomorphism test, c-GNNs will be severely limited on simple random graph models. For instance, they will fail to distinguish the communities of a well-separated Stochastic Block Model (SBM) with constant degree function. Thus, we consider recently proposed architectures that augment GNNs with unique node identifiers, referred to as Structural GNNs here (SGNNs). We study the convergence of SGNNs to their continuous counterpart (c-SGNNs) in the large random graph limit, under new conditions on the node identifiers. We then show that c-SGNNs are strictly more powerful than c-GNNs in the continuous limit, and prove their universality on several random graph models of interest, including most SBMs and a large class of random geometric graphs. Our results cover both permutation-invariant and permutation-equivariant architectures."
                    },
                    {
                        "title": "On Approximation in Deep Convolutional Networks: a Kernel Perspective",
                        "abstract": ","
                    },
                    {
                        "title": "Approximation and Learning with Deep Convolutional Models: a Kernel Perspective",
                        "abstract": "The empirical success of deep convolutional networks on tasks involving high-dimensional data such as images or audio suggests that they can efficiently approximate certain functions that are well-suited for such tasks. In this paper, we study this through the lens of kernel methods, by considering simple hierarchical kernels with two or three convolution and pooling layers, inspired by convolutional kernel networks. These achieve good empirical performance on standard vision datasets, while providing a precise description of their functional space that yields new insights on their inductive bias. We show that the RKHS consists of additive models of interaction terms between patches, and that its norm encourages spatial similarities between these terms through pooling layers. We then provide generalization bounds which illustrate how pooling and patches yield improved sample complexity guarantees when the target function presents such regularities."
                    },
                    {
                        "title": "On Energy-Based Models with Overparametrized Shallow Neural Networks",
                        "abstract": "Energy-based models (EBMs) are a simple yet powerful framework for generative modeling. They are based on a trainable energy function which defines an associated Gibbs measure, and they can be trained and sampled from via well-established statistical tools, such as MCMC. Neural networks may be used as energy function approximators, providing both a rich class of expressive models as well as a flexible device to incorporate data structure. In this work we focus on shallow neural networks. Building from the incipient theory of overparametrized neural networks, we show that models trained in the so-called\"active\"regime provide a statistical advantage over their associated\"lazy\"or kernel regime, leading to improved adaptivity to hidden low-dimensional structure in the data distribution, as already observed in supervised learning. Our study covers both maximum likelihood and Stein Discrepancy estimators, and we validate our theoretical results with numerical experiments on synthetic data."
                    },
                    {
                        "title": "On the Sample Complexity of Learning under Invariance and Geometric Stability",
                        "abstract": "Many supervised learning problems involve high-dimensional data such as images, text, or graphs. In order to make efficient use of data, it is often useful to leverage certain geometric priors in the problem at hand, such as invariance to translations, permutation subgroups, or stability to small deformations. We study the sample complexity of learning problems where the target function presents such invariance and stability properties, by considering spherical harmonic decompositions of such functions on the sphere. We provide non-parametric rates of convergence for kernel methods, and show improvements in sample complexity by a factor equal to the size of the group when using an invariant kernel over the group, compared to the corresponding non-invariant kernel. These improvements are valid when the sample size is large enough, with an asymptotic behavior that depends on spectral properties of the group. Finally, these gains are extended beyond invariance groups to also cover geometric stability to small deformations, modeled here as subsets (not necessarily subgroups) of permutations."
                    }
                ]
            }
        }
    },
    "2406.16778": {
        "paper_data": {
            "title": "Finding Transformer Circuits with Edge Pruning",
            "url": "http://arxiv.org/abs/2406.16778v1",
            "arxiv_id": "2406.16778",
            "authors": [
                "Adithya Bhaskar",
                "Alexander Wettig",
                "Dan Friedman",
                "Danqi Chen"
            ],
            "abstract": "The path to interpreting a language model often proceeds via analysis of circuits -- sparse computational subgraphs of the model that capture specific aspects of its behavior. Recent work has automated the task of discovering circuits. Yet, these methods have practical limitations, as they rely either on inefficient search algorithms or inaccurate approximations. In this paper, we frame automated circuit discovery as an optimization problem and propose *Edge Pruning* as an effective and scalable solution. Edge Pruning leverages gradient-based pruning techniques, but instead of removing neurons or components, it prunes the \\emph{edges} between components. Our method finds circuits in GPT-2 that use less than half the number of edges compared to circuits found by previous methods while being equally faithful to the full model predictions on standard circuit-finding tasks. Edge Pruning is efficient even with as many as 100K examples, outperforming previous methods in speed and producing substantially better circuits. It also perfectly recovers the ground-truth circuits in two models compiled with Tracr. Thanks to its efficiency, we scale Edge Pruning to CodeLlama-13B, a model over 100x the scale that prior methods operate on. We use this setting for a case study comparing the mechanisms behind instruction prompting and in-context learning. We find two circuits with more than 99.96% sparsity that match the performance of the full model and reveal that the mechanisms in the two settings overlap substantially. Our case study shows that Edge Pruning is a practical and scalable tool for interpretability and sheds light on behaviors that only emerge in large models.",
            "introduction": "   1 Introduction     (a) Regular Transformer     (b) Optimize edge masks     (c) Obtain sparse circuit    Figure 1: Edge Pruning disentangles the residual stream and optimizes continuous masks on the read operations via gradient descent. Discretizing the masks to {0,1}01\\{0,1\\}{ 0 , 1 } yields the final circuit. The full model corresponds to the case where all masks equal 1111.    Mechanistic interpretability strives to understand models via bottom-up descriptions of their components (e.g., attention heads and MLPs in Transformers\u00a0(Vaswani et\u00a0al., 2017)). This typically proceeds via the identification and analysis of a circuit\u00a0(Olah et\u00a0al., 2020; Elhage et\u00a0al., 2021)\u2014a sparse computational subgraph of the model that captures the aspects of its behavior we wish to study. The arduous process of identifying circuits was recently automated by ACDC\u00a0(Conmy et\u00a0al., 2023) and EAP\u00a0(Syed et\u00a0al., 2023). However, ACDC uses an expensive greedy search that ablates each edge to estimate its importance. It cannot scale to datasets beyond a few hundred examples or to billion-parameter models. EAP, on the other hand, uses gradient-based linear approximations of activation patching to estimate the importance of all edges simultaneously. While fast, these approximations often sacrifice faithfulness to the full model. Besides, this approach ignores the impact of the presence/absence of other edges on the score.   In this paper, we frame circuit discovery as an optimization problem and tackle it via gradient-based pruning, rather than discrete search or first-order approximations. As such, we adapt pruning for the goal of circuit discovery instead of model compression. Rather than components, we prune the edges between components and replace missing edges with counterfactual activations from corrupted examples. We enable this by replacing the residual stream of a Transformer (Figure\u00a01a) with a disentangled residual stream (Lindner et\u00a0al., 2023; Friedman et\u00a0al., 2023), which retains a list of all previous activations. This allows us to introduce edge masks that determine from which components to read. We then leverage discrete optimization techniques such as L0subscript\ud835\udc3f0L_{0}italic_L start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT regularization\u00a0(Louizos et\u00a0al., 2018) to optimize these edge masks and produce sparse circuits (Figure\u00a01c).   We evaluate our approach, Edge Pruning, on four fronts: (1) we measure how faithfully the discovered circuits describe the behavior of the full model, (2) we verify if it can recover ground-truth circuits in Tracr models\u00a0(Lindner et\u00a0al., 2023) compiled from known program descriptions, (3) we evaluate how the method scales to more examples and (4) we assess its ability to find extremely sparse circuits in multi-billion parameter models. On four standard circuit-finding tasks, Edge Pruning finds circuits in GPT-2 Small\u00a0(Radford et\u00a0al., 2019) that are consistently more faithful to the full model and have better task performance than circuits found by prior methods. The gap is especially pronounced on more complex tasks like multi-template IOI\u00a0(Wang et\u00a0al., 2023), where we find circuits that have 2.65\u00d72.65\\times2.65 \u00d7 fewer edges but describe model outputs just as faithfully as the circuit found by the next-best method. We show that Edge Pruning scales effectively to a version of IOI with 100K examples, where it outperforms prior methods in terms of speed and performance. Edge Pruning also perfectly recovers ground-truth circuits in two models compiled from known program descriptions with Tracr.   Finally, we establish that Edge Pruning scales to CodeLlama-13B\u00a0(Rozi\u00e8re et\u00a0al., 2024)\u2014100\u00d7\\times\u00d7 the size of models typically tackled by automated",
            "references": [
                {
                    "title": "Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanisms",
                    "abstract": "Many recent language model (LM) interpretability studies have adopted the circuits framework, which aims to find the minimal computational subgraph, or circuit, that explains LM behavior on a given task. Most studies determine which edges belong in a LM's circuit by performing causal interventions on each edge independently, but this scales poorly with model size. Edge attribution patching (EAP), gradient-based approximation to interventions, has emerged as a scalable but imperfect solution to this problem. In this paper, we introduce a new method - EAP with integrated gradients (EAP-IG) - that aims to better maintain a core property of circuits: faithfulness. A circuit is faithful if all model edges outside the circuit can be ablated without changing the model's performance on the task; faithfulness is what justifies studying circuits, rather than the full model. Our experiments demonstrate that circuits found using EAP are less faithful than those found using EAP-IG, even though both have high node overlap with circuits found previously using causal interventions. We conclude more generally that when using circuits to compare the mechanisms models use to solve tasks, faithfulness, not overlap, is what should be measured."
                },
                {
                    "title": "AtP*: An efficient and scalable method for localizing LLM behaviour to components",
                    "abstract": "Activation Patching is a method of directly computing causal attributions of behavior to model components. However, applying it exhaustively requires a sweep with cost scaling linearly in the number of model components, which can be prohibitively expensive for SoTA Large Language Models (LLMs). We investigate Attribution Patching (AtP), a fast gradient-based approximation to Activation Patching and find two classes of failure modes of AtP which lead to significant false negatives. We propose a variant of AtP called AtP*, with two changes to address these failure modes while retaining scalability. We present the first systematic study of AtP and alternative methods for faster activation patching and show that AtP significantly outperforms all other investigated methods, with AtP* providing further significant improvement. Finally, we provide a method to bound the probability of remaining false negatives of AtP* estimates."
                },
                {
                    "title": "Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking",
                    "abstract": "Fine-tuning on generalized tasks such as instruction following, code generation, and mathematics has been shown to enhance language models' performance on a range of tasks. Nevertheless, explanations of how such fine-tuning influences the internal computations in these models remain elusive. We study how fine-tuning affects the internal mechanisms implemented in language models. As a case study, we explore the property of entity tracking, a crucial facet of language comprehension, where models fine-tuned on mathematics have substantial performance gains. We identify the mechanism that enables entity tracking and show that (i) in both the original model and its fine-tuned versions primarily the same circuit implements entity tracking. In fact, the entity tracking circuit of the original model on the fine-tuned versions performs better than the full original model. (ii) The circuits of all the models implement roughly the same functionality: Entity tracking is performed by tracking the position of the correct entity in both the original model and its fine-tuned versions. (iii) Performance boost in the fine-tuned models is primarily attributed to its improved ability to handle the augmented positional information. To uncover these findings, we employ: Patch Patching, DCM, which automatically detects model components responsible for specific semantics, and CMAP, a new approach for patching activations across models to reveal improved mechanisms. Our findings suggest that fine-tuning enhances, rather than fundamentally alters, the mechanistic operation of the model."
                },
                {
                    "title": "Is This the Subspace You Are Looking for? An Interpretability Illusion for Subspace Activation Patching",
                    "abstract": "Mechanistic interpretability aims to understand model behaviors in terms of specific, interpretable features, often hypothesized to manifest as low-dimensional subspaces of activations. Specifically, recent studies have explored subspace interventions (such as activation patching) as a way to simultaneously manipulate model behavior and attribute the features behind it to given subspaces. In this work, we demonstrate that these two aims diverge, potentially leading to an illusory sense of interpretability. Counterintuitively, even if a subspace intervention makes the model's output behave as if the value of a feature was changed, this effect may be achieved by activating a dormant parallel pathway leveraging another subspace that is causally disconnected from model outputs. We demonstrate this phenomenon in a distilled mathematical example, in two real-world domains (the indirect object identification task and factual recall), and present evidence for its prevalence in practice. In the context of factual recall, we further show a link to rank-1 fact editing, providing a mechanistic explanation for previous work observing an inconsistency between fact editing performance and fact localization. However, this does not imply that activation patching of subspaces is intrinsically unfit for interpretability. To contextualize our findings, we also show what a success case looks like in a task (indirect object identification) where prior manual circuit analysis informs an understanding of the location of a feature. We explore the additional evidence needed to argue that a patched subspace is faithful."
                },
                {
                    "title": "Mechanistically analyzing the effects of fine-tuning on procedurally defined tasks",
                    "abstract": "Fine-tuning large pre-trained models has become the de facto strategy for developing both task-specific and general-purpose machine learning systems, including developing models that are safe to deploy. Despite its clear importance, there has been minimal work that explains how fine-tuning alters the underlying capabilities learned by a model during pretraining: does fine-tuning yield entirely novel capabilities or does it just modulate existing ones? We address this question empirically in synthetic, controlled settings where we can use mechanistic interpretability tools (e.g., network pruning and probing) to understand how the model's underlying capabilities are changing. We perform an extensive analysis of the effects of fine-tuning in these settings, and show that: (i) fine-tuning rarely alters the underlying model capabilities; (ii) a minimal transformation, which we call a 'wrapper', is typically learned on top of the underlying model capabilities, creating the illusion that they have been modified; and (iii) further fine-tuning on a task where such hidden capabilities are relevant leads to sample-efficient 'revival' of the capability, i.e., the model begins reusing these capability after only a few gradient steps. This indicates that practitioners can unintentionally remove a model's safety wrapper merely by fine-tuning it on a, e.g., superficially unrelated, downstream task. We additionally perform analysis on language models trained on the TinyStories dataset to support our claims in a more realistic setup."
                },
                {
                    "title": "Attention Lens: A Tool for Mechanistically Interpreting the Attention Head Information Retrieval Mechanism",
                    "abstract": "Transformer-based Large Language Models (LLMs) are the state-of-the-art for natural language tasks. Recent work has attempted to decode, by reverse engineering the role of linear layers, the internal mechanisms by which LLMs arrive at their final predictions for text completion tasks. Yet little is known about the specific role of attention heads in producing the final token prediction. We propose Attention Lens, a tool that enables researchers to translate the outputs of attention heads into vocabulary tokens via learned attention-head-specific transformations called lenses. Preliminary findings from our trained lenses indicate that attention heads play highly specialized roles in language models. The code for Attention Lens is available at github.com/msakarvadia/AttentionLens."
                },
                {
                    "title": "Attribution Patching Outperforms Automated Circuit Discovery",
                    "abstract": "Automated interpretability research has recently attracted attention as a potential research direction that could scale explanations of neural network behavior to large models. Existing automated circuit discovery work applies activation patching to identify subnetworks responsible for solving specific tasks (circuits). In this work, we show that a simple method based on attribution patching outperforms all existing methods while requiring just two forward passes and a backward pass. We apply a linear approximation to activation patching to estimate the importance of each edge in the computational subgraph. Using this approximation, we prune the least important edges of the network. We survey the performance and limitations of this method, finding that averaged over all tasks our method has greater AUC from circuit recovery than other methods."
                },
                {
                    "title": "Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning",
                    "abstract": "The popularity of LLaMA (Touvron et al., 2023a;b) and other recently emerged moderate-sized large language models (LLMs) highlights the potential of building smaller yet powerful LLMs. Regardless, the cost of training such models from scratch on trillions of tokens remains high. In this work, we study structured pruning as an effective means to develop smaller LLMs from pre-trained, larger models. Our approach employs two key techniques: (1) targeted structured pruning, which prunes a larger model to a specified target shape by removing layers, heads, and intermediate and hidden dimensions in an end-to-end manner, and (2) dynamic batch loading, which dynamically updates the composition of sampled data in each training batch based on varying losses across different domains. We demonstrate the efficacy of our approach by presenting the Sheared-LLaMA series, pruning the LLaMA2-7B model down to 1.3B and 2.7B parameters. Sheared-LLaMA models outperform state-of-the-art open-source models of equivalent sizes, such as Pythia, INCITE, OpenLLaMA and the concurrent TinyLlama models, on a wide range of downstream and instruction tuning evaluations, while requiring only 3% of compute compared to training such models from scratch. This work provides compelling evidence that leveraging existing LLMs with structured pruning is a far more cost-effective approach for building competitive small-scale LLMs"
                },
                {
                    "title": "Towards Best Practices of Activation Patching in Language Models: Metrics and Methods",
                    "abstract": "Mechanistic interpretability seeks to understand the internal mechanisms of machine learning models, where localization -- identifying the important model components -- is a key step. Activation patching, also known as causal tracing or interchange intervention, is a standard technique for this task (Vig et al., 2020), but the literature contains many variants with little consensus on the choice of hyperparameters or methodology. In this work, we systematically examine the impact of methodological details in activation patching, including evaluation metrics and corruption methods. In several settings of localization and circuit discovery in language models, we find that varying these hyperparameters could lead to disparate interpretability results. Backed by empirical observations, we give conceptual arguments for why certain metrics or methods may be preferred. Finally, we provide recommendations for the best practices of activation patching going forwards."
                },
                {
                    "title": "Sparse Autoencoders Find Highly Interpretable Features in Language Models",
                    "abstract": "One of the roadblocks to a better understanding of neural networks' internals is \\textit{polysemanticity}, where neurons appear to activate in multiple, semantically distinct contexts. Polysemanticity prevents us from identifying concise, human-understandable explanations for what neural networks are doing internally. One hypothesised cause of polysemanticity is \\textit{superposition}, where neural networks represent more features than they have neurons by assigning features to an overcomplete set of directions in activation space, rather than to individual neurons. Here, we attempt to identify those directions, using sparse autoencoders to reconstruct the internal activations of a language model. These autoencoders learn sets of sparsely activating features that are more interpretable and monosemantic than directions identified by alternative approaches, where interpretability is measured by automated methods. Moreover, we show that with our learned set of features, we can pinpoint the features that are causally responsible for counterfactual behaviour on the indirect object identification task \\citep{wang2022interpretability} to a finer degree than previous decompositions. This work indicates that it is possible to resolve superposition in language models using a scalable, unsupervised method. Our method may serve as a foundation for future mechanistic interpretability work, which we hope will enable greater model transparency and steerability."
                },
                {
                    "title": "The Hydra Effect: Emergent Self-repair in Language Model Computations",
                    "abstract": "We investigate the internal structure of language model computations using causal analysis and demonstrate two motifs: (1) a form of adaptive computation where ablations of one attention layer of a language model cause another layer to compensate (which we term the Hydra effect) and (2) a counterbalancing function of late MLP layers that act to downregulate the maximum-likelihood token. Our ablation studies demonstrate that language model layers are typically relatively loosely coupled (ablations to one layer only affect a small number of downstream layers). Surprisingly, these effects occur even in language models trained without any form of dropout. We analyse these effects in the context of factual recall and consider their implications for circuit-level attribution in language models."
                },
                {
                    "title": "Does Circuit Analysis Interpretability Scale? Evidence from Multiple Choice Capabilities in Chinchilla",
                    "abstract": "\\emph{Circuit analysis} is a promising technique for understanding the internal mechanisms of language models. However, existing analyses are done in small models far from the state of the art. To address this, we present a case study of circuit analysis in the 70B Chinchilla model, aiming to test the scalability of circuit analysis. In particular, we study multiple-choice question answering, and investigate Chinchilla's capability to identify the correct answer \\emph{label} given knowledge of the correct answer \\emph{text}. We find that the existing techniques of logit attribution, attention pattern visualization, and activation patching naturally scale to Chinchilla, allowing us to identify and categorize a small set of `output nodes' (attention heads and MLPs). We further study the `correct letter' category of attention heads aiming to understand the semantics of their features, with mixed results. For normal multiple-choice question answers, we significantly compress the query, key and value subspaces of the head without loss of performance when operating on the answer labels for multiple-choice questions, and we show that the query and key subspaces represent an `Nth item in an enumeration' feature to at least some extent. However, when we attempt to use this explanation to understand the heads' behaviour on a more general distribution including randomized answer labels, we find that it is only a partial explanation, suggesting there is more to learn about the operation of `correct letter' heads on multiple choice question answering."
                },
                {
                    "title": "Learning Transformer Programs",
                    "abstract": "Recent research in mechanistic interpretability has attempted to reverse-engineer Transformer models by carefully inspecting network weights and activations. However, these approaches require considerable manual effort and still fall short of providing complete, faithful descriptions of the underlying algorithms. In this work, we introduce a procedure for training Transformers that are mechanistically interpretable by design. We build on RASP [Weiss et al., 2021], a programming language that can be compiled into Transformer weights. Instead of compiling human-written programs into Transformers, we design a modified Transformer that can be trained using gradient-based optimization and then automatically converted into a discrete, human-readable program. We refer to these models as Transformer Programs. To validate our approach, we learn Transformer Programs for a variety of problems, including an in-context learning task, a suite of algorithmic problems (e.g. sorting, recognizing Dyck languages), and NLP tasks including named entity recognition and text classification. The Transformer Programs can automatically find reasonable solutions, performing on par with standard Transformers of comparable size; and, more importantly, they are easy to interpret. To demonstrate these advantages, we convert Transformers into Python programs and use off-the-shelf code analysis tools to debug model errors and identify the\"circuits\"used to solve different sub-problems. We hope that Transformer Programs open a new path toward the goal of intrinsically interpretable machine learning."
                },
                {
                    "title": "VISIT: Visualizing and Interpreting the Semantic Information Flow of Transformers",
                    "abstract": "Recent advances in interpretability suggest we can project weights and hidden states of transformer-based language models (LMs) to their vocabulary, a transformation that makes them more human interpretable. In this paper, we investigate LM attention heads and memory values, the vectors the models dynamically create and recall while processing a given input. By analyzing the tokens they represent through this projection, we identify patterns in the information flow inside the attention mechanism. Based on our discoveries, we create a tool to visualize a forward pass of Generative Pre-trained Transformers (GPTs) as an interactive flow graph, with nodes representing neurons or hidden states and edges representing the interactions between them. Our visualization simplifies huge amounts of data into easy-to-read plots that can reflect the models' internal processing, uncovering the contribution of each component to the models' final prediction. Our visualization also unveils new insights about the role of layer norms as semantic filters that influence the models' output, and about neurons that are always activated during forward passes and act as regularization vectors."
                },
                {
                    "title": "Interpretability at Scale: Identifying Causal Mechanisms in Alpaca",
                    "abstract": "Obtaining human-interpretable explanations of large, general-purpose language models is an urgent goal for AI safety. However, it is just as important that our interpretability methods are faithful to the causal dynamics underlying model behavior and able to robustly generalize to unseen inputs. Distributed Alignment Search (DAS) is a powerful gradient descent method grounded in a theory of causal abstraction that has uncovered perfect alignments between interpretable symbolic algorithms and small deep learning models fine-tuned for specific tasks. In the present paper, we scale DAS significantly by replacing the remaining brute-force search steps with learned parameters -- an approach we call Boundless DAS. This enables us to efficiently search for interpretable causal structure in large language models while they follow instructions. We apply Boundless DAS to the Alpaca model (7B parameters), which, off the shelf, solves a simple numerical reasoning problem. With Boundless DAS, we discover that Alpaca does this by implementing a causal model with two interpretable boolean variables. Furthermore, we find that the alignment of neural representations with these variables is robust to changes in inputs and instructions. These findings mark a first step toward faithfully understanding the inner-workings of our ever-growing and most widely deployed language models. Our tool is extensible to larger LLMs and is released publicly at `https://github.com/stanfordnlp/pyvene`."
                },
                {
                    "title": "How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language model",
                    "abstract": "Pre-trained language models can be surprisingly adept at tasks they were not explicitly trained on, but how they implement these capabilities is poorly understood. In this paper, we investigate the basic mathematical abilities often acquired by pre-trained language models. Concretely, we use mechanistic interpretability techniques to explain the (limited) mathematical abilities of GPT-2 small. As a case study, we examine its ability to take in sentences such as\"The war lasted from the year 1732 to the year 17\", and predict valid two-digit end years (years>32). We first identify a circuit, a small subset of GPT-2 small's computational graph that computes this task's output. Then, we explain the role of each circuit component, showing that GPT-2 small's final multi-layer perceptrons boost the probability of end years greater than the start year. Finally, we find related tasks that activate our circuit. Our results suggest that GPT-2 small computes greater-than using a complex but general mechanism that activates across diverse contexts."
                },
                {
                    "title": "Towards Automated Circuit Discovery for Mechanistic Interpretability",
                    "abstract": "Through considerable effort and intuition, several recent works have reverse-engineered nontrivial behaviors of transformer models. This paper systematizes the mechanistic interpretability process they followed. First, researchers choose a metric and dataset that elicit the desired model behavior. Then, they apply activation patching to find which abstract neural network units are involved in the behavior. By varying the dataset, metric, and units under investigation, researchers can understand the functionality of each component. We automate one of the process' steps: to identify the circuit that implements the specified behavior in the model's computational graph. We propose several algorithms and reproduce previous interpretability results to validate them. For example, the ACDC algorithm rediscovered 5/5 of the component types in a circuit in GPT-2 Small that computes the Greater-Than operation. ACDC selected 68 of the 32,000 edges in GPT-2 Small, all of which were manually found by previous work. Our code is available at https://github.com/ArthurConmy/Automatic-Circuit-Discovery."
                },
                {
                    "title": "Localizing Model Behavior with Path Patching",
                    "abstract": "Localizing behaviors of neural networks to a subset of the network's components or a subset of interactions between components is a natural first step towards analyzing network mechanisms and possible failure modes. Existing work is often qualitative and ad-hoc, and there is no consensus on the appropriate way to evaluate localization claims. We introduce path patching, a technique for expressing and quantitatively testing a natural class of hypotheses expressing that behaviors are localized to a set of paths. We refine an explanation of induction heads, characterize a behavior of GPT-2, and open source a framework for efficiently running similar experiments."
                },
                {
                    "title": "Finding Alignments Between Interpretable Causal Variables and Distributed Neural Representations",
                    "abstract": "Causal abstraction is a promising theoretical framework for explainable artificial intelligence that defines when an interpretable high-level causal model is a faithful simplification of a low-level deep learning system. However, existing causal abstraction methods have two major limitations: they require a brute-force search over alignments between the high-level model and the low-level one, and they presuppose that variables in the high-level model will align with disjoint sets of neurons in the low-level one. In this paper, we present distributed alignment search (DAS), which overcomes these limitations. In DAS, we find the alignment between high-level and low-level models using gradient descent rather than conducting a brute-force search, and we allow individual neurons to play multiple distinct roles by analyzing representations in non-standard bases-distributed representations. Our experiments show that DAS can discover internal structure that prior approaches miss. Overall, DAS removes previous obstacles to conducting causal abstraction analyses and allows us to find conceptual structure in trained neural nets."
                },
                {
                    "title": "Tracr: Compiled Transformers as a Laboratory for Interpretability",
                    "abstract": "We show how to\"compile\"human-readable programs into standard decoder-only transformer models. Our compiler, Tracr, generates models with known structure. This structure can be used to design experiments. For example, we use it to study\"superposition\"in transformers that execute multi-step algorithms. Additionally, the known structure of Tracr-compiled models can serve as ground-truth for evaluating interpretability methods. Commonly, because the\"programs\"learned by transformers are unknown it is unclear whether an interpretation succeeded. We demonstrate our approach by implementing and examining programs including computing token frequencies, sorting, and parenthesis checking. We provide an open-source implementation of Tracr at https://github.com/deepmind/tracr."
                },
                {
                    "title": "Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small",
                    "abstract": "Research in mechanistic interpretability seeks to explain behaviors of machine learning models in terms of their internal components. However, most previous work either focuses on simple behaviors in small models, or describes complicated behaviors in larger models with broad strokes. In this work, we bridge this gap by presenting an explanation for how GPT-2 small performs a natural language task called indirect object identification (IOI). Our explanation encompasses 26 attention heads grouped into 7 main classes, which we discovered using a combination of interpretability approaches relying on causal interventions. To our knowledge, this investigation is the largest end-to-end attempt at reverse-engineering a natural behavior\"in the wild\"in a language model. We evaluate the reliability of our explanation using three quantitative criteria--faithfulness, completeness and minimality. Though these criteria support our explanation, they also point to remaining gaps in our understanding. Our work provides evidence that a mechanistic understanding of large ML models is feasible, opening opportunities to scale our understanding to both larger models and more complex tasks."
                },
                {
                    "title": "Structured Pruning Learns Compact and Accurate Models",
                    "abstract": "The growing size of neural language models has led to increased attention in model compression. The two predominant approaches are pruning, which gradually removes weights from a pre-trained model, and distillation, which trains a smaller compact model to match a larger one. Pruning methods can significantly reduce the model size but hardly achieve large speedups as distillation. However, distillation methods require large amounts of unlabeled data and are expensive to train. In this work, we propose a task-specific structured pruning method CoFi (Coarse- and Fine-grained Pruning), which delivers highly parallelizable subnetworks and matches the distillation methods in both accuracy and latency, without resorting to any unlabeled data. Our key insight is to jointly prune coarse-grained (e.g., layers) and fine-grained (e.g., heads and hidden units) modules, which controls the pruning decision of each parameter with masks of different granularity. We also devise a layerwise distillation strategy to transfer knowledge from unpruned to pruned models during optimization. Our experiments on GLUE and SQuAD datasets show that CoFi yields models with over 10X speedups with a small accuracy drop, showing its effectiveness and efficiency compared to previous pruning and distillation approaches."
                },
                {
                    "title": "Thinking Like Transformers",
                    "abstract": "What is the computational model behind a Transformer? Where recurrent neural networks have direct parallels in finite state machines, allowing clear discussion and thought around architecture variants or trained models, Transformers have no such familiar parallel. In this paper we aim to change that, proposing a computational model for the transformer-encoder in the form of a programming language. We map the basic components of a transformer-encoder -- attention and feed-forward computation -- into simple primitives, around which we form a programming language: the Restricted Access Sequence Processing Language (RASP). We show how RASP can be used to program solutions to tasks that could conceivably be learned by a Transformer, and how a Transformer can be trained to mimic a RASP solution. In particular, we provide RASP programs for histograms, sorting, and Dyck-languages. We further use our model to relate their difficulty in terms of the number of required layers and attention heads: analyzing a RASP program implies a maximum number of heads and layers necessary to encode a task in a transformer. Finally, we see how insights gained from our abstraction might be used to explain phenomena seen in recent works."
                },
                {
                    "title": "Low-Complexity Probing via Finding Subnetworks",
                    "abstract": "The dominant approach in probing neural networks for linguistic properties is to train a new shallow multi-layer perceptron (MLP) on top of the model\u2019s internal representations. This approach can detect properties encoded in the model, but at the cost of adding new parameters that may learn the task directly. We instead propose a subtractive pruning-based probe, where we find an existing subnetwork that performs the linguistic task of interest. Compared to an MLP, the subnetwork probe achieves both higher accuracy on pre-trained models and lower accuracy on random models, so it is both better at finding properties of interest and worse at learning on its own. Next, by varying the complexity of each probe, we show that subnetwork probing Pareto-dominates MLP probing in that it achieves higher accuracy given any budget of probe complexity. Finally, we analyze the resulting subnetworks across various tasks to locate where each task is encoded, and we find that lower-level tasks are captured in lower layers, reproducing similar findings in past work."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Neural Natural Language Inference Models Partially Embed Theories of Lexical Entailment and Negation",
                    "abstract": "We address whether neural models for Natural Language Inference (NLI) can learn the compositional interactions between lexical entailment and negation, using four methods: the behavioral evaluation methods of (1) challenge test sets and (2) systematic generalization tasks, and the structural evaluation methods of (3) probes and (4) interventions. To facilitate this holistic evaluation, we present Monotonicity NLI (MoNLI), a new naturalistic dataset focused on lexical entailment and negation. In our behavioral evaluations, we find that models trained on general-purpose NLI datasets fail systematically on MoNLI examples containing negation, but that MoNLI fine-tuning addresses this failure. In our structural evaluations, we look for evidence that our top-performing BERT-based model has learned to implement the monotonicity algorithm behind MoNLI. Probes yield evidence consistent with this conclusion, and our intervention experiments bolster this, showing that the causal dynamics of the model mirror the causal dynamics of this algorithm on subsets of MoNLI. This suggests that the BERT model at least partially embeds a theory of lexical entailment and negation at an algorithmic level."
                },
                {
                    "title": "Structured Pruning of Large Language Models",
                    "abstract": "Large language models have recently achieved state of the art performance across a wide variety of natural language tasks. Meanwhile, the size of these models and their latency have significantly increased, which makes their usage costly, and raises an interesting question: do language models need to be large? We study this question through the lens of model compression. We present a novel, structured pruning approach based on low rank factorization and augmented Lagrangian L0 norm regularization. Our structured approach achieves significant inference speedups while matching or outperforming our unstructured pruning baseline at various sparsity levels. We apply our method to state of the art models on the enwiki8 dataset and obtain a 1.19 perplexity score with just 5M parameters, vastly outperforming a model of the same size trained from scratch. We also demonstrate that our method can be applied to language model fine-tuning by pruning the BERT model on several downstream classification benchmarks."
                },
                {
                    "title": "Are Sixteen Heads Really Better than One?",
                    "abstract": "Attention is a powerful and ubiquitous mechanism for allowing neural models to focus on particular salient pieces of information by taking their weighted average when making predictions. In particular, multi-headed attention is a driving force behind many recent state-of-the-art NLP models such as Transformer-based MT models and BERT. These models apply multiple attention mechanisms in parallel, with each attention \"head\" potentially focusing on different parts of the input, which makes it possible to express sophisticated functions beyond the simple weighted average. In this paper we make the surprising observation that even if models have been trained using multiple heads, in practice, a large percentage of attention heads can be removed at test time without significantly impacting performance. In fact, some layers can even be reduced to a single head. We further examine greedy algorithms for pruning down models, and the potential speed, memory efficiency, and accuracy improvements obtainable therefrom. Finally, we analyze the results with respect to which parts of the model are more reliant on having multiple heads, and provide precursory evidence that training dynamics play a role in the gains provided by multi-head attention."
                },
                {
                    "title": "Learning Sparse Neural Networks through L0 Regularization",
                    "abstract": "We propose a practical method for $L_0$ norm regularization for neural networks: pruning the network during training by encouraging weights to become exactly zero. Such regularization is interesting since (1) it can greatly speed up training and inference, and (2) it can improve generalization. AIC and BIC, well-known model selection criteria, are special cases of $L_0$ regularization. However, since the $L_0$ norm of weights is non-differentiable, we cannot incorporate it directly as a regularization term in the objective function. We propose a solution through the inclusion of a collection of non-negative stochastic gates, which collectively determine which weights to set to zero. We show that, somewhat surprisingly, for certain distributions over the gates, the expected $L_0$ norm of the resulting gated weights is differentiable with respect to the distribution parameters. We further propose the \\emph{hard concrete} distribution for the gates, which is obtained by \"stretching\" a binary concrete distribution and then transforming its samples with a hard-sigmoid. The parameters of the distribution over the gates can then be jointly optimized with the original network parameters. As a result our method allows for straightforward and efficient learning of model structures with stochastic gradient descent and allows for conditional computation in a principled way. We perform various experiments to demonstrate the effectiveness of the resulting approach and regularizer."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Second Order Derivatives for Network Pruning: Optimal Brain Surgeon",
                    "abstract": "We investigate the use of information from all second order derivatives of the error function to perform network pruning (i.e., removing unimportant weights from a trained network) in order to improve generalization, simplify networks, reduce hardware or storage requirements, increase the speed of further training, and in some cases enable rule extraction. Our method, Optimal Brain Surgeon (OBS), is Significantly better than magnitude-based methods and Optimal Brain Damage [Le Cun, Denker and Solla, 1990], which often remove the wrong weights. OBS permits the pruning of more weights than other methods (for the same error on the training set), and thus yields better generalization on test data. Crucial to OBS is a recursion relation for calculating the inverse Hessian matrix H-1 from training data and structural information of the net. OBS permits a 90%, a 76%, and a 62% reduction in weights over backpropagation with weight decay on three benchmark MONK's problems [Thrun et al., 1991]. Of OBS, Optimal Brain Damage, and magnitude-based methods, only OBS deletes the correct weights from a trained XOR network in every case. Finally, whereas Sejnowski and Rosenberg [1987] used 18,000 weights in their NETtalk network, we used OBS to prune a network to just 1560 weights, yielding better generalization."
                },
                {
                    "title": "Investigating Gender Bias in Language Models Using Causal Mediation Analysis",
                    "abstract": "Many interpretation methods for neural models in natural language processing investigate how information is encoded inside hidden representations. However, these methods can only measure whether the information exists, not whether it is actually used by the model. We propose a methodology grounded in the theory of causal mediation analysis for interpreting which parts of a model are causally implicated in its behavior. The approach enables us to analyze the mechanisms that facilitate the \ufb02ow of information from input to output through various model components, known as mediators. As a case study, we apply this methodology to analyzing gender bias in pre-trained Transformer language models. We study the role of individual neurons and attention heads in mediating gender bias across three datasets designed to gauge a model\u2019s sensitivity to gender bias. Our mediation analysis reveals that gender bias effects are concentrated in speci\ufb01c components of the model that may exhibit highly specialized behavior."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Optimal Brain Damage",
                    "abstract": "We have used information-theoretic ideas to derive a class of practical and nearly optimal schemes for adapting the size of a neural network. By removing unimportant weights from a network, several improvements can be expected: better generalization, fewer training examples required, and improved speed of learning and/or classification. The basic idea is to use second-derivative information to make a tradeoff between network complexity and training set error. Experiments confirm the usefulness of the methods on a real-world application."
                }
            ],
            "categories": [
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively discover and optimize sparse circuits in large-scale Transformer models using gradient-based pruning techniques?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing mechanistic interpretability in machine learning, particularly in understanding the inner workings of complex models like Transformers. By improving circuit discovery methods, we can enhance our ability to analyze model behavior, leading to better model design and optimization. This research could pave the way for more efficient model architectures, facilitate debugging, and contribute to the development of interpretable AI systems, ultimately influencing future research directions in both theoretical and practical applications.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of large-scale models and the limitations of existing methods. Naive approaches may fail due to their inability to scale effectively or accurately capture the interactions between edges in the model. Technical obstacles include the need for efficient optimization techniques that can handle the high dimensionality of the problem space, as well as the requirement for methods that maintain faithfulness to the full model's behavior while discovering sparse circuits. Additionally, the reliance on discrete search methods or first-order approximations can lead to suboptimal results.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the computational expense of greedy search methods, which cannot scale to larger datasets or models, and by the shortcomings of gradient-based approximations that sacrifice accuracy for speed. These barriers have hindered the development of effective circuit discovery techniques. Our approach differs by framing circuit discovery as an optimization problem and utilizing gradient-based pruning, which allows for a more scalable and faithful discovery of circuits compared to prior methods.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Edge Pruning, involves replacing the residual stream of a Transformer with a disentangled residual stream to introduce edge masks that determine the reading components. We employ L0 regularization to optimize these masks and produce sparse circuits. We will evaluate our approach using metrics that measure the faithfulness of the discovered circuits, recovery of ground-truth circuits, scalability to larger datasets, and performance in multi-billion parameter models. Expected outcomes include the discovery of circuits that are more faithful and efficient than those found by existing methods, particularly in complex tasks."
            }
        },
        "author_data": {
            "288312a0-63bc-43d0-b5be-468783f4014a": {
                "pk": "288312a0-63bc-43d0-b5be-468783f4014a",
                "name": "Adithya Bhaskar",
                "collaborators": [
                    "Danqi Chen",
                    "Alexander R. Fabbri",
                    "Greg Durrett",
                    "Dan Friedman",
                    "Tushar Tomar",
                    "Ashutosh Sathe",
                    "Sunita Sarawagi",
                    "Tianyu Gao",
                    "Zirui Wang",
                    "Richeek Das"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Language Models",
                    "Text-to-SQL"
                ],
                "publications": [
                    {
                        "title": "Prompted Opinion Summarization with GPT-3.5",
                        "abstract": "Large language models have shown impressive performance across a wide variety of tasks, including text summarization. In this paper, we show that this strong performance extends to opinion summarization. We explore several pipeline methods for applying GPT-3.5 to summarize a large collection of user reviews in a prompted fashion. To handle arbitrarily large numbers of user reviews, we explore recursive summarization as well as methods for selecting salient content to summarize through supervised clustering or extraction. On two datasets, an aspect-oriented summarization dataset of hotel reviews (SPACE) and a generic summarization dataset of Amazon and Yelp reviews (FewSum), we show that GPT-3.5 models achieve very strong performance in human evaluation. We argue that standard evaluation metrics do not reflect this, and introduce three new metrics targeting faithfulness, factuality, and genericity to contrast these different methods."
                    },
                    {
                        "title": "The Heuristic Core: Understanding Subnetwork Generalization in Pretrained Language Models",
                        "abstract": "Prior work has found that pretrained language models (LMs) fine-tuned with different random seeds can achieve similar in-domain performance but generalize differently on tests of syntactic generalization. In this work, we show that, even within a single model, we can find multiple subnetworks that perform similarly in-domain, but generalize vastly differently. To better understand these phenomena, we investigate if they can be understood in terms of \"competing subnetworks\": the model initially represents a variety of distinct algorithms, corresponding to different subnetworks, and generalization occurs when it ultimately converges to one. This explanation has been used to account for generalization in simple algorithmic tasks (\"grokking\"). Instead of finding competing subnetworks, we find that all subnetworks -- whether they generalize or not -- share a set of attention heads, which we refer to as the heuristic core. Further analysis suggests that these attention heads emerge early in training and compute shallow, non-generalizing features. The model learns to generalize by incorporating additional attention heads, which depend on the outputs of the \"heuristic\" heads to compute higher-level features. Overall, our results offer a more detailed picture of the mechanisms for syntactic generalization in pretrained LMs."
                    },
                    {
                        "title": "Benchmarking and Improving Text-to-SQL Generation under Ambiguity",
                        "abstract": "Research in Text-to-SQL conversion has been largely benchmarked against datasets where each text query corresponds to one correct SQL. However, natural language queries over real-life databases frequently involve significant ambiguity about the intended SQL due to overlapping schema names and multiple confusing relationship paths. To bridge this gap, we develop a novel benchmark called AmbiQT with over 3000 examples where each text is interpretable as two plausible SQLs due to lexical and/or structural ambiguity.   When faced with ambiguity, an ideal top-$k$ decoder should generate all valid interpretations for possible disambiguation by the user. We evaluate several Text-to-SQL systems and decoding algorithms, including those employing state-of-the-art LLMs, and find them to be far from this ideal. The primary reason is that the prevalent beam search algorithm and its variants, treat SQL queries as a string and produce unhelpful token-level diversity in the top-$k$.   We propose LogicalBeam, a new decoding algorithm that navigates the SQL logic space using a blend of plan-based template generation and constrained infilling. Counterfactually generated plans diversify templates while in-filling with a beam-search that branches solely on schema names provides value diversity. LogicalBeam is up to $2.5$ times more effective than state-of-the-art models at generating all candidate SQLs in the top-$k$ ranked outputs. It also enhances the top-$5$ Exact and Execution Match Accuracies on SPIDER and Kaggle DBQA."
                    },
                    {
                        "title": "Improving Language Understanding from Screenshots",
                        "abstract": "An emerging family of language models (LMs), capable of processing both text and images within a single visual view, has the promise to unlock complex tasks such as chart understanding and UI navigation. We refer to these models as screenshot language models. Despite their appeal, existing screenshot LMs substantially lag behind text-only models on language understanding tasks. To close this gap, we adopt a simplified setting where the model inputs are plain-text-rendered screenshots, and we focus on improving the text ability of screenshot LMs. We propose a novel Patch-and-Text Prediction (PTP) objective, which masks and recovers both image patches of screenshots and text within screenshots. We also conduct extensive ablation studies on masking rates and patch sizes, as well as designs for improving training stability. Our pre-trained model, while solely taking visual inputs, achieves comparable performance with BERT on 6 out of 8 GLUE tasks (within 2%) and improves up to 8% over prior work. Additionally, we extend PTP to train autoregressive screenshot LMs and demonstrate its effectiveness--our models can significantly reduce perplexity by utilizing the screenshot context. Together, we hope our findings can inspire future research on developing powerful screenshot LMs and extending their reach to broader applications."
                    },
                    {
                        "title": "Performance Bounds for LASSO under Multiplicative Noise: Applications to Pooled RT-PCR Testing",
                        "abstract": "Group testing is a technique which avoids individually testing $n$ samples for a rare disease and instead tests $n < p$ pools, where a pool consists of a mixture of small, equal portions of a subset of the $p$ samples. Group testing saves testing time and resources in many applications, including RT-PCR, with guarantees for the recovery of the status of the $p$ samples from results on $n$ pools. The noise in quantitative RT- PCR is inherently known to follow a multiplicative data-dependent model. In recent literature, the corresponding linear systems for inferring the health status of $p$ samples from results on $n$ pools have been solved using the Lasso estimator and its variants, which have been typically used in additive Gaussian noise settings. There is no existing literature which establishes performance bounds for Lasso for the multiplicative noise model associated with RT-PCR. After noting that a recent general technique, Hunt et al., works for Poisson inverse problems, we adapt it to handle sparse signal reconstruction from compressive measurements with multiplicative noise: we present high probability performance bounds and data-dependent weights for the Lasso and its weighted version. We also show numerical results on simulated pooled RT-PCR data to empirically validate our bounds."
                    },
                    {
                        "title": "Unintentional Unalignment: Likelihood Displacement in Direct Preference Optimization",
                        "abstract": "Direct Preference Optimization (DPO) and its variants are increasingly used for aligning language models with human preferences. Although these methods are designed to teach a model to generate preferred responses more frequently relative to dispreferred responses, prior work has observed that the likelihood of preferred responses often decreases during training. The current work sheds light on the causes and implications of this counter-intuitive phenomenon, which we term likelihood displacement. We demonstrate that likelihood displacement can be catastrophic, shifting probability mass from preferred responses to responses with an opposite meaning. As a simple example, training a model to prefer $\\texttt{No}$ over $\\texttt{Never}$ can sharply increase the probability of $\\texttt{Yes}$. Moreover, when aligning the model to refuse unsafe prompts, we show that such displacement can unintentionally lead to unalignment, by shifting probability mass from preferred refusal responses to harmful responses (e.g., reducing the refusal rate of Llama-3-8B-Instruct from 74.4% to 33.4%). We theoretically characterize that likelihood displacement is driven by preferences that induce similar embeddings, as measured by a centered hidden embedding similarity (CHES) score. Empirically, the CHES score enables identifying which training samples contribute most to likelihood displacement in a given dataset. Filtering out these samples effectively mitigated unintentional unalignment in our experiments. More broadly, our results highlight the importance of curating data with sufficiently distinct preferences, for which we believe the CHES score may prove valuable."
                    }
                ]
            },
            "b09aff15-e13e-4080-b565-00e836f730bd": {
                "pk": "b09aff15-e13e-4080-b565-00e836f730bd",
                "name": "Alexander Wettig",
                "collaborators": [
                    "Danqi Chen",
                    "Alexis Chevalier",
                    "Sanjeev Arora",
                    "Dan Friedman",
                    "Tianyu Gao",
                    "Zexuan Zhong",
                    "Carlos E. Jimenez",
                    "John Yang",
                    "Shunyu Yao",
                    "Ofir Press"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Language Models",
                    "Information Retrieval",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Adapting Language Models to Compress Contexts",
                        "abstract": "Transformer-based language models (LMs) are powerful and widely-applicable tools, but their usefulness is constrained by a finite context window and the expensive computational cost of processing long text documents. We propose to adapt pre-trained LMs into AutoCompressors. These language models are capable of compressing long contexts into compact summary vectors, which are then accessible to the model as soft prompts. Summary vectors are trained with an unsupervised objective, whereby long documents are processed in segments, and summary vectors from all previous segments are used in language modeling. We fine-tune OPT and Llama-2 models on sequences of up to 30,720 tokens and show that AutoCompressors can utilize long contexts to improve perplexity. We evaluate AutoCompressors on in-context learning by compressing task demonstrations and find that summary vectors are good substitutes for plain-text demonstrations, increasing accuracy while reducing inference costs. Finally, we explore the benefits of pre-computing summary vectors for large corpora by applying summary vectors to retrievalaugmented language modeling and a passage re-ranking task. Overall, AutoCompressors emerge as a simple and inexpensive solution to extend the context window of LMs while speeding up inference over long contexts."
                    },
                    {
                        "title": "QuRating: Selecting High-Quality Data for Training Language Models",
                        "abstract": "Selecting high-quality pre-training data is important for creating capable language models, but existing methods rely on simple heuristics. We introduce QuRating, a method for selecting pre-training data that can capture human intuitions about data quality. In this paper, we investigate four qualities - writing style, required expertise, facts & trivia, and educational value - and find that LLMs are able to discern these qualities, especially when making pairwise judgments of texts. We train a QuRater model to learn scalar ratings from pairwise judgments, and use it to annotate a 260B training corpus with quality ratings for each of the four criteria. In our experiments, we select 30B tokens according to the different quality ratings and train 1.3B-parameter language models on the selected data. We find that it is important to balance quality and diversity. When we sample using quality ratings as logits over documents, our models obtain lower perplexity and stronger in-context learning performance than baselines. Our best model is based on educational value and performs similarly to a model trained with uniform sampling for 50% more steps. Beyond data selection, we use the quality ratings to construct a training curriculum which improves performance without changing the training dataset. We extensively analyze the quality ratings and discuss their characteristics, biases, and wider implications."
                    },
                    {
                        "title": "A Kernel-Based View of Language Model Fine-Tuning",
                        "abstract": "It has become standard to solve NLP tasks by fine-tuning pre-trained language models (LMs), especially in low-data settings. There is minimal theoretical understanding of empirical success, e.g., why fine-tuning a model with $10^8$ or more parameters on a couple dozen training points does not result in overfitting. We investigate whether the Neural Tangent Kernel (NTK) - which originated as a model to study the gradient descent dynamics of infinitely wide networks with suitable random initialization - describes fine-tuning of pre-trained LMs. This study was inspired by the decent performance of NTK for computer vision tasks (Wei et al., 2022). We extend the NTK formalism to Adam and use Tensor Programs (Yang, 2020) to characterize conditions under which the NTK lens may describe fine-tuning updates to pre-trained language models. Extensive experiments on 14 NLP tasks validate our theory and show that formulating the downstream task as a masked word prediction problem through prompting often induces kernel-based dynamics during fine-tuning. Finally, we use this kernel view to propose an explanation for the success of parameter-efficient subspace-based fine-tuning methods."
                    },
                    {
                        "title": "Phrase Retrieval Learns Passage Retrieval, Too",
                        "abstract": "Dense retrieval methods have shown great promise over sparse retrieval methods in a range of NLP problems. Among them, dense phrase retrieval-the most fine-grained retrieval unit-is appealing because phrases can be directly used as the output for question answering and slot filling tasks. In this work, we follow the intuition that retrieving phrases naturally entails retrieving larger text blocks and study whether phrase retrieval can serve as the basis for coarse-level retrieval including passages and documents. We first observe that a dense phrase-retrieval system, without any retraining, already achieves better passage retrieval accuracy (+3-5% in top-5 accuracy) compared to passage retrievers, which also helps achieve superior end-to-end QA performance with fewer passages. Then, we provide an interpretation for why phrase-level supervision helps learn better fine-grained entailment compared to passage-level supervision, and also show that phrase retrieval can be improved to achieve competitive performance in document-retrieval tasks such as entity linking and knowledge-grounded dialogue. Finally, we demonstrate how phrase filtering and vector quantization can reduce the size of our index by 4-10x, making dense phrase retrieval a practical and versatile solution in multi-granularity retrieval."
                    },
                    {
                        "title": "Finding Dataset Shortcuts with Grammar Induction",
                        "abstract": "Many NLP datasets have been found to contain shortcuts: simple decision rules that achieve surprisingly high accuracy. However, it is difficult to discover shortcuts automatically. Prior work on automatic shortcut detection has focused on enumerating features like unigrams or bigrams, which can find only low-level shortcuts, or relied on post-hoc model interpretability methods like saliency maps, which reveal qualitative patterns without a clear statistical interpretation. In this work, we propose to use probabilistic grammars to characterize and discover shortcuts in NLP datasets. Specifically, we use a context-free grammar to model patterns in sentence classification datasets and use a synchronous context-free grammar to model datasets involving sentence pairs. The resulting grammars reveal interesting shortcut features in a number of datasets, including both simple and high-level features, and automatically identify groups of test examples on which conventional classifiers fail. Finally, we show that the features we discover can be used to generate diagnostic contrast examples and incorporated into standard robust optimization methods to improve worst-group accuracy."
                    },
                    {
                        "title": "Should You Mask 15% in Masked Language Modeling?",
                        "abstract": "Masked language models (MLMs) conventionally mask 15% of tokens due to the belief that more masking would leave insufficient context to learn good representations; this masking rate has been widely used, regardless of model sizes or masking strategies. In this work, we revisit this important choice of MLM pre-training. We first establish that 15% is not universally optimal, and larger models should adopt a higher masking rate. Specifically, we find that masking 40% outperforms 15% for BERT-large size models on GLUE and SQuAD. Interestingly, an extremely high masking rate of 80% can still preserve 95% fine-tuning performance and most of the accuracy in linguistic probing, challenging the conventional wisdom about the role of the masking rate. We then examine the interplay between masking rates and masking strategies and find that uniform masking requires a higher masking rate compared to sophisticated masking strategies such as span or PMI masking. Finally, we argue that increasing the masking rate has two distinct effects: it leads to more corruption, which makes the prediction task more difficult; it also enables more predictions, which benefits optimization. Using this framework, we revisit BERT's 80-10-10 corruption strategy. Together, our results contribute to a better understanding of MLM pre-training."
                    },
                    {
                        "title": "Learning Transformer Programs",
                        "abstract": "Recent research in mechanistic interpretability has attempted to reverse-engineer Transformer models by carefully inspecting network weights and activations. However, these approaches require considerable manual effort and still fall short of providing complete, faithful descriptions of the underlying algorithms. In this work, we introduce a procedure for training Transformers that are mechanistically interpretable by design. We build on RASP [Weiss et al., 2021], a programming language that can be compiled into Transformer weights. Instead of compiling human-written programs into Transformers, we design a modified Transformer that can be trained using gradient-based optimization and then automatically converted into a discrete, human-readable program. We refer to these models as Transformer Programs. To validate our approach, we learn Transformer Programs for a variety of problems, including an in-context learning task, a suite of algorithmic problems (e.g. sorting, recognizing Dyck languages), and NLP tasks including named entity recognition and text classification. The Transformer Programs can automatically find reasonable solutions, performing on par with standard Transformers of comparable size; and, more importantly, they are easy to interpret. To demonstrate these advantages, we convert Transformers into Python programs and use off-the-shelf code analysis tools to debug model errors and identify the \"circuits\" used to solve different sub-problems. We hope that Transformer Programs open a new path toward the goal of intrinsically interpretable machine learning."
                    },
                    {
                        "title": "Poisoning Retrieval Corpora by Injecting Adversarial Passages",
                        "abstract": "Dense retrievers have achieved state-of-the-art performance in various information retrieval tasks, but to what extent can they be safely deployed in real-world applications? In this work, we propose a novel attack for dense retrieval systems in which a malicious user generates a small number of adversarial passages by perturbing discrete tokens to maximize similarity with a provided set of training queries. When these adversarial passages are inserted into a large retrieval corpus, we show that this attack is highly effective in fooling these systems to retrieve them for queries that were not seen by the attacker. More surprisingly, these adversarial passages can directly generalize to out-of-domain queries and corpora with a high success attack rate -- for instance, we find that 50 generated passages optimized on Natural Questions can mislead >94% of questions posed in financial documents or online forums. We also benchmark and compare a range of state-of-the-art dense retrievers, both unsupervised and supervised. Although different systems exhibit varying levels of vulnerability, we show they can all be successfully attacked by injecting up to 500 passages, a small fraction compared to a retrieval corpus of millions of passages."
                    },
                    {
                        "title": "How to Train Long-Context Language Models (Effectively)",
                        "abstract": "We study continued training and supervised fine-tuning (SFT) of a language model (LM) to make effective use of long-context information. We first establish a reliable evaluation protocol to guide model development -- Instead of perplexity or simple needle-in-a-haystack (NIAH) tests, we use a broad set of long-context tasks, and we evaluate models after SFT with instruction data as this better reveals long-context abilities. Supported by our robust evaluations, we run thorough experiments to decide the data mix for continued pre-training, the instruction tuning dataset, and many other design choices. We find that (1) code repositories and books are excellent sources of long data, but it is crucial to combine them with high-quality short data; (2) training with a sequence length beyond the evaluation length boosts long-context performance; (3) for SFT, using only short instruction datasets yields strong performance on long-context tasks. Our final model, ProLong-8B, which is initialized from Llama-3 and trained on 40B tokens, demonstrates state-of-the-art long-context performance among similarly sized models at a length of 128K. ProLong outperforms Llama-3.18B-Instruct on the majority of long-context tasks despite having seen only 5% as many tokens during long-context training. Additionally, ProLong can effectively process up to 512K tokens, one of the longest context windows of publicly available LMs."
                    },
                    {
                        "title": "SWE-bench: Can Language Models Resolve Real-World GitHub Issues?",
                        "abstract": "Language models have outpaced our ability to evaluate them effectively, but for their future development it is essential to study the frontier of their capabilities. We find real-world software engineering to be a rich, sustainable, and challenging testbed for evaluating the next generation of language models. To this end, we introduce SWE-bench, an evaluation framework consisting of $2,294$ software engineering problems drawn from real GitHub issues and corresponding pull requests across $12$ popular Python repositories. Given a codebase along with a description of an issue to be resolved, a language model is tasked with editing the codebase to address the issue. Resolving issues in SWE-bench frequently requires understanding and coordinating changes across multiple functions, classes, and even files simultaneously, calling for models to interact with execution environments, process extremely long contexts and perform complex reasoning that goes far beyond traditional code generation tasks. Our evaluations show that both state-of-the-art proprietary models and our fine-tuned model SWE-Llama can resolve only the simplest issues. The best-performing model, Claude 2, is able to solve a mere $1.96$% of the issues. Advances on SWE-bench represent steps towards LMs that are more practical, intelligent, and autonomous."
                    },
                    {
                        "title": "SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering",
                        "abstract": "Language model (LM) agents are increasingly being used to automate complicated tasks in digital environments. Just as humans benefit from powerful software applications, such as integrated development environments, for complex tasks like software engineering, we posit that LM agents represent a new category of end users with their own needs and abilities, and would benefit from specially-built interfaces to the software they use. We investigate how interface design affects the performance of language model agents. As a result of this exploration, we introduce SWE-agent: a system that facilitates LM agents to autonomously use computers to solve software engineering tasks. SWE-agent's custom agent-computer interface (ACI) significantly enhances an agent's ability to create and edit code files, navigate entire repositories, and execute tests and other programs. We evaluate SWE-agent on SWE-bench and HumanEvalFix, achieving state-of-the-art performance on both with a pass@1 rate of 12.5% and 87.7%, respectively, far exceeding the previous state-of-the-art achieved with non-interactive LMs. Finally, we provide insight on how the design of the ACI can impact agents' behavior and performance."
                    },
                    {
                        "title": "OLMoE: Open Mixture-of-Experts Language Models",
                        "abstract": "We introduce OLMoE, a fully open, state-of-the-art language model leveraging sparse Mixture-of-Experts (MoE). OLMoE-1B-7B has 7 billion (B) parameters but uses only 1B per input token. We pretrain it on 5 trillion tokens and further adapt it to create OLMoE-1B-7B-Instruct. Our models outperform all available models with similar active parameters, even surpassing larger ones like Llama2-13B-Chat and DeepSeekMoE-16B. We present various experiments on MoE training, analyze routing in our model showing high specialization, and open-source all aspects of our work: model weights, training data, code, and logs."
                    },
                    {
                        "title": "Language Models as Science Tutors",
                        "abstract": "NLP has recently made exciting progress toward training language models (LMs) with strong scientific problem-solving skills. However, model development has not focused on real-life use-cases of LMs for science, including applications in education that require processing long scientific documents. To address this, we introduce TutorEval and TutorChat. TutorEval is a diverse question-answering benchmark consisting of questions about long chapters from STEM textbooks, written by experts. TutorEval helps measure real-life usability of LMs as scientific assistants, and it is the first benchmark combining long contexts, free-form generation, and multi-disciplinary scientific knowledge. Moreover, we show that fine-tuning base models with existing dialogue datasets leads to poor performance on TutorEval. Therefore, we create TutorChat, a dataset of 80,000 long synthetic dialogues about textbooks. We use TutorChat to fine-tune Llemma models with 7B and 34B parameters. These LM tutors specialized in math have a 32K-token context window, and they excel at TutorEval while performing strongly on GSM8K and MATH. Our datasets build on open-source materials, and we release our models, data, and evaluations."
                    }
                ]
            },
            "1d0050c8-9c69-4752-93ed-670318536f00": {
                "pk": "1d0050c8-9c69-4752-93ed-670318536f00",
                "name": "Dan Friedman",
                "collaborators": [
                    "Danqi Chen",
                    "Alexander Wettig",
                    "Ben Dodge",
                    "Abhishek Panigrahi",
                    "Adithya Bhaskar",
                    "Andrew Lampinen",
                    "Lucas Dixon",
                    "Asma Ghandeharioun",
                    "Zexuan Zhong",
                    "Adji Bousso Dieng"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Interpretability",
                    "Recommendation Systems"
                ],
                "publications": [
                    {
                        "title": "Single-dataset Experts for Multi-dataset Question Answering",
                        "abstract": "Many datasets have been created for training reading comprehension models, and a natural question is whether we can combine them to build models that (1) perform better on all of the training datasets and (2) generalize and transfer better to new datasets. Prior work has addressed this goal by training one network simultaneously on multiple datasets, which works well on average but is prone to over- or under-fitting different sub-distributions and might transfer worse compared to source models with more overlap with the target dataset. Our approach is to model multi-dataset question answering with a collection of single-dataset experts, by training a collection of lightweight, dataset-specific adapter modules (Houlsby et al., 2019) that share an underlying Transformer model. We find that these Multi-Adapter Dataset Experts (MADE) outperform all our baselines in terms of in-distribution accuracy, and simple methods based on parameter-averaging lead to better zero-shot generalization and few-shot transfer performance, offering a strong and versatile starting point for building new reading comprehension systems."
                    },
                    {
                        "title": "Finding Dataset Shortcuts with Grammar Induction",
                        "abstract": "Many NLP datasets have been found to contain shortcuts: simple decision rules that achieve surprisingly high accuracy. However, it is difficult to discover shortcuts automatically. Prior work on automatic shortcut detection has focused on enumerating features like unigrams or bigrams, which can find only low-level shortcuts, or relied on post-hoc model interpretability methods like saliency maps, which reveal qualitative patterns without a clear statistical interpretation. In this work, we propose to use probabilistic grammars to characterize and discover shortcuts in NLP datasets. Specifically, we use a context-free grammar to model patterns in sentence classification datasets and use a synchronous context-free grammar to model datasets involving sentence pairs. The resulting grammars reveal interesting shortcut features in a number of datasets, including both simple and high-level features, and automatically identify groups of test examples on which conventional classifiers fail. Finally, we show that the features we discover can be used to generate diagnostic contrast examples and incorporated into standard robust optimization methods to improve worst-group accuracy."
                    },
                    {
                        "title": "Learning Transformer Programs",
                        "abstract": "Recent research in mechanistic interpretability has attempted to reverse-engineer Transformer models by carefully inspecting network weights and activations. However, these approaches require considerable manual effort and still fall short of providing complete, faithful descriptions of the underlying algorithms. In this work, we introduce a procedure for training Transformers that are mechanistically interpretable by design. We build on RASP [Weiss et al., 2021], a programming language that can be compiled into Transformer weights. Instead of compiling human-written programs into Transformers, we design a modified Transformer that can be trained using gradient-based optimization and then automatically converted into a discrete, human-readable program. We refer to these models as Transformer Programs. To validate our approach, we learn Transformer Programs for a variety of problems, including an in-context learning task, a suite of algorithmic problems (e.g. sorting, recognizing Dyck languages), and NLP tasks including named entity recognition and text classification. The Transformer Programs can automatically find reasonable solutions, performing on par with standard Transformers of comparable size; and, more importantly, they are easy to interpret. To demonstrate these advantages, we convert Transformers into Python programs and use off-the-shelf code analysis tools to debug model errors and identify the \"circuits\" used to solve different sub-problems. We hope that Transformer Programs open a new path toward the goal of intrinsically interpretable machine learning."
                    },
                    {
                        "title": "Representing Rule-based Chatbots with Transformers",
                        "abstract": "Transformer-based chatbots can conduct fluent, natural-sounding conversations, but we have limited understanding of the mechanisms underlying their behavior. Prior work has taken a bottom-up approach to understanding Transformers by constructing Transformers for various synthetic and formal language tasks, such as regular expressions and Dyck languages. However, it is not obvious how to extend this approach to understand more naturalistic conversational agents. In this work, we take a step in this direction by constructing a Transformer that implements the ELIZA program, a classic, rule-based chatbot. ELIZA illustrates some of the distinctive challenges of the conversational setting, including both local pattern matching and long-term dialog state tracking. We build on constructions from prior work -- in particular, for simulating finite-state automata -- showing how simpler constructions can be composed and extended to give rise to more sophisticated behavior. Next, we train Transformers on a dataset of synthetically generated ELIZA conversations and investigate the mechanisms the models learn. Our analysis illustrates the kinds of mechanisms these models tend to prefer -- for example, models favor an induction head mechanism over a more precise, position based copying mechanism; and using intermediate generations to simulate recurrent data structures, like ELIZA's memory mechanisms. Overall, by drawing an explicit connection between neural chatbots and interpretable, symbolic mechanisms, our results offer a new setting for mechanistic analysis of conversational agents."
                    },
                    {
                        "title": "The Heuristic Core: Understanding Subnetwork Generalization in Pretrained Language Models",
                        "abstract": "Prior work has found that pretrained language models (LMs) fine-tuned with different random seeds can achieve similar in-domain performance but generalize differently on tests of syntactic generalization. In this work, we show that, even within a single model, we can find multiple subnetworks that perform similarly in-domain, but generalize vastly differently. To better understand these phenomena, we investigate if they can be understood in terms of \"competing subnetworks\": the model initially represents a variety of distinct algorithms, corresponding to different subnetworks, and generalization occurs when it ultimately converges to one. This explanation has been used to account for generalization in simple algorithmic tasks (\"grokking\"). Instead of finding competing subnetworks, we find that all subnetworks -- whether they generalize or not -- share a set of attention heads, which we refer to as the heuristic core. Further analysis suggests that these attention heads emerge early in training and compute shallow, non-generalizing features. The model learns to generalize by incorporating additional attention heads, which depend on the outputs of the \"heuristic\" heads to compute higher-level features. Overall, our results offer a more detailed picture of the mechanisms for syntactic generalization in pretrained LMs."
                    },
                    {
                        "title": "Interpretability Illusions in the Generalization of Simplified Models",
                        "abstract": "A common method to study deep learning systems is to use simplified model representations--for example, using singular value decomposition to visualize the model's hidden states in a lower dimensional space. This approach assumes that the results of these simplifications are faithful to the original model. Here, we illustrate an important caveat to this assumption: even if the simplified representations can accurately approximate the full model on the training set, they may fail to accurately capture the model's behavior out of distribution. We illustrate this by training Transformer models on controlled datasets with systematic generalization splits, including the Dyck balanced-parenthesis languages and a code completion task. We simplify these models using tools like dimensionality reduction and clustering, and then explicitly test how these simplified proxies match the behavior of the original model. We find consistent generalization gaps: cases in which the simplified proxies are more faithful to the original model on the in-distribution evaluations and less faithful on various tests of systematic generalization. This includes cases where the original model generalizes systematically but the simplified proxies fail, and cases where the simplified proxies generalize better. Together, our results raise questions about the extent to which mechanistic interpretations derived using tools like SVD can reliably predict what a model will do in novel situations."
                    },
                    {
                        "title": "Factual Probing Is [MASK]: Learning vs. Learning to Recall",
                        "abstract": "Petroni et al. (2019) demonstrated that it is possible to retrieve world facts from a pre-trained language model by expressing them as cloze-style prompts and interpret the model's prediction accuracy as a lower bound on the amount of factual information it encodes. Subsequent work has attempted to tighten the estimate by searching for better prompts, using a disjoint set of facts as training data. In this work, we make two complementary contributions to better understand these factual probing techniques. First, we propose OptiPrompt, a novel and efficient method which directly optimizes in continuous embedding space. We find this simple method is able to predict an additional 6.4% of facts in the LAMA benchmark. Second, we raise a more important question: Can we really interpret these probing results as a lower bound? Is it possible that these prompt-search methods learn from the training data too? We find, somewhat surprisingly, that the training data used by these methods contains certain regularities of the underlying fact distribution, and all the existing prompt methods, including ours, are able to exploit them for better fact prediction. We conduct a set of control experiments to disentangle \"learning\" from \"learning to recall\", providing a more detailed picture of what different prompts can reveal about pre-trained language models."
                    },
                    {
                        "title": "The Vendi Score: A Diversity Evaluation Metric for Machine Learning",
                        "abstract": "Diversity is an important criterion for many areas of machine learning (ML), including generative modeling and dataset curation. However, existing metrics for measuring diversity are often domain-specific and limited in flexibility. In this paper, we address the diversity evaluation problem by proposing the Vendi Score, which connects and extends ideas from ecology and quantum statistical mechanics to ML. The Vendi Score is defined as the exponential of the Shannon entropy of the eigenvalues of a similarity matrix. This matrix is induced by a user-defined similarity function applied to the sample to be evaluated for diversity. In taking a similarity function as input, the Vendi Score enables its user to specify any desired form of diversity. Importantly, unlike many existing metrics in ML, the Vendi Score does not require a reference dataset or distribution over samples or labels, it is therefore general and applicable to any generative model, decoding algorithm, and dataset from any domain where similarity can be defined. We showcase the Vendi Score on molecular generative modeling where we found it addresses shortcomings of the current diversity metric of choice in that domain. We also applied the Vendi Score to generative models of images and decoding algorithms of text where we found it confirms known results about diversity in those domains. Furthermore, we used the Vendi Score to measure mode collapse, a known shortcoming of generative adversarial networks (GANs). In particular, the Vendi Score revealed that even GANs that capture all the modes of a labeled dataset can be less diverse than the original dataset. Finally, the interpretability of the Vendi Score allowed us to diagnose several benchmark ML datasets for diversity, opening the door for diversity-informed data augmentation."
                    },
                    {
                        "title": "Syntax-aware Neural Semantic Role Labeling with Supertags",
                        "abstract": "We introduce a new syntax-aware model for dependency-based semantic role labeling that outperforms syntax-agnostic models for English and Spanish. We use a BiLSTM to tag the text with supertags extracted from dependency parses, and we feed these supertags, along with words and parts of speech, into a deep highway BiLSTM for semantic role labeling. Our model combines the strengths of earlier models that performed SRL on the basis of a full dependency parse with more recent models that use no syntactic information at all. Our local and non-ensemble model achieves state-of-the-art performance on the CoNLL 09 English and Spanish datasets. SRL models benefit from syntactic information, and we show that supertagging is a simple, powerful, and robust way to incorporate syntax into a neural SRL system."
                    },
                    {
                        "title": "Deep density networks and uncertainty in recommender systems",
                        "abstract": "Building robust online content recommendation systems requires learning complex interactions between user preferences and content features. The field has evolved rapidly in recent years from traditional multi-arm bandit and collaborative filtering techniques, with new methods employing Deep Learning models to capture non-linearities. Despite progress, the dynamic nature of online recommendations still poses great challenges, such as finding the delicate balance between exploration and exploitation. In this paper we show how uncertainty estimations can be incorporated by employing them in an optimistic exploitation/exploration strategy for more efficient exploration of new recommendations. We provide a novel hybrid deep neural network model, Deep Density Networks (DDN), which integrates content-based deep learning models with a collaborative scheme that is able to robustly model and estimate uncertainty. Finally, we present online and offline results after incorporating DNN into a real world content recommendation system that serves billions of recommendations per day, and show the benefit of using DDN in practice."
                    },
                    {
                        "title": "When a language model is optimized for reasoning, does it still show embers of autoregression? An analysis of OpenAI o1",
                        "abstract": "In \"Embers of Autoregression\" (McCoy et al., 2023), we showed that several large language models (LLMs) have some important limitations that are attributable to their origins in next-word prediction. Here we investigate whether these issues persist with o1, a new system from OpenAI that differs from previous LLMs in that it is optimized for reasoning. We find that o1 substantially outperforms previous LLMs in many cases, with particularly large improvements on rare variants of common tasks (e.g., forming acronyms from the second letter of each word in a list, rather than the first letter). Despite these quantitative improvements, however, o1 still displays the same qualitative trends that we observed in previous systems. Specifically, o1 -- like previous LLMs -- is sensitive to the probability of examples and tasks, performing better and requiring fewer \"thinking tokens\" in high-probability settings than in low-probability ones. These results show that optimizing a language model for reasoning can mitigate but might not fully overcome the language model's probability sensitivity."
                    }
                ]
            },
            "92148ae6-3604-4783-b1be-90701c5deeb0": {
                "pk": "92148ae6-3604-4783-b1be-90701c5deeb0",
                "name": "Danqi Chen",
                "collaborators": [
                    "Howard Chen",
                    "Zexuan Zhong",
                    "Jinhyuk Lee",
                    "Alexander Wettig",
                    "Dan Friedman",
                    "Jacqueline He",
                    "Karthik Narasimhan",
                    "Mengzhou Xia",
                    "Huihan Li",
                    "Tianyu Gao"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Question Answering",
                    "Machine Learning",
                    "Language Models"
                ],
                "publications": [
                    {
                        "title": "A Frustratingly Easy Approach for Entity and Relation Extraction",
                        "abstract": "End-to-end relation extraction aims to identify named entities and extract relations between them. Most recent work models these two subtasks jointly, either by casting them in one structured prediction framework, or performing multi-task learning through shared representations. In this work, we present a simple pipelined approach for entity and relation extraction, and establish the new state-of-the-art on standard benchmarks (ACE04, ACE05 and SciERC), obtaining a 1.7%-2.8% absolute improvement in relation F1 over previous joint models with the same pre-trained encoders. Our approach essentially builds on two independent encoders and merely uses the entity model to construct the input for the relation model. Through a series of careful examinations, we validate the importance of learning distinct contextual representations for entities and relations, fusing entity information early in the relation model, and incorporating global context. Finally, we also present an efficient approximation to our approach which requires only one pass of both entity and relation encoders at inference time, achieving an 8-16$\\times$ speedup with a slight reduction in accuracy."
                    },
                    {
                        "title": "Can Rationalization Improve Robustness?",
                        "abstract": "A growing line of work has investigated the development of neural NLP models that can produce rationales--subsets of input that can explain their model predictions. In this paper, we ask whether such rationale models can also provide robustness to adversarial attacks in addition to their interpretable nature. Since these models need to first generate rationales (\"rationalizer\") before making predictions (\"predictor\"), they have the potential to ignore noise or adversarially added text by simply masking it out of the generated rationale. To this end, we systematically generate various types of 'AddText' attacks for both token and sentence-level rationalization tasks, and perform an extensive empirical evaluation of state-of-the-art rationale models across five different tasks. Our experiments reveal that the rationale models show the promise to improve robustness, while they struggle in certain scenarios--when the rationalizer is sensitive to positional bias or lexical choices of attack text. Further, leveraging human rationale as supervision does not always translate to better performance. Our study is a first step towards exploring the interplay between interpretability and robustness in the rationalize-then-predict framework."
                    },
                    {
                        "title": "Non-Parametric Few-Shot Learning for Word Sense Disambiguation",
                        "abstract": "Word sense disambiguation (WSD) is a long-standing problem in natural language processing. One significant challenge in supervised all-words WSD is to classify among senses for a majority of words that lie in the long-tail distribution. For instance, 84% of the annotated words have less than 10 examples in the SemCor training data. This issue is more pronounced as the imbalance occurs in both word and sense distributions. In this work, we propose MetricWSD, a non-parametric few-shot learning approach to mitigate this data imbalance issue. By learning to compute distances among the senses of a given word through episodic training, MetricWSD transfers knowledge (a learned metric space) from high-frequency words to infrequent ones. MetricWSD constructs the training episodes tailored to word frequencies and explicitly addresses the problem of the skewed distribution, as opposed to mixing all the words trained with parametric models in previous work. Without resorting to any lexical resources, MetricWSD obtains strong performance against parametric alternatives, achieving a 75.1 F1 score on the unified WSD evaluation benchmark (Raganato et al., 2017b). Our analysis further validates that infrequent words and senses enjoy significant improvement."
                    },
                    {
                        "title": "Controllable Text Generation with Language Constraints",
                        "abstract": "We consider the task of text generation in language models with constraints specified in natural language. To this end, we first create a challenging benchmark Cognac that provides as input to the model a topic with example text, along with a constraint on text to be avoided. Unlike prior work, our benchmark contains knowledge-intensive constraints sourced from databases like Wordnet and Wikidata, which allows for straightforward evaluation while striking a balance between broad attribute-level and narrow lexical-level controls. We find that even state-of-the-art language models like GPT-3 fail often on this task, and propose a solution to leverage a language model's own internal knowledge to guide generation. Our method, called CognacGen, first queries the language model to generate guidance terms for a specified topic or constraint, and uses the guidance to modify the model's token generation probabilities. We propose three forms of guidance (binary verifier, top-k tokens, textual example), and employ prefix-tuning approaches to distill the guidance to tackle diverse natural language constraints. Through extensive empirical evaluations, we demonstrate that CognacGen can successfully generalize to unseen instructions and outperform competitive baselines in generating constraint conforming text."
                    },
                    {
                        "title": "A Thorough Examination of the CNN/Daily Mail Reading Comprehension Task",
                        "abstract": "Enabling a computer to understand a document so that it can answer comprehension questions is a central, yet unsolved goal of NLP. A key factor impeding its solution by machine learned systems is the limited availability of human-annotated data. Hermann et al. (2015) seek to solve this problem by creating over a million training examples by pairing CNN and Daily Mail news articles with their summarized bullet points, and show that a neural network can then be trained to give good performance on this task. In this paper, we conduct a thorough examination of this new reading comprehension task. Our primary aim is to understand what depth of language understanding is required to do well on this task. We approach this from one side by doing a careful hand-analysis of a small subset of the problems and from the other by showing that simple, carefully designed systems can obtain accuracies of 73.6% and 76.6% on these two datasets, exceeding current state-of-the-art results by 7-10% and approaching what we believe is the ceiling for performance on this task."
                    },
                    {
                        "title": "What In-Context Learning \"Learns\" In-Context: Disentangling Task Recognition and Task Learning",
                        "abstract": "Large language models (LLMs) exploit in-context learning (ICL) to solve tasks with only a few demonstrations, but its mechanisms are not yet well-understood. Some works suggest that LLMs only recall already learned concepts from pre-training, while others hint that ICL performs implicit learning over demonstrations. We characterize two ways through which ICL leverages demonstrations. Task recognition (TR) captures the extent to which LLMs can recognize a task through demonstrations -- even without ground-truth labels -- and apply their pre-trained priors, whereas task learning (TL) is the ability to capture new input-label mappings unseen in pre-training. Using a wide range of classification datasets and three LLM families (GPT-3, LLaMA and OPT), we design controlled experiments to disentangle the roles of TR and TL in ICL. We show that (1) models can achieve non-trivial performance with only TR, and TR does not further improve with larger models or more demonstrations; (2) LLMs acquire TL as the model scales, and TL's performance consistently improves with more demonstrations in context. Our findings unravel two different forces behind ICL and we advocate for discriminating them in future ICL research due to their distinct nature."
                    },
                    {
                        "title": "A Discrete Hard EM Approach for Weakly Supervised Question Answering",
                        "abstract": "Many question answering (QA) tasks only provide weak supervision for how the answer should be computed. For example, TriviaQA answers are entities that can be mentioned multiple times in supporting documents, while DROP answers can be computed by deriving many different equations from numbers in the reference text. In this paper, we show it is possible to convert such tasks into discrete latent variable learning problems with a precomputed, task-specific set of possible \"solutions\" (e.g. different mentions or equations) that contains one correct option. We then develop a hard EM learning scheme that computes gradients relative to the most likely solution at each update. Despite its simplicity, we show that this approach significantly outperforms previous methods on six QA tasks, including absolute gains of 2--10%, and achieves the state-of-the-art on five of them. Using hard updates instead of maximizing marginal likelihood is key to these results as it encourages the model to find the one correct answer, which we show through detailed qualitative analysis."
                    },
                    {
                        "title": "Knowledge Guided Text Retrieval and Reading for Open Domain Question Answering",
                        "abstract": "We introduce an approach for open-domain question answering (QA) that retrieves and reads a passage graph, where vertices are passages of text and edges represent relationships that are derived from an external knowledge base or co-occurrence in the same article. Our goals are to boost coverage by using knowledge-guided retrieval to find more relevant passages than text-matching methods, and to improve accuracy by allowing for better knowledge-guided fusion of information across related passages. Our graph retrieval method expands a set of seed keyword-retrieved passages by traversing the graph structure of the knowledge base. Our reader extends a BERT-based architecture and updates passage representations by propagating information from related passages and their relations, instead of reading each passage in isolation. Experiments on three open-domain QA datasets, WebQuestions, Natural Questions and TriviaQA, show improved performance over non-graph baselines by 2-11% absolute. Our approach also matches or exceeds the state-of-the-art in every case, without using an expensive end-to-end training regime."
                    },
                    {
                        "title": "Phrase Retrieval Learns Passage Retrieval, Too",
                        "abstract": "Dense retrieval methods have shown great promise over sparse retrieval methods in a range of NLP problems. Among them, dense phrase retrieval-the most fine-grained retrieval unit-is appealing because phrases can be directly used as the output for question answering and slot filling tasks. In this work, we follow the intuition that retrieving phrases naturally entails retrieving larger text blocks and study whether phrase retrieval can serve as the basis for coarse-level retrieval including passages and documents. We first observe that a dense phrase-retrieval system, without any retraining, already achieves better passage retrieval accuracy (+3-5% in top-5 accuracy) compared to passage retrievers, which also helps achieve superior end-to-end QA performance with fewer passages. Then, we provide an interpretation for why phrase-level supervision helps learn better fine-grained entailment compared to passage-level supervision, and also show that phrase retrieval can be improved to achieve competitive performance in document-retrieval tasks such as entity linking and knowledge-grounded dialogue. Finally, we demonstrate how phrase filtering and vector quantization can reduce the size of our index by 4-10x, making dense phrase retrieval a practical and versatile solution in multi-granularity retrieval."
                    },
                    {
                        "title": "Simple Entity-Centric Questions Challenge Dense Retrievers",
                        "abstract": "Open-domain question answering has exploded in popularity recently due to the success of dense retrieval models, which have surpassed sparse models using only a few supervised training examples. However, in this paper, we demonstrate current dense models are not yet the holy grail of retrieval. We first construct EntityQuestions, a set of simple, entity-rich questions based on facts from Wikidata (e.g., \"Where was Arve Furset born?\"), and observe that dense retrievers drastically underperform sparse methods. We investigate this issue and uncover that dense retrievers can only generalize to common entities unless the question pattern is explicitly observed during training. We discuss two simple solutions towards addressing this critical problem. First, we demonstrate that data augmentation is unable to fix the generalization problem. Second, we argue a more robust passage encoder helps facilitate better question adaptation using specialized question encoders. We hope our work can shed light on the challenges in creating a robust, universal dense retriever that works well across different input distributions."
                    },
                    {
                        "title": "Single-dataset Experts for Multi-dataset Question Answering",
                        "abstract": "Many datasets have been created for training reading comprehension models, and a natural question is whether we can combine them to build models that (1) perform better on all of the training datasets and (2) generalize and transfer better to new datasets. Prior work has addressed this goal by training one network simultaneously on multiple datasets, which works well on average but is prone to over- or under-fitting different sub-distributions and might transfer worse compared to source models with more overlap with the target dataset. Our approach is to model multi-dataset question answering with a collection of single-dataset experts, by training a collection of lightweight, dataset-specific adapter modules (Houlsby et al., 2019) that share an underlying Transformer model. We find that these Multi-Adapter Dataset Experts (MADE) outperform all our baselines in terms of in-distribution accuracy, and simple methods based on parameter-averaging lead to better zero-shot generalization and few-shot transfer performance, offering a strong and versatile starting point for building new reading comprehension systems."
                    },
                    {
                        "title": "Ditch the Gold Standard: Re-evaluating Conversational Question Answering",
                        "abstract": "Conversational question answering aims to provide natural-language answers to users in information-seeking conversations. Existing conversational QA benchmarks compare models with pre-collected human-human conversations, using ground-truth answers provided in conversational history. It remains unclear whether we can rely on this static evaluation for model development and whether current systems can well generalize to real-world human-machine conversations. In this work, we conduct the first large-scale human evaluation of state-of-the-art conversational QA systems, where human evaluators converse with models and judge the correctness of their answers. We find that the distribution of human machine conversations differs drastically from that of human-human conversations, and there is a disagreement between human and gold-history evaluation in terms of model ranking. We further investigate how to improve automatic evaluations, and propose a question rewriting mechanism based on predicted history, which better correlates with human judgments. Finally, we analyze the impact of various modeling strategies and discuss future directions towards building better conversational question answering systems."
                    },
                    {
                        "title": "Training Language Models with Memory Augmentation",
                        "abstract": "Recent work has improved language models (LMs) remarkably by equipping them with a non-parametric memory component. However, most existing approaches only introduce mem-ories at testing time or represent them using a separately trained encoder, resulting in suboptimal training of the language model. In this work, we present TRIME, a novel yet simple training approach designed for training LMs with memory augmentation. Our approach uses a training objective that directly takes in-batch examples as accessible memory. We also present new methods for memory construction and data batching, which are used for adapting to different sets of memories--local, long-term, and external memory--at testing time. We evaluate TRIME on multiple language modeling and machine translation benchmarks and show that it is able to achieve significant improvements across all the settings. Concretely, TRIME reduces the perplexity from 18.70 to 15.37 on WIKITEXT-103, by effectively leveraging a large memory set from the training corpus. Compared to standard LM training, TRIME adds negligible computational overhead and is compatible with different neural architectures, making it a versatile solution for training memory-augmented LMs."
                    },
                    {
                        "title": "Finding Dataset Shortcuts with Grammar Induction",
                        "abstract": "Many NLP datasets have been found to contain shortcuts: simple decision rules that achieve surprisingly high accuracy. However, it is difficult to discover shortcuts automatically. Prior work on automatic shortcut detection has focused on enumerating features like unigrams or bigrams, which can find only low-level shortcuts, or relied on post-hoc model interpretability methods like saliency maps, which reveal qualitative patterns without a clear statistical interpretation. In this work, we propose to use probabilistic grammars to characterize and discover shortcuts in NLP datasets. Specifically, we use a context-free grammar to model patterns in sentence classification datasets and use a synchronous context-free grammar to model datasets involving sentence pairs. The resulting grammars reveal interesting shortcut features in a number of datasets, including both simple and high-level features, and automatically identify groups of test examples on which conventional classifiers fail. Finally, we show that the features we discover can be used to generate diagnostic contrast examples and incorporated into standard robust optimization methods to improve worst-group accuracy."
                    },
                    {
                        "title": "MABEL: Attenuating Gender Bias using Textual Entailment Data",
                        "abstract": "Pre-trained language models encode undesirable social biases, which are further exacerbated in downstream use. To this end, we propose MABEL (a Method for Attenuating Gender Bias using Entailment Labels), an intermediate pre-training approach for mitigating gender bias in contextualized representations. Key to our approach is the use of a contrastive learning objective on counterfactually augmented, gender-balanced entailment pairs from natural language inference (NLI) datasets. We also introduce an alignment regularizer that pulls identical entailment pairs along opposite gender directions closer. We extensively evaluate our approach on intrinsic and extrinsic metrics, and show that MABEL outperforms previous task-agnostic debiasing approaches in terms of fairness. It also preserves task performance after fine-tuning on downstream tasks. Together, these findings demonstrate the suitability of NLI data as an effective means of bias mitigation, as opposed to only using unlabeled sentences in the literature. Finally, we identify that existing approaches often use evaluation settings that are insufficient or inconsistent. We make an effort to reproduce and compare previous methods, and call for unifying the evaluation settings across gender debiasing methods for better future comparison."
                    },
                    {
                        "title": "Adapting Language Models to Compress Contexts",
                        "abstract": "Transformer-based language models (LMs) are powerful and widely-applicable tools, but their usefulness is constrained by a finite context window and the expensive computational cost of processing long text documents. We propose to adapt pre-trained LMs into AutoCompressors. These language models are capable of compressing long contexts into compact summary vectors, which are then accessible to the model as soft prompts. Summary vectors are trained with an unsupervised objective, whereby long documents are processed in segments, and summary vectors from all previous segments are used in language modeling. We fine-tune OPT and Llama-2 models on sequences of up to 30,720 tokens and show that AutoCompressors can utilize long contexts to improve perplexity. We evaluate AutoCompressors on in-context learning by compressing task demonstrations and find that summary vectors are good substitutes for plain-text demonstrations, increasing accuracy while reducing inference costs. Finally, we explore the benefits of pre-computing summary vectors for large corpora by applying summary vectors to retrievalaugmented language modeling and a passage re-ranking task. Overall, AutoCompressors emerge as a simple and inexpensive solution to extend the context window of LMs while speeding up inference over long contexts."
                    },
                    {
                        "title": "The Heuristic Core: Understanding Subnetwork Generalization in Pretrained Language Models",
                        "abstract": "Prior work has found that pretrained language models (LMs) fine-tuned with different random seeds can achieve similar in-domain performance but generalize differently on tests of syntactic generalization. In this work, we show that, even within a single model, we can find multiple subnetworks that perform similarly in-domain, but generalize vastly differently. To better understand these phenomena, we investigate if they can be understood in terms of \"competing subnetworks\": the model initially represents a variety of distinct algorithms, corresponding to different subnetworks, and generalization occurs when it ultimately converges to one. This explanation has been used to account for generalization in simple algorithmic tasks (\"grokking\"). Instead of finding competing subnetworks, we find that all subnetworks -- whether they generalize or not -- share a set of attention heads, which we refer to as the heuristic core. Further analysis suggests that these attention heads emerge early in training and compute shallow, non-generalizing features. The model learns to generalize by incorporating additional attention heads, which depend on the outputs of the \"heuristic\" heads to compute higher-level features. Overall, our results offer a more detailed picture of the mechanisms for syntactic generalization in pretrained LMs."
                    },
                    {
                        "title": "Learning Dense Representations of Phrases at Scale",
                        "abstract": "Open-domain question answering can be reformulated as a phrase retrieval problem, without the need for processing documents on-demand during inference (Seo et al., 2019). However, current phrase retrieval models heavily depend on sparse representations and still underperform retriever-reader approaches. In this work, we show for the first time that we can learn dense representations of phrases alone that achieve much stronger performance in open-domain QA. We present an effective method to learn phrase representations from the supervision of reading comprehension tasks, coupled with novel negative sampling methods. We also propose a query-side fine-tuning strategy, which can support transfer learning and reduce the discrepancy between training and inference. On five popular open-domain QA datasets, our model DensePhrases improves over previous phrase retrieval models by 15%-25% absolute accuracy and matches the performance of state-of-the-art retriever-reader models. Our model is easy to parallelize due to pure dense representations and processes more than 10 questions per second on CPUs. Finally, we directly use our pre-indexed dense phrase representations for two slot filling tasks, showing the promise of utilizing DensePhrases as a dense knowledge base for downstream tasks."
                    }
                ]
            }
        }
    },
    "2406.08316": {
        "paper_data": {
            "title": "Is Programming by Example solved by LLMs?",
            "url": "http://arxiv.org/abs/2406.08316v2",
            "arxiv_id": "2406.08316",
            "authors": [
                "Wen-Ding Li",
                "Kevin Ellis"
            ],
            "abstract": "Programming-by-Examples (PBE) aims to generate an algorithm from input-output examples. Such systems are practically and theoretically important: from an end-user perspective, they are deployed to millions of people, and from an AI perspective, PBE corresponds to a very general form of few-shot inductive inference. Given the success of Large Language Models (LLMs) in code-generation tasks, we investigate here the extent to which LLMs can be said to have `solved' PBE. We experiment on classic domains such as lists and strings, and an uncommon graphics programming domain not well represented in typical pretraining data. We find that pretrained models are not effective at PBE, but that they can be fine-tuned for much higher performance, provided the test problems are in-distribution. We analyze empirically what causes these models to succeed and fail, and take steps toward understanding how to achieve better out-of-distribution generalization. Collectively these results suggest that LLMs make strong progress toward solving the typical suite of PBE tasks, potentially increasing the flexibility and applicability of PBE systems, while also identifying ways in which LLMs still fall short.",
            "introduction": " Introduction Programming-by-Example (PBE) systems solve a challenging task: Given input-output examples of a hidden algorithm, they seek to construct the source code of the underlying function [ 1,2]. PBE is deployed to millions of users [ 3,4,5,6], lies near the heart of core AI challenges [ 7,8,9,10], and is a qualitatively different problem from the bulk of recent work on LLM code generation, because rather than generate source code from natural language [ 11], PBE is instead fundamentally about few-shot inductive inference: Given a handful of examples, inferring the program that will generalize to new inputs, or which captures the \u2018true\u2019 latent regularity, without relying on natural-language guidance. We investigate here the extent to which large language models pretrained on source code can solve PBE. If they can, this unlocks the ability to do PBE in general-purpose Turing complete languages like Python, unlike the restricted domain-specific languages which have so far dominated PBE [ 4,12,13,14, i.a.], thereby increasing the scope and power of this paradigm. If LLMs cannot perform PBE, then this highlights a deficit of inductive reasoning and problem solving, and suggests LLMs lean too heavily on natural language cues to generate code. We find that pretrained and instruction-tuned models serve as poor PBE systems, a finding also supported by recent work [ 15,12]. But our investigation further finds that LLMs can be fine-tuned for significantly higher performance, provided they are not asked to generalize far beyond the fine-tuning data. To address this failure of generalization we give an algorithm for taking a small unlabeled dataset of problems and adapting the LLM to it, which we find narrows this domain gap. The resulting recipe allows PBE over Turing-complete languages across three qualitatively different domains (Fig. 1): algorithms on vectors of numbers, string manipulation macros, and graphics programs in LOGO/Turtle. In every case, our final model is at least as effective as custom symbolic search algorithms operating over domain-specific languages, and surpasses powerful closed-source Preprint. Under review.arXiv:2406.08316v2  [cs.CL]  13 Jun 2024models such as GPT4 [ 16]. We also find that the resulting system can cover a broader scope of problems than classic symbolic methods can, within their domains, synthesize programs perfectly (i.e. always find a program that fits the training input-outputs). In contrast, neural networks can struggle to generalize beyond the training distribution. We therefore consider train/test splits that force the model to generalize beyond the distribution of its training data (beyond Dseed). On text editing, we seed with SyGuS problems, and perform out- of-distribution testing on PROSE problems (PROSE is much harder than SyGuS). On list functions, we seed with problems from Rule et al. 2024 and test on Shi et al. 2023 (the Shi dataset contains unusual combinators, such as Scan ). On LOGO, we seed with short programs ( \u226412lines of code), and test on long programs ( >12lines of code). Using these splits we also measure the ability of the adaptation method in Sec. 3 to improve out-of-distribution generalization.3 Fig. 8 shows that there is nontrivial degradation when testing out of distribution. For example, a 7B model seeded with PROSE problems and tested on a different subset of PROSE has an accuracy of 76% (Fig. 3b), but this degrades to 59% when seeded with SyGuS problems, which follow a different distribution and",
            "references": [
                {
                    "title": "Code Repair with LLMs gives an Exploration-Exploitation Tradeoff",
                    "abstract": "Iteratively improving and repairing source code with large language models (LLMs), known as refinement, has emerged as a popular way of generating programs that would be too complex to construct in one shot. Given a bank of test cases, together with a candidate program, an LLM can improve that program by being prompted with failed test cases. But it remains an open question how to best iteratively refine code, with prior work employing simple greedy or breadth-first strategies. We show here that refinement exposes an explore-exploit tradeoff: exploit by refining the program that passes the most test cases, or explore by refining a lesser considered program. We frame this as an arm-acquiring bandit problem, which we solve with Thompson Sampling. The resulting LLM-based program synthesis algorithm is broadly applicable: Across loop invariant synthesis, visual reasoning puzzles, and competition programming problems, we find that our new method can solve more problems using fewer language model calls."
                },
                {
                    "title": "DataDreamer: A Tool for Synthetic Data Generation and Reproducible LLM Workflows",
                    "abstract": "Large language models (LLMs) have become a dominant and important tool for NLP researchers in a wide range of tasks. Today, many researchers use LLMs in synthetic data generation, task evaluation, fine-tuning, distillation, and other model-in-the-loop research workflows. However, challenges arise when using these models that stem from their scale, their closed source nature, and the lack of standardized tooling for these new and emerging workflows. The rapid rise to prominence of these models and these unique challenges has had immediate adverse impacts on open science and on the reproducibility of work that uses them. In this paper, we introduce DataDreamer, an open source Python library that allows researchers to write simple code to implement powerful LLM workflows. DataDreamer also helps researchers adhere to best practices that we propose to encourage open science and reproducibility. The library and documentation are available at https://github.com/datadreamer-dev/DataDreamer ."
                },
                {
                    "title": "Doing Experiments and Revising Rules with Natural Language and Probabilistic Reasoning",
                    "abstract": "We build a computational model of how humans actively infer hidden rules by doing experiments. The basic principles behind the model is that, even if the rule is deterministic, the learner considers a broader space of fuzzy probabilistic rules, which it represents in natural language, and updates its hypotheses online after each experiment according to approximately Bayesian principles. In the same framework we also model experiment design according to information-theoretic criteria. We find that the combination of these three principles\u2014explicit hypotheses, probabilistic rules, and online updates\u2014can explain human performance on a Zendo-style task, and that removing any of these components leaves the model unable to account for the data."
                },
                {
                    "title": "ReGAL: Refactoring Programs to Discover Generalizable Abstractions",
                    "abstract": "While large language models (LLMs) are increasingly being used for program synthesis, they lack the global view needed to develop useful abstractions; they generally predict programs one at a time, often repeating the same functionality. Generating redundant code from scratch is both inefficient and error-prone. To address this, we propose Refactoring for Generalizable Abstraction Learning (ReGAL), a gradient-free method for learning a library of reusable functions via code refactorization, i.e., restructuring code without changing its execution output. ReGAL learns from a small set of existing programs, iteratively verifying and refining its abstractions via execution. We find that the shared function libraries discovered by ReGAL make programs easier to predict across diverse domains. On five datasets -- LOGO graphics generation, Date reasoning, TextCraft (a Minecraft-based text-game) MATH, and TabMWP -- both open-source and proprietary LLMs improve in accuracy when predicting programs with ReGAL functions. For CodeLlama-13B, ReGAL results in absolute accuracy increases of 11.5% on LOGO, 26.1% on date understanding, and 8.1% on TextCraft, outperforming GPT-3.5 in two of three domains. Our analysis reveals ReGAL's abstractions encapsulate frequently-used subroutines as well as environment dynamics."
                },
                {
                    "title": "DeepSeek-Coder: When the Large Language Model Meets Programming - The Rise of Code Intelligence",
                    "abstract": "The rapid development of large language models has revolutionized code intelligence in software development. However, the predominance of closed-source models has restricted extensive research and development. To address this, we introduce the DeepSeek-Coder series, a range of open-source code models with sizes from 1.3B to 33B, trained from scratch on 2 trillion tokens. These models are pre-trained on a high-quality project-level code corpus and employ a fill-in-the-blank task with a 16K window to enhance code generation and infilling. Our extensive evaluations demonstrate that DeepSeek-Coder not only achieves state-of-the-art performance among open-source code models across multiple benchmarks but also surpasses existing closed-source models like Codex and GPT-3.5. Furthermore, DeepSeek-Coder models are under a permissive license that allows for both research and unrestricted commercial use."
                },
                {
                    "title": "Phenomenal Yet Puzzling: Testing Inductive Reasoning Capabilities of Language Models with Hypothesis Refinement",
                    "abstract": "The ability to derive underlying principles from a handful of observations and then generalize to novel situations -- known as inductive reasoning -- is central to human intelligence. Prior work suggests that language models (LMs) often fall short on inductive reasoning, despite achieving impressive success on research benchmarks. In this work, we conduct a systematic study of the inductive reasoning capabilities of LMs through iterative hypothesis refinement, a technique that more closely mirrors the human inductive process than standard input-output prompting. Iterative hypothesis refinement employs a three-step process: proposing, selecting, and refining hypotheses in the form of textual rules. By examining the intermediate rules, we observe that LMs are phenomenal hypothesis proposers (i.e., generating candidate rules), and when coupled with a (task-specific) symbolic interpreter that is able to systematically filter the proposed set of rules, this hybrid approach achieves strong results across inductive reasoning benchmarks that require inducing causal relations, language-like instructions, and symbolic concepts. However, they also behave as puzzling inductive reasoners, showing notable performance gaps between rule induction (i.e., identifying plausible rules) and rule application (i.e., applying proposed rules to instances), suggesting that LMs are proposing hypotheses without being able to actually apply the rules. Through empirical and human analyses, we further reveal several discrepancies between the inductive reasoning processes of LMs and humans, shedding light on both the potentials and limitations of using LMs in inductive reasoning tasks."
                },
                {
                    "title": "Hypothesis Search: Inductive Reasoning with Language Models",
                    "abstract": "Inductive reasoning is a core problem-solving capacity: humans can identify underlying principles from a few examples, which robustly generalize to novel scenarios. Recent work evaluates large language models (LLMs) on inductive reasoning tasks by directly prompting them yielding\"in context learning.\"This works well for straightforward inductive tasks but performs poorly on complex tasks such as the Abstraction and Reasoning Corpus (ARC). In this work, we propose to improve the inductive reasoning ability of LLMs by generating explicit hypotheses at multiple levels of abstraction: we prompt the LLM to propose multiple abstract hypotheses about the problem, in natural language, then implement the natural language hypotheses as concrete Python programs. These programs can be verified by running on observed examples and generalized to novel inputs. To reduce the hypothesis search space, we explore steps to filter the set of hypotheses to implement: we either ask the LLM to summarize them into a smaller set of hypotheses or ask human annotators to select a subset. We verify our pipeline's effectiveness on the ARC visual inductive reasoning benchmark, its variant 1D-ARC, string transformation dataset SyGuS, and list transformation dataset List Functions. On a random 100-problem subset of ARC, our automated pipeline using LLM summaries achieves 30% accuracy, outperforming the direct prompting baseline (accuracy of 17%). With the minimal human input of selecting from LLM-generated candidates, performance is boosted to 33%. Our ablations show that both abstract hypothesis generation and concrete program representations benefit LLMs on inductive reasoning tasks."
                },
                {
                    "title": "Is Self-Repair a Silver Bullet for Code Generation?",
                    "abstract": "Large language models have shown remarkable aptitude in code generation, but still struggle to perform complex tasks. Self-repair -- in which the model debugs and repairs its own code -- has recently become a popular way to boost performance in these settings. However, despite its increasing popularity, existing studies of self-repair have been limited in scope; in many settings, its efficacy thus remains poorly understood. In this paper, we analyze Code Llama, GPT-3.5 and GPT-4's ability to perform self-repair on problems taken from HumanEval and APPS. We find that when the cost of carrying out repair is taken into account, performance gains are often modest, vary a lot between subsets of the data, and are sometimes not present at all. We hypothesize that this is because self-repair is bottlenecked by the model's ability to provide feedback on its own code; using a stronger model to artificially boost the quality of the feedback, we observe substantially larger performance gains. Similarly, a small-scale study in which we provide GPT-4 with feedback from human participants suggests that even for the strongest models, self-repair still lags far behind what can be achieved with human-level debugging."
                },
                {
                    "title": "Human-like Few-Shot Learning via Bayesian Reasoning over Natural Language",
                    "abstract": "A core tension in models of concept learning is that the model must carefully balance the tractability of inference against the expressivity of the hypothesis class. Humans, however, can efficiently learn a broad range of concepts. We introduce a model of inductive learning that seeks to be human-like in that sense. It implements a Bayesian reasoning process where a language model first proposes candidate hypotheses expressed in natural language, which are then re-weighed by a prior and a likelihood. By estimating the prior from human data, we can predict human judgments on learning problems involving numbers and sets, spanning concepts that are generative, discriminative, propositional, and higher-order."
                },
                {
                    "title": "LambdaBeam: Neural Program Search with Higher-Order Functions and Lambdas",
                    "abstract": "Search is an important technique in program synthesis that allows for adaptive strategies such as focusing on particular search directions based on execution results. Several prior works have demonstrated that neural models are effective at guiding program synthesis searches. However, a common drawback of those approaches is the inability to handle iterative loops, higher-order functions, or lambda functions, thus limiting prior neural searches from synthesizing longer and more general programs. We address this gap by designing a search algorithm called LambdaBeam that can construct arbitrary lambda functions that compose operations within a given DSL. We create semantic vector representations of the execution behavior of the lambda functions and train a neural policy network to choose which lambdas to construct during search, and pass them as arguments to higher-order functions to perform looping computations. Our experiments show that LambdaBeam outperforms neural, symbolic, and LLM-based techniques in an integer list manipulation domain."
                },
                {
                    "title": "Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generation",
                    "abstract": "Program synthesis has been long studied with recent approaches focused on directly using the power of Large Language Models (LLMs) to generate code. Programming benchmarks, with curated synthesis problems and test-cases, are used to measure the performance of various LLMs on code synthesis. However, these test-cases can be limited in both quantity and quality for fully assessing the functional correctness of the generated code. Such limitation in the existing benchmarks begs the following question: In the era of LLMs, is the code generated really correct? To answer this, we propose EvalPlus -- a code synthesis evaluation framework to rigorously benchmark the functional correctness of LLM-synthesized code. EvalPlus augments a given evaluation dataset with large amounts of test-cases newly produced by an automatic test input generator, powered by both LLM- and mutation-based strategies. While EvalPlus is general, we extend the test-cases of the popular HumanEval benchmark by 80x to build HumanEval+. Our extensive evaluation across 26 popular LLMs (e.g., GPT-4 and ChatGPT) demonstrates that HumanEval+ is able to catch significant amounts of previously undetected wrong code synthesized by LLMs, reducing the pass@k by up-to 19.3-28.9%. We also surprisingly found that test insufficiency can lead to mis-ranking. For example, both WizardCoder-CodeLlama and Phind-CodeLlama now outperform ChatGPT on HumanEval+, while none of them could on HumanEval. Our work not only indicates that prior popular code synthesis evaluation results do not accurately reflect the true performance of LLMs for code synthesis, but also opens up a new direction to improve such programming benchmarks through automated testing. We have open-sourced our tools, enhanced datasets as well as all LLM-generated code at https://github.com/evalplus/evalplus to facilitate and accelerate future LLM-for-code research."
                },
                {
                    "title": "Teaching Large Language Models to Self-Debug",
                    "abstract": "Large language models (LLMs) have achieved impressive performance on code generation. However, for complex programming tasks, generating the correct solution in one go becomes challenging, thus some prior works have designed program repair approaches to improve code generation performance. In this work, we propose Self-Debugging, which teaches a large language model to debug its predicted program via few-shot demonstrations. In particular, we demonstrate that Self-Debugging can teach the large language model to perform rubber duck debugging; i.e., without any human feedback on the code correctness or error messages, the model is able to identify its mistakes by investigating the execution results and explaining the generated code in natural language. Self-Debugging achieves the state-of-the-art performance on several code generation benchmarks, including the Spider dataset for text-to-SQL generation, TransCoder for C++-to-Python translation, and MBPP for text-to-Python generation. On the Spider benchmark where there are no unit tests to verify the correctness of predictions, Self-Debugging with code explanation consistently improves the baseline by 2-3%, and improves the prediction accuracy on problems of the hardest level by 9%. On TransCoder and MBPP where unit tests are available, Self-Debugging improves the baseline accuracy by up to 12%. Meanwhile, by leveraging feedback messages and reusing failed predictions, Self-Debugging notably improves sample efficiency, and can match or outperform baseline models that generate more than 10x candidate programs."
                },
                {
                    "title": "Self-Refine: Iterative Refinement with Self-Feedback",
                    "abstract": "Like humans, large language models (LLMs) do not always generate the best output on their first try. Motivated by how humans refine their written text, we introduce Self-Refine, an approach for improving initial outputs from LLMs through iterative feedback and refinement. The main idea is to generate an initial output using an LLMs; then, the same LLMs provides feedback for its output and uses it to refine itself, iteratively. Self-Refine does not require any supervised training data, additional training, or reinforcement learning, and instead uses a single LLM as the generator, refiner, and feedback provider. We evaluate Self-Refine across 7 diverse tasks, ranging from dialog response generation to mathematical reasoning, using state-of-the-art (GPT-3.5, ChatGPT, and GPT-4) LLMs. Across all evaluated tasks, outputs generated with Self-Refine are preferred by humans and automatic metrics over those generated with the same LLM using conventional one-step generation, improving by ~20% absolute on average in task performance. Our work demonstrates that even state-of-the-art LLMs like GPT-4 can be further improved at test time using our simple, standalone approach."
                },
                {
                    "title": "LEVER: Learning to Verify Language-to-Code Generation with Execution",
                    "abstract": "The advent of large language models trained on code (code LLMs) has led to significant progress in language-to-code generation. State-of-the-art approaches in this area combine LLM decoding with sample pruning and reranking using test cases or heuristics based on the execution results. However, it is challenging to obtain test cases for many real-world language-to-code applications, and heuristics cannot well capture the semantic features of the execution results, such as data type and value range, which often indicates the correctness of the program. In this work, we propose LEVER, a simple approach to improve language-to-code generation by learning to verify the generated programs with their execution results. Specifically, we train verifiers to determine whether a program sampled from the LLMs is correct or not based on the natural language input, the program itself and its execution results. The sampled programs are reranked by combining the verification score with the LLM generation probability, and marginalizing over programs with the same execution results. On four datasets across the domains of table QA, math QA and basic Python programming, LEVER consistently improves over the base code LLMs(4.6% to 10.9% with code-davinci-002) and achieves new state-of-the-art results on all of them."
                },
                {
                    "title": "FlashFill++: Scaling Programming by Example by Cutting to the Chase",
                    "abstract": "Programming-by-Examples (PBE) involves synthesizing an \"intended program\" from a small set of user-provided input-output examples. A key PBE strategy has been to restrict the search to a carefully designed small domain-specific language (DSL) with \"effectively-invertible\" (EI) operators at the top and \"effectively-enumerable\" (EE) operators at the bottom. This facilitates an effective combination of top-down synthesis strategy (which backpropagates outputs over various paths in the DSL using inverse functions) with a bottom-up synthesis strategy (which propagates inputs over various paths in the DSL). We address the problem of scaling synthesis to large DSLs with several non-EI/EE operators. This is motivated by the need to support a richer class of transformations and the need for readable code generation. We propose a novel solution strategy that relies on propagating fewer values and over fewer paths. Our first key idea is that of \"cut functions\" that prune the set of values being propagated by using knowledge of the sub-DSL on the other side. Cuts can be designed to preserve completeness of synthesis; however, DSL designers may use incomplete cuts to have finer control over the kind of programs synthesized. In either case, cuts make search feasible for non-EI/EE operators and efficient for deep DSLs. Our second key idea is that of \"guarded DSLs\" that allow a precedence on DSL operators, which dynamically controls exploration of various paths in the DSL. This makes search efficient over grammars with large fanouts without losing recall. It also makes ranking simpler yet more effective in learning an intended program from very few examples. Both cuts and precedence provide a mechanism to the DSL designer to restrict search to a reasonable, and possibly incomplete, space of programs. Using cuts and gDSLs, we have built FlashFill++, an industrial-strength PBE engine for performing rich string transformations, including datetime and number manipulations. The FlashFill++ gDSL is designed to enable readable code generation in different target languages including Excel's formula language, PowerFx, and Python. We show FlashFill++ is more expressive, more performant, and generates better quality code than comparable existing PBE systems. FlashFill++ is being deployed in several mass-market products ranging from spreadsheet software to notebooks and business intelligence applications, each with millions of users."
                },
                {
                    "title": "Parsel\ud83e\udd86: Algorithmic Reasoning with Language Models by Composing Decompositions",
                    "abstract": "Despite recent success in large language model (LLM) reasoning, LLMs struggle with hierarchical multi-step reasoning tasks like generating complex programs. For these tasks, humans often start with a high-level algorithmic design and implement each part gradually. We introduce Parsel, a framework enabling automatic implementation and validation of complex algorithms with code LLMs. With Parsel, we automatically decompose algorithmic tasks into hierarchical natural language function descriptions and then search over combinations of possible function implementations using tests. We show that Parsel can be used across domains requiring hierarchical reasoning, including program synthesis and robotic planning. We find that, using Parsel, LLMs solve more competition-level problems in the APPS dataset, resulting in pass rates over 75\\% higher than prior results from directly sampling AlphaCode and Codex, while often using a smaller sample budget. Moreover, with automatically generated tests, we find that Parsel can improve the state-of-the-art pass@1 performance on HumanEval from 67\\% to 85\\%. We also find that LLM-generated robotic plans using Parsel are more than twice as likely to be considered accurate than directly generated plans. Lastly, we explore how Parsel addresses LLM limitations and discuss how Parsel may be useful for human programmers. We release our code at https://github.com/ezelikman/parsel"
                },
                {
                    "title": "Self-Instruct: Aligning Language Models with Self-Generated Instructions",
                    "abstract": "Large \u201cinstruction-tuned\u201d language models (i.e., finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is often limited in quantity, diversity, and creativity, therefore hindering the generality of the tuned model. We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off their own generations. Our pipeline generates instructions, input, and output samples from a language model, then filters invalid or similar ones before using them to finetune the original model. Applying our method to the vanilla GPT3, we demonstrate a 33% absolute improvement over the original model on Super-NaturalInstructions, on par with the performance of InstructGPT-001, which was trained with private user data and human annotations. For further evaluation, we curate a set of expert-written instructions for novel tasks, and show through human evaluation that tuning GPT3 with Self-Instruct outperforms using existing public instruction datasets by a large margin, leaving only a 5% absolute gap behind InstructGPT-001. Self-Instruct provides an almost annotation-free method for aligning pre-trained language models with instructions, and we release our large synthetic dataset to facilitate future studies on instruction tuning."
                },
                {
                    "title": "Generating Sequences by Learning to Self-Correct",
                    "abstract": "Sequence generation applications require satisfying semantic constraints, such as ensuring that programs are correct, using certain keywords, or avoiding undesirable content. Language models, whether fine-tuned or prompted with few-shot demonstrations, frequently violate these constraints, and lack a mechanism to iteratively revise their outputs. Moreover, some powerful language models are of extreme scale or inaccessible, making it inefficient, if not infeasible, to update their parameters for task-specific adaptation. We present Self-Correction, an approach that decouples an imperfect base generator (an off-the-shelf language model or supervised sequence-to-sequence model) from a separate corrector that learns to iteratively correct imperfect generations. To train the corrector, we propose an online training procedure that can use either scalar or natural language feedback on intermediate imperfect generations. We show that Self-Correction improves upon the base generator in three diverse generation tasks - mathematical program synthesis, lexically-constrained generation, and toxicity control - even when the corrector is much smaller than the base generator."
                },
                {
                    "title": "CodeT: Code Generation with Generated Tests",
                    "abstract": "The task of generating code solutions for a given programming problem can benefit from the use of pre-trained language models such as Codex, which can produce multiple diverse samples. However, a major challenge for this task is to select the most appropriate solution from the multiple samples generated by the pre-trained language models. A natural way to evaluate the quality and correctness of a code solution is to run it against a set of test cases, but the manual creation of such test cases is often costly and time-consuming. In this paper, we propose a novel method, CodeT, that leverages the same pre-trained language models to automatically generate test cases for the code samples, thus reducing the human effort and increasing the coverage of the test scenarios. CodeT then executes the code samples using the generated test cases, and performs a dual execution agreement, which considers both the consistency of the outputs against the generated test cases and the agreement of the outputs with other code samples. We conduct comprehensive experiments on four benchmarks, HumanEval, MBPP, APPS and CodeContests, using five different pre-trained language models with varying sizes and capabilities. Our results show that CodeT can significantly improve the performance of code solution selection over previous methods, achieving remarkable and consistent gains across different models and benchmarks. For instance, CodeT improves the pass@1 metric on HumanEval to 65.8%, which represents an absolute improvement of 18.8% over the code-davinci-002 model, and an absolute improvement of more than 20% over the previous state-of-the-art results."
                },
                {
                    "title": "CodeRL: Mastering Code Generation through Pretrained Models and Deep Reinforcement Learning",
                    "abstract": "Program synthesis or code generation aims to generate a program that satisfies a problem specification. Recent approaches using large-scale pretrained language models (LMs) have shown promising results, yet they have some critical limitations. In particular, they often follow a standard supervised fine-tuning procedure to train a code generation model only from the pairs of natural-language problem descriptions and ground-truth programs. Such paradigm largely ignores some important but potentially useful signals in the problem specification such as unit tests, which thus often results in poor performance when solving complex unseen coding tasks. To address the limitations, we propose\"CodeRL\", a new framework for program synthesis tasks through pretrained LMs and deep reinforcement learning (RL). Specifically, during training, we treat the code-generating LM as an actor network, and introduce a critic network that is trained to predict the functional correctness of generated programs and provide dense feedback signals to the actor. During inference, we introduce a new generation procedure with a critical sampling strategy that allows a model to automatically regenerate programs based on feedback from example unit tests and critic scores. For the model backbones, we extended the encoder-decoder architecture of CodeT5 with enhanced learning objectives, larger model sizes, and better pretraining data. Our method not only achieves new SOTA results on the challenging APPS benchmark, but also shows strong zero-shot transfer capability with new SOTA results on the simpler MBPP benchmark."
                },
                {
                    "title": "Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models",
                    "abstract": "Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit\"breakthrough\"behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting."
                },
                {
                    "title": "Fault-Aware Neural Code Rankers",
                    "abstract": "Large language models (LLMs) have demonstrated an impressive ability to generate code for various programming tasks. In many instances, LLMs can generate a correct program for a task when given numerous trials. Consequently, a recent trend is to do large scale sampling of programs using a model and then filtering/ranking the programs based on the program execution on a small number of known unit tests to select one candidate solution. However, these approaches assume that the unit tests are given and assume the ability to safely execute the generated programs (which can do arbitrary dangerous operations such as file manipulations). Both of the above assumptions are impractical in real-world software development. In this paper, we propose CodeRanker, a neural ranker that can predict the correctness of a sampled program without executing it. Our CodeRanker is fault-aware i.e., it is trained to predict different kinds of execution information such as predicting the exact compile/runtime error type (e.g., an IndexError or a TypeError). We show that CodeRanker can significantly increase the pass@1 accuracy of various code generation models (including Codex, GPT-Neo, GPT-J) on APPS, HumanEval and MBPP datasets."
                },
                {
                    "title": "CrossBeam: Learning to Search in Bottom-Up Program Synthesis",
                    "abstract": "Many approaches to program synthesis perform a search within an enormous space of programs to find one that satisfies a given specification. Prior works have used neural models to guide combinatorial search algorithms, but such approaches still explore a huge portion of the search space and quickly become intractable as the size of the desired program increases. To tame the search space blowup, we propose training a neural model to learn a hands-on search policy for bottom-up synthesis, instead of relying on a combinatorial search algorithm. Our approach, called CrossBeam, uses the neural model to choose how to combine previously-explored programs into new programs, taking into account the search history and partial program executions. Motivated by work in structured prediction on learning to search, CrossBeam is trained on-policy using data extracted from its own bottom-up searches on training tasks. We evaluate CrossBeam in two very different domains, string manipulation and logic programming. We observe that CrossBeam learns to search efficiently, exploring much smaller portions of the program space compared to the state-of-the-art."
                },
                {
                    "title": "ZeroGen: Efficient Zero-shot Learning via Dataset Generation",
                    "abstract": "There is a growing interest in dataset generation recently due to the superior generative capacity of large pre-trained language models (PLMs). In this paper, we study a flexible and efficient zero-short learning method, ZeroGen.Given a zero-shot task, we first generate a dataset from scratch using PLMs in an unsupervised manner. Then, we train a tiny task model (e.g., LSTM) under the supervision of the synthesized dataset. This approach allows highly efficient inference as the final task model only has orders of magnitude fewer parameters comparing to PLMs (e.g., GPT2-XL).Apart from being annotation-free and efficient, we argue that ZeroGen can also provide useful insights from the perspective of data-free model-agnostic knowledge distillation, and unreferenced text generation evaluation. Experiments and analysis on different NLP tasks, namely, text classification, question answering, and natural language inference, show the effectiveness of ZeroGen."
                },
                {
                    "title": "Generating Training Data with Language Models: Towards Zero-Shot Language Understanding",
                    "abstract": "Pretrained language models (PLMs) have demonstrated remarkable performance in various natural language processing tasks: Unidirectional PLMs (e.g., GPT) are well known for their superior text generation capabilities; bidirectional PLMs (e.g., BERT) have been the prominent choice for natural language understanding (NLU) tasks. While both types of models have achieved promising few-shot learning performance, their potential for zero-shot learning has been underexplored. In this paper, we present a simple approach that uses both types of PLMs for fully zero-shot learning of NLU tasks without requiring any task-specific data: A unidirectional PLM generates class-conditioned texts guided by prompts, which are used as the training data for fine-tuning a bidirectional PLM. With quality training data selected based on the generation probability and regularization techniques (label smoothing and temporal ensembling) applied to the fine-tuning stage for better generalization and stability, our approach demonstrates strong performance across seven classification tasks of the GLUE benchmark (e.g., 72.3/73.8 on MNLI-m/mm and 92.8 on SST-2), significantly outperforming zero-shot prompting methods and achieving even comparable results to strong few-shot approaches using 32 training samples per class."
                },
                {
                    "title": "Competition-level code generation with AlphaCode",
                    "abstract": "Programming is a powerful and ubiquitous problem-solving tool. Systems that can assist programmers or even generate programs themselves could make programming more productive and accessible. Recent transformer-based neural network models show impressive code generation abilities yet still perform poorly on more complex tasks requiring problem-solving skills, such as competitive programming problems. Here, we introduce AlphaCode, a system for code generation that achieved an average ranking in the top 54.3% in simulated evaluations on recent programming competitions on the Codeforces platform. AlphaCode solves problems by generating millions of diverse programs using specially trained transformer-based networks and then filtering and clustering those programs to a maximum of just 10 submissions. This result marks the first time an artificial intelligence system has performed competitively in programming competitions. Description Machine learning systems can program too Computer programming competitions are popular tests among programmers that require critical thinking informed by experience and creating solutions to unforeseen problems, both of which are key aspects of human intelligence but challenging to mimic by machine learning models. Using self-supervised learning and an encoder-decoder transformer architecture, Li et al. developed AlphaCode, a deep-learning model that can achieve approximately human-level performance on the Codeforces platform, which regularly hosts these competitions and attracts numerous participants worldwide (see the Perspective by Kolter). The development of such coding platforms could have a huge impact on programmers\u2019 productivity. It may even change the culture of programming by shifting human work to formulating problems, with machine learning being the main one responsible for generating and executing codes. \u2014YS Modern machine learning systems can achieve average human-level performance in popular competitive programming contests."
                },
                {
                    "title": "Scaling Neural Program Synthesis with Distribution-based Search",
                    "abstract": "We consider the problem of automatically constructing computer programs from input-output examples. We investigate how to augment probabilistic and neural program synthesis methods with new search algorithms, proposing a framework called distribution-based search. Within this framework, we introduce two new search algorithms: Heap Search, an enumerative method, and SQRT Sampling, a probabilistic method. We prove certain optimality guarantees for both methods, show how they integrate with probabilistic and neural techniques, and demonstrate how they can operate at scale across parallel compute environments. Collectively these findings offer theoretical and applied studies of search algorithms for program synthesis that integrate with recent developments in machine-learned program synthesizers."
                },
                {
                    "title": "Evaluating Large Language Models Trained on Code",
                    "abstract": "We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics."
                },
                {
                    "title": "SpreadsheetCoder: Formula Prediction from Semi-structured Context",
                    "abstract": "Spreadsheet formula prediction has been an important program synthesis problem with many real-world applications. Previous works typically utilize input-output examples as the specification for spreadsheet formula synthesis, where each input-output pair simulates a separate row in the spreadsheet. However, this formulation does not fully capture the rich context in real-world spreadsheets. First, spreadsheet data entries are organized as tables, thus rows and columns are not necessarily independent from each other. In addition, many spreadsheet tables include headers, which provide high-level descriptions of the cell data. However, previous synthesis approaches do not consider headers as part of the specification. In this work, we present the first approach for synthesizing spreadsheet formulas from tabular context, which includes both headers and semi-structured tabular data. In particular, we propose SpreadsheetCoder, a BERT-based model architecture to represent the tabular context in both row-based and column-based formats. We train our model on a large dataset of spreadsheets, and demonstrate that SpreadsheetCoder achieves top-1 prediction accuracy of 42.51%, which is a considerable improvement over baselines that do not employ rich tabular context. Compared to the rule-based system, SpreadsheetCoder assists 82% more users in composing formulas on Google Sheets."
                },
                {
                    "title": "Leveraging Language to Learn Program Abstractions and Search Heuristics",
                    "abstract": "Inductive program synthesis, or inferring programs from examples of desired behavior, offers a general paradigm for building interpretable, robust, and generalizable machine learning systems. Effective program synthesis depends on two key ingredients: a strong library of functions from which to build programs, and an efficient search strategy for finding programs that solve a given task. We introduce LAPS (Language for Abstraction and Program Search), a technique for using natural language annotations to guide joint learning of libraries and neurally-guided search models for synthesis. When integrated into a state-of-the-art library learning system (DreamCoder), LAPS produces higher-quality libraries and improves search efficiency and generalization on three domains -- string editing, image composition, and abstract reasoning about scenes -- even when no natural language hints are available at test time."
                },
                {
                    "title": "DreamCoder: bootstrapping inductive program synthesis with wake-sleep library learning",
                    "abstract": "We present a system for inductive program synthesis called DreamCoder, which inputs a corpus of synthesis problems each specified by one or a few examples, and automatically derives a library of program components and a neural search policy that can be used to efficiently solve other similar synthesis problems. The library and search policy bootstrap each other iteratively through a variant of \"wake-sleep\" approximate Bayesian learning. A new refactoring algorithm based on E-graph matching identifies common sub-components across synthesized programs, building a progressively deepening library of abstractions capturing the structure of the input domain. We evaluate on eight domains including classic program synthesis areas and AI tasks such as planning, inverse graphics, and equation discovery. We show that jointly learning the library and neural search policy leads to solving more problems, and solving them more quickly."
                },
                {
                    "title": "Measuring Coding Challenge Competence With APPS",
                    "abstract": "While programming is one of the most broadly applicable skills in modern society, modern machine learning models still cannot code solutions to basic problems. Despite its importance, there has been surprisingly little work on evaluating code generation, and it can be difficult to accurately assess code generation performance rigorously. To meet this challenge, we introduce APPS, a benchmark for code generation. Unlike prior work in more restricted settings, our benchmark measures the ability of models to take an arbitrary natural language specification and generate satisfactory Python code. Similar to how companies assess candidate software developers, we then evaluate models by checking their generated code on test cases. Our benchmark includes 10,000 problems, which range from having simple one-line solutions to being substantial algorithmic challenges. We fine-tune large language models on both GitHub and our training set, and we find that the prevalence of syntax errors is decreasing exponentially as models improve. Recent models such as GPT-Neo can pass approximately 20% of the test cases of introductory problems, so we find that machine learning models are now beginning to learn how to code. As the social significance of automatic code generation increases over the coming years, our benchmark can provide an important measure for tracking advancements."
                },
                {
                    "title": "Just-in-time learning for bottom-up enumerative synthesis",
                    "abstract": "A key challenge in program synthesis is the astronomical size of the search space the synthesizer has to explore. In response to this challenge, recent work proposed to guide synthesis using learned probabilistic models. Obtaining such a model, however, might be infeasible for a problem domain where no high-quality training data is available. In this work we introduce an alternative approach to guided program synthesis: instead of training a model ahead of time we show how to bootstrap one just in time, during synthesis, by learning from partial solutions encountered along the way. To make the best use of the model, we also propose a new program enumeration algorithm we dub guided bottom-up search, which extends the efficient bottom-up search with guidance from probabilistic models. We implement this approach in a tool called Probe, which targets problems in the popular syntax-guided synthesis (SyGuS) format. We evaluate Probe on benchmarks from the literature and show that it achieves significant performance gains both over unguided bottom-up search and over a state-of-the-art probability-guided synthesizer, which had been trained on a corpus of existing solutions. Moreover, we show that these performance gains do not come at the cost of solution quality: programs generated by Probe are only slightly more verbose than the shortest solutions and perform no unnecessary case-splitting."
                },
                {
                    "title": "Synthesize, Execute and Debug: Learning to Repair for Neural Program Synthesis",
                    "abstract": "The use of deep learning techniques has achieved significant progress for program synthesis from input-output examples. However, when the program semantics become more complex, it still remains a challenge to synthesize programs consistent with the specification. In this work, we propose SED, a neural program generation framework that incorporates synthesis, execution, and debugging stages. Instead of purely relying on the neural program synthesizer to generate the final program, SED first produces initial programs using the neural program synthesizer component, then utilizes a neural program debugger to iteratively repair the generated programs. The integration of the debugger component enables SED to modify the programs based on the execution results and specification, which resembles the coding process of human programmers. On Karel, a challenging input-output program synthesis benchmark, SED reduces the error rate of the neural program synthesizer itself by at least 7.5%, and outperforms the standard beam search for decoding."
                },
                {
                    "title": "Program-Guided Image Manipulators",
                    "abstract": "Humans are capable of building holistic representations for images at various levels, from local objects, to pairwise relations, to global structures. The interpretation of structures involves reasoning over repetition and symmetry of the objects in the image. In this paper, we present the Program-Guided Image Manipulator (PG-IM), inducing neuro-symbolic program-like representations to represent and manipulate images. Given an image, PG-IM detects repeated patterns, induces symbolic programs, and manipulates the image using a neural network that is guided by the program. PG-IM learns from a single image, exploiting its internal statistics. Despite trained only on image inpainting, PG-IM is directly capable of extrapolation and regularity editing in a unified framework. Extensive experiments show that PG-IM achieves superior performance on all the tasks."
                },
                {
                    "title": "Write, Execute, Assess: Program Synthesis with a REPL",
                    "abstract": "We present a neural program synthesis approach integrating components which write, execute, and assess code to navigate the search space of possible programs. We equip the search process with an interpreter or a read-eval-print-loop (REPL), which immediately executes partially written programs, exposing their semantics. The REPL addresses a basic challenge of program synthesis: tiny changes in syntax can lead to huge changes in semantics. We train a pair of models, a policy that proposes the new piece of code to write, and a value function that assesses the prospects of the code written so-far. At test time we can combine these models with a Sequential Monte Carlo algorithm. We apply our approach to two domains: synthesizing text editing programs and inferring 2D and 3D graphics programs."
                },
                {
                    "title": "Execution-Guided Neural Program Synthesis",
                    "abstract": "Neural"
                },
                {
                    "title": "Neural-Guided Deductive Search for Real-Time Program Synthesis from Examples",
                    "abstract": "Synthesizing user-intended programs from a small number of input-output examples is a challenging problem with several important applications like spreadsheet manipulation, data wrangling and code refactoring. Existing synthesis systems either completely rely on deductive logic techniques that are extensively hand-engineered or on purely statistical models that need massive amounts of data, and in general fail to provide real-time synthesis on challenging benchmarks. In this work, we propose Neural Guided Deductive Search (NGDS), a hybrid synthesis technique that combines the best of both symbolic logic techniques and statistical models. Thus, it produces programs that satisfy the provided specifications by construction and generalize well on unseen examples, similar to data-driven systems. Our technique effectively utilizes the deductive search framework to reduce the learning problem of the neural component to a simple supervised learning setup. Further, this allows us to both train on sparingly available real-world data and still leverage powerful recurrent neural network encoders. We demonstrate the effectiveness of our method by evaluating on real-world customer scenarios by synthesizing accurate programs with up to 12x speed-up compared to state-of-the-art systems."
                },
                {
                    "title": "RobustFill: Neural Program Learning under Noisy I/O",
                    "abstract": "The problem of automatically generating a computer program from some specification has been studied since the early days of AI. Recently, two competing approaches for automatic program learning have received significant attention: (1) neural program synthesis, where a neural network is conditioned on input/output (I/O) examples and learns to generate a program, and (2) neural program induction, where a neural network generates new outputs directly using a latent program representation. Here, for the first time, we directly compare both approaches on a large-scale, real-world learning task and we additionally contrast to rule-based program synthesis, which uses hand-crafted semantics to guide the program generation. Our neural models use a modified attention RNN to allow encoding of variable-sized sets of I/O pairs, which achieve 92% accuracy on a real-world test set, compared to the 34% accuracy of the previous best neural synthesis approach. The synthesis model also outperforms a comparable induction model on this task, but we more importantly demonstrate that the strength of each approach is highly dependent on the evaluation metric and end-user application. Finally, we show that we can train our neural models to remain very robust to the type of noise expected in real-world data (e.g., typos), while a highly-engineered rule-based system fails entirely."
                },
                {
                    "title": "DeepCoder: Learning to Write Programs",
                    "abstract": "We develop a first line of attack for solving programming competition-style problems from input-output examples using deep learning. The approach is to train a neural network to predict properties of the program that generated the outputs from the inputs. We use the neural network\u2019s predictions to augment search techniques from the programming languages community, including enumerative search and an SMT-based solver. Empirically, we show that our approach leads to an order of magnitude speedup over the strong non-augmented baselines and a Recurrent Neural Network approach, and that we are able to solve problems of difficulty comparable to the simplest problems on programming competition websites."
                },
                {
                    "title": "Human-level concept learning through probabilistic program induction",
                    "abstract": "Handwritten characters drawn by a model Not only do children learn effortlessly, they do so quickly and with a remarkable ability to use what they have learned as the raw material for creating new stuff. Lake et al. describe a computational model that learns in a similar fashion and does so better than current deep learning algorithms. The model classifies, parses, and recreates handwritten characters, and can generate new letters of the alphabet that look \u201cright\u201d as judged by Turing-like tests of the model's output in comparison to what real humans produce. Science, this issue p. 1332 Combining the capacity to handle noise with probabilistic learning yields humanlike performance in a computational model. People learning new concepts can often generalize successfully from just a single example, yet machine learning algorithms typically require tens or hundreds of examples to perform with similar accuracy. People can also use learned concepts in richer ways than conventional algorithms\u2014for action, imagination, and explanation. We present a computational model that captures these human learning abilities for a large class of simple visual concepts: handwritten characters from the world\u2019s alphabets. The model represents concepts as simple programs that best explain observed examples under a Bayesian criterion. On a challenging one-shot classification task, the model achieves human-level performance while outperforming recent deep learning approaches. We also present several \u201cvisual Turing tests\u201d probing the model\u2019s creative generalization abilities, which in many cases are indistinguishable from human behavior."
                },
                {
                    "title": "FlashMeta: a framework for inductive program synthesis",
                    "abstract": "Inductive synthesis, or programming-by-examples (PBE) is gaining prominence with disruptive applications for automating repetitive tasks in end-user programming. However, designing, developing, and maintaining an effective industrial-quality inductive synthesizer is an intellectual and engineering challenge, requiring 1-2 man-years of effort. Our novel observation is that many PBE algorithms are a natural fall-out of one generic meta-algorithm and the domain-specific properties of the operators in the underlying domain-specific language (DSL). The meta-algorithm propagates example-based constraints on an expression to its subexpressions by leveraging associated witness functions, which essentially capture the inverse semantics of the underlying operator. This observation enables a novel program synthesis methodology called data-driven domain-specific deduction (D4), where domain-specific insight, provided by the DSL designer, is separated from the synthesis algorithm. Our FlashMeta framework implements this methodology, allowing synthesizer developers to generate an efficient synthesizer from the mere DSL definition (if properties of the DSL operators have been modeled). In our case studies, we found that 10+ existing industrial-quality mass-market applications based on PBE can be cast as instances of D4. Our evaluation includes reimplementation of some prior works, which in FlashMeta become more efficient, maintainable, and extensible. As a result, FlashMeta-based PBE tools are deployed in several industrial products, including Microsoft PowerShell 3.0 for Windows 10, Azure Operational Management Suite, and Microsoft Cortana digital assistant."
                },
                {
                    "title": "Inductive programming meets the real world",
                    "abstract": "Inductive programming can liberate users from performing tedious and repetitive tasks."
                },
                {
                    "title": "Type-and-example-directed program synthesis",
                    "abstract": "This paper presents an algorithm for synthesizing recursive functions that process algebraic datatypes. It is founded on proof-theoretic techniques that exploit both type information and input\u2013output examples to prune the search space. The algorithm uses refinement trees, a data structure that succinctly represents constraints on the shape of generated code. We evaluate the algorithm by using a prototype implementation to synthesize more than 40 benchmarks and several non-trivial larger examples. Our results demonstrate that the approach meets or outperforms the state-of-the-art for this domain, in terms of synthesis time or attainable size of the generated programs."
                },
                {
                    "title": "Synthesizing data structure transformations from input-output examples",
                    "abstract": "We present a method for example-guided synthesis of functional programs over recursive data structures. Given a set of input-output examples, our method synthesizes a program in a functional language with higher-order combinators like map and fold. The synthesized program is guaranteed to be the simplest program in the language to fit the examples. Our approach combines three technical ideas: inductive generalization, deduction, and enumerative search. First, we generalize the input-output examples into hypotheses about the structure of the target program. For each hypothesis, we use deduction to infer new input/output examples for the missing subexpressions. This leads to a new subproblem where the goal is to synthesize expressions within each hypothesis. Since not every hypothesis can be realized into a program that fits the examples, we use a combination of best-first enumeration and deduction to search for a hypothesis that meets our needs. We have implemented our method in a tool called \u03bb2, and we evaluate this tool on a large set of synthesis problems involving lists, trees, and nested data structures. The experiments demonstrate the scalability and broad scope of \u03bb2. A highlight is the synthesis of a program believed to be the world's earliest functional pearl."
                },
                {
                    "title": "Syntax-guided synthesis",
                    "abstract": "The classical formulation of the program-synthesis problem is to find a program that meets a correctness specification given as a logical formula. Recent work on program synthesis and program optimization illustrates many potential benefits of allowing the user to supplement the logical specification with a syntactic template that constrains the space of allowed implementations. Our goal is to identify the core computational problem common to these proposals in a logical framework. The input to the syntax-guided synthesis problem (SyGuS) consists of a background theory, a semantic correctness specification for the desired program given by a logical formula, and a syntactic set of candidate implementations given by a grammar. The computational problem then is to find an implementation from the set of candidate expressions so that it satisfies the specification in the given theory. We describe three different instantiations of the counter-example-guided-inductive-synthesis (CEGIS) strategy for solving the synthesis problem, report on prototype implementations, and present experimental results on an initial set of benchmarks."
                },
                {
                    "title": "Learning Programs: A Hierarchical Bayesian Approach",
                    "abstract": "We are interested in learning programs for multiple related tasks given only a few training examples per task. Since the program for a single task is underdetermined by its data, we introduce a nonparametric hierarchical Bayesian prior over programs which shares statistical strength across multiple tasks. The key challenge is to parametrize this multi-task sharing. For this, we introduce a new representation of programs based on combinatory logic and provide an MCMC algorithm that can perform safe program transformations on this representation to reveal shared inter-program substructures."
                },
                {
                    "title": "Your Wish is My Command: Programming By Example",
                    "abstract": "Foreword Ben Shneiderman Introduction Henry Lieberman 1) Novice Programming Comes of Age David Canfield Smith, Allen Cypher, Larry Tesler 2) Generalizing by Removing Detail: How Any Program Can Be Created by Working with Examples Ken Kahn 3) Demonstrational Interfaces: Sometimes You Need a Little Intelligence Sometimes You Need a Lot Brad A. Myers, Richard McDaniel 4) Web Browsing by Demonstration Atsushi Sugiura 5) Programming by Demonstration for Information Agents Mathias Bauer, Dietmar Dengler, Gabriele Paul 6) End Users and GIS: A Demonstration is Worth a Thousand Words Carol Traynor and Marian Williams 7) Bring Programming by Demonstration to CAD Users Patrick Girard 8) Demonstrating the Hidden Features That Make an Application Work Richard McDaniel 9) A reporting tool using programming by example for format designation Tetsuya Masuishi and Nobuo Takahashi 10) Composition by Example Toshiyuki Masui 11) Learning Repetitive Text-editing Procedures with SMARTedit Tessa Lau, Steve Wolfman, Pedro Domingos and Daniel S. Weld 12) Training Agents to Recognize Text by Example Henry Lieberman, Bonnie A. Nardi and David J. Wright 13) SWYN: A Visual Representation for Regular Expressions Alan Blackwell 14) Learning Users' Habits to Automate Repetitive Tasks Jean-David Ruvini and Christophe Dony 15) Domain-independent programming by demonstration in existing applications Gordon W. Paynter and Ian H. Witten 16) Stimulus-Response PBD: Demonstrating When as Well as What David Wolber and Brad A. Myers 17) Pavlov: Where PBD Meets Macromedia's Director David Wolber 18) Programming by Analogous Examples Alexander Repenning and Corrina Perrone-Smith 19) Visual Generalization in Programming by Example Robert St. Amant, Henry Lieberman, Richard Potter, and Luke Zettlemoyer"
                },
                {
                    "title": "The \"wake-sleep\" algorithm for unsupervised neural networks.",
                    "abstract": "An unsupervised learning algorithm for a multilayer network of stochastic neurons is described. Bottom-up \"recognition\" connections convert the input into representations in successive hidden layers, and top-down \"generative\" connections reconstruct the representation in one layer from the representation in the layer above. In the \"wake\" phase, neurons are driven by recognition connections, and generative connections are adapted to increase the probability that they would reconstruct the correct activity vector in the layer below. In the \"sleep\" phase, neurons are driven by generative connections, and recognition connections are adapted to increase the probability that they would produce the correct activity vector in the layer above."
                },
                {
                    "title": "Reflexion: an autonomous agent with dynamic memory and self-reflection",
                    "abstract": "Recent advancements in decision-making large language model (LLM) agents have demonstrated impressive performance across various benchmarks. However, these state-of-the-art approaches typically necessitate internal model fine-tuning, external model fine-tuning, or policy optimization over a defined state space. Implementing these methods can prove challenging due to the scarcity of high-quality training data or the lack of well-defined state space. Moreover, these agents do not possess certain qualities inherent to human decision-making processes, specifically the ability to learn from mistakes. Self-reflection allows humans to efficiently solve novel problems through a process of trial and error. Building on recent research, we propose Reflexion, an approach that endows an agent with dynamic memory and self-reflection capabilities to enhance its existing reasoning trace and task-specific action choice abilities. To achieve full automation, we introduce a straightforward yet effective heuristic that enables the agent to pinpoint hallucination instances, avoid repetition in action sequences, and, in some environments, construct an internal memory map of the given environment. To assess our approach, we evaluate the agent\u2019s ability to complete decision-making tasks in AlfWorld environments and knowledge-intensive, search-based question-and-answer tasks in HotPotQA environments. We observe success rates of 97% and 51%, respectively, and provide a discussion on the emergent property of self-reflection."
                },
                {
                    "title": "Ultra-Strong Machine Learning \u2013 Comprehensibility of Programs Learned with ILP",
                    "abstract": "During the 1980s Michie de\ufb01ned Machine Learning in terms of two orthogonal axes of performance: predictive accuracy and comprehensibility of generated hypotheses. Since predictive accuracy was readily measurable and comprehensibility not so, later de\ufb01nitions in the 1990s, such as that of Mitchell, tended to use a one-dimensional approach to Machine Learning based solely on predictive accuracy, ultimately favour-ing statistical over symbolic Machine Learning approaches. In this paper we provide a de\ufb01nition of comprehensibility of hypotheses which can be estimated using human participant trials. We present two sets of experiments testing human comprehensibility of logic programs. In the \ufb01rst experiment we test human comprehensibility with and without predicate invention. Results indicate that comprehensibility is a\ufb00ected not only by the complexity of the presented program but also by the existence of anonymous predicate symbols. In the second experiment we directly test whether any state-of-the-art ILP systems are ultra-strong learners in Michie\u2019s sense, and select the Metagol system for use in humans trials. Results show that participants were not able to learn the relational concept on their own from a set of examples but they were able to apply the relational de\ufb01nition provided by the ILP system correctly. This implies the existence of a class of relational concepts which are hard to acquire for humans, though easy to understand given an abstract explanation. We believe improved understanding of this class could have potential relevance to contexts involving human learning, teaching and verbal interaction."
                },
                {
                    "title": "Learning to Learn Programs from Examples : Going Beyond Program Structure",
                    "abstract": "Programming-by-example technologies let end users construct and run new programs by providing examples of the intended program behavior. But, the few provided examples seldom uniquely determine the intended program. Previous approaches to picking a program used a bias toward shorter or more naturally structured programs. Our work here gives a machine learning approach for learning to learn programs that departs from previous work by relying upon features that are independent of the program structure, instead relying upon a learned bias over program behaviors, and more generally over program execution traces. Our approach leverages abundant unlabeled data for semisupervised learning, and incorporates simple kinds of world knowledge for common-sense reasoning during program induction. These techniques are evaluated in two programming-by-example domains, improving the accuracy of program learners."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG",
                "cs.PL",
                "cs.SE"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can large language models (LLMs) be effectively fine-tuned to perform Programming-by-Example (PBE) tasks in general-purpose Turing-complete languages, given their current limitations in generalization and inductive reasoning?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning and artificial intelligence, particularly in the area of program synthesis. If LLMs can be adapted to perform PBE effectively, it would significantly broaden the applicability of these models, allowing them to tackle a wider range of programming tasks without relying on domain-specific languages. This could lead to practical applications in software development, education, and automation, ultimately enhancing productivity and creativity in coding. Furthermore, it would provide insights into the capabilities and limitations of LLMs, guiding future research in improving their inductive reasoning and problem-solving skills.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexities of few-shot inductive inference, where the model must generalize from a limited number of input-output examples to infer the underlying program. Naive approaches may fail because LLMs often rely heavily on natural language cues, which may not be present in PBE tasks. Additionally, the technical obstacles include the need for effective fine-tuning strategies that can adapt LLMs to diverse programming problems while maintaining generalization capabilities. Theoretical challenges also arise from the need to understand the limitations of LLMs in capturing the true latent regularities of programming tasks.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on LLMs generating code from natural language rather than addressing the unique challenges of PBE. Existing solutions have been limited by their reliance on domain-specific languages, which restricts their applicability. Barriers such as the lack of effective fine-tuning methods for general-purpose languages and the difficulty in achieving robust generalization beyond training distributions have hindered progress. Our approach differs by introducing a novel algorithm for adapting LLMs to small unlabeled datasets, which addresses these gaps and improves upon prior work by enabling broader problem coverage and better performance in Turing-complete languages.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves fine-tuning pretrained LLMs on a small unlabeled dataset of PBE problems, followed by"
            }
        },
        "author_data": {
            "ff8c29cc-ed05-4033-90a1-36d96afc2711": {
                "pk": "ff8c29cc-ed05-4033-90a1-36d96afc2711",
                "name": "Wen-Ding Li",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "21213f49-a59e-43a9-a2f3-da2b3e9e4423": {
                "pk": "21213f49-a59e-43a9-a2f3-da2b3e9e4423",
                "name": "Kevin Ellis",
                "collaborators": [
                    "Armando Solar-Lezama",
                    "Joshua B. Tenenbaum",
                    "Yewen Pu",
                    "Hao Tang",
                    "Darren Key",
                    "Marta Kryven",
                    "Josh Tenenbaum",
                    "Maxwell Nye",
                    "Leopoldo Sarra",
                    "Florian Marquardt"
                ],
                "domain": [
                    "Program Synthesis",
                    "Inductive Learning",
                    "Neural Networks",
                    "Human-AI Interaction"
                ],
                "publications": [
                    {
                        "title": "Human-like Few-Shot Learning via Bayesian Reasoning over Natural Language",
                        "abstract": "A core tension in models of concept learning is that the model must carefully balance the tractability of inference against the expressivity of the hypothesis class. Humans, however, can efficiently learn a broad range of concepts. We introduce a model of inductive learning that seeks to be human-like in that sense. It implements a Bayesian reasoning process where a language model first proposes candidate hypotheses expressed in natural language, which are then re-weighed by a prior and a likelihood. By estimating the prior from human data, we can predict human judgments on learning problems involving numbers and sets, spanning concepts that are generative, discriminative, propositional, and higher-order."
                    },
                    {
                        "title": "From Perception to Programs: Regularize, Overparameterize, and Amortize",
                        "abstract": "Toward combining inductive reasoning with perception abilities, we develop techniques for neurosymbolic program synthesis where perceptual input is first parsed by neural nets into a low-dimensional interpretable representation, which is then processed by a synthesized program. We explore several techniques for relaxing the problem and jointly learning all modules end-to-end with gradient descent: multitask learning; amortized inference; overparameterization; and a differentiable strategy for penalizing lengthy programs. Collectedly this toolbox improves the stability of gradient-guided program search, and suggests ways of learning both how to perceive input as discrete abstractions, and how to symbolically process those abstractions as programs."
                    },
                    {
                        "title": "Discovering Quantum Circuit Components with Program Synthesis",
                        "abstract": "Despite rapid progress in the field, it is still challenging to discover new ways to take advantage of quantum computation: all quantum algorithms need to be designed by hand, and quantum mechanics is notoriously counterintuitive. In this paper, we study how artificial intelligence, in the form of program synthesis, may help to overcome some of these difficulties, by showing how a computer can incrementally learn concepts relevant for quantum circuit synthesis with experience, and reuse them in unseen tasks. In particular, we focus on the decomposition of unitary matrices into quantum circuits, and we show how, starting from a set of elementary gates, we can automatically discover a library of new useful composite gates and use them to decompose more and more complicated unitaries."
                    },
                    {
                        "title": "Learning to Infer Graphics Programs from Hand-Drawn Images",
                        "abstract": "We introduce a model that learns to convert simple hand drawings into graphics programs written in a subset of \\LaTeX. The model combines techniques from deep learning and program synthesis. We learn a convolutional neural network that proposes plausible drawing primitives that explain an image. These drawing primitives are like a trace of the set of primitive commands issued by a graphics program. We learn a model that uses program synthesis techniques to recover a graphics program from that trace. These programs have constructs like variable bindings, iterative loops, or simple kinds of conditionals. With a graphics program in hand, we can correct errors made by the deep network, measure similarity between drawings by use of similar high-level geometric structures, and extrapolate drawings. Taken together these results are a step towards agents that induce useful, human-readable programs from perceptual input."
                    },
                    {
                        "title": "CrossBeam: Learning to Search in Bottom-Up Program Synthesis",
                        "abstract": "Many approaches to program synthesis perform a search within an enormous space of programs to find one that satisfies a given specification. Prior works have used neural models to guide combinatorial search algorithms, but such approaches still explore a huge portion of the search space and quickly become intractable as the size of the desired program increases. To tame the search space blowup, we propose training a neural model to learn a hands-on search policy for bottom-up synthesis, instead of relying on a combinatorial search algorithm. Our approach, called CrossBeam, uses the neural model to choose how to combine previously-explored programs into new programs, taking into account the search history and partial program executions. Motivated by work in structured prediction on learning to search, CrossBeam is trained on-policy using data extracted from its own bottom-up searches on training tasks. We evaluate CrossBeam in two very different domains, string manipulation and logic programming. We observe that CrossBeam learns to search efficiently, exploring much smaller portions of the program space compared to the state-of-the-art."
                    },
                    {
                        "title": "Efficient Pragmatic Program Synthesis with Informative Specifications",
                        "abstract": "Providing examples is one of the most common way for end-users to interact with program synthesizers. However, program synthesis systems assume that examples consistent with the program are chosen at random, and do not exploit the fact that users choose examples pragmatically. Prior work modeled program synthesis as pragmatic communication, but required an inefficient enumeration of the entire program space. In this paper, we show that it is possible to build a program synthesizer that is both pragmatic and efficient by approximating the joint distribution of programs with a product of independent factors, and performing pragmatic inference on each factor separately. This factored distribution approximates the exact joint distribution well when the examples are given pragmatically, and is compatible with a basic neuro-symbolic program synthesis algorithm. Surprisingly, we find that the synthesizer assuming a factored approximation performs better than a synthesizer assuming an exact joint distribution when evaluated on natural human inputs. This suggests that humans may be assuming a factored distribution while communicating programs."
                    },
                    {
                        "title": "Toward Trustworthy Neural Program Synthesis",
                        "abstract": "We develop an approach to estimate the probability that a program sampled from a large language model is correct. Given a natural language description of a programming problem, our method samples both candidate programs as well as candidate predicates specifying how the program should behave. This allows learning a model that forms a well-calibrated probabilistic prediction of program correctness. Our system also infers which predicates are useful to explain the behavior of the generated code, and humans preferred these in a human study over raw language model outputs. Our method is simple, easy to implement, and maintains state of the art generation accuracy results."
                    },
                    {
                        "title": "Rapid Motor Adaptation for Robotic Manipulator Arms",
                        "abstract": "Developing generalizable manipulation skills is a core challenge in embodied AI. This includes generalization across diverse task configurations, encompassing variations in object shape, density, friction coefficient, and external disturbances such as forces applied to the robot. Rapid Motor Adaptation (RMA) offers a promising solution to this challenge. It posits that essential hidden variables influencing an agent's task performance, such as object mass and shape, can be effectively inferred from the agent's action and proprioceptive history. Drawing inspiration from RMA in locomotion and in-hand rotation, we use depth perception to develop agents tailored for rapid motor adaptation in a variety of manipulation tasks. We evaluated our agents on four challenging tasks from the Maniskill2 benchmark, namely pick-and-place operations with hundreds of objects from the YCB and EGAD datasets, peg insertion with precise position and orientation, and operating a variety of faucets and handles, with customized environment variations. Empirical results demonstrate that our agents surpass state-of-the-art methods like automatic domain randomization and vision-based policies, obtaining better generalization performance and sample efficiency."
                    },
                    {
                        "title": "Active Preference Inference using Language Models and Probabilistic Reasoning",
                        "abstract": "Actively inferring user preferences, for example by asking good questions, is important for any human-facing decision-making system. Active inference allows such systems to adapt and personalize themselves to nuanced individual preferences. To enable this ability for instruction-tuned large language models (LLMs), one may prompt them to ask users questions to infer their preferences, transforming the language models into more robust, interactive systems. However, out of the box, these models are not efficient at extracting preferences: the questions they generate are not informative, requiring a high number of user interactions and impeding the usability of the downstream system. In this work, we introduce an inference-time algorithm that helps LLMs quickly infer preferences by using more informative questions. Our algorithm uses a probabilistic model whose conditional distributions are defined by prompting an LLM, and returns questions that optimize expected entropy and expected model change. Results in a simplified interactive web shopping setting with real product items show that an LLM equipped with our entropy reduction algorithm outperforms baselines with the same underlying LLM on task performance while using fewer user interactions."
                    },
                    {
                        "title": "WorldCoder, a Model-Based LLM Agent: Building World Models by Writing Code and Interacting with the Environment",
                        "abstract": "We give a model-based agent that builds a Python program representing its knowledge of the world based on its interactions with the environment. The world model tries to explain its interactions, while also being optimistic about what reward it can achieve. We define this optimism as a logical constraint between a program and a planner. We study our agent on gridworlds, and on task planning, finding our approach is more sample-efficient compared to deep RL, more compute-efficient compared to ReAct-style agents, and that it can transfer its knowledge across environments by editing its code."
                    },
                    {
                        "title": "Program Synthesis with Pragmatic Communication",
                        "abstract": "Program synthesis techniques construct or infer programs from user-provided specifications, such as input-output examples. Yet most specifications, especially those given by end-users, leave the synthesis problem radically ill-posed, because many programs may simultaneously satisfy the specification. Prior work resolves this ambiguity by using various inductive biases, such as a preference for simpler programs. This work introduces a new inductive bias derived by modeling the program synthesis task as rational communication, drawing insights from recursive reasoning models of pragmatics. Given a specification, we score a candidate program both on its consistency with the specification, and also whether a rational speaker would chose this particular specification to communicate that program. We develop efficient algorithms for such an approach when learning from input-output examples, and build a pragmatic program synthesizer over a simple grid-like layout domain. A user study finds that end-user participants communicate more effectively with the pragmatic program synthesizer over a non-pragmatic one."
                    },
                    {
                        "title": "DreamCoder: Growing generalizable, interpretable knowledge with wake-sleep Bayesian program learning",
                        "abstract": "Expert problem-solving is driven by powerful languages for thinking about problems and their solutions. Acquiring expertise means learning these languages -- systems of concepts, alongside the skills to use them. We present DreamCoder, a system that learns to solve problems by writing programs. It builds expertise by creating programming languages for expressing domain concepts, together with neural networks to guide the search for programs within these languages. A ``wake-sleep'' learning algorithm alternately extends the language with new symbolic abstractions and trains the neural network on imagined and replayed problems. DreamCoder solves both classic inductive programming tasks and creative tasks such as drawing pictures and building scenes. It rediscovers the basics of modern functional programming, vector algebra and classical physics, including Newton's and Coulomb's laws. Concepts are built compositionally from those learned earlier, yielding multi-layered symbolic representations that are interpretable and transferrable to new tasks, while still growing scalably and flexibly with experience."
                    },
                    {
                        "title": "Write, Execute, Assess: Program Synthesis with a REPL",
                        "abstract": "We present a neural program synthesis approach integrating components which write, execute, and assess code to navigate the search space of possible programs. We equip the search process with an interpreter or a read-eval-print-loop (REPL), which immediately executes partially written programs, exposing their semantics. The REPL addresses a basic challenge of program synthesis: tiny changes in syntax can lead to huge changes in semantics. We train a pair of models, a policy that proposes the new piece of code to write, and a value function that assesses the prospects of the code written so-far. At test time we can combine these models with a Sequential Monte Carlo algorithm. We apply our approach to two domains: synthesizing text editing programs and inferring 2D and 3D graphics programs."
                    },
                    {
                        "title": "Learning abstract structure for drawing by efficient motor program induction",
                        "abstract": "Humans flexibly solve new problems that differ qualitatively from those they were trained on. This ability to generalize is supported by learned concepts that capture structure common across different problems. Here we develop a naturalistic drawing task to study how humans rapidly acquire structured prior knowledge. The task requires drawing visual objects that share underlying structure, based on a set of composable geometric rules. We show that people spontaneously learn abstract drawing procedures that support generalization, and propose a model of how learners can discover these reusable drawing programs. Trained in the same setting as humans, and constrained to produce efficient motor actions, this model discovers new drawing routines that transfer to test objects and resemble learned features of human sequences. These results suggest that two principles guiding motor program induction in the model - abstraction (general programs that ignore object-specific details) and compositionality (recombining previously learned programs) - are key for explaining how humans learn structured internal representations that guide flexible reasoning and learning."
                    }
                ]
            }
        }
    },
    "2305.19094": {
        "paper_data": {
            "title": "Diffusion Model for Dense Matching",
            "url": "http://arxiv.org/abs/2305.19094v2",
            "arxiv_id": "2305.19094",
            "authors": [
                "Jisu Nam",
                "Gyuseong Lee",
                "Sunwoo Kim",
                "Hyeonsu Kim",
                "Hyoungwon Cho",
                "Seyeon Kim",
                "Seungryong Kim"
            ],
            "abstract": "The objective for establishing dense correspondence between paired images consists of two terms: a data term and a prior term. While conventional techniques focused on defining hand-designed prior terms, which are difficult to formulate, recent approaches have focused on learning the data term with deep neural networks without explicitly modeling the prior, assuming that the model itself has the capacity to learn an optimal prior from a large-scale dataset. The performance improvement was obvious, however, they often fail to address inherent ambiguities of matching, such as textureless regions, repetitive patterns, and large displacements. To address this, we propose DiffMatch, a novel conditional diffusion-based framework designed to explicitly model both the data and prior terms. Unlike previous approaches, this is accomplished by leveraging a conditional denoising diffusion model. DiffMatch consists of two main components: conditional denoising diffusion module and cost injection module. We stabilize the training process and reduce memory usage with a stage-wise training strategy. Furthermore, to boost performance, we introduce an inference technique that finds a better path to the accurate matching field. Our experimental results demonstrate significant performance improvements of our method over existing approaches, and the ablation studies validate our design choices along with the effectiveness of each component. Project page is available at https://ku-cvlab.github.io/DiffMatch/.",
            "introduction": "   1 Introduction  Establishing pixel-wise correspondences between pairs of images has been one of the crucial problems, as it supports a wide range of applications, including structure from motion (SfM)\u00a0(Schonberger & Frahm, 2016), simultaneous localization and mapping (SLAM)\u00a0(Durrant-Whyte & Bailey, 2006; Bailey & Durrant-Whyte, 2006), image editing\u00a0(Barnes et\u00a0al., 2009; Cheng et\u00a0al., 2010; Zhang et\u00a0al., 2020), and video analysis\u00a0(Hu et\u00a0al., 2018; Lai & Xie, 2019). In contrast to sparse correspondence\u00a0(Calonder et\u00a0al., 2010; Lowe, 2004; Sarlin et\u00a0al., 2020) which detects and matches only a sparse set of key points, dense correspondence\u00a0(P\u00e9rez et\u00a0al., 2013; Rocco et\u00a0al., 2017; Kim et\u00a0al., 2018; Cho et\u00a0al., 2021) aims to match all the points between input images.   In the probabilistic interpretation, the objective for dense correspondence can be defined with a data term, measuring matching evidence between source and target features, and a prior term, encoding prior knowledge of correspondence. Traditional methods\u00a0(P\u00e9rez et\u00a0al., 2013; Drulea & Nedevschi, 2011; Werlberger et\u00a0al., 2010; Lhuillier & Quan, 2000; Liu et\u00a0al., 2010; Ham et\u00a0al., 2016) explicitly incorporated hand-designed prior terms to achieve smoother correspondence, such as total variation (TV) or image discontinuity-aware smoothness. However, the formulation of the hand-crafted prior term is notoriously challenging and may vary depending on the specific dense correspondence tasks, such as geometric matching\u00a0(Liu et\u00a0al., 2010; Duchenne et\u00a0al., 2011; Kim et\u00a0al., 2013) or optical flow\u00a0(Weinzaepfel et\u00a0al., 2013; Revaud et\u00a0al., 2015).   Unlike them, recent approaches\u00a0(Kim et\u00a0al., 2017a; Sun et\u00a0al., 2018; Rocco et\u00a0al., 2017; 2020; Truong et\u00a0al., 2020b; Min & Cho, 2021; Kim et\u00a0al., 2018; Jiang et\u00a0al., 2021; Cho et\u00a0al., 2021; 2022) have focused on solely learning the data term with deep neural networks. However, despite demonstrating certain performance improvements, these methods still struggle with effectively addressing the inherent ambiguities encountered in dense correspondence, including challenges posed by textureless regions, repetitive patterns, large displacements, or noises. We argue that it is because they concentrate on maximizing the likelihood, which corresponds to learning the data term only, and do not explicitly consider the matching prior. This limits their ability to learn ideal matching field manifold, and leads to poor generalization.   On the other hand, diffusion models\u00a0(Ho et\u00a0al., 2020; Song et\u00a0al., 2020a; Song & Ermon, 2019; Song et\u00a0al., 2020b) have recently demonstrated a powerful capability for learning posterior distribution and have achieved considerable success in the field of generative models\u00a0(Karras et\u00a0al., 2020). Building on these advancements, recent studies\u00a0(Rombach et\u00a0al., 2022; Seo et\u00a0al., 2023; Saharia et\u00a0al., 2022a; Lugmayr et\u00a0al., 2022) have focused on controllable image synthesis by leveraging external conditions. Moreover, these advances in diffusion models have also led to successful applications in numerous discriminative tasks, such as depth estimation\u00a0(Saxena et\u00a0al., 2023b; Kim et\u00a0al., 2022; Duan et\u00a0al., 2023), object detection\u00a0(Chen et\u00a0al., 2022a), segmentation\u00a0(Gu et\u00a0al., 2022; Giannone et\u00a0al., 2022), and human pose estimation\u00a0(Holmquist & Wandt, 2022).   Inspired by the recent success of the diffusion model\u00a0(Ho et\u00a0al., 2020; Song et\u00a0al., 2020a; Song & Ermon, 2019; Song et\u00a0al., 2020b), we introduce DiffMatch, a conditional diffusion-based framework designed to explicitly model the matching field distribution within diffusion process.   Unlike existing discriminative learning-based methods\u00a0(Kim et\u00a0al., 2017a; Jiang et\u00a0al., 2021; Rocco et\u00a0al., 2017; 2020; Teed & Deng, 2020) that focus solely on maximizing the likelihood, DiffMatch aims to",
            "references": [
                {
                    "title": "The Surprising Effectiveness of Diffusion Models for Optical Flow and Monocular Depth Estimation",
                    "abstract": "Denoising diffusion probabilistic models have transformed image generation with their impressive fidelity and diversity. We show that they also excel in estimating optical flow and monocular depth, surprisingly, without task-specific architectures and loss functions that are predominant for these tasks. Compared to the point estimates of conventional regression-based methods, diffusion models also enable Monte Carlo inference, e.g., capturing uncertainty and ambiguity in flow and depth. With self-supervised pre-training, the combined use of synthetic and real data for supervised training, and technical innovations (infilling and step-unrolled denoising diffusion training) to handle noisy-incomplete training data, and a simple form of coarse-to-fine refinement, one can train state-of-the-art diffusion models for depth and optical flow estimation. Extensive experiments focus on quantitative performance against benchmarks, ablations, and the model's ability to capture uncertainty and multimodality, and impute missing values. Our model, DDVM (Denoising Diffusion Vision Model), obtains a state-of-the-art relative depth error of 0.074 on the indoor NYU benchmark and an Fl-all outlier rate of 3.26\\% on the KITTI optical flow benchmark, about 25\\% better than the best published method. For an overview see https://diffusion-vision.github.io."
                },
                {
                    "title": "DDP: Diffusion Model for Dense Visual Prediction",
                    "abstract": "We propose a simple, efficient, yet powerful framework for dense visual predictions based on the conditional diffusion pipeline. Our approach follows a \"noise-to-map\" generative paradigm for prediction by progressively removing noise from a random Gaussian distribution, guided by the image. The method, called DDP, efficiently extends the denoising diffusion process into the modern perception pipeline. Without task-specific design and architecture customization, DDP is easy to generalize to most dense prediction tasks, e.g., semantic segmentation and depth estimation. In addition, DDP shows attractive properties such as dynamic inference and uncertainty awareness, in contrast to previous single-step discriminative methods. We show top results on three representative tasks with six diverse benchmarks, without tricks, DDP achieves state-of-the-art or competitive performance on each task compared to the specialist counterparts. For example, semantic segmentation (83.9 mIoU on Cityscapes), BEV map segmentation (70.6 mIoU on nuScenes), and depth estimation (0.05 REL on KITTI). We hope that our approach will serve as a solid baseline and facilitate future research."
                },
                {
                    "title": "DiffusionDepth: Diffusion Denoising Approach for Monocular Depth Estimation",
                    "abstract": "Monocular depth estimation is a challenging task that predicts the pixel-wise depth from a single 2D image. Current methods typically model this problem as a regression or classification task. We propose DiffusionDepth, a new approach that reformulates monocular depth estimation as a denoising diffusion process. It learns an iterative denoising process to `denoise' random depth distribution into a depth map with the guidance of monocular visual conditions. The process is performed in the latent space encoded by a dedicated depth encoder and decoder. Instead of diffusing ground truth (GT) depth, the model learns to reverse the process of diffusing the refined depth of itself into random depth distribution. This self-diffusion formulation overcomes the difficulty of applying generative models to sparse GT depth scenarios. The proposed approach benefits this task by refining depth estimation step by step, which is superior for generating accurate and highly detailed depth maps. Experimental results on KITTI and NYU-Depth-V2 datasets suggest that a simple yet efficient diffusion approach could reach state-of-the-art performance in both indoor and outdoor scenarios with acceptable inference time."
                },
                {
                    "title": "Monocular Depth Estimation using Diffusion Models",
                    "abstract": "We formulate monocular depth estimation using denoising diffusion models, inspired by their recent successes in high fidelity image generation. To that end, we introduce innovations to address problems arising due to noisy, incomplete depth maps in training data, including step-unrolled denoising diffusion, an $L_1$ loss, and depth infilling during training. To cope with the limited availability of data for supervised training, we leverage pre-training on self-supervised image-to-image translation tasks. Despite the simplicity of the approach, with a generic loss and architecture, our DepthGen model achieves SOTA performance on the indoor NYU dataset, and near SOTA results on the outdoor KITTI dataset. Further, with a multimodal posterior, DepthGen naturally represents depth ambiguity (e.g., from transparent surfaces), and its zero-shot performance combined with depth imputation, enable a simple but effective text-to-3D pipeline. Project page: https://depth-gen.github.io"
                },
                {
                    "title": "DAG: Depth-Aware Guidance with Denoising Diffusion Probabilistic Models",
                    "abstract": "Generative models have recently undergone significant advancement due to the diffusion models. The success of these models can be often attributed to their use of guidance techniques, such as classifier or classifier-free guidance, which provide effective mechanisms to trade-off between fidelity and diversity. However, these methods are not capable of guiding a generated image to be aware of its geometric configuration, e.g., depth, which hinders their application to areas that require a certain level of depth awareness. To address this limitation, we propose a novel guidance method for diffusion models that uses estimated depth information derived from the rich intermediate representations of diffusion models. We first present label-efficient depth estimation framework using internal representations of diffusion models. Subsequently, we propose the incorporation of two guidance techniques based on pseudo-labeling and depth-domain diffusion prior during the sampling phase to self-condition the generated image using the estimated depth map. Experiments and comprehensive ablation studies demonstrate the effectiveness of our method in guiding the diffusion models towards the generation of geometrically plausible images."
                },
                {
                    "title": "Diffusioninst: Diffusion Model for Instance Segmentation",
                    "abstract": "Diffusion frameworks have achieved comparable performance with previous state-of-the-art image generation models. This paper proposes DiffusionInst, a novel framework representing instances as vectors and formulates instance segmentation as a noise-to-vector denoising process. The model is trained to reverse the noisy groundtruth mask without any inductive bias from RPN. It takes a randomly generated vector as input and outputs mask with multi-step denoising during inference. Extensive experimental results on COCO and LVIS show that DiffusionInst achieves competitive performance. Our code is available at https://github.com/chenhaoxing/DiffusionInst."
                },
                {
                    "title": "DiffPose: Multi-hypothesis Human Pose Estimation using Diffusion Models",
                    "abstract": "Traditionally, monocular 3D human pose estimation employs a machine learning model to predict the most likely 3D pose for a given input image. However, a single image can be highly ambiguous and induces multiple plausible solutions for the 2D-3D lifting step, which results in overly confident 3D pose predictors. To this end, we propose DiffPose, a conditional diffusion model that predicts multiple hypotheses for a given input image. Compared to similar approaches, our diffusion model is straightforward and avoids intensive hyperparameter tuning, complex network structures, mode collapse, and unstable training. Moreover, we tackle the problem of over-simplification of the intermediate representation of the common two-step approaches which first estimate a distribution of 2D joint locations via joint-wise heatmaps and consecutively use their maximum argument for the 3D pose estimation step. Since such a simplification of the heatmaps removes valid information about possibly correct, though labeled unlikely, joint locations, we propose to represent the heatmaps as a set of 2D joint candidate samples. To extract information about the original distribution from these samples, we introduce our embedding transformer which conditions the diffusion model. Experimentally, we show that DiffPose improves upon the state of the art for multi-hypothesis pose estimation by 3-5% for simple poses and outperforms it by a large margin for highly ambiguous poses.1"
                },
                {
                    "title": "DiffusionDet: Diffusion Model for Object Detection",
                    "abstract": "We propose DiffusionDet, a new framework that formulates object detection as a denoising diffusion process from noisy boxes to object boxes. During the training stage, object boxes diffuse from ground-truth boxes to random distribution, and the model learns to reverse this noising process. In inference, the model refines a set of randomly generated boxes to the output results in a progressive way. Our work possesses an appealing property of flexibility, which enables the dynamic number of boxes and iterative evaluation. The extensive experiments on the standard benchmarks show that DiffusionDet achieves favorable performance compared to previous well-established detectors. For example, DiffusionDet achieves 5.3 AP and 4.8 AP gains when evaluated with more boxes and iteration steps, under a zero-shot transfer setting from COCO to CrowdHuman. Our code is available at https://github.com/ShoufaChen/DiffusionDet."
                },
                {
                    "title": "InstructPix2Pix: Learning to Follow Image Editing Instructions",
                    "abstract": "We propose a method for editing images from human instructions: given an input image and a written instruction that tells the model what to do, our model follows these instructions to edit the image. To obtain training data for this problem, we combine the knowledge of two large pretrained models\u2014a language model (GPT-3) and a text-to-image model (Stable Diffusion)\u2014to generate a large dataset of image editing examples. Our conditional diffusion model, InstructPix2Pix, is trained on our generated data, and generalizes to real images and user-written instructions at inference time. Since it performs edits in the forward pass and does not require per-example fine-tuning or inversion, our model edits images quickly, in a matter of seconds. We show compelling editing results for a diverse collection of input images and written instructions."
                },
                {
                    "title": "eDiff-I: Text-to-Image Diffusion Models with an Ensemble of Expert Denoisers",
                    "abstract": "Large-scale diffusion-based generative models have led to breakthroughs in text-conditioned high-resolution image synthesis. Starting from random noise, such text-to-image diffusion models gradually synthesize images in an iterative fashion while conditioning on text prompts. We find that their synthesis behavior qualitatively changes throughout this process: Early in sampling, generation strongly relies on the text prompt to generate text-aligned content, while later, the text conditioning is almost entirely ignored. This suggests that sharing model parameters throughout the entire generation process may not be ideal. Therefore, in contrast to existing works, we propose to train an ensemble of text-to-image diffusion models specialized for different synthesis stages. To maintain training efficiency, we initially train a single model, which is then split into specialized models that are trained for the specific stages of the iterative generation process. Our ensemble of diffusion models, called eDiff-I, results in improved text alignment while maintaining the same inference computation cost and preserving high visual quality, outperforming previous large-scale text-to-image diffusion models on the standard benchmark. In addition, we train our model to exploit a variety of embeddings for conditioning, including the T5 text, CLIP text, and CLIP image embeddings. We show that these different embeddings lead to different behaviors. Notably, the CLIP image embedding allows an intuitive way of transferring the style of a reference image to the target text-to-image output. Lastly, we show a technique that enables eDiff-I's\"paint-with-words\"capability. A user can select the word in the input text and paint it in a canvas to control the output, which is very handy for crafting the desired image in mind. The project page is available at https://deepimagination.cc/eDiff-I/"
                },
                {
                    "title": "A Generalist Framework for Panoptic Segmentation of Images and Videos",
                    "abstract": "Panoptic segmentation assigns semantic and instance ID labels to every pixel of an image. As permutations of instance IDs are also valid solutions, the task requires learning of high-dimensional one-to-many mapping. As a result, state-of-the-art approaches use customized architectures and task-specific loss functions. We formulate panoptic segmentation as a discrete data generation problem, without relying on inductive bias of the task. A diffusion model is proposed to model panoptic masks, with a simple architecture and generic loss function. By simply adding past predictions as a conditioning signal, our method is capable of modeling video (in a streaming setting) and thereby learns to track object instances automatically. With extensive experiments, we demonstrate that our simple approach can perform competitively to state-of-the-art specialist methods in similar settings. 1"
                },
                {
                    "title": "Human Motion Diffusion Model",
                    "abstract": "Natural and expressive human motion generation is the holy grail of computer animation. It is a challenging task, due to the diversity of possible motion, human perceptual sensitivity to it, and the difficulty of accurately describing it. Therefore, current generative solutions are either low-quality or limited in expressiveness. Diffusion models, which have already shown remarkable generative capabilities in other domains, are promising candidates for human motion due to their many-to-many nature, but they tend to be resource hungry and hard to control. In this paper, we introduce Motion Diffusion Model (MDM), a carefully adapted classifier-free diffusion-based generative model for the human motion domain. MDM is transformer-based, combining insights from motion generation literature. A notable design-choice is the prediction of the sample, rather than the noise, in each diffusion step. This facilitates the use of established geometric losses on the locations and velocities of the motion, such as the foot contact loss. As we demonstrate, MDM is a generic approach, enabling different modes of conditioning, and different generation tasks. We show that our model is trained with lightweight resources and yet achieves state-of-the-art results on leading benchmarks for text-to-motion and action-to-motion. https://guytevet.github.io/mdm-page/ ."
                },
                {
                    "title": "MIDMs: Matching Interleaved Diffusion Models for Exemplar-based Image Translation",
                    "abstract": "We present a novel method for exemplar-based image translation, called matching interleaved diffusion models (MIDMs). Most existing methods for this task were formulated as GAN-based matching-then-generation framework. However, in this framework, matching errors induced by the difficulty of semantic matching across cross-domain, e.g., sketch and photo, can be easily propagated to the generation step, which in turn leads to the degenerated results. Motivated by the recent success of diffusion models, overcoming the shortcomings of GANs, we incorporate the diffusion models to overcome these limitations. Specifically, we formulate a diffusion-based matching-and-generation framework that interleaves cross-domain matching and diffusion steps in the latent space by iteratively feeding the intermediate warp into the noising process and denoising it to generate a translated image. In addition, to improve the reliability of diffusion process, we design confidence-aware process using cycle-consistency to consider only confident regions during translation. Experimental results show that our MIDMs generate more plausible images than state-of-the-art methods."
                },
                {
                    "title": "Pyramidal Denoising Diffusion Probabilistic Models",
                    "abstract": "Recently, diffusion model have demonstrated impressive image generation performances, and have been extensively studied in various computer vision tasks. Unfortunately, training and evaluating diffusion models consume a lot of time and computational resources. To address this problem, here we present a novel pyramidal diffusion model that can generate high resolution images starting from much coarser resolution images using a {\\em single} score function trained with a positional embedding. This enables a neural network to be much lighter and also enables time-efficient image generation without compromising its performances. Furthermore, we show that the proposed approach can be also efficiently used for multi-scale super-resolution problem using a single score function."
                },
                {
                    "title": "Few-Shot Diffusion Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) are powerful hierarchical latent variable models with remarkable sample generation quality and training stability. These properties can be attributed to parameter sharing in the generative hierarchy, as well as a parameter-free diffusion-based inference procedure. In this paper, we present Few-Shot Diffusion Models (FSDM), a framework for few-shot generation leveraging conditional DDPMs. FSDMs are trained to adapt the generative process conditioned on a small set of images from a given class by aggregating image patch information using a set-based Vision Transformer (ViT). At test time, the model is able to generate samples from previously unseen classes conditioned on as few as 5 samples from that class. We empirically show that FSDM can perform few-shot generation and transfer to new datasets. We benchmark variants of our method on complex vision datasets for few-shot learning and compare to unconditional and conditional DDPM baselines. Additionally, we show how conditioning the model on patch-based input set information improves training convergence."
                },
                {
                    "title": "CATs++: Boosting Cost Aggregation With Convolutions and Transformers",
                    "abstract": "Cost aggregation is a process in image matching tasks that aims to disambiguate the noisy matching scores. Existing methods generally tackle this by hand-crafted or CNN-based methods, which either lack robustness to severe deformations or inherit the limitation of CNNs that fail to discriminate incorrect matches due to limited receptive fields and inadaptability. In this paper, we introduce Cost Aggregation with Transformers (CATs) to tackle this by exploring global consensus among initial correlation map with the help of some architectural designs that allow us to benefit from global receptive fields of self-attention mechanism. To this end, we include appearance affinity modeling, which helps to disambiguate the noisy initial correlation maps. Furthermore, we introduce some techniques, including multi-level aggregation to exploit rich semantics prevalent at different feature levels and swapping self-attention to obtain reciprocal matching scores to act as a regularization. Although CATs can attain competitive performance, it may face some limitations, i.e., high computational costs, which may restrict its applicability only at limited resolution and hurt performance. To overcome this, we propose CATs++, an extension of CATs. Concretely, we introduce early convolutions prior to cost aggregation with a transformer to control the number of tokens and inject some convolutional inductive bias, then propose a novel transformer architecture for both efficient and effective cost aggregation, which results in apparent performance boost and cost reduction. With the reduced costs, we are able to compose our network with a hierarchical structure to process higher-resolution inputs. We show that the proposed method with these integrated outperforms the previous state-of-the-art methods by large margins. Codes and pretrained weights are available at: https://ku-cvlab.github.io/CATs-PlusPlus-Project-Page/."
                },
                {
                    "title": "DKM: Dense Kernelized Feature Matching for Geometry Estimation",
                    "abstract": "Feature matching is a challenging computer vision task that involves finding correspondences between two images of a 3D scene. In this paper we consider the dense approach instead of the more common sparse paradigm, thus striving to find all correspondences. Perhaps counter-intuitively, dense methods have previously shown inferior performance to their sparse and semi-sparse counterparts for estimation of two-view geometry. This changes with our novel dense method, which outperforms both dense and sparse methods on geometry estimation. The novelty is threefold: First, we propose a kernel regression global matcher. Secondly, we propose warp refinement through stacked feature maps and depthwise convolution kernels. Thirdly, we propose learning dense confidence through consistent depth and a balanced sampling approach for dense confidence maps. Through extensive experiments we confirm that our proposed dense method, Dense Kernelized Feature Matching, sets a new state-of-the-art on multiple geometry estimation benchmarks. In particular, we achieve an improvement on MegaDepth-1500 of +4.9 and +8.9 AUC@5\u00b0 compared to the best previous sparse method and dense method respectively. Our code is provided at the following repository: https://github.com/Parskatt/DKM."
                },
                {
                    "title": "RePaint: Inpainting using Denoising Diffusion Probabilistic Models",
                    "abstract": "Free-form inpainting is the task of adding new content to an image in the regions specified by an arbitrary binary mask. Most existing approaches train for a certain distribution of masks, which limits their generalization capabilities to unseen mask types. Furthermore, training with pixel-wise and perceptual losses often leads to simple textural extensions towards the missing areas instead of semantically meaningful generation. In this work, we propose RePaint: A Denoising Diffusion Probabilistic Model (DDPM) based inpainting approach that is applicable to even extreme masks. We employ a pretrained unconditional DDPM as the generative prior. To condition the generation process, we only alter the reverse diffusion iterations by sampling the unmasked regions using the given image infor-mation. Since this technique does not modify or condition the original DDPM network itself, the model produces high-quality and diverse output images for any inpainting form. We validate our method for both faces and general-purpose image inpainting using standard and extreme masks. Re-Paint outperforms state-of-the-art Autoregressive, and GAN approaches for at least five out of six mask distributions. Github Repository: git.io/RePaint"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Conditional Image Generation with Score-Based Diffusion Models",
                    "abstract": "Score-based diffusion models have emerged as one of the most promising frameworks for deep generative modelling. In this work we conduct a systematic comparison and theoretical analysis of different approaches to learning conditional probability distributions with score-based diffusion models. In particular, we prove results which provide a theoretical justification for one of the most successful estimators of the conditional score. Moreover, we introduce a multi-speed diffusion framework, which leads to a new estimator for the conditional score, performing on par with previous state-of-the-art approaches. Our theoretical and experimental findings are accompanied by an open source library MSDiff which allows for application and further research of multi-speed diffusion models."
                },
                {
                    "title": "Palette: Image-to-Image Diffusion Models",
                    "abstract": "This paper develops a unified framework for image-to-image translation based on conditional diffusion models and evaluates this framework on four challenging image-to-image translation tasks, namely colorization, inpainting, uncropping, and JPEG restoration. Our simple implementation of image-to-image diffusion models outperforms strong GAN and regression baselines on all tasks, without task-specific hyper-parameter tuning, architecture customization, or any auxiliary loss or sophisticated new techniques needed. We uncover the impact of an L2 vs. L1 loss in the denoising diffusion objective on sample diversity, and demonstrate the importance of self-attention in the neural architecture through empirical studies. Importantly, we advocate a unified evaluation protocol based on ImageNet, with human evaluation and sample quality scores (FID, Inception Score, Classification Accuracy of a pre-trained ResNet-50, and Perceptual Distance against original images). We expect this standardized evaluation protocol to play a role in advancing image-to-image translation research. Finally, we show that a generalist, multi-task diffusion model performs as well or better than task-specific specialist counterparts. Check out https://diffusion-palette.github.io/ for an overview of the results and code."
                },
                {
                    "title": "PDC-Net+: Enhanced Probabilistic Dense Correspondence Network",
                    "abstract": "Establishing robust and accurate correspondences between a pair of images is a long-standing computer vision problem with numerous applications. While classically dominated by sparse methods, emerging dense approaches offer a compelling alternative paradigm that avoids the keypoint detection step. However, dense flow estimation is often inaccurate in the case of large displacements, occlusions, or homogeneous regions. In order to apply dense methods to real-world applications, such as pose estimation, image manipulation, or 3D reconstruction, it is therefore crucial to estimate the confidence of the predicted matches. We propose the Enhanced Probabilistic Dense Correspondence Network, PDC-Net+, capable of estimating accurate dense correspondences along with a reliable confidence map. We develop a flexible probabilistic approach that jointly learns the flow prediction and its uncertainty. In particular, we parametrize the predictive distribution as a constrained mixture model, ensuring better modelling of both accurate flow predictions and outliers. Moreover, we develop an architecture and an enhanced training strategy tailored for robust and generalizable uncertainty prediction in the context of self-supervised training. Our approach obtains state-of-the-art results on multiple challenging geometric matching and optical flow datasets. We further validate the usefulness of our probabilistic confidence estimation for the tasks of pose estimation, 3D reconstruction, image-based localization, and image retrieval."
                },
                {
                    "title": "Deep Matching Prior: Test-Time Optimization for Dense Correspondence",
                    "abstract": "Conventional techniques to establish dense correspondences across visually or semantically similar images focused on designing a task-specific matching prior, which is difficult to model in general. To overcome this, recent learning-based methods have attempted to learn a good matching prior within a model itself on large training data. The performance improvement was apparent, but the need for sufficient training data and intensive learning hinders their applicability. Moreover, using the fixed model at test time does not account for the fact that a pair of images may require their own prior, thus providing limited performance and poor generalization to unseen images.In this paper, we show that an image pair-specific prior can be captured by solely optimizing the untrained matching networks on an input pair of images. Tailored for such test-time optimization for dense correspondence, we present a residual matching network and a confidence-aware contrastive loss to guarantee a meaningful convergence. Experiments demonstrate that our framework, dubbed Deep Matching Prior (DMP), is competitive, or even outperforms, against the latest learning-based methods on several benchmarks for geometric matching and semantic matching, even though it requires neither large training data nor intensive learning. With the networks pre-trained, DMP attains state-of-the-art performance on all benchmarks."
                },
                {
                    "title": "CATs: Cost Aggregation Transformers for Visual Correspondence",
                    "abstract": "We propose a novel cost aggregation network, called Cost Aggregation Transformers (CATs), to find dense correspondences between semantically similar images with additional challenges posed by large intra-class appearance and geometric variations. Cost aggregation is a highly important process in matching tasks, which the matching accuracy depends on the quality of its output. Compared to hand-crafted or CNN-based methods addressing the cost aggregation, in that either lacks robustness to severe deformations or inherit the limitation of CNNs that fail to discriminate incorrect matches due to limited receptive fields, CATs explore global consensus among initial correlation map with the help of some architectural designs that allow us to fully leverage self-attention mechanism. Specifically, we include appearance affinity modeling to aid the cost aggregation process in order to disambiguate the noisy initial correlation maps and propose multi-level aggregation to efficiently capture different semantics from hierarchical feature representations. We then combine with swapping self-attention technique and residual connections not only to enforce consistent matching but also to ease the learning process, which we find that these result in an apparent performance boost. We conduct experiments to demonstrate the effectiveness of the proposed model over the latest methods and provide extensive ablation studies. Project page is available at : https://sunghwanhong.github.io/CATs/."
                },
                {
                    "title": "PatchMatch-Based Neighborhood Consensus for Semantic Correspondence",
                    "abstract": "We address estimating dense correspondences between two images depicting different but semantically related scenes. End-to-end trainable deep neural networks incorporating neighborhood consensus cues are currently the best methods for this task. However, these architectures require exhaustive matching and 4D convolutions over matching costs for all pairs of feature map pixels. This makes them computationally expensive. We present a more efficient neighborhood consensus approach based on PatchMatch. For higher accuracy, we propose to use a learned local 4D scoring function for evaluating candidates during the PatchMatch iterations. We have devised an approach to jointly train the scoring function and the feature extraction modules by embedding them into a proxy model which is end-to-end differentiable. The modules are trained in a supervised setting using a cross-entropy loss to directly incorporate sparse keypoint supervision. Our evaluation on PF-Pascal and SPair-71K shows that our method significantly outperforms the state-of-the-art on both datasets while also being faster and using less memory."
                },
                {
                    "title": "Cascaded Diffusion Models for High Fidelity Image Generation",
                    "abstract": "We show that cascaded diffusion models are capable of generating high fidelity images on the class-conditional ImageNet generation benchmark, without any assistance from auxiliary image classifiers to boost sample quality. A cascaded diffusion model comprises a pipeline of multiple diffusion models that generate images of increasing resolution, beginning with a standard diffusion model at the lowest resolution, followed by one or more super-resolution diffusion models that successively upsample the image and add higher resolution details. We find that the sample quality of a cascading pipeline relies crucially on conditioning augmentation, our proposed method of data augmentation of the lower resolution conditioning inputs to the super-resolution models. Our experiments show that conditioning augmentation prevents compounding error during sampling in a cascaded model, helping us to train cascading pipelines achieving FID scores of 1.48 at 64x64, 3.52 at 128x128 and 4.88 at 256x256 resolutions, outperforming BigGAN-deep, and classification accuracy scores of 63.02% (top-1) and 84.06% (top-5) at 256x256, outperforming VQ-VAE-2."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "Image Super-Resolution via Iterative Refinement",
                    "abstract": "We present SR3, an approach to image Super-Resolution via Repeated Refinement. SR3 adapts denoising diffusion probabilistic models (Ho et al. 2020), (Sohl-Dickstein et al. 2015) to image-to-image translation, and performs super-resolution through a stochastic iterative denoising process. Output images are initialized with pure Gaussian noise and iteratively refined using a U-Net architecture that is trained on denoising at various noise levels, conditioned on a low-resolution input image. SR3 exhibits strong performance on super-resolution tasks at different magnification factors, on faces and natural images. We conduct human evaluation on a standard 8\u00d7 face super-resolution task on CelebA-HQ for which SR3 achieves a fool rate close to 50%, suggesting photo-realistic outputs, while GAN baselines do not exceed a fool rate of 34%. We evaluate SR3 on a 4\u00d7 super-resolution task on ImageNet, where SR3 outperforms baselines in human evaluation and classification accuracy of a ResNet-50 classifier trained on high-resolution images. We further show the effectiveness of SR3 in cascaded image generation, where a generative model is chained with super-resolution models to synthesize high-resolution images with competitive FID scores on the class-conditional 256\u00d7256 ImageNet generation challenge."
                },
                {
                    "title": "Convolutional Hough Matching Networks",
                    "abstract": "Despite advances in feature representation, leveraging geometric relations is crucial for establishing reliable visual correspondences under large variations of images. In this work we introduce a Hough transform perspective on convolutional matching and propose an effective geometric matching algorithm, dubbed Convolutional Hough Matching (CHM). The method distributes similarities of candidate matches over a geometric transformation space and evaluate them in a convolutional manner. We cast it into a trainable neural layer with a semi-isotropic high-dimensional kernel, which learns non-rigid matching with a small number of interpretable parameters. To validate the effect, we develop the neural network with CHM layers that perform convolutional matching in the space of translation and scaling. Our method sets a new state of the art on standard benchmarks for semantic visual correspondence, proving its strong robustness to challenging intra-class variations."
                },
                {
                    "title": "COTR: Correspondence Transformer for Matching Across Images",
                    "abstract": "We propose a novel framework for finding correspondences in images based on a deep neural network that, given two images and a query point in one of them, finds its correspondence in the other. By doing so, one has the option to query only the points of interest and retrieve sparse correspondences, or to query all points in an image and obtain dense mappings. Importantly, in order to capture both local and global priors, and to let our model relate between image regions using the most relevant among said priors, we realize our network using a transformer. At inference time, we apply our correspondence network by recursively zooming in around the estimates, yielding a multiscale pipeline able to provide highly-accurate correspondences. Our method significantly outperforms the state of the art on both sparse and dense correspondence problems on multiple datasets and tasks, ranging from wide-baseline stereo to optical flow, without any retraining for a specific dataset. We commit to releasing data, code, and all the tools necessary to train from scratch and ensure reproducibility."
                },
                {
                    "title": "Improved Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) are a class of generative models which have recently been shown to produce excellent samples. We show that with a few simple modifications, DDPMs can also achieve competitive log-likelihoods while maintaining high sample quality. Additionally, we find that learning variances of the reverse diffusion process allows sampling with an order of magnitude fewer forward passes with a negligible difference in sample quality, which is important for the practical deployment of these models. We additionally use precision and recall to compare how well DDPMs and GANs cover the target distribution. Finally, we show that the sample quality and likelihood of these models scale smoothly with model capacity and training compute, making them easily scalable. We release our code at https://github.com/openai/improved-diffusion"
                },
                {
                    "title": "Learning Accurate Dense Correspondences and When to Trust Them",
                    "abstract": "Establishing dense correspondences between a pair of images is an important and general problem. However, dense flow estimation is often inaccurate in the case of large displacements or homogeneous regions. For most applications and down-stream tasks, such as pose estimation, image manipulation, or 3D reconstruction, it is crucial to know when and where to trust the estimated matches.In this work, we aim to estimate a dense flow field relating two images, coupled with a robust pixel-wise confidence map indicating the reliability and accuracy of the prediction. We develop a flexible probabilistic approach that jointly learns the flow prediction and its uncertainty. In particular, we parametrize the predictive distribution as a constrained mixture model, ensuring better modelling of both accurate flow predictions and outliers. Moreover, we develop an architecture and training strategy tailored for robust and generalizable uncertainty prediction in the context of self-supervised training. Our approach obtains state- of-the-art results on multiple challenging geometric matching and optical flow datasets. We further validate the usefulness of our probabilistic confidence estimation for the task of pose estimation. Code and models are available at https://github.com/PruneTruong/PDCNet."
                },
                {
                    "title": "Score-Based Generative Modeling through Stochastic Differential Equations",
                    "abstract": "Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "GOCor: Bringing Globally Optimized Correspondence Volumes into Your Neural Network",
                    "abstract": "The feature correlation layer serves as a key neural network module in numerous computer vision problems that involve dense correspondences between image pairs. It predicts a correspondence volume by evaluating dense scalar products between feature vectors extracted from pairs of locations in two images. However, this point-to-point feature comparison is insufficient when disambiguating multiple similar regions in an image, severely affecting the performance of the end task. We propose GOCor, a fully differentiable dense matching module, acting as a direct replacement to the feature correlation layer. The correspondence volume generated by our module is the result of an internal optimization procedure that explicitly accounts for similar regions in the scene. Moreover, our approach is capable of effectively learning spatial matching priors to resolve further matching ambiguities. We analyze our GOCor module in extensive ablative experiments. When integrated into state-of-the-art networks, our approach significantly outperforms the feature correlation layer for the tasks of geometric matching, optical flow, and dense semantic matching. The code and trained models will be made available at this http URL."
                },
                {
                    "title": "NCNet: Neighbourhood Consensus Networks for Estimating Image Correspondences",
                    "abstract": "We address the problem of finding reliable dense correspondences between a pair of images. This is a challenging task due to strong appearance differences between the corresponding scene elements and ambiguities generated by repetitive patterns. The contributions of this work are threefold. First, inspired by the classic idea of disambiguating feature matches using semi-local constraints, we develop an end-to-end trainable convolutional neural network architecture that identifies sets of spatially consistent matches by analyzing neighbourhood consensus patterns in the 4D space of all possible correspondences between a pair of images without the need for a global geometric model. Second, we demonstrate that the model can be trained effectively from weak supervision in the form of matching and non-matching image pairs without the need for costly manual annotation of point to point correspondences. Third, we show the proposed neighbourhood consensus network can be applied to a range of matching tasks including both category- and instance-level matching, obtaining the state-of-the-art results on the PF, TSS, InLoc, and HPatches benchmarks."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Cross-Domain Correspondence Learning for Exemplar-Based Image Translation",
                    "abstract": "We present a general framework for exemplar-based image translation, which synthesizes a photo-realistic image from the input in a distinct domain (e.g., semantic segmentation mask, or edge map, or pose keypoints), given an exemplar image. The output has the style (e.g., color, texture) in consistency with the semantically corresponding objects in the exemplar. We propose to jointly learn the cross-domain correspondence and the image translation, where both tasks facilitate each other and thus can be learned with weak supervision. The images from distinct domains are first aligned to an intermediate domain where dense correspondence is established. Then, the network synthesizes images based on the appearance of semantically corresponding patches in the exemplar. We demonstrate the effectiveness of our approach in several image translation tasks. Our method is superior to state-of-the-art methods in terms of image quality significantly, with the image style faithful to the exemplar with semantic consistency. Moreover, we show the utility of our method for several applications."
                },
                {
                    "title": "GLU-Net: Global-Local Universal Network for Dense Flow and Correspondences",
                    "abstract": "Establishing dense correspondences between a pair of images is an important and general problem, covering geometric matching, optical flow and semantic correspondences. While these applications share fundamental challenges, such as large displacements, pixel-accuracy, and appearance changes, they are currently addressed with specialized network architectures, designed for only one particular task. This severely limits the generalization capabilities of such networks to new scenarios, where e.g. robustness to larger displacements or higher accuracy is required. In this work, we propose a universal network architecture that is directly applicable to all the aforementioned dense correspondence problems. We achieve both high accuracy and robustness to large displacements by investigating the combined use of global and local correlation layers. We further propose an adaptive resolution strategy, allowing our network to operate on virtually any input image resolution. The proposed GLU-Net achieves state-of-the-art performance for geometric and semantic matching as well as optical flow, when using the same network and weights. Code and trained models are available at https://github.com/PruneTruong/GLU-Net."
                },
                {
                    "title": "Analyzing and Improving the Image Quality of StyleGAN",
                    "abstract": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign the generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent codes to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably attribute a generated image to a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "SuperGlue: Learning Feature Matching With Graph Neural Networks",
                    "abstract": "This paper introduces SuperGlue, a neural network that matches two sets of local features by jointly finding correspondences and rejecting non-matchable points. Assignments are estimated by solving a differentiable optimal transport problem, whose costs are predicted by a graph neural network. We introduce a flexible context aggregation mechanism based on attention, enabling SuperGlue to reason about the underlying 3D scene and feature assignments jointly. Compared to traditional, hand-designed heuristics, our technique learns priors over geometric transformations and regularities of the 3D world through end-to-end training from image pairs. SuperGlue outperforms other learned approaches and achieves state-of-the-art results on the task of pose estimation in challenging real-world indoor and outdoor environments. The proposed method performs matching in real-time on a modern GPU and can be readily integrated into modern SfM or SLAM systems. The code and trained weights are publicly available at github.com/magicleap/SuperGluePretrainedNetwork."
                },
                {
                    "title": "Self-Supervised Learning of Depth and Motion Under Photometric Inconsistency",
                    "abstract": "The self-supervised learning of depth and pose from monocular sequences provides an attractive solution by using the photometric consistency of nearby frames as it depends much less on the ground-truth data. In this paper, we address the issue when previous assumptions of the self-supervised approaches are violated due to the dynamic nature of real-world scenes. Different from handling the noise as uncertainty, our key idea is to incorporate more robust geometric quantities and enforce internal consistency in the temporal image sequence. As demonstrated on commonly used benchmark datasets, the proposed method substantially improves the state-of-the-art methods on both depth and relative pose estimation for monocular image sequences, without adding inference overhead."
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "Benchmarking Neural Network Robustness to Common Corruptions and Perturbations",
                    "abstract": "In this paper we establish rigorous benchmarks for image classifier robustness. Our first benchmark, ImageNet-C, standardizes and expands the corruption robustness topic, while showing which classifiers are preferable in safety-critical applications. Then we propose a new dataset called ImageNet-P which enables researchers to benchmark a classifier's robustness to common perturbations. Unlike recent robustness research, this benchmark evaluates performance on common corruptions and perturbations not worst-case adversarial perturbations. We find that there are negligible changes in relative corruption robustness from AlexNet classifiers to ResNet classifiers. Afterward we discover ways to enhance corruption and perturbation robustness. We even find that a bypassed adversarial defense provides substantial common perturbation robustness. Together our benchmarks may aid future work toward networks that robustly generalize."
                },
                {
                    "title": "Recurrent Transformer Networks for Semantic Correspondence",
                    "abstract": "We present recurrent transformer networks (RTNs) for obtaining dense correspondences between semantically similar images. Our networks accomplish this through an iterative process of estimating spatial transformations between the input images and using these transformations to generate aligned convolutional activations. By directly estimating the transformations between an image pair, rather than employing spatial transformer networks to independently normalize each individual image, we show that greater accuracy can be achieved. This process is conducted in a recursive manner to refine both the transformation estimates and the feature representations. In addition, a technique is presented for weakly-supervised training of RTNs that is based on a proposed classification loss. With RTNs, state-of-the-art performance is attained on several benchmarks for semantic correspondence."
                },
                {
                    "title": "DGC-Net: Dense Geometric Correspondence Network",
                    "abstract": "This paper addresses the challenge of dense pixel correspondence estimation between two images. This problem is closely related to optical flow estimation task where ConvNets (CNNs) have recently achieved significant progress. While optical flow methods produce very accurate results for the small pixel translation and limited appearance variation scenarios, they hardly deal with the strong geometric transformations that we consider in this work. In this paper, we propose a coarse-to-fine CNN-based framework that can leverage the advantages of optical flow approaches and extend them to the case of large transformations providing dense and subpixel accurate estimates. It is trained on synthetic transformations and demonstrates very good performance to unseen, realistic, data. Further, we apply our method to the problem of relative camera pose estimation and demonstrate that the model outperforms existing dense approaches."
                },
                {
                    "title": "MegaDepth: Learning Single-View Depth Prediction from Internet Photos",
                    "abstract": "Single-view depth prediction is a fundamental problem in computer vision. Recently, deep learning methods have led to significant progress, but such methods are limited by the available training data. Current datasets based on 3D sensors have key limitations, including indoor-only images (NYU), small numbers of training examples (Make3D), and sparse sampling (KITTI). We propose to use multi-view Internet photo collections, a virtually unlimited data source, to generate training data via modern structure-from-motion and multi-view stereo (MVS) methods, and present a large depth dataset called MegaDepth based on this idea. Data derived from MVS comes with its own challenges, including noise and unreconstructable objects. We address these challenges with new data cleaning methods, as well as automatically augmenting our data with ordinal depth relations generated using semantic segmentation. We validate the use of large amounts of Internet data by showing that models trained on MegaDepth exhibit strong generalization-not only to novel scenes, but also to other diverse datasets including Make3D, KITTI, and DIW, even when no images from those datasets are seen during training.1"
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume",
                    "abstract": "We present a compact but effective CNN model for optical flow, called PWC-Net. PWC-Net has been designed according to simple and well-established principles: pyramidal processing, warping, and the use of a cost volume. Cast in a learnable feature pyramid, PWC-Net uses the current optical flow estimate to warp the CNN features of the second image. It then uses the warped features and features of the first image to construct a cost volume, which is processed by a CNN to estimate the optical flow. PWC-Net is 17 times smaller in size and easier to train than the recent FlowNet2 model. Moreover, it outperforms all published optical flow methods on the MPI Sintel final pass and KITTI 2015 benchmarks, running at about 35 fps on Sintel resolution (1024 \u00c3\u2014 436) images. Our models are available on our project website."
                },
                {
                    "title": "ProbFlow: Joint Optical Flow and Uncertainty Estimation",
                    "abstract": "Optical flow estimation remains challenging due to untextured areas, motion boundaries, occlusions, and more. Thus, the estimated flow is not equally reliable across the image. To that end, post-hoc confidence measures have been introduced to assess the per-pixel reliability of the flow. We overcome the artificial separation of optical flow and confidence estimation by introducing a method that jointly predicts optical flow and its underlying uncertainty. Starting from common energy-based formulations, we rely on the corresponding posterior distribution of the flow given the images. We derive a variational inference scheme based on mean field, which incorporates best practices from energy minimization. An uncertainty measure is obtained along the flow at every pixel as the (marginal) entropy of the variational distribution. We demonstrate the flexibility of our probabilistic approach by applying it to two different energies and on two benchmarks. We not only obtain flow results that are competitive with the underlying energy minimization approach, but also a reliable uncertainty measure that significantly outperforms existing post-hoc approaches."
                },
                {
                    "title": "A Multi-view Stereo Benchmark with High-Resolution Images and Multi-camera Videos",
                    "abstract": "Motivated by the limitations of existing multi-view stereo benchmarks, we present a novel dataset for this task. Towards this goal, we recorded a variety of indoor and outdoor scenes using a high-precision laser scanner and captured both high-resolution DSLR imagery as well as synchronized low-resolution stereo videos with varying fields-of-view. To align the images with the laser scans, we propose a robust technique which minimizes photometric errors conditioned on the geometry. In contrast to previous datasets, our benchmark provides novel challenges and covers a diverse set of viewpoints and scene types, ranging from natural scenes to man-made indoor and outdoor environments. Furthermore, we provide data at significantly higher temporal and spatial resolution. Our benchmark is the first to cover the important use case of hand-held mobile devices while also providing high-resolution DSLR camera images. We make our datasets and an online evaluation server available at http://www.eth3d.net."
                },
                {
                    "title": "DCTM: Discrete-Continuous Transformation Matching for Semantic Flow",
                    "abstract": "Techniques for dense semantic correspondence have provided limited ability to deal with the geometric variations that commonly exist between semantically similar images. While variations due to scale and rotation have been examined, there is a lack of practical solutions for more complex deformations such as affine transformations because of the tremendous size of the associated solution space. To address this problem, we present a discrete-continuous transformation matching (DCTM) framework where dense affine transformation fields are inferred through a discrete label optimization in which the labels are iteratively updated via continuous regularization. In this way, our approach draws solutions from the continuous space of affine transformations in a manner that can be computed efficiently through constant-time edge-aware filtering and a proposed affine-varying CNN-based descriptor. Experimental results show that this model outperforms the state-of-the-art methods for dense semantic correspondence on various benchmarks."
                },
                {
                    "title": "HPatches: A Benchmark and Evaluation of Handcrafted and Learned Local Descriptors",
                    "abstract": "In this paper, we propose a novel benchmark for evaluating local image descriptors. We demonstrate that the existing datasets and evaluation protocols do not specify unambiguously all aspects of evaluation, leading to ambiguities and inconsistencies in results reported in the literature. Furthermore, these datasets are nearly saturated due to the recent improvements in local descriptors obtained by learning them from large annotated datasets. Therefore, we introduce a new large dataset suitable for training and testing modern descriptors, together with strictly defined evaluation protocols in several tasks such as matching, retrieval and classification. This allows for more realistic, and thus more reliable comparisons in different application scenarios. We evaluate the performance of several state-of-the-art descriptors and analyse their properties. We show that a simple normalisation of traditional hand-crafted descriptors can boost their performance to the level of deep learning based descriptors within a realistic benchmarks evaluation."
                },
                {
                    "title": "DSLR-Quality Photos on Mobile Devices with Deep Convolutional Networks",
                    "abstract": "Despite a rapid rise in the quality of built-in smartphone cameras, their physical limitations \u2013 small sensor size, compact lenses and the lack of specific hardware, \u2013 impede them to achieve the quality results of DSLR cameras. In this work we present an end-to-end deep learning approach that bridges this gap by translating ordinary photos into DSLR-quality images. We propose learning the translation function using a residual convolutional neural network that improves both color rendition and image sharpness. Since the standard mean squared loss is not well suited for measuring perceptual image quality, we introduce a composite perceptual error function that combines content, color and texture losses. The first two losses are defined analytically, while the texture loss is learned in an adversarial fashion. We also present DPED, a large-scale dataset that consists of real photos captured from three different phones and one high-end reflex camera. Our quantitative and qualitative assessments reveal that the enhanced image quality is comparable to that of DSLR-taken photos, while the methodology is generalized to any type of digital camera."
                },
                {
                    "title": "Convolutional Neural Network Architecture for Geometric Matching",
                    "abstract": "We address the problem of determining correspondences between two images in agreement with a geometric model such as an affine or thin-plate spline transformation, and estimating its parameters. The contributions of this work are three-fold. First, we propose a convolutional neural network architecture for geometric matching. The architecture is based on three main components that mimic the standard steps of feature extraction, matching and simultaneous inlier detection and model parameter estimation, while being trainable end-to-end. Second, we demonstrate that the network parameters can be trained from synthetically generated imagery without the need for manual annotation and that our matching layer significantly increases generalization capabilities to never seen before images. Finally, we show that the same model can perform both instance-level and category-level matching giving state-of-the-art results on the challenging Proposal Flow dataset."
                },
                {
                    "title": "FCSS: Fully Convolutional Self-Similarity for Dense Semantic Correspondence",
                    "abstract": "We present a descriptor, called fully convolutional self-similarity (FCSS), for dense semantic correspondence. To robustly match points among different instances within the same object class, we formulate FCSS using local self-similarity (LSS) within a fully convolutional network. In contrast to existing CNN-based descriptors, FCSS is inherently insensitive to intra-class appearance variations because of its LSS-based structure, while maintaining the precise localization ability of deep neural networks. The sampling patterns of local structure and the self-similarity measure are jointly learned within the proposed network in an end-to-end and multi-scale manner. As training data for semantic correspondence is rather limited, we propose to leverage object candidate priors provided in existing image datasets and also correspondence consistency between object pairs to enable weakly-supervised learning. Experiments demonstrate that FCSS outperforms conventional handcrafted descriptors and CNN-based descriptors on various benchmarks."
                },
                {
                    "title": "DeMoN: Depth and Motion Network for Learning Monocular Stereo",
                    "abstract": "In this paper we formulate structure from motion as a learning problem. We train a convolutional network end-to-end to compute depth and camera motion from successive, unconstrained image pairs. The architecture is composed of multiple stacked encoder-decoder networks, the core part being an iterative network that is able to improve its own predictions. The network estimates not only depth and motion, but additionally surface normals, optical flow between the images and confidence of the matching. A crucial component of the approach is a training loss based on spatial relative differences. Compared to traditional two-frame structure from motion methods, results are more accurate and more robust. In contrast to the popular depth-from-single-image networks, DeMoN learns the concept of matching and, thus, better generalizes to structures not seen during training."
                },
                {
                    "title": "Optical Flow Estimation Using a Spatial Pyramid Network",
                    "abstract": "We learn to compute optical flow by combining a classical spatial-pyramid formulation with deep learning. This estimates large motions in a coarse-to-fine approach by warping one image of a pair at each pyramid level by the current flow estimate and computing an update to the flow. Instead of the standard minimization of an objective function at each pyramid level, we train one deep network per level to compute the flow update. Unlike the recent FlowNet approach, the networks do not need to deal with large motions, these are dealt with by the pyramid. This has several advantages. First, our Spatial Pyramid Network (SPyNet) is much simpler and 96% smaller than FlowNet in terms of model parameters. This makes it more efficient and appropriate for embedded applications. Second, since the flow at each pyramid level is small ("
                },
                {
                    "title": "Structure-from-Motion Revisited",
                    "abstract": "Incremental Structure-from-Motion is a prevalent strategy for 3D reconstruction from unordered image collections. While incremental reconstruction systems have tremendously advanced in all regards, robustness, accuracy, completeness, and scalability remain the key problems towards building a truly general-purpose pipeline. We propose a new SfM technique that improves upon the state of the art to make a further step towards this ultimate goal. The full reconstruction pipeline is released to the public as an open-source implementation."
                },
                {
                    "title": "Monocular 3D Object Detection for Autonomous Driving",
                    "abstract": "The goal of this paper is to perform 3D object detection from a single monocular image in the domain of autonomous driving. Our method first aims to generate a set of candidate class-specific object proposals, which are then run through a standard CNN pipeline to obtain high-quality object detections. The focus of this paper is on proposal generation. In particular, we propose an energy minimization approach that places object candidates in 3D using the fact that objects should be on the ground-plane. We then score each candidate box projected to the image plane via several intuitive potentials encoding semantic segmentation, contextual information, size and location priors and typical object shape. Our experimental evaluation demonstrates that our object proposal generation approach significantly outperforms all monocular approaches, and achieves the best detection performance on the challenging KITTI benchmark, among published monocular competitors."
                },
                {
                    "title": "Joint Recovery of Dense Correspondence and Cosegmentation in Two Images",
                    "abstract": "We propose a new technique to jointly recover cosegmentation and dense per-pixel correspondence in two images. Our method parameterizes the correspondence field using piecewise similarity transformations and recovers a mapping between the estimated common \"foreground\" regions in the two images allowing them to be precisely aligned. Our formulation is based on a hierarchical Markov random field model with segmentation and transformation labels. The hierarchical structure uses nested image regions to constrain inference across multiple scales. Unlike prior hierarchical methods which assume that the structure is given, our proposed iterative technique dynamically recovers the structure along with the labeling. This joint inference is performed in an energy minimization framework using iterated graph cuts. We evaluate our method on a new dataset of 400 image pairs with manually obtained ground truth, where it outperforms state-of-the-art methods designed specifically for either cosegmentation or correspondence estimation."
                },
                {
                    "title": "The Cityscapes Dataset for Semantic Urban Scene Understanding",
                    "abstract": "Visual understanding of complex urban street scenes is an enabling factor for a wide range of applications. Object detection has benefited enormously from large-scale datasets, especially in the context of deep learning. For semantic urban scene understanding, however, no current dataset adequately captures the complexity of real-world urban scenes. To address this, we introduce Cityscapes, a benchmark suite and large-scale dataset to train and test approaches for pixel-level and instance-level semantic labeling. Cityscapes is comprised of a large, diverse set of stereo video sequences recorded in streets from 50 different cities. 5000 of these images have high quality pixel-level annotations, 20 000 additional images have coarse annotations to enable methods that leverage large volumes of weakly-labeled data. Crucially, our effort exceeds previous attempts in terms of dataset size, annotation richness, scene variability, and complexity. Our accompanying empirical study provides an in-depth analysis of the dataset characteristics, as well as a performance evaluation of several state-of-the-art approaches based on our benchmark."
                },
                {
                    "title": "Proposal Flow",
                    "abstract": "Finding image correspondences remains a challenging problem in the presence of intra-class variations and large changes in scene layout. Semantic flow methods are designed to handle images depicting different instances of the same object or scene category. We introduce a novel approach to semantic flow, dubbed proposal flow, that establishes reliable correspondences using object proposals. Unlike prevailing semantic flow approaches that operate on pixels or regularly sampled local regions, proposal flow benefits from the characteristics of modern object proposals, that exhibit high repeatability at multiple scales, and can take advantage of both local and geometric consistency constraints among proposals. We also show that proposal flow can effectively be transformed into a conventional dense flow field. We introduce a new dataset that can be used to evaluate both general semantic flow techniques and region-based approaches such as proposal flow. We use this benchmark to compare different matching algorithms, object proposals, and region features within proposal flow, to the state of the art in semantic flow. This comparison, along with experiments on standard datasets, demonstrates that proposal flow significantly outperforms existing semantic flow methods in various settings."
                },
                {
                    "title": "FlowNet: Learning Optical Flow with Convolutional Networks",
                    "abstract": "Convolutional neural networks (CNNs) have recently been very successful in a variety of computer vision tasks, especially on those linked to recognition. Optical flow estimation has not been among the tasks CNNs succeeded at. In this paper we construct CNNs which are capable of solving the optical flow estimation problem as a supervised learning task. We propose and compare two architectures: a generic architecture and another one including a layer that correlates feature vectors at different image locations. Since existing ground truth data sets are not sufficiently large to train a CNN, we generate a large synthetic Flying Chairs dataset. We show that networks trained on this unrealistic data still generalize very well to existing datasets such as Sintel and KITTI, achieving competitive accuracy at frame rates of 5 to 10 fps."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "EpicFlow: Edge-preserving interpolation of correspondences for optical flow",
                    "abstract": "We propose a novel approach for optical flow estimation, targeted at large displacements with significant occlusions. It consists of two steps: (i) dense matching by edge-preserving interpolation from a sparse set of matches; (ii) variational energy minimization initialized with the dense matches. The sparse-to-dense interpolation relies on an appropriate choice of the distance, namely an edge-aware geodesic distance. This distance is tailored to handle occlusions and motion boundaries - two common and difficult issues for optical flow computation. We also propose an approximation scheme for the geodesic distance to allow fast computation without loss of performance. Subsequent to the dense interpolation step, standard one-level variational energy minimization is carried out on the dense matches to obtain the final flow estimation. The proposed approach, called Edge-Preserving Interpolation of Correspondences (EpicFlow) is fast and robust to large displacements. It significantly outperforms the state of the art on MPI-Sintel and performs on par on Kitti and Middlebury."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "Return of the Devil in the Details: Delving Deep into Convolutional Nets",
                    "abstract": "The latest generation of Convolutional Neural Networks (CNN) have achieved impressive results in challenging benchmarks on image recognition and object detection, significantly raising the interest of the community in these methods. Nevertheless, it is still unclear how different CNN methods compare with each other and with previous state-of-the-art shallow representations such as the Bag-of-Visual-Words and the Improved Fisher Vector. This paper conducts a rigorous evaluation of these new techniques, exploring different deep architectures and comparing them on a common ground, identifying and disclosing important implementation details. We identify several useful properties of CNN-based representations, including the fact that the dimensionality of the CNN output layer can be reduced significantly without having an adverse effect on performance. We also identify aspects of deep and shallow methods that can be successfully shared. In particular, we show that the data augmentation techniques commonly applied to CNN-based methods can also be applied to shallow methods, and result in an analogous performance boost. Source code and models to reproduce the experiments in the paper is made publicly available."
                },
                {
                    "title": "DeepFlow: Large Displacement Optical Flow with Deep Matching",
                    "abstract": "Optical flow computation is a key component in many computer vision systems designed for tasks such as action detection or activity recognition. However, despite several major advances over the last decade, handling large displacement in optical flow remains an open problem. Inspired by the large displacement optical flow of Brox and Malik, our approach, termed Deep Flow, blends a matching algorithm with a variational approach for optical flow. We propose a descriptor matching algorithm, tailored to the optical flow problem, that allows to boost performance on fast motions. The matching algorithm builds upon a multi-stage architecture with 6 layers, interleaving convolutions and max-pooling, a construction akin to deep convolutional nets. Using dense sampling, it allows to efficiently retrieve quasi-dense correspondences, and enjoys a built-in smoothing effect on descriptors matches, a valuable asset for integration into an energy minimization framework for optical flow estimation. Deep Flow efficiently handles large displacements occurring in realistic videos, and shows competitive performance on optical flow benchmarks. Furthermore, it sets a new state-of-the-art on the MPI-Sintel dataset."
                },
                {
                    "title": "TV-L1 Optical Flow Estimation",
                    "abstract": "This article describes an implementation of the optical flow estimation method introduced by Zach, Pock and Bischof in 2007. This method is based on the minimization of a functional containing a data term using the L 1 norm and a regularization term using the total variation of the flow. The main feature of this formulation is that it allows discontinuities in the flow field, while being more robust to noise than the classical approach by Horn and Schunck. The algorithm is an efficient numerical scheme, which solves a relaxed version of the problem by alternate minimization. Source Code A C implementation of this algorithm is provided. The source code and an online demo are accessible at the web page of this article 1 ."
                },
                {
                    "title": "Deformable Spatial Pyramid Matching for Fast Dense Correspondences",
                    "abstract": "We introduce a fast deformable spatial pyramid (DSP) matching algorithm for computing dense pixel correspondences. Dense matching methods typically enforce both appearance agreement between matched pixels as well as geometric smoothness between neighboring pixels. Whereas the prevailing approaches operate at the pixel level, we propose a pyramid graph model that simultaneously regularizes match consistency at multiple spatial extents-ranging from an entire image, to coarse grid cells, to every single pixel. This novel regularization substantially improves pixel-level matching in the face of challenging image variations, while the \"deformable\" aspect of our model overcomes the strict rigidity of traditional spatial pyramids. Results on Label Me and Caltech show our approach outperforms state-of-the-art methods (SIFT Flow [15] and Patch-Match [2]), both in terms of accuracy and run time."
                },
                {
                    "title": "Total variation regularization of local-global optical flow",
                    "abstract": "More data fidelity terms in variational optical flow methods improve the estimation's robustness. A robust and anisotropic smoother enhances the specific fill-in process. This work presents a combined local-global (CLG) approach with total variation regularization. The combination of bilateral filtering and anisotropic (image driven) regularization is used to control the propagation phenomena. The resulted method, CLG-TV, is able to compute larger displacements in a reasonable time. The numerical scheme is highly parallelizable and runs in real-time on current generation graphic processing units."
                },
                {
                    "title": "A graph-matching kernel for object categorization",
                    "abstract": "This paper addresses the problem of category-level image classification. The underlying image model is a graph whose nodes correspond to a dense set of regions, and edges reflect the underlying grid structure of the image and act as springs to guarantee the geometric consistency of nearby regions during matching. A fast approximate algorithm for matching the graphs associated with two images is presented. This algorithm is used to construct a kernel appropriate for SVM-based image classification, and experiments with the Caltech 101, Caltech 256, and Scenes datasets demonstrate performance that matches or exceeds the state of the art for methods using a single type of features."
                },
                {
                    "title": "SIFT Flow: Dense Correspondence across Scenes and Its Applications",
                    "abstract": "While image alignment has been studied in different areas of computer vision for decades, aligning images depicting different scenes remains a challenging problem. Analogous to optical flow, where an image is aligned to its temporally adjacent frame, we propose SIFT flow, a method to align an image to its nearest neighbors in a large image corpus containing a variety of scenes. The SIFT flow algorithm consists of matching densely sampled, pixelwise SIFT features between two images while preserving spatial discontinuities. The SIFT features allow robust matching across different scene/object appearances, whereas the discontinuity-preserving spatial model allows matching of objects located at different parts of the scene. Experiments show that the proposed approach robustly aligns complex scene pairs containing significant spatial differences. Based on SIFT flow, we propose an alignment-based large database framework for image analysis and synthesis, where image information is transferred from the nearest neighbors to a query image according to the dense scene correspondence. This framework is demonstrated through concrete applications such as motion field prediction from a single image, motion synthesis via object transfer, satellite image registration, and face recognition."
                },
                {
                    "title": "RepFinder: finding approximately repeated scene elements for image editing",
                    "abstract": "Repeated elements are ubiquitous and abundant in both manmade and natural scenes. Editing such images while preserving the repetitions and their relations is nontrivial due to overlap, missing parts, deformation across instances, illumination variation, etc. Manually enforcing such relations is laborious and error-prone. We propose a novel framework where user scribbles are used to guide detection and extraction of such repeated elements. Our detection process, which is based on a novel boundary band method, robustly extracts the repetitions along with their deformations. The algorithm only considers the shape of the elements, and ignores similarity based on color, texture, etc. We then use topological sorting to establish a partial depth ordering of overlapping repeated instances. Missing parts on occluded instances are completed using information from other instances. The extracted repeated instances can then be seamlessly edited and manipulated for a variety of high level tasks that are otherwise difficult to perform. We demonstrate the versatility of our framework on a large set of inputs of varying complexity, showing applications to image rearrangement, edit transfer, deformation propagation, and instance replacement."
                },
                {
                    "title": "Motion estimation with non-local total variation regularization",
                    "abstract": "State-of-the-art motion estimation algorithms suffer from three major problems: Poorly textured regions, occlusions and small scale image structures. Based on the Gestalt principles of grouping we propose to incorporate a low level image segmentation process in order to tackle these problems. Our new motion estimation algorithm is based on non-local total variation regularization which allows us to integrate the low level image segmentation process in a unified variational framework. Numerical results on the Mid-dlebury optical flow benchmark data set demonstrate that we can cope with the aforementioned problems."
                },
                {
                    "title": "Secrets of optical flow estimation and their principles",
                    "abstract": "The accuracy of optical flow estimation algorithms has been improving steadily as evidenced by results on the Middlebury optical flow benchmark. The typical formulation, however, has changed little since the work of Horn and Schunck. We attempt to uncover what has made recent advances possible through a thorough analysis of how the objective function, the optimization method, and modern implementation practices influence accuracy. We discover that \u201cclassical\u201d flow formulations perform surprisingly well when combined with modern optimization and implementation techniques. Moreover, we find that while median filtering of intermediate flow fields during optimization is a key to recent performance gains, it leads to higher energy solutions. To understand the principles behind this phenomenon, we derive a new objective that formalizes the median filtering heuristic. This objective includes a nonlocal term that robustly integrates flow estimates over large spatial neighborhoods. By modifying this new term to include information about flow and image boundaries we develop a method that ranks at the top of the Middlebury benchmark."
                },
                {
                    "title": "PatchMatch: a randomized correspondence algorithm for structural image editing",
                    "abstract": "This paper presents interactive image editing tools using a new randomized algorithm for quickly finding approximate nearest-neighbor matches between image patches. Previous research in graphics and vision has leveraged such nearest-neighbor searches to provide a variety of high-level digital image editing tools. However, the cost of computing a field of such matches for an entire image has eluded previous efforts to provide interactive performance. Our algorithm offers substantial performance improvements over the previous state of the art (20-100x), enabling its use in interactive editing tools. The key insights driving the algorithm are that some good patch matches can be found via random sampling, and that natural coherence in the imagery allows us to propagate such matches quickly to surrounding areas. We offer theoretical analysis of the convergence properties of the algorithm, as well as empirical and practical evidence for its high quality and performance. This one simple algorithm forms the basis for a variety of tools -- image retargeting, completion and reshuffling -- that can be used together in the context of a high-level image editing application. Finally, we propose additional intuitive constraints on the synthesis process that offer the user a level of control unavailable in previous methods."
                },
                {
                    "title": "Simultaneous localization and mapping (SLAM): part II",
                    "abstract": "This paper discusses the recursive Bayesian formulation of the simultaneous localization and mapping (SLAM) problem in which probability distributions or estimates of absolute or relative locations of landmarks and vehicle pose are obtained. The paper focuses on three key areas: computational complexity; data association; and environment representation"
                },
                {
                    "title": "Simultaneous localization and mapping: part I",
                    "abstract": "This paper describes the simultaneous localization and mapping (SLAM) problem and the essential methods for solving the SLAM problem and summarizes key implementations and demonstrations of the method. While there are still many practical issues to overcome, especially in more complex outdoor environments, the general SLAM method is now a well understood and established part of robotics. Another part of the tutorial summarized more recent works in addressing some of the remaining issues in SLAM, including computation, feature representation, and data association"
                },
                {
                    "title": "Robust dense matching using local and global geometric constraints",
                    "abstract": "A new robust dense matching algorithm is introduced. The algorithm starts from matching the most textured points, then a match propagation algorithm is developed with the best first strategy to dense matching. Next, the matching map is regularised by using the local geometric constraints encoded by planar affine applications and by using the global geometric constraint encoded by the fundamental matrix. Two most distinctive features are a match propagation strategy developed by analogy to region growing and a successive regularisation by local and global geometric constraints. The algorithm is efficient, robust and can cope with wide disparity. The algorithm is demonstrated on many real image pairs, and applications on image interpolation and a creation of novel views are also presented."
                },
                {
                    "title": "Probability distributions of optical flow",
                    "abstract": "Gradient methods are widely used in the computation of optical flow. The authors discuss extensions of these methods which compute probability distributions of optical flow. The use of distributions allows representation of the uncertainties inherent in the optical flow computation, facilitating the combination with information from other sources. Distributed optical flow for a synthetic image sequence is computed, and it is demonstrated that the probabilistic model accounts for the errors in the flow estimates. The distributed optical flow for a real image sequence is computed.<<ETX>>"
                },
                {
                    "title": "An Iterative Image Registration Technique with an Application to Stereo Vision",
                    "abstract": "Image registration finds a variety of applications in computer vision. Unfortunately, traditional image registration techniques tend to be costly. We present a new image registration technique that makes use of the spatial intensity gradient of the images to find a good match using a type of Newton-Raphson iteration. Our technique is taster because it examines far fewer potential matches between the images than existing techniques Furthermore, this registration technique can be generalized to handle rotation, scaling and shearing. We show how our technique can be adapted tor use in a stereo vision system."
                },
                {
                    "title": "Determining Optical Flow",
                    "abstract": "Optical flow cannot be computed locally, since only one independent measurement is available from the image sequence at a point, while the flow velocity has two components. A second constraint is needed. A method for finding the optical flow pattern is presented which assumes that the apparent velocity of the brightness pattern varies smoothly almost everywhere in the image. An iterative implementation is shown which successfully computes the optical flow for a number of synthetic image sequences. The algorithm is robust in that it can handle image sequences that are quantized rather coarsely in space and time. It is also insensitive to quantization of brightness levels and additive noise. Examples are included where the assumption of smoothness is violated at singular points or along lines in the image."
                },
                {
                    "title": "Bayes Theorem",
                    "abstract": "P(hjD) = P(Djh)P(h) P(D) P(h) = prior probability o f h ypothesis h P(D) = prior probability of training data D P(hjD) = probability o f h given D P(Djh) = probability o f D given h 127 lecture slides for textbook Machine Learning,"
                },
                {
                    "title": "Supplementary Material for LiteFlowNet: A Lightweight Convolutional Neural Network for Optical Flow Estimation",
                    "abstract": "Figures S1 to S3 show some examples of flow fields of different best performing networks on Sintel clean and final passes [2], KITTI12 [4], KITTI15 [7], and Middlebury [1] benchmarks. A video clip (https://www. youtube.com/watch?v=pfQ0zFwv-hM) is available on our project page (http://mmlab.ie.cuhk.edu. hk/projects/LiteFlowNet/) to showcase the performance of LiteFlowNet [5] and the effectiveness of the proposed components in our network. Flow fields of the compared methods, FlowNet1 [3], FlowNet22 [6] and SPyNet3 [8], are generated using the trained models and the code packages provided by the corresponding authors."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively model the matching field distribution for dense correspondence between pairs of images using a conditional diffusion-based framework?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of dense correspondence, which underpins various applications such as structure from motion, SLAM, image editing, and video analysis. By addressing the inherent ambiguities in dense correspondence, such as textureless regions and large displacements, this research could lead to more robust and generalizable models. The introduction of DiffMatch could inspire future research to explore diffusion models in other areas of computer vision, potentially leading to breakthroughs in generative and discriminative tasks.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities of accurately modeling the matching field distribution, particularly in the presence of ambiguities like repetitive patterns and noise. Naive approaches that focus solely on maximizing likelihood may fail because they do not account for the necessary prior knowledge of correspondences, which is essential for achieving smooth and accurate matching. Overcoming these technical obstacles requires a sophisticated understanding of both the data term and the prior term in the context of dense correspondence.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on hand-designed prior terms or on learning the data term without adequately addressing the matching prior, leading to limitations in performance and generalization. The lack of effective methods to model the matching field distribution has been a significant barrier. DiffMatch differs from prior work by explicitly incorporating a conditional diffusion process to model the matching field distribution, thereby addressing the shortcomings of existing approaches.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves using a conditional diffusion model to learn the matching field distribution for dense correspondence. The dataset will consist of pairs of images with known correspondences, and the evaluation metric will focus on the accuracy of pixel-wise correspondences. The expected outcomes include improved performance in dense correspondence tasks, particularly in challenging scenarios with textureless regions and large displacements, leading to more robust applications in computer vision."
            }
        },
        "author_data": {
            "d0a1c30c-785a-4368-b2f1-f8a1d90eccc3": {
                "pk": "d0a1c30c-785a-4368-b2f1-f8a1d90eccc3",
                "name": "Jisu Nam",
                "collaborators": [
                    "Seokju Cho",
                    "Seungryong Kim",
                    "Seung Wook Kim",
                    "Siyoon Jin",
                    "Seyeon Kim",
                    "Kihong Kim",
                    "Sung\u2010Jin Hong",
                    "Susung Hong",
                    "Sunghwan Hong",
                    "Jiahui Huang",
                    "Honggyu An",
                    "Joon-Young Lee",
                    "Heesu Kim",
                    "Dongjae Lee",
                    "Seunggyu Chang",
                    "Jihye Park",
                    "Jiyoung Kim",
                    "Gyuseong Lee",
                    "Sunwoo Kim",
                    "Ines Hyeonsu Kim",
                    "Hyoungwon Cho",
                    "Sangryul Jeon",
                    "Dongbo Min",
                    "Yunho Kim",
                    "Junyoung Seo",
                    "Kychul Lee",
                    "Kwanghee Lee",
                    "Stephen Lin",
                    "Seungjun Lee",
                    "Seungjun An"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Semantic Correspondence",
                    "Video Tracking"
                ],
                "publications": [
                    {
                        "title": "Local All-Pair Correspondence for Point Tracking",
                        "abstract": "We introduce LocoTrack, a highly accurate and efficient model designed for the task of tracking any point (TAP) across video sequences. Previous approaches in this task often rely on local 2D correlation maps to establish correspondences from a point in the query image to a local region in the target image, which often struggle with homogeneous regions or repetitive features, leading to matching ambiguities. LocoTrack overcomes this challenge with a novel approach that utilizes all-pair correspondences across regions, i.e., local 4D correlation, to establish precise correspondences, with bidirectional correspondence and matching smoothness significantly enhancing robustness against ambiguities. We also incorporate a lightweight correlation encoder to enhance computational efficiency, and a compact Transformer architecture to integrate long-term temporal information. LocoTrack achieves unmatched accuracy on all TAP-Vid benchmarks and operates at a speed almost 6 times faster than the current state-of-the-art."
                    },
                    {
                        "title": "DreamMatcher: Appearance Matching Self-Attention for Semantically-Consistent Text-to-Image Personalization",
                        "abstract": "The objective of text-to-image (T2I) personalization is to customize a diffusion model to a user-provided reference concept, generating diverse images of the concept aligned with the target prompts. Conventional methods representing the reference concepts using unique text embeddings often fail to accurately mimic the appearance of the reference. To address this, one solution may be explicitly conditioning the reference images into the target denoising process, known as key-value replacement. However, prior works are constrained to local editing since they disrupt the structure path of the pretrained T2I model. To overcome this, we propose a novel plug-in method, called DreamMatcher, which reformulates T2I personalization as semantic matching. Specifically, DreamMatcher replaces the target values with reference values aligned by semantic matching, while leaving the structure path unchanged to preserve the versatile capability of pretrained T2I models for generating diverse structures. We also introduce a semantic-consistent masking strategy to isolate the personalized concept from irrelevant regions introduced by the target prompts. Compatible with existing T2I models, DreamMatcher shows significant improvements in complex scenarios. Intensive analyses demonstrate the effectiveness of our approach."
                    },
                    {
                        "title": "MoDiTalker: Motion-Disentangled Diffusion Model for High-Fidelity Talking Head Generation",
                        "abstract": "Conventional GAN-based models for talking head generation often suffer from limited quality and unstable training. Recent approaches based on diffusion models aimed to address these limitations and improve fidelity. However, they still face challenges, including extensive sampling times and difficulties in maintaining temporal consistency due to the high stochasticity of diffusion models. To overcome these challenges, we propose a novel motion-disentangled diffusion model for high-quality talking head generation, dubbed MoDiTalker. We introduce the two modules: audio-to-motion (AToM), designed to generate a synchronized lip motion from audio, and motion-to-video (MToV), designed to produce high-quality head video following the generated motion. AToM excels in capturing subtle lip movements by leveraging an audio attention mechanism. In addition, MToV enhances temporal consistency by leveraging an efficient tri-plane representation. Our experiments conducted on standard benchmarks demonstrate that our model achieves superior performance compared to existing models. We also provide comprehensive ablation studies and user study results."
                    },
                    {
                        "title": "DiffMatch: Diffusion Model for Dense Matching",
                        "abstract": ","
                    },
                    {
                        "title": "Neural Matching Fields: Implicit Representation of Matching Fields for Visual Correspondence",
                        "abstract": "Existing pipelines of semantic correspondence commonly include extracting high-level semantic features for the invariance against intra-class variations and background clutters. This architecture, however, inevitably results in a low-resolution matching field that additionally requires an ad-hoc interpolation process as a post-processing for converting it into a high-resolution one, certainly limiting the overall performance of matching results. To overcome this, inspired by recent success of implicit neural representation, we present a novel method for semantic correspondence, called Neural Matching Field (NeMF). However, complicacy and high-dimensionality of a 4D matching field are the major hindrances, which we propose a cost embedding network to process a coarse cost volume to use as a guidance for establishing high-precision matching field through the following fully-connected network. Nevertheless, learning a high-dimensional matching field remains challenging mainly due to computational complexity, since a naive exhaustive inference would require querying from all pixels in the 4D space to infer pixel-wise correspondences. To overcome this, we propose adequate training and inference procedures, which in the training phase, we randomly sample matching candidates and in the inference phase, we iteratively performs PatchMatch-based inference and coordinate optimization at test time. With these combined, competitive results are attained on several standard benchmarks for semantic correspondence. Code and pre-trained weights are available at https://ku-cvlab.github.io/NeMF/."
                    },
                    {
                        "title": "DiffFace: Diffusion-based Face Swapping with Facial Guidance",
                        "abstract": "In this paper, we propose a diffusion-based face swapping framework for the first time, called DiffFace, composed of training ID conditional DDPM, sampling with facial guidance, and a target-preserving blending. In specific, in the training process, the ID conditional DDPM is trained to generate face images with the desired identity. In the sampling process, we use the off-the-shelf facial expert models to make the model transfer source identity while preserving target attributes faithfully. During this process, to preserve the background of the target image and obtain the desired face swapping result, we additionally propose a target-preserving blending strategy. It helps our model to keep the attributes of the target face from noise while transferring the source facial identity. In addition, without any re-training, our model can flexibly apply additional facial guidance and adaptively control the ID-attributes trade-off to achieve the desired results. To the best of our knowledge, this is the first approach that applies the diffusion model in face swapping task. Compared with previous GAN-based approaches, by taking advantage of the diffusion model for the face swapping task, DiffFace achieves better benefits such as training stability, high fidelity, diversity of the samples, and controllability. Extensive experiments show that our DiffFace is comparable or superior to the state-of-the-art methods on several standard face swapping benchmarks."
                    },
                    {
                        "title": "Cost Aggregation with 4D Convolutional Swin Transformer for Few-Shot Segmentation",
                        "abstract": "This paper presents a novel cost aggregation network, called Volumetric Aggregation with Transformers (VAT), for few-shot segmentation. The use of transformers can benefit correlation map aggregation through self-attention over a global receptive field. However, the tokenization of a correlation map for transformer processing can be detrimental, because the discontinuity at token boundaries reduces the local context available near the token edges and decreases inductive bias. To address this problem, we propose a 4D Convolutional Swin Transformer, where a high-dimensional Swin Transformer is preceded by a series of small-kernel convolutions that impart local context to all pixels and introduce convolutional inductive bias. We additionally boost aggregation performance by applying transformers within a pyramidal structure, where aggregation at a coarser level guides aggregation at a finer level. Noise in the transformer output is then filtered in the subsequent decoder with the help of the query's appearance embedding. With this model, a new state-of-the-art is set for all the standard benchmarks in few-shot segmentation. It is shown that VAT attains state-of-the-art performance for semantic correspondence as well, where cost aggregation also plays a central role."
                    },
                    {
                        "title": "Cost Aggregation with Transformers for Sparse Correspondence",
                        "abstract": "In this work, we introduce a novel network, namely SuperCATs, which aims to find a correspondence field between visually similar images. SuperCATs stands on the shoulder of the recently proposed matching networks, SuperGlue and CATs, taking the merits of both for constructing an integrative framework. Specifically, given keypoints and corresponding descriptors, we first apply attentional aggregation consisting of self- and cross-graph neural network to obtain feature descriptors. Subsequently, we construct a cost volume using the descriptors, which then undergoes a tranformer aggregator for cost aggregation. With this approach, we manage to replace the hand-crafted module based on solving an optimal transport problem initially included in SuperGlue with a transformer well known for its global receptive fields, making our approach more robust to severe deformations. We conduct experiments to demonstrate the effectiveness of the proposed method, and show that the proposed model is on par with SuperGlue for both indoor and outdoor scenes."
                    },
                    {
                        "title": "Cost Aggregation Is All You Need for Few-Shot Segmentation",
                        "abstract": "We introduce a novel cost aggregation network, dubbed Volumetric Aggregation with Transformers (VAT), to tackle the few-shot segmentation task by using both convolutions and transformers to efficiently handle high dimensional correlation maps between query and support. In specific, we propose our encoder consisting of volume embedding module to not only transform the correlation maps into more tractable size but also inject some convolutional inductive bias and volumetric transformer module for the cost aggregation. Our encoder has a pyramidal structure to let the coarser level aggregation to guide the finer level and enforce to learn complementary matching scores. We then feed the output into our affinity-aware decoder along with the projected feature maps for guiding the segmentation process. Combining these components, we conduct experiments to demonstrate the effectiveness of the proposed method, and our method sets a new state-of-the-art for all the standard benchmarks in few-shot segmentation task. Furthermore, we find that the proposed method attains state-of-the-art performance even for the standard benchmarks in semantic correspondence task although not specifically designed for this task. We also provide an extensive ablation study to validate our architectural choices. The trained weights and codes are available at: https://seokju-cho.github.io/VAT/."
                    }
                ]
            },
            "c70d34cd-4939-4ccc-8052-a168b6c8e0ab": {
                "pk": "c70d34cd-4939-4ccc-8052-a168b6c8e0ab",
                "name": "Gyuseong Lee",
                "collaborators": [
                    "Seung Wook Kim",
                    "Junyoung Seo",
                    "Kwang-seok Ryoo",
                    "Wooseok Jang",
                    "Jiwon Kim",
                    "Daehwan Kim",
                    "Seungryong Kim",
                    "Susung Hong",
                    "Seokju Cho",
                    "Hansang Cho",
                    "Jin Hyeon Kim",
                    "Jaewoo Jung",
                    "Jisu Nam",
                    "Sunwoo Kim",
                    "Ines Hyeonsu Kim",
                    "Hyoungwon Cho",
                    "Seyeon Kim",
                    "Youngjo Min",
                    "Jiyoung Lee",
                    "Yunsung Lee",
                    "Hyojun Go",
                    "Jihye Park",
                    "Gyeongnyeon Kim"
                ],
                "domain": [
                    "Domain Generalization",
                    "Semi-Supervised Learning",
                    "Generative Models",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Domain Generalization Using Large Pretrained Models with Mixture-of-Adapters",
                        "abstract": "Learning a robust vision model despite large distribution shift is essential for model deployment in real-world settings. Especially, domain generalization (DG) algorithm aims to maintain the performance of a trained model on different distributions which were not seen during training. One of the most effective methods has been leveraging the already learned rich knowledge of large pretrained models. However, naively fine-tuning large models to DG tasks is often practically infeasible due to memory limitations, extensive time requirements for training, and the risk of learned knowledge deterioration. Recently, parameter-efficient fine-tuning (PEFT) methods have been proposed to reduce the high computational cost during training and efficiently adapt large models to downstream tasks. In this work, for the first time, we find that the use of adapters in PEFT methods not only reduce high computational cost during training but also serve as an effective regularizer for DG tasks. Surprisingly, a naive adapter implementation for large models achieve superior performance on common datasets. However, in situations of large distribution shifts, additional factors such as optimal amount of regularization due to the strength of distribution shifts should be considered for a sophisticated adapter implementation. To address this, we propose a mixture-of-expert based adapter fine-tuning method, dubbed as mixture-of-adapters (MoA). Specifically, we employ multiple adapters that have varying capacities, and by using learnable routers, we allocate each token to a proper adapter. By using both PEFT and MoA methods, we effectively alleviate the performance deterioration caused by distribution shifts and achieve state-of-the-art performance on diverse DG benchmarks."
                    },
                    {
                        "title": "DiffMatch: Diffusion Model for Dense Matching",
                        "abstract": ","
                    },
                    {
                        "title": "Improving Sample Quality of Diffusion Models Using Self-Attention Guidance",
                        "abstract": "Denoising diffusion models (DDMs) have attracted attention for their exceptional generation quality and diversity. This success is largely attributed to the use of class- or text-conditional diffusion guidance methods, such as classifier and classifier-free guidance. In this paper, we present a more comprehensive perspective that goes beyond the traditional guidance methods. From this generalized perspective, we introduce novel condition- and training-free strategies to enhance the quality of generated images. As a simple solution, blur guidance improves the suitability of intermediate samples for their fine-scale information and structures, enabling diffusion models to generate higher quality samples with a moderate guidance scale. Improving upon this, Self-Attention Guidance (SAG) uses the intermediate self-attention maps of diffusion models to enhance their stability and efficacy. Specifically, SAG adversarially blurs only the regions that diffusion models attend to at each iteration and guides them accordingly. Our experimental re sults show that our SAG improves the performance of various diffusion models, including ADM, IDDPM, Stable Diffusion, and DiT. Moreover, combining SAG with conventional guidance methods leads to further improvement."
                    },
                    {
                        "title": "ConMatch: Semi-Supervised Learning with Confidence-Guided Consistency Regularization",
                        "abstract": "We present a novel semi-supervised learning framework that intelligently leverages the consistency regularization between the model's predictions from two strongly-augmented views of an image, weighted by a confidence of pseudo-label, dubbed ConMatch. While the latest semi-supervised learning methods use weakly- and strongly-augmented views of an image to define a directional consistency loss, how to define such direction for the consistency regularization between two strongly-augmented views remains unexplored. To account for this, we present novel confidence measures for pseudo-labels from strongly-augmented views by means of weakly-augmented view as an anchor in non-parametric and parametric approaches. Especially, in parametric approach, we present, for the first time, to learn the confidence of pseudo-label within the networks, which is learned with backbone model in an end-to-end manner. In addition, we also present a stage-wise training to boost the convergence of training. When incorporated in existing semi-supervised learners, ConMatch consistently boosts the performance. We conduct experiments to demonstrate the effectiveness of our ConMatch over the latest methods and provide extensive ablation studies. Code has been made publicly available at https://github.com/JiwonCocoder/ConMatch."
                    },
                    {
                        "title": "MIDMs: Matching Interleaved Diffusion Models for Exemplar-based Image Translation",
                        "abstract": "We present a novel method for exemplar-based image translation, called matching interleaved diffusion models (MIDMs). Most existing methods for this task were formulated as GAN-based matching-then-generation framework. However, in this framework, matching errors induced by the difficulty of semantic matching across cross-domain, e.g., sketch and photo, can be easily propagated to the generation step, which in turn leads to the degenerated results. Motivated by the recent success of diffusion models, overcoming the shortcomings of GANs, we incorporate the diffusion models to overcome these limitations. Specifically, we formulate a diffusion-based matching-and-generation framework that interleaves cross-domain matching and diffusion steps in the latent space by iteratively feeding the intermediate warp into the noising process and denoising it to generate a translated image. In addition, to improve the reliability of diffusion process, we design confidence-aware process using cycle-consistency to consider only confident regions during translation. Experimental results show that our MIDMs generate more plausible images than state-of-the-art methods."
                    },
                    {
                        "title": "Towards Flexible Inductive Bias via Progressive Reparameterization Scheduling",
                        "abstract": "There are two de facto standard architectures in recent computer vision: Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs). Strong inductive biases of convolutions help the model learn sample effectively, but such strong biases also limit the upper bound of CNNs when sufficient data are available. On the contrary, ViT is inferior to CNNs for small data but superior for sufficient data. Recent approaches attempt to combine the strengths of these two architectures. However, we show these approaches overlook that the optimal inductive bias also changes according to the target data scale changes by comparing various models' accuracy on subsets of sampled ImageNet at different ratios. In addition, through Fourier analysis of feature maps, the model's response patterns according to signal frequency changes, we observe which inductive bias is advantageous for each data scale. The more convolution-like inductive bias is included in the model, the smaller the data scale is required where the ViT-like model outperforms the ResNet performance. To obtain a model with flexible inductive bias on the data scale, we show reparameterization can interpolate inductive bias between convolution and self-attention. By adjusting the number of epochs the model stays in the convolution, we show that reparameterization from convolution to self-attention interpolates the Fourier analysis pattern between CNNs and ViTs. Adapting these findings, we propose Progressive Reparameterization Scheduling (PRS), in which reparameterization adjusts the required amount of convolution-like or self-attention-like inductive bias per layer. For small-scale datasets, our PRS performs reparameterization from convolution to self-attention linearly faster at the late stage layer. PRS outperformed previous studies on the small-scale dataset, e.g., CIFAR-100."
                    },
                    {
                        "title": "Semi-Supervised Learning of Semantic Correspondence with Pseudo-Labels",
                        "abstract": "Establishing dense correspondences across semantically similar images remains a challenging task due to the significant intra-class variations and background clutters. Traditionally, a supervised learning was used for training the models, which required tremendous manually-labeled data, while some methods suggested a self-supervised or weakly-supervised learning to mitigate the reliance on the labeled data, but with limited performance. In this paper, we present a simple, but effective solution for semantic correspondence that learns the networks in a semi-supervised manner by supplementing few ground-truth correspondences via utilization of a large amount of confident correspondences as pseudo-labels, called SemiMatch. Specifically, our framework generates the pseudo-labels using the model's prediction itself between source and weakly-augmented target, and uses pseudo-labels to learn the model again between source and strongly-augmented target, which improves the robustness of the model. We also present a novel confidence measure for pseudo-labels and data augmentation tailored for semantic correspondence. In experiments, SemiMatch achieves state-of-the-art performance on various benchmarks."
                    },
                    {
                        "title": "DAG: Depth-Aware Guidance with Denoising Diffusion Probabilistic Models",
                        "abstract": "Generative models have recently undergone significant advancement due to the diffusion models. The success of these models can be often attributed to their use of guidance techniques, such as classifier or classifier-free guidance, which provide effective mechanisms to trade-off between fidelity and diversity. However, these methods are not capable of guiding a generated image to be aware of its geometric configuration, e.g., depth, which hinders their application to areas that require a certain level of depth awareness. To address this limitation, we propose a novel guidance method for diffusion models that uses estimated depth information derived from the rich intermediate representations of diffusion models. We first present label-efficient depth estimation framework using internal representations of diffusion models. Subsequently, we propose the incorporation of two guidance techniques based on pseudo-labeling and depth-domain diffusion prior during the sampling phase to self-condition the generated image using the estimated depth map. Experiments and comprehensive ablation studies demonstrate the effectiveness of our method in guiding the diffusion models towards the generation of geometrically plausible images."
                    },
                    {
                        "title": "AggMatch: Aggregating Pseudo Labels for Semi-Supervised Learning",
                        "abstract": "Semi-supervised learning (SSL) has recently proven to be an effective paradigm for leveraging a huge amount of unlabeled data while mitigating the reliance on large labeled data. Conventional methods focused on extracting a pseudo label from individual unlabeled data sample and thus they mostly struggled to handle inaccurate or noisy pseudo labels, which degenerate performance. In this paper, we address this limitation with a novel SSL framework for aggregating pseudo labels, called AggMatch, which refines initial pseudo labels by using different confident instances. Specifically, we introduce an aggregation module for consistency regularization framework that aggregates the initial pseudo labels based on the similarity between the instances. To enlarge the aggregation candidates beyond the mini-batch, we present a class-balanced confidence-aware queue built with the momentum model, encouraging to provide more stable and consistent aggregation. We also propose a novel uncertainty-based confidence measure for the pseudo label by considering the consensus among multiple hypotheses with different subsets of the queue. We conduct experiments to demonstrate the effectiveness of AggMatch over the latest methods on standard benchmarks and provide extensive analyses."
                    }
                ]
            },
            "9baa0294-9c87-4ed5-9a02-1441f0e7705e": {
                "pk": "9baa0294-9c87-4ed5-9a02-1441f0e7705e",
                "name": "Sunwoo Kim",
                "collaborators": [
                    "Minje Kim",
                    "Seung Wook Kim",
                    "Minyoung Choe",
                    "Jaemin Yoo",
                    "Kijung Shin",
                    "B. Shim",
                    "Soohyun Kim",
                    "Wooseok Jang",
                    "Hyunsung Kim",
                    "Junho Kim",
                    "Yunjey Choi",
                    "Gayeong Lee",
                    "Fanchen Bu",
                    "Kyuhong Shim",
                    "L. Nguyen",
                    "Youngjo Min",
                    "Younghun Jung",
                    "Jaehyuk Heo",
                    "YongGi Jeong",
                    "Jaehee Kim",
                    "Pilsung Kang",
                    "Suhwan Song",
                    "Jaewon Hur",
                    "Philip Rogers",
                    "Byoungyoung Lee",
                    "Jihye Park",
                    "Jaejun Yoo",
                    "Youngjung Uh",
                    "K. Suh",
                    "Yongjun Ahn",
                    "Seungnyun Kim",
                    "Hyungyu Ju",
                    "B. Kim",
                    "Minhyeong Yu",
                    "J. Yoon",
                    "Seungjun Baek",
                    "Min Jung",
                    "Jong Seob Song",
                    "Seongmin Hong",
                    "Sang-Uk Go",
                    "Y. Lim",
                    "Juhan Park",
                    "Sung Goo Park",
                    "Yong-Min Kim",
                    "Matthew Muckley",
                    "Bruno Riemenschneider",
                    "A. Radmanesh",
                    "Geunu Jeong",
                    "Jingyu Ko",
                    "Yohan Jun",
                    "Hyungseob Shin",
                    "D. Hwang",
                    "Mahmoud Mostapha",
                    "S. Arberet",
                    "D. Nickel",
                    "Zaccharie Ramzi",
                    "P. Ciuciu",
                    "Jean-Luc Starck",
                    "J. Teuwen",
                    "D. Karkalousos",
                    "Chaoping Zhang",
                    "Anuroop Sriram",
                    "Zhengnan Huang",
                    "N. Yakubova",
                    "Y. Lui",
                    "F. Knoll",
                    "Aswin Sivaraman",
                    "Seungryong Kim"
                ],
                "domain": [
                    "Image Processing",
                    "Deep Learning",
                    "Hypergraph Neural Network",
                    "Speech Enhancement"
                ],
                "publications": [
                    {
                        "title": "User-friendly Image Editing with Minimal Text Input: Leveraging Captioning and Injection Techniques",
                        "abstract": "Recent text-driven image editing in diffusion models has shown remarkable success. However, the existing methods assume that the user's description sufficiently grounds the contexts in the source image, such as objects, background, style, and their relations. This assumption is unsuitable for real-world applications because users have to manually engineer text prompts to find optimal descriptions for different images. From the users' standpoint, prompt engineering is a labor-intensive process, and users prefer to provide a target word for editing instead of a full sentence. To address this problem, we first demonstrate the importance of a detailed text description of the source image, by dividing prompts into three categories based on the level of semantic details. Then, we propose simple yet effective methods by combining prompt generation frameworks, thereby making the prompt engineering process more user-friendly. Extensive qualitative and quantitative experiments demonstrate the importance of prompts in text-driven image editing and our method is comparable to ground-truth prompts."
                    },
                    {
                        "title": "Classification of Edge-dependent Labels of Nodes in Hypergraphs",
                        "abstract": "A hypergraph is a data structure composed of nodes and hyperedges, where each hyperedge is an any-sized subset of nodes. Due to the flexibility in hyperedge size, hypergraphs represent group interactions (e.g., co-authorship by more than two authors) more naturally and accurately than ordinary graphs. Interestingly, many real-world systems modeled as hypergraphs contain edge-dependent node labels, i.e., node labels that vary depending on hyperedges. For example, on co-authorship datasets, the same author (i.e., a node) can be the primary author in a paper (i.e., a hyperedge) but the corresponding author in another paper (i.e., another hyperedge). In this work, we introduce a classification of edge-dependent node labels as a new problem. This problem can be used as a benchmark task for hypergraph neural networks, which recently have attracted great attention, and also the usefulness of edge-dependent node labels has been verified in various applications. To tackle this problem, we propose WHATsNet, a novel hypergraph neural network that represents the same node differently depending on the hyperedges it participates in by reflecting its varying importance in the hyperedges. To this end, WHATsNet models the relations between nodes within each hyperedge, using their relative centrality as positional encodings. In our experiments, we demonstrate that WHATsNet significantly and consistently outperforms ten competitors on six real-world hypergraphs, and we also show successful applications of WHATsNet to (a) ranking aggregation, (b) node clustering, and (c) product return prediction."
                    },
                    {
                        "title": "How Transitive Are Real-World Group Interactions? - Measurement and Reproduction",
                        "abstract": "Many real-world interactions (e.g., researcher collaborations and email communication) occur among multiple entities. These group interactions are naturally modeled as hypergraphs. In graphs, transitivity is helpful to understand the connections between node pairs sharing a neighbor, and it has extensive applications in various domains. Hypergraphs, an extension of graphs, are designed to represent group relations. However, to the best of our knowledge, there has been no examination regarding the transitivity of real-world group interactions. In this work, we investigate the transitivity of group interactions in real-world hypergraphs. We first suggest intuitive axioms as necessary characteristics of hypergraph transitivity measures. Then, we propose a principled hypergraph transitivity measure HyperTrans, which satisfies all the proposed axioms, with a fast computation algorithm Fast-HyperTrans. After that, we analyze the transitivity patterns in real-world hypergraphs distinguished from those in random hypergraphs. Lastly, we propose a scalable hypergraph generator THera. It reproduces the observed transitivity patterns by leveraging community structures, which are pervasive in real-world hypergraphs. Our code and datasets are available at https://github.com/kswoo97/hypertrans."
                    },
                    {
                        "title": "Semantic-Preserving Augmentation for Robust Image-Text Retrieval",
                        "abstract": "Image-text retrieval is a task to search for the proper textual descriptions of the visual world and vice versa. One challenge of this task is the vulnerability to input image/text corruptions. Such corruptions are often unobserved during the training, and degrade the retrieval model\u2019s decision quality substantially. In this paper, we propose a novel image-text retrieval technique, referred to as robust visual semantic embedding (RVSE), which consists of novel image-based and text-based augmentation techniques called semantic-preserving augmentation for image (SPAug-I) and text (SPAug-T). Since SPAug-I and SPAug-T change the original data in a way that its semantic information is preserved, we enforce the feature extractors to generate semantic-aware embedding vectors regardless of the corruption, improving the model\u2019s robustness significantly. From extensive experiments using benchmark datasets, we show that RVSE outperforms conventional retrieval schemes in terms of image-text retrieval performance."
                    },
                    {
                        "title": "Controllable Style Transfer via Test-time Training of Implicit Neural Representation",
                        "abstract": "We propose a controllable style transfer framework based on Implicit Neural Representation that pixel-wisely controls the stylized output via test-time training. Unlike traditional image optimization methods that often suffer from unstable convergence and learning-based methods that require intensive training and have limited generalization ability, we present a model optimization framework that optimizes the neural networks during test-time with explicit loss functions for style transfer. After being test-time trained once, thanks to the flexibility of the INR-based model, our framework can precisely control the stylized images in a pixel-wise manner and freely adjust image resolution without further optimization or training. We demonstrate several applications."
                    },
                    {
                        "title": "REVECA - Rich Encoder-decoder framework for Video Event CAptioner",
                        "abstract": "We describe an approach used in the Generic Boundary Event Captioning challenge at the Long-Form Video Understanding Workshop held at CVPR 2022. We designed a Rich Encoder\u2013decoder framework for Video Event CAptioner (REVECA) that utilizes spatial and temporal information from the video to generate a caption for the corresponding the event boundary. REVECA uses frame position embedding to incorporate information before and after the event boundary. Furthermore, it employs features extracted using the temporal segment network and temporal-based pairwise difference method to learn temporal information. A semantic segmentation mask for the attentional pooling process is adopted to learn the subject of an event. Finally, LoRA is applied to \ufb01ne-tune the image encoder to enhance the learning ef\ufb01ciency. REVECA yielded an average score of 50.97 on the Kinetics-GEBC test data, which is an improvement of 10.17 over the baseline method. Our code is available in https://github.com/TooTouch/REVECA ."
                    },
                    {
                        "title": "R2Z2: Detecting Rendering Regressions in Web Browsers through Differential Fuzz Testing",
                        "abstract": "A rendering regression is a bug introduced by a web browser where a web page no longer functions as users expect. Such rendering bugs critically harm the usability of web browsers as well as web applications. The unique aspect of rendering bugs is that they affect the presented visual appearance of web pages, but those web pages have no pre-defined correct appearance. Therefore, it is challenging to automatically detect errors in their appearance. In practice, web browser vendors rely on non-trivial and time-prohibitive manual analysis to detect and handle rendering regressions. This paper proposes R2Z2, an automated tool to find rendering regressions. R2Z2 uses the differential fuzz testing approach, which repeatedly compares the rendering results of two different versions of a browser while providing the same HTML as input. If the rendering results are different, R2Z2 further performs cross browser compatibility testing to check if the rendering difference is indeed a rendering regression. After identifying a rendering regression, R2Z2 will perform an in-depth analysis to aid in fixing the regression. Specifically, R2Z2 performs a delta-debugging-like analysis to pinpoint the exact browser source code commit causing the regression, as well as inspecting the rendering pipeline stages to pinpoint which pipeline stage is responsible. We implemented a prototype of R2Z2 particularly targeting the Chrome browser. So far, R2Z2 found 11 previously undiscovered rendering regressions in Chrome, all of which were confirmed by the Chrome developers. Importantly, in each case, R2Z2 correctly reported the culprit commit. Moreover, R2Z2 correctly pin-pointed the culprit rendering pipeline stage in all but one case."
                    },
                    {
                        "title": "Boosted Locality Sensitive Hashing: Discriminative, Efficient, and Scalable Binary Codes for Source Separation",
                        "abstract": "We propose a novel adaptive boosting approach to learn discriminative binary hash codes, boosted locality sensitive hashing (BLSH), that can represent audio spectra efficiently. We aim to use the learned hash codes in the single-channel speech denoising task by designing a nearest neighborhood search method that operates in the hashed feature space. To achieve the optimal denoising results given the highly compact binary feature representation, our proposed BLSH algorithm learns simple logistic regressors as the weak learners in an incremental way (i.e., one by one) so that each weak learner is trained to complement the mistake its predecessors have made. Upon testing, their binary classification results transform each spectrum of noisy speech into a bit string, where the bits are ordered based on their significance, adding scalability to the denoising system. Simple bitwise operations calculate Hamming distance to find the $\\boldsymbol{K}$-nearest matching hashed frames in the dictionary of training noisy speech spectra, whose associated ideal binary masks are averaged to estimate the denoising mask for that test mixture. In contrast to the locality sensitive hashing method's random projections, our proposed supervised learning algorithm trains the projections such that the distance between the self-similarity matrix of the hash codes and that of the original spectra is minimized. Likewise, the process conceptually aligns to the Adaboost algorithm, although ours is specialized in learning binary features for source separation rather than classification. Experimental results on speech denoising suggest that the BLSH algorithm learns more discriminative representations than Fourier or mel spectra and the nonlinear kernels derived from them. Our compact binary representation is expected to facilitate model deployment onto resource-constrained environments, where comprehensive models (e.g., deep neural networks) are unaffordable."
                    },
                    {
                        "title": "LANIT: Language-Driven Image-to-Image Translation for Unlabeled Data",
                        "abstract": "Existing techniques for image-to-image translation commonly have suffered from two critical problems: heavy reliance on per-sample domain annotation and/or inability to handle multiple attributes per image. Recent truly-unsupervised methods adopt clustering approaches to easily provide per-sample one-hot domain labels. However, they cannot account for the real-world setting: one sample may have multiple attributes. In addition, the semantics of the clusters are not easily coupled to human understanding. To overcome these, we present LANguage-driven Image-to-image Translation model, dubbed LANIT. We leverage easy-to-obtain candidate attributes given in texts for a dataset: the similarity between images and attributes indicates per-sample domain labels. This formulation naturally enables multi-hot labels so that users can specify the target domain with a set of attributes in language. To account for the case that the initial prompts are inaccurate, we also present prompt learning. We further present domain regularization loss that enforces translated images to be mapped to the corresponding domain. Experiments on several standard benchmarks demonstrate that LANIT achieves comparable or superior performance to existing models. The code is available at github.com/KU-CVLAB/LANIT."
                    },
                    {
                        "title": "Deep Reinforcement Learning-based Network Slicing for Beyond 5G",
                        "abstract": "With the advent of 5G era, network slicing has received a great deal of attention as a means to support a variety of wireless services in a flexible manner. Network slicing is a technique to divide a single physical resource network into multiple slices supporting independent services. In beyond 5G (B5G) systems, the main goal of network slicing is to assign the physical resource blocks (RBs) such that the quality of service (QoS) requirements of eMBB, URLLC, and mMTC services are satisfied. Since the goal of each service category is dearly distinct and the computational burden caused by the increased number of time slots is huge, it is in general very difficult to assign each RB to a certain service properly. In this paper, we propose a deep reinforcement learning (DRL)-based network slicing technique to find out the resource allocation policy maximizing the long-term throughput while satisfying the QoS requirements in the B5G systems. Key ingredient of the proposed technique is to use action elimination to eliminate undesirable actions that cannot satisfy the QoS requirements. Numerical results demonstrate that the proposed technique is effective in maximizing the long-term throughput and handling the coexistence of use cases in the B5G environments."
                    },
                    {
                        "title": "Scale-attentional U-Net for the segmentation of the median nerve in ultrasound images",
                        "abstract": "Purpose The aim of this study was to develop a neural network that accurately and effectively segments the median nerve in ultrasound (US) images. Methods In total, 1,305 images of the median nerve of 123 normal subjects were used to train and evaluate the model. Four datasets from two measurement regions (wrist and forearm) of the nerve and two US machines were used. The neural network was designed for high accuracy by combining information at multiple scales, as well as for high efficiency to prevent overfitting. The model was designed in two parts (cascaded and factorized convolutions), followed by self-attention over scale and channel features. The precision, recall, dice similarity coefficient (DSC), and Hausdorff distance (HD) were used as performance metrics. The area under the receiver operating characteristic curve (AUC) was also assessed. Results In the wrist datasets, the proposed network achieved 92.7% and 90.3% precision, 92.4% and 89.8% recall, DSCs of 92.3% and 89.7%, HDs of 5.158 and 4.966, and AUCs of 0.9755 and 0.9399 on two machines. In the forearm datasets, 79.3% and 87.8% precision, 76.0% and 85.0% recall, DSCs of 76.1% and 85.8%, HDs of 5.206 and 4.527, and AUCs of 0.8846 and 0.9150 were achieved. In all datasets, the model developed herein achieved better performance in terms of DSC than previous U-Net-based systems. Conclusion The proposed neural network yields accurate segmentation results to assist clinicians in identifying the median nerve."
                    },
                    {
                        "title": "Deep Learning Algorithms Correctly Classify Brassica rapa Varieties Using Digital Images",
                        "abstract": "Efficient and accurate methods of analysis are needed for the huge amount of biological data that have accumulated in various research fields, including genomics, phenomics, and genetics. Artificial intelligence (AI)-based analysis is one promising method to manipulate biological data. To this end, various algorithms have been developed and applied in fields such as disease diagnosis, species classification, and object prediction. In the field of phenomics, classification of accessions and variants is important for basic science and industrial applications. To construct AI-based classification models, three types of phenotypic image data were generated from 156 Brassica rapa core collections, and classification analyses were carried out using four different convolutional neural network architectures. The results of lateral view data showed higher accuracy compared with top view data. Furthermore, the relatively low accuracy of ResNet50 architecture suggested that definition and estimation of similarity index of phenotypic data were required before the selection of deep learning architectures."
                    },
                    {
                        "title": "Results of the 2020 fastMRI Challenge for Machine Learning MR Image Reconstruction",
                        "abstract": "Accelerating MRI scans is one of the principal outstanding problems in the MRI research community. Towards this goal, we hosted the second fastMRI competition targeted towards reconstructing MR images with subsampled k-space data. We provided participants with data from 7,299 clinical brain scans (de-identified via a HIPAA-compliant procedure by NYU Langone Health), holding back the fully-sampled data from 894 of these scans for challenge evaluation purposes. In contrast to the 2019 challenge, we focused our radiologist evaluations on pathological assessment in brain images. We also debuted a new Transfer track that required participants to submit models evaluated on MRI scanners from outside the training set. We received 19 submissions from eight different groups. Results showed one team scoring best in both SSIM scores and qualitative radiologist evaluations. We also performed analysis on alternative metrics to mitigate the effects of background noise and collected feedback from the participants to inform future challenges. Lastly, we identify common failure modes across the submissions, highlighting areas of need for future research in the MRI reconstruction community."
                    },
                    {
                        "title": "Bloom-Net: Blockwise Optimization for Masking Networks Toward Scalable and Efficient Speech Enhancement",
                        "abstract": "In this paper, we present a blockwise optimization method for masking-based networks (BLOOM-Net) for training scalable speech enhancement networks. Here, we design our network with a residual learning scheme and train the internal separator blocks sequentially to obtain a scalable masking-based deep neural network for speech enhancement. Its scalability lets it dynamically adjust the run-time complexity depending on the test time environment. To this end, we modularize our models in that they can flexibly accommodate varying needs for enhancement performance and constraints on the resources, incurring minimal memory or training overhead due to the added scalability. Our experiments on speech enhancement demonstrate that the proposed blockwise optimization method achieves the desired scalability with only a slight performance degradation compared to corresponding models trained end-to-end."
                    },
                    {
                        "title": "Personalized Speech Enhancement through Self-Supervised Data Augmentation and Purification",
                        "abstract": "Training personalized speech enhancement models is innately a no-shot learning problem due to privacy constraints and limited access to noise-free speech from the target user. If there is an abundance of unlabeled noisy speech from the test-time user, a personalized speech enhancement model can be trained using self-supervised learning. One straightforward approach to model personalization is to use the target speaker's noisy recordings as pseudo-sources. Then, a pseudo denoising model learns to remove injected training noises and recover the pseudo-sources. However, this approach is volatile as it depends on the quality of the pseudo-sources, which may be too noisy. As a remedy, we propose an improvement to the self-supervised approach through data purification. We first train an SNR predictor model to estimate the frame-by-frame SNR of the pseudo-sources. Then, the predictor's estimates are converted into weights which adjust the frame-by-frame contribution of the pseudo-sources towards training the personalized model. We empirically show that the proposed data purification step improves the usability of the speaker-specific noisy data in the context of personalized speech enhancement. Without relying on any clean speech recordings or speaker embeddings, our approach may be seen as privacy-preserving."
                    },
                    {
                        "title": "Test-Time Adaptation Toward Personalized Speech Enhancement: Zero-Shot Learning with Knowledge Distillation",
                        "abstract": "In realistic speech enhancement settings for end-user devices, we often encounter only a few speakers and noise types that tend to reoccur in the specific acoustic environment. We propose a novel personalized speech enhancement method to adapt a compact denoising model to the test-time specificity. Our goal in this test-time adaptation is to utilize no clean speech target of the test speaker, thus fulfilling the requirement for zero-shot learning. To complement the lack of clean speech, we employ the knowledge distillation framework: we distill the more advanced denoising results from an overly large teacher model, and use them as the pseudo target to train the small student model. This zero-shot learning procedure circumvents the process of collecting users' clean speech, a process that users are reluctant to comply due to privacy concerns and technical difficulty of recording clean voice. Experiments on various test-time conditions show that the proposed personalization method can significantly improve the compact models' performance during the test time. Furthermore, since the personalized models outperform larger non-personalized baseline models, we claim that personalization achieves model compression with no loss of denoising performance. As expected, the student models underperform the state-of-the-art teacher models."
                    },
                    {
                        "title": "Deep Translation Prior: Test-time Training for Photorealistic Style Transfer",
                        "abstract": "Recent techniques to solve photorealistic style transfer within deep convolutional neural networks (CNNs) generally require intensive training from large-scale datasets, thus having limited applicability and poor generalization ability to unseen images or styles. To overcome this, we propose a novel framework, dubbed Deep Translation Prior (DTP), to accomplish photorealistic style transfer through test-time training on given input image pair with untrained networks, which learns an image pair-specific translation prior and thus yields better performance and generalization. Tailored for such test-time training for style transfer, we present novel network architectures, with two sub-modules of correspondence and generation modules, and loss functions consisting of contrastive content, style, and cycle consistency losses. Our framework does not require offline training phase for style transfer, which has been one of the main challenges in existing methods, but the networks are to be solely learned during test time. Experimental results prove that our framework has a better generalization ability to unseen image pairs and even outperforms the state-of-the-art methods."
                    }
                ]
            },
            "5c876041-ae11-4b15-b805-7f8e49b32b7d": {
                "pk": "5c876041-ae11-4b15-b805-7f8e49b32b7d",
                "name": "Hyeonsu Kim",
                "collaborators": [
                    "Minseop Kwak",
                    "Seungryong Kim",
                    "Junyoung Seo",
                    "Wooseok Jang",
                    "Donghoon Ahn",
                    "Jin-wha Kim",
                    "Joungbin Lee",
                    "Soowon Son",
                    "Woojeong Jin",
                    "Kyusun Cho",
                    "Seokju Cho",
                    "Jeongyeol Baek",
                    "Byeongwon Lee",
                    "Susung Hong",
                    "Doyup Lee",
                    "Jaehoon Ko",
                    "Junho Kim",
                    "Jin-Hwa Kim",
                    "Jiyoung Lee",
                    "Seung Wook Kim",
                    "Jisu Nam",
                    "Gyuseong Lee",
                    "Sunwoo Kim",
                    "Hyoungwon Cho",
                    "Seyeon Kim"
                ],
                "domain": [
                    "Text-to-3D Generation",
                    "Diffusion Models",
                    "Computer Vision",
                    "Data Augmentation"
                ],
                "publications": [
                    {
                        "title": "Geometry-Aware Score Distillation via 3D Consistent Noising and Gradient Consistency Modeling",
                        "abstract": "Score distillation sampling (SDS), the methodology in which the score from pretrained 2D diffusion models is distilled into 3D representation, has recently brought significant advancements in text-to-3D generation task. However, this approach is still confronted with critical geometric inconsistency problems such as the Janus problem. Starting from a hypothesis that such inconsistency problems may be induced by multiview inconsistencies between 2D scores predicted from various viewpoints, we introduce GSD, a simple and general plug-and-play framework for incorporating 3D consistency and therefore geometry awareness into the SDS process. Our methodology is composed of three components: 3D consistent noising, designed to produce 3D consistent noise maps that perfectly follow the standard Gaussian distribution, geometry-based gradient warping for identifying correspondences between predicted gradients of different viewpoints, and novel gradient consistency loss to optimize the scene geometry toward producing more consistent gradients. We demonstrate that our method significantly improves performance, successfully addressing the geometric inconsistency problems in text-to-3D generation task with minimal computation cost and being compatible with existing score distillation-based models. Our project page is available at https://ku-cvlab.github.io/GSD/."
                    },
                    {
                        "title": "Pose-Diversified Augmentation with Diffusion Model for Person Re-Identification",
                        "abstract": "Person re-identification (Re-ID) often faces challenges due to variations in human poses and camera viewpoints, which significantly affect the appearance of individuals across images. Existing datasets frequently lack diversity and scalability in these aspects, hindering the generalization of Re-ID models to new camera systems. We propose Pose-dIVE, a novel data augmentation approach that incorporates sparse and underrepresented human pose and camera viewpoint examples into the training data, addressing the limited diversity in the original training data distribution. Our objective is to augment the training dataset to enable existing Re-ID models to learn features unbiased by human pose and camera viewpoint variations. To achieve this, we leverage the knowledge of pre-trained large-scale diffusion models. By conditioning the diffusion model on both the human pose and camera viewpoint concurrently through the SMPL model, we generate training data with diverse human poses and camera viewpoints. Experimental results demonstrate the effectiveness of our method in addressing human pose bias and enhancing the generalizability of Re-ID models compared to other data augmentation-based Re-ID approaches."
                    },
                    {
                        "title": "Retrieval-Augmented Score Distillation for Text-to-3D Generation",
                        "abstract": "Text-to-3D generation has achieved significant success by incorporating powerful 2D diffusion models, but insufficient 3D prior knowledge also leads to the inconsistency of 3D geometry. Recently, since large-scale multi-view datasets have been released, fine-tuning the diffusion model on the multi-view datasets becomes a mainstream to solve the 3D inconsistency problem. However, it has confronted with fundamental difficulties regarding the limited quality and diversity of 3D data, compared with 2D data. To sidestep these trade-offs, we explore a retrieval-augmented approach tailored for score distillation, dubbed ReDream. We postulate that both expressiveness of 2D diffusion models and geometric consistency of 3D assets can be fully leveraged by employing the semantically relevant assets directly within the optimization process. To this end, we introduce novel framework for retrieval-based quality enhancement in text-to-3D generation. We leverage the retrieved asset to incorporate its geometric prior in the variational objective and adapt the diffusion model's 2D prior toward view consistency, achieving drastic improvements in both geometry and fidelity of generated scenes. We conduct extensive experiments to demonstrate that ReDream exhibits superior quality with increased geometric consistency. Project page is available at https://ku-cvlab.github.io/ReDream/."
                    },
                    {
                        "title": "Let 2D Diffusion Model Know 3D-Consistency for Robust Text-to-3D Generation",
                        "abstract": "Text-to-3D generation has shown rapid progress in recent days with the advent of score distillation, a methodology of using pretrained text-to-2D diffusion models to optimize neural radiance field (NeRF) in the zero-shot setting. However, the lack of 3D awareness in the 2D diffusion models destabilizes score distillation-based methods from reconstructing a plausible 3D scene. To address this issue, we propose 3DFuse, a novel framework that incorporates 3D awareness into pretrained 2D diffusion models, enhancing the robustness and 3D consistency of score distillation-based methods. We realize this by first constructing a coarse 3D structure of a given text prompt and then utilizing projected, view-specific depth map as a condition for the diffusion model. Additionally, we introduce a training strategy that enables the 2D diffusion model learns to handle the errors and sparsity within the coarse 3D structure for robust generation, as well as a method for ensuring semantic consistency throughout all viewpoints of the scene. Our framework surpasses the limitations of prior arts, and has significant implications for 3D consistent generation of 2D diffusion models."
                    },
                    {
                        "title": "DiffMatch: Diffusion Model for Dense Matching",
                        "abstract": ","
                    }
                ]
            },
            "c7e714cc-dd51-4771-b94c-0d673bbc7368": {
                "pk": "c7e714cc-dd51-4771-b94c-0d673bbc7368",
                "name": "Hyoungwon Cho",
                "collaborators": [
                    "Seungryong Kim",
                    "Donghoon Ahn",
                    "Jaewon Min",
                    "Wooseok Jang",
                    "Jungwoo Kim",
                    "SeonHwa Kim",
                    "Hyun Hee Park",
                    "Kyong Hwan Jin",
                    "Jisu Nam",
                    "Gyuseong Lee",
                    "Sunwoo Kim",
                    "Ines Hyeonsu Kim",
                    "Seyeon Kim"
                ],
                "domain": [
                    "Diffusion Models",
                    "Image Generation",
                    "Sampling Techniques",
                    "Image Restoration"
                ],
                "publications": [
                    {
                        "title": "Self-Rectifying Diffusion Sampling with Perturbed-Attention Guidance",
                        "abstract": "Recent studies have demonstrated that diffusion models are capable of generating high-quality samples, but their quality heavily depends on sampling guidance techniques, such as classifier guidance (CG) and classifier-free guidance (CFG). These techniques are often not applicable in unconditional generation or in various downstream tasks such as image restoration. In this paper, we propose a novel sampling guidance, called Perturbed-Attention Guidance (PAG), which improves diffusion sample quality across both unconditional and conditional settings, achieving this without requiring additional training or the integration of external modules. PAG is designed to progressively enhance the structure of samples throughout the denoising process. It involves generating intermediate samples with degraded structure by substituting selected self-attention maps in diffusion U-Net with an identity matrix, by considering the self-attention mechanisms' ability to capture structural information, and guiding the denoising process away from these degraded samples. In both ADM and Stable Diffusion, PAG surprisingly improves sample quality in conditional and even unconditional scenarios. Moreover, PAG significantly improves the baseline performance in various downstream tasks where existing guidances such as CG or CFG cannot be fully utilized, including ControlNet with empty prompts and image restoration such as inpainting and deblurring."
                    },
                    {
                        "title": "DiffMatch: Diffusion Model for Dense Matching",
                        "abstract": ","
                    }
                ]
            },
            "53c765b8-126d-420f-bc70-993e92e5f667": {
                "pk": "53c765b8-126d-420f-bc70-993e92e5f667",
                "name": "Seyeon Kim",
                "collaborators": [
                    "Sangil Na",
                    "Hayoon Kim",
                    "Moonki Kim",
                    "Byoung-Ki Jeon",
                    "Taewan Kim",
                    "Beyeongki Jeon"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "E-commerce",
                    "Visual Search"
                ],
                "publications": [
                    {
                        "title": "M ULTI-LABEL L EARNING WITH THE RNN S FOR F ASHION S EARCH",
                        "abstract": "We build a large-scale visual search system which finds similar product images given a fashion item. Defining similarity among arbitrary fashion-products is still remains a challenging problem, even there is no exact ground-truth. To resolve this problem, we define more than 90 fashion-related attributes, and combination of these attributes can represent thousands of unique fashion-styles. We then introduce to use the recurrent neural networks (RNNs) recognising multiple fashion-attributes with the end-to-end manner. To build our system at scale, these fashion-attributes are again used to build an inverted indexing scheme. In addition to these fashion-attributes for semantic similarity, we extract colour and appearance features in a region-of-interest (ROI) of a fashion item for visual similarity. By sharing our approach, we expect active discussion on that how to apply current deep learning researches into the e-commerce industry."
                    },
                    {
                        "title": "Visual Fashion-Product Search at SK Planet",
                        "abstract": "We build a large-scale visual search system which finds similar product images given a fashion item. Defining similarity among arbitrary fashion-products is still remains a challenging problem, even there is no exact ground-truth. To resolve this problem, we define more than 90 fashion-related attributes, and combination of these attributes can represent thousands of unique fashion-styles. The fashion-attributes are one of the ingredients to define semantic similarity among fashion-product images. To build our system at scale, these fashion-attributes are again used to build an inverted indexing scheme. In addition to these fashion-attributes for semantic similarity, we extract colour and appearance features in a region-of-interest (ROI) of a fashion item for visual similarity. By sharing our approach, we expect active discussion on that how to apply current computer vision research into the e-commerce industry."
                    }
                ]
            },
            "e08d8e19-83da-406f-9419-911f7640c668": {
                "pk": "e08d8e19-83da-406f-9419-911f7640c668",
                "name": "Seungryong Kim",
                "collaborators": [
                    "Wooseok Jang",
                    "Junho Kim",
                    "Susung Hong",
                    "Minseop Kwak",
                    "Jiyoung Lee",
                    "Karsten Kreis",
                    "S. Fidler",
                    "Sunwoo Kim",
                    "Honglak Lee",
                    "Junyoung Seo",
                    "Jaehoon Ko",
                    "Jin-Hwa Kim",
                    "Seokju Cho",
                    "Heeseong Shin",
                    "Sung\u2010Jin Hong",
                    "Jiuhn Song",
                    "Daiqing Li",
                    "Robin Rombach",
                    "A. Torralba",
                    "Huan Ling",
                    "J. Lim",
                    "Myunghoon Kang",
                    "Yuna Hur",
                    "Seung-Ju Jung",
                    "Jinsung Kim",
                    "Yoonna Jang",
                    "Dongyub Lee",
                    "Hyesung Ji",
                    "Heu-Jeoung Lim",
                    "Hyunsung Kim",
                    "Yunjey Choi",
                    "Gayeong Lee",
                    "Donghoon Ahn",
                    "Daechul Ahn",
                    "Daneul Kim",
                    "Gwangmo Song",
                    "Dongyeop Kang",
                    "Jonghyun Choi",
                    "Ines Hyeonsu Kim",
                    "Seungjun An",
                    "Seungjun Lee",
                    "Anurag Arnab",
                    "P. H. Seo",
                    "GiJae Lee",
                    "Jun-Sik Kim",
                    "Kanggeon Kim",
                    "Soohyun Kim",
                    "Taekyung Kim",
                    "Hwan Heo",
                    "Honggyu An",
                    "Sung-Jin Cho",
                    "B. Brown",
                    "K. Yin",
                    "Katja Schwarz",
                    "Amlan Kar",
                    "David Acuna",
                    "Taehoon Kim",
                    "Pyunghwan Ahn",
                    "Sangyun Kim",
                    "Sihaeng Lee",
                    "Mark Marsden",
                    "Alessandra Sala",
                    "Bohyung Han",
                    "Kyoung Mu Lee",
                    "Kyounghoon Bae",
                    "Xiangyu Wu",
                    "Yi Gao",
                    "Hailiang Zhang",
                    "Yang Yang",
                    "Weili Guo",
                    "Jianfeng Lu",
                    "Youngtaek Oh",
                    "Jae-Won Cho",
                    "Dong-Jin Kim",
                    "In-So Kweon",
                    "Junmo Kim",
                    "Woohyun Kang",
                    "Won Young Jhoo",
                    "Byungseok Roh",
                    "Jonghwan Mun",
                    "Solgil Oh",
                    "Kenan E. Ak",
                    "Gwang Lee",
                    "Yan Xu",
                    "Mingwei Shen",
                    "Kyomin Hwang",
                    "Wonsik Shin",
                    "Kamin Lee",
                    "Wonhark Park",
                    "Dongkwan Lee",
                    "N. Kwak",
                    "Yujin Wang",
                    "Yimu Wang",
                    "Tiancheng Gu",
                    "Xingchang Lv",
                    "Mingmao Sun",
                    "Seunghyun Won",
                    "Joonseok Lee",
                    "M. Ghassemi",
                    "James Thorne"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Computer Vision",
                    "3D Generation",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "You Truly Understand What I Need: Intellectual and Friendly Dialogue Agents grounding Knowledge and Persona",
                        "abstract": "To build a conversational agent that interacts fluently with humans, previous studies blend knowledge or personal profile into the pre-trained language model. However, the model that considers knowledge and persona at the same time is still limited, leading to hallucination and a passive way of using personas. We propose an effective dialogue agent that grounds external knowledge and persona simultaneously. The agent selects the proper knowledge and persona to use for generating the answers with our candidate scoring implemented with a poly-encoder. Then, our model generates the utterance with lesser hallucination and more engagingness utilizing retrieval augmented generation with knowledge-persona enhanced query. We conduct experiments on the persona-knowledge chat and achieve state-of-the-art performance in grounding and generation tasks on the automatic metrics. Moreover, we validate the answers from the models regarding hallucination and engagingness through human evaluation and qualitative results. We show our retriever's effectiveness in extracting relevant documents compared to the other previous retrievers, along with the comparison of multiple candidate scoring methods. Code is available at https://github.com/dlawjddn803/INFO"
                    },
                    {
                        "title": "User-friendly Image Editing with Minimal Text Input: Leveraging Captioning and Injection Techniques",
                        "abstract": "Recent text-driven image editing in diffusion models has shown remarkable success. However, the existing methods assume that the user's description sufficiently grounds the contexts in the source image, such as objects, background, style, and their relations. This assumption is unsuitable for real-world applications because users have to manually engineer text prompts to find optimal descriptions for different images. From the users' standpoint, prompt engineering is a labor-intensive process, and users prefer to provide a target word for editing instead of a full sentence. To address this problem, we first demonstrate the importance of a detailed text description of the source image, by dividing prompts into three categories based on the level of semantic details. Then, we propose simple yet effective methods by combining prompt generation frameworks, thereby making the prompt engineering process more user-friendly. Extensive qualitative and quantitative experiments demonstrate the importance of prompts in text-driven image editing and our method is comparable to ground-truth prompts."
                    },
                    {
                        "title": "Debiasing Scores and Prompts of 2D Diffusion for Robust Text-to-3D Generation",
                        "abstract": "The view inconsistency problem in score-distilling text-to-3D generation, also known as the Janus problem, arises from the intrinsic bias of 2D diffusion models, which leads to the unrealistic generation of 3D objects. In this work, we explore score-distilling text-to-3D generation and identify the main causes of the Janus problem. Based on these \ufb01ndings, we propose two approaches to debias the score-distillation frameworks for robust text-to-3D generation. Our \ufb01rst approach, called score debiasing, involves gradually increasing the truncation value for the score estimated by 2D diffusion models throughout the optimization process. Our second approach, called prompt debiasing, identi\ufb01es con\ufb02icting words between user prompts and view prompts utilizing a language model and adjusts the discrepancy between view prompts and object-space camera poses. Our experimental results show that our methods improve realism by signi\ufb01cantly reducing artifacts and achieve a good trade-off between faithfulness to the 2D diffusion models and 3D consistency with little overhead."
                    },
                    {
                        "title": "Story Visualization by Online Text Augmentation with Context Memory",
                        "abstract": "Story visualization (SV) is a challenging text-to-image generation task for the difficulty of not only rendering visual details from the text descriptions but also encoding a long-term context across multiple sentences. While prior efforts mostly focus on generating a semantically relevant image for each sentence, encoding a context spread across the given paragraph to generate contextually convincing images (e.g., with a correct character or with a proper background of the scene) remains a challenge. To this end, we propose a novel memory architecture for the Bi-directional Transformer framework with an online text augmentation that generates multiple pseudo-descriptions as supplementary supervision during training for better generalization to the language variation at inference. In extensive experiments on the two popular SV benchmarks, i.e., the Pororo-SV and Flintstones-SV, the proposed method significantly outperforms the state of the arts in various metrics including FID, character F1, frame accuracy, BLEU-2/3, and R-precision with similar or less computational complexity."
                    },
                    {
                        "title": "Let 2D Diffusion Model Know 3D-Consistency for Robust Text-to-3D Generation",
                        "abstract": "Text-to-3D generation has shown rapid progress in recent days with the advent of score distillation, a methodology of using pretrained text-to-2D diffusion models to optimize neural radiance field (NeRF) in the zero-shot setting. However, the lack of 3D awareness in the 2D diffusion models destabilizes score distillation-based methods from reconstructing a plausible 3D scene. To address this issue, we propose 3DFuse, a novel framework that incorporates 3D awareness into pretrained 2D diffusion models, enhancing the robustness and 3D consistency of score distillation-based methods. We realize this by first constructing a coarse 3D structure of a given text prompt and then utilizing projected, view-specific depth map as a condition for the diffusion model. Additionally, we introduce a training strategy that enables the 2D diffusion model learns to handle the errors and sparsity within the coarse 3D structure for robust generation, as well as a method for ensuring semantic consistency throughout all viewpoints of the scene. Our framework surpasses the limitations of prior arts, and has significant implications for 3D consistent generation of 2D diffusion models."
                    },
                    {
                        "title": "CAT-Seg: Cost Aggregation for Open-Vocabulary Semantic Segmentation",
                        "abstract": "Open-vocabulary semantic segmentation presents the challenge of labeling each pixel within an image based on a wide range of text descriptions. In this work, we introduce a novel cost-based approach to adapt vision-language foundation models, notably CLIP, for the intricate task of semantic segmentation. Through aggregating the cosine similarity score, i. e., the cost volume between image and text embeddings, our method potently adapts CLIP for segmenting seen and unseen classes by fine-tuning its encoders, addressing the challenges faced by existing methods in handling unseen classes. Building upon this, we explore methods to effectively aggregate the cost volume considering its multi-modal nature of being established between image and text embeddings. Furthermore, we examine various methods for efficiently fine-tuning CLIP."
                    },
                    {
                        "title": "6D Object Pose Estimation Using a Particle Filter With Better Initialization",
                        "abstract": "Estimation of 6D object poses is a key issue in robotic grasping tasks. Recently, many high-performance learning-based methods have been introduced using robust deep learning techniques; however, applying these methods to real robot environments requires many ground truth 6D pose annotations for training. To address this problem, we propose a template matching-based particle filter approach for 6D pose estimation; the proposed method does not require ground truth 6D poses. Although particle filter approaches can stochastically avoid local optima, they require adequate initial pose hypotheses for estimating an accurate 6D object pose. Therefore, we estimated an initial translation of the target object for accurately initializing a particle filter by developing a new deep network. Once the proposed centroid prediction network (CPN) is trained with a specific dataset, no additional training is required for new objects not in the dataset. We evaluated the performance of the CPN and the proposed 6D pose estimation method on benchmark datasets, which demonstrated that the CPN can predict the centroid for any object, including those not in the training data, and that our 6D pose estimation method outperforms existing methods for partially occluded objects. Finally, we tested a grasping task based on our proposed method using a real robot platform to demonstrate an application of our method to a downstream task. This experiment shows that our method can be applied to part assembly, bin picking, and object manipulation without large training datasets with 6D pose annotations. The code and models are available at: https://github.com/oorrppp2/Particle_filter_approach_6D_pose_estimation."
                    },
                    {
                        "title": "Panoramic Image-to-Image Translation",
                        "abstract": "In this paper, we tackle the challenging task of Panoramic Image-to-Image translation (Pano-I2I) for the first time. This task is difficult due to the geometric distortion of panoramic images and the lack of a panoramic image dataset with diverse conditions, like weather or time. To address these challenges, we propose a panoramic distortion-aware I2I model that preserves the structure of the panoramic images while consistently translating their global style referenced from a pinhole image. To mitigate the distortion issue in naive 360 panorama translation, we adopt spherical positional embedding to our transformer encoders, introduce a distortion-free discriminator, and apply sphere-based rotation for augmentation and its ensemble. We also design a content encoder and a style encoder to be deformation-aware to deal with a large domain gap between panoramas and pinhole images, enabling us to work on diverse conditions of pinhole images. In addition, considering the large discrepancy between panoramas and pinhole images, our framework decouples the learning procedure of the panoramic reconstruction stage from the translation stage. We show distinct improvements over existing I2I models in translating the StreetLearn dataset in the daytime into diverse conditions. The code will be publicly available online for our community."
                    },
                    {
                        "title": "Large Language Models are Frame-level Directors for Zero-shot Text-to-Video Generation",
                        "abstract": "In the paradigm of AI-generated content (AIGC), there has been increasing attention to transferring knowledge from pre-trained text-to-image (T2I) models to text-to-video (T2V) generation. Despite their effectiveness, these frameworks face challenges in maintaining consistent narratives and handling shifts in scene composition or object placement from a single abstract user prompt. Exploring the ability of large language models (LLMs) to generate time-dependent, frame-by-frame prompts, this paper introduces a new framework, dubbed DirecT2V. DirecT2V leverages instruction-tuned LLMs as directors, enabling the inclusion of time-varying content and facilitating consistent video generation. To maintain temporal consistency and prevent mapping the value to a different object, we equip a diffusion model with a novel value mapping method and dual-softmax filtering, which do not require any additional training. The experimental results validate the effectiveness of our framework in producing visually coherent and storyful videos from abstract user prompts, successfully addressing the challenges of zero-shot video generation."
                    },
                    {
                        "title": "GeCoNeRF: Few-shot Neural Radiance Fields via Geometric Consistency",
                        "abstract": "We present a novel framework to regularize Neural Radiance Field (NeRF) in a few-shot setting with a geometry-aware consistency regularization. The proposed approach leverages a rendered depth map at unobserved viewpoint to warp sparse input images to the unobserved viewpoint and impose them as pseudo ground truths to facilitate learning of NeRF. By encouraging such geometry-aware consistency at a feature-level instead of using pixel-level reconstruction loss, we regularize the NeRF at semantic and structural levels while allowing for modeling view dependent radiance to account for color variations across viewpoints. We also propose an effective method to filter out erroneous warped solutions, along with training strategies to stabilize training during optimization. We show that our model achieves competitive results compared to state-of-the-art few-shot NeRF models. Project page is available at https://ku-cvlab.github.io/GeCoNeRF/."
                    },
                    {
                        "title": "D\u00e4RF: Boosting Radiance Fields from Sparse Inputs with Monocular Depth Adaptation",
                        "abstract": "Neural radiance fields (NeRF) shows powerful performance in novel view synthesis and 3D geometry reconstruction, but it suffers from critical performance degradation when the number of known viewpoints is drastically reduced. Existing works attempt to overcome this problem by employing external priors, but their success is limited to certain types of scenes or datasets. Employing monocular depth estimation (MDE) networks, pretrained on large-scale RGB-D datasets, with powerful generalization capability would be a key to solving this problem: however, using MDE in conjunction with NeRF comes with a new set of challenges due to various ambiguity problems exhibited by monocular depths. In this light, we propose a novel framework, dubbed D\\\"aRF, that achieves robust NeRF reconstruction with a handful of real-world images by combining the strengths of NeRF and monocular depth estimation through online complementary training. Our framework imposes the MDE network's powerful geometry prior to NeRF representation at both seen and unseen viewpoints to enhance its robustness and coherence. In addition, we overcome the ambiguity problems of monocular depths through patch-wise scale-shift fitting and geometry distillation, which adapts the MDE network to produce depths aligned accurately with NeRF geometry. Experiments show our framework achieves state-of-the-art results both quantitatively and qualitatively, demonstrating consistent and reliable performance in both indoor and outdoor real-world datasets. Project page is available at https://ku-cvlab.github.io/DaRF/."
                    },
                    {
                        "title": "NeuralField-LDM: Scene Generation with Hierarchical Latent Diffusion Models",
                        "abstract": "Automatically generating high-quality real world 3D scenes is of enormous interest for applications such as virtual reality and robotics simulation. Towards this goal, we introduce NeuralField-LDM, a generative model capable of synthesizing complex 3D environments. We leverage Latent Diffusion Models that have been successfully utilized for efficient high-quality 2D content creation. We first train a scene auto-encoder to express a set of image and pose pairs as a neural field, represented as density and feature voxel grids that can be projected to produce novel views of the scene. To further compress this representation, we train a latent-autoencoder that maps the voxel grids to a set of latent representations. A hierarchical diffusion model is then fit to the latents to complete the scene generation pipeline. We achieve a substantial improvement over existing state-of-the-art scene generation models. Additionally, we show how NeuralField-LDM can be used for a variety of 3D content creation applications, including conditional scene generation, scene inpainting and scene style manipulation."
                    },
                    {
                        "title": "DreamTeacher: Pretraining Image Backbones with Deep Generative Models",
                        "abstract": "In this work, we introduce a self-supervised feature representation learning framework DreamTeacher that utilizes generative networks for pre-training downstream image backbones. We propose to distill knowledge from a trained generative model into standard image backbones that have been well engineered for specific perception tasks. We investigate two types of knowledge distillation: 1) distilling learned generative features onto target image backbones as an alternative to pretraining these backbones on large labeled datasets such as ImageNet, and 2) distilling labels obtained from generative networks with task heads onto logits of target backbones. We perform extensive analyses on multiple generative models, dense prediction benchmarks, and several pre-training regimes. We empirically find that our DreamTeacher significantly outperforms existing self-supervised representation learning approaches across the board. Unsupervised ImageNet pre-training with DreamTeacher leads to significant improvements over ImageNet classification pre-training on downstream datasets, showcasing generative models, and diffusion generative models specifically, as a promising approach to representation learning on large, diverse datasets without requiring manual annotation."
                    },
                    {
                        "title": "NICE: CVPR 2023 Challenge on Zero-shot Image Captioning",
                        "abstract": "In this report, we introduce NICE (New frontiers for zero-shot Image Captioning Evaluation) project1 and share the results and outcomes of 2023 challenge. This project is designed to challenge the computer vision community to develop robust image captioning models that advance the state-of-the-art both in terms of accuracy and fairness. Through the challenge, the image captioning models were tested using a new evaluation dataset that includes a large variety of visual concepts from many domains. There was no specific training data provided for the challenge, and therefore the challenge entries were required to adapt to new types of image descriptions that had not been seen during training. This report includes information on the newly proposed NICE dataset, evaluation methods, challenge results, and technical details of top-ranking entries. We expect that the outcomes of the challenge will contribute to the improvement of AI models on various vision-language tasks."
                    },
                    {
                        "title": "VisAlign: Dataset for Measuring the Degree of Alignment between AI and Humans in Visual Perception",
                        "abstract": "AI alignment refers to models acting towards human-intended goals, preferences, or ethical principles. Given that most large-scale deep learning models act as black boxes and cannot be manually controlled, analyzing the similarity between models and humans can be a proxy measure for ensuring AI safety. In this paper, we focus on the models' visual perception alignment with humans, further referred to as AI-human visual alignment. Specifically, we propose a new dataset for measuring AI-human visual alignment in terms of image classification, a fundamental task in machine perception. In order to evaluate AI-human visual alignment, a dataset should encompass samples with various scenarios that may arise in the real world and have gold human perception labels. Our dataset consists of three groups of samples, namely Must-Act (i.e., Must-Classify), Must-Abstain, and Uncertain, based on the quantity and clarity of visual information in an image and further divided into eight categories. All samples have a gold human perception label; even Uncertain (severely blurry) sample labels were obtained via crowd-sourcing. The validity of our dataset is verified by sampling theory, statistical theories related to survey design, and experts in the related fields. Using our dataset, we analyze the visual alignment and reliability of five popular visual perception models and seven abstention methods. Our code and data is available at https://github.com/jiyounglee-0523/VisAlign."
                    },
                    {
                        "title": "Align Your Latents: High-Resolution Video Synthesis with Latent Diffusion Models",
                        "abstract": "Latent Diffusion Models (LDMs) enable high-quality image synthesis while avoiding excessive compute demands by training a diffusion model in a compressed lower-dimensional latent space. Here, we apply the LDM paradigm to high-resolution video generation, a particularly resource-intensive task. We first pre-train an LDM on images only; then, we turn the image generator into a video generator by introducing a temporal dimension to the latent space diffusion model and finetuning on encoded image sequences, i.e., videos. Similarly, we temporally align diffusion model upsamplers, turning them into temporally consistent video super resolution models. We focus on two relevant real-world applications: Simulation of in-the-wild driving data and creative content creation with text-to-video modeling. In particular, we validate our Video LDM on real driving videos of resolution $512 \\times 1024$, achieving state-of-the-art performance. Furthermore, our approach can easily leverage off-the-shelf pretrained image LDMs, as we only need to train a temporal alignment model in that case. Doing so, we turn the publicly available, state-of-the-art text-to-image LDM Stable Diffusion into an efficient and expressive text-to-video model with resolution up to $1280 \\times 2048$. We show that the temporal layers trained in this way generalize to different finetuned text-to-image LDMs. Utilizing this property, we show the first results for personalized text-to-video generation, opening exciting directions for future content creation. Project page: https://nv-tlabs.github.io/VideoLDM/"
                    },
                    {
                        "title": "Controllable Style Transfer via Test-time Training of Implicit Neural Representation",
                        "abstract": "We propose a controllable style transfer framework based on Implicit Neural Representation that pixel-wisely controls the stylized output via test-time training. Unlike traditional image optimization methods that often suffer from unstable convergence and learning-based methods that require intensive training and have limited generalization ability, we present a model optimization framework that optimizes the neural networks during test-time with explicit loss functions for style transfer. After being test-time trained once, thanks to the flexibility of the INR-based model, our framework can precisely control the stylized images in a pixel-wise manner and freely adjust image resolution without further optimization or training. We demonstrate several applications."
                    },
                    {
                        "title": "AE-NeRF: Auto-Encoding Neural Radiance Fields for 3D-Aware Object Manipulation",
                        "abstract": "We propose a novel framework for 3D-aware object manipulation, called Auto-Encoding Neural Radiance Fields (AE-NeRF). Our model, which is formulated in an auto-encoder architecture, extracts disentangled 3D attributes such as 3D shape, appearance, and camera pose from an image, and a high-quality image is rendered from the attributes through disentangled generative Neural Radiance Fields (NeRF). To improve the disentanglement ability, we present two losses, global-local attribute consistency loss defined between input and output, and swapped-attribute classification loss. Since training such auto-encoding networks from scratch without ground-truth shape and appearance information is non-trivial, we present a stage-wise training scheme, which dramatically helps to boost the performance. We conduct experiments to demonstrate the effectiveness of the proposed model over the latest methods and provide extensive ablation studies."
                    },
                    {
                        "title": "Improving Sample Quality of Diffusion Models Using Self-Attention Guidance",
                        "abstract": "Denoising diffusion models (DDMs) have attracted attention for their exceptional generation quality and diversity. This success is largely attributed to the use of class- or text-conditional diffusion guidance methods, such as classifier and classifier-free guidance. In this paper, we present a more comprehensive perspective that goes beyond the traditional guidance methods. From this generalized perspective, we introduce novel condition- and training-free strategies to enhance the quality of generated images. As a simple solution, blur guidance improves the suitability of intermediate samples for their fine-scale information and structures, enabling diffusion models to generate higher quality samples with a moderate guidance scale. Improving upon this, Self-Attention Guidance (SAG) uses the intermediate self-attention maps of diffusion models to enhance their stability and efficacy. Specifically, SAG adversarially blurs only the regions that diffusion models attend to at each iteration and guides them accordingly. Our experimental re sults show that our SAG improves the performance of various diffusion models, including ADM, IDDPM, Stable Diffusion, and DiT. Moreover, combining SAG with conventional guidance methods leads to further improvement."
                    }
                ]
            }
        }
    },
    "2406.10019": {
        "paper_data": {
            "title": "Group and Shuffle: Efficient Structured Orthogonal Parametrization",
            "url": "http://arxiv.org/abs/2406.10019v1",
            "arxiv_id": "2406.10019",
            "authors": [
                "Mikhail Gorbunov",
                "Nikolay Yudin",
                "Vera Soboleva",
                "Aibek Alanov",
                "Alexey Naumov",
                "Maxim Rakhuba"
            ],
            "abstract": "The increasing size of neural networks has led to a growing demand for methods of efficient fine-tuning. Recently, an orthogonal fine-tuning paradigm was introduced that uses orthogonal matrices for adapting the weights of a pretrained model. In this paper, we introduce a new class of structured matrices, which unifies and generalizes structured classes from previous works. We examine properties of this class and build a structured orthogonal parametrization upon it. We then use this parametrization to modify the orthogonal fine-tuning framework, improving parameter and computational efficiency. We empirically validate our method on different domains, including adapting of text-to-image diffusion models and downstream task fine-tuning in language modeling. Additionally, we adapt our construction for orthogonal convolutions and conduct experiments with 1-Lipschitz neural networks.",
            "introduction": "   1 Introduction  Orthogonal transforms have proven useful in different deep learning tasks. For example, they were shown to stabilize CNNs\u00a0(Li et\u00a0al., 2019; Singla and Feizi, 2021) or used in RNNs to combat the problem of exploding/vanishing gradients (Arjovsky et\u00a0al., 2016). Recent works OFT (Orthogonal Fine-Tuning) and BOFT (Butterfly Orthogonal Fine-Tuning) (Qiu et\u00a0al., 2023; Liu et\u00a0al., 2024b) use learnable orthogonal matrices for parameter-efficient fine-tuning of neural networks, which prevents training instabilities and overfitting that alternative methods like LoRA (Hu et\u00a0al., 2022) suffer from.   Nevertheless, parametrization of orthogonal matrices is a challenging task, and the existing methods typically lack in either computational efficiency or expressiveness. Classical methods like Cayley parametrization and matrix exponential map cannot operate under low parameter budget, while Givens rotations and Householder reflections requires computing products of several matrices, which makes their use less efficient in deep learning tasks. Alternative approach in OFT method uses block-diagonal matrix structure in an attempt to be more computationally efficient and use less trainable parameters. Unfortunately, this simple structure can be too restrictive. Thus, arises the problem of constructing dense orthogonal matrix while still being parameter-efficient. While attempting to tackle this task, BOFT method uses a variation of butterfly matrices, parametrizing orthogonal matrices as a product of several matrices with different sparsity patterns, enforcing orthogonality on each of them. This parametrization is able to construct dense matrices while still being parameter-efficient. However it requires to compute a product of multiple matrices (typically up to 6666) which can be computationally expensive. In this paper, we aim to overcome these issues and build dense orthogonal matrices in a more efficient way.   We present a novel structured matrix class parametrized by an alternating product of block-diagonal matrices and several permutations. Multiplying by these matrices can be seen as a consecutive application of independent linear transforms within certain small groups and then shuffling the elements between them, hence the name Group-and-Shuffle matrices (or \ud835\udca2\u2062\ud835\udcae\ud835\udca2\ud835\udcae\\mathcal{GS}caligraphic_G caligraphic_S-matrices for short). This class generalizes Monarch matrices\u00a0(Dao et\u00a0al., 2022) and with the right permutation choices, is able to form dense orthogonal matrices more effectively compared to approach proposed in BOFT, decreasing number of matrices in the product as well as the number of trainable parameters. We build efficient structured orthogonal parametrization with this class and use it to construct a new parameter-efficient fine-tuning method named GSOFT.   Our contributions:   \u2022  We introduce a new class of structured matrices, called \ud835\udca2\u2062\ud835\udcae\ud835\udca2\ud835\udcae\\mathcal{GS}caligraphic_G caligraphic_S, that is more effective at forming dense matrices than block butterfly matrices from the BOFT method.    \u2022  Using \ud835\udca2\u2062\ud835\udcae\ud835\udca2\ud835\udcae\\mathcal{GS}caligraphic_G caligraphic_S-matrices, we propose an efficient structured orthogonal parametrization, provide theoretical insights and study its performance in the orthogonal fine-tuning framework.    \u2022  We adapt our ideas for convolutional architectures, providing a framework to compress and speed-up orthogonal convolution layers.        2 Orthogonal Fine-tuning  Orthogonal Fine-tuning method (OFT) introduced in (Qiu et\u00a0al., 2023) is a Parameter-Efficient Fine-Tuning (PEFT) method which fine-tunes pre-trained weight matrices through a learnable orthogonal block-diagonal matrix. Some of the properties that make orthogonal transforms desirable are preservation of pair-wise angles of neurons, spectral properties and hyperspherical energy. More precisely, OFT optimizes an orthogonal matrix Q\u2208\u211dd\u00d7d\ud835\udc44superscript\u211d\ud835\udc51\ud835\udc51Q\\in\\mathbb{R}^{d\\times d}italic_Q \u2208 blackboard_R start_POSTSUPERSCRIPT italic_d \u00d7 italic_d end_POSTSUPERSCRIPT for a pre-trained frozen weight matrix W0\u2208\u211dd\u00d7nsuperscript\ud835\udc4a0superscript\u211d\ud835\udc51\ud835\udc5bW^{0}\\in\\mathbb{R}^{d\\times n}italic_W start_POSTSUPERSCRIPT 0 end_POSTSUPERSCRIPT \u2208 blackboard_R start_POSTSUPERSCRIPT italic_d \u00d7 italic_n",
            "references": [
                {
                    "title": "DoRA: Weight-Decomposed Low-Rank Adaptation",
                    "abstract": "Among the widely used parameter-efficient fine-tuning (PEFT) methods, LoRA and its variants have gained considerable popularity because of avoiding additional inference costs. However, there still often exists an accuracy gap between these methods and full fine-tuning (FT). In this work, we first introduce a novel weight decomposition analysis to investigate the inherent differences between FT and LoRA. Aiming to resemble the learning capacity of FT from the findings, we propose Weight-Decomposed Low-Rank Adaptation (DoRA). DoRA decomposes the pre-trained weight into two components, magnitude and direction, for fine-tuning, specifically employing LoRA for directional updates to efficiently minimize the number of trainable parameters. By employing \\ours, we enhance both the learning capacity and training stability of LoRA while avoiding any additional inference overhead. \\ours~consistently outperforms LoRA on fine-tuning LLaMA, LLaVA, and VL-BART on various downstream tasks, such as commonsense reasoning, visual instruction tuning, and image/video-text understanding. Code is available at https://github.com/NVlabs/DoRA."
                },
                {
                    "title": "Parameter-Efficient Orthogonal Finetuning via Butterfly Factorization",
                    "abstract": "Large foundation models are becoming ubiquitous, but training them from scratch is prohibitively expensive. Thus, efficiently adapting these powerful models to downstream tasks is increasingly important. In this paper, we study a principled finetuning paradigm -- Orthogonal Finetuning (OFT) -- for downstream task adaptation. Despite demonstrating good generalizability, OFT still uses a fairly large number of trainable parameters due to the high dimensionality of orthogonal matrices. To address this, we start by examining OFT from an information transmission perspective, and then identify a few key desiderata that enable better parameter-efficiency. Inspired by how the Cooley-Tukey fast Fourier transform algorithm enables efficient information transmission, we propose an efficient orthogonal parameterization using butterfly structures. We apply this parameterization to OFT, creating a novel parameter-efficient finetuning method, called Orthogonal Butterfly (BOFT). By subsuming OFT as a special case, BOFT introduces a generalized orthogonal finetuning framework. Finally, we conduct an extensive empirical study of adapting large vision transformers, large language models, and text-to-image diffusion models to various downstream tasks in vision and language."
                },
                {
                    "title": "Controlling Text-to-Image Diffusion by Orthogonal Finetuning",
                    "abstract": "Large text-to-image diffusion models have impressive capabilities in generating photorealistic images from text prompts. How to effectively guide or control these powerful models to perform different downstream tasks becomes an important open problem. To tackle this challenge, we introduce a principled finetuning method -- Orthogonal Finetuning (OFT), for adapting text-to-image diffusion models to downstream tasks. Unlike existing methods, OFT can provably preserve hyperspherical energy which characterizes the pairwise neuron relationship on the unit hypersphere. We find that this property is crucial for preserving the semantic generation ability of text-to-image diffusion models. To improve finetuning stability, we further propose Constrained Orthogonal Finetuning (COFT) which imposes an additional radius constraint to the hypersphere. Specifically, we consider two important finetuning text-to-image tasks: subject-driven generation where the goal is to generate subject-specific images given a few images of a subject and a text prompt, and controllable generation where the goal is to enable the model to take in additional control signals. We empirically show that our OFT framework outperforms existing methods in generation quality and convergence speed."
                },
                {
                    "title": "QLoRA: Efficient Finetuning of Quantized LLMs",
                    "abstract": "We present QLoRA, an efficient finetuning approach that reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU while preserving full 16-bit finetuning task performance. QLoRA backpropagates gradients through a frozen, 4-bit quantized pretrained language model into Low Rank Adapters~(LoRA). Our best model family, which we name Guanaco, outperforms all previous openly released models on the Vicuna benchmark, reaching 99.3% of the performance level of ChatGPT while only requiring 24 hours of finetuning on a single GPU. QLoRA introduces a number of innovations to save memory without sacrificing performance: (a) 4-bit NormalFloat (NF4), a new data type that is information theoretically optimal for normally distributed weights (b) double quantization to reduce the average memory footprint by quantizing the quantization constants, and (c) paged optimziers to manage memory spikes. We use QLoRA to finetune more than 1,000 models, providing a detailed analysis of instruction following and chatbot performance across 8 instruction datasets, multiple model types (LLaMA, T5), and model scales that would be infeasible to run with regular finetuning (e.g. 33B and 65B parameter models). Our results show that QLoRA finetuning on a small high-quality dataset leads to state-of-the-art results, even when using smaller models than the previous SoTA. We provide a detailed analysis of chatbot performance based on both human and GPT-4 evaluations showing that GPT-4 evaluations are a cheap and reasonable alternative to human evaluation. Furthermore, we find that current chatbot benchmarks are not trustworthy to accurately evaluate the performance levels of chatbots. A lemon-picked analysis demonstrates where Guanaco fails compared to ChatGPT. We release all of our models and code, including CUDA kernels for 4-bit training."
                },
                {
                    "title": "Key-Locked Rank One Editing for Text-to-Image Personalization",
                    "abstract": "Text-to-image models (T2I) offer a new level of flexibility by allowing users to guide the creative process through natural language. However, personalizing these models to align with user-provided visual concepts remains a challenging problem. The task of T2I personalization poses multiple hard challenges, such as maintaining high visual fidelity while allowing creative control, combining multiple personalized concepts in a single image, and keeping a small model size. We present Perfusion, a T2I personalization method that addresses these challenges using dynamic rank-1 updates to the underlying T2I model. Perfusion avoids overfitting by introducing a new mechanism that \u201clocks\u201d new concepts\u2019 cross-attention Keys to their superordinate category. Additionally, we develop a gated rank-1 approach that enables us to control the influence of a learned concept during inference time and to combine multiple concepts. This allows runtime efficient balancing of visual-fidelity and textual-alignment with a single 100KB trained model. Importantly, it can span different operating points across the Pareto front without additional training. We compare our approach to strong baselines and demonstrate its qualitative and quantitative strengths."
                },
                {
                    "title": "SVDiff: Compact Parameter Space for Diffusion Fine-Tuning",
                    "abstract": "Diffusion models have achieved remarkable success in text-to-image generation, enabling the creation of high-quality images from text prompts or other modalities. However, existing methods for customizing these models are limited by handling multiple personalized subjects and the risk of overfitting. Moreover, their large number of parameters is inefficient for model storage. In this paper, we propose a novel approach to address these limitations in existing text-to-image diffusion models for personalization. Our method involves fine-tuning the singular values of the weight matrices, leading to a compact and efficient parameter space that reduces the risk of overfitting and language-drifting. We also propose a Cut-Mix-Unmix data-augmentation technique to enhance the quality of multi-subject image generation and a simple text-based image editing framework. Our proposed SVDiff method has a significantly smaller model size compared to existing methods (\u22482,200 times fewer parameters compared with vanilla DreamBooth), making it more practical for real-world applications."
                },
                {
                    "title": "ELITE: Encoding Visual Concepts into Textual Embeddings for Customized Text-to-Image Generation",
                    "abstract": "In addition to the unprecedented ability in imaginary creation, large text-to-image models are expected to take customized concepts in image generation. Existing works generally learn such concepts in an optimization-based manner, yet bringing excessive computation or memory burden. In this paper, we instead propose a learning-based encoder, which consists of a global and a local mapping networks for fast and accurate customized text-to-image generation. In specific, the global mapping network projects the hierarchical features of a given image into multiple \"new\" words in the textual word embedding space, i.e., one primary word for well-editable concept and other auxiliary words to exclude irrelevant disturbances (e.g., background). In the meantime, a local mapping network injects the encoded patch features into cross attention layers to provide omitted details, without sacrificing the editability of primary concepts. We compare our method with existing optimization-based approaches on a variety of user-defined concepts, and demonstrate that our method enables high-fidelity inversion and more robust editability with a significantly faster encoding process. Our code is publicly available at https://github.com/csyxwei/ELITE."
                },
                {
                    "title": "KronA: Parameter Efficient Tuning with Kronecker Adapter",
                    "abstract": "Fine-tuning a Pre-trained Language Model (PLM) on a specific downstream task has been a well-known paradigm in Natural Language Processing. However, with the ever-growing size of PLMs, training the entire model on several downstream tasks becomes very expensive and resource-hungry. Recently, different Parameter Efficient Tuning (PET) techniques are proposed to improve the efficiency of fine-tuning PLMs. One popular category of PET methods is the low-rank adaptation methods which insert learnable truncated SVD modules into the original model either sequentially or in parallel. However, low-rank decomposition suffers from limited representation power. In this work, we address this problem using the Kronecker product instead of the low-rank representation. We introduce KronA, a Kronecker product-based adapter module for efficient fine-tuning of Transformer-based PLMs. We apply the proposed methods for fine-tuning T5 on the GLUE benchmark to show that incorporating the Kronecker-based modules can outperform state-of-the-art PET methods."
                },
                {
                    "title": "Multi-Concept Customization of Text-to-Image Diffusion",
                    "abstract": "While generative models produce high-quality images of concepts learned from a large-scale database, a user often wishes to synthesize instantiations of their own concepts (for example, their family, pets, or items). Can we teach a model to quickly acquire a new concept, given a few examples? Furthermore, can we compose multiple new concepts together? We propose Custom Diffusion, an efficient method for augmenting existing text-to-image models. We find that only optimizing a few parameters in the text-to-image conditioning mechanism is sufficiently powerful to represent new concepts while enabling fast tuning (~ 6 minutes). Additionally, we can jointly train for multiple concepts or combine multiple fine-tuned models into one via closed-form constrained optimization. Our fine-tuned model generates variations of multiple new concepts and seamlessly composes them with existing concepts in novel settings. Our method outperforms or performs on par with several baselines and concurrent works in both qualitative and quantitative evaluations, while being memory and computationally efficient."
                },
                {
                    "title": "DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation",
                    "abstract": "Large text-to-image models achieved a remarkable leap in the evolution of AI, enabling high-quality and diverse synthesis of images from a given text prompt. However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts. In this work, we present a new approach for \u201cpersonalization\u201d of text-to-image diffusion models. Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject. Once the subject is embedded in the output domain of the model, the unique identifier can be used to synthesize novel photorealistic images of the subject contextualized in different scenes. By leveraging the semantic prior embedded in the model with a new autogenous class-specific prior preservation loss, our technique enables synthesizing the subject in diverse scenes, poses, views and lighting conditions that do not appear in the reference images. We apply our technique to several previously-unassailable tasks, including subject recontextualization, text-guided view synthesis, and artistic rendering, all while preserving the subject's key features. We also provide a new dataset and evaluation protocol for this new task of subject-driven generation. Project page: https://dreambooth.github.io/"
                },
                {
                    "title": "An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion",
                    "abstract": "Text-to-image models offer unprecedented freedom to guide creation through natural language. Yet, it is unclear how such freedom can be exercised to generate images of specific unique concepts, modify their appearance, or compose them in new roles and novel scenes. In other words, we ask: how can we use language-guided models to turn our cat into a painting, or imagine a new product based on our favorite toy? Here we present a simple approach that allows such creative freedom. Using only 3-5 images of a user-provided concept, like an object or a style, we learn to represent it through new\"words\"in the embedding space of a frozen text-to-image model. These\"words\"can be composed into natural language sentences, guiding personalized creation in an intuitive way. Notably, we find evidence that a single word embedding is sufficient for capturing unique and varied concepts. We compare our approach to a wide range of baselines, and demonstrate that it can more faithfully portray the concepts across a range of applications and tasks. Our code, data and new words will be available at: https://textual-inversion.github.io"
                },
                {
                    "title": "Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding",
                    "abstract": "We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding. Imagen builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation. Our key discovery is that generic large language models (e.g. T5), pretrained on text-only corpora, are surprisingly effective at encoding text for image synthesis: increasing the size of the language model in Imagen boosts both sample fidelity and image-text alignment much more than increasing the size of the image diffusion model. Imagen achieves a new state-of-the-art FID score of 7.27 on the COCO dataset, without ever training on COCO, and human raters find Imagen samples to be on par with the COCO data itself in image-text alignment. To assess text-to-image models in greater depth, we introduce DrawBench, a comprehensive and challenging benchmark for text-to-image models. With DrawBench, we compare Imagen with recent methods including VQ-GAN+CLIP, Latent Diffusion Models, and DALL-E 2, and find that human raters prefer Imagen over other models in side-by-side comparisons, both in terms of sample quality and image-text alignment. See https://imagen.research.google/ for an overview of the results."
                },
                {
                    "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
                    "abstract": "Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples."
                },
                {
                    "title": "Monarch: Expressive Structured Matrices for Efficient and Accurate Training",
                    "abstract": "Large neural networks excel in many domains, but they are expensive to train and fine-tune. A popular approach to reduce their compute or memory requirements is to replace dense weight matrices with structured ones (e.g., sparse, low-rank, Fourier transform). These methods have not seen widespread adoption (1) in end-to-end training due to unfavorable efficiency--quality tradeoffs, and (2) in dense-to-sparse fine-tuning due to lack of tractable algorithms to approximate a given dense weight matrix. To address these issues, we propose a class of matrices (Monarch) that is hardware-efficient (they are parameterized as products of two block-diagonal matrices for better hardware utilization) and expressive (they can represent many commonly used transforms). Surprisingly, the problem of approximating a dense weight matrix with a Monarch matrix, though nonconvex, has an analytical optimal solution. These properties of Monarch matrices unlock new ways to train and fine-tune sparse and dense models. We empirically validate that Monarch can achieve favorable accuracy-efficiency tradeoffs in several end-to-end sparse training applications: speeding up ViT and GPT-2 training on ImageNet classification and Wikitext-103 language modeling by 2x with comparable model quality, and reducing the error on PDE solving and MRI reconstruction tasks by 40%. In sparse-to-dense training, with a simple technique called\"reverse sparsification,\"Monarch matrices serve as a useful intermediate representation to speed up GPT-2 pretraining on OpenWebText by 2x without quality drop. The same technique brings 23% faster BERT pretraining than even the very optimized implementation from Nvidia that set the MLPerf 1.1 record. In dense-to-sparse fine-tuning, as a proof-of-concept, our Monarch approximation algorithm speeds up BERT fine-tuning on GLUE by 1.7x with comparable accuracy."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Pixelated Butterfly: Simple and Efficient Sparse training for Neural Network Models",
                    "abstract": "Overparameterized neural networks generalize well but are expensive to train. Ideally, one would like to reduce their computational cost while retaining their generalization benefits. Sparse model training is a simple and promising approach to achieve this, but there remain challenges as existing methods struggle with accuracy loss, slow training runtime, or difficulty in sparsifying all model components. The core problem is that searching for a sparsity mask over a discrete set of sparse matrices is difficult and expensive. To address this, our main insight is to optimize over a continuous superset of sparse matrices with a fixed structure known as products of butterfly matrices. As butterfly matrices are not hardware efficient, we propose simple variants of butterfly (block and flat) to take advantage of modern hardware. Our method (Pixelated Butterfly) uses a simple fixed sparsity pattern based on flat block butterfly and low-rank matrices to sparsify most network layers (e.g., attention, MLP). We empirically validate that Pixelated Butterfly is 3x faster than butterfly and speeds up training to achieve favorable accuracy--efficiency tradeoffs. On the ImageNet classification and WikiText-103 language modeling tasks, our sparse models train up to 2.5x faster than the dense MLP-Mixer, Vision Transformer, and GPT-2 medium with no drop in accuracy."
                },
                {
                    "title": "Improved deterministic l2 robustness on CIFAR-10 and CIFAR-100",
                    "abstract": "Training convolutional neural networks (CNNs) with a strict Lipschitz constraint under the l2 norm is useful for provable adversarial robustness, interpretable gradients and stable training. While 1-Lipschitz CNNs can be designed by enforcing a 1-Lipschitz constraint on each layer, training such networks requires each layer to have an orthogonal Jacobian matrix (for all inputs) to prevent the gradients from vanishing during backpropagation. A layer with this property is said to be Gradient Norm Preserving (GNP). In this work, we introduce a procedure to certify the robustness of 1-Lipschitz CNNs by relaxing the orthogonalization of the last linear layer of the network that significantly advances the state of the art for both standard and provable robust accuracies on CIFAR-100 (gains of 4.80% and 4.71%, respectively). We further boost their robustness by introducing (i) a novel Gradient Norm preserving activation function called the Householder activation function (that includes every GroupSort activation) and (ii) a certificate regularization. On CIFAR-10, we achieve significant improvements over prior works in provable robust accuracy (5.81%) with only a minor drop in standard accuracy (\u22120.29%). Code for reproducing all experiments in the paper is available at https://github.com/singlasahil14/SOC."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Compacter: Efficient Low-Rank Hypercomplex Adapter Layers",
                    "abstract": "Adapting large-scale pretrained language models to downstream tasks via fine-tuning is the standard method for achieving state-of-the-art performance on NLP benchmarks. However, fine-tuning all weights of models with millions or billions of parameters is sample-inefficient, unstable in low-resource settings, and wasteful as it requires storing a separate copy of the model for each task. Recent work has developed parameter-efficient fine-tuning methods, but these approaches either still require a relatively large number of parameters or underperform standard fine-tuning. In this work, we propose Compacter, a method for fine-tuning large-scale language models with a better trade-off between task performance and the number of trainable parameters than prior work. Compacter accomplishes this by building on top of ideas from adapters, low-rank optimization, and parameterized hypercomplex multiplication layers. Specifically, Compacter inserts task-specific weight matrices into a pretrained model's weights, which are computed efficiently as a sum of Kronecker products between shared\"slow\"weights and\"fast\"rank-one matrices defined per Compacter layer. By only training 0.047% of a pretrained model's parameters, Compacter performs on par with standard fine-tuning on GLUE and outperforms standard fine-tuning on SuperGLUE and low-resource settings. Our code is publicly available at~\\url{https://github.com/rabeehk/compacter}."
                },
                {
                    "title": "Skew Orthogonal Convolutions",
                    "abstract": "Training convolutional neural networks with a Lipschitz constraint under the $l_{2}$ norm is useful for provable adversarial robustness, interpretable gradients, stable training, etc. While 1-Lipschitz networks can be designed by imposing a 1-Lipschitz constraint on each layer, training such networks requires each layer to be gradient norm preserving (GNP) to prevent gradients from vanishing. However, existing GNP convolutions suffer from slow training, lead to significant reduction in accuracy and provide no guarantees on their approximations. In this work, we propose a GNP convolution layer called Skew Orthogonal Convolution (SOC) that uses the following mathematical property: when a matrix is {\\it Skew-Symmetric}, its exponential function is an {\\it orthogonal} matrix. To use this property, we first construct a convolution filter whose Jacobian is Skew-Symmetric. Then, we use the Taylor series expansion of the Jacobian exponential to construct the SOC layer that is orthogonal. To efficiently implement SOC, we keep a finite number of terms from the Taylor series and provide a provable guarantee on the approximation error. Our experiments on CIFAR-10 and CIFAR-100 show that SOC allows us to train provably Lipschitz, large convolutional neural networks significantly faster than prior works while achieving significant improvements for both standard and certified robust accuracies."
                },
                {
                    "title": "The Power of Scale for Parameter-Efficient Prompt Tuning",
                    "abstract": "In this work, we explore \u201cprompt tuning,\u201d a simple yet effective mechanism for learning \u201csoft prompts\u201d to condition frozen language models to perform specific downstream tasks. Unlike the discrete text prompts used by GPT-3, soft prompts are learned through backpropagation and can be tuned to incorporate signals from any number of labeled examples. Our end-to-end learned approach outperforms GPT-3\u2019s few-shot learning by a large margin. More remarkably, through ablations on model size using T5, we show that prompt tuning becomes more competitive with scale: as models exceed billions of parameters, our method \u201ccloses the gap\u201d and matches the strong performance of model tuning (where all model weights are tuned). This finding is especially relevant because large models are costly to share and serve and the ability to reuse one frozen model for multiple downstream tasks can ease this burden. Our method can be seen as a simplification of the recently proposed \u201cprefix tuning\u201d of Li and Liang (2021) and we provide a comparison to this and other similar approaches. Finally, we show that conditioning a frozen model with soft prompts confers benefits in robustness to domain transfer and enables efficient \u201cprompt ensembling.\u201d We release code and model checkpoints to reproduce our experiments."
                },
                {
                    "title": "Zero-Shot Text-to-Image Generation",
                    "abstract": "Text-to-image generation has traditionally focused on finding better modeling assumptions for training on a fixed dataset. These assumptions might involve complex architectures, auxiliary losses, or side information such as object part labels or segmentation masks supplied during training. We describe a simple approach for this task based on a transformer that autoregressively models the text and image tokens as a single stream of data. With sufficient data and scale, our approach is competitive with previous domain-specific models when evaluated in a zero-shot fashion."
                },
                {
                    "title": "Preventing Gradient Attenuation in Lipschitz Constrained Convolutional Networks",
                    "abstract": "Lipschitz constraints under L2 norm on deep neural networks are useful for provable adversarial robustness bounds, stable training, and Wasserstein distance estimation. While heuristic approaches such as the gradient penalty have seen much practical success, it is challenging to achieve similar practical performance while provably enforcing a Lipschitz constraint. In principle, one can design Lipschitz constrained architectures using the composition property of Lipschitz functions, but Anil et al. recently identified a key obstacle to this approach: gradient norm attenuation. They showed how to circumvent this problem in the case of fully connected networks by designing each layer to be gradient norm preserving. We extend their approach to train scalable, expressive, provably Lipschitz convolutional networks. In particular, we present the Block Convolution Orthogonal Parameterization (BCOP), an expressive parameterization of orthogonal convolution operations. We show that even though the space of orthogonal convolutions is disconnected, the largest connected component of BCOP with 2n channels can represent arbitrary BCOP convolutions over n channels. Our BCOP parameterization allows us to train large convolutional networks with provable Lipschitz bounds. Empirically, we find that it is competitive with existing approaches to provable adversarial robustness and Wasserstein distance estimation."
                },
                {
                    "title": "RoBERTa: A Robustly Optimized BERT Pretraining Approach",
                    "abstract": "Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code."
                },
                {
                    "title": "Parameter-Efficient Transfer Learning for NLP",
                    "abstract": "Fine-tuning large pre-trained models is an effective transfer mechanism in NLP. However, in the presence of many downstream tasks, fine-tuning is parameter inefficient: an entire new model is required for every task. As an alternative, we propose transfer with adapter modules. Adapter modules yield a compact and extensible model; they add only a few trainable parameters per task, and new tasks can be added without revisiting previous ones. The parameters of the original network remain fixed, yielding a high degree of parameter sharing. To demonstrate adapter's effectiveness, we transfer the recently proposed BERT Transformer model to 26 diverse text classification tasks, including the GLUE benchmark. Adapters attain near state-of-the-art performance, whilst adding only a few parameters per task. On GLUE, we attain within 0.4% of the performance of full fine-tuning, adding only 3.6% parameters per task. By contrast, fine-tuning trains 100% of the parameters per task."
                },
                {
                    "title": "Sorting out Lipschitz function approximation",
                    "abstract": "Training neural networks under a strict Lipschitz constraint is useful for provable adversarial robustness, generalization bounds, interpretable gradients, and Wasserstein distance estimation. By the composition property of Lipschitz functions, it suffices to ensure that each individual affine transformation or nonlinear activation is 1-Lipschitz. The challenge is to do this while maintaining the expressive power. We identify a necessary property for such an architecture: each of the layers must preserve the gradient norm during backpropagation. Based on this, we propose to combine a gradient norm preserving activation function, GroupSort, with norm-constrained weight matrices. We show that norm-constrained GroupSort architectures are universal Lipschitz function approximators. Empirically, we show that norm-constrained GroupSort networks achieve tighter estimates of Wasserstein distance than their ReLU counterparts and can achieve provable adversarial robustness guarantees with little cost to accuracy."
                },
                {
                    "title": "GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding",
                    "abstract": "Human ability to understand language is general, flexible, and robust. In contrast, most NLU models above the word level are designed for a specific task and struggle with out-of-domain data. If we aspire to develop models with understanding beyond the detection of superficial correspondences between inputs and outputs, then it is critical to develop a unified model that can execute a range of linguistic tasks across different domains. To facilitate research in this direction, we present the General Language Understanding Evaluation (GLUE, gluebenchmark.com): a benchmark of nine diverse NLU tasks, an auxiliary dataset for probing models for understanding of specific linguistic phenomena, and an online platform for evaluating and comparing models. For some benchmark tasks, training data is plentiful, but for others it is limited or does not match the genre of the test set. GLUE thus favors models that can represent linguistic knowledge in a way that facilitates sample-efficient learning and effective knowledge-transfer across tasks. While none of the datasets in GLUE were created from scratch for the benchmark, four of them feature privately-held test data, which is used to ensure that the benchmark is used fairly. We evaluate baselines that use ELMo (Peters et al., 2018), a powerful transfer learning technique, as well as state-of-the-art sentence representation models. The best models still achieve fairly low absolute scores. Analysis with our diagnostic dataset yields similarly weak performance over all phenomena tested, with some exceptions."
                },
                {
                    "title": "ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices",
                    "abstract": "We introduce an extremely computation-efficient CNN architecture named ShuffleNet, which is designed specially for mobile devices with very limited computing power (e.g., 10-150 MFLOPs). The new architecture utilizes two new operations, pointwise group convolution and channel shuffle, to greatly reduce computation cost while maintaining accuracy. Experiments on ImageNet classification and MS COCO object detection demonstrate the superior performance of ShuffleNet over other structures, e.g. lower top-1 error (absolute 7.8%) than recent MobileNet [12] on ImageNet classification task, under the computation budget of 40 MFLOPs. On an ARM-based mobile device, ShuffleNet achieves ~13\u00c3\u2014 actual speedup over AlexNet while maintaining comparable accuracy."
                },
                {
                    "title": "Unitary Evolution Recurrent Neural Networks",
                    "abstract": "Recurrent neural networks (RNNs) are notoriously difficult to train. When the eigenvalues of the hidden to hidden weight matrix deviate from absolute value 1, optimization becomes difficult due to the well studied issue of vanishing and exploding gradients, especially when trying to learn long-term dependencies. To circumvent this problem, we propose a new architecture that learns a unitary weight matrix, with eigenvalues of absolute value exactly 1. The challenge we address is that of parametrizing unitary matrices in a way that does not require expensive computations (such as eigendecomposition) after each weight update. We construct an expressive unitary weight matrix by composing several structured matrices that act as building blocks with parameters to be learned. Optimization with this parameterization becomes feasible only when considering hidden states in the complex domain. We demonstrate the potential of this architecture by achieving state of the art results in several hard tasks involving very long-term dependencies."
                },
                {
                    "title": "Tensorizing Neural Networks",
                    "abstract": "Deep neural networks currently demonstrate state-of-the-art performance in several domains. At the same time, models of this class are very demanding in terms of computational resources. In particular, a large amount of memory is required by commonly used fully-connected layers, making it hard to use the models on low-end devices and stopping the further increase of the model size. In this paper we convert the dense weight matrices of the fully-connected layers to the Tensor Train [17] format such that the number of parameters is reduced by a huge factor and at the same time the expressive power of the layer is preserved. In particular, for the Very Deep VGG networks [21] we report the compression factor of the dense weight matrix of a fully-connected layer up to 200000 times leading to the compression factor of the whole network up to 7 times."
                },
                {
                    "title": "Speeding-up Convolutional Neural Networks Using Fine-tuned CP-Decomposition",
                    "abstract": "We propose a simple two-step approach for speeding up convolution layers within large convolutional neural networks based on tensor decomposition and discriminative fine-tuning. Given a layer, we use non-linear least squares to compute a low-rank CP-decomposition of the 4D convolution kernel tensor into a sum of a small number of rank-one tensors. At the second step, this decomposition is used to replace the original convolutional layer with a sequence of four convolutional layers with small kernels. After such replacement, the entire network is fine-tuned on the training data using standard backpropagation process. \nWe evaluate this approach on two CNNs and show that it is competitive with previous approaches, leading to higher obtained CPU speedups at the cost of lower accuracy drops for the smaller of the two networks. Thus, for the 36-class character classification CNN, our approach obtains a 8.5x CPU speedup of the whole network with only minor accuracy drop (1% from 91% to 90%). For the standard ImageNet architecture (AlexNet), the approach speeds up the second convolution layer by a factor of 4x at the cost of $1\\%$ increase of the overall top-5 classification error."
                },
                {
                    "title": "Adaptive Budget Allocation for Parameter-Efficient Fine-Tuning",
                    "abstract": ", question answering and natural language generation tasks. Results show that AdaLoRA outperforms existing approaches."
                },
                {
                    "title": "Prefix-Tuning: Optimizing Continuous Prompts for Generation",
                    "abstract": "Fine-tuning is the de facto way of leveraging large pretrained language models for downstream tasks. However, fine-tuning modifies all the language model parameters and therefore necessitates storing a full copy for each task. In this paper, we propose prefix-tuning, a lightweight alternative to fine-tuning for natural language generation tasks, which keeps language model parameters frozen and instead optimizes a sequence of continuous task-specific vectors, which we call the prefix. Prefix-tuning draws inspiration from prompting for language models, allowing subsequent tokens to attend to this prefix as if it were \u201cvirtual tokens\u201d. We apply prefix-tuning to GPT-2 for table-to-text generation and to BART for summarization. We show that by learning only 0.1% of the parameters, prefix-tuning obtains comparable performance in the full data setting, outperforms fine-tuning in low-data settings, and extrapolates better to examples with topics that are unseen during training."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CL",
                "cs.CV",
                "cs.NA",
                "math.NA"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we construct dense orthogonal matrices in a parameter-efficient manner while overcoming the computational inefficiencies and expressiveness limitations of existing methods?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning, particularly in improving the stability and performance of neural networks during fine-tuning. By developing a more efficient method for constructing dense orthogonal matrices, we can enhance the capabilities of existing fine-tuning techniques, leading to better model performance and generalization. This research could pave the way for new applications in various domains, such as computer vision and natural language processing, where robust and efficient models are essential.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of parametrizing orthogonal matrices while maintaining computational efficiency. Naive approaches, such as classical methods like Cayley parametrization or matrix exponential maps, fail under low parameter budgets, while methods like Givens rotations and Householder reflections require extensive matrix multiplications, making them impractical for deep learning tasks. Additionally, existing structured methods, like block-diagonal matrices, can be overly restrictive, limiting their expressiveness and effectiveness in forming dense orthogonal matrices.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either computational efficiency or expressiveness, but not both simultaneously. Existing methods have limitations, such as the inability to operate under low parameter budgets or the computational burden of matrix products. Barriers include the lack of a suitable framework that balances these needs. Our approach differs by introducing a novel class of structured matrices (Group-and-Shuffle matrices) that generalizes previous work and allows for a more effective construction of dense orthogonal matrices with fewer parameters and reduced computational costs.\n\n**[Question 5] - What are the key components of my approach and results?**  \nWe propose a new methodology that utilizes Group-and-Shuffle matrices, which are parametrized by an alternating product of block-diagonal matrices and permutations. This approach allows for efficient construction of dense orthogonal matrices while minimizing the number of trainable parameters. We will evaluate our method using standard datasets and metrics for fine-tuning performance, expecting to demonstrate improved efficiency and effectiveness in orthogonal fine-tuning compared to existing methods. Additionally, we will adapt our framework for convolutional architectures to enhance the performance of orthogonal convolution layers."
            }
        },
        "author_data": {
            "a7c6cadf-f76b-41bf-9507-678f0708fbba": {
                "pk": "a7c6cadf-f76b-41bf-9507-678f0708fbba",
                "name": "Mikhail Gorbunov",
                "collaborators": [
                    "Ekaterina Grishina",
                    "Maxim Rakhuba"
                ],
                "domain": [
                    "Convolutional Neural Networks",
                    "Robustness",
                    "Generalization",
                    "Spectral Norm"
                ],
                "publications": [
                    {
                        "title": "Tight and Efficient Upper Bound on Spectral Norm of Convolutional Layers",
                        "abstract": "Controlling the spectral norm of the Jacobian matrix, which is related to the convolution operation, has been shown to improve generalization, training stability and robustness in CNNs. Existing methods for computing the norm either tend to overestimate it or their performance may deteriorate quickly with increasing the input and kernel sizes. In this paper, we demonstrate that the tensor version of the spectral norm of a four-dimensional convolution kernel, up to a constant factor, serves as an upper bound for the spectral norm of the Jacobian matrix associated with the convolution operation. This new upper bound is independent of the input image resolution, differentiable and can be efficiently calculated during training. Through experiments, we demonstrate how this new bound can be used to improve the performance of convolutional architectures."
                    }
                ]
            },
            "26f1be35-b4b0-4ee5-9e76-0dad4fd2aab6": {
                "pk": "26f1be35-b4b0-4ee5-9e76-0dad4fd2aab6",
                "name": "Nikolay Yudin",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "76d839e7-5bd1-45f6-9db1-2769e3e12f8c": {
                "pk": "76d839e7-5bd1-45f6-9db1-2769e3e12f8c",
                "name": "Vera Soboleva",
                "collaborators": [
                    "Oleg Shipitko",
                    "Mikhail Goncharov",
                    "Anvar Kurmukov",
                    "Maxim Pisov",
                    "Mikhail Belyaev"
                ],
                "domain": [
                    "Computer Vision",
                    "Self-Supervised Learning",
                    "Medical Imaging",
                    "Autonomous Vehicles"
                ],
                "publications": [
                    {
                        "title": "Raindrops on Windshield: Dataset and Lightweight Gradient-Based Detection Algorithm",
                        "abstract": "Autonomous vehicles use cameras as one of the primary sources of information about the environment. Adverse weather conditions such as raindrops, snow, mud, and others, can lead to various image artifacts. Such artifacts significantly degrade the quality and reliability of the obtained visual data and can lead to accidents if they are not detected in time. This paper presents ongoing work on a new dataset for training and assessing vision algorithms' performance for different tasks of image artifacts detection on either camera lens or windshield. At the moment, we present a publicly available set of images containing $8190$ images, of which $3390$ contain raindrops. Images are annotated with the binary mask representing areas with raindrops. We demonstrate the applicability of the dataset in the problems of raindrops presence detection and raindrop region segmentation. To augment the data, we also propose an algorithm for data augmentation which allows the generation of synthetic raindrops on images. Apart from the dataset, we present a novel gradient-based algorithm for raindrop presence detection in a video sequence. The experimental evaluation proves that the algorithm reliably detects raindrops. Moreover, compared with the state-of-the-art cross-correlation-based algorithm \\cite{Einecke2014}, the proposed algorithm showed a higher quality of raindrop presence detection and image processing speed, making it applicable for the self-check procedure of real autonomous systems. The dataset is available at \\href{https://github.com/EvoCargo/RaindropsOnWindshield}{$github.com/EvoCargo/RaindropsOnWindshield$}."
                    },
                    {
                        "title": "vox2vec: A Framework for Self-supervised Contrastive Learning of Voxel-level Representations in Medical Images",
                        "abstract": "This paper introduces vox2vec - a contrastive method for self-supervised learning (SSL) of voxel-level representations. vox2vec representations are modeled by a Feature Pyramid Network (FPN): a voxel representation is a concatenation of the corresponding feature vectors from different pyramid levels. The FPN is pre-trained to produce similar representations for the same voxel in different augmented contexts and distinctive representations for different voxels. This results in unified multi-scale representations that capture both global semantics (e.g., body part) and local semantics (e.g., different small organs or healthy versus tumor tissue). We use vox2vec to pre-train a FPN on more than 6500 publicly available computed tomography images. We evaluate the pre-trained representations by attaching simple heads on top of them and training the resulting models for 22 segmentation tasks. We show that vox2vec outperforms existing medical imaging SSL techniques in three evaluation setups: linear and non-linear probing and end-to-end fine-tuning. Moreover, a non-linear head trained on top of the frozen vox2vec representations achieves competitive performance with the FPN trained from scratch while having 50 times fewer trainable parameters. The code is available at https://github.com/mishgon/vox2vec ."
                    }
                ]
            },
            "cf6dd1e7-23d8-4ecd-9115-824550e52f12": {
                "pk": "cf6dd1e7-23d8-4ecd-9115-824550e52f12",
                "name": "Aibek Alanov",
                "collaborators": [
                    "Dmitry Vetrov",
                    "Vadim Titov",
                    "Pavel Andreev",
                    "Max Kochurov",
                    "Daniil Yashkov",
                    "Oleg Ivanov",
                    "Denis Bobkov",
                    "Ivan Shchekotov",
                    "Maksim Nakhodnov",
                    "Aleksandr Beznosikov"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Domain Adaptation",
                    "Machine Learning",
                    "Image Processing"
                ],
                "publications": [
                    {
                        "title": "Pairwise Augmented GANs with Adversarial Reconstruction Loss",
                        "abstract": "We propose a novel autoencoding model called Pairwise Augmented GANs. We train a generator and an encoder jointly and in an adversarial manner. The generator network learns to sample realistic objects. In turn, the encoder network at the same time is trained to map the true data distribution to the prior in latent space. To ensure good reconstructions, we introduce an augmented adversarial reconstruction loss. Here we train a discriminator to distinguish two types of pairs: an object with its augmentation and the one with its reconstruction. We show that such adversarial loss compares objects based on the content rather than on the exact match. We experimentally demonstrate that our model generates samples and reconstructions of quality competitive with state-of-the-art on datasets MNIST, CIFAR10, CelebA and achieves good quantitative results on CIFAR10."
                    },
                    {
                        "title": "HiFi++: a Unified Framework for Bandwidth Extension and Speech Enhancement",
                        "abstract": "Generative adversarial networks have recently demonstrated outstanding performance in neural vocoding outperforming best autoregressive and flow-based models. In this paper, we show that this success can be extended to other tasks of conditional audio generation. In particular, building upon HiFi vocoders, we propose a novel HiFi++ general framework for bandwidth extension and speech enhancement. We show that with the improved generator architecture, HiFi++ performs better or comparably with the state-of-the-art in these tasks while spending significantly less computational resources. The effectiveness of our approach is validated through a series of extensive experiments."
                    },
                    {
                        "title": "HyperDomainNet: Universal Domain Adaptation for Generative Adversarial Networks",
                        "abstract": "Domain adaptation framework of GANs has achieved great progress in recent years as a main successful approach of training contemporary GANs in the case of very limited training data. In this work, we significantly improve this framework by proposing an extremely compact parameter space for fine-tuning the generator. We introduce a novel domain-modulation technique that allows to optimize only 6 thousand-dimensional vector instead of 30 million weights of StyleGAN2 to adapt to a target domain. We apply this parameterization to the state-of-art domain adaptation methods and show that it has almost the same expressiveness as the full parameter space. Additionally, we propose a new regularization loss that considerably enhances the diversity of the fine-tuned generator. Inspired by the reduction in the size of the optimizing parameter space we consider the problem of multi-domain adaptation of GANs, i.e. setting when the same model can adapt to several domains depending on the input query. We propose the HyperDomainNet that is a hypernetwork that predicts our parameterization given the target domain. We empirically confirm that it can successfully learn a number of domains at once and may even generalize to unseen domains. Source code can be found at https://github.com/MACderRu/HyperDomainNet"
                    },
                    {
                        "title": "StyleDomain: Efficient and Lightweight Parameterizations of StyleGAN for One-shot and Few-shot Domain Adaptation",
                        "abstract": "Domain adaptation of GANs is a problem of fine-tuning GAN models pretrained on a large dataset (e.g. StyleGAN) to a specific domain with few samples (e.g. painting faces, sketches, etc.). While there are many methods that tackle this problem in different ways, there are still many important questions that remain unanswered. In this paper, we provide a systematic and in-depth analysis of the domain adaptation problem of GANs, focusing on the StyleGAN model. We perform a detailed exploration of the most important parts of StyleGAN that are responsible for adapting the generator to a new domain depending on the similarity between the source and target domains. As a result of this study, we propose new efficient and lightweight parameterizations of StyleGAN for domain adaptation. Particularly, we show that there exist directions in StyleSpace (StyleDomain directions) that are sufficient for adapting to similar domains. For dissimilar domains, we propose Affine+ and AffineLight+ parameterizations that allows us to outperform existing baselines in few-shot adaptation while having significantly less training parameters. Finally, we examine StyleDomain directions and discover their many surprising properties that we apply for domain mixing and cross-domain image morphing. Source code can be found at https://github.com/AIRI-Institute/StyleDomain."
                    },
                    {
                        "title": "The Devil is in the Details: StyleFeatureEditor for Detail-Rich StyleGAN Inversion and High Quality Image Editing",
                        "abstract": "The task of manipulating real image attributes through StyleGAN inversion has been extensively researched. This process involves searching latent variables from a well-trained StyleGAN generator that can synthesize a real image, modifying these latent variables, and then synthesizing an image with the desired edits. A balance must be struck between the quality of the reconstruction and the ability to edit. Earlier studies utilized the low-dimensional W-space for latent search, which facilitated effective editing but struggled with reconstructing intricate details. More recent research has turned to the high-dimensional feature space F, which successfully inverses the input image but loses much of the detail during editing. In this paper, we introduce StyleFeatureEditor -- a novel method that enables editing in both w-latents and F-latents. This technique not only allows for the reconstruction of finer image details but also ensures their preservation during editing. We also present a new training pipeline specifically designed to train our model to accurately edit F-latents. Our method is compared with state-of-the-art encoding approaches, demonstrating that our model excels in terms of reconstruction quality and is capable of editing even challenging out-of-domain examples. Code is available at https://github.com/AIRI-Institute/StyleFeatureEditor."
                    },
                    {
                        "title": "FFC-SE: Fast Fourier Convolution for Speech Enhancement",
                        "abstract": "Fast Fourier convolution (FFC) is the recently proposed neural operator showing promising performance in several computer vision problems. The FFC operator allows employing large receptive field operations within early layers of the neural network. It was shown to be especially helpful for inpainting of periodic structures which are common in audio processing. In this work, we design neural network architectures which adapt FFC for speech enhancement. We hypothesize that a large receptive field allows these networks to produce more coherent phases than vanilla convolutional models, and validate this hypothesis experimentally. We found that neural networks based on Fast Fourier convolution outperform analogous convolutional models and show better or comparable results with other speech enhancement baselines."
                    },
                    {
                        "title": "On Scaled Methods for Saddle Point Problems",
                        "abstract": "Methods with adaptive scaling of different features play a key role in solving saddle point problems, primarily due to Adam's popularity for solving adversarial machine learning problems, including GANS training. This paper carries out a theoretical analysis of the following scaling techniques for solving SPPs: the well-known Adam and RmsProp scaling and the newer AdaHessian and OASIS based on Hutchison approximation. We use the Extra Gradient and its improved version with negative momentum as the basic method. Experimental studies on GANs show good applicability not only for Adam, but also for other less popular methods."
                    },
                    {
                        "title": "Iterative autoregression: a novel trick to improve your low-latency speech enhancement model",
                        "abstract": "Streaming models are an essential component of real-time speech enhancement tools. The streaming regime constrains speech enhancement models to use only a tiny context of future information. As a result, the low-latency streaming setup is generally considered a challenging task and has a significant negative impact on the model's quality. However, the sequential nature of streaming generation offers a natural possibility for autoregression, that is, utilizing previous predictions while making current ones. The conventional method for training autoregressive models is teacher forcing, but its primary drawback lies in the training-inference mismatch that can lead to a substantial degradation in quality. In this study, we propose a straightforward yet effective alternative technique for training autoregressive low-latency speech enhancement models. We demonstrate that the proposed approach leads to stable improvement across diverse architectures and training scenarios."
                    },
                    {
                        "title": "User-Controllable Multi-Texture Synthesis with Generative Adversarial Networks",
                        "abstract": "We propose a novel multi-texture synthesis model based on generative adversarial networks (GANs) with a user-controllable mechanism. The user control ability allows to explicitly specify the texture which should be generated by the model. This property follows from using an encoder part which learns a latent representation for each texture from the dataset. To ensure a dataset coverage, we use an adversarial loss function that penalizes for incorrect reproductions of a given texture. In experiments, we show that our model can learn descriptive texture manifolds for large datasets and from raw data such as a collection of high-resolution photos. Moreover, we apply our method to produce 3D textures and show that it outperforms existing baselines."
                    },
                    {
                        "title": "HairFastGAN: Realistic and Robust Hair Transfer with a Fast Encoder-Based Approach",
                        "abstract": "Our paper addresses the complex task of transferring a hairstyle from a reference image to an input photo for virtual hair try-on. This task is challenging due to the need to adapt to various photo poses, the sensitivity of hairstyles, and the lack of objective metrics. The current state of the art hairstyle transfer methods use an optimization process for different parts of the approach, making them inexcusably slow. At the same time, faster encoder-based models are of very low quality because they either operate in StyleGAN's W+ space or use other low-dimensional image generators. Additionally, both approaches have a problem with hairstyle transfer when the source pose is very different from the target pose, because they either don't consider the pose at all or deal with it inefficiently. In our paper, we present the HairFast model, which uniquely solves these problems and achieves high resolution, near real-time performance, and superior reconstruction compared to optimization problem-based methods. Our solution includes a new architecture operating in the FS latent space of StyleGAN, an enhanced inpainting approach, and improved encoders for better alignment, color transfer, and a new encoder for post-processing. The effectiveness of our approach is demonstrated on realism metrics after random hairstyle transfer and reconstruction when the original hairstyle is transferred. In the most difficult scenario of transferring both shape and color of a hairstyle from different images, our method performs in less than a second on the Nvidia V100. Our code is available at https://github.com/AIRI-Institute/HairFastGAN."
                    },
                    {
                        "title": "Star-Shaped Denoising Diffusion Probabilistic Models",
                        "abstract": "Denoising Diffusion Probabilistic Models (DDPMs) provide the foundation for the recent breakthroughs in generative modeling. Their Markovian structure makes it difficult to define DDPMs with distributions other than Gaussian or discrete. In this paper, we introduce Star-Shaped DDPM (SS-DDPM). Its star-shaped diffusion process allows us to bypass the need to define the transition probabilities or compute posteriors. We establish duality between star-shaped and specific Markovian diffusions for the exponential family of distributions and derive efficient algorithms for training and sampling from SS-DDPMs. In the case of Gaussian distributions, SS-DDPM is equivalent to DDPM. However, SS-DDPMs provide a simple recipe for designing diffusion models with distributions such as Beta, von Mises$\\unicode{x2013}$Fisher, Dirichlet, Wishart and others, which can be especially useful when data lies on a constrained manifold. We evaluate the model in different settings and find it competitive even on image data, where Beta SS-DDPM achieves results comparable to a Gaussian DDPM. Our implementation is available at https://github.com/andrey-okhotin/star-shaped ."
                    },
                    {
                        "title": "Guide-and-Rescale: Self-Guidance Mechanism for Effective Tuning-Free Real Image Editing",
                        "abstract": "Despite recent advances in large-scale text-to-image generative models, manipulating real images with these models remains a challenging problem. The main limitations of existing editing methods are that they either fail to perform with consistent quality on a wide range of image edits or require time-consuming hyperparameter tuning or fine-tuning of the diffusion model to preserve the image-specific appearance of the input image. We propose a novel approach that is built upon a modified diffusion sampling process via the guidance mechanism. In this work, we explore the self-guidance technique to preserve the overall structure of the input image and its local regions appearance that should not be edited. In particular, we explicitly introduce layout-preserving energy functions that are aimed to save local and global structures of the source image. Additionally, we propose a noise rescaling mechanism that allows to preserve noise distribution by balancing the norms of classifier-free guidance and our proposed guiders during generation. Such a guiding approach does not require fine-tuning the diffusion model and exact inversion process. As a result, the proposed method provides a fast and high-quality editing mechanism. In our experiments, we show through human evaluation and quantitative analysis that the proposed method allows to produce desired editing which is more preferable by humans and also achieves a better trade-off between editing quality and preservation of the original image. Our code is available at https://github.com/MACderRu/Guide-and-Rescale."
                    },
                    {
                        "title": "CLEAR: Character Unlearning in Textual and Visual Modalities",
                        "abstract": "Machine Unlearning (MU) is critical for enhancing privacy and security in deep learning models, particularly in large multimodal language models (MLLMs), by removing specific private or hazardous information. While MU has made significant progress in textual and visual modalities, multimodal unlearning (MMU) remains significantly underexplored, partially due to the absence of a suitable open-source benchmark. To address this, we introduce CLEAR, a new benchmark designed to evaluate MMU methods. CLEAR contains 200 fictitious individuals and 3,700 images linked with corresponding question-answer pairs, enabling a thorough evaluation across modalities. We assess 10 MU methods, adapting them for MMU, and highlight new challenges specific to multimodal forgetting. We also demonstrate that simple $\\ell_1$ regularization on LoRA weights significantly mitigates catastrophic forgetting, preserving model performance on retained data. The dataset is available at https://huggingface.co/datasets/therem/CLEAR"
                    }
                ]
            },
            "5fae6399-6062-4738-902e-db90a8e99f3e": {
                "pk": "5fae6399-6062-4738-902e-db90a8e99f3e",
                "name": "Alexey Naumov",
                "collaborators": [
                    "Friedrich G\u00f6tze",
                    "Alexander Tikhomirov",
                    "Sergey Samsonov",
                    "Vladimir Ulyanov",
                    "Eric Moulines",
                    "Denis Belomestny",
                    "Daniil Tiapkin",
                    "Vladimir Spokoiny",
                    "Alain Durmus",
                    "Nikita Morozov"
                ],
                "domain": [
                    "Random Matrix Theory",
                    "Probability Theory",
                    "Reinforcement Learning",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Elliptic law for real random matrices",
                        "abstract": "In this paper we consider ensemble of random matrices $\\X_n$ with independent identically distributed vectors $(X_{ij}, X_{ji})_{i \\neq j}$ of entries. Under assumption of finite fourth moment of matrix entries it is proved that empirical spectral distribution of eigenvalues converges in probability to a uniform distribution on the ellipse. The axis of the ellipse are determined by correlation between $X_{12}$ and $X_{21}$. This result is called Elliptic Law. Limit distribution doesn't depend on distribution of matrix elements and the result in this sence is universal."
                    },
                    {
                        "title": "Asymptotic analysis of symmetric functions",
                        "abstract": "In this paper we consider asymptotic expansions for a class of sequences of symmetric functions of many variables. Applications to classical and free probability theory are discussed."
                    },
                    {
                        "title": "On one generalization of the elliptic law for random matrices",
                        "abstract": "We consider the products of $m\\ge 2$ independent large real random matrices with independent vectors $(X_{jk}^{(q)},X_{kj}^{(q)})$ of entries. The entries $X_{jk}^{(q)},X_{kj}^{(q)}$ are correlated with $\\rho=\\mathbb E X_{jk}^{(q)}X_{kj}^{(q)}$. The limit distribution of the empirical spectral distribution of the eigenvalues of such products doesn't depend on $\\rho$ and equals to the distribution of $m$th power of the random variable uniformly distributed on the unit disc."
                    },
                    {
                        "title": "Distribution of Linear Statistics of Singular Values of the Product of Random Matrices",
                        "abstract": "In this paper we consider the product of two independent random matrices $\\mathbb X^{(1)}$ and $\\mathbb X^{(2)}$. Assume that $X_{jk}^{(q)}, 1 \\le j,k \\le n, q = 1, 2,$ are i.i.d. random variables with $\\mathbb E X_{jk}^{(q)} = 0, \\mathbb E (X_{jk}^{(q)})^2 = 1$. Denote by $s_1, ..., s_n$ the singular values of $\\mathbb W: = \\frac{1}{n} \\mathbb X^{(1)} \\mathbb X^{(2)}$. We prove the central limit theorem for linear statistics of the squared singular values $s_1^2, ..., s_n^2$ showing that the limiting variance depends on $\\kappa_4: = \\mathbb E (X_{11}^{1})^4 - 3$."
                    },
                    {
                        "title": "Local semicircle law under fourth moment condition",
                        "abstract": "We consider a random symmetric matrix ${\\bf X} = [X_{jk}]_{j,k=1}^n$ with upper triangular entries being independent random variables with mean zero and unit variance. Assuming that $\\max_{jk} {\\mathbb E} |X_{jk}|^{4+\\delta} < \\infty, \\delta > 0$, it was proved in [G\\\"otze, Naumov and Tikhomirov, Bernoulli, 2018] that with high probability the typical distance between the Stieltjes transforms $m_n(z), z = u + i v$, of the empirical spectral distribution (ESD) and the Stieltjes transforms $m_{sc}(z)$ of the semicircle law is of order $(nv)^{-1} \\log n$. The aim of this paper is to remove $\\delta>0$ and show that this result still holds if we assume that $\\max_{jk} {\\mathbb E} |X_{jk}|^{4} < \\infty$. We also discuss applications to the rate of convergence of the ESD to the semicircle law in the Kolmogorov distance, rates of localization of the eigenvalues around the classical positions and rates of delocalization of eigenvectors."
                    },
                    {
                        "title": "On the Local Semicircular Law for Wigner Ensembles",
                        "abstract": "We consider a random symmetric matrix ${\\bf X} = [X_{jk}]_{j,k=1}^n$ with upper triangular entries being i.i.d. random variables with mean zero and unit variance. We additionally suppose that $\\mathbb E |X_{11}|^{4 + \\delta} =: \\mu_{4+\\delta} < \\infty$ for some $\\delta > 0$. The aim of this paper is to significantly extend recent result of the authors [18] and show that with high probability the typical distance between the Stieltjes transform of the empirical spectral distribution (ESD) of the matrix $n^{-\\frac{1}{2}} {\\bf X}$ and Wigner's semicircle law is of order $(nv)^{-1} \\log n$, where $v$ denotes the distance to the real line in the complex plane. We apply this result to the rate of convergence of the ESD to the distribution function of the semicircle law as well as to rigidity of eigenvalues and eigenvector delocalization significantly extending a recent result by G\\\"otze, Naumov and Tikhomirov [19]. The result on delocalization is optimal by comparison with GOE ensembles. Furthermore the techniques of this paper provide a new shorter proof for the optimal $O(n^{-1})$ rate of convergence of the expected ESD to the semicircle law."
                    },
                    {
                        "title": "On minimal singular values of random matrices with correlated entries",
                        "abstract": "Let $\\mathbf X$ be a random matrix whose pairs of entries $X_{jk}$ and $X_{kj}$ are correlated and vectors $ (X_{jk},X_{kj})$, for $1\\le j<k\\le n$, are mutually independent. Assume that the diagonal entries are independent from off-diagonal entries as well. We assume that $\\mathbb{E} X_{jk}=0$, $\\mathbb{E} X_{jk}^2=1$, for any $j,k=1,\\ldots,n$ and $\\mathbb{E} X_{jk}X_{kj}=\\rho$ for $1\\le j<k\\le n$. Let $\\mathbf M_n$ be a non-random $n\\times n$ matrix with $\\|\\mathbf M_n\\|\\le Kn^Q$, for some positive constants $K>0$ and $Q\\ge 0$. Let $s_n(\\mathbf X+\\mathbf M_n)$ denote the least singular value of the matrix $\\mathbf X+\\mathbf M_n$. It is shown that there exist positive constants $A$ and $B$ depending on $K,Q,\\rho$ only such that $$ \\mathbb{P}(s_n(\\mathbf X+\\mathbf M_n)\\le n^{-A})\\le n^{-B}. $$ As an application of this result we prove the elliptic law for this class of matrices with non identically distributed correlated entries."
                    },
                    {
                        "title": "Local semicircle law under moment conditions. Part II: Localization and delocalization",
                        "abstract": "We consider a random symmetric matrix ${\\bf X} = [X_{jk}]_{j,k=1}^n$ with upper triangular entries being independent identically distributed random variables with mean zero and unit variance. We additionally suppose that $\\mathbb E |X_{11}|^{4 + \\delta} =: \\mu_{4+\\delta} < C$ for some $\\delta > 0$ and some absolute constant $C$. Under these conditions we show that the typical Kolmogorov distance between the empirical spectral distribution function of eigenvalues of $n^{-1/2} {\\bf X}$ and Wigner's semicircle law is of order $1/n$ up to some logarithmic correction factor. As a direct consequence of this result we establish that the semicircle law holds on a short scale. Furthermore, we show for this finite moment ensemble rigidity of eigenvalues and delocalization properties of the eigenvectors. Some numerical experiments are included illustrating the influence of the tail behavior of the matrix entries when only a small number of moments exist."
                    },
                    {
                        "title": "Local semicircle law under moment conditions. Part I: The Stieltjes transform",
                        "abstract": "We consider a random symmetric matrix ${\\bf X} = [X_{jk}]_{j,k=1}^n$ in which the upper triangular entries are independent identically distributed random variables with mean zero and unit variance. We additionally suppose that $\\mathbb E |X_{11}|^{4 + \\delta} =: \\mu_4 < \\infty$ for some $\\delta > 0$. Under these conditions we show that the typical distance between the Stieltjes transform of the empirical spectral distribution (ESD) of the matrix $n^{-\\frac{1}{2}} {\\bf X}$ and Wigner's semicircle law is of order $(nv)^{-1}$, where $v$ is the distance in the complex plane to the real line. Furthermore we outline applications which are deferred to a subsequent paper, such as the rate of convergence in probability of the ESD to the distribution function of the semicircle law, rigidity of the eigenvalues and eigenvector delocalization."
                    },
                    {
                        "title": "Bootstrap confidence sets for spectral projectors of sample covariance",
                        "abstract": "Let $X_{1},\\ldots,X_{n}$ be i.i.d. sample in $\\mathbb{R}^{p}$ with zero mean and the covariance matrix $\\mathbf{\\Sigma}$. The problem of recovering the projector onto an eigenspace of $\\mathbf{\\Sigma}$ from these observations naturally arises in many applications. Recent technique from [Koltchinskii, Lounici, 2015] helps to study the asymptotic distribution of the distance in the Frobenius norm $\\| \\mathbf{P}_r - \\widehat{\\mathbf{P}}_r \\|_{2}$ between the true projector $\\mathbf{P}_r$ on the subspace of the $r$-th eigenvalue and its empirical counterpart $\\widehat{\\mathbf{P}}_r$ in terms of the effective rank of $\\mathbf{\\Sigma}$. This paper offers a bootstrap procedure for building sharp confidence sets for the true projector $\\mathbf{P}_r$ from the given data. This procedure does not rely on the asymptotic distribution of $\\| \\mathbf{P}_r - \\widehat{\\mathbf{P}}_r \\|_{2}$ and its moments. It could be applied for small or moderate sample size $n$ and large dimension $p$. The main result states the validity of the proposed procedure for finite samples with an explicit error bound for the error of bootstrap approximation. This bound involves some new sharp results on Gaussian comparison and Gaussian anti-concentration in high-dimensional spaces. Numeric results confirm a good performance of the method in realistic examples."
                    },
                    {
                        "title": "On Local laws for non-Hermitian random matrices and their products",
                        "abstract": "The aim of this paper is to prove a local version of the circular law for non-Hermitian random matrices and its generalization to the product of non-Hermitian random matrices under weak moment conditions. More precisely we assume that the entries $X_{jk}^{(q)}$ of non-Hermitian random matrices ${\\bf X}^{(q)}, 1 \\le j,k \\le n, q = 1, \\ldots, m, m \\geq 1$ are i.i.d. r.v. with $\\mathbb E X_{jk} =0, \\mathbb E X_{jk}^2 = 1$ and $\\mathbb E |X_{jk}|^{4+\\delta} < \\infty$ for some $\\delta > 0$. It is shown that the local law holds on the optimal scale $n^{-1+2a}, a > 0$, up to some logarithmic factor. We further develop a Stein type method to estimate the perturbation of the equations for the Stieltjes transform of the limiting distribution. We also generalize the recent results [Bourgade--Yau-Yin, 2014], [Tao--Vu, 2015] and [Nemish, 2017]. An extension to the case of non-i.i.d. entries is discussed."
                    },
                    {
                        "title": "Large ball probability, Gaussian comparison and anti-concentration",
                        "abstract": "We derive tight non-asymptotic bounds for the Kolmogorov distance between the probabilities of two Gaussian elements to hit a ball in a Hilbert space. The key property of these bounds is that they are dimension-free and depend on the nuclear (Schatten-one) norm of the difference between the covariance operators of the elements and on the norm of the mean shift. The obtained bounds significantly improve the bound based on Pinsker's inequality via the Kullback-Leibler divergence. We also establish an anti-concentration bound for a squared norm of a non-centered Gaussian element in Hilbert space. The paper presents a number of examples motivating our results and applications of the obtained bounds to statistical inference and to high-dimensional CLT."
                    },
                    {
                        "title": "Probability and moment inequalities for additive functionals of geometrically ergodic Markov chains",
                        "abstract": "In this paper, we establish moment and Bernstein-type inequalities for additive functionals of geometrically ergodic Markov chains. These inequalities extend the corresponding inequalities for independent random variables. Our conditions cover Markov chains converging geometrically to the stationary distribution either in $V$-norms or in weighted Wasserstein distances. Our inequalities apply to unbounded functions and depend explicitly on constants appearing in the conditions that we consider."
                    },
                    {
                        "title": "Simultaneous approximation of a smooth function and its derivatives by deep neural networks with piecewise-polynomial activations",
                        "abstract": "This paper investigates the approximation properties of deep neural networks with piecewise-polynomial activation functions. We derive the required depth, width, and sparsity of a deep neural network to approximate any H\\\"{o}lder smooth function up to a given approximation error in H\\\"{o}lder norms in such a way that all weights of this neural network are bounded by $1$. The latter feature is essential to control generalization errors in many statistical and machine learning applications."
                    },
                    {
                        "title": "Theoretical guarantees for neural control variates in MCMC",
                        "abstract": "In this paper, we propose a variance reduction approach for Markov chains based on additive control variates and the minimization of an appropriate estimate for the asymptotic variance. We focus on the particular case when control variates are represented as deep neural networks. We derive the optimal convergence rate of the asymptotic variance under various ergodicity assumptions on the underlying Markov chain. The proposed approach relies upon recent results on the stochastic errors of variance reduction algorithms and function approximation theory."
                    },
                    {
                        "title": "UVIP: Model-Free Approach to Evaluate Reinforcement Learning Algorithms",
                        "abstract": "Policy evaluation is an important instrument for the comparison of different algorithms in Reinforcement Learning (RL). Yet even a precise knowledge of the value function $V^{\\pi}$ corresponding to a policy $\\pi$ does not provide reliable information on how far is the policy $\\pi$ from the optimal one. We present a novel model-free upper value iteration procedure $({\\sf UVIP})$ that allows us to estimate the suboptimality gap $V^{\\star}(x) - V^{\\pi}(x)$ from above and to construct confidence intervals for $V^\\star$. Our approach relies on upper bounds to the solution of the Bellman optimality equation via martingale approach. We provide theoretical guarantees for ${\\sf UVIP}$ under general assumptions and illustrate its performance on a number of benchmark RL problems."
                    },
                    {
                        "title": "Generative Flow Networks as Entropy-Regularized RL",
                        "abstract": "The recently proposed generative flow networks (GFlowNets) are a method of training a policy to sample compositional discrete objects with probabilities proportional to a given reward via a sequence of actions. GFlowNets exploit the sequential nature of the problem, drawing parallels with reinforcement learning (RL). Our work extends the connection between RL and GFlowNets to a general case. We demonstrate how the task of learning a generative flow network can be efficiently redefined as an entropy-regularized RL problem with a specific reward and regularizer structure. Furthermore, we illustrate the practical efficiency of this reformulation by applying standard soft RL algorithms to GFlowNet training across several probabilistic modeling tasks. Contrary to previously reported results, we show that entropic RL approaches can be competitive against established GFlowNet training methods. This perspective opens a direct path for integrating RL principles into the realm of generative flow networks."
                    },
                    {
                        "title": "Improved High-Probability Bounds for the Temporal Difference Learning Algorithm via Exponential Stability",
                        "abstract": "In this paper we consider the problem of obtaining sharp bounds for the performance of temporal difference (TD) methods with linear function approximation for policy evaluation in discounted Markov decision processes. We show that a simple algorithm with a universal and instance-independent step size together with Polyak-Ruppert tail averaging is sufficient to obtain near-optimal variance and bias terms. We also provide the respective sample complexity bounds. Our proof technique is based on refined error bounds for linear stochastic approximation together with the novel stability result for the product of random matrices that arise from the TD-type recurrence."
                    },
                    {
                        "title": "Improving GFlowNets with Monte Carlo Tree Search",
                        "abstract": "Generative Flow Networks (GFlowNets) treat sampling from distributions over compositional discrete spaces as a sequential decision-making problem, training a stochastic policy to construct objects step by step. Recent studies have revealed strong connections between GFlowNets and entropy-regularized reinforcement learning. Building on these insights, we propose to enhance planning capabilities of GFlowNets by applying Monte Carlo Tree Search (MCTS). Specifically, we show how the MENTS algorithm (Xiao et al., 2019) can be adapted for GFlowNets and used during both training and inference. Our experiments demonstrate that this approach improves the sample efficiency of GFlowNet training and the generation fidelity of pre-trained GFlowNet models."
                    },
                    {
                        "title": "Finite-time High-probability Bounds for Polyak-Ruppert Averaged Iterates of Linear Stochastic Approximation",
                        "abstract": "This paper provides a finite-time analysis of linear stochastic approximation (LSA) algorithms with fixed step size, a core method in statistics and machine learning. LSA is used to compute approximate solutions of a $d$-dimensional linear system $\\bar{\\mathbf{A}} \\theta = \\bar{\\mathbf{b}}$ for which $(\\bar{\\mathbf{A}}, \\bar{\\mathbf{b}})$ can only be estimated by (asymptotically) unbiased observations $\\{(\\mathbf{A}(Z_n),\\mathbf{b}(Z_n))\\}_{n \\in \\mathbb{N}}$. We consider here the case where $\\{Z_n\\}_{n \\in \\mathbb{N}}$ is an i.i.d. sequence or a uniformly geometrically ergodic Markov chain. We derive $p$-th moment and high-probability deviation bounds for the iterates defined by LSA and its Polyak-Ruppert-averaged version. Our finite-time instance-dependent bounds for the averaged LSA iterates are sharp in the sense that the leading term we obtain coincides with the local asymptotic minimax limit. Moreover, the remainder terms of our bounds admit a tight dependence on the mixing time $t_{\\operatorname{mix}}$ of the underlying chain and the norm of the noise variables. We emphasize that our result requires the SA step size to scale only with logarithm of the problem dimension $d$."
                    }
                ]
            },
            "773d7b44-0643-4295-b0b1-f6c744c46ac4": {
                "pk": "773d7b44-0643-4295-b0b1-f6c744c46ac4",
                "name": "Maxim Rakhuba",
                "collaborators": [
                    "Ivan Oseledets",
                    "Anton Obukhov",
                    "Alexander Novikov",
                    "Carlo Marcati",
                    "Stamatios Georgoulis",
                    "Dengxin Dai",
                    "Luc Van Gool",
                    "Lev Vysotsky",
                    "Nikita Puchkin",
                    "Andr\u00e9 Uschmajew"
                ],
                "domain": [
                    "Tensor Decomposition",
                    "Low-Rank Approximation",
                    "Scientific Computing",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Jacobi-Davidson method on low-rank matrix manifolds",
                        "abstract": "In this work we generalize the Jacobi-Davidson method to the case when eigenvector can be reshaped into a low-rank matrix. In this setting the proposed method inherits advantages of the original Jacobi-Davidson method, has lower complexity and requires less storage. We also introduce low-rank version of the Rayleigh quotient iteration which naturally arises in the Jacobi-Davidson method."
                    },
                    {
                        "title": "Calculating vibrational spectra of molecules using tensor train decomposition",
                        "abstract": "We propose a new algorithm for calculation of vibrational spectra of molecules using tensor train decomposition. Under the assumption that eigenfunctions lie on a low-parametric manifold of low-rank tensors we suggest using well-known iterative methods that utilize matrix inversion (LOBPCG, inverse iteration) and solve corresponding linear systems inexactly along this manifold. As an application, we accurately compute vibrational spectra (84 states) of acetonitrile molecule CH$_3$CN on a laptop in one hour using only $100$ MB of memory to represent all computed eigenfunctions."
                    },
                    {
                        "title": "Grid-based electronic structure calculations: the tensor decomposition approach",
                        "abstract": "We present a fully grid-based approach for solving Hartree-Fock and all-electron Kohn-Sham equations based on low-rank approximation of three-dimensional electron orbitals. Due to the low-rank structure the total complexity of the algorithm depends linearly with respect to the one-dimensional grid size. Linear complexity allows for the usage of fine grids, e.g. $8192^3$ and, thus, cheap extrapolation procedure.   We test the proposed approach on closed-shell atoms up to the argon, several molecules and clusters of hydrogen atoms. All tests show systematical convergence with the required accuracy."
                    },
                    {
                        "title": "Tensor rank bounds and explicit QTT representations for the inverses of circulant matrices",
                        "abstract": "In this paper, we are concerned with the inversion of circulant matrices and their quantized tensor-train (QTT) structure. In particular, we show that the inverse of a complex circulant matrix $A$, generated by the first column of the form $(a_0,\\dots,a_{m-1},0,\\dots,0,a_{-n},\\dots, a_{-1})^\\top$ admits a QTT representation with the QTT ranks bounded by $(m+n)$. Under certain assumptions on the entries of $A$, we also derive an explicit QTT representation of $A^{-1}$. The latter can be used, for instance, to overcome stability issues arising when numerically solving differential equations with periodic boundary conditions in the QTT format."
                    },
                    {
                        "title": "Dimension-free Structured Covariance Estimation",
                        "abstract": "Given a sample of i.i.d. high-dimensional centered random vectors, we consider a problem of estimation of their covariance matrix $\\Sigma$ with an additional assumption that $\\Sigma$ can be represented as a sum of a few Kronecker products of smaller matrices. Under mild conditions, we derive the first non-asymptotic dimension-free high-probability bound on the Frobenius distance between $\\Sigma$ and a widely used penalized permuted least squares estimate. Because of the hidden structure, the established rate of convergence is faster than in the standard covariance estimation problem."
                    },
                    {
                        "title": "Low-rank Riemannian eigensolver for high-dimensional Hamiltonians",
                        "abstract": "Such problems as computation of spectra of spin chains and vibrational spectra of molecules can be written as high-dimensional eigenvalue problems, i.e., when the eigenvector can be naturally represented as a multidimensional tensor. Tensor methods have proven to be an efficient tool for the approximation of solutions of high-dimensional eigenvalue problems, however, their performance deteriorates quickly when the number of eigenstates to be computed increases. We address this issue by designing a new algorithm motivated by the ideas of Riemannian optimization (optimization on smooth manifolds) for the approximation of multiple eigenstates in the tensor-train format, which is also known as matrix product state representation. The proposed algorithm is implemented in TensorFlow, which allows for both CPU and GPU parallelization."
                    },
                    {
                        "title": "Alternating least squares as moving subspace correction",
                        "abstract": "In this note we take a new look at the local convergence of alternating optimization methods for low-rank matrices and tensors. Our abstract interpretation as sequential optimization on moving subspaces yields insightful reformulations of some known convergence conditions that focus on the interplay between the contractivity of classical multiplicative Schwarz methods with overlapping subspaces and the curvature of low-rank matrix and tensor manifolds. While the verification of the abstract conditions in concrete scenarios remains open in most cases, we are able to provide an alternative and conceptually simple derivation of the asymptotic convergence rate of the two-sided block power method of numerical algebra for computing the dominant singular subspaces of a rectangular matrix. This method is equivalent to an alternating least squares method applied to a distance function. The theoretical results are illustrated and validated by numerical experiments."
                    },
                    {
                        "title": "Vico-Greengard-Ferrando quadratures in the tensor solver for integral equations",
                        "abstract": "Convolution with Green's function of a differential operator appears in a lot of applications e.g. Lippmann-Schwinger integral equation. Algorithms for computing such are usually non-trivial and require non-uniform mesh. However, recently Vico, Greengard and Ferrando developed method for computing convolution with smooth functions with compact support with spectral accuracy, requiring nothing more than Fast Fourier Transform (FFT). Their approach is very suitable for the low-rank tensor implementation which we develop using Quantized Tensor Train (QTT) decomposition."
                    },
                    {
                        "title": "Tensor Rank bounds for Point Singularities in $\\mathbb{R}^3$",
                        "abstract": "We analyze rates of approximation by quantized, tensor-structured representations of functions with isolated point singularities in ${\\mathbb R}^3$. We consider functions in countably normed Sobolev spaces with radial weights and analytic- or Gevrey-type control of weighted semi-norms. Several classes of boundary value and eigenvalue problems from science and engineering are discussed whose solutions belong to the countably normed spaces.   It is shown that quantized, tensor-structured approximations of functions in these classes exhibit tensor ranks bounded polylogarithmically with respect to the accuracy $\\epsilon\\in(0,1)$ in the Sobolev space $H^1$. We prove exponential convergence rates of three specific types of quantized tensor decompositions: quantized tensor train (QTT), transposed QTT and Tucker-QTT. In addition, the bounds for the patchwise decompositions are uniform with respect to the position of the point singularity. An auxiliary result of independent interest is the proof of exponential convergence of $hp$-finite element approximations for Gevrey-regular functions with point singularities in the unit cube $Q=(0,1)^3$. Numerical examples of function approximations and of Schr\\\"odinger-type eigenvalue problems illustrate the theoretical results."
                    },
                    {
                        "title": "Automatic differentiation for Riemannian optimization on low-rank matrix and tensor-train manifolds",
                        "abstract": "In scientific computing and machine learning applications, matrices and more general multidimensional arrays (tensors) can often be approximated with the help of low-rank decompositions. Since matrices and tensors of fixed rank form smooth Riemannian manifolds, one of the popular tools for finding low-rank approximations is to use Riemannian optimization. Nevertheless, efficient implementation of Riemannian gradients and Hessians, required in Riemannian optimization algorithms, can be a nontrivial task in practice. Moreover, in some cases, analytic formulas are not even available. In this paper, we build upon automatic differentiation and propose a method that, given an implementation of the function to be minimized, efficiently computes Riemannian gradients and matrix-by-vector products between an approximate Riemannian Hessian and a given vector."
                    },
                    {
                        "title": "Towards Practical Control of Singular Values of Convolutional Layers",
                        "abstract": "In general, convolutional neural networks (CNNs) are easy to train, but their essential properties, such as generalization error and adversarial robustness, are hard to control. Recent research demonstrated that singular values of convolutional layers significantly affect such elusive properties and offered several methods for controlling them. Nevertheless, these methods present an intractable computational challenge or resort to coarse approximations. In this paper, we offer a principled approach to alleviating constraints of the prior art at the expense of an insignificant reduction in layer expressivity. Our method is based on the tensor-train decomposition; it retains control over the actual singular values of convolutional mappings while providing structurally sparse and hardware-friendly representation. We demonstrate the improved properties of modern CNNs with our method and analyze its impact on the model performance, calibration, and adversarial robustness. The source code is available at: https://github.com/WhiteTeaDragon/practical_svd_conv"
                    },
                    {
                        "title": "Tight and Efficient Upper Bound on Spectral Norm of Convolutional Layers",
                        "abstract": "Controlling the spectral norm of the Jacobian matrix, which is related to the convolution operation, has been shown to improve generalization, training stability and robustness in CNNs. Existing methods for computing the norm either tend to overestimate it or their performance may deteriorate quickly with increasing the input and kernel sizes. In this paper, we demonstrate that the tensor version of the spectral norm of a four-dimensional convolution kernel, up to a constant factor, serves as an upper bound for the spectral norm of the Jacobian matrix associated with the convolution operation. This new upper bound is independent of the input image resolution, differentiable and can be efficiently calculated during training. Through experiments, we demonstrate how this new bound can be used to improve the performance of convolutional architectures."
                    },
                    {
                        "title": "Low rank tensor approximation of singularly perturbed partial differential equations in one dimension",
                        "abstract": "We derive rank bounds on the quantized tensor train (QTT) compressed approximation of singularly perturbed reaction diffusion partial differential equations (PDEs) in one dimension. Specifically, we show that, independently of the scale of the singular perturbation parameter, a numerical solution with accuracy $0<\\epsilon<1$ can be represented in QTT format with a number of parameters that depends only polylogarithmically on $\\epsilon$. In other words, QTT compressed solutions converge exponentially to the exact solution, with respect to a root of the number of parameters. We also verify the rank bound estimates numerically, and overcome known stability issues of the QTT based solution of PDEs by adapting a preconditioning strategy to obtain stable schemes at all scales. We find, therefore, that the QTT based strategy is a rapidly converging algorithm for the solution of singularly perturbed PDEs, which does not require prior knowledge on the scale of the singular perturbation and on the shape of the boundary layers."
                    },
                    {
                        "title": "T-Basis: a Compact Representation for Neural Networks",
                        "abstract": "We introduce T-Basis, a novel concept for a compact representation of a set of tensors, each of an arbitrary shape, which is often seen in Neural Networks. Each of the tensors in the set is modeled using Tensor Rings, though the concept applies to other Tensor Networks. Owing its name to the T-shape of nodes in diagram notation of Tensor Rings, T-Basis is simply a list of equally shaped three-dimensional tensors, used to represent Tensor Ring nodes. Such representation allows us to parameterize the tensor set with a small number of parameters (coefficients of the T-Basis tensors), scaling logarithmically with each tensor's size in the set and linearly with the dimensionality of T-Basis. We evaluate the proposed approach on the task of neural network compression and demonstrate that it reaches high compression rates at acceptable performance drops. Finally, we analyze memory and operation requirements of the compressed networks and conclude that T-Basis networks are equally well suited for training and inference in resource-constrained environments and usage on the edge devices."
                    },
                    {
                        "title": "Cherry-Picking Gradients: Learning Low-Rank Embeddings of Visual Data via Differentiable Cross-Approximation",
                        "abstract": "We propose an end-to-end trainable framework that processes large-scale visual data tensors by looking at a fraction of their entries only. Our method combines a neural network encoder with a tensor train decomposition to learn a low-rank latent encoding, coupled with cross-approximation (CA) to learn the representation through a subset of the original samples. CA is an adaptive sampling algorithm that is native to tensor decompositions and avoids working with the full high-resolution data explicitly. Instead, it actively selects local representative samples that we fetch out-of-core and on-demand. The required number of samples grows only logarithmically with the size of the input. Our implicit representation of the tensor in the network enables processing large grids that could not be otherwise tractable in their uncompressed form. The proposed approach is particularly useful for large-scale multidimensional grid data (e.g., 3D tomography), and for tasks that require context over a large receptive field (e.g., predicting the medical condition of entire organs). The code is available at https://github.com/aelphy/c-pic."
                    },
                    {
                        "title": "Spectral Tensor Train Parameterization of Deep Learning Layers",
                        "abstract": "We study low-rank parameterizations of weight matrices with embedded spectral properties in the Deep Learning context. The low-rank property leads to parameter efficiency and permits taking computational shortcuts when computing mappings. Spectral properties are often subject to constraints in optimization problems, leading to better models and stability of optimization. We start by looking at the compact SVD parameterization of weight matrices and identifying redundancy sources in the parameterization. We further apply the Tensor Train (TT) decomposition to the compact SVD components, and propose a non-redundant differentiable parameterization of fixed TT-rank tensor manifolds, termed the Spectral Tensor Train Parameterization (STTP). We demonstrate the effects of neural network compression in the image classification setting and both compression and improved training stability in the generative adversarial training setting."
                    }
                ]
            }
        }
    },
    "2305.18475": {
        "paper_data": {
            "title": "Approximation Rate of the Transformer Architecture for Sequence Modeling",
            "url": "http://arxiv.org/abs/2305.18475v2",
            "arxiv_id": "2305.18475",
            "authors": [
                "Haotian Jiang",
                "Qianxiao Li"
            ],
            "abstract": "The Transformer architecture is widely applied in sequence modeling applications, yet the theoretical understanding of its working principles remains limited. In this work, we investigate the approximation rate for single-layer Transformers with one head. We consider a class of non-linear relationships and identify a novel notion of complexity measures to establish an explicit Jackson-type approximation rate estimate for the Transformer. This rate reveals the structural properties of the Transformer and suggests the types of sequential relationships it is best suited for approximating. In particular, the results on approximation rates enable us to concretely analyze the differences between the Transformer and classical sequence modeling methods, such as recurrent neural networks.",
            "introduction": " Introduction The Transformer network, as introduced by Vaswani et al. (2017) has become immensely popular in the field of sequence modeling. Variants such as BERT Devlin et al. (2019) and GPT Brown et al. (2020) have achieved excellent performance, becoming the default choices for natural language pro- cessing (NLP) problems. Concurrently, Dosovitskiy et al. (2020) successfully applied the Transformer to image classification problems by flattening the image into a sequence of patches. Despite the success across various fields of practical applications, many theoretical questions remain unanswered. Among these, we focus on two essential questions in this work: firstly, the approximation rate of the Transformer on sequence modeling; secondly, the comparative advantages and disadvantages of the Transformer with recurrent neural networks (RNNs) on different temporal structures. The concept of Jackson-type approximation rates is derived from the Jackson Theorem for polyno- mials Jackson (1930). This involves defining an appropriate approximation space, accompanied by specific complexity measures, precisely the Sobolev space. According to the Jackson theorem, a func- tion with a small Sobolev norm can be efficiently approximated by a polynomial. We aim to establish similar approximation rate Experiments in Section 6 Comparison of the Transformer and RNNs We consider the following linear target relationship: Ht(x) =tX t=0\u03c1(s)x(t\u2212s), (69) with kernel \u03c1(s) = exp ( \u2212s). The input sequence of length 32 is generated using a uniform distribution on [0,1]. For the RNN, we utilize a one-layer vanilla RNN architecture with 128 hidden units, employing a linear activation function. For the Transformer, we use the simplified structure as presented in Equation (8) with position encoding added. For parameter settings, we have 21n=dv=mh= 32 . The feed-forward part is constructed using a dense network with a width 128 and a depth 3. To change the temporal ordering, we permute the input while keeping the output unchanged, this leads to a change in the temporal ordering of the target relationship. For all the input sequences, we exchange the first 10 values with the last 10 values, resulting in a change in the underlying temporal structure. For the temporal mixing operation, we use a randomly generated length 5filter and the operation\u2217described in Section 6.2. In Table 2 it is also interesting to note that for this target, the RNN performs better than the Transformer. This is potentially because temporal ordering is inherently important in this target, and the Transformer is not good at capturing such relationships. Some variants of the Transformer, such as Informer Zhou et al. (2021) and Autoformer Wu et al. (2022), have been proposed specifically to address temporal relationships, particularly in sequence forecasting tasks. However, empirical Related work We first review the approximation Appendix C. The references. This flexibility allows for a broader application of the complexity measure in different scenarios and settings. Next, we proceed to define the complexity measure for H\u2208 H. This measure takes into account all the components that need to be approximated using the feed-forward network. C(\u03b2) 2(H, k) = inf F,f,\u03c1\u0010 C(\u03b2) FF(F) +C(\u03b2) FF(f) +C(\u03b2) FF(\u03c1, k)\u0011 , (14) where C(\u03b2) FF(\u03c1, k) =kX i=1(C(\u03b2) FF(\u03d5i) +C(\u03b2) FF(\u03c8i))considers the approximation of the approximation of the POD bases \u03d5iand\u03c8ifor the target function \u03c1. Notably, this complexity measure is dependent on a parameter k, which determines the number of bases we want to consider in the approximation of \u03c1. Finally, we define a complexity measure considering the norm of the target. C0(H) =",
            "references": [
                {
                    "title": "Are Transformers with One Layer Self-Attention Using Low-Rank Weight Matrices Universal Approximators?",
                    "abstract": "Existing analyses of the expressive capacity of Transformer models have required excessively deep layers for data memorization, leading to a discrepancy with the Transformers actually used in practice. This is primarily due to the interpretation of the softmax function as an approximation of the hardmax function. By clarifying the connection between the softmax function and the Boltzmann operator, we prove that a single layer of self-attention with low-rank weight matrices possesses the capability to perfectly capture the context of an entire input sequence. As a consequence, we show that one-layer and single-head Transformers have a memorization capacity for finite samples, and that Transformers consisting of one self-attention layer with two feed-forward neural networks are universal approximators for continuous permutation equivariant functions on a compact domain."
                },
                {
                    "title": "Representational Strengths and Limitations of Transformers",
                    "abstract": "Attention layers, as commonly used in transformers, form the backbone of modern deep learning, yet there is no mathematical description of their benefits and deficiencies as compared with other architectures. In this work we establish both positive and negative results on the representation power of attention layers, with a focus on intrinsic complexity parameters such as width, depth, and embedding dimension. On the positive side, we present a sparse averaging task, where recurrent networks and feedforward networks all have complexity scaling polynomially in the input size, whereas transformers scale merely logarithmically in the input size; furthermore, we use the same construction to show the necessity and role of a large embedding dimension in a transformer. On the negative side, we present a triple detection task, where attention layers in turn have complexity scaling linearly in the input size; as this scenario seems rare in practice, we also present natural variants that can be efficiently solved by attention layers. The proof techniques emphasize the value of communication complexity in the analysis of transformers and related models, and the role of sparse averaging as a prototypical attention task, which even finds use in the analysis of triple detection."
                },
                {
                    "title": "Approximation and Estimation Ability of Transformers for Sequence-to-Sequence Functions with Infinite Dimensional Input",
                    "abstract": "Despite the great success of Transformer networks in various applications such as natural language processing and computer vision, their theoretical aspects are not well understood. In this paper, we study the approximation and estimation ability of Transformers as sequence-to-sequence functions with infinite dimensional inputs. Although inputs and outputs are both infinite dimensional, we show that when the target function has anisotropic smoothness, Transformers can avoid the curse of dimensionality due to their feature extraction ability and parameter sharing property. In addition, we show that even if the smoothness changes depending on each input, Transformers can estimate the importance of features for each input and extract important features dynamically. Then, we proved that Transformers achieve similar convergence rate as in the case of the fixed smoothness. Our theoretical results support the practical success of Transformers for high dimensional data."
                },
                {
                    "title": "Are Transformers Effective for Time Series Forecasting?",
                    "abstract": "Recently, there has been a surge of Transformer-based solutions for the long-term time series forecasting (LTSF) task. Despite the growing performance over the past few years, we question the validity of this line of research in this work. Specifically, Transformers is arguably the most successful solution to extract the semantic correlations among the elements in a long sequence. However, in time series modeling, we are to extract the temporal relations in an ordered set of continuous points. While employing positional encoding and using tokens to embed sub-series in Transformers facilitate preserving some ordering information, the nature of the permutation-invariant self-attention mechanism inevitably results in temporal information loss. \nTo validate our claim, we introduce a set of embarrassingly simple one-layer linear models named LTSF-Linear for comparison. Experimental results on nine real-life datasets show that LTSF-Linear surprisingly outperforms existing sophisticated Transformer-based LTSF models in all cases, and often by a large margin. Moreover, we conduct comprehensive empirical studies to explore the impacts of various design elements of LTSF models on their temporal relation extraction capability. We hope this surprising finding opens up new research directions for the LTSF task. We also advocate revisiting the validity of Transformer-based solutions for other time series analysis tasks (e.g., anomaly detection) in the future."
                },
                {
                    "title": "Your Transformer May Not be as Powerful as You Expect",
                    "abstract": "Relative Positional Encoding (RPE), which encodes the relative distance between any pair of tokens, is one of the most successful modifications to the original Transformer. As far as we know, theoretical understanding of the RPE-based Transformers is largely unexplored. In this work, we mathematically analyze the power of RPE-based Transformers regarding whether the model is capable of approximating any continuous sequence-to-sequence functions. One may naturally assume the answer is in the affirmative -- RPE-based Transformers are universal function approximators. However, we present a negative result by showing there exist continuous sequence-to-sequence functions that RPE-based Transformers cannot approximate no matter how deep and wide the neural network is. One key reason lies in that most RPEs are placed in the softmax attention that always generates a right stochastic matrix. This restricts the network from capturing positional information in the RPEs and limits its capacity. To overcome the problem and make the model more powerful, we first present sufficient conditions for RPE-based Transformers to achieve universal function approximation. With the theoretical guidance, we develop a novel attention module, called Universal RPE-based (URPE) Attention, which satisfies the conditions. Therefore, the corresponding URPE-based Transformers become universal function approximators. Extensive experiments covering typical architectures and tasks demonstrate that our model is parameter-efficient and can achieve superior performance to strong baselines in a wide range of applications. The code will be made publicly available at https://github.com/lsj2408/URPE."
                },
                {
                    "title": "On the Rate of Convergence of a Classifier Based on a Transformer Encoder",
                    "abstract": "Pattern recognition based on a high-dimensional predictor is considered. A classifier is defined which is based on a Transformer encoder. The rate of convergence of the misclassification probability of the classifier towards the optimal misclassification probability is analyzed. It is shown that this classifier is able to circumvent the curse of dimensionality provided the a posteriori probability satisfies a suitable hierarchical composition model. Furthermore, the difference between the Transformer classifiers theoretically analyzed in this paper and the ones used in practice today is illustrated by means of classification problems in natural language processing."
                },
                {
                    "title": "Can Vision Transformers Perform Convolution?",
                    "abstract": "Several recent studies have demonstrated that attention-based networks, such as Vision Transformer (ViT), can outperform Convolutional Neural Networks (CNNs) on several computer vision tasks without using convolutional layers. This naturally leads to the following questions: Can a self-attention layer of ViT express any convolution operation? In this work, we prove that a single ViT layer with image patches as the input can perform any convolution operation constructively, where the multi-head attention mechanism and the relative positional encoding play essential roles. We further provide a lower bound on the number of heads for Vision Transformers to express CNNs. Corresponding with our analysis, experimental results show that the construction in our proof can help inject convolutional bias into Transformers and significantly improve the performance of ViT in low data regimes."
                },
                {
                    "title": "Inductive Biases and Variable Creation in Self-Attention Mechanisms",
                    "abstract": "Self-attention, an architectural motif designed to model long-range interactions in sequential data, has driven numerous recent breakthroughs in natural language processing and beyond. This work provides a theoretical analysis of the inductive biases of self-attention modules. Our focus is to rigorously establish which functions and long-range dependencies self-attention blocks prefer to represent. Our main result shows that bounded-norm Transformer networks\"create sparse variables\": a single self-attention head can represent a sparse function of the input sequence, with sample complexity scaling only logarithmically with the context length. To support our analysis, we present synthetic experiments to probe the sample complexity of learning sparse Boolean functions with Transformers."
                },
                {
                    "title": "Universal Approximation Under Constraints is Possible with Transformers",
                    "abstract": "Many practical problems need the output of a machine learning model to satisfy a set of constraints, $K$. Nevertheless, there is no known guarantee that classical neural network architectures can exactly encode constraints while simultaneously achieving universality. We provide a quantitative constrained universal approximation theorem which guarantees that for any non-convex compact set $K$ and any continuous function $f:\\mathbb{R}^n\\rightarrow K$, there is a probabilistic transformer $\\hat{F}$ whose randomized outputs all lie in $K$ and whose expected output uniformly approximates $f$. Our second main result is a\"deep neural version\"of Berge's Maximum Theorem (1963). The result guarantees that given an objective function $L$, a constraint set $K$, and a family of soft constraint sets, there is a probabilistic transformer $\\hat{F}$ that approximately minimizes $L$ and whose outputs belong to $K$; moreover, $\\hat{F}$ approximately satisfies the soft constraints. Our results imply the first universal approximation theorem for classical transformers with exact convex constraint satisfaction. They also yield that a chart-free universal approximation theorem for Riemannian manifold-valued functions subject to suitable geodesically convex constraints."
                },
                {
                    "title": "Approximation Theory of Convolutional Architectures for Time Series Modelling",
                    "abstract": "We study the approximation properties of convolutional architectures applied to time series modelling, which can be formulated mathematically as a functional approximation problem. In the recurrent setting, recent results reveal an intricate connection between approximation efficiency and memory structures in the data generation process. In this paper, we derive parallel results for convolutional architectures, with WaveNet being a prime example. Our results reveal that in this new setting, approximation efficiency is not only characterised by memory, but also additional fine structures in the target relationship. This leads to a novel definition of spectrum-based regularity that measures the complexity of temporal relationships under the convolutional approximation scheme. These analyses provide a foundation to understand the differences between architectural choices for time series modelling and can give theoretically grounded guidance for practical applications."
                },
                {
                    "title": "Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting",
                    "abstract": "Extending the forecasting time is a critical demand for real applications, such as extreme weather early warning and long-term energy consumption planning. This paper studies the long-term forecasting problem of time series. Prior Transformer-based models adopt various self-attention mechanisms to discover the long-range dependencies. However, intricate temporal patterns of the long-term future prohibit the model from finding reliable dependencies. Also, Transformers have to adopt the sparse versions of point-wise self-attentions for long series efficiency, resulting in the information utilization bottleneck. Going beyond Transformers, we design Autoformer as a novel decomposition architecture with an Auto-Correlation mechanism. We break with the pre-processing convention of series decomposition and renovate it as a basic inner block of deep models. This design empowers Autoformer with progressive decomposition capacities for complex time series. Further, inspired by the stochastic process theory, we design the Auto-Correlation mechanism based on the series periodicity, which conducts the dependencies discovery and representation aggregation at the sub-series level. Auto-Correlation outperforms self-attention in both efficiency and accuracy. In long-term forecasting, Autoformer yields state-of-the-art accuracy, with a 38% relative improvement on six benchmarks, covering five practical applications: energy, traffic, economics, weather and disease. Code is available at this repository: \\url{https://github.com/thuml/Autoformer}."
                },
                {
                    "title": "Attention is Not All You Need: Pure Attention Loses Rank Doubly Exponentially with Depth",
                    "abstract": "Attention-based architectures have become ubiquitous in machine learning, yet our understanding of the reasons for their effectiveness remains limited. This work proposes a new way to understand self-attention networks: we show that their output can be decomposed into a sum of smaller terms, each involving the operation of a sequence of attention heads across layers. Using this decomposition, we prove that self-attention possesses a strong inductive bias towards\"token uniformity\". Specifically, without skip connections or multi-layer perceptrons (MLPs), the output converges doubly exponentially to a rank-1 matrix. On the other hand, skip connections and MLPs stop the output from degeneration. Our experiments verify the identified convergence phenomena on different variants of standard transformer architectures."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "On the Curse of Memory in Recurrent Neural Networks: Approximation and Optimization Analysis",
                    "abstract": "We study the approximation properties and optimization dynamics of recurrent neural networks (RNNs) when applied to learn input-output relationships in temporal data. We consider the simple but representative setting of using continuous-time linear RNNs to learn from data generated by linear relationships. Mathematically, the latter can be understood as a sequence of linear functionals. We prove a universal approximation theorem of such linear functionals, and characterize the approximation rate and its relation with memory. Moreover, we perform a fine-grained dynamical analysis of training linear RNNs, which further reveal the intricate interactions between memory and learning. A unifying theme uncovered is the non-trivial effect of memory, a notion that can be made precise in our framework, on approximation and optimization: when there is long term memory in the target, it takes a large number of neurons to approximate it. Moreover, the training process will suffer from slow downs. In particular, both of these effects become exponentially more pronounced with memory - a phenomenon we call the \"curse of memory\". These analyses represent a basic step towards a concrete mathematical understanding of new phenomenon that may arise in learning temporal relationships using recurrent architectures."
                },
                {
                    "title": "Limits to Depth Efficiencies of Self-Attention",
                    "abstract": "Self-attention architectures, which are rapidly pushing the frontier in natural language processing, demonstrate a surprising depth-inefficient behavior: Empirical signals indicate that increasing the internal representation (network width) is just as useful as increasing the number of self-attention layers (network depth). In this paper, we theoretically study the interplay between depth and width in self-attention, and shed light on the root of the above phenomenon. We invalidate the seemingly plausible hypothesis by which widening is as effective as deepening for self-attention, and show that in fact stacking self-attention layers is so effective that it quickly saturates a capacity of the network width. Specifically, we pinpoint a \"depth threshold\" that is logarithmic in $d_x$, the network width: $L_{\\textrm{th}}=\\log_{3}(d_x)$. For networks of depth that is below the threshold, we establish a double-exponential depth-efficiency of the self-attention operation, while for depths over the threshold we show that depth-inefficiency kicks in. Our predictions strongly accord with extensive empirical ablations in Kaplan et al. (2020), accounting for the different behaviors in the two depth-(in)efficiency regimes. By identifying network width as a limiting factor, our analysis indicates that solutions for dramatically increasing the width can facilitate the next leap in self-attention expressivity."
                },
                {
                    "title": "$O(n)$ Connections are Expressive Enough: Universal Approximability of Sparse Transformers",
                    "abstract": "Transformer networks use pairwise attention to compute contextual embeddings of inputs, and have redefined the state of the art in many NLP tasks. However, these models suffer from quadratic computational cost in the input sequence length $n$ to compute attention in each layer. This has prompted recent research into faster attention models, with a predominant approach involving sparsifying the connections in the attention layers. While empirically promising for long sequences, fundamental questions remain unanswered: Can sparse transformers approximate any arbitrary sequence-to-sequence function, similar to their dense counterparts? How does the sparsity pattern and the sparsity level affect their performance? In this paper, we address these questions and provide a unifying framework that captures existing sparse attention models. Our analysis proposes sufficient conditions under which we prove that a sparse attention model can universally approximate any sequence-to-sequence function. Surprisingly, our results show the existence of models with only $O(n)$ connections per attention layer that can approximate the same function class as the dense model with $n^2$ connections. Lastly, we present experiments comparing different patterns/levels of sparsity on standard NLP tasks."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Low-Rank Bottleneck in Multi-head Attention Models",
                    "abstract": "Attention based Transformer architecture has enabled significant advances in the field of natural language processing. In addition to new pre-training techniques, recent improvements crucially rely on working with a relatively larger embedding dimension for tokens. Unfortunately, this leads to models that are prohibitively large to be employed in the downstream tasks. In this paper we identify one of the important factors contributing to the large embedding size requirement. In particular, our analysis highlights that the scaling between the number of heads and the size of each head in the current architecture gives rise to a low-rank bottleneck in attention heads, causing this limitation. We further validate this in our experiments. As a solution we propose to set the head size of an attention unit to input sequence length, and independent of the number of heads, resulting in multi-head attention layers with provably more expressive power. We empirically show that this allows us to train models with a relatively smaller embedding dimension and with better performance scaling."
                },
                {
                    "title": "Are Transformers universal approximators of sequence-to-sequence functions?",
                    "abstract": "Despite the widespread adoption of Transformer models for NLP tasks, the expressive power of these models is not well-understood. In this paper, we establish that Transformer models are universal approximators of continuous permutation equivariant sequence-to-sequence functions with compact support, which is quite surprising given the amount of shared parameters in these models. Furthermore, using positional encodings, we circumvent the restriction of permutation equivariance, and show that Transformer models can universally approximate arbitrary continuous sequence-to-sequence functions on a compact domain. Interestingly, our proof techniques clearly highlight the different roles of the self-attention and the feed-forward layers in Transformers. In particular, we prove that fixed width self-attention layers can compute contextual mappings of the input sequences, playing a key role in the universal approximation property of Transformers. Based on this insight from our analysis, we consider other architectures that can compute contextual mappings and empirically evaluate them."
                },
                {
                    "title": "On the Relationship between Self-Attention and Convolutional Layers",
                    "abstract": "Recent trends of incorporating attention mechanisms in vision have led researchers to reconsider the supremacy of convolutional layers as a primary building block. Beyond helping CNNs to handle long-range dependencies, Ramachandran et al. (2019) showed that attention can completely replace convolution and achieve state-of-the-art performance on vision tasks. This raises the question: do learned attention layers operate similarly to convolutional layers? This work provides evidence that attention layers can perform convolution and, indeed, they often learn to do so in practice. Specifically, we prove that a multi-head self-attention layer with sufficient number of heads is at least as expressive as any convolutional layer. Our numerical experiments then show that self-attention layers attend to pixel-grid patterns similarly to CNN layers, corroborating our analysis. Our code is publicly available."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "On the approximation properties of recurrent encoder-decoder architectures",
                    "abstract": "Encoder-decoder architectures have recently gained popularity in sequence to sequence modelling, featuring in state-of-the-art models such as trans-formers. However, a mathematical understanding of their working principles still remains limited. In this paper, we study the approximation properties of recurrent encoder-decoder architectures. Prior work established theoretical results for RNNs in the linear setting, where approximation capabilities can be related to smoothness and memory of target temporal relationships. Here, we uncover that the encoder and decoder together form a particular \u201ctemporal product structure\u201d which determines the approximation e\ufb03ciency. Moreover, the encoder-decoder architecture generalises RNNs with the capability to learn time-inhomogeneous relationships. Our results provide the theoretical understanding of approximation properties of the recurrent encoder-decoder architecture, which precisely characterises, in the considered setting, the types of temporal relationships that can be e\ufb03ciently learned."
                },
                {
                    "title": "Spectral properties of the kernel matrix and their relation to kernel methods in machine learning",
                    "abstract": "This chapter serves as a brief introduction to the supervised learning setting and kernel methods. Moreover, several results from linear algebra, probability theory, and functional analysis are reviewed which will be used throughout the thesis. 2.1 Some notational conventions We begin by introducing some basic notational conventions. The sets N, Z, R, C denote the natural, integer, real, and complex numbers. Vectors will be denoted by lowercase letters, whereas matrices will be denoted by bold uppercase letters. Random variables will be denoted by uppercase letters. The individual entries of vectors and matrices are denoted by square brackets. For example, x \u2208 R is a vector with coefficients [x]i. The matrix A has entries [A]ij . Vector and matrix transpose is denoted by x>. Sometimes, the set of square n \u00d7 n matrices are denoted by Mn, and the set of general n \u00d7 m matrices by Mn,m. The set of eigenvalues of a square matrix A is denoted by \u03bb(A). For a symmetric n \u00d7 n matrix A, we will always assume that the eigenvalues and eigenvectors are sorted in non-increasing order with eigenvalues repeated according to their multiplicity. The eigenvalues of A are thus \u03bb1(A) \u2265 . . . \u2265 \u03bbn(A). We use the following standard norms on finite-dimensional vector spaces. Let x \u2208 R and A \u2208Mn. Then, \u2016x\u2016 = \u221a\u221a\u221a\u221a n \u2211 i=1 [x]i , \u2016A\u2016 = max x : \u2016x\u20166=0 \u2016Ax\u2016 \u2016x\u2016 . (2.1) A useful upper bound on \u2016A\u2016 is given by \u2016A\u2016 \u2264 n max 1\u2264i,j\u2264n |[A]ij |. (2.2) Another matrix norm we will encounter is the Frobenius norm \u2016A\u2016F = \u221a\u221a\u221a\u221a n \u2211"
                },
                {
                    "title": "The Proper Orthogonal Decomposition in the Analysis of Turbulent Flows",
                    "abstract": "It has often been remarked that turbulence is a subject of great scientific and technological importance, and yet one of the least understood (e.g. McComb 1990). To an outsider this may seem strange, since the basic physical laws of fluid mechanics are well established, an excellent mathematical model is available in the Navier-Stokes equations, and the results of well over a century of increasingly sophisticated experiments are at our disposal. One major difficulty, of course, is that the governing equations are nonlinear and little is known about their solutions at high Reynolds number, even in simple geometries. Even mathematical questions as basic as existence and uniqueness are unsettled in three spatial dimensions (cf Temam 1988). A second problem, more important from the physical viewpoint, is that experiments and the available mathematical evidence all indicate that turbulence involves the interaction of many degrees of freedom over broad ranges of spatial and temporal scales. One of the problems of turbulence is to derive this complex picture from the simple laws of mass and momentum balance enshrined in the NavierStokes equations. It was to this that Ruelle & Takens (1971) contributed with their suggestion that turbulence might be a manifestation in physical"
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nWhat is the approximation rate of the Transformer network in sequence modeling compared to recurrent neural networks (RNNs) across different temporal structures?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses fundamental theoretical questions regarding the capabilities of Transformer networks in sequence modeling, which have become the standard in natural language processing and image classification. Understanding the approximation rates can lead to improved model designs and better performance in various applications, including time-series forecasting and sequential data analysis. This research could pave the way for future studies that explore the limitations and strengths of different architectures, ultimately advancing knowledge in deep learning and leading to more effective practical applications.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities of defining appropriate approximation spaces and complexity measures, particularly in the context of Sobolev spaces. Naive approaches may fail because they do not account for the intricate relationships in temporal data that RNNs are designed to capture. Additionally, the theoretical underpinnings of approximation rates require rigorous mathematical analysis, which can be technically demanding. Overcoming these obstacles involves not only a deep understanding of the mathematical foundations but also the ability to effectively compare the performance of Transformers and RNNs under varying temporal structures.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on empirical performance rather than theoretical analysis, leading to a lack of understanding of the underlying approximation capabilities of Transformers compared to RNNs. Existing solutions may have overlooked the importance of temporal ordering in sequence data, which is critical for accurate modeling. Barriers such as the complexity of defining and measuring approximation rates and the absence of a comprehensive framework for comparison have prevented this problem from being adequately addressed. Our approach differs by explicitly focusing on the theoretical aspects of approximation rates and providing a structured comparison between the two architectures.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves defining the approximation space using Sobolev norms and establishing a framework for measuring the approximation rates of both Transformers and RNNs. We will utilize a dataset generated from a linear target relationship with a uniform distribution, employing a one-layer vanilla RNN and a simplified Transformer structure for comparison. The performance will be evaluated based on the accuracy of approximating the target function under different temporal structures. We expect to demonstrate that while"
            }
        },
        "author_data": {
            "c71aeda3-86bd-4d42-b408-8109cfc60085": {
                "pk": "c71aeda3-86bd-4d42-b408-8109cfc60085",
                "name": "Haotian Jiang",
                "collaborators": [
                    "Qianxiao Li",
                    "Nikhil Bansal",
                    "Sahil Singla",
                    "Victor Reis",
                    "Andrei Graur",
                    "Aaron Sidford",
                    "Zhong Li",
                    "Jian Li",
                    "Venkatesan Guruswami",
                    "Deeparnab Chakrabarty"
                ],
                "domain": [
                    "Optimization",
                    "Approximation Algorithms",
                    "Combinatorial Algorithms",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "An Attempt to Design a Better Algorithm for the Uncapacitated Facility Location Problem",
                        "abstract": "The uncapacitated facility location has always been an important problem due to its connection to operational research and infrastructure planning. Byrka obtained an algorithm that is parametrized by $\\gamma$ and proved that it is optimal when $\\gamma>1.6774$. He also proved that the algorithm achieved an approximation ratio of 1.50. A later work by Shi Li achieved an approximation factor of 1.488. In this research, we studied these algorithms and several related works. Although we didn't improve upon the algorithm of Shi Li, our work did provide some insight into the problem. We also reframed the problem as a vector game, which provided a framework to design balanced algorithms for this problem."
                    },
                    {
                        "title": "Two Stage Optimization with Recourse and Revocation",
                        "abstract": "Two-stage optimization with recourse model is an important and widely used model, which has been studied extensively these years. In this article, we will look at a new variant of it, called the two-stage optimization with recourse and revocation model. This new model differs from the traditional one in that one is allowed to revoke some of his earlier decisions and withdraw part of the earlier costs, which is not unlikely in many real applications, and is therefore considered to be more realistic under many situations. We will adopt several approaches to study this model. In fact, we will develop an LP rounding scheme for some cover problems and show that they can be solved using this scheme and an adaptation of the rounding approach for the deterministic counterpart, provided the polynomial scenario assumption. Stochastic uncapacitated facility location problem will also be studied to show that the approximation algorithm that worked for the two-stage with recourse model worked for this model as well. In addition, we will use other methods to study the model."
                    },
                    {
                        "title": "Minimizing Convex Functions with Rational Minimizers",
                        "abstract": "Given a separation oracle $\\mathsf{SO}$ for a convex function $f$ defined on $\\mathbb{R}^n$ that has an integral minimizer inside a box with radius $R$, we show how to find an exact minimizer of $f$ using at most (a) $O(n (n \\log \\log (n)/\\log (n) + \\log(R)))$ calls to $\\mathsf{SO}$ and $\\mathsf{poly}(n, \\log(R))$ arithmetic operations, or (b) $O(n \\log(nR))$ calls to $\\mathsf{SO}$ and $\\exp(O(n)) \\cdot \\mathsf{poly}(\\log(R))$ arithmetic operations. When the set of minimizers of $f$ has integral extreme points, our algorithm outputs an integral minimizer of $f$. This improves upon the previously best oracle complexity of $O(n^2 (n + \\log(R)))$ for polynomial time algorithms and $O(n^2\\log(nR))$ for exponential time algorithms obtained by [Gr\\\"otschel, Lov\\'asz and Schrijver, Prog. Comb. Opt. 1984, Springer 1988] over thirty years ago. Our improvement on Gr\\\"otschel, Lov\\'asz and Schrijver's result generalizes to the setting where the set of minimizers of $f$ is a rational polyhedron with bounded vertex complexity.   For the Submodular Function Minimization problem, our result immediately implies a strongly polynomial algorithm that makes at most $O(n^3 \\log \\log (n)/\\log (n))$ calls to an evaluation oracle, and an exponential time algorithm that makes at most $O(n^2 \\log(n))$ calls to an evaluation oracle. These improve upon the previously best $O(n^3 \\log^2(n))$ oracle complexity for strongly polynomial algorithms given in [Lee, Sidford and Wong, FOCS 2015] and [Dadush, V\\'egh and Zambelli, SODA 2018], and an exponential time algorithm with oracle complexity $O(n^3 \\log(n))$ given in the former work.   Our result is achieved via a reduction to the Shortest Vector Problem in lattices. We analyze its oracle complexity using a potential function that simultaneously captures the size of the search set and the density of the lattice."
                    },
                    {
                        "title": "Forward and Inverse Approximation Theory for Linear Temporal Convolutional Networks",
                        "abstract": "We present a theoretical analysis of the approximation properties of convolutional architectures when applied to the modeling of temporal sequences. Specifically, we prove an approximation rate estimate (Jackson-type result) and an inverse approximation theorem (Bernstein-type result), which together provide a comprehensive characterization of the types of sequential relationships that can be efficiently captured by a temporal convolutional architecture. The rate estimate improves upon a previous result via the introduction of a refined complexity measure, whereas the inverse approximation theorem is new."
                    },
                    {
                        "title": "Near-optimal Repair of Reed-Solomon Codes with Low Sub-packetization",
                        "abstract": "Minimum storage regenerating (MSR) codes are MDS codes which allow for recovery of any single erased symbol with optimal repair bandwidth, based on the smallest possible fraction of the contents downloaded from each of the other symbols. Recently, certain Reed-Solomon codes were constructed which are MSR. However, the sub-packetization of these codes is exponentially large, growing like $n^{\\Omega(n)}$ in the constant-rate regime. In this work, we study the relaxed notion of $\\epsilon$-MSR codes, which incur a factor of $(1+\\epsilon)$ higher than the optimal repair bandwidth, in the context of Reed-Solomon codes. We give constructions of constant-rate $\\epsilon$-MSR Reed-Solomon codes with polynomial sub-packetization of $n^{O(1/\\epsilon)}$ and thereby giving an explicit tradeoff between the repair bandwidth and sub-packetization."
                    },
                    {
                        "title": "A Tighter Relation Between Hereditary Discrepancy and Determinant Lower Bound",
                        "abstract": "In seminal work, Lov\\'asz, Spencer, and Vesztergombi [European J. Combin., 1986] proved a lower bound for the hereditary discrepancy of a matrix $A \\in \\mathbb{R}^{m \\times n}$ in terms of the maximum $|\\det(B)|^{1/k}$ over all $k \\times k$ submatrices $B$ of $A$. We show algorithmically that this determinant lower bound can be off by at most a factor of $O(\\sqrt{\\log (m) \\cdot \\log (n)})$, improving over the previous bound of $O(\\log(mn) \\cdot \\sqrt{\\log (n)})$ given by Matou\\v{s}ek [Proc. of the AMS, 2013]. Our result immediately implies $\\mathrm{herdisc}(\\mathcal{F}_1 \\cup \\mathcal{F}_2) \\leq O(\\sqrt{\\log (m) \\cdot \\log (n)}) \\cdot \\max(\\mathrm{herdisc}(\\mathcal{F}_1), \\mathrm{herdisc}(\\mathcal{F}_2))$, for any two set systems $\\mathcal{F}_1, \\mathcal{F}_2$ over $[n]$ satisfying $|\\mathcal{F}_1 \\cup \\mathcal{F}_2| = m$. Our bounds are tight up to constants when $m = O(\\mathrm{poly}(n))$ due to a construction of P\\'alv\\\"olgyi [Discrete Comput. Geom., 2010] or the counterexample to Beck's three permutation conjecture by Newman, Neiman and Nikolov [FOCS, 2012]."
                    },
                    {
                        "title": "Quasi-Monte Carlo Beyond Hardy-Krause",
                        "abstract": "The classical approaches to numerically integrating a function $f$ are Monte Carlo (MC) and quasi-Monte Carlo (QMC) methods. MC methods use random samples to evaluate $f$ and have error $O(\\sigma(f)/\\sqrt{n})$, where $\\sigma(f)$ is the standard deviation of $f$. QMC methods are based on evaluating $f$ at explicit point sets with low discrepancy, and as given by the classical Koksma-Hlawka inequality, they have error $\\widetilde{O}(\\sigma_{\\mathsf{HK}}(f)/n)$, where $\\sigma_{\\mathsf{HK}}(f)$ is the variation of $f$ in the sense of Hardy and Krause. These two methods have distinctive advantages and shortcomings, and a fundamental question is to find a method that combines the advantages of both.   In this work, we give a simple randomized algorithm that produces QMC point sets with the following desirable features: (1) It achieves substantially better error than given by the classical Koksma-Hlawka inequality. In particular, it has error $\\widetilde{O}(\\sigma_{\\mathsf{SO}}(f)/n)$, where $\\sigma_{\\mathsf{SO}}(f)$ is a new measure of variation that we introduce, which is substantially smaller than the Hardy-Krause variation. (2) The algorithm only requires random samples from the underlying distribution, which makes it as flexible as MC. (3) It automatically achieves the best of both MC and QMC (and the above improvement over Hardy-Krause variation) in an optimal way. (4) The algorithm is extremely efficient, with an amortized $\\widetilde{O}(1)$ runtime per sample.   Our method is based on the classical transference principle in geometric discrepancy, combined with recent algorithmic innovations in combinatorial discrepancy that besides producing low-discrepancy colorings, also guarantee certain subgaussian properties. This allows us to bypass several limitations of previous works in bridging the gap between MC and QMC methods and go beyond the Hardy-Krause variation."
                    },
                    {
                        "title": "Improved Lower Bounds for Submodular Function Minimization",
                        "abstract": "We provide a generic technique for constructing families of submodular functions to obtain lower bounds for submodular function minimization (SFM). Applying this technique, we prove that any deterministic SFM algorithm on a ground set of $n$ elements requires at least $\\Omega(n \\log n)$ queries to an evaluation oracle. This is the first super-linear query complexity lower bound for SFM and improves upon the previous best lower bound of $2n$ given by [Graur et al., ITCS 2020]. Using our construction, we also prove that any (possibly randomized) parallel SFM algorithm, which can make up to $\\mathsf{poly}(n)$ queries per round, requires at least $\\Omega(n / \\log n)$ rounds to minimize a submodular function. This improves upon the previous best lower bound of $\\tilde{\\Omega}(n^{1/3})$ rounds due to [Chakrabarty et al., FOCS 2021], and settles the parallel complexity of query-efficient SFM up to logarithmic factors due to a recent advance in [Jiang, SODA 2021]."
                    },
                    {
                        "title": "A Brief Survey on the Approximation Theory for Sequence Modelling",
                        "abstract": "We survey current developments in the approximation theory of sequence modelling in machine learning. Particular emphasis is placed on classifying existing results for various model architectures through the lens of classical approximation paradigms, and the insights one can gain from these results. We also outline some future research directions towards building a theory of sequence modelling."
                    },
                    {
                        "title": "Online Geometric Discrepancy for Stochastic Arrivals with Applications to Envy Minimization",
                        "abstract": "Consider a unit interval $[0,1]$ in which $n$ points arrive one-by-one independently and uniformly at random. On arrival of a point, the problem is to immediately and irrevocably color it in $\\{+1,-1\\}$ while ensuring that every interval $[a,b] \\subseteq [0,1]$ is nearly-balanced. We define \\emph{discrepancy} as the largest imbalance of any interval during the entire process. If all the arriving points were known upfront then we can color them alternately to achieve a discrepancy of $1$. What is the minimum possible expected discrepancy when we color the points online?   We show that the discrepancy of the above problem is sub-polynomial in $n$ and that no algorithm can achieve a constant discrepancy. This is a substantial improvement over the trivial random coloring that only gets an $\\widetilde{O}(\\sqrt n)$ discrepancy. We then obtain similar results for a natural generalization of this problem to $2$-dimensions where the points arrive uniformly at random in a unit square. This generalization allows us to improve recent results of Benade et al.\\cite{BenadeKPP-EC18} for the online envy minimization problem when the arrivals are stochastic."
                    },
                    {
                        "title": "Approximation Theory of Convolutional Architectures for Time Series Modelling",
                        "abstract": "We study the approximation properties of convolutional architectures applied to time series modelling, which can be formulated mathematically as a functional approximation problem. In the recurrent setting, recent results reveal an intricate connection between approximation efficiency and memory structures in the data generation process. In this paper, we derive parallel results for convolutional architectures, with WaveNet being a prime example. Our results reveal that in this new setting, approximation efficiency is not only characterised by memory, but also additional fine structures in the target relationship. This leads to a novel definition of spectrum-based regularity that measures the complexity of temporal relationships under the convolutional approximation scheme. These analyses provide a foundation to understand the differences between architectural choices for time series modelling and can give theoretically grounded guidance for practical applications."
                    },
                    {
                        "title": "Resolving Matrix Spencer Conjecture Up to Poly-logarithmic Rank",
                        "abstract": "We give a simple proof of the matrix Spencer conjecture up to poly-logarithmic rank: given symmetric $d \\times d$ matrices $A_1,\\ldots,A_n$ each with $\\|A_i\\|_{\\mathsf{op}} \\leq 1$ and rank at most $n/\\log^3 n$, one can efficiently find $\\pm 1$ signs $x_1,\\ldots,x_n$ such that their signed sum has spectral norm $\\|\\sum_{i=1}^n x_i A_i\\|_{\\mathsf{op}} = O(\\sqrt{n})$. This result also implies a $\\log n - \\Omega( \\log \\log n)$ qubit lower bound for quantum random access codes encoding $n$ classical bits with advantage $\\gg 1/\\sqrt{n}$.   Our proof uses the recent refinement of the non-commutative Khintchine inequality in [Bandeira, Boedihardjo, van Handel, 2022] for random matrices with correlated Gaussian entries."
                    },
                    {
                        "title": "Practical Algorithms for Best-K Identification in Multi-Armed Bandits",
                        "abstract": "In the Best-$K$ identification problem (Best-$K$-Arm), we are given $N$ stochastic bandit arms with unknown reward distributions. Our goal is to identify the $K$ arms with the largest means with high confidence, by drawing samples from the arms adaptively. This problem is motivated by various practical applications and has attracted considerable attention in the past decade. In this paper, we propose new practical algorithms for the Best-$K$-Arm problem, which have nearly optimal sample complexity bounds (matching the lower bound up to logarithmic factors) and outperform the state-of-the-art algorithms for the Best-$K$-Arm problem (even for $K=1$) in practice."
                    },
                    {
                        "title": "Online Vector Balancing and Geometric Discrepancy",
                        "abstract": "We consider an online vector balancing question where $T$ vectors, chosen from an arbitrary distribution over $[-1,1]^n$, arrive one-by-one and must be immediately given a $\\pm$ sign. The goal is to keep the discrepancy small as possible. A concrete example is the online interval discrepancy problem where T points are sampled uniformly in [0,1], and the goal is to immediately color them $\\pm$ such that every sub-interval remains nearly balanced. As random coloring incurs $\\Omega(T^{1/2})$ discrepancy, while the offline bounds are $\\Theta(\\sqrt{n \\log (T/n)})$ for vector balancing and $1$ for interval balancing, a natural question is whether one can (nearly) match the offline bounds in the online setting for these problems. One must utilize the stochasticity as in the worst-case scenario it is known that discrepancy is $\\Omega(T^{1/2})$ for any online algorithm.   Bansal and Spencer recently show an $O(\\sqrt{n}\\log T)$ bound when each coordinate is independent. When there are dependencies among the coordinates, the problem becomes much more challenging, as evidenced by a recent work of Jiang, Kulkarni, and Singla that gives a non-trivial $O(T^{1/\\log\\log T})$ bound for online interval discrepancy. Although this beats random coloring, it is still far from the offline bound.   In this work, we introduce a new framework for online vector balancing when the input distribution has dependencies across coordinates. This lets us obtain a $poly(n, \\log T)$ bound for online vector balancing under arbitrary input distributions, and a $poly(\\log T)$ bound for online interval discrepancy. Our framework is powerful enough to capture other well-studied geometric discrepancy problems; e.g., a $poly(\\log^d (T))$ bound for the online $d$-dimensional Tusn\\'ady's problem. A key new technical ingredient is an {anti-concentration} inequality for sums of pairwise uncorrelated random variables."
                    },
                    {
                        "title": "Algorithms and Adaptivity Gaps for Stochastic $k$-TSP",
                        "abstract": "Given a metric $(V,d)$ and a $\\textsf{root} \\in V$, the classic $\\textsf{$k$-TSP}$ problem is to find a tour originating at the $\\textsf{root}$ of minimum length that visits at least $k$ nodes in $V$. In this work, motivated by applications where the input to an optimization problem is uncertain, we study two stochastic versions of $\\textsf{$k$-TSP}$.   In Stoch-Reward $k$-TSP, originally defined by Ene-Nagarajan-Saket [ENS17], each vertex $v$ in the given metric $(V,d)$ contains a stochastic reward $R_v$. The goal is to adaptively find a tour of minimum expected length that collects at least reward $k$; here \"adaptively\" means our next decision may depend on previous outcomes. Ene et al. give an $O(\\log k)$-approximation adaptive algorithm for this problem, and left open if there is an $O(1)$-approximation algorithm. We totally resolve their open question and even give an $O(1)$-approximation \\emph{non-adaptive} algorithm for this problem.   We also introduce and obtain similar results for the Stoch-Cost $k$-TSP problem. In this problem each vertex $v$ has a stochastic cost $C_v$, and the goal is to visit and select at least $k$ vertices to minimize the expected \\emph{sum} of tour length and cost of selected vertices. This problem generalizes the Price of Information framework [Singla18] from deterministic probing costs to metric probing costs.   Our techniques are based on two crucial ideas: \"repetitions\" and \"critical scaling\". We show using Freedman's and Jogdeo-Samuels' inequalities that for our problems, if we truncate the random variables at an ideal threshold and repeat, then their expected values form a good surrogate. Unfortunately, this ideal threshold is adaptive as it depends on how far we are from achieving our target $k$, so we truncate at various different scales and identify a \"critical\" scale."
                    },
                    {
                        "title": "Sparse Submodular Function Minimization",
                        "abstract": "In this paper we study the problem of minimizing a submodular function $f : 2^V \\rightarrow \\mathbb{R}$ that is guaranteed to have a $k$-sparse minimizer. We give a deterministic algorithm that computes an additive $\\epsilon$-approximate minimizer of such $f$ in $\\widetilde{O}(\\mathsf{poly}(k) \\log(|f|/\\epsilon))$ parallel depth using a polynomial number of queries to an evaluation oracle of $f$, where $|f| = \\max_{S \\subseteq V} |f(S)|$. Further, we give a randomized algorithm that computes an exact minimizer of $f$ with high probability using $\\widetilde{O}(|V| \\cdot \\mathsf{poly}(k))$ queries and polynomial time. When $k = \\widetilde{O}(1)$, our algorithms use either nearly-constant parallel depth or a nearly-linear number of evaluation oracle queries. All previous algorithms for this problem either use $\\Omega(|V|)$ parallel depth or $\\Omega(|V|^2)$ queries.   In contrast to state-of-the-art weakly-polynomial and strongly-polynomial time algorithms for SFM, our algorithms use first-order optimization methods, e.g., mirror descent and follow the regularized leader. We introduce what we call {\\em sparse dual certificates}, which encode information on the structure of sparse minimizers, and both our parallel and sequential algorithms provide new algorithmic tools for allowing first-order optimization methods to efficiently compute them. Correspondingly, our algorithm does not invoke fast matrix multiplication or general linear system solvers and in this sense is more combinatorial than previous state-of-the-art methods."
                    },
                    {
                        "title": "A New Framework for Matrix Discrepancy: Partial Coloring Bounds via Mirror Descent",
                        "abstract": "Motivated by the Matrix Spencer conjecture, we study the problem of finding signed sums of matrices with a small matrix norm. A well-known strategy to obtain these signs is to prove, given matrices $A_1, \\dots, A_n \\in \\mathbb{R}^{m \\times m}$, a Gaussian measure lower bound of $2^{-O(n)}$ for a scaling of the discrepancy body $\\{x \\in \\mathbb{R}^n: \\| \\sum_{i=1}^n x_i A_i\\| \\leq 1\\}$. We show this is equivalent to covering its polar with $2^{O(n)}$ translates of the cube $\\frac{1}{n} B^n_\\infty$, and construct such a cover via mirror descent. As applications of our framework, we show:   $\\bullet$ Matrix Spencer for Low-Rank Matrices. If the matrices satisfy $\\|A_i\\|_{\\mathrm{op}} \\leq 1$ and $\\mathrm{rank}(A_i) \\leq r$, we can efficiently find a coloring $x \\in \\{\\pm 1\\}^n$ with discrepancy $\\|\\sum_{i=1}^n x_i A_i \\|_{\\mathrm{op}} \\lesssim \\sqrt{n \\log (\\min(rm/n, r))}$. This improves upon the naive $O(\\sqrt{n \\log r})$ bound for random coloring and proves the matrix Spencer conjecture when $r m \\leq n$.   $\\bullet$ Matrix Spencer for Block Diagonal Matrices. For block diagonal matrices with $\\|A_i\\|_{\\mathrm{op}} \\leq 1$ and block size $h$, we can efficiently find a coloring $x \\in \\{\\pm 1\\}^n$ with $\\|\\sum_{i=1}^n x_i A_i \\|_{\\mathrm{op}} \\lesssim \\sqrt{n \\log (hm/n)}$. Using our proof, we reduce the matrix Spencer conjecture to the existence of a $O(\\log(m/n))$ quantum relative entropy net on the spectraplex.   $\\bullet$ Matrix Discrepancy for Schatten Norms. We generalize our discrepancy bound for matrix Spencer to Schatten norms $2 \\le p \\leq q$. Given $\\|A_i\\|_{S_p} \\leq 1$ and $\\mathrm{rank}(A_i) \\leq r$, we can efficiently find a partial coloring $x \\in [-1,1]^n$ with $|\\{i : |x_i| = 1\\}| \\ge n/2$ and $\\|\\sum_{i=1}^n x_i A_i\\|_{S_q} \\lesssim \\sqrt{n \\min(p, \\log(rk))} \\cdot k^{1/p-1/q}$, where $k := \\min(1,m/n)$."
                    },
                    {
                        "title": "Convex Minimization with Integer Minima in $\\widetilde O(n^4)$ Time",
                        "abstract": "Given a convex function $f$ on $\\mathbb{R}^n$ with an integer minimizer, we show how to find an exact minimizer of $f$ using $O(n^2 \\log n)$ calls to a separation oracle and $O(n^4 \\log n)$ time. The previous best polynomial time algorithm for this problem given in [Jiang, SODA 2021, JACM 2022] achieves $O(n^2\\log\\log n/\\log n)$ oracle complexity. However, the overall runtime of Jiang's algorithm is at least $\\widetilde{\\Omega}(n^8)$, due to expensive sub-routines such as the Lenstra-Lenstra-Lov\\'asz (LLL) algorithm [Lenstra, Lenstra, Lov\\'asz, Math. Ann. 1982] and random walk based cutting plane method [Bertsimas, Vempala, JACM 2004]. Our significant speedup is obtained by a nontrivial combination of a faster version of the LLL algorithm due to [Neumaier, Stehl\\'e, ISSAC 2016] that gives similar guarantees, the volumetric center cutting plane method (CPM) by [Vaidya, FOCS 1989] and its fast implementation given in [Jiang, Lee, Song, Wong, STOC 2020].   For the special case of submodular function minimization (SFM), our result implies a strongly polynomial time algorithm for this problem using $O(n^3 \\log n)$ calls to an evaluation oracle and $O(n^4 \\log n)$ additional arithmetic operations. Both the oracle complexity and the number of arithmetic operations of our more general algorithm are better than the previous best-known runtime algorithms for this specific problem given in [Lee, Sidford, Wong, FOCS 2015] and [Dadush, V\\'egh, Zambelli, SODA 2018, MOR 2021]."
                    },
                    {
                        "title": "A Unified PTAS for Prize Collecting TSP and Steiner Tree Problem in Doubling Metrics",
                        "abstract": "We present a unified polynomial-time approximation scheme (PTAS) for the prize collecting traveling salesman problem (PCTSP) and the prize collecting Steiner tree problem (PCSTP) in doubling metrics. Given a metric space and a penalty function on a subset of points known as terminals, a solution is a subgraph on points in the metric space, whose cost is the weight of its edges plus the penalty due to terminals not covered by the subgraph. Under our unified framework, the solution subgraph needs to be Eulerian for PCTSP, while it needs to be connected for PCSTP. Before our work, even a QPTAS for the problems in doubling metrics is not known.   Our unified PTAS is based on the previous dynamic programming frameworks proposed in [Talwar STOC 2004] and [Bartal, Gottlieb, Krauthgamer STOC 2012]. However, since it is unknown which part of the optimal cost is due to edge lengths and which part is due to penalties of uncovered terminals, we need to develop new techniques to apply previous divide-and-conquer strategies and sparse instance decompositions."
                    }
                ]
            },
            "6fe8e452-7368-4f13-a3c2-5125c60cc7a9": {
                "pk": "6fe8e452-7368-4f13-a3c2-5125c60cc7a9",
                "name": "Qianxiao Li",
                "collaborators": [
                    "Weinan E",
                    "Cheng Tai",
                    "Haotian Jiang",
                    "Sohei Arisaka",
                    "Shida Wang",
                    "Zhong Li",
                    "Zuowei Shen",
                    "Lianhai Ren",
                    "Shuji Hao",
                    "Chi Zhang"
                ],
                "domain": [
                    "Deep Learning",
                    "Control Theory",
                    "Optimization",
                    "Stochastic Processes"
                ],
                "publications": [
                    {
                        "title": "Unifying back-propagation and forward-forward algorithms through model predictive control",
                        "abstract": "We introduce a Model Predictive Control (MPC) framework for training deep neural networks, systematically unifying the Back-Propagation (BP) and Forward-Forward (FF) algorithms. At the same time, it gives rise to a range of intermediate training algorithms with varying look-forward horizons, leading to a performance-efficiency trade-off. We perform a precise analysis of this trade-off on a deep linear network, where the qualitative conclusions carry over to general networks. Based on our analysis, we propose a principled method to choose the optimization horizon based on given objectives and model specifications. Numerical results on various models and tasks demonstrate the versatility of our method."
                    },
                    {
                        "title": "An Optimal Control Approach to Deep Learning and Applications to Discrete-Weight Neural Networks",
                        "abstract": "Deep learning is formulated as a discrete-time optimal control problem. This allows one to characterize necessary conditions for optimality and develop training algorithms that do not rely on gradients with respect to the trainable parameters. In particular, we introduce the discrete-time method of successive approximations (MSA), which is based on the Pontryagin's maximum principle, for training neural networks. A rigorous error estimate for the discrete MSA is obtained, which sheds light on its dynamics and the means to stabilize the algorithm. The developed methods are applied to train, in a rather principled way, neural networks with weights that are constrained to take values in a discrete set. We obtain competitive performance and interestingly, very sparse weights in the case of ternary networks, which may be useful in model deployment in low-memory devices."
                    },
                    {
                        "title": "Forward and Inverse Approximation Theory for Linear Temporal Convolutional Networks",
                        "abstract": "We present a theoretical analysis of the approximation properties of convolutional architectures when applied to the modeling of temporal sequences. Specifically, we prove an approximation rate estimate (Jackson-type result) and an inverse approximation theorem (Bernstein-type result), which together provide a comprehensive characterization of the types of sequential relationships that can be efficiently captured by a temporal convolutional architecture. The rate estimate improves upon a previous result via the introduction of a refined complexity measure, whereas the inverse approximation theorem is new."
                    },
                    {
                        "title": "Distributed Optimization for Over-Parameterized Learning",
                        "abstract": "Distributed optimization often consists of two updating phases: local optimization and inter-node communication. Conventional approaches require working nodes to communicate with the server every one or few iterations to guarantee convergence. In this paper, we establish a completely different conclusion that each node can perform an arbitrary number of local optimization steps before communication. Moreover, we show that the more local updating can reduce the overall communication, even for an infinity number of steps where each node is free to update its local model to near-optimality before exchanging information. The extra assumption we make is that the optimal sets of local loss functions have a non-empty intersection, which is inspired by the over-paramterization phenomenon in large-scale optimization and deep learning. Our theoretical findings are confirmed by both distributed convex optimization and deep learning experiments."
                    },
                    {
                        "title": "Principled Acceleration of Iterative Numerical Methods Using Machine Learning",
                        "abstract": "Iterative methods are ubiquitous in large-scale scientific computing applications, and a number of approaches based on meta-learning have been recently proposed to accelerate them. However, a systematic study of these approaches and how they differ from meta-learning is lacking. In this paper, we propose a framework to analyze such learning-based acceleration approaches, where one can immediately identify a departure from classical meta-learning. We show that this departure may lead to arbitrary deterioration of model performance. Based on our analysis, we introduce a novel training method for learning-based acceleration of iterative methods. Furthermore, we theoretically prove that the proposed method improves upon the existing methods, and demonstrate its significant advantage and versatility through various numerical applications."
                    },
                    {
                        "title": "StableSSM: Alleviating the Curse of Memory in State-space Models through Stable Reparameterization",
                        "abstract": "In this paper, we investigate the long-term memory learning capabilities of state-space models (SSMs) from the perspective of parameterization. We prove that state-space models without any reparameterization exhibit a memory limitation similar to that of traditional RNNs: the target relationships that can be stably approximated by state-space models must have an exponential decaying memory. Our analysis identifies this \"curse of memory\" as a result of the recurrent weights converging to a stability boundary, suggesting that a reparameterization technique can be effective. To this end, we introduce a class of reparameterization techniques for SSMs that effectively lift its memory limitations. Besides improving approximation capabilities, we further illustrate that a principled choice of reparameterization scheme can also enhance optimization stability. We validate our findings using synthetic datasets, language models and image classifications."
                    },
                    {
                        "title": "Mitigating distribution shift in machine learning-augmented hybrid simulation",
                        "abstract": "We study the problem of distribution shift generally arising in machine-learning augmented hybrid simulation, where parts of simulation algorithms are replaced by data-driven surrogates. We first establish a mathematical framework to understand the structure of machine-learning augmented hybrid simulation problems, and the cause and effect of the associated distribution shift. We show correlations between distribution shift and simulation error both numerically and theoretically. Then, we propose a simple methodology based on tangent-space regularized estimator to control the distribution shift, thereby improving the long-term accuracy of the simulation results. In the linear dynamics case, we provide a thorough theoretical analysis to quantify the effectiveness of the proposed method. Moreover, we conduct several numerical experiments, including simulating a partially known reaction-diffusion equation and solving Navier-Stokes equations using the projection method with a data-driven pressure solver. In all cases, we observe marked improvements in simulation accuracy under the proposed method, especially for systems with high degrees of distribution shift, such as those with relatively strong non-linear reaction mechanisms, or flows at large Reynolds numbers."
                    },
                    {
                        "title": "DynGMA: a robust approach for learning stochastic differential equations from data",
                        "abstract": "Learning unknown stochastic differential equations (SDEs) from observed data is a significant and challenging task with applications in various fields. Current approaches often use neural networks to represent drift and diffusion functions, and construct likelihood-based loss by approximating the transition density to train these networks. However, these methods often rely on one-step stochastic numerical schemes, necessitating data with sufficiently high time resolution. In this paper, we introduce novel approximations to the transition density of the parameterized SDE: a Gaussian density approximation inspired by the random perturbation theory of dynamical systems, and its extension, the dynamical Gaussian mixture approximation (DynGMA). Benefiting from the robust density approximation, our method exhibits superior accuracy compared to baseline methods in learning the fully unknown drift and diffusion functions and computing the invariant distribution from trajectory data. And it is capable of handling trajectory data with low time resolution and variable, even uncontrollable, time step sizes, such as data generated from Gillespie's stochastic simulations. We then conduct several experiments across various scenarios to verify the advantages and robustness of the proposed method."
                    },
                    {
                        "title": "From Generalization Analysis to Optimization Designs for State Space Models",
                        "abstract": "A State Space Model (SSM) is a foundation model in time series analysis, which has recently been shown as an alternative to transformers in sequence modeling. In this paper, we theoretically study the generalization of SSMs and propose improvements to training algorithms based on the generalization results. Specifically, we give a \\textit{data-dependent} generalization bound for SSMs, showing an interplay between the SSM parameters and the temporal dependencies of the training sequences. Leveraging the generalization bound, we (1) set up a scaling rule for model initialization based on the proposed generalization measure, which significantly improves the robustness of the output value scales on SSMs to different temporal patterns in the sequence data; (2) introduce a new regularization method for training SSMs to enhance the generalization performance. Numerical results are conducted to validate our results."
                    },
                    {
                        "title": "Accelerating Legacy Numerical Solvers by Non-intrusive Gradient-based Meta-solving",
                        "abstract": "Scientific computing is an essential tool for scientific discovery and engineering design, and its computational cost is always a main concern in practice. To accelerate scientific computing, it is a promising approach to use machine learning (especially meta-learning) techniques for selecting hyperparameters of traditional numerical methods. There have been numerous proposals to this direction, but many of them require automatic-differentiable numerical methods. However, in reality, many practical applications still depend on well-established but non-automatic-differentiable legacy codes, which prevents practitioners from applying the state-of-the-art research to their own problems. To resolve this problem, we propose a non-intrusive methodology with a novel gradient estimation technique to combine machine learning and legacy numerical codes without any modification. We theoretically and numerically show the advantage of the proposed method over other baselines and present applications of accelerating established non-automatic-differentiable numerical solvers implemented in PETSc, a widely used open-source numerical software library."
                    },
                    {
                        "title": "Dynamics of Taxi-like Logistics Systems: Theory and Microscopic Simulations",
                        "abstract": "In this paper we study the dynamics of a class of bi-agent logistics systems consisting of two types of agents interacting on an arbitrary complex network. By approximating the system with simple microscopic models and solving them analytically, we reveal some universal dynamical features of such logistics systems, and propose the applications of such features for system optimisations. Large scale agent-based numerical simulations are also carried out to explore more realistic and complicated systems, with interesting emergent behaviours that can be well understood from our analytical studies. Using the taxi system as a typical logistics system with commuters and empty taxis as two types of agents, we illustrate two dynamical phases with distinct behaviours, separated by a phase boundary that can be identified as the optimal number of taxis for a particular taxi system. We show that these features, and the tuning of the optimal number of taxis, can be applied to various situations, including taxi systems allowing real-time dynamical ride-sharing. Our studies could lead to a theoretical basis for the understanding of a large class of bi-agent logistics systems, that can be useful for systematic optimisations via judicious benchmarking of routing and resource allocation strategies."
                    },
                    {
                        "title": "A Brief Survey on the Approximation Theory for Sequence Modelling",
                        "abstract": "We survey current developments in the approximation theory of sequence modelling in machine learning. Particular emphasis is placed on classifying existing results for various model architectures through the lens of classical approximation paradigms, and the insights one can gain from these results. We also outline some future research directions towards building a theory of sequence modelling."
                    },
                    {
                        "title": "Well-Posedness of the Limiting Equation of a Noisy Consensus Model in Opinion Dynamics",
                        "abstract": "This paper establishes the global well-posedness of the nonlinear Fokker-Planck equation for a noisy version of the Hegselmann-Krause model. The equation captures the mean-field behavior of a classic multiagent system for opinion dynamics. We prove the global existence, uniqueness, nonnegativity and regularity of the weak solution. We also exhibit a global stability condition, which delineates a forbidden region for consensus formation. This is the first nonlinear stability result derived for the Hegselmann-Krause model."
                    },
                    {
                        "title": "Stochastic modified equations and adaptive stochastic gradient algorithms",
                        "abstract": "We develop the method of stochastic modified equations (SME), in which stochastic gradient algorithms are approximated in the weak sense by continuous-time stochastic differential equations. We exploit the continuous formulation together with optimal control theory to derive novel adaptive hyper-parameter adjustment policies. Our algorithms have competitive performance with the added benefit of being robust to varying models and datasets. This provides a general methodology for the analysis and design of stochastic gradient algorithms."
                    },
                    {
                        "title": "Deep Learning via Dynamical Systems: An Approximation Perspective",
                        "abstract": "We build on the dynamical systems approach to deep learning, where deep residual networks are idealized as continuous-time dynamical systems, from the approximation perspective. In particular, we establish general sufficient conditions for universal approximation using continuous-time deep residual networks, which can also be understood as approximation theories in $L^p$ using flow maps of dynamical systems. In specific cases, rates of approximation in terms of the time horizon are also established. Overall, these results reveal that composition function approximation through flow maps present a new paradigm in approximation theory and contributes to building a useful mathematical framework to investigate deep learning."
                    },
                    {
                        "title": "On the Curse of Memory in Recurrent Neural Networks: Approximation and Optimization Analysis",
                        "abstract": "We study the approximation properties and optimization dynamics of recurrent neural networks (RNNs) when applied to learn input-output relationships in temporal data. We consider the simple but representative setting of using continuous-time linear RNNs to learn from data generated by linear relationships. Mathematically, the latter can be understood as a sequence of linear functionals. We prove a universal approximation theorem of such linear functionals, and characterize the approximation rate and its relation with memory. Moreover, we perform a fine-grained dynamical analysis of training linear RNNs, which further reveal the intricate interactions between memory and learning. A unifying theme uncovered is the non-trivial effect of memory, a notion that can be made precise in our framework, on approximation and optimization: when there is long term memory in the target, it takes a large number of neurons to approximate it. Moreover, the training process will suffer from slow downs. In particular, both of these effects become exponentially more pronounced with memory - a phenomenon we call the \"curse of memory\". These analyses represent a basic step towards a concrete mathematical understanding of new phenomenon that may arise in learning temporal relationships using recurrent architectures."
                    },
                    {
                        "title": "Maximum Principle Based Algorithms for Deep Learning",
                        "abstract": "The continuous dynamical system approach to deep learning is explored in order to devise alternative frameworks for training algorithms. Training is recast as a control problem and this allows us to formulate necessary optimality conditions in continuous time using the Pontryagin's maximum principle (PMP). A modification of the method of successive approximations is then used to solve the PMP, giving rise to an alternative training algorithm for deep learning. This approach has the advantage that rigorous error estimates and convergence results can be established. We also show that it may avoid some pitfalls of gradient-based methods, such as slow convergence on flat landscapes near saddle points. Furthermore, we demonstrate that it obtains favorable initial convergence rate per-iteration, provided Hamiltonian maximization can be efficiently carried out - a step which is still in need of improvement. Overall, the approach opens up new avenues to attack problems associated with deep learning, such as trapping in slow manifolds and inapplicability of gradient-based methods for discrete trainable variables."
                    },
                    {
                        "title": "A Quantitative Analysis of the Effect of Batch Normalization on Gradient Descent",
                        "abstract": "Despite its empirical success and recent theoretical progress, there generally lacks a quantitative analysis of the effect of batch normalization (BN) on the convergence and stability of gradient descent. In this paper, we provide such an analysis on the simple problem of ordinary least squares (OLS). Since precise dynamical properties of gradient descent (GD) is completely known for the OLS problem, it allows us to isolate and compare the additional effects of BN. More precisely, we show that unlike GD, gradient descent with BN (BNGD) converges for arbitrary learning rates for the weights, and the convergence remains linear under mild conditions. Moreover, we quantify two different sources of acceleration of BNGD over GD -- one due to over-parameterization which improves the effective condition number and another due having a large range of learning rates giving rise to fast descent. These phenomena set BNGD apart from GD and could account for much of its robustness properties. These findings are confirmed quantitatively by numerical experiments, which further show that many of the uncovered properties of BNGD in OLS are also observed qualitatively in more complex supervised learning problems."
                    },
                    {
                        "title": "Stochastic Modified Equations and Dynamics of Stochastic Gradient Algorithms I: Mathematical Foundations",
                        "abstract": "We develop the mathematical foundations of the stochastic modified equations (SME) framework for analyzing the dynamics of stochastic gradient algorithms, where the latter is approximated by a class of stochastic differential equations with small noise parameters. We prove that this approximation can be understood mathematically as an weak approximation, which leads to a number of precise and useful results on the approximations of stochastic gradient descent (SGD), momentum SGD and stochastic Nesterov's accelerated gradient method in the general setting of stochastic objectives. We also demonstrate through explicit calculations that this continuous-time approach can uncover important analytical insights into the stochastic gradient algorithms under consideration that may not be easy to obtain in a purely discrete-time setting."
                    }
                ]
            }
        }
    },
    "2404.12754": {
        "paper_data": {
            "title": "Adaptive Regularization of Representation Rank as an Implicit Constraint of Bellman Equation",
            "url": "http://arxiv.org/abs/2404.12754v1",
            "arxiv_id": "2404.12754",
            "authors": [
                "Qiang He",
                "Tianyi Zhou",
                "Meng Fang",
                "Setareh Maghsudi"
            ],
            "abstract": "Representation rank is an important concept for understanding the role of Neural Networks (NNs) in Deep Reinforcement learning (DRL), which measures the expressive capacity of value networks. Existing studies focus on unboundedly maximizing this rank; nevertheless, that approach would introduce overly complex models in the learning, thus undermining performance. Hence, fine-tuning representation rank presents a challenging and crucial optimization problem. To address this issue, we find a guiding principle for adaptive control of the representation rank. We employ the Bellman equation as a theoretical foundation and derive an upper bound on the cosine similarity of consecutive state-action pairs representations of value networks. We then leverage this upper bound to propose a novel regularizer, namely BEllman Equation-based automatic rank Regularizer (BEER). This regularizer adaptively regularizes the representation rank, thus improving the DRL agent's performance. We first validate the effectiveness of automatic control of rank on illustrative experiments. Then, we scale up BEER to complex continuous control tasks by combining it with the deterministic policy gradient method. Among 12 challenging DeepMind control tasks, BEER outperforms the baselines by a large margin. Besides, BEER demonstrates significant advantages in Q-value approximation. Our code is available at https://github.com/sweetice/BEER-ICLR2024.",
            "introduction": "   1 Introduction  Deep reinforcement learning (DRL), empowered by the large capacity of neural networks (NNs), has made notable strides in a variety of sequential decision-making tasks, ranging from games\u00a0(Mnih et\u00a0al., 2015; Silver et\u00a0al., 2018; Vinyals et\u00a0al., 2019), to robotics control\u00a0(Haarnoja et\u00a0al., 2018), and large language models\u00a0(Christiano et\u00a0al., 2017; OpenAI, 2023). Yet, despite these advancements, NNs within DRL systems are still largely treated as black-box function approximators\u00a0(Schulman et\u00a0al., 2017; Lillicrap et\u00a0al., 2016; Fujimoto et\u00a0al., 2018; Haarnoja et\u00a0al., 2018). Investigating the specific roles played by NNs in DRL may offer insights into their function, enabling further optimization of DRL agents from the aspect of NNs. Recent work\u00a0(Kumar et\u00a0al., 2021; Lyle et\u00a0al., 2022; He et\u00a0al., 2023a) has begun to shed light on the role of NNs in RL, proposing to leverage some properties (e.g., capacity\u00a0(Lyle et\u00a0al., 2022) and plasticity\u00a0(Lyle et\u00a0al., 2023; Nikishin et\u00a0al., 2022; Sokar et\u00a0al., 2023)) of NNs to improve DRL agents.   A core concept to study the properties of NNs in the DRL setting is the representation rank\u00a0(Kumar et\u00a0al., 2021). It measures the representation capacity of a NN by performing a singular value decomposition on the output of its representation layer, which serves as the input representation for value functions. Previous studies (e.g., InFeR\u00a0(Lyle et\u00a0al., 2022) and DR3\u00a0(Kumar et\u00a0al., 2021)) heavily focus on unbounded maximizing this metric. Yet, that approach might cause overfitting by constructing an overly complex function approximation, thus inhibiting the model\u2019s ability to generalize to new data. The resulting complexity demands more data and makes the model susceptible to noise that deters sampling efficiency and robustness in RL. As such, an indiscriminate maximization of representation rank is not advisable. A more desirable method is to adaptively control the representation rank. However, empirically fine-tuning the balance of representation rank can be tricky. An excessively complex model induces the issues mentioned above, whereas overly simple models lose the capability to obtain the optimal policy. As a result, it is imperative to search for a guiding principle that allows for the adaptive control of the representation rank.   A primary discovery of this paper is that an adaptive control of the representation rank can be derived from Bellman equation\u00a0(Sutton & Barto, 2018) as an implicitly imposed constraint on NNs. Specifically, this can be achieved by an upper bound on the cosine similarity between the representations of consecutive state-action pairs, determined by factors like the discount factor, the weights of the last layer of the neural network, and the representations norms. Cosine similarity measures the linear dependent relationship between the composition vectors, thus affecting the rank. The implication is profound: The cosine similarity is constrained, thereby restraining the representation rank itself\u00a0(Kumar et\u00a0al., 2021; Lyle et\u00a0al., 2022; 2023). However, a previous work\u00a0(Kumar et\u00a0al., 2021) shows that the dynamics of NNs result in feature co-adaptation, which potentially makes DRL agents fail to hold the bound. Thus, these derived upper bounds provide a criterion to adaptively control the representation rank. Motivated by these findings, we introduce a novel regularizer, namely BEllman Equation-based automatic rank Regularizer (BEER), designed to control the representation rank by regularizing the similarity between adjacent representations of the value",
            "references": [
                {
                    "title": "Eigensubspace of Temporal-Difference Dynamics and How It Improves Value Approximation in Reinforcement Learning",
                    "abstract": "We propose a novel value approximation method, namely Eigensubspace Regularized Critic (ERC) for deep reinforcement learning (RL). ERC is motivated by an analysis of the dynamics of Q-value approximation error in the Temporal-Difference (TD) method, which follows a path defined by the 1-eigensubspace of the transition kernel associated with the Markov Decision Process (MDP). It reveals a fundamental property of TD learning that has remained unused in previous deep RL approaches. In ERC, we propose a regularizer that guides the approximation error tending towards the 1-eigensubspace, resulting in a more efficient and stable path of value approximation. Moreover, we theoretically prove the convergence of the ERC method. Besides, theoretical analysis and experiments demonstrate that ERC effectively reduces the variance of value functions. Among 26 tasks in the DMControl benchmark, ERC outperforms state-of-the-art methods for 20. Besides, it shows significant advantages in Q-value approximation and variance reduction. Our code is available at https://sites.google.com/view/erc-ecml23/."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Understanding plasticity in neural networks",
                    "abstract": "Plasticity, the ability of a neural network to quickly change its predictions in response to new information, is essential for the adaptability and robustness of deep reinforcement learning systems. Deep neural networks are known to lose plasticity over the course of training even in relatively simple learning problems, but the mechanisms driving this phenomenon are still poorly understood. This paper conducts a systematic empirical analysis into plasticity loss, with the goal of understanding the phenomenon mechanistically in order to guide the future development of targeted solutions. We find that loss of plasticity is deeply connected to changes in the curvature of the loss landscape, but that it often occurs in the absence of saturated units. Based on this insight, we identify a number of parameterization and optimization design choices which enable networks to better preserve plasticity over the course of training. We validate the utility of these findings on larger-scale RL benchmarks in the Arcade Learning Environment."
                },
                {
                    "title": "The Dormant Neuron Phenomenon in Deep Reinforcement Learning",
                    "abstract": "In this work we identify the dormant neuron phenomenon in deep reinforcement learning, where an agent's network suffers from an increasing number of inactive neurons, thereby affecting network expressivity. We demonstrate the presence of this phenomenon across a variety of algorithms and environments, and highlight its effect on learning. To address this issue, we propose a simple and effective method (ReDo) that Recycles Dormant neurons throughout training. Our experiments demonstrate that ReDo maintains the expressive power of networks by reducing the number of dormant neurons and results in improved performance."
                },
                {
                    "title": "Frustratingly Easy Regularization on Representation Can Boost Deep Reinforcement Learning",
                    "abstract": "Deep reinforcement learning (DRL) gives the promise that an agent learns good policy from high-dimensional information, whereas representation learning removes irrelevant and redundant information and retains pertinent information. In this work, we demonstrate that the learned representation of the Q-network and its target Q-network should, in theory, satisfy a favorable distinguishable representation property. Specifically, there exists an upper bound on the representation similarity of the value functions of two adjacent time steps in a typical DRL setting. However, through illustrative experiments, we show that the learned DRL agent may violate this property and lead to a sub-optimal policy. Therefore, we propose a simple yet effective regularizer called P_olicy E_valuation with E_asy Regularization on Representation (PEER), which aims to maintain the distinguishable representation property via explicit regularization on internal representations. And we provide the convergence rate guarantee of PEER. Implementing PEER requires only one line of code. Our experiments demonstrate that incorporating PEER into DRL can significantly improve performance and sample efficiency. Comprehensive experiments show that PEER achieves state-of-the-art performance on all 4 environments on PyBullet, 9 out of 12 tasks on DM-Control, and 19 out of 26 games on Atari. To the best of our knowledge, PEER is the first work to study the inherent representation property of Q-network and its target. Our code is available at https://sites.google.com/view/peer-cvpr2023/."
                },
                {
                    "title": "The Primacy Bias in Deep Reinforcement Learning",
                    "abstract": "This work identifies a common flaw of deep reinforcement learning (RL) algorithms: a tendency to rely on early interactions and ignore useful evidence encountered later. Because of training on progressively growing datasets, deep RL agents incur a risk of overfitting to earlier experiences, negatively affecting the rest of the learning process. Inspired by cognitive science, we refer to this effect as the primacy bias. Through a series of experiments, we dissect the algorithmic aspects of deep RL that exacerbate this bias. We then propose a simple yet generally-applicable mechanism that tackles the primacy bias by periodically resetting a part of the agent. We apply this mechanism to algorithms in both discrete (Atari 100k) and continuous action (DeepMind Control Suite) domains, consistently improving their performance."
                },
                {
                    "title": "Understanding and Preventing Capacity Loss in Reinforcement Learning",
                    "abstract": "The reinforcement learning (RL) problem is rife with sources of non-stationarity, making it a notoriously difficult problem domain for the application of neural networks. We identify a mechanism by which non-stationary prediction targets can prevent learning progress in deep RL agents: \\textit{capacity loss}, whereby networks trained on a sequence of target values lose their ability to quickly update their predictions over time. We demonstrate that capacity loss occurs in a range of RL agents and environments, and is particularly damaging to performance in sparse-reward tasks. We then present a simple regularizer, Initial Feature Regularization (InFeR), that mitigates this phenomenon by regressing a subspace of features towards its value at initialization, leading to significant performance improvements in sparse-reward environments such as Montezuma's Revenge. We conclude that preventing capacity loss is crucial to enable agents to maximally benefit from the learning signals they obtain throughout the entire training trajectory."
                },
                {
                    "title": "DR3: Value-Based Deep Reinforcement Learning Requires Explicit Regularization",
                    "abstract": "Despite overparameterization, deep networks trained via supervised learning are easy to optimize and exhibit excellent generalization. One hypothesis to explain this is that overparameterized deep networks enjoy the benefits of implicit regularization induced by stochastic gradient descent, which favors parsimonious solutions that generalize well on test inputs. It is reasonable to surmise that deep reinforcement learning (RL) methods could also benefit from this effect. In this paper, we discuss how the implicit regularization effect of SGD seen in supervised learning could in fact be harmful in the offline deep RL setting, leading to poor generalization and degenerate feature representations. Our theoretical analysis shows that when existing models of implicit regularization are applied to temporal difference learning, the resulting derived regularizer favors degenerate solutions with excessive\"aliasing\", in stark contrast to the supervised learning case. We back up these findings empirically, showing that feature representations learned by a deep network value function trained via bootstrapping can indeed become degenerate, aliasing the representations for state-action pairs that appear on either side of the Bellman backup. To address this issue, we derive the form of this implicit regularizer and, inspired by this derivation, propose a simple and effective explicit regularizer, called DR3, that counteracts the undesirable effects of this implicit regularizer. When combined with existing offline RL methods, DR3 substantially improves performance and stability, alleviating unlearning in Atari 2600 games, D4RL domains and robotic manipulation from images."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "Reinforcement Learning in Linear MDPs: Constant Regret and Representation Selection",
                    "abstract": "We study the role of the representation of state-action value functions in regret minimization in finite-horizon Markov Decision Processes (MDPs) with linear structure. We first derive a necessary condition on the representation, called universally spanning optimal features (UNISOFT), to achieve constant regret in any MDP with linear reward function. This result encompasses the well-known settings of low-rank MDPs and, more generally, zero inherent Bellman error (also known as the Bellman closure assumption). We then demonstrate that this condition is also sufficient for these classes of problems by deriving a constant regret bound for two optimistic algorithms (LSVI-UCB and ELEANOR). Finally, we propose an algorithm for representation selection and we prove that it achieves constant regret when one of the given representations, or a suitable combination of them, satisfies the UNISOFT condition."
                },
                {
                    "title": "Learning Dynamics Models for Model Predictive Agents",
                    "abstract": "Model-Based Reinforcement Learning involves learning a \\textit{dynamics model} from data, and then using this model to optimise behaviour, most often with an online \\textit{planner}. Much of the recent research along these lines presents a particular set of design choices, involving problem definition, model learning and planning. Given the multiple contributions, it is difficult to evaluate the effects of each. This paper sets out to disambiguate the role of different design choices for learning dynamics models, by comparing their performance to planning with a ground-truth model -- the simulator. First, we collect a rich dataset from the training sequence of a model-free agent on 5 domains of the DeepMind Control Suite. Second, we train feed-forward dynamics models in a supervised fashion, and evaluate planner performance while varying and analysing different model design choices, including ensembling, stochasticity, multi-step training and timestep size. Besides the quantitative analysis, we describe a set of qualitative findings, rules of thumb, and future research directions for planning with learned dynamics models. Videos of the results are available at https://sites.google.com/view/learning-better-models."
                },
                {
                    "title": "Provably efficient representation selection in Low-rank Markov Decision Processes: from online to offline RL",
                    "abstract": "The success of deep reinforcement learning (DRL) lies in its ability to learn a representation that is well-suited for the exploration and exploitation task. To understand how the choice of representation can improve the efficiency of reinforcement learning (RL), we study representation selection for a class of low-rank Markov Decision Processes (MDPs) where the transition kernel can be represented in a bilinear form. We propose an efficient algorithm, called ReLEX, for representation learning in both online and offline RL. Specifically, we show that the online version of ReLEX, called ReLEX-UCB, always performs no worse than the state-of-the-art algorithm without representation selection, and achieves a strictly better constant regret if the representation function class has a\"coverage\"property over the entire state-action space. For the offline counterpart, ReLEX-LCB, we show that the algorithm can find the optimal policy if the representation class can cover the state-action space and achieves gap-dependent sample complexity. This is the first result with constant sample complexity for representation learning in offline RL."
                },
                {
                    "title": "Leveraging Good Representations in Linear Contextual Bandits",
                    "abstract": "The linear contextual bandit literature is mostly focused on the design of efficient learning algorithms for a given representation. However, a contextual bandit problem may admit multiple linear representations, each one with different characteristics that directly impact the regret of the learning algorithm. In particular, recent works showed that there exist\"good\"representations for which constant problem-dependent regret can be achieved. In this paper, we first provide a systematic analysis of the different definitions of\"good\"representations proposed in the literature. We then propose a novel selection algorithm able to adapt to the best representation in a set of $M$ candidates. We show that the regret is indeed never worse than the regret obtained by running LinUCB on the best representation (up to a $\\ln M$ factor). As a result, our algorithm achieves constant regret whenever a\"good\"representation is available in the set. Furthermore, we show that the algorithm may still achieve constant regret by implicitly constructing a\"good\"representation, even when none of the initial representations is\"good\". Finally, we empirically validate our theoretical findings in a number of standard contextual bandit problems."
                },
                {
                    "title": "On The Effect of Auxiliary Tasks on Representation Dynamics",
                    "abstract": "While auxiliary tasks play a key role in shaping the representations learnt by reinforcement learning agents, much is still unknown about the mechanisms through which this is achieved. This work develops our understanding of the relationship between auxiliary tasks, environment structure, and representations by analysing the dynamics of temporal difference algorithms. Through this approach, we establish a connection between the spectral decomposition of the transition operator and the representations induced by a variety of auxiliary tasks. We then leverage insights from these theoretical results to inform the selection of auxiliary tasks for deep reinforcement learning agents in sparse-reward environments."
                },
                {
                    "title": "Randomized Ensembled Double Q-Learning: Learning Fast Without a Model",
                    "abstract": "Using a high Update-To-Data (UTD) ratio, model-based methods have recently achieved much higher sample efficiency than previous model-free methods for continuous-action DRL benchmarks. In this paper, we introduce a simple model-free algorithm, Randomized Ensembled Double Q-Learning (REDQ), and show that its performance is just as good as, if not better than, a state-of-the-art model-based algorithm for the MuJoCo benchmark. Moreover, REDQ can achieve this performance using fewer parameters than the model-based method, and with less wall-clock run time. REDQ has three carefully integrated ingredients which allow it to achieve its high performance: (i) a UTD ratio>>1; (ii) an ensemble of Q functions; (iii) in-target minimization across a random subset of Q functions from the ensemble. Through carefully designed experiments, we provide a detailed analysis of REDQ and related model-free algorithms. To our knowledge, REDQ is the first successful model-free DRL algorithm for continuous-action spaces using a UTD ratio>>1."
                },
                {
                    "title": "Towards Continual Reinforcement Learning: A Review and Perspectives",
                    "abstract": "In this article, we aim to provide a literature review of different formulations and approaches to continual reinforcement learning (RL), also known as lifelong or non-stationary RL. We begin by discussing our perspective on why RL is a natural fit for studying continual learning. We then provide a taxonomy of different continual RL formulations by mathematically characterizing two key properties of non-stationarity, namely, the scope and driver non-stationarity. This offers a unified view of various formulations. Next, we review and present a taxonomy of continual RL approaches. We go on to discuss evaluation of continual RL agents, providing an overview of benchmarks used in the literature and important metrics for understanding agent performance. Finally, we highlight open problems and challenges in bridging the gap between the current state of continual RL and findings in neuroscience. While still in its early days, the study of continual RL has the promise to develop better incremental reinforcement learners that can function in increasingly realistic applications where non-stationarity plays a vital role. These include applications such as those in the fields of healthcare, education, logistics, and robotics."
                },
                {
                    "title": "Representations for Stable Off-Policy Reinforcement Learning",
                    "abstract": "Reinforcement learning with function approximation can be unstable and even divergent, especially when combined with off-policy learning and Bellman updates. In deep reinforcement learning, these issues have been dealt with empirically by adapting and regularizing the representation, in particular with auxiliary tasks. This suggests that representation learning may provide a means to guarantee stability. In this paper, we formally show that there are indeed nontrivial state representations under which the canonical TD algorithm is stable, even when learning off-policy. We analyze representation learning schemes that are based on the transition matrix of a policy, such as proto-value functions, along three axes: approximation error, stability, and ease of estimation. In the most general case, we show that a Schur basis provides convergence guarantees, but is difficult to estimate from samples. For a fixed reward function, we find that an orthogonal basis of the corresponding Krylov subspace is an even better choice. We conclude by empirically demonstrating that these stable representations can be learned using stochastic gradient descent, opening the door to improved techniques for representation learning with deep networks."
                },
                {
                    "title": "WD3: Taming the Estimation Bias in Deep Reinforcement Learning",
                    "abstract": "The overestimation phenomenon caused by function approximation is a well-known issue in value-based reinforcement learning algorithms such as deep Q-networks and DDPG, which could lead to suboptimal policies. To address this issue, TD3 takes the minimum value between a pair of critics, which introduces underestimation bias. By unifying these two opposites, we propose a novel Weighted Delayed Deep Deterministic Policy Gradient algorithm, which can reduce the estimation error and further improve the performance by weighting a pair of critics. We compare the learning process of value function between DDPG, TD3, and our proposed algorithm, which verifies that our algorithm could indeed eliminate the estimation error of value function. We evaluate our algorithm in the OpenAI Gym continuous control tasks, outperforming the state-of-the-art algorithms on every environment tested."
                },
                {
                    "title": "Bootstrap Latent-Predictive Representations for Multitask Reinforcement Learning",
                    "abstract": "Learning a good representation is an essential component for deep reinforcement learning (RL). Representation learning is especially important in multitask and partially observable settings where building a representation of the unknown environment is crucial to solve the tasks. Here we introduce Prediction of Bootstrap Latents (PBL), a simple and flexible self-supervised representation learning algorithm for multitask deep RL. PBL builds on multistep predictive representations of future observations, and focuses on capturing structured information about environment dynamics. Specifically, PBL trains its representation by predicting latent embeddings of future observations. These latent embeddings are themselves trained to be predictive of the aforementioned representations. These predictions form a bootstrapping effect, allowing the agent to learn more about the key aspects of the environment dynamics. In addition, by defining prediction tasks completely in latent space, PBL provides the flexibility of using multimodal observations involving pixel images, language instructions, rewards and more. We show in our experiments that PBL delivers across-the-board improved performance over state of the art deep RL agents in the DMLab-30 and Atari-57 multitask setting."
                },
                {
                    "title": "Reinforcement Learning with Augmented Data",
                    "abstract": "Learning from visual observations is a fundamental yet challenging problem in Reinforcement Learning (RL). Although algorithmic advances combined with convolutional neural networks have proved to be a recipe for success, current methods are still lacking on two fronts: (a) data-efficiency of learning and (b) generalization to new environments. To this end, we present Reinforcement Learning with Augmented Data (RAD), a simple plug-and-play module that can enhance most RL algorithms. We perform the first extensive study of general data augmentations for RL on both pixel-based and state-based inputs, and introduce two new data augmentations - random translate and random amplitude scale. We show that augmentations such as random translate, crop, color jitter, patch cutout, random convolutions, and amplitude scale can enable simple RL algorithms to outperform complex state-of-the-art methods across common benchmarks. RAD sets a new state-of-the-art in terms of data-efficiency and final performance on the DeepMind Control Suite benchmark for pixel-based control as well as OpenAI Gym benchmark for state-based control. We further demonstrate that RAD significantly improves test-time generalization over existing methods on several OpenAI ProcGen benchmarks. Our RAD module and training code are available at this https URL."
                },
                {
                    "title": "Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixels",
                    "abstract": "We propose a simple data augmentation technique that can be applied to standard model-free reinforcement learning algorithms, enabling robust learning directly from pixels without the need for auxiliary losses or pre-training. The approach leverages input perturbations commonly used in computer vision tasks to regularize the value function. Existing model-free approaches, such as Soft Actor-Critic (SAC), are not able to train deep networks effectively from image pixels. However, the addition of our augmentation method dramatically improves SAC's performance, enabling it to reach state-of-the-art performance on the DeepMind control suite, surpassing model-based (Dreamer, PlaNet, and SLAC) methods and recently proposed contrastive learning (CURL). Our approach can be combined with any model-free reinforcement learning algorithm, requiring only minor modifications. An implementation can be found at this https URL."
                },
                {
                    "title": "CURL: Contrastive Unsupervised Representations for Reinforcement Learning",
                    "abstract": "We present CURL: Contrastive Unsupervised Representations for Reinforcement Learning. CURL extracts high-level features from raw pixels using contrastive learning and performs off-policy control on top of the extracted features. CURL outperforms prior pixel-based methods, both model-based and model-free, on complex tasks in the DeepMind Control Suite and Atari Games showing 1.9x and 1.2x performance gains at the 100K environment and interaction steps benchmarks respectively. On the DeepMind Control Suite, CURL is the first image-based algorithm to nearly match the sample-efficiency of methods that use state-based features. Our code is open-sourced and available at this https URL."
                },
                {
                    "title": "Improved Baselines with Momentum Contrastive Learning",
                    "abstract": "Contrastive unsupervised learning has recently shown encouraging progress, e.g., in Momentum Contrast (MoCo) and SimCLR. In this note, we verify the effectiveness of two of SimCLR's design improvements by implementing them in the MoCo framework. With simple modifications to MoCo---namely, using an MLP projection head and more data augmentation---we establish stronger baselines that outperform SimCLR and do not require large training batches. We hope this will make state-of-the-art unsupervised learning research more accessible. Code will be made public."
                },
                {
                    "title": "Maxmin Q-learning: Controlling the Estimation Bias of Q-learning",
                    "abstract": "Q-learning suffers from overestimation bias, because it approximates the maximum action value using the maximum estimated action value. Algorithms have been proposed to reduce overestimation bias, but we lack an understanding of how bias interacts with performance, and the extent to which existing algorithms mitigate bias. In this paper, we 1) highlight that the effect of overestimation bias on learning efficiency is environment-dependent; 2) propose a generalization of Q-learning, called \\emph{Maxmin Q-learning}, which provides a parameter to flexibly control bias; 3) show theoretically that there exists a parameter choice for Maxmin Q-learning that leads to unbiased estimation with a lower approximation variance than Q-learning; and 4) prove the convergence of our algorithm in the tabular case, as well as convergence of several previous Q-learning variants, using a novel Generalized Q-learning framework. We empirically verify that our algorithm better controls estimation bias in toy environments, and that it achieves superior performance on several benchmark problems."
                },
                {
                    "title": "Momentum Contrast for Unsupervised Visual Representation Learning",
                    "abstract": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a moving-averaged encoder. This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning. MoCo provides competitive results under the common linear protocol on ImageNet classification. More importantly, the representations learned by MoCo transfer well to downstream tasks. MoCo can outperform its supervised pre-training counterpart in 7 detection/segmentation tasks on PASCAL VOC, COCO, and other datasets, sometimes surpassing it by large margins. This suggests that the gap between unsupervised and supervised representation learning has been largely closed in many vision tasks."
                },
                {
                    "title": "A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play",
                    "abstract": "One program to rule them all Computers can beat humans at increasingly complex games, including chess and Go. However, these programs are typically constructed for a particular game, exploiting its properties, such as the symmetries of the board on which it is played. Silver et al. developed a program called AlphaZero, which taught itself to play Go, chess, and shogi (a Japanese version of chess) (see the Editorial, and the Perspective by Campbell). AlphaZero managed to beat state-of-the-art programs specializing in these three games. The ability of AlphaZero to adapt to various game rules is a notable step toward achieving a general game-playing system. Science, this issue p. 1140; see also pp. 1087 and 1118 AlphaZero teaches itself to play three different board games and beats state-of-the-art programs in each. The game of chess is the longest-studied domain in the history of artificial intelligence. The strongest programs are based on a combination of sophisticated search techniques, domain-specific adaptations, and handcrafted evaluation functions that have been refined by human experts over several decades. By contrast, the AlphaGo Zero program recently achieved superhuman performance in the game of Go by reinforcement learning from self-play. In this paper, we generalize this approach into a single AlphaZero algorithm that can achieve superhuman performance in many challenging games. Starting from random play and given no domain knowledge except the game rules, AlphaZero convincingly defeated a world champion program in the games of chess and shogi (Japanese chess), as well as Go."
                },
                {
                    "title": "Representation Learning with Contrastive Predictive Coding",
                    "abstract": "While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments."
                },
                {
                    "title": "Addressing Function Approximation Error in Actor-Critic Methods",
                    "abstract": "In value-based reinforcement learning methods such as deep Q-learning, function approximation errors are known to lead to overestimated value estimates and suboptimal policies. We show that this problem persists in an actor-critic setting and propose novel mechanisms to minimize its effects on both the actor and the critic. Our algorithm builds on Double Q-learning, by taking the minimum value between a pair of critics to limit overestimation. We draw the connection between target networks and overestimation bias, and suggest delaying policy updates to reduce per-update error and further improve performance. We evaluate our method on the suite of OpenAI gym tasks, outperforming the state of the art in every environment tested."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "Deep Reinforcement Learning that Matters",
                    "abstract": "\n \n In recent years, significant progress has been made in solving challenging problems across various domains using deep reinforcement learning (RL). Reproducing existing work and accurately judging the improvements offered by novel methods is vital to sustaining this progress. Unfortunately, reproducing results for state-of-the-art deep RL methods is seldom straightforward. In particular, non-determinism in standard benchmark environments, combined with variance intrinsic to the methods, can make reported results tough to interpret. Without significance metrics and tighter standardization of experimental reporting, it is difficult to determine whether improvements over the prior state-of-the-art are meaningful. In this paper, we investigate challenges posed by reproducibility, proper experimental techniques, and reporting procedures. We illustrate the variability in reported metrics and results when comparing against common baselines and suggest guidelines to make future results in deep RL more reproducible. We aim to spur discussion about how to ensure continued progress in the field by minimizing wasted effort stemming from results that are non-reproducible and easily misinterpreted.\n \n"
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "Deep Reinforcement Learning from Human Preferences",
                    "abstract": "For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems. In this work, we explore goals defined in terms of (non-expert) human preferences between pairs of trajectory segments. We show that this approach can effectively solve complex RL tasks without access to the reward function, including Atari games and simulated robot locomotion, while providing feedback on less than one percent of our agent's interactions with the environment. This reduces the cost of human oversight far enough that it can be practically applied to state-of-the-art RL systems. To demonstrate the flexibility of our approach, we show that we can successfully train complex novel behaviors with about an hour of human time. These behaviors and environments are considerably more complex than any that have been previously learned from human feedback."
                },
                {
                    "title": "Averaged-DQN: Variance Reduction and Stabilization for Deep Reinforcement Learning",
                    "abstract": "Instability and variability of Deep Reinforcement Learning (DRL) algorithms tend to adversely affect their performance. Averaged-DQN is a simple extension to the DQN algorithm, based on averaging previously learned Q-values estimates, which leads to a more stable training procedure and improved performance by reducing approximation error variance in the target values. To understand the effect of the algorithm, we examine the source of value function estimation errors and provide an analytical comparison within a simplified model. We further present experiments on the Arcade Learning Environment benchmark that demonstrate significantly improved stability and performance due to the proposed extension."
                },
                {
                    "title": "Reinforcement Learning with Unsupervised Auxiliary Tasks",
                    "abstract": "Deep reinforcement learning agents have achieved state-of-the-art results by directly maximising cumulative reward. However, environments contain a much wider variety of possible training signals. In this paper, we introduce an agent that also maximises many other pseudo-reward functions simultaneously by reinforcement learning. All of these tasks share a common representation that, like unsupervised learning, continues to develop in the absence of extrinsic rewards. We also introduce a novel mechanism for focusing this representation upon extrinsic rewards, so that learning can rapidly adapt to the most relevant aspects of the actual task. Our agent significantly outperforms the previous state-of-the-art on Atari, averaging 880\\% expert human performance, and a challenging suite of first-person, three-dimensional \\emph{Labyrinth} tasks leading to a mean speedup in learning of 10$\\times$ and averaging 87\\% expert human performance on Labyrinth."
                },
                {
                    "title": "Successor Features for Transfer in Reinforcement Learning",
                    "abstract": "Transfer in reinforcement learning refers to the notion that generalization should occur not only within a task but also across tasks. We propose a transfer framework for the scenario where the reward function changes between tasks but the environment's dynamics remain the same. Our approach rests on two key ideas: \"successor features\", a value function representation that decouples the dynamics of the environment from the rewards, and \"generalized policy improvement\", a generalization of dynamic programming's policy improvement operation that considers a set of policies rather than a single one. Put together, the two ideas lead to an approach that integrates seamlessly within the reinforcement learning framework and allows the free exchange of information across tasks. The proposed method also provides performance guarantees for the transferred policy even before any learning has taken place. We derive two theorems that set our approach in firm theoretical ground and present experiments that show that it successfully promotes transfer in practice, significantly outperforming alternative methods in a sequence of navigation tasks and in the control of a simulated robotic arm."
                },
                {
                    "title": "Continuous control with deep reinforcement learning",
                    "abstract": "We adapt the ideas underlying the success of Deep Q-Learning to the continuous action domain. We present an actor-critic, model-free algorithm based on the deterministic policy gradient that can operate over continuous action spaces. Using the same learning algorithm, network architecture and hyper-parameters, our algorithm robustly solves more than 20 simulated physics tasks, including classic problems such as cartpole swing-up, dexterous manipulation, legged locomotion and car driving. Our algorithm is able to find policies whose performance is competitive with those found by a planning algorithm with full access to the dynamics of the domain and its derivatives. We further demonstrate that for many of the tasks the algorithm can learn policies end-to-end: directly from raw pixel inputs."
                },
                {
                    "title": "Deterministic Policy Gradient Algorithms",
                    "abstract": "In this paper we consider deterministic policy gradient algorithms for reinforcement learning with continuous actions. The deterministic policy gradient has a particularly appealing form: it is the expected gradient of the action-value function. This simple form means that the deterministic policy gradient can be estimated much more efficiently than the usual stochastic policy gradient. To ensure adequate exploration, we introduce an off-policy actor-critic algorithm that learns a deterministic target policy from an exploratory behaviour policy. We demonstrate that deterministic policy gradient algorithms can significantly outperform their stochastic counterparts in high-dimensional action spaces."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we adaptively control the representation rank of neural networks in deep reinforcement learning to improve generalization and robustness without overfitting?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a fundamental limitation in deep reinforcement learning (DRL) where neural networks are often treated as black boxes. By understanding and controlling the representation rank, we can enhance the performance of DRL agents, leading to more efficient learning and better generalization to new environments. This advancement could pave the way for practical applications in various fields, such as robotics and autonomous systems, where reliable decision-making is essential. Furthermore, it could inspire future research to explore the interplay between neural network properties and reinforcement learning dynamics, potentially leading to novel architectures and training methodologies.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complex relationship between representation rank, model complexity, and generalization. Naive approaches that focus solely on maximizing representation rank can lead to overfitting, making the model less robust and efficient in sampling. Additionally, empirically fine-tuning the balance of representation rank is difficult, as it requires navigating the trade-off between model complexity and the ability to learn optimal policies. The technical obstacles include deriving effective constraints from the Bellman equation and ensuring that the neural network dynamics do not violate these constraints, which is complicated by feature co-adaptation phenomena in DRL.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on maximizing representation rank without considering the implications of overfitting and generalization. There has been a lack of understanding of how to derive adaptive control mechanisms for representation rank from foundational principles like the Bellman equation. Barriers include insufficient exploration of the relationship between cosine similarity and representation rank, as well as the complexities introduced by neural network dynamics. Our approach differs by introducing the BEllman Equation-based automatic rank Regularizer (BEER), which provides a systematic way to control representation rank based on derived constraints, addressing the limitations of prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the introduction of the BEER regularizer, which controls the representation rank by regularizing the cosine similarity between adjacent state-action representations. We will evaluate this approach using standard DRL benchmarks and datasets, measuring performance through metrics such as sample efficiency"
            }
        },
        "author_data": {
            "c9545b73-d115-4fea-a8a7-3736303666e0": {
                "pk": "c9545b73-d115-4fea-a8a7-3736303666e0",
                "name": "Qiang He",
                "collaborators": [
                    "Xinwen Hou",
                    "Chen Gong",
                    "Yu Liu",
                    "Yunpeng Bai",
                    "Guoliang Fan",
                    "Xiaoyu Chen",
                    "Chen Zhang",
                    "Zhou Yuan",
                    "Elvis S. Liu",
                    "Hong Wang"
                ],
                "domain": [
                    "Deep Reinforcement Learning",
                    "Ensemble Learning",
                    "Policy Optimization",
                    "Game AI"
                ],
                "publications": [
                    {
                        "title": "Advancing DRL Agents in Commercial Fighting Games: Training, Integration, and Agent-Human Alignment",
                        "abstract": "Deep Reinforcement Learning (DRL) agents have demonstrated impressive success in a wide range of game genres. However, existing research primarily focuses on optimizing DRL competence rather than addressing the challenge of prolonged player interaction. In this paper, we propose a practical DRL agent system for fighting games named Sh\\=ukai, which has been successfully deployed to Naruto Mobile, a popular fighting game with over 100 million registered users. Sh\\=ukai quantifies the state to enhance generalizability, introducing Heterogeneous League Training (HELT) to achieve balanced competence, generalizability, and training efficiency. Furthermore, Sh\\=ukai implements specific rewards to align the agent's behavior with human expectations. Sh\\=ukai's ability to generalize is demonstrated by its consistent competence across all characters, even though it was trained on only 13% of them. Additionally, HELT exhibits a remarkable 22% improvement in sample efficiency. Sh\\=ukai serves as a valuable training partner for players in Naruto Mobile, enabling them to enhance their abilities and skills."
                    },
                    {
                        "title": "MEPE: A Minimalist Ensemble Policy Evaluation Operator for Deep Reinforcement Learning",
                        "abstract": "Ensemble deep reinforcement learning (DRL) is a popular approach to mitigate the risk of overspecialization to a particular distribution and improve performance. However, it is commonly plagued by high computational resource requirements that arise from the introduction of multiple value and policy functions. To avoid the notorious resource consumption issue, we design a simple yet effective ensemble policy evaluation operator, termed Minimalist Ensemble Policy Evaluation (MEPE), which leverages a modified dropout operator combined with the Bellman equation. MEPE is equivalent to integrating multiple models into a single model. The MEPE operator holds ensemble property by keeping the dropout consistency of both sides of the Bellman equation and can be combined with any DRL algorithms if they have a policy evaluation phase. To verify the MEPE\u2019s ability, we perform experiments on both low-dimensional and high-dimensional environments, which presents that the algorithms combined with the MEPE operator outperform or achieve a similar level of performance as the current state-of-the-art ensemble methods and model-free methods without increasing additional computational resource costs. To the best of our knowledge, our work is the first that applies the dropout operator both in low-dimensional control tasks and high-dimensional games. Our code is available at https://github.com/sweetice/MEPE."
                    },
                    {
                        "title": "Eigensubspace of Temporal-Difference Dynamics and How It Improves Value Approximation in Reinforcement Learning",
                        "abstract": "We propose a novel value approximation method, namely Eigensubspace Regularized Critic (ERC) for deep reinforcement learning (RL). ERC is motivated by an analysis of the dynamics of Q-value approximation error in the Temporal-Difference (TD) method, which follows a path defined by the 1-eigensubspace of the transition kernel associated with the Markov Decision Process (MDP). It reveals a fundamental property of TD learning that has remained unused in previous deep RL approaches. In ERC, we propose a regularizer that guides the approximation error tending towards the 1-eigensubspace, resulting in a more efficient and stable path of value approximation. Moreover, we theoretically prove the convergence of the ERC method. Besides, theoretical analysis and experiments demonstrate that ERC effectively reduces the variance of value functions. Among 26 tasks in the DMControl benchmark, ERC outperforms state-of-the-art methods for 20. Besides, it shows significant advantages in Q-value approximation and variance reduction. Our code is available at https://sites.google.com/view/erc-ecml23/."
                    },
                    {
                        "title": "Keep Various Trajectories: Promoting Exploration of Ensemble Policies in Continuous Control",
                        "abstract": "The combination of deep reinforcement learning (DRL) with ensemble methods has been proved to be highly effective in addressing complex sequential decision-making problems. This success can be primarily attributed to the utilization of multiple models, which enhances both the robustness of the policy and the accuracy of value function estimation. However, there has been limited analysis of the empirical success of current ensemble RL methods thus far. Our new analysis reveals that the sample efficiency of previous ensemble DRL algorithms may be limited by sub-policies that are not as diverse as they could be. Motivated by these findings, our study introduces a new ensemble RL algorithm, termed \\textbf{T}rajectories-awar\\textbf{E} \\textbf{E}nsemble exploratio\\textbf{N} (TEEN). The primary goal of TEEN is to maximize the expected return while promoting more diverse trajectories. Through extensive experiments, we demonstrate that TEEN not only enhances the sample diversity of the ensemble policy compared to using sub-policies alone but also improves the performance over ensemble RL algorithms. On average, TEEN outperforms the baseline ensemble DRL algorithms by 41\\% in performance on the tested representative environments."
                    },
                    {
                        "title": "Frustratingly Easy Regularization on Representation Can Boost Deep Reinforcement Learning",
                        "abstract": "Deep reinforcement learning (DRL) gives the promise that an agent learns good policy from high-dimensional information, whereas representation learning removes irrelevant and redundant information and retains pertinent information. In this work, we demonstrate that the learned representation of the Q-network and its target Q-network should, in theory, satisfy a favorable distinguishable representation property. Specifically, there exists an upper bound on the representation similarity of the value functions of two adjacent time steps in a typical DRL setting. However, through illustrative experiments, we show that the learned DRL agent may violate this property and lead to a sub-optimal policy. Therefore, we propose a simple yet effective regularizer called P_olicy E_valuation with E_asy Regularization on Representation (PEER), which aims to maintain the distinguishable representation property via explicit regularization on internal representations. And we provide the convergence rate guarantee of PEER. Implementing PEER requires only one line of code. Our experiments demonstrate that incorporating PEER into DRL can significantly improve performance and sample efficiency. Comprehensive experiments show that PEER achieves state-of-the-art performance on all 4 environments on PyBullet, 9 out of 12 tasks on DM-Control, and 19 out of 26 games on Atari. To the best of our knowledge, PEER is the first work to study the inherent representation property of Q-network and its target. Our code is available at https://sites.google.com/view/peer-cvpr2023/."
                    },
                    {
                        "title": "Wide-Sense Stationary Policy Optimization with Bellman Residual on Video Games",
                        "abstract": "Deep Reinforcement Learning (DRL) has an increasing application in video games. However, it usually suffers from unstable training, low sampling efficiency, etc. Under the assumption that Bellman residual follows a stationary random process when the training process is convergent, we propose the Wide-sense Stationary Policy Optimization (WSPO) framework, which leverages the Wasserstein distance from the Bellman Residual Distribution (BRD) between two adjacent time steps, to stabilize the training stage and improve the sampling efficiency. We minimize the Wasserstein distance with Quantile Regression, where the specific form of BRD is not needed. Finally, we combine WSPO with Advantage Actor-Critic (A2C) algorithm and Deep Deterministic Policy Gradient (DDPG) algorithm. We evaluate WSPO on Atari 2600 video games and continuous control tasks, illustrating that WSPO compares or outperforms the state-of-the-art algorithms we tested."
                    },
                    {
                        "title": "MEPG: A Minimalist Ensemble Policy Gradient Framework for Deep Reinforcement Learning",
                        "abstract": "During the training of a reinforcement learning (RL) agent, the distribution of training data is non-stationary as the agent's behavior changes over time. Therefore, there is a risk that the agent is overspecialized to a particular distribution and its performance suffers in the larger picture. Ensemble RL can mitigate this issue by learning a robust policy. However, it suffers from heavy computational resource consumption due to the newly introduced value and policy functions. In this paper, to avoid the notorious resources consumption issue, we design a novel and simple ensemble deep RL framework that integrates multiple models into a single model. Specifically, we propose the \\underline{M}inimalist \\underline{E}nsemble \\underline{P}olicy \\underline{G}radient framework (MEPG), which introduces minimalist ensemble consistent Bellman update by utilizing a modified dropout operator. MEPG holds ensemble property by keeping the dropout consistency of both sides of the Bellman equation. Additionally, the dropout operator also increases MEPG's generalization capability. Moreover, we theoretically show that the policy evaluation phase in the MEPG maintains two synchronized deep Gaussian Processes. To verify the MEPG framework's ability to generalize, we perform experiments on the gym simulator, which presents that the MEPG framework outperforms or achieves a similar level of performance as the current state-of-the-art ensemble methods and model-free methods without increasing additional computational resource costs."
                    },
                    {
                        "title": "The $f$-Divergence Reinforcement Learning Framework",
                        "abstract": "The framework of deep reinforcement learning (DRL) provides a powerful and widely applicable mathematical formalization for sequential decision-making. This paper present a novel DRL framework, termed \\emph{$f$-Divergence Reinforcement Learning (FRL)}. In FRL, the policy evaluation and policy improvement phases are simultaneously performed by minimizing the $f$-divergence between the learning policy and sampling policy, which is distinct from conventional DRL algorithms that aim to maximize the expected cumulative rewards. We theoretically prove that minimizing such $f$-divergence can make the learning policy converge to the optimal policy. Besides, we convert the process of training agents in FRL framework to a saddle-point optimization problem with a specific $f$ function through Fenchel conjugate, which forms new methods for policy evaluation and policy improvement. Through mathematical proofs and empirical evaluation, we demonstrate that the FRL framework has two advantages: (1) policy evaluation and policy improvement processes are performed simultaneously and (2) the issues of overestimating value function are naturally alleviated. To evaluate the effectiveness of the FRL framework, we conduct experiments on Atari 2600 video games and show that agents trained in the FRL framework match or surpass the baseline DRL algorithms."
                    },
                    {
                        "title": "WD3: Taming the Estimation Bias in Deep Reinforcement Learning",
                        "abstract": "The overestimation phenomenon caused by function approximation is a well-known issue in value-based reinforcement learning algorithms such as deep Q-networks and DDPG, which could lead to suboptimal policies. To address this issue, TD3 takes the minimum value between a pair of critics, which introduces underestimation bias. By unifying these two opposites, we propose a novel Weighted Delayed Deep Deterministic Policy Gradient algorithm, which can reduce the estimation error and further improve the performance by weighting a pair of critics. We compare the learning process of value function between DDPG, TD3, and our proposed algorithm, which verifies that our algorithm could indeed eliminate the estimation error of value function. We evaluate our algorithm in the OpenAI Gym continuous control tasks, outperforming the state-of-the-art algorithms on every environment tested."
                    },
                    {
                        "title": "POPO: Pessimistic Offline Policy Optimization",
                        "abstract": "Offline reinforcement learning (RL) aims to optimize policy from large pre-recorded datasets without interaction with the environment. This setting offers the promise of utilizing diverse and static datasets to obtain policies without costly, risky, active exploration. However, commonly used off-policy deep RL methods perform poorly when facing arbitrary off-policy datasets. In this work, we show that there exists an estimation gap of value-based deep RL algorithms in the offline setting. To eliminate the estimation gap, we propose a novel offline RL algorithm that we term Pessimistic Offline Policy Optimization (POPO), which learns a pessimistic value function. To demonstrate the effectiveness of POPO, we perform experiments on various quality datasets. And we find that POPO performs surprisingly well and scales to tasks with high-dimensional state and action space, comparing or outperforming tested state-of-the-art offline RL algorithms on benchmark tasks."
                    },
                    {
                        "title": "Reducing Estimation Bias via Weighted Delayed Deep Deterministic Policy Gradient",
                        "abstract": "The overestimation phenomenon caused by function approximation is a well-known issue in value-based reinforcement learning algorithms such as deep Q-networks and DDPG, which could lead to suboptimal policies. To address this issue, TD3 takes the minimum value between a pair of critics, which introduces underestimation bias. By unifying these two opposites, we propose a novel Weighted Delayed Deep Deterministic Policy Gradient algorithm, which can reduce the estimation error and further improve the performance by weighting a pair of critics. We compare the learning process of value function between DDPG, TD3, and our proposed algorithm, which verifies that our algorithm could indeed eliminate the estimation error of value function. We evaluate our algorithm in the OpenAI Gym continuous control tasks, outperforming the state-of-the-art algorithms on every environment tested."
                    }
                ]
            },
            "c5ed6359-6ca5-4d9d-9022-4eeb3dcd978a": {
                "pk": "c5ed6359-6ca5-4d9d-9022-4eeb3dcd978a",
                "name": "Tianyi Zhou",
                "collaborators": [
                    "Yucheng Yang",
                    "Lei Han",
                    "Mykola Pechenizkiy",
                    "Meng Fang",
                    "Qiang He"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Unsupervised Learning",
                    "Curriculum Learning"
                ],
                "publications": [
                    {
                        "title": "Task Adaptation from Skills: Information Geometry, Disentanglement, and New Objectives for Unsupervised Reinforcement Learning",
                        "abstract": "."
                    },
                    {
                        "title": "Automatic Curriculum for Unsupervised Reinforcement Learning",
                        "abstract": "Unsupervised reinforcement learning (URL) relies on carefully designed training objectives rather than task rewards to learn general skills. However, we lack quantitative evaluation metrics for URL but mainly rely on visualizations of trajectories for comparison. Moreover, most URL methods choose to optimize a single training objective, which may hinder later-stage learning and the development of new skills. To bridge these gaps, we first introduce a combination of metrics that can evaluate diverse properties of URL. We show that balancing these metrics in URL leads to better performance and trajectories with empirical evidence and theoretical insights. Next, we develop an automatic curriculum that uses a non-stationary multi-armed bandit algorithm to select URL objectives for different learning episodes, resulting in a balanced improvement on all the metrics. Extensive experiments in different environments demonstrate the advantages of our method in achieving promising and balanced performance on multiple metrics when compared to recent URL methods."
                    }
                ]
            },
            "5e3143e4-edd1-47a9-96ff-9146cac06f51": {
                "pk": "5e3143e4-edd1-47a9-96ff-9146cac06f51",
                "name": "Meng Fang",
                "collaborators": [
                    "Yucheng Yang",
                    "Tianyi Zhou",
                    "Qiang He",
                    "Lei Han",
                    "Mykola Pechenizkiy"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ],
                "publications": [
                    {
                        "title": "Task Adaptation from Skills: Information Geometry, Disentanglement, and New Objectives for Unsupervised Reinforcement Learning",
                        "abstract": "."
                    }
                ]
            },
            "79553e44-e963-4f03-919c-5aa8b6a1a391": {
                "pk": "79553e44-e963-4f03-919c-5aa8b6a1a391",
                "name": "Setareh Maghsudi",
                "collaborators": [
                    "Xiaotong Cheng",
                    "Saeed Ghoorchian",
                    "Evgenii Kortukov",
                    "Steven Bilaj",
                    "Manuel Eberhardinger",
                    "Johannes Maucher",
                    "S\u0142awomir Sta\u0144czak",
                    "Behzad Nourani-Koliji",
                    "A. R. Balef",
                    "Ioannis Tsetis"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Bandit Algorithms",
                    "Multi-Agent Systems",
                    "Resource Management"
                ],
                "publications": [
                    {
                        "title": "Distributed Management of Fluctuating Energy Resources in Dynamic Networked Systems",
                        "abstract": "Modern power systems integrate renewable distributed energy resources (DERs) as an environment-friendly enhancement to meet the ever-increasing demands. However, the inherent unreliability of renewable energy renders developing DER management algorithms imperative. We study the energy-sharing problem in a system consisting of several DERs. Each agent harvests and distributes renewable energy in its neighborhood to optimize the network's performance while minimizing energy waste. We model this problem as a bandit convex optimization problem with constraints that correspond to each node's limitations for energy production. We propose distributed decision-making policies to solve the formulated problem, where we utilize the notion of dynamic regret as the performance metric. We also include an adjustment strategy in our developed algorithm to reduce the constraint violations. Besides, we design a policy that deals with the non-stationary environment. Theoretical analysis shows the effectiveness of our proposed algorithm. Numerical experiments using a real-world dataset show superior performance of our proposal compared to state-of-the-art methods."
                    },
                    {
                        "title": "Non-Stationary Linear Bandits With Dimensionality Reduction for Large-Scale Recommender Systems",
                        "abstract": "Taking advantage ofcontextual information can potentially boost the performance of recommender systems. In the era of Big Data, such side information often has several dimensions. Thus, developing decision-making algorithms to cope with such a high-dimensional context in real time is essential. That is specifically challenging when the decision-maker has a variety of items to recommend. In addition, changes in items' popularity or users' preferences can hinder the performance of the deployed recommender system due to a lack of robustness to distribution shifts in the environment. In this paper, we build upon the linear contextual multi-armed bandit framework to address this problem. We develop a decision-making policy for a linear bandit problem with high-dimensional feature vectors, a large set of arms, and non-stationary reward-generating processes. Our Thompson sampling-based policy reduces the dimension of feature vectors using random projection and uses exponentially increasing weights to decrease the influence of past observations with time. Our proposed recommender system employs this policy to learn the users' item preferences online while minimizing runtime. We prove a regret bound that scales as a factor of the reduced dimension instead of the original one. To evaluate our proposed recommender system numerically, we apply it to three real-world datasets. The theoretical and numerical results demonstrate the effectiveness of our proposed algorithm in making a trade-off between computational complexity and regret performance compared to the state-of-the-art."
                    },
                    {
                        "title": "Meta Learning in Bandits within Shared Affine Subspaces",
                        "abstract": "We study the problem of meta-learning several contextual stochastic bandits tasks by leveraging their concentration around a low-dimensional affine subspace, which we learn via online principal component analysis to reduce the expected regret over the encountered bandits. We propose and theoretically analyze two strategies that solve the problem: One based on the principle of optimism in the face of uncertainty and the other via Thompson sampling. Our framework is generic and includes previously proposed approaches as special cases. Besides, the empirical results show that our methods significantly reduce the regret on several bandit tasks."
                    },
                    {
                        "title": "Unveiling the Decision-Making Process in Reinforcement Learning with Genetic Programming",
                        "abstract": "Despite tremendous progress, machine learning and deep learning still suffer from incomprehensible predictions. Incomprehensibility, however, is not an option for the use of (deep) reinforcement learning in the real world, as unpredictable actions can seriously harm the involved individuals. In this work, we propose a genetic programming framework to generate explanations for the decision-making process of already trained agents by imitating them with programs. Programs are interpretable and can be executed to generate explanations of why the agent chooses a particular action. Furthermore, we conduct an ablation study that investigates how extending the domain-specific language by using library learning alters the performance of the method. We compare our results with the previous state of the art for this problem and show that we are comparable in performance but require much less hardware resources and computation time."
                    },
                    {
                        "title": "Gradual Change Detection in Covariance Matrix: A Lazy Approach",
                        "abstract": "Thanks to its slow-varying characteristic and relatively low requirement for estimation overhead, the covariance matrix has been extensively researched in sixth-generation (6G) wireless systems. Nevertheless, user mobility in practice will cause a gradual change in the covariance matrix, thereby deteriorating the system's performance if no update of the covariance matrix is applied. In this paper, we study the problem of efficient detection of gradual changes in the covariance matrix. We first introduce four change-point detectors that directly map the observations to change in our target KPI. Then, we propose a low-overhead detection algorithm that omits unnecessary channel estimations by adapting an AoA-based estimation trigger. Simulation results show that our proposed scheme can provide near-optimal performance while drastically reducing the estimation and computation overhead."
                    },
                    {
                        "title": "Online Influence Maximization with Semi-Bandit Feedback under Corruptions",
                        "abstract": "In this work, we investigate the online influence maximization in social networks. Most prior research studies on online influence maximization assume that the nodes are fully cooperative and act according to their stochastically generated influence probabilities on others. In contrast, we study the online influence maximization problem in the presence of some corrupted nodes whose damaging effects diffuse throughout the network. We propose a novel bandit algorithm, CW-IMLinUCB, which robustly learns and finds the optimal seed set in the presence of corrupted users. Theoretical analyses establish that the regret performance of our proposed algorithm is better than the state-of-the-art online influence maximization algorithms. Extensive empirical evaluations on synthetic and real-world datasets also show the superior performance of our proposed algorithm."
                    },
                    {
                        "title": "A Repeated Auction Model for Load-Aware Dynamic Resource Allocation in Multi-Access Edge Computing",
                        "abstract": "Multi-access edge computing (MEC) is one of the enabling technologies for high-performance computing at the edge of the 6 G networks, supporting high data rates and ultra-low service latency. Although MEC is a remedy to meet the growing demand for computation-intensive applications, the scarcity of resources at the MEC servers degrades its performance. Hence, effective resource management is essential; nevertheless, state-of-the-art research lacks efficient economic models to support the exponential growth of the MEC-enabled applications market. We focus on designing a MEC offloading service market based on a repeated auction model with multiple resource sellers (e.g., network operators and service providers) that compete to sell their computing resources to the offloading users. We design a computationally-efficient modified Generalized Second Price (GSP)-based algorithm that decides on pricing and resource allocation by considering the dynamic offloading requests arrival and the servers\u2019 computational workloads. Besides, we propose adaptive best-response bidding strategies for the resource sellers, satisfying the symmetric Nash equilibrium (SNE) and individual rationality properties. Finally, via intensive numerical results1, we show the effectiveness of our proposed resource allocation mechanism."
                    },
                    {
                        "title": "Age-Based Federated Learning Approach to In-Network Caching: An Online Scheduling Policy",
                        "abstract": "We develop an accurate real-time scheduling framework for federated learning (FL) in wireless caching networks to guarantee the successful delivery of files at a low cost and with a short delay. The following persisting challenges motivated our work: i) Enforcing an excessive number of FL model update per communication round is infeasible due to the limited backhaul spectrum; ii) Naive scheduling policy accounting for FL model update renders service backlogs, thus leading to network parameter staleness and in-network caching utility (ICU) deterioration. Optimal scheduling in FL is challenging, as the mobile users' preferences for content, request patterns, and network traffic are dynamic and unknown. To tackle that challenge, we first formulate an instantaneous ICU optimization problem against the stale FL models. Afterward, based on the concept of age-of-update (AoU), we propose a federated learning with an unsatisfactory set selection (FedUSS) approach capable of executing the multiple-tasks of short-term predictions and making cache replacement decisions at low cost. Theoretical and numerical analyses manifest the effectiveness of our approach."
                    },
                    {
                        "title": "Decentralized Task Offloading and Load-Balancing for Mobile Edge Computing in Dense Networks",
                        "abstract": "We study the problem of decentralized task offloading and load-balancing in a dense network with numerous devices and a set of edge servers. Solving this problem optimally is complicated due to the unknown network information and random task sizes. The shared network resources also influence the users\u2019 decisions and resource distribution. Our solution combines the mean field multi-agent multi-armed bandit (MAB) game with a load-balancing technique that adjusts the servers\u2019 rewards to achieve a target population profile despite the distributed user decision-making. Numerical results demonstrate the efficacy of our approach and the convergence to the target load distribution."
                    },
                    {
                        "title": "Contextual Multi-Armed Bandit With Costly Feature Observation in Non-Stationary Environments",
                        "abstract": "Maximizing long-term rewards is the primary goal in sequential decision-making problems. The majority of existing methods assume that side information is freely available, enabling the learning agent to observe all features' states before making a decision. In real-world problems, however, collecting beneficial information is often costly. That implies that, besides individual arms' reward, learning the observations of the features' states is essential to improve the decision-making strategy. The problem is aggravated in a non-stationary environment where reward and cost distributions undergo abrupt changes over time. To address the aforementioned dual learning problem, we extend the contextual bandit setting and allow the agent to observe subsets of features' states. The objective is to maximize the long-term average gain, which is the difference between the accumulated rewards and the paid costs on average. Therefore, the agent faces a trade-off between minimizing the cost of information acquisition and possibly improving the decision-making process using the obtained information. To this end, we develop an algorithm that guarantees a sublinear regret in time. Numerical results demonstrate the superiority of our proposed policy in a real-world scenario."
                    },
                    {
                        "title": "Multi-Agent Learning from Learners",
                        "abstract": "A large body of the \u201cInverse Reinforcement Learning\u201d (IRL) literature focuses on recovering the reward function from a set of demonstrations of an expert agent who acts optimally or noisily optimally. Nevertheless, some recent works move away from the optimality assumption to study the \u201cLearning from a Learner (LfL)\u201d problem, where the challenge is inferring the reward function of a learning agent from a sequence of demonstrations produced by progressively improving policies. In this work, we take one of the initial steps in addressing the multi-agent version of this problem and propose a new algorithm, MA-LfL (Multi-agent Learning from a Learner). Unlike the state-of-the-art literature, which recovers the reward functions from trajectories produced by agents in some equilibrium, we study the problem of inferring the reward functions of interacting agents in a general sum stochastic game without assuming any equilibrium state. The MA-LfL algorithm is rigorously built on a theoretical result that ensures its validity in the case of agents learning according to a multi-agent soft policy iteration scheme. We empirically test MA-LfL and we observe high positive correlation between the recovered reward functions and the ground truth."
                    },
                    {
                        "title": "Piecewise-Stationary Multi-Objective Multi-Armed Bandit With Application to Joint Communications and Sensing",
                        "abstract": "We study a multi-objective multi-armed bandit problem in a dynamic environment. The problem portrays a decision-maker that sequentially selects an arm from a given set. If selected, each action produces a reward vector, where every element follows a piecewise-stationary Bernoulli distribution. The agent aims at choosing an arm among the Pareto optimal set of arms to minimize its regret. We propose a Pareto generic upper confidence bound (UCB)-based algorithm with change detection to solve this problem. By developing the essential inequalities for multi-dimensional spaces, we establish that our proposal guarantees a regret bound in the order of <inline-formula> <tex-math notation=\"LaTeX\">$\\gamma _{T}\\log (T/{\\gamma _{T}})$ </tex-math></inline-formula> when the number of breakpoints <inline-formula> <tex-math notation=\"LaTeX\">$\\gamma _{T}$ </tex-math></inline-formula> is known. Without this assumption, the regret bound of our algorithm is <inline-formula> <tex-math notation=\"LaTeX\">$\\gamma _{T}\\log (T)$ </tex-math></inline-formula>. Finally, we formulate an energy-efficient waveform design problem in an integrated communication and sensing system as a toy example. Numerical experiments on the toy example and synthetic and real-world datasets demonstrate the efficiency of our policy compared to the current methods."
                    },
                    {
                        "title": "Piecewise-Stationary Combinatorial Semi-Bandit with Causally Related Rewards",
                        "abstract": "We study the piecewise stationary combinatorial semi-bandit problem with causally related rewards. In our nonstationary environment, variations in the base arms' distributions, causal relationships between rewards, or both, change the reward generation process. In such an environment, an optimal decision-maker must follow both sources of change and adapt accordingly. The problem becomes aggravated in the combinatorial semi-bandit setting, where the decision-maker only observes the outcome of the selected bundle of arms. The core of our proposed policy is the Upper Confidence Bound (UCB) algorithm. We assume the agent relies on an adaptive approach to overcome the challenge. More specifically, it employs a change-point detector based on the Generalized Likelihood Ratio (GLR) test. Besides, we introduce the notion of group restart as a new alternative restarting strategy in the decision making process in structured environments. Finally, our algorithm integrates a mechanism to trace the variations of the underlying graph structure, which captures the causal relationships between the rewards in the bandit setting. Theoretically, we establish a regret upper bound that reflects the effects of the number of structural- and distribution changes on the performance. The outcome of our numerical experiments in real-world scenarios exhibits applicability and superior performance of our proposal compared to the state-of-the-art benchmarks."
                    },
                    {
                        "title": "Eigensubspace of Temporal-Difference Dynamics and How It Improves Value Approximation in Reinforcement Learning",
                        "abstract": "We propose a novel value approximation method, namely Eigensubspace Regularized Critic (ERC) for deep reinforcement learning (RL). ERC is motivated by an analysis of the dynamics of Q-value approximation error in the Temporal-Difference (TD) method, which follows a path defined by the 1-eigensubspace of the transition kernel associated with the Markov Decision Process (MDP). It reveals a fundamental property of TD learning that has remained unused in previous deep RL approaches. In ERC, we propose a regularizer that guides the approximation error tending towards the 1-eigensubspace, resulting in a more efficient and stable path of value approximation. Moreover, we theoretically prove the convergence of the ERC method. Besides, theoretical analysis and experiments demonstrate that ERC effectively reduces the variance of value functions. Among 26 tasks in the DMControl benchmark, ERC outperforms state-of-the-art methods for 20. Besides, it shows significant advantages in Q-value approximation and variance reduction. Our code is available at https://sites.google.com/view/erc-ecml23/."
                    },
                    {
                        "title": "Learning of Generalizable and Interpretable Knowledge in Grid-Based Reinforcement Learning Environments",
                        "abstract": "Understanding the interactions of agents trained with deep reinforcement learning is crucial for deploying agents in games or the real world. In the former, unreasonable actions confuse players. In the latter, that effect is even more significant, as unexpected behavior cause accidents with potentially grave and long-lasting consequences for the involved individuals. In this work, we propose using program synthesis to imitate reinforcement learning policies after seeing a trajectory of the action sequence. Programs have the advantage that they are inherently interpretable and verifiable for correctness. We adapt the state-of-the-art program synthesis system DreamCoder for learning concepts in grid-based environments, specifically, a navigation task and two miniature versions of Atari games, Space Invaders and Asterix. By inspecting the generated libraries, we can make inferences about the concepts the black-box agent has learned and better understand the agent's behavior. We achieve the same by visualizing the agent's decision-making process for the imitated sequences. We evaluate our approach with different types of program synthesizers based on a search-only method, a neural-guided search, and a language model fine-tuned on code."
                    },
                    {
                        "title": "Distributed Consensus Algorithm for Decision-Making in Multi-agent Multi-armed Bandit",
                        "abstract": "We study a structured multi-agent multi-armed bandit (MAMAB) problem in a dynamic environment. A graph reflects the information-sharing structure among agents, and the arms' reward distributions are piecewise-stationary with several unknown change points. The agents face the identical piecewise-stationary MAB problem. The goal is to develop a decision-making policy for the agents that minimizes the regret, which is the expected total loss of not playing the optimal arm at each time step. Our proposed solution, Restarted Bayesian Online Change Point Detection in Cooperative Upper Confidence Bound Algorithm (RBO-Coop-UCB), involves an efficient multi-agent UCB algorithm as its core enhanced with a Bayesian change point detector. We also develop a simple restart decision cooperation that improves decision-making. Theoretically, we establish that the expected group regret of RBO-Coop-UCB is upper bounded by $\\mathcal{O}(KNM\\log T + K\\sqrt{MT\\log T})$, where K is the number of agents, M is the number of arms, and T is the number of time steps. Numerical experiments on synthetic and real-world datasets demonstrate that our proposed method outperforms the state-of-the-art algorithms."
                    },
                    {
                        "title": "Online Learning with Costly Features in Non-stationary Environments",
                        "abstract": "Maximizing long-term rewards is the primary goal in sequential decision-making problems. The majority of existing methods assume that side information is freely available, enabling the learning agent to observe all features' states before making a decision. In real-world problems, however, collecting beneficial information is often costly. That implies that, besides individual arms' reward, learning the observations of the features' states is essential to improve the decision-making strategy. The problem is aggravated in a non-stationary environment where reward and cost distributions undergo abrupt changes over time. To address the aforementioned dual learning problem, we extend the contextual bandit setting and allow the agent to observe subsets of features' states. The objective is to maximize the long-term average gain, which is the difference between the accumulated rewards and the paid costs on average. Therefore, the agent faces a trade-off between minimizing the cost of information acquisition and possibly improving the decision-making process using the obtained information. To this end, we develop an algorithm that guarantees a sublinear regret in time. Numerical results demonstrate the superiority of our proposed policy in a real-world scenario."
                    },
                    {
                        "title": "A Deep Reinforcement Learning Approach for Load Balancing in Open Radio Access Networks",
                        "abstract": "The Open RAN paradigm offers data-driven, intelligent optimization of the radio access network (RAN). The disaggregated nature of the Open RAN combined with virtualization on general-purpose CPUs with limited computation capacity creates different load types at multiple levels, making it more challenging to balance the load within the network. This paper proposes a learning framework that learns the assignment of users (UEs) to network nodes to balance the communication and computation load in the network. The framework incorporates communication resources consumed by the users in the radio unit (RU), and computation resources needed for baseband processing in the virtualized distributed unit (DU). The goal is thus to balance the communication load between RUs and the computation load between DUs to avoid overloading network elements or to handle higher peak data rate demands when new users arrive in the network. We apply a novel utility-based approach to jointly optimize the UE-RU and RU-DU assignments taking into account the users' QoS (quality of service) requirements. Simulations demonstrate that the proposed method generates the assignments that significantly improve the network load conditions compared to baseline schemes, thereby enabling more available communication and computation resources for incoming peak data rate users in the network."
                    }
                ]
            }
        }
    },
    "2309.05196": {
        "paper_data": {
            "title": "Does Writing with Language Models Reduce Content Diversity?",
            "url": "http://arxiv.org/abs/2309.05196v3",
            "arxiv_id": "2309.05196",
            "authors": [
                "Vishakh Padmakumar",
                "He He"
            ],
            "abstract": "Large language models (LLMs) have led to a surge in collaborative writing with model assistance. As different users incorporate suggestions from the same model, there is a risk of decreased diversity in the produced content, potentially limiting diverse perspectives in public discourse. In this work, we measure the impact of co-writing on diversity via a controlled experiment, where users write argumentative essays in three setups -- using a base LLM (GPT3), a feedback-tuned LLM (InstructGPT), and writing without model help. We develop a set of diversity metrics and find that writing with InstructGPT (but not the GPT3) results in a statistically significant reduction in diversity. Specifically, it increases the similarity between the writings of different authors and reduces the overall lexical and content diversity. We additionally find that this effect is mainly attributable to InstructGPT contributing less diverse text to co-written essays. In contrast, the user-contributed text remains unaffected by model collaboration. This suggests that the recent improvement in generation quality from adapting models to human feedback might come at the cost of more homogeneous and less diverse content.",
            "introduction": "   1 Introduction  Large language models (LLMs) are rapidly changing how people create content (Lee et\u00a0al., 2022a; Mirowski et\u00a0al., 2023). While LLM-based writing assistants have the potential to improve writing quality and increase author productivity, they also introduce an algorithmic monoculture (Kleinberg and Raghavan, 2021). As millions of users rely on the same underlying model to produce text, there is a potential risk of shifting the content towards the mode\u2014good but not colorful writing. In this work, we aim to assess whether writing with LLMs unintentionally reduces content diversity.   Existing evidence has already hinted that LLMs may influence users\u2019 opinions. Jakesch et\u00a0al. (2023) and Bhat et\u00a0al. (2023) find that users\u2019 opinions expressed in their writing can be biased by controlling LLMs to promote a specific argument (e.g., social media is good/bad for society). Additionally, it has been shown that current LLMs do not equally represent views from various demographic groups (Santurkar et\u00a0al., 2023; Durmus et\u00a0al., 2023). These results suggest that writing with LLMs may limit the perspectives expressed in the writing. We hypothesize that incorporating model suggestions dilutes the writer\u2019s unique voice, leading to different writers producing similar content, and this homogenization in turn reduces the overall diversity of content produced by many writers.   Figure 1: We measure the content diversity of essays written by three groups of users: a control group writing without model help (Solo), a treatment group writing with a base language model (GPT3), and second treatment group writing with a feedback-tuned language model (InstructGPT). Essays co-written with InstructGPT are more similar to each other and exhibit the least lexical and content diversity, whereas those co-written with GPT3 are not significantly different to the Solo essays.     To test our hypotheses, we design a controlled experiment (Figure\u00a01) where users are asked to write an argumentative essay given a topic from the New York Times student opinion series following Lee et\u00a0al. (2022a), e.g., \u201cWhat are the most important things students learn at school?\u201d We assign participants to three groups: a control group where participants write essays without model help, an LLM treatment group where participants write essays with a base language model (GPT3), and a feedback-tuned LLM treatment group where participants write essays with a language model finetuned with human feedback (Ouyang et\u00a0al., 2022) (InstructGPT). We hired 38383838 writers from Upwork. For each group, we collected 100100100100 essays on 10101010 topics. We then develop a set of metrics and measure the effect of LLMs on content diversity at both the individual level and the collective level:   \u2022  Homogenization: Do users write more similarly to each other when writing with LLMs?   We find that essays written by the group using InstructGPT exhibit a higher degree of homogenization compared to the control group and the GPT3 group (Section\u00a04). In particular, by matching model-contributed text to the summarized key points of each essay, we find that the key points contributed by InstructGPT result in increased homogenization (Section\u00a06).    \u2022  Diversity: Does writing with LLMs reduce the diversity of content produced by a group of users?   We find that the set of essays written with InstructGPT show both lower lexical diversity and reduced diversity in the key points expressed as compared to the other two groups (Section\u00a05). This is seen in the increased repetition of phrases (higher-order n\ud835\udc5bnitalic_n-grams) introduced",
            "references": [
                {
                    "title": "Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback",
                    "abstract": "Reinforcement learning from human feedback (RLHF) is a technique for training AI systems to align with human goals. RLHF has emerged as the central method used to finetune state-of-the-art large language models (LLMs). Despite this popularity, there has been relatively little public work systematizing its flaws. In this paper, we (1) survey open problems and fundamental limitations of RLHF and related methods; (2) overview techniques to understand, improve, and complement RLHF in practice; and (3) propose auditing and disclosure standards to improve societal oversight of RLHF systems. Our work emphasizes the limitations of RLHF and highlights the importance of a multi-faceted approach to the development of safer AI systems."
                },
                {
                    "title": "Towards Measuring the Representation of Subjective Global Opinions in Language Models",
                    "abstract": "Large language models (LLMs) may not equitably represent diverse global perspectives on societal issues. In this paper, we develop a quantitative framework to evaluate whose opinions model-generated responses are more similar to. We first build a dataset, GlobalOpinionQA, comprised of questions and answers from cross-national surveys designed to capture diverse opinions on global issues across different countries. Next, we define a metric that quantifies the similarity between LLM-generated survey responses and human responses, conditioned on country. With our framework, we run three experiments on an LLM trained to be helpful, honest, and harmless with Constitutional AI. By default, LLM responses tend to be more similar to the opinions of certain populations, such as those from the USA, and some European and South American countries, highlighting the potential for biases. When we prompt the model to consider a particular country's perspective, responses shift to be more similar to the opinions of the prompted populations, but can reflect harmful cultural stereotypes. When we translate GlobalOpinionQA questions to a target language, the model's responses do not necessarily become the most similar to the opinions of speakers of those languages. We release our dataset for others to use and build on. Our data is at https://huggingface.co/datasets/Anthropic/llm_global_opinions. We also provide an interactive visualization at https://llmglobalvalues.anthropic.com."
                },
                {
                    "title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society",
                    "abstract": "Generative AI systems across modalities, ranging from text (including code), image, audio, and video, have broad social impacts, but there is no official standard for means of evaluating those impacts or for which impacts should be evaluated. In this paper, we present a guide that moves toward a standard approach in evaluating a base generative AI system for any modality in two overarching categories: what can be evaluated in a base system independent of context and what can be evaluated in a societal context. Importantly, this refers to base systems that have no predetermined application or deployment context, including a model itself, as well as system components, such as training data. Our framework for a base system defines seven categories of social impact: bias, stereotypes, and representational harms; cultural values and sensitive content; disparate performance; privacy and data protection; financial costs; environmental costs; and data and content moderation labor costs. Suggested methods for evaluation apply to listed generative modalities and analyses of the limitations of existing evaluations serve as a starting point for necessary investment in future evaluations. We offer five overarching categories for what can be evaluated in a broader societal context, each with its own subcategories: trustworthiness and autonomy; inequality, marginalization, and violence; concentration of authority; labor and creativity; and ecosystem and environment. Each subcategory includes recommendations for mitigating harm."
                },
                {
                    "title": "Rethinking Model Evaluation as Narrowing the Socio-Technical Gap",
                    "abstract": "The recent development of generative and large language models (LLMs) poses new challenges for model evaluation that the research community and industry are grappling with. While the versatile capabilities of these models ignite excitement, they also inevitably make a leap toward homogenization: powering a wide range of applications with a single, often referred to as ``general-purpose'', model. In this position paper, we argue that model evaluation practices must take on a critical task to cope with the challenges and responsibilities brought by this homogenization: providing valid assessments for whether and how much human needs in downstream use cases can be satisfied by the given model (socio-technical gap). By drawing on lessons from the social sciences, human-computer interaction (HCI), and the interdisciplinary field of explainable AI (XAI), we urge the community to develop evaluation methods based on real-world socio-requirements and embrace diverse evaluation methods with an acknowledgment of trade-offs between realism to socio-requirements and pragmatic costs to conduct the evaluation. By mapping HCI and current NLG evaluation methods, we identify opportunities for evaluation methods for LLMs to narrow the socio-technical gap and pose open questions."
                },
                {
                    "title": "Whose Opinions Do Language Models Reflect?",
                    "abstract": "Language models (LMs) are increasingly being used in open-ended contexts, where the opinions reflected by LMs in response to subjective queries can have a profound impact, both on user satisfaction, as well as shaping the views of society at large. In this work, we put forth a quantitative framework to investigate the opinions reflected by LMs -- by leveraging high-quality public opinion polls and their associated human responses. Using this framework, we create OpinionsQA, a new dataset for evaluating the alignment of LM opinions with those of 60 US demographic groups over topics ranging from abortion to automation. Across topics, we find substantial misalignment between the views reflected by current LMs and those of US demographic groups: on par with the Democrat-Republican divide on climate change. Notably, this misalignment persists even after explicitly steering the LMs towards particular demographic groups. Our analysis not only confirms prior observations about the left-leaning tendencies of some human feedback-tuned LMs, but also surfaces groups whose opinions are poorly reflected by current LMs (e.g., 65+ and widowed individuals). Our code and data are available at https://github.com/tatsu-lab/opinions_qa."
                },
                {
                    "title": "Co-Writing with Opinionated Language Models Affects Users\u2019 Views",
                    "abstract": "If large language models like GPT-3 preferably produce a particular point of view, they may influence people\u2019s opinions on an unknown scale. This study investigates whether a language-model-powered writing assistant that generates some opinions more often than others impacts what users write \u2013 and what they think. In an online experiment, we asked participants (N=1,506) to write a post discussing whether social media is good for society. Treatment group participants used a language-model-powered writing assistant configured to argue that social media is good or bad for society. Participants then completed a social media attitude survey, and independent judges (N=500) evaluated the opinions expressed in their writing. Using the opinionated language model affected the opinions expressed in participants\u2019 writing and shifted their opinions in the subsequent attitude survey. We discuss the wider implications of our results and argue that the opinions built into AI language technologies need to be monitored and engineered more carefully."
                },
                {
                    "title": "Evaluating Human-Language Model Interaction",
                    "abstract": "Many real-world applications of language models (LMs), such as writing assistance and code autocomplete, involve human-LM interaction. However, most benchmarks are non-interactive in that a model produces output without human involvement. To evaluate human-LM interaction, we develop a new framework, Human-AI Language-based Interaction Evaluation (HALIE), that defines the components of interactive systems and dimensions to consider when designing evaluation metrics. Compared to standard, non-interactive evaluation, HALIE captures (i) the interactive process, not only the final output; (ii) the first-person subjective experience, not just a third-party assessment; and (iii) notions of preference beyond quality (e.g., enjoyment and ownership). We then design five tasks to cover different forms of interaction: social dialogue, question answering, crossword puzzles, summarization, and metaphor generation. With four state-of-the-art LMs (three variants of OpenAI's GPT-3 and AI21 Labs' Jurassic-1), we find that better non-interactive performance does not always translate to better human-LM interaction. In particular, we highlight three cases where the results from non-interactive and interactive metrics diverge and underscore the importance of human-LM interaction for LM evaluation."
                },
                {
                    "title": "Picking on the Same Person: Does Algorithmic Monoculture lead to Outcome Homogenization?",
                    "abstract": "As the scope of machine learning broadens, we observe a recurring theme of algorithmic monoculture: the same systems, or systems that share components (e.g. training data), are deployed by multiple decision-makers. While sharing offers clear advantages (e.g. amortizing costs), does it bear risks? We introduce and formalize one such risk, outcome homogenization: the extent to which particular individuals or groups experience negative outcomes from all decision-makers. If the same individuals or groups exclusively experience undesirable outcomes, this may institutionalize systemic exclusion and reinscribe social hierarchy. To relate algorithmic monoculture and outcome homogenization, we propose the component-sharing hypothesis: if decision-makers share components like training data or specific models, then they will produce more homogeneous outcomes. We test this hypothesis on algorithmic fairness benchmarks, demonstrating that sharing training data reliably exacerbates homogenization, with individual-level effects generally exceeding group-level effects. Further, given the dominant paradigm in AI of foundation models, i.e. models that can be adapted for myriad downstream tasks, we test whether model sharing homogenizes outcomes across tasks. We observe mixed results: we find that for both vision and language settings, the specific methods for adapting a foundation model significantly influence the degree of outcome homogenization. We conclude with philosophical analyses of and societal challenges for outcome homogenization, with an eye towards implications for deployed machine learning systems."
                },
                {
                    "title": "Creative Writing with an AI-Powered Writing Assistant: Perspectives from Professional Writers",
                    "abstract": "Recent developments in natural language generation (NLG) using neural language models have brought us closer than ever to the goal of building AI-powered creative writing tools. However, most prior work on human-AI collaboration in the creative writing domain has evaluated new systems with amateur writers, typically in contrived user studies of limited scope. In this work, we commissioned 13 professional, published writers from a diverse set of creative writing backgrounds to craft stories using Wordcraft, a text editor with built-in AI-powered writing assistance tools. Using interviews and participant journals, we discuss the potential of NLG to have significant impact in the creative writing domain--especially with respect to brainstorming, generation of story details, world-building, and research assistance. Experienced writers, more so than amateurs, typically have well-developed systems and methodologies for writing, as well as distinctive voices and target audiences. Our work highlights the challenges in building for these writers; NLG technologies struggle to preserve style and authorial voice, and they lack deep understanding of story contents. In order for AI-powered writing assistants to realize their full potential, it is essential that they take into account the diverse goals and expertise of human writers."
                },
                {
                    "title": "Contrastive Decoding: Open-ended Text Generation as Optimization",
                    "abstract": "Given a language model (LM), maximum probability is a poor decoding objective for open-ended generation, because it produces short and repetitive text. On the other hand, sampling can often produce incoherent text that drifts from the original topics. We propose contrastive decoding (CD), a reliable decoding approach that optimizes a contrastive objective subject to a plausibility constraint. The contrastive objective returns the difference between the likelihood under a large LM (called the expert, e.g. OPT-13B) and a small LM (called the amateur, e.g. OPT-125M), and the constraint ensures that the outputs are plausible. CD is inspired by the fact that the failures of larger LMs (e.g., repetition, inco- herence) are even more prevalent in smaller LMs, and that this difference signals which texts should be preferred. CD requires zero additional training, and produces higher quality text than decoding from the larger LM alone. It also works across model scales (OPT-13B and GPT2-1.5B) and significantly outperforms four strong decoding algorithms (e.g., nucleus, top-k) in automatic and human evaluations across wikipedia, news and story domains."
                },
                {
                    "title": "Help me write a Poem - Instruction Tuning as a Vehicle for Collaborative Poetry Writing",
                    "abstract": "Recent work in training large language models (LLMs) to follow natural language instructions has opened up exciting opportunities for natural language interface design. Building on the prior success of large language models in the realm of computer assisted creativity, in this work, we present CoPoet, a collaborative poetry writing system, with the goal of to study if LLM\u2019s actually improve the quality of the generated content. In contrast to auto-completing a user\u2019s text, CoPoet is controlled by user instructions that specify the attributes of the desired text, such as Write a sentence about \u2018love\u2019 or Write a sentence ending in \u2018fly\u2019. The core component of our system is a language model fine-tuned on a diverse collection of instructions for poetry writing. Our model is not only competitive to publicly available LLMs trained on instructions (InstructGPT), but also capable of satisfying unseen compositional instructions. A study with 15 qualified crowdworkers shows that users successfully write poems with CoPoet on diverse topics ranging from Monarchy to Climate change, which are preferred by third-party evaluators over poems written without the system."
                },
                {
                    "title": "Co-Writing Screenplays and Theatre Scripts with Language Models: Evaluation by Industry Professionals",
                    "abstract": "Language models are increasingly attracting interest from writers. However, such models lack long-range semantic coherence, limiting their usefulness for longform creative writing. We address this limitation by applying language models hierarchically, in a system we call Dramatron. By building structural context via prompt chaining, Dramatron can generate coherent scripts and screenplays complete with title, characters, story beats, location descriptions, and dialogue. We illustrate Dramatron\u2019s usefulness as an interactive co-creative system with a user study of 15 theatre and film industry professionals. Participants co-wrote theatre scripts and screenplays with Dramatron and engaged in open-ended interviews. We report reflections both from our interviewees and from independent reviewers who critiqued performances of several of the scripts to illustrate how both Dramatron and hierarchical text generation could be useful for human-machine co-creativity. Finally, we discuss the suitability of Dramatron for co-creativity, ethical considerations\u2014including plagiarism and bias\u2014and participatory models for the design and deployment of such tools."
                },
                {
                    "title": "News Summarization and Evaluation in the Era of GPT-3",
                    "abstract": "The recent success of prompting large language models like GPT-3 has led to a paradigm shift in NLP research. In this paper, we study its impact on text summarization, focusing on the classic benchmark domain of news summarization. First, we investigate how GPT-3 compares against fine-tuned models trained on large summarization datasets. We show that not only do humans overwhelmingly prefer GPT-3 summaries, prompted using only a task description, but these also do not suffer from common dataset-specific issues such as poor factuality. Next, we study what this means for evaluation, particularly the role of gold standard test sets. Our experiments show that both reference-based and reference-free automatic metrics cannot reliably evaluate GPT-3 summaries. Finally, we evaluate models on a setting beyond generic summarization, specifically keyword-based summarization, and show how dominant fine-tuning approaches compare to prompting. To support further research, we release: (a) a corpus of 10K generated summaries from fine-tuned and prompt-based models across 4 standard summarization benchmarks, (b) 1K human preference judgments comparing different systems for generic- and keyword-based summarization."
                },
                {
                    "title": "Data Feedback Loops: Model-driven Amplification of Dataset Biases",
                    "abstract": "Datasets scraped from the internet have been critical to the successes of large-scale machine learning. Yet, this very success puts the utility of future internet-derived datasets at potential risk, as model outputs begin to replace human annotations as a source of supervision. In this work, we first formalize a system where interactions with one model are recorded as history and scraped as training data in the future. We then analyze its stability over time by tracking changes to a test-time bias statistic (e.g. gender bias of model predictions). We find that the degree of bias amplification is closely linked to whether the model's outputs behave like samples from the training distribution, a behavior which we characterize and define as consistent calibration. Experiments in three conditional prediction scenarios - image classification, visual role-labeling, and language generation - demonstrate that models that exhibit a sampling-like behavior are more calibrated and thus more stable. Based on this insight, we propose an intervention to help calibrate and stabilize unstable feedback systems. Code is available at https://github.com/rtaori/data_feedback."
                },
                {
                    "title": "Interacting with Next-Phrase Suggestions: How Suggestion Systems Aid and Influence the Cognitive Processes of Writing",
                    "abstract": "Writing with next-phrase suggestions powered by large language models is becoming more pervasive by the day. However, research to understand writers\u2019 interaction and decision-making processes while engaging with such systems is still emerging. We conducted a qualitative study to shed light on writers\u2019 cognitive processes while writing with next-phrase suggestion systems. To do so, we recruited 14 amateur writers to write two movie reviews each, one without suggestions and one with suggestions. Additionally, we also positively and negatively biased the suggestion system to get a diverse range of instances where writers\u2019 opinions and the bias in the language model align or misalign to varying degrees. We found that writers interact with next-phrase suggestions in various complex ways: Writers abstracted and extracted multiple parts of the suggestions and incorporated them within their writing, even when they disagreed with the suggestion as a whole; along with evaluating the suggestions on various criteria. The suggestion system also had various effects on the writing process, such as altering the writer\u2019s usual writing plans, leading to higher levels of distraction etc. Based on our qualitative analysis using the cognitive process model of writing by Hayes [35] as a lens, we propose a theoretical model of \u2019writer-suggestion interaction\u2019 for writing with GPT-2 (and causal language models in general) for a movie review writing task, followed by directions for future research and design."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "Wordcraft: Story Writing With Large Language Models",
                    "abstract": "The latest generation of large neural language models such as GPT-3 have achieved new levels of performance on benchmarks for language understanding and generation. These models have even demonstrated an ability to perform arbitrary tasks without explicit training. In this work, we sought to learn how people might use such models in the process of creative writing. We built Wordcraft, a text editor in which users collaborate with a generative language model to write a story. We evaluated Wordcraft with a user study in which participants wrote short stories with and without the tool. Our results show that large language models enable novel co-writing experiences. For example, the language model is able to engage in open-ended conversation about the story, respond to writers\u2019 custom requests expressed in natural language (such as \u201drewrite this text to be more Dickensian\u201d), and generate suggestions that serve to unblock writers in the creative process. Based on these results, we discuss design implications for future human-AI co-writing systems."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Repairing the Cracked Foundation: A Survey of Obstacles in Evaluation Practices for Generated Text",
                    "abstract": "Evaluation practices in natural language generation (NLG) have many known flaws, but improved evaluation approaches are rarely widely adopted. This issue has become more urgent, since neural generation models have improved to the point where their outputs can often no longer be distinguished based on the surface-level features that older metrics rely on. This paper surveys the issues with human and automatic model evaluations and with commonly used datasets in NLG that have been pointed out over the past 20 years. We summarize, categorize, and discuss how researchers have been addressing these issues and what their findings mean for the current state of model evaluations. Building on those insights, we lay out a long-term vision for evaluation research and propose concrete steps for researchers to improve their evaluation processes. Finally, we analyze 66 generation papers from recent NLP conferences in how well they already follow these suggestions and identify which areas require more drastic changes to the status quo."
                },
                {
                    "title": "Locally Typical Sampling",
                    "abstract": "Today\u2019s probabilistic language generators fall short when it comes to producing coherent and fluent text despite the fact that the underlying models perform well under standard metrics (e.g., perplexity). This discrepancy has puzzled the language generation community for the last few years. In this work, we posit that the abstraction of natural language generation as a discrete stochastic process\u2014which allows for an information-theoretic analysis\u2014can provide new insights into the behavior of probabilistic language generators, for example, why high-probability texts can be dull or repetitive. Humans use language as a means of communicating information, aiming to do so in a simultaneously efficient and error-minimizing manner; in fact, psycholinguistics research suggests humans choose each word in a string with this subconscious goal in mind. We formally define the set of strings that meet this criterion: Those for which each word has an information content close to the expected information content, namely, the conditional entropy of our model. We then propose a simple and efficient procedure for enforcing this criterion when generating from probabilistic models, which we call locally typical sampling. Automatic and human evaluations show that, in comparison to nucleus and top-k sampling, locally typical sampling offers competitive performance (in both abstractive summarization and story generation) in terms of quality while consistently reducing degenerate repetitions."
                },
                {
                    "title": "CoAuthor: Designing a Human-AI Collaborative Writing Dataset for Exploring Language Model Capabilities",
                    "abstract": "Large language models (LMs) offer unprecedented language generation capabilities and exciting opportunities for interaction design. However, their highly context-dependent capabilities are difficult to grasp and are often subjectively interpreted. In this paper, we argue that by curating and analyzing large interaction datasets, the HCI community can foster more incisive examinations of LMs\u2019 generative capabilities. Exemplifying this approach, we present CoAuthor, a dataset designed for revealing GPT-3\u2019s capabilities in assisting creative and argumentative writing. CoAuthor captures rich interactions between 63 writers and four instances of GPT-3 across 1445 writing sessions. We demonstrate that CoAuthor can address questions about GPT-3\u2019s language, ideation, and collaboration capabilities, and reveal its contribution as a writing \u201ccollaborator\u201d under various definitions of good collaboration. Finally, we discuss how this work may facilitate a more principled discussion around LMs\u2019 promises and pitfalls in relation to interaction design. The dataset and an interface for replaying the writing sessions are publicly available at https://coauthor.stanford.edu."
                },
                {
                    "title": "Machine-in-the-Loop Rewriting for Creative Image Captioning",
                    "abstract": "Machine-in-the-loop writing aims to enable humans to collaborate with models to complete their writing tasks more effectively. Prior work has found that providing humans a machine-written draft or sentence-level continuations has limited success since the generated text tends to deviate from humans' intention. To allow the user to retain control over the content, we train a rewriting model that, when prompted, modifies specified spans of text within the user's original draft to introduce descriptive and figurative elements locally in the text. We evaluate the model on its ability to collaborate with humans on the task of creative image captioning. On a user study through Amazon Mechanical Turk, our model is rated to be more helpful than a baseline infilling language model. In addition, third-party evaluation shows that users write more descriptive and figurative captions when collaborating with our model compared to completing the task alone."
                },
                {
                    "title": "On the Opportunities and Risks of Foundation Models",
                    "abstract": "AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature."
                },
                {
                    "title": "Divergence Frontiers for Generative Models: Sample Complexity, Quantization Effects, and Frontier Integrals",
                    "abstract": "The spectacular success of deep generative models calls for quantitative tools to measure their statistical performance. Divergence frontiers have recently been proposed as an evaluation framework for generative models, due to their ability to measure the quality-diversity trade-off inherent to deep generative modeling. We establish non-asymptotic bounds on the sample complexity of divergence frontiers. We also introduce frontier integrals which provide summary statistics of divergence frontiers. We show how smoothed estimators such as Good-Turing or Krichevsky-Trofimov can overcome the missing mass problem and lead to faster rates of convergence. We illustrate the theoretical results with numerical examples from natural language processing and computer vision."
                },
                {
                    "title": "The Algorithmic Leviathan: Arbitrariness, Fairness, and Opportunity in Algorithmic Decision Making Systems",
                    "abstract": "Automated decision-making systems implemented in public life are typically standardized. One algorithmic decision-making system can replace thousands of human deciders. Each of the humans so replaced had her own decision-making criteria: some good, some bad, and some arbitrary. Is such arbitrariness of moral concern? We argue that an isolated arbitrary decision need not morally wrong the individual whom it misclassifies. However, if the same algorithms are applied across a public sphere, such as hiring or lending, a person could be excluded from a large number of opportunities. This harm persists even when the automated decision-making systems are \"fair\" on standard metrics of fairness. We argue that such arbitrariness at scale is morally problematic and propose technically informed solutions that can lessen the impact of algorithms at scale and so mitigate or avoid the moral harms we identify."
                },
                {
                    "title": "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? \ud83e\udd9c",
                    "abstract": "The past 3 years of work in NLP have been characterized by the development and deployment of ever larger language models, especially for English. BERT, its variants, GPT-2/3, and others, most recently Switch-C, have pushed the boundaries of the possible both through architectural innovations and through sheer size. Using these pretrained models and the methodology of fine-tuning them for specific tasks, researchers have extended the state of the art on a wide array of tasks as measured by leaderboards on specific benchmarks for English. In this paper, we take a step back and ask: How big is too big? What are the possible risks associated with this technology and what paths are available for mitigating those risks? We provide recommendations including weighing the environmental and financial costs first, investing resources into curating and carefully documenting datasets rather than ingesting everything on the web, carrying out pre-development exercises evaluating how the planned approach fits into research and development goals and supports stakeholder values, and encouraging research directions beyond ever larger language models."
                },
                {
                    "title": "MAUVE: Measuring the Gap Between Neural Text and Human Text using Divergence Frontiers",
                    "abstract": "As major progress is made in open-ended text generation, measuring how close machine-generated text is to human language remains a critical open problem. We introduce MAUVE, a comparison measure for open-ended text generation, which directly compares the learnt distribution from a text generation model to the distribution of human-written text using divergence frontiers. MAUVE scales up to modern text generation models by computing information divergences in a quantized embedding space. Through an extensive empirical study on three open-ended generation tasks, we find that MAUVE identifies known properties of generated text, scales naturally with model size, and correlates with human judgments, with fewer restrictions than existing distributional evaluation metrics."
                },
                {
                    "title": "The Impact of Multiple Parallel Phrase Suggestions on Email Input and Composition Behaviour of Native and Non-Native English Writers",
                    "abstract": "We present an in-depth analysis of the impact of multi-word suggestion choices from a neural language model on user behaviour regarding input and text composition in email writing. Our study for the first time compares different numbers of parallel suggestions, and use by native and non-native English writers, to explore a trade-off of \u201cefficiency vs ideation\u201d, emerging from recent literature. We built a text editor prototype with a neural language model (GPT-2), refined in a prestudy with 30 people. In an online study (N=156), people composed emails in four conditions (0/1/3/6 parallel suggestions). Our results reveal (1) benefits for ideation, and costs for efficiency, when suggesting multiple phrases; (2) that non-native speakers benefit more from more suggestions; and (3) further insights into behaviour patterns. We discuss implications for research, the design of interactive suggestion systems, and the vision of supporting writers with AI instead of replacing them."
                },
                {
                    "title": "Algorithmic monoculture and social welfare",
                    "abstract": "Significance Algorithmic monoculture is a growing concern in the use of algorithms for high-stakes screening decisions in areas such as employment and lending. If many firms use the same algorithm, even if it is more accurate than the alternatives, the resulting \u201cmonoculture\u201d may be susceptible to correlated failures, much as a monocultural system is in biological settings. To investigate this concern, we develop a model of selection under monoculture. We find that even without any assumption of shocks or correlated failures\u2014i.e., under \u201cnormal operations\u201d\u2014the quality of decisions may decrease when multiple firms use the same algorithm. Thus, the introduction of a more accurate algorithm may decrease social welfare\u2014a kind of \u201cBraess\u2019 paradox\u201d for algorithmic decision-making. As algorithms are increasingly applied to screen applicants for high-stakes decisions in employment, lending, and other domains, concerns have been raised about the effects of algorithmic monoculture, in which many decision-makers all rely on the same algorithm. This concern invokes analogies to agriculture, where a monocultural system runs the risk of severe harm from unexpected shocks. Here, we show that the dangers of algorithmic monoculture run much deeper, in that monocultural convergence on a single algorithm by a group of decision-making agents, even when the algorithm is more accurate for any one agent in isolation, can reduce the overall quality of the decisions being made by the full collection of agents. Unexpected shocks are therefore not needed to expose the risks of monoculture; it can hurt accuracy even under \u201cnormal\u201d operations and even for algorithms that are more accurate when used by only a single decision-maker. Our results rely on minimal assumptions and involve the development of a probabilistic framework for analyzing systems that use multiple noisy estimates of a set of alternatives."
                },
                {
                    "title": "Langsmith: An Interactive Academic Text Revision System",
                    "abstract": "Despite the current diversity and inclusion initiatives in the academic community, researchers with a non-native command of English still face significant obstacles when writing papers in English. This paper presents the Langsmith editor, which assists inexperienced, non-native researchers to write English papers, especially in the natural language processing (NLP) field. Our system can suggest fluent, academic-style sentences to writers based on their rough, incomplete phrases or sentences. The system also encourages interaction between human writers and the computerized revision system. The experimental results demonstrated that Langsmith helps non-native English-speaker students write papers in English. The system is available at https://emnlp-demo.editor. langsmith.co.jp/."
                },
                {
                    "title": "BLEURT: Learning Robust Metrics for Text Generation",
                    "abstract": "Text generation has made significant advances in the last few years. Yet, evaluation metrics have lagged behind, as the most popular choices (e.g., BLEU and ROUGE) may correlate poorly with human judgment. We propose BLEURT, a learned evaluation metric for English based on BERT. BLEURT can model human judgment with a few thousand possibly biased training examples. A key aspect of our approach is a novel pre-training scheme that uses millions of synthetic examples to help the model generalize. BLEURT provides state-of-the-art results on the last three years of the WMT Metrics shared task and the WebNLG data set. In contrast to a vanilla BERT-based approach, it yields superior results even when the training data is scarce and out-of-distribution."
                },
                {
                    "title": "Evaluating the Evaluation of Diversity in Natural Language Generation",
                    "abstract": "Despite growing interest in natural language generation (NLG) models that produce diverse outputs, there is currently no principled method for evaluating the diversity of an NLG system. In this work, we propose a framework for evaluating diversity metrics. The framework measures the correlation between a proposed diversity metric and a diversity parameter, a single parameter that controls some aspect of diversity in generated text. For example, a diversity parameter might be a binary variable used to instruct crowdsourcing workers to generate text with either low or high content diversity. We demonstrate the utility of our framework by: (a) establishing best practices for eliciting diversity judgments from humans, (b) showing that humans substantially outperform automatic metrics in estimating content diversity, and (c) demonstrating that existing methods for controlling diversity by tuning a \u201cdecoding parameter\u201d mostly affect form but not meaning. Our framework can advance the understanding of different diversity metrics, an essential step on the road towards better NLG systems."
                },
                {
                    "title": "Predictive text encourages predictable writing",
                    "abstract": "Intelligent text entry systems, including the now-ubiquitous predictive keyboard, can make text entry more efficient, but little is known about how these systems affect the content that people write. To study how predictive text systems affect content, we compared image captions written with different kinds of predictive text suggestions. Our key findings were that captions written with suggestions were shorter and that they included fewer words that that the system did not predict. Suggestions also boosted text entry speed, but with diminishing benefit for faster typists. Our findings imply that text entry systems should be evaluated not just by speed and accuracy but also by their effect on the content written."
                },
                {
                    "title": "AI-Mediated Communication: Definition, Research Agenda, and Ethical Considerations",
                    "abstract": "\n We define Artificial Intelligence-Mediated Communication (AI-MC) as interpersonal communication in which an intelligent agent operates on behalf of a communicator by modifying, augmenting, or generating messages to accomplish communication goals. The recent advent of AI-MC raises new questions about how technology may shape human communication and requires re-evaluation \u2013 and potentially expansion \u2013 of many of Computer-Mediated Communication\u2019s (CMC) key theories, frameworks, and findings. A research agenda around AI-MC should consider the design of these technologies and the psychological, linguistic, relational, policy and ethical implications of introducing AI into human\u2013human communication. This article aims to articulate such an agenda."
                },
                {
                    "title": "Do Massively Pretrained Language Models Make Better Storytellers?",
                    "abstract": "Large neural language models trained on massive amounts of text have emerged as a formidable strategy for Natural Language Understanding tasks. However, the strength of these models as Natural Language Generators is less clear. Though anecdotal evidence suggests that these models generate better quality text, there has been no detailed study characterizing their generation abilities. In this work, we compare the performance of an extensively pretrained model, OpenAI GPT2-117 (Radford et al., 2019), to a state-of-the-art neural story generation model (Fan et al., 2018). By evaluating the generated text across a wide variety of automatic metrics, we characterize the ways in which pretrained models do, and do not, make better storytellers. We find that although GPT2-117 conditions more strongly on context, is more sensitive to ordering of events, and uses more unusual words, it is just as likely to produce repetitive and under-diverse text when using likelihood-maximizing decoding algorithms."
                },
                {
                    "title": "Metaphoria: An Algorithmic Companion for Metaphor Creation",
                    "abstract": "Creative writing, from poetry to journalism, is at the crux of human ingenuity and social interaction. Existing creative writing support tools produce entire passages or fully formed sentences, but these approaches fail to adapt to the writer's own ideas and intentions. Instead we posit to build tools that generate ideas coherent with the writer's context and encourage writers to produce divergent outcomes. To explore this, we focus on supporting metaphor creation. We present Metaphoria, an interactive system that generates metaphorical connections based on an input word from the writer. Our studies show that Metaphoria provides more coherent suggestions than existing systems, and supports the expression of writers' unique intentions. We discuss the complex issue of ownership in human-machine collaboration and how to build adaptive creativity support tools in other domains."
                },
                {
                    "title": "BERTScore: Evaluating Text Generation with BERT",
                    "abstract": "We propose BERTScore, an automatic evaluation metric for text generation. Analogously to common metrics, BERTScore computes a similarity score for each token in the candidate sentence with each token in the reference sentence. However, instead of exact matches, we compute token similarity using contextual embeddings. We evaluate using the outputs of 363 machine translation and image captioning systems. BERTScore correlates better with human judgments and provides stronger model selection performance than existing metrics. Finally, we use an adversarial paraphrase detection task to show that BERTScore is more robust to challenging examples when compared to existing metrics."
                },
                {
                    "title": "Creative Writing with a Machine in the Loop: Case Studies on Slogans and Stories",
                    "abstract": "As the quality of natural language generated by artificial intelligence systems improves, writing interfaces can support interventions beyond grammar-checking and spell-checking, such as suggesting content to spark new ideas. To explore the possibility of machine-in-the-loop creative writing, we performed two case studies using two system prototypes, one for short story writing and one for slogan writing. Participants in our studies were asked to write with a machine in the loop or alone (control condition). They assessed their writing and experience through surveys and an open-ended interview. We collected additional assessments of the writing from Amazon Mechanical Turk crowdworkers. Our findings indicate that participants found the process fun and helpful and could envision use cases for future systems. At the same time, machine suggestions do not necessarily lead to better written artifacts. We therefore suggest novel natural language models and design choices that may better support creative writing."
                },
                {
                    "title": "Automated Assistance for Creative Writing with an RNN Language Model",
                    "abstract": "This work demonstrates an interface, Creative Help, that assists people with creative writing by automatically suggesting new sentences in a story. Authors can freely edit the generated suggestions, and the application tracks their modifications. We make use of a Recurrent Neural Network language model to generate suggestions in a simple probabilistic way. Motivated by the theorized role of unpredictability in creativity, we vary the degree of randomness in the probability distribution used to generate the sentences, and find that authors' interactions with the suggestions are influenced by this randomness."
                },
                {
                    "title": "Texygen: A Benchmarking Platform for Text Generation Models",
                    "abstract": "We introduce Texygen, a benchmarking platform to support research on open-domain text generation models. Texygen has not only implemented a majority of text generation models, but also covered a set of metrics that evaluate the diversity, the quality and the consistency of the generated texts. The Texygen platform could help standardize the research on text generation and improve the reproductivity and reliability of future research work in text generation."
                },
                {
                    "title": "Hafez: an Interactive Poetry Generation System",
                    "abstract": "Hafez is an automatic poetry generation system that integrates a Recurrent Neural Network (RNN) with a Finite State Accep-tor (FSA). It generates sonnets given arbitrary topics. Furthermore, Hafez enables users to revise and polish generated poems by adjusting various style con\ufb01gurations. Experiments demonstrate that such \u201cpol-ish\u201d mechanisms consider the user\u2019s intention and lead to a better poem. For evaluation, we build a web interface where users can rate the quality of each poem from 1 to 5 stars. We also speed up the whole system by a factor of 10, via vocabulary pruning and GPU computation, so that adequate feedback can be collected at a fast pace. Based on such feedback, the system learns to adjust its parameters to improve poetry quality."
                },
                {
                    "title": "Poetry of the Crowd: A Human Computation Algorithm to Convert Prose into Rhyming Verse",
                    "abstract": "\n \n Poetry composition is a very complex task that requires a poet to satisfy multiple constraints concurrently. We believe that the task can be augmented by combining the creative abilities of humans with computational algorithms that efficiently constrain and permute available choices. We present a hybrid method for generating poetry from prose that combines crowdsourcing with natural language processing (NLP) machinery. We test the ability of crowd workers to accomplish the technically challenging and creative task of composing poems.\n \n"
                },
                {
                    "title": "Say Anything: Using Textual Case-Based Reasoning to Enable Open-Domain Interactive Storytelling",
                    "abstract": "We describe Say Anything, a new interactive storytelling system that collaboratively writes textual narratives with human users. Unlike previous attempts, this interactive storytelling system places no restrictions on the content or direction of the user\u2019s contribution to the emerging storyline. In response to these contributions, the computer continues the storyline with narration that is both coherent and entertaining. This capacity for open-domain interactive storytelling is enabled by an extremely large repository of nonfiction personal stories, which is used as a knowledge base in a case-based reasoning architecture. In this article, we describe the three main components of our case-based reasoning approach: a million-item corpus of personal stories mined from internet weblogs, a case retrieval strategy that is optimized for narrative coherence, and an adaptation strategy that ensures that repurposed sentences from the case base are appropriate for the user\u2019s emerging fiction. We describe a series of evaluations of the system\u2019s ability to produce coherent and entertaining stories, and we compare these narratives with single-author stories posted to internet weblogs."
                },
                {
                    "title": "Scikit-learn: Machine Learning in Python",
                    "abstract": "Scikit-learn is a Python module integrating a wide range of state-of-the-art machine learning algorithms for medium-scale supervised and unsupervised problems. This package focuses on bringing machine learning to non-specialists using a general-purpose high-level language. Emphasis is put on ease of use, performance, documentation, and API consistency. It has minimal dependencies and is distributed under the simplified BSD license, encouraging its use in both academic and commercial settings. Source code, binaries, and documentation can be downloaded from http://scikit-learn.sourceforge.net."
                },
                {
                    "title": "ROUGE: A Package for Automatic Evaluation of Summaries",
                    "abstract": "ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST."
                },
                {
                    "title": "Bleu: a Method for Automatic Evaluation of Machine Translation",
                    "abstract": "Human evaluations of machine translation are extensive but expensive. Human evaluations can take months to finish and involve human labor that can not be reused. We propose a method of automatic machine translation evaluation that is quick, inexpensive, and language-independent, that correlates highly with human evaluation, and that has little marginal cost per run. We present this method as an automated understudy to skilled human judges which substitutes for them when there is need for quick or frequent evaluations."
                },
                {
                    "title": "Compression of individual sequences via variable-rate coding",
                    "abstract": "Compressibility of individual sequences by the class of generalized finite-state information-lossless encoders is investigated. These encoders can operate in a variable-rate mode as well as a fixed-rate one, and they allow for any finite-state scheme of variable-length-to-variable-length coding. For every individual infinite sequence x a quantity \\rho(x) is defined, called the compressibility of x , which is shown to be the asymptotically attainable lower bound on the compression ratio that can be achieved for x by any finite-state encoder. This is demonstrated by means of a constructive coding theorem and its converse that, apart from their asymptotic significance, also provide useful performance criteria for finite and practical data-compression tasks. The proposed concept of compressibility is also shown to play a role analogous to that of entropy in classical information theory where one deals with probabilistic ensembles of sequences rather than with individual sequences. While the definition of \\rho(x) allows a different machine for each different sequence to be compressed, the constructive coding theorem leads to a universal algorithm that is asymptotically optimal for all sequences."
                },
                {
                    "title": "A universal algorithm for sequential data compression",
                    "abstract": "A universal algorithm for sequential data compression is presented. Its performance is investigated with respect to a nonprobabilistic model of constrained sources. The compression ratio achieved by the proposed universal code uniformly approaches the lower bounds on the compression ratios attainable by block-to-variable codes and variable-to-block codes designed to match a completely specified source."
                },
                {
                    "title": "Story Centaur: Large Language Model Few Shot Learning as a Creative Writing Tool",
                    "abstract": "Few shot learning with large language models has the potential to give individuals without formal machine learning training the access to a wide range of text to text models. We consider how this applies to creative writers and present Story Centaur, a user interface for prototyping few shot models and a set of recombinable web components that deploy them. Story Centaur\u2019s goal is to expose creative writers to few shot learning with a simple but powerful interface that lets them compose their own co-creation tools that further their own unique artistic directions. We build out several examples of such tools, and in the process probe the boundaries and issues surrounding generation with large language models."
                },
                {
                    "title": "STORIUM: A Dataset and Evaluation Platform for Machine-in-the-Loop Story Generation",
                    "abstract": "Systems for story generation are asked to produce plausible and enjoyable stories given an input context. This task is underspecified, as a vast number of diverse stories can originate from a single input. The large output space makes it difficult to build and evaluate story generation models, as (1) existing datasets lack rich enough contexts to meaningfully guide models, and (2) existing evaluations (both crowdsourced and automatic) are unreliable for assessing long-form creative text. To address these issues, we introduce a dataset and evaluation platform built from STORIUM, an online collaborative storytelling community. Our author-generated dataset contains 6K lengthy stories (125M tokens) with fine-grained natural language annotations (e.g., character goals and attributes) interspersed throughout each narrative, forming a robust source for guiding models. We evaluate language models fine-tuned on our dataset by integrating them onto STORIUM, where real authors can query a model for suggested story continuations and then edit them. Automatic metrics computed over these edits correlate well with both user ratings of generated stories and qualitative feedback from semi-structured user interviews. We release both the STORIUM dataset and evaluation platform to spur more principled research into story generation."
                },
                {
                    "title": "How to Compare Summarizers without Target Length? Pitfalls, Solutions and Re-Examination of the Neural Summarization Literature",
                    "abstract": "We show that plain ROUGE F1 scores are not ideal for comparing current neural systems which on average produce different lengths. This is due to a non-linear pattern between ROUGE F1 and summary length. To alleviate the effect of length during evaluation, we have proposed a new method which normalizes the ROUGE F1 scores of a system by that of a random system with same average output length. A pilot human evaluation has shown that humans prefer short summaries in terms of the verbosity of a summary but overall consider longer summaries to be of higher quality. While human evaluations are more expensive in time and resources, it is clear that normalization, such as the one we proposed for automatic evaluation, will make human evaluations more meaningful."
                },
                {
                    "title": "Evaluating Story Generation Systems Using Automated Linguistic Analyses",
                    "abstract": "Story generation is a well-recognized task in computational creativity research, but one that can be di\ufb03cult to evaluate empirically. It is often ine\ufb03cient and costly to rely solely on human feedback for judging the quality of generated stories. We address this by examining the use of linguistic analyses for automated evaluation, using metrics from existing work on predicting writing quality. We apply these metrics speci\ufb01cally to story continuation, where a model is given the beginning of a story and generates the next sentence, which is useful for systems that interactively support authors\u2019 creativity in writing. We compare sentences generated by different existing models to human-authored ones according to the analyses. The results show some meaningful di\ufb00er-ences between the models, suggesting that this evaluation approach may be advantageous for future research."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.CY",
                "cs.HC",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nDoes writing with large language models (LLMs) unintentionally reduce content diversity among users?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the implications of LLMs on content creation and diversity. Understanding how LLMs influence writing can lead to better design and implementation of these models, ensuring they enhance rather than homogenize creative expression. This research could advance knowledge in the fields of natural language processing and human-computer interaction, leading to practical applications that promote diverse perspectives in generated content, which is vital in a world where information is increasingly generated by algorithms.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the complexity of measuring content diversity and the subjective nature of writing quality. Naive approaches may fail because they might not account for the nuanced ways in which LLMs influence individual writing styles and the collective output of groups. Technical obstacles include developing robust metrics for diversity and homogenization, while theoretical challenges involve understanding the cognitive and social dynamics of how users interact with LLMs. Additionally, isolating the effects of LLMs from other variables in writing behavior complicates the analysis.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the capabilities and biases of LLMs without specifically addressing their impact on content diversity. Limitations in existing studies include a lack of controlled experiments that isolate the effects of LLMs on writing diversity and the absence of comprehensive metrics to evaluate this aspect. Barriers such as the complexity of human writing behavior and the evolving nature of LLMs have also hindered progress. Our approach differs by employing a controlled experimental design with clear metrics to assess the impact of LLMs on both individual and collective writing diversity.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves a controlled experiment where participants are divided into three groups: a control group writing without LLM assistance, a treatment group using a base language model (GPT3), and a second treatment group using a feedback-tuned model (InstructGPT). We will collect 100 essays from each group on 10 topics, measuring content diversity through metrics such as homogenization and lexical diversity. We expect to find that essays co-written with InstructGPT exhibit higher homogenization and lower diversity compared to those written"
            }
        },
        "author_data": {
            "d404bc7f-c3e7-4eb5-ba1f-10afb4653599": {
                "pk": "d404bc7f-c3e7-4eb5-ba1f-10afb4653599",
                "name": "Vishakh Padmakumar",
                "collaborators": [
                    "He He",
                    "Jason Phang",
                    "Alicia Parrish",
                    "Angelica Chen",
                    "Sam Bowman",
                    "Richard Yuanzhe Pang",
                    "Nikita Nangia",
                    "Tuhin Chakrabarty",
                    "Jana Thompson",
                    "Faeze Brahman",
                    "S. Muresan",
                    "Behnam Hedayatnia",
                    "Nanyun Peng",
                    "Samuel R. Bowman",
                    "Nitish Joshi",
                    "Ankur P. Parikh",
                    "Phu Mon Htut",
                    "Yu Hou",
                    "Di Jin",
                    "Patrick Lange",
                    "Seokhwan Kim",
                    "Yang Liu",
                    "Dilek Z. Hakkani-T\u00fcr",
                    "Chen Zhao",
                    "Kyunghyun Cho",
                    "Shuo Guan",
                    "Julian Michael",
                    "Salsabila Mahdi",
                    "David Rein",
                    "Jackson Petty",
                    "Julien Dirani",
                    "Abulhair Saparov",
                    "Seyed Mehran Kazemi",
                    "Najoung Kim",
                    "Leonard Lausen",
                    "Miguel Ballesteros",
                    "Sheng Zha",
                    "G. Karypis",
                    "Hengxing He",
                    "Aarohi Srivastava",
                    "Abhinav Rastogi",
                    "Abhishek Rao",
                    "Abu Awal Md Shoeb",
                    "Abubakar Abid",
                    "Adam Fisch",
                    "Adam R. Brown",
                    "Adam Santoro",
                    "Aditya Gupta",
                    "Adri\u00e0 Garriga-Alonso",
                    "Agnieszka Kluska",
                    "Aitor Lewkowycz",
                    "Akshat Agarwal",
                    "Alethea Power",
                    "Alex Ray",
                    "Alex Warstadt",
                    "Alexander W. Kocurek",
                    "Ali Safaya",
                    "Ali Tazarv",
                    "Alice Xiang",
                    "Allen Nie",
                    "Aman Hussain",
                    "Amanda Askell",
                    "Amanda Dsouza",
                    "Ambrose Slone",
                    "Ameet Rahane",
                    "Anantharaman S. Iyer",
                    "Anders Andreassen",
                    "Andrea Madotto",
                    "Andrea Santilli",
                    "Andreas Stuhlmuller",
                    "Andrew M. Dai",
                    "Andrew La",
                    "Andrew Kyle Lampinen",
                    "Andy Zou",
                    "Angela Jiang",
                    "Anh Vuong",
                    "Animesh Gupta",
                    "Anna Gottardi",
                    "Antonio Norelli",
                    "Anu Venkatesh",
                    "Arash Gholamidavoodi",
                    "Arfa Tabassum",
                    "Arul Menezes",
                    "Arun Kirubarajan",
                    "A. Mullokandov",
                    "Ashish Sabharwal",
                    "Austin Herrick",
                    "Avia Efrat",
                    "Aykut Erdem",
                    "Ayla Karakacs",
                    "B. R. Roberts",
                    "B. S. Loe",
                    "Barret Zoph",
                    "Bartlomiej Bojanowski",
                    "Batuhan Ozyurt",
                    "Behnam Neyshabur",
                    "Benjamin Inden",
                    "Benno Stein",
                    "Berk Ekmekci",
                    "Bill Yuchen Lin"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Models",
                    "Text Generation",
                    "Summarization"
                ],
                "publications": [
                    {
                        "title": "Creativity Support in the Age of Large Language Models: An Empirical Study Involving Professional Writers",
                        "abstract": "The development of large language models (LLMs) capable of following instructions and engaging in conversational interactions has led to increased interest in their use across various support tools. We investigate the effectiveness of contemporary LLMs in assisting professional writers via an empirical user study (n=30). The design of our collaborative writing interface is grounded in the cognitive process model of writing [17]. This allows writers to obtain model help in each of the three non-linear cognitive activities in the writing process: planning, translating and reviewing. Participants write short fiction/non-fiction with model help and are subsequently asked to submit a post-completion survey to provide qualitative feedback on the potential and pitfalls of LLMs as writing collaborators. Upon analyzing the writer-LLM interactions, we find that while seeking help across all three types of cognitive activities, writers find LLMs more helpful in translation and reviewing. Our findings from analyzing both the interactions and the survey responses highlight future research directions in creative writing assistance using LLMs."
                    },
                    {
                        "title": "Investigating the Representation of Open Domain Dialogue Context for Transformer Models",
                        "abstract": "The bulk of work adapting transformer models to open-domain dialogue represents dialogue context as the concatenated set of turns in natural language. However, it is unclear if this is the best approach. In this work, we investigate this question by means of an empirical controlled experiment varying the dialogue context format from text-only formats (all recent utterances, summaries, selected utterances) as well as variants that are more structurally different (triples, AMR). We compare these formats based on fine-tuned model performance on two downstream tasks\u2014knowledge selection and response generation. We find that simply concatenating the utterances works as a strong baseline in most cases, but is outperformed in longer contexts by a hybrid approach of combining a summary of the context with recent utterances. Through empirical analysis, our work highlights the need to examine the format of context representation and offers recommendations on adapting general-purpose language models to dialogue tasks."
                    },
                    {
                        "title": "Two Failures of Self-Consistency in the Multi-Step Reasoning of LLMs",
                        "abstract": "Large language models (LLMs) have achieved widespread success on a variety of in-context few-shot tasks, but this success is typically evaluated via correctness rather than consistency. We argue that self-consistency is an important criteria for valid multi-step reasoning in tasks where the solution is composed of the answers to multiple sub-steps. We propose two types of self-consistency that are particularly important for multi-step reasoning -- hypothetical consistency (a model's ability to predict what its output would be in a hypothetical other context) and compositional consistency (consistency of a model's final outputs when intermediate sub-steps are replaced with the model's outputs for those steps). We demonstrate that multiple variants of the GPT-3/-4 models exhibit poor consistency rates across both types of consistency on a variety of tasks."
                    },
                    {
                        "title": "Extract, Select and Rewrite: A Modular Sentence Summarization Method",
                        "abstract": "A modular approach has the advantage of being compositional and controllable, comparing to most end-to-end models. In this paper we propose Extract-Select-Rewrite (ESR), a three-phase abstractive sentence summarization method. We decompose summarization into three stages: (i) knowledge extraction, where we extract relation triples from the text using off-the-shelf tools; (ii) content selection, where a subset of triples are selected; and (iii) rewriting, where the selected triple are realized into natural language. Our results demonstrates that ESR is competitive with the best end-to-end models while being more faithful. %than these baseline models. Being modular, ESR\u2019s modules can be trained on separate data which is beneficial in low-resource settings and enhancing the style controllability on text generation."
                    },
                    {
                        "title": "Creativity Support in the Age of Large Language Models: An Empirical Study Involving Emerging Writers",
                        "abstract": "The development of large language models (LLMs) capable of following instructions and engaging in conversational interactions sparked increased interest in their utilization across various support tools. We investigate the utility of modern LLMs in assisting professional writers via an empirical user study (n=30). The design of our collaborative writing interface is grounded in the cognitive process model of writing that views writing as a goal-oriented thinking process encompassing non-linear cognitive activities: planning, translating, and reviewing. Participants are asked to submit a post-completion survey to provide feedback on the potential and pitfalls of LLMs as writing collaborators. Upon analyzing the writer-LLM interactions, we find that while writers seek LLM's help across all three types of cognitive activities, they find LLMs more helpful in translation and reviewing. Our findings from analyzing both the interactions and the survey responses highlight future research directions in creative writing assistance using LLMs."
                    },
                    {
                        "title": "Debate Helps Supervise Unreliable Experts",
                        "abstract": "As AI systems are used to answer more difficult questions and potentially help create new knowledge, judging the truthfulness of their outputs becomes more difficult and more important. How can we supervise unreliable experts, which have access to the truth but may not accurately report it, to give answers that are systematically true and don't just superficially seem true, when the supervisor can't tell the difference between the two on their own? In this work, we show that debate between two unreliable experts can help a non-expert judge more reliably identify the truth. We collect a dataset of human-written debates on hard reading comprehension questions where the judge has not read the source passage, only ever seeing expert arguments and short quotes selectively revealed by 'expert' debaters who have access to the passage. In our debates, one expert argues for the correct answer, and the other for an incorrect answer. Comparing debate to a baseline we call consultancy, where a single expert argues for only one answer which is correct half of the time, we find that debate performs significantly better, with 84% judge accuracy compared to consultancy's 74%. Debates are also more efficient, being 68% of the length of consultancies. By comparing human to AI debaters, we find evidence that with more skilled (in this case, human) debaters, the performance of debate goes up but the performance of consultancy goes down. Our error analysis also supports this trend, with 46% of errors in human debate attributable to mistakes by the honest debater (which should go away with increased skill); whereas 52% of errors in human consultancy are due to debaters obfuscating the relevant evidence from the judge (which should become worse with increased skill). Overall, these results show that debate is a promising approach for supervising increasingly capable but potentially unreliable AI systems."
                    },
                    {
                        "title": "Testing the General Deductive Reasoning Capacity of Large Language Models Using OOD Examples",
                        "abstract": "Given the intractably large size of the space of proofs, any model that is capable of general deductive reasoning must generalize to proofs of greater complexity. Recent studies have shown that large language models (LLMs) possess some abstract deductive reasoning ability given chain-of-thought prompts. However, they have primarily been tested on proofs using modus ponens or of a specific size, and from the same distribution as the in-context examples. To measure the general deductive reasoning ability of LLMs, we test on a broad set of deduction rules and measure their ability to generalize to more complex proofs from simpler demonstrations from multiple angles: depth-, width-, and compositional generalization. To facilitate systematic exploration, we construct a new synthetic and programmable reasoning dataset that enables control over deduction rules and proof complexity. Our experiments on four LLMs of various sizes and training objectives show that they are able to generalize to compositional proofs. However, they have difficulty generalizing to longer proofs, and they require explicit demonstrations to produce hypothetical subproofs, specifically in proof by cases and proof by contradiction."
                    },
                    {
                        "title": "Extrapolative Controlled Sequence Generation via Iterative Refinement",
                        "abstract": "We study the problem of extrapolative controlled generation, i.e., generating sequences with attribute values beyond the range seen in training. This task is of significant importance in automated design, especially drug discovery, where the goal is to design novel proteins that are \\textit{better} (e.g., more stable) than existing sequences. Thus, by definition, the target sequences and their attribute values are out of the training distribution, posing challenges to existing methods that aim to directly generate the target sequence. Instead, in this work, we propose Iterative Controlled Extrapolation (ICE) which iteratively makes local edits to a sequence to enable extrapolation. We train the model on synthetically generated sequence pairs that demonstrate small improvement in the attribute value. Results on one natural language task (sentiment analysis) and two protein engineering tasks (ACE2 stability and AAV fitness) show that ICE considerably outperforms state-of-the-art approaches despite its simplicity. Our code and models are available at: https://github.com/vishakhpk/iter-extrapolation."
                    },
                    {
                        "title": "Exploring the Role of Task Transferability in Large-Scale Multi-Task Learning",
                        "abstract": "Recent work has found that multi-task training with a large number of diverse tasks can uniformly improve downstream performance on unseen target tasks. In contrast, literature on task transferability has established that the choice of intermediate tasks can heavily affect downstream task performance. In this work, we aim to disentangle the effect of scale and relatedness of tasks in multi-task representation learning. We find that, on average, increasing the scale of multi-task learning, in terms of the number of tasks, indeed results in better learned representations than smaller multi-task setups. However, if the target tasks are known ahead of time, then training on a smaller set of related tasks is competitive to the large-scale multi-task training at a reduced computational cost."
                    },
                    {
                        "title": "Help me write a Poem - Instruction Tuning as a Vehicle for Collaborative Poetry Writing",
                        "abstract": "Recent work in training large language models (LLMs) to follow natural language instructions has opened up exciting opportunities for natural language interface design. Building on the prior success of large language models in the realm of computer assisted creativity, in this work, we present CoPoet, a collaborative poetry writing system, with the goal of to study if LLM\u2019s actually improve the quality of the generated content. In contrast to auto-completing a user\u2019s text, CoPoet is controlled by user instructions that specify the attributes of the desired text, such as Write a sentence about \u2018love\u2019 or Write a sentence ending in \u2018fly\u2019. The core component of our system is a language model fine-tuned on a diverse collection of instructions for poetry writing. Our model is not only competitive to publicly available LLMs trained on instructions (InstructGPT), but also capable of satisfying unseen compositional instructions. A study with 15 qualified crowdworkers shows that users successfully write poems with CoPoet on diverse topics ranging from Monarchy to Climate change, which are preferred by third-party evaluators over poems written without the system."
                    },
                    {
                        "title": "Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models",
                        "abstract": "Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit\"breakthrough\"behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting."
                    },
                    {
                        "title": "Reward Gaming in Conditional Text Generation",
                        "abstract": "To align conditional text generation model outputs with desired behaviors, there has been an increasing focus on training the model using reinforcement learning (RL) with reward functions learned from human annotations. Under this framework, we identify three common cases where high rewards are incorrectly assigned to undesirable patterns: noise-induced spurious correlation, naturally occurring spurious correlation, and covariate shift. We show that even though learned metrics achieve high performance on the distribution of the data used to train the reward function, the undesirable patterns may be amplified during RL training of the text generation model. While there has been discussion about reward gaming in the RL or safety community, in this discussion piece, we would like to highlight reward gaming in the natural language generation (NLG) community using concrete conditional text generation examples and discuss potential fixes and areas for future work."
                    },
                    {
                        "title": "Two-Turn Debate Doesn't Help Humans Answer Hard Reading Comprehension Questions",
                        "abstract": "The use of language-model-based question-answering systems to aid humans in completing difficult tasks is limited, in part, by the unreliability of the text these systems generate. Using hard multiple-choice reading comprehension questions as a testbed, we assess whether presenting humans with arguments for two competing answer options, where one is correct and the other is incorrect, allows human judges to perform more accurately, even when one of the arguments is unreliable and deceptive. If this is helpful, we may be able to increase our justified trust in language-model-based systems by asking them to produce these arguments where needed. Previous research has shown that just a single turn of arguments in this format is not helpful to humans. However, as debate settings are characterized by a back-and-forth dialogue, we follow up on previous results to test whether adding a second round of counter-arguments is helpful to humans. We find that, regardless of whether they have access to arguments or not, humans perform similarly on our task. These findings suggest that, in the case of answering reading comprehension questions, debate is not a helpful format."
                    },
                    {
                        "title": "Machine-in-the-Loop Rewriting for Creative Image Captioning",
                        "abstract": "Machine-in-the-loop writing aims to enable humans to collaborate with models to complete their writing tasks more effectively. Prior work has found that providing humans a machine-written draft or sentence-level continuations has limited success since the generated text tends to deviate from humans' intention. To allow the user to retain control over the content, we train a rewriting model that, when prompted, modifies specified spans of text within the user's original draft to introduce descriptive and figurative elements locally in the text. We evaluate the model on its ability to collaborate with humans on the task of creative image captioning. On a user study through Amazon Mechanical Turk, our model is rated to be more helpful than a baseline infilling language model. In addition, third-party evaluation shows that users write more descriptive and figurative captions when collaborating with our model compared to completing the task alone."
                    },
                    {
                        "title": "QuALITY: Question Answering with Long Input Texts, Yes!",
                        "abstract": "To enable building and testing models on long-document comprehension, we introduce QuALITY, a multiple-choice QA dataset with context passages in English that have an average length of about 5,000 tokens, much longer than typical current models can process. Unlike in prior work with passages, our questions are written and validated by contributors who have read the entire passage, rather than relying on summaries or excerpts. In addition, only half of the questions are answerable by annotators working under tight time constraints, indicating that skimming and simple search are not enough to consistently perform well. Our baseline models perform poorly on this task (55.4%) and significantly lag behind human performance (93.5%)."
                    },
                    {
                        "title": "BBQ: A hand-built bias benchmark for question answering",
                        "abstract": "It is well documented that NLP models learn social biases, but little work has been done on how these biases manifest in model outputs for applied tasks like question answering (QA). We introduce the Bias Benchmark for QA (BBQ), a dataset of question-sets constructed by the authors that highlight attested social biases against people belonging to protected classes along nine social dimensions relevant for U.S. English-speaking contexts. Our task evaluate model responses at two levels: (i) given an under-informative context, we test how strongly responses reflect social biases, and (ii) given an adequately informative context, we test whether the model\u2019s biases override a correct answer choice. We find that models often rely on stereotypes when the context is under-informative, meaning the model\u2019s outputs consistently reproduce harmful biases in this setting. Though models are more accurate when the context provides an informative answer, they still rely on stereotypes and average up to 3.4 percentage points higher accuracy when the correct answer aligns with a social bias than when it conflicts, with this difference widening to over 5 points on examples targeting gender for most models tested."
                    },
                    {
                        "title": "Unsupervised Extractive Summarization using Pointwise Mutual Information",
                        "abstract": "Unsupervised approaches to extractive summarization usually rely on a notion of sentence importance defined by the semantic similarity between a sentence and the document. We propose new metrics of relevance and redundancy using pointwise mutual information (PMI) between sentences, which can be easily computed by a pre-trained language model. Intuitively, a relevant sentence allows readers to infer the document content (high PMI with the document), and a redundant sentence can be inferred from the summary (high PMI with the summary). We then develop a greedy sentence selection algorithm to maximize relevance and minimize redundancy of extracted sentences. We show that our method outperforms similarity-based methods on datasets in a range of domains including news, medical journal articles, and personal anecdotes."
                    }
                ]
            },
            "5f9e9636-d44a-49f0-b2ac-ff92b26ef017": {
                "pk": "5f9e9636-d44a-49f0-b2ac-ff92b26ef017",
                "name": "He He",
                "collaborators": [
                    "J. Hansen",
                    "J. Yin",
                    "Lifeng Song",
                    "Qiong Deng",
                    "Saranya Venkatraman",
                    "D. Reitter",
                    "A. Heister",
                    "S. Schael",
                    "R. Barate",
                    "R. Bruneli\u00e8re",
                    "I. Bonis",
                    "D. Decamp",
                    "C. Goy",
                    "S. J\u00e9z\u00e9quel",
                    "J. Lees",
                    "F. Martin",
                    "E. Merle",
                    "M. Minard",
                    "B. Pietrzyk",
                    "B. Trocm\u00b6e",
                    "S. Bravo",
                    "M. Casado",
                    "M. Chmeissani",
                    "J. M. Crespo",
                    "E. Fern\u00e1ndez",
                    "M. Fernandez-Bosman",
                    "Ll. Garrido",
                    "Miguel \u00c1ngel Mart\u00ednez",
                    "A. Pacheco",
                    "H. Ruiz",
                    "A. Colaleo",
                    "D. Creanza",
                    "Nicola De Filippis",
                    "M. Palma",
                    "G. Iaselli",
                    "G. Maggi",
                    "M. Maggi",
                    "S. Nuzzo",
                    "A. Ranieri",
                    "G. Raso",
                    "F. Ruggieri",
                    "G. Selvaggi",
                    "L. Silvestris",
                    "P. Tempesta",
                    "A. Tricomi",
                    "G. Zito",
                    "X. Huang",
                    "J. Lin",
                    "Q. Ouyang",
                    "T. Wang",
                    "Y. Xie",
                    "R. Xu",
                    "S. Xue",
                    "J. Zhang",
                    "Ling Zhang",
                    "Wei Zhao",
                    "D. Abbaneo",
                    "P. Azzurri",
                    "T. Barklow",
                    "O. Buchm\u02dculler",
                    "M. Cattaneo",
                    "F. Cerutti",
                    "B. Clerbaux",
                    "H. Drevermann",
                    "R. Forty",
                    "M. Frank",
                    "F. Gianotti",
                    "J. Harvey",
                    "D. Hutchcroft",
                    "P. Janot",
                    "B. Jost",
                    "M. Kado",
                    "P. Mato",
                    "A. Moutoussi",
                    "F. Ranjard",
                    "L. Rolandi",
                    "D. Schlatter",
                    "G. Sguazzoni",
                    "W. Tejessy",
                    "F. Teubert",
                    "A. Valassi",
                    "I. Videau",
                    "J. Ward",
                    "F. Badaud",
                    "S. Dessagne",
                    "A. Falvard",
                    "D. Fayolle",
                    "P. Gay",
                    "J. Jousset",
                    "B. Michel",
                    "S. Monteil",
                    "D. Pallin",
                    "J. M. Pascolo",
                    "P. Perret",
                    "J. D. Hansen",
                    "P. Hansen",
                    "A. Kraan",
                    "B. Nilsson",
                    "A. Kyriakis",
                    "C. Markou"
                ],
                "domain": [
                    "Material Science",
                    "Natural Language Processing",
                    "Medical Imaging",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "High-temperature mechanical responses of the polyimide at a high strain rate",
                        "abstract": "Polyimide is one of the best super engineering materials in the world at present, and it is required to have high bearing capacity at specific temperature and strain rate because of the application environment. Therefore, the compressive mechanical properties of polyimide at different strain rates (10\u22123 s\u22121\u20134000 s\u22121) and temperatures (20 \u00b0C\u2013350 \u00b0C) were researched. The experimental results show the sensitivity to temperatures and strain rates. However, the strain hardening rates decrease with the increase in temperature, and even the flow stress under dynamic compression is lower than that under quasi-static compression at large strain. In addition, it is found that the cracks show different characteristics under 4000 s\u22121 at different loading temperatures, which are tiny cracks at room temperature and deep and wide cracks due to the coupling of strain rate and thermal at 350 \u00b0C."
                    },
                    {
                        "title": "How do decoding algorithms distribute information in dialogue responses?",
                        "abstract": "Humans tend to follow the Uniform Information Density (UID) principle by distributing information evenly in utterances. We study if decoding algorithms implicitly follow this UID principle, and under what conditions adherence to UID might be desirable for dialogue generation. We generate responses using different decoding algorithms with GPT-2 on the Persona-Chat dataset and collect human judgments on their quality using Amazon Mechanical Turk. We find that (i) surprisingly, model-generated responses follow the UID principle to a greater extent than human responses, and (ii) decoding algorithms that promote UID do not generate higher-quality responses. Instead, when we control for surprisal, non-uniformity of information density correlates with the quality of responses with very low/high surprisal. Our findings indicate that encouraging non-uniform responses is a potential solution to the \u201clikelihood trap\u201d problem (quality degradation in very high-likelihood text). Our dataset containing multiple candidate responses per dialog history along with human-annotated quality ratings is available at: https://huggingface.co/datasets/saranya132/dialog_uid_gpt2."
                    },
                    {
                        "title": "Measurement of the inclusive D*+- production in gamma-gamma collisions at LEP",
                        "abstract": "The inclusive production of D \u2044\u00a7 mesons in two-photon collisions is measured with the ALEPH detector at e + e \u00a1 centre-of-mass energies from 183 GeV to 209 GeV. A total of 360 \u00a7 27 D \u2044\u00a7 meson events were observed from an integrated luminosity of 699 pb \u00a1 1 . Contributions from direct and single-resolved processes are separated using the ratio of the transverse momentum p D \u2044\u00a7 t of the D \u2044\u00a7 to the visible invariant mass W vis of the event. Di\ufb01erential cross sections of D \u2044\u00a7 production as functions of p D \u2044\u00a7 t and the pseudorapidity j \u00b7 D \u2044\u00a7 j are measured in the range 2 GeV =c < p D \u2044\u00a7 t < 12 GeV =c and j \u00b7 D \u2044\u00a7 j < 1 : 5. They are compared to next-to-leading order (NLO) perturbative QCD calculations. The extrapolation of the integrated visible D \u2044\u00a7 cross section to the total charm cross section, based on the Pythia Monte Carlo program, yields (cid:190) (e + e \u00a1 ! e + e \u00a1 c\u201ec) < p s> =197GeV = 731 \u00a7 74 stat \u00a7 47 syst \u00a7 157 extr pb."
                    },
                    {
                        "title": "Leveraging Implicit Feedback from Deployment Data in Dialogue",
                        "abstract": "We study improving social conversational agents by learning from natural dialogue between users and a deployed model, without extra annotations. To implicitly measure the quality of a machine-generated utterance, we leverage signals like user response length, sentiment and reaction of the future human utterances in the collected dialogue episodes. Our experiments use the publicly released deployment data from BlenderBot (Xu et al., 2023). Human evaluation indicates improvements in our new models over baseline responses; however, we find that some proxy signals can lead to more generations with undesirable properties as well. For example, optimizing for conversation length can lead to more controversial or unfriendly generations compared to the baseline, whereas optimizing for positive sentiment or reaction can decrease these behaviors."
                    },
                    {
                        "title": "POS1277\u2005LUNG FUNCTION ASSOCIATED WITH SUBCLINICAL MYOCARDIAL IMPAIRMENT IN SYSTEMIC SCLEROSIS: A CARDIAC MAGNETIC RESONANCE STUDY",
                        "abstract": "Systemic sclerosis (SSc) is an autoimmune disease and microvasculopathy is its major pathophysiological process. Patients with SSc usually have abnormal immune activation, neovascularization, and vascular remodeling. Progressive fibrosis and dysfunction will occur in multiple organs along with the progression of diseases, such as interstitial lung disease and cardiomyopathy [1]. Myocardial impairment is the major cause of death in patients with SSc [2]. Cardiac magnetic resonance (CMR) has been demonstrated to be an ideal non-invasive modality to assess myocardial pathology, particularly quantitative T1 [3]. However, CMR is not applicable for individuals with contraindications to MR examination.This study aimed to investigate the association between lung function and myocardial T1 values using cardiac magnetic resonance (CMR) in SSc patients without cardiovascular symptoms.The SSc patients without cardiovascular symptoms and age- and sex-matched healthy subjects were recruited and underwent CMR examination to acquire CINE and T1 mapping images. Myocardial native T1 values and lung function were evaluated. Spearman\u2019s rank correlations and linear regression analyses were conducted to determine the association between lung function and myocardial native T1 values.Forty SSc patients (mean age 47.7\u00b113.8 years, 35 females) and 13 (mean age 43.2\u00b13.3 years,10 females) healthy subjects were enrolled. SSc patients showed significantly higher native T1 value in myocardium compared to healthy subjects (1305\u00b148.3\u2009ms vs. 1273.7\u00b135.5\u2009ms, P=0.036). Diffusing capacity of lungs for carbon monoxide (DLCO) was significantly associated with myocardial native T1 value before (\u03b2, -0.958; 95%CI, -1.822 to -0.094; P=0.031) and after adjusted for confounding factors (\u03b2, -1.223; 95%CI, -2.155 to -0.291; P=0.012). Moderate-to-severe decrease of DLCO (DLCO<60%) was independently associated with myocardial native T1 value (\u03b2, 33.189; 95%CI, 16.332 to 50.046; P=0.001). Figure 1 presents a typical case about myocardial native T1 value in a SSc patient (DLCO<60%) and an age- and sex-matched healthy subject.Lung function is independently associated with myocardial native T1 values in SSc patients, particularly for moderate-to-severe decrease of DLCO, suggesting that lung function measurements might be a potential indicator for subclinical myocardial impairment in SSc patients. It is well established that chronic inflammation, the major pathophysiological process of SSc in early stage, persistently activates interstitial fibroblasts of multiple organs including lung and heart, leading to the irreversible fibrosis and subsequent decline of function [4].[1]Katsumoto TR, Whitfield ML, Connolly MK. The pathogenesis of systemic sclerosis. Annu Rev Pathol 2011;6:509-37.[2]Kom\u00f3csi A, Vorobcsuk A, Faludi R, et al. The impact of cardiopulmonary manifestations on the mortality of SSc: A systematic review and meta-analysis of observational studies. Rheumatology (Oxford) 2012;51:1027-36.[3]Agoston-Coldea L, Zlibut A, Revnic R, Florea M, Muntean L. Current advances in cardiac magnetic resonance imaging in systemic sclerosis. Eur Rev Med Pharmacol Sci 2021;25:3718-36.[4]Asano Y. The pathogenesis of systemic sclerosis: An understanding based on a common pathologic cascade across multiple organs and additional Organ-Specific pathologies. J Clin Med 2020;9:2687.Figure 1: A 59 years old female SSc patient with declined DLCO (54%) and elevated myocardial native T1 (1327ms) (a-c); Images d-f showed the native T1 maps of myocardium of a 57 years old female healthy subject with the mean T1 value of 1255ms. The MR images of T1 maps of myocardium were taken from short axis slices of left ventricle. nT1, native T1 value.Figure 1 Native T1 value measured by cardiac magnetic resonance in a SSc patient and a healthy subject.NIL.None Declared."
                    },
                    {
                        "title": "Evaluation of a new dengue 3 controlled human infection model for use in the evaluation of candidate dengue vaccines.",
                        "abstract": "All four serotypes of dengue virus (DENV) cause the full spectrum of disease. Therefore, vaccines must protect against all serotypes. To evaluate candidate vaccines, a human challenge model of dengue serotype 3 (rDEN30{triangleup}30) was developed. All challenge virus recipients safely met the primary endpoint of viremia and secondary endpoints of rash and seroconversion to DENV-3."
                    },
                    {
                        "title": "Efficient Shapley Values Estimation by Amortization for Text Classification",
                        "abstract": "Despite the popularity of Shapley Values in explaining neural text classification models, computing them is prohibitive for large pretrained models due to a large number of model evaluations. In practice, Shapley Values are often estimated with a small number of stochastic model evaluations. However, we show that the estimated Shapley Values are sensitive to random seed choices \u2013 the top-ranked features often have little overlap across different seeds, especially on examples with longer input texts. This can only be mitigated by aggregating thousands of model evaluations, which on the other hand, induces substantial computational overheads. To mitigate the trade-off between stability and efficiency, we develop an amortized model that directly predicts each input feature\u2019s Shapley Value without additional model evaluations. It is trained on a set of examples whose Shapley Values are estimated from a large number of model evaluations to ensure stability. Experimental results on two text classification datasets demonstrate that our amortized model estimates Shapley Values accurately with up to 60 times speedup compared to traditional methods. Further, our model does not suffer from stability issues as inference is deterministic. We release our code at https://github.com/yangalan123/Amortized-Interpretability."
                    }
                ]
            }
        }
    },
    "2406.18664": {
        "paper_data": {
            "title": "Evaluating Copyright Takedown Methods for Language Models",
            "url": "http://arxiv.org/abs/2406.18664v4",
            "arxiv_id": "2406.18664",
            "authors": [
                "Boyi Wei",
                "Weijia Shi",
                "Yangsibo Huang",
                "Noah A. Smith",
                "Chiyuan Zhang",
                "Luke Zettlemoyer",
                "Kai Li",
                "Peter Henderson"
            ],
            "abstract": "Language models (LMs) derive their capabilities from extensive training on diverse data, including potentially copyrighted material. These models can memorize and generate content similar to their training data, posing potential concerns. Therefore, model creators are motivated to develop mitigation methods that prevent generating protected content. We term this procedure as copyright takedowns for LMs, noting the conceptual similarity to (but legal distinction from) the DMCA takedown This paper introduces the first evaluation of the feasibility and side effects of copyright takedowns for LMs. We propose CoTaEval, an evaluation framework to assess the effectiveness of copyright takedown methods, the impact on the model's ability to retain uncopyrightable factual knowledge from the training data whose recitation is embargoed, and how well the model maintains its general utility and efficiency. We examine several strategies, including adding system prompts, decoding-time filtering interventions, and unlearning approaches. Our findings indicate that no tested method excels across all metrics, showing significant room for research in this unique problem setting and indicating potential unresolved challenges for live policy proposals.",
            "introduction": "   1 Introduction  Large language models (LLMs) are trained on massive amounts of data, largely drawn from across the web\u00a0(Bommasani et\u00a0al., 2021). In most countries, explicit policies regarding training on copyrighted material have been lagging behind the development of LLM training techniques. In the US, model creators often cite the fair use doctrine, a legal defense (developed before the LLM era) that allows the use of copyrighted data without permission under certain circumstances\u00a0(Lemley & Casey, 2021). Nonetheless, litigation has swept the United States and abroad as copyright owners challenge the use of their content for training and deploying foundation models\u2014e.g., \u00a0Tremblay v. OpenAI, Inc., (2023); Kadrey v. Meta Platforms, Inc. (2023). Generally, there is less legal risk, and a more likely fair use defense, if models do not output content substantially similar to the training data \u00a0(Henderson et\u00a0al., 2023; Sag, 2023; Lee et\u00a0al., 2024).   Thus, model creators increasingly seek to use guardrails that prevent their models from regurgitating content. An example is Github Copilot, a code completion model, provides a duplication detection filter. When turned on, \u201cGitHub Copilot checks code completion suggestions with their surrounding code of about 150 characters against public code on GitHub. If there is a match, or a near match, the suggestion is not shown\u201d\u00a0(GitHub, 2023b). OpenAI\u2019s ChatGPT appears to have a similar filter for some types of content, as well as training the model to reject requests that may ask for infringing outputs\u00a0(Henderson et\u00a0al., 2023). Such post-training mitigation strategies will be an essential aspect of model deployments. Even if model creators possess licenses and filter pre-training data, they may unwittingly include copyrighted material that the model could regurgitate. For example, consider if a company licenses Reddit data for training. There is no guarantee that Reddit posts are not themselves infringing, and tracing the provenance of every piece of content is nearly impossible. Therefore, model deployers require a strategy to prevent models from outputting content that are too similar to specific copyrighted data, which they may only notice after training is complete. Noting the conceptual similarity to DMCA Takedown, we refer to this procedure as a copyright takedown for LMs, or simply takedown when there is no ambiguity. Note unlike a DMCA Takedown, copyright takedown is not a formally defined legal term, and in this paper specifically refers to the post-training procedures applied to prevent an LM from generating texts that are too similar to specific contents. Legal scholars suggest that a takedown mechanism may be a necessary and effective part of future policymaking\u00a0(Henderson et\u00a0al., 2023; Pasquale & Sun, 2024; Lee et\u00a0al., 2024). Yet, a key question remains: Can \u201ctakedown\u201d of copyrighted content be operationalized in the context of large language models?   Figure 1:  Effective takedown methods should prevent models from generating text matching the blocklisted content (low similarity) while preserving uncopyrightable facts and fair use information (high utility).    This paper introduces the first evaluation of the feasibility and side effects of copyright takedowns in language models. Our benchmark, CoTaEval, considers potential regurgitation of blocklisted content due to both memorized content and content retrieved through retrieval-augmented generation (RAG,  Lewis et\u00a0al., 2020; Shi et\u00a0al., 2024c) or tool-based approaches (Thoppilan et\u00a0al., 2022).111Both",
            "references": [
                {
                    "title": "Gemma 2: Improving Open Language Models at a Practical Size",
                    "abstract": "In this work, we introduce Gemma 2, a new addition to the Gemma family of lightweight, state-of-the-art open models, ranging in scale from 2 billion to 27 billion parameters. In this new version, we apply several known technical modifications to the Transformer architecture, such as interleaving local-global attentions (Beltagy et al., 2020a) and group-query attention (Ainslie et al., 2023). We also train the 2B and 9B models with knowledge distillation (Hinton et al., 2015) instead of next token prediction. The resulting models deliver the best performance for their size, and even offer competitive alternatives to models that are 2-3 times bigger. We release all our models to the community."
                },
                {
                    "title": "CopyBench: Measuring Literal and Non-Literal Reproduction of Copyright-Protected Text in Language Model Generation",
                    "abstract": "Evaluating the degree of reproduction of copyright-protected content by language models (LMs) is of significant interest to the AI and legal communities. Although both literal and non-literal similarities are considered by courts when assessing the degree of reproduction, prior research has focused only on literal similarities. To bridge this gap, we introduce CopyBench, a benchmark designed to measure both literal and non-literal copying in LM generations. Using copyrighted fiction books as text sources, we provide automatic evaluation protocols to assess literal and non-literal copying, balanced against the model utility in terms of the ability to recall facts from the copyrighted works and generate fluent completions. We find that, although literal copying is relatively rare, two types of non-literal copying -- event copying and character copying -- occur even in models as small as 7B parameters. Larger models demonstrate significantly more copying, with literal copying rates increasing from 0.2\\% to 10.5\\% and non-literal copying from 2.3\\% to 5.9\\% when comparing Llama3-8B and 70B models, respectively. We further evaluate the effectiveness of current strategies for mitigating copying and show that (1) training-time alignment can reduce literal copying but may increase non-literal copying, and (2) current inference-time mitigation methods primarily reduce literal but not non-literal copying."
                },
                {
                    "title": "MUSE: Machine Unlearning Six-Way Evaluation for Language Models",
                    "abstract": "Language models (LMs) are trained on vast amounts of text data, which may include private and copyrighted content. Data owners may request the removal of their data from a trained model due to privacy or copyright concerns. However, exactly unlearning only these datapoints (i.e., retraining with the data removed) is intractable in modern-day models. This has led to the development of many approximate unlearning algorithms. The evaluation of the efficacy of these algorithms has traditionally been narrow in scope, failing to precisely quantify the success and practicality of the algorithm from the perspectives of both the model deployers and the data owners. We address this issue by proposing MUSE, a comprehensive machine unlearning evaluation benchmark that enumerates six diverse desirable properties for unlearned models: (1) no verbatim memorization, (2) no knowledge memorization, (3) no privacy leakage, (4) utility preservation on data not intended for removal, (5) scalability with respect to the size of removal requests, and (6) sustainability over sequential unlearning requests. Using these criteria, we benchmark how effectively eight popular unlearning algorithms on 7B-parameter LMs can unlearn Harry Potter books and news articles. Our results demonstrate that most algorithms can prevent verbatim memorization and knowledge memorization to varying degrees, but only one algorithm does not lead to severe privacy leakage. Furthermore, existing algorithms fail to meet deployer's expectations because they often degrade general model utility and also cannot sustainably accommodate successive unlearning requests or large-scale content removal. Our findings identify key issues with the practicality of existing unlearning algorithms on language models, and we release our benchmark to facilitate further evaluations: muse-bench.github.io"
                },
                {
                    "title": "Fantastic Copyrighted Beasts and How (Not) to Generate Them",
                    "abstract": "Recent studies show that image and video generation models can be prompted to reproduce copyrighted content from their training data, raising serious legal concerns around copyright infringement. Copyrighted characters, in particular, pose a difficult challenge for image generation services, with at least one lawsuit already awarding damages based on the generation of these characters. Yet, little research has empirically examined this issue. We conduct a systematic evaluation to fill this gap. First, we build CopyCat, an evaluation suite consisting of diverse copyrighted characters and a novel evaluation pipeline. Our evaluation considers both the detection of similarity to copyrighted characters and generated image's consistency with user input. Our evaluation systematically shows that both image and video generation models can still generate characters even if characters' names are not explicitly mentioned in the prompt, sometimes with only two generic keywords (e.g., prompting with\"videogame, plumber\"consistently generates Nintendo's Mario character). We then introduce techniques to semi-automatically identify such keywords or descriptions that trigger character generation. Using our evaluation suite, we study runtime mitigation strategies, including both existing methods and new strategies we propose. Our findings reveal that commonly employed strategies, such as prompt rewriting in the DALL-E system, are not sufficient as standalone guardrails. These strategies must be coupled with other approaches, like negative prompting, to effectively reduce the unintended generation of copyrighted characters. Our work provides empirical grounding to the discussion of copyright mitigation strategies and offers actionable insights for model deployers actively implementing them."
                },
                {
                    "title": "SHIELD: Evaluation and Defense Strategies for Copyright Compliance in LLM Text Generation",
                    "abstract": "Large Language Models (LLMs) have transformed machine learning but raised significant legal concerns due to their potential to produce text that infringes on copyrights, resulting in several high-profile lawsuits. The legal landscape is struggling to keep pace with these rapid advancements, with ongoing debates about whether generated text might plagiarize copyrighted materials. Current LLMs may infringe on copyrights or overly restrict non-copyrighted texts, leading to these challenges: (i) the need for a comprehensive evaluation benchmark to assess copyright compliance from multiple aspects; (ii) evaluating robustness against safeguard bypassing attacks; and (iii) developing effective defense targeted against the generation of copyrighted text. To tackle these challenges, we introduce a curated dataset to evaluate methods, test attack strategies, and propose lightweight, real-time defense to prevent the generation of copyrighted text, ensuring the safe and lawful use of LLMs. Our experiments demonstrate that current LLMs frequently output copyrighted text, and that jailbreaking attacks can significantly increase the volume of copyrighted output. Our proposed defense mechanism significantly reduces the volume of copyrighted text generated by LLMs by effectively refusing malicious requests. Code is publicly available at https://github.com/xz-liu/SHIELD"
                },
                {
                    "title": "Be like a Goldfish, Don't Memorize! Mitigating Memorization in Generative LLMs",
                    "abstract": "Large language models can memorize and repeat their training data, causing privacy and copyright risks. To mitigate memorization, we introduce a subtle modification to the next-token training objective that we call the goldfish loss. During training, a randomly sampled subset of tokens are excluded from the loss computation. These dropped tokens are not memorized by the model, which prevents verbatim reproduction of a complete chain of tokens from the training set. We run extensive experiments training billion-scale Llama-2 models, both pre-trained and trained from scratch, and demonstrate significant reductions in extractable memorization with little to no impact on downstream benchmarks."
                },
                {
                    "title": "Safety Alignment Should Be Made More Than Just a Few Tokens Deep",
                    "abstract": "The safety alignment of current Large Language Models (LLMs) is vulnerable. Relatively simple attacks, or even benign fine-tuning, can jailbreak aligned models. We argue that many of these vulnerabilities are related to a shared underlying issue: safety alignment can take shortcuts, wherein the alignment adapts a model's generative distribution primarily over only its very first few output tokens. We refer to this issue as shallow safety alignment. In this paper, we present case studies to explain why shallow safety alignment can exist and provide evidence that current aligned LLMs are subject to this issue. We also show how these findings help explain multiple recently discovered vulnerabilities in LLMs, including the susceptibility to adversarial suffix attacks, prefilling attacks, decoding parameter attacks, and fine-tuning attacks. Importantly, we discuss how this consolidated notion of shallow safety alignment sheds light on promising research directions for mitigating these vulnerabilities. For instance, we show that deepening the safety alignment beyond just the first few tokens can often meaningfully improve robustness against some common exploits. Finally, we design a regularized finetuning objective that makes the safety alignment more persistent against fine-tuning attacks by constraining updates on initial tokens. Overall, we advocate that future safety alignment should be made more than just a few tokens deep."
                },
                {
                    "title": "Automatic Jailbreaking of the Text-to-Image Generative AI Systems",
                    "abstract": "Recent AI systems have shown extremely powerful performance, even surpassing human performance, on various tasks such as information retrieval, language generation, and image generation based on large language models (LLMs). At the same time, there are diverse safety risks that can cause the generation of malicious contents by circumventing the alignment in LLMs, which are often referred to as jailbreaking. However, most of the previous works only focused on the text-based jailbreaking in LLMs, and the jailbreaking of the text-to-image (T2I) generation system has been relatively overlooked. In this paper, we first evaluate the safety of the commercial T2I generation systems, such as ChatGPT, Copilot, and Gemini, on copyright infringement with naive prompts. From this empirical study, we find that Copilot and Gemini block only 12% and 17% of the attacks with naive prompts, respectively, while ChatGPT blocks 84% of them. Then, we further propose a stronger automated jailbreaking pipeline for T2I generation systems, which produces prompts that bypass their safety guards. Our automated jailbreaking framework leverages an LLM optimizer to generate prompts to maximize degree of violation from the generated images without any weight updates or gradient computation. Surprisingly, our simple yet effective approach successfully jailbreaks the ChatGPT with 11.0% block rate, making it generate copyrighted contents in 76% of the time. Finally, we explore various defense strategies, such as post-generation filtering and machine unlearning techniques, but found that they were inadequate, which suggests the necessity of stronger defense mechanisms."
                },
                {
                    "title": "Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning",
                    "abstract": "Large Language Models (LLMs) often memorize sensitive, private, or copyrighted data during pre-training. LLM unlearning aims to eliminate the influence of undesirable data from the pre-trained model while preserving the model's utilities on other tasks. Several practical methods have recently been proposed for LLM unlearning, mostly based on gradient ascent (GA) on the loss of undesirable data. However, on certain unlearning tasks, these methods either fail to effectively unlearn the target data or suffer from catastrophic collapse -- a drastic degradation of the model's utilities. In this paper, we propose Negative Preference Optimization (NPO), a simple alignment-inspired method that could efficiently and effectively unlearn a target dataset. We theoretically show that the progression toward catastrophic collapse by minimizing the NPO loss is exponentially slower than GA. Through experiments on synthetic data and the benchmark TOFU dataset, we demonstrate that NPO-based methods achieve a better balance between unlearning the undesirable data and maintaining the model's utilities. We also observe that NPO-based methods generate more sensible outputs than GA-based methods, whose outputs are often gibberish. Remarkably, on TOFU, NPO-based methods are the first to achieve reasonable unlearning results in forgetting 50% (or more) of the training data, whereas existing methods already struggle with forgetting 10% of training data."
                },
                {
                    "title": "The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning",
                    "abstract": "The White House Executive Order on Artificial Intelligence highlights the risks of large language models (LLMs) empowering malicious actors in developing biological, cyber, and chemical weapons. To measure these risks of malicious use, government institutions and major AI labs are developing evaluations for hazardous capabilities in LLMs. However, current evaluations are private, preventing further research into mitigating risk. Furthermore, they focus on only a few, highly specific pathways for malicious use. To fill these gaps, we publicly release the Weapons of Mass Destruction Proxy (WMDP) benchmark, a dataset of 3,668 multiple-choice questions that serve as a proxy measurement of hazardous knowledge in biosecurity, cybersecurity, and chemical security. WMDP was developed by a consortium of academics and technical consultants, and was stringently filtered to eliminate sensitive information prior to public release. WMDP serves two roles: first, as an evaluation for hazardous knowledge in LLMs, and second, as a benchmark for unlearning methods to remove such hazardous knowledge. To guide progress on unlearning, we develop RMU, a state-of-the-art unlearning method based on controlling model representations. RMU reduces model performance on WMDP while maintaining general capabilities in areas such as biology and computer science, suggesting that unlearning may be a concrete path towards reducing malicious use from LLMs. We release our benchmark and code publicly at https://wmdp.ai"
                },
                {
                    "title": "Proving membership in LLM pretraining data via data watermarks",
                    "abstract": "Detecting whether copyright holders' works were used in LLM pretraining is poised to be an important problem. This work proposes using data watermarks to enable principled detection with only black-box model access, provided that the rightholder contributed multiple training documents and watermarked them before public release. By applying a randomly sampled data watermark, detection can be framed as hypothesis testing, which provides guarantees on the false detection rate. We study two watermarks: one that inserts random sequences, and another that randomly substitutes characters with Unicode lookalikes. We first show how three aspects of watermark design -- watermark length, number of duplications, and interference -- affect the power of the hypothesis test. Next, we study how a watermark's detection strength changes under model and dataset scaling: while increasing the dataset size decreases the strength of the watermark, watermarks remain strong if the model size also increases. Finally, we view SHA hashes as natural watermarks and show that we can robustly detect hashes from BLOOM-176B's training data, as long as they occurred at least 90 times. Together, our results point towards a promising future for data watermarks in real world use."
                },
                {
                    "title": "Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modifications",
                    "abstract": "Large language models (LLMs) show inherent brittleness in their safety mechanisms, as evidenced by their susceptibility to jailbreaking and even non-malicious fine-tuning. This study explores this brittleness of safety alignment by leveraging pruning and low-rank modifications. We develop methods to identify critical regions that are vital for safety guardrails, and that are disentangled from utility-relevant regions at both the neuron and rank levels. Surprisingly, the isolated regions we find are sparse, comprising about $3\\%$ at the parameter level and $2.5\\%$ at the rank level. Removing these regions compromises safety without significantly impacting utility, corroborating the inherent brittleness of the model's safety mechanisms. Moreover, we show that LLMs remain vulnerable to low-cost fine-tuning attacks even when modifications to the safety-critical regions are restricted. These findings underscore the urgent need for more robust safety strategies in LLMs."
                },
                {
                    "title": "TOFU: A Task of Fictitious Unlearning for LLMs",
                    "abstract": "Large language models trained on massive corpora of data from the web can memorize and reproduce sensitive or private data raising both legal and ethical concerns. Unlearning, or tuning models to forget information present in their training data, provides us with a way to protect private data after training. Although several methods exist for such unlearning, it is unclear to what extent they result in models equivalent to those where the data to be forgotten was never learned in the first place. To address this challenge, we present TOFU, a Task of Fictitious Unlearning, as a benchmark aimed at helping deepen our understanding of unlearning. We offer a dataset of 200 diverse synthetic author profiles, each consisting of 20 question-answer pairs, and a subset of these profiles called the forget set that serves as the target for unlearning. We compile a suite of metrics that work together to provide a holistic picture of unlearning efficacy. Finally, we provide a set of baseline results from existing unlearning algorithms. Importantly, none of the baselines we consider show effective unlearning motivating continued efforts to develop approaches for unlearning that effectively tune models so that they truly behave as if they were never trained on the forget data at all."
                },
                {
                    "title": "Scalable Extraction of Training Data from (Production) Language Models",
                    "abstract": "This paper studies extractable memorization: training data that an adversary can efficiently extract by querying a machine learning model without prior knowledge of the training dataset. We show an adversary can extract gigabytes of training data from open-source language models like Pythia or GPT-Neo, semi-open models like LLaMA or Falcon, and closed models like ChatGPT. Existing techniques from the literature suffice to attack unaligned models; in order to attack the aligned ChatGPT, we develop a new divergence attack that causes the model to diverge from its chatbot-style generations and emit training data at a rate 150x higher than when behaving properly. Our methods show practical attacks can recover far more data than previously thought, and reveal that current alignment techniques do not eliminate memorization."
                },
                {
                    "title": "What's In My Big Data?",
                    "abstract": "Large text corpora are the backbone of language models. However, we have a limited understanding of the content of these corpora, including general statistics, quality, social factors, and inclusion of evaluation data (contamination). In this work, we propose What's In My Big Data? (WIMBD), a platform and a set of sixteen analyses that allow us to reveal and compare the contents of large text corpora. WIMBD builds on two basic capabilities -- count and search -- at scale, which allows us to analyze more than 35 terabytes on a standard compute node. We apply WIMBD to ten different corpora used to train popular language models, including C4, The Pile, and RedPajama. Our analysis uncovers several surprising and previously undocumented findings about these corpora, including the high prevalence of duplicate, synthetic, and low-quality content, personally identifiable information, toxic language, and benchmark contamination. For instance, we find that about 50% of the documents in RedPajama and LAION-2B-en are duplicates. In addition, several datasets used for benchmarking models trained on such corpora are contaminated with respect to important benchmarks, including the Winograd Schema Challenge and parts of GLUE and SuperGLUE. We open-source WIMBD's code and artifacts to provide a standard set of evaluations for new text-based corpora and to encourage more analyses and transparency around them."
                },
                {
                    "title": "DEPN: Detecting and Editing Privacy Neurons in Pretrained Language Models",
                    "abstract": "Large language models pretrained on a huge amount of data capture rich knowledge and information in the training data. The ability of data memorization and regurgitation in pretrained language models, revealed in previous studies, brings the risk of data leakage. In order to effectively reduce these risks, we propose a framework DEPN to Detect and Edit Privacy Neurons in pretrained language models, partially inspired by knowledge neurons and model editing. In DEPN, we introduce a novel method, termed as privacy neuron detector, to locate neurons associated with private information, and then edit these detected privacy neurons by setting their activations to zero. Furthermore, we propose a privacy neuron aggregator dememorize private information in a batch processing manner. Experimental results show that our method can significantly and efficiently reduce the exposure of private data leakage without deteriorating the performance of the model. Additionally, we empirically demonstrate the relationship between model memorization and privacy neurons, from multiple perspectives, including model size, training time, prompts, privacy neuron distribution, illustrating the robustness of our approach."
                },
                {
                    "title": "Detecting Pretraining Data from Large Language Models",
                    "abstract": "Although large language models (LLMs) are widely deployed, the data used to train them is rarely disclosed. Given the incredible scale of this data, up to trillions of tokens, it is all but certain that it includes potentially problematic text such as copyrighted materials, personally identifiable information, and test data for widely reported reference benchmarks. However, we currently have no way to know which data of these types is included or in what proportions. In this paper, we study the pretraining data detection problem: given a piece of text and black-box access to an LLM without knowing the pretraining data, can we determine if the model was trained on the provided text? To facilitate this study, we introduce a dynamic benchmark WIKIMIA that uses data created before and after model training to support gold truth detection. We also introduce a new detection method Min-K% Prob based on a simple hypothesis: an unseen example is likely to contain a few outlier words with low probabilities under the LLM, while a seen example is less likely to have words with such low probabilities. Min-K% Prob can be applied without any knowledge about the pretraining corpus or any additional training, departing from previous detection methods that require training a reference model on data that is similar to the pretraining data. Moreover, our experiments demonstrate that Min-K% Prob achieves a 7.4% improvement on WIKIMIA over these previous methods. We apply Min-K% Prob to two real-world scenarios, copyrighted book detection, and contaminated downstream example detection, and find it a consistently effective solution."
                },
                {
                    "title": "The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI",
                    "abstract": "The race to train language models on vast, diverse, and inconsistently documented datasets has raised pressing concerns about the legal and ethical risks for practitioners. To remedy these practices threatening data transparency and understanding, we convene a multi-disciplinary effort between legal and machine learning experts to systematically audit and trace 1800+ text datasets. We develop tools and standards to trace the lineage of these datasets, from their source, creators, series of license conditions, properties, and subsequent use. Our landscape analysis highlights the sharp divides in composition and focus of commercially open vs closed datasets, with closed datasets monopolizing important categories: lower resource languages, more creative tasks, richer topic variety, newer and more synthetic training data. This points to a deepening divide in the types of data that are made available under different license conditions, and heightened implications for jurisdictional legal interpretations of copyright and fair use. We also observe frequent miscategorization of licenses on widely used dataset hosting sites, with license omission of 72%+ and error rates of 50%+. This points to a crisis in misattribution and informed use of the most popular datasets driving many recent breakthroughs. As a contribution to ongoing improvements in dataset transparency and responsible use, we release our entire audit, with an interactive UI, the Data Provenance Explorer, which allows practitioners to trace and filter on data provenance for the most popular open source finetuning data collections: www.dataprovenance.org."
                },
                {
                    "title": "Copyright Violations and Large Language Models",
                    "abstract": "Language models may memorize more than just facts, including entire chunks of texts seen during training. Fair use exemptions to copyright laws typically allow for limited use of copyrighted material without permission from the copyright holder, but typically for extraction of information from copyrighted materials, rather than {\\em verbatim} reproduction. This work explores the issue of copyright violations and large language models through the lens of verbatim memorization, focusing on possible redistribution of copyrighted text. We present experiments with a range of language models over a collection of popular books and coding problems, providing a conservative characterization of the extent to which language models can redistribute these materials. Overall, this research highlights the need for further examination and the potential impact on future developments in natural language processing to ensure adherence to copyright regulations. Code is at \\url{https://github.com/coastalcph/CopyrightLLMs}."
                },
                {
                    "title": "The Foundation Model Transparency Index",
                    "abstract": "Foundation models have rapidly permeated society, catalyzing a wave of generative AI applications spanning enterprise and consumer-facing contexts. While the societal impact of foundation models is growing, transparency is on the decline, mirroring the opacity that has plagued past digital technologies (e.g. social media). Reversing this trend is essential: transparency is a vital precondition for public accountability, scientific innovation, and effective governance. To assess the transparency of the foundation model ecosystem and help improve transparency over time, we introduce the Foundation Model Transparency Index. The Foundation Model Transparency Index specifies 100 fine-grained indicators that comprehensively codify transparency for foundation models, spanning the upstream resources used to build a foundation model (e.g data, labor, compute), details about the model itself (e.g. size, capabilities, risks), and the downstream use (e.g. distribution channels, usage policies, affected geographies). We score 10 major foundation model developers (e.g. OpenAI, Google, Meta) against the 100 indicators to assess their transparency. To facilitate and standardize assessment, we score developers in relation to their practices for their flagship foundation model (e.g. GPT-4 for OpenAI, PaLM 2 for Google, Llama 2 for Meta). We present 10 top-level findings about the foundation model ecosystem: for example, no developer currently discloses significant information about the downstream impact of its flagship model, such as the number of users, affected market sectors, or how users can seek redress for harm. Overall, the Foundation Model Transparency Index establishes the level of transparency today to drive progress on foundation model governance via industry standards and regulatory intervention."
                },
                {
                    "title": "Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection",
                    "abstract": "Despite their remarkable capabilities, large language models (LLMs) often produce responses containing factual inaccuracies due to their sole reliance on the parametric knowledge they encapsulate. Retrieval-Augmented Generation (RAG), an ad hoc approach that augments LMs with retrieval of relevant knowledge, decreases such issues. However, indiscriminately retrieving and incorporating a fixed number of retrieved passages, regardless of whether retrieval is necessary, or passages are relevant, diminishes LM versatility or can lead to unhelpful response generation. We introduce a new framework called Self-Reflective Retrieval-Augmented Generation (Self-RAG) that enhances an LM's quality and factuality through retrieval and self-reflection. Our framework trains a single arbitrary LM that adaptively retrieves passages on-demand, and generates and reflects on retrieved passages and its own generations using special tokens, called reflection tokens. Generating reflection tokens makes the LM controllable during the inference phase, enabling it to tailor its behavior to diverse task requirements. Experiments show that Self-RAG (7B and 13B parameters) significantly outperforms state-of-the-art LLMs and retrieval-augmented models on a diverse set of tasks. Specifically, Self-RAG outperforms ChatGPT and retrieval-augmented Llama2-chat on Open-domain QA, reasoning and fact verification tasks, and it shows significant gains in improving factuality and citation accuracy for long-form generations relative to these models."
                },
                {
                    "title": "In-Context Unlearning: Language Models as Few Shot Unlearners",
                    "abstract": "Machine unlearning, the study of efficiently removing the impact of specific training instances on a model, has garnered increased attention in recent years due to regulatory guidelines such as the \\emph{Right to be Forgotten}. Achieving precise unlearning typically involves fully retraining the model and is computationally infeasible in case of very large models such as Large Language Models (LLMs). To this end, recent work has proposed several algorithms which approximate the removal of training data without retraining the model. These algorithms crucially rely on access to the model parameters in order to update them, an assumption that may not hold in practice due to computational constraints or having only query access to the LLMs. In this work, we propose a new class of unlearning methods for LLMs called ``In-Context Unlearning.'' This method unlearns instances from the model by simply providing specific kinds of inputs in context, without the need to update model parameters. To unlearn specific training instances, we present these instances to the LLMs at inference time along with labels that differ from their ground truth. Our experimental results demonstrate that in-context unlearning performs on par with, or in some cases outperforms other state-of-the-art methods that require access to model parameters, effectively removing the influence of specific instances on the model while preserving test accuracy."
                },
                {
                    "title": "Who's Harry Potter? Approximate Unlearning in LLMs",
                    "abstract": "Large language models (LLMs) are trained on massive internet corpora that often contain copyrighted content. This poses legal and ethical challenges for the developers and users of these models, as well as the original authors and publishers. In this paper, we propose a novel technique for unlearning a subset of the training data from a LLM, without having to retrain it from scratch. We evaluate our technique on the task of unlearning the Harry Potter books from the Llama2-7b model (a generative language model recently open-sourced by Meta). While the model took over 184K GPU-hours to pretrain, we show that in about 1 GPU hour of finetuning, we effectively erase the model's ability to generate or recall Harry Potter-related content, while its performance on common benchmarks (such as Winogrande, Hellaswag, arc, boolq and piqa) remains almost unaffected. We make our fine-tuned model publicly available on HuggingFace for community evaluation. To the best of our knowledge, this is the first paper to present an effective technique for unlearning in generative language models. Our technique consists of three main components: First, we use a reinforced model that is further trained on the target data to identify the tokens that are most related to the unlearning target, by comparing its logits with those of a baseline model. Second, we replace idiosyncratic expressions in the target data with generic counterparts, and leverage the model's own predictions to generate alternative labels for every token. These labels aim to approximate the next-token predictions of a model that has not been trained on the target data. Third, we finetune the model on these alternative labels, which effectively erases the original text from the model's memory whenever it is prompted with its context."
                },
                {
                    "title": "RA-DIT: Retrieval-Augmented Dual Instruction Tuning",
                    "abstract": "Retrieval-augmented language models (RALMs) improve performance by accessing long-tail and up-to-date knowledge from external data stores, but are challenging to build. Existing approaches require either expensive retrieval-specific modifications to LM pre-training or use post-hoc integration of the data store that leads to suboptimal performance. We introduce Retrieval-Augmented Dual Instruction Tuning (RA-DIT), a lightweight fine-tuning methodology that provides a third option by retrofitting any LLM with retrieval capabilities. Our approach operates in two distinct fine-tuning steps: (1) one updates a pre-trained LM to better use retrieved information, while (2) the other updates the retriever to return more relevant results, as preferred by the LM. By fine-tuning over tasks that require both knowledge utilization and contextual awareness, we demonstrate that each stage yields significant performance improvements, and using both leads to additional gains. Our best model, RA-DIT 65B, achieves state-of-the-art performance across a range of knowledge-intensive zero- and few-shot learning benchmarks, significantly outperforming existing in-context RALM approaches by up to +8.9% in 0-shot setting and +1.4% in 5-shot setting on average."
                },
                {
                    "title": "Can Sensitive Information Be Deleted From LLMs? Objectives for Defending Against Extraction Attacks",
                    "abstract": "Pretrained language models sometimes possess knowledge that we do not wish them to, including memorized personal information and knowledge that could be used to harm people. They can also output toxic or harmful text. To mitigate these safety and informational issues, we propose an attack-and-defense framework for studying the task of deleting sensitive information directly from model weights. We study direct edits to model weights because (1) this approach should guarantee that particular deleted information is never extracted by future prompt attacks, and (2) it should protect against whitebox attacks, which is necessary for making claims about safety/privacy in a setting where publicly available model weights could be used to elicit sensitive information. Our threat model assumes that an attack succeeds if the answer to a sensitive question is located among a set of B generated candidates, based on scenarios where the information would be insecure if the answer is among B candidates. Experimentally, we show that even state-of-the-art model editing methods such as ROME struggle to truly delete factual information from models like GPT-J, as our whitebox and blackbox attacks can recover\"deleted\"information from an edited model 38% of the time. These attacks leverage two key observations: (1) that traces of deleted information can be found in intermediate model hidden states, and (2) that applying an editing method for one question may not delete information across rephrased versions of the question. Finally, we provide new defense methods that protect against some extraction attacks, but we do not find a single universally effective defense method. Our results suggest that truly deleting sensitive information is a tractable but difficult problem, since even relatively low attack success rates have potentially severe societal implications for real-world deployment of language models."
                },
                {
                    "title": "Talkin' 'Bout AI Generation: Copyright and the Generative-AI Supply Chain (The Short Version)",
                    "abstract": "\"Does generative AI infringe copyright?\" is an urgent question. It is also a difficult question, for two reasons. First, \"generative AI\" is not just one product from one company. It is a catch-all name for a massive ecosystem of loosely related technologies. These systems behave differently and raise different legal issues. Second, copyright law is notoriously complicated, and generative-AI systems manage to touch on a great many corners of it. They raise issues of authorship, similarity, direct and indirect liability, and fair use, among much else. These issues cannot be analyzed in isolation, because there are connections everywhere. We aim to bring order to the chaos. To do so, we introduce the generative-AI supply chain: an interconnected set of stages that transform training data into generations. The supply chain reveals all of the places at which companies and users make choices that have copyright consequences. It enables us to trace the effects of upstream technical designs on downstream uses, and to assess who in these complicated sociotechnical systems bears responsibility for infringement when it happens. Because we engage so closely with the technology of generative AI, we are able to shed more light on the copyright questions. We identify the key decisions that courts will need to make as they grapple with these issues, and point out the consequences that would likely flow from different liability regimes. This article is a much-abbreviated version of a forthcoming law review article at The Journal of the Copyright Society."
                },
                {
                    "title": "How to Protect Copyright Data in Optimization of Large Language Models?",
                    "abstract": "Large language models (LLMs) and generative AI have played a transformative role in computer research and applications. Controversy has arisen as to whether these models output copyrighted data, which can occur if the data the models are trained on is copyrighted. LLMs are built on the transformer neural network architecture, which in turn relies on a mathematical computation called Attention that uses the softmax function.\n\nIn this paper, we observe that large language model training and optimization can be seen as a softmax regression problem. We then establish a method of efficiently performing softmax regression, in a way that prevents the regression function from generating copyright data. This establishes a theoretical method of training large language models in a way that avoids generating copyright data."
                },
                {
                    "title": "SILO Language Models: Isolating Legal Risk In a Nonparametric Datastore",
                    "abstract": "The legality of training language models (LMs) on copyrighted or otherwise restricted data is under intense debate. However, as we show, model performance significantly degrades if trained only on low-risk text (e.g., out-of-copyright books or government documents), due to its limited size and domain coverage. We present SILO, a new language model that manages this risk-performance tradeoff during inference. SILO is built by (1) training a parametric LM on Open License Corpus (OLC), a new corpus we curate with 228B tokens of public domain and permissively licensed text and (2) augmenting it with a more general and easily modifiable nonparametric datastore (e.g., containing copyrighted books or news) that is only queried during inference. The datastore allows use of high-risk data without training on it, supports sentence-level data attribution, and enables data producers to opt out from the model by removing content from the store. These capabilities can foster compliance with data-use regulations such as the fair use doctrine in the United States and the GDPR in the European Union. Our experiments show that the parametric LM struggles on domains not covered by OLC. However, access to the datastore greatly improves out of domain performance, closing 90% of the performance gap with an LM trained on the Pile, a more diverse corpus with mostly high-risk text. We also analyze which nonparametric approach works best, where the remaining errors lie, and how performance scales with datastore size. Our results suggest that it is possible to build high quality language models while mitigating their legal risk."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "Holistic Evaluation of Language Models",
                    "abstract": "Language models (LMs) like GPT\u20103, PaLM, and ChatGPT are the foundation for almost all major language technologies, but their capabilities, limitations, and risks are not well understood. We present Holistic Evaluation of Language Models (HELM) to improve the transparency of LMs. LMs can serve many purposes and their behavior should satisfy many desiderata. To navigate the vast space of potential scenarios and metrics, we taxonomize the space and select representative subsets. We evaluate models on 16 core scenarios and 7 metrics, exposing important trade\u2010offs. We supplement our core evaluation with seven targeted evaluations to deeply analyze specific aspects (including world knowledge, reasoning, regurgitation of copyrighted content, and generation of disinformation). We benchmark 30 LMs, from OpenAI, Microsoft, Google, Meta, Cohere, AI21 Labs, and others. Prior to HELM, models were evaluated on just 17.9% of the core HELM scenarios, with some prominent models not sharing a single scenario in common. We improve this to 96.0%: all 30 models are now benchmarked under the same standardized conditions. Our evaluation surfaces 25 top\u2010level findings. For full transparency, we release all raw model prompts and completions publicly. HELM is a living benchmark for the community, continuously updated with new scenarios, metrics, and models https://crfm.stanford.edu/helm/latest/."
                },
                {
                    "title": "Trusting Your Evidence: Hallucinate Less with Context-aware Decoding",
                    "abstract": "Language models (LMs) often struggle to pay enough attention to the input context, and generate texts that are unfaithful or contain hallucinations. To mitigate this issue, we present context-aware decoding (CAD), which follows a contrastive output distribution that amplifies the difference between the output probabilities when a model is used with and without context. Our experiments show that CAD, without additional training, significantly improves the faithfulness of different LM families, including OPT, GPT, LLaMA, and FLAN-T5 for summarization tasks (e.g., 14.3% gain for LLaMA in factuality metrics). Furthermore, CAD is particularly effective in overriding a model\u2019s prior knowledge when it contradicts the provided context, leading to substantial improvements in tasks where resolving the knowledge conflict is essential. Our code is publicly released at https://github.com/xhan77/context-aware-decoding."
                },
                {
                    "title": "Speak, Memory: An Archaeology of Books Known to ChatGPT/GPT-4",
                    "abstract": "In this work, we carry out a data archaeology to infer books that are known to ChatGPT and GPT-4 using a name cloze membership inference query. We find that OpenAI models have memorized a wide collection of copyrighted materials, and that the degree of memorization is tied to the frequency with which passages of those books appear on the web. The ability of these models to memorize an unknown set of books complicates assessments of measurement validity for cultural analytics by contaminating test data; we show that models perform much better on memorized books than on non-memorized books for downstream tasks. We argue that this supports a case for open models whose training data is known."
                },
                {
                    "title": "Foundation Models and Fair Use",
                    "abstract": "Existing foundation models are trained on copyrighted material. Deploying these models can pose both legal and ethical risks when data creators fail to receive appropriate attribution or compensation. In the United States and several other countries, copyrighted content may be used to build foundation models without incurring liability due to the fair use doctrine. However, there is a caveat: If the model produces output that is similar to copyrighted data, particularly in scenarios that affect the market of that data, fair use may no longer apply to the output of the model. In this work, we emphasize that fair use is not guaranteed, and additional work may be necessary to keep model development and deployment squarely in the realm of fair use. First, we survey the potential risks of developing and deploying foundation models based on copyrighted content. We review relevant U.S. case law, drawing parallels to existing and potential applications for generating text, source code, and visual art. Experiments confirm that popular foundation models can generate content considerably similar to copyrighted material. Second, we discuss technical mitigations that can help foundation models stay in line with fair use. We argue that more research is needed to align mitigation strategies with the current state of the law. Lastly, we suggest that the law and technical mitigations should co-evolve. For example, coupled with other policy mechanisms, the law could more explicitly consider safe harbors when strong technical tools are used to mitigate infringement harms. This co-evolution may help strike a balance between intellectual property and innovation, which speaks to the original goal of fair use. But we emphasize that the strategies we describe here are not a panacea and more work is needed to develop policies that address the potential harms of foundation models."
                },
                {
                    "title": "Data Portraits: Recording Foundation Model Training Data",
                    "abstract": "Foundation models are trained on increasingly immense and opaque datasets. Even while these models are now key in AI system building, it can be difficult to answer the straightforward question: has the model already encountered a given example during training? We therefore propose a widespread adoption of Data Portraits: artifacts that record training data and allow for downstream inspection. First we outline the properties of such an artifact and discuss how existing solutions can be used to increase transparency. We then propose and implement a solution based on data sketching, stressing fast and space efficient querying. Using our tools, we document a popular language modeling corpus (The Pile) and a recently released code modeling dataset (The Stack). We show that our solution enables answering questions about test set leakage and model plagiarism. Our tool is lightweight and fast, costing only 3% of the dataset size in overhead. We release a live interface of our tools at https://dataportraits.org/ and call on dataset and model creators to release Data Portraits as a complement to current documentation practices."
                },
                {
                    "title": "REPLUG: Retrieval-Augmented Black-Box Language Models",
                    "abstract": "We introduce REPLUG, a retrieval-augmented language modeling framework that treats the language model (LM) as a black box and augments it with a tuneable retrieval model. Unlike prior retrieval-augmented LMs that train language models with special cross-attention mechanisms to encode the retrieved text, REPLUG simply prepends retrieved documents to the input for the frozen black-box LM. This simple design can be easily applied to any existing language models. Furthermore, we show that the LM can be used to supervise the retrieval model, which can then find documents that help the LM make better predictions. Our experiments demonstrate that REPLUG with the tuned retriever significantly improves the performance of GPT-3 (175B) on language modeling by 6.3%, as well as the performance of Codex on five-shot MMLU by 5.1%. Code is publicly released at github.com/swj0419/REPLUG."
                },
                {
                    "title": "Editing Models with Task Arithmetic",
                    "abstract": "Changing how pre-trained models behave -- e.g., improving their performance on a downstream task or mitigating biases learned during pre-training -- is a common practice when developing machine learning systems. In this work, we propose a new paradigm for steering the behavior of neural networks, centered around \\textit{task vectors}. A task vector specifies a direction in the weight space of a pre-trained model, such that movement in that direction improves performance on the task. We build task vectors by subtracting the weights of a pre-trained model from the weights of the same model after fine-tuning on a task. We show that these task vectors can be modified and combined together through arithmetic operations such as negation and addition, and the behavior of the resulting model is steered accordingly. Negating a task vector decreases performance on the target task, with little change in model behavior on control tasks. Moreover, adding task vectors together can improve performance on multiple tasks at once. Finally, when tasks are linked by an analogy relationship of the form ``A is to B as C is to D\", combining task vectors from three of the tasks can improve performance on the fourth, even when no data from the fourth task is used for training. Overall, our experiments with several models, modalities and tasks show that task arithmetic is a simple, efficient and effective way of editing models."
                },
                {
                    "title": "A Dataset Auditing Method for Collaboratively Trained Machine Learning Models",
                    "abstract": "Dataset auditing for machine learning (ML) models is a method to evaluate if a given dataset is used in training a model. In a Federated Learning setting where multiple institutions collaboratively train a model with their decentralized private datasets, dataset auditing can facilitate the enforcement of regulations, which provide rules for preserving privacy, but also allow users to revoke authorizations and remove their data from collaboratively trained models. This paper first proposes a set of requirements for a practical dataset auditing method, and then present a novel dataset auditing method called Ensembled Membership Auditing (<inline-formula> <tex-math notation=\"LaTeX\">$\\mathsf{EMA}$ </tex-math></inline-formula>). Its key idea is to leverage previously proposed Membership Inference Attack methods and to aggregate data-wise membership scores using statistic testing to audit a dataset for a ML model. We have experimentally evaluated the proposed approach with benchmark datasets, as well as 4 X-ray datasets (CBIS-DDSM, COVIDx, Child-XRay, and CXR-NIH) and 3 dermatology datasets (DERM7pt, HAM10000, and PAD-UFES-20). Our results show that <inline-formula> <tex-math notation=\"LaTeX\">$\\mathsf{EMA}$ </tex-math></inline-formula> meet the requirements substantially better than the previous state-of-the-art method. Our code is at:<uri>https://github.com/Hazelsuko07/EMA</uri>."
                },
                {
                    "title": "Preventing Generation of Verbatim Memorization in Language Models Gives a False Sense of Privacy",
                    "abstract": "Studying data memorization in neural language models helps us understand the risks (e.g., to privacy or copyright) associated with models regurgitating training data and aids in the development of countermeasures. Many prior works\u2014and some recently deployed defenses\u2014focus on \u201cverbatim memorization\u201d, defined as a model generation that exactly matches a substring from the training set. We argue that verbatim memorization definitions are too restrictive and fail to capture more subtle forms of memorization. Specifically, we design and implement an efficient defense that _perfectly_ prevents all verbatim memorization. And yet, we demonstrate that this \u201cperfect\u201d filter does not prevent the leakage of training data. Indeed, it is easily circumvented by plausible and minimally modified \u201cstyle-transfer\u201d prompts\u2014and in some cases even the non-modified original prompts\u2014to extract memorized information. We conclude by discussing potential alternative definitions and why defining memorization is a difficult yet crucial open question for neural language models."
                },
                {
                    "title": "Continual Learning and Private Unlearning",
                    "abstract": "As intelligent agents become autonomous over longer periods of time, they may eventually become lifelong counterparts to specific people. If so, it may be common for a user to want the agent to master a task temporarily but later on to forget the task due to privacy concerns. However enabling an agent to \\emph{forget privately} what the user specified without degrading the rest of the learned knowledge is a challenging problem. With the aim of addressing this challenge, this paper formalizes this continual learning and private unlearning (CLPU) problem. The paper further introduces a straightforward but exactly private solution, CLPU-DER++, as the first step towards solving the CLPU problem, along with a set of carefully designed benchmark problems to evaluate the effectiveness of the proposed solution. The code is available at https://github.com/Cranial-XIX/Continual-Learning-Private-Unlearning."
                },
                {
                    "title": "Quantifying Memorization Across Neural Language Models",
                    "abstract": "Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations."
                },
                {
                    "title": "Locating and Editing Factual Associations in GPT",
                    "abstract": "We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's factual predictions. This reveals a distinct set of steps in middle-layer feed-forward modules that mediate factual predictions while processing subject tokens. To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update specific factual associations using Rank-One Model Editing (ROME). We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another. Our results confirm an important role for mid-layer feed-forward modules in storing factual associations and suggest that direct manipulation of computational mechanisms may be a feasible approach for model editing. The code, dataset, visualizations, and an interactive demo notebook are available at https://rome.baulab.info/"
                },
                {
                    "title": "LaMDA: Language Models for Dialog Applications",
                    "abstract": "We present LaMDA: Language Models for Dialog Applications. LaMDA is a family of Transformer-based neural language models specialized for dialog, which have up to 137B parameters and are pre-trained on 1.56T words of public dialog data and web text. While model scaling alone can improve quality, it shows less improvements on safety and factual grounding. We demonstrate that fine-tuning with annotated data and enabling the model to consult external knowledge sources can lead to significant improvements towards the two key challenges of safety and factual grounding. The first challenge, safety, involves ensuring that the model's responses are consistent with a set of human values, such as preventing harmful suggestions and unfair bias. We quantify safety using a metric based on an illustrative set of human values, and we find that filtering candidate responses using a LaMDA classifier fine-tuned with a small amount of crowdworker-annotated data offers a promising approach to improving model safety. The second challenge, factual grounding, involves enabling the model to consult external knowledge sources, such as an information retrieval system, a language translator, and a calculator. We quantify factuality using a groundedness metric, and we find that our approach enables the model to generate responses grounded in known sources, rather than responses that merely sound plausible. Finally, we explore the use of LaMDA in the domains of education and content recommendations, and analyze their helpfulness and role consistency."
                },
                {
                    "title": "Counterfactual Memorization in Neural Language Models",
                    "abstract": "Modern neural language models that are widely used in various NLP tasks risk memorizing sensitive information from their training data. Understanding this memorization is important in real world applications and also from a learning-theoretical perspective. An open question in previous studies of language model memorization is how to filter out\"common\"memorization. In fact, most memorization criteria strongly correlate with the number of occurrences in the training set, capturing memorized familiar phrases, public knowledge, templated texts, or other repeated data. We formulate a notion of counterfactual memorization which characterizes how a model's predictions change if a particular document is omitted during training. We identify and study counterfactually-memorized training examples in standard text datasets. We estimate the influence of each memorized training example on the validation set and on generated texts, showing how this can provide direct evidence of the source of memorization at test time."
                },
                {
                    "title": "Unrolling SGD: Understanding Factors Influencing Machine Unlearning",
                    "abstract": "Machine unlearning is the process through which a deployed machine learning model is made to forget about some of its training data points. While naively retraining the model from scratch is an option, it is almost always associated with large computational overheads for deep learning models. Thus, several approaches to approximately unlearn have been proposed along with corresponding metrics that formalize what it means for a model to forget about a data point. In this work, we first taxonomize approaches and metrics of approximate unlearning. As a result, we identify verification error, i.e., the $\\ell_{2}$ difference between the weights of an approximately unlearned and a naively retrained model, as an approximate unlearning metric that should be optimized for as it subsumes a large class of other metrics. We theoretically analyze the canonical training algorithm, stochastic gradient descent (SGD), to surface the variables which are relevant to reducing the verification error of approximate unlearning for SGD. From this analysis, we first derive an easy-to-compute proxy for verification error (termed unlearning error). The analysis also informs the design of a new training objective penalty that limits the overall change in weights during SGD and as a result facilitates approximate unlearning with lower verification error. We validate our theoretical work through an empirical evaluation on learning with CIFAR-10, CIFAR-100, and IMDB sentiment analysis."
                },
                {
                    "title": "On the Opportunities and Risks of Foundation Models",
                    "abstract": "AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature."
                },
                {
                    "title": "BookSum: A Collection of Datasets for Long-form Narrative Summarization",
                    "abstract": "The majority of available text summarization datasets include short-form source documents that lack long-range causal and temporal dependencies, and often contain strong layout and stylistic biases. While relevant, such datasets will offer limited challenges for future generations of text summarization systems. We address these issues by introducing BookSum, a collection of datasets for long-form narrative summarization. Our dataset covers source documents from the literature domain, such as novels, plays and stories, and includes highly abstractive, human written summaries on three levels of granularity of increasing difficulty: paragraph-, chapter-, and book-level. The domain and structure of our dataset poses a unique set of challenges for summarization systems, which include: processing very long documents, non-trivial causal and temporal dependencies, and rich discourse structures. To facilitate future work, we trained and evaluated multiple extractive and abstractive summarization models as baselines for our dataset."
                },
                {
                    "title": "Extracting Training Data from Large Language Models",
                    "abstract": "It has become common to publish large (billion parameter) language models that have been trained on private datasets. This paper demonstrates that in such settings, an adversary can perform a training data extraction attack to recover individual training examples by querying the language model. \nWe demonstrate our attack on GPT-2, a language model trained on scrapes of the public Internet, and are able to extract hundreds of verbatim text sequences from the model's training data. These extracted examples include (public) personally identifiable information (names, phone numbers, and email addresses), IRC conversations, code, and 128-bit UUIDs. Our attack is possible even though each of the above sequences are included in just one document in the training data. \nWe comprehensively evaluate our extraction attack to understand the factors that contribute to its success. For example, we find that larger models are more vulnerable than smaller models. We conclude by drawing lessons and discussing possible safeguards for training large language models."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks",
                    "abstract": "Large pre-trained language models have been shown to store factual knowledge in their parameters, and achieve state-of-the-art results when fine-tuned on downstream NLP tasks. However, their ability to access and precisely manipulate knowledge is still limited, and hence on knowledge-intensive tasks, their performance lags behind task-specific architectures. Additionally, providing provenance for their decisions and updating their world knowledge remain open research problems. Pre-trained models with a differentiable access mechanism to explicit non-parametric memory can overcome this issue, but have so far been only investigated for extractive downstream tasks. We explore a general-purpose fine-tuning recipe for retrieval-augmented generation (RAG) -- models which combine pre-trained parametric and non-parametric memory for language generation. We introduce RAG models where the parametric memory is a pre-trained seq2seq model and the non-parametric memory is a dense vector index of Wikipedia, accessed with a pre-trained neural retriever. We compare two RAG formulations, one which conditions on the same retrieved passages across the whole generated sequence, the other can use different passages per token. We fine-tune and evaluate our models on a wide range of knowledge-intensive NLP tasks and set the state-of-the-art on three open domain QA tasks, outperforming parametric seq2seq models and task-specific retrieve-and-extract architectures. For language generation tasks, we find that RAG models generate more specific, diverse and factual language than a state-of-the-art parametric-only seq2seq baseline."
                },
                {
                    "title": "Fair Learning",
                    "abstract": "Neural network and machine learning artificial intelligences (AIs) need comprehensive data sets to train on. Those data sets will often be composed of images, videos, audio, or text. All those things are copyrighted. Copyright law thus stands as an enormous potential obstacle to training AIs. Not only might the aggregate data sets themselves be copyrighted, but each individual image, video, and text in the data set is likely to be copyrighted too. It\u2019s not clear that the use of these databases of copyrighted works to build self-driving cars, or to learn natural languages by analyzing the content in them, will be treated as a fair use under current law. Fair use doctrine in the last quarter century has focused on the transformation of the copyrighted work. AIs aren\u2019t transforming the databases they train on; they are using the entire database, and for a commercial purpose at that. Courts may view that as a kind of free riding they should prohibit. In this Article, we argue that AIs should generally be able to use databases for training whether or not the contents of that database are copyrighted. There are good policy reasons to do so. And because training data sets are likely to contain millions of different works with thousands of different owners, there is no plausible option simply to license all the underlying photographs or texts for the new use. So allowing a copyright claim is tantamount to saying, not that copyright owners will get paid, but that no one will get the benefit of this new use. There is another, deeper reason to permit such uses, one that has implications far beyond training AIs. Understanding why the use of copyrighted works by AIs should be fair actually reveals a significant issue at the heart of copyright law. Sometimes people (or machines) copy expression but they are only interested in learning the ideas conveyed by that expression. That\u2019s what is going on with training data in most cases. The AI wants photos of stop signs so it can learn to recognize stop signs, not because of whatever artistic choices you made in lighting or composing your photo. Similarly, it wants to see what you wrote to learn how words are sequenced in ordinary conversation, not because your prose is particularly expressive. AIs are not alone in wanting just the facts. The issue arises in lots of other contexts. In American Geophysical Union v. Texaco, for example, the defendants were interested only in the ideas in scientific journal articles; photocopying the article was simply the most convenient way of gaining access to those ideas. Other examples include copyright disputes over software interoperability cases like Google v. Oracle, current disputes over copyright in state statutes and rules adopted into law, and perhaps even Bikram yoga poses and the tangled morass of cases around copyright protection for the artistic aspects of utilitarian works like clothing and bike racks. In all of these cases, copyright law is being used to target defendants who actually want something the law is not supposed to protect \u2013 the underlying ideas, facts, or functions of the work. Copyright law should permit copying of works for non-expressive purposes. When the defendant copies a work for reasons other than to have access to the protectable expression in that work, fair use should consider under both factors one and two whether the purpose of the defendant\u2019s copying was to appropriate the plaintiff\u2019s expression or just the ideas. We don\u2019t want to allow the copyright on the creative pieces to end up controlling the unprotectable elements."
                },
                {
                    "title": "Eternal Sunshine of the Spotless Net: Selective Forgetting in Deep Networks",
                    "abstract": "We explore the problem of selectively forgetting a particular subset of the data used for training a deep neural network. While the effects of the data to be forgotten can be hidden from the output of the network, insights may still be gleaned by probing deep into its weights. We propose a method for \"scrubbing\" the weights clean of information about a particular set of training data. The method does not require retraining from scratch, nor access to the data originally used for training. Instead, the weights are modified so that any probing function of the weights is indistinguishable from the same function applied to the weights of a network trained without the data to be forgotten. This condition is a generalized and weaker form of Differential Privacy. Exploiting ideas related to the stability of stochastic gradient descent, we introduce an upper-bound on the amount of information remaining in the weights, which can be estimated efficiently even for deep neural networks."
                },
                {
                    "title": "Certified Data Removal from Machine Learning Models",
                    "abstract": "Good data stewardship requires removal of data at the request of the data's owner. This raises the question if and how a trained machine-learning model, which implicitly stores information about its training data, should be affected by such a removal request. Is it possible to \"remove\" data from a machine-learning model? We study this problem by defining certified removal: a very strong theoretical guarantee that a model from which data is removed cannot be distinguished from a model that never observed the data to begin with. We develop a certified-removal mechanism for linear classifiers and empirically study learning settings in which this mechanism is practical."
                },
                {
                    "title": "The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks",
                    "abstract": "This paper describes a testing methodology for quantitatively assessing the risk that rare or unique training-data sequences are unintentionally memorized by generative sequence models---a common type of machine-learning model. Because such models are sometimes trained on sensitive data (e.g., the text of users' private messages), this methodology can benefit privacy by allowing deep-learning practitioners to select means of training that minimize such memorization. \nIn experiments, we show that unintended memorization is a persistent, hard-to-avoid issue that can have serious consequences. Specifically, for models trained without consideration of memorization, we describe new, efficient procedures that can extract unique, secret sequences, such as credit card numbers. We show that our testing strategy is a practical and easy-to-use first line of defense, e.g., by describing its application to quantitatively limit data exposure in Google's Smart Compose, a commercial text-completion neural network trained on millions of users' email messages."
                },
                {
                    "title": "NewsQA: A Machine Comprehension Dataset",
                    "abstract": "We present NewsQA, a challenging machine comprehension dataset of over 100,000 human-generated question-answer pairs. Crowdworkers supply questions and answers based on a set of over 10,000 news articles from CNN, with answers consisting of spans of text in the articles. We collect this dataset through a four-stage process designed to solicit exploratory questions that require reasoning. Analysis confirms that NewsQA demands abilities beyond simple word matching and recognizing textual entailment. We measure human performance on the dataset and compare it to several strong neural models. The performance gap between humans and machines (13.3% F1) indicates that significant progress can be made on NewsQA through future research. The dataset is freely available online."
                },
                {
                    "title": "Towards Making Systems Forget with Machine Unlearning",
                    "abstract": "Today's systems produce a rapidly exploding amount of data, and the data further derives more data, forming a complex data propagation network that we call the data's lineage. There are many reasons that users want systems to forget certain data including its lineage. From a privacy perspective, users who become concerned with new privacy risks of a system often want the system to forget their data and lineage. From a security perspective, if an attacker pollutes an anomaly detector by injecting manually crafted data into the training data set, the detector must forget the injected data to regain security. From a usability perspective, a user can remove noise and incorrect entries so that a recommendation engine gives useful recommendations. Therefore, we envision forgetting systems, capable of forgetting certain data and their lineages, completely and quickly. This paper focuses on making learning systems forget, the process of which we call machine unlearning, or simply unlearning. We present a general, efficient unlearning approach by transforming learning algorithms used by a system into a summation form. To forget a training data sample, our approach simply updates a small number of summations -- asymptotically faster than retraining from scratch. Our approach is general, because the summation form is from the statistical query learning in which many machine learning algorithms can be implemented. Our approach also applies to all stages of machine learning, including feature selection and modeling. Our evaluation, on four diverse learning systems and real-world workloads, shows that our approach is general, effective, fast, and easy to use."
                },
                {
                    "title": "ROUGE: A Package for Automatic Evaluation of Summaries",
                    "abstract": "ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST."
                },
                {
                    "title": "On the resemblance and containment of documents",
                    "abstract": "Given two documents A and B we define two mathematical notions: their resemblance r(A, B) and their containment c(A, B) that seem to capture well the informal notions of \"roughly the same\" and \"roughly contained.\" The basic idea is to reduce these issues to set intersection problems that can be easily evaluated by a process of random sampling that can be done independently for each document. Furthermore, the resemblance can be evaluated using a fixed size sample for each document. This paper discusses the mathematical properties of these measures and the efficient implementation of the sampling process using Rabin (1981) fingerprints."
                },
                {
                    "title": "Retrieval-Augmented Multimodal Language Modeling",
                    "abstract": "Recent multimodal models such as DALL-E and\r\nCM3 have achieved remarkable progress in textto-image and image-to-text generation. However, these models store all learned knowledge (e.g., the appearance of the Eiffel Tower) in the\r\nmodel parameters, requiring increasingly larger models and training data to capture more knowledge. To integrate knowledge in a more scalable and modular way, we propose a retrievalaugmented multimodal model, which enables a base multimodal model (generator) to refer to relevant knowledge fetched by a retriever from external memory (e.g., multimodal documents on the web). Specifically, we implement a retriever using the pretrained CLIP model and a generator using the CM3 Transformer architecture, and\r\ntrain this model using the LAION dataset. Our resulting model, named Retrieval-Augmented CM3 (RA-CM3), is the first multimodal model that can retrieve and generate mixtures of text and images.\r\nWe show that RA-CM3 significantly outperforms baseline multimodal models such as DALL-E and CM3 on both image and caption generation tasks (12 FID and 17 CIDEr improvements on MSCOCO), while requiring much less compute for training (<30% of DALL-E). Moreover, we show that RA-CM3 exhibits novel capabilities such as knowledge-intensive image generation and multimodal in-context learning"
                },
                {
                    "title": "GitHub Copilot",
                    "abstract": "Durante el programa de ingenier\u00eda de sistemas se ha podido trabajar con muchas herramientas las cuales facilitan las ense\u00f1anzas de los temas que est\u00e1n enfocados al desarrollo del software al igual que el cloud computing. En esta ocasi\u00f3n se hablar\u00e1 de un nuevo m\u00e9todo que sali\u00f3 recientemente, se llama GitHub Copilot; es un asistente que sirve para escribir c\u00f3digo basado en machine learn, o como les gusta decirle en GitHub, es una aplicaci\u00f3n de pair programming basado en IA (inteligencia artificial). Para comprender m\u00e1s este software, se relaciona m\u00e1s como un Intellisense (auto completa el c\u00f3digo) pero m\u00e1s eficiente y va mejorando a medida que transcurren sus actualizaciones. Por otro lado, no solamente hace peque\u00f1as sugerencias sino predice lo que quieres completar y te brinda funciones complementarias con m\u00faltiples variantes. Incluso pude llegar a escribir el c\u00f3digo a partir de comentarios en el lenguaje que se requiera. En el aula ha sido de gran ayuda para explicar los temas que se enfocan en la inteligencia artificial y programaci\u00f3n a nuevos estudiantes, debido a que con el lenguaje \u201ccom\u00fan\u201d se puede realizar una estructura similar al pseudoc\u00f3digo. De igual manera, se puede comprender los modelos de servicios en la nube, como, por ejemplo: Software as service (SaaS), porque copilot recopila todos los repositorios p\u00fablicos de GitHub que es similar a un data warehouse el cual abastece a la IA. El objetivo principal de la presentaci\u00f3n es demostrar c\u00f3mo se puede optimizar el desarrollo de un software con la ayuda de GitHub Copilot, tambi\u00e9n explicar sus ventajas al igual que su arquitectura en la nube, con el objetivo de que todos los ingenieros se puedan apoyar a trav\u00e9s de esta herramienta. Esta nueva tecnolog\u00eda a\u00fan no ha completado su funcionalidad a los usuarios en general, sin embargo, para acceder se requiere solicitar su uso a GitHub Copilot. Actualmente muy pocos desarrolladores tienen posibilidades de acceso a este programa; se espera que en los pr\u00f3ximos meses salga su producci\u00f3n a los dem\u00e1s usuarios. Hoy en d\u00eda con este software durante 6 meses aproximadamente, se ha demostrado que casi ha mejorado desde sus etapas iniciales de lanzamiento, es por tal motivo que hay una mejor experiencia en el uso diario de la misma."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan \u201ctakedown\u201d of copyrighted content be operationalized in the context of large language models?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the legal and ethical implications of using copyrighted material in training large language models (LLMs). By developing effective takedown mechanisms, researchers can ensure that LLMs operate within legal boundaries, thereby fostering trust and compliance in AI technologies. This work could lead to advancements in model deployment strategies, influencing future research on copyright compliance and the development of more robust models that respect intellectual property rights. Additionally, practical applications of this research could include the creation of tools that help organizations navigate copyright issues, ultimately promoting responsible AI usage.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities of distinguishing between copyrighted and non-copyrighted content within the vast datasets used for training LLMs. Naive approaches may fail because they might not adequately account for the nuances of copyright law, such as fair use, or the intricacies of how models memorize and generate content. Technical obstacles include developing algorithms that can effectively filter out specific copyrighted material without compromising the model's performance or utility. Theoretical challenges involve understanding the implications of copyright law in the context of AI and ensuring that takedown mechanisms do not inadvertently suppress legitimate outputs.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often overlooked the specific operationalization of copyright takedowns in LLMs, focusing instead on broader issues of copyright and AI. Limitations in existing solutions include a lack of comprehensive frameworks for evaluating the effectiveness of takedown mechanisms and insufficient understanding of how LLMs memorize and reproduce content. Barriers such as the rapid evolution of AI technologies and the lagging development of legal frameworks have also contributed to this gap. This paper proposes a novel approach by introducing the CoTaEval benchmark, which specifically evaluates the feasibility and side effects of copyright takedowns, thus filling a critical void in the literature.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves the development of the CoTaEval benchmark, which will assess the effectiveness of copyright takedown mechanisms in LLMs. This evaluation will consider both memorized content and content generated through retrieval-augmented generation (RAG) techniques. The dataset will"
            }
        },
        "author_data": {
            "968968ae-0db5-44ed-8367-78ed019b2fa5": {
                "pk": "968968ae-0db5-44ed-8367-78ed019b2fa5",
                "name": "Boyi Wei",
                "collaborators": [
                    "Chenyang Shi",
                    "Hanxiao Liu",
                    "Yibo Zhang",
                    "Jing Jin",
                    "Yangsibo Huang",
                    "Peter Henderson",
                    "Yuchen Gui",
                    "Wei Yuan",
                    "Xi Jin",
                    "Shasha Guo"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Event Cameras",
                    "Large Language Models",
                    "Video Frame Interpolation"
                ],
                "publications": [
                    {
                        "title": "Hardware Acceleration of Sampling Algorithms in Sample and Aggregate Graph Neural Networks",
                        "abstract": "Sampling is an important process in many GNN structures in order to train larger datasets with a smaller computational complexity. However, compared to other processes in GNN (such as aggregate, backward propagation), the sampling process still costs tremendous time, which limits the speed of training. To reduce the time of sampling, hardware acceleration is an ideal choice. However, state of the art GNN acceleration proposal did not specify how to accelerate the sampling process. What's more, directly accelerating traditional sampling algorithms will make the structure of the accelerator very complicated.   In this work, we made two contributions: (1) Proposed a new neighbor sampler: CONCAT Sampler, which can be easily accelerated on hardware level while guaranteeing the test accuracy. (2) Designed a CONCAT-sampler-accelerator based on FPGA, with which the neighbor sampling process boosted to about 300-1000 times faster compared to the sampling process without it."
                    },
                    {
                        "title": "A Label-Free and Non-Monotonic Metric for Evaluating Denoising in Event Cameras",
                        "abstract": "Event cameras are renowned for their high efficiency due to outputting a sparse, asynchronous stream of events. However, they are plagued by noisy events, especially in low light conditions. Denoising is an essential task for event cameras, but evaluating denoising performance is challenging. Label-dependent denoising metrics involve artificially adding noise to clean sequences, complicating evaluations. Moreover, the majority of these metrics are monotonic, which can inflate scores by removing substantial noise and valid events. To overcome these limitations, we propose the first label-free and non-monotonic evaluation metric, the area of the continuous contrast curve (AOCC), which utilizes the area enclosed by event frame contrast curves across different time intervals. This metric is inspired by how events capture the edge contours of scenes or objects with high temporal resolution. An effective denoising method removes noise without eliminating these edge-contour events, thus preserving the contrast of event frames. Consequently, contrast across various time ranges serves as a metric to assess denoising effectiveness. As the time interval lengthens, the curve will initially rise and then fall. The proposed metric is validated through both theoretical and experimental evidence."
                    },
                    {
                        "title": "An Adversarial Perspective on Machine Unlearning for AI Safety",
                        "abstract": "Large language models are finetuned to refuse questions about hazardous knowledge, but these protections can often be bypassed. Unlearning methods aim at completely removing hazardous capabilities from models and make them inaccessible to adversaries. This work challenges the fundamental differences between unlearning and traditional safety post-training from an adversarial perspective. We demonstrate that existing jailbreak methods, previously reported as ineffective against unlearning, can be successful when applied carefully. Furthermore, we develop a variety of adaptive methods that recover most supposedly unlearned capabilities. For instance, we show that finetuning on 10 unrelated examples or removing specific directions in the activation space can recover most hazardous capabilities for models edited with RMU, a state-of-the-art unlearning method. Our findings challenge the robustness of current unlearning approaches and question their advantages over safety training."
                    },
                    {
                        "title": "IDO-VFI: Identifying Dynamics via Optical Flow Guidance for Video Frame Interpolation with Events",
                        "abstract": "Video frame interpolation aims to generate high-quality intermediate frames from boundary frames and increase frame rate. While existing linear, symmetric and nonlinear models are used to bridge the gap from the lack of inter-frame motion, they cannot reconstruct real motions. Event cameras, however, are ideal for capturing inter-frame dynamics with their extremely high temporal resolution. In this paper, we propose an event-and-frame-based video frame interpolation method named IDO-VFI that assigns varying amounts of computation for different sub-regions via optical flow guidance. The proposed method first estimates the optical flow based on frames and events, and then decides whether to further calculate the residual optical flow in those sub-regions via a Gumbel gating module according to the optical flow amplitude. Intermediate frames are eventually generated through a concise Transformer-based fusion network. Our proposed method maintains high-quality performance while reducing computation time and computational effort by 10% and 17% respectively on Vimeo90K datasets, compared with a unified process on the whole region. Moreover, our method outperforms state-of-the-art frame-only and frames-plus-events methods on multiple video frame interpolation benchmarks. Codes and models are available at https://github.com/shicy17/IDO-VFI."
                    },
                    {
                        "title": "Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modifications",
                        "abstract": "Large language models (LLMs) show inherent brittleness in their safety mechanisms, as evidenced by their susceptibility to jailbreaking and even non-malicious fine-tuning. This study explores this brittleness of safety alignment by leveraging pruning and low-rank modifications. We develop methods to identify critical regions that are vital for safety guardrails, and that are disentangled from utility-relevant regions at both the neuron and rank levels. Surprisingly, the isolated regions we find are sparse, comprising about $3\\%$ at the parameter level and $2.5\\%$ at the rank level. Removing these regions compromises safety without significantly impacting utility, corroborating the inherent brittleness of the model's safety mechanisms. Moreover, we show that LLMs remain vulnerable to low-cost fine-tuning attacks even when modifications to the safety-critical regions are restricted. These findings underscore the urgent need for more robust safety strategies in LLMs."
                    }
                ]
            },
            "6d443bbd-7423-4e76-a754-0b561410ad82": {
                "pk": "6d443bbd-7423-4e76-a754-0b561410ad82",
                "name": "Weijia Shi",
                "collaborators": [
                    "Muhao Chen",
                    "Luke Zettlemoyer",
                    "Dan Roth",
                    "Kai-Wei Chang",
                    "Mandar Joshi",
                    "Julian Michael",
                    "Suchin Gururangan",
                    "Ben Zhou",
                    "Fangyuan Xu",
                    "Eunsol Choi"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Knowledge Graph",
                    "Entity Linking",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "DESCGEN: A Distantly Supervised Dataset for Generating Abstractive Entity Descriptions",
                        "abstract": "Short textual descriptions of entities provide summaries of their key attributes and have been shown to be useful sources of background knowledge for tasks such as entity linking and question answering. However, generating entity descriptions, especially for new and long-tail entities, can be challenging since relevant information is often scattered across multiple sources with varied content and style. We introduce DESCGEN: given mentions spread over multiple documents, the goal is to generate an entity summary description. DESCGEN consists of 37K entity descriptions from Wikipedia and Fandom, each paired with nine evidence documents on average. The documents were collected using a combination of entity linking and hyperlinks to the Wikipedia and Fandom entity pages, which together provide high-quality distant supervision. The resulting summaries are more abstractive than those found in existing datasets and provide a better proxy for the challenge of describing new and emerging entities. We also propose a two-stage extract-then-generate baseline and show that there exists a large gap (19.9% in ROUGE-L) between state-of-the-art models and human performance, suggesting that the data will support significant future work."
                    },
                    {
                        "title": "kNN-Prompt: Nearest Neighbor Zero-Shot Inference",
                        "abstract": "Retrieval-augmented language models (LMs) use non-parametric memory to substantially outperform their non-retrieval counterparts on perplexity-based evaluations, but it is an open question whether they achieve similar gains in few- and zero-shot end-task accuracy. We extensively study one such model, the k-nearest neighbor LM (kNN-LM), showing that the gains marginally transfer. The main challenge is to achieve coverage of the verbalizer tokens that define the different end-task class labels. To address this challenge, we also introduce kNN-Prompt, a simple and effective kNN-LM with automatically expanded fuzzy verbalizers (e.g. to expand terrible to also include silly and other task-specific synonyms for sentiment classification). Across nine diverse end-tasks, using kNN-Prompt with GPT-2 large yields significant performance boosts over strong zero-shot baselines (13.4% absolute improvement over the base LM on average). We also show that other advantages of non-parametric augmentation hold for end tasks; kNN-Prompt is effective for domain adaptation with no further training, and gains increase with the size of the retrieval model."
                    },
                    {
                        "title": "Cross-lingual Entity Alignment with Incidental Supervision",
                        "abstract": "Much research effort has been put to multilingual knowledge graph (KG) embedding methods to address the entity alignment task, which seeks to match entities in different languagespecific KGs that refer to the same real-world object. Such methods are often hindered by the insufficiency of seed alignment provided between KGs. Therefore, we propose an incidentally supervised model, JEANS , which jointly represents multilingual KGs and text corpora in a shared embedding scheme, and seeks to improve entity alignment with incidental supervision signals from text. JEANS first deploys an entity grounding process to combine each KG with the monolingual text corpus. Then, two learning processes are conducted: (i) an embedding learning process to encode the KG and text of each language in one embedding space, and (ii) a selflearning based alignment learning process to iteratively induce the matching of entities and that of lexemes between embeddings. Experiments on benchmark datasets show that JEANS leads to promising improvement on entity alignment with incidental supervision, and significantly outperforms state-of-the-art methods that solely rely on internal information of KGs."
                    },
                    {
                        "title": "RECOMP: Improving Retrieval-Augmented LMs with Compression and Selective Augmentation",
                        "abstract": "Retrieving documents and prepending them in-context at inference time improves performance of language model (LMs) on a wide range of tasks. However, these documents, often spanning hundreds of words, make inference substantially more expensive. We propose compressing the retrieved documents into textual summaries prior to in-context integration. This not only reduces the computational costs but also relieves the burden of LMs to identify relevant information in long retrieved documents. We present two compressors -- an extractive compressor which selects useful sentences from retrieved documents and an abstractive compressor which generates summaries by synthesizing information from multiple documents. Both compressors are trained to improve LMs' performance on end tasks when the generated summaries are prepended to the LMs' input, while keeping the summary concise.If the retrieved documents are irrelevant to the input or offer no additional information to LM, our compressor can return an empty string, implementing selective augmentation.We evaluate our approach on language modeling task and open domain question answering task. We achieve a compression rate of as low as 6% with minimal loss in performance for both tasks, significantly outperforming the off-the-shelf summarization models. We show that our compressors trained for one LM can transfer to other LMs on the language modeling task and provide summaries largely faithful to the retrieved documents."
                    },
                    {
                        "title": "Retrofitting Contextualized Word Embeddings with Paraphrases",
                        "abstract": "Contextualized word embedding models, such as ELMo, generate meaningful representations of words and their context. These models have been shown to have a great impact on downstream applications. However, in many cases, the contextualized embedding of a word changes drastically when the context is paraphrased. As a result, the downstream model is not robust to paraphrasing and other linguistic variations. To enhance the stability of contextualized word embedding models, we propose an approach to retrofitting contextualized embedding models with paraphrase contexts. Our method learns an orthogonal transformation on the input space, which seeks to minimize the variance of word representations on paraphrased contexts. Experiments show that the retrofitted model significantly outperforms the original ELMo on various sentence classification and language inference tasks."
                    },
                    {
                        "title": "Learning Bilingual Word Embeddings Using Lexical Definitions",
                        "abstract": "Bilingual word embeddings, which representlexicons of different languages in a shared em-bedding space, are essential for supporting se-mantic and knowledge transfers in a variety ofcross-lingual NLP tasks. Existing approachesto training bilingual word embeddings requireoften require pre-defined seed lexicons that areexpensive to obtain, or parallel sentences thatcomprise coarse and noisy alignment. In con-trast, we propose BilLex that leverages pub-licly available lexical definitions for bilingualword embedding learning. Without the needof predefined seed lexicons, BilLex comprisesa novel word pairing strategy to automati-cally identify and propagate the precise fine-grained word alignment from lexical defini-tions. We evaluate BilLex in word-level andsentence-level translation tasks, which seek tofind the cross-lingual counterparts of wordsand sentences respectively.BilLex signifi-cantly outperforms previous embedding meth-ods on both tasks."
                    },
                    {
                        "title": "Design Challenges in Low-resource Cross-lingual Entity Linking",
                        "abstract": "Cross-lingual Entity Linking (XEL), the problem of grounding mentions of entities in a foreign language text into an English knowledge base such as Wikipedia, has seen a lot of research in recent years, with a range of promising techniques. However, current techniques do not rise to the challenges introduced by text in low-resource languages (LRL) and, surprisingly, fail to generalize to text not taken from Wikipedia, on which they are usually trained.   This paper provides a thorough analysis of low-resource XEL techniques, focusing on the key step of identifying candidate English Wikipedia titles that correspond to a given foreign language mention. Our analysis indicates that current methods are limited by their reliance on Wikipedia's interlanguage links and thus suffer when the foreign language's Wikipedia is small. We conclude that the LRL setting requires the use of outside-Wikipedia cross-lingual resources and present a simple yet effective zero-shot XEL system, QuEL, that utilizes search engines query logs. With experiments on 25 languages, QuEL~shows an average increase of 25\\% in gold candidate recall and of 13\\% in end-to-end linking accuracy over state-of-the-art baselines."
                    },
                    {
                        "title": "On Tractable Representations of Binary Neural Networks",
                        "abstract": "We consider the compilation of a binary neural network's decision function into tractable representations such as Ordered Binary Decision Diagrams (OBDDs) and Sentential Decision Diagrams (SDDs). Obtaining this function as an OBDD/SDD facilitates the explanation and formal verification of a neural network's behavior. First, we consider the task of verifying the robustness of a neural network, and show how we can compute the expected robustness of a neural network, given an OBDD/SDD representation of it. Next, we consider a more efficient approach for compiling neural networks, based on a pseudo-polynomial time algorithm for compiling a neuron. We then provide a case study in a handwritten digits dataset, highlighting how two neural networks trained from the same dataset can have very high accuracies, yet have very different levels of robustness. Finally, in experiments, we show that it is feasible to obtain compact representations of neural networks as SDDs."
                    },
                    {
                        "title": "Predicting vs. Acting: A Trade-off Between World Modeling & Agent Modeling",
                        "abstract": "RLHF-aligned LMs have shown unprecedented ability on both benchmarks and long-form text generation, yet they struggle with one foundational task: next-token prediction. As RLHF models become agent models aimed at interacting with humans, they seem to lose their world modeling -- the ability to predict what comes next in arbitrary documents, which is the foundational training objective of the Base LMs that RLHF adapts.   Besides empirically demonstrating this trade-off, we propose a potential explanation: to perform coherent long-form generation, RLHF models restrict randomness via implicit blueprints. In particular, RLHF models concentrate probability on sets of anchor spans that co-occur across multiple generations for the same prompt, serving as textual scaffolding but also limiting a model's ability to generate documents that do not include these spans. We study this trade-off on the most effective current agent models, those aligned with RLHF, while exploring why this may remain a fundamental trade-off between models that act and those that predict, even as alignment techniques improve."
                    }
                ]
            },
            "c227896f-97f0-4c9a-91b2-e980cec60000": {
                "pk": "c227896f-97f0-4c9a-91b2-e980cec60000",
                "name": "Yangsibo Huang",
                "collaborators": [
                    "Kai Li",
                    "Zhao Song",
                    "Sanjeev Arora",
                    "Samyak Gupta",
                    "Danqi Chen",
                    "Xiaoxiao Li",
                    "Zexuan Zhong",
                    "Daogao Liu",
                    "Mengzhou Xia",
                    "Badih Ghazi"
                ],
                "domain": [
                    "Privacy Preservation",
                    "Federated Learning",
                    "Generative Models",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "IFGAN: Missing Value Imputation using Feature-specific Generative Adversarial Networks",
                        "abstract": "Missing value imputation is a challenging and well-researched topic in data mining. In this paper, we propose IFGAN, a missing value imputation algorithm based on Feature-specific Generative Adversarial Networks (GAN). Our idea is intuitive yet effective: a feature-specific generator is trained to impute missing values, while a discriminator is expected to distinguish the imputed values from observed ones. The proposed architecture is capable of handling different data types, data distributions, missing mechanisms, and missing rates. It also improves post-imputation analysis by preserving inter-feature correlations. We empirically show on several real-life datasets that IFGAN outperforms current state-of-the-art algorithm under various missing conditions."
                    },
                    {
                        "title": "InstaHide: Instance-hiding Schemes for Private Distributed Learning",
                        "abstract": "How can multiple distributed entities collaboratively train a shared deep net on their private data while preserving privacy? This paper introduces InstaHide, a simple encryption of training images, which can be plugged into existing distributed deep learning pipelines. The encryption is efficient and applying it during training has minor effect on test accuracy.   InstaHide encrypts each training image with a \"one-time secret key\" which consists of mixing a number of randomly chosen images and applying a random pixel-wise mask. Other contributions of this paper include: (a) Using a large public dataset (e.g. ImageNet) for mixing during its encryption, which improves security. (b) Experimental results to show effectiveness in preserving privacy against known attacks with only minor effects on accuracy. (c) Theoretical analysis showing that successfully attacking privacy requires attackers to solve a difficult computational problem. (d) Demonstrating that use of the pixel-wise mask is important for security, since Mixup alone is shown to be insecure to some some efficient attacks. (e) Release of a challenge dataset https://github.com/Hazelsuko07/InstaHide_Challenge   Our code is available at https://github.com/Hazelsuko07/InstaHide"
                    },
                    {
                        "title": "EMA: Auditing Data Removal from Trained Models",
                        "abstract": "Data auditing is a process to verify whether certain data have been removed from a trained model. A recently proposed method (Liu et al. 20) uses Kolmogorov-Smirnov (KS) distance for such data auditing. However, it fails under certain practical conditions. In this paper, we propose a new method called Ensembled Membership Auditing (EMA) for auditing data removal to overcome these limitations. We compare both methods using benchmark datasets (MNIST and SVHN) and Chest X-ray datasets with multi-layer perceptrons (MLP) and convolutional neural networks (CNN). Our experiments show that EMA is robust under various conditions, including the failure cases of the previously proposed method. Our code is available at: https://github.com/Hazelsuko07/EMA."
                    },
                    {
                        "title": "DeepMCDose: A Deep Learning Method for Efficient Monte Carlo Beamlet Dose Calculation by Predictive Denoising in MR-Guided Radiotherapy",
                        "abstract": "The next great leap toward improving treatment of cancer with radiation will require the combined use of online adaptive and magnetic resonance guided radiation therapy techniques with automatic X-ray beam orientation selection. Unfortunately, by uniting these advancements, we are met with a substantial expansion in the required dose information and consequential increase to the overall computational time imposed during radiation treatment planning, which cannot be handled by existing techniques for accelerating Monte Carlo dose calculation. We propose a deep convolutional neural network approach that unlocks new levels of acceleration and accuracy with regards to post-processed Monte Carlo dose results by relying on data-driven learned representations of low-level beamlet dose distributions instead of more limited filter-based denoising techniques that only utilize the information in a single dose input. Our method uses parallel UNET branches acting on three input channels before mixing latent understanding to produce noise-free dose predictions. Our model achieves a normalized mean absolute error of only 0.106% compared with the ground truth dose contrasting the 25.7% error of the under sampled MC dose fed into the network at prediction time. Our model's per-beamlet prediction time is ~220ms, including Monte Carlo simulation and network prediction, with substantial additional acceleration expected from batched processing and combination with existing Monte Carlo acceleration techniques. Our method shows promise toward enabling clinical practice of advanced treatment technologies."
                    },
                    {
                        "title": "Evaluating Gradient Inversion Attacks and Defenses in Federated Learning",
                        "abstract": "Gradient inversion attack (or input recovery from gradient) is an emerging threat to the security and privacy preservation of Federated learning, whereby malicious eavesdroppers or participants in the protocol can recover (partially) the clients' private data. This paper evaluates existing attacks and defenses. We find that some attacks make strong assumptions about the setup. Relaxing such assumptions can substantially weaken these attacks. We then evaluate the benefits of three proposed defense mechanisms against gradient inversion attacks. We show the trade-offs of privacy leakage and data utility of these defense methods, and find that combining them in an appropriate manner makes the attack less effective, even under the original strong assumptions. We also estimate the computation cost of end-to-end recovery of a single image under each evaluated defense. Our findings suggest that the state-of-the-art attacks can currently be defended against with minor data utility loss, as summarized in a list of potential strategies. Our code is available at: https://github.com/Princeton-SysML/GradAttack."
                    },
                    {
                        "title": "MixCon: Adjusting the Separability of Data Representations for Harder Data Recovery",
                        "abstract": "To address the issue that deep neural networks (DNNs) are vulnerable to model inversion attacks, we design an objective function, which adjusts the separability of the hidden data representations, as a way to control the trade-off between data utility and vulnerability to inversion attacks. Our method is motivated by the theoretical insights of data separability in neural networking training and results on the hardness of model inversion. Empirically, by adjusting the separability of data representation, we show that there exist sweet-spots for data separability such that it is difficult to recover data during inference while maintaining data utility."
                    },
                    {
                        "title": "Privacy Implications of Retrieval-Based Language Models",
                        "abstract": "Retrieval-based language models (LMs) have demonstrated improved interpretability, factuality, and adaptability compared to their parametric counterparts, by incorporating retrieved text from external datastores. While it is well known that parametric models are prone to leaking private data, it remains unclear how the addition of a retrieval datastore impacts model privacy. In this work, we present the first study of privacy risks in retrieval-based LMs, particularly $k$NN-LMs. Our goal is to explore the optimal design and training procedure in domains where privacy is of concern, aiming to strike a balance between utility and privacy. Crucially, we find that $k$NN-LMs are more susceptible to leaking private information from their private datastore than parametric models. We further explore mitigations of privacy risks. When privacy information is targeted and readily detected in the text, we find that a simple sanitization step would completely eliminate the risks, while decoupling query and key encoders achieves an even better utility-privacy trade-off. Otherwise, we consider strategies of mixing public and private data in both datastore and encoder training. While these methods offer modest improvements, they leave considerable room for future work. Together, our findings provide insights for practitioners to better understand and mitigate privacy risks in retrieval-based LMs. Our code is available at: https://github.com/Princeton-SysML/kNNLM_privacy ."
                    },
                    {
                        "title": "Catastrophic Jailbreak of Open-source LLMs via Exploiting Generation",
                        "abstract": "The rapid progress in open-source large language models (LLMs) is significantly advancing AI development. Extensive efforts have been made before model release to align their behavior with human values, with the primary goal of ensuring their helpfulness and harmlessness. However, even carefully aligned models can be manipulated maliciously, leading to unintended behaviors, known as \"jailbreaks\". These jailbreaks are typically triggered by specific text inputs, often referred to as adversarial prompts. In this work, we propose the generation exploitation attack, an extremely simple approach that disrupts model alignment by only manipulating variations of decoding methods. By exploiting different generation strategies, including varying decoding hyper-parameters and sampling methods, we increase the misalignment rate from 0% to more than 95% across 11 language models including LLaMA2, Vicuna, Falcon, and MPT families, outperforming state-of-the-art attacks with $30\\times$ lower computational cost. Finally, we propose an effective alignment method that explores diverse generation strategies, which can reasonably reduce the misalignment rate under our attack. Altogether, our study underscores a major failure in current safety evaluation and alignment procedures for open-source LLMs, strongly advocating for more comprehensive red teaming and better alignment before releasing such models. Our code is available at https://github.com/Princeton-SysML/Jailbreak_LLM."
                    },
                    {
                        "title": "Deep Q Learning Driven CT Pancreas Segmentation with Geometry-Aware U-Net",
                        "abstract": "Segmentation of pancreas is important for medical image analysis, yet it faces great challenges of class imbalance, background distractions and non-rigid geometrical features. To address these difficulties, we introduce a Deep Q Network(DQN) driven approach with deformable U-Net to accurately segment the pancreas by explicitly interacting with contextual information and extract anisotropic features from pancreas. The DQN based model learns a context-adaptive localization policy to produce a visually tightened and precise localization bounding box of the pancreas. Furthermore, deformable U-Net captures geometry-aware information of pancreas by learning geometrically deformable filters for feature extraction. Experiments on NIH dataset validate the effectiveness of the proposed framework in pancreas segmentation."
                    },
                    {
                        "title": "TextHide: Tackling Data Privacy in Language Understanding Tasks",
                        "abstract": "An unsolved challenge in distributed or federated learning is to effectively mitigate privacy risks without slowing down training or reducing accuracy. In this paper, we propose TextHide aiming at addressing this challenge for natural language understanding tasks. It requires all participants to add a simple encryption step to prevent an eavesdropping attacker from recovering private text data. Such an encryption step is efficient and only affects the task performance slightly. In addition, TextHide fits well with the popular framework of fine-tuning pre-trained language models (e.g., BERT) for any sentence or sentence-pair task. We evaluate TextHide on the GLUE benchmark, and our experiments show that TextHide can effectively defend attacks on shared gradients or representations and the averaged accuracy reduction is only $1.9\\%$. We also present an analysis of the security of TextHide using a conjecture about the computational intractability of a mathematical problem.   Our code is available at https://github.com/Hazelsuko07/TextHide"
                    },
                    {
                        "title": "ConceptMix: A Compositional Image Generation Benchmark with Controllable Difficulty",
                        "abstract": "Compositionality is a critical capability in Text-to-Image (T2I) models, as it reflects their ability to understand and combine multiple concepts from text descriptions. Existing evaluations of compositional capability rely heavily on human-designed text prompts or fixed templates, limiting their diversity and complexity, and yielding low discriminative power. We propose ConceptMix, a scalable, controllable, and customizable benchmark which automatically evaluates compositional generation ability of T2I models. This is done in two stages. First, ConceptMix generates the text prompts: concretely, using categories of visual concepts (e.g., objects, colors, shapes, spatial relationships), it randomly samples an object and k-tuples of visual concepts, then uses GPT4-o to generate text prompts for image generation based on these sampled concepts. Second, ConceptMix evaluates the images generated in response to these prompts: concretely, it checks how many of the k concepts actually appeared in the image by generating one question per visual concept and using a strong VLM to answer them. Through administering ConceptMix to a diverse set of T2I models (proprietary as well as open ones) using increasing values of k, we show that our ConceptMix has higher discrimination power than earlier benchmarks. Specifically, ConceptMix reveals that the performance of several models, especially open models, drops dramatically with increased k. Importantly, it also provides insight into the lack of prompt diversity in widely-used training datasets. Additionally, we conduct extensive human studies to validate the design of ConceptMix and compare our automatic grading with human judgement. We hope it will guide future T2I model development."
                    },
                    {
                        "title": "Privacy-preserving Learning via Deep Net Pruning",
                        "abstract": "This paper attempts to answer the question whether neural network pruning can be used as a tool to achieve differential privacy without losing much data utility. As a first step towards understanding the relationship between neural network pruning and differential privacy, this paper proves that pruning a given layer of the neural network is equivalent to adding a certain amount of differentially private noise to its hidden-layer activations. The paper also presents experimental results to show the practical implications of the theoretical finding and the key parameter values in a simple practical setting. These results show that neural network pruning can be a more effective alternative to adding differentially private noise for neural networks."
                    },
                    {
                        "title": "$k$NN-Adapter: Efficient Domain Adaptation for Black-Box Language Models",
                        "abstract": "Fine-tuning a language model on a new domain is standard practice for domain adaptation. However, it can be infeasible when it comes to modern large-scale language models such as GPT-3, which can only be accessed through APIs, making it difficult to access the internal parameters of the model. In this paper, we propose $k$NN-Adapter, a method to effectively adapt these black-box large language models (LLMs) to a new domain. The $k$NN-Adapter builds on top of the retrieval-augmented language model, and adaptively learns to interpolate the output of the language model with retrieval results from a datastore consisting of the target domain data. Our experiments on four different domains demonstrate that $k$NN-Adapter significantly improves perplexity, and works particularly well in settings with limited access to LLMs. Additionally, we show that $k$NN-Adapter is more effective than fine-tuning when the amount of training data is limited. We also release a dataset to encourage further study."
                    },
                    {
                        "title": "Sparsity-Preserving Differentially Private Training of Large Embedding Models",
                        "abstract": "As the use of large embedding models in recommendation systems and language applications increases, concerns over user data privacy have also risen. DP-SGD, a training algorithm that combines differential privacy with stochastic gradient descent, has been the workhorse in protecting user privacy without compromising model accuracy by much. However, applying DP-SGD naively to embedding models can destroy gradient sparsity, leading to reduced training efficiency. To address this issue, we present two new algorithms, DP-FEST and DP-AdaFEST, that preserve gradient sparsity during private training of large embedding models. Our algorithms achieve substantial reductions ($10^6 \\times$) in gradient size, while maintaining comparable levels of accuracy, on benchmark real-world datasets."
                    },
                    {
                        "title": "Learning across Data Owners with Joint Differential Privacy",
                        "abstract": "In this paper, we study the setting in which data owners train machine learning models collaboratively under a privacy notion called joint differential privacy [Kearns et al., 2018]. In this setting, the model trained for each data owner $j$ uses $j$'s data without privacy consideration and other owners' data with differential privacy guarantees. This setting was initiated in [Jain et al., 2021] with a focus on linear regressions. In this paper, we study this setting for stochastic convex optimization (SCO). We present an algorithm that is a variant of DP-SGD [Song et al., 2013; Abadi et al., 2016] and provides theoretical bounds on its population loss. We compare our algorithm to several baselines and discuss for what parameter setups our algorithm is more preferred. We also empirically study joint differential privacy in the multi-class classification problem over two public datasets. Our empirical findings are well-connected to the insights from our theoretical results."
                    },
                    {
                        "title": "An Adversarial Perspective on Machine Unlearning for AI Safety",
                        "abstract": "Large language models are finetuned to refuse questions about hazardous knowledge, but these protections can often be bypassed. Unlearning methods aim at completely removing hazardous capabilities from models and make them inaccessible to adversaries. This work challenges the fundamental differences between unlearning and traditional safety post-training from an adversarial perspective. We demonstrate that existing jailbreak methods, previously reported as ineffective against unlearning, can be successful when applied carefully. Furthermore, we develop a variety of adaptive methods that recover most supposedly unlearned capabilities. For instance, we show that finetuning on 10 unrelated examples or removing specific directions in the activation space can recover most hazardous capabilities for models edited with RMU, a state-of-the-art unlearning method. Our findings challenge the robustness of current unlearning approaches and question their advantages over safety training."
                    },
                    {
                        "title": "Recovering Private Text in Federated Learning of Language Models",
                        "abstract": "Federated learning allows distributed users to collaboratively train a model while keeping each user's data private. Recently, a growing body of work has demonstrated that an eavesdropping attacker can effectively recover image data from gradients transmitted during federated learning. However, little progress has been made in recovering text data. In this paper, we present a novel attack method FILM for federated learning of language models (LMs). For the first time, we show the feasibility of recovering text from large batch sizes of up to 128 sentences. Unlike image-recovery methods that are optimized to match gradients, we take a distinct approach that first identifies a set of words from gradients and then directly reconstructs sentences based on beam search and a prior-based reordering strategy. We conduct the FILM attack on several large-scale datasets and show that it can successfully reconstruct single sentences with high fidelity for large batch sizes and even multiple sentences if applied iteratively. We evaluate three defense methods: gradient pruning, DPSGD, and a simple approach to freeze word embeddings that we propose. We show that both gradient pruning and DPSGD lead to a significant drop in utility. However, if we fine-tune a public pre-trained LM on private text without updating word embeddings, it can effectively defend the attack with minimal data utility loss. Together, we hope that our results can encourage the community to rethink the privacy concerns of LM training and its standard practices in the future."
                    },
                    {
                        "title": "GMValuator: Similarity-based Data Valuation for Generative Models",
                        "abstract": "Data valuation plays a crucial role in machine learning. Existing data valuation methods have primarily focused on discriminative models, neglecting generative models that have recently gained considerable attention. A very few existing attempts of data valuation method designed for deep generative models either concentrates on specific models or lacks robustness in their outcomes. Moreover, efficiency still reveals vulnerable shortcomings. To bridge the gaps, we formulate the data valuation problem in generative models from a similarity-matching perspective. Specifically, we introduce Generative Model Valuator (GMValuator), the first training-free and model-agnostic approach to provide data valuation for generation tasks. It empowers efficient data valuation through our innovatively similarity matching module, calibrates biased contribution by incorporating image quality assessment, and attributes credits to all training samples based on their contributions to the generated samples. Additionally, we introduce four evaluation criteria for assessing data valuation methods in generative models, aligning with principles of plausibility and truthfulness. GMValuator is extensively evaluated on various datasets and generative architectures to demonstrate its effectiveness."
                    },
                    {
                        "title": "Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modifications",
                        "abstract": "Large language models (LLMs) show inherent brittleness in their safety mechanisms, as evidenced by their susceptibility to jailbreaking and even non-malicious fine-tuning. This study explores this brittleness of safety alignment by leveraging pruning and low-rank modifications. We develop methods to identify critical regions that are vital for safety guardrails, and that are disentangled from utility-relevant regions at both the neuron and rank levels. Surprisingly, the isolated regions we find are sparse, comprising about $3\\%$ at the parameter level and $2.5\\%$ at the rank level. Removing these regions compromises safety without significantly impacting utility, corroborating the inherent brittleness of the model's safety mechanisms. Moreover, we show that LLMs remain vulnerable to low-cost fine-tuning attacks even when modifications to the safety-critical regions are restricted. These findings underscore the urgent need for more robust safety strategies in LLMs."
                    },
                    {
                        "title": "Mind the Privacy Unit! User-Level Differential Privacy for Language Model Fine-Tuning",
                        "abstract": "Large language models (LLMs) have emerged as powerful tools for tackling complex tasks across diverse domains, but they also raise privacy concerns when fine-tuned on sensitive data due to potential memorization. While differential privacy (DP) offers a promising solution by ensuring models are 'almost indistinguishable' with or without any particular privacy unit, current evaluations on LLMs mostly treat each example (text record) as the privacy unit. This leads to uneven user privacy guarantees when contributions per user vary. We therefore study user-level DP motivated by applications where it necessary to ensure uniform privacy protection across users. We present a systematic evaluation of user-level DP for LLM fine-tuning on natural language generation tasks. Focusing on two mechanisms for achieving user-level DP guarantees, Group Privacy and User-wise DP-SGD, we investigate design choices like data selection strategies and parameter tuning for the best privacy-utility tradeoff."
                    }
                ]
            },
            "1e42ad0f-1332-4dbd-8fe1-522b8bbad514": {
                "pk": "1e42ad0f-1332-4dbd-8fe1-522b8bbad514",
                "name": "Noah A. Smith",
                "collaborators": [],
                "domain": [
                    "Natural Language Processing",
                    "Word Embeddings",
                    "Evaluation Framework"
                ],
                "publications": [
                    {
                        "title": "Contextual Word Representations: A Contextual Introduction",
                        "abstract": "This introduction aims to tell the story of how we put words into computers. It is part of the story of the field of natural language processing (NLP), a branch of artificial intelligence. It targets a wide audience with a basic understanding of computer programming, but avoids a detailed mathematical treatment, and it does not present any algorithms. It also does not focus on any particular application of NLP such as translation, question answering, or information extraction. The ideas presented here were developed by many researchers over many decades, so the citations are not exhaustive but rather direct the reader to a handful of papers that are, in the author's view, seminal. After reading this document, you should have a general understanding of word vectors (also known as word embeddings): why they exist, what problems they solve, where they come from, how they have changed over time, and what some of the open questions about them are. Readers already familiar with word vectors are advised to skip to Section 5 for the discussion of the most recent advance, contextual word vectors."
                    },
                    {
                        "title": "Adversarial Evaluation for Models of Natural Language",
                        "abstract": "We now have a rich and growing set of modeling tools and algorithms for inducing linguistic structure from text that is less than fully annotated. In this paper, we discuss some of the weaknesses of our current methodology. We present a new abstract framework for evaluating natural language processing (NLP) models in general and unsupervised NLP models in particular. The central idea is to make explicit certain adversarial roles among researchers, so that the different roles in an evaluation are more clearly defined and performers of all roles are offered ways to make measurable contributions to the larger goal. Adopting this approach may help to characterize model successes and failures by encouraging earlier consideration of error analysis. The framework can be instantiated in a variety of ways, simulating some familiar intrinsic and extrinsic evaluations as well as some new evaluations."
                    }
                ]
            },
            "689e8f56-c9c0-4aab-83f1-3278e7a35c72": {
                "pk": "689e8f56-c9c0-4aab-83f1-3278e7a35c72",
                "name": "Chiyuan Zhang",
                "collaborators": [
                    "Samy Bengio",
                    "Tomaso Poggio",
                    "Maithra Raghu",
                    "Michael C. Mozer",
                    "Yoram Singer",
                    "Jon Kleinberg",
                    "Badih Ghazi",
                    "Noah Golowich",
                    "Ravi Kumar",
                    "Pasin Manurangsi"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Transfer Learning",
                    "Generalization"
                ],
                "publications": [
                    {
                        "title": "What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation",
                        "abstract": "Deep learning algorithms are well-known to have a propensity for fitting the training data very well and often fit even outliers and mislabeled data points. Such fitting requires memorization of training data labels, a phenomenon that has attracted significant research interest but has not been given a compelling explanation so far. A recent work of Feldman (2019) proposes a theoretical explanation for this phenomenon based on a combination of two insights. First, natural image and data distributions are (informally) known to be long-tailed, that is have a significant fraction of rare and atypical examples. Second, in a simple theoretical model such memorization is necessary for achieving close-to-optimal generalization error when the data distribution is long-tailed. However, no direct empirical evidence for this explanation or even an approach for obtaining such evidence were given.   In this work we design experiments to test the key ideas in this theory. The experiments require estimation of the influence of each training example on the accuracy at each test example as well as memorization values of training examples. Estimating these quantities directly is computationally prohibitive but we show that closely-related subsampled influence and memorization values can be estimated much more efficiently. Our experiments demonstrate the significant benefits of memorization for generalization on several standard benchmarks. They also provide quantitative and visually compelling evidence for the theory put forth in (Feldman, 2019)."
                    },
                    {
                        "title": "Are All Layers Created Equal?",
                        "abstract": "Understanding deep neural networks is a major research objective with notable experimental and theoretical attention in recent years. The practical success of excessively large networks underscores the need for better theoretical analyses and justifications. In this paper we focus on layer-wise functional structure and behavior in overparameterized deep models. To do so, we study empirically the layers' robustness to post-training re-initialization and re-randomization of the parameters. We provide experimental results which give evidence for the heterogeneity of layers. Morally, layers of large deep neural networks can be categorized as either \"robust\" or \"critical\". Resetting the robust layers to their initial values does not result in adverse decline in performance. In many cases, robust layers hardly change throughout training. In contrast, re-initializing critical layers vastly degrades the performance of the network with test error essentially dropping to random guesses. Our study provides further evidence that mere parameter counting or norm calculations are too coarse in studying generalization of deep models, and \"flatness\" and robustness analysis of trained models need to be examined while taking into account the respective network architectures."
                    },
                    {
                        "title": "Pointer Value Retrieval: A new benchmark for understanding the limits of neural network generalization",
                        "abstract": "Central to the success of artificial neural networks is their ability to generalize. But does neural network generalization primarily rely on seeing highly similar training examples (memorization)? Or are neural networks capable of human-intelligence styled reasoning, and if so, to what extent? These remain fundamental open questions on artificial neural networks. In this paper, as steps towards answering these questions, we introduce a new benchmark, Pointer Value Retrieval (PVR) to study the limits of neural network reasoning. The PVR suite of tasks is based on reasoning about indirection, a hallmark of human intelligence, where a first stage (task) contains instructions for solving a second stage (task). In PVR, this is done by having one part of the task input act as a pointer, giving instructions on a different input location, which forms the output. We show this simple rule can be applied to create a diverse set of tasks across different input modalities and configurations. Importantly, this use of indirection enables systematically varying task difficulty through distribution shifts and increasing functional complexity. We conduct a detailed empirical study of different PVR tasks, discovering large variations in performance across dataset sizes, neural network architectures and task complexity. Further, by incorporating distribution shift and increased functional complexity, we develop nuanced tests for reasoning, revealing subtle failures and surprising successes, suggesting many promising directions of exploration on this benchmark."
                    },
                    {
                        "title": "What is being transferred in transfer learning?",
                        "abstract": "One desired capability for machines is the ability to transfer their knowledge of one domain to another where data is (usually) scarce. Despite ample adaptation of transfer learning in various deep learning applications, we yet do not understand what enables a successful transfer and which part of the network is responsible for that. In this paper, we provide new tools and analyses to address these fundamental questions. Through a series of analyses on transferring to block-shuffled images, we separate the effect of feature reuse from learning low-level statistics of data and show that some benefit of transfer learning comes from the latter. We present that when training from pre-trained weights, the model stays in the same basin in the loss landscape and different instances of such model are similar in feature space and close in parameter space."
                    },
                    {
                        "title": "Localizing Paragraph Memorization in Language Models",
                        "abstract": "Can we localize the weights and mechanisms used by a language model to memorize and recite entire paragraphs of its training data? In this paper, we show that while memorization is spread across multiple layers and model components, gradients of memorized paragraphs have a distinguishable spatial pattern, being larger in lower model layers than gradients of non-memorized examples. Moreover, the memorized examples can be unlearned by fine-tuning only the high-gradient weights. We localize a low-layer attention head that appears to be especially involved in paragraph memorization. This head is predominantly focusing its attention on distinctive, rare tokens that are least frequent in a corpus-level unigram distribution. Next, we study how localized memorization is across the tokens in the prefix by perturbing tokens and measuring the caused change in the decoding. A few distinctive tokens early in a prefix can often corrupt the entire continuation. Overall, memorized continuations are not only harder to unlearn, but also to corrupt than non-memorized ones."
                    },
                    {
                        "title": "Transfusion: Understanding Transfer Learning for Medical Imaging",
                        "abstract": "Transfer learning from natural image datasets, particularly ImageNet, using standard large models and corresponding pretrained weights has become a de-facto method for deep learning applications to medical imaging. However, there are fundamental differences in data sizes, features and task specifications between natural image classification and the target medical tasks, and there is little understanding of the effects of transfer. In this paper, we explore properties of transfer learning for medical imaging. A performance evaluation on two large scale medical imaging tasks shows that surprisingly, transfer offers little benefit to performance, and simple, lightweight models can perform comparably to ImageNet architectures. Investigating the learned representations and features, we find that some of the differences from transfer learning are due to the over-parametrization of standard models rather than sophisticated feature reuse. We isolate where useful feature reuse occurs, and outline the implications for more efficient model exploration. We also explore feature independent benefits of transfer arising from weight scalings."
                    },
                    {
                        "title": "Characterizing Structural Regularities of Labeled Data in Overparameterized Models",
                        "abstract": "Humans are accustomed to environments that contain both regularities and exceptions. For example, at most gas stations, one pays prior to pumping, but the occasional rural station does not accept payment in advance. Likewise, deep neural networks can generalize across instances that share common patterns or structures, yet have the capacity to memorize rare or irregular forms. We analyze how individual instances are treated by a model via a consistency score. The score characterizes the expected accuracy for a held-out instance given training sets of varying size sampled from the data distribution. We obtain empirical estimates of this score for individual instances in multiple data sets, and we show that the score identifies out-of-distribution and mislabeled examples at one end of the continuum and strongly regular examples at the other end. We identify computationally inexpensive proxies to the consistency score using statistics collected during training. We show examples of potential applications to the analysis of deep-learning systems."
                    },
                    {
                        "title": "Deep Learning with Label Differential Privacy",
                        "abstract": "The Randomized Response (RR) algorithm is a classical technique to improve robustness in survey aggregation, and has been widely adopted in applications with differential privacy guarantees. We propose a novel algorithm, Randomized Response with Prior (RRWithPrior), which can provide more accurate results while maintaining the same level of privacy guaranteed by RR. We then apply RRWithPrior to learn neural networks with label differential privacy (LabelDP), and show that when only the label needs to be protected, the model performance can be significantly improved over the previous state-of-the-art private baselines. Moreover, we study different ways to obtain priors, which when used with RRWithPrior can additionally improve the model performance, further reducing the accuracy gap between private and non-private models. We complement the empirical results with theoretical analysis showing that LabelDP is provably easier than protecting both the inputs and labels."
                    },
                    {
                        "title": "Understanding deep learning requires rethinking generalization",
                        "abstract": "Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family, or to the regularization techniques used during training.   Through extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization, and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice.   We interpret our experimental findings by comparison with traditional models."
                    },
                    {
                        "title": "A Study on Overfitting in Deep Reinforcement Learning",
                        "abstract": "Recent years have witnessed significant progresses in deep Reinforcement Learning (RL). Empowered with large scale neural networks, carefully designed architectures, novel training algorithms and massively parallel computing devices, researchers are able to attack many challenging RL problems. However, in machine learning, more training power comes with a potential risk of more overfitting. As deep RL techniques are being applied to critical problems such as healthcare and finance, it is important to understand the generalization behaviors of the trained agents. In this paper, we conduct a systematic study of standard RL agents and find that they could overfit in various ways. Moreover, overfitting could happen \"robustly\": commonly used techniques in RL that add stochasticity do not necessarily prevent or detect overfitting. In particular, the same agents and learning algorithms could have drastically different test performance, even when all of them achieve optimal rewards during training. The observations call for more principled and careful evaluation protocols in RL. We conclude with a general discussion on overfitting in RL and a study of the generalization behaviors from the perspective of inductive bias."
                    },
                    {
                        "title": "A Deep Representation for Invariance And Music Classification",
                        "abstract": "Representations in the auditory cortex might be based on mechanisms similar to the visual ventral stream; modules for building invariance to transformations and multiple layers for compositionality and selectivity. In this paper we propose the use of such computational modules for extracting invariant and discriminative audio representations. Building on a theory of invariance in hierarchical architectures, we propose a novel, mid-level representation for acoustical signals, using the empirical distributions of projections on a set of templates and their transformations. Under the assumption that, by construction, this dictionary of templates is composed from similar classes, and samples the orbit of variance-inducing signal transformations (such as shift and scale), the resulting signature is theoretically guaranteed to be unique, invariant to transformations and stable to deformations. Modules of projection and pooling can then constitute layers of deep networks, for learning composite representations. We present the main theoretical and computational aspects of a framework for unsupervised learning of invariant audio representations, empirically evaluated on music genre classification."
                    },
                    {
                        "title": "Learning with a Wasserstein Loss",
                        "abstract": "Learning to predict multi-label outputs is challenging, but in many problems there is a natural metric on the outputs that can be used to improve predictions. In this paper we develop a loss function for multi-label learning, based on the Wasserstein distance. The Wasserstein distance provides a natural notion of dissimilarity for probability measures. Although optimizing with respect to the exact Wasserstein distance is costly, recent work has described a regularized approximation that is efficiently computed. We describe an efficient learning algorithm based on this regularization, as well as a novel extension of the Wasserstein distance from probability measures to unnormalized measures. We also describe a statistical learning bound for the loss. The Wasserstein loss can encourage smoothness of the predictions with respect to a chosen metric on the output space. We demonstrate this property on a real-data tag prediction problem, using the Yahoo Flickr Creative Commons dataset, outperforming a baseline that doesn't use the metric."
                    },
                    {
                        "title": "Do Vision Transformers See Like Convolutional Neural Networks?",
                        "abstract": "Convolutional neural networks (CNNs) have so far been the de-facto model for visual data. Recent work has shown that (Vision) Transformer models (ViT) can achieve comparable or even superior performance on image classification tasks. This raises a central question: how are Vision Transformers solving these tasks? Are they acting like convolutional networks, or learning entirely different visual representations? Analyzing the internal representation structure of ViTs and CNNs on image classification benchmarks, we find striking differences between the two architectures, such as ViT having more uniform representations across all layers. We explore how these differences arise, finding crucial roles played by self-attention, which enables early aggregation of global information, and ViT residual connections, which strongly propagate features from lower to higher layers. We study the ramifications for spatial localization, demonstrating ViTs successfully preserve input spatial information, with noticeable effects from different classification methods. Finally, we study the effect of (pretraining) dataset scale on intermediate features and transfer learning, and conclude with a discussion on connections to new architectures such as the MLP-Mixer."
                    },
                    {
                        "title": "Learning An Invariant Speech Representation",
                        "abstract": "Recognition of speech, and in particular the ability to generalize and learn from small sets of labelled examples like humans do, depends on an appropriate representation of the acoustic input. We formulate the problem of finding robust speech features for supervised learning with small sample complexity as a problem of learning representations of the signal that are maximally invariant to intraclass transformations and deformations. We propose an extension of a theory for unsupervised learning of invariant visual representations to the auditory domain and empirically evaluate its validity for voiced speech sound classification. Our version of the theory requires the memory-based, unsupervised storage of acoustic templates -- such as specific phones or words -- together with all the transformations of each that normally occur. A quasi-invariant representation for a speech segment can be obtained by projecting it to each template orbit, i.e., the set of transformed signals, and computing the associated one-dimensional empirical probability distributions. The computations can be performed by modules of filtering and pooling, and extended to hierarchical architectures. In this paper, we apply a single-layer, multicomponent representation for phonemes and demonstrate improved accuracy and decreased sample complexity for vowel classification compared to standard spectral, cepstral and perceptual features."
                    },
                    {
                        "title": "Training Deep Nets with Sublinear Memory Cost",
                        "abstract": "We propose a systematic approach to reduce the memory consumption of deep neural network training. Specifically, we design an algorithm that costs O(sqrt(n)) memory to train a n layer network, with only the computational cost of an extra forward pass per mini-batch. As many of the state-of-the-art models hit the upper bound of the GPU memory, our algorithm allows deeper and more complex models to be explored, and helps advance the innovations in deep learning research. We focus on reducing the memory cost to store the intermediate feature maps and gradients during training. Computation graph analysis is used for automatic in-place operation and memory sharing optimizations. We show that it is possible to trade computation for memory - giving a more memory efficient training algorithm with a little extra computation cost. In the extreme case, our analysis also shows that the memory consumption can be reduced to O(log n) with as little as O(n log n) extra cost for forward computation. Our experiments show that we can reduce the memory cost of a 1,000-layer deep residual network from 48G to 7G with only 30 percent additional running time cost on ImageNet problems. Similarly, significant memory cost reduction is observed in training complex recurrent neural networks on very long sequences."
                    },
                    {
                        "title": "Understanding Invariance via Feedforward Inversion of Discriminatively Trained Classifiers",
                        "abstract": "A discriminatively trained neural net classifier can fit the training data perfectly if all information about its input other than class membership has been discarded prior to the output layer. Surprisingly, past research has discovered that some extraneous visual detail remains in the logit vector. This finding is based on inversion techniques that map deep embeddings back to images. We explore this phenomenon further using a novel synthesis of methods, yielding a feedforward inversion model that produces remarkably high fidelity reconstructions, qualitatively superior to those of past efforts. When applied to an adversarially robust classifier model, the reconstructions contain sufficient local detail and global structure that they might be confused with the original image in a quick glance, and the object category can clearly be gleaned from the reconstruction. Our approach is based on BigGAN (Brock, 2019), with conditioning on logits instead of one-hot class labels. We use our reconstruction model as a tool for exploring the nature of representations, including: the influence of model architecture and training objectives (specifically robust losses), the forms of invariance that networks achieve, representational differences between correctly and incorrectly classified images, and the effects of manipulating logits and images. We believe that our method can inspire future investigations into the nature of information flow in a neural net and can provide diagnostics for improving discriminative models."
                    },
                    {
                        "title": "Counterfactual Memorization in Neural Language Models",
                        "abstract": "Modern neural language models that are widely used in various NLP tasks risk memorizing sensitive information from their training data. Understanding this memorization is important in real world applications and also from a learning-theoretical perspective. An open question in previous studies of language model memorization is how to filter out \"common\" memorization. In fact, most memorization criteria strongly correlate with the number of occurrences in the training set, capturing memorized familiar phrases, public knowledge, templated texts, or other repeated data. We formulate a notion of counterfactual memorization which characterizes how a model's predictions change if a particular document is omitted during training. We identify and study counterfactually-memorized training examples in standard text datasets. We estimate the influence of each memorized training example on the validation set and on generated texts, showing how this can provide direct evidence of the source of memorization at test time."
                    },
                    {
                        "title": "Theory of Deep Learning IIb: Optimization Properties of SGD",
                        "abstract": "In Theory IIb we characterize with a mix of theory and experiments the optimization of deep convolutional networks by Stochastic Gradient Descent. The main new result in this paper is theoretical and experimental evidence for the following conjecture about SGD: SGD concentrates in probability -- like the classical Langevin equation -- on large volume, \"flat\" minima, selecting flat minimizers which are with very high probability also global minimizers"
                    },
                    {
                        "title": "On Convex Optimization with Semi-Sensitive Features",
                        "abstract": "We study the differentially private (DP) empirical risk minimization (ERM) problem under the semi-sensitive DP setting where only some features are sensitive. This generalizes the Label DP setting where only the label is sensitive. We give improved upper and lower bounds on the excess risk for DP-ERM. In particular, we show that the error only scales polylogarithmically in terms of the sensitive domain size, improving upon previous results that scale polynomially in the sensitive domain size (Ghazi et al., 2021)."
                    },
                    {
                        "title": "Identity Crisis: Memorization and Generalization under Extreme Overparameterization",
                        "abstract": "We study the interplay between memorization and generalization of overparameterized networks in the extreme case of a single training example and an identity-mapping task. We examine fully-connected and convolutional networks (FCN and CNN), both linear and nonlinear, initialized randomly and then trained to minimize the reconstruction error. The trained networks stereotypically take one of two forms: the constant function (memorization) and the identity function (generalization). We formally characterize generalization in single-layer FCNs and CNNs. We show empirically that different architectures exhibit strikingly different inductive biases. For example, CNNs of up to 10 layers are able to generalize from a single example, whereas FCNs cannot learn the identity function reliably from 60k examples. Deeper CNNs often fail, but nonetheless do astonishing work to memorize the training output: because CNN biases are location invariant, the model must progressively grow an output pattern from the image boundaries via the coordination of many layers. Our work helps to quantify and visualize the sensitivity of inductive biases to architectural choices such as depth, kernel width, and number of channels."
                    }
                ]
            },
            "ee5c7a10-bea7-490f-8553-c47e055b555f": {
                "pk": "ee5c7a10-bea7-490f-8553-c47e055b555f",
                "name": "Luke Zettlemoyer",
                "collaborators": [
                    "Kenton Lee",
                    "Terra Blevins",
                    "Luheng He",
                    "Omer Levy",
                    "Tim Dettmers",
                    "Mike Lewis",
                    "Nicholas FitzGerald",
                    "Srinivasan Iyer",
                    "Alvin Cheung",
                    "Michael Petrochuk"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Deep Learning",
                    "Information Extraction",
                    "Coreference Resolution"
                ],
                "publications": [
                    {
                        "title": "Better Character Language Modeling Through Morphology",
                        "abstract": "We incorporate morphological supervision into character language models (CLMs) via multitasking and show that this addition improves bits-per-character (BPC) performance across 24 languages, even when the morphology data and language modeling data are disjoint. Analyzing the CLMs shows that inflected words benefit more from explicitly modeling morphology than uninflected words, and that morphological supervision improves performance even as the amount of language modeling data grows. We then transfer morphological supervision across languages to improve language modeling performance in the low-resource setting."
                    },
                    {
                        "title": "Sparse Networks from Scratch: Faster Training without Losing Performance",
                        "abstract": "We demonstrate the possibility of what we call sparse learning: accelerated training of deep neural networks that maintain sparse weights throughout training while achieving dense performance levels. We accomplish this by developing sparse momentum, an algorithm which uses exponentially smoothed gradients (momentum) to identify layers and weights which reduce the error efficiently. Sparse momentum redistributes pruned weights across layers according to the mean momentum magnitude of each layer. Within a layer, sparse momentum grows weights according to the momentum magnitude of zero-valued weights. We demonstrate state-of-the-art sparse performance on MNIST, CIFAR-10, and ImageNet, decreasing the mean error by a relative 8%, 15%, and 6% compared to other sparse algorithms. Furthermore, we show that sparse momentum reliably reproduces dense performance levels while providing up to 5.61x faster training. In our analysis, ablations show that the benefits of momentum redistribution and growth increase with the depth and size of the network. Additionally, we find that sparse momentum is insensitive to the choice of its hyperparameters suggesting that sparse momentum is robust and easy to use."
                    },
                    {
                        "title": "Moving Down the Long Tail of Word Sense Disambiguation with Gloss-Informed Biencoders",
                        "abstract": "A major obstacle in Word Sense Disambiguation (WSD) is that word senses are not uniformly distributed, causing existing models to generally perform poorly on senses that are either rare or unseen during training. We propose a bi-encoder model that independently embeds (1) the target word with its surrounding context and (2) the dictionary definition, or gloss, of each sense. The encoders are jointly optimized in the same representation space, so that sense disambiguation can be performed by finding the nearest sense embedding for each target word embedding. Our system outperforms previous state-of-the-art models on English all-words WSD; these gains predominantly come from improved performance on rare senses, leading to a 31.1% error reduction on less frequent senses over prior work. This demonstrates that rare senses can be more effectively disambiguated by modeling their definitions."
                    },
                    {
                        "title": "Language Contamination Helps Explain the Cross-lingual Capabilities of English Pretrained Models",
                        "abstract": "English pretrained language models, which make up the backbone of many modern NLP systems, require huge amounts of unlabeled training data. These models are generally presented as being trained only on English text but have been found to transfer surprisingly well to other languages. We investigate this phenomenon and find that common English pretraining corpora actually contain significant amounts of non-English text: even when less than 1% of data is not English (well within the error rate of strong language classifiers), this leads to hundreds of millions of foreign language tokens in large-scale datasets. We then demonstrate that even these small percentages of non-English data facilitate cross-lingual transfer for models trained on them, with target language performance strongly correlated to the amount of in-language data seen during pretraining. In light of these findings, we argue that no model is truly monolingual when pretrained at scale, which should be considered when evaluating cross-lingual transfer."
                    },
                    {
                        "title": "SimpleQuestions Nearly Solved: A New Upperbound and Baseline Approach",
                        "abstract": "The SimpleQuestions dataset is one of the most commonly used benchmarks for studying single-relation factoid questions. In this paper, we present new evidence that this benchmark can be nearly solved by standard methods. First we show that ambiguity in the data bounds performance on this benchmark at 83.4%; there are often multiple answers that cannot be disambiguated from the linguistic signal alone. Second we introduce a baseline that sets a new state-of-the-art performance level at 78.1% accuracy, despite using standard methods. Finally, we report an empirical analysis showing that the upperbound is loose; roughly a third of the remaining errors are also not resolvable from the linguistic signal. Together, these results suggest that the SimpleQuestions dataset is nearly solved."
                    },
                    {
                        "title": "E3: Entailment-driven Extracting and Editing for Conversational Machine Reading",
                        "abstract": "Conversational machine reading systems help users answer high-level questions (e.g. determine if they qualify for particular government benefits) when they do not know the exact rules by which the determination is made(e.g. whether they need certain income levels or veteran status). The key challenge is that these rules are only provided in the form of a procedural text (e.g. guidelines from government website) which the system must read to figure out what to ask the user. We present a new conversational machine reading model that jointly extracts a set of decision rules from the procedural text while reasoning about which are entailed by the conversational history and which still need to be edited to create questions for the user. On the recently introduced ShARC conversational machine reading dataset, our Entailment-driven Extract and Edit network (E3) achieves a new state-of-the-art, outperforming existing systems as well as a new BERT-based baseline. In addition, by explicitly highlighting which information still needs to be gathered, E3 provides a more explainable alternative to prior work. We release source code for our models and experiments at https://github.com/vzhong/e3."
                    },
                    {
                        "title": "The case for 4-bit precision: k-bit Inference Scaling Laws",
                        "abstract": "Quantization methods reduce the number of bits required to represent each parameter in a model, trading accuracy for smaller memory footprints and inference latencies. However, the final model size depends on both the number of parameters of the original model and the rate of compression. For example, a 30B 8-bit model and a 60B 4-bit model have the same number of bits but may have very different zero-shot accuracies. In this work, we study this trade-off by developing inference scaling laws of zero-shot performance in Large Language Models (LLMs) to determine the bit-precision and model size that maximizes zero-shot performance. We run more than 35,000 experiments with 16-bit inputs and k-bit parameters to examine which zero-shot quantization methods improve scaling for 3 to 8-bit precision at scales of 19M to 176B parameters across the LLM families BLOOM, OPT, NeoX/Pythia, and GPT-2. We find that it is challenging to improve the bit-level scaling trade-off, with the only improvements being the use of a small block size -- splitting the parameters into small independently quantized blocks -- and the quantization data type being used (e.g., Int vs Float). Overall, our findings show that {4-bit} precision is almost universally optimal for total model bits and zero-shot accuracy."
                    },
                    {
                        "title": "Jointly Predicting Predicates and Arguments in Neural Semantic Role Labeling",
                        "abstract": "Recent BIO-tagging-based neural semantic role labeling models are very high performing, but assume gold predicates as part of the input and cannot incorporate span-level features. We propose an end-to-end approach for jointly predicting all predicates, arguments spans, and the relations between them. The model makes independent decisions about what relationship, if any, holds between every possible word-span pair, and learns contextualized span representations that provide rich, shared input features for each decision. Experiments demonstrate that this approach sets a new state of the art on PropBank SRL without gold predicates."
                    },
                    {
                        "title": "Semi-Autoregressive Training Improves Mask-Predict Decoding",
                        "abstract": "The recently proposed mask-predict decoding algorithm has narrowed the performance gap between semi-autoregressive machine translation models and the traditional left-to-right approach. We introduce a new training method for conditional masked language models, SMART, which mimics the semi-autoregressive behavior of mask-predict, producing training examples that contain model predictions as part of their inputs. Models trained with SMART produce higher-quality translations when using mask-predict decoding, effectively closing the remaining performance gap with fully autoregressive models."
                    },
                    {
                        "title": "Higher-order Coreference Resolution with Coarse-to-fine Inference",
                        "abstract": "We introduce a fully differentiable approximation to higher-order inference for coreference resolution. Our approach uses the antecedent distribution from a span-ranking architecture as an attention mechanism to iteratively refine span representations. This enables the model to softly consider multiple hops in the predicted clusters. To alleviate the computational cost of this iterative process, we introduce a coarse-to-fine approach that incorporates a less accurate but more efficient bilinear factor, enabling more aggressive pruning without hurting accuracy. Compared to the existing state-of-the-art span-ranking approach, our model significantly improves accuracy on the English OntoNotes benchmark, while being far more computationally efficient."
                    },
                    {
                        "title": "Extreme Extraction: Only One Hour per Relation",
                        "abstract": "Information Extraction (IE) aims to automatically generate a large knowledge base from natural language text, but progress remains slow. Supervised learning requires copious human annotation, while unsupervised and weakly supervised approaches do not deliver competitive accuracy. As a result, most fielded applications of IE, as well as the leading TAC-KBP systems, rely on significant amounts of manual engineering. Even \"Extreme\" methods, such as those reported in Freedman et al. 2011, require about 10 hours of expert labor per relation.   This paper shows how to reduce that effort by an order of magnitude. We present a novel system, InstaRead, that streamlines authoring with an ensemble of methods: 1) encoding extraction rules in an expressive and compositional representation, 2) guiding the user to promising rules based on corpus statistics and mined resources, and 3) introducing a new interactive development cycle that provides immediate feedback --- even on large datasets. Experiments show that experts can create quality extractors in under an hour and even NLP novices can author good extractors. These extractors equal or outperform ones obtained by comparably supervised and state-of-the-art distantly supervised approaches."
                    },
                    {
                        "title": "Global Neural CCG Parsing with Optimality Guarantees",
                        "abstract": "We introduce the first global recursive neural parsing model with optimality guarantees during decoding. To support global features, we give up dynamic programs and instead search directly in the space of all possible subtrees. Although this space is exponentially large in the sentence length, we show it is possible to learn an efficient A* parser. We augment existing parsing models, which have informative bounds on the outside score, with a global model that has loose bounds but only needs to model non-local phenomena. The global model is trained with a new objective that encourages the parser to explore a tiny fraction of the search space. The approach is applied to CCG parsing, improving state-of-the-art accuracy by 0.4 F1. The parser finds the optimal parse for 99.9% of held-out sentences, exploring on average only 190 subtrees."
                    },
                    {
                        "title": "End-to-end Neural Coreference Resolution",
                        "abstract": "We introduce the first end-to-end coreference resolution model and show that it significantly outperforms all previous work without using a syntactic parser or hand-engineered mention detector. The key idea is to directly consider all spans in a document as potential mentions and learn distributions over possible antecedents for each. The model computes span embeddings that combine context-dependent boundary representations with a head-finding attention mechanism. It is trained to maximize the marginal likelihood of gold antecedent spans from coreference clusters and is factored to enable aggressive pruning of potential mentions. Experiments demonstrate state-of-the-art performance, with a gain of 1.5 F1 on the OntoNotes benchmark and by 3.1 F1 using a 5-model ensemble, despite the fact that this is the first approach to be successfully trained with no external resources."
                    },
                    {
                        "title": "Long Short-Term Memory as a Dynamically Computed Element-wise Weighted Sum",
                        "abstract": "LSTMs were introduced to combat vanishing gradients in simple RNNs by augmenting them with gated additive recurrent connections. We present an alternative view to explain the success of LSTMs: the gates themselves are versatile recurrent models that provide more representational power than previously appreciated. We do this by decoupling the LSTM's gates from the embedded simple RNN, producing a new class of RNNs where the recurrence computes an element-wise weighted sum of context-independent functions of the input. Ablations on a range of problems demonstrate that the gating mechanism alone performs as well as an LSTM in most settings, strongly suggesting that the gates are doing much more in practice than just alleviating vanishing gradients."
                    },
                    {
                        "title": "Deep RNNs Encode Soft Hierarchical Syntax",
                        "abstract": "We present a set of experiments to demonstrate that deep recurrent neural networks (RNNs) learn internal representations that capture soft hierarchical notions of syntax from highly varied supervision. We consider four syntax tasks at different depths of the parse tree; for each word, we predict its part of speech as well as the first (parent), second (grandparent) and third level (great-grandparent) constituent labels that appear above it. These predictions are made from representations produced at different depths in networks that are pretrained with one of four objectives: dependency parsing, semantic role labeling, machine translation, or language modeling. In every case, we find a correspondence between network depth and syntactic depth, suggesting that a soft syntactic hierarchy emerges. This effect is robust across all conditions, indicating that the models encode significant amounts of syntax even in the absence of an explicit syntactic training supervision."
                    },
                    {
                        "title": "Large-Scale QA-SRL Parsing",
                        "abstract": "We present a new large-scale corpus of Question-Answer driven Semantic Role Labeling (QA-SRL) annotations, and the first high-quality QA-SRL parser. Our corpus, QA-SRL Bank 2.0, consists of over 250,000 question-answer pairs for over 64,000 sentences across 3 domains and was gathered with a new crowd-sourcing scheme that we show has high precision and good recall at modest cost. We also present neural models for two QA-SRL subtasks: detecting argument spans for a predicate and generating questions to label the semantic relationship. The best models achieve question accuracy of 82.6% and span-level accuracy of 77.6% (under human evaluation) on the full pipelined QA-SRL prediction task. They can also, as we show, be used to gather additional annotations at low cost."
                    },
                    {
                        "title": "Mapping Language to Code in Programmatic Context",
                        "abstract": "Source code is rarely written in isolation. It depends significantly on the programmatic context, such as the class that the code would reside in. To study this phenomenon, we introduce the task of generating class member functions given English documentation and the programmatic context provided by the rest of the class. This task is challenging because the desired code can vary greatly depending on the functionality the class provides (e.g., a sort function may or may not be available when we are asked to \"return the smallest element\" in a particular member variable list). We introduce CONCODE, a new large dataset with over 100,000 examples consisting of Java classes from online code repositories, and develop a new encoder-decoder architecture that models the interaction between the method documentation and the class environment. We also present a detailed error analysis suggesting that there is significant room for future work on this task."
                    },
                    {
                        "title": "The Referential Reader: A Recurrent Entity Network for Anaphora Resolution",
                        "abstract": "We present a new architecture for storing and accessing entity mentions during online text processing. While reading the text, entity references are identified, and may be stored by either updating or overwriting a cell in a fixed-length memory. The update operation implies coreference with the other mentions that are stored in the same cell; the overwrite operation causes these mentions to be forgotten. By encoding the memory operations as differentiable gates, it is possible to train the model end-to-end, using both a supervised anaphora resolution objective as well as a supplementary language modeling objective. Evaluation on a dataset of pronoun-name anaphora demonstrates strong performance with purely incremental text processing."
                    },
                    {
                        "title": "Learning Programmatic Idioms for Scalable Semantic Parsing",
                        "abstract": "Programmers typically organize executable source code using high-level coding patterns or idiomatic structures such as nested loops, exception handlers and recursive blocks, rather than as individual code tokens. In contrast, state of the art (SOTA) semantic parsers still map natural language instructions to source code by building the code syntax tree one node at a time. In this paper, we introduce an iterative method to extract code idioms from large source code corpora by repeatedly collapsing most-frequent depth-2 subtrees of their syntax trees, and train semantic parsers to apply these idioms during decoding. Applying idiom-based decoding on a recent context-dependent semantic parsing task improves the SOTA by 2.2\\% BLEU score while reducing training time by more than 50\\%. This improved speed enables us to scale up the model by training on an extended training set that is 5$\\times$ larger, to further move up the SOTA by an additional 2.3\\% BLEU and 0.9\\% exact match. Finally, idioms also significantly improve accuracy of semantic parsing to SQL on the ATIS-SQL dataset, when training data is limited."
                    }
                ]
            },
            "0acd2fee-b208-4cdf-adb6-5d598bd5241c": {
                "pk": "0acd2fee-b208-4cdf-adb6-5d598bd5241c",
                "name": "Kai Li",
                "collaborators": [
                    "Kia Kai Li",
                    "Wenjin Li",
                    "Dan Crisan",
                    "Lili Yan",
                    "Haixiao Deng",
                    "Yi Luo"
                ],
                "domain": [
                    "Research Instruments",
                    "Autonomous Driving",
                    "Stochastic Filtering",
                    "Audio Restoration"
                ],
                "publications": [
                    {
                        "title": "The (re-)instrumentalization of the Diagnostic and Statistical Manual of Mental Disorders (DSM) in psychological publications: a citation context analysis",
                        "abstract": "Research instruments play significant roles in the construction of scientific knowledge, even though we have only acquired very limited knowledge about their lifecycles from quantitative studies. This paper aims to address this gap by quantitatively examining the citation contexts of an exemplary research instrument, the Diagnostic and Statistical Manual of Mental Disorders (DSM), in full-text psychological publications. We investigated the relationship between the citation contexts of the DSM and its status as a valid instrument being used and described by psychological researchers. We specifically focused on how this relationship has changed over the DSM's citation histories, especially through the temporal framework of its versions. We found that a new version of the DSM is increasingly regarded as a valid instrument after its publication; this is reflected in various key citation contexts, such as the use of hedges, attention markers, and the verb profile in sentences where the DSM is cited. We call this process the re-instrumentalization of the DSM in the space of scientific publications. Our findings bridge an important gap between quantitative and qualitative science studies and shed light on an aspect of the social process of scientific instrument development that is not addressed by the current qualitative literature."
                    },
                    {
                        "title": "Deep Reinforcement Learning for Adverse Garage Scenario Generation",
                        "abstract": "Autonomous vehicles need to travel over 11 billion miles to ensure their safety. Therefore, the importance of simulation testing before real-world testing is self-evident. In recent years, the release of 3D simulators for autonomous driving, represented by Carla and CarSim, marks the transition of autonomous driving simulation testing environments from simple 2D overhead views to complex 3D models. During simulation testing, experimenters need to build static scenes and dynamic traffic flows, pedestrian flows, and other experimental elements to construct experimental scenarios. When building static scenes in 3D simulators, experimenters often need to manually construct 3D models, set parameters and attributes, which is time-consuming and labor-intensive. This thesis proposes an automated program generation framework. Based on deep reinforcement learning, this framework can generate different 2D ground script codes, on which 3D model files and map model files are built. The generated 3D ground scenes are displayed in the Carla simulator, where experimenters can use this scene for navigation algorithm simulation testing."
                    },
                    {
                        "title": "$\u03b7$-pairing in correlated fermion models with spin-orbit coupling",
                        "abstract": "We generalize the $\\eta$-pairing theory in Hubbard models to the ones with spin-orbit coupling (SOC) and obtain the conditions under which the $\\eta$-pairing operator is an eigenoperator of the Hamiltonian. The $\\eta$ pairing thus reveals an exact $SU(2)$ pseudospin symmetry in our spin-orbit coupled Hubbard model, even though the $SU(2)$ spin symmetry is explicitly broken by the SOC. In particular, these exact results can be applied to a variety of Hubbard models with SOC on either bipartite or non-bipartite lattices, whose noninteracting limit can be a Dirac semimetal, a Weyl semimetal, a nodal-line semimetal, and a Chern insulator. The $\\eta$ pairing conditions also impose constraints on the band topology of these systems. We then construct and focus on an interacting Dirac-semimetal model, which exhibits an exact pseudospin symmetry with fine-tuned parameters. The stability regions for the exact $\\eta$-pairing ground states (with momentum $\\bm{\\pi}$ or $\\bm{0}$) and the exact charge-density-wave ground states are established. Between these distinct symmetry-breaking phases, there exists an exactly solvable multicritical line. In the end, we discuss possible experimental realizations of our results."
                    },
                    {
                        "title": "Exploring k-Colorability",
                        "abstract": "An introductory paper to the graph k-colorability problem."
                    },
                    {
                        "title": "PCG-based Static Underground Garage Scenario Generation",
                        "abstract": "Autonomous driving technology has five levels, from L0 to L5. Currently, only the L2 level (partial automation) can be achieved, and there is a long way to go before reaching the final level of L5 (full automation). The key to crossing these levels lies in training the autonomous driving model. However, relying solely on real-world road data to train the model is far from enough and consumes a great deal of resources. Although there are already examples of training autonomous driving models through simulators that simulate real-world scenarios, these scenarios require complete manual construction. Directly converting 3D scenes from road network formats will lack a large amount of detail and cannot be used as training sets. Underground parking garage static scenario simulation is regarded as a procedural content generation (PCG) problem. This paper will use the Sarsa algorithm to solve procedural content generation on underground garage structures."
                    },
                    {
                        "title": "Generalised Particle Filters with Gaussian Mixtures",
                        "abstract": "Stochastic filtering is defined as the estimation of a partially observed dynamical system. A massive scientific and computational effort is dedicated to the development of numerical methods for approximating the solution of the filtering problem. Approximating the solution of the filtering problem with Gaussian mixtures has been a very popular method since the 1970s (see [1],[2],[46],[49]). Despite nearly fifty years of development, the existing work is based on the success of the numerical implementation and is not theoretically justified. This paper fills this gap and contains a rigorous analysis of a new Gaussian mixture approximation to the solution of the filtering problem. We deduce the L^2-convergence rate for the approximating system and show some numerical example to test the new algorithm."
                    },
                    {
                        "title": "Interactive Summarizing -- Automatic Slide Localization Technology as Generative Learning Tool",
                        "abstract": "Making a summary is a common learning strategy in lecture learning. It is an effective way for learners to engage in both traditional and video lectures. Video summarization is an effective technology applied to enhance learners' summarizing experience in a video lecture. In this article, we propose to apply cutting-edge automatic slide localization technology to lecture video learning experience. An interactive summarizing model is designed to explain how learners are engaged in the video lecture learning process supported by convolutional neural network and the possibility of related learning analytics."
                    },
                    {
                        "title": "Systematical design and three-dimensional simulation of X-ray FEL oscillator for Shanghai Coherent Light Facility",
                        "abstract": "Shanghai Coherent Light Facility (SCLF) is a quasi-CW hard X-ray free electron laser user facility which is recently proposed. Due to the high repetition rate, high quality electron beams, it is straightforward to consider an X-ray free electron laser oscillator (XFELO) operation for SCLF. The main processes for XFELO design, and parameters optimization of the undulator, X-ray cavity and electron beam are described. The first three-dimensional X-ray crystal Bragg diffraction code, named BRIGHT is built, which collaborates closely with GENESIS and OPC for numerical simulations of XFELO. The XFELO performances of SCLF is investigated and optimized by theoretical analysis and numerical simulation."
                    },
                    {
                        "title": "Apollo: Band-sequence Modeling for High-Quality Audio Restoration",
                        "abstract": "Audio restoration has become increasingly significant in modern society, not only due to the demand for high-quality auditory experiences enabled by advanced playback devices, but also because the growing capabilities of generative audio models necessitate high-fidelity audio. Typically, audio restoration is defined as a task of predicting undistorted audio from damaged input, often trained using a GAN framework to balance perception and distortion. Since audio degradation is primarily concentrated in mid- and high-frequency ranges, especially due to codecs, a key challenge lies in designing a generator capable of preserving low-frequency information while accurately reconstructing high-quality mid- and high-frequency content. Inspired by recent advancements in high-sample-rate music separation, speech enhancement, and audio codec models, we propose Apollo, a generative model designed for high-sample-rate audio restoration. Apollo employs an explicit frequency band split module to model the relationships between different frequency bands, allowing for more coherent and higher-quality restored audio. Evaluated on the MUSDB18-HQ and MoisesDB datasets, Apollo consistently outperforms existing SR-GAN models across various bit rates and music genres, particularly excelling in complex scenarios involving mixtures of multiple instruments and vocals. Apollo significantly improves music restoration quality while maintaining computational efficiency. The source code for Apollo is publicly available at https://github.com/JusperLee/Apollo."
                    }
                ]
            },
            "bac051d6-086a-48f2-a6a3-aab12385b471": {
                "pk": "bac051d6-086a-48f2-a6a3-aab12385b471",
                "name": "Peter Henderson",
                "collaborators": [
                    "Joelle Pineau",
                    "David Meger",
                    "Riashat Islam",
                    "Joshua Romoff",
                    "Emma Brunskill",
                    "Mark Lemley",
                    "Matthew Vertescher",
                    "Doina Precup",
                    "Muthucumaru Maheswaran",
                    "Thang Doan"
                ],
                "domain": [
                    "Embedded Systems",
                    "Reinforcement Learning",
                    "Machine Learning",
                    "Legal AI"
                ],
                "publications": [
                    {
                        "title": "Chaotic Memory Randomization for Securing Embedded Systems",
                        "abstract": "Embedded systems permeate through nearly all aspects of modern society. From cars to refrigerators to nuclear refineries, securing these systems has never been more important. Intrusions, such as the Stuxnet malware which broke the centrifuges in Iran's Natanz refinery, can be catastrophic to not only the infected systems, but even to the wellbeing of the surrounding population. Modern day protection mechanisms for these embedded systems generally look only at protecting the network layer, and those that try to discover malware already existing on a system typically aren't efficient enough to run on a standalone embedded system. As such, we present a novel way to ensure that no malware has been inserted into an embedded system. We chaotically randomize the entire memory space of the application, interspersing watchdog-monitor programs throughout, to monitor that the core application hasn't been infiltrated. By validating the original program through conventional methods and creating a clean reset, we can ensure that any inserted malware is purged from the system with minimal effect on the given system. We also present a software prototype to validate the possibility of this approach, but given the limitations and vulnerabilities of the prototype, we also suggest a hardware alternative to the system."
                    },
                    {
                        "title": "Distilling Information from a Flood: A Possibility for the Use of Meta-Analysis and Systematic Review in Machine Learning Research",
                        "abstract": "The current flood of information in all areas of machine learning research, from computer vision to reinforcement learning, has made it difficult to make aggregate scientific inferences. It can be challenging to distill a myriad of similar papers into a set of useful principles, to determine which new methodologies to use for a particular application, and to be confident that one has compared against all relevant related work when developing new ideas. However, such a rapidly growing body of research literature is a problem that other fields have already faced - in particular, medicine and epidemiology. In those fields, systematic reviews and meta-analyses have been used exactly for dealing with these issues and it is not uncommon for entire journals to be dedicated to such analyses. Here, we suggest the field of machine learning might similarly benefit from meta-analysis and systematic review, and we encourage further discussion and development along this direction."
                    },
                    {
                        "title": "Bayesian Policy Gradients via Alpha Divergence Dropout Inference",
                        "abstract": "Policy gradient methods have had great success in solving continuous control tasks, yet the stochastic nature of such problems makes deterministic value estimation difficult. We propose an approach which instead estimates a distribution by fitting the value function with a Bayesian Neural Network. We optimize an $\\alpha$-divergence objective with Bayesian dropout approximation to learn and estimate this distribution. We show that using the Monte Carlo posterior mean of the Bayesian value function distribution, rather than a deterministic network, improves stability and performance of policy gradient methods in continuous control MuJoCo simulations."
                    },
                    {
                        "title": "Freedom of Speech and AI Output",
                        "abstract": "Is the output of generative AI entitled to First Amendment protection? We're inclined to say yes. Even though current AI programs are of course not people and do not themselves have constitutional rights, their speech may potentially be protected because of the rights of the programs' creators. But beyond that, and likely more significantly, AI programs' speech should be protected because of the rights of their users-both the users' rights to listen and their rights to speak. In this short Article, we sketch the outlines of this analysis."
                    },
                    {
                        "title": "An Analysis of Parallelized Motion Masking Using Dual-Mode Single Gaussian Models",
                        "abstract": "Motion detection in video is important for a number of applications and fields. In video surveillance, motion detection is an essential accompaniment to activity recognition for early warning systems. Robotics also has much to gain from motion detection and segmentation, particularly in high speed motion tracking for tactile systems. There are a myriad of techniques for detecting and masking motion in an image. Successful systems have used Gaussian Models to discern background from foreground in an image (motion from static imagery). However, particularly in the case of a moving camera or frame of reference, it is necessary to compensate for the motion of the camera when attempting to discern objects moving in the foreground. For example, it is possible to estimate motion of the camera through optical flow methods or temporal differencing and then compensate for this motion in a background subtraction model. We selection a method by Yi et al. using Dual-Mode Single Gaussian Models which does just this. We implement the technique in Intel's Thread Building Blocks (TBB) and NVIDIA's CUDA libraries. We then compare parallelization improvements with a theoretical analysis of speedups based on the characteristics of our selected model and attributes of both TBB and CUDA. We make our implementation available to the public."
                    },
                    {
                        "title": "Reproducibility of Benchmarked Deep Reinforcement Learning Tasks for Continuous Control",
                        "abstract": "Policy gradient methods in reinforcement learning have become increasingly prevalent for state-of-the-art performance in continuous control tasks. Novel methods typically benchmark against a few key algorithms such as deep deterministic policy gradients and trust region policy optimization. As such, it is important to present and use consistent baselines experiments. However, this can be difficult due to general variance in the algorithms, hyper-parameter tuning, and environment stochasticity. We investigate and discuss: the significance of hyper-parameters in policy gradients for continuous control, general variance in the algorithms, and reproducibility of reported results. We provide guidelines on reporting novel results as comparisons against baseline methods such that future researchers can make informed decisions when investigating novel methods."
                    },
                    {
                        "title": "Cost Adaptation for Robust Decentralized Swarm Behaviour",
                        "abstract": "Decentralized receding horizon control (D-RHC) provides a mechanism for coordination in multi-agent settings without a centralized command center. However, combining a set of different goals, costs, and constraints to form an efficient optimization objective for D-RHC can be difficult. To allay this problem, we use a meta-learning process -- cost adaptation -- which generates the optimization objective for D-RHC to solve based on a set of human-generated priors (cost and constraint functions) and an auxiliary heuristic. We use this adaptive D-RHC method for control of mesh-networked swarm agents. This formulation allows a wide range of tasks to be encoded and can account for network delays, heterogeneous capabilities, and increasingly large swarms through the adaptation mechanism. We leverage the Unity3D game engine to build a simulator capable of introducing artificial networking failures and delays in the swarm. Using the simulator we validate our method on an example coordinated exploration task. We demonstrate that cost adaptation allows for more efficient and safer task completion under varying environment conditions and increasingly large swarm sizes. We release our simulator and code to the community for future work."
                    },
                    {
                        "title": "Where Did My Optimum Go?: An Empirical Analysis of Gradient Descent Optimization in Policy Gradient Methods",
                        "abstract": "Recent analyses of certain gradient descent optimization methods have shown that performance can degrade in some settings - such as with stochasticity or implicit momentum. In deep reinforcement learning (Deep RL), such optimization methods are often used for training neural networks via the temporal difference error or policy gradient. As an agent improves over time, the optimization target changes and thus the loss landscape (and local optima) change. Due to the failure modes of those methods, the ideal choice of optimizer for Deep RL remains unclear. As such, we provide an empirical analysis of the effects that a wide range of gradient descent optimizers and their hyperparameters have on policy gradient methods, a subset of Deep RL algorithms, for benchmark continuous control tasks. We find that adaptive optimizers have a narrow window of effective learning rates, diverging in other cases, and that the effectiveness of momentum varies depending on the properties of the environment. Our analysis suggests that there is significant interplay between the dynamics of the environment and Deep RL algorithm properties which aren't necessarily accounted for by traditional adaptive gradient methods. We provide suggestions for optimal settings of current methods and further lines of research based on our findings."
                    },
                    {
                        "title": "Where's the Liability in Harmful AI Speech?",
                        "abstract": "Generative AI, in particular text-based \"foundation models\" (large models trained on a huge variety of information including the internet), can generate speech that could be problematic under a wide range of liability regimes. Machine learning practitioners regularly \"red team\" models to identify and mitigate such problematic speech: from \"hallucinations\" falsely accusing people of serious misconduct to recipes for constructing an atomic bomb. A key question is whether these red-teamed behaviors actually present any liability risk for model creators and deployers under U.S. law, incentivizing investments in safety mechanisms. We examine three liability regimes, tying them to common examples of red-teamed model behaviors: defamation, speech integral to criminal conduct, and wrongful death. We find that any Section 230 immunity analysis or downstream liability analysis is intimately wrapped up in the technical details of algorithm design. And there are many roadblocks to truly finding models (and their associated parties) liable for generated speech. We argue that AI should not be categorically immune from liability in these scenarios and that as courts grapple with the already fine-grained complexities of platform algorithms, the technical details of generative AI loom above with thornier questions. Courts and policymakers should think carefully about what technical design incentives they create as they evaluate these issues."
                    },
                    {
                        "title": "Entropy Regularization for Population Estimation",
                        "abstract": "Entropy regularization is known to improve exploration in sequential decision-making problems. We show that this same mechanism can also lead to nearly unbiased and lower-variance estimates of the mean reward in the optimize-and-estimate structured bandit setting. Mean reward estimation (i.e., population estimation) tasks have recently been shown to be essential for public policy settings where legal constraints often require precise estimates of population metrics. We show that leveraging entropy and KL divergence can yield a better trade-off between reward and estimator variance than existing baselines, all while remaining nearly unbiased. These properties of entropy regularization illustrate an exciting potential for bridging the optimal exploration and estimation literatures."
                    },
                    {
                        "title": "Reward Estimation for Variance Reduction in Deep Reinforcement Learning",
                        "abstract": "Reinforcement Learning (RL) agents require the specification of a reward signal for learning behaviours. However, introduction of corrupt or stochastic rewards can yield high variance in learning. Such corruption may be a direct result of goal misspecification, randomness in the reward signal, or correlation of the reward with external factors that are not known to the agent. Corruption or stochasticity of the reward signal can be especially problematic in robotics, where goal specification can be particularly difficult for complex tasks. While many variance reduction techniques have been studied to improve the robustness of the RL process, handling such stochastic or corrupted reward structures remains difficult. As an alternative for handling this scenario in model-free RL methods, we suggest using an estimator for both rewards and value functions. We demonstrate that this improves performance under corrupted stochastic rewards in both the tabular and non-linear function approximation settings for a variety of noise types and environments. The use of reward estimation is a robust and easy-to-implement improvement for handling corrupted reward signals in model-free RL."
                    },
                    {
                        "title": "Adversarial Gain",
                        "abstract": "Adversarial examples can be defined as inputs to a model which induce a mistake - where the model output is different than that of an oracle, perhaps in surprising or malicious ways. Original models of adversarial attacks are primarily studied in the context of classification and computer vision tasks. While several attacks have been proposed in natural language processing (NLP) settings, they often vary in defining the parameters of an attack and what a successful attack would look like. The goal of this work is to propose a unifying model of adversarial examples suitable for NLP tasks in both generative and classification settings. We define the notion of adversarial gain: based in control theory, it is a measure of the change in the output of a system relative to the perturbation of the input (caused by the so-called adversary) presented to the learner. This definition, as we show, can be used under different feature spaces and distance conditions to determine attack or defense effectiveness across different intuitive manifolds. This notion of adversarial gain not only provides a useful way for evaluating adversaries and defenses, but can act as a building block for future work in robustness under adversaries due to its rooted nature in stability and manifold theory."
                    },
                    {
                        "title": "Benchmark Environments for Multitask Learning in Continuous Domains",
                        "abstract": "As demand drives systems to generalize to various domains and problems, the study of multitask, transfer and lifelong learning has become an increasingly important pursuit. In discrete domains, performance on the Atari game suite has emerged as the de facto benchmark for assessing multitask learning. However, in continuous domains there is a lack of agreement on standard multitask evaluation environments which makes it difficult to compare different approaches fairly. In this work, we describe a benchmark set of tasks that we have developed in an extendable framework based on OpenAI Gym. We run a simple baseline using Trust Region Policy Optimization and release the framework publicly to be expanded and used for the systematic comparison of multitask, transfer, and lifelong learning in continuous domains."
                    },
                    {
                        "title": "Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning",
                        "abstract": "Accurate reporting of energy and carbon usage is essential for understanding the potential climate impacts of machine learning research. We introduce a framework that makes this easier by providing a simple interface for tracking realtime energy consumption and carbon emissions, as well as generating standardized online appendices. Utilizing this framework, we create a leaderboard for energy efficient reinforcement learning algorithms to incentivize responsible research in this area as an example for other areas of machine learning. Finally, based on case studies using our framework, we propose strategies for mitigation of carbon emissions and reduction of energy consumption. By making accounting easier, we hope to further the sustainable development of machine learning experiments and spur more research into energy efficient algorithms."
                    },
                    {
                        "title": "Deep Reinforcement Learning that Matters",
                        "abstract": "In recent years, significant progress has been made in solving challenging problems across various domains using deep reinforcement learning (RL). Reproducing existing work and accurately judging the improvements offered by novel methods is vital to sustaining this progress. Unfortunately, reproducing results for state-of-the-art deep RL methods is seldom straightforward. In particular, non-determinism in standard benchmark environments, combined with variance intrinsic to the methods, can make reported results tough to interpret. Without significance metrics and tighter standardization of experimental reporting, it is difficult to determine whether improvements over the prior state-of-the-art are meaningful. In this paper, we investigate challenges posed by reproducibility, proper experimental techniques, and reporting procedures. We illustrate the variability in reported metrics and results when comparing against common baselines and suggest guidelines to make future results in deep RL more reproducible. We aim to spur discussion about how to ensure continued progress in the field by minimizing wasted effort stemming from results that are non-reproducible and easily misinterpreted."
                    },
                    {
                        "title": "An Information-Theoretic Perspective on Credit Assignment in Reinforcement Learning",
                        "abstract": "How do we formalize the challenge of credit assignment in reinforcement learning? Common intuition would draw attention to reward sparsity as a key contributor to difficult credit assignment and traditional heuristics would look to temporal recency for the solution, calling upon the classic eligibility trace. We posit that it is not the sparsity of the reward itself that causes difficulty in credit assignment, but rather the \\emph{information sparsity}. We propose to use information theory to define this notion, which we then use to characterize when credit assignment is an obstacle to efficient learning. With this perspective, we outline several information-theoretic mechanisms for measuring credit under a fixed behavior policy, highlighting the potential of information theory as a key tool towards provably-efficient credit assignment."
                    },
                    {
                        "title": "Promises and pitfalls of artificial intelligence for legal applications",
                        "abstract": "Is AI set to redefine the legal profession? We argue that this claim is not supported by the current evidence. We dive into AI's increasingly prevalent roles in three types of legal tasks: information processing; tasks involving creativity, reasoning, or judgment; and predictions about the future. We find that the ease of evaluating legal applications varies greatly across legal tasks, based on the ease of identifying correct answers and the observability of information relevant to the task at hand. Tasks that would lead to the most significant changes to the legal profession are also the ones most prone to overoptimism about AI capabilities, as they are harder to evaluate. We make recommendations for better evaluation and deployment of AI in legal contexts."
                    }
                ]
            }
        }
    },
    "2403.13117": {
        "paper_data": {
            "title": "Optimal Flow Matching: Learning Straight Trajectories in Just One Step",
            "url": "http://arxiv.org/abs/2403.13117v2",
            "arxiv_id": "2403.13117",
            "authors": [
                "Nikita Kornilov",
                "Petr Mokrov",
                "Alexander Gasnikov",
                "Alexander Korotin"
            ],
            "abstract": "Over the several recent years, there has been a boom in development of Flow Matching (FM) methods for generative modeling. One intriguing property pursued by the community is the ability to learn flows with straight trajectories which realize the Optimal Transport (OT) displacements. Straightness is crucial for the fast integration (inference) of the learned flow's paths. Unfortunately, most existing flow straightening methods are based on non-trivial iterative FM procedures which accumulate the error during training or exploit heuristics based on minibatch OT. To address these issues, we develop and theoretically justify the novel Optimal Flow Matching approach which allows recovering the straight OT displacement for the quadratic transport in just one FM step. The main idea of our approach is the employment of vector field for FM which are parameterized by convex functions.",
            "introduction": "   1 Introduction  Recent success in generative modeling Liu et\u00a0al. (2023); Esser et\u00a0al. (2024); Cao et\u00a0al. (2024) is mostly driven by Flow Matching (FM) Lipman et\u00a0al. (2022) models. These models move a known distribution to a target one via ordinary differential equations (ODE) describing the mass movement. However, such processes usually have curved trajectories, resulting in time-consuming ODE integration for sampling. To overcome this issue, researches developed several improvements of the FM Liu (2022); Liu et\u00a0al. (2022); Pooladian et\u00a0al. (2023), which aim to recover more straight paths.   Rectified Flow (RF) method Liu (2022); Liu et\u00a0al. (2022) iteratively solves FM and gradually rectifies trajectories. Unfortunately, in each FM iteration, it accumulates the error, see (Liu et\u00a0al., 2022, \u00a72.2) and (Liu, 2022, \u00a76). This may spoil the performance of the method. The other popular branch of approaches to straighten trajectories is based on the connection between straight paths and Optimal Transport (OT) Villani (2021). The main goal of OT is to find the way to move one probability distribution to another with the minimal effort. Such OT maps are usually described by ODEs with straight trajectories. In OT Conditional Flow Matching (OT-CFM) Pooladian et\u00a0al. (2023); Tong et\u00a0al. (2023), the authors propose to apply FM on top of OT solution between batches from considered distributions. Unfortunately, such a heuristic does not guarantee straight paths because of minibatch OT biases, see, e.g., (Tong et\u00a0al., 2023, Figure 1, right) for the practical illustration.   Contributions. In this paper, we fix the above-mentioned problems of the straightening methods. We propose a novel Optimal Flow Matching (OFM) approach (\u00a7\u00a7\\S\u00a73) that after a single FM iteration obtains straight trajectories which can be simulated without ODE solving. It recovers OT flow for the quadratic transport cost function, i.e., it solves the Benamou\u2013Brenier problem. We demonstrate the potential of OFM in the series of experiments and benchmarks (\u00a7\u00a7\\S\u00a74).   The main idea of our OFM is to consider during FM only specific vector fields which yield straight paths by design. These vector fields are the gradients of convex functions, which in practice are parametrized by Input Convex Neural Networks Amos et\u00a0al. (2017). In OFM, one can optionally use minibatch OT or any other transport plan as the input, and this is completely theoretically justified.     2 Background and Related Works  In this section, we provide all necessary backgrounds for the theory. First, we recall static (\u00a7\u00a7\\S\u00a72.1) and dynamic (\u00a7\u00a7\\S\u00a72.2) formulations of Optimal Transport and solvers (\u00a7\u00a7\\S\u00a72.3) for them. Then, we recall Flow Matching (\u00a7\u00a7\\S\u00a72.4.1) and flow straightening approaches: OT-CFM (\u00a7\u00a7\\S\u00a72.4.2) and Rectified Flow (\u00a7\u00a7\\S\u00a72.4.3).   Notations. For vectors x,y\u2208\u211dD\ud835\udc65\ud835\udc66superscript\u211d\ud835\udc37x,y\\in\\mathbb{R}^{D}italic_x , italic_y \u2208 blackboard_R start_POSTSUPERSCRIPT italic_D end_POSTSUPERSCRIPT, we denote the inner product by \u27e8x,y\u27e9\ud835\udc65\ud835\udc66\\langle x,y\\rangle\u27e8 italic_x , italic_y \u27e9 and the corresponding \u21132subscript\u21132\\ell_{2}roman_\u2113 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT norm by \u2016x\u2016:=\u27e8x,x\u27e9assignnorm\ud835\udc65\ud835\udc65\ud835\udc65\\|x\\|:=\\sqrt{\\langle x,x\\rangle}\u2225 italic_x \u2225 := square-root start_ARG \u27e8 italic_x , italic_x \u27e9 end_ARG. We use \ud835\udcab2,a\u2062c\u2062(\u211dD)subscript\ud835\udcab2\ud835\udc4e\ud835\udc50superscript\u211d\ud835\udc37\\mathcal{P}_{2,ac}(\\mathbb{R}^{D})caligraphic_P start_POSTSUBSCRIPT 2 , italic_a italic_c end_POSTSUBSCRIPT ( blackboard_R start_POSTSUPERSCRIPT italic_D end_POSTSUPERSCRIPT ) to refer to the set of absolute continuous probability distributions with the finite second moment. For vector x\u2208\u211dD\ud835\udc65superscript\u211d\ud835\udc37x\\in\\mathbb{R}^{D}italic_x \u2208 blackboard_R start_POSTSUPERSCRIPT italic_D end_POSTSUPERSCRIPT and distribution p\u2208\ud835\udcab2,a\u2062c\u2062(\u211dD)\ud835\udc5dsubscript\ud835\udcab2\ud835\udc4e\ud835\udc50superscript\u211d\ud835\udc37p\\in\\mathcal{P}_{2,ac}(\\mathbb{R}^{D})italic_p \u2208 caligraphic_P start_POSTSUBSCRIPT 2 , italic_a italic_c end_POSTSUBSCRIPT ( blackboard_R start_POSTSUPERSCRIPT italic_D end_POSTSUPERSCRIPT ), notation x\u223cpsimilar-to\ud835\udc65\ud835\udc5dx\\sim",
            "references": [
                {
                    "title": "Scaling Rectified Flow Transformers for High-Resolution Image Synthesis",
                    "abstract": "Diffusion models create data from noise by inverting the forward paths of data towards noise and have emerged as a powerful generative modeling technique for high-dimensional, perceptual data such as images and videos. Rectified flow is a recent generative model formulation that connects data and noise in a straight line. Despite its better theoretical properties and conceptual simplicity, it is not yet decisively established as standard practice. In this work, we improve existing noise sampling techniques for training rectified flow models by biasing them towards perceptually relevant scales. Through a large-scale study, we demonstrate the superior performance of this approach compared to established diffusion formulations for high-resolution text-to-image synthesis. Additionally, we present a novel transformer-based architecture for text-to-image generation that uses separate weights for the two modalities and enables a bidirectional flow of information between image and text tokens, improving text comprehension, typography, and human preference ratings. We demonstrate that this architecture follows predictable scaling trends and correlates lower validation loss to improved text-to-image synthesis as measured by various metrics and human evaluations. Our largest models outperform state-of-the-art models, and we will make our experimental data, code, and model weights publicly available."
                },
                {
                    "title": "Estimating Barycenters of Distributions with Neural Optimal Transport",
                    "abstract": "Given a collection of probability measures, a practitioner sometimes needs to find an\"average\"distribution which adequately aggregates reference distributions. A theoretically appealing notion of such an average is the Wasserstein barycenter, which is the primal focus of our work. By building upon the dual formulation of Optimal Transport (OT), we propose a new scalable approach for solving the Wasserstein barycenter problem. Our methodology is based on the recent Neural OT solver: it has bi-level adversarial learning objective and works for general cost functions. These are key advantages of our method since the typical adversarial algorithms leveraging barycenter tasks utilize tri-level optimization and focus mostly on quadratic cost. We also establish theoretical error bounds for our proposed approach and showcase its applicability and effectiveness in illustrative scenarios and image data setups. Our source code is available at https://github.com/justkolesov/NOTBarycenters."
                },
                {
                    "title": "Light and Optimal Schr\u00f6dinger Bridge Matching",
                    "abstract": "Schr\\\"odinger Bridges (SB) have recently gained the attention of the ML community as a promising extension of classic diffusion models which is also interconnected to the Entropic Optimal Transport (EOT). Recent solvers for SB exploit the pervasive bridge matching procedures. Such procedures aim to recover a stochastic process transporting the mass between distributions given only a transport plan between them. In particular, given the EOT plan, these procedures can be adapted to solve SB. This fact is heavily exploited by recent works giving rise to matching-based SB solvers. The cornerstone here is recovering the EOT plan: recent works either use heuristical approximations (e.g., the minibatch OT) or establish iterative matching procedures which by the design accumulate the error during the training. We address these limitations and propose a novel procedure to learn SB which we call the \\textbf{optimal Schr\\\"odinger bridge matching}. It exploits the optimal parameterization of the diffusion process and provably recovers the SB process \\textbf{(a)} with a single bridge matching step and \\textbf{(b)} with arbitrary transport plan as the input. Furthermore, we show that the optimal bridge matching objective coincides with the recently discovered energy-based modeling (EBM) objectives to learn EOT/SB. Inspired by this observation, we develop a light solver (which we call LightSB-M) to implement optimal matching in practice using the Gaussian mixture parameterization of the adjusted Schr\\\"odinger potential. We experimentally showcase the performance of our solver in a range of practical tasks. The code for our solver can be found at https://github.com/SKholkin/LightSB-Matching."
                },
                {
                    "title": "Principled Weight Initialisation for Input-Convex Neural Networks",
                    "abstract": "Input-Convex Neural Networks (ICNNs) are networks that guarantee convexity in their input-output mapping. These networks have been successfully applied for energy-based modelling, optimal transport problems and learning invariances. The convexity of ICNNs is achieved by using non-decreasing convex activation functions and non-negative weights. Because of these peculiarities, previous initialisation strategies, which implicitly assume centred weights, are not effective for ICNNs. By studying signal propagation through layers with non-negative weights, we are able to derive a principled weight initialisation for ICNNs. Concretely, we generalise signal propagation theory by removing the assumption that weights are sampled from a centred distribution. In a set of experiments, we demonstrate that our principled initialisation effectively accelerates learning in ICNNs and leads to better generalisation. Moreover, we find that, in contrast to common belief, ICNNs can be trained without skip-connections when initialised correctly. Finally, we apply ICNNs to a real-world drug discovery task and show that they allow for more effective molecular latent space exploration."
                },
                {
                    "title": "Light Schr\u00f6dinger Bridge",
                    "abstract": "Despite the recent advances in the field of computational Schr\\\"odinger Bridges (SB), most existing SB solvers are still heavy-weighted and require complex optimization of several neural networks. It turns out that there is no principal solver which plays the role of simple-yet-effective baseline for SB just like, e.g., $k$-means method in clustering, logistic regression in classification or Sinkhorn algorithm in discrete optimal transport. We address this issue and propose a novel fast and simple SB solver. Our development is a smart combination of two ideas which recently appeared in the field: (a) parameterization of the Schr\\\"odinger potentials with sum-exp quadratic functions and (b) viewing the log-Schr\\\"odinger potentials as the energy functions. We show that combined together these ideas yield a lightweight, simulation-free and theoretically justified SB solver with a simple straightforward optimization objective. As a result, it allows solving SB in moderate dimensions in a matter of minutes on CPU without a painful hyperparameter selection. Our light solver resembles the Gaussian mixture model which is widely used for density estimation. Inspired by this similarity, we also prove an important theoretical result showing that our light solver is a universal approximator of SBs. Furthemore, we conduct the analysis of the generalization error of our light solver. The code for our solver can be found at https://github.com/ngushchin/LightSB"
                },
                {
                    "title": "InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generation",
                    "abstract": "Diffusion models have revolutionized text-to-image generation with its exceptional quality and creativity. However, its multi-step sampling process is known to be slow, often requiring tens of inference steps to obtain satisfactory results. Previous attempts to improve its sampling speed and reduce computational costs through distillation have been unsuccessful in achieving a functional one-step model. In this paper, we explore a recent method called Rectified Flow, which, thus far, has only been applied to small datasets. The core of Rectified Flow lies in its \\emph{reflow} procedure, which straightens the trajectories of probability flows, refines the coupling between noises and images, and facilitates the distillation process with student models. We propose a novel text-conditioned pipeline to turn Stable Diffusion (SD) into an ultra-fast one-step model, in which we find reflow plays a critical role in improving the assignment between noise and images. Leveraging our new pipeline, we create, to the best of our knowledge, the first one-step diffusion-based text-to-image generator with SD-level image quality, achieving an FID (Frechet Inception Distance) of $23.3$ on MS COCO 2017-5k, surpassing the previous state-of-the-art technique, progressive distillation, by a significant margin ($37.2$ $\\rightarrow$ $23.3$ in FID). By utilizing an expanded network with 1.7B parameters, we further improve the FID to $22.4$. We call our one-step models \\emph{InstaFlow}. On MS COCO 2014-30k, InstaFlow yields an FID of $13.1$ in just $0.09$ second, the best in $\\leq 0.1$ second regime, outperforming the recent StyleGAN-T ($13.9$ in $0.1$ second). Notably, the training of InstaFlow only costs 199 A100 GPU days. Codes and pre-trained models are available at \\url{github.com/gnobitab/InstaFlow}."
                },
                {
                    "title": "Generative Modeling through the Semi-dual Formulation of Unbalanced Optimal Transport",
                    "abstract": "Optimal Transport (OT) problem investigates a transport map that bridges two distributions while minimizing a given cost function. In this regard, OT between tractable prior distribution and data has been utilized for generative modeling tasks. However, OT-based methods are susceptible to outliers and face optimization challenges during training. In this paper, we propose a novel generative model based on the semi-dual formulation of Unbalanced Optimal Transport (UOT). Unlike OT, UOT relaxes the hard constraint on distribution matching. This approach provides better robustness against outliers, stability during training, and faster convergence. We validate these properties empirically through experiments. Moreover, we study the theoretical upper-bound of divergence between distributions in UOT. Our model outperforms existing OT-based generative models, achieving FID scores of 2.97 on CIFAR-10 and 6.36 on CelebA-HQ-256. The code is available at \\url{https://github.com/Jae-Moo/UOTM}."
                },
                {
                    "title": "Multisample Flow Matching: Straightening Flows with Minibatch Couplings",
                    "abstract": "Simulation-free methods for training continuous-time generative models construct probability paths that go between noise distributions and individual data samples. Recent works, such as Flow Matching, derived paths that are optimal for each data sample. However, these algorithms rely on independent data and noise samples, and do not exploit underlying structure in the data distribution for constructing probability paths. We propose Multisample Flow Matching, a more general framework that uses non-trivial couplings between data and noise samples while satisfying the correct marginal constraints. At very small overhead costs, this generalization allows us to (i) reduce gradient variance during training, (ii) obtain straighter flows for the learned vector field, which allows us to generate high-quality samples using fewer function evaluations, and (iii) obtain transport maps with lower cost in high dimensions, which has applications beyond generative modeling. Importantly, we do so in a completely simulation-free manner with a simple minimization objective. We show that our proposed methods improve sample consistency on downsampled ImageNet data sets, and lead to better low-cost sample generation."
                },
                {
                    "title": "Energy-guided Entropic Neural Optimal Transport",
                    "abstract": "Energy-based models (EBMs) are known in the Machine Learning community for decades. Since the seminal works devoted to EBMs dating back to the noughties, there have been a lot of efficient methods which solve the generative modelling problem by means of energy potentials (unnormalized likelihood functions). In contrast, the realm of Optimal Transport (OT) and, in particular, neural OT solvers is much less explored and limited by few recent works (excluding WGAN-based approaches which utilize OT as a loss function and do not model OT maps themselves). In our work, we bridge the gap between EBMs and Entropy-regularized OT. We present a novel methodology which allows utilizing the recent developments and technical improvements of the former in order to enrich the latter. From the theoretical perspective, we prove generalization bounds for our technique. In practice, we validate its applicability in toy 2D and image domains. To showcase the scalability, we empower our method with a pre-trained StyleGAN and apply it to high-res AFHQ $512\\times 512$ unpaired I2I translation. For simplicity, we choose simple short- and long-run EBMs as a backbone of our Energy-guided Entropic OT approach, leaving the application of more sophisticated EBMs for future research. Our code is available at: https://github.com/PetrMokrov/Energy-guided-Entropic-OT"
                },
                {
                    "title": "Diffusion Schr\u00f6dinger Bridge Matching",
                    "abstract": "Solving transport problems, i.e. finding a map transporting one given distribution to another, has numerous applications in machine learning. Novel mass transport methods motivated by generative modeling have recently been proposed, e.g. Denoising Diffusion Models (DDMs) and Flow Matching Models (FMMs) implement such a transport through a Stochastic Differential Equation (SDE) or an Ordinary Differential Equation (ODE). However, while it is desirable in many applications to approximate the deterministic dynamic Optimal Transport (OT) map which admits attractive properties, DDMs and FMMs are not guaranteed to provide transports close to the OT map. In contrast, Schr\\\"odinger bridges (SBs) compute stochastic dynamic mappings which recover entropy-regularized versions of OT. Unfortunately, existing numerical methods approximating SBs either scale poorly with dimension or accumulate errors across iterations. In this work, we introduce Iterative Markovian Fitting (IMF), a new methodology for solving SB problems, and Diffusion Schr\\\"odinger Bridge Matching (DSBM), a novel numerical algorithm for computing IMF iterates. DSBM significantly improves over previous SB numerics and recovers as special/limiting cases various recent transport methods. We demonstrate the performance of DSBM on a variety of problems."
                },
                {
                    "title": "The Monge Gap: A Regularizer to Learn All Transport Maps",
                    "abstract": "Optimal transport (OT) theory has been been used in machine learning to study and characterize maps that can push-forward efficiently a probability measure onto another. Recent works have drawn inspiration from Brenier's theorem, which states that when the ground cost is the squared-Euclidean distance, the ``best'' map to morph a continuous measure in $\\mathcal{P}(\\Rd)$ into another must be the gradient of a convex function. To exploit that result, [Makkuva+ 2020, Korotin+2020] consider maps $T=\\nabla f_\\theta$, where $f_\\theta$ is an input convex neural network (ICNN), as defined by Amos+2017, and fit $\\theta$ with SGD using samples. Despite their mathematical elegance, fitting OT maps with ICNNs raises many challenges, due notably to the many constraints imposed on $\\theta$; the need to approximate the conjugate of $f_\\theta$; or the limitation that they only work for the squared-Euclidean cost. More generally, we question the relevance of using Brenier's result, which only applies to densities, to constrain the architecture of candidate maps fitted on samples. Motivated by these limitations, we propose a radically different approach to estimating OT maps: Given a cost $c$ and a reference measure $\\rho$, we introduce a regularizer, the Monge gap $\\mathcal{M}^c_{\\rho}(T)$ of a map $T$. That gap quantifies how far a map $T$ deviates from the ideal properties we expect from a $c$-OT map. In practice, we drop all architecture requirements for $T$ and simply minimize a distance (e.g., the Sinkhorn divergence) between $T\\sharp\\mu$ and $\\nu$, regularized by $\\mathcal{M}^c_\\rho(T)$. We study $\\mathcal{M}^c_{\\rho}$, and show how our simple pipeline outperforms significantly other baselines in practice."
                },
                {
                    "title": "Entropic Neural Optimal Transport via Diffusion Processes",
                    "abstract": "We propose a novel neural algorithm for the fundamental problem of computing the entropic optimal transport (EOT) plan between continuous probability distributions which are accessible by samples. Our algorithm is based on the saddle point reformulation of the dynamic version of EOT which is known as the Schr\\\"odinger Bridge problem. In contrast to the prior methods for large-scale EOT, our algorithm is end-to-end and consists of a single learning step, has fast inference procedure, and allows handling small values of the entropy regularization coefficient which is of particular importance in some applied problems. Empirically, we show the performance of the method on several large-scale EOT tasks. https://github.com/ngushchin/EntropicNeuralOptimalTransport"
                },
                {
                    "title": "On amortizing convex conjugates for optimal transport",
                    "abstract": "This paper focuses on computing the convex conjugate operation that arises when solving Euclidean Wasserstein-2 optimal transport problems. This conjugation, which is also referred to as the Legendre-Fenchel conjugate or c-transform,is considered difficult to compute and in practice,Wasserstein-2 methods are limited by not being able to exactly conjugate the dual potentials in continuous space. To overcome this, the computation of the conjugate can be approximated with amortized optimization, which learns a model to predict the conjugate. I show that combining amortized approximations to the conjugate with a solver for fine-tuning significantly improves the quality of transport maps learned for the Wasserstein-2 benchmark by Korotin et al. (2021a) and is able to model many 2-dimensional couplings and flows considered in the literature. All of the baselines, methods, and solvers in this paper are available at http://github.com/facebookresearch/w2ot."
                },
                {
                    "title": "Numerical Methods for Large-Scale Optimal Transport",
                    "abstract": "The optimal transport (OT) problem is a classical optimization problem having the form of linear programming. Machine learning applications put forward new computational challenges in its solution. In particular, the OT problem defines a distance between real-world objects such as images, videos, texts, etc., modeled as probability distributions. In this case, the large dimension of the corresponding optimization problem does not allow applying classical methods such as network simplex or interior-point methods. This challenge was overcome by introducing entropic regularization and using the efficient Sinkhorn's algorithm to solve the regularized problem. A flexible alternative is the accelerated primal-dual gradient method, which can use any strongly-convex regularization. We discuss these algorithms and other related problems such as approximating the Wasserstein barycenter together with efficient algorithms for its solution, including decentralized distributed algorithms."
                },
                {
                    "title": "Flow Matching for Generative Modeling",
                    "abstract": "We introduce a new paradigm for generative modeling built on Continuous Normalizing Flows (CNFs), allowing us to train CNFs at unprecedented scale. Specifically, we present the notion of Flow Matching (FM), a simulation-free approach for training CNFs based on regressing vector fields of fixed conditional probability paths. Flow Matching is compatible with a general family of Gaussian probability paths for transforming between noise and data samples -- which subsumes existing diffusion paths as specific instances. Interestingly, we find that employing FM with diffusion paths results in a more robust and stable alternative for training diffusion models. Furthermore, Flow Matching opens the door to training CNFs with other, non-diffusion probability paths. An instance of particular interest is using Optimal Transport (OT) displacement interpolation to define the conditional probability paths. These paths are more efficient than diffusion paths, provide faster training and sampling, and result in better generalization. Training CNFs using Flow Matching on ImageNet leads to consistently better performance than alternative diffusion-based methods in terms of both likelihood and sample quality, and allows fast and reliable sample generation using off-the-shelf numerical ODE solvers."
                },
                {
                    "title": "Rectified Flow: A Marginal Preserving Approach to Optimal Transport",
                    "abstract": "We present a \ufb02ow-based approach to the optimal transport (OT) problem between two continuous distributions \u03c0 0 , \u03c0 1 on R d , of minimizing a transport cost E [ c ( X 1 \u2212 X 0 )] in the set of couplings ( X 0 , X 1 ) whose marginal distributions on X 0 , X 1 equals \u03c0 0 , \u03c0 1 , respectively, where c is a cost function. Our method iteratively constructs a sequence of neural ordinary differentiable equations (ODE), each learned by solving a simple unconstrained regression problem, which monotonically reduce the transport cost while automatically preserving the marginal constraints. This yields a monotonic interior approach that traverses inside the set of valid couplings to decrease the transport cost, which distinguishes itself from most existing approaches that enforce the coupling constraints from the outside. The main idea of the method draws from recti\ufb01ed \ufb02ow [LGL22], a recent approach that simultaneously decreases the whole family of transport costs induced by convex functions c (and is hence multi-objective in nature), but is not tailored to minimize a speci\ufb01c transport cost. Our method is a single-object variant of recti\ufb01ed \ufb02ow that guarantees to solve the OT problem for a \ufb01xed, user-speci\ufb01ed convex cost function c ."
                },
                {
                    "title": "Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow",
                    "abstract": "We present rectified flow, a surprisingly simple approach to learning (neural) ordinary differential equation (ODE) models to transport between two empirically observed distributions \\pi_0 and \\pi_1, hence providing a unified solution to generative modeling and domain transfer, among various other tasks involving distribution transport. The idea of rectified flow is to learn the ODE to follow the straight paths connecting the points drawn from \\pi_0 and \\pi_1 as much as possible. This is achieved by solving a straightforward nonlinear least squares optimization problem, which can be easily scaled to large models without introducing extra parameters beyond standard supervised learning. The straight paths are special and preferred because they are the shortest paths between two points, and can be simulated exactly without time discretization and hence yield computationally efficient models. We show that the procedure of learning a rectified flow from data, called rectification, turns an arbitrary coupling of \\pi_0 and \\pi_1 to a new deterministic coupling with provably non-increasing convex transport costs. In addition, recursively applying rectification allows us to obtain a sequence of flows with increasingly straight paths, which can be simulated accurately with coarse time discretization in the inference phase. In empirical studies, we show that rectified flow performs superbly on image generation, image-to-image translation, and domain adaptation. In particular, on image generation and translation, our method yields nearly straight flows that give high quality results even with a single Euler discretization step."
                },
                {
                    "title": "A Survey on Generative Diffusion Models",
                    "abstract": "Deep generative models have unlocked another profound realm of human creativity. By capturing and generalizing patterns within data, we have entered the epoch of all-encompassing Artificial Intelligence for General Creativity (AIGC). Notably, diffusion models, recognized as one of the paramount generative models, materialize human ideation into tangible instances across diverse domains, encompassing imagery, text, speech, biology, and healthcare. To provide advanced and comprehensive insights into diffusion, this survey comprehensively elucidates its developmental trajectory and future directions from three distinct angles: the fundamental formulation of diffusion, algorithmic enhancements, and the manifold applications of diffusion. Each layer is meticulously explored to offer a profound comprehension of its evolution. Structured and summarized approaches are presented here."
                },
                {
                    "title": "Supervised Training of Conditional Monge Maps",
                    "abstract": "Optimal transport (OT) theory describes general principles to define and select, among many possible choices, the most efficient way to map a probability measure onto another. That theory has been mostly used to estimate, given a pair of source and target probability measures $(\\mu, \\nu)$, a parameterized map $T_\\theta$ that can efficiently map $\\mu$ onto $\\nu$. In many applications, such as predicting cell responses to treatments, pairs of input/output data measures $(\\mu, \\nu)$ that define optimal transport problems do not arise in isolation but are associated with a context $c$, as for instance a treatment when comparing populations of untreated and treated cells. To account for that context in OT estimation, we introduce CondOT, a multi-task approach to estimate a family of OT maps conditioned on a context variable, using several pairs of measures $\\left(\\mu_i, \\nu_i\\right)$ tagged with a context label $c_i$. CondOT learns a global map $\\mathcal{T}_\\theta$ conditioned on context that is not only expected to fit all labeled pairs in the dataset $\\left\\{\\left(c_i,\\left(\\mu_i, \\nu_i\\right)\\right)\\right\\}$, i.e., $\\mathcal{T}_\\theta\\left(c_i\\right) \\sharp \\mu_i \\approx \\nu_i$, but should also generalize to produce meaningful maps $\\mathcal{T}_\\theta\\left(c_{\\text {new }}\\right)$ when conditioned on unseen contexts $c_{\\text {new }}$. Our approach harnesses and provides a novel usage for partially input convex neural networks, for which we introduce a robust and efficient initialization strategy inspired by Gaussian approximations. We demonstrate the ability of CondOT to infer the effect of an arbitrary combination of genetic or therapeutic perturbations on single cells, using only observations of the effects of said perturbations separately."
                },
                {
                    "title": "Neural Optimal Transport with General Cost Functionals",
                    "abstract": "We introduce a novel neural network-based algorithm to compute optimal transport (OT) plans for general cost functionals. In contrast to common Euclidean costs, i.e., $\\ell^1$ or $\\ell^2$, such functionals provide more flexibility and allow using auxiliary information, such as class labels, to construct the required transport map. Existing methods for general costs are discrete and have limitations in practice, i.e. they do not provide an out-of-sample estimation. We address the challenge of designing a continuous OT approach for general costs that generalizes to new data points in high-dimensional spaces, such as images. Additionally, we provide the theoretical error analysis for our recovered transport plans. As an application, we construct a cost functional to map data distributions while preserving the class-wise structure."
                },
                {
                    "title": "Kernel Neural Optimal Transport",
                    "abstract": "We study the Neural Optimal Transport (NOT) algorithm which uses the general optimal transport formulation and learns stochastic transport plans. We show that NOT with the weak quadratic cost might learn fake plans which are not optimal. To resolve this issue, we introduce kernel weak quadratic costs. We show that they provide improved theoretical guarantees and practical performance. We test NOT with kernel costs on the unpaired image-to-image translation task."
                },
                {
                    "title": "Neural Optimal Transport",
                    "abstract": "We present a novel neural-networks-based algorithm to compute optimal transport maps and plans for strong and weak transport costs. To justify the usage of neural networks, we prove that they are universal approximators of transport plans between probability distributions. We evaluate the performance of our optimal transport algorithm on toy examples and on the unpaired image-to-image translation."
                },
                {
                    "title": "Wasserstein Iterative Networks for Barycenter Estimation",
                    "abstract": "Wasserstein barycenters have become popular due to their ability to represent the average of probability measures in a geometrically meaningful way. In this paper, we present an algorithm to approximate the Wasserstein-2 barycenters of continuous measures via a generative model. Previous approaches rely on regularization (entropic/quadratic) which introduces bias or on input convex neural networks which are not expressive enough for large-scale tasks. In contrast, our algorithm does not introduce bias and allows using arbitrary neural networks. In addition, based on the celebrity faces dataset, we construct Ave, celeba! dataset which can be used for quantitative evaluation of barycenter algorithms by using standard metrics of generative models such as FID."
                },
                {
                    "title": "Input Convex Gradient Networks",
                    "abstract": "The gradients of convex functions are expressive models of non-trivial vector fields. For example, Brenier's theorem yields that the optimal transport map between any two measures on Euclidean space under the squared distance is realized as a convex gradient, which is a key insight used in recent generative flow models. In this paper, we study how to model convex gradients by integrating a Jacobian-vector product parameterized by a neural network, which we call the Input Convex Gradient Network (ICGN). We theoretically study ICGNs and compare them to taking the gradient of an Input-Convex Neural Network (ICNN), empirically demonstrating that a single layer ICGN can fit a toy example better than a single layer ICNN. Lastly, we explore extensions to deeper networks and connections to constructions from Riemannian geometry."
                },
                {
                    "title": "Score-based Generative Neural Networks for Large-Scale Optimal Transport",
                    "abstract": "We consider the fundamental problem of sampling the optimal transport coupling between given source and target distributions. In certain cases, the optimal transport plan takes the form of a one-to-one mapping from the source support to the target support, but learning or even approximating such a map is computationally challenging for large and high-dimensional datasets due to the high cost of linear programming routines and an intrinsic curse of dimensionality. We study instead the Sinkhorn problem, a regularized form of optimal transport whose solutions are couplings between the source and the target distribution. We introduce a novel framework for learning the Sinkhorn coupling between two distributions in the form of a score-based generative model. Conditioned on source data, our procedure iterates Langevin Dynamics to sample target data according to the regularized optimal coupling. Key to this approach is a neural network parametrization of the Sinkhorn problem, and we prove convergence of gradient descent with respect to network parameters in this formulation. We demonstrate its empirical success on a variety of large scale optimal transport tasks."
                },
                {
                    "title": "Generative Modeling with Optimal Transport Maps",
                    "abstract": "With the discovery of Wasserstein GANs, Optimal Transport (OT) has become a powerful tool for large-scale generative modeling tasks. In these tasks, OT cost is typically used as the loss for training GANs. In contrast to this approach, we show that the OT map itself can be used as a generative model, providing comparable performance. Previous analogous approaches consider OT maps as generative models only in the latent spaces due to their poor performance in the original high-dimensional ambient space. In contrast, we apply OT maps directly in the ambient space, e.g., a space of high-dimensional images. First, we derive a min-max optimization algorithm to efficiently compute OT maps for the quadratic cost (Wasserstein-2 distance). Next, we extend the approach to the case when the input and output distributions are located in the spaces of different dimensions and derive error bounds for the computed OT map. We evaluate the algorithm on image generation and unpaired image restoration tasks. In particular, we consider denoising, colorization, and inpainting, where the optimality of the restoration map is a desired attribute, since the output (restored) image is expected to be close to the input (degraded) one."
                },
                {
                    "title": "Neural Monge Map estimation and its applications",
                    "abstract": "Monge map refers to the optimal transport map between two probability distributions and provides a principled approach to transform one distribution to another. Neural network based optimal transport map solver has gained great attention in recent years. Along this line, we present a scalable algorithm for computing the neural Monge map between two probability distributions. Our algorithm is based on a weak form of the optimal transport problem, thus it only requires samples from the marginals instead of their analytic expressions, and can accommodate optimal transport between two distributions with different dimensions. Our algorithm is suitable for general cost functions, compared with other existing methods for estimating Monge maps using samples, which are usually for quadratic costs. The performance of our algorithms is demonstrated through a series of experiments with both synthetic and realistic data, including text-to-image generation and image inpainting tasks."
                },
                {
                    "title": "Solving Schr\u00f6dinger Bridges via Maximum Likelihood",
                    "abstract": "The Schr\u00f6dinger bridge problem (SBP) finds the most likely stochastic evolution between two probability distributions given a prior stochastic evolution. As well as applications in the natural sciences, problems of this kind have important applications in machine learning such as dataset alignment and hypothesis testing. Whilst the theory behind this problem is relatively mature, scalable numerical recipes to estimate the Schr\u00f6dinger bridge remain an active area of research. Our main contribution is the proof of equivalence between solving the SBP and an autoregressive maximum likelihood estimation objective. This formulation circumvents many of the challenges of density estimation and enables direct application of successful machine learning techniques. We propose a numerical procedure to estimate SBPs using Gaussian process and demonstrate the practical usage of our approach in numerical simulations and experiments."
                },
                {
                    "title": "Do Neural Optimal Transport Solvers Work? A Continuous Wasserstein-2 Benchmark",
                    "abstract": "Despite the recent popularity of neural network-based solvers for optimal transport (OT), there is no standard quantitative way to evaluate their performance. In this paper, we address this issue for quadratic-cost transport -- specifically, computation of the Wasserstein-2 distance, a commonly-used formulation of optimal transport in machine learning. To overcome the challenge of computing ground truth transport maps between continuous measures needed to assess these solvers, we use input-convex neural networks (ICNN) to construct pairs of measures whose ground truth OT maps can be obtained analytically. This strategy yields pairs of continuous benchmark measures in high-dimensional spaces such as spaces of images. We thoroughly evaluate existing optimal transport solvers using these benchmark measures. Even though these solvers perform well in downstream tasks, many do not faithfully recover optimal transport maps. To investigate the cause of this discrepancy, we further test the solvers in a setting of image generation. Our study reveals crucial limitations of existing solvers and shows that increased OT accuracy does not necessarily correlate to better results downstream."
                },
                {
                    "title": "Unbalanced Feature Transport for Exemplar-based Image Translation",
                    "abstract": "Despite the great success of GANs in images translation with different conditioned inputs such as semantic segmentation and edge maps, generating high-fidelity realistic images with reference styles remains a grand challenge in conditional image-to-image translation. This paper presents a general image translation framework that incorporates optimal transport for feature alignment between conditional inputs and style exemplars in image translation. The introduction of optimal transport mitigates the constraint of many-to-one feature matching significantly while building up accurate semantic correspondences between conditional inputs and exemplars. We design a novel unbalanced optimal transport to address the transport between features with deviational distributions which exists widely between conditional inputs and exemplars. In addition, we design a semantic-activation normalization scheme that injects style features of exemplars into the image translation process successfully. Extensive experiments over multiple image translation tasks show that our method achieves superior image translation qualitatively and quantitatively as compared with the state-of-the-art."
                },
                {
                    "title": "Diffusion Schr\u00f6dinger Bridge with Applications to Score-Based Generative Modeling",
                    "abstract": "Progressively applying Gaussian noise transforms complex data distributions to approximately Gaussian. Reversing this dynamic defines a generative model. When the forward noising process is given by a Stochastic Differential Equation (SDE), Song et al. (2021) demonstrate how the time inhomogeneous drift of the associated reverse-time SDE may be estimated using score-matching. A limitation of this approach is that the forward-time SDE must be run for a sufficiently long time for the final distribution to be approximately Gaussian. In contrast, solving the Schr\\\"odinger Bridge problem (SB), i.e. an entropy-regularized optimal transport problem on path spaces, yields diffusions which generate samples from the data distribution in finite time. We present Diffusion SB (DSB), an original approximation of the Iterative Proportional Fitting (IPF) procedure to solve the SB problem, and provide theoretical analysis along with generative modeling experiments. The first DSB iteration recovers the methodology proposed by Song et al. (2021), with the flexibility of using shorter time intervals, as subsequent DSB iterations reduce the discrepancy between the final-time marginal of the forward (resp. backward) SDE with respect to the prior (resp. data) distribution. Beyond generative modeling, DSB offers a widely applicable computational optimal transport tool as the continuous state-space analogue of the popular Sinkhorn algorithm (Cuturi, 2013)."
                },
                {
                    "title": "Wasserstein GANs Work Because They Fail (to Approximate the Wasserstein Distance)",
                    "abstract": "Wasserstein GANs are based on the idea of minimising the Wasserstein distance between a real and a generated distribution. We provide an in-depth mathematical analysis of differences between the theoretical setup and the reality of training Wasserstein GANs. In this work, we gather both theoretical and empirical evidence that the WGAN loss is not a meaningful approximation of the Wasserstein distance. Moreover, we argue that the Wasserstein distance is not even a desirable loss function for deep generative models, and conclude that the success of Wasserstein GANs can in truth be attributed to a failure to approximate the Wasserstein distance."
                },
                {
                    "title": "Continuous Wasserstein-2 Barycenter Estimation without Minimax Optimization",
                    "abstract": "Wasserstein barycenters provide a geometric notion of the weighted average of probability measures based on optimal transport. In this paper, we present a scalable algorithm to compute Wasserstein-2 barycenters given sample access to the input measures, which are not restricted to being discrete. While past approaches rely on entropic or quadratic regularization, we employ input convex neural networks and cycle-consistency regularization to avoid introducing bias. As a result, our approach does not resort to minimax optimization. We provide theoretical analysis on error bounds as well as empirical evidence of the effectiveness of the proposed approach in low-dimensional qualitative scenarios and high-dimensional quantitative experiments."
                },
                {
                    "title": "Convex Potential Flows: Universal Probability Distributions with Optimal Transport and Convex Optimization",
                    "abstract": "Flow-based models are powerful tools for designing probabilistic models with tractable density. This paper introduces Convex Potential Flows (CP-Flow), a natural and efficient parameterization of invertible models inspired by the optimal transport (OT) theory. CP-Flows are the gradient map of a strongly convex neural potential function. The convexity implies invertibility and allows us to resort to convex optimization to solve the convex conjugate for efficient inversion. To enable maximum likelihood training, we derive a new gradient estimator of the log-determinant of the Jacobian, which involves solving an inverse-Hessian vector product using the conjugate gradient method. The gradient estimator has constant-memory cost, and can be made effectively unbiased by reducing the error tolerance level of the convex optimization routine. Theoretically, we prove that CP-Flows are universal density approximators and are optimal in the OT sense. Our empirical results show that CP-Flow performs competitively on standard benchmarks of density estimation and variational inference."
                },
                {
                    "title": "Scalable Computations of Wasserstein Barycenter via Input Convex Neural Networks",
                    "abstract": "Wasserstein Barycenter is a principled approach to represent the weighted mean of a given set of probability distributions, utilizing the geometry induced by optimal transport. In this work, we present a novel scalable algorithm to approximate the Wasserstein Barycenters aiming at high-dimensional applications in machine learning. Our proposed algorithm is based on the Kantorovich dual formulation of the 2-Wasserstein distance as well as a recent neural network architecture, input convex neural network, that is known to parametrize convex functions. The distinguishing features of our method are: i) it only requires samples from the marginal distributions; ii) unlike the existing semi-discrete approaches, it represents the Barycenter with a generative model; iii) it allows to compute the barycenter with arbitrary weights after one training session. We demonstrate the efficacy of our algorithm by comparing it with the state-of-art methods in multiple experiments."
                },
                {
                    "title": "Adversarial Latent Autoencoders",
                    "abstract": "Autoencoder networks are unsupervised approaches aiming at combining generative and representational properties by learning simultaneously an encoder-generator map. Although studied extensively, the issues of whether they have the same generative power of GANs, or learn disentangled representations, have not been fully addressed. We introduce an autoencoder that tackles these issues jointly, which we call Adversarial Latent Autoencoder (ALAE). It is a general architecture that can leverage recent improvements on GAN training procedures. We designed two autoencoders: one based on a MLP encoder, and another based on a StyleGAN generator, which we call StyleALAE. We verify the disentanglement properties of both architectures. We show that StyleALAE can not only generate 1024x1024 face images with comparable quality of StyleGAN, but at the same resolution can also produce face reconstructions and manipulations based on real images. This makes ALAE the first autoencoder able to compare with, and go beyond the capabilities of a generator-only type of architecture."
                },
                {
                    "title": "Learning with minibatch Wasserstein : asymptotic and gradient properties",
                    "abstract": "Optimal transport distances are powerful tools to compare probability distributions and have found many applications in machine learning. Yet their algorithmic complexity prevents their direct use on large scale datasets. To overcome this challenge, practitioners compute these distances on minibatches {\\em i.e.} they average the outcome of several smaller optimal transport problems. We propose in this paper an analysis of this practice, which effects are not well understood so far. We notably argue that it is equivalent to an implicit regularization of the original problem, with appealing properties such as unbiased estimators, gradients and a concentration bound around the expectation, but also with defects such as loss of distance property. Along with this theoretical analysis, we also conduct empirical experiments on gradient flows, GANs or color transfer that highlight the practical interest of this strategy."
                },
                {
                    "title": "Wasserstein-2 Generative Networks",
                    "abstract": "Generative Adversarial Networks training is not easy due to the minimax nature of the optimization objective. In this paper, we propose a novel end-to-end algorithm for training generative models which uses a non-minimax objective simplifying model training. The proposed algorithm uses the approximation of Wasserstein-2 distance by Input Convex Neural Networks. From the theoretical side, we estimate the properties of the generative mapping fitted by the algorithm. From the practical side, we conduct computational experiments which confirm the efficiency of our algorithm in various applied problems: image-to-image color transfer, latent space optimal transport, image-to-image style transfer, and domain adaptation."
                },
                {
                    "title": "Optimal transport mapping via input convex neural networks",
                    "abstract": "In this paper, we present a novel and principled approach to learn the optimal transport between two distributions, from samples. Guided by the optimal transport theory, we learn the optimal Kantorovich potential which induces the optimal transport map. This involves learning two convex functions, by solving a novel minimax optimization. Building upon recent advances in the field of input convex neural networks, we propose a new framework where the gradient of one convex function represents the optimal transport mapping. Numerical experiments confirm that we learn the optimal transport mapping. This approach ensures that the transport mapping we find is optimal independent of how we initialize the neural networks. Further, target distributions from a discontinuous support can be easily captured, as gradient of a convex function naturally models a {\\em discontinuous} transport mapping."
                },
                {
                    "title": "2-Wasserstein Approximation via Restricted Convex Potentials with Application to Improved Training for GANs",
                    "abstract": "We provide a framework to approximate the 2-Wasserstein distance and the optimal transport map, amenable to efficient training as well as statistical and geometric analysis. With the quadratic cost and considering the Kantorovich dual form of the optimal transportation problem, the Brenier theorem states that the optimal potential function is convex and the optimal transport map is the gradient of the optimal potential function. Using this geometric structure, we restrict the optimization problem to different parametrized classes of convex functions and pay special attention to the class of input-convex neural networks. We analyze the statistical generalization and the discriminative power of the resulting approximate metric, and we prove a restricted moment-matching property for the approximate optimal map. Finally, we discuss a numerical algorithm to solve the restricted optimization problem and provide numerical experiments to illustrate and compare the proposed approach with the established regularization-based approaches. We further discuss practical implications of our proposal in a modular and interpretable design for GANs which connects the generator training with discriminator computations to allow for learning an overall composite generator."
                },
                {
                    "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                    "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                },
                {
                    "title": "Optimal Control Via Neural Networks: A Convex Approach",
                    "abstract": "Control of complex systems involves both system identification and controller design. Deep neural networks have proven to be successful in many identification tasks, however, from model-based control perspective, these networks are difficult to work with because they are typically nonlinear and nonconvex. Therefore many systems are still identified and controlled based on simple linear models despite their poor representation capability. In this paper we bridge the gap between model accuracy and control tractability faced by neural networks, by explicitly constructing networks that are convex with respect to their inputs. We show that these input convex networks can be trained to obtain accurate models of complex physical systems. In particular, we design input convex recurrent neural networks to capture temporal behavior of dynamical systems. Then optimal controllers can be achieved via solving a convex model predictive control problem. Experiment results demonstrate the good potential of the proposed input convex neural network based approach in a variety of control applications. In particular we show that in the MuJoCo locomotion tasks, we could achieve over 10% higher performance using 5* less time compared with state-of-the-art model-based reinforcement learning method; and in the building HVAC control example, our method achieved up to 20% energy reduction compared with classic linear models."
                },
                {
                    "title": "Computational Optimal Transport: Complexity by Accelerated Gradient Descent Is Better Than by Sinkhorn's Algorithm",
                    "abstract": "We analyze two algorithms for approximating the general optimal transport (OT) distance between two discrete distributions of size $n$, up to accuracy $\\varepsilon$. For the first algorithm, which is based on the celebrated Sinkhorn's algorithm, we prove the complexity bound $\\widetilde{O}\\left({n^2/\\varepsilon^2}\\right)$ arithmetic operations. For the second one, which is based on our novel Adaptive Primal-Dual Accelerated Gradient Descent (APDAGD) algorithm, we prove the complexity bound $\\widetilde{O}\\left(\\min\\left\\{n^{9/4}/\\varepsilon, n^{2}/\\varepsilon^2 \\right\\}\\right)$ arithmetic operations. Both bounds have better dependence on $\\varepsilon$ than the state-of-the-art result given by $\\widetilde{O}\\left({n^2/\\varepsilon^3}\\right)$. Our second algorithm not only has better dependence on $\\varepsilon$ in the complexity bound, but also is not specific to entropic regularization and can solve the OT problem with different regularizers."
                },
                {
                    "title": "Large Scale Optimal Transport and Mapping Estimation",
                    "abstract": "This paper presents a novel two-step approach for the fundamental problem of learning an optimal map from one distribution to another. First, we learn an optimal transport (OT) plan, which can be thought as a one-to-many map between the two distributions. To that end, we propose a stochastic dual approach of regularized OT, and show empirically that it scales better than a recent related approach when the amount of samples is very large. Second, we estimate a \\textit{Monge map} as a deep neural network learned by approximating the barycentric projection of the previously-obtained OT plan. This parameterization allows generalization of the mapping outside the support of the input measure. We prove two theoretical stability results of regularized OT which show that our estimations converge to the OT plan and Monge map between the underlying continuous measures. We showcase our proposed approach on two applications: domain adaptation and generative modeling."
                },
                {
                    "title": "Wasserstein Distance Guided Representation Learning for Domain Adaptation",
                    "abstract": "\n \n Domain adaptation aims at generalizing a high-performance learner on a target domain via utilizing the knowledge distilled from a source domain which has a different but related data distribution. One solution to domain adaptation is to learn domain invariant feature representations while the learned representations should also be discriminative in prediction. To learn such representations, domain adaptation frameworks usually include a domain invariant representation learning approach to measure and reduce the domain discrepancy, as well as a discriminator for classification. Inspired by Wasserstein GAN, in this paper we propose a novel approach to learn domain invariant feature representations, namely Wasserstein Distance Guided Representation Learning (WDGRL). WDGRL utilizes a neural network, denoted by the domain critic, to estimate empirical Wasserstein distance between the source and target samples and optimizes the feature extractor network to minimize the estimated Wasserstein distance in an adversarial manner. The theoretical advantages of Wasserstein distance for domain adaptation lie in its gradient property and promising generalization bound. Empirical studies on common sentiment and image classification adaptation datasets demonstrate that our proposed WDGRL outperforms the state-of-the-art domain invariant representation learning approaches.\n \n"
                },
                {
                    "title": "On parameter estimation with the Wasserstein distance",
                    "abstract": "\n Statistical inference can be performed by minimizing, over the parameter space, the Wasserstein distance between model distributions and the empirical distribution of the data. We study asymptotic properties of such minimum Wasserstein distance estimators, complementing results derived by Bassetti, Bodini and Regazzini in 2006. In particular, our results cover the misspecified setting, in which the data-generating process is not assumed to be part of the family of distributions described by the model. Our results are motivated by recent applications of minimum Wasserstein estimators to complex generative models. We discuss some difficulties arising in the numerical approximation of these estimators. Two of our numerical examples ($g$-and-$\\kappa$ and sum of log-normals) are taken from the literature on approximate Bayesian computation and have likelihood functions that are not analytically tractable. Two other examples involve misspecified models."
                },
                {
                    "title": "An overview of gradient descent optimization algorithms",
                    "abstract": "The statistical testing technique is considered to compare the metrics values of machine learning models on a test set. Since the values of metrics depend not only on the models, but also on the data, it may turn out that different models are the best on different test sets. For this reason, the traditional approach to comparing the values of metrics on a test set is often not enough. Sometimes a statistical comparison of the results obtained on the basis of cross-validation is used, but in this case it is impossible to guarantee the independence of the obtained measurements, which does not allow the use of the Student's t-test. There are criteria that do not require independent measurements, but they have less power. For additive metrics, a technique is proposed in this paper, when a test sample is divided into N parts, on each of which the values of the metrics are calculated. Since the value on each part is obtained as the sum of independent random variables, according to the central limit theorem, the obtained metrics values on each of the N parts are realizations of the normally distributed random variable. To estimate the required sample size, it is proposed to use normality tests and build quantile\u2013 quantile plots. You can then use a modification of the Student's t-test to conduct a statistical test comparing the mean values of the metrics. A simplified approach is also considered, in which confidence intervals are built for the base model. A model whose metric values do not fall into this interval works differently from the base model. This approach reduces the amount of computations needed, however, an experimental analysis of the binary cross-entropy metric for CTR (Click-Through Rate) prediction models showed that it is more rough than the first one."
                },
                {
                    "title": "Input Convex Neural Networks",
                    "abstract": "This paper presents the input convex neural network architecture. These are scalar-valued (potentially deep) neural networks with constraints on the network parameters such that the output of the network is a convex function of (some of) the inputs. The networks allow for efficient inference via optimization over some inputs to the network given others, and can be applied to settings including structured prediction, data imputation, reinforcement learning, and others. In this paper we lay the basic groundwork for these models, proposing methods for inference, optimization and learning, and analyze their representational power. We show that many existing neural network architectures can be made input-convex with a minor modification, and develop specialized optimization algorithms tailored to this setting. Finally, we highlight the performance of the methods on multi-label prediction, image completion, and reinforcement learning problems, where we show improvement over the existing state of the art in many cases."
                },
                {
                    "title": "Stochastic Optimization for Large-scale Optimal Transport",
                    "abstract": "Optimal transport (OT) defines a powerful framework to compare probability distributions in a geometrically faithful way. However, the practical impact of OT is still limited because of its computational burden. We propose a new class of stochastic optimization algorithms to cope with large-scale problems routinely encountered in machine learning applications. These methods are able to manipulate arbitrary distributions (either discrete or continuous) by simply requiring to be able to draw samples from them, which is the typical setup in high-dimensional learning problems. This alleviates the need to discretize these densities, while giving access to provably convergent methods that output the correct distance without discretization error. These algorithms rely on two main ideas: (a) the dual OT problem can be re-cast as the maximization of an expectation ; (b) entropic regularization of the primal OT problem results in a smooth dual optimization optimization which can be addressed with algorithms that have a provably faster convergence. We instantiate these ideas in three different setups: (i) when comparing a discrete distribution to another, we show that incremental stochastic optimization schemes can beat Sinkhorn's algorithm, the current state-of-the-art finite dimensional OT solver; (ii) when comparing a discrete distribution to a continuous density, a semi-discrete reformulation of the dual program is amenable to averaged stochastic gradient descent, leading to better performance than approximately solving the problem by discretization ; (iii) when dealing with two continuous densities, we propose a stochastic gradient descent over a reproducing kernel Hilbert space (RKHS). This is currently the only known method to solve this problem, apart from computing OT on finite samples. We backup these claims on a set of discrete, semi-discrete and continuous benchmark problems."
                },
                {
                    "title": "A survey of the Schr\\\"odinger problem and some of its connections with optimal transport",
                    "abstract": "This article is aimed at presenting the Schr\\\"odinger problem and some of its connections with optimal transport. We hope that it can be used as a basic user's guide to Schr\\\"odinger problem. We also give a survey of the related literature. In addition, some new results are proved."
                },
                {
                    "title": "Sinkhorn Distances: Lightspeed Computation of Optimal Transport",
                    "abstract": "Optimal transportation distances are a fundamental family of parameterized distances for histograms. Despite their appealing theoretical properties, excellent performance in retrieval tasks and intuitive formulation, their computation involves the resolution of a linear program whose cost is prohibitive whenever the histograms' dimension exceeds a few hundreds. We propose in this work a new family of optimal transportation distances that look at transportation problems from a maximum-entropy perspective. We smooth the classical optimal transportation problem with an entropic regularization term, and show that the resulting optimum is also a distance which can be computed through Sinkhorn-Knopp's matrix scaling algorithm at a speed that is several orders of magnitude faster than that of transportation solvers. We also report improved performance over classical optimal transportation distances on the MNIST benchmark problem."
                },
                {
                    "title": "Gradient networks",
                    "abstract": "Gradient networks are defined (Toroczkai and Bassler 2004 Nature 428 716) as directed graphs formed by local gradients of a scalar field distributed on the nodes of a substrate network G. We present the derivation for some of the general properties of gradient graphs and give an exact expression for the in-degree distribution of the gradient network when the substrate is a binomial (Erds\u2013R\u00e9nyi) random graph, , and the scalars are independent identically distributed (i.i.d.) random variables. We show that in the limit for , i.e., gradient networks become scale-free graphs up to a cut-off degree. This paper presents the detailed derivation of the results announced in Toroczkai and Bassler (2004 Nature 428 716)."
                },
                {
                    "title": "Conditional Flow Matching: Simulation-Free Dynamic Optimal Transport",
                    "abstract": "Continuous normalizing \ufb02ows (CNFs) are an attractive generative modeling technique, but they have thus far been held back by limitations in their simulation-based maximum likelihood training. In this paper, we introduce a new technique called conditional \ufb02ow matching (CFM), a simulation-free training objective for CNFs. CFM features a stable regression objective like that used to train the stochastic \ufb02ow in diffusion models but enjoys the ef\ufb01cient inference of deterministic \ufb02ow models. In contrast to both diffusion models and prior CNF training algorithms, our CFM objec-tive does not require the source distribution to be Gaussian or require evaluation of its density. Based on this new objective, we also introduce optimal transport CFM (OT-CFM), which creates simpler \ufb02ows that are more stable to train and lead to faster inference, as evaluated in our experiments. Training CNFs with CFM improves results on a variety of conditional and unconditional generation tasks such as inferring single cell dynamics, unsupervised image translation, and Schr \u00a8 odinger bridge inference. Code is available at https://github.com/atong01/ conditional-flow-matching ."
                },
                {
                    "title": "Most: multi-source domain adaptation via optimal transport for student-teacher learning",
                    "abstract": "Multi-source domain adaptation (DA) is more challenging than conventional DA because the knowledge is transferred from several source domains to a target domain. To this end, we propose in this paper a novel model for multi-source DA using the theory of optimal transport and imitation learning. More speci\ufb01cally, our approach consists of two cooperative agents: a teacher classi\ufb01er and a student classi\ufb01er. The teacher classi\ufb01er is a combined expert that leverages knowledge of domain experts that can be theoretically guaranteed to handle perfectly source examples, while the student classi\ufb01er acting on the target domain tries to imitate the teacher classi\ufb01er acting on the source domains. Our rigorous theory developed based on optimal transport makes this cross-domain imitation possible and also helps to mitigate not only the data shift but also the label shift, which are inherently thorny issues in DA research. We conduct comprehensive experiments on real-world datasets to demonstrate the merit of our approach and its optimal transport based imitation learning view-point. Experimental results show that our proposed method achieves state-of-the-art performance on benchmark datasets for multi-source domain adaptation including Digits-\ufb01ve, Of\ufb01ce-Caltech10, and Of\ufb01ce-31 to the best of our knowledge."
                },
                {
                    "title": "The Fastest Known Globally Convergent First-Order Method for Minimizing Strongly Convex Functions",
                    "abstract": "We design and analyze a novel gradient-based algorithm for unconstrained convex optimization. When the objective function is <inline-formula> <tex-math notation=\"LaTeX\">$m$ </tex-math></inline-formula>-strongly convex and its gradient is <inline-formula> <tex-math notation=\"LaTeX\">$L$ </tex-math></inline-formula>-Lipschitz continuous, the iterates and function values converge linearly to the optimum at rates <inline-formula> <tex-math notation=\"LaTeX\">$\\rho $ </tex-math></inline-formula> and <inline-formula> <tex-math notation=\"LaTeX\">$\\rho ^{2}$ </tex-math></inline-formula>, respectively, where <inline-formula> <tex-math notation=\"LaTeX\">$\\rho = 1-\\sqrt {m/L}$ </tex-math></inline-formula>. These are the fastest known guaranteed linear convergence rates for globally convergent first-order methods, and for high desired accuracies the corresponding iteration complexity is within a factor of two of the theoretical lower bound. We use a simple graphical design procedure based on integral quadratic constraints to derive closed-form expressions for the algorithm parameters. The new algorithm, which we call the triple momentum method, can be seen as an extension of methods such as gradient descent, Nesterov\u2019s accelerated gradient descent, and the heavy-ball method."
                },
                {
                    "title": "Parametric computation of the Legendre-Fenchel conjugate with application to the computation of the Moreau envelope",
                    "abstract": "A new algorithm, named the Parametric Legendre Transform (PLT) algorithm, to compute the Legendre-Fenchel conjugate of a convex function of one variable is presented. It returns a parameterization of the graph of the conjugate except for some affine parts corresponding to nondifferentiable points of the function. The approach is extended to the computation of the Moreau envelope, resulting in a simple yet efficient algorithm. Theoretical results, the description (and extension) of the algorithm, its approximation error and the convergence, as well as the comparison with known algorithms are included."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a method that guarantees straight trajectories in Flow Matching models without the need for time-consuming ODE integration?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the inefficiencies in current generative modeling techniques, particularly those relying on Flow Matching. By ensuring straight trajectories, we can significantly reduce computational costs and improve the speed of sampling processes. This advancement could lead to more efficient generative models, enabling broader applications in fields such as image synthesis, natural language processing, and beyond. Furthermore, it could inspire future research to explore new methodologies in generative modeling and optimal transport, potentially leading to breakthroughs in how we understand and implement these concepts.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexity of ensuring straight trajectories while maintaining the accuracy of the Flow Matching process. Naive approaches may fail because they do not account for the accumulation of errors during iterations, which can degrade performance. Additionally, the connection between Flow Matching and Optimal Transport introduces theoretical complexities, as existing methods do not guarantee straight paths due to biases in minibatch OT. Overcoming these technical obstacles requires a deep understanding of both the mathematical foundations of optimal transport and the practical implications of implementing these methods in generative models.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on improving Flow Matching and related methods without adequately addressing the issue of trajectory straightening. Limitations in earlier approaches, such as the accumulation of errors in Rectified Flow and the heuristic nature of OT-CFM, have prevented a comprehensive solution. Additionally, the lack of a theoretical framework that guarantees straight paths has been a significant barrier. Our approach differs by proposing the Optimal Flow Matching (OFM) method, which directly targets the generation of straight trajectories through the use of specific vector fields, thereby providing a more robust solution.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Optimal Flow Matching (OFM), involves a single iteration of Flow Matching that yields straight trajectories without the need for ODE solving. We will utilize Input Convex Neural Networks to parameterize the gradients of convex functions, which are essential for generating the desired vector fields. The dataset will consist of various probability distributions, and we will evaluate the performance using metrics that assess the efficiency and accuracy of the generated samples. We"
            }
        },
        "author_data": {
            "b4f7e1a3-a85d-4298-83e0-eea2500ac2e4": {
                "pk": "b4f7e1a3-a85d-4298-83e0-eea2500ac2e4",
                "name": "Nikita Kornilov",
                "collaborators": [
                    "Alexander Gasnikov",
                    "Eduard Gorbunov",
                    "Pavel Dvurechensky",
                    "Darina Dvinskikh",
                    "Yuriy Dorn",
                    "Nikolay Kutuzov",
                    "Alexander Nazin",
                    "Aleksandr Lobanov",
                    "Innokentiy Shibaev",
                    "Sergei Kholkin"
                ],
                "domain": [
                    "Optimization",
                    "Stochastic Methods",
                    "Machine Learning",
                    "Multi-Armed Bandit"
                ],
                "publications": [
                    {
                        "title": "Gradient-Free Methods for Non-Smooth Convex Stochastic Optimization with Heavy-Tailed Noise on Convex Compact",
                        "abstract": "We present two easy-to-implement gradient-free/zeroth-order methods to optimize a stochastic non-smooth function accessible only via a black-box. The methods are built upon efficient first-order methods in the heavy-tailed case, i.e., when the gradient noise has infinite variance but bounded $(1+\\kappa)$-th moment for some $\\kappa \\in(0,1]$. The first algorithm is based on the stochastic mirror descent with a particular class of uniformly convex mirror maps which is robust to heavy-tailed noise. The second algorithm is based on the stochastic mirror descent and gradient clipping technique. Additionally, for the objective functions satisfying the $r$-growth condition, faster algorithms are proposed based on these methods and the restart technique."
                    },
                    {
                        "title": "Diffusion & Adversarial Schr\u00f6dinger Bridges via Iterative Proportional Markovian Fitting",
                        "abstract": "The Iterative Markovian Fitting (IMF) procedure based on iterative reciprocal and Markovian projections has recently been proposed as a powerful method for solving the Schr\\\"odinger Bridge problem. However, it has been observed that for the practical implementation of this procedure, it is crucial to alternate between fitting a forward and backward time diffusion at each iteration. Such implementation is thought to be a practical heuristic, which is required to stabilize training and obtain good results in applications such as unpaired domain translation. In our work, we show that this heuristic closely connects with the pioneer approaches for the Schr\\\"odinger Bridge based on the Iterative Proportional Fitting (IPF) procedure. Namely, we find that the practical implementation of IMF is, in fact, a combination of IMF and IPF procedures, and we call this combination the Iterative Proportional Markovian Fitting (IPMF) procedure. We show both theoretically and practically that this combined IPMF procedure can converge under more general settings, thus, showing that the IPMF procedure opens a door towards developing a unified framework for solving Schr\\\"odinger Bridge problems."
                    },
                    {
                        "title": "Intermediate Gradient Methods with Relative Inexactness",
                        "abstract": "This paper is devoted to first-order algorithms for smooth convex optimization with inexact gradients. Unlike the majority of the literature on this topic, we consider the setting of relative rather than absolute inexactness. More precisely, we assume that an additive error in the gradient is proportional to the gradient norm, rather than being globally bounded by some small quantity. We propose a novel analysis of the accelerated gradient method under relative inexactness and strong convexity and improve the bound on the maximum admissible error that preserves the linear convergence of the algorithm. In other words, we analyze how robust is the accelerated gradient method to the relative inexactness of the gradient information. Moreover, based on the Performance Estimation Problem (PEP) technique, we show that the obtained result is optimal for the family of accelerated algorithms we consider. Motivated by the existing intermediate methods with absolute error, i.e., the methods with convergence rates that interpolate between slower but more robust non-accelerated algorithms and faster, but less robust accelerated algorithms, we propose an adaptive variant of the intermediate gradient method with relative error in the gradient."
                    },
                    {
                        "title": "Implicitly normalized forecaster with clipping for linear and non-linear heavy-tailed multi-armed bandits",
                        "abstract": "The Implicitly Normalized Forecaster (INF) algorithm is considered to be an optimal solution for adversarial multi-armed bandit (MAB) problems. However, most of the existing complexity results for INF rely on restrictive assumptions, such as bounded rewards. Recently, a related algorithm was proposed that works for both adversarial and stochastic heavy-tailed MAB settings. However, this algorithm fails to fully exploit the available data.   In this paper, we propose a new version of INF called the Implicitly Normalized Forecaster with clipping (INF-clip) for MAB problems with heavy-tailed reward distributions. We establish convergence results under mild assumptions on the rewards distribution and demonstrate that INF-clip is optimal for linear heavy-tailed stochastic MAB problems and works well for non-linear ones. Furthermore, we show that INF-clip outperforms the best-of-both-worlds algorithm in cases where it is difficult to distinguish between different arms."
                    },
                    {
                        "title": "Median Clipping for Zeroth-order Non-Smooth Convex Optimization and Multi-Armed Bandit Problem with Heavy-tailed Symmetric Noise",
                        "abstract": "In this paper, we consider non-smooth convex optimization with a zeroth-order oracle corrupted by symmetric stochastic noise. Unlike the existing high-probability results requiring the noise to have bounded $\\kappa$-th moment with $\\kappa \\in (1,2]$, our results allow even heavier noise with any $\\kappa > 0$, e.g., the noise distribution can have unbounded expectation. Our convergence rates match the best-known ones for the case of the bounded variance. To achieve this, we build the median gradient estimate with bounded second moment as the mini-batched median of the sampled gradient differences. We apply this technique to the stochastic multi-armed bandit problem with heavy-tailed distribution of rewards and achieve $\\tilde{O}(\\sqrt{dT})$ regret. We demonstrate the performance of our zeroth-order and MAB algorithms for different $\\kappa$ on synthetic and real-world data. Our methods do not lose to SOTA approaches, moreover, they dramatically outperform SOTA for $\\kappa \\leq 1$."
                    },
                    {
                        "title": "Accelerated Zeroth-order Method for Non-Smooth Stochastic Convex Optimization Problem with Infinite Variance",
                        "abstract": "In this paper, we consider non-smooth stochastic convex optimization with two function evaluations per round under infinite noise variance. In the classical setting when noise has finite variance, an optimal algorithm, built upon the batched accelerated gradient method, was proposed in (Gasnikov et. al., 2022). This optimality is defined in terms of iteration and oracle complexity, as well as the maximal admissible level of adversarial noise. However, the assumption of finite variance is burdensome and it might not hold in many practical scenarios. To address this, we demonstrate how to adapt a refined clipped version of the accelerated gradient (Stochastic Similar Triangles) method from (Sadiev et al., 2023) for a two-point zero-order oracle. This adaptation entails extending the batching technique to accommodate infinite variance -- a non-trivial task that stands as a distinct contribution of this paper."
                    }
                ]
            },
            "c5450154-2d31-40b0-b71e-3cba0f5d8878": {
                "pk": "c5450154-2d31-40b0-b71e-3cba0f5d8878",
                "name": "Petr Mokrov",
                "collaborators": [
                    "Alexander Korotin",
                    "Evgeny Burnaev",
                    "Alexander Kolesov",
                    "Milena Gazdieva",
                    "Igor Udovichenko",
                    "Gudmund Pammer",
                    "Nikita Gushchin",
                    "Xavier Aramayo Carrasco",
                    "Maksim Nekrashevich",
                    "Polina Karpikova"
                ],
                "domain": [
                    "Optimal Transport",
                    "Machine Learning",
                    "Generative Models",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Estimating Barycenters of Distributions with Neural Optimal Transport",
                        "abstract": "Given a collection of probability measures, a practitioner sometimes needs to find an \"average\" distribution which adequately aggregates reference distributions. A theoretically appealing notion of such an average is the Wasserstein barycenter, which is the primal focus of our work. By building upon the dual formulation of Optimal Transport (OT), we propose a new scalable approach for solving the Wasserstein barycenter problem. Our methodology is based on the recent Neural OT solver: it has bi-level adversarial learning objective and works for general cost functions. These are key advantages of our method since the typical adversarial algorithms leveraging barycenter tasks utilize tri-level optimization and focus mostly on quadratic cost. We also establish theoretical error bounds for our proposed approach and showcase its applicability and effectiveness in illustrative scenarios and image data setups. Our source code is available at https://github.com/justkolesov/NOTBarycenters."
                    },
                    {
                        "title": "Uncovering Challenges of Solving the Continuous Gromov-Wasserstein Problem",
                        "abstract": "Recently, the Gromov-Wasserstein Optimal Transport (GWOT) problem has attracted the special attention of the ML community. In this problem, given two distributions supported on two (possibly different) spaces, one has to find the most isometric map between them. In the discrete variant of GWOT, the task is to learn an assignment between given discrete sets of points. In the more advanced continuous formulation, one aims at recovering a parametric mapping between unknown continuous distributions based on i.i.d. samples derived from them. The clear geometrical intuition behind the GWOT makes it a natural choice for several practical use cases, giving rise to a number of proposed solvers. Some of them claim to solve the continuous version of the problem. At the same time, GWOT is notoriously hard, both theoretically and numerically. Moreover, all existing continuous GWOT solvers still heavily rely on discrete techniques. Natural questions arise: to what extent existing methods unravel GWOT problem, what difficulties they encounter, and under which conditions they are successful. Our benchmark paper is an attempt to answer these questions. We specifically focus on the continuous GWOT as the most interesting and debatable setup. We crash-test existing continuous GWOT approaches on different scenarios, carefully record and analyze the obtained results, and identify issues. Our findings experimentally testify that the scientific community is still missing a reliable continuous GWOT solver, which necessitates further research efforts. As the first step in this direction, we propose a new continuous GWOT method which does not rely on discrete techniques and partially solves some of the problems of the competitors. Our code is available at https://github.com/Ark-130994/GW-Solvers."
                    },
                    {
                        "title": "Energy-guided Entropic Neural Optimal Transport",
                        "abstract": "Energy-based models (EBMs) are known in the Machine Learning community for decades. Since the seminal works devoted to EBMs dating back to the noughties, there have been a lot of efficient methods which solve the generative modelling problem by means of energy potentials (unnormalized likelihood functions). In contrast, the realm of Optimal Transport (OT) and, in particular, neural OT solvers is much less explored and limited by few recent works (excluding WGAN-based approaches which utilize OT as a loss function and do not model OT maps themselves). In our work, we bridge the gap between EBMs and Entropy-regularized OT. We present a novel methodology which allows utilizing the recent developments and technical improvements of the former in order to enrich the latter. From the theoretical perspective, we prove generalization bounds for our technique. In practice, we validate its applicability in toy 2D and image domains. To showcase the scalability, we empower our method with a pre-trained StyleGAN and apply it to high-res AFHQ $512\\times 512$ unpaired I2I translation. For simplicity, we choose simple short- and long-run EBMs as a backbone of our Energy-guided Entropic OT approach, leaving the application of more sophisticated EBMs for future research. Our code is available at: https://github.com/PetrMokrov/Energy-guided-Entropic-OT"
                    },
                    {
                        "title": "Building the Bridge of Schr\u00f6dinger: A Continuous Entropic Optimal Transport Benchmark",
                        "abstract": "Over the last several years, there has been significant progress in developing neural solvers for the Schr\\\"odinger Bridge (SB) problem and applying them to generative modelling. This new research field is justifiably fruitful as it is interconnected with the practically well-performing diffusion models and theoretically grounded entropic optimal transport (EOT). Still, the area lacks non-trivial tests allowing a researcher to understand how well the methods solve SB or its equivalent continuous EOT problem. We fill this gap and propose a novel way to create pairs of probability distributions for which the ground truth OT solution is known by the construction. Our methodology is generic and works for a wide range of OT formulations, in particular, it covers the EOT which is equivalent to SB (the main interest of our study). This development allows us to create continuous benchmark distributions with the known EOT and SB solutions on high-dimensional spaces such as spaces of images. As an illustration, we use these benchmark pairs to test how well existing neural EOT/SB solvers actually compute the EOT solution. Our code for constructing benchmark pairs under different setups is available at: https://github.com/ngushchin/EntropicOTBenchmark."
                    },
                    {
                        "title": "Neural Optimal Transport with General Cost Functionals",
                        "abstract": "We introduce a novel neural network-based algorithm to compute optimal transport (OT) plans for general cost functionals. In contrast to common Euclidean costs, i.e., $\\ell^1$ or $\\ell^2$, such functionals provide more flexibility and allow using auxiliary information, such as class labels, to construct the required transport map. Existing methods for general costs are discrete and have limitations in practice, i.e. they do not provide an out-of-sample estimation. We address the challenge of designing a continuous OT approach for general costs that generalizes to new data points in high-dimensional spaces, such as images. Additionally, we provide the theoretical error analysis for our recovered transport plans. As an application, we construct a cost functional to map data distributions while preserving the class-wise structure."
                    },
                    {
                        "title": "Energy-Guided Continuous Entropic Barycenter Estimation for General Costs",
                        "abstract": "Optimal transport (OT) barycenters are a mathematically grounded way of averaging probability distributions while capturing their geometric properties. In short, the barycenter task is to take the average of a collection of probability distributions w.r.t. given OT discrepancies. We propose a novel algorithm for approximating the continuous Entropic OT (EOT) barycenter for arbitrary OT cost functions. Our approach is built upon the dual reformulation of the EOT problem based on weak OT, which has recently gained the attention of the ML community. Beyond its novelty, our method enjoys several advantageous properties: (i) we establish quality bounds for the recovered solution; (ii) this approach seamlessly interconnects with the Energy-Based Models (EBMs) learning procedure enabling the use of well-tuned algorithms for the problem of interest; (iii) it provides an intuitive optimization scheme avoiding min-max, reinforce and other intricate technical tricks. For validation, we consider several low-dimensional scenarios and image-space setups, including non-Euclidean cost functions. Furthermore, we investigate the practical task of learning the barycenter on an image manifold generated by a pretrained generative model, opening up new directions for real-world applications."
                    },
                    {
                        "title": "Large-Scale Wasserstein Gradient Flows",
                        "abstract": "Wasserstein gradient flows provide a powerful means of understanding and solving many diffusion equations. Specifically, Fokker-Planck equations, which model the diffusion of probability measures, can be understood as gradient descent over entropy functionals in Wasserstein space. This equivalence, introduced by Jordan, Kinderlehrer and Otto, inspired the so-called JKO scheme to approximate these diffusion processes via an implicit discretization of the gradient flow in Wasserstein space. Solving the optimization problem associated to each JKO step, however, presents serious computational challenges. We introduce a scalable method to approximate Wasserstein gradient flows, targeted to machine learning applications. Our approach relies on input-convex neural networks (ICNNs) to discretize the JKO steps, which can be optimized by stochastic gradient descent. Unlike previous work, our method does not require domain discretization or particle simulation. As a result, we can sample from the measure at each time step of the diffusion and compute its probability density. We demonstrate our algorithm's performance by computing diffusions following the Fokker-Planck equation and apply it to unnormalized density sampling as well as nonlinear filtering."
                    },
                    {
                        "title": "Robust Barycenter Estimation using Semi-Unbalanced Neural Optimal Transport",
                        "abstract": "A common challenge in aggregating data from multiple sources can be formalized as an \\textit{Optimal Transport} (OT) barycenter problem, which seeks to compute the average of probability distributions with respect to OT discrepancies. However, the presence of outliers and noise in the data measures can significantly hinder the performance of traditional statistical methods for estimating OT barycenters. To address this issue, we propose a novel, scalable approach for estimating the \\textit{robust} continuous barycenter, leveraging the dual formulation of the \\textit{(semi-)unbalanced} OT problem. To the best of our knowledge, this paper is the first attempt to develop an algorithm for robust barycenters under the continuous distribution setup. Our method is framed as a $\\min$-$\\max$ optimization problem and is adaptable to \\textit{general} cost function. We rigorously establish the theoretical underpinnings of the proposed method and demonstrate its robustness to outliers and class imbalance through a number of illustrative experiments."
                    }
                ]
            },
            "2920031d-3c92-4945-ad70-3db928b3b031": {
                "pk": "2920031d-3c92-4945-ad70-3db928b3b031",
                "name": "Alexander Gasnikov",
                "collaborators": [
                    "Alexander Rogozin",
                    "Evgenia Gasnikova",
                    "Anton Anikin",
                    "Alexander Gornov",
                    "Darina Dvinskikh",
                    "Alexander Beznosikov",
                    "Alexander Shananin",
                    "Michael Vaschenko",
                    "Evgeny Molchanov",
                    "Ludmila Pospelova"
                ],
                "domain": [
                    "Optimization",
                    "Traffic Assignment",
                    "Game Theory",
                    "Distributed Systems"
                ],
                "publications": [
                    {
                        "title": "Newton-Kantorovich-Vietoris method",
                        "abstract": "In this short work we consider nonlinear equation in Banach space $x = A(x)$. We combine classical Newton method with Vietoris method to propose new more general method. Under rather general conditions about the noise we investigate local convergence of the method."
                    },
                    {
                        "title": "Searching equillibriums in large transport networks",
                        "abstract": "In this Doctoral Dissertation we propose new variational principles for traffic assignment problems. So to find equillibrium we have to solve large-scale convex optimization problem of special (multilevel) type. We propose different numerical approaches for this problem."
                    },
                    {
                        "title": "Parallel and Distributed algorithms for ML problems",
                        "abstract": "In this paper we make a survey of modern parallel and distributed approaches to solve sum-type convex minimization problems come from ML applications."
                    },
                    {
                        "title": "About reduction of searching competetive equillibrium to the minimax problem in application to different network problems",
                        "abstract": "In the paper we show that an important class of multistage traffic equillibrium models (including correspondence matrix calculation, traffic assignment problem etc) and their economic generalizations can be considered as proper population games with the minimax structure of equillibriums. This is not trivial because at the first sight we have to consider competetive (Valras) equilibrium and Nash-Vardroop equillibrium."
                    },
                    {
                        "title": "Comparision of the definitions of generalized solution of the Cauchy problem for quasi-linear equation",
                        "abstract": "In preprint we consider and compare different definitions of generalized solution of the Cauchy problem for 1d-scalar quasilinear equation (conservation law). We start from the classical approaches goes back to I.M. Gelfand, O.A. Oleinik, S.N. Kruzhkov and move to the modern finite-difference approximations approaches belongs to A.A. Shananin and G.M. Henkin. We discuss the conditions that provide definitions to be equivalent."
                    },
                    {
                        "title": "Markov models of macrosystems",
                        "abstract": "We collect different examples reflect Bolzman--Jaynes theory of maximum entropy principle. This principle proposed that equillibrium of macrosystem (most probable macrostate of the invariant measure of macrosystem) can be find as a solution of the entropy linear programming problem. Among this examples one can find applications to the economics, sociology, linguistics, traffic flow theory, biology e.t.c. We also describe on examples the connection between Nash's equillibrium, Darwin's principle of natural selection and conception of equillibrium of macrosystem."
                    },
                    {
                        "title": "Efficient numerical methods for searching equillibriums in large transport networks",
                        "abstract": "We propose new variational principles for traffic assignment problems. So to find equillibrium we have to solve large-scale convex optimization problem of special type. We propose some kind of \"algebra\" on different models and corresponding numerical algorithms that allows us to combine the optimal method from the basis ones and predescribed operations on methods."
                    },
                    {
                        "title": "Analysis of tariff policy of a railway cargo transportation",
                        "abstract": "We describe a competetive equillibrium in a railway cargo transportation model. We reduce the problem of finding this equillibrium to the solution of to mutually dual convex optimization problems. According to L.V. Kantorvich we interpret an optimal traffic policy for the model in terms of Lagrange multipliers."
                    },
                    {
                        "title": "Universal gradient descent",
                        "abstract": "In this book we collect many different and useful facts around gradient descent method. First of all we consider gradient descent with inexact oracle. We build a general model of optimized function that include composite optimization approach, level's methods, proximal methods etc. Then we investigate primal-dual properties of the gradient descent in general model set-up. At the end we generalize method to universal one."
                    },
                    {
                        "title": "Non accelerated efficient numerical methods for sparse quadratic optimization problems and its generalizations",
                        "abstract": "We investigate primal gradient method with l1-norm and conditional gradient method (both methods are non accelerated). We show that these methods can outperform well known accelerated approaches for some classes of sparse quadratic problems. Moreover we discuss some generalizations."
                    },
                    {
                        "title": "Strongly convex stochastic online optimization on a unit simplex with application to the mixing least square regression",
                        "abstract": "In this paper we propose a new approach to obtain mixing least square regression estimate by means of stochastic online mirror descent in non-euclidian set-up."
                    },
                    {
                        "title": "Fast gradient descent method for convex optimization problems with an oracle that generates a $(\u03b4,L)$-model of a function in a requested point",
                        "abstract": "In this article we propose a new concept of a $(\\delta,L)$-model of a function which generalizes the concept of the $(\\delta,L)$-oracle (Devolder-Glineur-Nesterov). Using this concept we describe the gradient descent method and the fast gradient descent method and we show that many popular methods (composite optimization methods, level methods, conjugate gradient methods, proximal methods) are special cases of proposed methods in this article."
                    },
                    {
                        "title": "About universality of entropy in stochastic chemical kinetic models",
                        "abstract": "This paper studies the relationship between the Lyapunov function of a macrosystem whose dynamics is governed by the laws of stochastic chemical kinetics and the invariant measure of this macrosystem arising at large times. A necessary and sufficient condition for the reduction of the search problem for the equilibrium of the macrosystem (the most probable macrostate of the invariant measure of this macrosystem) to an entropy-linear programming problem is given."
                    },
                    {
                        "title": "Traffic assignment models. Numerical aspects",
                        "abstract": "In this book we describe BMW traffic assignment model and Nesterov-dePalma model. We consider Entropy model for demand matrix. Based on this models we build multi-stage traffic assignment models. The equilibrium in such models can be found from convex-concave saddle-point problem. We show how to solve this problem by using special combination of universal gradient method and Sinkhorn's algorithm."
                    },
                    {
                        "title": "An evolutionary view on equilibrium models of transport flows",
                        "abstract": "In this short paper we describe natural logit population games dynamics that explain equilibrium models of origin-destination matrix estimation and (stochastic) traffic assignment models (Beckmann, Nesterov--de Palma). Composition of the proposed dynamics allows to explain two-stages traffic assignment models."
                    },
                    {
                        "title": "On Decentralized Nonsmooth Optimization",
                        "abstract": "In decentralized optimization, several nodes connected by a network collaboratively minimize some objective function. For minimization of Lipschitz functions lower bounds are known along with optimal algorithms. We study a specific class of problems: linear models with nonsmooth loss functions. Our algorithm combines regularization and dual reformulation to get an effective optimization method with complexity better than the lower bounds."
                    },
                    {
                        "title": "Decentralized conditional gradient method over time-varying graphs",
                        "abstract": "In this paper we study a generalization of distributed conditional gradient method to time-varying network architectures. We theoretically analyze convergence properties of the algorithm and provide numerical experiments. The time-varying network is modeled as a deterministic of a stochastic sequence of graphs."
                    },
                    {
                        "title": "Randomization and sparsity in huge-scale optimization on the Mirror Descent example",
                        "abstract": "We investigate different randomizations for mirror descent method. We try to propose such a randomization that allows us to use sparsity of the problem as much as it possible. In the paper one can also find a generalization of randomizaed mirror descent for the convex optimization problems with functional restrictions."
                    },
                    {
                        "title": "Projected Gradient Method for Decentralized Optimization over Time-Varying Networks",
                        "abstract": "Decentralized distributed optimization over time-varying graphs (networks) is nowadays a very popular branch of research in optimization theory and consensus theory. One of the motivations to consider such networks is an application to drone networks. However, the first theoretical results in this branch appeared only five years ago (Nedic, 2014). The first results about the possibility of geometric rates of convergence for strongly convex smooth optimization problems on such networks were obtained only two years ago (Nedic, 2017). In this paper, we improve the rate of geometric convergence in the latter paper for the considered class of problems, using an original penalty method trick and robustness of projected gradient descent."
                    },
                    {
                        "title": "Entropy linear programming",
                        "abstract": "We propose an efficient dual algorithm for ELP based on Fast Gradient Method. The basic idea - to solve properly regularized dual problem."
                    }
                ]
            },
            "e27a5c9a-3584-4148-a751-26a868e6a62e": {
                "pk": "e27a5c9a-3584-4148-a751-26a868e6a62e",
                "name": "Alexander Korotin",
                "collaborators": [
                    "Evgeny Burnaev",
                    "Alexander Kolesov",
                    "Lingxiao Li",
                    "Vladimir V'yugin",
                    "Justin Solomon",
                    "Vage Egiazarian",
                    "Arip Asadulaev",
                    "Petr Mokrov",
                    "Milena Gazdieva",
                    "Daniil Selikhanovych"
                ],
                "domain": [
                    "Optimal Transport",
                    "Generative Models",
                    "Neural Networks",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Kantorovich Strikes Back! Wasserstein GANs are not Optimal Transport?",
                        "abstract": "Wasserstein Generative Adversarial Networks (WGANs) are the popular generative models built on the theory of Optimal Transport (OT) and the Kantorovich duality. Despite the success of WGANs, it is still unclear how well the underlying OT dual solvers approximate the OT cost (Wasserstein-1 distance, $\\mathbb{W}_{1}$) and the OT gradient needed to update the generator. In this paper, we address these questions. We construct 1-Lipschitz functions and use them to build ray monotone transport plans. This strategy yields pairs of continuous benchmark distributions with the analytically known OT plan, OT cost and OT gradient in high-dimensional spaces such as spaces of images. We thoroughly evaluate popular WGAN dual form solvers (gradient penalty, spectral normalization, entropic regularization, etc.) using these benchmark pairs. Even though these solvers perform well in WGANs, none of them faithfully compute $\\mathbb{W}_{1}$ in high dimensions. Nevertheless, many provide a meaningful approximation of the OT gradient. These observations suggest that these solvers should not be treated as good estimators of $\\mathbb{W}_{1}$, but to some extent they indeed can be used in variational problems requiring the minimization of $\\mathbb{W}_{1}$."
                    },
                    {
                        "title": "Continuous Wasserstein-2 Barycenter Estimation without Minimax Optimization",
                        "abstract": "Wasserstein barycenters provide a geometric notion of the weighted average of probability measures based on optimal transport. In this paper, we present a scalable algorithm to compute Wasserstein-2 barycenters given sample access to the input measures, which are not restricted to being discrete. While past approaches rely on entropic or quadratic regularization, we employ input convex neural networks and cycle-consistency regularization to avoid introducing bias. As a result, our approach does not resort to minimax optimization. We provide theoretical analysis on error bounds as well as empirical evidence of the effectiveness of the proposed approach in low-dimensional qualitative scenarios and high-dimensional quantitative experiments."
                    },
                    {
                        "title": "Neural Optimal Transport",
                        "abstract": "We present a novel neural-networks-based algorithm to compute optimal transport maps and plans for strong and weak transport costs. To justify the usage of neural networks, we prove that they are universal approximators of transport plans between probability distributions. We evaluate the performance of our optimal transport algorithm on toy examples and on the unpaired image-to-image translation."
                    },
                    {
                        "title": "Kernel Neural Optimal Transport",
                        "abstract": "We study the Neural Optimal Transport (NOT) algorithm which uses the general optimal transport formulation and learns stochastic transport plans. We show that NOT with the weak quadratic cost might learn fake plans which are not optimal. To resolve this issue, we introduce kernel weak quadratic costs. We show that they provide improved theoretical guarantees and practical performance. We test NOT with kernel costs on the unpaired image-to-image translation task."
                    },
                    {
                        "title": "Wasserstein-2 Generative Networks",
                        "abstract": "We propose a novel end-to-end non-minimax algorithm for training optimal transport mappings for the quadratic cost (Wasserstein-2 distance). The algorithm uses input convex neural networks and a cycle-consistency regularization to approximate Wasserstein-2 distance. In contrast to popular entropic and quadratic regularizers, cycle-consistency does not introduce bias and scales well to high dimensions. From the theoretical side, we estimate the properties of the generative mapping fitted by our algorithm. From the practical side, we evaluate our algorithm on a wide range of tasks: image-to-image color transfer, latent space optimal transport, image-to-image style transfer, and domain adaptation."
                    },
                    {
                        "title": "Do Neural Optimal Transport Solvers Work? A Continuous Wasserstein-2 Benchmark",
                        "abstract": "Despite the recent popularity of neural network-based solvers for optimal transport (OT), there is no standard quantitative way to evaluate their performance. In this paper, we address this issue for quadratic-cost transport -- specifically, computation of the Wasserstein-2 distance, a commonly-used formulation of optimal transport in machine learning. To overcome the challenge of computing ground truth transport maps between continuous measures needed to assess these solvers, we use input-convex neural networks (ICNN) to construct pairs of measures whose ground truth OT maps can be obtained analytically. This strategy yields pairs of continuous benchmark measures in high-dimensional spaces such as spaces of images. We thoroughly evaluate existing optimal transport solvers using these benchmark measures. Even though these solvers perform well in downstream tasks, many do not faithfully recover optimal transport maps. To investigate the cause of this discrepancy, we further test the solvers in a setting of image generation. Our study reveals crucial limitations of existing solvers and shows that increased OT accuracy does not necessarily correlate to better results downstream."
                    },
                    {
                        "title": "Integral Mixability: a Tool for Efficient Online Aggregation of Functional and Probabilistic Forecasts",
                        "abstract": "In this paper we extend the setting of the online prediction with expert advice to function-valued forecasts. At each step of the online game several experts predict a function, and the learner has to efficiently aggregate these functional forecasts into a single forecast. We adapt basic mixable (and exponentially concave) loss functions to compare functional predictions and prove that these adaptations are also mixable (exp-concave). We call this phenomena integral mixability (exp-concavity). As an application of our main result, we prove that various loss functions used for probabilistic forecasting are mixable (exp-concave). The considered losses include Sliced Continuous Ranking Probability Score, Energy-Based Distance, Optimal Transport Costs & Sliced Wasserstein-2 distance, Beta-2 & Kullback-Leibler divergences, Characteristic function and Maximum Mean Discrepancies."
                    },
                    {
                        "title": "Adaptive Hedging under Delayed Feedback",
                        "abstract": "The article is devoted to investigating the application of hedging strategies to online expert weight allocation under delayed feedback. As the main result, we develop the General Hedging algorithm $\\mathcal{G}$ based on the exponential reweighing of experts' losses. We build the artificial probabilistic framework and use it to prove the adversarial loss bounds for the algorithm $\\mathcal{G}$ in the delayed feedback setting. The designed algorithm $\\mathcal{G}$ can be applied to both countable and continuous sets of experts. We also show how algorithm $\\mathcal{G}$ extends classical Hedge (Multiplicative Weights) and adaptive Fixed Share algorithms to the delayed feedback and derive their regret bounds for the delayed setting by using our main result."
                    },
                    {
                        "title": "Wasserstein Iterative Networks for Barycenter Estimation",
                        "abstract": "Wasserstein barycenters have become popular due to their ability to represent the average of probability measures in a geometrically meaningful way. In this paper, we present an algorithm to approximate the Wasserstein-2 barycenters of continuous measures via a generative model. Previous approaches rely on regularization (entropic/quadratic) which introduces bias or on input convex neural networks which are not expressive enough for large-scale tasks. In contrast, our algorithm does not introduce bias and allows using arbitrary neural networks. In addition, based on the celebrity faces dataset, we construct Ave, celeba! dataset which can be used for quantitative evaluation of barycenter algorithms by using standard metrics of generative models such as FID."
                    },
                    {
                        "title": "Long-Term Online Smoothing Prediction Using Expert Advice",
                        "abstract": "For the prediction with experts' advice setting, we construct forecasting algorithms that suffer loss not much more than any expert in the pool. In contrast to the standard approach, we investigate the case of long-term forecasting of time series and consider two scenarios. In the first one, at each step $t$ the learner has to combine the point forecasts of the experts issued for the time interval $[t+1, t+d]$ ahead. Our approach implies that at each time step experts issue point forecasts for arbitrary many steps ahead and then the learner (algorithm) combines these forecasts and the forecasts made earlier into one vector forecast for steps $[t+1,t+d]$. By combining past and the current long-term forecasts we obtain a smoothing mechanism that protects our algorithm from temporary trend changes, noise and outliers. In the second scenario, at each step $t$ experts issue a prediction function, and the learner has to combine these functions into the single one, which will be used for long-term time-series prediction. For each scenario, we develop an algorithm for combining experts forecasts and prove $O(\\ln T)$ adversarial regret upper bound for both algorithms."
                    },
                    {
                        "title": "Connecting adversarial attacks and optimal transport for domain adaptation",
                        "abstract": "We present a novel algorithm for domain adaptation using optimal transport. In domain adaptation, the goal is to adapt a classifier trained on the source domain samples to the target domain. In our method, we use optimal transport to map target samples to the domain named source fiction. This domain differs from the source but is accurately classified by the source domain classifier. Our main idea is to generate a source fiction by c-cyclically monotone transformation over the target domain. If samples with the same labels in two domains are c-cyclically monotone, the optimal transport map between these domains preserves the class-wise structure, which is the main goal of domain adaptation. To generate a source fiction domain, we propose an algorithm that is based on our finding that adversarial attacks are a c-cyclically monotone transformation of the dataset. We conduct experiments on Digits and Modern Office-31 datasets and achieve improvement in performance for simple discrete optimal transport solvers for all adaptation tasks."
                    },
                    {
                        "title": "Entropic Neural Optimal Transport via Diffusion Processes",
                        "abstract": "We propose a novel neural algorithm for the fundamental problem of computing the entropic optimal transport (EOT) plan between continuous probability distributions which are accessible by samples. Our algorithm is based on the saddle point reformulation of the dynamic version of EOT which is known as the Schr\\\"odinger Bridge problem. In contrast to the prior methods for large-scale EOT, our algorithm is end-to-end and consists of a single learning step, has fast inference procedure, and allows handling small values of the entropy regularization coefficient which is of particular importance in some applied problems. Empirically, we show the performance of the method on several large-scale EOT tasks. https://github.com/ngushchin/EntropicNeuralOptimalTransport"
                    },
                    {
                        "title": "Large-Scale Wasserstein Gradient Flows",
                        "abstract": "Wasserstein gradient flows provide a powerful means of understanding and solving many diffusion equations. Specifically, Fokker-Planck equations, which model the diffusion of probability measures, can be understood as gradient descent over entropy functionals in Wasserstein space. This equivalence, introduced by Jordan, Kinderlehrer and Otto, inspired the so-called JKO scheme to approximate these diffusion processes via an implicit discretization of the gradient flow in Wasserstein space. Solving the optimization problem associated to each JKO step, however, presents serious computational challenges. We introduce a scalable method to approximate Wasserstein gradient flows, targeted to machine learning applications. Our approach relies on input-convex neural networks (ICNNs) to discretize the JKO steps, which can be optimized by stochastic gradient descent. Unlike previous work, our method does not require domain discretization or particle simulation. As a result, we can sample from the measure at each time step of the diffusion and compute its probability density. We demonstrate our algorithm's performance by computing diffusions following the Fokker-Planck equation and apply it to unnormalized density sampling as well as nonlinear filtering."
                    },
                    {
                        "title": "An Optimal Transport Perspective on Unpaired Image Super-Resolution",
                        "abstract": "Real-world image super-resolution (SR) tasks often do not have paired datasets, which limits the application of supervised techniques. As a result, the tasks are usually approached by unpaired techniques based on Generative Adversarial Networks (GANs), which yield complex training losses with several regularization terms, e.g., content or identity losses. We theoretically investigate optimization problems which arise in such models and find two surprizing observations. First, the learned SR map is always an optimal transport (OT) map. Second, we theoretically prove and empirically show that the learned map is biased, i.e., it does not actually transform the distribution of low-resolution images to high-resolution ones. Inspired by these findings, we propose an algorithm for unpaired SR which learns an unbiased OT map for the perceptual transport cost. Unlike the existing GAN-based alternatives, our algorithm has a simple optimization objective reducing the need for complex hyperparameter selection and an application of additional regularizations. At the same time, it provides a nearly state-of-the-art performance on the large-scale unpaired AIM19 dataset."
                    },
                    {
                        "title": "Estimating Barycenters of Distributions with Neural Optimal Transport",
                        "abstract": "Given a collection of probability measures, a practitioner sometimes needs to find an \"average\" distribution which adequately aggregates reference distributions. A theoretically appealing notion of such an average is the Wasserstein barycenter, which is the primal focus of our work. By building upon the dual formulation of Optimal Transport (OT), we propose a new scalable approach for solving the Wasserstein barycenter problem. Our methodology is based on the recent Neural OT solver: it has bi-level adversarial learning objective and works for general cost functions. These are key advantages of our method since the typical adversarial algorithms leveraging barycenter tasks utilize tri-level optimization and focus mostly on quadratic cost. We also establish theoretical error bounds for our proposed approach and showcase its applicability and effectiveness in illustrative scenarios and image data setups. Our source code is available at https://github.com/justkolesov/NOTBarycenters."
                    },
                    {
                        "title": "Robust Barycenter Estimation using Semi-Unbalanced Neural Optimal Transport",
                        "abstract": "A common challenge in aggregating data from multiple sources can be formalized as an \\textit{Optimal Transport} (OT) barycenter problem, which seeks to compute the average of probability distributions with respect to OT discrepancies. However, the presence of outliers and noise in the data measures can significantly hinder the performance of traditional statistical methods for estimating OT barycenters. To address this issue, we propose a novel, scalable approach for estimating the \\textit{robust} continuous barycenter, leveraging the dual formulation of the \\textit{(semi-)unbalanced} OT problem. To the best of our knowledge, this paper is the first attempt to develop an algorithm for robust barycenters under the continuous distribution setup. Our method is framed as a $\\min$-$\\max$ optimization problem and is adaptable to \\textit{general} cost function. We rigorously establish the theoretical underpinnings of the proposed method and demonstrate its robustness to outliers and class imbalance through a number of illustrative experiments."
                    },
                    {
                        "title": "Rethinking Optimal Transport in Offline Reinforcement Learning",
                        "abstract": "We propose a novel algorithm for offline reinforcement learning using optimal transport. Typically, in offline reinforcement learning, the data is provided by various experts and some of them can be sub-optimal. To extract an efficient policy, it is necessary to \\emph{stitch} the best behaviors from the dataset. To address this problem, we rethink offline reinforcement learning as an optimal transportation problem. And based on this, we present an algorithm that aims to find a policy that maps states to a \\emph{partial} distribution of the best expert actions for each given state. We evaluate the performance of our algorithm on continuous control problems from the D4RL suite and demonstrate improvements over existing methods."
                    },
                    {
                        "title": "Aggregating Strategies for Long-term Forecasting",
                        "abstract": "The article is devoted to investigating the application of aggregating algorithms to the problem of the long-term forecasting. We examine the classic aggregating algorithms based on the exponential reweighing. For the general Vovk's aggregating algorithm we provide its generalization for the long-term forecasting. For the special basic case of Vovk's algorithm we provide its two modifications for the long-term forecasting. The first one is theoretically close to an optimal algorithm and is based on replication of independent copies. It provides the time-independent regret bound with respect to the best expert in the pool. The second one is not optimal but is more practical and has $O(\\sqrt{T})$ regret bound, where $T$ is the length of the game."
                    },
                    {
                        "title": "Generative Modeling with Optimal Transport Maps",
                        "abstract": "With the discovery of Wasserstein GANs, Optimal Transport (OT) has become a powerful tool for large-scale generative modeling tasks. In these tasks, OT cost is typically used as the loss for training GANs. In contrast to this approach, we show that the OT map itself can be used as a generative model, providing comparable performance. Previous analogous approaches consider OT maps as generative models only in the latent spaces due to their poor performance in the original high-dimensional ambient space. In contrast, we apply OT maps directly in the ambient space, e.g., a space of high-dimensional images. First, we derive a min-max optimization algorithm to efficiently compute OT maps for the quadratic cost (Wasserstein-2 distance). Next, we extend the approach to the case when the input and output distributions are located in the spaces of different dimensions and derive error bounds for the computed OT map. We evaluate the algorithm on image generation and unpaired image restoration tasks. In particular, we consider denoising, colorization, and inpainting, where the optimality of the restoration map is a desired attribute, since the output (restored) image is expected to be close to the input (degraded) one."
                    }
                ]
            }
        }
    },
    "2401.09742": {
        "paper_data": {
            "title": "Image Translation as Diffusion Visual Programmers",
            "url": "http://arxiv.org/abs/2401.09742v2",
            "arxiv_id": "2401.09742",
            "authors": [
                "Cheng Han",
                "James C. Liang",
                "Qifan Wang",
                "Majid Rabbani",
                "Sohail Dianat",
                "Raghuveer Rao",
                "Ying Nian Wu",
                "Dongfang Liu"
            ],
            "abstract": "We introduce the novel Diffusion Visual Programmer (DVP), a neuro-symbolic image translation framework. Our proposed DVP seamlessly embeds a condition-flexible diffusion model within the GPT architecture, orchestrating a coherent sequence of visual programs (i.e., computer vision models) for various pro-symbolic steps, which span RoI identification, style transfer, and position manipulation, facilitating transparent and controllable image translation processes. Extensive experiments demonstrate DVP's remarkable performance, surpassing concurrent arts. This success can be attributed to several key features of DVP: First, DVP achieves condition-flexible translation via instance normalization, enabling the model to eliminate sensitivity caused by the manual guidance and optimally focus on textual descriptions for high-quality content generation. Second, the framework enhances in-context reasoning by deciphering intricate high-dimensional concepts in feature spaces into more accessible low-dimensional symbols (e.g., [Prompt], [RoI object]), allowing for localized, context-free editing while maintaining overall coherence. Last but not least, DVP improves systemic controllability and explainability by offering explicit symbolic representations at each programming stage, empowering users to intuitively interpret and modify results. Our research marks a substantial step towards harmonizing artificial image translation processes with cognitive intelligence, promising broader applications.",
            "introduction": "   1 Introduction  Figure 1:  Working pipeline showcase. DVP represents a solution rooted in visual programming, demonstrating pronounced capabilities in-context reasoning and explainable control, in addition to its remarkable efficacy in style transfer.    Concurrent state-of-the-art image translation methods predominantly follow the paradigm of connectionism\u00a0(Garcez et\u00a0al., 2022), setting their goal as answering \u201cwhat\u201d \u2014 they aim to translate the image from the source domain to the designated target domain with fidelity. These methods can be broadly categorized into Generative Adversarial Network (GAN)-based and Diffusion-based methods (see \u00a72). Diffusion-based approaches usually demonstrate superior capabilities when compared to their GAN-based counterparts, especially in the domains such as image quality and coherence, training stability, and fine-grained control\u00a0(Dhariwal & Nichol, 2021). As such, they are perceived to harbor more auspicious potential in image translation\u00a0(Mokady et\u00a0al., 2023).   Despite their huge success\u00a0(Epstein et\u00a0al., 2023; Dhariwal & Nichol, 2021), these diffusion-based methods exhibit a series of limitations: \u2460 Condition-rigid learning. Existing arts following the principles in classifier-free guidance\u00a0(Ho & Salimans, 2021; Mokady et\u00a0al., 2023) face a significant challenge in effectively achieving a harmonious equilibrium between unconditional and conditional predictions. Typically, these methods rely on manually crafted guidance scale parameters to supervise the process of each individual image translation. This inherent restriction curtails the algorithmic scalability, thereby impeding their potential for achieving holistic automation for real-world applications; \u2461 Context-free incompetence. Concurrent methods primarily engage in the global manipulation of diverse attributes (e.g., stylistic and content-related elements) within images, which prioritize the maintenance of contextual integrity over local alterations. Nevertheless, the absence of context-free reasoning hinders the precision required to realize specific RoI modifications, while simultaneously upholding a broader sense coherence. Such contextual understanding capacities necessitate a high degree of semantic acumen and a robust mastery of image structure, which is sorely lacking within contemporary image translation solutions\u00a0(Hitzler et\u00a0al., 2022; Oltramari et\u00a0al., 2020; Mao et\u00a0al., 2019). \u2462 System opacity. Due to their black box nature, diffusion-based methods \u2014 coined by connectionism\u00a0(Wang et\u00a0al., 2023) \u2014 often exhibit a level of abstraction that distances them from the intrinsic physical characteristics of the problem they aim to model\u00a0(Angelov & Soares, 2020). As a result, users encounter limited control over the model\u2019s behaviors before reaching the final outputs, rendering them rudderless in establishing trustworthiness in decision-making or pursuing systemic enhancement.   In this work, we revolutionize the focus of \u201cwhat\u201d into a more flexible image translation paradigm of \u201cwhere-then-what.\u201d Under this paradigm, we first answer the question of \u201cwhere\u201d by finding the instructed target region accurately. After getting the target region (i.e., Region of Interest (RoI)), we answers the question of \u201cwhat\u201d by translating it into the targeted domain with high-fidelity. Considering the aforementioned discussions, we are approaching the task of image translation from a fresh neuro-symbolic perspective, presenting a novel method named the Diffusion Visual Programmer (DVP). More concretely, we architect an condition-flexible diffusion model\u00a0(Song et\u00a0al., 2021), harmoniously integrated with the foundation of GPT\u00a0(Brown et\u00a0al., 2020), which orchestrates a concatenation of off-the-shelf computer vision models to yield the coveted outcomes. For example in Fig.\u00a01, given an input instruction that specifies the translation from the \u201cleft dog\u201d to a \u201csheep,\u201d DVP utilizes GPT as an AI agent to plan a sequence of programs with operations, subsequently invoked in the following procedure. It first addresses the",
            "references": [
                {
                    "title": "Facing the Elephant in the Room: Visual Prompt Tuning or Full Finetuning?",
                    "abstract": "As the scale of vision models continues to grow, the emergence of Visual Prompt Tuning (VPT) as a parameter-efficient transfer learning technique has gained attention due to its superior performance compared to traditional full-finetuning. However, the conditions favoring VPT (the ``when\") and the underlying rationale (the ``why\") remain unclear. In this paper, we conduct a comprehensive analysis across 19 distinct datasets and tasks. To understand the ``when\"aspect, we identify the scenarios where VPT proves favorable by two dimensions: task objectives and data distributions. We find that VPT is preferrable when there is 1) a substantial disparity between the original and the downstream task objectives (e.g., transitioning from classification to counting), or 2) a similarity in data distributions between the two tasks (e.g., both involve natural images). In exploring the ``why\"dimension, our results indicate VPT's success cannot be attributed solely to overfitting and optimization considerations. The unique way VPT preserves original features and adds parameters appears to be a pivotal factor. Our study provides insights into VPT's mechanisms, and offers guidance for its optimal utilization."
                },
                {
                    "title": "Prompt Learns Prompt: Exploring Knowledge-Aware Generative Prompt Collaboration For Video Captioning",
                    "abstract": "Fine-tuning large vision-language models is a challenging task. Prompt tuning approaches have been introduced to learn fixed textual or visual prompts while freezing the pre-trained model in downstream tasks. Despite the effectiveness of prompt tuning, what do those learnable prompts learn remains unexplained. In this work, we explore whether prompts in the fine-tuning can learn knowledge-aware prompts from the pre-training, by designing two different sets of prompts in pre-training and fine-tuning phases respectively. Specifically, we present a Video-Language Prompt tuning (VL-Prompt) approach for video captioning, which first efficiently pre-train a video-language model to extract key information (e.g., actions and objects) with flexibly generated Knowledge-Aware Prompt (KAP). Then, we design a Video-Language Prompt (VLP) to transfer the knowledge from the knowledge-aware prompts and fine-tune the model to generate full captions. Experimental results show the superior performance of our approach over several state-of-the-art baselines. We further demonstrate that the video-language prompts are well learned from the knowledge-aware prompts."
                },
                {
                    "title": "E2VPT: An Effective and Efficient Approach for Visual Prompt Tuning",
                    "abstract": "As the size of transformer-based, models continues to grow, fine-tuning these large-scale pretrained vision models for new tasks has become increasingly parameter-intensive. Parameter-efficient learning has been developed to reduce the number of tunable parameters during fine-tuning. Although these methods show promising results, there is still a significant performance gap compared to full fine-tuning. To address this challenge, we propose an Effective and Efficient Visual Prompt Tuning (E2VPT) approach for large-scale transformer-based model adaptation. Specifically, we introduce a set of learnable key-value prompts and visual prompts into self-attention and input layers, respectively, to improve the effectiveness of model fine-tuning. Moreover, we design a prompt pruning procedure to systematically prune low importance prompts while preserving model performance, which largely enhances the model\u2019s efficiency. Empirical results demonstrate that our approach outperforms several state-of-the-art baselines on two benchmarks, with considerably low parameter usage (e.g., 0.32% of model parameters on VTAB-1k). Our code is available at https://github.com/ChengHan111/E2VPT."
                },
                {
                    "title": "AnyDoor: Zero-shot Object-level Image Customization",
                    "abstract": "This work presents AnyDoor, a diffusion-based image generator with the power to teleport target objects to new scenes at user-specified locations with desired shapes. Instead of tuning parameters for each object, our model is trained only once and effortlessly generalizes to diverse object-scene combinations at the inference stage. Such a challenging zero-shot setting requires an adequate characterization of a certain object. To this end, we complement the commonly used identity feature with detail features, which are carefully designed to maintain appearance details yet allow versatile local variations (e.g., lighting, orientation, posture, etc.), supporting the object in favorably blending with different surroundings. We further propose to borrow knowledge from video datasets, where we can observe various forms (i.e., along the time axis) of a single object, leading to stronger model generalizability and robustness. Extensive experiments demonstrate the superiority of our approach over existing alternatives as well as its great potential in real-world applications, such as virtual try-on, shape editing, and object swapping. Code is released at github.com/ali-vilab/AnyDoor."
                },
                {
                    "title": "Diffusion Self-Guidance for Controllable Image Generation",
                    "abstract": "Large-scale generative models are capable of producing high-quality images from detailed text descriptions. However, many aspects of an image are difficult or impossible to convey through text. We introduce self-guidance, a method that provides greater control over generated images by guiding the internal representations of diffusion models. We demonstrate that properties such as the shape, location, and appearance of objects can be extracted from these representations and used to steer sampling. Self-guidance works similarly to classifier guidance, but uses signals present in the pretrained model itself, requiring no additional models or training. We show how a simple set of properties can be composed to perform challenging image manipulations, such as modifying the position or size of objects, merging the appearance of objects in one image with the layout of another, composing objects from many images into one, and more. We also show that self-guidance can be used to edit real images. For results and an interactive demo, see our project page at https://dave.ml/selfguidance/"
                },
                {
                    "title": "Coarse-to-Fine Video Instance Segmentation With Factorized Conditional Appearance Flows",
                    "abstract": "We introduce a novel method using a new generative model that automatically learns effective representations of the target and background appearance to detect, segment and track each instance in a video sequence. Differently from current discriminative tracking-by-detection solutions, our proposed hierarchical structural embedding learning can predict more high-quality masks with accurate boundary details over spatio-temporal space via the normalizing flows. We formulate the instance inference procedure as a hierarchical spatio-temporal embedded learning across time and space. Given the video clip, our method first coarsely locates pixels belonging to a particular instance with Gaussian distribution and then builds a novel mixing distribution to promote the instance boundary by fusing hierarchical appearance embedding information in a coarse-to-fine manner. For the mixing distribution, we utilize a factorization condition normalized flow fashion to estimate the distribution parameters to improve the segmentation performance. Comprehensive qualitative, quantitative, and ablation experiments are performed on three representative video instance segmentation benchmarks (i.e., YouTube-VIS19, YouTube-VIS21, and OVIS) and the effectiveness of the proposed method is demonstrated. More impressively, the superior performance of our model on an unsupervised video object segmentation dataset (i.e., DAVIS19) proves its generalizability. Our algorithm implementations are publicly available at https://github.com/zyqin19/HEVis."
                },
                {
                    "title": "Segment Anything",
                    "abstract": "We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive \u2013 often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images at segment-anything.com to foster research into foundation models for computer vision. We recommend reading the full paper at: arxiv.org/abs/2304.02643."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Bi-Noising Diffusion: Towards Conditional Diffusion Models with Generative Restoration Priors",
                    "abstract": "Conditional diffusion probabilistic models can model the distribution of natural images and can generate diverse and realistic samples based on given conditions. However, oftentimes their results can be unrealistic with observable color shifts and textures. We believe that this issue results from the divergence between the probabilistic distribution learned by the model and the distribution of natural images. The delicate conditions gradually enlarge the divergence during each sampling timestep. To address this issue, we introduce a new method that brings the predicted samples to the training data manifold using a pretrained unconditional diffusion model. The unconditional model acts as a regularizer and reduces the divergence introduced by the conditional model at each sampling step. We perform comprehensive experiments to demonstrate the effectiveness of our approach on super-resolution, colorization, turbulence removal, and image-deraining tasks. The improvements obtained by our method suggest that the priors can be incorporated as a general plugin for improving conditional diffusion models."
                },
                {
                    "title": "Plug-and-Play Diffusion Features for Text-Driven Image-to-Image Translation",
                    "abstract": "Large-scale text-to-image generative models have been a revolutionary breakthrough in the evolution of generative AI, synthesizing diverse images with highly complex visual concepts. However, a pivotal challenge in leveraging such models for real-world content creation is providing users with control over the generated content. In this paper, we present a new framework that takes text-to- image synthesis to the realm of image-to-image translation - given a guidance image and a target text prompt as input, our method harnesses the power of a pre-trained text-to-image diffusion model to generate a new image that complies with the target text, while preserving the semantic layout of the guidance image. Specifically, we observe and empirically demonstrate that fine-grained control over the generated structure can be achieved by manipulating spatial features and their self-attention inside the model. This results in a simple and effective approach, where features extracted from the guidance image are directly injected into the generation process of the translated image, requiring no training or fine-tuning. We demonstrate high-quality results on versatile text-guided image translation tasks, including translating sketches, rough drawings and animations into realistic images, changing the class and appearance of objects in a given image, and modifying global qualities such as lighting and color."
                },
                {
                    "title": "Visual Programming: Compositional visual reasoning without training",
                    "abstract": "We present Visprog, a neuro-symbolic approach to solving complex and compositional visual tasks given natural language instructions. Visprog avoids the need for any task-specific training. Instead, it uses the incontext learning ability of large language models to generate python-like modular programs, which are then executed to get both the solution and a comprehensive and interpretable rationale. Each line of the generated program may invoke one of several off-the-shelf computer vision models, image processing subroutines, or python functions to produce intermediate outputs that may be consumed by subsequent parts of the program. We demonstrate the flexibility of VIsPROG on 4 diverse tasks - compositional visual question answering, zero-shot reasoning on image pairs, factual knowledge object tagging, and language-guided image editing. We believe neuro-symbolic approaches like Visprog are an exciting avenue to easily and effectively expand the scope of AI systems to serve the long tail of complex tasks that people may wish to perform."
                },
                {
                    "title": "Null-text Inversion for Editing Real Images using Guided Diffusion Models",
                    "abstract": "Recent large-scale text-guided diffusion models provide powerful image generation capabilities. Currently, a massive effort is given to enable the modification of these images using text only as means to offer intuitive and versatile editing tools. To edit a real image using these state-of-the-art tools, one must first invert the image with a meaningful text prompt into the pretrained model's domain. In this paper, we introduce an accurate inversion technique and thus facilitate an intuitive text-based modification of the image. Our proposed inversion consists of two key novel components: (i) Pivotal inversion for diffusion models. While current methods aim at mapping random noise samples to a single input image, we use a single pivotal noise vector for each timestamp and optimize around it. We demonstrate that a direct DDIM inversion is inadequate on its own, but does provide a rather good anchor for our optimization. (ii) Null-text optimization, where we only modify the unconditional textual embedding that is used for classifier-free guidance, rather than the input text embedding. This allows for keeping both the model weights and the conditional embedding intact and hence enables applying prompt-based editing while avoiding the cumbersome tuning of the model's weights. Our null-text inversion, based on the publicly available Stable Diffusion model, is extensively evaluated on a variety of images and various prompt editing, showing high-fidelity editing of real images."
                },
                {
                    "title": "InstructPix2Pix: Learning to Follow Image Editing Instructions",
                    "abstract": "We propose a method for editing images from human instructions: given an input image and a written instruction that tells the model what to do, our model follows these instructions to edit the image. To obtain training data for this problem, we combine the knowledge of two large pretrained models\u2014a language model (GPT-3) and a text-to-image model (Stable Diffusion)\u2014to generate a large dataset of image editing examples. Our conditional diffusion model, InstructPix2Pix, is trained on our generated data, and generalizes to real images and user-written instructions at inference time. Since it performs edits in the forward pass and does not require per-example fine-tuning or inversion, our model edits images quickly, in a matter of seconds. We show compelling editing results for a diverse collection of input images and written instructions."
                },
                {
                    "title": "Learning Equivariant Segmentation with Instance-Unique Querying",
                    "abstract": "Prevalent state-of-the-art instance segmentation methods fall into a query-based scheme, in which instance masks are derived by querying the image feature using a set of instance-aware embeddings. In this work, we devise a new training framework that boosts query-based models through discriminative query embedding learning. It explores two essential properties, namely dataset-level uniqueness and transformation equivariance, of the relation between queries and instances. First, our algorithm uses the queries to retrieve the corresponding instances from the whole training dataset, instead of only searching within individual scenes. As querying instances across scenes is more challenging, the segmenters are forced to learn more discriminative queries for effective instance separation. Second, our algorithm encourages both image (instance) representations and queries to be equivariant against geometric transformations, leading to more robust, instance-query matching. On top of four famous, query-based models ($i.e.,$ CondInst, SOLOv2, SOTR, and Mask2Former), our training algorithm provides significant performance gains ($e.g.,$ +1.6 - 3.2 AP) on COCO dataset. In addition, our algorithm promotes the performance of SOLOv2 by 2.7 AP, on LVISv1 dataset."
                },
                {
                    "title": "Diffusion-based Image Translation using Disentangled Style and Content Representation",
                    "abstract": "Diffusion-based image translation guided by semantic texts or a single target image has enabled flexible style transfer which is not limited to the specific domains. Unfortunately, due to the stochastic nature of diffusion models, it is often difficult to maintain the original content of the image during the reverse diffusion. To address this, here we present a novel diffusion-based unsupervised image translation method using disentangled style and content representation. Specifically, inspired by the splicing Vision Transformer, we extract intermediate keys of multihead self attention layer from ViT model and used them as the content preservation loss. Then, an image guided style transfer is performed by matching the [CLS] classification token from the denoised samples and target image, whereas additional CLIP loss is used for the text-driven style transfer. To further accelerate the semantic change during the reverse diffusion, we also propose a novel semantic divergence loss and resampling strategy. Our experimental results show that the proposed method outperforms state-of-the-art baseline models in both text-guided and image-guided translation tasks."
                },
                {
                    "title": "Visual Recognition with Deep Nearest Centroids",
                    "abstract": "We devise deep nearest centroids (DNC), a conceptually elegant yet surprisingly effective network for large-scale visual recognition, by revisiting Nearest Centroids, one of the most classic and simple classifiers. Current deep models learn the classifier in a fully parametric manner, ignoring the latent data structure and lacking simplicity and explainability. DNC instead conducts nonparametric, case-based reasoning; it utilizes sub-centroids of training samples to describe class distributions and clearly explains the classification as the proximity of test data and the class sub-centroids in the feature space. Due to the distance-based nature, the network output dimensionality is flexible, and all the learnable parameters are only for data embedding. That means all the knowledge learnt for ImageNet classification can be completely transferred for pixel recognition learning, under the\"pre-training and fine-tuning\"paradigm. Apart from its nested simplicity and intuitive decision-making mechanism, DNC can even possess ad-hoc explainability when the sub-centroids are selected as actual training images that humans can view and inspect. Compared with parametric counterparts, DNC performs better on image classification (CIFAR-10, ImageNet) and greatly boots pixel recognition (ADE20K, Cityscapes), with improved transparency and fewer learnable parameters, using various network architectures (ResNet, Swin) and segmentation models (FCN, DeepLabV3, Swin). We feel this work brings fundamental insights into related fields."
                },
                {
                    "title": "DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation",
                    "abstract": "Large text-to-image models achieved a remarkable leap in the evolution of AI, enabling high-quality and diverse synthesis of images from a given text prompt. However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts. In this work, we present a new approach for \u201cpersonalization\u201d of text-to-image diffusion models. Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject. Once the subject is embedded in the output domain of the model, the unique identifier can be used to synthesize novel photorealistic images of the subject contextualized in different scenes. By leveraging the semantic prior embedded in the model with a new autogenous class-specific prior preservation loss, our technique enables synthesizing the subject in diverse scenes, poses, views and lighting conditions that do not appear in the reference images. We apply our technique to several previously-unassailable tasks, including subject recontextualization, text-guided view synthesis, and artistic rendering, all while preserving the subject's key features. We also provide a new dataset and evaluation protocol for this new task of subject-driven generation. Project page: https://dreambooth.github.io/"
                },
                {
                    "title": "Prompt-to-Prompt Image Editing with Cross Attention Control",
                    "abstract": "Recent large-scale text-driven synthesis models have attracted much attention thanks to their remarkable capabilities of generating highly diverse images that follow given text prompts. Such text-based synthesis methods are particularly appealing to humans who are used to verbally describe their intent. Therefore, it is only natural to extend the text-driven image synthesis to text-driven image editing. Editing is challenging for these generative models, since an innate property of an editing technique is to preserve most of the original image, while in the text-based models, even a small modification of the text prompt often leads to a completely different outcome. State-of-the-art methods mitigate this by requiring the users to provide a spatial mask to localize the edit, hence, ignoring the original structure and content within the masked region. In this paper, we pursue an intuitive prompt-to-prompt editing framework, where the edits are controlled by text only. To this end, we analyze a text-conditioned model in depth and observe that the cross-attention layers are the key to controlling the relation between the spatial layout of the image to each word in the prompt. With this observation, we present several applications which monitor the image synthesis by editing the textual prompt only. This includes localized editing by replacing a word, global editing by adding a specification, and even delicately controlling the extent to which a word is reflected in the image. We present our results over diverse images and prompts, demonstrating high-quality synthesis and fidelity to the edited prompts."
                },
                {
                    "title": "An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion",
                    "abstract": "Text-to-image models offer unprecedented freedom to guide creation through natural language. Yet, it is unclear how such freedom can be exercised to generate images of specific unique concepts, modify their appearance, or compose them in new roles and novel scenes. In other words, we ask: how can we use language-guided models to turn our cat into a painting, or imagine a new product based on our favorite toy? Here we present a simple approach that allows such creative freedom. Using only 3-5 images of a user-provided concept, like an object or a style, we learn to represent it through new\"words\"in the embedding space of a frozen text-to-image model. These\"words\"can be composed into natural language sentences, guiding personalized creation in an intuitive way. Notably, we find evidence that a single word embedding is sufficient for capturing unique and varied concepts. We compare our approach to a wide range of baselines, and demonstrate that it can more faithfully portray the concepts across a range of applications and tasks. Our code, data and new words will be available at: https://textual-inversion.github.io"
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "BBDM: Image-to-Image Translation with Brownian Bridge Diffusion Models",
                    "abstract": "Image-to-image translation is an important and challenging problem in computer vision and image processing. Diffusion models (DM) have shown great potentials for high-quality image synthesis, and have gained competitive performance on the task of image-to-image translation. However, most of the existing diffusion models treat image-to-image translation as conditional generation processes, and suffer heavily from the gap between distinct domains. In this paper, a novel image-to-image translation method based on the Brownian Bridge Diffusion Model (BBDM) is proposed, which models image-to-image translation as a stochastic Brownian Bridge process, and learns the translation between two domains directly through the bidirectional diffusion process rather than a conditional generation process. To the best of our knowledge, it is the first work that proposes Brownian Bridge diffusion process for image-to-image translation. Experimental results on various benchmarks demonstrate that the proposed BBDM model achieves competitive performance through both visual inspection and measurable metrics."
                },
                {
                    "title": "VQGAN-CLIP: Open Domain Image Generation and Editing with Natural Language Guidance",
                    "abstract": "Generating and editing images from open domain text prompts is a challenging task that heretofore has required expensive and specially trained models. We demonstrate a novel methodology for both tasks which is capable of producing images of high visual quality from text prompts of significant semantic complexity without any training by using a multimodal encoder to guide image generations. We demonstrate on a variety of tasks how using CLIP [37] to guide VQGAN [11] produces higher visual quality outputs than prior, less flexible approaches like DALL-E [38], GLIDE [33] and Open-Edit [24], despite not being trained for the tasks presented. Our code is available in a public repository."
                },
                {
                    "title": "A Style-aware Discriminator for Controllable Image Translation",
                    "abstract": "Current image-to-image translations do not control the output domain beyond the classes used during training, nor do they interpolate between different domains well, leading to implausible results. This limitation largely arises because labels do not consider the semantic distance. To mitigate such problems, we propose a style-aware discriminator that acts as a critic as well as a style encoder to provide conditions. The style-aware discriminator learns a controllable style space using prototype-based self-supervised learning and simultaneously guides the generator. Experiments on multiple datasets verify that the proposed model outperforms current state-of-the-art image-to-image translation methods. In contrast with current methods, the proposed approach supports various applications, including style interpolation, content transplantation, and local image translation. The code is available at github.com/kunheek/style-aware-discriminator."
                },
                {
                    "title": "Neuro-symbolic approaches in artificial intelligence",
                    "abstract": "artificial"
                },
                {
                    "title": "BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation",
                    "abstract": "Vision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has been largely achieved by scaling up the dataset with noisy image-text pairs collected from the web, which is a suboptimal source of supervision. In this paper, we propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks. BLIP effectively utilizes the noisy web data by bootstrapping the captions, where a captioner generates synthetic captions and a filter removes the noisy ones. We achieve state-of-the-art results on a wide range of vision-language tasks, such as image-text retrieval (+2.7% in average recall@1), image captioning (+2.8% in CIDEr), and VQA (+1.6% in VQA score). BLIP also demonstrates strong generalization ability when directly transferred to video-language tasks in a zero-shot manner. Code, models, and datasets are released at https://github.com/salesforce/BLIP."
                },
                {
                    "title": "RePaint: Inpainting using Denoising Diffusion Probabilistic Models",
                    "abstract": "Free-form inpainting is the task of adding new content to an image in the regions specified by an arbitrary binary mask. Most existing approaches train for a certain distribution of masks, which limits their generalization capabilities to unseen mask types. Furthermore, training with pixel-wise and perceptual losses often leads to simple textural extensions towards the missing areas instead of semantically meaningful generation. In this work, we propose RePaint: A Denoising Diffusion Probabilistic Model (DDPM) based inpainting approach that is applicable to even extreme masks. We employ a pretrained unconditional DDPM as the generative prior. To condition the generation process, we only alter the reverse diffusion iterations by sampling the unmasked regions using the given image infor-mation. Since this technique does not modify or condition the original DDPM network itself, the model produces high-quality and diverse output images for any inpainting form. We validate our method for both faces and general-purpose image inpainting using standard and extreme masks. Re-Paint outperforms state-of-the-art Autoregressive, and GAN approaches for at least five out of six mask distributions. Github Repository: git.io/RePaint"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Masked-attention Mask Transformer for Universal Image Segmentation",
                    "abstract": "Image segmentation groups pixels with different semantics, e.g., category or instance membership. Each choice of semantics defines a task. While only the semantics of each task differ, current research focuses on designing spe-cialized architectures for each task. We present Masked- attention Mask Transformer (Mask2Former), a new archi-tecture capable of addressing any image segmentation task (panoptic, instance or semantic). Its key components in-clude masked attention, which extracts localized features by constraining cross-attention within predicted mask regions. In addition to reducing the research effort by at least three times, it outperforms the best specialized architectures by a significant margin on four popular datasets. Most no-tably, Mask2Former sets a new state-of-the-art for panoptic segmentation (57.8 PQ on COCO), instance segmentation (50.1 AP on COCO) and semantic segmentation (57.7 mIoU onADE20K)."
                },
                {
                    "title": "Conditional Image Generation with Score-Based Diffusion Models",
                    "abstract": "Score-based diffusion models have emerged as one of the most promising frameworks for deep generative modelling. In this work we conduct a systematic comparison and theoretical analysis of different approaches to learning conditional probability distributions with score-based diffusion models. In particular, we prove results which provide a theoretical justification for one of the most successful estimators of the conditional score. Moreover, we introduce a multi-speed diffusion framework, which leads to a new estimator for the conditional score, performing on par with previous state-of-the-art approaches. Our theoretical and experimental findings are accompanied by an open source library MSDiff which allows for application and further research of multi-speed diffusion models."
                },
                {
                    "title": "Palette: Image-to-Image Diffusion Models",
                    "abstract": "This paper develops a unified framework for image-to-image translation based on conditional diffusion models and evaluates this framework on four challenging image-to-image translation tasks, namely colorization, inpainting, uncropping, and JPEG restoration. Our simple implementation of image-to-image diffusion models outperforms strong GAN and regression baselines on all tasks, without task-specific hyper-parameter tuning, architecture customization, or any auxiliary loss or sophisticated new techniques needed. We uncover the impact of an L2 vs. L1 loss in the denoising diffusion objective on sample diversity, and demonstrate the importance of self-attention in the neural architecture through empirical studies. Importantly, we advocate a unified evaluation protocol based on ImageNet, with human evaluation and sample quality scores (FID, Inception Score, Classification Accuracy of a pre-trained ResNet-50, and Perceptual Distance against original images). We expect this standardized evaluation protocol to play a role in advancing image-to-image translation research. Finally, we show that a generalist, multi-task diffusion model performs as well or better than task-specific specialist counterparts. Check out https://diffusion-palette.github.io/ for an overview of the results and code."
                },
                {
                    "title": "xGQA: Cross-Lingual Visual Question Answering",
                    "abstract": "Recent advances in multimodal vision and language modeling have predominantly focused on the English language, mostly due to the lack of multilingual multimodal datasets to steer modeling efforts. In this work, we address this gap and provide xGQA, a new multilingual evaluation benchmark for the visual question answering task. We extend the established English GQA dataset to 7 typologically diverse languages, enabling us to detect and explore crucial challenges in cross-lingual visual question answering. We further propose new adapter-based approaches to adapt multimodal transformer-based models to become multilingual, and\u2014vice versa\u2014multilingual models to become multimodal. Our proposed methods outperform current state-of-the-art multilingual multimodal models (e.g., M3P) in zero-shot cross-lingual settings, but the accuracy remains low across the board; a performance drop of around 38 accuracy points in target languages showcases the difficulty of zero-shot cross-lingual transfer for this task. Our results suggest that simple cross-lingual transfer of multimodal models yields latent multilingual multimodal misalignment, calling for more sophisticated methods for vision and multilingual language modeling."
                },
                {
                    "title": "ILVR: Conditioning Method for Denoising Diffusion Probabilistic Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPM) have shown remarkable performance in unconditional image generation. However, due to the stochasticity of the generative process in DDPM, it is challenging to generate images with the desired semantics. In this work, we propose Iterative Latent Variable Refinement (ILVR), a method to guide the generative process in DDPM to generate high-quality images based on a given reference image. Here, the refinement of the generative process in DDPM enables a single DDPM to sample images from various sets directed by the reference image. The proposed ILVR method generates high-quality images while controlling the generation. The controllability of our method allows adaptation of a single DDPM without any additional learning in various image generation tasks, such as generation from various downsampling factors, multi-domain image translation, paint-to-image, and editing with scribbles."
                },
                {
                    "title": "SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations",
                    "abstract": "Guided image synthesis enables everyday users to create and edit photo-realistic images with minimum effort. The key challenge is balancing faithfulness to the user input (e.g., hand-drawn colored strokes) and realism of the synthesized image. Existing GAN-based methods attempt to achieve such balance using either conditional GANs or GAN inversions, which are challenging and often require additional training data or loss functions for individual applications. To address these issues, we introduce a new image synthesis and editing method, Stochastic Differential Editing (SDEdit), based on a diffusion model generative prior, which synthesizes realistic images by iteratively denoising through a stochastic differential equation (SDE). Given an input image with user guide of any type, SDEdit first adds noise to the input, then subsequently denoises the resulting image through the SDE prior to increase its realism. SDEdit does not require task-specific training or inversions and can naturally achieve the balance between realism and faithfulness. SDEdit significantly outperforms state-of-the-art GAN-based methods by up to 98.09% on realism and 91.72% on overall satisfaction scores, according to a human perception study, on multiple tasks, including stroke-based image synthesis and editing as well as image compositing."
                },
                {
                    "title": "Diffusion Models Beat GANs on Image Synthesis",
                    "abstract": "We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\\times$128, 4.59 on ImageNet 256$\\times$256, and 7.72 on ImageNet 512$\\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\\times$256 and 3.85 on ImageNet 512$\\times$512. We release our code at https://github.com/openai/guided-diffusion"
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "CLIPScore: A Reference-free Evaluation Metric for Image Captioning",
                    "abstract": "Image captioning has conventionally relied on reference-based automatic evaluations, where machine captions are compared against captions written by humans. This is in contrast to the reference-free manner in which humans assess caption quality. In this paper, we report the surprising empirical finding that CLIP (Radford et al., 2021), a cross-modal model pretrained on 400M image+caption pairs from the web, can be used for robust automatic evaluation of image captioning without the need for references. Experiments spanning several corpora demonstrate that our new reference-free metric, CLIPScore, achieves the highest correlation with human judgements, outperforming existing reference-based metrics like CIDEr and SPICE. Information gain experiments demonstrate that CLIPScore, with its tight focus on image-text compatibility, is complementary to existing reference-based metrics that emphasize text-text similarities. Thus, we also present a reference-augmented version, RefCLIPScore, which achieves even higher correlation. Beyond literal description tasks, several case studies reveal domains where CLIPScore performs well (clip-art images, alt-text rating), but also where it is relatively weaker in comparison to reference-based metrics, e.g., news captions that require richer contextual knowledge."
                },
                {
                    "title": "SG-Net: Spatial Granularity Network for One-Stage Video Instance Segmentation",
                    "abstract": "Video instance segmentation (VIS) is a new and critical task in computer vision. To date, top-performing VIS methods extend the two-stage Mask R-CNN by adding a tracking branch, leaving plenty of room for improvement. In contrast, we approach the VIS task from a new perspective and propose a one-stage spatial granularity network (SG-Net). Compared to the conventional two-stage methods, SG-Net demonstrates four advantages: 1) Our method has a one-stage compact architecture and each task head (detection, segmentation, and tracking) is crafted interdependently so they can effectively share features and enjoy the joint optimization; 2) Our mask prediction is dynamically performed on the sub-regions of each detected instance, leading to high-quality masks of fine granularity; 3) Each of our task predictions avoids using expensive proposal-based RoI features, resulting in much reduced runtime complexity per instance; 4) Our tracking head models objects\u2019 centerness movements for tracking, which effectively enhances the tracking robustness to different object appearances. In evaluation, we present state-of-the-art comparisons on the YouTube-VIS dataset. Extensive experiments demonstrate that our compact one-stage method can achieve improved performance in both accuracy and inference speed. We hope our SG-Net could serve as a strong and flexible base-line for the VIS task. Our code will be available here 1."
                },
                {
                    "title": "Cascaded Parsing of Human-Object Interaction Recognition",
                    "abstract": "This paper addresses the task of detecting and recognizing human-object interactions (HOI) in images. Considering the intrinsic complexity and structural nature of the task, we introduce a cascaded parsing network (CP-HOI) for a multi-stage, structured HOI understanding. At each cascade stage, an instance detection module progressively refines HOI proposals and feeds them into a structured interaction reasoning module. Each of the two modules is also connected to its predecessor in the previous stage, enabling efficient cross-stage information propagation. The structured interaction reasoning module is built upon a graph parsing neural network (GPNN), which efficiently models potential HOI structures as graphs and mines rich context for comprehensive relation understanding. In particular, GPNN infers a parse graph that i) interprets meaningful HOI structures by a learnable adjacency matrix, and ii) predicts action (edge) labels. Within an end-to-end, message-passing framework, GPNN blends learning and inference, iteratively parsing HOI structures and reasoning HOI representations (i.e., instance and relation features). Further beyond relation detection at a bounding-box level, we make our framework flexible to perform fine-grained pixel-wise relation segmentation; this provides a new glimpse into better relation modeling. A preliminary version of our CP-HOI model reached 1st place in the ICCV2019 Person in Context Challenge, on both relation detection and segmentation. In addition, our CP-HOI shows promising results on two popular HOI recognition benchmarks, i.e., V-COCO and HICO-DET."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "Swapping Autoencoder for Deep Image Manipulation",
                    "abstract": "Deep generative models have become increasingly effective at producing realistic images from randomly sampled seeds, but using such models for controllable manipulation of existing images remains challenging. We propose the Swapping Autoencoder, a deep model designed specifically for image manipulation, rather than random sampling. The key idea is to encode an image with two independent components and enforce that any swapped combination maps to a realistic image. In particular, we encourage the components to represent structure and texture, by enforcing one component to encode co-occurrent patch statistics across different parts of an image. As our method is trained with an encoder, finding the latent codes for a new input image becomes trivial, rather than cumbersome. As a result, it can be used to manipulate real input images in various ways, including texture swapping, local and global editing, and latent code vector arithmetic. Experiments on multiple datasets show that our model produces better results and is substantially more efficient compared to recent generative models."
                },
                {
                    "title": "The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization",
                    "abstract": "We introduce four new real-world distribution shift datasets consisting of changes in image style, image blurriness, geographic location, camera operation, and more. With our new datasets, we take stock of previously proposed methods for improving out-of-distribution robustness and put them to the test. We find that using larger models and artificial data augmentations can improve robustness on real-world distribution shifts, contrary to claims in prior work. We find improvements in artificial robustness benchmarks can transfer to real-world distribution shifts, contrary to claims in prior work. Motivated by our observation that data augmentations can help with real-world distribution shifts, we also introduce a new data augmentation method which advances the state-of-the-art and outperforms models pre-trained with 1000\u00d7 more labeled data. Overall we find that some methods consistently help with distribution shifts in texture and local image statistics, but these methods do not help with some other distribution shifts like geographic changes. Our results show that future research must study multiple distribution shifts simultaneously, as we demonstrate that no evaluated method consistently improves robustness."
                },
                {
                    "title": "Conversational Neuro-Symbolic Commonsense Reasoning",
                    "abstract": "In order for conversational AI systems to hold more natural and broad-ranging conversations, they will require much more commonsense, including the ability to identify unstated presumptions of their conversational partners. For example, in the command \"If it snows at night then wake me up early because I don't want to be late for work\" the speaker relies on commonsense reasoning of the listener to infer the implicit presumption that they wish to be woken only if it snows enough to cause traffic slowdowns. We consider here the problem of understanding such imprecisely stated natural language commands given in the form of if-(state), then-(action), because-(goal) statements. More precisely, we consider the problem of identifying the unstated presumptions of the speaker that allow the requested action to achieve the desired goal from the given state (perhaps elaborated by making the implicit presumptions explicit). We release a benchmark data set for this task, collected from humans and annotated with commonsense presumptions. We present a neuro-symbolic theorem prover that extracts multi-hop reasoning chains, and apply it to this problem. Furthermore, to accommodate the reality that current AI commonsense systems lack full coverage, we also present an interactive conversational framework built on our neuro-symbolic system, that conversationally evokes commonsense knowledge from humans to complete its reasoning chains."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Neuro-symbolic Architectures for Context Understanding",
                    "abstract": "Computational context understanding refers to an agent's ability to fuse disparate sources of information for decision-making and is, therefore, generally regarded as a prerequisite for sophisticated machine reasoning capabilities, such as in artificial intelligence (AI). Data-driven and knowledge-driven methods are two classical techniques in the pursuit of such machine sense-making capability. However, while data-driven methods seek to model the statistical regularities of events by making observations in the real-world, they remain difficult to interpret and they lack mechanisms for naturally incorporating external knowledge. Conversely, knowledge-driven methods, combine structured knowledge bases, perform symbolic reasoning based on axiomatic principles, and are more interpretable in their inferential processing; however, they often lack the ability to estimate the statistical salience of an inference. To combat these issues, we propose the use of hybrid AI methodology as a general framework for combining the strengths of both approaches. Specifically, we inherit the concept of neuro-symbolism as a way of using knowledge-bases to guide the learning progress of deep neural networks. We further ground our discussion in two applications of neuro-symbolism and, in both cases, show that our systems maintain interpretability while achieving comparable performance, relative to the state-of-the-art."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Learning Semantics-aware Distance Map with Semantics Layering Network for Amodal Instance Segmentation",
                    "abstract": "In this work, we demonstrate yet another approach to tackle the amodal segmentation problem. Specifically, we first introduce a new representation, namely a semantics-aware distance map (sem-dist map), to serve as our target for amodal segmentation instead of the commonly used masks and heatmaps. The sem-dist map is a kind of level-set representation, of which the different regions of an object are placed into different levels on the map according to their visibility. It is a natural extension of masks and heatmaps, where modal, amodal segmentation, as well as depth order information, are all well-described. Then we also introduce a novel convolutional neural network (CNN) architecture, which we refer to as semantic layering network, to estimate sem-dist maps layer by layer, from the global-level to the instance-level, for all objects in an image. Extensive experiments on the COCOA and D2SA datasets have demonstrated that our framework can predict amodal segmentation, occlusion, and depth order with state-of-the-art performance."
                },
                {
                    "title": "Classification Accuracy Score for Conditional Generative Models",
                    "abstract": "Deep generative models (DGMs) of images are now sufficiently mature that they produce nearly photorealistic samples and obtain scores similar to the data distribution on heuristics such as Frechet Inception Distance (FID). These results, especially on large-scale datasets such as ImageNet, suggest that DGMs are learning the data distribution in a perceptually meaningful space and can be used in downstream tasks. To test this latter hypothesis, we use class-conditional generative models from a number of model classes---variational autoencoders, autoregressive models, and generative adversarial networks (GANs)---to infer the class labels of real data. We perform this inference by training an image classifier using only synthetic data and using the classifier to predict labels on real data. The performance on this task, which we call Classification Accuracy Score (CAS), reveals some surprising results not identified by traditional metrics and constitute our contributions. First, when using a state-of-the-art GAN (BigGAN-deep), Top-1 and Top-5 accuracy decrease by 27.9\\% and 41.6\\%, respectively, compared to the original data; and conditional generative models from other model classes, such as Vector-Quantized Variational Autoencoder-2 (VQ-VAE-2) and Hierarchical Autoregressive Models (HAMs), substantially outperform GANs on this benchmark. Second, CAS automatically surfaces particular classes for which generative models failed to capture the data distribution, and were previously unknown in the literature. Third, we find traditional GAN metrics such as Inception Score (IS) and FID neither predictive of CAS nor useful when evaluating non-GAN models. Furthermore, in order to facilitate better diagnoses of generative models, we open-source the proposed metric."
                },
                {
                    "title": "The Neuro-Symbolic Concept Learner: Interpreting Scenes, Words, and Sentences From Natural Supervision",
                    "abstract": "We propose the Neuro-Symbolic Concept Learner (NS-CL), a model that learns visual concepts, words, and semantic parsing of sentences without explicit supervision on any of them; instead, our model learns by simply looking at images and reading paired questions and answers. Our model builds an object-based scene representation and translates sentences into executable, symbolic programs. To bridge the learning of two modules, we use a neuro-symbolic reasoning module that executes these programs on the latent scene representation. Analogical to human concept learning, the perception module learns visual concepts based on the language description of the object being referred to. Meanwhile, the learned visual concepts facilitate learning new words and parsing new sentences. We use curriculum learning to guide the searching over the large compositional space of images and language. Extensive experiments demonstrate the accuracy and efficiency of our model on learning visual concepts, word representations, and semantic parsing of sentences. Further, our method allows easy generalization to new object attributes, compositions, language concepts, scenes and questions, and even new program domains. It also empowers applications including visual question answering and bidirectional image-text retrieval."
                },
                {
                    "title": "Neural-Symbolic VQA: Disentangling Reasoning from Vision and Language Understanding",
                    "abstract": "We marry two powerful ideas: deep representation learning for visual recognition and language understanding, and symbolic program execution for reasoning. Our neural-symbolic visual question answering (NS-VQA) system first recovers a structural scene representation from the image and a program trace from the question. It then executes the program on the scene representation to obtain an answer. Incorporating symbolic structure as prior knowledge offers three unique advantages. First, executing programs on a symbolic space is more robust to long program traces; our model can solve complex reasoning tasks better, achieving an accuracy of 99.8% on the CLEVR dataset. Second, the model is more data- and memory-efficient: it performs well after learning on a small number of training data; it can also encode an image into a compact representation, requiring less storage than existing methods for offline question answering. Third, symbolic program execution offers full transparency to the reasoning process; we are thus able to interpret and diagnose each execution step."
                },
                {
                    "title": "Vision-and-Language Navigation: Interpreting Visually-Grounded Navigation Instructions in Real Environments",
                    "abstract": "A robot that can carry out a natural-language instruction has been a dream since before the Jetsons cartoon series imagined a life of leisure mediated by a fleet of attentive robot helpers. It is a dream that remains stubbornly distant. However, recent advances in vision and language methods have made incredible progress in closely related areas. This is significant because a robot interpreting a natural-language navigation instruction on the basis of what it sees is carrying out a vision and language process that is similar to Visual Question Answering. Both tasks can be interpreted as visually grounded sequence-to-sequence translation problems, and many of the same methods are applicable. To enable and encourage the application of vision and language methods to the problem of interpreting visually-grounded navigation instructions, we present the Matter-port3D Simulator - a large-scale reinforcement learning environment based on real imagery [11]. Using this simulator, which can in future support a range of embodied vision and language tasks, we provide the first benchmark dataset for visually-grounded natural language navigation in real buildings - the Room-to-Room (R2R) dataset1."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Neural-Symbolic Learning and Reasoning: A Survey and Interpretation",
                    "abstract": "The study and understanding of human behaviour is relevant to computer science, artificial intelligence, neural computation, cognitive science, philosophy, psychology, and several other areas. Presupposing cognition as basis of behaviour, among the most prominent tools in the modelling of behaviour are computational-logic systems, connectionist models of cognition, and models of uncertainty. Recent studies in cognitive science, artificial intelligence, and psychology have produced a number of cognitive models of reasoning, learning, and language that are underpinned by computation. In addition, efforts in computer science research have led to the development of cognitive computational systems integrating machine learning and automated reasoning. Such systems have shown promise in a range of applications, including computational biology, fault diagnosis, training and assessment in simulators, and software verification. This joint survey reviews the personal ideas and views of several researchers on neural-symbolic learning and reasoning. The article is organised in three parts: Firstly, we frame the scope and goals of neural-symbolic computation and have a look at the theoretical foundations. We then proceed to describe the realisations of neural-symbolic computation, systems, and applications. Finally we present the challenges facing the area and avenues for further research."
                },
                {
                    "title": "Wasserstein Generative Adversarial Networks",
                    "abstract": "We introduce a new algorithm named WGAN, an alternative to traditional GAN training. In this new model, we show that we can improve the stability of learning, get rid of problems like mode collapse, and provide meaningful learning curves useful for debugging and hyperparameter searches. Furthermore, we show that the corresponding optimization problem is sound, and provide extensive theoretical work highlighting the deep connections to different distances between distributions."
                },
                {
                    "title": "Logic Tensor Networks for Semantic Image Interpretation",
                    "abstract": "Semantic Image Interpretation (SII) is the task of extracting structured semantic descriptions from images. It is widely agreed that the combined use of visual data and background knowledge is of great importance for SII. Recently, Statistical Relational Learning (SRL) approaches have been developed for reasoning under uncertainty and learning in the presence of data and rich knowledge. Logic Tensor Networks (LTNs) are an SRL framework which integrates neural networks with first-order fuzzy logic to allow (i) efficient learning from noisy data in the presence of logical constraints, and (ii) reasoning with logical formulas describing general properties of the data. In this paper, we develop and apply LTNs to two of the main tasks of SII, namely, the classification of an image's bounding boxes and the detection of the relevant part-of relations between objects. To the best of our knowledge, this is the first successful application of SRL to such SII tasks. The proposed approach is evaluated on a standard image processing benchmark. Experiments show that the use of background knowledge in the form of logical constraints can improve the performance of purely data-driven approaches, including the state-of-the-art Fast Region-based Convolutional Neural Networks (Fast R-CNN). Moreover, we show that the use of logical background knowledge adds robustness to the learning system when errors are present in the labels of the training data."
                },
                {
                    "title": "Improved Training of Wasserstein GANs",
                    "abstract": "Generative Adversarial Networks (GANs) are powerful generative models, but suffer from training instability. The recently proposed Wasserstein GAN (WGAN) makes progress toward stable training of GANs, but sometimes can still generate only low-quality samples or fail to converge. We find that these problems are often due to the use of weight clipping in WGAN to enforce a Lipschitz constraint on the critic, which can lead to undesired behavior. We propose an alternative to clipping weights: penalize the norm of gradient of the critic with respect to its input. Our proposed method performs better than standard WGAN and enables stable training of a wide variety of GAN architectures with almost no hyperparameter tuning, including 101-layer ResNets and language models over discrete data. We also achieve high quality generations on CIFAR-10 and LSUN bedrooms."
                },
                {
                    "title": "Improved Texture Networks: Maximizing Quality and Diversity in Feed-Forward Stylization and Texture Synthesis",
                    "abstract": "The recent work of Gatys et al., who characterized the style of an image by the statistics of convolutional neural network filters, ignited a renewed interest in the texture generation and image stylization problems. While their image generation technique uses a slow optimization process, recently several authors have proposed to learn generator neural networks that can produce similar outputs in one quick forward pass. While generator networks are promising, they are still inferior in visual quality and diversity compared to generation-by-optimization. In this work, we advance them in two significant ways. First, we introduce an instance normalization module to replace batch normalization with significant improvements to the quality of image stylization. Second, we improve diversity by introducing a new learning formulation that encourages generators to sample unbiasedly from the Julesz texture ensemble, which is the equivalence class of all images characterized by certain filter responses. Together, these two improvements take feed forward texture synthesis and image stylization much closer to the quality of generation-via-optimization, while retaining the speed advantage."
                },
                {
                    "title": "Image-to-Image Translation with Conditional Adversarial Networks",
                    "abstract": "We investigate conditional adversarial networks as a general-purpose solution to image-to-image translation problems. These networks not only learn the mapping from input image to output image, but also learn a loss function to train this mapping. This makes it possible to apply the same generic approach to problems that traditionally would require very different loss formulations. We demonstrate that this approach is effective at synthesizing photos from label maps, reconstructing objects from edge maps, and colorizing images, among other tasks. Moreover, since the release of the pix2pix software associated with this paper, hundreds of twitter users have posted their own artistic experiments using our system. As a community, we no longer hand-engineer our mapping functions, and this work suggests we can achieve reasonable results without handengineering our loss functions either."
                },
                {
                    "title": "Unrolled Generative Adversarial Networks",
                    "abstract": "We introduce a method to stabilize Generative Adversarial Networks (GANs) by defining the generator objective with respect to an unrolled optimization of the discriminator. This allows training to be adjusted between using the optimal discriminator in the generator's objective, which is ideal but infeasible in practice, and using the current value of the discriminator, which is often unstable and leads to poor solutions. We show how this technique solves the common problem of mode collapse, stabilizes training of GANs with complex recurrent generators, and increases diversity and coverage of the data distribution by the generator."
                },
                {
                    "title": "Instance Normalization: The Missing Ingredient for Fast Stylization",
                    "abstract": "It this paper we revisit the fast stylization method introduced in Ulyanov et. al. (2016). We show how a small change in the stylization architecture results in a significant qualitative improvement in the generated images. The change is limited to swapping batch normalization with instance normalization, and to apply the latter both at training and testing times. The resulting method can be used to train high-performance architectures for real-time image generation. The code will is made available on github at this https URL. Full paper can be found at arXiv:1701.02096."
                },
                {
                    "title": "Semantic Amodal Segmentation",
                    "abstract": "Common visual recognition tasks such as classification, object detection, and semantic segmentation are rapidly reaching maturity, and given the recent rate of progress, it is not unreasonable to conjecture that techniques for many of these problems will approach human levels of performance in the next few years. In this paper we look to the future: what is the next frontier in visual recognition? We offer one possible answer to this question. We propose a detailed image annotation that captures information beyond the visible pixels and requires complex reasoning about full scene structure. Specifically, we create an amodal segmentation of each image: the full extent of each region is marked, not just the visible pixels. Annotators outline and name all salient regions in the image and specify a partial depth order. The result is a rich scene structure, including visible and occluded portions of each region, figure-ground edge information, semantic labels, and object overlap. We create two datasets for semantic amodal segmentation. First, we label 500 images in the BSDS dataset with multiple annotators per image, allowing us to study the statistics of human annotations. We show that the proposed full scene annotation is surprisingly consistent between annotators, including for regions and edges. Second, we annotate 5000 images from COCO. This larger dataset allows us to explore a number of algorithmic ideas for amodal segmentation and depth ordering. We introduce novel metrics for these tasks, and along with our strong baselines, define concrete new challenges for the community."
                },
                {
                    "title": "Analyzing Likert Data",
                    "abstract": "This article provides information for Extension professionals on the correct analysis of Likert data. The analyses of Likert-type and Likert scale data require unique data analysis procedures, and as a result, misuses and/or mistakes often occur. This article discusses the differences between Likert-type and Likert scale data and provides recommendations for descriptive statistics to be used during the analysis. Once a researcher understands the difference between Likert-type and Likert scale data, the decision on appropriate statistical procedures will be apparent."
                },
                {
                    "title": "The conference paper",
                    "abstract": "Purpose - A satire on conference paper presentations. Design/methodology/approach - A poetic critique of conference presenters and their occasional failure to engage with their audience. Findings - What may be fascinating and of assumed and obvious importance to a research presenter may be quite obscure to their audience, who nonetheless participate in the ritual of conference attendance and polite \u201clistening\u201d. Research limitations/implications - The reader may discover even more meaning in this poem than its original author envisaged. Originality/value - A poignant reminder of the perils of many conference presentations."
                },
                {
                    "title": "Learning where and when to reason in neuro-symbolic inference",
                    "abstract": "The integration of hard constraints on neural network outputs is a very desirable capability. This allows to instill trust in AI by guaranteeing the sanity of that neural network predictions with respect to domain knowledge. Recently, this topic has received a lot of attention. However, all the existing methods usually either impose the constraints in a \u201cweak\u201d form at training time, with no guarantees at inference, or fail to provide a general framework that supports different tasks and constraint types. We tackle this open problem from a neuro-symbolic perspective. Our pipeline enhances a conventional neural predictor with (1) a symbolic reasoning module capable of correcting structured prediction errors and (2) a neural attention module that learns to direct the reasoning effort to focus on potential prediction errors, while keeping other outputs unchanged. This framework provides an appealing trade-off between the efficiency of constraint-free neural inference and the prohibitive cost of exhaustive reasoning at inference time. We show that our method outperforms the state of the art on visual-Sudoku, and can also benefit visual scene graph prediction. Furthermore, it can improve the performance of existing neuro-symbolic systems that lack our explicit reasoning during inference."
                },
                {
                    "title": "\u201cTechnique for the Measurement of Attitudes, A\u201d",
                    "abstract": "THIS paper describes a technique which has been developed for the measurement of race prejudice. This technique differs from most prejudice inventories in that it avoids the following assumptions: (a) that the individual can say, to his own or the investigator's satisfaction, \"This is how prejudiced I am,\" and (b) that, to the extent that the individual can accurately assess his degree of antipathy, he will report honestly the findings of such introspection. Most sociologists would perhaps agree that race attitudes rarely reside on a completely articulate level. Even where the individual holds to intellectual or ideological convictions which would seem to leave no room for out-group antipathies, such do persevere. Thus, we may expect the number of Americans who honestly think themselves \"unprejudiced\" to be considerably larger than effective research would reveal. Moreover, the number who present themselves as unprejudiced probably exceeds considerably the number who honestly, though often inaccurately, see themselves in this light. Most indirect techniques for the measurement of attitudes have their rationale in observations such as these. The instrument to be described here is not, however, indirect in the usual sense of the word; it does not seek responses to items apparently unrelated to the attitudes investigated. We do, however, seek to measure prejudice in a manner less direct than is true of the usual prejudice scale. In our instrument we seek to measure anti-Negro prejudice. Persons are called upon to respond on social distance scales to whites and Negroes who occupy a variety of occupational positions. The measure of prejudice is derived through the summation of the differences in distance responses to Negroes as opposed to whites in the same occupations. Thus, for lack of a better label,"
                },
                {
                    "title": "Agent-based software engineering",
                    "abstract": "The technology of intelligent agents and multi-agent systems is expected to alter radically the way in which complex, distributed, open systems are conceptualised and implemented. The paper considers the problem of building a multi-agent system as a software engineering enterprise. Three issues are focused on: how agents might be specified; how these specifications might be refined or otherwise transformed into efficient implementations: and how implemented agents and multi-agent systems might subsequently be verified, to show that they are correct with respect to their specifications. These issues are discussed with reference to a number of case studies. The paper concludes by setting out some issues and open problems for future research."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a flexible image translation paradigm that effectively identifies and modifies specific regions of interest (RoI) in images while maintaining high fidelity and coherence in the output?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of image translation, as it addresses the limitations of current diffusion-based methods, such as condition-rigid learning and context-free incompetence. By enabling more precise control over image modifications, this research could lead to significant improvements in applications like content creation, augmented reality, and automated image editing. Furthermore, it could inspire future research into neuro-symbolic approaches, enhancing our understanding of how to integrate symbolic reasoning with machine learning techniques.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the need for a robust mechanism to balance conditional and unconditional predictions, which current methods struggle with due to their reliance on manually crafted parameters. Additionally, achieving context-aware modifications requires a deep understanding of image semantics and structure, which is lacking in existing solutions. The black box nature of diffusion models further complicates the task, as it limits user control and understanding of the model's decision-making processes, making it difficult to establish trust and improve the system.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either global manipulation of image attributes or on rigid conditional learning, leading to a lack of flexibility and precision in image translation tasks. The barriers to solving this problem include the absence of effective frameworks that combine symbolic reasoning with machine learning, as well as the limitations of existing models in handling specific RoI modifications. Our approach differs by integrating a condition-flexible diffusion model with GPT, allowing for a more dynamic and context-aware image translation process.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the Diffusion Visual Programmer (DVP), which utilizes a condition-flexible diffusion model integrated with GPT to plan and execute image translation tasks. The DVP will leverage off-the-shelf computer vision models to identify the target RoI and perform the necessary modifications. We will evaluate the model using metrics such as image fidelity, coherence, and user control over the translation process. Expected outcomes include improved precision in RoI modifications and enhanced user trust in the model's outputs, paving the way for practical applications in various domains."
            }
        },
        "author_data": {
            "abadf98e-03d1-49bb-b680-f8795057ba7c": {
                "pk": "abadf98e-03d1-49bb-b680-f8795057ba7c",
                "name": "Cheng Han",
                "collaborators": [
                    "Dongfang Liu",
                    "Qifan Wang",
                    "Yiming Cui",
                    "Lifu Huang",
                    "Yawen Lu",
                    "Zhiwen Cao",
                    "James Liang",
                    "Wenguan Wang",
                    "S. Dianat",
                    "Raghuveer M. Rao"
                ],
                "domain": [
                    "Computer Vision",
                    "Transformer",
                    "Multi-task Learning",
                    "Knowledge Distillation"
                ],
                "publications": [
                    {
                        "title": "AMD: Automatic Multi-step Distillation of Large-scale Vision Models",
                        "abstract": "Transformer-based architectures have become the de-facto standard models for diverse vision tasks owing to their superior performance. As the size of the models continues to scale up, model distillation becomes extremely important in various real applications, particularly on devices limited by computational resources. However, prevailing knowledge distillation methods exhibit diminished efficacy when confronted with a large capacity gap between the teacher and the student, e.g, 10x compression rate. In this paper, we present a novel approach named Automatic Multi-step Distillation (AMD) for large-scale vision model compression. In particular, our distillation process unfolds across multiple steps. Initially, the teacher undergoes distillation to form an intermediate teacher-assistant model, which is subsequently distilled further to the student. An efficient and effective optimization framework is introduced to automatically identify the optimal teacher-assistant that leads to the maximal student performance. We conduct extensive experiments on multiple image classification datasets, including CIFAR-10, CIFAR-100, and ImageNet. The findings consistently reveal that our approach outperforms several established baselines, paving a path for future knowledge distillation methods on large-scale vision models."
                    },
                    {
                        "title": "ProMotion: Prototypes as Motion Learners",
                        "abstract": "In this work, we introduce PRoMoTION, a unified proto-typical transformer-based framework engineered to model fundamental motion tasks. PRoMoTION offers a range of compelling attributes that set it apart from current task-specific paradigms. (1) We adopt a prototypical perspective, establishing a unified paradigm that harmonizes disparate motion learning approaches. This novel paradigm stream-lines the architectural design, enabling the simultaneous assimilation of diverse motion information. (2) We capitalize on a dual mechanism involving the feature denoiser and the prototypical learner to decipher the intricacies of motion. This approach effectively circumvents the pitfalls of ambiguity in pixel-wise feature matching, significantly bolstering the robustness of motion representation. (3)) We demon-strate a profound degree of transferability across distinct motion patterns. This inherent versatility reverberates robustly across a comprehensive spectrum of both 2D and 3D downstream tasks. Empirical results demonstrate that PRoMOTION outperforms various well-known specialized architectures, achieving 0.54 and 0.054 $AbsRel$ error on the Sintel and KITTI depth datasets, 1.04 and 2.01 average endpoint error on the clean and final pass of Sintel flow benchmark, and 4.30 F1-all error on the KITTI flow bench-mark. For its efficacy, we hope our work can catalyze a paradigm shift in universal models in computer vision."
                    },
                    {
                        "title": "Optical Flow as Spatial-Temporal Attention Learners.",
                        "abstract": "Optical flow is an indispensable building block for various important computer vision tasks, including motion estimation, object tracking, and disparity measurement. To date, the dominant methods are CNN-based, leaving plenty of room for improvement. In\u00a0this\u00a0work, we propose TransFlow, a transformer architecture for optical flow estimation. Compared to dominant CNN-based methods, TransFlow demonstrates three advantages. First, it provides more accurate correlation and trustworthy matching in flow estimation by utilizing spatial self-attention and cross-attention mechanisms between adjacent frames to effectively capture global dependencies; Second, it recovers more compromised information (e.g. occlusion and motion blur) in flow estimation through long-range temporal association in dynamic scenes; Third, it introduces a concise self-learning paradigm, eliminating the need for complex and laborious multi-stage pre-training procedures. The versatility and superiority of TransFlow extend seamlessly to 3D scene motion, yielding competitive outcomes in 3D scene flow estimation. Our approach attains state-of-the-art results on benchmark datasets such as Sintel and KITTI-15, while also exhibiting exceptional performance on downstream tasks, including video object detection using the ImageNet VID dataset, video frame interpolation using the GoPro dataset, and video stabilization using the DeepStab dataset. We believe that the effectiveness of TransFlow positions it as a flexible baseline for both optical flow and scene flow estimation, offering promising avenues for future research and development."
                    },
                    {
                        "title": "M2PT: Multimodal Prompt Tuning for Zero-shot Instruction Learning",
                        "abstract": "Multimodal Large Language Models (MLLMs) demonstrate remarkable performance across a wide range of domains, with increasing emphasis on enhancing their zero-shot generalization capabilities for unseen tasks across various modalities. Instruction tuning has emerged as an effective strategy for achieving zero-shot generalization by finetuning pretrained models on diverse multimodal tasks. As the scale of MLLMs continues to grow, parameter-efficient finetuning becomes increasingly critical. However, most existing parameter-efficient approaches focus only on single modalities and often overlook the multimodal characteristics during finetuning. In this work, we introduce a novel Multimodal Prompt Tuning (M$^2$PT) approach for efficient instruction tuning of MLLMs. M$^2$PT effectively integrates visual and textual prompts into the vision encoder and language processor respectively during finetuning, facilitating the extraction and alignment of features across modalities. Empirical results on various multimodal evaluation datasets demonstrate the superior performance of our approach compared to several state-of-the-art baselines. A comprehensive set of ablation studies validates the effectiveness of our prompt design and the efficiency of our approach."
                    },
                    {
                        "title": "Prototypical Transformer as Unified Motion Learners",
                        "abstract": "In this work, we introduce the Prototypical Transformer (ProtoFormer), a general and unified framework that approaches various motion tasks from a prototype perspective. ProtoFormer seamlessly integrates prototype learning with Transformer by thoughtfully considering motion dynamics, introducing two innovative designs. First, Cross-Attention Prototyping discovers prototypes based on signature motion patterns, providing transparency in understanding motion scenes. Second, Latent Synchronization guides feature representation learning via prototypes, effectively mitigating the problem of motion uncertainty. Empirical results demonstrate that our approach achieves competitive performance on popular motion tasks such as optical flow and scene depth. Furthermore, it exhibits generality across various downstream tasks, including object tracking and video stabilization."
                    },
                    {
                        "title": "Self-supervised Adversarial Training of Monocular Depth Estimation against Physical-World Attacks",
                        "abstract": "Monocular Depth Estimation (MDE) plays a vital role in applications such as autonomous driving. However, various attacks target MDE models, with physical attacks posing significant threats to system security. Traditional adversarial training methods, which require ground-truth labels, are not directly applicable to MDE models that lack ground-truth depth. Some self-supervised model hardening techniques (e.g., contrastive learning) overlook the domain knowledge of MDE, resulting in suboptimal performance. In this work, we introduce a novel self-supervised adversarial training approach for MDE models, leveraging view synthesis without the need for ground-truth depth. We enhance adversarial robustness against real-world attacks by incorporating L0-norm-bounded perturbation during training. We evaluate our method against supervised learning-based and contrastive learning-based approaches specifically designed for MDE. Our experiments with two representative MDE networks demonstrate improved robustness against various adversarial attacks, with minimal impact on benign performance. Our code: https://github.com/Bob-cheng/DepthModelHardening."
                    },
                    {
                        "title": "Stepwise Spatial Global-local Aggregation Networks for Autonomous Driving",
                        "abstract": "Visual-based perception is the key module for autonomous driving. Among those visual perception tasks, video object detection is a primary yet challenging one because of feature degradation caused by fast motion or multiple poses. Current models usually aggregate features from the neighboring frames to enhance the object representations for the task heads to generate more accurate predictions. Though getting better performance, these methods rely on the information from the future frames and suffer from high computational complexity. Meanwhile, the aggregation process is not reconfigurable during the inference time. These issues make most of the existing models infeasible for online applications. To solve these problems, we introduce a stepwise spatial global-local aggregation network. Our proposed models mainly contain three parts: 1). Multi-stage stepwise network gradually refines the predictions and object representations from the previous stage; 2). Spatial global-local aggregation fuses the local information from the neighboring frames and global semantics from the current frame to eliminate the feature degradation; 3). Dynamic aggregation strategy stops the aggregation process early based on the refinement results to remove redundancy and improve efficiency. Extensive experiments on the ImageNet VID benchmark validate the effectiveness and efficiency of our proposed models."
                    },
                    {
                        "title": "Facing the Elephant in the Room: Visual Prompt Tuning or Full Finetuning?",
                        "abstract": "As the scale of vision models continues to grow, the emergence of Visual Prompt Tuning (VPT) as a parameter-efficient transfer learning technique has gained attention due to its superior performance compared to traditional full-finetuning. However, the conditions favoring VPT (the ``when\") and the underlying rationale (the ``why\") remain unclear. In this paper, we conduct a comprehensive analysis across 19 distinct datasets and tasks. To understand the ``when\"aspect, we identify the scenarios where VPT proves favorable by two dimensions: task objectives and data distributions. We find that VPT is preferrable when there is 1) a substantial disparity between the original and the downstream task objectives (e.g., transitioning from classification to counting), or 2) a similarity in data distributions between the two tasks (e.g., both involve natural images). In exploring the ``why\"dimension, our results indicate VPT's success cannot be attributed solely to overfitting and optimization considerations. The unique way VPT preserves original features and adds parameters appears to be a pivotal factor. Our study provides insights into VPT's mechanisms, and offers guidance for its optimal utilization."
                    },
                    {
                        "title": "SSGA-Net: Stepwise Spatial Global-local Aggregation Networks for for Autonomous Driving",
                        "abstract": "Visual-based perception is the key module for autonomous driving. Among those visual perception tasks, video object detection is a primary yet challenging one because of feature degradation caused by fast motion or multiple poses. Current models usually aggregate features from the neighboring frames to enhance the object representations for the task heads to generate more accurate predictions. Though getting better performance, these methods rely on the information from the future frames and suffer from high computational complexity. Meanwhile, the aggregation process is not reconfigurable during the inference time. These issues make most of the existing models infeasible for online applications. To solve these problems, we introduce a stepwise spatial global-local aggregation network. Our proposed models mainly contain three parts: 1). Multi-stage stepwise network gradually refines the predictions and object representations from the previous stage; 2). Spatial global-local aggregation fuses the local information from the neighboring frames and global semantics from the current frame to eliminate the feature degradation; 3). Dynamic aggregation strategy stops the aggregation process early based on the refinement results to remove redundancy and improve efficiency. Extensive experiments on the ImageNet VID benchmark validate the effectiveness and efficiency of our proposed models."
                    },
                    {
                        "title": "Collaborative Multi-task Learning for Multi-Object Tracking and Segmentation",
                        "abstract": "The advancement of computer vision has pushed visual analysis tasks from still images to the video domain. In recent years, video instance segmentation, which aims at tracking and segment multiple objects in video frames, has drawn much attention for its potential applications in various emerging areas such as autonomous driving, intelligent transportation, and smart retail. In this article, we propose an effective framework for instance-level visual analysis on video frames, which can simultaneously conduct object detection, instance segmentation, and multi-object tracking. The core idea of our method is collaborative multi-task learning which is achieved by a novel structure, named associative connections among detection, segmentation, and tracking task heads in an end-to-end learnable CNN. These additional connections allow information propagation across multiple related tasks, so as to benefit these tasks simultaneously. We evaluate the proposed method extensively on KITTI MOTS and MOTS Challenge datasets and obtain quite encouraging results."
                    },
                    {
                        "title": "Prompt Learns Prompt: Exploring Knowledge-Aware Generative Prompt Collaboration For Video Captioning",
                        "abstract": "Fine-tuning large vision-language models is a challenging task. Prompt tuning approaches have been introduced to learn fixed textual or visual prompts while freezing the pre-trained model in downstream tasks. Despite the effectiveness of prompt tuning, what do those learnable prompts learn remains unexplained. In this work, we explore whether prompts in the fine-tuning can learn knowledge-aware prompts from the pre-training, by designing two different sets of prompts in pre-training and fine-tuning phases respectively. Specifically, we present a Video-Language Prompt tuning (VL-Prompt) approach for video captioning, which first efficiently pre-train a video-language model to extract key information (e.g., actions and objects) with flexibly generated Knowledge-Aware Prompt (KAP). Then, we design a Video-Language Prompt (VLP) to transfer the knowledge from the knowledge-aware prompts and fine-tune the model to generate full captions. Experimental results show the superior performance of our approach over several state-of-the-art baselines. We further demonstrate that the video-language prompts are well learned from the knowledge-aware prompts."
                    },
                    {
                        "title": "E2VPT: An Effective and Efficient Approach for Visual Prompt Tuning",
                        "abstract": "As the size of transformer-based, models continues to grow, fine-tuning these large-scale pretrained vision models for new tasks has become increasingly parameter-intensive. Parameter-efficient learning has been developed to reduce the number of tunable parameters during fine-tuning. Although these methods show promising results, there is still a significant performance gap compared to full fine-tuning. To address this challenge, we propose an Effective and Efficient Visual Prompt Tuning (E2VPT) approach for large-scale transformer-based model adaptation. Specifically, we introduce a set of learnable key-value prompts and visual prompts into self-attention and input layers, respectively, to improve the effectiveness of model fine-tuning. Moreover, we design a prompt pruning procedure to systematically prune low importance prompts while preserving model performance, which largely enhances the model\u2019s efficiency. Empirical results demonstrate that our approach outperforms several state-of-the-art baselines on two benchmarks, with considerably low parameter usage (e.g., 0.32% of model parameters on VTAB-1k). Our code is available at https://github.com/ChengHan111/E2VPT."
                    },
                    {
                        "title": "Visual Recognition with Deep Nearest Centroids",
                        "abstract": "We devise deep nearest centroids (DNC), a conceptually elegant yet surprisingly effective network for large-scale visual recognition, by revisiting Nearest Centroids, one of the most classic and simple classifiers. Current deep models learn the classifier in a fully parametric manner, ignoring the latent data structure and lacking simplicity and explainability. DNC instead conducts nonparametric, case-based reasoning; it utilizes sub-centroids of training samples to describe class distributions and clearly explains the classification as the proximity of test data and the class sub-centroids in the feature space. Due to the distance-based nature, the network output dimensionality is flexible, and all the learnable parameters are only for data embedding. That means all the knowledge learnt for ImageNet classification can be completely transferred for pixel recognition learning, under the\"pre-training and fine-tuning\"paradigm. Apart from its nested simplicity and intuitive decision-making mechanism, DNC can even possess ad-hoc explainability when the sub-centroids are selected as actual training images that humans can view and inspect. Compared with parametric counterparts, DNC performs better on image classification (CIFAR-10, ImageNet) and greatly boots pixel recognition (ADE20K, Cityscapes), with improved transparency and fewer learnable parameters, using various network architectures (ResNet, Swin) and segmentation models (FCN, DeepLabV3, Swin). We feel this work brings fundamental insights into related fields."
                    },
                    {
                        "title": "YOLOPv2: Better, Faster, Stronger for Panoptic Driving Perception",
                        "abstract": "Over the last decade, multi-tasking learning approaches have achieved promising results in solving panoptic driving perception problems, providing both high-precision and high-efficiency performance. It has become a popular paradigm when designing networks for real-time practical autonomous driving system, where computation resources are limited. This paper proposed an effective and efficient multi-task learning network to simultaneously perform the task of traffic object detection, drivable road area segmentation and lane detection. Our model achieved the new state-of-the-art (SOTA) performance in terms of accuracy and speed on the challenging BDD100K dataset. Especially, the inference time is reduced by half compared to the previous SOTA model. Code will be released in the near future."
                    }
                ]
            },
            "49f1b286-007d-443b-b03e-616c91383aaf": {
                "pk": "49f1b286-007d-443b-b03e-616c91383aaf",
                "name": "James C. Liang",
                "collaborators": [
                    "Dongfang Liu",
                    "Qifan Wang",
                    "Zhiyuan Cheng",
                    "Xiangyu Zhang",
                    "Tong Geng",
                    "Guanhong Tao",
                    "Cheng Han",
                    "Wenguan Wang",
                    "Hongjun Choi",
                    "Chuan Liu"
                ],
                "domain": [
                    "Nuclear Fusion",
                    "Machine Learning",
                    "Computer Vision",
                    "Adversarial Attacks"
                ],
                "publications": [
                    {
                        "title": "Diff-PIC: Revolutionizing Particle-In-Cell Nuclear Fusion Simulation with Diffusion Models",
                        "abstract": "The rapid development of AI highlights the pressing need for sustainable energy, a critical global challenge for decades. Nuclear fusion, generally seen as an ultimate solution, has been the focus of intensive research for nearly a century, with investments reaching hundreds of billions of dollars. Recent advancements in Inertial Confinement Fusion have drawn significant attention to fusion research, in which Laser-Plasma Interaction (LPI) is critical for ensuring fusion stability and efficiency. However, the complexity of LPI upon fusion ignition makes analytical approaches impractical, leaving researchers depending on extremely computation-demanding Particle-in-Cell (PIC) simulations to generate data, presenting a significant bottleneck to advancing fusion research. In response, this work introduces Diff-PIC, a novel framework that leverages conditional diffusion models as a computationally efficient alternative to PIC simulations for generating high-fidelity scientific LPI data. In this work, physical patterns captured by PIC simulations are distilled into diffusion models associated with two tailored enhancements: (1) To effectively capture the complex relationships between physical parameters and corresponding outcomes, the parameters are encoded in a physically-informed manner. (2) To further enhance efficiency while maintaining high fidelity and physical validity, the rectified flow technique is employed to transform our model into a one-step conditional diffusion model. Experimental results show that Diff-PIC achieves 16,200$\\times$ speedup compared to traditional PIC on a 100 picosecond simulation, with an average reduction in MAE / RMSE / FID of 59.21% / 57.15% / 39.46% with respect to two other SOTA data generation approaches."
                    },
                    {
                        "title": "M2PT: Multimodal Prompt Tuning for Zero-shot Instruction Learning",
                        "abstract": "Multimodal Large Language Models (MLLMs) demonstrate remarkable performance across a wide range of domains, with increasing emphasis on enhancing their zero-shot generalization capabilities for unseen tasks across various modalities. Instruction tuning has emerged as an effective strategy for achieving zero-shot generalization by finetuning pretrained models on diverse multimodal tasks. As the scale of MLLMs continues to grow, parameter-efficient finetuning becomes increasingly critical. However, most existing parameter-efficient approaches focus only on single modalities and often overlook the multimodal characteristics during finetuning. In this work, we introduce a novel Multimodal Prompt Tuning (M$^2$PT) approach for efficient instruction tuning of MLLMs. M$^2$PT effectively integrates visual and textual prompts into the vision encoder and language processor respectively during finetuning, facilitating the extraction and alignment of features across modalities. Empirical results on various multimodal evaluation datasets demonstrate the superior performance of our approach compared to several state-of-the-art baselines. A comprehensive set of ablation studies validates the effectiveness of our prompt design and the efficiency of our approach."
                    },
                    {
                        "title": "Prototypical Transformer as Unified Motion Learners",
                        "abstract": "In this work, we introduce the Prototypical Transformer (ProtoFormer), a general and unified framework that approaches various motion tasks from a prototype perspective. ProtoFormer seamlessly integrates prototype learning with Transformer by thoughtfully considering motion dynamics, introducing two innovative designs. First, Cross-Attention Prototyping discovers prototypes based on signature motion patterns, providing transparency in understanding motion scenes. Second, Latent Synchronization guides feature representation learning via prototypes, effectively mitigating the problem of motion uncertainty. Empirical results demonstrate that our approach achieves competitive performance on popular motion tasks such as optical flow and scene depth. Furthermore, it exhibits generality across various downstream tasks, including object tracking and video stabilization."
                    },
                    {
                        "title": "Self-supervised Adversarial Training of Monocular Depth Estimation against Physical-World Attacks",
                        "abstract": "Monocular Depth Estimation (MDE) plays a vital role in applications such as autonomous driving. However, various attacks target MDE models, with physical attacks posing significant threats to system security. Traditional adversarial training methods, which require ground-truth labels, are not directly applicable to MDE models that lack ground-truth depth. Some self-supervised model hardening techniques (e.g., contrastive learning) overlook the domain knowledge of MDE, resulting in suboptimal performance. In this work, we introduce a novel self-supervised adversarial training approach for MDE models, leveraging view synthesis without the need for ground-truth depth. We enhance adversarial robustness against real-world attacks by incorporating L0-norm-bounded perturbation during training. We evaluate our method against supervised learning-based and contrastive learning-based approaches specifically designed for MDE. Our experiments with two representative MDE networks demonstrate improved robustness against various adversarial attacks, with minimal impact on benign performance. Our code: https://github.com/Bob-cheng/DepthModelHardening."
                    },
                    {
                        "title": "Inertial Confinement Fusion Forecasting via Large Language Models",
                        "abstract": "Controlled fusion energy is deemed pivotal for the advancement of human civilization. In this study, we introduce $\\textbf{LPI-LLM}$, a novel integration of Large Language Models (LLMs) with classical reservoir computing paradigms tailored to address a critical challenge, Laser-Plasma Instabilities ($\\texttt{LPI}$), in Inertial Confinement Fusion ($\\texttt{ICF}$). Our approach offers several key contributions: Firstly, we propose the $\\textit{LLM-anchored Reservoir}$, augmented with a $\\textit{Fusion-specific Prompt}$, enabling accurate forecasting of $\\texttt{LPI}$-generated-hot electron dynamics during implosion. Secondly, we develop $\\textit{Signal-Digesting Channels}$ to temporally and spatially describe the driver laser intensity across time, capturing the unique characteristics of $\\texttt{ICF}$ inputs. Lastly, we design the $\\textit{Confidence Scanner}$ to quantify the confidence level in forecasting, providing valuable insights for domain experts to design the $\\texttt{ICF}$ process. Extensive experiments demonstrate the superior performance of our method, achieving 1.90 CAE, 0.14 $\\texttt{top-1}$ MAE, and 0.11 $\\texttt{top-5}$ MAE in predicting Hard X-ray ($\\texttt{HXR}$) energies emitted by the hot electrons in $\\texttt{ICF}$ implosions, which presents state-of-the-art comparisons against concurrent best systems. Additionally, we present $\\textbf{LPI4AI}$, the first $\\texttt{LPI}$ benchmark based on physical experiments, aimed at fostering novel ideas in $\\texttt{LPI}$ research and enhancing the utility of LLMs in scientific exploration. Overall, our work strives to forge an innovative synergy between AI and $\\texttt{ICF}$ for advancing fusion energy."
                    },
                    {
                        "title": "Adversarial Training of Self-supervised Monocular Depth Estimation against Physical-World Attacks",
                        "abstract": "Monocular Depth Estimation (MDE) is a critical component in applications such as autonomous driving. There are various attacks against MDE networks. These attacks, especially the physical ones, pose a great threat to the security of such systems. Traditional adversarial training method requires ground-truth labels hence cannot be directly applied to self-supervised MDE that does not have ground-truth depth. Some self-supervised model hardening techniques (e.g., contrastive learning) ignore the domain knowledge of MDE and can hardly achieve optimal performance. In this work, we propose a novel adversarial training method for self-supervised MDE models based on view synthesis without using ground-truth depth. We improve adversarial robustness against physical-world attacks using L0-norm-bounded perturbation in training. We compare our method with supervised learning based and contrastive learning based methods that are tailored for MDE. Results on two representative MDE networks show that we achieve better robustness against various adversarial attacks with nearly no benign performance degradation."
                    },
                    {
                        "title": "CLUSTSEG: Clustering for Universal Segmentation",
                        "abstract": "We present CLUSTSEG, a general, transformer-based framework that tackles different image segmentation tasks (i.e., superpixel, semantic, instance, and panoptic) through a unified neural clustering scheme. Regarding queries as cluster centers, CLUSTSEG is innovative in two aspects:1) cluster centers are initialized in heterogeneous ways so as to pointedly address task-specific demands (e.g., instance- or category-level distinctiveness), yet without modifying the architecture; and 2) pixel-cluster assignment, formalized in a cross-attention fashion, is alternated with cluster center update, yet without learning additional parameters. These innovations closely link CLUSTSEG to EM clustering and make it a transparent and powerful framework that yields superior results across the above segmentation tasks."
                    },
                    {
                        "title": "Fusion Is Not Enough: Single Modal Attacks on Fusion Models for 3D Object Detection",
                        "abstract": "Multi-sensor fusion (MSF) is widely used in autonomous vehicles (AVs) for perception, particularly for 3D object detection with camera and LiDAR sensors. The purpose of fusion is to capitalize on the advantages of each modality while minimizing its weaknesses. Advanced deep neural network (DNN)-based fusion techniques have demonstrated the exceptional and industry-leading performance. Due to the redundant information in multiple modalities, MSF is also recognized as a general defence strategy against adversarial attacks. In this paper, we attack fusion models from the camera modality that is considered to be of lesser importance in fusion but is more affordable for attackers. We argue that the weakest link of fusion models depends on their most vulnerable modality, and propose an attack framework that targets advanced camera-LiDAR fusion-based 3D object detection models through camera-only adversarial attacks. Our approach employs a two-stage optimization-based strategy that first thoroughly evaluates vulnerable image areas under adversarial attacks, and then applies dedicated attack strategies for different fusion models to generate deployable patches. The evaluations with six advanced camera-LiDAR fusion models and one camera-only model indicate that our attacks successfully compromise all of them. Our approach can either decrease the mean average precision (mAP) of detection performance from 0.824 to 0.353, or degrade the detection score of a target object from 0.728 to 0.156, demonstrating the efficacy of our proposed attack framework. Code is available."
                    },
                    {
                        "title": "ClusterFormer: Clustering As A Universal Visual Learner",
                        "abstract": "This paper presents CLUSTERFORMER, a universal vision model that is based on the CLUSTERing paradigm with TransFORMER. It comprises two novel designs: 1. recurrent cross-attention clustering, which reformulates the cross-attention mechanism in Transformer and enables recursive updates of cluster centers to facilitate strong representation learning; and 2. feature dispatching, which uses the updated cluster centers to redistribute image features through similarity-based metrics, resulting in a transparent pipeline. This elegant design streamlines an explainable and transferable workflow, capable of tackling heterogeneous vision tasks (i.e., image classification, object detection, and image segmentation) with varying levels of clustering granularity (i.e., image-, box-, and pixel-level). Empirical results demonstrate that CLUSTERFORMER outperforms various well-known specialized architectures, achieving 83.41% top-1 acc. over ImageNet-1K for image classification, 54.2% and 47.0% mAP over MS COCO for object detection and instance segmentation, 52.4% mIoU over ADE20K for semantic segmentation, and 55.8% PQ over COCO Panoptic for panoptic segmentation. For its efficacy, we hope our work can catalyze a paradigm shift in universal models in computer vision."
                    },
                    {
                        "title": "Tripartite Feature Enhanced Pyramid Network for Dense Prediction",
                        "abstract": "Learning pyramidal feature representations is important for many dense prediction tasks (e.g., object detection, semantic segmentation) that demand multi-scale visual understanding. Feature Pyramid Network (FPN) is a well-known architecture for multi-scale feature learning, however, intrinsic weaknesses in feature extraction and fusion impede the production of informative features. This work addresses the weaknesses of FPN through a novel tripartite feature enhanced pyramid network (TFPN), with three distinct and effective designs. First, we develop a feature reference module with lateral connections to adaptively extract bottom-up features with richer details for feature pyramid construction. Second, we design a feature calibration module between adjacent layers that calibrates the upsampled features to be spatially aligned, allowing for feature fusion with accurate correspondences. Third, we introduce a feature feedback module in FPN, which creates a communication channel from the feature pyramid back to the bottom-up backbone and doubles the encoding capacity, enabling the entire architecture to generate incrementally more powerful representations. The TFPN is extensively evaluated over four popular dense prediction tasks, i.e., object detection, instance segmentation, panoptic segmentation, and semantic segmentation. The results demonstrate that TFPN consistently and significantly outperforms the vanilla FPN. Our code is available at https://github.com/jamesliang819."
                    },
                    {
                        "title": "Fusion is Not Enough: Single-Modal Attacks to Compromise Fusion Models in Autonomous Driving",
                        "abstract": "\u2014Multi-sensor fusion (MSF) is widely adopted for perception in autonomous vehicles (AVs), particularly for the task of 3D object detection with camera and LiDAR sensors. The rationale behind fusion is to capitalize on the strengths of each modality while mitigating their limitations. The exceptional and leading performance of fusion models has been demonstrated by advanced deep neural network (DNN)- based fusion techniques. Fusion models are also perceived as more robust to attacks compared to single-modal ones due to the redundant information in multiple modalities. In this work, we challenge this perspective with single-modal attacks that targets the camera modality, which is considered less signi\ufb01cant in fusion but more affordable for attackers. We argue that the weakest link of fusion models depends on their most vulnerable modality, and propose an attack framework that targets advanced camera-LiDAR fusion models with adversarial patches. Our approach employs a two-stage optimization-based strategy that \ufb01rst comprehensively assesses vulnerable image areas under adversarial attacks, and then applies customized attack strategies to different fusion models, generating deployable patches. Evaluations with \ufb01ve state-of-the-art camera-LiDAR fusion models on a real-world dataset show that our attacks successfully compromise all models. Our approach can either reduce the mean average precision (mAP) of detection performance from 0.824 to 0.353 or degrade the detection score of the target object from 0.727 to 0.151 on average, demonstrating the effectiveness and practicality of our proposed attack framework."
                    },
                    {
                        "title": "Learning Equivariant Segmentation with Instance-Unique Querying",
                        "abstract": "Prevalent state-of-the-art instance segmentation methods fall into a query-based scheme, in which instance masks are derived by querying the image feature using a set of instance-aware embeddings. In this work, we devise a new training framework that boosts query-based models through discriminative query embedding learning. It explores two essential properties, namely dataset-level uniqueness and transformation equivariance, of the relation between queries and instances. First, our algorithm uses the queries to retrieve the corresponding instances from the whole training dataset, instead of only searching within individual scenes. As querying instances across scenes is more challenging, the segmenters are forced to learn more discriminative queries for effective instance separation. Second, our algorithm encourages both image (instance) representations and queries to be equivariant against geometric transformations, leading to more robust, instance-query matching. On top of four famous, query-based models ($i.e.,$ CondInst, SOLOv2, SOTR, and Mask2Former), our training algorithm provides significant performance gains ($e.g.,$ +1.6 - 3.2 AP) on COCO dataset. In addition, our algorithm promotes the performance of SOLOv2 by 2.7 AP, on LVISv1 dataset."
                    },
                    {
                        "title": "Physical Attack on Monocular Depth Estimation with Optimal Adversarial Patches",
                        "abstract": "Deep learning has substantially boosted the performance of Monocular Depth Estimation (MDE), a critical component in fully vision-based autonomous driving (AD) systems (e.g., Tesla and Toyota). In this work, we develop an attack against learning-based MDE. In particular, we use an optimization-based method to systematically generate stealthy physical-object-oriented adversarial patches to attack depth estimation. We balance the stealth and effectiveness of our attack with object-oriented adversarial design, sensitive region localization, and natural style camouflage. Using real-world driving scenarios, we evaluate our attack on concurrent MDE models and a representative downstream task for AD (i.e., 3D object detection). Experimental results show that our method can generate stealthy, effective, and robust adversarial patches for different target objects and models and achieves more than 6 meters mean depth estimation error and 93% attack success rate (ASR) in object detection with a patch of 1/9 of the vehicle's rear area. Field tests on three different driving routes with a real vehicle indicate that we cause over 6 meters mean depth estimation error and reduce the object detection rate from 90.70% to 5.16% in continuous video frames."
                    }
                ]
            },
            "b3814dba-546e-4685-91a6-6df937f7060d": {
                "pk": "b3814dba-546e-4685-91a6-6df937f7060d",
                "name": "Qifan Wang",
                "collaborators": [
                    "Dongfang Liu",
                    "Yiming Cui",
                    "Wenguan Wang",
                    "Cheng Han",
                    "Siyuan Qi",
                    "Tong Geng",
                    "Siqi Ma",
                    "Victor Y. Chen",
                    "Zhiwen Cao",
                    "Lifu Huang"
                ],
                "domain": [
                    "Computer Vision",
                    "Transformer",
                    "Visual Prompt Tuning",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Facing the Elephant in the Room: Visual Prompt Tuning or Full Finetuning?",
                        "abstract": "As the scale of vision models continues to grow, the emergence of Visual Prompt Tuning (VPT) as a parameter-efficient transfer learning technique has gained attention due to its superior performance compared to traditional full-finetuning. However, the conditions favoring VPT (the ``when\") and the underlying rationale (the ``why\") remain unclear. In this paper, we conduct a comprehensive analysis across 19 distinct datasets and tasks. To understand the ``when\"aspect, we identify the scenarios where VPT proves favorable by two dimensions: task objectives and data distributions. We find that VPT is preferrable when there is 1) a substantial disparity between the original and the downstream task objectives (e.g., transitioning from classification to counting), or 2) a similarity in data distributions between the two tasks (e.g., both involve natural images). In exploring the ``why\"dimension, our results indicate VPT's success cannot be attributed solely to overfitting and optimization considerations. The unique way VPT preserves original features and adds parameters appears to be a pivotal factor. Our study provides insights into VPT's mechanisms, and offers guidance for its optimal utilization."
                    },
                    {
                        "title": "ClusterFormer: Clustering As A Universal Visual Learner",
                        "abstract": "This paper presents CLUSTERFORMER, a universal vision model that is based on the CLUSTERing paradigm with TransFORMER. It comprises two novel designs: 1. recurrent cross-attention clustering, which reformulates the cross-attention mechanism in Transformer and enables recursive updates of cluster centers to facilitate strong representation learning; and 2. feature dispatching, which uses the updated cluster centers to redistribute image features through similarity-based metrics, resulting in a transparent pipeline. This elegant design streamlines an explainable and transferable workflow, capable of tackling heterogeneous vision tasks (i.e., image classification, object detection, and image segmentation) with varying levels of clustering granularity (i.e., image-, box-, and pixel-level). Empirical results demonstrate that CLUSTERFORMER outperforms various well-known specialized architectures, achieving 83.41% top-1 acc. over ImageNet-1K for image classification, 54.2% and 47.0% mAP over MS COCO for object detection and instance segmentation, 52.4% mIoU over ADE20K for semantic segmentation, and 55.8% PQ over COCO Panoptic for panoptic segmentation. For its efficacy, we hope our work can catalyze a paradigm shift in universal models in computer vision."
                    },
                    {
                        "title": "TransFlow: Transformer as Flow Learner",
                        "abstract": "Optical flow is an indispensable building block for various important computer vision tasks, including motion estimation, object tracking, and disparity measurement. In this work, we propose TransFlow, a pure transformer architecture for optical flow estimation. Compared to dominant CNN-based methods, TransFlow demonstrates three advantages. First, it provides more accurate correlation and trustworthy matching in flow estimation by utilizing spatial self-attention and crossattention mechanisms between adjacent frames to effectively capture global dependencies; Second, it recovers more compromised information (e.g., occlusion and motion blur) in flow estimation through long-range temporal association in dynamic scenes; Third, it enables a concise self-learning paradigm and effectively eliminate the complex and laborious multi-stage pre-training procedures. We achieve the state-of-the-art results on the Sintel, KITTI-15, as well as several downstream tasks, including video object detection, interpolation and stabilization. For its efficacy, we hope TransFlow could serve as a flexible baseline for optical flow estimation."
                    },
                    {
                        "title": "MMViT: Multiscale Multiview Vision Transformers",
                        "abstract": "We present Multiscale Multiview Vision Transformers (MMViT), which introduces multiscale feature maps and multiview encodings to transformer models. Our model encodes different views of the input signal and builds several channel-resolution feature stages to process the multiple views of the input at different resolutions in parallel. At each scale stage, we use a cross-attention block to fuse information across different views. This enables the MMViT model to acquire complex high-dimensional representations of the input at different resolutions. The proposed model can serve as a backbone model in multiple domains. We demonstrate the effectiveness of MMViT on audio and image classification tasks, achieving state-of-the-art results."
                    },
                    {
                        "title": "E2VPT: An Effective and Efficient Approach for Visual Prompt Tuning",
                        "abstract": "As the size of transformer-based, models continues to grow, fine-tuning these large-scale pretrained vision models for new tasks has become increasingly parameter-intensive. Parameter-efficient learning has been developed to reduce the number of tunable parameters during fine-tuning. Although these methods show promising results, there is still a significant performance gap compared to full fine-tuning. To address this challenge, we propose an Effective and Efficient Visual Prompt Tuning (E2VPT) approach for large-scale transformer-based model adaptation. Specifically, we introduce a set of learnable key-value prompts and visual prompts into self-attention and input layers, respectively, to improve the effectiveness of model fine-tuning. Moreover, we design a prompt pruning procedure to systematically prune low importance prompts while preserving model performance, which largely enhances the model\u2019s efficiency. Empirical results demonstrate that our approach outperforms several state-of-the-art baselines on two benchmarks, with considerably low parameter usage (e.g., 0.32% of model parameters on VTAB-1k). Our code is available at https://github.com/ChengHan111/E2VPT."
                    },
                    {
                        "title": "Video Captioning Using Global-Local Representation",
                        "abstract": "Video captioning is a challenging task as it needs to accurately transform visual understanding into natural language description. To date, state-of-the-art methods inadequately model global-local vision representation for sentence generation, leaving plenty of room for improvement. In this work, we approach the video captioning task from a new perspective and propose a GLR framework, namely a global-local representation granularity. Our GLR demonstrates three advantages over the prior efforts. First, we propose a simple solution, which exploits extensive vision representations from different video ranges to improve linguistic expression. Second, we devise a novel global-local encoder, which encodes different video representations including long-range, short-range and local-keyframe, to produce rich semantic vocabulary for obtaining a descriptive granularity of video contents across frames. Finally, we introduce the progressive training strategy which can effectively organize feature learning to incur optimal captioning behavior. Evaluated on the MSR-VTT and MSVD dataset, we outperform recent state-of-the-art methods including a well-tuned SA-LSTM baseline by a significant margin, with shorter training schedules. Because of its simplicity and efficacy, we hope that our GLR could serve as a strong baseline for many video understanding tasks besides video captioning. Code will be available."
                    },
                    {
                        "title": "Towards Unbiased Label Distribution Learning for Facial Pose Estimation Using Anisotropic Spherical Gaussian",
                        "abstract": "Facial pose estimation refers to the task of predicting face orientation from a single RGB image. It is an important research topic with a wide range of applications in computer vision. Label distribution learning (LDL) based methods have been recently proposed for facial pose estimation, which achieve promising results. However, there are two major issues in existing LDL methods. First, the expectations of label distributions are biased, leading to a biased pose estimation. Second, fixed distribution parameters are applied for all learning samples, severely limiting the model capability. In this paper, we propose an Anisotropic Spherical Gaussian (ASG)-based LDL approach for facial pose estimation. In particular, our approach adopts the spherical Gaussian distribution on a unit sphere which constantly generates unbiased expectation. Meanwhile, we introduce a new loss function that allows the network to learn the distribution parameter for each learning sample flexibly. Extensive experimental results show that our method sets new state-of-the-art records on AFLW2000 and BIWI datasets."
                    },
                    {
                        "title": "Fall Detection from Audios with Audio Transformers",
                        "abstract": "Fall detection for the elderly is a well-researched problem with several proposed solutions, including wearable and non-wearable techniques. While the existing techniques have excellent detection rates, their adoption by the target population is lacking due to the need for wearing devices and user privacy concerns. Our paper provides a novel, non-wearable, non-intrusive, and scalable solution for fall detection, deployed on an autonomous mobile robot equipped with a microphone. The proposed method uses ambient sound input recorded in people's homes. We specifically target the bathroom environment as it is highly prone to falls and where existing techniques cannot be deployed without jeopardizing user privacy. The present work develops a solution based on a Transformer architecture that takes noisy sound input from bathrooms and classifies it into fall/no-fall class with an accuracy of 0.8673. Further, the proposed approach is extendable to other indoor environments, besides bathrooms and is suitable for deploying in elderly homes, hospitals, and rehabilitation facilities without requiring the user to wear any device or be constantly\"watched\"by the sensors."
                    },
                    {
                        "title": "A Fast PRM Planner for Car-like Vehicles",
                        "abstract": "We present a probabilistic roadmap method (PRM) for Ackermann vehicles by exploring the non-holonomic constraint of the Ackermann linakge. Similar to most PRM methods, the configuration space is first sampled in order to guarantee no-collision of the vehicle. Then, the sampled states are connected using proximity and our own algorithm for modeling the non-holonomic properties of the Ackermann. Finally, a path-searching algorithm is applied to find the shortest path from the start to goal state. We show that the path planner is successful in finding the shortest path while incorporating the non-holonomic constraints, and discuss some important computational properties of this method."
                    }
                ]
            },
            "f709a293-ca5d-432e-bad5-dc4b73a5e734": {
                "pk": "f709a293-ca5d-432e-bad5-dc4b73a5e734",
                "name": "Majid Rabbani",
                "collaborators": [
                    "Raghuveer M. Rao",
                    "S. Dianat",
                    "Prasanna Reddy Pulakurthi",
                    "Dongfang Liu",
                    "Celso M. De Melo",
                    "Cheng Han",
                    "Qifan Wang",
                    "Yi Fang",
                    "Qiang Guan",
                    "Lifu Huang"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Knowledge Distillation",
                    "Action Recognition",
                    "Text-Video Retrieval"
                ],
                "publications": [
                    {
                        "title": "Enhancing human action recognition with GAN-based data augmentation",
                        "abstract": "Deep Neural Networks (DNNs) have emerged as a powerful tool for human action recognition, yet their reliance on vast amounts of high-quality labeled data poses significant challenges. A promising alternative is to train the network on generated synthetic data. However, existing synthetic data generation pipelines require complex simulation environments. Our novel solution bypasses this requirement by employing Generative Adversarial Networks (GANs) to generate synthetic data from only a small existing real-world dataset. Our training pipeline extracts the motion from each training video and augments it across various subject appearances within the training set. This approach increases the diversity in both motion and subject representations, thus significantly enhancing the model's performance. A rigorous evaluation of the model's performance is presented under diverse scenarios, including ground and aerial views. Moreover, an insightful analysis of critical factors influencing human action recognition performance, such as gesture motion diversity and subject appearance, is presented."
                    },
                    {
                        "title": "AMD: Automatic Multi-step Distillation of Large-scale Vision Models",
                        "abstract": "Transformer-based architectures have become the de-facto standard models for diverse vision tasks owing to their superior performance. As the size of the models continues to scale up, model distillation becomes extremely important in various real applications, particularly on devices limited by computational resources. However, prevailing knowledge distillation methods exhibit diminished efficacy when confronted with a large capacity gap between the teacher and the student, e.g, 10x compression rate. In this paper, we present a novel approach named Automatic Multi-step Distillation (AMD) for large-scale vision model compression. In particular, our distillation process unfolds across multiple steps. Initially, the teacher undergoes distillation to form an intermediate teacher-assistant model, which is subsequently distilled further to the student. An efficient and effective optimization framework is introduced to automatically identify the optimal teacher-assistant that leads to the maximal student performance. We conduct extensive experiments on multiple image classification datasets, including CIFAR-10, CIFAR-100, and ImageNet. The findings consistently reveal that our approach outperforms several established baselines, paving a path for future knowledge distillation methods on large-scale vision models."
                    },
                    {
                        "title": "Enhancing GAN Performance Through Neural Architecture Search and Tensor Decomposition",
                        "abstract": "Generative Adversarial Networks (GANs) have emerged as a powerful tool for generating high-fidelity content. This paper presents a new training procedure that leverages Neural Architecture Search (NAS) to discover the optimal architecture for image generation while employing the Maximum Mean Discrepancy (MMD) repulsive loss for adversarial training. Moreover, the generator network is compressed using tensor decomposition to reduce its computational footprint and inference time while preserving its generative performance. Experimental results show improvements of 34% and 28% in the FID score on the CIFAR-10 and STL-10 datasets, respectively, with corresponding footprint reductions of 14\u00d7 and 31\u00d7 compared to the best FID score method reported in the literature. The implementation code is available at: https://github.com/PrasannaPulakurthi/MMD-AdversarialNAS."
                    },
                    {
                        "title": "Text Is MASS: Modeling as Stochastic Embedding for Text-Video Retrieval",
                        "abstract": "The increasing prevalence of video clips has sparked growing interest in text-video retrieval. Recent advances focus on establishing a joint embedding space for text and video, relying on consistent embedding representations to compute similarity. However, the text content in existing datasets is generally short and concise, making it hard to fully describe the redundant semantics of a video. Correspondingly, a single text embedding may be less expressive to capture the video embedding and empower the retrieval. In this study, we propose a new stochastic text modeling method T-MASS, i.e., text is modeled as a stochastic embedding, to enrich text embedding with a flexible and re-silient semantic range, yielding a text mass. To be specific, we introduce a similarity-aware radius module to adapt the scale of the text mass upon the given text-video pairs. Plus, we design and develop a support text regularization to further control the text mass during the training. The inference pipeline is also tailored to fully exploit the text mass for accurate retrieval. Empirical evidence suggests that T-MASS not only effectively attracts relevant text-video pairs while distancing irrelevant ones, but also enables the de-termination of precise text embeddings for relevant pairs. Our experimental results show a substantial improvement of T-MASS over baseline (3% ~ 6.3% by R@1). Also, T-MASS achieves state-of-the-art performance on five bench-mark datasets, including MSRVTT, LSMDC, DiDeMo, VA-TEX, and Charades. Code and models are available here."
                    }
                ]
            },
            "d8253aca-c049-4263-8520-6697de270571": {
                "pk": "d8253aca-c049-4263-8520-6697de270571",
                "name": "Sohail Dianat",
                "collaborators": [
                    "Raghuveer M. Rao",
                    "Majid Rabbani",
                    "Dongfang Liu",
                    "Cheng Han",
                    "Qifan Wang",
                    "Qiang Guan",
                    "Prasanna Reddy Pulakurthi",
                    "Guohao Sun",
                    "Zhiqiang Tao",
                    "R. Rao"
                ],
                "domain": [
                    "Computer Vision",
                    "Knowledge Distillation",
                    "Generative Adversarial Networks",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "AMD: Automatic Multi-step Distillation of Large-scale Vision Models",
                        "abstract": "Transformer-based architectures have become the de-facto standard models for diverse vision tasks owing to their superior performance. As the size of the models continues to scale up, model distillation becomes extremely important in various real applications, particularly on devices limited by computational resources. However, prevailing knowledge distillation methods exhibit diminished efficacy when confronted with a large capacity gap between the teacher and the student, e.g, 10x compression rate. In this paper, we present a novel approach named Automatic Multi-step Distillation (AMD) for large-scale vision model compression. In particular, our distillation process unfolds across multiple steps. Initially, the teacher undergoes distillation to form an intermediate teacher-assistant model, which is subsequently distilled further to the student. An efficient and effective optimization framework is introduced to automatically identify the optimal teacher-assistant that leads to the maximal student performance. We conduct extensive experiments on multiple image classification datasets, including CIFAR-10, CIFAR-100, and ImageNet. The findings consistently reveal that our approach outperforms several established baselines, paving a path for future knowledge distillation methods on large-scale vision models."
                    },
                    {
                        "title": "Enhancing GAN Performance Through Neural Architecture Search and Tensor Decomposition",
                        "abstract": "Generative Adversarial Networks (GANs) have emerged as a powerful tool for generating high-fidelity content. This paper presents a new training procedure that leverages Neural Architecture Search (NAS) to discover the optimal architecture for image generation while employing the Maximum Mean Discrepancy (MMD) repulsive loss for adversarial training. Moreover, the generator network is compressed using tensor decomposition to reduce its computational footprint and inference time while preserving its generative performance. Experimental results show improvements of 34% and 28% in the FID score on the CIFAR-10 and STL-10 datasets, respectively, with corresponding footprint reductions of 14\u00d7 and 31\u00d7 compared to the best FID score method reported in the literature. The implementation code is available at: https://github.com/PrasannaPulakurthi/MMD-AdversarialNAS."
                    },
                    {
                        "title": "Text Is MASS: Modeling as Stochastic Embedding for Text-Video Retrieval",
                        "abstract": "The increasing prevalence of video clips has sparked growing interest in text-video retrieval. Recent advances focus on establishing a joint embedding space for text and video, relying on consistent embedding representations to compute similarity. However, the text content in existing datasets is generally short and concise, making it hard to fully describe the redundant semantics of a video. Correspondingly, a single text embedding may be less expressive to capture the video embedding and empower the retrieval. In this study, we propose a new stochastic text modeling method T-MASS, i.e., text is modeled as a stochastic embedding, to enrich text embedding with a flexible and re-silient semantic range, yielding a text mass. To be specific, we introduce a similarity-aware radius module to adapt the scale of the text mass upon the given text-video pairs. Plus, we design and develop a support text regularization to further control the text mass during the training. The inference pipeline is also tailored to fully exploit the text mass for accurate retrieval. Empirical evidence suggests that T-MASS not only effectively attracts relevant text-video pairs while distancing irrelevant ones, but also enables the de-termination of precise text embeddings for relevant pairs. Our experimental results show a substantial improvement of T-MASS over baseline (3% ~ 6.3% by R@1). Also, T-MASS achieves state-of-the-art performance on five bench-mark datasets, including MSRVTT, LSMDC, DiDeMo, VA-TEX, and Charades. Code and models are available here."
                    },
                    {
                        "title": "Prototypical Transformer as Unified Motion Learners",
                        "abstract": "In this work, we introduce the Prototypical Transformer (ProtoFormer), a general and unified framework that approaches various motion tasks from a prototype perspective. ProtoFormer seamlessly integrates prototype learning with Transformer by thoughtfully considering motion dynamics, introducing two innovative designs. First, Cross-Attention Prototyping discovers prototypes based on signature motion patterns, providing transparency in understanding motion scenes. Second, Latent Synchronization guides feature representation learning via prototypes, effectively mitigating the problem of motion uncertainty. Empirical results demonstrate that our approach achieves competitive performance on popular motion tasks such as optical flow and scene depth. Furthermore, it exhibits generality across various downstream tasks, including object tracking and video stabilization."
                    },
                    {
                        "title": "Unsupervised domain adaptation using feature aligned maximum classifier discrepancy",
                        "abstract": "The maximum classifier discrepancy method has achieved great success in solving unsupervised domain adaptation tasks for image classification in recent years. Its basic structure consists of a feature generator and two classifiers that aim to maximize the classifier discrepancy while minimizing the generator discrepancy of the target samples. This method improves the performance of the existing adversarial training methods by employing task-specific classifiers that remove the ambiguity in classifying the target samples near the class boundaries. In this paper, we propose a modified network architecture and two training objectives to further boost the performance of the maximum classifier discrepancy method. The first training objective minimizes the feature level discrepancy and forces the generator to generate domain invariant features. This training objective is particularly beneficial when the source and the target domain distributions are vastly different. The second training objective that works at the mini-batch level aims at creating a uniform distribution of the target class predictions by maximizing the entropy of the expectation of the target class predictions. We show through extensive empirical evaluations that the proposed architecture and training objectives significantly improve the performance of the original algorithm. Furthermore, this method also outperforms the state-of-the-art techniques in most unsupervised domain adaptation tasks."
                    },
                    {
                        "title": "Advanced Linear Algebra for Engineers with Matlab",
                        "abstract": "Matrices, Matrix Algebra, and Elementary Matrix Operations Basic Concepts and Notation Matrix Algebra Elementary Row Operations Solution of System of Linear Equations Matrix Partitions Block Multiplication Inner, Outer, and Kronecker Products Determinants, Matrix Inversion and Solutions to Systems of Linear Equations Determinant of a Matrix Matrix Inversion Solution of Simultaneous Linear Equations Applications: Circuit Analysis Homogeneous Coordinates System Rank, Null Space and Invertibility of Matrices Special Matrices with Applications Derivatives and Gradients Linear Vector Spaces Linear Vector Space Span of a Set of Vectors Normed Vector Spaces Inner Product Spaces Orthogonality Matrix Factorization Eigenvalues and Eigenvectors Matrices as Linear Transformations Eigenvalues and Eigenvectors Matrix Diagonalization Special Matrices Singular Value Decomposition (SVD) Numerical Computation of Eigenvalues and Eigenvectors Properties of Eigenvalues and Eigenvectors of Different Classes of Matrices Applications Matrix Polynomials and Functions of Square Matrices Matrix Polynomials Cayley-Hamilton Theorem Functions of Matrices The State Space Modeling of Linear Continuous-time Systems State Space Representation of Discrete-time Systems Controllability of LTI Systems Observability of LTI Systems Introduction to Optimization Stationary Points of Functions of Several Variables Least-Square (LS) Technique Total Least-Squares (TLS) Eigen Filters Stationary Points with Equality Constraints Appendix A: The Laplace Transform Appendix B: The z-Transform Bibliography Index"
                    },
                    {
                        "title": "End-to-end system of license plate localization and recognition",
                        "abstract": "Abstract. An end-to-end license plate recognition system is proposed. It is composed of preprocessing, detection, segmentation, and character recognition to find and recognize plates from camera-based still images. The system utilizes connected component (CC) properties to quickly extract the license plate region. A two-stage CC filtering is utilized to address both shape and spatial relationship information to produce high precision and to recall values for detection. Floating peak and valleys of projection profiles are used to cut the license plates into individual characters. A turning function-based method is proposed to quickly and accurately recognize each character. It is further accelerated using curvature histogram-based support vector machine. The INFTY dataset is used to train the recognition system, and MediaLab license plate dataset is used for testing. The proposed system achieved 89.45% F-measure for detection and 87.33% accuracy for overall recognition rate which is comparable to current state-of-the-art systems."
                    },
                    {
                        "title": "Synthesis of intensity gradient and texture information for efficient three-dimensional segmentation of medical volumes",
                        "abstract": "Abstract. We propose a framework that efficiently employs intensity, gradient, and textural features for three-dimensional (3-D) segmentation of medical (MRI/CT) volumes. Our methodology commences by determining the magnitude of intensity variations across the input volume using a 3-D gradient detection scheme. The resultant gradient volume is utilized in a dynamic volume growing/formation process that is initiated in voxel locations with small gradient magnitudes and is concluded at sites with large gradient magnitudes, yielding a map comprising an initial set of partitions (or subvolumes). This partition map is combined with an entropy-based texture descriptor along with intensity and gradient attributes in a multivariate analysis-based volume merging procedure that fuses subvolumes with similar characteristics to yield a final/refined segmentation output. Additionally, a semiautomated version of the aforestated algorithm that allows a user to interactively segment a desired subvolume of interest as opposed to the entire volume is also discussed. Our approach was tested on several MRI and CT datasets and the results show favorable performance in comparison to the state-of-the-art ITK-SNAP technique."
                    },
                    {
                        "title": "Multispectral Image Denoising With Optimized Vector Bilateral Filter",
                        "abstract": "Vector bilateral filtering has been shown to provide good tradeoff between noise removal and edge degradation when applied to multispectral/hyperspectral image denoising. It has also been demonstrated to provide dynamic range enhancement of bands that have impaired signal to noise ratios (SNRs). Typical vector bilateral filtering described in the literature does not use parameters satisfying optimality criteria. We introduce an approach for selection of the parameters of a vector bilateral filter through an optimization procedure rather than by ad hoc means. The approach is based on posing the filtering problem as one of nonlinear estimation and minimization of the Stein's unbiased risk estimate of this nonlinear estimator. Along the way, we provide a plausibility argument through an analytical example as to why vector bilateral filtering outperforms band-wise 2D bilateral filtering in enhancing SNR. Experimental results show that the optimized vector bilateral filter provides improved denoising performance on multispectral images when compared with several other approaches."
                    },
                    {
                        "title": "Developing a linear model of RF power generators with pseudo random binary signals (PRBS)",
                        "abstract": "In this paper, we will present an approach developing a linear model of a radio frequency (RF) power generator by using pseudo random binary signals (PRBS). We will compare two linear models obtained respectively by the PRBS approach and a traditional modeling approach. The result shows that both approaches achieve a very similar model of the RF power generator. Moreover, it can be shown that the PRBS approach is easily implemented in FPGA (Field-Programmable Gate Array) and can be adapted for the on-line system identification."
                    }
                ]
            },
            "2ea6e3b4-9c4d-43ad-9b39-ce9c095c3368": {
                "pk": "2ea6e3b4-9c4d-43ad-9b39-ce9c095c3368",
                "name": "Raghuveer Rao",
                "collaborators": [
                    "Majid Rabbani",
                    "S. Dianat",
                    "Dongfang Liu",
                    "Prasanna Reddy Pulakurthi",
                    "Cheng Han",
                    "Qifan Wang",
                    "Qiang Guan",
                    "Guohao Sun",
                    "Zhiqiang Tao",
                    "Celso M. De Melo"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Knowledge Distillation",
                    "Video Retrieval",
                    "Transformer Models"
                ],
                "publications": [
                    {
                        "title": "Enhancing human action recognition with GAN-based data augmentation",
                        "abstract": "Deep Neural Networks (DNNs) have emerged as a powerful tool for human action recognition, yet their reliance on vast amounts of high-quality labeled data poses significant challenges. A promising alternative is to train the network on generated synthetic data. However, existing synthetic data generation pipelines require complex simulation environments. Our novel solution bypasses this requirement by employing Generative Adversarial Networks (GANs) to generate synthetic data from only a small existing real-world dataset. Our training pipeline extracts the motion from each training video and augments it across various subject appearances within the training set. This approach increases the diversity in both motion and subject representations, thus significantly enhancing the model's performance. A rigorous evaluation of the model's performance is presented under diverse scenarios, including ground and aerial views. Moreover, an insightful analysis of critical factors influencing human action recognition performance, such as gesture motion diversity and subject appearance, is presented."
                    },
                    {
                        "title": "AMD: Automatic Multi-step Distillation of Large-scale Vision Models",
                        "abstract": "Transformer-based architectures have become the de-facto standard models for diverse vision tasks owing to their superior performance. As the size of the models continues to scale up, model distillation becomes extremely important in various real applications, particularly on devices limited by computational resources. However, prevailing knowledge distillation methods exhibit diminished efficacy when confronted with a large capacity gap between the teacher and the student, e.g, 10x compression rate. In this paper, we present a novel approach named Automatic Multi-step Distillation (AMD) for large-scale vision model compression. In particular, our distillation process unfolds across multiple steps. Initially, the teacher undergoes distillation to form an intermediate teacher-assistant model, which is subsequently distilled further to the student. An efficient and effective optimization framework is introduced to automatically identify the optimal teacher-assistant that leads to the maximal student performance. We conduct extensive experiments on multiple image classification datasets, including CIFAR-10, CIFAR-100, and ImageNet. The findings consistently reveal that our approach outperforms several established baselines, paving a path for future knowledge distillation methods on large-scale vision models."
                    },
                    {
                        "title": "Enhancing GAN Performance Through Neural Architecture Search and Tensor Decomposition",
                        "abstract": "Generative Adversarial Networks (GANs) have emerged as a powerful tool for generating high-fidelity content. This paper presents a new training procedure that leverages Neural Architecture Search (NAS) to discover the optimal architecture for image generation while employing the Maximum Mean Discrepancy (MMD) repulsive loss for adversarial training. Moreover, the generator network is compressed using tensor decomposition to reduce its computational footprint and inference time while preserving its generative performance. Experimental results show improvements of 34% and 28% in the FID score on the CIFAR-10 and STL-10 datasets, respectively, with corresponding footprint reductions of 14\u00d7 and 31\u00d7 compared to the best FID score method reported in the literature. The implementation code is available at: https://github.com/PrasannaPulakurthi/MMD-AdversarialNAS."
                    },
                    {
                        "title": "Text Is MASS: Modeling as Stochastic Embedding for Text-Video Retrieval",
                        "abstract": "The increasing prevalence of video clips has sparked growing interest in text-video retrieval. Recent advances focus on establishing a joint embedding space for text and video, relying on consistent embedding representations to compute similarity. However, the text content in existing datasets is generally short and concise, making it hard to fully describe the redundant semantics of a video. Correspondingly, a single text embedding may be less expressive to capture the video embedding and empower the retrieval. In this study, we propose a new stochastic text modeling method T-MASS, i.e., text is modeled as a stochastic embedding, to enrich text embedding with a flexible and re-silient semantic range, yielding a text mass. To be specific, we introduce a similarity-aware radius module to adapt the scale of the text mass upon the given text-video pairs. Plus, we design and develop a support text regularization to further control the text mass during the training. The inference pipeline is also tailored to fully exploit the text mass for accurate retrieval. Empirical evidence suggests that T-MASS not only effectively attracts relevant text-video pairs while distancing irrelevant ones, but also enables the de-termination of precise text embeddings for relevant pairs. Our experimental results show a substantial improvement of T-MASS over baseline (3% ~ 6.3% by R@1). Also, T-MASS achieves state-of-the-art performance on five bench-mark datasets, including MSRVTT, LSMDC, DiDeMo, VA-TEX, and Charades. Code and models are available here."
                    },
                    {
                        "title": "Prototypical Transformer as Unified Motion Learners",
                        "abstract": "In this work, we introduce the Prototypical Transformer (ProtoFormer), a general and unified framework that approaches various motion tasks from a prototype perspective. ProtoFormer seamlessly integrates prototype learning with Transformer by thoughtfully considering motion dynamics, introducing two innovative designs. First, Cross-Attention Prototyping discovers prototypes based on signature motion patterns, providing transparency in understanding motion scenes. Second, Latent Synchronization guides feature representation learning via prototypes, effectively mitigating the problem of motion uncertainty. Empirical results demonstrate that our approach achieves competitive performance on popular motion tasks such as optical flow and scene depth. Furthermore, it exhibits generality across various downstream tasks, including object tracking and video stabilization."
                    }
                ]
            },
            "9edca9d0-29a7-4f30-9e24-df7bf04811ff": {
                "pk": "9edca9d0-29a7-4f30-9e24-df7bf04811ff",
                "name": "Ying Nian Wu",
                "collaborators": [
                    "Dongfang Liu",
                    "Chuan Liu",
                    "Chunshu Wu",
                    "Shihui Cao",
                    "Mingkai Chen",
                    "James Liang",
                    "Ang Li",
                    "Michael C. Huang",
                    "Chuang Ren",
                    "Tong Geng"
                ],
                "domain": [
                    "Artificial Intelligence",
                    "Nuclear Fusion",
                    "Machine Learning",
                    "Cognitive Systems"
                ],
                "publications": [
                    {
                        "title": "Diff-PIC: Revolutionizing Particle-In-Cell Nuclear Fusion Simulation with Diffusion Models",
                        "abstract": "The rapid development of AI highlights the pressing need for sustainable energy, a critical global challenge for decades. Nuclear fusion, generally seen as an ultimate solution, has been the focus of intensive research for nearly a century, with investments reaching hundreds of billions of dollars. Recent advancements in Inertial Confinement Fusion have drawn significant attention to fusion research, in which Laser-Plasma Interaction (LPI) is critical for ensuring fusion stability and efficiency. However, the complexity of LPI upon fusion ignition makes analytical approaches impractical, leaving researchers depending on extremely computation-demanding Particle-in-Cell (PIC) simulations to generate data, presenting a significant bottleneck to advancing fusion research. In response, this work introduces Diff-PIC, a novel framework that leverages conditional diffusion models as a computationally efficient alternative to PIC simulations for generating high-fidelity scientific LPI data. In this work, physical patterns captured by PIC simulations are distilled into diffusion models associated with two tailored enhancements: (1) To effectively capture the complex relationships between physical parameters and corresponding outcomes, the parameters are encoded in a physically-informed manner. (2) To further enhance efficiency while maintaining high fidelity and physical validity, the rectified flow technique is employed to transform our model into a one-step conditional diffusion model. Experimental results show that Diff-PIC achieves 16,200$\\times$ speedup compared to traditional PIC on a 100 picosecond simulation, with an average reduction in MAE / RMSE / FID of 59.21% / 57.15% / 39.46% with respect to two other SOTA data generation approaches."
                    },
                    {
                        "title": "Visual Agents as Fast and Slow Thinkers",
                        "abstract": "Achieving human-level intelligence requires refining cognitive distinctions between System 1 and System 2 thinking. While contemporary AI, driven by large language models, demonstrates human-like traits, it falls short of genuine cognition. Transitioning from structured benchmarks to real-world scenarios presents challenges for visual agents, often leading to inaccurate and overly confident responses. To address the challenge, we introduce FaST, which incorporates the Fast and Slow Thinking mechanism into visual agents. FaST employs a switch adapter to dynamically select between System 1/2 modes, tailoring the problem-solving approach to different task complexity. It tackles uncertain and unseen objects by adjusting model confidence and integrating new contextual data. With this novel design, we advocate a flexible system, hierarchical reasoning capabilities, and a transparent decision-making pipeline, all of which contribute to its ability to emulate human-like cognitive processes in visual intelligence. Empirical results demonstrate that FaST outperforms various well-known baselines, achieving 80.8% accuracy over VQA^{v2} for visual question answering and 48.7% GIoU score over ReasonSeg for reasoning segmentation, demonstrate FaST's superior performance. Extensive testing validates the efficacy and robustness of FaST's core components, showcasing its potential to advance the development of cognitive visual agents in AI systems. The code is available at ttps://github.com/GuangyanS/Sys2-LLaVA."
                    },
                    {
                        "title": "Inertial Confinement Fusion Forecasting via Large Language Models",
                        "abstract": "Controlled fusion energy is deemed pivotal for the advancement of human civilization. In this study, we introduce $\\textbf{LPI-LLM}$, a novel integration of Large Language Models (LLMs) with classical reservoir computing paradigms tailored to address a critical challenge, Laser-Plasma Instabilities ($\\texttt{LPI}$), in Inertial Confinement Fusion ($\\texttt{ICF}$). Our approach offers several key contributions: Firstly, we propose the $\\textit{LLM-anchored Reservoir}$, augmented with a $\\textit{Fusion-specific Prompt}$, enabling accurate forecasting of $\\texttt{LPI}$-generated-hot electron dynamics during implosion. Secondly, we develop $\\textit{Signal-Digesting Channels}$ to temporally and spatially describe the driver laser intensity across time, capturing the unique characteristics of $\\texttt{ICF}$ inputs. Lastly, we design the $\\textit{Confidence Scanner}$ to quantify the confidence level in forecasting, providing valuable insights for domain experts to design the $\\texttt{ICF}$ process. Extensive experiments demonstrate the superior performance of our method, achieving 1.90 CAE, 0.14 $\\texttt{top-1}$ MAE, and 0.11 $\\texttt{top-5}$ MAE in predicting Hard X-ray ($\\texttt{HXR}$) energies emitted by the hot electrons in $\\texttt{ICF}$ implosions, which presents state-of-the-art comparisons against concurrent best systems. Additionally, we present $\\textbf{LPI4AI}$, the first $\\texttt{LPI}$ benchmark based on physical experiments, aimed at fostering novel ideas in $\\texttt{LPI}$ research and enhancing the utility of LLMs in scientific exploration. Overall, our work strives to forge an innovative synergy between AI and $\\texttt{ICF}$ for advancing fusion energy."
                    }
                ]
            },
            "f52bc7ea-8afc-427b-85c9-2a884204060c": {
                "pk": "f52bc7ea-8afc-427b-85c9-2a884204060c",
                "name": "Dongfang Liu",
                "collaborators": [
                    "Qifan Wang",
                    "Cheng Han",
                    "Lifu Huang",
                    "Yiming Cui",
                    "Heng Fan",
                    "Fuli Feng",
                    "Zenglin Xu",
                    "Yi Fang",
                    "Qiang Guan",
                    "Yawen Lu"
                ],
                "domain": [
                    "Computer Vision",
                    "Knowledge Distillation",
                    "Multimodal Learning",
                    "Depth Estimation"
                ],
                "publications": [
                    {
                        "title": "AMD: Automatic Multi-step Distillation of Large-scale Vision Models",
                        "abstract": "Transformer-based architectures have become the de-facto standard models for diverse vision tasks owing to their superior performance. As the size of the models continues to scale up, model distillation becomes extremely important in various real applications, particularly on devices limited by computational resources. However, prevailing knowledge distillation methods exhibit diminished efficacy when confronted with a large capacity gap between the teacher and the student, e.g, 10x compression rate. In this paper, we present a novel approach named Automatic Multi-step Distillation (AMD) for large-scale vision model compression. In particular, our distillation process unfolds across multiple steps. Initially, the teacher undergoes distillation to form an intermediate teacher-assistant model, which is subsequently distilled further to the student. An efficient and effective optimization framework is introduced to automatically identify the optimal teacher-assistant that leads to the maximal student performance. We conduct extensive experiments on multiple image classification datasets, including CIFAR-10, CIFAR-100, and ImageNet. The findings consistently reveal that our approach outperforms several established baselines, paving a path for future knowledge distillation methods on large-scale vision models."
                    },
                    {
                        "title": "ProMotion: Prototypes as Motion Learners",
                        "abstract": "In this work, we introduce PRoMoTION, a unified proto-typical transformer-based framework engineered to model fundamental motion tasks. PRoMoTION offers a range of compelling attributes that set it apart from current task-specific paradigms. (1) We adopt a prototypical perspective, establishing a unified paradigm that harmonizes disparate motion learning approaches. This novel paradigm stream-lines the architectural design, enabling the simultaneous assimilation of diverse motion information. (2) We capitalize on a dual mechanism involving the feature denoiser and the prototypical learner to decipher the intricacies of motion. This approach effectively circumvents the pitfalls of ambiguity in pixel-wise feature matching, significantly bolstering the robustness of motion representation. (3)) We demon-strate a profound degree of transferability across distinct motion patterns. This inherent versatility reverberates robustly across a comprehensive spectrum of both 2D and 3D downstream tasks. Empirical results demonstrate that PRoMOTION outperforms various well-known specialized architectures, achieving 0.54 and 0.054 $AbsRel$ error on the Sintel and KITTI depth datasets, 1.04 and 2.01 average endpoint error on the clean and final pass of Sintel flow benchmark, and 4.30 F1-all error on the KITTI flow bench-mark. For its efficacy, we hope our work can catalyze a paradigm shift in universal models in computer vision."
                    },
                    {
                        "title": "EAVE: Efficient Product Attribute Value Extraction via Lightweight Sparse-layer Interaction",
                        "abstract": "Product attribute value extraction involves identifying the specific values associated with various attributes from a product profile. While existing methods often prioritize the development of effective models to improve extraction performance, there has been limited emphasis on extraction efficiency. However, in real-world scenarios, products are typically associated with multiple attributes, necessitating multiple extractions to obtain all corresponding values. In this work, we propose an Efficient product Attribute Value Extraction (EAVE) approach via lightweight sparse-layer interaction. Specifically, we employ a heavy encoder to separately encode the product context and attribute. The resulting non-interacting heavy representations of the context can be cached and reused for all attributes. Additionally, we introduce a light encoder to jointly encode the context and the attribute, facilitating lightweight interactions between them. To enrich the interaction within the lightweight encoder, we design a sparse-layer interaction module to fuse the non-interacting heavy representation into the lightweight encoder. Comprehensive evaluation on two benchmarks demonstrate that our method achieves significant efficiency gains with neutral or marginal loss in performance when the context is long and number of attributes is large. Our code is available \\href{https://anonymous.4open.science/r/EAVE-EA18}{here}."
                    },
                    {
                        "title": "Optical Flow as Spatial-Temporal Attention Learners.",
                        "abstract": "Optical flow is an indispensable building block for various important computer vision tasks, including motion estimation, object tracking, and disparity measurement. To date, the dominant methods are CNN-based, leaving plenty of room for improvement. In\u00a0this\u00a0work, we propose TransFlow, a transformer architecture for optical flow estimation. Compared to dominant CNN-based methods, TransFlow demonstrates three advantages. First, it provides more accurate correlation and trustworthy matching in flow estimation by utilizing spatial self-attention and cross-attention mechanisms between adjacent frames to effectively capture global dependencies; Second, it recovers more compromised information (e.g. occlusion and motion blur) in flow estimation through long-range temporal association in dynamic scenes; Third, it introduces a concise self-learning paradigm, eliminating the need for complex and laborious multi-stage pre-training procedures. The versatility and superiority of TransFlow extend seamlessly to 3D scene motion, yielding competitive outcomes in 3D scene flow estimation. Our approach attains state-of-the-art results on benchmark datasets such as Sintel and KITTI-15, while also exhibiting exceptional performance on downstream tasks, including video object detection using the ImageNet VID dataset, video frame interpolation using the GoPro dataset, and video stabilization using the DeepStab dataset. We believe that the effectiveness of TransFlow positions it as a flexible baseline for both optical flow and scene flow estimation, offering promising avenues for future research and development."
                    },
                    {
                        "title": "M2PT: Multimodal Prompt Tuning for Zero-shot Instruction Learning",
                        "abstract": "Multimodal Large Language Models (MLLMs) demonstrate remarkable performance across a wide range of domains, with increasing emphasis on enhancing their zero-shot generalization capabilities for unseen tasks across various modalities. Instruction tuning has emerged as an effective strategy for achieving zero-shot generalization by finetuning pretrained models on diverse multimodal tasks. As the scale of MLLMs continues to grow, parameter-efficient finetuning becomes increasingly critical. However, most existing parameter-efficient approaches focus only on single modalities and often overlook the multimodal characteristics during finetuning. In this work, we introduce a novel Multimodal Prompt Tuning (M$^2$PT) approach for efficient instruction tuning of MLLMs. M$^2$PT effectively integrates visual and textual prompts into the vision encoder and language processor respectively during finetuning, facilitating the extraction and alignment of features across modalities. Empirical results on various multimodal evaluation datasets demonstrate the superior performance of our approach compared to several state-of-the-art baselines. A comprehensive set of ablation studies validates the effectiveness of our prompt design and the efficiency of our approach."
                    },
                    {
                        "title": "Self-supervised Adversarial Training of Monocular Depth Estimation against Physical-World Attacks",
                        "abstract": "Monocular Depth Estimation (MDE) plays a vital role in applications such as autonomous driving. However, various attacks target MDE models, with physical attacks posing significant threats to system security. Traditional adversarial training methods, which require ground-truth labels, are not directly applicable to MDE models that lack ground-truth depth. Some self-supervised model hardening techniques (e.g., contrastive learning) overlook the domain knowledge of MDE, resulting in suboptimal performance. In this work, we introduce a novel self-supervised adversarial training approach for MDE models, leveraging view synthesis without the need for ground-truth depth. We enhance adversarial robustness against real-world attacks by incorporating L0-norm-bounded perturbation during training. We evaluate our method against supervised learning-based and contrastive learning-based approaches specifically designed for MDE. Our experiments with two representative MDE networks demonstrate improved robustness against various adversarial attacks, with minimal impact on benign performance. Our code: https://github.com/Bob-cheng/DepthModelHardening."
                    },
                    {
                        "title": "Facing the Elephant in the Room: Visual Prompt Tuning or Full Finetuning?",
                        "abstract": "As the scale of vision models continues to grow, the emergence of Visual Prompt Tuning (VPT) as a parameter-efficient transfer learning technique has gained attention due to its superior performance compared to traditional full-finetuning. However, the conditions favoring VPT (the ``when\") and the underlying rationale (the ``why\") remain unclear. In this paper, we conduct a comprehensive analysis across 19 distinct datasets and tasks. To understand the ``when\"aspect, we identify the scenarios where VPT proves favorable by two dimensions: task objectives and data distributions. We find that VPT is preferrable when there is 1) a substantial disparity between the original and the downstream task objectives (e.g., transitioning from classification to counting), or 2) a similarity in data distributions between the two tasks (e.g., both involve natural images). In exploring the ``why\"dimension, our results indicate VPT's success cannot be attributed solely to overfitting and optimization considerations. The unique way VPT preserves original features and adds parameters appears to be a pivotal factor. Our study provides insights into VPT's mechanisms, and offers guidance for its optimal utilization."
                    },
                    {
                        "title": "SSGA-Net: Stepwise Spatial Global-local Aggregation Networks for for Autonomous Driving",
                        "abstract": "Visual-based perception is the key module for autonomous driving. Among those visual perception tasks, video object detection is a primary yet challenging one because of feature degradation caused by fast motion or multiple poses. Current models usually aggregate features from the neighboring frames to enhance the object representations for the task heads to generate more accurate predictions. Though getting better performance, these methods rely on the information from the future frames and suffer from high computational complexity. Meanwhile, the aggregation process is not reconfigurable during the inference time. These issues make most of the existing models infeasible for online applications. To solve these problems, we introduce a stepwise spatial global-local aggregation network. Our proposed models mainly contain three parts: 1). Multi-stage stepwise network gradually refines the predictions and object representations from the previous stage; 2). Spatial global-local aggregation fuses the local information from the neighboring frames and global semantics from the current frame to eliminate the feature degradation; 3). Dynamic aggregation strategy stops the aggregation process early based on the refinement results to remove redundancy and improve efficiency. Extensive experiments on the ImageNet VID benchmark validate the effectiveness and efficiency of our proposed models."
                    },
                    {
                        "title": "Radiance Field Learners As UAV First-Person Viewers",
                        "abstract": "First-Person-View (FPV) holds immense potential for revolutionizing the trajectory of Unmanned Aerial Vehicles (UAVs), offering an exhilarating avenue for navigating complex building structures. Yet, traditional Neural Radiance Field (NeRF) methods face challenges such as sampling single points per iteration and requiring an extensive array of views for supervision. UAV videos exacerbate these issues with limited viewpoints and significant spatial scale variations, resulting in inadequate detail rendering across diverse scales. In response, we introduce FPV-NeRF, addressing these challenges through three key facets: (1) Temporal consistency. Leveraging spatio-temporal continuity ensures seamless coherence between frames; (2) Global structure. Incorporating various global features during point sampling preserves space integrity; (3) Local granularity. Employing a comprehensive framework and multi-resolution supervision for multi-scale scene feature representation tackles the intricacies of UAV video spatial scales. Additionally, due to the scarcity of publicly available FPV videos, we introduce an innovative view synthesis method using NeRF to generate FPV perspectives from UAV footage, enhancing spatial perception for drones. Our novel dataset spans diverse trajectories, from outdoor to indoor environments, in the UAV domain, differing significantly from traditional NeRF scenarios. Through extensive experiments encompassing both interior and exterior building structures, FPV-NeRF demonstrates a superior understanding of the UAV flying space, outperforming state-of-the-art methods in our curated UAV dataset. Explore our project page for further insights: https://fpv-nerf.github.io/."
                    },
                    {
                        "title": "Collaborative Multi-task Learning for Multi-Object Tracking and Segmentation",
                        "abstract": "The advancement of computer vision has pushed visual analysis tasks from still images to the video domain. In recent years, video instance segmentation, which aims at tracking and segment multiple objects in video frames, has drawn much attention for its potential applications in various emerging areas such as autonomous driving, intelligent transportation, and smart retail. In this article, we propose an effective framework for instance-level visual analysis on video frames, which can simultaneously conduct object detection, instance segmentation, and multi-object tracking. The core idea of our method is collaborative multi-task learning which is achieved by a novel structure, named associative connections among detection, segmentation, and tracking task heads in an end-to-end learnable CNN. These additional connections allow information propagation across multiple related tasks, so as to benefit these tasks simultaneously. We evaluate the proposed method extensively on KITTI MOTS and MOTS Challenge datasets and obtain quite encouraging results."
                    },
                    {
                        "title": "Efficient Multimodal Semantic Segmentation via Dual-Prompt Learning",
                        "abstract": "Multimodal (e.g., RGB-Depth/RGB-Thermal) fusion has shown great potential for improving semantic segmentation in complex scenes (e.g., indoor/low-light conditions). Existing approaches often fully fine-tune a dual-branch encoder-decoder framework with a complicated feature fusion strategy for achieving multimodal semantic segmentation, which is training-costly due to the massive parameter updates in feature extraction and fusion. To address this issue, we propose a surprisingly simple yet effective dual-prompt learning network (dubbed DPLNet) for training-efficient multimodal (e.g., RGB-D/T) semantic segmentation. The core of DPLNet is to directly adapt a frozen pre-trained RGB model to multimodal semantic segmentation, reducing parameter updates. For this purpose, we present two prompt learning modules, comprising multimodal prompt generator (MPG) and multimodal feature adapter (MFA). MPG works to fuse the features from different modalities in a compact manner and is inserted from shadow to deep stages to generate the multi-level multimodal prompts that are injected into the frozen backbone, while MPG adapts prompted multimodal features in the frozen backbone for better multimodal semantic segmentation. Since both the MPG and MFA are lightweight, only a few trainable parameters (3.88M, 4.4% of the pre-trained backbone parameters) are introduced for multimodal feature fusion and learning. Using a simple decoder (3.27M parameters), DPLNet achieves new state-of-the-art performance or is on a par with other complex approaches on four RGB-D/T semantic segmentation datasets while satisfying parameter efficiency. Moreover, we show that DPLNet is general and applicable to other multimodal tasks such as salient object detection and video semantic segmentation. Without special design, DPLNet outperforms many complicated models. Our code will be available at github.com/ShaohuaDong2021/DPLNet."
                    },
                    {
                        "title": "APrompt: Attention Prompt Tuning for Efficient Adaptation of Pre-trained Language Models",
                        "abstract": "With the continuous growth of large language models, the process of fine-tuning these models for new tasks has become increasingly parameter-intensive. Prompt tuning, a method that involves tuning a small set of soft prompts, has emerged as an effective and efficient approach for adapting large pre-trained language models. However, most existing prompt tuning approaches only introduce prompts at the input layer, limiting their performance and leaving large rooms for improvement. In this work, we propose a novel Attention Prompt tuning method, namely AP ROMPT , for efficient adaptation of pre-trained language models. We first demonstrate that existing prompt tuning can be considered as a special case of attention prompt tuning. We then formally introduce AP ROMPT , which incorporates query, key, and value prompts into the attention layer to guide the attention computation during fine-tuning. Experimental results on the SuperGLUE benchmark consistently demonstrate that our proposed approach outperforms state-of-the-art baselines and full fine-tuning method with pre-trained models at different scales. In addition, a comprehensive set of ablation studies validate the effectiveness of the prompt design, as well as the efficiency of our approach."
                    }
                ]
            }
        }
    },
    "2405.15223": {
        "paper_data": {
            "title": "iVideoGPT: Interactive VideoGPTs are Scalable World Models",
            "url": "http://arxiv.org/abs/2405.15223v2",
            "arxiv_id": "2405.15223",
            "authors": [
                "Jialong Wu",
                "Shaofeng Yin",
                "Ningya Feng",
                "Xu He",
                "Dong Li",
                "Jianye Hao",
                "Mingsheng Long"
            ],
            "abstract": "World models empower model-based agents to interactively explore, reason, and plan within imagined environments for real-world decision-making. However, the high demand for interactivity poses challenges in harnessing recent advancements in video generative models for developing world models at scale. This work introduces Interactive VideoGPT (iVideoGPT), a scalable autoregressive transformer framework that integrates multimodal signals--visual observations, actions, and rewards--into a sequence of tokens, facilitating an interactive experience of agents via next-token prediction. iVideoGPT features a novel compressive tokenization technique that efficiently discretizes high-dimensional visual observations. Leveraging its scalable architecture, we are able to pre-train iVideoGPT on millions of human and robotic manipulation trajectories, establishing a versatile foundation that is adaptable to serve as interactive world models for a wide range of downstream tasks. These include action-conditioned video prediction, visual planning, and model-based reinforcement learning, where iVideoGPT achieves competitive performance compared with state-of-the-art methods. Our work advances the development of interactive general world models, bridging the gap between generative video models and practical model-based reinforcement learning applications.",
            "introduction": "   1 Introduction  Recent years have witnessed remarkable advancements in generative models of multimodal contents, including text [1], audio [8], and images [21], with video generation now emerging as a new frontier [10]. A particularly significant application of these generative video models, learned in an unsupervised way on diverse Internet-scale data, is to construct predictive world models [51, 27] at scale. These world models are expected to accumulate commonsense knowledge about how the world works, enabling the prediction of potential future outcomes (e.g., visual observations and reward signals) based on the actions of agents. By leveraging these world models, agents employing model-based reinforcement learning (RL) can imagine, reason, and plan inside world models [19, 28], thus acquiring new skills more safely and efficiently with a handful of trials in the real world.   Despite the fundamental connection, significant gaps remain between the development of generative models for video generation and world models for agent learning. One primary challenge is achieving the best of both interactivity and scalability. In the realm of model-based RL, world models predominantly utilize recurrent network architecture. This design naturally allows the transition of observations or latent states conditioned on actions in each step, facilitating interactive behavior learning [28, 77, 32]. However, these models mostly focus on games or simulated environments with visually simple data and have limited capability to model complex, in-the-wild data at scale. In contrast, Internet-scale video generative models [35, 6, 10] can synthesize realistic long videos that are controllable via text descriptions [104] or future action sequences [97]. Although such models allow for high-level, long-horizon planning [18], their trajectory-level interactivity does not provide sufficient granularity for agents to efficiently learn precise behavior as basic skills step-by-step.   In this work, we explore the development of world models that are both interactive and scalable within a GPT-like autoregressive transformer framework [86, 72]. Pioneering efforts have been made recently through diffusion models [98] and masked generative models [11]. Nevertheless, utilizing autoregressive transformers for world models offers distinct advantages such as seamless integration with the established Large Language Model (LLM) ecosystem [105] and greater flexibility in handling diverse conditions without the need for specific architectural modifications like adapter modules [74, 102]. We present Interactive VideoGPT (iVideoGPT), a scalable world model architecture that incorporates multimodal signals, including visual observations, actions, and rewards, in an interactively autoregressive manner. Unlike multimodal LLMs that discretize visual observations into tokens frame-by-frame using image tokenizers [53], a key innovation of iVideoGPT is to learn compressive tokenization for each observation conditioned on rich contextual observations, achieving an asymptotic 16\u00d7\\times\u00d7 reduction in token sequence length. We highlight that more compact video tokenization could not only facilitate more efficient training and generation but also enhance video quality. This is achieved by decoupling context from dynamics, allowing the model to focus on predicting the motion of objects while maintaining temporal consistency within the scene [95].   We demonstrate a series of practical applications of iVideoGPT for visual robotic manipulation, as illustrated in Figure 1. Mirroring the two-phase approach popularized by Large Language Models (LLMs), our methodology involves pre-training followed by domain-specific adaptation. During pre-training, iVideoGPT",
            "references": [
                {
                    "title": "Octo: An Open-Source Generalist Robot Policy",
                    "abstract": "Large policies pretrained on diverse robot datasets have the potential to transform robotic learning: instead of training new policies from scratch, such generalist robot policies may be finetuned with only a little in-domain data, yet generalize broadly. However, to be widely applicable across a range of robotic learning scenarios, environments, and tasks, such policies need to handle diverse sensors and action spaces, accommodate a variety of commonly used robotic platforms, and finetune readily and efficiently to new domains. In this work, we aim to lay the groundwork for developing open-source, widely applicable, generalist policies for robotic manipulation. As a first step, we introduce Octo, a large transformer-based policy trained on 800k trajectories from the Open X-Embodiment dataset, the largest robot manipulation dataset to date. It can be instructed via language commands or goal images and can be effectively finetuned to robot setups with new sensory inputs and action spaces within a few hours on standard consumer GPUs. In experiments across 9 robotic platforms, we demonstrate that Octo serves as a versatile policy initialization that can be effectively finetuned to new observation and action spaces. We also perform detailed ablations of design decisions for the Octo model, from architecture to training data, to guide future research on building generalist robot models."
                },
                {
                    "title": "Genie: Generative Interactive Environments",
                    "abstract": "We introduce Genie, the first generative interactive environment trained in an unsupervised manner from unlabelled Internet videos. The model can be prompted to generate an endless variety of action-controllable virtual worlds described through text, synthetic images, photographs, and even sketches. At 11B parameters, Genie can be considered a foundation world model. It is comprised of a spatiotemporal video tokenizer, an autoregressive dynamics model, and a simple and scalable latent action model. Genie enables users to act in the generated environments on a frame-by-frame basis despite training without any ground-truth action labels or other domain-specific requirements typically found in the world model literature. Further the resulting learned latent action space facilitates training agents to imitate behaviors from unseen videos, opening the path for training generalist agents of the future."
                },
                {
                    "title": "World Model on Million-Length Video And Language With Blockwise RingAttention",
                    "abstract": "Current language models fall short in understanding aspects of the world not easily described in words, and struggle with complex, long-form tasks. Video sequences offer valuable temporal information absent in language and static images, making them attractive for joint modeling with language. Such models could develop a understanding of both human textual knowledge and the physical world, enabling broader AI capabilities for assisting humans. However, learning from millions of tokens of video and language sequences poses challenges due to memory constraints, computational complexity, and limited datasets. To address these challenges, we curate a large dataset of diverse videos and books, utilize the Blockwise RingAttention technique to scalably train on long sequences, and gradually increase context size from 4K to 1M tokens. This paper makes the following contributions: (a) Largest context size neural network: We train one of the largest context size transformers on long video and language sequences, setting new benchmarks in difficult retrieval tasks and long video understanding. (b) Solutions for overcoming vision-language training challenges, including using masked sequence packing for mixing different sequence lengths, loss weighting to balance language and vision, and model-generated QA dataset for long sequence chat. (c) A highly-optimized implementation with RingAttention, Blockwise Transformers, masked sequence packing, and other key features for training on millions-length multimodal sequences. (d) Fully open-sourced a family of 7B parameter models capable of processing long text documents (LWM-Text, LWM-Text-Chat) and videos (LWM, LWM-Chat) of over 1M tokens. This work paves the way for training on massive datasets of long video and language to develop understanding of both human knowledge and the multimodal world, and broader capabilities."
                },
                {
                    "title": "Video-LaVIT: Unified Video-Language Pre-training with Decoupled Visual-Motional Tokenization",
                    "abstract": "In light of recent advances in multimodal Large Language Models (LLMs), there is increasing attention to scaling them from image-text data to more informative real-world videos. Compared to static images, video poses unique challenges for effective large-scale pre-training due to the modeling of its spatiotemporal dynamics. In this paper, we address such limitations in video-language pre-training with an efficient video decomposition that represents each video as keyframes and temporal motions. These are then adapted to an LLM using well-designed tokenizers that discretize visual and temporal information as a few tokens, thus enabling unified generative pre-training of videos, images, and text. At inference, the generated tokens from the LLM are carefully recovered to the original continuous pixel space to create various video content. Our proposed framework is both capable of comprehending and generating image and video content, as demonstrated by its competitive performance across 13 multimodal benchmarks in image and video understanding and generation. Our code and models are available at https://video-lavit.github.io."
                },
                {
                    "title": "aMUSEd: An Open MUSE Reproduction",
                    "abstract": "We present aMUSEd, an open-source, lightweight masked image model (MIM) for text-to-image generation based on MUSE. With 10 percent of MUSE's parameters, aMUSEd is focused on fast image generation. We believe MIM is under-explored compared to latent diffusion, the prevailing approach for text-to-image generation. Compared to latent diffusion, MIM requires fewer inference steps and is more interpretable. Additionally, MIM can be fine-tuned to learn additional styles with only a single image. We hope to encourage further exploration of MIM by demonstrating its effectiveness on large-scale text-to-image generation and releasing reproducible training code. We also release checkpoints for two models which directly produce images at 256x256 and 512x512 resolutions."
                },
                {
                    "title": "VideoPoet: A Large Language Model for Zero-Shot Video Generation",
                    "abstract": "We present VideoPoet, a language model capable of synthesizing high-quality video, with matching audio, from a large variety of conditioning signals. VideoPoet employs a decoder-only transformer architecture that processes multimodal inputs -- including images, videos, text, and audio. The training protocol follows that of Large Language Models (LLMs), consisting of two stages: pretraining and task-specific adaptation. During pretraining, VideoPoet incorporates a mixture of multimodal generative objectives within an autoregressive Transformer framework. The pretrained LLM serves as a foundation that can be adapted for a range of video generation tasks. We present empirical results demonstrating the model's state-of-the-art capabilities in zero-shot video generation, specifically highlighting VideoPoet's ability to generate high-fidelity motions. Project page: http://sites.research.google/videopoet/"
                },
                {
                    "title": "PlayFusion: Skill Acquisition via Diffusion from Language-Annotated Play",
                    "abstract": "Learning from unstructured and uncurated data has become the dominant paradigm for generative approaches in language and vision. Such unstructured and unguided behavior data, commonly known as play, is also easier to collect in robotics but much more difficult to learn from due to its inherently multimodal, noisy, and suboptimal nature. In this paper, we study this problem of learning goal-directed skill policies from unstructured play data which is labeled with language in hindsight. Specifically, we leverage advances in diffusion models to learn a multi-task diffusion model to extract robotic skills from play data. Using a conditional denoising diffusion process in the space of states and actions, we can gracefully handle the complexity and multimodality of play data and generate diverse and interesting robot behaviors. To make diffusion models more useful for skill learning, we encourage robotic agents to acquire a vocabulary of skills by introducing discrete bottlenecks into the conditional behavior generation process. In our experiments, we demonstrate the effectiveness of our approach across a wide variety of environments in both simulation and the real world. Results visualizations and videos at https://play-fusion.github.io"
                },
                {
                    "title": "Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets",
                    "abstract": "We present Stable Video Diffusion - a latent video diffusion model for high-resolution, state-of-the-art text-to-video and image-to-video generation. Recently, latent diffusion models trained for 2D image synthesis have been turned into generative video models by inserting temporal layers and finetuning them on small, high-quality video datasets. However, training methods in the literature vary widely, and the field has yet to agree on a unified strategy for curating video data. In this paper, we identify and evaluate three different stages for successful training of video LDMs: text-to-image pretraining, video pretraining, and high-quality video finetuning. Furthermore, we demonstrate the necessity of a well-curated pretraining dataset for generating high-quality videos and present a systematic curation process to train a strong base model, including captioning and filtering strategies. We then explore the impact of finetuning our base model on high-quality data and train a text-to-video model that is competitive with closed-source video generation. We also show that our base model provides a powerful motion representation for downstream tasks such as image-to-video generation and adaptability to camera motion-specific LoRA modules. Finally, we demonstrate that our model provides a strong multi-view 3D-prior and can serve as a base to finetune a multi-view diffusion model that jointly generates multiple views of objects in a feedforward fashion, outperforming image-based methods at a fraction of their compute budget. We release code and model weights at https://github.com/Stability-AI/generative-models ."
                },
                {
                    "title": "Finetuning Offline World Models in the Real World",
                    "abstract": "Reinforcement Learning (RL) is notoriously data-inefficient, which makes training on a real robot difficult. While model-based RL algorithms (world models) improve data-efficiency to some extent, they still require hours or days of interaction to learn skills. Recently, offline RL has been proposed as a framework for training RL policies on pre-existing datasets without any online interaction. However, constraining an algorithm to a fixed dataset induces a state-action distribution shift between training and inference, and limits its applicability to new tasks. In this work, we seek to get the best of both worlds: we consider the problem of pretraining a world model with offline data collected on a real robot, and then finetuning the model on online data collected by planning with the learned model. To mitigate extrapolation errors during online interaction, we propose to regularize the planner at test-time by balancing estimated returns and (epistemic) model uncertainty. We evaluate our method on a variety of visuo-motor control tasks in simulation and on a real robot, and find that our method enables few-shot finetuning to seen and unseen tasks even when offline data is limited. Videos, code, and data are available at https://yunhaifeng.com/FOWM ."
                },
                {
                    "title": "Video Language Planning",
                    "abstract": "We are interested in enabling visual planning for complex long-horizon tasks in the space of generated videos and language, leveraging recent advances in large generative models pretrained on Internet-scale data. To this end, we present video language planning (VLP), an algorithm that consists of a tree search procedure, where we train (i) vision-language models to serve as both policies and value functions, and (ii) text-to-video models as dynamics models. VLP takes as input a long-horizon task instruction and current image observation, and outputs a long video plan that provides detailed multimodal (video and language) specifications that describe how to complete the final task. VLP scales with increasing computation budget where more computation time results in improved video plans, and is able to synthesize long-horizon video plans across different robotics domains: from multi-object rearrangement, to multi-camera bi-arm dexterous manipulation. Generated video plans can be translated into real robot actions via goal-conditioned policies, conditioned on each intermediate frame of the generated video. Experiments show that VLP substantially improves long-horizon task success rates compared to prior methods on both simulated and real robots (across 3 hardware platforms)."
                },
                {
                    "title": "Open X-Embodiment: Robotic Learning Datasets and RT-X Models",
                    "abstract": "Large, high-capacity models trained on diverse datasets have shown remarkable successes on efficiently tackling downstream applications. In domains from NLP to Computer Vision, this has led to a consolidation of pretrained models, with general pretrained backbones serving as a starting point for many applications. Can such a consolidation happen in robotics? Conventionally, robotic learning methods train a separate model for every application, every robot, and even every environment. Can we instead train generalist X-robot policy that can be adapted efficiently to new robots, tasks, and environments? In this paper, we provide datasets in standardized data formats and models to make it possible to explore this possibility in the context of robotic manipulation, alongside experimental results that provide an example of effective X-robot policies. We assemble a dataset from 22 different robots collected through a collaboration between 21 institutions, demonstrating 527 skills (160266 tasks). We show that a high-capacity model trained on this data, which we call RT-X, exhibits positive transfer and improves the capabilities of multiple robots by leveraging experience from other platforms. More details can be found on the project website https://robotics-transformer-x.github.io."
                },
                {
                    "title": "Learning Interactive Real-World Simulators",
                    "abstract": "Generative models trained on internet data have revolutionized how text, image, and video content can be created. Perhaps the next milestone for generative models is to simulate realistic experience in response to actions taken by humans, robots, and other interactive agents. Applications of a real-world simulator range from controllable content creation in games and movies, to training embodied agents purely in simulation that can be directly deployed in the real world. We explore the possibility of learning a universal simulator (UniSim) of real-world interaction through generative modeling. We first make the important observation that natural datasets available for learning a real-world simulator are often rich along different dimensions (e.g., abundant objects in image data, densely sampled actions in robotics data, and diverse movements in navigation data). With careful orchestration of diverse datasets, each providing a different aspect of the overall experience, we can simulate the visual outcome of both high-level instructions such as\"open the drawer\"and low-level controls from otherwise static scenes and objects. We use the simulator to train both high-level vision-language policies and low-level reinforcement learning policies, each of which can be deployed in the real world in zero shot after training purely in simulation. We also show that other types of intelligence such as video captioning models can benefit from training with simulated experience, opening up even wider applications. Video demos can be found at https://universal-simulator.github.io."
                },
                {
                    "title": "HarmonyDream: Task Harmonization Inside World Models",
                    "abstract": "Model-based reinforcement learning (MBRL) holds the promise of sample-efficient learning by utilizing a world model, which models how the environment works and typically encompasses components for two tasks: observation modeling and reward modeling. In this paper, through a dedicated empirical investigation, we gain a deeper understanding of the role each task plays in world models and uncover the overlooked potential of sample-efficient MBRL by mitigating the domination of either observation or reward modeling. Our key insight is that while prevalent approaches of explicit MBRL attempt to restore abundant details of the environment via observation models, it is difficult due to the environment's complexity and limited model capacity. On the other hand, reward models, while dominating implicit MBRL and adept at learning compact task-centric dynamics, are inadequate for sample-efficient learning without richer learning signals. Motivated by these insights and discoveries, we propose a simple yet effective approach, HarmonyDream, which automatically adjusts loss coefficients to maintain task harmonization, i.e. a dynamic equilibrium between the two tasks in world model learning. Our experiments show that the base MBRL method equipped with HarmonyDream gains 10%-69% absolute performance boosts on visual robotic tasks and sets a new state-of-the-art result on the Atari 100K benchmark. Code is available at https://github.com/thuml/HarmonyDream."
                },
                {
                    "title": "D$^3$Fields: Dynamic 3D Descriptor Fields for Zero-Shot Generalizable Rearrangement",
                    "abstract": "Scene representation is a crucial design choice in robotic manipulation systems. An ideal representation is expected to be 3D, dynamic, and semantic to meet the demands of diverse manipulation tasks. However, previous works often lack all three properties simultaneously. In this work, we introduce D$^3$Fields -- dynamic 3D descriptor fields. These fields are implicit 3D representations that take in 3D points and output semantic features and instance masks. They can also capture the dynamics of the underlying 3D environments. Specifically, we project arbitrary 3D points in the workspace onto multi-view 2D visual observations and interpolate features derived from visual foundational models. The resulting fused descriptor fields allow for flexible goal specifications using 2D images with varied contexts, styles, and instances. To evaluate the effectiveness of these descriptor fields, we apply our representation to rearrangement tasks in a zero-shot manner. Through extensive evaluation in real worlds and simulations, we demonstrate that D$^3$Fields are effective for zero-shot generalizable rearrangement tasks. We also compare D$^3$Fields with state-of-the-art implicit 3D representations and show significant improvements in effectiveness and efficiency."
                },
                {
                    "title": "MUTEX: Learning Unified Policies from Multimodal Task Specifications",
                    "abstract": "Humans use different modalities, such as speech, text, images, videos, etc., to communicate their intent and goals with teammates. For robots to become better assistants, we aim to endow them with the ability to follow instructions and understand tasks specified by their human partners. Most robotic policy learning methods have focused on one single modality of task specification while ignoring the rich cross-modal information. We present MUTEX, a unified approach to policy learning from multimodal task specifications. It trains a transformer-based architecture to facilitate cross-modal reasoning, combining masked modeling and cross-modal matching objectives in a two-stage training procedure. After training, MUTEX can follow a task specification in any of the six learned modalities (video demonstrations, goal images, text goal descriptions, text instructions, speech goal descriptions, and speech instructions) or a combination of them. We systematically evaluate the benefits of MUTEX in a newly designed dataset with 100 tasks in simulation and 50 tasks in the real world, annotated with multiple instances of task specifications in different modalities, and observe improved performance over methods trained specifically for any single modality. More information at https://ut-austin-rpl.github.io/MUTEX/"
                },
                {
                    "title": "Pre-and Post-Contact Policy Decomposition for Non-Prehensile Manipulation with Zero-Shot Sim-To-Real Transfer",
                    "abstract": "We present a system for non-prehensile manipulation that require a significant number of contact mode transitions and the use of environmental contacts to successfully manipulate an object to a target location. Our method is based on deep reinforcement learning which, unlike state-of-the-art planning algorithms, does not require apriori knowledge of the physical parameters of the object or environment such as friction coefficients or centers of mass. The planning time is reduced to the simple feed-forward prediction time on a neural network. We propose a computational structure, action space design, and curriculum learning scheme that facilitates efficient exploration and sim-to-real transfer. In challenging real-world non-prehensile manipulation tasks, we show that our method can generalize over different objects, and succeed even for novel objects not seen during training. Project website: https://sites.google.com/view/nonprenehsile-decomposition"
                },
                {
                    "title": "BridgeData V2: A Dataset for Robot Learning at Scale",
                    "abstract": "We introduce BridgeData V2, a large and diverse dataset of robotic manipulation behaviors designed to facilitate research on scalable robot learning. BridgeData V2 contains 60,096 trajectories collected across 24 environments on a publicly available low-cost robot. BridgeData V2 provides extensive task and environment variability, leading to skills that can generalize across environments, domains, and institutions, making the dataset a useful resource for a broad range of researchers. Additionally, the dataset is compatible with a wide variety of open-vocabulary, multi-task learning methods conditioned on goal images or natural language instructions. In our experiments, we train 6 state-of-the-art imitation learning and offline reinforcement learning methods on our dataset, and find that they succeed on a suite of tasks requiring varying amounts of generalization. We also demonstrate that the performance of these methods improves with more data and higher capacity models, and that training on a greater variety of skills leads to improved generalization. By publicly sharing BridgeData V2 and our pre-trained models, we aim to accelerate research in scalable robot learning methods. Project page at https://rail-berkeley.github.io/bridgedata"
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Structured World Models from Human Videos",
                    "abstract": "We tackle the problem of learning complex, general behaviors directly in the real world. We propose an approach for robots to efficiently learn manipulation skills using only a handful of real-world interaction trajectories from many different settings. Inspired by the success of learning from large-scale datasets in the fields of computer vision and natural language, our belief is that in order to efficiently learn, a robot must be able to leverage internet-scale, human video data. Humans interact with the world in many interesting ways, which can allow a robot to not only build an understanding of useful actions and affordances but also how these actions affect the world for manipulation. Our approach builds a structured, human-centric action space grounded in visual affordances learned from human videos. Further, we train a world model on human videos and fine-tune on a small amount of robot interaction data without any task supervision. We show that this approach of affordance-space world models enables different robots to learn various manipulation skills in complex settings, in under 30 minutes of interaction. Videos can be found at https://human-world-model.github.io"
                },
                {
                    "title": "HYDRA: Hybrid Robot Actions for Imitation Learning",
                    "abstract": "Imitation Learning (IL) is a sample efficient paradigm for robot learning using expert demonstrations. However, policies learned through IL suffer from state distribution shift at test time, due to compounding errors in action prediction which lead to previously unseen states. Choosing an action representation for the policy that minimizes this distribution shift is critical in imitation learning. Prior work propose using temporal action abstractions to reduce compounding errors, but they often sacrifice policy dexterity or require domain-specific knowledge. To address these trade-offs, we introduce HYDRA, a method that leverages a hybrid action space with two levels of action abstractions: sparse high-level waypoints and dense low-level actions. HYDRA dynamically switches between action abstractions at test time to enable both coarse and fine-grained control of a robot. In addition, HYDRA employs action relabeling to increase the consistency of actions in the dataset, further reducing distribution shift. HYDRA outperforms prior imitation learning methods by 30-40% on seven challenging simulation and real world environments, involving long-horizon tasks in the real world like making coffee and toasting bread. Videos are found on our website: https://tinyurl.com/3mc6793z"
                },
                {
                    "title": "Pre-training Contextualized World Models with In-the-wild Videos for Reinforcement Learning",
                    "abstract": "Unsupervised pre-training methods utilizing large and diverse datasets have achieved tremendous success across a range of domains. Recent work has investigated such unsupervised pre-training methods for model-based reinforcement learning (MBRL) but is limited to domain-specific or simulated data. In this paper, we study the problem of pre-training world models with abundant in-the-wild videos for efficient learning of downstream visual control tasks. However, in-the-wild videos are complicated with various contextual factors, such as intricate backgrounds and textured appearance, which precludes a world model from extracting shared world knowledge to generalize better. To tackle this issue, we introduce Contextualized World Models (ContextWM) that explicitly separate context and dynamics modeling to overcome the complexity and diversity of in-the-wild videos and facilitate knowledge transfer between distinct scenes. Specifically, a contextualized extension of the latent dynamics model is elaborately realized by incorporating a context encoder to retain contextual information and empower the image decoder, which encourages the latent dynamics model to concentrate on essential temporal variations. Our experiments show that in-the-wild video pre-training equipped with ContextWM can significantly improve the sample efficiency of MBRL in various domains, including robotic manipulation, locomotion, and autonomous driving. Code is available at this repository: https://github.com/thuml/ContextWM."
                },
                {
                    "title": "Train Offline, Test Online: A Real Robot Learning Benchmark",
                    "abstract": "Three challenges limit the progress of robot learning research: robots are expensive (few labs can participate), everyone uses different robots (findings do not generalize across labs), and we lack internet-scale robotics data. We take on these challenges via a new benchmark: Train Offline, Test Online (TOTO). TOTO provides remote users with access to shared robots for evaluating methods on common tasks and an open-source dataset of these tasks for offline training. Its manipulation task suite requires challenging generalization to unseen objects, positions, and lighting. We present initial results on TOTO comparing five pretrained visual representations and four offline policy learning baselines, remotely contributed by five institutions. The real promise of TOTO, however, lies in the future: we release the benchmark for additional submissions from any user, enabling easy, direct comparison to several methods without the need to obtain hardware or collect data."
                },
                {
                    "title": "Guiding Reinforcement Learning with Shared Control Templates",
                    "abstract": "Purposeful interaction with objects usually requires certain constraints to be respected, e.g. keeping a bottle upright to avoid spilling. In reinforcement learning, such constraints are typically encoded in the reward function. As a consequence, constraints can only be learned by violating them. This often precludes learning on the physical robot, as it may take many trials to learn the constraints, and the necessity to violate them during the trial-and-error learning may be unsafe. We have serendipitously discovered that constraint representations for shared control - in particular Shared Control Templates (SCTs) - are ideally suited for safely guiding RL. Representing constraints explicitly, rather than implicitly in the reward function, also simplifies the design of the reward function. The main advantage of the approach is safer, faster learning without constraint violations (even with sparse reward functions). We demonstrate this in a pouring task in simulation and on a real robot, where learning the task requires only 65 episodes in 16 minutes."
                },
                {
                    "title": "ControlVideo: Training-free Controllable Text-to-Video Generation",
                    "abstract": "Text-driven diffusion models have unlocked unprecedented abilities in image generation, whereas their video counterpart still lags behind due to the excessive training cost of temporal modeling. Besides the training burden, the generated videos also suffer from appearance inconsistency and structural flickers, especially in long video synthesis. To address these challenges, we design a \\emph{training-free} framework called \\textbf{ControlVideo} to enable natural and efficient text-to-video generation. ControlVideo, adapted from ControlNet, leverages coarsely structural consistency from input motion sequences, and introduces three modules to improve video generation. Firstly, to ensure appearance coherence between frames, ControlVideo adds fully cross-frame interaction in self-attention modules. Secondly, to mitigate the flicker effect, it introduces an interleaved-frame smoother that employs frame interpolation on alternated frames. Finally, to produce long videos efficiently, it utilizes a hierarchical sampler that separately synthesizes each short clip with holistic coherency. Empowered with these modules, ControlVideo outperforms the state-of-the-arts on extensive motion-prompt pairs quantitatively and qualitatively. Notably, thanks to the efficient designs, it generates both short and long videos within several minutes using one NVIDIA 2080Ti. Code is available at https://github.com/YBYBZhang/ControlVideo."
                },
                {
                    "title": "FurnitureBench: Reproducible Real-World Benchmark for Long-Horizon Complex Manipulation",
                    "abstract": "Reinforcement learning (RL), imitation learning (IL), and task and motion planning (TAMP) have demonstrated impressive performance across various robotic manipulation tasks. However, these approaches have been limited to learning simple behaviors in current real-world manipulation benchmarks, such as pushing or pick-and-place. To enable more complex, long-horizon behaviors of an autonomous robot, we propose to focus on real-world furniture assembly, a complex, long-horizon robot manipulation task that requires addressing many current robotic manipulation challenges to solve. We present FurnitureBench, a reproducible real-world furniture assembly benchmark aimed at providing a low barrier for entry and being easily reproducible, so that researchers across the world can reliably test their algorithms and compare them against prior work. For ease of use, we provide 200+ hours of pre-collected data (5000+ demonstrations), 3D printable furniture models, a robotic environment setup guide, and systematic task initialization. Furthermore, we provide FurnitureSim, a fast and realistic simulator of FurnitureBench. We benchmark the performance of offline RL and IL algorithms on our assembly tasks and demonstrate the need to improve such algorithms to be able to solve our tasks in the real world, providing ample opportunities for future research."
                },
                {
                    "title": "A Control-Centric Benchmark for Video Prediction",
                    "abstract": "Video is a promising source of knowledge for embodied agents to learn models of the world's dynamics. Large deep networks have become increasingly effective at modeling complex video data in a self-supervised manner, as evaluated by metrics based on human perceptual similarity or pixel-wise comparison. However, it remains unclear whether current metrics are accurate indicators of performance on downstream tasks. We find empirically that for planning robotic manipulation, existing metrics can be unreliable at predicting execution success. To address this, we propose a benchmark for action-conditioned video prediction in the form of a control benchmark that evaluates a given model for simulated robotic manipulation through sampling-based planning. Our benchmark, Video Prediction for Visual Planning ($VP^2$), includes simulated environments with 11 task categories and 310 task instance definitions, a full planning implementation, and training datasets containing scripted interaction trajectories for each task category. A central design goal of our benchmark is to expose a simple interface -- a single forward prediction call -- so it is straightforward to evaluate almost any action-conditioned video prediction model. We then leverage our benchmark to study the effects of scaling model size, quantity of training data, and model ensembling by analyzing five highly-performant video prediction models, finding that while scale can improve perceptual quality when modeling visually diverse settings, other attributes such as uncertainty awareness can also aid planning performance."
                },
                {
                    "title": "Align Your Latents: High-Resolution Video Synthesis with Latent Diffusion Models",
                    "abstract": "Latent Diffusion Models (LDMs) enable high-quality image synthesis while avoiding excessive compute demands by training a diffusion model in a compressed lower-dimensional latent space. Here, we apply the LDM paradigm to high-resolution video generation, a particularly resource-intensive task. We first pre-train an LDM on images only; then, we turn the image generator into a video generator by introducing a temporal dimension to the latent space diffusion model and finetuning on encoded image sequences, i.e., videos. Similarly, we temporally align diffusion model upsamplers, turning them into temporally consistent video super resolution models. We focus on two relevant real-world applications: Simulation of in-the-wild driving data and creative content creation with text-to-video modeling. In particular, we validate our Video LDM on real driving videos of resolution $512 \\times 1024$, achieving state-of-the-art performance. Furthermore, our approach can easily leverage off-the-shelf pretrained image LDMs, as we only need to train a temporal alignment model in that case. Doing so, we turn the publicly available, state-of-the-art text-to-image LDM Stable Diffusion into an efficient and expressive text-to-video model with resolution up to $1280 \\times 2048$. We show that the temporal layers trained in this way generalize to different finetuned text-to-image LDMs. Utilizing this property, we show the first results for personalized text-to-video generation, opening exciting directions for future content creation. Project page: https://nv-tlabs.github.io/VideoLDM/"
                },
                {
                    "title": "Affordances from Human Videos as a Versatile Representation for Robotics",
                    "abstract": "Building a robot that can understand and learn to interact by watching humans has inspired several vision problems. However, despite some successful results on static datasets, it remains unclear how current models can be used on a robot directly. In this paper, we aim to bridge this gap by leveraging videos of human interactions in an environment centric manner. Utilizing internet videos of human behavior, we train a visual affordance model that estimates where and how in the scene a human is likely to interact. The structure of these behavioral affordances directly enables the robot to perform many complex tasks. We show how to seamlessly integrate our affordance model with four robot learning paradigms including offline imitation learning, exploration, goal-conditioned learning, and action parameterization for reinforcement learning. We show the efficacy of our approach, which we call Vision-Robotics Bridge (VRB) across 4 real world environments, over 10 different tasks, and 2 robotic platforms operating in the wild."
                },
                {
                    "title": "A Survey of Large Language Models",
                    "abstract": "Language is essentially a complex, intricate system of human expressions governed by grammatical rules. It poses a significant challenge to develop capable AI algorithms for comprehending and grasping a language. As a major approach, language modeling has been widely studied for language understanding and generation in the past two decades, evolving from statistical language models to neural language models. Recently, pre-trained language models (PLMs) have been proposed by pre-training Transformer models over large-scale corpora, showing strong capabilities in solving various NLP tasks. Since researchers have found that model scaling can lead to performance improvement, they further study the scaling effect by increasing the model size to an even larger size. Interestingly, when the parameter scale exceeds a certain level, these enlarged language models not only achieve a significant performance improvement but also show some special abilities that are not present in small-scale language models. To discriminate the difference in parameter scale, the research community has coined the term large language models (LLM) for the PLMs of significant size. Recently, the research on LLMs has been largely advanced by both academia and industry, and a remarkable progress is the launch of ChatGPT, which has attracted widespread attention from society. The technical evolution of LLMs has been making an important impact on the entire AI community, which would revolutionize the way how we develop and use AI algorithms. In this survey, we review the recent advances of LLMs by introducing the background, key findings, and mainstream techniques. In particular, we focus on four major aspects of LLMs, namely pre-training, adaptation tuning, utilization, and capacity evaluation. Besides, we also summarize the available resources for developing LLMs and discuss the remaining issues for future directions."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Diffusion Policy: Visuomotor Policy Learning via Action Diffusion",
                    "abstract": "This paper introduces Diffusion Policy, a new way of generating robot behavior by representing a robot\u2019s visuomotor policy as a conditional denoising diffusion process. We benchmark Diffusion Policy across 15 different tasks from 4 different robot manipulation benchmarks and find that it consistently outperforms existing state-of-the-art robot learning methods with an average improvement of 46.9%. Diffusion Policy learns the gradient of the action-distribution score function and iteratively optimizes with respect to this gradient field during inference via a series of stochastic Langevin dynamics steps. We find that the diffusion formulation yields powerful advantages when used for robot policies, including gracefully handling multimodal action distributions, being suitable for high-dimensional action spaces, and exhibiting impressive training stability. To fully unlock the potential of diffusion models for visuomotor policy learning on physical robots, this paper presents a set of key technical contributions including the incorporation of receding horizon control, visual conditioning, and the time-series diffusion transformer. We hope this work will help motivate a new generation of policy learning techniques that are able to leverage the powerful generative modeling capabilities of diffusion models. Code, data, and training details are available (diffusion-policy.cs.columbia.edu)."
                },
                {
                    "title": "Adding Conditional Control to Text-to-Image Diffusion Models",
                    "abstract": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models."
                },
                {
                    "title": "Mastering Diverse Domains through World Models",
                    "abstract": "Developing a general algorithm that learns to solve tasks across a wide range of applications has been a fundamental challenge in artificial intelligence. Although current reinforcement learning algorithms can be readily applied to tasks similar to what they have been developed for, configuring them for new application domains requires significant human expertise and experimentation. We present DreamerV3, a general algorithm that outperforms specialized methods across over 150 diverse tasks, with a single configuration. Dreamer learns a model of the environment and improves its behavior by imagining future scenarios. Robustness techniques based on normalization, balancing, and transformations enable stable learning across domains. Applied out of the box, Dreamer is the first algorithm to collect diamonds in Minecraft from scratch without human data or curricula. This achievement has been posed as a significant challenge in artificial intelligence that requires exploring farsighted strategies from pixels and sparse rewards in an open world. Our work allows solving challenging control problems without extensive experimentation, making reinforcement learning broadly applicable."
                },
                {
                    "title": "RT-1: Robotics Transformer for Real-World Control at Scale",
                    "abstract": "By transferring knowledge from large, diverse, task-agnostic datasets, modern machine learning models can solve specific downstream tasks either zero-shot or with small task-specific datasets to a high level of performance. While this capability has been demonstrated in other fields such as computer vision, natural language processing or speech recognition, it remains to be shown in robotics, where the generalization capabilities of the models are particularly critical due to the difficulty of collecting real-world robotic data. We argue that one of the keys to the success of such general robotic models lies with open-ended task-agnostic training, combined with high-capacity architectures that can absorb all of the diverse, robotic data. In this paper, we present a model class, dubbed Robotics Transformer, that exhibits promising scalable model properties. We verify our conclusions in a study of different model classes and their ability to generalize as a function of the data size, model size, and data diversity based on a large-scale data collection on real robots performing real-world tasks. The project's website and videos can be found at robotics-transformer1.github.io"
                },
                {
                    "title": "MAGVIT: Masked Generative Video Transformer",
                    "abstract": "We introduce the MAsked Generative VIdeo Transformer, MAGVIT, to tackle various video synthesis tasks with a single model. We introduce a 3D tokenizer to quantize a video into spatial-temporal visual tokens and propose an embedding method for masked video token modeling to facilitate multi-task learning. We conduct extensive experiments to demonstrate the quality, efficiency, and flexibility of MAGVIT. Our experiments show that (i) MAGVIT performs favorably against state-of-the-art approaches and establishes the best-published FVD on three video generation benchmarks, including the challenging Kinetics-600. (ii) MAGVIT outperforms existing methods in inference time by two orders of magnitude against diffusion models and by 60x against autoregressive models. (iii) A single MAGVIT model supports ten diverse generation tasks and generalizes across videos from different visual domains. The source code and trained models will be released to the public at https://magvit.cs.cmu.edu."
                },
                {
                    "title": "Modularity through Attention: Efficient Training and Transfer of Language-Conditioned Policies for Robot Manipulation",
                    "abstract": "Language-conditioned policies allow robots to interpret and execute human instructions. Learning such policies requires a substantial investment with regards to time and compute resources. Still, the resulting controllers are highly device-specific and cannot easily be transferred to a robot with different morphology, capability, appearance or dynamics. In this paper, we propose a sample-efficient approach for training language-conditioned manipulation policies that allows for rapid transfer across different types of robots. By introducing a novel method, namely Hierarchical Modularity, and adopting supervised attention across multiple sub-modules, we bridge the divide between modular and end-to-end learning and enable the reuse of functional building blocks. In both simulated and real world robot manipulation experiments, we demonstrate that our method outperforms the current state-of-the-art methods and can transfer policies across 4 different robots in a sample-efficient manner. Finally, we show that the functionality of learned sub-modules is maintained beyond the training process and can be used to introspect the robot decision-making process. Code is available at https://github.com/ir-lab/ModAttn."
                },
                {
                    "title": "VIOLA: Imitation Learning for Vision-Based Manipulation with Object Proposal Priors",
                    "abstract": "We introduce VIOLA, an object-centric imitation learning approach to learning closed-loop visuomotor policies for robot manipulation. Our approach constructs object-centric representations based on general object proposals from a pre-trained vision model. VIOLA uses a transformer-based policy to reason over these representations and attend to the task-relevant visual factors for action prediction. Such object-based structural priors improve deep imitation learning algorithm's robustness against object variations and environmental perturbations. We quantitatively evaluate VIOLA in simulation and on real robots. VIOLA outperforms the state-of-the-art imitation learning methods by $45.8\\%$ in success rate. It has also been deployed successfully on a physical robot to solve challenging long-horizon tasks, such as dining table arrangement and coffee making. More videos and model details can be found in supplementary material and the project website: https://ut-austin-rpl.github.io/VIOLA ."
                },
                {
                    "title": "Learning and Retrieval from Prior Data for Skill-based Imitation Learning",
                    "abstract": "Imitation learning offers a promising path for robots to learn general-purpose behaviors, but traditionally has exhibited limited scalability due to high data supervision requirements and brittle generalization. Inspired by recent advances in multi-task imitation learning, we investigate the use of prior data from previous tasks to facilitate learning novel tasks in a robust, data-efficient manner. To make effective use of the prior data, the robot must internalize knowledge from past experiences and contextualize this knowledge in novel tasks. To that end, we develop a skill-based imitation learning framework that extracts temporally extended sensorimotor skills from prior data and subsequently learns a policy for the target task that invokes these learned skills. We identify several key design choices that significantly improve performance on novel tasks, namely representation learning objectives to enable more predictable skill representations and a retrieval-based data augmentation mechanism to increase the scope of supervision for policy training. On a collection of simulated and real-world manipulation domains, we demonstrate that our method significantly outperforms existing imitation learning and offline reinforcement learning approaches. Videos and code are available at https://ut-austin-rpl.github.io/sailor"
                },
                {
                    "title": "From Play to Policy: Conditional Behavior Generation from Uncurated Robot Data",
                    "abstract": "While large-scale sequence modeling from offline data has led to impressive performance gains in natural language and image generation, directly translating such ideas to robotics has been challenging. One critical reason for this is that uncurated robot demonstration data, i.e. play data, collected from non-expert human demonstrators are often noisy, diverse, and distributionally multi-modal. This makes extracting useful, task-centric behaviors from such data a difficult generative modeling problem. In this work, we present Conditional Behavior Transformers (C-BeT), a method that combines the multi-modal generation ability of Behavior Transformer with future-conditioned goal specification. On a suite of simulated benchmark tasks, we find that C-BeT improves upon prior state-of-the-art work in learning from play data by an average of 45.7%. Further, we demonstrate for the first time that useful task-centric behaviors can be learned on a real-world robot purely from play data without any task labels or reward information. Robot videos are best viewed on our project website: https://play-to-policy.github.io"
                },
                {
                    "title": "Interactive Language: Talking to Robots in Real Time",
                    "abstract": "We present a framework for building interactive, real-time, natural language-instructable robots in the real world, and we open source related assets (dataset, environment, benchmark, and policies). Trained with behavioral cloning on a dataset of hundreds of thousands of language-annotated trajectories, a produced policy can proficiently execute an order of magnitude more commands than previous works: specifically we estimate a 93.5% success rate on a set of 87,000 unique natural language strings specifying raw end-to-end visuo-linguo-motor skills in the real world. We find that the same policy is capable of being guided by a human via real-time language to address a wide range of precise long-horizon rearrangement goals, e.g.\"make a smiley face out of blocks\". The dataset we release comprises nearly 600,000 language-labeled trajectories, an order of magnitude larger than prior available datasets. We hope the demonstrated results and associated assets enable further advancement of helpful, capable, natural-language-interactable robots. See videos at https://interactive-language.github.io."
                },
                {
                    "title": "Imagen Video: High Definition Video Generation with Diffusion Models",
                    "abstract": "We present Imagen Video, a text-conditional video generation system based on a cascade of video diffusion models. Given a text prompt, Imagen Video generates high definition videos using a base video generation model and a sequence of interleaved spatial and temporal video super-resolution models. We describe how we scale up the system as a high definition text-to-video model including design decisions such as the choice of fully-convolutional temporal and spatial super-resolution models at certain resolutions, and the choice of the v-parameterization of diffusion models. In addition, we confirm and transfer findings from previous work on diffusion-based image generation to the video generation setting. Finally, we apply progressive distillation to our video models with classifier-free guidance for fast, high quality sampling. We find Imagen Video not only capable of generating videos of high fidelity, but also having a high degree of controllability and world knowledge, including the ability to generate diverse videos and text animations in various artistic styles and with 3D object understanding. See https://imagen.research.google/video/ for samples."
                },
                {
                    "title": "Grounding Language with Visual Affordances over Unstructured Data",
                    "abstract": "Recent works have shown that Large Language Models (LLMs) can be applied to ground natural language to a wide variety of robot skills. However, in practice, learning multi-task, language-conditioned robotic skills typically requires large-scale data collection and frequent human intervention to reset the environment or help correcting the current policies. In this work, we propose a novel approach to efficiently learn general-purpose language-conditioned robot skills from unstructured, offline and reset-free data in the real world by exploiting a self-supervised visuo-lingual affordance model, which requires annotating as little as 1% of the total data with language. We evaluate our method in extensive experiments both in simulated and real-world robotic tasks, achieving state-of-the-art performance on the challenging CALVIN benchmark and learning over 25 distinct visuomotor manipulation tasks with a single policy in the real world. We find that when paired with LLMs to break down abstract natural language instructions into subgoals via few-shot prompting, our method is capable of completing long-horizon, multi-tier tasks in the real world, while requiring an order of magnitude less data than previous approaches. Code and videos are available at http://hulc2.cs.uni-freiburg.de."
                },
                {
                    "title": "Latent Plans for Task-Agnostic Offline Reinforcement Learning",
                    "abstract": "Everyday tasks of long-horizon and comprising a sequence of multiple implicit subtasks still impose a major challenge in offline robot control. While a number of prior methods aimed to address this setting with variants of imitation and offline reinforcement learning, the learned behavior is typically narrow and often struggles to reach configurable long-horizon goals. As both paradigms have complementary strengths and weaknesses, we propose a novel hierarchical approach that combines the strengths of both methods to learn task-agnostic long-horizon policies from high-dimensional camera observations. Concretely, we combine a low-level policy that learns latent skills via imitation learning and a high-level policy learned from offline reinforcement learning for skill-chaining the latent behavior priors. Experiments in various simulated and real robot control tasks show that our formulation enables producing previously unseen combinations of skills to reach temporally extended goals by\"stitching\"together latent skills through goal chaining with an order-of-magnitude improvement in performance upon state-of-the-art baselines. We even learn one multi-task visuomotor policy for 25 distinct manipulation tasks in the real world which outperforms both imitation learning and offline reinforcement learning techniques."
                },
                {
                    "title": "AudioLM: A Language Modeling Approach to Audio Generation",
                    "abstract": "We introduce AudioLM, a framework for high-quality audio generation with long-term consistency. AudioLM maps the input audio to a sequence of discrete tokens and casts audio generation as a language modeling task in this representation space. We show how existing audio tokenizers provide different trade-offs between reconstruction quality and long-term structure, and we propose a hybrid tokenization scheme to achieve both objectives. Namely, we leverage the discretized activations of a masked language model pre-trained on audio to capture long-term structure and the discrete codes produced by a neural audio codec to achieve high-quality synthesis. By training on large corpora of raw audio waveforms, AudioLM learns to generate natural and coherent continuations given short prompts. When trained on speech, and without any transcript or annotation, AudioLM generates syntactically and semantically plausible speech continuations while also maintaining speaker identity and prosody for unseen speakers. Furthermore, we demonstrate how our approach extends beyond speech by generating coherent piano music continuations, despite being trained without any symbolic representation of music."
                },
                {
                    "title": "Transformers are Sample Efficient World Models",
                    "abstract": "Deep reinforcement learning agents are notoriously sample inefficient, which considerably limits their application to real-world problems. Recently, many model-based methods have been designed to address this issue, with learning in the imagination of a world model being one of the most prominent approaches. However, while virtually unlimited interaction with a simulated environment sounds appealing, the world model has to be accurate over extended periods of time. Motivated by the success of Transformers in sequence modeling tasks, we introduce IRIS, a data-efficient agent that learns in a world model composed of a discrete autoencoder and an autoregressive Transformer. With the equivalent of only two hours of gameplay in the Atari 100k benchmark, IRIS achieves a mean human normalized score of 1.046, and outperforms humans on 10 out of 26 games, setting a new state of the art for methods without lookahead search. To foster future research on Transformers and world models for sample-efficient reinforcement learning, we release our code and models at https://github.com/eloialonso/iris."
                },
                {
                    "title": "MaskViT: Masked Visual Pre-Training for Video Prediction",
                    "abstract": "The ability to predict future visual observations conditioned on past observations and motor commands can enable embodied agents to plan solutions to a variety of tasks in complex environments. This work shows that we can create good video prediction models by pre-training transformers via masked visual modeling. Our approach, named MaskViT, is based on two simple design decisions. First, for memory and training efficiency, we use two types of window attention: spatial and spatiotemporal. Second, during training, we mask a variable percentage of tokens instead of a fixed mask ratio. For inference, MaskViT generates all tokens via iterative refinement where we incrementally decrease the masking ratio following a mask scheduling function. On several datasets we demonstrate that MaskViT outperforms prior works in video prediction, is parameter efficient, and can generate high-resolution videos (256x256). Further, we demonstrate the benefits of inference speedup (up to 512x) due to iterative decoding by using MaskViT for planning on a real robot. Our work suggests that we can endow embodied agents with powerful predictive models by leveraging the general framework of masked visual modeling with minimal domain knowledge."
                },
                {
                    "title": "MCVD: Masked Conditional Video Diffusion for Prediction, Generation, and Interpolation",
                    "abstract": "Video prediction is a challenging task. The quality of video frames from current state-of-the-art (SOTA) generative models tends to be poor and generalization beyond the training data is difficult. Furthermore, existing prediction frameworks are typically not capable of simultaneously handling other video-related tasks such as unconditional generation or interpolation. In this work, we devise a general-purpose framework called Masked Conditional Video Diffusion (MCVD) for all of these video synthesis tasks using a probabilistic conditional score-based denoising diffusion model, conditioned on past and/or future frames. We train the model in a manner where we randomly and independently mask all the past frames or all the future frames. This novel but straightforward setup allows us to train a single model that is capable of executing a broad range of video tasks, specifically: future/past prediction -- when only future/past frames are masked; unconditional generation -- when both past and future frames are masked; and interpolation -- when neither past nor future frames are masked. Our experiments show that this approach can generate high-quality frames for diverse types of videos. Our MCVD models are built from simple non-recurrent 2D-convolutional architectures, conditioning on blocks of frames and generating blocks of frames. We generate videos of arbitrary lengths autoregressively in a block-wise manner. Our approach yields SOTA results across standard video prediction and interpolation benchmarks, with computation times for training models measured in 1-12 days using $\\le$ 4 GPUs. Project page: https://mask-cond-video-diffusion.github.io ; Code : https://github.com/voletiv/mcvd-pytorch"
                },
                {
                    "title": "Video Diffusion Models",
                    "abstract": "Generating temporally coherent high fidelity video is an important milestone in generative modeling research. We make progress towards this milestone by proposing a diffusion model for video generation that shows very promising initial results. Our model is a natural extension of the standard image diffusion architecture, and it enables jointly training from image and video data, which we find to reduce the variance of minibatch gradients and speed up optimization. To generate long and higher resolution videos we introduce a new conditional sampling technique for spatial and temporal video extension that performs better than previously proposed methods. We present the first results on a large text-conditioned video generation task, as well as state-of-the-art results on established benchmarks for video prediction and unconditional video generation. Supplementary material is available at https://video-diffusion.github.io/"
                },
                {
                    "title": "Reinforcement Learning with Action-Free Pre-Training from Videos",
                    "abstract": "Recent unsupervised pre-training methods have shown to be effective on language and vision domains by learning useful representations for multiple downstream tasks. In this paper, we investigate if such unsupervised pre-training methods can also be effective for vision-based reinforcement learning (RL). To this end, we introduce a framework that learns representations useful for understanding the dynamics via generative pre-training on videos. Our framework consists of two phases: we pre-train an action-free latent video prediction model, and then utilize the pre-trained representations for efficiently learning action-conditional world models on unseen environments. To incorporate additional action inputs during fine-tuning, we introduce a new architecture that stacks an action-conditional latent prediction model on top of the pre-trained action-free prediction model. Moreover, for better exploration, we propose a video-based intrinsic bonus that leverages pre-trained representations. We demonstrate that our framework significantly improves both final performances and sample-efficiency of vision-based RL in a variety of manipulation and locomotion tasks. Code is available at https://github.com/younggyoseo/apv."
                },
                {
                    "title": "Temporal Difference Learning for Model Predictive Control",
                    "abstract": "Data-driven model predictive control has two key advantages over model-free methods: a potential for improved sample efficiency through model learning, and better performance as computational budget for planning increases. However, it is both costly to plan over long horizons and challenging to obtain an accurate model of the environment. In this work, we combine the strengths of model-free and model-based methods. We use a learned task-oriented latent dynamics model for local trajectory optimization over a short horizon, and use a learned terminal value function to estimate long-term return, both of which are learned jointly by temporal difference learning. Our method, TD-MPC, achieves superior sample efficiency and asymptotic performance over prior work on both state and image-based continuous control tasks from DMControl and Meta-World. Code and video results are available at https://nicklashansen.github.io/td-mpc."
                },
                {
                    "title": "BC-Z: Zero-Shot Task Generalization with Robotic Imitation Learning",
                    "abstract": "In this paper, we study the problem of enabling a vision-based robotic manipulation system to generalize to novel tasks, a long-standing challenge in robot learning. We approach the challenge from an imitation learning perspective, aiming to study how scaling and broadening the data collected can facilitate such generalization. To that end, we develop an interactive and flexible imitation learning system that can learn from both demonstrations and interventions and can be conditioned on different forms of information that convey the task, including pre-trained embeddings of natural language or videos of humans performing the task. When scaling data collection on a real robot to more than 100 distinct tasks, we find that this system can perform 24 unseen manipulation tasks with an average success rate of 44%, without any robot demonstrations for those tasks."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Ego4D: Around the World in 3,000 Hours of Egocentric Video",
                    "abstract": "We introduce Ego4D, a massive-scale egocentric video dataset and benchmark suite. It offers 3,670 hours of dailylife activity video spanning hundreds of scenarios (household, outdoor, workplace, leisure, etc.) captured by 931 unique camera wearers from 74 worldwide locations and 9 different countries. The approach to collection is designed to uphold rigorous privacy and ethics standards, with consenting participants and robust de-identification procedures where relevant. Ego4D dramatically expands the volume of diverse egocentric video footage publicly available to the research community. Portions of the video are accompanied by audio, 3D meshes of the environment, eye gaze, stereo, and/or synchronized videos from multiple egocentric cameras at the same event. Furthermore, we present a host of new benchmark challenges centered around understanding the first-person visual experience in the past (querying an episodic memory), present (analyzing hand-object manipulation, audio-visual conversation, and social interactions), and future (forecasting activities). By publicly sharing this massive annotated dataset and benchmark suite, we aim to push the frontier of first-person perception. Project page: https://ego4d-data.org/"
                },
                {
                    "title": "Bottom-Up Skill Discovery From Unsegmented Demonstrations for Long-Horizon Robot Manipulation",
                    "abstract": "We tackle real-world long-horizon robot manipulation tasks through skill discovery. We present a bottom-up approach to learning a library of reusable skills from unsegmented demonstrations and use these skills to synthesize prolonged robot behaviors. Our method starts with constructing a hierarchical task structure from each demonstration through agglomerative clustering. From the task structures of multi-task demonstrations, we identify skills based on the recurring patterns and train goal-conditioned sensorimotor policies with hierarchical imitation learning. Finally, we train a meta controller to compose these skills to solve long-horizon manipulation tasks. The entire model can be trained on a small set of human demonstrations collected within 30 minutes without further annotations, making it amendable to real-world deployment. We systematically evaluated our method in simulation environments and on a real robot. Our method has shown superior performance over state-of-the-art imitation learning methods in multi-stage manipulation tasks. Furthermore, skills discovered from multi-task demonstrations boost the average task success by 8% compared to those discovered from individual tasks."
                },
                {
                    "title": "Mastering Visual Continuous Control: Improved Data-Augmented Reinforcement Learning",
                    "abstract": "We present DrQ-v2, a model-free reinforcement learning (RL) algorithm for visual continuous control. DrQ-v2 builds on DrQ, an off-policy actor-critic approach that uses data augmentation to learn directly from pixels. We introduce several improvements that yield state-of-the-art results on the DeepMind Control Suite. Notably, DrQ-v2 is able to solve complex humanoid locomotion tasks directly from pixel observations, previously unattained by model-free RL. DrQ-v2 is conceptually simple, easy to implement, and provides significantly better computational footprint compared to prior work, with the majority of tasks taking just 8 hours to train on a single GPU. Finally, we publicly release DrQ-v2's implementation to provide RL practitioners with a strong and computationally efficient baseline."
                },
                {
                    "title": "CCVS: Context-aware Controllable Video Synthesis",
                    "abstract": "This presentation introduces a self-supervised learning approach to the synthesis of new video clips from old ones, with several new key elements for improved spatial resolution and realism: It conditions the synthesis process on contextual information for temporal continuity and ancillary information for fine control. The prediction model is doubly autoregressive, in the latent space of an autoencoder for forecasting, and in image space for updating contextual information, which is also used to enforce spatio-temporal consistency through a learnable optical flow module. Adversarial training of the autoencoder in the appearance and temporal domains is used to further improve the realism of its output. A quantizer inserted between the encoder and the transformer in charge of forecasting future frames in latent space (and its inverse inserted between the transformer and the decoder) adds even more flexibility by affording simple mechanisms for handling multimodal ancillary information for controlling the synthesis process (eg, a few sample frames, an audio track, a trajectory in image space) and taking into account the intrinsically uncertain nature of the future by allowing multiple predictions. Experiments with an implementation of the proposed approach give very good qualitative and quantitative results on multiple tasks and standard benchmarks."
                },
                {
                    "title": "FitVid: Overfitting in Pixel-Level Video Prediction",
                    "abstract": "An agent that is capable of predicting what happens next can perform a variety of tasks through planning with no additional training. Furthermore, such an agent can internally represent the complex dynamics of the real-world and therefore can acquire a representation useful for a variety of visual perception tasks. This makes predicting the future frames of a video, conditioned on the observed past and potentially future actions, an interesting task which remains exceptionally challenging despite many recent advances. Existing video prediction models have shown promising results on simple narrow benchmarks but they generate low quality predictions on real-life datasets with more complicated dynamics or broader domain. There is a growing body of evidence that underfitting on the training data is one of the primary causes for the low quality predictions. In this paper, we argue that the inefficient use of parameters in the current video models is the main reason for underfitting. Therefore, we introduce a new architecture, named FitVid, which is capable of severe overfitting on the common benchmarks while having similar parameter count as the current state-of-the-art models. We analyze the consequences of overfitting, illustrating how it can produce unexpected outcomes such as generating high quality output by repeating the training data, and how it can be mitigated using existing image augmentation techniques. As a result, FitVid outperforms the current state-of-the-art models across four different video prediction benchmarks on four different metrics."
                },
                {
                    "title": "VideoGPT: Video Generation using VQ-VAE and Transformers",
                    "abstract": "We present VideoGPT: a conceptually simple architecture for scaling likelihood based generative modeling to natural videos. VideoGPT uses VQ-VAE that learns downsampled discrete latent representations of a raw video by employing 3D convolutions and axial self-attention. A simple GPT-like architecture is then used to autoregressively model the discrete latents using spatio-temporal position encodings. Despite the simplicity in formulation and ease of training, our architecture is able to generate samples competitive with state-of-the-art GAN models for video generation on the BAIR Robot dataset, and generate high fidelity natural videos from UCF-101 and Tumbler GIF Dataset (TGIF). We hope our proposed architecture serves as a reproducible reference for a minimalistic implementation of transformer based video generation models. Samples and code are available at https://wilson1yan.github.io/videogpt/index.html"
                },
                {
                    "title": "PredRNN: A Recurrent Neural Network for Spatiotemporal Predictive Learning",
                    "abstract": "The predictive learning of spatiotemporal sequences aims to generate future images by learning from the historical context, where the visual dynamics are believed to have modular structures that can be learned with compositional subsystems. This paper models these structures by presenting PredRNN, a new recurrent network, in which a pair of memory cells are explicitly decoupled, operate in nearly independent transition manners, and finally form unified representations of the complex environment. Concretely, besides the original memory cell of LSTM, this network is featured by a zigzag memory flow that propagates in both bottom-up and top-down directions across all layers, enabling the learned visual dynamics at different levels of RNNs to communicate. It also leverages a memory decoupling loss to keep the memory cells from learning redundant features. We further propose a new curriculum learning strategy to force PredRNN to learn long-term dynamics from context frames, which can be generalized to most sequence-to-sequence models. We provide detailed ablation studies to verify the effectiveness of each component. Our approach is shown to obtain highly competitive results on five datasets for both action-free and action-conditioned predictive learning scenarios."
                },
                {
                    "title": "Greedy Hierarchical Variational Autoencoders for Large-Scale Video Prediction",
                    "abstract": "A video prediction model that generalizes to diverse scenes would enable intelligent agents such as robots to perform a variety of tasks via planning with the model. However, while existing video prediction models have produced promising results on small datasets, they suffer from severe underfitting when trained on large and diverse datasets. To address this underfitting challenge, we first observe that the ability to train larger video prediction models is often bottlenecked by the memory constraints of GPUs or TPUs. In parallel, deep hierarchical latent variable models can produce higher quality predictions by capturing the multi-level stochasticity of future observations, but end-to-end optimization of such models is notably difficult. Our key insight is that greedy and modular optimization of hierarchical autoencoders can simultaneously address both the memory constraints and the optimization challenges of large-scale video prediction. We introduce Greedy Hierarchical Variational Autoencoders (GHVAEs), a method that learns highfidelity video predictions by greedily training each level of a hierarchical autoencoder. In comparison to state- of-the-art models, GHVAEs provide 17-55% gains in prediction performance on four video datasets, a 35\u201340% higher success rate on real robot tasks, and can improve performance monotonically by simply adding more modules. Visualization and more details are at https://sites.google.com/view/ghvae."
                },
                {
                    "title": "Taming Transformers for High-Resolution Image Synthesis",
                    "abstract": "Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks. In contrast to CNNs, they contain no inductive bias that prioritizes local interactions. This makes them expressive, but also computationally infeasible for long sequences, such as high-resolution images. We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images. We show how to (i) use CNNs to learn a context-rich vocabulary of image constituents, and in turn (ii) utilize transformers to efficiently model their composition within high-resolution images. Our approach is readily applied to conditional synthesis tasks, where both non-spatial information, such as object classes, and spatial information, such as segmentations, can control the generated image. In particular, we present the first results on semantically-guided synthesis of megapixel images with transformers. Project page at https://git.io/JLlvY."
                },
                {
                    "title": "EDAN: An EMG-controlled Daily Assistant to Help People With Physical Disabilities",
                    "abstract": "Injuries, accidents, strokes, and other diseases can significantly degrade the capabilities to perform even the most simple activities in daily life. A large share of these cases involves neuromuscular diseases, which lead to severely reduced muscle function. However, even though affected people are no longer able to move their limbs, residual muscle function can still be existent. Previous work has shown that this residual muscular activity can suffice to apply an EMG-based user interface. In this paper, we introduce DLR\u2019s robotic wheelchair EDAN (EMG-controlled Daily Assistant), which is equipped with a torque-controlled, eight degree-of-freedom light-weight arm and a dexterous, five-fingered robotic hand. Using electromyography, muscular activity of the user is measured, processed and utilized to control both the wheelchair and the robotic manipulator. This EMG-based interface is enhanced with shared control functionality to allow for efficient and safe physical interaction with the environment."
                },
                {
                    "title": "Mastering Atari with Discrete World Models",
                    "abstract": "Intelligent agents need to generalize from past experience to achieve goals in complex environments. World models facilitate such generalization and allow learning behaviors from imagined outcomes to increase sample-efficiency. While learning world models from image inputs has recently become feasible for some tasks, modeling Atari games accurately enough to derive successful behaviors has remained an open challenge for many years. We introduce DreamerV2, a reinforcement learning agent that learns behaviors purely from predictions in the compact latent space of a powerful world model. The world model uses discrete representations and is trained separately from the policy. DreamerV2 constitutes the first agent that achieves human-level performance on the Atari benchmark of 55 tasks by learning behaviors inside a separately trained world model. With the same computational budget and wall-clock time, DreamerV2 reaches 200M frames and exceeds the final performance of the top single-GPU agents IQN and Rainbow."
                },
                {
                    "title": "Shared Control Templates for Assistive Robotics",
                    "abstract": "Light-weight robotic manipulators can be used to restore the manipulation capability of people with a motor disability. However, manipulating the environment poses a complex task, especially when the control interface is of low bandwidth, as may be the case for users with impairments. Therefore, we propose a constraint-based shared control scheme to define skills which provide support during task execution. This is achieved by representing a skill as a sequence of states, with specific user command mappings and different sets of constraints being applied in each state. New skills are defined by combining different types of constraints and conditions for state transitions, in a human-readable format. We demonstrate its versatility in a pilot experiment with three activities of daily living. Results show that even complex, high-dimensional tasks can be performed with a low-dimensional interface using our shared control approach."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Dream to Control: Learning Behaviors by Latent Imagination",
                    "abstract": "To select effective actions in complex environments, intelligent agents need to generalize from past experience. World models can represent knowledge about the environment to facilitate such generalization. While learning world models from high-dimensional sensory inputs is becoming feasible through deep learning, there are many potential ways for deriving behaviors from them. We present Dreamer, a reinforcement learning agent that solves long-horizon tasks purely by latent imagination. We efficiently learn behaviors by backpropagating analytic gradients of learned state values through trajectories imagined in the compact state space of a learned world model. On 20 challenging visual control tasks, Dreamer exceeds existing approaches in data-efficiency, computation time, and final performance."
                },
                {
                    "title": "High Fidelity Video Prediction with Large Stochastic Recurrent Neural Networks",
                    "abstract": "Predicting future video frames is extremely challenging, as there are many factors of variation that make up the dynamics of how frames change through time. Previously proposed solutions require complex inductive biases inside network architectures with highly specialized computation, including segmentation masks, optical flow, and foreground and background separation. In this work, we question if such handcrafted architectures are necessary and instead propose a different approach: finding minimal inductive bias for video prediction while maximizing network capacity. We investigate this question by performing the first large-scale empirical study and demonstrate state-of-the-art performance by learning large models on three different datasets: one for modeling object interactions, one for modeling human motion, and one for modeling car driving."
                },
                {
                    "title": "Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning",
                    "abstract": "Meta-reinforcement learning algorithms can enable robots to acquire new skills much more quickly, by leveraging prior experience to learn how to learn. However, much of the current research on meta-reinforcement learning focuses on task distributions that are very narrow. For example, a commonly used meta-reinforcement learning benchmark uses different running velocities for a simulated robot as different tasks. When policies are meta-trained on such narrow task distributions, they cannot possibly generalize to more quickly acquire entirely new tasks. Therefore, if the aim of these methods is to enable faster acquisition of entirely new behaviors, we must evaluate them on task distributions that are sufficiently broad to enable generalization to new behaviors. In this paper, we propose an open-source simulated benchmark for meta-reinforcement learning and multi-task learning consisting of 50 distinct robotic manipulation tasks. Our aim is to make it possible to develop algorithms that generalize to accelerate the acquisition of entirely new, held-out tasks. We evaluate 6 state-of-the-art meta-reinforcement learning and multi-task learning algorithms on these tasks. Surprisingly, while each task and its variations (e.g., with different object positions) can be learned with reasonable success, these algorithms struggle to learn with multiple tasks at the same time, even with as few as ten distinct training tasks. Our analysis and open-source environments pave the way for future research in multi-task learning and meta-learning that can enable meaningful generalization, thereby unlocking the full potential of these methods."
                },
                {
                    "title": "RoboNet: Large-Scale Multi-Robot Learning",
                    "abstract": "Robot learning has emerged as a promising tool for taming the complexity and diversity of the real world. Methods based on high-capacity models, such as deep networks, hold the promise of providing effective generalization to a wide range of open-world environments. However, these same methods typically require large amounts of diverse training data to generalize effectively. In contrast, most robotic learning experiments are small-scale, single-domain, and single-robot. This leads to a frequent tension in robotic learning: how can we learn generalizable robotic controllers without having to collect impractically large amounts of data for each separate experiment? In this paper, we propose RoboNet, an open database for sharing robotic experience, which provides an initial pool of 15 million video frames, from 7 different robot platforms, and study how it can be used to learn generalizable models for vision-based robotic manipulation. We combine the dataset with two different learning algorithms: visual foresight, which uses forward video prediction models, and supervised inverse models. Our experiments test the learned algorithms' ability to work across new objects, new tasks, new scenes, new camera viewpoints, new grippers, or even entirely new robots. In our final experiment, we find that by pre-training on RoboNet and fine-tuning on data from a held-out Franka or Kuka robot, we can exceed the performance of a robot-specific training approach that uses 4x-20x more data. For videos and data, see the project webpage: https://www.robonet.wiki/"
                },
                {
                    "title": "Unsupervised Learning of Object Structure and Dynamics from Videos",
                    "abstract": "Extracting and predicting object structure and dynamics from videos without supervision is a major challenge in machine learning. To address this challenge, we adopt a keypoint-based image representation and learn a stochastic dynamics model of the keypoints. Future frames are reconstructed from the keypoints and a reference frame. By modeling dynamics in the keypoint coordinate space, we achieve stable learning and avoid compounding of errors in pixel space. Our method improves upon unstructured representations both for pixel-level video prediction and for downstream tasks requiring object-level understanding of motion dynamics. We evaluate our model on diverse datasets: a multi-agent sports dataset, the Human3.6M dataset, and datasets based on continuous control tasks from the DeepMind Control Suite. The spatially structured representation outperforms unstructured representations on a range of motion-related tasks such as object tracking, action recognition and reward prediction."
                },
                {
                    "title": "When to Trust Your Model: Model-Based Policy Optimization",
                    "abstract": "Designing effective model-based reinforcement learning algorithms is difficult because the ease of data generation must be weighed against the bias of model-generated data. In this paper, we study the role of model usage in policy optimization both theoretically and empirically. We first formulate and analyze a model-based reinforcement learning algorithm with a guarantee of monotonic improvement at each step. In practice, this analysis is overly pessimistic and suggests that real off-policy data is always preferable to model-generated on-policy data, but we show that an empirical estimate of model generalization can be incorporated into such analysis to justify model usage. Motivated by this analysis, we then demonstrate that a simple procedure of using short model-generated rollouts branched from real data has the benefits of more complicated model-based algorithms without the usual pitfalls. In particular, this approach surpasses the sample efficiency of prior model-based methods, matches the asymptotic performance of the best model-free algorithms, and scales to horizons that cause other model-based methods to fail entirely."
                },
                {
                    "title": "Model-Based Reinforcement Learning for Atari",
                    "abstract": "Model-free reinforcement learning (RL) can be used to learn effective policies for complex tasks, such as Atari games, even from image observations. However, this typically requires very large amounts of interaction -- substantially more, in fact, than a human would need to learn the same games. How can people learn so quickly? Part of the answer may be that people can learn how the game works and predict which actions will lead to desirable outcomes. In this paper, we explore how video prediction models can similarly enable agents to solve Atari games with fewer interactions than model-free methods. We describe Simulated Policy Learning (SimPLe), a complete model-based deep RL algorithm based on video prediction models and present a comparison of several model architectures, including a novel architecture that yields the best results in our setting. Our experiments evaluate SimPLe on a range of Atari games in low data regime of 100k interactions between the agent and the environment, which corresponds to two hours of real-time play. In most games SimPLe outperforms state-of-the-art model-free algorithms, in some games by over an order of magnitude."
                },
                {
                    "title": "Towards Accurate Generative Models of Video: A New Metric & Challenges",
                    "abstract": "Recent advances in deep generative models have lead to remarkable progress in synthesizing high quality images. Following their successful application in image processing and representation learning, an important next step is to consider videos. Learning generative models of video is a much harder task, requiring a model to capture the temporal dynamics of a scene, in addition to the visual presentation of objects. While recent attempts at formulating generative models of video have had some success, current progress is hampered by (1) the lack of qualitative metrics that consider visual quality, temporal coherence, and diversity of samples, and (2) the wide gap between purely synthetic video data sets and challenging real-world data sets in terms of complexity. To this extent we propose Fr\\'{e}chet Video Distance (FVD), a new metric for generative models of video, and StarCraft 2 Videos (SCV), a benchmark of game play from custom starcraft 2 scenarios that challenge the current capabilities of generative models of video. We contribute a large-scale human study, which confirms that FVD correlates well with qualitative human judgment of generated videos, and provide initial benchmark results on SCV."
                },
                {
                    "title": "Visual Foresight: Model-Based Deep Reinforcement Learning for Vision-Based Robotic Control",
                    "abstract": "Deep reinforcement learning (RL) algorithms can learn complex robotic skills from raw sensory inputs, but have yet to achieve the kind of broad generalization and applicability demonstrated by deep learning methods in supervised domains. We present a deep RL method that is practical for real-world robotics tasks, such as robotic manipulation, and generalizes effectively to never-before-seen tasks and objects. In these settings, ground truth reward signals are typically unavailable, and we therefore propose a self-supervised model-based approach, where a predictive model learns to directly predict the future from raw sensory readings, such as camera images. At test time, we explore three distinct goal specification methods: designated pixels, where a user specifies desired object manipulation tasks by selecting particular pixels in an image and corresponding goal positions, goal images, where the desired goal state is specified with an image, and image classifiers, which define spaces of goal states. Our deep predictive models are trained using data collected autonomously and continuously by a robot interacting with hundreds of objects, without human supervision. We demonstrate that visual MPC can generalize to never-before-seen objects---both rigid and deformable---and solve a range of user-defined object manipulation tasks using the same model."
                },
                {
                    "title": "Learning Latent Dynamics for Planning from Pixels",
                    "abstract": "Planning has been very successful for control tasks with known environment dynamics. To leverage planning in unknown environments, the agent needs to learn the dynamics from interactions with the world. However, learning dynamics models that are accurate enough for planning has been a long-standing challenge, especially in image-based domains. We propose the Deep Planning Network (PlaNet), a purely model-based agent that learns the environment dynamics from images and chooses actions through fast online planning in latent space. To achieve high performance, the dynamics model must accurately predict the rewards ahead for multiple time steps. We approach this using a latent dynamics model with both deterministic and stochastic transition components. Moreover, we propose a multi-step variational inference objective that we name latent overshooting. Using only pixel observations, our agent solves continuous control tasks with contact dynamics, partial observability, and sparse rewards, which exceed the difficulty of tasks that were previously solved by planning with learned models. PlaNet uses substantially fewer episodes and reaches final performance close to and sometimes higher than strong model-free algorithms."
                },
                {
                    "title": "Making Sense of Vision and Touch: Self-Supervised Learning of Multimodal Representations for Contact-Rich Tasks",
                    "abstract": "Contact-rich manipulation tasks in unstructured environments often require both haptic and visual feedback. However, it is non-trivial to manually design a robot controller that combines modalities with very different characteristics. While deep reinforcement learning has shown success in learning control policies for high-dimensional inputs, these algorithms are generally intractable to deploy on real robots due to sample complexity. We use self-supervision to learn a compact and multimodal representation of our sensory inputs, which can then be used to improve the sample efficiency of our policy learning. We evaluate our method on a peg insertion task, generalizing over different geometry, configurations, and clearances, while being robust to external perturbations. We present results in simulation and on a real robot."
                },
                {
                    "title": "Recurrent World Models Facilitate Policy Evolution",
                    "abstract": "A generative recurrent neural network is quickly trained in an unsupervised manner to model popular reinforcement learning environments through compressed spatio-temporal representations. The world model's extracted features are fed into compact and simple policies trained by evolution, achieving state of the art results in various environments. We also train our agent entirely inside of an environment generated by its own internal world model, and transfer this policy back into the actual environment. Interactive version of this paper is available at https://worldmodels.github.io"
                },
                {
                    "title": "QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation",
                    "abstract": "In this paper, we study the problem of learning vision-based dynamic manipulation skills using a scalable reinforcement learning approach. We study this problem in the context of grasping, a longstanding challenge in robotic manipulation. In contrast to static learning behaviors that choose a grasp point and then execute the desired grasp, our method enables closed-loop vision-based control, whereby the robot continuously updates its grasp strategy based on the most recent observations to optimize long-horizon grasp success. To that end, we introduce QT-Opt, a scalable self-supervised vision-based reinforcement learning framework that can leverage over 580k real-world grasp attempts to train a deep neural network Q-function with over 1.2M parameters to perform closed-loop, real-world grasping that generalizes to 96% grasp success on unseen objects. Aside from attaining a very high success rate, our method exhibits behaviors that are quite distinct from more standard grasping systems: using only RGB vision-based perception from an over-the-shoulder camera, our method automatically learns regrasping strategies, probes objects to find the most effective grasps, learns to reposition objects and perform other non-prehensile pre-grasp manipulations, and responds dynamically to disturbances and perturbations."
                },
                {
                    "title": "Stochastic Video Generation with a Learned Prior",
                    "abstract": "Generating video frames that accurately predict future world states is challenging. Existing approaches either fail to capture the full distribution of outcomes, or yield blurry generations, or both. In this paper we introduce an unsupervised video generation model that learns a prior model of uncertainty in a given environment. Video frames are generated by drawing samples from this prior and combining them with a deterministic estimate of the future frame. The approach is simple and easily trained end-to-end on a variety of datasets. Sample generations are both varied and sharp, even many frames into the future, and compare favorably to those from existing approaches."
                },
                {
                    "title": "The Unreasonable Effectiveness of Deep Features as a Perceptual Metric",
                    "abstract": "While it is nearly effortless for humans to quickly assess the perceptual similarity between two images, the underlying processes are thought to be quite complex. Despite this, the most widely used perceptual metrics today, such as PSNR and SSIM, are simple, shallow functions, and fail to account for many nuances of human perception. Recently, the deep learning community has found that features of the VGG network trained on ImageNet classification has been remarkably useful as a training loss for image synthesis. But how perceptual are these so-called \"perceptual losses\"? What elements are critical for their success? To answer these questions, we introduce a new dataset of human perceptual similarity judgments. We systematically evaluate deep features across different architectures and tasks and compare them with classic metrics. We find that deep features outperform all previous metrics by large margins on our dataset. More surprisingly, this result is not restricted to ImageNet-trained VGG features, but holds across different deep architectures and levels of supervision (supervised, self-supervised, or even unsupervised). Our results suggest that perceptual similarity is an emergent property shared across deep visual representations."
                },
                {
                    "title": "Neural Discrete Representation Learning",
                    "abstract": "Learning useful representations without supervision remains a key challenge in machine learning. In this paper, we propose a simple yet powerful generative model that learns such discrete representations. Our model, the Vector Quantised-Variational AutoEncoder (VQ-VAE), differs from VAEs in two key ways: the encoder network outputs discrete, rather than continuous, codes; and the prior is learnt rather than static. In order to learn a discrete latent representation, we incorporate ideas from vector quantisation (VQ). Using the VQ method allows the model to circumvent issues of \"posterior collapse\" -- where the latents are ignored when they are paired with a powerful autoregressive decoder -- typically observed in the VAE framework. Pairing these representations with an autoregressive prior, the model can generate high quality images, videos, and speech as well as doing high quality speaker conversion and unsupervised learning of phonemes, providing further evidence of the utility of the learnt representations."
                },
                {
                    "title": "Stochastic Variational Video Prediction",
                    "abstract": "Predicting the future in real-world settings, particularly from raw sensory observations such as images, is exceptionally challenging. Real-world events can be stochastic and unpredictable, and the high dimensionality and complexity of natural images requires the predictive model to build an intricate understanding of the natural world. Many existing methods tackle this problem by making simplifying assumptions about the environment. One common assumption is that the outcome is deterministic and there is only one plausible future. This can lead to low-quality predictions in real-world settings with stochastic dynamics. In this paper, we develop a stochastic variational video prediction (SV2P) method that predicts a different possible future for each sample of its latent variables. To the best of our knowledge, our model is the first to provide effective stochastic multi-frame prediction for real-world video. We demonstrate the capability of the proposed method in predicting detailed future frames of videos on multiple real-world datasets, both action-free and action-conditioned. We find that our proposed method produces substantially improved video predictions when compared to the same model without stochasticity, and to other stochastic video prediction methods. Our SV2P implementation will be open sourced upon publication."
                },
                {
                    "title": "Self-Supervised Visual Planning with Temporal Skip Connections",
                    "abstract": "In order to autonomously learn wide repertoires of complex skills, robots must be able to learn from their own autonomously collected data, without human supervision. One learning signal that is always available for autonomously collected data is prediction: if a robot can learn to predict the future, it can use this predictive model to take actions to produce desired outcomes, such as moving an object to a particular location. However, in complex open-world scenarios, designing a representation for prediction is difficult. In this work, we instead aim to enable self-supervised robotic learning through direct video prediction: instead of attempting to design a good representation, we directly predict what the robot will see next, and then use this model to achieve desired goals. A key challenge in video prediction for robotic manipulation is handling complex spatial arrangements such as occlusions. To that end, we introduce a video prediction model that can keep track of objects through occlusion by incorporating temporal skip-connections. Together with a novel planning criterion and action space formulation, we demonstrate that this model substantially outperforms prior work on video prediction-based control. Our results show manipulation of objects not seen during training, handling multiple objects, and pushing objects around obstructions. These results represent a significant advance in the range and complexity of skills that can be performed entirely with self-supervised robotic learning."
                },
                {
                    "title": "The \u201cSomething Something\u201d Video Database for Learning and Evaluating Visual Common Sense",
                    "abstract": "Neural networks trained on datasets such as ImageNet have led to major advances in visual object classification. One obstacle that prevents networks from reasoning more deeply about complex scenes and situations, and from integrating visual knowledge with natural language, like humans do, is their lack of common sense knowledge about the physical world. Videos, unlike still images, contain a wealth of detailed information about the physical world. However, most labelled video datasets represent high-level concepts rather than detailed physical aspects about actions and scenes. In this work, we describe our ongoing collection of the \u201csomething-something\u201d database of video prediction tasks whose solutions require a common sense understanding of the depicted situation. The database currently contains more than 100,000 videos across 174 classes, which are defined as caption-templates. We also describe the challenges in crowd-sourcing this data at scale."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Deep Q-learning From Demonstrations",
                    "abstract": "\n \n Deep reinforcement learning (RL) has achieved several high profile successes in difficult decision-making problems. However, these algorithms typically require a huge amount of data before they reach reasonable performance. In fact, their performance during learning can be extremely poor. This may be acceptable for a simulator, but it severely limits the applicability of deep RL to many real-world tasks, where the agent must learn in the real environment. In this paper we study a setting where the agent may access data from previous control of the system. We present an algorithm, Deep Q-learning from Demonstrations (DQfD), that leverages small sets of demonstration data to massively accelerate the learning process even from relatively small amounts of demonstration data and is able to automatically assess the necessary ratio of demonstration data while learning thanks to a prioritized replay mechanism. DQfD works by combining temporal difference updates with supervised classification of the demonstrator\u2019s actions. We show that DQfD has better initial performance than Prioritized Dueling Double Deep Q-Networks (PDD DQN) as it starts with better scores on the first million steps on 41 of 42 games and on average it takes PDD DQN 83 million steps to catch up to DQfD\u2019s performance. DQfD learns to out-perform the best demonstration given in 14 of 42 games. In addition, DQfD leverages human demonstrations to achieve state-of-the-art results for 11 games. Finally, we show that DQfD performs better than three related algorithms for incorporating demonstration data into DQN.\n \n"
                },
                {
                    "title": "Deep visual foresight for planning robot motion",
                    "abstract": "A key challenge in scaling up robot learning to many skills and environments is removing the need for human supervision, so that robots can collect their own data and improve their own performance without being limited by the cost of requesting human feedback. Model-based reinforcement learning holds the promise of enabling an agent to learn to predict the effects of its actions, which could provide flexible predictive models for a wide range of tasks and environments, without detailed human supervision. We develop a method for combining deep action-conditioned video prediction models with model-predictive control that uses entirely unlabeled training data. Our approach does not require a calibrated camera, an instrumented training set-up, nor precise sensing and actuation. Our results show that our method enables a real robot to perform nonprehensile manipulation \u2014 pushing objects \u2014 and can handle novel objects not seen during training."
                },
                {
                    "title": "Action-Conditional Video Prediction using Deep Networks in Atari Games",
                    "abstract": "Motivated by vision-based reinforcement learning (RL) problems, in particular Atari games from the recent benchmark Aracade Learning Environment (ALE), we consider spatio-temporal prediction problems where future (image-)frames are dependent on control variables or actions as well as previous frames. While not composed of natural scenes, frames in Atari games are high-dimensional in size, can involve tens of objects with one or more objects being controlled by the actions directly and many other objects being influenced indirectly, can involve entry and departure of objects, and can involve deep partial observability. We propose and evaluate two deep neural network architectures that consist of encoding, action-conditional transformation, and decoding layers based on convolutional neural networks and recurrent neural networks. Experimental results show that the proposed architectures are able to generate visually-realistic frames that are also useful for control over approximately 100-step action-conditional futures in some games. To the best of our knowledge, this paper is the first to make and evaluate long-term predictions on high-dimensional video conditioned by control inputs."
                },
                {
                    "title": "Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting",
                    "abstract": "The goal of precipitation nowcasting is to predict the future rainfall intensity in a local region over a relatively short period of time. Very few previous studies have examined this crucial and challenging weather forecasting problem from the machine learning perspective. In this paper, we formulate precipitation nowcasting as a spatiotemporal sequence forecasting problem in which both the input and the prediction target are spatiotemporal sequences. By extending the fully connected LSTM (FC-LSTM) to have convolutional structures in both the input-to-state and state-to-state transitions, we propose the convolutional LSTM (ConvLSTM) and use it to build an end-to-end trainable model for the precipitation nowcasting problem. Experiments show that our ConvLSTM network captures spatiotemporal correlations better and consistently outperforms FC-LSTM and the state-of-the-art operational ROVER algorithm for precipitation nowcasting."
                },
                {
                    "title": "Deep neural networks are easily fooled: High confidence predictions for unrecognizable images",
                    "abstract": "Deep neural networks (DNNs) have recently been achieving state-of-the-art performance on a variety of pattern-recognition tasks, most notably visual classification problems. Given that DNNs are now able to classify objects in images with near-human-level performance, questions naturally arise as to what differences remain between computer and human vision. A recent study [30] revealed that changing an image (e.g. of a lion) in a way imperceptible to humans can cause a DNN to label the image as something else entirely (e.g. mislabeling a lion a library). Here we show a related result: it is easy to produce images that are completely unrecognizable to humans, but that state-of-the-art DNNs believe to be recognizable objects with 99.99% confidence (e.g. labeling with certainty that white noise static is a lion). Specifically, we take convolutional neural networks trained to perform well on either the ImageNet or MNIST datasets and then find images with evolutionary algorithms or gradient ascent that DNNs label with high confidence as belonging to each dataset class. It is possible to produce images totally unrecognizable to human eyes that DNNs believe with near certainty are familiar objects, which we call \u201cfooling images\u201d (more generally, fooling examples). Our results shed light on interesting differences between human vision and current DNNs, and raise questions about the generality of DNN computer vision."
                },
                {
                    "title": "Scope of validity of PSNR in image/video quality assessment",
                    "abstract": "Experimental data are presented that clearly demonstrate the scope of application of peak signal-to-noise ratio (PSNR) as a video quality metric. It is shown that as long as the video content and the codec type are not changed, PSNR is a valid quality measure. However, when the content is changed, correlation between subjective quality and PSNR is highly reduced. Hence PSNR cannot be a reliable method for assessing the video quality across different video contents."
                },
                {
                    "title": "Image quality assessment: from error visibility to structural similarity",
                    "abstract": "Objective methods for assessing perceptual image quality traditionally attempted to quantify the visibility of errors (differences) between a distorted image and a reference image using a variety of known properties of the human visual system. Under the assumption that human visual perception is highly adapted for extracting structural information from a scene, we introduce an alternative complementary framework for quality assessment based on the degradation of structural information. As a specific example of this concept, we develop a structural similarity index and demonstrate its promise through a set of intuitive examples, as well as comparison to both subjective ratings and state-of-the-art objective methods on a database of images compressed with JPEG and JPEG2000. A MATLAB implementation of the proposed algorithm is available online at http://www.cns.nyu.edu//spl sim/lcv/ssim/."
                },
                {
                    "title": "A Path Towards Autonomous Machine Intelligence Version 0.9.2, 2022-06-27",
                    "abstract": "How could machines learn as efficiently as humans and animals? How could machines learn to reason and plan? How could machines learn representations of percepts and action plans at multiple levels of abstraction, enabling them to reason, predict, and plan at multiple time horizons? This position paper proposes an architecture and training paradigms with which to construct autonomous intelligent agents. It combines concepts such as configurable predictive world model, behavior driven through intrinsic motivation, and hierarchical joint embedding architectures trained with self-supervised learning."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.LG",
                "cs.RO"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop an interactive and scalable world model for video generation that effectively integrates multimodal signals to enhance model-based reinforcement learning?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of generative models and their applications in reinforcement learning. By bridging the gap between generative video models and world models, this research could lead to more sophisticated agents capable of learning complex behaviors in real-world scenarios. The implications extend to various fields, including robotics, autonomous systems, and interactive AI, where improved predictive capabilities can enhance decision-making and efficiency. Addressing this question could pave the way for practical applications that require high-level planning and real-time interaction, ultimately transforming how agents learn and operate in dynamic environments.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to balance interactivity and scalability in world models. Current model-based reinforcement learning approaches primarily utilize recurrent networks, which are limited in their ability to handle complex, real-world data at scale. Naive approaches may fail due to their inability to provide the necessary granularity for agents to learn precise behaviors step-by-step. Additionally, the integration of multimodal signals in a coherent manner poses technical obstacles, such as ensuring temporal consistency and efficient tokenization of video data. Overcoming these complexities requires innovative architectural designs and training methodologies that can effectively manage diverse input conditions.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either generative video models or world models in isolation, leading to a lack of comprehensive solutions that integrate both aspects. Limitations in existing models often stem from their reliance on specific architectures that do not scale well or adapt to diverse conditions. Barriers such as the complexity of modeling interactions in high-dimensional video data and the inefficiencies in current tokenization methods have hindered progress. Our approach differs by leveraging an autoregressive transformer framework that allows for seamless integration with existing large language models, enabling more flexible and efficient handling of multimodal signals without the need for extensive architectural modifications.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, Interactive VideoGPT (iVideoGPT), involves a two-phase approach: pre-training followed by domain-specific adaptation. The model will utilize a dataset of diverse video content to learn compressive tokenization for visual observations, actions"
            }
        },
        "author_data": {
            "015ec252-d93a-4b68-b8c6-059b7fe0e211": {
                "pk": "015ec252-d93a-4b68-b8c6-059b7fe0e211",
                "name": "Jialong Wu",
                "collaborators": [
                    "Mingsheng Long",
                    "Deyu Zhou",
                    "Jianmin Wang",
                    "Linhai Zhang",
                    "Guoqiang Xu",
                    "Chaoyi Deng",
                    "Haixu Wu",
                    "Yilong Lai",
                    "Congzhi Zhang",
                    "Mirko Meuter"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Reinforcement Learning",
                    "Computer Vision",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "DINER: Debiasing Aspect-based Sentiment Analysis with Multi-variable Causal Inference",
                        "abstract": "Though notable progress has been made, neural-based aspect-based sentiment analysis (ABSA) models are prone to learn spurious correlations from annotation biases, resulting in poor robustness on adversarial data transformations. Among the debiasing solutions, causal inference-based methods have attracted much research attention, which can be mainly categorized into causal intervention methods and counterfactual reasoning methods. However, most of the present debiasing methods focus on single-variable causal inference, which is not suitable for ABSA with two input variables (the target aspect and the review). In this paper, we propose a novel framework based on multi-variable causal inference for debiasing ABSA. In this framework, different types of biases are tackled based on different causal intervention methods. For the review branch, the bias is modeled as indirect confounding from context, where backdoor adjustment intervention is employed for debiasing. For the aspect branch, the bias is described as a direct correlation with labels, where counterfactual reasoning is adopted for debiasing. Extensive experiments demonstrate the effectiveness of the proposed method compared to various baselines on the two widely used real-world aspect robustness test set datasets."
                    },
                    {
                        "title": "SparseRadNet: Sparse Perception Neural Network on Subsampled Radar Data",
                        "abstract": "Radar-based perception has gained increasing attention in autonomous driving, yet the inherent sparsity of radars poses challenges. Radar raw data often contains excessive noise, whereas radar point clouds retain only limited information. In this work, we holistically treat the sparse nature of radar data by introducing an adaptive subsampling method together with a tailored network architecture that exploits the sparsity patterns to discover global and local dependencies in the radar signal. Our subsampling module selects a subset of pixels from range-doppler (RD) spectra that contribute most to the downstream perception tasks. To improve the feature extraction on sparse subsampled data, we propose a new way of applying graph neural networks on radar data and design a novel two-branch backbone to capture both global and local neighbor information. An attentive fusion module is applied to combine features from both branches. Experiments on the RADIal dataset show that our SparseRadNet exceeds state-of-the-art (SOTA) performance in object detection and achieves close to SOTA accuracy in freespace segmentation, meanwhile using sparse subsampled input data."
                    },
                    {
                        "title": "Pre-training Contextualized World Models with In-the-wild Videos for Reinforcement Learning",
                        "abstract": "Unsupervised pre-training methods utilizing large and diverse datasets have achieved tremendous success across a range of domains. Recent work has investigated such unsupervised pre-training methods for model-based reinforcement learning (MBRL) but is limited to domain-specific or simulated data. In this paper, we study the problem of pre-training world models with abundant in-the-wild videos for efficient learning of downstream visual control tasks. However, in-the-wild videos are complicated with various contextual factors, such as intricate backgrounds and textured appearance, which precludes a world model from extracting shared world knowledge to generalize better. To tackle this issue, we introduce Contextualized World Models (ContextWM) that explicitly separate context and dynamics modeling to overcome the complexity and diversity of in-the-wild videos and facilitate knowledge transfer between distinct scenes. Specifically, a contextualized extension of the latent dynamics model is elaborately realized by incorporating a context encoder to retain contextual information and empower the image decoder, which encourages the latent dynamics model to concentrate on essential temporal variations. Our experiments show that in-the-wild video pre-training equipped with ContextWM can significantly improve the sample efficiency of MBRL in various domains, including robotic manipulation, locomotion, and autonomous driving. Code is available at this repository: https://github.com/thuml/ContextWM."
                    },
                    {
                        "title": "STAR: Constraint LoRA with Dynamic Active Learning for Data-Efficient Fine-Tuning of Large Language Models",
                        "abstract": "Though Large Language Models (LLMs) have demonstrated the powerful capabilities of few-shot learning through prompting methods, supervised training is still necessary for complex reasoning tasks. Because of their extensive parameters and memory consumption, both Parameter-Efficient Fine-Tuning (PEFT) methods and Memory-Efficient Fine-Tuning methods have been proposed for LLMs. Nevertheless, the issue of large annotated data consumption, the aim of Data-Efficient Fine-Tuning, remains unexplored. One obvious way is to combine the PEFT method with active learning. However, the experimental results show that such a combination is not trivial and yields inferior results. Through probe experiments, such observation might be explained by two main reasons: uncertainty gap and poor model calibration. Therefore, in this paper, we propose a novel approach to effectively integrate uncertainty-based active learning and LoRA. Specifically, for the uncertainty gap, we introduce a dynamic uncertainty measurement that combines the uncertainty of the base model and the uncertainty of the full model during the iteration of active learning. For poor model calibration, we incorporate the regularization method during LoRA training to keep the model from being over-confident, and the Monte-Carlo dropout mechanism is employed to enhance the uncertainty estimation. Experimental results show that the proposed approach outperforms existing baseline models on three complex reasoning tasks."
                    },
                    {
                        "title": "Out-of-Dynamics Imitation Learning from Multimodal Demonstrations",
                        "abstract": "Existing imitation learning works mainly assume that the demonstrator who collects demonstrations shares the same dynamics as the imitator. However, the assumption limits the usage of imitation learning, especially when collecting demonstrations for the imitator is difficult. In this paper, we study out-of-dynamics imitation learning (OOD-IL), which relaxes the assumption to that the demonstrator and the imitator have the same state spaces but could have different action spaces and dynamics. OOD-IL enables imitation learning to utilize demonstrations from a wide range of demonstrators but introduces a new challenge: some demonstrations cannot be achieved by the imitator due to the different dynamics. Prior works try to filter out such demonstrations by feasibility measurements, but ignore the fact that the demonstrations exhibit a multimodal distribution since the different demonstrators may take different policies in different dynamics. We develop a better transferability measurement to tackle this newly-emerged challenge. We firstly design a novel sequence-based contrastive clustering algorithm to cluster demonstrations from the same mode to avoid the mutual interference of demonstrations from different modes, and then learn the transferability of each demonstration with an adversarial-learning based algorithm in each cluster. Experiment results on several MuJoCo environments, a driving environment, and a simulated robot environment show that the proposed transferability measurement more accurately finds and down-weights non-transferable demonstrations and outperforms prior works on the final imitation learning performance. We show the videos of our experiment results on our website."
                    },
                    {
                        "title": "Supported Policy Optimization for Offline Reinforcement Learning",
                        "abstract": "Policy constraint methods to offline reinforcement learning (RL) typically utilize parameterization or regularization that constrains the policy to perform actions within the support set of the behavior policy. The elaborative designs of parameterization methods usually intrude into the policy networks, which may bring extra inference cost and cannot take full advantage of well-established online methods. Regularization methods reduce the divergence between the learned policy and the behavior policy, which may mismatch the inherent density-based definition of support set thereby failing to avoid the out-of-distribution actions effectively. This paper presents Supported Policy OpTimization (SPOT), which is directly derived from the theoretical formalization of the density-based support constraint. SPOT adopts a VAE-based density estimator to explicitly model the support set of behavior policy and presents a simple but effective density-based regularization term, which can be plugged non-intrusively into off-the-shelf off-policy RL algorithms. SPOT achieves the state-of-the-art performance on standard benchmarks for offline RL. Benefiting from the pluggable design, offline pretrained models from SPOT can also be applied to perform online fine-tuning seamlessly."
                    },
                    {
                        "title": "Flowformer: Linearizing Transformers with Conservation Flows",
                        "abstract": "Transformers based on the attention mechanism have achieved impressive success in various areas. However, the attention mechanism has a quadratic complexity, significantly impeding Transformers from dealing with numerous tokens and scaling up to bigger models. Previous methods mainly utilize the similarity decomposition and the associativity of matrix multiplication to devise linear-time attention mechanisms. They avoid degeneration of attention to a trivial distribution by reintroducing inductive biases such as the locality, thereby at the expense of model generality and expressiveness. In this paper, we linearize Transformers free from specific inductive biases based on the flow network theory. We cast attention as the information flow aggregated from the sources (values) to the sinks (results) through the learned flow capacities (attentions). Within this framework, we apply the property of flow conservation into attention and propose the Flow-Attention mechanism of linear complexity. By respectively conserving the incoming flow of sinks for source competition and the outgoing flow of sources for sink allocation, Flow-Attention inherently generates informative attentions without using specific inductive biases. Empowered by the Flow-Attention, Flowformer yields strong performance in linear time for wide areas, including long sequence, time series, vision, natural language, and reinforcement learning. The code and settings are available at this repository: https://github.com/thuml/Flowformer."
                    },
                    {
                        "title": "CompilerDream: Learning a Compiler World Model for General Code Optimization",
                        "abstract": "Effective code optimization in compilers is crucial for computer and software engineering. The success of these optimizations primarily depends on the selection and ordering of the optimization passes applied to the code. While most compilers rely on a fixed sequence of optimization passes, current methods to find the optimal sequence either employ impractically slow search algorithms or learning methods that struggle to generalize to code unseen during training. We introduce CompilerDream, a model-based reinforcement learning approach to general code optimization. CompilerDream comprises a compiler world model that accurately simulates the intrinsic properties of optimization passes and an agent trained on this model to produce effective optimization strategies. By training on a large-scale program dataset, CompilerDream is equipped to serve as a general code optimizer across various application scenarios and source-code languages. Our extensive experiments first highlight CompilerDream's strong optimization capabilities for autotuning, where it leads the CompilerGym leaderboard. More importantly, the zero-shot generalization ability of large-scale trained compiler world model and agent, excels across diverse datasets, surpassing LLVM's built-in optimizations and other state-of-the-art methods in both settings of value prediction and end-to-end code optimization."
                    },
                    {
                        "title": "Constituency Parsing using LLMs",
                        "abstract": "Constituency parsing is a fundamental yet unsolved natural language processing task. In this paper, we explore the potential of recent large language models (LLMs) that have exhibited remarkable performance across various domains and tasks to tackle this task. We employ three linearization strategies to transform output trees into symbol sequences, such that LLMs can solve constituency parsing by generating linearized trees. We conduct experiments using a diverse range of LLMs, including ChatGPT, GPT-4, OPT, LLaMA, and Alpaca, comparing their performance against the state-of-the-art constituency parsers. Our experiments encompass zero-shot, few-shot, and full-training learning settings, and we evaluate the models on one in-domain and five out-of-domain test datasets. Our findings reveal insights into LLMs' performance, generalization abilities, and challenges in constituency parsing."
                    },
                    {
                        "title": "SEED: Accelerating Reasoning Tree Construction via Scheduled Speculative Decoding",
                        "abstract": "Large Language Models (LLMs) demonstrate remarkable emergent abilities across various tasks, yet fall short of complex reasoning and planning tasks. The tree-search-based reasoning methods address this by surpassing the capabilities of chain-of-thought prompting, encouraging exploration of intermediate steps. However, such methods introduce significant inference latency due to the systematic exploration and evaluation of multiple thought paths. This paper introduces SeeD, a novel and efficient inference framework to optimize runtime speed and GPU memory management concurrently. By employing a scheduled speculative execution, SeeD efficiently handles multiple iterations for the thought generation and the state evaluation, leveraging a rounds-scheduled strategy to manage draft model dispatching. Extensive experimental evaluations on three reasoning datasets demonstrate superior speedup performance of SeeD, providing a viable path for batched inference in training-free speculative decoding."
                    },
                    {
                        "title": "AdaCQR: Enhancing Query Reformulation for Conversational Search via Sparse and Dense Retrieval Alignment",
                        "abstract": "Conversational Query Reformulation (CQR) has significantly advanced in addressing the challenges of conversational search, particularly those stemming from the latent user intent and the need for historical context. Recent works aimed to boost the performance of CRQ through alignment. However, they are designed for one specific retrieval system, which potentially results in poor generalization. To overcome this limitation, we present a novel framework AdaCQR. By aligning reformulation models with both term-based and semantic-based retrieval systems, AdaCQR enhances the generalizability of information-seeking queries across diverse retrieval environments through a dual-phase training strategy. We also developed two effective approaches for acquiring superior labels and diverse input candidates, boosting the efficiency and robustness of the framework. Experimental evaluations on the TopiOCQA and QReCC datasets demonstrate that AdaCQR significantly outperforms existing methods, offering both quantitative and qualitative improvements in conversational query reformulation."
                    },
                    {
                        "title": "Real-Time And Robust 3D Object Detection with Roadside LiDARs",
                        "abstract": "This work aims to address the challenges in autonomous driving by focusing on the 3D perception of the environment using roadside LiDARs. We design a 3D object detection model that can detect traffic participants in roadside LiDARs in real-time. Our model uses an existing 3D detector as a baseline and improves its accuracy. To prove the effectiveness of our proposed modules, we train and evaluate the model on three different vehicle and infrastructure datasets. To show the domain adaptation ability of our detector, we train it on an infrastructure dataset from China and perform transfer learning on a different dataset recorded in Germany. We do several sets of experiments and ablation studies for each module in the detector that show that our model outperforms the baseline by a significant margin, while the inference speed is at 45 Hz (22 ms). We make a significant contribution with our LiDAR-based 3D detector that can be used for smart city applications to provide connected and automated vehicles with a far-reaching view. Vehicles that are connected to the roadside sensors can get information about other vehicles around the corner to improve their path and maneuver planning and to increase road traffic safety."
                    }
                ]
            },
            "5b37ca8b-8d54-49d9-ab49-39bce32c0a77": {
                "pk": "5b37ca8b-8d54-49d9-ab49-39bce32c0a77",
                "name": "Shaofeng Yin",
                "collaborators": [
                    "Ting Lei",
                    "Yang Liu",
                    "Yuxin Peng"
                ],
                "domain": [
                    "Zero-shot Learning",
                    "Human-Object Interaction",
                    "Visual-Language Models",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Exploring the Potential of Large Foundation Models for Open-Vocabulary HOI Detection",
                        "abstract": "Open-vocabulary human-object interaction (HOI) detection, which is concerned with the problem of detecting novel HOIs guided by natural language, is crucial for understanding human-centric scenes. However, prior zero-shot HOI detectors often employ the same levels of feature maps to model HOIs with varying distances, leading to suboptimal performance in scenes containing human-object pairs with a wide range of distances. In addition, these detectors primarily rely on category names and overlook the rich contextual information that language can provide, which is essential for capturing open vocabulary concepts that are typically rare and not well-represented by category names alone. In this paper, we introduce a novel end-to-end open vocabulary HOI detection framework with conditional multi-level decoding and fine-grained semantic enhancement (CMD-SE), harnessing the potential of Visual-Language Models (VLMs). Specifically, we propose to model human-object pairs with different distances with different levels of feature maps by incorporating a soft constraint during the bipartite matching process. Furthermore, by leveraging large language models (LLMs) such as GPT models, we exploit their extensive world knowledge to generate descriptions of human body part states for various interactions. Then we integrate the generalizable and fine-grained semantics of human body parts to improve interaction recognition. Experimental results on two datasets, SWIG-HOI and HICO-DET, demonstrate that our proposed method achieves state-of-the-art results in open vocabulary HOI detection. The code and models are available at https://github.com/ltttpku/CMD-SE-release."
                    },
                    {
                        "title": "Exploring Conditional Multi-Modal Prompts for Zero-shot HOI Detection",
                        "abstract": "Zero-shot Human-Object Interaction (HOI) detection has emerged as a frontier topic due to its capability to detect HOIs beyond a predefined set of categories. This task entails not only identifying the interactiveness of human-object pairs and localizing them but also recognizing both seen and unseen interaction categories. In this paper, we introduce a novel framework for zero-shot HOI detection using Conditional Multi-Modal Prompts, namely CMMP. This approach enhances the generalization of large foundation models, such as CLIP, when fine-tuned for HOI detection. Unlike traditional prompt-learning methods, we propose learning decoupled vision and language prompts for interactiveness-aware visual feature extraction and generalizable interaction classification, respectively. Specifically, we integrate prior knowledge of different granularity into conditional vision prompts, including an input-conditioned instance prior and a global spatial pattern prior. The former encourages the image encoder to treat instances belonging to seen or potentially unseen HOI concepts equally while the latter provides representative plausible spatial configuration of the human and object under interaction. Besides, we employ language-aware prompt learning with a consistency constraint to preserve the knowledge of the large foundation model to enable better generalization in the text branch. Extensive experiments demonstrate the efficacy of our detector with conditional multi-modal prompts, outperforming previous state-of-the-art on unseen classes of various zero-shot settings. The code and models are available at \\url{https://github.com/ltttpku/CMMP}."
                    }
                ]
            },
            "a25ba826-7dc5-48ba-b62d-1b6039d2bda9": {
                "pk": "a25ba826-7dc5-48ba-b62d-1b6039d2bda9",
                "name": "Ningya Feng",
                "collaborators": [
                    "Jialong Wu",
                    "Mingsheng Long",
                    "Jianmin Wang",
                    "Chaoyi Deng",
                    "Junwei Pan",
                    "Baixu Chen",
                    "Ximei Wang",
                    "Qian Li",
                    "Xian Hu",
                    "Jie Jiang"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Code Optimization",
                    "Recommendation Systems",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "CompilerDream: Learning a Compiler World Model for General Code Optimization",
                        "abstract": "Effective code optimization in compilers is crucial for computer and software engineering. The success of these optimizations primarily depends on the selection and ordering of the optimization passes applied to the code. While most compilers rely on a fixed sequence of optimization passes, current methods to find the optimal sequence either employ impractically slow search algorithms or learning methods that struggle to generalize to code unseen during training. We introduce CompilerDream, a model-based reinforcement learning approach to general code optimization. CompilerDream comprises a compiler world model that accurately simulates the intrinsic properties of optimization passes and an agent trained on this model to produce effective optimization strategies. By training on a large-scale program dataset, CompilerDream is equipped to serve as a general code optimizer across various application scenarios and source-code languages. Our extensive experiments first highlight CompilerDream's strong optimization capabilities for autotuning, where it leads the CompilerGym leaderboard. More importantly, the zero-shot generalization ability of large-scale trained compiler world model and agent, excels across diverse datasets, surpassing LLVM's built-in optimizations and other state-of-the-art methods in both settings of value prediction and end-to-end code optimization."
                    },
                    {
                        "title": "Long-Sequence Recommendation Models Need Decoupled Embeddings",
                        "abstract": "Lifelong user behavior sequences, comprising up to tens of thousands of history behaviors, are crucial for capturing user interests and predicting user responses in modern recommendation systems. A two-stage paradigm is typically adopted to handle these long sequences: a few relevant behaviors are first searched from the original long sequences via an attention mechanism in the first stage and then aggregated with the target item to construct a discriminative representation for prediction in the second stage. In this work, we identify and characterize, for the first time, a neglected deficiency in existing long-sequence recommendation models: a single set of embeddings struggles with learning both attention and representation, leading to interference between these two processes. Initial attempts to address this issue using linear projections -- a technique borrowed from language processing -- proved ineffective, shedding light on the unique challenges of recommendation models. To overcome this, we propose the Decoupled Attention and Representation Embeddings (DARE) model, where two distinct embedding tables are initialized and learned separately to fully decouple attention and representation. Extensive experiments and analysis demonstrate that DARE provides more accurate search of correlated behaviors and outperforms baselines with AUC gains up to 0.9% on public datasets and notable online system improvements. Furthermore, decoupling embedding spaces allows us to reduce the attention embedding dimension and accelerate the search procedure by 50% without significant performance impact, enabling more efficient, high-performance online serving."
                    },
                    {
                        "title": "HarmonyDream: Task Harmonization Inside World Models",
                        "abstract": "Model-based reinforcement learning (MBRL) holds the promise of sample-efficient learning by utilizing a world model, which models how the environment works and typically encompasses components for two tasks: observation modeling and reward modeling. In this paper, through a dedicated empirical investigation, we gain a deeper understanding of the role each task plays in world models and uncover the overlooked potential of sample-efficient MBRL by mitigating the domination of either observation or reward modeling. Our key insight is that while prevalent approaches of explicit MBRL attempt to restore abundant details of the environment via observation models, it is difficult due to the environment's complexity and limited model capacity. On the other hand, reward models, while dominating implicit MBRL and adept at learning compact task-centric dynamics, are inadequate for sample-efficient learning without richer learning signals. Motivated by these insights and discoveries, we propose a simple yet effective approach, HarmonyDream, which automatically adjusts loss coefficients to maintain task harmonization, i.e. a dynamic equilibrium between the two tasks in world model learning. Our experiments show that the base MBRL method equipped with HarmonyDream gains 10%-69% absolute performance boosts on visual robotic tasks and sets a new state-of-the-art result on the Atari 100K benchmark. Code is available at https://github.com/thuml/HarmonyDream."
                    }
                ]
            },
            "dc5fe171-d791-4a17-b216-d5fd2cb42010": {
                "pk": "dc5fe171-d791-4a17-b216-d5fd2cb42010",
                "name": "Xu He",
                "collaborators": [
                    "Herbert Jaeger"
                ],
                "domain": [
                    "Experimental Design",
                    "Computer Experiments",
                    "Continual Learning",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Interleaved lattice-based maximin distance designs",
                        "abstract": "We propose a new method to construct maximin distance designs with arbitrary number of dimensions and points. The proposed designs hold interleaved-layer structures and are by far the best maximin distance designs in four or more dimensions. Applicable to distance measures with equal or unequal weights, our method is useful for emulating computer experiments when a relatively accurate priori guess on the variable importance is available."
                    },
                    {
                        "title": "Lattice-based designs possessing quasi-optimal separation distance on all projections",
                        "abstract": "Experimental designs that spread out points apart from each other on projections are important for computer experiments when not necessarily all factors have substantial influence on the response. We provide a theoretical framework to generate designs that possess quasi-optimal separation distance on all of the projections and quasi-optimal fill distance on univariate margins. The key is to use special techniques to rotate certain lattices. One such type of design is densest packing-based maximum projection designs, which outperform existing types of space-filling designs in many scenarios. Computer code to generate these designs is provided in R package LatticeDesign."
                    },
                    {
                        "title": "Rotated sphere packing designs",
                        "abstract": "We propose a new class of space-filling designs called rotated sphere packing designs for computer experiments. The approach starts from the asymptotically optimal positioning of identical balls that covers the unit cube. Properly scaled, rotated, translated and extracted, such designs are excellent in maximin distance criterion, low in discrepancy, good in projective uniformity and thus useful in both prediction and numerical integration purposes. We provide a fast algorithm to construct such designs for any numbers of dimensions and points with R codes available online. Theoretical and numerical results are also provided."
                    },
                    {
                        "title": "Sliced rotated sphere packing designs",
                        "abstract": "Space-filling designs are popular choices for computer experiments. A sliced design is a design that can be partitioned into several subdesigns. We propose a new type of sliced space-filling design called sliced rotated sphere packing designs. Their full designs and subdesigns are rotated sphere packing designs. They are constructed by rescaling, rotating, translating and extracting the points from a sliced lattice. We provide two fast algorithms to generate such designs. Furthermore, we propose a strategy to use sliced rotated sphere packing designs adaptively. Under this strategy, initial runs are uniformly distributed in the design space, follow-up runs are added by incorporating information gained from initial runs, and the combined design is space-filling for any local region. Examples are given to illustrate its potential application."
                    },
                    {
                        "title": "Overcoming Catastrophic Interference by Conceptors",
                        "abstract": "Catastrophic interference has been a major roadblock in the research of continual learning. Here we propose a variant of the back-propagation algorithm, \"conceptor-aided back-prop\" (CAB), in which gradients are shielded by conceptors against degradation of previously learned tasks. Conceptors have their origin in reservoir computing, where they have been previously shown to overcome catastrophic forgetting. CAB extends these results to deep feedforward networks. On the disjoint MNIST task CAB outperforms two other methods for coping with catastrophic interference that have recently been proposed in the deep learning field."
                    }
                ]
            },
            "eebc902c-c17b-4cac-af51-d0adbc196e09": {
                "pk": "eebc902c-c17b-4cac-af51-d0adbc196e09",
                "name": "Dong Li",
                "collaborators": [
                    "Hongjie Dong"
                ],
                "domain": [
                    "Partial Differential Equations",
                    "Mathematical Physics",
                    "Fluid Dynamics",
                    "Nonlinear Analysis"
                ],
                "publications": [
                    {
                        "title": "A regularization-free approach to the Cahn-Hilliard equation with logarithmic potentials",
                        "abstract": "We introduce a regularization-free approach for the wellposedness of the classic Cahn-Hilliard equation with logarithmic potentials."
                    },
                    {
                        "title": "On a nonlinear recurrent relation",
                        "abstract": "We study the limiting behavior for the solutions of a nonlinear recurrent relation which arises from the study of Navier-Stokes equations. Some stability theorems are also shown concerning a related class of linear recurrent relations."
                    },
                    {
                        "title": "On fractional smoothness of modulus of functions",
                        "abstract": "We consider the Nemytskii operators $u\\to |u|$ and $u\\to u^{\\pm}$ in a bounded domain $\\Omega$ with $C^2$ boundary. We give elementary proofs of the boundedness in $H^s(\\Omega)$ with $0\\le s<3/2$."
                    },
                    {
                        "title": "Global wellposedness of hedgehog solutions for the (3+1) Skyrme model",
                        "abstract": "We consider the dynamics of hedgehog solutions in the $(3+1)$-dimensional Skyrme model which is an energy-supercritical problem. We introduce a new strategy to prove global wellposedness for arbitrarily large initial data."
                    },
                    {
                        "title": "Optimal Gevrey regularity for supercritical quasi-geostrophic equations",
                        "abstract": "We consider the two dimensional surface quasi-geostrophic equations with super-critical dissipation. For large initial data in critical Sobolev and Besov spaces, we prove optimal Gevrey regularity with the same decay exponent as the linear part. This settles several open problems in \\cite{Bis14, BMS15}."
                    },
                    {
                        "title": "Uniform estimates for 2D quasilinear wave",
                        "abstract": "We consider two-dimensional quasilinear wave equations with standard null-type quadratic nonlinearities. In 2001 Alinhac proved that such systems possess global in time solutions for compactly supported initial data with sufficiently small Sobolev norm. The highest norm of the constructed solution grows polynomially in time. In this work we develop a new strategy and prove uniform boundedness of the highest order norm of the solution for all time."
                    },
                    {
                        "title": "Moving structures in ultraviolet bright points: observations from Solar Orbiter/EUI",
                        "abstract": "Moving structures have been detected in coronal bright points and in a solar flare in active regions, which were bi-directional, symmetrical, simultaneous, and quasi-periodic (Ning & Guo 2014; Ning 2016; Li et al. 2016a). They could be regarded as observational evidence of plasma outflows via magnetic reconnection. In this article, we explored pairs of moving structures in fifteen ultraviolet bright points (UBPs), which were observed in the quiet Sun or inside a small active region on 19 November 2020, and measured by the High Resolution (HRI) Telescopes of the Extreme Ultraviolet Imager (EUI) on board the Solar Orbiter (SolO) in two passbands, HRIEUV 174 {\\AA} and HRILy{\\alpha} 1216 {\\AA}. Moving structures observed in ten UBPs as starting from their bright cores and propagating toward two ends, are interpreted as diverging motions of bi-directional moving structures. These moving structures are also characterized by simultaneity and symmetry and in the case of seven UBPs they exhibit quasi-periodicity. They could be generated by outflows after magnetic reconnections. Moving structures seen in another five UBPs as originating from double ends and moving closer, and merging together, are manifested as converging motions, which might be caused by inflows through the magnetic reconnection, or might be interpreted as upflows driven by the chromospheric evaporation."
                    },
                    {
                        "title": "On a frequency localized Bernstein inequality and some generalized Poincare-type inequalities",
                        "abstract": "We consider a frequency localized Bernstein inequality for the fractional Laplacian operator which has wide applications in fluid dynamics such as dissipative surface quasi-geostrophic equations. We use a heat flow reformulation and prove the inequality for the full range of parameters and in all dimensions. A crucial observation is that after frequency projection the zero frequency part of the L\\'{e}vy semigroup does not participate in the inequality and therefore can be freely adjusted. Our proof is based on this idea and a careful perturbation of the L\\'{e}vy semigroup near the zero frequency which preserves the positivity and improves the time decay. Several alternative proofs (with weaker results) are also included. As an application we also give new proofs of some generalized Poincare type inequalities."
                    },
                    {
                        "title": "On Kato-Ponce and fractional Leibniz",
                        "abstract": "We show that in the Kato-Ponce inequality $\\|J^s(fg)-fJ^s g\\|_p \\lesssim \\| \\partial f \\|_{\\infty} \\| J^{s-1} g \\|_p + \\| J^s f \\|_p \\|g\\|_{\\infty}$, the $J^s f$ term on the RHS can be replaced by $J^{s-1} \\partial f$. This solves a question raised in Kato-Ponce \\cite{KP88}. We propose and prove a new fractional Leibniz rule for $D^s=(-\\Delta)^{s/2}$ and similar operators, generalizing the Kenig-Ponce-Vega estimate \\cite{KPV93} to all $s>0$. We also prove a family of generalized and refined Kato-Ponce type inequalities which include many commutator estimates as special cases. To showcase the sharpness of the estimates at various endpoint cases, we construct several counterexamples. In particular, we show that in the original Kato-Ponce inequality, the $L^{\\infty}$-norm on the RHS cannot be replaced by the weaker BMO norm. Some divergence-free counterexamples are also included."
                    },
                    {
                        "title": "Transition-region explosive events produced by plasmoid instability",
                        "abstract": "Magnetic reconnection is thought to be a key process in most of solar eruptions. Thanks to high-resolution observations and simulations, the studied scale of reconnection process has become smaller and smaller. Spectroscopic observations show that the reconnection site can be very small, which always exhibits a bright core and two extended wings with fast speeds, i.e., transition-region explosive events. In this paper, using the PLUTO code, we perform a 2-D magnetohydrodynamic simulation to investigate the small-scale reconnection in double current sheets. Based on our simulation results, such as the line-of-sight velocity, number density and plasma temperature, we can synthesize the line profile of Si IV 1402.77 A which is a well known emission line to study the transition-region explosive events on the Sun. The synthetic line profile of Si IV 1402.77 A is complex with a bright core and two broad wings which can extend to be nearly 200 km/s. Our simulation results suggest that the transition-region explosive events on the Sun are produced by plasmoid instability during the small-scale magnetic reconnection."
                    },
                    {
                        "title": "Why large time-stepping methods for the Cahn-Hilliard equation is stable",
                        "abstract": "We consider the Cahn-Hilliard equation with standard double-well potential. We employ a prototypical class of first order in time semi-implicit methods with implicit treatment of the linear dissipation term and explicit extrapolation of the nonlinear term. When the dissipation coefficient is held small, a conventional wisdom is to add a judiciously chosen stabilization term in order to afford relatively large time stepping and speed up the simulation. In practical numerical implementations it has been long observed that the resulting system exhibits remarkable stability properties in the regime where the stabilization parameter is $\\mathcal O(1)$, the dissipation coefficient is vanishingly small and the size of the time step is moderately large. In this work we develop a new stability theory to address this perplexing phenomenon."
                    },
                    {
                        "title": "Statistical properties of the Bipolar Magnetic Regions",
                        "abstract": "Using the observations from Michelson Doppler Imager (MDI) onboard Solar and Heliospheric Observatory (SOHO), we develop a computational algorithm to automatically identify the bipolar magnetic regions (BMRs) in the active regions, and then study their statistical properties. The individual magnetic (positive or negative) pole of the BMR is determined from the region with an absolute strength above 55 G and with an area above 250 pixel^2 (495 Mm^2), while a BMR is identified as a pair of positive and negative poles with a shortest area-weight distance between them. Based on this method, 2234 BMRs are identified from the MDI synoptic magnetograms between the Carrington Rotation 1909 (1996 May 06) and 2104 (2010 December 10). 1005 of them are located in the northern hemisphere, while the other 1229 are in the southern hemisphere.We find that the BMR paraments (e.g., latitudes, separations, fragments, and strength) are similar to those of active regions (ARs). Moreover, based on the maximum likelihood estimation (MLE)method, the frequency distributions of these BMRs occurrence as functions of area size and magnetic flux exhibit a power-law behavior. We also find that their orientation angles follow the Hale's Polarity Laws and deviate slightly to the solar equator direction. Consistent with the previous findings, we obtain the orientation angles dependence on the latitudes for the normal BMRs during the 23rd solar cycle. The north-south asymmetry of these BMRs is also detected here."
                    },
                    {
                        "title": "Effective maximum principles for spectral methods",
                        "abstract": "Many physical problems such as Allen-Cahn flows have natural maximum principles which yield strong point-wise control of the physical solutions in terms of the boundary data, the initial conditions and the operator coefficients. Sharp/strict maximum principles insomuch of fundamental importance for the continuous problem often do not persist under numerical discretization. A lot of past research concentrates on designing fine numerical schemes which preserves the sharp maximum principles especially for nonlinear problems. However these sharp principles not only sometimes introduce unwanted stringent conditions on the numerical schemes but also completely leaves many powerful frequency-based methods unattended and rarely analyzed directly in the sharp maximum norm topology. A prominent example is the spectral methods in the family of weighted residual methods.   In this work we introduce and develop a new framework of almost sharp maximum principles which allow the numerical solutions to deviate from the sharp bound by a controllable discretization error: we call them effective maximum principles. We showcase the analysis for the classical Fourier spectral methods including Fourier Galerkin and Fourier collocation in space with forward Euler in time or second order Strang splitting. The model equations include the Allen-Cahn equations with double well potential, the Burgers equation and the Navier-Stokes equations. We give a comprehensive proof of the effective maximum principles under very general parametric conditions."
                    },
                    {
                        "title": "On the Analyticity of Solutions to the Navier-Stokes Equations with Fractional Dissipation",
                        "abstract": "By using a new bilinear estimate, a pointwise estimate of the generalized Oseen kernel and an idea of fractional bootstrap, we show in this note that solutions to the Navier-Stokes equations with fractional dissipation are analytic in space variables."
                    },
                    {
                        "title": "Global $\\dot H^1 \\cap \\dot H^{-1}$ solutions to a logarithmically regularized 2D Euler equation",
                        "abstract": "We construct global $\\dot H^1\\cap \\dot H^{-1}$ solutions to a logarithmically modified 2D Euler vorticity equation. Our main tool is a new logarithm interpolation inequality which exploits the $L^{\\infty-}$-conservation of the vorticity."
                    },
                    {
                        "title": "On a generalized maximum principle for a transport-diffusion model with $\\log$-modulated fractional dissipation",
                        "abstract": "We consider a transport-diffusion equation of the form $\\partial_t \\theta +v \\cdot \\nabla \\theta + \\nu \\A \\theta =0$, where $v$ is a given time-dependent vector field on $\\mathbb R^d$. The operator $\\A$ represents log-modulated fractional dissipation: $\\A=\\frac {|\\nabla|^{\\gamma}}{\\log^{\\beta}(\\lambda+|\\nabla|)}$ and the parameters $\\nu\\ge 0$, $\\beta\\ge 0$, $0\\le \\gamma \\le 2$, $\\lambda>1$. We introduce a novel nonlocal decomposition of the operator $\\A$ in terms of a weighted integral of the usual fractional operators $|\\nabla|^{s}$, $0\\le s \\le \\gamma$ plus a smooth remainder term which corresponds to an $L^1$ kernel. For a general vector field $v$ (possibly non-divergence-free) we prove a generalized $L^\\infty$ maximum principle of the form $ |\\theta(t)|_\\infty \\le e^{Ct} |\\theta_0|_{\\infty}$ where the constant $C=C(\\nu,\\beta,\\gamma)>0$. In the case $\\text{div}(v)=0$ the same inequality holds for $|\\theta(t)|_p$ with $1\\le p \\le \\infty$. At the cost of an exponential factor, this extends a recent result of Hmidi (2011) to the full regime $d\\ge 1$, $0\\le \\gamma \\le 2$ and removes the incompressibility assumption in the $L^\\infty$ case."
                    },
                    {
                        "title": "On a one-dimensional \u03b1-patch model with nonlocal drift and fractional dissipation",
                        "abstract": "We consider a one-dimensional nonlocal nonlinear equation of the form: $\\partial_t u = (\\Lambda^{-\\alpha} u)\\partial_x u - \\nu \\Lambda^{\\beta}u$ where $\\Lambda =(-\\partial_{xx})^{\\frac 12}$ is the fractional Laplacian and $\\nu\\ge 0$ is the viscosity coefficient. We consider primarily the regime $0<\\alpha<1$ and $0\\le \\beta \\le 2$ for which the model has nonlocal drift, fractional dissipation, and captures essential features of the 2D $\\alpha$-patch models. In the critical and subcritical range $1-\\alpha\\le \\beta \\le 2$, we prove global wellposedness for arbitrarily large initial data in Sobolev spaces. In the full supercritical range $0 \\le \\beta<1-\\alpha$, we prove formation of singularities in finite time for a class of smooth initial data. Our proof is based on a novel nonlocal weighted inequality which can be of independent interest."
                    }
                ]
            },
            "6d3eafa7-a8bc-4393-b09b-125fa1b16516": {
                "pk": "6d3eafa7-a8bc-4393-b09b-125fa1b16516",
                "name": "Jianye Hao",
                "collaborators": [
                    "Hongyao Tang",
                    "Yan Zheng",
                    "Yifu Yuan",
                    "Weixun Wang",
                    "Min Zhang",
                    "Dongbin Zhao",
                    "Mengchen Zhao",
                    "Yixi Wang",
                    "Matthew Taylor",
                    "Zongzhang Zhang"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Multiagent Systems",
                    "Deep Learning",
                    "Game Theory"
                ],
                "publications": [
                    {
                        "title": "Towards Cooperation in Sequential Prisoner's Dilemmas: a Deep Multiagent Reinforcement Learning Approach",
                        "abstract": "The Iterated Prisoner's Dilemma has guided research on social dilemmas for decades. However, it distinguishes between only two atomic actions: cooperate and defect. In real-world prisoner's dilemmas, these choices are temporally extended and different strategies may correspond to sequences of actions, reflecting grades of cooperation. We introduce a Sequential Prisoner's Dilemma (SPD) game to better capture the aforementioned characteristics. In this work, we propose a deep multiagent reinforcement learning approach that investigates the evolution of mutual cooperation in SPD games. Our approach consists of two phases. The first phase is offline: it synthesizes policies with different cooperation degrees and then trains a cooperation degree detection network. The second phase is online: an agent adaptively selects its policy based on the detected degree of opponent cooperation. The effectiveness of our approach is demonstrated in two representative SPD 2D games: the Apple-Pear game and the Fruit Gathering game. Experimental results show that our strategy can avoid being exploited by exploitative opponents and achieve cooperation with cooperative opponents."
                    },
                    {
                        "title": "Weighted Double Deep Multiagent Reinforcement Learning in Stochastic Cooperative Environments",
                        "abstract": "Recently, multiagent deep reinforcement learning (DRL) has received increasingly wide attention. Existing multiagent DRL algorithms are inefficient when facing with the non-stationarity due to agents update their policies simultaneously in stochastic cooperative environments. This paper extends the recently proposed weighted double estimator to the multiagent domain and propose a multiagent DRL framework, named weighted double deep Q-network (WDDQN). By utilizing the weighted double estimator and the deep neural network, WDDQN can not only reduce the bias effectively but also be extended to scenarios with raw visual inputs. To achieve efficient cooperation in the multiagent domain, we introduce the lenient reward network and the scheduled replay strategy. Experiments show that the WDDQN outperforms the existing DRL and multiaent DRL algorithms, i.e., double DQN and lenient Q-learning, in terms of the average reward and the convergence rate in stochastic cooperative environments."
                    },
                    {
                        "title": "Critic PI2: Master Continuous Planning via Policy Improvement with Path Integrals and Deep Actor-Critic Reinforcement Learning",
                        "abstract": "Constructing agents with planning capabilities has long been one of the main challenges in the pursuit of artificial intelligence. Tree-based planning methods from AlphaGo to Muzero have enjoyed huge success in discrete domains, such as chess and Go. Unfortunately, in real-world applications like robot control and inverted pendulum, whose action space is normally continuous, those tree-based planning techniques will be struggling. To address those limitations, in this paper, we present a novel model-based reinforcement learning frameworks called Critic PI2, which combines the benefits from trajectory optimization, deep actor-critic learning, and model-based reinforcement learning. Our method is evaluated for inverted pendulum models with applicability to many continuous control systems. Extensive experiments demonstrate that Critic PI2 achieved a new state of the art in a range of challenging continuous domains. Furthermore, we show that planning with a critic significantly increases the sample efficiency and real-time performance. Our work opens a new direction toward learning the components of a model-based planning system and how to use them."
                    },
                    {
                        "title": "ED2: Environment Dynamics Decomposition World Models for Continuous Control",
                        "abstract": "Model-based reinforcement learning (MBRL) achieves significant sample efficiency in practice in comparison to model-free RL, but its performance is often limited by the existence of model prediction error. To reduce the model error, standard MBRL approaches train a single well-designed network to fit the entire environment dynamics, but this wastes rich information on multiple sub-dynamics which can be modeled separately, allowing us to construct the world model more accurately. In this paper, we propose the Environment Dynamics Decomposition (ED2), a novel world model construction framework that models the environment in a decomposing manner. ED2 contains two key components: sub-dynamics discovery (SD2) and dynamics decomposition prediction (D2P). SD2 discovers the sub-dynamics in an environment automatically and then D2P constructs the decomposed world model following the sub-dynamics. ED2 can be easily combined with existing MBRL algorithms and empirical results show that ED2 significantly reduces the model error, increases the sample efficiency, and achieves higher asymptotic performance when combined with the state-of-the-art MBRL algorithms on various continuous control tasks. Our code is open source and available at https://github.com/ED2-source-code/ED2."
                    },
                    {
                        "title": "Reweighted Interacting Langevin Diffusions: an Accelerated Sampling Methodfor Optimization",
                        "abstract": "We proposed a new technique to accelerate sampling methods for solving difficult optimization problems. Our method investigates the intrinsic connection between posterior distribution sampling and optimization with Langevin dynamics, and then we propose an interacting particle scheme that approximates a Reweighted Interacting Langevin Diffusion system (RILD). The underlying system is designed by adding a multiplicative source term into the classical Langevin operator, leading to a higher convergence rate and a more concentrated invariant measure. We analyze the convergence rate of our algorithm and the improvement compared to existing results in the asymptotic situation. We also design various tests to verify our theoretical results, showing the advantages of accelerating convergence and breaking through barriers of suspicious local minimums, especially in high-dimensional non-convex settings. Our algorithms and analysis shed some light on combining gradient and genetic algorithms using Partial Differential Equations (PDEs) with provable guarantees."
                    },
                    {
                        "title": "Spectral Augmentations for Graph Contrastive Learning",
                        "abstract": "Contrastive learning has emerged as a premier method for learning representations with or without supervision. Recent studies have shown its utility in graph representation learning for pre-training. Despite successes, the understanding of how to design effective graph augmentations that can capture structural properties common to many different types of downstream graphs remains incomplete. We propose a set of well-motivated graph transformation operations derived via graph spectral analysis to provide a bank of candidates when constructing augmentations for a graph contrastive objective, enabling contrastive learning to capture useful structural representation from pre-training graph datasets. We first present a spectral graph cropping augmentation that involves filtering nodes by applying thresholds to the eigenvalues of the leading Laplacian eigenvectors. Our second novel augmentation reorders the graph frequency components in a structural Laplacian-derived position graph embedding. Further, we introduce a method that leads to improved views of local subgraphs by performing alignment via global random walk embeddings. Our experimental results indicate consistent improvements in out-of-domain graph data transfer compared to state-of-the-art graph contrastive learning methods, shedding light on how to design a graph learner that is able to learn structural properties common to diverse graph types."
                    },
                    {
                        "title": "Rethinking Decision Transformer via Hierarchical Reinforcement Learning",
                        "abstract": "Decision Transformer (DT) is an innovative algorithm leveraging recent advances of the transformer architecture in reinforcement learning (RL). However, a notable limitation of DT is its reliance on recalling trajectories from datasets, losing the capability to seamlessly stitch sub-optimal trajectories together. In this work we introduce a general sequence modeling framework for studying sequential decision making through the lens of Hierarchical RL. At the time of making decisions, a high-level policy first proposes an ideal prompt for the current state, a low-level policy subsequently generates an action conditioned on the given prompt. We show DT emerges as a special case of this framework with certain choices of high-level and low-level policies, and discuss the potential failure of these choices. Inspired by these observations, we study how to jointly optimize the high-level and low-level policies to enable the stitching ability, which further leads to the development of new offline RL algorithms. Our empirical results clearly show that the proposed algorithms significantly surpass DT on several control and navigation benchmarks. We hope our contributions can inspire the integration of transformer architectures within the field of RL."
                    },
                    {
                        "title": "Towards A Unified Policy Abstraction Theory and Representation Learning Approach in Markov Decision Processes",
                        "abstract": "Lying on the heart of intelligent decision-making systems, how policy is represented and optimized is a fundamental problem. The root challenge in this problem is the large scale and the high complexity of policy space, which exacerbates the difficulty of policy learning especially in real-world scenarios. Towards a desirable surrogate policy space, recently policy representation in a low-dimensional latent space has shown its potential in improving both the evaluation and optimization of policy. The key question involved in these studies is by what criterion we should abstract the policy space for desired compression and generalization. However, both the theory on policy abstraction and the methodology on policy representation learning are less studied in the literature. In this work, we make very first efforts to fill up the vacancy. First, we propose a unified policy abstraction theory, containing three types of policy abstraction associated to policy features at different levels. Then, we generalize them to three policy metrics that quantify the distance (i.e., similarity) of policies, for more convenient use in learning policy representation. Further, we propose a policy representation learning approach based on deep metric learning. For the empirical study, we investigate the efficacy of the proposed policy metrics and representations, in characterizing policy difference and conveying policy generalization respectively. Our experiments are conducted in both policy optimization and evaluation problems, containing trust-region policy optimization (TRPO), diversity-guided evolution strategy (DGES) and off-policy evaluation (OPE). Somewhat naturally, the experimental results indicate that there is no a universally optimal abstraction for all downstream learning problems; while the influence-irrelevance policy abstraction can be a generally preferred choice."
                    },
                    {
                        "title": "The Ladder in Chaos: A Simple and Effective Improvement to General DRL Algorithms by Policy Path Trimming and Boosting",
                        "abstract": "Knowing the learning dynamics of policy is significant to unveiling the mysteries of Reinforcement Learning (RL). It is especially crucial yet challenging to Deep RL, from which the remedies to notorious issues like sample inefficiency and learning instability could be obtained. In this paper, we study how the policy networks of typical DRL agents evolve during the learning process by empirically investigating several kinds of temporal change for each policy parameter. On typical MuJoCo and DeepMind Control Suite (DMC) benchmarks, we find common phenomena for TD3 and RAD agents: 1) the activity of policy network parameters is highly asymmetric and policy networks advance monotonically along very few major parameter directions; 2) severe detours occur in parameter update and harmonic-like changes are observed for all minor parameter directions. By performing a novel temporal SVD along policy learning path, the major and minor parameter directions are identified as the columns of right unitary matrix associated with dominant and insignificant singular values respectively. Driven by the discoveries above, we propose a simple and effective method, called Policy Path Trimming and Boosting (PPTB), as a general plug-in improvement to DRL algorithms. The key idea of PPTB is to periodically trim the policy learning path by canceling the policy updates in minor parameter directions, while boost the learning path by encouraging the advance in major directions. In experiments, we demonstrate the general and significant performance improvements brought by PPTB, when combined with TD3 and RAD in MuJoCo and DMC environments respectively."
                    },
                    {
                        "title": "MENTOR: Guiding Hierarchical Reinforcement Learning with Human Feedback and Dynamic Distance Constraint",
                        "abstract": "Hierarchical reinforcement learning (HRL) provides a promising solution for complex tasks with sparse rewards of intelligent agents, which uses a hierarchical framework that divides tasks into subgoals and completes them sequentially. However, current methods struggle to find suitable subgoals for ensuring a stable learning process. Without additional guidance, it is impractical to rely solely on exploration or heuristics methods to determine subgoals in a large goal space. To address the issue, We propose a general hierarchical reinforcement learning framework incorporating human feedback and dynamic distance constraints (MENTOR). MENTOR acts as a \"mentor\", incorporating human feedback into high-level policy learning, to find better subgoals. As for low-level policy, MENTOR designs a dual policy for exploration-exploitation decoupling respectively to stabilize the training. Furthermore, although humans can simply break down tasks into subgoals to guide the right learning direction, subgoals that are too difficult or too easy can still hinder downstream learning efficiency. We propose the Dynamic Distance Constraint (DDC) mechanism dynamically adjusting the space of optional subgoals. Thus MENTOR can generate subgoals matching the low-level policy learning process from easy to hard. Extensive experiments demonstrate that MENTOR uses a small amount of human feedback to achieve significant improvement in complex tasks with sparse rewards."
                    },
                    {
                        "title": "MODULI: Unlocking Preference Generalization via Diffusion Models for Offline Multi-Objective Reinforcement Learning",
                        "abstract": "Multi-objective Reinforcement Learning (MORL) seeks to develop policies that simultaneously optimize multiple conflicting objectives, but it requires extensive online interactions. Offline MORL provides a promising solution by training on pre-collected datasets to generalize to any preference upon deployment. However, real-world offline datasets are often conservatively and narrowly distributed, failing to comprehensively cover preferences, leading to the emergence of out-of-distribution (OOD) preference areas. Existing offline MORL algorithms exhibit poor generalization to OOD preferences, resulting in policies that do not align with preferences. Leveraging the excellent expressive and generalization capabilities of diffusion models, we propose MODULI (Multi-objective Diffusion Planner with Sliding Guidance), which employs a preference-conditioned diffusion model as a planner to generate trajectories that align with various preferences and derive action for decision-making. To achieve accurate generation, MODULI introduces two return normalization methods under diverse preferences for refining guidance. To further enhance generalization to OOD preferences, MODULI proposes a novel sliding guidance mechanism, which involves training an additional slider adapter to capture the direction of preference changes. Incorporating the slider, it transitions from in-distribution (ID) preferences to generating OOD preferences, patching, and extending the incomplete Pareto front. Extensive experiments on the D4MORL benchmark demonstrate that our algorithm outperforms state-of-the-art Offline MORL baselines, exhibiting excellent generalization to OOD preferences."
                    },
                    {
                        "title": "Independent Generative Adversarial Self-Imitation Learning in Cooperative Multiagent Systems",
                        "abstract": "Many tasks in practice require the collaboration of multiple agents through reinforcement learning. In general, cooperative multiagent reinforcement learning algorithms can be classified into two paradigms: Joint Action Learners (JALs) and Independent Learners (ILs). In many practical applications, agents are unable to observe other agents' actions and rewards, making JALs inapplicable. In this work, we focus on independent learning paradigm in which each agent makes decisions based on its local observations only. However, learning is challenging in independent settings due to the local viewpoints of all agents, which perceive the world as a non-stationary environment due to the concurrently exploring teammates. In this paper, we propose a novel framework called Independent Generative Adversarial Self-Imitation Learning (IGASIL) to address the coordination problems in fully cooperative multiagent environments. To the best of our knowledge, we are the first to combine self-imitation learning with generative adversarial imitation learning (GAIL) and apply it to cooperative multiagent systems. Besides, we put forward a Sub-Curriculum Experience Replay mechanism to pick out the past beneficial experiences as much as possible and accelerate the self-imitation learning process. Evaluations conducted in the testbed of StarCraft unit micromanagement and a commonly adopted benchmark show that our IGASIL produces state-of-the-art results and even outperforms JALs in terms of both convergence speed and final performance."
                    },
                    {
                        "title": "Diverse Behavior Is What Game AI Needs: Generating Varied Human-Like Playing Styles Using Evolutionary Multi-Objective Deep Reinforcement Learning",
                        "abstract": "this paper has been withdrawn"
                    },
                    {
                        "title": "GFlowNets with Human Feedback",
                        "abstract": "We propose the GFlowNets with Human Feedback (GFlowHF) framework to improve the exploration ability when training AI models. For tasks where the reward is unknown, we fit the reward function through human evaluations on different trajectories. The goal of GFlowHF is to learn a policy that is strictly proportional to human ratings, instead of only focusing on human favorite ratings like RLHF. Experiments show that GFlowHF can achieve better exploration ability than RLHF."
                    },
                    {
                        "title": "Falsification of Cyber-Physical Systems Using Deep Reinforcement Learning",
                        "abstract": "With the rapid development of software and distributed computing, Cyber-Physical Systems (CPS) are widely adopted in many application areas, e.g., smart grid, autonomous automobile. It is difficult to detect defects in CPS models due to the complexities involved in the software and physical systems. To find defects in CPS models efficiently, robustness guided falsification of CPS is introduced. Existing methods use several optimization techniques to generate counterexamples, which falsify the given properties of a CPS. However those methods may require a large number of simulation runs to find the counterexample and is far from practical. In this work, we explore state-of-the-art Deep Reinforcement Learning (DRL) techniques to reduce the number of simulation runs required to find such counterexamples. We report our method and the preliminary evaluation results."
                    },
                    {
                        "title": "An Optimal Rewiring Strategy for Reinforcement Social Learning in Cooperative Multiagent Systems",
                        "abstract": "Multiagent coordination in cooperative multiagent systems (MASs) has been widely studied in both fixed-agent repeated interaction setting and the static social learning framework. However, two aspects of dynamics in real-world multiagent scenarios are currently missing in existing works. First, the network topologies can be dynamic where agents may change their connections through rewiring during the course of interactions. Second, the game matrix between each pair of agents may not be static and usually not known as a prior. Both the network dynamic and game uncertainty increase the coordination difficulty among agents. In this paper, we consider a multiagent dynamic social learning environment in which each agent can choose to rewire potential partners and interact with randomly chosen neighbors in each round. We propose an optimal rewiring strategy for agents to select most beneficial peers to interact with for the purpose of maximizing the accumulated payoff in repeated interactions. We empirically demonstrate the effectiveness and robustness of our approach through comparing with benchmark strategies. The performance of three representative learning strategies under our social learning framework with our optimal rewiring is investigated as well."
                    },
                    {
                        "title": "Dynamic Horizon Value Estimation for Model-based Reinforcement Learning",
                        "abstract": "Existing model-based value expansion methods typically leverage a world model for value estimation with a fixed rollout horizon to assist policy learning. However, the fixed rollout with an inaccurate model has a potential to harm the learning process. In this paper, we investigate the idea of using the model knowledge for value expansion adaptively. We propose a novel method called Dynamic-horizon Model-based Value Expansion (DMVE) to adjust the world model usage with different rollout horizons. Inspired by reconstruction-based techniques that can be applied for visual data novelty detection, we utilize a world model with a reconstruction module for image feature extraction, in order to acquire more precise value estimation. The raw and the reconstructed images are both used to determine the appropriate horizon for adaptive value expansion. On several benchmark visual control tasks, experimental results show that DMVE outperforms all baselines in sample efficiency and final performance, indicating that DMVE can achieve more effective and accurate value estimation than state-of-the-art model-based methods."
                    },
                    {
                        "title": "MGHRL: Meta Goal-generation for Hierarchical Reinforcement Learning",
                        "abstract": "Most meta reinforcement learning (meta-RL) methods learn to adapt to new tasks by directly optimizing the parameters of policies over primitive action space. Such algorithms work well in tasks with relatively slight difference. However, when the task distribution becomes wider, it would be quite inefficient to directly learn such a meta-policy. In this paper, we propose a new meta-RL algorithm called Meta Goal-generation for Hierarchical RL (MGHRL). Instead of directly generating policies over primitive action space for new tasks, MGHRL learns to generate high-level meta strategies over subgoals given past experience and leaves the rest of how to achieve subgoals as independent RL subtasks. Our empirical results on several challenging simulated robotics environments show that our method enables more efficient and generalized meta-learning from past experience."
                    },
                    {
                        "title": "Event-Triggered Multi-agent Reinforcement Learning with Communication under Limited-bandwidth Constraint",
                        "abstract": "Communicating with each other in a distributed manner and behaving as a group are essential in multi-agent reinforcement learning. However, real-world multi-agent systems suffer from restrictions on limited-bandwidth communication. If the bandwidth is fully occupied, some agents are not able to send messages promptly to others, causing decision delay and impairing cooperative effects. Recent related work has started to address the problem but still fails in maximally reducing the consumption of communication resources. In this paper, we propose Event-Triggered Communication Network (ETCNet) to enhance the communication efficiency in multi-agent systems by sending messages only when necessary. According to the information theory, the limited bandwidth is translated to the penalty threshold of an event-triggered strategy, which determines whether an agent at each step sends a message or not. Then the design of the event-triggered strategy is formulated as a constrained Markov decision problem, and reinforcement learning finds the best communication protocol that satisfies the limited bandwidth constraint. Experiments on typical multi-agent tasks demonstrate that ETCNet outperforms other methods in terms of the reduction of bandwidth occupancy and still preserves the cooperative performance of multi-agent systems at the most."
                    }
                ]
            },
            "256c3862-d807-4818-ac90-e4a65f07005b": {
                "pk": "256c3862-d807-4818-ac90-e4a65f07005b",
                "name": "Mingsheng Long",
                "collaborators": [
                    "Jianmin Wang",
                    "Junguang Jiang",
                    "Zhangjie Cao",
                    "Yong Liu",
                    "Haixu Wu",
                    "Jiehui Xu",
                    "Yang Shu",
                    "Liangbin Hu",
                    "W. LiMing",
                    "Chao Huang"
                ],
                "domain": [
                    "Time Series Forecasting",
                    "Transfer Learning",
                    "Deep Learning",
                    "Generative Modeling"
                ],
                "publications": [
                    {
                        "title": "Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting",
                        "abstract": "Extending the forecasting time is a critical demand for real applications, such as extreme weather early warning and long-term energy consumption planning. This paper studies the long-term forecasting problem of time series. Prior Transformer-based models adopt various self-attention mechanisms to discover the long-range dependencies. However, intricate temporal patterns of the long-term future prohibit the model from finding reliable dependencies. Also, Transformers have to adopt the sparse versions of point-wise self-attentions for long series efficiency, resulting in the information utilization bottleneck. Going beyond Transformers, we design Autoformer as a novel decomposition architecture with an Auto-Correlation mechanism. We break with the pre-processing convention of series decomposition and renovate it as a basic inner block of deep models. This design empowers Autoformer with progressive decomposition capacities for complex time series. Further, inspired by the stochastic process theory, we design the Auto-Correlation mechanism based on the series periodicity, which conducts the dependencies discovery and representation aggregation at the sub-series level. Auto-Correlation outperforms self-attention in both efficiency and accuracy. In long-term forecasting, Autoformer yields state-of-the-art accuracy, with a 38% relative improvement on six benchmarks, covering five practical applications: energy, traffic, economics, weather and disease. Code is available at this repository: \\url{https://github.com/thuml/Autoformer}."
                    },
                    {
                        "title": "Transferability in Deep Learning: A Survey",
                        "abstract": "The success of deep learning algorithms generally depends on large-scale data, while humans appear to have inherent ability of knowledge transfer, by recognizing and applying relevant knowledge from previous learning experiences when encountering and solving unseen tasks. Such an ability to acquire and reuse knowledge is known as transferability in deep learning. It has formed the long-term quest towards making deep learning as data-efficient as human learning, and has been motivating fruitful design of more powerful deep learning algorithms. We present this survey to connect different isolated areas in deep learning with their relation to transferability, and to provide a unified and complete view to investigating transferability through the whole lifecycle of deep learning. The survey elaborates the fundamental goals and challenges in parallel with the core principles and methods, covering recent cornerstones in deep architectures, pre-training, task adaptation and domain adaptation. This highlights unanswered questions on the appropriate objectives for learning transferable knowledge and for adapting the knowledge to new tasks and domains, avoiding catastrophic forgetting and negative transfer. Finally, we implement a benchmark and an open-source library, enabling a fair evaluation of deep learning methods in terms of transferability."
                    },
                    {
                        "title": "Spin density wave in oxypnictide superconductors in a three-band model",
                        "abstract": "The spin density wave and its temperature dependence in oxypnictide are studied in a three-band model. The spin susceptibilities with various interactions are calculated in the random phase approximation(PPA). It is found that the spin susceptibility peaks around the M point show a spin density wave(SDW) with momentum (0, $\\pi$) and a clear stripe-like spin configuration. The intra-band Coulomb repulsion enhances remarkably the SDW but the Hund's coupling weakens it. It is shown that a new resonance appears at higher temperatures at the $\\Gamma$ point indicating the formation of a paramagnetic phase. There is a clear transition from the SDW phase to the paramagnetic phase."
                    },
                    {
                        "title": "Transfer Adversarial Hashing for Hamming Space Retrieval",
                        "abstract": "Hashing is widely applied to large-scale image retrieval due to the storage and retrieval efficiency. Existing work on deep hashing assumes that the database in the target domain is identically distributed with the training set in the source domain. This paper relaxes this assumption to a transfer retrieval setting, which allows the database and the training set to come from different but relevant domains. However, the transfer retrieval setting will introduce two technical difficulties: first, the hash model trained on the source domain cannot work well on the target domain due to the large distribution gap; second, the domain gap makes it difficult to concentrate the database points to be within a small Hamming ball. As a consequence, transfer retrieval performance within Hamming Radius 2 degrades significantly in existing hashing methods. This paper presents Transfer Adversarial Hashing (TAH), a new hybrid deep architecture that incorporates a pairwise $t$-distribution cross-entropy loss to learn concentrated hash codes and an adversarial network to align the data distributions between the source and target domains. TAH can generate compact transfer hash codes for efficient image retrieval on both source and target domains. Comprehensive experiments validate that TAH yields state of the art Hamming space retrieval performance on standard datasets."
                    },
                    {
                        "title": "Unsupervised Transfer Learning for Spatiotemporal Predictive Networks",
                        "abstract": "This paper explores a new research problem of unsupervised transfer learning across multiple spatiotemporal prediction tasks. Unlike most existing transfer learning methods that focus on fixing the discrepancy between supervised tasks, we study how to transfer knowledge from a zoo of unsupervisedly learned models towards another predictive network. Our motivation is that models from different sources are expected to understand the complex spatiotemporal dynamics from different perspectives, thereby effectively supplementing the new task, even if the task has sufficient training samples. Technically, we propose a differentiable framework named transferable memory. It adaptively distills knowledge from a bank of memory states of multiple pretrained RNNs, and applies it to the target network via a novel recurrent structure called the Transferable Memory Unit (TMU). Compared with finetuning, our approach yields significant improvements on three benchmarks for spatiotemporal prediction, and benefits the target task even from less relevant pretext ones."
                    },
                    {
                        "title": "Transitive Hashing Network for Heterogeneous Multimedia Retrieval",
                        "abstract": "Hashing has been widely applied to large-scale multimedia retrieval due to the storage and retrieval efficiency. Cross-modal hashing enables efficient retrieval from database of one modality in response to a query of another modality. Existing work on cross-modal hashing assumes heterogeneous relationship across modalities for hash function learning. In this paper, we relax the strong assumption by only requiring such heterogeneous relationship in an auxiliary dataset different from the query/database domain. We craft a hybrid deep architecture to simultaneously learn the cross-modal correlation from the auxiliary dataset, and align the dataset distributions between the auxiliary dataset and the query/database domain, which generates transitive hash codes for heterogeneous multimedia retrieval. Extensive experiments exhibit that the proposed approach yields state of the art multimedia retrieval performance on public datasets, i.e. NUS-WIDE, ImageNet-YahooQA."
                    },
                    {
                        "title": "Decoupled Adaptation for Cross-Domain Object Detection",
                        "abstract": "Cross-domain object detection is more challenging than object classification since multiple objects exist in an image and the location of each object is unknown in the unlabeled target domain. As a result, when we adapt features of different objects to enhance the transferability of the detector, the features of the foreground and the background are easy to be confused, which may hurt the discriminability of the detector. Besides, previous methods focused on category adaptation but ignored another important part for object detection, i.e., the adaptation on bounding box regression. To this end, we propose D-adapt, namely Decoupled Adaptation, to decouple the adversarial adaptation and the training of the detector. Besides, we fill the blank of regression domain adaptation in object detection by introducing a bounding box adaptor. Experiments show that D-adapt achieves state-of-the-art results on four cross-domain object detection tasks and yields 17% and 21% relative improvement on benchmark datasets Clipart1k and Comic2k in particular."
                    },
                    {
                        "title": "Koopa: Learning Non-stationary Time Series Dynamics with Koopman Predictors",
                        "abstract": "Real-world time series are characterized by intrinsic non-stationarity that poses a principal challenge for deep forecasting models. While previous models suffer from complicated series variations induced by changing temporal distribution, we tackle non-stationary time series with modern Koopman theory that fundamentally considers the underlying time-variant dynamics. Inspired by Koopman theory of portraying complex dynamical systems, we disentangle time-variant and time-invariant components from intricate non-stationary series by Fourier Filter and design Koopman Predictor to advance respective dynamics forward. Technically, we propose Koopa as a novel Koopman forecaster composed of stackable blocks that learn hierarchical dynamics. Koopa seeks measurement functions for Koopman embedding and utilizes Koopman operators as linear portraits of implicit transition. To cope with time-variant dynamics that exhibits strong locality, Koopa calculates context-aware operators in the temporal neighborhood and is able to utilize incoming ground truth to scale up forecast horizon. Besides, by integrating Koopman Predictors into deep residual structure, we ravel out the binding reconstruction loss in previous Koopman forecasters and achieve end-to-end forecasting objective optimization. Compared with the state-of-the-art model, Koopa achieves competitive performance while saving 77.3% training time and 76.0% memory."
                    },
                    {
                        "title": "Diffusion Tuning: Transferring Diffusion Models via Chain of Forgetting",
                        "abstract": "Diffusion models have significantly advanced the field of generative modeling. However, training a diffusion model is computationally expensive, creating a pressing need to adapt off-the-shelf diffusion models for downstream generation tasks. Current fine-tuning methods focus on parameter-efficient transfer learning but overlook the fundamental transfer characteristics of diffusion models. In this paper, we investigate the transferability of diffusion models and observe a monotonous chain of forgetting trend of transferability along the reverse process. Based on this observation and novel theoretical insights, we present Diff-Tuning, a frustratingly simple transfer approach that leverages the chain of forgetting tendency. Diff-Tuning encourages the fine-tuned model to retain the pre-trained knowledge at the end of the denoising chain close to the generated data while discarding the other noise side. We conduct comprehensive experiments to evaluate Diff-Tuning, including the transfer of pre-trained Diffusion Transformer models to eight downstream generations and the adaptation of Stable Diffusion to five control conditions with ControlNet. Diff-Tuning achieves a 26% improvement over standard fine-tuning and enhances the convergence speed of ControlNet by 24%. Notably, parameter-efficient transfer learning techniques for diffusion models can also benefit from Diff-Tuning."
                    },
                    {
                        "title": "LogME: Practical Assessment of Pre-trained Models for Transfer Learning",
                        "abstract": "This paper studies task adaptive pre-trained model selection, an underexplored problem of assessing pre-trained models for the target task and select best ones from the model zoo \\emph{without fine-tuning}. A few pilot works addressed the problem in transferring supervised pre-trained models to classification tasks, but they cannot handle emerging unsupervised pre-trained models or regression tasks. In pursuit of a practical assessment method, we propose to estimate the maximum value of label evidence given features extracted by pre-trained models. Unlike the maximum likelihood, the maximum evidence is \\emph{immune to over-fitting}, while its expensive computation can be dramatically reduced by our carefully designed algorithm. The Logarithm of Maximum Evidence (LogME) can be used to assess pre-trained models for transfer learning: a pre-trained model with a high LogME value is likely to have good transfer performance. LogME is \\emph{fast, accurate, and general}, characterizing itself as the first practical method for assessing pre-trained models. Compared with brute-force fine-tuning, LogME brings at most $3000\\times$ speedup in wall-clock time and requires only $1\\%$ memory footprint. It outperforms prior methods by a large margin in their setting and is applicable to new settings. It is general enough for diverse pre-trained models (supervised pre-trained and unsupervised pre-trained), downstream tasks (classification and regression), and modalities (vision and language). Code is available at this repository: \\href{https://github.com/thuml/LogME}{https://github.com/thuml/LogME}."
                    }
                ]
            }
        }
    },
    "2309.06599": {
        "paper_data": {
            "title": "Reasoning with Latent Diffusion in Offline Reinforcement Learning",
            "url": "http://arxiv.org/abs/2309.06599v1",
            "arxiv_id": "2309.06599",
            "authors": [
                "Siddarth Venkatraman",
                "Shivesh Khaitan",
                "Ravi Tej Akella",
                "John Dolan",
                "Jeff Schneider",
                "Glen Berseth"
            ],
            "abstract": "Offline reinforcement learning (RL) holds promise as a means to learn high-reward policies from a static dataset, without the need for further environment interactions. However, a key challenge in offline RL lies in effectively stitching portions of suboptimal trajectories from the static dataset while avoiding extrapolation errors arising due to a lack of support in the dataset. Existing approaches use conservative methods that are tricky to tune and struggle with multi-modal data (as we show) or rely on noisy Monte Carlo return-to-go samples for reward conditioning. In this work, we propose a novel approach that leverages the expressiveness of latent diffusion to model in-support trajectory sequences as compressed latent skills. This facilitates learning a Q-function while avoiding extrapolation error via batch-constraining. The latent space is also expressive and gracefully copes with multi-modal data. We show that the learned temporally-abstract latent space encodes richer task-specific information for offline RL tasks as compared to raw state-actions. This improves credit assignment and facilitates faster reward propagation during Q-learning. Our method demonstrates state-of-the-art performance on the D4RL benchmarks, particularly excelling in long-horizon, sparse-reward tasks.",
            "introduction": " Introduction Offline reinforcement learning (RL) offers a promising approach to learning policies from static datasets. These datasets are often comprised of undirected demonstrations and suboptimal sequences collected using different behavior policies . Several results in Figure 8. Figure 8: D4RL score for LDCQ with varying diffusion steps Tin locomotion tasks. For the long horizon tasks, we found that increasing diffusion steps resulted in an initial trend upward in performance. Beyond this, the performance does not improve with additional diffusion steps (Figure 9). We also used additional diffusion steps at time t= 0similar to Diffusion-X (Pearce et al. 2023). This means that after the DDPM sampling of diffusion from time Tto 1, we run Xadditional diffusion steps to further denoise the sample, assuming time-step t= 1. Pearce et al. 2023 reasoned that this pushes the samples further towards higher-likelihood regions. We used 10 additional steps across Related Work Offline RL. As discussed previously, offline RL poses the challenge of distributional shift while stitching suboptimal trajectories together. Conservative Q-Learning (CQL) (Kumar et al. 2020) tries to constrain the policy to the behavioral support by learning a pessimistic Q-function that lower- bounds the optimal value function. Implicit Q-Learning (IQL) (Vieillard et al. 2022) tries to avoid extrapolation error by performing a trade-off between SARSA and DQN using expectile regression. However, it achieves the optimal batch-constrained policy only as their expectile parameter \u03c4\u21921, which leads to an increasingly difficult-to-optimize objective. Our method instead learns the optimal batch-constrained Q-function without introducing any pessimism or trade-off. Inspired by notable achievements of generative models in various domains including text-generation (Vaswani et al. 2017), speech synthesis (Kong et al.) and image-generation (Ramesh et al. 2022; Saharia et al. 2022a), Chen et al. 2021 proposed to use a generative model for offline RL and bypass the need for Q-learning or bootstrapping altogether with return-conditioning (Srivastava et al. 2019; Kumar et al. 2019). While these ideas have found success, getting a good return estimate for arbitrary states is not trivial and conditioning on returns outside the support of the training dataset can lead to the generative model producing low-value out-of-distribution sequences. Our method instead avoids return-conditioning and formulates a solution with batch-constraining which uses generative models to model the data distribution and use it to generate candidate actions to learn a Q-function without extrapolation-error (Fujimoto et al. 2019). This formulation relies on the assumption that sampling from the generative model does not sample out-of-support samples, which has been difficult to achieve with previously used generative models in offline RL. Our method circumvents this problem with the latent diffusion model. Further, to effectively address the problem of stitching, Pertsch et al. 2021 and Ajay et al. b proposed learning policies in latent-trajectory spaces. However, they have to rely on a highly constrained latent space which is not rich enough for the downstream policy. This is due to the limitations of the generative model used like V AEs. Our proposed method to use latent diffusion, which can model complex distributions, allows for the needed flexibility in the latent space for effective Q-learning and the final policy. Diffusion Probabilistic Models. Recently, diffusion models (Sohl-Dickstein et al. 2015; Song and Ermon 2019) have emerged as state-of-the-art generative models for conditional image-generation (Ramesh et al. 2022; Saharia et al. 2022a), super-resolution (Saharia et al. 2022b) and",
            "references": [
                {
                    "title": "Shap-E: Generating Conditional 3D Implicit Functions",
                    "abstract": "We present Shap-E, a conditional generative model for 3D assets. Unlike recent work on 3D generative models which produce a single output representation, Shap-E directly generates the parameters of implicit functions that can be rendered as both textured meshes and neural radiance fields. We train Shap-E in two stages: first, we train an encoder that deterministically maps 3D assets into the parameters of an implicit function; second, we train a conditional diffusion model on outputs of the encoder. When trained on a large dataset of paired 3D and text data, our resulting models are capable of generating complex and diverse 3D assets in a matter of seconds. When compared to Point-E, an explicit generative model over point clouds, Shap-E converges faster and reaches comparable or better sample quality despite modeling a higher-dimensional, multi-representation output space. We release model weights, inference code, and samples at https://github.com/openai/shap-e."
                },
                {
                    "title": "Efficient Diffusion Training via Min-SNR Weighting Strategy",
                    "abstract": "Denoising diffusion models have been a mainstream approach for image generation, however, training these models often suffers from slow convergence. In this paper, we discovered that the slow convergence is partly due to conflicting optimization directions between timesteps. To address this issue, we treat the diffusion training as a multi-task learning problem, and introduce a simple yet effective approach referred to as Min-SNR-\u03b3. This method adapts loss weights of timesteps based on clamped signal-to-noise ratios, which effectively balances the conflicts among timesteps. Our results demonstrate a significant improvement in converging speed, 3.4\u00d7 faster than previous weighting strategies. It is also more effective, achieving a new record FID score of 2.06 on the ImageNet 256 \u00d7 256 benchmark using smaller architectures than that employed in previous state-of-the-art. The code is available at https://github.com/TiankaiHang/Min-SNR-Diffusion-Training."
                },
                {
                    "title": "Consistency Models",
                    "abstract": "Diffusion models have significantly advanced the fields of image, audio, and video generation, but they depend on an iterative sampling process that causes slow generation. To overcome this limitation, we propose consistency models, a new family of models that generate high quality samples by directly mapping noise to data. They support fast one-step generation by design, while still allowing multistep sampling to trade compute for sample quality. They also support zero-shot data editing, such as image inpainting, colorization, and super-resolution, without requiring explicit training on these tasks. Consistency models can be trained either by distilling pre-trained diffusion models, or as standalone generative models altogether. Through extensive experiments, we demonstrate that they outperform existing distillation techniques for diffusion models in one- and few-step sampling, achieving the new state-of-the-art FID of 3.55 on CIFAR-10 and 6.20 on ImageNet 64x64 for one-step generation. When trained in isolation, consistency models become a new family of generative models that can outperform existing one-step, non-adversarial generative models on standard benchmarks such as CIFAR-10, ImageNet 64x64 and LSUN 256x256."
                },
                {
                    "title": "Imitating Human Behaviour with Diffusion Models",
                    "abstract": "Diffusion models have emerged as powerful generative models in the text-to-image domain. This paper studies their application as observation-to-action models for imitating human behaviour in sequential environments. Human behaviour is stochastic and multimodal, with structured correlations between action dimensions. Meanwhile, standard modelling choices in behaviour cloning are limited in their expressiveness and may introduce bias into the cloned policy. We begin by pointing out the limitations of these choices. We then propose that diffusion models are an excellent fit for imitating human behaviour, since they learn an expressive distribution over the joint action space. We introduce several innovations to make diffusion models suitable for sequential environments; designing suitable architectures, investigating the role of guidance, and developing reliable sampling strategies. Experimentally, diffusion models closely match human demonstrations in a simulated robotic control task and a modern 3D gaming environment."
                },
                {
                    "title": "Flow to Control: Offline Reinforcement Learning with Lossless Primitive Discovery",
                    "abstract": "Offline reinforcement learning (RL) enables the agent to effectively learn from logged data, which significantly extends the applicability of RL algorithms in real-world scenarios where exploration can be expensive or unsafe. Previous works have shown that extracting primitive skills from the recurring and temporally extended structures in the logged data yields better learning. However, these methods suffer greatly when the primitives have limited representation ability to recover the original policy space, especially in offline settings. In this paper, we give a quantitative characterization of the performance of offline hierarchical learning and highlight the importance of learning lossless primitives. To this end, we propose to use a flow-based structure as the representation for low-level policies. This allows us to represent the behaviors in the dataset faithfully while keeping the expression ability to recover the whole policy space. We show that such lossless primitives can drastically improve the performance of hierarchical policies. The experimental results and extensive ablation studies on the standard D4RL benchmark show that our method has a good representation ability for policies and achieves superior performance in most tasks."
                },
                {
                    "title": "Is Conditional Generative Modeling all you need for Decision-Making?",
                    "abstract": "Recent improvements in conditional generative modeling have made it possible to generate high-quality images from language descriptions alone. We investigate whether these methods can directly address the problem of sequential decision-making. We view decision-making not through the lens of reinforcement learning (RL), but rather through conditional generative modeling. To our surprise, we find that our formulation leads to policies that can outperform existing offline RL approaches across standard benchmarks. By modeling a policy as a return-conditional diffusion model, we illustrate how we may circumvent the need for dynamic programming and subsequently eliminate many of the complexities that come with traditional offline RL. We further demonstrate the advantages of modeling policies as conditional diffusion models by considering two other conditioning variables: constraints and skills. Conditioning on a single constraint or skill during training leads to behaviors at test-time that can satisfy several constraints together or demonstrate a composition of skills. Our results illustrate that conditional generative modeling is a powerful tool for decision-making."
                },
                {
                    "title": "Classifier-Free Diffusion Guidance",
                    "abstract": "Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance."
                },
                {
                    "title": "DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps",
                    "abstract": "Diffusion probabilistic models (DPMs) are emerging powerful generative models. Despite their high-quality generation performance, DPMs still suffer from their slow sampling as they generally need hundreds or thousands of sequential function evaluations (steps) of large neural networks to draw a sample. Sampling from DPMs can be viewed alternatively as solving the corresponding diffusion ordinary differential equations (ODEs). In this work, we propose an exact formulation of the solution of diffusion ODEs. The formulation analytically computes the linear part of the solution, rather than leaving all terms to black-box ODE solvers as adopted in previous works. By applying change-of-variable, the solution can be equivalently simplified to an exponentially weighted integral of the neural network. Based on our formulation, we propose DPM-Solver, a fast dedicated high-order solver for diffusion ODEs with the convergence order guarantee. DPM-Solver is suitable for both discrete-time and continuous-time DPMs without any further training. Experimental results show that DPM-Solver can generate high-quality samples in only 10 to 20 function evaluations on various datasets. We achieve 4.70 FID in 10 function evaluations and 2.87 FID in 20 function evaluations on the CIFAR10 dataset, and a $4\\sim 16\\times$ speedup compared with previous state-of-the-art training-free samplers on various datasets."
                },
                {
                    "title": "Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding",
                    "abstract": "We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding. Imagen builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation. Our key discovery is that generic large language models (e.g. T5), pretrained on text-only corpora, are surprisingly effective at encoding text for image synthesis: increasing the size of the language model in Imagen boosts both sample fidelity and image-text alignment much more than increasing the size of the image diffusion model. Imagen achieves a new state-of-the-art FID score of 7.27 on the COCO dataset, without ever training on COCO, and human raters find Imagen samples to be on par with the COCO data itself in image-text alignment. To assess text-to-image models in greater depth, we introduce DrawBench, a comprehensive and challenging benchmark for text-to-image models. With DrawBench, we compare Imagen with recent methods including VQ-GAN+CLIP, Latent Diffusion Models, and DALL-E 2, and find that human raters prefer Imagen over other models in side-by-side comparisons, both in terms of sample quality and image-text alignment. See https://imagen.research.google/ for an overview of the results."
                },
                {
                    "title": "Planning with Diffusion for Flexible Behavior Synthesis",
                    "abstract": "Model-based reinforcement learning methods often use learning only for the purpose of estimating an approximate dynamics model, offloading the rest of the decision-making work to classical trajectory optimizers. While conceptually simple, this combination has a number of empirical shortcomings, suggesting that learned models may not be well-suited to standard trajectory optimization. In this paper, we consider what it would look like to fold as much of the trajectory optimization pipeline as possible into the modeling problem, such that sampling from the model and planning with it become nearly identical. The core of our technical approach lies in a diffusion probabilistic model that plans by iteratively denoising trajectories. We show how classifier-guided sampling and image inpainting can be reinterpreted as coherent planning strategies, explore the unusual and useful properties of diffusion-based planning methods, and demonstrate the effectiveness of our framework in control settings that emphasize long-horizon decision-making and test-time flexibility."
                },
                {
                    "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
                    "abstract": "Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples."
                },
                {
                    "title": "RePaint: Inpainting using Denoising Diffusion Probabilistic Models",
                    "abstract": "Free-form inpainting is the task of adding new content to an image in the regions specified by an arbitrary binary mask. Most existing approaches train for a certain distribution of masks, which limits their generalization capabilities to unseen mask types. Furthermore, training with pixel-wise and perceptual losses often leads to simple textural extensions towards the missing areas instead of semantically meaningful generation. In this work, we propose RePaint: A Denoising Diffusion Probabilistic Model (DDPM) based inpainting approach that is applicable to even extreme masks. We employ a pretrained unconditional DDPM as the generative prior. To condition the generation process, we only alter the reverse diffusion iterations by sampling the unmasked regions using the given image infor-mation. Since this technique does not modify or condition the original DDPM network itself, the model produces high-quality and diverse output images for any inpainting form. We validate our method for both faces and general-purpose image inpainting using standard and extreme masks. Re-Paint outperforms state-of-the-art Autoregressive, and GAN approaches for at least five out of six mask distributions. Github Repository: git.io/RePaint"
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Tackling the Generative Learning Trilemma with Denoising Diffusion GANs",
                    "abstract": "A wide variety of deep generative models has been developed in the past decade. Yet, these models often struggle with simultaneously addressing three key requirements including: high sample quality, mode coverage, and fast sampling. We call the challenge imposed by these requirements the generative learning trilemma, as the existing models often trade some of them for others. Particularly, denoising diffusion models have shown impressive sample quality and diversity, but their expensive sampling does not yet allow them to be applied in many real-world applications. In this paper, we argue that slow sampling in these models is fundamentally attributed to the Gaussian assumption in the denoising step which is justified only for small step sizes. To enable denoising with large steps, and hence, to reduce the total number of denoising steps, we propose to model the denoising distribution using a complex multimodal distribution. We introduce denoising diffusion generative adversarial networks (denoising diffusion GANs) that model each denoising step using a multimodal conditional GAN. Through extensive evaluations, we show that denoising diffusion GANs obtain sample quality and diversity competitive with original diffusion models while being 2000$\\times$ faster on the CIFAR-10 dataset. Compared to traditional GANs, our model exhibits better mode coverage and sample diversity. To the best of our knowledge, denoising diffusion GAN is the first model that reduces sampling cost in diffusion models to an extent that allows them to be applied to real-world applications inexpensively. Project page and code can be found at https://nvlabs.github.io/denoising-diffusion-gan"
                },
                {
                    "title": "Score-Based Generative Modeling with Critically-Damped Langevin Diffusion",
                    "abstract": "Score-based generative models (SGMs) have demonstrated remarkable synthesis quality. SGMs rely on a diffusion process that gradually perturbs the data towards a tractable distribution, while the generative model learns to denoise. The complexity of this denoising task is, apart from the data distribution itself, uniquely determined by the diffusion process. We argue that current SGMs employ overly simplistic diffusions, leading to unnecessarily complex denoising processes, which limit generative modeling performance. Based on connections to statistical mechanics, we propose a novel critically-damped Langevin diffusion (CLD) and show that CLD-based SGMs achieve superior performance. CLD can be interpreted as running a joint diffusion in an extended space, where the auxiliary variables can be considered \u201cvelocities\u201d that are coupled to the data variables as in Hamiltonian dynamics. We derive a novel score matching objective for CLD and show that the model only needs to learn the score function of the conditional distribution of the velocity given data, an easier task than learning scores of the data directly. We also derive a new sampling scheme for ef\ufb01cient synthesis from CLD-based diffusion models. We \ufb01nd that CLD outperforms previous SGMs in synthesis quality for similar network architectures and sampling compute budgets. We show that our novel sampler for CLD signi\ufb01cantly outperforms solvers such as Euler\u2013Maruyama. Our framework provides new insights into score-based denoising diffusion models and can be readily used for high-resolution image synthesis. VPSDE. We leave the study of CLD with maximum likelihood training for high-dimensional (image) datasets to future work."
                },
                {
                    "title": "Temporal Abstraction in Reinforcement Learning with the Successor Representation",
                    "abstract": "Reasoning at multiple levels of temporal abstraction is one of the key attributes of intelligence. In reinforcement learning, this is often modeled through temporally extended courses of actions called options. Options allow agents to make predictions and to operate at different levels of abstraction within an environment. Nevertheless, approaches based on the options framework often start with the assumption that a reasonable set of options is known beforehand. When this is not the case, there are no definitive answers for which options one should consider. In this paper, we argue that the successor representation (SR), which encodes states based on the pattern of state visitation that follows them, can be seen as a natural substrate for the discovery and use of temporal abstractions. To support our claim, we take a big picture view of recent results, showing how the SR can be used to discover options that facilitate either temporally-extended exploration or planning. We cast these results as instantiations of a general framework for option discovery in which the agent's representation is used to identify useful options, which are then used to further improve its representation. This results in a virtuous, never-ending, cycle in which both the representation and the options are constantly refined based on each other. Beyond option discovery itself, we also discuss how the SR allows us to augment a set of options into a combinatorially large counterpart without additional learning. This is achieved through the combination of previously learned options. Our empirical evaluation focuses on options discovered for exploration and on the use of the SR to combine them. The results of our experiments shed light on important design decisions involved in the definition of options and demonstrate the synergy of different methods based on the SR, such as eigenoptions and the option keyboard."
                },
                {
                    "title": "Offline Reinforcement Learning with Implicit Q-Learning",
                    "abstract": "Offline reinforcement learning requires reconciling two conflicting aims: learning a policy that improves over the behavior policy that collected the dataset, while at the same time minimizing the deviation from the behavior policy so as to avoid errors due to distributional shift. This trade-off is critical, because most current offline reinforcement learning methods need to query the value of unseen actions during training to improve the policy, and therefore need to either constrain these actions to be in-distribution, or else regularize their values. We propose an offline RL method that never needs to evaluate actions outside of the dataset, but still enables the learned policy to improve substantially over the best behavior in the data through generalization. The main insight in our work is that, instead of evaluating unseen actions from the latest policy, we can approximate the policy improvement step implicitly by treating the state value function as a random variable, with randomness determined by the action (while still integrating over the dynamics to avoid excessive optimism), and then taking a state conditional upper expectile of this random variable to estimate the value of the best actions in that state. This leverages the generalization capacity of the function approximator to estimate the value of the best available action at a given state without ever directly querying a Q-function with this unseen action. Our algorithm alternates between fitting this upper expectile value function and backing it up into a Q-function. Then, we extract the policy via advantage-weighted behavioral cloning. We dub our method implicit Q-learning (IQL). IQL demonstrates the state-of-the-art performance on D4RL, a standard benchmark for offline reinforcement learning. We also demonstrate that IQL achieves strong performance fine-tuning using online interaction after offline initialization."
                },
                {
                    "title": "Implicitly Regularized RL with Implicit Q-Values",
                    "abstract": "The $Q$-function is a central quantity in many Reinforcement Learning (RL) algorithms for which RL agents behave following a (soft)-greedy policy w.r.t. to $Q$. It is a powerful tool that allows action selection without a model of the environment and even without explicitly modeling the policy. Yet, this scheme can only be used in discrete action tasks, with small numbers of actions, as the softmax cannot be computed exactly otherwise. Especially the usage of function approximation, to deal with continuous action spaces in modern actor-critic architectures, intrinsically prevents the exact computation of a softmax. We propose to alleviate this issue by parametrizing the $Q$-function implicitly, as the sum of a log-policy and of a value function. We use the resulting parametrization to derive a practical off-policy deep RL algorithm, suitable for large action spaces, and that enforces the softmax relation between the policy and the $Q$-value. We provide a theoretical analysis of our algorithm: from an Approximate Dynamic Programming perspective, we show its equivalence to a regularized version of value iteration, accounting for both entropy and Kullback-Leibler regularization, and that enjoys beneficial error propagation results. We then evaluate our algorithm on classic control tasks, where its results compete with state-of-the-art methods."
                },
                {
                    "title": "Offline Reinforcement Learning as One Big Sequence Modeling Problem",
                    "abstract": "Reinforcement learning (RL) is typically concerned with estimating stationary policies or single-step models, leveraging the Markov property to factorize problems in time. However, we can also view RL as a generic sequence modeling problem, with the goal being to produce a sequence of actions that leads to a sequence of high rewards. Viewed in this way, it is tempting to consider whether high-capacity sequence prediction models that work well in other domains, such as natural-language processing, can also provide effective solutions to the RL problem. To this end, we explore how RL can be tackled with the tools of sequence modeling, using a Transformer architecture to model distributions over trajectories and repurposing beam search as a planning algorithm. Framing RL as sequence modeling problem simplifies a range of design decisions, allowing us to dispense with many of the components common in offline RL algorithms. We demonstrate the flexibility of this approach across long-horizon dynamics prediction, imitation learning, goal-conditioned RL, and offline RL. Further, we show that this approach can be combined with existing model-free algorithms to yield a state-of-the-art planner in sparse-reward, long-horizon tasks."
                },
                {
                    "title": "Decision Transformer: Reinforcement Learning via Sequence Modeling",
                    "abstract": "We introduce a framework that abstracts Reinforcement Learning (RL) as a sequence modeling problem. This allows us to draw upon the simplicity and scalability of the Transformer architecture, and associated advances in language modeling such as GPT-x and BERT. In particular, we present Decision Transformer, an architecture that casts the problem of RL as conditional sequence modeling. Unlike prior approaches to RL that fit value functions or compute policy gradients, Decision Transformer simply outputs the optimal actions by leveraging a causally masked Transformer. By conditioning an autoregressive model on the desired return (reward), past states, and actions, our Decision Transformer model can generate future actions that achieve the desired return. Despite its simplicity, Decision Transformer matches or exceeds the performance of state-of-the-art model-free offline RL baselines on Atari, OpenAI Gym, and Key-to-Door tasks."
                },
                {
                    "title": "Image Super-Resolution via Iterative Refinement",
                    "abstract": "We present SR3, an approach to image Super-Resolution via Repeated Refinement. SR3 adapts denoising diffusion probabilistic models (Ho et al. 2020), (Sohl-Dickstein et al. 2015) to image-to-image translation, and performs super-resolution through a stochastic iterative denoising process. Output images are initialized with pure Gaussian noise and iteratively refined using a U-Net architecture that is trained on denoising at various noise levels, conditioned on a low-resolution input image. SR3 exhibits strong performance on super-resolution tasks at different magnification factors, on faces and natural images. We conduct human evaluation on a standard 8\u00d7 face super-resolution task on CelebA-HQ for which SR3 achieves a fool rate close to 50%, suggesting photo-realistic outputs, while GAN baselines do not exceed a fool rate of 34%. We evaluate SR3 on a 4\u00d7 super-resolution task on ImageNet, where SR3 outperforms baselines in human evaluation and classification accuracy of a ResNet-50 classifier trained on high-resolution images. We further show the effectiveness of SR3 in cascaded image generation, where a generative model is chained with super-resolution models to synthesize high-resolution images with competitive FID scores on the class-conditional 256\u00d7256 ImageNet generation challenge."
                },
                {
                    "title": "PLAS: Latent Action Space for Offline Reinforcement Learning",
                    "abstract": "The goal of offline reinforcement learning is to learn a policy from a fixed dataset, without further interactions with the environment. This setting will be an increasingly more important paradigm for real-world applications of reinforcement learning such as robotics, in which data collection is slow and potentially dangerous. Existing off-policy algorithms have limited performance on static datasets due to extrapolation errors from out-of-distribution actions. This leads to the challenge of constraining the policy to select actions within the support of the dataset during training. We propose to simply learn the Policy in the Latent Action Space (PLAS) such that this requirement is naturally satisfied. We evaluate our method on continuous control benchmarks in simulation and a deformable object manipulation task with a physical robot. We demonstrate that our method provides competitive performance consistently across various continuous control tasks and different types of datasets, outperforming existing offline reinforcement learning methods with explicit constraints. Videos and code are available at this https URL."
                },
                {
                    "title": "OPAL: Offline Primitive Discovery for Accelerating Offline Reinforcement Learning",
                    "abstract": "Reinforcement learning (RL) has achieved impressive performance in a variety of online settings in which an agent's ability to query the environment for transitions and rewards is effectively unlimited. However, in many practical applications, the situation is reversed: an agent may have access to large amounts of undirected offline experience data, while access to the online environment is severely limited. In this work, we focus on this offline setting. Our main insight is that, when presented with offline data composed of a variety of behaviors, an effective way to leverage this data is to extract a continuous space of recurring and temporally extended primitive behaviors before using these primitives for downstream task learning. Primitives extracted in this way serve two purposes: they delineate the behaviors that are supported by the data from those that are not, making them useful for avoiding distributional shift in offline RL; and they provide a degree of temporal abstraction, which reduces the effective horizon yielding better learning in theory, and improved offline RL in practice. In addition to benefiting offline policy optimization, we show that performing offline primitive learning in this way can also be leveraged for improving few-shot imitation learning as well as exploration and transfer in online RL on a variety of benchmark domains. Visualizations are available at this https URL"
                },
                {
                    "title": "Accelerating Reinforcement Learning with Learned Skill Priors",
                    "abstract": "Intelligent agents rely heavily on prior experience when learning a new task, yet most modern reinforcement learning (RL) approaches learn every task from scratch. One approach for leveraging prior knowledge is to transfer skills learned on prior tasks to the new task. However, as the amount of prior experience increases, the number of transferable skills grows too, making it challenging to explore the full set of available skills during downstream learning. Yet, intuitively, not all skills should be explored with equal probability; for example information about the current state can hint which skills are promising to explore. In this work, we propose to implement this intuition by learning a prior over skills. We propose a deep latent variable model that jointly learns an embedding space of skills and the skill prior from offline agent experience. We then extend common maximum-entropy RL approaches to use skill priors to guide downstream learning. We validate our approach, SPiRL (Skill-Prior RL), on complex navigation and robotic manipulation tasks and show that learned skill priors are essential for effective skill transfer from rich datasets. Videos and code are available at this https URL."
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "DiffWave: A Versatile Diffusion Model for Audio Synthesis",
                    "abstract": "In this work, we propose DiffWave, a versatile Diffusion probabilistic model for conditional and unconditional Waveform generation. The model is non-autoregressive, and converts the white noise signal into structured waveform through a Markov chain with a constant number of steps at synthesis. It is efficiently trained by optimizing a variant of variational bound on the data likelihood. DiffWave produces high-fidelity audios in Different Waveform generation tasks, including neural vocoding conditioned on mel spectrogram, class-conditional generation, and unconditional generation. We demonstrate that DiffWave matches a strong WaveNet vocoder in terms of speech quality~(MOS: 4.44 versus 4.43), while synthesizing orders of magnitude faster. In particular, it significantly outperforms autoregressive and GAN-based waveform models in the challenging unconditional generation task in terms of audio quality and sample diversity from various automatic and human evaluations."
                },
                {
                    "title": "EMaQ: Expected-Max Q-Learning Operator for Simple Yet Effective Offline and Online RL",
                    "abstract": "Off-policy reinforcement learning (RL) holds the promise of sample-efficient learning of decision-making policies by leveraging past experience. However, in the offline RL setting -- where a fixed collection of interactions are provided and no further interactions are allowed -- it has been shown that standard off-policy RL methods can significantly underperform. Recently proposed methods aim to address this shortcoming by regularizing learned policies to remain close to the given dataset of interactions. However, these methods involve several configurable components such as learning a separate policy network on top of a behavior cloning actor, and explicitly constraining action spaces through clipping or reward penalties. Striving for simultaneous simplicity and performance, in this work we present a novel backup operator, Expected-Max Q-Learning (EMaQ), which naturally restricts learned policies to remain within the support of the offline dataset \\emph{without any explicit regularization}, while retaining desirable theoretical properties such as contraction. We demonstrate that EMaQ is competitive with Soft Actor Critic (SAC) in online RL, and surpasses SAC in the deployment-efficient setting. In the offline RL setting -- the main focus of this work -- through EMaQ we are able to make important observations regarding key components of offline RL, and the nature of standard benchmark tasks. Lastly but importantly, we observe that EMaQ achieves state-of-the-art performance with fewer moving parts such as one less function approximation, making it a strong, yet easy to implement baseline for future work."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Conservative Q-Learning for Offline Reinforcement Learning",
                    "abstract": "Effectively leveraging large, previously collected datasets in reinforcement learning (RL) is a key challenge for large-scale real-world applications. Offline RL algorithms promise to learn effective policies from previously-collected, static datasets without further interaction. However, in practice, offline RL presents a major challenge, and standard off-policy RL methods can fail due to overestimation of values induced by the distributional shift between the dataset and the learned policy, especially when training on complex and multi-modal data distributions. In this paper, we propose conservative Q-learning (CQL), which aims to address these limitations by learning a conservative Q-function such that the expected value of a policy under this Q-function lower-bounds its true value. We theoretically show that CQL produces a lower bound on the value of the current policy and that it can be incorporated into a policy learning procedure with theoretical improvement guarantees. In practice, CQL augments the standard Bellman error objective with a simple Q-value regularizer which is straightforward to implement on top of existing deep Q-learning and actor-critic implementations. On both discrete and continuous control domains, we show that CQL substantially outperforms existing offline RL methods, often learning policies that attain 2-5 times higher final return, especially when learning from complex and multi-modal data distributions."
                },
                {
                    "title": "Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems",
                    "abstract": "In this tutorial article, we aim to provide the reader with the conceptual tools needed to get started on research on offline reinforcement learning algorithms: reinforcement learning algorithms that utilize previously collected data, without additional online data collection. Offline reinforcement learning algorithms hold tremendous promise for making it possible to turn large datasets into powerful decision making engines. Effective offline reinforcement learning methods would be able to extract policies with the maximum possible utility out of the available data, thereby allowing automation of a wide range of decision-making domains, from healthcare and education to robotics. However, the limitations of current algorithms make this difficult. We will aim to provide the reader with an understanding of these challenges, particularly in the context of modern deep reinforcement learning methods, and describe some potential solutions that have been explored in recent work to mitigate these challenges, along with recent applications, and a discussion of perspectives on open problems in the field."
                },
                {
                    "title": "D4RL: Datasets for Deep Data-Driven Reinforcement Learning",
                    "abstract": "The offline reinforcement learning (RL) problem, also known as batch RL, refers to the setting where a policy must be learned from a static dataset, without additional online data collection. This setting is compelling as potentially it allows RL methods to take advantage of large, pre-collected datasets, much like how the rise of large datasets has fueled results in supervised learning in recent years. However, existing online RL benchmarks are not tailored towards the offline setting, making progress in offline RL difficult to measure. In this work, we introduce benchmarks specifically designed for the offline setting, guided by key properties of datasets relevant to real-world applications of offline RL. Examples of such properties include: datasets generated via hand-designed controllers and human demonstrators, multi-objective datasets where an agent can perform different tasks in the same environment, and datasets consisting of a mixtures of policies. To facilitate research, we release our benchmark tasks and datasets with a comprehensive evaluation of existing algorithms and an evaluation protocol together with an open-source codebase. We hope that our benchmark will focus research effort on methods that drive improvements not just on simulated tasks, but ultimately on the kinds of real-world problems where offline RL will have the largest impact."
                },
                {
                    "title": "Reward-Conditioned Policies",
                    "abstract": "Reinforcement learning offers the promise of automating the acquisition of complex behavioral skills. However, compared to commonly used and well-understood supervised learning methods, reinforcement learning algorithms can be brittle, difficult to use and tune, and sensitive to seemingly innocuous implementation decisions. In contrast, imitation learning utilizes standard and well-understood supervised learning methods, but requires near-optimal expert data. Can we learn effective policies via supervised learning without demonstrations? The main idea that we explore in this work is that non-expert trajectories collected from sub-optimal policies can be viewed as optimal supervision, not for maximizing the reward, but for matching the reward of the given trajectory. By then conditioning the policy on the numerical value of the reward, we can obtain a policy that generalizes to larger returns. We show how such an approach can be derived as a principled method for policy search, discuss several variants, and compare the method experimentally to a variety of current reinforcement learning methods on standard benchmarks."
                },
                {
                    "title": "Training Agents using Upside-Down Reinforcement Learning",
                    "abstract": "Traditional Reinforcement Learning (RL) algorithms either predict rewards with value functions or maximize them using policy search. We study an alternative: Upside-Down Reinforcement Learning (Upside-Down RL or UDRL), that solves RL problems primarily using supervised learning techniques. Many of its main principles are outlined in a companion report [34]. Here we present the first concrete implementation of UDRL and demonstrate its feasibility on certain episodic learning problems. Experimental results show that its performance can be surprisingly competitive with, and even exceed that of traditional baseline algorithms developed over decades of research."
                },
                {
                    "title": "Generative Modeling by Estimating Gradients of the Data Distribution",
                    "abstract": "We introduce a new generative model where samples are produced via Langevin dynamics using gradients of the data distribution estimated with score matching. Because gradients can be ill-defined and hard to estimate when the data resides on low-dimensional manifolds, we perturb the data with different levels of Gaussian noise, and jointly estimate the corresponding scores, i.e., the vector fields of gradients of the perturbed data distribution for all noise levels. For sampling, we propose an annealed Langevin dynamics where we use gradients corresponding to gradually decreasing noise levels as the sampling process gets closer to the data manifold. Our framework allows flexible model architectures, requires no sampling during training or the use of adversarial methods, and provides a learning objective that can be used for principled model comparisons. Our models produce samples comparable to GANs on MNIST, CelebA and CIFAR-10 datasets, achieving a new state-of-the-art inception score of 8.87 on CIFAR-10. Additionally, we demonstrate that our models learn effective representations via image inpainting experiments."
                },
                {
                    "title": "Off-Policy Deep Reinforcement Learning without Exploration",
                    "abstract": "Many practical applications of reinforcement learning constrain agents to learn from a fixed batch of data which has already been gathered, without offering further possibility for data collection. In this paper, we demonstrate that due to errors introduced by extrapolation, standard off-policy deep reinforcement learning algorithms, such as DQN and DDPG, are incapable of learning with data uncorrelated to the distribution under the current policy, making them ineffective for this fixed batch setting. We introduce a novel class of off-policy algorithms, batch-constrained reinforcement learning, which restricts the action space in order to force the agent towards behaving close to on-policy with respect to a subset of the given data. We present the first continuous control deep reinforcement learning algorithm which can learn effectively from arbitrary, fixed batch data, and empirically demonstrate the quality of its behavior in several tasks."
                },
                {
                    "title": "Addressing Function Approximation Error in Actor-Critic Methods",
                    "abstract": "In value-based reinforcement learning methods such as deep Q-learning, function approximation errors are known to lead to overestimated value estimates and suboptimal policies. We show that this problem persists in an actor-critic setting and propose novel mechanisms to minimize its effects on both the actor and the critic. Our algorithm builds on Double Q-learning, by taking the minimum value between a pair of critics to limit overestimation. We draw the connection between target networks and overestimation bias, and suggest delaying policy updates to reduce per-update error and further improve performance. We evaluate our method on the suite of OpenAI gym tasks, outperforming the state of the art in every environment tested."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework",
                    "abstract": "an"
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Prioritized Experience Replay",
                    "abstract": "Experience replay lets online reinforcement learning agents remember and reuse experiences from the past. In prior work, experience transitions were uniformly sampled from a replay memory. However, this approach simply replays transitions at the same frequency that they were originally experienced, regardless of their significance. In this paper we develop a framework for prioritizing experience, so as to replay important transitions more frequently, and therefore learn more efficiently. We use prioritized experience replay in Deep Q-Networks (DQN), a reinforcement learning algorithm that achieved human-level performance across many Atari games. DQN with prioritized experience replay achieves a new state-of-the-art, outperforming DQN with uniform replay on 41 out of 49 games."
                },
                {
                    "title": "Deep Unsupervised Learning using Nonequilibrium Thermodynamics",
                    "abstract": "A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable. Here, we develop an approach that simultaneously achieves both flexibility and tractability. The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data. This approach allows us to rapidly learn, sample from, and evaluate probabilities in deep generative models with thousands of layers or time steps, as well as to compute conditional and posterior probabilities under the learned model. We additionally release an open source reference implementation of the algorithm."
                },
                {
                    "title": "Scaling Up Approximate Value Iteration with Options: Better Policies with Fewer Iterations",
                    "abstract": "We show how options, a class of control structures encompassing primitive and temporally extended actions, can play a valuable role in planning in MDPs with continuous state-spaces. Analyzing the convergence rate of Approximate Value Iteration with options reveals that for pessimistic initial value function estimates, options can speed up convergence compared to planning with only primitive actions even when the temporally extended actions are suboptimal and sparsely scattered throughout the state-space. Our experimental results in an optimal replacement task and a complex inventory management task demonstrate the potential for options to speed up convergence in practice. We show that options induce faster convergence to the optimal value function, which implies deriving better policies with fewer iterations."
                },
                {
                    "title": "AUTO-ENCODING VARIATIONAL BAYES",
                    "abstract": "To make decisions based on a model fit by Auto-Encoding Variational Bayes (AEVB), practitioners typically use importance sampling to estimate a functional of the posterior distribution. The variational distribution found by AEVB serves as the proposal distribution for importance sampling. However, this proposal distribution may give unreliable (high variance) importance sampling estimates, thus leading to poor decisions. We explore how changing the objective function for learning the variational distribution, while continuing to learn the generative model based on the ELBO, affects the quality of downstream decisions. For a particular model, we characterize the error of importance sampling as a function of posterior variance and show that proposal distributions learned with evidence upper bounds are better. Motivated by these theoretical results, we propose a novel variant of the VAE. In addition to experimenting with MNIST, we present a full-fledged application of the proposed method to single-cell RNA sequencing. In this challenging instance of multiple hypothesis testing, the proposed method surpasses the current state of the art."
                },
                {
                    "title": "GENERATIVE ADVERSARIAL NETS",
                    "abstract": "Estimating individualized treatment effects (ITE) is a challenging task due to the need for an individual\u2019s potential outcomes to be learned from biased data and without having access to the counterfactuals. We propose a novel method for inferring ITE based on the Generative Adversarial Nets (GANs) framework. Our method, termed Generative Adversarial Nets for inference of Individualized Treatment Effects (GANITE), is motivated by the possibility that we can capture the uncertainty in the counterfactual distributions by attempting to learn them using a GAN. We generate proxies of the counterfactual outcomes using a counterfactual generator, G, and then pass these proxies to an ITE generator, I, in order to train it. By modeling both of these using the GAN framework, we are able to infer based on the factual data, while still accounting for the unseen counterfactuals. We test our method on three real-world datasets (with both binary and multiple treatments) and show that GANITE outperforms state-of-the-art methods."
                },
                {
                    "title": "Reinforcement Learning: An Introduction",
                    "abstract": "Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives when interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the key ideas and algorithms of reinforcement learning. Their discussion ranges from the history of the field's intellectual foundations to the most recent developments and applications. The only necessary mathematical background is familiarity with elementary concepts of probability. The book is divided into three parts. Part I defines the reinforcement learning problem in terms of Markov decision processes. Part II provides basic solution methods: dynamic programming, Monte Carlo methods, and temporal-difference learning. Part III presents a unified view of the solution methods and incorporates artificial neural networks, eligibility traces, and planning; the two final chapters present case studies and consider the future of reinforcement learning."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively learn optimal policies in offline reinforcement learning without suffering from distributional shift and extrapolation errors, while leveraging generative models to model data distributions?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing offline reinforcement learning, as it addresses the limitations of existing methods that struggle with suboptimal trajectories and distributional shifts. By developing a robust approach that utilizes generative models without pessimism or trade-offs, this research could lead to more effective learning algorithms that can be applied in various domains, such as robotics, healthcare, and autonomous systems. The implications of this work could foster new research directions in offline RL and enhance the practical applicability of RL in real-world scenarios.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities of offline RL, particularly the distributional shift that occurs when stitching together suboptimal trajectories. Naive approaches may fail because they do not adequately account for the limitations of the training data, leading to extrapolation errors when estimating Q-values. Additionally, achieving a good return estimate for arbitrary states is non-trivial, and conditioning on out-of-support returns can result in low-value sequences. Overcoming these technical and theoretical obstacles requires innovative methodologies that can effectively model complex data distributions without introducing biases.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the reliance on pessimistic Q-learning methods, which can hinder performance by constraining the policy too much. Existing generative models have struggled to avoid sampling out-of-support sequences, leading to ineffective learning. Additionally, prior approaches that utilized latent spaces often lacked the richness needed for effective policy learning. Our approach differs by employing a latent diffusion model that can better capture complex distributions, allowing for more flexible and effective Q-learning without the pitfalls of previous methods.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using a latent diffusion model to generate candidate actions while learning a Q-function that is batch-constrained. We will utilize a dataset of offline trajectories and evaluate our approach using metrics such as average return and policy performance on benchmark tasks. The expected outcomes include improved policy learning that effectively addresses distributional shift and extrapolation errors, leading to higher performance in offline RL tasks compared to existing methods."
            }
        },
        "author_data": {
            "c33d6db4-03ab-46fc-84d6-4193d7718744": {
                "pk": "c33d6db4-03ab-46fc-84d6-4193d7718744",
                "name": "Siddarth Venkatraman",
                "collaborators": [
                    "J. Schneider",
                    "S. Daftry",
                    "Neil Abcouwer",
                    "Jialin Song",
                    "Yisong Yue",
                    "M. Ono",
                    "Shivam Agarwal",
                    "Moksh Jain",
                    "Luca Scimeca",
                    "Minsu Kim"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Robotics",
                    "Generative Models",
                    "Steganography"
                ],
                "publications": [
                    {
                        "title": "Amortizing intractable inference in diffusion models for vision, language, and control",
                        "abstract": "Diffusion models have emerged as effective distribution estimators in vision, language, and reinforcement learning, but their use as priors in downstream tasks poses an intractable posterior inference problem. This paper studies amortized sampling of the posterior over data, $\\mathbf{x}\\sim p^{\\rm post}(\\mathbf{x})\\propto p(\\mathbf{x})r(\\mathbf{x})$, in a model that consists of a diffusion generative model prior $p(\\mathbf{x})$ and a black-box constraint or likelihood function $r(\\mathbf{x})$. We state and prove the asymptotic correctness of a data-free learning objective, relative trajectory balance, for training a diffusion model that samples from this posterior, a problem that existing methods solve only approximately or in restricted cases. Relative trajectory balance arises from the generative flow network perspective on diffusion models, which allows the use of deep reinforcement learning techniques to improve mode coverage. Experiments illustrate the broad potential of unbiased inference of arbitrary posteriors under diffusion priors: in vision (classifier guidance), language (infilling under a discrete diffusion LLM), and multimodal data (text-to-image generation). Beyond generative modeling, we apply relative trajectory balance to the problem of continuous control with a score-based behavior prior, achieving state-of-the-art results on benchmarks in offline reinforcement learning."
                    },
                    {
                        "title": "Multi-Alpha Soft Actor-Critic: Overcoming Stochastic Biases in Maximum Entropy Reinforcement Learning",
                        "abstract": "The successful application of robotic control requires intelligent decision-making to handle the long tail of complex scenarios that arise in real-world environments. Recently, Deep Reinforcement Learning (DRL) has provided a data-driven framework to automatically learn effective policies in such complex settings. Since its introduction in 2018, Soft Actor-Critic (SAC) remains as one of the most popular off-policy DRL algorithms and has been used extensively to learn performant robotic control policies. However, in this paper we argue that by relying on the maximum entropy formalism to define learning objectives, previous work introduces a significant bias away from optimal decision making, which often requires near-deterministic behaviour for high-precision tasks. Moreover, we show that when training with the original variants of SAC, overcoming this bias by reducing entropy budgets or entropy coefficients introduces separate issues that lead to slow or unstable learning. We address these shortcomings by treating the entropy coefficient $\\alpha$ as a random variable and introduce Multi-Alpha Soft Actor-Critic (MAS). We show how MAS overcomes the stochastic bias of SAC in a variety of robotic control tasks including the CARLA urban-driving simulator, while maintaining the stability and sample efficiency of the original algorithms."
                    },
                    {
                        "title": "Learning Temporally AbstractWorld Models without Online Experimentation",
                        "abstract": "Agents that can build temporally abstract representations of their environment are better able to understand their world and make plans on extended time scales, with limited computational power and modeling capacity. However, existing methods for automatically learning temporally abstract world models usually require millions of online environmental interactions and incentivize agents to reach every accessible environmental state, which is infeasible for most real-world robots both in terms of data efficiency and hardware safety. In this paper, we present an approach for simultaneously learning sets of skills and temporally abstract, skill-conditioned world models purely from offline data, enabling agents to perform zero-shot online planning of skill sequences for new tasks. We show that our approach performs comparably to or better than a wide array of state-of-the-art offline RL algorithms on a number of simulated robotics locomotion and manipulation benchmarks, while offering a higher degree of adaptability to new goals. Finally, we show that our approach offers a much higher degree of robustness to perturbations in environmental dynamics, compared to policy-based methods."
                    },
                    {
                        "title": "MLNav: Learning to Safely Navigate on Martian Terrains",
                        "abstract": "We present MLNav, a learning-enhanced path planning framework for safety-critical and resource-limited systems operating in complex environments, such as rovers navigating on Mars. MLNav makes judicious use of machine learning to enhance the efficiency of path planning while fully respecting safety constraints. In particular, the dominant computational cost in such safety-critical settings is running a model-based safety checker on the proposed paths. Our learned search heuristic can simultaneously predict the feasibility for all path options in a single run, and the model-based safety checker is only invoked on the top-scoring paths. We validate in high-fidelity simulations using both real Martian terrain data collected by the Perseverance rover, as well as a suite of challenging synthetic terrains. Our experiments show that: (i) compared to the baseline ENav path planner on board the Perserverance rover, MLNav can provide a significant improvement in multiple key metrics, such as a 10x reduction in collision checks when navigating real Martian terrains, despite being trained with synthetic terrains; and (ii) MLNav can successfully navigate highly challenging terrains where the baseline ENav fails to find a feasible path before timing out."
                    },
                    {
                        "title": "Deep Residual Neural Networks for Image in Audio Steganography (Workshop Paper)",
                        "abstract": "Steganography is the art of hiding a secret message inside a publicly visible carrier message. Ideally, it is done without modifying the carrier, and with minimal loss of information in the secret message. Recently, various deep learning based approaches to steganography have been applied to different message types. We propose a deep learning based technique to hide a source RGB image message inside finite length speech segments without perceptual loss. To achieve this, we train three neural networks; an encoding network to hide the message in the carrier, a decoding network to reconstruct the message from the carrier and an additional image enhancer network to further improve the reconstructed message. We also discuss future improvements to the algorithm proposed."
                    },
                    {
                        "title": "Deep Residual Neural Networks for Image in Speech Steganography",
                        "abstract": "Steganography is the art of hiding a secret message inside a publicly visible carrier message. Ideally, it is done without modifying the carrier, and with minimal loss of information in the secret message. Recently, various deep learning based approaches to steganography have been applied to different message types. We propose a deep learning based technique to hide a source RGB image message inside finite length speech segments without perceptual loss. To achieve this, we train three neural networks; an encoding network to hide the message in the carrier, a decoding network to reconstruct the message from the carrier and an additional image enhancer network to further improve the reconstructed message. We also discuss future improvements to the algorithm proposed."
                    },
                    {
                        "title": "Machine Learning Based Path Planning for Improved Rover Navigation",
                        "abstract": "Enhanced AutoNav (ENav), the baseline surface navigation software for NASA's Perseverance rover, sorts a list of candidate paths for the rover to traverse, then uses the Approximate Clearance Evaluation (ACE) algorithm to evaluate whether the most highly ranked paths are safe. ACE is crucial for maintaining the safety of the rover, but is computationally expensive. If the most promising candidates in the list of paths are all found to be infeasible, ENav must continue to search the list and run time-consuming ACE evaluations until a feasible path is found. In this paper, we present two heuristics that, given a terrain heightmap around the rover, produce cost estimates that more effectively rank the candidate paths before ACE evaluation. The first heuristic uses Sobel operators and convolution to incorporate the cost of traversing high-gradient terrain. The second heuristic uses a machine learning (ML) model to predict areas that will be deemed untraversable by ACE. We used physics simulations to collect training data for the ML model and to run Monte Carlo trials to quantify navigation performance across a variety of terrains with various slopes and rock distributions. Compared to ENav's baseline performance, integrating the heuristics can lead to a significant reduction in ACE evaluations and average computation time per planning cycle, increase path efficiency, and maintain or improve the rate of successful traverses. This strategy of targeting specific bottlenecks with ML while maintaining the original ACE safety checks provides an example of how ML can be infused into planetary science missions and other safety-critical software."
                    }
                ]
            },
            "a5868c08-9a47-4261-bec1-dc0dcbdcc909": {
                "pk": "a5868c08-9a47-4261-bec1-dc0dcbdcc909",
                "name": "Shivesh Khaitan",
                "collaborators": [
                    "J. Dolan",
                    "Qin Lin"
                ],
                "domain": [
                    "Autonomous Driving",
                    "Deep Reinforcement Learning",
                    "Motion Planning",
                    "Uncertainty Modeling"
                ],
                "publications": [
                    {
                        "title": "State Dropout-Based Curriculum Reinforcement Learning for Self-Driving at Unsignalized Intersections",
                        "abstract": "Traversing intersections is a challenging problem for autonomous vehicles, especially when the intersections do not have traffic control. Recently deep reinforcement learning has received massive attention due to its success in dealing with autonomous driving tasks. In this work, we address the problem of traversing unsignalized intersections using a novel curriculum for deep reinforcement learning. The proposed curriculum leads to: 1) A faster training process for the reinforcement learning agent, and 2) Better performance compared to an agent trained without curriculum. Our main contribution is two-fold: 1) Presenting a unique curriculum for training deep reinforcement learning agents, and 2) demonstrating the performance improvement using the proposed curriculum in the unsignalized intersection traversal task. The framework expects processed observations of the surroundings from the perception system of the autonomous vehicle. We test our method in the CommonRoad motion planning simulator on T-intersections and four-way intersections."
                    },
                    {
                        "title": "Safe Planning and Control Under Uncertainty for Self-Driving",
                        "abstract": "Motion planning under uncertainty is critical for safe self-driving. This paper proposes a unified obstacle avoidance framework that deals with 1) uncertainty in ego-vehicle motion; and 2) prediction uncertainty of dynamic obstacles from the environment. A two-stage traffic participant trajectory predictor comprising short-term and long-term prediction is used in the planning layer to generate safe but not over-conservative trajectories for the ego-vehicle. The prediction module cooperates well with existing planning approaches. Our work showcases its effectiveness in a Frenet frame planner. A robust controller using tube MPC guarantees safe execution of the trajectory in the presence of state noise and dynamic model uncertainty. A Gaussian process regression model is used for on-line identification of the uncertainty's bound. We demonstrate the effectiveness, safety, and real-time performance of our framework in the CARLA simulator."
                    }
                ]
            },
            "53e95279-f554-4905-a629-23d2a14a3710": {
                "pk": "53e95279-f554-4905-a629-23d2a14a3710",
                "name": "Ravi Tej Akella",
                "collaborators": [
                    "Vinod Pankajakshan",
                    "Raviteja Rekula",
                    "Shirsendu Sukanta Halder",
                    "Arunav Shandeelya",
                    "Benjamin Eysenbach",
                    "Jeff Schneider",
                    "R. Salakhutdinov"
                ],
                "domain": [
                    "Computer Vision",
                    "Image Processing",
                    "Reinforcement Learning",
                    "Security"
                ],
                "publications": [
                    {
                        "title": "Enhancing Perceptual Loss with Adversarial Feature Matching for Super-Resolution",
                        "abstract": "Single image super-resolution (SISR) is an ill-posed problem with an indeterminate number of valid solutions. Solving this problem with neural networks would require access to extensive experience, either presented as a large training set over natural images or a condensed representation from another pre-trained network. Perceptual loss functions, which belong to the latter category, have achieved breakthrough success in SISR and several other computer vision tasks. While perceptual loss plays a central role in the generation of photo-realistic images, it also produces undesired pattern artifacts in the super-resolved outputs. In this paper, we show that the root cause of these pattern artifacts can be traced back to a mismatch between the pre-training objective of perceptual loss and the super-resolution objective. To address this issue, we propose to augment the existing perceptual loss formulation with a novel content loss function that uses the latent features of a discriminator network to filter the unwanted artifacts across several levels of adversarial similarity. Further, our modification has a stabilizing effect on non-convex optimization in adversarial training. The proposed approach offers notable gains in perceptual quality based on an extensive human evaluation study and a competent reconstruction fidelity when tested on objective evaluation metrics."
                    },
                    {
                        "title": "A Randomized Kernel-Based Secret Image Sharing Scheme \u2014Supplementary\u2014",
                        "abstract": "The proposed ( k; n ) scheme from share generation to secret reconstruction is described in the figure. Sharing the kernel not just prevents it from becoming SPOF, but also makes the scheme more secure. The kernel plays a vital role in the reconstruction of the secret image."
                    },
                    {
                        "title": "A Randomized Kernel-Based Secret Image Sharing Scheme",
                        "abstract": "This paper proposes a (k, n)-threshold secret image sharing scheme that offers flexibility in terms of meeting contrasting demands such as information security and storage efficiency with the help of a randomized kernel (binary matrix) operation. A secret image is split into n shares such that any k or more shares (k $\\leq$ n) can be used to reconstruct the image. Each share has a size less than or at most equal to the size of the secret image. Security and share sizes are solely determined by the kernel of the scheme. The kernel operation is optimized in terms of the security and computational requirements. The storage overhead of the kernel can further be made independent of its size by efficiently storing it as a sparse matrix. Moreover, the scheme is free from any kind of single point of failure (SPOF)."
                    },
                    {
                        "title": "Distributional Distance Classifiers for Goal-Conditioned Reinforcement Learning",
                        "abstract": "Autonomous systems are increasingly being deployed in stochastic real-world environments. Often, these agents are trying to find the shortest path to a commanded goal. But what does it mean to find the shortest path in stochastic environments, where every strategy has a non-zero probability of failing? At the core of this question is a conflict between two seemingly-natural notions of planning: maximizing the probability of reaching a goal state, and minimizing the expected number of steps to reach that goal state. Reinforcement learning (RL) methods based on minimizing the steps to a goal make an implicit assumption: that the goal is always reached, at least within some finite horizon. This assumption is violated in practical settings and can lead to very suboptimal strategies. In this paper, we bridge the gap between these two notions of planning by estimating the probability of reaching the goal at different future timesteps. This is not the same as estimating the distance to the goal \u2013 rather, probabilities convey uncertainty in ever reaching the goal at all. Our value function will resemble that used in distribution RL, but will be used to solve (reward-free) goal-reaching tasks rather than (single) reward-maximization tasks. Taken together, we believe that our results provide a cogent framework for thinking about probabilities and distances in stochastic settings, along with a practical and effective algorithm for goal-conditioned RL."
                    }
                ]
            },
            "816af70a-bc6f-4955-b3bd-f23949a14fa4": {
                "pk": "816af70a-bc6f-4955-b3bd-f23949a14fa4",
                "name": "John Dolan",
                "collaborators": [
                    "Nathan Ludlow",
                    "Yiwei Lyu",
                    "Adam R. Villaflor",
                    "Brian Yang",
                    "Huangyuan Su",
                    "Katerina Fragkiadaki",
                    "Jeff Schneider"
                ],
                "domain": [
                    "Autonomous Driving",
                    "Trajectory Prediction",
                    "Deep Learning",
                    "Human Behavior Modeling"
                ],
                "publications": [
                    {
                        "title": "Hierarchical Learned Risk-Aware Planning Framework for Human Driving Modeling",
                        "abstract": "This paper presents a novel approach to modeling human driving behavior, designed for use in evaluating autonomous vehicle control systems in a simulation environments. Our methodology leverages a hierarchical forward-looking, risk-aware estimation framework with learned parameters to generate human-like driving trajectories, accommodating multiple driver levels determined by model parameters. This approach is grounded in multimodal trajectory prediction, using a deep neural network with LSTM-based social pooling to predict the trajectories of surrounding vehicles. These trajectories are used to compute forward-looking risk assessments along the ego vehicle\u2019s path, guiding its navigation. Our method aims to replicate human driving behaviors by learning parameters that emulate human decision-making during driving. We ensure that our model exhibits robust generalization capabilities by conducting simulations, employing real-world driving data to validate the accuracy of our approach in modeling human behavior. The results reveal that our model effectively captures human behavior, showcasing its versatility in modeling human drivers in diverse highway scenarios."
                    },
                    {
                        "title": "Tractable Joint Prediction and Planning over Discrete Behavior Modes for Urban Driving",
                        "abstract": "Significant progress has been made in training multimodal trajectory forecasting models for autonomous driving. However, effectively integrating these models with downstream planners and model-based control approaches is still an open problem. Although these models have conventionally been evaluated for open-loop prediction, we show that they can be used to parameterize autoregressive closed-loop models without retraining. We consider recent trajectory prediction approaches which leverage learned anchor embeddings to predict multiple trajectories, finding that these anchor embeddings can parameterize discrete and distinct modes representing high-level driving behaviors. We propose to perform fully reactive closed-loop planning over these discrete latent modes, allowing us to tractably model the causal interactions between agents at each step. We validate our approach on a suite of more dynamic merging scenarios, finding that our approach avoids the frozen robot problem which is pervasive in conventional planners. Our approach also outperforms the previous state-of-the-art in CARLA on challenging dense traffic scenarios when evaluated at realistic speeds."
                    }
                ]
            },
            "ae694b66-addd-4479-b566-d3c4e07d63c2": {
                "pk": "ae694b66-addd-4479-b566-d3c4e07d63c2",
                "name": "Jeff Schneider",
                "collaborators": [
                    "Ravi Tej Akella",
                    "Benjamin Eysenbach",
                    "R. Salakhutdinov"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Autonomous Systems",
                    "Stochastic Environments"
                ],
                "publications": [
                    {
                        "title": "Distributional Distance Classifiers for Goal-Conditioned Reinforcement Learning",
                        "abstract": "Autonomous systems are increasingly being deployed in stochastic real-world environments. Often, these agents are trying to find the shortest path to a commanded goal. But what does it mean to find the shortest path in stochastic environments, where every strategy has a non-zero probability of failing? At the core of this question is a conflict between two seemingly-natural notions of planning: maximizing the probability of reaching a goal state, and minimizing the expected number of steps to reach that goal state. Reinforcement learning (RL) methods based on minimizing the steps to a goal make an implicit assumption: that the goal is always reached, at least within some finite horizon. This assumption is violated in practical settings and can lead to very suboptimal strategies. In this paper, we bridge the gap between these two notions of planning by estimating the probability of reaching the goal at different future timesteps. This is not the same as estimating the distance to the goal \u2013 rather, probabilities convey uncertainty in ever reaching the goal at all. Our value function will resemble that used in distribution RL, but will be used to solve (reward-free) goal-reaching tasks rather than (single) reward-maximization tasks. Taken together, we believe that our results provide a cogent framework for thinking about probabilities and distances in stochastic settings, along with a practical and effective algorithm for goal-conditioned RL."
                    }
                ]
            },
            "d25da009-420f-4327-af18-5a56d90f9505": {
                "pk": "d25da009-420f-4327-af18-5a56d90f9505",
                "name": "Glen Berseth",
                "collaborators": [
                    "Roger Creus Castanyer",
                    "Adriana Hugessen",
                    "Raj Ghugare",
                    "Faisal Mohamed",
                    "Mingqi Yuan",
                    "Bo Li",
                    "Xin Jin",
                    "Wenjun Zeng",
                    "Matthieu Geist",
                    "Benjamin Eysenbach"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Intrinsic Motivation",
                    "Generative Models",
                    "Molecular Design"
                ],
                "publications": [
                    {
                        "title": "Surprise-Adaptive Intrinsic Motivation for Unsupervised Reinforcement Learning",
                        "abstract": "Both entropy-minimizing and entropy-maximizing (curiosity) objectives for unsupervised reinforcement learning (RL) have been shown to be effective in different environments, depending on the environment's level of natural entropy. However, neither method alone results in an agent that will consistently learn intelligent behavior across environments. In an effort to find a single entropy-based method that will encourage emergent behaviors in any environment, we propose an agent that can adapt its objective online, depending on the entropy conditions by framing the choice as a multi-armed bandit problem. We devise a novel intrinsic feedback signal for the bandit, which captures the agent's ability to control the entropy in its environment. We demonstrate that such agents can learn to control entropy and exhibit emergent behaviors in both high- and low-entropy regimes and can learn skillful behaviors in benchmark tasks. Videos of the trained agents and summarized findings can be found on our project page https://sites.google.com/view/surprise-adaptive-agents"
                    },
                    {
                        "title": "RLeXplore: Accelerating Research in Intrinsically-Motivated Reinforcement Learning",
                        "abstract": "Extrinsic rewards can effectively guide reinforcement learning (RL) agents in specific tasks. However, extrinsic rewards frequently fall short in complex environments due to the significant human effort needed for their design and annotation. This limitation underscores the necessity for intrinsic rewards, which offer auxiliary and dense signals and can enable agents to learn in an unsupervised manner. Although various intrinsic reward formulations have been proposed, their implementation and optimization details are insufficiently explored and lack standardization, thereby hindering research progress. To address this gap, we introduce RLeXplore, a unified, highly modularized, and plug-and-play framework offering reliable implementations of eight state-of-the-art intrinsic reward algorithms. Furthermore, we conduct an in-depth study that identifies critical implementation details and establishes well-justified standard practices in intrinsically-motivated RL. The source code for RLeXplore is available at https://github.com/RLE-Foundation/RLeXplore."
                    },
                    {
                        "title": "Closing the Gap between TD Learning and Supervised Learning - A Generalisation Point of View",
                        "abstract": "Some reinforcement learning (RL) algorithms can stitch pieces of experience to solve a task never seen before during training. This oft-sought property is one of the few ways in which RL methods based on dynamic-programming differ from RL methods based on supervised-learning (SL). Yet, certain RL methods based on off-the-shelf SL algorithms achieve excellent results without an explicit mechanism for stitching; it remains unclear whether those methods forgo this important stitching property. This paper studies this question for the problems of achieving a target goal state and achieving a target return value. Our main result is to show that the stitching property corresponds to a form of combinatorial generalization: after training on a distribution of (state, goal) pairs, one would like to evaluate on (state, goal) pairs not seen together in the training data. Our analysis shows that this sort of generalization is different from i.i.d. generalization. This connection between stitching and generalisation reveals why we should not expect SL-based RL methods to perform stitching, even in the limit of large datasets and models. Based on this analysis, we construct new datasets to explicitly test for this property, revealing that SL-based methods lack this stitching property and hence fail to perform combinatorial generalization. Nonetheless, the connection between stitching and combinatorial generalisation also suggests a simple remedy for improving generalisation in SL: data augmentation. We propose a temporal data augmentation and demonstrate that adding it to SL-based methods enables them to successfully complete tasks not seen together during training. On a high level, this connection illustrates the importance of combinatorial generalization for data efficiency in time-series data beyond tasks beyond RL, like audio, video, or text."
                    },
                    {
                        "title": "Amortizing intractable inference in diffusion models for vision, language, and control",
                        "abstract": "Diffusion models have emerged as effective distribution estimators in vision, language, and reinforcement learning, but their use as priors in downstream tasks poses an intractable posterior inference problem. This paper studies amortized sampling of the posterior over data, $\\mathbf{x}\\sim p^{\\rm post}(\\mathbf{x})\\propto p(\\mathbf{x})r(\\mathbf{x})$, in a model that consists of a diffusion generative model prior $p(\\mathbf{x})$ and a black-box constraint or likelihood function $r(\\mathbf{x})$. We state and prove the asymptotic correctness of a data-free learning objective, relative trajectory balance, for training a diffusion model that samples from this posterior, a problem that existing methods solve only approximately or in restricted cases. Relative trajectory balance arises from the generative flow network perspective on diffusion models, which allows the use of deep reinforcement learning techniques to improve mode coverage. Experiments illustrate the broad potential of unbiased inference of arbitrary posteriors under diffusion priors: in vision (classifier guidance), language (infilling under a discrete diffusion LLM), and multimodal data (text-to-image generation). Beyond generative modeling, we apply relative trajectory balance to the problem of continuous control with a score-based behavior prior, achieving state-of-the-art results on benchmarks in offline reinforcement learning."
                    },
                    {
                        "title": "Searching for High-Value Molecules Using Reinforcement Learning and Transformers",
                        "abstract": "Reinforcement learning (RL) over text representations can be effective for finding high-value policies that can search over graphs. However, RL requires careful structuring of the search space and algorithm design to be effective in this challenge. Through extensive experiments, we explore how different design choices for text grammar and algorithmic choices for training can affect an RL policy's ability to generate molecules with desired properties. We arrive at a new RL-based molecular design algorithm (ChemRLformer) and perform a thorough analysis using 25 molecule design tasks, including computationally complex protein docking simulations. From this analysis, we discover unique insights in this problem space and show that ChemRLformer achieves state-of-the-art performance while being more straightforward than prior work by demystifying which design choices are actually helpful for text-based molecule design."
                    },
                    {
                        "title": "Improving Intrinsic Exploration by Creating Stationary Objectives",
                        "abstract": "Exploration bonuses in reinforcement learning guide long-horizon exploration by defining custom intrinsic objectives. Count-based methods use the frequency of state visits to derive an exploration bonus. In this paper, we identify that any intrinsic reward function derived from count-based methods is non-stationary and hence induces a difficult objective to optimize for the agent. The key contribution of our work lies in transforming the original non-stationary rewards into stationary rewards through an augmented state representation. For this purpose, we introduce the Stationary Objectives For Exploration (SOFE) framework. SOFE requires identifying sufficient statistics for different exploration bonuses and finding an efficient encoding of these statistics to use as input to a deep network. SOFE is based on proposing state augmentations that expand the state space but hold the promise of simplifying the optimization of the agent's objective. Our experiments show that SOFE improves the agents' performance in challenging exploration problems, including sparse-reward tasks, pixel-based observations, 3D navigation, and procedurally generated environments."
                    }
                ]
            }
        }
    },
    "2312.02438": {
        "paper_data": {
            "title": "Adaptive Instrument Design for Indirect Experiments",
            "url": "http://arxiv.org/abs/2312.02438v1",
            "arxiv_id": "2312.02438",
            "authors": [
                "Yash Chandak",
                "Shiv Shankar",
                "Vasilis Syrgkanis",
                "Emma Brunskill"
            ],
            "abstract": "Indirect experiments provide a valuable framework for estimating treatment effects in situations where conducting randomized control trials (RCTs) is impractical or unethical. Unlike RCTs, indirect experiments estimate treatment effects by leveraging (conditional) instrumental variables, enabling estimation through encouragement and recommendation rather than strict treatment assignment. However, the sample efficiency of such estimators depends not only on the inherent variability in outcomes but also on the varying compliance levels of users with the instrumental variables and the choice of estimator being used, especially when dealing with numerous instrumental variables. While adaptive experiment design has a rich literature for direct experiments, in this paper we take the initial steps towards enhancing sample efficiency for indirect experiments by adaptively designing a data collection policy over instrumental variables. Our main contribution is a practical computational procedure that utilizes influence functions to search for an optimal data collection policy, minimizing the mean-squared error of the desired (non-linear) estimator. Through experiments conducted in various domains inspired by real-world applications, we showcase how our method can significantly improve the sample efficiency of indirect experiments.",
            "introduction": " introduction to influence functions). Therefore, when all the higher order in- fluence function exists, we can use 8 to recover an estimate of MSE( \u03b8(Dn))\u2212MSE( \u03b8(Dn\\i))with- outre-training. In practice, approximating 8 with a K-th order expansion often suffices, which per- mits a gradient estimator \u02c6\u2207IFL(\u03c0)that avoids any re-computation of the mean-square-error, \u02c6\u2207IFL(\u03c0):=nX i=1\u2202log\u03c0(Zi|Xi) \u2202\u03c0IMSE(Si),where, IMSE(Si):=\u2212KX k=1\u03f5k k!I(k) MSE(Si).(9) In fact, as we illustrate in Figure 3 and state formally in Section 4, when using finite Kterms, the bias of the gradient in 9 is of the order n\u2212K, while still providing significant variance reduction. This en- ables the bias to be neglected even when using K= 1. Intuitively, for k= 1,I(k) MSE operationalizes the concept of the first-order derivative for the functional MSE( \u03b8(\u00b7)), i.e., it characterizes the effect onMSE when the training sample Siis infinitesimaly up-weighted during optimization, I(1) MSE(Si):=d dtMSE\u0010 \u03b8\u0010 (1\u2212t)D+t\u03b4(Si)\u0011\u0011\f\f\f t=0. (10) A key contribution is to derive (Theorem 2in Results for the TripAdvisor domain (Syrgkanis et al., 2019) (using neural network based estimator). Error bars correspond to one stderr. Unlike Property 1, Property 2holds irrespective of \u03b1(n), thus providing a reliable gradient estimator even if the function approximator is mis-specified, or optimization is susceptible to local optima, or if 2 is ill-posed when dealing with non-parametric estimators. Further, our loss using influence functions 9 can also yield similar variance reduction, at much lower computational cost: Property 3 (Informal) .If the Von-Mises expansion in 8 exists, E[\u02c6\u2207IFL(\u03c0)] =\u2207L(\u03c0) +O(n\u2212K). IfMSE( \u03b8(Dn))is leave-one-out stable then Var(\u02c6\u2207IFL(\u03c0)) =O(1)(c.f. Theorem 5 in REFERENCES Karim Abou-Moustafa and Csaba Szepesv\u00e1ri. An exponential efron-stein inequality for l_qstable learning rules. In Algorithmic Learning Theory , pp. 31\u201363. PMLR, 2019.[Page(s): 27] Ahmed Alaa and Mihaela Van Der Schaar. Validating causal inference models via influence func- tions. In International Conference on Machine Learning , pp. 191\u2013201. PMLR, 2019.[Page(s): 17, 25] Ahmed Alaa and Mihaela Van Der Schaar. Discriminative jackknife: Quantifying uncertainty in deep learning via higher-order influence functions. In International Conference on Machine Learning , pp. 165\u2013174. PMLR, 2020.[Page(s): 17] Joshua D Angrist, Guido W Imbens, and Donald B Rubin. Identification of causal effects using instrumental variables. Journal of the American statistical Association , 91(434):444\u2013455, 1996. [Page(s): 2, 3, 17] Susan Athey and Stefan Wager. Policy learning with observational data. Econometrica , 89(1): 133\u2013161, 2021.[Page(s): 17] Juhan Bae, Nathan Ng, Alston Lo, Marzyeh Ghassemi, and Roger B Grosse. If influence functions are the answer, then what is the question? Advances in Neural Information Processing Systems , 35:17953\u201317967, 2022.[Page(s): 24, 46, 47] Samyadeep Basu, Philip Pope, and Soheil Feizi. Influence functions in deep learning are fragile. arXiv preprint arXiv:2006.14651 , 2020.[Page(s): 46] James Bradbury, Roy Frostig, Peter Hawkins, Matthew James Johnson, Chris Leary, Dougal Maclaurin, George Necula, Adam Paszke, Jake VanderPlas, Skye Wanderman-Milne, and Qiao Zhang. JAX: composable transformations of Python+NumPy programs, 2018. URL http: //github.com/google/jax .[Page(s): 5] Patrick Breheny. Statistical functionals and influence functions. Technical report, The University of Iowa, 2020. https://myweb.uiowa.edu/pbreheny/uk/teaching/621/notes/ 8-28.pdf .[Page(s): 24] Erik Brynjolfsson, Danielle Li, and Lindsey R Raymond. Generative ai at work. Technical report, National Bureau of Economic Research, 2023.[Page(s): 1] Brian S Caffo, James G Booth, and AC Davison. Empirical supremum rejection sampling. Biometrika , 89(4):745\u2013754, 2002.[Page(s): 45] Ricardo Cao. Bootstrapping the mean integrated squared error. Journal of Multivariate Analysis , 45 (1):137\u2013160, 1993.[Page(s): 17] Alain Celisse and Benjamin Guedj. Stability revisited: new generalisation bounds for the leave-one- out. arXiv preprint arXiv:1608.06412 , 2016.[Page(s): 27] Kathryn Chaloner and Isabella Verdinelli. Bayesian experimental design:",
            "references": [
                {
                    "title": "Double machine learning and design in batch adaptive experiments",
                    "abstract": "We consider an experiment with at least two stages or batches and $O(N)$ subjects per batch. First, we propose a semiparametric treatment effect estimator that efficiently pools information across the batches, and show it asymptotically dominates alternatives that aggregate single batch estimates. Then, we consider the design problem of learning propensity scores for assigning treatment in the later batches of the experiment to maximize the asymptotic precision of this estimator. For two common causal estimands, we estimate this precision using observations from previous batches, and then solve a finite-dimensional concave maximization problem to adaptively learn flexible propensity scores that converge to suitably defined optima in each batch at rate $O_p(N^{-1/4})$. By extending the framework of double machine learning, we show this rate suffices for our pooled estimator to attain the targeted precision after each batch, as long as nuisance function estimates converge at rate $o_p(N^{-1/4})$. These relatively weak rate requirements enable the investigator to avoid the common practice of discretizing the covariate space for design and estimation in batch adaptive experiments while maintaining the advantages of pooling. Our numerical study shows that such discretization often leads to substantial asymptotic and finite sample precision losses outweighing any gains from design."
                },
                {
                    "title": "Studying Large Language Model Generalization with Influence Functions",
                    "abstract": "When trying to gain better visibility into a machine learning model in order to understand and mitigate the associated risks, a potentially valuable source of evidence is: which training examples most contribute to a given behavior? Influence functions aim to answer a counterfactual: how would the model's parameters (and hence its outputs) change if a given sequence were added to the training set? While influence functions have produced insights for small models, they are difficult to scale to large language models (LLMs) due to the difficulty of computing an inverse-Hessian-vector product (IHVP). We use the Eigenvalue-corrected Kronecker-Factored Approximate Curvature (EK-FAC) approximation to scale influence functions up to LLMs with up to 52 billion parameters. In our experiments, EK-FAC achieves similar accuracy to traditional influence function estimators despite the IHVP computation being orders of magnitude faster. We investigate two algorithmic techniques to reduce the cost of computing gradients of candidate training sequences: TF-IDF filtering and query batching. We use influence functions to investigate the generalization patterns of LLMs, including the sparsity of the influence patterns, increasing abstraction with scale, math and programming abilities, cross-lingual generalization, and role-playing behavior. Despite many apparently sophisticated forms of generalization, we identify a surprising limitation: influences decay to near-zero when the order of key phrases is flipped. Overall, influence functions give us a powerful new tool for studying the generalization properties of LLMs."
                },
                {
                    "title": "ACE: Active Learning for Causal Inference with Expensive Experiments",
                    "abstract": "Experiments are the gold standard for causal inference. In many applications, experimental units can often be recruited or chosen sequentially, and the adaptive execution of such experiments may offer greatly improved inference of causal quantities over non-adaptive approaches, particularly when experiments are expensive. We thus propose a novel active learning method called ACE (Active learning for Causal inference with Expensive experiments), which leverages Gaussian process modeling of the conditional mean functions to guide an informed sequential design of costly experiments. In particular, we develop new acquisition functions for sequential design via the minimization of the posterior variance of a desired causal estimand. Our approach facilitates targeted learning of a variety of causal estimands, such as the average treatment effect (ATE), the average treatment effect on the treated (ATTE), and individualized treatment effects (ITE), and can be used for adaptive selection of an experimental unit and/or the applied treatment. We then demonstrate in a suite of numerical experiments the improved performance of ACE over baseline methods for estimating causal estimands given a limited number of experiments."
                },
                {
                    "title": "Accounting For Informative Sampling When Learning to Forecast Treatment Outcomes Over Time",
                    "abstract": "Machine learning (ML) holds great potential for accurately forecasting treatment outcomes over time, which could ultimately enable the adoption of more individualized treatment strategies in many practical applications. However, a significant challenge that has been largely overlooked by the ML literature on this topic is the presence of informative sampling in observational data. When instances are observed irregularly over time, sampling times are typically not random, but rather informative -- depending on the instance's characteristics, past outcomes, and administered treatments. In this work, we formalize informative sampling as a covariate shift problem and show that it can prohibit accurate estimation of treatment outcomes if not properly accounted for. To overcome this challenge, we present a general framework for learning treatment outcomes in the presence of informative sampling using inverse intensity-weighting, and propose a novel method, TESAR-CDE, that instantiates this framework using Neural CDEs. Using a simulation environment based on a clinical use case, we demonstrate the effectiveness of our approach in learning under informative sampling."
                },
                {
                    "title": "Helping the Helper: Supporting Peer Counselors via AI-Empowered Practice and Feedback",
                    "abstract": "Millions of users come to online peer counseling platforms to seek support on diverse topics ranging from relationship stress to anxiety. However, studies show that online peer support groups are not always as effective as expected largely due to users' negative experiences with unhelpful counselors. Peer counselors are key to the success of online peer counseling platforms, but most of them often do not have systematic ways to receive guidelines or supervision. In this work, we introduce CARE: an interactive AI-based tool to empower peer counselors through automatic suggestion generation. During the practical training stage, CARE helps diagnose which specific counseling strategies are most suitable in the given context and provides tailored example responses as suggestions. Counselors can choose to select, modify, or ignore any suggestion before replying to the support seeker. Building upon the Motivational Interviewing framework, CARE utilizes large-scale counseling conversation data together with advanced natural language generation techniques to achieve these functionalities. We demonstrate the efficacy of CARE by performing both quantitative evaluations and qualitative user studies through simulated chats and semi-structured interviews. We also find that CARE especially helps novice counselors respond better in challenging situations."
                },
                {
                    "title": "Generative AI at Work",
                    "abstract": "We study the staggered introduction of a generative AI-based conversational assistant using data from 5,000 customer support agents. Access to the tool increases productivity, as measured by issues resolved per hour, by 14 percent on average, with the greatest impact on novice and low-skilled workers, and minimal impact on experienced and highly skilled workers. We provide suggestive evidence that the AI model disseminates the potentially tacit knowledge of more able workers and helps newer workers move down the experience curve. In addition, we show that AI assistance improves customer sentiment, reduces requests for managerial intervention, and improves employee retention."
                },
                {
                    "title": "Modern Bayesian Experimental Design",
                    "abstract": "Bayesian experimental design (BED) provides a powerful and general framework for optimizing the design of experiments. However, its deployment often poses substantial computational challenges that can undermine its practical use. In this review, we outline how recent advances have transformed our ability to overcome these challenges and thus utilize BED effectively, before discussing some key areas for future development in the field."
                },
                {
                    "title": "Stochastic Online Instrumental Variable Regression: Regrets for Endogeneity and Bandit Feedback",
                    "abstract": "Endogeneity, i.e. the dependence of noise and covariates, is a common phenomenon in real data due to omitted variables, strategic behaviours, measurement errors etc. In contrast, the existing analyses of stochastic online linear regression with unbounded noise and linear bandits depend heavily on exogeneity, i.e. the independence of noise and covariates. Motivated by this gap, we study the over- and just-identified Instrumental Variable (IV) regression, specifically Two-Stage Least Squares, for stochastic online learning, and propose to use an online variant of Two-Stage Least Squares, namely O2SLS. We show that O2SLS achieves $\\mathcal O(d_{x}d_{z}\\log^2 T)$ identification and $\\widetilde{\\mathcal O}(\\gamma \\sqrt{d_{z} T})$ oracle regret after $T$ interactions, where $d_{x}$ and $d_{z}$ are the dimensions of covariates and IVs, and $\\gamma$ is the bias due to endogeneity. For $\\gamma=0$, i.e. under exogeneity, O2SLS exhibits $\\mathcal O(d_{x}^2 \\log^2 T)$ oracle regret, which is of the same order as that of the stochastic online ridge. Then, we leverage O2SLS as an oracle to design OFUL-IV, a stochastic linear bandit algorithm to tackle endogeneity. OFUL-IV yields $\\widetilde{\\mathcal O}(\\sqrt{d_{x}d_{z}T})$ regret that matches the regret lower bound under exogeneity. For different datasets with endogeneity, we experimentally show efficiencies of O2SLS and OFUL-IV."
                },
                {
                    "title": "Dataset Distillation with Convexified Implicit Gradients",
                    "abstract": "We propose a new dataset distillation algorithm using reparameterization and convexification of implicit gradients (RCIG), that substantially improves the state-of-the-art. To this end, we first formulate dataset distillation as a bi-level optimization problem. Then, we show how implicit gradients can be effectively used to compute meta-gradient updates. We further equip the algorithm with a convexified approximation that corresponds to learning on top of a frozen finite-width neural tangent kernel. Finally, we improve bias in implicit gradients by parameterizing the neural network to enable analytical computation of final-layer parameters given the body parameters. RCIG establishes the new state-of-the-art on a diverse series of dataset distillation tasks. Notably, with one image per class, on resized ImageNet, RCIG sees on average a 108\\% improvement over the previous state-of-the-art distillation algorithm. Similarly, we observed a 66\\% gain over SOTA on Tiny-ImageNet and 37\\% on CIFAR-100."
                },
                {
                    "title": "Instrumental Variables in Causal Inference and Machine Learning: A Survey",
                    "abstract": "Causal inference is the process of using assumptions, study designs, and estimation strategies to draw conclusions about the causal relationships between variables based on data. This allows researchers to better understand the underlying mechanisms at work in complex systems and make more informed decisions. In many settings, we may not fully observe all the confounders that affect both the treatment and outcome variables, complicating the estimation of causal effects. To address this problem, a growing literature in both causal inference and machine learning proposes to use Instrumental Variables (IV). This paper serves as the first effort to systematically and comprehensively introduce and discuss the IV methods and their applications in both causal inference and machine learning. First, we provide the formal definition of IVs and discuss the identification problem of IV regression methods under different assumptions. Second, we categorize the existing work on IV methods into three streams according to the focus on the proposed methods, including two-stage least squares with IVs, control function with IVs, and evaluation of IVs. For each stream, we present both the classical causal inference methods, and recent developments in the machine learning literature. Then, we introduce a variety of applications of IV methods in real-world scenarios and provide a summary of the available datasets and algorithms. Finally, we summarize the literature, discuss the open problems and suggest promising future research directions for IV methods and their applications. We also develop a toolkit of IVs methods reviewed in this survey at https://github.com/causal-machine-learning-lab/mliv."
                },
                {
                    "title": "If Influence Functions are the Answer, Then What is the Question?",
                    "abstract": "Influence functions efficiently estimate the effect of removing a single training data point on a model's learned parameters. While influence estimates align well with leave-one-out retraining for linear models, recent works have shown this alignment is often poor in neural networks. In this work, we investigate the specific factors that cause this discrepancy by decomposing it into five separate terms. We study the contributions of each term on a variety of architectures and datasets and how they vary with factors such as network width and training time. While practical influence function estimates may be a poor match to leave-one-out retraining for nonlinear networks, we show they are often a good approximation to a different object we term the proximal Bregman response function (PBRF). Since the PBRF can still be used to answer many of the questions motivating influence functions, such as identifying influential or mislabeled examples, our results suggest that current algorithms for influence function estimation give more informative results than previous error analyses would suggest."
                },
                {
                    "title": "Estimating individual treatment effects under unobserved confounding using binary instruments",
                    "abstract": "Estimating conditional average treatment effects (CATEs) from observational data is relevant in many fields such as personalized medicine. However, in practice, the treatment assignment is usually confounded by unobserved variables and thus introduces bias. A remedy to remove the bias is the use of instrumental variables (IVs). Such settings are widespread in medicine (e.g., trials where the treatment assignment is used as binary IV). In this paper, we propose a novel, multiply robust machine learning framework, called MRIV, for estimating CATEs using binary IVs and thus yield an unbiased CATE estimator. Different from previous work for binary IVs, our framework estimates the CATE directly via a pseudo outcome regression. (1)~We provide a theoretical analysis where we show that our framework yields multiple robust convergence rates: our CATE estimator achieves fast convergence even if several nuisance estimators converge slowly. (2)~We further show that our framework asymptotically outperforms state-of-the-art plug-in IV methods for CATE estimation, in the sense that it achieves a faster rate of convergence if the CATE is smoother than the individual outcome surfaces. (3)~We build upon our theoretical results and propose a tailored deep neural network architecture called MRIV-Net for CATE estimation using binary IVs. Across various computational experiments, we demonstrate empirically that our MRIV-Net achieves state-of-the-art performance. To the best of our knowledge, our MRIV is the first multiply robust machine learning framework tailored to estimating CATEs in the binary IV setting."
                },
                {
                    "title": "1009-P: The Association between Patient Characteristics and the Efficacy of Remote Patient Monitoring and Messaging",
                    "abstract": "In the Pilot 4T study (n=135) , remote patient monitoring (RPM) was associated with improved time in range (TIR) and lower A1c. Measuring differences in how patients respond to messages from the care team is essential for the design of effective, personalized care models. We analyzed electronic health record (EHR) and continuous glucose monitor (CGM) data to estimate the week-over-week impact of RPM messages on patients\u2019 TIR. We applied statistical clustering methods to divide patients who received messages into two groups based on their changes in TIR and compared the characteristics of these groups. Receiving a message was associated with a greater mean week-to-week improvement in TIR after a low-TIR week [4.9 percentage points (pp) after a message vs. 2.5pp without a message; p < 0.001]. A patient\u2019s TIR improvement after receiving a message is greater on average if the same patient\u2019s TIR improved following previous messages. We identified two groups of patients with significantly different responses to messages. The group with greater mean TIR improvement after messages contains a larger proportion of non-white, non-English speaking, and publicly insured patients. We found that messages were associated with different magnitudes of TIR improvement across two clusters of patients. Identifying patients who benefit more from RPM could facilitate the personalization of management strategies.\n \n \n \n J.Ferstad: None. P.Prahalad: None. D.M.Maahs: Advisory Panel; Abbott Diabetes, Eli Lilly and Company, Medtronic, Novo Nordisk, Sanofi, Consultant; Aditx Therapeutics, Inc., Biospex. E.Fox: None. R.Johari: None. D.Scheinker: None.\n \n \n \n Helmsley Charitable Trust, ISPAD-JDRF Fellowship,NIH R18DK122422,Stanford Diabetes Research Center,LPCH Auxiliaries,Stanford Maternal and Child Health Research Institute, The Stanford REDCap platform (http://redcap.stanford.edu) is developed and operated by Stanford Medicine Research IT team. The REDCap platform services at Stanford are subsidized by a) Stanford School of Medicine Research Office, and b) the National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health, through grant UL1 TR001085.\n"
                },
                {
                    "title": "Gradient Estimation with Discrete Stein Operators",
                    "abstract": "Gradient estimation -- approximating the gradient of an expectation with respect to the parameters of a distribution -- is central to the solution of many machine learning problems. However, when the distribution is discrete, most common gradient estimators suffer from excessive variance. To improve the quality of gradient estimation, we introduce a variance reduction technique based on Stein operators for discrete distributions. We then use this technique to build flexible control variates for the REINFORCE leave-one-out estimator. Our control variates can be adapted online to minimize variance and do not require extra evaluations of the target function. In benchmark generative modeling tasks such as training binary variational autoencoders, our gradient estimator achieves substantially lower variance than state-of-the-art estimators with the same number of function evaluations."
                },
                {
                    "title": "Biased Gradient Estimate with Drastic Variance Reduction for Meta Reinforcement Learning",
                    "abstract": "Despite the empirical success of meta reinforcement learning (meta-RL), there are still a number poorly-understood discrepancies between theory and practice. Critically, biased gradient estimates are almost always implemented in practice, whereas prior theory on meta-RL only establishes convergence under unbiased gradient estimates. In this work, we investigate such a discrepancy. In particular, (1) We show that unbiased gradient estimates have variance $\\Theta(N)$ which linearly depends on the sample size $N$ of the inner loop updates; (2) We propose linearized score function (LSF) gradient estimates, which have bias $\\mathcal{O}(1/\\sqrt{N})$ and variance $\\mathcal{O}(1/N)$; (3) We show that most empirical prior work in fact implements variants of the LSF gradient estimates. This implies that practical algorithms\"accidentally\"introduce bias to achieve better performance; (4) We establish theoretical guarantees for the LSF gradient estimates in meta-RL regarding its convergence to stationary points, showing better dependency on $N$ than prior work when $N$ is large."
                },
                {
                    "title": "Lifelong Hyper-Policy Optimization with Multiple Importance Sampling Regularization",
                    "abstract": "Learning in a lifelong setting, where the dynamics continually evolve, is a hard challenge for current reinforcement learning algorithms. Yet this would be a much needed feature for practical applications. \n In this paper, we propose an approach which learns a hyper-policy, whose input is time, that outputs the parameters of the policy to be queried at that time. \n This hyper-policy is trained to maximize the estimated future performance, efficiently reusing past data by means of importance sampling, at the cost of introducing a controlled bias. We combine the future performance estimate with the past performance to mitigate catastrophic forgetting.\n To avoid overfitting the collected data, we derive a differentiable variance bound that we embed as a penalization term. Finally, we empirically validate our approach, in comparison with state-of-the-art algorithms, on realistic environments, including water resource management and trading."
                },
                {
                    "title": "Scaling Up Influence Functions",
                    "abstract": "We address efficient calculation of influence functions for tracking predictions back to the training data. We propose and analyze a new approach to speeding up the inverse Hessian calculation based on Arnoldi iteration. With this improvement, we achieve, to the best of our knowledge, the first successful implementation of influence functions that scales to full-size (language and vision) Transformer models with several hundreds of millions of parameters. We evaluate our approach in image classification and sequence-to-sequence tasks with tens to a hundred of millions of training examples. Our code is available at https://github.com/google-research/jax-influence."
                },
                {
                    "title": "Double Control Variates for Gradient Estimation in Discrete Latent Variable Models",
                    "abstract": "Stochastic gradient-based optimization for discrete latent variable models is challenging due to the high variance of gradients. We introduce a variance reduction technique for score function estimators that makes use of double control variates . These control variates act on top of a main control variate, and try to further reduce the variance of the overall estimator. We develop a double control variate for the REINFORCE leave-one-out estimator using Taylor expansions. For training discrete latent variable models, such as variational autoencoders with binary latent variables, our approach adds no extra computational cost compared to standard training with the REINFORCE leave-one-out estimator. We apply our method to challenging high-dimensional toy examples and for training variational autoencoders with binary latent variables. We show that our estimator can have lower variance compared to other state-of-the-art estimators."
                },
                {
                    "title": "Efficient Online Estimation of Causal Effects by Deciding What to Observe",
                    "abstract": "Researchers often face data fusion problems, where multiple data sources are available, each capturing a distinct subset of variables. While problem formulations typically take the data as given, in practice, data acquisition can be an ongoing process. In this paper, we aim to estimate any functional of a probabilistic model (e.g., a causal effect) as efficiently as possible, by deciding, at each time, which data source to query. We propose online moment selection (OMS), a framework in which structural assumptions are encoded as moment conditions. The optimal action at each step depends, in part, on the very moments that identify the functional of interest. Our algorithms balance exploration with choosing the best action as suggested by current estimates of the moments. We propose two selection strategies: (1) explore-then-commit (OMS-ETC) and (2) explore-then-greedy (OMS-ETG), proving that both achieve zero asymptotic regret as assessed by MSE. We instantiate our setup for average treatment effect estimation, where structural assumptions are given by a causal graph and data sources may include subsets of mediators, confounders, and instrumental variables."
                },
                {
                    "title": "Incentivizing Compliance with Algorithmic Instruments",
                    "abstract": "Randomized experiments can be susceptible to selection bias due to potential non-compliance by the participants. While much of the existing work has studied compliance as a static behavior, we propose a game-theoretic model to study compliance as dynamic behavior that may change over time. In rounds, a social planner interacts with a sequence of heterogeneous agents who arrive with their unobserved private type that determines both their prior preferences across the actions (e.g., control and treatment) and their baseline rewards without taking any treatment. The planner provides each agent with a randomized recommendation that may alter their beliefs and their action selection. We develop a novel recommendation mechanism that views the planner\u2019s recommendation as a form of instrumental variable (IV) that only affects an agents\u2019 action selection, but not the observed rewards. We construct such IVs by carefully mapping the history \u2013the interactions between the planner and the previous agents\u2013 to a random recommendation. Even though the initial agents may be completely non-compliant, our mechanism can incentivize compliance over time, thereby enabling the estimation of the treatment effect of each treatment, and minimizing the cumulative regret of the planner whose goal is to identify the optimal treatment."
                },
                {
                    "title": "Auto IV: Counterfactual Prediction via Automatic Instrumental Variable Decomposition",
                    "abstract": "Instrumental variables (IVs), sources of treatment randomization that are conditionally independent of the outcome, play an important role in causal inference with unobserved confounders. However, the existing IV-based counterfactual prediction methods need well-predefined IVs, while it\u2019s an art rather than science to find valid IVs in many real-world scenes. Moreover, the predefined hand-made IVs could be weak or erroneous by violating the conditions of valid IVs. These thorny facts hinder the application of the IV-based counterfactual prediction methods. In this article, we propose a novel Automatic Instrumental Variable decomposition (AutoIV) algorithm to automatically generate representations serving the role of IVs from observed variables (IV candidates). Specifically, we let the learned IV representations satisfy the relevance condition with the treatment and exclusion condition with the outcome via mutual information maximization and minimization constraints, respectively. We also learn confounder representations by encouraging them to be relevant to both the treatment and the outcome. The IV and confounder representations compete for the information with their constraints in an adversarial game, which allows us to get valid IV representations for IV-based counterfactual prediction. Extensive experiments demonstrate that our method generates valid IV representations for accurate IV-based counterfactual prediction."
                },
                {
                    "title": "Demystifying Statistical Learning Based on Efficient Influence Functions",
                    "abstract": "Abstract Evaluation of treatment effects and more general estimands is typically achieved via parametric modeling, which is unsatisfactory since model misspecification is likely. Data-adaptive model building (e.g., statistical/machine learning) is commonly employed to reduce the risk of misspecification. Na\u00efve use of such methods, however, delivers estimators whose bias may shrink too slowly with sample size for inferential methods to perform well, including those based on the bootstrap. Bias arises because standard data-adaptive methods are tuned toward minimal prediction error as opposed to, for example, minimal MSE in the estimator. This may cause excess variability that is difficult to acknowledge, due to the complexity of such strategies. Building on results from nonparametric statistics, targeted learning and debiased machine learning overcome these problems by constructing estimators using the estimand\u2019s efficient influence function under the nonparametric model. These increasingly popular methodologies typically assume that the efficient influence function is given, or that the reader is familiar with its derivation. In this article, we focus on derivation of the efficient influence function and explain how it may be used to construct statistical/machine-learning-based estimators. We discuss the requisite conditions for these estimators to perform well and use diverse examples to convey the broad applicability of the theory."
                },
                {
                    "title": "Statistical Inference with M-Estimators on Adaptively Collected Data",
                    "abstract": "Bandit algorithms are increasingly used in real-world sequential decision-making problems. Associated with this is an increased desire to be able to use the resulting datasets to answer scientific questions like: Did one type of ad lead to more purchases? In which contexts is a mobile health intervention effective? However, classical statistical approaches fail to provide valid confidence intervals when used with data collected with bandit algorithms. Alternative methods have recently been developed for simple models (e.g., comparison of means). Yet there is a lack of general methods for conducting statistical inference using more complex models on data collected with (contextual) bandit algorithms; for example, current methods cannot be used for valid inference on parameters in a logistic regression model for a binary reward. In this work, we develop theory justifying the use of M-estimators-which includes estimators based on empirical risk minimization as well as maximum likelihood-on data collected with adaptive algorithms, including (contextual) bandit algorithms. Specifically, we show that M-estimators, modified with particular adaptive weights, can be used to construct asymptotically valid confidence regions for a variety of inferential targets."
                },
                {
                    "title": "VarGrad: A Low-Variance Gradient Estimator for Variational Inference",
                    "abstract": "We analyse the properties of an unbiased gradient estimator of the ELBO for variational inference, based on the score function method with leave-one-out control variates. We show that this gradient estimator can be obtained using a new loss, defined as the variance of the log-ratio between the exact posterior and the variational approximation, which we call the $\\textit{log-variance loss}$. Under certain conditions, the gradient of the log-variance loss equals the gradient of the (negative) ELBO. We show theoretically that this gradient estimator, which we call $\\textit{VarGrad}$ due to its connection to the log-variance loss, exhibits lower variance than the score function method in certain settings, and that the leave-one-out control variate coefficients are close to the optimal ones. We empirically demonstrate that VarGrad offers a favourable variance versus computation trade-off compared to other state-of-the-art estimators on a discrete VAE."
                },
                {
                    "title": "Learning Deep Features in Instrumental Variable Regression",
                    "abstract": "Instrumental variable (IV) regression is a standard strategy for learning causal relationships between confounded treatment and outcome variables from observational data by utilizing an instrumental variable, which affects the outcome only through the treatment. In classical IV regression, learning proceeds in two stages: stage 1 performs linear regression from the instrument to the treatment; and stage 2 performs linear regression from the treatment to the outcome, conditioned on the instrument. We propose a novel method, deep feature instrumental variable regression (DFIV), to address the case where relations between instruments, treatments, and outcomes may be nonlinear. In this case, deep neural nets are trained to define informative nonlinear features on the instruments and treatments. We propose an alternating training regime for these features to ensure good end-to-end performance when composing stages 1 and 2, thus obtaining highly flexible feature maps in a computationally efficient manner. DFIV outperforms recent state-of-the-art methods on challenging IV benchmarks, including settings involving high dimensional image data. DFIV also exhibits competitive performance in off-policy policy evaluation for reinforcement learning, which can be understood as an IV regression task."
                },
                {
                    "title": "Estimation of local treatment effects under the binary instrumental variable model",
                    "abstract": "\n Instrumental variables are widely used to deal with unmeasured confounding in observational studies and imperfect randomized controlled trials. In these studies, researchers often target the so-called local average treatment effect as it is identifiable under mild conditions. In this paper we consider estimation of the local average treatment effect under the binary instrumental variable model. We discuss the challenges of causal estimation with a binary outcome and show that, surprisingly, it can be more difficult than in the case with a continuous outcome. We propose novel modelling and estimation procedures that improve upon existing proposals in terms of model congeniality, interpretability, robustness and efficiency. Our approach is illustrated via simulation studies and a real data analysis."
                },
                {
                    "title": "Discriminative Jackknife: Quantifying Uncertainty in Deep Learning via Higher-Order Influence Functions",
                    "abstract": "Deep learning models achieve high predictive accuracy across a broad spectrum of tasks, but rigorously quantifying their predictive uncertainty remains challenging. Usable estimates of predictive uncertainty should (1) cover the true prediction targets with high probability, and (2) discriminate between high- and low-confidence prediction instances. Existing methods for uncertainty quantification are based predominantly on Bayesian neural networks; these may fall short of (1) and (2) -- i.e., Bayesian credible intervals do not guarantee frequentist coverage, and approximate posterior inference undermines discriminative accuracy. In this paper, we develop the discriminative jackknife (DJ), a frequentist procedure that utilizes influence functions of a model's loss functional to construct a jackknife (or leave-one-out) estimator of predictive confidence intervals. The DJ satisfies (1) and (2), is applicable to a wide range of deep learning models, is easy to implement, and can be applied in a post-hoc fashion without interfering with model training or compromising its accuracy. Experiments demonstrate that DJ performs competitively compared to existing Bayesian and non-Bayesian regression baselines."
                },
                {
                    "title": "Influence Functions in Deep Learning Are Fragile",
                    "abstract": "Influence functions approximate the effect of training samples in test-time predictions and have a wide variety of applications in machine learning interpretability and uncertainty estimation. A commonly-used (first-order) influence function can be implemented efficiently as a post-hoc method requiring access only to the gradients and Hessian of the model. For linear models, influence functions are well-defined due to the convexity of the underlying loss function and are generally accurate even across difficult settings where model changes are fairly large such as estimating group influences. Influence functions, however, are not well-understood in the context of deep learning with non-convex loss functions. In this paper, we provide a comprehensive and large-scale empirical study of successes and failures of influence functions in neural network models trained on datasets such as Iris, MNIST, CIFAR-10 and ImageNet. Through our extensive experiments, we show that the network architecture, its depth and width, as well as the extent of model parameterization and regularization techniques have strong effects in the accuracy of influence functions. In particular, we find that (i) influence estimates are fairly accurate for shallow networks, while for deeper networks the estimates are often erroneous; (ii) for certain network architectures and datasets, training with weight-decay regularization is important to get high-quality influence estimates; and (iii) the accuracy of influence estimates can vary significantly depending on the examined test points. These results suggest that in general influence functions in deep learning are fragile and call for developing improved influence estimation methods to mitigate these issues in non-convex setups."
                },
                {
                    "title": "Efficient Adaptive Experimental Design for Average Treatment Effect Estimation",
                    "abstract": "The goal of many scientific experiments including A/B testing is to estimate the average treatment effect (ATE), which is defined as the difference between the expected outcomes of two or more treatments. In this paper, we consider a situation where an experimenter can assign a treatment to research subjects sequentially. In adaptive experimental design, the experimenter is allowed to change the probability of assigning a treatment using past observations for estimating the ATE efficiently. However, with this approach, it is difficult to apply a standard statistical method to construct an estimator because the observations are not independent and identically distributed. We thus propose an algorithm for efficient experiments with estimators constructed from dependent samples. We also introduce a sequential testing framework using the proposed estimator. To justify our proposed approach, we provide finite and infinite sample analyses. Finally, we experimentally show that the proposed algorithm exhibits preferable performance."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Optimal Experimental Design for Staggered Rollouts",
                    "abstract": "In this paper, we study the design and analysis of experiments conducted on a set of units over multiple time periods in which the starting time of the treatment may vary by unit. The design problem involves selecting an initial treatment time for each unit in order to most precisely estimate both the instantaneous and cumulative effects of the treatment. We first consider nonadaptive experiments, in which all treatment assignment decisions are made prior to the start of the experiment. For this case, we show that the optimization problem is generally NP-hard, and we propose a near-optimal solution. Under this solution, the fraction entering treatment each period is initially low, then high, and finally low again. Next, we study an adaptive experimental design problem, in which both the decision to continue the experiment and treatment assignment decisions are updated after each period\u2019s data are collected. For the adaptive case, we propose a new algorithm, the precision-guided adaptive experiment algorithm, which addresses the challenges at both the design stage and the stage of estimating treatment effects, ensuring valid post-experiment inference, accounting for the adaptive nature of the design. Using realistic settings, we demonstrate that our proposed solutions can reduce the opportunity cost of the experiments by more than 50%, compared with static design benchmarks. This paper was accepted by George Shanthikumar, data science. Funding: S. Athey and G. Imbens were supported by the Office of Naval Research [Grant N00014-19-1-2468]. M. Bayati was supported by the National Science Foundation [Grant CMMI 1554140]. Supplemental Material: The e-companion and data are available at https://doi.org/10.1287/mnsc.2023.4928 ."
                },
                {
                    "title": "Optimizing Millions of Hyperparameters by Implicit Differentiation",
                    "abstract": "We propose an algorithm for inexpensive gradient-based hyperparameter optimization that combines the implicit function theorem (IFT) with efficient inverse Hessian approximations. We present results about the relationship between the IFT and differentiating through optimization, motivating our algorithm. We use the proposed approach to train modern network architectures with millions of weights and millions of hyper-parameters. For example, we learn a data-augmentation network - where every weight is a hyperparameter tuned for validation performance - outputting augmented training examples. Jointly tuning weights and hyperparameters with our approach is only a few times more costly in memory and compute than standard training."
                },
                {
                    "title": "Machine Learning Estimation of Heterogeneous Treatment Effects with Instruments",
                    "abstract": "We consider the estimation of heterogeneous treatment effects with arbitrary machine learning methods in the presence of unobserved confounders with the aid of a valid instrument. Such settings arise in A/B tests with an intent-to-treat structure, where the experimenter randomizes over which user will receive a recommendation to take an action, and we are interested in the effect of the downstream action. We develop a statistical learning approach to the estimation of heterogeneous effects, reducing the problem to the minimization of an appropriate loss function that depends on a set of auxiliary models (each corresponding to a separate prediction task). The reduction enables the use of all recent algorithmic advances (e.g. neural nets, forests). We show that the estimated effect model is robust to estimation errors in the auxiliary models, by showing that the loss satisfies a Neyman orthogonality criterion. Our approach can be used to estimate projections of the true effect model on simpler hypothesis spaces. When these spaces are parametric, then the parameter estimates are asymptotically normal, which enables construction of confidence sets. We applied our method to estimate the effect of membership on downstream webpage engagement on TripAdvisor, using as an instrument an intent-to-treat A/B test among 4 million TripAdvisor users, where some users received an easier membership sign-up process. We also validate our method on synthetic data and on public datasets for the effects of schooling on income."
                },
                {
                    "title": "Validating Causal Inference Models via Influence Functions",
                    "abstract": "The problem of estimating causal effects of treatments from observational data falls beyond the realm of supervised learning \u2014 because counter-factual data is inaccessible, we can never observe the true causal effects. In the absence of \u201csuper-vision\u201d, how can we evaluate the performance of causal inference methods? In this paper, we use in\ufb02uence functions \u2014 the functional derivatives of a loss function \u2014 to develop a model validation procedure that estimates the estimation error of causal inference methods. Our procedure utilizes a Taylor-like expansion to approximate the loss function of a method on a given dataset in terms of the in\ufb02uence functions of its loss on a \u201csynthesized\u201d, proximal dataset with known causal effects. Under minimal regularity assumptions, we show that our procedure is \u221a n -consistent and ef\ufb01cient. Experiments on 77 benchmark datasets show that using our procedure, we can accurately predict the comparative performances of state-of-the-art causal inference methods applied to a given observational study."
                },
                {
                    "title": "Optimistic Policy Optimization via Multiple Importance Sampling",
                    "abstract": "Policy Search (PS) is an effective approach to Reinforcement Learning (RL) for solving control tasks with continuous state-action spaces. In this paper, we address the exploration-exploitation trade-off in PS by proposing an approach based on Optimism in the Face of Uncertainty. We cast the PS problem as a suitable Multi Armed Bandit (MAB) problem, defined over the policy parameter space, and we propose a class of algorithms that effectively exploit the problem structure, by leveraging Multiple Importance Sampling to perform an off-policy estimation of the expected return. We show that the regret of the proposed approach is bounded by r Op ? T q for both discrete and continuous parameter spaces. Finally, we evaluate our algorithms on tasks of varying difficulty, comparing them with existing MAB and RL algorithms."
                },
                {
                    "title": "An Exponential Efron-Stein Inequality for Lq Stable Learning Rules",
                    "abstract": "There is accumulating evidence in the literature that stability of learning algorithms is a key characteristic that permits a learning algorithm to generalize. Despite various insightful results in this direction, there seems to be an overlooked dichotomy in the type of stability-based generalization bounds we have in the literature. On one hand, the literature seems to suggest that exponential generalization bounds for the estimated risk, which are optimal, can be only obtained through stringent, distribution independent and computationally intractable notions of stability such as uniform stability. On the other hand, it seems that weaker notions of stability such as hypothesis stability, although it is distribution dependent and more amenable to computation, can only yield polynomial generalization bounds for the estimated risk, which are suboptimal. \nIn this paper, we address the gap between these two regimes of results. In particular, the main question we address here is whether it is possible to derive exponential generalization bounds for the estimated risk using a notion of stability that is computationally tractable and distribution dependent, but weaker than uniform stability. Using recent advances in concentration inequalities, and using a notion of stability that is weaker than uniform stability but distribution dependent and amenable to computation, we derive an exponential tail bound for the concentration of the estimated risk of a hypothesis returned by a general learning rule, where the estimated risk is expressed in terms of either the resubstitution estimate (empirical error), or the deleted (or, leave-one-out) estimate. As an illustration, we derive exponential tail bounds for ridge regression with unbounded responses -- a setting where uniform stability results of Bousquet and Elisseeff (2002) are not applicable."
                },
                {
                    "title": "Visually Communicating and Teaching Intuition for Influence Functions",
                    "abstract": "Abstract Estimators based on influence functions (IFs) have been shown to be effective in many settings, especially when combined with machine learning techniques. By focusing on estimating a specific target of interest (e.g., the average effect of a treatment), rather than on estimating the full underlying data generating distribution, IF-based estimators are often able to achieve asymptotically optimal mean-squared error. Still, many researchers find IF-based estimators to be opaque or overly technical, which makes their use less prevalent and their benefits less available. To help foster understanding and trust in IF-based estimators, we present tangible, visual illustrations of when and how IF-based estimators can outperform standard \u201cplug-in\u201d estimators. The figures we show are based on connections between IFs, gradients, linear approximations, and Newton\u2013Raphson."
                },
                {
                    "title": "DNN: A Two-Scale Distributional Tale of Heterogeneous Treatment Effect Inference",
                    "abstract": "Heterogeneous treatment effects are the center of gravity in many modern causal inference applications. In this paper, we investigate the estimation and inference of heterogeneous treatment effects with precision in a general non-parametric setting. To this end, we enhance the classical $k$-nearest neighbor method with a simple algorithm, extend it to a distributional setting, and suggest the two-scale distributional nearest neighbors (DNN) estimator with reduced finite-sample bias. Our recipe is first to subsample the data and average the 1-nearest neighbor estimators from each subsample. With appropriately chosen subsampling scale, the resulting DNN estimator is proved to be asymptotically unbiased and normal under mild regularity conditions. We then proceed with combining DNN estimators with different subsampling scales to further reduce bias. Our theoretical results on the advantages of the new two-scale DNN framework are well supported by several Monte Carlo simulations. The newly suggested method is also applied to a real-life data set to study the heterogeneity of treatment effects of smoking on children's birth weights across mothers' ages."
                },
                {
                    "title": "Variance Reduction for Policy Gradient with Action-Dependent Factorized Baselines",
                    "abstract": "Policy gradient methods have enjoyed great success in deep reinforcement learning but suffer from high variance of gradient estimates. The high variance problem is particularly exasperated in problems with long horizons or high-dimensional action spaces. To mitigate this issue, we derive a bias-free action-dependent baseline for variance reduction which fully exploits the structural form of the stochastic policy itself and does not make any additional assumptions about the MDP. We demonstrate and quantify the benefit of the action-dependent baseline through both theoretical analysis as well as numerical results, including an analysis of the suboptimality of the optimal state-dependent baseline. The result is a computationally efficient policy gradient algorithm, which scales to high-dimensional control problems, as demonstrated by a synthetic 2000-dimensional target matching task. Our experimental results indicate that action-dependent baselines allow for faster learning on standard reinforcement learning benchmarks and high-dimensional hand manipulation and synthetic tasks. Finally, we show that the general idea of including additional information in baselines for improved variance reduction can be extended to partially observed and multi-agent tasks."
                },
                {
                    "title": "Action-depedent Control Variates for Policy Optimization via Stein's Identity",
                    "abstract": "Policy gradient methods have achieved remarkable successes in solving challenging reinforcement learning problems. However, it still often suffers from the large variance issue on policy gradient estimation, which leads to poor sample efficiency during training. In this work, we propose a control variate method to effectively reduce variance for policy gradient methods. Motivated by the Stein's identity, our method extends the previous control variate methods used in REINFORCE and advantage actor-critic by introducing more general action-dependent baseline functions. Empirical studies show that our method significantly improves the sample efficiency of the state-of-the-art policy gradient approaches."
                },
                {
                    "title": "Deep IV: A Flexible Approach for Counterfactual Prediction",
                    "abstract": "Counterfactual prediction requires understanding causal relationships between so-called treatment and outcome variables. This paper provides a recipe for augmenting deep learning methods to accurately characterize such relationships in the presence of instrument variables (IVs)\u2014sources of treatment randomization that are conditionally independent from the outcomes. Our IV specification resolves into two prediction tasks that can be solved with deep neural nets: a first-stage network for treatment prediction and a second-stage network whose loss function involves integration over the conditional treatment distribution. This Deep IV framework allows us to take advantage of off-the-shelf supervised learning techniques to estimate causal effects by adapting the loss function. Experiments show that it outperforms existing machine learning approaches."
                },
                {
                    "title": "Instrument-Armed Bandits",
                    "abstract": "We extend the classic multi-armed bandit (MAB) model to the setting of noncompliance, where the arm pull is a mere instrument and the treatment applied may differ from it, which gives rise to the instrument-armed bandit (IAB) problem. The IAB setting is relevant whenever the experimental units are human since free will, ethics, and the law may prohibit unrestricted or forced application of treatment. In particular, the setting is relevant in bandit models of dynamic clinical trials and other controlled trials on human interventions. Nonetheless, the setting has not been fully investigate in the bandit literature. We show that there are various and divergent notions of regret in this setting, all of which coincide only in the classic MAB setting. We characterize the behavior of these regrets and analyze standard MAB algorithms. We argue for a particular kind of regret that captures the causal effect of treatments but show that standard MAB algorithms cannot achieve sublinear control on this regret. Instead, we develop new algorithms for the IAB problem, prove new regret bounds for them, and compare them to standard MAB algorithms in numerical examples."
                },
                {
                    "title": "Understanding Black-box Predictions via Influence Functions",
                    "abstract": "How can we explain the predictions of a black-box model? In this paper, we use influence functions \u2014 a classic technique from robust statistics \u2014 to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. We show that even on non-convex and non-differentiable models where the theory breaks down, approximations to influence functions can still provide valuable information. On linear models and convolutional neural networks, we demonstrate that influence functions are useful for multiple purposes: understanding model behavior, debugging models, detecting dataset errors, and even creating visually-indistinguishable training-set attacks."
                },
                {
                    "title": "Policy Learning With Observational Data",
                    "abstract": "In many areas, practitioners seek to use observational data to learn a treatment assignment policy that satisfies application\u2010specific constraints, such as budget, fairness, simplicity, or other functional form constraints. For example, policies may be restricted to take the form of decision trees based on a limited set of easily observable individual characteristics. We propose a new approach to this problem motivated by the theory of semiparametrically efficient estimation. Our method can be used to optimize either binary treatments or infinitesimal nudges to continuous treatments, and can leverage observational data where causal effects are identified using a variety of strategies, including selection on observables and instrumental variables. Given a doubly robust estimator of the causal effect of assigning everyone to treatment, we develop an algorithm for choosing whom to treat, and establish strong guarantees for the asymptotic utilitarian regret of the resulting policy."
                },
                {
                    "title": "Stability revisited: new generalisation bounds for the Leave-one-Out",
                    "abstract": "The present paper provides a new generic strategy leading to non-asymptotic theoretical guarantees on the Leave-one-Out procedure applied to a broad class of learning algorithms. This strategy relies on two main ingredients: the new notion of $L^q$ stability, and the strong use of moment inequalities. $L^q$ stability extends the ongoing notion of hypothesis stability while remaining weaker than the uniform stability. It leads to new PAC exponential generalisation bounds for Leave-one-Out under mild assumptions. In the literature, such bounds are available only for uniform stable algorithms under boundedness for instance. Our generic strategy is applied to the Ridge regression algorithm as a first step."
                },
                {
                    "title": "Estimation and Inference of Heterogeneous Treatment Effects using Random Forests",
                    "abstract": "ABSTRACT Many scientific and engineering challenges\u2014ranging from personalized medicine to customized marketing recommendations\u2014require an understanding of treatment effect heterogeneity. In this article, we develop a nonparametric causal forest for estimating heterogeneous treatment effects that extends Breiman\u2019s widely used random forest algorithm. In the potential outcomes framework with unconfoundedness, we show that causal forests are pointwise consistent for the true treatment effect and have an asymptotically Gaussian and centered sampling distribution. We also discuss a practical method for constructing asymptotic confidence intervals for the true treatment effect that are centered at the causal forest estimates. Our theoretical results rely on a generic Gaussian theory for a large family of random forest algorithms. To our knowledge, this is the first set of results that allows any type of random forest, including classification and regression forests, to be used for provably valid statistical inference. In experiments, we find causal forests to be substantially more powerful than classical methods based on nearest-neighbor matching, especially in the presence of irrelevant covariates."
                },
                {
                    "title": "The influence function of semiparametric estimators",
                    "abstract": "There are many economic parameters that depend on nonparametric first steps. Examples include games, dynamic discrete choice, average exact consumer surplus, and treatment effects. Often estimators of these parameters are asymptotically equivalent to a sample average of an object referred to as the influence function. The influence function is useful in local policy analysis, in evaluating local sensitivity of estimators, and constructing debiased machine learning estimators. We show that the influence function is a Gateaux derivative with respect to a smooth deviation evaluated at a point mass. This result generalizes the classic Von Mises (1947) and Hampel (1974) calculation to estimators that depend on smooth nonparametric first steps. We give explicit influence functions for first steps that satisfy exogenous or endogenous orthogonality conditions. We use these results to generalize the omitted variable bias formula for regression to policy analysis for and sensitivity to structural changes. We apply this analysis and find no sensitivity to endogeneity of average equivalent variation estimates in a gasoline demand application."
                },
                {
                    "title": "Instrumental Variables: An Econometrician's Perspective",
                    "abstract": "I review recent work in the statistics literature on instrumental variables methods from an econometrics perspective. I discuss some of the older, economic, applications including supply and demand models and relate them to the recent applications in settings of randomized experiments with noncompliance. I discuss the assumptions underlying instrumental variables methods and in what settings these may be plausible. By providing context to the current applications a better understanding of the applicability of these methods may arise."
                },
                {
                    "title": "On Using Control Variates with Stochastic Approximation for Variational Bayes and its Connection to Stochastic Linear Regression",
                    "abstract": "Recently, we and several other authors have written about the possibilities of using stochastic approximation techniques for fitting variational approximations to intractable Bayesian posterior distributions. Naive implementations of stochastic approximation suffer from high variance in this setting. Several authors have therefore suggested using control variates to reduce this variance, while we have taken a different but analogous approach to reducing the variance which we call stochastic linear regression. In this note we take the former perspective and derive the ideal set of control variates for stochastic approximation variational Bayes under a certain set of assumptions. We then show that using these control variates is closely related to using the stochastic linear regression approximation technique we proposed earlier. A simple example shows that our method for constructing control variates leads to stochastic estimators with much lower variance compared to other approaches."
                },
                {
                    "title": "Instrumental Variable Estimators for Binary Outcomes",
                    "abstract": "Instrumental variables (IVs) can be used to construct estimators of exposure effects on the outcomes of studies affected by nonignorable selection of the exposure. Estimators that fail to adjust for the effects of nonignorable selection will be biased and inconsistent. Such situations commonly arise in observational studies, but are also a problem for randomized experiments affected by nonignorable noncompliance. In this article, we review IV estimators for studies in which the outcome is binary, and consider the links between different approaches developed in the statistics and econometrics literatures. The implicit assumptions made by each method are highlighted and compared within our framework. We illustrate our findings through the reanalysis of a randomized placebo-controlled trial, and highlight important directions for future work in this area."
                },
                {
                    "title": "Adaptive Experimental Design Using the Propensity Score",
                    "abstract": "Many social experiments are run in multiple waves or replicate earlier social experiments. In principle, the sampling design can be modified in later stages or replications to allow for more efficient estimation of causal effects. We consider the design of a two-stage experiment for estimating an average treatment effect when covariate information is available for experimental subjects. We use data from the first stage to choose a conditional treatment assignment rule for units in the second stage of the experiment. This amounts to choosing the propensity score, the conditional probability of treatment given covariates. We propose to select the propensity score to minimize the asymptotic variance bound for estimating the average treatment effect. Our procedure can be implemented simply using standard statistical software and has attractive large-sample properties."
                },
                {
                    "title": "Semiparametrically Efficient Estimation of Conditional Instrumental Variables Parameters",
                    "abstract": "In this paper, I propose a set of parameters designed to identify the slope of structural relationships based on a combination of conditioning on covariates and the use of an exogenous instrument. After giving structural interpretations to these parameters in the context of specific semiparametric models, I derive their efficient influence curves in a fully nonparametric context as well as under imposition of restrictions on the instrument. These influence curves give the semiparametric efficiency bounds for regular asymptotically linear estimators of the parameters and allow the construction of asymptotically efficient estimators. Monte Carlo experiments finally demonstrate the good finite sample performance of such estimators."
                },
                {
                    "title": "Better Late than Nothing: Some Comments on Deaton (2009) and Heckman and Urzua (2009)",
                    "abstract": "Two recent papers, Deaton (2009), and Heckman and Urzua (2009), argue against what they see as an excessive and inappropriate use of experimental and quasi-experimental methods in empirical work in economics in the last decade. They specifically question the increased use of instrumental variables and natural experiments in labor economics, and of randomized experiments in development economics. In these comments I will make the case that this move towards shoring up the internal validity of estimates, and towards clarifying the description of the population these estimates are relevant for, has been important and beneficial in increasing the credibility of empirical work in economics. I also address some other concerns raised by the Deaton and Heckman-Urzua papers."
                },
                {
                    "title": "Model Selection in Kernel Based Regression using the Influence Function",
                    "abstract": "Recent results about the robustness of kernel methods involve the analysis of influence functions. By definition the influence function is closely related to leave-one-out criteria. In statistical learning, the latter is often used to assess the generalization of a method. In statistics, the influence function is used in a similar way to analyze the statistical efficiency of a method. Links between both worlds are explored. The influence function is related to the first term of a Taylor expansion. Higher order influence functions are calculated. A recursive relation between these terms is found characterizing the full Taylor expansion. It is shown how to evaluate influence functions at a specific sample distribution to obtain an approximation of the leave-one-out error. A specific implementation is proposed using a L1 loss in the selection of the hyperparameters and a Huber loss in the estimation procedure. The parameter in the Huber loss controlling the degree of robustness is optimized as well. The resulting procedure gives good results, even when outliers are present in the data."
                },
                {
                    "title": "Higher order influence functions and minimax estimation of nonlinear functionals",
                    "abstract": "We present a theory of point and interval estimation for nonlinear functionals in parametric, semi-, and non-parametric models based on higher order influence functions (Robins (2004), Section 9; Li et al. (2004), Tchetgen et al. (2006), Robins et al. (2007)). Higher order influence functions are higher order U-statistics. Our theory extends the first order semiparametric theory of Bickel et al. (1993) and van der Vaart (1991) by incorporating the theory of higher order scores considered by Pfanzagl (1990), Small and McLeish (1994) and Lindsay and Waterman (1996). The theory reproduces many previous results, produces new non-$\\sqrt{n}$ results, and opens up the ability to perform optimal non-$\\sqrt{n}$ inference in complex high dimensional models. We present novel rate-optimal point and interval estimators for various functionals of central importance to biostatistics in settings in which estimation at the expected $\\sqrt{n}$ rate is not possible, owing to the curse of dimensionality. We also show that our higher order influence functions have a multi-robustness property that extends the double robustness property of first order influence functions described by Robins and Rotnitzky (2001) and van der Laan and Robins (2003)."
                },
                {
                    "title": "The Impact of Homework on Student Achievement",
                    "abstract": "Utilizing parametric and nonparametric techniques, we assess the role of a heretofore relatively unexplored \"input\" in the educational process, homework, on academic achievement. Our results indicate that homework is an important determinant of student test scores. Relative to more standard spending related measures, extra homework has a larger and more significant impact on test scores. However, the effects are not uniform across different subpopulations. Specifically, we find additional homework to be most effective for high and low achievers, which is further confirmed by stochastic dominance analysis. Moreover, the parametric estimates of the educational production function overstate the impact of schooling related inputs. In all estimates, the homework coefficient from the parametric model maps to the upper deciles of the nonparametric coefficient distribution and as a by-product the parametric model understates the percentage of students with negative responses to additional homework. Copyright \u00a9 2008 The Author(s). Journal compilation \u00a9 Royal Economic Society 2008"
                },
                {
                    "title": "An experimental design for the development of adaptive treatment strategies",
                    "abstract": "In adaptive treatment strategies, the treatment level and type is repeatedly adjusted according to ongoing individual response. Since past treatment may have delayed effects, the development of these treatment strategies is challenging. This paper advocates the use of sequential multiple assignment randomized trials in the development of adaptive treatment strategies. Both a simple ad hoc method for ascertaining sample sizes and simple analysis methods are provided. Copyright \u00a9 2004 John Wiley & Sons, Ltd."
                },
                {
                    "title": "Instrumental variable estimation of nonparametric models",
                    "abstract": "In econometrics there are many occasions where knowledge of the structural relationship among dependent variables is required to answer questions of interest. This paper gives identification and estimation results for nonparametric conditional moment restrictions. We characterize identification of structural functions as completeness of certain conditional distributions, and give sufficient identification conditions for exponential families and discrete variables. We also give a consistent, nonparametric estimator of the structural function. The estimator is nonparametric two-stage least squares based on series approximation, which overcomes an ill-posed inverse problem by placing bounds on integrals of higher-order derivatives. Copyright The Econometric Society 2003."
                },
                {
                    "title": "Empirical supremum rejection sampling",
                    "abstract": "Rejection sampling thins out samples from a candidate density from which it is easy to simulate, to obtain samples from a more awkward target density. A prerequisite is knowledge of the finite supremum of the ratio of the target and candidate densities. This severely restricts application of the method because it can be difficult to calculate the supremum. We use theoretical argument and numerical work to show that a practically perfect sample may be obtained by replacing the exact supremum with the maximum obtained from simulated candidates. We also provide diagnostics for failure of the method caused by a bad choice of candidate distribution. The implication is that essentially no theoretical work is required to apply rejection sampling in many practical cases. Copyright Biometrika Trust 2002, Oxford University Press."
                },
                {
                    "title": "Subsampling",
                    "abstract": "level or as a useful reference book for the applied researcher or practitioner. Speaking as a practitioner, I found many parts of the book to be extremely useful. The main aims of the book are to give a brief coverage of the major MCMC methods for sampling from posterior distributions and then focus extensively on methods for computing posterior quantities of interest. Topics such as marginal densities, ratios of normalizing constants, constrained parameter estimation, highest posterior density intervals, and methods for model comparison and selection are given thorough treatment in separate chapters of the book. The book begins by describing several datasets that are used throughout the book as examples. These cover a variety of types of data and are useful support for a number of diverse types of models throughout the book. A number of references point the interested reader to standard analyses of these data that can be used for comparison with the Bayesian techniques described later in the text. Chapter 2 gives an overview of the main techniques used in MCMC sampling. The \u008e rst section starts out with the Gibbs sampler. Detailed examples of its use for a bivariate normal model and then for a constrained regression problem in which the model coef\u008e cients are constrained to be increasing are thorough and comprehensive. Equally thorough discussions of the Metropolis\u2013 Hastings algorithm, the hit-and-run algorithm, and the multiple-try Metropolis algorithm follow. A nice feature of this section is the use of different algorithms on the same set of example data. The chapter continues covering methods used to improve the ef\u008e ciency of MCMC sampling. The section on grouping, collapsing, and reparameterization gives several different algorithms for dealing with ordinal response models. I found this section of the book to be the most useful because of the clear exposition of the algorithms and the detailed description of the steps taken in the MCMC sampling. The chapter goes on to discuss acceleration, a random direction importance sampler that is a step toward a \u201cblack box\u201d sampling algorithm, and \u008e nally, some material on convergence diagnostics. Chapter 3 begins the main part of the text, covering basic Monte Carlo methods for estimating posterior quantities. Chapter 4 covers methods for estimating posterior marginal densities and includes several detailed proofs. Chapter 5 gives an overview of methods used for estimating ratios of normalizing constants. Importance sampling, path sampling, bridge sampling, and variations of each are all discussed. Extensions to problems of differing dimensions are also covered. A detailed application of these techniques to choose between various link functions in a generalized linear model is presented. Chapter 6 extends the concepts explored in Chapter 5 to problems that involve constraints on parameters. This chapter gives a brief overview of the bene\u008e ts of calculating normalizing constants for constrained parameter problems and then launches into two detailed examples. Chapter 7 proceeds similarly covering calculation of Bayesian credible and HPD intervals. The remaining chapters treat a variety of special topics in MCMC modeling. Chapter 8 covers approaches for comparing nonnested models, Chapter 9 gives an extensive treatment of Bayesian variable selection techniques, and Chapter 10 gives a sampling of other topics. The book ends with over 280 references. Every chapter ends with several exercises. Unfortunately, many of them look as if they would take an entire semester to complete. Typically they involve constructing several alternative algorithms for analyzing the data and comparing them. Given that excellent programming pro\u008e ciency is necessary, not to mention the time required to run the analyses an adequate number of iterations, it seems a student would have to devote full time to this single course to have any hope of completing many of the exercises. The book \u008e lls an important niche by pulling together a wide variety of material that is usually only available in diverse journals and technical reports. The topics are given a thorough mathematical treatment and the detailed examples make it possible for someone without rigorous mathematical training to implement many of the described algorithms. The step-by-step logic behind each method and the wide variety of topics covered makes this a valuable reference book for anyone who regularly works with Bayesian models."
                },
                {
                    "title": "Eligibility Traces for Off-Policy Policy Evaluation",
                    "abstract": "Eligibility traces have been shown to speed reinforcement learning, to make it more robust to hidden states, and to provide a link between Monte Carlo and temporal-difference methods. Here we generalize eligibility traces to off-policy learning, in which one learns about a policy different from the policy that generates the data. Off-policy methods can greatly multiply learning, as many policies can be learned about from the same data stream, and have been identified as particularly useful for learning about subgoals and temporally extended macro-actions. In this paper we consider the off-policy version of the policy evaluation problem, for which only one eligibility trace algorithm is known, a Monte Carlo method. We analyze and compare this and four new eligibility trace algorithms, emphasizing their relationships to the classical statistical technique known as importance sampling. Our main results are 1) to establish the consistency and bias properties of the new methods and 2) to empirically rank the new methods, showing improvement over one-step and Monte Carlo methods. Our results are restricted to model-free, table-lookup methods and to offline updating (at the end of each episode) although several of the algorithms could be applied more generally."
                },
                {
                    "title": "On Subsampling Estimators with Unknown Rate of Convergence",
                    "abstract": "Abstract Politis and Romano have put forth a general subsampling methodology for the construction of large-sample confidence regions for a general unknown parameter \u03b8 associated with the probability distribution generating the stationary sequence X 1,\u2026,X n . The subsampling methodology hinges on approximating the large-sample distribution of a statistic T n = T n (X 1,\u2026, X n ) that is consistent for \u03b8 at some known rate \u03c4 n . Although subsampling has been shown to yield confidence regions for \u03b8 of asymptotically correct coverage under very weak assumptions, the applicability of the methodology as it has been presented so far is limited if the rate of convergence \u03c4 n happens to be unknown or intractable in a particular setting. In this article we show how it is possible to circumvent this limitation by (a) using the subsampling methodology to derive a consistent estimator of the rate \u03c4 n , and (b) using the estimated rate to construct asymptotically correct confidence regions for \u03b8 based on subsampling."
                },
                {
                    "title": "Algorithmic Stability and Sanity-Check Bounds for Leave-One-Out Cross-Validation",
                    "abstract": "In this article we prove sanity-check bounds for the error of the leave-oneout cross-validation estimate of the generalization error: that is, bounds showing that the worst-case error of this estimate is not much worse than that of the training error estimate. The name sanity check refers to the fact that although we often expect the leave-one-out estimate to perform considerably better than the training error estimate, we are here only seeking assurance that its performance will not be considerably worse. Perhaps surprisingly, such assurance has been given only for limited cases in the prior literature on cross-validation. Any nontrivial bound on the error of leave-one-out must rely on some notion of algorithmic stability. Previous bounds relied on the rather strong notion of hypothesis stability, whose application was primarily limited to nearest-neighbor and other local algorithms. Here we introduce the new and weaker notion of error stability and apply it to obtain sanity-check bounds for leave-one-out for other classes of learning algorithms, including training error minimization procedures and Bayesian algorithms. We also provide lower bounds demonstrating the necessity of some form of error stability for proving bounds on the error of the leave-one-out estimate, and the fact that for training error minimization algorithms, in the worst case such bounds must still depend on the Vapnik-Chervonenkis dimension of the hypothesis class."
                },
                {
                    "title": "Bayesian Experimental Design: A Review",
                    "abstract": "This paper reviews the literature on Bayesian experimental design, both for linear and nonlinear models. A uniied view of the topic is presented by putting experimental design in a decision theoretic framework. This framework justiies many optimality criteria, and opens new possibilities. Various design criteria become part of a single, coherent approach."
                },
                {
                    "title": "Time Series Analysis",
                    "abstract": "A ordered sequence of events or observations having a time component is called as a time series. Some good examples of time series are daily opening and closing stock prices, daily humidity, temperature, pressure, annual gross domestic product (GDP) of a country and so on."
                },
                {
                    "title": "Fast Exact Multiplication by the Hessian",
                    "abstract": "Just storing the Hessian H (the matrix of second derivatives 2E/wiwj of the error E with respect to each pair of weights) of a large neural network is difficult. Since a common use of a large matrix like H is to compute its product with various vectors, we derive a technique that directly calculates Hv, where v is an arbitrary vector. To calculate Hv, we first define a differential operator Rv{f(w)} = (/r)f(w rv)|r=0, note that Rv{w} = Hv and Rv{w} = v, and then apply Rv{} to the equations used to compute w. The result is an exact and numerically stable procedure for computing Hv, which takes about as much computation, and is about as local, as a gradient evaluation. We then apply the technique to a one pass gradient calculation algorithm (backpropagation), a relaxation gradient calculation algorithm (recurrent backpropagation), and two stochastic gradient calculation algorithms (Boltzmann machines and weight perturbation). Finally, we show that this technique can be used at the heart of many iterative techniques for computing various properties of H, obviating any need to calculate the full Hessian."
                },
                {
                    "title": "Identification of Causal Effects Using Instrumental Variables",
                    "abstract": "Abstract We outline a framework for causal inference in settings where assignment to a binary treatment is ignorable, but compliance with the assignment is not perfect so that the receipt of treatment is nonignorable. To address the problems associated with comparing subjects by the ignorable assignment\u2014an \u201cintention-to-treat analysis\u201d\u2014we make use of instrumental variables, which have long been used by economists in the context of regression models with constant treatment effects. We show that the instrumental variables (IV) estimand can be embedded within the Rubin Causal Model (RCM) and that under some simple and easily interpretable assumptions, the IV estimand is the average causal effect for a subgroup of units, the compliers. Without these assumptions, the IV estimand is simply the ratio of intention-to-treat causal estimands with no interpretation as an average causal effect. The advantages of embedding the IV approach in the RCM are that it clarifies the nature of critical assumptions needed for a..."
                },
                {
                    "title": "Bootstrapping the mean integrated squared error",
                    "abstract": "A smooth bootstrap method is used to find an estimator of the mean integrated squared error in density estimation. This provides a natural bootstrap selector for the bandwidth. Some rates of convergence and limit distributions are presented for this new selector as well as for some modification of it."
                },
                {
                    "title": "Causality",
                    "abstract": "IT is a rather depressing task for a Catholic to write a technical philosophical essay, for it is improbable that the philosophers whom he criticises wi l l read it . F r Hawkins therefore deserves our thanks in a special w a y ; such work as this is necessary and its effects, though slow, are sure. There are a number of excellences about the book. I t reveals a wide knowledge of contem\u00ad porary English thought on the subject ; an index would have shown many references to Mind, to the Proceedings of the Aristotelian Society and to works of modern logic (the well-arranged table of contents gives some idea of the book's scope in this respect). Recent attempts to eliminate the notion of causality are very carefully repulsed. The writing is almost invariably attractive, and there are no slips or misprints obscuring the argument, which must be something like a record for Catholic philosophical books in recent years. These last virtues unfortunately are not shared by the only review (to the writer's present knowledge) which has had anything in particular to say about F r Hawkins's book, that in the Tablet (October 23rd, 1937). I t wi l l be of advantage to consider this article, for the criticisms which it em\u00ad bodies deserve to be made clear, and it wi l l therefore be quoted later in some detail. T h e main criticism that we have to offer here affects the policy which F r Hawkins has adopted. He has written a small book (as someone said of Kant's Critique, it would have been quicker to read had it been longer) and by far the greater part is taken up by the historical setting of the question and"
                },
                {
                    "title": "Adaptive Experimentation at Scale: Bayesian Algorithms for Flexible Batches",
                    "abstract": "Standard bandit algorithms that assume continual reallocation of measurement effort are challenging to implement due to delayed feedback and infrastructural/organizational difficulties. Motivated by practical instances involving a handful of reallocation epochs in which outcomes are measured in batches, we develop a new adaptive experimentation framework that can flexibly handle any batch size. Our main observation is that normal approximations, which are universal in statistical inference, can also guide the design of scalable adaptive designs. By deriving an asymptotic sequential experiment, we formulate a dynamic program that can leverage prior information on average rewards. We propose a simple iterative planning method, Residual Horizon Optimization, which selects sampling allocations by optimizing a planning objective with stochastic gradient descent. Our method significantly improves statistical power over standard methods, even when compared to Bayesian bandit algorithms (e.g., Thompson sampling) that require full distributional knowledge of individual rewards. Overall, we expand the scope of adaptive experimentation to settings that are difficult for standard methods, involving a small number of reallocation epochs, low signal-to-noise ratio, and unknown reward distributions."
                },
                {
                    "title": "Adaptively Identifying Patient Populations With Treatment Benefit in Clinical Trials",
                    "abstract": "We study the problem of adaptively identifying patient subpopulations that bene\ufb01t from a given treatment during a con\ufb01rmatory clinical trial. This type of adaptive clinical trial, often referred to as adaptive enrichment design , has been thoroughly studied in biostatistics with a focus on a limited number of subgroups (typically two) which make up (sub)populations, and a small number of interim analysis points. In this paper, we aim to relax classical restrictions on such designs and investigate how to incorporate ideas from the recent machine learning literature on adaptive and online experimentation to make trials more \ufb02exible and e\ufb03cient. We \ufb01nd that the unique characteristics of the subpopulation selection problem \u2013 most importantly that (i) one is usually interested in \ufb01nding subpopulations with any treatment bene\ufb01t (and not necessarily the single subgroup with largest e\ufb00ect) given a limited budget and that (ii) e\ufb00ectiveness only has to be demonstrated across the subpopulation on average \u2013 give rise to interesting challenges and new desiderata when designing algorithmic solutions. Building on these \ufb01ndings, we propose AdaGGI and AdaGCPI, two meta-algorithms for subpopulation construction, which focus on identifying good subgroups and good composite subpopulations, respectively. We empirically investigate their performance across a range of simulation scenarios and derive insights into their (dis)advantages across di\ufb00erent settings."
                },
                {
                    "title": "Policy Choice and Best Arm Identification: Comments on \"Adaptive Treatment Assignment in Experiments for Policy Choice\"",
                    "abstract": "Adaptive experimental design for efficient decision-making is an important problem in economics. The purpose of this paper is to connect the \u201cpolicy choice\u201d problem, proposed in Kasy and Sautmann (2021) as an instance of adaptive experimental design, to the frontiers of the bandit literature in machine learning. We discuss how the policy choice problem can be framed in a way such that it is identical to what is called the \u201cbest arm identification\u201d (BAI) problem. By connecting the literature, we identify that the asymptotic optimality of policy choice algorithms tackled in Kasy and Sautmann (2021) is a long-standing open question in the literature. While Kasy and Sautmann (2021) presents an interesting and important empirical study, unfortunately, this connection highlights several major issues with the theoretical results. In particular, we show that Theorem 1 in Kasy and Sautmann (2021) is false. We find that the proofs of statements (1) and (2) of Theorem 1 are incorrect. Although the statements themselves may be true, they are non-trivial to fix. Statement (3), and its proof, on the other hand, is false, which we show by utilizing existing theoretical results in the bandit literature. As this question is critically important, garnering much interest in the last decade within the bandit community, we provide a review of recent developments in the BAI literature. We hope this serves to highlight the relevance to economic problems and stimulate methodological and theoretical developments in the econometric community."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively utilize higher-order influence functions to improve gradient estimation in machine learning models without the need for retraining?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem has significant implications for the research community as it addresses the computational inefficiencies associated with traditional gradient estimation methods in machine learning. By leveraging higher-order influence functions, researchers can achieve more accurate gradient estimates with reduced variance, leading to improved model performance and stability. This advancement could pave the way for more efficient training algorithms, enabling the application of complex models in resource-constrained environments. Furthermore, it could enhance our understanding of model behavior and robustness, fostering further research into causal inference and uncertainty quantification in machine learning.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the complexities of accurately approximating higher-order influence functions and their interactions with model parameters. Naive approaches may fail due to the intricate nature of the mean-square-error (MSE) landscape, which can be highly non-linear and sensitive to local optima. Additionally, the computational burden of estimating these influence functions increases with model complexity and dataset size, making it difficult to achieve reliable estimates without extensive retraining. Overcoming these technical obstacles requires innovative methodologies that can balance accuracy and computational efficiency.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on first-order influence functions, neglecting the potential benefits of higher-order approximations. Limitations in computational resources and the complexity of deriving and implementing these higher-order functions have hindered progress. Additionally, existing solutions may not adequately address the stability and reliability of gradient estimates in the presence of model mis-specification or local optima. Our approach differs by providing a systematic framework for deriving and utilizing higher-order influence functions, thus improving upon prior work by ensuring robustness and reducing computational costs.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves deriving a K-th order expansion of influence functions to create a gradient estimator that avoids retraining. We will utilize a dataset from the TripAdvisor domain, applying neural network-based estimators to evaluate the performance of our approach. The key metric for assessment will be the mean-square-error (MSE) of the gradient estimates. We expect our method to yield significant variance reduction while maintaining a bias of the order n^(-K), thus providing reliable gradient estimates even in challenging optimization scenarios. The anticipated outcomes include improved training efficiency"
            }
        },
        "author_data": {
            "371e54a6-f63e-444a-90bb-85e0282eb7af": {
                "pk": "371e54a6-f63e-444a-90bb-85e0282eb7af",
                "name": "Yash Chandak",
                "collaborators": [
                    "E. Brunskill",
                    "P. Thomas",
                    "Allen Nie",
                    "Yannis Flet-Berliac",
                    "Shiv Shankar",
                    "M. G. Azar",
                    "B. C. Silva",
                    "Emma Brunskil",
                    "Nathan Grinsztajn",
                    "Florian Strub",
                    "Eugene Choi",
                    "Chris Cremer",
                    "Arash Ahmadian",
                    "O. Pietquin",
                    "Matthieu Geist",
                    "S. Thakoor",
                    "Z. Guo",
                    "Yunhao Tang",
                    "R. Munos",
                    "Will Dabney",
                    "Dhawal Gupta",
                    "Scott M. Jordan",
                    "Martha White",
                    "Christina J. Yuan",
                    "Anirudhan Badrinath",
                    "Miroslav Suzara",
                    "Malika Ali",
                    "Juliette Woodrow",
                    "Matt Peng",
                    "Mehran Sahami",
                    "Chris Piech",
                    "Hyunji Alex Nam",
                    "Ritwik Sinha",
                    "Saayan Mitra",
                    "M. Fiterau",
                    "Bill Wu",
                    "Amelia Leon",
                    "V. Gandikota",
                    "A. Mazumdar",
                    "Jonathan Lee",
                    "Annie Xie",
                    "Aldo Pacchiano",
                    "Chelsea Finn",
                    "Ofir Nachum",
                    "Diana Borsa",
                    "Philip S. Thomas",
                    "Nathaniel D. Bastian",
                    "Vincent Liu",
                    "James E. Kostas",
                    "Georgios Theocharous",
                    "Cheril Shah",
                    "Manan Suri",
                    "Tong Mu",
                    "Tatsunori B. Hashimoto",
                    "Ali T. Habiboglu",
                    "Pravin Khanal",
                    "B. Zhou",
                    "Carter Eckel",
                    "Jacob Cutshall Kennedy Warrilow",
                    "J. O'Brien",
                    "J. Schaibley",
                    "B. LeRoy",
                    "Wei-Gang Wang",
                    "Pierre H. Richemond",
                    "B. '. Pires",
                    "Mark Rowland",
                    "Charline Le Lan",
                    "Clare Lyle",
                    "Andr'as Gyorgy",
                    "Bilal Piot",
                    "Daniele Calandriello",
                    "M. Va\u013ako",
                    "Priyesh Vijayan",
                    "Mitesh M. Khapra",
                    "Srinivas Parthasarathy",
                    "B. Ravindran"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Offline Policy Evaluation",
                    "Large Language Models",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "OPERA: Automatic Offline Policy Evaluation with Re-weighted Aggregates of Multiple Estimators",
                        "abstract": "Offline policy evaluation (OPE) allows us to evaluate and estimate a new sequential decision-making policy's performance by leveraging historical interaction data collected from other policies. Evaluating a new policy online without a confident estimate of its performance can lead to costly, unsafe, or hazardous outcomes, especially in education and healthcare. Several OPE estimators have been proposed in the last decade, many of which have hyperparameters and require training. Unfortunately, choosing the best OPE algorithm for each task and domain is still unclear. In this paper, we propose a new algorithm that adaptively blends a set of OPE estimators given a dataset without relying on an explicit selection using a statistical procedure. We prove that our estimator is consistent and satisfies several desirable properties for policy evaluation. Additionally, we demonstrate that when compared to alternative approaches, our estimator can be used to select higher-performing policies in healthcare and robotics. Our work contributes to improving ease of use for a general-purpose, estimator-agnostic, off-policy evaluation framework for offline RL."
                    },
                    {
                        "title": "The GPT Surprise: Offering Large Language Model Chat in a Massive Coding Class Reduced Engagement but Increased Adopters Exam Performances",
                        "abstract": "Large language models (LLMs) are quickly being adopted in a wide range of learning experiences, especially via ubiquitous and broadly accessible chat interfaces like ChatGPT and Copilot. This type of interface is readily available to students and teachers around the world, yet relatively little research has been done to assess the impact of such generic tools on student learning. Coding education is an interesting test case, both because LLMs have strong performance on coding tasks, and because LLM-powered support tools are rapidly becoming part of the workflow of professional software engineers. To help understand the impact of generic LLM use on coding education, we conducted a large-scale randomized control trial with 5,831 students from 146 countries in an online coding class in which we provided some students with access to a chat interface with GPT-4. We estimate positive benefits on exam performance for adopters, the students who used the tool, but over all students, the advertisement of GPT-4 led to a significant average decrease in exam participation. We observe similar decreases in other forms of course engagement. However, this decrease is modulated by the student's country of origin. Offering access to LLMs to students from low human development index countries increased their exam participation rate on average. Our results suggest there may be promising benefits to using LLMs in an introductory coding class, but also potential harms for engagement, which makes their longer term impact on student success unclear. Our work highlights the need for additional investigations to help understand the potential impact of future adoption and integration of LLMs into classrooms."
                    },
                    {
                        "title": "Short-Long Policy Evaluation with Novel Actions",
                        "abstract": "From incorporating LLMs in education, to identifying new drugs and improving ways to charge batteries, innovators constantly try new strategies in search of better long-term outcomes for students, patients and consumers. One major bottleneck in this innovation cycle is the amount of time it takes to observe the downstream effects of a decision policy that incorporates new interventions. The key question is whether we can quickly evaluate long-term outcomes of a new decision policy without making long-term observations. Organizations often have access to prior data about past decision policies and their outcomes, evaluated over the full horizon of interest. Motivated by this, we introduce a new setting for short-long policy evaluation for sequential decision making tasks. Our proposed methods significantly outperform prior results on simulators of HIV treatment, kidney dialysis and battery charging. We also demonstrate that our methods can be useful for applications in AI safety by quickly identifying when a new decision policy is likely to have substantially lower performance than past policies."
                    },
                    {
                        "title": "A/B testing under Interference with Partial Network Information",
                        "abstract": "A/B tests are often required to be conducted on subjects that might have social connections. For e.g., experiments on social media, or medical and social interventions to control the spread of an epidemic. In such settings, the SUTVA assumption for randomized-controlled trials is violated due to network interference, or spill-over effects, as treatments to group A can potentially also affect the control group B. When the underlying social network is known exactly, prior works have demonstrated how to conduct A/B tests adequately to estimate the global average treatment effect (GATE). However, in practice, it is often impossible to obtain knowledge about the exact underlying network. In this paper, we present UNITE: a novel estimator that relax this assumption and can identify GATE while only relying on knowledge of the superset of neighbors for any subject in the graph. Through theoretical analysis and extensive experiments, we show that the proposed approach performs better in comparison to standard estimators."
                    },
                    {
                        "title": "Averaging log-likelihoods in direct alignment",
                        "abstract": "To better align Large Language Models (LLMs) with human judgment, Reinforcement Learning from Human Feedback (RLHF) learns a reward model and then optimizes it using regularized RL. Recently, direct alignment methods were introduced to learn such a fine-tuned model directly from a preference dataset without computing a proxy reward function. These methods are built upon contrastive losses involving the log-likelihood of (dis)preferred completions according to the trained model. However, completions have various lengths, and the log-likelihood is not length-invariant. On the other side, the cross-entropy loss used in supervised training is length-invariant, as batches are typically averaged token-wise. To reconcile these approaches, we introduce a principled approach for making direct alignment length-invariant. Formally, we introduce a new averaging operator, to be composed with the optimality operator giving the best policy for the underlying RL problem. It translates into averaging the log-likelihood within the loss. We empirically study the effect of such averaging, observing a trade-off between the length of generations and their scores."
                    },
                    {
                        "title": "Estimating the Causal Treatment Effect of Unproductive Persistence",
                        "abstract": "There has been considerable work in classifying and predicting unproductive persistence, but much less in understanding its causal impact on downstream outcomes of interest, like external assessments. In general, it is experimentally challenging to understand the causal impact because, unlike in many other settings, we cannot directly intervene (to conduct a randomized control trial) and cause students to struggle unproductively in an authentic manner. In this work, we use data from a prior study that used virtual reality headsets to alert teacher\u2019s attention to students who were unproductively struggling. We show that we can use this as an instrumental variable, and use a two-stage least squares analysis to provide a causal estimate of the treatment effect of unproductive persistence on post-test performance. Our results further strengthen the importance of unproductive struggle and highlight the potential of leveraging instruments to identify causal treatment effects of student behaviors during the use of educational technology."
                    },
                    {
                        "title": "Contrastive Policy Gradient: Aligning LLMs on sequence-level scores in a supervised-friendly fashion",
                        "abstract": "Reinforcement Learning (RL) has been used to finetune Large Language Models (LLMs) using a reward model trained from preference data, to better align with human judgment. The recently introduced direct alignment methods, which are often simpler, more stable, and computationally lighter, can more directly achieve this. However, these approaches cannot optimize arbitrary rewards, and the preference-based ones are not the only rewards of interest for LLMs (eg., unit tests for code generation or textual entailment for summarization, among others). RL-finetuning is usually done with a variation of policy gradient, which calls for on-policy or near-on-policy samples, requiring costly generations. We introduce Contrastive Policy Gradient, or CoPG, a simple and mathematically principled new RL algorithm that can estimate the optimal policy even from off-policy data. It can be seen as an off-policy policy gradient approach that does not rely on important sampling techniques and highlights the importance of using (the right) state baseline. We show this approach to generalize the direct alignment method IPO (identity preference optimization) and classic policy gradient. We experiment with the proposed CoPG on a toy bandit problem to illustrate its properties, as well as for finetuning LLMs on a summarization task, using a learned reward function considered as ground truth for the purpose of the experiments."
                    },
                    {
                        "title": "Optimization using Parallel Gradient Evaluations on Multiple Parameters",
                        "abstract": "We propose a first-order method for convex optimization, where instead of being restricted to the gradient from a single parameter, gradients from multiple parameters can be used during each step of gradient descent. This setup is particularly useful when a few processors are available that can be used in parallel for optimization. Our method uses gradients from multiple parameters in synergy to update these parameters together towards the optima. While doing so, it is ensured that the computational and memory complexity is of the same order as that of gradient descent. Empirical results demonstrate that even using gradients from as low as \\textit{two} parameters, our method can often obtain significant acceleration and provide robustness to hyper-parameter settings. We remark that the primary goal of this work is less theoretical, and is instead aimed at exploring the understudied case of using multiple gradients during each step of optimization."
                    },
                    {
                        "title": "Supervised Pretraining Can Learn In-Context Reinforcement Learning",
                        "abstract": "Large transformer models trained on diverse datasets have shown a remarkable ability to learn in-context, achieving high few-shot performance on tasks they were not explicitly trained to solve. In this paper, we study the in-context learning capabilities of transformers in decision-making problems, i.e., reinforcement learning (RL) for bandits and Markov decision processes. To do so, we introduce and study Decision-Pretrained Transformer (DPT), a supervised pretraining method where the transformer predicts an optimal action given a query state and an in-context dataset of interactions, across a diverse set of tasks. This procedure, while simple, produces a model with several surprising capabilities. We find that the pretrained transformer can be used to solve a range of RL problems in-context, exhibiting both exploration online and conservatism offline, despite not being explicitly trained to do so. The model also generalizes beyond the pretraining distribution to new tasks and automatically adapts its decision-making strategies to unknown structure. Theoretically, we show DPT can be viewed as an efficient implementation of Bayesian posterior sampling, a provably sample-efficient RL algorithm. We further leverage this connection to provide guarantees on the regret of the in-context algorithm yielded by DPT, and prove that it can learn faster than algorithms used to generate the pretraining data. These results suggest a promising yet simple path towards instilling strong in-context decision-making abilities in transformers."
                    },
                    {
                        "title": "Representations and Exploration for Deep Reinforcement Learning using Singular Value Decomposition",
                        "abstract": "Representation learning and exploration are among the key challenges for any deep reinforcement learning agent. In this work, we provide a singular value decomposition based method that can be used to obtain representations that preserve the underlying transition structure in the domain. Perhaps interestingly, we show that these representations also capture the relative frequency of state visitations, thereby providing an estimate for pseudo-counts for free. To scale this decomposition method to large-scale domains, we provide an algorithm that never requires building the transition matrix, can make use of deep networks, and also permits mini-batch training. Further, we draw inspiration from predictive state representations and extend our decomposition method to partially observable environments. With experiments on multi-task settings with partially observable domains, we show that the proposed method can not only learn useful representation on DM-Lab-30 environments (that have inputs involving language instructions, pixel images, and rewards, among others) but it can also be effective at hard exploration tasks in DM-Hard-8 environments."
                    },
                    {
                        "title": "Behavior Alignment via Reward Function Optimization",
                        "abstract": "Designing reward functions for efficiently guiding reinforcement learning (RL) agents toward specific behaviors is a complex task. This is challenging since it requires the identification of reward structures that are not sparse and that avoid inadvertently inducing undesirable behaviors. Naively modifying the reward structure to offer denser and more frequent feedback can lead to unintended outcomes and promote behaviors that are not aligned with the designer's intended goal. Although potential-based reward shaping is often suggested as a remedy, we systematically investigate settings where deploying it often significantly impairs performance. To address these issues, we introduce a new framework that uses a bi-level objective to learn \\emph{behavior alignment reward functions}. These functions integrate auxiliary rewards reflecting a designer's heuristics and domain knowledge with the environment's primary rewards. Our approach automatically determines the most effective way to blend these types of feedback, thereby enhancing robustness against heuristic reward misspecification. Remarkably, it can also adapt an agent's policy optimization process to mitigate suboptimalities resulting from limitations and biases inherent in the underlying RL algorithms. We evaluate our method's efficacy on a diverse set of tasks, from small-scale experiments to high-dimensional control challenges. We investigate heuristic auxiliary rewards of varying quality -- some of which are beneficial and others detrimental to the learning process. Our results show that our framework offers a robust and principled way to integrate designer-specified heuristics. It not only addresses key shortcomings of existing approaches but also consistently leads to high-performing solutions, even when given misaligned or poorly-specified auxiliary reward functions."
                    },
                    {
                        "title": "Off-Policy Evaluation for Action-Dependent Non-Stationary Environments",
                        "abstract": "Methods for sequential decision-making are often built upon a foundational assumption that the underlying decision process is stationary. This limits the application of such methods because real-world problems are often subject to changes due to external factors (passive non-stationarity), changes induced by interactions with the system itself (active non-stationarity), or both (hybrid non-stationarity). In this work, we take the first steps towards the fundamental challenge of on-policy and off-policy evaluation amidst structured changes due to active, passive, or hybrid non-stationarity. Towards this goal, we make a higher-order stationarity assumption such that non-stationarity results in changes over time, but the way changes happen is fixed. We propose, OPEN, an algorithm that uses a double application of counterfactual reasoning and a novel importance-weighted instrument-variable regression to obtain both a lower bias and a lower variance estimate of the structure in the changes of a policy's past performances. Finally, we show promising results on how OPEN can be used to predict future performances for several domains inspired by real-world applications that exhibit non-stationarity."
                    },
                    {
                        "title": "Asymptotically Unbiased Off-Policy Policy Evaluation when Reusing Old Data in Nonstationary Environments",
                        "abstract": "In this work, we consider the off-policy policy evaluation problem for contextual bandits and finite horizon reinforcement learning in the nonstationary setting. Reusing old data is critical for policy evaluation, but existing estimators that reuse old data introduce large bias such that we can not obtain a valid confidence interval. Inspired from a related field called survey sampling, we introduce a variant of the doubly robust (DR) estimator, called the regression-assisted DR estimator, that can incorporate the past data without introducing a large bias. The estimator unifies several existing off-policy policy evaluation methods and improves on them with the use of auxiliary information and a regression approach. We prove that the new estimator is asymptotically unbiased, and provide a consistent variance estimator to a construct a large sample confidence interval. Finally, we empirically show that the new estimator improves estimation for the current and future policy values, and provides a tight and valid interval estimation in several nonstationary recommendation environments."
                    },
                    {
                        "title": "Coagent Networks: Generalized and Scaled",
                        "abstract": "Coagent networks for reinforcement learning (RL) [Thomas and Barto, 2011] provide a powerful and flexible framework for deriving principled learning rules for arbitrary stochastic neural networks. The coagent framework offers an alternative to backpropagation-based deep learning (BDL) that overcomes some of backpropagation's main limitations. For example, coagent networks can compute different parts of the network \\emph{asynchronously} (at different rates or at different times), can incorporate non-differentiable components that cannot be used with backpropagation, and can explore at levels higher than their action spaces (that is, they can be designed as hierarchical networks for exploration and/or temporal abstraction). However, the coagent framework is not just an alternative to BDL; the two approaches can be blended: BDL can be combined with coagent learning rules to create architectures with the advantages of both approaches. This work generalizes the coagent theory and learning rules provided by previous works; this generalization provides more flexibility for network architecture design within the coagent framework. This work also studies one of the chief disadvantages of coagent networks: high variance updates for networks that have many coagents and do not use backpropagation. We show that a coagent algorithm with a policy network that does not use backpropagation can scale to a challenging RL domain with a high-dimensional state and action space (the MuJoCo Ant environment), learning reasonable (although not state-of-the-art) policies. These contributions motivate and provide a more general theoretical foundation for future work that studies coagent networks."
                    },
                    {
                        "title": "The Geometry of Multilingual Language Models: An Equality Lens",
                        "abstract": "Understanding the representations of different languages in multilingual language models is essential for comprehending their cross-lingual properties, predicting their performance on downstream tasks, and identifying any biases across languages. In our study, we analyze the geometry of three multilingual language models in Euclidean space and find that all languages are represented by unique geometries. Using a geometric separability index we find that although languages tend to be closer according to their linguistic family, they are almost separable with languages from other families. We also introduce a Cross-Lingual Similarity Index to measure the distance of languages with each other in the semantic space. Our findings indicate that the low-resource languages are not represented as good as high resource languages in any of the models"
                    },
                    {
                        "title": "Factored DRO: Factored Distributionally Robust Policies for Contextual Bandits",
                        "abstract": "While there has been extensive work on learning from of\ufb02ine data for contextual multi-armed bandit settings, existing methods typically assume there is no environment shift: that the learned policy will operate in the same environmental process as that of data collection. However, this assumption may limit the use of these methods for many practical situations where there may be distribution shifts. In this work we propose Factored Distributionally Robust Optimization (Factored-DRO) 1 , which is able to separately handle distribution shifts in the context distribution and shifts in the reward generating process. Prior work that either ignores potential shifts in the context, or considers them jointly, can lead to performance that is too conservative, especially under certain forms of reward feedback. Our Factored-DRO objective mitigates this by considering the shifts separately, and our proposed estimators are consistent and converge asymptotically. We also introduce a practical algorithm and demonstrate promising empirical results in environments based on real-world datasets, such as voting outcomes and scene classi\ufb01cation."
                    },
                    {
                        "title": "The impact of boron doping in the tunneling magnetoresistance of Heusler alloy Co2FeAl",
                        "abstract": "Heusler alloy-based magnetic tunnel junctions have the potential to provide high spin polarization, small damping, and fast switching. In this study, junctions with a ferromagnetic electrode of Co2FeAl were fabricated via room-temperature sputtering on Si/SiO2 substrates. The effect of boron doping on Co2FeAl magnetic tunnel junctions was investigated for different boron concentrations. The surface roughness determined by atomic force microscope, and the analysis of x-ray diffraction measurement on the Co2FeAl thin film reveals critical information about the interface. The Co2FeAl layer was deposited on the bottom and on the top of the insulating MgO layer as two different sample structures to compare the impact of the boron doping on different layers through tunneling magnetoresistance measurements. The doping of boron in Co2FeAl had a large positive impact on the structural and magneto-transport properties of the junctions, with reduced interfacial roughness and substantial improvement in tunneling magnetoresistance. In samples annealed at low temperature, a two-level magnetoresistance was also observed. This is believed to be related to the memristive effect of the tunnel barrier. The findings of this study have practical uses for the design and fabrication of magnetic tunnel junctions with improved magneto-transport properties."
                    },
                    {
                        "title": "Understanding Self-Predictive Learning for Reinforcement Learning",
                        "abstract": "We study the learning dynamics of self-predictive learning for reinforcement learning, a family of algorithms that learn representations by minimizing the prediction error of their own future latent representations. Despite its recent empirical success, such algorithms have an apparent defect: trivial representations (such as constants) minimize the prediction error, yet it is obviously undesirable to converge to such solutions. Our central insight is that careful designs of the optimization dynamics are critical to learning meaningful representations. We identify that a faster paced optimization of the predictor and semi-gradient updates on the representation, are crucial to preventing the representation collapse. Then in an idealized setup, we show self-predictive learning dynamics carries out spectral decomposition on the state transition matrix, effectively capturing information of the transition dynamics. Building on the theoretical insights, we propose bidirectional self-predictive learning, a novel self-predictive algorithm that learns two representations simultaneously. We examine the robustness of our theoretical insights with a number of small-scale experiments and showcase the promise of the novel representation learning algorithm with large-scale experiments."
                    },
                    {
                        "title": "Scaling Graph Propagation Kernels for Predictive Learning",
                        "abstract": "Many real-world applications deal with data that have an underlying graph structure associated with it. To perform downstream analysis on such data, it is crucial to capture relational information of nodes over their expanded neighborhood efficiently. Herein, we focus on the problem of Collective Classification (CC) for assigning labels to unlabeled nodes. Most deep learning models for CC heavily rely on differentiable variants of Weisfeiler-Lehman (WL) kernels. However, due to current computing architectures' limitations, WL kernels and their differentiable variants are limited in their ability to capture useful relational information only over a small expanded neighborhood of a node. To address this concern, we propose the framework, I-HOP, that couples differentiable kernels with an iterative inference mechanism to scale to larger neighborhoods. I-HOP scales differentiable graph kernels to capture and summarize information from a larger neighborhood in each iteration by leveraging a historical neighborhood summary obtained in the previous iteration. This recursive nature of I-HOP provides an exponential reduction in time and space complexity over straightforward differentiable graph kernels. Additionally, we point out a limitation of WL kernels where the node's original information is decayed exponentially with an increase in neighborhood size and provide a solution to address it. Finally, extensive evaluation across 11 datasets showcases the improved results and robustness of our proposed iterative framework, I-HOP."
                    }
                ]
            },
            "5c42973b-46c4-474a-ba29-1bbaeaf6dc9d": {
                "pk": "5c42973b-46c4-474a-ba29-1bbaeaf6dc9d",
                "name": "Shiv Shankar",
                "collaborators": [
                    "Yash Chandak",
                    "Ritwik Sinha",
                    "M. Fiterau",
                    "P. Thomas",
                    "Saayan Mitra",
                    "Moumita Sinha",
                    "V. Gandikota",
                    "A. Mazumdar",
                    "Nathaniel D. Bastian",
                    "B. C. Silva",
                    "Emma Brunskil",
                    "Laure Thompson",
                    "Pushpa L. Trivedi",
                    "Viswanathan Swaminathan",
                    "Sridhar Mahadevan",
                    "D. Towsley",
                    "Vihari Piratla",
                    "Frank Bentley",
                    "Josh Jacobson",
                    "Charlotte Sperling",
                    "Christiaan Royer",
                    "I. McCarthy",
                    "Georgios Theocharous",
                    "S. Mahadevan",
                    "Martha White"
                ],
                "domain": [
                    "Causal Inference",
                    "Machine Learning",
                    "Reinforcement Learning",
                    "Multimodal Integration"
                ],
                "publications": [
                    {
                        "title": "A/B testing under Interference with Partial Network Information",
                        "abstract": "A/B tests are often required to be conducted on subjects that might have social connections. For e.g., experiments on social media, or medical and social interventions to control the spread of an epidemic. In such settings, the SUTVA assumption for randomized-controlled trials is violated due to network interference, or spill-over effects, as treatments to group A can potentially also affect the control group B. When the underlying social network is known exactly, prior works have demonstrated how to conduct A/B tests adequately to estimate the global average treatment effect (GATE). However, in practice, it is often impossible to obtain knowledge about the exact underlying network. In this paper, we present UNITE: a novel estimator that relax this assumption and can identify GATE while only relying on knowledge of the superset of neighbors for any subject in the graph. Through theoretical analysis and extensive experiments, we show that the proposed approach performs better in comparison to standard estimators."
                    },
                    {
                        "title": "On Online Experimentation without Device Identifiers",
                        "abstract": "Measuring human feedback via randomized experimentation is a cornerstone of data-driven decision-making. The methodology used to estimate user preferences from their online behaviours is critically dependent on user identifiers. However, in today\u2019s digital landscape, consumers frequently interact with content across multiple devices, which are often recorded with different identifiers for the same consumer. The inability to match different device identities across consumers poses significant challenges for accurately estimating human preferences and other causal effects. Moreover, without strong assumptions about the device-user graph, the causal effects might not be identifiable. In this paper, we pro-pose HIFIVE, a variational method to solve the problem of estimating global average treatment effects (GATE) from a fragmented view of exposures and outcomes. Experiments show that our estimator is superior to standard estimators, with a lower bias and greater robustness to network uncertainty."
                    },
                    {
                        "title": "Optimization using Parallel Gradient Evaluations on Multiple Parameters",
                        "abstract": "We propose a first-order method for convex optimization, where instead of being restricted to the gradient from a single parameter, gradients from multiple parameters can be used during each step of gradient descent. This setup is particularly useful when a few processors are available that can be used in parallel for optimization. Our method uses gradients from multiple parameters in synergy to update these parameters together towards the optima. While doing so, it is ensured that the computational and memory complexity is of the same order as that of gradient descent. Empirical results demonstrate that even using gradients from as low as \\textit{two} parameters, our method can often obtain significant acceleration and provide robustness to hyper-parameter settings. We remark that the primary goal of this work is less theoretical, and is instead aimed at exploring the understudied case of using multiple gradients during each step of optimization."
                    },
                    {
                        "title": "Direct Inference of Effect of Treatment (DIET) for a Cookieless World",
                        "abstract": "Brands use cookies and device identifiers to link different web visits to the same consumer. However, with increasing demands for privacy, these identifiers are about to be phased out, making identity fragmentation a permanent feature of the online world. Assessing treatment effects via randomized experiments (A/B testing) in such a scenario is challenging because identity fragmentation causes a) users to receive hybrid/mixed treatments, and b) hides the causal link between the historical treatments and the outcome. In this work, we address the problem of estimating treatment effects when a lack of identification leads to incomplete knowledge of historical treatments. This is a challenging problem which has not been addressed in literature yet. We develop a new method called DIET, which can adjust for users being exposed to mixed treatments without the entire history of treatments being available . Our method takes inspiration from the Cox model, and uses a proportional outcome approach under which we prove that one can obtain consistent estimates of treatment effects even under identity fragmentation. Our experiments, on one simulated and two real datasets, show that our method leads to up to 20% reduction in error and 25% reduction in bias over the naive estimate."
                    },
                    {
                        "title": "Off-Policy Evaluation for Action-Dependent Non-Stationary Environments",
                        "abstract": "Methods for sequential decision-making are often built upon a foundational assumption that the underlying decision process is stationary. This limits the application of such methods because real-world problems are often subject to changes due to external factors (passive non-stationarity), changes induced by interactions with the system itself (active non-stationarity), or both (hybrid non-stationarity). In this work, we take the first steps towards the fundamental challenge of on-policy and off-policy evaluation amidst structured changes due to active, passive, or hybrid non-stationarity. Towards this goal, we make a higher-order stationarity assumption such that non-stationarity results in changes over time, but the way changes happen is fixed. We propose, OPEN, an algorithm that uses a double application of counterfactual reasoning and a novel importance-weighted instrument-variable regression to obtain both a lower bias and a lower variance estimate of the structure in the changes of a policy's past performances. Finally, we show promising results on how OPEN can be used to predict future performances for several domains inspired by real-world applications that exhibit non-stationarity."
                    },
                    {
                        "title": "Implicit Training of Inference Network Models for Structured Prediction",
                        "abstract": "Most research in deep learning has predominantly focused on the development of new models and training procedures. In contrast, the exploration of training objectives has received considerably less attention, often limited to combinations of standard losses. When dealing with complex structured outputs, the effectiveness of conventional objectives as proxies for the true objective becomes can be questionable. In this study, we propose that existing inference network-based methods for structured prediction, as observed in previous works [Tu and Gimpel, 2018, Tu et al., 2020a], indirectly learn to optimize a dynamic loss objective parameterized by the energy model. Based on this insight, we propose a method that treats the energy network as a trainable loss function and employs an implicit-gradient-based technique to learn the corresponding dynamic objective. We experiment with multiple tasks such as multi-label classi\ufb01cation, entity recognition, etc. and \ufb01nd signi\ufb01cant performance improvements over baseline approaches. Our results demonstrate that implicitly learning a dynamic loss landscape proves to be an effective approach for enhancing model performance in structured prediction tasks."
                    },
                    {
                        "title": "Progressive Fusion for Multimodal Integration",
                        "abstract": "Integration of multimodal information from various sources has been shown to boost the performance of machine learning models and thus has received increased attention in recent years. Often such models use deep modality-specific networks to obtain unimodal features which are combined to obtain\"late-fusion\"representations. However, these designs run the risk of information loss in the respective unimodal pipelines. On the other hand,\"early-fusion\"methodologies, which combine features early, suffer from the problems associated with feature heterogeneity and high sample complexity. In this work, we present an iterative representation refinement approach, called Progressive Fusion, which mitigates the issues with late fusion representations. Our model-agnostic technique introduces backward connections that make late stage fused representations available to early layers, improving the expressiveness of the representations at those stages, while retaining the advantages of late fusion designs. We test Progressive Fusion on tasks including affective sentiment detection, multimedia analysis, and time series fusion with different models, demonstrating its versatility. We show that our approach consistently improves performance, for instance attaining a 5% reduction in MSE and 40% improvement in robustness on multimodal time series prediction."
                    },
                    {
                        "title": "Evaluating the Long-run Sustainability of India\u2019s Fiscal Management with Structural Change",
                        "abstract": "Amidst output moderation, rising deficits and increasing debt, India\u2019s macro-fiscal arithmetic witnesses severe strain and often invites downward rating pressures by sovereign rating agencies. The article aims to examine India\u2019s fiscal sustainability during the past four decades employing time series integration and cointegration techniques, with structural breaks in a sequential schematic framework under intertemporal government budget constraints. It examines stationarity with exogenous and endogenous structure breakpoint(s) at the level and slope for government revenue and expenditure data-generating process following Narayan and Popp (2010), Lee and Strazicich (2003), Zivot and Andrews (1992) and Perron (1989). Furthermore, the cointegration vectors of these fiscal variables in Fully Modified Ordinary Least Squares (FMOLS), Dynamic Ordinary Least Squares (DOLS) and a generic cointegration framework following Gregory and Hansen (1996) with structural shifts confirm the sustainability of India\u2019s fiscal management. However, a less-than-unity estimate of the DOLS cointegrating slope parameter with few exogenous breakpoints signifies a weak form of sustainability and hence emphasises a credible commitment to fiscal consolidation going forward for India. JEL Codes: C32, H50, E62, H62"
                    },
                    {
                        "title": "Privacy Aware Experiments without Cookies",
                        "abstract": "Consider two brands that want to jointly test alternate web experiences for their customers with an A/B test. Such collaborative tests are today enabled usingthird-party cookies, where each brand has information on the identity of visitors to another website, ensuring a consistent treatment experience. With the imminent elimination of third-party cookies, such A/B tests will become untenable. We propose a two-stage experimental design, where the two brands only need to agree on high-level aggregate parameters of the experiment to test the alternate experiences. Our design respects the privacy of customers. We propose an unbiased estimator of the Average Treatment Effect (ATE), and provide a way to use regression adjustment to improve this estimate. On real and simulated data, we show that the approach provides valid estimate of the ATE and is robust to the proportion of visitors overlapping across the brands. Our demonstration describes how a marketer can design such an experiment and analyze the results."
                    },
                    {
                        "title": "Variational Boson Sampling",
                        "abstract": ". Boson Samplers are near-term quantum devices based on photonic quantum technology, which can outperform classical computing systems. This paper takes a hybrid circuit learning approach to utilize boson samplers as a generative model called Variational Boson Sampling (VBS). VBS introduces an optimizable parametric structure into the evolution operator for boson sampling and uses the complete model as a variational ansatz. To simulate working with real quantum devices, we use gradient free-optimization methods to optimize the resultant circuit. We experiment with this framework for problems in optimization and generative modeling."
                    },
                    {
                        "title": "Implicit Training of Energy Model for Structure Prediction",
                        "abstract": "Most deep learning research has focused on developing new model and training procedures. On the other hand the training objective has usually been restricted to combinations of standard losses. When the objective aligns well with the evaluation metric, this is not a major issue. However when dealing with complex structured outputs, the ideal objective can be hard to optimize and the efficacy of usual objectives as a proxy for the true objective can be questionable. In this work, we argue that the existing inference network based structure prediction methods ( Tu and Gimpel 2018; Tu, Pang, and Gimpel 2020) are indirectly learning to optimize a dynamic loss objective parameterized by the energy model. We then explore using implicit-gradient based technique to learn the corresponding dynamic objectives. Our experiments show that implicitly learning a dynamic loss landscape is an effective method for improving model performance in structure prediction."
                    },
                    {
                        "title": "Multimodal fusion via cortical network inspired losses",
                        "abstract": "Information integration from different modalities is an active area of research. Human beings and, in general, biological neural systems are quite adept at using a multitude of signals from different sensory perceptive fields to interact with the environment and each other. Recent work in deep fusion models via neural networks has led to substantial improvements over unimodal approaches in areas like speech recognition, emotion recognition and analysis, captioning and image description. However, such research has mostly focused on architectural changes allowing for fusion of different modalities while keeping the model complexity manageable.Inspired by neuroscientific ideas about multisensory integration and processing, we investigate the effect of introducing neural dependencies in the loss functions. Experiments on multimodal sentiment analysis tasks with different models show that our approach provides a consistent performance boost."
                    },
                    {
                        "title": "High-Confidence Off-Policy (or Counterfactual) Variance Estimation",
                        "abstract": "Many sequential decision-making systems leverage data collected using prior policies to propose a new policy. For critical applications, it is important that high-confidence guarantees on the new policy\u2019s behavior are provided before deployment, to ensure that the policy will behave as desired. Prior works have studied high-confidence off-policy estimation of the expected return, however, high-confidence off-policy estimation of the variance of returns can be equally critical for high-risk applications. In this paper we tackle the previously open problem of estimating and bounding, with high confidence, the variance of returns from off-policy data."
                    },
                    {
                        "title": "Adversarial Stein Training for Graph Energy Models",
                        "abstract": "Learning distributions over graph-structured data is a challenging task with many applications in biology and chemistry. In this work we use an energy-based model (EBM) based on multi-channel graph neural networks (GNN) to learn permutation invariant unnormalized density functions on graphs. Unlike standard EBM training methods our approach is to learn the model via minimizing adversarial stein discrepancy. Samples from the model can be obtained via Langevin dynamics based MCMC. We find that this approach achieves competitive results on graph generation compared to benchmark models."
                    },
                    {
                        "title": "Neural Dependency Coding inspired Multimodal Fusion",
                        "abstract": "Information integration from different modalities is an active area of research. Human beings and, in general, biological neural systems are quite adept at using a multitude of signals from different sensory perceptive fields to interact with the environment and each other. Recent work in deep fusion models via neural networks has led to substantial improvements over unimodal approaches in areas like speech recognition, emotion recognition and analysis, captioning and image description. However, such research has mostly focused on architectural changes allowing for fusion of different modalities while keeping the model complexity manageable. Inspired by recent neuroscience ideas about multisensory integration and processing, we investigate the effect of synergy maximizing loss functions. Experiments on multimodal sentiment analysis tasks: CMU-MOSI and CMU-MOSEI with different models show that our approach provides a consistent performance boost."
                    },
                    {
                        "title": "Rethinking Consumer Email: The Research Process for Yahoo Mail 6",
                        "abstract": "This case study follows the research process of rethinking the design and functionality of a personal email client, Yahoo Mail. Over three years, we changed the focus of the product from composing emails towards automatically organizing specific categories of business to consumer email (such as deals, receipts, and travel) and creating experiences unique to each category. To achieve this, we employed iterative user research with over 1,500 in-person interviews in six countries and surveys to many thousands of people around the world. This research process culminated in the launch of Yahoo Mail 6.0 for iOS and Android devices in the fall of 2019."
                    },
                    {
                        "title": "Optimizing for the Future in Non-Stationary MDPs",
                        "abstract": "Most reinforcement learning methods are based upon the key assumption that the transition dynamics and reward functions are fixed, that is, the underlying Markov decision process is stationary. However, in many real-world applications, this assumption is violated, and using existing algorithms may result in a performance lag. To proactively search for a good future policy, we present a policy gradient algorithm that maximizes a forecast of future performance. This forecast is obtained by fitting a curve to the counter-factual estimates of policy performance over time, without explicitly modeling the underlying non-stationarity. The resulting algorithm amounts to a non-uniform reweighting of past data, and we observe that minimizing performance over some of the data from past episodes can be beneficial when searching for a policy that maximizes future performance. We show that our algorithm, called Prognosticator, is more robust to non-stationarity than two online adaptation techniques, on three simulated problems motivated by real-world applications."
                    }
                ]
            },
            "01545f3b-4a07-42ae-b500-c561c8c45797": {
                "pk": "01545f3b-4a07-42ae-b500-c561c8c45797",
                "name": "Vasilis Syrgkanis",
                "collaborators": [
                    "Nathan Kallus",
                    "Masatoshi Uehara",
                    "Keertana Chidambaram",
                    "Hui Lan",
                    "Christopher Jung",
                    "Zhiwei Steven Wu",
                    "Rahul G. Krishnan",
                    "Keegan Harris",
                    "Jikai Jin",
                    "Andrew Bennett",
                    "Xiaojie Mao",
                    "Whitney Newey",
                    "V. Chernozhukov",
                    "Christian Hansen",
                    "Martin Spindler",
                    "Zihao Li",
                    "Mengdi Wang",
                    "Yifan Wu",
                    "Ramesh Johari",
                    "G. Weintraub",
                    "Ravi B. Sojitra",
                    "J. Whitehouse",
                    "Bryan Wilder",
                    "Jiyuan Tan",
                    "Jose Blanchet",
                    "Karthik Vinay Seetharaman",
                    "Jabari Hastings",
                    "Charlotte Peale",
                    "Allison Lau",
                    "Younwoo Choi",
                    "Vahid Balazadeh",
                    "David M. Ritzwoller",
                    "Vahid Balazadeh Meresht",
                    "Viet Nguyen",
                    "Qizhao Chen",
                    "Morgane Austern",
                    "Daniel Ngo",
                    "Anish Agarwal"
                ],
                "domain": [
                    "Causal Inference",
                    "Machine Learning",
                    "Instrumental Variables",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Applied Causal Inference Powered by ML and AI",
                        "abstract": "An introduction to the emerging fusion of machine learning and causal inference. The book presents ideas from classical structural equation models (SEMs) and their modern AI equivalent, directed acyclical graphs (DAGs) and structural causal models (SCMs), and covers Double/Debiased Machine Learning methods to do inference in such models using modern predictive tools."
                    },
                    {
                        "title": "Regularized DeepIV with Model Selection",
                        "abstract": "In this paper, we study nonparametric estimation of instrumental variable (IV) regressions. While recent advancements in machine learning have introduced flexible methods for IV estimation, they often encounter one or more of the following limitations: (1) restricting the IV regression to be uniquely identified; (2) requiring minimax computation oracle, which is highly unstable in practice; (3) absence of model selection procedure. In this paper, we present the first method and analysis that can avoid all three limitations, while still enabling general function approximation. Specifically, we propose a minimax-oracle-free method called Regularized DeepIV (RDIV) regression that can converge to the least-norm IV solution. Our method consists of two stages: first, we learn the conditional distribution of covariates, and by utilizing the learned distribution, we learn the estimator by minimizing a Tikhonov-regularized loss function. We further show that our method allows model selection procedures that can achieve the oracle rates in the misspecified regime. When extended to an iterative estimator, our method matches the current state-of-the-art convergence rate. Our method is a Tikhonov regularized variant of the popular DeepIV method with a non-parametric MLE first-stage estimator, and our results provide the first rigorous guarantees for this empirically used method, showcasing the importance of regularization which was absent from the original work."
                    },
                    {
                        "title": "Switchback Price Experiments with Forward-Looking Demand",
                        "abstract": "We consider a retailer running a switchback experiment for the price of a single product, with infinite supply. In each period, the seller chooses a price $p$ from a set of predefined prices that consist of a reference price and a few discounted price levels. The goal is to estimate the demand gradient at the reference price point, with the goal of adjusting the reference price to improve revenue after the experiment. In our model, in each period, a unit mass of buyers arrives on the market, with values distributed based on a time-varying process. Crucially, buyers are forward looking with a discounted utility and will choose to not purchase now if they expect to face a discounted price in the near future. We show that forward-looking demand introduces bias in naive estimators of the demand gradient, due to intertemporal interference. Furthermore, we prove that there is no estimator that uses data from price experiments with only two price points that can recover the correct demand gradient, even in the limit of an infinitely long experiment with an infinitesimal price discount. Moreover, we characterize the form of the bias of naive estimators. Finally, we show that with a simple three price level experiment, the seller can remove the bias due to strategic forward-looking behavior and construct an estimator for the demand gradient that asymptotically recovers the truth."
                    },
                    {
                        "title": "Dynamic Local Average Treatment Effects",
                        "abstract": "We consider Dynamic Treatment Regimes (DTRs) with One Sided Noncompliance that arise in applications such as digital recommendations and adaptive medical trials. These are settings where decision makers encourage individuals to take treatments over time, but adapt encouragements based on previous encouragements, treatments, states, and outcomes. Importantly, individuals may not comply with encouragements based on unobserved confounders. For settings with binary treatments and encouragements, we provide nonparametric identification, estimation, and inference for Dynamic Local Average Treatment Effects (LATEs), which are expected values of multiple time period treatment contrasts for the respective complier subpopulations. Under standard assumptions in the Instrumental Variable and DTR literature, we show that one can identify Dynamic LATEs that correspond to treating at single time steps. Under an additional cross-period effect-compliance independence assumption, which is satisfied in Staggered Adoption settings and a generalization of them, which we define as Staggered Compliance settings, we identify Dynamic LATEs for treating in multiple time periods."
                    },
                    {
                        "title": "Orthogonal Causal Calibration",
                        "abstract": "Estimates of causal parameters such as conditional average treatment effects and conditional quantile treatment effects play an important role in real-world decision making. Given this importance, one should ensure these estimators are calibrated. While there is a rich literature on calibrating estimators of non-causal parameters, very few methods have been derived for calibrating estimators of causal parameters, or more generally estimators of quantities involving nuisance parameters. In this work, we provide a general framework for calibrating predictors involving nuisance estimation. We consider a notion of calibration defined with respect to an arbitrary, nuisance-dependent loss $\\ell$, under which we say an estimator $\\theta$ is calibrated if its predictions cannot be changed on any level set to decrease loss. We prove generic upper bounds on the calibration error of any causal parameter estimate $\\theta$ with respect to any loss $\\ell$ using a concept called Neyman Orthogonality. Our bounds involve two decoupled terms - one measuring the error in estimating the unknown nuisance parameters, and the other representing the calibration error in a hypothetical world where the learned nuisance estimates were true. We use our bound to analyze the convergence of two sample splitting algorithms for causal calibration. One algorithm, which applies to universally orthogonalizable loss functions, transforms the data into generalized pseudo-outcomes and applies an off-the-shelf calibration procedure. The other algorithm, which applies to conditionally orthogonalizable loss functions, extends the classical uniform mass binning algorithm to include nuisance estimation. Our results are exceedingly general, showing that essentially any existing calibration algorithm can be used in causal settings, with additional loss only arising from errors in nuisance estimation."
                    },
                    {
                        "title": "Consistency of Neural Causal Partial Identification",
                        "abstract": "Recent progress in Neural Causal Models (NCMs) showcased how identification and partial identification of causal effects can be automatically carried out via training of neural generative models that respect the constraints encoded in a given causal graph [Xia et al. 2022, Balazadeh et al. 2022]. However, formal consistency of these methods has only been proven for the case of discrete variables or only for linear causal models. In this work, we prove consistency of partial identification via NCMs in a general setting with both continuous and categorical variables. Further, our results highlight the impact of the design of the underlying neural network architecture in terms of depth and connectivity as well as the importance of applying Lipschitz regularization in the training phase. In particular, we provide a counterexample showing that without Lipschitz regularization the NCM may not be asymptotically consistent. Our results are enabled by new results on the approximability of structural causal models via neural generative models, together with an analysis of the sample complexity of the resulting architectures and how that translates into an error in the constrained optimization problem that defines the partial identification bounds."
                    },
                    {
                        "title": "Direct Preference Optimization With Unobserved Preference Heterogeneity",
                        "abstract": "RLHF has emerged as a pivotal step in aligning language models with human objectives and values. It typically involves learning a reward model from human preference data and then using reinforcement learning to update the generative model accordingly. Conversely, Direct Preference Optimization (DPO) directly optimizes the generative model with preference data, skipping reinforcement learning. However, both RLHF and DPO assume uniform preferences, overlooking the reality of diverse human annotators. This paper presents a new method to align generative models with varied human preferences. We propose an Expectation-Maximization adaptation to DPO, generating a mixture of models based on latent preference types of the annotators. We then introduce a min-max regret ensemble learning model to produce a single generative method to minimize worst-case regret among annotator subgroups with similar latent factors. Our algorithms leverage the simplicity of DPO while accommodating diverse preferences. Experimental results validate the effectiveness of our approach in producing equitable generative policies."
                    },
                    {
                        "title": "Taking a Moment for Distributional Robustness",
                        "abstract": "A rich line of recent work has studied distributionally robust learning approaches that seek to learn a hypothesis that performs well, in the worst-case, on many different distributions over a population. We argue that although the most common approaches seek to minimize the worst-case loss over distributions, a more reasonable goal is to minimize the worst-case distance to the true conditional expectation of labels given each covariate. Focusing on the minmax loss objective can dramatically fail to output a solution minimizing the distance to the true conditional expectation when certain distributions contain high levels of label noise. We introduce a new min-max objective based on what is known as the adversarial moment violation and show that minimizing this objective is equivalent to minimizing the worst-case $\\ell_2$-distance to the true conditional expectation if we take the adversary's strategy space to be sufficiently rich. Previous work has suggested minimizing the maximum regret over the worst-case distribution as a way to circumvent issues arising from differential noise levels. We show that in the case of square loss, minimizing the worst-case regret is also equivalent to minimizing the worst-case $\\ell_2$-distance to the true conditional expectation. Although their objective and our objective both minimize the worst-case distance to the true conditional expectation, we show that our approach provides large empirical savings in computational cost in terms of the number of groups, while providing the same noise-oblivious worst-distribution guarantee as the minimax regret approach, thus making positive progress on an open question posed by Agarwal and Zhang (2022)."
                    },
                    {
                        "title": "Personalized Adaptation via In-Context Preference Learning",
                        "abstract": "Reinforcement Learning from Human Feedback (RLHF) is widely used to align Language Models (LMs) with human preferences. However, existing approaches often neglect individual user preferences, leading to suboptimal personalization. We present the Preference Pretrained Transformer (PPT), a novel approach for adaptive personalization using online user feedback. PPT leverages the in-context learning capabilities of transformers to dynamically adapt to individual preferences. Our approach consists of two phases: (1) an offline phase where we train a single policy model using a history-dependent loss function, and (2) an online phase where the model adapts to user preferences through in-context learning. We demonstrate PPT's effectiveness in a contextual bandit setting, showing that it achieves personalized adaptation superior to existing methods while significantly reducing the computational costs. Our results suggest the potential of in-context learning for scalable and efficient personalization in large language models."
                    },
                    {
                        "title": "Causal Inference under Incentives: An Annotated Reading List",
                        "abstract": "We provide an overview of research on causal inference in the presence of strategic agents. Work in this area uses tools from econometrics, statistics, machine learning, and game theory to infer causal relationships between treatments and outcomes of interest when the treated individuals have an incentive to behave strategically."
                    },
                    {
                        "title": "Simultaneous Inference for Local Structural Parameters with Random Forests",
                        "abstract": "We construct simultaneous confidence intervals for solutions to conditional moment equations. The intervals are built around a class of nonparametric regression algorithms based on subsampled kernels. This class encompasses various forms of subsampled random forest regression, including Generalized Random Forests (Athey et al., 2019). Although simultaneous validity is often desirable in practice -- for example, for fine-grained characterization of treatment effect heterogeneity -- only confidence intervals that confer pointwise guarantees were previously available. Our work closes this gap. As a by-product, we obtain several new order-explicit results on the concentration and normal approximation of high-dimensional U-statistics."
                    },
                    {
                        "title": "Sequential Decision Making with Expert Demonstrations under Unobserved Heterogeneity",
                        "abstract": "We study the problem of online sequential decision-making given auxiliary demonstrations from experts who made their decisions based on unobserved contextual information. These demonstrations can be viewed as solving related but slightly different tasks than what the learner faces. This setting arises in many application domains, such as self-driving cars, healthcare, and finance, where expert demonstrations are made using contextual information, which is not recorded in the data available to the learning agent. We model the problem as a zero-shot meta-reinforcement learning setting with an unknown task distribution and a Bayesian regret minimization objective, where the unobserved tasks are encoded as parameters with an unknown prior. We propose the Experts-as-Priors algorithm (ExPerior), a non-parametric empirical Bayes approach that utilizes the principle of maximum entropy to establish an informative prior over the learner's decision-making problem. This prior enables the application of any Bayesian approach for online decision-making, such as posterior sampling. We demonstrate that our strategy surpasses existing behaviour cloning and online algorithms for multi-armed bandits and reinforcement learning, showcasing the utility of our approach in leveraging expert demonstrations across different decision-making setups."
                    },
                    {
                        "title": "Structure-agnostic Optimality of Doubly Robust Learning for Treatment Effect Estimation",
                        "abstract": "Average treatment effect estimation is the most central problem in causal inference with application to numerous disciplines. While many estimation strategies have been proposed in the literature, the statistical optimality of these methods has still remained an open area of investigation, especially in regimes where these methods do not achieve parametric rates. In this paper, we adopt the recently introduced structure-agnostic framework of statistical lower bounds, which poses no structural properties on the nuisance functions other than access to black-box estimators that achieve some statistical estimation rate. This framework is particularly appealing when one is only willing to consider estimation strategies that use non-parametric regression and classification oracles as black-box sub-processes. Within this framework, we prove the statistical optimality of the celebrated and widely used doubly robust estimators for both the Average Treatment Effect (ATE) and the Average Treatment Effect on the Treated (ATT), as well as weighted variants of the former, which arise in policy evaluation."
                    },
                    {
                        "title": "Inference on Optimal Dynamic Policies via Softmax Approximation",
                        "abstract": "Estimating optimal dynamic policies from offline data is a fundamental problem in dynamic decision making. In the context of causal inference, the problem is known as estimating the optimal dynamic treatment regime. Even though there exists a plethora of methods for estimation, constructing confidence intervals for the value of the optimal regime and structural parameters associated with it is inherently harder, as it involves non-linear and non-differentiable functionals of unknown quantities that need to be estimated. Prior work resorted to sub-sample approaches that can deteriorate the quality of the estimate. We show that a simple soft-max approximation to the optimal treatment regime, for an appropriately fast growing temperature parameter, can achieve valid inference on the truly optimal regime. We illustrate our result for a two-period optimal dynamic regime, though our approach should directly extend to the finite horizon case. Our work combines techniques from semi-parametric inference and $g$-estimation, together with an appropriate triangular array central limit theorem, as well as a novel analysis of the asymptotic influence and asymptotic bias of softmax approximations."
                    },
                    {
                        "title": "Incentive-Aware Synthetic Control: Accurate Counterfactual Estimation via Incentivized Exploration",
                        "abstract": "We consider the setting of synthetic control methods (SCMs), a canonical approach used to estimate the treatment effect on the treated in a panel data setting. We shed light on a frequently overlooked but ubiquitous assumption made in SCMs of\"overlap\": a treated unit can be written as some combination -- typically, convex or linear combination -- of the units that remain under control. We show that if units select their own interventions, and there is sufficiently large heterogeneity between units that prefer different interventions, overlap will not hold. We address this issue by proposing a framework which incentivizes units with different preferences to take interventions they would not normally consider. Specifically, leveraging tools from information design and online learning, we propose a SCM that incentivizes exploration in panel data settings by providing incentive-compatible intervention recommendations to units. We establish this estimator obtains valid counterfactual estimates without the need for an a priori overlap assumption. We extend our results to the setting of synthetic interventions, where the goal is to produce counterfactual outcomes under all interventions, not just control. Finally, we provide two hypothesis tests for determining whether unit overlap holds for a given panel dataset."
                    },
                    {
                        "title": "Source Condition Double Robust Inference on Functionals of Inverse Problems",
                        "abstract": "We consider estimation of parameters defined as linear functionals of solutions to linear inverse problems. Any such parameter admits a doubly robust representation that depends on the solution to a dual linear inverse problem, where the dual solution can be thought as a generalization of the inverse propensity function. We provide the first source condition double robust inference method that ensures asymptotic normality around the parameter of interest as long as either the primal or the dual inverse problem is sufficiently well-posed, without knowledge of which inverse problem is the more well-posed one. Our result is enabled by novel guarantees for iterated Tikhonov regularized adversarial estimators for linear inverse problems, over general hypothesis spaces, which are developments of independent interest."
                    },
                    {
                        "title": "Learning Causal Representations from General Environments: Identifiability and Intrinsic Ambiguity",
                        "abstract": "We study causal representation learning, the task of recovering high-level latent variables and their causal relationships in the form of a causal graph from low-level observed data (such as text and images), assuming access to observations generated from multiple environments. Prior results on the identifiability of causal representations typically assume access to single-node interventions which is rather unrealistic in practice, since the latent variables are unknown in the first place. In this work, we provide the first identifiability results based on data that stem from general environments. We show that for linear causal models, while the causal graph can be fully recovered, the latent variables are only identified up to the surrounded-node ambiguity (SNA) \\citep{varici2023score}. We provide a counterpart of our guarantee, showing that SNA is basically unavoidable in our setting. We also propose an algorithm, \\texttt{LiNGCReL} which provably recovers the ground-truth model up to SNA, and we demonstrate its effectiveness via numerical experiments. Finally, we consider general non-parametric causal models and show that the same identification barrier holds when assuming access to groups of soft single-node interventions."
                    },
                    {
                        "title": "Causal Q-Aggregation for CATE Model Selection",
                        "abstract": "Accurate estimation of conditional average treatment effects (CATE) is at the core of personalized decision making. While there is a plethora of models for CATE estimation, model selection is a nontrivial task, due to the fundamental problem of causal inference. Recent empirical work provides evidence in favor of proxy loss metrics with double robust properties and in favor of model ensembling. However, theoretical understanding is lacking. Direct application of prior theoretical work leads to suboptimal oracle model selection rates due to the non-convexity of the model selection problem. We provide regret rates for the major existing CATE ensembling approaches and propose a new CATE model ensembling approach based on Q-aggregation using the doubly robust loss. Our main result shows that causal Q-aggregation achieves statistically optimal oracle model selection regret rates of $\\frac{\\log(M)}{n}$ (with $M$ models and $n$ samples), with the addition of higher-order estimation error terms related to products of errors in the nuisance functions. Crucially, our regret rate does not require that any of the candidate CATE models be close to the truth. We validate our new method on many semi-synthetic datasets and also provide extensions of our work to CATE model selection with instrumental variables and unobserved confounding."
                    },
                    {
                        "title": "Minimax Instrumental Variable Regression and $L_2$ Convergence Guarantees without Identification or Closedness",
                        "abstract": "In this paper, we study nonparametric estimation of instrumental variable (IV) regressions. Recently, many flexible machine learning methods have been developed for instrumental variable estimation. However, these methods have at least one of the following limitations: (1) restricting the IV regression to be uniquely identified; (2) only obtaining estimation error rates in terms of pseudometrics (\\emph{e.g.,} projected norm) rather than valid metrics (\\emph{e.g.,} $L_2$ norm); or (3) imposing the so-called closedness condition that requires a certain conditional expectation operator to be sufficiently smooth. In this paper, we present the first method and analysis that can avoid all three limitations, while still permitting general function approximation. Specifically, we propose a new penalized minimax estimator that can converge to a fixed IV solution even when there are multiple solutions, and we derive a strong $L_2$ error rate for our estimator under lax conditions. Notably, this guarantee only needs a widely-used source condition and realizability assumptions, but not the so-called closedness condition. We argue that the source condition and the closedness condition are inherently conflicting, so relaxing the latter significantly improves upon the existing literature that requires both conditions. Our estimator can achieve this improvement because it builds on a novel formulation of the IV estimation problem as a constrained optimization problem."
                    }
                ]
            },
            "1c67e828-6898-4653-b6a1-969ebd02acd8": {
                "pk": "1c67e828-6898-4653-b6a1-969ebd02acd8",
                "name": "Emma Brunskill",
                "collaborators": [
                    "Allen Nie",
                    "Yash Chandak",
                    "Joy He-Yueya",
                    "Noah D. Goodman",
                    "Juliette Woodrow",
                    "Jonathan Lee",
                    "Aldo Pacchiano",
                    "Yannis Flet-Berliac",
                    "Sanath Kumar Krishnamurthy",
                    "Susan Athey",
                    "Miroslav Suzara",
                    "Malika Ali",
                    "Matt Peng",
                    "Mehran Sahami",
                    "Chris Piech",
                    "Hyunji Alex Nam",
                    "Alan Y. Cheng",
                    "Ellie Tanimura",
                    "Joseph Tey",
                    "Andrew C. Wu",
                    "Ryan Louie",
                    "Ananjan Nandi",
                    "William Fang",
                    "Cheng Chang",
                    "Diyi Yang",
                    "Amelia Leon",
                    "E. Liu",
                    "David Yuan",
                    "Ahmed Ahmed",
                    "Elyse Cornwall",
                    "Kaylee Burns",
                    "C. Piech",
                    "Chelsea Finn",
                    "Wanjing Anya Ma",
                    "Kanishk Gandhi",
                    "Benjamin W. Domingue",
                    "Yuchen Hu",
                    "Henry Zhu",
                    "Stefan Wager",
                    "Matthew J\u00f6rke",
                    "Shardul Sapkota",
                    "Lyndsea Warkenthien",
                    "Niklas Vainio",
                    "Paul Schmiedmayer",
                    "James Landay",
                    "Weihao Kong",
                    "Vidya Muthukumar",
                    "Danielle R. Thomas",
                    "Jionghao Lin",
                    "Erin Gatz",
                    "Ashish Gurung",
                    "Shivang Gupta",
                    "Kole A. Norberg",
                    "Stephen E. Fancsali",
                    "Vincent Aleven",
                    "Lee Branstetter",
                    "K. Koedinger",
                    "Anirudhan Badrinath",
                    "S. Ruan",
                    "William Steenbergen",
                    "Jiayu He",
                    "JQ Zhang",
                    "Meng Guo",
                    "Yao Liu",
                    "Kyle Dang Nguyen",
                    "Catherine Y Wang",
                    "Rui Ying",
                    "J. Landay",
                    "Kefan Dong",
                    "Aishwarya Mandyam",
                    "Matthew Joerke",
                    "Barbara E. Engelhardt"
                ],
                "domain": [
                    "Machine Learning",
                    "Educational Technology",
                    "Reinforcement Learning",
                    "Health Informatics"
                ],
                "publications": [
                    {
                        "title": "Data-driven Error Estimation: Upper Bounding Multiple Errors with No Technical Debt",
                        "abstract": "We formulate the problem of constructing multiple simultaneously valid confidence intervals (CIs) as estimating a high probability upper bound on the maximum error for a class/set of estimate-estimand-error tuples, and refer to this as the error estimation problem. For a single such tuple, data-driven confidence intervals can often be used to bound the error in our estimate. However, for a class of estimate-estimand-error tuples, nontrivial high probability upper bounds on the maximum error often require class complexity as input -- limiting the practicality of such methods and often resulting in loose bounds. Rather than deriving theoretical class complexity-based bounds, we propose a completely data-driven approach to estimate an upper bound on the maximum error. The simple and general nature of our solution to this fundamental challenge lends itself to several applications including: multiple CI construction, multiple hypothesis testing, estimating excess risk bounds (a fundamental measure of uncertainty in machine learning) for any training/fine-tuning algorithm, and enabling the development of a contextual bandit pipeline that can leverage any reward model estimation procedure as input (without additional mathematical analysis)."
                    },
                    {
                        "title": "Evaluating and Optimizing Educational Content with Large Language Model Judgments",
                        "abstract": "Creating effective educational materials generally requires expensive and time-consuming studies of student learning outcomes. To overcome this barrier, one idea is to build computational models of student learning and use them to optimize instructional materials. However, it is difficult to model the cognitive processes of learning dynamics. We propose an alternative approach that uses Language Models (LMs) as educational experts to assess the impact of various instructions on learning outcomes. Specifically, we use GPT-3.5 to evaluate the overall effect of instructional materials on different student groups and find that it can replicate well-established educational findings such as the Expertise Reversal Effect and the Variability Effect. This demonstrates the potential of LMs as reliable evaluators of educational content. Building on this insight, we introduce an instruction optimization approach in which one LM generates instructional materials using the judgments of another LM as a reward function. We apply this approach to create math word problem worksheets aimed at maximizing student learning gains. Human teachers' evaluations of these LM-generated worksheets show a significant alignment between the LM judgments and human teacher preferences. We conclude by discussing potential divergences between human and LM opinions and the resulting pitfalls of automating instructional design."
                    },
                    {
                        "title": "The GPT Surprise: Offering Large Language Model Chat in a Massive Coding Class Reduced Engagement but Increased Adopters Exam Performances",
                        "abstract": "Large language models (LLMs) are quickly being adopted in a wide range of learning experiences, especially via ubiquitous and broadly accessible chat interfaces like ChatGPT and Copilot. This type of interface is readily available to students and teachers around the world, yet relatively little research has been done to assess the impact of such generic tools on student learning. Coding education is an interesting test case, both because LLMs have strong performance on coding tasks, and because LLM-powered support tools are rapidly becoming part of the workflow of professional software engineers. To help understand the impact of generic LLM use on coding education, we conducted a large-scale randomized control trial with 5,831 students from 146 countries in an online coding class in which we provided some students with access to a chat interface with GPT-4. We estimate positive benefits on exam performance for adopters, the students who used the tool, but over all students, the advertisement of GPT-4 led to a significant average decrease in exam participation. We observe similar decreases in other forms of course engagement. However, this decrease is modulated by the student's country of origin. Offering access to LLMs to students from low human development index countries increased their exam participation rate on average. Our results suggest there may be promising benefits to using LLMs in an introductory coding class, but also potential harms for engagement, which makes their longer term impact on student success unclear. Our work highlights the need for additional investigations to help understand the potential impact of future adoption and integration of LLMs into classrooms."
                    },
                    {
                        "title": "Short-Long Policy Evaluation with Novel Actions",
                        "abstract": "From incorporating LLMs in education, to identifying new drugs and improving ways to charge batteries, innovators constantly try new strategies in search of better long-term outcomes for students, patients and consumers. One major bottleneck in this innovation cycle is the amount of time it takes to observe the downstream effects of a decision policy that incorporates new interventions. The key question is whether we can quickly evaluate long-term outcomes of a new decision policy without making long-term observations. Organizations often have access to prior data about past decision policies and their outcomes, evaluated over the full horizon of interest. Motivated by this, we introduce a new setting for short-long policy evaluation for sequential decision making tasks. Our proposed methods significantly outperform prior results on simulators of HIV treatment, kidney dialysis and battery charging. We also demonstrate that our methods can be useful for applications in AI safety by quickly identifying when a new decision policy is likely to have substantially lower performance than past policies."
                    },
                    {
                        "title": "Brief, Just-in-Time Teaching Tips to Support Computer Science Tutors",
                        "abstract": "As enrollments in computing-related programs continue to rise, computer science departments are increasingly relying on teaching assistants (TAs) to provide additional educational support to students, such as one-on-one tutoring or office hours. Tutoring is more effective with highly trained tutors, but most TAs receive little to no training in pedagogical skills. How might we provide support to TAs working with students one-on-one, especially in online settings? We propose a just-in-time intervention that shows a tutor actionable teaching tips and relevant information right before they begin an online tutoring session with a student. We conducted a crossover experiment (n = 46) where participants engaged in two tutoring roleplays for an introductory computer science programming task and found that participants demonstrated effective instructional strategies for much longer periods of time after receiving the intervention. We discuss the implications of these findings for both educators looking to support tutors and researchers seeking to build technology for tutors."
                    },
                    {
                        "title": "Roleplay-doh: Enabling Domain-Experts to Create LLM-simulated Patients via Eliciting and Adhering to Principles",
                        "abstract": "Recent works leverage LLMs to roleplay realistic social scenarios, aiding novices in practicing their social skills. However, simulating sensitive interactions, such as in mental health, is challenging. Privacy concerns restrict data access, and collecting expert feedback, although vital, is laborious. To address this, we develop Roleplay-doh, a novel human-LLM collaboration pipeline that elicits qualitative feedback from a domain-expert, which is transformed into a set of principles, or natural language rules, that govern an LLM-prompted roleplay. We apply this pipeline to enable senior mental health supporters to create customized AI patients for simulated practice partners for novice counselors. After uncovering issues in GPT-4 simulations not adhering to expert-defined principles, we also introduce a novel principle-adherence prompting pipeline which shows 30% improvements in response quality and principle following for the downstream task. Via a user study with 25 counseling experts, we demonstrate that the pipeline makes it easy and effective to create AI patients that more faithfully resemble real patients, as judged by creators and third-party counselors. See our project website at https://roleplay-doh.github.io/ for code and data."
                    },
                    {
                        "title": "Estimating the Causal Treatment Effect of Unproductive Persistence",
                        "abstract": "There has been considerable work in classifying and predicting unproductive persistence, but much less in understanding its causal impact on downstream outcomes of interest, like external assessments. In general, it is experimentally challenging to understand the causal impact because, unlike in many other settings, we cannot directly intervene (to conduct a randomized control trial) and cause students to struggle unproductively in an authentic manner. In this work, we use data from a prior study that used virtual reality headsets to alert teacher\u2019s attention to students who were unproductively struggling. We show that we can use this as an instrumental variable, and use a two-stage least squares analysis to provide a causal estimate of the treatment effect of unproductive persistence on post-test performance. Our results further strengthen the importance of unproductive struggle and highlight the potential of leveraging instruments to identify causal treatment effects of student behaviors during the use of educational technology."
                    },
                    {
                        "title": "A Fast and Accurate Machine Learning Autograder for the Breakout Assignment",
                        "abstract": "In this paper, we detail the successful deployment of a machine learning autograder that significantly decreases the grading labor required in the Breakout computer science assignment. This assignment - which tasks students with programming a game consisting of a controllable paddle and a ball that bounces off the paddle to break bricks - is popular for engaging students with introductory computer science concepts, but creates a large grading burden. Due to the game's interactive nature, grading defies traditional unit tests and instead typically requires 8+ minutes of manually playing each student's game to search for bugs. This amounts to 45+ hours of grading in a standard course offering and prevents further widespread adoption of the assignment. Our autograder alleviates this burden by playing each student's game with a reinforcement learning agent and providing videos of discovered bugs to instructors. In an A/B test with manual grading, we find that our human-in-the-loop AI autograder reduces grading time by 44%, while slightly improving grading accuracy by 6%, ultimately saving roughly 30 hours over our deployment in two offerings of the assignment. Our results further suggest the practicality of grading other interactive assignments (e.g., other games or building websites) via similar machine learning techniques. Live demo at https://ezliu.github.io/breakoutgrader."
                    },
                    {
                        "title": "Psychometric Alignment: Capturing Human Knowledge Distributions via Language Models",
                        "abstract": "Language models (LMs) are increasingly used to simulate human-like responses in scenarios where accurately mimicking a population's behavior can guide decision-making, such as in developing educational materials and designing public policies. The objective of these simulations is for LMs to capture the variations in human responses, rather than merely providing the expected correct answers. Prior work has shown that LMs often generate unrealistically accurate responses, but there are no established metrics to quantify how closely the knowledge distribution of LMs aligns with that of humans. To address this, we introduce\"psychometric alignment,\"a metric that measures the extent to which LMs reflect human knowledge distribution. Assessing this alignment involves collecting responses from both LMs and humans to the same set of test items and using Item Response Theory to analyze the differences in item functioning between the groups. We demonstrate that our metric can capture important variations in populations that traditional metrics, like differences in accuracy, fail to capture. We apply this metric to assess existing LMs for their alignment with human knowledge distributions across three real-world domains. We find significant misalignment between LMs and human populations, though using persona-based prompts can improve alignment. Interestingly, smaller LMs tend to achieve greater psychometric alignment than larger LMs. Further, training LMs on human response data from the target distribution enhances their psychometric alignment on unseen test items, but the effectiveness of such training varies across domains."
                    },
                    {
                        "title": "Minimax-Regret Sample Selection in Randomized Experiments (preprint)",
                        "abstract": "Randomized controlled trials (RCTs) are often run in settings with many subpopulations that may have differential benefits from the treatment being evaluated. We consider the problem of sample selection, i.e., whom to enroll in an RCT, such as to optimize welfare in a heterogeneous population. We formalize this problem within the minimax-regret framework, and derive optimal sample-selection schemes under a variety of conditions. We also highlight how different objectives and decisions can lead to notably different guidance regarding optimal sample allocation through a synthetic experiment leveraging historical COVID-19 trial data."
                    },
                    {
                        "title": "Supporting Physical Activity Behavior Change with LLM-Based Conversational Agents",
                        "abstract": "Physical activity has significant benefits to health, yet large portions of the population remain physically inactive. Mobile health applications show promising potential for low-cost, scalable physical activity promotion, but existing approaches are often insufficiently personalized to a user's context and life circumstances. In this work, we explore the potential for large language model (LLM) based conversational agents to motivate physical activity behavior change. Through formative interviews with 12 health professionals and 10 non-experts, we identify design considerations and opportunities for LLM health coaching. We present GPTCoach, a chatbot that implements an evidence-based health coaching program, uses counseling strategies from motivational interviewing, and can query and visualize health data from a wearable through tool use. We evaluate GPTCoach as a technology probe in a user study with 16 participants. Through quantitive and qualitative analyses, we find promising evidence that GPTCoach can adhere to a health coaching program while adopting a facilitative, supportive, and non-judgmental tone. We find more variable support for GPTCoach's ability to proactively make use of data in ways that foster motivation and empowerment. We conclude with a discussion of our findings, implications for future research, as well as risks and limitations."
                    },
                    {
                        "title": "Estimating Optimal Policy Value in Linear Contextual Bandits Beyond Gaussianity",
                        "abstract": "While much of bandit learning focuses on minimizing regret while learning an optimal policy, it is often of interest to estimate the maximum value achievable before learning the optimal policy, which can be of use as an input to down-stream tasks like model selection. Prior work in contextual bandits has considered this in the Gaussian setting. Here, we study the problem of approximating the optimal policy value in the more general linear contextual bandit problem, and we focus on whether it is possible to do so with less data than what is needed to learn the optimal policy. We consider two objectives: (1) estimating upper bounds on the value and (2) estimating the value directly. For the \ufb01rst, we present an adaptive upper bound that is at most logarithmic factor larger than the value and tight when the data is Gaussian and show that it is possible to estimate this upper bound in (cid:101) O ( \u221a d ) samples where d is the number of parameters. As a consequence of this bound, we show improved regret bounds for model selection. For the second objective, we present a moment-based algorithm for estimating the optimal policy value with sample complexity (cid:101) O ( \u221a d ) for sub-Gaussian context distributions whose low order moments are known."
                    },
                    {
                        "title": "Experiment Planning with Function Approximation",
                        "abstract": "We study the problem of experiment planning with function approximation in contextual bandit problems. In settings where there is a significant overhead to deploying adaptive algorithms -- for example, when the execution of the data collection policies is required to be distributed, or a human in the loop is needed to implement these policies -- producing in advance a set of policies for data collection is paramount. We study the setting where a large dataset of contexts but not rewards is available and may be used by the learner to design an effective data collection strategy. Although when rewards are linear this problem has been well studied, results are still missing for more complex reward models. In this work we propose two experiment planning strategies compatible with function approximation. The first is an eluder planning and sampling procedure that can recover optimality guarantees depending on the eluder dimension of the reward function class. For the second, we show that a uniform sampler achieves competitive optimality rates in the setting where the number of actions is small. We finalize our results introducing a statistical gap fleshing out the fundamental differences between planning and adaptive learning and provide results for planning with model selection."
                    },
                    {
                        "title": "Improving Student Learning with Hybrid Human-AI Tutoring: A Three-Study Quasi-Experimental Investigation",
                        "abstract": "Artificial intelligence (AI) applications to support human tutoring have potential to significantly improve learning outcomes, but engagement issues persist, especially among students from low-income backgrounds. We introduce an AI-assisted tutoring model that combines human and AI tutoring and hypothesize this synergy will have positive impacts on learning processes. To investigate this hypothesis, we conduct a three-study quasi-experiment across three urban and low-income middle schools: 1) 125 students in a Pennsylvania school; 2) 385 students (50% Latinx) in a California school, and 3) 75 students (100% Black) in a Pennsylvania charter school, all implementing analogous tutoring models. We compare learning analytics of students engaged in human-AI tutoring compared to students using math software only. We find human-AI tutoring has positive effects, particularly in student\u2019s proficiency and usage, with evidence suggesting lower achieving students may benefit more compared to higher achieving students. We illustrate the use of quasi-experimental methods adapted to the particulars of different schools and data-availability contexts so as to achieve the rapid data-driven iteration needed to guide an inspired creation into effective innovation. Future work focuses on improving the tutor dashboard and optimizing tutor-student ratios, while maintaining annual costs per student of approximately $700 annually."
                    },
                    {
                        "title": "Waypoint Transformer: Reinforcement Learning via Supervised Learning with Intermediate Targets",
                        "abstract": "Despite the recent advancements in offline reinforcement learning via supervised learning (RvS) and the success of the decision transformer (DT) architecture in various domains, DTs have fallen short in several challenging benchmarks. The root cause of this underperformance lies in their inability to seamlessly connect segments of suboptimal trajectories. To overcome this limitation, we present a novel approach to enhance RvS methods by integrating intermediate targets. We introduce the Waypoint Transformer (WT), using an architecture that builds upon the DT framework and conditioned on automatically-generated waypoints. The results show a significant increase in the final return compared to existing RvS methods, with performance on par or greater than existing state-of-the-art temporal difference learning-based methods. Additionally, the performance and stability improvements are largest in the most challenging environments and data configurations, including AntMaze Large Play/Diverse and Kitchen Mixed/Partial."
                    },
                    {
                        "title": "Reinforcement Learning Tutor Better Supported Lower Performers in a Math Task",
                        "abstract": "Resource limitations make it challenging to provide all students with one of the most effective educational interventions: personalized instruction. Reinforcement learning could be a pivotal tool to decrease the development costs and enhance the effectiveness of intelligent tutoring software, that aims to provide the right support, at the right time, to a student. Here we illustrate that deep reinforcement learning can be used to provide adaptive pedagogical support to students learning about the concept of volume in a narrative storyline software. Using explainable artificial intelligence tools, we extracted interpretable insights about the pedagogical policy learned and demonstrated that the resulting policy had similar performance in a different student population. Most importantly, in both studies, the reinforcement-learning narrative system had the largest benefit for those students with the lowest initial pretest scores, suggesting the opportunity for AI to adapt and provide support for those most in need."
                    },
                    {
                        "title": "Model-based Offline Reinforcement Learning with Local Misspecification",
                        "abstract": "We present a model-based offline reinforcement learning policy performance lower bound that explicitly captures dynamics model misspecification and distribution mismatch and we propose an empirical algorithm for optimal offline policy selection. Theoretically, we prove a novel safe policy improvement theorem by establishing pessimism approximations to the value function. Our key insight is to jointly consider selecting over dynamics models and policies: as long as a dynamics model can accurately represent the dynamics of the state-action pairs visited by a given policy, it is possible to approximate the value of that particular policy. We analyze our lower bound in the LQR setting and also show competitive performance to previous lower bounds on policy selection across a set of D4RL tasks."
                    },
                    {
                        "title": "Adaptive Interventions with User-Defined Goals for Health Behavior Change",
                        "abstract": "Promoting healthy lifestyle behaviors remains a major public health concern, particularly due to their crucial role in preventing chronic conditions such as cancer, heart disease, and type 2 diabetes. Mobile health applications present a promising avenue for low-cost, scalable health behavior change promotion. Researchers are increasingly exploring adaptive algorithms that personalize interventions to each person's unique context. However, in empirical studies, mobile health applications often suffer from small effect sizes and low adherence rates, particularly in comparison to human coaching. Tailoring advice to a person's unique goals, preferences, and life circumstances is a critical component of health coaching that has been underutilized in adaptive algorithms for mobile health interventions. To address this, we introduce a new Thompson sampling algorithm that can accommodate personalized reward functions (i.e., goals, preferences, and constraints), while also leveraging data sharing across individuals to more quickly be able to provide effective recommendations. We prove that our modification incurs only a constant penalty on cumulative regret while preserving the sample complexity benefits of data sharing. We present empirical results on synthetic and semi-synthetic physical activity simulators, where in the latter we conducted an online survey to solicit preference data relating to physical activity, which we use to construct realistic reward models that leverages historical data from another study. Our algorithm achieves substantial performance improvements compared to baselines that do not share data or do not optimize for individualized rewards."
                    }
                ]
            }
        }
    },
    "2409.19659": {
        "paper_data": {
            "title": "Flipped Classroom: Aligning Teacher Attention with Student in Generalized Category Discovery",
            "url": "http://arxiv.org/abs/2409.19659v1",
            "arxiv_id": "2409.19659",
            "authors": [
                "Haonan Lin",
                "Wenbin An",
                "Jiahao Wang",
                "Yan Chen",
                "Feng Tian",
                "Mengmeng Wang",
                "Guang Dai",
                "Qianying Wang",
                "Jingdong Wang"
            ],
            "abstract": "Recent advancements have shown promise in applying traditional Semi-Supervised Learning strategies to the task of Generalized Category Discovery (GCD). Typically, this involves a teacher-student framework in which the teacher imparts knowledge to the student to classify categories, even in the absence of explicit labels. Nevertheless, GCD presents unique challenges, particularly the absence of priors for new classes, which can lead to the teacher's misguidance and unsynchronized learning with the student, culminating in suboptimal outcomes. In our work, we delve into why traditional teacher-student designs falter in open-world generalized category discovery as compared to their success in closed-world semi-supervised learning. We identify inconsistent pattern learning across attention layers as the crux of this issue and introduce FlipClass, a method that dynamically updates the teacher to align with the student's attention, instead of maintaining a static teacher reference. Our teacher-student attention alignment strategy refines the teacher's focus based on student feedback from an energy perspective, promoting consistent pattern recognition and synchronized learning across old and new classes. Extensive experiments on a spectrum of benchmarks affirm that FlipClass significantly surpasses contemporary GCD methods, establishing new standards for the field.",
            "introduction": "   1 Introduction  Teacher-Student architecture has proved its effectiveness in Semi-Supervised Learning (SSL) [65, 32, 52], which aims to take advantage of a large collection of unlabeled data, reducing the expensive costs of annotation [56, 89, 88]. Previous approaches tend to model p\u2062(student|teacher)\ud835\udc5dconditionalstudentteacherp(\\text{student}|\\text{teacher})italic_p ( student | teacher ), where the teacher typically acts as a fixed point of reference, providing a form of \u201csupervision prior\" to guide the student [25, 41, 59]. This supervision comes from labeled data and is asymmetrical: while the teacher has robust prior knowledge and provides a stable learning signal, the student\u2019s knowledge is incomplete and evolving. The student learns from both the teacher\u2019s supervision and the data it is exposed to, trying to emulate the teacher by aligning its predictions with those of the teacher.   Teacher-Student designs traditionally rely on a closed-world assumption, where it is expected that the teachers have supervision prior to all classes they will face while instructing students [60, 3]. However, real-world applications often involve dynamic and open environments, where instances belonging to new classes may appear [83, 30, 31, 40]. In such cases, discovering novelties could enable models to adapt new information and evolve continually as biological systems do [14, 48, 39]. Recently, Generalized Category Discovery (GCD) [68] stands out by challenging models to categorize unlabeled data containing both old and new classes using only partial labels for training. Although recent GCD methods have adapted closed-world Teacher-Student strategies with notable success [74, 49], the transition is not seamless and presents several challenges.   Challenge I: Learning gap. Fig.\u00a01 top left illustrates the learning evolution of student and teacher. The previous teacher-student models result in unsynchronized learning and a significant learning gap in new classes. The ideal learning dynamic between the teacher and student should be cohesive, which requires teaching students in accordance with their aptitude.   Challenge II: Discrepancies in features. The learning gap arises from the teacher\u2019s fast pace, leading to large discrepancies in representations between weakly-augmented data (teacher) and strongly-augmented data (student), especially for new classes (Fig.\u00a01 middle). This causes significant prediction differences, making consistency loss optimization difficult and hindering effective student learning. Over time, the iterative learning process exacerbates this misalignment.   Challenge III: Attention inconsistency. Inadequate supervision for new classes leads to inaccurate instructions from the teacher, causing the student to focus on different parts than the teacher (Fig.\u00a01 right). Imagine a classroom transitioning to a new subject. Without proper guidance, students\u2019 attention diverges from the teacher\u2019s, resulting in confusion and ineffective learning.   Figure 1: Left: Learning effects of traditional Teacher-Student Consistency Model (TSCM, e.g., SimGCD [74]) and our Flipped Classroom Consistency Model (FlipClass) on Stanford Cars [33]. Middle: Model comparison between TSCM and our FlipClass, where \ud835\udd3bnewsuperscript\ud835\udd3bnew\\mathbb{D}^{\\text{new}}blackboard_D start_POSTSUPERSCRIPT new end_POSTSUPERSCRIPT refers to data belonging to new classes. Right: Illustration of the inner feedback mechanism in FlipClass, where teacher attention is adapted to the student, leading to the alignment of attention.   To sum up, the challenges of previous teacher-student models arise from inadequate supervision on new classes and the gap between weakly and strongly augmented data. This results in attention inconsistency (Chall. III), which leads to discrepancies in predictions and representations (Chall. II), ultimately",
            "references": [
                {
                    "title": "Contrastive Mean-Shift Learning for Generalized Category Discovery",
                    "abstract": "We address the problem of generalized category discovery (GCD) that aims to partition a partially labeled collection of images; only a small part of the collection is labeled and the total number of target classes is unknown. To address this generalized image clustering problem, we revisit the mean-shift algorithm, i.e., a classic, powerful technique for mode seeking, and incorporate it into a contrastive learning framework. The proposed method, dubbed Contrastive Mean-Shift (CMS) learning, trains an embedding network to produce representations with better clustering properties by an iterative process of mean shift and contrastive update. Experiments demonstrate that our method, both in settings with and without the total number of clusters being known, achieves state-of-the-art performance on six public GCD benchmarks without bells and whistles."
                },
                {
                    "title": "Beyond Known Clusters: Probe New Prototypes for Efficient Generalized Class Discovery",
                    "abstract": "Generalized Class Discovery (GCD) aims to dynamically assign labels to unlabelled data partially based on knowledge learned from labelled data, where the unlabelled data may come from known or novel classes. The prevailing approach generally involves clustering across all data and learning conceptions by prototypical contrastive learning. However, existing methods largely hinge on the performance of clustering algorithms and are thus subject to their inherent limitations. Firstly, the estimated cluster number is often smaller than the ground truth, making the existing methods suffer from the lack of prototypes for comprehensive conception learning. To address this issue, we propose an adaptive probing mechanism that introduces learnable potential prototypes to expand cluster prototypes (centers). As there is no ground truth for the potential prototype, we develop a self-supervised prototype learning framework to optimize the potential prototype in an end-to-end fashion. Secondly, clustering is computationally intensive, and the conventional strategy of clustering both labelled and unlabelled instances exacerbates this issue. To counteract this inefficiency, we opt to cluster only the unlabelled instances and subsequently expand the cluster prototypes with our introduced potential prototypes to fast explore novel classes. Despite the simplicity of our proposed method, extensive empirical analysis on a wide range of datasets confirms that our method consistently delivers state-of-the-art results. Specifically, our method surpasses the nearest competitor by a significant margin of 9.7% within the Stanford Cars dataset and 12x clustering efficiency within the Herbarium 19 dataset. We will make the code and checkpoints publicly available at https://github.com/xjtuYW/PNP.git."
                },
                {
                    "title": "SPTNet: An Efficient Alternative Framework for Generalized Category Discovery with Spatial Prompt Tuning",
                    "abstract": "Generalized Category Discovery (GCD) aims to classify unlabelled images from both `seen' and `unseen' classes by transferring knowledge from a set of labelled `seen' class images. A key theme in existing GCD approaches is adapting large-scale pre-trained models for the GCD task. An alternate perspective, however, is to adapt the data representation itself for better alignment with the pre-trained model. As such, in this paper, we introduce a two-stage adaptation approach termed SPTNet, which iteratively optimizes model parameters (i.e., model-finetuning) and data parameters (i.e., prompt learning). Furthermore, we propose a novel spatial prompt tuning method (SPT) which considers the spatial property of image data, enabling the method to better focus on object parts, which can transfer between seen and unseen classes. We thoroughly evaluate our SPTNet on standard benchmarks and demonstrate that our method outperforms existing GCD methods. Notably, we find our method achieves an average accuracy of 61.4% on the SSB, surpassing prior state-of-the-art methods by approximately 10%. The improvement is particularly remarkable as our method yields extra parameters amounting to only 0.117% of those in the backbone architecture. Project page: https://visual-ai.github.io/sptnet."
                },
                {
                    "title": "Active Generalized Category Discovery",
                    "abstract": "Generalized Category Discovery (GCD) is a pragmatic and challenging open-world task, which endeavors to cluster unlabeled samples from both novel and old classes, leveraging some labeled data of old classes. Given that knowledge learned from old classes is not fully transferable to new classes, and that novel categories are fully unlabeled, GCD inherently faces intractable problems, including imbalanced classification performance and inconsistent confidence between old and new classes, especially in the low-labeling regime. Hence, some annotations of new classes are deemed necessary. However, labeling new classes is extremely costly. To address this issue, we take the spirit of active learning and propose a new setting called Active Generalized Category Discovery (AGCD). The goal is to improve the performance of GCD by actively selecting a limited amount of valuable samples for labeling from the oracle. To solve this problem, we devise an adaptive sampling strategy, which jointly considers novelty, informativeness and diversity to adaptively select novel samples with proper uncertainty. However, owing to the varied orderings of label indices caused by the clustering of novel classes, the queried labels are not directly applicable to subsequent training. To overcome this issue, we further propose a stable label mapping algorithm that transforms ground truth labels to the label space of the classifier, thereby ensuring consistent training across different active selection stages. Our method achieves state-of-the-art performance on both generic and fine-grained datasets. Our code is available at https://github.com/mashijie1028/ActiveGCD"
                },
                {
                    "title": "Guided Cluster Aggregation: A Hierarchical Approach to Generalized Category Discovery",
                    "abstract": "Despite advances in image recognition, recognizing novel categories in unlabeled data remains challenging for machine learning methods, even though humans can perform this task with ease. A recently developed setting to tackle this problem is Generalized Category Discovery (GCD), in which the task is to, given a labeled dataset, classify an unlabeled dataset, where the unlabeled dataset contains both known classes and novel classes that do not appear in the labeled data. Existing GCD methods mostly focus on learning strong image representations, on which they then apply a clustering algorithm such as k-means. Despite obtaining good performance, they do not fully exploit the potential of the learned features due to the simple nature of the clustering mechanism. To address this issue, we make use of the fact that local neighborhoods in self-supervised feature spaces are highly homogeneous. We leverage this observation to develop Guided Cluster Aggregation (GCA), a hierarchical approach that first groups the data into small clusters of high purity, then aggregates them into larger clusters. Experiments show that GCA outperforms semi-supervised k-means in most cases, especially in fine-grained classification tasks. Code available at https://github.com/J-L-O/guidedcluster-aggregation."
                },
                {
                    "title": "AMEND: Adaptive Margin and Expanded Neighborhood for Efficient Generalized Category Discovery",
                    "abstract": "Generalized Category Discovery aims to discover and cluster images from previously unseen classes, in addition to classifying images from seen classes correctly. In this work, we propose a simple, yet effective framework for this task, which not only performs on-par or better with the current approaches but is also significantly more efficient in terms of computational requirements. Our first contribution is to use expanded neighborhood information in contrastive learning to generate robust and generalizable features. To generate more discriminative feature representations, especially for fine-grained datasets and confusing classes, we propose a class-wise adaptive margin regularizer that aims at increasing the angular separation among the prototypes of all classes. Extensive experiments on three generic as well as four fine-grained benchmark datasets show the usefulness of the proposed Adaptive Margin and Expanded Neighborhood (AMEND) framework."
                },
                {
                    "title": "Transfer and Alignment Network for Generalized Category Discovery",
                    "abstract": "Generalized Category Discovery (GCD) is a crucial real-world task that aims to recognize both known and novel categories from an unlabeled dataset by leveraging another labeled dataset with only known categories. Despite the improved performance on known categories, current methods perform poorly on novel categories. We attribute the poor performance to two reasons: biased knowledge transfer between labeled and unlabeled data and noisy representation learning on the unlabeled data. The former leads to unreliable estimation of learning targets for novel categories and the latter hinders models from learning discriminative features. To mitigate these two issues, we propose a Transfer and Alignment Network (TAN), which incorporates two knowledge transfer mechanisms to calibrate the biased knowledge and two feature alignment mechanisms to learn discriminative features.\nSpecifically, we model different categories with prototypes and transfer the prototypes in labeled data to correct model bias towards known categories. On the one hand, we pull instances with known categories in unlabeled data closer to these prototypes to form more compact clusters and avoid boundary overlap between known and novel categories. On the other hand, we use these prototypes to calibrate noisy prototypes estimated from unlabeled data based on category similarities, which allows for more accurate estimation of prototypes for novel categories that can be used as reliable learning targets later. After knowledge transfer, we further propose two feature alignment mechanisms to acquire both instance- and category-level knowledge from unlabeled data by aligning instance features with both augmented features and the calibrated prototypes, which can boost model performance on both known and novel categories with less noise. Experiments on three benchmark datasets show that our model outperforms SOTA methods, especially on novel categories. Theoretical analysis is provided for an in-depth understanding of our model in general.\nOur code and data are available at https://github.com/Lackel/TAN."
                },
                {
                    "title": "Generalized Category Discovery with Large Language Models in the Loop",
                    "abstract": "Generalized Category Discovery (GCD) is a crucial task that aims to recognize both known and novel categories from a set of unlabeled data by utilizing a few labeled data with only known categories. Due to the lack of supervision and category information, current methods usually perform poorly on novel categories and struggle to reveal semantic meanings of the discovered clusters, which limits their applications in the real world. To mitigate the above issues, we propose Loop, an end-to-end active-learning framework that introduces Large Language Models (LLMs) into the training loop, which can boost model performance and generate category names without relying on any human efforts. Specifically, we first propose Local Inconsistent Sampling (LIS) to select samples that have a higher probability of falling to wrong clusters, based on neighborhood prediction consistency and entropy of cluster assignment probabilities. Then we propose a Scalable Query strategy to allow LLMs to choose true neighbors of the selected samples from multiple candidate samples. Based on the feedback from LLMs, we perform Refined Neighborhood Contrastive Learning (RNCL) to pull samples and their neighbors closer to learn clustering-friendly representations. Finally, we select representative samples from clusters corresponding to novel categories to allow LLMs to generate category names for them. Extensive experiments on three benchmark datasets show that Loop outperforms SOTA models by a large margin and generates accurate category names for the discovered clusters. Code and data are available at https://github.com/Lackel/LOOP."
                },
                {
                    "title": "No Representation Rules Them All in Category Discovery",
                    "abstract": "In this paper we tackle the problem of Generalized Category Discovery (GCD). Specifically, given a dataset with labelled and unlabelled images, the task is to cluster all images in the unlabelled subset, whether or not they belong to the labelled categories. Our first contribution is to recognize that most existing GCD benchmarks only contain labels for a single clustering of the data, making it difficult to ascertain whether models are using the available labels to solve the GCD task, or simply solving an unsupervised clustering problem. As such, we present a synthetic dataset, named 'Clevr-4', for category discovery. Clevr-4 contains four equally valid partitions of the data, i.e based on object shape, texture, color or count. To solve the task, models are required to extrapolate the taxonomy specified by the labelled set, rather than simply latching onto a single natural grouping of the data. We use this dataset to demonstrate the limitations of unsupervised clustering in the GCD setting, showing that even very strong unsupervised models fail on Clevr-4. We further use Clevr-4 to examine the weaknesses of existing GCD algorithms, and propose a new method which addresses these shortcomings, leveraging consistent findings from the representation learning literature to do so. Our simple solution, which is based on 'mean teachers' and termed $\\mu$GCD, substantially outperforms implemented baselines on Clevr-4. Finally, when we transfer these findings to real data on the challenging Semantic Shift Benchmark (SSB), we find that $\\mu$GCD outperforms all prior work, setting a new state-of-the-art. For the project webpage, see https://www.robots.ox.ac.uk/~vgg/data/clevr4/"
                },
                {
                    "title": "Learn to Categorize or Categorize to Learn? Self-Coding for Generalized Category Discovery",
                    "abstract": "In the quest for unveiling novel categories at test time, we confront the inherent limitations of traditional supervised recognition models that are restricted by a predefined category set. While strides have been made in the realms of self-supervised and open-world learning towards test-time category discovery, a crucial yet often overlooked question persists: what exactly delineates a \\textit{category}? In this paper, we conceptualize a \\textit{category} through the lens of optimization, viewing it as an optimal solution to a well-defined problem. Harnessing this unique conceptualization, we propose a novel, efficient and self-supervised method capable of discovering previously unknown categories at test time. A salient feature of our approach is the assignment of minimum length category codes to individual data instances, which encapsulates the implicit category hierarchy prevalent in real-world datasets. This mechanism affords us enhanced control over category granularity, thereby equipping our model to handle fine-grained categories adeptly. Experimental evaluations, bolstered by state-of-the-art benchmark comparisons, testify to the efficacy of our solution in managing unknown categories at test time. Furthermore, we fortify our proposition with a theoretical foundation, providing proof of its optimality. Our code is available at: \\url{https://github.com/SarahRastegar/InfoSieve}."
                },
                {
                    "title": "Towards Distribution-Agnostic Generalized Category Discovery",
                    "abstract": "Data imbalance and open-ended distribution are two intrinsic characteristics of the real visual world. Though encouraging progress has been made in tackling each challenge separately, few works dedicated to combining them towards real-world scenarios. While several previous works have focused on classifying close-set samples and detecting open-set samples during testing, it's still essential to be able to classify unknown subjects as human beings. In this paper, we formally define a more realistic task as distribution-agnostic generalized category discovery (DA-GCD): generating fine-grained predictions for both close- and open-set classes in a long-tailed open-world setting. To tackle the challenging problem, we propose a Self-Balanced Co-Advice contrastive framework (BaCon), which consists of a contrastive-learning branch and a pseudo-labeling branch, working collaboratively to provide interactive supervision to resolve the DA-GCD task. In particular, the contrastive-learning branch provides reliable distribution estimation to regularize the predictions of the pseudo-labeling branch, which in turn guides contrastive learning through self-balanced knowledge transfer and a proposed novel contrastive loss. We compare BaCon with state-of-the-art methods from two closely related fields: imbalanced semi-supervised learning and generalized category discovery. The effectiveness of BaCon is demonstrated with superior performance over all baselines and comprehensive analysis across various datasets. Our code is publicly available."
                },
                {
                    "title": "CiPR: An Efficient Framework with Cross-instance Positive Relations for Generalized Category Discovery",
                    "abstract": "We tackle the issue of generalized category discovery (GCD). GCD considers the open-world problem of automatically clustering a partially labelled dataset, in which the unlabelled data may contain instances from both novel categories and labelled classes. In this paper, we address the GCD problem with an unknown category number for the unlabelled data. We propose a framework, named CiPR, to bootstrap the representation by exploiting Cross-instance Positive Relations in the partially labelled data for contrastive learning, which have been neglected in existing methods. To obtain reliable cross-instance relations to facilitate representation learning, we introduce a semi-supervised hierarchical clustering algorithm, named selective neighbor clustering (SNC), which can produce a clustering hierarchy directly from the connected components of a graph constructed from selective neighbors. We further present a method to estimate the unknown class number using SNC with a joint reference score that considers clustering indexes of both labelled and unlabelled data, and extend SNC to allow label assignment for the unlabelled instances with a given class number. We thoroughly evaluate our framework on public generic image recognition datasets and challenging fine-grained datasets, and establish a new state-of-the-art. Code: https://github.com/haoosz/CiPR"
                },
                {
                    "title": "DINOv2: Learning Robust Visual Features without Supervision",
                    "abstract": "The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels."
                },
                {
                    "title": "Random Teachers are Good Teachers",
                    "abstract": "In this work, we investigate the implicit regularization induced by teacher-student learning dynamics in self-distillation. To isolate its effect, we describe a simple experiment where we consider teachers at random initialization instead of trained teachers. Surprisingly, when distilling a student into such a random teacher, we observe that the resulting model and its representations already possess very interesting characteristics; (1) we observe a strong improvement of the distilled student over its teacher in terms of probing accuracy. (2) The learned representations are data-dependent and transferable between different tasks but deteriorate strongly if trained on random inputs. (3) The student checkpoint contains sparse subnetworks, so-called lottery tickets, and lies on the border of linear basins in the supervised loss landscape. These observations have interesting consequences for several important areas in machine learning: (1) Self-distillation can work solely based on the implicit regularization present in the gradient dynamics without relying on any dark knowledge, (2) self-supervised learning can learn features even in the absence of data augmentation and (3) training dynamics during the early phase of supervised training do not necessarily require label information. Finally, we shed light on an intriguing local property of the loss landscape: the process of feature learning is strongly amplified if the student is initialized closely to the teacher. These results raise interesting questions about the nature of the landscape that have remained unexplored so far. Code is available at https://github.com/safelix/dinopl."
                },
                {
                    "title": "PromptCAL: Contrastive Affinity Learning via Auxiliary Prompts for Generalized Novel Category Discovery",
                    "abstract": "Although existing semi-supervised learning models achieve remarkable success in learning with unannotated in-distribution data, they mostly fail to learn on unlabeled data sampled from novel semantic classes due to their closed-set assumption. In this work, we target a pragmatic but under-explored Generalized Novel Category Discovery (GNCD) setting. The GNCD setting aims to categorize unlabeled training data coming from known and novel classes by leveraging the information of partially labeled known classes. We propose a two-stage Contrastive Affinity Learning method with auxiliary visual Prompts, dubbed PromptCAL, to address this challenging problem. Our approach discovers reliable pairwise sample affinities to learn better semantic clustering of both known and novel classes for the class token and visual prompts. First, we propose a discriminative prompt regularization loss to reinforce semantic discriminativeness of prompt-adapted pre-trained vision transformer for refined affinity relationships. Besides, we propose contrastive affinity learning to calibrate semantic representations based on our iterative semi-supervised affinity graph generation method for semantically-enhanced supervision. Extensive experimental evaluation demonstrates that our PromptCAL method is more effective in discovering novel classes even with limited annotations and surpasses the current state-of-the-art on generic and fine-grained benchmarks (e.g., with nearly 11% gain on CUB-200, and 9% on ImageNet-100) on overall accuracy. Our code is available at https://github.com/sheng-eatamath/PromptCAL."
                },
                {
                    "title": "Generalized Category Discovery with Decoupled Prototypical Network",
                    "abstract": "Generalized Category Discovery (GCD) aims to recognize both known and novel categories from a set of unlabeled data, based on another dataset labeled with only known categories. Without considering differences between known and novel categories, current methods learn about them in a coupled manner, which can hurt model's generalization and discriminative ability. Furthermore, the coupled training approach prevents these models transferring category-specific knowledge explicitly from labeled data to unlabeled data, which can lose high-level semantic information and impair model performance. To mitigate above limitations, we present a novel model called Decoupled Prototypical Network (DPN). By formulating a bipartite matching problem for category prototypes, DPN can not only decouple known and novel categories to achieve different training targets effectively, but also align known categories in labeled and unlabeled data to transfer category-specific knowledge explicitly and capture high-level semantics. Furthermore, DPN can learn more discriminative features for both known and novel categories through our proposed Semantic-aware Prototypical Learning (SPL). Besides capturing meaningful semantic information, SPL can also alleviate the noise of hard pseudo labels through semantic-weighted soft assignment. Extensive experiments show that DPN outperforms state-of-the-art models by a large margin on all evaluation metrics across multiple benchmark datasets. Code and data are available at https://github.com/Lackel/DPN."
                },
                {
                    "title": "Parametric Classification for Generalized Category Discovery: A Baseline Study",
                    "abstract": "Generalized Category Discovery (GCD) aims to discover novel categories in unlabelled datasets using knowledge learned from labelled samples. Previous studies argued that parametric classifiers are prone to overfitting to seen categories, and endorsed using a non-parametric classifier formed with semi-supervised k-means. However, in this study, we investigate the failure of parametric classifiers, verify the effectiveness of previous design choices when high-quality supervision is available, and identify unreliable pseudo-labels as a key problem. We demonstrate that two prediction biases exist: the classifier tends to predict seen classes more often, and produces an imbalanced distribution across seen and novel categories. Based on these findings, we propose a simple yet effective parametric classification method that benefits from entropy regularisation, achieves state-of-the-art performance on multiple GCD benchmarks and shows strong robustness to unknown class numbers. We hope the investigation and proposed simple framework can serve as a strong baseline to facilitate future studies in this field. Our code is available at: https://github.com/CVMI-Lab/SimGCD."
                },
                {
                    "title": "Modeling Inter-Class and Intra-Class Constraints in Novel Class Discovery",
                    "abstract": "Novel class discovery (NCD) aims at learning a model that transfers the common knowledge from a class-disjoint labelled dataset to another unlabelled dataset and discovers new classes (clusters) within it. Many methods, as well as elaborate training pipelines and appropriate objectives, have been proposed and considerably boosted performance on NCD tasks. Despite all this, we find that the existing methods do not sufficiently take advantage of the essence of the NCD setting. To this end, in this paper, we propose to model both inter-class and intra-class constraints in NCD based on the symmetric Kullback-Leibler divergence (sKLD). Specifically, we propose an inter-class sKLD constraint to effectively exploit the disjoint relationship between labelled and unlabelled classes, enforcing the separability for different classes in the embedding space. In addition, we present an intra-class sKLD constraint to explicitly constrain the intra-relationship between a sample and its augmentations and ensure the stability of the training process at the same time. We conduct extensive experiments on the popular CIFAR10, CIFAR100 and ImageNet benchmarks and successfully demonstrate that our method can establish a new state of the art and can achieve significant performance improvements, e.g., 3.5%/3.7% clustering accuracy improvements on CIFAR100-50 dataset split under the task-aware/-agnostic evaluation protocol, over previous state-of-the-art methods. Code is available at https://github.com/FanZhichen/NCD-IIC."
                },
                {
                    "title": "OpenCon: Open-world Contrastive Learning",
                    "abstract": "Machine learning models deployed in the wild naturally encounter unlabeled samples from both known and novel classes. Challenges arise in learning from both the labeled and unlabeled data, in an open-world semi-supervised manner. In this paper, we introduce a new learning framework, open-world contrastive learning (OpenCon). OpenCon tackles the challenges of learning compact representations for both known and novel classes and facilitates novelty discovery along the way. We demonstrate the effectiveness of OpenCon on challenging benchmark datasets and establish competitive performance. On the ImageNet dataset, OpenCon significantly outperforms the current best method by 11.9% and 7.4% on novel and overall classification accuracy, respectively. Theoretically, OpenCon can be rigorously interpreted from an EM algorithm perspective--minimizing our contrastive loss partially maximizes the likelihood by clustering similar samples in the embedding space. The code is available at https://github.com/deeplearning-wisc/opencon."
                },
                {
                    "title": "XCon: Learning with Experts for Fine-grained Category Discovery",
                    "abstract": "We address the problem of generalized category discovery (GCD) in this paper, i.e. clustering the unlabeled images leveraging the information from a set of seen classes, where the unlabeled images could contain both seen classes and unseen classes. The seen classes can be seen as an implicit criterion of classes, which makes this setting different from unsupervised clustering where the cluster criteria may be ambiguous. We mainly concern the problem of discovering categories within a fine-grained dataset since it is one of the most direct applications of category discovery, i.e. helping experts discover novel concepts within an unlabeled dataset using the implicit criterion set forth by the seen classes. State-of-the-art methods for generalized category discovery leverage contrastive learning to learn the representations, but the large inter-class similarity and intra-class variance pose a challenge for the methods because the negative examples may contain irrelevant cues for recognizing a category so the algorithms may converge to a local-minima. We present a novel method called Expert-Contrastive Learning (XCon) to help the model to mine useful information from the images by first partitioning the dataset into sub-datasets using k-means clustering and then performing contrastive learning on each of the sub-datasets to learn fine-grained discriminative features. Experiments on fine-grained datasets show a clear improved performance over the previous best methods, indicating the effectiveness of our method."
                },
                {
                    "title": "Towards Realistic Semi-Supervised Learning",
                    "abstract": "Deep learning is pushing the state-of-the-art in many computer vision applications. However, it relies on large annotated data repositories, and capturing the unconstrained nature of the real-world data is yet to be solved. Semi-supervised learning (SSL) complements the annotated training data with a large corpus of unlabeled data to reduce annotation cost. The standard SSL approach assumes unlabeled data are from the same distribution as annotated data. Recently, a more realistic SSL problem, called open-world SSL, is introduced, where the unannotated data might contain samples from unknown classes. In this paper, we propose a novel pseudo-label based approach to tackle SSL in open-world setting. At the core of our method, we utilize sample uncertainty and incorporate prior knowledge about class distribution to generate reliable class-distribution-aware pseudo-labels for unlabeled data belonging to both known and unknown classes. Our extensive experimentation showcases the effectiveness of our approach on several benchmark datasets, where it substantially outperforms the existing state-of-the-art on seven diverse datasets including CIFAR-100 (~17%), ImageNet-100 (~5%), and Tiny ImageNet (~9%). We also highlight the flexibility of our approach in solving novel class discovery task, demonstrate its stability in dealing with imbalanced data, and complement our approach with a technique to estimate the number of novel classes"
                },
                {
                    "title": "Self-Labeling Framework for Novel Category Discovery over Domains",
                    "abstract": "Unsupervised domain adaptation (UDA) has been highly successful in transferring knowledge acquired from a label-rich source domain to a label-scarce target domain. Open-set domain adaptation (open-set DA) and universal domain adaptation (UniDA) have been proposed as solutions to the problem concerning the presence of additional novel categories in the target domain. Existing open-set DA and UniDA approaches treat all novel categories as one unified unknown class and attempt to detect this unknown class during the training process. However, the features of the novel categories learned by these methods are not discriminative. This limits the applicability of UDA in the further classification of these novel categories into their original categories, rather than assigning them to a single unified class. In this paper, we propose a self-labeling framework to cluster all target samples, including those in the ''unknown'' categories. We train the network to learn the representations of target samples via self-supervised learning (SSL) and to identify the seen and unseen (novel) target-sample categories simultaneously by maximizing the mutual information between labels and input data. We evaluated our approach under different DA settings and concluded that our method generally outperformed existing ones by a wide margin."
                },
                {
                    "title": "Masked Siamese Networks for Label-Efficient Learning",
                    "abstract": "We propose Masked Siamese Networks (MSN), a self-supervised learning framework for learning image representations. Our approach matches the representation of an image view containing randomly masked patches to the representation of the original unmasked image. This self-supervised pre-training strategy is particularly scalable when applied to Vision Transformers since only the unmasked patches are processed by the network. As a result, MSNs improve the scalability of joint-embedding architectures, while producing representations of a high semantic level that perform competitively on low-shot image classification. For instance, on ImageNet-1K, with only 5,000 annotated images, our base MSN model achieves 72.4% top-1 accuracy, and with 1% of ImageNet-1K labels, we achieve 75.7% top-1 accuracy, setting a new state-of-the-art for self-supervised learning on this benchmark. Our code is publicly available."
                },
                {
                    "title": "SimMatch: Semi-supervised Learning with Similarity Matching",
                    "abstract": "Learning with few labeled data has been a longstanding problem in the computer vision and machine learning research community. In this paper, we introduced a new semi-supervised learning framework, SimMatch, which simulta-neously considers semantic similarity and instance similarity. In SimMatch, the consistency regularization will be applied on both semantic-level and instance-level. The different augmented views of the same instance are encouraged to have the same class prediction and similar similarity re-lationship respected to other instances. Next, we instanti-ated a labeled memory buffer to fully leverage the ground truth labels on instance-level and bridge the gaps between the semantic and instance similarities. Finally, we proposed the unfolding and aggregation operation which allows these two similarities be isomorphically transformed with each other. In this way, the semantic and instance pseudo-labels can be mutually propagated to generate more high-quality and reliable matching targets. Extensive ex-perimental results demonstrate that SimMatch improves the performance of semi-supervised learning tasks across dif-ferent benchmark datasets and different settings. Notably, with 400 epochs of training, SimMatch achieves 67.2%, and 74.4% Top-1 Accuracy with 1% and 10% labeled examples on ImageNet, which significantly outperforms the baseline methods and is better than previous semi-supervised learning frameworks."
                },
                {
                    "title": "Generalized Category Discovery",
                    "abstract": "In this paper, we consider a highly general image recognition setting wherein, given a labelled and unlabelled set of images, the task is to categorize all images in the unlabelled set. Here, the unlabelled images may come from labelled classes or from novel ones. Existing recognition methods are not able to deal with this setting, because they make several restrictive assumptions, such as the unlabelled instances only coming from known \u2013 or unknown \u2013 classes, and the number of unknown classes being known a-priori. We address the more unconstrained setting, naming it \u2018Generalized Category Discovery\u2019, and challenge all these assumptions. We first establish strong baselines by taking state-of-the-art algorithms from novel category discovery and adapting them for this task. Next, we propose the use of vision transformers with contrastive representation learning for this open-world setting. We then introduce a simple yet effective semi-supervised k-means method to cluster the unlabelled data into seen and unseen classes automatically, substantially outperforming the baselines. Finally, we also propose a new approach to estimate the number of classes in the unlabelled data. We thoroughly evaluate our approach on public datasets for generic object classification and on fine-grained datasets, leveraging the recent Semantic Shift Benchmark suite. Code: https://www.robots.ox.ac.uk/~vgg/research/gcd"
                },
                {
                    "title": "FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labeling",
                    "abstract": "The recently proposed FixMatch achieved state-of-the-art results on most semi-supervised learning (SSL) benchmarks. However, like other modern SSL algorithms, FixMatch uses a pre-defined constant threshold for all classes to select unlabeled data that contribute to the training, thus failing to consider different learning status and learning difficulties of different classes. To address this issue, we propose Curriculum Pseudo Labeling (CPL), a curriculum learning approach to leverage unlabeled data according to the model's learning status. The core of CPL is to flexibly adjust thresholds for different classes at each time step to let pass informative unlabeled data and their pseudo labels. CPL does not introduce additional parameters or computations (forward or backward propagation). We apply CPL to FixMatch and call our improved algorithm FlexMatch. FlexMatch achieves state-of-the-art performance on a variety of SSL benchmarks, with especially strong performances when the labeled data are extremely limited or when the task is challenging. For example, FlexMatch achieves 13.96% and 18.96% error rate reduction over FixMatch on CIFAR-100 and STL-10 datasets respectively, when there are only 4 labels per class. CPL also significantly boosts the convergence speed, e.g., FlexMatch can use only 1/5 training time of FixMatch to achieve even better performance. Furthermore, we show that CPL can be easily adapted to other SSL algorithms and remarkably improve their performances. We open-source our code at https://github.com/TorchSSL/TorchSSL."
                },
                {
                    "title": "Open-Set Recognition: A Good Closed-Set Classifier is All You Need",
                    "abstract": "The ability to identify whether or not a test sample belongs to one of the semantic classes in a classifier's training set is critical to practical deployment of the model. This task is termed open-set recognition (OSR) and has received significant attention in recent years. In this paper, we first demonstrate that the ability of a classifier to make the 'none-of-above' decision is highly correlated with its accuracy on the closed-set classes. We find that this relationship holds across loss objectives and architectures, and further demonstrate the trend both on the standard OSR benchmarks as well as on a large-scale ImageNet evaluation. Second, we use this correlation to boost the performance of a maximum logit score OSR 'baseline' by improving its closed-set accuracy, and with this strong baseline achieve state-of-the-art on a number of OSR benchmarks. Similarly, we boost the performance of the existing state-of-the-art method by improving its closed-set accuracy, but the resulting discrepancy with the strong baseline is marginal. Our third contribution is to present the 'Semantic Shift Benchmark' (SSB), which better respects the task of detecting semantic novelty, in contrast to other forms of distribution shift also considered in related sub-fields, such as out-of-distribution detection. On this new evaluation, we again demonstrate that there is negligible difference between the strong baseline and the existing state-of-the-art. Project Page: https://www.robots.ox.ac.uk/~vgg/research/osr/"
                },
                {
                    "title": "Dash: Semi-Supervised Learning with Dynamic Thresholding",
                    "abstract": "While semi-supervised learning (SSL) has received tremendous attentions in many machine learning tasks due to its successful use of unlabeled data, existing SSL algorithms use either all unlabeled examples or the unlabeled examples with a fixed high-confidence prediction during the training progress. However, it is possible that too many correct/wrong pseudo labeled examples are eliminated/selected. In this work we develop a simple yet powerful framework, whose key idea is to select a subset of training examples from the unlabeled data when performing existing SSL methods so that only the unlabeled examples with pseudo labels related to the labeled data will be used to train models. The selection is performed at each updating iteration by only keeping the examples whose losses are smaller than a given threshold that is dynamically adjusted through the iteration. Our proposed approach, Dash, enjoys its adaptivity in terms of unlabeled data selection and its theoretical guarantee. Specifically, we theoretically establish the convergence rate of Dash from the view of non-convex optimization. Finally, we empirically demonstrate the effectiveness of the proposed method in comparison with state-of-the-art over benchmarks."
                },
                {
                    "title": "A Unified Objective for Novel Class Discovery",
                    "abstract": "In this paper, we study the problem of Novel Class Discovery (NCD). NCD aims at inferring novel object categories in an unlabeled set by leveraging from prior knowledge of a labeled set containing different, but related classes. Existing approaches tackle this problem by considering multiple objective functions, usually involving specialized loss terms for the labeled and the unlabeled samples respectively, and often requiring auxiliary regularization terms. In this paper we depart from this traditional scheme and introduce a UNified Objective function (UNO) for discovering novel classes, with the explicit purpose of favoring synergy between supervised and unsupervised learning. Using a multi-view self-labeling strategy, we generate pseudo-labels that can be treated homogeneously with ground truth labels. This leads to a single classification objective operating on both known and unknown classes. De-spite its simplicity, UNO outperforms the state of the art by a significant margin on several benchmarks (\u2248 +10% on CIFAR-100 and +8% on ImageNet). The project page is available at : https://ncd-uno.github.io."
                },
                {
                    "title": "Novel Visual Category Discovery with Dual Ranking Statistics and Mutual Knowledge Distillation",
                    "abstract": "In this paper, we tackle the problem of novel visual category discovery, i.e., grouping unlabelled images from new classes into different semantic partitions by leveraging a labelled dataset that contains images from other different but relevant categories. This is a more realistic and challenging setting than conventional semi-supervised learning. We propose a two-branch learning framework for this problem, with one branch focusing on local part-level information and the other branch focusing on overall characteristics. To transfer knowledge from the labelled data to the unlabelled, we propose using dual ranking statistics on both branches to generate pseudo labels for training on the unlabelled data. We further introduce a mutual knowledge distillation method to allow information exchange and encourage agreement between the two branches for discovering new categories, allowing our model to enjoy the benefits of global and local features. We comprehensively evaluate our method on public benchmarks for generic object classification, as well as the more challenging datasets for fine-grained visual recognition, achieving state-of-the-art performance."
                },
                {
                    "title": "Neighborhood Contrastive Learning for Novel Class Discovery",
                    "abstract": "In this paper, we address Novel Class Discovery (NCD), the task of unveiling new classes in a set of unlabeled samples given a labeled dataset with known classes. We exploit the peculiarities of NCD to build a new framework, named Neighborhood Contrastive Learning (NCL), to learn discriminative representations that are important to clustering performance. Our contribution is twofold. First, we find that a feature extractor trained on the labeled set generates representations in which a generic query sample and its neighbors are likely to share the same class. We exploit this observation to retrieve and aggregate pseudo-positive pairs with contrastive learning, thus encouraging the model to learn more discriminative representations. Second, we notice that most of the instances are easily discriminated by the network, contributing less to the contrastive loss. To overcome this issue, we propose to generate hard negatives by mixing labeled and unlabeled samples in the feature space. We experimentally demonstrate that these two ingredients significantly contribute to clustering performance and lead our model to outperform state-of-the-art methods by a large margin (e.g., clustering accuracy +13% on CIFAR-100 and +8% on ImageNet)."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "Learning curves of generic features maps for realistic datasets with a teacher-student model",
                    "abstract": "Teacher-student models provide a framework in which the typical-case performance of high-dimensional supervised learning can be described in closed form. The assumptions of Gaussian i.i.d. input data underlying the canonical teacher-student model may, however, be perceived as too restrictive to capture the behaviour of realistic data sets. In this paper, we introduce a Gaussian covariate generalisation of the model where the teacher and student can act on different spaces, generated with fixed, but generic feature maps. While still solvable in a closed form, this generalization is able to capture the learning curves for a broad range of realistic data sets, thus redeeming the potential of the teacher-student framework. Our contribution is then two-fold: first, we prove a rigorous formula for the asymptotic training loss and generalisation error. Second, we present a number of situations where the learning curve of the model captures the one of a realistic data set learned with kernel regression and classification, with out-of-the-box feature maps such as random projections or scattering transforms, or with pre-learned ones\u2014such as the features learned by training multi-layer neural networks. We discuss both the power and the limitations of the framework."
                },
                {
                    "title": "Open-World Semi-Supervised Learning",
                    "abstract": "A fundamental limitation of applying semi-supervised learning in real-world settings is the assumption that unlabeled test data contains only classes previously encountered in the labeled training data. However, this assumption rarely holds for data in-the-wild, where instances belonging to novel classes may appear at testing time. Here, we introduce a novel open-world semi-supervised learning setting that formalizes the notion that novel classes may appear in the unlabeled test data. In this novel setting, the goal is to solve the class distribution mismatch between labeled and unlabeled data, where at the test time every input instance either needs to be classified into one of the existing classes or a new unseen class needs to be initialized. To tackle this challenging problem, we propose ORCA, an end-to-end deep learning approach that introduces uncertainty adaptive margin mechanism to circumvent the bias towards seen classes caused by learning discriminative features for seen classes faster than for the novel classes. In this way, ORCA reduces the gap between intra-class variance of seen with respect to novel classes. Experiments on image classification datasets and a single-cell annotation dataset demonstrate that ORCA consistently outperforms alternative baselines, achieving 25% improvement on seen and 96% improvement on novel classes of the ImageNet dataset."
                },
                {
                    "title": "Exponential Moving Average Normalization for Self-supervised and Semi-supervised Learning",
                    "abstract": "We present a plug-in replacement for batch normalization (BN) called exponential moving average normalization (EMAN), which improves the performance of existing student-teacher based self- and semi-supervised learning techniques. Unlike the standard BN, where the statistics are computed within each batch, EMAN, used in the teacher, updates its statistics by exponential moving average from the BN statistics of the student. This design reduces the intrinsic cross-sample dependency of BN and enhances the generalization of the teacher. EMAN improves strong baselines for self-supervised learning by 4-6/1-2 points and semi-supervised learning by about 7/2 points, when 1%/10% supervised labels are available on ImageNet. These improvements are consistent across methods, network architectures, training duration, and datasets, demonstrating the general effectiveness of this technique. The code will be made available online."
                },
                {
                    "title": "Energy-based Out-of-distribution Detection",
                    "abstract": "Determining whether inputs are out-of-distribution (OOD) is an essential building block for safely deploying machine learning models in the open world. However, previous methods relying on the softmax confidence score suffer from overconfident posterior distributions for OOD data. We propose a unified framework for OOD detection that uses an energy score. We show that energy scores better distinguish in- and out-of-distribution samples than the traditional approach using the softmax scores. Unlike softmax confidence scores, energy scores are theoretically aligned with the probability density of the inputs and are less susceptible to the overconfidence issue. Within this framework, energy can be flexibly used as a scoring function for any pre-trained neural classifier as well as a trainable cost function to shape the energy surface explicitly for OOD detection. On a CIFAR-10 pre-trained WideResNet, using the energy score reduces the average FPR (at TPR 95%) by 18.03% compared to the softmax confidence score. With energy-based training, our method outperforms the state-of-the-art on common benchmarks."
                },
                {
                    "title": "Hopfield Networks is All You Need",
                    "abstract": "We show that the transformer attention mechanism is the update rule of a modern Hopfield network with continuous states. This new Hopfield network can store exponentially (with the dimension) many patterns, converges with one update, and has exponentially small retrieval errors. The number of stored patterns is traded off against convergence speed and retrieval error. The new Hopfield network has three types of energy minima (fixed points of the update): (1) global fixed point averaging over all patterns, (2) metastable states averaging over a subset of patterns, and (3) fixed points which store a single pattern. Transformer and BERT models operate in their first layers preferably in the global averaging regime, while they operate in higher layers in metastable states. The gradient in transformers is maximal for metastable states, is uniformly distributed for global averaging, and vanishes for a fixed point near a stored pattern. Using the Hopfield network interpretation, we analyzed learning of transformer and BERT models. Learning starts with attention heads that average and then most of them switch to metastable states. However, the majority of heads in the first layers still averages and can be replaced by averaging, e.g. our proposed Gaussian weighting. In contrast, heads in the last layers steadily learn and seem to use metastable states to collect information created in lower layers. These heads seem to be a promising target for improving transformers. Neural networks with Hopfield networks outperform other methods on immune repertoire classification, where the Hopfield net stores several hundreds of thousands of patterns. We provide a new PyTorch layer called \"Hopfield\", which allows to equip deep learning architectures with modern Hopfield networks as a new powerful concept comprising pooling, memory, and attention. GitHub: this https URL"
                },
                {
                    "title": "Meta Pseudo Labels",
                    "abstract": "We present Meta Pseudo Labels, a semi-supervised learning method that achieves a new state-of-the-art top-1 accuracy of 90.2% on ImageNet, which is 1.6% better than the existing state-of-the-art [16]. Like Pseudo Labels, Meta Pseudo Labels has a teacher network to generate pseudo labels on unlabeled data to teach a student network. However, unlike Pseudo Labels where the teacher is fixed, the teacher in Meta Pseudo Labels is constantly adapted by the feedback of the student\u2019s performance on the labeled dataset. As a result, the teacher generates better pseudo labels to teach the student.1"
                },
                {
                    "title": "Automatically Discovering and Learning New Visual Categories with Ranking Statistics",
                    "abstract": "We tackle the problem of discovering novel classes in an image collection given labelled examples of other classes. This setting is similar to semi-supervised learning, but significantly harder because there are no labelled examples for the new classes. The challenge, then, is to leverage the information contained in the labelled images in order to learn a general-purpose clustering model and use the latter to identify the new classes in the unlabelled data. In this work we address this problem by combining three ideas: (1) we suggest that the common approach of bootstrapping an image representation using the labeled data only introduces an unwanted bias, and that this can be avoided by using self-supervised learning to train the representation from scratch on the union of labelled and unlabelled data; (2) we use rank statistics to transfer the model's knowledge of the labelled classes to the problem of clustering the unlabelled images; and, (3) we train the data representation by optimizing a joint objective function on the labelled and unlabelled subsets of the data, improving both the supervised classification of the labelled data, and the clustering of the unlabelled data. We evaluate our approach on standard classification benchmarks and outperform current methods for novel category discovery by a significant margin."
                },
                {
                    "title": "FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence",
                    "abstract": "Semi-supervised learning (SSL) provides an effective means of leveraging unlabeled data to improve a model's performance. In this paper, we demonstrate the power of a simple combination of two common SSL methods: consistency regularization and pseudo-labeling. Our algorithm, FixMatch, first generates pseudo-labels using the model's predictions on weakly-augmented unlabeled images. For a given image, the pseudo-label is only retained if the model produces a high-confidence prediction. The model is then trained to predict the pseudo-label when fed a strongly-augmented version of the same image. Despite its simplicity, we show that FixMatch achieves state-of-the-art performance across a variety of standard semi-supervised learning benchmarks, including 94.93% accuracy on CIFAR-10 with 250 labels and 88.61% accuracy with 40 -- just 4 labels per class. Since FixMatch bears many similarities to existing SSL methods that achieve worse performance, we carry out an extensive ablation study to tease apart the experimental factors that are most important to FixMatch's success. We make our code available at this https URL."
                },
                {
                    "title": "AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty",
                    "abstract": "Modern deep neural networks can achieve high accuracy when the training distribution and test distribution are identically distributed, but this assumption is frequently violated in practice. When the train and test distributions are mismatched, accuracy can plummet. Currently there are few techniques that improve robustness to unforeseen data shifts encountered during deployment. In this work, we propose a technique to improve the robustness and uncertainty estimates of image classifiers. We propose AugMix, a data processing technique that is simple to implement, adds limited computational overhead, and helps models withstand unforeseen corruptions. AugMix significantly improves robustness and uncertainty measures on challenging image classification benchmarks, closing the gap between previous methods and the best possible performance in some cases by more than half."
                },
                {
                    "title": "ReMixMatch: Semi-Supervised Learning with Distribution Alignment and Augmentation Anchoring",
                    "abstract": "We improve the recently-proposed \"MixMatch\" semi-supervised learning algorithm by introducing two new techniques: distribution alignment and augmentation anchoring. Distribution alignment encourages the marginal distribution of predictions on unlabeled data to be close to the marginal distribution of ground-truth labels. Augmentation anchoring feeds multiple strongly augmented versions of an input into the model and encourages each output to be close to the prediction for a weakly-augmented version of the same input. To produce strong augmentations, we propose a variant of AutoAugment which learns the augmentation policy while the model is being trained. Our new algorithm, dubbed ReMixMatch, is significantly more data-efficient than prior work, requiring between $5\\times$ and $16\\times$ less data to reach the same accuracy. For example, on CIFAR-10 with 250 labeled examples we reach $93.73\\%$ accuracy (compared to MixMatch's accuracy of $93.58\\%$ with $4{,}000$ examples) and a median accuracy of $84.92\\%$ with just four labels per class. We make our code and data open-source at this https URL."
                },
                {
                    "title": "Self-Training With Noisy Student Improves ImageNet Classification",
                    "abstract": "We present a simple self-training method that achieves 88.4% top-1 accuracy on ImageNet, which is 2.0% better than the state-of-the-art model that requires 3.5B weakly labeled Instagram images. On robustness test sets, it improves ImageNet-A top-1 accuracy from 61.0% to 83.7%, reduces ImageNet-C mean corruption error from 45.7 to 28.3, and reduces ImageNet-P mean flip rate from 27.8 to 12.2. To achieve this result, we first train an EfficientNet model on labeled ImageNet images and use it as a teacher to generate pseudo labels on 300M unlabeled images. We then train a larger EfficientNet as a student model on the combination of labeled and pseudo labeled images. We iterate this process by putting back the student as the teacher. During the generation of the pseudo labels, the teacher is not noised so that the pseudo labels are as accurate as possible. However, during the learning of the student, we inject noise such as dropout, stochastic depth and data augmentation via RandAugment to the student so that the student generalizes better than the teacher."
                },
                {
                    "title": "Randaugment: Practical automated data augmentation with a reduced search space",
                    "abstract": "Recent work on automated augmentation strategies has led to state-of-the-art results in image classification and object detection. An obstacle to a large-scale adoption of these methods is that they require a separate and expensive search phase. A common way to overcome the expense of the search phase was to use a smaller proxy task. However, it was not clear if the optimized hyperparameters found on the proxy task are also optimal for the actual task. In this work, we rethink the process of designing automated augmentation strategies. We find that while previous work required a search for both magnitude and probability of each operation independently, it is sufficient to only search for a single distortion magnitude that jointly controls all operations. We hence propose a simplified search space that vastly reduces the computational expense of automated augmentation, and permits the removal of a separate proxy task.Despite the simplifications, our method achieves equal or better performance over previous automated augmentation strategies on on CIFAR-10/100, SVHN, ImageNet and COCO datasets. EfficientNet-B7, we achieve 85.0% accuracy, a 1.0% increase over baseline augmentation, a 0.6% improvement over AutoAugment on the ImageNet dataset. With EfficientNet-B8, we achieve 85.4% accuracy on ImageNet, which matches a previous result that used 3.5B extra images. On object detection, the same method as classification leads to 1.0-1.3% improvement over baseline augmentation. Code will be made available online."
                },
                {
                    "title": "Learning to Discover Novel Visual Categories via Deep Transfer Clustering",
                    "abstract": "We consider the problem of discovering novel object categories in an image collection. While these images are unlabelled, we also assume prior knowledge of related but different image classes. We use such prior knowledge to reduce the ambiguity of clustering, and improve the quality of the newly discovered classes. Our contributions are twofold. The first contribution is to extend Deep Embedded Clustering to a transfer learning setting; we also improve the algorithm by introducing a representation bottleneck, temporal ensembling, and consistency. The second contribution is a method to estimate the number of classes in the unlabelled data. This also transfers knowledge from the known classes, using them as probes to diagnose different choices for the number of classes in the unlabelled subset. We thoroughly evaluate our method, substantially outperforming state-of-the-art techniques in a large number of benchmarks, including ImageNet, OmniGlot, CIFAR-100, CIFAR-10, and SVHN."
                },
                {
                    "title": "The Herbarium Challenge 2019 Dataset",
                    "abstract": "Herbarium sheets are invaluable for botanical research, and considerable time and effort is spent by experts to label and identify specimens on them. In view of recent advances in computer vision and deep learning, developing an automated approach to help experts identify specimens could significantly accelerate research in this area. Whereas most existing botanical datasets comprise photos of specimens in the wild, herbarium sheets exhibit dried specimens, which poses new challenges. We present a challenge dataset of herbarium sheet images labeled by experts, with the intent of facilitating the development of automated identification techniques for this challenging scenario."
                },
                {
                    "title": "CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features",
                    "abstract": "Regional dropout strategies have been proposed to enhance performance of convolutional neural network classifiers. They have proved to be effective for guiding the model to attend on less discriminative parts of objects (e.g. leg as opposed to head of a person), thereby letting the network generalize better and have better object localization capabilities. On the other hand, current methods for regional dropout removes informative pixels on training images by overlaying a patch of either black pixels or random noise. Such removal is not desirable because it suffers from information loss causing inefficiency in training. We therefore propose the CutMix augmentation strategy: patches are cut and pasted among training images where the ground truth labels are also mixed proportionally to the area of the patches. By making efficient use of training pixels and retaining the regularization effect of regional dropout, CutMix consistently outperforms state-of-the-art augmentation strategies on CIFAR and ImageNet classification tasks, as well as on ImageNet weakly-supervised localization task. Moreover, unlike previous augmentation methods, our CutMix-trained ImageNet classifier, when used as a pretrained model, results in consistent performance gain in Pascal detection and MS-COCO image captioning benchmarks. We also show that CutMix can improve the model robustness against input corruptions and its out-of distribution detection performance."
                },
                {
                    "title": "Unsupervised Data Augmentation for Consistency Training",
                    "abstract": "Semi-supervised learning lately has shown much promise in improving deep learning models when labeled data is scarce. Common among recent approaches is the use of consistency training on a large amount of unlabeled data to constrain model predictions to be invariant to input noise. In this work, we present a new perspective on how to effectively noise unlabeled examples and argue that the quality of noising, specifically those produced by advanced data augmentation methods, plays a crucial role in semi-supervised learning. By substituting simple noising operations with advanced data augmentation methods such as RandAugment and back-translation, our method brings substantial improvements across six language and three vision tasks under the same consistency training framework. On the IMDb text classification dataset, with only 20 labeled examples, our method achieves an error rate of 4.20, outperforming the state-of-the-art model trained on 25,000 labeled examples. On a standard semi-supervised learning benchmark, CIFAR-10, our method outperforms all previous approaches and achieves an error rate of 5.43 with only 250 examples. Our method also combines well with transfer learning, e.g., when finetuning from BERT, and yields improvements in high-data regime, such as ImageNet, whether when there is only 10% labeled data or when a full labeled set with 1.3M extra unlabeled examples is used. Code is available at this https URL."
                },
                {
                    "title": "Multi-class Classification without Multi-class Labels",
                    "abstract": "This work presents a new strategy for multi-class classification that requires no class-specific labels, but instead leverages pairwise similarity between examples, which is a weaker form of annotation. The proposed method, meta classification learning, optimizes a binary classifier for pairwise similarity prediction and through this process learns a multi-class classifier as a submodule. We formulate this approach, present a probabilistic graphical model for it, and derive a surprisingly simple loss function that can be used to learn neural network-based models. We then demonstrate that this same framework generalizes to the supervised, unsupervised cross-task, and semi-supervised settings. Our method is evaluated against state of the art in all three learning paradigms and shows a superior or comparable accuracy, providing evidence that learning multi-class classification without multi-class labels is a viable learning option."
                },
                {
                    "title": "Co-teaching: Robust training of deep neural networks with extremely noisy labels",
                    "abstract": "Deep learning with noisy labels is practically challenging, as the capacity of deep models is so high that they can totally memorize these noisy labels sooner or later during training. Nonetheless, recent studies on the memorization effects of deep neural networks show that they would first memorize training data of clean labels and then those of noisy labels. Therefore in this paper, we propose a new deep learning paradigm called Co-teaching for combating with noisy labels. Namely, we train two deep neural networks simultaneously, and let them teach each other given every mini-batch: firstly, each network feeds forward all data and selects some data of possibly clean labels; secondly, two networks communicate with each other what data in this mini-batch should be used for training; finally, each network back propagates the data selected by its peer network and updates itself. Empirical results on noisy versions of MNIST, CIFAR-10 and CIFAR-100 demonstrate that Co-teaching is much superior to the state-of-the-art methods in the robustness of trained deep models."
                },
                {
                    "title": "Learning to cluster in order to Transfer across domains and tasks",
                    "abstract": "This paper introduces a novel method to perform transfer learning across domains and tasks, formulating it as a problem of learning to cluster. The key insight is that, in addition to features, we can transfer similarity information and this is sufficient to learn a similarity function and clustering network to perform both domain adaptation and cross-task transfer learning. We begin by reducing categorical information to pairwise constraints, which only considers whether two instances belong to the same class or not. This similarity is category-agnostic and can be learned from data in the source domain using a similarity network. We then present two novel approaches for performing transfer learning using this similarity function. First, for unsupervised domain adaptation, we design a new loss function to regularize classification with a constrained clustering loss, hence learning a clustering network with the transferred similarity metric generating the training inputs. Second, for cross-task learning (i.e., unsupervised clustering with unseen categories), we propose a framework to reconstruct and estimate the number of semantic clusters, again using the clustering network. Since the similarity network is noisy, the key is to use a robust clustering algorithm, and we show that our formulation is more robust than the alternative constrained and unconstrained clustering approaches. Using this method, we first show state of the art results for the challenging cross-task problem, applied on Omniglot and ImageNet. Our results show that we can reconstruct semantic clusters with high accuracy. We then evaluate the performance of cross-domain transfer using images from the Office-31 and SVHN-MNIST tasks and present top accuracy on both datasets. Our approach doesn't explicitly deal with domain discrepancy. If we combine with a domain adaptation loss, it shows further improvement."
                },
                {
                    "title": "Smooth Neighbors on Teacher Graphs for Semi-Supervised Learning",
                    "abstract": "The recently proposed self-ensembling methods have achieved promising results in deep semi-supervised learning, which penalize inconsistent predictions of unlabeled data under different perturbations. However, they only consider adding perturbations to each single data point, while ignoring the connections between data samples. In this paper, we propose a novel method, called Smooth Neighbors on Teacher Graphs (SNTG). In SNTG, a graph is constructed based on the predictions of the teacher model, i.e., the implicit self-ensemble of models. Then the graph serves as a similarity measure with respect to which the representations of \"similar\" neighboring points are learned to be smooth on the low-dimensional manifold. We achieve state-of-the-art results on semi-supervised learning benchmarks. The error rates are 9.89%, 3.99% for CIFAR-10 with 4000 labels, SVHN with 500 labels, respectively. In particular, the improvements are significant when the labels are fewer. For the non-augmented MNIST with only 20 labels, the error rate is reduced from previous 4.81% to 1.36%. Our method also shows robustness to noisy labels."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Sparse Representation-Based Open Set Recognition",
                    "abstract": "We propose a generalized Sparse Representation-based Classification (SRC) algorithm for open set recognition where not all classes presented during testing are known during training. The SRC algorithm uses class reconstruction errors for classification. As most of the discriminative information for open set recognition is hidden in the tail part of the matched and sum of non-matched reconstruction error distributions, we model the tail of those two error distributions using the statistical Extreme Value Theory (EVT). Then we simplify the open set recognition problem into a set of hypothesis testing problems. The confidence scores corresponding to the tail distributions of a novel test sample are then fused to determine its identity. The effectiveness of the proposed method is demonstrated using four publicly available image and object classification datasets and it is shown that this method can perform significantly better than many competitive open set recognition algorithms."
                },
                {
                    "title": "Virtual Adversarial Training: A Regularization Method for Supervised and Semi-Supervised Learning",
                    "abstract": "We propose a new regularization method based on virtual adversarial loss: a new measure of local smoothness of the conditional label distribution given input. Virtual adversarial loss is defined as the robustness of the conditional label distribution around each input data point against local perturbation. Unlike adversarial training, our method defines the adversarial direction without label information and is hence applicable to semi-supervised learning. Because the directions in which we smooth the model are only \u201cvirtually\u201d adversarial, we call our method virtual adversarial training (VAT). The computational cost of VAT is relatively low. For neural networks, the approximated gradient of virtual adversarial loss can be computed with no more than two pairs of forward- and back-propagations. In our experiments, we applied VAT to supervised and semi-supervised learning tasks on multiple benchmark datasets. With a simple enhancement of the algorithm based on the entropy minimization principle, our VAT achieves state-of-the-art performance for semi-supervised learning tasks on SVHN and CIFAR-10."
                },
                {
                    "title": "Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results",
                    "abstract": "The recently proposed Temporal Ensembling has achieved state-of-the-art results in several semi-supervised learning benchmarks. It maintains an exponential moving average of label predictions on each training example, and penalizes predictions that are inconsistent with this target. However, because the targets change only once per epoch, Temporal Ensembling becomes unwieldy when learning large datasets. To overcome this problem, we propose Mean Teacher, a method that averages model weights instead of label predictions. As an additional benefit, Mean Teacher improves test accuracy and enables training with fewer labels than Temporal Ensembling. Without changing the network architecture, Mean Teacher achieves an error rate of 4.35% on SVHN with 250 labels, outperforming Temporal Ensembling trained with 1000 labels. We also show that a good network architecture is crucial to performance. Combining Mean Teacher and Residual Networks, we improve the state of the art on CIFAR-10 with 4000 labels from 10.55% to 6.28%, and on ImageNet 2012 with 10% of the labels from 35.24% to 9.11%."
                },
                {
                    "title": "Temporal Ensembling for Semi-Supervised Learning",
                    "abstract": "In this paper, we present a simple and efficient method for training deep neural networks in a semi-supervised setting where only a small portion of training data is labeled. We introduce self-ensembling, where we form a consensus prediction of the unknown labels using the outputs of the network-in-training on different epochs, and most importantly, under different regularization and input augmentation conditions. This ensemble prediction can be expected to be a better predictor for the unknown labels than the output of the network at the most recent training epoch, and can thus be used as a target for training. Using our method, we set new records for two standard semi-supervised learning benchmarks, reducing the (non-augmented) classification error rate from 18.44% to 7.05% in SVHN with 500 labels and from 18.63% to 16.55% in CIFAR-10 with 4000 labels, and further to 5.12% and 12.16% by enabling the standard augmentations. We additionally obtain a clear improvement in CIFAR-100 classification accuracy by using random images from the Tiny Images dataset as unlabeled extra inputs during training. Finally, we demonstrate good tolerance to incorrect labels."
                },
                {
                    "title": "Regularization With Stochastic Transformations and Perturbations for Deep Semi-Supervised Learning",
                    "abstract": "Effective convolutional neural networks are trained on large sets of labeled data. However, creating large labeled datasets is a very costly and time-consuming task. Semi-supervised learning uses unlabeled data to train a model with higher accuracy when there is a limited set of labeled data available. In this paper, we consider the problem of semi-supervised learning with convolutional neural networks. Techniques such as randomized data augmentation, dropout and random max-pooling provide better generalization and stability for classifiers that are trained using gradient descent. Multiple passes of an individual sample through the network might lead to different predictions due to the non-deterministic behavior of these techniques. We propose an unsupervised loss function that takes advantage of the stochastic nature of these methods and minimizes the difference between the predictions of multiple passes of a training sample through the network. We evaluate the proposed method on several benchmark datasets."
                },
                {
                    "title": "Learning with Pseudo-Ensembles",
                    "abstract": "We formalize the notion of a pseudo-ensemble, a (possibly infinite) collection of child models spawned from a parent model by perturbing it according to some noise process. E.g., dropout [9] in a deep neural network trains a pseudo-ensemble of child subnetworks generated by randomly masking nodes in the parent network. We examine the relationship of pseudo-ensembles, which involve perturbation in model-space, to standard ensemble methods and existing notions of robustness, which focus on perturbation in observation-space. We present a novel regularizer based on making the behavior of a pseudo-ensemble robust with respect to the noise process generating it. In the fully-supervised setting, our regularizer matches the performance of dropout. But, unlike dropout, our regularizer naturally extends to the semi-supervised setting, where it produces state-of-the-art results. We provide a case study in which we transform the Recursive Neural Tensor Network of [19] into a pseudo-ensemble, which significantly improves its performance on a real-world sentiment analysis benchmark."
                },
                {
                    "title": "3D Object Representations for Fine-Grained Categorization",
                    "abstract": "While 3D object representations are being revived in the context of multi-view object class detection and scene understanding, they have not yet attained wide-spread use in fine-grained categorization. State-of-the-art approaches achieve remarkable performance when training data is plentiful, but they are typically tied to flat, 2D representations that model objects as a collection of unconnected views, limiting their ability to generalize across viewpoints. In this paper, we therefore lift two state-of-the-art 2D object representations to 3D, on the level of both local feature appearance and location. In extensive experiments on existing and newly proposed datasets, we show our 3D object representations outperform their state-of-the-art 2D counterparts for fine-grained categorization and demonstrate their efficacy for estimating 3D geometry from images via ultra-wide baseline matching and 3D reconstruction."
                },
                {
                    "title": "Toward Open Set Recognition",
                    "abstract": "To date, almost all experimental evaluations of machine learning-based recognition algorithms in computer vision have taken the form of \u201cclosed set\u201d recognition, whereby all testing classes are known at training time. A more realistic scenario for vision applications is \u201copen set\u201d recognition, where incomplete knowledge of the world is present at training time, and unknown classes can be submitted to an algorithm during testing. This paper explores the nature of open set recognition and formalizes its definition as a constrained minimization problem. The open set recognition problem is not well addressed by existing algorithms because it requires strong generalization. As a step toward a solution, we introduce a novel \u201c1-vs-set machine,\u201d which sculpts a decision space from the marginal distances of a 1-class or binary SVM with a linear kernel. This methodology applies to several different applications in computer vision where open set recognition is a challenging problem, including object recognition and face verification. We consider both in this work, with large scale cross-dataset experiments performed over the Caltech 256 and ImageNet sets, as well as face matching experiments performed over the Labeled Faces in the Wild set. The experiments highlight the effectiveness of machines adapted for open set evaluation compared to existing 1-class and binary SVMs for the same tasks."
                },
                {
                    "title": "Fine-Grained Visual Classification of Aircraft",
                    "abstract": "This paper introduces FGVC-Aircraft, a new dataset containing 10,000 images of aircraft spanning 100 aircraft models, organised in a three-level hierarchy. At the finer level, differences between models are often subtle but always visually measurable, making visual recognition challenging but possible. A benchmark is obtained by defining corresponding classification tasks and evaluation protocols, and baseline results are presented. The construction of this dataset was made possible by the work of aircraft enthusiasts, a strategy that can extend to the study of number of other object classes. Compared to the domains usually considered in fine-grained visual classification (FGVC), for example animals, aircraft are rigid and hence less deformable. They, however, present other interesting modes of variation, including purpose, size, designation, structure, historical style, and branding."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "The Caltech-UCSD Birds-200-2011 Dataset",
                    "abstract": "CUB-200-2011 is an extended version of CUB-200 [7], a challenging dataset of 200 bird species. The extended version roughly doubles the number of images per category and adds new part localization annotations. All images are annotated with bounding boxes, part locations, and at- tribute labels. Images and annotations were filtered by mul- tiple users of Mechanical Turk. We introduce benchmarks and baseline experiments for multi-class categorization and part localization."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Semi-Supervised Self-Training of Object Detection Models",
                    "abstract": "The construction of appearance-based object detection systems is time-consuming and difficult because a large number of training examples must be collected and manually labeled in order to capture variations in object appearance. Semi-supervised training is a means for reducing the effort needed to prepare the training set by training the model with a small number of fully labeled examples and an additional set of unlabeled or weakly labeled examples. In this work we present a semi-supervised approach to training object detection systems based on self-training. We implement our approach as a wrapper around the training process of an existing object detector and present empirical results. The key contributions of this empirical study is to demonstrate that a model trained in this manner can achieve results comparable to a model trained in the traditional manner using a much larger set of fully labeled data, and that a training data selection metric that is defined independently of the detector greatly outperforms a selection metric based on the detection confidence generated by the detector."
                },
                {
                    "title": "Semi-supervised Learning by Entropy Minimization",
                    "abstract": "We consider the semi-supervised learning problem, where a decision rule is to be learned from labeled and unlabeled data. In this framework, we motivate minimum entropy regularization, which enables to incorporate unlabeled data in the standard supervised learning. Our approach includes other approaches to the semi-supervised problem as particular or limiting cases. A series of experiments illustrates that the proposed solution benefits from unlabeled data. The method challenges mixture models when the data are sampled from the distribution class spanned by the generative model. The performances are definitely in favor of minimum entropy regularization when generative models are misspecified, and the weighting of unlabeled data provides robustness to the violation of the \"cluster assumption\". Finally, we also illustrate that the method can also be far superior to manifold learning in high dimension spaces."
                },
                {
                    "title": "The Big Book of Concepts",
                    "abstract": "Concepts embody our knowledge of the kinds of things there are in the world. Tying our past experiences to our present interactions with the environment, they enable us to recognize and understand new objects and events. Concepts are also relevant to understanding domains such as social situations, personality types, and even artistic styles. Yet like other phenomenologically simple cognitive processes such as walking or understanding speech, concept formation and use are maddeningly complex. Research since the 1970s and the decline of the \"classical view\" of concepts have greatly illuminated the psychology of concepts. But persistent theoretical disputes have sometimes obscured this progress. The Big Book of Concepts goes beyond those disputes to reveal the advances that have been made, focusing on the major empirical discoveries. By reviewing and evaluating research on diverse topics such as category learning, word meaning, conceptual development in infants and children, and the basic level of categorization, the book develops a much broader range of criteria than is usual for evaluating theories of concepts."
                },
                {
                    "title": "The Concave-Convex Procedure (CCCP)",
                    "abstract": "We introduce the Concave-Convex procedure (CCCP) which constructs discrete time iterative dynamical systems which are guaranteed to monotonically decrease global optimization/energy functions. It can be applied to (almost) any optimization problem and many existing algorithms can be interpreted in terms of CCCP. In particular, we prove relationships to some applications of Legendre transform techniques. We then illustrate CCCP by applications to Potts models, linear assignment, EM algorithms, and Generalized Iterative Scaling (GIS). CCCP can be used both as a new way to understand existing optimization algorithms and as a procedure for generating new algorithms."
                },
                {
                    "title": "The Origin of Concepts",
                    "abstract": "Only human beings have a rich conceptual repertoire with concepts like tort, entropy, Abelian group, mannerism, icon and deconstruction. How have humans constructed these concepts? And once they have been constructed by adults, how do children acquire them? While primarily focusing on the second question, in The Origin of Concepts, Susan Carey shows that the answers to both overlap substantially. Carey begins by characterizing the innate starting point for conceptual development, namely systems of core cognition. Representations of core cognition are the output of dedicated input analyzers, as with perceptual representations, but these core representations differ from perceptual representations in having more abstract contents and richer functional roles. Carey argues that the key to understanding cognitive development lies in recognizing conceptual discontinuities in which new representational systems emerge that have more expressive power than core cognition and are also incommensurate with core cognition and other earlier representational systems. Finally, Carey fleshes out Quinian bootstrapping, a learning mechanism that has been repeatedly sketched in the literature on the history and philosophy of science. She demonstrates that Quinian bootstrapping is a major mechanism in the construction of new representational resources over the course of childrens cognitive development. Carey shows how developmental cognitive science resolves aspects of long-standing philosophical debates about the existence, nature, content, and format of innate knowledge. She also shows that understanding the processes of conceptual development in children illuminates the historical process by which concepts are constructed, and transforms the way we think about philosophical problems about the nature of concepts and the relations between language and thought."
                },
                {
                    "title": "Learning Patterns and Pattern Sequences by Self-Organizing Nets of Threshold Elements",
                    "abstract": "Various information-processing capabilities of self-organizing nets of threshold elements are studied. A self-organizing net, learning from patterns or pattern sequences given from outside as stimuli, \"remembers\" some of them as stable equilibrium states or state-transition sequences of the net. A condition where many patterns and pattern sequences are remembered in a net at the same time is shown. The stability degree of their remembrance and recalling under noise disturbances is investigated theoretically. For this purpose, the stability of state transition in an autonomous logical net of threshold elements is studied by the use of characteristics of threshold elements."
                },
                {
                    "title": "Probability of error of some adaptive pattern-recognition machines",
                    "abstract": "A simple taught pattern-recognition machine for detecting an unknown, fixed, randomly occurring pattern is derived using a Bayes' approach, and its probability of error is analyzed. It is shown that with probability one, the machine converges to the optimal detector (a matched filter) for the unknown pattern, that the asymptotic decision function statistics are Gaussian, and that for an important class of problems, the central-limit theorem can be invoked to calculate the approximate probability of error at any stage of convergence. An untaught adaptive pattern-recognition machine may be made from the taught machine by using its own output instead of a teacher, and the asymptotic probability of error of this device is derived. It is shown that it does not converge to a matched filter for the unknown pattern, but that in any practical case it performs almost as well. Finally, the results of an experimental simulation of both machines are presented as curves of the relative frequency of error vs. time, and are compared with the values calculated by theory."
                },
                {
                    "title": "Robust Semi-Supervised Learning when Not All Classes have Labels",
                    "abstract": "Semi-supervised learning (SSL) provides a powerful framework for leveraging unlabeled data. Existing SSL typically requires all classes have labels. However, in many real-world applications, there may exist some classes that are difficult to label or newly occurred classes that cannot be labeled in time, resulting in there are unseen classes in unlabeled data. Unseen classes will be misclassified as seen classes, causing poor classification performance. The performance of seen classes is also harmed by the existence of unseen classes. This limits the practical and wider application of SSL. To address this problem, this paper proposes a new SSL approach that can classify not only seen classes but also unseen classes. Our approach consists of two modules: unseen class classification and learning pace synchronization . Specifically, we first enable the SSL methods to classify unseen classes by exploiting pairwise similarity between examples and then synchronize the learning pace between seen and unseen classes by proposing an adaptive threshold with distribution alignment. Extensive empirical results show our approach achieves significant performance improvement in both seen and unseen classes compared with previous studies."
                },
                {
                    "title": "Mutual Information-based Generalized Category Discovery",
                    "abstract": "We introduce an information-maximization approach for the Generalized Category Discovery (GCD) problem. Speci\ufb01cally, we explore a parametric family of loss functions evaluating the mutual information between the features and the labels, and \ufb01nd automatically the one that maximizes the predictive performances. Furthermore, we introduce the Elbow Maximum Centroid-Shift (EMaCS) technique, which estimates the number of classes in the unlabeled set. We report comprehensive experiments, which show that our mutual information-based approach (MIB) is both versatile and highly competitive under various GCD scenarios. The gap between the proposed approach and the existing methods is signi\ufb01cant, more so when dealing with \ufb01ne-grained classi\ufb01cation problems. Our code:"
                },
                {
                    "title": "Appendix of Divide and Conquer: Compositional Experts for Generalized Novel Class Discovery",
                    "abstract": "In GNCD, base and novel sets contain different training targets \u2014 for base set we have accurate and consistent ground-truth labels, while for novel set the training targets (generated pseudo labels) are less accurate and may change during training. We aim to build a model able to handle both base and novel classes, which is challenging due to (1) the ambiguity between base and novel classes, and (2) the lack of supervision on novel data. To address (1), we propose to model base and novel data using two groups of compositional experts with complementary specialties from both batch-wise and class-wise perspectives. For (2), we propose a Local Aggregation strategy to refine the pseudo labels by encouraging local consistency among novel data. Our idea of the compositional structure comes from the observations towards GNCD output space, which can be clearly characterized with a batch-class view (cf. Fig. 1b in our paper). From this view the GNCD output space exhibits a compositional nature in both batch-wise and classwise perspectives. We propose to model this compositional nature using a group of batch-wise experts (Fig. 1c) and another group of class-wise experts (Fig. 1d). Each group of experts is capable of characterizing the whole dataset, yet presents different specialties \u2014 with batch-wise experts capturing separability between base and novel sets, and class-wise experts modeling discriminability within each set of classes. Considering the challenges of GNCD, we keep both groups of experts to make the most of their learning abilities by allowing complementary collaborations. We further propose to address the aforementioned inconsistent training targets using local aggregation for novel samples, which complement the global-to-local pseudolabeling with local-to-local regularization. Extensive experiments have validated the effectiveness of the above ideas, which compose our proposed ComEx and contribute to its superior performance in GNCD."
                },
                {
                    "title": "Dual Student: Breaking the Limits of the Teacher in Semi-supervised Learning",
                    "abstract": "Recently, consistency-based methods have achieved state-of-the-art results in semi-supervised learning (SSL). These methods always involve two roles, an explicit or implicit teacher model and a student model, and penalize predictions under different perturbations by a consistency constraint. However, the weights of these two roles are tightly coupled since the teacher is essentially an exponential moving average (EMA) of the student. In this work, we show that the coupled EMA teacher causes a performance bottleneck. To address this problem, we introduce Dual Student, which replaces the teacher with another student. We also define a novel concept, stable sample, following which a stabilization constraint is designed for our structure to be trainable. Further, we discuss two variants of our method, which produce even higher performance. Extensive experiments show that our method improves the classification performance significantly on several main SSL benchmarks. Specifically, it reduces the error rate of the 13-layer CNN from 16.84% to 12.39% on CIFAR-10 with 1k labels and from 34.10% to 31.56% on CIFAR-100 with 10k labels. In addition, our method also achieves a clear improvement in domain adaptation."
                },
                {
                    "title": "A brief introduction to weakly supervised learning",
                    "abstract": "Supervised learning techniques construct predictive models by learning from a large number of training examples, where each training example has a label indicating its ground-truth output. Though current techniques have achieved great success, it is noteworthy that in many tasks it is difficult to get strong supervision information like fully ground-truth labels due to the high cost of the data-labeling process. Thus, it is desirable for machine-learning techniques to work with weak supervision. This article reviews some research progress of weakly supervised learning, focusing on three typical types of weak supervision: incomplete supervision, where only a subset of training data is given with labels; inexact supervision, where the training data are given with only coarse-grained labels; and inaccurate supervision, where the given labels are not always ground-truth."
                },
                {
                    "title": "Pseudo-Label : The Simple and Efficient Semi-Supervised Learning Method for Deep Neural Networks",
                    "abstract": "We propose the simple and e\ufb03cient method of semi-supervised learning for deep neural networks. Basically, the proposed network is trained in a supervised fashion with labeled and unlabeled data simultaneously. For unlabeled data, Pseudo-Label s, just picking up the class which has the maximum network output, are used as if they were true labels. Without any unsupervised pre-training method, this simple method with dropout shows the state-of-the-art performance."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Iterative Reclassification Procedure for Constructing An Asymptotically Optimal Rule of Allocation in Discriminant-Analysis",
                    "abstract": "Abstract The construction of a suitable rule of allocation in the two-population discrimination problem is considered in the case where there are initially available from the populations II1, II2, n 1, n 2 observations and M unclassified observations. An iterative reclassification procedure based on the n 1 + n 3 + M observations is proposed and found asymptotically optimal when M \u2192 \u221e and n 1 and n 2 are moderately large. The case of finite M is evaluated by a Monte Carlo experiment which suggests that the proposed procedure, after only one iteration, gives a rule with smaller average risk than the usual rule based on just the n 1 + n 2 classified observations."
                },
                {
                    "title": "Some methods for classification and analysis of multivariate observations",
                    "abstract": "The main purpose of this paper is to describe a process for partitioning an N-dimensional population into k sets on the basis of a sample. The process, which is called 'k-means,' appears to give partitions which are reasonably efficient in the sense of within-class variance. That is, if p is the probability mass function for the population, S = {S1, S2, * *, Sk} is a partition of EN, and ui, i = 1, 2, * , k, is the conditional mean of p over the set Si, then W2(S) = ff=ISi f z u42 dp(z) tends to be low for the partitions S generated by the method. We say 'tends to be low,' primarily because of intuitive considerations, corroborated to some extent by mathematical analysis and practical computational experience. Also, the k-means procedure is easily programmed and is computationally economical, so that it is feasible to process very large samples on a digital computer. Possible applications include methods for similarity grouping, nonlinear prediction, approximating multivariate distributions, and nonparametric tests for independence among several variables. In addition to suggesting practical classification methods, the study of k-means has proved to be theoretically interesting. The k-means concept represents a generalization of the ordinary sample mean, and one is naturally led to study the pertinent asymptotic behavior, the object being to establish some sort of law of large numbers for the k-means. This problem is sufficiently interesting, in fact, for us to devote a good portion of this paper to it. The k-means are defined in section 2.1, and the main results which have been obtained on the asymptotic behavior are given there. The rest of section 2 is devoted to the proofs of these results. Section 3 describes several specific possible applications, and reports some preliminary results from computer experiments conducted to explore the possibilities inherent in the k-means idea. The extension to general metric spaces is indicated briefly in section 4. The original point of departure for the work described here was a series of problems in optimal classification (MacQueen [9]) which represented special"
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively adapt Teacher-Student architectures in Semi-Supervised Learning to address the challenges of Generalized Category Discovery in dynamic environments with new classes?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of Semi-Supervised Learning, particularly in real-world applications where new classes frequently emerge. By improving the adaptability of Teacher-Student models, we can enhance their performance in dynamic settings, leading to more robust machine learning systems. This research could pave the way for future studies focused on continuous learning and adaptability, ultimately contributing to the development of intelligent systems that can learn and evolve like biological organisms. Practical applications may include areas such as autonomous driving, where models must recognize and adapt to new vehicle types or traffic scenarios.\n\n**[Question 3] - Why is it hard?**  \nThe problem is challenging due to several complexities:  \n1. **Learning Gap**: Traditional Teacher-Student models often result in unsynchronized learning, leading to a significant gap in knowledge transfer, especially for new classes.  \n2. **Discrepancies in Features**: The teacher's rapid learning pace can create large discrepancies between the representations of weakly-augmented (teacher) and strongly-augmented (student) data, complicating consistency loss optimization.  \n3. **Attention Inconsistency**: Inadequate supervision for new classes can lead to misalignment in focus between the teacher and student, resulting in ineffective learning. Naive approaches may fail to address these issues, as they do not account for the dynamic nature of learning and the need for synchronized attention.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on closed-world assumptions, which do not account for the emergence of new classes in real-world scenarios. Existing solutions have limitations in their ability to adapt to dynamic environments, often leading to unsynchronized learning and attention misalignment. Barriers such as the lack of effective methods for bridging the learning gap and addressing feature discrepancies have hindered progress. Our approach differs by proposing a Flipped Classroom Consistency Model (FlipClass) that aims to synchronize learning and adapt teacher attention to the student's needs, thereby improving the learning process for new classes.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the following key components:  \n- **Method**: Implementing the Flipped Classroom Cons"
            }
        },
        "author_data": {
            "630a31c6-8178-428b-abff-fe43b3b0397e": {
                "pk": "630a31c6-8178-428b-abff-fe43b3b0397e",
                "name": "Haonan Lin",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "27492235-2dac-4505-bdfc-01559e6b1660": {
                "pk": "27492235-2dac-4505-bdfc-01559e6b1660",
                "name": "Wenbin An",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "ec29cdb7-b7e5-4edf-bfac-f7e9740a1afa": {
                "pk": "ec29cdb7-b7e5-4edf-bfac-f7e9740a1afa",
                "name": "Jiahao Wang",
                "collaborators": [
                    "Amer Shalaby",
                    "Yunhong Wang",
                    "Sheng Liu",
                    "Annan Li",
                    "Mingjin Zhang",
                    "Jianming Wang",
                    "Qi Wang",
                    "Renyuan Zhang",
                    "Zenghui Wang",
                    "Kai Cai"
                ],
                "domain": [
                    "Public Transit",
                    "Large Language Models",
                    "Deep Learning",
                    "Action Recognition"
                ],
                "publications": [
                    {
                        "title": "Leveraging Large Language Models for Enhancing Public Transit Services",
                        "abstract": "Public transit systems play a crucial role in providing efficient and sustainable transportation options in urban areas. However, these systems face various challenges in meeting commuters' needs. On the other hand, despite the rapid development of Large Language Models (LLMs) worldwide, their integration into transit systems remains relatively unexplored.   The objective of this paper is to explore the utilization of LLMs in the public transit system, with a specific focus on improving the customers' experience and transit staff performance. We present a general framework for developing LLM applications in transit systems, wherein the LLM serves as the intermediary for information communication between natural language content and the resources within the database. In this context, the LLM serves a multifaceted role, including understanding users' requirements, retrieving data from the dataset in response to user queries, and tailoring the information to align with the users' specific needs. Three transit LLM applications are presented: Tweet Writer, Trip Advisor, and Policy Navigator. Tweet Writer automates updates to the transit system alerts on social media, Trip Advisor offers customized transit trip suggestions, and Policy Navigator provides clear and personalized answers to policy queries. Leveraging LLMs in these applications enhances seamless communication with their capabilities of understanding and generating human-like languages. With the help of these three LLM transit applications, transit system media personnel can provide system updates more efficiently, and customers can access travel information and policy answers in a more user-friendly manner."
                    },
                    {
                        "title": "DST-TransitNet: A Dynamic Spatio-Temporal Deep Learning Model for Scalable and Efficient Network-Wide Prediction of Station-Level Transit Ridership",
                        "abstract": "Accurate prediction of public transit ridership is vital for efficient planning and management of transit in rapidly growing urban areas in Canada. Unexpected increases in passengers can cause overcrowded vehicles, longer boarding times, and service disruptions. Traditional time series models like ARIMA and SARIMA face limitations, particularly in short-term predictions and integration of spatial and temporal features. These models struggle with the dynamic nature of ridership patterns and often ignore spatial correlations between nearby stops. Deep Learning (DL) models present a promising alternative, demonstrating superior performance in short-term prediction tasks by effectively capturing both spatial and temporal features. However, challenges such as dynamic spatial feature extraction, balancing accuracy with computational efficiency, and ensuring scalability remain.   This paper introduces DST-TransitNet, a hybrid DL model for system-wide station-level ridership prediction. This proposed model uses graph neural networks (GNN) and recurrent neural networks (RNN) to dynamically integrate the changing temporal and spatial correlations within the stations. The model also employs a precise time series decomposition framework to enhance accuracy and interpretability. Tested on Bogota's BRT system data, with three distinct social scenarios, DST-TransitNet outperformed state-of-the-art models in precision, efficiency and robustness. Meanwhile, it maintains stability over long prediction intervals, demonstrating practical applicability."
                    },
                    {
                        "title": "Transit Pulse: Utilizing Social Media as a Source for Customer Feedback and Information Extraction with Large Language Model",
                        "abstract": "Users of the transit system flood social networks daily with messages that contain valuable insights crucial for improving service quality. These posts help transit agencies quickly identify emerging issues. Parsing topics and sentiments is key to gaining comprehensive insights to foster service excellence. However, the volume of messages makes manual analysis impractical, and standard NLP techniques like Term Frequency-Inverse Document Frequency (TF-IDF) fall short in nuanced interpretation. Traditional sentiment analysis separates topics and sentiments before integrating them, often missing the interaction between them. This incremental approach complicates classification and reduces analytical productivity. To address these challenges, we propose a novel approach to extracting and analyzing transit-related information, including sentiment and sarcasm detection, identification of unusual system problems, and location data from social media. Our method employs Large Language Models (LLM), specifically Llama 3, for a streamlined analysis free from pre-established topic labels. To enhance the model's domain-specific knowledge, we utilize Retrieval-Augmented Generation (RAG), integrating external knowledge sources into the information extraction pipeline. We validated our method through extensive experiments comparing its performance with traditional NLP approaches on user tweet data from the real world transit system. Our results demonstrate the potential of LLMs to transform social media data analysis in the public transit domain, providing actionable insights and enhancing transit agencies' responsiveness by extracting a broader range of information."
                    },
                    {
                        "title": "Few-Shot Fine-Grained Action Recognition via Bidirectional Attention and Contrastive Meta-Learning",
                        "abstract": "Fine-grained action recognition is attracting increasing attention due to the emerging demand of specific action understanding in real-world applications, whereas the data of rare fine-grained categories is very limited. Therefore, we propose the few-shot fine-grained action recognition problem, aiming to recognize novel fine-grained actions with only few samples given for each class. Although progress has been made in coarse-grained actions, existing few-shot recognition methods encounter two issues handling fine-grained actions: the inability to capture subtle action details and the inadequacy in learning from data with low inter-class variance. To tackle the first issue, a human vision inspired bidirectional attention module (BAM) is proposed. Combining top-down task-driven signals with bottom-up salient stimuli, BAM captures subtle action details by accurately highlighting informative spatio-temporal regions. To address the second issue, we introduce contrastive meta-learning (CML). Compared with the widely adopted ProtoNet-based method, CML generates more discriminative video representations for low inter-class variance data, since it makes full use of potential contrastive pairs in each training episode. Furthermore, to fairly compare different models, we establish specific benchmark protocols on two large-scale fine-grained action recognition datasets. Extensive experiments show that our method consistently achieves state-of-the-art performance across evaluated tasks."
                    },
                    {
                        "title": "Research on gesture recognition method based on SEDCNN-SVM",
                        "abstract": "Gesture recognition based on surface electromyographic signal (sEMG) is one of the most used methods. The traditional manual feature extraction can only extract some low-level signal features, this causes poor classifier performance and low recognition accuracy when dealing with some complex signals. A recognition method, namely SEDCNN-SVM, is proposed to recognize sEMG of different gestures. SEDCNN-SVM consists of an improved deep convolutional neural network (DCNN) and a support vector machine (SVM). The DCNN can automatically extract and learn the feature information of sEMG through the convolution operation of the convolutional layer, so that it can capture the complex and high-level features in the data. The Squeeze and Excitation Networks (SE-Net) and the residual module were added to the model, so that the feature representation of each channel could be improved, the loss of feature information in convolutional operations was reduced, useful feature information was captured, and the problem of network gradient vanishing was eased. The SVM can improve the generalization ability and classification accuracy of the model by constructing an optimal hyperplane of the feature space. Hence, the SVM was used to replace the full connection layer and the Softmax function layer of the DCNN, the use of a suitable kernel function in SVM can improve the model's generalization ability and classification accuracy. To verify the effectiveness of the proposed classification algorithm, this method is analyzed and compared with other comparative classification methods. The recognition accuracy of SEDCNN-SVM can reach 0.955, it is significantly improved compared with other classification methods, the SEDCNN-SVM model is recognized online in real time."
                    },
                    {
                        "title": "Quantitatively Nonblocking Supervisory Control of Discrete-Event Systems",
                        "abstract": "In this paper, we propose two new nonblocking properties of automata as quantitative measures of maximal distances to marker states. The first property, called {\\em quantitative nonblockingness}, captures the practical requirement that at least one of the marker states (representing e.g. task completion) be reachable within a prescribed number of steps from any non-marker state and following any trajectory of the system. The second property, called {\\em heterogeneously quantitative nonblockingness}, distinguishes individual marker states and requires that each marker state be reached within a given bounded number of steps from any other state and following any trajectory of the system. Accordingly, we formulate two new problems of quantitatively nonblocking supervisory control and heterogeneously quantitatively nonblocking supervisory control, and characterize their solvabilities in terms of new concepts of quantitative language completability and heterogeneously quantitative language completability, respectively. It is proved that there exists the unique supremal (heterogeneously) quantitatively completable sublanguage of a given language, and we develop effective algorithms to compute the supremal sublanguages. Finally, combining with the algorithm of computing the supremal controllable sublanguage, we design algorithms to compute the maximally permissive solutions to the formulated (heterogeneously) quantitatively nonblocking supervisory control problems."
                    }
                ]
            },
            "b2cd7510-f987-4c36-90d8-a50401ef2360": {
                "pk": "b2cd7510-f987-4c36-90d8-a50401ef2360",
                "name": "Yan Chen",
                "collaborators": [
                    "C. S. Ting",
                    "Yan Song",
                    "Jie Lou",
                    "Chun Chen",
                    "Wei Yan"
                ],
                "domain": [
                    "Urban Planning",
                    "Quantum Physics",
                    "Superconductivity",
                    "Tensor Networks"
                ],
                "publications": [
                    {
                        "title": "From Measuring Land Use Mix to Measuring Land Use Pattern -- New Methods for Assessing Land Use",
                        "abstract": "Land use mix is one of the central concepts in the urban planning field, though its measure has been found to have many fallacies. In this study, we propose multiple alternative methods to the Conventional Shannon Entropy land use mix index, which is generally employed to measure land use diversity. The study attempted to measure land use mix from two new perspectives and test the performance of the new measures with economic, transportation, and social outcomes. We first offered an improved single measure, entropy-based land use mix index weighted by adding surrounding land use attributes and the regional land use context to mitigate the modifiable area unit problem (MAUP) and equal composition issue. Second, we designed a multi-measure method to describe land use patterns via clustering. We found that the new single measure method is more effective than the existing Conventional Shannon Entropy index in accurately delivering the vision of land use mix that Jane Jacobs originally proposed. We also found the multi-measure clustering method performed much better in the regression model on various effect variables, compared with both conventional and other new index. Upon further review, we found that land use pattern complexity's relationship with various outcomes makes a single definition measure very difficult to represent true land use fully. Therefore, we recommend that future research measure whether or not land use is mixed to then assess land use pattern."
                    },
                    {
                        "title": "Combining Grassmann algebra with entanglement renormalization method",
                        "abstract": "By combining the Grassmann algebra with multi-scale entanglement renormalization ansatz (MERA), we introduce a new unbiased and effective numerical method for simulating 2D strongly correlated electronic systems. The new GMERA method inherits all the advantages of MERA, which constructs the variational wave function based on complicated tensor network. Besides it can deal with fermionic properties of the system due to Grassmann algebra through local tensor contractions. This general method can treat different tensor network structures in a universal way. We show several benchmark calculations of the GMERA method, including the free fermion model, tight binding model, as well as the t-J model with hole doping."
                    },
                    {
                        "title": "Type-II Topological Meissner States",
                        "abstract": "We study the \\emph{orbital effects} of the synthetic magnetic fields in an interacting square lattice two-leg fermionic ladder model with the number-conserving pair hopping that hosting the Majorana bound states. By utilizing density matrix renormalization group and exact diagonalization, we identify a novel type-II topological Meissner $($\\emph{topo}-Meissner$)$ phase $($as distinguished from the low-field type-I \\emph{topo}-Meissner state$)$ when threading a \\emph{high} magnetic flux through the plaquette of the ladder, which not only exhibits a large uniformly circulating chiral current along the legs, the characteristics of the celebrated Meissner state, but also accommodates a topologically protected ground-state manifold due to the \\emph{reentrant} emergence of the Majorana end modes. Our work reveals some interesting interference effects resulting from the interplay between the gauge fields and the strong interactions in establishing the intrinsic topological states of matter in low-dimensional quantum systems."
                    },
                    {
                        "title": "Theory of Magnetic Field Induced Spin Density Wave in High Temperature Superconductors",
                        "abstract": "The induction of spin density wave (SDW) and charge density wave (CDW) orderings in the mixed state of high $T_c$ superconductors (HTS) is investigated by using the self-consistent Bogoliubov-de Gennes equations based upon an effective model Hamiltonian with competing SDW and d-wave superconductivity interactions. For optimized doping sample, the modulation of the induced SDW and its associated CDW is determined by the vortex lattice and their patterns obey the four-fold symmetry. By deceasing doping level, both SDW and CDW show quasi-one dimensional like behavior, and the CDW has a period just half that of the SDW along one direction. From the calculation of the local density of states (LDOS), we found that the majority of the quasi-particles inside the vortex core are localized. All these results are consistent with several recent experiments on HTS."
                    }
                ]
            },
            "f80a8c22-a4a2-46e2-98e9-c8586656bf09": {
                "pk": "f80a8c22-a4a2-46e2-98e9-c8586656bf09",
                "name": "Feng Tian",
                "collaborators": [
                    "Palle Jorgensen",
                    "Steen Pedersen",
                    "Palle E. T. Jorgensen"
                ],
                "domain": [
                    "Functional Analysis",
                    "Spectral Theory",
                    "Kernel Methods",
                    "Graph Theory"
                ],
                "publications": [
                    {
                        "title": "Scripting Relational Database Engine Using Transducer",
                        "abstract": "We allow database user to script a parallel relational database engine with a procedural language. Procedural language code is executed as a user defined relational query operator called transducer. Transducer is tightly integrated with relation engine, including query optimizer, query executor and can be executed in parallel like other query operators. With transducer, we can efficiently execute queries that are very difficult to express in SQL. As example, we show how to run time series and graph queries, etc, within a parallel relational database."
                    },
                    {
                        "title": "Spectral pairs and positive definite tempered distributions",
                        "abstract": "The present paper presents two new approaches to Fourier series and spectral analysis of singular measures."
                    },
                    {
                        "title": "Momentum Operators in The Unit Square",
                        "abstract": "We investigate the skew-adjoint extensions of a partial derivative operator acting in the direction of one of the sides a unit square. We investigate the unitary equivalence of such extensions and the spectra of such extensions. It follows from our results, that such extensions need not have discete spectrum. We apply our techniques to the problem of finding commuting skew-adjoint extensions of the partial derivative operators acting in the directions of the sides of the unit square. While our results are most easily stated for the unit square, they are established for a larger class of domains, including certain fractal domains."
                    },
                    {
                        "title": "von Neumann indices and classes of positive definite functions",
                        "abstract": "With view to applications, we establish a correspondence between two problems: (i) the problem of finding continuous positive definite extensions of functions $F$ which are defined on open bounded domains $\\Omega$ in $\\mathbb{R}$, on the one hand; and (ii) spectral theory for elliptic differential operators acting on $\\Omega$, (constant coefficients.) A novelty in our approach is the use of a reproducing kernel Hilbert space $\\mathscr{H}_{F}$ computed directly from $\\left(\\Omega,F\\right)$, as well as algorithms for computing relevant orthonormal bases in $\\mathscr{H}_{F}$."
                    },
                    {
                        "title": "Unbounded operators, Lie algebras, and local representations",
                        "abstract": "We prove a number of results on integrability and extendability of Lie algebras of unbounded skew-symmetric operators with common dense domain in Hilbert space. By integrability for a Lie algebra $\\mathfrak{g}$, we mean that there is an associated unitary representation $\\mathcal{U}$ of the corresponding simply connected Lie group such that $\\mathfrak{g}$ is the differential of $\\mathcal{U}$. Our results extend earlier integrability results in the literature; and are new even in the case of a single operator. Our applications include a new invariant for certain Riemann surfaces."
                    },
                    {
                        "title": "Dynamical properties of endomorphisms, multiresolutions, similarity-and orthogonality relations",
                        "abstract": "We study positive transfer operators $R$ in the setting of general measure spaces $\\left(X,\\mathscr{B}\\right)$. For each $R$, we compute associated path-space probability spaces $\\left(\\Omega,\\mathbb{P}\\right)$. When the transfer operator $R$ is compatible with an endomorphism in $\\left(X,\\mathscr{B}\\right)$, we get associated multiresolutions for the Hilbert spaces $L^{2}\\left(\\Omega,\\mathbb{P}\\right)$ where the path-space $\\Omega$ may then be taken to be a solenoid. Our multiresolutions include both orthogonality relations and self-similarity algorithms for standard wavelets and for generalized wavelet-resolutions. Applications are given to topological dynamics, ergodic theory, and spectral theory, in general; to iterated function systems (IFSs), and to Markov chains in particular."
                    },
                    {
                        "title": "Reproducing kernels and choices of associated feature spaces, in the form of $L^{2}$-spaces",
                        "abstract": "Motivated by applications to the study of stochastic processes, we introduce a new analysis of positive definite kernels $K$, their reproducing kernel Hilbert spaces (RKHS), and an associated family of feature spaces that may be chosen in the form $L^{2}\\left(\\mu\\right)$; and we study the question of which measures $\\mu$ are right for a particular kernel $K$. The answer to this depends on the particular application at hand. Such applications are the focus of the separate sections in the paper."
                    },
                    {
                        "title": "Superposition, reduction of multivariable problems, and approximation",
                        "abstract": "We study reduction schemes for functions of \"many\" variables into system of functions in one variable. Our setting includes infinite-dimensions. Following Cybenko-Kolmogorov, the outline for our results is as follows: We present explicit reductions schemes for multivariable problems, covering both a finite, and an infinite, number of variables. Starting with functions in \"many\" variables, we offer constructive reductions into superposition, with component terms, that make use of only functions in one variable, and specified choices of coordinate directions. Our proofs are transform based, using explicit transforms, Fourier and Radon; as well as multivariable Shannon interpolation."
                    },
                    {
                        "title": "Graph Laplacians and discrete reproducing kernel Hilbert spaces from restrictions",
                        "abstract": "We study kernel functions, and associated reproducing kernel Hilbert spaces $\\mathscr{H}$ over infinite, discrete and countable sets $V$. Numerical analysis builds discrete models (e.g., finite element) for the purpose of finding approximate solutions to boundary value problems; using multiresolution-subdivision schemes in continuous domains. In this paper, we turn the tables: our object of study is realistic infinite discrete models in their own right; and we then use an analysis of suitable continuous counterpart problems, but now serving as a tool for obtaining solutions in the discrete world."
                    },
                    {
                        "title": "Induced representations arising from a character with finite orbit in a semidirect product",
                        "abstract": "Making use of a unified approach to certain classes of induced representations, we establish here a number of detailed spectral theoretic decomposition results. They apply to specific problems from non-commutative harmonic analysis, ergodic theory, and dynamical systems. Our analysis is in the setting of semidirect products, discrete subgroups, and solenoids. Our applications include analysis and ergodic theory of Bratteli diagrams and their compact duals; of wavelet sets, and wavelet representations."
                    },
                    {
                        "title": "Sampling with positive definite kernels and an associated dichotomy",
                        "abstract": "We study classes of reproducing kernels $K$ on general domains; these are kernels which arise commonly in machine learning models; models based on certain families of reproducing kernel Hilbert spaces. They are the positive definite kernels $K$ with the property that there are countable discrete sample-subsets $S$; i.e., proper subsets $S$ having the property that every function in $\\mathscr{H}\\left(K\\right)$ admits an $S$-sample representation. We give a characterizations of kernels which admit such non-trivial countable discrete sample-sets. A number of applications and concrete kernels are given in the second half of the paper."
                    },
                    {
                        "title": "Functional Analysis",
                        "abstract": "Notes from a course taught by Palle Jorgensen in the fall semester of 2009. The course covered central themes in functional analysis and operator theory, with an emphasis on topics of special relevance to such applications as representation theory, harmonic analysis, mathematical physics, and stochastic integration."
                    },
                    {
                        "title": "Infinite networks and variation of conductance functions in discrete Laplacians",
                        "abstract": "For a given infinite connected graph $G=(V,E)$ and an arbitrary but fixed conductance function $c$, we study an associated graph Laplacian $\\Delta_{c}$; it is a generalized difference operator where the differences are measured across the edges $E$ in $G$; and the conductance function $c$ represents the corresponding coefficients. The graph Laplacian (a key tool in the study of infinite networks) acts in an energy Hilbert space $\\mathscr{H}_{E}$ computed from $c$. Using a certain Parseval frame, we study the spectral theoretic properties of graph Laplacians. In fact, for fixed $c$, there are two versions of the graph Laplacian, one defined naturally in the $l^{2}$ space of $V$, and the other in $\\mathscr{H}_{E}$. The first is automatically selfadjoint, but the second involves a Krein extension. We prove that, as sets, the two spectra are the same, aside from the point 0. The point zero may be in the spectrum of the second, but not the first.   We further study the fine structure of the respective spectra as the conductance function varies; showing now how the spectrum changes subject to variations in the function $c$."
                    }
                ]
            },
            "000e0353-80e9-416f-9598-f3d5733565b2": {
                "pk": "000e0353-80e9-416f-9598-f3d5733565b2",
                "name": "Mengmeng Wang",
                "collaborators": [
                    "Yong Liu",
                    "Jiazheng Xing",
                    "Chungang Shi",
                    "Yan Lv",
                    "Wei Wang",
                    "Qinghua Chen",
                    "Yan Wang",
                    "Xiaomeng Li",
                    "Zeyi Huang",
                    "Xiaoqian Yang"
                ],
                "domain": [
                    "Object Tracking",
                    "Action Recognition",
                    "Language Processing",
                    "Energy Harvesting"
                ],
                "publications": [
                    {
                        "title": "Robust Object Tracking with a Hierarchical Ensemble Framework",
                        "abstract": "Autonomous robots enjoy a wide popularity nowadays and have been applied in many applications, such as home security, entertainment, delivery, navigation and guidance. It is vital to robots to track objects accurately in these applications, so it is necessary to focus on tracking algorithms to improve the robustness and accuracy. In this paper, we propose a robust object tracking algorithm based on a hierarchical ensemble framework which can incorporate information including individual pixel features, local patches and holistic target models. The framework combines multiple ensemble models simultaneously instead of using a single ensemble model individually. A discriminative model which accounts for the matching degree of local patches is adopted via a bottom ensemble layer, and a generative model which exploits holistic templates is used to search for the object through the middle ensemble layer as well as an adaptive Kalman filter. We test the proposed tracker on challenging benchmark image sequences. Both qualitative and quantitative evaluations demonstrate that the proposed tracker performs superiorly against several state-of-the-art algorithms, especially when the appearance changes dramatically and the occlusions occur."
                    },
                    {
                        "title": "The small mass limit for a McKean-Vlasov equation with state-dependent friction",
                        "abstract": "The small mass limit is derived for a McKean-Vlasov equation with state-dependent friction in $d$-dimensional space. By applying the averaging approach to a non-autonomous slow-fast system with the microscopic and macroscopic scales, the convergence in distribution is obtained."
                    },
                    {
                        "title": "The 'Letter' Distribution in the Chinese Language",
                        "abstract": "Corpus-based statistical analysis plays a significant role in linguistic research, and ample evidence has shown that different languages exhibit some common laws. Studies have found that letters in some alphabetic writing languages have strikingly similar statistical usage frequency distributions. Does this hold for Chinese, which employs ideogram writing? We obtained letter frequency data of some alphabetic writing languages and found the common law of the letter distributions. In addition, we collected Chinese literature corpora for different historical periods from the Tang Dynasty to the present, and we dismantled the Chinese written language into three kinds of basic particles: characters, strokes and constructive parts. The results of the statistical analysis showed that, in different historical periods, the intensity of the use of basic particles in Chinese writing varied, but the form of the distribution was consistent. In particular, the distributions of the Chinese constructive parts are certainly consistent with those alphabetic writing languages. This study provides new evidence of the consistency of human languages."
                    },
                    {
                        "title": "Large Margin Object Tracking with Circulant Feature Maps",
                        "abstract": "Structured output support vector machine (SVM) based tracking algorithms have shown favorable performance recently. Nonetheless, the time-consuming candidate sampling and complex optimization limit their real-time applications. In this paper, we propose a novel large margin object tracking method which absorbs the strong discriminative ability from structured output SVM and speeds up by the correlation filter algorithm significantly. Secondly, a multimodal target detection technique is proposed to improve the target localization precision and prevent model drift introduced by similar objects or background noise. Thirdly, we exploit the feedback from high-confidence tracking results to avoid the model corruption problem. We implement two versions of the proposed tracker with the representations from both conventional hand-crafted and deep convolution neural networks (CNNs) based features to validate the strong compatibility of the algorithm. The experimental results demonstrate that the proposed tracker performs superiorly against several state-of-the-art algorithms on the challenging benchmark sequences while runs at speed in excess of 80 frames per second. The source code and experimental results will be made publicly available."
                    },
                    {
                        "title": "ActionCLIP: A New Paradigm for Video Action Recognition",
                        "abstract": "The canonical approach to video action recognition dictates a neural model to do a classic and standard 1-of-N majority vote task. They are trained to predict a fixed set of predefined categories, limiting their transferable ability on new datasets with unseen concepts. In this paper, we provide a new perspective on action recognition by attaching importance to the semantic information of label texts rather than simply mapping them into numbers. Specifically, we model this task as a video-text matching problem within a multimodal learning framework, which strengthens the video representation with more semantic language supervision and enables our model to do zero-shot action recognition without any further labeled data or parameters requirements. Moreover, to handle the deficiency of label texts and make use of tremendous web data, we propose a new paradigm based on this multimodal learning framework for action recognition, which we dub \"pre-train, prompt and fine-tune\". This paradigm first learns powerful representations from pre-training on a large amount of web image-text or video-text data. Then it makes the action recognition task to act more like pre-training problems via prompt engineering. Finally, it end-to-end fine-tunes on target datasets to obtain strong performance. We give an instantiation of the new paradigm, ActionCLIP, which not only has superior and flexible zero-shot/few-shot transfer ability but also reaches a top performance on general action recognition task, achieving 83.8% top-1 accuracy on Kinetics-400 with a ViT-B/16 as the backbone. Code is available at https://github.com/sallymmx/ActionCLIP.git"
                    },
                    {
                        "title": "Explicitly Modeling the Discriminability for Instance-Aware Visual Object Tracking",
                        "abstract": "Visual object tracking performance has been dramatically improved in recent years, but some severe challenges remain open, like distractors and occlusions. We suspect the reason is that the feature representations of the tracking targets are only expressively learned but not fully discriminatively modeled. In this paper, we propose a novel Instance-Aware Tracker (IAT) to explicitly excavate the discriminability of feature representations, which improves the classical visual tracking pipeline with an instance-level classifier. First, we introduce a contrastive learning mechanism to formulate the classification task, ensuring that every training sample could be uniquely modeled and be highly distinguishable from plenty of other samples. Besides, we design an effective negative sample selection scheme to contain various intra and inter classes in the instance classification branch. Furthermore, we implement two variants of the proposed IAT, including a video-level one and an object-level one. They realize the concept of \\textbf{instance} in different granularity as videos and target bounding boxes, respectively. The former enhances the ability to recognize the target from the background while the latter boosts the discriminative power for mitigating the target-distractor dilemma. Extensive experimental evaluations on 8 benchmark datasets show that both two versions of the proposed IAT achieve leading results against state-of-the-art methods while running at 30FPS. Code will be available when it is published."
                    },
                    {
                        "title": "Extended Feature Pyramid Network for Small Object Detection",
                        "abstract": "Small object detection remains an unsolved challenge because it is hard to extract information of small objects with only a few pixels. While scale-level corresponding detection in feature pyramid network alleviates this problem, we find feature coupling of various scales still impairs the performance of small objects. In this paper, we propose extended feature pyramid network (EFPN) with an extra high-resolution pyramid level specialized for small object detection. Specifically, we design a novel module, named feature texture transfer (FTT), which is used to super-resolve features and extract credible regional details simultaneously. Moreover, we design a foreground-background-balanced loss function to alleviate area imbalance of foreground and background. In our experiments, the proposed EFPN is efficient on both computation and memory, and yields state-of-the-art results on small traffic-sign dataset Tsinghua-Tencent 100K and small category of general object detection dataset MS COCO."
                    },
                    {
                        "title": "BSNet: Lane Detection via Draw B-spline Curves Nearby",
                        "abstract": "Curve-based methods are one of the classic lane detection methods. They learn the holistic representation of lane lines, which is intuitive and concise. However, their performance lags behind the recent state-of-the-art methods due to the limitation of their lane representation and optimization. In this paper, we revisit the curve-based lane detection methods from the perspectives of the lane representations' globality and locality. The globality of lane representation is the ability to complete invisible parts of lanes with visible parts. The locality of lane representation is the ability to modify lanes locally which can simplify parameter optimization. Specifically, we first propose to exploit the b-spline curve to fit lane lines since it meets the locality and globality. Second, we design a simple yet efficient network BSNet to ensure the acquisition of global and local features. Third, we propose a new curve distance to make the lane detection optimization objective more reasonable and alleviate ill-conditioned problems. The proposed methods achieve state-of-the-art performance on the Tusimple, CULane, and LLAMAS datasets, which dramatically improved the accuracy of curve-based methods in the lane detection task while running far beyond real-time (197FPS)."
                    },
                    {
                        "title": "Stochastic Duty Cycling for Heterogenous Energy Harvesting Networks",
                        "abstract": "In recent years, there have been several kinds of energy harvesting networks containing some tiny devices, such as ambient backscatter, ring and renewable sensor networks. During energy harvesting, such networks suffer from the energy heterogeneity, dynamics and prediction hardness because the access to natural resources is often spatiotemporal different and timely changing among the devices. Meanwhile, the charging efficiency is quite low especially when the power of the harvested energy is weak. It results in the energy waste to store the harvested energy indirectly. These features bring challenging and interesting issues on efficient allocation of the harvested energy. This paper studies the \\emph{stochastic duty cycling} by considering these features with the objective characterized by maximizing the common active time. We consider two cases: offline and online stochastic duty cycling. For the offline case, we design an optimal solution: offline duty cycling algorithm. For the online case, we design an online duty cycling algorithm, which achieves the approximation ratio with at least $1-e^{-\\gamma^2}$, where $\\gamma$ is the probability able to harvest energy. We also evaluate our algorithms with the experiment on a real energy harvesting network. The experiment results show that the performance of the online algorithm can be very close to the offline algorithm."
                    },
                    {
                        "title": "MaIL: A Unified Mask-Image-Language Trimodal Network for Referring Image Segmentation",
                        "abstract": "Referring image segmentation is a typical multi-modal task, which aims at generating a binary mask for referent described in given language expressions. Prior arts adopt a bimodal solution, taking images and languages as two modalities within an encoder-fusion-decoder pipeline. However, this pipeline is sub-optimal for the target task for two reasons. First, they only fuse high-level features produced by uni-modal encoders separately, which hinders sufficient cross-modal learning. Second, the uni-modal encoders are pre-trained independently, which brings inconsistency between pre-trained uni-modal tasks and the target multi-modal task. Besides, this pipeline often ignores or makes little use of intuitively beneficial instance-level features. To relieve these problems, we propose MaIL, which is a more concise encoder-decoder pipeline with a Mask-Image-Language trimodal encoder. Specifically, MaIL unifies uni-modal feature extractors and their fusion model into a deep modality interaction encoder, facilitating sufficient feature interaction across different modalities. Meanwhile, MaIL directly avoids the second limitation since no uni-modal encoders are needed anymore. Moreover, for the first time, we propose to introduce instance masks as an additional modality, which explicitly intensifies instance-level features and promotes finer segmentation results. The proposed MaIL set a new state-of-the-art on all frequently-used referring image segmentation datasets, including RefCOCO, RefCOCO+, and G-Ref, with significant gains, 3%-10% against previous best methods. Code will be released soon."
                    },
                    {
                        "title": "Revisiting the Spatial and Temporal Modeling for Few-shot Action Recognition",
                        "abstract": "Spatial and temporal modeling is one of the most core aspects of few-shot action recognition. Most previous works mainly focus on long-term temporal relation modeling based on high-level spatial representations, without considering the crucial low-level spatial features and short-term temporal relations. Actually, the former feature could bring rich local semantic information, and the latter feature could represent motion characteristics of adjacent frames, respectively. In this paper, we propose SloshNet, a new framework that revisits the spatial and temporal modeling for few-shot action recognition in a finer manner. First, to exploit the low-level spatial features, we design a feature fusion architecture search module to automatically search for the best combination of the low-level and high-level spatial features. Next, inspired by the recent transformer, we introduce a long-term temporal modeling module to model the global temporal relations based on the extracted spatial appearance features. Meanwhile, we design another short-term temporal modeling module to encode the motion characteristics between adjacent frame representations. After that, the final predictions can be obtained by feeding the embedded rich spatial-temporal features to a common frame-level class prototype matcher. We extensively validate the proposed SloshNet on four few-shot action recognition datasets, including Something-Something V2, Kinetics, UCF101, and HMDB51. It achieves favorable results against state-of-the-art methods in all datasets."
                    },
                    {
                        "title": "LooGLE: Can Long-Context Language Models Understand Long Contexts?",
                        "abstract": "Large language models (LLMs), despite their impressive performance in various language tasks, are typically limited to processing texts within context-window size. This limitation has spurred significant research efforts to enhance LLMs' long-context understanding with high-quality long-sequence benchmarks. However, prior datasets in this regard suffer from shortcomings, such as short context length compared to the context window of modern LLMs; outdated documents that have data leakage problems; and an emphasis on short dependency tasks rather than long dependency tasks. In this paper, we present LooGLE, a Long Context Generic Language Evaluation benchmark for LLMs' long context understanding. LooGLE features relatively new documents post-2022, with over 24,000 tokens per document and 6,000 newly generated questions spanning diverse domains. Human annotators meticulously crafted more than 1,100 high-quality question-answer pairs to meet the long dependency requirements. These pairs underwent thorough cross-validation, yielding the most precise assessment of LLMs' long dependency capabilities. The evaluation of eight state-of-the-art LLMs on LooGLE revealed key findings: (i) commercial models outperformed open-sourced models; (ii) LLMs excelled in short dependency tasks like short question-answering and cloze tasks but struggled with more intricate long dependency tasks; (iii) in-context learning and chaining thoughts offered only marginal improvements; (iv) retrieval-based techniques demonstrated substantial benefits for short question-answering, while strategies for extending context window length had limited impact on long context understanding. As such, LooGLE not only provides a systematic and comprehensive evaluation schema on long-context LLMs, but also sheds light on future development of enhanced models towards \"true long-context understanding\"."
                    }
                ]
            },
            "add22d55-09f4-4251-bf36-3d7d08e027e4": {
                "pk": "add22d55-09f4-4251-bf36-3d7d08e027e4",
                "name": "Guang Dai",
                "collaborators": [
                    "Zhihua Zhang",
                    "Michael I. Jordan"
                ],
                "domain": [
                    "Support Vector Machines",
                    "Machine Learning",
                    "Optimization",
                    "Classification"
                ],
                "publications": [
                    {
                        "title": "Coherence Functions with Applications in Large-Margin Classification Methods",
                        "abstract": "Support vector machines (SVMs) naturally embody sparseness due to their use of hinge loss functions. However, SVMs can not directly estimate conditional class probabilities. In this paper we propose and study a family of coherence functions, which are convex and differentiable, as surrogates of the hinge function. The coherence function is derived by using the maximum-entropy principle and is characterized by a temperature parameter. It bridges the hinge function and the logit function in logistic regression. The limit of the coherence function at zero temperature corresponds to the hinge function, and the limit of the minimizer of its expected error is the minimizer of the expected error of the hinge loss. We refer to the use of the coherence function in large-margin classification as C-learning, and we present efficient coordinate descent algorithms for the training of regularized ${\\cal C}$-learning models."
                    }
                ]
            },
            "09308843-0eb0-41b5-a05b-5b395355b5ac": {
                "pk": "09308843-0eb0-41b5-a05b-5b395355b5ac",
                "name": "Qianying Wang",
                "collaborators": [
                    "Siling Guo",
                    "Tianchen Sun",
                    "Jiangtao Gong",
                    "Zhicong Lu",
                    "Liuxin Zhang"
                ],
                "domain": [
                    "Education Technology",
                    "ICT4D",
                    "Co-Teaching",
                    "Remote Learning"
                ],
                "publications": [
                    {
                        "title": "Remote Co-teaching in Rural Classroom: Current Practices, Impacts, and Challenges",
                        "abstract": "The shortage of high-quality teachers is one of the biggest educational problems faced by underdeveloped areas. With the development of information and communication technologies (ICTs), China has begun a remote co-teaching intervention program using ICTs for rural classes, forming a unique co-teaching classroom. We conducted semi-structured interviews with nine remote urban teachers and twelve local rural teachers. We identified the remote co-teaching classes standard practices and co-teachers collaborative work process. We also found that remote teachers high-quality class directly impacted local teachers and students. Furthermore, interestingly, local teachers were also actively involved in making indirect impacts on their students by deeply coordinating with remote teachers and adapting the resources offered by the remote teachers. We conclude by summarizing and discussing the challenges faced by teachers, lessons learned from the current program, and related design implications to achieve a more adaptive and sustainable ICT4D program design."
                    }
                ]
            },
            "1369ce5a-bb37-4e27-9b64-e888b821c5d6": {
                "pk": "1369ce5a-bb37-4e27-9b64-e888b821c5d6",
                "name": "Jingdong Wang",
                "collaborators": [
                    "Ting Zhang",
                    "Yuhui Yuan",
                    "Xiaokang Chen",
                    "Wenjun Zeng",
                    "Xilin Chen",
                    "Gang Zeng",
                    "Yifan Liu",
                    "Chunhua Shen",
                    "Changqian Yu",
                    "Chao Du"
                ],
                "domain": [
                    "Deep Learning",
                    "Computer Vision",
                    "Nearest Neighbor Search",
                    "Semantic Segmentation"
                ],
                "publications": [
                    {
                        "title": "Inner Product Similarity Search using Compositional Codes",
                        "abstract": "This paper addresses the nearest neighbor search problem under inner product similarity and introduces a compact code-based approach. The idea is to approximate a vector using the composition of several elements selected from a source dictionary and to represent this vector by a short code composed of the indices of the selected elements. The inner product between a query vector and a database vector is efficiently estimated from the query vector and the short code of the database vector. We show the superior performance of the proposed group $M$-selection algorithm that selects $M$ elements from $M$ source dictionaries for vector approximation in terms of search accuracy and efficiency for compact codes of the same length via theoretical and empirical analysis. Experimental results on large-scale datasets ($1M$ and $1B$ SIFT features, $1M$ linear models and Netflix) demonstrate the superiority of the proposed approach."
                    },
                    {
                        "title": "Composite Quantization",
                        "abstract": "This paper studies the compact coding approach to approximate nearest neighbor search. We introduce a composite quantization framework. It uses the composition of several ($M$) elements, each of which is selected from a different dictionary, to accurately approximate a $D$-dimensional vector, thus yielding accurate search, and represents the data vector by a short code composed of the indices of the selected elements in the corresponding dictionaries. Our key contribution lies in introducing a near-orthogonality constraint, which makes the search efficiency is guaranteed as the cost of the distance computation is reduced to $O(M)$ from $O(D)$ through a distance table lookup scheme. The resulting approach is called near-orthogonal composite quantization. We theoretically justify the equivalence between near-orthogonal composite quantization and minimizing an upper bound of a function formed by jointly considering the quantization error and the search cost according to a generalized triangle inequality. We empirically show the efficacy of the proposed approach over several benchmark datasets. In addition, we demonstrate the superior performances in other three applications: combination with inverted multi-index, quantizing the query for mobile search, and inner-product similarity search."
                    },
                    {
                        "title": "Collaborative Quantization for Cross-Modal Similarity Search",
                        "abstract": "Cross-modal similarity search is a problem about designing a search system supporting querying across content modalities, e.g., using an image to search for texts or using a text to search for images. This paper presents a compact coding solution for efficient search, with a focus on the quantization approach which has already shown the superior performance over the hashing solutions in the single-modal similarity search. We propose a cross-modal quantization approach, which is among the early attempts to introduce quantization into cross-modal search. The major contribution lies in jointly learning the quantizers for both modalities through aligning the quantized representations for each pair of image and text belonging to a document. In addition, our approach simultaneously learns the common space for both modalities in which quantization is conducted to enable efficient and effective search using the Euclidean distance computed in the common space with fast distance table lookup. Experimental results compared with several competitive algorithms over three benchmark datasets demonstrate that the proposed approach achieves the state-of-the-art performance."
                    },
                    {
                        "title": "Admix: Enhancing the Transferability of Adversarial Attacks",
                        "abstract": "Deep neural networks are known to be extremely vulnerable to adversarial examples under white-box setting. Moreover, the malicious adversaries crafted on the surrogate (source) model often exhibit black-box transferability on other models with the same learning task but having different architectures. Recently, various methods are proposed to boost the adversarial transferability, among which the input transformation is one of the most effective approaches. We investigate in this direction and observe that existing transformations are all applied on a single image, which might limit the adversarial transferability. To this end, we propose a new input transformation based attack method called Admix that considers the input image and a set of images randomly sampled from other categories. Instead of directly calculating the gradient on the original input, Admix calculates the gradient on the input image admixed with a small portion of each add-in image while using the original label of the input to craft more transferable adversaries. Empirical evaluations on standard ImageNet dataset demonstrate that Admix could achieve significantly better transferability than existing input transformation methods under both single model setting and ensemble-model setting. By incorporating with existing input transformations, our method could further improve the transferability and outperforms the state-of-the-art combination of input transformations by a clear margin when attacking nine advanced defense models under ensemble-model setting. Code is available at https://github.com/JHL-HUST/Admix."
                    },
                    {
                        "title": "Deeply-Fused Nets",
                        "abstract": "In this paper, we present a novel deep learning approach, deeply-fused nets. The central idea of our approach is deep fusion, i.e., combine the intermediate representations of base networks, where the fused output serves as the input of the remaining part of each base network, and perform such combinations deeply over several intermediate representations. The resulting deeply fused net enjoys several benefits. First, it is able to learn multi-scale representations as it enjoys the benefits of more base networks, which could form the same fused network, other than the initial group of base networks. Second, in our suggested fused net formed by one deep and one shallow base networks, the flows of the information from the earlier intermediate layer of the deep base network to the output and from the input to the later intermediate layer of the deep base network are both improved. Last, the deep and shallow base networks are jointly learnt and can benefit from each other. More interestingly, the essential depth of a fused net composed from a deep base network and a shallow base network is reduced because the fused net could be composed from a less deep base network, and thus training the fused net is less difficult than training the initial deep base network. Empirical results demonstrate that our approach achieves superior performance over two closely-related methods, ResNet and Highway, and competitive performance compared to the state-of-the-arts."
                    },
                    {
                        "title": "Deeply-Learned Part-Aligned Representations for Person Re-Identification",
                        "abstract": "In this paper, we address the problem of person re-identification, which refers to associating the persons captured from different cameras. We propose a simple yet effective human part-aligned representation for handling the body part misalignment problem. Our approach decomposes the human body into regions (parts) which are discriminative for person matching, accordingly computes the representations over the regions, and aggregates the similarities computed between the corresponding regions of a pair of probe and gallery images as the overall matching score. Our formulation, inspired by attention models, is a deep neural network modeling the three steps together, which is learnt through minimizing the triplet loss function without requiring body part labeling information. Unlike most existing deep learning algorithms that learn a global or spatial partition-based local representation, our approach performs human body partition, and thus is more robust to pose changes and various human spatial distributions in the person bounding box. Our approach shows state-of-the-art results over standard datasets, Market-$1501$, CUHK$03$, CUHK$01$ and VIPeR."
                    },
                    {
                        "title": "Weakly-Supervised Action Localization by Generative Attention Modeling",
                        "abstract": "Weakly-supervised temporal action localization is a problem of learning an action localization model with only video-level action labeling available. The general framework largely relies on the classification activation, which employs an attention model to identify the action-related frames and then categorizes them into different classes. Such method results in the action-context confusion issue: context frames near action clips tend to be recognized as action frames themselves, since they are closely related to the specific classes. To solve the problem, in this paper we propose to model the class-agnostic frame-wise probability conditioned on the frame attention using conditional Variational Auto-Encoder (VAE). With the observation that the context exhibits notable difference from the action at representation level, a probabilistic model, i.e., conditional VAE, is learned to model the likelihood of each frame given the attention. By maximizing the conditional probability with respect to the attention, the action and non-action frames are well separated. Experiments on THUMOS14 and ActivityNet1.2 demonstrate advantage of our method and effectiveness in handling action-context confusion problem. Code is now available on GitHub."
                    },
                    {
                        "title": "SegFix: Model-Agnostic Boundary Refinement for Segmentation",
                        "abstract": "We present a model-agnostic post-processing scheme to improve the boundary quality for the segmentation result that is generated by any existing segmentation model. Motivated by the empirical observation that the label predictions of interior pixels are more reliable, we propose to replace the originally unreliable predictions of boundary pixels by the predictions of interior pixels. Our approach processes only the input image through two steps: (i) localize the boundary pixels and (ii) identify the corresponding interior pixel for each boundary pixel. We build the correspondence by learning a direction away from the boundary pixel to an interior pixel. Our method requires no prior information of the segmentation models and achieves nearly real-time speed. We empirically verify that our SegFix consistently reduces the boundary errors for segmentation results generated from various state-of-the-art models on Cityscapes, ADE20K and GTA5. Code is available at: https://github.com/openseg-group/openseg.pytorch."
                    },
                    {
                        "title": "Distillation Guided Residual Learning for Binary Convolutional Neural Networks",
                        "abstract": "It is challenging to bridge the performance gap between Binary CNN (BCNN) and Floating point CNN (FCNN). We observe that, this performance gap leads to substantial residuals between intermediate feature maps of BCNN and FCNN. To minimize the performance gap, we enforce BCNN to produce similar intermediate feature maps with the ones of FCNN. This training strategy, i.e., optimizing each binary convolutional block with block-wise distillation loss derived from FCNN, leads to a more effective optimization to BCNN. It also motivates us to update the binary convolutional block architecture to facilitate the optimization of block-wise distillation loss. Specifically, a lightweight shortcut branch is inserted into each binary convolutional block to complement residuals at each block. Benefited from its Squeeze-and-Interaction (SI) structure, this shortcut branch introduces a fraction of parameters, e.g., 10\\% overheads, but effectively complements the residuals. Extensive experiments on ImageNet demonstrate the superior performance of our method in both classification efficiency and accuracy, e.g., BCNN trained with our methods achieves the accuracy of 60.45\\% on ImageNet."
                    },
                    {
                        "title": "Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision",
                        "abstract": "In this paper, we study the semi-supervised semantic segmentation problem via exploring both labeled data and extra unlabeled data. We propose a novel consistency regularization approach, called cross pseudo supervision (CPS). Our approach imposes the consistency on two segmentation networks perturbed with different initialization for the same input image. The pseudo one-hot label map, output from one perturbed segmentation network, is used to supervise the other segmentation network with the standard cross-entropy loss, and vice versa. The CPS consistency has two roles: encourage high similarity between the predictions of two perturbed networks for the same input image, and expand training data by using the unlabeled data with pseudo labels. Experiment results show that our approach achieves the state-of-the-art semi-supervised segmentation performance on Cityscapes and PASCAL VOC 2012. Code is available at https://git.io/CPS."
                    },
                    {
                        "title": "Rating the Crisis of Online Public Opinion Using a Multi-Level Index System",
                        "abstract": "Online public opinion usually spreads rapidly and widely, thus a small incident probably evolves into a large social crisis in a very short time, and results in a heavy loss in credit or economic aspects. We propose a method to rate the crisis of online public opinion based on a multi-level index system to evaluate the impact of events objectively. Firstly, the dissemination mechanism of online public opinion is explained from the perspective of information ecology. According to the mechanism, some evaluation indexes are selected through correlation analysis and principal component analysis. Then, a classification model of text emotion is created via the training by deep learning to achieve the accurate quantification of the emotional indexes in the index system. Finally, based on the multi-level evaluation index system and grey correlation analysis, we propose a method to rate the crisis of online public opinion. The experiment with the real-time incident show that this method can objectively evaluate the emotional tendency of Internet users and rate the crisis in different dissemination stages of online public opinion. It is helpful to realizing the crisis warning of online public opinion and timely blocking the further spread of the crisis."
                    },
                    {
                        "title": "Assessing Model Generalization in Vicinity",
                        "abstract": "This paper evaluates the generalization ability of classification models on out-of-distribution test sets without depending on ground truth labels. Common approaches often calculate an unsupervised metric related to a specific model property, like confidence or invariance, which correlates with out-of-distribution accuracy. However, these metrics are typically computed for each test sample individually, leading to potential issues caused by spurious model responses, such as overly high or low confidence. To tackle this challenge, we propose incorporating responses from neighboring test samples into the correctness assessment of each individual sample. In essence, if a model consistently demonstrates high correctness scores for nearby samples, it increases the likelihood of correctly predicting the target sample, and vice versa. The resulting scores are then averaged across all test samples to provide a holistic indication of model accuracy. Developed under the vicinal risk formulation, this approach, named vicinal risk proxy (VRP), computes accuracy without relying on labels. We show that applying the VRP method to existing generalization indicators, such as average confidence and effective invariance, consistently improves over these baselines both methodologically and experimentally. This yields a stronger correlation with model accuracy, especially on challenging out-of-distribution test sets."
                    },
                    {
                        "title": "Efficient Semantic Video Segmentation with Per-frame Inference",
                        "abstract": "For semantic segmentation, most existing real-time deep models trained with each frame independently may produce inconsistent results for a video sequence. Advanced methods take into considerations the correlations in the video sequence, e.g., by propagating the results to the neighboring frames using optical flow, or extracting the frame representations with other frames, which may lead to inaccurate results or unbalanced latency. In this work, we process efficient semantic video segmentation in a per-frame fashion during the inference process. Different from previous per-frame models, we explicitly consider the temporal consistency among frames as extra constraints during the training process and embed the temporal consistency into the segmentation network. Therefore, in the inference process, we can process each frame independently with no latency, and improve the temporal consistency with no extra computational cost and post-processing. We employ compact models for real-time execution. To narrow the performance gap between compact models and large models, new knowledge distillation methods are designed. Our results outperform previous keyframe based methods with a better trade-off between the accuracy and the inference speed on popular benchmarks, including the Cityscapes and Camvid. The temporal consistency is also improved compared with corresponding baselines which are trained with each frame independently. Code is available at: https://tinyurl.com/segment-video"
                    },
                    {
                        "title": "Efficient Video Segmentation Models with Per-frame Inference",
                        "abstract": "Most existing real-time deep models trained with each frame independently may produce inconsistent results across the temporal axis when tested on a video sequence. A few methods take the correlations in the video sequence into account,e.g., by propagating the results to the neighboring frames using optical flow or extracting frame representations using multi-frame information, which may lead to inaccurate results or unbalanced latency. In this work, we focus on improving the temporal consistency without introducing computation overhead in inference. To this end, we perform inference at each frame. Temporal consistency is achieved by learning from video frames with extra constraints during the training phase. introduced for inference. We propose several techniques to learn from the video sequence, including a temporal consistency loss and online/offline knowledge distillation methods. On the task of semantic video segmentation, weighing among accuracy, temporal smoothness, and efficiency, our proposed method outperforms keyframe-based methods and a few baseline methods that are trained with each frame independently, on datasets including Cityscapes, Camvid, and 300VW-Mask. We further apply our training method to video instance segmentation on YouTubeVISand develop an application of portrait matting in video sequences, by segmenting temporally consistent instance-level trimaps across frames. Experiments show superior qualitative and quantitative results. Code is available at: https://git.io/vidseg."
                    },
                    {
                        "title": "Conditional DETR V2: Efficient Detection Transformer with Box Queries",
                        "abstract": "In this paper, we are interested in Detection Transformer (DETR), an end-to-end object detection approach based on a transformer encoder-decoder architecture without hand-crafted postprocessing, such as NMS. Inspired by Conditional DETR, an improved DETR with fast training convergence, that presented box queries (originally called spatial queries) for internal decoder layers, we reformulate the object query into the format of the box query that is a composition of the embeddings of the reference point and the transformation of the box with respect to the reference point. This reformulation indicates the connection between the object query in DETR and the anchor box that is widely studied in Faster R-CNN. Furthermore, we learn the box queries from the image content, further improving the detection quality of Conditional DETR still with fast training convergence. In addition, we adopt the idea of axial self-attention to save the memory cost and accelerate the encoder. The resulting detector, called Conditional DETR V2, achieves better results than Conditional DETR, saves the memory cost and runs more efficiently. For example, for the DC$5$-ResNet-$50$ backbone, our approach achieves $44.8$ AP with $16.4$ FPS on the COCO $val$ set and compared to Conditional DETR, it runs $1.6\\times$ faster, saves $74$\\% of the overall memory cost, and improves $1.0$ AP score."
                    },
                    {
                        "title": "Segmentation Transformer: Object-Contextual Representations for Semantic Segmentation",
                        "abstract": "In this paper, we address the semantic segmentation problem with a focus on the context aggregation strategy. Our motivation is that the label of a pixel is the category of the object that the pixel belongs to. We present a simple yet effective approach, object-contextual representations, characterizing a pixel by exploiting the representation of the corresponding object class. First, we learn object regions under the supervision of ground-truth segmentation. Second, we compute the object region representation by aggregating the representations of the pixels lying in the object region. Last, % the representation similarity we compute the relation between each pixel and each object region and augment the representation of each pixel with the object-contextual representation which is a weighted aggregation of all the object region representations according to their relations with the pixel. We empirically demonstrate that the proposed approach achieves competitive performance on various challenging semantic segmentation benchmarks: Cityscapes, ADE20K, LIP, PASCAL-Context, and COCO-Stuff. Cityscapes, ADE20K, LIP, PASCAL-Context, and COCO-Stuff. Our submission \"HRNet + OCR + SegFix\" achieves 1-st place on the Cityscapes leaderboard by the time of submission. Code is available at: https://git.io/openseg and https://git.io/HRNet.OCR. We rephrase the object-contextual representation scheme using the Transformer encoder-decoder framework. The details are presented in~Section3.3."
                    },
                    {
                        "title": "Cross View Fusion for 3D Human Pose Estimation",
                        "abstract": "We present an approach to recover absolute 3D human poses from multi-view images by incorporating multi-view geometric priors in our model. It consists of two separate steps: (1) estimating the 2D poses in multi-view images and (2) recovering the 3D poses from the multi-view 2D poses. First, we introduce a cross-view fusion scheme into CNN to jointly estimate 2D poses for multiple views. Consequently, the 2D pose estimation for each view already benefits from other views. Second, we present a recursive Pictorial Structure Model to recover the 3D pose from the multi-view 2D poses. It gradually improves the accuracy of 3D pose with affordable computational cost. We test our method on two public datasets H36M and Total Capture. The Mean Per Joint Position Errors on the two datasets are 26mm and 29mm, which outperforms the state-of-the-arts remarkably (26mm vs 52mm, 29mm vs 35mm). Our code is released at \\url{https://github.com/microsoft/multiview-human-pose-estimation-pytorch}."
                    },
                    {
                        "title": "Deep Regression for Face Alignment",
                        "abstract": "In this paper, we present a deep regression approach for face alignment. The deep architecture consists of a global layer and multi-stage local layers. We apply the back-propagation algorithm with the dropout strategy to jointly optimize the regression parameters. We show that the resulting deep regressor gradually and evenly approaches the true facial landmarks stage by stage, avoiding the tendency to yield over-strong early stage regressors while over-weak later stage regressors. Experimental results show that our approach achieves the state-of-the-art"
                    }
                ]
            }
        }
    },
    "2410.05601": {
        "paper_data": {
            "title": "ReFIR: Grounding Large Restoration Models with Retrieval Augmentation",
            "url": "http://arxiv.org/abs/2410.05601v1",
            "arxiv_id": "2410.05601",
            "authors": [
                "Hang Guo",
                "Tao Dai",
                "Zhihao Ouyang",
                "Taolin Zhang",
                "Yaohua Zha",
                "Bin Chen",
                "Shu-tao Xia"
            ],
            "abstract": "Recent advances in diffusion-based Large Restoration Models (LRMs) have significantly improved photo-realistic image restoration by leveraging the internal knowledge embedded within model weights. However, existing LRMs often suffer from the hallucination dilemma, i.e., producing incorrect contents or textures when dealing with severe degradations, due to their heavy reliance on limited internal knowledge. In this paper, we propose an orthogonal solution called the Retrieval-augmented Framework for Image Restoration (ReFIR), which incorporates retrieved images as external knowledge to extend the knowledge boundary of existing LRMs in generating details faithful to the original scene. Specifically, we first introduce the nearest neighbor lookup to retrieve content-relevant high-quality images as reference, after which we propose the cross-image injection to modify existing LRMs to utilize high-quality textures from retrieved images. Thanks to the additional external knowledge, our ReFIR can well handle the hallucination challenge and facilitate faithfully results. Extensive experiments demonstrate that ReFIR can achieve not only high-fidelity but also realistic restoration results. Importantly, our ReFIR requires no training and is adaptable to various LRMs.",
            "introduction": "   1 Introduction  Restoring a high-quality image (HQ) from its low-quality counterpart (LQ) is a well-known ill-posed problem and has been studied over the years\u00a0[1, 2, 3, 4, 5, 6, 7, 8, 9]. Previous efforts attempt to handle this problem through employing various neural network architectures, including CNNs, GANs and Transformers. Recently, diffusion models\u00a0[10, 11] have emerged as a promising alternative, delivering noteworthy results in real-world image restoration\u00a0[12, 13, 14]. In particular, some works\u00a0[15, 16, 17, 18, 19, 20] have successfully leveraged the powerful generative prior of pre-trained text-to-image (T2I) diffusion models for scaling up, to obtain the Large Restoration Model (LRM) with billions of parameters, bringing significant progress in restoring photo-realistic images.   Although scaling up restoration models has achieved remarkable success, existing LRMs may not always produce results that are faithful to the original scene, particularly when faced with heavily degraded images that surpass the LRMs\u2019 capabilities (see Fig.\u00a01). This issue is similar to the hallucination problem observed in large language models (LLMs)\u00a0[21, 22], e.g. ChatGPT might generate nonsense responses when highly specialized questions exceed its knowledge boundary. Similarly, if one LRM has never seen a specific scene, it will struggle to restore corresponding images faithfully. By analogizing LLM to LRM, we define the phenomenon where LRMs generate textures inconsistent with the original scene when facing hard samples as the hallucination of LRMs.   To address the hallucination problem in LRMs, simply expanding the internal knowledge through additional training data and parameters might seem straightforward, but it can significantly increase computational and storage costs. Instead, this work considers another orthogonal strategy that enhances the external knowledge of LRMs without adding parameter counts. Drawing inspiration from the retrieval-augmented generation (RAG) used in LLMs\u00a0[23, 24, 25], we aim to use the retrieved high-quality content-relevant images as external knowledge to alleviate the hallucination of LRMs. However, applying RAG to image restoration poses specific challenges. Specifically, in natural language, simply feeding the retrieved documents along with the original user query to LLMs can allow it to produce grounded responses. However, in the context of image restoration, allowing low-quality images to attend to retrieved images during their restoration process is non-trivial, which motivates us to develop novel techniques to enable LRMs to utilize external knowledge in restoration.   Figure 1: Existing LRMs encounter hallucination issues, i.e., generating contents or details that deviate from the original scene, when dealing with challenging degradations. By incorporating the proposed ReFIR to existing LRMs\u00a0[20] without any training, the additional external knowledge facilitates producing more faithful results. Please zoom in for better visualization.   To this end, we delve deep into the working mechanisms of LRMs for insightful observations. Details of the experimental setup are described in Sec.\u00a03. Our key findings indicate that the workflow of LRMs can be divided into two distinct stages: the Denoising Structure Reconstruction stage, during which the self-attention in the ControlNet\u00a0[26] reconstructs a clear overall structure from the noised representation. After that, in the Detail Texture Restoration stage, the self-attention in the UNet\u00a0[27] decoder fills scene-specific textures based on the denoised structure map. Based on these findings, a natural solution emerged: we can transfer high-quality,",
            "references": [
                {
                    "title": "MambaIR: A Simple Baseline for Image Restoration with State-Space Model",
                    "abstract": "Recent years have seen significant advancements in image restoration, largely attributed to the development of modern deep neural networks, such as CNNs and Transformers. However, existing restoration backbones often face the dilemma between global receptive fields and efficient computation, hindering their application in practice. Recently, the Selective Structured State Space Model, especially the improved version Mamba, has shown great potential for long-range dependency modeling with linear complexity, which offers a way to resolve the above dilemma. However, the standard Mamba still faces certain challenges in low-level vision such as local pixel forgetting and channel redundancy. In this work, we introduce a simple but effective baseline, named MambaIR, which introduces both local enhancement and channel attention to improve the vanilla Mamba. In this way, our MambaIR takes advantage of the local pixel similarity and reduces the channel redundancy. Extensive experiments demonstrate the superiority of our method, for example, MambaIR outperforms SwinIR by up to 0.45dB on image SR, using similar computational cost but with a global receptive field. Code is available at \\url{https://github.com/csguoh/MambaIR}."
                },
                {
                    "title": "Scaling Up to Excellence: Practicing Model Scaling for Photo-Realistic Image Restoration In the Wild",
                    "abstract": "We introduce SUPIR (Scaling-UP Image Restoration), a groundbreaking image restoration method that harnesses generative prior and the power of model scaling up. Lever-aging multi-modal techniques and advanced generative prior, SUPIR marks a significant advance in intelligent and realistic image restoration. As a pivotal catalyst within SUPIR, model scaling dramatically enhances its capabil-ities and demonstrates new potential for image restoration. We collect a dataset comprising 20 million high-resolution, high-quality images for model training, each en-riched with descriptive text annotations. SUPIR provides the capability to restore images guided by textual prompts, broadening its application scope and potential. Moreover, we introduce negative-quality prompts to further improve perceptual quality. We also develop a restoration-guided sampling method to suppress the fidelity issue encountered in generative-based restoration. Experiments demonstrate SUPIR's exceptional restoration effects and its novel capac-ity to manipulate restoration through textual prompts."
                },
                {
                    "title": "SeeSR: Towards Semantics-Aware Real-World Image Super-Resolution",
                    "abstract": "Owe to the powerful generative priors, the pretrained text-to-image (T2I) diffusion models have become increasingly popular in solving the real-world image super-resolution problem. However, as a consequence of the heavy quality degradation of input low-resolution (LR) images, the destruction of local structures can lead to ambiguous image semantics. As a result, the content of reproduced high-resolution image may have semantic errors, deteriorating the super-resolution performance. To address this issue, we present a semantics-aware approach to better preserve the semantic fidelity of generative real-world image super-resolution. First, we train a degradation-aware prompt extractor, which can generate accurate soft and hard semantic prompts even under strong degradation. The hard semantic prompts refer to the image tags, aiming to enhance the local perception ability of the T2I model, while the soft semantic prompts compensate for the hard ones to provide additional representation information. These semantic prompts encourage the T2I model to generate detailed and semantically accurate results. Further-more, during the inference process, we integrate the LR images into the initial sampling noise to mitigate the diffusion model's tendency to generate excessive random details. The experiments show that our method can reproduce more realistic image details and hold better the semantics. The source code of our method can be found at https://github.com/cswry/SeeSR."
                },
                {
                    "title": "Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection",
                    "abstract": "Despite their remarkable capabilities, large language models (LLMs) often produce responses containing factual inaccuracies due to their sole reliance on the parametric knowledge they encapsulate. Retrieval-Augmented Generation (RAG), an ad hoc approach that augments LMs with retrieval of relevant knowledge, decreases such issues. However, indiscriminately retrieving and incorporating a fixed number of retrieved passages, regardless of whether retrieval is necessary, or passages are relevant, diminishes LM versatility or can lead to unhelpful response generation. We introduce a new framework called Self-Reflective Retrieval-Augmented Generation (Self-RAG) that enhances an LM's quality and factuality through retrieval and self-reflection. Our framework trains a single arbitrary LM that adaptively retrieves passages on-demand, and generates and reflects on retrieved passages and its own generations using special tokens, called reflection tokens. Generating reflection tokens makes the LM controllable during the inference phase, enabling it to tailor its behavior to diverse task requirements. Experiments show that Self-RAG (7B and 13B parameters) significantly outperforms state-of-the-art LLMs and retrieval-augmented models on a diverse set of tasks. Specifically, Self-RAG outperforms ChatGPT and retrieval-augmented Llama2-chat on Open-domain QA, reasoning and fact verification tasks, and it shows significant gains in improving factuality and citation accuracy for long-form generations relative to these models."
                },
                {
                    "title": "DiffBIR: Towards Blind Image Restoration with Generative Diffusion Prior",
                    "abstract": "We present DiffBIR, a general restoration pipeline that could handle different blind image restoration tasks in a unified framework. DiffBIR decouples blind image restoration problem into two stages: 1) degradation removal: removing image-independent content; 2) information regeneration: generating the lost image content. Each stage is developed independently but they work seamlessly in a cascaded manner. In the first stage, we use restoration modules to remove degradations and obtain high-fidelity restored results. For the second stage, we propose IRControlNet that leverages the generative ability of latent diffusion models to generate realistic details. Specifically, IRControlNet is trained based on specially produced condition images without distracting noisy content for stable generation performance. Moreover, we design a region-adaptive restoration guidance that can modify the denoising process during inference without model re-training, allowing users to balance realness and fidelity through a tunable guidance scale. Extensive experiments have demonstrated DiffBIR's superiority over state-of-the-art approaches for blind image super-resolution, blind face restoration and blind image denoising tasks on both synthetic and real-world datasets. The code is available at https://github.com/XPixelGroup/DiffBIR."
                },
                {
                    "title": "Pixel-Aware Stable Diffusion for Realistic Image Super-resolution and Personalized Stylization",
                    "abstract": "Diffusion models have demonstrated impressive performance in various image generation, editing, enhancement and translation tasks. In particular, the pre-trained text-to-image stable diffusion models provide a potential solution to the challenging realistic image super-resolution (Real-ISR) and image stylization problems with their strong generative priors. However, the existing methods along this line often fail to keep faithful pixel-wise image structures. If extra skip connections between the encoder and the decoder of a VAE are used to reproduce details, additional training in image space will be required, limiting the application to tasks in latent space such as image stylization. In this work, we propose a pixel-aware stable diffusion (PASD) network to achieve robust Real-ISR and personalized image stylization. Specifically, a pixel-aware cross attention module is introduced to enable diffusion models perceiving image local structures in pixel-wise level, while a degradation removal module is used to extract degradation insensitive features to guide the diffusion process together with image high level information. An adjustable noise schedule is introduced to further improve the image restoration results. By simply replacing the base diffusion model with a stylized one, PASD can generate diverse stylized images without collecting pairwise training data, and by shifting the base model with an aesthetic one, PASD can bring old photos back to life. Extensive experiments in a variety of image enhancement and stylization tasks demonstrate the effectiveness of our proposed PASD approach. Our source codes are available at \\url{https://github.com/yangxy/PASD/}."
                },
                {
                    "title": "ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shifting",
                    "abstract": "Diffusion-based image super-resolution (SR) methods are mainly limited by the low inference speed due to the requirements of hundreds or even thousands of sampling steps. Existing acceleration sampling techniques inevitably sacrifice performance to some extent, leading to over-blurry SR results. To address this issue, we propose a novel and efficient diffusion model for SR that significantly reduces the number of diffusion steps, thereby eliminating the need for post-acceleration during inference and its associated performance deterioration. Our method constructs a Markov chain that transfers between the high-resolution image and the low-resolution image by shifting the residual between them, substantially improving the transition efficiency. Additionally, an elaborate noise schedule is developed to flexibly control the shifting speed and the noise strength during the diffusion process. Extensive experiments demonstrate that the proposed method obtains superior or at least comparable performance to current state-of-the-art methods on both synthetic and real-world datasets, even only with 15 sampling steps. Our code and model are available at https://github.com/zsyOAOA/ResShift."
                },
                {
                    "title": "A Unified Conditional Framework for Diffusion-based Image Restoration",
                    "abstract": "Diffusion Probabilistic Models (DPMs) have recently shown remarkable performance in image generation tasks, which are capable of generating highly realistic images. When adopting DPMs for image restoration tasks, the crucial aspect lies in how to integrate the conditional information to guide the DPMs to generate accurate and natural output, which has been largely overlooked in existing works. In this paper, we present a unified conditional framework based on diffusion models for image restoration. We leverage a lightweight UNet to predict initial guidance and the diffusion model to learn the residual of the guidance. By carefully designing the basic module and integration module for the diffusion model block, we integrate the guidance and other auxiliary conditional information into every block of the diffusion model to achieve spatially-adaptive generation conditioning. To handle high-resolution images, we propose a simple yet effective inter-step patch-splitting strategy to produce arbitrary-resolution images without grid artifacts. We evaluate our conditional framework on three challenging tasks: extreme low-light denoising, deblurring, and JPEG restoration, demonstrating its significant improvements in perceptual quality and the generalization to restoration tasks."
                },
                {
                    "title": "Real-World Image Variation by Aligning Diffusion Inversion Chain",
                    "abstract": "Recent diffusion model advancements have enabled high-fidelity images to be generated using text prompts. However, a domain gap exists between generated images and real-world images, which poses a challenge in generating high-quality variations of real-world images. Our investigation uncovers that this domain gap originates from a latents' distribution gap in different diffusion processes. To address this issue, we propose a novel inference pipeline called Real-world Image Variation by ALignment (RIVAL) that utilizes diffusion models to generate image variations from a single image exemplar. Our pipeline enhances the generation quality of image variations by aligning the image generation process to the source image's inversion chain. Specifically, we demonstrate that step-wise latent distribution alignment is essential for generating high-quality variations. To attain this, we design a cross-image self-attention injection for feature interaction and a step-wise distribution normalization to align the latent features. Incorporating these alignment processes into a diffusion model allows RIVAL to generate high-quality image variations without further parameter optimization. Our experimental results demonstrate that our proposed approach outperforms existing methods with respect to semantic-condition similarity and perceptual quality. Furthermore, this generalized inference pipeline can be easily applied to other diffusion-based generation tasks, such as image-conditioned text-to-image generation and example-based image inpainting."
                },
                {
                    "title": "Photoswap: Personalized Subject Swapping in Images",
                    "abstract": "In an era where images and visual content dominate our digital landscape, the ability to manipulate and personalize these images has become a necessity. Envision seamlessly substituting a tabby cat lounging on a sunlit window sill in a photograph with your own playful puppy, all while preserving the original charm and composition of the image. We present Photoswap, a novel approach that enables this immersive image editing experience through personalized subject swapping in existing images. Photoswap first learns the visual concept of the subject from reference images and then swaps it into the target image using pre-trained diffusion models in a training-free manner. We establish that a well-conceptualized visual subject can be seamlessly transferred to any image with appropriate self-attention and cross-attention manipulation, maintaining the pose of the swapped subject and the overall coherence of the image. Comprehensive experiments underscore the efficacy and controllability of Photoswap in personalized subject swapping. Furthermore, Photoswap significantly outperforms baseline methods in human ratings across subject swapping, background preservation, and overall quality, revealing its vast application potential, from entertainment to professional editing."
                },
                {
                    "title": "FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation",
                    "abstract": "Evaluating the factuality of long-form text generated by large language models (LMs) is non-trivial because (1) generations often contain a mixture of supported and unsupported pieces of information, making binary judgments of quality inadequate, and (2) human evaluation is time-consuming and costly. In this paper, we introduce FACTSCORE, a new evaluation that breaks a generation into a series of atomic facts and computes the percentage of atomic facts supported by a reliable knowledge source. We conduct an extensive human evaluation to obtain FACTSCOREs of people biographies generated by several state-of-the-art commercial LMs -- InstructGPT, ChatGPT, and the retrieval-augmented PerplexityAI -- and report new analysis demonstrating the need for such a fine-grained score (e.g., ChatGPT only achieves 58%). Since human evaluation is costly, we also introduce an automated model that estimates FACTSCORE using retrieval and a strong language model, with less than a 2% error rate. Finally, we use this automated metric to evaluate 6,500 generations from a new set of 13 recent LMs that would have cost $26K if evaluated by humans, with various findings: GPT-4 and ChatGPT are more factual than public models, and Vicuna and Alpaca are some of the best public models. FACTSCORE is available for public use via `pip install factscore`."
                },
                {
                    "title": "Exploiting Diffusion Prior for Real-World Image Super-Resolution",
                    "abstract": "We present a novel approach to leverage prior knowledge encapsulated in pre-trained text-to-image diffusion models for blind super-resolution (SR). Specifically, by employing our time-aware encoder, we can achieve promising restoration results without altering the pre-trained synthesis model, thereby preserving the generative prior and minimizing training cost. To remedy the loss of fidelity caused by the inherent stochasticity of diffusion models, we introduce a controllable feature wrapping module that allows users to balance quality and fidelity by simply adjusting a scalar value during the inference process. Moreover, we develop a progressive aggregation sampling strategy to overcome the fixed-size constraints of pre-trained diffusion models, enabling adaptation to resolutions of any size. A comprehensive evaluation of our method using both synthetic and real-world benchmarks demonstrates its superiority over current state-of-the-art approaches."
                },
                {
                    "title": "MasaCtrl: Tuning-Free Mutual Self-Attention Control for Consistent Image Synthesis and Editing",
                    "abstract": "Despite the success in large-scale text-to-image generation and text-conditioned image editing, existing methods still struggle to produce consistent generation and editing results. For example, generation approaches usually fail to synthesize multiple images of the same objects/characters but with different views or poses. Meanwhile, existing editing methods either fail to achieve effective complex nonrigid editing while maintaining the overall textures and identity, or require time-consuming fine-tuning to capture the image-specific appearance. In this paper, we develop MasaCtrl, a tuning-free method to achieve consistent image generation and complex non-rigid image editing simultaneously. Specifically, MasaCtrl converts existing self-attention in diffusion models into mutual self-attention, so that it can query correlated local contents and textures from source images for consistency. To further alleviate the query confusion between foreground and background, we propose a mask-guided mutual self-attention strategy, where the mask can be easily extracted from the cross-attention maps. Extensive experiments show that the proposed MasaCtrl can produce impressive results in both consistent image generation and complex non-rigid real image editing."
                },
                {
                    "title": "Generative Diffusion Prior for Unified Image Restoration and Enhancement",
                    "abstract": "Existing image restoration methods mostly leverage the posterior distribution of natural images. However, they often assume known degradation and also require supervised training, which restricts their adaptation to complex real applications. In this work, we propose the Generative Diffusion Prior (GDP) to effectively model the posterior distributions in an unsupervised sampling manner. GDP utilizes a pre-train denoising diffusion generative model (DDPM) for solving linear inverse, non-linear, or blind problems. Specifically, GDP systematically explores a protocol of conditional guidance, which is verified more practical than the commonly used guidance way. Furthermore, GDP is strength at optimizing the parameters of degradation model during the denoising process, achieving blind image restoration. Besides, we devise hierarchical guidance and patch-based methods, enabling the GDP to generate images of arbitrary resolutions. Experimentally, we demonstrate GDP's versatility on several image datasets for linear problems, such as super-resolution, deblurring, inpainting, and colorization, as well as non-linear and blind issues, such as low-light enhancement and HDR image recovery. GDP outperforms the current leading unsupervised methods on the diverse benchmarks in reconstruction quality and perceptual quality. Moreover, GDP also generalizes well for natural images or synthesized images with arbitrary sizes from various tasks out of the distribution of the ImageNet training set. The project page is available at https://generativediffusionprior.github.io/"
                },
                {
                    "title": "ResDiff: Combining CNN and Diffusion Model for Image Super-Resolution",
                    "abstract": "Adapting the Diffusion Probabilistic Model (DPM) for direct image super-resolution is wasteful, given that a simple Convolutional Neural Network (CNN) can recover the main low-frequency content. Therefore, we present ResDiff, a novel Diffusion Probabilistic Model based on Residual structure for Single Image Super-Resolution (SISR). ResDiff utilizes a combination of a CNN, which restores primary low-frequency components, and a DPM, which predicts the residual between the ground-truth image and the CNN predicted image. In contrast to the common diffusion-based methods that directly use LR space to guide the noise towards HR space, ResDiff utilizes the CNN\u2019s initial prediction to direct the noise towards the residual space between HR space and CNN-predicted space, which not only accelerates the generation process but also acquires superior sample quality. Additionally, a frequency-domain-based loss function for CNN is introduced to facilitate its restoration, and a frequency-domain guided diffusion is designed for DPM on behalf of predicting high-frequency details. The extensive experiments on multiple benchmark datasets demonstrate that ResDiff outperforms previous diffusion based methods in terms of shorter model convergence time, superior generation quality, and more diverse samples."
                },
                {
                    "title": "LMR: A Large-Scale Multi-Reference Dataset for Reference-based Super-Resolution",
                    "abstract": "It is widely agreed that reference-based super-resolution (RefSR) achieves superior results by referring to similar high quality images, compared to single image super-resolution (SISR). Intuitively, the more references, the better performance. However, previous RefSR methods have all focused on single-reference image training, while multiple reference images are often available in testing or practical applications. The root cause of such training-testing mismatch is the absence of publicly available multi-reference SR training datasets, which greatly hinders research efforts on multi-reference super-resolution. To this end, we construct a large-scale, multi-reference super-resolution dataset, named LMR. It contains 112, 142 groups of 300\u00d7300 training images, which is 10\u00d7 of the existing largest RefSR dataset. The image size is also some times larger. More importantly, each group is equipped with 5 reference images with different similarity levels. Furthermore, we propose a new baseline method for multi-reference super-resolution: MRefSR, including a Multi-Reference Attention Module (MAM) for feature fusion of an arbitrary number of reference images, and a Spatial Aware Filtering Module (SAFM) for the fused feature selection. The proposed MRefSR achieves significant improvements over state-of-the-art approaches on both quantitative and qualitative evaluations. Our code and data are available at: https://github.com/wdmwhh/MRefSR."
                },
                {
                    "title": "Adding Conditional Control to Text-to-Image Diffusion Models",
                    "abstract": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models."
                },
                {
                    "title": "In-Context Retrieval-Augmented Language Models",
                    "abstract": "Abstract Retrieval-Augmented Language Modeling (RALM) methods, which condition a language model (LM) on relevant documents from a grounding corpus during generation, were shown to significantly improve language modeling performance. In addition, they can mitigate the problem of factually inaccurate text generation and provide natural source attribution mechanism. Existing RALM approaches focus on modifying the LM architecture in order to facilitate the incorporation of external information, significantly complicating deployment. This paper considers a simple alternative, which we dub In-Context RALM: leaving the LM architecture unchanged and prepending grounding documents to the input, without any further training of the LM. We show that In-Context RALM that builds on off-the-shelf general purpose retrievers provides surprisingly large LM gains across model sizes and diverse corpora. We also demonstrate that the document retrieval and ranking mechanism can be specialized to the RALM setting to further boost performance. We conclude that In-Context RALM has considerable potential to increase the prevalence of LM grounding, particularly in settings where a pretrained LM must be used without modification or even via API access.1"
                },
                {
                    "title": "When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memories",
                    "abstract": "Despite their impressive performance on diverse tasks, large language models (LMs) still struggle with tasks requiring rich world knowledge, implying the difficulty of encoding a wealth of world knowledge in their parameters. This paper aims to understand LMs\u2019 strengths and limitations in memorizing factual knowledge, by conducting large-scale knowledge probing experiments on two open-domain entity-centric QA datasets: PopQA, our new dataset with 14k questions about long-tail entities, and EntityQuestions, a widely used open-domain QA dataset. We find that LMs struggle with less popular factual knowledge, and that retrieval augmentation helps significantly in these cases. Scaling, on the other hand, mainly improves memorization of popular knowledge, and fails to appreciably improve memorization of factual knowledge in the tail. Based on those findings, we devise a new method for retrieval-augmentation that improves performance and reduces inference costs by only retrieving non-parametric memories when necessary."
                },
                {
                    "title": "Scalable Diffusion Models with Transformers",
                    "abstract": "We explore a new class of diffusion models based on the transformer architecture. We train latent diffusion models of images, replacing the commonly-used U-Net backbone with a transformer that operates on latent patches. We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops. We find that DiTs with higher Gflops\u2014through increased transformer depth/width or increased number of input tokens\u2014consistently have lower FID. In addition to possessing good scalability properties, our largest DiT-XL/2 models outperform all prior diffusion models on the class-conditional ImageNet 512\u00d7512 and 256\u00d7256 benchmarks, achieving a state-of-the-art FID of 2.27 on the latter."
                },
                {
                    "title": "Zero-Shot Image Restoration Using Denoising Diffusion Null-Space Model",
                    "abstract": "Most existing Image Restoration (IR) models are task-specific, which can not be generalized to different degradation operators. In this work, we propose the Denoising Diffusion Null-Space Model (DDNM), a novel zero-shot framework for arbitrary linear IR problems, including but not limited to image super-resolution, colorization, inpainting, compressed sensing, and deblurring. DDNM only needs a pre-trained off-the-shelf diffusion model as the generative prior, without any extra training or network modifications. By refining only the null-space contents during the reverse diffusion process, we can yield diverse results satisfying both data consistency and realness. We further propose an enhanced and robust version, dubbed DDNM+, to support noisy restoration and improve restoration quality for hard tasks. Our experiments on several IR tasks reveal that DDNM outperforms other state-of-the-art zero-shot IR methods. We also demonstrate that DDNM+ can solve complex real-world applications, e.g., old photo restoration."
                },
                {
                    "title": "VoP: Text-Video Co-Operative Prompt Tuning for Cross-Modal Retrieval",
                    "abstract": "Many recent studies leverage the pre-trained CLIP for text-video cross-modal retrieval by tuning the backbone with additional heavy modules, which not only brings huge computational burdens with much more parameters, but also leads to the knowledge forgetting from upstream models. In this work, we propose the VoP: Text-Video Cooperative Prompt Tuning for efficient tuning on the text-video retrieval task. The proposed VoP is an end-to-end framework with both video & text prompts introducing, which can be regarded as a powerful baseline with only 0.1% trainable parameters. Further, based on the spatiotemporal characteristics of videos, we develop three novel video prompt mechanisms to improve the performance with different scales of trainable parameters. The basic idea of the VoP enhancement is to model the frame position, frame context, and layer function with specific trainable prompts, respectively. Extensive experiments show that compared to full fine-tuning, the enhanced VoP achieves a 1.4% average R@1 gain across five text-video retrieval benchmarks with 6\u00d7 less parameter overhead. The code will be available at https://github.com/bighuang624/VoP."
                },
                {
                    "title": "Prompt-to-Prompt Image Editing with Cross Attention Control",
                    "abstract": "Recent large-scale text-driven synthesis models have attracted much attention thanks to their remarkable capabilities of generating highly diverse images that follow given text prompts. Such text-based synthesis methods are particularly appealing to humans who are used to verbally describe their intent. Therefore, it is only natural to extend the text-driven image synthesis to text-driven image editing. Editing is challenging for these generative models, since an innate property of an editing technique is to preserve most of the original image, while in the text-based models, even a small modification of the text prompt often leads to a completely different outcome. State-of-the-art methods mitigate this by requiring the users to provide a spatial mask to localize the edit, hence, ignoring the original structure and content within the masked region. In this paper, we pursue an intuitive prompt-to-prompt editing framework, where the edits are controlled by text only. To this end, we analyze a text-conditioned model in depth and observe that the cross-attention layers are the key to controlling the relation between the spatial layout of the image to each word in the prompt. With this observation, we present several applications which monitor the image synthesis by editing the textual prompt only. This includes localized editing by replacing a word, global editing by adding a specification, and even delicately controlling the extent to which a word is reflected in the image. We present our results over diverse images and prompts, demonstrating high-quality synthesis and fidelity to the edited prompts."
                },
                {
                    "title": "Exploring CLIP for Assessing the Look and Feel of Images",
                    "abstract": "Measuring the perception of visual content is a long-standing problem in computer vision. Many mathematical models have been developed to evaluate the look or quality of an image. Despite the effectiveness of such tools in quantifying degradations such as noise and blurriness levels, such quantification is loosely coupled with human language. When it comes to more abstract perception about the feel of visual content, existing methods can only rely on supervised models that are explicitly trained with labeled data collected via laborious user study. In this paper, we go beyond the conventional paradigms by exploring the rich visual language prior encapsulated in Contrastive Language-Image Pre-training (CLIP) models for assessing both the quality perception (look) and abstract perception (feel) of images without explicit task-specific training. In particular, we discuss effective prompt designs and show an effective prompt pairing strategy to harness the prior. We also provide extensive experiments on controlled datasets and Image Quality Assessment (IQA) benchmarks. Our results show that CLIP captures meaningful priors that generalize well to different perceptual assessments."
                },
                {
                    "title": "Reference-based Image Super-Resolution with Deformable Attention Transformer",
                    "abstract": "Reference-based image super-resolution (RefSR) aims to exploit auxiliary reference (Ref) images to super-resolve low-resolution (LR) images. Recently, RefSR has been attracting great attention as it provides an alternative way to surpass single image SR. However, addressing the RefSR problem has two critical challenges: (i) It is difficult to match the correspondence between LR and Ref images when they are significantly different; (ii) How to transfer the relevant texture from Ref images to compensate the details for LR images is very challenging. To address these issues of RefSR, this paper proposes a deformable attention Transformer, namely DATSR, with multiple scales, each of which consists of a texture feature encoder (TFE) module, a reference-based deformable attention (RDA) module and a residual feature aggregation (RFA) module. Specifically, TFE first extracts image transformation (e.g., brightness) insensitive features for LR and Ref images, RDA then can exploit multiple relevant textures to compensate more information for LR features, and RFA lastly aggregates LR features and relevant textures to get a more visually pleasant result. Extensive experiments demonstrate that our DATSR achieves state-of-the-art performance on benchmark datasets quantitatively and qualitatively."
                },
                {
                    "title": "How Do Vision Transformers Work?",
                    "abstract": "The success of multi-head self-attentions (MSAs) for computer vision is now indisputable. However, little is known about how MSAs work. We present fundamental explanations to help better understand the nature of MSAs. In particular, we demonstrate the following properties of MSAs and Vision Transformers (ViTs): (1) MSAs improve not only accuracy but also generalization by flattening the loss landscapes. Such improvement is primarily attributable to their data specificity, not long-range dependency. On the other hand, ViTs suffer from non-convex losses. Large datasets and loss landscape smoothing methods alleviate this problem; (2) MSAs and Convs exhibit opposite behaviors. For example, MSAs are low-pass filters, but Convs are high-pass filters. Therefore, MSAs and Convs are complementary; (3) Multi-stage neural networks behave like a series connection of small individual models. In addition, MSAs at the end of a stage play a key role in prediction. Based on these insights, we propose AlterNet, a model in which Conv blocks at the end of a stage are replaced with MSA blocks. AlterNet outperforms CNNs not only in large data regimes but also in small data regimes. The code is available at https://github.com/xxxnell/how-do-vits-work."
                },
                {
                    "title": "Denoising Diffusion Restoration Models",
                    "abstract": "Many interesting tasks in image restoration can be cast as linear inverse problems. A recent family of approaches for solving these problems uses stochastic algorithms that sample from the posterior distribution of natural images given the measurements. However, efficient solutions often require problem-specific supervised training to model the posterior, whereas unsupervised methods that are not problem-specific typically rely on inefficient iterative methods. This work addresses these issues by introducing Denoising Diffusion Restoration Models (DDRM), an efficient, unsupervised posterior sampling method. Motivated by variational inference, DDRM takes advantage of a pre-trained denoising diffusion generative model for solving any linear inverse problem. We demonstrate DDRM's versatility on several image datasets for super-resolution, deblurring, inpainting, and colorization under various amounts of measurement noise. DDRM outperforms the current leading unsupervised methods on the diverse ImageNet dataset in reconstruction quality, perceptual quality, and runtime, being 5x faster than the nearest competitor. DDRM also generalizes well for natural images out of the distribution of the observed ImageNet training set."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "On Efficient Transformer-Based Image Pre-training for Low-Level Vision",
                    "abstract": "Pre-training has marked numerous state of the arts in high-level computer vision, while few attempts have ever been made to investigate how pre-training acts in image processing systems. In this paper, we tailor transformer-based pre-training regimes that boost various low-level tasks. To comprehensively diagnose the influence of pre-training, we design a whole set of principled evaluation tools that uncover its effects on internal representations. The observations demonstrate that pre-training plays strikingly different roles in low-level tasks. For example, pre-training introduces more local information to intermediate layers in super-resolution (SR), yielding significant performance gains, while pre-training hardly affects internal feature representations in denoising, resulting in limited gains. Further, we explore different methods of pre-training, revealing that multi-related-task pre-training is more effective and data-efficient than other alternatives. Finally, we extend our study to varying data scales and model sizes, as well as comparisons between transformers and CNNs. Based on the study, we successfully develop state-of-the-art models for multiple low-level tasks."
                },
                {
                    "title": "SwinIR: Image Restoration Using Swin Transformer",
                    "abstract": "Image restoration is a long-standing low-level vision problem that aims to restore high-quality images from low-quality images (e.g., downscaled, noisy and compressed images). While state-of-the-art image restoration methods are based on convolutional neural networks, few attempts have been made with Transformers which show impressive performance on high-level vision tasks. In this paper, we propose a strong baseline model SwinIR for image restoration based on the Swin Transformer. SwinIR consists of three parts: shallow feature extraction, deep feature extraction and high-quality image reconstruction. In particular, the deep feature extraction module is composed of several residual Swin Transformer blocks (RSTB), each of which has several Swin Transformer layers together with a residual connection. We conduct experiments on three representative tasks: image super-resolution (including classical, lightweight and real-world image super-resolution), image denoising (including grayscale and color image denoising) and JPEG compression artifact reduction. Experimental results demonstrate that SwinIR outperforms state-of-the-art methods on different tasks by up to 0.14\u223c0.45dB, while the total number of parameters can be reduced by up to 67%."
                },
                {
                    "title": "MUSIQ: Multi-scale Image Quality Transformer",
                    "abstract": "Image quality assessment (IQA) is an important research topic for understanding and improving visual experience. The current state-of-the-art IQA methods are based on convolutional neural networks (CNNs). The performance of CNN-based models is often compromised by the fixed shape constraint in batch training. To accommodate this, the input images are usually resized and cropped to a fixed shape, causing image quality degradation. To address this, we design a multi-scale image quality Transformer (MUSIQ) to process native resolution images with varying sizes and aspect ratios. With a multi-scale image representation, our proposed method can capture image quality at different granularities. Furthermore, a novel hash-based 2D spatial embedding and a scale embedding is proposed to support the positional embedding in the multi-scale representation. Experimental results verify that our method can achieve state-of-the-art performance on multiple large scale IQA datasets such as PaQ-2-PiQ [41], SPAQ [11], and KonIQ-10k [16]. 1"
                },
                {
                    "title": "Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data",
                    "abstract": "Though many attempts have been made in blind super-resolution to restore low-resolution images with unknown and complex degradations, they are still far from addressing general real-world degraded images. In this work, we extend the powerful ESRGAN to a practical restoration application (namely, Real-ESRGAN), which is trained with pure synthetic data. Specifically, a high-order degradation modeling process is introduced to better simulate complex real-world degradations. We also consider the common ringing and overshoot artifacts in the synthesis process. In addition, we employ a U-Net discriminator with spectral normalization to increase discriminator capability and stabilize the training dynamics. Extensive comparisons have shown its superior visual performance than prior works on various real datasets. We also provide efficient implementations to synthesize training pairs on the fly."
                },
                {
                    "title": "Robust Reference-based Super-Resolution via C2-Matching",
                    "abstract": "Reference-based Super-Resolution (Ref-SR) has recently emerged as a promising paradigm to enhance a low-resolution (LR) input image by introducing an additional high-resolution (HR) reference image. Existing Ref-SR methods mostly rely on implicit correspondence matching to borrow HR textures from reference images to compensate for the information loss in input images. However, performing local transfer is difficult because of two gaps between input and reference images: the transformation gap (e.g. scale and rotation) and the resolution gap (e.g. HR and LR). To tackle these challenges, we propose C2-Matching in this work, which produces explicit robust matching crossing transformation and resolution. 1) For the transformation gap, we propose a contrastive correspondence network, which learns transformation-robust correspondences using augmented views of the input image. 2) For the resolution gap, we adopt a teacher-student correlation distillation, which distills knowledge from the easier HR-HR matching to guide the more ambiguous LR-HR matching. 3) Finally, we design a dynamic aggregation module to address the potential misalignment issue. In addition, to faithfully evaluate the performance of Ref-SR under a realistic setting, we contribute the Webly-Referenced SR (WR-SR) dataset, mimicking the practical usage scenario. Extensive experiments demonstrate that our proposed C2-Matching significantly outperforms state of the arts by over 1dB on the standard CUFED5 benchmark. Notably, it also shows great generalizability on WR-SR dataset as well as robustness across large scale and rotation transformations 1."
                },
                {
                    "title": "MASA-SR: Matching Acceleration and Spatial Adaptation for Reference-Based Image Super-Resolution",
                    "abstract": "Reference-based image super-resolution (RefSR) has shown promising success in recovering high-frequency details by utilizing an external reference image (Ref). In this task, texture details are transferred from the Ref image to the low-resolution (LR) image according to their point- or patch-wise correspondence. Therefore, high-quality correspondence matching is critical. It is also desired to be computationally efficient. Besides, existing RefSR methods tend to ignore the potential large disparity in distributions between the LR and Ref images, which hurts the effectiveness of the information utilization. In this paper, we propose the MASA network for RefSR, where two novel modules are designed to address these problems. The proposed Match & Extraction Module significantly reduces the computational cost by a coarse-to-fine correspondence matching scheme. The Spatial Adaptation Module learns the difference of distribution between the LR and Ref images, and remaps the distribution of Ref features to that of LR features in a spatially adaptive way. This scheme makes the network robust to handle different reference images. Extensive quantitative and qualitative experiments validate the effectiveness of our proposed model."
                },
                {
                    "title": "Designing a Practical Degradation Model for Deep Blind Image Super-Resolution",
                    "abstract": "It is widely acknowledged that single image super-resolution (SISR) methods would not perform well if the assumed degradation model deviates from those in real images. Although several degradation models take additional factors into consideration, such as blur, they are still not effective enough to cover the diverse degradations of real images. To address this issue, this paper proposes to design a more complex but practical degradation model that consists of randomly shuffled blur, downsampling and noise degradations. Specifically, the blur is approximated by two convolutions with isotropic and anisotropic Gaussian kernels; the downsampling is randomly chosen from nearest, bilinear and bicubic interpolations; the noise is synthesized by adding Gaussian noise with different noise levels, adopting JPEG compression with different quality factors, and generating processed camera sensor noise via reverse-forward camera image signal processing (ISP) pipeline model and RAW image noise model. To verify the effectiveness of the new degradation model, we have trained a deep blind ES-RGAN super-resolver and then applied it to super-resolve both synthetic and real images with diverse degradations. The experimental results demonstrate that the new degradation model can help to significantly improve the practicability of deep super-resolvers, thus providing a powerful alternative solution for real SISR applications."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Pre-Trained Image Processing Transformer",
                    "abstract": "As the computing power of modern hardware is increasing strongly, pre-trained deep learning models (e.g., BERT, GPT-3) learned on large-scale datasets have shown their effectiveness over conventional methods. The big progress is mainly contributed to the representation ability of transformer and its variant architectures. In this paper, we study the low-level computer vision task (e.g., denoising, super-resolution and deraining) and develop a new pre-trained model, namely, image processing transformer (IPT). To maximally excavate the capability of transformer, we present to utilize the well-known ImageNet benchmark for generating a large amount of corrupted image pairs. The IPT model is trained on these images with multi-heads and multi-tails. In addition, the contrastive learning is introduced for well adapting to different image processing tasks. The pre-trained model can therefore efficiently employed on desired task after fine-tuning. With only one pre-trained model, IPT outperforms the current state-of-the-art methods on various low-level benchmarks. Code is available at https://github.com/huawei-noah/Pretrained-IPT and https://gitee.com/mindspore/mindspore/tree/master/model_zoo/research/cv/IPT"
                },
                {
                    "title": "Denoising Diffusion Implicit Models",
                    "abstract": "Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples $10 \\times$ to $50 \\times$ faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space."
                },
                {
                    "title": "Denoising Diffusion Probabilistic Models",
                    "abstract": "We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN. Our implementation is available at this https URL"
                },
                {
                    "title": "Learning Texture Transformer Network for Image Super-Resolution",
                    "abstract": "We study on image super-resolution (SR), which aims to recover realistic textures from a low-resolution (LR) image. Recent progress has been made by taking high-resolution images as references (Ref), so that relevant textures can be transferred to LR images. However, existing SR approaches neglect to use attention mechanisms to transfer high-resolution (HR) textures from Ref images, which limits these approaches in challenging cases. In this paper, we propose a novel Texture Transformer Network for Image Super-Resolution (TTSR), in which the LR and Ref images are formulated as queries and keys in a transformer, respectively. TTSR consists of four closely-related modules optimized for image generation tasks, including a learnable texture extractor by DNN, a relevance embedding module, a hard-attention module for texture transfer, and a soft-attention module for texture synthesis. Such a design encourages joint feature learning across LR and Ref images, in which deep feature correspondences can be discovered by attention, and thus accurate texture features can be transferred. The proposed texture transformer can be further stacked in a cross-scale way, which enables texture recovery from different levels (e.g., from 1x to 4x magnification). Extensive experiments show that TTSR achieves significant improvements over state-of-the-art approaches on both quantitative and qualitative evaluations."
                },
                {
                    "title": "Deep Multi-View Enhancement Hashing for Image Retrieval",
                    "abstract": "Hashing is an efficient method for nearest neighbor search in large-scale data space by embedding high-dimensional feature descriptors into a similarity preserving Hamming space with a low dimension. However, large-scale high-speed retrieval through binary code has a certain degree of reduction in retrieval accuracy compared to traditional retrieval methods. We have noticed that multi-view methods can well preserve the diverse characteristics of data. Therefore, we try to introduce the multi-view deep neural network into the hash learning field, and design an efficient and innovative retrieval model, which has achieved a significant improvement in retrieval performance. In this paper, we propose a supervised multi-view hash model which can enhance the multi-view information through neural networks. This is a completely new hash learning method that combines multi-view and deep learning methods. The proposed method utilizes an effective view stability evaluation method to actively explore the relationship among views, which will affect the optimization direction of the entire network. We have also designed a variety of multi-data fusion methods in the Hamming space to preserve the advantages of both convolution and multi-view. In order to avoid excessive computing resources on the enhancement procedure during retrieval, we set up a separate structure called memory network which participates in training together. The proposed method is systematically evaluated on the CIFAR-10, NUS-WIDE and MS-COCO datasets, and the results show that our method significantly outperforms the state-of-the-art single-view and multi-view hashing methods."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Second-Order Attention Network for Single Image Super-Resolution",
                    "abstract": "Recently, deep convolutional neural networks (CNNs) have been widely explored in single image super-resolution (SISR) and obtained remarkable performance. However, most of the existing CNN-based SISR methods mainly focus on wider or deeper architecture design, neglecting to explore the feature correlations of intermediate layers, hence hindering the representational power of CNNs. To address this issue, in this paper, we propose a second-order attention network (SAN) for more powerful feature expression and feature correlation learning. Specifically, a novel train- able second-order channel attention (SOCA) module is developed to adaptively rescale the channel-wise features by using second-order feature statistics for more discriminative representations. Furthermore, we present a non-locally enhanced residual group (NLRG) structure, which not only incorporates non-local operations to capture long-distance spatial contextual information, but also contains repeated local-source residual attention groups (LSRAG) to learn increasingly abstract feature representations. Experimental results demonstrate the superiority of our SAN network over state-of-the-art SISR methods in terms of both quantitative metrics and visual quality."
                },
                {
                    "title": "Image Super-Resolution by Neural Texture Transfer",
                    "abstract": "Due to the significant information loss in low-resolution (LR) images, it has become extremely challenging to further advance the state-of-the-art of single image super-resolution (SISR). Reference-based super-resolution (RefSR), on the other hand, has proven to be promising in recovering high-resolution (HR) details when a reference (Ref) image with similar content as that of the LR input is given. However, the quality of RefSR can degrade severely when Ref is less similar. This paper aims to unleash the potential of RefSR by leveraging more texture details from Ref images with stronger robustness even when irrelevant Ref images are provided. Inspired by the recent work on image stylization, we formulate the RefSR problem as neural texture transfer. We design an end-to-end deep model which enriches HR details by adaptively transferring the texture from Ref images according to their textural similarity. Instead of matching content in the raw pixel space as done by previous methods, our key contribution is a multi-level matching conducted in the neural space. This matching scheme facilitates multi-scale neural transfer that allows the model to benefit more from those semantically related Ref patches, and gracefully degrade to SISR performance on the least relevant Ref inputs. We build a benchmark dataset for the general research of RefSR, which contains Ref images paired with LR inputs with varying levels of similarity. Both quantitative and qualitative evaluations demonstrate the superiority of our method over state-of-the-art."
                },
                {
                    "title": "The Unreasonable Effectiveness of Deep Features as a Perceptual Metric",
                    "abstract": "While it is nearly effortless for humans to quickly assess the perceptual similarity between two images, the underlying processes are thought to be quite complex. Despite this, the most widely used perceptual metrics today, such as PSNR and SSIM, are simple, shallow functions, and fail to account for many nuances of human perception. Recently, the deep learning community has found that features of the VGG network trained on ImageNet classification has been remarkably useful as a training loss for image synthesis. But how perceptual are these so-called \"perceptual losses\"? What elements are critical for their success? To answer these questions, we introduce a new dataset of human perceptual similarity judgments. We systematically evaluate deep features across different architectures and tasks and compare them with classic metrics. We find that deep features outperform all previous metrics by large margins on our dataset. More surprisingly, this result is not restricted to ImageNet-trained VGG features, but holds across different deep architectures and levels of supervision (supervised, self-supervised, or even unsupervised). Our results suggest that perceptual similarity is an emergent property shared across deep visual representations."
                },
                {
                    "title": "NTIRE 2017 Challenge on Single Image Super-Resolution: Methods and Results",
                    "abstract": "This paper reviews the first challenge on single image super-resolution (restoration of rich details in an low resolution image) with focus on proposed solutions and results. A new DIVerse 2K resolution image dataset (DIV2K) was employed. The challenge had 6 competitions divided into 2 tracks with 3 magnification factors each. Track 1 employed the standard bicubic downscaling setup, while Track 2 had unknown downscaling operators (blur kernel and decimation) but learnable through low and high res train images. Each competition had \u223d100 registered participants and 20 teams competed in the final testing phase. They gauge the state-of-the-art in single image super-resolution."
                },
                {
                    "title": "GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium",
                    "abstract": "Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the \"Frechet Inception Distance\" (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark."
                },
                {
                    "title": "Learning Deep CNN Denoiser Prior for Image Restoration",
                    "abstract": "Model-based optimization methods and discriminative learning methods have been the two dominant strategies for solving various inverse problems in low-level vision. Typically, those two kinds of methods have their respective merits and drawbacks, e.g., model-based optimization methods are flexible for handling different inverse problems but are usually time-consuming with sophisticated priors for the purpose of good performance, in the meanwhile, discriminative learning methods have fast testing speed but their application range is greatly restricted by the specialized task. Recent works have revealed that, with the aid of variable splitting techniques, denoiser prior can be plugged in as a modular part of model-based optimization methods to solve other inverse problems (e.g., deblurring). Such an integration induces considerable advantage when the denoiser is obtained via discriminative learning. However, the study of integration with fast discriminative denoiser prior is still lacking. To this end, this paper aims to train a set of fast and effective CNN (convolutional neural network) denoisers and integrate them into model-based optimization method to solve other inverse problems. Experimental results demonstrate that the learned set of denoisers can not only achieve promising Gaussian denoising results but also can be used as prior to deliver good performance for various low-level vision applications."
                },
                {
                    "title": "Arbitrary Style Transfer in Real-Time with Adaptive Instance Normalization",
                    "abstract": "Gatys et al. recently introduced a neural algorithm that renders a content image in the style of another image, achieving so-called style transfer. However, their framework requires a slow iterative optimization process, which limits its practical application. Fast approximations with feed-forward neural networks have been proposed to speed up neural style transfer. Unfortunately, the speed improvement comes at a cost: the network is usually tied to a fixed set of styles and cannot adapt to arbitrary new styles. In this paper, we present a simple yet effective approach that for the first time enables arbitrary style transfer in real-time. At the heart of our method is a novel adaptive instance normalization (AdaIN) layer that aligns the mean and variance of the content features with those of the style features. Our method achieves speed comparable to the fastest existing approach, without the restriction to a pre-defined set of styles. In addition, our approach allows flexible user controls such as content-style trade-off, style interpolation, color & spatial controls, all using a single feed-forward neural network."
                },
                {
                    "title": "Event-Specific Image Importance",
                    "abstract": "When creating a photo album of an event, people typically select a few important images to keep or share. There is some consistency in the process of choosing the important images, and discarding the unimportant ones. Modeling this selection process will assist automatic photo selection and album summarization. In this paper, we show that the selection of important images is consistent among different viewers, and that this selection process is related to the event type of the album. We introduce the concept of event-specific image importance. We collected a new event album dataset with human annotation of the relative image importance with each event album. We also propose a Convolutional Neural Network (CNN) based method to predict the image importance score of a given event album, using a novel rank loss function and a progressive training scheme. Results demonstrate that our method significantly outperforms various baseline methods."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "Making a \u201cCompletely Blind\u201d Image Quality Analyzer",
                    "abstract": "An important aim of research on the blind image quality assessment (IQA) problem is to devise perceptual models that can predict the quality of distorted images with as little prior knowledge of the images or their distortions as possible. Current state-of-the-art \u201cgeneral purpose\u201d no reference (NR) IQA algorithms require knowledge about anticipated distortions in the form of training examples and corresponding human opinion scores. However we have recently derived a blind IQA model that only makes use of measurable deviations from statistical regularities observed in natural images, without training on human-rated distorted images, and, indeed without any exposure to distorted images. Thus, it is \u201ccompletely blind.\u201d The new IQA model, which we call the Natural Image Quality Evaluator (NIQE) is based on the construction of a \u201cquality aware\u201d collection of statistical features based on a simple and successful space domain natural scene statistic (NSS) model. These features are derived from a corpus of natural, undistorted images. Experimental results show that the new index delivers performance comparable to top performing NR IQA models that require training on large databases of human opinions of distorted images. A software release is available at http://live.ece.utexas.edu/research/quality/niqe_release.zip."
                },
                {
                    "title": "Retrieval-based Language Models and Applications",
                    "abstract": "Retrieval-based language models (LMs) have shown impressive performance on diverse NLP tasks. In this tutorial, we will provide a comprehensive and coherent overview of recent advances in retrieval-based LMs. We will start by providing preliminaries covering the foundation of LMs (e.g., masked LMs, autoregressive LMs) and retrieval systems (e.g., nearest-neighbor search). We will then detail recent progress in retrieval-based models, focusing on their model architectures and learning approaches. Finally, we will show how retrieval-based LMs are adapted to downstream applications, and extended to multilingual and multi-modal settings. Finally, we will use an exercise to showcase the effectiveness of retrieval-based LMs."
                },
                {
                    "title": "Retrieval-Augmented Diffusion Models",
                    "abstract": "Generative image synthesis with diffusion models has recently achieved excellent visual quality in several tasks such as text-based or class-conditional image synthesis. Much of this success is due to a dramatic increase in the computational capacity invested in training these models. This work presents an alternative approach: inspired by its successful application in natural language processing, we propose to complement the diffusion model with a retrieval-based approach and to introduce an explicit memory in the form of an external database. During training, our diffusion model is trained with similar visual features retrieved via CLIP and from the neighborhood of each training instance. By leveraging CLIP\u2019s joint image-text embedding space, our model achieves highly competitive performance on tasks for which it has not been explicitly trained, such as class-conditional or text-image synthesis, and can be conditioned on both text and image embeddings. Moreover, we can apply our approach to unconditional generation, where it achieves state-of-the-art performance. Our approach incurs low computational and memory overheads and is easy to implement. We discuss its relationship to concurrent work and will publish code and pretrained models soon."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively mitigate the hallucination problem in Large Restoration Models (LRMs) when restoring heavily degraded images without increasing computational and storage costs?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the hallucination problem in LRMs is crucial for advancing the field of image restoration, as it directly impacts the quality and fidelity of restored images. Addressing this issue could lead to more reliable applications in various domains, such as medical imaging, satellite imagery, and digital forensics, where accurate image restoration is essential. Furthermore, this research could inspire future studies on integrating external knowledge into machine learning models, potentially leading to breakthroughs in other areas of artificial intelligence.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing the hallucination problem stem from the inherent complexities of image restoration, particularly when dealing with low-quality inputs that may not align with the training data of LRMs. Naive approaches, such as simply increasing the model size or training data, may fail due to the prohibitive computational and storage costs involved. Additionally, the non-trivial task of allowing low-quality images to effectively utilize retrieved high-quality images during the restoration process presents significant technical and theoretical obstacles that need to be overcome.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on enhancing the internal capabilities of LRMs through larger models or more extensive datasets, overlooking the potential of external knowledge integration. Barriers such as the lack of effective methodologies for incorporating retrieved images into the restoration process have prevented this problem from being adequately addressed. Our approach differs by proposing a novel technique that leverages retrieval-augmented generation (RAG) principles, allowing LRMs to utilize external knowledge without the need for additional training or increased parameters.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves integrating a retrieval-augmented generation (RAG) framework into existing LRMs to enhance their performance on challenging image restoration tasks. We will utilize a dataset of high-quality images relevant to the restoration context and employ metrics such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) to evaluate the effectiveness of our approach. The expected outcomes include improved fidelity in restored images, particularly for heavily degraded inputs, demonstrating that external knowledge can significantly alleviate the hallucination problem in LRMs."
            }
        },
        "author_data": {
            "f1d1eb9b-9db1-4e48-b4b6-c60157ff48de": {
                "pk": "f1d1eb9b-9db1-4e48-b4b6-c60157ff48de",
                "name": "Hang Guo",
                "collaborators": [
                    "Zhengxi Hu",
                    "Jingtai Liu",
                    "Tao Dai",
                    "Guanghao Meng",
                    "Shu-Tao Xia",
                    "Xun Fan",
                    "Anh Cao",
                    "Geoff Outhred",
                    "John Heidemann"
                ],
                "domain": [
                    "Computer Vision",
                    "Anomaly Detection",
                    "Machine Learning",
                    "Image Processing"
                ],
                "publications": [
                    {
                        "title": "MGTR: End-to-End Mutual Gaze Detection with Transformer",
                        "abstract": "People's looking at each other or mutual gaze is ubiquitous in our daily interactions, and detecting mutual gaze is of great significance for understanding human social scenes. Current mutual gaze detection methods focus on two-stage methods, whose inference speed is limited by the two-stage pipeline and the performance in the second stage is affected by the first one. In this paper, we propose a novel one-stage mutual gaze detection framework called Mutual Gaze TRansformer or MGTR to perform mutual gaze detection in an end-to-end manner. By designing mutual gaze instance triples, MGTR can detect each human head bounding box and simultaneously infer mutual gaze relationship based on global image information, which streamlines the whole process with simplicity. Experimental results on two mutual gaze datasets show that our method is able to accelerate mutual gaze detection process without losing performance. Ablation study shows that different components of MGTR can capture different levels of semantic information in images. Code is available at https://github.com/Gmbition/MGTR"
                    },
                    {
                        "title": "Towards Robust Scene Text Image Super-resolution via Explicit Location Enhancement",
                        "abstract": "Scene text image super-resolution (STISR), aiming to improve image quality while boosting downstream scene text recognition accuracy, has recently achieved great success. However, most existing methods treat the foreground (character regions) and background (non-character regions) equally in the forward process, and neglect the disturbance from the complex background, thus limiting the performance. To address these issues, in this paper, we propose a novel method LEMMA that explicitly models character regions to produce high-level text-specific guidance for super-resolution. To model the location of characters effectively, we propose the location enhancement module to extract character region features based on the attention map sequence. Besides, we propose the multi-modal alignment module to perform bidirectional visual-semantic alignment to generate high-quality prior guidance, which is then incorporated into the super-resolution branch in an adaptive manner using the proposed adaptive fusion module. Experiments on TextZoom and four scene text recognition benchmarks demonstrate the superiority of our method over other state-of-the-art methods. Code is available at https://github.com/csguoh/LEMMA."
                    },
                    {
                        "title": "Peek Inside the Closed World: Evaluating Autoencoder-Based Detection of DDoS to Cloud",
                        "abstract": "Machine-learning-based anomaly detection (ML-based AD) has been successful at detecting DDoS events in the lab. However published evaluations of ML-based AD have used only limited data and provided minimal insight into why it works. To address limited evaluation against real-world data, we apply autoencoder, an existing ML-AD model, to 57 DDoS attack events captured at 5 cloud IPs from a major cloud provider. We show that our models detect nearly all malicious flows for 2 of the 4 cloud IPs under attack (at least 99.99%) and detect most malicious flows (94.75% and 91.37%) for the remaining 2 IPs. Our models also maintain near-zero false positives on benign flows to all 5 IPs. Our primary contribution is to improve our understanding for why ML-based AD works on some malicious flows but not others. We interpret our detection results with feature attribution and counterfactual explanation. We show that our models are better at detecting malicious flows with anomalies on allow-listed features (those with only a few benign values) than flows with anomalies on deny-listed features (those with mostly benign values) because our models are more likely to learn correct normality for allow-listed features. We then show that our models are better at detecting malicious flows with anomalies on unordered features (that have no ordering among their values) than flows with anomalies on ordered features because even with incomplete normality, our models could still detect anomalies on unordered feature with high recall. Lastly, we summarize the implications of what we learn on applying autoencoder-based AD in production: training with noisy real-world data is possible, autoencoder can reliably detect real-world anomalies on well-represented unordered features and combinations of autoencoder-based AD and heuristic-based filters can help both."
                    }
                ]
            },
            "c3200df3-fb75-47fc-9657-6c77e3ef78b1": {
                "pk": "c3200df3-fb75-47fc-9657-6c77e3ef78b1",
                "name": "Tao Dai",
                "collaborators": [
                    "Qiang Li",
                    "Shu-Tao Xia",
                    "Yan Feng",
                    "Bin Chen",
                    "Shutao Xia"
                ],
                "domain": [
                    "Deep Learning",
                    "Point Cloud",
                    "Adversarial Machine Learning",
                    "Image Retrieval"
                ],
                "publications": [
                    {
                        "title": "A Comparative Study of Feature Expansion Unit for 3D Point Cloud Upsampling",
                        "abstract": "Recently, deep learning methods have shown great success in 3D point cloud upsampling. Among these methods, many feature expansion units were proposed to complete point expansion at the end. In this paper, we compare various feature expansion units by both theoretical analysis and quantitative experiments. We show that most of the existing feature expansion units process each point feature independently, while ignoring the feature interaction among different points. Further, inspired by upsampling module of image super-resolution and recent success of dynamic graph CNN on point clouds, we propose a novel feature expansion units named ProEdgeShuffle. Experiments show that our proposed method can achieve considerable improvement over previous feature expansion units."
                    },
                    {
                        "title": "Adversarial Attack on Deep Product Quantization Network for Image Retrieval",
                        "abstract": "Deep product quantization network (DPQN) has recently received much attention in fast image retrieval tasks due to its efficiency of encoding high-dimensional visual features especially when dealing with large-scale datasets. Recent studies show that deep neural networks (DNNs) are vulnerable to input with small and maliciously designed perturbations (a.k.a., adversarial examples). This phenomenon raises the concern of security issues for DPQN in the testing/deploying stage as well. However, little effort has been devoted to investigating how adversarial examples affect DPQN. To this end, we propose product quantization adversarial generation (PQ-AG), a simple yet effective method to generate adversarial examples for product quantization based retrieval systems. PQ-AG aims to generate imperceptible adversarial perturbations for query images to form adversarial queries, whose nearest neighbors from a targeted product quantizaiton model are not semantically related to those from the original queries. Extensive experiments show that our PQ-AQ successfully creates adversarial examples to mislead targeted product quantization retrieval models. Besides, we found that our PQ-AG significantly degrades retrieval performance in both white-box and black-box settings."
                    }
                ]
            },
            "c28645f4-5a16-4978-8902-091ae26caac8": {
                "pk": "c28645f4-5a16-4978-8902-091ae26caac8",
                "name": "Zhihao Ouyang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "0e15dde0-a0d7-47da-b183-79acd6d77c43": {
                "pk": "0e15dde0-a0d7-47da-b183-79acd6d77c43",
                "name": "Taolin Zhang",
                "collaborators": [
                    "Xiaofeng He",
                    "Chengyu Wang",
                    "Jun Huang",
                    "Dongyang Li",
                    "Longtao Huang",
                    "Hui Xue",
                    "Minghui Qiu",
                    "Qizhou Chen",
                    "Shu-Tao Xia",
                    "Nan Hu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Knowledge Graph",
                    "Machine Learning",
                    "Model Editing"
                ],
                "publications": [
                    {
                        "title": "CIDR: A Cooperative Integrated Dynamic Refining Method for Minimal Feature Removal Problem",
                        "abstract": "The minimal feature removal problem in the post-hoc explanation area aims to identify the minimal feature set (MFS). Prior studies using the greedy algorithm to calculate the minimal feature set lack the exploration of feature interactions under a monotonic assumption which cannot be satisfied in general scenarios. In order to address the above limitations, we propose a Cooperative Integrated Dynamic Refining method (CIDR) to efficiently discover minimal feature sets. Specifically, we design Cooperative Integrated Gradients (CIG) to detect interactions between features. By incorporating CIG and characteristics of the minimal feature set, we transform the minimal feature removal problem into a knapsack problem. Additionally, we devise an auxiliary Minimal Feature Refinement algorithm to determine the minimal feature set from numerous candidate sets. To the best of our knowledge, our work is the first to address the minimal feature removal problem in the field of natural language processing. Extensive experiments demonstrate that CIDR is capable of tracing representative minimal feature sets with improved interpretability across various models and datasets."
                    },
                    {
                        "title": "HiCLRE: A Hierarchical Contrastive Learning Framework for Distantly Supervised Relation Extraction",
                        "abstract": "Distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships. Previous works of distantly supervised relation extraction (DSRE) task generally focus on sentence-level or bag-level de-noising techniques independently, neglecting the explicit interaction with cross levels. In this paper, we propose a hierarchical contrastive learning Framework for Distantly Supervised relation extraction (HiCLRE) to reduce noisy sentences, which integrate the global structural information and local fine-grained interaction. Specifically, we propose a three-level hierarchical learning framework to interact with cross levels, generating the de-noising context-aware representations via adapting the existing multi-head self-attention, named Multi-Granularity Recontextualization. Meanwhile, pseudo positive samples are also provided in the specific level for contrastive learning via a dynamic gradient-based data augmentation strategy, named Dynamic Gradient Adversarial Perturbation. Experiments demonstrate that HiCLRE significantly outperforms strong baselines in various mainstream DSRE datasets."
                    },
                    {
                        "title": "SMedBERT: A Knowledge-Enhanced Pre-trained Language Model with Structured Semantics for Medical Text Mining",
                        "abstract": "Recently, the performance of Pre-trained Language Models (PLMs) has been significantly improved by injecting knowledge facts to enhance their abilities of language understanding. For medical domains, the background knowledge sources are especially useful, due to the massive medical terms and their complicated relations are difficult to understand in text. In this work, we introduce SMedBERT, a medical PLM trained on large-scale medical corpora, incorporating deep structured semantic knowledge from neighbors of linked-entity.In SMedBERT, the mention-neighbor hybrid attention is proposed to learn heterogeneous-entity information, which infuses the semantic representations of entity types into the homogeneous neighboring entity structure. Apart from knowledge integration as external features, we propose to employ the neighbors of linked-entities in the knowledge graph as additional global contexts of text mentions, allowing them to communicate via shared neighbors, thus enrich their semantic representations. Experiments demonstrate that SMedBERT significantly outperforms strong baselines in various knowledge-intensive Chinese medical tasks. It also improves the performance of other tasks such as question answering, question matching and natural language inference."
                    },
                    {
                        "title": "Knowledge-Empowered Representation Learning for Chinese Medical Reading Comprehension: Task, Model and Resources",
                        "abstract": "Machine Reading Comprehension (MRC) aims to extract answers to questions given a passage. It has been widely studied recently, especially in open domains. However, few efforts have been made on closed-domain MRC, mainly due to the lack of large-scale training data. In this paper, we introduce a multi-target MRC task for the medical domain, whose goal is to predict answers to medical questions and the corresponding support sentences from medical information sources simultaneously, in order to ensure the high reliability of medical knowledge serving. A high-quality dataset is manually constructed for the purpose, named Multi-task Chinese Medical MRC dataset (CMedMRC), with detailed analysis conducted. We further propose the Chinese medical BERT model for the task (CMedBERT), which fuses medical knowledge into pre-trained language models by the dynamic fusion mechanism of heterogeneous features and the multi-task learning strategy. Experiments show that CMedBERT consistently outperforms strong baselines by fusing context-aware and knowledge-aware token representations."
                    },
                    {
                        "title": "Vision-Language Pre-training with Object Contrastive Learning for 3D Scene Understanding",
                        "abstract": "In recent years, vision language pre-training frameworks have made significant progress in natural language processing and computer vision, achieving remarkable performance improvement on various downstream tasks. However, when extended to point cloud data, existing works mainly focus on building task-specific models, and fail to extract universal 3D vision-language embedding that generalize well. We carefully investigate three common tasks in semantic 3D scene understanding, and derive key insights into the development of a pre-training model. Motivated by these observations, we propose a vision-language pre-training framework 3DVLP (3D vision-language pre-training with object contrastive learning), which transfers flexibly on 3D vision-language downstream tasks. 3DVLP takes visual grounding as the proxy task and introduces Object-level IoU-guided Detection (OID) loss to obtain high-quality proposals in the scene. Moreover, we design Object-level Cross-Contrastive alignment (OCC) task and Object-level Self-Contrastive learning (OSC) task to align the objects with descriptions and distinguish different objects in the scene, respectively. Extensive experiments verify the excellent performance of 3DVLP on three 3D vision-language tasks, reflecting its superiority in semantic 3D scene understanding."
                    },
                    {
                        "title": "From Complex to Simple: Unraveling the Cognitive Tree for Reasoning with Small Language Models",
                        "abstract": "Reasoning is a distinctive human capacity, enabling us to address complex problems by breaking them down into a series of manageable cognitive steps. Yet, complex logical reasoning is still cumbersome for language models. Based on the dual process theory in cognitive science, we are the first to unravel the cognitive reasoning abilities of language models. Our framework employs an iterative methodology to construct a Cognitive Tree (CogTree). The root node of this tree represents the initial query, while the leaf nodes consist of straightforward questions that can be answered directly. This construction involves two main components: the implicit extraction module (referred to as the intuitive system) and the explicit reasoning module (referred to as the reflective system). The intuitive system rapidly generates multiple responses by utilizing in-context examples, while the reflective system scores these responses using comparative learning. The scores guide the intuitive system in its subsequent generation step. Our experimental results on two popular and challenging reasoning tasks indicate that it is possible to achieve a performance level comparable to that of GPT-3.5 (with 175B parameters), using a significantly smaller language model that contains fewer parameters (<=7B) than 5% of GPT-3.5."
                    },
                    {
                        "title": "DKPLM: Decomposable Knowledge-enhanced Pre-trained Language Model for Natural Language Understanding",
                        "abstract": "Knowledge-Enhanced Pre-trained Language Models (KEPLMs) are pre-trained models with relation triples injecting from knowledge graphs to improve language understanding abilities. To guarantee effective knowledge injection, previous studies integrate models with knowledge encoders for representing knowledge retrieved from knowledge graphs. The operations for knowledge retrieval and encoding bring significant computational burdens, restricting the usage of such models in real-world applications that require high inference speed. In this paper, we propose a novel KEPLM named DKPLM that Decomposes Knowledge injection process of the Pre-trained Language Models in pre-training, fine-tuning and inference stages, which facilitates the applications of KEPLMs in real-world scenarios. Specifically, we first detect knowledge-aware long-tail entities as the target for knowledge injection, enhancing the KEPLMs' semantic understanding abilities and avoiding injecting redundant information. The embeddings of long-tail entities are replaced by \"pseudo token representations\" formed by relevant knowledge triples. We further design the relational knowledge decoding task for pre-training to force the models to truly understand the injected knowledge by relation triple reconstruction. Experiments show that our model outperforms other KEPLMs significantly over zero-shot knowledge probing tasks and multiple knowledge-aware language understanding tasks. We further show that DKPLM has a higher inference speed than other competing models due to the decomposing mechanism."
                    },
                    {
                        "title": "FedEgo: Privacy-preserving Personalized Federated Graph Learning with Ego-graphs",
                        "abstract": "As special information carriers containing both structure and feature information, graphs are widely used in graph mining, e.g., Graph Neural Networks (GNNs). However, in some practical scenarios, graph data are stored separately in multiple distributed parties, which may not be directly shared due to conflicts of interest. Hence, federated graph neural networks are proposed to address such data silo problems while preserving the privacy of each party (or client). Nevertheless, different graph data distributions among various parties, which is known as the statistical heterogeneity, may degrade the performance of naive federated learning algorithms like FedAvg. In this paper, we propose FedEgo, a federated graph learning framework based on ego-graphs to tackle the challenges above, where each client will train their local models while also contributing to the training of a global model. FedEgo applies GraphSAGE over ego-graphs to make full use of the structure information and utilizes Mixup for privacy concerns. To deal with the statistical heterogeneity, we integrate personalization into learning and propose an adaptive mixing coefficient strategy that enables clients to achieve their optimal personalization. Extensive experimental results and in-depth analysis demonstrate the effectiveness of FedEgo."
                    },
                    {
                        "title": "BoostAdapter: Improving Vision-Language Test-Time Adaptation via Regional Bootstrapping",
                        "abstract": "Adaptation of pretrained vision-language models such as CLIP to various downstream tasks have raised great interest in recent researches. Previous works have proposed a variety of test-time adaptation (TTA) methods to achieve strong generalization without any knowledge of the target domain. However, existing training-required TTA approaches like TPT necessitate entropy minimization that involves large computational overhead, while training-free methods like TDA overlook the potential for information mining from the test samples themselves. In this paper, we break down the design of existing popular training-required and training-free TTA methods and bridge the gap between them within our framework. Specifically, we maintain a light-weight key-value memory for feature retrieval from instance-agnostic historical samples and instance-aware boosting samples. The historical samples are filtered from the testing data stream and serve to extract useful information from the target distribution, while the boosting samples are drawn from regional bootstrapping and capture the knowledge of the test sample itself. We theoretically justify the rationality behind our method and empirically verify its effectiveness on both the out-of-distribution and the cross-domain datasets, showcasing its applicability in real-world situations."
                    },
                    {
                        "title": "KEHRL: Learning Knowledge-Enhanced Language Representations with Hierarchical Reinforcement Learning",
                        "abstract": "Knowledge-enhanced pre-trained language models (KEPLMs) leverage relation triples from knowledge graphs (KGs) and integrate these external data sources into language models via self-supervised learning. Previous works treat knowledge enhancement as two independent operations, i.e., knowledge injection and knowledge integration. In this paper, we propose to learn Knowledge-Enhanced language representations with Hierarchical Reinforcement Learning (KEHRL), which jointly addresses the problems of detecting positions for knowledge injection and integrating external knowledge into the model in order to avoid injecting inaccurate or irrelevant knowledge. Specifically, a high-level reinforcement learning (RL) agent utilizes both internal and prior knowledge to iteratively detect essential positions in texts for knowledge injection, which filters out less meaningful entities to avoid diverting the knowledge learning direction. Once the entity positions are selected, a relevant triple filtration module is triggered to perform low-level RL to dynamically refine the triples associated with polysemic entities through binary-valued actions. Experiments validate KEHRL's effectiveness in probing factual knowledge and enhancing the model's performance on various natural language understanding tasks."
                    },
                    {
                        "title": "Multimodal Label Relevance Ranking via Reinforcement Learning",
                        "abstract": "Conventional multi-label recognition methods often focus on label confidence, frequently overlooking the pivotal role of partial order relations consistent with human preference. To resolve these issues, we introduce a novel method for multimodal label relevance ranking, named Label Relevance Ranking with Proximal Policy Optimization (LR\\textsuperscript{2}PPO), which effectively discerns partial order relations among labels. LR\\textsuperscript{2}PPO first utilizes partial order pairs in the target domain to train a reward model, which aims to capture human preference intrinsic to the specific scenario. Furthermore, we meticulously design state representation and a policy loss tailored for ranking tasks, enabling LR\\textsuperscript{2}PPO to boost the performance of label relevance ranking model and largely reduce the requirement of partial order annotation for transferring to new scenes. To assist in the evaluation of our approach and similar methods, we further propose a novel benchmark dataset, LRMovieNet, featuring multimodal labels and their corresponding partial order data. Extensive experiments demonstrate that our LR\\textsuperscript{2}PPO algorithm achieves state-of-the-art performance, proving its effectiveness in addressing the multimodal label relevance ranking problem. Codes and the proposed LRMovieNet dataset are publicly available at \\url{https://github.com/ChazzyGordon/LR2PPO}."
                    },
                    {
                        "title": "Lifelong Knowledge Editing for LLMs with Retrieval-Augmented Continuous Prompt Learning",
                        "abstract": "Model editing aims to correct outdated or erroneous knowledge in large language models (LLMs) without the need for costly retraining. Lifelong model editing is the most challenging task that caters to the continuous editing requirements of LLMs. Prior works primarily focus on single or batch editing; nevertheless, these methods fall short in lifelong editing scenarios due to catastrophic knowledge forgetting and the degradation of model performance. Although retrieval-based methods alleviate these issues, they are impeded by slow and cumbersome processes of integrating the retrieved knowledge into the model. In this work, we introduce RECIPE, a RetriEval-augmented ContInuous Prompt lEarning method, to boost editing efficacy and inference efficiency in lifelong learning. RECIPE first converts knowledge statements into short and informative continuous prompts, prefixed to the LLM's input query embedding, to efficiently refine the response grounded on the knowledge. It further integrates the Knowledge Sentinel (KS) that acts as an intermediary to calculate a dynamic threshold, determining whether the retrieval repository contains relevant knowledge. Our retriever and prompt encoder are jointly trained to achieve editing properties, i.e., reliability, generality, and locality. In our experiments, RECIPE is assessed extensively across multiple LLMs and editing datasets, where it achieves superior editing performance. RECIPE also demonstrates its capability to maintain the overall performance of LLMs alongside showcasing fast editing and inference speed."
                    },
                    {
                        "title": "UniPSDA: Unsupervised Pseudo Semantic Data Augmentation for Zero-Shot Cross-Lingual Natural Language Understanding",
                        "abstract": "Cross-lingual representation learning transfers knowledge from resource-rich data to resource-scarce ones to improve the semantic understanding abilities of different languages. However, previous works rely on shallow unsupervised data generated by token surface matching, regardless of the global context-aware semantics of the surrounding text tokens. In this paper, we propose an Unsupervised Pseudo Semantic Data Augmentation (UniPSDA) mechanism for cross-lingual natural language understanding to enrich the training data without human interventions. Specifically, to retrieve the tokens with similar meanings for the semantic data augmentation across different languages, we propose a sequential clustering process in 3 stages: within a single language, across multiple languages of a language family, and across languages from multiple language families. Meanwhile, considering the multi-lingual knowledge infusion with context-aware semantics while alleviating computation burden, we directly replace the key constituents of the sentences with the above-learned multi-lingual family knowledge, viewed as pseudo-semantic. The infusion process is further optimized via three de-biasing techniques without introducing any neural parameters. Extensive experiments demonstrate that our model consistently improves the performance on general zero-shot cross-lingual natural language understanding tasks, including sequence classification, information extraction, and question answering."
                    },
                    {
                        "title": "Attribution Analysis Meets Model Editing: Advancing Knowledge Correction in Vision Language Models with VisEdit",
                        "abstract": "Model editing aims to correct outdated or erroneous knowledge in large models without costly retraining. Recent research discovered that the mid-layer representation of the subject's final token in a prompt has a strong influence on factual predictions, and developed Large Language Model (LLM) editing techniques based on this observation. However, for Vision-LLMs (VLLMs), how visual representations impact the predictions from a decoder-only language model remains largely unexplored. To the best of our knowledge, model editing for VLLMs has not been extensively studied in the literature. In this work, we employ the contribution allocation and noise perturbation methods to measure the contributions of visual representations for token predictions. Our attribution analysis shows that visual representations in mid-to-later layers that are highly relevant to the prompt contribute significantly to predictions. Based on these insights, we propose VisEdit, a novel model editor for VLLMs that effectively corrects knowledge by editing intermediate visual representations in regions important to the edit prompt. We evaluated VisEdit using multiple VLLM backbones and public VLLM editing benchmark datasets. The results show the superiority of VisEdit over the strong baselines adapted from existing state-of-the-art editors for LLMs."
                    },
                    {
                        "title": "Revisiting and Advancing Chinese Natural Language Understanding with Accelerated Heterogeneous Knowledge Pre-training",
                        "abstract": "Recently, knowledge-enhanced pre-trained language models (KEPLMs) improve context-aware representations via learning from structured relations in knowledge graphs, and/or linguistic knowledge from syntactic or dependency analysis. Unlike English, there is a lack of high-performing open-source Chinese KEPLMs in the natural language processing (NLP) community to support various language understanding applications. In this paper, we revisit and advance the development of Chinese natural language understanding with a series of novel Chinese KEPLMs released in various parameter sizes, namely CKBERT (Chinese knowledge-enhanced BERT).Specifically, both relational and linguistic knowledge is effectively injected into CKBERT based on two novel pre-training tasks, i.e., linguistic-aware masked language modeling and contrastive multi-hop relation modeling. Based on the above two pre-training paradigms and our in-house implemented TorchAccelerator, we have pre-trained base (110M), large (345M) and huge (1.3B) versions of CKBERT efficiently on GPU clusters. Experiments demonstrate that CKBERT outperforms strong baselines for Chinese over various benchmark NLP tasks and in terms of different model sizes."
                    },
                    {
                        "title": "TRELM: Towards Robust and Efficient Pre-training for Knowledge-Enhanced Language Models",
                        "abstract": "KEPLMs are pre-trained models that utilize external knowledge to enhance language understanding. Previous language models facilitated knowledge acquisition by incorporating knowledge-related pre-training tasks learned from relation triples in knowledge graphs. However, these models do not prioritize learning embeddings for entity-related tokens. Moreover, updating the entire set of parameters in KEPLMs is computationally demanding. This paper introduces TRELM, a Robust and Efficient Pre-training framework for Knowledge-Enhanced Language Models. We observe that entities in text corpora usually follow the long-tail distribution, where the representations of some entities are suboptimally optimized and hinder the pre-training process for KEPLMs. To tackle this, we employ a robust approach to inject knowledge triples and employ a knowledge-augmented memory bank to capture valuable information. Furthermore, updating a small subset of neurons in the feed-forward networks (FFNs) that store factual knowledge is both sufficient and efficient. Specifically, we utilize dynamic knowledge routing to identify knowledge paths in FFNs and selectively update parameters during pre-training. Experimental results show that TRELM reduces pre-training time by at least 50% and outperforms other KEPLMs in knowledge probing tasks and multiple knowledge-aware language understanding tasks."
                    },
                    {
                        "title": "R4: Reinforced Retriever-Reorder-Responder for Retrieval-Augmented Large Language Models",
                        "abstract": "Retrieval-augmented large language models (LLMs) leverage relevant content retrieved by information retrieval systems to generate correct responses, aiming to alleviate the hallucination problem. However, existing retriever-responder methods typically append relevant documents to the prompt of LLMs to perform text generation tasks without considering the interaction of fine-grained structural semantics between the retrieved documents and the LLMs. This issue is particularly important for accurate response generation as LLMs tend to \"lose in the middle\" when dealing with input prompts augmented with lengthy documents. In this work, we propose a new pipeline named \"Reinforced Retriever-Reorder-Responder\" (R$^4$) to learn document orderings for retrieval-augmented LLMs, thereby further enhancing their generation abilities while the large numbers of parameters of LLMs remain frozen. The reordering learning process is divided into two steps according to the quality of the generated responses: document order adjustment and document representation enhancement. Specifically, document order adjustment aims to organize retrieved document orderings into beginning, middle, and end positions based on graph attention learning, which maximizes the reinforced reward of response quality. Document representation enhancement further refines the representations of retrieved documents for responses of poor quality via document-level gradient adversarial learning. Extensive experiments demonstrate that our proposed pipeline achieves better factual question-answering performance on knowledge-intensive tasks compared to strong baselines across various public datasets. The source codes and trained models will be released upon paper acceptance."
                    },
                    {
                        "title": "DAFNet: Dynamic Auxiliary Fusion for Sequential Model Editing in Large Language Models",
                        "abstract": "Recently, while large language models (LLMs) have demonstrated impressive results, they still suffer from hallucination, i.e., the generation of false information. Model editing is the task of fixing factual mistakes in LLMs; yet, most previous works treat it as a one-time task, paying little attention to ever-emerging mistakes generated by LLMs. We address the task of sequential model editing (SME) that aims to rectify mistakes continuously. A Dynamic Auxiliary Fusion Network (DAFNet) is designed to enhance the semantic interaction among the factual knowledge within the entire sequence, preventing catastrophic forgetting during the editing process of multiple knowledge triples. Specifically, (1) for semantic fusion within a relation triple, we aggregate the intra-editing attention flow into auto-regressive self-attention with token-level granularity in LLMs. We further leverage multi-layer diagonal inter-editing attention flow to update the weighted representations of the entire sequence-level granularity. (2) Considering that auxiliary parameters are required to store the knowledge for sequential editing, we construct a new dataset named \\textbf{DAFSet}, fulfilling recent, popular, long-tail and robust properties to enhance the generality of sequential editing. Experiments show DAFNet significantly outperforms strong baselines in single-turn and sequential editing. The usage of DAFSet also consistently improves the performance of other auxiliary network-based methods in various scenarios"
                    },
                    {
                        "title": "Parameter-Efficient and Memory-Efficient Tuning for Vision Transformer: A Disentangled Approach",
                        "abstract": "Recent works on parameter-efficient transfer learning (PETL) show the potential to adapt a pre-trained Vision Transformer to downstream recognition tasks with only a few learnable parameters. However, since they usually insert new structures into the pre-trained model, entire intermediate features of that model are changed and thus need to be stored to be involved in back-propagation, resulting in memory-heavy training. We solve this problem from a novel disentangled perspective, i.e., dividing PETL into two aspects: task-specific learning and pre-trained knowledge utilization. Specifically, we synthesize the task-specific query with a learnable and lightweight module, which is independent of the pre-trained model. The synthesized query equipped with task-specific knowledge serves to extract the useful features for downstream tasks from the intermediate representations of the pre-trained model in a query-only manner. Built upon these features, a customized classification head is proposed to make the prediction for the input sample. lightweight architecture and avoids the use of heavy intermediate features for running gradient descent, it demonstrates limited memory usage in training. Extensive experiments manifest that our method achieves state-of-the-art performance under memory constraints, showcasing its applicability in real-world situations."
                    },
                    {
                        "title": "EasyNLP: A Comprehensive and Easy-to-use Toolkit for Natural Language Processing",
                        "abstract": "The success of Pre-Trained Models (PTMs) has reshaped the development of Natural Language Processing (NLP). Yet, it is not easy to obtain high-performing models and deploy them online for industrial practitioners. To bridge this gap, EasyNLP is designed to make it easy to build NLP applications, which supports a comprehensive suite of NLP algorithms. It further features knowledge-enhanced pre-training, knowledge distillation and few-shot learning functionalities for large-scale PTMs, and provides a unified framework of model training, inference and deployment for real-world applications. Currently, EasyNLP has powered over ten business units within Alibaba Group and is seamlessly integrated to the Platform of AI (PAI) products on Alibaba Cloud. The source code of our EasyNLP toolkit is released at GitHub (https://github.com/alibaba/EasyNLP)."
                    }
                ]
            },
            "c6304b43-e3dd-452c-8843-257d25400fe1": {
                "pk": "c6304b43-e3dd-452c-8843-257d25400fe1",
                "name": "Yaohua Zha",
                "collaborators": [
                    "Tao Dai",
                    "Shu-Tao Xia",
                    "Bin Chen",
                    "Zhi Wang",
                    "Yanzi Wang",
                    "Jinpeng Wang",
                    "Jinmin Li",
                    "Hang Guo",
                    "Zhihao Ouyang",
                    "Longfei Lu"
                ],
                "domain": [
                    "3D Representation",
                    "Point Cloud",
                    "Machine Learning",
                    "Autoencoder"
                ],
                "publications": [
                    {
                        "title": "Pre-training Point Cloud Compact Model with Partial-aware Reconstruction",
                        "abstract": "The pre-trained point cloud model based on Masked Point Modeling (MPM) has exhibited substantial improvements across various tasks. However, two drawbacks hinder their practical application. Firstly, the positional embedding of masked patches in the decoder results in the leakage of their central coordinates, leading to limited 3D representations. Secondly, the excessive model size of existing MPM methods results in higher demands for devices. To address these, we propose to pre-train Point cloud Compact Model with Partial-aware \\textbf{R}econstruction, named Point-CPR. Specifically, in the decoder, we couple the vanilla masked tokens with their positional embeddings as randomly masked queries and introduce a partial-aware prediction module before each decoder layer to predict them from the unmasked partial. It prevents the decoder from creating a shortcut between the central coordinates of masked patches and their reconstructed coordinates, enhancing the robustness of models. We also devise a compact encoder composed of local aggregation and MLPs, reducing the parameters and computational requirements compared to existing Transformer-based encoders. Extensive experiments demonstrate that our model exhibits strong performance across various tasks, especially surpassing the leading MPM-based model PointGPT-B with only 2% of its parameters."
                    },
                    {
                        "title": "Instance-aware Dynamic Prompt Tuning for Pre-trained Point Cloud Models",
                        "abstract": "Pre-trained point cloud models have found extensive applications in 3D understanding tasks like object classification and part segmentation. However, the prevailing strategy of full fine-tuning in downstream tasks leads to large per-task storage overhead for model parameters, which limits the efficiency when applying large-scale pre-trained models. Inspired by the recent success of visual prompt tuning (VPT), this paper attempts to explore prompt tuning on pre-trained point cloud models, to pursue an elegant balance between performance and parameter efficiency. We find while instance-agnostic static prompting, e.g. VPT, shows some efficacy in downstream transfer, it is vulnerable to the distribution diversity caused by various types of noises in real-world point cloud data. To conquer this limitation, we propose a novel Instance-aware Dynamic Prompt Tuning (IDPT) strategy for pre-trained point cloud models. The essence of IDPT is to develop a dynamic prompt generation module to perceive semantic prior features of each point cloud instance and generate adaptive prompt tokens to enhance the model's robustness. Notably, extensive experiments demonstrate that IDPT outperforms full fine-tuning in most tasks with a mere 7% of the trainable parameters, providing a promising solution to parameter-efficient learning for pre-trained point cloud models. Code is available at \\url{https://github.com/zyh16143998882/ICCV23-IDPT}."
                    },
                    {
                        "title": "Towards Compact 3D Representations via Point Feature Enhancement Masked Autoencoders",
                        "abstract": "Learning 3D representation plays a critical role in masked autoencoder (MAE) based pre-training methods for point cloud, including single-modal and cross-modal based MAE. Specifically, although cross-modal MAE methods learn strong 3D representations via the auxiliary of other modal knowledge, they often suffer from heavy computational burdens and heavily rely on massive cross-modal data pairs that are often unavailable, which hinders their applications in practice. Instead, single-modal methods with solely point clouds as input are preferred in real applications due to their simplicity and efficiency. However, such methods easily suffer from limited 3D representations with global random mask input. To learn compact 3D representations, we propose a simple yet effective Point Feature Enhancement Masked Autoencoders (Point-FEMAE), which mainly consists of a global branch and a local branch to capture latent semantic features. Specifically, to learn more compact features, a share-parameter Transformer encoder is introduced to extract point features from the global and local unmasked patches obtained by global random and local block mask strategies, followed by a specific decoder to reconstruct. Meanwhile, to further enhance features in the local branch, we propose a Local Enhancement Module with local patch convolution to perceive fine-grained local context at larger scales. Our method significantly improves the pre-training efficiency compared to cross-modal alternatives, and extensive downstream experiments underscore the state-of-the-art effectiveness, particularly outperforming our baseline (Point-MAE) by 5.16%, 5.00%, and 5.04% in three variants of ScanObjectNN, respectively. The code is available at https://github.com/zyh16143998882/AAAI24-PointFEMAE."
                    },
                    {
                        "title": "LCM: Locally Constrained Compact Point Cloud Model for Masked Point Modeling",
                        "abstract": "The pre-trained point cloud model based on Masked Point Modeling (MPM) has exhibited substantial improvements across various tasks. However, these models heavily rely on the Transformer, leading to quadratic complexity and limited decoder, hindering their practice application. To address this limitation, we first conduct a comprehensive analysis of existing Transformer-based MPM, emphasizing the idea that redundancy reduction is crucial for point cloud analysis. To this end, we propose a Locally constrained Compact point cloud Model (LCM) consisting of a locally constrained compact encoder and a locally constrained Mamba-based decoder. Our encoder replaces self-attention with our local aggregation layers to achieve an elegant balance between performance and efficiency. Considering the varying information density between masked and unmasked patches in the decoder inputs of MPM, we introduce a locally constrained Mamba-based decoder. This decoder ensures linear complexity while maximizing the perception of point cloud geometry information from unmasked patches with higher information density. Extensive experimental results show that our compact model significantly surpasses existing Transformer-based models in both performance and efficiency, especially our LCM-based Point-MAE model, compared to the Transformer-based model, achieved an improvement of 2.24%, 0.87%, and 0.94% in performance on the three variants of ScanObjectNN while reducing parameters by 88% and computation by 73%."
                    },
                    {
                        "title": "Large Point-to-Gaussian Model for Image-to-3D Generation",
                        "abstract": "Recently, image-to-3D approaches have significantly advanced the generation quality and speed of 3D assets based on large reconstruction models, particularly 3D Gaussian reconstruction models. Existing large 3D Gaussian models directly map 2D image to 3D Gaussian parameters, while regressing 2D image to 3D Gaussian representations is challenging without 3D priors. In this paper, we propose a large Point-to-Gaussian model, that inputs the initial point cloud produced from large 3D diffusion model conditional on 2D image to generate the Gaussian parameters, for image-to-3D generation. The point cloud provides initial 3D geometry prior for Gaussian generation, thus significantly facilitating image-to-3D Generation. Moreover, we present the \\textbf{A}ttention mechanism, \\textbf{P}rojection mechanism, and \\textbf{P}oint feature extractor, dubbed as \\textbf{APP} block, for fusing the image features with point cloud features. The qualitative and quantitative experiments extensively demonstrate the effectiveness of the proposed approach on GSO and Objaverse datasets, and show the proposed method achieves state-of-the-art performance."
                    },
                    {
                        "title": "Invertible Residual Rescaling Models",
                        "abstract": "Invertible Rescaling Networks (IRNs) and their variants have witnessed remarkable achievements in various image processing tasks like image rescaling. However, we observe that IRNs with deeper networks are difficult to train, thus hindering the representational ability of IRNs. To address this issue, we propose Invertible Residual Rescaling Models (IRRM) for image rescaling by learning a bijection between a high-resolution image and its low-resolution counterpart with a specific distribution. Specifically, we propose IRRM to build a deep network, which contains several Residual Downscaling Modules (RDMs) with long skip connections. Each RDM consists of several Invertible Residual Blocks (IRBs) with short connections. In this way, RDM allows rich low-frequency information to be bypassed by skip connections and forces models to focus on extracting high-frequency information from the image. Extensive experiments show that our IRRM performs significantly better than other state-of-the-art methods with much fewer parameters and complexity. Particularly, our IRRM has respectively PSNR gains of at least 0.3 dB over HCFlow and IRN in the x4 rescaling while only using 60% parameters and 50% FLOPs. The code will be available at https://github.com/THU-Kingmin/IRRM."
                    },
                    {
                        "title": "Block-to-Scene Pre-training for Point Cloud Hybrid-Domain Masked Autoencoders",
                        "abstract": "Point clouds, as a primary representation of 3D data, can be categorized into scene domain point clouds and object domain point clouds based on the modeled content. Masked autoencoders (MAE) have become the mainstream paradigm in point clouds self-supervised learning. However, existing MAE-based methods are domain-specific, limiting the model's generalization. In this paper, we propose to pre-train a general Point cloud Hybrid-Domain Masked AutoEncoder (PointHDMAE) via a block-to-scene pre-training strategy. We first propose a hybrid-domain masked autoencoder consisting of an encoder and decoder belonging to the scene domain and object domain, respectively. The object domain encoder specializes in handling object point clouds and multiple shared object encoders assist the scene domain encoder in analyzing the scene point clouds. Furthermore, we propose a block-to-scene strategy to pre-train our hybrid-domain model. Specifically, we first randomly select point blocks within a scene and apply a set of transformations to convert each point block coordinates from the scene space to the object space. Then, we employ an object-level mask and reconstruction pipeline to recover the masked points of each block, enabling the object encoder to learn a universal object representation. Finally, we introduce a scene-level block position regression pipeline, which utilizes the blocks' features in the object space to regress these blocks' initial positions within the scene space, facilitating the learning of scene representations. Extensive experiments across different datasets and tasks demonstrate the generalization and superiority of our hybrid-domain model."
                    },
                    {
                        "title": "Towards Scalable Semantic Representation for Recommendation",
                        "abstract": "With recent advances in large language models (LLMs), there has been emerging numbers of research in developing Semantic IDs based on LLMs to enhance the performance of recommendation systems. However, the dimension of these embeddings needs to match that of the ID embedding in recommendation, which is usually much smaller than the original length. Such dimension compression results in inevitable losses in discriminability and dimension robustness of the LLM embeddings, which motivates us to scale up the semantic representation. In this paper, we propose Mixture-of-Codes, which first constructs multiple independent codebooks for LLM representation in the indexing stage, and then utilizes the Semantic Representation along with a fusion module for the downstream recommendation stage. Extensive analysis and experiments demonstrate that our method achieves superior discriminability and dimension robustness scalability, leading to the best scale-up performance in recommendations."
                    }
                ]
            },
            "59e18a2b-3bc9-451d-a6f2-856176fcffb3": {
                "pk": "59e18a2b-3bc9-451d-a6f2-856176fcffb3",
                "name": "Bin Chen",
                "collaborators": [
                    "Xiao Liu",
                    "Jiang Long",
                    "Bin Bin Chen",
                    "Pascale Primet",
                    "Peibiao Zhao",
                    "Wei He",
                    "Xi Li",
                    "Xinmin Hou",
                    "Zhibo Xu",
                    "Ronald Kantowski"
                ],
                "domain": [
                    "Theoretical Physics",
                    "String Theory",
                    "General Relativity",
                    "Mathematical Physics"
                ],
                "publications": [
                    {
                        "title": "The Time-dependent Supersymmetric Configurations in M-theory and Matrix Models",
                        "abstract": "In this paper, we study the half-supersymmetric time-dependent configurations in M-theory and their matrix models. We find a large class of 11D supergravity solutions, which keeps sixteen supersymmetries. Furthermore, we investigate the isometries of these configurations and show that in general these configurations have no supernumerary supersymmetries. And also we define the Matrix models in these backgrounds following Discrete Light-Cone Quantization (DLCQ) prescription."
                    },
                    {
                        "title": "On almost-prime $k$-tuples",
                        "abstract": "Let $\\tau$ denote the divisor function and $\\mathcal{H}=\\{h_{1},...,h_{k}\\}$ be an admissible set. We prove that there are infinitely many $n$ for which the product $\\prod_{i=1}^{k}(n+h_{i})$ is square-free and $\\sum_{i=1}^{k}\\tau(n+h_{i})\\leq \\lfloor \\rho_{k}\\rfloor$, where $\\rho_{k}$ is asymptotic to $\\frac{2126}{2853} k^{2}$. It improves a previous result of M. Ram Murty and A. Vatwani, replacing $2126/2853$ by $3/4$. The main ingredients in our proof are the higher rank Selberg sieve and Irving-Wu-Xi estimate for the divisor function in arithmetic progressions to smooth moduli."
                    },
                    {
                        "title": "Colliding Plane Wave Solutions in String theory Revisited",
                        "abstract": "We construct the colliding plane wave solutions in the higher-dimensional gravity theory with fluxes and dilaton, with a more general ansatz on the metric. We consider two classes of solutions to the equations of motions and after imposing the junction conditions we find that they are all physically acceptable. In particular, we manage to obtain the higher-dimensional Bell-Szekeres solutions in the Maxwell-Einstein gravity theory, and the flux-CPW solutions in the eleven-dimensional supergravity theory. All the solutions have been shown to develop the late time curvature singularity."
                    },
                    {
                        "title": "The Self-dual String Soliton in AdS_4\\times S^7 spacetime",
                        "abstract": "We construct self-dual string soliton solutions in $AdS_4\\times   S^7$ spacetime, starting from the covariant equations of motion of M5-brane. We study the properties of the solutions and find that their action are linearized, indicating the BPS nature of the solutions, and they have the same electric and magnetic charge. The straight string soliton solution represents the configuration of the membranes ending on M5-brane with a straight string intersection, and it behaves like the spiky solution in flat spacetime. The spherical string soliton solution, which could be related to the straight one by a conformal transformation, represents the membranes ending on M5-brane with a spherical intersection."
                    },
                    {
                        "title": "Probing the gravitational Faraday rotation using quasar X-ray microlensing",
                        "abstract": "The effect of gravitational Faraday rotation was predicted in the 1950s, but there is currently no practical method for measuring this effect. Measuring this effect is important because it will provide new evidence for correctness of general relativity, in particular, in the strong field limit. We predict that the observed degree and angle of the X-ray polarization of a cosmologically distant quasar microlensed by the random star field in a foreground galaxy or cluster lens vary rapidly and concurrently with flux during caustic-crossing events using the first simulation of quasar X-ray microlensing polarization light curves. Therefore, it is possible to detect gravitational Faraday rotation by monitoring the X-ray polarization of gravitationally microlensed quasars. Detecting this effect will also confirm the strong gravity nature of quasar X-ray emission."
                    },
                    {
                        "title": "Small gaps between almost-twin primes",
                        "abstract": "Let $m \\in \\mathbb{N}$ be large. We show that there exist infinitely many primes $q_{1}< \\cdot\\cdot\\cdot < q_{m+1}$ such that \\[ q_{m+1}-q_{1}=O(e^{7.63m}) \\] and $q_{j}+2$ has at most \\[ \\frac{7.36m}{\\log 2} + \\frac{4\\log m}{\\log 2} + 21 \\] prime factors for each $1 \\leq j \\leq m+1$. This improves the previous result of Li and Pan, replacing $e^{7.63m}$ by $m^{4}e^{8m}$ and $\\frac{7.36m}{\\log 2} + \\frac{4\\log m}{\\log 2} + 21$ by $\\frac{16m}{\\log 2} + \\frac{5\\log m}{\\log 2} + 37$. The main inputs are the Maynard-Tao sieve, a minorant for the indicator function of the primes constructed by Baker and Irving, for which a stronger equidistribution theorem in arithmatic progressions to smooth moduli is applicable, and Tao's approach previously used to estimate $\\sum_{x \\leq n < 2x} \\mathbf{1}_{\\mathbb{P}}(n)\\mathbf{1}_{\\mathbb{P}}(n+12)\\omega_{n}$, where $\\mathbf{1}_{\\mathbb{P}}$ stands for the characteristic function of the primes and $\\omega_{n}$ are multidimensional sieve weights."
                    },
                    {
                        "title": "A Flexible Bandwidth Reservation Framework for Bulk Data Transfers in Grid Networks",
                        "abstract": "In grid networks, distributed resources are interconnected by wide area network to support compute and data-intensive applications, which require reliable and efficient transfer of gigabits (even terabits) of data. Different from best-effort traffic in Internet, bulk data transfer in grid requires bandwidth reservation as a fundamental service. Existing reservation schemes such as RSVP are designed for real-time traffic specified by reservation rate, transfer start time but with unknown lifetime. In comparison, bulk data transfer requests are defined in terms of volume and deadline, which provide more information, and allow more flexibility in reservation schemes, i.e., transfer start time can be flexibly chosen, and reservation for a single request can be divided into multiple intervals with different reservation rates. We define a flexible reservation framework using time-rate function algebra, and identify a series of practical reservation scheme families with increasing generality and potential performance, namely, FixTime-FixRate, FixTime-FlexRate, FlexTime-FlexRate, and Multi-Interval. Simple heuristics are used to select representative scheme from each family for performance comparison. Simulation results show that the increasing flexibility can potentially improve system performance, minimizing both blocking probability and mean flow time. We also discuss the distributed implementation of proposed framework."
                    },
                    {
                        "title": "A Gauss curvature flow related to the Orlicz-Aleksandrov problem",
                        "abstract": "In this paper we first obtain the existence of smooth solutions to Orlicz-Aleksandrov problem via a Gauss-like curvature flow."
                    },
                    {
                        "title": "On 1/2-BPS Wilson-'t Hooft loops",
                        "abstract": "We investigate the 1/2-BPS Wilson-'t Hooft loops in ${\\cal N}=4$ Super-Yang-Mills theory. We use the bulk D-brane with both electric and magnetic charges to calculate the all genus contribution of the circular loops. The expectation value of Wilson-'t Hooft loops are in perfect agreement with the result through supersymmetric condition and duality transformation in the gauge theory."
                    },
                    {
                        "title": "Tiny Graviton Matrix Theory On Time-Dependent Background",
                        "abstract": "In this article we construct a tiny graviton matrix model for type IIB string theory on a plane-wave background with null dilaton. For the linear null dilaton case, we analyze its vacuum and the excitation spectrum around the vacuum, and discuss the time-dependent fuzzy three-sphere solutions and their evolution. It turns out that at very late time the non-Abelian fuzzy degrees of freedom disappear, which indicates the appearance of perturbative strings."
                    },
                    {
                        "title": "D1-D3 (or $\\bar{\\textrm{D}3}$) Systems with Fluxes",
                        "abstract": "In this article we study D1-D3 (or $\\bar{\\textrm{D}3}$) brane systems with generic constant electric and magnetic fluxes in IIB string theory. We work out all possible supersymmetric configurations and find out via T-duality all of them and corresponding supersymmetry conditions could be related to the supersymmetric intersecting D1-D1 pairs. And we do D1-D3 (or $\\bar{\\textrm{D}3}$) open string quantization for a class of configurations. We find that there are many near massless states in NS sector for near-BPS configurations. Furthermore we calculate open string pair creation rate in generic nonsupersymmetric configurations."
                    },
                    {
                        "title": "Co-regulation of multi-cellular swirling by individual cellular chirality and boundaries",
                        "abstract": "To understand the relationship between the chirality of individual cells and that of tissues and organisms, we have developed a chiral polarized particle model to investigate the movement of cell populations on substrates. Our model analysis indicates that cells with the same chirality can form distinct chiral patterns on ring-shaped or rectangular substrates. Our model analysis also reveals the importance of coordination between boundary features and individual cellular chirality in regulating the movement of cell populations. This work provides valuable insights into comprehending the intricate connection between the chirality of single cells and that of tissues and organisms."
                    },
                    {
                        "title": "Maximum size of $C_{\\leq k}$-free strong digraphs with out-degree at least two",
                        "abstract": "Let $\\mathscr{H}$ be a family of digraphs. A digraph $D$ is \\emph{$\\mathscr{H}$-free} if it contains no isomorphic copy of any member of $\\mathscr{H}$. For $k\\geq2$, we set $C_{\\leq k}=\\{C_{2}, C_{3},\\ldots,C_{k}\\}$, where $C_{\\ell}$ is a directed cycle of length $\\ell\\in\\{2,3,\\ldots,k\\}$. Let $D_{n}^{k}(\\xi,\\zeta)$ denote the family of \\emph{${C}_{\\le k}$-free} strong digraphs on $n$ vertices with every vertex having out-degree at least $\\xi$ and in-degree at least $\\zeta$, where both $\\xi$ and $\\zeta$ are positive integers. Let $\\varphi_{n}^{k}(\\xi,\\zeta)=\\max\\{|A(D)|:\\;D\\in D_{n}^{k}(\\xi,\\zeta)\\}$ and $\\Phi_{n}^{k}(\\xi,\\zeta)=\\{D\\in D_{n}^{k}(\\xi,\\zeta): |A(D)|=\\varphi_{n}^{k}(\\xi,\\zeta)\\}$. Bermond et al.\\;(1980) verified that $\\varphi_{n}^{k}(1,1)=\\binom{n-k+2}{2}+k-2$. Chen and Chang\\;(2021) showed that $\\binom{n-1}{2}-2\\leq\\varphi_{n}^{3}(2,1)\\leq\\binom{n-1}{2}$. This upper bound was further improved to $\\binom{n-1}{2}-1$ by Chen and Chang\\;(DAM, 2022), furthermore, they also gave the exact values of $\\varphi_{n}^{3}(2,1)$ for $n\\in \\{7,8,9\\}$. In this paper, we continue to determine the exact values of $\\varphi_{n}^{3}(2,1)$ for $n\\ge 10$, i.e., $\\varphi_{n}^{3}(2,1)=\\binom{n-1}{2}-2$ for $n\\geq10$."
                    },
                    {
                        "title": "Hawking Radiation from General Kerr-(anti)de Sitter Black Holes",
                        "abstract": "We calculate the total flux of Hawking radiation from Kerr-(anti)de Sitter black holes by using gravitational anomaly method developed in gr-qc/0502074. We consider the general Kerr-(anti)de Sitter black holes in arbitrary $D$ dimensions with the maximal number [D/2] of independent rotating parameters. We find that the physics near the horizon can be described by an infinite collection of $(1+1)$-dimensional quantum fields coupled to a set of gauge fields with charges proportional to the azimuthal angular momentums $m_i$. With the requirement of anomaly cancellation and regularity at the horizon, the Hawking radiation is determined."
                    },
                    {
                        "title": "Distance-redshift from an optical metric that includes absorption",
                        "abstract": "We show that it is possible to equate the intensity reduction of a light wave caused by weak absorption with a geometrical reduction in intensity caused by a \"transverse\" conformal transformation of the spacetime metric in which the wave travels. We are consequently able to modify Gordon's optical metric to account for electromagnetic properties of ponderable material whose properties include both refraction and absorption. Unlike refraction alone however, including absorption requires a modification of the optical metric that depends on the eikonal of the wave itself. We derive the distance-redshift relation from the modified optical metric for Friedman-Lema\\^itre-Robertson-Walker spacetimes whose cosmic fluid has associated refraction and absorption coefficients. We then fit the current supernovae data and provide an alternate explanation (other than dark energy) of the apparent acceleration of the universe."
                    },
                    {
                        "title": "Real-time Correlators and Hidden Conformal Symmetry in Kerr/CFT Correspondence",
                        "abstract": "In this paper, we study the real-time correlators in Kerr/CFT, in the low frequency limit of generic non-extremal Kerr(-Newman) black holes. From the low frequency scattering of Kerr-Newman black holes, we show that for the uncharged scalar scattering, there exists hidden conformal symmetry on the solution space. Similar to Kerr case, this suggests that the Kerr-Newman black hole is dual to a two-dimensional CFT with central charges $c_L=c_R=12J$ and temperatures $T_L=\\frac{(r_++r_-)-Q^2/M}{4\\pi a}, T_R=\\frac{r_+-r_-}{4\\pi a}$. Using the Minkowski prescription, we compute the real-time correlators of charged scalar and find perfect match with CFT prediction. We further discuss the low-frequency scattering of photons and gravitons by Kerr black hole and find that their retarded Green's functions are in good agreement with CFT prediction. Our study supports the idea that the hidden conformal symmetry in the solution space is essential to Kerr/CFT correspondence."
                    },
                    {
                        "title": "Hidden Conformal Symmetry and Quasi-normal Modes",
                        "abstract": "We provide an algebraic way to calculate the quasi-normal modes of a black hole, which possesses a hidden conformal symmetry. We construct an infinite tower of quasi-normal modes from the highest-weight mode, in a simple and elegant way. For the scalar, the hidden conformal symmetry manifest itself in the fact that the scalar Laplacian could be rewritten in terms of the $SL(2,R)$ quadratic Casimir. For the vector and the tensor, the hidden conformal symmetry acts on them through Lie derivatives. We show that for three-dimensional black holes, with appropriate combination of the components the radial equations of the vector and the tensor could be written in terms of the Lie-induced quadratic Casimir. This allows the algebraic construction of the quasi-normal modes feasible. Our results are in good agreement with the previous study."
                    }
                ]
            },
            "9b064559-f522-47d6-b5b0-5dfa06efeefa": {
                "pk": "9b064559-f522-47d6-b5b0-5dfa06efeefa",
                "name": "Shu-tao Xia",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            }
        }
    },
    "2404.05595": {
        "paper_data": {
            "title": "UniFL: Improve Stable Diffusion via Unified Feedback Learning",
            "url": "http://arxiv.org/abs/2404.05595v2",
            "arxiv_id": "2404.05595",
            "authors": [
                "Jiacheng Zhang",
                "Jie Wu",
                "Yuxi Ren",
                "Xin Xia",
                "Huafeng Kuang",
                "Pan Xie",
                "Jiashi Li",
                "Xuefeng Xiao",
                "Min Zheng",
                "Lean Fu",
                "Guanbin Li"
            ],
            "abstract": "Diffusion models have revolutionized the field of image generation, leading to the proliferation of high-quality models and diverse downstream applications. However, despite these significant advancements, the current competitive solutions still suffer from several limitations, including inferior visual quality, a lack of aesthetic appeal, and inefficient inference, without a comprehensive solution in sight. To address these challenges, we present UniFL, a unified framework that leverages feedback learning to enhance diffusion models comprehensively. UniFL stands out as a universal, effective, and generalizable solution applicable to various diffusion models, such as SD1.5 and SDXL. Notably, UniFL incorporates three key components: perceptual feedback learning, which enhances visual quality; decoupled feedback learning, which improves aesthetic appeal; and adversarial feedback learning, which optimizes inference speed. In-depth experiments and extensive user studies validate the superior performance of our proposed method in enhancing both the quality of generated models and their acceleration. For instance, UniFL surpasses ImageReward by 17% user preference in terms of generation quality and outperforms LCM and SDXL Turbo by 57% and 20% in 4-step inference. Moreover, we have verified the efficacy of our approach in downstream tasks, including Lora, ControlNet, and AnimateDiff.",
            "introduction": "   1 Introduction  The emergence of diffusion models has catapulted the field of Text-to-Image (T2I) into a realm of unparalleled progress, marked by notable contributions like DALLE-3\u00a0[34], Imagen\u00a0[38], Midjourney\u00a0[49], and etc,. In particular, the introduction of open-source image generation models, exemplified by stable diffusion\u00a0[35], has inaugurated a transformative era of Text-to-Image, giving rise to numerous downstream applications such as T2I Personalization\u00a0[37, 15, 21, 60], Controllable Generation\u00a0[61, 33, 29] and Text-to-Video (T2V) Generation\u00a0[53, 16, 18]. Despite the remarkable progress achieved thus far, current stable diffusion-based image generation models still exhibit certain limitations. i) Inferior Quality: the generated images often suffer from poor quality and lack authenticity. Examples include characters with incomplete limbs or distorted body parts, as well as limited fidelity in terms of style representation. ii) Lack Aesthetics: there is a notable bias in the aesthetic appeal of the generated images, as they often fail to align with human preferences. Deficiencies in crucial aspects such as details, lighting, and atmosphere further contribute to this aesthetic disparity. iii) Inference Inefficiency: the iterative denoising process employed by diffusion models introduces inefficiencies that significantly impede inference speed, thereby limiting the practicality of these models in various application scenarios.   Recently, numerous works have endeavored to address the aforementioned challenges. For instance, SDXL\u00a0[32] enhances the generation quality of diffusion models by refining training strategies, while RAPHAEL\u00a0[59] resorts to the techniques of Mixture of Experts (MoE)\u00a0[44, 63, 14]. RAFT\u00a0[11], HPS\u00a0[55, 54], ImageReward\u00a0[57], and DPO\u00a0[50] propose incorporating human feedback to guide diffusion models toward aligning with human preferences. SDXL Turbo\u00a0[40], PGD\u00a0[39], and LCM\u00a0[27, 28], on the other hand, tackle the issue of inference acceleration through techniques like distillation and consistency models\u00a0[46]. However, these methods primarily concentrate on tackling individual problems through specialized designs, which poses a significant challenge for the direct integration of these techniques. For example, MoE significantly complicates the pipeline, making the acceleration method infeasible, and the consistency models\u00a0[46] alter the denoising process of the diffusion model, making it arduous to directly apply the ReFL framework proposed by ImageReward\u00a0[57]. Naturally, a pertinent question arises: Can we devise a more effective approach that comprehensively enhances diffusion models in terms of image quality, aesthetic appearance, and generation speed?   In this paper, we present UniFL, an innovative approach that offers a comprehensive improvement to diffusion models through unified feedback learning. UniFL aims to elevate the visual generation quality, enhance preference aesthetics, and accelerate the inference process. To achieve these objectives, we propose three key components. Firstly, we introduce a pioneering perceptual feedback learning (PeFL) framework that effectively harnesses the extensive knowledge embedded within diverse existing perceptual models to improve visual generation quality. This framework enables the provision of more precise and targeted feedback signals, ultimately enhancing the quality of visual generation in various aspects.   Secondly, we employ decoupled feedback learning to optimize aesthetic quality. By breaking down the coarse aesthetic concept into distinct aspects such as color, atmosphere, and texture, UniFL simplifies the challenge of aesthetic optimization. Additionally, we introduce an active prompt selection strategy to choose the prompts that are more informative and diverse to facilitate more efficient aesthetics preference feedback learning.   Lastly, UniFL develops adversarial feedback learning, wherein the reward model",
            "references": [
                {
                    "title": "Confronting Reward Overoptimization for Diffusion Models: A Perspective of Inductive and Primacy Biases",
                    "abstract": "Bridging the gap between diffusion models and human preferences is crucial for their integration into practical generative workflows. While optimizing downstream reward models has emerged as a promising alignment strategy, concerns arise regarding the risk of excessive optimization with learned reward models, which potentially compromises ground-truth performance. In this work, we confront the reward overoptimization problem in diffusion model alignment through the lenses of both inductive and primacy biases. We first identify a mismatch between current methods and the temporal inductive bias inherent in the multi-step denoising process of diffusion models, as a potential source of reward overoptimization. Then, we surprisingly discover that dormant neurons in our critic model act as a regularization against reward overoptimization while active neurons reflect primacy bias. Motivated by these observations, we propose Temporal Diffusion Policy Optimization with critic active neuron Reset (TDPO-R), a policy gradient algorithm that exploits the temporal inductive bias of diffusion models and mitigates the primacy bias stemming from active neurons. Empirical results demonstrate the superior efficacy of our methods in mitigating reward overoptimization. Code is avaliable at https://github.com/ZiyiZhang27/tdpo."
                },
                {
                    "title": "Adversarial Diffusion Distillation",
                    "abstract": "We introduce Adversarial Diffusion Distillation (ADD), a novel training approach that efficiently samples large-scale foundational image diffusion models in just 1-4 steps while maintaining high image quality. We use score distillation to leverage large-scale off-the-shelf image diffusion models as a teacher signal in combination with an adversarial loss to ensure high image fidelity even in the low-step regime of one or two sampling steps. Our analyses show that our model clearly outperforms existing few-step methods (GANs, Latent Consistency Models) in a single step and reaches the performance of state-of-the-art diffusion models (SDXL) in only four steps. ADD is the first method to unlock single-step, real-time image synthesis with foundation models. Code and weights available under https://github.com/Stability-AI/generative-models and https://huggingface.co/stabilityai/ ."
                },
                {
                    "title": "Diffusion Model Alignment Using Direct Preference Optimization",
                    "abstract": "Large language models (LLMs) are fine-tuned using human comparison data with Reinforcement Learning from Human Feedback (RLHF) methods to make them better aligned with users' preferences. In contrast to LLMs, human preference learning has not been widely explored in text-to-image diffusion models; the best existing approach is to fine-tune a pretrained model using carefully curated high quality images and captions to improve visual appeal and text alignment. We propose Diffusion-DPO, a method to align diffusion models to human preferences by directly optimizing on human comparison data. Diffusion-DPO is adapted from the recently developed Direct Preference Optimization (DPO) [36], a simpler alternative to RLHF which directly optimizes a policy that best satisfies human preferences under a classification objective. We re-formulate DPO to account for a diffusion model notion of likelihood, utilizing the evidence lower bound to derive a differentiable objective. Using the Pick-a-Pic dataset of 851K crowdsourced pairwise preferences, we fine-tune the base model of the state-of-the-art Stable Diffusion XL (SDXL)-1.0 model with Diffusion-DPO. Our fine-tuned base model significantly outperforms both base SDXL-1.0 and the larger SDXL-1.0 model consisting of an additional refinement model in human evaluation, improving visual appeal and prompt alignment. We also develop a variant that uses AI feedback and has comparable performance to training on human preferences, opening the door for scaling of diffusion model alignment methods."
                },
                {
                    "title": "Emu Video: Factorizing Text-to-Video Generation by Explicit Image Conditioning",
                    "abstract": "We present Emu Video, a text-to-video generation model that factorizes the generation into two steps: first generating an image conditioned on the text, and then generating a video conditioned on the text and the generated image. We identify critical design decisions--adjusted noise schedules for diffusion, and multi-stage training that enable us to directly generate high quality and high resolution videos, without requiring a deep cascade of models as in prior work. In human evaluations, our generated videos are strongly preferred in quality compared to all prior work--81% vs. Google's Imagen Video, 90% vs. Nvidia's PYOCO, and 96% vs. Meta's Make-A-Video. Our model outperforms commercial solutions such as RunwayML's Gen2 and Pika Labs. Finally, our factorizing approach naturally lends itself to animating images based on a user's text prompt, where our generations are preferred 96% over prior work."
                },
                {
                    "title": "UFOGen: You Forward Once Large Scale Text-to-Image Generation via Diffusion GANs",
                    "abstract": "Text-to-image diffusion models have demonstrated re-markable capabilities in transforming text prompts into co-herent images, yet the computational cost of the multi-step inference remains a persistent challenge. To address this issue, we present UFOGen, a novel generative model de-signed for ultra-fast, one-step text-to-image generation. In contrast to conventional approaches that focus on improving samplers or employing distillation techniques for diffusion models, UFOGen adopts a hybrid methodology, inte-grating diffusion models with a GAN objective. Leveraging a newly introduced diffusion-GAN objective and initialization with pre-trained diffusion models, UFOGen excels in efficiently generating high-quality images conditioned on textual descriptions in a single step. Beyond traditional text-to-image generation, UFOGen showcases versatility in applications. Notably, UFOGen stands among the pioneering models enabling one-step text-to-image generation and diverse downstream tasks, presenting a significant advance-ment in the landscape of efficient generative models."
                },
                {
                    "title": "LCM-LoRA: A Universal Stable-Diffusion Acceleration Module",
                    "abstract": "Latent Consistency Models (LCMs) have achieved impressive performance in accelerating text-to-image generative tasks, producing high-quality images with minimal inference steps. LCMs are distilled from pre-trained latent diffusion models (LDMs), requiring only ~32 A100 GPU training hours. This report further extends LCMs' potential in two aspects: First, by applying LoRA distillation to Stable-Diffusion models including SD-V1.5, SSD-1B, and SDXL, we have expanded LCM's scope to larger models with significantly less memory consumption, achieving superior image generation quality. Second, we identify the LoRA parameters obtained through LCM distillation as a universal Stable-Diffusion acceleration module, named LCM-LoRA. LCM-LoRA can be directly plugged into various Stable-Diffusion fine-tuned models or LoRAs without training, thus representing a universally applicable accelerator for diverse image generation tasks. Compared with previous numerical PF-ODE solvers such as DDIM, DPM-Solver, LCM-LoRA can be viewed as a plug-in neural PF-ODE solver that possesses strong generalization abilities. Project page: https://github.com/luosiallen/latent-consistency-model."
                },
                {
                    "title": "Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inference",
                    "abstract": "Latent Diffusion models (LDMs) have achieved remarkable results in synthesizing high-resolution images. However, the iterative sampling process is computationally intensive and leads to slow generation. Inspired by Consistency Models (song et al.), we propose Latent Consistency Models (LCMs), enabling swift inference with minimal steps on any pre-trained LDMs, including Stable Diffusion (rombach et al). Viewing the guided reverse diffusion process as solving an augmented probability flow ODE (PF-ODE), LCMs are designed to directly predict the solution of such ODE in latent space, mitigating the need for numerous iterations and allowing rapid, high-fidelity sampling. Efficiently distilled from pre-trained classifier-free guided diffusion models, a high-quality 768 x 768 2~4-step LCM takes only 32 A100 GPU hours for training. Furthermore, we introduce Latent Consistency Fine-tuning (LCF), a novel method that is tailored for fine-tuning LCMs on customized image datasets. Evaluation on the LAION-5B-Aesthetics dataset demonstrates that LCMs achieve state-of-the-art text-to-image generation performance with few-step inference. Project Page: https://latent-consistency-models.github.io/"
                },
                {
                    "title": "PixArt-\u03b1: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis",
                    "abstract": "The most advanced text-to-image (T2I) models require significant training costs (e.g., millions of GPU hours), seriously hindering the fundamental innovation for the AIGC community while increasing CO2 emissions. This paper introduces PIXART-$\\alpha$, a Transformer-based T2I diffusion model whose image generation quality is competitive with state-of-the-art image generators (e.g., Imagen, SDXL, and even Midjourney), reaching near-commercial application standards. Additionally, it supports high-resolution image synthesis up to 1024px resolution with low training cost, as shown in Figure 1 and 2. To achieve this goal, three core designs are proposed: (1) Training strategy decomposition: We devise three distinct training steps that separately optimize pixel dependency, text-image alignment, and image aesthetic quality; (2) Efficient T2I Transformer: We incorporate cross-attention modules into Diffusion Transformer (DiT) to inject text conditions and streamline the computation-intensive class-condition branch; (3) High-informative data: We emphasize the significance of concept density in text-image pairs and leverage a large Vision-Language model to auto-label dense pseudo-captions to assist text-image alignment learning. As a result, PIXART-$\\alpha$'s training speed markedly surpasses existing large-scale T2I models, e.g., PIXART-$\\alpha$ only takes 10.8% of Stable Diffusion v1.5's training time (675 vs. 6,250 A100 GPU days), saving nearly \\$300,000 (\\$26,000 vs. \\$320,000) and reducing 90% CO2 emissions. Moreover, compared with a larger SOTA model, RAPHAEL, our training cost is merely 1%. Extensive experiments demonstrate that PIXART-$\\alpha$ excels in image quality, artistry, and semantic control. We hope PIXART-$\\alpha$ will provide new insights to the AIGC community and startups to accelerate building their own high-quality yet low-cost generative models from scratch."
                },
                {
                    "title": "IP-Adapter: Text Compatible Image Prompt Adapter for Text-to-Image Diffusion Models",
                    "abstract": "Recent years have witnessed the strong power of large text-to-image diffusion models for the impressive generative capability to create high-fidelity images. However, it is very tricky to generate desired images using only text prompt as it often involves complex prompt engineering. An alternative to text prompt is image prompt, as the saying goes:\"an image is worth a thousand words\". Although existing methods of direct fine-tuning from pretrained models are effective, they require large computing resources and are not compatible with other base models, text prompt, and structural controls. In this paper, we present IP-Adapter, an effective and lightweight adapter to achieve image prompt capability for the pretrained text-to-image diffusion models. The key design of our IP-Adapter is decoupled cross-attention mechanism that separates cross-attention layers for text features and image features. Despite the simplicity of our method, an IP-Adapter with only 22M parameters can achieve comparable or even better performance to a fully fine-tuned image prompt model. As we freeze the pretrained diffusion model, the proposed IP-Adapter can be generalized not only to other custom models fine-tuned from the same base model, but also to controllable generation using existing controllable tools. With the benefit of the decoupled cross-attention strategy, the image prompt can also work well with the text prompt to achieve multimodal image generation. The project page is available at \\url{https://ip-adapter.github.io}."
                },
                {
                    "title": "AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning",
                    "abstract": "With the advance of text-to-image (T2I) diffusion models (e.g., Stable Diffusion) and corresponding personalization techniques such as DreamBooth and LoRA, everyone can manifest their imagination into high-quality images at an affordable cost. However, adding motion dynamics to existing high-quality personalized T2Is and enabling them to generate animations remains an open challenge. In this paper, we present AnimateDiff, a practical framework for animating personalized T2I models without requiring model-specific tuning. At the core of our framework is a plug-and-play motion module that can be trained once and seamlessly integrated into any personalized T2Is originating from the same base T2I. Through our proposed training strategy, the motion module effectively learns transferable motion priors from real-world videos. Once trained, the motion module can be inserted into a personalized T2I model to form a personalized animation generator. We further propose MotionLoRA, a lightweight fine-tuning technique for AnimateDiff that enables a pre-trained motion module to adapt to new motion patterns, such as different shot types, at a low training and data collection cost. We evaluate AnimateDiff and MotionLoRA on several public representative personalized T2I models collected from the community. The results demonstrate that our approaches help these models generate temporally smooth animation clips while preserving the visual quality and motion diversity. Codes and pre-trained weights are available at https://github.com/guoyww/AnimateDiff."
                },
                {
                    "title": "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis",
                    "abstract": "We present SDXL, a latent diffusion model for text-to-image synthesis. Compared to previous versions of Stable Diffusion, SDXL leverages a three times larger UNet backbone: The increase of model parameters is mainly due to more attention blocks and a larger cross-attention context as SDXL uses a second text encoder. We design multiple novel conditioning schemes and train SDXL on multiple aspect ratios. We also introduce a refinement model which is used to improve the visual fidelity of samples generated by SDXL using a post-hoc image-to-image technique. We demonstrate that SDXL shows drastically improved performance compared the previous versions of Stable Diffusion and achieves results competitive with those of black-box state-of-the-art image generators. In the spirit of promoting open research and fostering transparency in large model training and evaluation, we provide access to code and model weights at https://github.com/Stability-AI/generative-models"
                },
                {
                    "title": "JourneyDB: A Benchmark for Generative Image Understanding",
                    "abstract": "While recent advancements in vision-language models have had a transformative impact on multi-modal comprehension, the extent to which these models possess the ability to comprehend generated images remains uncertain. Synthetic images, in comparison to real data, encompass a higher level of diversity in terms of both content and style, thereby presenting significant challenges for the models to fully grasp. In light of this challenge, we introduce a comprehensive dataset, referred to as JourneyDB, that caters to the domain of generative images within the context of multi-modal visual understanding. Our meticulously curated dataset comprises 4 million distinct and high-quality generated images, each paired with the corresponding text prompts that were employed in their creation. Furthermore, we additionally introduce an external subset with results of another 22 text-to-image generative models, which makes JourneyDB a comprehensive benchmark for evaluating the comprehension of generated images. On our dataset, we have devised four benchmarks to assess the performance of generated image comprehension in relation to both content and style interpretation. These benchmarks encompass prompt inversion, style retrieval, image captioning, and visual question answering. Lastly, we evaluate the performance of state-of-the-art multi-modal models when applied to the JourneyDB dataset, providing a comprehensive analysis of their strengths and limitations in comprehending generated content. We anticipate that the proposed dataset and benchmarks will facilitate further research in the field of generative content understanding. The dataset is publicly available at https://journeydb.github.io."
                },
                {
                    "title": "Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis",
                    "abstract": "Recent text-to-image generative models can generate high-fidelity images from text inputs, but the quality of these generated images cannot be accurately evaluated by existing evaluation metrics. To address this issue, we introduce Human Preference Dataset v2 (HPD v2), a large-scale dataset that captures human preferences on images from a wide range of sources. HPD v2 comprises 798,090 human preference choices on 433,760 pairs of images, making it the largest dataset of its kind. The text prompts and images are deliberately collected to eliminate potential bias, which is a common issue in previous datasets. By fine-tuning CLIP on HPD v2, we obtain Human Preference Score v2 (HPS v2), a scoring model that can more accurately predict human preferences on generated images. Our experiments demonstrate that HPS v2 generalizes better than previous metrics across various image distributions and is responsive to algorithmic improvements of text-to-image generative models, making it a preferable evaluation metric for these models. We also investigate the design of the evaluation prompts for text-to-image generative models, to make the evaluation stable, fair and easy-to-use. Finally, we establish a benchmark for text-to-image generative models using HPS v2, which includes a set of recent text-to-image models from the academic, community and industry. The code and dataset is available at https://github.com/tgxs002/HPSv2 ."
                },
                {
                    "title": "Training Diffusion Models with Reinforcement Learning",
                    "abstract": "Diffusion models are a class of flexible generative models trained with an approximation to the log-likelihood objective. However, most use cases of diffusion models are not concerned with likelihoods, but instead with downstream objectives such as human-perceived image quality or drug effectiveness. In this paper, we investigate reinforcement learning methods for directly optimizing diffusion models for such objectives. We describe how posing denoising as a multi-step decision-making problem enables a class of policy gradient algorithms, which we refer to as denoising diffusion policy optimization (DDPO), that are more effective than alternative reward-weighted likelihood approaches. Empirically, DDPO is able to adapt text-to-image diffusion models to objectives that are difficult to express via prompting, such as image compressibility, and those derived from human feedback, such as aesthetic quality. Finally, we show that DDPO can improve prompt-image alignment using feedback from a vision-language model without the need for additional data collection or human annotation. The project's website can be found at http://rl-diffusion.github.io ."
                },
                {
                    "title": "UniControl: A Unified Diffusion Model for Controllable Visual Generation In the Wild",
                    "abstract": "Achieving machine autonomy and human control often represent divergent objectives in the design of interactive AI systems. Visual generative foundation models such as Stable Diffusion show promise in navigating these goals, especially when prompted with arbitrary languages. However, they often fall short in generating images with spatial, structural, or geometric controls. The integration of such controls, which can accommodate various visual conditions in a single unified model, remains an unaddressed challenge. In response, we introduce UniControl, a new generative foundation model that consolidates a wide array of controllable condition-to-image (C2I) tasks within a singular framework, while still allowing for arbitrary language prompts. UniControl enables pixel-level-precise image generation, where visual conditions primarily influence the generated structures and language prompts guide the style and context. To equip UniControl with the capacity to handle diverse visual conditions, we augment pretrained text-to-image diffusion models and introduce a task-aware HyperNet to modulate the diffusion models, enabling the adaptation to different C2I tasks simultaneously. Trained on nine unique C2I tasks, UniControl demonstrates impressive zero-shot generation abilities with unseen visual conditions. Experimental results show that UniControl often surpasses the performance of single-task-controlled methods of comparable model sizes. This control versatility positions UniControl as a significant advancement in the realm of controllable visual generation."
                },
                {
                    "title": "RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment",
                    "abstract": "Generative foundation models are susceptible to implicit biases that can arise from extensive unsupervised training data. Such biases can produce suboptimal samples, skewed outcomes, and unfairness, with potentially serious consequences. Consequently, aligning these models with human ethics and preferences is an essential step toward ensuring their responsible and effective deployment in real-world applications. Prior research has primarily employed Reinforcement Learning from Human Feedback (RLHF) to address this problem, where generative models are fine-tuned with RL algorithms guided by a human-feedback-informed reward model. However, the inefficiencies and instabilities associated with RL algorithms frequently present substantial obstacles to the successful alignment, necessitating the development of a more robust and streamlined approach. To this end, we introduce a new framework, Reward rAnked FineTuning (RAFT), designed to align generative models effectively. Utilizing a reward model and a sufficient number of samples, our approach selects the high-quality samples, discarding those that exhibit undesired behavior, and subsequently enhancing the model by fine-tuning on these filtered samples. Our studies show that RAFT can effectively improve the model performance in both reward learning and other automated metrics in both large language models and diffusion models."
                },
                {
                    "title": "Human Preference Score: Better Aligning Text-to-image Models with Human Preference",
                    "abstract": "Recent years have witnessed a rapid growth of deep generative models, with text-to-image models gaining significant attention from the public. However, existing models often generate images that do not align well with human preferences, such as awkward combinations of limbs and facial expressions. To address this issue, we collect a dataset of human choices on generated images from the Stable Foundation Discord channel. Our experiments demonstrate that current evaluation metrics for generative models do not correlate well with human choices. Thus, we train a human preference classifier with the collected dataset and derive a Human Preference Score (HPS) based on the classifier. Using HPS, we propose a simple yet effective method to adapt Stable Diffusion to better align with human preferences. Our experiments show that HPS outperforms CLIP in predicting human choices and has good generalization capability toward images generated from other models. By tuning Stable Diffusion with the guidance of HPS, the adapted model is able to generate images that are more preferred by human users. The project page is available here: https://tgxs002.github.io/alignsd-web/."
                },
                {
                    "title": "Adding Conditional Control to Text-to-Image Diffusion Models",
                    "abstract": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models."
                },
                {
                    "title": "Tune-A-Video: One-Shot Tuning of Image Diffusion Models for Text-to-Video Generation",
                    "abstract": "To replicate the success of text-to-image (T2I) generation, recent works employ large-scale video datasets to train a text-to-video (T2V) generator. Despite their promising results, such paradigm is computationally expensive. In this work, we propose a new T2V generation setting\u2014One-Shot Video Tuning, where only one text-video pair is presented. Our model is built on state-of-the-art T2I diffusion models pre-trained on massive image data. We make two key observations: 1) T2I models can generate still images that represent verb terms; 2) extending T2I models to generate multiple images concurrently exhibits surprisingly good content consistency. To further learn continuous motion, we introduce Tune-A-Video, which involves a tailored spatio-temporal attention mechanism and an efficient one-shot tuning strategy. At inference, we employ DDIM inversion to provide structure guidance for sampling. Extensive qualitative and numerical experiments demonstrate the remarkable ability of our method across various applications."
                },
                {
                    "title": "InstructPix2Pix: Learning to Follow Image Editing Instructions",
                    "abstract": "We propose a method for editing images from human instructions: given an input image and a written instruction that tells the model what to do, our model follows these instructions to edit the image. To obtain training data for this problem, we combine the knowledge of two large pretrained models\u2014a language model (GPT-3) and a text-to-image model (Stable Diffusion)\u2014to generate a large dataset of image editing examples. Our conditional diffusion model, InstructPix2Pix, is trained on our generated data, and generalizes to real images and user-written instructions at inference time. Since it performs edits in the forward pass and does not require per-example fine-tuning or inversion, our model edits images quickly, in a matter of seconds. We show compelling editing results for a diverse collection of input images and written instructions."
                },
                {
                    "title": "LAION-5B: An open large-scale dataset for training next generation image-text models",
                    "abstract": "Groundbreaking language-vision architectures like CLIP and DALL-E proved the utility of training on large amounts of noisy image-text data, without relying on expensive accurate labels used in standard vision unimodal supervised learning. The resulting models showed capabilities of strong text-guided image generation and transfer to downstream tasks, while performing remarkably at zero-shot classification with noteworthy out-of-distribution robustness. Since then, large-scale language-vision models like ALIGN, BASIC, GLIDE, Flamingo and Imagen made further improvements. Studying the training and capabilities of such models requires datasets containing billions of image-text pairs. Until now, no datasets of this size have been made openly available for the broader research community. To address this problem and democratize research on large-scale multi-modal models, we present LAION-5B - a dataset consisting of 5.85 billion CLIP-filtered image-text pairs, of which 2.32B contain English language. We show successful replication and fine-tuning of foundational models like CLIP, GLIDE and Stable Diffusion using the dataset, and discuss further experiments enabled with an openly available dataset of this scale. Additionally we provide several nearest neighbor indices, an improved web-interface for dataset exploration and subset generation, and detection scores for watermark, NSFW, and toxic content detection. Announcement page https://laion.ai/laion-5b-a-new-era-of-open-large-scale-multi-modal-datasets/"
                },
                {
                    "title": "An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion",
                    "abstract": "Text-to-image models offer unprecedented freedom to guide creation through natural language. Yet, it is unclear how such freedom can be exercised to generate images of specific unique concepts, modify their appearance, or compose them in new roles and novel scenes. In other words, we ask: how can we use language-guided models to turn our cat into a painting, or imagine a new product based on our favorite toy? Here we present a simple approach that allows such creative freedom. Using only 3-5 images of a user-provided concept, like an object or a style, we learn to represent it through new\"words\"in the embedding space of a frozen text-to-image model. These\"words\"can be composed into natural language sentences, guiding personalized creation in an intuitive way. Notably, we find evidence that a single word embedding is sufficient for capturing unique and varied concepts. We compare our approach to a wide range of baselines, and demonstrate that it can more faithfully portray the concepts across a range of applications and tasks. Our code, data and new words will be available at: https://textual-inversion.github.io"
                },
                {
                    "title": "Prompt-to-Prompt Image Editing with Cross Attention Control",
                    "abstract": "Recent large-scale text-driven synthesis models have attracted much attention thanks to their remarkable capabilities of generating highly diverse images that follow given text prompts. Such text-based synthesis methods are particularly appealing to humans who are used to verbally describe their intent. Therefore, it is only natural to extend the text-driven image synthesis to text-driven image editing. Editing is challenging for these generative models, since an innate property of an editing technique is to preserve most of the original image, while in the text-based models, even a small modification of the text prompt often leads to a completely different outcome. State-of-the-art methods mitigate this by requiring the users to provide a spatial mask to localize the edit, hence, ignoring the original structure and content within the masked region. In this paper, we pursue an intuitive prompt-to-prompt editing framework, where the edits are controlled by text only. To this end, we analyze a text-conditioned model in depth and observe that the cross-attention layers are the key to controlling the relation between the spatial layout of the image to each word in the prompt. With this observation, we present several applications which monitor the image synthesis by editing the textual prompt only. This includes localized editing by replacing a word, global editing by adding a specification, and even delicately controlling the extent to which a word is reflected in the image. We present our results over diverse images and prompts, demonstrating high-quality synthesis and fidelity to the edited prompts."
                },
                {
                    "title": "Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding",
                    "abstract": "We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding. Imagen builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation. Our key discovery is that generic large language models (e.g. T5), pretrained on text-only corpora, are surprisingly effective at encoding text for image synthesis: increasing the size of the language model in Imagen boosts both sample fidelity and image-text alignment much more than increasing the size of the image diffusion model. Imagen achieves a new state-of-the-art FID score of 7.27 on the COCO dataset, without ever training on COCO, and human raters find Imagen samples to be on par with the COCO data itself in image-text alignment. To assess text-to-image models in greater depth, we introduce DrawBench, a comprehensive and challenging benchmark for text-to-image models. With DrawBench, we compare Imagen with recent methods including VQ-GAN+CLIP, Latent Diffusion Models, and DALL-E 2, and find that human raters prefer Imagen over other models in side-by-side comparisons, both in terms of sample quality and image-text alignment. See https://imagen.research.google/ for an overview of the results."
                },
                {
                    "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
                    "abstract": "Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Mixture-of-Experts with Expert Choice Routing",
                    "abstract": "Sparsely-activated Mixture-of-experts (MoE) models allow the number of parameters to greatly increase while keeping the amount of computation for a given token or a given sample unchanged. However, a poor expert routing strategy (e.g. one resulting in load imbalance) can cause certain experts to be under-trained, leading to an expert being under or over-specialized. Prior work allocates a fixed number of experts to each token using a top-k function regardless of the relative importance of different tokens. To address this, we propose a heterogeneous mixture-of-experts employing an expert choice method. Instead of letting tokens select the top-k experts, we have experts selecting the top-k tokens. As a result, each token can be routed to a variable number of experts and each expert can have a fixed bucket size. We systematically study pre-training speedups using the same computational resources of the Switch Transformer top-1 and GShard top-2 gating of prior work and find that our method improves training convergence time by more than 2x. For the same computational cost, our method demonstrates higher performance in fine-tuning 11 selected tasks in the GLUE and SuperGLUE benchmarks. For a smaller activation cost, our method outperforms the T5 dense model in 7 out of the 11 tasks."
                },
                {
                    "title": "GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models",
                    "abstract": "Diffusion models have recently been shown to generate high-quality synthetic images, especially when paired with a guidance technique to trade off diversity for fidelity. We explore diffusion models for the problem of text-conditional image synthesis and compare two different guidance strategies: CLIP guidance and classifier-free guidance. We find that the latter is preferred by human evaluators for both photorealism and caption similarity, and often produces photorealistic samples. Samples from a 3.5 billion parameter text-conditional diffusion model using classifier-free guidance are favored by human evaluators to those from DALL-E, even when the latter uses expensive CLIP reranking. Additionally, we find that our models can be fine-tuned to perform image inpainting, enabling powerful text-driven image editing. We train a smaller model on a filtered dataset and release the code and weights at https://github.com/openai/glide-text2im."
                },
                {
                    "title": "Masked-attention Mask Transformer for Universal Image Segmentation",
                    "abstract": "Image segmentation groups pixels with different semantics, e.g., category or instance membership. Each choice of semantics defines a task. While only the semantics of each task differ, current research focuses on designing spe-cialized architectures for each task. We present Masked- attention Mask Transformer (Mask2Former), a new archi-tecture capable of addressing any image segmentation task (panoptic, instance or semantic). Its key components in-clude masked attention, which extracts localized features by constraining cross-attention within predicted mask regions. In addition to reducing the research effort by at least three times, it outperforms the best specialized architectures by a significant margin on four popular datasets. Most no-tably, Mask2Former sets a new state-of-the-art for panoptic segmentation (57.8 PQ on COCO), instance segmentation (50.1 AP on COCO) and semantic segmentation (57.7 mIoU onADE20K)."
                },
                {
                    "title": "SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations",
                    "abstract": "Guided image synthesis enables everyday users to create and edit photo-realistic images with minimum effort. The key challenge is balancing faithfulness to the user input (e.g., hand-drawn colored strokes) and realism of the synthesized image. Existing GAN-based methods attempt to achieve such balance using either conditional GANs or GAN inversions, which are challenging and often require additional training data or loss functions for individual applications. To address these issues, we introduce a new image synthesis and editing method, Stochastic Differential Editing (SDEdit), based on a diffusion model generative prior, which synthesizes realistic images by iteratively denoising through a stochastic differential equation (SDE). Given an input image with user guide of any type, SDEdit first adds noise to the input, then subsequently denoises the resulting image through the SDE prior to increase its realism. SDEdit does not require task-specific training or inversions and can naturally achieve the balance between realism and faithfulness. SDEdit significantly outperforms state-of-the-art GAN-based methods by up to 98.09% on realism and 91.72% on overall satisfaction scores, according to a human perception study, on multiple tasks, including stroke-based image synthesis and editing as well as image compositing."
                },
                {
                    "title": "Reward-Weighted Regression Converges to a Global Optimum",
                    "abstract": "Reward-Weighted Regression (RWR) belongs to a family of widely known iterative Reinforcement Learning algorithms based on the Expectation-Maximization framework. In this family, learning at each iteration consists of sampling a batch of trajectories using the current policy and fitting a new policy to maximize a return-weighted log-likelihood of actions. Although RWR is known to yield monotonic improvement of the policy under certain circumstances, whether and under which conditions RWR converges to the optimal policy have remained open questions. In this paper, we provide for the first time a proof that RWR converges to a global optimum when no function approximation is used, in a general compact setting. Furthermore, for the simpler case with finite state and action spaces we prove R-linear convergence of the state-value function to the optimum."
                },
                {
                    "title": "SOLO: A Simple Framework for Instance Segmentation",
                    "abstract": "Compared to many other dense prediction tasks, e.g., semantic segmentation, it is the arbitrary number of instances that has made instance segmentation much more challenging. In order to predict a mask for each instance, mainstream approaches either follow the \u201cdetect-then-segment\u201d strategy (e.g., Mask R-CNN), or predict embedding vectors first then cluster pixels into individual instances. In this paper, we view the task of instance segmentation from a completely new perspective by introducing the notion of \u201cinstance categories\u201d, which assigns categories to each pixel within an instance according to the instance's location. With this notion, we propose segmenting objects by locations (SOLO), a simple, direct, and fast framework for instance segmentation with strong performance. We derive a few SOLO variants (e.g., Vanilla SOLO, Decoupled SOLO, Dynamic SOLO) following the basic principle. Our method directly maps a raw input image to the desired object categories and instance masks, eliminating the need for the grouping post-processing or the bounding box detection. Our approach achieves state-of-the-art results for instance segmentation in terms of both speed and accuracy, while being considerably simpler than the existing methods. Besides instance segmentation, our method yields state-of-the-art results in object detection (from our mask byproduct) and panoptic segmentation. We further demonstrate the flexibility and high-quality segmentation of SOLO by extending it to perform one-stage instance-level image matting. Code is available at: https://git.io/AdelaiDet."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "CogView: Mastering Text-to-Image Generation via Transformers",
                    "abstract": "Text-to-Image generation in the general domain has long been an open problem, which requires both a powerful generative model and cross-modal understanding. We propose CogView, a 4-billion-parameter Transformer with VQ-VAE tokenizer to advance this problem. We also demonstrate the finetuning strategies for various downstream tasks, e.g. style learning, super-resolution, text-image ranking and fashion design, and methods to stabilize pretraining, e.g. eliminating NaN losses. CogView achieves the state-of-the-art FID on the blurred MS COCO dataset, outperforming previous GAN-based models and a recent similar work DALL-E."
                },
                {
                    "title": "Read Like Humans: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Recognition",
                    "abstract": "Linguistic knowledge is of great benefit to scene text recognition. However, how to effectively model linguistic rules in end-to-end deep networks remains a research challenge. In this paper, we argue that the limited capacity of language models comes from: 1) implicitly language modeling; 2) unidirectional feature representation; and 3) language model with noise input. Correspondingly, we propose an autonomous, bidirectional and iterative ABINet for scene text recognition. Firstly, the autonomous suggests to block gradient flow between vision and language models to enforce explicitly language modeling. Secondly, a novel bidirectional cloze network (BCN) as the language model is proposed based on bidirectional feature representation. Thirdly, we propose an execution manner of iterative correction for language model which can effectively alleviate the impact of noise input. Additionally, based on the ensemble of iterative predictions, we propose a self-training method which can learn from unlabeled images effectively. Extensive experiments indicate that ABINet has superiority on low-quality images and achieves state-of-the-art results on several mainstream benchmarks. Besides, the ABINet trained with ensemble self-training shows promising improvement in realizing human-level recognition. Code is available at https://github.com/FangShancheng/ABINet."
                },
                {
                    "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity",
                    "abstract": "In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with outrageous numbers of parameters -- but a constant computational cost. However, despite several notable successes of MoE, widespread adoption has been hindered by complexity, communication costs and training instability -- we address these with the Switch Transformer. We simplify the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs. Our proposed training techniques help wrangle the instabilities and we show large sparse models may be trained, for the first time, with lower precision (bfloat16) formats. We design models based off T5-Base and T5-Large to obtain up to 7x increases in pre-training speed with the same computational resources. These improvements extend into multilingual settings where we measure gains over the mT5-Base version across all 101 languages. Finally, we advance the current scale of language models by pre-training up to trillion parameter models on the\"Colossal Clean Crawled Corpus\"and achieve a 4x speedup over the T5-XXL model."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Real-time Scene Text Detection with Differentiable Binarization",
                    "abstract": "Recently, segmentation-based methods are quite popular in scene text detection, as the segmentation results can more accurately describe scene text of various shapes such as curve text. However, the post-processing of binarization is essential for segmentation-based detection, which converts probability maps produced by a segmentation method into bounding boxes/regions of text. In this paper, we propose a module named Differentiable Binarization (DB), which can perform the binarization process in a segmentation network. Optimized along with a DB module, a segmentation network can adaptively set the thresholds for binarization, which not only simplifies the post-processing but also enhances the performance of text detection. Based on a simple segmentation network, we validate the performance improvements of DB on five benchmark datasets, which consistently achieves state-of-the-art results, in terms of both detection accuracy and speed. In particular, with a light-weight backbone, the performance improvements by DB are significant so that we can look for an ideal tradeoff between detection accuracy and efficiency. Specifically, with a backbone of ResNet-18, our detector achieves an F-measure of 82.8, running at 62 FPS, on the MSRA-TD500 dataset. Code is available at: https://github.com/MhLiao/DB."
                },
                {
                    "title": "GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium",
                    "abstract": "Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the \"Frechet Inception Distance\" (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark."
                },
                {
                    "title": "Rethinking Atrous Convolution for Semantic Image Segmentation",
                    "abstract": "In this work, we revisit atrous convolution, a powerful tool to explicitly adjust filter's field-of-view as well as control the resolution of feature responses computed by Deep Convolutional Neural Networks, in the application of semantic image segmentation. To handle the problem of segmenting objects at multiple scales, we design modules which employ atrous convolution in cascade or in parallel to capture multi-scale context by adopting multiple atrous rates. Furthermore, we propose to augment our previously proposed Atrous Spatial Pyramid Pooling module, which probes convolutional features at multiple scales, with image-level features encoding global context and further boost performance. We also elaborate on implementation details and share our experience on training our system. The proposed `DeepLabv3' system significantly improves over our previous DeepLab versions without DenseCRF post-processing and attains comparable performance with other state-of-art models on the PASCAL VOC 2012 semantic image segmentation benchmark."
                },
                {
                    "title": "Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer",
                    "abstract": "The capacity of a neural network to absorb information is limited by its number of parameters. Conditional computation, where parts of the network are active on a per-example basis, has been proposed in theory as a way of dramatically increasing model capacity without a proportional increase in computation. In practice, however, there are significant algorithmic and performance challenges. In this work, we address these challenges and finally realize the promise of conditional computation, achieving greater than 1000x improvements in model capacity with only minor losses in computational efficiency on modern GPU clusters. We introduce a Sparsely-Gated Mixture-of-Experts layer (MoE), consisting of up to thousands of feed-forward sub-networks. A trainable gating network determines a sparse combination of these experts to use for each example. We apply the MoE to the tasks of language modeling and machine translation, where model capacity is critical for absorbing the vast quantities of knowledge available in the training corpora. We present model architectures in which a MoE with up to 137 billion parameters is applied convolutionally between stacked LSTM layers. On large language modeling and machine translation benchmarks, these models achieve significantly better results than state-of-the-art at lower computational cost."
                },
                {
                    "title": "COCO-Stuff: Thing and Stuff Classes in Context",
                    "abstract": "Semantic classes can be either things (objects with a well-defined shape, e.g. car, person) or stuff (amorphous background regions, e.g. grass, sky). While lots of classification and detection works focus on thing classes, less attention has been given to stuff classes. Nonetheless, stuff classes are important as they allow to explain important aspects of an image, including (1) scene type; (2) which thing classes are likely to be present and their location (through contextual reasoning); (3) physical attributes, material types and geometric properties of the scene. To understand stuff and things in context we introduce COCO-Stuff1, which augments all 164K images of the COCO 2017 dataset with pixel-wise annotations for 91 stuff classes. We introduce an efficient stuff annotation protocol based on superpixels, which leverages the original thing annotations. We quantify the speed versus quality trade-off of our protocol and explore the relation between annotation time and boundary complexity. Furthermore, we use COCO-Stuff to analyze: (a) the importance of stuff and thing classes in terms of their surface cover and how frequently they are mentioned in image captions; (b) the spatial relations between stuff and things, highlighting the rich contextual relations that make our dataset unique; (c) the performance of a modern semantic segmentation method on stuff and thing classes, and whether stuff is easier to segment than things."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "A Computational Approach to Edge Detection",
                    "abstract": "This paper describes a computational approach to edge detection. The success of the approach depends on the definition of a comprehensive set of goals for the computation of edge points. These goals must be precise enough to delimit the desired behavior of the detector while making minimal assumptions about the form of the solution. We define detection and localization criteria for a class of edges, and present mathematical forms for these criteria as functionals on the operator impulse response. A third criterion is then added to ensure that the detector has only one response to a single edge. We use the criteria in numerical optimization to derive detectors for several common image features, including step edges. On specializing the analysis to step edges, we find that there is a natural uncertainty principle between detection and localization performance, which are the two main goals. With this principle we derive a single operator shape which is optimal at any scale. The optimal detector has a simple approximate implementation in which edges are marked at maxima in gradient magnitude of a Gaussian-smoothed image. We extend this simple detector using operators of several widths to cope with different signal-to-noise ratios in the image. We present a general method, called feature synthesis, for the fine-to-coarse integration of information from operators at different scales. Finally we show that step edge detector performance improves considerably as the operator point spread function is extended along the edge."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan we devise a more effective approach that comprehensively enhances diffusion models in terms of image quality, aesthetic appearance, and generation speed?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of Text-to-Image (T2I) generation, as it addresses significant limitations in current diffusion models, such as poor image quality, lack of aesthetic appeal, and slow inference speeds. By improving these aspects, the research community can unlock new applications and enhance existing ones, leading to more realistic and visually appealing image generation. This could pave the way for practical applications in various domains, including art, design, and entertainment, ultimately influencing future research directions and methodologies in generative models.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexities of diffusion models, which involve iterative denoising processes that can lead to inefficiencies and suboptimal outputs. Naive approaches may fail because they often address individual issues in isolation, without considering the interdependencies between image quality, aesthetics, and inference speed. Additionally, the integration of various techniques, such as Mixture of Experts and consistency models, complicates the pipeline and can hinder performance. Overcoming these technical and theoretical obstacles requires a holistic approach that effectively balances these competing demands.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on isolated aspects of diffusion models, such as improving generation quality or accelerating inference, without a unified framework to address all challenges simultaneously. Limitations in existing solutions often arise from their specialized designs, which complicate integration and application. Barriers such as the complexity of the feedback mechanisms and the lack of comprehensive methodologies have prevented a holistic solution. Our approach, UniFL, differs by proposing a unified feedback learning framework that simultaneously enhances visual quality, aesthetic appeal, and inference speed, addressing the limitations of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, UniFL, consists of three key components:  \n1. **Perceptual Feedback Learning (PeFL)**: This framework leverages knowledge from existing perceptual models to provide precise feedback signals, enhancing visual generation quality.  \n2. **Decoupled Feedback Learning**: This component breaks down aesthetic optimization into distinct aspects (color, atmosphere, texture) and employs an active prompt selection strategy to facilitate efficient learning"
            }
        },
        "author_data": {
            "9a3b22f1-5eaa-4830-ba8f-8633f9370ede": {
                "pk": "9a3b22f1-5eaa-4830-ba8f-8633f9370ede",
                "name": "Jiacheng Zhang",
                "collaborators": [
                    "Daniel Lacker",
                    "Xin Guo",
                    "Mykhaylo Shkolnikov",
                    "Thibaut Mastrolia",
                    "Changxiao Cai",
                    "Wenyi Yan",
                    "Ye Zhang",
                    "Anran Hu",
                    "Lalit Rajendran",
                    "Sally Bane"
                ],
                "domain": [
                    "Stochastic Processes",
                    "Mean Field Games",
                    "Machine Learning",
                    "Control Theory"
                ],
                "publications": [
                    {
                        "title": "Ito's formula for flows of conditional measures on semimartingales",
                        "abstract": "Motivated by recent development of mean-field systems with common noise, this paper establishes Ito's formula for flows of conditional probability measures under a common filtration associated with general semimartingales. This generalizes existing works on flows of conditional measures on Ito processes and flows of deterministic measure on general semimartingales. The key technical components involve constructing conditional independent copies and establishing the equivalence between stochastic integrals with respect to the conditional law of semimartingales and the conditional expectation of stochastic integrals with respect to copies of semimartingales. Ito's formula is then established for cylindrical functions through conditional independent copies, and extended to the general case through function approximations."
                    },
                    {
                        "title": "Stationary solutions and local equations for interacting diffusions on regular trees",
                        "abstract": "We study the invariant measures of infinite systems of stochastic differential equations (SDEs) indexed by the vertices of a regular tree. These invariant measures correspond to Gibbs measures associated with certain continuous specifications, and we focus specifically on those measures which are homogeneous Markov random fields. We characterize the joint law at any two adjacent vertices in terms of a new two-dimensional SDE system, called the \"local equation\", which exhibits an unusual dependence on a conditional law. Exploiting an alternative characterization in terms of an eigenfunction-type fixed point problem, we derive existence and uniqueness results for invariant measures of the local equation and infinite SDE system. This machinery is put to use in two examples. First, we give a detailed analysis of the surprisingly subtle case of linear coefficients, which yields a new way to derive the famous Kesten-McKay law for the spectral measure of the regular tree. Second, we construct solutions of tree-indexed SDE systems with nearest-neighbor repulsion effects, similar to Dyson's Brownian motion."
                    },
                    {
                        "title": "Agency problem and mean field system of agents with moral hazard, synergistic effects and accidents",
                        "abstract": "We investigate the existence of an optimal policy to monitor a mean field systems of agents managing a risky project under moral hazard with accidents modeled by L\\'evy processes magnified by the law of the project. We provide a general method to find both a mean field equilibrium for the agents and the optimal compensation policy under general, sufficient and necessary assumptions on all the parameters. We formalize the problem as a bilevel optimization with the probabilistic version of a mean field games which can be reduced to a controlled McKean-Vlasov SDE with jumps. We apply our results to an optimal energy demand-response problem with a crowd of consumers subjected to power cut/shortage when the variability of the energy consumption is too high under endogenous or exogenous strains. In this example, we get explicit solution to the mean field game and to the McKean-Vlasov equation with jumps"
                    },
                    {
                        "title": "Minimax-optimal trust-aware multi-armed bandits",
                        "abstract": "Multi-armed bandit (MAB) algorithms have achieved significant success in sequential decision-making applications, under the premise that humans perfectly implement the recommended policy. However, existing methods often overlook the crucial factor of human trust in learning algorithms. When trust is lacking, humans may deviate from the recommended policy, leading to undesired learning performance. Motivated by this gap, we study the trust-aware MAB problem by integrating a dynamic trust model into the standard MAB framework. Specifically, it assumes that the recommended and actually implemented policy differs depending on human trust, which in turn evolves with the quality of the recommended policy. We establish the minimax regret in the presence of the trust issue and demonstrate the suboptimality of vanilla MAB algorithms such as the upper confidence bound (UCB) algorithm. To overcome this limitation, we introduce a novel two-stage trust-aware procedure that provably attains near-optimal statistical guarantees. A simulation study is conducted to illustrate the benefits of our proposed algorithm when dealing with the trust issue."
                    },
                    {
                        "title": "A new Speech Feature Fusion method with cross gate parallel CNN for Speaker Recognition",
                        "abstract": "In this paper, a new speech feature fusion method is proposed for speaker recognition on the basis of the cross gate parallel convolutional neural network (CG-PCNN). The Mel filter bank features (MFBFs) of different frequency resolutions can be extracted from each speech frame of a speaker's speech by several Mel filter banks, where the numbers of the triangular filters in the Mel filter banks are different. Due to the frequency resolutions of these MFBFs are different, there are some complementaries for these MFBFs. The CG-PCNN is utilized to extract the deep features from these MFBFs, which applies a cross gate mechanism to capture the complementaries for improving the performance of the speaker recognition system. Then, the fusion feature can be obtained by concatenating these deep features for speaker recognition. The experimental results show that the speaker recognition system with the proposed speech feature fusion method is effective, and marginally outperforms the existing state-of-the-art systems."
                    },
                    {
                        "title": "Optimization frameworks and sensitivity analysis of Stackelberg mean-field games",
                        "abstract": "This paper proposes and studies a class of discrete-time finite-time-horizon Stackelberg mean-field games, with one leader and an infinite number of identical and indistinguishable followers. In this game, the objective of the leader is to maximize her reward considering the worst-case cost over all possible $\\epsilon$-Nash equilibria among followers. A new analytical paradigm is established by showing the equivalence between this Stackelberg mean-field game and a minimax optimization problem. This optimization framework facilitates studying both analytically and numerically the set of Nash equilibria for the game; and leads to the sensitivity and the robustness analysis of the game value. In particular, when there is model uncertainty, the game value for the leader suffers non-vanishing sub-optimality as the perturbed model converges to the true model. In order to obtain a near-optimal solution, the leader needs to be more pessimistic with anticipation of model errors and adopts a relaxed version of the original Stackelberg game."
                    },
                    {
                        "title": "Inverting the Markovian projection, with an application to local stochastic volatility models",
                        "abstract": "We study two-dimensional stochastic differential equations (SDEs) of McKean--Vlasov type in which the conditional distribution of the second component of the solution given the first enters the equation for the first component of the solution. Such SDEs arise when one tries to invert the Markovian projection developed by Gy\\\"ongy (1986), typically to produce an It\\^o process with the fixed-time marginal distributions of a given one-dimensional diffusion but richer dynamical features. We prove the strong existence of stationary solutions for these SDEs, as well as their strong uniqueness in an important special case. Variants of the SDEs discussed in this paper enjoy frequent application in the calibration of local stochastic volatility models in finance, despite the very limited theoretical understanding."
                    },
                    {
                        "title": "Superposition and mimicking theorems for conditional McKean-Vlasov equations",
                        "abstract": "We consider conditional McKean-Vlasov stochastic differential equations (SDEs), such as the ones arising in the large-system limit of mean field games and particle systems with mean field interactions when common noise is present. The conditional time-marginals of the solutions to these SDEs satisfy non-linear stochastic partial differential equations (SPDEs) of the second order, whereas the laws of the conditional time-marginals follow Fokker-Planck equations on the space of probability measures. We prove two superposition principles: The first establishes that any solution of the SPDE can be lifted to a solution of the conditional McKean-Vlasov SDE, and the second guarantees that any solution of the Fokker-Planck equation on the space of probability measures can be lifted to a solution of the SPDE. We use these results to obtain a mimicking theorem which shows that the conditional time-marginals of an Ito process can be emulated by those of a solution to a conditional McKean-Vlasov SDE with Markovian coefficients. This yields, in particular, a tool for converting open-loop controls into Markovian ones in the context of controlled McKean-Vlasov dynamics."
                    },
                    {
                        "title": "Weighted Least Squares (WLS) Density Integration for Background Oriented Schlieren (BOS)",
                        "abstract": "We propose an improved density integration methodology for Background Oriented Schlieren (BOS) measurements that overcomes the noise sensitivity of the commonly used Poisson solver. The method employs a weighted least-squares (WLS) optimization of the 2D integration of the density gradient field by solving an over-determined system of equations. Weights are assigned to the grid points based on density gradient uncertainties to ensure that a less reliable measurement point has less effect on the integration procedure. Synthetic image analysis with a Gaussian density field shows that WLS constrains the propagation of random error and reduces it by 80% in comparison to Poisson for the highest noise level. Using WLS with experimental BOS measurements of flow induced by a spark plasma discharge show a 30% reduction in density uncertainty in comparison to Poisson, thereby increasing the overall precision of the BOS density measurements."
                    },
                    {
                        "title": "Deep Learning for Population-Dependent Controls in Mean Field Control Problems with Common Noise",
                        "abstract": "In this paper, we propose several approaches to learn the optimal population-dependent controls in order to solve mean field control problems (MFC). Such policies enable us to solve MFC problems with forms of common noises at a level of generality that was not covered by existing methods. We analyze rigorously the theoretical convergence of the proposed approximation algorithms. Of particular interest for its simplicity of implementation is the $N$-particle approximation. The effectiveness and the flexibility of our algorithms is supported by numerical experiments comparing several combinations of distribution approximation techniques and neural network architectures. We use three different benchmark problems from the literature: a systemic risk model, a price impact model, and a crowd motion model. We first show that our proposed algorithms converge to the correct solution in an explicitly solvable MFC problem. Then, we show that population-dependent controls outperform state-dependent controls. Along the way, we show that specific neural network architectures can improve the learning further."
                    },
                    {
                        "title": "Generating Visual Scenes from Touch",
                        "abstract": "An emerging line of work has sought to generate plausible imagery from touch. Existing approaches, however, tackle only narrow aspects of the visuo-tactile synthesis problem, and lag significantly behind the quality of cross-modal synthesis methods in other domains. We draw on recent advances in latent diffusion to create a model for synthesizing images from tactile signals (and vice versa) and apply it to a number of visuo-tactile synthesis tasks. Using this model, we significantly outperform prior work on the tactile-driven stylization problem, i.e., manipulating an image to match a touch signal, and we are the first to successfully generate images from touch without additional sources of information about the scene. We also successfully use our model to address two novel synthesis problems: generating images that do not contain the touch sensor or the hand holding it, and estimating an image's shading from its reflectance and touch."
                    }
                ]
            },
            "f92a03ee-cf74-4a9a-90ec-d193c0057bef": {
                "pk": "f92a03ee-cf74-4a9a-90ec-d193c0057bef",
                "name": "Jie Wu",
                "collaborators": [
                    "Liqun Zhang",
                    "Qiang Wu",
                    "Hongze Li",
                    "Jie Qing",
                    "Yuguang Shi",
                    "Roman Mikhailov"
                ],
                "domain": [
                    "Homotopy Theory",
                    "Algebraic Topology",
                    "Differential Geometry",
                    "Graph Theory"
                ],
                "publications": [
                    {
                        "title": "A braided simplicial group",
                        "abstract": "By studying braid group actions on Milnor's construction of the 1-sphere, we show that the general higher homotopy group of the 3-sphere is the fixed set of the pure braid group action on certain combinatorially described group. We also give certain representation of higher differentials in the Adams spectral sequence for the homotopy groups of the 2-sphere."
                    },
                    {
                        "title": "The Functor $A^{\\min}$ for $(p-1)$-cell Complexes and $\\EHP$ Sequences",
                        "abstract": "Let $X$ be a co-$H$-space of $(p-1)$-cell complex with all cells in even dimensions. Then the loop space $\\Omega X$ admits a retract $\\bar A^{\\min}(X)$ that is the evaluation of the functor $\\bar A^{\\min}$ on $X$. In this paper, we determine the homology $H_*(\\bar A^{\\min}(X))$ and give the $\\EHP$ sequence for the spaces $\\bar A^{\\min}(X)$."
                    },
                    {
                        "title": "A note on new weighted geometric inequalities for hypersurfaces in $\\mathbb R^n$",
                        "abstract": "In this note, we prove a family of sharp weighed inequality which involves weighted $k$-th mean curvature integral and two distinct quermassintegrals for closed hypersurfaces in $\\mathbb{R}^n$. This inequality generalizes the corresponding result of Wei and Zhou \\cite{WZ} where their proof is based on earlier results of Kwong-Miao \\cite{KM1,KM2}. Here we present a proof which does not rely on Kwong-Miao's results."
                    },
                    {
                        "title": "On Combinatorial Descriptions of Homotopy Groups of $\u03a3K(\u03c0,1)$",
                        "abstract": "We give a combinatorial description of homotopy groups of $\\Sigma K(\\pi,1)$. In particular, all of the homotopy groups of the $3$-sphere are combinatorially given."
                    },
                    {
                        "title": "On Combinatorial Calculations for the James--Hopf maps",
                        "abstract": "We give some formulas of the James-Hopf maps by using combinatorial methods. An application is to give a product decomposition of the spaces $\\Omega\\Sigma^2(X)$."
                    },
                    {
                        "title": "A Product Decomposition of $\u03a9^3_0\u03a3${\\bf R}$P^2$",
                        "abstract": "We give a specific product decomposition of the base-point path connected component of the triple loop space of the suspension of the projective plane."
                    },
                    {
                        "title": "A new characterization of geodesic spheres in the Hyperbolic space",
                        "abstract": "This paper gives a new characterization of geodesic spheres in the hyperbolic space in terms of a ``weighted'' higher order mean curvature. Precisely, we show that a compact hypersurface $\\Sigma^{n-1}$ embedded in $\\H^n$ with $VH_k$ being constant for some $k=1,\\cdots,n-1$ is a centered geodesic sphere. Here $H_k$ is the $k$-th normalized mean curvature of $\\Sigma$ induced from $\\H^n$ and $V=\\cosh r$, where $r$ is a hyperbolic distance to a fixed point in $\\H^n$. Moreover, this result can be generalized to a compact hypersurface $\\Sigma$ embedded in $\\H^n$ with the ratio $V\\left(\\frac{H_k}{H_j}\\right)\\equiv\\mbox{constant},\\;0\\leq j< k\\leq n-1$ and $H_j$ not vanishing on $\\Sigma$."
                    },
                    {
                        "title": "On the Homology of Configuration Spaces $C((M,M_o)\\times {\\bold R}^n; X)$",
                        "abstract": "The homology with coefficients in a field of the configuration spaces $C(M\\times \\bold R ^n,M_o\\times \\bold R ^n;X)$ is determined in this paper."
                    },
                    {
                        "title": "On Brunnian-type links and the link invariants given by homotopy groups of spheres",
                        "abstract": "We introduce the (general) homotopy groups of spheres as link invariants for Brunnian-type links through the investigations on the intersection subgroup of the normal closures of the meridians of strongly nonsplittable links. The homotopy groups measure the difference between the intersection subgroup and symmetric commutator subgroup of the normal closures of the meridians and give the invariants of the links obtained in this way. Moreover the higher homotopy-group invariants can produce some links that could not be detected by the Milnor invariants. Furthermore all homotopy groups of spheres can be obtained from the geometric Massey products on links."
                    },
                    {
                        "title": "A sufficient condition for the existence of fractional $(g,f,n)$-critical covered graphs",
                        "abstract": "In data transmission networks, the availability of data transmission is equivalent to the existence of the fractional factor of the corresponding graph which is generated by the network. Research on the existence of fractional factors under specific network structures can help scientists design and construct networks with high data transmission rates. A graph $G$ is called a fractional $(g,f)$-covered graph if for any $e\\in E(G)$, $G$ admits a fractional $(g,f)$-factor covering $e$. A graph $G$ is called a fractional $(g,f,n)$-critical covered graph if after removing any $n$ vertices of $G$, the resulting graph of $G$ is a fractional $(g,f)$-covered graph. In this paper, we verify that if a graph $G$ of order $p$ satisfies $p\\geq\\frac{(a+b-1)(a+b-2)+(a+d)n+1}{a+d}$, $\\delta(G)\\geq\\frac{(b-d-1)p+(a+d)n+a+b+1}{a+b-1}$ and $\\delta(G)>\\frac{(b-d-2)p+2\\alpha(G)+(a+d)n+1}{a+b-2}$, then $G$ is a fractional $(g,f,n)$-critical covered graph, where $g,f:V(G)\\rightarrow Z^{+}$ be two functions such that $a\\leq g(x)\\leq f(x)-d\\leq b-d$ for all $x\\in V(G)$, which is a generalization of Zhou's previous result [S. Zhou, Some new sufficient conditions for graphs to have fractional $k$-factors, International Journal of Computer Mathematics 88(3)(2011)484--490]."
                    },
                    {
                        "title": "Chen's double sieve, Goldbach's conjecture and the twin prime problem",
                        "abstract": "We give a more comrehensive treatment of Chen's double sieve and improve related constants in Goldbach's conjecture and the twin prime problem."
                    },
                    {
                        "title": "Chen's double sieve, Goldbach's conjecture and the twin prime problem, 2",
                        "abstract": "For every even integer N, denote by D_{1,2}(N) the number of representations of N as a sum of a prime and an integer having at most two prime factors. In this paper, we give a new lower bound for D_{1,2}(N)."
                    },
                    {
                        "title": "Power sums of Hecke's eigenvalues and application",
                        "abstract": "We sharpen some estimates of Rankin on power sums of Hecke eigenvalues, by using Kim & Shahidi's recent results on higher order symmetric powers. As an application, we improve Kohnen, Lau & Shparlinski's lower bound for the number of Hecke eigenvalues of same signs."
                    },
                    {
                        "title": "The average exponent of elliptic curves modulo $p$",
                        "abstract": "Let $E$ be an elliptic curve defined over ${\\mathbb Q}$. For a prime $p$ of good reduction for $E$, denote by $e_p$ the exponent of the reduction of $E$ modulo $p$. Under GRH, we prove that there is a constant $C_E\\in (0, 1)$ such that $$ \\frac{1}{\\pi(x)} \\sum_{p\\le x} e_p = 1/2 C_E x + O_E\\big(x^{5/6} (\\log x)^{4/3}\\big) $$ for all $x\\ge 2$, where the implied constant depends on $E$ at most. When $E$ has complex multiplication, the same asymptotic formula with a weaker error term $O_E(1/(\\log x)^{1/14})$ is established unconditionally. These improve some recent results of Freiberg and Kurlberg."
                    },
                    {
                        "title": "Mean Values for a Class of Arithmetic Functions in Short Intervals",
                        "abstract": "In this paper, we shall establish a rather general asymptotic formula in short intervals for a classe of arithmetic functions and announce two applications about the distribution of divisors of square-full numbers and integers representable as sums of two squares."
                    },
                    {
                        "title": "The universality of symmetric power L-functions and their Rankin-Selberg L-functions",
                        "abstract": "We establish the universality theorem for the first four symmetric power L-functions of automorphic forms and their associated Rankin-Selberg L-functions. This generalizes some results of Laurincikas & Matsumoto and Matsumoto respectively."
                    },
                    {
                        "title": "Normalized Ricci flows and conformally compact Einstein metrics",
                        "abstract": "In this paper, we investigate the behavior of the normalized Ricci flow on asymptotically hyperbolic manifolds. We show that the normalized Ricci flow exists globally and converges to an Einstein metric when starting from a non-degenerate and sufficiently Ricci pinched metric. More importantly we use maximum principles to establish the regularity of conformal compactness along the normalized Ricci flow including that of the limit metric at time infinity. Therefore we are able to recover the existence results in \\cite{GL} \\cite{Lee} \\cite{Bi} of conformally compact Einstein metrics with conformal infinities which are perturbations of that of given non-degenerate conformally compact Einstein metrics."
                    },
                    {
                        "title": "A combinatorial description of homotopy groups of spheres",
                        "abstract": "We give a combinatorial description of general homotopy groups of $k$-dimensional spheres with $k\\geq3$ as well as those of Moore spaces.   For $n>k\\geq 3,$ we construct a finitely generated group defined by explicit generators and relations, whose center is exactly $\\pi_n(S^k)$."
                    },
                    {
                        "title": "Backward uniqueness for general parabolic operators in the whole space",
                        "abstract": "We prove the backward uniqueness for general parabolic operators of second order in the whole space under assumptions that the leading coefficients of the operator are Lipschitz and their gradients satisfy certain decay conditions. This result extends in some ways a classical result of Lions and Malgrange [12] and a recent result of the authors [10]."
                    },
                    {
                        "title": "Landis-Oleinik Conjecture in the Exterior Domain",
                        "abstract": "In 1974, Landis and Oleinik conjectured that if a bounded solution of a parabolic equation decays fast at a time, then the solution must vanish identically before that time, provided the coefficients of the equation satisfy appropriate conditions at infinity. We prove this conjecture under some reasonable assumptions on the coefficients which improved the earlier results."
                    }
                ]
            },
            "11f36ab5-5bc4-46c4-934e-5e2ace68fb6c": {
                "pk": "11f36ab5-5bc4-46c4-934e-5e2ace68fb6c",
                "name": "Yuxi Ren",
                "collaborators": [
                    "Jie Wu",
                    "Xuefeng Xiao",
                    "Rui Wang",
                    "Min Zheng",
                    "Xin Pan",
                    "Jie Qin",
                    "Huixia Li",
                    "Pan Xie",
                    "Xin Xia",
                    "Jianchao Yang"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Image Generation",
                    "Machine Learning",
                    "Distillation Techniques"
                ],
                "publications": [
                    {
                        "title": "Online Multi-Granularity Distillation for GAN Compression",
                        "abstract": "Generative Adversarial Networks (GANs) have witnessed prevailing success in yielding outstanding images, however, they are burdensome to deploy on resource-constrained devices due to ponderous computational costs and hulking memory usage. Although recent efforts on compressing GANs have acquired remarkable results, they still exist potential model redundancies and can be further compressed. To solve this issue, we propose a novel online multi-granularity distillation (OMGD) scheme to obtain lightweight GANs, which contributes to generating high-fidelity images with low computational demands. We offer the first attempt to popularize single-stage online distillation for GAN-oriented compression, where the progressively promoted teacher generator helps to refine the discriminator-free based student generator. Complementary teacher generators and network layers provide comprehensive and multi-granularity concepts to enhance visual fidelity from diverse dimensions. Experimental results on four benchmark datasets demonstrate that OMGD successes to compress 40x MACs and 82.5X parameters on Pix2Pix and CycleGAN, without loss of image quality. It reveals that OMGD provides a feasible solution for the deployment of real-time image translation on resource-constrained devices. Our code and models are made public at: https://github.com/bytedance/OMGD."
                    },
                    {
                        "title": "UGC: Unified GAN Compression for Efficient Image-to-Image Translation",
                        "abstract": "Recent years have witnessed the prevailing progress of Generative Adversarial Networks (GANs) in image-to-image translation. However, the success of these GAN models hinges on ponderous computational costs and labor-expensive training data. Current efficient GAN learning techniques often fall into two orthogonal aspects: i) model slimming via reduced calculation costs; ii)data/label-efficient learning with fewer training data/labels. To combine the best of both worlds, we propose a new learning paradigm, Unified GAN Compression (UGC), with a unified optimization objective to seamlessly prompt the synergy of model-efficient and label-efficient learning. UGC sets up semi-supervised-driven network architecture search and adaptive online semi-supervised distillation stages sequentially, which formulates a heterogeneous mutual learning scheme to obtain an architecture-flexible, label-efficient, and performance-excellent model."
                    },
                    {
                        "title": "Parallel Pre-trained Transformers (PPT) for Synthetic Data-based Instance Segmentation",
                        "abstract": "Recently, Synthetic data-based Instance Segmentation has become an exceedingly favorable optimization paradigm since it leverages simulation rendering and physics to generate high-quality image-annotation pairs. In this paper, we propose a Parallel Pre-trained Transformers (PPT) framework to accomplish the synthetic data-based Instance Segmentation task. Specifically, we leverage the off-the-shelf pre-trained vision Transformers to alleviate the gap between natural and synthetic data, which helps to provide good generalization in the downstream synthetic data scene with few samples. Swin-B-based CBNet V2, SwinL-based CBNet V2 and Swin-L-based Uniformer are employed for parallel feature learning, and the results of these three models are fused by pixel-level Non-maximum Suppression (NMS) algorithm to obtain more robust results. The experimental results reveal that PPT ranks first in the CVPR2022 AVA Accessibility Vision and Autonomy Challenge, with a 65.155% mAP."
                    },
                    {
                        "title": "DiffusionGPT: LLM-Driven Text-to-Image Generation System",
                        "abstract": "Diffusion models have opened up new avenues for the field of image generation, resulting in the proliferation of high-quality models shared on open-source platforms. However, a major challenge persists in current text-to-image systems are often unable to handle diverse inputs, or are limited to single model results. Current unified attempts often fall into two orthogonal aspects: i) parse Diverse Prompts in input stage; ii) activate expert model to output. To combine the best of both worlds, we propose DiffusionGPT, which leverages Large Language Models (LLM) to offer a unified generation system capable of seamlessly accommodating various types of prompts and integrating domain-expert models. DiffusionGPT constructs domain-specific Trees for various generative models based on prior knowledge. When provided with an input, the LLM parses the prompt and employs the Trees-of-Thought to guide the selection of an appropriate model, thereby relaxing input constraints and ensuring exceptional performance across diverse domains. Moreover, we introduce Advantage Databases, where the Tree-of-Thought is enriched with human feedback, aligning the model selection process with human preferences. Through extensive experiments and comparisons, we demonstrate the effectiveness of DiffusionGPT, showcasing its potential for pushing the boundaries of image synthesis in diverse domains."
                    },
                    {
                        "title": "ResAdapter: Domain Consistent Resolution Adapter for Diffusion Models",
                        "abstract": "Recent advancement in text-to-image models (e.g., Stable Diffusion) and corresponding personalized technologies (e.g., DreamBooth and LoRA) enables individuals to generate high-quality and imaginative images. However, they often suffer from limitations when generating images with resolutions outside of their trained domain. To overcome this limitation, we present the Resolution Adapter (ResAdapter), a domain-consistent adapter designed for diffusion models to generate images with unrestricted resolutions and aspect ratios. Unlike other multi-resolution generation methods that process images of static resolution with complex post-process operations, ResAdapter directly generates images with the dynamical resolution. Especially, after learning a deep understanding of pure resolution priors, ResAdapter trained on the general dataset, generates resolution-free images with personalized diffusion models while preserving their original style domain. Comprehensive experiments demonstrate that ResAdapter with only 0.5M can process images with flexible resolutions for arbitrary diffusion models. More extended experiments demonstrate that ResAdapter is compatible with other modules (e.g., ControlNet, IP-Adapter and LCM-LoRA) for image generation across a broad range of resolutions, and can be integrated into other multi-resolution model (e.g., ElasticDiffusion) for efficiently generating higher-resolution images. Project link is https://res-adapter.github.io"
                    },
                    {
                        "title": "Hyper-SD: Trajectory Segmented Consistency Model for Efficient Image Synthesis",
                        "abstract": "Recently, a series of diffusion-aware distillation algorithms have emerged to alleviate the computational overhead associated with the multi-step inference process of Diffusion Models (DMs). Current distillation techniques often dichotomize into two distinct aspects: i) ODE Trajectory Preservation; and ii) ODE Trajectory Reformulation. However, these approaches suffer from severe performance degradation or domain shifts. To address these limitations, we propose Hyper-SD, a novel framework that synergistically amalgamates the advantages of ODE Trajectory Preservation and Reformulation, while maintaining near-lossless performance during step compression. Firstly, we introduce Trajectory Segmented Consistency Distillation to progressively perform consistent distillation within pre-defined time-step segments, which facilitates the preservation of the original ODE trajectory from a higher-order perspective. Secondly, we incorporate human feedback learning to boost the performance of the model in a low-step regime and mitigate the performance loss incurred by the distillation process. Thirdly, we integrate score distillation to further improve the low-step generation capability of the model and offer the first attempt to leverage a unified LoRA to support the inference process at all steps. Extensive experiments and user studies demonstrate that Hyper-SD achieves SOTA performance from 1 to 8 inference steps for both SDXL and SD1.5. For example, Hyper-SDXL surpasses SDXL-Lightning by +0.68 in CLIP Score and +0.51 in Aes Score in the 1-step inference."
                    }
                ]
            },
            "7627b996-f890-4eba-bf8d-dc13970a1c4f": {
                "pk": "7627b996-f890-4eba-bf8d-dc13970a1c4f",
                "name": "Xin Xia",
                "collaborators": [
                    "Xuefeng Xiao",
                    "Xing Wang",
                    "Min Zheng",
                    "Yansen Su",
                    "Chunhou Zheng",
                    "Xiangxiang Zeng",
                    "Chao Ni",
                    "Xin Yin",
                    "Kaiwen Yang",
                    "Dehai Zhao"
                ],
                "domain": [
                    "Neural Architecture Search",
                    "Machine Learning",
                    "Vulnerability Detection",
                    "Multi-objective Optimization"
                ],
                "publications": [
                    {
                        "title": "Progressive Automatic Design of Search Space for One-Shot Neural Architecture Search",
                        "abstract": "Neural Architecture Search (NAS) has attracted growing interest. To reduce the search cost, recent work has explored weight sharing across models and made major progress in One-Shot NAS. However, it has been observed that a model with higher one-shot model accuracy does not necessarily perform better when stand-alone trained. To address this issue, in this paper, we propose Progressive Automatic Design of search space, named PAD-NAS. Unlike previous approaches where the same operation search space is shared by all the layers in the supernet, we formulate a progressive search strategy based on operation pruning and build a layer-wise operation search space. In this way, PAD-NAS can automatically design the operations for each layer and achieve a trade-off between search space quality and model diversity. During the search, we also take the hardware platform constraints into consideration for efficient neural network model deployment. Extensive experiments on ImageNet show that our method can achieve state-of-the-art performance."
                    },
                    {
                        "title": "Molecule optimization via multi-objective evolutionary in implicit chemical space",
                        "abstract": "Machine learning methods have been used to accelerate the molecule optimization process. However, efficient search for optimized molecules satisfying several properties with scarce labeled data remains a challenge for machine learning molecule optimization. In this study, we propose MOMO, a multi-objective molecule optimization framework to address the challenge by combining learning of chemical knowledge with Pareto-based multi-objective evolutionary search. To learn chemistry, it employs a self-supervised codec to construct an implicit chemical space and acquire the continues representation of molecules. To explore the established chemical space, MOMO uses multi-objective evolution to comprehensively and efficiently search for similar molecules with multiple desirable properties. We demonstrate the high performance of MOMO on four multi-objective property and similarity optimization tasks, and illustrate the search capability of MOMO through case studies. Remarkably, our approach significantly outperforms previous approaches in optimizing three objectives simultaneously. The results show the optimization capability of MOMO, suggesting to improve the success rate of lead molecule optimization."
                    },
                    {
                        "title": "Distinguishing Look-Alike Innocent and Vulnerable Code by Subtle Semantic Representation Learning and Explanation",
                        "abstract": "Though many deep learning (DL)-based vulnerability detection approaches have been proposed and indeed achieved remarkable performance, they still have limitations in the generalization as well as the practical usage. More precisely, existing DL-based approaches (1) perform negatively on prediction tasks among functions that are lexically similar but have contrary semantics; (2) provide no intuitive developer-oriented explanations to the detected results. In this paper, we propose a novel approach named SVulD, a function-level Subtle semantic embedding for Vulnerability Detection along with intuitive explanations, to alleviate the above limitations. Specifically, SVulD firstly trains a model to learn distinguishing semantic representations of functions regardless of their lexical similarity. Then, for the detected vulnerable functions, SVulD provides natural language explanations (e.g., root cause) of results to help developers intuitively understand the vulnerabilities. To evaluate the effectiveness of SVulD, we conduct large-scale experiments on a widely used practical vulnerability dataset and compare it with four state-of-the-art (SOTA) approaches by considering five performance measures. The experimental results indicate that SVulD outperforms all SOTAs with a substantial improvement (i.e., 23.5%-68.0% in terms of F1-score, 15.9%-134.8% in terms of PR-AUC and 7.4%-64.4% in terms of Accuracy). Besides, we conduct a user-case study to evaluate the usefulness of SVulD for developers on understanding the vulnerable code and the participants' feedback demonstrates that SVulD is helpful for development practice."
                    }
                ]
            },
            "4f7e8c59-aa76-452d-935f-28ec4133790b": {
                "pk": "4f7e8c59-aa76-452d-935f-28ec4133790b",
                "name": "Huafeng Kuang",
                "collaborators": [
                    "Rongrong Ji",
                    "Xingbin Liu",
                    "Xianming Lin",
                    "Yongjian Wu",
                    "Xiaoshuai Sun",
                    "Hong Liu",
                    "Ming Li",
                    "Taojiannan Yang",
                    "Jie Wu",
                    "Zhaoning Wang"
                ],
                "domain": [
                    "Adversarial Training",
                    "Generative Models",
                    "Computer Vision",
                    "Robustness"
                ],
                "publications": [
                    {
                        "title": "CAT:Collaborative Adversarial Training",
                        "abstract": "Adversarial training can improve the robustness of neural networks. Previous methods focus on a single adversarial training strategy and do not consider the model property trained by different strategies. By revisiting the previous methods, we find different adversarial training methods have distinct robustness for sample instances. For example, a sample instance can be correctly classified by a model trained using standard adversarial training (AT) but not by a model trained using TRADES, and vice versa. Based on this observation, we propose a collaborative adversarial training framework to improve the robustness of neural networks. Specifically, we use different adversarial training methods to train robust models and let models interact with their knowledge during the training process. Collaborative Adversarial Training (CAT) can improve both robustness and accuracy. Extensive experiments on various networks and datasets validate the effectiveness of our method. CAT achieves state-of-the-art adversarial robustness without using any additional data on CIFAR-10 under the Auto-Attack benchmark. Code is available at https://github.com/liuxingbin/CAT."
                    },
                    {
                        "title": "Latent Feature Relation Consistency for Adversarial Robustness",
                        "abstract": "Deep neural networks have been applied in many computer vision tasks and achieved state-of-the-art performance. However, misclassification will occur when DNN predicts adversarial examples which add human-imperceptible adversarial noise to natural examples. This limits the application of DNN in security-critical fields. To alleviate this problem, we first conducted an empirical analysis of the latent features of both adversarial and natural examples and found the similarity matrix of natural examples is more compact than those of adversarial examples. Motivated by this observation, we propose \\textbf{L}atent \\textbf{F}eature \\textbf{R}elation \\textbf{C}onsistency (\\textbf{LFRC}), which constrains the relation of adversarial examples in latent space to be consistent with the natural examples. Importantly, our LFRC is orthogonal to the previous method and can be easily combined with them to achieve further improvement. To demonstrate the effectiveness of LFRC, we conduct extensive experiments using different neural networks on benchmark datasets. For instance, LFRC can bring 0.78\\% further improvement compared to AT, and 1.09\\% improvement compared to TRADES, against AutoAttack on CIFAR10. Code is available at https://github.com/liuxingbin/LFRC."
                    },
                    {
                        "title": "ControlNet++: Improving Conditional Controls with Efficient Consistency Feedback",
                        "abstract": "To enhance the controllability of text-to-image diffusion models, existing efforts like ControlNet incorporated image-based conditional controls. In this paper, we reveal that existing methods still face significant challenges in generating images that align with the image conditional controls. To this end, we propose ControlNet++, a novel approach that improves controllable generation by explicitly optimizing pixel-level cycle consistency between generated images and conditional controls. Specifically, for an input conditional control, we use a pre-trained discriminative reward model to extract the corresponding condition of the generated images, and then optimize the consistency loss between the input conditional control and extracted condition. A straightforward implementation would be generating images from random noises and then calculating the consistency loss, but such an approach requires storing gradients for multiple sampling timesteps, leading to considerable time and memory costs. To address this, we introduce an efficient reward strategy that deliberately disturbs the input images by adding noise, and then uses the single-step denoised images for reward fine-tuning. This avoids the extensive costs associated with image sampling, allowing for more efficient reward fine-tuning. Extensive experiments show that ControlNet++ significantly improves controllability under various conditional controls. For example, it achieves improvements over ControlNet by 11.1% mIoU, 13.4% SSIM, and 7.6% RMSE, respectively, for segmentation mask, line-art edge, and depth conditions. All the code, models, demo and organized data have been open sourced on our Github Repo."
                    },
                    {
                        "title": "StealthDiffusion: Towards Evading Diffusion Forensic Detection through Diffusion Model",
                        "abstract": "The rapid progress in generative models has given rise to the critical task of AI-Generated Content Stealth (AIGC-S), which aims to create AI-generated images that can evade both forensic detectors and human inspection. This task is crucial for understanding the vulnerabilities of existing detection methods and developing more robust techniques. However, current adversarial attacks often introduce visible noise, have poor transferability, and fail to address spectral differences between AI-generated and genuine images. To address this, we propose StealthDiffusion, a framework based on stable diffusion that modifies AI-generated images into high-quality, imperceptible adversarial examples capable of evading state-of-the-art forensic detectors. StealthDiffusion comprises two main components: Latent Adversarial Optimization, which generates adversarial perturbations in the latent space of stable diffusion, and Control-VAE, a module that reduces spectral differences between the generated adversarial images and genuine images without affecting the original diffusion model's generation process. Extensive experiments show that StealthDiffusion is effective in both white-box and black-box settings, transforming AI-generated images into high-quality adversarial forgeries with frequency spectra similar to genuine images. These forgeries are classified as genuine by advanced forensic classifiers and are difficult for humans to distinguish."
                    },
                    {
                        "title": "TraDiffusion: Trajectory-Based Training-Free Image Generation",
                        "abstract": "In this work, we propose a training-free, trajectory-based controllable T2I approach, termed TraDiffusion. This novel method allows users to effortlessly guide image generation via mouse trajectories. To achieve precise control, we design a distance awareness energy function to effectively guide latent variables, ensuring that the focus of generation is within the areas defined by the trajectory. The energy function encompasses a control function to draw the generation closer to the specified trajectory and a movement function to diminish activity in areas distant from the trajectory. Through extensive experiments and qualitative assessments on the COCO dataset, the results reveal that TraDiffusion facilitates simpler, more natural image control. Moreover, it showcases the ability to manipulate salient regions, attributes, and relationships within the generated images, alongside visual input based on arbitrary or enhanced trajectories."
                    }
                ]
            },
            "2b1db83a-8d18-42bb-b462-5ea4da8f2583": {
                "pk": "2b1db83a-8d18-42bb-b462-5ea4da8f2583",
                "name": "Pan Xie",
                "collaborators": [
                    "Xiaohui Hu",
                    "Mengyi Zhao",
                    "Zexian Li",
                    "Taiyi Peng",
                    "Yao Du",
                    "Qipeng Zhang",
                    "Hengnian Gu",
                    "Zhiyi Duan",
                    "Dongdai Zhou"
                ],
                "domain": [
                    "Sign Language Recognition",
                    "Machine Translation",
                    "Graph Neural Network",
                    "Contrastive Learning"
                ],
                "publications": [
                    {
                        "title": "PiSLTRc: Position-informed Sign Language Transformer with Content-aware Convolution",
                        "abstract": "Since the superiority of Transformer in learning long-term dependency, the sign language Transformer model achieves remarkable progress in Sign Language Recognition (SLR) and Translation (SLT). However, there are several issues with the Transformer that prevent it from better sign language understanding. The first issue is that the self-attention mechanism learns sign video representation in a frame-wise manner, neglecting the temporal semantic structure of sign gestures. Secondly, the attention mechanism with absolute position encoding is direction and distance unaware, thus limiting its ability. To address these issues, we propose a new model architecture, namely PiSLTRc, with two distinctive characteristics: (i) content-aware and position-aware convolution layers. Specifically, we explicitly select relevant features using a novel content-aware neighborhood gathering method. Then we aggregate these features with position-informed temporal convolution layers, thus generating robust neighborhood-enhanced sign representation. (ii) injecting the relative position information to the attention mechanism in the encoder, decoder, and even encoder-decoder cross attention. Compared with the vanilla Transformer model, our model performs consistently better on three large-scale sign language benchmarks: PHOENIX-2014, PHOENIX-2014-T and CSL. Furthermore, extensive experiments demonstrate that the proposed method achieves state-of-the-art performance on translation quality with $+1.6$ BLEU improvements."
                    },
                    {
                        "title": "MvSR-NAT: Multi-view Subset Regularization for Non-Autoregressive Machine Translation",
                        "abstract": "Conditional masked language models (CMLM) have shown impressive progress in non-autoregressive machine translation (NAT). They learn the conditional translation model by predicting the random masked subset in the target sentence. Based on the CMLM framework, we introduce Multi-view Subset Regularization (MvSR), a novel regularization method to improve the performance of the NAT model. Specifically, MvSR consists of two parts: (1) \\textit{shared mask consistency}: we forward the same target with different mask strategies, and encourage the predictions of shared mask positions to be consistent with each other. (2) \\textit{model consistency}, we maintain an exponential moving average of the model weights, and enforce the predictions to be consistent between the average model and the online model. Without changing the CMLM-based architecture, our approach achieves remarkable performance on three public benchmarks with 0.36-1.14 BLEU gains over previous NAT models. Moreover, compared with the stronger Transformer baseline, we reduce the gap to 0.01-0.44 BLEU scores on small datasets (WMT16 RO$\\leftrightarrow$EN and IWSLT DE$\\rightarrow$EN)."
                    },
                    {
                        "title": "Sign Language Production with Latent Motion Transformer",
                        "abstract": "Sign Language Production (SLP) is the tough task of turning sign language into sign videos. The main goal of SLP is to create these videos using a sign gloss. In this research, we've developed a new method to make high-quality sign videos without using human poses as a middle step. Our model works in two main parts: first, it learns from a generator and the video's hidden features, and next, it uses another model to understand the order of these hidden features. To make this method even better for sign videos, we make several significant improvements. (i) In the first stage, we take an improved 3D VQ-GAN to learn downsampled latent representations. (ii) In the second stage, we introduce sequence-to-sequence attention to better leverage conditional information. (iii) The separated two-stage training discards the realistic visual semantic of the latent codes in the second stage. To endow the latent sequences semantic information, we extend the token-level autoregressive latent codes learning with perceptual loss and reconstruction loss for the prior model with visual perception. Compared with previous state-of-the-art approaches, our model performs consistently better on two word-level sign language datasets, i.e., WLASL and NMFs-CSL."
                    },
                    {
                        "title": "Modeling Balanced Explicit and Implicit Relations with Contrastive Learning for Knowledge Concept Recommendation in MOOCs",
                        "abstract": "The knowledge concept recommendation in Massive Open Online Courses (MOOCs) is a significant issue that has garnered widespread attention. Existing methods primarily rely on the explicit relations between users and knowledge concepts on the MOOC platforms for recommendation. However, there are numerous implicit relations (e.g., shared interests or same knowledge levels between users) generated within the users' learning activities on the MOOC platforms. Existing methods fail to consider these implicit relations, and these relations themselves are difficult to learn and represent, causing poor performance in knowledge concept recommendation and an inability to meet users' personalized needs. To address this issue, we propose a novel framework based on contrastive learning, which can represent and balance the explicit and implicit relations for knowledge concept recommendation in MOOCs (CL-KCRec). Specifically, we first construct a MOOCs heterogeneous information network (HIN) by modeling the data from the MOOC platforms. Then, we utilize a relation-updated graph convolutional network and stacked multi-channel graph neural network to represent the explicit and implicit relations in the HIN, respectively. Considering that the quantity of explicit relations is relatively fewer compared to implicit relations in MOOCs, we propose a contrastive learning with prototypical graph to enhance the representations of both relations to capture their fruitful inherent relational knowledge, which can guide the propagation of students' preferences within the HIN. Based on these enhanced representations, to ensure the balanced contribution of both towards the final recommendation, we propose a dual-head attention mechanism for balanced fusion. Experimental results demonstrate that CL-KCRec outperforms several state-of-the-art baselines on real-world datasets in terms of HR, NDCG and MRR."
                    }
                ]
            },
            "4d6e7a9e-ba6d-4d2f-b441-9967a6d48d5c": {
                "pk": "4d6e7a9e-ba6d-4d2f-b441-9967a6d48d5c",
                "name": "Jiashi Li",
                "collaborators": [
                    "Shanchuan Lin",
                    "Bingchen Liu",
                    "Xiao Yang"
                ],
                "domain": [
                    "Diffusion Models",
                    "Image Generation",
                    "Machine Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Common Diffusion Noise Schedules and Sample Steps are Flawed",
                        "abstract": "We discover that common diffusion noise schedules do not enforce the last timestep to have zero signal-to-noise ratio (SNR), and some implementations of diffusion samplers do not start from the last timestep. Such designs are flawed and do not reflect the fact that the model is given pure Gaussian noise at inference, creating a discrepancy between training and inference. We show that the flawed design causes real problems in existing implementations. In Stable Diffusion, it severely limits the model to only generate images with medium brightness and prevents it from generating very bright and dark samples. We propose a few simple fixes: (1) rescale the noise schedule to enforce zero terminal SNR; (2) train the model with v prediction; (3) change the sampler to always start from the last timestep; (4) rescale classifier-free guidance to prevent over-exposure. These simple changes ensure the diffusion process is congruent between training and inference and allow the model to generate samples more faithful to the original data distribution."
                    }
                ]
            },
            "01f2183e-8cea-4d80-94a6-1ea6c3e5a66a": {
                "pk": "01f2183e-8cea-4d80-94a6-1ea6c3e5a66a",
                "name": "Xuefeng Xiao",
                "collaborators": [
                    "Xin Xia",
                    "Xing Wang",
                    "Min Zheng",
                    "Yuxi Ren",
                    "Jie Wu",
                    "Jianchao Yang"
                ],
                "domain": [
                    "Neural Architecture Search",
                    "Generative Adversarial Networks",
                    "Model Compression",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Progressive Automatic Design of Search Space for One-Shot Neural Architecture Search",
                        "abstract": "Neural Architecture Search (NAS) has attracted growing interest. To reduce the search cost, recent work has explored weight sharing across models and made major progress in One-Shot NAS. However, it has been observed that a model with higher one-shot model accuracy does not necessarily perform better when stand-alone trained. To address this issue, in this paper, we propose Progressive Automatic Design of search space, named PAD-NAS. Unlike previous approaches where the same operation search space is shared by all the layers in the supernet, we formulate a progressive search strategy based on operation pruning and build a layer-wise operation search space. In this way, PAD-NAS can automatically design the operations for each layer and achieve a trade-off between search space quality and model diversity. During the search, we also take the hardware platform constraints into consideration for efficient neural network model deployment. Extensive experiments on ImageNet show that our method can achieve state-of-the-art performance."
                    },
                    {
                        "title": "Online Multi-Granularity Distillation for GAN Compression",
                        "abstract": "Generative Adversarial Networks (GANs) have witnessed prevailing success in yielding outstanding images, however, they are burdensome to deploy on resource-constrained devices due to ponderous computational costs and hulking memory usage. Although recent efforts on compressing GANs have acquired remarkable results, they still exist potential model redundancies and can be further compressed. To solve this issue, we propose a novel online multi-granularity distillation (OMGD) scheme to obtain lightweight GANs, which contributes to generating high-fidelity images with low computational demands. We offer the first attempt to popularize single-stage online distillation for GAN-oriented compression, where the progressively promoted teacher generator helps to refine the discriminator-free based student generator. Complementary teacher generators and network layers provide comprehensive and multi-granularity concepts to enhance visual fidelity from diverse dimensions. Experimental results on four benchmark datasets demonstrate that OMGD successes to compress 40x MACs and 82.5X parameters on Pix2Pix and CycleGAN, without loss of image quality. It reveals that OMGD provides a feasible solution for the deployment of real-time image translation on resource-constrained devices. Our code and models are made public at: https://github.com/bytedance/OMGD."
                    }
                ]
            },
            "d83bb12e-7e1c-49ab-9175-95a48657dc98": {
                "pk": "d83bb12e-7e1c-49ab-9175-95a48657dc98",
                "name": "Min Zheng",
                "collaborators": [
                    "Chong Tan",
                    "Chen Mao",
                    "Jingqi Hu",
                    "Xuemei Chen",
                    "Yang Chu",
                    "Jichen Bian",
                    "Peiyao Tang",
                    "Xin Xia",
                    "Xuefeng Xiao",
                    "Xing Wang"
                ],
                "domain": [
                    "Signal Processing",
                    "Wireless Sensing",
                    "Neural Architecture Search",
                    "Person Re-identification"
                ],
                "publications": [
                    {
                        "title": "Identify Equivalent Frames",
                        "abstract": "A frame is an overcomplete set that can represent vectors(signals) faithfully and stably. Two frames are equivalent if signals can be essentially represented in the same way, which means two frames differ by a permutation, sign change or orthogonal transformation. Since these operations are combinatorial in nature, it is infeasible to check whether two frames are equivalent by exhaustive search. In this note, we present an algorithm that can check this equivalence in polynomial time. Theoretical guarantees are provided for special cases."
                    },
                    {
                        "title": "Wi-Fi-based Personnel Identity Recognition: Addressing Dataset Imbalance with C-DDPMs",
                        "abstract": "Wireless sensing technologies become increasingly prevalent due to the ubiquitous nature of wireless signals and their inherent privacy-friendly characteristics. Device-free personnel identity recognition, a prevalent application in wireless sensing, is susceptibly challenged by imbalanced channel state information (CSI) datasets. This letter proposes a novel method for CSI dataset augmentation that employs Conditional Denoising Diffusion Probabilistic Models (C-DDPMs) to generate additional samples that address class imbalance issues. The augmentation markedly improves classification accuracies on our homemade dataset, elevating all classes to above 94%."
                    },
                    {
                        "title": "Progressive Automatic Design of Search Space for One-Shot Neural Architecture Search",
                        "abstract": "Neural Architecture Search (NAS) has attracted growing interest. To reduce the search cost, recent work has explored weight sharing across models and made major progress in One-Shot NAS. However, it has been observed that a model with higher one-shot model accuracy does not necessarily perform better when stand-alone trained. To address this issue, in this paper, we propose Progressive Automatic Design of search space, named PAD-NAS. Unlike previous approaches where the same operation search space is shared by all the layers in the supernet, we formulate a progressive search strategy based on operation pruning and build a layer-wise operation search space. In this way, PAD-NAS can automatically design the operations for each layer and achieve a trade-off between search space quality and model diversity. During the search, we also take the hardware platform constraints into consideration for efficient neural network model deployment. Extensive experiments on ImageNet show that our method can achieve state-of-the-art performance."
                    },
                    {
                        "title": "Time-Frequency Analysis of Variable-Length WiFi CSI Signals for Person Re-Identification",
                        "abstract": "Person re-identification (ReID), as a crucial technology in the field of security, plays an important role in security detection and people counting. Current security and monitoring systems largely rely on visual information, which may infringe on personal privacy and be susceptible to interference from pedestrian appearances and clothing in certain scenarios. Meanwhile, the widespread use of routers offers new possibilities for ReID. This letter introduces a method using WiFi Channel State Information (CSI), leveraging the multipath propagation characteristics of WiFi signals as a basis for distinguishing different pedestrian features. We propose a two-stream network structure capable of processing variable-length data, which analyzes the amplitude in the time domain and the phase in the frequency domain of WiFi signals, fuses time-frequency information through continuous lateral connections, and employs advanced objective functions for representation and metric learning. Tested on a dataset collected in the real world, our method achieves 93.68% mAP and 98.13% Rank-1."
                    },
                    {
                        "title": "ViFi-ReID: A Two-Stream Vision-WiFi Multimodal Approach for Person Re-identification",
                        "abstract": "Person re-identification(ReID), as a crucial technology in the field of security, plays a vital role in safety inspections, personnel counting, and more. Most current ReID approaches primarily extract features from images, which are easily affected by objective conditions such as clothing changes and occlusions. In addition to cameras, we leverage widely available routers as sensing devices by capturing gait information from pedestrians through the Channel State Information (CSI) in WiFi signals and contribute a multimodal dataset. We employ a two-stream network to separately process video understanding and signal analysis tasks, and conduct multi-modal fusion and contrastive learning on pedestrian video and WiFi data. Extensive experiments in real-world scenarios demonstrate that our method effectively uncovers the correlations between heterogeneous data, bridges the gap between visual and signal modalities, significantly expands the sensing range, and improves ReID accuracy across multiple sensors."
                    }
                ]
            },
            "82f2efad-29b4-4563-9f2b-a65b33f02474": {
                "pk": "82f2efad-29b4-4563-9f2b-a65b33f02474",
                "name": "Lean Fu",
                "collaborators": [
                    "Songwei Liu",
                    "Hang Dong",
                    "Fei Wang",
                    "Fangmin Chen",
                    "Kai Chen",
                    "Jinshan Pan",
                    "Ding Liu",
                    "Chenqian Yan",
                    "Jingwen He",
                    "Chao Zhu"
                ],
                "domain": [
                    "Image Super Resolution",
                    "Video Processing",
                    "Deep Learning",
                    "Model Compression"
                ],
                "publications": [
                    {
                        "title": "GCFSR: a Generative and Controllable Face Super Resolution Method Without Facial and GAN Priors",
                        "abstract": "Face image super resolution (face hallucination) usually relies on facial priors to restore realistic details and preserve identity information. Recent advances can achieve impressive results with the help of GAN prior. They either design complicated modules to modify the fixed GAN prior or adopt complex training strategies to finetune the generator. In this work, we propose a generative and controllable face SR framework, called GCFSR, which can reconstruct images with faithful identity information without any additional priors. Generally, GCFSR has an encoder-generator architecture. Two modules called style modulation and feature modulation are designed for the multi-factor SR task. The style modulation aims to generate realistic face details and the feature modulation dynamically fuses the multi-level encoded features and the generated ones conditioned on the upscaling factor. The simple and elegant architecture can be trained from scratch in an end-to-end manner. For small upscaling factors (<=8), GCFSR can produce surprisingly good results with only adversarial loss. After adding L1 and perceptual losses, GCFSR can outperform state-of-the-art methods for large upscaling factors (16, 32, 64). During the test phase, we can modulate the generative strength via feature modulation by changing the conditional upscaling factor continuously to achieve various generative effects."
                    },
                    {
                        "title": "Deep Recurrent Neural Network with Multi-scale Bi-directional Propagation for Video Deblurring",
                        "abstract": "The success of the state-of-the-art video deblurring methods stems mainly from implicit or explicit estimation of alignment among the adjacent frames for latent video restoration. However, due to the influence of the blur effect, estimating the alignment information from the blurry adjacent frames is not a trivial task. Inaccurate estimations will interfere the following frame restoration. Instead of estimating alignment information, we propose a simple and effective deep Recurrent Neural Network with Multi-scale Bi-directional Propagation (RNN-MBP) to effectively propagate and gather the information from unaligned neighboring frames for better video deblurring. Specifically, we build a Multi-scale Bi-directional Propagation~(MBP) module with two U-Net RNN cells which can directly exploit the inter-frame information from unaligned neighboring hidden states by integrating them in different scales. Moreover, to better evaluate the proposed algorithm and existing state-of-the-art methods on real-world blurry scenes, we also create a Real-World Blurry Video Dataset (RBVD) by a well-designed Digital Video Acquisition System (DVAS) and use it as the training and evaluation dataset. Extensive experimental results demonstrate that the proposed RBVD dataset effectively improves the performance of existing algorithms on real-world blurry videos, and the proposed algorithm performs favorably against the state-of-the-art methods on three typical benchmarks. The code is available at https://github.com/XJTU-CVLAB-LOWLEVEL/RNN-MBP."
                    },
                    {
                        "title": "Fast and Memory-Efficient Network Towards Efficient Image Super-Resolution",
                        "abstract": "Runtime and memory consumption are two important aspects for efficient image super-resolution (EISR) models to be deployed on resource-constrained devices. Recent advances in EISR exploit distillation and aggregation strategies with plenty of channel split and concatenation operations to make full use of limited hierarchical features. In contrast, sequential network operations avoid frequently accessing preceding states and extra nodes, and thus are beneficial to reducing the memory consumption and runtime overhead. Following this idea, we design our lightweight network backbone by mainly stacking multiple highly optimized convolution and activation layers and decreasing the usage of feature fusion. We propose a novel sequential attention branch, where every pixel is assigned an important factor according to local and global contexts, to enhance high-frequency details. In addition, we tailor the residual block for EISR and propose an enhanced residual block (ERB) to further accelerate the network inference. Finally, combining all the above techniques, we construct a fast and memory-efficient network (FMEN) and its small version FMEN-S, which runs 33% faster and reduces 74% memory consumption compared with the state-of-the-art EISR model: E-RFDN, the champion in AIM 2020 efficient super-resolution challenge. Besides, FMEN-S achieves the lowest memory consumption and the second shortest runtime in NTIRE 2022 challenge on efficient super-resolution. Code is available at https://github.com/NJU-Jet/FMEN."
                    },
                    {
                        "title": "FoldGPT: Simple and Effective Large Language Model Compression Scheme",
                        "abstract": "The demand for deploying large language models(LLMs) on mobile devices continues to increase, driven by escalating data security concerns and cloud costs. However, network bandwidth and memory limitations pose challenges for deploying billion-level models on mobile devices. In this study, we investigate the outputs of different layers across various scales of LLMs and found that the outputs of most layers exhibit significant similarity. Moreover, this similarity becomes more pronounced as the model size increases, indicating substantial redundancy in the depth direction of the LLMs. Based on this observation, we propose an efficient model volume compression strategy, termed FoldGPT, which combines block removal and block parameter sharing.This strategy consists of three parts: (1) Based on the learnable gating parameters, we determine the block importance ranking while modeling the coupling effect between blocks. Then we delete some redundant layers based on the given removal rate. (2) For the retained blocks, we apply a specially designed group parameter sharing strategy, where blocks within the same group share identical weights, significantly compressing the number of parameters and slightly reducing latency overhead. (3) After sharing these Blocks, we \"cure\" the mismatch caused by sparsity with a minor amount of fine-tuning and introduce a tail-layer distillation strategy to improve the performance. Experiments demonstrate that FoldGPT outperforms previous state-of-the-art(SOTA) methods in efficient model compression, demonstrating the feasibility of achieving model lightweighting through straightforward block removal and parameter sharing."
                    },
                    {
                        "title": "Hybrid SD: Edge-Cloud Collaborative Inference for Stable Diffusion Models",
                        "abstract": "Stable Diffusion Models (SDMs) have shown remarkable proficiency in image synthesis. However, their broad application is impeded by their large model sizes and intensive computational requirements, which typically require expensive cloud servers for deployment. On the flip side, while there are many compact models tailored for edge devices that can reduce these demands, they often compromise on semantic integrity and visual quality when compared to full-sized SDMs. To bridge this gap, we introduce Hybrid SD, an innovative, training-free SDMs inference framework designed for edge-cloud collaborative inference. Hybrid SD distributes the early steps of the diffusion process to the large models deployed on cloud servers, enhancing semantic planning. Furthermore, small efficient models deployed on edge devices can be integrated for refining visual details in the later stages. Acknowledging the diversity of edge devices with differing computational and storage capacities, we employ structural pruning to the SDMs U-Net and train a lightweight VAE. Empirical evaluations demonstrate that our compressed models achieve state-of-the-art parameter efficiency (225.8M) on edge devices with competitive image quality. Additionally, Hybrid SD reduces the cloud cost by 66% with edge-cloud collaborative inference."
                    },
                    {
                        "title": "Residual Local Feature Network for Efficient Super-Resolution",
                        "abstract": "Deep learning based approaches has achieved great performance in single image super-resolution (SISR). However, recent advances in efficient super-resolution focus on reducing the number of parameters and FLOPs, and they aggregate more powerful features by improving feature utilization through complex layer connection strategies. These structures may not be necessary to achieve higher running speed, which makes them difficult to be deployed to resource-constrained devices. In this work, we propose a novel Residual Local Feature Network (RLFN). The main idea is using three convolutional layers for residual local feature learning to simplify feature aggregation, which achieves a good trade-off between model performance and inference time. Moreover, we revisit the popular contrastive loss and observe that the selection of intermediate features of its feature extractor has great influence on the performance. Besides, we propose a novel multi-stage warm-start training strategy. In each stage, the pre-trained weights from previous stages are utilized to improve the model performance. Combined with the improved contrastive loss and training strategy, the proposed RLFN outperforms all the state-of-the-art efficient image SR models in terms of runtime while maintaining both PSNR and SSIM for SR. In addition, we won the first place in the runtime track of the NTIRE 2022 efficient super-resolution challenge. Code will be available at https://github.com/fyan111/RLFN."
                    },
                    {
                        "title": "MSTRIQ: No Reference Image Quality Assessment Based on Swin Transformer with Multi-Stage Fusion",
                        "abstract": "Measuring the perceptual quality of images automatically is an essential task in the area of computer vision, as degradations on image quality can exist in many processes from image acquisition, transmission to enhancing. Many Image Quality Assessment(IQA) algorithms have been designed to tackle this problem. However, it still remains un settled due to the various types of image distortions and the lack of large-scale human-rated datasets. In this paper, we propose a novel algorithm based on the Swin Transformer [31] with fused features from multiple stages, which aggregates information from both local and global features to better predict the quality. To address the issues of small-scale datasets, relative rankings of images have been taken into account together with regression loss to simultaneously optimize the model. Furthermore, effective data augmentation strategies are also used to improve the performance. In comparisons with previous works, experiments are carried out on two standard IQA datasets and a challenge dataset. The results demonstrate the effectiveness of our work. The proposed method outperforms other methods on standard datasets and ranks 2nd in the no-reference track of NTIRE 2022 Perceptual Image Quality Assessment Challenge [53]. It verifies that our method is promising in solving diverse IQA problems and thus can be used to real-word applications."
                    },
                    {
                        "title": "Boosting Video Super Resolution with Patch-Based Temporal Redundancy Optimization",
                        "abstract": "The success of existing video super-resolution (VSR) algorithms stems mainly exploiting the temporal information from the neighboring frames. However, none of these methods have discussed the influence of the temporal redundancy in the patches with stationary objects and background and usually use all the information in the adjacent frames without any discrimination. In this paper, we observe that the temporal redundancy will bring adverse effect to the information propagation,which limits the performance of the most existing VSR methods. Motivated by this observation, we aim to improve existing VSR algorithms by handling the temporal redundancy patches in an optimized manner. We develop two simple yet effective plug and play methods to improve the performance of existing local and non-local propagation-based VSR algorithms on widely-used public videos. For more comprehensive evaluating the robustness and performance of existing VSR algorithms, we also collect a new dataset which contains a variety of public videos as testing set. Extensive evaluations show that the proposed methods can significantly improve the performance of existing VSR methods on the collected videos from wild scenarios while maintain their performance on existing commonly used datasets. The code is available at https://github.com/HYHsimon/Boosted-VSR."
                    },
                    {
                        "title": "Unfolding Once is Enough: A Deployment-Friendly Transformer Unit for Super-Resolution",
                        "abstract": "Recent years have witnessed a few attempts of vision transformers for single image super-resolution (SISR). Since the high resolution of intermediate features in SISR models increases memory and computational requirements, efficient SISR transformers are more favored. Based on some popular transformer backbone, many methods have explored reasonable schemes to reduce the computational complexity of the self-attention module while achieving impressive performance. However, these methods only focus on the performance on the training platform (e.g., Pytorch/Tensorflow) without further optimization for the deployment platform (e.g., TensorRT). Therefore, they inevitably contain some redundant operators, posing challenges for subsequent deployment in real-world applications. In this paper, we propose a deployment-friendly transformer unit, namely UFONE (i.e., UnFolding ONce is Enough), to alleviate these problems. In each UFONE, we introduce an Inner-patch Transformer Layer (ITL) to efficiently reconstruct the local structural information from patches and a Spatial-Aware Layer (SAL) to exploit the long-range dependencies between patches. Based on UFONE, we propose a Deployment-friendly Inner-patch Transformer Network (DITN) for the SISR task, which can achieve favorable performance with low latency and memory usage on both training and deployment platforms. Furthermore, to further boost the deployment efficiency of the proposed DITN on TensorRT, we also provide an efficient substitution for layer normalization and propose a fusion optimization strategy for specific operators. Extensive experiments show that our models can achieve competitive results in terms of qualitative and quantitative performance with high deployment efficiency. Code is available at \\url{https://github.com/yongliuy/DITN}."
                    },
                    {
                        "title": "SparseByteNN: A Novel Mobile Inference Acceleration Framework Based on Fine-Grained Group Sparsity",
                        "abstract": "To address the challenge of increasing network size, researchers have developed sparse models through network pruning. However, maintaining model accuracy while achieving significant speedups on general computing devices remains an open problem. In this paper, we present a novel mobile inference acceleration framework SparseByteNN, which leverages fine-grained kernel sparsity to achieve real-time execution as well as high accuracy. Our framework consists of two parts: (a) A fine-grained kernel sparsity schema with a sparsity granularity between structured pruning and unstructured pruning. It designs multiple sparse patterns for different operators. Combined with our proposed whole network rearrangement strategy, the schema achieves a high compression rate and high precision at the same time. (b) Inference engine co-optimized with the sparse pattern. The conventional wisdom is that this reduction in theoretical FLOPs does not translate into real-world efficiency gains. We aim to correct this misconception by introducing a family of efficient sparse kernels for ARM and WebAssembly. Equipped with our efficient implementation of sparse primitives, we show that sparse versions of MobileNet-v1 outperform strong dense baselines on the efficiency-accuracy curve. Experimental results on Qualcomm 855 show that for 30% sparse MobileNet-v1, SparseByteNN achieves 1.27x speedup over the dense version and 1.29x speedup over the state-of-the-art sparse inference engine MNN with a slight accuracy drop of 0.224%. The source code of SparseByteNN will be available at https://github.com/lswzjuer/SparseByteNN"
                    },
                    {
                        "title": "ResAdapter: Domain Consistent Resolution Adapter for Diffusion Models",
                        "abstract": "Recent advancement in text-to-image models (e.g., Stable Diffusion) and corresponding personalized technologies (e.g., DreamBooth and LoRA) enables individuals to generate high-quality and imaginative images. However, they often suffer from limitations when generating images with resolutions outside of their trained domain. To overcome this limitation, we present the Resolution Adapter (ResAdapter), a domain-consistent adapter designed for diffusion models to generate images with unrestricted resolutions and aspect ratios. Unlike other multi-resolution generation methods that process images of static resolution with complex post-process operations, ResAdapter directly generates images with the dynamical resolution. Especially, after learning a deep understanding of pure resolution priors, ResAdapter trained on the general dataset, generates resolution-free images with personalized diffusion models while preserving their original style domain. Comprehensive experiments demonstrate that ResAdapter with only 0.5M can process images with flexible resolutions for arbitrary diffusion models. More extended experiments demonstrate that ResAdapter is compatible with other modules (e.g., ControlNet, IP-Adapter and LCM-LoRA) for image generation across a broad range of resolutions, and can be integrated into other multi-resolution model (e.g., ElasticDiffusion) for efficiently generating higher-resolution images. Project link is https://res-adapter.github.io"
                    },
                    {
                        "title": "PatchScaler: An Efficient Patch-Independent Diffusion Model for Super-Resolution",
                        "abstract": "Diffusion models significantly improve the quality of super-resolved images with their impressive content generation capabilities. However, the huge computational costs limit the applications of these methods.Recent efforts have explored reasonable inference acceleration to reduce the number of sampling steps, but the computational cost remains high as each step is performed on the entire image.This paper introduces PatchScaler, a patch-independent diffusion-based single image super-resolution (SR) method, designed to enhance the efficiency of the inference process.The proposed method is motivated by the observation that not all the image patches within an image need the same sampling steps for reconstructing high-resolution images.Based on this observation, we thus develop a Patch-adaptive Group Sampling (PGS) to divide feature patches into different groups according to the patch-level reconstruction difficulty and dynamically assign an appropriate sampling configuration for each group so that the inference speed can be better accelerated.In addition, to improve the denoising ability at each step of the sampling, we develop a texture prompt to guide the estimations of the diffusion model by retrieving high-quality texture priors from a patch-independent reference texture memory.Experiments show that our PatchScaler achieves favorable performance in both quantitative and qualitative evaluations with fast inference speed.Our code and model are available at \\url{https://github.com/yongliuy/PatchScaler}."
                    }
                ]
            },
            "92264184-3ef1-472d-878c-3efe57b4d827": {
                "pk": "92264184-3ef1-472d-878c-3efe57b4d827",
                "name": "Guanbin Li",
                "collaborators": [
                    "Yizhou Yu",
                    "Haofeng Li",
                    "Jichang Li",
                    "Liang Lin",
                    "Xiang Wan",
                    "Guanqi Chen",
                    "Junjia Huang",
                    "Yuan Xie",
                    "Feng Liu",
                    "Wei Lou"
                ],
                "domain": [
                    "Computer Vision",
                    "Deep Learning",
                    "Salient Object Detection",
                    "Semi-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Deep Contrast Learning for Salient Object Detection",
                        "abstract": "Salient object detection has recently witnessed substantial progress due to powerful features extracted using deep convolutional neural networks (CNNs). However, existing CNN-based methods operate at the patch level instead of the pixel level. Resulting saliency maps are typically blurry, especially near the boundary of salient objects. Furthermore, image patches are treated as independent samples even when they are overlapping, giving rise to significant redundancy in computation and storage. In this CVPR 2016 paper, we propose an end-to-end deep contrast network to overcome the aforementioned limitations. Our deep network consists of two complementary components, a pixel-level fully convolutional stream and a segment-wise spatial pooling stream. The first stream directly produces a saliency map with pixel-level accuracy from an input image. The second stream extracts segment-wise features very efficiently, and better models saliency discontinuities along object boundaries. Finally, a fully connected CRF model can be optionally incorporated to improve spatial coherence and contour localization in the fused result from these two streams. Experimental results demonstrate that our deep model significantly improves the state of the art."
                    },
                    {
                        "title": "Visual Saliency Detection Based on Multiscale Deep CNN Features",
                        "abstract": "Visual saliency is a fundamental problem in both cognitive and computational sciences, including computer vision. In this paper, we discover that a high-quality visual saliency model can be learned from multiscale features extracted using deep convolutional neural networks (CNNs), which have had many successes in visual recognition tasks. For learning such saliency models, we introduce a neural network architecture, which has fully connected layers on top of CNNs responsible for feature extraction at three different scales. The penultimate layer of our neural network has been confirmed to be a discriminative high-level feature vector for saliency detection, which we call deep contrast feature. To generate a more robust feature, we integrate handcrafted low-level features with our deep contrast feature. To promote further research and evaluation of visual saliency models, we also construct a new large database of 4447 challenging images and their pixelwise saliency annotations. Experimental results demonstrate that our proposed method is capable of achieving state-of-the-art performance on all public benchmarks, improving the F- measure by 6.12% and 10.0% respectively on the DUT-OMRON dataset and our new dataset (HKU-IS), and lowering the mean absolute error by 9% and 35.3% respectively on these two datasets."
                    },
                    {
                        "title": "Visual Saliency Based on Multiscale Deep Features",
                        "abstract": "Visual saliency is a fundamental problem in both cognitive and computational sciences, including computer vision. In this CVPR 2015 paper, we discover that a high-quality visual saliency model can be trained with multiscale features extracted using a popular deep learning architecture, convolutional neural networks (CNNs), which have had many successes in visual recognition tasks. For learning such saliency models, we introduce a neural network architecture, which has fully connected layers on top of CNNs responsible for extracting features at three different scales. We then propose a refinement method to enhance the spatial coherence of our saliency results. Finally, aggregating multiple saliency maps computed for different levels of image segmentation can further boost the performance, yielding saliency maps better than those generated from a single segmentation. To promote further research and evaluation of visual saliency models, we also construct a new large database of 4447 challenging images and their pixelwise saliency annotation. Experimental results demonstrate that our proposed method is capable of achieving state-of-the-art performance on all public benchmarks, improving the F-Measure by 5.0% and 13.2% respectively on the MSRA-B dataset and our new dataset (HKU-IS), and lowering the mean absolute error by 5.7% and 35.1% respectively on these two datasets."
                    },
                    {
                        "title": "Contrast-Oriented Deep Neural Networks for Salient Object Detection",
                        "abstract": "Deep convolutional neural networks have become a key element in the recent breakthrough of salient object detection. However, existing CNN-based methods are based on either patch-wise (region-wise) training and inference or fully convolutional networks. Methods in the former category are generally time-consuming due to severe storage and computational redundancies among overlapping patches. To overcome this deficiency, methods in the second category attempt to directly map a raw input image to a predicted dense saliency map in a single network forward pass. Though being very efficient, it is arduous for these methods to detect salient objects of different scales or salient regions with weak semantic information. In this paper, we develop hybrid contrast-oriented deep neural networks to overcome the aforementioned limitations. Each of our deep networks is composed of two complementary components, including a fully convolutional stream for dense prediction and a segment-level spatial pooling stream for sparse saliency inference. We further propose an attentional module that learns weight maps for fusing the two saliency predictions from these two streams. A tailored alternate scheme is designed to train these deep networks by fine-tuning pre-trained baseline models. Finally, a customized fully connected CRF model incorporating a salient contour feature embedding can be optionally applied as a post-processing step to improve spatial coherence and contour positioning in the fused result from these two streams. Extensive experiments on six benchmark datasets demonstrate that our proposed model can significantly outperform the state of the art in terms of all popular evaluation metrics."
                    },
                    {
                        "title": "Semantic-aware Temporal Channel-wise Attention for Cardiac Function Assessment",
                        "abstract": "Cardiac function assessment aims at predicting left ventricular ejection fraction (LVEF) given an echocardiogram video, which requests models to focus on the changes in the left ventricle during the cardiac cycle. How to assess cardiac function accurately and automatically from an echocardiogram video is a valuable topic in intelligent assisted healthcare. Existing video-based methods do not pay much attention to the left ventricular region, nor the left ventricular changes caused by motion. In this work, we propose a semi-supervised auxiliary learning paradigm with a left ventricular segmentation task, which contributes to the representation learning for the left ventricular region. To better model the importance of motion information, we introduce a temporal channel-wise attention (TCA) module to excite those channels used to describe motion. Furthermore, we reform the TCA module with semantic perception by taking the segmentation map of the left ventricle as input to focus on the motion patterns of the left ventricle. Finally, to reduce the difficulty of direct LVEF regression, we utilize an anchor-based classification and regression method to predict LVEF. Our approach achieves state-of-the-art performance on the Stanford dataset with an improvement of 0.22 MAE, 0.26 RMSE, and 1.9% $R^2$."
                    },
                    {
                        "title": "Neighborhood Collective Estimation for Noisy Label Identification and Correction",
                        "abstract": "Learning with noisy labels (LNL) aims at designing strategies to improve model performance and generalization by mitigating the effects of model overfitting to noisy labels. The key success of LNL lies in identifying as many clean samples as possible from massive noisy data, while rectifying the wrongly assigned noisy labels. Recent advances employ the predicted label distributions of individual samples to perform noise verification and noisy label correction, easily giving rise to confirmation bias. To mitigate this issue, we propose Neighborhood Collective Estimation, in which the predictive reliability of a candidate sample is re-estimated by contrasting it against its feature-space nearest neighbors. Specifically, our method is divided into two steps: 1) Neighborhood Collective Noise Verification to separate all training samples into a clean or noisy subset, 2) Neighborhood Collective Label Correction to relabel noisy samples, and then auxiliary techniques are used to assist further model optimization. Extensive experiments on four commonly used benchmark datasets, i.e., CIFAR-10, CIFAR-100, Clothing-1M and Webvision-1.0, demonstrate that our proposed method considerably outperforms state-of-the-art methods."
                    },
                    {
                        "title": "Context-Aware Semantic Inpainting",
                        "abstract": "Recently image inpainting has witnessed rapid progress due to generative adversarial networks (GAN) that are able to synthesize realistic contents. However, most existing GAN-based methods for semantic inpainting apply an auto-encoder architecture with a fully connected layer, which cannot accurately maintain spatial information. In addition, the discriminator in existing GANs struggle to understand high-level semantics within the image context and yield semantically consistent content. Existing evaluation criteria are biased towards blurry results and cannot well characterize edge preservation and visual authenticity in the inpainting results. In this paper, we propose an improved generative adversarial network to overcome the aforementioned limitations. Our proposed GAN-based framework consists of a fully convolutional design for the generator which helps to better preserve spatial structures and a joint loss function with a revised perceptual loss to capture high-level semantics in the context. Furthermore, we also introduce two novel measures to better assess the quality of image inpainting results. Experimental results demonstrate that our method outperforms the state of the art under a wide range of criteria."
                    },
                    {
                        "title": "UniCell: Universal Cell Nucleus Classification via Prompt Learning",
                        "abstract": "The recognition of multi-class cell nuclei can significantly facilitate the process of histopathological diagnosis. Numerous pathological datasets are currently available, but their annotations are inconsistent. Most existing methods require individual training on each dataset to deduce the relevant labels and lack the use of common knowledge across datasets, consequently restricting the quality of recognition. In this paper, we propose a universal cell nucleus classification framework (UniCell), which employs a novel prompt learning mechanism to uniformly predict the corresponding categories of pathological images from different dataset domains. In particular, our framework adopts an end-to-end architecture for nuclei detection and classification, and utilizes flexible prediction heads for adapting various datasets. Moreover, we develop a Dynamic Prompt Module (DPM) that exploits the properties of multiple datasets to enhance features. The DPM first integrates the embeddings of datasets and semantic categories, and then employs the integrated prompts to refine image representations, efficiently harvesting the shared knowledge among the related cell types and data sources. Experimental results demonstrate that the proposed method effectively achieves the state-of-the-art results on four nucleus detection and classification benchmarks. Code and models are available at https://github.com/lhaof/UniCell"
                    },
                    {
                        "title": "Cell Graph Transformer for Nuclei Classification",
                        "abstract": "Nuclei classification is a critical step in computer-aided diagnosis with histopathology images. In the past, various methods have employed graph neural networks (GNN) to analyze cell graphs that model inter-cell relationships by considering nuclei as vertices. However, they are limited by the GNN mechanism that only passes messages among local nodes via fixed edges. To address the issue, we develop a cell graph transformer (CGT) that treats nodes and edges as input tokens to enable learnable adjacency and information exchange among all nodes. Nevertheless, training the transformer with a cell graph presents another challenge. Poorly initialized features can lead to noisy self-attention scores and inferior convergence, particularly when processing the cell graphs with numerous connections. Thus, we further propose a novel topology-aware pretraining method that leverages a graph convolutional network (GCN) to learn a feature extractor. The pre-trained features may suppress unreasonable correlations and hence ease the finetuning of CGT. Experimental results suggest that the proposed cell graph transformer with topology-aware pretraining significantly improves the nuclei classification results, and achieves the state-of-the-art performance. Code and models are available at https://github.com/lhaof/CGT"
                    },
                    {
                        "title": "ROSA: Robust Salient Object Detection against Adversarial Attacks",
                        "abstract": "Recently salient object detection has witnessed remarkable improvement owing to the deep convolutional neural networks which can harvest powerful features for images. In particular, state-of-the-art salient object detection methods enjoy high accuracy and efficiency from fully convolutional network (FCN) based frameworks which are trained from end to end and predict pixel-wise labels. However, such framework suffers from adversarial attacks which confuse neural networks via adding quasi-imperceptible noises to input images without changing the ground truth annotated by human subjects. To our knowledge, this paper is the first one that mounts successful adversarial attacks on salient object detection models and verifies that adversarial samples are effective on a wide range of existing methods. Furthermore, this paper proposes a novel end-to-end trainable framework to enhance the robustness for arbitrary FCN-based salient object detection models against adversarial attacks. The proposed framework adopts a novel idea that first introduces some new generic noise to destroy adversarial perturbations, and then learns to predict saliency maps for input images with the introduced noise. Specifically, our proposed method consists of a segment-wise shielding component, which preserves boundaries and destroys delicate adversarial noise patterns and a context-aware restoration component, which refines saliency maps through global contrast modeling. Experimental results suggest that our proposed framework improves the performance significantly for state-of-the-art models on a series of datasets."
                    },
                    {
                        "title": "Motion Guided Attention for Video Salient Object Detection",
                        "abstract": "Video salient object detection aims at discovering the most visually distinctive objects in a video. How to effectively take object motion into consideration during video salient object detection is a critical issue. Existing state-of-the-art methods either do not explicitly model and harvest motion cues or ignore spatial contexts within optical flow images. In this paper, we develop a multi-task motion guided video salient object detection network, which learns to accomplish two sub-tasks using two sub-networks, one sub-network for salient object detection in still images and the other for motion saliency detection in optical flow images. We further introduce a series of novel motion guided attention modules, which utilize the motion saliency sub-network to attend and enhance the sub-network for still images. These two sub-networks learn to adapt to each other by end-to-end training. Experimental results demonstrate that the proposed method significantly outperforms existing state-of-the-art algorithms on a wide range of benchmarks. We hope our simple and effective approach will serve as a solid baseline and help ease future research in video salient object detection. Code and models will be made available."
                    },
                    {
                        "title": "Cross-Domain Adaptive Clustering for Semi-Supervised Domain Adaptation",
                        "abstract": "In semi-supervised domain adaptation, a few labeled samples per class in the target domain guide features of the remaining target samples to aggregate around them. However, the trained model cannot produce a highly discriminative feature representation for the target domain because the training data is dominated by labeled samples from the source domain. This could lead to disconnection between the labeled and unlabeled target samples as well as misalignment between unlabeled target samples and the source domain. In this paper, we propose a novel approach called Cross-domain Adaptive Clustering to address this problem. To achieve both inter-domain and intra-domain adaptation, we first introduce an adversarial adaptive clustering loss to group features of unlabeled target data into clusters and perform cluster-wise feature alignment across the source and target domains. We further apply pseudo labeling to unlabeled samples in the target domain and retain pseudo-labels with high confidence. Pseudo labeling expands the number of ``labeled\" samples in each class in the target domain, and thus produces a more robust and powerful cluster core for each class to facilitate adversarial learning. Extensive experiments on benchmark datasets, including DomainNet, Office-Home and Office, demonstrate that our proposed approach achieves the state-of-the-art performance in semi-supervised domain adaptation."
                    },
                    {
                        "title": "Attentive Symmetric Autoencoder for Brain MRI Segmentation",
                        "abstract": "Self-supervised learning methods based on image patch reconstruction have witnessed great success in training auto-encoders, whose pre-trained weights can be transferred to fine-tune other downstream tasks of image understanding. However, existing methods seldom study the various importance of reconstructed patches and the symmetry of anatomical structures, when they are applied to 3D medical images. In this paper we propose a novel Attentive Symmetric Auto-encoder (ASA) based on Vision Transformer (ViT) for 3D brain MRI segmentation tasks. We conjecture that forcing the auto-encoder to recover informative image regions can harvest more discriminative representations, than to recover smooth image patches. Then we adopt a gradient based metric to estimate the importance of each image patch. In the pre-training stage, the proposed auto-encoder pays more attention to reconstruct the informative patches according to the gradient metrics. Moreover, we resort to the prior of brain structures and develop a Symmetric Position Encoding (SPE) method to better exploit the correlations between long-range but spatially symmetric regions to obtain effective features. Experimental results show that our proposed attentive symmetric auto-encoder outperforms the state-of-the-art self-supervised learning methods and medical image segmentation models on three brain MRI segmentation benchmarks."
                    },
                    {
                        "title": "Adaptive Betweenness Clustering for Semi-Supervised Domain Adaptation",
                        "abstract": "Compared to unsupervised domain adaptation, semi-supervised domain adaptation (SSDA) aims to significantly improve the classification performance and generalization capability of the model by leveraging the presence of a small amount of labeled data from the target domain. Several SSDA approaches have been developed to enable semantic-aligned feature confusion between labeled (or pseudo labeled) samples across domains; nevertheless, owing to the scarcity of semantic label information of the target domain, they were arduous to fully realize their potential. In this study, we propose a novel SSDA approach named Graph-based Adaptive Betweenness Clustering (G-ABC) for achieving categorical domain alignment, which enables cross-domain semantic alignment by mandating semantic transfer from labeled data of both the source and target domains to unlabeled target samples. In particular, a heterogeneous graph is initially constructed to reflect the pairwise relationships between labeled samples from both domains and unlabeled ones of the target domain. Then, to degrade the noisy connectivity in the graph, connectivity refinement is conducted by introducing two strategies, namely Confidence Uncertainty based Node Removal and Prediction Dissimilarity based Edge Pruning. Once the graph has been refined, Adaptive Betweenness Clustering is introduced to facilitate semantic transfer by using across-domain betweenness clustering and within-domain betweenness clustering, thereby propagating semantic label information from labeled samples across domains to unlabeled target data. Extensive experiments on three standard benchmark datasets, namely DomainNet, Office-Home, and Office-31, indicated that our method outperforms previous state-of-the-art SSDA approaches, demonstrating the superiority of the proposed G-ABC algorithm."
                    },
                    {
                        "title": "Inter-Domain Mixup for Semi-Supervised Domain Adaptation",
                        "abstract": "Semi-supervised domain adaptation (SSDA) aims to bridge source and target domain distributions, with a small number of target labels available, achieving better classification performance than unsupervised domain adaptation (UDA). However, existing SSDA work fails to make full use of label information from both source and target domains for feature alignment across domains, resulting in label mismatch in the label space during model testing. This paper presents a novel SSDA approach, Inter-domain Mixup with Neighborhood Expansion (IDMNE), to tackle this issue. Firstly, we introduce a cross-domain feature alignment strategy, Inter-domain Mixup, that incorporates label information into model adaptation. Specifically, we employ sample-level and manifold-level data mixing to generate compatible training samples. These newly established samples, combined with reliable and actual label information, display diversity and compatibility across domains, while such extra supervision thus facilitates cross-domain feature alignment and mitigates label mismatch. Additionally, we utilize Neighborhood Expansion to leverage high-confidence pseudo-labeled samples in the target domain, diversifying the label information of the target domain and thereby further increasing the performance of the adaptation model. Accordingly, the proposed approach outperforms existing state-of-the-art methods, achieving significant accuracy improvements on popular SSDA benchmarks, including DomainNet, Office-Home, and Office-31."
                    },
                    {
                        "title": "Instance-Level Salient Object Segmentation",
                        "abstract": "Image saliency detection has recently witnessed rapid progress due to deep convolutional neural networks. However, none of the existing methods is able to identify object instances in the detected salient regions. In this paper, we present a salient instance segmentation method that produces a saliency mask with distinct object instance labels for an input image. Our method consists of three steps, estimating saliency map, detecting salient object contours and identifying salient object instances. For the first two steps, we propose a multiscale saliency refinement network, which generates high-quality salient region masks and salient object contours. Once integrated with multiscale combinatorial grouping and a MAP-based subset optimization framework, our method can generate very promising salient object instance segmentation results. To promote further research and evaluation of salient instance segmentation, we also construct a new database of 1000 images and their pixelwise salient instance annotations. Experimental results demonstrate that our proposed method is capable of achieving state-of-the-art performance on all public benchmarks for salient region detection as well as on our new dataset for salient instance segmentation."
                    },
                    {
                        "title": "Tree-Structured Policy based Progressive Reinforcement Learning for Temporally Language Grounding in Video",
                        "abstract": "Temporally language grounding in untrimmed videos is a newly-raised task in video understanding. Most of the existing methods suffer from inferior efficiency, lacking interpretability, and deviating from the human perception mechanism. Inspired by human's coarse-to-fine decision-making paradigm, we formulate a novel Tree-Structured Policy based Progressive Reinforcement Learning (TSP-PRL) framework to sequentially regulate the temporal boundary by an iterative refinement process. The semantic concepts are explicitly represented as the branches in the policy, which contributes to efficiently decomposing complex policies into an interpretable primitive action. Progressive reinforcement learning provides correct credit assignment via two task-oriented rewards that encourage mutual promotion within the tree-structured policy. We extensively evaluate TSP-PRL on the Charades-STA and ActivityNet datasets, and experimental results show that TSP-PRL achieves competitive performance over existing state-of-the-art methods."
                    },
                    {
                        "title": "Weakly Supervised Salient Object Detection Using Image Labels",
                        "abstract": "Deep learning based salient object detection has recently achieved great success with its performance greatly outperforms any other unsupervised methods. However, annotating per-pixel saliency masks is a tedious and inefficient procedure. In this paper, we note that superior salient object detection can be obtained by iteratively mining and correcting the labeling ambiguity on saliency maps from traditional unsupervised methods. We propose to use the combination of a coarse salient object activation map from the classification network and saliency maps generated from unsupervised methods as pixel-level annotation, and develop a simple yet very effective algorithm to train fully convolutional networks for salient object detection supervised by these noisy annotations. Our algorithm is based on alternately exploiting a graphical model and training a fully convolutional network for model updating. The graphical model corrects the internal labeling ambiguity through spatial consistency and structure preserving while the fully convolutional network helps to correct the cross-image semantic ambiguity and simultaneously update the coarse activation map for next iteration. Experimental results demonstrate that our proposed method greatly outperforms all state-of-the-art unsupervised saliency detection methods and can be comparable to the current best strongly-supervised methods training with thousands of pixel-level saliency map annotations on all public benchmarks."
                    },
                    {
                        "title": "Improved Distribution Matching for Dataset Condensation",
                        "abstract": "Dataset Condensation aims to condense a large dataset into a smaller one while maintaining its ability to train a well-performing model, thus reducing the storage cost and training effort in deep learning applications. However, conventional dataset condensation methods are optimization-oriented and condense the dataset by performing gradient or parameter matching during model optimization, which is computationally intensive even on small datasets and models. In this paper, we propose a novel dataset condensation method based on distribution matching, which is more efficient and promising. Specifically, we identify two important shortcomings of naive distribution matching (i.e., imbalanced feature numbers and unvalidated embeddings for distance computation) and address them with three novel techniques (i.e., partitioning and expansion augmentation, efficient and enriched model sampling, and class-aware distribution regularization). Our simple yet effective method outperforms most previous optimization-oriented methods with much fewer computational resources, thereby scaling data condensation to larger datasets and models. Extensive experiments demonstrate the effectiveness of our method. Codes are available at https://github.com/uitrbn/IDM"
                    },
                    {
                        "title": "Semi- and Weakly-Supervised Learning for Mammogram Mass Segmentation with Limited Annotations",
                        "abstract": "Accurate identification of breast masses is crucial in diagnosing breast cancer; however, it can be challenging due to their small size and being camouflaged in surrounding normal glands. Worse still, it is also expensive in clinical practice to obtain adequate pixel-wise annotations for training deep neural networks. To overcome these two difficulties with one stone, we propose a semi- and weakly-supervised learning framework for mass segmentation that utilizes limited strongly-labeled samples and sufficient weakly-labeled samples to achieve satisfactory performance. The framework consists of an auxiliary branch to exclude lesion-irrelevant background areas, a segmentation branch for final prediction, and a spatial prompting module to integrate the complementary information of the two branches. We further disentangle encoded obscure features into lesion-related and others to boost performance. Experiments on CBIS-DDSM and INbreast datasets demonstrate the effectiveness of our method."
                    }
                ]
            }
        }
    },
    "2309.03160": {
        "paper_data": {
            "title": "ResFields: Residual Neural Fields for Spatiotemporal Signals",
            "url": "http://arxiv.org/abs/2309.03160v5",
            "arxiv_id": "2309.03160",
            "authors": [
                "Marko Mihajlovic",
                "Sergey Prokudin",
                "Marc Pollefeys",
                "Siyu Tang"
            ],
            "abstract": "Neural fields, a category of neural networks trained to represent high-frequency signals, have gained significant attention in recent years due to their impressive performance in modeling complex 3D data, such as signed distance (SDFs) or radiance fields (NeRFs), via a single multi-layer perceptron (MLP). However, despite the power and simplicity of representing signals with an MLP, these methods still face challenges when modeling large and complex temporal signals due to the limited capacity of MLPs. In this paper, we propose an effective approach to address this limitation by incorporating temporal residual layers into neural fields, dubbed ResFields. It is a novel class of networks specifically designed to effectively represent complex temporal signals. We conduct a comprehensive analysis of the properties of ResFields and propose a matrix factorization technique to reduce the number of trainable parameters and enhance generalization capabilities. Importantly, our formulation seamlessly integrates with existing MLP-based neural fields and consistently improves results across various challenging tasks: 2D video approximation, dynamic shape modeling via temporal SDFs, and dynamic NeRF reconstruction. Lastly, we demonstrate the practical utility of ResFields by showcasing its effectiveness in capturing dynamic 3D scenes from sparse RGBD cameras of a lightweight capture system.",
            "introduction": "   1 Introduction  Multi-layer Perceptron (MLP) is a common neural network architecture used for representing continuous spatiotemporal signals, known as neural fields. Its popularity stems from its capacity to encode continuous signals across arbitrary dimensions (Kim & Adal\u0131, 2003). Additionally, inherent implicit regularization (Goodfellow et\u00a0al., 2016; Neyshabur et\u00a0al., 2014) and spectral bias (Rahaman et\u00a0al., 2019) equip MLPs with excellent interpolation capabilities. Due to these remarkable properties, MLPs have achieved widespread success in many applications such as image synthesis, animation, texture generation, and novel view synthesis (Tewari et\u00a0al., 2022; Xie et\u00a0al., 2022).   However, the spectral bias of MLPs (Rahaman et\u00a0al., 2019), which refers to the tendency of neural networks to learn functions with low frequencies, presents a challenge when it comes to accurately representing complex real-world signals and capturing fine-grained details. Previous efforts have aimed to address the spectral bias by utilizing techniques like positional encoding (Vaswani et\u00a0al., 2017; Mildenhall et\u00a0al., 2020; Zhong et\u00a0al., 2019; M\u00fcller et\u00a0al., 2022) or special activation functions (Sitzmann et\u00a0al., 2020b; Fathony et\u00a0al., 2020). However, even with these methods, representing fine-grained details remains a challenge, particularly when dealing with large spatiotemporal signals such as long videos or dynamic 3D scenes.   A straightforward way of increasing the capacity of MLPs is to increase the network complexity in terms of the total number of neurons. However, such an approach would make the inference and optimization slower and more GPU memory expensive, as time and memory complexity scales with respect to the total number of parameters. Another possibility is to meta-learn MLP weights (Sitzmann et\u00a0al., 2020a) and maintain specialized independent parameters, but this imposes slow training that does not scale to photo-realistic reconstructions (Tancik et\u00a0al., 2021). By far the most popular approach for increasing modeling capacity is to partition the spatiotemporal field and fit separate/local neural fields (Reiser et\u00a0al., 2021; M\u00fcller et\u00a0al., 2022; Chen et\u00a0al., 2022). However, these approaches hinder global reasoning and generalization due to local gradient updates of grid structures (Peng et\u00a0al., 2023).  \\begin{overpic}[width=397.48499pt,keepaspectratio]{figures/resfields_teaser_v2% _notxt_bigger.pdf} \\put(6.5,48.0){Siren~{}\\cite[citep]{(\\@@bibref{AuthorsPhrase1Year}{sitzmann202% 0implicit}{\\@@citephrase{, }}{})}} \\put(60.0,48.0){TNeRF~{}\\cite[citep]{(\\@@bibref{AuthorsPhrase1Year}{DyNeRF}{% \\@@citephrase{, }}{})}} \\put(0.0,33.0){\\rotatebox{90.0}{Baseline}} \\put(0.0,5.0){\\rotatebox{90.0}{+ResFields}} \\\\ \\put(18.0,-2.0){{(a)}} \\put(55.0,-2.0){{(b)}} \\put(86.0,-2.0){{(c)}} \\end{overpic}  Figure 1:  ResField extends an MLP architecture to effectively represent complex temporal signals by replacing the conventional linear layers with Residual Field Layers. As such, ResField is versatile and straightforwardly compatible with most existing temporal neural fields. Here we demonstrate its applicability on three challenging tasks by extending Siren\u00a0(Sitzmann et\u00a0al., 2020b) and TNeRF\u00a0(Li et\u00a0al., 2022): (a) learning temporal signed distance fields and (b) neural radiance fields from four RGB views and (c) from three time-synchronized RGBD views captured by our lightweight rig. The figure is best viewed in electronic format on a color screen, please zoom-in to observe details.    Figure 2: ResField MLP Architecture.    The challenge that we aim to address is how to increase the model capacity in a way that is agnostic to the design choices of MLP neural fields. This includes architecture, input encoding, and activation functions. At the same time, we must maintain the implicit regularization property of neural networks and retain compatibility with existing techniques developed for reducing the spectral bias (Mildenhall et\u00a0al., 2020; Sitzmann et\u00a0al., 2020b). Our key idea is to substitute MLP layers with time-dependent layers (see Fig.\u00a02) whose weights are modeled as trainable residual parameters \ud835\udce6i\u2062(t)subscript\ud835\udce6\ud835\udc56\ud835\udc61\\bm{\\mathcal{W}}_{i}(t)bold_caligraphic_W start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT (",
            "references": [
                {
                    "title": "Tracking Everything Everywhere All at Once",
                    "abstract": "We present a new test-time optimization method for estimating dense and long-range motion from a video sequence. Prior optical flow or particle video tracking algorithms typically operate within limited temporal windows, struggling to track through occlusions and maintain global consistency of estimated motion trajectories. We propose a complete and globally consistent motion representation, dubbed OmniMotion, that allows for accurate, full-length motion estimation of every pixel in a video. OmniMotion represents a video using a quasi-3D canonical volume and performs pixel-wise tracking via bijections between local and canonical space. This representation allows us to ensure global consistency, track through occlusions, and model any combination of camera and object motion. Extensive evaluations on the TAP-Vid benchmark and real-world footage show that our approach outperforms prior state-of-the-art methods by a large margin both quantitatively and qualitatively. See our project page for more results: omnimotion.github.io."
                },
                {
                    "title": "QLoRA: Efficient Finetuning of Quantized LLMs",
                    "abstract": "We present QLoRA, an efficient finetuning approach that reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU while preserving full 16-bit finetuning task performance. QLoRA backpropagates gradients through a frozen, 4-bit quantized pretrained language model into Low Rank Adapters~(LoRA). Our best model family, which we name Guanaco, outperforms all previous openly released models on the Vicuna benchmark, reaching 99.3% of the performance level of ChatGPT while only requiring 24 hours of finetuning on a single GPU. QLoRA introduces a number of innovations to save memory without sacrificing performance: (a) 4-bit NormalFloat (NF4), a new data type that is information theoretically optimal for normally distributed weights (b) double quantization to reduce the average memory footprint by quantizing the quantization constants, and (c) paged optimziers to manage memory spikes. We use QLoRA to finetune more than 1,000 models, providing a detailed analysis of instruction following and chatbot performance across 8 instruction datasets, multiple model types (LLaMA, T5), and model scales that would be infeasible to run with regular finetuning (e.g. 33B and 65B parameter models). Our results show that QLoRA finetuning on a small high-quality dataset leads to state-of-the-art results, even when using smaller models than the previous SoTA. We provide a detailed analysis of chatbot performance based on both human and GPT-4 evaluations showing that GPT-4 evaluations are a cheap and reasonable alternative to human evaluation. Furthermore, we find that current chatbot benchmarks are not trustworthy to accurately evaluate the performance levels of chatbots. A lemon-picked analysis demonstrates where Guanaco fails compared to ChatGPT. We release all of our models and code, including CUDA kernels for 4-bit training."
                },
                {
                    "title": "HumanRF: High-Fidelity Neural Radiance Fields for Humans in Motion",
                    "abstract": "Representing human performance at high-fidelity is an essential building block in diverse applications, such as film production, computer games or videoconferencing. To close the gap to production-level quality, we introduce HumanRF1, a 4D dynamic neural scene representation that captures full-body appearance in motion from multi-view video input, and enables playback from novel, unseen viewpoints. Our novel representation acts as a dynamic video encoding that captures fine details at high compression rates by factorizing space-time into a temporal matrix-vector decomposition. This allows us to obtain temporally coherent reconstructions of human actors for long sequences, while representing high-resolution details even in the context of challenging motion. While most research focuses on synthesizing at resolutions of 4MP or lower, we address the challenge of operating at 12MP. To this end, we introduce ActorsHQ, a novel multi-view dataset that provides 12MP footage from 160 cameras for 16 sequences with high-fidelity, per-frame mesh reconstructions2. We demonstrate challenges that emerge from using such high-resolution data and show that our newly introduced HumanRF effectively leverages this data, making a significant step towards production-level quality novel view synthesis."
                },
                {
                    "title": "PET-NeuS: Positional Encoding Tri-Planes for Neural Surfaces",
                    "abstract": "A signed distance function (SDF) parametrized by an MLP is a common ingredient of neural surface reconstruction. We build on the successful recent method NeuS to extend it by three new components. The first component is to borrow the tri-plane representation from EG3D and represent signed distance fields as a mixture of tri-planes and MLPs instead of representing it with MLPs only. Using tri-planes leads to a more expressive data structure but will also introduce noise in the reconstructed surface. The second component is to use a new type of positional encoding with learnable weights to combat noise in the reconstruction process. We divide the features in the tri-plane into multiple frequency scales and modulate them with sin and cos functions of different frequencies. The third component is to use learnable convolution operations on the tri-plane features using self-attention convolution to produce features with different frequency bands. The experiments show that PET-NeuS achieves high-fidelity surface reconstruction on standard datasets. Following previous work and using the Chamfer metric as the most important way to measure surface reconstruction quality, we are able to improve upon the NeuS baseline by 57% on Nerf-synthetic (0.84 compared to 1.97) and by 15.5% on DTU (0.71 compared to 0.84). The qualitative evaluation reveals how our method can better control the interference of high-frequency noise."
                },
                {
                    "title": "Representing Volumetric Videos as Dynamic MLP Maps",
                    "abstract": "This paper introduces a novel representation of volumetric videos for real-time view synthesis of dynamic scenes. Recent advances in neural scene representations demonstrate their remarkable capability to model and render complex static scenes, but extending them to represent dynamic scenes is not straightforward due to their slow rendering speed or high storage cost. To solve this problem, our key idea is to represent the radiance field of each frame as a set of shallow MLP networks whose parameters are stored in 2D grids, called MLP maps, and dynamically predicted by a 2D CNN decoder shared by all frames. Representing 3D scenes with shallow MLPs significantly improves the rendering speed, while dynamically predicting MLP parameters with a shared 2D CNN instead of explicitly storing them leads to low storage cost. Experiments show that the proposed approach achieves state-of-the-art rendering quality on the NHR and ZJU-MoCap datasets, while being efficient for real-time rendering with a speed of 41.7 fps for $512\\times 512$ images on an RTX 3090 GPU. The code is available at https://zju3dv.github.io/mlp_maps/."
                },
                {
                    "title": "Flow Supervision for Deformable NeRF",
                    "abstract": "In this paper we present a new method for deformable NeRF that can directly use optical flow as supervision. We overcome the major challenge with respect to the computationally inefficiency of enforcing the flow constraints to the backward deformation field, used by deformable NeRFs. Specifically, we show that inverting the backward deformation function is actually not needed for computing scene flows between frames. This insight dramatically simplifies the problem, as one is no longer constrained to deformation functions that can be analytically inverted. Instead, thanks to the weak assumptions required by our derivation based on the inverse function theorem, our approach can be extended to a broad class of commonly used backward deformation field. We present results on monocular novel view synthesis with rapid object motion, and demonstrate significant improvements over baselines without flow supervision."
                },
                {
                    "title": "K-Planes: Explicit Radiance Fields in Space, Time, and Appearance",
                    "abstract": "We introduce k-planes, a white-box model for radiance fields in arbitrary dimensions. Our model uses planes to represent a d-dimensional scene, providing a seamless way to go from static (d = 3) to dynamic (d= 4) scenes. This planar factorization makes adding dimension-specific priors easy, e.g. temporal smoothness and multi-resolution spatial structure, and induces a natural decomposition of static and dynamic components of a scene. We use a linear feature decoder with a learned color basis that yields similar performance as a nonlinear black-box MLP decoder. Across a range of synthetic and real, static and dynamic, fixed and varying appearance scenes, k-planes yields competitive and often state-of-the-art recon- struction fidelity with low memory usage, achieving 1000x compression over a full 4D grid, and fast optimization with a pure PyTorch implementation. For video results and code, please see sarafridov.github.io/K-Planes."
                },
                {
                    "title": "HexPlane: A Fast Representation for Dynamic Scenes",
                    "abstract": "Modeling and re-rendering dynamic 3D scenes is a challenging task in 3D vision. Prior approaches build on NeRF and rely on implicit representations. This is slow since it requires many MLP evaluations, constraining real-world applications. We show that dynamic 3D scenes can be explicitly represented by six planes of learned features, leading to an elegant solution we call HexPlane. A HexPlane computes features for points in spacetime by fusing vectors extracted from each plane, which is highly efficient. Pairing a HexPlane with a tiny MLP to regress output colors and training via volume rendering gives impressive results for novel view synthesis on dynamic scenes, matching the image quality of prior work but reducing training time by more than 100\u00d7. Extensive ablations confirm our HexPlane design and show that it is robust to different feature fusion mechanisms, coordinate systems, and decoding mechanisms. HexPlane is a simple and effective solution for representing 4D volumes, and we hope they can broadly contribute to modeling spacetime for dynamic 3D scenes.11Project page: https://caoang327.github.io/HexPlane."
                },
                {
                    "title": "NeuS2: Fast Learning of Neural Implicit Surfaces for Multi-view Reconstruction",
                    "abstract": "Recent methods for neural surface representation and rendering, for example NeuS [59], have demonstrated the remarkably high-quality reconstruction of static scenes. However, the training of NeuS takes an extremely long time (8 hours), which makes it almost impossible to apply them to dynamic scenes with thousands of frames. We propose a fast neural surface reconstruction approach, called NeuS2, which achieves two orders of magnitude improvement in terms of acceleration without compromising reconstruction quality. To accelerate the training process, we parameterize a neural surface representation by multi-resolution hash encodings and present a novel lightweight calculation of second-order derivatives tailored to our networks to leverage CUDA parallelism, achieving a factor two speed up. To further stabilize and expedite training, a progressive learning strategy is proposed to optimize multi-resolution hash encodings from coarse to fine. We extend our method for fast training of dynamic scenes, with a proposed incremental training strategy and a novel global transformation prediction component, which allow our method to handle challenging long sequences with large movements and deformations. Our experiments on various datasets demonstrate that NeuS2 significantly outperforms the state-of-the-arts in both surface reconstruction accuracy and training speed for both static and dynamic scenes. The code is available at our website: https://vcai.mpi-inf.mpg.de/projects/NeuS2/."
                },
                {
                    "title": "Tensor4D: Efficient Neural 4D Decomposition for High-Fidelity Dynamic Reconstruction and Rendering",
                    "abstract": "We present Tensor4D, an efficient yet effective approach to dynamic scene modeling. The key of our solution is an efficient 4D tensor decomposition method so that the dynamic scene can be directly represented as a 4D spatio-temporal tensor. To tackle the accompanying memory issue, we decompose the 4D tensor hierarchically by projecting it first into three time-aware volumes and then nine compact feature planes. In this way, spatial information over time can be simultaneously captured in a compact and memory-efficient manner. When applying Tensor4D for dynamic scene reconstruction and rendering, we further factorize the 4D fields to different scales in the sense that structural motions and dynamic detailed changes can be learned from coarse to fine. The effectiveness of our method is validated on both synthetic and real-world scenes. Extensive experiments show that our method is able to achieve high-quality dynamic reconstruction and rendering from sparse-view camera rigs or even a monocular camera. The code and dataset will be released at https://github.com/DSaurus/Tensor4D."
                },
                {
                    "title": "DynIBaR: Neural Dynamic Image-Based Rendering",
                    "abstract": "We address the problem of synthesizing novel views from a monocular video depicting a complex dynamic scene. State-of-the-art methods based on temporally varying Neural Radiance Fields (aka dynamic NeRFs) have shown impressive results on this task. However, for long videos with complex object motions and uncontrolled camera trajectories, these methods can produce blurry or inaccurate renderings, hampering their use in real-world applications. Instead of encoding the entire dynamic scene within the weights of MLPs, we present a new approach that addresses these limitations by adopting a volumetric image-based rendering framework that synthesizes new viewpoints by aggregating features from nearby views in a scene motion-aware manner. Our system retains the advantages of prior methods in its ability to model complex scenes and view-dependent effects, but also enables synthesizing photo-realistic novel views from long videos featuring complex scene dynamics with unconstrained camera trajectories. We demonstrate significant improvements over state-of-the-art methods on dynamic scene datasets, and also apply our approach to in-the-wild videos with challenging camera and object motion, where prior methods fail to produce high-quality renderings."
                },
                {
                    "title": "Neural Surface Reconstruction of Dynamic Scenes with Monocular RGB-D Camera",
                    "abstract": "We propose Neural-DynamicReconstruction (NDR), a template-free method to recover high-fidelity geometry and motions of a dynamic scene from a monocular RGB-D camera. In NDR, we adopt the neural implicit function for surface representation and rendering such that the captured color and depth can be fully utilized to jointly optimize the surface and deformations. To represent and constrain the non-rigid deformations, we propose a novel neural invertible deforming network such that the cycle consistency between arbitrary two frames is automatically satisfied. Considering that the surface topology of dynamic scene might change over time, we employ a topology-aware strategy to construct the topology-variant correspondence for the fused frames. NDR also further refines the camera poses in a global optimization manner. Experiments on public datasets and our collected dataset demonstrate that NDR outperforms existing monocular dynamic reconstruction methods."
                },
                {
                    "title": "SparseNeuS: Fast Generalizable Neural Surface Reconstruction from Sparse views",
                    "abstract": "We introduce SparseNeuS, a novel neural rendering based method for the task of surface reconstruction from multi-view images. This task becomes more difficult when only sparse images are provided as input, a scenario where existing neural reconstruction approaches usually produce incomplete or distorted results. Moreover, their inability of generalizing to unseen new scenes impedes their application in practice. Contrarily, SparseNeuS can generalize to new scenes and work well with sparse images (as few as 2 or 3). SparseNeuS adopts signed distance function (SDF) as the surface representation, and learns generalizable priors from image features by introducing geometry encoding volumes for generic surface prediction. Moreover, several strategies are introduced to effectively leverage sparse views for high-quality reconstruction, including 1) a multi-level geometry reasoning framework to recover the surfaces in a coarse-to-fine manner; 2) a multi-scale color blending scheme for more reliable color prediction; 3) a consistency-aware fine-tuning scheme to control the inconsistent regions caused by occlusion and noise. Extensive experiments demonstrate that our approach not only outperforms the state-of-the-art methods, but also exhibits good efficiency, generalizability, and flexibility."
                },
                {
                    "title": "Residual Multiplicative Filter Networks for Multiscale Reconstruction",
                    "abstract": "Coordinate networks like Multiplicative Filter Networks (MFNs) and BACON offer some control over the frequency spectrum used to represent continuous signals such as images or 3D volumes. Yet, they are not readily applicable to problems for which coarse-to-fine estimation is required, including various inverse problems in which coarse-to-fine optimization plays a key role in avoiding poor local minima. We introduce a new coordinate network architecture and training scheme that enables coarse-to-fine optimization with fine-grained control over the frequency support of learned reconstructions. This is achieved with two key innovations. First, we incorporate skip connections so that structure at one scale is preserved when fitting finer-scale structure. Second, we propose a novel initialization scheme to provide control over the model frequency spectrum at each stage of optimization. We demonstrate how these modifications enable multiscale optimization for coarse-to-fine fitting to natural images. We then evaluate our model on synthetically generated datasets for the the problem of single-particle cryo-EM reconstruction. We learn high resolution multiscale structures, on par with the state-of-the art."
                },
                {
                    "title": "TensoRF: Tensorial Radiance Fields",
                    "abstract": "We present TensoRF, a novel approach to model and reconstruct radiance fields. Unlike NeRF that purely uses MLPs, we model the radiance field of a scene as a 4D tensor, which represents a 3D voxel grid with per-voxel multi-channel features. Our central idea is to factorize the 4D scene tensor into multiple compact low-rank tensor components. We demonstrate that applying traditional CP decomposition -- that factorizes tensors into rank-one components with compact vectors -- in our framework leads to improvements over vanilla NeRF. To further boost performance, we introduce a novel vector-matrix (VM) decomposition that relaxes the low-rank constraints for two modes of a tensor and factorizes tensors into compact vector and matrix factors. Beyond superior rendering quality, our models with CP and VM decompositions lead to a significantly lower memory footprint in comparison to previous and concurrent works that directly optimize per-voxel features. Experimentally, we demonstrate that TensoRF with CP decomposition achieves fast reconstruction (<30 min) with better rendering quality and even a smaller model size (<4 MB) compared to NeRF. Moreover, TensoRF with VM decomposition further boosts rendering quality and outperforms previous state-of-the-art methods, while reducing the reconstruction time (<10 min) and retaining a compact model size (<75 MB)."
                },
                {
                    "title": "Instant neural graphics primitives with a multiresolution hash encoding",
                    "abstract": "Neural graphics primitives, parameterized by fully connected neural networks, can be costly to train and evaluate. We reduce this cost with a versatile new input encoding that permits the use of a smaller network without sacrificing quality, thus significantly reducing the number of floating point and memory access operations: a small neural network is augmented by a multiresolution hash table of trainable feature vectors whose values are optimized through stochastic gradient descent. The multiresolution structure allows the network to disambiguate hash collisions, making for a simple architecture that is trivial to parallelize on modern GPUs. We leverage this parallelism by implementing the whole system using fully-fused CUDA kernels with a focus on minimizing wasted bandwidth and compute operations. We achieve a combined speedup of several orders of magnitude, enabling training of high-quality neural graphics primitives in a matter of seconds, and rendering in tens of milliseconds at a resolution of 1920\u00d71080."
                },
                {
                    "title": "Efficient Geometry-aware 3D Generative Adversarial Networks",
                    "abstract": "Unsupervised generation of high-quality multi-view-consistent images and 3D shapes using only collections of single-view 2D photographs has been a long-standing challenge. Existing 3D GANs are either compute intensive or make approximations that are not 3D-consistent; the former limits quality and resolution of the generated images and the latter adversely affects multi-view consistency and shape quality. In this work, we improve the computational efficiency and image quality of 3D GANs without overly relying on these approximations. We introduce an expressive hybrid explicit implicit network architecture that, together with other design choices, synthesizes not only high-resolution multi-view-consistent images in real time but also produces high-quality 3D geometry. By decoupling feature generation and neural rendering, our framework is able to leverage state-of-the-art 2D CNN generators, such as StyleGAN2, and inherit their efficiency and expressiveness. We demonstrate state-of-the-art 3D-aware synthesis with FFHQ and AFHQ Cats, among other experiments."
                },
                {
                    "title": "Bacon: Band-limited Coordinate Networks for Multiscale Scene Representation",
                    "abstract": "Coordinate-based networks have emerged as a powerful tool for 3D representation and scene reconstruction. These networks are trained to map continuous input coordinates to the value of a signal at each point. Still, current architectures are black boxes: their spectral characteristics cannot be easily analyzed, and their behavior at unsupervised points is difficult to predict. Moreover, these networks are typically trained to represent a signal at a single scale, so naive downsampling or upsampling results in artifacts. We introduce band-limited coordinate networks (BACON), a network architecture with an analytical Fourier spectrum. Bacon has constrained behavior at unsupervised points, can be designed based on the spectral characteristics of the represented signal, and can represent signals at multiple scales without per-scale supervision. We demonstrate Bacon for multiscale neural representation of images, radiance fields, and 3D scenes using signed distance functions and show that it outperforms conventional single-scale coordinate networks in terms of interpretability and quality."
                },
                {
                    "title": "Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields",
                    "abstract": "Though neural radiance fields (NeRF) have demon-strated impressive view synthesis results on objects and small bounded regions of space, they struggle on \u201cun-bounded\u201d scenes, where the camera may point in any di-rection and content may exist at any distance. In this set-ting, existing NeRF-like models often produce blurry or low-resolution renderings (due to the unbalanced detail and scale of nearby and distant objects), are slow to train, and may exhibit artifacts due to the inherent ambiguity of the task of reconstructing a large scene from a small set of images. We present an extension of mip-NeRF (a NeRF variant that addresses sampling and aliasing) that uses a non-linear scene parameterization, online distillation, and a novel distortion-based regularizer to overcome the chal-lenges presented by unbounded scenes. Our model, which we dub \u201cmip-NeRF 360\u201d as we target scenes in which the camera rotates 360 degrees around a point, reduces mean-squared error by 57% compared to mip-NeRF, and is able to produce realistic synthesized views and detailed depth maps for highly intricate, unbounded real-world scenes."
                },
                {
                    "title": "Neural Fields in Visual Computing and Beyond",
                    "abstract": "Recent advances in machine learning have led to increased interest in solving visual computing problems using methods that employ coordinate\u2010based neural networks. These methods, which we call neural fields, parameterize physical properties of scenes or objects across space and time. They have seen widespread success in problems such as 3D shape and image synthesis, animation of human bodies, 3D reconstruction, and pose estimation. Rapid progress has led to numerous papers, but a consolidation of the discovered knowledge has not yet emerged. We provide context, mathematical grounding, and a review of over 250 papers in the literature on neural fields. In Part I, we focus on neural field techniques by identifying common components of neural field methods, including different conditioning, representation, forward map, architecture, and manipulation methods. In Part II, we focus on applications of neural fields to different problems in visual computing, and beyond (e.g., robotics, audio). Our review shows the breadth of topics already covered in visual computing, both historically and in current incarnations, and highlights the improved quality, flexibility, and capability brought by neural field methods. Finally, we present a companion website that acts as a living database that can be continually updated by the community."
                },
                {
                    "title": "The Power of Points for Modeling Humans in Clothing",
                    "abstract": "Currently it requires an artist to create 3D human avatars with realistic clothing that can move naturally. Despite progress on 3D scanning and modeling of human bodies, there is still no technology that can easily turn a static scan into an animatable avatar. Automating the creation of such avatars would enable many applications in games, social networking, animation, and AR/VR to name a few. The key problem is one of representation. Standard 3D meshes are widely used in modeling the minimally-clothed body but do not readily capture the complex topology of clothing. Recent interest has shifted to implicit surface models for this task but they are computationally heavy and lack compatibility with existing 3D tools. What is needed is a 3D representation that can capture varied topology at high resolution and that can be learned from data. We argue that this representation has been with us all along \u2014 the point cloud. Point clouds have properties of both implicit and explicit representations that we exploit to model 3D garment geometry on a human body. We train a neural network with a novel local clothing geometric feature to represent the shape of different outfits. The network is trained from 3D point clouds of many types of clothing, on many bodies, in many poses, and learns to model pose-dependent clothing deformations. The geometry feature can be optimized to fit a previously unseen scan of a person in clothing, enabling the scan to be reposed realistically. Our model demonstrates superior quantitative and qualitative results in both multi-outfit modeling and unseen outfit animation. The code is available for research purposes at https://qianlim.github.io/POP."
                },
                {
                    "title": "Advances in Neural Rendering",
                    "abstract": "Synthesizing photo\u2010realistic images and videos is at the heart of computer graphics and has been the focus of decades of research. Traditionally, synthetic images of a scene are generated using rendering algorithms such as rasterization or ray tracing, which take specifically defined representations of geometry and material properties as input. Collectively, these inputs define the actual scene and what is rendered, and are referred to as the scene representation (where a scene consists of one or more objects). Example scene representations are triangle meshes with accompanied textures (e.g., created by an artist), point clouds (e.g., from a depth sensor), volumetric grids (e.g., from a CT scan), or implicit surface functions (e.g., truncated signed distance fields). The reconstruction of such a scene representation from observations using differentiable rendering losses is known as inverse graphics or inverse rendering. Neural rendering is closely related, and combines ideas from classical computer graphics and machine learning to create algorithms for synthesizing images from real\u2010world observations. Neural rendering is a leap forward towards the goal of synthesizing photo\u2010realistic image and video content. In recent years, we have seen immense progress in this field through hundreds of publications that show different ways to inject learnable components into the rendering pipeline. This state\u2010of\u2010the\u2010art report on advances in neural rendering focuses on methods that combine classical rendering principles with learned 3D scene representations, often now referred to as neural scene representations. A key advantage of these methods is that they are 3D\u2010consistent by design, enabling applications such as novel viewpoint synthesis of a captured scene. In addition to methods that handle static scenes, we cover neural scene representations for modeling non\u2010rigidly deforming objects and scene editing and composition. While most of these approaches are scene\u2010specific, we also discuss techniques that generalize across object classes and can be used for generative tasks. In addition to reviewing these state\u2010of\u2010the\u2010art methods, we provide an overview of fundamental concepts and definitions used in the current literature. We conclude with a discussion on open challenges and social implications."
                },
                {
                    "title": "Volume Rendering of Neural Implicit Surfaces",
                    "abstract": "Neural volume rendering became increasingly popular recently due to its success in synthesizing novel views of a scene from a sparse set of input images. So far, the geometry learned by neural volume rendering techniques was modeled using a generic density function. Furthermore, the geometry itself was extracted using an arbitrary level set of the density function leading to a noisy, often low fidelity reconstruction. The goal of this paper is to improve geometry representation and reconstruction in neural volume rendering. We achieve that by modeling the volume density as a function of the geometry. This is in contrast to previous work modeling the geometry as a function of the volume density. In more detail, we define the volume density function as Laplace's cumulative distribution function (CDF) applied to a signed distance function (SDF) representation. This simple density representation has three benefits: (i) it provides a useful inductive bias to the geometry learned in the neural volume rendering process; (ii) it facilitates a bound on the opacity approximation error, leading to an accurate sampling of the viewing ray. Accurate sampling is important to provide a precise coupling of geometry and radiance; and (iii) it allows efficient unsupervised disentanglement of shape and appearance in volume rendering. Applying this new density representation to challenging scene multiview datasets produced high quality geometry reconstructions, outperforming relevant baselines. Furthermore, switching shape and appearance between scenes is possible due to the disentanglement of the two."
                },
                {
                    "title": "MetaAvatar: Learning Animatable Clothed Human Models from Few Depth Images",
                    "abstract": "In this paper, we aim to create generalizable and controllable neural signed distance fields (SDFs) that represent clothed humans from monocular depth observations. Recent advances in deep learning, especially neural implicit representations, have enabled human shape reconstruction and controllable avatar generation from different sensor inputs. However, to generate realistic cloth deformations from novel input poses, watertight meshes or dense full-body scans are usually needed as inputs. Furthermore, due to the difficulty of effectively modeling pose-dependent cloth deformations for diverse body shapes and cloth types, existing approaches resort to per-subject/cloth-type optimization from scratch, which is computationally expensive. In contrast, we propose an approach that can quickly generate realistic clothed human avatars, represented as controllable neural SDFs, given only monocular depth images. We achieve this by using meta-learning to learn an initialization of a hypernetwork that predicts the parameters of neural SDFs. The hypernetwork is conditioned on human poses and represents a clothed neural avatar that deforms non-rigidly according to the input poses. Meanwhile, it is meta-learned to effectively incorporate priors of diverse body shapes and cloth types and thus can be much faster to fine-tune, compared to models trained from scratch. We qualitatively and quantitatively show that our approach outperforms state-of-the-art approaches that require complete meshes as inputs while our approach requires only depth frames as inputs and runs orders of magnitudes faster. Furthermore, we demonstrate that our meta-learned hypernetwork is very robust, being the first to generate avatars with realistic dynamic cloth deformations given as few as 8 monocular depth frames."
                },
                {
                    "title": "NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstruction",
                    "abstract": "We present a novel neural surface reconstruction method, called NeuS, for reconstructing objects and scenes with high fidelity from 2D image inputs. Existing neural surface reconstruction approaches, such as DVR and IDR, require foreground mask as supervision, easily get trapped in local minima, and therefore struggle with the reconstruction of objects with severe self-occlusion or thin structures. Meanwhile, recent neural methods for novel view synthesis, such as NeRF and its variants, use volume rendering to produce a neural scene representation with robustness of optimization, even for highly complex objects. However, extracting high-quality surfaces from this learned implicit representation is difficult because there are not sufficient surface constraints in the representation. In NeuS, we propose to represent a surface as the zero-level set of a signed distance function (SDF) and develop a new volume rendering method to train a neural SDF representation. We observe that the conventional volume rendering method causes inherent geometric errors (i.e. bias) for surface reconstruction, and therefore propose a new formulation that is free of bias in the first order of approximation, thus leading to more accurate surface reconstruction even without the mask supervision. Experiments on the DTU dataset and the BlendedMVS dataset show that NeuS outperforms the state-of-the-arts in high-quality surface reconstruction, especially for objects and scenes with complex structures and self-occlusion."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Compacter: Efficient Low-Rank Hypercomplex Adapter Layers",
                    "abstract": "Adapting large-scale pretrained language models to downstream tasks via fine-tuning is the standard method for achieving state-of-the-art performance on NLP benchmarks. However, fine-tuning all weights of models with millions or billions of parameters is sample-inefficient, unstable in low-resource settings, and wasteful as it requires storing a separate copy of the model for each task. Recent work has developed parameter-efficient fine-tuning methods, but these approaches either still require a relatively large number of parameters or underperform standard fine-tuning. In this work, we propose Compacter, a method for fine-tuning large-scale language models with a better trade-off between task performance and the number of trainable parameters than prior work. Compacter accomplishes this by building on top of ideas from adapters, low-rank optimization, and parameterized hypercomplex multiplication layers. Specifically, Compacter inserts task-specific weight matrices into a pretrained model's weights, which are computed efficiently as a sum of Kronecker products between shared\"slow\"weights and\"fast\"rank-one matrices defined per Compacter layer. By only training 0.047% of a pretrained model's parameters, Compacter performs on par with standard fine-tuning on GLUE and outperforms standard fine-tuning on SuperGLUE and low-resource settings. Our code is publicly available at~\\url{https://github.com/rabeehk/compacter}."
                },
                {
                    "title": "Neural Trajectory Fields for Dynamic Novel View Synthesis",
                    "abstract": "Recent approaches to render photorealistic views from a limited set of photographs have pushed the boundaries of our interactions with pictures of static scenes. The ability to recreate moments, that is, time-varying sequences, is perhaps an even more interesting scenario, but it remains largely unsolved. We introduce DCT-NeRF, a coordinatebased neural representation for dynamic scenes. DCTNeRF learns smooth and stable trajectories over the input sequence for each point in space. This allows us to enforce consistency between any two frames in the sequence, which results in high quality reconstruction, particularly in dynamic regions."
                },
                {
                    "title": "4DComplete: Non-Rigid Motion Estimation Beyond the Observable Surface",
                    "abstract": "Tracking non-rigidly deforming scenes using range sensors has numerous applications including computer vision, AR/VR, and robotics. However, due to occlusions and physical limitations of range sensors, existing methods only handle the visible surface, thus causing discontinuities and in-completeness in the motion field. To this end, we introduce 4DComplete, a novel data-driven approach that estimates the non-rigid motion for the unobserved geometry. 4DComplete takes as input a partial shape and motion observation, extracts 4D time-space embedding, and jointly infers the missing geometry and motion field using a sparse fully-convolutional network. For network training, we constructed a large-scale synthetic dataset called DeformingThings4D, which consists of 1,972 animation sequences spanning 31 different animals or humanoid categories with dense 4D annotation. Experiments show that 4DComplete 1) reconstructs high-resolution volumetric shape and motion field from a partial observation, 2) learns an entangled 4D feature representation that benefits both shape and motion estimation, 3) yields more accurate and natural deformation than classic non-rigid priors such as As-Rigid-As-Possible (ARAP) deformation, and 4) generalizes well to unseen objects in real-world sequences."
                },
                {
                    "title": "Initialization and Regularization of Factorized Neural Layers",
                    "abstract": "Factorized layers--operations parameterized by products of two or more matrices--occur in a variety of deep learning contexts, including compressed model training, certain types of knowledge distillation, and multi-head self-attention architectures. We study how to initialize and regularize deep nets containing such layers, examining two simple, understudied schemes, spectral initialization and Frobenius decay, for improving their performance. The guiding insight is to design optimization routines for these networks that are as close as possible to that of their well-tuned, non-decomposed counterparts; we back this intuition with an analysis of how the initialization and regularization schemes impact training with gradient descent, drawing on modern attempts to understand the interplay of weight-decay and batch-normalization. Empirically, we highlight the benefits of spectral initialization and Frobenius decay across a variety of settings. In model compression, we show that they enable low-rank methods to significantly outperform both unstructured sparsity and tensor methods on the task of training low-memory residual networks; analogs of the schemes also improve the performance of tensor decomposition techniques. For knowledge distillation, Frobenius decay enables a simple, overcomplete baseline that yields a compact model from over-parameterized training without requiring retraining with or pruning a teacher network. Finally, we show how both schemes applied to multi-head attention lead to improved performance on both translation and unsupervised pre-training."
                },
                {
                    "title": "Modulated Periodic Activations for Generalizable Local Functional Representations",
                    "abstract": "Multi-Layer Perceptrons (MLPs) make powerful functional representations for sampling and reconstruction problems involving low-dimensional signals like images, shapes and light fields. Recent works have significantly improved their ability to represent high-frequency content by using periodic activations or positional encodings. This often came at the expense of generalization: modern methods are typically optimized for a single signal. We present a new representation that generalizes to multiple instances and achieves state-of-the-art fidelity. We use a dual-MLP architecture to encode the signals. A synthesis network creates a functional mapping from a low-dimensional input (e.g. pixel-position) to the output domain (e.g. RGB color). A modulation network maps a latent code corresponding to the target signal to parameters that modulate the periodic activations of the synthesis network. We also propose a local-functional representation which enables generalization. The signal\u2019s domain is partitioned into a regular grid, with each tile represented by a latent code. At test time, the signal is encoded with high-fidelity by inferring (or directly optimizing) the latent code-book. Our approach produces generalizable functional representations of images, videos and shapes, and achieves higher reconstruction quality than prior works that are optimized for a single signal."
                },
                {
                    "title": "KiloNeRF: Speeding up Neural Radiance Fields with Thousands of Tiny MLPs",
                    "abstract": "NeRF synthesizes novel views of a scene with unprecedented quality by fitting a neural radiance field to RGB images. However, NeRF requires querying a deep Multi-Layer Perceptron (MLP) millions of times, leading to slow rendering times, even on modern GPUs. In this paper, we demonstrate that real-time rendering is possible by utilizing thousands of tiny MLPs instead of one single large MLP. In our setting, each individual MLP only needs to represent parts of the scene, thus smaller and faster-to-evaluate MLPs can be used. By combining this divide-and-conquer strategy with further optimizations, rendering is accelerated by three orders of magnitude compared to the original NeRF model without incurring high storage costs. Further, using teacher-student distillation for training, we show that this speed-up can be achieved without sacrificing visual quality."
                },
                {
                    "title": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields",
                    "abstract": "The rendering procedure used by neural radiance fields (NeRF) samples a scene with a single ray per pixel and may therefore produce renderings that are excessively blurred or aliased when training or testing images observe scene content at different resolutions. The straightforward solution of supersampling by rendering with multiple rays per pixel is impractical for NeRF, because rendering each ray requires querying a multilayer perceptron hundreds of times. Our solution, which we call \"mip-NeRF\" (\u00e0 la \"mipmap\"), extends NeRF to represent the scene at a continuously-valued scale. By efficiently rendering anti-aliased conical frustums instead of rays, mip-NeRF reduces objectionable aliasing artifacts and significantly improves NeRF\u2019s ability to represent fine details, while also being 7% faster than NeRF and half the size. Compared to NeRF, mip-NeRF reduces average error rates by 17% on the dataset presented with NeRF and by 60% on a challenging multiscale variant of that dataset that we present. Mip-NeRF is also able to match the accuracy of a brute-force supersampled NeRF on our multiscale dataset while being 22\u00d7 faster."
                },
                {
                    "title": "Neural Geometric Level of Detail: Real-time Rendering with Implicit 3D Shapes",
                    "abstract": "Neural signed distance functions (SDFs) are emerging as an effective representation for 3D shapes. State-of-the-art methods typically encode the SDF with a large, fixed-size neural network to approximate complex shapes with implicit surfaces. Rendering with these large networks is, however, computationally expensive since it requires many forward passes through the network for every pixel, making these representations impractical for real-time graphics. We introduce an efficient neural representation that, for the first time, enables real-time rendering of high-fidelity neural SDFs, while achieving state-of-the-art geometry reconstruction quality. We represent implicit surfaces using an octree-based feature volume which adaptively fits shapes with multiple discrete levels of detail (LODs), and enables continuous LOD with SDF interpolation. We further develop an efficient algorithm to directly render our novel neural SDF representation in real-time by querying only the necessary LODs with sparse octree traversal. We show that our representation is 2\u20133 orders of magnitude more efficient in terms of rendering speed compared to previous works. Furthermore, it produces state-of-the-art reconstruction quality for complex shapes under both 3D geometric and 2D image-space metrics."
                },
                {
                    "title": "Learned Initializations for Optimizing Coordinate-Based Neural Representations",
                    "abstract": "Coordinate-based neural representations have shown significant promise as an alternative to discrete, array-based representations for complex low dimensional signals. However, optimizing a coordinate-based network from randomly initialized weights for each new signal is inefficient. We propose applying standard meta-learning algorithms to learn the initial weight parameters for these fully-connected networks based on the underlying class of signals being represented (e.g., images of faces or 3D models of chairs). Despite requiring only a minor change in implementation, using these learned initial weights enables faster convergence during optimization and can serve as a strong prior over the signal class being modeled, resulting in better generalization when only partial observations of a given signal are available. We explore these benefits across a variety of tasks, including representing 2D images, reconstructing CT scans, and recovering 3D shapes and scenes from 2D image observations."
                },
                {
                    "title": "D-NeRF: Neural Radiance Fields for Dynamic Scenes",
                    "abstract": "Neural rendering techniques combining machine learning with geometric reasoning have arisen as one of the most promising approaches for synthesizing novel views of a scene from a sparse set of images. Among these, stands out the Neural radiance fields (NeRF) [31], which trains a deep network to map 5D input coordinates (representing spatial location and viewing direction) into a volume density and view-dependent emitted radiance. However, despite achieving an unprecedented level of photorealism on the generated images, NeRF is only applicable to static scenes, where the same spatial location can be queried from different images. In this paper we introduce D-NeRF, a method that extends neural radiance fields to a dynamic domain, allowing to reconstruct and render novel images of objects under rigid and non-rigid motions from a single camera moving around the scene. For this purpose we consider time as an additional input to the system, and split the learning process in two main stages: one that encodes the scene into a canonical space and another that maps this canonical representation into the deformed scene at a particular time. Both mappings are simultaneously learned using fully-connected networks. Once the networks are trained, D-NeRF can render novel images, controlling both the camera view and the time variable, and thus, the object movement. We demonstrate the effectiveness of our approach on scenes with objects under rigid, articulated and non-rigid motions. Code, model weights and the dynamic scenes dataset will be available at [1]."
                },
                {
                    "title": "Neural Scene Flow Fields for Space-Time View Synthesis of Dynamic Scenes",
                    "abstract": "We present a method to perform novel view and time synthesis of dynamic scenes, requiring only a monocular video with known camera poses as input. To do this, we introduce Neural Scene Flow Fields, a new representation that models the dynamic scene as a time-variant continuous function of appearance, geometry, and 3D scene motion. Our representation is optimized through a neural network to fit the observed input views. We show that our representation can be used for varieties of in-the-wild scenes, including thin structures, view-dependent effects, and complex degrees of motion. We conduct a number of experiments that demonstrate our approach significantly outperforms recent monocular view synthesis methods, and show qualitative results of space-time view synthesis on a variety of real-world videos."
                },
                {
                    "title": "Nerfies: Deformable Neural Radiance Fields",
                    "abstract": "We present the first method capable of photorealistically reconstructing deformable scenes using photos/videos captured casually from mobile phones. Our approach augments neural radiance fields (NeRF) by optimizing an additional continuous volumetric deformation field that warps each observed point into a canonical 5D NeRF. We observe that these NeRF-like deformation fields are prone to local minima, and propose a coarse-to-fine optimization method for coordinate-based models that allows for more robust optimization. By adapting principles from geometry processing and physical simulation to NeRF-like models, we propose an elastic regularization of the deformation field that further improves robustness. We show that our method can turn casually captured selfie photos/videos into deformable NeRF models that allow for photorealistic renderings of the subject from arbitrary viewpoints, which we dub \"nerfies.\" We evaluate our method by collecting time-synchronized data using a rig with two mobile phones, yielding train/validation images of the same pose at different viewpoints. We show that our method faithfully reconstructs non-rigidly deforming scenes and reproduces unseen views with high fidelity."
                },
                {
                    "title": "Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains",
                    "abstract": "We show that passing input points through a simple Fourier feature mapping enables a multilayer perceptron (MLP) to learn high-frequency functions in low-dimensional problem domains. These results shed light on recent advances in computer vision and graphics that achieve state-of-the-art results by using MLPs to represent complex 3D objects and scenes. Using tools from the neural tangent kernel (NTK) literature, we show that a standard MLP fails to learn high frequencies both in theory and in practice. To overcome this spectral bias, we use a Fourier feature mapping to transform the effective NTK into a stationary kernel with a tunable bandwidth. We suggest an approach for selecting problem-specific Fourier features that greatly improves the performance of MLPs for low-dimensional regression tasks relevant to the computer vision and graphics communities."
                },
                {
                    "title": "MetaSDF: Meta-learning Signed Distance Functions",
                    "abstract": "Neural implicit shape representations are an emerging paradigm that offers many potential benefits over conventional discrete representations, including memory efficiency at a high spatial resolution. Generalizing across shapes with such neural implicit representations amounts to learning priors over the respective function space and enables geometry reconstruction from partial or noisy observations. Existing generalization methods rely on conditioning a neural network on a low-dimensional latent code that is either regressed by an encoder or jointly optimized in the auto-decoder framework. Here, we formalize learning of a shape space as a meta-learning problem and leverage gradient-based meta-learning algorithms to solve this task. We demonstrate that this approach performs on par with auto-decoder based approaches while being an order of magnitude faster at test-time inference. We further demonstrate that the proposed gradient-based method outperforms encoder-decoder based methods that leverage pooling-based set encoders."
                },
                {
                    "title": "Implicit Neural Representations with Periodic Activation Functions",
                    "abstract": "Implicitly defined, continuous, differentiable signal representations parameterized by neural networks have emerged as a powerful paradigm, offering many possible benefits over conventional representations. However, current network architectures for such implicit neural representations are incapable of modeling signals with fine detail, and fail to represent a signal's spatial and temporal derivatives, despite the fact that these are essential to many physical signals defined implicitly as the solution to partial differential equations. We propose to leverage periodic activation functions for implicit neural representations and demonstrate that these networks, dubbed sinusoidal representation networks or Sirens, are ideally suited for representing complex natural signals and their derivatives. We analyze Siren activation statistics to propose a principled initialization scheme and demonstrate the representation of images, wavefields, video, sound, and their derivatives. Further, we show how Sirens can be leveraged to solve challenging boundary value problems, such as particular Eikonal equations (yielding signed distance functions), the Poisson equation, and the Helmholtz and wave equations. Lastly, we combine Sirens with hypernetworks to learn priors over the space of Siren functions."
                },
                {
                    "title": "NeRF",
                    "abstract": "We present a method that achieves state-of-the-art results for synthesizing novel views of complex scenes by optimizing an underlying continuous volumetric scene function using a sparse set of input views. Our algorithm represents a scene using a fully connected (nonconvolutional) deep network, whose input is a single continuous 5D coordinate (spatial location (x, y, z) and viewing direction (\u03b8, \u03d5)) and whose output is the volume density and view-dependent emitted radiance at that spatial location. We synthesize views by querying 5D coordinates along camera rays and use classic volume rendering techniques to project the output colors and densities into an image. Because volume rendering is naturally differentiable, the only input required to optimize our representation is a set of images with known camera poses. We describe how to effectively optimize neural radiance fields to render photorealistic novel views of scenes with complicated geometry and appearance, and demonstrate results that outperform prior work on neural rendering and view synthesis."
                },
                {
                    "title": "Implicit Geometric Regularization for Learning Shapes",
                    "abstract": "Representing shapes as level sets of neural networks has been recently proved to be useful for different shape analysis and reconstruction tasks. So far, such representations were computed using either: (i) pre-computed implicit shape representations; or (ii) loss functions explicitly defined over the neural level sets. In this paper we offer a new paradigm for computing high fidelity implicit neural representations directly from raw data (i.e., point clouds, with or without normal information). We observe that a rather simple loss function, encouraging the neural network to vanish on the input point cloud and to have a unit norm gradient, possesses an implicit geometric regularization property that favors smooth and natural zero level set surfaces, avoiding bad zero-loss solutions. We provide a theoretical analysis of this property for the linear case, and show that, in practice, our method leads to state of the art implicit neural representations with higher level-of-details and fidelity compared to previous methods."
                },
                {
                    "title": "Volume Rendering",
                    "abstract": "This paper shows topical research on volume rendering and volume rendering techniques. Different areas of applications are presented and related problems are discussed."
                },
                {
                    "title": "Analyzing and Improving the Image Quality of StyleGAN",
                    "abstract": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign the generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent codes to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably attribute a generated image to a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Reconstructing continuous distributions of 3D protein structure from cryo-EM images",
                    "abstract": "Cryo-electron microscopy (cryo-EM) is a powerful technique for determining the structure of proteins and other macromolecular complexes at near-atomic resolution. In single particle cryo-EM, the central problem is to reconstruct the 3D structure of a macromolecule from $10^{4-7}$ noisy and randomly oriented 2D projection images. However, the imaged protein complexes may exhibit structural variability, which complicates reconstruction and is typically addressed using discrete clustering approaches that fail to capture the full range of protein dynamics. Here, we introduce a novel method for cryo-EM reconstruction that extends naturally to modeling continuous generative factors of structural heterogeneity. This method encodes structures in Fourier space using coordinate-based deep neural networks, and trains these networks from unlabeled 2D cryo-EM images by combining exact inference over image orientation with variational inference for structural heterogeneity. We demonstrate that the proposed method, termed cryoDRGN, can perform ab-initio reconstruction of 3D protein complexes from simulated and real 2D cryo-EM image data. To our knowledge, cryoDRGN is the first neural network-based approach for cryo-EM reconstruction and the first end-to-end method for directly reconstructing continuous ensembles of protein structures from cryo-EM images."
                },
                {
                    "title": "On the Spectral Bias of Neural Networks",
                    "abstract": "Neural networks are known to be a class of highly expressive functions able to fit even random input-output mappings with $100\\%$ accuracy. In this work, we present properties of neural networks that complement this aspect of expressivity. By using tools from Fourier analysis, we show that deep ReLU networks are biased towards low frequency functions, meaning that they cannot have local fluctuations without affecting their global behavior. Intuitively, this property is in line with the observation that over-parameterized networks find simple patterns that generalize across data samples. We also investigate how the shape of the data manifold affects expressivity by showing evidence that learning high frequencies gets \\emph{easier} with increasing manifold complexity, and present a theoretical understanding of this behavior. Finally, we study the robustness of the frequency components with respect to parameter perturbation, to develop the intuition that the parameters must be finely tuned to express high frequency functions."
                },
                {
                    "title": "The Unreasonable Effectiveness of Deep Features as a Perceptual Metric",
                    "abstract": "While it is nearly effortless for humans to quickly assess the perceptual similarity between two images, the underlying processes are thought to be quite complex. Despite this, the most widely used perceptual metrics today, such as PSNR and SSIM, are simple, shallow functions, and fail to account for many nuances of human perception. Recently, the deep learning community has found that features of the VGG network trained on ImageNet classification has been remarkably useful as a training loss for image synthesis. But how perceptual are these so-called \"perceptual losses\"? What elements are critical for their success? To answer these questions, we introduce a new dataset of human perceptual similarity judgments. We systematically evaluate deep features across different architectures and tasks and compare them with classic metrics. We find that deep features outperform all previous metrics by large margins on our dataset. More surprisingly, this result is not restricted to ImageNet-trained VGG features, but holds across different deep architectures and levels of supervision (supervised, self-supervised, or even unsupervised). Our results suggest that perceptual similarity is an emergent property shared across deep visual representations."
                },
                {
                    "title": "NTIRE 2017 Challenge on Single Image Super-Resolution: Dataset and Study",
                    "abstract": "This paper introduces a novel large dataset for example-based single image super-resolution and studies the state-of-the-art as emerged from the NTIRE 2017 challenge. The challenge is the first challenge of its kind, with 6 competitions, hundreds of participants and tens of proposed solutions. Our newly collected DIVerse 2K resolution image dataset (DIV2K) was employed by the challenge. In our study we compare the solutions from the challenge to a set of representative methods from the literature and evaluate them using diverse measures on our proposed DIV2K dataset. Moreover, we conduct a number of experiments and draw conclusions on several topics of interest. We conclude that the NTIRE 2017 challenge pushes the state-of-the-art in single-image super-resolution, reaching the best results to date on the popular Set5, Set14, B100, Urban100 datasets and on our newly proposed DIV2K."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "HyperNetworks",
                    "abstract": "This work explores hypernetworks: an approach of using one network, also known as a hypernetwork, to generate the weights for another network. We apply hypernetworks to generate adaptive weights for recurrent networks. In this case, hypernetworks can be viewed as a relaxed form of weight-sharing across layers. In our implementation, hypernetworks are are trained jointly with the main network in an end-to-end fashion. Our main result is that hypernetworks can generate non-shared weights for LSTM and achieve state-of-the-art results on a variety of sequence modelling tasks including character-level language modelling, handwriting generation and neural machine translation, challenging the weight-sharing paradigm for recurrent networks."
                },
                {
                    "title": "SGDR: Stochastic Gradient Descent with Warm Restarts",
                    "abstract": "Restart techniques are common in gradient-free optimization to deal with multimodal functions. Partial warm restarts are also gaining popularity in gradient-based optimization to improve the rate of convergence in accelerated gradient schemes to deal with ill-conditioned functions. In this paper, we propose a simple warm restart technique for stochastic gradient descent to improve its anytime performance when training deep neural networks. We empirically study its performance on the CIFAR-10 and CIFAR-100 datasets, where we demonstrate new state-of-the-art results at 3.14% and 16.21%, respectively. We also demonstrate its advantages on a dataset of EEG recordings and on a downsampled version of the ImageNet dataset. Our source code is available at this https URL"
                },
                {
                    "title": "Training CNNs with Low-Rank Filters for Efficient Image Classification",
                    "abstract": "We propose a new method for creating computationally efficient convolutional neural networks (CNNs) by using low-rank representations of convolutional filters. Rather than approximating filters in previously-trained networks with more efficient versions, we learn a set of small basis filters from scratch; during training, the network learns to combine these basis filters into more complex filters that are discriminative for image classification. To train such networks, a novel weight initialization scheme is used. This allows effective initialization of connection weights in convolutional layers composed of groups of differently-shaped filters. We validate our approach by applying it to several existing CNN architectures and training these networks from scratch using the CIFAR, ILSVRC and MIT Places datasets. Our results show similar or higher accuracy than conventional CNNs with much less compute. Applying our method to an improved version of VGG-11 network using global max-pooling, we achieve comparable validation accuracy using 41% less compute and only 24% of the original VGG-11 model parameters; another variant of our method gives a 1 percentage point increase in accuracy over our improved VGG-11 model, giving a top-5 center-crop validation accuracy of 89.7% while reducing computation by 16% relative to the original VGG-11 model. Applying our method to the GoogLeNet architecture for ILSVRC, we achieved comparable accuracy with 26% less compute and 41% fewer model parameters. Applying our method to a near state-of-the-art network for CIFAR, we achieved comparable accuracy with 46% less compute and 55% fewer parameters."
                },
                {
                    "title": "Highway Networks",
                    "abstract": "There is plenty of theoretical and empirical evidence that depth of neural networks is a crucial ingredient for their success. However, network training becomes more difficult with increasing depth and training of very deep networks remains an open problem. In this extended abstract, we introduce a new architecture designed to ease gradient-based training of very deep networks. We refer to networks with this architecture as highway networks, since they allow unimpeded information flow across several layers on\"information highways\". The architecture is characterized by the use of gating units which learn to regulate the flow of information through a network. Highway networks with hundreds of layers can be trained directly using stochastic gradient descent and with a variety of activation functions, opening up the possibility of studying extremely deep and efficient architectures."
                },
                {
                    "title": "Adam: A Method for Stochastic Optimization",
                    "abstract": "We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
                },
                {
                    "title": "In Search of the Real Inductive Bias: On the Role of Implicit Regularization in Deep Learning",
                    "abstract": "We present experiments demonstrating that some other form of capacity control, different from network size, plays a central role in learning multilayer feed-forward networks. We argue, partially through analogy to matrix factorization, that this is an inductive bias that can help shed light on deep learning."
                },
                {
                    "title": "Predicting Parameters in Deep Learning",
                    "abstract": "We demonstrate that there is significant redundancy in the parameterization of several deep learning models. Given only a few weight values for each feature it is possible to accurately predict the remaining values. Moreover, we show that not only can the parameter values be predicted, but many of them need not be learned at all. We train several different architectures by learning only a small number of weights and predicting the rest. In the best case we are able to predict more than 95% of the weights of a network without any drop in accuracy."
                },
                {
                    "title": "Understanding the difficulty of training deep feedforward neural networks",
                    "abstract": "Whereas before 2006 it appears that deep multilayer neural networks were not successfully trained, since then several algorithms have been shown to successfully train them, with experimental results showing the superiority of deeper vs less deep architectures. All these experimental results were obtained with new initialization or training mechanisms. Our objective here is to understand better why standard gradient descent from random initialization is doing so poorly with deep neural networks, to better understand these recent relative successes and help design better algorithms in the future. We first observe the influence of the non-linear activations functions. We find that the logistic sigmoid activation is unsuited for deep networks with random initialization because of its mean value, which can drive especially the top hidden layer into saturation. Surprisingly, we find that saturated units can move out of saturation by themselves, albeit slowly, and explaining the plateaus sometimes seen when training neural networks. We find that a new non-linearity that saturates less can often be beneficial. Finally, we study how activations and gradients vary across layers and during training, with the idea that training may be more difficult when the singular values of the Jacobian associated with each layer are far from 1. Based on these considerations, we propose a new initialization scheme that brings substantially faster convergence. 1 Deep Neural Networks Deep learning methods aim at learning feature hierarchies with features from higher levels of the hierarchy formed by the composition of lower level features. They include Appearing in Proceedings of the 13 International Conference on Artificial Intelligence and Statistics (AISTATS) 2010, Chia Laguna Resort, Sardinia, Italy. Volume 9 of JMLR: WC Weston et al., 2008). Much attention has recently been devoted to them (see (Bengio, 2009) for a review), because of their theoretical appeal, inspiration from biology and human cognition, and because of empirical success in vision (Ranzato et al., 2007; Larochelle et al., 2007; Vincent et al., 2008) and natural language processing (NLP) (Collobert & Weston, 2008; Mnih & Hinton, 2009). Theoretical results reviewed and discussed by Bengio (2009), suggest that in order to learn the kind of complicated functions that can represent high-level abstractions (e.g. in vision, language, and other AI-level tasks), one may need deep architectures. Most of the recent experimental results with deep architecture are obtained with models that can be turned into deep supervised neural networks, but with initialization or training schemes different from the classical feedforward neural networks (Rumelhart et al., 1986). Why are these new algorithms working so much better than the standard random initialization and gradient-based optimization of a supervised training criterion? Part of the answer may be found in recent analyses of the effect of unsupervised pretraining (Erhan et al., 2009), showing that it acts as a regularizer that initializes the parameters in a \u201cbetter\u201d basin of attraction of the optimization procedure, corresponding to an apparent local minimum associated with better generalization. But earlier work (Bengio et al., 2007) had shown that even a purely supervised but greedy layer-wise procedure would give better results. So here instead of focusing on what unsupervised pre-training or semi-supervised criteria bring to deep architectures, we focus on analyzing what may be going wrong with good old (but deep) multilayer neural networks. Our analysis is driven by investigative experiments to monitor activations (watching for saturation of hidden units) and gradients, across layers and across training iterations. We also evaluate the effects on these of choices of activation function (with the idea that it might affect saturation) and initialization procedure (since unsupervised pretraining is a particular form of initialization and it has a drastic impact)."
                },
                {
                    "title": "Approximation by Fully Complex Multilayer Perceptrons",
                    "abstract": "We investigate the approximation ability of a multi layer perceptron (MLP) network when it is extended to the complex domain. The main challenge for processing complex data with neural networks has been the lack of bounded and analytic complex nonlinear activation functions in the complex domain, as stated by Liouville's theorem. To avoid the conflict between the boundedness and the analyticity of a nonlinear complex function in the complex domain, a number of ad hoc MLPs that include using two real-valued MLPs, one processing the real part and the other processing the imaginary part, have been traditionally employed. However, since nonanalytic functions do not meet the Cauchy-Riemann conditions, they render themselves into degenerative backpropagation algorithms that compromise the efficiency of nonlinear approximation and learning in the complex vector field. A number of elementary transcendental functions (ETFs) derivable from the entire exponential functionez that are analytic are defined as fully complex activation functions and are shown to provide a parsimonious structure for processing data in the complex domain and address most of the shortcomings of the traditional approach. The introduction of ETFs, however, raises a new question in the approximation capability of this fully complex MLP. In this letter, three proofs of the approximation capability of the fully complex MLP are provided based on the characteristics of singularity among ETFs. First, the fully complex MLPs with continuous ETFs over a compact set in the complex vector field are shown to be the universal approximator of any continuous complex mappings. The complex universal approximation theorem extends to bounded measurable ETFs possessing a removable singularity. Finally, it is shown that the output of complex MLPs using ETFs with isolated and essential singularities uniformly converges to any nonlinear mapping in the deleted annulus of singularity nearest to the origin."
                },
                {
                    "title": "The Vanishing Gradient Problem During Learning Recurrent Neural Nets and Problem Solutions",
                    "abstract": "Recurrent nets are in principle capable to store past inputs to produce the currently desired output. Because of this property recurrent nets are used in time series prediction and process control. Practical applications involve temporal dependencies spanning many time steps, e.g. between relevant inputs and desired outputs. In this case, however, gradient based learning methods take too much time. The extremely increased learning time arises because the error vanishes as it gets propagated back. In this article the de-caying error flow is theoretically analyzed. Then methods trying to overcome vanishing gradients are briefly discussed. Finally, experiments comparing conventional algorithms and alternative methods are presented. With advanced methods long time lag problems can be solved in reasonable time."
                },
                {
                    "title": "Marching cubes: A high resolution 3D surface construction algorithm",
                    "abstract": "We present a new algorithm, called marching cubes, that creates triangle models of constant density surfaces from 3D medical data. Using a divide-and-conquer approach to generate inter-slice connectivity, we create a case table that defines triangle topology. The algorithm processes the 3D medical data in scan-line order and calculates triangle vertices using linear interpolation. We find the gradient of the original data, normalize it, and use it as a basis for shading the models. The detail in images produced from the generated surface models is the result of maintaining the inter-slice connectivity, surface data, and gradient information present in the original 3D data. Results from computed tomography (CT), magnetic resonance (MR), and single-photon emission computed tomography (SPECT) illustrate the quality and functionality of marching cubes. We also discuss improvements that decrease processing time and add solid modeling capabilities."
                },
                {
                    "title": "PRINCIPLES OF NEURODYNAMICS. PERCEPTRONS AND THE THEORY OF BRAIN MECHANISMS",
                    "abstract": "Abstract : Part I attempts to review the background, basic sources of data, concepts, and methodology to be employed in the study of perceptrons. In Chapter 2, a brief review of the main alternative approaches to the development of brain models is presented. Chapter 3 considers the physiological and psychological criteria for a suitable model, and attempts to evaluate the empirical evidence which is available on several important issues. Chapter 4 contains basic definitions and some of the notation to be used in later sections are presented. Parts II and III are devoted to a summary of the established theoretical results obtained to date. Part II (Chapters 5 through 14) deals with the theory of three-layer series-coupled perceptrons, on which most work has been done to date. Part III (Chapters 15 through 20) deals with the theory of multi-layer and cross-coupled perceptrons. Part IV is concerned with more speculative models and problems for future analysis. Of necessity, the final chapters become increasingly heuristic in character, as the theory of perceptrons is not yet complete, and new possibilities are continually coming to light."
                },
                {
                    "title": "Implicit Neural Representations with Levels-of-Experts",
                    "abstract": "Coordinate-based networks, usually in the forms of MLPs, have been successfully applied to the task of predicting high-frequency but low-dimensional signals using coordinate inputs. To scale them to model large-scale signals, previous works resort to hybrid representations, combining a coordinate-based network with a grid-based representation, such as sparse voxels. However, such approaches lack a compact global latent representation in its grid, making it difficult to model a distribution of signals, which is important for generalization tasks. To address the limitation, we propose the Levels-of-Experts (LoE) framework, which is a novel coordinate-based representation consisting of an MLP with periodic, position-dependent weights arranged hierarchically. For each linear layer of the MLP, multiple candidate values of its weight matrix are tiled and replicated across the input space, with different layers replicating at different frequencies. Based on the input, only one of the weight matrices is chosen for each layer. This greatly increases the model capacity without incurring extra computation or compromising generalization capability. We show that the new representation is an efficient and competitive drop-in replacement for a wide range of tasks, including signal fitting, novel view synthesis, and generative modeling."
                },
                {
                    "title": "Supplemental Materials: Neural 3D Video Synthesis from Multi-view Video",
                    "abstract": "Our collected data can provide sufficient synchronized camera views for high quality 4D reconstruction of challenging dynamic objects and view-dependent effects in a natural daily indoor environment, which did not exist in public 4D datasets. Our captured data demonstrates a variety of challenges for video synthesis, including objects of high specularity, translucency and transparency. It also contains scene changes and motions with changing topology (poured liquid), self-cast moving shadows, volumetric effects (fire flame), and an entangled moving object with strong view-dependent effects (the torch gun and the pan), various lighting conditions (daytime, night, spotlight from the side), multiple people moving around in open living room space with outdoor scenes seen through transparent windows with relatively dark indoor illumination. We visualize one snapshot of the sequence in Fig. 1. Unless otherwise stated, we use keyframes that are 30 frames apart. In total, we trained our methods on a 60 second video sequence (flame salmon) in 6 chunks with each 10 seconds in length, five other 10 seconds cooking videos captured at different time with different motion and lighting, and one 25 seconds video in indoor videos in 5 chunks. We also trained a few additional videos of outdoor scenes in chunks of 5 seconds with denser keyframes, which are 10 frames apart. In the end, we employ a subset of 18 camera views for training, and 1 view for quantitative evaluation for all datasets except one sequence observing multiple people moving, which only uses 14 cameras views for training. We calculate a continuous interpolated spiral trajectory based on the training camera views, which we employ for qualitative novel view evaluation."
                },
                {
                    "title": "Multiplicative Filter Networks",
                    "abstract": "Although deep networks are typically used to approximate functions over high dimensional inputs, recent work has increased interest in neural networks as function approximators for low-dimensional-but-complex functions, such as representing images as a function of pixel coordinates, solving differential equations, or representing signed distance functions or neural radiance fields. Key to these recent successes has been the use of new elements such as sinusoidal nonlinearities or Fourier features in positional encodings, which vastly outperform simple ReLU networks. In this paper, we propose and empirically demonstrate that an arguably simpler class of function approximators can work just as well for such problems: multiplicative filter networks. In these networks, we avoid traditional compositional depth altogether, and simply multiply together (linear functions of) sinusoidal or Gabor wavelet functions applied to the input. This representation has the notable advantage that the entire function can simply be viewed as a linear function approximator over an exponential number of Fourier or Gabor basis functions, respectively. Despite this simplicity, when compared to recent approaches that use Fourier features with ReLU networks or sinusoidal activation networks, we show that these multiplicative filter networks largely outperform or match the performance of these approaches on the domains highlighted in these past works."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively increase the model capacity of Multi-layer Perceptrons (MLPs) for representing complex spatiotemporal signals while maintaining their implicit regularization properties and compatibility with existing techniques to reduce spectral bias?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the capabilities of neural fields in accurately representing complex real-world signals, particularly in applications such as video synthesis, dynamic scene reconstruction, and other areas requiring fine-grained detail. By addressing the limitations of current MLP architectures, this research could lead to significant improvements in the performance of neural networks, fostering further innovations in machine learning and computer vision. The findings could inspire new methodologies and frameworks that enhance the efficiency and effectiveness of neural networks, ultimately impacting future research directions and practical applications across various domains.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in balancing increased model capacity with the need to maintain efficient inference and optimization processes. Naive approaches, such as simply increasing the number of neurons, lead to slower inference times and higher memory costs, making them impractical for large-scale applications. Additionally, existing methods like meta-learning or partitioning spatiotemporal fields introduce their own complexities, such as slow training times and limited global reasoning capabilities. Overcoming these technical and practical obstacles requires innovative solutions that can enhance model capacity without compromising performance or scalability.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on addressing spectral bias through techniques like positional encoding and specialized activation functions, but these methods have not fully resolved the issue of representing fine-grained details in complex signals. The limitations of existing solutions stem from their inability to effectively increase model capacity without incurring significant computational costs or sacrificing the implicit regularization properties of MLPs. Our approach differs by introducing time-dependent layers that utilize trainable residual parameters, allowing for greater flexibility and compatibility with existing techniques while addressing the shortcomings of prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves replacing conventional MLP layers with Residual Field Layers that are time-dependent, allowing for the modeling of weights as trainable residual parameters. We will evaluate our approach using datasets that include complex spatiotemporal signals, such as dynamic 3D scenes and long videos. The performance will be measured using metrics that assess the accuracy of signal representation and"
            }
        },
        "author_data": {
            "9f84e22a-2614-4cbe-9995-da0fd130cb25": {
                "pk": "9f84e22a-2614-4cbe-9995-da0fd130cb25",
                "name": "Marko Mihajlovic",
                "collaborators": [
                    "Siyu Tang",
                    "Shaofei Wang",
                    "Sergey Prokudin",
                    "Yan Zhang",
                    "Qianli Ma",
                    "Andreas Geiger",
                    "Aayush Bansal",
                    "Michael Zollhoefer",
                    "Shunsuke Saito",
                    "Deheng Zhang"
                ],
                "domain": [
                    "Computer Vision",
                    "Neural Rendering",
                    "Machine Learning",
                    "3D Reconstruction"
                ],
                "publications": [
                    {
                        "title": "RISE-SDF: a Relightable Information-Shared Signed Distance Field for Glossy Object Inverse Rendering",
                        "abstract": "In this paper, we propose a novel end-to-end relightable neural inverse rendering system that achieves high-quality reconstruction of geometry and material properties, thus enabling high-quality relighting. The cornerstone of our method is a two-stage approach for learning a better factorization of scene parameters. In the first stage, we develop a reflection-aware radiance field using a neural signed distance field (SDF) as the geometry representation and deploy an MLP (multilayer perceptron) to estimate indirect illumination. In the second stage, we introduce a novel information-sharing network structure to jointly learn the radiance field and the physically based factorization of the scene. For the physically based factorization, to reduce the noise caused by Monte Carlo sampling, we apply a split-sum approximation with a simplified Disney BRDF and cube mipmap as the environment light representation. In the relighting phase, to enhance the quality of indirect illumination, we propose a second split-sum algorithm to trace secondary rays under the split-sum rendering framework. Furthermore, there is no dataset or protocol available to quantitatively evaluate the inverse rendering performance for glossy objects. To assess the quality of material reconstruction and relighting, we have created a new dataset with ground truth BRDF parameters and relighting results. Our experiments demonstrate that our algorithm achieves state-of-the-art performance in inverse rendering and relighting, with particularly strong results in the reconstruction of highly reflective objects."
                    },
                    {
                        "title": "SplatFields: Neural Gaussian Splats for Sparse 3D and 4D Reconstruction",
                        "abstract": "Digitizing 3D static scenes and 4D dynamic events from multi-view images has long been a challenge in computer vision and graphics. Recently, 3D Gaussian Splatting (3DGS) has emerged as a practical and scalable reconstruction method, gaining popularity due to its impressive reconstruction quality, real-time rendering capabilities, and compatibility with widely used visualization tools. However, the method requires a substantial number of input views to achieve high-quality scene reconstruction, introducing a significant practical bottleneck. This challenge is especially severe in capturing dynamic scenes, where deploying an extensive camera array can be prohibitively costly. In this work, we identify the lack of spatial autocorrelation of splat features as one of the factors contributing to the suboptimal performance of the 3DGS technique in sparse reconstruction settings. To address the issue, we propose an optimization strategy that effectively regularizes splat features by modeling them as the outputs of a corresponding implicit neural field. This results in a consistent enhancement of reconstruction quality across various scenarios. Our approach effectively handles static and dynamic cases, as demonstrated by extensive testing across different setups and scene complexities."
                    },
                    {
                        "title": "Degrees of Freedom Matter: Inferring Dynamics from Point Trajectories",
                        "abstract": "Understanding the dynamics of generic 3D scenes is fundamentally challenging in computer vision, essential in enhancing applications related to scene reconstruction, motion tracking, and avatar creation. In this work, we address the task as the problem of inferring dense, long-range motion of 3D points. By observing a set of point trajectories, we aim to learn an implicit motion field parameterized by a neural network to predict the movement of novel points within the same domain, without relying on any data-driven or scene-specific priors. To achieve this, our approach builds upon the recently introduced dynamic point field model [48] that learns smooth deformation fields between the canonical frame and individual observation frames. However, temporal consistency between consecutive frames is neglected, and the number of required parameters increases linearly with the sequence length due to per-frame modeling. To address these shortcomings, we exploit the intrinsic regularization provided by SIREN [53], and modify the input layer to produce a spatiotemporally smooth motion field. Additionally, we analyze the motion field Jacobian matrix, and discover that the motion degrees of freedom (DOFs) in an infinitesimal area around a point and the network hidden variables have different behaviors to affect the model's representational power. This enables us to improve the model representation capability while retaining the model compactness. Furthermore, to reduce the risk of overfitting, we introduce a regularization term based on the assumption of piece-wise motion smoothness. Our experiments assess the model's performance in predicting unseen point trajectories and its application in temporal mesh alignment with guidance. The results demonstrate its superiority and effectiveness. The code and data for the project are publicly available1 1https://yz-cnsdqz.github.io/eigenmotion/DOMA/."
                    },
                    {
                        "title": "3DGS-Avatar: Animatable Avatars via Deformable 3D Gaussian Splatting",
                        "abstract": "We introduce an approach that creates animatable hu-man avatars from monocular videos using 3D Gaussian Splatting (3DGS). Existing methods based on neural radi-ance fields (NeRFs) achieve high-quality novel-viewlnovel-pose image synthesis but often require days of training, and are extremely slow at inference time. Recently, the com-munity has explored fast grid structures for efficient training of clothed avatars. Albeit being extremely fast at training, these methods can barely achieve an interactive ren-de ring frame rate with around 15 FPS. In this paper, we use 3D Gaussian Splatting and learn a non-rigid deformation network to reconstruct animatable clothed human avatars that can be trained within 30 minutes and rendered at real-time frame rates (50+ FPS). Given the explicit nature of our representation, we further introduce as-isometric-as-possible regularizations on both the Gaussian mean vectors and the covariance matrices, enhancing the generalization of our model on highly articulated unseen poses. Experi-mental results show that our method achieves comparable and even better performance compared to state-of-the-art approaches on animatable avatar creation from a monoc-ular input, while being 400x and 250x faster in training and inference, respectively. Please see our project page at https://neuralbodies.github.ioI3DGS-Avatar."
                    },
                    {
                        "title": "KeypointNeRF: Generalizing Image-based Volumetric Avatars using Relative Spatial Encoding of Keypoints",
                        "abstract": "Image-based volumetric humans using pixel-aligned features promise generalization to unseen poses and identities. Prior work leverages global spatial encodings and multi-view geometric consistency to reduce spatial ambiguity. However, global encodings often suffer from overfitting to the distribution of the training data, and it is difficult to learn multi-view consistent reconstruction from sparse views. In this work, we investigate common issues with existing spatial encodings and propose a simple yet highly effective approach to modeling high-fidelity volumetric humans from sparse views. One of the key ideas is to encode relative spatial 3D information via sparse 3D keypoints. This approach is robust to the sparsity of viewpoints and cross-dataset domain gap. Our approach outperforms state-of-the-art methods for head reconstruction. On human body reconstruction for unseen subjects, we also achieve performance comparable to prior work that uses a parametric human body model and temporal feature aggregation. Our experiments show that a majority of errors in prior work stem from an inappropriate choice of spatial encoding and thus we suggest a new direction for high-fidelity image-based human modeling. https://markomih.github.io/KeypointNeRF"
                    },
                    {
                        "title": "COAP: Compositional Articulated Occupancy of People",
                        "abstract": "We present a novel neural implicit representation for articulated human bodies. Compared to explicit template meshes, neural implicit body representations provide an efficient mechanism for modeling interactions with the environment, which is essential for human motion reconstruction and synthesis in 3D scenes. However, existing neural implicit bodies suffer from either poor generalization on highly articulated poses or slow inference time. In this work, we observe that prior knowledge about the human body's shape and kinematic structure can be leveraged to improve generalization and efficiency. We decompose the full-body geometry into local body parts and employ a part-aware encoder-decoder architecture to learn neural articulated occupancy that models complex deformations locally. Our local shape encoder represents the body deformation of not only the corresponding body part but also the neighboring body parts. The decoder incorporates the geometric constraints of local body shape which significantly improves pose generalization. We demonstrate that our model is suitable for resolving self-intersections and collisions with 3D environments. Quantitative and qualitative experiments show that our method largely outperforms existing solutions in terms of both efficiency and accuracy."
                    },
                    {
                        "title": "LEAP: Learning Articulated Occupancy of People",
                        "abstract": "Substantial progress has been made on modeling rigid 3D objects using deep implicit representations. Yet, extending these methods to learn neural models of human shape is still in its infancy. Human bodies are complex and the key challenge is to learn a representation that generalizes such that it can express body shape deformations for unseen subjects in unseen, highly-articulated, poses. To address this challenge, we introduce LEAP (LEarning Articulated occupancy of People), a novel neural occupancy representation of the human body. Given a set of bone transformations (i.e. joint locations and rotations) and a query point in space, LEAP first maps the query point to a canonical space via learned linear blend skinning (LBS) functions and then efficiently queries the occupancy value via an occupancy network that models accurate identity- and pose-dependent deformations in the canonical space. Experiments show that our canonicalized occupancy estimation with the learned LBS functions greatly improves the generalization capability of the learned occupancy representation across various human shapes and poses, outperforming existing solutions in all settings."
                    },
                    {
                        "title": "MetaAvatar: Learning Animatable Clothed Human Models from Few Depth Images",
                        "abstract": "In this paper, we aim to create generalizable and controllable neural signed distance fields (SDFs) that represent clothed humans from monocular depth observations. Recent advances in deep learning, especially neural implicit representations, have enabled human shape reconstruction and controllable avatar generation from different sensor inputs. However, to generate realistic cloth deformations from novel input poses, watertight meshes or dense full-body scans are usually needed as inputs. Furthermore, due to the difficulty of effectively modeling pose-dependent cloth deformations for diverse body shapes and cloth types, existing approaches resort to per-subject/cloth-type optimization from scratch, which is computationally expensive. In contrast, we propose an approach that can quickly generate realistic clothed human avatars, represented as controllable neural SDFs, given only monocular depth images. We achieve this by using meta-learning to learn an initialization of a hypernetwork that predicts the parameters of neural SDFs. The hypernetwork is conditioned on human poses and represents a clothed neural avatar that deforms non-rigidly according to the input poses. Meanwhile, it is meta-learned to effectively incorporate priors of diverse body shapes and cloth types and thus can be much faster to fine-tune, compared to models trained from scratch. We qualitatively and quantitatively show that our approach outperforms state-of-the-art approaches that require complete meshes as inputs while our approach requires only depth frames as inputs and runs orders of magnitudes faster. Furthermore, we demonstrate that our meta-learned hypernetwork is very robust, being the first to generate avatars with realistic dynamic cloth deformations given as few as 8 monocular depth frames."
                    },
                    {
                        "title": "DeepSurfels: Learning Online Appearance Fusion",
                        "abstract": "We present DeepSurfels, a novel hybrid scene representation for geometry and appearance information. DeepSurfels combines explicit and neural building blocks to jointly encode geometry and appearance information. In contrast to established representations, DeepSurfels better represents high-frequency textures, is well-suited for online updates of appearance information, and can be easily combined with machine learning methods. We further present an end-to-end trainable online appearance fusion pipeline that fuses information from RGB images into the proposed scene representation and is trained using self-supervision imposed by the reprojection error with respect to the input images. Our method compares favorably to classical texture mapping approaches as well as recent learning-based techniques. Moreover, we demonstrate lower runtime, improved generalization capabilities, and better scalability to larger scenes compared to existing methods."
                    },
                    {
                        "title": "Finding the most similar textual documents using Case-Based Reasoning",
                        "abstract": "In recent years, huge amounts of unstructured textual data on the Internet are a big difficulty for AI algorithms to provide the best recommendations for users and their search queries. Since the Internet became widespread, a lot of research has been done in the field of Natural Language Processing (NLP) and machine learning. Almost every solution transforms documents into Vector Space Models (VSM) in order to apply AI algorithms over them. One such approach is based on Case-Based Reasoning (CBR). Therefore, the most important part of those systems is to compute the similarity between numerical data points. In 2016, the new similarity TS-SS metric is proposed, which showed state-of-the-art results in the field of textual mining for unsupervised learning. However, no one before has investigated its performances for supervised learning (classification task). In this work, we devised a CBR system capable of finding the most similar documents for a given query aiming to investigate performances of the new state-of-the-art metric, TS-SS, in addition to the two other geometrical similarity measures --- Euclidean distance and Cosine similarity --- that showed the best predictive results over several benchmark corpora. The results show surprising inappropriateness of TS-SS measure for high dimensional features."
                    },
                    {
                        "title": "A Machine Learning Approach to Classify Pedestrians\u2019 Event based on IMU and GPS",
                        "abstract": "This paper investigates and implements six Machine Learning (ML) algorithms, i.e. Artificial Neural Network (ANN), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Extra Tree ( ..."
                    },
                    {
                        "title": "Fooling a neural network with common adversarial noise",
                        "abstract": "These days deep Neural Networks (NN) show exceptional performance on speech and visual recognition tasks. These systems are still considered a black box without deep understanding why they perform in such a manner. This lack of understanding makes NNs vulnerable to specially crafted adversarial examples \u2014 inputs with small perturbations that make the model misclassify. In this paper, we generated adversarial examples that will fool a NN used for classifying handwritten digits. We start by generating additive adversarial noise for each image. Afterwards, we propose an algorithm for crafting a single adversarial noise for misclassifying different members of the same class."
                    }
                ]
            },
            "3b89e4b3-0fe5-4214-b2aa-a8382b4e1902": {
                "pk": "3b89e4b3-0fe5-4214-b2aa-a8382b4e1902",
                "name": "Sergey Prokudin",
                "collaborators": [
                    "Marko Mihajlovic",
                    "Siyu Tang",
                    "Shaofei Wang",
                    "Deheng Zhang",
                    "Jingyu Wang",
                    "Hendrik P. A. Lensch",
                    "Adam Kania",
                    "Jacek Tabor",
                    "P. Spurek",
                    "Jonas Hein"
                ],
                "domain": [
                    "Neural Rendering",
                    "3D Reconstruction",
                    "Implicit Neural Representations",
                    "Motion Tracking"
                ],
                "publications": [
                    {
                        "title": "RISE-SDF: a Relightable Information-Shared Signed Distance Field for Glossy Object Inverse Rendering",
                        "abstract": "In this paper, we propose a novel end-to-end relightable neural inverse rendering system that achieves high-quality reconstruction of geometry and material properties, thus enabling high-quality relighting. The cornerstone of our method is a two-stage approach for learning a better factorization of scene parameters. In the first stage, we develop a reflection-aware radiance field using a neural signed distance field (SDF) as the geometry representation and deploy an MLP (multilayer perceptron) to estimate indirect illumination. In the second stage, we introduce a novel information-sharing network structure to jointly learn the radiance field and the physically based factorization of the scene. For the physically based factorization, to reduce the noise caused by Monte Carlo sampling, we apply a split-sum approximation with a simplified Disney BRDF and cube mipmap as the environment light representation. In the relighting phase, to enhance the quality of indirect illumination, we propose a second split-sum algorithm to trace secondary rays under the split-sum rendering framework. Furthermore, there is no dataset or protocol available to quantitatively evaluate the inverse rendering performance for glossy objects. To assess the quality of material reconstruction and relighting, we have created a new dataset with ground truth BRDF parameters and relighting results. Our experiments demonstrate that our algorithm achieves state-of-the-art performance in inverse rendering and relighting, with particularly strong results in the reconstruction of highly reflective objects."
                    },
                    {
                        "title": "FreSh: Frequency Shifting for Accelerated Neural Representation Learning",
                        "abstract": "Implicit Neural Representations (INRs) have recently gained attention as a powerful approach for continuously representing signals such as images, videos, and 3D shapes using multilayer perceptrons (MLPs). However, MLPs are known to exhibit a low-frequency bias, limiting their ability to capture high-frequency details accurately. This limitation is typically addressed by incorporating high-frequency input embeddings or specialized activation layers. In this work, we demonstrate that these embeddings and activations are often configured with hyperparameters that perform well on average but are suboptimal for specific input signals under consideration, necessitating a costly grid search to identify optimal settings. Our key observation is that the initial frequency spectrum of an untrained model's output correlates strongly with the model's eventual performance on a given target signal. Leveraging this insight, we propose frequency shifting (or FreSh), a method that selects embedding hyperparameters to align the frequency spectrum of the model's initial output with that of the target signal. We show that this simple initialization technique improves performance across various neural representation methods and tasks, achieving results comparable to extensive hyperparameter sweeps but with only marginal computational overhead compared to training a single model with default hyperparameters."
                    },
                    {
                        "title": "Creating a Digital Twin of Spinal Surgery: A Proof of Concept",
                        "abstract": "Surgery digitalization is the process of creating a virtual replica of real-world surgery, also referred to as a surgical digital twin (SDT). It has significant applications in various fields such as education and training, surgical planning, and automation of surgical tasks. In addition, SDTs are an ideal foundation for machine learning methods, enabling the automatic generation of training data. In this paper, we present a proof of concept (PoC) for surgery digitalization that is applied to an ex-vivo spinal surgery. The proposed digitalization focuses on the acquisition and modelling of the geometry and appearance of the entire surgical scene. We employ five RGB-D cameras for dynamic 3D reconstruction of the surgeon, a high-end camera for 3D reconstruction of the anatomy, an infrared stereo camera for surgical instrument tracking, and a laser scanner for 3D reconstruction of the operating room and data fusion. We justify the proposed methodology, discuss the challenges faced and further extensions of our prototype. While our PoC partially relies on manual data curation, its high quality and great potential motivate the development of automated methods for the creation of SDTs."
                    },
                    {
                        "title": "Morphable Diffusion: 3D-Consistent Diffusion for Single-image Avatar Creation",
                        "abstract": "Recent advances in generative diffusion models have enabled the previously unfeasible capability of generating 3D assets from a single input image or a text prompt. In this work, we aim to enhance the quality and functionality of these models for the task of creating controllable, photorealistic human avatars. We achieve this by integrating a 3D morphable model into the state-of-the-art multi-view-consistent diffusion approach. We demonstrate that accurate conditioning of a generative pipeline on the articulated 3D model enhances the baseline model performance on the task of novel view synthesis from a single image. More importantly, this integration facilitates a seamless and accurate incorporation of facial expression and body pose control into the generation process. To the best of our knowledge, our proposed framework is the first diffusion model to enable the creation of fully 3D-consistent, animatable, and photorealistic human avatars from a single image of an unseen subject; extensive quantitative and qualitative evaluations demonstrate the advantages of our approach over existing state-of-the-art avatar creation models on both novel view and novel expression synthesis tasks. The code for our project is publicly available."
                    },
                    {
                        "title": "SplatFields: Neural Gaussian Splats for Sparse 3D and 4D Reconstruction",
                        "abstract": "Digitizing 3D static scenes and 4D dynamic events from multi-view images has long been a challenge in computer vision and graphics. Recently, 3D Gaussian Splatting (3DGS) has emerged as a practical and scalable reconstruction method, gaining popularity due to its impressive reconstruction quality, real-time rendering capabilities, and compatibility with widely used visualization tools. However, the method requires a substantial number of input views to achieve high-quality scene reconstruction, introducing a significant practical bottleneck. This challenge is especially severe in capturing dynamic scenes, where deploying an extensive camera array can be prohibitively costly. In this work, we identify the lack of spatial autocorrelation of splat features as one of the factors contributing to the suboptimal performance of the 3DGS technique in sparse reconstruction settings. To address the issue, we propose an optimization strategy that effectively regularizes splat features by modeling them as the outputs of a corresponding implicit neural field. This results in a consistent enhancement of reconstruction quality across various scenarios. Our approach effectively handles static and dynamic cases, as demonstrated by extensive testing across different setups and scene complexities."
                    },
                    {
                        "title": "Degrees of Freedom Matter: Inferring Dynamics from Point Trajectories",
                        "abstract": "Understanding the dynamics of generic 3D scenes is fundamentally challenging in computer vision, essential in enhancing applications related to scene reconstruction, motion tracking, and avatar creation. In this work, we address the task as the problem of inferring dense, long-range motion of 3D points. By observing a set of point trajectories, we aim to learn an implicit motion field parameterized by a neural network to predict the movement of novel points within the same domain, without relying on any data-driven or scene-specific priors. To achieve this, our approach builds upon the recently introduced dynamic point field model [48] that learns smooth deformation fields between the canonical frame and individual observation frames. However, temporal consistency between consecutive frames is neglected, and the number of required parameters increases linearly with the sequence length due to per-frame modeling. To address these shortcomings, we exploit the intrinsic regularization provided by SIREN [53], and modify the input layer to produce a spatiotemporally smooth motion field. Additionally, we analyze the motion field Jacobian matrix, and discover that the motion degrees of freedom (DOFs) in an infinitesimal area around a point and the network hidden variables have different behaviors to affect the model's representational power. This enables us to improve the model representation capability while retaining the model compactness. Furthermore, to reduce the risk of overfitting, we introduce a regularization term based on the assumption of piece-wise motion smoothness. Our experiments assess the model's performance in predicting unseen point trajectories and its application in temporal mesh alignment with guidance. The results demonstrate its superiority and effectiveness. The code and data for the project are publicly available1 1https://yz-cnsdqz.github.io/eigenmotion/DOMA/."
                    }
                ]
            },
            "be373c1e-46fa-4c5b-b7bb-9246b6ce7cb7": {
                "pk": "be373c1e-46fa-4c5b-b7bb-9246b6ce7cb7",
                "name": "Marc Pollefeys",
                "collaborators": [
                    "Iro Armeni",
                    "D. Bar\u00e1th",
                    "Shaohui Liu",
                    "R\u00e9mi Pautrat",
                    "Viktor Larsson",
                    "Tobias Fischer",
                    "Shengze Jin",
                    "Daniel Barath",
                    "D. Rozumnyi",
                    "Martin R. Oswald"
                ],
                "domain": [
                    "Computer Vision",
                    "3D Reconstruction",
                    "Domain Adaptation",
                    "Augmented Reality"
                ],
                "publications": [
                    {
                        "title": "MICDrop: Masking Image and Depth Features via Complementary Dropout for Domain-Adaptive Semantic Segmentation",
                        "abstract": "Unsupervised Domain Adaptation (UDA) is the task of bridging the domain gap between a labeled source domain, e.g., synthetic data, and an unlabeled target domain. We observe that current UDA methods show inferior results on fine structures and tend to oversegment objects with ambiguous appearance. To address these shortcomings, we propose to leverage geometric information, i.e., depth predictions, as depth discontinuities often coincide with segmentation boundaries. We show that naively incorporating depth into current UDA methods does not fully exploit the potential of this complementary information. To this end, we present MICDrop, which learns a joint feature representation by masking image encoder features while inversely masking depth encoder features. With this simple yet effective complementary masking strategy, we enforce the use of both modalities when learning the joint feature representation. To aid this process, we propose a feature fusion module to improve both global as well as local information sharing while being robust to errors in the depth predictions. We show that our method can be plugged into various recent UDA methods and consistently improve results across standard UDA benchmarks, obtaining new state-of-the-art performances."
                    },
                    {
                        "title": "Multiway Point Cloud Mosaicking with Diffusion and Global Optimization",
                        "abstract": "We introduce a novel framework for multiway point cloud mosaicking (named Wednesday), designed to co-align sets of partially overlapping point clouds - typically obtained from 3D scanners or moving RGB-D cameras - into a unified coordinate system. At the core of our approach is ODIN, a learned pairwise registration algorithm that iteratively identifies overlaps and refines attention scores, employing a diffusion-based process for denoising pairwise correlation matrices to enhance matching accuracy. Further steps include constructing a pose graph from all point clouds, performing rotation averaging, a novel robust algorithm for re-estimating translations optimally in terms of consensus maximization and translation optimization. Finally, the point cloud rotations and positions are optimized jointly by a diffusion-based approach. Tested on four diverse, large-scale datasets, our method achieves state-of-the-art pairwise and multiway registration results by a large margin on all benchmarks. Our code and models are available at https://github.com/jinsz/Multiway-Point-Cloud-Mosaicking-with-Diffusion-and-Global-Optimization."
                    },
                    {
                        "title": "Semantically Guided Feature Matching for Visual SLAM",
                        "abstract": "We introduce a new algorithm that utilizes semantic information to enhance feature matching in visual SLAM pipelines. The proposed method constructs a high-dimensional semantic descriptor for each detected ORB feature. When integrated with traditional visual ones, these descriptors aid in establishing accurate tentative point correspondences between consecutive frames. Additionally, our semantic descriptors enrich 3D map points, enhancing loop closure detection by providing deeper insights into the underlying map regions. Experiments on public large-scale datasets demonstrate that our technique surpasses the accuracy of established methods. Importantly, given its detector-agnostic nature, our algorithm also amplifies the efficacy of modern keypoint detectors, such as SuperPoint. The implementation of our algorithm can be found on Github 3."
                    },
                    {
                        "title": "Global Structure-from-Motion Revisited",
                        "abstract": "Recovering 3D structure and camera motion from images has been a long-standing focus of computer vision research and is known as Structure-from-Motion (SfM). Solutions to this problem are categorized into incremental and global approaches. Until now, the most popular systems follow the incremental paradigm due to its superior accuracy and robustness, while global approaches are drastically more scalable and efficient. With this work, we revisit the problem of global SfM and propose GLOMAP as a new general-purpose system that outperforms the state of the art in global SfM. In terms of accuracy and robustness, we achieve results on-par or superior to COLMAP, the most widely used incremental SfM, while being orders of magnitude faster. We share our system as an open-source implementation at {https://github.com/colmap/glomap}."
                    },
                    {
                        "title": "Retrieval Robust to Object Motion Blur",
                        "abstract": "Moving objects are frequently seen in daily life and usually appear blurred in images due to their motion. While general object retrieval is a widely explored area in computer vision, it primarily focuses on sharp and static objects, and retrieval of motion-blurred objects in large image collections remains unexplored. We propose a method for object retrieval in images that are affected by motion blur. The proposed method learns a robust representation capable of matching blurred objects to their deblurred versions and vice versa. To evaluate our approach, we present the first large-scale datasets for blurred object retrieval, featuring images with objects exhibiting varying degrees of blur in various poses and scales. We conducted extensive experiments, showing that our method outperforms state-of-the-art retrieval methods on the new blur-retrieval datasets, which validates the effectiveness of the proposed approach. Code, data, and model are available at https://github.com/Rong-Zou/Retrieval-Robust-to-Object-Motion-Blur."
                    },
                    {
                        "title": "Efficient Solution of Point-Line Absolute Pose",
                        "abstract": "We revisit certain problems of pose estimation based on 3D-2D correspondences between features which may be points or lines. Specifically, we address the two previously-studied minimal problems of estimating camera extrinsics from $p\\in\\{1,2\\}$ point-point correspondences and $l=3-p$ line-line correspondences. To the best of our knowledge, all of the previously-known practical solutions to these problems required computing the roots of degree $\\geq 4$ (univariate) polynomials when $p=2$, or degree $\\geq 8$ polynomials when $p=1$. We describe and implement two elementary solutions which reduce the degrees of the needed polynomials from 4 to 2 and from 8 to 4, respectively. We show experimentally that the resulting solvers are numerically stable and fast: when compared to the previous state-of-the art, we may obtain nearly an order of magnitude speedup. The code is available at https://github.com/petrhruby97/efficient_absolute"
                    },
                    {
                        "title": "EgoGen: An Egocentric Synthetic Data Generator",
                        "abstract": "Understanding the world in first-person view is fundamental in Augmented Reality (AR). This immersive perspective brings dramatic visual changes and unique challenges compared to third-person views. Synthetic data has empowered third-person-view vision models, but its application to embodied egocentric perception tasks remains largely unexplored. A critical challenge lies in simulating natural human movements and behaviors that effectively steer the embodied cameras to capture a faithful egocentric representation of the 3D world. To address this challenge, we introduce EgoGen, a new synthetic data generator that can produce accurate and rich ground-truth training data for egocentric perception tasks. At the heart of EgoGen is a novel human motion synthesis model that directly leverages egocentric visual inputs of a virtual human to sense the 3D environment. Combined with collision-avoiding motion primitives and a two-stage reinforcement learning approach, our motion synthesis model offers a closed-loop solution where the embodied perception and movement of the virtual human are seamlessly coupled. Compared to previous works, our model eliminates the need for a predefined global path, and is directly applicable to dynamic environments. Combined with our easy-to-use and scalable data generation pipeline, we demonstrate EgoGen's efficacy in three tasks: mapping and localization for head-mounted cameras, egocentric camera tracking, and human mesh recovery from egocentric views. EgoGen will be fully open-sourced, offering a practical solution for creating realistic egocentric training data and aiming to serve as a useful tool for egocentric computer vision research."
                    },
                    {
                        "title": "Nothing Stands Still: A Spatiotemporal Benchmark on 3D Point Cloud Registration Under Large Geometric and Temporal Change",
                        "abstract": "Building 3D geometric maps of man-made spaces is a well-established and active field that is fundamental to computer vision and robotics. However, considering the evolving nature of built environments, it is essential to question the capabilities of current mapping efforts in handling temporal changes. In addition, spatiotemporal mapping holds significant potential for achieving sustainability and circularity goals. Existing mapping approaches focus on small changes, such as object relocation or self-driving car operation; in all cases where the main structure of the scene remains fixed. Consequently, these approaches fail to address more radical changes in the structure of the built environment, such as geometry and topology. To this end, we introduce the Nothing Stands Still (NSS) benchmark, which focuses on the spatiotemporal registration of 3D scenes undergoing large spatial and temporal change, ultimately creating one coherent spatiotemporal map. Specifically, the benchmark involves registering two or more partial 3D point clouds (fragments) from the same scene but captured from different spatiotemporal views. In addition to the standard pairwise registration, we assess the multi-way registration of multiple fragments that belong to any temporal stage. As part of NSS, we introduce a dataset of 3D point clouds recurrently captured in large-scale building indoor environments that are under construction or renovation. The NSS benchmark presents three scenarios of increasing difficulty, to quantify the generalization ability of point cloud registration methods over space (within one building and across buildings) and time. We conduct extensive evaluations of state-of-the-art methods on NSS. The results demonstrate the necessity for novel methods specifically designed to handle large spatiotemporal changes. The homepage of our benchmark is at http://nothing-stands-still.com."
                    },
                    {
                        "title": "ARrow: A Real-Time AR Rowing Coach",
                        "abstract": "Rowing requires physical strength and endurance in athletes as well as a precise rowing technique. The ideal rowing stroke is based on biomechanical principles and typically takes years to master. Except for time-consuming video analysis after practice, coaches currently have no means to quantitatively analyze a rower\u2019s stroke sequence and body movement. We propose ARrow , an AR application for coaches and athletes that provides real-time and situated feedback on a rower\u2019s body position and stroke. We use computer vision techniques to extract the rower\u2019s 3D skeleton and to detect the rower\u2019s stroke cycle. ARrow provides visual feedback on three levels: Tracking of basic performance metrics over time, visual feedback and guidance on a rower\u2019s stroke sequence, and a rowing ghost view that helps synchronize the body movement of two rowers. We developed ARrow in close colaboration with international rowing coaches and demonstrate its usefulness in a user study with athletes and coaches."
                    },
                    {
                        "title": "Tracking by 3D Model Estimation of Unknown Objects in Videos",
                        "abstract": "Most model-free visual object tracking methods formulate the tracking task as object location estimation given by a 2D segmentation or a bounding box in each video frame. We argue that this representation is limited and instead propose to guide and improve 2D tracking with an explicit object representation, namely the textured 3D shape and 6DoF pose in each video frame. Our representation tackles a complex long-term dense correspondence problem between all 3D points on the object for all video frames, including frames where some points are invisible. To achieve that, the estimation is driven by re-rendering the input video frames as well as possible through differentiable rendering, which has not been used for tracking before. The proposed optimization minimizes a novel loss function to estimate the best 3D shape, texture, and 6DoF pose. We improve the state-of-the-art in 2D segmentation tracking on three different datasets with mostly rigid objects."
                    },
                    {
                        "title": "Q-REG: End-to-End Trainable Point Cloud Registration with Surface Curvature",
                        "abstract": "Point cloud registration has seen recent success with several learning-based methods that focus on correspondence matching, and as such, optimize only for this objective. Following the learning step of correspondence matching, they evaluate the estimated rigid transformation with a RANSAC-like framework. While it is an indispensable component of these methods, it prevents a fully end-to-end training, leaving the objective to minimize the pose error non-served. We present a novel solution, Q-REG, which utilizes rich geometric information to estimate the rigid pose from a single correspondence. Q-REG allows to formalize the robust estimation as an exhaustive search, hence enabling end-to-end training that optimizes over both objectives of correspondence matching and rigid pose estimation. We demonstrate in the experiments that Q-REG is agnostic to the correspondence matching method and provides consistent improvement both when used only in inference and in end-to-end training. It sets a new state-of-the-art on the 3DMatch, KITTI and ModelNet benchmarks. Our code and models are available at https://github.com/jinsz/Q-REG."
                    },
                    {
                        "title": "OpenMask3D: Open-Vocabulary 3D Instance Segmentation",
                        "abstract": "We introduce the task of open-vocabulary 3D instance segmentation. Current approaches for 3D instance segmentation can typically only recognize object categories from a pre-defined closed set of classes that are annotated in the training datasets. This results in important limitations for real-world applications where one might need to perform tasks guided by novel, open-vocabulary queries related to a wide variety of objects. Recently, open-vocabulary 3D scene understanding methods have emerged to address this problem by learning queryable features for each point in the scene. While such a representation can be directly employed to perform semantic segmentation, existing methods cannot separate multiple object instances. In this work, we address this limitation, and propose OpenMask3D, which is a zero-shot approach for open-vocabulary 3D instance segmentation. Guided by predicted class-agnostic 3D instance masks, our model aggregates per-mask features via multi-view fusion of CLIP-based image embeddings. Experiments and ablation studies on ScanNet200 and Replica show that OpenMask3D outperforms other open-vocabulary methods, especially on the long-tail distribution. Qualitative experiments further showcase OpenMask3D's ability to segment object properties based on free-form queries describing geometry, affordances, and materials."
                    },
                    {
                        "title": "3D Line Mapping Revisited",
                        "abstract": "In contrast to sparse keypoints, a handful of line segments can concisely encode the high-level scene layout, as they often delineate the main structural elements. In addition to offering strong geometric cues, they are also omnipresent in urban landscapes and indoor scenes. Despite their apparent advantages, current line-based reconstruction methods are far behind their point-based counterparts. In this paper we aim to close the gap by introducing LIMAP, a library for 3D line mapping that robustly and efficiently creates 3D line maps from multi-view imagery. This is achieved through revisiting the degeneracy problem of line triangulation, carefully crafted scoring and track building, and exploiting structural priors such as line coincidence, parallelism, and orthogonality. Our code integrates seamlessly with existing point-based Structure-from-Motion methods and can leverage their 3D points to further improve the line reconstruction. Furthermore, as a byproduct, the method is able to recover 3D association graphs between lines and points / vanishing points (VPs). In thorough experiments, we show that LIMAP significantly outperforms existing approaches for 3D line mapping. Our robust 3D line maps also open up new research directions. We show two example applications: visual localization and bundle adjustment, where integrating lines alongside points yields the best results. Code is available at https://github.com/cvg/limap."
                    },
                    {
                        "title": "Next-generation Surgical Navigation: Multi-view Marker-less 6DoF Pose Estimation of Surgical Instruments",
                        "abstract": ","
                    },
                    {
                        "title": "Consensus-Adaptive RANSAC",
                        "abstract": "RANSAC and its variants are widely used for robust estimation, however, they commonly follow a greedy approach to finding the highest scoring model while ignoring other model hypotheses. In contrast, Iteratively Reweighted Least Squares (IRLS) techniques gradually approach the model by iteratively updating the weight of each correspondence based on the residuals from previous iterations. Inspired by these methods, we propose a new RANSAC framework that learns to explore the parameter space by considering the residuals seen so far via a novel attention layer. The attention mechanism operates on a batch of point-to-model residuals, and updates a per-point estimation state to take into account the consensus found through a lightweight one-step transformer. This rich state then guides the minimal sampling between iterations as well as the model refinement. We evaluate the proposed approach on essential and fundamental matrix estimation on a number of indoor and outdoor datasets. It outperforms state-of-the-art estimators by a significant margin adding only a small runtime overhead. Moreover, we demonstrate good generalization properties of our trained model, indicating its effectiveness across different datasets and tasks. The proposed attention mechanism and one-step transformer provide an adaptive behavior that enhances the performance of RANSAC, making it a more effective tool for robust estimation. Code is available at https://github.com/cavalli1234/CA-RANSAC."
                    },
                    {
                        "title": "Supplementary Material for \u201c3D Line Mapping Revisited\u201d",
                        "abstract": "Full Acknowledgements. We sincerely thank the reviewers for their constructive feedback. We are grateful to Philipp Lindenberger, Paul-Edouard Sarlin, Johannes Sch\u00f6nberger for their open-source projects, Iago Su\u00e1rez, Hengkai Guo, Daniel Barath, Siyan Dong, Martin Oswald, Jing Ren, Iro Armeni for helpful discussions, Marcel Geppert, Daniel Thul, Peizhuo Li for technical support on Blender. Special thanks to Mihai Dusmanu and Wang Zhao for proof-reading. Viktor Larsson received funding by the strategic research project ELLIIT."
                    },
                    {
                        "title": "A 3D Mixed Reality Interface for Human-Robot Teaming",
                        "abstract": "This paper presents a mixed-reality human-robot teaming system. It allows human operators to see in real-time where robots are located, even if they are not in line of sight. The operator can also visualize the map that the robots create of their environment and can easily send robots to new goal positions. The system mainly consists of a mapping and a control module. The mapping module is a real-time multi-agent visual SLAM system that co-localizes all robots and mixed-reality devices to a common reference frame. Visualizations in the mixed-reality device then allow operators to see a virtual life-sized representation of the cumulative 3D map overlaid onto the real environment. As such, the operator can effectively \"see through\" walls into other rooms. To control robots and send them to new locations, we propose a drag-and-drop interface. An operator can grab any robot hologram in a 3D mini map and drag it to a new desired goal pose. We validate the proposed system through a user study and real-world deployments. We make the mixed-reality application publicly available at github.com/cvg/HoloLens_ros."
                    },
                    {
                        "title": "R3D3: Dense 3D Reconstruction of Dynamic Scenes from Multiple Cameras",
                        "abstract": "Dense 3D reconstruction and ego-motion estimation are key challenges in autonomous driving and robotics. Compared to the complex, multi-modal systems deployed today, multi-camera systems provide a simpler, low-cost alternative. However, camera-based 3D reconstruction of complex dynamic scenes has proven extremely difficult, as existing solutions often produce incomplete or incoherent results. We propose R3D3, a multi-camera system for dense 3D reconstruction and ego-motion estimation. Our approach iterates between geometric estimation that exploits spatial-temporal information from multiple cameras, and monocular depth refinement. We integrate multi-camera feature correlation and dense bundle adjustment operators that yield robust geometric depth and pose estimates. To improve reconstruction where geometric depth is unreliable, e.g. for moving objects or low-textured regions, we introduce learnable scene priors via a depth refinement network. We show that this design enables a dense, consistent 3D reconstruction of challenging, dynamic outdoor environments. Consequently, we achieve state-of-the-art dense depth prediction on the DDAD and NuScenes benchmarks."
                    },
                    {
                        "title": "Vanishing Point Estimation in Uncalibrated Images with Prior Gravity Direction",
                        "abstract": "We tackle the problem of estimating a Manhattan frame, i.e. three orthogonal vanishing points, and the unknown focal length of the camera, leveraging a prior vertical direction. The direction can come from an Inertial Measurement Unit that is a standard component of recent consumer devices, e.g., smartphones. We provide an exhaustive analysis of minimal line configurations and derive two new 2-line solvers, one of which does not suffer from singularities affecting existing solvers. Additionally, we design a new non-minimal method, running on an arbitrary number of lines, to boost the performance in local optimization. Combining all solvers in a hybrid robust estimator, our method achieves increased accuracy even with a rough prior. Experiments on synthetic and real-world datasets demonstrate the superior accuracy of our method compared to the state of the art, while having comparable runtimes. We further demonstrate the applicability of our solvers for relative rotation estimation. The code is available at https://github.com/cvg/VP-Estimation-with-Prior-Gravity."
                    }
                ]
            },
            "8a5f3a0c-61f0-40b6-80bf-74cb0e407d89": {
                "pk": "8a5f3a0c-61f0-40b6-80bf-74cb0e407d89",
                "name": "Siyu Tang",
                "collaborators": [
                    "Marko Mihajlovic",
                    "Shaofei Wang",
                    "Sergey Prokudin",
                    "Andreas Geiger",
                    "Deheng Zhang",
                    "Jingyu Wang",
                    "Hendrik P. A. Lensch",
                    "Anpei Chen",
                    "Haofei Xu",
                    "Stefano Esposito"
                ],
                "domain": [
                    "Neural Rendering",
                    "3D Reconstruction",
                    "Generative Models",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "RISE-SDF: a Relightable Information-Shared Signed Distance Field for Glossy Object Inverse Rendering",
                        "abstract": "In this paper, we propose a novel end-to-end relightable neural inverse rendering system that achieves high-quality reconstruction of geometry and material properties, thus enabling high-quality relighting. The cornerstone of our method is a two-stage approach for learning a better factorization of scene parameters. In the first stage, we develop a reflection-aware radiance field using a neural signed distance field (SDF) as the geometry representation and deploy an MLP (multilayer perceptron) to estimate indirect illumination. In the second stage, we introduce a novel information-sharing network structure to jointly learn the radiance field and the physically based factorization of the scene. For the physically based factorization, to reduce the noise caused by Monte Carlo sampling, we apply a split-sum approximation with a simplified Disney BRDF and cube mipmap as the environment light representation. In the relighting phase, to enhance the quality of indirect illumination, we propose a second split-sum algorithm to trace secondary rays under the split-sum rendering framework. Furthermore, there is no dataset or protocol available to quantitatively evaluate the inverse rendering performance for glossy objects. To assess the quality of material reconstruction and relighting, we have created a new dataset with ground truth BRDF parameters and relighting results. Our experiments demonstrate that our algorithm achieves state-of-the-art performance in inverse rendering and relighting, with particularly strong results in the reconstruction of highly reflective objects."
                    },
                    {
                        "title": "LaRa: Efficient Large-Baseline Radiance Fields",
                        "abstract": "Radiance field methods have achieved photorealistic novel view synthesis and geometry reconstruction. But they are mostly applied in per-scene optimization or small-baseline settings. While several recent works investigate feed-forward reconstruction with large baselines by utilizing transformers, they all operate with a standard global attention mechanism and hence ignore the local nature of 3D reconstruction. We propose a method that unifies local and global reasoning in transformer layers, resulting in improved quality and faster convergence. Our model represents scenes as Gaussian Volumes and combines this with an image encoder and Group Attention Layers for efficient feed-forward reconstruction. Experimental results demonstrate that our model, trained for two days on four GPUs, demonstrates high fidelity in reconstructing 360 deg radiance fields, and robustness to zero-shot and out-of-domain testing. Our project Page: https://apchenstu.github.io/LaRa/."
                    },
                    {
                        "title": "Morphable Diffusion: 3D-Consistent Diffusion for Single-image Avatar Creation",
                        "abstract": "Recent advances in generative diffusion models have enabled the previously unfeasible capability of generating 3D assets from a single input image or a text prompt. In this work, we aim to enhance the quality and functionality of these models for the task of creating controllable, photorealistic human avatars. We achieve this by integrating a 3D morphable model into the state-of-the-art multi-view-consistent diffusion approach. We demonstrate that accurate conditioning of a generative pipeline on the articulated 3D model enhances the baseline model performance on the task of novel view synthesis from a single image. More importantly, this integration facilitates a seamless and accurate incorporation of facial expression and body pose control into the generation process. To the best of our knowledge, our proposed framework is the first diffusion model to enable the creation of fully 3D-consistent, animatable, and photorealistic human avatars from a single image of an unseen subject; extensive quantitative and qualitative evaluations demonstrate the advantages of our approach over existing state-of-the-art avatar creation models on both novel view and novel expression synthesis tasks. The code for our project is publicly available."
                    },
                    {
                        "title": "SplatFields: Neural Gaussian Splats for Sparse 3D and 4D Reconstruction",
                        "abstract": "Digitizing 3D static scenes and 4D dynamic events from multi-view images has long been a challenge in computer vision and graphics. Recently, 3D Gaussian Splatting (3DGS) has emerged as a practical and scalable reconstruction method, gaining popularity due to its impressive reconstruction quality, real-time rendering capabilities, and compatibility with widely used visualization tools. However, the method requires a substantial number of input views to achieve high-quality scene reconstruction, introducing a significant practical bottleneck. This challenge is especially severe in capturing dynamic scenes, where deploying an extensive camera array can be prohibitively costly. In this work, we identify the lack of spatial autocorrelation of splat features as one of the factors contributing to the suboptimal performance of the 3DGS technique in sparse reconstruction settings. To address the issue, we propose an optimization strategy that effectively regularizes splat features by modeling them as the outputs of a corresponding implicit neural field. This results in a consistent enhancement of reconstruction quality across various scenarios. Our approach effectively handles static and dynamic cases, as demonstrated by extensive testing across different setups and scene complexities."
                    },
                    {
                        "title": "Degrees of Freedom Matter: Inferring Dynamics from Point Trajectories",
                        "abstract": "Understanding the dynamics of generic 3D scenes is fundamentally challenging in computer vision, essential in enhancing applications related to scene reconstruction, motion tracking, and avatar creation. In this work, we address the task as the problem of inferring dense, long-range motion of 3D points. By observing a set of point trajectories, we aim to learn an implicit motion field parameterized by a neural network to predict the movement of novel points within the same domain, without relying on any data-driven or scene-specific priors. To achieve this, our approach builds upon the recently introduced dynamic point field model [48] that learns smooth deformation fields between the canonical frame and individual observation frames. However, temporal consistency between consecutive frames is neglected, and the number of required parameters increases linearly with the sequence length due to per-frame modeling. To address these shortcomings, we exploit the intrinsic regularization provided by SIREN [53], and modify the input layer to produce a spatiotemporally smooth motion field. Additionally, we analyze the motion field Jacobian matrix, and discover that the motion degrees of freedom (DOFs) in an infinitesimal area around a point and the network hidden variables have different behaviors to affect the model's representational power. This enables us to improve the model representation capability while retaining the model compactness. Furthermore, to reduce the risk of overfitting, we introduce a regularization term based on the assumption of piece-wise motion smoothness. Our experiments assess the model's performance in predicting unseen point trajectories and its application in temporal mesh alignment with guidance. The results demonstrate its superiority and effectiveness. The code and data for the project are publicly available1 1https://yz-cnsdqz.github.io/eigenmotion/DOMA/."
                    },
                    {
                        "title": "IntrinsicAvatar: Physically Based Inverse Rendering of Dynamic Humans from Monocular Videos via Explicit Ray Tracing",
                        "abstract": "We present IntrinsicAvatar, a novel approach to recov-ering the intrinsic properties of clothed human avatars in-cluding geometry, albedo, material, and environment lighting from only monocular videos. Recent advancements in human-based neural rendering have enabled high-quality geometry and appearance reconstruction of clothed humans from just monocular videos. However, these methods bake intrinsic properties such as albedo, material, and environment lighting into a single entangled neural representation. On the other hand, only a handful of works tackle the problem of estimating geometry and disentangled appearance properties of clothed humans from monocular videos. They usually achieve limited quality and disentanglement due to approximations of secondary shading effects via learned MLPs. In this work, we propose to model secondary shading effects explicitly via Monte-Carlo ray tracing. We model the rendering process of clothed humans as a volumetric scattering process, and combine ray tracing with body ar-ticulation. Our approach can recover high-quality geom-etry, albedo, material, and lighting properties of clothed humans from a single monocular video, without requiring supervised pretraining using ground truth materials. Fur-thermore, since we explicitly model the volumetric scattering process and ray tracing, our model naturally generalizes to novel poses, enabling animation of the reconstructed avatar in novel lighting conditions."
                    },
                    {
                        "title": "3DGS-Avatar: Animatable Avatars via Deformable 3D Gaussian Splatting",
                        "abstract": "We introduce an approach that creates animatable hu-man avatars from monocular videos using 3D Gaussian Splatting (3DGS). Existing methods based on neural radi-ance fields (NeRFs) achieve high-quality novel-viewlnovel-pose image synthesis but often require days of training, and are extremely slow at inference time. Recently, the com-munity has explored fast grid structures for efficient training of clothed avatars. Albeit being extremely fast at training, these methods can barely achieve an interactive ren-de ring frame rate with around 15 FPS. In this paper, we use 3D Gaussian Splatting and learn a non-rigid deformation network to reconstruct animatable clothed human avatars that can be trained within 30 minutes and rendered at real-time frame rates (50+ FPS). Given the explicit nature of our representation, we further introduce as-isometric-as-possible regularizations on both the Gaussian mean vectors and the covariance matrices, enhancing the generalization of our model on highly articulated unseen poses. Experi-mental results show that our method achieves comparable and even better performance compared to state-of-the-art approaches on animatable avatar creation from a monoc-ular input, while being 400x and 250x faster in training and inference, respectively. Please see our project page at https://neuralbodies.github.ioI3DGS-Avatar."
                    }
                ]
            }
        }
    },
    "2405.14473": {
        "paper_data": {
            "title": "Poisson Variational Autoencoder",
            "url": "http://arxiv.org/abs/2405.14473v1",
            "arxiv_id": "2405.14473",
            "authors": [
                "Hadi Vafaii",
                "Dekel Galor",
                "Jacob L. Yates"
            ],
            "abstract": "Variational autoencoders (VAE) employ Bayesian inference to interpret sensory inputs, mirroring processes that occur in primate vision across both ventral (Higgins et al., 2021) and dorsal (Vafaii et al., 2023) pathways. Despite their success, traditional VAEs rely on continuous latent variables, which deviates sharply from the discrete nature of biological neurons. Here, we developed the Poisson VAE (P-VAE), a novel architecture that combines principles of predictive coding with a VAE that encodes inputs into discrete spike counts. Combining Poisson-distributed latent variables with predictive coding introduces a metabolic cost term in the model loss function, suggesting a relationship with sparse coding which we verify empirically. Additionally, we analyze the geometry of learned representations, contrasting the P-VAE to alternative VAE models. We find that the P-VAEencodes its inputs in relatively higher dimensions, facilitating linear separability of categories in a downstream classification task with a much better (5x) sample efficiency. Our work provides an interpretable computational framework to study brain-like sensory processing and paves the way for a deeper understanding of perception as an inferential process.",
            "introduction": "   1 Introduction  The study of artificial neural networks (ANN) and neuroscience has always been closely linked, driving advancements in both fields [mcculloch1943logical, churchland1988perspectives, thomas2008connectionist, kriegeskorte2015deep, lindsay2021convolutional, marino2022predictive]. Despite the close proximity of the two fields, most ANN models deviate substantially from biological brains [bowers2023deep, wichmann2023are]. A major challenge is designing models that not only perform well computationally but also exhibit \\saybrain-like structure and function. This is seen both as a goal for improving ANNs [zador2023catalyzing, sinz2019engineering, hassabis2017neuroscience], and better understanding biological brains [lindsay2021convolutional, kriegeskorte2015deep, kanwisher2023using, richards2022application, richards2019deep, barrett2019analyzing], which has recently been referred to as the neuroconnectionist research programme [doerig2023neuroconnectionist].   Drawing from neuroscience, a major guiding idea is that perception is a process of inference [alhazen, helmholtz1867handbuch], where the brain constructs a representation of the external world by inferring the causes of sensory inputs [lee2003hierarchical, olshausen2014, boring1946perception, friston2010free]. This concept is mirrored in \\saygenerative AI where models learn the generative process underlying their inputs [bond2021deep, chan2024tutorial, zhao2023survey]. However, in this vein, there is a tension between small well-understood models that are directly inspired by cortex, such as sparse coding [olshausen1996emergence] and predictive coding [rao1999predictive], and deep generative models that perform well [rombach2022high, karras2019style, vahdat2020nvae, child2021vdvae].   The variational autoencoder (VAE; [kingma2014auto, rezende2014stochastic]) model family is a promising candidate for neuroconnectionist goals for multiple reasons. First, VAEs learn probabilistic generative models of their inputs and are grounded in Bayesian probability theory, providing a solid theoretical foundation that directly incorporates the concept of perceptual inference [marino2022predictive, helmholtz1867handbuch]. Second, the VAE model family, specifically hierarchical VAEs, is broad with other generative models, such as diffusion models, understood as special cases of hierarchical VAEs [kingma2023understanding, ldm_tutorial_2023, kingma2021variational]. Finally, VAEs learn representations that are similar to cortex [higgins2021unsupervised, vafaii2023hierarchical, csikor2023topdown], exhibit cortex-like topographic organization [keller2021modeling, keller2021topographic], and make perceptual errors that mimic those of humans [storrs2021unsupervised], indicating a significant degree of neural, organizational, and psychophysical alignment with the brain.   However, standard VAEs diverge from brains in the way they encode information. Biological neurons fire all-or-none action potentials [adrian1932nobel], and are thought to represent information via firing rate [Adrian1926TheIP, perkel1968neural, barlow1972single, zohary1994correlated, rieke1999spikes]. These firing rates must be positive and generate discrete \\sayspike counts, which exhibit conditionally Poisson-like statistics in small counting windows [teich1989fractal, rieke1999spikes, goris2014partitioning]. In contrast, VAEs are typically parameterized with real-valued, continuous, Gaussian distributions [kingma2019introduction].   Contributions.  In this work, we address this discrepancy by introducing the Poisson Variational Autoencoder (\ud835\udcab\ud835\udcab\\operatorname{\\mathcal{P}}caligraphic_P-VAE), a novel architecture that combines perceptual inference with two other inspirations from neuroscience (Fig.\u00a0). First, that information is encoded in the rates of discrete spike counts, which are approximately Poisson-distributed on short time intervals. And second, that feedforward connections encode deviations from expectations contained in feedback connections (Fig.\u00a01a; [rao1999predictive, gilbert2013top]). We introduce a reparameterization trick for Poisson samples (Algorithm\u00a01), and derive the evidence lower bound (ELBO) objective for the \ud835\udcab\ud835\udcab\\operatorname{\\mathcal{P}}caligraphic_P-VAE (eq.\u00a03). Overall, we believe \ud835\udcab\ud835\udcab\\operatorname{\\mathcal{P}}caligraphic_P-VAE introduces a promising new model at the intersection of computational neuroscience and machine learning that offers several appealing features over existing VAE architectures:     \u2022  The \ud835\udcab\ud835\udcab\\operatorname{\\mathcal{P}}caligraphic_P-VAE loss derivation (eq.\u00a03) naturally results in a metabolic cost term that penalizes high firing rates, such that \ud835\udcab\ud835\udcab\\operatorname{\\mathcal{P}}caligraphic_P-VAE with a linear decoder",
            "references": [
                {
                    "title": "Tutorial on Diffusion Models for Imaging and Vision",
                    "abstract": "The astonishing growth of generative tools in recent years has empowered many exciting applications in text-to-image generation and text-to-video generation. The underlying principle behind these generative tools is the concept of diffusion, a particular sampling mechanism that has overcome some shortcomings that were deemed difficult in the previous approaches. The goal of this tutorial is to discuss the essential ideas underlying the diffusion models. The target audience of this tutorial includes undergraduate and graduate students who are interested in doing research on diffusion models or applying these models to solve other problems."
                },
                {
                    "title": "Predictive coding networks for temporal prediction",
                    "abstract": "One of the key problems the brain faces is inferring the state of the world from a sequence of dynamically changing stimuli, and it is not yet clear how the sensory system achieves this task. A well-established computational framework for describing perceptual processes in the brain is provided by the theory of predictive coding. Although the original proposals of predictive coding have discussed temporal prediction, later work developing this theory mostly focused on static stimuli, and key questions on neural implementation and computational properties of temporal predictive coding networks remain open. Here, we address these questions and present a formulation of the temporal predictive coding model that can be naturally implemented in recurrent networks, in which activity dynamics rely only on local inputs to the neurons, and learning only utilises local Hebbian plasticity. Additionally, we show that temporal predictive coding networks can approximate the performance of the Kalman filter in predicting behaviour of linear systems, and behave as a variant of a Kalman filter which does not track its own subjective posterior variance. Importantly, temporal predictive coding networks can achieve similar accuracy as the Kalman filter without performing complex mathematical operations, but just employing simple computations that can be implemented by biological networks. Moreover, when trained with natural dynamic inputs, we found that temporal predictive coding can produce Gabor-like, motion-sensitive receptive fields resembling those observed in real neurons in visual areas. In addition, we demonstrate how the model can be effectively generalized to nonlinear systems. Overall, models presented in this paper show how biologically plausible circuits can predict future stimuli and may guide research on understanding specific neural circuits in brain areas involved in temporal prediction. Author summary While significant advances have been made in the neuroscience of how the brain processes static stimuli, the time dimension has often been relatively neglected. However, time is crucial since the stimuli perceived by our senses typically dynamically vary in time, and the cortex needs to make sense of these changing inputs. This paper describes a computational model of cortical networks processing temporal stimuli. This model is able to infer and track the state of the environment based on noisy inputs, and predict future sensory stimuli. By ensuring that these predictions match the incoming stimuli, the model is able to learn the structure and statistics of its temporal inputs and produces responses of neurons resembling those in the brain. The model may help in further understanding neural circuits in sensory cortical areas."
                },
                {
                    "title": "A doubly stochastic renewal framework for partitioning spiking variability",
                    "abstract": "The firing rate is a prevalent concept used to describe neural computations, but estimating dynamically changing firing rates from irregular spikes is challenging. An inhomogeneous Poisson process, the standard model for partitioning firing rate and spiking irregularity, cannot account for diverse spike statistics observed across neurons. We introduce a doubly stochastic renewal point process, a flexible mathematical framework for partitioning spiking variability, which captures the broad spectrum of spiking irregularity from periodic to super-Poisson. We validate our partitioning framework using intracellular voltage recordings and develop a method for estimating spiking irregularity from data. We find that the spiking irregularity of cortical neurons decreases from sensory to association areas and is nearly constant for each neuron under many conditions but can also change across task epochs. A spiking network model shows that spiking irregularity depends on connectivity and can change with external input. These results help improve the precision of estimating firing rates on single trials and constrain mechanistic models of neural circuits."
                },
                {
                    "title": "Task structure and nonlinearity jointly determine learned representational geometry",
                    "abstract": "The utility of a learned neural representation depends on how well its geometry supports performance in downstream tasks. This geometry depends on the structure of the inputs, the structure of the target outputs, and the architecture of the network. By studying the learning dynamics of networks with one hidden layer, we discovered that the network's activation function has an unexpectedly strong impact on the representational geometry: Tanh networks tend to learn representations that reflect the structure of the target outputs, while ReLU networks retain more information about the structure of the raw inputs. This difference is consistently observed across a broad class of parameterized tasks in which we modulated the degree of alignment between the geometry of the task inputs and that of the task labels. We analyzed the learning dynamics in weight space and show how the differences between the networks with Tanh and ReLU nonlinearities arise from the asymmetric asymptotic behavior of ReLU, which leads feature neurons to specialize for different regions of input space. By contrast, feature neurons in Tanh networks tend to inherit the task label structure. Consequently, when the target outputs are low dimensional, Tanh networks generate neural representations that are more disentangled than those obtained with a ReLU nonlinearity. Our findings shed light on the interplay between input-output geometry, nonlinearity, and learned representations in neural networks."
                },
                {
                    "title": "High-performing neural network models of visual cortex benefit from high latent dimensionality",
                    "abstract": "Geometric descriptions of deep neural networks (DNNs) have the potential to uncover core representational principles of computational models in neuroscience. Here we examined the geometry of DNN models of visual cortex by quantifying the latent dimensionality of their natural image representations. A popular view holds that optimal DNNs compress their representations onto low-dimensional subspaces to achieve invariance and robustness, which suggests that better models of visual cortex should have lower dimensional geometries. Surprisingly, we found a strong trend in the opposite direction\u2014neural networks with high-dimensional image subspaces tended to have better generalization performance when predicting cortical responses to held-out stimuli in both monkey electrophysiology and human fMRI data. Moreover, we found that high dimensionality was associated with better performance when learning new categories of stimuli, suggesting that higher dimensional representations are better suited to generalize beyond their training domains. These findings suggest a general principle whereby high-dimensional geometry confers computational benefits to DNN models of visual cortex."
                },
                {
                    "title": "Top-down perceptual inference shaping the activity of early visual cortex",
                    "abstract": "Deep discriminative models provide remarkable insights into hierarchical processing in the brain by predicting neural activity along the visual pathway. However, these models differ from biological systems in their computational and architectural properties. Unlike biological systems, they require teaching signals for supervised learning. Moreover, they rely on feed-forward processing of stimuli, which contrasts with the extensive top-down connections in the ventral pathway. Here, we address both issues by developing a hierarchical deep generative model and show that it predicts an extensive set of experimental results in the primary and secondary visual cortices (V1 and V2). We show that the widely documented nonlinear sensitivity of V2 neurons to texture statistics is a consequence of learning a hierarchical representation of natural images. Further, we show that top-down influences are inherent to inference in hierarchical generative models, and explain neuronal responses to illusory contours and systematic modulations of noise correlations in V1."
                },
                {
                    "title": "Hierarchical temporal prediction captures motion processing along the visual pathway",
                    "abstract": "Visual neurons respond selectively to features that become increasingly complex from the eyes to the cortex. Retinal neurons prefer flashing spots of light, primary visual cortical (V1) neurons prefer moving bars, and those in higher cortical areas favor complex features like moving textures. Previously, we showed that V1 simple cell tuning can be accounted for by a basic model implementing temporal prediction \u2013 representing features that predict future sensory input from past input (Singer et al., 2018). Here, we show that hierarchical application of temporal prediction can capture how tuning properties change across at least two levels of the visual system. This suggests that the brain does not efficiently represent all incoming information; instead, it selectively represents sensory inputs that help in predicting the future. When applied hierarchically, temporal prediction extracts time-varying features that depend on increasingly high-level statistics of the sensory input."
                },
                {
                    "title": "Hierarchical VAEs provide a normative account of motion processing in the primate brain",
                    "abstract": "The relationship between perception and inference, as postulated by Helmholtz in the 19th century, is paralleled in modern machine learning by generative models like Variational Autoencoders (VAEs) and their hierarchical variants. Here, we evaluate the role of hierarchical inference and its alignment with brain function in the domain of motion perception. We first introduce a novel synthetic data framework, Retinal Optic Flow Learning (ROFL), which enables control over motion statistics and their causes. We then present a new hierarchical VAE and test it against alternative models on two downstream tasks: (i) predicting ground truth causes of retinal optic flow (e.g., self-motion); and (ii) predicting the responses of neurons in the motion processing pathway of primates. We manipulate the model architectures (hierarchical versus non-hierarchical), loss functions, and the causal structure of the motion stimuli. We find that hierarchical latent structure in the model leads to several improvements. First, it improves the linear decodability of ground truth factors and does so in a sparse and disentangled manner. Second, our hierarchical VAE outperforms previous state-of-the-art models in predicting neuronal responses and exhibits sparse latent-to-neuron relationships. These results depend on the causal structure of the world, indicating that alignment between brains and artificial neural networks depends not only on architecture but also on matching ecologically relevant stimulus statistics. Taken together, our results suggest that hierarchical Bayesian inference underlines the brain\u2019s understanding of the world, and hierarchical VAEs can effectively model this understanding."
                },
                {
                    "title": "What can 1.8 billion regressions tell us about the pressures shaping high-level visual representation in brains and machines?",
                    "abstract": "The rapid development and open-source release of highly performant computer vision models offers new potential for examining how different inductive biases impact representation learning and emergent alignment with the high-level human ventral visual system. Here, we assess a diverse set of 224 models, curated to enable controlled comparison of different model properties, testing their brain predictivity using large-scale functional magnetic resonance imaging data. We find that models with qualitatively different architectures (e.g. CNNs versus Transformers) and markedly different task objectives (e.g. purely visual contrastive learning versus vision-language alignment) achieve near equivalent degrees of brain predictivity, when other factors are held constant. Instead, variation across model visual training diets yields the largest, most consistent effect on emergent brain predictivity. Overarching model properties commonly suspected to increase brain predictivity (e.g. greater effective dimensionality; learnable parameter count) were not robust indicators across this more extensive survey. We highlight that standard model-to-brain linear re-weighting methods may be too flexible, as most performant models have very similar brain-predictivity scores, despite significant variation in their underlying representations. Broadly, our findings point to the importance of visual diet, challenge common assumptions about the methods used to link models to brains, and more concretely outline future directions for leveraging the full diversity of existing open-source models as tools to probe the common computational principles underlying biological and artificial visual systems."
                },
                {
                    "title": "A Survey of Large Language Models",
                    "abstract": "Language is essentially a complex, intricate system of human expressions governed by grammatical rules. It poses a significant challenge to develop capable AI algorithms for comprehending and grasping a language. As a major approach, language modeling has been widely studied for language understanding and generation in the past two decades, evolving from statistical language models to neural language models. Recently, pre-trained language models (PLMs) have been proposed by pre-training Transformer models over large-scale corpora, showing strong capabilities in solving various NLP tasks. Since researchers have found that model scaling can lead to performance improvement, they further study the scaling effect by increasing the model size to an even larger size. Interestingly, when the parameter scale exceeds a certain level, these enlarged language models not only achieve a significant performance improvement but also show some special abilities that are not present in small-scale language models. To discriminate the difference in parameter scale, the research community has coined the term large language models (LLM) for the PLMs of significant size. Recently, the research on LLMs has been largely advanced by both academia and industry, and a remarkable progress is the launch of ChatGPT, which has attracted widespread attention from society. The technical evolution of LLMs has been making an important impact on the entire AI community, which would revolutionize the way how we develop and use AI algorithms. In this survey, we review the recent advances of LLMs by introducing the background, key findings, and mainstream techniques. In particular, we focus on four major aspects of LLMs, namely pre-training, adaptation tuning, utilization, and capacity evaluation. Besides, we also summarize the available resources for developing LLMs and discuss the remaining issues for future directions."
                },
                {
                    "title": "SC-VAE: Sparse Coding-based Variational Autoencoder",
                    "abstract": "Learning rich data representations from unlabeled data is a key challenge towards applying deep learning algorithms in downstream tasks. Several variants of variational autoencoders (VAEs) have been proposed to learn compact data representations by encoding high-dimensional data in a lower dimensional space. Two main classes of VAEs methods may be distinguished depending on the characteristics of the meta-priors that are enforced in the representation learning step. The first class of methods derives a continuous encoding by assuming a static prior distribution in the latent space. The second class of methods learns instead a discrete latent representation using vector quantization (VQ) along with a codebook. However, both classes of methods suffer from certain challenges, which may lead to suboptimal image reconstruction results. The first class suffers from posterior collapse, whereas the second class suffers from codebook collapse. To address these challenges, we introduce a new VAE variant, termed sparse coding-based VAE with learned ISTA (SC-VAE), which integrates sparse coding within variational autoencoder framework. The proposed method learns sparse data representations that consist of a linear combination of a small number of predetermined orthogonal atoms. The sparse coding problem is solved using a learnable version of the iterative shrinkage thresholding algorithm (ISTA). Experiments on two image datasets demonstrate that our model achieves improved image reconstruction results compared to state-of-the-art methods. Moreover, we demonstrate that the use of learned sparse code vectors allows us to perform downstream tasks like image generation and unsupervised image segmentation through clustering image patches."
                },
                {
                    "title": "A polar prediction model for learning to represent visual transformations",
                    "abstract": "All organisms make temporal predictions, and their evolutionary fitness level depends on the accuracy of these predictions. In the context of visual perception, the motions of both the observer and objects in the scene structure the dynamics of sensory signals, allowing for partial prediction of future signals based on past ones. Here, we propose a self-supervised representation-learning framework that extracts and exploits the regularities of natural videos to compute accurate predictions. We motivate the polar architecture by appealing to the Fourier shift theorem and its group-theoretic generalization, and we optimize its parameters on next-frame prediction. Through controlled experiments, we demonstrate that this approach can discover the representation of simple transformation groups acting in data. When trained on natural video datasets, our framework achieves better prediction performance than traditional motion compensation and rivals conventional deep networks, while maintaining interpretability and speed. Furthermore, the polar computations can be restructured into components resembling normalized simple and direction-selective complex cell models of primate V1 neurons. Thus, polar prediction offers a principled framework for understanding how the visual system represents sensory inputs in a form that simplifies temporal prediction."
                },
                {
                    "title": "Understanding Diffusion Objectives as the ELBO with Simple Data Augmentation",
                    "abstract": "To achieve the highest perceptual quality, state-of-the-art diffusion models are optimized with objectives that typically look very different from the maximum likelihood and the Evidence Lower Bound (ELBO) objectives. In this work, we reveal that diffusion model objectives are actually closely related to the ELBO. Specifically, we show that all commonly used diffusion model objectives equate to a weighted integral of ELBOs over different noise levels, where the weighting depends on the specific objective used. Under the condition of monotonic weighting, the connection is even closer: the diffusion objective then equals the ELBO, combined with simple data augmentation, namely Gaussian noise perturbation. We show that this condition holds for a number of state-of-the-art diffusion models. In experiments, we explore new monotonic weightings and demonstrate their effectiveness, achieving state-of-the-art FID scores on the high-resolution ImageNet benchmark."
                },
                {
                    "title": "Deep problems with neural network models of human vision",
                    "abstract": "Abstract Deep neural networks (DNNs) have had extraordinary successes in classifying photographic images of objects and are often described as the best models of biological vision. This conclusion is largely based on three sets of findings: (1) DNNs are more accurate than any other model in classifying images taken from various datasets, (2) DNNs do the best job in predicting the pattern of human errors in classifying objects taken from various behavioral datasets, and (3) DNNs do the best job in predicting brain signals in response to images taken from various brain datasets (e.g., single cell responses or fMRI data). However, these behavioral and brain datasets do not test hypotheses regarding what features are contributing to good predictions and we show that the predictions may be mediated by DNNs that share little overlap with biological vision. More problematically, we show that DNNs account for almost no results from psychological research. This contradicts the common claim that DNNs are good, let alone the best, models of human object recognition. We argue that theorists interested in developing biologically plausible models of human vision need to direct their attention to explaining psychological findings. More generally, theorists need to build models that explain the results of experiments that manipulate independent variables designed to test hypotheses rather than compete on making the best predictions. We conclude by briefly summarizing various promising modeling approaches that focus on psychological data."
                },
                {
                    "title": "Disentanglement with Biological Constraints: A Theory of Functional Cell Types",
                    "abstract": "Neurons in the brain are often finely tuned for specific task variables. Moreover, such disentangled representations are highly sought after in machine learning. Here we mathematically prove that simple biological constraints on neurons, namely nonnegativity and energy efficiency in both activity and weights, promote such sought after disentangled representations by enforcing neurons to become selective for single factors of task variation. We demonstrate these constraints lead to disentanglement in a variety of tasks and architectures, including variational autoencoders. We also use this theory to explain why the brain partitions its cells into distinct cell types such as grid and object-vector cells, and also explain when the brain instead entangles representations in response to entangled task factors. Overall, this work provides a mathematical understanding of why single neurons in the brain often represent single human-interpretable factors, and steps towards an understanding task structure shapes the structure of brain representation."
                },
                {
                    "title": "Amortized Variational Inference: A Systematic Review",
                    "abstract": "The core principle of Variational Inference (VI) is to convert the statistical inference problem of computing complex posterior probability densities into a tractable optimization problem. This property enables VI to be faster than several sampling-based techniques. However, the traditional VI algorithm is not scalable to large data sets and is unable to readily infer out-of-bounds data points without re-running the optimization process. Recent developments in the field, like stochastic-, black box-, and amortized-VI, have helped address these issues. Generative modeling tasks nowadays widely make use of amortized VI for its efficiency and scalability, as it utilizes a parameterized function to learn the approximate posterior density parameters. In this paper, we review the mathematical foundations of various VI techniques to form the basis for understanding amortized VI. Additionally, we provide an overview of the recent trends that address several issues of amortized VI, such as the amortization gap, generalization issues, inconsistent representation learning, and posterior collapse. Finally, we analyze alternate divergence measures that improve VI optimization."
                },
                {
                    "title": "Tutorial on Amortized Optimization",
                    "abstract": "Optimization is a ubiquitous modeling tool and is often deployed in settings which repeatedly solve similar instances of the same problem. Amortized optimization methods use learning to predict the solutions to problems in these settings, exploiting the shared structure between similar problem instances. These methods have been crucial in variational inference and reinforcement learning and are capable of solving optimization problems many orders of magnitudes times faster than traditional optimization methods that do not use amortization. This tutorial presents an introduction to the amortized optimization foundations behind these advancements and overviews their applications in variational inference, sparse coding, gradient-based meta-learning, control, reinforcement learning, convex optimization, optimal transport, and deep equilibrium networks. The source code for this tutorial is available at https://github.com/facebookresearch/amortized-optimization-tutorial."
                },
                {
                    "title": "High-Resolution Image Synthesis with Latent Diffusion Models",
                    "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs."
                },
                {
                    "title": "Modeling Category-Selective Cortical Regions with Topographic Variational Autoencoders",
                    "abstract": "Category-selectivity in the brain describes the observation that certain spatially localized areas of the cerebral cortex tend to respond robustly and selectively to stimuli from specific limited categories. One of the most well known examples of category-selectivity is the Fusiform Face Area (FFA), an area of the inferior temporal cortex in primates which responds preferentially to images of faces when compared with objects or other generic stimuli. In this work, we leverage the newly introduced Topographic Variational Autoencoder to model the emergence of such localized category-selectivity in an unsupervised manner. Experimentally, we demonstrate our model yields spatially dense neural clusters selective to faces, bodies, and places through visualized maps of Cohen's d metric. We compare our model with related supervised approaches, namely the Topographic Deep Artificial Neural Network (TDANN) of Lee et al., and discuss both theoretical and empirical similarities. Finally, we show preliminary results suggesting that our model yields a nested spatial hierarchy of increasingly abstract categories, analogous to observations from the human ventral temporal cortex."
                },
                {
                    "title": "Fully Spiking Variational Autoencoder",
                    "abstract": "Spiking neural networks (SNNs) can be run on neuromorphic devices with ultra-high speed and ultra-low energy consumption because of their binary and event-driven nature. Therefore, SNNs are expected to have various applications, including as generative models being running on edge devices to create high-quality images. In this study, we build a variational autoencoder (VAE) with SNN to enable image generation. VAE is known for its stability among generative models; recently, its quality advanced. In vanilla VAE, the latent space is represented as a normal distribution, and floating-point calculations are required in sampling. However, this is not possible in SNNs because all features must be binary time series data. Therefore, we constructed the latent space with an autoregressive SNN model, and randomly selected samples from its output to sample the latent variables. This allows the latent variables to follow the Bernoulli process and allows variational learning. Thus, we build the Fully Spiking Variational Autoencoder where all modules are constructed with SNN. To the best of our knowledge, we are the first to build a VAE only with SNN layers. We experimented with several datasets, and confirmed that it can generate images with the same or better quality compared to conventional ANNs. The code is available at https://github.com/kamata1729/FullySpikingVAE."
                },
                {
                    "title": "Neuromodulatory Control of Early Visual Processing in Macaque.",
                    "abstract": "Visual processing is dynamically controlled by multiple neuromodulatory molecules that modify the responsiveness of neurons and the strength of the connections between them. In particular, modulatory control of processing in the lateral geniculate nucleus of the thalamus, V1, and V2 will alter the outcome of all subsequent processing of visual information, including the extent to and manner in which individual inputs contribute to perception and decision making and are stored in memory. This review addresses five small-molecule neuromodulators-acetylcholine, dopamine, serotonin, noradrenaline, and histamine-considering the structural basis for their action, and the effects of their release, in the early visual pathway of the macaque monkey. Traditionally, neuromodulators are studied in isolation and in discrete circuits; this review makes a case for considering the joint action of modulatory molecules and differences in modulatory effects across brain areas as a better means of understanding the diverse roles that these molecules serve."
                },
                {
                    "title": "Topographic VAEs learn Equivariant Capsules",
                    "abstract": "In this work we seek to bridge the concepts of topographic organization and equivariance in neural networks. To accomplish this, we introduce the Topographic VAE: a novel method for efficiently training deep generative models with topographically organized latent variables. We show that such a model indeed learns to organize its activations according to salient characteristics such as digit class, width, and style on MNIST. Furthermore, through topographic organization over time (i.e. temporal coherence), we demonstrate how predefined latent space transformation operators can be encouraged for observed transformed input sequences -- a primitive form of unsupervised learned equivariance. We demonstrate that this model successfully learns sets of approximately equivariant features (i.e.\"capsules\") directly from sequences and achieves higher likelihood on correspondingly transforming test sequences. Equivariance is verified quantitatively by measuring the approximate commutativity of the inference network and the sequence transformations. Finally, we demonstrate approximate equivariance to complex transformations, expanding upon the capabilities of existing group equivariant neural networks."
                },
                {
                    "title": "Variational Diffusion Models",
                    "abstract": "Diffusion-based generative models have demonstrated a capacity for perceptually impressive synthesis, but can they also be great likelihood-based models? We answer this in the affirmative, and introduce a family of diffusion-based generative models that obtain state-of-the-art likelihoods on standard image density estimation benchmarks. Unlike other diffusion-based models, our method allows for efficient optimization of the noise schedule jointly with the rest of the model. We show that the variational lower bound (VLB) simplifies to a remarkably short expression in terms of the signal-to-noise ratio of the diffused data, thereby improving our theoretical understanding of this model class. Using this insight, we prove an equivalence between several models proposed in the literature. In addition, we show that the continuous-time VLB is invariant to the noise schedule, except for the signal-to-noise ratio at its endpoints. This enables us to learn a noise schedule that minimizes the variance of the resulting VLB estimator, leading to faster optimization. Combining these advances with architectural improvements, we obtain state-of-the-art likelihoods on image density estimation benchmarks, outperforming autoregressive models that have dominated these benchmarks for many years, with often significantly faster optimization. In addition, we show how to use the model as part of a bits-back compression scheme, and demonstrate lossless compression rates close to the theoretical optimum. Code is available at https://github.com/google-research/vdm ."
                },
                {
                    "title": "Deep Generative Modelling: A Comparative Review of VAEs, GANs, Normalizing Flows, Energy-Based and Autoregressive Models",
                    "abstract": "Deep generative models are a class of techniques that train deep neural networks to model the distribution of training samples. Research has fragmented into various interconnected approaches, each of which make trade-offs including run-time, diversity, and architectural restrictions. In particular, this compendium covers energy-based models, variational autoencoders, generative adversarial networks, autoregressive models, normalizing flows, in addition to numerous hybrid approaches. These techniques are compared and contrasted, explaining the premises behind each and how they are interrelated, while reviewing current state-of-the-art advances and implementations."
                },
                {
                    "title": "Very Deep VAEs Generalize Autoregressive Models and Can Outperform Them on Images",
                    "abstract": "We present a hierarchical VAE that, for the first time, outperforms the PixelCNN in log-likelihood on all natural image benchmarks. We begin by observing that VAEs can actually implement autoregressive models, and other, more efficient generative models, if made sufficiently deep. Despite this, autoregressive models have traditionally outperformed VAEs. We test if insufficient depth explains the performance gap by by scaling a VAE to greater stochastic depth than previously explored and evaluating it on CIFAR-10, ImageNet, and FFHQ. We find that, in comparison to the PixelCNN, these very deep VAEs achieve higher likelihoods, use fewer parameters, generate samples thousands of times faster, and are more easily applied to high-resolution images. We visualize the generative process and show the VAEs learn efficient hierarchical visual representations. We release our source code and models at https://github.com/openai/vdvae."
                },
                {
                    "title": "NVAE: A Deep Hierarchical Variational Autoencoder",
                    "abstract": "Normalizing flows, autoregressive models, variational autoencoders (VAEs), and deep energy-based models are among competing likelihood-based frameworks for deep generative learning. Among them, VAEs have the advantage of fast and tractable sampling and easy-to-access encoding networks. However, they are currently outperformed by other models such as normalizing flows and autoregressive models. While the majority of the research in VAEs is focused on the statistical challenges, we explore the orthogonal direction of carefully designing neural architectures for hierarchical VAEs. We propose Nouveau VAE (NVAE), a deep hierarchical VAE built for image generation using depth-wise separable convolutions and batch normalization. NVAE is equipped with a residual parameterization of Normal distributions and its training is stabilized by spectral regularization. We show that NVAE achieves state-of-the-art results among non-autoregressive likelihood-based models on the MNIST, CIFAR-10, CelebA 64, and CelebA HQ datasets and it provides a strong baseline on FFHQ. For example, on CIFAR-10, NVAE pushes the state-of-the-art from 2.98 to 2.91 bits per dimension, and it produces high-quality images on CelebA HQ. To the best of our knowledge, NVAE is the first successful VAE applied to natural images as large as 256$\\times$256 pixels. The source code is available at this https URL ."
                },
                {
                    "title": "Fast and Flexible Temporal Point Processes with Triangular Maps",
                    "abstract": "Temporal point process (TPP) models combined with recurrent neural networks provide a powerful framework for modeling continuous-time event data. While such models are flexible, they are inherently sequential and therefore cannot benefit from the parallelism of modern hardware. By exploiting the recent developments in the field of normalizing flows, we design TriTPP -- a new class of non-recurrent TPP models, where both sampling and likelihood computation can be done in parallel. TriTPP matches the flexibility of RNN-based methods but permits orders of magnitude faster sampling. This enables us to use the new model for variational inference in continuous-time discrete-state systems. We demonstrate the advantages of the proposed framework on synthetic and real-world datasets."
                },
                {
                    "title": "Convolutional Neural Networks as a Model of the Visual System: Past, Present, and Future",
                    "abstract": "Abstract Convolutional neural networks (CNNs) were inspired by early findings in the study of biological vision. They have since become successful tools in computer vision and state-of-the-art models of both neural activity and behavior on visual tasks. This review highlights what, in the context of CNNs, it means to be a good model in computational neuroscience and the various ways models can provide insight. Specifically, it covers the origins of CNNs and the methods by which we validate them as models of biological vision. It then goes on to elaborate on what we can learn about biological vision by understanding and experimenting on CNNs and discusses emerging opportunities for the use of CNNs in vision research beyond basic object recognition."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Predictive Coding, Variational Autoencoders, and Biological Connections",
                    "abstract": "Abstract We present a review of predictive coding, from theoretical neuroscience, and variational autoencoders, from machine learning, identifying the common origin and mathematical framework underlying both areas. As each area is prominent within its respective field, more firmly connecting these areas could prove useful in the dialogue between neuroscience and machine learning. After reviewing each area, we discuss two possible correspondences implied by this perspective: cortical pyramidal dendrites as analogous to (nonlinear) deep networks and lateral inhibition as analogous to normalizing flows. These connections may provide new directions for further investigations in each field."
                },
                {
                    "title": "Variational Sparse Coding",
                    "abstract": "Unsupervised discovery of interpretable features and controllable generation with highdimensional data are currently major challenges in machine learning, with applications \nin data visualisation, clustering and artificial \ndata synthesis. We propose a model based \non variational auto-encoders (VAEs) in which \ninterpretation is induced through latent space \nsparsity with a mixture of Spike and Slab distributions as prior. We derive an evidence \nlower bound for this model and propose a specific training method for recovering disentangled features as sparse elements in latent vectors. In our experiments, we demonstrate superior disentanglement performance to standard \nVAE approaches when an estimate of the number of true sources of variation is not available \nand objects display different combinations of \nattributes. Furthermore, the new model provides unique capabilities, such as recovering \nfeature exploitation, synthesising samples that \nshare attributes with a given input object and \ncontrolling both discrete and continuous features upon generation."
                },
                {
                    "title": "Monte Carlo Gradient Estimation in Machine Learning",
                    "abstract": "This paper is a broad and accessible survey of the methods we have at our disposal for Monte Carlo gradient estimation in machine learning and across the statistical sciences: the problem of computing the gradient of an expectation of a function with respect to parameters defining the distribution that is integrated; the problem of sensitivity analysis. In machine learning research, this gradient problem lies at the core of many learning problems, in supervised, unsupervised and reinforcement learning. We will generally seek to rewrite such gradients in a form that allows for Monte Carlo estimation, allowing them to be easily and efficiently used and analysed. We explore three strategies--the pathwise, score function, and measure-valued gradient estimators--exploring their historical developments, derivation, and underlying assumptions. We describe their use in other fields, show how they are related and can be combined, and expand on their possible generalisations. Wherever Monte Carlo gradient estimators have been derived and deployed in the past, important advances have followed. A deeper and more widely-held understanding of this problem will lead to further advances, and it is these advances that we wish to support."
                },
                {
                    "title": "An Introduction to Variational Autoencoders",
                    "abstract": "Variational autoencoders provide a principled framework for learning deep latent-variable models and corresponding inference models. In this work, we provide an introduction to variational autoencoders and some important extensions."
                },
                {
                    "title": "Cyclical Annealing Schedule: A Simple Approach to Mitigating KL Vanishing",
                    "abstract": "Variational autoencoders (VAE) with an auto-regressive decoder have been applied for many natural language processing (NLP) tasks. VAE objective consists of two terms, the KL regularization term and the reconstruction term, balanced by a weighting hyper-parameter \\beta. One notorious training difficulty is that the KL term tends to vanish. In this paper we study different scheduling schemes for \\beta, and show that KL vanishing is caused by the lack of good latent codes in training decoder at the beginning of optimization. To remedy the issue, we propose a cyclical annealing schedule, which simply repeats the process of increasing \\beta multiple times. This new procedure allows us to learn more meaningful latent codes progressively by leveraging the results of previous learning cycles as warm re-restart. The effectiveness of cyclical annealing schedule is validated on a broad range of NLP tasks, including language modeling, dialog response generation and semi-supervised text classification."
                },
                {
                    "title": "Sparse deep predictive coding captures contour integration capabilities of the early visual system",
                    "abstract": "Both neurophysiological and psychophysical experiments have pointed out the crucial role of recurrent and feedback connections to process context-dependent information in the early visual cortex. While numerous models have accounted for feedback effects at either neural or representational level, none of them were able to bind those two levels of analysis. Is it possible to describe feedback effects at both levels using the same model? We answer this question by combining Predictive Coding (PC) and Sparse Coding (SC) into a hierarchical and convolutional framework applied to realistic problems. In the Sparse Deep Predictive Coding (SDPC) model, the SC component models the internal recurrent processing within each layer, and the PC component describes the interactions between layers using feedforward and feedback connections. Here, we train a 2-layered SDPC on two different databases of images, and we interpret it as a model of the early visual system (V1 & V2). We first demonstrate that once the training has converged, SDPC exhibits oriented and localized receptive fields in V1 and more complex features in V2. Second, we analyze the effects of feedback on the neural organization beyond the classical receptive field of V1 neurons using interaction maps. These maps are similar to association fields and reflect the Gestalt principle of good continuation. We demonstrate that feedback signals reorganize interaction maps and modulate neural activity to promote contour integration. Third, we demonstrate at the representational level that the SDPC feedback connections are able to overcome noise in input images. Therefore, the SDPC captures the association field principle at the neural level which results in a better reconstruction of blurred images at the representational level."
                },
                {
                    "title": "A Style-Based Generator Architecture for Generative Adversarial Networks",
                    "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces."
                },
                {
                    "title": "Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations",
                    "abstract": "The key idea behind the unsupervised learning of disentangled representations is that real-world data is generated by a few explanatory factors of variation which can be recovered by unsupervised learning algorithms. In this paper, we provide a sober look at recent progress in the field and challenge some common assumptions. We first theoretically show that the unsupervised learning of disentangled representations is fundamentally impossible without inductive biases on both the models and the data. Then, we train more than 12000 models covering most prominent methods and evaluation metrics in a reproducible large-scale experimental study on seven different data sets. We observe that while the different methods successfully enforce properties ``encouraged'' by the corresponding losses, well-disentangled models seemingly cannot be identified without supervision. Furthermore, increased disentanglement does not seem to lead to a decreased sample complexity of learning for downstream tasks. Our results suggest that future work on disentanglement learning should be explicit about the role of inductive biases and (implicit) supervision, investigate concrete benefits of enforcing disentanglement of the learned representations, and consider a reproducible experimental setup covering several data sets."
                },
                {
                    "title": "Sparse-Coding Variational Auto-Encoders",
                    "abstract": "The sparse coding model posits that the visual system has evolved to efficiently code natural stimuli using a sparse set of features from an overcomplete dictionary. The original sparse coding model suffered from two key limitations, however: (1) computing the neural response to an image patch required minimizing a nonlinear objective function via recurrent dynamics; (2) fitting relied on approximate inference methods that ignored uncertainty. Although subsequent work has developed several methods to overcome these obstacles, we propose a novel solution inspired by the variational auto-encoder (VAE) framework. We introduce the sparse-coding variational auto-encoder (SVAE), which augments the sparse coding model with a probabilistic recognition model parametrized by a deep neural network. This recognition model provides a neurally plausible feedforward implementation for the mapping from image patches to neural activities, and enables a principled method for fitting the sparse coding model to data via maximization of the evidence lower bound (ELBO). The SVAE differs from standard VAEs in three key respects: the latent representation is overcomplete (there are more latent dimensions than image pixels), the prior is sparse or heavy-tailed instead of Gaussian, and the decoder network is a linear projection instead of a deep network. We fit the SVAE to natural image data under different assumed prior distributions, and show that it obtains higher test performance than previous fitting methods. Finally, we examine the response properties of the recognition network and show that it captures important nonlinear properties of neurons in the early visual pathway."
                },
                {
                    "title": "Inference Suboptimality in Variational Autoencoders",
                    "abstract": "Amortized inference allows latent-variable models trained via variational learning to scale to large datasets. The quality of approximate inference is determined by two factors: a) the capacity of the variational distribution to match the true posterior and b) the ability of the recognition network to produce good variational parameters for each datapoint. We examine approximate inference in variational autoencoders in terms of these factors. We find that divergence from the true posterior is often due to imperfect recognition networks, rather than the limited complexity of the approximating distribution. We show that this is due partly to the generator learning to accommodate the choice of approximation. Furthermore, we show that the parameters used to increase the expressiveness of the approximation play a role in generalizing inference rather than simply improving the complexity of the approximation."
                },
                {
                    "title": "Neural Discrete Representation Learning",
                    "abstract": "Learning useful representations without supervision remains a key challenge in machine learning. In this paper, we propose a simple yet powerful generative model that learns such discrete representations. Our model, the Vector Quantised-Variational AutoEncoder (VQ-VAE), differs from VAEs in two key ways: the encoder network outputs discrete, rather than continuous, codes; and the prior is learnt rather than static. In order to learn a discrete latent representation, we incorporate ideas from vector quantisation (VQ). Using the VQ method allows the model to circumvent issues of \"posterior collapse\" -- where the latents are ignored when they are paired with a powerful autoregressive decoder -- typically observed in the VAE framework. Pairing these representations with an autoregressive prior, the model can generate high quality images, videos, and speech as well as doing high quality speaker conversion and unsupervised learning of phonemes, providing further evidence of the utility of the learnt representations."
                },
                {
                    "title": "Fixing a Broken ELBO",
                    "abstract": "Recent work in unsupervised representation learning has focused on learning deep directed latent-variable models. Fitting these models by maximizing the marginal likelihood or evidence is typically intractable, thus a common approximation is to maximize the evidence lower bound (ELBO) instead. However, maximum likelihood training (whether exact or approximate) does not necessarily result in a good latent representation, as we demonstrate both theoretically and empirically. In particular, we derive variational lower and upper bounds on the mutual information between the input and the latent variable, and use these bounds to derive a rate-distortion curve that characterizes the tradeoff between compression and reconstruction accuracy. Using this framework, we demonstrate that there is a family of models with identical ELBO, but different quantitative and qualitative characteristics. Our framework also suggests a simple new method to ensure that latent variable models with powerful stochastic decoders do not ignore their latent code."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Categorical Reparameterization with Gumbel-Softmax",
                    "abstract": "Categorical variables are a natural choice for representing discrete structure in the world. However, stochastic neural networks rarely use categorical latent variables due to the inability to backpropagate through samples. In this work, we present an efficient gradient estimator that replaces the non-differentiable sample from a categorical distribution with a differentiable sample from a novel Gumbel-Softmax distribution. This distribution has the essential property that it can be smoothly annealed into a categorical distribution. We show that our Gumbel-Softmax estimator outperforms state-of-the-art gradient estimators on structured output prediction and unsupervised generative modeling tasks with categorical latent variables, and enables large speedups on semi-supervised classification."
                },
                {
                    "title": "The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables",
                    "abstract": "The reparameterization trick enables optimizing large scale stochastic computation graphs via gradient descent. The essence of the trick is to refactor each stochastic node into a differentiable function of its parameters and a random variable with fixed distribution. After refactoring, the gradients of the loss propagated by the chain rule through the graph are low variance unbiased estimators of the gradients of the expected loss. While many continuous random variables have such reparameterizations, discrete random variables lack useful reparameterizations due to the discontinuous nature of discrete states. In this work we introduce Concrete random variables---continuous relaxations of discrete random variables. The Concrete distribution is a new family of distributions with closed form densities and a simple reparameterization. Whenever a discrete stochastic node of a computation graph can be refactored into a one-hot bit representation that is treated continuously, Concrete stochastic nodes can be used with automatic differentiation to produce low-variance biased gradients of objectives (including objectives that depend on the log-probability of latent stochastic nodes) on the corresponding discrete graph. We demonstrate the effectiveness of Concrete relaxations on density estimation and structured prediction tasks using neural networks."
                },
                {
                    "title": "SGDR: Stochastic Gradient Descent with Warm Restarts",
                    "abstract": "Restart techniques are common in gradient-free optimization to deal with multimodal functions. Partial warm restarts are also gaining popularity in gradient-based optimization to improve the rate of convergence in accelerated gradient schemes to deal with ill-conditioned functions. In this paper, we propose a simple warm restart technique for stochastic gradient descent to improve its anytime performance when training deep neural networks. We empirically study its performance on the CIFAR-10 and CIFAR-100 datasets, where we demonstrate new state-of-the-art results at 3.14% and 16.21%, respectively. We also demonstrate its advantages on a dataset of EEG recordings and on a downsampled version of the ImageNet dataset. Our source code is available at this https URL"
                },
                {
                    "title": "Deep Predictive Coding Networks for Video Prediction and Unsupervised Learning",
                    "abstract": "While great strides have been made in using deep learning algorithms to solve supervised learning tasks, the problem of unsupervised learning - leveraging unlabeled examples to learn about the structure of a domain - remains a difficult unsolved challenge. Here, we explore prediction of future frames in a video sequence as an unsupervised learning rule for learning about the structure of the visual world. We describe a predictive neural network (\"PredNet\") architecture that is inspired by the concept of \"predictive coding\" from the neuroscience literature. These networks learn to predict future frames in a video sequence, with each layer in the network making local predictions and only forwarding deviations from those predictions to subsequent network layers. We show that these networks are able to robustly learn to predict the movement of synthetic (rendered) objects, and that in doing so, the networks learn internal representations that are useful for decoding latent object parameters (e.g. pose) that support object recognition with fewer training views. We also show that these networks can scale to complex natural image streams (car-mounted camera videos), capturing key aspects of both egocentric movement and the movement of objects in the visual scene, and the representation learned in this setting is useful for estimating the steering angle. Altogether, these results suggest that prediction represents a powerful framework for unsupervised learning, allowing for implicit learning of object and scene structure."
                },
                {
                    "title": "Ladder Variational Autoencoders",
                    "abstract": "Variational Autoencoders are powerful models for unsupervised learning. However deep models with several layers of dependent stochastic variables are difficult to train which limits the improvements obtained using these highly expressive models. We propose a new inference model, the Ladder Variational Autoencoder, that recursively corrects the generative distribution by a data dependent approximate likelihood in a process resembling the recently proposed Ladder Network. We show that this model provides state of the art predictive log-likelihood and tighter log-likelihood lower bound compared to the purely bottom-up inference in layered Variational Autoencoders and other generative models. We provide a detailed analysis of the learned hierarchical latent representation and show that our new inference model is qualitatively different and utilizes a deeper more distributed hierarchy of latent variables. Finally, we observe that batch normalization and deterministic warm-up (gradually turning on the KL-term) are crucial for training variational models with many stochastic layers."
                },
                {
                    "title": "Generating Sentences from a Continuous Space",
                    "abstract": "The standard recurrent neural network language model (RNNLM) generates sentences one word at a time and does not work from an explicit global sentence representation. In this work, we introduce and study an RNN-based variational autoencoder generative model that incorporates distributed latent representations of entire sentences. This factorization allows it to explicitly model holistic properties of sentences such as style, topic, and high-level syntactic features. Samples from the prior over these sentence representations remarkably produce diverse and well-formed sentences through simple deterministic decoding. By examining paths through this latent space, we are able to generate coherent novel sentences that interpolate between known sentences. We present techniques for solving the difficult learning problem presented by this model, demonstrate its effectiveness in imputing missing words, explore many interesting properties of the model's latent sentence space, and present negative results on the use of the model in language modeling."
                },
                {
                    "title": "Deep neural networks: a new framework for modelling biological vision and brain information processing",
                    "abstract": "Recent advances in neural network modelling have enabled major strides in computer vision and other artificial intelligence applications. Human-level visual recognition abilities are coming within reach of artificial systems. Artificial neural networks are inspired by the brain and their computations could be implemented in biological neurons. Convolutional feedforward networks, which now dominate computer vision, take further inspiration from the architecture of the primate visual hierarchy. However, the current models are designed with engineering goals and not to model brain computations. Nevertheless, initial studies comparing internal representations between these models and primate brains find surprisingly similar representational spaces. With human-level performance no longer out of reach, we are entering an exciting new era, in which we will be able to build neurobiologically faithful feedforward and recurrent computational models of how biological brains perform high-level feats of intelligence, including vision."
                },
                {
                    "title": "Adam: A Method for Stochastic Optimization",
                    "abstract": "We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
                },
                {
                    "title": "Stochastic Backpropagation and Approximate Inference in Deep Generative Models",
                    "abstract": "We marry ideas from deep neural networks and approximate Bayesian inference to derive a generalised class of deep, directed generative models, endowed with a new algorithm for scalable inference and learning. Our algorithm introduces a recognition model to represent approximate posterior distributions, and that acts as a stochastic encoder of the data. We develop stochastic back-propagation -- rules for back-propagation through stochastic variables -- and use this to develop an algorithm that allows for joint optimisation of the parameters of both the generative and recognition model. We demonstrate on several real-world data sets that the model generates realistic samples, provides accurate imputations of missing data and is a useful tool for high-dimensional data visualisation."
                },
                {
                    "title": "Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation",
                    "abstract": "Stochastic neurons and hard non-linearities can be useful for a number of reasons in deep learning models, but in many cases they pose a challenging problem: how to estimate the gradient of a loss function with respect to the input of such stochastic or non-smooth neurons? I.e., can we \"back-propagate\" through these stochastic neurons? We examine this question, existing approaches, and compare four families of solutions, applicable in different settings. One of them is the minimum variance unbiased gradient estimator for stochatic binary neurons (a special case of the REINFORCE algorithm). A second approach, introduced here, decomposes the operation of a binary stochastic neuron into a stochastic binary part and a smooth differentiable part, which approximates the expected effect of the pure stochatic binary neuron to first order. A third approach involves the injection of additive or multiplicative noise in a computational graph that is otherwise differentiable. A fourth approach heuristically copies the gradient with respect to the stochastic output directly as an estimator of the gradient with respect to the sigmoid argument (we call this the straight-through estimator). To explore a context where these estimators are useful, we consider a small-scale version of {\\em conditional computation}, where sparse stochastic units form a distributed representation of gaters that can turn off in combinatorially many ways large chunks of the computation performed in the rest of the neural network. In this case, it is important that the gating units produce an actual 0 most of the time. The resulting sparsity can be potentially be exploited to greatly reduce the computational cost of large deep networks for which conditional computation would be useful."
                },
                {
                    "title": "Possible Principles Underlying the Transformations of Sensory Messages",
                    "abstract": "and"
                },
                {
                    "title": "Sparse coding in striate and extrastriate visual cortex.",
                    "abstract": "Theoretical studies of mammalian cortex argue that efficient neural codes should be sparse. However, theoretical and experimental studies have used different definitions of the term \"sparse\" leading to three assumptions about the nature of sparse codes. First, codes that have high lifetime sparseness require few action potentials. Second, lifetime-sparse codes are also population-sparse. Third, neural codes are optimized to maximize lifetime sparseness. Here, we examine these assumptions in detail and test their validity in primate visual cortex. We show that lifetime and population sparseness are not necessarily correlated and that a code may have high lifetime sparseness regardless of how many action potentials it uses. We measure lifetime sparseness during presentation of natural images in three areas of macaque visual cortex, V1, V2, and V4. We find that lifetime sparseness does not increase across the visual hierarchy. This suggests that the neural code is not simply optimized to maximize lifetime sparseness. We also find that firing rates during a challenging visual task are higher than theoretical values based on metabolic limits and that responses in V1, V2, and V4 are well-described by exponential distributions. These findings are consistent with the hypothesis that neurons are optimized to maximize information transmission subject to metabolic constraints on mean firing rate."
                },
                {
                    "title": "Learning Fast Approximations of Sparse Coding",
                    "abstract": "In Sparse Coding (SC), input vectors are reconstructed using a sparse linear combination of basis vectors. SC has become a popular method for extracting features from data. For a given input, SC minimizes a quadratic reconstruction error with an L1 penalty term on the code. The process is often too slow for applications such as real-time pattern recognition. We proposed two versions of a very fast algorithm that produces approximate estimates of the sparse code that can be used to compute good visual features, or to initialize exact iterative algorithms. The main idea is to train a non-linear, feed-forward predictor with a specific architecture and a fixed depth to produce the best possible approximation of the sparse code. A version of the method, which can be seen as a trainable version of Li and Osher's coordinate descent method, is shown to produce approximate solutions with 10 times less computation than Li and Os-her's for the same approximation error. Unlike previous proposals for sparse code predictors, the system allows a kind of approximate \"explaining away\" to take place during inference. The resulting predictor is differentiable and can be included into globally-trained recognition systems."
                },
                {
                    "title": "Sparse Coding via Thresholding and Local Competition in Neural Circuits",
                    "abstract": "While evidence indicates that neural systems may be employing sparse approximations to represent sensed stimuli, the mechanisms underlying this ability are not understood. We describe a locally competitive algorithm (LCA) that solves a collection of sparse coding principles minimizing a weighted combination of mean-squared error and a coefficient cost function. LCAs are designed to be implemented in a dynamical system composed of many neuron-like elements operating in parallel. These algorithms use thresholding functions to induce local (usually one-way) inhibitory competitions between nodes to produce sparse representations. LCAs produce coefficients with sparsity levels comparable to the most popular centralized sparse coding algorithms while being readily suited for neural implementation. Additionally, LCA coefficients for video sequences demonstrate inertial properties that are both qualitatively and quantitatively more regular (i.e., smoother and more predictable) than the coefficients produced by greedy algorithms."
                },
                {
                    "title": "Connectionist models of cognition.",
                    "abstract": "Book synopsis: This book is a definitive reference source for the growing, increasingly more important, and interdisciplinary field of computational cognitive modeling, that is, computational psychology. It combines breadth of coverage with definitive statements by leading scientists in this field. Research in computational cognitive modeling explores the essence of cognition through developing detailed, process-based understanding by specifying computational mechanisms, structures, and processes. Computational models provide both conceptual clarity and precision at the same time. This book substantiates this approach through overviews and many examples."
                },
                {
                    "title": "Sparse Representation of Sounds in the Unanesthetized Auditory Cortex",
                    "abstract": "How do neuronal populations in the auditory cortex represent acoustic stimuli? Although sound-evoked neural responses in the anesthetized auditory cortex are mainly transient, recent experiments in the unanesthetized preparation have emphasized subpopulations with other response properties. To quantify the relative contributions of these different subpopulations in the awake preparation, we have estimated the representation of sounds across the neuronal population using a representative ensemble of stimuli. We used cell-attached recording with a glass electrode, a method for which single-unit isolation does not depend on neuronal activity, to quantify the fraction of neurons engaged by acoustic stimuli (tones, frequency modulated sweeps, white-noise bursts, and natural stimuli) in the primary auditory cortex of awake head-fixed rats. We find that the population response is sparse, with stimuli typically eliciting high firing rates (>20 spikes/second) in less than 5% of neurons at any instant. Some neurons had very low spontaneous firing rates (<0.01 spikes/second). At the other extreme, some neurons had driven rates in excess of 50 spikes/second. Interestingly, the overall population response was well described by a lognormal distribution, rather than the exponential distribution that is often reported. Our results represent, to our knowledge, the first quantitative evidence for sparse representations of sounds in the unanesthetized auditory cortex. Our results are compatible with a model in which most neurons are silent much of the time, and in which representations are composed of small dynamic subsets of highly active neurons."
                },
                {
                    "title": "Distance Metric Learning for Large Margin Nearest Neighbor Classification",
                    "abstract": "The accuracy of k-nearest neighbor (kNN) classification depends significantly on the metric used to compute distances between different examples. In this paper, we show how to learn a Mahalanobis distance metric for kNN classification from labeled examples. The Mahalanobis metric can equivalently be viewed as a global linear transformation of the input space that precedes kNN classification using Euclidean distances. In our approach, the metric is trained with the goal that the k-nearest neighbors always belong to the same class while examples from different classes are separated by a large margin. As in support vector machines (SVMs), the margin criterion leads to a convex optimization based on the hinge loss. Unlike learning in SVMs, however, our approach requires no modification or extension for problems in multiway (as opposed to binary) classification. In our framework, the Mahalanobis distance metric is obtained as the solution to a semidefinite program. On several data sets of varying size and difficulty, we find that metrics trained in this way lead to significant improvements in kNN classification. Sometimes these results can be further improved by clustering the training examples and learning an individual metric within each cluster. We show how to learn and combine these local metrics in a globally integrated manner."
                },
                {
                    "title": "A theory of cortical responses",
                    "abstract": "This article concerns the nature of evoked brain responses and the principles underlying their generation. We start with the premise that the sensory brain has evolved to represent or infer the causes of changes in its sensory inputs. The problem of inference is well formulated in statistical terms. The statistical fundaments of inference may therefore afford important constraints on neuronal implementation. By formulating the original ideas of Helmholtz on perception, in terms of modern-day statistical theories, one arrives at a model of perceptual inference and learning that can explain a remarkable range of neurobiological facts. It turns out that the problems of inferring the causes of sensory input (perceptual inference) and learning the relationship between input and cause (perceptual learning) can be resolved using exactly the same principle. Specifically, both inference and learning rest on minimizing the brain's free energy, as defined in statistical physics. Furthermore, inference and learning can proceed in a biologically plausible fashion. Cortical responses can be seen as the brain\u2019s attempt to minimize the free energy induced by a stimulus and thereby encode the most likely cause of that stimulus. Similarly, learning emerges from changes in synaptic efficacy that minimize the free energy, averaged over all stimuli encountered. The underlying scheme rests on empirical Bayes and hierarchical models of how sensory input is caused. The use of hierarchical models enables the brain to construct prior expectations in a dynamic and context-sensitive fashion. This scheme provides a principled way to understand many aspects of cortical organization and responses. The aim of this article is to encompass many apparently unrelated anatomical, physiological and psychophysical attributes of the brain within a single theoretical perspective. In terms of cortical architectures, the theoretical treatment predicts that sensory cortex should be arranged hierarchically, that connections should be reciprocal and that forward and backward connections should show a functional asymmetry (forward connections are driving, whereas backward connections are both driving and modulatory). In terms of synaptic physiology, it predicts associative plasticity and, for dynamic models, spike-timing-dependent plasticity. In terms of electrophysiology, it accounts for classical and extra classical receptive field effects and long-latency or endogenous components of evoked cortical responses. It predicts the attenuation of responses encoding prediction error with perceptual learning and explains many phenomena such as repetition suppression, mismatch negativity (MMN) and the P300 in electroencephalography. In psychophysical terms, it accounts for the behavioural correlates of these physiological phenomena, for example, priming and global precedence. The final focus of this article is on perceptual learning as measured with the MMN and the implications for empirical studies of coupling among cortical areas using evoked sensory responses."
                },
                {
                    "title": "An iterative thresholding algorithm for linear inverse problems with a sparsity constraint",
                    "abstract": "We consider linear inverse problems where the solution is assumed to have a sparse expansion on an arbitrary preassigned orthonormal basis. We prove that replacing the usual quadratic regularizing penalties by weighted \ud835\udcc1p\u2010penalties on the coefficients of such expansions, with 1 \u2264 p \u2264 2, still regularizes the problem. Use of such \ud835\udcc1p\u2010penalized problems with p < 2 is often advocated when one expects the underlying ideal noiseless solution to have a sparse expansion with respect to the basis under consideration. To compute the corresponding regularized solutions, we analyze an iterative algorithm that amounts to a Landweber iteration with thresholding (or nonlinear shrinkage) applied at each iteration step. We prove that this algorithm converges in norm. \u00a9 2004 Wiley Periodicals, Inc."
                },
                {
                    "title": "Hierarchical Bayesian inference in the visual cortex.",
                    "abstract": "Traditional views of visual processing suggest that early visual neurons in areas V1 and V2 are static spatiotemporal filters that extract local features from a visual scene. The extracted information is then channeled through a feedforward chain of modules in successively higher visual areas for further analysis. Recent electrophysiological recordings from early visual neurons in awake behaving monkeys reveal that there are many levels of complexity in the information processing of the early visual cortex, as seen in the long-latency responses of its neurons. These new findings suggest that activity in the early visual cortex is tightly coupled and highly interactive with the rest of the visual system. They lead us to propose a new theoretical setting based on the mathematical framework of hierarchical Bayesian inference for reasoning about the visual system. In this framework, the recurrent feedforward/feedback loops in the cortex serve to integrate top-down contextual priors and bottom-up observations so as to implement concurrent probabilistic inference along the visual hierarchy. We suggest that the algorithms of particle filtering and Bayesian-belief propagation might model these interactive cortical computations. We review some recent neurophysiological evidences that support the plausibility of these ideas."
                },
                {
                    "title": "Sparse coding and decorrelation in primary visual cortex during natural vision.",
                    "abstract": "Theoretical studies suggest that primary visual cortex (area V1) uses a sparse code to efficiently represent natural scenes. This issue was investigated by recording from V1 neurons in awake behaving macaques during both free viewing of natural scenes and conditions simulating natural vision. Stimulation of the nonclassical receptive field increases the selectivity and sparseness of individual V1 neurons, increases the sparseness of the population response distribution, and strongly decorrelates the responses of neuron pairs. These effects are due to both excitatory and suppressive modulation of the classical receptive field by the nonclassical receptive field and do not depend critically on the spatiotemporal structure of the stimuli. During natural vision, the classical and nonclassical receptive fields function together to form a sparse representation of the visual world. This sparse code may be computationally efficient for both early vision and higher visual processing."
                },
                {
                    "title": "Spikes: Exploring the Neural Code",
                    "abstract": "Our perception of the world is driven by input from the sensory nerves. This input arrives encoded as sequences of identical spikes. Much of neural computation involves processing these spike trains. What does it mean to say that a certain set of spikes is the right answer to a computational problem? In what sense does a spike train convey information about the sensory world? Spikes begins by providing precise formulations of these and related questions about the representation of sensory signals in neural spike trains. The answers to these questions are then pursued in experiments on sensory neurons.The authors invite the reader to play the role of a hypothetical observer inside the brain who makes decisions based on the incoming spike trains. Rather than asking how a neuron responds to a given stimulus, the authors ask how the brain could make inferences about an unknown stimulus from a given neural response. The flavor of some problems faced by the organism is captured by analyzing the way in which the observer can make a running reconstruction of the sensory stimulus as it evolves in time. These ideas are illustrated by examples from experiments on several biological systems. Intended for neurobiologists with an interest in mathematical analysis of neural data as well as the growing number of physicists and mathematicians interested in information processing by \"real\" nervous systems, Spikes provides a self-contained review of relevant concepts in information theory and statistical decision theory. A quantitative framework is used to pose precise questions about the structure of the neural code. These questions in turn influence both the design and analysis of experiments on sensory neurons."
                },
                {
                    "title": "The Helmholtz Machine",
                    "abstract": "Discovering the structure inherent in a set of patterns is a fundamental aim of statistical inference or learning. One fruitful approach is to build a parameterized stochastic generative model, independent draws from which are likely to produce the patterns. For all but the simplest generative models, each pattern can be generated in exponentially many ways. It is thus intractable to adjust the parameters to maximize the probability of the observed patterns. We describe a way of finessing this combinatorial explosion by maximizing an easily computed lower bound on the probability of the observations. Our method can be viewed as a form of hierarchical self-supervised learning that may relate to the function of bottom-up and top-down cortical processing pathways."
                },
                {
                    "title": "Perspectives on cognitive neuroscience.",
                    "abstract": "How is it that we can perceive, learn and be aware of the world? The development of new techniques for studying large-scale brain activity, together with insights from computational modeling and a better understanding of cognitive processes, have opened the door for collaborative research that could lead to major advances in our understanding of ourselves."
                },
                {
                    "title": "Fractal character of the auditory neural spike train",
                    "abstract": "Long-counting-time pulse-number distributions (PNDs) were measured from a broad variety of cat primary auditory fibers using different tone and noise stimuli, counting times T, and number of samples N/sub T/. Whereas short-counting-time PNDs (T approximately 50 ms) manifest the presence of spike pairs (an enhancement of even- over odd-count probabilities), the irregular shapes of the long-counting-time PNDs (T>or approximately=0.1 s) reveal that the underlying sequence of action potentials consists of spike clusters when viewed on a longer time scale. For all units measured, the count variance-to-mean ratio (Fano factor) F(T) varied little over some 90 dB change in the stimulus level. On the other hand, F(T) increased substantially as T and/or N/sub T/ were increased, corresponding to the capture of larger and larger spike clusters in the counting time. A relationship is developed between the Fano-time function F(T) and the normalized coincidence rate function, g ( tau ) versus delay time tau . The behavior of the PNDs and the scale invariance implicit in fractional-power-law relationships suggest that the neural events on all primary auditory fibers exhibit fractal behavior for sufficiently large times (sufficiently low frequencies).<<ETX>>"
                },
                {
                    "title": "Predictive coding: a fresh view of inhibition in the retina",
                    "abstract": "Interneurons exhibiting centre-surround antagonism within their receptive fields are commonly found in peripheral visual pathways. We propose that this organization enables the visual system to encode spatial detail in a manner that minimizes the deleterious effects of intrinsic noise, by exploiting the spatial correlation that exists within natural scenes. The antagonistic surround takes a weighted mean of the signals in neighbouring receptors to generate a statistical prediction of the signal at the centre. The predicted value is subtracted from the actual centre signal, thus minimizing the range of outputs transmitted by the centre. In this way the entire dynamic range of the interneuron can be devoted to encoding a small range of intensities, thus rendering fine detail detectable against intrinsic noise injected at later stages in processing. This predictive encoding scheme also reduces spatial redundancy, thereby enabling the array of interneurons to transmit a larger number of distinguishable images, taking into account the expected structure of the visual world. The profile of the required inhibitory field is derived from statistical estimation theory. This profile depends strongly upon the signal: noise ratio and weakly upon the extent of lateral spatial correlation. The receptive fields that are quantitatively predicted by the theory resemble those of X-type retinal ganglion cells and show that the inhibitory surround should become weaker and more diffuse at low intensities. The latter property is unequivocally demonstrated in the first-order interneurons of the fly\u2019s compound eye. The theory is extended to the time domain to account for the phasic responses of fly interneurons. These comparisons suggest that, in the early stages of processing, the visual system is concerned primarily with coding the visual image to protect against subsequent intrinsic noise, rather than with reconstructing the scene or extracting specific features from it. The treatment emphasizes that a neuron\u2019s dynamic range should be matched to both its receptive field and the statistical properties of the visual pattern expected within this field. Finally, the analysis is synthetic because it is an extension of the background suppression hypothesis (Barlow & Levick 1976), satisfies the redundancy reduction hypothesis (Barlow 1961 a, b) and is equivalent to deblurring under certain conditions (Ratliff 1965)."
                },
                {
                    "title": "Single Units and Sensation: A Neuron Doctrine for Perceptual Psychology?",
                    "abstract": "The problem discussed is the relationship between the firing of single neurons in sensory pathways and subjectively experienced sensations. The conclusions are formulated as the following five dogmas: To understand nervous function one needs to look at interactions at a cellular level, rather than either a more macroscopic or microscopic level, because behaviour depends upon the organized pattern of these intercellular interactions. The sensory system is organized to achieve as complete a representation of the sensory stimulus as possible with the minimum number of active neurons. Trigger features of sensory neurons are matched to redundant patterns of stimulation by experience as well as by developmental processes. Perception corresponds to the activity of a small selection from the very numerous high-level neurons, each of which corresponds to a pattern of external events of the order of complexity of the events symbolized by a word. High impulse frequency in such neurons corresponds to high certainty that the trigger feature is present. The development of the concepts leading up to these speculative dogmas, their experimental basis, and some of their limitations are discussed."
                },
                {
                    "title": "Some informational aspects of visual perception.",
                    "abstract": "xThe ideas of information theory are at present stimulating many different areas of psychological inquiry. In providing techniques for quantifying situations which have hitherto been difficult or impossible to quantify, they suggest new and more precise ways of conceptualizing these situations (see Miller [12] for a general discussion and bibliography). Events ordered in time are particularly amenable to informational analysis; thus language sequences are being extensively studied, and other sequences, such as those of music, plainly invite research. In this paper I shall indicate some of the ways in which the concepts and techniques of information theory may clarify our understanding of visual perception. When we begin to consider perception as an information-handling process, it quickly becomes clear that much of the information received by any higher organism is redundant. Sensory events are highly interdependent in both space and time: if we know at a given moment the states of a limited number of receptors (i.e., whether they are firing or not firing), we can make better-than-chance inferences with respect to the prior and subsequent states of these receptors, and also with respect to the present, prior, and subsequent states of other receptors. The preceding statement, taken in its broadest im1 The experimental work for this study was performed as part of the United States Air Force Human Resources Research and Development Program. The opinions and conclusions contained in this report are those of the author. They are not to be construed as reflecting the views or indorsement of the Department of the Air Force. plications, is precisely equivalent to an assertion that the world as we know it is lawful. In the present discussion, however, we shall restrict our attention to special types of lawfulness which may exist in space at a fixed time, and which seem particularly relevant to processes of visual perception."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "q-bio.NC"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we design a variational autoencoder that better aligns with the biological principles of neural information encoding, specifically by incorporating Poisson-distributed spike counts?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for bridging the gap between artificial neural networks and biological neural systems. By developing a model that reflects the brain's structure and function, we can enhance our understanding of both fields. This research could lead to more efficient and interpretable machine learning models, potentially advancing applications in areas such as cognitive computing, neuroprosthetics, and brain-computer interfaces. Furthermore, it may inspire future research directions that explore the intersection of neuroscience and machine learning, fostering a deeper understanding of perceptual inference.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in accurately modeling the discrete nature of neural firing and the statistical properties of spike counts, which differ fundamentally from the continuous representations used in standard VAEs. Naive approaches that simply adapt existing VAE architectures may fail to capture the essential characteristics of biological neurons, such as the all-or-none firing mechanism and the Poisson-like statistics of spike counts. Additionally, the need to incorporate feedback mechanisms and metabolic cost considerations adds layers of complexity that require innovative solutions.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on continuous representations in neural networks, overlooking the discrete and probabilistic nature of biological neural encoding. Existing models have not adequately addressed the need for a framework that integrates both perceptual inference and the statistical properties of neural firing. Barriers include a lack of theoretical foundations that connect these concepts and the difficulty in deriving effective training objectives for such models. Our approach differs by introducing the Poisson Variational Autoencoder (\ud835\udcab\ud835\udcab\\operatorname{\\mathcal{P}}caligraphic_P-VAE), which explicitly incorporates these biological principles into its architecture and training process.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of the Poisson Variational Autoencoder (\ud835\udcab\ud835\udcab\\operatorname{\\mathcal{P}}caligraphic_P-VAE), which utilizes a reparameterization trick for Poisson samples and derives a new evidence lower bound (ELBO) objective that includes a metabolic cost term. We will evaluate the model using benchmark datasets relevant to generative modeling and perceptual tasks,"
            }
        },
        "author_data": {
            "982ad544-c17e-4a82-b9fc-a3af1d265953": {
                "pk": "982ad544-c17e-4a82-b9fc-a3af1d265953",
                "name": "Hadi Vafaii",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "7cf84e58-a24e-4e1b-bd37-74e62d88fd59": {
                "pk": "7cf84e58-a24e-4e1b-bd37-74e62d88fd59",
                "name": "Dekel Galor",
                "collaborators": [
                    "Laura Waller",
                    "Guanghan Meng",
                    "Martin S. Banks",
                    "Ruiming Cao",
                    "Amit Kohli",
                    "Jacob L Yates"
                ],
                "domain": [
                    "Computer Vision",
                    "Motion Artifacts",
                    "Event Cameras",
                    "Image Processing"
                ],
                "publications": [
                    {
                        "title": "BiPMAP: A Toolbox for Predictions of Perceived Motion Artifacts on Modern Displays",
                        "abstract": "Presenting dynamic scenes without incurring motion artifacts visible to observers requires sustained effort from the display industry. A tool that predicts motion artifacts and simulates artifact elimination through optimizing the display configuration is highly desired to guide the design and manufacture of modern displays. Despite the popular demands, there is no such tool available in the market. In this study, we deliver an interactive toolkit, Binocular Perceived Motion Artifact Predictor (BiPMAP), as an executable file with GPU acceleration. BiPMAP accounts for an extensive collection of user-defined parameters and directly visualizes a variety of motion artifacts by presenting the perceived continuous and sampled moving stimuli side-by-side. For accurate artifact predictions, BiPMAP utilizes a novel model of the human contrast sensitivity function to effectively imitate the frequency modulation of the human visual system. In addition, BiPMAP is capable of deriving various in-plane motion artifacts for 2D displays and depth distortion in 3D stereoscopic displays."
                    },
                    {
                        "title": "Noise2Image: Noise-Enabled Static Scene Recovery for Event Cameras",
                        "abstract": "Event cameras capture changes of intensity over time as a stream of 'events' and generally cannot measure intensity itself; hence, they are only used for imaging dynamic scenes. However, fluctuations due to random photon arrival inevitably trigger noise events, even for static scenes. While previous efforts have been focused on filtering out these undesirable noise events to improve signal quality, we find that, in the photon-noise regime, these noise events are correlated with the static scene intensity. We analyze the noise event generation and model its relationship to illuminance. Based on this understanding, we propose a method, called Noise2Image, to leverage the illuminance-dependent noise characteristics to recover the static parts of a scene, which are otherwise invisible to event cameras. We experimentally collect a dataset of noise events on static scenes to train and validate Noise2Image. Our results show that Noise2Image can robustly recover intensity images solely from noise events, providing a novel approach for capturing static scenes in event cameras, without additional hardware."
                    }
                ]
            },
            "accc63d8-9da4-434b-92f9-8ff8c5883615": {
                "pk": "accc63d8-9da4-434b-92f9-8ff8c5883615",
                "name": "Jacob L. Yates",
                "collaborators": [
                    "Stephen L. Keeley",
                    "David M. Zoltowski",
                    "Yiyi Yu",
                    "Spencer L. Smith",
                    "Jonathan W. Pillow"
                ],
                "domain": [
                    "Gaussian Processes",
                    "Neural Data Analysis",
                    "Statistical Inference"
                ],
                "publications": [
                    {
                        "title": "Efficient non-conjugate Gaussian process factor models for spike count data using polynomial approximations",
                        "abstract": "Gaussian Process Factor Analysis (GPFA) has been broadly applied to the problem of identifying smooth, low-dimensional temporal structure underlying large-scale neural recordings. However, spike trains are non-Gaussian, which motivates combining GPFA with discrete observation models for binned spike count data. The drawback to this approach is that GPFA priors are not conjugate to count model likelihoods, which makes inference challenging. Here we address this obstacle by introducing a fast, approximate inference method for non-conjugate GPFA models. Our approach uses orthogonal second-order polynomials to approximate the nonlinear terms in the non-conjugate log-likelihood, resulting in a method we refer to as \\textit{polynomial approximate log-likelihood} (PAL) estimators. This approximation allows for accurate closed-form evaluation of marginal likelihoods and fast numerical optimization for parameters and hyperparameters. We derive PAL estimators for GPFA models with binomial, Poisson, and negative binomial observations and find the PAL estimation is highly accurate, and achieves faster convergence times compared to existing state-of-the-art inference methods. We also find that PAL hyperparameters can provide sensible initialization for black box variational inference (BBVI), which improves BBVI accuracy. We demonstrate that PAL estimators achieve fast and accurate extraction of latent structure from multi-neuron spike train data."
                    }
                ]
            }
        }
    },
    "2310.02117": {
        "paper_data": {
            "title": "Symmetric Single Index Learning",
            "url": "http://arxiv.org/abs/2310.02117v1",
            "arxiv_id": "2310.02117",
            "authors": [
                "Aaron Zweig",
                "Joan Bruna"
            ],
            "abstract": "Few neural architectures lend themselves to provable learning with gradient based methods. One popular model is the single-index model, in which labels are produced by composing an unknown linear projection with a possibly unknown scalar link function. Learning this model with SGD is relatively well-understood, whereby the so-called information exponent of the link function governs a polynomial sample complexity rate. However, extending this analysis to deeper or more complicated architectures remains challenging.   In this work, we consider single index learning in the setting of symmetric neural networks. Under analytic assumptions on the activation and maximum degree assumptions on the link function, we prove that gradient flow recovers the hidden planted direction, represented as a finitely supported vector in the feature space of power sum polynomials. We characterize a notion of information exponent adapted to our setting that controls the efficiency of learning.",
            "introduction": "   1 Introduction  Quantifying the advantage of neural networks over simpler learning systems remains a primary question in deep learning theory. Specifically, understanding their ability to discover relevant low-dimensional features out of high-dimensional inputs is a particularly important topic of study. One facet of the challenge is explicitly characterizing the evolution of neural network weights through gradient-based methods, owing to the nonconvexity of the optimization landscape.   The single index setting, long studied in economics and biostatistics\u00a0[25] offers the simplest setting where non-linear feature learning can be characterized explicitly. In this setting, functions of the form x\u21a6f\u2062(\u27e8x,\u03b8*\u27e9)maps-to\ud835\udc65\ud835\udc53\ud835\udc65superscript\ud835\udf03x\\mapsto f(\\langle x,\\theta^{*}\\rangle)italic_x \u21a6 italic_f ( \u27e8 italic_x , italic_\u03b8 start_POSTSUPERSCRIPT * end_POSTSUPERSCRIPT \u27e9 ) where \u03b8*\u2208\ud835\udcaed\u22121superscript\ud835\udf03subscript\ud835\udcae\ud835\udc511\\theta^{*}\\in\\mathcal{S}_{d-1}italic_\u03b8 start_POSTSUPERSCRIPT * end_POSTSUPERSCRIPT \u2208 caligraphic_S start_POSTSUBSCRIPT italic_d - 1 end_POSTSUBSCRIPT represents a hidden direction in high-dimensional space, and f\ud835\udc53fitalic_f a certain non-linear link function, are learned via a student with an identical architecture x\u21a6f\u2062(\u27e8x,\u03b8\u27e9)maps-to\ud835\udc65\ud835\udc53\ud835\udc65\ud835\udf03x\\mapsto f(\\langle x,\\theta\\rangle)italic_x \u21a6 italic_f ( \u27e8 italic_x , italic_\u03b8 \u27e9 ), under certain data distribution assumptions, such as Gaussian data. Gradient flow and gradient descent\u00a0[29, 4, 13] in this setting can be analyzed by reducing the high-dimensional dynamics of \u03b8\ud835\udf03\\thetaitalic_\u03b8 to dimension-free dynamics of appropriate summary statistics, given in this case by the scalar correlation \u27e8\u03b8,\u03b8*\u27e9\ud835\udf03superscript\ud835\udf03\\langle\\theta,\\theta^{*}\\rangle\u27e8 italic_\u03b8 , italic_\u03b8 start_POSTSUPERSCRIPT * end_POSTSUPERSCRIPT \u27e9.   The efficiency of gradient methods in this setting, measured either in continuous time or independent samples, is controlled by two main properties. First, the correlation initialization, which typically scales as 1d1\ud835\udc51\\frac{1}{\\sqrt{d}}divide start_ARG 1 end_ARG start_ARG square-root start_ARG italic_d end_ARG end_ARG for standard assumptions. Second, the information exponent sfsubscript\ud835\udc60\ud835\udc53s_{f}italic_s start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT of f\ud835\udc53fitalic_f \u00a0[4, 13, 7, 10, 11, 2], which measures the number of effective vanishing moments of the link function \u2014 leading to a sample complexity of the form O\u2062(ds\u22121)\ud835\udc42superscript\ud835\udc51\ud835\udc601O(d^{s-1})italic_O ( italic_d start_POSTSUPERSCRIPT italic_s - 1 end_POSTSUPERSCRIPT ) for generic values of s\ud835\udc60sitalic_s.   While this basic setup has been extended along certain directions, e.g. relaxing the structure on the input data distribution [29, 9], considering the multi-index counterpart [11, 1, 2, 3], or learning the link function with semi-parametric methods [7, 21], they are all fundamentally associated with fully-connected shallow neural networks. Such architecture, for all its rich mathematical structure, also comes with important shortcomings. In particular, it is unable to account for predefined symmetries in the target function that the learner wishes to exploit. This requires specialized neural architectures enforcing particular invariances, setting up novel technical challenges to carry out the program outlined above.   In this work, we consider arguably the easiest form of symmetry, given by permutation invariance. The primary architecture for this invariance is DeepSets\u00a0[30], which is necessarily three layers by definition and therefore not a simple extension of the two layer setting. In order to quantify the notion of \u2018symmetric\u2019 feature learning in this setting, we introduce a symmetric single index target, and analyze the ability of gradient descent over a DeepSet architecture to recover it. Under appropriate assumptions on the model, initialization and data distribution, we combine the previous analyses with tools from symmetric polynomial theory to characterize the dynamics of this learning problem. Our primary theorem is a proof of efficient learning under gradient",
            "references": [
                {
                    "title": "On Single Index Models beyond Gaussian Data",
                    "abstract": "Sparse high-dimensional functions have arisen as a rich framework to study the behavior of gradient-descent methods using shallow neural networks, showcasing their ability to perform feature learning beyond linear models. Amongst those functions, the simplest are single-index models $f(x) = \\phi( x \\cdot \\theta^*)$, where the labels are generated by an arbitrary non-linear scalar link function $\\phi$ applied to an unknown one-dimensional projection $\\theta^*$ of the input data. By focusing on Gaussian data, several recent works have built a remarkable picture, where the so-called information exponent (related to the regularity of the link function) controls the required sample complexity. In essence, these tools exploit the stability and spherical symmetry of Gaussian distributions. In this work, building from the framework of \\cite{arous2020online}, we explore extensions of this picture beyond the Gaussian setting, where both stability or symmetry might be violated. Focusing on the planted setting where $\\phi$ is known, our main results establish that Stochastic Gradient Descent can efficiently recover the unknown direction $\\theta^*$ in the high-dimensional regime, under assumptions that extend previous works \\cite{yehudai2020learning,wu2022learning}."
                },
                {
                    "title": "Beyond NTK with Vanilla Gradient Descent: A Mean-Field Analysis of Neural Networks with Polynomial Width, Samples, and Time",
                    "abstract": "Despite recent theoretical progress on the non-convex optimization of two-layer neural networks, it is still an open question whether gradient descent on neural networks without unnatural modifications can achieve better sample complexity than kernel methods. This paper provides a clean mean-field analysis of projected gradient flow on polynomial-width two-layer neural networks. Different from prior works, our analysis does not require unnatural modifications of the optimization algorithm. We prove that with sample size $n = O(d^{3.1})$ where $d$ is the dimension of the inputs, the network trained with projected gradient flow converges in $\\text{poly}(d)$ time to a non-trivial error that is not achievable by kernel methods using $n \\ll d^4$ samples, hence demonstrating a clear separation between unmodified gradient descent and NTK. As a corollary, we show that projected gradient descent with a positive learning rate and a polynomial number of iterations converges to low error with the same sample complexity."
                },
                {
                    "title": "Smoothing the Landscape Boosts the Signal for SGD: Optimal Sample Complexity for Learning Single Index Models",
                    "abstract": "We focus on the task of learning a single index model $\\sigma(w^\\star \\cdot x)$ with respect to the isotropic Gaussian distribution in $d$ dimensions. Prior work has shown that the sample complexity of learning $w^\\star$ is governed by the information exponent $k^\\star$ of the link function $\\sigma$, which is defined as the index of the first nonzero Hermite coefficient of $\\sigma$. Ben Arous et al. (2021) showed that $n \\gtrsim d^{k^\\star-1}$ samples suffice for learning $w^\\star$ and that this is tight for online SGD. However, the CSQ lower bound for gradient based methods only shows that $n \\gtrsim d^{k^\\star/2}$ samples are necessary. In this work, we close the gap between the upper and lower bounds by showing that online SGD on a smoothed loss learns $w^\\star$ with $n \\gtrsim d^{k^\\star/2}$ samples. We also draw connections to statistical analyses of tensor PCA and to the implicit regularization effects of minibatch SGD on empirical losses."
                },
                {
                    "title": "SGD learning on neural networks: leap complexity and saddle-to-saddle dynamics",
                    "abstract": "We investigate the time complexity of SGD learning on fully-connected neural networks with isotropic data. We put forward a complexity measure -- the leap -- which measures how\"hierarchical\"target functions are. For $d$-dimensional uniform Boolean or isotropic Gaussian data, our main conjecture states that the time complexity to learn a function $f$ with low-dimensional support is $\\tilde\\Theta (d^{\\max(\\mathrm{Leap}(f),2)})$. We prove a version of this conjecture for a class of functions on Gaussian isotropic data and 2-layer neural networks, under additional technical assumptions on how SGD is run. We show that the training sequentially learns the function support with a saddle-to-saddle dynamic. Our result departs from [Abbe et al. 2022] by going beyond leap 1 (merged-staircase functions), and by going beyond the mean-field and gradient flow approximations that prohibit the full complexity control obtained here. Finally, we note that this gives an SGD complexity for the full training trajectory that matches that of Correlational Statistical Query (CSQ) lower-bounds."
                },
                {
                    "title": "From high-dimensional & mean-field dynamics to dimensionless ODEs: A unifying approach to SGD in two-layers networks",
                    "abstract": "This manuscript investigates the one-pass stochastic gradient descent (SGD) dynamics of a two-layer neural network trained on Gaussian data and labels generated by a similar, though not necessarily identical, target function. We rigorously analyse the limiting dynamics via a deterministic and low-dimensional description in terms of the sufficient statistics for the population risk. Our unifying analysis bridges different regimes of interest, such as the classical gradient-flow regime of vanishing learning rate, the high-dimensional regime of large input dimension, and the overparameterised\"mean-field\"regime of large network width, covering as well the intermediate regimes where the limiting dynamics is determined by the interplay between these behaviours. In particular, in the high-dimensional limit, the infinite-width dynamics is found to remain close to a low-dimensional subspace spanned by the target principal directions. Our results therefore provide a unifying picture of the limiting SGD dynamics with synthetic data."
                },
                {
                    "title": "Learning Single-Index Models with Shallow Neural Networks",
                    "abstract": "Single-index models are a class of functions given by an unknown univariate ``link'' function applied to an unknown one-dimensional projection of the input. These models are particularly relevant in high dimension, when the data might present low-dimensional structure that learning algorithms should adapt to. While several statistical aspects of this model, such as the sample complexity of recovering the relevant (one-dimensional) subspace, are well-understood, they rely on tailored algorithms that exploit the specific structure of the target function. In this work, we introduce a natural class of shallow neural networks and study its ability to learn single-index models via gradient flow. More precisely, we consider shallow networks in which biases of the neurons are frozen at random initialization. We show that the corresponding optimization landscape is benign, which in turn leads to generalization guarantees that match the near-optimal sample complexity of dedicated semi-parametric methods."
                },
                {
                    "title": "Neural Networks can Learn Representations with Gradient Descent",
                    "abstract": "Significant theoretical work has established that in specific regimes, neural networks trained by gradient descent behave like kernel methods. However, in practice, it is known that neural networks strongly outperform their associated kernels. In this work, we explain this gap by demonstrating that there is a large class of functions which cannot be efficiently learned by kernel methods but can be easily learned with gradient descent on a two layer neural network outside the kernel regime by learning representations that are relevant to the target task. We also demonstrate that these representations allow for efficient transfer learning, which is impossible in the kernel regime. Specifically, we consider the problem of learning polynomials which depend on only a few relevant directions, i.e. of the form $f^\\star(x) = g(Ux)$ where $U: \\R^d \\to \\R^r$ with $d \\gg r$. When the degree of $f^\\star$ is $p$, it is known that $n \\asymp d^p$ samples are necessary to learn $f^\\star$ in the kernel regime. Our primary result is that gradient descent learns a representation of the data which depends only on the directions relevant to $f^\\star$. This results in an improved sample complexity of $n\\asymp d^2 r + dr^p$. Furthermore, in a transfer learning setup where the data distributions in the source and target domain share the same representation $U$ but have different polynomial heads we show that a popular heuristic for transfer learning has a target sample complexity independent of $d$."
                },
                {
                    "title": "High\u2010dimensional limit theorems for SGD: Effective dynamics and critical scaling",
                    "abstract": "We study the scaling limits of stochastic gradient descent (SGD) with constant step\u2010size in the high\u2010dimensional regime. We prove limit theorems for the trajectories of summary statistics (i.e., finite\u2010dimensional functions) of SGD as the dimension goes to infinity. Our approach allows one to choose the summary statistics that are tracked, the initialization, and the step\u2010size. It yields both ballistic (ODE) and diffusive (SDE) limits, with the limit depending dramatically on the former choices. We show a critical scaling regime for the step\u2010size, below which the effective ballistic dynamics matches gradient flow for the population loss, but at which, a new correction term appears which changes the phase diagram. About the fixed points of this effective dynamics, the corresponding diffusive limits can be quite complex and even degenerate. We demonstrate our approach on popular examples including estimation for spiked matrix and tensor models and classification via two\u2010layer networks for binary and XOR\u2010type Gaussian mixture models. These examples exhibit surprising phenomena including multimodal timescales to convergence as well as convergence to sub\u2010optimal solutions with probability bounded away from zero from random (e.g., Gaussian) initializations. At the same time, we demonstrate the benefit of overparametrization by showing that the latter probability goes to zero as the second layer width grows."
                },
                {
                    "title": "Exponential Separations in Symmetric Neural Networks",
                    "abstract": "In this work we demonstrate a novel separation between symmetric neural network architectures. Specifically, we consider the Relational Network~\\parencite{santoro2017simple} architecture as a natural generalization of the DeepSets~\\parencite{zaheer2017deep} architecture, and study their representational gap. Under the restriction to analytic activation functions, we construct a symmetric function acting on sets of size $N$ with elements in dimension $D$, which can be efficiently approximated by the former architecture, but provably requires width exponential in $N$ and $D$ for the latter."
                },
                {
                    "title": "The merged-staircase property: a necessary and nearly sufficient condition for SGD learning of sparse functions on two-layer neural networks",
                    "abstract": "It is currently known how to characterize functions that neural networks can learn with SGD for two extremal parameterizations: neural networks in the linear regime, and neural networks with no structural constraints. However, for the main parametrization of interest (non-linear but regular networks) no tight characterization has yet been achieved, despite significant developments. We take a step in this direction by considering depth-2 neural networks trained by SGD in the mean-field regime. We consider functions on binary inputs that depend on a latent low-dimensional subspace (i.e., small number of coordinates). This regime is of interest since it is poorly understood how neural networks routinely tackle high-dimensional datasets and adapt to latent low-dimensional structure without suffering from the curse of dimensionality. Accordingly, we study SGD-learnability with $O(d)$ sample complexity in a large ambient dimension $d$. Our main results characterize a hierarchical property, the\"merged-staircase property\", that is both necessary and nearly sufficient for learning in this setting. We further show that non-linear training is necessary: for this class of functions, linear methods on any feature map (e.g., the NTK) are not capable of learning efficiently. The key tools are a new\"dimension-free\"dynamics approximation result that applies to functions defined on a latent space of low-dimension, a proof of global convergence based on polynomial identity testing, and an improvement of lower bounds against linear methods for non-almost orthogonal functions."
                },
                {
                    "title": "Parent Hamiltonians of Jastrow wavefunctions",
                    "abstract": "We find the complete family of many-body quantum Hamiltonians with\nground-state of Jastrow form involving the pairwise product of a pair\nfunction in an arbitrary spatial dimension. The parent Hamiltonian\ngenerally includes a two-body pairwise potential as well as a three-body\npotential. We thus generalize the Calogero-Marchioro construction for\nthe three-dimensional case to an arbitrary spatial dimension. The\nresulting family of models is further extended to include a one-body\nterm representing an external potential, which gives rise to an\nadditional long-range two-body interaction. Using this framework, we\nprovide the generalization to an arbitrary spatial dimension of\nwell-known systems such as the Calogero-Sutherland and Calogero-Moser\nmodels. We also introduce novel models, generalizing the McGuire\nmany-body quantum bright soliton solution to higher dimensions and\nconsidering ground-states which involve e.g., polynomial, Gaussian,\nexponential, and hyperbolic pair functions. Finally, we show how the\npair function can be reverse-engineered to construct models with a given\npotential, such as a pair-wise Yukawa potential, and to identify models\ngoverned exclusively by three-body interactions."
                },
                {
                    "title": "High-Dimensional Probability: An Introduction with Applications in Data Science",
                    "abstract": "\u00a9 2018, Cambridge University Press Let us summarize our findings. A random projection of a set T in R n onto an m-dimensional subspace approximately preserves the geometry of T if m \u2a86 d ( T ) . For..."
                },
                {
                    "title": "Online stochastic gradient descent on non-convex losses from high-dimensional inference",
                    "abstract": "Stochastic gradient descent (SGD) is a popular algorithm for optimization problems arising in high-dimensional inference tasks. Here one produces an estimator of an unknown parameter from independent samples of data by iteratively optimizing a loss function. This loss function is random and often non-convex. We study the performance of the simplest version of SGD, namely online SGD, from a random start in the setting where the parameter space is high-dimensional. \nWe develop nearly sharp thresholds for the number of samples needed for consistent estimation as one varies the dimension. Our thresholds depend only on an intrinsic property of the population loss which we call the information exponent. In particular, our results do not assume uniform control on the loss itself, such as convexity or uniform derivative bounds. The thresholds we obtain are polynomial in the dimension and the precise exponent depends explicitly on the information exponent. As a consequence of our results, we find that except for the simplest tasks, almost all of the data is used simply in the initial search phase to obtain non-trivial correlation with the ground truth. Upon attaining non-trivial correlation, the descent is rapid and exhibits law of large numbers type behaviour. \nWe illustrate our approach by applying it to a wide set of inference tasks such as phase retrieval, parameter estimation for generalized linear models, spiked matrix models, and spiked tensor models, as well as for supervised learning for single-layer networks with general activation functions."
                },
                {
                    "title": "Learning a Single Neuron with Gradient Methods",
                    "abstract": "We consider the fundamental problem of learning a single neuron $x \\mapsto\\sigma(w^\\top x)$ using standard gradient methods. As opposed to previous works, which considered specific (and not always realistic) input distributions and activation functions $\\sigma(\\cdot)$, we ask whether a more general result is attainable, under milder assumptions. On the one hand, we show that some assumptions on the distribution and the activation function are necessary. On the other hand, we prove positive guarantees under mild assumptions, which go beyond those studied in the literature so far. We also point out and study the challenges in further strengthening and generalizing our results."
                },
                {
                    "title": "Learning Single-Index Models in Gaussian Space",
                    "abstract": "We consider regression problems where the response is a smooth but non-linear function of a kdimensional projection of p normally-distributed covariates, contaminated with additive Gaussian noise. The goal is to recover the range of the k-dimensional projection, i.e., the index space. This model is called the multi-index model, and the k = 1 case is called the single-index model. For the single-index model, we characterize the population landscape of a natural semi-parametric maximum likelihood objective in terms of the link function and prove that it has no spurious local minima. We also propose and analyze an efficient iterative procedure that recovers the index space up to error using a sample size \u00d5(p 2/\u03bc)+p/ ), whereR and \u03bc, respectively, parameterize the smoothness of the link function and the signal strength. When a multi-index model is incorrectly specified as a single-index model, we prove that essentially the same procedure, with sample size \u00d5(p /\u03bc) + p/ ), returns a vector that is -close to being completely in the index space."
                },
                {
                    "title": "A simple neural network module for relational reasoning",
                    "abstract": "Relational reasoning is a central component of generally intelligent behavior, but has proven difficult for neural networks to learn. In this paper we describe how to use Relation Networks (RNs) as a simple plug-and-play module to solve problems that fundamentally hinge on relational reasoning. We tested RN-augmented networks on three tasks: visual question answering using a challenging dataset called CLEVR, on which we achieve state-of-the-art, super-human performance; text-based question answering using the bAbI suite of tasks; and complex reasoning about dynamic physical systems. Then, using a curated dataset called Sort-of-CLEVR we show that powerful convolutional networks do not have a general capacity to solve relational questions, but can gain this capacity when augmented with RNs. Our work shows how a deep learning architecture equipped with an RN module can implicitly discover and learn to reason about entities and their relations."
                },
                {
                    "title": "Deep Sets",
                    "abstract": "In this paper, we study the problem of designing objective functions for machine learning problems defined on finite \\emph{sets}. In contrast to traditional objective functions defined for machine learning problems operating on finite dimensional vectors, the new objective functions we propose are operating on finite sets and are invariant to permutations. Such problems are widespread, ranging from estimation of population statistics \\citep{poczos13aistats}, via anomaly detection in piezometer data of embankment dams \\citep{Jung15Exploration}, to cosmology \\citep{Ntampaka16Dynamical,Ravanbakhsh16ICML1}. Our main theorem characterizes the permutation invariant objective functions and provides a family of functions to which any permutation invariant objective function must belong. This family of functions has a special structure which enables us to design a deep network architecture that can operate on sets and which can be deployed on a variety of scenarios including both unsupervised and supervised learning tasks. We demonstrate the applicability of our method on population statistic estimation, point cloud classification, set expansion, and image tagging."
                },
                {
                    "title": "PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation",
                    "abstract": "Point cloud is an important type of geometric data structure. Due to its irregular format, most researchers transform such data to regular 3D voxel grids or collections of images. This, however, renders data unnecessarily voluminous and causes issues. In this paper, we design a novel type of neural network that directly consumes point clouds, which well respects the permutation invariance of points in the input. Our network, named PointNet, provides a unified architecture for applications ranging from object classification, part segmentation, to scene semantic parsing. Though simple, PointNet is highly efficient and effective. Empirically, it shows strong performance on par or even better than state of the art. Theoretically, we provide analysis towards understanding of what the network has learnt and why the network is robust with respect to input perturbation and corruption."
                },
                {
                    "title": "Analysis of Boolean Functions",
                    "abstract": "Scribe notes from the 2012 Barbados Workshop on Computational Complexity. A series of lectures on Analysis of Boolean Functions by Ryan O'Donnell, with a guest lecture by Per Austrin."
                },
                {
                    "title": "Efficient Learning of Generalized Linear and Single Index Models with Isotonic Regression",
                    "abstract": "Generalized Linear Models (GLMs) and Single Index Models (SIMs) provide powerful generalizations of linear regression, where the target variable is assumed to be a (possibly unknown) 1-dimensional function of a linear predictor. In general, these problems entail non-convex estimation procedures, and, in practice, iterative local search heuristics are often used. Kalai and Sastry (2009) provided the first provably efficient method, the Isotron algorithm, for learning SIMs and GLMs, under the assumption that the data is in fact generated under a GLM and under certain monotonicity and Lipschitz (bounded slope) constraints. The Isotron algorithm interleaves steps of perceptron-like updates with isotonic regression (fitting a one-dimensional non-decreasing function). However, to obtain provable performance, the method requires a fresh sample every iteration. In this paper, we provide algorithms for learning GLMs and SIMs, which are both computationally and statistically efficient. We modify the isotonic regression step in Isotron to fit a Lipschitz monotonic function, and also provide an efficient O(n log(n)) algorithm for this step, improving upon the previous O(n2) algorithm. We provide a brief empirical study, demonstrating the feasibility of our algorithms in practice."
                },
                {
                    "title": "Best Arm Identification in Multi-Armed Bandits",
                    "abstract": "We consider the problem of finding the best arm in a stochastic multi-armed bandit game. The regret of a forecaster is here defined by the gap between the mean reward of the optimal arm and the mean reward of the ultimately chosen arm. We propose a highly exploring UCB policy and a new algorithm based on successive rejects. We show that these algorithms are essentially optimal since their regret decreases exponentially at a rate which is, up to a logarithmic factor, the best possible. However, while the UCB policy needs the tuning of a parameter depending on the unobservable hardness of the task, the successive rejects policy benefits from being parameter-free, and also independent of the scaling of the rewards. As a by-product of our analysis, we show that identifying the best arm (when it is unique) requires a number of samples of order (up to a log(K) factor) \u03a3 i 1/\u03942i, where the sum is on the suboptimal arms and\u0394i represents the difference between the mean reward of the best arm and the one of arm i. This generalizes the well-known fact that one needs of order of 1/\u03942 samples to differentiate the means of two distributions with gap \u0394."
                },
                {
                    "title": "The Isotron Algorithm: High-Dimensional Isotonic Regression",
                    "abstract": "The Perceptron algorithm elegantly solves binary classification problems that have a margin between positive and negative examples. Isotonic regression (fitting an arbitrary increasing function in one dimension) is also a natural problem with a simple solution. By combining the two, we get a new but very simple algorithm with strong guarantees. Our ISOTRON algorithm provably learns Single Index Models (SIM), a generalization of linear and logistic regression, generalized linear models, as well as binary classification by linear threshold functions. In particular, it provably learns SIMs with unknown mean functions that are nondecreasing and Lipschitz-continuous, thereby generalizing linear and logistic regression and linear-threshold functions (with a margin). Like the Perceptron, it is straightforward to implement and kernelize. Hence, the ISOTRON provides a very simple yet flexible and principled approach to regression."
                },
                {
                    "title": "A method to derive explicit formulas for an elliptic generalization of the Jack polynomials",
                    "abstract": "We review a method providing explicit formulas for the Jack polynomials. Our method is based on the relation of the Jack polynomials to the eigenfunctions of a well-known exactly solvable quantum many-body system of Calogero-Sutherland type. We also sketch a generalization of our method allowing to find the exact solution of the elliptic generalization of the Calogero-Sutherland model. We present the resulting explicit formulas for certain symmetric functions generalizing the Jack polynomials to the elliptic case."
                },
                {
                    "title": "Precoding and Signal Shaping for Digital Transmission",
                    "abstract": "Preface. Introduction. Digital Communications via Linear, Distorting Channels. Precoding Schemes. Signal Shaping. Combined Precoding and Signal Shaping. Appendix A: Wirtinger Calculus. Appendix B: Parameters of the Numerical Examples. Appendix C: Introduction to Lattices. Appendix D: Calculation of Shell Frequency Distribution. Appendix E: Precoding for MIMO Channels. Appendix F: List of Symbols, Variables, and Acronyms. Index."
                },
                {
                    "title": "Laplace and the origin of the Ornstein-Uhlenbeck process",
                    "abstract": "It is shown that a second-order partial differential equation, found by Laplace in 1810, is the FokkerPlanck equation for a one-dimensional Ornstein-Uhlenbeck process. It is argued that Laplace's reasoning, although not rigorous, can be entirely justified by using the modern theory of weak convergence of stochastic processes. The solutions to the differential equation found by Laplace and others, using expansions in terms of Hermite polynomials, are discussed."
                },
                {
                    "title": "On the eigenvalues of random matrices",
                    "abstract": "Let M be a random matrix chosen from Haar measure on the unitary group Un. Let Z = X + iY be a standard complex normal random variable with X and Y independent, mean 0 and variance \u00bd normal variables. We show that for j = 1, 2, \u2026, Tr(Mj ) are independent and distributed as \u221ajZ asymptotically as n \u2192\u221e. This result is used to study the set of eigenvalues of M. Similar results are given for the orthogonal and symplectic and symmetric groups."
                },
                {
                    "title": "Symmetric functions and Hall polynomials",
                    "abstract": "I. Symmetric functions II. Hall polynomials III. HallLittlewood symmetric functions IV. The characters of GLn over a finite field V. The Hecke ring of GLn over a finite field VI. Symmetric functions with two parameters VII. Zonal polynomials"
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively characterize the learning dynamics of DeepSet architectures under gradient descent for symmetric feature learning in high-dimensional spaces?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing our understanding of how neural networks can leverage symmetry in data, which has significant implications for various fields such as computer vision, natural language processing, and bioinformatics. By addressing this question, we can enhance the theoretical foundations of deep learning, leading to the development of more efficient and robust neural architectures that can exploit predefined symmetries. This could pave the way for practical applications where symmetry plays a critical role, such as in set-based data representations and permutation-invariant tasks.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the nonconvex nature of the optimization landscape associated with neural networks, particularly in the context of DeepSet architectures. Naive approaches may fail due to the complexity of capturing permutation invariance and the intricacies of gradient descent dynamics in high-dimensional spaces. Additionally, the need to analyze the interplay between model initialization, data distribution, and the specific architecture introduces significant theoretical and practical obstacles that complicate the learning process.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on fully-connected shallow neural networks, which do not adequately address the need for specialized architectures that can exploit symmetries in target functions. The limitations of these prior works include a lack of consideration for permutation invariance and the complexities involved in extending existing analyses to more sophisticated architectures like DeepSets. Our approach differs by explicitly incorporating symmetric polynomial theory and focusing on the unique dynamics of gradient descent in the context of DeepSet architectures, which has not been thoroughly explored in earlier studies.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves analyzing the learning dynamics of DeepSet architectures under gradient descent by introducing a symmetric single index target. We will utilize a dataset that reflects permutation-invariant properties and measure performance using metrics that capture the efficiency of learning, such as convergence rates and sample complexity. The expected outcomes include a comprehensive characterization of the learning dynamics, demonstrating efficient learning under the specified conditions, and providing insights into the role of symmetry in feature learning within neural networks."
            }
        },
        "author_data": {
            "7d5935cd-b3ed-48b9-89ce-6b7770d7ea90": {
                "pk": "7d5935cd-b3ed-48b9-89ce-6b7770d7ea90",
                "name": "Aaron Zweig",
                "collaborators": [
                    "Joan Bruna",
                    "Aditya Grover",
                    "Stefano Ermon",
                    "Loucas Pillaud-Vivien",
                    "Nesreen Ahmed",
                    "Ted Willke",
                    "Guixiang Ma",
                    "Eric Wang"
                ],
                "domain": [
                    "Neural Networks",
                    "Graph Learning",
                    "Reinforcement Learning",
                    "Domain Adaptation"
                ],
                "publications": [
                    {
                        "title": "On Single Index Models beyond Gaussian Data",
                        "abstract": "Sparse high-dimensional functions have arisen as a rich framework to study the behavior of gradient-descent methods using shallow neural networks, showcasing their ability to perform feature learning beyond linear models. Amongst those functions, the simplest are single-index models $f(x) = \\phi( x \\cdot \\theta^*)$, where the labels are generated by an arbitrary non-linear scalar link function $\\phi$ applied to an unknown one-dimensional projection $\\theta^*$ of the input data. By focusing on Gaussian data, several recent works have built a remarkable picture, where the so-called information exponent (related to the regularity of the link function) controls the required sample complexity. In essence, these tools exploit the stability and spherical symmetry of Gaussian distributions. In this work, building from the framework of \\cite{arous2020online}, we explore extensions of this picture beyond the Gaussian setting, where both stability or symmetry might be violated. Focusing on the planted setting where $\\phi$ is known, our main results establish that Stochastic Gradient Descent can efficiently recover the unknown direction $\\theta^*$ in the high-dimensional regime, under assumptions that extend previous works \\cite{yehudai2020learning,wu2022learning}."
                    },
                    {
                        "title": "Exponential Separations in Symmetric Neural Networks",
                        "abstract": "In this work we demonstrate a novel separation between symmetric neural network architectures. Specifically, we consider the Relational Network~\\parencite{santoro2017simple} architecture as a natural generalization of the DeepSets~\\parencite{zaheer2017deep} architecture, and study their representational gap. Under the restriction to analytic activation functions, we construct a symmetric function acting on sets of size $N$ with elements in dimension $D$, which can be efficiently approximated by the former architecture, but provably requires width exponential in $N$ and $D$ for the latter."
                    },
                    {
                        "title": "Towards Antisymmetric Neural Ansatz Separation",
                        "abstract": "We study separations between two fundamental models (or \\emph{Ans\\\"atze}) of antisymmetric functions, that is, functions $f$ of the form $f(x_{\\sigma(1)}, \\ldots, x_{\\sigma(N)}) = \\text{sign}(\\sigma)f(x_1, \\ldots, x_N)$, where $\\sigma$ is any permutation. These arise in the context of quantum chemistry, and are the basic modeling tool for wavefunctions of Fermionic systems. Specifically, we consider two popular antisymmetric Ans\\\"atze: the Slater representation, which leverages the alternating structure of determinants, and the Jastrow ansatz, which augments Slater determinants with a product by an arbitrary symmetric function. We construct an antisymmetric function in $N$ dimensions that can be efficiently expressed in Jastrow form, yet provably cannot be approximated by Slater determinants unless there are exponentially (in $N^2$) many terms. This represents the first explicit quantitative separation between these two Ans\\\"atze."
                    },
                    {
                        "title": "A Functional Perspective on Learning Symmetric Functions with Neural Networks",
                        "abstract": "Symmetric functions, which take as input an unordered, fixed-size set, are known to be universally representable by neural networks that enforce permutation invariance. However, these architectures only give guarantees for fixed input sizes, yet in many practical scenarios, such as particle physics, a relevant notion of generalization should include varying the input size.  In this paper, we embed symmetric functions (of any size) as functions over probability measures, and study the ability of neural networks defined over this space of measures to represent and learn in that space. By focusing on shallow architectures, we establish approximation and generalization bounds under different choices of regularization (such as RKHS and variation norms), that capture a hierarchy of functional spaces with increasing amount of non-linear learning. The resulting models can be learnt efficiently and enjoy generalization guarantees that extend across input sizes, as we verify empirically."
                    },
                    {
                        "title": "Provably Efficient Third-Person Imitation from Offline Observation",
                        "abstract": "Domain adaptation in imitation learning represents an essential step towards improving generalizability. However, even in the restricted setting of third-person imitation where transfer is between isomorphic Markov Decision Processes, there are no strong guarantees on the performance of transferred policies. We present problem-dependent, statistical learning guarantees for third-person imitation from observation in an offline setting, and a lower bound on performance in the online setting."
                    },
                    {
                        "title": "Neural Algorithms for Graph Navigation",
                        "abstract": "The application of deep reinforcement learning (RL) to graph learning and meta-learning admits challenges from both topics. We consider the task of one-shot, partially observed graph navigation, acknowledging and addressing the dif\ufb01culties of partially observed graph environments. In this work, we present a framework for graph meta-learning"
                    },
                    {
                        "title": "Stochastic Optimization of Sorting Networks via Continuous Relaxations",
                        "abstract": "Sorting input objects is an important step in many machine learning pipelines. However, the sorting operator is non-differentiable with respect to its inputs, which prohibits end-to-end gradient-based optimization. In this work, we propose NeuralSort, a general-purpose continuous relaxation of the output of the sorting operator from permutation matrices to the set of unimodal row-stochastic matrices, where every row sums to one and has a distinct arg max. This relaxation permits straight-through optimization of any computational graph involve a sorting operation. Further, we use this relaxation to enable gradient-based stochastic optimization over the combinatorially large space of permutations by deriving a reparameterized gradient estimator for the Plackett-Luce family of distributions over permutations. We demonstrate the usefulness of our framework on three tasks that require learning semantic orderings of high-dimensional objects, including a fully differentiable, parameterized extension of the k-nearest neighbors algorithm."
                    },
                    {
                        "title": "Graphite: Iterative Generative Modeling of Graphs",
                        "abstract": "Graphs are a fundamental abstraction for modeling relational data. However, graphs are discrete and combinatorial in nature, and learning representations suitable for machine learning tasks poses statistical and computational challenges. In this work, we propose Graphite, an algorithmic framework for unsupervised learning of representations over nodes in large graphs using deep latent variable generative models. Our model parameterizes variational autoencoders (VAE) with graph neural networks, and uses a novel iterative graph refinement strategy inspired by low-rank approximations for decoding. On a wide variety of synthetic and benchmark datasets, Graphite outperforms competing approaches for the tasks of density estimation, link prediction, and node classification. Finally, we derive a theoretical connection between message passing in graph neural networks and mean-field variational inference."
                    }
                ]
            },
            "81537ad6-c14b-4024-9764-e43324906062": {
                "pk": "81537ad6-c14b-4024-9764-e43324906062",
                "name": "Joan Bruna",
                "collaborators": [
                    "A. Bietti",
                    "Loucas Pillaud-Vivien",
                    "Lei Chen",
                    "Alex Damian",
                    "Jason D. Lee"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Statistical Learning",
                    "High-Dimensional Regression",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "How Truncating Weights Improves Reasoning in Language Models",
                        "abstract": "In addition to the ability to generate fluent text in various languages, large language models have been successful at tasks that involve basic forms of logical\"reasoning\"over their context. Recent work found that selectively removing certain components from weight matrices in pre-trained models can improve such reasoning capabilities. We investigate this phenomenon further by carefully studying how certain global associations tend to be stored in specific weight components or Transformer blocks, in particular feed-forward layers. Such associations may hurt predictions in reasoning tasks, and removing the corresponding components may then improve performance. We analyze how this arises during training, both empirically and theoretically, on a two-layer Transformer trained on a basic reasoning task with noise, a toy associative memory model, and on the Pythia family of pre-trained models tested on simple reasoning tasks."
                    },
                    {
                        "title": "Computational-Statistical Gaps in Gaussian Single-Index Models",
                        "abstract": "Single-Index Models are high-dimensional regression problems with planted structure, whereby labels depend on an unknown one-dimensional projection of the input via a generic, non-linear, and potentially non-deterministic transformation. As such, they encompass a broad class of statistical inference tasks, and provide a rich template to study statistical and computational trade-offs in the high-dimensional regime. While the information-theoretic sample complexity to recover the hidden direction is linear in the dimension $d$, we show that computationally efficient algorithms, both within the Statistical Query (SQ) and the Low-Degree Polynomial (LDP) framework, necessarily require $\\Omega(d^{k^\\star/2})$ samples, where $k^\\star$ is a\"generative\"exponent associated with the model that we explicitly characterize. Moreover, we show that this sample complexity is also sufficient, by establishing matching upper bounds using a partial-trace algorithm. Therefore, our results provide evidence of a sharp computational-to-statistical gap (under both the SQ and LDP class) whenever $k^\\star>2$. To complete the study, we provide examples of smooth and Lipschitz deterministic target functions with arbitrarily large generative exponents $k^\\star$."
                    },
                    {
                        "title": "On Learning Gaussian Multi-index Models with Gradient Flow",
                        "abstract": "We study gradient flow on the multi-index regression problem for high-dimensional Gaussian data. Multi-index functions consist of a composition of an unknown low-rank linear projection and an arbitrary unknown, low-dimensional link function. As such, they constitute a natural template for feature learning in neural networks. We consider a two-timescale algorithm, whereby the low-dimensional link function is learnt with a non-parametric model infinitely faster than the subspace parametrizing the low-rank projection. By appropriately exploiting the matrix semigroup structure arising over the subspace correlation matrices, we establish global convergence of the resulting Grassmannian population gradient flow dynamics, and provide a quantitative description of its associated `saddle-to-saddle' dynamics. Notably, the timescales associated with each saddle can be explicitly characterized in terms of an appropriate Hermite decomposition of the target link function. In contrast with these positive results, we also show that the related \\emph{planted} problem, where the link function is known and fixed, in fact has a rough optimization landscape, in which gradient flow dynamics might get trapped with high probability."
                    }
                ]
            }
        }
    },
    "2410.15010": {
        "paper_data": {
            "title": "FlexMol: A Flexible Toolkit for Benchmarking Molecular Relational Learning",
            "url": "http://arxiv.org/abs/2410.15010v1",
            "arxiv_id": "2410.15010",
            "authors": [
                "Sizhe Liu",
                "Jun Xia",
                "Lecheng Zhang",
                "Yuchen Liu",
                "Yue Liu",
                "Wenjie Du",
                "Zhangyang Gao",
                "Bozhen Hu",
                "Cheng Tan",
                "Hongxin Xiang",
                "Stan Z. Li"
            ],
            "abstract": "Molecular relational learning (MRL) is crucial for understanding the interaction behaviors between molecular pairs, a critical aspect of drug discovery and development. However, the large feasible model space of MRL poses significant challenges to benchmarking, and existing MRL frameworks face limitations in flexibility and scope. To address these challenges, avoid repetitive coding efforts, and ensure fair comparison of models, we introduce FlexMol, a comprehensive toolkit designed to facilitate the construction and evaluation of diverse model architectures across various datasets and performance metrics. FlexMol offers a robust suite of preset model components, including 16 drug encoders, 13 protein sequence encoders, 9 protein structure encoders, and 7 interaction layers. With its easy-to-use API and flexibility, FlexMol supports the dynamic construction of over 70, 000 distinct combinations of model architectures. Additionally, we provide detailed benchmark results and code examples to demonstrate FlexMol's effectiveness in simplifying and standardizing MRL model development and comparison.",
            "introduction": "   1 Introduction  Molecular relational learning (MRL) aims to understand the interaction behavior between molecular pairs[23]. Among all interaction types, those involving drugs and proteins are of particular interest due to their significant impact on therapeutic discovery and development. Drug-target interactions (DTIs) play a crucial role in various aspects of drug development, such as virtual screening, drug repurposing, and predicting potential side effects [56]. Protein-protein interactions (PPIs) reveal new potential therapeutic targets by enhancing our understanding of protein structural characteristics and cellular molecular machinery [42, 27]. Drug-drug interactions (DDIs) are vital for understanding the effects of concurrent drug use, which can inform strategies to prevent adverse drug reactions and ensure patient safety [5, 12].   MRL has significantly advanced through the integration of deep learning models and substantial AI-ready datasets [14, 54]. MRL models typically consist of several key components: modules for encoding molecule 1, modules for encoding molecule 2, modules for modeling interactions between molecules, and additional auxiliary modules [38, 22, 60, 30, 16, 46, 59, 55]. The variety of encoding methods and interaction layers available allows for the easy construction of new models by recombining these components. This flexibility, however, results in a substantial model space, posing significant challenges for benchmarking processes.   To address these challenges and facilitate benchmarking in MRL, DeepPurpose\u00a0[15] emerged as the the first and, to our knowledge, the only MRL library that enables dynamically built models. It offers a user-friendly API, allowing users to create DTI models with customizable drug and protein encoders. However, DeepPurpose inherits several limitations from early MRL models.     \u2022  Insufficient Input Support: Early models typically relied on amino acid sequences as protein input data due to the limited availability of protein structure data [22, 60, 30]. Consequently, DeepPurpose supports only protein sequences as input types for proteins. However, advances in structural biology have enabled more models to use protein structures\u2014either experimentally verified or AlphaFold-generated\u2014as inputs to enhance performance [46, 11, 58, 55]. This shift underscores the need for libraries that can handle both sequence and structure data for proteins.    \u2022  Lack of Flexible Architectures: Early machine learning models for MRL often had rigid and simplistic architectures. For instance, DeepDDI uses a Structural Similarity Profile (SSP) to encode each drug [38], while DeepDTI employs convolutional neural networks (CNNs) to encode both drug and protein sequences [60]. These encoded vectors are then concatenated and fed into a multi-layer perceptron (MLP) predictor. Similarly, while DeepPurpose can dynamically construct models, it remains confined to a dual-encoder plus MLP architecture. Recent advances in MRL have introduced more complex and effective model architectures that surpass the capabilities of DeepPurpose. Many models often employ multiple encoding methods for comprehensive molecular representation. Examples include DataDTA [59], which uses pocket structure information along with protein sequences to encode drug targets, and 3DProtDTA [46], which combines Morgan fingerprints and graph neural networks for drug encoding. Finally, interaction layers have become critical components of modern models. For instance, DrugBan [1] uses a bilinear interaction layer, and MCL-DTI [34] implements bi-directional cross-attention to model molecular relations.      In this paper, we introduce FlexMol, a comprehensive toolkit designed to overcome the limitations of existing MRL",
            "references": [
                {
                    "title": "DataDTA: a multi-feature and dual-interaction aggregation framework for drug\u2013target binding affinity prediction",
                    "abstract": "Abstract Motivation Accurate prediction of drug\u2013target binding affinity (DTA) is crucial for drug discovery. The increase in the publication of large-scale DTA datasets enables the development of various computational methods for DTA prediction. Numerous deep learning-based methods have been proposed to predict affinities, some of which only utilize original sequence information or complex structures, but the effective combination of various information and protein-binding pockets have not been fully mined. Therefore, a new method that integrates available key information is urgently needed to predict DTA and accelerate the drug discovery process. Results In this study, we propose a novel deep learning-based predictor termed DataDTA to estimate the affinities of drug\u2013target pairs. DataDTA utilizes descriptors of predicted pockets and sequences of proteins, as well as low-dimensional molecular features and SMILES strings of compounds as inputs. Specifically, the pockets were predicted from the three-dimensional structure of proteins and their descriptors were extracted as the partial input features for DTA prediction. The molecular representation of compounds based on algebraic graph features was collected to supplement the input information of targets. Furthermore, to ensure effective learning of multiscale interaction features, a dual-interaction aggregation neural network strategy was developed. DataDTA was compared with state-of-the-art methods on different datasets, and the results showed that DataDTA is a reliable prediction tool for affinities estimation. Specifically, the concordance index (CI) of DataDTA is 0.806 and the Pearson correlation coefficient (R) value is 0.814 on the test dataset, which is higher than other methods. Availability and implementation The codes and datasets of DataDTA are available at https://github.com/YanZhu06/DataDTA."
                },
                {
                    "title": "From Proteins to Ligands: Decoding Deep Learning Methods for Binding Affinity Prediction",
                    "abstract": "Accurate in silico prediction of protein-ligand binding affinity is important in the early stages of drug discovery. Deep learning-based methods exist but have yet to overtake more conventional methods such as giga-docking largely due to their lack of generalisability. To improve generalizability we need to understand what these models learn from input protein and ligand data. We systematically investigated a sequence-based deep learning framework to assess the impact of protein and ligand encodings on predicting binding affinities for commonly used kinase data sets. The role of proteins is studied using convolutional neural network-based encodings obtained from sequences and graph neural network-based encodings enriched with structural information from contact maps. Ligand-based encodings are generated from graph-neural networks. We test different ligand perturbations by randomizing node and edge properties. For proteins we make use of 3 different protein contact generation methods (AlphaFold2, Pconsc4, and ESM-1b) and compare these with a random control. Our investigation shows that protein encodings do not substantially impact the binding predictions, with no statistically significant difference in binding affinity for KIBA in the investigated metrics (concordance index, Pearson\u2019s R Spearman\u2019s Rank, and RMSE). Significant differences are seen for ligand encodings with random ligands and random ligand node properties, suggesting a much bigger reliance on ligand data for the learning tasks. Using different ways to combine protein and ligand encodings, did not show a significant change in performance. TOC Graphic"
                },
                {
                    "title": "Shift-Robust Molecular Relational Learning with Causal Substructure",
                    "abstract": "Recently, molecular relational learning, whose goal is to predict the interaction behavior between molecular pairs, got a surge of interest in molecular sciences due to its wide range of applications. In this work, we propose CMRL that is robust to the distributional shift in molecular relational learning by detecting the core substructure that is causally related to chemical reactions. To do so, we first assume a causal relationship based on the domain knowledge of molecular sciences and construct a structural causal model (SCM) that reveals the relationship between variables. Based on the SCM, we introduce a novel conditional intervention framework whose intervention is conditioned on the paired molecule. With the conditional intervention framework, our model successfully learns from the causal substructure and alleviates the confounding effect of shortcut substructures that are spuriously correlated to chemical reactions. Extensive experiments on various tasks with real-world and synthetic datasets demonstrate the superiority of CMRL over state-of-the-art baseline models."
                },
                {
                    "title": "3DProtDTA: a deep learning model for drug-target affinity prediction based on residue-level protein graphs",
                    "abstract": "Accurate prediction of the drug-target affinity (DTA) in silico is of critical importance for modern drug discovery. Computational methods of DTA prediction, applied in the early stages of drug development, are able to speed it up and cut its cost significantly. A wide range of approaches based on machine learning were recently proposed for DTA assessment. The most promising of them are based on deep learning techniques and graph neural networks to encode molecular structures. The recent breakthrough in protein structure prediction made by AlphaFold made an unprecedented amount of proteins without experimentally defined structures accessible for computational DTA prediction. In this work, we propose a new deep learning DTA model 3DProtDTA, which utilises AlphaFold structure predictions in conjunction with the graph representation of proteins. The model is superior to its rivals on common benchmarking datasets and has potential for further improvement."
                },
                {
                    "title": "Machine learning on protein-protein interaction prediction: models, challenges and trends",
                    "abstract": "Protein-protein interactions (PPIs) carry out the cellular processes of all living organisms. Experimental methods for PPI detection suffer from high cost and false-positive rate, hence efficient computational methods are highly desirable for facilitating PPI detection. In recent years, benefiting from the enormous amount of protein data produced by advanced high-throughput technologies, machine learning models have been well developed in the field of PPI prediction. In this paper, we present a comprehensive survey of the recently proposed machine learning-based prediction methods. The machine learning models applied in these methods and details of protein data representation are also outlined. To understand the potential improvements in PPI prediction, we discuss the trend in the development of machine learning-based methods. Finally, we highlight potential directions in PPI prediction, such as the use of computationally predicted protein structures to extend the data source for machine learning models. This review is supposed to serve as a companion for further improvements in this field."
                },
                {
                    "title": "MARPPI: boosting prediction of protein-protein interactions with multi-scale architecture residual network",
                    "abstract": "Protein-protein interactions (PPIs) are a major component of the cellular biochemical reaction network. Rich sequence information and machine learning techniques reduce the dependence of exploring PPIs on wet experiments, which are costly and time-consuming. This paper proposes a PPI prediction model, multi-scale architecture residual network for PPIs (MARPPI), based on dual-channel and multi-feature. Multi-feature leverages Res2vec to obtain the association information between residues, and utilizes pseudo amino acid composition, autocorrelation descriptors and multivariate mutual information to achieve the amino acid composition and order information, physicochemical properties and information entropy, respectively. Dual channel utilizes multi-scale architecture improved ResNet network which extracts protein sequence features to reduce protein feature loss. Compared with other advanced methods, MARPPI achieves 96.03%, 99.01% and 91.80% accuracy in the intraspecific datasets of Saccharomyces cerevisiae, Human and Helicobacter pylori, respectively. The accuracy on the two interspecific datasets of Human-Bacillus anthracis and Human-Yersinia pestis is 97.29%, and 95.30%, respectively. In addition, results on specific datasets of disease (neurodegenerative and metabolic disorders) demonstrate the ability to detect hidden interactions. To better illustrate the performance of MARPPI, evaluations on independent datasets and PPIs network suggest that MARPPI can be used to predict cross-species interactions. The above shows that MARPPI can be regarded as a concise, efficient and accurate tool for PPI datasets."
                },
                {
                    "title": "A Systematic Survey of Chemical Pre-trained Models",
                    "abstract": "Deep learning has achieved remarkable success in learning representations for molecules, which is crucial for various biochemical applications, ranging from property prediction to drug design. However, training Deep Neural Networks (DNNs) from scratch often requires abundant labeled molecules, which are expensive to acquire in the real world. To alleviate this issue, tremendous efforts have been devoted to Chemical Pre-trained Models (CPMs), where DNNs are pre-trained using large-scale unlabeled molecular databases and then fine-tuned over specific downstream tasks. Despite the prosperity, there lacks a systematic review of this fast-growing field. In this paper, we present the first survey that summarizes the current progress of CPMs. We first highlight the limitations of training molecular representation models from scratch to motivate CPM studies. Next, we systematically review recent advances on this topic from several key perspectives, including molecular descriptors, encoder architectures, pre-training strategies, and applications. We also highlight the challenges and promising avenues for future research, providing a useful resource for both machine learning and scientific communities."
                },
                {
                    "title": "Tartarus: A Benchmarking Platform for Realistic And Practical Inverse Molecular Design",
                    "abstract": "The efficient exploration of chemical space to design molecules with intended properties enables the accelerated discovery of drugs, materials, and catalysts, and is one of the most important outstanding challenges in chemistry. Encouraged by the recent surge in computer power and artificial intelligence development, many algorithms have been developed to tackle this problem. However, despite the emergence of many new approaches in recent years, comparatively little progress has been made in developing realistic benchmarks that reflect the complexity of molecular design for real-world applications. In this work, we develop a set of practical benchmark tasks relying on physical simulation of molecular systems mimicking real-life molecular design problems for materials, drugs, and chemical reactions. Additionally, we demonstrate the utility and ease of use of our new benchmark set by demonstrating how to compare the performance of several well-established families of algorithms. Surprisingly, we find that model performance can strongly depend on the benchmark domain. We believe that our benchmark suite will help move the field towards more realistic molecular design benchmarks, and move the development of inverse molecular design algorithms closer to designing molecules that solve existing problems in both academia and industry alike."
                },
                {
                    "title": "AttentionSiteDTI: an interpretable graph-based model for drug-target interaction prediction using NLP sentence-level relation classification",
                    "abstract": "Abstract In this study, we introduce an interpretable graph-based deep learning prediction model, AttentionSiteDTI, which utilizes protein binding sites along with a self-attention mechanism to address the problem of drug\u2013target interaction prediction. Our proposed model is inspired by sentence classification models in the field of Natural Language Processing, where the drug\u2013target complex is treated as a sentence with relational meaning between its biochemical entities a.k.a. protein pockets and drug molecule. AttentionSiteDTI enables interpretability by identifying the protein binding sites that contribute the most toward the drug\u2013target interaction. Results on three benchmark datasets show improved performance compared with the current state-of-the-art models. More significantly, unlike previous studies, our model shows superior performance, when tested on new proteins (i.e. high generalizability). Through multidisciplinary collaboration, we further experimentally evaluate the practical potential of our proposed approach. To achieve this, we first computationally predict the binding interactions between some candidate compounds and a target protein, then experimentally validate the binding interactions for these pairs in the laboratory. The high agreement between the computationally predicted and experimentally observed (measured) drug\u2013target interactions illustrates the potential of our method as an effective pre-screening tool in drug repurposing applications."
                },
                {
                    "title": "Protein Representation Learning by Geometric Structure Pretraining",
                    "abstract": "Learning effective protein representations is critical in a variety of tasks in biology such as predicting protein function or structure. Existing approaches usually pretrain protein language models on a large number of unlabeled amino acid sequences and then finetune the models with some labeled data in downstream tasks. Despite the effectiveness of sequence-based approaches, the power of pretraining on known protein structures, which are available in smaller numbers only, has not been explored for protein property prediction, though protein structures are known to be determinants of protein function. In this paper, we propose to pretrain protein representations according to their 3D structures. We first present a simple yet effective encoder to learn the geometric features of a protein. We pretrain the protein graph encoder by leveraging multiview contrastive learning and different self-prediction tasks. Experimental results on both function prediction and fold classification tasks show that our proposed pretraining methods outperform or are on par with the state-of-the-art sequence-based methods, while using much less pretraining data. Our implementation is available at https://github.com/DeepGraphLearning/GearNet."
                },
                {
                    "title": "A Review of Approaches for Predicting Drug\u2013Drug Interactions Based on Machine Learning",
                    "abstract": "Drug\u2013drug interactions play a vital role in drug research. However, they may also cause adverse reactions in patients, with serious consequences. Manual detection of drug\u2013drug interactions is time-consuming and expensive, so it is urgent to use computer methods to solve the problem. There are two ways for computers to identify drug interactions: one is to identify known drug interactions, and the other is to predict unknown drug interactions. In this paper, we review the research progress of machine learning in predicting unknown drug interactions. Among these methods, the literature-based method is special because it combines the extraction method of DDI and the prediction method of DDI. We first introduce the common databases, then briefly describe each method, and summarize the advantages and disadvantages of some prediction models. Finally, we discuss the challenges and prospects of machine learning methods in predicting drug interactions. This review aims to provide useful guidance for interested researchers to further promote bioinformatics algorithms to predict DDI."
                },
                {
                    "title": "Learning spatial structures of proteins improves protein-protein interaction prediction",
                    "abstract": "Spatial structures of proteins are closely related to protein functions. Integrating protein structures improves the performance of protein-protein interaction (PPI) prediction. However, the limited quantity of known protein structures restricts the application of structure-based prediction methods. Utilizing the predicted protein structure information is a promising method to improve the performance of sequence-based prediction methods. We propose a novel end-to-end framework, TAGPPI, to predict PPIs using protein sequence alone. TAGPPI extracts multi-dimensional features by employing 1D convolution operation on protein sequences and graph learning method on contact maps constructed from AlphaFold. A contact map contains abundant spatial structure information, which is difficult to obtain from 1D sequence data directly. We further demonstrate that the spatial information learned from contact maps improves the ability of TAGPPI in PPI prediction tasks. We compare the performance of TAGPPI with those of nine state-of-the-art sequence-based methods, and TAGPPI outperforms such methods in all metrics. To the best of our knowledge, this is the first method to use the predicted protein topology structure graph for sequence-based PPI prediction. More importantly, our proposed architecture could be extended to other prediction tasks related to proteins."
                },
                {
                    "title": "Drug-drug interaction prediction with learnable size-adaptive molecular substructures",
                    "abstract": "Drug-drug interactions (DDIs) are interactions with adverse effects on the body, manifested when two or more incompatible drugs are taken together. They can be caused by the chemical compositions of the drugs involved. We introduce gated message passing neural network (GMPNN), a message passing neural network which learns chemical substructures with different sizes and shapes from the molecular graph representations of drugs for DDI prediction between a pair of drugs. In GMPNN, edges are considered as gates which control the flow of message passing, and therefore delimiting the substructures in a learnable way. The final DDI prediction between a drug pair is based on the interactions between pairs of their (learned) substructures, each pair weighted by a relevance score to the final DDI prediction output. Our proposed method GMPNN-CS (i.e. GMPNN + prediction module) is evaluated on two real-world datasets, with competitive results on one, and improved performance on the other compared with previous methods. Source code is freely available at https://github.com/kanz76/GMPNN-CS."
                },
                {
                    "title": "Improved prediction of protein-protein interaction using a hybrid of functional-link Siamese neural network and gradient boosting machines",
                    "abstract": "In this paper, for accurate prediction of protein-protein interaction (PPI), a novel hybrid classifier is developed by combining the functional-link Siamese neural network (FSNN) with the light gradient boosting machine (LGBM) classifier. The hybrid classifier (FSNN-LGBM) uses the fusion of features derived using pseudo amino acid composition and conjoint triad descriptors. The FSNN extracts the high-level abstraction features from the raw features and LGBM performs the PPI prediction task using these abstraction features. On performing 5-fold cross-validation experiments, the proposed hybrid classifier provides average accuracies of 98.70 and 98.38%, respectively, on the intraspecies PPI data sets of Saccharomyces cerevisiae and Helicobacter pylori. Similarly, the average accuracies for the interspecies PPI data sets of the Human-Bacillus and Human-Yersinia data sets are 98.52 and 97.40%, respectively. Compared with the existing methods, the hybrid classifier achieves higher prediction accuracy on the independent test sets and network data sets. The improved prediction performance obtained by the FSNN-LGBM makes it a flexible and effective PPI prediction model."
                },
                {
                    "title": "DGL-LifeSci: An Open-Source Toolkit for Deep Learning on Graphs in Life Science",
                    "abstract": "Graph neural networks (GNNs) constitute a class of deep learning methods for graph data. They have wide applications in chemistry and biology, such as molecular property prediction, reaction prediction, and drug\u2013target interaction prediction. Despite the interest, GNN-based modeling is challenging as it requires graph data preprocessing and modeling in addition to programming and deep learning. Here, we present Deep Graph Library (DGL)-LifeSci, an open-source package for deep learning on graphs in life science. Deep Graph Library (DGL)-LifeSci is a python toolkit based on RDKit, PyTorch, and Deep Graph Library (DGL). DGL-LifeSci allows GNN-based modeling on custom datasets for molecular property prediction, reaction prediction, and molecule generation. With its command-line interfaces, users can perform modeling without any background in programming and deep learning. We test the command-line interfaces using standard benchmarks MoleculeNet, USPTO, and ZINC. Compared with previous implementations, DGL-LifeSci achieves a speed up by up to 6\u00d7. For modeling flexibility, DGL-LifeSci provides well-optimized modules for various stages of the modeling pipeline. In addition, DGL-LifeSci provides pretrained models for reproducing the test experiment results and applying models without training. The code is distributed under an Apache-2.0 License and is freely accessible at https://github.com/awslabs/dgl-lifesci."
                },
                {
                    "title": "Application of Machine Learning for Drug\u2013Target Interaction Prediction",
                    "abstract": "Exploring drug\u2013target interactions by biomedical experiments requires a lot of human, financial, and material resources. To save time and cost to meet the needs of the present generation, machine learning methods have been introduced into the prediction of drug\u2013target interactions. The large amount of available drug and target data in existing databases, the evolving and innovative computer technologies, and the inherent characteristics of various types of machine learning have made machine learning techniques the mainstream method for drug\u2013target interaction prediction research. In this review, details of the specific applications of machine learning in drug\u2013target interaction prediction are summarized, the characteristics of each algorithm are analyzed, and the issues that need to be further addressed and explored for future research are discussed. The aim of this review is to provide a sound basis for the construction of high-performance models."
                },
                {
                    "title": "Assigning confidence to molecular property prediction",
                    "abstract": "ABSTRACT Introduction: Computational modeling has rapidly advanced over the last decades. Recently, machine learning has emerged as a powerful and cost-effective strategy to learn from existing datasets and perform predictions on unseen molecules. Accordingly, the explosive rise of data-driven techniques raises an important question: What confidence can be assigned to molecular property predictions and what techniques can be used? Areas covered: The authors discuss popular strategies for predicting molecular properties, their corresponding uncertainty sources and methods to quantify uncertainty. First, the authors\u2019 considerations for assessing confidence begin with dataset bias and size, data-driven property prediction and feature design. Next, the authors discuss property simulation via computations of binding affinity in detail. Lastly, they investigate how these uncertainties propagate to generative models, as they are usually coupled with property predictors. Expert opinion: Computational techniques are paramount to reduce the prohibitive cost of brute-force experimentation during exploration. The authors believe that assessing uncertainty in property prediction models is essential whenever closed-loop drug design campaigns relying on high-throughput virtual screening are deployed. Accordingly, considering sources of uncertainty leads to better-informed validations, more reliable predictions and more realistic expectations of the entire workflow. Overall, this increases confidence in the predictions and, ultimately, accelerates drug design."
                },
                {
                    "title": "Therapeutics Data Commons: Machine Learning Datasets and Tasks for Drug Discovery and Development",
                    "abstract": "Therapeutics machine learning is an emerging field with incredible opportunities for innovatiaon and impact. However, advancement in this field requires formulation of meaningful learning tasks and careful curation of datasets. Here, we introduce Therapeutics Data Commons (TDC), the first unifying platform to systematically access and evaluate machine learning across the entire range of therapeutics. To date, TDC includes 66 AI-ready datasets spread across 22 learning tasks and spanning the discovery and development of safe and effective medicines. TDC also provides an ecosystem of tools and community resources, including 33 data functions and types of meaningful data splits, 23 strategies for systematic model evaluation, 17 molecule generation oracles, and 29 public leaderboards. All resources are integrated and accessible via an open Python library. We carry out extensive experiments on selected datasets, demonstrating that even the strongest algorithms fall short of solving key therapeutics challenges, including real dataset distributional shifts, multi-scale modeling of heterogeneous data, and robust generalization to novel data points. We envision that TDC can facilitate algorithmic and scientific advances and considerably accelerate machine-learning model development, validation and transition into biomedical and clinical implementation. TDC is an open-science initiative available at https://tdcommons.ai."
                },
                {
                    "title": "DeepPurpose: a deep learning library for drug\u2013target interaction prediction",
                    "abstract": "Abstract Summary Accurate prediction of drug\u2013target interactions (DTI) is crucial for drug discovery. Recently, deep learning (DL) models for show promising performance for DTI prediction. However, these models can be difficult to use for both computer scientists entering the biomedical field and bioinformaticians with limited DL experience. We present DeepPurpose, a comprehensive and easy-to-use DL library for DTI prediction. DeepPurpose supports training of customized DTI prediction models by implementing 15 compound and protein encoders and over 50 neural architectures, along with providing many other useful features. We demonstrate state-of-the-art performance of DeepPurpose on several benchmark datasets. Availability and implementation https://github.com/kexinhuang12345/DeepPurpose. Supplementary information Supplementary data are available at Bioinformatics online."
                },
                {
                    "title": "Learning from Protein Structure with Geometric Vector Perceptrons",
                    "abstract": "Learning on 3D structures of large biomolecules is emerging as a distinct area in machine learning, but there has yet to emerge a unifying network architecture that simultaneously leverages the graph-structured and geometric aspects of the problem domain. To address this gap, we introduce geometric vector perceptrons, which extend standard dense layers to operate on collections of Euclidean vectors. Graph neural networks equipped with such layers are able to perform both geometric and relational reasoning on efficient and natural representations of macromolecular structure. We demonstrate our approach on two important problems in learning from protein structure: model quality assessment and computational protein design. Our approach improves over existing classes of architectures, including state-of-the-art graph-based and voxel-based methods."
                },
                {
                    "title": "Pushing the boundaries of molecular representation for drug discovery with graph attention mechanism.",
                    "abstract": "Hunting for chemicals with favourable pharmacological, toxicological and pharmacokinetic properties remains a formidable challenge for drug discovery. Deep learning provides us with powerful tools to build predictive models that are appropriate for the rising amounts of data, but the gap between what these neural networks learn and what human beings can comprehend is growing. Moreover, this gap may induce distrust and restrict deep learning applications in practice. Here, we introduce a new graph neural network architecture called Attentive FP for molecular representation that uses a graph attention mechanism to learn from relevant drug discovery datasets. We demonstrate that Attentive FP achieves state-of-the-art predictive performances on a variety of datasets and that what it learns is interpretable. The feature visualization for Attentive FP suggests that it automatically learns non-local intramolecular interactions from specified tasks, which can help us gain chemical insights directly from data beyond human perception."
                },
                {
                    "title": "Graphein - a Python Library for Geometric Deep Learning and Network Analysis on Protein Structures and Interaction Networks",
                    "abstract": "Geometric deep learning has well-motivated applications in the context of biology, a domain where relational structure in datasets can be meaningfully leveraged. Currently, efforts in both geometric deep learning and, more broadly, deep learning applied to biomolecular tasks have been hampered by a scarcity of appropriate datasets accessible to domain specialists and machine learning researchers alike. However, there has been little exploration of how to best to integrate and construct geometric representations of these datatypes. To address this, we introduce Graphein as a turn-key tool for transforming raw data from widely-used bioinformatics databases into machine learning-ready datasets in a high-throughput and flexible manner. Graphein is a Python library for constructing graph and surface-mesh representations of protein structures and biological interaction networks for computational analysis. Graphein provides utilities for data retrieval from widely-used bioinformatics databases for structural data, including the Protein Data Bank, the recently-released AlphaFold Structure Database, and for biomolecular interaction networks from STRINGdb, BioGrid, TRRUST and RegNetwork. The library interfaces with popular geometric deep learning libraries: DGL, PyTorch Geometric and PyTorch3D though remains framework agnostic as it is built on top of the PyData ecosystem to enable inter-operability with scientific computing tools and libraries. Graphein is designed to be highly flexible, allowing the user to specify each step of the data preparation, scalable to facilitate working with large protein complexes and interaction graphs, and contains useful pre-processing tools for preparing experimental files. Graphein facilitates network-based, graph-theoretic and topological analyses of structural and interaction datasets in a high-throughput manner. As example workflows, we make available two new protein structure-related datasets, previously unused by the geometric deep learning community. We envision that Graphein will facilitate developments in computational biology, graph representation learning and drug discovery. Availability and implementation Graphein is written in Python. Source code, example usage and tutorials, datasets, and documentation are made freely available under the MIT License at the following URL: graphein.ai"
                },
                {
                    "title": "ProtTrans: Towards Cracking the Language of Life\u2019s Code Through Self-Supervised Deep Learning and High Performance Computing",
                    "abstract": "Computational biology and bioinformatics provide vast data gold-mines from protein sequences, ideal for Language Models (LMs) taken from Natural Language Processing (NLP). These LMs reach for new prediction frontiers at low inference costs. Here, we trained two auto-regressive language models (Transformer-XL, XLNet) and two auto-encoder models (Bert, Albert) on data from UniRef and BFD containing up to 393 billion amino acids (words) from 2.1 billion protein sequences (22- and 112-times the entire English Wikipedia). The LMs were trained on the Summit supercomputer at Oak Ridge National Laboratory (ORNL), using 936 nodes (total 5616 GPUs) and one TPU Pod (V3-512 or V3-1024). We validated the advantage of up-scaling LMs to larger models supported by bigger data by predicting secondary structure (3-states: Q3=76-84, 8-states: Q8=65-73), sub-cellular localization for 10 cellular compartments (Q10=74) and whether a protein is membrane-bound or water-soluble (Q2=89). Dimensionality reduction revealed that the LM-embeddings from unlabeled data (only protein sequences) captured important biophysical properties governing protein shape. This implied learning some of the grammar of the language of life realized in protein sequences. The successful up-scaling of protein LMs through HPC to larger data sets slightly reduced the gap between models trained on evolutionary information and LMs."
                },
                {
                    "title": "MolTrans: Molecular Interaction Transformer for drug\u2013target interaction prediction",
                    "abstract": "Abstract Motivation Drug\u2013target interaction (DTI) prediction is a foundational task for in-silico drug discovery, which is costly and time-consuming due to the need of experimental search over large drug compound space. Recent years have witnessed promising progress for deep learning in DTI predictions. However, the following challenges are still open: (i) existing molecular representation learning approaches ignore the sub-structural nature of DTI, thus produce results that are less accurate and difficult to explain and (ii) existing methods focus on limited labeled data while ignoring the value of massive unlabeled molecular data. Results We propose a Molecular Interaction Transformer (MolTrans) to address these limitations via: (i) knowledge inspired sub-structural pattern mining algorithm and interaction modeling module for more accurate and interpretable DTI prediction and (ii) an augmented transformer encoder to better extract and capture the semantic relations among sub-structures extracted from massive unlabeled biomedical data. We evaluate MolTrans on real-world data and show it improved DTI prediction performance compared to state-of-the-art baselines. Availability and implementation The model scripts are available at https://github.com/kexinhuang12345/moltrans. Supplementary information Supplementary data are available at Bioinformatics online."
                },
                {
                    "title": "Multifaceted protein\u2013protein interaction prediction based on Siamese residual RCNN",
                    "abstract": "Abstract Motivation Sequence-based protein\u2013protein interaction (PPI) prediction represents a fundamental computational biology problem. To address this problem, extensive research efforts have been made to extract predefined features from the sequences. Based on these features, statistical algorithms are learned to classify the PPIs. However, such explicit features are usually costly to extract, and typically have limited coverage on the PPI information. Results We present an end-to-end framework, PIPR (Protein\u2013Protein Interaction Prediction Based on Siamese Residual RCNN), for PPI predictions using only the protein sequences. PIPR incorporates a deep residual recurrent convolutional neural network in the Siamese architecture, which leverages both robust local features and contextualized information, which are significant for capturing the mutual influence of proteins sequences. PIPR relieves the data pre-processing efforts that are required by other systems, and generalizes well to different application scenarios. Experimental evaluations show that PIPR outperforms various state-of-the-art systems on the binary PPI prediction problem. Moreover, it shows a promising performance on more challenging problems of interaction type prediction and binding affinity estimation, where existing approaches fall short. Availability and implementation The implementation is available at https://github.com/muhaochen/seq_ppi.git. Supplementary information Supplementary data are available at Bioinformatics online."
                },
                {
                    "title": "Molecular Property Prediction: A Multilevel Quantum Interactions Modeling Perspective",
                    "abstract": "Predicting molecular properties (e.g., atomization energy) is an essential issue in quantum chemistry, which could speed up much research progress, such as drug designing and substance discovery. Traditional studies based on density functional theory (DFT) in physics are proved to be time-consuming for predicting large number of molecules. Recently, the machine learning methods, which consider much rule-based information, have also shown potentials for this issue. However, the complex inherent quantum interactions of molecules are still largely underexplored by existing solutions. In this paper, we propose a generalizable and transferable Multilevel Graph Convolutional neural Network (MGCN) for molecular property prediction. Specifically, we represent each molecule as a graph to preserve its internal structure. Moreover, the well-designed hierarchical graph neural network directly extracts features from the conformation and spatial information followed by the multilevel interactions. As a consequence, the multilevel overall representations can be utilized to make the prediction. Extensive experiments on both datasets of equilibrium and off-equilibrium molecules demonstrate the effectiveness of our model. Furthermore, the detailed results also prove that MGCN is generalizable and transferable for the prediction."
                },
                {
                    "title": "MR-GNN: Multi-Resolution and Dual Graph Neural Network for Predicting Structured Entity Interactions",
                    "abstract": "Predicting interactions between structured entities lies at the core of numerous tasks such as drug regimen and new material design. In recent years, graph neural networks have become attractive. They represent structured entities as graphs, and then extract features from each individual graph using graph convolution operations. However, these methods have some limitations: i) their networks only extract features from a fix-sized subgraph structure (i.e., a fix-sized receptive field) of each node, and ignore features in substructures of different sizes, and ii) features are extracted by considering each entity independently, which may not effectively reflect the interaction between two entities. To resolve these problems, we present {\\em MR-GNN}, an end-to-end graph neural network with the following features: i) it uses a multi-resolution based architecture to extract node features from different neighborhoods of each node, and, ii) it uses dual graph-state long short-term memory networks (LSTMs) to summarize local features of each graph and extracts the interaction features between pairwise graphs. Experiments conducted on real-world datasets show that MR-GNN improves the prediction of state-of-the-art methods."
                },
                {
                    "title": "Drug-Drug Adverse Effect Prediction with Graph Co-Attention",
                    "abstract": "Complex or co-existing diseases are commonly treated using drug combinations, which can lead to higher risk of adverse side effects. The detection of polypharmacy side effects is usually done in Phase IV clinical trials, but there are still plenty which remain undiscovered when the drugs are put on the market. Such accidents have been affecting an increasing proportion of the population (15% in the US now) and it is thus of high interest to be able to predict the potential side effects as early as possible. Systematic combinatorial screening of possible drug-drug interactions (DDI) is challenging and expensive. However, the recent significant increases in data availability from pharmaceutical research and development efforts offer a novel paradigm for recovering relevant insights for DDI prediction. Accordingly, several recent approaches focus on curating massive DDI datasets (with millions of examples) and training machine learning models on them. Here we propose a neural network architecture able to set state-of-the-art results on this task---using the type of the side-effect and the molecular structure of the drugs alone---by leveraging a co-attentional mechanism. In particular, we show the importance of integrating joint information from the drug pairs early on when learning each drug's representation."
                },
                {
                    "title": "Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences",
                    "abstract": "Significance Learning biological properties from sequence data is a logical step toward generative and predictive artificial intelligence for biology. Here, we propose scaling a deep contextual language model with unsupervised learning to sequences spanning evolutionary diversity. We find that without prior knowledge, information emerges in the learned representations on fundamental properties of proteins such as secondary structure, contacts, and biological activity. We show the learned representations are useful across benchmarks for remote homology detection, prediction of secondary structure, long-range residue\u2013residue contacts, and mutational effect. Unsupervised representation learning enables state-of-the-art supervised prediction of mutational effect and secondary structure and improves state-of-the-art features for long-range contact prediction. In the field of artificial intelligence, a combination of scale in data and model capacity enabled by unsupervised learning has led to major advances in representation learning and statistical generation. In the life sciences, the anticipated growth of sequencing promises unprecedented data on natural sequence diversity. Protein language modeling at the scale of evolution is a logical step toward predictive and generative artificial intelligence for biology. To this end, we use unsupervised learning to train a deep contextual language model on 86 billion amino acids across 250 million protein sequences spanning evolutionary diversity. The resulting model contains information about biological properties in its representations. The representations are learned from sequence data alone. The learned representation space has a multiscale organization reflecting structure from the level of biochemical properties of amino acids to remote homology of proteins. Information about secondary and tertiary structure is encoded in the representations and can be identified by linear projections. Representation learning produces features that generalize across a range of applications, enabling state-of-the-art supervised prediction of mutational effect and secondary structure and improving state-of-the-art features for long-range contact prediction."
                },
                {
                    "title": "DeepConv-DTI: Prediction of drug-target interactions via deep learning with convolution on protein sequences",
                    "abstract": "Identification of drug-target interactions (DTIs) plays a key role in drug discovery. The high cost and labor-intensive nature of in vitro and in vivo experiments have highlighted the importance of in silico-based DTI prediction approaches. In several computational models, conventional protein descriptors have been shown to not be sufficiently informative to predict accurate DTIs. Thus, in this study, we propose a deep learning based DTI prediction model capturing local residue patterns of proteins participating in DTIs. When we employ a convolutional neural network (CNN) on raw protein sequences, we perform convolution on various lengths of amino acids subsequences to capture local residue patterns of generalized protein classes. We train our model with large-scale DTI information and demonstrate the performance of the proposed model using an independent dataset that is not seen during the training phase. As a result, our model performs better than previous protein descriptor-based models. Also, our model performs better than the recently developed deep learning models for massive prediction of DTIs. By examining pooled convolution results, we confirmed that our model can detect binding sites of proteins for DTIs. In conclusion, our prediction model for detecting local residue patterns of target proteins successfully enriches the protein features of a raw protein sequence, yielding better prediction results than previous approaches. Our code is available at https://github.com/GIST-CSBL/DeepConv-DTI."
                },
                {
                    "title": "PubChem 2019 update: improved access to chemical data",
                    "abstract": "Abstract PubChem (https://pubchem.ncbi.nlm.nih.gov) is a key chemical information resource for the biomedical research community. Substantial improvements were made in the past few years. New data content was added, including spectral information, scientific articles mentioning chemicals, and information for food and agricultural chemicals. PubChem released new web interfaces, such as PubChem Target View page, Sources page, Bioactivity dyad pages and Patent View page. PubChem also released a major update to PubChem Widgets and introduced a new programmatic access interface, called PUG-View. This paper describes these new developments in PubChem."
                },
                {
                    "title": "How Powerful are Graph Neural Networks?",
                    "abstract": "Graph Neural Networks (GNNs) are an effective framework for representation learning of graphs. GNNs follow a neighborhood aggregation scheme, where the representation vector of a node is computed by recursively aggregating and transforming representation vectors of its neighboring nodes. Many GNN variants have been proposed and have achieved state-of-the-art results on both node and graph classification tasks. However, despite GNNs revolutionizing graph representation learning, there is limited understanding of their representational properties and limitations. Here, we present a theoretical framework for analyzing the expressive power of GNNs to capture different graph structures. Our results characterize the discriminative power of popular GNN variants, such as Graph Convolutional Networks and GraphSAGE, and show that they cannot learn to distinguish certain simple graph structures. We then develop a simple architecture that is provably the most expressive among the class of GNNs and is as powerful as the Weisfeiler-Lehman graph isomorphism test. We empirically validate our theoretical findings on a number of graph classification benchmarks, and demonstrate that our model achieves state-of-the-art performance."
                },
                {
                    "title": "Compound\u2010protein interaction prediction with end\u2010to\u2010end learning of neural networks for graphs and sequences",
                    "abstract": "Motivation: In bioinformatics, machine learning\u2010based methods that predict the compound\u2010protein interactions (CPIs) play an important role in the virtual screening for drug discovery. Recently, end\u2010to\u2010end representation learning for discrete symbolic data (e.g. words in natural language processing) using deep neural networks has demonstrated excellent performance on various difficult problems. For the CPI problem, data are provided as discrete symbolic data, i.e. compounds are represented as graphs where the vertices are atoms, the edges are chemical bonds, and proteins are sequences in which the characters are amino acids. In this study, we investigate the use of end\u2010to\u2010end representation learning for compounds and proteins, integrate the representations, and develop a new CPI prediction approach by combining a graph neural network (GNN) for compounds and a convolutional neural network (CNN) for proteins. Results: Our experiments using three CPI datasets demonstrated that the proposed end\u2010to\u2010end approach achieves competitive or higher performance as compared to various existing CPI prediction methods. In addition, the proposed approach significantly outperformed existing methods on an unbalanced dataset. This suggests that data\u2010driven representations of compounds and proteins obtained by end\u2010to\u2010end GNNs and CNNs are more robust than traditional chemical and biological features obtained from databases. Although analyzing deep learning models is difficult due to their black\u2010box nature, we address this issue using a neural attention mechanism, which allows us to consider which subsequences in a protein are more important for a drug compound when predicting its interaction. The neural attention mechanism also provides effective visualization, which makes it easier to analyze a model even when modeling is performed using real\u2010valued representations instead of discrete features. Availability and implementation: https://github.com/masashitsubaki Supplementary information: Supplementary data are available at Bioinformatics online."
                },
                {
                    "title": "Deep learning improves prediction of drug\u2013drug and drug\u2013food interactions",
                    "abstract": "Significance Drug interactions, including drug\u2013drug interactions (DDIs) and drug\u2013food constituent interactions, can trigger unexpected pharmacological effects such as adverse drug events (ADEs). Several existing methods predict drug interactions, but require detailed, but often unavailable drug information as inputs, such as drug targets. To this end, we present a computational framework DeepDDI that accurately predicts DDI types for given drug pairs and drug\u2013food constituent pairs using only name and structural information as inputs. We show four applications of DeepDDI to better understand drug interactions, including prediction of DDI mechanisms causing ADEs, suggestion of alternative drug members for the intended pharmacological effects without negative health effects, prediction of the effects of food constituents on interacting drugs, and prediction of bioactivities of food constituents. Drug interactions, including drug\u2013drug interactions (DDIs) and drug\u2013food constituent interactions (DFIs), can trigger unexpected pharmacological effects, including adverse drug events (ADEs), with causal mechanisms often unknown. Several computational methods have been developed to better understand drug interactions, especially for DDIs. However, these methods do not provide sufficient details beyond the chance of DDI occurrence, or require detailed drug information often unavailable for DDI prediction. Here, we report development of a computational framework DeepDDI that uses names of drug\u2013drug or drug\u2013food constituent pairs and their structural information as inputs to accurately generate 86 important DDI types as outputs of human-readable sentences. DeepDDI uses deep neural network with its optimized prediction performance and predicts 86 DDI types with a mean accuracy of 92.4% using the DrugBank gold standard DDI dataset covering 192,284 DDIs contributed by 191,878 drug pairs. DeepDDI is used to suggest potential causal mechanisms for the reported ADEs of 9,284 drug pairs, and also predict alternative drug candidates for 62,707 drug pairs having negative health effects. Furthermore, DeepDDI is applied to 3,288,157 drug\u2013food constituent pairs (2,159 approved drugs and 1,523 well-characterized food constituents) to predict DFIs. The effects of 256 food constituents on pharmacological effects of interacting drugs and bioactivities of 149 food constituents are predicted. These results suggest that DeepDDI can provide important information on drug prescription and even dietary suggestions while taking certain drugs and also guidelines during drug development."
                },
                {
                    "title": "DeepDTA: deep drug\u2013target binding affinity prediction",
                    "abstract": "Motivation The identification of novel drug\u2010target (DT) interactions is a substantial part of the drug discovery process. Most of the computational methods that have been proposed to predict DT interactions have focused on binary classification, where the goal is to determine whether a DT pair interacts or not. However, protein\u2010ligand interactions assume a continuum of binding strength values, also called binding affinity and predicting this value still remains a challenge. The increase in the affinity data available in DT knowledge\u2010bases allows the use of advanced learning techniques such as deep learning architectures in the prediction of binding affinities. In this study, we propose a deep\u2010learning based model that uses only sequence information of both targets and drugs to predict DT interaction binding affinities. The few studies that focus on DT binding affinity prediction use either 3D structures of protein\u2010ligand complexes or 2D features of compounds. One novel approach used in this work is the modeling of protein sequences and compound 1D representations with convolutional neural networks (CNNs). Results The results show that the proposed deep learning based model that uses the 1D representations of targets and drugs is an effective approach for drug target binding affinity prediction. The model in which high\u2010level representations of a drug and a target are constructed via CNNs achieved the best Concordance Index (CI) performance in one of our larger benchmark datasets, outperforming the KronRLS algorithm and SimBoost, a state\u2010of\u2010the\u2010art method for DT binding affinity prediction. Availability and implementation https://github.com/hkmztrk/DeepDTA Supplementary information Supplementary data are available at Bioinformatics online."
                },
                {
                    "title": "Graph Attention Networks",
                    "abstract": "We present graph attention networks (GATs), novel neural network architectures that operate on graph-structured data, leveraging masked self-attentional layers to address the shortcomings of prior methods based on graph convolutions or their approximations. By stacking layers in which nodes are able to attend over their neighborhoods' features, we enable (implicitly) specifying different weights to different nodes in a neighborhood, without requiring any kind of costly matrix operation (such as inversion) or depending on knowing the graph structure upfront. In this way, we address several key challenges of spectral-based graph neural networks simultaneously, and make our model readily applicable to inductive as well as transductive problems. Our GAT models have achieved or matched state-of-the-art results across four established transductive and inductive graph benchmarks: the Cora, Citeseer and Pubmed citation network datasets, as well as a protein-protein interaction dataset (wherein test graphs remain unseen during training)."
                },
                {
                    "title": "SchNet: A continuous-filter convolutional neural network for modeling quantum interactions",
                    "abstract": "Deep learning has the potential to revolutionize quantum chemistry as it is ideally suited to learn representations for structured data and speed up the exploration of chemical space. While convolutional neural networks have proven to be the first choice for images, audio and video data, the atoms in molecules are not restricted to a grid. Instead, their precise locations contain essential physical information, that would get lost if discretized. Thus, we propose to use continuous-filter convolutional layers to be able to model local correlations without requiring the data to lie on a grid. We apply those layers in SchNet: a novel deep learning architecture modeling quantum interactions in molecules. We obtain a joint model for the total energy and interatomic forces that follows fundamental quantum-chemical principles. This includes rotationally invariant energy predictions and a smooth, differentiable potential energy surface. Our architecture achieves state-of-the-art performance for benchmarks of equilibrium molecules and molecular dynamics trajectories. Finally, we introduce a more challenging benchmark with chemical and structural variations that suggests the path for further work."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Neural Message Passing for Quantum Chemistry",
                    "abstract": "Supervised learning on molecules has incredible potential to be useful in chemistry, drug discovery, and materials science. Luckily, several promising and closely related neural network models invariant to molecular symmetries have already been described in the literature. These models learn a message passing algorithm and aggregation procedure to compute a function of their entire input graph. At this point, the next step is to find a particularly effective variant of this general approach and apply it to chemical prediction benchmarks until we either solve them or reach the limits of the approach. In this paper, we reformulate existing models into a single common framework we call Message Passing Neural Networks (MPNNs) and explore additional novel variations within this framework. Using MPNNs we demonstrate state of the art results on an important molecular property prediction benchmark; these results are strong enough that we believe future work should focus on datasets with larger molecules or more accurate ground truth labels."
                },
                {
                    "title": "Deep-Learning-Based Drug-Target Interaction Prediction.",
                    "abstract": "Identifying interactions between known drugs and targets is a major challenge in drug repositioning. In silico prediction of drug-target interaction (DTI) can speed up the expensive and time-consuming experimental work by providing the most potent DTIs. In silico prediction of DTI can also provide insights about the potential drug-drug interaction and promote the exploration of drug side effects. Traditionally, the performance of DTI prediction depends heavily on the descriptors used to represent the drugs and the target proteins. In this paper, to accurately predict new DTIs between approved drugs and targets without separating the targets into different classes, we developed a deep-learning-based algorithmic framework named DeepDTIs. It first abstracts representations from raw input descriptors using unsupervised pretraining and then applies known label pairs of interaction to build a classification model. Compared with other methods, it is found that DeepDTIs reaches or outperforms other state-of-the-art methods. The DeepDTIs can be further used to predict whether a new drug targets to some existing targets or whether a new target interacts with some existing drugs."
                },
                {
                    "title": "Gated Multimodal Units for Information Fusion",
                    "abstract": "This paper presents a novel model for multimodal learning based on gated neural networks. The Gated Multimodal Unit (GMU) model is intended to be used as an internal unit in a neural network architecture whose purpose is to find an intermediate representation based on a combination of data from different modalities. The GMU learns to decide how modalities influence the activation of the unit using multiplicative gates. It was evaluated on a multilabel scenario for genre classification of movies using the plot and the poster. The GMU improved the macro f-score performance of single-modality approaches and outperformed other fusion strategies, including mixture of experts models. Along with this work, the MM-IMDb dataset is released which, to the best of our knowledge, is the largest publicly available multimodal dataset for genre prediction on movies."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Convolutional Networks on Graphs for Learning Molecular Fingerprints",
                    "abstract": "We introduce a convolutional neural network that operates directly on graphs. These networks allow end-to-end learning of prediction pipelines whose inputs are graphs of arbitrary size and shape. The architecture we present generalizes standard molecular feature extraction methods based on circular fingerprints. We show that these data-driven features are more interpretable, and have better predictive performance on a variety of tasks."
                },
                {
                    "title": "Bilinear CNN Models for Fine-Grained Visual Recognition",
                    "abstract": "We propose bilinear models, a recognition architecture that consists of two feature extractors whose outputs are multiplied using outer product at each location of the image and pooled to obtain an image descriptor. This architecture can model local pairwise feature interactions in a translationally invariant manner which is particularly useful for fine-grained categorization. It also generalizes various orderless texture descriptors such as the Fisher vector, VLAD and O2P. We present experiments with bilinear models where the feature extractors are based on convolutional neural networks. The bilinear form simplifies gradient computation and allows end-to-end training of both networks using image labels only. Using networks initialized from the ImageNet dataset followed by domain specific fine-tuning we obtain 84.1% accuracy of the CUB-200-2011 dataset requiring only category labels at training time. We present experiments and visualizations that analyze the effects of fine-tuning and the choice two networks on the speed and accuracy of the models. Results show that the architecture compares favorably to the existing state of the art on a number of fine-grained datasets while being substantially simpler and easier to train. Moreover, our most accurate model is fairly efficient running at 8 frames/sec on a NVIDIA Tesla K40 GPU. The source code for the complete system will be made available at http://vis-www.cs.umass.edu/bcnn."
                },
                {
                    "title": "propy: a tool to generate various modes of Chou's PseAAC",
                    "abstract": "SUMMARY\nSequence-derived structural and physiochemical features have been frequently used for analysing and predicting structural, functional, expression and interaction profiles of proteins and peptides. To facilitate extensive studies of proteins and peptides, we developed a freely available, open source python package called protein in python (propy) for calculating the widely used structural and physicochemical features of proteins and peptides from amino acid sequence. It computes five feature groups composed of 13 features, including amino acid composition, dipeptide composition, tripeptide composition, normalized Moreau-Broto autocorrelation, Moran autocorrelation, Geary autocorrelation, sequence-order-coupling number, quasi-sequence-order descriptors, composition, transition and distribution of various structural and physicochemical properties and two types of pseudo amino acid composition (PseAAC) descriptors. These features could be generally regarded as different Chou's PseAAC modes. In addition, it can also easily compute the previous descriptors based on user-defined properties, which are automatically available from the AAindex database.\n\n\nAVAILABILITY\nThe python package, propy, is freely available via http://code.google.com/p/protpy/downloads/list, and it runs on Linux and MS-Windows.\n\n\nSUPPLEMENTARY INFORMATION\nSupplementary data are available at Bioinformatics online."
                },
                {
                    "title": "Extended-Connectivity Fingerprints",
                    "abstract": "Extended-connectivity fingerprints (ECFPs) are a novel class of topological fingerprints for molecular characterization. Historically, topological fingerprints were developed for substructure and similarity searching. ECFPs were developed specifically for structure-activity modeling. ECFPs are circular fingerprints with a number of useful qualities: they can be very rapidly calculated; they are not predefined and can represent an essentially infinite number of different molecular features (including stereochemical information); their features represent the presence of particular substructures, allowing easier interpretation of analysis results; and the ECFP algorithm can be tailored to generate different types of circular fingerprints, optimized for different uses. While the use of ECFPs has been widely adopted and validated, a description of their implementation has not previously been presented in the literature."
                },
                {
                    "title": "Using support vector machine combined with auto covariance to predict protein\u2013protein interactions from protein sequences",
                    "abstract": "Compared to the available protein sequences of different organisms, the number of revealed protein\u2013protein interactions (PPIs) is still very limited. So many computational methods have been developed to facilitate the identification of novel PPIs. However, the methods only using the information of protein sequences are more universal than those that depend on some additional information or predictions about the proteins. In this article, a sequence-based method is proposed by combining a new feature representation using auto covariance (AC) and support vector machine (SVM). AC accounts for the interactions between residues a certain distance apart in the sequence, so this method adequately takes the neighbouring effect into account. When performed on the PPI data of yeast Saccharomyces cerevisiae, the method achieved a very promising prediction result. An independent data set of 11 474 yeast PPIs was used to evaluate this prediction model and the prediction accuracy is 88.09%. The performance of this method is superior to those of the existing sequence-based methods, so it can be a useful supplementary tool for future proteomics studies. The prediction software and all data sets used in this article are freely available at http://www.scucic.cn/Predict_PPI/index.htm."
                },
                {
                    "title": "Predicting protein\u2013protein interactions based only on sequences information",
                    "abstract": "Protein\u2013protein interactions (PPIs) are central to most biological processes. Although efforts have been devoted to the development of methodology for predicting PPIs and protein interaction networks, the application of most existing methods is limited because they need information about protein homology or the interaction marks of the protein partners. In the present work, we propose a method for PPI prediction using only the information of protein sequences. This method was developed based on a learning algorithm-support vector machine combined with a kernel function and a conjoint triad feature for describing amino acids. More than 16,000 diverse PPI pairs were used to construct the universal model. The prediction ability of our approach is better than that of other sequence-based PPI prediction methods because it is able to predict PPI networks. Different types of PPI networks have been effectively mapped with our method, suggesting that, even with only sequence information, this method could be applied to the exploration of networks for any newly discovered protein with unknown biological relativity. In addition, such supplementary experimental information can enhance the prediction ability of the method."
                },
                {
                    "title": "Prediction of protein cellular attributes using pseudo\u2010amino acid composition",
                    "abstract": "The cellular attributes of a protein, such as which compartment of a cell it belongs to and how it is associated with the lipid bilayer of an organelle, are closely correlated with its biological functions. The success of human genome project and the rapid increase in the number of protein sequences entering into data bank have stimulated a challenging frontier: How to develop a fast and accurate method to predict the cellular attributes of a protein based on its amino acid sequence? The existing algorithms for predicting these attributes were all based on the amino acid composition in which no sequence order effect was taken into account. To improve the prediction quality, it is necessary to incorporate such an effect. However, the number of possible patterns for protein sequences is extremely large, which has posed a formidable difficulty for realizing this goal. To deal with such a difficulty, the pseudo\u2010amino acid composition is introduced. It is a combination of a set of discrete sequence correlation factors and the 20 components of the conventional amino acid composition. A remarkable improvement in prediction quality has been observed by using the pseudo\u2010amino acid composition. The success rates of prediction thus obtained are so far the highest for the same classification schemes and same data sets. It has not escaped from our notice that the concept of pseudo\u2010amino acid composition as well as its mathematical framework and biochemical implication may also have a notable impact on improving the prediction quality of other protein features. Proteins 2001;43:246\u2013255. \u00a9 2001 Wiley\u2010Liss, Inc."
                },
                {
                    "title": "Prediction of protein subcellular locations by incorporating quasi-sequence-order effect.",
                    "abstract": "How to incorporate the sequence order effect is a key and logical step for improving the prediction quality of protein subcellular location, but meanwhile it is a very difficult problem as well. This is because the number of possible sequence order patterns in proteins is extremely large, which has posed a formidable barrier to construct an effective training data set for statistical treatment based on the current knowledge. That is why most of the existing prediction algorithms are operated based on the amino-acid composition alone. In this paper, based on the physicochemical distance between amino acids, a set of sequence-order-coupling numbers was introduced to reflect the sequence order effect, or in a rigorous term, the quasi-sequence-order effect. Furthermore, the covariant discriminant algorithm by Chou and Elrod (Protein Eng. 12, 107-118, 1999) developed recently was augmented to allow the prediction performed by using the input of both the sequence-order-coupling numbers and amino-acid composition. A remarkable improvement was observed in the prediction quality using the augmented covariant discriminant algorithm. The approach described here represents one promising step forward in the efforts of incorporating sequence order effect in protein subcellular location prediction. It is anticipated that the current approach may also have a series of impacts on the prediction of other protein features by statistical approaches."
                },
                {
                    "title": "Adverse drug events in hospitalized patients. Excess length of stay, extra costs, and attributable mortality.",
                    "abstract": "OBJECTIVE\nTo determine the excess length of stay, extra costs, and mortality attributable to adverse drug events (ADEs) in hospitalized patients.\n\n\nDESIGN\nMatched case-control study.\n\n\nSETTING\nThe LDS Hospital, a tertiary care health care institution.\n\n\nPATIENTS\nAll patients admitted to LDS Hospital from January 1, 1990, to December 31, 1993, were eligible. Cases were defined as patients with ADEs that occurred during hospitalization; controls were selected according to matching variables in a stepwise fashion.\n\n\nMETHODS\nControls were matched to cases on primary discharge diagnosis related group (DRG), age, sex, acuity, and year of admission; varying numbers of controls were matched to each case. Matching was successful for 71% of the cases, leading to 1580 cases and 20,197 controls.\n\n\nMAIN OUTCOME MEASURES\nCrude and attributable mortality, crude and attributable length of stay, and cost of hospitalization.\n\n\nRESULTS\nADEs complicated 2.43 per 100 admissions to the LDS Hospital during the study period. The crude mortality rates for the cases and matched controls were 3.5% and 1.05%, respectively (P<.001). The mean length of hospital stay significantly differed between the cases and matched controls (7.69 vs 4.46 days; P<.001) as did the mean cost of hospitalization ($10,010 vs $5355; P<.001). The extra length of hospital stay attributable to an ADE was 1.74 days (P<.001). The excess cost of hospitalization attributable to an ADE was $2013 (P<.001). A linear regression analysis for length of stay and cost controlling for all matching variables revealed that the occurrence of an ADE was associated with increased length of stay of 1.91 days and an increased cost of $2262 (P<.001). In a similar logistic regression analysis for mortality, the increased risk of death among patients experiencing an ADE was 1.88 (95% confidence interval, 1.54-2.22; P<.001).\n\n\nCONCLUSION\nThe attributable lengths of stay and costs of hospitalization for ADEs are substantial. An ADE is associated with a significantly prolonged length of stay, increased economic burden, and an almost 2-fold increased risk of death."
                },
                {
                    "title": "Prediction of protein folding class using global description of amino acid sequence.",
                    "abstract": "We present a method for predicting protein folding class based on global protein chain description and a voting process. Selection of the best descriptors was achieved by a computer-simulated neural network trained on a data base consisting of 83 folding classes. Protein-chain descriptors include overall composition, transition, and distribution of amino acid attributes, such as relative hydrophobicity, predicted secondary structure, and predicted solvent exposure. Cross-validation testing was performed on 15 of the largest classes. The test shows that proteins were assigned to the correct class (correct positive prediction) with an average accuracy of 71.7%, whereas the inverse prediction of proteins as not belonging to a particular class (correct negative prediction) was 90-95% accurate. When tested on 254 structures used in this study, the top two predictions contained the correct class in 91% of the cases."
                },
                {
                    "title": "Prediction of protein helix content from an autocorrelation analysis of sequence hydrophobicities",
                    "abstract": "It is demonstrated that protein \u03b1\u2010helix content can be predicted from an autocorrelation analysis of the protein hydrophobicity sequence. The Fourier transform of the autocorrelation function yields the spectral densities or weights of the various frequencies contributing to the autocorrelation function. Using sequence and secondary structure data from more than 160 proteins and domains, a linear relationship was found between spectral density at periodicity 3.7 and protein \u03b1\u2010helix content (r = 0.83). This relation permits prediction of the helix content (x) of proteins of known sequence to within \u00b1 15%, i.e., as (x \u00b1 15)%. Predictions based on the autocorrelation procedure are compared with values obtained by other methods."
                },
                {
                    "title": "Mole-BERT: Rethinking Pre-training Graph Neural Networks for Molecules",
                    "abstract": "Recent years have witnessed the prosperity of pre-training graph neural networks (GNNs) for molecules. Typically, atom types as node attributes are randomly masked, and GNNs are then trained to predict masked types as in AttrMask (Hu et al., 2020), following the Masked Language Modeling (MLM) task of BERT (Devlin et al., 2019). However, unlike MLM with a large vocabulary, the AttrMask pre-training does not learn informative molecular representations due to small and unbalanced atom \u2018vocabulary\u2019. To amend this problem, we propose a variant of VQ-VAE (Van Den Oord et al., 2017) as a context-aware tokenizer to encode atom attributes into chemically meaningful discrete codes. This can enlarge the atom vocabulary size and mitigate the quantitative divergence between dominant (e.g., carbons) and rare atoms (e.g., phosphorus). With the enlarged atom \u2018vocabulary\u2019, we propose a novel node-level pre-training task, dubbed Masked Atoms Modeling (MAM), to mask some discrete codes randomly and then pre-train GNNs to predict them. MAM also mitigates another issue of AttrMask, namely the negative transfer. It can be easily combined with various pre-training tasks to improve their performance. Furthermore, we propose triplet masked contrastive learning (TMCL) for graph-level pre-training to model the heterogeneous semantic similarity between molecules for effective molecule retrieval. MAM and TMCL constitute a novel pre-training framework, Mole-BERT, which can match or outperform state-of-the-art methods in a fully data-driven manner. We release the code at https://github.com/junxia97/Mole-BERT."
                },
                {
                    "title": "Understanding the Limitations of Deep Models for Molecular property prediction: Insights and Solutions",
                    "abstract": "Molecular Property Prediction (MPP) is a crucial task in the AI-driven Drug Discovery (AIDD) pipeline, which has recently gained considerable attention thanks to advancements in deep learning. However, recent research has revealed that deep models struggle to beat traditional non-deep ones on MPP. In this study, we benchmark 12 representative models (3 non-deep models and 9 deep models) on 15 molecule datasets. Through the most comprehensive study to date, we make the following key observations: (i) Deep models are generally unable to outperform non-deep ones; (ii) The failure of deep models on MPP cannot be solely attributed to the small size of molecular datasets; (iii) In particular, some traditional models including XGB and RF that use molecular fingerprints as inputs tend to perform better than other competitors. Furthermore, we conduct extensive empirical investigations into the unique patterns of molecule data and inductive biases of various models underlying these phenomena. These findings stimulate us to develop a simple-yet-effective feature mapping method for molecule data prior to feeding them into deep models. Empirically, deep models equipped with this mapping method can beat non-deep ones in most MoleculeNet datasets. Notably, the effectiveness is further corroborated by extensive experiments on cutting-edge dataset related to COVID-19 and activity cliff datasets."
                },
                {
                    "title": "ErG: 2D Pharmacophore Descriptions for Scaffold Hopping",
                    "abstract": "An extended reduced graph approach (ErG) is presented that uses pharmacophore-type node descriptions to encode the relevant molecular properties. The basic idea of the method can be described as a hybrid approach of reduced graphs (Gillet et al. J. Chem. Inf. Comput. Sci. 2003, 43, 338-345) and binding property pairs (Kearsley et al. J. Chem. Inf. Comput. Sci. 1996, 36, 118-127). However, specific extension modifications to correctly describe the pharmacophoric properties, size, and shape of the molecules under study result in a very stable and good performance as compared to DAYLIGHT fingerprints (DFP). This is exemplified for 11 activity classes of the MDL Drug Data Report database, for which ErG performs as well or better than DFP in 10 cases. On the basis of the example data sets, the ability of ErG to switch from one chemotype to another (often referred to as \"scaffold hopping\") is highlighted. Additionally, possible pitfalls of reduced graph approaches as well as suitable solutions are discussed with the help of example structures. Overall, it is shown that ErG is a widely applicable method capable of identifying structurally diverse actives for a given active search query. This diversity is achieved by a high degree of molecular abstraction, which in turn results in a low dimensional descriptor vector that allows very low computation times for similarity searches."
                },
                {
                    "title": "An update of the DEF database of protein fold class predictions",
                    "abstract": "An update is given on the Database of Expected Fold classes (DEF) that contains a collection of fold-class predictions made from protein sequences and a mail server that provides new predictions for new sequences. To any given sequence one of 49 fold-classes is chosen to classify the structure related to the sequence with high accuracy. The updated prediction system is developed using data from the new version of the 3D-ALI database of aligned protein structures and thus is giving more reliable and more detailed predictions than the previous DEF system."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a flexible and comprehensive molecular relational learning (MRL) toolkit that effectively integrates both sequence and structural data for drug-target interactions, while accommodating advanced model architectures and interaction layers?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of existing MRL models, particularly in their ability to utilize diverse input types and complex architectures. By enhancing the capabilities of MRL, this research could lead to more accurate predictions of drug-target interactions, thereby accelerating therapeutic discovery and development. The advancements could also facilitate better understanding of protein-protein and drug-drug interactions, ultimately improving patient safety and treatment efficacy. This work could inspire future research to explore novel architectures and methodologies in MRL, leading to practical applications in drug design and personalized medicine.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to integrate various types of molecular data (both sequence and structure) while also accommodating complex model architectures. Naive approaches may fail due to the inherent complexity of molecular interactions and the vast model space created by different encoding methods and interaction layers. Additionally, technical obstacles include the need for efficient algorithms to process and learn from high-dimensional data, as well as the challenge of benchmarking diverse models effectively. Theoretical complexities arise from understanding the relationships between molecular features and their interactions, which are often non-linear and context-dependent.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the reliance on amino acid sequences as the primary input for protein data, due to a lack of accessible structural data. This has restricted the development of more sophisticated models that can leverage structural information. Additionally, existing MRL libraries, such as DeepPurpose, have rigid architectures that do not accommodate the latest advancements in model design. Barriers such as the complexity of integrating multiple data types and the need for flexible architectures have hindered progress. Our approach differs by introducing FlexMol, which is designed to support both sequence and structural data while allowing for customizable and advanced model architectures.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing FlexMol, a toolkit that enables the dynamic construction of MRL models with support for both protein sequences and structures. We will utilize a diverse dataset that includes both types of input data, and we will benchmark our models using metrics"
            }
        },
        "author_data": {
            "abd1cb7f-316a-4c8b-a373-cf85b1eda727": {
                "pk": "abd1cb7f-316a-4c8b-a373-cf85b1eda727",
                "name": "Sizhe Liu",
                "collaborators": [
                    "Shujian Huang",
                    "Wenhao Zhu",
                    "Jiajun Chen",
                    "Jingbo Zhou",
                    "Shaorong Chen",
                    "Jun Xia",
                    "Tianze Ling",
                    "Wenjie Du",
                    "Stan Z. Li",
                    "Kyle C. Smith"
                ],
                "domain": [
                    "Machine Learning",
                    "Natural Language Processing",
                    "Proteomics",
                    "Electrochemistry"
                ],
                "publications": [
                    {
                        "title": "Quantifying the Trade-offs between Energy Consumption and Salt Removal Rate in Membrane-free Cation Intercalation Desalination",
                        "abstract": "Electrochemical desalination devices that use redox-active cation intercalation electrodes show promise for desalination of salt-rich water resources with high water recovery and low energy consumption. While previous modeling and experiments used ion-exchange membranes to maximize charge efficiency, here a membrane-free alternative is evaluated to reduce capital cost by using a porous diaphragm to separate Na$_{1+x}$NiFe(CN)$_6$ electrodes. Two-dimensional porous-electrode modeling shows that, while charge efficiency losses are inherent to a diaphragm-based architecture, charge efficiency values approaching the anion transference number (61$\\%$ for NaCl) are achievable for diaphragms with sufficiently low salt conductance. Closed-form equations are thereby derived that relate charge efficiency to the non-dimensional P\\`{e}clet and Damk\\\"ohler numbers that enable the selection of current and flow velocity to produce a desired degree of desalination. Simulations using these conditions are used to quantify the tradeoffs between energy consumption and salt removal rate for diaphragm-based cells operated at a range of currents. The simulated distributions of reactions are shown to result from the local salt concentration variations within electrodes using diffusion-potential theory. We also simulate the cycling dynamics of various flow configurations and show that flow-through electrodes exceed the degree-of-desalination compared with flow-by and flow-behind configurations due to solution stagnation within electrodes."
                    },
                    {
                        "title": "kNN-BOX: A Unified Framework for Nearest Neighbor Generation",
                        "abstract": "Augmenting the base neural model with a token-level symbolic datastore is a novel generation paradigm and has achieved promising results in machine translation (MT). In this paper, we introduce a unified framework kNN-BOX, which enables quick development and interactive analysis for this novel paradigm. kNN-BOX decomposes the datastore-augmentation approach into three modules: datastore, retriever and combiner, thus putting diverse kNN generation methods into a unified way. Currently, kNN-BOX has provided implementation of seven popular kNN-MT variants, covering research from performance enhancement to efficiency optimization. It is easy for users to reproduce these existing works or customize their own models. Besides, users can interact with their kNN generation systems with kNN-BOX to better understand the underlying inference process in a visualized way. In the experiment section, we apply kNN-BOX for machine translation and three other seq2seq generation tasks, namely, text simplification, paraphrase generation and question generation. Experiment results show that augmenting the base neural model with kNN-BOX leads to a large performance improvement in all these tasks. The code and document of kNN-BOX is available at https://github.com/NJUNLP/knn-box."
                    },
                    {
                        "title": "Multilingual Contrastive Decoding via Language-Agnostic Layers Skipping",
                        "abstract": "Decoding by contrasting layers (DoLa), is designed to improve the generation quality of large language models (LLMs) by contrasting the prediction probabilities between an early exit output (amateur logits) and the final output (expert logits). However, we find that this approach does not work well on non-English tasks. Inspired by previous interpretability work on language transition during the model's forward pass, we discover that this issue arises from a language mismatch between early exit output and final output. In this work, we propose an improved contrastive decoding algorithm that is effective for diverse languages beyond English. To obtain more helpful amateur logits, we devise two strategies to skip a set of bottom, language-agnostic layers based on our preliminary analysis. Experimental results on multilingual reasoning benchmarks demonstrate that our proposed method outperforms previous contrastive decoding baselines and substantially improves LLM's chain-of-thought reasoning accuracy across 11 languages. The project will be available at: https://github.com/NJUNLP/SkipLayerCD."
                    },
                    {
                        "title": "NovoBench: Benchmarking Deep Learning-based De Novo Peptide Sequencing Methods in Proteomics",
                        "abstract": "Tandem mass spectrometry has played a pivotal role in advancing proteomics, enabling the high-throughput analysis of protein composition in biological tissues. Many deep learning methods have been developed for \\emph{de novo} peptide sequencing task, i.e., predicting the peptide sequence for the observed mass spectrum. However, two key challenges seriously hinder the further advancement of this important task. Firstly, since there is no consensus for the evaluation datasets, the empirical results in different research papers are often not comparable, leading to unfair comparison. Secondly, the current methods are usually limited to amino acid-level or peptide-level precision and recall metrics. In this work, we present the first unified benchmark NovoBench for \\emph{de novo} peptide sequencing, which comprises diverse mass spectrum data, integrated models, and comprehensive evaluation metrics. Recent impressive methods, including DeepNovo, PointNovo, Casanovo, InstaNovo, AdaNovo and $\\pi$-HelixNovo are integrated into our framework. In addition to amino acid-level and peptide-level precision and recall, we evaluate the models' performance in terms of identifying post-tranlational modifications (PTMs), efficiency and robustness to peptide length, noise peaks and missing fragment ratio, which are important influencing factors while seldom be considered. Leveraging this benchmark, we conduct a large-scale study of current methods, report many insightful findings that open up new possibilities for future development. The benchmark will be open-sourced to facilitate future research and application."
                    },
                    {
                        "title": "Large Language Models Are Cross-Lingual Knowledge-Free Reasoners",
                        "abstract": "Large Language Models have demonstrated impressive reasoning capabilities across multiple languages. However, the relationship between capabilities in different languages is less explored. In this work, we decompose the process of reasoning tasks into two separated components: knowledge retrieval and knowledge-free reasoning, and analyze the relationship between cross-lingual transferability and these two components. With adapted commonsense reasoning datasets and constructed knowledge-free reasoning datasets, we show that the knowledge-free reasoning capability can be nearly perfectly transferred across various source-target language directions despite the secondary impact of resource in some specific target languages, while cross-lingual knowledge retrieval significantly hinders the transfer. Moreover, by analyzing the hidden states and feed-forward network neuron activation during the reasoning, we show that higher similarity of hidden representations and larger overlap of activated neurons could explain the better cross-lingual transferability of knowledge-free reasoning than knowledge retrieval. Thus, we hypothesize that knowledge-free reasoning shares similar neurons in different languages for reasoning, while knowledge is stored separately in different languages. Our code and data is available at: https://github.com/NJUNLP/Knowledge-Free-Reasoning."
                    },
                    {
                        "title": "AdaNovo: Adaptive \\emph{De Novo} Peptide Sequencing with Conditional Mutual Information",
                        "abstract": "Tandem mass spectrometry has played a pivotal role in advancing proteomics, enabling the analysis of protein composition in biological samples. Despite the development of various deep learning methods for identifying amino acid sequences (peptides) responsible for observed spectra, challenges persist in \\emph{de novo} peptide sequencing. Firstly, prior methods struggle to identify amino acids with post-translational modifications (PTMs) due to their lower frequency in training data compared to canonical amino acids, further resulting in decreased peptide-level identification precision. Secondly, diverse types of noise and missing peaks in mass spectra reduce the reliability of training data (peptide-spectrum matches, PSMs). To address these challenges, we propose AdaNovo, a novel framework that calculates conditional mutual information (CMI) between the spectrum and each amino acid/peptide, using CMI for adaptive model training. Extensive experiments demonstrate AdaNovo's state-of-the-art performance on a 9-species benchmark, where the peptides in the training set are almost completely disjoint from the peptides of the test sets. Moreover, AdaNovo excels in identifying amino acids with PTMs and exhibits robustness against data noise. The supplementary materials contain the official code."
                    }
                ]
            },
            "8135783b-e1ce-41f7-846e-ae9b9c20b90f": {
                "pk": "8135783b-e1ce-41f7-846e-ae9b9c20b90f",
                "name": "Jun Xia",
                "collaborators": [
                    "Stan Z. Li",
                    "Yanqiao Zhu",
                    "Yuanqi Du",
                    "Wen-Gao Long",
                    "Yi Zhang",
                    "Yiyu Shi",
                    "Lecheng Zhang",
                    "Xiao Zhu"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Federated Learning",
                    "Natural Language Processing",
                    "Molecular Property Prediction"
                ],
                "publications": [
                    {
                        "title": "On the asymptotics of real solutions for the Painlev\u00e9 I equation",
                        "abstract": "In this paper, we revisit the asymptotic formulas of real Painlev\\'e I transcendents as the independent variable tends to negative infinity, which were initially derived by Kapaev with the complex WKB method. Using the Riemann-Hilbert method, we improve the error estimates of the oscillatory type asymptotics and provide precise error estimates of the singular type asymptotics. We also establish the corresponding asymptotics for the associated Hamiltonians of real Painlev\\'e I transcendents. In addition, two typos in the mentioned asymptotic behaviors in literature are corrected."
                    },
                    {
                        "title": "Towards Energy-Aware Federated Learning via MARL: A Dual-Selection Approach for Model and Client",
                        "abstract": "Although Federated Learning (FL) is promising in knowledge sharing for heterogeneous Artificial Intelligence of Thing (AIoT) devices, their training performance and energy efficacy are severely restricted in practical battery-driven scenarios due to the ``wooden barrel effect'' caused by the mismatch between homogeneous model paradigms and heterogeneous device capability. As a result, due to various kinds of differences among devices, it is hard for existing FL methods to conduct training effectively in energy-constrained scenarios, such as battery constraints of devices. To tackle the above issues, we propose an energy-aware FL framework named DR-FL, which considers the energy constraints in both clients and heterogeneous deep learning models to enable energy-efficient FL. Unlike Vanilla FL, DR-FL adopts our proposed Muti-Agents Reinforcement Learning (MARL)-based dual-selection method, which allows participated devices to make contributions to the global model effectively and adaptively based on their computing capabilities and energy capacities in a MARL-based manner. Experiments conducted with various widely recognized datasets demonstrate that DR-FL has the capability to optimize the exchange of knowledge among diverse models in large-scale AIoT systems while adhering to energy limitations. Additionally, it improves the performance of each individual heterogeneous device's model."
                    },
                    {
                        "title": "A Survey of Pretraining on Graphs: Taxonomy, Methods, and Applications",
                        "abstract": "Pretrained Language Models (PLMs) such as BERT have revolutionized the landscape of Natural Language Processing (NLP). Inspired by their proliferation, tremendous efforts have been devoted to Pretrained Graph Models (PGMs). Owing to the powerful model architectures of PGMs, abundant knowledge from massive labeled and unlabeled graph data can be captured. The knowledge implicitly encoded in model parameters can benefit various downstream tasks and help to alleviate several fundamental issues of learning on graphs. In this paper, we provide the first comprehensive survey for PGMs. We firstly present the limitations of graph representation learning and thus introduce the motivation for graph pre-training. Then, we systematically categorize existing PGMs based on a taxonomy from four different perspectives. Next, we present the applications of PGMs in social recommendation and drug discovery. Finally, we outline several promising research directions that can serve as a guideline for future research."
                    },
                    {
                        "title": "A Systematic Survey of Chemical Pre-trained Models",
                        "abstract": "Deep learning has achieved remarkable success in learning representations for molecules, which is crucial for various biochemical applications, ranging from property prediction to drug design. However, training Deep Neural Networks (DNNs) from scratch often requires abundant labeled molecules, which are expensive to acquire in the real world. To alleviate this issue, tremendous efforts have been devoted to Molecular Pre-trained Models (CPMs), where DNNs are pre-trained using large-scale unlabeled molecular databases and then fine-tuned over specific downstream tasks. Despite the prosperity, there lacks a systematic review of this fast-growing field. In this paper, we present the first survey that summarizes the current progress of CPMs. We first highlight the limitations of training molecular representation models from scratch to motivate CPM studies. Next, we systematically review recent advances on this topic from several key perspectives, including molecular descriptors, encoder architectures, pre-training strategies, and applications. We also highlight the challenges and promising avenues for future research, providing a useful resource for both machine learning and scientific communities."
                    },
                    {
                        "title": "Why Deep Models Often cannot Beat Non-deep Counterparts on Molecular Property Prediction?",
                        "abstract": "Molecular property prediction (MPP) is a crucial task in the drug discovery pipeline, which has recently gained considerable attention thanks to advances in deep neural networks. However, recent research has revealed that deep models struggle to beat traditional non-deep ones on MPP. In this study, we benchmark 12 representative models (3 non-deep models and 9 deep models) on 14 molecule datasets. Through the most comprehensive study to date, we make the following key observations: \\textbf{(\\romannumeral 1)} Deep models are generally unable to outperform non-deep ones; \\textbf{(\\romannumeral 2)} The failure of deep models on MPP cannot be solely attributed to the small size of molecular datasets. What matters is the irregular molecule data pattern; \\textbf{(\\romannumeral 3)} In particular, tree models using molecular fingerprints as inputs tend to perform better than other competitors. Furthermore, we conduct extensive empirical investigations into the unique patterns of molecule data and inductive biases of various models underlying these phenomena."
                    }
                ]
            },
            "08c30b3a-6ecd-4824-ab28-1397c9b02e0a": {
                "pk": "08c30b3a-6ecd-4824-ab28-1397c9b02e0a",
                "name": "Lecheng Zhang",
                "collaborators": [
                    "Jun Xia",
                    "Xiao Zhu",
                    "Stan Z. Li"
                ],
                "domain": [
                    "Molecular Property Prediction",
                    "Deep Learning",
                    "Drug Discovery",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Why Deep Models Often cannot Beat Non-deep Counterparts on Molecular Property Prediction?",
                        "abstract": "Molecular property prediction (MPP) is a crucial task in the drug discovery pipeline, which has recently gained considerable attention thanks to advances in deep neural networks. However, recent research has revealed that deep models struggle to beat traditional non-deep ones on MPP. In this study, we benchmark 12 representative models (3 non-deep models and 9 deep models) on 14 molecule datasets. Through the most comprehensive study to date, we make the following key observations: \\textbf{(\\romannumeral 1)} Deep models are generally unable to outperform non-deep ones; \\textbf{(\\romannumeral 2)} The failure of deep models on MPP cannot be solely attributed to the small size of molecular datasets. What matters is the irregular molecule data pattern; \\textbf{(\\romannumeral 3)} In particular, tree models using molecular fingerprints as inputs tend to perform better than other competitors. Furthermore, we conduct extensive empirical investigations into the unique patterns of molecule data and inductive biases of various models underlying these phenomena."
                    }
                ]
            },
            "472ff194-3b41-4c16-b886-ec61718e68de": {
                "pk": "472ff194-3b41-4c16-b886-ec61718e68de",
                "name": "Yuchen Liu",
                "collaborators": [
                    "Ziquan Zhuang",
                    "Chenyang Xu",
                    "Harold Blum",
                    "Ziwen Zhu",
                    "Yueqiao Wu",
                    "Zheng Zhu",
                    "Andrea Petracci",
                    "Junyan Zhao"
                ],
                "domain": [
                    "Algebraic Geometry",
                    "K-stability",
                    "Moduli Spaces",
                    "Fano Varieties"
                ],
                "publications": [
                    {
                        "title": "Semi-continuity of Complex Singularity Exponents in Singular Central Fiber Cases",
                        "abstract": "In this new version, we add the proof of the main theorem when the central fiber is not necessarily simple normal crossing. We also correct some typos."
                    },
                    {
                        "title": "K-stability of cubic fourfolds",
                        "abstract": "We prove that the K-moduli space of cubic fourfolds is identical to their GIT moduli space. More precisely, the K-(semi/poly)stability of cubic fourfolds coincide to the corresponding GIT stabilities, which was studied in detail by Laza. In particular, this implies that all smooth cubic fourfolds admit K\\\"ahler-Einstein metrics. Key ingredients are local volume estimates in dimension three due to Liu-Xu, and Ambro-Kawamata's non-vanishing theorem for Fano fourfolds."
                    },
                    {
                        "title": "Construction of hyperbolic Horikawa surfaces",
                        "abstract": "We construct a Brody hyperbolic Horikawa surface that is a double cover of $\\mathbb{P}^2$ branched along a smooth curve of degree $10$. We also construct Brody hyperbolic double covers of Hirzebruch surfaces with branch loci of the lowest possible bidegree."
                    },
                    {
                        "title": "K-stability of Fano threefolds of rank 2 and degree 14 as double covers",
                        "abstract": "We prove that every smooth Fano threefold from the family No 2.8 is K-stable. Such a Fano threefold is a double cover of the blow-up of $\\mathbb{P}^3$ at one point branched along an anti-canonical divisor."
                    },
                    {
                        "title": "Hyperbolicity of cyclic covers and complements",
                        "abstract": "We prove that a cyclic cover of a smooth complex projective variety is Brody hyperbolic if its branch divisor is a generic small deformation of a large enough multiple of a Brody hyperbolic base-point-free ample divisor. We also show the hyperbolicity of complements of those branch divisors. As an application, we find new examples of Brody hyperbolic hypersurfaces in $\\mathbb{P}^{n+1}$ that are cyclic covers of $\\mathbb{P}^n$."
                    },
                    {
                        "title": "On the semi-continuity problem of normalized volumes of singularities",
                        "abstract": "We show that in any $\\mathbb{Q}$-Gorenstein flat family of klt singularities, normalized volumes can only jump down at countably many subvarieties. A quick consequence is that smooth points have the largest normalized volume among all klt singularities. Using an alternative characterization of K-semistability developed by Li, Xu and the author, we show that K-semistability is a very generic or empty condition in any $\\mathbb{Q}$-Gorenstein flat family of log Fano pairs."
                    },
                    {
                        "title": "The volume of singular K\u00e4hler-Einstein Fano varieties",
                        "abstract": "We show that the anti-canonical volume of an $n$-dimensional K\\\"ahler-Einstein $\\mathbb{Q}$-Fano variety is bounded from above by certain invariants of the local singularities, namely $\\mathrm{lct}^n\\cdot\\mathrm{mult}$ for ideals and the normalized volume function for real valuations. This refines a recent result by Fujita. As an application, we get sharp volume upper bounds for K\\\"ahler-Einstein Fano varieties with quotient singularities. Based on very recent results by Li and the author, we show that a Fano manifold is K-semistable if and only if a de Fernex-Ein-Musta\\c{t}\\u{a} type inequality holds on its affine cone."
                    },
                    {
                        "title": "Court Judgement Labeling Using Topic Modeling and Syntactic Parsing",
                        "abstract": "In regions that practice common law, relevant historical cases are essential references for sentencing. To help legal practitioners find previous judgement easier, this paper aims to label each court judgement by some tags. These tags are legally important to summarize the judgement and can guide the user to similar judgements. We introduce a heuristic system to solve the problem, which starts from Aspect-driven Topic Modeling and uses Dependency Parsing and Constituency Parsing for phrase generation. We also construct a legal term tree for Hong Kong and implemented a sentence simplification module to support the system. Finally, we propose a similar document recommendation algorithm based on the generated tags. It enables users to find similar documents based on a few selected aspects rather than the whole passage. Experiment results show that this system is the best approach for this specific task. It is better than simple term extraction method in terms of summarizing the document, and the recommendation algorithm is more effective than full-text comparison approaches. We believe that the system has huge potential in law as well as in other areas."
                    },
                    {
                        "title": "A note on Koll\u00e1r valuations",
                        "abstract": "We prove the set of Koll\\'{a}r valuations in the dual complex of a klt singularity with a fixed complement is path connected. We also classify the case when the dual complex is one dimensional."
                    },
                    {
                        "title": "Equivariant K-stability under finite group action",
                        "abstract": "We show that G-equivariant K-semistability (resp. G-equivariant K-polystability) implies K-semistability (resp. K-polystability) for log Fano pairs when G is a finite group."
                    },
                    {
                        "title": "The normalized volume of a singularity is lower semicontinuous",
                        "abstract": "We show that in any $\\mathbb{Q}$-Gorenstein flat family of klt singularities, normalized volumes are lower semicontinuous with respect to the Zariski topology. A quick consequence is that smooth points have the largest normalized volume among all klt singularities. Using an alternative characterization of K-semistability developed by Li, Liu and Xu, we show that K-semistability is a very generic or empty condition in any $\\mathbb{Q}$-Gorenstein flat family of log Fano pairs."
                    },
                    {
                        "title": "Characterization of projective spaces by Seshadri constants",
                        "abstract": "We prove that an $n$-dimensional complex projective variety is isomorphic to $\\mathbb{P}^n$ if the Seshadri constant of the anti-canonical divisor at some smooth point is greater than $n$. We also classify complex projective varieties with Seshadri constants equal to $n$."
                    },
                    {
                        "title": "On the sharpness of Tian's criterion for K-stability",
                        "abstract": "Tian's criterion for K-stability states that a Fano variety of dimension $n$ whose alpha invariant is greater than $\\frac{n}{n+1}$ is K-stable. We show that this criterion is sharp by constructing singular Fano varieties with alpha invariants $\\frac{n}{n+1}$ that are not K-polystable for sufficiently large $n$. We also construct K-unstable Fano varieties with alpha invariants $\\frac{n-1}{n}$."
                    },
                    {
                        "title": "K-semistability of log Fano cone singularities",
                        "abstract": "We give a non-Archimedean characterization of K-semistability of log Fano cone singularities, and show that it agrees with the definition originally defined by Collins--Sz\\'ekelyhidi. As an application, we show that to test K-semistability, it suffices to test special test configurations. We also show that special test configurations give rise to lc places of torus equivariant bounded complements."
                    },
                    {
                        "title": "K-stability of cubic threefolds",
                        "abstract": "We prove the K-moduli space of cubic threefolds is identical to their GIT moduli. More precisely, the K-(semi,poly)-stability of cubic threefolds coincide to the corresponding GIT stabilities, which could be explicitly calculated. In particular, this implies that all smooth cubic threefolds admit K\\\"ahler-Einstein metric as well as provides a precise list of singular KE ones. To achieve this, our main new contribution is an estimate in dimension three of the volumes of kawamata log terminal singularities introduced by Chi Li. This is obtained via a detailed study of the classification of three dimensional canonical and terminal singularities, which was established during the study of the explicit three dimensional minimal model program."
                    },
                    {
                        "title": "Engineering Fractional Chern Insulators through Periodic Strain in Monolayer Graphene and Transition Metal Dichalcogenides",
                        "abstract": "We propose the realization of interaction-driven insulators in periodically strained monolayer graphene and transition metal dichalcogenides (TMDs). By analyzing the tunable band structure and band geometry via strain, and performing extensive many-body exact diagonalization of a realistic model, we present compelling evidence for realizing various fractional Chern insulators in both strained monolayer graphene and TMDs. Our thorough analysis across different strain parameters, accounting for experimental variability, reveals that a broad spectrum of fractional Chern insulators, including the Laughlin states, Halperin 112, 332 and 111 states, and Chern number |C| = 2 states, can be stabilized in distinct regions of the phase diagram. These findings suggest that periodically strained monolayer graphene and TMDs provide promising platforms for engineering fractional Chern insulators."
                    },
                    {
                        "title": "Openness of uniform K-stability in families of $\\mathbb{Q}$-Fano varieties",
                        "abstract": "We show that uniform K-stability is a Zariski open condition in Q-Gorenstein families of Q-Fano varieties. To prove this result, we consider the behavior of the stability threshold in families. The stability threshold (also known as the delta-invariant) is a recently introduced invariant that is known to detect the K-semistability and uniform K-stability of a Q-Fano variety. We show that the stability threshold is lower semicontinuous in families and provide an interpretation of the invariant in terms of the K-stability of log pairs."
                    },
                    {
                        "title": "Birational superrigidity and K-stability of singular Fano complete intersections",
                        "abstract": "We introduce an inductive argument for proving birational superrigidity and K-stability of singular Fano complete intersections of index one, using the same types of information from lower dimensions. In particular, we prove that a hypersurface in $\\mathbb{P}^{n+1}$ of degree $n+1$ with only ordinary singularities of multiplicity at most $n-5$ is birationally superrigid and K-stable if $n\\gg0$. As part of the argument, we also establish an adjunction type result for local volumes of singularities."
                    },
                    {
                        "title": "On K-stability of some del Pezzo surfaces of Fano index 2",
                        "abstract": "For every integer $a \\geq 2$, we relate the K-stability of hypersurfaces in the weighted projective space $\\mathbb{P}(1,1,a,a)$ of degree $2a$ with the GIT stability of binary forms of degree $2a$. Moreover, we prove that such a hypersurface is K-polystable and not K-stable if it is quasi-smooth."
                    },
                    {
                        "title": "K-moduli of Fano threefolds and genus four curves",
                        "abstract": "In this article, we study the K-moduli space of Fano threefolds obtained by blowing up $\\mathbb{P}^3$ along $(2,3)$-complete intersection curves. This K-moduli space is a two-step birational modification of the GIT moduli space of $(3,3)$-curves on $\\mathbb{P}^1 \\times \\mathbb{P}^1$. As an application, we show that our K-moduli space appears as one model of the Hassett--Keel program for $\\overline{M}_4$. In particular, we classify all K-(semi/poly)stable members in this deformation family of Fano varieties. We follow the moduli continuity method with moduli of lattice-polarized K3 surfaces, general elephants and Sarkisov links as new ingredients."
                    }
                ]
            },
            "a3258444-6ac7-4525-9635-6f073c2ea85c": {
                "pk": "a3258444-6ac7-4525-9635-6f073c2ea85c",
                "name": "Yue Liu",
                "collaborators": [
                    "Nicolas Privault",
                    "Yu Liu",
                    "Kang Yue",
                    "Changyu Ren",
                    "Jianqing Chen",
                    "Zhiwu Lin",
                    "Dahai He",
                    "Huiqun Jiang",
                    "Minghua Li"
                ],
                "domain": [
                    "Mathematical Modeling",
                    "Graph Theory",
                    "Immersive Learning",
                    "Stochastic Processes"
                ],
                "publications": [
                    {
                        "title": "On the Algebraic Representation of One-Tape Deterministic Turing Machine",
                        "abstract": "An algebraic representation of the Turing machines is given, where the configurations of Turing machines are represented by 4 order tensors, and the transition functions by 8 order tensors. Two types of tensor product are defined, one is to model the evolution of the Turing machines, and the other is to model the compositions of transition functions. It is shown that the two types of tensor product are harmonic in the sense that the associate law is obeyed."
                    },
                    {
                        "title": "Homomorphism of independent random variable convolution and matrix multiplication",
                        "abstract": "A map is given showing that convolutions of independent random variables over a finite group and matrix multiplications of doubly stochastic matrices are homomorphic. As an application, a short proof is given to the theorem that the limiting distributions of stochastic processes with stationary independent increments over a finite group are always uniform."
                    },
                    {
                        "title": "Behavioral analysis in immersive learning environments: A systematic literature review and research agenda",
                        "abstract": "The rapid growth of immersive technologies in educational areas has increased research interest in analyzing the specific behavioral patterns of learners in immersive learning environments. Considering the fact that research on the technical affordances of immersive technologies and the pedagogical affordances of behavioral analysis remains fragmented, this study first contributes by developing a conceptual framework that amalgamates learning requirements, specification, evaluation, and iteration into an integrated model to identify learning benefits and potential hurdles of behavioral analysis in immersive learning environments. Then, a systematic review was conducted underpinning the proposed conceptual framework to retrieve valuable empirical evidence from the 40 eligible articles during the last decade. The review findings suggest that (1) there is an essential need to sufficiently prepare the salient pedagogical requirements to define the specific learning stage, envisage intended cognitive objectives, and specify an appropriate set of learning activities, when developing comprehensive plans on behavioral analysis in immersive learning environments. (2) Researchers could customize the unique immersive experimental implementation by considering factors from four dimensions: learner, pedagogy, context, and representation. (3) The behavioral patterns constructed in immersive learning environments vary by considering the influence of behavioral analysis techniques, research themes, and immersive technical features. (4) The use of behavioral analysis in immersive learning environments faces several challenges from technical, implementation, and data processing perspectives. This study also articulates critical research agenda that could drive future investigation on behavioral analysis in immersive learning environments."
                    },
                    {
                        "title": "Pogorelov type $C^2$ estimates for Sum Hessian equations and a rigidity theorem",
                        "abstract": "We mainly study Pogorelov type $C^2$ estimates for solutions to the Dirichlet problem of Sum Hessian equations. We establish respectively Pogorelov type $C^2$ estimates for $k$-convex solutions and admissible solutions under some conditions. Furthermore, we apply such estimates to obtain a rigidity theorem for $k$-convex solutions of Sum Hessian equations in Euclidean space."
                    },
                    {
                        "title": "Instability of Standing Waves to the Inhomogeneous Nonlinear Schr\u00f6dinger Equation with Harmonic Potential",
                        "abstract": "We study the instability of standing-wave solutions $e^{i\\omega t}\\phi_{\\omega}(x)$ to the inhomogeneous nonlinear Schr\\\"{o}dinger equation $$i\\phi_t=-\\triangle\\phi+|x|^2\\phi-|x|^b|\\phi|^{p-1}\\phi, \\qquad \\in\\mathbb{R}^N, $$ where $ b > 0 $ and $ \\phi_{\\omega} $ is a ground-state solution. The results of the instability of standing-wave solutions reveal a balance between the frequency $\\omega $ of wave and the power of nonlinearity $p $ for any fixed $ b > 0. $"
                    },
                    {
                        "title": "Stability of peakons for the Degasperis-Procesi equation",
                        "abstract": "The Degasperis-Procesi equation can be derived as a member of a one-parameter family of asymptotic shallow water approximations to the Euler equations with the same asymptotic accuracy as that of the Camassa-Holm equation. In this paper, we study the orbital stability problem of the peaked solitons to the Degasperis-Procesi equation on the line. By constructing a Liapunov function, we prove that the shapes of these peakon solitons are stable under small perturbations."
                    },
                    {
                        "title": "Selling at the ultimate maximum in a regime switching model",
                        "abstract": "This paper deals with optimal prediction in a regime-switching model driven by a continuous-time Markov chain. We extend existing results for geometric Brownian motion by deriving optimal stopping strategies that depend on the current regime state, and prove a number of continuity properties relating to optimal value and boundary functions. Our approach replaces the use of closed form expressions, which are not available in our setting, with PDE arguments that also simplify the approach of [2] in the classical Brownian case."
                    },
                    {
                        "title": "A recursive algorithm for selling at the ultimate maximum in regime-switching models",
                        "abstract": "We propose a recursive algorithm for the numerical computation of the optimal value function $\\inf_{t\\le\\tau\\le T} E \\Big[\\sup_{0\\le s\\le T } Y_s / Y_{\\tau} \\big| {\\cal F}_t\\Big]$ over the stopping times $\\tau$ with respect to the filtration of a geometric Brownian motion $Y_t$ with Markovian regime switching. This method allows us to determine the boundary functions of the optimal stopping set when no associated Volterra integral equation is available. It applies in particular when regime-switching drifts have mixed signs, in which case the boundary functions may not be monotone."
                    },
                    {
                        "title": "Anomalous interfacial temperature profile induced by phonon localization",
                        "abstract": "Through the integration of the power spectral density, we obtain temperature profiles of both multi-segment harmonic and anharmonic systems, showing the presence of an anomalous negative temperature gradient inside the interfacial segment. Via investigating patterns of the power spectral density, we found that the counterintuitive phenomenon comes from the presence of interfacial localized phonon modes. Two out-band localized modes of the harmonic model, which make no contributions to local temperature due to the absence of phonon interactions, result in the concave temperature profile and over-cooling effect. For the anharmonic model, thanks to the phonon-phonon interactions, the localized modes are excited and make considerable contributions to interfacial temperature, which is clearly shown by examining the temperature accumulation function. When anharmonicity is considerably large, the negative temperature gradient is absent since the localized phonon modes are fully mixed. The presence of localized modes are evidently demonstrated by the inverse participation ratio and normal mode analysis for the isolated harmonic model."
                    },
                    {
                        "title": "Maximal signed bipartite graphs with totally disconnected graphs as star complements",
                        "abstract": "Let $\\dot{\\mathscr{B}}\\triangleq \\dot{\\mathscr{B}}_{\\dot{H}, \\mu}$ denote an arbitrary signed bipartite graph with $\\dot{H}$ as a star complement for an eigenvalue $\\mu$, where $\\dot{H}$ is a totally disconnected graph of order $s$. In this paper, by using Hadamard and Conference matrices as tools, the maximum order of $\\dot{\\mathscr{B}}$ and the extremal graphs are studied. It is shown that $\\dot{\\mathscr{B}}$ exists if and only if $\\mu^2$ is a positive integer. A formula of the maximum order of $\\dot{\\mathscr{B}}$ is given in the case of $\\mu^2=p\\times q$ such that $p$, $q$ are integers and there exists a $p$-order Hadamard or $(p+1)$-order Conference matrix. In particular, it is proved the maximum order of $\\dot{\\mathscr{B}}$ is $2s$ when either $q=1$, $s=c\\mu^2=cp$ or $q=1$, $s=c(\\mu^2+1)=c(p+1)$, $ c=1,2,3,\\cdots$. Futhermore, some extremal graphs are characterized."
                    },
                    {
                        "title": "Some Bounds on the Energy of Graphs with Self-Loops regarding $\u03bb_{1}$ and $\u03bb_{n}$",
                        "abstract": "Let $G_{S}$ be a graph with $n$ vertices obtained from a simple graph $G$ by attaching one self-loop at each vertex in $S \\subseteq V(G)$. The energy of $G_{S}$ is defined by Gutman et al. as $E(G_{S})=\\sum_{i=1}^{n}\\left| \\lambda_{i} -\\frac{\\sigma}{n} \\right|$, where $\\lambda_{1},\\dots,\\lambda_{n}$ are the adjacency eigenvalues of $G_{S}$ and $\\sigma$ is the number of self-loops of $G_{S}$. In this paper, several upper and lower bounds of $E(G_{S})$ regarding $\\lambda_{1}$ and $\\lambda_{n}$ are obtained. Especially, the upper bound $E(G_{S}) \\leq \\sqrt{n\\left(2m+\\sigma-\\frac{\\sigma^{2}}{n}\\right)}$ $(\\ast)$ given by Gutman et al. is improved to the following bound   \\begin{align*}   E(G_{S})\\leq \\sqrt{n\\left(2m+\\sigma-\\frac{\\sigma^{2}}{n}\\right)-\\frac{n}{2}\\left(\\left |\\lambda_{1}-\\frac{\\sigma}{n}\\right |-\\left |\\lambda_{n}-\\frac{\\sigma}{n}\\right |\\right)^{2}},   \\end{align*} where $\\left| \\lambda_{1}-\\frac{\\sigma}{n}\\right| \\geq \\dots \\geq \\left| \\lambda_{n}-\\frac{\\sigma}{n}\\right|$. Moreover, all graphs are characterized when the equality holds in Gutmans' bound $(\\ast)$ by using this new bound."
                    }
                ]
            },
            "5dd49d36-90fe-4f76-a6b5-01f736e43450": {
                "pk": "5dd49d36-90fe-4f76-a6b5-01f736e43450",
                "name": "Wenjie Du",
                "collaborators": [
                    "Yuxin Deng",
                    "David Cote",
                    "Yan Liu",
                    "Junfeng Sun",
                    "Yueling Yang",
                    "Huilan Ma",
                    "Daniel Gebler",
                    "Zezhong Chen",
                    "Chris Barber"
                ],
                "domain": [
                    "Time Series Analysis",
                    "Machine Learning",
                    "Probabilistic Processes",
                    "Software Engineering"
                ],
                "publications": [
                    {
                        "title": "PyPOTS: A Python Toolbox for Data Mining on Partially-Observed Time Series",
                        "abstract": "PyPOTS is an open-source Python library dedicated to data mining and analysis on multivariate partially-observed time series, i.e. incomplete time series with missing values, A.K.A. irregularlysampled time series. Particularly, it provides easy access to diverse algorithms categorized into four tasks: imputation, classification, clustering, and forecasting. The included models contain probabilistic approaches as well as neural-network methods, with a well-designed and fully-documented programming interface for both academic researchers and industrial professionals to use. With robustness and scalability in its design philosophy, best practices of software construction, for example, unit testing, continuous integration (CI) and continuous delivery (CD), code coverage, maintainability evaluation, interactive tutorials, and parallelization, are carried out as principles during the development of PyPOTS. The toolkit is available on both Python Package Index (PyPI) and Anaconda. PyPOTS is open-source and publicly available on GitHub https://github.com/WenjieDu/PyPOTS."
                    },
                    {
                        "title": "Logical, Metric, and Algorithmic Characterisations of Probabilistic Bisimulation",
                        "abstract": "Many behavioural equivalences or preorders for probabilistic processes involve a lifting operation that turns a relation on states into a relation on distributions of states. We show that several existing proposals for lifting relations can be reconciled to be different presentations of essentially the same lifting operation. More interestingly, this lifting operation nicely corresponds to the Kantorovich metric, a fundamental concept used in mathematics to lift a metric on states to a metric on distributions of states, besides the fact the lifting operation is related to the maximum flow problem in optimisation theory.   The lifting operation yields a neat notion of probabilistic bisimulation, for which we provide logical, metric, and algorithmic characterisations. Specifically, we extend the Hennessy-Milner logic and the modal mu-calculus with a new modality, resulting in an adequate and an expressive logic for probabilistic bisimilarity, respectively. The correspondence of the lifting operation and the Kantorovich metric leads to a natural characterisation of bisimulations as pseudometrics which are post-fixed points of a monotone function. We also present an \"on the fly\" algorithm to check if two states in a finitary system are related by probabilistic bisimilarity, exploiting the close relationship between the lifting operation and the maximum flow problem."
                    },
                    {
                        "title": "Study of Bc -> BP,BV Decays with QCD Factorization",
                        "abstract": "The Bc -> BP,BV decays are studied with the QCD factorization approach (where P and V denote pseudoscalar and vector mesons, respectively). Considering the contributions of both current-current and penguin operators, the amplitudes of branching ratios are estimated at the leading approximation. We find that the contributions of the penguin operators are very small due to the serious suppression by the CKM elements. The most promising decay modes are Bc -> Bs Pi, Bs rho, which might be easily detected at hadron colliders."
                    },
                    {
                        "title": "Modal Characterisations of Behavioural Pseudometrics",
                        "abstract": "For the model of probabilistic labelled transition systems that allow for the co-existence of nondeterminism and probabilities, we present two notions of bisimulation metrics: one is state-based and the other is distribution-based. We provide a sound and complete modal characterisation for each of them, using real-valued modal logics based on the Hennessy-Milner logic. The logic for characterising the state-based metric is much simpler than an earlier logic by Desharnais et al. as it uses only two non-expansive operators rather than the general class of non-expansive operators."
                    },
                    {
                        "title": "SAITS: Self-Attention-based Imputation for Time Series",
                        "abstract": "Missing data in time series is a pervasive problem that puts obstacles in the way of advanced analysis. A popular solution is imputation, where the fundamental challenge is to determine what values should be filled in. This paper proposes SAITS, a novel method based on the self-attention mechanism for missing value imputation in multivariate time series. Trained by a joint-optimization approach, SAITS learns missing values from a weighted combination of two diagonally-masked self-attention (DMSA) blocks. DMSA explicitly captures both the temporal dependencies and feature correlations between time steps, which improves imputation accuracy and training speed. Meanwhile, the weighted-combination design enables SAITS to dynamically assign weights to the learned representations from two DMSA blocks according to the attention map and the missingness information. Extensive experiments quantitatively and qualitatively demonstrate that SAITS outperforms the state-of-the-art methods on the time-series imputation task efficiently and reveal SAITS' potential to improve the learning performance of pattern recognition models on incomplete time-series data from the real world. The code is open source on GitHub at https://github.com/WenjieDu/SAITS."
                    },
                    {
                        "title": "Trusta: Reasoning about Assurance Cases with Formal Methods and Large Language Models",
                        "abstract": "Assurance cases can be used to argue for the safety of products in safety engineering. In safety-critical areas, the construction of assurance cases is indispensable. Trustworthiness Derivation Trees (TDTs) enhance assurance cases by incorporating formal methods, rendering it possible for automatic reasoning about assurance cases. We present Trustworthiness Derivation Tree Analyzer (Trusta), a desktop application designed to automatically construct and verify TDTs. The tool has a built-in Prolog interpreter in its backend, and is supported by the constraint solvers Z3 and MONA. Therefore, it can solve constraints about logical formulas involving arithmetic, sets, Horn clauses etc. Trusta also utilizes large language models to make the creation and evaluation of assurance cases more convenient. It allows for interactive human examination and modification. We evaluated top language models like ChatGPT-3.5, ChatGPT-4, and PaLM 2 for generating assurance cases. Our tests showed a 50%-80% similarity between machine-generated and human-created cases. In addition, Trusta can extract formal constraints from text in natural languages, facilitating an easier interpretation and validation process. This extraction is subject to human review and correction, blending the best of automated efficiency with human insight. To our knowledge, this marks the first integration of large language models in automatic creating and reasoning about assurance cases, bringing a novel approach to a traditional challenge. Through several industrial case studies, Trusta has proven to quickly find some subtle issues that are typically missed in manual inspection, demonstrating its practical value in enhancing the assurance case development process."
                    },
                    {
                        "title": "Forecasting Loss of Signal in Optical Networks with Machine Learning",
                        "abstract": "Loss of Signal (LOS) represents a significant cost for operators of optical networks. By studying large sets of real-world Performance Monitoring (PM) data collected from six international optical networks, we find that it is possible to forecast LOS events with good precision 1-7 days before they occur, albeit at relatively low recall, with supervised machine learning (ML). Our study covers twelve facility types, including 100G lines and ETH10G clients. We show that the precision for a given network improves when training on multiple networks simultaneously relative to training on an individual network. Furthermore, we show that it is possible to forecast LOS from all facility types and all networks with a single model, whereas fine-tuning for a particular facility or network only brings modest improvements. Hence our ML models remain effective for optical networks previously unknown to the model, which makes them usable for commercial applications."
                    }
                ]
            },
            "ae96b20c-e545-4475-ad6e-f168bded2719": {
                "pk": "ae96b20c-e545-4475-ad6e-f168bded2719",
                "name": "Zhangyang Gao",
                "collaborators": [
                    "Cheng Tan",
                    "Stan Z. Li",
                    "Lirong Wu",
                    "Haitao Lin",
                    "Stan. Z Li",
                    "Pablo Chac\u00f3n",
                    "Yuqi Hu",
                    "Xingran Chen",
                    "Siyuan Li",
                    "Yijie Zhang"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Protein Design",
                    "Machine Learning",
                    "Clustering"
                ],
                "publications": [
                    {
                        "title": "LookHops: light multi-order convolution and pooling for graph classification",
                        "abstract": "Convolution and pooling are the key operations to learn hierarchical representation for graph classification, where more expressive $k$-order($k>1$) method requires more computation cost, limiting the further applications. In this paper, we investigate the strategy of selecting $k$ via neighborhood information gain and propose light $k$-order convolution and pooling requiring fewer parameters while improving the performance. Comprehensive and fair experiments through six graph classification benchmarks show: 1) the performance improvement is consistent to the $k$-order information gain. 2) the proposed convolution requires fewer parameters while providing competitive results. 3) the proposed pooling outperforms SOTA algorithms in terms of efficiency and performance."
                    },
                    {
                        "title": "AlphaDesign: A graph protein design method and benchmark on AlphaFoldDB",
                        "abstract": "While DeepMind has tentatively solved protein folding, its inverse problem -- protein design which predicts protein sequences from their 3D structures -- still faces significant challenges. Particularly, the lack of large-scale standardized benchmark and poor accuray hinder the research progress. In order to standardize comparisons and draw more research interest, we use AlphaFold DB, one of the world's largest protein structure databases, to establish a new graph-based benchmark -- AlphaDesign. Based on AlphaDesign, we propose a new method called ADesign to improve accuracy by introducing protein angles as new features, using a simplified graph transformer encoder (SGT), and proposing a confidence-aware protein decoder (CPD). Meanwhile, SGT and CPD also improve model efficiency by simplifying the training and testing procedures. Experiments show that ADesign significantly outperforms previous graph models, e.g., the average accuracy is improved by 8\\%, and the inference speed is 40+ times faster than before."
                    },
                    {
                        "title": "Knowledge-Design: Pushing the Limit of Protein Design via Knowledge Refinement",
                        "abstract": "Recent studies have shown competitive performance in protein design that aims to find the amino acid sequence folding into the desired structure. However, most of them disregard the importance of predictive confidence, fail to cover the vast protein space, and do not incorporate common protein knowledge. After witnessing the great success of pretrained models on diverse protein-related tasks and the fact that recovery is highly correlated with confidence, we wonder whether this knowledge can push the limits of protein design further. As a solution, we propose a knowledge-aware module that refines low-quality residues. We also introduce a memory-retrieval mechanism to save more than 50\\% of the training time. We extensively evaluate our proposed method on the CATH, TS50, and TS500 datasets and our results show that our Knowledge-Design method outperforms the previous PiFold method by approximately 9\\% on the CATH dataset. Specifically, Knowledge-Design is the first method that achieves 60+\\% recovery on CATH, TS50 and TS500 benchmarks. We also provide additional analysis to demonstrate the effectiveness of our proposed method. The code will be publicly available."
                    },
                    {
                        "title": "VQPL: Vector Quantized Protein Language",
                        "abstract": "Is there a foreign language describing protein sequences and structures simultaneously? Protein structures, represented by continuous 3D points, have long posed a challenge due to the contrasting modeling paradigms of discrete sequences. To represent protein sequence-structure as discrete symbols, we propose a VQProteinformer to project residue types and structures into a discrete space, supervised by a reconstruction loss to ensure information preservation. The sequential latent codes of residues introduce a new quantized protein language, transforming the protein sequence-structure into a unified modality. We demonstrate the potential of the created protein language on predictive and generative tasks, which may not only advance protein research but also establish a connection between the protein-related and NLP-related fields. The proposed method will be continually improved to unify more protein modalities, including text and point cloud."
                    },
                    {
                        "title": "FoldToken2: Learning compact, invariant and generative protein structure language",
                        "abstract": "The equivalent nature of 3D coordinates has posed long term challenges in protein structure representation learning, alignment, and generation. Can we create a compact and invariant language that equivalently represents protein structures? Towards this goal, we propose FoldToken2 to transfer equivariant structures into discrete tokens, while maintaining the recoverability of the original structures. From FoldToken1 to FoldToken2, we improve three key components: (1) invariant structure encoder, (2) vector-quantized compressor, and (3) equivalent structure decoder. We evaluate FoldToken2 on the protein structure reconstruction task and show that it outperforms previous FoldToken1 by 20\\% in TMScore and 81\\% in RMSD. FoldToken2 probably be the first method that works well on both single-chain and multi-chain protein structures quantization. We believe that FoldToken2 will inspire further improvement in protein structure representation learning, structure alignment, and structure generation tasks."
                    },
                    {
                        "title": "DiffSDS: A language diffusion model for protein backbone inpainting under geometric conditions and constraints",
                        "abstract": "Have you ever been troubled by the complexity and computational cost of SE(3) protein structure modeling and been amazed by the simplicity and power of language modeling? Recent work has shown promise in simplifying protein structures as sequences of protein angles; therefore, language models could be used for unconstrained protein backbone generation. Unfortunately, such simplification is unsuitable for the constrained protein inpainting problem, where the model needs to recover masked structures conditioned on unmasked ones, as it dramatically increases the computing cost of geometric constraints. To overcome this dilemma, we suggest inserting a hidden \\textbf{a}tomic \\textbf{d}irection \\textbf{s}pace (\\textbf{ADS}) upon the language model, converting invariant backbone angles into equivalent direction vectors and preserving the simplicity, called Seq2Direct encoder ($\\text{Enc}_{s2d}$). Geometric constraints could be efficiently imposed on the newly introduced direction space. A Direct2Seq decoder ($\\text{Dec}_{d2s}$) with mathematical guarantees is also introduced to develop a \\textbf{SDS} ($\\text{Enc}_{s2d}$+$\\text{Dec}_{d2s}$) model. We apply the SDS model as the denoising neural network during the conditional diffusion process, resulting in a constrained generative model--\\textbf{DiffSDS}. Extensive experiments show that the plug-and-play ADS could transform the language model into a strong structural model without loss of simplicity. More importantly, the proposed DiffSDS outperforms previous strong baselines by a large margin on the task of protein inpainting."
                    },
                    {
                        "title": "Clustering Based on Graph of Density Topology",
                        "abstract": "Data clustering with uneven distribution in high level noise is challenging. Currently, HDBSCAN is considered as the SOTA algorithm for this problem. In this paper, we propose a novel clustering algorithm based on what we call graph of density topology (GDT). GDT jointly considers the local and global structures of data samples: firstly forming local clusters based on a density growing process with a strategy for properly noise handling as well as cluster boundary detection; and then estimating a GDT from relationship between local clusters in terms of a connectivity measure, givingglobal topological graph. The connectivity, measuring similarity between neighboring local clusters, is based on local clusters rather than individual points, ensuring its robustness to even very large noise. Evaluation results on both toy and real-world datasets show that GDT achieves the SOTA performance by far on almost all the popular datasets, and has a low time complexity of O(nlogn). The code is available at https://github.com/gaozhangyang/DGC.git."
                    },
                    {
                        "title": "Target-aware Molecular Graph Generation",
                        "abstract": "Generating molecules with desired biological activities has attracted growing attention in drug discovery. Previous molecular generation models are designed as chemocentric methods that hardly consider the drug-target interaction, limiting their practical applications. In this paper, we aim to generate molecular drugs in a target-aware manner that bridges biological activity and molecular design. To solve this problem, we compile a benchmark dataset from several publicly available datasets and build baselines in a unified framework. Building on the recent advantages of flow-based molecular generation models, we propose SiamFlow, which forces the flow to fit the distribution of target sequence embeddings in latent space. Specifically, we employ an alignment loss and a uniform loss to bring target sequence embeddings and drug graph embeddings into agreements while avoiding collapse. Furthermore, we formulate the alignment into a one-to-many problem by learning spaces of target sequence embeddings. Experiments quantitatively show that our proposed method learns meaningful representations in the latent space toward the target-aware molecular graph generation and provides an alternative approach to bridge biology and chemistry in drug discovery."
                    },
                    {
                        "title": "SemiRetro: Semi-template framework boosts deep retrosynthesis prediction",
                        "abstract": "Recently, template-based (TB) and template-free (TF) molecule graph learning methods have shown promising results to retrosynthesis. TB methods are more accurate using pre-encoded reaction templates, and TF methods are more scalable by decomposing retrosynthesis into subproblems, i.e., center identification and synthon completion. To combine both advantages of TB and TF, we suggest breaking a full-template into several semi-templates and embedding them into the two-step TF framework. Since many semi-templates are reduplicative, the template redundancy can be reduced while the essential chemical knowledge is still preserved to facilitate synthon completion. We call our method SemiRetro, introduce a new GNN layer (DRGAT) to enhance center identification, and propose a novel self-correcting module to improve semi-template classification. Experimental results show that SemiRetro significantly outperforms both existing TB and TF methods. In scalability, SemiRetro covers 98.9\\% data using 150 semi-templates, while previous template-based GLN requires 11,647 templates to cover 93.3\\% data. In top-1 accuracy, SemiRetro exceeds template-free G2G 4.8\\% (class known) and 6.0\\% (class unknown). Besides, SemiRetro has better training efficiency than existing methods."
                    },
                    {
                        "title": "SimVP: Simpler yet Better Video Prediction",
                        "abstract": "From CNN, RNN, to ViT, we have witnessed remarkable advancements in video prediction, incorporating auxiliary inputs, elaborate neural architectures, and sophisticated training strategies. We admire these progresses but are confused about the necessity: is there a simple method that can perform comparably well? This paper proposes SimVP, a simple video prediction model that is completely built upon CNN and trained by MSE loss in an end-to-end fashion. Without introducing any additional tricks and complicated strategies, we can achieve state-of-the-art performance on five benchmark datasets. Through extended experiments, we demonstrate that SimVP has strong generalization and extensibility on real-world datasets. The significant reduction of training cost makes it easier to scale to complex scenarios. We believe SimVP can serve as a solid baseline to stimulate the further development of video prediction. The code is available at \\href{https://github.com/gaozhangyang/SimVP-Simpler-yet-Better-Video-Prediction}{Github}."
                    },
                    {
                        "title": "CoSP: Co-supervised pretraining of pocket and ligand",
                        "abstract": "Can we inject the pocket-ligand interaction knowledge into the pre-trained model and jointly learn their chemical space? Pretraining molecules and proteins has attracted considerable attention in recent years, while most of these approaches focus on learning one of the chemical spaces and lack the injection of biological knowledge. We propose a co-supervised pretraining (CoSP) framework to simultaneously learn 3D pocket and ligand representations. We use a gated geometric message passing layer to model both 3D pockets and ligands, where each node's chemical features, geometric position and orientation are considered. To learn biological meaningful embeddings, we inject the pocket-ligand interaction knowledge into the pretraining model via contrastive loss. Considering the specificity of molecules, we further propose a chemical similarity-enhanced negative sampling strategy to improve the contrastive learning performance. Through extensive experiments, we conclude that CoSP can achieve competitive results in pocket matching, molecule property predictions, and virtual screening."
                    },
                    {
                        "title": "PiFold: Toward effective and efficient protein inverse folding",
                        "abstract": "How can we design protein sequences folding into the desired structures effectively and efficiently? AI methods for structure-based protein design have attracted increasing attention in recent years; however, few methods can simultaneously improve the accuracy and efficiency due to the lack of expressive features and autoregressive sequence decoder. To address these issues, we propose PiFold, which contains a novel residue featurizer and PiGNN layers to generate protein sequences in a one-shot way with improved recovery. Experiments show that PiFold could achieve 51.66\\% recovery on CATH 4.2, while the inference speed is 70 times faster than the autoregressive competitors. In addition, PiFold achieves 58.72\\% and 60.42\\% recovery scores on TS50 and TS500, respectively. We conduct comprehensive ablation studies to reveal the role of different types of protein features and model designs, inspiring further simplification and improvement. The PyTorch code is available at \\href{https://github.com/A4Bio/PiFold}{GitHub}."
                    },
                    {
                        "title": "PrefixMol: Target- and Chemistry-aware Molecule Design via Prefix Embedding",
                        "abstract": "Is there a unified model for generating molecules considering different conditions, such as binding pockets and chemical properties? Although target-aware generative models have made significant advances in drug design, they do not consider chemistry conditions and cannot guarantee the desired chemical properties. Unfortunately, merging the target-aware and chemical-aware models into a unified model to meet customized requirements may lead to the problem of negative transfer. Inspired by the success of multi-task learning in the NLP area, we use prefix embeddings to provide a novel generative model that considers both the targeted pocket's circumstances and a variety of chemical properties. All conditional information is represented as learnable features, which the generative model subsequently employs as a contextual prompt. Experiments show that our model exhibits good controllability in both single and multi-conditional molecular generation. The controllability enables us to outperform previous structure-based drug design methods. More interestingly, we open up the attention mechanism and reveal coupling relationships between conditions, providing guidance for multi-conditional molecule generation."
                    },
                    {
                        "title": "MotifRetro: Exploring the Combinability-Consistency Trade-offs in retrosynthesis via Dynamic Motif Editing",
                        "abstract": "Is there a unified framework for graph-based retrosynthesis prediction? Through analysis of full-, semi-, and non-template retrosynthesis methods, we discovered that they strive to strike an optimal balance between combinability and consistency: \\textit{Should atoms be combined as motifs to simplify the molecular editing process, or should motifs be broken down into atoms to reduce the vocabulary and improve predictive consistency?}   Recent works have studied several specific cases, while none of them explores different combinability-consistency trade-offs. Therefore, we propose MotifRetro, a dynamic motif editing framework for retrosynthesis prediction that can explore the entire trade-off space and unify graph-based models. MotifRetro comprises two components: RetroBPE, which controls the combinability-consistency trade-off, and a motif editing model, where we introduce a novel LG-EGAT module to dynamiclly add motifs to the molecule. We conduct extensive experiments on USPTO-50K to explore how the trade-off affects the model performance and finally achieve state-of-the-art performance."
                    },
                    {
                        "title": "SimVP: Towards Simple yet Powerful Spatiotemporal Predictive Learning",
                        "abstract": "Recent years have witnessed remarkable advances in spatiotemporal predictive learning, incorporating auxiliary inputs, elaborate neural architectures, and sophisticated training strategies. Although impressive, the system complexity of mainstream methods is increasing as well, which may hinder the convenient applications. This paper proposes SimVP, a simple spatiotemporal predictive baseline model that is completely built upon convolutional networks without recurrent architectures and trained by common mean squared error loss in an end-to-end fashion. Without introducing any extra tricks and strategies, SimVP can achieve superior performance on various benchmark datasets. To further improve the performance, we derive variants with the gated spatiotemporal attention translator from SimVP that can achieve better performance. We demonstrate that SimVP has strong generalization and extensibility on real-world datasets through extensive experiments. The significant reduction in training cost makes it easier to scale to complex scenarios. We believe SimVP can serve as a solid baseline to benefit the spatiotemporal predictive learning community."
                    },
                    {
                        "title": "Efficiently Predicting Protein Stability Changes Upon Single-point Mutation with Large Language Models",
                        "abstract": "Predicting protein stability changes induced by single-point mutations has been a persistent challenge over the years, attracting immense interest from numerous researchers. The ability to precisely predict protein thermostability is pivotal for various subfields and applications in biochemistry, including drug development, protein evolution analysis, and enzyme synthesis. Despite the proposition of multiple methodologies aimed at addressing this issue, few approaches have successfully achieved optimal performance coupled with high computational efficiency. Two principal hurdles contribute to the existing challenges in this domain. The first is the complexity of extracting and aggregating sufficiently representative features from proteins. The second refers to the limited availability of experimental data for protein mutation analysis, further complicating the comprehensive evaluation of model performance on unseen data samples. With the advent of Large Language Models(LLM), such as the ESM models in protein research, profound interpretation of protein features is now accessibly aided by enormous training data. Therefore, LLMs are indeed to facilitate a wide range of protein research. In our study, we introduce an ESM-assisted efficient approach that integrates protein sequence and structural features to predict the thermostability changes in protein upon single-point mutations. Furthermore, we have curated a dataset meticulously designed to preclude data leakage, corresponding to two extensively employed test datasets, to facilitate a more equitable model comparison."
                    },
                    {
                        "title": "Git: Clustering Based on Graph of Intensity Topology",
                        "abstract": "\\textbf{A}ccuracy, \\textbf{R}obustness to noises and scales, \\textbf{I}nterpretability, \\textbf{S}peed, and \\textbf{E}asy to use (ARISE) are crucial requirements of a good clustering algorithm. However, achieving these goals simultaneously is challenging, and most advanced approaches only focus on parts of them. Towards an overall consideration of these aspects, we propose a novel clustering algorithm, namely GIT (Clustering Based on \\textbf{G}raph of \\textbf{I}ntensity \\textbf{T}opology). GIT considers both local and global data structures: firstly forming local clusters based on intensity peaks of samples, and then estimating the global topological graph (topo-graph) between these local clusters. We use the Wasserstein Distance between the predicted and prior class proportions to automatically cut noisy edges in the topo-graph and merge connected local clusters as final clusters. Then, we compare GIT with seven competing algorithms on five synthetic datasets and nine real-world datasets. With fast local cluster detection, robust topo-graph construction and accurate edge-cutting, GIT shows attractive ARISE performance and significantly exceeds other non-convex clustering methods. For example, GIT outperforms its counterparts about $10\\%$ (F1-score) on MNIST and FashionMNIST. Code is available at \\color{red}{https://github.com/gaozhangyang/GIT}."
                    },
                    {
                        "title": "FoldToken: Learning Protein Language via Vector Quantization and Beyond",
                        "abstract": "Is there a foreign language describing protein sequences and structures simultaneously? Protein structures, represented by continuous 3D points, have long posed a challenge due to the contrasting modeling paradigms of discrete sequences. We introduce \\textbf{FoldTokenizer} to represent protein sequence-structure as discrete symbols. This innovative approach involves projecting residue types and structures into a discrete space, guided by a reconstruction loss for information preservation. We refer to the learned discrete symbols as \\textbf{FoldToken}, and the sequence of FoldTokens serves as a new protein language, transforming the protein sequence-structure into a unified modality. We apply the created protein language on general backbone inpainting and antibody design tasks, building the first GPT-style model (\\textbf{FoldGPT}) for sequence-structure co-generation with promising results. Key to our success is the substantial enhancement of the vector quantization module, Soft Conditional Vector Quantization (\\textbf{SoftCVQ})."
                    },
                    {
                        "title": "A Graph is Worth $K$ Words: Euclideanizing Graph using Pure Transformer",
                        "abstract": "Can we model Non-Euclidean graphs as pure language or even Euclidean vectors while retaining their inherent information? The Non-Euclidean property have posed a long term challenge in graph modeling. Despite recent graph neural networks and graph transformers efforts encoding graphs as Euclidean vectors, recovering the original graph from vectors remains a challenge. In this paper, we introduce GraphsGPT, featuring an Graph2Seq encoder that transforms Non-Euclidean graphs into learnable Graph Words in the Euclidean space, along with a GraphGPT decoder that reconstructs the original graph from Graph Words to ensure information equivalence. We pretrain GraphsGPT on $100$M molecules and yield some interesting findings: (1) The pretrained Graph2Seq excels in graph representation learning, achieving state-of-the-art results on $8/9$ graph classification and regression tasks. (2) The pretrained GraphGPT serves as a strong graph generator, demonstrated by its strong ability to perform both few-shot and conditional graph generation. (3) Graph2Seq+GraphGPT enables effective graph mixup in the Euclidean space, overcoming previously known Non-Euclidean challenges. (4) The edge-centric pretraining framework GraphsGPT demonstrates its efficacy in graph domain tasks, excelling in both representation and generation. Code is available at \\href{https://github.com/A4Bio/GraphsGPT}{GitHub}."
                    }
                ]
            },
            "5ea76fdf-ed9a-45a3-ac78-d0e876e5eaa2": {
                "pk": "5ea76fdf-ed9a-45a3-ac78-d0e876e5eaa2",
                "name": "Bozhen Hu",
                "collaborators": [
                    "Stan Z. Li",
                    "Cheng Tan",
                    "Jun Xia",
                    "Zhangyang Gao",
                    "Lirong Wu",
                    "Jiangbin Zheng",
                    "Zicheng Liu",
                    "Zelin Zang",
                    "Yufei Huang",
                    "Bin Gao"
                ],
                "domain": [
                    "Protein Representation Learning",
                    "Graph Neural Network",
                    "Deep Learning",
                    "Drug Discovery"
                ],
                "publications": [
                    {
                        "title": "Deep Manifold Transformation for Protein Representation Learning",
                        "abstract": "Protein representation learning is critical in various tasks in biology, such as drug design and protein structure or function prediction, which has primarily benefited from protein language models and graph neural networks. These models can capture intrinsic patterns from protein sequences and structures through masking and task-related losses. However, the learned protein representations are usually not well optimized, leading to performance degradation due to limited data, difficulty adapting to new tasks, etc. To address this, we propose a new \\underline{d}eep \\underline{m}anifold \\underline{t}ransformation approach for universal \\underline{p}rotein \\underline{r}epresentation \\underline{l}earning (DMTPRL). It employs manifold learning strategies to improve the quality and adaptability of the learned embeddings. Specifically, we apply a novel manifold learning loss during training based on the graph inter-node similarity. Our proposed DMTPRL method outperforms state-of-the-art baselines on diverse downstream tasks across popular datasets. This validates our approach for learning universal and robust protein representations. We promise to release the code after acceptance."
                    },
                    {
                        "title": "Segment Anything in Defect Detection",
                        "abstract": "Defect detection plays a crucial role in infrared non-destructive testing systems, offering non-contact, safe, and efficient inspection capabilities. However, challenges such as low resolution, high noise, and uneven heating in infrared thermal images hinder comprehensive and accurate defect detection. In this study, we propose DefectSAM, a novel approach for segmenting defects on highly noisy thermal images based on the widely adopted model, Segment Anything (SAM)\\cite{kirillov2023segany}. Harnessing the power of a meticulously curated dataset generated through labor-intensive lab experiments and valuable prompts from experienced experts, DefectSAM surpasses existing state-of-the-art segmentation algorithms and achieves significant improvements in defect detection rates. Notably, DefectSAM excels in detecting weaker and smaller defects on complex and irregular surfaces, reducing the occurrence of missed detections and providing more accurate defect size estimations. Experimental studies conducted on various materials have validated the effectiveness of our solutions in defect detection, which hold significant potential to expedite the evolution of defect detection tools, enabling enhanced inspection capabilities and accuracy in defect identification."
                    },
                    {
                        "title": "Deep Manifold Graph Auto-Encoder for Attributed Graph Embedding",
                        "abstract": "Representing graph data in a low-dimensional space for subsequent tasks is the purpose of attributed graph embedding. Most existing neural network approaches learn latent representations by minimizing reconstruction errors. Rare work considers the data distribution and the topological structure of latent codes simultaneously, which often results in inferior embeddings in real-world graph data. This paper proposes a novel Deep Manifold (Variational) Graph Auto-Encoder (DMVGAE/DMGAE) method for attributed graph data to improve the stability and quality of learned representations to tackle the crowding problem. The node-to-node geodesic similarity is preserved between the original and latent space under a pre-defined distribution. The proposed method surpasses state-of-the-art baseline algorithms by a significant margin on different downstream tasks across popular datasets, which validates our solutions. We promise to release the code after acceptance."
                    },
                    {
                        "title": "Generative De Novo Protein Design with Global Context",
                        "abstract": "The linear sequence of amino acids determines protein structure and function. Protein design, known as the inverse of protein structure prediction, aims to obtain a novel protein sequence that will fold into the defined structure. Recent works on computational protein design have studied designing sequences for the desired backbone structure with local positional information and achieved competitive performance. However, similar local environments in different backbone structures may result in different amino acids, indicating that protein structure's global context matters. Thus, we propose the Global-Context Aware generative de novo protein design method (GCA), consisting of local and global modules. While local modules focus on relationships between neighbor amino acids, global modules explicitly capture non-local contexts. Experimental results demonstrate that the proposed GCA method outperforms state-of-the-arts on de novo protein design. Our code and pretrained model will be released."
                    },
                    {
                        "title": "A Graph is Worth $K$ Words: Euclideanizing Graph using Pure Transformer",
                        "abstract": "Can we model Non-Euclidean graphs as pure language or even Euclidean vectors while retaining their inherent information? The Non-Euclidean property have posed a long term challenge in graph modeling. Despite recent graph neural networks and graph transformers efforts encoding graphs as Euclidean vectors, recovering the original graph from vectors remains a challenge. In this paper, we introduce GraphsGPT, featuring an Graph2Seq encoder that transforms Non-Euclidean graphs into learnable Graph Words in the Euclidean space, along with a GraphGPT decoder that reconstructs the original graph from Graph Words to ensure information equivalence. We pretrain GraphsGPT on $100$M molecules and yield some interesting findings: (1) The pretrained Graph2Seq excels in graph representation learning, achieving state-of-the-art results on $8/9$ graph classification and regression tasks. (2) The pretrained GraphGPT serves as a strong graph generator, demonstrated by its strong ability to perform both few-shot and conditional graph generation. (3) Graph2Seq+GraphGPT enables effective graph mixup in the Euclidean space, overcoming previously known Non-Euclidean challenges. (4) The edge-centric pretraining framework GraphsGPT demonstrates its efficacy in graph domain tasks, excelling in both representation and generation. Code is available at \\href{https://github.com/A4Bio/GraphsGPT}{GitHub}."
                    },
                    {
                        "title": "SimGRACE: A Simple Framework for Graph Contrastive Learning without Data Augmentation",
                        "abstract": "Graph contrastive learning (GCL) has emerged as a dominant technique for graph representation learning which maximizes the mutual information between paired graph augmentations that share the same semantics. Unfortunately, it is difficult to preserve semantics well during augmentations in view of the diverse nature of graph data. Currently, data augmentations in GCL that are designed to preserve semantics broadly fall into three unsatisfactory ways. First, the augmentations can be manually picked per dataset by trial-and-errors. Second, the augmentations can be selected via cumbersome search. Third, the augmentations can be obtained by introducing expensive domain-specific knowledge as guidance. All of these limit the efficiency and more general applicability of existing GCL methods. To circumvent these crucial issues, we propose a \\underline{Sim}ple framework for \\underline{GRA}ph \\underline{C}ontrastive l\\underline{E}arning, \\textbf{SimGRACE} for brevity, which does not require data augmentations. Specifically, we take original graph as input and GNN model with its perturbed version as two encoders to obtain two correlated views for contrast. SimGRACE is inspired by the observation that graph data can preserve their semantics well during encoder perturbations while not requiring manual trial-and-errors, cumbersome search or expensive domain knowledge for augmentations selection. Also, we explain why SimGRACE can succeed. Furthermore, we devise adversarial training scheme, dubbed \\textbf{AT-SimGRACE}, to enhance the robustness of graph contrastive learning and theoretically explain the reasons. Albeit simple, we show that SimGRACE can yield competitive or better performance compared with state-of-the-art methods in terms of generalizability, transferability and robustness, while enjoying unprecedented degree of flexibility and efficiency."
                    },
                    {
                        "title": "Protein Language Models and Structure Prediction: Connection and Progression",
                        "abstract": "The prediction of protein structures from sequences is an important task for function prediction, drug design, and related biological processes understanding. Recent advances have proved the power of language models (LMs) in processing the protein sequence databases, which inherit the advantages of attention networks and capture useful information in learning representations for proteins. The past two years have witnessed remarkable success in tertiary protein structure prediction (PSP), including evolution-based and single-sequence-based PSP. It seems that instead of using energy-based models and sampling procedures, protein language model (pLM)-based pipelines have emerged as mainstream paradigms in PSP. Despite the fruitful progress, the PSP community needs a systematic and up-to-date survey to help bridge the gap between LMs in the natural language processing (NLP) and PSP domains and introduce their methodologies, advancements and practical applications. To this end, in this paper, we first introduce the similarities between protein and human languages that allow LMs extended to pLMs, and applied to protein databases. Then, we systematically review recent advances in LMs and pLMs from the perspectives of network architectures, pre-training strategies, applications, and commonly-used protein databases. Next, different types of methods for PSP are discussed, particularly how the pLM-based architectures function in the process of protein folding. Finally, we identify challenges faced by the PSP community and foresee promising research directions along with the advances of pLMs. This survey aims to be a hands-on guide for researchers to understand PSP methods, develop pLMs and tackle challenging problems in this field for practical purposes."
                    },
                    {
                        "title": "RDesign: Hierarchical Data-efficient Representation Learning for Tertiary Structure-based RNA Design",
                        "abstract": "While artificial intelligence has made remarkable strides in revealing the relationship between biological macromolecules' primary sequence and tertiary structure, designing RNA sequences based on specified tertiary structures remains challenging. Though existing approaches in protein design have thoroughly explored structure-to-sequence dependencies in proteins, RNA design still confronts difficulties due to structural complexity and data scarcity. Moreover, direct transplantation of protein design methodologies into RNA design fails to achieve satisfactory outcomes although sharing similar structural components. In this study, we aim to systematically construct a data-driven RNA design pipeline. We crafted a large, well-curated benchmark dataset and designed a comprehensive structural modeling approach to represent the complex RNA tertiary structure. More importantly, we proposed a hierarchical data-efficient representation learning framework that learns structural representations through contrastive learning at both cluster-level and sample-level to fully leverage the limited data. By constraining data representations within a limited hyperspherical space, the intrinsic relationships between data points could be explicitly imposed. Moreover, we incorporated extracted secondary structures with base pairs as prior knowledge to facilitate the RNA design process. Extensive experiments demonstrate the effectiveness of our proposed method, providing a reliable baseline for future RNA design tasks. The source code and benchmark dataset are available at https://github.com/A4Bio/RDesign."
                    },
                    {
                        "title": "PSC-CPI: Multi-Scale Protein Sequence-Structure Contrasting for Efficient and Generalizable Compound-Protein Interaction Prediction",
                        "abstract": "Compound-Protein Interaction (CPI) prediction aims to predict the pattern and strength of compound-protein interactions for rational drug discovery. Existing deep learning-based methods utilize only the single modality of protein sequences or structures and lack the co-modeling of the joint distribution of the two modalities, which may lead to significant performance drops in complex real-world scenarios due to various factors, e.g., modality missing and domain shifting. More importantly, these methods only model protein sequences and structures at a single fixed scale, neglecting more fine-grained multi-scale information, such as those embedded in key protein fragments. In this paper, we propose a novel multi-scale Protein Sequence-structure Contrasting framework for CPI prediction (PSC-CPI), which captures the dependencies between protein sequences and structures through both intra-modality and cross-modality contrasting. We further apply length-variable protein augmentation to allow contrasting to be performed at different scales, from the amino acid level to the sequence level. Finally, in order to more fairly evaluate the model generalizability, we split the test data into four settings based on whether compounds and proteins have been observed during the training stage. Extensive experiments have shown that PSC-CPI generalizes well in all four settings, particularly in the more challenging ``Unseen-Both\" setting, where neither compounds nor proteins have been observed during training. Furthermore, even when encountering a situation of modality missing, i.e., inference with only single-modality protein data, PSC-CPI still exhibits comparable or even better performance than previous approaches."
                    },
                    {
                        "title": "Cross-Gate MLP with Protein Complex Invariant Embedding is A One-Shot Antibody Designer",
                        "abstract": "Antibodies are crucial proteins produced by the immune system in response to foreign substances or antigens. The specificity of an antibody is determined by its complementarity-determining regions (CDRs), which are located in the variable domains of the antibody chains and form the antigen-binding site. Previous studies have utilized complex techniques to generate CDRs, but they suffer from inadequate geometric modeling. Moreover, the common iterative refinement strategies lead to an inefficient inference. In this paper, we propose a \\textit{simple yet effective} model that can co-design 1D sequences and 3D structures of CDRs in a one-shot manner. To achieve this, we decouple the antibody CDR design problem into two stages: (i) geometric modeling of protein complex structures and (ii) sequence-structure co-learning. We develop a novel macromolecular structure invariant embedding, typically for protein complexes, that captures both intra- and inter-component interactions among the backbone atoms, including C$\\alpha$, N, C, and O atoms, to achieve comprehensive geometric modeling. Then, we introduce a simple cross-gate MLP for sequence-structure co-learning, allowing sequence and structure representations to implicitly refine each other. This enables our model to design desired sequences and structures in a one-shot manner. Extensive experiments are conducted to evaluate our results at both the sequence and structure levels, which demonstrate that our model achieves superior performance compared to the state-of-the-art antibody CDR design methods."
                    },
                    {
                        "title": "Advances of Deep Learning in Protein Science: A Comprehensive Survey",
                        "abstract": "Protein representation learning plays a crucial role in understanding the structure and function of proteins, which are essential biomolecules involved in various biological processes. In recent years, deep learning has emerged as a powerful tool for protein modeling due to its ability to learn complex patterns and representations from large-scale protein data. This comprehensive survey aims to provide an overview of the recent advances in deep learning techniques applied to protein science. The survey begins by introducing the developments of deep learning based protein models and emphasizes the importance of protein representation learning in drug discovery, protein engineering, and function annotation. It then delves into the fundamentals of deep learning, including convolutional neural networks, recurrent neural networks, attention models, and graph neural networks in modeling protein sequences, structures, and functions, and explores how these techniques can be used to extract meaningful features and capture intricate relationships within protein data. Next, the survey presents various applications of deep learning in the field of proteins, including protein structure prediction, protein-protein interaction prediction, protein function prediction, etc. Furthermore, it highlights the challenges and limitations of these deep learning techniques and also discusses potential solutions and future directions for overcoming these challenges. This comprehensive survey provides a valuable resource for researchers and practitioners in the field of proteins who are interested in harnessing the power of deep learning techniques. By consolidating the latest advancements and discussing potential avenues for improvement, this review contributes to the ongoing progress in protein research and paves the way for future breakthroughs in the field."
                    }
                ]
            },
            "a5f75ab4-9aee-49b1-bfc0-241bdd383c58": {
                "pk": "a5f75ab4-9aee-49b1-bfc0-241bdd383c58",
                "name": "Cheng Tan",
                "collaborators": [
                    "Wing Shing Wong",
                    "Andrew Yuan",
                    "Alina Oprea",
                    "Huanshui Zhang",
                    "Zhangyang Gao",
                    "Stan Z. Li"
                ],
                "domain": [
                    "Deep Learning",
                    "Adversarial Attacks",
                    "Drug Discovery",
                    "Online Optimization"
                ],
                "publications": [
                    {
                        "title": "Dropout Attacks",
                        "abstract": "Dropout is a common operator in deep learning, aiming to prevent overfitting by randomly dropping neurons during training. This paper introduces a new family of poisoning attacks against neural networks named DROPOUTATTACK. DROPOUTATTACK attacks the dropout operator by manipulating the selection of neurons to drop instead of selecting them uniformly at random. We design, implement, and evaluate four DROPOUTATTACK variants that cover a broad range of scenarios. These attacks can slow or stop training, destroy prediction accuracy of target classes, and sabotage either precision or recall of a target class. In our experiments of training a VGG-16 model on CIFAR-100, our attack can reduce the precision of the victim class by 34.6% (from 81.7% to 47.1%) without incurring any degradation in model accuracy"
                    },
                    {
                        "title": "Learning Based Control Policy and Regret Analysis for Online Quadratic Optimization with Asymmetric Information Structure",
                        "abstract": "In this paper, we propose a learning approach to analyze dynamic systems with asymmetric information structure. Instead of adopting a game theoretic setting, we investigate an online quadratic optimization problem driven by system noises with unknown statistics. Due to information asymmetry, it is infeasible to use classic Kalman filter nor optimal control strategies for such systems. It is necessary and beneficial to develop a robust approach that learns the probability statistics as time goes forward. Motivated by online convex optimization (OCO) theory, we introduce the notion of regret, which is defined as the cumulative performance loss difference between the optimal offline known statistics cost and the optimal online unknown statistics cost. By utilizing dynamic programming and linear minimum mean square biased estimate (LMMSUE), we propose a new type of online state feedback control policies and characterize the behavior of regret in finite time regime. The regret is shown to be sub-linear and bounded by O(ln T). Moreover, we address an online optimization problem with output feedback control policies."
                    },
                    {
                        "title": "Integrated Stabilization Policy over a Software Defined Network",
                        "abstract": "In this paper, we mainly investigate an integrated system operating under a software defined network (SDN) protocol. SDN is a new networking paradigm in which network intelligence is centrally administered and data is communicated via channels that are physically separated from those conveying user data. Under the SDN architecture, it is feasible to set up multiple flows for transmitting control signals to an actuator with high priority for each individual application. While each flow may suffer random transmission delay, we focus on the stabilization problem under the joint design of the event-driven strategy in actuator and the control policy in decision-maker. By introducing a predefined application time, the integrated system can be reformulated as the form of stochastic system with input delay and multiplicative noise. For such system, we propose a set of necessary and sufficient stabilization conditions. Specifically, for the scalar system, we derive the allowable sampling period bound that can guarantee stabilization in terms of the probability distributions of the random transmission delays. A simple example is included to show the performance of our theoretic results."
                    },
                    {
                        "title": "Target-aware Molecular Graph Generation",
                        "abstract": "Generating molecules with desired biological activities has attracted growing attention in drug discovery. Previous molecular generation models are designed as chemocentric methods that hardly consider the drug-target interaction, limiting their practical applications. In this paper, we aim to generate molecular drugs in a target-aware manner that bridges biological activity and molecular design. To solve this problem, we compile a benchmark dataset from several publicly available datasets and build baselines in a unified framework. Building on the recent advantages of flow-based molecular generation models, we propose SiamFlow, which forces the flow to fit the distribution of target sequence embeddings in latent space. Specifically, we employ an alignment loss and a uniform loss to bring target sequence embeddings and drug graph embeddings into agreements while avoiding collapse. Furthermore, we formulate the alignment into a one-to-many problem by learning spaces of target sequence embeddings. Experiments quantitatively show that our proposed method learns meaningful representations in the latent space toward the target-aware molecular graph generation and provides an alternative approach to bridge biology and chemistry in drug discovery."
                    }
                ]
            },
            "dbdc0dcb-fd60-4be0-b37c-cb5c0acb7a0d": {
                "pk": "dbdc0dcb-fd60-4be0-b37c-cb5c0acb7a0d",
                "name": "Hongxin Xiang",
                "collaborators": [
                    "Cheng Xie",
                    "Ting Zeng",
                    "Yun Yang",
                    "Qing Liu",
                    "Keqin Li",
                    "Beibei Yu",
                    "Zhixiang Cheng",
                    "Pengsen Ma",
                    "Li Zeng",
                    "Xin Jin"
                ],
                "domain": [
                    "Zero-Shot Learning",
                    "Generative Models",
                    "Molecular Representation Learning",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Multi-Knowledge Fusion for New Feature Generation in Generalized Zero-Shot Learning",
                        "abstract": "Suffering from the semantic insufficiency and domain-shift problems, most of existing state-of-the-art methods fail to achieve satisfactory results for Zero-Shot Learning (ZSL). In order to alleviate these problems, we propose a novel generative ZSL method to learn more generalized features from multi-knowledge with continuously generated new semantics in semantic-to-visual embedding. In our approach, the proposed Multi-Knowledge Fusion Network (MKFNet) takes different semantic features from multi-knowledge as input, which enables more relevant semantic features to be trained for semantic-to-visual embedding, and finally generates more generalized visual features by adaptively fusing visual features from different knowledge domain. The proposed New Feature Generator (NFG) with adaptive genetic strategy is used to enrich semantic information on the one hand, and on the other hand it greatly improves the intersection of visual feature generated by MKFNet and unseen visual faetures. Empirically, we show that our approach can achieve significantly better performance compared to existing state-of-the-art methods on a large number of benchmarks for several ZSL tasks, including traditional ZSL, generalized ZSL and zero-shot retrieval."
                    },
                    {
                        "title": "Class Knowledge Overlay to Visual Feature Learning for Zero-Shot Image Classification",
                        "abstract": "New categories can be discovered by transforming semantic features into synthesized visual features without corresponding training samples in zero-shot image classification. Although significant progress has been made in generating high-quality synthesized visual features using generative adversarial networks, guaranteeing semantic consistency between the semantic features and visual features remains very challenging. In this paper, we propose a novel zero-shot learning approach, GAN-CST, based on class knowledge to visual feature learning to tackle the problem. The approach consists of three parts, class knowledge overlay, semi-supervised learning and triplet loss. It applies class knowledge overlay (CKO) to obtain knowledge not only from the corresponding class but also from other classes that have the knowledge overlay. It ensures that the knowledge-to-visual learning process has adequate information to generate synthesized visual features. The approach also applies a semi-supervised learning process to re-train knowledge-to-visual model. It contributes to reinforcing synthesized visual features generation as well as new category prediction. We tabulate results on a number of benchmark datasets demonstrating that the proposed model delivers superior performance over state-of-the-art approaches."
                    },
                    {
                        "title": "Cross Knowledge-based Generative Zero-Shot Learning Approach with Taxonomy Regularization",
                        "abstract": "Although zero-shot learning (ZSL) has an inferential capability of recognizing new classes that have never been seen before, it always faces two fundamental challenges of the cross modality and crossdomain challenges. In order to alleviate these problems, we develop a generative network-based ZSL approach equipped with the proposed Cross Knowledge Learning (CKL) scheme and Taxonomy Regularization (TR). In our approach, the semantic features are taken as inputs, and the output is the synthesized visual features generated from the corresponding semantic features. CKL enables more relevant semantic features to be trained for semantic-to-visual feature embedding in ZSL, while Taxonomy Regularization (TR) significantly improves the intersections with unseen images with more generalized visual features generated from generative network. Extensive experiments on several benchmark datasets (i.e., AwA1, AwA2, CUB, NAB and aPY) show that our approach is superior to these state-of-the-art methods in terms of ZSL image classification and retrieval."
                    },
                    {
                        "title": "MaskMol: Knowledge-guided Molecular Image Pre-Training Framework for Activity Cliffs",
                        "abstract": "Activity cliffs, which refer to pairs of molecules that are structurally similar but show significant differences in their potency, can lead to model representation collapse and make the model challenging to distinguish them. Our research indicates that as molecular similarity increases, graph-based methods struggle to capture these nuances, whereas image-based approaches effectively retain the distinctions. Thus, we developed MaskMol, a knowledge-guided molecular image self-supervised learning framework. MaskMol accurately learns the representation of molecular images by considering multiple levels of molecular knowledge, such as atoms, bonds, and substructures. By utilizing pixel masking tasks, MaskMol extracts fine-grained information from molecular images, overcoming the limitations of existing deep learning models in identifying subtle structural changes. Experimental results demonstrate MaskMol's high accuracy and transferability in activity cliff estimation and compound potency prediction across 20 different macromolecular targets, outperforming 25 state-of-the-art deep learning and machine learning approaches. Visualization analyses reveal MaskMol's high biological interpretability in identifying activity cliff-relevant molecular substructures. Notably, through MaskMol, we identified candidate EP4 inhibitors that could be used to treat tumors. This study not only raises awareness about activity cliffs but also introduces a novel method for molecular image representation learning and virtual screening, advancing drug discovery and providing new insights into structure-activity relationships (SAR)."
                    }
                ]
            },
            "47fe8e96-9ddd-4e2d-ba2e-ec2397c7e5f4": {
                "pk": "47fe8e96-9ddd-4e2d-ba2e-ec2397c7e5f4",
                "name": "Stan Z. Li",
                "collaborators": [
                    "Zhangyang Gao",
                    "Cheng Tan",
                    "Zelin Zang",
                    "Fang Wu",
                    "Lirong Wu",
                    "Jiangbin Zheng",
                    "Jianwei Yang",
                    "Zhen Lei",
                    "Dragomir Radev",
                    "Di Wu"
                ],
                "domain": [
                    "Deep Learning",
                    "Manifold Learning",
                    "Protein Design",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Markov-Lipschitz Deep Learning",
                        "abstract": "We propose a novel framework, called Markov-Lipschitz deep learning (MLDL), to tackle geometric deterioration caused by collapse, twisting, or crossing in vector-based neural network transformations for manifold-based representation learning and manifold data generation. A prior constraint, called locally isometric smoothness (LIS), is imposed across-layers and encoded into a Markov random field (MRF)-Gibbs distribution. This leads to the best possible solutions for local geometry preservation and robustness as measured by locally geometric distortion and locally bi-Lipschitz continuity. Consequently, the layer-wise vector transformations are enhanced into well-behaved, LIS-constrained metric homeomorphisms. Extensive experiments, comparisons, and ablation study demonstrate significant advantages of MLDL for manifold learning and manifold data generation. MLDL is general enough to enhance any vector transformation-based networks. The code is available at https://github.com/westlake-cairi/Markov-Lipschitz-Deep-Learning."
                    },
                    {
                        "title": "Consistent Representation Learning for High Dimensional Data Analysis",
                        "abstract": "High dimensional data analysis for exploration and discovery includes three fundamental tasks: dimensionality reduction, clustering, and visualization. When the three associated tasks are done separately, as is often the case thus far, inconsistencies can occur among the tasks in terms of data geometry and others. This can lead to confusing or misleading data interpretation. In this paper, we propose a novel neural network-based method, called Consistent Representation Learning (CRL), to accomplish the three associated tasks end-to-end and improve the consistencies. The CRL network consists of two nonlinear dimensionality reduction (NLDR) transformations: (1) one from the input data space to the latent feature space for clustering, and (2) the other from the clustering space to the final 2D or 3D space for visualization. Importantly, the two NLDR transformations are performed to best satisfy local geometry preserving (LGP) constraints across the spaces or network layers, to improve data consistencies along with the processing flow. Also, we propose a novel metric, clustering-visualization inconsistency (CVI), for evaluating the inconsistencies. Extensive comparative results show that the proposed CRL neural network method outperforms the popular t-SNE and UMAP-based and other contemporary clustering and visualization algorithms in terms of evaluation metrics and visualization."
                    },
                    {
                        "title": "Deep Manifold Transformation for Nonlinear Dimensionality Reduction",
                        "abstract": "Manifold learning-based encoders have been playing important roles in nonlinear dimensionality reduction (NLDR) for data exploration. However, existing methods can often fail to preserve geometric, topological and/or distributional structures of data. In this paper, we propose a deep manifold learning framework, called deep manifold transformation (DMT) for unsupervised NLDR and embedding learning. DMT enhances deep neural networks by using cross-layer local geometry-preserving (LGP) constraints. The LGP constraints constitute the loss for deep manifold learning and serve as geometric regularizers for NLDR network training. Extensive experiments on synthetic and real-world data demonstrate that DMT networks outperform existing leading manifold-based NLDR methods in terms of preserving the structures of data."
                    },
                    {
                        "title": "InsertGNN: Can Graph Neural Networks Outperform Humans in TOEFL Sentence Insertion Problem?",
                        "abstract": "Sentence insertion is an interesting NLP problem but received insufficient attention. Existing approaches in sentence ordering, text coherence, and question answering are neither suitable nor good enough at solving it. To bridge this gap, we propose InsertGNN, a simple yet effective model that represents the problem as a graph and adopts a hierarchical graph neural network (GNN) to learn the connection between sentences. We evaluate our method in our newly collected TOEFL dataset and further verify its effectiveness on the larger arXiv dataset using cross-domain learning. Extensive experiments demonstrate that InsertGNN outperforms all baselines by a large margin with an accuracy of 70\\%, rivaling the average human test scores."
                    },
                    {
                        "title": "Learn Convolutional Neural Network for Face Anti-Spoofing",
                        "abstract": "Though having achieved some progresses, the hand-crafted texture features, e.g., LBP [23], LBP-TOP [11] are still unable to capture the most discriminative cues between genuine and fake faces. In this paper, instead of designing feature by ourselves, we rely on the deep convolutional neural network (CNN) to learn features of high discriminative ability in a supervised manner. Combined with some data pre-processing, the face anti-spoofing performance improves drastically. In the experiments, over 70% relative decrease of Half Total Error Rate (HTER) is achieved on two challenging datasets, CASIA [36] and REPLAY-ATTACK [7] compared with the state-of-the-art. Meanwhile, the experimental results from inter-tests between two datasets indicates CNN can obtain features with better generalization ability. Moreover, the nets trained using combined data from two datasets have less biases between two datasets."
                    },
                    {
                        "title": "AlphaDesign: A graph protein design method and benchmark on AlphaFoldDB",
                        "abstract": "While DeepMind has tentatively solved protein folding, its inverse problem -- protein design which predicts protein sequences from their 3D structures -- still faces significant challenges. Particularly, the lack of large-scale standardized benchmark and poor accuray hinder the research progress. In order to standardize comparisons and draw more research interest, we use AlphaFold DB, one of the world's largest protein structure databases, to establish a new graph-based benchmark -- AlphaDesign. Based on AlphaDesign, we propose a new method called ADesign to improve accuracy by introducing protein angles as new features, using a simplified graph transformer encoder (SGT), and proposing a confidence-aware protein decoder (CPD). Meanwhile, SGT and CPD also improve model efficiency by simplifying the training and testing procedures. Experiments show that ADesign significantly outperforms previous graph models, e.g., the average accuracy is improved by 8\\%, and the inference speed is 40+ times faster than before."
                    },
                    {
                        "title": "Target-aware Molecular Graph Generation",
                        "abstract": "Generating molecules with desired biological activities has attracted growing attention in drug discovery. Previous molecular generation models are designed as chemocentric methods that hardly consider the drug-target interaction, limiting their practical applications. In this paper, we aim to generate molecular drugs in a target-aware manner that bridges biological activity and molecular design. To solve this problem, we compile a benchmark dataset from several publicly available datasets and build baselines in a unified framework. Building on the recent advantages of flow-based molecular generation models, we propose SiamFlow, which forces the flow to fit the distribution of target sequence embeddings in latent space. Specifically, we employ an alignment loss and a uniform loss to bring target sequence embeddings and drug graph embeddings into agreements while avoiding collapse. Furthermore, we formulate the alignment into a one-to-many problem by learning spaces of target sequence embeddings. Experiments quantitatively show that our proposed method learns meaningful representations in the latent space toward the target-aware molecular graph generation and provides an alternative approach to bridge biology and chemistry in drug discovery."
                    },
                    {
                        "title": "CCPL: Cross-modal Contrastive Protein Learning",
                        "abstract": "Effective protein representation learning is crucial for predicting protein functions. Traditional methods often pretrain protein language models on large, unlabeled amino acid sequences, followed by finetuning on labeled data. While effective, these methods underutilize the potential of protein structures, which are vital for function determination. Common structural representation techniques rely heavily on annotated data, limiting their generalizability. Moreover, structural pretraining methods, similar to natural language pretraining, can distort actual protein structures. In this work, we introduce a novel unsupervised protein structure representation pretraining method, cross-modal contrastive protein learning (CCPL). CCPL leverages a robust protein language model and uses unsupervised contrastive alignment to enhance structure learning, incorporating self-supervised structural constraints to maintain intrinsic structural information. We evaluated our model across various benchmarks, demonstrating the framework's superiority."
                    },
                    {
                        "title": "Knowledge-Design: Pushing the Limit of Protein Design via Knowledge Refinement",
                        "abstract": "Recent studies have shown competitive performance in protein design that aims to find the amino acid sequence folding into the desired structure. However, most of them disregard the importance of predictive confidence, fail to cover the vast protein space, and do not incorporate common protein knowledge. After witnessing the great success of pretrained models on diverse protein-related tasks and the fact that recovery is highly correlated with confidence, we wonder whether this knowledge can push the limits of protein design further. As a solution, we propose a knowledge-aware module that refines low-quality residues. We also introduce a memory-retrieval mechanism to save more than 50\\% of the training time. We extensively evaluate our proposed method on the CATH, TS50, and TS500 datasets and our results show that our Knowledge-Design method outperforms the previous PiFold method by approximately 9\\% on the CATH dataset. Specifically, Knowledge-Design is the first method that achieves 60+\\% recovery on CATH, TS50 and TS500 benchmarks. We also provide additional analysis to demonstrate the effectiveness of our proposed method. The code will be publicly available."
                    },
                    {
                        "title": "VQPL: Vector Quantized Protein Language",
                        "abstract": "Is there a foreign language describing protein sequences and structures simultaneously? Protein structures, represented by continuous 3D points, have long posed a challenge due to the contrasting modeling paradigms of discrete sequences. To represent protein sequence-structure as discrete symbols, we propose a VQProteinformer to project residue types and structures into a discrete space, supervised by a reconstruction loss to ensure information preservation. The sequential latent codes of residues introduce a new quantized protein language, transforming the protein sequence-structure into a unified modality. We demonstrate the potential of the created protein language on predictive and generative tasks, which may not only advance protein research but also establish a connection between the protein-related and NLP-related fields. The proposed method will be continually improved to unify more protein modalities, including text and point cloud."
                    },
                    {
                        "title": "A Hierarchical Training Paradigm for Antibody Structure-sequence Co-design",
                        "abstract": "Therapeutic antibodies are an essential and rapidly expanding drug modality. The binding specificity between antibodies and antigens is decided by complementarity-determining regions (CDRs) at the tips of these Y-shaped proteins. In this paper, we propose a hierarchical training paradigm (HTP) for the antibody sequence-structure co-design. HTP consists of four levels of training stages, each corresponding to a specific protein modality within a particular protein domain. Through carefully crafted tasks in different stages, HTP seamlessly and effectively integrates geometric graph neural networks (GNNs) with large-scale protein language models to excavate evolutionary information from not only geometric structures but also vast antibody and non-antibody sequence databases, which determines ligand binding pose and strength. Empirical experiments show that HTP sets the new state-of-the-art performance in the co-design problem as well as the fix-backbone design. Our research offers a hopeful path to unleash the potential of deep generative architectures and seeks to illuminate the way forward for the antibody sequence and structure co-design challenge."
                    },
                    {
                        "title": "Progressive Multi-Modality Learning for Inverse Protein Folding",
                        "abstract": "While deep generative models show promise for learning inverse protein folding directly from data, the lack of publicly available structure-sequence pairings limits their generalization. Previous improvements and data augmentation efforts to overcome this bottleneck have been insufficient. To further address this challenge, we propose a novel protein design paradigm called MMDesign, which leverages multi-modality transfer learning. To our knowledge, MMDesign is the first framework that combines a pretrained structural module with a pretrained contextual module, using an auto-encoder (AE) based language model to incorporate prior protein semantic knowledge. Experimental results, only training with the small dataset, demonstrate that MMDesign consistently outperforms baselines on various public benchmarks. To further assess the biological plausibility, we present systematic quantitative analysis techniques that provide interpretability and reveal more about the laws of protein design."
                    },
                    {
                        "title": "FoldToken2: Learning compact, invariant and generative protein structure language",
                        "abstract": "The equivalent nature of 3D coordinates has posed long term challenges in protein structure representation learning, alignment, and generation. Can we create a compact and invariant language that equivalently represents protein structures? Towards this goal, we propose FoldToken2 to transfer equivariant structures into discrete tokens, while maintaining the recoverability of the original structures. From FoldToken1 to FoldToken2, we improve three key components: (1) invariant structure encoder, (2) vector-quantized compressor, and (3) equivalent structure decoder. We evaluate FoldToken2 on the protein structure reconstruction task and show that it outperforms previous FoldToken1 by 20\\% in TMScore and 81\\% in RMSD. FoldToken2 probably be the first method that works well on both single-chain and multi-chain protein structures quantization. We believe that FoldToken2 will inspire further improvement in protein structure representation learning, structure alignment, and structure generation tasks."
                    },
                    {
                        "title": "Molformer: Motif-based Transformer on 3D Heterogeneous Molecular Graphs",
                        "abstract": "Procuring expressive molecular representations underpins AI-driven molecule design and scientific discovery. The research mainly focuses on atom-level homogeneous molecular graphs, ignoring the rich information in subgraphs or motifs. However, it has been widely accepted that substructures play a dominant role in identifying and determining molecular properties. To address such issues, we formulate heterogeneous molecular graphs (HMGs), and introduce a novel architecture to exploit both molecular motifs and 3D geometry. Precisely, we extract functional groups as motifs for small molecules and employ reinforcement learning to adaptively select quaternary amino acids as motif candidates for proteins. Then HMGs are constructed with both atom-level and motif-level nodes. To better accommodate those HMGs, we introduce a variant of Transformer named Molformer, which adopts a heterogeneous self-attention layer to distinguish the interactions between multi-level nodes. Besides, it is also coupled with a multi-scale mechanism to capture fine-grained local patterns with increasing contextual scales. An attentive farthest point sampling algorithm is also proposed to obtain the molecular representations. We validate Molformer across a broad range of domains, including quantum chemistry, physiology, and biophysics. Extensive experiments show that Molformer outperforms or achieves the comparable performance of several state-of-the-art baselines. Our work provides a promising way to utilize informative motifs from the perspective of multi-level graph construction."
                    },
                    {
                        "title": "DiffMD: A Geometric Diffusion Model for Molecular Dynamics Simulations",
                        "abstract": "Molecular dynamics (MD) has long been the de facto choice for simulating complex atomistic systems from first principles. Recently deep learning models become a popular way to accelerate MD. Notwithstanding, existing models depend on intermediate variables such as the potential energy or force fields to update atomic positions, which requires additional computations to perform back-propagation. To waive this requirement, we propose a novel model called DiffMD by directly estimating the gradient of the log density of molecular conformations. DiffMD relies on a score-based denoising diffusion generative model that perturbs the molecular structure with a conditional noise depending on atomic accelerations and treats conformations at previous timeframes as the prior distribution for sampling. Another challenge of modeling such a conformation generation process is that a molecule is kinetic instead of static, which no prior works have strictly studied. To solve this challenge, we propose an equivariant geometric Transformer as the score function in the diffusion process to calculate corresponding gradients. It incorporates the directions and velocities of atomic motions via 3D spherical Fourier-Bessel representations. With multiple architectural improvements, we outperform state-of-the-art baselines on MD17 and isomers of C7O2H10 datasets. This work contributes to accelerating material and drug discovery."
                    },
                    {
                        "title": "DiffSDS: A language diffusion model for protein backbone inpainting under geometric conditions and constraints",
                        "abstract": "Have you ever been troubled by the complexity and computational cost of SE(3) protein structure modeling and been amazed by the simplicity and power of language modeling? Recent work has shown promise in simplifying protein structures as sequences of protein angles; therefore, language models could be used for unconstrained protein backbone generation. Unfortunately, such simplification is unsuitable for the constrained protein inpainting problem, where the model needs to recover masked structures conditioned on unmasked ones, as it dramatically increases the computing cost of geometric constraints. To overcome this dilemma, we suggest inserting a hidden \\textbf{a}tomic \\textbf{d}irection \\textbf{s}pace (\\textbf{ADS}) upon the language model, converting invariant backbone angles into equivalent direction vectors and preserving the simplicity, called Seq2Direct encoder ($\\text{Enc}_{s2d}$). Geometric constraints could be efficiently imposed on the newly introduced direction space. A Direct2Seq decoder ($\\text{Dec}_{d2s}$) with mathematical guarantees is also introduced to develop a \\textbf{SDS} ($\\text{Enc}_{s2d}$+$\\text{Dec}_{d2s}$) model. We apply the SDS model as the denoising neural network during the conditional diffusion process, resulting in a constrained generative model--\\textbf{DiffSDS}. Extensive experiments show that the plug-and-play ADS could transform the language model into a strong structural model without loss of simplicity. More importantly, the proposed DiffSDS outperforms previous strong baselines by a large margin on the task of protein inpainting."
                    },
                    {
                        "title": "Exploring Localization for Self-supervised Fine-grained Contrastive Learning",
                        "abstract": "Self-supervised contrastive learning has demonstrated great potential in learning visual representations. Despite their success in various downstream tasks such as image classification and object detection, self-supervised pre-training for fine-grained scenarios is not fully explored. We point out that current contrastive methods are prone to memorizing background/foreground texture and therefore have a limitation in localizing the foreground object. Analysis suggests that learning to extract discriminative texture information and localization are equally crucial for fine-grained self-supervised pre-training. Based on our findings, we introduce cross-view saliency alignment (CVSA), a contrastive learning framework that first crops and swaps saliency regions of images as a novel view generation and then guides the model to localize on foreground objects via a cross-view alignment loss. Extensive experiments on both small- and large-scale fine-grained classification benchmarks show that CVSA significantly improves the learned representation."
                    }
                ]
            }
        }
    },
    "2406.11741": {
        "paper_data": {
            "title": "Transcendence: Generative Models Can Outperform The Experts That Train Them",
            "url": "http://arxiv.org/abs/2406.11741v4",
            "arxiv_id": "2406.11741",
            "authors": [
                "Edwin Zhang",
                "Vincent Zhu",
                "Naomi Saphra",
                "Anat Kleiman",
                "Benjamin L. Edelman",
                "Milind Tambe",
                "Sham M. Kakade",
                "Eran Malach"
            ],
            "abstract": "Generative models are trained with the simple objective of imitating the conditional probability distribution induced by the data they are trained on. Therefore, when trained on data generated by humans, we may not expect the artificial model to outperform the humans on their original objectives. In this work, we study the phenomenon of transcendence: when a generative model achieves capabilities that surpass the abilities of the experts generating its data. We demonstrate transcendence by training an autoregressive transformer to play chess from game transcripts, and show that the trained model can sometimes achieve better performance than all players in the dataset. We theoretically prove that transcendence can be enabled by low-temperature sampling, and rigorously assess this claim experimentally. Finally, we discuss other sources of transcendence, laying the groundwork for future investigation of this phenomenon in a broader setting.",
            "introduction": "   1 Introduction  Generative models (GMs) are typically trained to mimic human behavior. These humans may be skilled in their various human objectives: answering a question, creating art, singing a song. The model has only one objective: minimizing the cross-entropy loss with respect to the output distribution, thereby adjusting it to match the distribution of human labels222Although chatbots are subject to a variety of post-training tuning methods, e.g., RLHF, we restrict our scope by assuming that the specialized knowledge and capacities are already provided by cross-entropy loss.. Therefore, one might assume the model can, at best, match the performance of an expert on their human objectives. Is it possible for these models to surpass\u2014to transcend\u2014their expert sources in some domains?   We illustrate an example of such transcendence\u00a0in Figure\u00a01, which measures the chess ratings (Glicko-2 [7]) of several transformer [35] models. Our experimental testbed is generative modeling on chess, which we choose as a domain for its well-understood, constrained nature. The transformer models are trained on public datasets of human chess transcripts, autoregressively predicting the next move in the game. To test for transcendence, we limit the maximal rating of the human players in the dataset below a specified score. We find that ChessFormer 1000100010001000 and ChessFormer 1300130013001300 (the latter number being the maximum rating seen during training) achieve significant levels of transcendence, surpassing the maximal rating seen in the dataset. Our focus is this capacity of a GM to transcend its expert sources by broadly outperforming any one expert. The key to our findings is the observation that GMs implicitly perform majority voting over the human experts. As these models are trained on a collection of many experts with diverse capacities, predilections, and biases, this majority vote oftentimes outperforms any individual expert, a phenomena that is known as \u201cwisdom of the crowd\u201d.   Our objective is to formalize the notion of transcendence and focus narrowly on this source of improvement over the experts: the removal of diverse human biases and errors. We prove that this form of denoising is enabled by low-temperature sampling, which implicitly induces a majority vote. Our result draws a subtle but deep connection from our new setting to a rich prior literature on model ensembling [1, 6, 19], enabling several key results. We precisely characterize the conditions under which transcendence is possible, and give a rigorous theoretical framework for enabling future study into the phenomenon. To test the predictive power of our theory, we then empirically demonstrate these effects. Digging deeper into the effects of majority voting, we show that its advantage is primarily due to performing much better on a small subset of states\u2014that is, under conditions that are likely key to determining the outcome of the game. We also find that diversity in the data is a necessary condition for practically effective majority voting, confirming our theoretical findings. In short:     \u2022  We formalize the notion of transcendence in generative models (Section 2).    \u2022  We find a key insight explaining one cause of transcendence by connecting the case of denoising experts to model ensembling. In low temperature sampling settings, we prove that a generative model can transcend",
            "references": [
                {
                    "title": "Emergent World Models and Latent Variable Estimation in Chess-Playing Language Models",
                    "abstract": "Language models have shown unprecedented capabilities, sparking debate over the source of their performance. Is it merely the outcome of learning syntactic patterns and surface level statistics, or do they extract semantics and a world model from the text? Prior work by Li et al. investigated this by training a GPT model on synthetic, randomly generated Othello games and found that the model learned an internal representation of the board state. We extend this work into the more complex domain of chess, training on real games and investigating our model's internal representations using linear probes and contrastive activations. The model is given no a priori knowledge of the game and is solely trained on next character prediction, yet we find evidence of internal representations of board state. We validate these internal representations by using them to make interventions on the model's activations and edit its internal board state. Unlike Li et al's prior synthetic dataset approach, our analysis finds that the model also learns to estimate latent variables like player skill to better predict the next character. We derive a player skill vector and add it to the model, improving the model's win rate by up to 2.6 times."
                },
                {
                    "title": "Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision",
                    "abstract": "Widely used alignment techniques, such as reinforcement learning from human feedback (RLHF), rely on the ability of humans to supervise model behavior - for example, to evaluate whether a model faithfully followed instructions or generated safe outputs. However, future superhuman models will behave in complex ways too difficult for humans to reliably evaluate; humans will only be able to weakly supervise superhuman models. We study an analogy to this problem: can weak model supervision elicit the full capabilities of a much stronger model? We test this using a range of pretrained language models in the GPT-4 family on natural language processing (NLP), chess, and reward modeling tasks. We find that when we naively finetune strong pretrained models on labels generated by a weak model, they consistently perform better than their weak supervisors, a phenomenon we call weak-to-strong generalization. However, we are still far from recovering the full capabilities of strong models with naive finetuning alone, suggesting that techniques like RLHF may scale poorly to superhuman models without further work. We find that simple methods can often significantly improve weak-to-strong generalization: for example, when finetuning GPT-4 with a GPT-2-level supervisor and an auxiliary confidence loss, we can recover close to GPT-3.5-level performance on NLP tasks. Our results suggest that it is feasible to make empirical progress today on a fundamental challenge of aligning superhuman models."
                },
                {
                    "title": "Leveraging Ensemble Diversity for Robust Self-Training in the Presence of Sample Selection Bias",
                    "abstract": "Self-training is a well-known approach for semi-supervised learning. It consists of iteratively assigning pseudo-labels to unlabeled data for which the model is confident and treating them as labeled examples. For neural networks, softmax prediction probabilities are often used as a confidence measure, although they are known to be overconfident, even for wrong predictions. This phenomenon is particularly intensified in the presence of sample selection bias, i.e., when data labeling is subject to some constraint. To address this issue, we propose a novel confidence measure, called $\\mathcal{T}$-similarity, built upon the prediction diversity of an ensemble of linear classifiers. We provide the theoretical analysis of our approach by studying stationary points and describing the relationship between the diversity of the individual members and their performance. We empirically demonstrate the benefit of our confidence measure for three different pseudo-labeling policies on classification datasets of various data modalities. The code is available at https://github.com/ambroiseodt/tsim."
                },
                {
                    "title": "ChessGPT: Bridging Policy Learning and Language Modeling",
                    "abstract": "When solving decision-making tasks, humans typically depend on information from two key sources: (1) Historical policy data, which provides interaction replay from the environment, and (2) Analytical insights in natural language form, exposing the invaluable thought process or strategic considerations. Despite this, the majority of preceding research focuses on only one source: they either use historical replay exclusively to directly learn policy or value functions, or engaged in language model training utilizing mere language corpus. In this paper, we argue that a powerful autonomous agent should cover both sources. Thus, we propose ChessGPT, a GPT model bridging policy learning and language modeling by integrating data from these two sources in Chess games. Specifically, we build a large-scale game and language dataset related to chess. Leveraging the dataset, we showcase two model examples ChessCLIP and ChessGPT, integrating policy learning and language modeling. Finally, we propose a full evaluation framework for evaluating language model's chess ability. Experimental results validate our model and dataset's effectiveness. We open source our code, model, and dataset at https://github.com/waterhorse1/ChessGPT."
                },
                {
                    "title": "Offline Reinforcement Learning with Closed-Form Policy Improvement Operators",
                    "abstract": "Behavior constrained policy optimization has been demonstrated to be a successful paradigm for tackling Offline Reinforcement Learning. By exploiting historical transitions, a policy is trained to maximize a learned value function while constrained by the behavior policy to avoid a significant distributional shift. In this paper, we propose our closed-form policy improvement operators. We make a novel observation that the behavior constraint naturally motivates the use of first-order Taylor approximation, leading to a linear approximation of the policy objective. Additionally, as practical datasets are usually collected by heterogeneous policies, we model the behavior policies as a Gaussian Mixture and overcome the induced optimization difficulties by leveraging the LogSumExp's lower bound and Jensen's Inequality, giving rise to a closed-form policy improvement operator. We instantiate offline RL algorithms with our novel policy improvement operators and empirically demonstrate their effectiveness over state-of-the-art algorithms on the standard D4RL benchmark. Our code is available at https://cfpi-icml23.github.io/."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "Learning from Disagreement: A Survey",
                    "abstract": "Many tasks in Natural Language Processing (NLP) and Computer Vision (CV) offer evidence that humans disagree, from objective tasks such as part-of-speech tagging to more subjective tasks such as classifying an image or deciding whether a proposition follows from certain premises. While most learning in artificial intelligence (AI) still relies on the assumption that a single (gold) interpretation exists for each item, a growing body of research aims to develop learning methods that do not rely on this assumption. In this survey, we review the evidence for disagreements on NLP and CV tasks, focusing on tasks for which substantial datasets containing this information have been created. We discuss the most popular approaches to training models from datasets containing multiple judgments potentially in disagreement. We systematically compare these different approaches by training them with each of the available datasets, considering several ways to evaluate the resulting models. Finally, we discuss the results in depth, focusing on four key research questions, and assess how the type of evaluation and the characteristics of a dataset determine the answers to these questions. Our results suggest, first of all, that even if we abandon the assumption of a gold standard, it is still essential to reach a consensus on how to evaluate models. This is because the relative performance of the various training methods is critically affected by the chosen form of evaluation. Secondly, we observed a strong dataset effect. With substantial datasets, providing many judgments by high-quality coders for each item, training directly with soft labels achieved better results than training from aggregated or even gold labels. This result holds for both hard and soft evaluation. But when the above conditions do not hold, leveraging both gold and soft labels generally achieved the best results in the hard evaluation. All datasets and models employed in this paper are freely available as supplementary materials."
                },
                {
                    "title": "Offline Reinforcement Learning as One Big Sequence Modeling Problem",
                    "abstract": "Reinforcement learning (RL) is typically concerned with estimating stationary policies or single-step models, leveraging the Markov property to factorize problems in time. However, we can also view RL as a generic sequence modeling problem, with the goal being to produce a sequence of actions that leads to a sequence of high rewards. Viewed in this way, it is tempting to consider whether high-capacity sequence prediction models that work well in other domains, such as natural-language processing, can also provide effective solutions to the RL problem. To this end, we explore how RL can be tackled with the tools of sequence modeling, using a Transformer architecture to model distributions over trajectories and repurposing beam search as a planning algorithm. Framing RL as sequence modeling problem simplifies a range of design decisions, allowing us to dispense with many of the components common in offline RL algorithms. We demonstrate the flexibility of this approach across long-horizon dynamics prediction, imitation learning, goal-conditioned RL, and offline RL. Further, we show that this approach can be combined with existing model-free algorithms to yield a state-of-the-art planner in sparse-reward, long-horizon tasks."
                },
                {
                    "title": "Decision Transformer: Reinforcement Learning via Sequence Modeling",
                    "abstract": "We introduce a framework that abstracts Reinforcement Learning (RL) as a sequence modeling problem. This allows us to draw upon the simplicity and scalability of the Transformer architecture, and associated advances in language modeling such as GPT-x and BERT. In particular, we present Decision Transformer, an architecture that casts the problem of RL as conditional sequence modeling. Unlike prior approaches to RL that fit value functions or compute policy gradients, Decision Transformer simply outputs the optimal actions by leveraging a causally masked Transformer. By conditioning an autoregressive model on the desired return (reward), past states, and actions, our Decision Transformer model can generate future actions that achieve the desired return. Despite its simplicity, Decision Transformer matches or exceeds the performance of state-of-the-art model-free offline RL baselines on Atari, OpenAI Gym, and Key-to-Door tasks."
                },
                {
                    "title": "Chess as a Testbed for Language Model State Tracking",
                    "abstract": "Transformer language models have made tremendous strides in natural language understanding tasks. However, the complexity of natural language makes it challenging to ascertain how accurately these models are tracking the world state underlying the text. Motivated by this issue, we consider the task of language modeling for the game of chess. Unlike natural language, chess notations describe a simple, constrained, and deterministic domain. Moreover, we observe that the appropriate choice of chess notation allows for directly probing the world state, without requiring any additional probing-related machinery. We find that: (a) With enough training data, transformer language models can learn to track pieces and predict legal moves with high accuracy when trained solely on move sequences. (b) For small training sets providing access to board state information during training can yield significant improvements. (c) The success of transformer language models is dependent on access to the entire game history i.e. \u201cfull attention\u201d. Approximating this full attention results in a significant performance drop. We propose this testbed as a benchmark for future work on the development and analysis of transformer language models."
                },
                {
                    "title": "The Chess Transformer: Mastering Play using Generative Language Models",
                    "abstract": "This work demonstrates that natural language transformers can support more generic strategic modeling, particularly for text-archived games. In addition to learning natural language skills, the abstract transformer architecture can generate meaningful moves on a chessboard. With further fine-tuning, the transformer learns complex gameplay by training on 2.8 million chess games in Portable Game Notation. After 30,000 training steps, OpenAI's Generative Pre-trained Transformer (GPT-2) optimizes weights for 774 million parameters. This fine-tuned Chess Transformer generates plausible strategies and displays game formations identifiable as classic openings, such as English or the Slav Exchange. Finally, in live play, the novel model demonstrates a human-to-transformer interface that correctly filters illegal moves and provides a novel method to challenge the transformer's chess strategies. We anticipate future work will build on this transformer's promise, particularly in other strategy games where features can capture the underlying complex rule syntax from simple but expressive player annotations."
                },
                {
                    "title": "Ensemble Distillation for Robust Model Fusion in Federated Learning",
                    "abstract": "Federated Learning (FL) is a machine learning setting where many devices collaboratively train a machine learning model while keeping the training data decentralized. In most of the current training schemes the central model is refined by averaging the parameters of the server model and the updated parameters from the client side. However, directly averaging model parameters is only possible if all models have the same structure and size, which could be a restrictive constraint in many scenarios. \nIn this work we investigate more powerful and more flexible aggregation schemes for FL. Specifically, we propose ensemble distillation for model fusion, i.e. training the central classifier through unlabeled data on the outputs of the models from the clients. This knowledge distillation technique mitigates privacy risk and cost to the same extent as the baseline FL algorithms, but allows flexible aggregation over heterogeneous client models that can differ e.g. in size, numerical precision or structure. We show in extensive empirical experiments on various CV/NLP datasets (CIFAR-10/100, ImageNet, AG News, SST2) and settings (heterogeneous models/data) that the server model can be trained much faster, requiring fewer communication rounds than any existing FL technique so far."
                },
                {
                    "title": "Aligning Superhuman AI with Human Behavior: Chess as a Model System",
                    "abstract": "As artificial intelligence becomes increasingly intelligent---in some cases, achieving superhuman performance---there is growing potential for humans to learn from and collaborate with algorithms. However, the ways in which AI systems approach problems are often different from the ways people do, and thus may be uninterpretable and hard to learn from. A crucial step in bridging this gap between human and artificial intelligence is modeling the granular actions that constitute human behavior, rather than simply matching aggregate human performance. We pursue this goal in a model system with a long history in artificial intelligence: chess. The aggregate performance of a chess player unfolds as they make decisions over the course of a game. The hundreds of millions of games played online by players at every skill level form a rich source of data in which these decisions, and their exact context, are recorded in minute detail. Applying existing chess engines to this data, including an open-source implementation of AlphaZero, we find that they do not predict human moves well. We develop and introduce Maia, a customized version of AlphaZero trained on human chess games, that predicts human moves at a much higher accuracy than existing engines, and can achieve maximum accuracy when predicting decisions made by players at a specific skill level in a tuneable way. For a dual task of predicting whether a human will make a large mistake on the next move, we develop a deep neural network that significantly outperforms competitive baselines. Taken together, our results suggest that there is substantial promise in designing artificial intelligence systems with human collaboration in mind by first accurately modeling granular human decision-making."
                },
                {
                    "title": "Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems",
                    "abstract": "In this tutorial article, we aim to provide the reader with the conceptual tools needed to get started on research on offline reinforcement learning algorithms: reinforcement learning algorithms that utilize previously collected data, without additional online data collection. Offline reinforcement learning algorithms hold tremendous promise for making it possible to turn large datasets into powerful decision making engines. Effective offline reinforcement learning methods would be able to extract policies with the maximum possible utility out of the available data, thereby allowing automation of a wide range of decision-making domains, from healthcare and education to robotics. However, the limitations of current algorithms make this difficult. We will aim to provide the reader with an understanding of these challenges, particularly in the context of modern deep reinforcement learning methods, and describe some potential solutions that have been explored in recent work to mitigate these challenges, along with recent applications, and a discussion of perspectives on open problems in the field."
                },
                {
                    "title": "Stable Policy Optimization via Off-Policy Divergence Regularization",
                    "abstract": "Trust Region Policy Optimization (TRPO) and Proximal Policy Optimization (PPO) are among the most successful policy gradient approaches in deep reinforcement learning (RL). While these methods achieve state-of-the-art performance across a wide range of challenging tasks, there is room for improvement in the stabilization of the policy learning and how the off-policy data are used. In this paper we revisit the theoretical foundations of these algorithms and propose a new algorithm which stabilizes the policy improvement through a proximity term that constrains the discounted state-action visitation distribution induced by consecutive policies to be close to one another. This proximity term, expressed in terms of the divergence between the visitation distributions, is learned in an off-policy and adversarial manner. We empirically show that our proposed method can have a beneficial effect on stability and improve final performance in benchmark high-dimensional control tasks."
                },
                {
                    "title": "Self-Training With Noisy Student Improves ImageNet Classification",
                    "abstract": "We present a simple self-training method that achieves 88.4% top-1 accuracy on ImageNet, which is 2.0% better than the state-of-the-art model that requires 3.5B weakly labeled Instagram images. On robustness test sets, it improves ImageNet-A top-1 accuracy from 61.0% to 83.7%, reduces ImageNet-C mean corruption error from 45.7 to 28.3, and reduces ImageNet-P mean flip rate from 27.8 to 12.2. To achieve this result, we first train an EfficientNet model on labeled ImageNet images and use it as a teacher to generate pseudo labels on 300M unlabeled images. We then train a larger EfficientNet as a student model on the combination of labeled and pseudo labeled images. We iterate this process by putting back the student as the teacher. During the generation of the pseudo labels, the teacher is not noised so that the pseudo labels are as accurate as possible. However, during the learning of the student, we inject noise such as dropout, stochastic depth and data augmentation via RandAugment to the student so that the student generalizes better than the teacher."
                },
                {
                    "title": "Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm",
                    "abstract": "The game of chess is the most widely-studied domain in the history of artificial intelligence. The strongest programs are based on a combination of sophisticated search techniques, domain-specific adaptations, and handcrafted evaluation functions that have been refined by human experts over several decades. In contrast, the AlphaGo Zero program recently achieved superhuman performance in the game of Go, by tabula rasa reinforcement learning from games of self-play. In this paper, we generalise this approach into a single AlphaZero algorithm that can achieve, tabula rasa, superhuman performance in many challenging domains. Starting from random play, and given no domain knowledge except the game rules, AlphaZero achieved within 24 hours a superhuman level of play in the games of chess and shogi (Japanese chess) as well as Go, and convincingly defeated a world-champion program in each case."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Safe and Efficient Off-Policy Reinforcement Learning",
                    "abstract": "In this work, we take a fresh look at some old and new algorithms for off-policy, return-based reinforcement learning. Expressing these in a common form, we derive a novel algorithm, Retrace($\\lambda$), with three desired properties: (1) it has low variance; (2) it safely uses samples collected from any behaviour policy, whatever its degree of \"off-policyness\"; and (3) it is efficient as it makes the best use of samples collected from near on-policy behaviour policies. We analyze the contractive nature of the related operator under both off-policy policy evaluation and control settings and derive online sample-based algorithms. We believe this is the first return-based off-policy control algorithm converging a.s. to $Q^*$ without the GLIE assumption (Greedy in the Limit with Infinite Exploration). As a corollary, we prove the convergence of Watkins' Q($\\lambda$), which was an open problem since 1989. We illustrate the benefits of Retrace($\\lambda$) on a standard suite of Atari 2600 games."
                },
                {
                    "title": "Communication-Efficient Learning of Deep Networks from Decentralized Data",
                    "abstract": "Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data center and training there using conventional approaches. We advocate an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. \nWe present a practical method for the federated learning of deep networks based on iterative model averaging, and conduct an extensive empirical evaluation, considering five different model architectures and four datasets. These experiments demonstrate the approach is robust to the unbalanced and non-IID data distributions that are a defining characteristic of this setting. Communication costs are the principal constraint, and we show a reduction in required communication rounds by 10-100x as compared to synchronized stochastic gradient descent."
                },
                {
                    "title": "Give a Hard Problem to a Diverse Team: Exploring Large Action Spaces",
                    "abstract": "\n \n Recent work has shown that diverse teams can outperform a uniform team made of copies of the best agent. However, there are fundamental questions that were not asked before. When should we use diverse or uniform teams? How does the performance change as the action space or the teams get larger? Hence, we present a new model of diversity for teams, that is more general than previous models. We prove that the performance of a diverse team improves as the size of the action space gets larger. Concerning the size of the diverse team, we show that the performance converges exponentially fast to the optimal one as we increase the number of agents. We present synthetic experiments that allow us to gain further insights: even though a diverse team outperforms a uniform team when the size of the action space increases, the uniform team will eventually again play better than the diverse team for a large enough action space. We verify our predictions in a system of Go playing agents, where we show a diverse team that improves in performance as the board size increases, and eventually overcomes a uniform team.\n \n"
                },
                {
                    "title": "Multi-Agent Team Formation: Diversity Beats Strength?",
                    "abstract": "Team formation is a critical step in deploying a multi-agent team. In some scenarios, agents coordinate by voting continuously. When forming such teams, should we focus on the diversity of the team or on the strength of each member? Can a team of diverse (and weak) agents outperform a uniform team of strong agents? We propose a new model to address these questions. Our key contributions include: (i) we show that a diverse team can overcome a uniform team and we give the necessary conditions for it to happen; (ii) we present optimal voting rules for a diverse team; (iii) we perform synthetic experiments that demonstrate that both diversity and strength contribute to the performance of a team; (iv) we show experiments that demonstrate the usefulness of our model in one of the most difficult challenges for Artificial Intelligence: Computer Go."
                },
                {
                    "title": "XXII. Programming a Computer for Playing Chess 1",
                    "abstract": "(1950). XXII. Programming a computer for playing chess. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science: Vol. 41, No. 314, pp. 256-275."
                },
                {
                    "title": "Visualizing Data using t-SNE",
                    "abstract": "We present a new technique called \u201ct-SNE\u201d that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to influence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are significantly better than those produced by the other techniques on almost all of the datasets."
                },
                {
                    "title": "A Short Introduction to Boosting",
                    "abstract": "Boosting is a general method for improving the accuracy of any given learning algorithm. This short overview paper introduces the boosting algorithm AdaBoost, and explains the underlying theory of boosting, including an explanation of why boosting often does not suffer from overfitting as well as boosting\u2019s relationship to support-vector machines. Some examples of recent applications of boosting are also described."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan generative models (GMs) transcend the performance of their expert sources in specific domains, such as chess, by leveraging the diversity of human expertise?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem has significant implications for the research community as it challenges the conventional understanding of generative models' capabilities. If GMs can indeed surpass expert performance, it could lead to advancements in AI applications across various fields, including game strategy, decision-making, and creative tasks. This research could inspire new methodologies for training models that harness the \"wisdom of the crowd,\" potentially leading to more robust and versatile AI systems. Furthermore, it opens avenues for future research into the mechanisms of model ensembling and the role of diversity in training data.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of formalizing the concept of transcendence and understanding the conditions under which it occurs. Naive approaches may fail because they do not account for the intricate dynamics of majority voting among diverse expert inputs, which can obscure the underlying patterns necessary for superior performance. Additionally, technical obstacles include the need for rigorous theoretical frameworks to characterize the conditions for transcendence and the empirical validation of these theories, particularly in constrained environments like chess.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on the limitations of generative models in mimicking expert behavior without exploring the potential for surpassing it. Gaps in understanding the role of diversity in training data and the mechanisms of model ensembling have hindered progress. Barriers include a lack of formal definitions and frameworks for transcendence, as well as insufficient empirical studies demonstrating the phenomenon. Our approach differs by explicitly formalizing transcendence, connecting it to model ensembling, and providing a theoretical basis for future exploration.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves training transformer models on public datasets of human chess transcripts, focusing on autoregressive prediction of moves. We will employ low-temperature sampling to facilitate majority voting among diverse expert inputs. The key metrics for evaluation will include chess ratings (Glicko-2) and performance on critical game states. We expect to demonstrate that GMs can achieve transcendence by outperforming the highest-rated human players in the dataset, confirming that diversity in training data is essential for effective majority voting and providing a theoretical framework for understanding this"
            }
        },
        "author_data": {
            "353c76dd-06b1-4d6a-bd11-c37413f724a8": {
                "pk": "353c76dd-06b1-4d6a-bd11-c37413f724a8",
                "name": "Edwin Zhang",
                "collaborators": [
                    "Yujie Lu",
                    "William Wang",
                    "Amy Zhang"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Language Models",
                    "Diffusion Models",
                    "Robotics"
                ],
                "publications": [
                    {
                        "title": "Language Control Diffusion: Efficiently Scaling through Space, Time, and Tasks",
                        "abstract": "Training generalist agents is difficult across several axes, requiring us to deal with high-dimensional inputs (space), long horizons (time), and generalization to novel tasks. Recent advances with architectures have allowed for improved scaling along one or two of these axes, but are still computationally prohibitive to use. In this paper, we propose to address all three axes by leveraging \\textbf{L}anguage to \\textbf{C}ontrol \\textbf{D}iffusion models as a hierarchical planner conditioned on language (LCD). We effectively and efficiently scale diffusion models for planning in extended temporal, state, and task dimensions to tackle long horizon control problems conditioned on natural language instructions, as a step towards generalist agents. Comparing LCD with other state-of-the-art models on the CALVIN language robotics benchmark finds that LCD outperforms other SOTA methods in multi-task success rates, whilst improving inference speed over other comparable diffusion models by 3.3x~15x. We show that LCD can successfully leverage the unique strength of diffusion models to produce coherent long range plans while addressing their weakness in generating low-level details and control."
                    }
                ]
            },
            "842fa3e5-e155-4938-84b8-b04ad22136aa": {
                "pk": "842fa3e5-e155-4938-84b8-b04ad22136aa",
                "name": "Vincent Zhu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "e23cfacf-6c42-4d24-a27b-cab4cd6e341b": {
                "pk": "e23cfacf-6c42-4d24-a27b-cab4cd6e341b",
                "name": "Naomi Saphra",
                "collaborators": [
                    "Adam Lopez",
                    "Ryan Cotterell",
                    "Kyunghyun Cho",
                    "Tiago Pimentel",
                    "Adina Williams",
                    "Sarah Wiegreffe",
                    "Eve Fleisig",
                    "Bingchen Zhao",
                    "Yuling Gu",
                    "Jessica Zosa Forde"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Interpretability",
                    "Machine Learning",
                    "Compositionality"
                ],
                "publications": [
                    {
                        "title": "Evaluating Informal-Domain Word Representations With UrbanDictionary",
                        "abstract": "Existing corpora for intrinsic evaluation are not targeted towards tasks in informal domains such as Twitter or news comment forums. We want to test whether a representation of informal words fulfills the promise of eliding explicit text normalization as a preprocessing step. One possible evaluation metric for such domains is the proximity of spelling variants. We propose how such a metric might be computed and how a spelling variant dataset can be collected using UrbanDictionary."
                    },
                    {
                        "title": "Sparsity Emerges Naturally in Neural Language Models",
                        "abstract": "Concerns about interpretability, computational resources, and principled inductive priors have motivated efforts to engineer sparse neural models for NLP tasks. If sparsity is important for NLP, might well-trained neural models naturally become roughly sparse? Using the Taxi-Euclidean norm to measure sparsity, we find that frequent input words are associated with concentrated or sparse activations, while frequent target words are associated with dispersed activations but concentrated gradients. We find that gradients associated with function words are more concentrated than the gradients of content words, even controlling for word frequency."
                    },
                    {
                        "title": "Understanding Learning Dynamics Of Language Models with SVCCA",
                        "abstract": "Research has shown that neural models implicitly encode linguistic features, but there has been no research showing \\emph{how} these encodings arise as the models are trained. We present the first study on the learning dynamics of neural language models, using a simple and flexible analysis method called Singular Vector Canonical Correlation Analysis (SVCCA), which enables us to compare learned representations across time and across models, without the need to evaluate directly on annotated data. We probe the evolution of syntactic, semantic, and topic representations and find that part-of-speech is learned earlier than topic; that recurrent layers become more similar to those of a tagger during training; and embedding layers less similar. Our results and methods could inform better learning algorithms for NLP models, possibly to incorporate linguistic information more effectively."
                    },
                    {
                        "title": "LSTMs Compose (and Learn) Bottom-Up",
                        "abstract": "Recent work in NLP shows that LSTM language models capture hierarchical structure in language data. In contrast to existing work, we consider the \\textit{learning} process that leads to their compositional behavior. For a closer look at how an LSTM's sequential representations are composed hierarchically, we present a related measure of Decompositional Interdependence (DI) between word meanings in an LSTM, based on their gate interactions. We connect this measure to syntax with experiments on English language data, where DI is higher on pairs of words with lower syntactic distance. To explore the inductive biases that cause these compositional representations to arise during training, we conduct simple experiments on synthetic data. These synthetic experiments support a specific hypothesis about how hierarchical structures are discovered over the course of training: that LSTM constituent representations are learned bottom-up, relying on effective representations of their shorter children, rather than learning the longer-range relations independently from children."
                    },
                    {
                        "title": "Word Interdependence Exposes How LSTMs Compose Representations",
                        "abstract": "Recent work in NLP shows that LSTM language models capture compositional structure in language data. For a closer look at how these representations are composed hierarchically, we present a novel measure of interdependence between word meanings in an LSTM, based on their interactions at the internal gates. To explore how compositional representations arise over training, we conduct simple experiments on synthetic data, which illustrate our measure by showing how high interdependence can hurt generalization. These synthetic experiments also illustrate a specific hypothesis about how hierarchical structures are discovered over the course of training: that parent constituents rely on effective representations of their children, rather than on learning long-range relations independently. We further support this measure with experiments on English language data, where interdependence is higher for more closely syntactically linked word pairs."
                    },
                    {
                        "title": "Mechanistic?",
                        "abstract": "The rise of the term \"mechanistic interpretability\" has accompanied increasing interest in understanding neural models -- particularly language models. However, this jargon has also led to a fair amount of confusion. So, what does it mean to be \"mechanistic\"? We describe four uses of the term in interpretability research. The most narrow technical definition requires a claim of causality, while a broader technical definition allows for any exploration of a model's internals. However, the term also has a narrow cultural definition describing a cultural movement. To understand this semantic drift, we present a history of the NLP interpretability community and the formation of the separate, parallel \"mechanistic\" interpretability community. Finally, we discuss the broad cultural definition -- encompassing the entire field of interpretability -- and why the traditional NLP interpretability community has come to embrace it. We argue that the polysemy of \"mechanistic\" is the product of a critical divide within the interpretability community."
                    },
                    {
                        "title": "First Tragedy, then Parse: History Repeats Itself in the New Era of Large Language Models",
                        "abstract": "Many NLP researchers are experiencing an existential crisis triggered by the astonishing success of ChatGPT and other systems based on large language models (LLMs). After such a disruptive change to our understanding of the field, what is left to do? Taking a historical lens, we look for guidance from the first era of LLMs, which began in 2005 with large $n$-gram models for machine translation (MT). We identify durable lessons from the first era, and more importantly, we identify evergreen problems where NLP researchers can continue to make meaningful contributions in areas where LLMs are ascendant. We argue that disparities in scale are transient and researchers can work to reduce them; that data, rather than hardware, is still a bottleneck for many applications; that meaningful realistic evaluation is still an open problem; and that there is still room for speculative approaches."
                    },
                    {
                        "title": "Pareto Probing: Trading Off Accuracy for Complexity",
                        "abstract": "The question of how to probe contextual word representations for linguistic structure in a way that is both principled and useful has seen significant attention recently in the NLP literature. In our contribution to this discussion, we argue for a probe metric that reflects the fundamental trade-off between probe complexity and performance: the Pareto hypervolume. To measure complexity, we present a number of parametric and non-parametric metrics. Our experiments using Pareto hypervolume as an evaluation metric show that probes often do not conform to our expectations -- e.g., why should the non-contextual fastText representations encode more morpho-syntactic information than the contextual BERT representations? These results suggest that common, simplistic probing tasks, such as part-of-speech labeling and dependency arc labeling, are inadequate to evaluate the linguistic structure encoded in contextual word representations. This leads us to propose full dependency parsing as a probing task. In support of our suggestion that harder probing tasks are necessary, our experiments with dependency parsing reveal a wide gap in syntactic knowledge between contextual and non-contextual representations."
                    },
                    {
                        "title": "One Venue, Two Conferences: The Separation of Chinese and American Citation Networks",
                        "abstract": "At NeurIPS, American and Chinese institutions cite papers from each other's regions substantially less than they cite endogamously. We build a citation graph to quantify this divide, compare it to European connectivity, and discuss the causes and consequences of the separation."
                    },
                    {
                        "title": "ChatGPT Doesn't Trust Chargers Fans: Guardrail Sensitivity in Context",
                        "abstract": "While the biases of language models in production are extensively documented, the biases of their guardrails have been neglected. This paper studies how contextual information about the user influences the likelihood of an LLM to refuse to execute a request. By generating user biographies that offer ideological and demographic information, we find a number of biases in guardrail sensitivity on GPT-3.5. Younger, female, and Asian-American personas are more likely to trigger a refusal guardrail when requesting censored or illegal information. Guardrails are also sycophantic, refusing to comply with requests for a political position the user is likely to disagree with. We find that certain identity groups and seemingly innocuous information, e.g., sports fandom, can elicit changes in guardrail sensitivity similar to direct statements of political ideology. For each demographic category and even for American football team fandom, we find that ChatGPT appears to infer a likely political ideology and modify guardrail behavior accordingly."
                    },
                    {
                        "title": "How to Evaluate Word Representations of Informal Domain?",
                        "abstract": "Diverse word representations have surged in most state-of-the-art natural language processing (NLP) applications. Nevertheless, how to efficiently evaluate such word embeddings in the informal domain such as Twitter or forums, remains an ongoing challenge due to the lack of sufficient evaluation dataset. We derived a large list of variant spelling pairs from UrbanDictionary with the automatic approaches of weakly-supervised pattern-based bootstrapping and self-training linear-chain conditional random field (CRF). With these extracted relation pairs we promote the odds of eliding the text normalization procedure of traditional NLP pipelines and directly adopting representations of non-standard words in the informal domain. Our code is available."
                    },
                    {
                        "title": "Towards out-of-distribution generalization in large-scale astronomical surveys: robust networks learn similar representations",
                        "abstract": "The generalization of machine learning (ML) models to out-of-distribution (OOD) examples remains a key challenge in extracting information from upcoming astronomical surveys. Interpretability approaches are a natural way to gain insights into the OOD generalization problem. We use Centered Kernel Alignment (CKA), a similarity measure metric of neural network representations, to examine the relationship between representation similarity and performance of pre-trained Convolutional Neural Networks (CNNs) on the CAMELS Multifield Dataset. We find that when models are robust to a distribution shift, they produce substantially different representations across their layers on OOD data. However, when they fail to generalize, these representations change less from layer to layer on OOD data. We discuss the potential application of similarity representation in guiding model design, training strategy, and mitigating the OOD problem by incorporating CKA as an inductive bias during training."
                    },
                    {
                        "title": "Fast Forwarding Low-Rank Training",
                        "abstract": "Parameter efficient finetuning methods like low-rank adaptation (LoRA) aim to reduce the computational costs of finetuning pretrained Language Models (LMs). Enabled by these low-rank settings, we propose an even more efficient optimization strategy: Fast Forward, a simple and effective approach to accelerate large segments of training. In a Fast Forward stage, we repeat the most recent optimizer step until the loss stops improving on a tiny validation set. By alternating between regular optimization steps and Fast Forward stages, Fast Forward provides up to an 87\\% reduction in FLOPs and up to an 81\\% reduction in train time over standard SGD with Adam. We validate Fast Forward by finetuning various models on different tasks and demonstrate that it speeds up training without compromising model performance. Additionally, we analyze when and how to apply Fast Forward."
                    },
                    {
                        "title": "TRAM: Bridging Trust Regions and Sharpness Aware Minimization",
                        "abstract": "Sharpness-aware minimization (SAM) reports improving domain generalization by reducing the loss surface curvature in the parameter space. However, generalization during fine-tuning is often more dependent on the transferability of representations in the function space. Trust-region methods (TR) target this goal by regularizing representation curvature to reduce catastrophic forgetting of pre-trained task-agnostic information while adopting task-specific skills. We consider unifying these strategies for low curvature in both parameter space and function space to improve out-of-domain (OOD) generalization. We propose Trust Region Aware Minimization (TRAM), a SAM algorithm fine-tuning for low parameter sharpness and smooth, informative representations preserving pre-trained structure. TRAM uses a trust region bound to inform the SAM adversarial neighborhood, introducing an awareness of function curvature within optimization for flatter minima. We empirically validate TRAM in vision (cross-dataset adaptation) and text (OOD language modeling, zero-shot cross-lingual transfer) tasks where robust domain transfer and representation generality are critical. TRAM outperforms SAM- and TR-based optimization across all tasks, notably surpassing competing methods for hard transfer between anticorrelated domains. TRAM establishes a novel standard in fine-tuning for domain-generalizable models with minimal additional computation over previous sharpness-aware methods."
                    },
                    {
                        "title": "A Non-Linear Structural Probe",
                        "abstract": "Probes are models devised to investigate the encoding of knowledge -- e.g. syntactic structure -- in contextual representations. Probes are often designed for simplicity, which has led to restrictions on probe design that may not allow for the full exploitation of the structure of encoded information; one such restriction is linearity. We examine the case of a structural probe (Hewitt and Manning, 2019), which aims to investigate the encoding of syntactic structure in contextual representations through learning only linear transformations. By observing that the structural probe learns a metric, we are able to kernelize it and develop a novel non-linear variant with an identical number of parameters. We test on 6 languages and find that the radial-basis function (RBF) kernel, in conjunction with regularization, achieves a statistically significant improvement over the baseline in all languages -- implying that at least part of the syntactic knowledge is encoded non-linearly. We conclude by discussing how the RBF kernel resembles BERT's self-attention layers and speculate that this resemblance leads to the RBF-based probe's stronger performance."
                    },
                    {
                        "title": "Benchmarking Compositionality with Formal Languages",
                        "abstract": "Recombining known primitive concepts into larger novel combinations is a quintessentially human cognitive capability. Whether large neural models in NLP can acquire this ability while learning from data is an open question. In this paper, we investigate this problem from the perspective of formal languages. We use deterministic finite-state transducers to make an unbounded number of datasets with controllable properties governing compositionality. By randomly sampling over many transducers, we explore which of their properties contribute to learnability of a compositional relation by a neural network. We find that the models either learn the relations completely or not at all. The key is transition coverage, setting a soft learnability limit at 400 examples per transition."
                    },
                    {
                        "title": "Latent State Models of Training Dynamics",
                        "abstract": "The impact of randomness on model training is poorly understood. How do differences in data order and initialization actually manifest in the model, such that some training runs outperform others or converge faster? Furthermore, how can we interpret the resulting training dynamics and the phase transitions that characterize different trajectories? To understand the effect of randomness on the dynamics and outcomes of neural network training, we train models multiple times with different random seeds and compute a variety of metrics throughout training, such as the $L_2$ norm, mean, and variance of the neural network's weights. We then fit a hidden Markov model (HMM) over the resulting sequences of metrics. The HMM represents training as a stochastic process of transitions between latent states, providing an intuitive overview of significant changes during training. Using our method, we produce a low-dimensional, discrete representation of training dynamics on grokking tasks, image classification, and masked language modeling. We use the HMM representation to study phase transitions and identify latent \"detour\" states that slow down convergence."
                    },
                    {
                        "title": "Knowing Your Nonlinearities: Shapley Interactions Reveal the Underlying Structure of Data",
                        "abstract": "Measuring nonlinear feature interaction is an established approach to understanding complex patterns of attribution in many models. In this paper, we use Shapley Taylor interaction indices (STII) to analyze the impact of underlying data structure on model representations in a variety of modalities, tasks, and architectures. Considering linguistic structure in masked and auto-regressive language models (MLMs and ALMs), we find that STII increases within idiomatic expressions and that MLMs scale STII with syntactic distance, relying more on syntax in their nonlinear structure than ALMs do. Our speech model findings reflect the phonetic principal that the openness of the oral cavity determines how much a phoneme varies based on its context. Finally, we study image classifiers and illustrate that feature interactions intuitively reflect object boundaries. Our wide range of results illustrates the benefits of interdisciplinary work and domain expertise in interpretability research."
                    },
                    {
                        "title": "Dynamic Masking Rate Schedules for MLM Pretraining",
                        "abstract": "Most works on transformers trained with the Masked Language Modeling (MLM) objective use the original BERT model's fixed masking rate of 15%. We propose to instead dynamically schedule the masking rate throughout training. We find that linearly decreasing the masking rate over the course of pretraining improves average GLUE accuracy by up to 0.46% and 0.25% in BERT-base and BERT-large, respectively, compared to fixed rate baselines. These gains come from exposure to both high and low masking rate regimes, providing benefits from both settings. Our results demonstrate that masking rate scheduling is a simple way to improve the quality of masked language models, achieving up to a 1.89x speedup in pretraining for BERT-base as well as a Pareto improvement for BERT-large."
                    },
                    {
                        "title": "Attribute Diversity Determines the Systematicity Gap in VQA",
                        "abstract": "Although modern neural networks often generalize to new combinations of familiar concepts, the conditions that enable such compositionality have long been an open question. In this work, we study the systematicity gap in visual question answering: the performance difference between reasoning on previously seen and unseen combinations of object attributes. To test, we introduce a novel diagnostic dataset, CLEVR-HOPE. We find that the systematicity gap is not reduced by increasing the quantity of training data, but is reduced by increasing the diversity of training data. In particular, our experiments suggest that the more distinct attribute type combinations are seen during training, the more systematic we can expect the resulting model to be."
                    }
                ]
            },
            "cae78ff7-72ae-4a61-9fba-e05dea9a6e64": {
                "pk": "cae78ff7-72ae-4a61-9fba-e05dea9a6e64",
                "name": "Anat Kleiman",
                "collaborators": [
                    "Chi Zhang",
                    "Ryan Marcus",
                    "Olga Papaemmanouil",
                    "Angelina Wang",
                    "Alexander Liu",
                    "Ryan Zhang",
                    "Leslie Kim",
                    "Dora Zhao",
                    "Iroha Shirai",
                    "Arvind Narayanan"
                ],
                "domain": [
                    "Machine Learning",
                    "Database Systems",
                    "Bias Mitigation",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Buffer Pool Aware Query Scheduling via Deep Reinforcement Learning",
                        "abstract": "In this extended abstract, we propose a new technique for query scheduling with the explicit goal of reducing disk reads and thus implicitly increasing query performance. We introduce SmartQueue, a learned scheduler that leverages overlapping data reads among incoming queries and learns a scheduling strategy that improves cache hits. SmartQueue relies on deep reinforcement learning to produce workload-specific scheduling strategies that focus on long-term performance benefits while being adaptive to previously-unseen data access patterns. We present results from a proof-of-concept prototype, demonstrating that learned schedulers can offer significant performance improvements over hand-crafted scheduling heuristics. Ultimately, we make the case that this is a promising research direction at the intersection of machine learning and databases."
                    },
                    {
                        "title": "REVISE: A Tool for Measuring and Mitigating Bias in Visual Datasets",
                        "abstract": "Machine learning models are known to perpetuate and even amplify the biases present in the data. However, these data biases frequently do not become apparent until after the models are deployed. Our work tackles this issue and enables the preemptive analysis of large-scale datasets. REVISE (REvealing VIsual biaSEs) is a tool that assists in the investigation of a visual dataset, surfacing potential biases along three dimensions: (1) object-based, (2) person-based, and (3) geography-based. Object-based biases relate to the size, context, or diversity of the depicted objects. Person-based metrics focus on analyzing the portrayal of people within the dataset. Geography-based analyses consider the representation of different geographic locations. These three dimensions are deeply intertwined in how they interact to bias a dataset, and REVISE sheds light on this; the responsibility then lies with the user to consider the cultural and historical context, and to determine which of the revealed biases may be problematic. The tool further assists the user by suggesting actionable steps that may be taken to mitigate the revealed biases. Overall, the key aim of our work is to tackle the machine learning bias problem early in the pipeline. REVISE is available at https://github.com/princetonvisualai/revise-tool"
                    }
                ]
            },
            "4934b971-c253-4a12-9a04-343aa01a56e2": {
                "pk": "4934b971-c253-4a12-9a04-343aa01a56e2",
                "name": "Benjamin L. Edelman",
                "collaborators": [
                    "Surbhi Goel",
                    "Sham Kakade",
                    "Eran Malach",
                    "Cyril Zhang",
                    "Boaz Barak",
                    "Enric Boix-Adser\u00e0",
                    "Siddhartha Jayanti",
                    "Ezra Edelman",
                    "Nikolaos Tsilivis",
                    "Depen Morwani"
                ],
                "domain": [
                    "Game Theory",
                    "Natural Language Processing",
                    "Neural Networks",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "The Multiplayer Colonel Blotto Game",
                        "abstract": "We initiate the study of the natural multiplayer generalization of the classic continuous Colonel Blotto game. The two-player Blotto game, introduced by Borel as a model of resource competition across $n$ simultaneous fronts, has been studied extensively for a century and seen numerous applications throughout the social sciences. Our work defines the multiplayer Colonel Blotto game and derives Nash equilibria for various settings of $k$ (number of players) and $n$. We also introduce a \"Boolean\" version of Blotto that becomes interesting in the multiplayer setting. The main technical difficulty of our work, as in the two-player theoretical literature, is the challenge of coupling various marginal distributions into a joint distribution satisfying a strict sum constraint. In contrast to previous works in the continuous setting, we derive our couplings algorithmically in the form of efficient sampling algorithms."
                    },
                    {
                        "title": "The Evolution of Statistical Induction Heads: In-Context Learning Markov Chains",
                        "abstract": "Large language models have the ability to generate text that mimics patterns in their inputs. We introduce a simple Markov Chain sequence modeling task in order to study how this in-context learning (ICL) capability emerges. In our setting, each example is sampled from a Markov chain drawn from a prior distribution over Markov chains. Transformers trained on this task form \\emph{statistical induction heads} which compute accurate next-token probabilities given the bigram statistics of the context. During the course of training, models pass through multiple phases: after an initial stage in which predictions are uniform, they learn to sub-optimally predict using in-context single-token statistics (unigrams); then, there is a rapid phase transition to the correct in-context bigram solution. We conduct an empirical and theoretical investigation of this multi-phase process, showing how successful learning results from the interaction between the transformer's layers, and uncovering evidence that the presence of the simpler unigram solution may delay formation of the final bigram solution. We examine how learning is affected by varying the prior distribution over Markov chains, and consider the generalization of our in-context learning of Markov chains (ICL-MC) task to $n$-grams for $n > 2$."
                    },
                    {
                        "title": "Inductive Biases and Variable Creation in Self-Attention Mechanisms",
                        "abstract": "Self-attention, an architectural motif designed to model long-range interactions in sequential data, has driven numerous recent breakthroughs in natural language processing and beyond. This work provides a theoretical analysis of the inductive biases of self-attention modules. Our focus is to rigorously establish which functions and long-range dependencies self-attention blocks prefer to represent. Our main result shows that bounded-norm Transformer networks \"create sparse variables\": a single self-attention head can represent a sparse function of the input sequence, with sample complexity scaling only logarithmically with the context length. To support our analysis, we present synthetic experiments to probe the sample complexity of learning sparse Boolean functions with Transformers."
                    },
                    {
                        "title": "Feature emergence via margin maximization: case studies in algebraic tasks",
                        "abstract": "Understanding the internal representations learned by neural networks is a cornerstone challenge in the science of machine learning. While there have been significant recent strides in some cases towards understanding how neural networks implement specific target functions, this paper explores a complementary question -- why do networks arrive at particular computational strategies? Our inquiry focuses on the algebraic learning tasks of modular addition, sparse parities, and finite group operations. Our primary theoretical findings analytically characterize the features learned by stylized neural networks for these algebraic tasks. Notably, our main technique demonstrates how the principle of margin maximization alone can be used to fully specify the features learned by the network. Specifically, we prove that the trained networks utilize Fourier features to perform modular addition and employ features corresponding to irreducible group-theoretic representations to perform compositions in general groups, aligning closely with the empirical observations of Nanda et al. and Chughtai et al. More generally, we hope our techniques can help to foster a deeper understanding of why neural networks adopt specific computational strategies."
                    },
                    {
                        "title": "Distinguishing the Knowable from the Unknowable with Language Models",
                        "abstract": "We study the feasibility of identifying epistemic uncertainty (reflecting a lack of knowledge), as opposed to aleatoric uncertainty (reflecting entropy in the underlying distribution), in the outputs of large language models (LLMs) over free-form text. In the absence of ground-truth probabilities, we explore a setting where, in order to (approximately) disentangle a given LLM's uncertainty, a significantly larger model stands in as a proxy for the ground truth. We show that small linear probes trained on the embeddings of frozen, pretrained models accurately predict when larger models will be more confident at the token level and that probes trained on one text domain generalize to others. Going further, we propose a fully unsupervised method that achieves non-trivial accuracy on the same task. Taken together, we interpret these results as evidence that LLMs naturally contain internal representations of different types of uncertainty that could potentially be leveraged to devise more informative indicators of model confidence in diverse practical settings."
                    },
                    {
                        "title": "Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limit",
                        "abstract": "There is mounting evidence of emergent phenomena in the capabilities of deep learning methods as we scale up datasets, model sizes, and training times. While there are some accounts of how these resources modulate statistical capacity, far less is known about their effect on the computational problem of model training. This work conducts such an exploration through the lens of learning a $k$-sparse parity of $n$ bits, a canonical discrete search problem which is statistically easy but computationally hard. Empirically, we find that a variety of neural networks successfully learn sparse parities, with discontinuous phase transitions in the training curves. On small instances, learning abruptly occurs at approximately $n^{O(k)}$ iterations; this nearly matches SQ lower bounds, despite the apparent lack of a sparse prior. Our theoretical analysis shows that these observations are not explained by a Langevin-like mechanism, whereby SGD \"stumbles in the dark\" until it finds the hidden set of features (a natural algorithm which also runs in $n^{O(k)}$ time). Instead, we show that SGD gradually amplifies the sparse solution via a Fourier gap in the population gradient, making continual progress that is invisible to loss and error metrics."
                    },
                    {
                        "title": "Pareto Frontiers in Neural Feature Learning: Data, Compute, Width, and Luck",
                        "abstract": "In modern deep learning, algorithmic choices (such as width, depth, and learning rate) are known to modulate nuanced resource tradeoffs. This work investigates how these complexities necessarily arise for feature learning in the presence of computational-statistical gaps. We begin by considering offline sparse parity learning, a supervised classification problem which admits a statistical query lower bound for gradient-based training of a multilayer perceptron. This lower bound can be interpreted as a multi-resource tradeoff frontier: successful learning can only occur if one is sufficiently rich (large model), knowledgeable (large dataset), patient (many training iterations), or lucky (many random guesses). We show, theoretically and experimentally, that sparse initialization and increasing network width yield significant improvements in sample efficiency in this setting. Here, width plays the role of parallel search: it amplifies the probability of finding \"lottery ticket\" neurons, which learn sparse features more sample-efficiently. Finally, we show that the synthetic sparse parity task can be useful as a proxy for real problems requiring axis-aligned feature learning. We demonstrate improved sample efficiency on tabular classification benchmarks by using wide, sparsely-initialized MLP models; these networks sometimes outperform tuned random forests."
                    }
                ]
            },
            "94a17269-78e6-4478-a2bd-737f8b9122d2": {
                "pk": "94a17269-78e6-4478-a2bd-737f8b9122d2",
                "name": "Milind Tambe",
                "collaborators": [
                    "Bryan Wilder",
                    "Arunesh Sinha",
                    "Bistra Dilkina",
                    "Andrew Perrault",
                    "Aparna Taneja",
                    "Fei Fang",
                    "Kai Wang",
                    "Lily Xu",
                    "Sanket Shah",
                    "Niclas Boehmer"
                ],
                "domain": [
                    "Game Theory",
                    "Machine Learning",
                    "Multi-Agent Systems",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Practical Scalability for Stackelberg Security Games",
                        "abstract": "Stackelberg Security Games (SSGs) have been adopted widely for modeling adversarial interactions. With increasing size of the applications of SSGs, scalability of equilibrium computation is an important research problem. While prior research has made progress with regards to scalability, many real world problems cannot be solved satisfactorily yet as per current requirements; these include the deployed federal air marshals (FAMS) application and the threat screening (TSG) problem at airports. Further, these problem domains are inherently limited by NP hardness shown in prior literature. We initiate a principled study of approximations in zero-sum SSGs. Our contribution includes the following: (1) a unified model of SSGs called adversarial randomized allocation (ARA) games that allows studying most SSGs in one model, (2) hardness of approximation results for zero-sum ARA, as well as for the sub-problem of allocating federal air marshal (FAMS) and threat screening problem (TSG) at airports, (3) an approximation framework for zero-sum ARA with provable approximation guarantees and (4) experiments demonstrating the significant scalability of up to 1000x improvement in runtime with an acceptable 5% solution quality loss."
                    },
                    {
                        "title": "End to end learning and optimization on graphs",
                        "abstract": "Real-world applications often combine learning and optimization problems on graphs. For instance, our objective may be to cluster the graph in order to detect meaningful communities (or solve other common graph optimization problems such as facility location, maxcut, and so on). However, graphs or related attributes are often only partially observed, introducing learning problems such as link prediction which must be solved prior to optimization. Standard approaches treat learning and optimization entirely separately, while recent machine learning work aims to predict the optimal solution directly from the inputs. Here, we propose an alternative decision-focused learning approach that integrates a differentiable proxy for common graph optimization problems as a layer in learned systems. The main idea is to learn a representation that maps the original optimization problem onto a simpler proxy problem that can be efficiently differentiated through. Experimental results show that our ClusterNet system outperforms both pure end-to-end approaches (that directly predict the optimal solution) and standard approaches that entirely separate learning and optimization. Code for our system is available at https://github.com/bwilder0/clusternet."
                    },
                    {
                        "title": "AI for Social Impact: Learning and Planning in the Data-to-Deployment Pipeline",
                        "abstract": "With the maturing of AI and multiagent systems research, we have a tremendous opportunity to direct these advances towards addressing complex societal problems. In pursuit of this goal of AI for Social Impact, we as AI researchers must go beyond improvements in computational methodology; it is important to step out in the field to demonstrate social impact. To this end, we focus on the problems of public safety and security, wildlife conservation, and public health in low-resource communities, and present research advances in multiagent systems to address one key cross-cutting challenge: how to effectively deploy our limited intervention resources in these problem domains. We present case studies from our deployments around the world as well as lessons learned that we hope are of use to researchers who are interested in AI for Social Impact. In pushing this research agenda, we believe AI can indeed play an important role in fighting social injustice and improving society."
                    },
                    {
                        "title": "Optimistic Whittle Index Policy: Online Learning for Restless Bandits",
                        "abstract": "Restless multi-armed bandits (RMABs) extend multi-armed bandits to allow for stateful arms, where the state of each arm evolves restlessly with different transitions depending on whether that arm is pulled. Solving RMABs requires information on transition dynamics, which are often unknown upfront. To plan in RMAB settings with unknown transitions, we propose the first online learning algorithm based on the Whittle index policy, using an upper confidence bound (UCB) approach to learn transition dynamics. Specifically, we estimate confidence bounds of the transition probabilities and formulate a bilinear program to compute optimistic Whittle indices using these estimates. Our algorithm, UCWhittle, achieves sublinear $O(H \\sqrt{T \\log T})$ frequentist regret to solve RMABs with unknown transitions in $T$ episodes with a constant horizon $H$. Empirically, we demonstrate that UCWhittle leverages the structure of RMABs and the Whittle index policy solution to achieve better performance than existing online learning baselines across three domains, including one constructed from a real-world maternal and childcare dataset."
                    },
                    {
                        "title": "Improved Policy Evaluation for Randomized Trials of Algorithmic Resource Allocation",
                        "abstract": "We consider the task of evaluating policies of algorithmic resource allocation through randomized controlled trials (RCTs). Such policies are tasked with optimizing the utilization of limited intervention resources, with the goal of maximizing the benefits derived. Evaluation of such allocation policies through RCTs proves difficult, notwithstanding the scale of the trial, because the individuals' outcomes are inextricably interlinked through resource constraints controlling the policy decisions. Our key contribution is to present a new estimator leveraging our proposed novel concept, that involves retrospective reshuffling of participants across experimental arms at the end of an RCT. We identify conditions under which such reassignments are permissible and can be leveraged to construct counterfactual trials, whose outcomes can be accurately ascertained, for free. We prove theoretically that such an estimator is more accurate than common estimators based on sample means -- we show that it returns an unbiased estimate and simultaneously reduces variance. We demonstrate the value of our approach through empirical experiments on synthetic, semi-synthetic as well as real case study data and show improved estimation accuracy across the board."
                    },
                    {
                        "title": "Leaving the Nest: Going Beyond Local Loss Functions for Predict-Then-Optimize",
                        "abstract": "Predict-then-Optimize is a framework for using machine learning to perform decision-making under uncertainty. The central research question it asks is, \"How can the structure of a decision-making task be used to tailor ML models for that specific task?\" To this end, recent work has proposed learning task-specific loss functions that capture this underlying structure. However, current approaches make restrictive assumptions about the form of these losses and their impact on ML model behavior. These assumptions both lead to approaches with high computational cost, and when they are violated in practice, poor performance. In this paper, we propose solutions to these issues, avoiding the aforementioned assumptions and utilizing the ML model's features to increase the sample efficiency of learning loss functions. We empirically show that our method achieves state-of-the-art results in four domains from the literature, often requiring an order of magnitude fewer samples than comparable methods from past work. Moreover, our approach outperforms the best existing method by nearly 200% when the localness assumption is broken."
                    },
                    {
                        "title": "Escape Sensing Games: Detection-vs-Evasion in Security Applications",
                        "abstract": "Traditional game-theoretic research for security applications primarily focuses on the allocation of external protection resources to defend targets. This work puts forward the study of a new class of games centered around strategically arranging targets to protect them against a constrained adversary, with motivations from varied domains such as peacekeeping resource transit and cybersecurity. Specifically, we introduce Escape Sensing Games (ESGs). In ESGs, a blue player manages the order in which targets pass through a channel, while her opponent tries to capture the targets using a set of sensors that need some time to recharge after each activation. We present a thorough computational study of ESGs. Among others, we show that it is NP-hard to compute best responses and equilibria. Nevertheless, we propose a variety of effective (heuristic) algorithms whose quality we demonstrate in extensive computational experiments."
                    },
                    {
                        "title": "Balancing Act: Prioritization Strategies for LLM-Designed Restless Bandit Rewards",
                        "abstract": "LLMs are increasingly used to design reward functions based on human preferences in Reinforcement Learning (RL). We focus on LLM-designed rewards for Restless Multi-Armed Bandits, a framework for allocating limited resources among agents. In applications such as public health, this approach empowers grassroots health workers to tailor automated allocation decisions to community needs. In the presence of multiple agents, altering the reward function based on human preferences can impact subpopulations very differently, leading to complex tradeoffs and a multi-objective resource allocation problem. We are the first to present a principled method termed Social Choice Language Model for dealing with these tradeoffs for LLM-designed rewards for multiagent planners in general and restless bandits in particular. The novel part of our model is a transparent and configurable selection component, called an adjudicator, external to the LLM that controls complex tradeoffs via a user-selected social welfare function. Our experiments demonstrate that our model reliably selects more effective, aligned, and balanced reward functions compared to purely LLM-based approaches."
                    },
                    {
                        "title": "Learning Adversary Behavior in Security Games: A PAC Model Perspective",
                        "abstract": "Recent applications of Stackelberg Security Games (SSG), from wildlife crime to urban crime, have employed machine learning tools to learn and predict adversary behavior using available data about defender-adversary interactions. Given these recent developments, this paper commits to an approach of directly learning the response function of the adversary. Using the PAC model, this paper lays a firm theoretical foundation for learning in SSGs (e.g., theoretically answer questions about the numbers of samples required to learn adversary behavior) and provides utility guarantees when the learned adversary model is used to plan the defender's strategy. The paper also aims to answer practical questions such as how much more data is needed to improve an adversary model's accuracy. Additionally, we explain a recently observed phenomenon that prediction accuracy of learned adversary behavior is not enough to discover the utility maximizing defender strategy. We provide four main contributions: (1) a PAC model of learning adversary response functions in SSGs; (2) PAC-model analysis of the learning of key, existing bounded rationality models in SSGs; (3) an entirely new approach to adversary modeling based on a non-parametric class of response functions with PAC-model analysis and (4) identification of conditions under which computing the best defender strategy against the learned adversary behavior is indeed the optimal strategy. Finally, we conduct experiments with real-world data from a national park in Uganda, showing the benefit of our new adversary modeling approach and verification of our PAC model predictions."
                    },
                    {
                        "title": "Melding the Data-Decisions Pipeline: Decision-Focused Learning for Combinatorial Optimization",
                        "abstract": "Creating impact in real-world settings requires artificial intelligence techniques to span the full pipeline from data, to predictive models, to decisions. These components are typically approached separately: a machine learning model is first trained via a measure of predictive accuracy, and then its predictions are used as input into an optimization algorithm which produces a decision. However, the loss function used to train the model may easily be misaligned with the end goal, which is to make the best decisions possible. Hand-tuning the loss function to align with optimization is a difficult and error-prone process (which is often skipped entirely).   We focus on combinatorial optimization problems and introduce a general framework for decision-focused learning, where the machine learning model is directly trained in conjunction with the optimization algorithm to produce high-quality decisions. Technically, our contribution is a means of integrating common classes of discrete optimization problems into deep learning or other predictive models, which are typically trained via gradient descent. The main idea is to use a continuous relaxation of the discrete problem to propagate gradients through the optimization procedure. We instantiate this framework for two broad classes of combinatorial problems: linear programs and submodular maximization. Experimental results across a variety of domains show that decision-focused learning often leads to improved optimization performance compared to traditional methods. We find that standard measures of accuracy are not a reliable proxy for a predictive model's utility in optimization, and our method's ability to specify the true goal as the model's training objective yields substantial dividends across a range of decision problems."
                    },
                    {
                        "title": "MIPaaL: Mixed Integer Program as a Layer",
                        "abstract": "Machine learning components commonly appear in larger decision-making pipelines; however, the model training process typically focuses only on a loss that measures accuracy between predicted values and ground truth values. Decision-focused learning explicitly integrates the downstream decision problem when training the predictive model, in order to optimize the quality of decisions induced by the predictions. It has been successfully applied to several limited combinatorial problem classes, such as those that can be expressed as linear programs (LP), and submodular optimization. However, these previous applications have uniformly focused on problems from specific classes with simple constraints. Here, we enable decision-focused learning for the broad class of problems that can be encoded as a Mixed Integer Linear Program (MIP), hence supporting arbitrary linear constraints over discrete and continuous variables. We show how to differentiate through a MIP by employing a cutting planes solution approach, which is an exact algorithm that iteratively adds constraints to a continuous relaxation of the problem until an integral solution is found. We evaluate our new end-to-end approach on several real world domains and show that it outperforms the standard two phase approaches that treat prediction and prescription separately, as well as a baseline approach of simply applying decision-focused learning to the LP relaxation of the MIP."
                    },
                    {
                        "title": "Automatically Learning Compact Quality-aware Surrogates for Optimization Problems",
                        "abstract": "Solving optimization problems with unknown parameters often requires learning a predictive model to predict the values of the unknown parameters and then solving the problem using these values. Recent work has shown that including the optimization problem as a layer in the model training pipeline results in predictions of the unobserved parameters that lead to higher decision quality. Unfortunately, this process comes at a large computational cost because the optimization problem must be solved and differentiated through in each training iteration; furthermore, it may also sometimes fail to improve solution quality due to non-smoothness issues that arise when training through a complex optimization layer. To address these shortcomings, we learn a low-dimensional surrogate model of a large optimization problem by representing the feasible space in terms of meta-variables, each of which is a linear combination of the original variables. By training a low-dimensional surrogate model end-to-end, and jointly with the predictive model, we achieve: i) a large reduction in training and inference time; and ii) improved performance by focusing attention on the more important variables in the optimization and learning in a smoother space. Empirically, we demonstrate these improvements on a non-convex adversary modeling task, a submodular recommendation task and a convex portfolio optimization task."
                    },
                    {
                        "title": "Tracking disease outbreaks from sparse data with Bayesian inference",
                        "abstract": "The COVID-19 pandemic provides new motivation for a classic problem in epidemiology: estimating the empirical rate of transmission during an outbreak (formally, the time-varying reproduction number) from case counts. While standard methods exist, they work best at coarse-grained national or state scales with abundant data, and struggle to accommodate the partial observability and sparse data common at finer scales (e.g., individual schools or towns). For example, case counts may be sparse when only a small fraction of infections are caught by a testing program. Or, whether an infected individual tests positive may depend on the kind of test and the point in time when they are tested. We propose a Bayesian framework which accommodates partial observability in a principled manner. Our model places a Gaussian process prior over the unknown reproduction number at each time step and models observations sampled from the distribution of a specific testing program. For example, our framework can accommodate a variety of kinds of tests (viral RNA, antibody, antigen, etc.) and sampling schemes (e.g., longitudinal or cross-sectional screening). Inference in this framework is complicated by the presence of tens or hundreds of thousands of discrete latent variables. To address this challenge, we propose an efficient stochastic variational inference method which relies on a novel gradient estimator for the variational objective. Experimental results for an example motivated by COVID-19 show that our method produces an accurate and well-calibrated posterior, while standard methods for estimating the reproduction number can fail badly."
                    },
                    {
                        "title": "Learn to Intervene: An Adaptive Learning Policy for Restless Bandits in Application to Preventive Healthcare",
                        "abstract": "In many public health settings, it is important for patients to adhere to health programs, such as taking medications and periodic health checks. Unfortunately, beneficiaries may gradually disengage from such programs, which is detrimental to their health. A concrete example of gradual disengagement has been observed by an organization that carries out a free automated call-based program for spreading preventive care information among pregnant women. Many women stop picking up calls after being enrolled for a few months. To avoid such disengagements, it is important to provide timely interventions. Such interventions are often expensive and can be provided to only a small fraction of the beneficiaries. We model this scenario as a restless multi-armed bandit (RMAB) problem, where each beneficiary is assumed to transition from one state to another depending on the intervention. Moreover, since the transition probabilities are unknown a priori, we propose a Whittle index based Q-Learning mechanism and show that it converges to the optimal solution. Our method improves over existing learning-based methods for RMABs on multiple benchmarks from literature and also on the maternal healthcare dataset."
                    },
                    {
                        "title": "Ranked Prioritization of Groups in Combinatorial Bandit Allocation",
                        "abstract": "Preventing poaching through ranger patrols protects endangered wildlife, directly contributing to the UN Sustainable Development Goal 15 of life on land. Combinatorial bandits have been used to allocate limited patrol resources, but existing approaches overlook the fact that each location is home to multiple species in varying proportions, so a patrol benefits each species to differing degrees. When some species are more vulnerable, we ought to offer more protection to these animals; unfortunately, existing combinatorial bandit approaches do not offer a way to prioritize important species. To bridge this gap, (1) We propose a novel combinatorial bandit objective that trades off between reward maximization and also accounts for prioritization over species, which we call ranked prioritization. We show this objective can be expressed as a weighted linear sum of Lipschitz-continuous reward functions. (2) We provide RankedCUCB, an algorithm to select combinatorial actions that optimize our prioritization-based objective, and prove that it achieves asymptotic no-regret. (3) We demonstrate empirically that RankedCUCB leads to up to 38% improvement in outcomes for endangered species using real-world wildlife conservation data. Along with adapting to other challenges such as preventing illegal logging and overfishing, our no-regret algorithm addresses the general combinatorial bandit problem with a weighted linear objective."
                    },
                    {
                        "title": "Indexability is Not Enough for Whittle: Improved, Near-Optimal Algorithms for Restless Bandits",
                        "abstract": "We study the problem of planning restless multi-armed bandits (RMABs) with multiple actions. This is a popular model for multi-agent systems with applications like multi-channel communication, monitoring and machine maintenance tasks, and healthcare. Whittle index policies, which are based on Lagrangian relaxations, are widely used in these settings due to their simplicity and near-optimality under certain conditions. In this work, we first show that Whittle index policies can fail in simple and practically relevant RMAB settings, even when the RMABs are indexable. We discuss why the optimality guarantees fail and why asymptotic optimality may not translate well to practically relevant planning horizons.   We then propose an alternate planning algorithm based on the mean-field method, which can provably and efficiently obtain near-optimal policies with a large number of arms, without the stringent structural assumptions required by the Whittle index policies. This borrows ideas from existing research with some improvements: our approach is hyper-parameter free, and we provide an improved non-asymptotic analysis which has: (a) no requirement for exogenous hyper-parameters and tighter polynomial dependence on known problem parameters; (b) high probability bounds which show that the reward of the policy is reliable; and (c) matching sub-optimality lower bounds for this algorithm with respect to the number of arms, thus demonstrating the tightness of our bounds. Our extensive experimental analysis shows that the mean-field approach matches or outperforms other baselines."
                    },
                    {
                        "title": "Efficient Public Health Intervention Planning Using Decomposition-Based Decision-Focused Learning",
                        "abstract": "The declining participation of beneficiaries over time is a key concern in public health programs. A popular strategy for improving retention is to have health workers `intervene' on beneficiaries at risk of dropping out. However, the availability and time of these health workers are limited resources. As a result, there has been a line of research on optimizing these limited intervention resources using Restless Multi-Armed Bandits (RMABs). The key technical barrier to using this framework in practice lies in the need to estimate the beneficiaries' RMAB parameters from historical data. Recent research has shown that Decision-Focused Learning (DFL), which focuses on maximizing the beneficiaries' adherence rather than predictive accuracy, improves the performance of intervention targeting using RMABs. Unfortunately, these gains come at a high computational cost because of the need to solve and evaluate the RMAB in each DFL training step. In this paper, we provide a principled way to exploit the structure of RMABs to speed up intervention planning by cleverly decoupling the planning for different beneficiaries. We use real-world data from an Indian NGO, ARMMAN, to show that our approach is up to two orders of magnitude faster than the state-of-the-art approach while also yielding superior model performance. This would enable the NGO to scale up deployments using DFL to potentially millions of mothers, ultimately advancing progress toward UNSDG 3.1."
                    },
                    {
                        "title": "Combining Diverse Information for Coordinated Action: Stochastic Bandit Algorithms for Heterogeneous Agents",
                        "abstract": "Stochastic multi-agent multi-armed bandits typically assume that the rewards from each arm follow a fixed distribution, regardless of which agent pulls the arm. However, in many real-world settings, rewards can depend on the sensitivity of each agent to their environment. In medical screening, disease detection rates can vary by test type; in preference matching, rewards can depend on user preferences; and in environmental sensing, observation quality can vary across sensors. Since past work does not specify how to allocate agents of heterogeneous but known sensitivity of these types in a stochastic bandit setting, we introduce a UCB-style algorithm, Min-Width, which aggregates information from diverse agents. In doing so, we address the joint challenges of (i) aggregating the rewards, which follow different distributions for each agent-arm pair, and (ii) coordinating the assignments of agents to arms. Min-Width facilitates efficient collaboration among heterogeneous agents, exploiting the known structure in the agents' reward functions to weight their rewards accordingly. We analyze the regret of Min-Width and conduct pseudo-synthetic and fully synthetic experiments to study the performance of different levels of information sharing. Our results confirm that the gains to modeling agent heterogeneity tend to be greater when the sensitivities are more varied across agents, while combining more information does not always improve performance."
                    },
                    {
                        "title": "Mitigating the Curse of Correlation in Security Games by Entropy Maximization",
                        "abstract": "In Stackelberg security games, a defender seeks to randomly allocate limited security resources to protect critical targets from an attack. In this paper, we study a fundamental, yet underexplored, phenomenon in security games, which we term the \\emph{Curse of Correlation} (CoC). Specifically, we observe that there are inevitable correlations among the protection status of different targets. Such correlation is a crucial concern, especially in \\emph{spatio-temporal} domains like conservation area patrolling, where attackers can surveil patrollers at certain areas and then infer their patrolling routes using such correlations. To mitigate this issue, we propose to design entropy-maximizing defending strategies for spatio-temporal security games, which frequently suffer from CoC. We prove that the problem is \\#P-hard in general. However, it admits efficient algorithms in well-motivated special settings. Our experiments show significant advantages of max-entropy algorithms over previous algorithms. A scalable implementation of our algorithm is currently under pre-deployment testing for integration into FAMS software to improve the scheduling of US federal air marshals."
                    },
                    {
                        "title": "Group-Fairness in Influence Maximization",
                        "abstract": "Influence maximization is a widely used model for information dissemination in social networks. Recent work has employed such interventions across a wide range of social problems, spanning public health, substance abuse, and international development (to name a few examples). A critical but understudied question is whether the benefits of such interventions are fairly distributed across different groups in the population; e.g., avoiding discrimination with respect to sensitive attributes such as race or gender. Drawing on legal and game-theoretic concepts, we introduce formal definitions of fairness in influence maximization. We provide an algorithmic framework to find solutions which satisfy fairness constraints, and in the process improve the state of the art for general multi-objective submodular maximization problems. Experimental results on real data from an HIV prevention intervention for homeless youth show that standard influence maximization techniques oftentimes neglect smaller groups which contribute less to overall utility, resulting in a disparity which our proposed algorithms substantially reduce."
                    }
                ]
            },
            "16b0e8dc-26b9-4548-aec9-d4f3c5bd1b54": {
                "pk": "16b0e8dc-26b9-4548-aec9-d4f3c5bd1b54",
                "name": "Sham M. Kakade",
                "collaborators": [
                    "Daniel Hsu",
                    "Tong Zhang",
                    "Ambuj Tewari",
                    "Qingqing Huang",
                    "Elad Hazan",
                    "Karan Singh",
                    "Shai Shalev-Shwartz",
                    "Ilan Lobel",
                    "Hamid Nazerzadeh",
                    "Animashree Anandkumar"
                ],
                "domain": [
                    "Statistical Learning",
                    "Machine Learning",
                    "Optimization",
                    "Random Matrix Theory"
                ],
                "publications": [
                    {
                        "title": "A tail inequality for quadratic forms of subgaussian random vectors",
                        "abstract": "We prove an exponential probability tail inequality for positive semidefinite quadratic forms in a subgaussian random vector. The bound is analogous to one that holds when the vector has independent Gaussian entries."
                    },
                    {
                        "title": "Dimension-free tail inequalities for sums of random matrices",
                        "abstract": "We derive exponential tail inequalities for sums of random matrices with no dependence on the explicit matrix dimensions. These are similar to the matrix versions of the Chernoff bound and Bernstein inequality except with the explicit matrix dimensions replaced by a trace quantity that can be small even when the dimension is large or infinite. Some applications to principal component analysis and approximate matrix multiplication are given to illustrate the utility of the new bounds."
                    },
                    {
                        "title": "Learning mixtures of spherical Gaussians: moment methods and spectral decompositions",
                        "abstract": "This work provides a computationally efficient and statistically consistent moment-based estimator for mixtures of spherical Gaussians. Under the condition that component means are in general position, a simple spectral decomposition technique yields consistent parameter estimates from low-order observable moments, without additional minimum separation assumptions needed by previous computationally efficient estimation procedures. Thus computational and information-theoretic barriers to efficient estimation in mixture models are precluded when the mixture components have means in general position and spherical covariances. Some connections are made to estimation problems related to independent component analysis."
                    },
                    {
                        "title": "The Nonstochastic Control Problem",
                        "abstract": "We consider the problem of controlling an unknown linear dynamical system in the presence of (nonstochastic) adversarial perturbations and adversarial convex loss functions. In contrast to classical control, the a priori determination of an optimal controller here is hindered by the latter's dependence on the yet unknown perturbations and costs. Instead, we measure regret against an optimal linear policy in hindsight, and give the first efficient algorithm that guarantees a sublinear regret bound, scaling as T^{2/3}, in this setting."
                    },
                    {
                        "title": "Analysis of a randomized approximation scheme for matrix multiplication",
                        "abstract": "This note gives a simple analysis of a randomized approximation scheme for matrix multiplication proposed by Sarlos (2006) based on a random rotation followed by uniform column sampling. The result follows from a matrix version of Bernstein's inequality and a tail inequality for quadratic forms in subgaussian random vectors."
                    },
                    {
                        "title": "Regularization Techniques for Learning with Matrices",
                        "abstract": "There is growing body of learning problems for which it is natural to organize the parameters into matrix, so as to appropriately regularize the parameters under some matrix norm (in order to impose some more sophisticated prior knowledge). This work describes and analyzes a systematic method for constructing such matrix-based, regularization methods. In particular, we focus on how the underlying statistical properties of a given problem can help us decide which regularization function is appropriate.   Our methodology is based on the known duality fact: that a function is strongly convex with respect to some norm if and only if its conjugate function is strongly smooth with respect to the dual norm. This result has already been found to be a key component in deriving and analyzing several learning algorithms. We demonstrate the potential of this framework by deriving novel generalization and regret bounds for multi-task learning, multi-class learning, and kernel learning."
                    },
                    {
                        "title": "An Optimal Dynamic Mechanism for Multi-Armed Bandit Processes",
                        "abstract": "We consider the problem of revenue-optimal dynamic mechanism design in settings where agents' types evolve over time as a function of their (both public and private) experience with items that are auctioned repeatedly over an infinite horizon. A central question here is understanding what natural restrictions on the environment permit the design of optimal mechanisms (note that even in the simpler static setting, optimal mechanisms are characterized only under certain restrictions). We provide a {\\em structural characterization} of a natural \"separable: multi-armed bandit environment (where the evolution and incentive structure of the a-priori type is decoupled from the subsequent experience in a precise sense) where dynamic optimal mechanism design is possible. Here, we present the Virtual Index Mechanism, an optimal dynamic mechanism, which maximizes the (long term) {\\em virtual surplus} using the classical Gittins algorithm. The mechanism optimally balances exploration and exploitation, taking incentives into account."
                    },
                    {
                        "title": "Robust Matrix Decomposition with Outliers",
                        "abstract": "Suppose a given observation matrix can be decomposed as the sum of a low-rank matrix and a sparse matrix (outliers), and the goal is to recover these individual components from the observed sum. Such additive decompositions have applications in a variety of numerical problems including system identification, latent variable graphical modeling, and principal components analysis. We study conditions under which recovering such a decomposition is possible via a combination of $\\ell_1$ norm and trace norm minimization. We are specifically interested in the question of how many outliers are allowed so that convex programming can still achieve accurate recovery, and we obtain stronger recovery guarantees than previous studies. Moreover, we do not assume that the spatial pattern of outliers is random, which stands in contrast to related analyses under such assumptions via matrix completion."
                    },
                    {
                        "title": "Random design analysis of ridge regression",
                        "abstract": "This work gives a simultaneous analysis of both the ordinary least squares estimator and the ridge regression estimator in the random design setting under mild assumptions on the covariate/response distributions. In particular, the analysis provides sharp results on the ``out-of-sample'' prediction error, as opposed to the ``in-sample'' (fixed design) error. The analysis also reveals the effect of errors in the estimated covariance structure, as well as the effect of modeling errors, neither of which effects are present in the fixed design setting. The proofs of the main results are based on a simple decomposition lemma combined with concentration inequalities for random vectors and matrices."
                    },
                    {
                        "title": "A Method of Moments for Mixture Models and Hidden Markov Models",
                        "abstract": "Mixture models are a fundamental tool in applied statistics and machine learning for treating data taken from multiple subpopulations. The current practice for estimating the parameters of such models relies on local search heuristics (e.g., the EM algorithm) which are prone to failure, and existing consistent methods are unfavorable due to their high computational and sample complexity which typically scale exponentially with the number of mixture components. This work develops an efficient method of moments approach to parameter estimation for a broad class of high-dimensional mixture models with many components, including multi-view mixtures of Gaussians (such as mixtures of axis-aligned Gaussians) and hidden Markov models. The new method leads to rigorous unsupervised learning results for mixture models that were not achieved by previous works; and, because of its simplicity, it offers a viable alternative to EM for practical deployment."
                    },
                    {
                        "title": "Planning in POMDPs Using Multiplicity Automata",
                        "abstract": "Planning and learning in Partially Observable MDPs (POMDPs) are among the most challenging tasks in both the AI and Operation Research communities. Although solutions to these problems are intractable in general, there might be special cases, such as structured POMDPs, which can be solved efficiently. A natural and possibly efficient way to represent a POMDP is through the predictive state representation (PSR) - a representation which recently has been receiving increasing attention. In this work, we relate POMDPs to multiplicity automata- showing that POMDPs can be represented by multiplicity automata with no increase in the representation size. Furthermore, we show that the size of the multiplicity automaton is equal to the rank of the predictive state representation. Therefore, we relate both the predictive state representation and POMDPs to the well-founded multiplicity automata literature. Based on the multiplicity automata representation, we provide a planning algorithm which is exponential only in the multiplicity automata rank rather than the number of states of the POMDP. As a result, whenever the predictive state representation is logarithmic in the standard POMDP representation, our planning algorithm is efficient."
                    },
                    {
                        "title": "Robust Aggregation for Federated Learning",
                        "abstract": "Federated learning is the centralized training of statistical models from decentralized data on mobile devices while preserving the privacy of each device. We present a robust aggregation approach to make federated learning robust to settings when a fraction of the devices may be sending corrupted updates to the server. The approach relies on a robust aggregation oracle based on the geometric median, which returns a robust aggregate using a constant number of iterations of a regular non-robust averaging oracle. The robust aggregation oracle is privacy-preserving, similar to the non-robust secure average oracle it builds upon. We establish its convergence for least squares estimation of additive models. We provide experimental results with linear models and deep networks for three tasks in computer vision and natural language processing. The robust aggregation approach is agnostic to the level of corruption; it outperforms the classical aggregation approach in terms of robustness when the level of corruption is high, while being competitive in the regime of low corruption. Two variants, a faster one with one-step robust aggregation and another one with on-device personalization, round off the paper."
                    },
                    {
                        "title": "A Spectral Algorithm for Learning Hidden Markov Models",
                        "abstract": "Hidden Markov Models (HMMs) are one of the most fundamental and widely used statistical tools for modeling discrete time series. In general, learning HMMs from data is computationally hard (under cryptographic assumptions), and practitioners typically resort to search heuristics which suffer from the usual local optima issues. We prove that under a natural separation condition (bounds on the smallest singular value of the HMM parameters), there is an efficient and provably correct algorithm for learning HMMs. The sample complexity of the algorithm does not explicitly depend on the number of distinct (discrete) observations---it implicitly depends on this quantity through spectral properties of the underlying HMM. This makes the algorithm particularly applicable to settings with a large number of observations, such as those in natural language processing where the space of observation is sometimes the words in a language. The algorithm is also simple, employing only a singular value decomposition and matrix multiplications."
                    },
                    {
                        "title": "Learning Mixtures of Gaussians in High Dimensions",
                        "abstract": "Efficiently learning mixture of Gaussians is a fundamental problem in statistics and learning theory. Given samples coming from a random one out of k Gaussian distributions in Rn, the learning problem asks to estimate the means and the covariance matrices of these Gaussians. This learning problem arises in many areas ranging from the natural sciences to the social sciences, and has also found many machine learning applications. Unfortunately, learning mixture of Gaussians is an information theoretically hard problem: in order to learn the parameters up to a reasonable accuracy, the number of samples required is exponential in the number of Gaussian components in the worst case. In this work, we show that provided we are in high enough dimensions, the class of Gaussian mixtures is learnable in its most general form under a smoothed analysis framework, where the parameters are randomly perturbed from an adversarial starting point. In particular, given samples from a mixture of Gaussians with randomly perturbed parameters, when n > {\\Omega}(k^2), we give an algorithm that learns the parameters with polynomial running time and using polynomial number of samples. The central algorithmic ideas consist of new ways to decompose the moment tensor of the Gaussian mixture by exploiting its structural properties. The symmetries of this tensor are derived from the combinatorial structure of higher order moments of Gaussian distributions (sometimes referred to as Isserlis' theorem or Wick's theorem). We also develop new tools for bounding smallest singular values of structured random matrices, which could be useful in other smoothed analysis settings."
                    },
                    {
                        "title": "Provable Efficient Online Matrix Completion via Non-convex Stochastic Gradient Descent",
                        "abstract": "Matrix completion, where we wish to recover a low rank matrix by observing a few entries from it, is a widely studied problem in both theory and practice with wide applications. Most of the provable algorithms so far on this problem have been restricted to the offline setting where they provide an estimate of the unknown matrix using all observations simultaneously. However, in many applications, the online version, where we observe one entry at a time and dynamically update our estimate, is more appealing. While existing algorithms are efficient for the offline setting, they could be highly inefficient for the online setting.   In this paper, we propose the first provable, efficient online algorithm for matrix completion. Our algorithm starts from an initial estimate of the matrix and then performs non-convex stochastic gradient descent (SGD). After every observation, it performs a fast update involving only one row of two tall matrices, giving near linear total runtime. Our algorithm can be naturally used in the offline setting as well, where it gives competitive sample complexity and runtime to state of the art algorithms. Our proofs introduce a general framework to show that SGD updates tend to stay away from saddle surfaces and could be of broader interests for other non-convex problems to prove tight rates."
                    },
                    {
                        "title": "Towards minimax policies for online linear optimization with bandit feedback",
                        "abstract": "We address the online linear optimization problem with bandit feedback. Our contribution is twofold. First, we provide an algorithm (based on exponential weights) with a regret of order $\\sqrt{d n \\log N}$ for any finite action set with $N$ actions, under the assumption that the instantaneous loss is bounded by 1. This shaves off an extraneous $\\sqrt{d}$ factor compared to previous works, and gives a regret bound of order $d \\sqrt{n \\log n}$ for any compact set of actions. Without further assumptions on the action set, this last bound is minimax optimal up to a logarithmic factor. Interestingly, our result also shows that the minimax regret for bandit linear optimization with expert advice in $d$ dimension is the same as for the basic $d$-armed bandit with expert advice. Our second contribution is to show how to use the Mirror Descent algorithm to obtain computationally efficient strategies with minimax optimal regret bounds in specific examples. More precisely we study two canonical action sets: the hypercube and the Euclidean ball. In the former case, we obtain the first computationally efficient algorithm with a $d \\sqrt{n}$ regret, thus improving by a factor $\\sqrt{d \\log n}$ over the best known result for a computationally efficient algorithm. In the latter case, our approach gives the first algorithm with a $\\sqrt{d n \\log n}$ regret, again shaving off an extraneous $\\sqrt{d}$ compared to previous works."
                    },
                    {
                        "title": "Super-Resolution Off the Grid",
                        "abstract": "Super-resolution is the problem of recovering a superposition of point sources using bandlimited measurements, which may be corrupted with noise. This signal processing problem arises in numerous imaging problems, ranging from astronomy to biology to spectroscopy, where it is common to take (coarse) Fourier measurements of an object. Of particular interest is in obtaining estimation procedures which are robust to noise, with the following desirable statistical and computational properties: we seek to use coarse Fourier measurements (bounded by some cutoff frequency); we hope to take a (quantifiably) small number of measurements; we desire our algorithm to run quickly.   Suppose we have k point sources in d dimensions, where the points are separated by at least \\Delta from each other (in Euclidean distance). This work provides an algorithm with the following favorable guarantees: - The algorithm uses Fourier measurements, whose frequencies are bounded by O(1/\\Delta) (up to log factors). Previous algorithms require a cutoff frequency which may be as large as {\\Omega}( d/\\Delta). - The number of measurements taken by and the computational complexity of our algorithm are bounded by a polynomial in both the number of points k and the dimension d, with no dependence on the separation \\Delta. In contrast, previous algorithms depended inverse polynomially on the minimal separation and exponentially on the dimension for both of these quantities.   Our estimation procedure itself is simple: we take random bandlimited measurements (as opposed to taking an exponential number of measurements on the hyper-grid). Furthermore, our analysis and algorithm are elementary (based on concentration bounds for sampling and the singular value decomposition)."
                    },
                    {
                        "title": "Learning Exponential Families in High-Dimensions: Strong Convexity and Sparsity",
                        "abstract": "The versatility of exponential families, along with their attendant convexity properties, make them a popular and effective statistical model. A central issue is learning these models in high-dimensions, such as when there is some sparsity pattern of the optimal parameter. This work characterizes a certain strong convexity property of general exponential families, which allow their generalization ability to be quantified. In particular, we show how this property can be used to analyze generic exponential families under L_1 regularization."
                    },
                    {
                        "title": "Identifiability and Unmixing of Latent Parse Trees",
                        "abstract": "This paper explores unsupervised learning of parsing models along two directions. First, which models are identifiable from infinite data? We use a general technique for numerically checking identifiability based on the rank of a Jacobian matrix, and apply it to several standard constituency and dependency parsing models. Second, for identifiable models, how do we estimate the parameters efficiently? EM suffers from local optima, while recent work using spectral methods cannot be directly applied since the topology of the parse tree varies across sentences. We develop a strategy, unmixing, which deals with this additional complexity for restricted classes of parsing models."
                    },
                    {
                        "title": "An Exponential Lower Bound for Linearly-Realizable MDPs with Constant Suboptimality Gap",
                        "abstract": "A fundamental question in the theory of reinforcement learning is: suppose the optimal $Q$-function lies in the linear span of a given $d$ dimensional feature mapping, is sample-efficient reinforcement learning (RL) possible? The recent and remarkable result of Weisz et al. (2020) resolved this question in the negative, providing an exponential (in $d$) sample size lower bound, which holds even if the agent has access to a generative model of the environment. One may hope that this information theoretic barrier for RL can be circumvented by further supposing an even more favorable assumption: there exists a \\emph{constant suboptimality gap} between the optimal $Q$-value of the best action and that of the second-best action (for all states). The hope is that having a large suboptimality gap would permit easier identification of optimal actions themselves, thus making the problem tractable; indeed, provided the agent has access to a generative model, sample-efficient RL is in fact possible with the addition of this more favorable assumption.   This work focuses on this question in the standard online reinforcement learning setting, where our main result resolves this question in the negative: our hardness result shows that an exponential sample complexity lower bound still holds even if a constant suboptimality gap is assumed in addition to having a linearly realizable optimal $Q$-function. Perhaps surprisingly, this implies an exponential separation between the online RL setting and the generative model setting. Complementing our negative hardness result, we give two positive results showing that provably sample-efficient RL is possible either under an additional low-variance assumption or under a novel hypercontractivity assumption (both implicitly place stronger conditions on the underlying dynamics model)."
                    }
                ]
            },
            "a0a87256-f715-4392-835e-7cccf7a25777": {
                "pk": "a0a87256-f715-4392-835e-7cccf7a25777",
                "name": "Eran Malach",
                "collaborators": [
                    "Shai Shalev-Shwartz",
                    "Amit Daniely",
                    "Alon Brutzkus",
                    "Gilad Yehudai",
                    "Ohad Shamir",
                    "David Brandfonbrener",
                    "Samy Jelassi",
                    "Gal Kaplun",
                    "Pritish Kamath",
                    "Emmanuel Abbe"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Language Models",
                    "Theoretical Computer Science"
                ],
                "publications": [
                    {
                        "title": "Auto-Regressive Next-Token Predictors are Universal Learners",
                        "abstract": "Large language models display remarkable capabilities in logical and mathematical reasoning, allowing them to solve complex tasks. Interestingly, these abilities emerge in networks trained on the simple task of next-token prediction. In this work, we present a theoretical framework for studying auto-regressive next-token predictors. We demonstrate that even simple models such as linear next-token predictors, trained on Chain-of-Thought (CoT) data, can approximate any function efficiently computed by a Turing machine. We introduce a new complexity measure -- length complexity -- which measures the number of intermediate tokens in a CoT sequence required to approximate some target function, and analyze the interplay between length complexity and other notions of complexity. Finally, we show experimentally that simple next-token predictors, such as linear networks and shallow Multi-Layer Perceptrons (MLPs), display non-trivial performance on text generation and arithmetic tasks. Our results demonstrate that the power of today's LLMs can be attributed, to a great extent, to the auto-regressive next-token training scheme, and not necessarily to a particular choice of architecture."
                    },
                    {
                        "title": "Computational Separation Between Convolutional and Fully-Connected Networks",
                        "abstract": "Convolutional neural networks (CNN) exhibit unmatched performance in a multitude of computer vision tasks. However, the advantage of using convolutional networks over fully-connected networks is not understood from a theoretical perspective. In this work, we show how convolutional networks can leverage locality in the data, and thus achieve a computational advantage over fully-connected networks. Specifically, we show a class of problems that can be efficiently solved using convolutional networks trained with gradient-descent, but at the same time is hard to learn using a polynomial-size fully-connected network."
                    },
                    {
                        "title": "Decoupling \"when to update\" from \"how to update\"",
                        "abstract": "Deep learning requires data. A useful approach to obtain data is to be creative and mine data from various sources, that were created for different purposes. Unfortunately, this approach often leads to noisy labels. In this paper, we propose a meta algorithm for tackling the noisy labels problem. The key idea is to decouple \"when to update\" from \"how to update\". We demonstrate the effectiveness of our algorithm by mining data for gender classification by combining the Labeled Faces in the Wild (LFW) face recognition dataset with a textual genderizing service, which leads to a noisy dataset. While our approach is very simple to implement, it leads to state-of-the-art results. We analyze some convergence properties of the proposed algorithm."
                    },
                    {
                        "title": "Is Deeper Better only when Shallow is Good?",
                        "abstract": "Understanding the power of depth in feed-forward neural networks is an ongoing challenge in the field of deep learning theory. While current works account for the importance of depth for the expressive power of neural-networks, it remains an open question whether these benefits are exploited during a gradient-based optimization process. In this work we explore the relation between expressivity properties of deep networks and the ability to train them efficiently using gradient-based algorithms. We give a depth separation argument for distributions with fractal structure, showing that they can be expressed efficiently by deep networks, but not with shallow ones. These distributions have a natural coarse-to-fine structure, and we show that the balance between the coarse and fine details has a crucial effect on whether the optimization process is likely to succeed. We prove that when the distribution is concentrated on the fine details, gradient-based algorithms are likely to fail. Using this result we prove that, at least in some distributions, the success of learning deep networks depends on whether the distribution can be well approximated by shallower networks, and we conjecture that this property holds in general."
                    },
                    {
                        "title": "Learning Parities with Neural Networks",
                        "abstract": "In recent years we see a rapidly growing line of research which shows learnability of various models via common neural network algorithms. Yet, besides a very few outliers, these results show learnability of models that can be learned using linear methods. Namely, such results show that learning neural-networks with gradient-descent is competitive with learning a linear classifier on top of a data-independent representation of the examples. This leaves much to be desired, as neural networks are far more successful than linear methods. Furthermore, on the more conceptual level, linear models don't seem to capture the \"deepness\" of deep networks. In this paper we make a step towards showing leanability of models that are inherently non-linear. We show that under certain distributions, sparse parities are learnable via gradient decent on depth-two network. On the other hand, under the same distributions, these parities cannot be learned efficiently by linear methods."
                    },
                    {
                        "title": "When Hardness of Approximation Meets Hardness of Learning",
                        "abstract": "A supervised learning algorithm has access to a distribution of labeled examples, and needs to return a function (hypothesis) that correctly labels the examples. The hypothesis of the learner is taken from some fixed class of functions (e.g., linear classifiers, neural networks etc.). A failure of the learning algorithm can occur due to two possible reasons: wrong choice of hypothesis class (hardness of approximation), or failure to find the best function within the hypothesis class (hardness of learning). Although both approximation and learnability are important for the success of the algorithm, they are typically studied separately. In this work, we show a single hardness property that implies both hardness of approximation using linear classes and shallow networks, and hardness of learning using correlation queries and gradient-descent. This allows us to obtain new results on hardness of approximation and learnability of parity functions, DNF formulas and $AC^0$ circuits."
                    },
                    {
                        "title": "A Provably Correct Algorithm for Deep Learning that Actually Works",
                        "abstract": "We describe a layer-by-layer algorithm for training deep convolutional networks, where each step involves gradient updates for a two layer network followed by a simple clustering algorithm. Our algorithm stems from a deep generative model that generates mages level by level, where lower resolution images correspond to latent semantic classes. We analyze the convergence rate of our algorithm assuming that the data is indeed generated according to this model (as well as additional assumptions). While we do not pretend to claim that the assumptions are realistic for natural images, we do believe that they capture some true properties of real data. Furthermore, we show that our algorithm actually works in practice (on the CIFAR dataset), achieving results in the same ballpark as that of vanilla convolutional neural networks that are being trained by stochastic gradient descent. Finally, our proof techniques may be of independent interest."
                    },
                    {
                        "title": "Learning Boolean Circuits with Neural Networks",
                        "abstract": "While on some natural distributions, neural-networks are trained efficiently using gradient-based algorithms, it is known that learning them is computationally hard in the worst-case. To separate hard from easy to learn distributions, we observe the property of local correlation: correlation between local patterns of the input and the target label. We focus on learning deep neural-networks using a gradient-based algorithm, when the target function is a tree-structured Boolean circuit. We show that in this case, the existence of correlation between the gates of the circuit and the target label determines whether the optimization succeeds or fails. Using this result, we show that neural-networks can learn the (log n)-parity problem for most product distributions. These results hint that local correlation may play an important role in separating easy/hard to learn distributions. We also obtain a novel depth separation result, in which we show that a shallow network cannot express some functions, while there exists an efficient gradient-based algorithm that can learn the very same functions using a deep network. The negative expressivity result for shallow networks is obtained by a reduction from results in communication complexity, that may be of independent interest."
                    },
                    {
                        "title": "ID3 Learns Juntas for Smoothed Product Distributions",
                        "abstract": "In recent years, there are many attempts to understand popular heuristics. An example of such a heuristic algorithm is the ID3 algorithm for learning decision trees. This algorithm is commonly used in practice, but there are very few theoretical works studying its behavior. In this paper, we analyze the ID3 algorithm, when the target function is a $k$-Junta, a function that depends on $k$ out of $n$ variables of the input. We prove that when $k = \\log n$, the ID3 algorithm learns in polynomial time $k$-Juntas, in the smoothed analysis model of Kalai & Teng. That is, we show a learnability result when the observed distribution is a \"noisy\" variant of the original distribution."
                    },
                    {
                        "title": "Proving the Lottery Ticket Hypothesis: Pruning is All You Need",
                        "abstract": "The lottery ticket hypothesis (Frankle and Carbin, 2018), states that a randomly-initialized network contains a small subnetwork such that, when trained in isolation, can compete with the performance of the original network. We prove an even stronger hypothesis (as was also conjectured in Ramanujan et al., 2019), showing that for every bounded distribution and every target network with bounded weights, a sufficiently over-parameterized neural network with random weights contains a subnetwork with roughly the same accuracy as the target network, without any further training."
                    },
                    {
                        "title": "Quantifying the Benefit of Using Differentiable Learning over Tangent Kernels",
                        "abstract": "We study the relative power of learning with gradient descent on differentiable models, such as neural networks, versus using the corresponding tangent kernels. We show that under certain conditions, gradient descent achieves small error only if a related tangent kernel method achieves a non-trivial advantage over random guessing (a.k.a. weak learning), though this advantage might be very small even when gradient descent can achieve arbitrarily high accuracy. Complementing this, we show that without these conditions, gradient descent can in fact learn with small error even when no kernel method, in particular using the tangent kernel, can achieve a non-trivial advantage over random guessing."
                    },
                    {
                        "title": "Decoupling Gating from Linearity",
                        "abstract": "ReLU neural-networks have been in the focus of many recent theoretical works, trying to explain their empirical success. Nonetheless, there is still a gap between current theoretical results and empirical observations, even in the case of shallow (one hidden-layer) networks. For example, in the task of memorizing a random sample of size $m$ and dimension $d$, the best theoretical result requires the size of the network to be $\\tilde{\\Omega}(\\frac{m^2}{d})$, while empirically a network of size slightly larger than $\\frac{m}{d}$ is sufficient. To bridge this gap, we turn to study a simplified model for ReLU networks. We observe that a ReLU neuron is a product of a linear function with a gate (the latter determines whether the neuron is active or not), where both share a jointly trained weight vector. In this spirit, we introduce the Gated Linear Unit (GaLU), which simply decouples the linearity from the gating by assigning different vectors for each role. We show that GaLU networks allow us to get optimization and generalization results that are much stronger than those available for ReLU networks. Specifically, we show a memorization result for networks of size $\\tilde{\\Omega}(\\frac{m}{d})$, and improved generalization bounds. Finally, we show that in some scenarios, GaLU networks behave similarly to ReLU networks, hence proving to be a good choice of a simplified model."
                    },
                    {
                        "title": "Universal Length Generalization with Turing Programs",
                        "abstract": "Length generalization refers to the ability to extrapolate from short training sequences to long test sequences and is a challenge for current large language models. While prior work has proposed some architecture or data format changes to achieve length generalization, these proposals typically apply to a limited set of tasks. Building on prior scratchpad and Chain-of-Thought (CoT) techniques, we propose Turing Programs, a novel CoT strategy that decomposes an algorithmic task into steps mimicking the computation of a Turing Machine. This framework is both universal, as it can accommodate any algorithmic task, and simple, requiring only copying text from the context with small modifications. We show that by using Turing Programs, we obtain robust length generalization on a range of algorithmic tasks: addition, multiplication and in-context SGD. We then demonstrate that transformers achieve length generalization on random Turing Programs, suggesting that length generalization is possible for any algorithmic task. Finally, we theoretically prove that transformers can implement Turing Programs, constructing a simple RASP (Weiss et al.) program that simulates an arbitrary Turing machine."
                    },
                    {
                        "title": "On the Optimality of Trees Generated by ID3",
                        "abstract": "Since its inception in the 1980s, ID3 has become one of the most successful and widely used algorithms for learning decision trees. However, its theoretical properties remain poorly understood. In this work, we introduce a novel metric of a decision tree algorithm's performance, called mean iteration statistical consistency (MIC), which measures optimality of trees generated by ID3. As opposed to previous metrics, MIC can differentiate between different decision tree algorithms and compare their performance. We provide theoretical and empirical evidence that the TopDown variant of ID3, introduced by Kearns and Mansour (1996), has near-optimal MIC in various settings for learning read-once DNFs under product distributions. In contrast, another widely used variant of ID3 has MIC which is not near-optimal. We show that the MIC analysis predicts well the performance of these algorithms in practice. Our results present a novel view of decision tree algorithms which may lead to better and more practical guarantees for these algorithms."
                    },
                    {
                        "title": "Knowledge Distillation: Bad Models Can Be Good Role Models",
                        "abstract": "Large neural networks trained in the overparameterized regime are able to fit noise to zero train error. Recent work \\citep{nakkiran2020distributional} has empirically observed that such networks behave as \"conditional samplers\" from the noisy distribution. That is, they replicate the noise in the train data to unseen examples. We give a theoretical framework for studying this conditional sampling behavior in the context of learning theory. We relate the notion of such samplers to knowledge distillation, where a student network imitates the outputs of a teacher on unlabeled data. We show that samplers, while being bad classifiers, can be good teachers. Concretely, we prove that distillation from samplers is guaranteed to produce a student which approximates the Bayes optimal classifier. Finally, we show that some common learning algorithms (e.g., Nearest-Neighbours and Kernel Machines) can generate samplers when applied in the overparameterized regime."
                    },
                    {
                        "title": "Corgi^2: A Hybrid Offline-Online Approach To Storage-Aware Data Shuffling For SGD",
                        "abstract": "When using Stochastic Gradient Descent (SGD) for training machine learning models, it is often crucial to provide the model with examples sampled at random from the dataset. However, for large datasets stored in the cloud, random access to individual examples is often costly and inefficient. A recent work \\cite{corgi}, proposed an online shuffling algorithm called CorgiPile, which greatly improves efficiency of data access, at the cost some performance loss, which is particularly apparent for large datasets stored in homogeneous shards (e.g., video datasets). In this paper, we introduce a novel two-step partial data shuffling strategy for SGD which combines an offline iteration of the CorgiPile method with a subsequent online iteration. Our approach enjoys the best of both worlds: it performs similarly to SGD with random access (even for homogenous data) without compromising the data access efficiency of CorgiPile. We provide a comprehensive theoretical analysis of the convergence properties of our method and demonstrate its practical advantages through experimental results."
                    },
                    {
                        "title": "On the Power of Decision Trees in Auto-Regressive Language Modeling",
                        "abstract": "Originally proposed for handling time series data, Auto-regressive Decision Trees (ARDTs) have not yet been explored for language modeling. This paper delves into both the theoretical and practical applications of ARDTs in this new context. We theoretically demonstrate that ARDTs can compute complex functions, such as simulating automata, Turing machines, and sparse circuits, by leveraging \"chain-of-thought\" computations. Our analysis provides bounds on the size, depth, and computational efficiency of ARDTs, highlighting their surprising computational power. Empirically, we train ARDTs on simple language generation tasks, showing that they can learn to generate coherent and grammatically correct text on par with a smaller Transformer model. Additionally, we show that ARDTs can be used on top of transformer representations to solve complex reasoning tasks. This research reveals the unique computational abilities of ARDTs, aiming to broaden the architectural diversity in language model development."
                    },
                    {
                        "title": "The Connection Between Approximation, Depth Separation and Learnability in Neural Networks",
                        "abstract": "Several recent works have shown separation results between deep neural networks, and hypothesis classes with inferior approximation capacity such as shallow networks or kernel classes. On the other hand, the fact that deep networks can efficiently express a target function does not mean that this target function can be learned efficiently by deep neural networks. In this work we study the intricate connection between learnability and approximation capacity. We show that learnability with deep networks of a target function depends on the ability of simpler classes to approximate the target. Specifically, we show that a necessary condition for a function to be learnable by gradient descent on deep neural networks is to be able to approximate the function, at least in a weak sense, with shallow neural networks. We also show that a class of functions can be learned by an efficient statistical query algorithm if and only if it can be approximated in a weak sense by some kernel class. We give several examples of functions which demonstrate depth separation, and conclude that they cannot be efficiently learned, even by a hypothesis class that can efficiently approximate them."
                    },
                    {
                        "title": "SGD Learns Over-parameterized Networks that Provably Generalize on Linearly Separable Data",
                        "abstract": "Neural networks exhibit good generalization behavior in the over-parameterized regime, where the number of network parameters exceeds the number of observations. Nonetheless, current generalization bounds for neural networks fail to explain this phenomenon. In an attempt to bridge this gap, we study the problem of learning a two-layer over-parameterized neural network, when the data is generated by a linearly separable function. In the case where the network has Leaky ReLU activations, we provide both optimization and generalization guarantees for over-parameterized networks. Specifically, we prove convergence rates of SGD to a global minimum and provide generalization guarantees for this global minimum that are independent of the network size. Therefore, our result clearly shows that the use of SGD for optimization both finds a global minimum, and avoids overfitting despite the high capacity of the model. This is the first theoretical demonstration that SGD can avoid overfitting, when learning over-specified neural network classifiers."
                    },
                    {
                        "title": "Repeat After Me: Transformers are Better than State Space Models at Copying",
                        "abstract": "Transformers are the dominant architecture for sequence modeling, but there is growing interest in models that use a fixed-size latent state that does not depend on the sequence length, which we refer to as \"generalized state space models\" (GSSMs). In this paper we show that while GSSMs are promising in terms of inference-time efficiency, they are limited compared to transformer models on tasks that require copying from the input context. We start with a theoretical analysis of the simple task of string copying and prove that a two layer transformer can copy strings of exponential length while GSSMs are fundamentally limited by their fixed-size latent state. Empirically, we find that transformers outperform GSSMs in terms of efficiency and generalization on synthetic tasks that require copying the context. Finally, we evaluate pretrained large language models and find that transformer models dramatically outperform state space models at copying and retrieving information from context. Taken together, these results suggest a fundamental gap between transformers and GSSMs on tasks of practical interest."
                    }
                ]
            }
        }
    },
    "2406.15916": {
        "paper_data": {
            "title": "Credit Attribution and Stable Compression",
            "url": "http://arxiv.org/abs/2406.15916v1",
            "arxiv_id": "2406.15916",
            "authors": [
                "Roi Livni",
                "Shay Moran",
                "Kobbi Nissim",
                "Chirag Pabbaraju"
            ],
            "abstract": "Credit attribution is crucial across various fields. In academic research, proper citation acknowledges prior work and establishes original contributions. Similarly, in generative models, such as those trained on existing artworks or music, it is important to ensure that any generated content influenced by these works appropriately credits the original creators.   We study credit attribution by machine learning algorithms. We propose new definitions--relaxations of Differential Privacy--that weaken the stability guarantees for a designated subset of $k$ datapoints. These $k$ datapoints can be used non-stably with permission from their owners, potentially in exchange for compensation. Meanwhile, the remaining datapoints are guaranteed to have no significant influence on the algorithm's output.   Our framework extends well-studied notions of stability, including Differential Privacy ($k = 0$), differentially private learning with public data (where the $k$ public datapoints are fixed in advance), and stable sample compression (where the $k$ datapoints are selected adaptively by the algorithm). We examine the expressive power of these stability notions within the PAC learning framework, provide a comprehensive characterization of learnability for algorithms adhering to these principles, and propose directions and questions for future research.",
            "introduction": "   1 Introduction  Many tasks that use machine learning algorithms require proper credit attribution. For example, consider a model trained on scientific papers that needs to reason about facts and figures based on existing literature. Most academic literature is protected under copyright licenses such as CC-BY\u00a04.0 which allows adapting, remixing, transforming, and to copy and redistribute in any medium or format, as long as attribution is given to the creator. In another setting, a learner generating content, such as images or music, may benefit from creating derivative works from copyrighted materials without violating the creator\u2019s rights (either through proper attribution or monetary compensation, depending on the context and licensing).   The increasing use of ML algorithms and the need for greater transparency is reflected by the recently implemented EU AI Act, which mandates the disclosure of training data [14]. However, disclosure of training data and proper attribution are not necessarily equivalent. In particular, mere transparency of the dataset does not reveal whether certain elements of certain content have been derived, nor does it provide proper attribution when particular content is heavily built upon. Therefore, there is a need to develop more nuanced notions and definitions that enable learning under the constraint that works are properly attributed. This paper focuses on this challenge, exploring theoretical models of credit attribution to provide rigorous and meaningful definitions for the task.   Credit attribution is part of a much larger problem of learning under copyright constraints. Copyright issues in machine learning models are becoming increasingly prominent as these models are trained on vast amounts of data, often some of which is copyrighted. Consequently, the resulting models might contain content from copyrighted data in their training sets. Previous work suggests it may be mathematically challenging to capture algorithms that protect copyright. Specifically, attempts to regulate copyright often focus on protecting against substantial similarity between output content and training data by, for example, employing stable algorithms that are not sensitive to individual training points [9, 24, 22]. This is an important aspect of copyright; however, substantial similarity is only one piece of the puzzle.   Another piece of the puzzle involves the use of original elements from copyrighted works in a legally permissible manner, such as through de minimis quotations, transformative use, and other types of fair uses, such as learning and research [13]. To fully utilize ML in many practical scenarios, it is desirable for learning models to be allowed to use original elements in a similar manner.   To address this second piece, we focus on designing algorithms that, while allowed to use and be influenced by copyrighted material, must provide proper attribution. Such models would enable users to inspect these influences and verify that they conform to legal standards, or take necessary measures (such as monetary compensation or requesting permission). Despite credit attribution being narrower in scope than copyright protection in general, even this concept may be nuanced to be captured mathematically. Therefore, we focus on formalizing a specific (but arguably basic) aspect of it \u2013 counterfactual attribution:  This principle asserts that any previous work that influenced the result should be credited. Counterfactually, if the creator of a work W\ud835\udc4aWitalic_W",
            "references": [
                {
                    "title": "Oracle-Efficient Differentially Private Learning with Public Data",
                    "abstract": "Due to statistical lower bounds on the learnability of many function classes under privacy constraints, there has been recent interest in leveraging public data to improve the performance of private learning algorithms. In this model, algorithms must always guarantee differential privacy with respect to the private samples while also ensuring learning guarantees when the private data distribution is sufficiently close to that of the public data. Previous work has demonstrated that when sufficient public, unlabelled data is available, private learning can be made statistically tractable, but the resulting algorithms have all been computationally inefficient. In this work, we present the first computationally efficient, algorithms to provably leverage public data to learn privately whenever a function class is learnable non-privately, where our notion of computational efficiency is with respect to the number of calls to an optimization oracle for the function class. In addition to this general result, we provide specialized algorithms with improved sample complexities in the special cases when the function class is convex or when the task is binary classification."
                },
                {
                    "title": "Boosting, Voting Classifiers and Randomized Sample Compression Schemes",
                    "abstract": "In boosting, we aim to leverage multiple weak learners to produce a strong learner. At the center of this paradigm lies the concept of building the strong learner as a voting classifier, which outputs a weighted majority vote of the weak learners. While many successful boosting algorithms, such as the iconic AdaBoost, produce voting classifiers, their theoretical performance has long remained sub-optimal: the best known bounds on the number of training examples necessary for a voting classifier to obtain a given accuracy has so far always contained at least two logarithmic factors above what is known to be achievable by general weak-to-strong learners. In this work, we break this barrier by proposing a randomized boosting algorithm that outputs voting classifiers whose generalization error contains a single logarithmic dependency on the sample size. We obtain this result by building a general framework that extends sample compression methods to support randomized learning algorithms based on sub-sampling."
                },
                {
                    "title": "Private Distribution Learning with Public Data: The View from Sample Compression",
                    "abstract": "We study the problem of private distribution learning with access to public data. In this setup, which we refer to as public-private learning, the learner is given public and private samples drawn from an unknown distribution $p$ belonging to a class $\\mathcal Q$, with the goal of outputting an estimate of $p$ while adhering to privacy constraints (here, pure differential privacy) only with respect to the private samples. We show that the public-private learnability of a class $\\mathcal Q$ is connected to the existence of a sample compression scheme for $\\mathcal Q$, as well as to an intermediate notion we refer to as list learning. Leveraging this connection: (1) approximately recovers previous results on Gaussians over $\\mathbb R^d$; and (2) leads to new ones, including sample complexity upper bounds for arbitrary $k$-mixtures of Gaussians over $\\mathbb R^d$, results for agnostic and distribution-shift resistant learners, as well as closure properties for public-private learnability under taking mixtures and products of distributions. Finally, via the connection to list learning, we show that for Gaussians in $\\mathbb R^d$, at least $d$ public samples are necessary for private learnability, which is close to the known upper bound of $d+1$ public samples."
                },
                {
                    "title": "PILLAR: How to make semi-private learning more effective",
                    "abstract": "In Semi-Supervised Semi-Private (SP) learning, the learner has access to both public unlabelled and private labelled data. We propose PILLAR, an easy-to-implement and computationally efficient algorithm that, under mild assumptions on the data, provably achieves significantly lower private labelled sample complexity and can be efficiently run on real-world datasets. The key idea is to use public data to estimate the principal components of the pre-trained features and subsequently project the private dataset onto the top-k Principal Components. We empirically validate the effectiveness of our algorithm in a wide variety of experiments under tight privacy constraints (\u03f5 < 1) and probe its effectiveness in low-data regimes and when the pre-training distribution significantly differs from the one on which SP learning is performed. Despite its simplicity, our algorithm exhibits significantly improved performance, in all of these settings, over all available baselines that use similar amounts of public data while often being more computationally expensive. For example, in the case of CIFAR-100 for \u03f5 = 0.1, our algorithm improves over the most competitive baselines by a factor of at least two."
                },
                {
                    "title": "Can Copyright be Reduced to Privacy?",
                    "abstract": "There is a growing concern that generative AI models will generate outputs closely resembling the copyrighted materials for which they are trained. This worry has intensified as the quality and complexity of generative models have immensely improved, and the availability of extensive datasets containing copyrighted material has expanded. Researchers are actively exploring strategies to mitigate the risk of generating infringing samples, with a recent line of work suggesting to employ techniques such as differential privacy and other forms of algorithmic stability to provide guarantees on the lack of infringing copying. In this work, we examine whether such algorithmic stability techniques are suitable to ensure the responsible use of generative models without inadvertently violating copyright laws. We argue that while these techniques aim to verify the presence of identifiable information in datasets, thus being privacy-oriented, copyright law aims to promote the use of original works for the benefit of society as a whole, provided that no unlicensed use of protected expression occurred. These fundamental differences between privacy and copyright must not be overlooked. In particular, we demonstrate that while algorithmic stability may be perceived as a practical tool to detect copying, such copying does not necessarily constitute copyright infringement. Therefore, if adopted as a standard for detecting an establishing copyright infringement, algorithmic stability may undermine the intended objectives of copyright law."
                },
                {
                    "title": "On Provable Copyright Protection for Generative Models",
                    "abstract": "There is a growing concern that learned conditional generative models may output samples that are substantially similar to some copyrighted data $C$ that was in their training set. We give a formal definition of $\\textit{near access-freeness (NAF)}$ and prove bounds on the probability that a model satisfying this definition outputs a sample similar to $C$, even if $C$ is included in its training set. Roughly speaking, a generative model $p$ is $\\textit{$k$-NAF}$ if for every potentially copyrighted data $C$, the output of $p$ diverges by at most $k$-bits from the output of a model $q$ that $\\textit{did not access $C$ at all}$. We also give generative model learning algorithms, which efficiently modify the original generative model learning algorithm in a black box manner, that output generative models with strong bounds on the probability of sampling protected content. Furthermore, we provide promising experiments for both language (transformers) and image (diffusion) generative models, showing minimal degradation in output quality while ensuring strong protections against sampling protected content."
                },
                {
                    "title": "Private Estimation with Public Data",
                    "abstract": "We initiate the study of differentially private (DP) estimation with access to a small amount of public data. For private estimation of d-dimensional Gaussians, we assume that the public data comes from a Gaussian that may have vanishing similarity in total variation distance with the underlying Gaussian of the private data. We show that under the constraints of pure or concentrated DP, d+1 public data samples are sufficient to remove any dependence on the range parameters of the private data distribution from the private sample complexity, which is known to be otherwise necessary without public data. For separated Gaussian mixtures, we assume that the underlying public and private distributions are the same, and we consider two settings: (1) when given a dimension-independent amount of public data, the private sample complexity can be improved polynomially in terms of the number of mixture components, and any dependence on the range parameters of the distribution can be removed in the approximate DP case; (2) when given an amount of public data linear in the dimension, the private sample complexity can be made independent of range parameters even under concentrated DP, and additional improvements can be made to the overall sample complexity."
                },
                {
                    "title": "Formalizing Human Ingenuity: A Quantitative Framework for Copyright Law's Substantial Similarity",
                    "abstract": "A central notion in U.S. copyright law is judging the substantial similarity between an original and an (allegedly) derived work. Capturing this notion has proven elusive, and the many approaches offered by case law and legal scholarship are often ill-defined, contradictory, or internally-inconsistent. This work suggests that key parts of the substantial-similarity puzzle are amenable to modeling inspired by theoretical computer science. Our proposed framework quantitatively evaluates how much ''novelty'' is needed to produce the derived work with access to the original work, versus reproducing it without access to the copyrighted elements of the original work. ''Novelty'' is captured by a computational notion of description length, in the spirit of Kolmogorov-Levin complexity, which is robust to mechanical transformations and availability of contextual information. This results in an actionable framework that could be used by courts as an aid for deciding substantial similarity. We evaluate it on several pivotal cases in copyright law and observe that the results are consistent with the rulings, and are philosophically aligned with the abstraction-filtration-comparison test of Altai."
                },
                {
                    "title": "Leveraging Public Data for Practical Private Query Release",
                    "abstract": "In many statistical problems, incorporating priors can significantly improve performance. However, the use of prior knowledge in differentially private query release has remained underexplored, despite such priors commonly being available in the form of public datasets, such as previous US Census releases. With the goal of releasing statistics about a private dataset, we present PMW^Pub, which -- unlike existing baselines -- leverages public data drawn from a related distribution as prior information. We provide a theoretical analysis and an empirical evaluation on the American Community Survey (ACS) and ADULT datasets, which shows that our method outperforms state-of-the-art methods. Furthermore, PMW^Pub scales well to high-dimensional data domains, where running many existing methods would be computationally infeasible."
                },
                {
                    "title": "Proper Learning, Helly Number, and an Optimal SVM Bound",
                    "abstract": "The classical PAC sample complexity bounds are stated for any Empirical Risk Minimizer (ERM) and contain an extra logarithmic factor $\\log(1/{\\epsilon})$ which is known to be necessary for ERM in general. It has been recently shown by Hanneke (2016) that the optimal sample complexity of PAC learning for any VC class C is achieved by a particular improper learning algorithm, which outputs a specific majority-vote of hypotheses in C. This leaves the question of when this bound can be achieved by proper learning algorithms, which are restricted to always output a hypothesis from C. \nIn this paper we aim to characterize the classes for which the optimal sample complexity can be achieved by a proper learning algorithm. We identify that these classes can be characterized by the dual Helly number, which is a combinatorial parameter that arises in discrete geometry and abstract convexity. In particular, under general conditions on C, we show that the dual Helly number is bounded if and only if there is a proper learner that obtains the optimal joint dependence on $\\epsilon$ and $\\delta$. \nAs further implications of our techniques we resolve a long-standing open problem posed by Vapnik and Chervonenkis (1974) on the performance of the Support Vector Machine by proving that the sample complexity of SVM in the realizable case is $\\Theta((n/{\\epsilon})+(1/{\\epsilon})\\log(1/{\\delta}))$, where $n$ is the dimension. This gives the first optimal PAC bound for Halfspaces achieved by a proper learning algorithm, and moreover is computationally efficient."
                },
                {
                    "title": "Private Query Release Assisted by Public Data",
                    "abstract": "We study the problem of differentially private query release assisted by access to public data. In this problem, the goal is to answer a large class $\\mathcal{H}$ of statistical queries with error no more than $\\alpha$ using a combination of public and private samples. The algorithm is required to satisfy differential privacy only with respect to the private samples. We study the limits of this task in terms of the private and public sample complexities. \nFirst, we show that we can solve the problem for any query class $\\mathcal{H}$ of finite VC-dimension using only $d/\\alpha$ public samples and $\\sqrt{p}d^{3/2}/\\alpha^2$ private samples, where $d$ and $p$ are the VC-dimension and dual VC-dimension of $\\mathcal{H}$, respectively. In comparison, with only private samples, this problem cannot be solved even for simple query classes with VC-dimension one, and without any private samples, a larger public sample of size $d/\\alpha^2$ is needed. Next, we give sample complexity lower bounds that exhibit tight dependence on $p$ and $\\alpha$. For the class of decision stumps, we give a lower bound of $\\sqrt{p}/\\alpha$ on the private sample complexity whenever the public sample size is less than $1/\\alpha^2$. Given our upper bounds, this shows that the dependence on $\\sqrt{p}$ is necessary in the private sample complexity. We also give a lower bound of $1/\\alpha$ on the public sample complexity for a broad family of query classes, which by our upper bound, is tight in $\\alpha$."
                },
                {
                    "title": "An Equivalence Between Private Classification and Online Prediction",
                    "abstract": "We prove that every concept class with finite Littlestone dimension can be learned by an (approximate) differentially-private algorithm. This answers an open question of Alon et al. (STOC 2019) who proved the converse statement (this question was also asked by Neel et al. (FOCS 2019)). Together these two results yield an equivalence between online learnability and private PAC learnability. We introduce a new notion of algorithmic stability called \u201cglobal stability\u201d which is essential to our proof and may be of independent interest. We also discuss an application of our results to boosting the privacy and accuracy parameters of differentially-private learners."
                },
                {
                    "title": "Limits of Private Learning with Access to Public Data",
                    "abstract": "We consider learning problems where the training set consists of two types of examples: private and public. The goal is to design a learning algorithm that satisfies differential privacy only with respect to the private examples. This setting interpolates between private learning (where all examples are private) and classical learning (where all examples are public). We study the limits of learning in this setting in terms of private and public sample complexities. We show that any hypothesis class of VC-dimension $d$ can be agnostically learned up to an excess error of $\\alpha$ using only (roughly) $d/\\alpha$ public examples and $d/\\alpha^2$ private labeled examples. This result holds even when the public examples are unlabeled. This gives a quadratic improvement over the standard $d/\\alpha^2$ upper bound on the public sample complexity (where private examples can be ignored altogether if the public examples are labeled). Furthermore, we give a nearly matching lower bound, which we prove via a generic reduction from this setting to the one of private learning without public data."
                },
                {
                    "title": "Private PAC learning implies finite Littlestone dimension",
                    "abstract": "We show that every approximately differentially private learning algorithm (possibly improper) for a class H with Littlestone dimension d requires \u03a9(log*(d)) examples. As a corollary it follows that the class of thresholds over \u2115 can not be learned in a private manner; this resolves open questions due to [Bun et al. 2015] and [Feldman and Xiao, 2015]. We leave as an open question whether every class with a finite Littlestone dimension can be learned by an approximately differentially private algorithm."
                },
                {
                    "title": "Optimal learning via local entropies and sample compression",
                    "abstract": "The aim of this paper is to provide several novel upper bounds on the excess risk with a primal focus on classification problems. We suggest two approaches and the obtained bounds are represented via the distribution dependent local entropies of the classes or the sizes of specific sample com- pression schemes. We show that in some cases, our guarantees are optimal up to constant factors and outperform previously known results. As an application of our results, we provide a new tight PAC bound for the hard-margin SVM, an extended analysis of certain empirical risk minimizers under log-concave distributions, a new variant of an online to batch conversion, and distribution dependent localized bounds in the aggregation framework. We also develop techniques that allow to replace empirical covering number or covering numbers with bracketing by the coverings with respect to the distribution of the data. The proofs for the sample compression schemes are based on the moment method combined with the analysis of voting algorithms."
                },
                {
                    "title": "Differentially Private Release and Learning of Threshold Functions",
                    "abstract": "We prove new upper and lower bounds on the sample complexity of (\u03b5, \u03b4) differentially private algorithms for releasing approximate answers to threshold functions. A threshold function c over a totally ordered domain X evaluates to cz(y) = 1 if y \u2264 x, and evaluates to 0 otherwise. We give the first nontrivial lower bound for releasing thresholds with (\u03b5, \u03b4) differential privacy, showing that the task is impossible over an infinite domain X, and moreover requires sample complexity n \u2265 \u03a9(log* |X|), which grows with the size of the domain. Inspired by the techniques used to prove this lower bound, we give an algorithm for releasing thresholds with n \u2264 2(1+\u03bf(1)) log* |X| samples. This improves the previous best upper bound of 8(1+\u03bf(1)) log* |X| (Beimel et al., RANDOM '13). Our sample complexity upper and lower bounds also apply to the tasks of learning distributions with respect to Kolmogorov distance and of properly PAC learning thresholds with differential privacy. The lower bound gives the first separation between the sample complexity of properly learning a concept class with (\u03b5, \u03b4) differential privacy and learning without privacy. For properly learning thresholds in \u2113 dimensions, this lower bound extends to n \u2265 \u03a9(\u2113 \u00b7 log* |X|). To obtain our results, we give reductions in both directions from releasing and properly learning thresholds and the simpler interior point problem. Given a database D of elements from X, the interior point problem asks for an element between the smallest and largest elements in D. We introduce new recursive constructions for bounding the sample complexity of the interior point problem, as well as further reductions and techniques for proving impossibility results for other basic problems in differential privacy."
                },
                {
                    "title": "What Can We Learn Privately?",
                    "abstract": "Learning problems form an important category of computational tasks that generalizes many of the computations researchers apply to large real-life data sets. We ask: what concept classes can be learned privately, namely, by an algorithm whose output does not depend too heavily on any one input or specific training example? More precisely, we investigate learning algorithms that satisfy differential privacy, a notion that provides strong confidentiality guarantees in the contexts where aggregate information is released about a database containing sensitive information about individuals. We present several basic results that demonstrate general feasibility of private learning and relate several models previously studied separately in the contexts of privacy and standard learning."
                },
                {
                    "title": "Synthetic Data Generators - Sequential and Private",
                    "abstract": "We study the sample complexity of private synthetic data generation over an unbounded sized class of statistical queries, and show that any class that is privately proper PAC learnable admits a private synthetic data generator (perhaps non-ef\ufb01cient). Previous work on synthetic data generators focused on the case that the query class D is \ufb01nite and obtained sample complexity bounds that scale logarithmically with the size |D| . Here we construct a private synthetic data generator whose sample complexity is independent of the domain size"
                },
                {
                    "title": "Relating Data Compression and Learnability",
                    "abstract": "We explore the learnability of two-valued functions from samples using the paradigm of Data Compression. A first algorithm (compression) choses a small subset of the sample which is called the kernel. A second algorithm predicts future values of the function from the kernel, i.e. the algorithm acts as an hypothesis for the function to be learned. The second algorithm must be able to reconstruct the correct function values when given a point of the original sample. We demonstrate that the existence of a suitable data compression scheme is sufficient to ensure learnability. We express the probability that the hypothesis predicts the function correctly on a random sample point as a function of the sample and kernel sizes. No assumptions are made on the probability distributions according to which the sample points are generated. This approach provides an alternative to that of [BEHW86], which uses the Vapnik-Chervonenkis dimension to classify learnable geometric concepts. Our bounds are derived directly from the kernel size of the algorithms rather than from the Vapnik-Chervonenkis dimension of the hypothesis class. The proofs are simpler and the introduced compression scheme provides a rigorous model for studying data compression in connection with machine learning."
                },
                {
                    "title": "A theory of the learnable",
                    "abstract": "Humans appear to be able to learn new concepts without needing to be programmed explicitly in any conventional sense. In this paper we regard learning as the phenomenon of knowledge acquisition in the absence of explicit programming. We give a precise methodology for studying this phenomenon from a computational viewpoint. It consists of choosing an appropriate information gathering mechanism, the learning protocol, and exploring the class of concepts that can be learned using it in a reasonable (polynomial) number of steps. Although inherent algorithmic complexity appears to set serious limits to the range of concepts that can be learned, we show that there are some important nontrivial classes of propositional concepts that can be learned in a realistic sense."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CR",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop rigorous and meaningful definitions and algorithms for credit attribution in machine learning models trained on copyrighted materials?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of credit attribution in machine learning is crucial for ensuring compliance with copyright laws and fostering ethical practices in AI development. By establishing clear guidelines for attribution, we can enhance transparency in machine learning processes, which is increasingly demanded by regulations like the EU AI Act. This research could lead to advancements in how machine learning models are trained and deployed, allowing for the responsible use of copyrighted materials while promoting innovation. Furthermore, it could pave the way for practical applications in various fields, such as content generation, where proper attribution is essential for legal and ethical considerations.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge of credit attribution in machine learning lies in the complexity of defining and implementing algorithms that can accurately identify and credit the influences of copyrighted works. Naive approaches may fail because they might not account for the nuanced ways in which copyrighted content can be transformed or integrated into new works. Additionally, technical obstacles include the difficulty of creating stable algorithms that are not overly sensitive to individual training data points while still being able to trace influences accurately. Theoretical challenges arise from the need to formalize concepts like counterfactual attribution in a mathematically rigorous way, which has not been adequately addressed in previous research.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on protecting against substantial similarity between output content and training data, often overlooking the need for proper credit attribution. Existing solutions have been limited by a lack of comprehensive frameworks that address the nuances of copyright law, such as transformative use and fair use. Barriers to solving this problem include the complexity of copyright regulations and the difficulty in mathematically capturing the concept of credit attribution. Our approach differs by specifically targeting the formalization of counterfactual attribution, providing a more focused and rigorous framework that has not been explored in depth in prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing algorithms that can identify and credit influences from copyrighted materials in machine learning models. We will utilize a dataset comprising various copyrighted works and their derivatives to train and evaluate our models. The key metric for success will be the accuracy and reliability of the attribution provided by our algorithms. We expect our outcomes to include"
            }
        },
        "author_data": {
            "3abf5397-a2f8-4935-879a-2277cef2c433": {
                "pk": "3abf5397-a2f8-4935-879a-2277cef2c433",
                "name": "Roi Livni",
                "collaborators": [
                    "Tomer Koren",
                    "Idan Amir",
                    "Yishay Mansour",
                    "Shay Moran",
                    "Shai Shalev-Shwartz",
                    "Ohad Shamir",
                    "Pravesh K. Kothari",
                    "Elad Hazan",
                    "Noah Golowich",
                    "Daniel Carmon"
                ],
                "domain": [
                    "Stochastic Optimization",
                    "Machine Learning",
                    "Generalization Bounds",
                    "Differential Privacy"
                ],
                "publications": [
                    {
                        "title": "Information Theoretic Lower Bounds for Information Theoretic Upper Bounds",
                        "abstract": "We examine the relationship between the mutual information between the output model and the empirical sample and the generalization of the algorithm in the context of stochastic convex optimization. Despite increasing interest in information-theoretic generalization bounds, it is uncertain if these bounds can provide insight into the exceptional performance of various learning algorithms. Our study of stochastic convex optimization reveals that, for true risk minimization, dimension-dependent mutual information is necessary. This indicates that existing information-theoretic generalization bounds fall short in capturing the generalization capabilities of algorithms like SGD and regularized ERM, which have dimension-independent sample complexity."
                    },
                    {
                        "title": "The Sample Complexity of Gradient Descent in Stochastic Convex Optimization",
                        "abstract": "We analyze the sample complexity of full-batch Gradient Descent (GD) in the setup of non-smooth Stochastic Convex Optimization. We show that the generalization error of GD, with common choice of hyper-parameters, can be $\\tilde \\Theta(d/m + 1/\\sqrt{m})$, where $d$ is the dimension and $m$ is the sample size. This matches the sample complexity of \\emph{worst-case} empirical risk minimizers. That means that, in contrast with other algorithms, GD has no advantage over naive ERMs. Our bound follows from a new generalization bound that depends on both the dimension as well as the learning rate and number of iterations. Our bound also shows that, for general hyper-parameters, when the dimension is strictly larger than number of samples, $T=\\Omega(1/\\epsilon^4)$ iterations are necessary to avoid overfitting. This resolves an open problem by Schlisserman et al.23 and Amir er Al.21, and improves over previous lower bounds that demonstrated that the sample size must be at least square root of the dimension."
                    },
                    {
                        "title": "Making Progress Based on False Discoveries",
                        "abstract": "The study of adaptive data analysis examines how many statistical queries can be answered accurately using a fixed dataset while avoiding false discoveries (statistically inaccurate answers). In this paper, we tackle a question that precedes the field of study: Is data only valuable when it provides accurate answers to statistical queries? To answer this question, we use Stochastic Convex Optimization as a case study.   In this model, algorithms are considered as analysts who query an estimate of the gradient of a noisy function at each iteration and move towards its minimizer. It is known that $O(1/\\epsilon^2)$ examples can be used to minimize the objective function, but none of the existing methods depend on the accuracy of the estimated gradients along the trajectory. Therefore, we ask: How many samples are needed to minimize a noisy convex function if we require $\\epsilon$-accurate estimates of $O(1/\\epsilon^2)$ gradients? Or, might it be that inaccurate gradient estimates are \\emph{necessary} for finding the minimum of a stochastic convex function at an optimal statistical rate?   We provide two partial answers to this question. First, we show that a general analyst (queries that may be maliciously chosen) requires $\\Omega(1/\\epsilon^3)$ samples, ruling out the possibility of a foolproof mechanism. Second, we show that, under certain assumptions on the oracle, $\\tilde \\Omega(1/\\epsilon^{2.5})$ samples are necessary for gradient descent to interact with the oracle. Our results are in contrast to classical bounds that show that $O(1/\\epsilon^2)$ samples can optimize the population risk to an accuracy of $O(\\epsilon)$, but with spurious gradients."
                    },
                    {
                        "title": "Graph-based Discriminators: Sample Complexity and Expressiveness",
                        "abstract": "A basic question in learning theory is to identify if two distributions are identical when we have access only to examples sampled from the distributions. This basic task is considered, for example, in the context of Generative Adversarial Networks (GANs), where a discriminator is trained to distinguish between a real-life distribution and a synthetic distribution. % Classically, we use a hypothesis class $H$ and claim that the two distributions are distinct if for some $h\\in H$ the expected value on the two distributions is (significantly) different. Our starting point is the following fundamental problem: \"is having the hypothesis dependent on more than a single random example beneficial\". To address this challenge we define $k$-ary based discriminators, which have a family of Boolean $k$-ary functions $\\mathcal{G}$. Each function $g\\in \\mathcal{G}$ naturally defines a hyper-graph, indicating whether a given hyper-edge exists. A function $g\\in \\mathcal{G}$ distinguishes between two distributions, if the expected value of $g$, on a $k$-tuple of i.i.d examples, on the two distributions is (significantly) different. We study the expressiveness of families of $k$-ary functions, compared to the classical hypothesis class $H$, which is $k=1$. We show a separation in expressiveness of $k+1$-ary versus $k$-ary functions. This demonstrate the great benefit of having $k\\geq 2$ as distinguishers. For $k\\geq 2$ we introduce a notion similar to the VC-dimension, and show that it controls the sample complexity. We proceed and provide upper and lower bounds as a function of our extended notion of VC-dimension."
                    },
                    {
                        "title": "Littlestone Classes are Privately Online Learnable",
                        "abstract": "We consider the problem of online classification under a privacy constraint. In this setting a learner observes sequentially a stream of labelled examples $(x_t, y_t)$, for $1 \\leq t \\leq T$, and returns at each iteration $t$ a hypothesis $h_t$ which is used to predict the label of each new example $x_t$. The learner's performance is measured by her regret against a known hypothesis class $\\mathcal{H}$. We require that the algorithm satisfies the following privacy constraint: the sequence $h_1, \\ldots, h_T$ of hypotheses output by the algorithm needs to be an $(\\epsilon, \\delta)$-differentially private function of the whole input sequence $(x_1, y_1), \\ldots, (x_T, y_T)$. We provide the first non-trivial regret bound for the realizable setting. Specifically, we show that if the class $\\mathcal{H}$ has constant Littlestone dimension then, given an oblivious sequence of labelled examples, there is a private learner that makes in expectation at most $O(\\log T)$ mistakes -- comparable to the optimal mistake bound in the non-private case, up to a logarithmic factor. Moreover, for general values of the Littlestone dimension $d$, the same mistake bound holds but with a doubly-exponential in $d$ factor. A recent line of work has demonstrated a strong connection between classes that are online learnable and those that are differentially-private learnable. Our results strengthen this connection and show that an online learning algorithm can in fact be directly privatized (in the realizable setting). We also discuss an adaptive setting and provide a sublinear regret bound of $O(\\sqrt{T})$."
                    },
                    {
                        "title": "On the Computational Efficiency of Training Neural Networks",
                        "abstract": "It is well-known that neural networks are computationally hard to train. On the other hand, in practice, modern day neural networks are trained efficiently using SGD and a variety of tricks that include different activation functions (e.g. ReLU), over-specification (i.e., train networks which are larger than needed), and regularization. In this paper we revisit the computational complexity of training neural networks from a modern perspective. We provide both positive and negative results, some of them yield new provably efficient and practical algorithms for training certain types of neural networks."
                    },
                    {
                        "title": "On the Expressive Power of Kernel Methods and the Efficiency of Kernel Learning by Association Schemes",
                        "abstract": "We study the expressive power of kernel methods and the algorithmic feasibility of multiple kernel learning for a special rich class of kernels.   Specifically, we define \\emph{Euclidean kernels}, a diverse class that includes most, if not all, families of kernels studied in literature such as polynomial kernels and radial basis functions. We then describe the geometric and spectral structure of this family of kernels over the hypercube (and to some extent for any compact domain). Our structural results allow us to prove meaningful limitations on the expressive power of the class as well as derive several efficient algorithms for learning kernels over different domains."
                    },
                    {
                        "title": "Classification with Low Rank and Missing Data",
                        "abstract": "We consider classification and regression tasks where we have missing data and assume that the (clean) data resides in a low rank subspace. Finding a hidden subspace is known to be computationally hard. Nevertheless, using a non-proper formulation we give an efficient agnostic algorithm that classifies as good as the best linear classifier coupled with the best low-dimensional subspace in which the data resides. A direct implication is that our algorithm can linearly (and non-linearly through kernels) classify provably as well as the best classifier that has access to the full data."
                    },
                    {
                        "title": "A Limitation of the PAC-Bayes Framework",
                        "abstract": "PAC-Bayes is a useful framework for deriving generalization bounds which was introduced by McAllester ('98). This framework has the flexibility of deriving distribution- and algorithm-dependent bounds, which are often tighter than VC-related uniform convergence bounds. In this manuscript we present a limitation for the PAC-Bayes framework. We demonstrate an easy learning task that is not amenable to a PAC-Bayes analysis.   Specifically, we consider the task of linear classification in 1D; it is well-known that this task is learnable using just $O(\\log(1/\\delta)/\\epsilon)$ examples. On the other hand, we show that this fact can not be proved using a PAC-Bayes analysis: for any algorithm that learns 1-dimensional linear classifiers there exists a (realizable) distribution for which the PAC-Bayes bound is arbitrarily large."
                    },
                    {
                        "title": "Learning Infinite-Layer Networks: Without the Kernel Trick",
                        "abstract": "Infinite--Layer Networks (ILN) have recently been proposed as an architecture that mimics neural networks while enjoying some of the advantages of kernel methods. ILN are networks that integrate over infinitely many nodes within a single hidden layer. It has been demonstrated by several authors that the problem of learning ILN can be reduced to the kernel trick, implying that whenever a certain integral can be computed analytically they are efficiently learnable.   In this work we give an online algorithm for ILN, which avoids the kernel trick assumption. More generally and of independent interest, we show that kernel methods in general can be exploited even when the kernel cannot be efficiently computed but can only be estimated via sampling.   We provide a regret analysis for our algorithm, showing that it matches the sample complexity of methods which have access to kernel values. Thus, our method is the first to demonstrate that the kernel trick is not necessary as such, and random features suffice to obtain comparable performance."
                    },
                    {
                        "title": "Synthetic Data Generators: Sequential and Private",
                        "abstract": "We study the sample complexity of private synthetic data generation over an unbounded sized class of statistical queries, and show that any class that is privately proper PAC learnable admits a private synthetic data generator (perhaps non-efficient). Previous work on synthetic data generators focused on the case that the query class $\\mathcal{D}$ is finite and obtained sample complexity bounds that scale logarithmically with the size $|\\mathcal{D}|$. Here we construct a private synthetic data generator whose sample complexity is independent of the domain size, and we replace finiteness with the assumption that $\\mathcal{D}$ is privately PAC learnable (a formally weaker task, hence we obtain equivalence between the two tasks)."
                    },
                    {
                        "title": "Agnostic Learning by Refuting",
                        "abstract": "The sample complexity of learning a Boolean-valued function class is precisely characterized by its Rademacher complexity. This has little bearing, however, on the sample complexity of \\emph{efficient} agnostic learning.   We introduce \\emph{refutation complexity}, a natural computational analog of Rademacher complexity of a Boolean concept class and show that it exactly characterizes the sample complexity of \\emph{efficient} agnostic learning. Informally, refutation complexity of a class $\\mathcal{C}$ is the minimum number of example-label pairs required to efficiently distinguish between the case that the labels correlate with the evaluation of some member of $\\mathcal{C}$ (\\emph{structure}) and the case where the labels are i.i.d. Rademacher random variables (\\emph{noise}). The easy direction of this relationship was implicitly used in the recent framework for improper PAC learning lower bounds of Daniely and co-authors via connections to the hardness of refuting random constraint satisfaction problems. Our work can be seen as making the relationship between agnostic learning and refutation implicit in their work into an explicit equivalence. In a recent, independent work, Salil Vadhan discovered a similar relationship between refutation and PAC-learning in the realizable (i.e. noiseless) case."
                    },
                    {
                        "title": "An Equivalence Between Private Classification and Online Prediction",
                        "abstract": "We prove that every concept class with finite Littlestone dimension can be learned by an (approximate) differentially-private algorithm. This answers an open question of Alon et al. (STOC 2019) who proved the converse statement (this question was also asked by Neel et al.~(FOCS 2019)). Together these two results yield an equivalence between online learnability and private PAC learnability.   We introduce a new notion of algorithmic stability called \"global stability\" which is essential to our proof and may be of independent interest. We also discuss an application of our results to boosting the privacy and accuracy parameters of differentially-private learners."
                    },
                    {
                        "title": "Online Learning with Low Rank Experts",
                        "abstract": "We consider the problem of prediction with expert advice when the losses of the experts have low-dimensional structure: they are restricted to an unknown $d$-dimensional subspace. We devise algorithms with regret bounds that are independent of the number of experts and depend only on the rank $d$. For the stochastic model we show a tight bound of $\\Theta(\\sqrt{dT})$, and extend it to a setting of an approximate $d$ subspace. For the adversarial model we show an upper bound of $O(d\\sqrt{T})$ and a lower bound of $\\Omega(\\sqrt{dT})$."
                    },
                    {
                        "title": "Never Go Full Batch (in Stochastic Convex Optimization)",
                        "abstract": "We study the generalization performance of $\\text{full-batch}$ optimization algorithms for stochastic convex optimization: these are first-order methods that only access the exact gradient of the empirical risk (rather than gradients with respect to individual data points), that include a wide range of algorithms such as gradient descent, mirror descent, and their regularized and/or accelerated variants. We provide a new separation result showing that, while algorithms such as stochastic gradient descent can generalize and optimize the population risk to within $\\epsilon$ after $O(1/\\epsilon^2)$ iterations, full-batch methods either need at least $\\Omega(1/\\epsilon^4)$ iterations or exhibit a dimension-dependent sample complexity."
                    },
                    {
                        "title": "Can Implicit Bias Explain Generalization? Stochastic Convex Optimization as a Case Study",
                        "abstract": "The notion of implicit bias, or implicit regularization, has been suggested as a means to explain the surprising generalization ability of modern-days overparameterized learning algorithms. This notion refers to the tendency of the optimization algorithm towards a certain structured solution that often generalizes well. Recently, several papers have studied implicit regularization and were able to identify this phenomenon in various scenarios. We revisit this paradigm in arguably the simplest non-trivial setup, and study the implicit bias of Stochastic Gradient Descent (SGD) in the context of Stochastic Convex Optimization. As a first step, we provide a simple construction that rules out the existence of a \\emph{distribution-independent} implicit regularizer that governs the generalization ability of SGD. We then demonstrate a learning problem that rules out a very general class of \\emph{distribution-dependent} implicit regularizers from explaining generalization, which includes strongly convex regularizers as well as non-degenerate norm-based regularizations. Certain aspects of our constructions point out to significant difficulties in providing a comprehensive explanation of an algorithm's generalization performance by solely arguing about its implicit regularization properties."
                    },
                    {
                        "title": "An Algorithm for Training Polynomial Networks",
                        "abstract": "We consider deep neural networks, in which the output of each node is a quadratic function of its inputs. Similar to other deep architectures, these networks can compactly represent any function on a finite training set. The main goal of this paper is the derivation of an efficient layer-by-layer algorithm for training such networks, which we denote as the \\emph{Basis Learner}. The algorithm is a universal learner in the sense that the training error is guaranteed to decrease at every iteration, and can eventually reach zero under mild conditions. We present practical implementations of this algorithm, as well as preliminary experimental results. We also compare our deep architecture to other shallow architectures for learning polynomials, in particular kernel learning."
                    },
                    {
                        "title": "SGD Generalizes Better Than GD (And Regularization Doesn't Help)",
                        "abstract": "We give a new separation result between the generalization performance of stochastic gradient descent (SGD) and of full-batch gradient descent (GD) in the fundamental stochastic convex optimization model. While for SGD it is well-known that $O(1/\\epsilon^2)$ iterations suffice for obtaining a solution with $\\epsilon$ excess expected risk, we show that with the same number of steps GD may overfit and emit a solution with $\\Omega(1)$ generalization error. Moreover, we show that in fact $\\Omega(1/\\epsilon^4)$ iterations are necessary for GD to match the generalization performance of SGD, which is also tight due to recent work by Bassily et al. (2020). We further discuss how regularizing the empirical risk minimized by GD essentially does not change the above result, and revisit the concepts of stability, implicit bias and the role of the learning algorithm in generalization."
                    },
                    {
                        "title": "Thinking Outside the Ball: Optimal Learning with Gradient Descent for Generalized Linear Stochastic Convex Optimization",
                        "abstract": "We consider linear prediction with a convex Lipschitz loss, or more generally, stochastic convex optimization problems of generalized linear form, i.e.~where each instantaneous loss is a scalar convex function of a linear function. We show that in this setting, early stopped Gradient Descent (GD), without any explicit regularization or projection, ensures excess error at most $\\epsilon$ (compared to the best possible with unit Euclidean norm) with an optimal, up to logarithmic factors, sample complexity of $\\tilde{O}(1/\\epsilon^2)$ and only $\\tilde{O}(1/\\epsilon^2)$ iterations. This contrasts with general stochastic convex optimization, where $\\Omega(1/\\epsilon^4)$ iterations are needed Amir et al. [2021b]. The lower iteration complexity is ensured by leveraging uniform convergence rather than stability. But instead of uniform convergence in a norm ball, which we show can guarantee suboptimal learning using $\\Theta(1/\\epsilon^4)$ samples, we rely on uniform convergence in a distribution-dependent ball."
                    },
                    {
                        "title": "Better Best of Both Worlds Bounds for Bandits with Switching Costs",
                        "abstract": "We study best-of-both-worlds algorithms for bandits with switching cost, recently addressed by Rouyer, Seldin and Cesa-Bianchi, 2021. We introduce a surprisingly simple and effective algorithm that simultaneously achieves minimax optimal regret bound of $\\mathcal{O}(T^{2/3})$ in the oblivious adversarial setting and a bound of $\\mathcal{O}(\\min\\{\\log (T)/\\Delta^2,T^{2/3}\\})$ in the stochastically-constrained regime, both with (unit) switching costs, where $\\Delta$ is the gap between the arms. In the stochastically constrained case, our bound improves over previous results due to Rouyer et al., that achieved regret of $\\mathcal{O}(T^{1/3}/\\Delta)$. We accompany our results with a lower bound showing that, in general, $\\tilde{\\Omega}(\\min\\{1/\\Delta^2,T^{2/3}\\})$ regret is unavoidable in the stochastically-constrained case for algorithms with $\\mathcal{O}(T^{2/3})$ worst-case regret."
                    }
                ]
            },
            "97959458-da96-4522-921f-e5fe9d368fbe": {
                "pk": "97959458-da96-4522-921f-e5fe9d368fbe",
                "name": "Shay Moran",
                "collaborators": [
                    "Amir Yehudayoff",
                    "Maryanthe Malliaris",
                    "Shlomo Moran",
                    "Rom Pinchasi",
                    "Zeev Dvir",
                    "Roi Livni",
                    "Tomer Moran",
                    "Laszlo Kozma",
                    "Cyrus Rashtchian",
                    "Steve Hanneke"
                ],
                "domain": [
                    "Statistical Learning",
                    "VC Dimension",
                    "Combinatorial Optimization",
                    "Online Learning"
                ],
                "publications": [
                    {
                        "title": "Shattering-Extremal Systems",
                        "abstract": "The Shatters relation and the VC dimension have been investigated since the early seventies. These concepts have found numerous applications in statistics, combinatorics, learning theory and computational geometry. Shattering extremal systems are set-systems with a very rich structure and many different characterizations. The goal of this thesis is to elaborate on the structure of these systems."
                    },
                    {
                        "title": "Matchings vs hitting sets among half-spaces in low dimensional euclidean spaces",
                        "abstract": "Let $\\mathcal{F}$ be any collection of linearly separable sets of a set $P$ of $n$ points either in $\\mathbb{R}^2$, or in $\\mathbb{R}^3$. We show that for every natural number $k$ either one can find $k$ pairwise disjoint sets in $\\mathcal{F}$, or there are $O(k)$ points in $P$ that together hit all sets in $\\mathcal{F}$. The proof is based on showing a similar result for families $\\mathcal{F}$ of sets separable by pseudo-discs in $\\mathbb{R}^2$. We complement these statements by showing that analogous result fails to hold for collections of linearly separable sets in $\\mathbb{R}^4$ and higher dimensional euclidean spaces."
                    },
                    {
                        "title": "On weak $\u03b5$-nets and the Radon number",
                        "abstract": "We show that the Radon number characterizes the existence of weak nets in separable convexity spaces (an abstraction of the euclidean notion of convexity). The construction of weak nets when the Radon number is finite is based on Helly's property and on metric properties of VC classes. The lower bound on the size of weak nets when the Radon number is large relies on the chromatic number of the Kneser graph. As an application, we prove a boosting-type result for weak $\\epsilon$-nets."
                    },
                    {
                        "title": "A Sauer-Shelah-Perles Lemma for Sumsets",
                        "abstract": "We show that any family of subsets $A\\subseteq 2^{[n]}$ satisfies $\\lvert A\\rvert \\leq O\\bigl(n^{\\lceil{d}/{2}\\rceil}\\bigr)$, where $d$ is the VC dimension of $\\{S\\triangle T \\,\\vert\\, S,T\\in A\\}$, and $\\triangle$ is the symmetric difference operator. We also observe that replacing $\\triangle$ by either $\\cup$ or $\\cap$ fails to satisfy an analogous statement. Our proof is based on the polynomial method; specifically, on an argument due to [Croot, Lev, Pach '17]."
                    },
                    {
                        "title": "The unstable formula theorem revisited via algorithms",
                        "abstract": "This paper is about the surprising interaction of a foundational result from model theory about stability of theories, which seems to be inherently about the infinite, with algorithmic stability in learning. Specifically, we develop a complete algorithmic analogue of Shelah's celebrated Unstable Formula Theorem, with algorithmic properties taking the place of the infinite. This draws on several new theorems as well as much recent work. In particular we introduce a new ``Probably Eventually Correct'' learning model, of independent interest, and characterize Littlestone (stable) classes in terms of this model; and we describe Littlestone classes via approximations, by analogy to definability of types in model theory."
                    },
                    {
                        "title": "Sample compression schemes for VC classes",
                        "abstract": "Sample compression schemes were defined by Littlestone and Warmuth (1986) as an abstraction of the structure underlying many learning algorithms. Roughly speaking, a sample compression scheme of size $k$ means that given an arbitrary list of labeled examples, one can retain only $k$ of them in a way that allows to recover the labels of all other examples in the list. They showed that compression implies PAC learnability for binary-labeled classes, and asked whether the other direction holds. We answer their question and show that every concept class $C$ with VC dimension $d$ has a sample compression scheme of size exponential in $d$. The proof uses an approximate minimax phenomenon for binary matrices of low VC dimension, which may be of interest in the context of game theory."
                    },
                    {
                        "title": "A Limitation of the PAC-Bayes Framework",
                        "abstract": "PAC-Bayes is a useful framework for deriving generalization bounds which was introduced by McAllester ('98). This framework has the flexibility of deriving distribution- and algorithm-dependent bounds, which are often tighter than VC-related uniform convergence bounds. In this manuscript we present a limitation for the PAC-Bayes framework. We demonstrate an easy learning task that is not amenable to a PAC-Bayes analysis.   Specifically, we consider the task of linear classification in 1D; it is well-known that this task is learnable using just $O(\\log(1/\\delta)/\\epsilon)$ examples. On the other hand, we show that this fact can not be proved using a PAC-Bayes analysis: for any algorithm that learns 1-dimensional linear classifiers there exists a (realizable) distribution for which the PAC-Bayes bound is arbitrarily large."
                    },
                    {
                        "title": "Agnostic Online Learning and Excellent Sets",
                        "abstract": "We use algorithmic methods from online learning to revisit a key idea from the interaction of model theory and combinatorics, the existence of large \"indivisible\" sets, called \"$\\epsilon$-excellent,\" in $k$-edge stable graphs (equivalently, Littlestone classes). These sets arise in the Stable Regularity Lemma, a theorem characterizing the appearance of irregular pairs in Szemer\\'edi's celebrated Regularity Lemma. Translating to the language of probability, we find a quite different existence proof for $\\epsilon$-excellent sets in Littlestone classes, using regret bounds in online learning. This proof applies to any $\\epsilon < {1}/{2}$, compared to $< {1}/{2^{2^k}}$ or so in the original proof. We include a second proof using closure properties and the VC theorem, with other advantages but weaker bounds. As a simple corollary, the Littlestone dimension remains finite under some natural modifications to the definition. A theme in these proofs is the interaction of two abstract notions of majority, arising from measure, and from rank or dimension; we prove that these densely often coincide and that this is characteristic of Littlestone (stable) classes. The last section lists several open problems."
                    },
                    {
                        "title": "Elementary Derivations of the Real Composition Algebras",
                        "abstract": "``Real Normed Algebras Revisited,'' the last paper of the late Gadi Moran, attempts to reconstruct the discovery of the complex numbers, the quaternions and the octonions, as well as proofs of their properties, using only what was known to 19th century mathematicians. In his research, Gadi had discovered simple and elegant proofs of the above-mentioned classical results using only basic properties of the geometry of Euclidean spaces and tools from high school geometry. His reconstructions underline an interesting connection between Euclidean geometry and these algebras, and avoid the advanced machinery used in previous derivations of these results. The goal of this article is to present Gadi's derivations in a way that is accessible to a wide audience of readers."
                    },
                    {
                        "title": "Shattering, Graph Orientations, and Connectivity",
                        "abstract": "We present a connection between two seemingly disparate fields: VC-theory and graph theory. This connection yields natural correspondences between fundamental concepts in VC-theory, such as shattering and VC-dimension, and well-studied concepts of graph theory related to connectivity, combinatorial optimization, forbidden subgraphs, and others.   In one direction, we use this connection to derive results in graph theory. Our main tool is a generalization of the Sauer-Shelah Lemma. Using this tool we obtain a series of inequalities and equalities related to properties of orientations of a graph. Some of these results appear to be new, for others we give new and simple proofs.   In the other direction, we present new illustrative examples of shattering-extremal systems - a class of set-systems in VC-theory whose understanding is considered by some authors to be incomplete. These examples are derived from properties of orientations related to distances and flows in networks."
                    },
                    {
                        "title": "Shattered Sets and the Hilbert Function",
                        "abstract": "We study complexity measures on subsets of the boolean hypercube and exhibit connections between algebra (the Hilbert function) and combinatorics (VC theory). These connections yield results in both directions. Our main complexity-theoretic result proves that most linear program feasibility problems cannot be computed by polynomial-sized constant-depth circuits. Moreover, our result applies to a stronger regime in which the hyperplanes are fixed and only the directions of the inequalities are given as input to the circuit. We derive this result by proving that a rich class of extremal functions in VC theory cannot be approximated by low-degree polynomials. We also present applications of algebra to combinatorics. We provide a new algebraic proof of the Sandwich Theorem, which is a generalization of the well-known Sauer-Perles-Shelah Lemma. Finally, we prove a structural result about downward-closed sets, related to the Chv\\'{a}tal conjecture in extremal combinatorics."
                    },
                    {
                        "title": "Universal Rates for Multiclass Learning",
                        "abstract": "We study universal rates for multiclass classification, establishing the optimal rates (up to log factors) for all hypothesis classes. This generalizes previous results on binary classification (Bousquet, Hanneke, Moran, van Handel, and Yehudayoff, 2021), and resolves an open question studied by Kalavasis, Velegkas, and Karbasi (2022) who handled the multiclass setting with a bounded number of class labels. In contrast, our result applies for any countable label space. Even for finite label space, our proofs provide a more precise bounds on the learning curves, as they do not depend on the number of labels. Specifically, we show that any class admits exponential rates if and only if it has no infinite Littlestone tree, and admits (near-)linear rates if and only if it has no infinite Daniely-Shalev-Shwartz-Littleston (DSL) tree, and otherwise requires arbitrarily slow rates. DSL trees are a new structure we define in this work, in which each node of the tree is given by a pseudo-cube of possible classifications of a given set of points. Pseudo-cubes are a structure, rooted in the work of Daniely and Shalev-Shwartz (2014), and recently shown by Brukhim, Carmon, Dinur, Moran, and Yehudayoff (2022) to characterize PAC learnability (i.e., uniform rates) for multiclass classification. We also resolve an open question of Kalavasis, Velegkas, and Karbasi (2022) regarding the equivalence of classes having infinite Graph-Littlestone (GL) trees versus infinite Natarajan-Littlestone (NL) trees, showing that they are indeed equivalent."
                    },
                    {
                        "title": "On the Perceptron's Compression",
                        "abstract": "We study and provide exposition to several phenomena that are related to the perceptron's compression. One theme concerns modifications of the perceptron algorithm that yield better guarantees on the margin of the hyperplane it outputs. These modifications can be useful in training neural networks as well, and we demonstrate them with some experimental data. In a second theme, we deduce conclusions from the perceptron's compression in various contexts."
                    },
                    {
                        "title": "Labeled compression schemes for extremal classes",
                        "abstract": "It is a long-standing open problem whether there always exists a compression scheme whose size is of the order of the Vapnik-Chervonienkis (VC) dimension $d$. Recently compression schemes of size exponential in $d$ have been found for any concept class of VC dimension $d$. Previously, compression schemes of size $d$ have been given for maximum classes, which are special concept classes whose size equals an upper bound due to Sauer-Shelah. We consider a generalization of maximum classes called extremal classes. Their definition is based on a powerful generalization of the Sauer-Shelah bound called the Sandwich Theorem, which has been studied in several areas of combinatorics and computer science. The key result of the paper is a construction of a sample compression scheme for extremal classes of size equal to their VC dimension. We also give a number of open problems concerning the combinatorial structure of extremal classes and the existence of unlabeled compression schemes for them."
                    },
                    {
                        "title": "Generalized comparison trees for point-location problems",
                        "abstract": "Let $H$ be an arbitrary family of hyper-planes in $d$-dimensions. We show that the point-location problem for $H$ can be solved by a linear decision tree that only uses a special type of queries called \\emph{generalized comparison queries}. These queries correspond to hyperplanes that can be written as a linear combination of two hyperplanes from $H$; in particular, if all hyperplanes in $H$ are $k$-sparse then generalized comparisons are $2k$-sparse. The depth of the obtained linear decision tree is polynomial in $d$ and logarithmic in $|H|$, which is comparable to previous results in the literature that use general linear queries.   This extends the study of comparison trees from a previous work by the authors [Kane {et al.}, FOCS 2017]. The main benefit is that using generalized comparison queries allows to overcome limitations that apply for the more restricted type of comparison queries.   Our analysis combines a seminal result of Forster regarding sets in isotropic position [Forster, JCSS 2002], the margin-based inference dimension analysis for comparison queries from [Kane {et al.}, FOCS 2017], and compactness arguments."
                    },
                    {
                        "title": "An adaptive nearest neighbor rule for classification",
                        "abstract": "We introduce a variant of the $k$-nearest neighbor classifier in which $k$ is chosen adaptively for each query, rather than supplied as a parameter. The choice of $k$ depends on properties of each neighborhood, and therefore may significantly vary between different points. (For example, the algorithm will use larger $k$ for predicting the labels of points in noisy regions.)   We provide theory and experiments that demonstrate that the algorithm performs comparably to, and sometimes better than, $k$-NN with an optimal choice of $k$. In particular, we derive bounds on the convergence rates of our classifier that depend on a local quantity we call the `advantage' which is significantly weaker than the Lipschitz conditions used in previous convergence rate proofs. These generalization bounds hinge on a variant of the seminal Uniform Convergence Theorem due to Vapnik and Chervonenkis; this variant concerns conditional probabilities and may be of independent interest."
                    },
                    {
                        "title": "Private Learning Implies Online Learning: An Efficient Reduction",
                        "abstract": "We study the relationship between the notions of differentially private learning and online learning in games. Several recent works have shown that differentially private learning implies online learning, but an open problem of Neel, Roth, and Wu \\cite{NeelAaronRoth2018} asks whether this implication is {\\it efficient}. Specifically, does an efficient differentially private learner imply an efficient online learner? In this paper we resolve this open question in the context of pure differential privacy. We derive an efficient black-box reduction from differentially private learning to online learning from expert advice."
                    },
                    {
                        "title": "A Resilient Distributed Boosting Algorithm",
                        "abstract": "Given a learning task where the data is distributed among several parties, communication is one of the fundamental resources which the parties would like to minimize. We present a distributed boosting algorithm which is resilient to a limited amount of noise. Our algorithm is similar to classical boosting algorithms, although it is equipped with a new component, inspired by Impagliazzo's hard-core lemma [Impagliazzo95], adding a robustness quality to the algorithm. We also complement this result by showing that resilience to any asymptotically larger noise is not achievable by a communication-efficient algorithm."
                    },
                    {
                        "title": "Simple and Optimal Randomized Fault-Tolerant Rumor Spreading",
                        "abstract": "We revisit the classic problem of spreading a piece of information in a group of $n$ fully connected processors. By suitably adding a small dose of randomness to the protocol of Gasienic and Pelc (1996), we derive for the first time protocols that (i) use a linear number of messages, (ii) are correct even when an arbitrary number of adversarially chosen processors does not participate in the process, and (iii) with high probability have the asymptotically optimal runtime of $O(\\log n)$ when at least an arbitrarily small constant fraction of the processors are working. In addition, our protocols do not require that the system is synchronized nor that all processors are simultaneously woken up at time zero, they are fully based on push-operations, and they do not need an a priori estimate on the number of failed nodes.   Our protocols thus overcome the typical disadvantages of the two known approaches, algorithms based on random gossip (typically needing a large number of messages due to their unorganized nature) and algorithms based on fair workload splitting (which are either not {time-efficient} or require intricate preprocessing steps plus synchronization)."
                    },
                    {
                        "title": "On statistical learning via the lens of compression",
                        "abstract": "This work continues the study of the relationship between sample compression schemes and statistical learning, which has been mostly investigated within the framework of binary classification. The central theme of this work is establishing equivalences between learnability and compressibility, and utilizing these equivalences in the study of statistical learning theory.   We begin with the setting of multiclass categorization (zero/one loss). We prove that in this case learnability is equivalent to compression of logarithmic sample size, and that uniform convergence implies compression of constant size.   We then consider Vapnik's general learning setting: we show that in order to extend the compressibility-learnability equivalence to this case, it is necessary to consider an approximate variant of compression.   Finally, we provide some applications of the compressibility-learnability equivalences:   (i) Agnostic-case learnability and realizable-case learnability are equivalent in multiclass categorization problems (in terms of sample complexity).   (ii) This equivalence between agnostic-case learnability and realizable-case learnability does not hold for general learning problems: There exists a learning problem whose loss function takes just three values, under which agnostic-case and realizable-case learnability are not equivalent.   (iii) Uniform convergence implies compression of constant size in multiclass categorization problems. Part of the argument includes an analysis of the uniform convergence rate in terms of the graph dimension, in which we improve upon previous bounds.   (iv) A dichotomy for sample compression in multiclass categorization problems: If a non-trivial compression exists then a compression of logarithmic size exists.   (v) A compactness theorem for multiclass categorization problems."
                    }
                ]
            },
            "fe5d6b89-78ba-46ce-814f-2e5dab9283b1": {
                "pk": "fe5d6b89-78ba-46ce-814f-2e5dab9283b1",
                "name": "Kobbi Nissim",
                "collaborators": [
                    "Uri Stemmer",
                    "Rann Smorodinsky",
                    "Amos Beimel",
                    "Aloni Cohen",
                    "Moshe Tennenholtz",
                    "Chao Yan",
                    "Hai Brenner",
                    "Moni Naor",
                    "Salil Vadhan",
                    "Haim Kaplan"
                ],
                "domain": [
                    "Differential Privacy",
                    "Secure Computation",
                    "Mechanism Design",
                    "Statistical Learning"
                ],
                "publications": [
                    {
                        "title": "Linear Program Reconstruction in Practice",
                        "abstract": "We briefly report on a successful linear program reconstruction attack performed on a production statistical queries system and using a real dataset. The attack was deployed in test environment in the course of the Aircloak Challenge bug bounty program and is based on the reconstruction algorithm of Dwork, McSherry, and Talwar. We empirically evaluate the effectiveness of the algorithm and a related algorithm by Dinur and Nissim with various dataset sizes, error rates, and numbers of queries in a Gaussian noise setting."
                    },
                    {
                        "title": "The Sample Complexity of Distribution-Free Parity Learning in the Robust Shuffle Model",
                        "abstract": "We provide a lowerbound on the sample complexity of distribution-free parity learning in the realizable case in the shuffle model of differential privacy. Namely, we show that the sample complexity of learning $d$-bit parity functions is $\\Omega(2^{d/2})$. Our result extends a recent similar lowerbound on the sample complexity of private agnostic learning of parity functions in the shuffle model by Cheu and Ullman. We also sketch a simple shuffle model protocol demonstrating that our results are tight up to $poly(d)$ factors."
                    },
                    {
                        "title": "Impossibility of Differentially Private Universally Optimal Mechanisms",
                        "abstract": "The notion of a universally utility-maximizing privacy mechanism was recently introduced by Ghosh, Roughgarden, and Sundararajan [STOC 2009]. These are mechanisms that guarantee optimal utility to a large class of information consumers, simultaneously, while preserving Differential Privacy [Dwork, McSherry, Nissim, and Smith, TCC 2006]. Ghosh et al. have demonstrated, quite surprisingly, a case where such a universally-optimal differentially-private mechanisms exists, when the information consumers are Bayesian. This result was recently extended by Gupte and Sundararajan [PODS 2010] to risk-averse consumers.   Both positive results deal with mechanisms (approximately) computing a single count query (i.e., the number of individuals satisfying a specific property in a given population), and the starting point of our work is a trial at extending these results to similar settings, such as sum queries with non-binary individual values, histograms, and two (or more) count queries. We show, however, that universally-optimal mechanisms do not exist for all these queries, both for Bayesian and risk-averse consumers.   For the Bayesian case, we go further, and give a characterization of those functions that admit universally-optimal mechanisms, showing that a universally-optimal mechanism exists, essentially, only for a (single) count query. At the heart of our proof is a representation of a query function $f$ by its privacy constraint graph $G_f$ whose edges correspond to values resulting by applying $f$ to neighboring databases."
                    },
                    {
                        "title": "Communication Complexity and Secure Function Evaluation",
                        "abstract": "We suggest two new methodologies for the design of efficient secure protocols, that differ with respect to their underlying computational models. In one methodology we utilize the communication complexity tree (or branching for f and transform it into a secure protocol. In other words, \"any function f that can be computed using communication complexity c can be can be computed securely using communication complexity that is polynomial in c and a security parameter\". The second methodology uses the circuit computing f, enhanced with look-up tables as its underlying computational model. It is possible to simulate any RAM machine in this model with polylogarithmic blowup. Hence it is possible to start with a computation of f on a RAM machine and transform it into a secure protocol.   We show many applications of these new methodologies resulting in protocols efficient either in communication or in computation. In particular, we exemplify a protocol for the \"millionaires problem\", where two participants want to compare their values but reveal no other information. Our protocol is more efficient than previously known ones in either communication or computation."
                    },
                    {
                        "title": "Clustering Algorithms for the Centralized and Local Models",
                        "abstract": "We revisit the problem of finding a minimum enclosing ball with differential privacy: Given a set of $n$ points in the Euclidean space $\\mathbb{R}^d$ and an integer $t\\leq n$, the goal is to find a ball of the smallest radius $r_{opt}$ enclosing at least $t$ input points. The problem is motivated by its various applications to differential privacy, including the sample and aggregate technique, private data exploration, and clustering.   Without privacy concerns, minimum enclosing ball has a polynomial time approximation scheme (PTAS), which computes a ball of radius almost $r_{opt}$ (the problem is NP-hard to solve exactly). In contrast, under differential privacy, until this work, only a $O(\\sqrt{\\log n})$-approximation algorithm was known.   We provide new constructions of differentially private algorithms for minimum enclosing ball achieving constant factor approximation to $r_{opt}$ both in the centralized model (where a trusted curator collects the sensitive information and analyzes it with differential privacy) and in the local model (where each respondent randomizes her answers to the data curator to protect her privacy).   We demonstrate how to use our algorithms as a building block for approximating $k$-means in both models."
                    },
                    {
                        "title": "Concentration Bounds for High Sensitivity Functions Through Differential Privacy",
                        "abstract": "A new line of work [Dwork et al. STOC 2015], [Hardt and Ullman FOCS 2014], [Steinke and Ullman COLT 2015], [Bassily et al. STOC 2016] demonstrates how differential privacy [Dwork et al. TCC 2006] can be used as a mathematical tool for guaranteeing generalization in adaptive data analysis. Specifically, if a differentially private analysis is applied on a sample S of i.i.d. examples to select a low-sensitivity function f, then w.h.p. f(S) is close to its expectation, although f is being chosen based on the data.   Very recently, Steinke and Ullman observed that these generalization guarantees can be used for proving concentration bounds in the non-adaptive setting, where the low-sensitivity function is fixed beforehand. In particular, they obtain alternative proofs for classical concentration bounds for low-sensitivity functions, such as the Chernoff bound and McDiarmid's Inequality.   In this work, we set out to examine the situation for functions with high-sensitivity, for which differential privacy does not imply generalization guarantees under adaptive analysis. We show that differential privacy can be used to prove concentration bounds for such functions in the non-adaptive setting."
                    },
                    {
                        "title": "On the Generalization Properties of Differential Privacy",
                        "abstract": "A new line of work, started with Dwork et al., studies the task of answering statistical queries using a sample and relates the problem to the concept of differential privacy. By the Hoeffding bound, a sample of size $O(\\log k/\\alpha^2)$ suffices to answer $k$ non-adaptive queries within error $\\alpha$, where the answers are computed by evaluating the statistical queries on the sample. This argument fails when the queries are chosen adaptively (and can hence depend on the sample). Dwork et al. showed that if the answers are computed with $(\\epsilon,\\delta)$-differential privacy then $O(\\epsilon)$ accuracy is guaranteed with probability $1-O(\\delta^\\epsilon)$. Using the Private Multiplicative Weights mechanism, they concluded that the sample size can still grow polylogarithmically with the $k$.   Very recently, Bassily et al. presented an improved bound and showed that (a variant of) the private multiplicative weights algorithm can answer $k$ adaptively chosen statistical queries using sample complexity that grows logarithmically in $k$. However, their results no longer hold for every differentially private algorithm, and require modifying the private multiplicative weights algorithm in order to obtain their high probability bounds.   We greatly simplify the results of Dwork et al. and improve on the bound by showing that differential privacy guarantees $O(\\epsilon)$ accuracy with probability $1-O(\\delta\\log(1/\\epsilon)/\\epsilon)$. It would be tempting to guess that an $(\\epsilon,\\delta)$-differentially private computation should guarantee $O(\\epsilon)$ accuracy with probability $1-O(\\delta)$. However, we show that this is not the case, and that our bound is tight (up to logarithmic factors)."
                    },
                    {
                        "title": "Towards Formalizing the GDPR's Notion of Singling Out",
                        "abstract": "There is a significant conceptual gap between legal and mathematical thinking around data privacy. The effect is uncertainty as to which technical offerings adequately match expectations expressed in legal standards. The uncertainty is exacerbated by a litany of successful privacy attacks, demonstrating that traditional statistical disclosure limitation techniques often fall short of the sort of privacy envisioned by legal standards.   We define predicate singling out, a new type of privacy attack intended to capture the concept of singling out appearing in the General Data Protection Regulation (GDPR). Informally, an adversary predicate singles out a dataset $X$ using the output of a data release mechanism $M(X)$ if it manages to find a predicate $p$ matching exactly one row $x \\in X$ with probability much better than a statistical baseline. A data release mechanism that precludes such attacks is secure against predicate singling out (PSO secure).   We argue that PSO security is a mathematical concept with legal consequences. Any data release mechanism that purports to \"render anonymous\" personal data under the GDPR must be secure against singling out, and hence must be PSO secure. We then analyze PSO security, showing that it fails to self-compose. Namely, a combination of $\\omega(\\log n)$ exact counts, each individually PSO secure, enables an attacker to predicate single out. In fact, the composition of just two PSO-secure mechanisms can fail to provide PSO security.   Finally, we ask whether differential privacy and $k$-anonymity are PSO secure. Leveraging a connection to statistical generalization, we show that differential privacy implies PSO security. However, $k$-anonymity does not: there exists a simple and general predicate singling out attack under mild assumptions on the $k$-anonymizer and the data distribution."
                    },
                    {
                        "title": "Locating a Small Cluster Privately",
                        "abstract": "We present a new algorithm for locating a small cluster of points with differential privacy [Dwork, McSherry, Nissim, and Smith, 2006]. Our algorithm has implications to private data exploration, clustering, and removal of outliers. Furthermore, we use it to significantly relax the requirements of the sample and aggregate technique [Nissim, Raskhodnikova, and Smith, 2007], which allows compiling of \"off the shelf\" (non-private) analyses into analyses that preserve differential privacy."
                    },
                    {
                        "title": "Separating Adaptive Streaming from Oblivious Streaming",
                        "abstract": "We present a streaming problem for which every adversarially-robust streaming algorithm must use polynomial space, while there exists a classical (oblivious) streaming algorithm that uses only polylogarithmic space. This is the first separation between oblivious streaming and adversarially-robust streaming, and resolves one of the central open questions in adversarial robust streaming."
                    },
                    {
                        "title": "Approximately Optimal Mechanism Design via Differential Privacy",
                        "abstract": "In this paper we study the implementation challenge in an abstract interdependent values model and an arbitrary objective function. We design a mechanism that allows for approximate optimal implementation of insensitive objective functions in ex-post Nash equilibrium. If, furthermore, values are private then the same mechanism is strategy proof. We cast our results onto two specific models: pricing and facility location. The mechanism we design is optimal up to an additive factor of the order of magnitude of one over the square root of the number of agents and involves no utility transfers.   Underlying our mechanism is a lottery between two auxiliary mechanisms: with high probability we actuate a mechanism that reduces players' influence on the choice of the social alternative, while choosing the optimal outcome with high probability. This is where the recent notion of differential privacy is employed. With the complementary probability we actuate a mechanism that is typically far from optimal but is incentive compatible. The joint mechanism inherits the desired properties from both."
                    },
                    {
                        "title": "Segmentation, Incentives and Privacy",
                        "abstract": "Data driven segmentation is the powerhouse behind the success of online advertising. Various underlying challenges for successful segmentation have been studied by the academic community, with one notable exception - consumers incentives have been typically ignored. This lacuna is troubling as consumers have much control over the data being collected. Missing or manipulated data could lead to inferior segmentation. The current work proposes a model of prior-free segmentation, inspired by models of facility location, and to the best of our knowledge provides the first segmentation mechanism that addresses incentive compatibility, efficient market segmentation and privacy in the absence of a common prior."
                    },
                    {
                        "title": "Distributed Private Data Analysis: On Simultaneously Solving How and What",
                        "abstract": "We examine the combination of two directions in the field of privacy concerning computations over distributed private inputs - secure function evaluation (SFE) and differential privacy. While in both the goal is to privately evaluate some function of the individual inputs, the privacy requirements are significantly different. The general feasibility results for SFE suggest a natural paradigm for implementing differentially private analyses distributively: First choose what to compute, i.e., a differentially private analysis; Then decide how to compute it, i.e., construct an SFE protocol for this analysis.   We initiate an examination whether there are advantages to a paradigm where both decisions are made simultaneously. In particular, we investigate under which accuracy requirements it is beneficial to adapt this paradigm for computing a collection of functions including binary sum, gap threshold, and approximate median queries. Our results imply that when computing the binary sum of $n$ distributed inputs then:   * When we require that the error is $o(\\sqrt{n})$ and the number of rounds is constant, there is no benefit in the new paradigm.   * When we allow an error of $O(\\sqrt{n})$, the new paradigm yields more efficient protocols when we consider protocols that compute symmetric functions.   Our results also yield new separations between the local and global models of computations for private data analysis."
                    },
                    {
                        "title": "Private Learning and Sanitization: Pure vs. Approximate Differential Privacy",
                        "abstract": "We compare the sample complexity of private learning [Kasiviswanathan et al. 2008] and sanitization~[Blum et al. 2008] under pure $\\epsilon$-differential privacy [Dwork et al. TCC 2006] and approximate $(\\epsilon,\\delta)$-differential privacy [Dwork et al. Eurocrypt 2006]. We show that the sample complexity of these tasks under approximate differential privacy can be significantly lower than that under pure differential privacy.   We define a family of optimization problems, which we call Quasi-Concave Promise Problems, that generalizes some of our considered tasks. We observe that a quasi-concave promise problem can be privately approximated using a solution to a smaller instance of a quasi-concave promise problem. This allows us to construct an efficient recursive algorithm solving such problems privately. Specifically, we construct private learners for point functions, threshold functions, and axis-aligned rectangles in high dimension. Similarly, we construct sanitizers for point functions and threshold functions.   We also examine the sample complexity of label-private learners, a relaxation of private learning where the learner is required to only protect the privacy of the labels in the sample. We show that the VC dimension completely characterizes the sample complexity of such learners, that is, the sample complexity of learning with label privacy is equal (up to constants) to learning without privacy."
                    },
                    {
                        "title": "Exploring Differential Obliviousness",
                        "abstract": "In a recent paper Chan et al. [SODA '19] proposed a relaxation of the notion of (full) memory obliviousness, which was introduced by Goldreich and Ostrovsky [J. ACM '96] and extensively researched by cryptographers. The new notion, differential obliviousness, requires that any two neighboring inputs exhibit similar memory access patterns, where the similarity requirement is that of differential privacy. Chan et al. demonstrated that differential obliviousness allows achieving improved efficiency for several algorithmic tasks, including sorting, merging of sorted lists, and range query data structures.   In this work, we continue the exploration and mapping of differential obliviousness, focusing on algorithms that do not necessarily examine all their input. This choice is motivated by the fact that the existence of logarithmic overhead ORAM protocols implies that differential obliviousness can yield at most a logarithmic improvement in efficiency for computations that need to examine all their input. In particular, we explore property testing, where we show that differential obliviousness yields an almost linear improvement in overhead in the dense graph model, and at most quadratic improvement in the bounded degree model. We also explore tasks where a non-oblivious algorithm would need to explore different portions of the input, where the latter would depend on the input itself, and where we show that such a behavior can be maintained under differential obliviousness, but not under full obliviousness. Our examples suggest that there would be benefits in further exploring which class of computational tasks are amenable to differential obliviousness."
                    },
                    {
                        "title": "Hard Instances of the Constrained Discrete Logarithm Problem",
                        "abstract": "The discrete logarithm problem (DLP) generalizes to the constrained DLP, where the secret exponent $x$ belongs to a set known to the attacker. The complexity of generic algorithms for solving the constrained DLP depends on the choice of the set. Motivated by cryptographic applications, we study sets with succinct representation for which the constrained DLP is hard. We draw on earlier results due to Erd\\\"os et al. and Schnorr, develop geometric tools such as generalized Menelaus' theorem for proving lower bounds on the complexity of the constrained DLP, and construct sets with succinct representation with provable non-trivial lower bounds."
                    },
                    {
                        "title": "Privacy-Aware Mechanism Design",
                        "abstract": "In traditional mechanism design, agents only care about the utility they derive from the outcome of the mechanism. We look at a richer model where agents also assign non-negative dis-utility to the information about their private types leaked by the outcome of the mechanism.   We present a new model for privacy-aware mechanism design, where we only assume an upper bound on the agents' loss due to leakage, as opposed to previous work where a full characterization of the loss was required.   In this model, under a mild assumption on the distribution of how agents value their privacy, we show a generic construction of privacy-aware mechanisms and demonstrate its applicability to electronic polling and pricing of a digital good."
                    },
                    {
                        "title": "Simultaneous Private Learning of Multiple Concepts",
                        "abstract": "We investigate the direct-sum problem in the context of differentially private PAC learning: What is the sample complexity of solving $k$ learning tasks simultaneously under differential privacy, and how does this cost compare to that of solving $k$ learning tasks without privacy? In our setting, an individual example consists of a domain element $x$ labeled by $k$ unknown concepts $(c_1,\\ldots,c_k)$. The goal of a multi-learner is to output $k$ hypotheses $(h_1,\\ldots,h_k)$ that generalize the input examples.   Without concern for privacy, the sample complexity needed to simultaneously learn $k$ concepts is essentially the same as needed for learning a single concept. Under differential privacy, the basic strategy of learning each hypothesis independently yields sample complexity that grows polynomially with $k$. For some concept classes, we give multi-learners that require fewer samples than the basic strategy. Unfortunately, however, we also give lower bounds showing that even for very simple concept classes, the sample cost of private multi-learning must grow polynomially in $k$."
                    },
                    {
                        "title": "Practical Locally Private Heavy Hitters",
                        "abstract": "We present new practical local differentially private heavy hitters algorithms achieving optimal or near-optimal worst-case error and running time -- TreeHist and Bitstogram. In both algorithms, server running time is $\\tilde O(n)$ and user running time is $\\tilde O(1)$, hence improving on the prior state-of-the-art result of Bassily and Smith [STOC 2015] requiring $O(n^{5/2})$ server time and $O(n^{3/2})$ user time. With a typically large number of participants in local algorithms ($n$ in the millions), this reduction in time complexity, in particular at the user side, is crucial for making locally private heavy hitters algorithms usable in practice. We implemented Algorithm TreeHist to verify our theoretical analysis and compared its performance with the performance of Google's RAPPOR code."
                    },
                    {
                        "title": "Differentially Private Summation with Multi-Message Shuffling",
                        "abstract": "In recent work, Cheu et al. (Eurocrypt 2019) proposed a protocol for $n$-party real summation in the shuffle model of differential privacy with $O_{\\epsilon, \\delta}(1)$ error and $\\Theta(\\epsilon\\sqrt{n})$ one-bit messages per party. In contrast, every local model protocol for real summation must incur error $\\Omega(1/\\sqrt{n})$, and there exist protocols matching this lower bound which require just one bit of communication per party. Whether this gap in number of messages is necessary was left open by Cheu et al.   In this note we show a protocol with $O(1/\\epsilon)$ error and $O(\\log(n/\\delta))$ messages of size $O(\\log(n))$ per party. This protocol is based on the work of Ishai et al.\\ (FOCS 2006) showing how to implement distributed summation from secure shuffling, and the observation that this allows simulating the Laplace mechanism in the shuffle model."
                    }
                ]
            },
            "524fd712-9706-4406-b26d-20e62af9cef6": {
                "pk": "524fd712-9706-4406-b26d-20e62af9cef6",
                "name": "Chirag Pabbaraju",
                "collaborators": [
                    "Moses Charikar",
                    "Holden Lee",
                    "Anish Sevekari",
                    "Andrej Risteski",
                    "Kirankumar Shiragur",
                    "Sahasrajit Sarmasarkar",
                    "Prateek Jain",
                    "Shivam Garg",
                    "Gregory Valiant",
                    "Spencer Compton"
                ],
                "domain": [
                    "Machine Learning",
                    "Statistical Learning Theory",
                    "Decision Trees",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Multiclass Learnability Does Not Imply Sample Compression",
                        "abstract": "A hypothesis class admits a sample compression scheme, if for every sample labeled by a hypothesis from the class, it is possible to retain only a small subsample, using which the labels on the entire sample can be inferred. The size of the compression scheme is an upper bound on the size of the subsample produced. Every learnable binary hypothesis class (which must necessarily have finite VC dimension) admits a sample compression scheme of size only a finite function of its VC dimension, independent of the sample size. For multiclass hypothesis classes, the analog of VC dimension is the DS dimension. We show that the analogous statement pertaining to sample compression is not true for multiclass hypothesis classes: every learnable multiclass hypothesis class, which must necessarily have finite DS dimension, does not admit a sample compression scheme of size only a finite function of its DS dimension."
                    },
                    {
                        "title": "A Characterization of List Learnability",
                        "abstract": "A classical result in learning theory shows the equivalence of PAC learnability of binary hypothesis classes and the finiteness of VC dimension. Extending this to the multiclass setting was an open problem, which was settled in a recent breakthrough result characterizing multiclass PAC learnability via the DS dimension introduced earlier by Daniely and Shalev-Shwartz. In this work we consider list PAC learning where the goal is to output a list of $k$ predictions. List learning algorithms have been developed in several settings before and indeed, list learning played an important role in the recent characterization of multiclass learnability. In this work we ask: when is it possible to $k$-list learn a hypothesis class? We completely characterize $k$-list learnability in terms of a generalization of DS dimension that we call the $k$-DS dimension. Generalizing the recent characterization of multiclass learnability, we show that a hypothesis class is $k$-list learnable if and only if the $k$-DS dimension is finite."
                    },
                    {
                        "title": "A Characterization of List Regression",
                        "abstract": "There has been a recent interest in understanding and characterizing the sample complexity of list learning tasks, where the learning algorithm is allowed to make a short list of $k$ predictions, and we simply require one of the predictions to be correct. This includes recent works characterizing the PAC sample complexity of standard list classification and online list classification.   Adding to this theme, in this work, we provide a complete characterization of list PAC regression. We propose two combinatorial dimensions, namely the $k$-OIG dimension and the $k$-fat-shattering dimension, and show that they optimally characterize realizable and agnostic $k$-list regression respectively. These quantities generalize known dimensions for standard regression. Our work thus extends existing list learning characterizations from classification to regression."
                    },
                    {
                        "title": "Learning Functions over Sets via Permutation Adversarial Networks",
                        "abstract": "In this paper, we consider the problem of learning functions over sets, i.e., functions that are invariant to permutations of input set items. Recent approaches of pooling individual element embeddings can necessitate extremely large embedding sizes for challenging functions. We address this challenge by allowing standard neural networks like LSTMs to succinctly capture the function over the set. However, to ensure invariance with respect to permutations of set elements, we propose a novel architecture called SPAN that simultaneously learns the function as well as adversarial or worst-case permutations for each input set. The learning problem reduces to a min-max optimization problem that is solved via a simple alternating block coordinate descent technique. We conduct extensive experiments on a variety of set-learning tasks and demonstrate that SPAN learns nearly permutation-invariant functions while still ensuring accuracy on test data. On a variety of tasks sampled from the domains of statistics, graph functions and linear algebra, we show that our method can significantly outperform state-of-the-art methods such as DeepSets and Janossy Pooling. Finally, we present a case study of how learning set-functions can help extract powerful features for recommendation systems, and show that such a method can be as much as 2% more accurate than carefully hand-tuned features on a real-world recommendation system."
                    },
                    {
                        "title": "Testing with Non-identically Distributed Samples",
                        "abstract": "We examine the extent to which sublinear-sample property testing and estimation applies to settings where samples are independently but not identically distributed. Specifically, we consider the following distributional property testing framework: Suppose there is a set of distributions over a discrete support of size $k$, $\\textbf{p}_1, \\textbf{p}_2,\\ldots,\\textbf{p}_T$, and we obtain $c$ independent draws from each distribution. Suppose the goal is to learn or test a property of the average distribution, $\\textbf{p}_{\\mathrm{avg}}$. This setup models a number of important practical settings where the individual distributions correspond to heterogeneous entities -- either individuals, chronologically distinct time periods, spatially separated data sources, etc. From a learning standpoint, even with $c=1$ samples from each distribution, $\\Theta(k/\\varepsilon^2)$ samples are necessary and sufficient to learn $\\textbf{p}_{\\mathrm{avg}}$ to within error $\\varepsilon$ in TV distance. To test uniformity or identity -- distinguishing the case that $\\textbf{p}_{\\mathrm{avg}}$ is equal to some reference distribution, versus has $\\ell_1$ distance at least $\\varepsilon$ from the reference distribution, we show that a linear number of samples in $k$ is necessary given $c=1$ samples from each distribution. In contrast, for $c \\ge 2$, we recover the usual sublinear sample testing of the i.i.d. setting: we show that $O(\\sqrt{k}/\\varepsilon^2 + 1/\\varepsilon^4)$ samples are sufficient, matching the optimal sample complexity in the i.i.d. case in the regime where $\\varepsilon \\ge k^{-1/4}$. Additionally, we show that in the $c=2$ case, there is a constant $\\rho > 0$ such that even in the linear regime with $\\rho k$ samples, no tester that considers the multiset of samples (ignoring which samples were drawn from the same $\\textbf{p}_i$) can perform uniformity testing."
                    },
                    {
                        "title": "Embedding Probability Distributions into Low Dimensional $\\ell_1$: Tree Ising Models via Truncated Metrics",
                        "abstract": "Given an arbitrary set of high dimensional points in $\\ell_1$, there are known negative results that preclude the possibility of always mapping them to a low dimensional $\\ell_1$ space while preserving distances with small multiplicative distortion. This is in stark contrast with dimension reduction in Euclidean space ($\\ell_2$) where such mappings are always possible. While the first non-trivial lower bounds for $\\ell_1$ dimension reduction were established almost 20 years ago, there has been limited progress in understanding what sets of points in $\\ell_1$ are conducive to a low-dimensional mapping.   In this work, we study a new characterization of $\\ell_1$ metrics that are conducive to dimension reduction in $\\ell_1$. Our characterization focuses on metrics that are defined by the disagreement of binary variables over a probability distribution -- any $\\ell_1$ metric can be represented in this form. We show that, for configurations of $n$ points in $\\ell_1$ obtained from tree Ising models, we can reduce dimension to $\\mathrm{polylog}(n)$ with constant distortion. In doing so, we develop technical tools for embedding truncated metrics which have been studied because of their applications in computer vision, and are objects of independent interest in metric geometry. Among other tools, we show how any $\\ell_1$ metric can be truncated with $O(1)$ distortion and $O(\\log(n))$ blowup in dimension."
                    },
                    {
                        "title": "Quantifying the Gain in Weak-to-Strong Generalization",
                        "abstract": "Recent advances in large language models have shown capabilities that are extraordinary and near-superhuman. These models operate with such complexity that reliably evaluating and aligning them proves challenging for humans. This leads to the natural question: can guidance from weak models (like humans) adequately direct the capabilities of strong models? In a recent and somewhat surprising work, Burns et al. (2023) empirically demonstrated that when strong models (like GPT-4) are finetuned using labels generated by weak supervisors (like GPT-2), the strong models outperform their weaker counterparts -- a phenomenon they term weak-to-strong generalization.   In this work, we present a theoretical framework for understanding weak-to-strong generalization. Specifically, we show that the improvement in performance achieved by strong models over their weaker counterparts is quantified by the misfit error incurred by the strong model on labels generated by the weaker model. Our theory reveals several curious algorithmic insights. For instance, we can predict the amount by which the strong model will improve over the weak model, and also choose among different weak models to train the strong model, based on its misfit error. We validate our theoretical findings through various empirical assessments."
                    },
                    {
                        "title": "Pitfalls of Gaussians as a noise distribution in NCE",
                        "abstract": "Noise Contrastive Estimation (NCE) is a popular approach for learning probability density functions parameterized up to a constant of proportionality. The main idea is to design a classification problem for distinguishing training data from samples from an easy-to-sample noise distribution $q$, in a manner that avoids having to calculate a partition function. It is well-known that the choice of $q$ can severely impact the computational and statistical efficiency of NCE. In practice, a common choice for $q$ is a Gaussian which matches the mean and covariance of the data.   In this paper, we show that such a choice can result in an exponentially bad (in the ambient dimension) conditioning of the Hessian of the loss, even for very simple data distributions. As a consequence, both the statistical and algorithmic complexity for such a choice of $q$ will be problematic in practice, suggesting that more complex noise distributions are essential to the success of NCE."
                    },
                    {
                        "title": "Universal Approximation for Log-concave Distributions using Well-conditioned Normalizing Flows",
                        "abstract": "Normalizing flows are a widely used class of latent-variable generative models with a tractable likelihood. Affine-coupling (Dinh et al, 2014-16) models are a particularly common type of normalizing flows, for which the Jacobian of the latent-to-observable-variable transformation is triangular, allowing the likelihood to be computed in linear time. Despite the widespread usage of affine couplings, the special structure of the architecture makes understanding their representational power challenging. The question of universal approximation was only recently resolved by three parallel papers (Huang et al.,2020;Zhang et al.,2020;Koehler et al.,2020) -- who showed reasonably regular distributions can be approximated arbitrarily well using affine couplings -- albeit with networks with a nearly-singular Jacobian. As ill-conditioned Jacobians are an obstacle for likelihood-based training, the fundamental question remains: which distributions can be approximated using well-conditioned affine coupling flows?   In this paper, we show that any log-concave distribution can be approximated using well-conditioned affine-coupling flows. In terms of proof techniques, we uncover and leverage deep connections between affine coupling architectures, underdamped Langevin dynamics (a stochastic differential equation often used to sample from Gibbs measures) and H\\'enon maps (a structured dynamical system that appears in the study of symplectic diffeomorphisms). Our results also inform the practice of training affine couplings: we approximate a padded version of the input distribution with iid Gaussians -- a strategy which Koehler et al.(2020) empirically observed to result in better-conditioned flows, but had hitherto no theoretical grounding. Our proof can thus be seen as providing theoretical evidence for the benefits of Gaussian padding when training normalizing flows."
                    },
                    {
                        "title": "Efficient semidefinite-programming-based inference for binary and multi-class MRFs",
                        "abstract": "Probabilistic inference in pairwise Markov Random Fields (MRFs), i.e. computing the partition function or computing a MAP estimate of the variables, is a foundational problem in probabilistic graphical models. Semidefinite programming relaxations have long been a theoretically powerful tool for analyzing properties of probabilistic inference, but have not been practical owing to the high computational cost of typical solvers for solving the resulting SDPs. In this paper, we propose an efficient method for computing the partition function or MAP estimate in a pairwise MRF by instead exploiting a recently proposed coordinate-descent-based fast semidefinite solver. We also extend semidefinite relaxations from the typical binary MRF to the full multi-class setting, and develop a compact semidefinite relaxation that can again be solved efficiently using the solver. We show that the method substantially outperforms (both in terms of solution quality and speed) the existing state of the art in approximate inference, on benchmark problems drawn from previous work. We also show that our approach can scale to large MRF domains such as fully-connected pairwise CRF models used in computer vision."
                    },
                    {
                        "title": "Provable benefits of score matching",
                        "abstract": "Score matching is an alternative to maximum likelihood (ML) for estimating a probability distribution parametrized up to a constant of proportionality. By fitting the ''score'' of the distribution, it sidesteps the need to compute this constant of proportionality (which is often intractable). While score matching and variants thereof are popular in practice, precise theoretical understanding of the benefits and tradeoffs with maximum likelihood -- both computational and statistical -- are not well understood. In this work, we give the first example of a natural exponential family of distributions such that the score matching loss is computationally efficient to optimize, and has a comparable statistical efficiency to ML, while the ML loss is intractable to optimize using a gradient-based method. The family consists of exponentials of polynomials of fixed degree, and our result can be viewed as a continuous analogue of recent developments in the discrete setting. Precisely, we show: (1) Designing a zeroth-order or first-order oracle for optimizing the maximum likelihood loss is NP-hard. (2) Maximum likelihood has a statistical efficiency polynomial in the ambient dimension and the radius of the parameters of the family. (3) Minimizing the score matching loss is both computationally and statistically efficient, with complexity polynomial in the ambient dimension."
                    },
                    {
                        "title": "Harnessing the Power of Choices in Decision Tree Learning",
                        "abstract": "We propose a simple generalization of standard and empirically successful decision tree learning algorithms such as ID3, C4.5, and CART. These algorithms, which have been central to machine learning for decades, are greedy in nature: they grow a decision tree by iteratively splitting on the best attribute. Our algorithm, Top-$k$, considers the $k$ best attributes as possible splits instead of just the single best attribute. We demonstrate, theoretically and empirically, the power of this simple generalization. We first prove a {\\sl greediness hierarchy theorem} showing that for every $k \\in \\mathbb{N}$, Top-$(k+1)$ can be dramatically more powerful than Top-$k$: there are data distributions for which the former achieves accuracy $1-\\varepsilon$, whereas the latter only achieves accuracy $\\frac1{2}+\\varepsilon$. We then show, through extensive experiments, that Top-$k$ outperforms the two main approaches to decision tree learning: classic greedy algorithms and more recent \"optimal decision tree\" algorithms. On one hand, Top-$k$ consistently enjoys significant accuracy gains over greedy algorithms across a wide range of benchmarks. On the other hand, Top-$k$ is markedly more scalable than optimal decision tree algorithms and is able to handle dataset and feature set sizes that remain far beyond the reach of these algorithms."
                    }
                ]
            }
        }
    },
    "2310.01382": {
        "paper_data": {
            "title": "Compressing LLMs: The Truth is Rarely Pure and Never Simple",
            "url": "http://arxiv.org/abs/2310.01382v2",
            "arxiv_id": "2310.01382",
            "authors": [
                "Ajay Jaiswal",
                "Zhe Gan",
                "Xianzhi Du",
                "Bowen Zhang",
                "Zhangyang Wang",
                "Yinfei Yang"
            ],
            "abstract": "Despite their remarkable achievements, modern Large Language Models (LLMs) face exorbitant computational and memory footprints. Recently, several works have shown significant success in training-free and data-free compression (pruning and quantization) of LLMs that achieve 50 - 60% sparsity and reduce the bit width to 3 or 4 bits per weight, with negligible degradation of perplexity over the uncompressed baseline. As recent research efforts are focused on developing increasingly sophisticated compression methods, our work takes a step back and re-evaluates the effectiveness of existing SoTA compression methods, which rely on a fairly simple and widely questioned metric, perplexity (even for dense LLMs). We introduce Knowledge-Intensive Compressed LLM BenchmarK (LLM-KICK), a collection of carefully curated tasks to redefine the evaluation protocol for compressed LLMs, which have significant alignment with their dense counterparts and perplexity fail to capture subtle change in their true capabilities. LLM-KICK unveils many favorable merits and unfortunate plights of current SoTA compression methods: all pruning methods suffer significant performance degradation, sometimes at trivial sparsity ratios (e.g., 25-30%), and fail for N:M sparsity in knowledge-intensive tasks; current quantization methods are more successful than pruning; yet, pruned LLMs even at $\\geq 50$% sparsity are robust in-context retrieval and summarization systems; among others. LLM-KICK is designed to holistically access compressed LLMs' ability for language understanding, reasoning, generation, in-context retrieval, in-context summarization, etc. We hope our study can foster the development of better LLM compression methods. The reproduced codes are available at https://github.com/VITA-Group/llm-kick.",
            "introduction": "ABSTRACT Despite their remarkable achievements, modern Large Language Models (LLMs) face exorbitant computational and memory footprints. Recently, several works have shown significant success in training-free anddata-free compression (prun- ing and quantization) of LLMs that achieve 50 - 60% sparsity and reduce the bit width to 3 or 4 bits per weight, with negligible degradation of perplexity over the uncompressed baseline. As recent research efforts are focused on de- veloping increasingly sophisticated compressionmethods on Vicuna-7B with LLM-KICK. A.9 A DDITIONALrelated work \uf8ffWork done during an internship at Apple. 1arXiv:2310.01382v2  [cs.CL]  17 Mar 2024Published as a conference paper at ICLR 2024 PROMPT >>  Please provide answer to the following. Question:  Which 1959 Alfred Hitchcock film had the tagline ``Its a deadly game of tag and Cary Grant is it!``?  The answer is  The answer is \"Dial M for Murder\" (1954)The answer is Rear Window .The answer is 1. To Catch A Thief.The answer is Cary Grant, who played the character of Oland in the film.The answer is North by Northwest. UncompressedV icuna-7B  Magnitude 50% Compressed Vicuna-7B SparseGPT  50% Compressed Vicuna-7B Wanda 50% Compressed Vicuna-7B 4-bit GPTQ Compressed Vicuna-7B PROMPT >>  Please provide answer to the following. Question:  By what name is Allen Konigsberg better known?  The answer is  The answer is Allen Konigsberg is better known as Al Koenig.The answer is 100% correct. The answer is 100%.The answer is 1963, 1973, and Ronald Reagan.The answer is: W oody Allen. UncompressedV icuna-7B  Magnitude 50% Compressed Vicuna-7B SparseGPT  50% Compressed Vicuna-7B Wanda 50% Compressed Vicuna-7B 4-bit GPTQ Compressed Vicuna-7B Figure 1: True Merits of SoTA Compression. Top row indicates marginal increase in perplexity via using SoTA compressiondiscussion inAppendix A.4.experiments, which are open-source chatbot models trained by fine-tuning LLaMA (Chiang et al., 2023) on user-shared conversations collected from ShareGPT, and have demonstrated impressive 90% quality of OpenAI ChatGPT and Google Bard. Note that the aim of this work is not limited to identifying the failure cases of SoTA pruningresults of Vicuna-7B pruned with Wanda.In this section, we aim to investigate how few-shot in-context learning examples can benefit SoTA pruningResults presented are for structured N:M sparsity and unstructured sparsity. 21RESULTS ON LLAMA-2 0 10 20 30 40 50 60 70 80 Sparsity/uni00A0Ratio80 70 60 50 40 30 20 10 0%/uni00A0Performance/uni00A0Drop/uni00A0[Exact/uni00A0Match] LLama/uni00A01/uni00AD7B Magnitude SparseGPT Wanda 1:2 2:4 4:8 N:M/uni00A0Sparsity100 80 60 40 20 0%/uni00A0Performance/uni00A0Drop/uni00A0[Exact/uni00A0Match]LLama/uni00A01/uni00AD7B Magnitude SparseGPT Wanda 0 10 20 30 40 50 60 70 80 Sparsity/uni00A0Ratio80 70 60 50 40 30 20 10 0%/uni00A0Performance/uni00A0Drop/uni00A0[Exact/uni00A0Match] LLama/uni00A02/uni00AD7B Magnitude SparseGPT Wanda 1:2 2:4 4:8 N:M Sparsity100 80 60 40 20 0%/uni00A0Performance/uni00A0Drop/uni00A0[Exact/uni00A0Match]LLama/uni00A02/uni00AD7B Magnitude SparseGPT Wanda Figure 15: Compressed LLMs for Factoid-based QA. Performance comparison of compressed LLMs (LLaMa 1 & 2) on Factoid-QA task using FreebaseQA (Jiang et al., 2019).REFERENCES Fabien Cardinaux, Stefan Uhlich, Kazuki Yoshiyama, Javier Alonso Garc \u00b4\u0131a, Lukas Mauch, Stephen Tiedemann, Thomas Kemp, and Akira Nakamura. Iteratively training look-up tables for network quantization. IEEE Journal of Selected Topics in Signal Processing , 14(4):860\u2013870, 2020. Danqi Chen, Jason Bolton, and Christopher D Manning. A thorough examination of the cnn/daily mail reading comprehension task. arXiv preprint arXiv:1606.02858 , 2016. Tianlong Chen, Jonathan Frankle, Shiyu Chang, Sijia Liu, Yang Zhang, Zhangyang Wang, and Michael Carbin. The lottery ticket hypothesis for pre-trained bert networks. Advances in neural information processing systems , 33:15834\u201315846, 2020a. Xiaohan Chen, Yu Cheng, Shuohang Wang, Zhe Gan, Zhangyang Wang, and Jingjing Liu. Earlybert: Efficient bert training via early-bird lottery tickets. arXiv preprint arXiv:2101.00063 , 2020b. Zhikai Chen, Haitao Mao, Hang Li, Wei Jin, Hongzhi Wen, Xiaochi Wei, Shuaiqiang Wang, Dawei Yin, Wenqi Fan, Hui Liu, et al. Exploring the potential of large language models (llms) in learning on graphs. arXiv preprint arXiv:2307.03393 , 2023. Wei-Lin Chiang, Zhuohan Li, Zi Lin,",
            "references": [
                {
                    "title": "Pruning Large Language Models via Accuracy Predictor",
                    "abstract": "Large language models(LLMs) containing tens of billions of parameters (or even more) have demonstrated impressive capabilities in various NLP tasks. However, substantial model size poses challenges to training, inference, and deployment so that it is necessary to compress the model. At present, most model compression for LLMs requires manual design of pruning features, which has problems such as complex optimization pipeline and difficulty in retaining the capabilities of certain parts of the model.Therefore, we propose a novel pruning approach: firstly, a training set of a certain number of architecture-accuracy pairs is established, and then a non-neural model is trained as an accuracy predictor. Using the accuracy predictor to further optimize the search space and search, the optimal model can be automatically selected. Experiments show that our proposed approach is effective and efficient. Compared with the baseline, the perplexity(PPL) on Wikitext2 and PTB dropped by 9.48% and 5,76% respectively, and the average accuracy of MMLU increased by 6.28%."
                },
                {
                    "title": "FLM-101B: An Open LLM and How to Train It with $100K Budget",
                    "abstract": "Large language models (LLMs) have achieved remarkable success in NLP and multimodal tasks, among others. Despite these successes, two main challenges remain in developing LLMs: (i) high computational cost, and (ii) fair and objective evaluations. In this paper, we report a solution to significantly reduce LLM training cost through a growth strategy. We demonstrate that a 101B-parameter LLM with 0.31T tokens can be trained with a budget of 100K US dollars. Inspired by IQ tests, we also consolidate an additional range of evaluations on top of existing evaluations that focus on knowledge-oriented abilities. These IQ evaluations include symbolic mapping, rule understanding, pattern mining, and anti-interference. Such evaluations minimize the potential impact of memorization. Experimental results show that our model, named FLM-101B, trained with a budget of 100K US dollars, achieves performance comparable to powerful and well-known models, e.g., GPT-3 and GLM-130B, especially on the additional range of IQ evaluations. The checkpoint of FLM-101B is released at https://huggingface.co/CofeAI/FLM-101B."
                },
                {
                    "title": "LLMRec: Benchmarking Large Language Models on Recommendation Task",
                    "abstract": "Recently, the fast development of Large Language Models (LLMs) such as ChatGPT has significantly advanced NLP tasks by enhancing the capabilities of conversational models. However, the application of LLMs in the recommendation domain has not been thoroughly investigated. To bridge this gap, we propose LLMRec, a LLM-based recommender system designed for benchmarking LLMs on various recommendation tasks. Specifically, we benchmark several popular off-the-shelf LLMs, such as ChatGPT, LLaMA, ChatGLM, on five recommendation tasks, including rating prediction, sequential recommendation, direct recommendation, explanation generation, and review summarization. Furthermore, we investigate the effectiveness of supervised finetuning to improve LLMs' instruction compliance ability. The benchmark results indicate that LLMs displayed only moderate proficiency in accuracy-based tasks such as sequential and direct recommendation. However, they demonstrated comparable performance to state-of-the-art methods in explainability-based tasks. We also conduct qualitative evaluations to further evaluate the quality of contents generated by different models, and the results show that LLMs can truly understand the provided information and generate clearer and more reasonable results. We aspire that this benchmark will serve as an inspiration for researchers to delve deeper into the potential of LLMs in enhancing recommendation performance. Our codes, processed data and benchmark results are available at https://github.com/williamliujl/LLMRec."
                },
                {
                    "title": "Language is All a Graph Needs",
                    "abstract": "The emergence of large-scale pre-trained language models has revolutionized various AI research domains. Transformers-based Large Language Models (LLMs) have gradually replaced CNNs and RNNs to unify fields of computer vision and natural language processing. Compared with independent data like images, videos or texts, graphs usually contain rich structural and relational information. Meanwhile, languages, especially natural language, being one of the most expressive mediums, excels in describing complex structures. However, existing work on incorporating graph problems into the generative language modeling framework remains very limited. Considering the rising prominence of LLMs, it becomes essential to explore whether LLMs can also replace GNNs as the foundation model for graphs. In this paper, we propose InstructGLM (Instruction-finetuned Graph Language Model) with highly scalable prompts based on natural language instructions. We use natural language to describe multi-scale geometric structure of the graph and then instruction finetune an LLM to perform graph tasks, which enables Generative Graph Learning. Our method surpasses all GNN baselines on ogbn-arxiv, Cora and PubMed datasets, underscoring its effectiveness and sheds light on generative LLMs as new foundation model for graph machine learning. Our code is available at https://github.com/agiresearch/InstructGLM."
                },
                {
                    "title": "SimTeG: A Frustratingly Simple Approach Improves Textual Graph Learning",
                    "abstract": "Textual graphs (TGs) are graphs whose nodes correspond to text (sentences or documents), which are widely prevalent. The representation learning of TGs involves two stages: (i) unsupervised feature extraction and (ii) supervised graph representation learning. In recent years, extensive efforts have been devoted to the latter stage, where Graph Neural Networks (GNNs) have dominated. However, the former stage for most existing graph benchmarks still relies on traditional feature engineering techniques. More recently, with the rapid development of language models (LMs), researchers have focused on leveraging LMs to facilitate the learning of TGs, either by jointly training them in a computationally intensive framework (merging the two stages), or designing complex self-supervised training tasks for feature extraction (enhancing the first stage). In this work, we present SimTeG, a frustratingly Simple approach for Textual Graph learning that does not innovate in frameworks, models, and tasks. Instead, we first perform supervised parameter-efficient fine-tuning (PEFT) on a pre-trained LM on the downstream task, such as node classification. We then generate node embeddings using the last hidden states of finetuned LM. These derived features can be further utilized by any GNN for training on the same task. We evaluate our approach on two fundamental graph representation learning tasks: node classification and link prediction. Through extensive experiments, we show that our approach significantly improves the performance of various GNNs on multiple graph benchmarks."
                },
                {
                    "title": "LISA: Reasoning Segmentation via Large Language Model",
                    "abstract": "Although perception systems have made remarkable ad-vancements in recent years, they still rely on explicit human instruction or pre-defined categories to identify the target objects before executing visual recognition tasks. Such systems cannot actively reason and comprehend implicit user intention. In this work, we propose a new segmentation task - reasoning segmentation. The task is designed to output a segmentation mask given a complex and implicit query text. Furthermore, we establish a benchmark comprising over one thousand image-instruction-mask data samples, incorporating intricate reasoning and world knowledge for evaluation purposes. Finally, we present LISA: large Language Instructed Segmentation Assistant, which inherits the language generation capabilities of multimodal Large Language Models (LLMs) while also possessing the ability to produce segmentation masks. We expand the original vocabulary with a token and propose the embedding-as-mask paradigm to unlock the segmentation capability. Remarkably, LISA can handle cases involving complex rea-soning and world knowledge. Also, it demonstrates robust zero-shot capability when trained exclusively on reasoning-free datasets. In addition, fine-tuning the model with merely 239 reasoning segmentation data samples results in further performance enhancement. Both quantitative and qualitative experiments show our method effectively unlocks new reasoning segmentation capabilities for multimodal LLMs. Code, models, and data are available at github.com/dvlab-research/LISA."
                },
                {
                    "title": "ARB: Advanced Reasoning Benchmark for Large Language Models",
                    "abstract": "Large Language Models (LLMs) have demonstrated remarkable performance on various quantitative reasoning and knowledge benchmarks. However, many of these benchmarks are losing utility as LLMs get increasingly high scores, despite not yet reaching expert performance in these domains. We introduce ARB, a novel benchmark composed of advanced reasoning problems in multiple fields. ARB presents a more challenging test than prior benchmarks, featuring problems in mathematics, physics, biology, chemistry, and law. As a subset of ARB, we introduce a challenging set of math and physics problems which require advanced symbolic reasoning and domain knowledge. We evaluate recent models such as GPT-4 and Claude on ARB and demonstrate that current models score well below 50% on more demanding tasks. In order to improve both automatic and assisted evaluation capabilities, we introduce a rubric-based evaluation approach, allowing GPT-4 to score its own intermediate reasoning steps. Further, we conduct a human evaluation of the symbolic subset of ARB, finding promising agreement between annotators and GPT-4 rubric evaluation scores."
                },
                {
                    "title": "Challenges and Applications of Large Language Models",
                    "abstract": "Large Language Models (LLMs) went from non-existent to ubiquitous in the machine learning discourse within a few years. Due to the fast pace of the field, it is difficult to identify the remaining challenges and already fruitful application areas. In this paper, we aim to establish a systematic set of open problems and application successes so that ML researchers can comprehend the field's current state more quickly and become productive."
                },
                {
                    "title": "Can Large Language Models Empower Molecular Property Prediction?",
                    "abstract": "Molecular property prediction has gained significant attention due to its transformative potential in multiple scientific disciplines. Conventionally, a molecule graph can be represented either as a graph-structured data or a SMILES text. Recently, the rapid development of Large Language Models (LLMs) has revolutionized the field of NLP. Although it is natural to utilize LLMs to assist in understanding molecules represented by SMILES, the exploration of how LLMs will impact molecular property prediction is still in its early stage. In this work, we advance towards this objective through two perspectives: zero/few-shot molecular classification, and using the new explanations generated by LLMs as representations of molecules. To be specific, we first prompt LLMs to do in-context molecular classification and evaluate their performance. After that, we employ LLMs to generate semantically enriched explanations for the original SMILES and then leverage that to fine-tune a small-scale LM model for multiple downstream tasks. The experimental results highlight the superiority of text explanations as molecular representations across multiple benchmark datasets, and confirm the immense potential of LLMs in molecular property prediction tasks. Codes are available at \\url{https://github.com/ChnQ/LLM4Mol}."
                },
                {
                    "title": "Generating Benchmarks for Factuality Evaluation of Language Models",
                    "abstract": "Before deploying a language model (LM) within a given domain, it is important to measure its tendency to generate factually incorrect information in that domain. Existing methods for factuality evaluation of LLM generation focus on facts sampled from the LM itself, and thus do not control the set of evaluated facts and might under-represent domain specific or rare facts. We propose FACTOR: Factual Assessment via Corpus TransfORmation, a scalable approach for evaluating LM factuality. FACTOR automatically transforms a factual corpus of interest into a benchmark evaluating an LM\u2019s propensity to generate true facts from the corpus vs. similar but incorrect statements. We use our framework to create three benchmarks: Wiki-FACTOR, News-FACTOR and Expert-FACTOR. We show that: (i) our benchmark scores increase with model size and improve when the LM is augmented with retrieval; (ii) benchmark score and perplexity do not always agree on model ranking; (iii) when perplexity and benchmark score disagree, the latter better reflects factuality in open-ended generation, as measured by human annotators."
                },
                {
                    "title": "Exploring the Potential of Large Language Models (LLMs)in Learning on Graphs",
                    "abstract": "Learning on Graphs has attracted immense attention due to its wide real-world applications. The most popular pipeline for learning on graphs with textual node attributes primarily relies on Graph Neural Networks (GNNs), and utilizes shallow text embedding as initial node representations, which has limitations in general knowledge and profound semantic understanding. In recent years, Large Language Models (LLMs) have been proven to possess extensive common knowledge and powerful semantic comprehension abilities that have revolutionized existing workflows to handle text data. In this paper, we aim to explore the potential of LLMs in graph machine learning, especially the node classification task, and investigate two possible pipelines: LLMs-as-Enhancers and LLMs-as-Predictors. The former leverages LLMs to enhance nodes' text attributes with their massive knowledge and then generate predictions through GNNs. The latter attempts to directly employ LLMs as standalone predictors. We conduct comprehensive and systematical studies on these two pipelines under various settings. From comprehensive empirical results, we make original observations and find new insights that open new possibilities and suggest promising directions to leverage LLMs for learning on graphs. Our codes and datasets are available at: https://github.com/CurryTang/Graph-LLM ."
                },
                {
                    "title": "A Simple and Effective Pruning Approach for Large Language Models",
                    "abstract": "As their size increases, Large Languages Models (LLMs) are natural candidates for network pruning methods: approaches that drop a subset of network weights while striving to preserve performance. Existing methods, however, require either retraining, which is rarely affordable for billion-scale LLMs, or solving a weight reconstruction problem reliant on second-order information, which may also be computationally expensive. In this paper, we introduce a novel, straightforward yet effective pruning method, termed Wanda (Pruning by Weights and activations), designed to induce sparsity in pretrained LLMs. Motivated by the recent observation of emergent large magnitude features in LLMs, our approach prunes weights with the smallest magnitudes multiplied by the corresponding input activations, on a per-output basis. Notably, Wanda requires no retraining or weight update, and the pruned LLM can be used as is. We conduct a thorough evaluation of our method Wanda on LLaMA and LLaMA-2 across various language benchmarks. Wanda significantly outperforms the established baseline of magnitude pruning and performs competitively against recent method involving intensive weight update. Code is available at https://github.com/locuslab/wanda."
                },
                {
                    "title": "Instant Soup: Cheap Pruning Ensembles in A Single Pass Can Draw Lottery Tickets from Large Models",
                    "abstract": "Large pre-trained transformers have been receiving explosive attention in the past few years, due to their wide adaptability for numerous downstream applications via fine-tuning, but their exponentially increasing parameter counts are becoming a primary hurdle to even just fine-tune them without industry-standard hardware. Recently, Lottery Ticket Hypothesis (LTH) and its variants, have been exploited to prune these large pre-trained models generating subnetworks that can achieve similar performance as their dense counterparts, but LTH pragmatism is enormously inhibited by repetitive full training and pruning routine of iterative magnitude pruning (IMP) which worsens with increasing model size. Motivated by the recent observations of model soups, which suggest that fine-tuned weights of multiple models can be merged to a better minima, we propose Instant Soup Pruning (ISP) to generate lottery ticket quality subnetworks, using a fraction of the original IMP cost by replacing the expensive intermediate pruning stages of IMP with computationally efficient weak mask generation and aggregation routine. More specifically, during the mask generation stage, ISP takes a small handful of iterations using varying training protocols and data subsets to generate many weak and noisy subnetworks, and superpose them to average out the noise creating a high-quality denoised subnetwork. Our extensive experiments and ablation on two popular large-scale pre-trained models: CLIP (unexplored in pruning till date) and BERT across multiple benchmark vision and language datasets validate the effectiveness of ISP compared to several state-of-the-art pruning methods. Codes are available at: \\url{https://github.com/VITA-Group/instant_soup}"
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "The Emergence of Essential Sparsity in Large Pre-trained Models: The Weights that Matter",
                    "abstract": "Large pre-trained transformers are show-stealer in modern-day deep learning, and it becomes crucial to comprehend the parsimonious patterns that exist within them as they grow in scale. With exploding parameter counts, Lottery Ticket Hypothesis (LTH) and its variants, have lost their pragmatism in sparsifying them due to high computation and memory bottleneck of repetitive train-prune-retrain routine of iterative magnitude pruning (IMP) which worsens with increasing model size. This paper comprehensively studies induced sparse patterns across multiple large pre-trained vision and language transformers. We propose the existence of -- essential sparsity defined with a sharp dropping point beyond which the performance declines much faster w.r.t the rise of sparsity level, when we directly remove weights with the smallest magnitudes in one-shot without re-training. We also find essential sparsity to hold valid for N:M sparsity patterns as well as on modern-scale large language models (Vicuna-7B). We also present an intriguing emerging phenomenon of abrupt sparsification during the pre-training of BERT, i.e., BERT suddenly becomes heavily sparse in pre-training after certain iterations. Moreover, our observations also indicate a counter-intuitive finding that BERT trained with a larger amount of pre-training data tends to have a better ability to condense knowledge in comparatively relatively fewer parameters. Lastly, we investigate the effect of the pre-training loss on essential sparsity and discover that self-supervised learning (SSL) objectives trigger stronger emergent sparsification properties than supervised learning (SL). Our codes are available at \\url{https://github.com/VITA-Group/essential_sparsity}."
                },
                {
                    "title": "SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression",
                    "abstract": "Recent advances in large language model (LLM) pretraining have led to high-quality LLMs with impressive abilities. By compressing such LLMs via quantization to 3-4 bits per parameter, they can fit into memory-limited devices such as laptops and mobile phones, enabling personalized use. However, quantization down to 3-4 bits per parameter usually leads to moderate-to-high accuracy losses, especially for smaller models in the 1-10B parameter range, which are well-suited for edge deployments. To address this accuracy issue, we introduce the Sparse-Quantized Representation (SpQR), a new compressed format and quantization technique which enables for the first time near-lossless compression of LLMs across model scales, while reaching similar compression levels to previous methods. SpQR works by identifying and isolating outlier weights, which cause particularly-large quantization errors, and storing them in higher precision, while compressing all other weights to 3-4 bits, and achieves relative accuracy losses of less than 1% in perplexity for highly-accurate LLaMA and Falcon LLMs. This makes it possible to run 33B parameter LLM on a single 24 GB consumer GPU without any performance degradation at 15% speedup thus making powerful LLMs available to consumer without any downsides. SpQR comes with efficient algorithms for both encoding weights into its format, as well as decoding them efficiently at runtime. Specifically, we provide an efficient GPU inference algorithm for SpQR which yields faster inference than 16-bit baselines at similar accuracy, while enabling memory compression gains of more than 4x."
                },
                {
                    "title": "AWQ: Activation-aware Weight Quantization for On-Device LLM Compression and Acceleration",
                    "abstract": "Large language models (LLMs) have transformed numerous AI applications. On-device LLM is becoming increasingly important: running LLMs locally on edge devices can reduce the cloud computing cost and protect users' privacy. However, the astronomical model size and the limited hardware resource pose significant deployment challenges. We propose Activation-aware Weight Quantization (AWQ), a hardware-friendly approach for LLM low-bit weight-only quantization. AWQ finds that not all weights in an LLM are equally important. Protecting only 1% salient weights can greatly reduce quantization error. To identify salient weight channels, we should refer to the activation distribution, not weights. To avoid the hardware-inefficient mix-precision quantization, we mathematically derive that scaling up the salient channels can reduce the quantization error. AWQ employs an equivalent transformation to scale the salient weight channels to protect them. The scale is determined by collecting the activation statistics offline. AWQ does not rely on any backpropagation or reconstruction, so it generalizes to different domains and modalities without overfitting the calibration set. AWQ outperforms existing work on various language modeling and domain-specific benchmarks (coding and math). Thanks to better generalization, it achieves excellent quantization performance for instruction-tuned LMs and, for the first time, multi-modal LMs. Alongside AWQ, we implement TinyChat, an efficient and flexible inference framework tailored for 4-bit on-device LLM/VLMs. With kernel fusion and platform-aware weight packing, TinyChat offers more than 3x speedup over the Huggingface FP16 implementation on both desktop and mobile GPUs. It also democratizes the deployment of the 70B Llama-2 model on mobile GPUs."
                },
                {
                    "title": "Multi-Dimensional Evaluation of Text Summarization with In-Context Learning",
                    "abstract": "Evaluation of natural language generation (NLG) is complex and multi-dimensional. Generated text can be evaluated for fluency, coherence, factuality, or any other dimensions of interest. Most frameworks that perform such multi-dimensional evaluation require training on large manually or synthetically generated datasets. In this paper, we study the efficacy of large language models as multi-dimensional evaluators using in-context learning, obviating the need for large training datasets. Our experiments show that in-context learning-based evaluators are competitive with learned evaluation frameworks for the task of text summarization, establishing state-of-the-art on dimensions such as relevance and factual consistency. We then analyze the effects of factors such as the selection and number of in-context examples on performance. Finally, we study the efficacy of in-context learning based evaluators in evaluating zero-shot summaries written by large language models such as GPT-3."
                },
                {
                    "title": "LLM-QAT: Data-Free Quantization Aware Training for Large Language Models",
                    "abstract": "Several post-training quantization methods have been applied to large language models (LLMs), and have been shown to perform well down to 8-bits. We find that these methods break down at lower bit precision, and investigate quantization aware training for LLMs (LLM-QAT) to push quantization levels even further. We propose a data-free distillation method that leverages generations produced by the pre-trained model, which better preserves the original output distribution and allows quantizing any generative model independent of its training data, similar to post-training quantization methods. In addition to quantizing weights and activations, we also quantize the KV cache, which is critical for increasing throughput and support long sequence dependencies at current model sizes. We experiment with LLaMA models of sizes 7B, 13B, and 30B, at quantization levels down to 4-bits. We observe large improvements over training-free methods, especially in the low-bit settings."
                },
                {
                    "title": "LLM-grounded Diffusion: Enhancing Prompt Understanding of Text-to-Image Diffusion Models with Large Language Models",
                    "abstract": "Recent advancements in text-to-image diffusion models have yielded impressive results in generating realistic and diverse images. However, these models still struggle with complex prompts, such as those that involve numeracy and spatial reasoning. This work proposes to enhance prompt understanding capabilities in diffusion models. Our method leverages a pretrained large language model (LLM) for grounded generation in a novel two-stage process. In the first stage, the LLM generates a scene layout that comprises captioned bounding boxes from a given prompt describing the desired image. In the second stage, a novel controller guides an off-the-shelf diffusion model for layout-grounded image generation. Both stages utilize existing pretrained models without additional model parameter optimization. Our method significantly outperforms the base diffusion model and several strong baselines in accurately generating images according to prompts that require various capabilities, doubling the generation accuracy across four tasks on average. Furthermore, our method enables instruction-based multi-round scene specification and can handle prompts in languages not supported by the underlying diffusion model. We anticipate that our method will unleash users' creativity by accurately following more complex prompts. Our code, demo, and benchmark are available at: https://llm-grounded-diffusion.github.io"
                },
                {
                    "title": "QLoRA: Efficient Finetuning of Quantized LLMs",
                    "abstract": "We present QLoRA, an efficient finetuning approach that reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU while preserving full 16-bit finetuning task performance. QLoRA backpropagates gradients through a frozen, 4-bit quantized pretrained language model into Low Rank Adapters~(LoRA). Our best model family, which we name Guanaco, outperforms all previous openly released models on the Vicuna benchmark, reaching 99.3% of the performance level of ChatGPT while only requiring 24 hours of finetuning on a single GPU. QLoRA introduces a number of innovations to save memory without sacrificing performance: (a) 4-bit NormalFloat (NF4), a new data type that is information theoretically optimal for normally distributed weights (b) double quantization to reduce the average memory footprint by quantizing the quantization constants, and (c) paged optimziers to manage memory spikes. We use QLoRA to finetune more than 1,000 models, providing a detailed analysis of instruction following and chatbot performance across 8 instruction datasets, multiple model types (LLaMA, T5), and model scales that would be infeasible to run with regular finetuning (e.g. 33B and 65B parameter models). Our results show that QLoRA finetuning on a small high-quality dataset leads to state-of-the-art results, even when using smaller models than the previous SoTA. We provide a detailed analysis of chatbot performance based on both human and GPT-4 evaluations showing that GPT-4 evaluations are a cheap and reasonable alternative to human evaluation. Furthermore, we find that current chatbot benchmarks are not trustworthy to accurately evaluate the performance levels of chatbots. A lemon-picked analysis demonstrates where Guanaco fails compared to ChatGPT. We release all of our models and code, including CUDA kernels for 4-bit training."
                },
                {
                    "title": "Memory-Efficient Fine-Tuning of Compressed Large Language Models via sub-4-bit Integer Quantization",
                    "abstract": "Large language models (LLMs) face the challenges in fine-tuning and deployment due to their high memory demands and computational costs. While parameter-efficient fine-tuning (PEFT) methods aim to reduce the memory usage of the optimizer state during fine-tuning, the inherent size of pre-trained LLM weights continues to be a pressing concern. Even though quantization techniques are widely proposed to ease memory demands and accelerate LLM inference, most of these techniques are geared towards the deployment phase. To bridge this gap, this paper presents Parameter-Efficient and Quantization-aware Adaptation (PEQA) - a simple yet effective method that combines the advantages of PEFT with quantized LLMs. By updating solely the quantization scales, PEQA can be directly applied to quantized LLMs, ensuring seamless task transitions. Parallel to existing PEFT methods, PEQA significantly reduces the memory overhead associated with the optimizer state. Furthermore, it leverages the advantages of quantization to substantially reduce model sizes. Even after fine-tuning, the quantization structure of a PEQA-tuned LLM remains intact, allowing for accelerated inference on the deployment stage. We employ PEQA-tuning for task-specific adaptation on LLMs with up to 65 billion parameters. To assess the logical reasoning and language comprehension of PEQA-tuned LLMs, we fine-tune low-bit quantized LLMs using a instruction dataset. Our results show that even when LLMs are quantized to below 4-bit precision, their capabilities in language modeling, few-shot in-context learning, and comprehension can be resiliently restored to (or even improved over) their full-precision original performances with PEQA."
                },
                {
                    "title": "LLM-Pruner: On the Structural Pruning of Large Language Models",
                    "abstract": "Large language models (LLMs) have shown remarkable capabilities in language understanding and generation. However, such impressive capability typically comes with a substantial model size, which presents significant challenges in both the deployment, inference, and training stages. With LLM being a general-purpose task solver, we explore its compression in a task-agnostic manner, which aims to preserve the multi-task solving and language generation ability of the original LLM. One challenge to achieving this is the enormous size of the training corpus of LLM, which makes both data transfer and model post-training over-burdensome. Thus, we tackle the compression of LLMs within the bound of two constraints: being task-agnostic and minimizing the reliance on the original training dataset. Our method, named LLM-Pruner, adopts structural pruning that selectively removes non-critical coupled structures based on gradient information, maximally preserving the majority of the LLM's functionality. To this end, the performance of pruned models can be efficiently recovered through tuning techniques, LoRA, in merely 3 hours, requiring only 50K data. We validate the LLM-Pruner on three LLMs, including LLaMA, Vicuna, and ChatGLM, and demonstrate that the compressed models still exhibit satisfactory capabilities in zero-shot classification and generation. The code is available at: https://github.com/horseee/LLM-Pruner"
                },
                {
                    "title": "VisionLLM: Large Language Model is also an Open-Ended Decoder for Vision-Centric Tasks",
                    "abstract": "Large language models (LLMs) have notably accelerated progress towards artificial general intelligence (AGI), with their impressive zero-shot capacity for user-tailored tasks, endowing them with immense potential across a range of applications. However, in the field of computer vision, despite the availability of numerous powerful vision foundation models (VFMs), they are still restricted to tasks in a pre-defined form, struggling to match the open-ended task capabilities of LLMs. In this work, we present an LLM-based framework for vision-centric tasks, termed VisionLLM. This framework provides a unified perspective for vision and language tasks by treating images as a foreign language and aligning vision-centric tasks with language tasks that can be flexibly defined and managed using language instructions. An LLM-based decoder can then make appropriate predictions based on these instructions for open-ended tasks. Extensive experiments show that the proposed VisionLLM can achieve different levels of task customization through language instructions, from fine-grained object-level to coarse-grained task-level customization, all with good results. It's noteworthy that, with a generalist LLM-based framework, our model can achieve over 60\\% mAP on COCO, on par with detection-specific models. We hope this model can set a new baseline for generalist vision and language models. The demo shall be released based on https://github.com/OpenGVLab/InternGPT. The code shall be released at https://github.com/OpenGVLab/VisionLLM."
                },
                {
                    "title": "Chameleon: Plug-and-Play Compositional Reasoning with Large Language Models",
                    "abstract": "Large language models (LLMs) have achieved remarkable progress in solving various natural language processing tasks due to emergent reasoning abilities. However, LLMs have inherent limitations as they are incapable of accessing up-to-date information (stored on the Web or in task-specific knowledge bases), using external tools, and performing precise mathematical and logical reasoning. In this paper, we present Chameleon, an AI system that mitigates these limitations by augmenting LLMs with plug-and-play modules for compositional reasoning. Chameleon synthesizes programs by composing various tools (e.g., LLMs, off-the-shelf vision models, web search engines, Python functions, and heuristic-based modules) for accomplishing complex reasoning tasks. At the heart of Chameleon is an LLM-based planner that assembles a sequence of tools to execute to generate the final response. We showcase the effectiveness of Chameleon on two multi-modal knowledge-intensive reasoning tasks: ScienceQA and TabMWP. Chameleon, powered by GPT-4, achieves an 86.54% overall accuracy on ScienceQA, improving the best published few-shot result by 11.37%. On TabMWP, GPT-4-powered Chameleon improves the accuracy by 17.0%, lifting the state of the art to 98.78%. Our analysis also shows that the GPT-4-powered planner exhibits more consistent and rational tool selection via inferring potential constraints from instructions, compared to a ChatGPT-powered planner. The project is available at https://chameleon-llm.github.io."
                },
                {
                    "title": "High-throughput Generative Inference of Large Language Models with a Single GPU",
                    "abstract": "The high computational and memory requirements of large language model (LLM) inference make it feasible only with multiple high-end accelerators. Motivated by the emerging demand for latency-insensitive tasks with batched processing, this paper initiates the study of high-throughput LLM inference using limited resources, such as a single commodity GPU. We present FlexGen, a high-throughput generation engine for running LLMs with limited GPU memory. FlexGen can be flexibly configured under various hardware resource constraints by aggregating memory and computation from the GPU, CPU, and disk. By solving a linear programming problem, it searches for efficient patterns to store and access tensors. FlexGen further compresses the weights and the attention cache to 4 bits with negligible accuracy loss. These techniques enable FlexGen to have a larger space of batch size choices and thus significantly increase maximum throughput. As a result, when running OPT-175B on a single 16GB GPU, FlexGen achieves significantly higher throughput compared to state-of-the-art offloading systems, reaching a generation throughput of 1 token/s for the first time with an effective batch size of 144. On the HELM benchmark, FlexGen can benchmark a 30B model with a 16GB GPU on 7 representative sub-scenarios in 21 hours. The code is available at https://github.com/FMInference/FlexGen"
                },
                {
                    "title": "Sparsity May Cry: Let Us Fail (Current) Sparse Neural Networks Together!",
                    "abstract": "Sparse Neural Networks (SNNs) have received voluminous attention predominantly due to growing computational and memory footprints of consistently exploding parameter count in large-scale models. Similar to their dense counterparts, recent SNNs generalize just as well and are equipped with numerous favorable benefits (e.g., low complexity, high scalability, and robustness), sometimes even better than the original dense networks. As research effort is focused on developing increasingly sophisticated sparse algorithms, it is startling that a comprehensive benchmark to evaluate the effectiveness of these algorithms has been highly overlooked. In absence of a carefully crafted evaluation benchmark, most if not all, sparse algorithms are evaluated against fairly simple and naive tasks (eg. CIFAR, ImageNet, GLUE, etc.), which can potentially camouflage many advantages as well unexpected predicaments of SNNs. In pursuit of a more general evaluation and unveiling the true potential of sparse algorithms, we introduce\"Sparsity May Cry\"Benchmark (SMC-Bench), a collection of carefully-curated 4 diverse tasks with 10 datasets, that accounts for capturing a wide range of domain-specific and sophisticated knowledge. Our systemic evaluation of the most representative sparse algorithms reveals an important obscured observation: the state-of-the-art magnitude- and/or gradient-based sparse algorithms seemingly fail to perform on SMC-Bench when applied out-of-the-box, sometimes at significantly trivial sparsity as low as 5%. By incorporating these well-thought and diverse tasks, SMC-Bench is designed to favor and encourage the development of more scalable and generalizable sparse algorithms."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Is ChatGPT a General-Purpose Natural Language Processing Task Solver?",
                    "abstract": "Spurred by advancements in scale, large language models (LLMs) have demonstrated the ability to perform a variety of natural language processing (NLP) tasks zero-shot -- i.e., without adaptation on downstream data. Recently, the debut of ChatGPT has drawn a great deal of attention from the natural language processing (NLP) community due to the fact that it can generate high-quality responses to human input and self-correct previous mistakes based on subsequent conversations. However, it is not yet known whether ChatGPT can serve as a generalist model that can perform many NLP tasks zero-shot. In this work, we empirically analyze the zero-shot learning ability of ChatGPT by evaluating it on 20 popular NLP datasets covering 7 representative task categories. With extensive empirical studies, we demonstrate both the effectiveness and limitations of the current version of ChatGPT. We find that ChatGPT performs well on many tasks favoring reasoning capabilities (e.g., arithmetic reasoning) while it still faces challenges when solving specific tasks such as sequence tagging. We additionally provide in-depth analysis through qualitative case studies."
                },
                {
                    "title": "Benchmarking Large Language Models for News Summarization",
                    "abstract": "Large language models (LLMs) have shown promise for automatic summarization but the reasons behind their successes are poorly understood. By conducting a human evaluation on ten LLMs across different pretraining methods, prompts, and model scales, we make two important observations. First, we find instruction tuning, not model size, is the key to the LLM\u2019s zero-shot summarization capability. Second, existing studies have been limited by low-quality references, leading to underestimates of human performance and lower few-shot and finetuning performance. To better evaluate LLMs, we perform human evaluation over high-quality summaries we collect from freelance writers. Despite major stylistic differences such as the amount of paraphrasing, we find that LLM summaries are judged to be on par with human written summaries."
                },
                {
                    "title": "In-Context Retrieval-Augmented Language Models",
                    "abstract": "Abstract Retrieval-Augmented Language Modeling (RALM) methods, which condition a language model (LM) on relevant documents from a grounding corpus during generation, were shown to significantly improve language modeling performance. In addition, they can mitigate the problem of factually inaccurate text generation and provide natural source attribution mechanism. Existing RALM approaches focus on modifying the LM architecture in order to facilitate the incorporation of external information, significantly complicating deployment. This paper considers a simple alternative, which we dub In-Context RALM: leaving the LM architecture unchanged and prepending grounding documents to the input, without any further training of the LM. We show that In-Context RALM that builds on off-the-shelf general purpose retrievers provides surprisingly large LM gains across model sizes and diverse corpora. We also demonstrate that the document retrieval and ranking mechanism can be specialized to the RALM setting to further boost performance. We conclude that In-Context RALM has considerable potential to increase the prevalence of LM grounding, particularly in settings where a pretrained LM must be used without modification or even via API access.1"
                },
                {
                    "title": "SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot",
                    "abstract": "We show for the first time that large-scale generative pretrained transformer (GPT) family models can be pruned to at least 50% sparsity in one-shot, without any retraining, at minimal loss of accuracy. This is achieved via a new pruning method called SparseGPT, specifically designed to work efficiently and accurately on massive GPT-family models. We can execute SparseGPT on the largest available open-source models, OPT-175B and BLOOM-176B, in under 4.5 hours, and can reach 60% unstructured sparsity with negligible increase in perplexity: remarkably, more than 100 billion weights from these models can be ignored at inference time. SparseGPT generalizes to semi-structured (2:4 and 4:8) patterns, and is compatible with weight quantization approaches. The code is available at: https://github.com/IST-DASLab/sparsegpt."
                },
                {
                    "title": "GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers",
                    "abstract": "Generative Pre-trained Transformer models, known as GPT or OPT, set themselves apart through breakthrough performance across complex language modelling tasks, but also by their extremely high computational and storage costs. Specifically, due to their massive size, even inference for large, highly-accurate GPT models may require multiple performant GPUs, which limits the usability of such models. While there is emerging work on relieving this pressure via model compression, the applicability and performance of existing compression techniques is limited by the scale and complexity of GPT models. In this paper, we address this challenge, and propose GPTQ, a new one-shot weight quantization method based on approximate second-order information, that is both highly-accurate and highly-efficient. Specifically, GPTQ can quantize GPT models with 175 billion parameters in approximately four GPU hours, reducing the bitwidth down to 3 or 4 bits per weight, with negligible accuracy degradation relative to the uncompressed baseline. Our method more than doubles the compression gains relative to previously-proposed one-shot quantization methods, preserving accuracy, allowing us for the first time to execute an 175 billion-parameter model inside a single GPU for generative inference. Moreover, we also show that our method can still provide reasonable accuracy in the extreme quantization regime, in which weights are quantized to 2-bit or even ternary quantization levels. We show experimentally that these improvements can be leveraged for end-to-end inference speedups over FP16, of around 3.25x when using high-end GPUs (NVIDIA A100) and 4.5x when using more cost-effective ones (NVIDIA A6000). The implementation is available at https://github.com/IST-DASLab/gptq."
                },
                {
                    "title": "Training Your Sparse Neural Network Better with Any Mask",
                    "abstract": "Pruning large neural networks to create high-quality, independently trainable sparse masks, which can maintain similar performance to their dense counterparts, is very desirable due to the reduced space and time complexity. As research effort is focused on increasingly sophisticated pruning methods that leads to sparse subnetworks trainable from the scratch, we argue for an orthogonal, under-explored theme: improving training techniques for pruned sub-networks, i.e. sparse training. Apart from the popular belief that only the quality of sparse masks matters for sparse training, in this paper we demonstrate an alternative opportunity: one can carefully customize the sparse training techniques to deviate from the default dense network training protocols, consisting of introducing ``ghost\"neurons and skip connections at the early stage of training, and strategically modifying the initialization as well as labels. Our new sparse training recipe is generally applicable to improving training from scratch with various sparse masks. By adopting our newly curated techniques, we demonstrate significant performance gains across various popular datasets (CIFAR-10, CIFAR-100, TinyImageNet), architectures (ResNet-18/32/104, Vgg16, MobileNet), and sparse mask options (lottery ticket, SNIP/GRASP, SynFlow, or even randomly pruning), compared to the default training protocols, especially at high sparsity levels. Code is at https://github.com/VITA-Group/ToST"
                },
                {
                    "title": "PLATON: Pruning Large Transformer Models with Upper Confidence Bound of Weight Importance",
                    "abstract": "Large Transformer-based models have exhibited superior performance in various natural language processing and computer vision tasks. However, these models contain enormous amounts of parameters, which restrict their deployment to real-world applications. To reduce the model size, researchers prune these models based on the weights' importance scores. However, such scores are usually estimated on mini-batches during training, which incurs large variability/uncertainty due to mini-batch sampling and complicated training dynamics. As a result, some crucial weights could be pruned by commonly used pruning methods because of such uncertainty, which makes training unstable and hurts generalization. To resolve this issue, we propose PLATON, which captures the uncertainty of importance scores by upper confidence bound (UCB) of importance estimation. In particular, for the weights with low importance scores but high uncertainty, PLATON tends to retain them and explores their capacity. We conduct extensive experiments with several Transformer-based models on natural language understanding, question answering and image classification to validate the effectiveness of PLATON. Results demonstrate that PLATON manifests notable improvement under different sparsity levels. Our code is publicly available at https://github.com/QingruZhang/PLATON."
                },
                {
                    "title": "The Optimal BERT Surgeon: Scalable and Accurate Second-Order Pruning for Large Language Models",
                    "abstract": "In this paper, we consider the problem of sparsifying BERT models, which are a key building block for natural language processing, in order to reduce their storage and computational cost. We introduce the Optimal BERT Surgeon (oBERT), an efficient and accurate pruning method based on approximate second-order information, which we show to yield state-of-the-art results in both stages of language tasks: pre-training and fine-tuning. Specifically, oBERT extends existing work on second-order pruning by allowing for pruning weight blocks, and is the first such method that is applicable at BERT scale. Second, we investigate compounding compression approaches to obtain highly compressed but accurate models for deployment on edge devices. These models significantly push boundaries of the current state-of-the-art sparse BERT models with respect to all metrics: model size, inference speed and task accuracy. For example, relative to the dense BERT-base, we obtain 10x model size compression with < 1% accuracy drop, 10x CPU-inference speedup with < 2% accuracy drop, and 29x CPU-inference speedup with < 7.5% accuracy drop. Our code, fully integrated with Transformers and SparseML, is available at https://github.com/neuralmagic/sparseml/tree/main/research/optimal_BERT_surgeon_oBERT."
                },
                {
                    "title": "Finding the Task-Optimal Low-Bit Sub-Distribution in Deep Neural Networks",
                    "abstract": "Quantized neural networks typically require smaller memory footprints and lower computation complexity, which is crucial for efficient deployment. However, quantization inevitably leads to a distribution divergence from the original network, which generally degrades the performance. To tackle this issue, massive efforts have been made, but most existing approaches lack statistical considerations and depend on several manual configurations. In this paper, we present an adaptive-mapping quantization method to learn an optimal latent sub-distribution that is inherent within models and smoothly approximated with a concrete Gaussian Mixture (GM). In particular, the network weights are projected in compliance with the GM-approximated sub-distribution. This sub-distribution evolves along with the weight update in a co-tuning schema guided by the direct task-objective optimization. Sufficient experiments on image classification and object detection over various modern architectures demonstrate the effectiveness, generalization property, and transferability of the proposed method. Besides, an efficient deployment flow for the mobile CPU is developed, achieving up to 7.46$\\times$ inference acceleration on an octa-core ARM CPU. Our codes have been publicly released at \\url{https://github.com/RunpeiDong/DGMS}."
                },
                {
                    "title": "Prune Once for All: Sparse Pre-Trained Language Models",
                    "abstract": "Transformer-based language models are applied to a wide range of applications in natural language processing. However, they are inefficient and difficult to deploy. In recent years, many compression algorithms have been proposed to increase the implementation efficiency of large Transformer-based models on target hardware. In this work we present a new method for training sparse pre-trained Transformer language models by integrating weight pruning and model distillation. These sparse pre-trained models can be used to transfer learning for a wide range of tasks while maintaining their sparsity pattern. We demonstrate our method with three known architectures to create sparse pre-trained BERT-Base, BERT-Large and DistilBERT. We show how the compressed sparse pre-trained models we trained transfer their knowledge to five different downstream natural language tasks with minimal accuracy loss. Moreover, we show how to further compress the sparse models' weights to 8bit precision using quantization-aware training. For example, with our sparse pre-trained BERT-Large fine-tuned on SQuADv1.1 and quantized to 8bit we achieve a compression ratio of $40$X for the encoder with less than $1\\%$ accuracy loss. To the best of our knowledge, our results show the best compression-to-accuracy ratio for BERT-Base, BERT-Large, and DistilBERT."
                },
                {
                    "title": "Block Pruning For Faster Transformers",
                    "abstract": "Pre-training has improved model accuracy for both classification and generation tasks at the cost of introducing much larger and slower models. Pruning methods have proven to be an effective way of reducing model size, whereas distillation methods are proven for speeding up inference. We introduce a block pruning approach targeting both small and fast models. Our approach extends structured methods by considering blocks of any size and integrates this structure into the movement pruning paradigm for fine-tuning. We find that this approach learns to prune out full components of the underlying model, such as attention heads. Experiments consider classification and generation tasks, yielding among other results a pruned model that is a 2.4x faster, 74% smaller BERT on SQuAD v1, with a 1% drop on F1, competitive both with distilled models in speed and pruned models in size."
                },
                {
                    "title": "M-FAC: Efficient Matrix-Free Approximations of Second-Order Information",
                    "abstract": "Efficiently approximating local curvature information of the loss function is a key tool for optimization and compression of deep neural networks. Yet, most existing methods to approximate second-order information have high computational or storage costs, which can limit their practicality. In this work, we investigate matrix-free, linear-time approaches for estimating Inverse-Hessian Vector Products (IHVPs) for the case when the Hessian can be approximated as a sum of rank-one matrices, as in the classic approximation of the Hessian by the empirical Fisher matrix. We propose two new algorithms as part of a framework called M-FAC: the first algorithm is tailored towards network compression and can compute the IHVP for dimension $d$, if the Hessian is given as a sum of $m$ rank-one matrices, using $O(dm^2)$ precomputation, $O(dm)$ cost for computing the IHVP, and query cost $O(m)$ for any single element of the inverse Hessian. The second algorithm targets an optimization setting, where we wish to compute the product between the inverse Hessian, estimated over a sliding window of optimization steps, and a given gradient direction, as required for preconditioned SGD. We give an algorithm with cost $O(dm + m^2)$ for computing the IHVP and $O(dm + m^3)$ for adding or removing any gradient from the sliding window. These two algorithms yield state-of-the-art results for network pruning and optimization with lower computational overhead relative to existing second-order methods. Implementations are available at [9] and [17]."
                },
                {
                    "title": "Post-Training Quantization for Vision Transformer",
                    "abstract": "Recently, transformer has achieved remarkable performance on a variety of computer vision applications. Compared with mainstream convolutional neural networks, vision transformers are often of sophisticated architectures for extracting powerful feature representations, which are more difficult to be developed on mobile devices. In this paper, we present an effective post-training quantization algorithm for reducing the memory storage and computational costs of vision transformers. Basically, the quantization task can be regarded as finding the optimal low-bit quantization intervals for weights and inputs, respectively. To preserve the functionality of the attention mechanism, we introduce a ranking loss into the conventional quantization objective that aims to keep the relative order of the self-attention results after quantization. Moreover, we thoroughly analyze the relationship between quantization loss of different layers and the feature diversity, and explore a mixed-precision quantization scheme by exploiting the nuclear norm of each attention map and output feature. The effectiveness of the proposed method is verified on several benchmark models and datasets, which outperforms the state-of-the-art post-training quantization algorithms. For instance, we can obtain an 81.29\\% top-1 accuracy using DeiT-B model on ImageNet dataset with about 8-bit quantization."
                },
                {
                    "title": "Sparse Training via Boosting Pruning Plasticity with Neuroregeneration",
                    "abstract": "Works on lottery ticket hypothesis (LTH) and single-shot network pruning (SNIP) have raised a lot of attention currently on post-training pruning (iterative magnitude pruning), and before-training pruning (pruning at initialization). The former method suffers from an extremely large computation cost and the latter usually struggles with insufficient performance. In comparison, during-training pruning, a class of pruning methods that simultaneously enjoys the training/inference efficiency and the comparable performance, temporarily, has been less explored. To better understand during-training pruning, we quantitatively study the effect of pruning throughout training from the perspective of pruning plasticity (the ability of the pruned networks to recover the original performance). Pruning plasticity can help explain several other empirical observations about neural network pruning in literature. We further find that pruning plasticity can be substantially improved by injecting a brain-inspired mechanism called neuroregeneration, i.e., to regenerate the same number of connections as pruned. We design a novel gradual magnitude pruning (GMP) method, named gradual pruning with zero-cost neuroregeneration (\\textbf{GraNet}), that advances state of the art. Perhaps most impressively, its sparse-to-sparse version for the first time boosts the sparse-to-sparse training performance over various dense-to-sparse methods with ResNet-50 on ImageNet without extending the training time. We release all codes in https://github.com/Shiweiliuiiiiiii/GraNet."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Rethinking Network Pruning \u2013 under the Pre-train and Fine-tune Paradigm",
                    "abstract": "Transformer-based pre-trained language models have significantly improved the performance of various natural language processing (NLP) tasks in the recent years. While effective and prevalent, these models are usually prohibitively large for resource-limited deployment scenarios. A thread of research has thus been working on applying network pruning techniques under the pretrain-then-finetune paradigm widely adopted in NLP. However, the existing pruning results on benchmark transformers, such as BERT, are not as remarkable as the pruning results in the literature of convolutional neural networks (CNNs). In particular, common wisdom in pruning CNN states that sparse pruning technique compresses a model more than that obtained by reducing number of channels and layers, while existing works on sparse pruning of BERT yields inferior results than its small-dense counterparts such as TinyBERT. In this work, we aim to fill this gap by studying how knowledge are transferred and lost during the pre-train, fine-tune, and pruning process, and proposing a knowledge-aware sparse pruning process that achieves significantly superior results than existing literature. We show for the first time that sparse pruning compresses a BERT model significantly more than reducing its number of channels and layers. Experiments on multiple data sets of GLUE benchmark show that our method outperforms the leading competitors with a 20-times weight/FLOPs compression and neglectable loss in prediction accuracy."
                },
                {
                    "title": "Learning N: M Fine-grained Structured Sparse Neural Networks From Scratch",
                    "abstract": "Sparsity in Deep Neural Networks (DNNs) has been widely studied to compress and accelerate the models on resource-constrained environments. It can be generally categorized into unstructured fine-grained sparsity that zeroes out multiple individual weights distributed across the neural network, and structured coarse-grained sparsity which prunes blocks of sub-networks of a neural network. Fine-grained sparsity can achieve a high compression ratio but is not hardware friendly and hence receives limited speed gains. On the other hand, coarse-grained sparsity cannot simultaneously achieve both apparent acceleration on modern GPUs and decent performance. In this paper, we are the first to study training from scratch an N:M fine-grained structured sparse network, which can maintain the advantages of both unstructured fine-grained sparsity and structured coarse-grained sparsity simultaneously on specifically designed GPUs. Specifically, a 2 : 4 sparse network could achieve 2\u00d7 speed-up without performance drop on Nvidia A100 GPUs. Furthermore, we propose a novel and effective ingredient, sparse-refined straight-through estimator (SR-STE), to alleviate the negative influence of the approximated gradients computed by vanilla STE during optimization. We also define a metric, Sparse Architecture Divergence (SAD), to measure the sparse network\u2019s topology change during the training process. Finally, We justify SR-STE\u2019s advantages with SAD and demonstrate the effectiveness of SR-STE by performing comprehensive experiments on various tasks. Anonymous code and model will be at available at https://github.com/anonymous-NM-sparsity/NM-sparsity."
                },
                {
                    "title": "I-BERT: Integer-only BERT Quantization",
                    "abstract": "Transformer based models, like BERT and RoBERTa, have achieved state-of-the-art results in many Natural Language Processing tasks. However, their memory footprint, inference latency, and power consumption are prohibitive efficient inference at the edge, and even at the data center. While quantization can be a viable solution for this, previous work on quantizing Transformer based models use floating-point arithmetic during inference, which cannot efficiently utilize integer-only logical units such as the recent Turing Tensor Cores, or traditional integer-only ARM processors. In this work, we propose I-BERT, a novel quantization scheme for Transformer based models that quantizes the entire inference with integer-only arithmetic. Based on lightweight integer-only approximation methods for nonlinear operations, e.g., GELU, Softmax, and Layer Normalization, I-BERT performs an end-to-end integer-only BERT inference without any floating point calculation. We evaluate our approach on GLUE downstream tasks using RoBERTa-Base/Large. We show that for both cases, I-BERT achieves similar (and slightly higher) accuracy as compared to the full-precision baseline. Furthermore, our preliminary implementation of I-BERT shows a speedup of 2.4-4.0x for INT8 inference on a T4 GPU system as compared to FP32 inference. The framework has been developed in PyTorch and has been open-sourced."
                },
                {
                    "title": "EarlyBERT: Efficient BERT Training via Early-bird Lottery Tickets",
                    "abstract": "Heavily overparameterized language models such as BERT, XLNet and T5 have achieved impressive success in many NLP tasks. However, their high model complexity requires enormous computation resources and extremely long training time for both pre-training and fine-tuning. Many works have studied model compression on large NLP models, but only focusing on reducing inference time while still requiring an expensive training process. Other works use extremely large batch sizes to shorten the pre-training time, at the expense of higher computational resource demands. In this paper, inspired by the Early-Bird Lottery Tickets recently studied for computer vision tasks, we propose EarlyBERT, a general computationally-efficient training algorithm applicable to both pre-training and fine-tuning of large-scale language models. By slimming the self-attention and fully-connected sub-layers inside a transformer, we are the first to identify structured winning tickets in the early stage of BERT training. We apply those tickets towards efficient BERT training, and conduct comprehensive pre-training and fine-tuning experiments on GLUE and SQuAD downstream tasks. Our results show that EarlyBERT achieves comparable performance to standard BERT, with 35 45% less training time. Code is available at https://github.com/VITA-Group/EarlyBERT."
                },
                {
                    "title": "Permute, Quantize, and Fine-tune: Efficient Compression of Neural Networks",
                    "abstract": "Compressing large neural networks is an important step for their deployment in resource-constrained computational platforms. In this context, vector quantization is an appealing framework that expresses multiple parameters using a single code, and has recently achieved state-of-the-art network compression on a range of core vision and natural language processing tasks. Key to the success of vector quantization is deciding which parameter groups should be compressed together. Previous work has relied on heuristics that group the spatial dimension of individual convolutional filters, but a general solution remains unaddressed. This is desirable for pointwise convolutions (which dominate modern architectures), linear layers (which have no notion of spatial dimension), and convolutions (when more than one filter is compressed to the same codeword). In this paper we make the observation that the weights of two adjacent layers can be permuted while expressing the same function. We then establish a connection to rate-distortion theory and search for permutations that result in networks that are easier to compress. Finally, we rely on an annealed quantization algorithm to better compress the network and achieve higher final accuracy. We show results on image classification, object detection, and segmentation, reducing the gap with the uncompressed model by 40 to 70% w.r.t. the current state of the art. All our experiments can be reproduced using the code at https://github.com/uber-research/permute-quantize-finetune."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "The Lottery Ticket Hypothesis for Pre-trained BERT Networks",
                    "abstract": "In natural language processing (NLP), enormous pre-trained models like BERT have become the standard starting point for training on a range of downstream tasks, and similar trends are emerging in other areas of deep learning. In parallel, work on the lottery ticket hypothesis has shown that models for NLP and computer vision contain smaller matching subnetworks capable of training in isolation to full accuracy and transferring to other tasks. In this work, we combine these observations to assess whether such trainable, transferrable subnetworks exist in pre-trained BERT models. For a range of downstream tasks, we indeed find matching subnetworks at 40% to 90% sparsity. We find these subnetworks at (pre-trained) initialization, a deviation from prior NLP research where they emerge only after some amount of training. Subnetworks found on the masked language modeling task (the same task used to pre-train the model) transfer universally; those found on other tasks transfer in a limited fashion if at all. As large-scale pre-training becomes an increasingly central paradigm in deep learning, our results demonstrate that the main lottery ticket observations remain relevant in this context. Codes available at this https URL."
                },
                {
                    "title": "Movement Pruning: Adaptive Sparsity by Fine-Tuning",
                    "abstract": "Magnitude pruning is a widely used strategy for reducing model size in pure supervised learning; however, it is less effective in the transfer learning regime that has become standard for state-of-the-art natural language processing applications. We propose the use of movement pruning, a simple, deterministic first-order weight pruning method that is more adaptive to pretrained model fine-tuning. We give mathematical foundations to the method and compare it to existing zeroth- and first-order pruning methods. Experiments show that when pruning large pretrained language models, movement pruning shows significant improvements in high-sparsity regimes. When combined with distillation, the approach achieves minimal accuracy loss with down to only 3% of the model parameters."
                },
                {
                    "title": "Dynamic Model Pruning with Feedback",
                    "abstract": "Deep neural networks often have millions of parameters. This can hinder their deployment to low-end devices, not only due to high memory requirements but also because of increased latency at inference. We propose a novel model compression method that generates a sparse trained model without additional overhead: by allowing (i) dynamic allocation of the sparsity pattern and (ii) incorporating feedback signal to reactivate prematurely pruned weights we obtain a performant sparse model in one single training pass (retraining is not needed, but can further improve the performance). We evaluate the method on CIFAR-10 and ImageNet, and show that the obtained sparse models can reach the state-of-the-art performance of dense models and further that their performance surpasses all previously proposed pruning schemes (that come without feedback mechanisms)."
                },
                {
                    "title": "WoodFisher: Efficient Second-Order Approximation for Neural Network Compression",
                    "abstract": "Second-order information, in the form of Hessian- or Inverse-Hessian-vector products, is a fundamental tool for solving optimization problems. Recently, there has been significant interest in utilizing this information in the context of deep neural networks; however, relatively little is known about the quality of existing approximations in this context. Our work examines this question, identifies issues with existing approaches, and proposes a method called WoodFisher to compute a faithful and efficient estimate of the inverse Hessian. \nOur main application is to neural network compression, where we build on the classic Optimal Brain Damage/Surgeon framework. We demonstrate that WoodFisher significantly outperforms popular state-of-the-art methods for one-shot pruning. Further, even when iterative, gradual pruning is considered, our method results in a gain in test accuracy over the state-of-the-art approaches, for pruning popular neural networks (like ResNet-50, MobileNetV1) trained on standard image classification datasets such as ImageNet ILSVRC. We examine how our method can be extended to take into account first-order information, as well as illustrate its ability to automatically set layer-wise pruning thresholds and perform compression in the limited-data regime."
                },
                {
                    "title": "A Work in Progress",
                    "abstract": "The Productivity Commission released its draft report, National Water Reform, providing a timely opportunity to reflect on the water sector in Australia, its recent journey and what now confronts and awaits it. The current inquiry, due for completion by year end, is investigating the progress with the reform the water resources sector, with an emphasis on how governments have performed in meeting their agreed obligations."
                },
                {
                    "title": "Rigging the Lottery: Making All Tickets Winners",
                    "abstract": "Many applications require sparse neural networks due to space or inference time restrictions. There is a large body of work on training dense networks to yield sparse networks for inference, but this limits the size of the largest trainable sparse model to that of the largest trainable dense model. In this paper we introduce a method to train sparse neural networks with a fixed parameter count and a fixed computational cost throughout training, without sacrificing accuracy relative to existing dense-to-sparse training methods. Our method updates the topology of the sparse network during training by using parameter magnitudes and infrequent gradient calculations. We show that this approach requires fewer floating-point operations (FLOPs) to achieve a given level of accuracy compared to prior techniques. We demonstrate state-of-the-art sparse training results on a variety of networks and datasets, including ResNet-50, MobileNets on Imagenet-2012, and RNNs on WikiText-103. Finally, we provide some insights into why allowing the topology to change during the optimization can overcome local minima encountered when the topology remains static. Code used in our work can be found in this http URL."
                },
                {
                    "title": "Sparse Networks from Scratch: Faster Training without Losing Performance",
                    "abstract": "We demonstrate the possibility of what we call sparse learning: accelerated training of deep neural networks that maintain sparse weights throughout training while achieving dense performance levels. We accomplish this by developing sparse momentum, an algorithm which uses exponentially smoothed gradients (momentum) to identify layers and weights which reduce the error efficiently. Sparse momentum redistributes pruned weights across layers according to the mean momentum magnitude of each layer. Within a layer, sparse momentum grows weights according to the momentum magnitude of zero-valued weights. We demonstrate state-of-the-art sparse performance on MNIST, CIFAR-10, and ImageNet, decreasing the mean error by a relative 8%, 15%, and 6% compared to other sparse algorithms. Furthermore, we show that sparse momentum reliably reproduces dense performance levels while providing up to 5.61x faster training. In our analysis, ablations show that the benefits of momentum redistribution and growth increase with the depth and size of the network. Additionally, we find that sparse momentum is insensitive to the choice of its hyperparameters suggesting that sparse momentum is robust and easy to use."
                },
                {
                    "title": "FreebaseQA: A New Factoid QA Data Set Matching Trivia-Style Question-Answer Pairs with Freebase",
                    "abstract": "In this paper, we present a new data set, named FreebaseQA, for open-domain factoid question answering (QA) tasks over structured knowledge bases, like Freebase. The data set is generated by matching trivia-type question-answer pairs with subject-predicate-object triples in Freebase. For each collected question-answer pair, we first tag all entities in each question and search for relevant predicates that bridge a tagged entity with the answer in Freebase. Finally, human annotation is used to remove any false positive in these matched triples. Using this method, we are able to efficiently generate over 54K matches from about 28K unique questions with minimal cost. Our analysis shows that this data set is suitable for model training in factoid QA tasks beyond simpler questions since FreebaseQA provides more linguistically sophisticated questions than other existing data sets."
                },
                {
                    "title": "The State of Sparsity in Deep Neural Networks",
                    "abstract": "We rigorously evaluate three state-of-the-art techniques for inducing sparsity in deep neural networks on two large-scale learning tasks: Transformer trained on WMT 2014 English-to-German, and ResNet-50 trained on ImageNet. Across thousands of experiments, we demonstrate that complex techniques (Molchanov et al., 2017; Louizos et al., 2017b) shown to yield high compression rates on smaller datasets perform inconsistently, and that simple magnitude pruning approaches achieve comparable or better results. Additionally, we replicate the experiments performed by (Frankle & Carbin, 2018) and (Liu et al., 2018) at scale and show that unstructured sparse architectures learned through pruning cannot be trained from scratch to the same test set performance as a model trained with joint sparsification and optimization. Together, these results highlight the need for large-scale benchmarks in the field of model compression. We open-source our code, top performing model checkpoints, and results of all hyperparameter configurations to establish rigorous baselines for future work on compression and sparsification."
                },
                {
                    "title": "Parameter Efficient Training of Deep Convolutional Neural Networks by Dynamic Sparse Reparameterization",
                    "abstract": "Modern deep neural networks are typically highly overparameterized. Pruning techniques are able to remove a significant fraction of network parameters with little loss in accuracy. Recently, techniques based on dynamic reallocation of non-zero parameters have emerged, allowing direct training of sparse networks without having to pre-train a large dense model. Here we present a novel dynamic sparse reparameterization method that addresses the limitations of previous techniques such as high computational cost and the need for manual configuration of the number of free parameters allocated to each layer. We evaluate the performance of dynamic reallocation methods in training deep convolutional networks and show that our method outperforms previous static and dynamic reparameterization methods, yielding the best accuracy for a fixed parameter budget, on par with accuracies obtained by iteratively pruning a pre-trained dense model. We further investigated the mechanisms underlying the superior generalization performance of the resultant sparse networks. We found that neither the structure, nor the initialization of the non-zero parameters were sufficient to explain the superior performance. Rather, effective learning crucially depended on the continuous exploration of the sparse network structure space during training. Our work suggests that exploring structural degrees of freedom during training is more effective than adding extra parameters to the network."
                },
                {
                    "title": "Iteratively Training Look-Up Tables for Network Quantization",
                    "abstract": "Operating deep neural networks (DNNs) on devices with limited resources requires the reduction of their memory as well as computational footprint. Popular reduction methods are network quantization or pruning, which either reduce the word length of the network parameters or remove weights from the network if they are not needed. In this article, we discuss a general framework for network reduction which we call Look-Up Table Quantization (LUT-Q). For each layer, we learn a value dictionary and an assignment matrix to represent the network weights. We propose a special solver which combines gradient descent and a one-step k-means update to learn both the value dictionaries and assignment matrices iteratively. This method is very flexible: by constraining the value dictionary, many different reduction problems such as non-uniform network quantization, training of multiplierless networks, network pruning, or simultaneous quantization and pruning can be implemented without changing the solver. This flexibility of the LUT-Q method allows us to use the same method to train networks for different hardware capabilities."
                },
                {
                    "title": "To prune, or not to prune: exploring the efficacy of pruning for model compression",
                    "abstract": "Model pruning seeks to induce sparsity in a deep neural network's various connection matrices, thereby reducing the number of nonzero-valued parameters in the model. Recent reports (Han et al., 2015; Narang et al., 2017) prune deep networks at the cost of only a marginal loss in accuracy and achieve a sizable reduction in model size. This hints at the possibility that the baseline models in these experiments are perhaps severely over-parameterized at the outset and a viable alternative for model compression might be to simply reduce the number of hidden units while maintaining the model's dense connection structure, exposing a similar trade-off in model size and accuracy. We investigate these two distinct paths for model compression within the context of energy-efficient inference in resource-constrained environments and propose a new gradual pruning technique that is simple and straightforward to apply across a variety of models/datasets with minimal tuning and can be seamlessly incorporated within the training process. We compare the accuracy of large, but pruned models (large-sparse) and their smaller, but dense (small-dense) counterparts with identical memory footprint. Across a broad range of neural network architectures (deep CNNs, stacked LSTM, and seq2seq LSTM models), we find large-sparse models to consistently outperform small-dense models and achieve up to 10x reduction in number of non-zero parameters with minimal loss in accuracy."
                },
                {
                    "title": "TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension",
                    "abstract": "We present TriviaQA, a challenging reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents, six per question on average, that provide high quality distant supervision for answering the questions. We show that, in comparison to other recently introduced large-scale datasets, TriviaQA (1) has relatively complex, compositional questions, (2) has considerable syntactic and lexical variability between questions and corresponding answer-evidence sentences, and (3) requires more cross sentence reasoning to find answers. We also present two baseline algorithms: a feature-based classifier and a state-of-the-art neural network, that performs well on SQuAD reading comprehension. Neither approach comes close to human performance (23% and 40% vs. 80%), suggesting that TriviaQA is a challenging testbed that is worth significant future study."
                },
                {
                    "title": "Pointer Sentinel Mixture Models",
                    "abstract": "Recent neural network sequence models with softmax classifiers have achieved their best language modeling performance only with very large hidden states and large vocabularies. Even then they struggle to predict rare or unseen words even if the context makes the prediction unambiguous. We introduce the pointer sentinel mixture architecture for neural sequence models which has the ability to either reproduce a word from the recent context or produce a word from a standard softmax classifier. Our pointer sentinel-LSTM model achieves state of the art language modeling performance on the Penn Treebank (70.9 perplexity) while using far fewer parameters than a standard softmax LSTM. In order to evaluate how well language models can exploit longer contexts and deal with more realistic vocabularies and larger corpora we also introduce the freely available WikiText corpus."
                },
                {
                    "title": "A Thorough Examination of the CNN/Daily Mail Reading Comprehension Task",
                    "abstract": "Enabling a computer to understand a document so that it can answer comprehension questions is a central, yet unsolved goal of NLP. A key factor impeding its solution by machine learned systems is the limited availability of human-annotated data. Hermann et al. (2015) seek to solve this problem by creating over a million training examples by pairing CNN and Daily Mail news articles with their summarized bullet points, and show that a neural network can then be trained to give good performance on this task. In this paper, we conduct a thorough examination of this new reading comprehension task. Our primary aim is to understand what depth of language understanding is required to do well on this task. We approach this from one side by doing a careful hand-analysis of a small subset of the problems and from the other by showing that simple, carefully designed systems can obtain accuracies of 73.6% and 76.6% on these two datasets, exceeding current state-of-the-art results by 7-10% and approaching what we believe is the ceiling for performance on this task."
                },
                {
                    "title": "Abstractive Text Summarization using Sequence-to-sequence RNNs and Beyond",
                    "abstract": "In this work, we model abstractive text summarization using Attentional Encoder-Decoder Recurrent Neural Networks, and show that they achieve state-of-the-art performance on two different corpora. We propose several novel models that address critical problems in summarization that are not adequately modeled by the basic architecture, such as modeling key-words, capturing the hierarchy of sentence-to-word structure, and emitting words that are rare or unseen at training time. Our work shows that many of our proposed models contribute to further improvement in performance. We also propose a new dataset consisting of multi-sentence summaries, and establish performance benchmarks for further research."
                },
                {
                    "title": "Deep Compression: Compressing Deep Neural Network with Pruning, Trained Quantization and Huffman Coding",
                    "abstract": "Neural networks are both computationally intensive and memory intensive, making them difficult to deploy on embedded systems with limited hardware resources. To address this limitation, we introduce \"deep compression\", a three stage pipeline: pruning, trained quantization and Huffman coding, that work together to reduce the storage requirement of neural networks by 35x to 49x without affecting their accuracy. Our method first prunes the network by learning only the important connections. Next, we quantize the weights to enforce weight sharing, finally, we apply Huffman coding. After the first two steps we retrain the network to fine tune the remaining connections and the quantized centroids. Pruning, reduces the number of connections by 9x to 13x; Quantization then reduces the number of bits that represent each connection from 32 to 5. On the ImageNet dataset, our method reduced the storage required by AlexNet by 35x, from 240MB to 6.9MB, without loss of accuracy. Our method reduced the size of VGG-16 by 49x from 552MB to 11.3MB, again with no loss of accuracy. This allows fitting the model into on-chip SRAM cache rather than off-chip DRAM memory. Our compression method also facilitates the use of complex neural networks in mobile applications where application size and download bandwidth are constrained. Benchmarked on CPU, GPU and mobile GPU, compressed network has 3x to 4x layerwise speedup and 3x to 7x better energy efficiency."
                },
                {
                    "title": "A Neural Network for Factoid Question Answering over Paragraphs",
                    "abstract": "Text classification methods for tasks like factoid question answering typically use manually defined string matching rules or bag of words representations. These methods are ineective when question text contains very few individual words (e.g., named entities) that are indicative of the answer. We introduce a recursive neural network (rnn) model that can reason over such input by modeling textual compositionality. We apply our model, qanta, to a dataset of questions from a trivia competition called quiz bowl. Unlike previous rnn models, qanta learns word and phrase-level representations that combine across sentences to reason about entities. The model outperforms multiple baselines and, when combined with information retrieval methods, rivals the best human players."
                },
                {
                    "title": "Large Language Models Are State-of-the-Art Evaluators of Code Generation",
                    "abstract": "Recent advancements in the field of natural language generation have facilitated the use of large language models to assess the quality of generated text. Although these models have shown promising results in tasks such as machine translation and summarization, their applicability in code generation tasks remains limited without human involvement. The complexity of programming concepts required for such tasks makes it difficult to develop evaluation metrics that align with human judgment. Token-matching-based metrics, such as BLEU, have demonstrated weak correlations with human practitioners in code generation tasks. Moreover, the utilization of human-written test suites to evaluate functional correctness can be challenging in domains with low resources. To overcome these obstacles, we propose a new evaluation framework based on the GPT-3.5 (GPT-3.5-turbo), for code generation assessments. Our framework addresses the limitations of existing approaches by achieving superior correlations with functional correctness and human preferences, without the need for test oracles or references. We evaluate the efficacy of our framework on two different aspects (human preference and execution success) and four programming languages, comparing its performance with the state-of-the-art CodeBERTScore metric, which relies on a pre-trained model. Our results demonstrate that our framework surpasses CodeBERTScore, delivering high levels of accuracy and consistency across various programming languages and tasks. We also make our evaluation framework and datasets available to the public at https://github.com/terryyz/llm-code-eval, encouraging further research in the evaluation of code generation.1"
                },
                {
                    "title": "Large Models are Parsimonious Learners: Activation Sparsity in Trained Transformers",
                    "abstract": ","
                },
                {
                    "title": "Optimal Brain Damage",
                    "abstract": "We have used information-theoretic ideas to derive a class of practical and nearly optimal schemes for adapting the size of a neural network. By removing unimportant weights from a network, several improvements can be expected: better generalization, fewer training examples required, and improved speed of learning and/or classification. The basic idea is to use second-derivative information to make a tradeoff between network complexity and training set error. Experiments confirm the usefulness of the methods on a real-world application."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively compress Large Language Models (LLMs) like Vicuna-7B while maintaining their performance and reducing computational and memory requirements?\n\n### [Question 2] - Why is it interesting and important?\nSolving the problem of compressing LLMs is crucial for making advanced AI technologies more accessible and efficient. As LLMs become increasingly integral to various applications, their high resource demands pose significant barriers to deployment, especially in resource-constrained environments. Addressing this issue could lead to broader adoption of LLMs in real-world applications, enhance research in model efficiency, and inspire new methodologies for model training and deployment. Furthermore, advancements in compression techniques could pave the way for future research focused on optimizing AI models, leading to innovations in areas such as edge computing, mobile applications, and real-time processing.\n\n### [Question 3] - Why is it hard?\nThe challenges in compressing LLMs stem from the delicate balance between reducing model size and maintaining performance. Naive approaches, such as simple pruning or quantization, often lead to significant degradation in model accuracy or perplexity. The complexities arise from the need to preserve the model's ability to generalize while reducing its parameters. Additionally, technical obstacles include the intricacies of sparsity patterns, the impact of compression on different layers of the model, and the need for sophisticated algorithms to ensure that the compressed model retains its capabilities. Theoretical challenges also exist in understanding the trade-offs between model size and performance metrics.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on either model performance or compression techniques in isolation, leading to gaps in understanding how to effectively combine the two. Many existing solutions have limitations in scalability or fail to generalize across different model architectures. Barriers such as the lack of comprehensive datasets for evaluating compressed models and insufficient exploration of advanced compression methods have hindered progress. Our approach aims to integrate state-of-the-art compression techniques with a focus on maintaining performance, addressing these gaps by leveraging recent advancements in model training and evaluation.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves a multi-faceted approach to compressing the Vicuna-7B model through advanced techniques such as structured N:M sparsity and quantization to 4 bits per weight. We will utilize a diverse dataset of user-shared conversations collected from ShareGPT to fine-tune the model and evaluate its performance. The primary metric for success"
            }
        },
        "author_data": {
            "856128ed-946d-4bd4-9a29-b9e8516ca831": {
                "pk": "856128ed-946d-4bd4-9a29-b9e8516ca831",
                "name": "Ajay Jaiswal",
                "collaborators": [
                    "Zhangyang Wang",
                    "Lu Yin",
                    "Shiwei Liu",
                    "Zhenyu (Allen) Zhang",
                    "Jiawei Zhao",
                    "Yuandong Tian",
                    "Souvik Kundu",
                    "Nurendra Choudhary",
                    "Ravinarayana Adkathimar",
                    "M. P. Alagappan",
                    "G. Hiranandani",
                    "Ying Ding",
                    "E-Wen Huang",
                    "Karthik Subbian",
                    "Runjin Chen",
                    "Tong Zhao",
                    "Neil Shah",
                    "Bodun Hu",
                    "Yeonju Ro",
                    "Tianlong Chen",
                    "Aditya Akella",
                    "Junyuan Hong",
                    "Jinhao Duan",
                    "Chenhui Zhang",
                    "Zhangheng Li",
                    "Chulin Xie",
                    "Kelsey Lieberman",
                    "James Diffenderfer",
                    "Brian Bartoldson",
                    "Kaidi Xu",
                    "B. Kailkhura",
                    "Dan Hendrycks",
                    "Dawn Song",
                    "Bo Li",
                    "G. Holste",
                    "Yiliang Zhou",
                    "Song Wang",
                    "Mingquan Lin",
                    "Sherry Zhuge",
                    "Yuzhe Yang",
                    "Dongkyun Kim",
                    "Trong-Hieu Nguyen-Mau",
                    "Minh-Triet Tran",
                    "Jaehyup Jeong",
                    "Wongi Park",
                    "Jongbin Ryu",
                    "Feng Hong",
                    "Arsh Verma",
                    "Yosuke Yamagishi",
                    "Changhyun Kim",
                    "Hyeryeong Seo",
                    "Myungjoo Kang",
                    "L. Celi",
                    "Zhiyong Lu",
                    "Ronald M. Summers",
                    "George Shih",
                    "Yifan Peng"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Large Language Models",
                    "Machine Learning",
                    "Trustworthiness"
                ],
                "publications": [
                    {
                        "title": "All Against Some: Efficient Integration of Large Language Models for Message Passing in Graph Neural Networks",
                        "abstract": "Graph Neural Networks (GNNs) have attracted immense attention in the past decade due to their numerous real-world applications built around graph-structured data. On the other hand, Large Language Models (LLMs) with extensive pretrained knowledge and powerful semantic comprehension abilities have recently shown a remarkable ability to benefit applications using vision and text data. In this paper, we investigate how LLMs can be leveraged in a computationally efficient fashion to benefit rich graph-structured data, a modality relatively unexplored in LLM literature. Prior works in this area exploit LLMs to augment every node features in an ad-hoc fashion (not scalable for large graphs), use natural language to describe the complex structural information of graphs, or perform computationally expensive finetuning of LLMs in conjunction with GNNs. We propose E-LLaGNN (Efficient LLMs augmented GNNs), a framework with an on-demand LLM service that enriches message passing procedure of graph learning by enhancing a limited fraction of nodes from the graph. More specifically, E-LLaGNN relies on sampling high-quality neighborhoods using LLMs, followed by on-demand neighborhood feature enhancement using diverse prompts from our prompt catalog, and finally information aggregation using message passing from conventional GNN architectures. We explore several heuristics-based active node selection strategies to limit the computational and memory footprint of LLMs when handling millions of nodes. Through extensive experiments&ablation on popular graph benchmarks of varying scales (Cora, PubMed, ArXiv,&Products), we illustrate the effectiveness of our E-LLaGNN framework and reveal many interesting capabilities such as improved gradient flow in deep GNNs, LLM-free inference ability etc."
                    },
                    {
                        "title": "LLaGA: Large Language and Graph Assistant",
                        "abstract": "Graph Neural Networks (GNNs) have empowered the advance in graph-structured data analysis. Recently, the rise of Large Language Models (LLMs) like GPT-4 has heralded a new era in deep learning. However, their application to graph data poses distinct challenges due to the inherent difficulty of translating graph structures to language. To this end, we introduce the Large Language and Graph Assistant (LLaGA), an innovative model that effectively integrates LLM capabilities to handle the complexities of graph-structured data. LLaGA retains the general-purpose nature of LLMs while adapting graph data into a format compatible with LLM input. LLaGA achieves this by reorganizing graph nodes to structure-aware sequences and then mapping these into the token embedding space through a versatile projector. LLaGA excels in versatility, generalizability and interpretability, allowing it to perform consistently well across different datasets and tasks, extend its ability to unseen datasets or tasks, and provide explanations for graphs. Our extensive experiments across popular graph benchmarks show that LLaGA delivers outstanding performance across four datasets and three tasks using one single model, surpassing state-of-the-art graph models in both supervised and zero-shot scenarios. Our code is available at \\url{https://github.com/VITA-Group/LLaGA}."
                    },
                    {
                        "title": "Q-GaLore: Quantized GaLore with INT4 Projection and Layer-Adaptive Low-Rank Gradients",
                        "abstract": "Training Large Language Models (LLMs) is memory-intensive due to the large number of parameters and associated optimization states. GaLore, a recent method, reduces memory usage by projecting weight gradients into a low-rank subspace without compromising performance. However, GaLore relies on time-consuming Singular Value Decomposition (SVD) operations to identify the subspace, and the frequent subspace updates lead to significant training time overhead. Moreover, GaLore offers minimal improvements in accuracy and efficiency compared to LoRA in more accessible fine-tuning scenarios. To address these limitations, we introduce Q-Galore, a novel approach that substantially reduces memory usage by combining quantization and low-rank projection, surpassing the benefits of GaLore. Our method is based on two key observations: (i) the gradient subspace exhibits diverse properties, with some layers converging early in training while others are subject to frequent changes; (ii) the projection matrices are highly resilient to low-bit quantization. Leveraging these insights, Q-GaLore adaptively updates the gradient subspace based on its convergence statistics, achieving comparable performance while significantly reducing the number of SVD operations. We maintain the projection matrices in INT4 format and weights in INT8 format, incorporating stochastic rounding to capture accumulated gradient information. This approach enables a high-precision training trajectory using only low-precision weights. We demonstrate that Q-GaLore achieves highly competitive performance with exceptional memory efficiency. At pre-training, Q-GaLore facilitates training a LLaMA-7B model from scratch on a single NVIDIA RTX 4060 Ti with only 16 GB memory. At fine-tuning, it reduces memory consumption by up to 50% compared to LoRA and GaLore, while consistently outperforming QLoRA at the same memory cost."
                    },
                    {
                        "title": "From GaLore to WeLore: How Low-Rank Weights Non-uniformly Emerge from Low-Rank Gradients",
                        "abstract": "Modern Large Language Models (LLMs) are composed of matrices with billions of elements, making their storage and processing quite demanding in terms of computational resources and memory usage. Being significantly large, such matrices can often be expressed in low-rank format with potential to relax resource requirements. Unlike prior works which focus on developing novel matrix decomposition algorithms, in this work we first study the emergence of low-rank structures across matrices within different layers of LLMs and establish a consequential relationship between the gradient dynamics and emerging low-rank expressiveness of matrices. Our findings reveal that different layers exhibit varying levels of converged low-rank structure, necessitating a non-uniform rank reduction across them to minimize performance drop due to compression. In view of that, we present Weight Low-Rank Projection (WeLore) that unifies weight compression and memory-efficient fine-tuning as ONE, in a data-agnostic and one-shot way. WeLore capitalizes the heavy-tail distribution of singular values to identify a suitable rank reduction ratio for matrices within LLMs. Going beyond only as a compression technique, WeLore categorizes weight matrices into Low-rank Components (LRCs) and Non-Low-rank Components (N-LRCs) based on their ability to express themselves as low-rank. Our gradient perspective and extensive experiments illustrate that LRCs tend to have better finetuning capabilities and can closely mimic (sometimes outperform) the training loss trajectory and performance of full-finetuning with notable memory and compute footprint reduction. For example, finetuning a 50\\% compressed LLaMa-2 7B model using only a fraction of parameters in LRCs (WeLore) can outperform its full finetuning with ~3x better throughput and ~0.6x GPU requirement. Our codes are available at \\url{https://github.com/VITA-Group/welore}"
                    },
                    {
                        "title": "FFN-SkipLLM: A Hidden Gem for Autoregressive Decoding with Adaptive Feed Forward Skipping",
                        "abstract": "Autoregressive Large Language Models (e.g., LLaMa, GPTs) are omnipresent achieving remarkable success in language understanding and generation. However, such impressive capability typically comes with a substantial model size, which presents significant challenges for autoregressive token-by-token generation. To mitigate computation overload incurred during generation, several early-exit and layer-dropping strategies have been proposed. Despite some promising success due to the redundancy across LLMs layers on metrics like Rough-L/BLUE, our careful knowledge-intensive evaluation unveils issues such as generation collapse, hallucination of wrong facts, and noticeable performance drop even at the trivial exit ratio of 10-15% of layers. We attribute these errors primarily to ineffective handling of the KV cache through state copying during early-exit. In this work, we observed the saturation of computationally expensive feed-forward blocks of LLM layers and proposed FFN-SkipLLM, which is a novel fine-grained skip strategy of autoregressive LLMs. More specifically, FFN-SkipLLM is an input-adaptive feed-forward skipping strategy that can skip 25-30% of FFN blocks of LLMs with marginal change in performance on knowledge-intensive generation tasks without any requirement to handle KV cache. Our extensive experiments and ablation across benchmarks like MT-Bench, Factoid-QA, and variable-length text summarization illustrate how our simple and ease-at-use method can facilitate faster autoregressive decoding."
                    },
                    {
                        "title": "Decoding Compressed Trust: Scrutinizing the Trustworthiness of Efficient LLMs Under Compression",
                        "abstract": "Compressing high-capability Large Language Models (LLMs) has emerged as a favored strategy for resource-efficient inferences. While state-of-the-art (SoTA) compression methods boast impressive advancements in preserving benign task performance, the potential risks of compression in terms of safety and trustworthiness have been largely neglected. This study conducts the first, thorough evaluation of three (3) leading LLMs using five (5) SoTA compression techniques across eight (8) trustworthiness dimensions. Our experiments highlight the intricate interplay between compression and trustworthiness, revealing some interesting patterns. We find that quantization is currently a more effective approach than pruning in achieving efficiency and trustworthiness simultaneously. For instance, a 4-bit quantized model retains the trustworthiness of its original counterpart, but model pruning significantly degrades trustworthiness, even at 50% sparsity. Moreover, employing quantization within a moderate bit range could unexpectedly improve certain trustworthiness dimensions such as ethics and fairness. Conversely, extreme quantization to very low bit levels (3 bits) tends to reduce trustworthiness significantly. This increased risk cannot be uncovered by looking at benign performance alone, in turn, mandating comprehensive trustworthiness evaluation in practice. These findings culminate in practical recommendations for simultaneously achieving high utility, efficiency, and trustworthiness in LLMs. Code and models are available at https://decoding-comp-trust.github.io."
                    },
                    {
                        "title": "Junk DNA Hypothesis: A Task-Centric Angle of LLM Pre-trained Weights through Sparsity",
                        "abstract": "validate"
                    },
                    {
                        "title": "Pruning Small Pre-Trained Weights Irreversibly and Monotonically Impairs\"Difficult\"Downstream Tasks in LLMs",
                        "abstract": "We present Junk DNA Hypothesis by adopting a novel task-centric angle for the pre-trained weights of large language models (LLMs). It has been believed that weights in LLMs contain significant redundancy, leading to the conception that a considerable chunk of the parameters can be removed by pruning without compromising performance. Contrary to this belief, this paper presents a counter-argument: small-magnitude weights of pre-trained model weights encode vital knowledge essential for tackling difficult downstream tasks - manifested as the monotonic relationship between the performance drop of downstream tasks across the difficulty spectrum, as we prune more pre-trained weights by magnitude. Moreover, we reveal that these seemingly inconsequential weights can result in irreparable loss of knowledge and performance degradation in difficult tasks, even when downstream continual training is allowed. Interestingly, our evaluations show that the other popular compression, namely quantization, fails to exhibit similar monotonic effect and does not as convincingly disentangle this task-difficulty information. To study formally, we introduce several quantifiable metrics to gauge the downstream task difficulty: (1) within the same task category, and (2) across different task categories. Our extensive experiments substantiate the Junk DNA Hypothesis across a diverse range of model sizes, tasks, datasets, and even pruning methods. Codes are available at: https://github.com/VITA-Group/Junk_DNA_Hypothesis.git."
                    },
                    {
                        "title": "Towards long-tailed, multi-label disease classification from chest X-ray: Overview of the CXR-LT challenge",
                        "abstract": "Many real-world image recognition problems, such as diagnostic medical imaging exams, are \u201clong-tailed\u201d \u2013 there are a few common findings followed by many more relatively rare conditions. In chest radiography, diagnosis is both a long-tailed and multi-label problem, as patients often present with multiple findings simultaneously. While researchers have begun to study the problem of long-tailed learning in medical image recognition, few have studied the interaction of label imbalance and label co-occurrence posed by long-tailed, multi-label disease classification. To engage with the research community on this emerging topic, we conducted an open challenge, CXR-LT, on long-tailed, multi-label thorax disease classification from chest X-rays (CXRs). We publicly release a large-scale benchmark dataset of over 350,000 CXRs, each labeled with at least one of 26 clinical findings following a long-tailed distribution. We synthesize common themes of top-performing solutions, providing practical recommendations for long-tailed, multi-label medical image classification. Finally, we use these insights to propose a path forward involving vision-language foundation models for few- and zero-shot disease classification."
                    }
                ]
            },
            "ee3a8f3c-3227-4fcf-85e1-017b5f5fed60": {
                "pk": "ee3a8f3c-3227-4fcf-85e1-017b5f5fed60",
                "name": "Zhe Gan",
                "collaborators": [
                    "Yinfei Yang",
                    "Haotian Zhang",
                    "Bowen Zhang",
                    "Hong-You Chen",
                    "Zhengfeng Lai",
                    "Haoxuan You",
                    "Peter Grasch",
                    "Keen You",
                    "Yanghao Li",
                    "J. Fauconnier",
                    "Zirui Wang",
                    "Afshin Dehghan",
                    "Yusu Qian",
                    "Meng Cao",
                    "Mingfei Gao",
                    "Philipp Dufter",
                    "Xianzhi Du",
                    "Mingze Xu",
                    "Chen Chen",
                    "Shih-Fu Chang",
                    "Hanrong Ye",
                    "Xinze Wang",
                    "Marcin Eichner",
                    "Nina Wenzel",
                    "Forrest Huang",
                    "Dhruti Shah",
                    "Sam Dodge",
                    "Zhen Yang",
                    "Aleksei Timofeev",
                    "E. Schoop",
                    "Floris Weers",
                    "Amanda Swearngin",
                    "Jeffrey Nichols",
                    "Vasileios Saveris",
                    "Juan Lao Tebar",
                    "Wenze Hu",
                    "Haiming Gang",
                    "Kai Kang",
                    "Tsu-Jui Fu",
                    "William Yang Wang",
                    "Erik Daxberger",
                    "Lin Chen",
                    "Zongyu Lin",
                    "Dan Xu",
                    "Jiasen Lu",
                    "Ruohong Zhang",
                    "Zhiqing Sun",
                    "Ruoming Pang",
                    "Yiming Yang",
                    "Elmira Amirloo",
                    "Christoph Roesmann",
                    "Christian Kerl",
                    "Rinu Boney",
                    "Liangliang Cao"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Large Language Models",
                    "Vision-Language Models",
                    "Image Understanding"
                ],
                "publications": [
                    {
                        "title": "Contrastive Localized Language-Image Pre-Training",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) has been a celebrated method for training vision encoders to generate image/text representations facilitating various applications. Recently, CLIP has been widely adopted as the vision backbone of multimodal large language models (MLLMs) to connect image inputs for language interactions. The success of CLIP as a vision-language foundation model relies on aligning web-crawled noisy text annotations at image levels. Nevertheless, such criteria may become insufficient for downstream tasks in need of fine-grained vision representations, especially when region-level understanding is demanding for MLLMs. In this paper, we improve the localization capability of CLIP with several advances. We propose a pre-training method called Contrastive Localized Language-Image Pre-training (CLOC) by complementing CLIP with region-text contrastive loss and modules. We formulate a new concept, promptable embeddings, of which the encoder produces image embeddings easy to transform into region representations given spatial hints. To support large-scale pre-training, we design a visually-enriched and spatially-localized captioning framework to effectively generate region-text pseudo-labels at scale. By scaling up to billions of annotated images, CLOC enables high-quality regional embeddings for image region recognition and retrieval tasks, and can be a drop-in replacement of CLIP to enhance MLLMs, especially on referring and grounding tasks."
                    },
                    {
                        "title": "MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning",
                        "abstract": "We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development."
                    },
                    {
                        "title": "Ferret-UI: Grounded Mobile UI Understanding with Multimodal LLMs",
                        "abstract": "Recent advancements in multimodal large language models (MLLMs) have been noteworthy, yet, these general-domain MLLMs often fall short in their ability to comprehend and interact effectively with user interface (UI) screens. In this paper, we present Ferret-UI, a new MLLM tailored for enhanced understanding of mobile UI screens, equipped with referring, grounding, and reasoning capabilities. Given that UI screens typically exhibit a more elongated aspect ratio and contain smaller objects of interest (e.g., icons, texts) than natural images, we incorporate\"any resolution\"on top of Ferret to magnify details and leverage enhanced visual features. Specifically, each screen is divided into 2 sub-images based on the original aspect ratio (i.e., horizontal division for portrait screens and vertical division for landscape screens). Both sub-images are encoded separately before being sent to LLMs. We meticulously gather training samples from an extensive range of elementary UI tasks, such as icon recognition, find text, and widget listing. These samples are formatted for instruction-following with region annotations to facilitate precise referring and grounding. To augment the model's reasoning ability, we further compile a dataset for advanced tasks, including detailed description, perception/interaction conversations, and function inference. After training on the curated datasets, Ferret-UI exhibits outstanding comprehension of UI screens and the capability to execute open-ended instructions. For model evaluation, we establish a comprehensive benchmark encompassing all the aforementioned tasks. Ferret-UI excels not only beyond most open-source UI MLLMs, but also surpasses GPT-4V on all the elementary UI tasks."
                    },
                    {
                        "title": "How Easy is It to Fool Your Multimodal LLMs? An Empirical Analysis on Deceptive Prompts",
                        "abstract": "The remarkable advancements in Multimodal Large Language Models (MLLMs) have not rendered them immune to challenges, particularly in the context of handling deceptive information in prompts, thus producing hallucinated responses under such conditions. To quantitatively assess this vulnerability, we present MAD-Bench, a carefully curated benchmark that contains 1000 test samples divided into 5 categories, such as non-existent objects, count of objects, and spatial relationship. We provide a comprehensive analysis of popular MLLMs, ranging from GPT-4v, Reka, Gemini-Pro, to open-sourced models, such as LLaVA-NeXT and MiniCPM-Llama3. Empirically, we observe significant performance gaps between GPT-4o and other models; and previous robust instruction-tuned models are not effective on this new benchmark. While GPT-4o achieves 82.82% accuracy on MAD-Bench, the accuracy of any other model in our experiments ranges from 9% to 50%. We further propose a remedy that adds an additional paragraph to the deceptive prompts to encourage models to think twice before answering the question. Surprisingly, this simple method can even double the accuracy; however, the absolute numbers are still too low to be satisfactory. We hope MAD-Bench can serve as a valuable benchmark to stimulate further research to enhance model resilience against deceptive prompts."
                    },
                    {
                        "title": "Revisit Large-Scale Image-Caption Data in Pre-training Multimodal Foundation Models",
                        "abstract": "Recent advancements in multimodal models highlight the value of rewritten captions for improving performance, yet key challenges remain. For example, while synthetic captions often provide superior quality and image-text alignment, it is not clear whether they can fully replace AltTexts: the role of synthetic captions and their interaction with original web-crawled AltTexts in pre-training is still not well understood. Moreover, different multimodal foundation models may have unique preferences for specific caption formats, but efforts to identify the optimal captions for each model remain limited. In this work, we propose a novel, controllable, and scalable captioning pipeline designed to generate diverse caption formats tailored to various multimodal models. By examining Short Synthetic Captions (SSC) towards Dense Synthetic Captions (DSC+) as case studies, we systematically explore their effects and interactions with AltTexts across models such as CLIP, multimodal LLMs, and diffusion models. Our findings reveal that a hybrid approach that keeps both synthetic captions and AltTexts can outperform the use of synthetic captions alone, improving both alignment and performance, with each model demonstrating preferences for particular caption formats. This comprehensive analysis provides valuable insights into optimizing captioning strategies, thereby advancing the pre-training of multimodal foundation models."
                    },
                    {
                        "title": "SlowFast-LLaVA: A Strong Training-Free Baseline for Video Large Language Models",
                        "abstract": "We propose SlowFast-LLaVA (or SF-LLaVA for short), a training-free video large language model (LLM) that can jointly capture detailed spatial semantics and long-range temporal context without exceeding the token budget of commonly used LLMs. This is realized by using a two-stream SlowFast design of inputs for Video LLMs to aggregate features from sampled frames in an effective way. Specifically, the Slow pathway extracts features at a low frame rate while keeping as much spatial detail as possible (e.g., with 12x24 tokens), and the Fast pathway operates on a high frame rate but uses a larger spatial pooling stride (e.g., downsampling 6x) to focus on the motion cues. As a result, this design allows us to adequately capture both spatial and temporal features that are beneficial for detailed video understanding. Experimental results show that SF-LLaVA outperforms existing training-free methods on a wide range of video tasks. On some benchmarks, it achieves comparable or even better performance compared to state-of-the-art Video LLMs that are fine-tuned on video datasets. Code has been made available at: https://github.com/apple/ml-slowfast-llava."
                    },
                    {
                        "title": "Ferret-v2: An Improved Baseline for Referring and Grounding with Large Language Models",
                        "abstract": "While Ferret seamlessly integrates regional understanding into the Large Language Model (LLM) to facilitate its referring and grounding capability, it poses certain limitations: constrained by the pre-trained fixed visual encoder and failed to perform well on broader tasks. In this work, we unveil Ferret-v2, a significant upgrade to Ferret, with three key designs. (1) Any resolution grounding and referring: A flexible approach that effortlessly handles higher image resolution, improving the model's ability to process and understand images in greater detail. (2) Multi-granularity visual encoding: By integrating the additional DINOv2 encoder, the model learns better and diverse underlying contexts for global and fine-grained visual information. (3) A three-stage training paradigm: Besides image-caption alignment, an additional stage is proposed for high-resolution dense alignment before the final instruction tuning. Experiments show that Ferret-v2 provides substantial improvements over Ferret and other state-of-the-art methods, thanks to its high-resolution scaling and fine-grained visual processing."
                    },
                    {
                        "title": "MM-Ego: Towards Building Egocentric Multimodal LLMs",
                        "abstract": "This research aims to comprehensively explore building a multimodal foundation model for egocentric video understanding. To achieve this goal, we work on three fronts. First, as there is a lack of QA data for egocentric video understanding, we develop a data engine that efficiently generates 7M high-quality QA samples for egocentric videos ranging from 30 seconds to one hour long, based on human-annotated data. This is currently the largest egocentric QA dataset. Second, we contribute a challenging egocentric QA benchmark with 629 videos and 7,026 questions to evaluate the models' ability in recognizing and memorizing visual details across videos of varying lengths. We introduce a new de-biasing evaluation method to help mitigate the unavoidable language bias present in the models being evaluated. Third, we propose a specialized multimodal architecture featuring a novel\"Memory Pointer Prompting\"mechanism. This design includes a global glimpse step to gain an overarching understanding of the entire video and identify key visual information, followed by a fallback step that utilizes the key visual information to generate responses. This enables the model to more effectively comprehend extended video content. With the data, benchmark, and model, we successfully build MM-Ego, an egocentric multimodal LLM that shows powerful performance on egocentric video understanding."
                    },
                    {
                        "title": "Improve Vision Language Model Chain-of-thought Reasoning",
                        "abstract": "Chain-of-thought (CoT) reasoning in vision language models (VLMs) is crucial for improving interpretability and trustworthiness. However, current training recipes lack robust CoT reasoning data, relying on datasets dominated by short annotations with minimal rationales. In this work, we show that training VLM on short answers does not generalize well to reasoning tasks that require more detailed responses. To address this, we propose a two-fold approach. First, we distill rationales from GPT-4o model to enrich the training data and fine-tune VLMs, boosting their CoT performance. Second, we apply reinforcement learning to further calibrate reasoning quality. Specifically, we construct positive (correct) and negative (incorrect) pairs of model-generated reasoning chains, by comparing their predictions with annotated short answers. Using this pairwise data, we apply the Direct Preference Optimization algorithm to refine the model's reasoning abilities. Our experiments demonstrate significant improvements in CoT reasoning on benchmark datasets and better generalization to direct answer prediction as well. This work emphasizes the importance of incorporating detailed rationales in training and leveraging reinforcement learning to strengthen the reasoning capabilities of VLMs."
                    },
                    {
                        "title": "MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMs",
                        "abstract": "We introduce MIA-Bench, a new benchmark designed to evaluate multimodal large language models (MLLMs) on their ability to strictly adhere to complex instructions. Our benchmark comprises a diverse set of 400 image-prompt pairs, each crafted to challenge the models' compliance with layered instructions in generating accurate responses that satisfy specific requested patterns. Evaluation results from a wide array of state-of-the-art MLLMs reveal significant variations in performance, highlighting areas for improvement in instruction fidelity. Additionally, we create extra training data and explore supervised fine-tuning to enhance the models' ability to strictly follow instructions without compromising performance on other tasks. We hope this benchmark not only serves as a tool for measuring MLLM adherence to instructions, but also guides future developments in MLLM training methods."
                    },
                    {
                        "title": "Understanding Alignment in Multimodal LLMs: A Comprehensive Study",
                        "abstract": "Preference alignment has become a crucial component in enhancing the performance of Large Language Models (LLMs), yet its impact in Multimodal Large Language Models (MLLMs) remains comparatively underexplored. Similar to language models, MLLMs for image understanding tasks encounter challenges like hallucination. In MLLMs, hallucination can occur not only by stating incorrect facts but also by producing responses that are inconsistent with the image content. A primary objective of alignment for MLLMs is to encourage these models to align responses more closely with image information. Recently, multiple works have introduced preference datasets for MLLMs and examined different alignment methods, including Direct Preference Optimization (DPO) and Proximal Policy Optimization (PPO). However, due to variations in datasets, base model types, and alignment methods, it remains unclear which specific elements contribute most significantly to the reported improvements in these works. In this paper, we independently analyze each aspect of preference alignment in MLLMs. We start by categorizing the alignment algorithms into two groups, offline (such as DPO), and online (such as online-DPO), and show that combining offline and online methods can improve the performance of the model in certain scenarios. We review a variety of published multimodal preference datasets and discuss how the details of their construction impact model performance. Based on these insights, we introduce a novel way of creating multimodal preference data called Bias-Driven Hallucination Sampling (BDHS) that needs neither additional annotation nor external models, and show that it can achieve competitive performance to previously published alignment work for multimodal models across a range of benchmarks."
                    },
                    {
                        "title": "Ferret: Refer and Ground Anything Anywhere at Any Granularity",
                        "abstract": "We introduce Ferret, a new Multimodal Large Language Model (MLLM) capable of understanding spatial referring of any shape or granularity within an image and accurately grounding open-vocabulary descriptions. To unify referring and grounding in the LLM paradigm, Ferret employs a novel and powerful hybrid region representation that integrates discrete coordinates and continuous features jointly to represent a region in the image. To extract the continuous features of versatile regions, we propose a spatial-aware visual sampler, adept at handling varying sparsity across different shapes. Consequently, Ferret can accept diverse region inputs, such as points, bounding boxes, and free-form shapes. To bolster the desired capability of Ferret, we curate GRIT, a comprehensive refer-and-ground instruction tuning dataset including 1.1M samples that contain rich hierarchical spatial knowledge, with 95K hard negative data to promote model robustness. The resulting model not only achieves superior performance in classical referring and grounding tasks, but also greatly outperforms existing MLLMs in region-based and localization-demanded multimodal chatting. Our evaluations also reveal a significantly improved capability of describing image details and a remarkable alleviation in object hallucination. Code and data will be available at https://github.com/apple/ml-ferret"
                    }
                ]
            },
            "6c312902-db95-4d61-adbb-88b8c70732f5": {
                "pk": "6c312902-db95-4d61-adbb-88b8c70732f5",
                "name": "Xianzhi Du",
                "collaborators": [
                    "Yinfei Yang",
                    "Zirui Wang",
                    "Bowen Zhang",
                    "Haotian Zhang",
                    "Tom Gunter",
                    "Ruoming Pang",
                    "Zhe Gan",
                    "Aonan Zhang",
                    "Floris Weers",
                    "Mark Lee",
                    "Nan Du",
                    "Xiang Kong",
                    "Zhengfeng Lai",
                    "Jiulong Shan",
                    "Meng Cao",
                    "Chong Wang",
                    "Guoli Yin",
                    "Jianyu Wang",
                    "Sam Wiseman",
                    "Tao Lei",
                    "Yanghao Li",
                    "Ankur Jain",
                    "Doug Kang",
                    "Karanjeet Singh",
                    "Liangliang Cao",
                    "Philipp Dufter",
                    "Dhruti Shah",
                    "Sam Dodge",
                    "J. Fauconnier",
                    "Haoxuan You",
                    "Peter Grasch",
                    "Alexander Toshev",
                    "Haoping Bai",
                    "Chen-Nee Chuah",
                    "Andy Narayanan",
                    "Chen Chen",
                    "Chung-Cheng Chiu",
                    "David Qiu",
                    "Deepak Gopinath",
                    "Dian Ang Yap",
                    "Dong Yin",
                    "Feng Nan",
                    "Haoshuo Huang",
                    "Jiarui Lu",
                    "John Peebles",
                    "Kewei Ye",
                    "Qibin Chen",
                    "Quentin Keunebroek",
                    "Syd Evans",
                    "Vivek Rathod",
                    "Yongqiang Wang",
                    "Yuan Gao",
                    "Zaid Ahmed",
                    "Zhaoyang Xu",
                    "Zhiyun Lu",
                    "Al Rashid",
                    "Albin Madappally Jose",
                    "Alec Doane",
                    "Alfredo Bencomo",
                    "Allison Vanderby",
                    "Andrew Hansen",
                    "A. Anupama",
                    "Areeba Kamal",
                    "Bugu Wu",
                    "Carolina Brum",
                    "Charlie Maalouf",
                    "Chinguun Erdenebileg",
                    "Chris Dulhanty",
                    "Dominik Moritz",
                    "Eduardo Jimenez",
                    "Evan Ladd",
                    "Fang Shi",
                    "Felix Bai",
                    "Frank Chu",
                    "Fred Hohman",
                    "Hadas Kotek",
                    "Hannah Gillis Coleman",
                    "Jane Li",
                    "Jeffrey P. Bigham",
                    "Jeffery Cao",
                    "Jeff Lai",
                    "Jessica Cheung",
                    "Joe Zhou",
                    "John Li",
                    "Jun Qin",
                    "Karla Vega",
                    "Kelvin Zou",
                    "Laura Heckman",
                    "Lauren Gardiner",
                    "Margit Bowler",
                    "Maria Cordell",
                    "Nicole Hay",
                    "Nilesh Shahdadpuri",
                    "Otto Godwin",
                    "Pranay Dighe",
                    "Pushyami Rachapudi",
                    "Ramsey Tantawi",
                    "Roman Frigg",
                    "Sam Davarnia",
                    "Sanskruti Shah"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Domain Adaptation",
                    "Large Language Models",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Empowering Unsupervised Domain Adaptation with Large-scale Pre-trained Vision-Language Models",
                        "abstract": "Unsupervised Domain Adaptation (UDA) aims to leverage the labeled source domain to solve the tasks on the unlabeled target domain. Traditional UDA methods face the challenge of the tradeoff between domain alignment and semantic class discriminability, especially when a large domain gap exists between the source and target domains. The efforts of applying large-scale pre-training to bridge the domain gaps remain limited. In this work, we propose that Vision-Language Models (VLMs) can empower UDA tasks due to their training pattern with language alignment and their large-scale pre-trained datasets. For example, CLIP and GLIP have shown promising zero-shot generalization in classification and detection tasks. However, directly fine-tuning these VLMs into downstream tasks may be computationally expensive and not scalable if we have multiple domains that need to be adapted. Therefore, in this work, we first study an efficient adaption of VLMs to preserve the original knowledge while maximizing its flexibility for learning new knowledge. Then, we design a domain-aware pseudo-labeling scheme tailored to VLMs for domain disentanglement. We show the superiority of the proposed methods in four UDA-classification and two UDA-detection benchmarks, with a significant improvement (+9.9%) on DomainNet."
                    },
                    {
                        "title": "Apple Intelligence Foundation Language Models",
                        "abstract": "We present foundation language models developed to power Apple Intelligence features, including a ~3 billion parameter model designed to run efficiently on devices and a large server-based language model designed for Private Cloud Compute. These models are designed to perform a wide range of tasks efficiently, accurately, and responsibly. This report describes the model architecture, the data used to train the model, the training process, how the models are optimized for inference, and the evaluation results. We highlight our focus on Responsible AI and how the principles are applied throughout the model development."
                    },
                    {
                        "title": "MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning",
                        "abstract": "We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development."
                    },
                    {
                        "title": "Revisiting MoE and Dense Speed-Accuracy Comparisons for LLM Training",
                        "abstract": "Mixture-of-Experts (MoE) enjoys performance gain by increasing model capacity while keeping computation cost constant. When comparing MoE to dense models, prior work typically adopt the following setting: 1) use FLOPs or activated parameters as a measure of model complexity; 2) train all models to the same number of tokens. We argue that this setting favors MoE as FLOPs and activated parameters do not accurately measure the communication overhead in sparse layers, leading to a larger actual training budget for MoE. In this work, we revisit the settings by adopting step time as a more accurate measure of model complexity, and by determining the total compute budget under the Chinchilla compute-optimal settings. To efficiently run MoE on modern accelerators, we adopt a 3D sharding method that keeps the dense-to-MoE step time increase within a healthy range. We evaluate MoE and dense LLMs on a set of nine 0-shot and two 1-shot English tasks, as well as MMLU 5-shot and GSM8K 8-shot across three model scales at 6.4B, 12.6B, and 29.6B. Experimental results show that even under these settings, MoE consistently outperform dense LLMs on the speed-accuracy trade-off curve with meaningful gaps. Our full model implementation and sharding strategy has been released at~\\url{https://github.com/apple/axlearn}"
                    },
                    {
                        "title": "MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training",
                        "abstract": "In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting."
                    },
                    {
                        "title": "Guiding Instruction-based Image Editing via Multimodal Large Language Models",
                        "abstract": "Instruction-based image editing improves the controllability and flexibility of image manipulation via natural commands without elaborate descriptions or regional masks. However, human instructions are sometimes too brief for current methods to capture and follow. Multimodal large language models (MLLMs) show promising capabilities in cross-modal understanding and visual-aware response generation via LMs. We investigate how MLLMs facilitate edit instructions and present MLLM-Guided Image Editing (MGIE). MGIE learns to derive expressive instructions and provides explicit guidance. The editing model jointly captures this visual imagination and performs manipulation through end-to-end training. We evaluate various aspects of Photoshop-style modification, global photo optimization, and local editing. Extensive experimental results demonstrate that expressive instructions are crucial to instruction-based image editing, and our MGIE can lead to a notable improvement in automatic metrics and human evaluation while maintaining competitive inference efficiency."
                    },
                    {
                        "title": "Mobile V-MoEs: Scaling Down Vision Transformers via Sparse Mixture-of-Experts",
                        "abstract": "Sparse Mixture-of-Experts models (MoEs) have recently gained popularity due to their ability to decouple model size from inference efficiency by only activating a small subset of the model parameters for any given input token. As such, sparse MoEs have enabled unprecedented scalability, resulting in tremendous successes across domains such as natural language processing and computer vision. In this work, we instead explore the use of sparse MoEs to scale-down Vision Transformers (ViTs) to make them more attractive for resource-constrained vision applications. To this end, we propose a simplified and mobile-friendly MoE design where entire images rather than individual patches are routed to the experts. We also propose a stable MoE training procedure that uses super-class information to guide the router. We empirically show that our sparse Mobile Vision MoEs (V-MoEs) can achieve a better trade-off between performance and efficiency than the corresponding dense ViTs. For example, for the ViT-Tiny model, our Mobile V-MoE outperforms its dense counterpart by 3.39% on ImageNet-1k. For an even smaller ViT variant with only 54M FLOPs inference cost, our MoE achieves an improvement of 4.66%."
                    },
                    {
                        "title": "Ferret: Refer and Ground Anything Anywhere at Any Granularity",
                        "abstract": "We introduce Ferret, a new Multimodal Large Language Model (MLLM) capable of understanding spatial referring of any shape or granularity within an image and accurately grounding open-vocabulary descriptions. To unify referring and grounding in the LLM paradigm, Ferret employs a novel and powerful hybrid region representation that integrates discrete coordinates and continuous features jointly to represent a region in the image. To extract the continuous features of versatile regions, we propose a spatial-aware visual sampler, adept at handling varying sparsity across different shapes. Consequently, Ferret can accept diverse region inputs, such as points, bounding boxes, and free-form shapes. To bolster the desired capability of Ferret, we curate GRIT, a comprehensive refer-and-ground instruction tuning dataset including 1.1M samples that contain rich hierarchical spatial knowledge, with 95K hard negative data to promote model robustness. The resulting model not only achieves superior performance in classical referring and grounding tasks, but also greatly outperforms existing MLLMs in region-based and localization-demanded multimodal chatting. Our evaluations also reveal a significantly improved capability of describing image details and a remarkable alleviation in object hallucination. Code and data will be available at https://github.com/apple/ml-ferret"
                    }
                ]
            },
            "45022483-e2a7-4f26-b58f-8a5614ec6fee": {
                "pk": "45022483-e2a7-4f26-b58f-8a5614ec6fee",
                "name": "Bowen Zhang",
                "collaborators": [
                    "Haotian Zhang",
                    "Yinfei Yang",
                    "Zhe Gan",
                    "Yanghao Li",
                    "Hong-You Chen",
                    "Zirui Wang",
                    "Ruoming Pang",
                    "Xianzhi Du",
                    "Haoxuan You",
                    "Zhengfeng Lai",
                    "Meng Cao",
                    "Chen Chen",
                    "Nan Du",
                    "Tao Lei",
                    "Philipp Dufter",
                    "Peter Grasch",
                    "Keen You",
                    "Tom Gunter",
                    "Chong Wang",
                    "Aonan Zhang",
                    "Floris Weers",
                    "Guoli Yin",
                    "Haoshuo Huang",
                    "Jianyu Wang",
                    "Mark Lee",
                    "Sam Wiseman",
                    "Xiang Kong",
                    "Ankur Jain",
                    "Doug Kang",
                    "Karanjeet Singh",
                    "Liangliang Cao",
                    "Dhruti Shah",
                    "Sam Dodge",
                    "J. Fauconnier",
                    "Shih-Fu Chang",
                    "Xinze Wang",
                    "Marcin Eichner",
                    "Andy Narayanan",
                    "Chung-Cheng Chiu",
                    "David Qiu",
                    "Deepak Gopinath",
                    "Dian Ang Yap",
                    "Dong Yin",
                    "Feng Nan",
                    "Jiarui Lu",
                    "John Peebles",
                    "Kewei Ye",
                    "Qibin Chen",
                    "Quentin Keunebroek",
                    "Syd Evans",
                    "Vivek Rathod",
                    "Yongqiang Wang",
                    "Yuan Gao",
                    "Zaid Ahmed",
                    "Zhaoyang Xu",
                    "Zhiyun Lu",
                    "Al Rashid",
                    "Albin Madappally Jose",
                    "Alec Doane",
                    "Alfredo Bencomo",
                    "Allison Vanderby",
                    "Andrew Hansen",
                    "A. Anupama",
                    "Areeba Kamal",
                    "Bugu Wu",
                    "Carolina Brum",
                    "Charlie Maalouf",
                    "Chinguun Erdenebileg",
                    "Chris Dulhanty",
                    "Dominik Moritz",
                    "Eduardo Jimenez",
                    "Evan Ladd",
                    "Fang Shi",
                    "Felix Bai",
                    "Frank Chu",
                    "Fred Hohman",
                    "Hadas Kotek",
                    "Hannah Gillis Coleman",
                    "Jane Li",
                    "Jeffrey P. Bigham",
                    "Jeffery Cao",
                    "Jeff Lai",
                    "Jessica Cheung",
                    "Jiulong Shan",
                    "Joe Zhou",
                    "John Li",
                    "Jun Qin",
                    "Karla Vega",
                    "Kelvin Zou",
                    "Laura Heckman",
                    "Lauren Gardiner",
                    "Margit Bowler",
                    "Maria Cordell",
                    "Nicole Hay",
                    "Nilesh Shahdadpuri",
                    "Otto Godwin",
                    "Pranay Dighe",
                    "Pushyami Rachapudi",
                    "Ramsey Tantawi",
                    "Roman Frigg"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Vision-Language Models",
                    "Foundation Models",
                    "Image Understanding"
                ],
                "publications": [
                    {
                        "title": "Contrastive Localized Language-Image Pre-Training",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) has been a celebrated method for training vision encoders to generate image/text representations facilitating various applications. Recently, CLIP has been widely adopted as the vision backbone of multimodal large language models (MLLMs) to connect image inputs for language interactions. The success of CLIP as a vision-language foundation model relies on aligning web-crawled noisy text annotations at image levels. Nevertheless, such criteria may become insufficient for downstream tasks in need of fine-grained vision representations, especially when region-level understanding is demanding for MLLMs. In this paper, we improve the localization capability of CLIP with several advances. We propose a pre-training method called Contrastive Localized Language-Image Pre-training (CLOC) by complementing CLIP with region-text contrastive loss and modules. We formulate a new concept, promptable embeddings, of which the encoder produces image embeddings easy to transform into region representations given spatial hints. To support large-scale pre-training, we design a visually-enriched and spatially-localized captioning framework to effectively generate region-text pseudo-labels at scale. By scaling up to billions of annotated images, CLOC enables high-quality regional embeddings for image region recognition and retrieval tasks, and can be a drop-in replacement of CLIP to enhance MLLMs, especially on referring and grounding tasks."
                    },
                    {
                        "title": "Apple Intelligence Foundation Language Models",
                        "abstract": "We present foundation language models developed to power Apple Intelligence features, including a ~3 billion parameter model designed to run efficiently on devices and a large server-based language model designed for Private Cloud Compute. These models are designed to perform a wide range of tasks efficiently, accurately, and responsibly. This report describes the model architecture, the data used to train the model, the training process, how the models are optimized for inference, and the evaluation results. We highlight our focus on Responsible AI and how the principles are applied throughout the model development."
                    },
                    {
                        "title": "EC-DIT: Scaling Diffusion Transformers with Adaptive Expert-Choice Routing",
                        "abstract": "Diffusion transformers have been widely adopted for text-to-image synthesis. While scaling these models up to billions of parameters shows promise, the effectiveness of scaling beyond current sizes remains underexplored and challenging. By explicitly exploiting the computational heterogeneity of image generations, we develop a new family of Mixture-of-Experts (MoE) models (EC-DIT) for diffusion transformers with expert-choice routing. EC-DIT learns to adaptively optimize the compute allocated to understand the input texts and generate the respective image patches, enabling heterogeneous computation aligned with varying text-image complexities. This heterogeneity provides an efficient way of scaling EC-DIT up to 97 billion parameters and achieving significant improvements in training convergence, text-to-image alignment, and overall generation quality over dense models and conventional MoE models. Through extensive ablations, we show that EC-DIT demonstrates superior scalability and adaptive compute allocation by recognizing varying textual importance through end-to-end training. Notably, in text-to-image alignment evaluation, our largest models achieve a state-of-the-art GenEval score of 71.68% and still maintain competitive inference speed with intuitive interpretability."
                    },
                    {
                        "title": "MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning",
                        "abstract": "We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development."
                    },
                    {
                        "title": "MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training",
                        "abstract": "In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting."
                    },
                    {
                        "title": "Revisit Large-Scale Image-Caption Data in Pre-training Multimodal Foundation Models",
                        "abstract": "Recent advancements in multimodal models highlight the value of rewritten captions for improving performance, yet key challenges remain. For example, while synthetic captions often provide superior quality and image-text alignment, it is not clear whether they can fully replace AltTexts: the role of synthetic captions and their interaction with original web-crawled AltTexts in pre-training is still not well understood. Moreover, different multimodal foundation models may have unique preferences for specific caption formats, but efforts to identify the optimal captions for each model remain limited. In this work, we propose a novel, controllable, and scalable captioning pipeline designed to generate diverse caption formats tailored to various multimodal models. By examining Short Synthetic Captions (SSC) towards Dense Synthetic Captions (DSC+) as case studies, we systematically explore their effects and interactions with AltTexts across models such as CLIP, multimodal LLMs, and diffusion models. Our findings reveal that a hybrid approach that keeps both synthetic captions and AltTexts can outperform the use of synthetic captions alone, improving both alignment and performance, with each model demonstrating preferences for particular caption formats. This comprehensive analysis provides valuable insights into optimizing captioning strategies, thereby advancing the pre-training of multimodal foundation models."
                    },
                    {
                        "title": "Ferret-v2: An Improved Baseline for Referring and Grounding with Large Language Models",
                        "abstract": "While Ferret seamlessly integrates regional understanding into the Large Language Model (LLM) to facilitate its referring and grounding capability, it poses certain limitations: constrained by the pre-trained fixed visual encoder and failed to perform well on broader tasks. In this work, we unveil Ferret-v2, a significant upgrade to Ferret, with three key designs. (1) Any resolution grounding and referring: A flexible approach that effortlessly handles higher image resolution, improving the model's ability to process and understand images in greater detail. (2) Multi-granularity visual encoding: By integrating the additional DINOv2 encoder, the model learns better and diverse underlying contexts for global and fine-grained visual information. (3) A three-stage training paradigm: Besides image-caption alignment, an additional stage is proposed for high-resolution dense alignment before the final instruction tuning. Experiments show that Ferret-v2 provides substantial improvements over Ferret and other state-of-the-art methods, thanks to its high-resolution scaling and fine-grained visual processing."
                    },
                    {
                        "title": "MM-Ego: Towards Building Egocentric Multimodal LLMs",
                        "abstract": "This research aims to comprehensively explore building a multimodal foundation model for egocentric video understanding. To achieve this goal, we work on three fronts. First, as there is a lack of QA data for egocentric video understanding, we develop a data engine that efficiently generates 7M high-quality QA samples for egocentric videos ranging from 30 seconds to one hour long, based on human-annotated data. This is currently the largest egocentric QA dataset. Second, we contribute a challenging egocentric QA benchmark with 629 videos and 7,026 questions to evaluate the models' ability in recognizing and memorizing visual details across videos of varying lengths. We introduce a new de-biasing evaluation method to help mitigate the unavoidable language bias present in the models being evaluated. Third, we propose a specialized multimodal architecture featuring a novel\"Memory Pointer Prompting\"mechanism. This design includes a global glimpse step to gain an overarching understanding of the entire video and identify key visual information, followed by a fallback step that utilizes the key visual information to generate responses. This enables the model to more effectively comprehend extended video content. With the data, benchmark, and model, we successfully build MM-Ego, an egocentric multimodal LLM that shows powerful performance on egocentric video understanding."
                    },
                    {
                        "title": "Improve Vision Language Model Chain-of-thought Reasoning",
                        "abstract": "Chain-of-thought (CoT) reasoning in vision language models (VLMs) is crucial for improving interpretability and trustworthiness. However, current training recipes lack robust CoT reasoning data, relying on datasets dominated by short annotations with minimal rationales. In this work, we show that training VLM on short answers does not generalize well to reasoning tasks that require more detailed responses. To address this, we propose a two-fold approach. First, we distill rationales from GPT-4o model to enrich the training data and fine-tune VLMs, boosting their CoT performance. Second, we apply reinforcement learning to further calibrate reasoning quality. Specifically, we construct positive (correct) and negative (incorrect) pairs of model-generated reasoning chains, by comparing their predictions with annotated short answers. Using this pairwise data, we apply the Direct Preference Optimization algorithm to refine the model's reasoning abilities. Our experiments demonstrate significant improvements in CoT reasoning on benchmark datasets and better generalization to direct answer prediction as well. This work emphasizes the importance of incorporating detailed rationales in training and leveraging reinforcement learning to strengthen the reasoning capabilities of VLMs."
                    },
                    {
                        "title": "Ferret: Refer and Ground Anything Anywhere at Any Granularity",
                        "abstract": "We introduce Ferret, a new Multimodal Large Language Model (MLLM) capable of understanding spatial referring of any shape or granularity within an image and accurately grounding open-vocabulary descriptions. To unify referring and grounding in the LLM paradigm, Ferret employs a novel and powerful hybrid region representation that integrates discrete coordinates and continuous features jointly to represent a region in the image. To extract the continuous features of versatile regions, we propose a spatial-aware visual sampler, adept at handling varying sparsity across different shapes. Consequently, Ferret can accept diverse region inputs, such as points, bounding boxes, and free-form shapes. To bolster the desired capability of Ferret, we curate GRIT, a comprehensive refer-and-ground instruction tuning dataset including 1.1M samples that contain rich hierarchical spatial knowledge, with 95K hard negative data to promote model robustness. The resulting model not only achieves superior performance in classical referring and grounding tasks, but also greatly outperforms existing MLLMs in region-based and localization-demanded multimodal chatting. Our evaluations also reveal a significantly improved capability of describing image details and a remarkable alleviation in object hallucination. Code and data will be available at https://github.com/apple/ml-ferret"
                    }
                ]
            },
            "8d590d32-3ad5-4853-a829-e530d1ec783b": {
                "pk": "8d590d32-3ad5-4853-a829-e530d1ec783b",
                "name": "Zhangyang Wang",
                "collaborators": [
                    "A. Jaiswal",
                    "Lu Yin",
                    "Shiwei Liu",
                    "Souvik Kundu",
                    "Nurendra Choudhary",
                    "Ravinarayana Adkathimar",
                    "M. P. Alagappan",
                    "G. Hiranandani",
                    "Ying Ding",
                    "E-Wen Huang",
                    "Karthik Subbian",
                    "Runjin Chen",
                    "Tong Zhao",
                    "Neil Shah",
                    "You Wu",
                    "Zhenyu (Allen) Zhang",
                    "Cheng-Yu Hsieh",
                    "Yaqing Wang",
                    "Yiling Jia",
                    "Mykola Pechenizkiy",
                    "Yi Liang"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Large Language Models",
                    "Model Compression",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "All Against Some: Efficient Integration of Large Language Models for Message Passing in Graph Neural Networks",
                        "abstract": "Graph Neural Networks (GNNs) have attracted immense attention in the past decade due to their numerous real-world applications built around graph-structured data. On the other hand, Large Language Models (LLMs) with extensive pretrained knowledge and powerful semantic comprehension abilities have recently shown a remarkable ability to benefit applications using vision and text data. In this paper, we investigate how LLMs can be leveraged in a computationally efficient fashion to benefit rich graph-structured data, a modality relatively unexplored in LLM literature. Prior works in this area exploit LLMs to augment every node features in an ad-hoc fashion (not scalable for large graphs), use natural language to describe the complex structural information of graphs, or perform computationally expensive finetuning of LLMs in conjunction with GNNs. We propose E-LLaGNN (Efficient LLMs augmented GNNs), a framework with an on-demand LLM service that enriches message passing procedure of graph learning by enhancing a limited fraction of nodes from the graph. More specifically, E-LLaGNN relies on sampling high-quality neighborhoods using LLMs, followed by on-demand neighborhood feature enhancement using diverse prompts from our prompt catalog, and finally information aggregation using message passing from conventional GNN architectures. We explore several heuristics-based active node selection strategies to limit the computational and memory footprint of LLMs when handling millions of nodes. Through extensive experiments&ablation on popular graph benchmarks of varying scales (Cora, PubMed, ArXiv,&Products), we illustrate the effectiveness of our E-LLaGNN framework and reveal many interesting capabilities such as improved gradient flow in deep GNNs, LLM-free inference ability etc."
                    },
                    {
                        "title": "LLaGA: Large Language and Graph Assistant",
                        "abstract": "Graph Neural Networks (GNNs) have empowered the advance in graph-structured data analysis. Recently, the rise of Large Language Models (LLMs) like GPT-4 has heralded a new era in deep learning. However, their application to graph data poses distinct challenges due to the inherent difficulty of translating graph structures to language. To this end, we introduce the Large Language and Graph Assistant (LLaGA), an innovative model that effectively integrates LLM capabilities to handle the complexities of graph-structured data. LLaGA retains the general-purpose nature of LLMs while adapting graph data into a format compatible with LLM input. LLaGA achieves this by reorganizing graph nodes to structure-aware sequences and then mapping these into the token embedding space through a versatile projector. LLaGA excels in versatility, generalizability and interpretability, allowing it to perform consistently well across different datasets and tasks, extend its ability to unseen datasets or tasks, and provide explanations for graphs. Our extensive experiments across popular graph benchmarks show that LLaGA delivers outstanding performance across four datasets and three tasks using one single model, surpassing state-of-the-art graph models in both supervised and zero-shot scenarios. Our code is available at \\url{https://github.com/VITA-Group/LLaGA}."
                    },
                    {
                        "title": "Junk DNA Hypothesis: A Task-Centric Angle of LLM Pre-trained Weights through Sparsity",
                        "abstract": "validate"
                    },
                    {
                        "title": "Pruning Small Pre-Trained Weights Irreversibly and Monotonically Impairs\"Difficult\"Downstream Tasks in LLMs",
                        "abstract": "We present Junk DNA Hypothesis by adopting a novel task-centric angle for the pre-trained weights of large language models (LLMs). It has been believed that weights in LLMs contain significant redundancy, leading to the conception that a considerable chunk of the parameters can be removed by pruning without compromising performance. Contrary to this belief, this paper presents a counter-argument: small-magnitude weights of pre-trained model weights encode vital knowledge essential for tackling difficult downstream tasks - manifested as the monotonic relationship between the performance drop of downstream tasks across the difficulty spectrum, as we prune more pre-trained weights by magnitude. Moreover, we reveal that these seemingly inconsequential weights can result in irreparable loss of knowledge and performance degradation in difficult tasks, even when downstream continual training is allowed. Interestingly, our evaluations show that the other popular compression, namely quantization, fails to exhibit similar monotonic effect and does not as convincingly disentangle this task-difficulty information. To study formally, we introduce several quantifiable metrics to gauge the downstream task difficulty: (1) within the same task category, and (2) across different task categories. Our extensive experiments substantiate the Junk DNA Hypothesis across a diverse range of model sizes, tasks, datasets, and even pruning methods. Codes are available at: https://github.com/VITA-Group/Junk_DNA_Hypothesis.git."
                    },
                    {
                        "title": "Outlier Weighed Layerwise Sparsity (OWL): A Missing Secret Sauce for Pruning LLMs to High Sparsity",
                        "abstract": "Large Language Models (LLMs), renowned for their remarkable performance across diverse domains, present a challenge when it comes to practical deployment due to their colossal model size. In response to this challenge, efforts have been directed toward the application of traditional network pruning techniques to LLMs, uncovering a massive number of parameters that can be pruned in one-shot without hurting performance. Prevailing LLM pruning strategies have consistently adhered to the practice of uniformly pruning all layers at equivalent sparsity, resulting in robust performance. However, this observation stands in contrast to the prevailing trends observed in the field of vision models, where non-uniform layerwise sparsity typically yields stronger results. To understand the underlying reasons for this disparity, we conduct a comprehensive study and discover a strong correlation with the emergence of activation outliers in LLMs. Inspired by this finding, we introduce a novel LLM pruning methodology that incorporates a tailored set of non-uniform layerwise sparsity ratios, termed as Outlier Weighed Layerwise sparsity (OWL). The sparsity ratio of OWL is proportional to the outlier ratio observed within each layer, facilitating a more effective alignment between layerwise weight sparsity and outlier ratios. Our empirical evaluation, conducted across the LLaMA-V1 family and OPT, spanning various benchmarks, demonstrates the distinct advantages offered by OWL over previous methods. For instance, OWL exhibits a remarkable performance gain, surpassing the state-of-the-art Wanda and SparseGPT by 61.22 and 6.80 perplexity at a high sparsity level of 70%, respectively, while delivering 2.6x end-to-end inference speed-up in the DeepSparse inference engine. Codes are available at https://github.com/luuyin/OWL."
                    }
                ]
            },
            "898de847-3e56-4379-bf70-01ea240bf5ee": {
                "pk": "898de847-3e56-4379-bf70-01ea240bf5ee",
                "name": "Yinfei Yang",
                "collaborators": [
                    "Zhe Gan",
                    "Haotian Zhang",
                    "Bowen Zhang",
                    "Xianzhi Du",
                    "Peter Grasch",
                    "Zhengfeng Lai",
                    "Hong-You Chen",
                    "Keen You",
                    "J. Fauconnier",
                    "Haoxuan You",
                    "Zirui Wang",
                    "Meng Cao",
                    "Philipp Dufter",
                    "Yanghao Li",
                    "Yusu Qian",
                    "Dhruti Shah",
                    "Sam Dodge",
                    "Afshin Dehghan",
                    "Floris Weers",
                    "Ruoming Pang",
                    "Chen Chen",
                    "Wenze Hu",
                    "Tsu-Jui Fu",
                    "William Yang Wang",
                    "Shih-Fu Chang",
                    "Hanrong Ye",
                    "Haoping Bai",
                    "Jiulong Shan",
                    "Chen-Nee Chuah",
                    "Xinze Wang",
                    "Marcin Eichner",
                    "Mingfei Gao",
                    "Nina Wenzel",
                    "Forrest Huang",
                    "Zhen Yang",
                    "Aleksei Timofeev",
                    "Mingze Xu",
                    "E. Schoop",
                    "Amanda Swearngin",
                    "Jeffrey Nichols",
                    "Zhangheng Li",
                    "Di Feng",
                    "Harsh Agrawal",
                    "Xiujun Li",
                    "M. Moorthy",
                    "Jeff Nichols",
                    "Brandon McKinzie",
                    "Futang Peng",
                    "Anton Belyi",
                    "Karanjeet Singh",
                    "Doug Kang",
                    "Ankur Jain",
                    "Hongyu He",
                    "Max Schwarzer",
                    "Tom Gunter",
                    "Xiang Kong",
                    "Aonan Zhang",
                    "Jianyu Wang",
                    "Chong Wang",
                    "Nan Du",
                    "Tao Lei",
                    "Sam Wiseman",
                    "Guoli Yin",
                    "Mark Lee",
                    "Alexander Toshev",
                    "Vasileios Saveris",
                    "Juan Lao Tebar",
                    "Erik Daxberger",
                    "Lin Chen",
                    "Zongyu Lin",
                    "Dan Xu",
                    "Jiasen Lu",
                    "Ruohong Zhang",
                    "Zhiqing Sun",
                    "Yiming Yang",
                    "Elmira Amirloo",
                    "Christoph Roesmann",
                    "Christian Kerl",
                    "Rinu Boney",
                    "Liangliang Cao"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Vision-Language Models",
                    "Domain Adaptation",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "Empowering Unsupervised Domain Adaptation with Large-scale Pre-trained Vision-Language Models",
                        "abstract": "Unsupervised Domain Adaptation (UDA) aims to leverage the labeled source domain to solve the tasks on the unlabeled target domain. Traditional UDA methods face the challenge of the tradeoff between domain alignment and semantic class discriminability, especially when a large domain gap exists between the source and target domains. The efforts of applying large-scale pre-training to bridge the domain gaps remain limited. In this work, we propose that Vision-Language Models (VLMs) can empower UDA tasks due to their training pattern with language alignment and their large-scale pre-trained datasets. For example, CLIP and GLIP have shown promising zero-shot generalization in classification and detection tasks. However, directly fine-tuning these VLMs into downstream tasks may be computationally expensive and not scalable if we have multiple domains that need to be adapted. Therefore, in this work, we first study an efficient adaption of VLMs to preserve the original knowledge while maximizing its flexibility for learning new knowledge. Then, we design a domain-aware pseudo-labeling scheme tailored to VLMs for domain disentanglement. We show the superiority of the proposed methods in four UDA-classification and two UDA-detection benchmarks, with a significant improvement (+9.9%) on DomainNet."
                    },
                    {
                        "title": "Contrastive Localized Language-Image Pre-Training",
                        "abstract": "Contrastive Language-Image Pre-training (CLIP) has been a celebrated method for training vision encoders to generate image/text representations facilitating various applications. Recently, CLIP has been widely adopted as the vision backbone of multimodal large language models (MLLMs) to connect image inputs for language interactions. The success of CLIP as a vision-language foundation model relies on aligning web-crawled noisy text annotations at image levels. Nevertheless, such criteria may become insufficient for downstream tasks in need of fine-grained vision representations, especially when region-level understanding is demanding for MLLMs. In this paper, we improve the localization capability of CLIP with several advances. We propose a pre-training method called Contrastive Localized Language-Image Pre-training (CLOC) by complementing CLIP with region-text contrastive loss and modules. We formulate a new concept, promptable embeddings, of which the encoder produces image embeddings easy to transform into region representations given spatial hints. To support large-scale pre-training, we design a visually-enriched and spatially-localized captioning framework to effectively generate region-text pseudo-labels at scale. By scaling up to billions of annotated images, CLOC enables high-quality regional embeddings for image region recognition and retrieval tasks, and can be a drop-in replacement of CLIP to enhance MLLMs, especially on referring and grounding tasks."
                    },
                    {
                        "title": "MM1.5: Methods, Analysis & Insights from Multimodal LLM Fine-tuning",
                        "abstract": "We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development."
                    },
                    {
                        "title": "Ferret-UI: Grounded Mobile UI Understanding with Multimodal LLMs",
                        "abstract": "Recent advancements in multimodal large language models (MLLMs) have been noteworthy, yet, these general-domain MLLMs often fall short in their ability to comprehend and interact effectively with user interface (UI) screens. In this paper, we present Ferret-UI, a new MLLM tailored for enhanced understanding of mobile UI screens, equipped with referring, grounding, and reasoning capabilities. Given that UI screens typically exhibit a more elongated aspect ratio and contain smaller objects of interest (e.g., icons, texts) than natural images, we incorporate\"any resolution\"on top of Ferret to magnify details and leverage enhanced visual features. Specifically, each screen is divided into 2 sub-images based on the original aspect ratio (i.e., horizontal division for portrait screens and vertical division for landscape screens). Both sub-images are encoded separately before being sent to LLMs. We meticulously gather training samples from an extensive range of elementary UI tasks, such as icon recognition, find text, and widget listing. These samples are formatted for instruction-following with region annotations to facilitate precise referring and grounding. To augment the model's reasoning ability, we further compile a dataset for advanced tasks, including detailed description, perception/interaction conversations, and function inference. After training on the curated datasets, Ferret-UI exhibits outstanding comprehension of UI screens and the capability to execute open-ended instructions. For model evaluation, we establish a comprehensive benchmark encompassing all the aforementioned tasks. Ferret-UI excels not only beyond most open-source UI MLLMs, but also surpasses GPT-4V on all the elementary UI tasks."
                    },
                    {
                        "title": "Ferret-UI 2: Mastering Universal User Interface Understanding Across Platforms",
                        "abstract": "Building a generalist model for user interface (UI) understanding is challenging due to various foundational issues, such as platform diversity, resolution variation, and data limitation. In this paper, we introduce Ferret-UI 2, a multimodal large language model (MLLM) designed for universal UI understanding across a wide range of platforms, including iPhone, Android, iPad, Webpage, and AppleTV. Building on the foundation of Ferret-UI, Ferret-UI 2 introduces three key innovations: support for multiple platform types, high-resolution perception through adaptive scaling, and advanced task training data generation powered by GPT-4o with set-of-mark visual prompting. These advancements enable Ferret-UI 2 to perform complex, user-centered interactions, making it highly versatile and adaptable for the expanding diversity of platform ecosystems. Extensive empirical experiments on referring, grounding, user-centric advanced tasks (comprising 9 subtasks $\\times$ 5 platforms), GUIDE next-action prediction dataset, and GUI-World multi-platform benchmark demonstrate that Ferret-UI 2 significantly outperforms Ferret-UI, and also shows strong cross-platform transfer capabilities."
                    },
                    {
                        "title": "MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training",
                        "abstract": "In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting."
                    },
                    {
                        "title": "How Easy is It to Fool Your Multimodal LLMs? An Empirical Analysis on Deceptive Prompts",
                        "abstract": "The remarkable advancements in Multimodal Large Language Models (MLLMs) have not rendered them immune to challenges, particularly in the context of handling deceptive information in prompts, thus producing hallucinated responses under such conditions. To quantitatively assess this vulnerability, we present MAD-Bench, a carefully curated benchmark that contains 1000 test samples divided into 5 categories, such as non-existent objects, count of objects, and spatial relationship. We provide a comprehensive analysis of popular MLLMs, ranging from GPT-4v, Reka, Gemini-Pro, to open-sourced models, such as LLaVA-NeXT and MiniCPM-Llama3. Empirically, we observe significant performance gaps between GPT-4o and other models; and previous robust instruction-tuned models are not effective on this new benchmark. While GPT-4o achieves 82.82% accuracy on MAD-Bench, the accuracy of any other model in our experiments ranges from 9% to 50%. We further propose a remedy that adds an additional paragraph to the deceptive prompts to encourage models to think twice before answering the question. Surprisingly, this simple method can even double the accuracy; however, the absolute numbers are still too low to be satisfactory. We hope MAD-Bench can serve as a valuable benchmark to stimulate further research to enhance model resilience against deceptive prompts."
                    },
                    {
                        "title": "Revisit Large-Scale Image-Caption Data in Pre-training Multimodal Foundation Models",
                        "abstract": "Recent advancements in multimodal models highlight the value of rewritten captions for improving performance, yet key challenges remain. For example, while synthetic captions often provide superior quality and image-text alignment, it is not clear whether they can fully replace AltTexts: the role of synthetic captions and their interaction with original web-crawled AltTexts in pre-training is still not well understood. Moreover, different multimodal foundation models may have unique preferences for specific caption formats, but efforts to identify the optimal captions for each model remain limited. In this work, we propose a novel, controllable, and scalable captioning pipeline designed to generate diverse caption formats tailored to various multimodal models. By examining Short Synthetic Captions (SSC) towards Dense Synthetic Captions (DSC+) as case studies, we systematically explore their effects and interactions with AltTexts across models such as CLIP, multimodal LLMs, and diffusion models. Our findings reveal that a hybrid approach that keeps both synthetic captions and AltTexts can outperform the use of synthetic captions alone, improving both alignment and performance, with each model demonstrating preferences for particular caption formats. This comprehensive analysis provides valuable insights into optimizing captioning strategies, thereby advancing the pre-training of multimodal foundation models."
                    },
                    {
                        "title": "Ferret-v2: An Improved Baseline for Referring and Grounding with Large Language Models",
                        "abstract": "While Ferret seamlessly integrates regional understanding into the Large Language Model (LLM) to facilitate its referring and grounding capability, it poses certain limitations: constrained by the pre-trained fixed visual encoder and failed to perform well on broader tasks. In this work, we unveil Ferret-v2, a significant upgrade to Ferret, with three key designs. (1) Any resolution grounding and referring: A flexible approach that effortlessly handles higher image resolution, improving the model's ability to process and understand images in greater detail. (2) Multi-granularity visual encoding: By integrating the additional DINOv2 encoder, the model learns better and diverse underlying contexts for global and fine-grained visual information. (3) A three-stage training paradigm: Besides image-caption alignment, an additional stage is proposed for high-resolution dense alignment before the final instruction tuning. Experiments show that Ferret-v2 provides substantial improvements over Ferret and other state-of-the-art methods, thanks to its high-resolution scaling and fine-grained visual processing."
                    },
                    {
                        "title": "MM-Ego: Towards Building Egocentric Multimodal LLMs",
                        "abstract": "This research aims to comprehensively explore building a multimodal foundation model for egocentric video understanding. To achieve this goal, we work on three fronts. First, as there is a lack of QA data for egocentric video understanding, we develop a data engine that efficiently generates 7M high-quality QA samples for egocentric videos ranging from 30 seconds to one hour long, based on human-annotated data. This is currently the largest egocentric QA dataset. Second, we contribute a challenging egocentric QA benchmark with 629 videos and 7,026 questions to evaluate the models' ability in recognizing and memorizing visual details across videos of varying lengths. We introduce a new de-biasing evaluation method to help mitigate the unavoidable language bias present in the models being evaluated. Third, we propose a specialized multimodal architecture featuring a novel\"Memory Pointer Prompting\"mechanism. This design includes a global glimpse step to gain an overarching understanding of the entire video and identify key visual information, followed by a fallback step that utilizes the key visual information to generate responses. This enables the model to more effectively comprehend extended video content. With the data, benchmark, and model, we successfully build MM-Ego, an egocentric multimodal LLM that shows powerful performance on egocentric video understanding."
                    },
                    {
                        "title": "Improve Vision Language Model Chain-of-thought Reasoning",
                        "abstract": "Chain-of-thought (CoT) reasoning in vision language models (VLMs) is crucial for improving interpretability and trustworthiness. However, current training recipes lack robust CoT reasoning data, relying on datasets dominated by short annotations with minimal rationales. In this work, we show that training VLM on short answers does not generalize well to reasoning tasks that require more detailed responses. To address this, we propose a two-fold approach. First, we distill rationales from GPT-4o model to enrich the training data and fine-tune VLMs, boosting their CoT performance. Second, we apply reinforcement learning to further calibrate reasoning quality. Specifically, we construct positive (correct) and negative (incorrect) pairs of model-generated reasoning chains, by comparing their predictions with annotated short answers. Using this pairwise data, we apply the Direct Preference Optimization algorithm to refine the model's reasoning abilities. Our experiments demonstrate significant improvements in CoT reasoning on benchmark datasets and better generalization to direct answer prediction as well. This work emphasizes the importance of incorporating detailed rationales in training and leveraging reinforcement learning to strengthen the reasoning capabilities of VLMs."
                    },
                    {
                        "title": "MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMs",
                        "abstract": "We introduce MIA-Bench, a new benchmark designed to evaluate multimodal large language models (MLLMs) on their ability to strictly adhere to complex instructions. Our benchmark comprises a diverse set of 400 image-prompt pairs, each crafted to challenge the models' compliance with layered instructions in generating accurate responses that satisfy specific requested patterns. Evaluation results from a wide array of state-of-the-art MLLMs reveal significant variations in performance, highlighting areas for improvement in instruction fidelity. Additionally, we create extra training data and explore supervised fine-tuning to enhance the models' ability to strictly follow instructions without compromising performance on other tasks. We hope this benchmark not only serves as a tool for measuring MLLM adherence to instructions, but also guides future developments in MLLM training methods."
                    },
                    {
                        "title": "Understanding Alignment in Multimodal LLMs: A Comprehensive Study",
                        "abstract": "Preference alignment has become a crucial component in enhancing the performance of Large Language Models (LLMs), yet its impact in Multimodal Large Language Models (MLLMs) remains comparatively underexplored. Similar to language models, MLLMs for image understanding tasks encounter challenges like hallucination. In MLLMs, hallucination can occur not only by stating incorrect facts but also by producing responses that are inconsistent with the image content. A primary objective of alignment for MLLMs is to encourage these models to align responses more closely with image information. Recently, multiple works have introduced preference datasets for MLLMs and examined different alignment methods, including Direct Preference Optimization (DPO) and Proximal Policy Optimization (PPO). However, due to variations in datasets, base model types, and alignment methods, it remains unclear which specific elements contribute most significantly to the reported improvements in these works. In this paper, we independently analyze each aspect of preference alignment in MLLMs. We start by categorizing the alignment algorithms into two groups, offline (such as DPO), and online (such as online-DPO), and show that combining offline and online methods can improve the performance of the model in certain scenarios. We review a variety of published multimodal preference datasets and discuss how the details of their construction impact model performance. Based on these insights, we introduce a novel way of creating multimodal preference data called Bias-Driven Hallucination Sampling (BDHS) that needs neither additional annotation nor external models, and show that it can achieve competitive performance to previously published alignment work for multimodal models across a range of benchmarks."
                    },
                    {
                        "title": "Guiding Instruction-based Image Editing via Multimodal Large Language Models",
                        "abstract": "Instruction-based image editing improves the controllability and flexibility of image manipulation via natural commands without elaborate descriptions or regional masks. However, human instructions are sometimes too brief for current methods to capture and follow. Multimodal large language models (MLLMs) show promising capabilities in cross-modal understanding and visual-aware response generation via LMs. We investigate how MLLMs facilitate edit instructions and present MLLM-Guided Image Editing (MGIE). MGIE learns to derive expressive instructions and provides explicit guidance. The editing model jointly captures this visual imagination and performs manipulation through end-to-end training. We evaluate various aspects of Photoshop-style modification, global photo optimization, and local editing. Extensive experimental results demonstrate that expressive instructions are crucial to instruction-based image editing, and our MGIE can lead to a notable improvement in automatic metrics and human evaluation while maintaining competitive inference efficiency."
                    },
                    {
                        "title": "Ferret: Refer and Ground Anything Anywhere at Any Granularity",
                        "abstract": "We introduce Ferret, a new Multimodal Large Language Model (MLLM) capable of understanding spatial referring of any shape or granularity within an image and accurately grounding open-vocabulary descriptions. To unify referring and grounding in the LLM paradigm, Ferret employs a novel and powerful hybrid region representation that integrates discrete coordinates and continuous features jointly to represent a region in the image. To extract the continuous features of versatile regions, we propose a spatial-aware visual sampler, adept at handling varying sparsity across different shapes. Consequently, Ferret can accept diverse region inputs, such as points, bounding boxes, and free-form shapes. To bolster the desired capability of Ferret, we curate GRIT, a comprehensive refer-and-ground instruction tuning dataset including 1.1M samples that contain rich hierarchical spatial knowledge, with 95K hard negative data to promote model robustness. The resulting model not only achieves superior performance in classical referring and grounding tasks, but also greatly outperforms existing MLLMs in region-based and localization-demanded multimodal chatting. Our evaluations also reveal a significantly improved capability of describing image details and a remarkable alleviation in object hallucination. Code and data will be available at https://github.com/apple/ml-ferret"
                    }
                ]
            }
        }
    },
    "2310.08513": {
        "paper_data": {
            "title": "How connectivity structure shapes rich and lazy learning in neural circuits",
            "url": "http://arxiv.org/abs/2310.08513v2",
            "arxiv_id": "2310.08513",
            "authors": [
                "Yuhan Helena Liu",
                "Aristide Baratin",
                "Jonathan Cornford",
                "Stefan Mihalas",
                "Eric Shea-Brown",
                "Guillaume Lajoie"
            ],
            "abstract": "In theoretical neuroscience, recent work leverages deep learning tools to explore how some network attributes critically influence its learning dynamics. Notably, initial weight distributions with small (resp. large) variance may yield a rich (resp. lazy) regime, where significant (resp. minor) changes to network states and representation are observed over the course of learning. However, in biology, neural circuit connectivity could exhibit a low-rank structure and therefore differs markedly from the random initializations generally used for these studies. As such, here we investigate how the structure of the initial weights -- in particular their effective rank -- influences the network learning regime. Through both empirical and theoretical analyses, we discover that high-rank initializations typically yield smaller network changes indicative of lazier learning, a finding we also confirm with experimentally-driven initial connectivity in recurrent neural networks. Conversely, low-rank initialization biases learning towards richer learning. Importantly, however, as an exception to this rule, we find lazier learning can still occur with a low-rank initialization that aligns with task and data statistics. Our research highlights the pivotal role of initial weight structures in shaping learning regimes, with implications for metabolic costs of plasticity and risks of catastrophic forgetting.",
            "introduction": "   1 Introduction  Structural variations can significantly impact learning dynamics in theoretical neuroscience studies of animals. For instance, studies have revealed that specific neural connectivity patterns can facilitate faster learning of certain tasks (Braun et\u00a0al., 2022; Raman & O\u2019Leary, 2021; Simard et\u00a0al., 2005; Canatar et\u00a0al., 2021; Xie et\u00a0al., 2022; Goudar et\u00a0al., 2023; Chang et\u00a0al., 2023). In deep learning, structure, encompassing architecture and initial connectivity, crucially dictates learning speed and effectiveness (Richards et\u00a0al., 2019; Zador, 2019; Yang & Molano-Maz\u00f3n, 2021; Braun et\u00a0al., 2022).   A key structural aspect is the initial connectivity prior to training. Specifically, the initial connection weight magnitude can significantly bias learning dynamics, pushing them towards either rich or lazy regimes (Chizat et\u00a0al., 2019; Flesch et\u00a0al., 2021). Lazy learning often induces minor changes in the network during the learning process. Such minimal adjustments are advantageous given that plasticity is metabolically costly (Mery & Kawecki, 2005; Pla\u00e7ais & Preat, 2013), and significant changes in representations might lead to issues like catastrophic forgetting (McCloskey & Cohen, 1989; Kirkpatrick et\u00a0al., 2017). On the other hand, the rich learning regime can significantly adapt the network\u2019s internal representations to task statistics, which can be advantageous for task feature acquisition and has implications for generalization (Flesch et\u00a0al., 2021; George et\u00a0al., 2022). Most research on initial weight magnitude\u2019s role in learning dynamics has focused on random Gaussian or Uniform initializations (Woodworth et\u00a0al., 2020; Flesch et\u00a0al., 2021; Braun et\u00a0al., 2022). These patterns stand in contrast to the connectivity structures observed in biological neural circuits, which could exhibit a more pronounced low-rank eigenstructure (Song et\u00a0al., 2005). This divergence prompts a pivotal question: how does the initial weight structure, given a fixed initial weight magnitude, bias the learning regime?   This study examines how initial weight structure, particularly the effective rank, modulates the effective richness or laziness of task learning within the standard training regime. We note that rich and lazy learning regimes have well established meanings in deep learning theory. The latter being defined as a situation where the Neural Tangent Kernel (NTK) stays stationary during training, while the former refers to the case where the NTK changes. In this work, we slightly extend these definitions and introduce effective learning richness/laziness. Unlike the traditional definition, which is based upon initialization, effective learning richness/laziness is defined in terms of post-training adjustment measurements. From this perspective, a learning process is deemed effectively \"lazy\" if the measured NTK movement is small. For example, consider a network whose initialization puts it in standard rich regime, but for a given task, its NTK moves very little during training. We define learning for this specific situation as effectively lazy. In other words, while the standard regime definition informs us (prior to training) whether the network can adapt significantly to task training or not, our \"effective\" definition lies in the post-training effects.    1.1 Contributions  Our main contributions and findings can be summarized as follows:   \u2022  Through theoretical derivation in two-layer feedforward linear network, we demonstrate that higher-rank initialization results in effectively lazier learning on average across tasks (Theorem\u00a01). We note that the emphasis of the theorem is on the expectation across tasks.    \u2022  We validate our theoretical findings",
            "references": [
                {
                    "title": "Exact learning dynamics of deep linear networks with prior knowledge",
                    "abstract": "Learning in deep neural networks is known to depend critically on the knowledge embedded in the initial network weights. However, few theoretical results have precisely linked prior knowledge to learning dynamics. Here we derive exact solutions to the dynamics of learning with rich prior knowledge in deep linear networks by generalising Fukumizu\u2019s matrix Riccati solution (Fukumizu 1998 Gen 1 1E\u201303). We obtain explicit expressions for the evolving network function, hidden representational similarity, and neural tangent kernel over training for a broad class of initialisations and tasks. The expressions reveal a class of task-independent initialisations that radically alter learning dynamics from slow non-linear dynamics to fast exponential trajectories while converging to a global optimum with identical representational similarity, dissociating learning trajectories from the structure of initial internal representations. We characterise how network weights dynamically align with task structure, rigorously justifying why previous solutions successfully described learning from small initial weights without incorporating their fine-scale structure. Finally, we discuss the implications of these findings for continual learning, reversal learning and learning of structured knowledge. Taken together, our results provide a mathematical toolkit for understanding the impact of prior knowledge on deep learning."
                },
                {
                    "title": "Aligned and oblique dynamics in recurrent neural networks",
                    "abstract": "The relation between neural activity and behaviorally relevant variables is at the heart of neuroscience research. When strong, this relation is termed a neural representation. There is increasing evidence, however, for partial dissociations between activity in an area and relevant external variables. While many explanations have been proposed, a theoretical framework for the relationship between external and internal variables is lacking. Here, we utilize recurrent neural networks (RNNs) to explore the question of when and how neural dynamics and the network's output are related from a geometrical point of view. We find that training RNNs can lead to two dynamical regimes: dynamics can either be aligned with the directions that generate output variables, or oblique to them. We show that the choice of readout weight magnitude before training can serve as a control knob between the regimes, similar to recent findings in feedforward networks. These regimes are functionally distinct. Oblique networks are more heterogeneous and suppress noise in their output directions. They are furthermore more robust to perturbations along the output directions. Crucially, the oblique regime is specific to recurrent (but not feedforward) networks, arising from dynamical stability considerations. Finally, we show that tendencies towards the aligned or the oblique regime can be dissociated in neural recordings. Altogether, our results open a new perspective for interpreting neural activity by relating network dynamics and their output."
                },
                {
                    "title": "Synaptic Weight Distributions Depend on the Geometry of Plasticity",
                    "abstract": "A growing literature in computational neuroscience leverages gradient descent and learning algorithms that approximate it to study synaptic plasticity in the brain. However, the vast majority of this work ignores a critical underlying assumption: the choice of distance for synaptic changes - i.e. the geometry of synaptic plasticity. Gradient descent assumes that the distance is Euclidean, but many other distances are possible, and there is no reason that biology necessarily uses Euclidean geometry. Here, using the theoretical tools provided by mirror descent, we show that the distribution of synaptic weights will depend on the geometry of synaptic plasticity. We use these results to show that experimentally-observed log-normal weight distributions found in several brain areas are not consistent with standard gradient descent (i.e. a Euclidean geometry), but rather with non-Euclidean distances. Finally, we show that it should be possible to experimentally test for different synaptic geometries by comparing synaptic weight distributions before and after learning. Overall, our work shows that the current paradigm in theoretical work on synaptic plasticity that assumes Euclidean synaptic geometry may be misguided and that it should be possible to experimentally determine the true geometry of synaptic plasticity in the brain."
                },
                {
                    "title": "De novo motor learning creates structure in neural activity space that shapes adaptation",
                    "abstract": "Animals can quickly adapt learned movements in response to external perturbations. Motor adaptation is likely influenced by an animal\u2019s existing movement repertoire, but the nature of this influence is unclear. Long-term learning causes lasting changes in neural connectivity which determine the activity patterns that can be produced. Here, we sought to understand how a neural population\u2019s activity repertoire, acquired through long-term learning, affects short-term adaptation by modeling motor cortical neural population dynamics during de novo learning and subsequent adaptation using recurrent neural networks. We trained these networks on different motor repertoires comprising varying numbers of movements. Networks with multiple movements had more constrained and robust dynamics, which were associated with more defined neural \u2018structure\u2019\u2014organization created by the neural population activity patterns corresponding to each movement. This structure facilitated adaptation, but only when small changes in motor output were required, and when the structure of the network inputs, the neural activity space, and the perturbation were congruent. These results highlight trade-offs in skill acquisition and demonstrate how prior experience and external cues during learning can shape the geometrical properties of neural population activity as well as subsequent adaptation."
                },
                {
                    "title": "The Forward-Forward Algorithm: Some Preliminary Investigations",
                    "abstract": "The aim of this paper is to introduce a new learning procedure for neural networks and to demonstrate that it works well enough on a few small problems to be worth further investigation. The Forward-Forward algorithm replaces the forward and backward passes of backpropagation by two forward passes, one with positive (i.e. real) data and the other with negative data which could be generated by the network itself. Each layer has its own objective function which is simply to have high goodness for positive data and low goodness for negative data. The sum of the squared activities in a layer can be used as the goodness but there are many other possibilities, including minus the sum of the squared activities. If the positive and negative passes could be separated in time, the negative passes could be done offline, which would make the learning much simpler in the positive pass and allow video to be pipelined through the network without ever storing activities or stopping to propagate derivatives."
                },
                {
                    "title": "Eigenvalue spectral properties of sparse random matrices obeying Dale's law",
                    "abstract": "This paper examines the relationship between sparse random network architectures and neural network stability by examining the eigenvalue spectral distribution. Specifically, we generalise classical eigenspectral results to sparse connectivity matrices obeying Dale's law: neurons function as either excitatory (E) or inhibitory (I). By defining sparsity as the probability that a neutron is connected to another neutron, we give explicit formulae that shows how sparsity interacts with the E/I population statistics to scale key features of the eigenspectrum, in both the balanced and unbalanced cases. Our results show that the eigenspectral outlier is linearly scaled by sparsity, but the eigenspectral radius and density now depend on a nonlinear interaction between sparsity, the E/I population means and variances. Contrary to previous results, we demonstrate that a non-uniform eigenspectral density results if any of the E/I population statistics differ, not just the E/I population variances. We also find that 'local' eigenvalue-outliers are present for sparse random matrices obeying Dale's law, and demonstrate that these eigenvalues can be controlled by a modified zero row-sum constraint for the balanced case, however, they persist in the unbalanced case. We examine all levels of connection (sparsity), and distributed E/I population weights, to describe a general class of sparse connectivity structures which unifies all the previous results as special cases of our framework. Sparsity and Dale's law are both fundamental anatomical properties of biological neural networks. We generalise their combined effects on the eigenspectrum of random neural networks, thereby gaining insight into network stability, state transitions and the structure-function relationship."
                },
                {
                    "title": "Second-order regression models exhibit progressive sharpening to the edge of stability",
                    "abstract": "Recent studies of gradient descent with large step sizes have shown that there is often a regime with an initial increase in the largest eigenvalue of the loss Hessian (progressive sharpening), followed by a stabilization of the eigenvalue near the maximum value which allows convergence (edge of stability). These phenomena are intrinsically non-linear and do not happen for models in the constant Neural Tangent Kernel (NTK) regime, for which the predictive function is approximately linear in the parameters. As such, we consider the next simplest class of predictive models, namely those that are quadratic in the parameters, which we call second-order regression models. For quadratic objectives in two dimensions, we prove that this second-order regression model exhibits progressive sharpening of the NTK eigenvalue towards a value that differs slightly from the edge of stability, which we explicitly compute. In higher dimensions, the model generically shows similar behavior, even without the specific structure of a neural network, suggesting that progressive sharpening and edge-of-stability behavior aren't unique features of neural networks, and could be a more general property of discrete learning algorithms in high-dimensional non-linear models."
                },
                {
                    "title": "The Influence of Learning Rule on Representation Dynamics in Wide Neural Networks",
                    "abstract": "It is unclear how changing the learning rule of a deep neural network alters its learning dynamics and representations. To gain insight into the relationship between learned features, function approximation, and the learning rule, we analyze infinite-width deep networks trained with gradient descent (GD) and biologically-plausible alternatives including feedback alignment (FA), direct feedback alignment (DFA), and error modulated Hebbian learning (Hebb), as well as gated linear networks (GLN). We show that, for each of these learning rules, the evolution of the output function at infinite width is governed by a time varying effective neural tangent kernel (eNTK). In the lazy training limit, this eNTK is static and does not evolve, while in the rich mean-field regime this kernel's evolution can be determined self-consistently with dynamical mean field theory (DMFT). This DMFT enables comparisons of the feature and prediction dynamics induced by each of these learning rules. In the lazy limit, we find that DFA and Hebb can only learn using the last layer features, while full FA can utilize earlier layers with a scale determined by the initial correlation between feedforward and feedback weight matrices. In the rich regime, DFA and FA utilize a temporally evolving and depth-dependent NTK. Counterintuitively, we find that FA networks trained in the rich regime exhibit more feature learning if initialized with smaller correlation between the forward and backward pass weights. GLNs admit a very simple formula for their lazy limit kernel and preserve conditional Gaussianity of their preactivations under gating functions. Error modulated Hebb rules show very small task-relevant alignment of their kernels and perform most task relevant learning in the last layer."
                },
                {
                    "title": "Lazy vs hasty: linearization in deep networks impacts learning schedule based on example difficulty",
                    "abstract": "Among attempts at giving a theoretical account of the success of deep neural networks, a recent line of work has identified a so-called lazy training regime in which the network can be well approximated by its linearization around initialization. Here we investigate the comparative effect of the lazy (linear) and feature learning (non-linear) regimes on subgroups of examples based on their difficulty. Specifically, we show that easier examples are given more weight in feature learning mode, resulting in faster training compared to more difficult ones. In other words, the non-linear dynamics tends to sequentialize the learning of examples of increasing difficulty. We illustrate this phenomenon across different ways to quantify example difficulty, including c-score, label noise, and in the presence of easy-to-learn spurious correlations. Our results reveal a new understanding of how deep networks prioritize resources across example difficulty."
                },
                {
                    "title": "Holomorphic Equilibrium Propagation Computes Exact Gradients Through Finite Size Oscillations",
                    "abstract": "Equilibrium propagation (EP) is an alternative to backpropagation (BP) that allows the training of deep neural networks with local learning rules. It thus provides a compelling framework for training neuromorphic systems and understanding learning in neurobiology. However, EP requires infinitesimal teaching signals, thereby limiting its applicability in noisy physical systems. Moreover, the algorithm requires separate temporal phases and has not been applied to large-scale problems. Here we address these issues by extending EP to holomorphic networks. We show analytically that this extension naturally leads to exact gradients even for finite-amplitude teaching signals. Importantly, the gradient can be computed as the first Fourier coefficient from finite neuronal activity oscillations in continuous time without requiring separate phases. Further, we demonstrate in numerical simulations that our approach permits robust estimation of gradients in the presence of noise and that deeper models benefit from the finite teaching signals. Finally, we establish the first benchmark for EP on the ImageNet 32x32 dataset and show that it matches the performance of an equivalent network trained with BP. Our work provides analytical insights that enable scaling EP to large-scale problems and establishes a formal framework for how oscillations could support learning in biological and neuromorphic systems."
                },
                {
                    "title": "Relating local connectivity and global dynamics in recurrent excitatory-inhibitory networks",
                    "abstract": "How the connectivity of cortical networks determines the neural dynamics and the resulting computations is one of the key questions in neuroscience. Previous works have pursued two complementary strategies to quantify the structure in connectivity, by specifying either the local statistics of connectivity motifs between small groups of neurons, or by defining network-wide low-rank patterns of connectivity that determine the resulting low-dimensional dynamics. A direct relationship between these two approaches is however currently missing, and in particular it remains to be clarified how local connectivity statistics are related to the global connectivity structure and shape the low-dimensional activity. To bridge this gap, here we develop a method for mapping local connectivity statistics onto an approximate global low-rank structure. Our method rests on approximating the global connectivity matrix using dominant eigenvectors, which we compute using perturbation theory for random matrices. This approach demonstrates that multi-population networks defined from local connectivity properties can in general be approximated by low-rank connectivity with Gaussian-mixture statistics. We specifically apply this method to excitatory-inhibitory networks, and show that it leads to accurate predictions for both the low-dimensional dynamics, and for the activity of individual neurons. Altogether, our approach allows us to disentangle the effects of mean connectivity and reciprocal motifs on the global recurrent feedback, and provides an intuitive picture of how local connectivity shapes global network dynamics. Author summary The structure of connections between neurons is believed to determine how cortical networks control behaviour. Current experimental methods typically measure connections between small numbers of simultaneously recorded neurons, and thereby provide information on statistics of local connectivity motifs. Collective network dynamics are however determined by network-wide patterns of connections. How these global patterns are related to local connectivity statistics and shape the dynamics is an open question that we address in this study. Starting from networks defined in terms of local statistics, we develop a method for approximating the resulting connectivity by global low-rank patterns. We apply this method to classical excitatory-inhibitory networks and show that it allows us to predict both collective and single-neuron activity. More generally, our approach provides a link between local connectivity statistics and global network dynamics."
                },
                {
                    "title": "Task-dependent optimal representations for cerebellar learning",
                    "abstract": "The cerebellar granule cell layer has inspired numerous theoretical models of neural representations that support learned behaviors, beginning with the work of Marr and Albus. In these models, granule cells form a sparse, combinatorial encoding of diverse sensorimotor inputs. Such sparse representations are optimal for learning to discriminate random stimuli. However, recent observations of dense, low-dimensional activity across granule cells have called into question the role of sparse coding in these neurons. Here, we generalize theories of cerebellar learning to determine the optimal granule cell representation for tasks beyond random stimulus discrimination, including continuous input-output transformations as required for smooth motor control. We show that for such tasks, the optimal granule cell representation is substantially denser than predicted by classic theories. Our results provide a general theory of learning in cerebellum-like systems and suggest that optimal cerebellar representations are task-dependent."
                },
                {
                    "title": "Winning the Lottery With Neural Connectivity Constraints: Faster Learning Across Cognitive Tasks With Spatially Constrained Sparse RNNs",
                    "abstract": "Abstract Recurrent neural networks (RNNs) are often used to model circuits in the brain and can solve a variety of difficult computational problems requiring memory, error correction, or selection (Hopfield, 1982; Maass et al., 2002; Maass, 2011). However, fully connected RNNs contrast structurally with their biological counterparts, which are extremely sparse (about 0.1%). Motivated by the neocortex, where neural connectivity is constrained by physical distance along cortical sheets and other synaptic wiring costs, we introduce locality masked RNNs (LM-RNNs) that use task-agnostic predetermined graphs with sparsity as low as 4%. We study LM-RNNs in a multitask learning setting relevant to cognitive systems neuroscience with a commonly used set of tasks, 20-Cog-tasks (Yang et al., 2019). We show through reductio ad absurdum that 20-Cog-tasks can be solved by a small pool of separated autapses that we can mechanistically analyze and understand. Thus, these tasks fall short of the goal of inducing complex recurrent dynamics and modular structure in RNNs. We next contribute a new cognitive multitask battery, Mod-Cog, consisting of up to 132 tasks that expands by about seven-fold the number of tasks and task complexity of 20-Cog-tasks. Importantly, while autapses can solve the simple 20-Cog-tasks, the expanded task set requires richer neural architectures and continuous attractor dynamics. On these tasks, we show that LM-RNNs with an optimal sparsity result in faster training and better data efficiency than fully connected networks."
                },
                {
                    "title": "The least-control principle for local learning at equilibrium",
                    "abstract": "Equilibrium systems are a powerful way to express neural computations. As special cases, they include models of great current interest in both neuroscience and machine learning, such as deep neural networks, equilibrium recurrent neural networks, deep equilibrium models, or meta-learning. Here, we present a new principle for learning such systems with a temporally- and spatially-local rule. Our principle casts learning as a least-control problem, where we first introduce an optimal controller to lead the system towards a solution state, and then define learning as reducing the amount of control needed to reach such a state. We show that incorporating learning signals within a dynamics as an optimal control enables transmitting activity-dependent credit assignment information, avoids storing intermediate states in memory, and does not rely on infinitesimal learning signals. In practice, our principle leads to strong performance matching that of leading gradient-based learning methods when applied to an array of problems involving recurrent neural networks and meta-learning. Our results shed light on how the brain might learn and offer new ways of approaching a broad class of machine learning problems."
                },
                {
                    "title": "Single-phase deep learning in cortico-cortical networks",
                    "abstract": "The error-backpropagation (backprop) algorithm remains the most common solution to the credit assignment problem in artificial neural networks. In neuroscience, it is unclear whether the brain could adopt a similar strategy to correctly modify its synapses. Recent models have attempted to bridge this gap while being consistent with a range of experimental observations. However, these models are either unable to effectively backpropagate error signals across multiple layers or require a multi-phase learning process, neither of which are reminiscent of learning in the brain. Here, we introduce a new model, Bursting Cortico-Cortical Networks (BurstCCN), which solves these issues by integrating known properties of cortical networks namely bursting activity, short-term plasticity (STP) and dendrite-targeting interneurons. BurstCCN relies on burst multiplexing via connection-type-specific STP to propagate backprop-like error signals within deep cortical networks. These error signals are encoded at distal dendrites and induce burst-dependent plasticity as a result of excitatory-inhibitory top-down inputs. First, we demonstrate that our model can effectively backpropagate errors through multiple layers using a single-phase learning process. Next, we show both empirically and analytically that learning in our model approximates backprop-derived gradients. Finally, we demonstrate that our model is capable of learning complex image classification tasks (MNIST and CIFAR-10). Overall, our results suggest that cortical features across sub-cellular, cellular, microcircuit and systems levels jointly underlie single-phase efficient deep learning in the brain."
                },
                {
                    "title": "Beyond accuracy: generalization properties of bio-plausible temporal credit assignment rules",
                    "abstract": "To unveil how the brain learns, ongoing work seeks biologically-plausible approximations of gradient descent algorithms for training recurrent neural networks (RNNs). Yet, beyond task accuracy, it is unclear if such learning rules converge to solutions that exhibit different levels of generalization than their nonbiologically-plausible counterparts. Leveraging results from deep learning theory based on loss landscape curvature, we ask: how do biologically-plausible gradient approximations affect generalization? We first demonstrate that state-of-the-art biologically-plausible learning rules for training RNNs exhibit worse and more variable generalization performance compared to their machine learning counterparts that follow the true gradient more closely. Next, we verify that such generalization performance is correlated significantly with loss landscape curvature, and we show that biologically-plausible learning rules tend to approach high-curvature regions in synaptic weight space. Using tools from dynamical systems, we derive theoretical arguments and present a theorem explaining this phenomenon. This predicts our numerical results, and explains why biologically-plausible rules lead to worse and more variable generalization properties. Finally, we suggest potential remedies that could be used by the brain to mitigate this effect. To our knowledge, our analysis is the first to identify the reason for this generalization gap between artificial and biologically-plausible learning rules, which can help guide future investigations into how the brain learns solutions that generalize."
                },
                {
                    "title": "Biologically-plausible backpropagation through arbitrary timespans via local neuromodulators",
                    "abstract": "The spectacular successes of recurrent neural network models where key parameters are adjusted via backpropagation-based gradient descent have inspired much thought as to how biological neuronal networks might solve the corresponding synaptic credit assignment problem. There is so far little agreement, however, as to how biological networks could implement the necessary backpropagation through time, given widely recognized constraints of biological synaptic network signaling architectures. Here, we propose that extra-synaptic diffusion of local neuromodulators such as neuropeptides may afford an effective mode of backpropagation lying within the bounds of biological plausibility. Going beyond existing temporal truncation-based gradient approximations, our approximate gradient-based update rule, ModProp, propagates credit information through arbitrary time steps. ModProp suggests that modulatory signals can act on receiving cells by convolving their eligibility traces via causal, time-invariant and synapse-type-specific filter taps. Our mathematical analysis of ModProp learning, together with simulation results on benchmark temporal tasks, demonstrate the advantage of ModProp over existing biologically-plausible temporal credit assignment rules. These results suggest a potential neuronal mechanism for signaling credit information related to recurrent interactions over a longer time horizon. Finally, we derive an in-silico implementation of ModProp that could serve as a low-complexity and causal alternative to backpropagation through time."
                },
                {
                    "title": "Maslow's Hammer for Catastrophic Forgetting: Node Re-Use vs Node Activation",
                    "abstract": "Continual learning - learning new tasks in sequence while maintaining performance on old tasks - remains particularly challenging for artificial neural networks. Surprisingly, the amount of forgetting does not increase with the dissimilarity between the learned tasks, but appears to be worst in an intermediate similarity regime. In this paper we theoretically analyse both a synthetic teacher-student framework and a real data setup to provide an explanation of this phenomenon that we name Maslow's hammer hypothesis. Our analysis reveals the presence of a trade-off between node activation and node re-use that results in worst forgetting in the intermediate regime. Using this understanding we reinterpret popular algorithmic interventions for catastrophic interference in terms of this trade-off, and identify the regimes in which they are most effective."
                },
                {
                    "title": "NeuroGym: An open resource for developing and sharing neuroscience tasks",
                    "abstract": "Artificial Neural Networks (ANNs) trained on specific cognitive tasks have re-emerged as a useful tool to study the brain. However, ANNs would better aid cognitive neuroscience if a given network could be easily trained on a wide range of tasks for which neural recordings are available. Moreover, unintentional divergent implementations of cognitive tasks can produce variable results, which limits their interpretability. Towards this goal, we present NeuroGym, an open-source Python package that provides a large collection of customizable neuroscience tasks to test and compare network models. Building upon the OpenAI Gym toolbox, NeuroGym tasks (1) are written in a high-level flexible Python framework; (2) possess a shared interface tailored to common needs of neuroscience tasks that facilitates their design and usage; (3) support the training of ANNs using both Reinforcement and Supervised Learning techniques. The toolbox allows easy assembly of new tasks by modifying existing ones in a hierarchical and modular fashion. These design features make it straightforward to take a network designed for one task and train it on many other tasks. NeuroGym is a community-driven effort that contributes to a rapidly evolving open ecosystem of neural network development, data analysis, and model-data comparison."
                },
                {
                    "title": "Nonlinear Initialization Methods for Low-Rank Neural Networks",
                    "abstract": "We propose a novel low-rank initialization framework for training low-rank deep neural networks -- networks where the weight parameters are re-parameterized by products of two low-rank matrices. The most successful prior existing approach, spectral initialization, draws a sample from the initialization distribution for the full-rank setting and then optimally approximates the full-rank initialization parameters in the Frobenius norm with a pair of low-rank initialization matrices via singular value decomposition. Our method is inspired by the insight that approximating the function corresponding to each layer is more important than approximating the parameter values. We provably demonstrate that there is a significant gap between these two approaches for ReLU networks, particularly as the desired rank of the approximating weights decreases, or as the dimension of the inputs to the layer increases (the latter point holds when the network width is super-linear in dimension). Along the way, we provide the first provably efficient algorithm for solving the ReLU low-rank approximation problem for fixed parameter rank $r$ -- previously, it was unknown that the problem was computationally tractable to solve even for rank $1$. We also provide a practical algorithm to solve this problem which is no more expensive than the existing spectral initialization approach, and validate our theory by training ResNet and EfficientNet models (He et al., 2016; Tan&Le, 2019) on ImageNet (Russakovsky et al., 2015)."
                },
                {
                    "title": "Neural Tangent Kernel Beyond the Infinite-Width Limit: Effects of Depth and Initialization",
                    "abstract": "Neural Tangent Kernel (NTK) is widely used to analyze overparametrized neural networks due to the famous result by Jacot et al. (2018): in the infinite-width limit, the NTK is deterministic and constant during training. However, this result cannot explain the behavior of deep networks, since it generally does not hold if depth and width tend to infinity simultaneously. In this paper, we study the NTK of fully-connected ReLU networks with depth comparable to width. We prove that the NTK properties depend significantly on the depth-to-width ratio and the distribution of parameters at initialization. In fact, our results indicate the importance of the three phases in the hyperparameter space identified in Poole et al. (2016): ordered, chaotic and the edge of chaos (EOC). We derive exact expressions for the NTK dispersion in the infinite-depth-and-width limit in all three phases and conclude that the NTK variability grows exponentially with depth at the EOC and in the chaotic phase but not in the ordered phase. We also show that the NTK of deep networks may stay constant during training only in the ordered phase and discuss how the structure of the NTK matrix changes during training."
                },
                {
                    "title": "Heterogeneity in Neuronal Dynamics is Learned by Gradient Descent for Temporal Processing Tasks",
                    "abstract": "Individual neurons in the brain have complex intrinsic dynamics that are highly diverse. We hypothesize that the complex dynamics produced by networks of complex and heterogeneous neurons may contribute to the brain\u2019s ability to process and respond to temporally complex data. To study the role of complex and heterogeneous neuronal dynamics in network computation, we develop a rate-based neuronal model, the generalized-leaky-integrate-and-firing-rate (GLIFR) model, which is a rate-equivalent of the generalized-leaky-integrate-and-fire model. The GLIFR model has multiple dynamical mechanisms which add to the complexity of its activity while maintaining differentiability. We focus on the role of after-spike currents, currents induced or modulated by neuronal spikes, in producing rich temporal dynamics. We use machine learning techniques to learn both synaptic weights and parameters underlying intrinsic dynamics to solve temporal tasks. The GLIFR model allows us to use standard gradient descent techniques rather than surrogate gradient descent, which has been utilized in spiking neural networks. After establishing the ability to optimize parameters using gradient descent in single neurons, we ask how networks of GLIFR neurons learn and perform on temporally challenging tasks, such as sinusoidal pattern generation and sequential MNIST. We find that these networks learn a diversity of parameters, which gives rise to diversity in neuronal dynamics. We also observe that training networks on the sequential MNIST task leads to formation of cell classes based on the clustering of neuronal parameters. GLIFR networks have mixed performance when compared to vanilla recurrent neural networks but appear to be more robust to random silencing. When we explore these performance gains further, we find that both the ability to learn heterogeneity and the presence of after-spike currents contribute. Our work both demonstrates the computational robustness of neuronal complexity and diversity in networks and demonstrates a feasible method of training such models using exact gradients."
                },
                {
                    "title": "Cell-type\u2013specific neuromodulation guides synaptic credit assignment in a spiking neural network",
                    "abstract": "Significance Synaptic connectivity provides the foundation for our present understanding of neuronal network function, but static connectivity cannot explain learning and memory. We propose a computational role for the diversity of cortical neuronal types and their associated cell-type\u2013specific neuromodulators in improving the efficiency of synaptic weight adjustments for task learning in neuronal networks. Brains learn tasks via experience-driven differential adjustment of their myriad individual synaptic connections, but the mechanisms that target appropriate adjustment to particular connections remain deeply enigmatic. While Hebbian synaptic plasticity, synaptic eligibility traces, and top-down feedback signals surely contribute to solving this synaptic credit-assignment problem, alone, they appear to be insufficient. Inspired by new genetic perspectives on neuronal signaling architectures, here, we present a normative theory for synaptic learning, where we predict that neurons communicate their contribution to the learning outcome to nearby neurons via cell-type\u2013specific local neuromodulation. Computational tests suggest that neuron-type diversity and neuron-type\u2013specific local neuromodulation may be critical pieces of the biological credit-assignment puzzle. They also suggest algorithms for improved artificial neural network learning efficiency."
                },
                {
                    "title": "Neural Networks as Kernel Learners: The Silent Alignment Effect",
                    "abstract": "Neural networks in the lazy training regime converge to kernel machines. Can neural networks in the rich feature learning regime learn a kernel machine with a data-dependent kernel? We demonstrate that this can indeed happen due to a phenomenon we term silent alignment, which requires that the tangent kernel of a network evolves in eigenstructure while small and before the loss appreciably decreases, and grows only in overall scale afterwards. We show that such an effect takes place in homogenous neural networks with small initialization and whitened data. We provide an analytical treatment of this effect in the linear network case. In general, we find that the kernel develops a low-rank contribution in the early phase of training, and then evolves in overall scale, yielding a function equivalent to a kernel regression solution with the final network's tangent kernel. The early spectral learning of the kernel depends on the depth. We also demonstrate that non-whitened data can weaken the silent alignment effect."
                },
                {
                    "title": "Functional connectomics spanning multiple areas of mouse visual cortex",
                    "abstract": "To understand the brain we must relate neurons\u2019 functional responses to the circuit architecture that shapes them. Here, we present a large functional connectomics dataset with dense calcium imaging of a millimeter scale volume. We recorded activity from approximately 75,000 neurons in primary visual cortex (VISp) and three higher visual areas (VISrl, VISal and VISlm) in an awake mouse viewing natural movies and synthetic stimuli. The functional data were co-registered with a volumetric electron microscopy (EM) reconstruction containing more than 200,000 cells and 0.5 billion synapses. Subsequent proofreading of a subset of neurons in this volume yielded reconstructions that include complete dendritic trees as well the local and inter-areal axonal projections that map up to thousands of cell-to-cell connections per neuron. Here, we release this dataset as an open-access resource to the scientific community including a set of tools that facilitate data retrieval and downstream analysis. In accompanying papers we describe our findings using the dataset to provide a comprehensive structural characterization of cortical cell types1\u20133 and the most detailed synaptic level connectivity diagram of a cortical column to date2, uncovering unique cell-type specific inhibitory motifs that can be linked to gene expression data4. Functionally, we identify new computational principles of how information is integrated across visual space5, characterize novel types of neuronal invariances6 and bring structure and function together to decipher a general principle that wires excitatory neurons within and across areas7, 8."
                },
                {
                    "title": "The functional specialization of visual cortex emerges from training parallel pathways with self-supervised predictive learning",
                    "abstract": "The visual system of mammals is comprised of parallel, hierarchical specialized pathways. Different pathways are specialized in so far as they use representations that are more suitable for supporting specific downstream behaviours. In particular, the clearest example is the specialization of the ventral (\u201cwhat\u201d) and dorsal (\u201cwhere\u201d) pathways of the visual cortex. These two pathways support behaviours related to visual recognition and movement, respectively. To-date, deep neural networks have mostly been used as models of the ventral, recognition pathway. However, it is unknown whether both pathways can be modelled with a single deep ANN. Here, we ask whether a single model with a single loss function can capture the properties of both the ventral and the dorsal pathways. We explore this question using data from mice, who like other mammals, have specialized pathways that appear to support recognition and movement behaviours. We show that when we train a deep neural network architecture with two parallel pathways using a self-supervised predictive loss function, we can outperform other models in fitting mouse visual cortex. Moreover, we can model both the dorsal and ventral pathways. These results demonstrate that a self-supervised predictive learning approach applied to parallel pathway architectures can account for some of the functional specialization seen in mammalian visual systems."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "Rich and lazy learning of task representations in brains and neural networks",
                    "abstract": "How do neural populations code for multiple, potentially conflicting tasks? Here, we used computational simulations involving neural networks to define \u201clazy\u201d and \u201crich\u201d coding solutions to this multitasking problem, which trade off learning speed for robustness. During lazy learning the input dimensionality is expanded by random projections to the network hidden layer, whereas in rich learning hidden units acquire structured representations that privilege relevant over irrelevant features. For context-dependent decision-making, one rich solution is to project task representations onto low-dimensional and orthogonal manifolds. Using behavioural testing and neuroimaging in humans, and analysis of neural signals from macaque prefrontal cortex, we report evidence for neural coding patterns in biological brains whose dimensionality and neural geometry are consistent with the rich learning regime."
                },
                {
                    "title": "Local Connectivity and Synaptic Dynamics in Mouse and Human Neocortex",
                    "abstract": "To elucidate cortical microcircuit structure and synaptic properties we present a unique, extensive, and public synaptic physiology dataset and analysis platform. Through its application, we reveal principles that relate cell type to synapse properties and intralaminar circuit organization in the mouse and human cortex. The dynamics of excitatory synapses align with the postsynaptic cell subclass, whereas inhibitory synapse dynamics partly align with presynaptic cell subclass but with considerable overlap. Despite these associations, synaptic properties are heterogeneous in most subclass to subclass connections. The two main axes of heterogeneity are strength and variability. Cell subclasses divide along the variability axis, while the strength axis accounts for significant heterogeneity within the subclass. In human cortex, excitatory to excitatory synapse dynamics are distinct from those in mouse and short-term plasticity varies with depth across layers 2 and 3. With a novel connectivity analysis that enables fair comparisons between circuit elements, we find that intralaminar connection probability among cell subclasses exhibits a strong layer dependence.These and other findings combined with the analysis platform create new opportunities for the neuroscience community to advance our understanding of cortical microcircuits."
                },
                {
                    "title": "Deep learning: a statistical viewpoint",
                    "abstract": "The remarkable practical success of deep learning has revealed some major surprises from a theoretical perspective. In particular, simple gradient methods easily find near-optimal solutions to non-convex optimization problems, and despite giving a near-perfect fit to training data without any explicit effort to control model complexity, these methods exhibit excellent predictive accuracy. We conjecture that specific principles underlie these phenomena: that overparametrization allows gradient methods to find interpolating solutions, that these methods implicitly impose regularization, and that overparametrization leads to benign overfitting, that is, accurate predictions despite overfitting training data. In this article, we survey recent progress in statistical learning theory that provides examples illustrating these principles in simpler settings. We first review classical uniform convergence results and why they fall short of explaining aspects of the behaviour of deep learning methods. We give examples of implicit regularization in simple settings, where gradient methods lead to minimal norm functions that perfectly fit the training data. Then we review prediction methods that exhibit benign overfitting, focusing on regression problems with quadratic loss. For these methods, we can decompose the prediction rule into a simple component that is useful for prediction and a spiky component that is useful for overfitting but, in a favourable setting, does not harm prediction accuracy. We focus specifically on the linear regime for neural networks, where the network can be approximated by a linear model. In this regime, we demonstrate the success of gradient flow, and we consider benign overfitting with two-layer networks, giving an exact asymptotic analysis that precisely demonstrates the impact of overparametrization. We conclude by highlighting the key challenges that arise in extending these insights to realistic deep learning settings."
                },
                {
                    "title": "On the Implicit Bias of Initialization Shape: Beyond Infinitesimal Mirror Descent",
                    "abstract": "Recent work has highlighted the role of initialization scale in determining the structure of the solutions that gradient methods converge to. In particular, it was shown that large initialization leads to the neural tangent kernel regime solution, whereas small initialization leads to so called\"rich regimes\". However, the initialization structure is richer than the overall scale alone and involves relative magnitudes of different weights and layers in the network. Here we show that these relative scales, which we refer to as initialization shape, play an important role in determining the learned model. We develop a novel technique for deriving the inductive bias of gradient-flow and use it to obtain closed-form implicit regularizers for multiple cases of interest."
                },
                {
                    "title": "Implicit Bias of Linear RNNs",
                    "abstract": "Contemporary wisdom based on empirical studies suggests that standard recurrent neural networks (RNNs) do not perform well on tasks requiring long-term memory. However, precise reasoning for this behavior is still unknown. This paper provides a rigorous explanation of this property in the special case of linear RNNs. Although this work is limited to linear RNNs, even these systems have traditionally been difficult to analyze due to their non-linear parameterization. Using recently-developed kernel regime analysis, our main result shows that linear RNNs learned from random initializations are functionally equivalent to a certain weighted 1D-convolutional network. Importantly, the weightings in the equivalent model cause an implicit bias to elements with smaller time lags in the convolution and hence, shorter memory. The degree of this bias depends on the variance of the transition kernel matrix at initialization and is related to the classic exploding and vanishing gradients problem. The theory is validated in both synthetic and real data experiments."
                },
                {
                    "title": "Gradient Starvation: A Learning Proclivity in Neural Networks",
                    "abstract": "We identify and formalize a fundamental gradient descent phenomenon resulting in a learning proclivity in over-parameterized neural networks. Gradient Starvation arises when cross-entropy loss is minimized by capturing only a subset of features relevant for the task, despite the presence of other predictive features that fail to be discovered. This work provides a theoretical explanation for the emergence of such feature imbalance in neural networks. Using tools from Dynamical Systems theory, we identify simple properties of learning dynamics during gradient descent that lead to this imbalance, and prove that such a situation can be expected given certain statistical structure in training data. Based on our proposed formalism, we develop guarantees for a novel regularization method aimed at decoupling feature learning dynamics, improving accuracy and robustness in cases hindered by gradient starvation. We illustrate our findings with simple and real-world out-of-distribution (OOD) generalization experiments."
                },
                {
                    "title": "A Theoretical Analysis of Catastrophic Forgetting through the NTK Overlap Matrix",
                    "abstract": "Continual learning (CL) is a setting in which an agent has to learn from an incoming stream of data during its entire lifetime. Although major advances have been made in the field, one recurring problem which remains unsolved is that of Catastrophic Forgetting (CF). While the issue has been extensively studied empirically, little attention has been paid from a theoretical angle. In this paper, we show that the impact of CF increases as two tasks increasingly align. We introduce a measure of task similarity called the NTK overlap matrix which is at the core of CF. We analyze common projected gradient algorithms and demonstrate how they mitigate forgetting. Then, we propose a variant of Orthogonal Gradient Descent (OGD) which leverages structure of the data through Principal Component Analysis (PCA). Experiments support our theoretical findings and show how our method reduces CF on classical CL datasets."
                },
                {
                    "title": "PsychRNN: An Accessible and Flexible Python Package for Training Recurrent Neural Network Models on Cognitive Tasks",
                    "abstract": "Visual Abstract Example workflow for using PsychRNN. First, the task of interest is defined, and a recurrent neural network (RNN) model is trained to perform the task, optionally with neurobiologically informed constraints on the network. After the network is trained, the researchers can investigate network properties including the synaptic connectivity patterns and the dynamics of neural population activity during task execution, and other studies, e.g., those on perturbations, can be explored. The dotted line shows the possible repetition of this cycle with one network, which allows investigation of training effects of task shaping, or curriculum learning, for closed-loop training of the network on a progression of tasks. Task-trained artificial recurrent neural networks (RNNs) provide a computational modeling framework of increasing interest and application in computational, systems, and cognitive neuroscience. RNNs can be trained, using deep-learning methods, to perform cognitive tasks used in animal and human experiments and can be studied to investigate potential neural representations and circuit mechanisms underlying cognitive computations and behavior. Widespread application of these approaches within neuroscience has been limited by technical barriers in use of deep-learning software packages to train network models. Here, we introduce PsychRNN, an accessible, flexible, and extensible Python package for training RNNs on cognitive tasks. Our package is designed for accessibility, for researchers to define tasks and train RNN models using only Python and NumPy, without requiring knowledge of deep-learning software. The training backend is based on TensorFlow and is readily extensible for researchers with TensorFlow knowledge to develop projects with additional customization. PsychRNN implements a number of specialized features to support applications in systems and cognitive neuroscience. Users can impose neurobiologically relevant constraints on synaptic connectivity patterns. Furthermore, specification of cognitive tasks has a modular structure, which facilitates parametric variation of task demands to examine their impact on model solutions. PsychRNN also enables task shaping during training, or curriculum learning, in which tasks are adjusted in closed-loop based on performance. Shaping is ubiquitous in training of animals in cognitive tasks, and PsychRNN allows investigation of how shaping trajectories impact learning and model solutions. Overall, the PsychRNN framework facilitates application of trained RNNs in neuroscience research."
                },
                {
                    "title": "Prevalence of neural collapse during the terminal phase of deep learning training",
                    "abstract": "Significance Modern deep neural networks for image classification have achieved superhuman performance. Yet, the complex details of trained networks have forced most practitioners and researchers to regard them as black boxes with little that could be understood. This paper considers in detail a now-standard training methodology: driving the cross-entropy loss to zero, continuing long after the classification error is already zero. Applying this methodology to an authoritative collection of standard deepnets and datasets, we observe the emergence of a simple and highly symmetric geometry of the deepnet features and of the deepnet classifier, and we document important benefits that the geometry conveys\u2014thereby helping us understand an important component of the modern deep learning training paradigm. Modern practice for training classification deepnets involves a terminal phase of training (TPT), which begins at the epoch where training error first vanishes. During TPT, the training error stays effectively zero, while training loss is pushed toward zero. Direct measurements of TPT, for three prototypical deepnet architectures and across seven canonical classification datasets, expose a pervasive inductive bias we call neural collapse (NC), involving four deeply interconnected phenomena. (NC1) Cross-example within-class variability of last-layer training activations collapses to zero, as the individual activations themselves collapse to their class means. (NC2) The class means collapse to the vertices of a simplex equiangular tight frame (ETF). (NC3) Up to rescaling, the last-layer classifiers collapse to the class means or in other words, to the simplex ETF (i.e., to a self-dual configuration). (NC4) For a given activation, the classifier\u2019s decision collapses to simply choosing whichever class has the closest train class mean (i.e., the nearest class center [NCC] decision rule). The symmetric and very simple geometry induced by the TPT confers important benefits, including better generalization performance, better robustness, and better interpretability."
                },
                {
                    "title": "Implicit Regularization via Neural Feature Alignment",
                    "abstract": "We approach the problem of implicit regularization in deep learning from a geometrical viewpoint. We highlight a regularization effect induced by a dynamical alignment of the neural tangent features introduced by Jacot et al, along a small number of task-relevant directions. This can be interpreted as a combined mechanism of feature selection and model compression. By extrapolating a new analysis of Rademacher complexity bounds for linear models, we motivate and study a heuristic complexity measure that captures this phenomenon, in terms of sequences of tangent kernel classes along the optimization paths."
                },
                {
                    "title": "Geometric compression of invariant manifolds in neural networks",
                    "abstract": "We study how neural networks compress uninformative input space in models where data lie in d dimensions, but the labels of which only vary within a linear manifold of dimension d \u2225 < d. We show that for a one-hidden-layer network initialized with infinitesimal weights (i.e. in the feature learning regime) trained with gradient descent, the first layer of weights evolves to become nearly insensitive to the d \u22a5 = d \u2212 d \u2225 uninformative directions. These are effectively compressed by a factor \u03bb\u223cp , where p is the size of the training set. We quantify the benefit of such a compression on the test error \u03f5. For large initialization of the weights (the lazy training regime), no compression occurs and for regular boundaries separating labels we find that \u03f5 \u223c p \u2212\u03b2 , with \u03b2 Lazy = d/(3d \u2212 2). Compression improves the learning curves so that \u03b2 Feature = (2d \u2212 1)/(3d \u2212 2) if d \u2225 = 1 and \u03b2 Feature = (d + d \u22a5/2)/(3d \u2212 2) if d \u2225 > 1. We test these predictions for a stripe model where boundaries are parallel interfaces (d \u2225 = 1) as well as for a cylindrical boundary (d \u2225 = 2). Next, we show that compression shapes the neural tangent kernel (NTK) evolution in time, so that its top eigenvectors become more informative and display a larger projection on the labels. Consequently, kernel learning with the frozen NTK at the end of training outperforms the initial NTK. We confirm these predictions both for a one-hidden-layer fully connected network trained on the stripe model and for a 16-layer convolutional neural network trained on the Modified National Institute of Standards and Technology database (MNIST), for which we also find \u03b2 Feature > \u03b2 Lazy. The great similarities found in these two cases support the idea that compression is central to the training of MNIST, and puts forward kernel principal component analysis on the evolving NTK as a useful diagnostic of compression in deep networks."
                },
                {
                    "title": "Tensor Programs II: Neural Tangent Kernel for Any Architecture",
                    "abstract": "We prove that a randomly initialized neural network of *any architecture* has its Tangent Kernel (NTK) converge to a deterministic limit, as the network widths tend to infinity. We demonstrate how to calculate this limit. In prior literature, the heuristic study of neural network gradients often assumes every weight matrix used in forward propagation is independent from its transpose used in backpropagation (Schoenholz et al. 2017). This is known as the *gradient independence assumption (GIA)*. We identify a commonly satisfied condition, which we call *Simple GIA Check*, such that the NTK limit calculation based on GIA is correct. Conversely, when Simple GIA Check fails, we show GIA can result in wrong answers. Our material here presents the NTK results of Yang (2019a) in a friendly manner and showcases the *tensor programs* technique for understanding wide neural networks. We provide reference implementations of infinite-width NTKs of recurrent neural network, transformer, and batch normalization at this https URL."
                },
                {
                    "title": "When do neural networks outperform kernel methods?",
                    "abstract": "For a certain scaling of the initialization of stochastic gradient descent (SGD), wide neural networks (NN) have been shown to be well approximated by reproducing kernel Hilbert space (RKHS) methods. Recent empirical work showed that, for some classification tasks, RKHS methods can replace NNs without a large loss in performance. On the other hand, two-layers NNs are known to encode richer smoothness classes than RKHS and we know of special examples for which SGD-trained NN provably outperform RKHS. This is true even in the wide network limit, for a different scaling of the initialization. How can we reconcile the above claims? For which tasks do NNs outperform RKHS? If covariates are nearly isotropic, RKHS methods suffer from the curse of dimensionality, while NNs can overcome it by learning the best low-dimensional representation. Here we show that this curse of dimensionality becomes milder if the covariates display the same low-dimensional structure as the target function, and we precisely characterize this tradeoff. Building on these results, we present the spiked covariates model that can capture in a unified framework both behaviors observed in earlier work. We hypothesize that such a latent low-dimensional structure is present in image classification. We test numerically this hypothesis by showing that specific perturbations of the training distribution degrade the performances of RKHS methods much more significantly than NNs."
                },
                {
                    "title": "The interplay between randomness and structure during learning in RNNs",
                    "abstract": "Recurrent neural networks (RNNs) trained on low-dimensional tasks have been widely used to model functional biological networks. However, the solutions found by learning and the effect of initial connectivity are not well understood. Here, we examine RNNs trained using gradient descent on different tasks inspired by the neuroscience literature. We find that the changes in recurrent connectivity can be described by low-rank matrices, despite the unconstrained nature of the learning algorithm. To identify the origin of the low-rank structure, we turn to an analytically tractable setting: training a linear RNN on a simplified task. We show how the low-dimensional task structure leads to low-rank changes to connectivity. This low-rank structure allows us to explain and quantify the phenomenon of accelerated learning in the presence of random initial connectivity. Altogether, our study opens a new perspective to understanding trained RNNs in terms of both the learning process and the resulting network structure."
                },
                {
                    "title": "The Recurrent Neural Tangent Kernel",
                    "abstract": "The study of deep networks (DNs) in the infinite-width limit, via the so-called Neural Tangent Kernel (NTK) approach, has provided new insights into the dynamics of learning, generalization, and the impact of initialization. One key DN architecture remains to be kernelized, namely, the Recurrent Neural Network (RNN). In this paper we introduce and study the Recurrent Neural Tangent Kernel (RNTK), which sheds new insights into the behavior of overparametrized RNNs, including how different time steps are weighted by the RNTK to form the output under different initialization parameters and nonlinearity choices, and how inputs of different lengths are treated. We demonstrate via a number of experiments that the RNTK offers significant performance gains over other kernels, including standard NTKs across a range of different data sets. A unique benefit of the RNTK is that it is agnostic to the length of the input, in stark contrast to other kernels."
                },
                {
                    "title": "Shape Matters: Understanding the Implicit Bias of the Noise Covariance",
                    "abstract": "The noise in stochastic gradient descent (SGD) provides a crucial implicit regularization effect for training overparameterized models. Prior theoretical work largely focuses on spherical Gaussian noise, whereas empirical studies demonstrate the phenomenon that parameter-dependent noise -- induced by mini-batches or label perturbation -- is far more effective than Gaussian noise. This paper theoretically characterizes this phenomenon on a quadratically-parameterized model introduced by Vaskevicius et el. and Woodworth et el. We show that in an over-parameterized setting, SGD with label noise recovers the sparse ground-truth with an arbitrary initialization, whereas SGD with Gaussian noise or gradient descent overfits to dense solutions with large norms. Our analysis reveals that parameter-dependent noise introduces a bias towards local minima with smaller noise variance, whereas spherical Gaussian noise does not. Code for our project is publicly available."
                },
                {
                    "title": "Spike frequency adaptation supports network computations on temporally dispersed information",
                    "abstract": "For solving tasks such as recognizing a song, answering a question, or inverting a sequence of symbols, cortical microcircuits need to integrate and manipulate information that was dispersed over time during the preceding seconds. Creating biologically realistic models for the underlying computations, especially with spiking neurons and for behaviorally relevant integration time spans, is notoriously difficult. We examine the role of spike frequency adaptation in such computations and find that it has a surprisingly large impact. The inclusion of this well known property of a substantial fraction of neurons in the neocortex \u2014 especially in higher areas of the human neocortex \u2014 moves the performance of spiking neural network models for computations on network inputs that are temporally dispersed from a fairly low level up to the performance level of the human brain."
                },
                {
                    "title": "A connectome and analysis of the adult Drosophila central brain",
                    "abstract": "The neural circuits responsible for animal behavior remain largely unknown. We summarize new methods and present the circuitry of a large fraction of the brain of the fruit fly Drosophila melanogaster. Improved methods include new procedures to prepare, image, align, segment, find synapses in, and proofread such large data sets. We define cell types, refine computational compartments, and provide an exhaustive atlas of cell examples and types, many of them novel. We provide detailed circuits consisting of neurons and their chemical synapses for most of the central brain. We make the data public and simplify access, reducing the effort needed to answer circuit questions, and provide procedures linking the neurons defined by our analysis with genetic reagents. Biologically, we examine distributions of connection strengths, neural motifs on different scales, electrical consequences of compartmentalization, and evidence that maximizing packing density is an important criterion in the evolution of the fly\u2019s brain."
                },
                {
                    "title": "Excitation and Inhibition Balance Underlying Epileptiform Activity",
                    "abstract": "<italic>Objective:</italic> The phenomenon of postictal generalized EEG suppression state (PGES) \u2014 a period with suppressed activity following seizure termination and has been found to be associated with sudden unexpected death in epilepsy \u2014 remains poorly understood. This article aims to examine the how the balance of excitation and inhibition (E/I balance) affect the dynamics of seizure and PGES. <italic>Methods:</italic> A network of 1000 Izhikevich model neurons was developed and only the strengths of synaptic connections were adjusted to recreate the dynamics observed in recordings of seizure and PGES from human patients. <italic>Results:</italic> A rapid rise followed by a slow decay of dominant frequency was observed in iEEG recordings of ictal periods and reproduced in the simulated local field potential by changing the E/I balance of the model network. The rate of this dominant frequency evolution was quantified by a single measure, <inline-formula><tex-math notation=\"LaTeX\">$\\beta$</tex-math></inline-formula>, which was found to have a significant rank correlation with the duration of PGES in iEEG data and the rate of E/I balance shift in the model. <italic>Significance and Conclusion:</italic> (i) highlighting the importance of E/I balance in the dynamics of seizure and PGES; (ii) suggesting the measure, <inline-formula><tex-math notation=\"LaTeX\">$\\beta$</tex-math></inline-formula>, as a marker for PGES and the shift in E/I balance as a neural correlate for this marker."
                },
                {
                    "title": "Disentangling Trainability and Generalization in Deep Neural Networks",
                    "abstract": "A longstanding goal in the theory of deep learning is to characterize the conditions under which a given neural network architecture will be trainable, and if so, how well it might generalize to unseen data. In this work, we provide such a characterization in the limit of very wide and very deep networks, for which the analysis simplifies considerably. For wide networks, the trajectory under gradient descent is governed by the Neural Tangent Kernel (NTK), and for deep networks the NTK itself maintains only weak data dependence. By analyzing the spectrum of the NTK, we formulate necessary conditions for trainability and generalization across a range of architectures, including Fully Connected Networks (FCNs) and Convolutional Neural Networks (CNNs). We identify large regions of hyperparameter space for which networks can memorize the training set but completely fail to generalize. We find that CNNs without global average pooling behave almost identically to FCNs, but that CNNs with pooling have markedly different and often better generalization performance. These theoretical results are corroborated experimentally on CIFAR10 for a variety of network architectures and we include a colab notebook that reproduces the essential results of the paper."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Consequences of Dale's law on the stability-complexity relationship of random neural networks.",
                    "abstract": "In the study of randomly connected neural network dynamics there is a phase transition from a simple state with few equilibria to a complex state characterized by the number of equilibria growing exponentially with the neuron population. Such phase transitions are often used to describe pathological brain state transitions observed in neurological diseases such as epilepsy. In this paper we investigate how more realistic heterogeneous network structures affect these phase transitions using techniques from random matrix theory. Specifically, we parametrize the network structure according to Dale's law and use the Kac-Rice formalism to compute the change in the number of equilibria when a phase transition occurs. We also examine the condition where the network is not balanced between excitation and inhibition causing outliers to appear in the eigenspectrum. This enables us to compute the effects of different heterogeneous network connectivities on brain state transitions, which can provide insights into pathological brain dynamics."
                },
                {
                    "title": "A Unified Framework of Online Learning Algorithms for Training Recurrent Neural Networks",
                    "abstract": "We present a framework for compactly summarizing many recent results in efficient and/or biologically plausible online training of recurrent neural networks (RNN). The framework organizes algorithms according to several criteria: (a) past vs. future facing, (b) tensor structure, (c) stochastic vs. deterministic, and (d) closed form vs. numerical. These axes reveal latent conceptual connections among several recent advances in online learning. Furthermore, we provide novel mathematical intuitions for their degree of success. Testing various algorithms on two synthetic tasks shows that performances cluster according to our criteria. Although a similar clustering is also observed for gradient alignment, alignment with exact methods does not alone explain ultimate performance, especially for stochastic algorithms. This suggests the need for better comparison metrics."
                },
                {
                    "title": "Disentangling feature and lazy training in deep neural networks",
                    "abstract": "Two distinct limits for deep learning have been derived as the network width h \u2192 \u221e, depending on how the weights of the last layer scale with h. In the neural tangent Kernel (NTK) limit, the dynamics becomes linear in the weights and is described by a frozen kernel \u0398 (the NTK). By contrast, in the mean-field limit, the dynamics can be expressed in terms of the distribution of the parameters associated with a neuron, that follows a partial differential equation. In this work we consider deep networks where the weights in the last layer scale as \u03b1h \u22121/2 at initialization. By varying \u03b1 and h, we probe the crossover between the two limits. We observe two the previously identified regimes of \u2018lazy training\u2019 and \u2018feature training\u2019. In the lazy-training regime, the dynamics is almost linear and the NTK barely changes after initialization. The feature-training regime includes the mean-field formulation as a limiting case and is characterized by a kernel that evolves in time, and thus learns some features. We perform numerical experiments on MNIST, Fashion-MNIST, EMNIST and CIFAR10 and consider various architectures. We find that: (i) the two regimes are separated by an \u03b1* that scales as 1h . (ii) Network architecture and data structure play an important role in determining which regime is better: in our tests, fully-connected networks perform generally better in the lazy-training regime, unlike convolutional networks. (iii) In both regimes, the fluctuations \u03b4F induced on the learned function by initial conditions decay as \u03b4F\u223c1/h , leading to a performance that increases with h. The same improvement can also be obtained at an intermediate width by ensemble-averaging several networks that are trained independently. (iv) In the feature-training regime we identify a time scale t1\u223ch\u03b1 , such that for t \u226a t 1 the dynamics is linear. At t \u223c t 1, the output has grown by a magnitude h and the changes of the tangent kernel | |\u0394\u0398| | become significant. Ultimately, it follows ||\u0394\u0398||\u223c(h\u03b1)\u2212a for ReLU and Softplus activation functions, with a < 2 and a \u2192 2 as depth grows. We provide scaling arguments supporting these findings."
                },
                {
                    "title": "Kernel and Rich Regimes in Overparametrized Models",
                    "abstract": "A recent line of work studies overparametrized neural networks in the \"kernel regime,\" i.e. when the network behaves during training as a kernelized linear predictor, and thus training with gradient descent has the effect of finding the minimum RKHS norm solution. This stands in contrast to other studies which demonstrate how gradient descent on overparametrized multilayer networks can induce rich implicit biases that are not RKHS norms. Building on an observation by Chizat and Bach, we show how the scale of the initialization controls the transition between the \"kernel\" (aka lazy) and \"rich\" (aka active) regimes and affects generalization properties in multilayer homogeneous models. We also highlight an interesting role for the width of a model in the case that the predictor is not identically zero at initialization. We provide a complete and detailed analysis for a family of simple depth-$D$ models that already exhibit an interesting and meaningful transition between the kernel and rich regimes, and we also demonstrate this transition empirically for more complex matrix factorization models and multilayer non-linear networks."
                },
                {
                    "title": "Reconstruction of 1,000 Projection Neurons Reveals New Cell Types and Organization of Long-Range Connectivity in the Mouse Brain",
                    "abstract": "Neuronal cell types are the nodes of neural circuits that determine the flow of information within the brain. Neuronal morphology, especially the shape of the axonal arbor, provides an essential descriptor of cell type and reveals how individual neurons route their output across the brain. Despite the importance of morphology, few projection neurons in the mouse brain have been reconstructed in their entirety. Here we present a robust and efficient platform for imaging and reconstructing complete neuronal morphologies, including axonal arbors that span substantial portions of the brain. We used this platform to reconstruct more than 1,000 projection neurons in the motor cortex, thalamus, subiculum, and hypothalamus. Together, the reconstructed neurons comprise more than 75 meters of axonal length and are available in a searchable online database. Axonal shapes revealed previously unknown subtypes of projection neurons and suggest organizational principles of long-range connectivity."
                },
                {
                    "title": "On Lazy Training in Differentiable Programming",
                    "abstract": "In a series of recent theoretical works, it was shown that strongly over-parameterized neural networks trained with gradient-based methods could converge exponentially fast to zero training loss, with their parameters hardly varying. In this work, we show that this \"lazy training\" phenomenon is not specific to over-parameterized neural networks, and is due to a choice of scaling, often implicit, that makes the model behave as its linearization around the initialization, thus yielding a model equivalent to learning with positive-definite kernels. Through a theoretical analysis, we exhibit various situations where this phenomenon arises in non-convex optimization and we provide bounds on the distance between the lazy and linearized optimization paths. Our numerical experiments bring a critical note, as we observe that the performance of commonly used non-linear deep convolutional neural networks in computer vision degrades when trained in the lazy regime. This makes it unlikely that \"lazy training\" is behind the many successes of neural networks in difficult high dimensional tasks."
                },
                {
                    "title": "Local online learning in recurrent networks with random feedback",
                    "abstract": "A longstanding challenge for computational neuroscience has been the development of biologically plausible learning rules for recurrent neural networks (RNNs) enabling the production and processing of time-dependent signals such as those that might drive movement or facilitate working memory. Classic gradient-based algorithms for training RNNs have been available for decades, but they are inconsistent with known biological features of the brain, such as causality and locality. In this work we derive an approximation to gradient-based learning that comports with these biologically motivated constraints. Specifically, the online learning rule for the synaptic weights involves only local information about the pre- and postsynaptic activities, in addition to a random feedback projection of the RNN output error. In addition to providing mathematical arguments for the effectiveness of the new learning rule, we show through simulations that it can be used to train an RNN to successfully perform a variety of tasks. Finally, to overcome the difficulty of training an RNN over a very large number of timesteps, we propose an augmented circuit architecture that allows the RNN to concatenate short-duration patterns into sequences of longer duration."
                },
                {
                    "title": "Dendritic cortical microcircuits approximate the backpropagation algorithm",
                    "abstract": "Deep learning has seen remarkable developments over the last years, many of them inspired by neuroscience. However, the main learning mechanism behind these advances - error backpropagation - appears to be at odds with neurobiology. Here, we introduce a multilayer neuronal network model with simplified dendritic compartments in which error-driven synaptic plasticity adapts the network towards a global desired output. In contrast to previous work our model does not require separate phases and synaptic learning is driven by local dendritic prediction errors continuously in time. Such errors originate at apical dendrites and occur due to a mismatch between predictive input from lateral interneurons and activity from actual top-down feedback. Through the use of simple dendritic compartments and different cell-types our model can represent both error and normal activity within a pyramidal neuron. We demonstrate the learning capabilities of the model in regression and classification tasks, and show analytically that it approximates the error backpropagation algorithm. Moreover, our framework is consistent with recent observations of learning between brain areas and the architecture of cortical microcircuits. Overall, we introduce a novel view of learning on dendritic cortical circuits and on how the brain may solve the long-standing synaptic credit assignment problem."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "On the information bottleneck theory of deep learning",
                    "abstract": "The practical successes of deep neural networks have not been matched by theoretical progress that satisfyingly explains their behavior. In this work, we study the information bottleneck (IB) theory of deep learning, which makes three specific claims: first, that deep networks undergo two distinct phases consisting of an initial fitting phase and a subsequent compression phase; second, that the compression phase is causally related to the excellent generalization performance of deep networks; and third, that the compression phase occurs due to the diffusion-like behavior of stochastic gradient descent. Here we show that none of these claims hold true in the general case, and instead reflect assumptions made to compute a finite mutual information metric in deterministic networks. When computed using simple binning, we demonstrate through a combination of analytical results and simulation that the information plane trajectory observed in prior work is predominantly a function of the neural nonlinearity employed: double-sided saturating nonlinearities like yield a compression phase as neural activations enter the saturation regime, but linear activation functions and single-sided saturating nonlinearities like the widely used ReLU in fact do not. Moreover, we find that there is no evident causal connection between compression and generalization: networks that do not compress are still capable of generalization, and vice versa. Next, we show that the compression phase, when it exists, does not arise from stochasticity in training by demonstrating that we can replicate the IB findings using full batch gradient descent rather than stochastic gradient descent. Finally, we show that when an input domain consists of a subset of task-relevant and task-irrelevant information, hidden representations do compress the task-irrelevant information, although the overall information about the input may monotonically increase with training time, and that this compression happens concurrently with the fitting process rather than during a subsequent compression period."
                },
                {
                    "title": "A theory of multineuronal dimensionality, dynamics and measurement",
                    "abstract": "In many experiments, neuroscientists tightly control behavior, record many trials, and obtain trial-averaged firing rates from hundreds of neurons in circuits containing billions of behaviorally relevant neurons. Di-mensionality reduction methods reveal a striking simplicity underlying such multi-neuronal data: they can be reduced to a low-dimensional space, and the resulting neural trajectories in this space yield a remarkably insightful dynamical portrait of circuit computation. This simplicity raises profound and timely conceptual questions. What are its origins and its implications for the complexity of neural dynamics? How would the situation change if we recorded more neurons? When, if at all, can we trust dynamical portraits obtained from measuring an infinitesimal fraction of task relevant neurons? We present a theory that answers these questions, and test it using physiological recordings from reaching monkeys. This theory reveals conceptual insights into how task complexity governs both neural dimensionality and accurate recovery of dynamic portraits, thereby providing quantitative guidelines for future large-scale experimental design."
                },
                {
                    "title": "Overcoming catastrophic forgetting in neural networks",
                    "abstract": "Significance Deep neural networks are currently the most successful machine-learning technique for solving a variety of tasks, including language translation, image classification, and image generation. One weakness of such models is that, unlike humans, they are unable to learn multiple tasks sequentially. In this work we propose a practical solution to train such models sequentially by protecting the weights important for previous tasks. This approach, inspired by synaptic consolidation in neuroscience, enables state of the art results on multiple reinforcement learning problems experienced sequentially. The ability to learn tasks in a sequential fashion is crucial to the development of artificial intelligence. Until now neural networks have not been capable of this and it has been widely thought that catastrophic forgetting is an inevitable feature of connectionist models. We show that it is possible to overcome this limitation and train networks that can maintain expertise on tasks that they have not experienced for a long time. Our approach remembers old tasks by selectively slowing down learning on the weights important for those tasks. We demonstrate our approach is scalable and effective by solving a set of classification tasks based on a hand-written digit dataset and by learning several Atari 2600 games sequentially."
                },
                {
                    "title": "Feedback through graph motifs relates structure and function in complex networks",
                    "abstract": "In physics, biology and engineering, network systems abound. How does the connectivity of a network system combine with the behavior of its individual components to determine its collective function? We approach this question for networks with linear time-invariant dynamics by relating internal network feedbacks to the statistical prevalence of connectivity motifs, a set of surprisingly simple and local statistics of connectivity. This results in a reduced order model of the network input-output dynamics in terms of motifs structures. As an example, the new formulation dramatically simplifies the classic Erdos-Renyi graph, reducing the overall network behavior to one proportional feedback wrapped around the dynamics of a single node. For general networks, higher-order motifs systematically provide further layers and types of feedback to regulate the network response. Thus, the local connectivity shapes temporal and spectral processing by the network as a whole, and we show how this enables robust, yet tunable, functionality such as extending the time constant with which networks remember past signals. The theory also extends to networks composed from heterogeneous nodes with distinct dynamics and connectivity, and patterned input to (and readout from) subsets of nodes. These statistical descriptions provide a powerful theoretical framework to understand the functionality of real-world network systems, as we illustrate with examples including the mouse brain connectome."
                },
                {
                    "title": "Equilibrium Propagation: Bridging the Gap between Energy-Based Models and Backpropagation",
                    "abstract": "We introduce Equilibrium Propagation, a learning framework for energy-based models. It involves only one kind of neural computation, performed in both the first phase (when the prediction is made) and the second phase of training (after the target or prediction error is revealed). Although this algorithm computes the gradient of an objective function just like Backpropagation, it does not need a special computation or circuit for the second phase, where errors are implicitly propagated. Equilibrium Propagation shares similarities with Contrastive Hebbian Learning and Contrastive Divergence while solving the theoretical issues of both algorithms: our algorithm computes the gradient of a well-defined objective function. Because the objective function is defined in terms of local perturbations, the second phase of Equilibrium Propagation corresponds to only nudging the prediction (fixed point or stationary distribution) toward a configuration that reduces prediction error. In the case of a recurrent multi-layer supervised network, the output units are slightly nudged toward their target in the second phase, and the perturbation introduced at the output layer propagates backward in the hidden layers. We show that the signal \u201cback-propagated\u201d during this second phase corresponds to the propagation of error derivatives and encodes the gradient of the objective function, when the synaptic update corresponds to a standard form of spike-timing dependent plasticity. This work makes it more plausible that a mechanism similar to Backpropagation could be implemented by brains, since leaky integrator neural computation performs both inference and error back-propagation in our model. The only local difference between the two phases is whether synaptic changes are allowed or not. We also show experimentally that multi-layer recurrently connected networks with 1, 2, and 3 hidden layers can be trained by Equilibrium Propagation on the permutation-invariant MNIST task."
                },
                {
                    "title": "Training Excitatory-Inhibitory Recurrent Neural Networks for Cognitive Tasks: A Simple and Flexible Framework",
                    "abstract": "The ability to simultaneously record from large numbers of neurons in behaving animals has ushered in a new era for the study of the neural circuit mechanisms underlying cognitive functions. One promising approach to uncovering the dynamical and computational principles governing population responses is to analyze model recurrent neural networks (RNNs) that have been optimized to perform the same tasks as behaving animals. Because the optimization of network parameters specifies the desired output but not the manner in which to achieve this output, \u201ctrained\u201d networks serve as a source of mechanistic hypotheses and a testing ground for data analyses that link neural computation to behavior. Complete access to the activity and connectivity of the circuit, and the ability to manipulate them arbitrarily, make trained networks a convenient proxy for biological circuits and a valuable platform for theoretical investigation. However, existing RNNs lack basic biological features such as the distinction between excitatory and inhibitory units (Dale\u2019s principle), which are essential if RNNs are to provide insights into the operation of biological circuits. Moreover, trained networks can achieve the same behavioral performance but differ substantially in their structure and dynamics, highlighting the need for a simple and flexible framework for the exploratory training of RNNs. Here, we describe a framework for gradient descent-based training of excitatory-inhibitory RNNs that can incorporate a variety of biological knowledge. We provide an implementation based on the machine learning library Theano, whose automatic differentiation capabilities facilitate modifications and extensions. We validate this framework by applying it to well-known experimental paradigms such as perceptual decision-making, context-dependent integration, multisensory integration, parametric working memory, and motor sequence generation. Our results demonstrate the wide range of neural activity patterns and behavior that can be modeled, and suggest a unified setting in which diverse cognitive computations and mechanisms can be studied."
                },
                {
                    "title": "Deep learning and the information bottleneck principle",
                    "abstract": "Deep Neural Networks (DNNs) are analyzed via the theoretical framework of the information bottleneck (IB) principle. We first show that any DNN can be quantified by the mutual information between the layers and the input and output variables. Using this representation we can calculate the optimal information theoretic limits of the DNN and obtain finite sample generalization bounds. The advantage of getting closer to the theoretical limit is quantifiable both by the generalization bound and by the network's simplicity. We argue that both the optimal architecture, number of layers and features/connections at each layer, are related to the bifurcation points of the information bottleneck tradeoff, namely, relevant compression of the input layer with respect to the output layer. The hierarchical representations at the layered network naturally correspond to the structural phase transitions along the information curve. We believe that this new insight can lead to new optimality bounds and deep learning algorithms."
                },
                {
                    "title": "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification",
                    "abstract": "Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on the learnable activation and advanced initialization, we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66% [33]). To our knowledge, our result is the first to surpass the reported human-level performance (5.1%, [26]) on this dataset."
                },
                {
                    "title": "Transition to chaos in random networks with cell-type-specific connectivity.",
                    "abstract": "In neural circuits, statistical connectivity rules strongly depend on cell-type identity. We study dynamics of neural networks with cell-type-specific connectivity by extending the dynamic mean-field method and find that these networks exhibit a phase transition between silent and chaotic activity. By analyzing the locus of this transition, we derive a new result in random matrix theory: the spectral radius of a random connectivity matrix with block-structured variances. We apply our results to show how a small group of hyperexcitable neurons within the network can significantly increase the network's computational capacity by bringing it into the chaotic regime."
                },
                {
                    "title": "The developmental stages of synaptic plasticity",
                    "abstract": "Abstract\u2002 The brain is programmed to drive behaviour by precisely wiring the appropriate neuronal circuits. Wiring and rewiring of neuronal circuits largely depends on the orchestrated changes in the strengths of synaptic contacts. Here, we review how the rules of synaptic plasticity change during development of the brain, from birth to independence. We focus on the changes that occur at the postsynaptic side of excitatory glutamatergic synapses in the rodent hippocampus and neocortex. First we summarize the current data on the structure of synapses and the developmental expression patterns of the key molecular players of synaptic plasticity, N\u2010methyl\u2010d\u2010aspartate (NMDA) and \u03b1\u2010amino\u20103\u2010hydroxy\u20105\u2010methyl\u20104\u2010isoxazolepropionic acid (AMPA) receptors, as well as pivotal kinases (Ca2+/calmodulin\u2010dependent protein kinase II, protein kinase A, protein kinase C) and phosphatases (PP1, PP2A, PP2B). In the second part we relate these findings to important characteristics of the emerging network. We argue that the concerted and gradual shifts in the usage of plasticity molecules comply with the changing need for (re)wiring neuronal circuits."
                },
                {
                    "title": "Exact solutions to the nonlinear dynamics of learning in deep linear neural networks",
                    "abstract": "Despite the widespread practical success of deep learning methods, our theoretical understanding of the dynamics of learning in deep neural networks remains quite sparse. We attempt to bridge the gap between the theory and practice of deep learning by systematically analyzing learning dynamics for the restricted case of deep linear neural networks. Despite the linearity of their input-output map, such networks have nonlinear gradient descent dynamics on weights that change with the addition of each new hidden layer. We show that deep linear networks exhibit nonlinear learning phenomena similar to those seen in simulations of nonlinear networks, including long plateaus followed by rapid transitions to lower error solutions, and faster convergence from greedy unsupervised pretraining initial conditions than from random initial conditions. We provide an analytical description of these phenomena by finding new exact solutions to the nonlinear dynamics of deep learning. Our theoretical analysis also reveals the surprising finding that as the depth of a network approaches infinity, learning speed can nevertheless remain finite: for a special class of initial conditions on the weights, very deep networks incur only a finite, depth independent, delay in learning speed relative to shallow networks. We show that, under certain conditions on the training data, unsupervised pretraining can find this special class of initial conditions, while scaled random Gaussian initializations cannot. We further exhibit a new class of random orthogonal initial conditions on weights that, like unsupervised pre-training, enjoys depth independent learning times. We further show that these initial conditions also lead to faithful propagation of gradients even in deep nonlinear networks, as long as they operate in a special regime known as the edge of chaos."
                },
                {
                    "title": "To Favor Survival Under Food Shortage, the Brain Disables Costly Memory",
                    "abstract": "Hunger and Memory During starvation, are all brain functions slowed down, or are specific functions disabled to save energy? Pla\u00e7ais and Preat (p. 440) investigated how the brain of Drosophila deals with severe resource limitation. The brain cut selected expenses to reduce the threat to survival and switched off the formation of aversive long-term memory that depends on costly protein synthesis. However, Hirano et al. (p. 443) focused on mild food-deprivation, which actually enhanced long-term memory formation. Presumably, improved memory should enhance survival when competing for limited food. After longer food deprivation, enhancement of aversive long-term memory decreased, while that of appetitive long-term memory remained high: Presumably, as starvation nears, it becomes more important to pursue food at all costs, and so appetitive memory takes precedence over aversive memories. Different types of memory interact in fed versus starved states to promote survival-oriented behavior. The brain regulates energy homeostasis in the organism. Under resource shortage, the brain takes priority over peripheral organs for energy supply. But can the brain also down-regulate its own consumption to favor survival? We show that the brain of Drosophila specifically disables the costly formation of aversive long-term memory (LTM) upon starvation, a physiological state required for appetitive LTM formation. At the neural circuit level, the slow oscillations normally triggered in two pairs of dopaminergic neurons to enable aversive LTM formation were abolished in starved flies. Transient artificial activation of these neurons during training restored LTM formation in starved flies but at the price of a reduced survival. LTM formation is thus subject to adaptive plasticity that helps survival under food shortage."
                },
                {
                    "title": "Synchronization from Second Order Network Connectivity Statistics",
                    "abstract": "We investigate how network structure can influence the tendency for a neuronal network to synchronize, or its synchronizability, independent of the dynamical model for each neuron. The synchrony analysis takes advantage of the framework of second order networks, which defines four second order connectivity statistics based on the relative frequency of two-connection network motifs. The analysis identifies two of these statistics, convergent connections, and chain connections, as highly influencing the synchrony. Simulations verify that synchrony decreases with the frequency of convergent connections and increases with the frequency of chain connections. These trends persist with simulations of multiple models for the neuron dynamics and for different types of networks. Surprisingly, divergent connections, which determine the fraction of shared inputs, do not strongly influence the synchrony. The critical role of chains, rather than divergent connections, in influencing synchrony can be explained by their increasing the effective coupling strength. The decrease of synchrony with convergent connections is primarily due to the resulting heterogeneity in firing rates."
                },
                {
                    "title": "Understanding the difficulty of training deep feedforward neural networks",
                    "abstract": "Whereas before 2006 it appears that deep multilayer neural networks were not successfully trained, since then several algorithms have been shown to successfully train them, with experimental results showing the superiority of deeper vs less deep architectures. All these experimental results were obtained with new initialization or training mechanisms. Our objective here is to understand better why standard gradient descent from random initialization is doing so poorly with deep neural networks, to better understand these recent relative successes and help design better algorithms in the future. We first observe the influence of the non-linear activations functions. We find that the logistic sigmoid activation is unsuited for deep networks with random initialization because of its mean value, which can drive especially the top hidden layer into saturation. Surprisingly, we find that saturated units can move out of saturation by themselves, albeit slowly, and explaining the plateaus sometimes seen when training neural networks. We find that a new non-linearity that saturates less can often be beneficial. Finally, we study how activations and gradients vary across layers and during training, with the idea that training may be more difficult when the singular values of the Jacobian associated with each layer are far from 1. Based on these considerations, we propose a new initialization scheme that brings substantially faster convergence. 1 Deep Neural Networks Deep learning methods aim at learning feature hierarchies with features from higher levels of the hierarchy formed by the composition of lower level features. They include Appearing in Proceedings of the 13 International Conference on Artificial Intelligence and Statistics (AISTATS) 2010, Chia Laguna Resort, Sardinia, Italy. Volume 9 of JMLR: WC Weston et al., 2008). Much attention has recently been devoted to them (see (Bengio, 2009) for a review), because of their theoretical appeal, inspiration from biology and human cognition, and because of empirical success in vision (Ranzato et al., 2007; Larochelle et al., 2007; Vincent et al., 2008) and natural language processing (NLP) (Collobert & Weston, 2008; Mnih & Hinton, 2009). Theoretical results reviewed and discussed by Bengio (2009), suggest that in order to learn the kind of complicated functions that can represent high-level abstractions (e.g. in vision, language, and other AI-level tasks), one may need deep architectures. Most of the recent experimental results with deep architecture are obtained with models that can be turned into deep supervised neural networks, but with initialization or training schemes different from the classical feedforward neural networks (Rumelhart et al., 1986). Why are these new algorithms working so much better than the standard random initialization and gradient-based optimization of a supervised training criterion? Part of the answer may be found in recent analyses of the effect of unsupervised pretraining (Erhan et al., 2009), showing that it acts as a regularizer that initializes the parameters in a \u201cbetter\u201d basin of attraction of the optimization procedure, corresponding to an apparent local minimum associated with better generalization. But earlier work (Bengio et al., 2007) had shown that even a purely supervised but greedy layer-wise procedure would give better results. So here instead of focusing on what unsupervised pre-training or semi-supervised criteria bring to deep architectures, we focus on analyzing what may be going wrong with good old (but deep) multilayer neural networks. Our analysis is driven by investigative experiments to monitor activations (watching for saturation of hidden units) and gradients, across layers and across training iterations. We also evaluate the effects on these of choices of activation function (with the idea that it might affect saturation) and initialization procedure (since unsupervised pretraining is a particular form of initialization and it has a drastic impact)."
                },
                {
                    "title": "Eigenvalue spectra of random matrices for neural networks.",
                    "abstract": "The dynamics of neural networks is influenced strongly by the spectrum of eigenvalues of the matrix describing their synaptic connectivity. In large networks, elements of the synaptic connectivity matrix can be chosen randomly from appropriate distributions, making results from random matrix theory highly relevant. Unfortunately, classic results on the eigenvalue spectra of random matrices do not apply to synaptic connectivity matrices because of the constraint that individual neurons are either excitatory or inhibitory. Therefore, we compute eigenvalue spectra of large random matrices with excitatory and inhibitory columns drawn from distributions with different means and equal or different variances."
                },
                {
                    "title": "A Cost of Long-Term Memory in Drosophila",
                    "abstract": "Two distinct forms of consolidated associative memory are known in Drosophila: long-term memory and so-called anesthesia-resistant memory. Long-term memory is more stable, but unlike anesthesia-resistant memory, its formation requires protein synthesis. We show that flies induced to form long-term memory become more susceptible to extreme stress (such as desiccation). In contrast, induction of anesthesia-resistant memory had no detectable effect on desiccation resistance. This finding may help to explain why evolution has maintained anesthesia-resistant memory as another form of consolidated memory, distinct from long-term memory."
                },
                {
                    "title": "Highly Nonrandom Features of Synaptic Connectivity in Local Cortical Circuits",
                    "abstract": "How different is local cortical circuitry from a random network? To answer this question, we probed synaptic connections with several hundred simultaneous quadruple whole-cell recordings from layer 5 pyramidal neurons in the rat visual cortex. Analysis of this dataset revealed several nonrandom features in synaptic connectivity. We confirmed previous reports that bidirectional connections are more common than expected in a random network. We found that several highly clustered three-neuron connectivity patterns are overrepresented, suggesting that connections tend to cluster together. We also analyzed synaptic connection strength as defined by the peak excitatory postsynaptic potential amplitude. We found that the distribution of synaptic connection strength differs significantly from the Poisson distribution and can be fitted by a lognormal distribution. Such a distribution has a heavier tail and implies that synaptic weight is concentrated among few synaptic connections. In addition, the strengths of synaptic connections sharing pre- or postsynaptic neurons are correlated, implying that strong connections are even more clustered than the weak ones. Therefore, the local cortical network structure can be viewed as a skeleton of stronger connections in a sea of weaker ones. Such a skeleton is likely to play an important role in network dynamics and should be investigated further."
                },
                {
                    "title": "Learning of correlated patterns in spin-glass networks by local learning rules.",
                    "abstract": "Deux prescriptions simples de mise en memoire sont presentees pour les modeles de reseaux imitant le systeme nerveux a N neurones a deux etats. Ces regles sont locales et permettent l'immersion de schemas correles sans erreurs dans un reseau de type verre de spins. A partir d'une configuration arbitraire de liaisons synaptiques, on peut mettre en memoire jusqu'a N schemas par des modifications successives des efficacites synaptiques. On demontre la convergence. Extensions des regles"
                }
            ],
            "categories": [
                "cs.NE",
                "cs.AI",
                "q-bio.NC"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does the initial weight structure, given a fixed initial weight magnitude, bias the learning regime in deep learning models?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing our understanding of how different initial weight structures influence learning dynamics in neural networks. By elucidating the relationship between initial weight structure and learning regimes, this research could lead to improved model architectures and training strategies, enhancing the efficiency and effectiveness of deep learning systems. Furthermore, insights gained could inform the design of biologically-inspired neural networks, bridging the gap between theoretical neuroscience and machine learning, and potentially leading to practical applications in various fields such as robotics, natural language processing, and computer vision.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complex interplay between initial weight structures and their impact on learning dynamics, which is not fully understood. Naive approaches that focus solely on random initializations may overlook the nuanced effects of structured initial weights, leading to incomplete or misleading conclusions. Additionally, the theoretical framework for analyzing learning regimes, particularly the effective richness or laziness of learning, requires sophisticated mathematical tools and a deep understanding of the Neural Tangent Kernel (NTK) dynamics. Overcoming these technical and theoretical obstacles is essential for accurately characterizing the learning behavior of neural networks.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily concentrated on random Gaussian or Uniform initializations, neglecting the potential influence of structured initial weight configurations that mimic biological neural circuits. This gap in the literature stems from a lack of comprehensive theoretical frameworks that connect initial weight structures to learning dynamics. Additionally, existing studies may not have adequately addressed the concept of effective learning richness/laziness, which extends beyond traditional definitions based on initialization alone. Our approach aims to fill these gaps by providing a theoretical basis and empirical validation for the role of effective rank in modulating learning dynamics.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves theoretical derivation within a two-layer feedforward linear network to demonstrate the relationship between higher-rank initialization and effectively lazier learning across tasks. We will utilize a range of datasets to empirically validate our theoretical findings, measuring the NTK movement during training as a key metric for assessing effective learning richness or laziness. The expected outcome is a clearer understanding of how initial weight structures influence learning dynamics, leading to actionable insights"
            }
        },
        "author_data": {
            "5eb3f424-52a3-4fc4-9b10-c326ff0af28c": {
                "pk": "5eb3f424-52a3-4fc4-9b10-c326ff0af28c",
                "name": "Yuhan Helena Liu",
                "collaborators": [
                    "E. Shea-Brown",
                    "Stefan Mihalas",
                    "U. S\u00fcmb\u00fcl",
                    "Arna Ghosh",
                    "Guillaume Lajoie",
                    "Stephen Smith",
                    "Stephen J. Smith",
                    "Dana Mastrovito",
                    "Lukasz Kusmierz",
                    "Christof Koch",
                    "Konrad K\u00a8ording",
                    "Blake A. Richards",
                    "James Hazelden",
                    "Eli Shlizerman",
                    "B. Richards",
                    "Uygar Sumbul",
                    "Si-Hyeon Lee",
                    "A. Khisti"
                ],
                "domain": [
                    "Neural Networks",
                    "Biologically Plausible Learning",
                    "Gradient Descent",
                    "Spiking Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Transition to chaos separates learning regimes and relates to measure of consciousness in recurrent neural networks",
                        "abstract": "Recurrent neural networks exhibit chaotic dynamics when the variance in their connection strengths exceed a critical value. Recent work indicates connection variance also modulates learning strategies; networks learn \u201drich\u201d representations when initialized with low coupling and \u201dlazier\u201dsolutions with larger variance. Using Watts-Strogatz networks of varying sparsity, structure, and hidden weight variance, we find that the critical coupling strength dividing chaotic from ordered dynamics also differentiates rich and lazy learning strategies. Training moves both stable and chaotic networks closer to the edge of chaos, with networks learning richer representations before the transition to chaos. In contrast, biologically realistic connectivity structures foster stability over a wide range of variances. The transition to chaos is also reflected in a measure that clinically discriminates levels of consciousness, the perturbational complexity index (PCIst). Networks with high values of PCIst exhibit stable dynamics and rich learning, suggesting a consciousness prior may promote rich learning. The results suggest a clear relationship between critical dynamics, learning regimes and complexity-based measures of consciousness."
                    },
                    {
                        "title": "H OW GRADIENT ESTIMATOR VARIANCE AND BIAS COULD IMPACT LEARNING IN NEURAL CIRCUITS",
                        "abstract": "There is growing interest in understanding how real brains may approximate gradients and how gradients can be used to train neuromorphic chips. However, neither real brains nor neuromorphic chips can perfectly follow the loss gradient, so parameter updates would necessarily use gradient estimators that have some variance and/or bias. Therefore, there is a need to understand better how variance and bias in gradient estimators impact learning dependent on network and task properties. Here, we show that variance and bias can impair learning on the training data, but some degree of variance and bias in a gradient estimator can be beneficial for generalization. We find that the ideal amount of variance and bias in a gradient estimator are dependent on several properties of the network and task: the size and activity sparsity of the network, the norm of the gradient, and the curvature of the loss landscape. As such, whether considering biologically-plausible learning algorithms or algorithms for training neuromorphic chips, researchers can analyze these properties to determine whether their approximation to gradient descent will be effective for learning given their network and task properties."
                    },
                    {
                        "title": "Evolutionary algorithms as an alternative to backpropagation for supervised training of Biophysical Neural Networks and Neural ODEs",
                        "abstract": "Training networks consisting of biophysically accurate neuron models could allow for new insights into how brain circuits can organize and solve tasks. We begin by analyzing the extent to which the central algorithm for neural network learning -- stochastic gradient descent through backpropagation (BP) -- can be used to train such networks. We find that properties of biophysically based neural network models needed for accurate modelling such as stiffness, high nonlinearity and long evaluation timeframes relative to spike times makes BP unstable and divergent in a variety of cases. To address these instabilities and inspired by recent work, we investigate the use of \"gradient-estimating\" evolutionary algorithms (EAs) for training biophysically based neural networks. We find that EAs have several advantages making them desirable over direct BP, including being forward-pass only, robust to noisy and rigid losses, allowing for discrete loss formulations, and potentially facilitating a more global exploration of parameters. We apply our method to train a recurrent network of Morris-Lecar neuron models on a stimulus integration and working memory task, and show how it can succeed in cases where direct BP is inapplicable. To expand on the viability of EAs in general, we apply them to a general neural ODE problem and a stiff neural ODE benchmark and find again that EAs can out-perform direct BP here, especially for the over-parameterized regime. Our findings suggest that biophysical neurons could provide useful benchmarks for testing the limits of BP-adjacent methods, and demonstrate the viability of EAs for training networks with complex components."
                    },
                    {
                        "title": "Beyond accuracy: generalization properties of bio-plausible temporal credit assignment rules",
                        "abstract": "To unveil how the brain learns, ongoing work seeks biologically-plausible approximations of gradient descent algorithms for training recurrent neural networks (RNNs). Yet, beyond task accuracy, it is unclear if such learning rules converge to solutions that exhibit different levels of generalization than their nonbiologically-plausible counterparts. Leveraging results from deep learning theory based on loss landscape curvature, we ask: how do biologically-plausible gradient approximations affect generalization? We first demonstrate that state-of-the-art biologically-plausible learning rules for training RNNs exhibit worse and more variable generalization performance compared to their machine learning counterparts that follow the true gradient more closely. Next, we verify that such generalization performance is correlated significantly with loss landscape curvature, and we show that biologically-plausible learning rules tend to approach high-curvature regions in synaptic weight space. Using tools from dynamical systems, we derive theoretical arguments and present a theorem explaining this phenomenon. This predicts our numerical results, and explains why biologically-plausible rules lead to worse and more variable generalization properties. Finally, we suggest potential remedies that could be used by the brain to mitigate this effect. To our knowledge, our analysis is the first to identify the reason for this generalization gap between artificial and biologically-plausible learning rules, which can help guide future investigations into how the brain learns solutions that generalize."
                    },
                    {
                        "title": "Biologically-plausible backpropagation through arbitrary timespans via local neuromodulators",
                        "abstract": "The spectacular successes of recurrent neural network models where key parameters are adjusted via backpropagation-based gradient descent have inspired much thought as to how biological neuronal networks might solve the corresponding synaptic credit assignment problem. There is so far little agreement, however, as to how biological networks could implement the necessary backpropagation through time, given widely recognized constraints of biological synaptic network signaling architectures. Here, we propose that extra-synaptic diffusion of local neuromodulators such as neuropeptides may afford an effective mode of backpropagation lying within the bounds of biological plausibility. Going beyond existing temporal truncation-based gradient approximations, our approximate gradient-based update rule, ModProp, propagates credit information through arbitrary time steps. ModProp suggests that modulatory signals can act on receiving cells by convolving their eligibility traces via causal, time-invariant and synapse-type-specific filter taps. Our mathematical analysis of ModProp learning, together with simulation results on benchmark temporal tasks, demonstrate the advantage of ModProp over existing biologically-plausible temporal credit assignment rules. These results suggest a potential neuronal mechanism for signaling credit information related to recurrent interactions over a longer time horizon. Finally, we derive an in-silico implementation of ModProp that could serve as a low-complexity and causal alternative to backpropagation through time."
                    },
                    {
                        "title": "Cell-type-speci\ufb01c modulatory signaling promotes learning in spiking neural networks",
                        "abstract": "Animals learn and form memories by jointly adjusting the e\ufb03cacy of their synapses. How they e\ufb03ciently solve the underlying temporal credit assignment problem remains elusive. Here, we re-analyze the mathematical basis of gradient descent learning in recurrent spiking neural networks (RSNNs) in light of the recent single-cell transcriptomic evidence for cell-type-speci\ufb01c local neuropeptide signaling in the cortex. Our normative theory posits an important role for the notion of neuronal cell types and local di\ufb00usive communication by enabling biologically plausible and e\ufb03cient weight update. While obeying fundamental biological constraints, including separating excitatory vs inhibitory cell types and observing connection sparsity, we trained RSNNs for temporal credit assignment tasks spanning seconds and observed that the inclusion of local modulatory signaling improved learning e\ufb03ciency. Our learning rule puts forth a novel form of interaction between modulatory signals and synaptic transmission. Moreover, it suggests a computationally e\ufb03cient learning method for bio-inspired arti\ufb01cial intelligence."
                    },
                    {
                        "title": "Cell-type\u2013specific neuromodulation guides synaptic credit assignment in a spiking neural network",
                        "abstract": "Significance Synaptic connectivity provides the foundation for our present understanding of neuronal network function, but static connectivity cannot explain learning and memory. We propose a computational role for the diversity of cortical neuronal types and their associated cell-type\u2013specific neuromodulators in improving the efficiency of synaptic weight adjustments for task learning in neuronal networks. Brains learn tasks via experience-driven differential adjustment of their myriad individual synaptic connections, but the mechanisms that target appropriate adjustment to particular connections remain deeply enigmatic. While Hebbian synaptic plasticity, synaptic eligibility traces, and top-down feedback signals surely contribute to solving this synaptic credit-assignment problem, alone, they appear to be insufficient. Inspired by new genetic perspectives on neuronal signaling architectures, here, we present a normative theory for synaptic learning, where we predict that neurons communicate their contribution to the learning outcome to nearby neurons via cell-type\u2013specific local neuromodulation. Computational tests suggest that neuron-type diversity and neuron-type\u2013specific local neuromodulation may be critical pieces of the biological credit-assignment puzzle. They also suggest algorithms for improved artificial neural network learning efficiency."
                    },
                    {
                        "title": "A solution to temporal credit assignment using cell-type-specific modulatory signals",
                        "abstract": "Animals learn and form memories by jointly adjusting the efficacy of their synapses. How they efficiently solve the underlying temporal credit assignment problem remains elusive. Here, we re-analyze the mathematical basis of gradient descent learning in recurrent spiking neural networks (RSNNs) in light of the recent single-cell transcriptomic evidence for cell-type-specific local neuropeptide signaling in the cortex. Our normative theory posits an important role for the notion of neuronal cell types and local diffusive communication by enabling biologically plausible and efficient weight update. While obeying fundamental biological constraints, including separating excitatory vs inhibitory cell types and observing connection sparsity, we trained RSNNs for temporal credit assignment tasks spanning seconds and observed that the inclusion of local modulatory signaling improved learning efficiency. Our learning rule puts forth a novel form of interaction between modulatory signals and synaptic transmission. Moreover, it suggests a computationally efficient on-chip learning method for bio-inspired artificial intelligence."
                    },
                    {
                        "title": "Information-Theoretic Privacy in Smart Metering Systems Using Cascaded Rechargeable Batteries",
                        "abstract": "A rechargeable battery may alleviate the issue of privacy loss in a smart metering system by distorting a household's load profile. However, existing studies involve a single rechargeable battery, whereas in a network scenario, there could be multiple batteries connected together. In this letter, we study the extension where a user's electricity load is input into a network of two rechargeable batteries, connected in series, and operating individually. This battery network attempts to mask the user load from the utility provider. We focus on the case of independent identically distributed load profile and a system of ideal batteries with no conversion loss, and use normalized mutual information (leakage rate) as the privacy metric. We derive upper and lower bounds on the leakage rate in terms of (single-letter) mutual information expressions. On the achievability side, our information-theoretic upper bound captures the novel tension between minimizing the leakage across each individual battery and the effect of their joint interaction. For the lower bound, we show that a system with a single battery, whose storage capacity is the sum of the two individual batteries, can achieve a leakage rate at least as small as our proposed setup. Furthermore, we use simulations to compare achievable leakage of our proposed scheme with several baseline schemes. The achievable leakage rates obtained in this study could help us to elucidate the privacy performance of a network of batteries."
                    }
                ]
            },
            "1c4e8eff-7380-4ad7-9c17-05c31ddd67cf": {
                "pk": "1c4e8eff-7380-4ad7-9c17-05c31ddd67cf",
                "name": "Aristide Baratin",
                "collaborators": [
                    "Simon Lacoste-Julien",
                    "Aaron C. Courville",
                    "R. Devon Hjelm",
                    "Yan Zhang",
                    "Thomas George",
                    "Guillaume Lajoie",
                    "Razvan Pascanu",
                    "Pranshu Malviya",
                    "Gon\u00e7alo Mordido",
                    "Sarath Chandar",
                    "Yuchen Lu",
                    "Zhen Liu",
                    "Alessandro Sordoni",
                    "James Vuckovic",
                    "R\u00e9mi Tachet des Combes",
                    "C\u00e9sar Laurent",
                    "Pascal Vincent",
                    "Sarthak Mittal",
                    "Vinayak Tantia",
                    "Matthew Scicluna",
                    "Ioannis Mitliagkas",
                    "Sai Rajeswar",
                    "Yoshua Bengio",
                    "Simon Dufort-Labb'e",
                    "P. D'Oro",
                    "Evgenii Nikishin",
                    "Pierre-Luc Bacon",
                    "Anchit Jain",
                    "Rozhin Nobahari",
                    "Stefano Sarao Mannelli",
                    "Alexia Jolicoeur-Martineau",
                    "Kisoo Kwon",
                    "Boris Knyazev",
                    "Md Rifat Arefin",
                    "Francesco Locatello",
                    "Irina Rish",
                    "Dianbo Liu",
                    "Kenji Kawaguchi",
                    "Reza Babanezhad Harikandeh",
                    "Jerry Huang",
                    "R. Laroche",
                    "Jihye Kim",
                    "Romain Laroche",
                    "Brady Neal",
                    "Mohamed Ishmael Belghazi",
                    "Sherjil Ozair",
                    "Nasim Rahaman",
                    "Devansh Arpit",
                    "Felix Dr\u00e4xler",
                    "Min Lin",
                    "F. Hamprecht",
                    "Ishmael Belghazi"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Machine Learning",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Maxwell's Demon at Work: Efficient Pruning by Leveraging Saturation of Neurons",
                        "abstract": "When training deep neural networks, the phenomenon of $\\textit{dying neurons}$ $\\unicode{x2013}$units that become inactive or saturated, output zero during training$\\unicode{x2013}$ has traditionally been viewed as undesirable, linked with optimization challenges, and contributing to plasticity loss in continual learning scenarios. In this paper, we reassess this phenomenon, focusing on sparsity and pruning. By systematically exploring the impact of various hyperparameter configurations on dying neurons, we unveil their potential to facilitate simple yet effective structured pruning algorithms. We introduce $\\textit{Demon Pruning}$ (DemP), a method that controls the proliferation of dead neurons, dynamically leading to network sparsity. Achieved through a combination of noise injection on active units and a one-cycled schedule regularization strategy, DemP stands out for its simplicity and broad applicability. Experiments on CIFAR10 and ImageNet datasets demonstrate that DemP surpasses existing structured pruning techniques, showcasing superior accuracy-sparsity tradeoffs and training speedups. These findings suggest a novel perspective on dying neurons as a valuable resource for efficient model compression and optimization."
                    },
                    {
                        "title": "Bias in Motion: Theoretical Insights into the Dynamics of Bias in SGD Training",
                        "abstract": "Machine learning systems often acquire biases by leveraging undesired features in the data, impacting accuracy variably across different sub-populations. Current understanding of bias formation mostly focuses on the initial and final stages of learning, leaving a gap in knowledge regarding the transient dynamics. To address this gap, this paper explores the evolution of bias in a teacher-student setup modeling different data sub-populations with a Gaussian-mixture model. We provide an analytical description of the stochastic gradient descent dynamics of a linear classifier in this setting, which we prove to be exact in high dimension. Notably, our analysis reveals how different properties of sub-populations influence bias at different timescales, showing a shifting preference of the classifier during training. Applying our findings to fairness and robustness, we delineate how and when heterogeneous data and spurious features can generate and amplify bias. We empirically validate our results in more complex scenarios by training deeper networks on synthetic and real datasets, including CIFAR10, MNIST, and CelebA."
                    },
                    {
                        "title": "Any-Property-Conditional Molecule Generation with Self-Criticism using Spanning Trees",
                        "abstract": "Generating novel molecules is challenging, with most representations leading to generative models producing many invalid molecules. Spanning Tree-based Graph Generation (STGG) is a promising approach to ensure the generation of valid molecules, outperforming state-of-the-art SMILES and graph diffusion models for unconditional generation. In the real world, we want to be able to generate molecules conditional on one or multiple desired properties rather than unconditionally. Thus, in this work, we extend STGG to multi-property-conditional generation. Our approach, STGG+, incorporates a modern Transformer architecture, random masking of properties during training (enabling conditioning on any subset of properties and classifier-free guidance), an auxiliary property-prediction loss (allowing the model to self-criticize molecules and select the best ones), and other improvements. We show that STGG+ achieves state-of-the-art performance on in-distribution and out-of-distribution conditional generation, and reward maximization."
                    },
                    {
                        "title": "Unsupervised Concept Discovery Mitigates Spurious Correlations",
                        "abstract": "Models prone to spurious correlations in training data often produce brittle predictions and introduce unintended biases. Addressing this challenge typically involves methods relying on prior knowledge and group annotation to remove spurious correlations, which may not be readily available in many applications. In this paper, we establish a novel connection between unsupervised object-centric learning and mitigation of spurious correlations. Instead of directly inferring subgroups with varying correlations with labels, our approach focuses on discovering concepts: discrete ideas that are shared across input samples. Leveraging existing object-centric representation learning, we introduce CoBalT: a concept balancing technique that effectively mitigates spurious correlations without requiring human labeling of subgroups. Evaluation across the benchmark datasets for sub-population shifts demonstrate superior or competitive performance compared state-of-the-art baselines, without the need for group annotation. Code is available at https://github.com/rarefin/CoBalT."
                    },
                    {
                        "title": "Promoting Exploration in Memory-Augmented Adam using Critical Momenta",
                        "abstract": "Adaptive gradient-based optimizers, notably Adam, have left their mark in training large-scale deep learning models, offering fast convergence and robustness to hyperparameter settings. However, they often struggle with generalization, attributed to their tendency to converge to sharp minima in the loss landscape. To address this, we propose a new memory-augmented version of Adam that encourages exploration towards flatter minima by incorporating a buffer of critical momentum terms during training. This buffer prompts the optimizer to overshoot beyond narrow minima, promoting exploration. Through comprehensive analysis in simple settings, we illustrate the efficacy of our approach in increasing exploration and bias towards flatter minima. We empirically demonstrate that it can improve model performance for image classification on ImageNet and CIFAR10/100, language modelling on Penn Treebank, and online learning tasks on TinyImageNet and 5-dataset. Our code is available at \\url{https://github.com/chandar-lab/CMOptimizer}."
                    },
                    {
                        "title": "Lookbehind-SAM: k steps back, 1 step forward",
                        "abstract": "Sharpness-aware minimization (SAM) methods have gained increasing popularity by formulating the problem of minimizing both loss value and loss sharpness as a minimax objective. In this work, we increase the efficiency of the maximization and minimization parts of SAM's objective to achieve a better loss-sharpness trade-off. By taking inspiration from the Lookahead optimizer, which uses multiple descent steps ahead, we propose Lookbehind, which performs multiple ascent steps behind to enhance the maximization step of SAM and find a worst-case perturbation with higher loss. Then, to mitigate the variance in the descent step arising from the gathered gradients across the multiple ascent steps, we employ linear interpolation to refine the minimization step. Lookbehind leads to a myriad of benefits across a variety of tasks. Particularly, we show increased generalization performance, greater robustness against noisy weights, as well as improved learning and less catastrophic forgetting in lifelong learning settings. Our code is available at https://github.com/chandar-lab/Lookbehind-SAM."
                    },
                    {
                        "title": "Expressiveness and Learnability: A Unifying View for Evaluating Self-Supervised Learning",
                        "abstract": "We propose a unifying view to analyze the representation quality of self-supervised learning (SSL) models without access to supervised labels, while being agnostic to the architecture, learning algorithm or data manipulation used during training. We argue that representations can be evaluated through the lens of expressiveness and learnability. We propose to use the Intrinsic Dimension (ID) to assess expressiveness and introduce Cluster Learnability (CL) to assess learnability. CL is measured as the learning speed of a KNN classifier trained to predict labels obtained by clustering the representations with K-means. We thus combine CL and ID into a single predictor: CLID. Through a large-scale empirical study with a diverse family of SSL algorithms, we find that CLID better correlates with in-distribution model performance than other competing recent evaluation schemes. We also benchmark CLID on out-of-domain generalization, where CLID serves as a predictor of the transfer performance of SSL models on several classification tasks, yielding improvements with respect to the competing baselines."
                    },
                    {
                        "title": "Lazy vs hasty: linearization in deep networks impacts learning schedule based on example difficulty",
                        "abstract": "Among attempts at giving a theoretical account of the success of deep neural networks, a recent line of work has identified a so-called lazy training regime in which the network can be well approximated by its linearization around initialization. Here we investigate the comparative effect of the lazy (linear) and feature learning (non-linear) regimes on subgroups of examples based on their difficulty. Specifically, we show that easier examples are given more weight in feature learning mode, resulting in faster training compared to more difficult ones. In other words, the non-linear dynamics tends to sequentialize the learning of examples of increasing difficulty. We illustrate this phenomenon across different ways to quantify example difficulty, including c-score, label noise, and in the presence of easy-to-learn spurious correlations. Our results reveal a new understanding of how deep networks prioritize resources across example difficulty."
                    },
                    {
                        "title": "CrossSplit: Mitigating Label Noise Memorization through Data Splitting",
                        "abstract": "We approach the problem of improving robustness of deep learning algorithms in the presence of label noise. Building upon existing label correction and co-teaching methods, we propose a novel training procedure to mitigate the memorization of noisy labels, called CrossSplit, which uses a pair of neural networks trained on two disjoint parts of the labelled dataset. CrossSplit combines two main ingredients: (i) Cross-split label correction. The idea is that, since the model trained on one part of the data cannot memorize example-label pairs from the other part, the training labels presented to each network can be smoothly adjusted by using the predictions of its peer network; (ii) Cross-split semi-supervised training. A network trained on one part of the data also uses the unlabeled inputs of the other part. Extensive experiments on CIFAR-10, CIFAR-100, Tiny-ImageNet and mini-WebVision datasets demonstrate that our method can outperform the current state-of-the-art in a wide range of noise ratios."
                    },
                    {
                        "title": "Using Representation Expressiveness and Learnability to Evaluate Self-Supervised Learning Methods",
                        "abstract": "We address the problem of evaluating the quality of self-supervised learning (SSL) models without access to supervised labels, while being agnostic to the architecture, learning algorithm or data manipulation used during training. We argue that representations can be evaluated through the lens of expressiveness and learnability. We propose to use the Intrinsic Dimension (ID) to assess expressiveness and introduce Cluster Learnability (CL) to assess learnability. CL is measured in terms of the performance of a KNN classifier trained to predict labels obtained by clustering the representations with K-means. We thus combine CL and ID into a single predictor -- CLID. Through a large-scale empirical study with a diverse family of SSL algorithms, we find that CLID better correlates with in-distribution model performance than other competing recent evaluation schemes. We also benchmark CLID on out-of-domain generalization, where CLID serves as a predictor of the transfer performance of SSL models on several visual classification tasks, yielding improvements with respect to the competing baselines."
                    },
                    {
                        "title": "On the Regularity of Attention",
                        "abstract": "Attention is a powerful component of modern neural networks across a wide variety of domains. In this paper, we seek to quantify the regularity (i.e. the amount of smoothness) of the attention operation. To accomplish this goal, we propose a new mathematical framework that uses measure theory and integral operators to model attention. We show that this framework is consistent with the usual definition, and that it captures the essential properties of attention. Then we use this framework to prove that, on compact domains, the attention operation is Lipschitz continuous and provide an estimate of its Lipschitz constant. Additionally, by focusing on a specific type of attention, we extend these Lipschitz continuity results to non-compact domains. We also discuss the effects regularity can have on NLP models, and applications to invertible and infinitely-deep networks."
                    },
                    {
                        "title": "A Mathematical Theory of Attention",
                        "abstract": "Attention is a powerful component of modern neural networks across a wide variety of domains. However, despite its ubiquity in machine learning, there is a gap in our understanding of attention from a theoretical point of view. We propose a framework to fill this gap by building a mathematically equivalent model of attention using measure theory. With this model, we are able to interpret self-attention as a system of self-interacting particles, we shed light on self-attention from a maximum entropy perspective, and we show that attention is actually Lipschitz-continuous (with an appropriate metric) under suitable assumptions. We then apply these insights to the problem of mis-specified input data; infinitely-deep, weight-sharing self-attention networks; and more general Lipschitz estimates for a specific type of attention studied in concurrent work."
                    },
                    {
                        "title": "Implicit Regularization in Deep Learning: A View from Function Space",
                        "abstract": "We approach the problem of implicit regularization in deep learning from a geometrical viewpoint. We highlight a possible regularization effect induced by a dynamical alignment of the neural tangent features introduced by Jacot et al, along a small number of task-relevant directions. By extrapolating a new analysis of Rademacher complexity bounds in linear models, we propose and study a new heuristic complexity measure for neural networks which captures this phenomenon, in terms of sequences of tangent kernel classes along in the learning trajectories."
                    },
                    {
                        "title": "Implicit Regularization via Neural Feature Alignment",
                        "abstract": "We approach the problem of implicit regularization in deep learning from a geometrical viewpoint. We highlight a regularization effect induced by a dynamical alignment of the neural tangent features introduced by Jacot et al, along a small number of task-relevant directions. This can be interpreted as a combined mechanism of feature selection and model compression. By extrapolating a new analysis of Rademacher complexity bounds for linear models, we motivate and study a heuristic complexity measure that captures this phenomenon, in terms of sequences of tangent kernel classes along the optimization paths."
                    },
                    {
                        "title": "A Modern Take on the Bias-Variance Tradeoff in Neural Networks",
                        "abstract": "The bias-variance tradeoff tells us that as model complexity increases, bias falls and variances increases, leading to a U-shaped test error curve. However, recent empirical results with over-parameterized neural networks are marked by a striking absence of the classic U-shaped test error curve: test error keeps decreasing in wider networks. This suggests that there might not be a bias-variance tradeoff in neural networks with respect to network width, unlike was originally claimed by, e.g., Geman et al. (1992). Motivated by the shaky evidence used to support this claim in neural networks, we measure bias and variance in the modern setting. We find that both bias and variance can decrease as the number of parameters grows. To better understand this, we introduce a new decomposition of the variance to disentangle the effects of optimization and data sampling. We also provide theoretical analysis in a simplified setting that is consistent with our empirical findings."
                    },
                    {
                        "title": "Mutual Information Neural Estimation",
                        "abstract": "We argue that the estimation of mutual information between high dimensional continuous random variables can be achieved by gradient descent over neural networks. We present a Mutual Information Neural Estimator (MINE) that is linearly scalable in dimensionality as well as in sample size, trainable through back-prop, and strongly consistent. We present a handful of applications on which MINE can be used to minimize or maximize mutual information. We apply MINE to improve adversarially trained generative models. We also use MINE to implement Information Bottleneck, applying it to supervised classification; our results demonstrate substantial improvement in flexibility and performance in these settings."
                    },
                    {
                        "title": "On the Spectral Bias of Neural Networks",
                        "abstract": "Neural networks are known to be a class of highly expressive functions able to fit even random input-output mappings with $100\\%$ accuracy. In this work, we present properties of neural networks that complement this aspect of expressivity. By using tools from Fourier analysis, we show that deep ReLU networks are biased towards low frequency functions, meaning that they cannot have local fluctuations without affecting their global behavior. Intuitively, this property is in line with the observation that over-parameterized networks find simple patterns that generalize across data samples. We also investigate how the shape of the data manifold affects expressivity by showing evidence that learning high frequencies gets \\emph{easier} with increasing manifold complexity, and present a theoretical understanding of this behavior. Finally, we study the robustness of the frequency components with respect to parameter perturbation, to develop the intuition that the parameters must be finely tuned to express high frequency functions."
                    },
                    {
                        "title": "A MODERN TAKE",
                        "abstract": "We revisit the bias-variance tradeoff for neural networks in light of modern empirical findings. The traditional bias-variance tradeoff in machine learning suggests that as model complexity grows, variance increases. Classical bounds in statistical learning theory point to the number of parameters in a model as a measure of model complexity, which means the tradeoff would indicate that variance increases with the size of neural networks. However, we empirically find that variance due to training set sampling is roughly constant (with both width and depth) in practice. Variance caused by the non-convexity of the loss landscape is different. We find that it decreases with width and increases with depth, in our setting. We provide theoretical analysis, in a simplified setting inspired by linear models, that is consistent with our empirical findings for width. We view bias-variance as a useful lens to study generalization through and encourage further theoretical explanation from this perspective."
                    },
                    {
                        "title": "MINE: Mutual Information Neural Estimation",
                        "abstract": "This paper presents a Mutual Information Neural Estimator (MINE) that is linearly scalable in dimensionality as well as in sample size. MINE is back-propable and we prove that it is strongly consistent. We illustrate a handful of applications in which MINE is succesfully applied to enhance the property of generative models in both unsupervised and supervised settings. We apply our framework to estimate the information bottleneck, and apply it in tasks related to supervised classification problems. Our results demonstrate substantial added flexibility and improvement in these settings."
                    }
                ]
            },
            "8138da9e-82cc-4a52-869b-ccdb093f833c": {
                "pk": "8138da9e-82cc-4a52-869b-ccdb093f833c",
                "name": "Jonathan Cornford",
                "collaborators": [
                    "Roman Pogodin",
                    "Arna Ghosh",
                    "Kaiwen Sheng",
                    "Brendan A. Bicknell",
                    "Olivier Codol",
                    "Beverley A. Clark",
                    "Guillaume Lajoie",
                    "Blake A. Richards"
                ],
                "domain": [
                    "Computational Neuroscience",
                    "Neural Networks",
                    "Learning Algorithms"
                ],
                "publications": [
                    {
                        "title": "Brain-like learning with exponentiated gradients",
                        "abstract": "Computational neuroscience relies on gradient descent (GD) for training artificial neural network (ANN) models of the brain. The advantage of GD is that it is effective at learning difficult tasks. However, it produces ANNs that are a poor phenomenological fit to biology, making them less relevant as models of the brain. Specifically, it violates Dale\u2019s law, by allowing synapses to change from excitatory to inhibitory, and leads to synaptic weights that are not log-normally distributed, contradicting experimental data. Here, starting from first principles of optimisation theory, we present an alternative learning algorithm, exponentiated gradient (EG), that respects Dale\u2019s Law and produces log-normal weights, without losing the power of learning with gradients. We also show that in biologically relevant settings EG outperforms GD, including learning from sparsely relevant signals and dealing with synaptic pruning. Altogether, our results show that EG is a superior learning algorithm for modelling the brain with ANNs."
                    }
                ]
            },
            "1c353396-983b-4553-b674-3af2f652d466": {
                "pk": "1c353396-983b-4553-b674-3af2f652d466",
                "name": "Stefan Mihalas",
                "collaborators": [
                    "E. Shea-Brown",
                    "Shawn R. Olsen",
                    "Yuhan Helena Liu",
                    "Christof Koch",
                    "C. Bennett",
                    "Samuel D. Gale",
                    "D. Mastrovito",
                    "Marina Garrett",
                    "M. Buice",
                    "Brian Hu",
                    "Hongkui Zeng",
                    "Kyle Aitken",
                    "Doris Voina",
                    "Stephen Smith",
                    "U. S\u00fcmb\u00fcl",
                    "Dana Mastrovito",
                    "Lukasz Kusmierz",
                    "Peter A. Groblewski",
                    "Alex T. Piet",
                    "D. Ollerenshaw",
                    "I. Yavorska",
                    "Farzaneh Najafi",
                    "Adam Amster",
                    "Shiella D. Caldejon",
                    "Linzy Casal",
                    "Florence D. D'Orazi",
                    "Scott Daniel",
                    "Saskia E. J. de Vries",
                    "D. Kapner",
                    "Justin Kiggins",
                    "J\u00e9r\u00f4me A. Lecoq",
                    "P. Ledochowitsch",
                    "Sahar Manavi",
                    "N. Mei",
                    "Christopher B. Morrison",
                    "Sarah R. Naylor",
                    "N. Orlova",
                    "Jed Perkins",
                    "Nicholas D. Ponvert",
                    "Clark Roll",
                    "Sam Seid",
                    "Derric Williams",
                    "A. Williford",
                    "Ruweida Ahmed",
                    "Daniel Amine",
                    "Yazan N. Billeh",
                    "C. Bowman",
                    "Nicholas Cain",
                    "Andrew Cho",
                    "Tim Dawe",
                    "Max Departee",
                    "M. Desoto",
                    "David Feng",
                    "Emily Gelfand",
                    "Nile Gradis",
                    "Conor Grasso",
                    "Nicole Hancock",
                    "Ross Hytnen",
                    "Xiaoxuan Jia",
                    "Tye Johnson",
                    "India Kato",
                    "Sara Kivikas",
                    "L. Kuan",
                    "Q. L\u2019Heureux",
                    "Sophie Lambert",
                    "Arielle Leon",
                    "Elizabeth Liang",
                    "Fuhui Long",
                    "Kyla Mace",
                    "Ildefons Magrans de Abril",
                    "Chris Mochizuki",
                    "Chelsea Nayan",
                    "Kat North",
                    "Lydia Ng",
                    "G. K. Ocker",
                    "M. Oliver",
                    "Paul A. Rhoads",
                    "K. Ronellenfitch",
                    "Kathryn Schelonka",
                    "Josh Sevigny",
                    "David Sullivan",
                    "Ben B. Sutton",
                    "Jackie Swapp",
                    "Thuyanh K Nguyen",
                    "Xana Waughman",
                    "Joshua Wilkes",
                    "Michael Wang",
                    "Colin Farrell",
                    "Wayne Wakeman",
                    "John W. Phillips",
                    "A. Arkhipov",
                    "Koosha Khalvati",
                    "S. Johnson",
                    "Luke Campagnola",
                    "Samson Koelle",
                    "Jennifer D. Whitesell",
                    "Karla E. Hirokawa",
                    "M. Meil\u0103",
                    "Julie A. Harris",
                    "Praveen Venkatesh"
                ],
                "domain": [
                    "Neuroscience",
                    "Neural Networks",
                    "Learning Mechanisms",
                    "Computational Modeling"
                ],
                "publications": [
                    {
                        "title": "Transition to chaos separates learning regimes and relates to measure of consciousness in recurrent neural networks",
                        "abstract": "Recurrent neural networks exhibit chaotic dynamics when the variance in their connection strengths exceed a critical value. Recent work indicates connection variance also modulates learning strategies; networks learn \u201drich\u201d representations when initialized with low coupling and \u201dlazier\u201dsolutions with larger variance. Using Watts-Strogatz networks of varying sparsity, structure, and hidden weight variance, we find that the critical coupling strength dividing chaotic from ordered dynamics also differentiates rich and lazy learning strategies. Training moves both stable and chaotic networks closer to the edge of chaos, with networks learning richer representations before the transition to chaos. In contrast, biologically realistic connectivity structures foster stability over a wide range of variances. The transition to chaos is also reflected in a measure that clinically discriminates levels of consciousness, the perturbational complexity index (PCIst). Networks with high values of PCIst exhibit stable dynamics and rich learning, suggesting a consciousness prior may promote rich learning. The results suggest a clear relationship between critical dynamics, learning regimes and complexity-based measures of consciousness."
                    },
                    {
                        "title": "Stimulus novelty uncovers coding diversity in visual cortical circuits",
                        "abstract": "The detection of novel stimuli is critical to learn and survive in a dynamic environment. Though novel stimuli powerfully affect brain activity, their impact on specific cell types and circuits is not well understood. Disinhibition is one candidate mechanism for novelty-induced enhancements in activity. Here we characterize the impact of stimulus novelty on disinhibitory circuit components using longitudinal 2-photon calcium imaging of Vip, Sst, and excitatory populations in the mouse visual cortex. Mice learn a behavioral task with stimuli that become highly familiar, then are tested on both familiar and novel stimuli. Mice consistently perform the task with novel stimuli, yet responses to stimulus presentations and stimulus omissions are dramatically altered. Further, we find that novelty modifies coding of visual as well as behavioral and task information. At the population level, the direction of these changes is consistent with engagement of the Vip-Sst disinhibitory circuit. At the single cell level, we identify separate clusters of Vip, Sst, and excitatory cells with unique patterns of novelty-induced coding changes. This study and the accompanying open-access dataset reveals the impact of novelty on sensory and behavioral representations in visual cortical circuits and establishes novelty as a key driver of cellular functional diversity."
                    },
                    {
                        "title": "Efficient approximation of neural population structure and correlations with probabilistic circuits",
                        "abstract": "We present a computationally efficient framework to model a wide range of population structures with high order correlations and a large number of neurons. Our method is based on a special type of Bayesian network that has linear inference time and is founded upon the concept of contextual independence. Our framework is both fast and accurate in approximating neural population structures. Furthermore, our approach enables us to reliably quantify higher order neural correlations. We test our method on simulated neural populations commonly used to generate higher order correlations, as well as on publicly available large-scale neural recordings from the Allen Brain Observatory. Our approach significantly outperforms other models both in terms of statistical measures and alignment with experimental evidence."
                    },
                    {
                        "title": "Simple synaptic modulations implement diverse novelty computations",
                        "abstract": "Since environments are constantly in flux, the brain\u2019s ability to identify novel stimuli that fall outside its own internal representation of the world is crucial for an organism\u2019s survival. Within the mammalian neocortex, inhibitory microcircuits are proposed to regulate activity in an experience-dependent manner and different inhibitory neuron subtypes exhibit distinct novelty responses. Discerning the function of diverse neural circuits and their modulation by experience can be daunting unless one has a biologically plausible mechanism to detect and learn from novel experiences that is both understandable and flexible. Here we introduce a learning mechanism, familiarity modulated synapses (FMSs), through which a network response that encodes novelty emerges from unsupervised multiplicative synaptic modifications depending only on the presynaptic or both the pre- and postsynaptic activity. FMSs stand apart from other familiarity mechanisms in their simplicity: they operate under continual learning, do not require specialized architecture, and can distinguish novelty rapidly without requiring feedback. Implementing FMSs within an experimentally-constrained model of a visual cortical circuit, we demonstrate the generalizability of FMSs by reproducing three distinct novelty effects recently observed in experiments: absolute, contextual (or oddball), and omission novelty. Additionally, our model reproduces functional diversity within cell subpopulations, leading to experimentally testable predictions about connectivity and synaptic dynamics that can produce both population-level novelty responses and heterogeneous individual neuron signals. Altogether, our findings demonstrate how simple plasticity mechanisms within the cortical circuit structure can give rise to qualitatively distinct novelty responses. The flexibility of FMSs opens the door to computationally and theoretically investigating how distinct synapse modulations can lead to a variety of experience-dependent responses in a simple, understandable, and biologically plausible setup."
                    },
                    {
                        "title": "Modeling the cell-type-specific mesoscale murine connectome with anterograde tracing experiments",
                        "abstract": "The Allen Mouse Brain Connectivity Atlas (MCA) consists of anterograde tracing experiments targeting diverse structures and classes of projecting neurons. Beyond regional anterograde tracing done in C57BL/6 wild type mice, a large fraction of experiments are performed using transgenic Cre-lines. This allows access to cell-class specific whole brain connectivity information, with class defined by the transgenic lines. However, even though the number of experiments is large, it does not come close to covering all existing cell classes in every area where they exist. Here, we study how much we can fill in these gaps and estimate the cell-class specific connectivity function given the simplifying assumptions that nearby voxels have smoothly varying projections, but that these projection tensors can change sharply depending on the region and class of the projecting cells. This paper describes the conversion of Cre-line tracer experiments into class-specific connectivity matrices representing the connection strengths between source and target structures. We introduce and validate a novel statistical model for creation of connectivity matrices. We extend the Nadaraya-Watson kernel learning method which we previously used to fill in spatial gaps to also fill in a gaps in cell-class connectivity information. To do this, we construct a \"cell-class space\" based on class-specific averaged regionalized projections and combine smoothing in 3D space as well as in this abstract space to share information between similar neuron classes. Using this method we construct a set of connectivity matrices using multiple levels of resolution at which discontinuities in connectivity are assumed. We show that the connectivities obtained from this model display expected cell-type and structure specific connectivities. We also show that the wild type connectivity matrix can be factored using a sparse set of factors, and analyze the informativeness of this latent variable model. AUTHOR SUMMARY Large-scale studies have described the connections between areas in multiple mammalian models in ever expanding detail. Standard connectivity studies focus on the connection strength between areas. However, when describing functions at a local circuit level, there is an increasing focus on cell types. We have recently described the importance of connection types in the cortico-thalamic system, which allows an unsupervised discovery of its hierarchical organization. In this study we focus on adding a dimension of connection type for a brain-wide mesoscopic connectivity model. Even with our relatively massive dataset, the data in the cell type direction for connectivity is quite sparse, and we had to develop methods to more reliably extrapolate in such directions, and to estimate when such extrapolations are impossible. This allows us to fill in such a connection type specific inter-areal connectivity matrix to the extent our data allows. While analyzing this complex connectivity, we observed that it can be described via a small set of factors. While not complete, this connectivity matrix represents a a categorical and quantitative improvement in mouse mesoscale connectivity models."
                    },
                    {
                        "title": "Gaussian Partial Information Decomposition: Bias Correction and Application to High-dimensional Data",
                        "abstract": "Recent advances in neuroscientific experimental techniques have enabled us to simultaneously record the activity of thousands of neurons across multiple brain regions. This has led to a growing need for computational tools capable of analyzing how task-relevant information is represented and communicated between several brain regions. Partial information decompositions (PIDs) have emerged as one such tool, quantifying how much unique, redundant and synergistic information two or more brain regions carry about a task-relevant message. However, computing PIDs is computationally challenging in practice, and statistical issues such as the bias and variance of estimates remain largely unexplored. In this paper, we propose a new method for efficiently computing and estimating a PID definition on multivariate Gaussian distributions. We show empirically that our method satisfies an intuitive additivity property, and recovers the ground truth in a battery of canonical examples, even at high dimensionality. We also propose and evaluate, for the first time, a method to correct the bias in PID estimates at finite sample sizes. Finally, we demonstrate that our Gaussian PID effectively characterizes inter-areal interactions in the mouse brain, revealing higher redundancy between visual areas when a stimulus is behaviorally relevant."
                    },
                    {
                        "title": "Single Circuit in V1 Capable of Switching Contexts During Movement Using an Inhibitory Population as a Switch",
                        "abstract": "Abstract As animals adapt to their environments, their brains are tasked with processing stimuli in different sensory contexts. Whether these computations are context dependent or independent, they are all implemented in the same neural tissue. A crucial question is what neural architectures can respond flexibly to a range of stimulus conditions and switch between them. This is a particular case of flexible architecture that permits multiple related computations within a single circuit. Here, we address this question in the specific case of the visual system circuitry, focusing on context integration, defined as the integration of feedforward and surround information across visual space. We show that a biologically inspired microcircuit with multiple inhibitory cell types can switch between visual processing of the static context and the moving context. In our model, the VIP population acts as the switch and modulates the visual circuit through a disinhibitory motif. Moreover, the VIP population is efficient, requiring only a relatively small number of neurons to switch contexts. This circuit eliminates noise in videos by using appropriate lateral connections for contextual spatiotemporal surround modulation, having superior denoising performance compared to circuits where only one context is learned. Our findings shed light on a minimally complex architecture that is capable of switching between two naturalistic contexts using few switching units."
                    },
                    {
                        "title": "MouseNet: A biologically constrained convolutional neural network model for the mouse visual cortex",
                        "abstract": "Convolutional neural networks trained on object recognition derive inspiration from the neural architecture of the visual system in mammals, and have been used as models of the feedforward computation performed in the primate ventral stream. In contrast to the deep hierarchical organization of primates, the visual system of the mouse has a shallower arrangement. Since mice and primates are both capable of visually guided behavior, this raises questions about the role of architecture in neural computation. In this work, we introduce a novel framework for building a biologically constrained convolutional neural network model of the mouse visual cortex. The architecture and structural parameters of the network are derived from experimental measurements, specifically the 100-micrometer resolution interareal connectome, the estimates of numbers of neurons in each area and cortical layer, and the statistics of connections between cortical layers. This network is constructed to support detailed task-optimized models of mouse visual cortex, with neural populations that can be compared to specific corresponding populations in the mouse brain. Using a well-studied image classification task as our working example, we demonstrate the computational capability of this mouse-sized network. Given its relatively small size, MouseNet achieves roughly 2/3rds the performance level on ImageNet as VGG16. In combination with the large scale Allen Brain Observatory Visual Coding dataset, we use representational similarity analysis to quantify the extent to which MouseNet recapitulates the neural representation in mouse visual cortex. Importantly, we provide evidence that optimizing for task performance does not improve similarity to the corresponding biological system beyond a certain point. We demonstrate that the distributions of some physiological quantities are closer to the observed distributions in the mouse brain after task training. We encourage the use of the MouseNet architecture by making the code freely available."
                    },
                    {
                        "title": "Neural population dynamics of computing with synaptic modulations",
                        "abstract": "In addition to long-time scale rewiring, synapses in the brain are subject to significant modulation that occurs at much shorter time scales and allows them to process short-term information. Despite this, models of the brain like recurrent neural networks (RNNs) often have their weights frozen after training, relying on an internal state stored in neuron activity to process temporal information. Although networks with dynamical synapses have been explored previously, often said dynamics are added to networks that also have recurrent connections and thus the short-time scale computational capabilities of synapse modulation alone remain unclear. In this work, we analyze the dynamics of a network that relies solely on synaptic modulations to process short-time scale information, the multi-plasticity network (MPN). We thoroughly examine the neural population dynamics of the MPN trained on integration-based tasks and compare it to known RNN dynamics, findings the two to have fundamentally different behavior and attractor structure. We find said differences in dynamics allow the MPN to outperform its RNN counterparts on several neuroscience-relevant tasks. Of note, the MPN has a significantly simpler attractor structure that allows it to be more flexible in training and sequential-learning settings. Lastly, how the dynamics change for MPNs trained on contextual and continuous integration tasks is also investigated."
                    },
                    {
                        "title": "Biologically-plausible backpropagation through arbitrary timespans via local neuromodulators",
                        "abstract": "The spectacular successes of recurrent neural network models where key parameters are adjusted via backpropagation-based gradient descent have inspired much thought as to how biological neuronal networks might solve the corresponding synaptic credit assignment problem. There is so far little agreement, however, as to how biological networks could implement the necessary backpropagation through time, given widely recognized constraints of biological synaptic network signaling architectures. Here, we propose that extra-synaptic diffusion of local neuromodulators such as neuropeptides may afford an effective mode of backpropagation lying within the bounds of biological plausibility. Going beyond existing temporal truncation-based gradient approximations, our approximate gradient-based update rule, ModProp, propagates credit information through arbitrary time steps. ModProp suggests that modulatory signals can act on receiving cells by convolving their eligibility traces via causal, time-invariant and synapse-type-specific filter taps. Our mathematical analysis of ModProp learning, together with simulation results on benchmark temporal tasks, demonstrate the advantage of ModProp over existing biologically-plausible temporal credit assignment rules. These results suggest a potential neuronal mechanism for signaling credit information related to recurrent interactions over a longer time horizon. Finally, we derive an in-silico implementation of ModProp that could serve as a low-complexity and causal alternative to backpropagation through time."
                    },
                    {
                        "title": "Heterogeneity in Neuronal Dynamics is Learned by Gradient Descent for Temporal Processing Tasks",
                        "abstract": "Individual neurons in the brain have complex intrinsic dynamics that are highly diverse. We hypothesize that the complex dynamics produced by networks of complex and heterogeneous neurons may contribute to the brain\u2019s ability to process and respond to temporally complex data. To study the role of complex and heterogeneous neuronal dynamics in network computation, we develop a rate-based neuronal model, the generalized-leaky-integrate-and-firing-rate (GLIFR) model, which is a rate-equivalent of the generalized-leaky-integrate-and-fire model. The GLIFR model has multiple dynamical mechanisms which add to the complexity of its activity while maintaining differentiability. We focus on the role of after-spike currents, currents induced or modulated by neuronal spikes, in producing rich temporal dynamics. We use machine learning techniques to learn both synaptic weights and parameters underlying intrinsic dynamics to solve temporal tasks. The GLIFR model allows us to use standard gradient descent techniques rather than surrogate gradient descent, which has been utilized in spiking neural networks. After establishing the ability to optimize parameters using gradient descent in single neurons, we ask how networks of GLIFR neurons learn and perform on temporally challenging tasks, such as sinusoidal pattern generation and sequential MNIST. We find that these networks learn a diversity of parameters, which gives rise to diversity in neuronal dynamics. We also observe that training networks on the sequential MNIST task leads to formation of cell classes based on the clustering of neuronal parameters. GLIFR networks have mixed performance when compared to vanilla recurrent neural networks but appear to be more robust to random silencing. When we explore these performance gains further, we find that both the ability to learn heterogeneity and the presence of after-spike currents contribute. Our work both demonstrates the computational robustness of neuronal complexity and diversity in networks and demonstrates a feasible method of training such models using exact gradients."
                    },
                    {
                        "title": "Mesolimbic dopamine release conveys causal associations",
                        "abstract": "Learning to predict rewards based on environmental cues is essential for survival. It is believed that animals learn to predict rewards by updating predictions whenever the outcome deviates from expectations, and that such reward prediction errors (RPEs) are signaled by the mesolimbic dopamine system\u2014a key controller of learning. However, instead of learning prospective predictions from RPEs, animals can infer predictions by learning the retrospective cause of rewards. Hence, whether mesolimbic dopamine instead conveys a causal associative signal that sometimes resembles RPE remains unknown. We developed an algorithm for retrospective causal learning and found that mesolimbic dopamine release conveys causal associations but not RPE, thereby challenging the dominant theory of reward learning. Our results reshape the conceptual and biological framework for associative learning. Description A framework for associative learning How do animals learn to associate environmental cues with delayed outcomes such as rewards? Temporal difference reinforcement learning (TDRL) is the most widely accepted model of reward learning. Jeong et al. now challenge this view by introducing a formal account and model of dopamine\u2019s role in learning. They present an algorithm that supports retrospective causal learning, allowing animals to infer the causes of an event. To directly compare the two competing hypotheses, the authors devised eight experimental tests. In each of the eight tests in every animal tested, they challenge the TDRL signaling model and show that every observation is consistent with their causal inference algorithm. These findings provide a fresh physiological and theoretical framework for associative learning in the brain. \u2014PRS Animals learns retrospectively that there are causal associations of conditioned stimuli and reward. INTRODUCTION How do animals learn to prospectively predict delayed outcomes such as rewards from environmental cues? Considerable evidence shows that the neuromodulator dopamine is critical for such associative learning. This evidence has supported the hypothesis that midbrain dopamine neurons and nucleus accumbens dopamine release signal a temporal difference (TD) reward prediction error (RPE)\u2014the difference between received and predicted reward. Hence, it is widely believed that animals use the TD algorithm to learn reward predictions. Recent results, however, suggest that dopamine signaling may not be fully consistent with RPEs. Despite this, the TDRPE hypothesis remains the best explanation of associative learning in the brain because no alternate model exists that successfully explains experimental observations inconsistent with TDRPE while also capturing experimental phenomena currently best explained by TDRPE. RATIONALE Here, we propose a new theory of associative learning. Since causes must precede outcomes, we propose that animals learn associations by looking back in memory to infer causes of meaningful outcomes such as rewards. For instance, imagine a child learning that the bell of an ice cream van predicts ice cream. Learning what cues precede ice cream (retrospective) is simpler than learning the future outcome of a near-infinite number of cues, only a few of which may result in ice cream (prospective). Thus, once a meaningful outcome such as ice cream is realized, the child can look back in memory to infer that the bell causes ice cream. Such learning only requires a memory of the previous bell when ice cream is received, and not active maintenance of predictions of ice cream prior to its receipt. We developed a model based on this concept\u2014adjusted net contingency for causal relations (ANCCR, pronounced \u201canchor\u201d)\u2014and tested whether this model is capable of learning reward predictions and explaining nucleus accumbens dopamine release better than TDRPE. In this model, a current event is considered a \u201cmeaningful causal target\u201d (i.e., an event whose cause should be learned) when it is either innately meaningful (e.g., ice cream) or when it has been learned to be a cause of other meaningful causal targets (e.g., bell of ice cream van after it has been learned to be a cause of ice cream). Our central hypothesis is that mesolimbic dopamine conveys the learned meaningfulness of events by signaling ANCCR and guides causal learning of the current event. RESULTS We first showed using simulations that ANCCR identifies relationships between events spread in time (e.g., cues and rewards) better than TD without loss of generality across timescales. Next, we showed that many classic experimental results supporting TDRPE coding are also consistent with ANCCR signaling. Thus, we sought to test dopamine activity during experiments that distinguish these two hypotheses. To this end, we devised eleven experimental tests that are capable of qualitatively distinguishing between both hypotheses. We performed these tests by using fiber photometry of an optical dopamine sensor (dLight1.3b) to measure nucleus accumbens dopamine release. We did so across tasks including repeated exposure to random rewards, variations of cue-reward learning, and sequential conditioning with and without optogenetic inhibition of dopamine release. In all cases, we observed that the results were inconsistent with TDRPE but consistent with ANCCR. CONCLUSION Collectively, these results demonstrate that dopamine function can be better understood as providing a signal to initiate learning of causes of a meaningful stimulus. These results reshape current understanding of the algorithms and neural mechanisms of associative learning in the brain. Retrospective causal learning of meaningful outcomes. How do we learn that environmental stimuli (e.g., bell of ice cream van) predict rewards (e.g., ice cream)? The standard model learns prospectively using prediction errors. Here, we present an alternative that learns predictions using retrospective identification of causes of meaningful outcomes. These models make different predictions about mesolimbic dopamine, a critical controller of learning. Our data support retrospective causal learning."
                    },
                    {
                        "title": "Cell-type-speci\ufb01c modulatory signaling promotes learning in spiking neural networks",
                        "abstract": "Animals learn and form memories by jointly adjusting the e\ufb03cacy of their synapses. How they e\ufb03ciently solve the underlying temporal credit assignment problem remains elusive. Here, we re-analyze the mathematical basis of gradient descent learning in recurrent spiking neural networks (RSNNs) in light of the recent single-cell transcriptomic evidence for cell-type-speci\ufb01c local neuropeptide signaling in the cortex. Our normative theory posits an important role for the notion of neuronal cell types and local di\ufb00usive communication by enabling biologically plausible and e\ufb03cient weight update. While obeying fundamental biological constraints, including separating excitatory vs inhibitory cell types and observing connection sparsity, we trained RSNNs for temporal credit assignment tasks spanning seconds and observed that the inclusion of local modulatory signaling improved learning e\ufb03ciency. Our learning rule puts forth a novel form of interaction between modulatory signals and synaptic transmission. Moreover, it suggests a computationally e\ufb03cient learning method for bio-inspired arti\ufb01cial intelligence."
                    },
                    {
                        "title": "Cell-type\u2013specific neuromodulation guides synaptic credit assignment in a spiking neural network",
                        "abstract": "Significance Synaptic connectivity provides the foundation for our present understanding of neuronal network function, but static connectivity cannot explain learning and memory. We propose a computational role for the diversity of cortical neuronal types and their associated cell-type\u2013specific neuromodulators in improving the efficiency of synaptic weight adjustments for task learning in neuronal networks. Brains learn tasks via experience-driven differential adjustment of their myriad individual synaptic connections, but the mechanisms that target appropriate adjustment to particular connections remain deeply enigmatic. While Hebbian synaptic plasticity, synaptic eligibility traces, and top-down feedback signals surely contribute to solving this synaptic credit-assignment problem, alone, they appear to be insufficient. Inspired by new genetic perspectives on neuronal signaling architectures, here, we present a normative theory for synaptic learning, where we predict that neurons communicate their contribution to the learning outcome to nearby neurons via cell-type\u2013specific local neuromodulation. Computational tests suggest that neuron-type diversity and neuron-type\u2013specific local neuromodulation may be critical pieces of the biological credit-assignment puzzle. They also suggest algorithms for improved artificial neural network learning efficiency."
                    },
                    {
                        "title": "A biologically inspired architecture with switching units can learn to generalize across backgrounds",
                        "abstract": "Humans and other animals navigate different landscapes and environments with ease, a feat that requires the brain\u2019s ability to rapidly and accurately adapt to different visual domains, generalizing across contexts/backgrounds. Despite recent progress in deep learning applied to classification and detection in the presence of multiple confounds including contextual ones [25, 30], there remain important challenges to address regarding how networks can perform context-dependent computations and how contextually-invariant visual concepts are formed. For instance, recent studies have shown artificial networks that repeatedly misclassified familiar objects set on new backgrounds, e.g. incorrectly labelling known animals when they appeared in a different setting [3]. Here, we show how a bio-inspired network motif can explicitly address this issue. We do this using a novel dataset which can be used as a benchmark for future studies probing invariance to backgrounds. The dataset consists of MNIST digits of varying transparency, set on one of two backgrounds with different statistics: a Gaussian noise or a more naturalistic background from the CIFAR-10 dataset. We use this dataset to learn digit classification when contexts are shown sequentially, and find that both shallow and deep networks have sharply decreased performance when returning to the first background after experience learning the second \u2013 the catastrophic forgetting phenomenon in continual learning. To overcome this, we propose an architecture with additional \u201c switching\u201d units that are activated in the presence of a new background. We find that the switching network can learn the new context even with very few switching units, while maintaining the performance in the previous context \u2013 but that they must be recurrently connected to network layers. When the task is difficult due to high transparency, the switching network trained on both contexts outperforms networks without switching trained on only one context. The switching mechanism leads to sparser activation patterns, and we provide intuition for why this helps to solve the task. We compare our architecture with other prominent learning methods, and find that elastic weight consolidation is not successful in our setting, while progressive nets are more complex but less effective. Our study therefore shows how a bio-inspired architectural motif can contribute to task generalization across context."
                    },
                    {
                        "title": "Geometry of inter-areal interactions in mouse visual cortex",
                        "abstract": "The response of a set of neurons in an area is the result of the sensory input, the interaction of the neurons within the area as well as the long range interactions between areas. We aimed to study the relation between interactions among multiple areas, and if they are fixed or dynamic. The structural connectivity provides a substrate for these interactions, but anatomical connectivity is not known in sufficient detail and it only gives us a static picture. Using the Allen Brain Observatory Visual Coding Neuropixels dataset, which includes simultaneous recordings of spiking activity from up to 6 hierarchically organized mouse cortical visual areas, we estimate the functional connectivity between neurons using a linear model of responses to flashed static grating stimuli. We characterize functional connectivity between populations via interaction subspaces. We find that distinct subspaces of a source area mediate interactions with distinct target areas, supporting the notion that cortical areas use distinct channels to communicate. Most importantly, using a piecewise linear model for activity within each trial, we find that these interactions evolve dynamically over tens of milliseconds following a stimulus presentation. Inter-areal subspaces become more aligned with the intra-areal subspaces during epochs in which a feedforward wave of activity propagates through visual cortical areas. When the short-term dynamics are averaged over, we find that the interaction subspaces are stable over multiple stimulus blocks. These findings have important implications for understanding how information flows through biological neural networks composed of interconnected modules, each of which may have a distinct functional specialization."
                    },
                    {
                        "title": "A Biologically Plausible Model for Continual Learning using Synaptic Weight Attractors",
                        "abstract": "The human brain readily learns tasks in sequence without forgetting previous ones. Artificial neural networks (ANNs), on the other hand, need to be modified to achieve similar performance. While effective, many algorithms that accomplish this are based on weight importance methods that do not correspond to biological mechanisms. Here we introduce a simple, biologically plausible method for enabling effective continual learning in ANNs. We show that it is possible to learn a weight-dependent plasticity function that prevents catastrophic forgetting over multiple tasks. We highlight the effectiveness of our method by evaluating it on a set of MNIST classification tasks. We further find that the use of our method promotes synaptic multi-modality, similar to that seen in biology."
                    },
                    {
                        "title": "Backward masking in mice requires visual cortex",
                        "abstract": "The authors introduce a novel visual masking task and use recordings and optogenetics to reveal the role of visual cortex. Visual masking can reveal the timescale of perception, but the underlying circuit mechanisms are not understood. Here we describe a backward masking task in mice and humans in which the location of a stimulus is potently masked. Humans report reduced subjective visibility that tracks behavioral deficits. In mice, both masking and optogenetic silencing of visual cortex (V1) reduce performance over a similar timecourse but have distinct effects on response rates and accuracy. Activity in V1 is consistent with masked behavior when quantified over long, but not short, time windows. A dual accumulator model recapitulates both mouse and human behavior. The model and subjects\u2019 performance imply that the initial spikes in V1 can trigger a correct response, but subsequent V1 activity degrades performance. Supporting this hypothesis, optogenetically suppressing mask-evoked activity in V1 fully restores accurate behavior. Together, these results demonstrate that mice, like humans, are susceptible to masking and that target and mask information is first confounded downstream of V1."
                    }
                ]
            },
            "b713cf68-212e-4915-9630-1a07701a51a9": {
                "pk": "b713cf68-212e-4915-9630-1a07701a51a9",
                "name": "Eric Shea-Brown",
                "collaborators": [
                    "Stefan Mihalas",
                    "Yuhan Helena Liu",
                    "Jianghong Shi",
                    "Shirui Chen",
                    "Stefano Recanatesi",
                    "Doris Voina",
                    "B. Tripp",
                    "M. Buice",
                    "Stephen Smith",
                    "U. S\u00fcmb\u00fcl",
                    "Dana Mastrovito",
                    "Lukasz Kusmierz",
                    "Christof Koch",
                    "James Hazelden",
                    "Eli Shlizerman",
                    "Linxing Jiang",
                    "Rajesh P. N. Rao",
                    "Brian Hu",
                    "Arna Ghosh",
                    "B. Richards",
                    "Guillaume Lajoie",
                    "Michael A Buice",
                    "Uygar Sumbul",
                    "Chloe N. Winston",
                    "D. Mastrovito",
                    "Julien A. Bloch",
                    "Alexander Greaves-Tunnell",
                    "Za\u00efd Harchaoui",
                    "A. Shojaie",
                    "A. Yazdan-Shahmorad",
                    "Stephen J. Smith"
                ],
                "domain": [
                    "Neuroscience",
                    "Deep Learning",
                    "Recurrent Neural Networks",
                    "Computational Models"
                ],
                "publications": [
                    {
                        "title": "Transition to chaos separates learning regimes and relates to measure of consciousness in recurrent neural networks",
                        "abstract": "Recurrent neural networks exhibit chaotic dynamics when the variance in their connection strengths exceed a critical value. Recent work indicates connection variance also modulates learning strategies; networks learn \u201drich\u201d representations when initialized with low coupling and \u201dlazier\u201dsolutions with larger variance. Using Watts-Strogatz networks of varying sparsity, structure, and hidden weight variance, we find that the critical coupling strength dividing chaotic from ordered dynamics also differentiates rich and lazy learning strategies. Training moves both stable and chaotic networks closer to the edge of chaos, with networks learning richer representations before the transition to chaos. In contrast, biologically realistic connectivity structures foster stability over a wide range of variances. The transition to chaos is also reflected in a measure that clinically discriminates levels of consciousness, the perturbational complexity index (PCIst). Networks with high values of PCIst exhibit stable dynamics and rich learning, suggesting a consciousness prior may promote rich learning. The results suggest a clear relationship between critical dynamics, learning regimes and complexity-based measures of consciousness."
                    },
                    {
                        "title": "A simple connection from loss flatness to compressed representations in neural networks",
                        "abstract": "The generalization capacity of deep neural networks has been studied in a variety of ways, including at least two distinct categories of approaches: one based on the shape of the loss landscape in parameter space, and the other based on the structure of the representation manifold in feature space (that is, in the space of unit activities). Although these two approaches are related, they are rarely studied together explicitly. Here, we present an analysis that bridges this gap. We show that in the final phase of learning in deep neural networks, the compression of the manifold of neural representations correlates with the flatness of the loss around the minima explored by SGD. This correlation is predicted by a relatively simple mathematical relationship: a flatter loss corresponds to a lower upper bound on the compression metrics of neural representations. Our work builds upon the linear stability insight by Ma and Ying, deriving inequalities between various compression metrics and quantities involving sharpness. Empirically, our derived inequality predicts a consistently positive correlation between representation compression and loss sharpness in multiple experimental settings. Overall, we advance a dual perspective on generalization in neural networks in both parameter and feature space."
                    },
                    {
                        "title": "Evolutionary algorithms as an alternative to backpropagation for supervised training of Biophysical Neural Networks and Neural ODEs",
                        "abstract": "Training networks consisting of biophysically accurate neuron models could allow for new insights into how brain circuits can organize and solve tasks. We begin by analyzing the extent to which the central algorithm for neural network learning -- stochastic gradient descent through backpropagation (BP) -- can be used to train such networks. We find that properties of biophysically based neural network models needed for accurate modelling such as stiffness, high nonlinearity and long evaluation timeframes relative to spike times makes BP unstable and divergent in a variety of cases. To address these instabilities and inspired by recent work, we investigate the use of \"gradient-estimating\" evolutionary algorithms (EAs) for training biophysically based neural networks. We find that EAs have several advantages making them desirable over direct BP, including being forward-pass only, robust to noisy and rigid losses, allowing for discrete loss formulations, and potentially facilitating a more global exploration of parameters. We apply our method to train a recurrent network of Morris-Lecar neuron models on a stimulus integration and working memory task, and show how it can succeed in cases where direct BP is inapplicable. To expand on the viability of EAs in general, we apply them to a general neural ODE problem and a stiff neural ODE benchmark and find again that EAs can out-perform direct BP here, especially for the over-parameterized regime. Our findings suggest that biophysical neurons could provide useful benchmarks for testing the limits of BP-adjacent methods, and demonstrate the viability of EAs for training networks with complex components."
                    },
                    {
                        "title": "Expressive probabilistic sampling in recurrent neural networks",
                        "abstract": "In sampling-based Bayesian models of brain function, neural activities are assumed to be samples from probability distributions that the brain uses for probabilistic computation. However, a comprehensive understanding of how mechanistic models of neural dynamics can sample from arbitrary distributions is still lacking. We use tools from functional analysis and stochastic differential equations to explore the minimum architectural requirements for $\\textit{recurrent}$ neural circuits to sample from complex distributions. We first consider the traditional sampling model consisting of a network of neurons whose outputs directly represent the samples (sampler-only network). We argue that synaptic current and firing-rate dynamics in the traditional model have limited capacity to sample from a complex probability distribution. We show that the firing rate dynamics of a recurrent neural circuit with a separate set of output units can sample from an arbitrary probability distribution. We call such circuits reservoir-sampler networks (RSNs). We propose an efficient training procedure based on denoising score matching that finds recurrent and output weights such that the RSN implements Langevin sampling. We empirically demonstrate our model's ability to sample from several complex data distributions using the proposed neural dynamics and discuss its applicability to developing the next generation of sampling-based brain models."
                    },
                    {
                        "title": "Single Circuit in V1 Capable of Switching Contexts During Movement Using an Inhibitory Population as a Switch",
                        "abstract": "Abstract As animals adapt to their environments, their brains are tasked with processing stimuli in different sensory contexts. Whether these computations are context dependent or independent, they are all implemented in the same neural tissue. A crucial question is what neural architectures can respond flexibly to a range of stimulus conditions and switch between them. This is a particular case of flexible architecture that permits multiple related computations within a single circuit. Here, we address this question in the specific case of the visual system circuitry, focusing on context integration, defined as the integration of feedforward and surround information across visual space. We show that a biologically inspired microcircuit with multiple inhibitory cell types can switch between visual processing of the static context and the moving context. In our model, the VIP population acts as the switch and modulates the visual circuit through a disinhibitory motif. Moreover, the VIP population is efficient, requiring only a relatively small number of neurons to switch contexts. This circuit eliminates noise in videos by using appropriate lateral connections for contextual spatiotemporal surround modulation, having superior denoising performance compared to circuits where only one context is learned. Our findings shed light on a minimally complex architecture that is capable of switching between two naturalistic contexts using few switching units."
                    },
                    {
                        "title": "Beyond accuracy: generalization properties of bio-plausible temporal credit assignment rules",
                        "abstract": "To unveil how the brain learns, ongoing work seeks biologically-plausible approximations of gradient descent algorithms for training recurrent neural networks (RNNs). Yet, beyond task accuracy, it is unclear if such learning rules converge to solutions that exhibit different levels of generalization than their nonbiologically-plausible counterparts. Leveraging results from deep learning theory based on loss landscape curvature, we ask: how do biologically-plausible gradient approximations affect generalization? We first demonstrate that state-of-the-art biologically-plausible learning rules for training RNNs exhibit worse and more variable generalization performance compared to their machine learning counterparts that follow the true gradient more closely. Next, we verify that such generalization performance is correlated significantly with loss landscape curvature, and we show that biologically-plausible learning rules tend to approach high-curvature regions in synaptic weight space. Using tools from dynamical systems, we derive theoretical arguments and present a theorem explaining this phenomenon. This predicts our numerical results, and explains why biologically-plausible rules lead to worse and more variable generalization properties. Finally, we suggest potential remedies that could be used by the brain to mitigate this effect. To our knowledge, our analysis is the first to identify the reason for this generalization gap between artificial and biologically-plausible learning rules, which can help guide future investigations into how the brain learns solutions that generalize."
                    },
                    {
                        "title": "MouseNet: A biologically constrained convolutional neural network model for the mouse visual cortex",
                        "abstract": "Convolutional neural networks trained on object recognition derive inspiration from the neural architecture of the visual system in mammals, and have been used as models of the feedforward computation performed in the primate ventral stream. In contrast to the deep hierarchical organization of primates, the visual system of the mouse has a shallower arrangement. Since mice and primates are both capable of visually guided behavior, this raises questions about the role of architecture in neural computation. In this work, we introduce a novel framework for building a biologically constrained convolutional neural network model of the mouse visual cortex. The architecture and structural parameters of the network are derived from experimental measurements, specifically the 100-micrometer resolution interareal connectome, the estimates of numbers of neurons in each area and cortical layer, and the statistics of connections between cortical layers. This network is constructed to support detailed task-optimized models of mouse visual cortex, with neural populations that can be compared to specific corresponding populations in the mouse brain. Using a well-studied image classification task as our working example, we demonstrate the computational capability of this mouse-sized network. Given its relatively small size, MouseNet achieves roughly 2/3rds the performance level on ImageNet as VGG16. In combination with the large scale Allen Brain Observatory Visual Coding dataset, we use representational similarity analysis to quantify the extent to which MouseNet recapitulates the neural representation in mouse visual cortex. Importantly, we provide evidence that optimizing for task performance does not improve similarity to the corresponding biological system beyond a certain point. We demonstrate that the distributions of some physiological quantities are closer to the observed distributions in the mouse brain after task training. We encourage the use of the MouseNet architecture by making the code freely available."
                    },
                    {
                        "title": "Learning dynamics of deep linear networks with multiple pathways",
                        "abstract": "Not only have deep networks become standard in machine learning, they are increasingly of interest in neuroscience as models of cortical computation that capture relationships between structural and functional properties. In addition they are a useful target of theoretical research into the properties of network computation. Deep networks typically have a serial or approximately serial organization across layers, and this is often mirrored in models that purport to represent computation in mammalian brains. There are, however, multiple examples of parallel pathways in mammalian brains. In some cases, such as the mouse, the entire visual system appears arranged in a largely parallel, rather than serial fashion. While these pathways may be formed by differing cost functions that drive different computations, here we present a new mathematical analysis of learning dynamics in networks that have parallel computational pathways driven by the same cost function. We use the approximation of deep linear networks with large hidden layer sizes to show that, as the depth of the parallel pathways increases, different features of the training set (defined by the singular values of the input-output correlation) will typically concentrate in one of the pathways. This result is derived analytically and demonstrated with numerical simulation with both linear and non-linear networks. Thus, rather than sharing stimulus and task features across multiple pathways, parallel network architectures learn to produce sharply diversified representations with specialized and specific pathways, a mechanism which may hold important consequences for codes in both biological and artificial systems."
                    },
                    {
                        "title": "Biologically-plausible backpropagation through arbitrary timespans via local neuromodulators",
                        "abstract": "The spectacular successes of recurrent neural network models where key parameters are adjusted via backpropagation-based gradient descent have inspired much thought as to how biological neuronal networks might solve the corresponding synaptic credit assignment problem. There is so far little agreement, however, as to how biological networks could implement the necessary backpropagation through time, given widely recognized constraints of biological synaptic network signaling architectures. Here, we propose that extra-synaptic diffusion of local neuromodulators such as neuropeptides may afford an effective mode of backpropagation lying within the bounds of biological plausibility. Going beyond existing temporal truncation-based gradient approximations, our approximate gradient-based update rule, ModProp, propagates credit information through arbitrary time steps. ModProp suggests that modulatory signals can act on receiving cells by convolving their eligibility traces via causal, time-invariant and synapse-type-specific filter taps. Our mathematical analysis of ModProp learning, together with simulation results on benchmark temporal tasks, demonstrate the advantage of ModProp over existing biologically-plausible temporal credit assignment rules. These results suggest a potential neuronal mechanism for signaling credit information related to recurrent interactions over a longer time horizon. Finally, we derive an in-silico implementation of ModProp that could serve as a low-complexity and causal alternative to backpropagation through time."
                    },
                    {
                        "title": "Heterogeneity in Neuronal Dynamics is Learned by Gradient Descent for Temporal Processing Tasks",
                        "abstract": "Individual neurons in the brain have complex intrinsic dynamics that are highly diverse. We hypothesize that the complex dynamics produced by networks of complex and heterogeneous neurons may contribute to the brain\u2019s ability to process and respond to temporally complex data. To study the role of complex and heterogeneous neuronal dynamics in network computation, we develop a rate-based neuronal model, the generalized-leaky-integrate-and-firing-rate (GLIFR) model, which is a rate-equivalent of the generalized-leaky-integrate-and-fire model. The GLIFR model has multiple dynamical mechanisms which add to the complexity of its activity while maintaining differentiability. We focus on the role of after-spike currents, currents induced or modulated by neuronal spikes, in producing rich temporal dynamics. We use machine learning techniques to learn both synaptic weights and parameters underlying intrinsic dynamics to solve temporal tasks. The GLIFR model allows us to use standard gradient descent techniques rather than surrogate gradient descent, which has been utilized in spiking neural networks. After establishing the ability to optimize parameters using gradient descent in single neurons, we ask how networks of GLIFR neurons learn and perform on temporally challenging tasks, such as sinusoidal pattern generation and sequential MNIST. We find that these networks learn a diversity of parameters, which gives rise to diversity in neuronal dynamics. We also observe that training networks on the sequential MNIST task leads to formation of cell classes based on the clustering of neuronal parameters. GLIFR networks have mixed performance when compared to vanilla recurrent neural networks but appear to be more robust to random silencing. When we explore these performance gains further, we find that both the ability to learn heterogeneity and the presence of after-spike currents contribute. Our work both demonstrates the computational robustness of neuronal complexity and diversity in networks and demonstrates a feasible method of training such models using exact gradients."
                    },
                    {
                        "title": "Cortical network structure mediates response to stimulation: an optogenetic study in non-human primates",
                        "abstract": "As aberrant network-level functional connectivity underlies a variety of neural disorders, the ability to induce targeted functional reorganization would be a profound development towards therapies for neural disorders. Brain stimulation has been shown to alter large-scale network-wide functional connectivity, but the mapping from stimulation to the modification is unclear. Here, we leverage advances in neural interfaces, interpretable machine learning, and graph theory to arrive at a model which accurately predicts stimulation-induced network-wide functional reorganization. The model jointly considers the stimulation protocol and the cortical network structure, departing from the standard approach which only considers the stimulation protocol. We validate our approach in the primary sensorimotor cortex of non-human primates using paired optogenetic stimulation through a large-scale optogenetic interface. We observe that the stimulation protocol only predicts a small portion of the induced functional connectivity changes while the network structure predicts much more, indicating that the underlying network is the primary mediator of the response to stimulation. We extract the relationships linking the stimulation and network characteristics to the functional connectivity changes and observe that the mappings diverge over frequency bands and successive stimulations. Finally, we uncover shared processes governing real-time and longer-term effects of stimulation. Our framework represents a paradigm shift for targeted neural stimulation and can be used to interrogate, improve, and develop stimulation-based interventions for neural disorders. Teaser Brain stimulation rewires the brain, but the pre-existing network structure of the brain controls the rewiring."
                    },
                    {
                        "title": "Cell-type-speci\ufb01c modulatory signaling promotes learning in spiking neural networks",
                        "abstract": "Animals learn and form memories by jointly adjusting the e\ufb03cacy of their synapses. How they e\ufb03ciently solve the underlying temporal credit assignment problem remains elusive. Here, we re-analyze the mathematical basis of gradient descent learning in recurrent spiking neural networks (RSNNs) in light of the recent single-cell transcriptomic evidence for cell-type-speci\ufb01c local neuropeptide signaling in the cortex. Our normative theory posits an important role for the notion of neuronal cell types and local di\ufb00usive communication by enabling biologically plausible and e\ufb03cient weight update. While obeying fundamental biological constraints, including separating excitatory vs inhibitory cell types and observing connection sparsity, we trained RSNNs for temporal credit assignment tasks spanning seconds and observed that the inclusion of local modulatory signaling improved learning e\ufb03ciency. Our learning rule puts forth a novel form of interaction between modulatory signals and synaptic transmission. Moreover, it suggests a computationally e\ufb03cient learning method for bio-inspired arti\ufb01cial intelligence."
                    },
                    {
                        "title": "Cell-type\u2013specific neuromodulation guides synaptic credit assignment in a spiking neural network",
                        "abstract": "Significance Synaptic connectivity provides the foundation for our present understanding of neuronal network function, but static connectivity cannot explain learning and memory. We propose a computational role for the diversity of cortical neuronal types and their associated cell-type\u2013specific neuromodulators in improving the efficiency of synaptic weight adjustments for task learning in neuronal networks. Brains learn tasks via experience-driven differential adjustment of their myriad individual synaptic connections, but the mechanisms that target appropriate adjustment to particular connections remain deeply enigmatic. While Hebbian synaptic plasticity, synaptic eligibility traces, and top-down feedback signals surely contribute to solving this synaptic credit-assignment problem, alone, they appear to be insufficient. Inspired by new genetic perspectives on neuronal signaling architectures, here, we present a normative theory for synaptic learning, where we predict that neurons communicate their contribution to the learning outcome to nearby neurons via cell-type\u2013specific local neuromodulation. Computational tests suggest that neuron-type diversity and neuron-type\u2013specific local neuromodulation may be critical pieces of the biological credit-assignment puzzle. They also suggest algorithms for improved artificial neural network learning efficiency."
                    },
                    {
                        "title": "A biologically inspired architecture with switching units can learn to generalize across backgrounds",
                        "abstract": "Humans and other animals navigate different landscapes and environments with ease, a feat that requires the brain\u2019s ability to rapidly and accurately adapt to different visual domains, generalizing across contexts/backgrounds. Despite recent progress in deep learning applied to classification and detection in the presence of multiple confounds including contextual ones [25, 30], there remain important challenges to address regarding how networks can perform context-dependent computations and how contextually-invariant visual concepts are formed. For instance, recent studies have shown artificial networks that repeatedly misclassified familiar objects set on new backgrounds, e.g. incorrectly labelling known animals when they appeared in a different setting [3]. Here, we show how a bio-inspired network motif can explicitly address this issue. We do this using a novel dataset which can be used as a benchmark for future studies probing invariance to backgrounds. The dataset consists of MNIST digits of varying transparency, set on one of two backgrounds with different statistics: a Gaussian noise or a more naturalistic background from the CIFAR-10 dataset. We use this dataset to learn digit classification when contexts are shown sequentially, and find that both shallow and deep networks have sharply decreased performance when returning to the first background after experience learning the second \u2013 the catastrophic forgetting phenomenon in continual learning. To overcome this, we propose an architecture with additional \u201c switching\u201d units that are activated in the presence of a new background. We find that the switching network can learn the new context even with very few switching units, while maintaining the performance in the previous context \u2013 but that they must be recurrently connected to network layers. When the task is difficult due to high transparency, the switching network trained on both contexts outperforms networks without switching trained on only one context. The switching mechanism leads to sparser activation patterns, and we provide intuition for why this helps to solve the task. We compare our architecture with other prominent learning methods, and find that elastic weight consolidation is not successful in our setting, while progressive nets are more complex but less effective. Our study therefore shows how a bio-inspired architectural motif can contribute to task generalization across context."
                    },
                    {
                        "title": "CNN MouseNet: A biologically constrained convolutional neural network model for mouse visual cortex",
                        "abstract": "Convolutional neural networks trained on object recognition derive inspiration from the neural architecture of the visual system in primates, and have been used as models of the feedforward computation performed in the primate ventral stream. In contrast to the deep hierarchical organization of primates, the visual system of the mouse has a shallower arrangement. Since mice and primates are both capable of visually guided behavior, this raises questions about the role of architecture in neural computation. In this work, we introduce a novel framework for building a biologically constrained convolutional neural network model of the mouse visual cortex. The architecture and structural parameters of the network are derived from experimental measurements, specifically the 100-micrometer resolution interareal connectome, the estimates of numbers of neurons in each area and cortical layer, and the statistics of connections between cortical layers. This network is constructed to support detailed task-optimized models of mouse visual cortex, with neural populations that can be compared to specific corresponding populations in the mouse brain. Using a well-studied image classification task as our working example, we demonstrate the computational capability of this mouse-sized network. Given its relatively small size, MouseNet achieves roughly 2/3rds the performance level on ImageNet as VGG16. In combination with the large scale Allen Brain Observatory Visual Coding dataset, we use representational similarity analysis to quantify the extent to which MouseNet recapitulates the neural representation in mouse visual cortex. Importantly, we provide evidence that optimizing for task performance does not improve similarity to the corresponding biological system beyond a certain point. We demonstrate that the distributions of some physiological quantities are closer to the observed distributions in the mouse brain after task training. We encourage the use of the MouseNet architecture by making the code freely available. Author summary Task-driven deep neural networks have shown great potential in predicting functional responses of biological neurons. Nevertheless, they are not precise biological analogues, raising questions about how they should be interpreted. Here, we build new deep neural network models of the mouse visual cortex (MouseNet) that are biologically constrained in detail, not only in terms of the basic structure of their connectivity, but also in terms of the count and hence density of neurons within each area, and the spatial extent of their projections. Equipped with the MouseNet model, we can address key questions about mesoscale brain architecture and its role in task learning and performance.We ask, and provide a first set of answers, to: What is the performance of a mouse brain-sized \u2013 and mouse brain-structured \u2013 model on benchmark image classification tasks? How does the training of a network on this task affect the functional properties of specified layers within the biologically constrained architecture \u2013 both overall, and in comparison with recorded function of mouse neurons? We anticipate much future work on allied questions, and the development of more sophisticated models in both mouse and other species, based on the freely available MouseNet model and code which we develop and provide here."
                    }
                ]
            },
            "2ef435e1-9c8e-409e-8225-d6c46a1bcef2": {
                "pk": "2ef435e1-9c8e-409e-8225-d6c46a1bcef2",
                "name": "Guillaume Lajoie",
                "collaborators": [
                    "Lorenz Wernisch",
                    "Tris Edwards",
                    "Antonin Berthon",
                    "Olivier Tessier-Larivi\u00e8re",
                    "Elvijs Sarkans",
                    "Myrta Stoukidi",
                    "Pascal Fortier-Poisson",
                    "Max Pinkney",
                    "Michael Thornton",
                    "Susannah Lee",
                    "Joel Jennings",
                    "Ben Appleton",
                    "P. Garsed",
                    "Bret Patterson",
                    "Samuel Gonshaw",
                    "Matja\u017e Jakopec",
                    "Sudha Shunmugam",
                    "J. Mamen",
                    "Aleksi Tukiainen",
                    "Oliver Armitage",
                    "Emil Hewage",
                    "Catherine Hanley",
                    "Will Buttinger",
                    "Luca Annecchino",
                    "Stefano Recanatesi",
                    "M. Farrell",
                    "Sophie Deneve",
                    "Mattia Rigotti",
                    "E. Shea-Brown"
                ],
                "domain": [
                    "Bioelectronic Medicine",
                    "Neural Stimulation",
                    "Machine Learning",
                    "Predictive Modeling"
                ],
                "publications": [
                    {
                        "title": "Online Bayesian optimization of vagus nerve stimulation",
                        "abstract": "Objective. In bioelectronic medicine, neuromodulation therapies induce neural signals to the brain or organs, modifying their function. Stimulation devices capable of triggering exogenous neural signals using electrical waveforms require a complex and multi-dimensional parameter space to control such waveforms. Determining the best combination of parameters (waveform optimization or dosing) for treating a particular patient\u2019s illness is therefore challenging. Comprehensive parameter searching for an optimal stimulation effect is often infeasible in a clinical setting due to the size of the parameter space. Restricting this space, however, may lead to suboptimal therapeutic results, reduced responder rates, and adverse effects. Approach. As an alternative to a full parameter search, we present a flexible machine learning, data acquisition, and processing framework for optimizing neural stimulation parameters, requiring as few steps as possible using Bayesian optimization. This optimization builds a model of the neural and physiological responses to stimulations, enabling it to optimize stimulation parameters and provide estimates of the accuracy of the response model. The vagus nerve (VN) innervates, among other thoracic and visceral organs, the heart, thus controlling heart rate (HR), making it an ideal candidate for demonstrating the effectiveness of our approach. Main results. The efficacy of our optimization approach was first evaluated on simulated neural responses, then applied to VN stimulation intraoperatively in porcine subjects. Optimization converged quickly on parameters achieving target HRs and optimizing neural B-fiber activations despite high intersubject variability. Significance. An optimized stimulation waveform was achieved in real time with far fewer stimulations than required by alternative optimization strategies, thus minimizing exposure to side effects. Uncertainty estimates helped avoiding stimulations outside a safe range. Our approach shows that a complex set of neural stimulation parameters can be optimized in real-time for a patient to achieve a personalized precision dosing."
                    },
                    {
                        "title": "Online Bayesian Optimization of Nerve Stimulation",
                        "abstract": "Objective In bioelectronic medicine, neuromodulation therapies induce neural signals to the brain or organs modifying their function. Stimulation devices, capable of triggering exogenous neural signals using electrical wave forms, require a complex and multi-dimensional parameter space in order to control such wave forms. Determining the best combination of parameters (wave form optimization, or dosing) for treating a particular patient\u2019s illness is therefore challenging. Comprehensive parameter searching for an optimal stimulation effect is often infeasible in a clinical setting, due to the size of the parameter space. Restricting this space, however, may lead to sub-optimal therapeutic results, reduced responder rates, and adverse effects. Approach As an alternative to a full parameter search, we present a flexible machine learning, data acquisition and processing framework for optimizing neural stimulation parameters requiring as few steps as possible using Bayesian optimization. Such optimization builds a model of the neural and physiological responses to stimulations enabling it to optimize stimulation parameters and to provide estimates of the accuracy of the response model. The vagus nerve innervates, among other thoracic and visceral organs, the heart, thus controlling heart rate and is therefore ideal for demonstrating the effectiveness of our approach. Main results. The efficacy of our optimization approach was first evaluated on simulated neural responses, then applied to vagus nerve stimulation intraoperatively in porcine subjects. Optimization converged quickly on parameters achieving target heart rates and optimizing neural B-fibre activations despite high intersubject variability. Significance An optimized stimulation waveform was achieved in real time with far fewer stimulations than required by alternative optimization strategies, thus minimizing exposure to side effects. Uncertainty estimates helped avoiding stimulations outside a safe range. Our approach shows that a complex set of neural stimulation parameters can be optimized in real-time for a patient to achieve a personalized precision dosing."
                    },
                    {
                        "title": "Using neural biomarkers to personalize dosing of vagus nerve stimulation",
                        "abstract": "Background Vagus nerve stimulation (VNS) is an established therapy for treating a variety of chronic diseases, such as epilepsy, depression, obesity, and for stroke rehabilitation. However, lack of precision and side-effects have hindered its efficacy and extension to new conditions. Objective To achieve a better understanding of the relationship between VNS parameters and neural and physiological responses to enable the design of personalized dosing procedures to improve precision and efficacy of VNS therapies. Methods We used biomarkers from recorded evoked neural activity and short-term physiological responses (throat muscle, cardiac and respiratory activity) to understand the response to a wide range of VNS parameters in anaesthetised pigs. Using signal processing, Gaussian processes (GP) and parametric regression models we analyse the relationship between VNS parameters and neural and physiological responses. Results Firstly, we observe inter-subject variability for both neural and physiological responses. Secondly, we illustrate how considering multiple stimulation parameters in VNS dosing can improve the efficacy and precision of VNS therapies. Thirdly, we describe the relationship between different VNS parameters and the evoked neural activity and show how spatially selective electrodes can be used to improve fibre recruitment. Fourthly, we provide a detailed exploration of the relationship between the activations of neural fibre types and different physiological effects, and show that recordings of evoked neural activity are powerful biomarkers for predicting the short-term physiological effects of VNS. Finally, based on these results, we discuss how recordings of evoked neural activity can help design VNS dosing procedures that optimize short-term physiological effects safely and efficiently. Conclusion Understanding of evoked neural activity during VNS provide powerful biomarkers that could improve the precision, safety and efficacy of VNS therapies."
                    },
                    {
                        "title": "Signatures and mechanisms of low-dimensional neural predictive manifolds",
                        "abstract": "Many of the recent advances of neural networks in sequential tasks such as natural language processing applications hinge on the use of representations obtained by predictive models. This success is usually ascribed to the emergence of neural representations that capture the low-dimensional latent structure implicit in the task. Motivated by the recent theoretical proposal that the hippocampus performs its role in sequential planning by organizing semantically related episodes in a relational network, we investigate the hypothesis that this organization results from learning a predictive representation of the world. Using an artificial recurrent neural network model trained with predictive learning on a simulated spatial navigation task, we show that network dynamics exhibit low dimensional but non-linearly transformed representations of sensory input statistics. These neural activations that are strongly reminiscent of the place-related neural activity that is experimentally observed in the hippocampus and in the entorhinal cortex. We quantify these results using measures of intrinsic dimensionality, which indeed confirm that the neural representations obtained with predictive learning reflect the low-dimensional latent structure of the spatial environment underlying the sensory input presented to the network. Moreover, the dimensionality gain of the neural representations, a measure of the discrepancy between linear and intrinsic dimensionality, allows us to follow how this process evolves as learning unfolds. Finally, we provide theoretical arguments as to how predictive learning can extract the latent manifold underlying sequential signals, and discuss how our results and methods can aid the analysis of experimental data."
                    }
                ]
            }
        }
    },
    "2402.03139": {
        "paper_data": {
            "title": "Enhancing Neural Subset Selection: Integrating Background Information into Set Representations",
            "url": "http://arxiv.org/abs/2402.03139v2",
            "arxiv_id": "2402.03139",
            "authors": [
                "Binghui Xie",
                "Yatao Bian",
                "Kaiwen zhou",
                "Yongqiang Chen",
                "Peilin Zhao",
                "Bo Han",
                "Wei Meng",
                "James Cheng"
            ],
            "abstract": "Learning neural subset selection tasks, such as compound selection in AI-aided drug discovery, have become increasingly pivotal across diverse applications. The existing methodologies in the field primarily concentrate on constructing models that capture the relationship between utility function values and subsets within their respective supersets. However, these approaches tend to overlook the valuable information contained within the superset when utilizing neural networks to model set functions. In this work, we address this oversight by adopting a probabilistic perspective. Our theoretical findings demonstrate that when the target value is conditioned on both the input set and subset, it is essential to incorporate an \\textit{invariant sufficient statistic} of the superset into the subset of interest for effective learning. This ensures that the output value remains invariant to permutations of the subset and its corresponding superset, enabling identification of the specific superset from which the subset originated. Motivated by these insights, we propose a simple yet effective information aggregation module designed to merge the representations of subsets and supersets from a permutation invariance perspective. Comprehensive empirical evaluations across diverse tasks and datasets validate the enhanced efficacy of our approach over conventional methods, underscoring the practicality and potency of our proposed strategies in real-world contexts.",
            "introduction": "   1 Introduction  The prediction of set-valued outputs plays a crucial role in various real-world applications. For instance, anomaly detection involves identifying outliers from a majority of data (Zhang et\u00a0al., 2020), and compound selection in drug discovery aims to extract the most effective compounds from a given compound database (Gimeno et\u00a0al., 2019). In these applications, there exists an implicit learning of a set function (Rezatofighi et\u00a0al., 2017; Zaheer et\u00a0al., 2017) that quantifies the utility of a given set input, where the highest utility value corresponds to the most desirable set output.   More formally, let\u2019s consider the compound selection task: given a compound database V\ud835\udc49Vitalic_V, the goal is to select a subset of compounds S\u2217\u2286Vsuperscript\ud835\udc46\ud835\udc49S^{*}\\subseteq Vitalic_S start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT \u2286 italic_V that exhibit the highest utility. This utility can be modeled by a parameterized utility function F\u03b8\u2062(S;V)subscript\ud835\udc39\ud835\udf03\ud835\udc46\ud835\udc49F_{\\theta}(S;V)italic_F start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( italic_S ; italic_V ), and the optimization criteria can be expressed as:    S\u2217=arg\u2062maxS\u22082V\u2061F\u03b8\u2062(S;V).superscript\ud835\udc46subscriptargmax\ud835\udc46superscript2\ud835\udc49subscript\ud835\udc39\ud835\udf03\ud835\udc46\ud835\udc49\\displaystyle S^{*}=\\operatorname*{arg\\,max}_{S\\in 2^{V}}F_{\\theta}(S;V).italic_S start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT = start_OPERATOR roman_arg roman_max end_OPERATOR start_POSTSUBSCRIPT italic_S \u2208 2 start_POSTSUPERSCRIPT italic_V end_POSTSUPERSCRIPT end_POSTSUBSCRIPT italic_F start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( italic_S ; italic_V ) .  (1)   One straightforward method is to explicitly model the utility by learning U=F\u03b8\u2062(S;V)\ud835\udc48subscript\ud835\udc39\ud835\udf03\ud835\udc46\ud835\udc49U=F_{\\theta}(S;V)italic_U = italic_F start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( italic_S ; italic_V ) using supervised data in the form of {(Si,Vi),Ui}i=1N,superscriptsubscriptsubscript\ud835\udc46\ud835\udc56subscript\ud835\udc49\ud835\udc56subscript\ud835\udc48\ud835\udc56\ud835\udc561\ud835\udc41\\{(S_{i},V_{i}),U_{i}\\}_{i=1}^{N},{ ( italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT , italic_V start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) , italic_U start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT } start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_N end_POSTSUPERSCRIPT , where Uisubscript\ud835\udc48\ud835\udc56U_{i}italic_U start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT represents the true utility value of subset Sisubscript\ud835\udc46\ud835\udc56S_{i}italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT given Visubscript\ud835\udc49\ud835\udc56V_{i}italic_V start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT. However, this training approach becomes prohibitively expensive due to the need for constructing a large amount of supervision signals (Balcan & Harvey, 2018).   To address this limitation, another way is to solve Eq.1 with an implicit learning approach from a probabilistic perspective. Specifically, it is required to utilize data in the form of {(Vi,Si\u2217)}i=1Nsuperscriptsubscriptsubscript\ud835\udc49\ud835\udc56superscriptsubscript\ud835\udc46\ud835\udc56\ud835\udc561\ud835\udc41\\{(V_{i},S_{i}^{*})\\}_{i=1}^{N}{ ( italic_V start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT , italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT ) } start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_N end_POSTSUPERSCRIPT, where Si\u2217superscriptsubscript\ud835\udc46\ud835\udc56S_{i}^{*}italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT represents the optimal subset corresponding to Visubscript\ud835\udc49\ud835\udc56V_{i}italic_V start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT. The goal is to estimate \u03b8\ud835\udf03\\thetaitalic_\u03b8 such that Eq.\u00a01 holds for all possible (Vi,Si)subscript\ud835\udc49\ud835\udc56subscript\ud835\udc46\ud835\udc56(V_{i},S_{i})( italic_V start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT , italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ). During practical training, with limited data (Si\u2217,Vi)i=1Nsuperscriptsubscriptsuperscriptsubscript\ud835\udc46\ud835\udc56subscript\ud835\udc49\ud835\udc56\ud835\udc561\ud835\udc41{(S_{i}^{*},V_{i})}_{i=1}^{N}( italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT , italic_V start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT ) start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_N end_POSTSUPERSCRIPT sampled from the underlying data distribution P\u2062(S,V)\ud835\udc43\ud835\udc46\ud835\udc49P(S,V)italic_P ( italic_S , italic_V ), the empirical log likelihood \u2211i=1N[log\u2061p\u03b8\u2062(S\u2217|V)]superscriptsubscript\ud835\udc561\ud835\udc41delimited-[]subscript\ud835\udc5d\ud835\udf03conditionalsuperscript\ud835\udc46\ud835\udc49\\sum_{i=1}^{N}[\\log p_{\\theta}(S^{*}|V)]\u2211 start_POSTSUBSCRIPT italic_i = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_N end_POSTSUPERSCRIPT [ roman_log italic_p start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( italic_S start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT | italic_V ) ] is maximized among all data pairs {S,V},\ud835\udc46\ud835\udc49\\{S,V\\},{ italic_S , italic_V } , where p\u03b8\u2062(S|V)\u221dF\u03b8\u2062(S;V)proportional-tosubscript\ud835\udc5d\ud835\udf03conditional\ud835\udc46\ud835\udc49subscript\ud835\udc39\ud835\udf03\ud835\udc46\ud835\udc49p_{\\theta}(S|V)\\propto F_{\\theta}(S;V)italic_p start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( italic_S | italic_V ) \u221d italic_F start_POSTSUBSCRIPT italic_\u03b8 end_POSTSUBSCRIPT ( italic_S ; italic_V ) for all S\u22082V\ud835\udc46superscript2\ud835\udc49S\\in 2^{V}italic_S \u2208 2 start_POSTSUPERSCRIPT italic_V end_POSTSUPERSCRIPT. To achieve this objective, Ou et\u00a0al. (2022) proposed to use a variational distribution q\u2062(Y|S,V)\ud835\udc5econditional\ud835\udc4c\ud835\udc46\ud835\udc49q(Y|S,V)italic_q ( italic_Y | italic_S , italic_V ) to approximate the distribution of P\u2062(S|V)\ud835\udc43conditional\ud835\udc46\ud835\udc49P(S|V)italic_P ( italic_S",
            "references": [
                {
                    "title": "Learning Neural Set Functions Under the Optimal Subset Oracle",
                    "abstract": "Learning neural set functions becomes increasingly more important in many applications like product recommendation and compound selection in AI-aided drug discovery. The majority of existing works study methodologies of set function learning under the function value oracle, which, however, requires expensive supervision signals. This renders it impractical for applications with only weak supervisions under the Optimal Subset (OS) oracle, the study of which is surprisingly overlooked. In this work, we present a principled yet practical maximum likelihood learning framework, termed as EquiVSet, that simultaneously meets the following desiderata of learning set functions under the OS oracle: i) permutation invariance of the set mass function being modeled; ii) permission of varying ground set; iii) minimum prior; and iv) scalability. The main components of our framework involve: an energy-based treatment of the set mass function, DeepSet-style architectures to handle permutation invariance, mean-field variational inference, and its amortized variants. Thanks to the elegant combination of these advanced architectures, empirical studies on three real-world applications (including Amazon product recommendation, set anomaly detection, and compound selection for virtual screening) demonstrate that EquiVSet outperforms the baselines by a large margin."
                },
                {
                    "title": "Equivariant Graph Hierarchy-Based Neural Networks",
                    "abstract": "Equivariant Graph neural Networks (EGNs) are powerful in characterizing the dynamics of multi-body physical systems. Existing EGNs conduct flat message passing, which, yet, is unable to capture the spatial/dynamical hierarchy for complex systems particularly, limiting substructure discovery and global information fusion. In this paper, we propose Equivariant Hierarchy-based Graph Networks (EGHNs) which consist of the three key components: generalized Equivariant Matrix Message Passing (EMMP) , E-Pool and E-UpPool. In particular, EMMP is able to improve the expressivity of conventional equivariant message passing, E-Pool assigns the quantities of the low-level nodes into high-level clusters, while E-UpPool leverages the high-level information to update the dynamics of the low-level nodes. As their names imply, both E-Pool and E-UpPool are guaranteed to be equivariant to meet physic symmetry. Considerable experimental evaluations verify the effectiveness of our EGHN on several applications including multi-object dynamics simulation, motion capture, and protein dynamics modeling."
                },
                {
                    "title": "DrugOOD: Out-of-Distribution (OOD) Dataset Curator and Benchmark for AI-aided Drug Discovery - A Focus on Affinity Prediction Problems with Noise Annotations",
                    "abstract": "AI-aided drug discovery (AIDD) is gaining increasing popularity due to its promise of making the search for new pharmaceuticals quicker, cheaper and more efficient. In spite of its extensive use in many fields, such as ADMET prediction, virtual screening, protein folding and generative chemistry, little has been explored in terms of the out-of-distribution (OOD) learning problem with \\emph{noise}, which is inevitable in real world AIDD applications. In this work, we present DrugOOD, a systematic OOD dataset curator and benchmark for AI-aided drug discovery, which comes with an open-source Python package that fully automates the data curation and OOD benchmarking processes. We focus on one of the most crucial problems in AIDD: drug target binding affinity prediction, which involves both macromolecule (protein target) and small-molecule (drug compound). In contrast to only providing fixed datasets, DrugOOD offers automated dataset curator with user-friendly customization scripts, rich domain annotations aligned with biochemistry knowledge, realistic noise annotations and rigorous benchmarking of state-of-the-art OOD algorithms. Since the molecular data is often modeled as irregular graphs using graph neural network (GNN) backbones, DrugOOD also serves as a valuable testbed for \\emph{graph OOD learning} problems. Extensive empirical studies have shown a significant performance gap between in-distribution and out-of-distribution experiments, which highlights the need to develop better schemes that can allow for OOD generalization under noise for AIDD."
                },
                {
                    "title": "Independent SE(3)-Equivariant Models for End-to-End Rigid Protein Docking",
                    "abstract": "Protein complex formation is a central problem in biology, being involved in most of the cell's processes, and essential for applications, e.g. drug design or protein engineering. We tackle rigid body protein-protein docking, i.e., computationally predicting the 3D structure of a protein-protein complex from the individual unbound structures, assuming no conformational change within the proteins happens during binding. We design a novel pairwise-independent SE(3)-equivariant graph matching network to predict the rotation and translation to place one of the proteins at the right docked position relative to the second protein. We mathematically guarantee a basic principle: the predicted complex is always identical regardless of the initial locations and orientations of the two structures. Our model, named EquiDock, approximates the binding pockets and predicts the docking poses using keypoint matching and alignment, achieved through optimal transport and a differentiable Kabsch algorithm. Empirically, we achieve significant running time improvements and often outperform existing docking software despite not relying on heavy candidate sampling, structure refinement, or templates."
                },
                {
                    "title": "Cloud\u2013Edge-Based Lightweight Temporal Convolutional Networks for Remaining Useful Life Prediction in IIoT",
                    "abstract": "Industrial Internet of Things (IIoT), as an important industrial branch of the Internet of Things (IoT), has an essential purpose to improve intelligent industrial production. For this purpose, IIoT big data should be efficiently processed to mine valuable information. In handing the IIoT big data, cloud\u2013edge computing is getting more attention to reduce the interaction latency to meet the real-time requirement, especially in the field of prognostic and health management (PHM). It is expected that artificial intelligence (AI) technologies will significantly change the manner of processing IIoT big data. Therefore, new methods about PHM, combining cloud\u2013edge computing with AI technologies, are required to process the IIoT big data for intelligent industrial manufacturing. As an essential element of PHM, predicting the remaining useful life (RUL) of industrial equipment plays an increasingly crucial role, especially for industrial intelligence. However, traditional methods pay much attention on prediction accuracy and neglect the influence of computing time. In this article, by combining cloud\u2013edge computing with AI technology, a new data-driven method, namely, cloud\u2013edge-based lightweight temporal convolutional networks (LTCNs), for RUL prediction is proposed. First, to meet the real-time requirement, a cloud\u2013edge computing and AI-based framework for RUL prediction is presented. Second, a new model structure named LTCN is proposed and applied in the framework. Real-time prediction results will be obtained in the edge plane and higher accuracy prediction results will be obtained through historical information in the cloud plane. Third, an incremental learning approach based on updating partial parameters of LTCN is discussed to improve the accuracy of prediction models with newly collected data. Experiments show that our method can improve the prediction accuracy and reduce the computational time of RUL."
                },
                {
                    "title": "Structure-aware Interactive Graph Neural Networks for the Prediction of Protein-Ligand Binding Affinity",
                    "abstract": "Drug discovery often relies on the successful prediction of protein-ligand binding affinity. Recent advances have shown great promise in applying graph neural networks (GNNs) for better affinity prediction by learning the representations of protein-ligand complexes. However, existing solutions usually treat protein-ligand complexes as topological graph data, thus the biomolecular structural information is not fully utilized. The essential long-range interactions among atoms are also neglected in GNN models. To this end, we propose a structure-aware interactive graph neural network (SIGN) which consists of two components: polar-inspired graph attention layers (PGAL) and pairwise interactive pooling (PiPool). Specifically, PGAL iteratively performs the node-edge aggregation process to update embeddings of nodes and edges while preserving the distance and angle information among atoms. Then, PiPool is adopted to gather interactive edges with a subsequent reconstruction loss to reflect the global interactions. Exhaustive experimental study on two benchmarks verifies the superiority of SIGN."
                },
                {
                    "title": "Mini-Batch Consistent Slot Set Encoder for Scalable Set Encoding",
                    "abstract": "Most existing set encoding algorithms operate under the implicit assumption that all the set elements are accessible, and that there are ample computational and memory resources to load the set into memory during training and inference. However, both assumptions fail when the set is excessively large such that it is impossible to load all set elements into memory, or when data arrives in a stream. To tackle such practical challenges in large-scale set encoding, the general set-function constraints of permutation invariance and equivariance are not sufficient. We introduce a new property termed Mini-Batch Consistency (MBC) that is required for large scale mini-batch set encoding. Additionally, we present a scalable and efficient attention-based set encoding mechanism that is amenable to mini-batch processing of sets, and capable of updating set representations as data arrives. The proposed method adheres to the required symmetries of invariance and equivariance as well as maintaining MBC for any partition of the input set. We perform extensive experiments and show that our method is computationally efficient and results in rich set encoding representations for set-structured data."
                },
                {
                    "title": "Set Prediction without Imposing Structure as Conditional Density Estimation",
                    "abstract": "Set prediction is about learning to predict a collection of unordered variables with unknown interrelations. Training such models with set losses imposes the structure of a metric space over sets. We focus on stochastic and underdefined cases, where an incorrectly chosen loss function leads to implausible predictions. Example tasks include conditional point-cloud reconstruction and predicting future states of molecules. In this paper, we propose an alternative to training via set losses by viewing learning as conditional density estimation. Our learning framework fits deep energy-based models and approximates the intractable likelihood with gradient-guided sampling. Furthermore, we propose a stochastically augmented prediction algorithm that enables multiple predictions, reflecting the possible variations in the target set. We empirically demonstrate on a variety of datasets the capability to learn multi-modal densities and produce multiple plausible predictions. Our approach is competitive with previous set prediction models on standard benchmarks. More importantly, it extends the family of addressable tasks beyond those that have unambiguous predictions."
                },
                {
                    "title": "Exchangeable Neural ODE for Set Modeling",
                    "abstract": "Reasoning over an instance composed of a set of vectors, like a point cloud, requires that one accounts for intra-set dependent features among elements. However, since such instances are unordered, the elements' features should remain unchanged when the input's order is permuted. This property, permutation equivariance, is a challenging constraint for most neural architectures. While recent work has proposed global pooling and attention-based solutions, these may be limited in the way that intradependencies are captured in practice. In this work we propose a more general formulation to achieve permutation equivariance through ordinary differential equations (ODE). Our proposed module, Exchangeable Neural ODE (ExNODE), can be seamlessly applied for both discriminative and generative tasks. We also extend set modeling in the temporal dimension and propose a VAE based model for temporal set modeling. Extensive experiments demonstrate the efficacy of our method over strong baselines."
                },
                {
                    "title": "Exploring Structural Knowledge for Automated Visual Inspection of Moving Trains",
                    "abstract": "Deep learning methods are becoming the de-facto standard for generic visual recognition in the literature. However, their adaptations to industrial scenarios, such as visual recognition for machines, product streamlines, etc., which consist of countless components, have not been investigated well yet. Compared with the generic object detection, there is some strong structural knowledge in these scenarios (e.g., fixed relative positions of components, component relationships, etc.). A case worth exploring could be automated visual inspection for trains, where there are various correlated components. However, the dominant object detection paradigm is limited by treating the visual features of each object region separately without considering common sense knowledge among objects. In this article, we propose a novel automated visual inspection framework for trains exploring structural knowledge for train component detection, which is called SKTCD. SKTCD is an end-to-end trainable framework, in which the visual features of train components and structural knowledge (including hierarchical scene contexts and spatial-aware component relationships) are jointly exploited for train component detection. We propose novel residual multiple gated recurrent units (Res-MGRUs) that can optimally fuse the visual features of train components and messages from the structural knowledge in a weighted-recurrent way. In order to verify the feasibility of SKTCD, a dataset that contains high-resolution images captured from moving trains has been collected, in which 18 590 critical train components are manually annotated. Extensive experiments on this dataset and on the PASCAL VOC dataset have demonstrated that SKTCD outperforms the existing challenging baselines significantly. The dataset as well as the source code can be downloaded online (https://github.com/smartprobe/SKCD)."
                },
                {
                    "title": "On Learning Sets of Symmetric Elements",
                    "abstract": "Learning from unordered sets is a fundamental learning setup, recently attracting increasing attention. Research in this area has focused on the case where elements of the set are represented by feature vectors, and far less emphasis has been given to the common case where set elements themselves adhere to their own symmetries. That case is relevant to numerous applications, from deblurring image bursts to multi-view 3D shape recognition and reconstruction. \nIn this paper, we present a principled approach to learning sets of general symmetric elements. We first characterize the space of linear layers that are equivariant both to element reordering and to the inherent symmetries of elements, like translation in the case of images. We further show that networks that are composed of these layers, called Deep Sets for Symmetric Elements layers (DSS), are universal approximators of both invariant and equivariant functions. DSS layers are also straightforward to implement. Finally, we show that they improve over existing set-learning architectures in a series of experiments with images, graphs, and point-clouds."
                },
                {
                    "title": "Hierarchical Generation of Molecular Graphs using Structural Motifs",
                    "abstract": "Graph generation techniques are increasingly being adopted for drug discovery. Previous graph generation approaches have utilized relatively small molecular building blocks such as atoms or simple cycles, limiting their effectiveness to smaller molecules. Indeed, as we demonstrate, their performance degrades significantly for larger molecules. In this paper, we propose a new hierarchical graph encoder-decoder that employs significantly larger and more flexible graph motifs as basic building blocks. Our encoder produces a multi-resolution representation for each molecule in a fine-to-coarse fashion, from atoms to connected motifs. Each level integrates the encoding of constituents below with the graph at that level. Our autoregressive coarse-to-fine decoder adds one motif at a time, interleaving the decision of selecting a new motif with the process of resolving its attachments to the emerging molecule. We evaluate our model on multiple molecule generation tasks, including polymers, and show that our model significantly outperforms previous state-of-the-art baselines."
                },
                {
                    "title": "AttPool: Towards Hierarchical Feature Representation in Graph Convolutional Networks via Attention Mechanism",
                    "abstract": "Graph convolutional networks (GCNs) are potentially short of the ability to learn hierarchical representation for graph embedding, which holds them back in the graph classification task. Here, we propose AttPool, which is a novel graph pooling module based on attention mechanism, to remedy the problem. It is able to select nodes that are significant for graph representation adaptively, and generate hierarchical features via aggregating the attention-weighted information in nodes. Additionally, we devise a hierarchical prediction architecture to sufficiently leverage the hierarchical representation and facilitate the model learning. The AttPool module together with the entire training structure can be integrated into existing GCNs, and is trained in an end-to-end fashion conveniently. The experimental results on several graph-classification benchmark datasets with various scales demonstrate the effectiveness of our method."
                },
                {
                    "title": "Set Functions for Time Series",
                    "abstract": "Despite the eminent successes of deep neural networks, many architectures are often hard to transfer to irregularly-sampled and asynchronous time series that commonly occur in real-world datasets, especially in healthcare applications. This paper proposes a novel approach for classifying irregularly-sampled time series with unaligned measurements, focusing on high scalability and data efficiency. Our method SeFT (Set Functions for Time Series) is based on recent advances in differentiable set function learning, extremely parallelizable with a beneficial memory footprint, thus scaling well to large datasets of long time series and online monitoring scenarios. Furthermore, our approach permits quantifying per-observation contributions to the classification outcome. We extensively compare our method with existing algorithms on multiple healthcare time series datasets and demonstrate that it performs competitively whilst significantly reducing runtime."
                },
                {
                    "title": "Graph U-Nets",
                    "abstract": "We consider the problem of representation learning for graph data. Given images are special cases of graphs with nodes lie on 2D lattices, graph embedding tasks have a natural correspondence with image pixel-wise prediction tasks such as segmentation. While encoder-decoder architectures like U-Nets have been successfully applied to image pixel-wise prediction tasks, similar methods are lacking for graph data. This is because pooling and up-sampling operations are not natural on graph data. To address these challenges, we propose novel graph pooling and unpooling operations. The gPool layer adaptively selects some nodes to form a smaller graph based on their scalar projection values. We further propose the gUnpool layer as the inverse operation of the gPool layer. Based on our proposed methods, we develop an encoder-decoder model, known as the graph U-Nets. Experimental results on node classification and graph classification tasks demonstrate that our methods achieve consistently better performance than previous models. Along this direction, we extend our methods by integrating attention mechanisms. Based on attention operators, we proposed attention-based pooling and unpooling layers, which can better capture graph topology information. The empirical results on graph classification tasks demonstrate the promising capability of our methods."
                },
                {
                    "title": "The Light and Dark Sides of Virtual Screening: What Is There to Know?",
                    "abstract": "Virtual screening consists of using computational tools to predict potentially bioactive compounds from files containing large libraries of small molecules. Virtual screening is becoming increasingly popular in the field of drug discovery as in silico techniques are continuously being developed, improved, and made available. As most of these techniques are easy to use, both private and public organizations apply virtual screening methodologies to save resources in the laboratory. However, it is often the case that the techniques implemented in virtual screening workflows are restricted to those that the research team knows. Moreover, although the software is often easy to use, each methodology has a series of drawbacks that should be avoided so that false results or artifacts are not produced. Here, we review the most common methodologies used in virtual screening workflows in order to both introduce the inexperienced researcher to new methodologies and advise the experienced researcher on how to prevent common mistakes and the improper usage of virtual screening methodologies."
                },
                {
                    "title": "Towards Sparse Hierarchical Graph Classifiers",
                    "abstract": "Recent advances in representation learning on graphs, mainly leveraging graph convolutional networks, have brought a substantial improvement on many graph-based benchmark tasks. While novel approaches to learning node embeddings are highly suitable for node classification and link prediction, their application to graph classification (predicting a single label for the entire graph) remains mostly rudimentary, typically using a single global pooling step to aggregate node features or a hand-designed, fixed heuristic for hierarchical coarsening of the graph structure. An important step towards ameliorating this is differentiable graph coarsening---the ability to reduce the size of the graph in an adaptive, data-dependent manner within a graph neural network pipeline, analogous to image downsampling within CNNs. However, the previous prominent approach to pooling has quadratic memory requirements during training and is therefore not scalable to large graphs. Here we combine several recent advances in graph neural network design to demonstrate that competitive hierarchical graph classification results are possible without sacrificing sparsity. Our results are verified on several established graph classification benchmarks, and highlight an important direction for future research in graph-based neural networks."
                },
                {
                    "title": "Janossy Pooling: Learning Deep Permutation-Invariant Functions for Variable-Size Inputs",
                    "abstract": "We consider a simple and overarching representation for permutation-invariant functions of sequences (or multiset functions). Our approach, which we call Janossy pooling, expresses a permutation-invariant function as the average of a permutation-sensitive function applied to all reorderings of the input sequence. This allows us to leverage the rich and mature literature on permutation-sensitive functions to construct novel and flexible permutation-invariant functions. If carried out naively, Janossy pooling can be computationally prohibitive. To allow computational tractability, we consider three kinds of approximations: canonical orderings of sequences, functions with $k$-order interactions, and stochastic optimization algorithms with random permutations. Our framework unifies a variety of existing work in the literature, and suggests possible modeling and algorithmic extensions. We explore a few in our experiments, which demonstrate improved performance over current state-of-the-art methods."
                },
                {
                    "title": "Hierarchical Graph Representation Learning with Differentiable Pooling",
                    "abstract": "Recently, graph neural networks (GNNs) have revolutionized the field of graph representation learning through effectively learned node embeddings, and achieved state-of-the-art results in tasks such as node classification and link prediction. However, current GNN methods are inherently flat and do not learn hierarchical representations of graphs---a limitation that is especially problematic for the task of graph classification, where the goal is to predict the label associated with an entire graph. Here we propose DiffPool, a differentiable graph pooling module that can generate hierarchical representations of graphs and can be combined with various graph neural network architectures in an end-to-end fashion. DiffPool learns a differentiable soft cluster assignment for nodes at each layer of a deep GNN, mapping nodes to a set of clusters, which then form the coarsened input for the next GNN layer. Our experimental results show that combining existing GNN methods with DiffPool yields an average improvement of 5-10% accuracy on graph classification benchmarks, compared to all existing pooling approaches, achieving a new state-of-the-art on four out of five benchmark datasets."
                },
                {
                    "title": "Differentiable Submodular Maximization",
                    "abstract": "We consider learning of submodular functions from data. These functions are important in machine learning and have a wide range of applications, e.g. data summarization, feature selection and active learning. Despite their combinatorial nature, submodular functions can be maximized approximately with strong theoretical guarantees in polynomial time. Typically, learning the submodular function and optimization of that function are treated separately, i.e. the function is first learned using a proxy objective and subsequently maximized. In contrast, we show how to perform learning and optimization jointly. By interpreting the output of greedy maximization algorithms as distributions over sequences of items and smoothening these distributions, we obtain a differentiable objective. In this way, we can differentiate through the maximization algorithms and optimize the model to work well with the optimization algorithm. We theoretically characterize the error made by our approach, yielding insights into the tradeoff of smoothness and accuracy. We demonstrate the effectiveness of our approach for jointly learning and optimizing on synthetic maxcut data, and on real world applications such as product recommendation and image collection summarization."
                },
                {
                    "title": "DeepDTA: deep drug\u2013target binding affinity prediction",
                    "abstract": "Motivation The identification of novel drug\u2010target (DT) interactions is a substantial part of the drug discovery process. Most of the computational methods that have been proposed to predict DT interactions have focused on binary classification, where the goal is to determine whether a DT pair interacts or not. However, protein\u2010ligand interactions assume a continuum of binding strength values, also called binding affinity and predicting this value still remains a challenge. The increase in the affinity data available in DT knowledge\u2010bases allows the use of advanced learning techniques such as deep learning architectures in the prediction of binding affinities. In this study, we propose a deep\u2010learning based model that uses only sequence information of both targets and drugs to predict DT interaction binding affinities. The few studies that focus on DT binding affinity prediction use either 3D structures of protein\u2010ligand complexes or 2D features of compounds. One novel approach used in this work is the modeling of protein sequences and compound 1D representations with convolutional neural networks (CNNs). Results The results show that the proposed deep learning based model that uses the 1D representations of targets and drugs is an effective approach for drug target binding affinity prediction. The model in which high\u2010level representations of a drug and a target are constructed via CNNs achieved the best Concordance Index (CI) performance in one of our larger benchmark datasets, outperforming the KronRLS algorithm and SimBoost, a state\u2010of\u2010the\u2010art method for DT binding affinity prediction. Availability and implementation https://github.com/hkmztrk/DeepDTA Supplementary information Supplementary data are available at Bioinformatics online."
                },
                {
                    "title": "Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms",
                    "abstract": "We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at this https URL"
                },
                {
                    "title": "PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space",
                    "abstract": "Few prior works study deep learning on point sets. PointNet by Qi et al. is a pioneer in this direction. However, by design PointNet does not capture local structures induced by the metric space points live in, limiting its ability to recognize fine-grained patterns and generalizability to complex scenes. In this work, we introduce a hierarchical neural network that applies PointNet recursively on a nested partitioning of the input point set. By exploiting metric space distances, our network is able to learn local features with increasing contextual scales. With further observation that point sets are usually sampled with varying densities, which results in greatly decreased performance for networks trained on uniform densities, we propose novel set learning layers to adaptively combine features from multiple scales. Experiments show that our network called PointNet++ is able to learn deep point set features efficiently and robustly. In particular, results significantly better than state-of-the-art have been obtained on challenging benchmarks of 3D point clouds."
                },
                {
                    "title": "Atomic Convolutional Networks for Predicting Protein-Ligand Binding Affinity",
                    "abstract": "Empirical scoring functions based on either molecular force fields or cheminformatics descriptors are widely used, in conjunction with molecular docking, during the early stages of drug discovery to predict potency and binding affinity of a drug-like molecule to a given target. These models require expert-level knowledge of physical chemistry and biology to be encoded as hand-tuned parameters or features rather than allowing the underlying model to select features in a data-driven procedure. Here, we develop a general 3-dimensional spatial convolution operation for learning atomic-level chemical interactions directly from atomic coordinates and demonstrate its application to structure-based bioactivity prediction. The atomic convolutional neural network is trained to predict the experimentally determined binding affinity of a protein-ligand complex by direct calculation of the energy associated with the complex, protein, and ligand given the crystal structure of the binding pose. Non-covalent interactions present in the complex that are absent in the protein-ligand sub-structures are identified and the model learns the interaction strength associated with these features. We test our model by predicting the binding free energy of a subset of protein-ligand complexes found in the PDBBind dataset and compare with state-of-the-art cheminformatics and machine learning-based approaches. We find that all methods achieve experimental accuracy and that atomic convolutional networks either outperform or perform competitively with the cheminformatics based methods. Unlike all previous protein-ligand prediction systems, atomic convolutional networks are end-to-end and fully-differentiable. They represent a new data-driven, physics-based deep learning model paradigm that offers a strong foundation for future improvements in structure-based bioactivity prediction."
                },
                {
                    "title": "Deep Sets",
                    "abstract": "In this paper, we study the problem of designing objective functions for machine learning problems defined on finite \\emph{sets}. In contrast to traditional objective functions defined for machine learning problems operating on finite dimensional vectors, the new objective functions we propose are operating on finite sets and are invariant to permutations. Such problems are widespread, ranging from estimation of population statistics \\citep{poczos13aistats}, via anomaly detection in piezometer data of embankment dams \\citep{Jung15Exploration}, to cosmology \\citep{Ntampaka16Dynamical,Ravanbakhsh16ICML1}. Our main theorem characterizes the permutation invariant objective functions and provides a family of functions to which any permutation invariant objective function must belong. This family of functions has a special structure which enables us to design a deep network architecture that can operate on sets and which can be deployed on a variety of scenarios including both unsupervised and supervised learning tasks. We demonstrate the applicability of our method on population statistic estimation, point cloud classification, set expansion, and image tagging."
                },
                {
                    "title": "Equivariance Through Parameter-Sharing",
                    "abstract": "We propose to study equivariance in deep neural networks through parameter symmetries. In particular, given a group G that acts discretely on the input and output of a standard neural network layer \u03d5W : \u211dM \u2192 \u211dN, we show that \u03d5W is equivariant with respect to G-action iff G explains the symmetries of the network parameters W. Inspired by this observation, we then propose two parameter-sharing schemes to induce the desirable symmetry on W. Our procedure for tying the parameters achieves G-equivariance and, under some conditions on the action of G, it guarantees sensitivity to all other permutation groups outside G."
                },
                {
                    "title": "PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation",
                    "abstract": "Point cloud is an important type of geometric data structure. Due to its irregular format, most researchers transform such data to regular 3D voxel grids or collections of images. This, however, renders data unnecessarily voluminous and causes issues. In this paper, we design a novel type of neural network that directly consumes point clouds, which well respects the permutation invariance of points in the input. Our network, named PointNet, provides a unified architecture for applications ranging from object classification, part segmentation, to scene semantic parsing. Though simple, PointNet is highly efficient and effective. Empirically, it shows strong performance on par or even better than state of the art. Theoretically, we provide analysis towards understanding of what the network has learnt and why the network is robust with respect to input perturbation and corruption."
                },
                {
                    "title": "DeepSetNet: Predicting Sets with Deep Neural Networks",
                    "abstract": "This paper addresses the task of set prediction using deep learning. This is important because the output of many computer vision tasks, including image tagging and object detection, are naturally expressed as sets of entities rather than vectors. As opposed to a vector, the size of a set is not fixed in advance, and it is invariant to the ordering of entities within it. We define a likelihood for a set distribution and learn its parameters using a deep neural network. We also derive a loss for predicting a discrete distribution corresponding to set cardinality. Set prediction is demonstrated on the problem of multi-class image classification. Moreover, we show that the proposed cardinality loss can also trivially be applied to the tasks of object counting and pedestrian detection. Our approach outperforms existing methods in all three cases on standard datasets."
                },
                {
                    "title": "Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering",
                    "abstract": "In this work, we are interested in generalizing convolutional neural networks (CNNs) from low-dimensional regular grids, where image, video and speech are represented, to high-dimensional irregular domains, such as social networks, brain connectomes or words' embedding, represented by graphs. We present a formulation of CNNs in the context of spectral graph theory, which provides the necessary mathematical background and efficient numerical schemes to design fast localized convolutional filters on graphs. Importantly, the proposed technique offers the same linear computational complexity and constant learning complexity as classical CNNs, while being universal to any graph structure. Experiments on MNIST and 20NEWS demonstrate the ability of this novel deep learning system to learn local, stationary, and compositional features on graphs."
                },
                {
                    "title": "Towards a Neural Statistician",
                    "abstract": "An efficient learner is one who reuses what they already know to tackle a new problem. For a machine learner, this means understanding the similarities amongst datasets. In order to do this, one must take seriously the idea of working with datasets, rather than datapoints, as the key objects to model. Towards this goal, we demonstrate an extension of a variational autoencoder that can learn a method for computing representations, or statistics, of datasets in an unsupervised fashion. The network is trained to produce statistics that encapsulate a generative model for each dataset. Hence the network enables efficient learning from new datasets for both unsupervised and supervised tasks. We show that we are able to learn statistics that can be used for: clustering datasets, transferring generative models to new datasets, selecting representative samples of datasets and classifying previously unseen classes. We refer to our model as a neural statistician, and by this we mean a neural network that can learn to compute summary statistics of datasets without supervision."
                },
                {
                    "title": "Deep Decision Network for Multi-class Image Classification",
                    "abstract": "In this paper, we present a novel Deep Decision Network (DDN) that provides an alternative approach towards building an efficient deep learning network. During the learning phase, starting from the root network node, DDN automatically builds a network that splits the data into disjoint clusters of classes which would be handled by the subsequent expert networks. This results in a tree-like structured network driven by the data. The proposed method provides an insight into the data by identifying the group of classes that are hard to classify and require more attention when compared to others. DDN also has the ability to make early decisions thus making it suitable for timesensitive applications. We validate DDN on two publicly available benchmark datasets: CIFAR-10 and CIFAR-100 and it yields state-of-the-art classification performance on both the datasets. The proposed algorithm has no limitations to be applied to any generic classification problems."
                },
                {
                    "title": "PDB-wide collection of binding data: current status of the PDBbind database",
                    "abstract": "MOTIVATION\nMolecular recognition between biological macromolecules and organic small molecules plays an important role in various life processes. Both structural information and binding data of biomolecular complexes are indispensable for depicting the underlying mechanism in such an event. The PDBbind database was created to collect experimentally measured binding data for the biomolecular complexes throughout the Protein Data Bank (PDB). It thus provides the linkage between structural information and energetic properties of biomolecular complexes, which is especially desirable for computational studies or statistical analyses.\n\n\nRESULTS\nSince its first public release in 2004, the PDBbind database has been updated on an annual basis. The latest release (version 2013) provides experimental binding affinity data for 10,776 biomolecular complexes in PDB, including 8302 protein-ligand complexes and 2474 other types of complexes. In this article, we will describe the current methods used for compiling PDBbind and the updated status of this database. We will also review some typical applications of PDBbind published in the scientific literature.\n\n\nAVAILABILITY AND IMPLEMENTATION\nAll contents of this database are freely accessible at the PDBbind-CN Web server at http://www.pdbbind-cn.org/.\n\n\nCONTACT\nwangrx@mail.sioc.ac.cn.\n\n\nSUPPLEMENTARY INFORMATION\nSupplementary data are available at Bioinformatics online."
                },
                {
                    "title": "Deep Learning Face Attributes in the Wild",
                    "abstract": "Predicting face attributes in the wild is challenging due to complex face variations. We propose a novel deep learning framework for attribute prediction in the wild. It cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently. LNet is pre-trained by massive general object categories for face localization, while ANet is pre-trained by massive face identities for attribute prediction. This framework not only outperforms the state-of-the-art with a large margin, but also reveals valuable facts on learning face representation. (1) It shows how the performances of face localization (LNet) and attribute prediction (ANet) can be improved by different pre-training strategies. (2) It reveals that although the filters of LNet are fine-tuned only with image-level attribute tags, their response maps over entire images have strong indication of face locations. This fact enables training LNet for face localization with only image-level annotations, but without face bounding boxes or landmarks, which are required by all attribute recognition works. (3) It also demonstrates that the high-level hidden neurons of ANet automatically discover semantic concepts after pre-training with massive face identities, and such concepts are significantly enriched after fine-tuning with attribute tags. Each attribute can be well explained with a sparse linear combination of these concepts."
                },
                {
                    "title": "Expectation-Maximization for Learning Determinantal Point Processes",
                    "abstract": "A determinantal point process (DPP) is a probabilistic model of set diversity compactly parameterized by a positive semi-definite kernel matrix. To fit a DPP to a given task, we would like to learn the entries of its kernel matrix by maximizing the log-likelihood of the available data. However, log-likelihood is non-convex in the entries of the kernel matrix, and this learning problem is conjectured to be NP-hard [1]. Thus, previous work has instead focused on more restricted convex learning settings: learning only a single weight for each row of the kernel matrix [2], or learning weights for a linear combination of DPPs with fixed kernel matrices [3]. In this work we propose a novel algorithm for learning the full kernel matrix. By changing the kernel parameterization from matrix entries to eigenvalues and eigenvectors, and then lower-bounding the likelihood in the manner of expectation-maximization algorithms, we obtain an effective optimization procedure. We test our method on a real-world product recommendation task, and achieve relative gains of up to 16.5% in test log-likelihood compared to the naive approach of maximizing likelihood by projected gradient ascent on the entries of the kernel matrix."
                },
                {
                    "title": "Error-Driven Incremental Learning in Deep Convolutional Neural Network for Large-Scale Image Classification",
                    "abstract": "Supervised learning using deep convolutional neural network has shown its promise in large-scale image classification task. As a building block, it is now well positioned to be part of a larger system that tackles real-life multimedia tasks. An unresolved issue is that such model is trained on a static snapshot of data. Instead, this paper positions the training as a continuous learning process as new classes of data arrive. A system with such capability is useful in practical scenarios, as it gradually expands its capacity to predict increasing number of new classes. It is also our attempt to address the more fundamental issue: a good learning system must deal with new knowledge that it is exposed to, much as how human do. We developed a training algorithm that grows a network not only incrementally but also hierarchically. Classes are grouped according to similarities, and self-organized into levels. The newly added capacities are divided into component models that predict coarse-grained superclasses and those return final prediction within a superclass. Importantly, all models are cloned from existing ones and can be trained in parallel. These models inherit features from existing ones and thus further speed up the learning. Our experiment points out advantages of this approach, and also yields a few important open questions."
                },
                {
                    "title": "The conference paper",
                    "abstract": "Purpose - A satire on conference paper presentations. Design/methodology/approach - A poetic critique of conference presenters and their occasional failure to engage with their audience. Findings - What may be fascinating and of assumed and obvious importance to a research presenter may be quite obscure to their audience, who nonetheless participate in the ritual of conference attendance and polite \u201clistening\u201d. Research limitations/implications - The reader may discover even more meaning in this poem than its original author envisaged. Originality/value - A poignant reminder of the perils of many conference presentations."
                },
                {
                    "title": "Submodular Functions: Learnability, Structure, and Optimization",
                    "abstract": "Submodular functions are discrete functions that model laws of diminishing returns and enjoy numerous algorithmic applications. They have been used in many areas, including combinatorial optimization, machine learning, and economics. In this work we study submodular functions from a learning theoretic angle. We provide algorithms for learning submodular functions, as well as lower bounds on their learnability. In doing so, we uncover several novel structural results revealing ways in which submodular functions can be both surprisingly structured and surprisingly unstructured. We provide several concrete implications of our work in other domains including algorithmic game theory and combinatorial optimization. \nAt a technical level, this research combines ideas from many areas, including learning theory (distributional learning and PAC-style analyses), combinatorics and optimization (matroids and submodular functions), and pseudorandomness (lossless expander graphs)."
                },
                {
                    "title": "BindingDB: a web-accessible database of experimentally determined protein\u2013ligand binding affinities",
                    "abstract": "BindingDB () is a publicly accessible database currently containing \u223c20\u2009000 experimentally determined binding affinities of protein\u2013ligand complexes, for 110 protein targets including isoforms and mutational variants, and \u223c11\u2009000 small molecule ligands. The data are extracted from the scientific literature, data collection focusing on proteins that are drug-targets or candidate drug-targets and for which structural data are present in the Protein Data Bank. The BindingDB website supports a range of query types, including searches by chemical structure, substructure and similarity; protein sequence; ligand and protein names; affinity ranges and molecular weight. Data sets generated by BindingDB queries can be downloaded in the form of annotated SDfiles for further analysis, or used as the basis for virtual screening of a compound database uploaded by the user. The data in BindingDB are linked both to structural data in the PDB via PDB IDs and chemical and sequence searches, and to the literature in PubMed via PubMed IDs."
                },
                {
                    "title": "Application of the Radon-Nikodym Theorem to the Theory of Sufficient Statistics",
                    "abstract": "The body of this paper is written in terms of very general and abstract ideas which have been popular in pure mathematical work on the theory of probability for the last two or three decades. It seems to us that these ideas, so fruitful in pure mathematics, have something to contribute to mathematical statistics also, and this paper is an attempt to illustrate the sort of contribution we have in mind. The purpose of generality here is not to solve immediate practical problems, but rather to capture the logical essence of an important concept (sufficient statistic), and in particular to disentangle that concept from such ideas as Euclidean space, dimensionality, partial differentiation, and the distinction between continuous and discrete distributions, which seem to us extraneous. In accordance with these principles the center of the stage is occupied by a completely abstract sample space--that is a set $X$ of objects $x$, to be thought of as possible outcomes of an experimental program, distributed according to an unknown one of a certain set of probability measures. Perhaps the most familiar concrete example in statistics is the one in which $X$ is $n$ dimensional Cartesian space, the points of which represent $n$ independent observations of a normally distributed random variable with unknown parameters, and in which the probability measures considered are those induced by the various common normal distributions of the individual observations. A statistic is defined, as usual, to be a function $T$ of the outcome, whose values, however, are not necessarily real numbers but may themselves be abstract entities. Thus, in the concrete example, the entire set of $n$ observations, or, less trivially, the sequence of all sample moments about the origin are statistics with values in an $n$ dimensional and in an infinite dimensional space respectively. Another illuminating and very general example of a statistic may be obtained as follows. Suppose that the outcomes of two not necessarily statistically independent programs are thought of as one united outcome--then the outcome $T$ of the first program alone is a statistic relative to the united program. A technical measure theoretic result, known as the Radon-Nikodym theorem, is important in the study of statistics such as $T$. It is, for example, essential to the very definition of the basic concept of conditional probability of a subset $E$ of $X$ given a value $y$ of $T$. The statistic $T$ is called sufficient for the given set $\\mathcal{M}$ of probability measures if (somewhat loosely speaking) the conditional probability of a subset $E$ of $X$ given a value $y$ of $T$ is the same for every probability measure in $\\mathcal{M}$. It is, for instance, well known that the sample mean and variance together form a sufficient statistic for the measures described in the concrete example. The theory of sufficiency is in an especially satisfactory state for the case in which the set $\\mathcal{M}$ of probability measures satisfies a certain condition described by the technical term dominated. A set $\\mathcal{M}$ of probability measures is called dominated if each measure in the set may be expressed as the indefinite integral of a density function with respect to a fixed measure which is not itself necessarily in the set. It is easy to verify that both classical extremes, commonly referred to as the discrete and continuous cases, are dominated. One possible formulation of the principal result concerning sufficiency for dominated sets is a direct generalization to the abstract case of the well known Fisher-Neyman result: $T$ is sufficient if and only if the densities can be written as products of two factors, the first of which depends on the outcome through $T$ only and the second of which is independent of the unknown measure. Another way of phrasing this result is to say that $T$ is sufficient if and only if the likelihood ratio of every pair of measures in $\\mathcal{M}$ depends on the outcome through $T$ only. The latter formulation makes sense even in the not necessarily dominated case but unfortunately it is not true in that case. The situation can be patched up somewhat by introducing a weaker notion called pairwise sufficiency. In ordinary statistical parlance one often speaks of a statistic sufficient for some of several parameters. The abstract results mentioned above can undoubtedly be extended to treat this concept."
                },
                {
                    "title": "Universal Mini-Batch Consistency for Set Encoding Functions",
                    "abstract": "Previous works have established solid foundations for neural set functions, as well as effective architectures which preserve the necessary properties for operating on sets, such as being invariant to permutations of the set elements. Subsequently, Mini-Batch Consistency (MBC), the ability to sequentially process any permutation of any random set partition scheme while maintaining consistency guarantees on the output, has been established but with limited options for network architectures. We further study the MBC property in neural set encoding functions, establishing a method for converting arbitrary non-MBC models to satisfy MBC. In doing so, we provide a framework for a universally-MBC (UMBC) class of set functions. Additionally, we explore an interesting dropout strategy made possible by our framework, and investigate its effects on probabilistic calibration under test-time distributional shifts. We validate UMBC with proofs backed by unit tests, also providing qualitative/quantitative experiments on toy data, clean and corrupted point cloud classi\ufb01cation, and amortized clustering on ImageNet. The results demonstrate the utility of UMBC, and we further discover that our dropout strategy improves uncertainty calibration."
                },
                {
                    "title": "Differentiable Expectation-Maximization for Set Representation Learning",
                    "abstract": "We tackle the set2vec problem, the task of extracting a vector representation from an input set comprised of a variable number of feature vectors. Although recent approaches based on self attention such as (Set)Transformers were very successful due to the capability of capturing complex interaction between set elements, the computational overhead is the well-known downside. The inducing-point attention and the latest optimal transport kernel embedding (OTKE) are promising remedies that attain comparable or better performance with reduced computational cost, by incorporating a fixed number of learnable queries in attention. In this paper we approach the set2vec problem from a completely different perspective. The elements of an input set are considered as i.i.d. samples from a mixture distribution, and we define our set embedding feed-forward network as the maximum-a-posterior (MAP) estimate of the mixture which is approximately attained by a few ExpectationMaximization (EM) steps. The whole MAP-EM steps are differentiable operations with a fixed number of mixture parameters, allowing efficient auto-diff backpropagation for any given downstream task. Furthermore, the proposed mixture set data fitting framework allows unsupervised set representation learning naturally via marginal likelihood maximization aka the empirical Bayes. Interestingly, we also find that OTKE can be seen as a special case of our framework, specifically a single-step EM with extra balanced assignment constraints on the E-step. Compared to OTKE, our approach provides more flexible set embedding as well as prior-induced model regularization. We evaluate our approach on various tasks demonstrating improved performance over the state-of-the-arts."
                },
                {
                    "title": "Probabilistic Symmetries and Invariant Neural Networks",
                    "abstract": "Treating neural network inputs and outputs as random variables, we characterize the structure of neural networks that can be used to model data that are invariant or equivariant under the action of a compact group. Much recent research has been devoted to encoding invariance under symmetry transformations into neural network architectures, in an effort to improve the performance of deep neural networks in data-scarce, non-i.i.d., or unsupervised settings. By considering group invariance from the perspective of probabilistic symmetry, we establish a link between functional and probabilistic symmetry, and obtain generative functional representations of probability distributions that are invariant or equivariant under the action of a compact group. Our representations completely characterize the structure of neural networks that can be used to model such distributions and yield a general program for constructing invariant stochastic or deterministic neural networks. We demonstrate that examples from the recent literature are special cases, and develop the details of the general program for exchangeable sequences and arrays."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively learn a utility function for set-valued outputs in tasks such as compound selection from a database, given the challenges of limited supervision and the need for implicit learning approaches?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the field of machine learning, particularly in applications like drug discovery and anomaly detection, where set-valued outputs are common. By developing methods that can learn utility functions with limited supervision, we can enhance the efficiency and effectiveness of these applications, leading to better decision-making processes. This research could pave the way for future studies that explore more complex set functions and their applications across various domains, ultimately contributing to the development of more robust machine learning models.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the need to model utility functions accurately while dealing with limited and potentially noisy data. Naive approaches that rely on explicit modeling of utility may fail due to the high cost of obtaining sufficient supervision signals, which can be prohibitively expensive. Additionally, the complexity of set-valued outputs introduces combinatorial challenges, as the number of possible subsets grows exponentially with the size of the input set. Overcoming these technical and practical obstacles requires innovative methodologies that can effectively leverage implicit learning and probabilistic modeling.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on explicit utility modeling, which necessitates extensive labeled data that is difficult to obtain in practice. This reliance on large amounts of supervision has limited the applicability of existing methods. Additionally, many approaches have not adequately addressed the combinatorial nature of set selection problems. Our approach differs by utilizing implicit learning techniques and probabilistic frameworks that can work with limited supervision, thereby addressing the gaps in prior work and providing a more scalable solution to the problem.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves estimating the parameters of a utility function using an implicit learning approach, where we maximize the empirical log likelihood of optimal subsets given the compound database. We will utilize a dataset consisting of compound databases and their corresponding optimal subsets, and we will evaluate our method using metrics such as the accuracy of the selected subsets and the utility values predicted by the learned function. The expected outcomes include a more efficient learning process for utility functions and improved performance in set-valued output tasks, particularly in compound selection scenarios."
            }
        },
        "author_data": {
            "23070752-5d2c-4e6f-9b97-3ea2b22f0236": {
                "pk": "23070752-5d2c-4e6f-9b97-3ea2b22f0236",
                "name": "Binghui Xie",
                "collaborators": [
                    "James Cheng",
                    "Bo Han",
                    "Yongqiang Chen",
                    "Kaiwen Zhou",
                    "Yatao Bian",
                    "Kaili Ma",
                    "Yonggang Zhang",
                    "Han Yang",
                    "Wei Meng",
                    "Tongliang Liu"
                ],
                "domain": [
                    "Domain Generalization",
                    "Out-of-Distribution Generalization",
                    "Graph Learning",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "Enhancing Evolving Domain Generalization through Dynamic Latent Representations",
                        "abstract": "Domain generalization is a critical challenge for machine learning systems. Prior domain generalization methods focus on extracting domain-invariant features across several stationary domains to enable generalization to new domains. However, in non-stationary tasks where new domains evolve in an underlying continuous structure, such as time, merely extracting the invariant features is insufficient for generalization to the evolving new domains. Nevertheless, it is non-trivial to learn both evolving and invariant features within a single model due to their conflicts. To bridge this gap, we build causal models to characterize the distribution shifts concerning the two patterns, and propose to learn both dynamic and invariant features via a new framework called Mutual Information-Based Sequential Autoencoders (MISTS). MISTS adopts information theoretic constraints onto sequential autoencoders to disentangle the dynamic and invariant features, and leverage an adaptive classifier to make predictions based on both evolving and invariant information. Our experimental results on both synthetic and real-world datasets demonstrate that MISTS succeeds in capturing both evolving and invariant information, and present promising results in evolving domain generalization tasks."
                    },
                    {
                        "title": "Does Invariant Graph Learning via Environment Augmentation Learn Invariance?",
                        "abstract": "Invariant graph representation learning aims to learn the invariance among data from different environments for out-of-distribution generalization on graphs. As the graph environment partitions are usually expensive to obtain, augmenting the environment information has become the de facto approach. However, the usefulness of the augmented environment information has never been verified. In this work, we find that it is fundamentally impossible to learn invariant graph representations via environment augmentation without additional assumptions. Therefore, we develop a set of minimal assumptions, including variation sufficiency and variation consistency, for feasible invariant graph learning. We then propose a new framework Graph invAriant Learning Assistant (GALA). GALA incorporates an assistant model that needs to be sensitive to graph environment changes or distribution shifts. The correctness of the proxy predictions by the assistant model hence can differentiate the variations in spurious subgraphs. We show that extracting the maximally invariant subgraph to the proxy predictions provably identifies the underlying invariant subgraph for successful OOD generalization under the established minimal assumptions. Extensive experiments on datasets including DrugOOD with various graph distribution shifts confirm the effectiveness of GALA."
                    },
                    {
                        "title": "Positional Information Matters for Invariant In-Context Learning: A Case Study of Simple Function Classes",
                        "abstract": "In-context learning (ICL) refers to the ability of a model to condition on a few in-context demonstrations (input-output examples of the underlying task) to generate the answer for a new query input, without updating parameters. Despite the impressive ICL ability of LLMs, it has also been found that ICL in LLMs is sensitive to input demonstrations and limited to short context lengths. To understand the limitations and principles for successful ICL, we conduct an investigation with ICL linear regression of transformers. We characterize several Out-of-Distribution (OOD) cases for ICL inspired by realistic LLM ICL failures and compare transformers with DeepSet, a simple yet powerful architecture for ICL. Surprisingly, DeepSet outperforms transformers across a variety of distribution shifts, implying that preserving permutation invariance symmetry to input demonstrations is crucial for OOD ICL. The phenomenon specifies a fundamental requirement by ICL, which we termed as ICL invariance. Nevertheless, the positional encodings in LLMs will break ICL invariance. To this end, we further evaluate transformers with identical positional encodings and find preserving ICL invariance in transformers achieves state-of-the-art performance across various ICL distribution shifts"
                    },
                    {
                        "title": "Fast and Reliable Evaluation of Adversarial Robustness with Minimum-Margin Attack",
                        "abstract": "The AutoAttack (AA) has been the most reliable method to evaluate adversarial robustness when considerable computational resources are available. However, the high computational cost (e.g., 100 times more than that of the project gradient descent attack) makes AA infeasible for practitioners with limited computational resources, and also hinders applications of AA in the adversarial training (AT). In this paper, we propose a novel method, minimum-margin (MM) attack, to fast and reliably evaluate adversarial robustness. Compared with AA, our method achieves comparable performance but only costs 3% of the computational time in extensive experiments. The reliability of our method lies in that we evaluate the quality of adversarial examples using the margin between two targets that can precisely identify the most adversarial example. The computational efficiency of our method lies in an effective Sequential TArget Ranking Selection (STARS) method, ensuring that the cost of the MM attack is independent of the number of classes. The MM attack opens a new way for evaluating adversarial robustness and provides a feasible and reliable way to generate high-quality adversarial examples in AT."
                    },
                    {
                        "title": "Pareto Invariant Risk Minimization",
                        "abstract": "Despite the success of invariant risk minimization (IRM) in tackling the Out-of-Distribution generalization problem, IRM can compromise the optimality when applied in practice. The practical variants of IRM, e.g., IRMv1, have been shown to have signi\ufb01cant gaps with IRM and thus could fail to capture the invariance even in simple problems. Moreover, the optimization procedure in IRMv1 involves two intrinsically con\ufb02icting objectives, and often requires careful tuning for the objec-tive weights. To remedy the above issues, we reformulate IRM as a multi-objective optimization problem, and propose a new optimization scheme for IRM, called PA reto I nvariant R isk Minimization ( PAIR ). PAIR can adaptively adjust the optimization direction under the objective con\ufb02icts. Furthermore, we show PAIR can empower the practical IRM variants to overcome the barriers with the original IRM when provided with proper guidance. We conduct experiments with ColoredMNIST to con\ufb01rm our theory and the effectiveness of PAIR ."
                    },
                    {
                        "title": "An Adaptive Incremental Gradient Method With Support for Non-Euclidean Norms",
                        "abstract": "Stochastic variance reduced methods have shown strong performance in solving finite-sum problems. However, these methods usually require the users to manually tune the step-size, which is time-consuming or even infeasible for some large-scale optimization tasks. To overcome the problem, we propose and analyze several novel adaptive variants of the popular SAGA algorithm. Eventually, we design a variant of Barzilai-Borwein step-size which is tailored for the incremental gradient method to ensure memory efficiency and fast convergence. We establish its convergence guarantees under general settings that allow non-Euclidean norms in the definition of smoothness and the composite objectives, which cover a broad range of applications in machine learning. We improve the analysis of SAGA to support non-Euclidean norms, which fills the void of existing work. Numerical experiments on standard datasets demonstrate a competitive performance of the proposed algorithm compared with existing variance-reduced methods and their adaptive variants."
                    },
                    {
                        "title": "Invariance Principle Meets Out-of-Distribution Generalization on Graphs",
                        "abstract": "Despite recent success in using the invariance principle for out-of-distribution (OOD) generalization on Euclidean data (e.g., images), studies on graph data are still limited. Different from images, the complex nature of graphs poses unique challenges to adopting the invariance principle. In particular, distribution shifts on graphs can appear in a variety of forms such as attributes and structures, making it dif\ufb01cult to identify the invariance. Moreover, domain or environment partitions, which are often required by OOD methods on Euclidean data, could be highly expensive to obtain for graphs. To bridge this gap, we pro-pose a new framework to capture the invariance of graphs for guaranteed OOD generalization under various distribution shifts. Speci\ufb01cally, we characterize potential distribution shifts on graphs with causal models, concluding that OOD generalization on graphs is achievable when models focus only on subgraphs containing the most information about the causes of labels. Accordingly, we propose an information-theoretic objective to extract the desired subgraphs that maximally preserve the invariant intra-class information. Learning with these subgraphs is immune to distribution shifts. Extensive experiments on both synthetic and real-world datasets, including a challenging setting in AI-aided drug discovery, validate the superior OOD generalization ability of our method."
                    },
                    {
                        "title": "Pareto Invariant Risk Minimization: Towards Mitigating the Optimization Dilemma in Out-of-Distribution Generalization",
                        "abstract": "Recently, there has been a growing surge of interest in enabling machine learning systems to generalize well to Out-of-Distribution (OOD) data. Most efforts are devoted to advancing optimization objectives that regularize models to capture the underlying invariance; however, there often are compromises in the optimization process of these OOD objectives: i) Many OOD objectives have to be relaxed as penalty terms of Empirical Risk Minimization (ERM) for the ease of optimization, while the relaxed forms can weaken the robustness of the original objective; ii) The penalty terms also require careful tuning of the penalty weights due to the intrinsic conflicts between ERM and OOD objectives. Consequently, these compromises could easily lead to suboptimal performance of either the ERM or OOD objective. To address these issues, we introduce a multi-objective optimization (MOO) perspective to understand the OOD optimization process, and propose a new optimization scheme called PAreto Invariant Risk Minimization (PAIR). PAIR improves the robustness of OOD objectives by cooperatively optimizing with other OOD objectives, thereby bridging the gaps caused by the relaxations. Then PAIR approaches a Pareto optimal solution that trades off the ERM and OOD objectives properly. Extensive experiments on challenging benchmarks, WILDS, show that PAIR alleviates the compromises and yields top OOD performances."
                    },
                    {
                        "title": "Learning Causally Invariant Representations for Out-of-Distribution Generalization on Graphs",
                        "abstract": "Despite recent success in using the invariance principle for out-of-distribution (OOD) generalization on Euclidean data (e.g., images), studies on graph data are still limited. Different from images, the complex nature of graphs poses unique challenges to adopting the invariance principle. In particular, distribution shifts on graphs can appear in a variety of forms such as attributes and structures, making it difficult to identify the invariance. Moreover, domain or environment partitions, which are often required by OOD methods on Euclidean data, could be highly expensive to obtain for graphs. To bridge this gap, we propose a new framework, called Causality Inspired Invariant Graph LeArning (CIGA), to capture the invariance of graphs for guaranteed OOD generalization under various distribution shifts. Specifically, we characterize potential distribution shifts on graphs with causal models, concluding that OOD generalization on graphs is achievable when models focus only on subgraphs containing the most information about the causes of labels. Accordingly, we propose an information-theoretic objective to extract the desired subgraphs that maximally preserve the invariant intra-class information. Learning with these subgraphs is immune to distribution shifts. Extensive experiments on 16 synthetic or real-world datasets, including a challenging setting -- DrugOOD, from AI-aided drug discovery, validate the superior OOD performance of CIGA."
                    }
                ]
            },
            "0bf8ccb8-fdf3-4b2f-9bc0-7e0bec3b48bc": {
                "pk": "0bf8ccb8-fdf3-4b2f-9bc0-7e0bec3b48bc",
                "name": "Yatao Bian",
                "collaborators": [
                    "Yongqiang Chen",
                    "Quanming Yao",
                    "James Cheng",
                    "Bo Han",
                    "Juzheng Zhang",
                    "Kaiwen Zhou",
                    "Binghui Xie",
                    "Haiquan Qiu",
                    "Shiguang Wu",
                    "Yaqing Wang"
                ],
                "domain": [
                    "Large Language Models",
                    "Graph Neural Network",
                    "Meta-Learning",
                    "Invariant Learning"
                ],
                "publications": [
                    {
                        "title": "UniMoT: Unified Molecule-Text Language Model with Discrete Token Representation",
                        "abstract": "The remarkable success of Large Language Models (LLMs) across diverse tasks has driven the research community to extend their capabilities to molecular applications. However, most molecular LLMs employ adapter-based architectures that do not treat molecule and text modalities equally and lack a supervision signal for the molecule modality. To address these issues, we introduce UniMoT, a Unified Molecule-Text LLM adopting a tokenizer-based architecture that expands the vocabulary of LLM with molecule tokens. Specifically, we introduce a Vector Quantization-driven tokenizer that incorporates a Q-Former to bridge the modality gap between molecule and text. This tokenizer transforms molecules into sequences of molecule tokens with causal dependency, encapsulating high-level molecular and textual information. Equipped with this tokenizer, UniMoT can unify molecule and text modalities under a shared token representation and an autoregressive training paradigm, enabling it to interpret molecules as a foreign language and generate them as text. Following a four-stage training scheme, UniMoT emerges as a multi-modal generalist capable of performing both molecule-to-text and text-to-molecule tasks. Extensive experiments demonstrate that UniMoT achieves state-of-the-art performance across a wide range of molecule comprehension and generation tasks."
                    },
                    {
                        "title": "Graph Unitary Message Passing",
                        "abstract": "Message passing mechanism contributes to the success of GNNs in various applications, but also brings the oversquashing problem. Recent works combat oversquashing by improving the graph spectrums with rewiring techniques, disrupting the structural bias in graphs, and having limited improvement on oversquashing in terms of oversquashing measure. Motivated by unitary RNN, we propose Graph Unitary Message Passing (GUMP) to alleviate oversquashing in GNNs by applying unitary adjacency matrix for message passing. To design GUMP, a transformation is first proposed to make general graphs have unitary adjacency matrix and keep its structural bias. Then, unitary adjacency matrix is obtained with a unitary projection algorithm, which is implemented by utilizing the intrinsic structure of unitary adjacency matrix and allows GUMP to be permutation-equivariant. Experimental results show the effectiveness of GUMP in improving the performance on various graph learning tasks."
                    },
                    {
                        "title": "How Interpretable Are Interpretable Graph Neural Networks?",
                        "abstract": "Interpretable graph neural networks (XGNNs ) are widely adopted in various scientific applications involving graph-structured data. Existing XGNNs predominantly adopt the attention-based mechanism to learn edge or node importance for extracting and making predictions with the interpretable subgraph. However, the representational properties and limitations of these methods remain inadequately explored. In this work, we present a theoretical framework that formulates interpretable subgraph learning with the multilinear extension of the subgraph distribution, coined as subgraph multilinear extension (SubMT). Extracting the desired interpretable subgraph requires an accurate approximation of SubMT, yet we find that the existing XGNNs can have a huge gap in fitting SubMT. Consequently, the SubMT approximation failure will lead to the degenerated interpretability of the extracted subgraphs. To mitigate the issue, we design a new XGNN architecture called Graph Multilinear neT (GMT), which is provably more powerful in approximating SubMT. We empirically validate our theoretical findings on a number of graph classification benchmarks. The results demonstrate that GMT outperforms the state-of-the-art up to 10% in terms of both interpretability and generalizability across 12 regular and geometric graph benchmarks."
                    },
                    {
                        "title": "HIGHT: Hierarchical Graph Tokenization for Graph-Language Alignment",
                        "abstract": "Recently there has been a surge of interest in extending the success of large language models (LLMs) to graph modality, such as social networks and molecules. As LLMs are predominantly trained with 1D text data, most existing approaches adopt a graph neural network to represent a graph as a series of node tokens and feed these tokens to LLMs for graph-language alignment. Despite achieving some successes, existing approaches have overlooked the hierarchical structures that are inherent in graph data. Especially, in molecular graphs, the high-order structural information contains rich semantics of molecular functional groups, which encode crucial biochemical functionalities of the molecules. We establish a simple benchmark showing that neglecting the hierarchical information in graph tokenization will lead to subpar graph-language alignment and severe hallucination in generated outputs. To address this problem, we propose a novel strategy called HIerarchical GrapH Tokenization (HIGHT). HIGHT employs a hierarchical graph tokenizer that extracts and encodes the hierarchy of node, motif, and graph levels of informative tokens to improve the graph perception of LLMs. HIGHT also adopts an augmented graph-language supervised fine-tuning dataset, enriched with the hierarchical graph information, to further enhance the graph-language alignment. Extensive experiments on 7 molecule-centric benchmarks confirm the effectiveness of HIGHT in reducing hallucination by 40%, as well as significant improvements in various molecule-language downstream tasks."
                    },
                    {
                        "title": "ConML: A Universal Meta-Learning Framework with Task-Level Contrastive Learning",
                        "abstract": "Meta-learning enables learning systems to adapt quickly to new tasks, similar to humans. To emulate this human-like rapid learning and enhance alignment and discrimination abilities, we propose ConML, a universal meta-learning framework that can be applied to various meta-learning algorithms without relying on specific model architectures nor target models. The core of ConML is task-level contrastive learning, which extends contrastive learning from the representation space in unsupervised learning to the model space in meta-learning. By leveraging task identity as an additional supervision signal during meta-training, we contrast the outputs of the meta-learner in the model space, minimizing inner-task distance (between models trained on different subsets of the same task) and maximizing inter-task distance (between models from different tasks). We demonstrate that ConML integrates seamlessly with optimization-based, metric-based, and amortization-based meta-learning algorithms, as well as in-context learning, resulting in performance improvements across diverse few-shot learning tasks."
                    },
                    {
                        "title": "Does Invariant Graph Learning via Environment Augmentation Learn Invariance?",
                        "abstract": "Invariant graph representation learning aims to learn the invariance among data from different environments for out-of-distribution generalization on graphs. As the graph environment partitions are usually expensive to obtain, augmenting the environment information has become the de facto approach. However, the usefulness of the augmented environment information has never been verified. In this work, we find that it is fundamentally impossible to learn invariant graph representations via environment augmentation without additional assumptions. Therefore, we develop a set of minimal assumptions, including variation sufficiency and variation consistency, for feasible invariant graph learning. We then propose a new framework Graph invAriant Learning Assistant (GALA). GALA incorporates an assistant model that needs to be sensitive to graph environment changes or distribution shifts. The correctness of the proxy predictions by the assistant model hence can differentiate the variations in spurious subgraphs. We show that extracting the maximally invariant subgraph to the proxy predictions provably identifies the underlying invariant subgraph for successful OOD generalization under the established minimal assumptions. Extensive experiments on datasets including DrugOOD with various graph distribution shifts confirm the effectiveness of GALA."
                    },
                    {
                        "title": "Positional Information Matters for Invariant In-Context Learning: A Case Study of Simple Function Classes",
                        "abstract": "In-context learning (ICL) refers to the ability of a model to condition on a few in-context demonstrations (input-output examples of the underlying task) to generate the answer for a new query input, without updating parameters. Despite the impressive ICL ability of LLMs, it has also been found that ICL in LLMs is sensitive to input demonstrations and limited to short context lengths. To understand the limitations and principles for successful ICL, we conduct an investigation with ICL linear regression of transformers. We characterize several Out-of-Distribution (OOD) cases for ICL inspired by realistic LLM ICL failures and compare transformers with DeepSet, a simple yet powerful architecture for ICL. Surprisingly, DeepSet outperforms transformers across a variety of distribution shifts, implying that preserving permutation invariance symmetry to input demonstrations is crucial for OOD ICL. The phenomenon specifies a fundamental requirement by ICL, which we termed as ICL invariance. Nevertheless, the positional encodings in LLMs will break ICL invariance. To this end, we further evaluate transformers with identical positional encodings and find preserving ICL invariance in transformers achieves state-of-the-art performance across various ICL distribution shifts"
                    }
                ]
            },
            "2fef46f7-5cda-4669-91cf-479acb027662": {
                "pk": "2fef46f7-5cda-4669-91cf-479acb027662",
                "name": "Kaiwen zhou",
                "collaborators": [
                    "James Cheng",
                    "Bo Han",
                    "Binghui Xie",
                    "Yongqiang Chen",
                    "Yatao Bian",
                    "Wei Meng",
                    "Ruize Gao",
                    "Feng Liu",
                    "Gang Niu",
                    "Kaili Ma"
                ],
                "domain": [
                    "Domain Generalization",
                    "Machine Learning",
                    "Adversarial Robustness",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Enhancing Evolving Domain Generalization through Dynamic Latent Representations",
                        "abstract": "Domain generalization is a critical challenge for machine learning systems. Prior domain generalization methods focus on extracting domain-invariant features across several stationary domains to enable generalization to new domains. However, in non-stationary tasks where new domains evolve in an underlying continuous structure, such as time, merely extracting the invariant features is insufficient for generalization to the evolving new domains. Nevertheless, it is non-trivial to learn both evolving and invariant features within a single model due to their conflicts. To bridge this gap, we build causal models to characterize the distribution shifts concerning the two patterns, and propose to learn both dynamic and invariant features via a new framework called Mutual Information-Based Sequential Autoencoders (MISTS). MISTS adopts information theoretic constraints onto sequential autoencoders to disentangle the dynamic and invariant features, and leverage an adaptive classifier to make predictions based on both evolving and invariant information. Our experimental results on both synthetic and real-world datasets demonstrate that MISTS succeeds in capturing both evolving and invariant information, and present promising results in evolving domain generalization tasks."
                    },
                    {
                        "title": "Genomic and transcriptomic profiling of inflammatory breast cancers and non-inflammatory breast cancers in Asian women.",
                        "abstract": "e12546 Background: Inflammatory breast cancer (IBC) is the most aggressive form of breast cancer. Despite its rare incidence, it accounts for 10% of breast cancer-caused mortality. Previous studies based on NGS have focused on western populations and found no common genomic and transcriptomic difference that could distinguish IBC from non-IBC. Methods: We performed WES on 15 IBC samples and their paired normal blood to identify somatic genomic alterations, copy number variants and large structural variants. RNA-seq was performed on 16 IBC samples to examine the RNA expression, pathway enrichment and gene fusions. WES and RNA-seq data were further investigated to discover genes that were dysregulated in both genomics and transcriptomics. For comparison, we performed WES and RNA-seq on 5 non-IBC samples and acquired additional 33 non-IBC samples of Asian women from the Cancer Genome Atlas Program (TCGA). Results: Apart from results similar to the previous reports in western IBC, ZNF74 and DYNC2H1 were identified as Asian-IBC-specific somatic mutations in our cohort. Elevated HRD mutational signature and scarHRD score indicated increased HRD events in Asian IBC patients. Structural variation analysis showed that deletions were found more frequent in IBC. RNA-seq indicated increased immune activation in IBC samples, but the immune cell infiltration within the tumor microenvironment were similar between IBC and non-IBC. Multiple gene fusions were more frequently found in IBC, including CTC-786C10.1\u2014RP11-680G10.1, TULP4\u2014RP11-732M18.3, RN7SL4P\u2014RN7SL173P. Combined analysis of WES and RNA-seq showed more frequent mutation and increased pathway enrichment of RAS pathway in IBC, which may explain the aggressive biological behavior of IBC, inhibitors such as sorafenib, tipifarnib and trametinib targeting RAS signaling and its downstream pathways were predicted to possess promising effects in IBC treatment. Conclusions: Our study is the first to provide a complete genomic and transcriptomic landscape of IBC based on combined analysis of WES and RNA-seq and the first to focus on Asian women. We discovered intriguing differences unique in Asian women that could potentially explain IBC etiology and presented RAS signaling pathway as a potential therapeutic target in IBC treatment."
                    },
                    {
                        "title": "Does Invariant Graph Learning via Environment Augmentation Learn Invariance?",
                        "abstract": "Invariant graph representation learning aims to learn the invariance among data from different environments for out-of-distribution generalization on graphs. As the graph environment partitions are usually expensive to obtain, augmenting the environment information has become the de facto approach. However, the usefulness of the augmented environment information has never been verified. In this work, we find that it is fundamentally impossible to learn invariant graph representations via environment augmentation without additional assumptions. Therefore, we develop a set of minimal assumptions, including variation sufficiency and variation consistency, for feasible invariant graph learning. We then propose a new framework Graph invAriant Learning Assistant (GALA). GALA incorporates an assistant model that needs to be sensitive to graph environment changes or distribution shifts. The correctness of the proxy predictions by the assistant model hence can differentiate the variations in spurious subgraphs. We show that extracting the maximally invariant subgraph to the proxy predictions provably identifies the underlying invariant subgraph for successful OOD generalization under the established minimal assumptions. Extensive experiments on datasets including DrugOOD with various graph distribution shifts confirm the effectiveness of GALA."
                    },
                    {
                        "title": "Modeling Wildfire Initial Attack Success Rate Based on Machine Learning in Liangshan, China",
                        "abstract": "The initial attack is a critical phase in firefighting efforts, where the first batch of resources are deployed to prevent the spread of the fire. This study aimed to analyze and understand the factors that impact the success of the initial attack, and used three machine learning models\u2014logistic regression, XGBoost, and artificial neural network\u2014to simulate the success rate of the initial attack in a specific region. The performance of each machine learning model was evaluated based on accuracy, AUC (Area Under the Curve), and F1 Score, with the results showing that the XGBoost model performed the best. In addition, the study also considered the impact of weather conditions on the initial attack success rate by dividing the scenario into normal weather and extreme weather conditions. This information can be useful for forest fire managers as they plan resource allocation, with the goal of improving the success rate of the initial attack in the area."
                    },
                    {
                        "title": "A Novel Extrapolation Technique to Accelerate WMMSE",
                        "abstract": "Precoding design is essential for massive multi-user multiple-input multiple-output (MU-MIMO) systems, which aims at maximizing the weighted sum-rate (WSR). This problem is known to be NP-hard, and iterative algorithms are typically used to approximately solve it. The weighted minimum mean-squared error (WMMSE) algorithm is a popular solver for WSR maximization, which efficiently finds a local maxima of WSR. In this work, we introduce a novel extrapolation technique to further accelerate WMMSE. This technique is inspired by the momentum technique in convex optimization, and can be interpreted as an accelerated second-order method. The merits of the proposed extrapolation technique are (i) lightweight, as it almost does not increase the iteration complexity, (ii) generic, since it works in various settings such as the sum power constraint or per-antenna power constraint cases and coordinated multi-point joint transmission networks, and (iii) effective, that our simulation results show it significantly accelerates the convergence of WMMSE in the high channel correlation regime."
                    },
                    {
                        "title": "Positional Information Matters for Invariant In-Context Learning: A Case Study of Simple Function Classes",
                        "abstract": "In-context learning (ICL) refers to the ability of a model to condition on a few in-context demonstrations (input-output examples of the underlying task) to generate the answer for a new query input, without updating parameters. Despite the impressive ICL ability of LLMs, it has also been found that ICL in LLMs is sensitive to input demonstrations and limited to short context lengths. To understand the limitations and principles for successful ICL, we conduct an investigation with ICL linear regression of transformers. We characterize several Out-of-Distribution (OOD) cases for ICL inspired by realistic LLM ICL failures and compare transformers with DeepSet, a simple yet powerful architecture for ICL. Surprisingly, DeepSet outperforms transformers across a variety of distribution shifts, implying that preserving permutation invariance symmetry to input demonstrations is crucial for OOD ICL. The phenomenon specifies a fundamental requirement by ICL, which we termed as ICL invariance. Nevertheless, the positional encodings in LLMs will break ICL invariance. To this end, we further evaluate transformers with identical positional encodings and find preserving ICL invariance in transformers achieves state-of-the-art performance across various ICL distribution shifts"
                    },
                    {
                        "title": "Understanding and Improving Feature Learning for Out-of-Distribution Generalization",
                        "abstract": "A common explanation for the failure of out-of-distribution (OOD) generalization is that the model trained with empirical risk minimization (ERM) learns spurious features instead of invariant features. However, several recent studies challenged this explanation and found that deep networks may have already learned sufficiently good features for OOD generalization. Despite the contradictions at first glance, we theoretically show that ERM essentially learns both spurious and invariant features, while ERM tends to learn spurious features faster if the spurious correlation is stronger. Moreover, when fed the ERM learned features to the OOD objectives, the invariant feature learning quality significantly affects the final OOD performance, as OOD objectives rarely learn new features. Therefore, ERM feature learning can be a bottleneck to OOD generalization. To alleviate the reliance, we propose Feature Augmented Training (FeAT), to enforce the model to learn richer features ready for OOD generalization. FeAT iteratively augments the model to learn new features while retaining the already learned features. In each round, the retention and augmentation operations are performed on different subsets of the training data that capture distinct features. Extensive experiments show that FeAT effectively learns richer features thus boosting the performance of various OOD objectives."
                    },
                    {
                        "title": "Fast and Reliable Evaluation of Adversarial Robustness with Minimum-Margin Attack",
                        "abstract": "The AutoAttack (AA) has been the most reliable method to evaluate adversarial robustness when considerable computational resources are available. However, the high computational cost (e.g., 100 times more than that of the project gradient descent attack) makes AA infeasible for practitioners with limited computational resources, and also hinders applications of AA in the adversarial training (AT). In this paper, we propose a novel method, minimum-margin (MM) attack, to fast and reliably evaluate adversarial robustness. Compared with AA, our method achieves comparable performance but only costs 3% of the computational time in extensive experiments. The reliability of our method lies in that we evaluate the quality of adversarial examples using the margin between two targets that can precisely identify the most adversarial example. The computational efficiency of our method lies in an effective Sequential TArget Ranking Selection (STARS) method, ensuring that the cost of the MM attack is independent of the number of classes. The MM attack opens a new way for evaluating adversarial robustness and provides a feasible and reliable way to generate high-quality adversarial examples in AT."
                    },
                    {
                        "title": "Pareto Invariant Risk Minimization",
                        "abstract": "Despite the success of invariant risk minimization (IRM) in tackling the Out-of-Distribution generalization problem, IRM can compromise the optimality when applied in practice. The practical variants of IRM, e.g., IRMv1, have been shown to have signi\ufb01cant gaps with IRM and thus could fail to capture the invariance even in simple problems. Moreover, the optimization procedure in IRMv1 involves two intrinsically con\ufb02icting objectives, and often requires careful tuning for the objec-tive weights. To remedy the above issues, we reformulate IRM as a multi-objective optimization problem, and propose a new optimization scheme for IRM, called PA reto I nvariant R isk Minimization ( PAIR ). PAIR can adaptively adjust the optimization direction under the objective con\ufb02icts. Furthermore, we show PAIR can empower the practical IRM variants to overcome the barriers with the original IRM when provided with proper guidance. We conduct experiments with ColoredMNIST to con\ufb01rm our theory and the effectiveness of PAIR ."
                    },
                    {
                        "title": "An Adaptive Incremental Gradient Method With Support for Non-Euclidean Norms",
                        "abstract": "Stochastic variance reduced methods have shown strong performance in solving finite-sum problems. However, these methods usually require the users to manually tune the step-size, which is time-consuming or even infeasible for some large-scale optimization tasks. To overcome the problem, we propose and analyze several novel adaptive variants of the popular SAGA algorithm. Eventually, we design a variant of Barzilai-Borwein step-size which is tailored for the incremental gradient method to ensure memory efficiency and fast convergence. We establish its convergence guarantees under general settings that allow non-Euclidean norms in the definition of smoothness and the composite objectives, which cover a broad range of applications in machine learning. We improve the analysis of SAGA to support non-Euclidean norms, which fills the void of existing work. Numerical experiments on standard datasets demonstrate a competitive performance of the proposed algorithm compared with existing variance-reduced methods and their adaptive variants."
                    },
                    {
                        "title": "Navigation as the Attacker Wishes? Towards Building Byzantine-Robust Embodied Agents under Federated Learning",
                        "abstract": "Federated embodied agent learning [39] protects the data privacy of individual visual environments by keeping data locally at each client (the individual environment) during training. However, since the local data is inaccessible to the server under federated learning, attackers may easily poison the training data of the local client to build a backdoor in the agent without notice. Deploying such an agent raises the risk of potential harm to humans, as the attackers may easily navigate and control the agent as they wish via the backdoor. Towards Byzantine-robust federated embodied agent learning, in this paper, we study the attack and defense for the task of vision-and-language navigation (VLN), where the agent is required to follow natural language instructions to navigate indoor environments. First, we introduce a simple but effective attack strategy, Navigation as Wish (NAW), in which the malicious client manipulates local trajectory data to implant a backdoor into the global model. Results on two VLN datasets (R2R [2] and RxR [19]) show that NAW can easily navigate the deployed VLN agent regardless of the language instruction, without affecting its performance on normal test sets. Then, we propose a new Prompt-Based Aggregation (PBA) to defend against the NAW attack in federated VLN, which provides the server with a \u201cprompt\u201d of the vision-and-language alignment variance between the benign and malicious clients so that they can be distinguished during training. We validate the effectiveness of the PBA method on protecting the global model from the NAW attack, which outperforms other state-of-the-art defense methods by a large margin in the defense metrics on R2R and RxR."
                    },
                    {
                        "title": "FedVLN: Privacy-preserving Federated Vision-and-Language Navigation",
                        "abstract": "Data privacy is a central problem for embodied agents that can perceive the environment, communicate with humans, and act in the real world. While helping humans complete tasks, the agent may observe and process sensitive information of users, such as house environments, human activities, etc. In this work, we introduce privacy-preserving embodied agent learning for the task of Vision-and-Language Navigation (VLN), where an embodied agent navigates house environments by following natural language instructions. We view each house environment as a local client, which shares nothing other than local updates with the cloud server and other clients, and propose a novel federated vision-and-language navigation (FedVLN) framework to protect data privacy during both training and pre-exploration. Particularly, we propose a decentralized training strategy to limit the data of each client to its local model training and a federated pre-exploration method to do partial model aggregation to improve model generalizability to unseen environments. Extensive results on R2R and RxR datasets show that under our FedVLN framework, decentralized VLN models achieve comparable results with centralized training while protecting seen environment privacy, and federated pre-exploration significantly outperforms centralized pre-exploration while preserving unseen environment privacy."
                    },
                    {
                        "title": "Pareto Invariant Risk Minimization: Towards Mitigating the Optimization Dilemma in Out-of-Distribution Generalization",
                        "abstract": "Recently, there has been a growing surge of interest in enabling machine learning systems to generalize well to Out-of-Distribution (OOD) data. Most efforts are devoted to advancing optimization objectives that regularize models to capture the underlying invariance; however, there often are compromises in the optimization process of these OOD objectives: i) Many OOD objectives have to be relaxed as penalty terms of Empirical Risk Minimization (ERM) for the ease of optimization, while the relaxed forms can weaken the robustness of the original objective; ii) The penalty terms also require careful tuning of the penalty weights due to the intrinsic conflicts between ERM and OOD objectives. Consequently, these compromises could easily lead to suboptimal performance of either the ERM or OOD objective. To address these issues, we introduce a multi-objective optimization (MOO) perspective to understand the OOD optimization process, and propose a new optimization scheme called PAreto Invariant Risk Minimization (PAIR). PAIR improves the robustness of OOD objectives by cooperatively optimizing with other OOD objectives, thereby bridging the gaps caused by the relaxations. Then PAIR approaches a Pareto optimal solution that trades off the ERM and OOD objectives properly. Extensive experiments on challenging benchmarks, WILDS, show that PAIR alleviates the compromises and yields top OOD performances."
                    },
                    {
                        "title": "Efficient Private SCO for Heavy-Tailed Data via Clipping",
                        "abstract": "We consider stochastic convex optimization for heavy-tailed data with the guarantee of being di\ufb00erentially private (DP). Prior work on this problem is restricted to the gradient descent (GD) method, which is ine\ufb03cient for large-scale problems. In this paper, we resolve this issue and derive the \ufb01rst high-probability bounds for private stochastic method with clipping. For general convex problems, we derive excess population risks \u02dc O"
                    },
                    {
                        "title": "Local Reweighting for Adversarial Training",
                        "abstract": "Instances-reweighted adversarial training (IRAT) can significantly boost the robustness of trained models, where data being less/more vulnerable to the given attack are assigned smaller/larger weights during training. However, when tested on attacks different from the given attack simulated in training, the robustness may drop significantly (e.g., even worse than no reweighting). In this paper, we study this problem and propose our solution--locally reweighted adversarial training (LRAT). The rationale behind IRAT is that we do not need to pay much attention to an instance that is already safe under the attack. We argue that the safeness should be attack-dependent, so that for the same instance, its weight can change given different attacks based on the same model. Thus, if the attack simulated in training is mis-specified, the weights of IRAT are misleading. To this end, LRAT pairs each instance with its adversarial variants and performs local reweighting inside each pair, while performing no global reweighting--the rationale is to fit the instance itself if it is immune to the attack, but not to skip the pair, in order to passively defend different attacks in future. Experiments show that LRAT works better than both IRAT (i.e., global reweighting) and the standard AT (i.e., no reweighting) when trained with an attack and tested on different attacks."
                    },
                    {
                        "title": "On the Finite-Time Complexity and Practical Computation of Approximate Stationarity Concepts of Lipschitz Functions",
                        "abstract": "We report a practical finite-time algorithmic scheme to compute approximately stationary points for nonconvex nonsmooth Lipschitz functions. In particular, we are interested in two kinds of approximate stationarity notions for nonconvex nonsmooth problems, i.e., Goldstein approximate stationarity (GAS) and near-approximate stationarity (NAS). For GAS, our scheme removes the unrealistic subgradient selection oracle assumption in (Zhang et al., 2020, Assumption 1) and computes GAS with the same finite-time complexity. For NAS, Davis&Drusvyatskiy (2019) showed that $\\rho$-weakly convex functions admit finite-time computation, while Tian&So (2021) provided the matching impossibility results of dimension-free finite-time complexity for first-order methods. Complement to these developments, in this paper, we isolate a new class of functions that could be Clarke irregular (and thus not weakly convex anymore) and show that our new algorithmic scheme can compute NAS points for functions in that class within finite time. To demonstrate the wide applicability of our new theoretical framework, we show that $\\rho$-margin SVM, $1$-layer, and $2$-layer ReLU neural networks, all being Clarke irregular, satisfy our new conditions."
                    },
                    {
                        "title": "Practical Schemes for Finding Near-Stationary Points of Convex Finite-Sums",
                        "abstract": "In convex optimization, the problem of finding near-stationary points has not been adequately studied yet, unlike other optimality measures such as the function value. Even in the deterministic case, the optimal method (OGM-G, due to Kim and Fessler (2021)) has just been discovered recently. In this work, we conduct a systematic study of algorithmic techniques for finding near-stationary points of convex finite-sums. Our main contributions are several algorithmic discoveries: (1) we discover a memory-saving variant of OGM-G based on the performance estimation problem approach (Drori and Teboulle, 2014); (2) we design a new accelerated SVRG variant that can simultaneously achieve fast rates for minimizing both the gradient norm and function value; (3) we propose an adaptively regularized accelerated SVRG variant, which does not require the knowledge of some unknown initial constants and achieves near-optimal complexities. We put an emphasis on the simplicity and practicality of the new schemes, which could facilitate future work."
                    }
                ]
            },
            "333b2d6b-477c-4d7e-8a08-c5c432e51290": {
                "pk": "333b2d6b-477c-4d7e-8a08-c5c432e51290",
                "name": "Yongqiang Chen",
                "collaborators": [
                    "James Cheng",
                    "Bo Han",
                    "Yatao Bian",
                    "Han Yang",
                    "Kaili Ma",
                    "Binghui Xie",
                    "Kaiwen Zhou",
                    "Yonggang Zhang",
                    "Tongliang Liu",
                    "Barakeel Fanseu Kamhoua"
                ],
                "domain": [
                    "Machine Learning",
                    "Graph Neural Network",
                    "Causal Inference",
                    "Domain Generalization"
                ],
                "publications": [
                    {
                        "title": "UniMoT: Unified Molecule-Text Language Model with Discrete Token Representation",
                        "abstract": "The remarkable success of Large Language Models (LLMs) across diverse tasks has driven the research community to extend their capabilities to molecular applications. However, most molecular LLMs employ adapter-based architectures that do not treat molecule and text modalities equally and lack a supervision signal for the molecule modality. To address these issues, we introduce UniMoT, a Unified Molecule-Text LLM adopting a tokenizer-based architecture that expands the vocabulary of LLM with molecule tokens. Specifically, we introduce a Vector Quantization-driven tokenizer that incorporates a Q-Former to bridge the modality gap between molecule and text. This tokenizer transforms molecules into sequences of molecule tokens with causal dependency, encapsulating high-level molecular and textual information. Equipped with this tokenizer, UniMoT can unify molecule and text modalities under a shared token representation and an autoregressive training paradigm, enabling it to interpret molecules as a foreign language and generate them as text. Following a four-stage training scheme, UniMoT emerges as a multi-modal generalist capable of performing both molecule-to-text and text-to-molecule tasks. Extensive experiments demonstrate that UniMoT achieves state-of-the-art performance across a wide range of molecule comprehension and generation tasks."
                    },
                    {
                        "title": "Enhancing Evolving Domain Generalization through Dynamic Latent Representations",
                        "abstract": "Domain generalization is a critical challenge for machine learning systems. Prior domain generalization methods focus on extracting domain-invariant features across several stationary domains to enable generalization to new domains. However, in non-stationary tasks where new domains evolve in an underlying continuous structure, such as time, merely extracting the invariant features is insufficient for generalization to the evolving new domains. Nevertheless, it is non-trivial to learn both evolving and invariant features within a single model due to their conflicts. To bridge this gap, we build causal models to characterize the distribution shifts concerning the two patterns, and propose to learn both dynamic and invariant features via a new framework called Mutual Information-Based Sequential Autoencoders (MISTS). MISTS adopts information theoretic constraints onto sequential autoencoders to disentangle the dynamic and invariant features, and leverage an adaptive classifier to make predictions based on both evolving and invariant information. Our experimental results on both synthetic and real-world datasets demonstrate that MISTS succeeds in capturing both evolving and invariant information, and present promising results in evolving domain generalization tasks."
                    },
                    {
                        "title": "Discovery of the Hidden World with Large Language Models",
                        "abstract": "Revealing the underlying causal mechanisms in the real world is the key to the development of science. Despite the progress in the past decades, traditional causal discovery approaches (CDs) mainly rely on high-quality measured variables, usually given by human experts, to find causal relations. The lack of well-defined high-level variables in many real-world applications has already been a longstanding roadblock to a broader application of CDs. To this end, this paper presents Causal representatiOn AssistanT (COAT) that introduces large language models (LLMs) to bridge the gap. LLMs are trained on massive observations of the world and have demonstrated great capability in extracting key information from unstructured data. Therefore, it is natural to employ LLMs to assist with proposing useful high-level factors and crafting their measurements. Meanwhile, COAT also adopts CDs to find causal relations among the identified variables as well as to provide feedback to LLMs to iteratively refine the proposed factors. We show that LLMs and CDs are mutually beneficial and the constructed feedback provably also helps with the factor proposal. We construct and curate several synthetic and real-world benchmarks including analysis of human reviews and diagnosis of neuropathic and brain tumors, to comprehensively evaluate COAT. Extensive empirical results confirm the effectiveness and reliability of COAT with significant improvements."
                    },
                    {
                        "title": "How Interpretable Are Interpretable Graph Neural Networks?",
                        "abstract": "Interpretable graph neural networks (XGNNs ) are widely adopted in various scientific applications involving graph-structured data. Existing XGNNs predominantly adopt the attention-based mechanism to learn edge or node importance for extracting and making predictions with the interpretable subgraph. However, the representational properties and limitations of these methods remain inadequately explored. In this work, we present a theoretical framework that formulates interpretable subgraph learning with the multilinear extension of the subgraph distribution, coined as subgraph multilinear extension (SubMT). Extracting the desired interpretable subgraph requires an accurate approximation of SubMT, yet we find that the existing XGNNs can have a huge gap in fitting SubMT. Consequently, the SubMT approximation failure will lead to the degenerated interpretability of the extracted subgraphs. To mitigate the issue, we design a new XGNN architecture called Graph Multilinear neT (GMT), which is provably more powerful in approximating SubMT. We empirically validate our theoretical findings on a number of graph classification benchmarks. The results demonstrate that GMT outperforms the state-of-the-art up to 10% in terms of both interpretability and generalizability across 12 regular and geometric graph benchmarks."
                    },
                    {
                        "title": "HIGHT: Hierarchical Graph Tokenization for Graph-Language Alignment",
                        "abstract": "Recently there has been a surge of interest in extending the success of large language models (LLMs) to graph modality, such as social networks and molecules. As LLMs are predominantly trained with 1D text data, most existing approaches adopt a graph neural network to represent a graph as a series of node tokens and feed these tokens to LLMs for graph-language alignment. Despite achieving some successes, existing approaches have overlooked the hierarchical structures that are inherent in graph data. Especially, in molecular graphs, the high-order structural information contains rich semantics of molecular functional groups, which encode crucial biochemical functionalities of the molecules. We establish a simple benchmark showing that neglecting the hierarchical information in graph tokenization will lead to subpar graph-language alignment and severe hallucination in generated outputs. To address this problem, we propose a novel strategy called HIerarchical GrapH Tokenization (HIGHT). HIGHT employs a hierarchical graph tokenizer that extracts and encodes the hierarchy of node, motif, and graph levels of informative tokens to improve the graph perception of LLMs. HIGHT also adopts an augmented graph-language supervised fine-tuning dataset, enriched with the hierarchical graph information, to further enhance the graph-language alignment. Extensive experiments on 7 molecule-centric benchmarks confirm the effectiveness of HIGHT in reducing hallucination by 40%, as well as significant improvements in various molecule-language downstream tasks."
                    },
                    {
                        "title": "Do CLIPs Always Generalize Better than ImageNet Models?",
                        "abstract": "Large vision language models, such as CLIPs, have revolutionized modern machine learning. CLIPs have demonstrated great generalizability under distribution shifts, supported by an increasing body of literature. However, the evaluation datasets for CLIPs are variations primarily designed for ImageNet benchmarks, which may not fully reflect the extent to which CLIPs, e.g., pre-trained on LAION, robust to spurious correlations. To bridge the gap, we collect a real-world dataset called CounterAnimal that contains realistic spurious features found in animal photos. CounterAnimal consists of a) the common group: comprising animals on common backgrounds, and b) the counter group: including animals on unusual backgrounds. The performance drops from the common to counter groups quantify the reliance of models on spurious features (i.e., backgrounds) to predict the animals. We find that CLIPs trained on either LAION or the OpenAI data exhibit notable performance drops on the counter group. Surprisingly, we observe that single-modal models trained on ImageNet are more robust than CLIPs. We provide both theoretical and empirical explanations for why CLIPs still learn spurious features. Our findings suggest that distribution shifts remain an open problem for CLIPs, and one needs to be cautious about test setups when evaluating foundation models pre-trained on a significantly different scale and distribution."
                    },
                    {
                        "title": "Does Invariant Graph Learning via Environment Augmentation Learn Invariance?",
                        "abstract": "Invariant graph representation learning aims to learn the invariance among data from different environments for out-of-distribution generalization on graphs. As the graph environment partitions are usually expensive to obtain, augmenting the environment information has become the de facto approach. However, the usefulness of the augmented environment information has never been verified. In this work, we find that it is fundamentally impossible to learn invariant graph representations via environment augmentation without additional assumptions. Therefore, we develop a set of minimal assumptions, including variation sufficiency and variation consistency, for feasible invariant graph learning. We then propose a new framework Graph invAriant Learning Assistant (GALA). GALA incorporates an assistant model that needs to be sensitive to graph environment changes or distribution shifts. The correctness of the proxy predictions by the assistant model hence can differentiate the variations in spurious subgraphs. We show that extracting the maximally invariant subgraph to the proxy predictions provably identifies the underlying invariant subgraph for successful OOD generalization under the established minimal assumptions. Extensive experiments on datasets including DrugOOD with various graph distribution shifts confirm the effectiveness of GALA."
                    },
                    {
                        "title": "Positional Information Matters for Invariant In-Context Learning: A Case Study of Simple Function Classes",
                        "abstract": "In-context learning (ICL) refers to the ability of a model to condition on a few in-context demonstrations (input-output examples of the underlying task) to generate the answer for a new query input, without updating parameters. Despite the impressive ICL ability of LLMs, it has also been found that ICL in LLMs is sensitive to input demonstrations and limited to short context lengths. To understand the limitations and principles for successful ICL, we conduct an investigation with ICL linear regression of transformers. We characterize several Out-of-Distribution (OOD) cases for ICL inspired by realistic LLM ICL failures and compare transformers with DeepSet, a simple yet powerful architecture for ICL. Surprisingly, DeepSet outperforms transformers across a variety of distribution shifts, implying that preserving permutation invariance symmetry to input demonstrations is crucial for OOD ICL. The phenomenon specifies a fundamental requirement by ICL, which we termed as ICL invariance. Nevertheless, the positional encodings in LLMs will break ICL invariance. To this end, we further evaluate transformers with identical positional encodings and find preserving ICL invariance in transformers achieves state-of-the-art performance across various ICL distribution shifts"
                    },
                    {
                        "title": "Towards out-of-distribution generalizable predictions of chemical kinetics properties",
                        "abstract": "Machine Learning (ML) techniques have found applications in estimating chemical kinetic properties. With the accumulated drug molecules identified through\"AI4drug discovery\", the next imperative lies in AI-driven design for high-throughput chemical synthesis processes, with the estimation of properties of unseen reactions with unexplored molecules. To this end, the existing ML approaches for kinetics property prediction are required to be Out-Of-Distribution (OOD) generalizable. In this paper, we categorize the OOD kinetic property prediction into three levels (structure, condition, and mechanism), revealing unique aspects of such problems. Under this framework, we create comprehensive datasets to benchmark (1) the state-of-the-art ML approaches for reaction prediction in the OOD setting and (2) the state-of-the-art graph OOD methods in kinetics property prediction problems. Our results demonstrated the challenges and opportunities in OOD kinetics property prediction. Our datasets and benchmarks can further support research in this direction."
                    },
                    {
                        "title": "Understanding and Improving Feature Learning for Out-of-Distribution Generalization",
                        "abstract": "A common explanation for the failure of out-of-distribution (OOD) generalization is that the model trained with empirical risk minimization (ERM) learns spurious features instead of invariant features. However, several recent studies challenged this explanation and found that deep networks may have already learned sufficiently good features for OOD generalization. Despite the contradictions at first glance, we theoretically show that ERM essentially learns both spurious and invariant features, while ERM tends to learn spurious features faster if the spurious correlation is stronger. Moreover, when fed the ERM learned features to the OOD objectives, the invariant feature learning quality significantly affects the final OOD performance, as OOD objectives rarely learn new features. Therefore, ERM feature learning can be a bottleneck to OOD generalization. To alleviate the reliance, we propose Feature Augmented Training (FeAT), to enforce the model to learn richer features ready for OOD generalization. FeAT iteratively augments the model to learn new features while retaining the already learned features. In each round, the retention and augmentation operations are performed on different subsets of the training data that capture distinct features. Extensive experiments show that FeAT effectively learns richer features thus boosting the performance of various OOD objectives."
                    },
                    {
                        "title": "Understanding and Improving Graph Injection Attack by Promoting Unnoticeability",
                        "abstract": "Recently Graph Injection Attack (GIA) emerges as a practical attack scenario on Graph Neural Networks (GNNs), where the adversary can merely inject few malicious nodes instead of modifying existing nodes or edges, i.e., Graph Modification Attack (GMA). Although GIA has achieved promising results, little is known about why it is successful and whether there is any pitfall behind the success. To understand the power of GIA, we compare it with GMA and find that GIA can be provably more harmful than GMA due to its relatively high flexibility. However, the high flexibility will also lead to great damage to the homophily distribution of the original graph, i.e., similarity among neighbors. Consequently, the threats of GIA can be easily alleviated or even prevented by homophily-based defenses designed to recover the original homophily. To mitigate the issue, we introduce a novel constraint -- homophily unnoticeability that enforces GIA to preserve the homophily, and propose Harmonious Adversarial Objective (HAO) to instantiate it. Extensive experiments verify that GIA with HAO can break homophily-based defenses and outperform previous GIA attacks by a significant margin. We believe our methods can serve for a more reliable evaluation of the robustness of GNNs."
                    },
                    {
                        "title": "Exact Shape Correspondence via 2D graph convolution",
                        "abstract": "For exact 3D shape correspondence (matching or alignment), i.e., the task of matching each point on a shape to its exact corresponding point on the other shape (or to be more specific, matching at geodesic error 0), most existing methods do not perform well due to two main problems. First, on nearly-isometric shapes (i.e., low noise levels), most existing methods use the eigen-vectors (eigen-functions) of the Laplace Beltrami Operator (LBO) or other shape descriptors to update an initialized correspondence which is not exact, leading to an accumulation of update errors. Thus, though the final correspondence may generally be smooth, it is generally inexact. Second, on non-isometric shapes (noisy shapes), existing methods are generally not robust to noise as they usually assume near-isometry. In addition, existing methods that attempt to address the non-isometric shape problem (e.g., GRAMPA) are generally computationally expensive and do not generalise to nearly-isometric shapes. To address these two problems, we propose a 2D graph convolution-based framework called 2D-GEM. 2D-GEM is robust to noise on non-isometric shapes and with a few additional constraints, it also addresses the errors in the update on nearly-isometric shapes. We demonstrate the effectiveness of 2D-GEM by achieving a high accuracy of 90.5 % at geodesic error 0 on the non-isometric benchmark SHREC16, i.e., TOPKIDS (while being much faster than GRAMPA), and on nearly-isometric benchmarks by achieving a high accuracy of 92.5 % on TOSCA and 84.9 % on SCAPE at geodesic error 0."
                    },
                    {
                        "title": "Pareto Invariant Risk Minimization",
                        "abstract": "Despite the success of invariant risk minimization (IRM) in tackling the Out-of-Distribution generalization problem, IRM can compromise the optimality when applied in practice. The practical variants of IRM, e.g., IRMv1, have been shown to have signi\ufb01cant gaps with IRM and thus could fail to capture the invariance even in simple problems. Moreover, the optimization procedure in IRMv1 involves two intrinsically con\ufb02icting objectives, and often requires careful tuning for the objec-tive weights. To remedy the above issues, we reformulate IRM as a multi-objective optimization problem, and propose a new optimization scheme for IRM, called PA reto I nvariant R isk Minimization ( PAIR ). PAIR can adaptively adjust the optimization direction under the objective con\ufb02icts. Furthermore, we show PAIR can empower the practical IRM variants to overcome the barriers with the original IRM when provided with proper guidance. We conduct experiments with ColoredMNIST to con\ufb01rm our theory and the effectiveness of PAIR ."
                    },
                    {
                        "title": "Invariance Principle Meets Out-of-Distribution Generalization on Graphs",
                        "abstract": "Despite recent success in using the invariance principle for out-of-distribution (OOD) generalization on Euclidean data (e.g., images), studies on graph data are still limited. Different from images, the complex nature of graphs poses unique challenges to adopting the invariance principle. In particular, distribution shifts on graphs can appear in a variety of forms such as attributes and structures, making it dif\ufb01cult to identify the invariance. Moreover, domain or environment partitions, which are often required by OOD methods on Euclidean data, could be highly expensive to obtain for graphs. To bridge this gap, we pro-pose a new framework to capture the invariance of graphs for guaranteed OOD generalization under various distribution shifts. Speci\ufb01cally, we characterize potential distribution shifts on graphs with causal models, concluding that OOD generalization on graphs is achievable when models focus only on subgraphs containing the most information about the causes of labels. Accordingly, we propose an information-theoretic objective to extract the desired subgraphs that maximally preserve the invariant intra-class information. Learning with these subgraphs is immune to distribution shifts. Extensive experiments on both synthetic and real-world datasets, including a challenging setting in AI-aided drug discovery, validate the superior OOD generalization ability of our method."
                    },
                    {
                        "title": "Pareto Invariant Risk Minimization: Towards Mitigating the Optimization Dilemma in Out-of-Distribution Generalization",
                        "abstract": "Recently, there has been a growing surge of interest in enabling machine learning systems to generalize well to Out-of-Distribution (OOD) data. Most efforts are devoted to advancing optimization objectives that regularize models to capture the underlying invariance; however, there often are compromises in the optimization process of these OOD objectives: i) Many OOD objectives have to be relaxed as penalty terms of Empirical Risk Minimization (ERM) for the ease of optimization, while the relaxed forms can weaken the robustness of the original objective; ii) The penalty terms also require careful tuning of the penalty weights due to the intrinsic conflicts between ERM and OOD objectives. Consequently, these compromises could easily lead to suboptimal performance of either the ERM or OOD objective. To address these issues, we introduce a multi-objective optimization (MOO) perspective to understand the OOD optimization process, and propose a new optimization scheme called PAreto Invariant Risk Minimization (PAIR). PAIR improves the robustness of OOD objectives by cooperatively optimizing with other OOD objectives, thereby bridging the gaps caused by the relaxations. Then PAIR approaches a Pareto optimal solution that trades off the ERM and OOD objectives properly. Extensive experiments on challenging benchmarks, WILDS, show that PAIR alleviates the compromises and yields top OOD performances."
                    },
                    {
                        "title": "Learning Causally Invariant Representations for Out-of-Distribution Generalization on Graphs",
                        "abstract": "Despite recent success in using the invariance principle for out-of-distribution (OOD) generalization on Euclidean data (e.g., images), studies on graph data are still limited. Different from images, the complex nature of graphs poses unique challenges to adopting the invariance principle. In particular, distribution shifts on graphs can appear in a variety of forms such as attributes and structures, making it difficult to identify the invariance. Moreover, domain or environment partitions, which are often required by OOD methods on Euclidean data, could be highly expensive to obtain for graphs. To bridge this gap, we propose a new framework, called Causality Inspired Invariant Graph LeArning (CIGA), to capture the invariance of graphs for guaranteed OOD generalization under various distribution shifts. Specifically, we characterize potential distribution shifts on graphs with causal models, concluding that OOD generalization on graphs is achievable when models focus only on subgraphs containing the most information about the causes of labels. Accordingly, we propose an information-theoretic objective to extract the desired subgraphs that maximally preserve the invariant intra-class information. Learning with these subgraphs is immune to distribution shifts. Extensive experiments on 16 synthetic or real-world datasets, including a challenging setting -- DrugOOD, from AI-aided drug discovery, validate the superior OOD performance of CIGA."
                    },
                    {
                        "title": "Calibrating and Improving Graph Contrastive Learning",
                        "abstract": "Graph contrastive learning algorithms have demonstrated remarkable success in various applications such as node classification, link prediction, and graph clustering. However, in unsupervised graph contrastive learning, some contrastive pairs may contradict the truths in downstream tasks and thus the decrease of losses on these pairs undesirably harms the performance in the downstream tasks. To assess the discrepancy between the prediction and the ground-truth in the downstream tasks for these contrastive pairs, we adapt the expected calibration error (ECE) to graph contrastive learning. The analysis of ECE motivates us to propose a novel regularization method, Contrast-Reg, to ensure that decreasing the contrastive loss leads to better performance in the downstream tasks. As a plug-in regularizer, Contrast-Reg effectively improves the performance of existing graph contrastive learning algorithms. We provide both theoretical and empirical results to demonstrate the effectiveness of Contrast-Reg in enhancing the generalizability of the Graph Neural Network(GNN) model and improving the performance of graph contrastive algorithms with different similarity definitions and encoder backbones across various downstream tasks."
                    },
                    {
                        "title": "Improving Graph Representation Learning by Contrastive Regularization",
                        "abstract": "Graph representation learning is an important task with applications in various areas such as online social networks, e-commerce networks, WWW and semantic webs. For unsupervised graph representation learning, many algorithms such as Node2Vec and Graph-SAGE make use of \u201cnegative sampling\u201d and/or noise contrastive estimation loss. This bears similar ideas to contrastive learning, which \u201ccontrasts\u201d the node representation similarities of semantically similar (positive) pairs against those of negative pairs. However, despite the success of contrastive learning, we found that directly applying this technique to graph representation learning models (e.g., graph convolutional networks) does not always work. We theoretically analyze the generalization performance and pro-pose a light-weight regularization term that avoids the high scales of node representations\u2019 norms and the high variance among them to improve the generalization performance. Our experimental re-sults further validate that this regularization term significantly improves the representation quality across different node similarity definitions and outperforms the state-of-the-art methods."
                    }
                ]
            },
            "0775ed47-bee8-4200-9f58-affd5e7ba7ed": {
                "pk": "0775ed47-bee8-4200-9f58-affd5e7ba7ed",
                "name": "Peilin Zhao",
                "collaborators": [
                    "Yatao Bian",
                    "Haoyu Wang",
                    "Bingzhe Wu",
                    "Xueqian Wang",
                    "Yongzhe Chang",
                    "Changqing Zhang",
                    "Guozheng Ma",
                    "Li Shen",
                    "Tingyang Xu",
                    "Liang Chen"
                ],
                "domain": [
                    "Large Language Models",
                    "Machine Learning",
                    "Drug Discovery",
                    "Content Moderation"
                ],
                "publications": [
                    {
                        "title": "Probing the Safety Response Boundary of Large Language Models via Unsafe Decoding Path Generation",
                        "abstract": "Large Language Models (LLMs) are implicit troublemakers. While they provide valuable insights and assist in problem-solving, they can also potentially serve as a resource for malicious activities. Implementing safety alignment could mitigate the risk of LLMs generating harmful responses. We argue that: even when an LLM appears to successfully block harmful queries, there may still be hidden vulnerabilities that could act as ticking time bombs. To identify these underlying weaknesses, we propose to use a cost value model as both a detector and an attacker. Trained on external or self-generated harmful datasets, the cost value model could successfully influence the original safe LLM to output toxic content in decoding process. For instance, LLaMA-2-chat 7B outputs 39.18% concrete toxic content, along with only 22.16% refusals without any harmful suffixes. These potential weaknesses can then be exploited via prompt optimization such as soft prompts on images. We name this decoding strategy: Jailbreak Value Decoding (JVD), emphasizing that seemingly secure LLMs may not be as safe as we initially believe. They could be used to gather harmful data or launch covert attacks."
                    },
                    {
                        "title": "COME: Test-time adaption by Conservatively Minimizing Entropy",
                        "abstract": "Machine learning models must continuously self-adjust themselves for novel data distribution in the open world. As the predominant principle, entropy minimization (EM) has been proven to be a simple yet effective cornerstone in existing test-time adaption (TTA) methods. While unfortunately its fatal limitation (i.e., overconfidence) tends to result in model collapse. For this issue, we propose to Conservatively Minimize the Entropy (COME), which is a simple drop-in replacement of traditional EM to elegantly address the limitation. In essence, COME explicitly models the uncertainty by characterizing a Dirichlet prior distribution over model predictions during TTA. By doing so, COME naturally regularizes the model to favor conservative confidence on unreliable samples. Theoretically, we provide a preliminary analysis to reveal the ability of COME in enhancing the optimization stability by introducing a data-adaptive lower bound on the entropy. Empirically, our method achieves state-of-the-art performance on commonly used benchmarks, showing significant improvements in terms of classification accuracy and uncertainty estimation under various settings including standard, life-long and open-world TTA, i.e., up to $34.5\\%$ improvement on accuracy and $15.1\\%$ on false positive rate."
                    },
                    {
                        "title": "Step-On-Feet Tuning: Scaling Self-Alignment of LLMs via Bootstrapping",
                        "abstract": "Self-alignment is an effective way to reduce the cost of human annotation while ensuring promising model capability. However, most current methods complete the data collection and training steps in a single round, which may overlook the continuously improving ability of self-aligned models. This gives rise to a key query: What if we do multi-time bootstrapping self-alignment? Does this strategy enhance model performance or lead to rapid degradation? In this paper, our pioneering exploration delves into the impact of bootstrapping self-alignment on large language models. Our findings reveal that bootstrapping self-alignment markedly surpasses the single-round approach, by guaranteeing data diversity from in-context learning. To further exploit the capabilities of bootstrapping, we investigate and adjust the training order of data, which yields improved performance of the model. Drawing on these findings, we propose Step-On-Feet Tuning (SOFT) which leverages model's continuously enhanced few-shot ability to boost zero or one-shot performance. Based on easy-to-hard training recipe, we propose SOFT+ which further boost self-alignment's performance. Our experiments demonstrate the efficiency of SOFT (SOFT+) across various classification and generation tasks, highlighting the potential of bootstrapping self-alignment on continually enhancing model alignment performance."
                    },
                    {
                        "title": "WatME: Towards Lossless Watermarking Through Lexical Redundancy",
                        "abstract": "Text watermarking has emerged as a pivotal technique for identifying machine-generated text. However, existing methods often rely on arbitrary vocabulary partitioning during decoding to embed watermarks, which compromises the availability of suitable tokens and significantly degrades the quality of responses. This study assesses the impact of watermarking on different capabilities of large language models (LLMs) from a cognitive science lens. Our finding highlights a significant disparity; knowledge recall and logical reasoning are more adversely affected than language generation. These results suggest a more profound effect of watermarking on LLMs than previously understood. To address these challenges, we introduce Watermarking with Mutual Exclusion (WatME), a novel approach leveraging linguistic prior knowledge of inherent lexical redundancy in LLM vocabularies to seamlessly integrate watermarks. Specifically, WatME dynamically optimizes token usage during the decoding process by applying a mutually exclusive rule to the identified lexical redundancies. This strategy effectively prevents the unavailability of appropriate tokens and preserves the expressive power of LLMs. We provide both theoretical analysis and empirical evidence showing that WatME effectively preserves the diverse capabilities of LLMs while ensuring watermark detectability."
                    },
                    {
                        "title": "Are Large Language Models Really Robust to Word-Level Perturbations?",
                        "abstract": "The swift advancement in the scales and capabilities of Large Language Models (LLMs) positions them as promising tools for a variety of downstream tasks. In addition to the pursuit of better performance and the avoidance of violent feedback on a certain prompt, to ensure the responsibility of the LLM, much attention is drawn to the robustness of LLMs. However, existing evaluation methods mostly rely on traditional question answering datasets with predefined supervised labels, which do not align with the superior generation capabilities of contemporary LLMs. To address this issue, we propose a novel rational evaluation approach that leverages pre-trained reward models as diagnostic tools to evaluate the longer conversation generated from more challenging open questions by LLMs, which we refer to as the Reward Model for Reasonable Robustness Evaluation (TREvaL). Longer conversations manifest the comprehensive grasp of language models in terms of their proficiency in understanding questions, a capability not entirely encompassed by individual words or letters, which may exhibit oversimplification and inherent biases. Our extensive empirical experiments demonstrate that TREvaL provides an innovative method for evaluating the robustness of an LLM. Furthermore, our results demonstrate that LLMs frequently exhibit vulnerability to word-level perturbations that are commonplace in daily language usage. Notably, we are surprised to discover that robustness tends to decrease as fine-tuning (SFT and RLHF) is conducted. The code of TREval is available in https://github.com/Harry-mic/TREvaL."
                    },
                    {
                        "title": "ETDock: A Novel Equivariant Transformer for Protein-Ligand Docking",
                        "abstract": "Predicting the docking between proteins and ligands is a crucial and challenging task for drug discovery. However, traditional docking methods mainly rely on scoring functions, and deep learning-based docking approaches usually neglect the 3D spatial information of proteins and ligands, as well as the graph-level features of ligands, which limits their performance. To address these limitations, we propose an equivariant transformer neural network for protein-ligand docking pose prediction. Our approach involves the fusion of ligand graph-level features by feature processing, followed by the learning of ligand and protein representations using our proposed TAMformer module. Additionally, we employ an iterative optimization approach based on the predicted distance matrix to generate refined ligand poses. The experimental results on real datasets show that our model can achieve state-of-the-art performance."
                    },
                    {
                        "title": "X-Mark: Towards Lossless Watermarking Through Lexical Redundancy",
                        "abstract": "Text watermarking has emerged as an important technique for detecting machine-generated text. However, existing methods can severely degrade text quality due to arbitrary vocabulary partitioning, which disrupts the language model\u2019s expressiveness and impedes textual coherence. To mitigate this, we introduce X-Mark, a novel approach that capitalizes on text redundancy within the lexical space. Specifically, X-Mark incorporates a mutually exclusive rule for synonyms during the language model decoding process, thereby integrating prior knowledge into vocabulary partitioning and preserving the capabilities of language generation. We present theoretical analyses and empirical evidence demonstrating that X-Mark substantially enhances text generation fluency while maintaining watermark detectability. Furthermore, we investigate watermark-ing\u2019s impact on the emergent abilities of large language models, including zero-shot and few-shot knowledge recall, logical reasoning, and instruction following. Our comprehensive experiments confirm that X-Mark consistently outperforms existing methods in retaining these crucial capabilities of LLMs."
                    },
                    {
                        "title": "Adapting Large Language Models for Content Moderation: Pitfalls in Data Engineering and Supervised Fine-tuning",
                        "abstract": "Nowadays, billions of people engage in communication and express their opinions on the internet daily. Unfortunately, not all of these expressions are friendly or compliant, making content moderation an indispensable task. A common approach is to use a discriminative model to classify the content, but this method often requires strict data engineering, otherwise it will face unacceptable overfitting. With the successful development of Large Language Models (LLMs) in recent years, LLM-based methods have become a feasible solution for handling tasks in various domains. Thanks to the knowledge of the foundation models, we can develop more robust privately deployed models with limited data via fine-tuning these foundation models. Moreover, as a generative model, it can provide detailed analysis of the review process, enhancing interpretability. In this paper, we introduce how to fine-tune a LLM model that can be privately deployed for content moderation. Specifically, we discuss the differences between discriminative and generative models using content moderation as an example. Additionally, we reveal that incorporating reasoning processes during the fine-tuning of LLMs can effectively alleviate overfitting, even if the model is not allowed to directly output reasoning processes during deployment. We present a complete process, from data collection and construction to model training and overfitting elimination, for fine-tuning LLMs in vertical domain deployments. We report the entire research process and the key findings in this paper, hoping to provide valuable experience for researchers who are fine-tuning privately deployed models in their domain-specific research."
                    },
                    {
                        "title": "DrugOOD: Out-of-Distribution (OOD) Dataset Curator and Benchmark for AI-aided Drug Discovery - A Focus on Affinity Prediction Problems with Noise Annotations",
                        "abstract": "AI-aided drug discovery (AIDD) is gaining increasing popularity due to its promise of making the search for new pharmaceuticals quicker, cheaper and more efficient. In spite of its extensive use in many fields, such as ADMET prediction, virtual screening, protein folding and generative chemistry, little has been explored in terms of the out-of-distribution (OOD) learning problem with \\emph{noise}, which is inevitable in real world AIDD applications. In this work, we present DrugOOD, a systematic OOD dataset curator and benchmark for AI-aided drug discovery, which comes with an open-source Python package that fully automates the data curation and OOD benchmarking processes. We focus on one of the most crucial problems in AIDD: drug target binding affinity prediction, which involves both macromolecule (protein target) and small-molecule (drug compound). In contrast to only providing fixed datasets, DrugOOD offers automated dataset curator with user-friendly customization scripts, rich domain annotations aligned with biochemistry knowledge, realistic noise annotations and rigorous benchmarking of state-of-the-art OOD algorithms. Since the molecular data is often modeled as irregular graphs using graph neural network (GNN) backbones, DrugOOD also serves as a valuable testbed for \\emph{graph OOD learning} problems. Extensive empirical studies have shown a significant performance gap between in-distribution and out-of-distribution experiments, which highlights the need to develop better schemes that can allow for OOD generalization under noise for AIDD."
                    }
                ]
            },
            "a5ec07d6-116c-417e-b220-6fe64f3ae35b": {
                "pk": "a5ec07d6-116c-417e-b220-6fe64f3ae35b",
                "name": "Bo Han",
                "collaborators": [
                    "Yongqiang Chen",
                    "James Cheng",
                    "Binghui Xie",
                    "Kaiwen Zhou",
                    "Yatao Bian",
                    "Jiaqi Wang",
                    "Wei Meng",
                    "Chenxi Liu",
                    "Tongliang Liu",
                    "Mingming Gong"
                ],
                "domain": [
                    "Domain Generalization",
                    "Causal Inference",
                    "Graph Neural Network",
                    "Out-of-Distribution Generalization"
                ],
                "publications": [
                    {
                        "title": "Enhancing Evolving Domain Generalization through Dynamic Latent Representations",
                        "abstract": "Domain generalization is a critical challenge for machine learning systems. Prior domain generalization methods focus on extracting domain-invariant features across several stationary domains to enable generalization to new domains. However, in non-stationary tasks where new domains evolve in an underlying continuous structure, such as time, merely extracting the invariant features is insufficient for generalization to the evolving new domains. Nevertheless, it is non-trivial to learn both evolving and invariant features within a single model due to their conflicts. To bridge this gap, we build causal models to characterize the distribution shifts concerning the two patterns, and propose to learn both dynamic and invariant features via a new framework called Mutual Information-Based Sequential Autoencoders (MISTS). MISTS adopts information theoretic constraints onto sequential autoencoders to disentangle the dynamic and invariant features, and leverage an adaptive classifier to make predictions based on both evolving and invariant information. Our experimental results on both synthetic and real-world datasets demonstrate that MISTS succeeds in capturing both evolving and invariant information, and present promising results in evolving domain generalization tasks."
                    },
                    {
                        "title": "Discovery of the Hidden World with Large Language Models",
                        "abstract": "Revealing the underlying causal mechanisms in the real world is the key to the development of science. Despite the progress in the past decades, traditional causal discovery approaches (CDs) mainly rely on high-quality measured variables, usually given by human experts, to find causal relations. The lack of well-defined high-level variables in many real-world applications has already been a longstanding roadblock to a broader application of CDs. To this end, this paper presents Causal representatiOn AssistanT (COAT) that introduces large language models (LLMs) to bridge the gap. LLMs are trained on massive observations of the world and have demonstrated great capability in extracting key information from unstructured data. Therefore, it is natural to employ LLMs to assist with proposing useful high-level factors and crafting their measurements. Meanwhile, COAT also adopts CDs to find causal relations among the identified variables as well as to provide feedback to LLMs to iteratively refine the proposed factors. We show that LLMs and CDs are mutually beneficial and the constructed feedback provably also helps with the factor proposal. We construct and curate several synthetic and real-world benchmarks including analysis of human reviews and diagnosis of neuropathic and brain tumors, to comprehensively evaluate COAT. Extensive empirical results confirm the effectiveness and reliability of COAT with significant improvements."
                    },
                    {
                        "title": "How Interpretable Are Interpretable Graph Neural Networks?",
                        "abstract": "Interpretable graph neural networks (XGNNs ) are widely adopted in various scientific applications involving graph-structured data. Existing XGNNs predominantly adopt the attention-based mechanism to learn edge or node importance for extracting and making predictions with the interpretable subgraph. However, the representational properties and limitations of these methods remain inadequately explored. In this work, we present a theoretical framework that formulates interpretable subgraph learning with the multilinear extension of the subgraph distribution, coined as subgraph multilinear extension (SubMT). Extracting the desired interpretable subgraph requires an accurate approximation of SubMT, yet we find that the existing XGNNs can have a huge gap in fitting SubMT. Consequently, the SubMT approximation failure will lead to the degenerated interpretability of the extracted subgraphs. To mitigate the issue, we design a new XGNN architecture called Graph Multilinear neT (GMT), which is provably more powerful in approximating SubMT. We empirically validate our theoretical findings on a number of graph classification benchmarks. The results demonstrate that GMT outperforms the state-of-the-art up to 10% in terms of both interpretability and generalizability across 12 regular and geometric graph benchmarks."
                    },
                    {
                        "title": "Do CLIPs Always Generalize Better than ImageNet Models?",
                        "abstract": "Large vision language models, such as CLIPs, have revolutionized modern machine learning. CLIPs have demonstrated great generalizability under distribution shifts, supported by an increasing body of literature. However, the evaluation datasets for CLIPs are variations primarily designed for ImageNet benchmarks, which may not fully reflect the extent to which CLIPs, e.g., pre-trained on LAION, robust to spurious correlations. To bridge the gap, we collect a real-world dataset called CounterAnimal that contains realistic spurious features found in animal photos. CounterAnimal consists of a) the common group: comprising animals on common backgrounds, and b) the counter group: including animals on unusual backgrounds. The performance drops from the common to counter groups quantify the reliance of models on spurious features (i.e., backgrounds) to predict the animals. We find that CLIPs trained on either LAION or the OpenAI data exhibit notable performance drops on the counter group. Surprisingly, we observe that single-modal models trained on ImageNet are more robust than CLIPs. We provide both theoretical and empirical explanations for why CLIPs still learn spurious features. Our findings suggest that distribution shifts remain an open problem for CLIPs, and one needs to be cautious about test setups when evaluating foundation models pre-trained on a significantly different scale and distribution."
                    },
                    {
                        "title": "Does Invariant Graph Learning via Environment Augmentation Learn Invariance?",
                        "abstract": "Invariant graph representation learning aims to learn the invariance among data from different environments for out-of-distribution generalization on graphs. As the graph environment partitions are usually expensive to obtain, augmenting the environment information has become the de facto approach. However, the usefulness of the augmented environment information has never been verified. In this work, we find that it is fundamentally impossible to learn invariant graph representations via environment augmentation without additional assumptions. Therefore, we develop a set of minimal assumptions, including variation sufficiency and variation consistency, for feasible invariant graph learning. We then propose a new framework Graph invAriant Learning Assistant (GALA). GALA incorporates an assistant model that needs to be sensitive to graph environment changes or distribution shifts. The correctness of the proxy predictions by the assistant model hence can differentiate the variations in spurious subgraphs. We show that extracting the maximally invariant subgraph to the proxy predictions provably identifies the underlying invariant subgraph for successful OOD generalization under the established minimal assumptions. Extensive experiments on datasets including DrugOOD with various graph distribution shifts confirm the effectiveness of GALA."
                    },
                    {
                        "title": "Positional Information Matters for Invariant In-Context Learning: A Case Study of Simple Function Classes",
                        "abstract": "In-context learning (ICL) refers to the ability of a model to condition on a few in-context demonstrations (input-output examples of the underlying task) to generate the answer for a new query input, without updating parameters. Despite the impressive ICL ability of LLMs, it has also been found that ICL in LLMs is sensitive to input demonstrations and limited to short context lengths. To understand the limitations and principles for successful ICL, we conduct an investigation with ICL linear regression of transformers. We characterize several Out-of-Distribution (OOD) cases for ICL inspired by realistic LLM ICL failures and compare transformers with DeepSet, a simple yet powerful architecture for ICL. Surprisingly, DeepSet outperforms transformers across a variety of distribution shifts, implying that preserving permutation invariance symmetry to input demonstrations is crucial for OOD ICL. The phenomenon specifies a fundamental requirement by ICL, which we termed as ICL invariance. Nevertheless, the positional encodings in LLMs will break ICL invariance. To this end, we further evaluate transformers with identical positional encodings and find preserving ICL invariance in transformers achieves state-of-the-art performance across various ICL distribution shifts"
                    },
                    {
                        "title": "Towards out-of-distribution generalizable predictions of chemical kinetics properties",
                        "abstract": "Machine Learning (ML) techniques have found applications in estimating chemical kinetic properties. With the accumulated drug molecules identified through\"AI4drug discovery\", the next imperative lies in AI-driven design for high-throughput chemical synthesis processes, with the estimation of properties of unseen reactions with unexplored molecules. To this end, the existing ML approaches for kinetics property prediction are required to be Out-Of-Distribution (OOD) generalizable. In this paper, we categorize the OOD kinetic property prediction into three levels (structure, condition, and mechanism), revealing unique aspects of such problems. Under this framework, we create comprehensive datasets to benchmark (1) the state-of-the-art ML approaches for reaction prediction in the OOD setting and (2) the state-of-the-art graph OOD methods in kinetics property prediction problems. Our results demonstrated the challenges and opportunities in OOD kinetics property prediction. Our datasets and benchmarks can further support research in this direction."
                    }
                ]
            },
            "059dcafc-ec2b-4756-883e-dd3c6eb1b7fa": {
                "pk": "059dcafc-ec2b-4756-883e-dd3c6eb1b7fa",
                "name": "Wei Meng",
                "collaborators": [
                    "Binghui Xie",
                    "Kaiwen Zhou",
                    "James Cheng",
                    "Yongqiang Chen",
                    "Jiaqi Wang",
                    "Bo Han",
                    "Chen Jin",
                    "Xuan Dong",
                    "Jiangtao Wen",
                    "Weixin Li"
                ],
                "domain": [
                    "Domain Generalization",
                    "Machine Learning",
                    "Video Enhancement",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Enhancing Evolving Domain Generalization through Dynamic Latent Representations",
                        "abstract": "Domain generalization is a critical challenge for machine learning systems. Prior domain generalization methods focus on extracting domain-invariant features across several stationary domains to enable generalization to new domains. However, in non-stationary tasks where new domains evolve in an underlying continuous structure, such as time, merely extracting the invariant features is insufficient for generalization to the evolving new domains. Nevertheless, it is non-trivial to learn both evolving and invariant features within a single model due to their conflicts. To bridge this gap, we build causal models to characterize the distribution shifts concerning the two patterns, and propose to learn both dynamic and invariant features via a new framework called Mutual Information-Based Sequential Autoencoders (MISTS). MISTS adopts information theoretic constraints onto sequential autoencoders to disentangle the dynamic and invariant features, and leverage an adaptive classifier to make predictions based on both evolving and invariant information. Our experimental results on both synthetic and real-world datasets demonstrate that MISTS succeeds in capturing both evolving and invariant information, and present promising results in evolving domain generalization tasks."
                    },
                    {
                        "title": "An Adaptive Incremental Gradient Method With Support for Non-Euclidean Norms",
                        "abstract": "Stochastic variance reduced methods have shown strong performance in solving finite-sum problems. However, these methods usually require the users to manually tune the step-size, which is time-consuming or even infeasible for some large-scale optimization tasks. To overcome the problem, we propose and analyze several novel adaptive variants of the popular SAGA algorithm. Eventually, we design a variant of Barzilai-Borwein step-size which is tailored for the incremental gradient method to ensure memory efficiency and fast convergence. We establish its convergence guarantees under general settings that allow non-Euclidean norms in the definition of smoothness and the composite objectives, which cover a broad range of applications in machine learning. We improve the analysis of SAGA to support non-Euclidean norms, which fills the void of existing work. Numerical experiments on standard datasets demonstrate a competitive performance of the proposed algorithm compared with existing variance-reduced methods and their adaptive variants."
                    },
                    {
                        "title": "An Efficient and Integrated Algorithm for Video Enhancement in Challenging Lighting Conditions",
                        "abstract": "We describe a novel integrated algorithm for real-time enhancement of video acquired under challenging lighting conditions. Such conditions include low lighting, haze, and high dynamic range situations. The algorithm automatically detects the dominate source of impairment, then depending on whether it is low lighting, haze or others, a corresponding pre-processing is applied to the input video, followed by the core enhancement algorithm. Temporal and spatial redundancies in the video input are utilized to facilitate real-time processing and to improve temporal and spatial consistency of the output. The proposed algorithm can be used as an independent module, or be integrated in either a video encoder or a video decoder for further optimizations."
                    }
                ]
            },
            "4d924c31-63f5-4b3d-ad77-ac9421a10945": {
                "pk": "4d924c31-63f5-4b3d-ad77-ac9421a10945",
                "name": "James Cheng",
                "collaborators": [
                    "Yongqiang Chen",
                    "Bo Han",
                    "Yatao Bian",
                    "Binghui Xie",
                    "Kaiwen Zhou",
                    "Jiaqi Wang",
                    "Wei Meng",
                    "Chenxi Liu",
                    "Tongliang Liu",
                    "Mingming Gong"
                ],
                "domain": [
                    "Domain Generalization",
                    "Causal Inference",
                    "Graph Neural Network",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Enhancing Evolving Domain Generalization through Dynamic Latent Representations",
                        "abstract": "Domain generalization is a critical challenge for machine learning systems. Prior domain generalization methods focus on extracting domain-invariant features across several stationary domains to enable generalization to new domains. However, in non-stationary tasks where new domains evolve in an underlying continuous structure, such as time, merely extracting the invariant features is insufficient for generalization to the evolving new domains. Nevertheless, it is non-trivial to learn both evolving and invariant features within a single model due to their conflicts. To bridge this gap, we build causal models to characterize the distribution shifts concerning the two patterns, and propose to learn both dynamic and invariant features via a new framework called Mutual Information-Based Sequential Autoencoders (MISTS). MISTS adopts information theoretic constraints onto sequential autoencoders to disentangle the dynamic and invariant features, and leverage an adaptive classifier to make predictions based on both evolving and invariant information. Our experimental results on both synthetic and real-world datasets demonstrate that MISTS succeeds in capturing both evolving and invariant information, and present promising results in evolving domain generalization tasks."
                    },
                    {
                        "title": "Discovery of the Hidden World with Large Language Models",
                        "abstract": "Revealing the underlying causal mechanisms in the real world is the key to the development of science. Despite the progress in the past decades, traditional causal discovery approaches (CDs) mainly rely on high-quality measured variables, usually given by human experts, to find causal relations. The lack of well-defined high-level variables in many real-world applications has already been a longstanding roadblock to a broader application of CDs. To this end, this paper presents Causal representatiOn AssistanT (COAT) that introduces large language models (LLMs) to bridge the gap. LLMs are trained on massive observations of the world and have demonstrated great capability in extracting key information from unstructured data. Therefore, it is natural to employ LLMs to assist with proposing useful high-level factors and crafting their measurements. Meanwhile, COAT also adopts CDs to find causal relations among the identified variables as well as to provide feedback to LLMs to iteratively refine the proposed factors. We show that LLMs and CDs are mutually beneficial and the constructed feedback provably also helps with the factor proposal. We construct and curate several synthetic and real-world benchmarks including analysis of human reviews and diagnosis of neuropathic and brain tumors, to comprehensively evaluate COAT. Extensive empirical results confirm the effectiveness and reliability of COAT with significant improvements."
                    },
                    {
                        "title": "How Interpretable Are Interpretable Graph Neural Networks?",
                        "abstract": "Interpretable graph neural networks (XGNNs ) are widely adopted in various scientific applications involving graph-structured data. Existing XGNNs predominantly adopt the attention-based mechanism to learn edge or node importance for extracting and making predictions with the interpretable subgraph. However, the representational properties and limitations of these methods remain inadequately explored. In this work, we present a theoretical framework that formulates interpretable subgraph learning with the multilinear extension of the subgraph distribution, coined as subgraph multilinear extension (SubMT). Extracting the desired interpretable subgraph requires an accurate approximation of SubMT, yet we find that the existing XGNNs can have a huge gap in fitting SubMT. Consequently, the SubMT approximation failure will lead to the degenerated interpretability of the extracted subgraphs. To mitigate the issue, we design a new XGNN architecture called Graph Multilinear neT (GMT), which is provably more powerful in approximating SubMT. We empirically validate our theoretical findings on a number of graph classification benchmarks. The results demonstrate that GMT outperforms the state-of-the-art up to 10% in terms of both interpretability and generalizability across 12 regular and geometric graph benchmarks."
                    },
                    {
                        "title": "HIGHT: Hierarchical Graph Tokenization for Graph-Language Alignment",
                        "abstract": "Recently there has been a surge of interest in extending the success of large language models (LLMs) to graph modality, such as social networks and molecules. As LLMs are predominantly trained with 1D text data, most existing approaches adopt a graph neural network to represent a graph as a series of node tokens and feed these tokens to LLMs for graph-language alignment. Despite achieving some successes, existing approaches have overlooked the hierarchical structures that are inherent in graph data. Especially, in molecular graphs, the high-order structural information contains rich semantics of molecular functional groups, which encode crucial biochemical functionalities of the molecules. We establish a simple benchmark showing that neglecting the hierarchical information in graph tokenization will lead to subpar graph-language alignment and severe hallucination in generated outputs. To address this problem, we propose a novel strategy called HIerarchical GrapH Tokenization (HIGHT). HIGHT employs a hierarchical graph tokenizer that extracts and encodes the hierarchy of node, motif, and graph levels of informative tokens to improve the graph perception of LLMs. HIGHT also adopts an augmented graph-language supervised fine-tuning dataset, enriched with the hierarchical graph information, to further enhance the graph-language alignment. Extensive experiments on 7 molecule-centric benchmarks confirm the effectiveness of HIGHT in reducing hallucination by 40%, as well as significant improvements in various molecule-language downstream tasks."
                    },
                    {
                        "title": "Does Invariant Graph Learning via Environment Augmentation Learn Invariance?",
                        "abstract": "Invariant graph representation learning aims to learn the invariance among data from different environments for out-of-distribution generalization on graphs. As the graph environment partitions are usually expensive to obtain, augmenting the environment information has become the de facto approach. However, the usefulness of the augmented environment information has never been verified. In this work, we find that it is fundamentally impossible to learn invariant graph representations via environment augmentation without additional assumptions. Therefore, we develop a set of minimal assumptions, including variation sufficiency and variation consistency, for feasible invariant graph learning. We then propose a new framework Graph invAriant Learning Assistant (GALA). GALA incorporates an assistant model that needs to be sensitive to graph environment changes or distribution shifts. The correctness of the proxy predictions by the assistant model hence can differentiate the variations in spurious subgraphs. We show that extracting the maximally invariant subgraph to the proxy predictions provably identifies the underlying invariant subgraph for successful OOD generalization under the established minimal assumptions. Extensive experiments on datasets including DrugOOD with various graph distribution shifts confirm the effectiveness of GALA."
                    },
                    {
                        "title": "Positional Information Matters for Invariant In-Context Learning: A Case Study of Simple Function Classes",
                        "abstract": "In-context learning (ICL) refers to the ability of a model to condition on a few in-context demonstrations (input-output examples of the underlying task) to generate the answer for a new query input, without updating parameters. Despite the impressive ICL ability of LLMs, it has also been found that ICL in LLMs is sensitive to input demonstrations and limited to short context lengths. To understand the limitations and principles for successful ICL, we conduct an investigation with ICL linear regression of transformers. We characterize several Out-of-Distribution (OOD) cases for ICL inspired by realistic LLM ICL failures and compare transformers with DeepSet, a simple yet powerful architecture for ICL. Surprisingly, DeepSet outperforms transformers across a variety of distribution shifts, implying that preserving permutation invariance symmetry to input demonstrations is crucial for OOD ICL. The phenomenon specifies a fundamental requirement by ICL, which we termed as ICL invariance. Nevertheless, the positional encodings in LLMs will break ICL invariance. To this end, we further evaluate transformers with identical positional encodings and find preserving ICL invariance in transformers achieves state-of-the-art performance across various ICL distribution shifts"
                    },
                    {
                        "title": "Towards out-of-distribution generalizable predictions of chemical kinetics properties",
                        "abstract": "Machine Learning (ML) techniques have found applications in estimating chemical kinetic properties. With the accumulated drug molecules identified through\"AI4drug discovery\", the next imperative lies in AI-driven design for high-throughput chemical synthesis processes, with the estimation of properties of unseen reactions with unexplored molecules. To this end, the existing ML approaches for kinetics property prediction are required to be Out-Of-Distribution (OOD) generalizable. In this paper, we categorize the OOD kinetic property prediction into three levels (structure, condition, and mechanism), revealing unique aspects of such problems. Under this framework, we create comprehensive datasets to benchmark (1) the state-of-the-art ML approaches for reaction prediction in the OOD setting and (2) the state-of-the-art graph OOD methods in kinetics property prediction problems. Our results demonstrated the challenges and opportunities in OOD kinetics property prediction. Our datasets and benchmarks can further support research in this direction."
                    },
                    {
                        "title": "Improving Graph Representation Learning by Contrastive Regularization",
                        "abstract": "Graph representation learning is an important task with applications in various areas such as online social networks, e-commerce networks, WWW and semantic webs. For unsupervised graph representation learning, many algorithms such as Node2Vec and Graph-SAGE make use of \u201cnegative sampling\u201d and/or noise contrastive estimation loss. This bears similar ideas to contrastive learning, which \u201ccontrasts\u201d the node representation similarities of semantically similar (positive) pairs against those of negative pairs. However, despite the success of contrastive learning, we found that directly applying this technique to graph representation learning models (e.g., graph convolutional networks) does not always work. We theoretically analyze the generalization performance and pro-pose a light-weight regularization term that avoids the high scales of node representations\u2019 norms and the high variance among them to improve the generalization performance. Our experimental re-sults further validate that this regularization term significantly improves the representation quality across different node similarity definitions and outperforms the state-of-the-art methods."
                    }
                ]
            }
        }
    },
    "2406.14991": {
        "paper_data": {
            "title": "SpreadsheetBench: Towards Challenging Real World Spreadsheet Manipulation",
            "url": "http://arxiv.org/abs/2406.14991v2",
            "arxiv_id": "2406.14991",
            "authors": [
                "Zeyao Ma",
                "Bohan Zhang",
                "Jing Zhang",
                "Jifan Yu",
                "Xiaokang Zhang",
                "Xiaohan Zhang",
                "Sijia Luo",
                "Xi Wang",
                "Jie Tang"
            ],
            "abstract": "We introduce SpreadsheetBench, a challenging spreadsheet manipulation benchmark exclusively derived from real-world scenarios, designed to immerse current large language models (LLMs) in the actual workflow of spreadsheet users. Unlike existing benchmarks that rely on synthesized queries and simplified spreadsheet files, SpreadsheetBench is built from 912 real questions gathered from online Excel forums, which reflect the intricate needs of users. The associated spreadsheets from the forums contain a variety of tabular data such as multiple tables, non-standard relational tables, and abundant non-textual elements. Furthermore, we propose a more reliable evaluation metric akin to online judge platforms, where multiple spreadsheet files are created as test cases for each instruction, ensuring the evaluation of robust solutions capable of handling spreadsheets with varying values. Our comprehensive evaluation of various LLMs under both single-round and multi-round inference settings reveals a substantial gap between the state-of-the-art (SOTA) models and human performance, highlighting the benchmark's difficulty.",
            "introduction": "   1 Introduction  Today, millions of users engage with spreadsheets across various fields, including management, marketing, and finance\u00a0[1, 2, 3]. Assisting users with spreadsheet manipulation can significantly reduce their workload, thereby enhancing productivity and efficiency\u00a0[4, 5]. Initially, program synthesis aided spreadsheet manipulation\u00a0[6, 7], followed by deep learning methods for automated manipulation\u00a0[8, 9]. More recently, leveraging the powerful capabilities of LLMs\u00a0[10], spreadsheet agents such as SheetCopilot[11] and SheetAgent\u00a0[12] have been developed to assist people in manipulating spreadsheets via natural language instructions. These agents employ LLMs to transform instructions into solutions, typically presented as executable code for spreadsheet operations.   However, current spreadsheet-related agents\u00a0[11, 12] are still far from being truly helpful assistants for automating spreadsheet tasks. Users of Copilot and similar agents find these tools useful for reducing the effort of searching online, but they do not necessarily improve task completion time or success rates\u00a0[13]. A critical reason for this shortcoming is that the benchmarks\u00a0[11, 12, 14] used to evaluate these agents do not truly reflect real and challenging user demands. The primary limitations could be summarized as follows.   Figure 1: Comparison of previous SheetCopilotBench and our benchmark. The instructions in our benchmark are more complex, reflecting real user demands, issues encountered, previous user attempts, and an output example. The spreadsheets in our benchmark organize data more flexibly. The figure provides a summary of the various data organization types used in our spreadsheets, with three non-standard relational tables featuring nested header, incomplete header and missing header, in addition to cells containing pure textual information and non-textual elements like colors.   First, these benchmarks either use the self-instruct technique\u00a0[15] to synthesize queries based on a limited set of real-world spreadsheet manipulation demands\u00a0[11, 12], or they rely on crowdworkers to manually create instructions for given spreadsheets based on a few manipulation examples\u00a0[14]. As illustrated in the left part of Figure\u00a01, real user questions collected from online forums (e.g., excelforum.com) differ significantly from synthetic queries, presenting complex demands that often require additional context, such as the users\u2019 previous attempts and encountered issues.   Second, the spreadsheets used in these benchmarks are overly simplified\u00a0[11, 12], typically containing only one regular relational table, similar to the tables used in TableQA tasks\u00a0[16, 17] and the tables in databases of Text2SQL tasks\u00a0[18, 19]. While both spreadsheets and relational tables store tabular data in a two-dimensional structure, there are significant differences that make spreadsheets more challenging to handle. As shown in the right part of Figure\u00a01, spreadsheets offer a flexible way of organizing data without distinct table boundaries or structures, permitting the inclusion of multiple tables in non-standard formats\u2014such as nested, incomplete, or missing headers\u2014and the presence of free-form text within a single cell. Additionally, spreadsheets support various non-textual elements (e.g., color, bold) compared to the relational tables in databases\u00a0[18] or the CSV/JSON format tables in TableQA tasks\u00a0[16, 20]. These distinct features are not fully utilized in current benchmarks.   Third, these benchmarks for LLMs typically involve a single test for each instruction, where the predicted answer, derived from the solution generated by the LLMs applied to the given spreadsheet, is compared to the ground truth answer\u00a0[11, 12]. This evaluation, however, may result in",
            "references": [
                {
                    "title": "FREB-TQA: A Fine-Grained Robustness Evaluation Benchmark for Table Question Answering",
                    "abstract": "Table Question Answering (TQA) aims at composing an answer to a question based on tabular data. While prior research has shown that TQA models lack robustness, understanding the underlying cause and nature of this issue remains predominantly unclear, posing a significant obstacle to the development of robust TQA systems. In this paper, we formalize three major desiderata for a fine-grained evaluation of robustness of TQA systems. They should (i) answer questions regardless of alterations in table structure, (ii) base their responses on the content of relevant cells rather than on biases, and (iii) demonstrate robust numerical reasoning capabilities. To investigate these aspects, we create and publish a novel TQA evaluation benchmark in English. Our extensive experimental analysis reveals that none of the examined state-of-the-art TQA systems consistently excels in these three aspects. Our benchmark is a crucial instrument for monitoring the behavior of TQA systems and paves the way for the development of robust TQA systems. We release our benchmark publicly."
                },
                {
                    "title": "Auto-Formula: Recommend Formulas in Spreadsheets using Contrastive Learning for Table Representations",
                    "abstract": "Spreadsheets are widely recognized as the most popular end-user programming tools, which blend the power of formula-based computation, with an intuitive table-based interface. Today, spreadsheets are used by billions of users to manipulate tables, most of whom are neither database experts nor professional programmers.\n Despite the success of spreadsheets, authoring complex formulas remains challenging, as non-technical users need to look up and understand non-trivial formula syntax. To address this pain point, we leverage the observation that there is often an abundance of similar-looking spreadsheets in the same organization, which not only have similar data, but also share similar computation logic encoded as formulas. We develop an Auto-Formula system that can accurately predict formulas that users want to author in a target spreadsheet cell, by learning and adapting formulas that already exist in similar spreadsheets, using contrastive-learning techniques inspired by \"similar-face recognition\" from compute vision. Extensive evaluations on over 2K test formulas extracted from real enterprise spreadsheets show the effectiveness of Auto-Formula over alternatives. Our benchmark data is available at https://github.com/microsoft/Auto-Formula to facilitate future research."
                },
                {
                    "title": "Tapilot-Crossing: Benchmarking and Evolving LLMs Towards Interactive Data Analysis Agents",
                    "abstract": "Interactive Data Analysis, the collaboration between humans and LLM agents, enables real-time data exploration for informed decision-making. The challenges and costs of collecting realistic interactive logs for data analysis hinder the quantitative evaluation of Large Language Model (LLM) agents in this task. To mitigate this issue, we introduce Tapilot-Crossing, a new benchmark to evaluate LLM agents on interactive data analysis. Tapilot-Crossing contains 1024 interactions, covering 4 practical scenarios: Normal, Action, Private, and Private Action. Notably, Tapilot-Crossing is constructed by an economical multi-agent environment, Decision Company, with few human efforts. We evaluate popular and advanced LLM agents in Tapilot-Crossing, which underscores the challenges of interactive data analysis. Furthermore, we propose Adaptive Interaction Reflection (AIR), a self-generated reflection strategy that guides LLM agents to learn from successful history. Experiments demonstrate that Air can evolve LLMs into effective interactive data analysis agents, achieving a relative performance improvement of up to 44.5%."
                },
                {
                    "title": "OpenCodeInterpreter: Integrating Code Generation with Execution and Refinement",
                    "abstract": "The introduction of large language models has significantly advanced code generation. However, open-source models often lack the execution capabilities and iterative refinement of advanced systems like the GPT-4 Code Interpreter. To address this, we introduce OpenCodeInterpreter, a family of open-source code systems designed for generating, executing, and iteratively refining code. Supported by Code-Feedback, a dataset featuring 68K multi-turn interactions, OpenCodeInterpreter integrates execution and human feedback for dynamic code refinement. Our comprehensive evaluation of OpenCodeInterpreter across key benchmarks such as HumanEval, MBPP, and their enhanced versions from EvalPlus reveals its exceptional performance. Notably, OpenCodeInterpreter-33B achieves an accuracy of 83.2 (76.4) on the average (and plus versions) of HumanEval and MBPP, closely rivaling GPT-4's 84.2 (76.2) and further elevates to 91.6 (84.6) with synthesized human feedback from GPT-4. OpenCodeInterpreter brings the gap between open-source code generation models and proprietary systems like GPT-4 Code Interpreter."
                },
                {
                    "title": "DeepSeek-Coder: When the Large Language Model Meets Programming - The Rise of Code Intelligence",
                    "abstract": "The rapid development of large language models has revolutionized code intelligence in software development. However, the predominance of closed-source models has restricted extensive research and development. To address this, we introduce the DeepSeek-Coder series, a range of open-source code models with sizes from 1.3B to 33B, trained from scratch on 2 trillion tokens. These models are pre-trained on a high-quality project-level code corpus and employ a fill-in-the-blank task with a 16K window to enhance code generation and infilling. Our extensive evaluations demonstrate that DeepSeek-Coder not only achieves state-of-the-art performance among open-source code models across multiple benchmarks but also surpasses existing closed-source models like Codex and GPT-3.5. Furthermore, DeepSeek-Coder models are under a permissive license that allows for both research and unrestricted commercial use."
                },
                {
                    "title": "InfiAgent-DABench: Evaluating Agents on Data Analysis Tasks",
                    "abstract": "In this paper, we introduce InfiAgent-DABench, the first benchmark specifically designed to evaluate LLM-based agents on data analysis tasks. These tasks require agents to end-to-end solving complex tasks by interacting with an execution environment. This benchmark contains DAEval, a dataset consisting of 257 data analysis questions derived from 52 CSV files, and an agent framework which incorporates LLMs to serve as data analysis agents for both serving and evaluation. Since data analysis questions are often open-ended and hard to evaluate without human supervision, we adopt a format-prompting technique to convert each question into a closed-form format so that they can be automatically evaluated. Our extensive benchmarking of 34 LLMs uncovers the current challenges encountered in data analysis tasks. In addition, building on top of our agent framework, we develop a specialized agent, DAAgent, which surpasses GPT-3.5 by 3.9% on DABench. Evaluation datasets and toolkits for InfiAgent-DABench are released at https://github.com/InfiAgent/InfiAgent ."
                },
                {
                    "title": "Mixtral of Experts",
                    "abstract": "We introduce Mixtral 8x7B, a Sparse Mixture of Experts (SMoE) language model. Mixtral has the same architecture as Mistral 7B, with the difference that each layer is composed of 8 feedforward blocks (i.e. experts). For every token, at each layer, a router network selects two experts to process the current state and combine their outputs. Even though each token only sees two experts, the selected experts can be different at each timestep. As a result, each token has access to 47B parameters, but only uses 13B active parameters during inference. Mixtral was trained with a context size of 32k tokens and it outperforms or matches Llama 2 70B and GPT-3.5 across all evaluated benchmarks. In particular, Mixtral vastly outperforms Llama 2 70B on mathematics, code generation, and multilingual benchmarks. We also provide a model fine-tuned to follow instructions, Mixtral 8x7B - Instruct, that surpasses GPT-3.5 Turbo, Claude-2.1, Gemini Pro, and Llama 2 70B - chat model on human benchmarks. Both the base and instruct models are released under the Apache 2.0 license."
                },
                {
                    "title": "Text2Analysis: A Benchmark of Table Question Answering with Advanced Data Analysis and Unclear Queries",
                    "abstract": "Tabular data analysis is crucial in various fields, and large language models show promise in this area. However, current research mostly focuses on rudimentary tasks like Text2SQL and TableQA, neglecting advanced analysis like forecasting and chart generation. To address this gap, we developed the Text2Analysis benchmark, incorporating advanced analysis tasks that go beyond the SQL-compatible operations and require more in-depth analysis. We also develop five innovative and effective annotation methods, harnessing the capabilities of large language models to enhance data quality and quantity. Additionally, we include unclear queries that resemble real-world user questions to test how well models can understand and tackle such challenges. Finally, we collect 2249 query-result pairs with 347 tables. We evaluate five state-of-the-art models using three different metrics and the results show that our benchmark presents introduces considerable challenge in the field of tabular data analysis, paving the way for more advanced research opportunities."
                },
                {
                    "title": "InstructExcel: A Benchmark for Natural Language Instruction in Excel",
                    "abstract": "With the evolution of Large Language Models (LLMs) we can solve increasingly more complex NLP tasks across various domains, including spreadsheets. This work investigates whether LLMs can generate code (Excel OfficeScripts, a TypeScript API for executing many tasks in Excel) that solves Excel specific tasks provided via natural language user instructions. To do so we introduce a new large-scale benchmark, InstructExcel, created by leveraging the 'Automate' feature in Excel to automatically generate OfficeScripts from users' actions. Our benchmark includes over 10k samples covering 170+ Excel operations across 2,000 publicly available Excel spreadsheets. Experiments across various zero-shot and few-shot settings show that InstructExcel is a hard benchmark for state of the art models like GPT-4. We observe that (1) using GPT-4 over GPT-3.5, (2) providing more in-context examples, and (3) dynamic prompting can help improve performance on this benchmark."
                },
                {
                    "title": "Qwen Technical Report",
                    "abstract": "Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural language processing tasks that were previously thought to be exclusive to humans. In this work, we introduce Qwen, the first installment of our large language model series. Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts. It includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques. The base language models consistently demonstrate superior performance across a multitude of downstream tasks, and the chat models, particularly those trained using Reinforcement Learning from Human Feedback (RLHF), are highly competitive. The chat models possess advanced tool-use and planning capabilities for creating agent applications, showcasing impressive performance even when compared to bigger models on complex tasks like utilizing a code interpreter. Furthermore, we have developed coding-specialized models, Code-Qwen and Code-Qwen-Chat, as well as mathematics-focused models, Math-Qwen-Chat, which are built upon base language models. These models demonstrate significantly improved performance in comparison with open-source models, and slightly fall behind the proprietary models."
                },
                {
                    "title": "SheetCopilot: Bringing Software Productivity to the Next Level through Large Language Models",
                    "abstract": "Computer end users have spent billions of hours completing daily tasks like tabular data processing and project timeline scheduling. Most of these tasks are repetitive and error-prone, yet most end users lack the skill to automate these burdensome works. With the advent of large language models (LLMs), directing software with natural language user requests become a reachable goal. In this work, we propose a SheetCopilot agent that takes natural language task and control spreadsheet to fulfill the requirements. We propose a set of atomic actions as an abstraction of spreadsheet software functionalities. We further design a state machine-based task planning framework for LLMs to robustly interact with spreadsheets. We curate a representative dataset containing 221 spreadsheet control tasks and establish a fully automated evaluation pipeline for rigorously benchmarking the ability of LLMs in software control tasks. Our SheetCopilot correctly completes 44.3\\% of tasks for a single generation, outperforming the strong code generation baseline by a wide margin. Our project page:https://sheetcopilot.github.io/."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "LoFT: Enhancing Faithfulness and Diversity for Table-to-Text Generation via Logic Form Control",
                    "abstract": "Logical Table-to-Text (LT2T) generation is tasked with generating logically faithful sentences from tables. There currently exists two challenges in the field: 1) Faithfulness: how to generate sentences that are factually correct given the table content; 2) Diversity: how to generate multiple sentences that offer different perspectives on the table. This work proposes LoFT, which utilizes logic forms as fact verifiers and content planners to control LT2T generation. Experimental results on the LogicNLG dataset demonstrate that LoFT is the first model that addresses unfaithfulness and lack of diversity issues simultaneously. Our code is publicly available at https://github.com/Yale-LILY/LoFT."
                },
                {
                    "title": "Self-Instruct: Aligning Language Models with Self-Generated Instructions",
                    "abstract": "Large \u201cinstruction-tuned\u201d language models (i.e., finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is often limited in quantity, diversity, and creativity, therefore hindering the generality of the tuned model. We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off their own generations. Our pipeline generates instructions, input, and output samples from a language model, then filters invalid or similar ones before using them to finetune the original model. Applying our method to the vanilla GPT3, we demonstrate a 33% absolute improvement over the original model on Super-NaturalInstructions, on par with the performance of InstructGPT-001, which was trained with private user data and human annotations. For further evaluation, we curate a set of expert-written instructions for novel tasks, and show through human evaluation that tuning GPT3 with Self-Instruct outperforms using existing public instruction datasets by a large margin, leaving only a 5% absolute gap behind InstructGPT-001. Self-Instruct provides an almost annotation-free method for aligning pre-trained language models with instructions, and we release our large synthetic dataset to facilitate future studies on instruction tuning."
                },
                {
                    "title": "DS-1000: A Natural and Reliable Benchmark for Data Science Code Generation",
                    "abstract": "We introduce DS-1000, a code generation benchmark with a thousand data science problems spanning seven Python libraries, such as NumPy and Pandas. Compared to prior works, DS-1000 incorporates three core features. First, our problems reflect diverse, realistic, and practical use cases since we collected them from StackOverflow. Second, our automatic evaluation is highly specific (reliable) -- across all Codex-002-predicted solutions that our evaluation accept, only 1.8% of them are incorrect; we achieve this with multi-criteria metrics, checking both functional correctness by running test cases and surface-form constraints by restricting API usages or keywords. Finally, we proactively defend against memorization by slightly modifying our problems to be different from the original StackOverflow source; consequently, models cannot answer them correctly by memorizing the solutions from pre-training. The current best public system (Codex-002) achieves 43.3% accuracy, leaving ample room for improvement. We release our benchmark at https://ds1000-code-gen.github.io."
                },
                {
                    "title": "Binding Language Models in Symbolic Languages",
                    "abstract": "Though end-to-end neural approaches have recently been dominating NLP tasks in both performance and ease-of-use, they lack interpretability and robustness. We propose Binder, a training-free neural-symbolic framework that maps the task input to a program, which (1) allows binding a unified API of language model (LM) functionalities to a programming language (e.g., SQL, Python) to extend its grammar coverage and thus tackle more diverse questions, (2) adopts an LM as both the program parser and the underlying model called by the API during execution, and (3) requires only a few in-context exemplar annotations. Specifically, we employ GPT-3 Codex as the LM. In the parsing stage, with only a few in-context exemplars, Codex is able to identify the part of the task input that cannot be answerable by the original programming language, correctly generate API calls to prompt Codex to solve the unanswerable part, and identify where to place the API calls while being compatible with the original grammar. In the execution stage, Codex can perform versatile functionalities (e.g., commonsense QA, information extraction) given proper prompts in the API calls. Binder achieves state-of-the-art results on WikiTableQuestions and TabFact datasets, with explicit output programs that benefit human debugging. Note that previous best systems are all finetuned on tens of thousands of task-specific samples, while Binder only uses dozens of annotations as in-context exemplars without any training. Our code is available at https://github.com/HKUNLP/Binder ."
                },
                {
                    "title": "ReAct: Synergizing Reasoning and Acting in Language Models",
                    "abstract": "While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources, such as knowledge bases or environments, to gather additional information. We apply our approach, named ReAct, to a diverse set of language and decision making tasks and demonstrate its effectiveness over state-of-the-art baselines, as well as improved human interpretability and trustworthiness over methods without reasoning or acting components. Concretely, on question answering (HotpotQA) and fact verification (Fever), ReAct overcomes issues of hallucination and error propagation prevalent in chain-of-thought reasoning by interacting with a simple Wikipedia API, and generates human-like task-solving trajectories that are more interpretable than baselines without reasoning traces. On two interactive decision making benchmarks (ALFWorld and WebShop), ReAct outperforms imitation and reinforcement learning methods by an absolute success rate of 34% and 10% respectively, while being prompted with only one or two in-context examples. Project site with code: https://react-lm.github.io"
                },
                {
                    "title": "Expectation vs.\u00a0Experience: Evaluating the Usability of Code Generation Tools Powered by Large Language Models",
                    "abstract": "Recent advances in Large Language Models (LLM) have made automatic code generation possible for real-world programming tasks in general-purpose programming languages such as Python. However, there are few human studies on the usability of these tools and how they fit the programming workflow. In this work, we conducted a within-subjects user study with 24 participants to understand how programmers use and perceive Copilot, a LLM-based code generation tool. We found that, while Copilot did not necessarily improve the task completion time or success rate, most participants preferred to use Copilot in daily programming tasks, since Copilot often provided a useful starting point and saved the effort of searching online. However, participants did face difficulties in understanding, editing, and debugging code snippets generated by Copilot, which significantly hindered their task-solving effectiveness. Finally, we highlighted several promising directions for improving the design of Copilot based on our observations and participants\u2019 feedback."
                },
                {
                    "title": "HiTab: A Hierarchical Table Dataset for Question Answering and Natural Language Generation",
                    "abstract": "Tables are often created with hierarchies, but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables. Hierarchical tables challenge numerical reasoning by complex hierarchical indexing, as well as implicit relationships of calculation and semantics. We present a new dataset, HiTab, to study question answering (QA) and natural language generation (NLG) over hierarchical tables. HiTab is a cross-domain dataset constructed from a wealth of statistical reports and Wikipedia pages, and has unique characteristics: (1) nearly all tables are hierarchical, and (2) QA pairs are not proposed by annotators from scratch, but are revised from real and meaningful sentences authored by analysts. (3) to reveal complex numerical reasoning in statistical reports, we provide fine-grained annotations of quantity and entity alignment. Experiments suggest that this HiTab presents a strong challenge for existing baselines and a valuable benchmark for future research. Targeting hierarchical structure, we devise a hierarchy-aware logical form for symbolic reasoning over tables, which shows high effectiveness. Targeting table reasoning, we leverage entity and quantity alignment to explore partially supervised training in QA and conditional generation in NLG, and largely reduce spurious predictions in QA and produce better descriptions in NLG."
                },
                {
                    "title": "TAPEX: Table Pre-training via Learning a Neural SQL Executor",
                    "abstract": "Recent progress in language model pre-training has achieved a great success via leveraging large-scale unstructured textual data. However, it is still a challenge to apply pre-training on structured tabular data due to the absence of large-scale high-quality tabular data. In this paper, we propose TAPEX to show that table pre-training can be achieved by learning a neural SQL executor over a synthetic corpus, which is obtained by automatically synthesizing executable SQL queries and their execution outputs. TAPEX addresses the data scarcity challenge via guiding the language model to mimic a SQL executor on the diverse, large-scale and high-quality synthetic corpus. We evaluate TAPEX on four benchmark datasets. Experimental results demonstrate that TAPEX outperforms previous table pre-training approaches by a large margin and achieves new state-of-the-art results on all of them. This includes the improvements on the weakly-supervised WikiSQL denotation accuracy to 89.5% (+2.3%), the WikiTableQuestions denotation accuracy to 57.5% (+4.8%), the SQA denotation accuracy to 74.5% (+3.5%), and the TabFact accuracy to 84.2% (+3.2%). To our knowledge, this is the first work to exploit table pre-training via synthetic executable programs and to achieve new state-of-the-art results on various downstream tasks. Our code can be found at https://github.com/microsoft/Table-Pretraining."
                },
                {
                    "title": "SpreadsheetCoder: Formula Prediction from Semi-structured Context",
                    "abstract": "Spreadsheet formula prediction has been an important program synthesis problem with many real-world applications. Previous works typically utilize input-output examples as the specification for spreadsheet formula synthesis, where each input-output pair simulates a separate row in the spreadsheet. However, this formulation does not fully capture the rich context in real-world spreadsheets. First, spreadsheet data entries are organized as tables, thus rows and columns are not necessarily independent from each other. In addition, many spreadsheet tables include headers, which provide high-level descriptions of the cell data. However, previous synthesis approaches do not consider headers as part of the specification. In this work, we present the first approach for synthesizing spreadsheet formulas from tabular context, which includes both headers and semi-structured tabular data. In particular, we propose SpreadsheetCoder, a BERT-based model architecture to represent the tabular context in both row-based and column-based formats. We train our model on a large dataset of spreadsheets, and demonstrate that SpreadsheetCoder achieves top-1 prediction accuracy of 42.51%, which is a considerable improvement over baselines that do not employ rich tabular context. Compared to the rule-based system, SpreadsheetCoder assists 82% more users in composing formulas on Google Sheets."
                },
                {
                    "title": "AIT-QA: Question Answering Dataset over Complex Tables in the Airline Industry",
                    "abstract": "Table Question Answering (Table QA) systems have been shown to be highly accurate when trained and tested on open-domain datasets built on top of Wikipedia tables. However, it is not clear whether their performance remains the same when applied to domain-specific scientific and business documents, encountered in industrial settings, which exhibit some unique characteristics: (a) they contain tables with a much more complex layout than Wikipedia tables (including hierarchical row and column headers), (b) they contain domain-specific terms, and (c) they are typically not accompanied by domain-specific labeled data that can be used to train Table QA models.To understand the performance of Table QA approaches in this setting, we introduce AIT-QA; a domain-specific Table QA test dataset. While focusing on the airline industry, AIT-QA reflects the challenges that domain-specific documents pose to Table QA, outlined above. In this work, we describe the creation of the dataset and report zero-shot experimental results of three SOTA Table QA methods. The results clearly expose the limitations of current methods with a best accuracy of just 51.8%. We also present pragmatic table pre-processing steps to pivot and project complex tables into a layout suitable for the SOTA Table QA models. Finally, we provide data-driven insights on how different aspects of this setting (including hierarchical headers, domain-specific terminology, and paraphrasing) affect Table QA methods, in order to help the community develop improved methods for domain-specific Table QA."
                },
                {
                    "title": "TAT-QA: A Question Answering Benchmark on a Hybrid of Tabular and Textual Content in Finance",
                    "abstract": "Hybrid data combining both tabular and textual content (e.g., financial reports) are quite pervasive in the real world. However, Question Answering (QA) over such hybrid data is largely neglected in existing research. In this work, we extract samples from real financial reports to build a new large-scale QA dataset containing both Tabular And Textual data, named TAT-QA, where numerical reasoning is usually required to infer the answer, such as addition, subtraction, multiplication, division, counting, comparison/sorting, and the compositions. We further propose a novel QA model termed TAGOP, which is capable of reasoning over both tables and text. It adopts sequence tagging to extract relevant cells from the table along with relevant spans from the text to infer their semantics, and then applies symbolic reasoning over them with a set of aggregation operators to arrive at the final answer. TAGOP achieves 58.0% inF1, which is an 11.1% absolute increase over the previous best baseline model, according to our experiments on TAT-QA. But this result still lags far behind performance of expert human, i.e.90.8% in F1. It is demonstrated that our TAT-QA is very challenging and can serve as a benchmark for training and testing powerful QA models that address hybrid form data."
                },
                {
                    "title": "FeTaQA: Free-form Table Question Answering",
                    "abstract": "Existing table question answering datasets contain abundant factual questions that primarily evaluate a QA system\u2019s comprehension of query and tabular data. However, restricted by their short-form answers, these datasets fail to include question\u2013answer interactions that represent more advanced and naturally occurring information needs: questions that ask for reasoning and integration of information pieces retrieved from a structured knowledge source. To complement the existing datasets and to reveal the challenging nature of the table-based question answering task, we introduce FeTaQA, a new dataset with 10K Wikipedia-based table, question, free-form answer, supporting table cells pairs. FeTaQA is collected from noteworthy descriptions of Wikipedia tables that contain information people tend to seek; generation of these descriptions requires advanced processing that humans perform on a daily basis: Understand the question and table, retrieve, integrate, infer, and conduct text planning and surface realization to generate an answer. We provide two benchmark methods for the proposed task: a pipeline method based on semantic parsing-based QA systems and an end-to-end method based on large pretrained text generation models, and show that FeTaQA poses a challenge for both methods."
                },
                {
                    "title": "Memorization vs. Generalization : Quantifying Data Leakage in NLP Performance Evaluation",
                    "abstract": "Public datasets are often used to evaluate the efficacy and generalizability of state-of-the-art methods for many tasks in natural language processing (NLP). However, the presence of overlap between the train and test datasets can lead to inflated results, inadvertently evaluating the model\u2019s ability to memorize and interpreting it as the ability to generalize. In addition, such data sets may not provide an effective indicator of the performance of these methods in real world scenarios. We identify leakage of training data into test data on several publicly available datasets used to evaluate NLP tasks, including named entity recognition and relation extraction, and study them to assess the impact of that leakage on the model\u2019s ability to memorize versus generalize."
                },
                {
                    "title": "Open Question Answering over Tables and Text",
                    "abstract": "In open question answering (QA), the answer to a question is produced by retrieving and then analyzing documents that might contain answers to the question. Most open QA systems have considered only retrieving information from unstructured text. Here we consider for the first time open QA over both tabular and textual data and present a new large-scale dataset Open Table-and-Text Question Answering (OTT-QA) to evaluate performance on this task. Most questions in OTT-QA require multi-hop inference across tabular data and unstructured text, and the evidence required to answer a question can be distributed in different ways over these two types of input, making evidence retrieval challenging -- our baseline model using an iterative retriever and BERT-based reader achieves an exact match score less than 10%. We then propose two novel techniques to address the challenge of retrieving and aggregating evidence for OTT-QA. The first technique is to use \"early fusion\" to group multiple highly relevant tabular and textual units into a fused block, which provides more context for the retriever to search for. The second technique is to use a cross-block reader to model the cross-dependency between multiple retrieved evidence with global-local sparse attention. Combining these two techniques improves the score significantly, to above 27%."
                },
                {
                    "title": "TaPas: Weakly Supervised Table Parsing via Pre-training",
                    "abstract": "Answering natural language questions over tables is usually seen as a semantic parsing task. To alleviate the collection cost of full logical forms, one popular approach focuses on weak supervision consisting of denotations instead of logical forms. However, training semantic parsers from weak supervision poses difficulties, and in addition, the generated logical forms are only used as an intermediate step prior to retrieving the denotation. In this paper, we present TaPas, an approach to question answering over tables without generating logical forms. TaPas trains from weak supervision, and predicts the denotation by selecting table cells and optionally applying a corresponding aggregation operator to such selection. TaPas extends BERT\u2019s architecture to encode tables as input, initializes from an effective joint pre-training of text segments and tables crawled from Wikipedia, and is trained end-to-end. We experiment with three different semantic parsing datasets, and find that TaPas outperforms or rivals semantic parsing models by improving state-of-the-art accuracy on SQA from 55.1 to 67.2 and performing on par with the state-of-the-art on WikiSQL and WikiTQ, but with a simpler model architecture. We additionally find that transfer learning, which is trivial in our setting, from WikiSQL to WikiTQ, yields 48.7 accuracy, 4.2 points above the state-of-the-art."
                },
                {
                    "title": "Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task",
                    "abstract": "We present Spider, a large-scale complex and cross-domain semantic parsing and text-to-SQL dataset annotated by 11 college students. It consists of 10,181 questions and 5,693 unique complex SQL queries on 200 databases with multiple tables covering 138 different domains. We define a new complex and cross-domain semantic parsing and text-to-SQL task so that different complicated SQL queries and databases appear in train and test sets. In this way, the task requires the model to generalize well to both new SQL queries and new database schemas. Therefore, Spider is distinct from most of the previous semantic parsing tasks because they all use a single database and have the exact same program in the train set and the test set. We experiment with various state-of-the-art models and the best model achieves only 9.7% exact matching accuracy on a database split setting. This shows that Spider presents a strong challenge for future research. Our dataset and task with the most recent updates are publicly available at https://yale-lily.github.io/seq2sql/spider."
                },
                {
                    "title": "A Survey on Online Judge Systems and Their Applications",
                    "abstract": "Online judges are systems designed for the reliable evaluation of algorithm source code submitted by users, which is next compiled and tested in a homogeneous environment. Online judges are becoming popular in various applications. Thus, we would like to review the state of the art for these systems. We classify them according to their principal objectives into systems supporting organization of competitive programming contests, enhancing education and recruitment processes, facilitating the solving of data mining challenges, online compilers and development platforms integrated as components of other custom systems. Moreover, we introduce a formal definition of an online judge system and summarize the common evaluation methodology supported by such systems. Finally, we briefly discuss an Optil.io platform as an example of an online judge system, which has been proposed for the solving of complex optimization problems. We also analyze the competition results conducted using this platform. The competition proved that online judge systems, strengthened by crowdsourcing concepts, can be successfully applied to accurately and efficiently solve complex industrial- and science-driven challenges."
                },
                {
                    "title": "Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning",
                    "abstract": "Relational databases store a significant amount of the worlds data. However, accessing this data currently requires users to understand a query language such as SQL. We propose Seq2SQL, a deep neural network for translating natural language questions to corresponding SQL queries. Our model uses rewards from in the loop query execution over the database to learn a policy to generate the query, which contains unordered parts that are less suitable for optimization via cross entropy loss. Moreover, Seq2SQL leverages the structure of SQL to prune the space of generated queries and significantly simplify the generation problem. In addition to the model, we release WikiSQL, a dataset of 80654 hand-annotated examples of questions and SQL queries distributed across 24241 tables fromWikipedia that is an order of magnitude larger than comparable datasets. By applying policy based reinforcement learning with a query execution environment to WikiSQL, Seq2SQL outperforms a state-of-the-art semantic parser, improving execution accuracy from 35.9% to 59.4% and logical form accuracy from 23.4% to 48.3%."
                },
                {
                    "title": "CUSTODES: Automatic Spreadsheet Cell Clustering and Smell Detection Using Strong and Weak Features",
                    "abstract": "Various techniques have been proposed to detect smells in spreadsheets, which are susceptible to errors. These techniques typically detect spreadsheet smells through a mechanism based on a fixed set of patterns or metric thresholds. Unlike conventional programs, tabulation styles vary greatly across spreadsheets. Smell detection based on fixed patterns or metric thresholds, which are insensitive to the varying tabulation styles, can miss many smells in one spreadsheet while reporting many spurious smells in another. In this paper, we propose CUSTODES to effectively cluster spreadsheet cells and detect smells in these clusters. The clustering mechanism can automatically adapt to the tabulation styles of each spreadsheet using strong and weak features. These strong and weak features capture the invariant and variant parts of tabulation styles, respectively. As smelly cells in a spreadsheet normally occur in minority, they can be mechanically detected as clusters\u2019 outliers in feature spaces. We implemented and applied CUSTODES to 70 spreadsheets files randomly sampled from the EUSES corpus. These spreadsheets contain 1,610 formula cell clusters. Experimental results confirmed that CUSTODES is effective. It successfully detected harmful smells that can induce computation anomalies in spreadsheets with an F-measure of 0.72, outperforming state-of-the-art techniques."
                },
                {
                    "title": "Compositional Semantic Parsing on Semi-Structured Tables",
                    "abstract": "Two important aspects of semantic parsing for question answering are the breadth of the knowledge source and the depth of logical compositionality. While existing work trades off one aspect for another, this paper simultaneously makes progress on both fronts through a new task: answering complex questions on semi-structured tables using question-answer pairs as supervision. The central challenge arises from two compounding factors: the broader domain results in an open-ended set of relations, and the deeper compositionality results in a combinatorial explosion in the space of logical forms. We propose a logical-form driven parsing algorithm guided by strong typing constraints and show that it obtains significant improvements over natural baselines. For evaluation, we created a new dataset of 22,033 complex questions on Wikipedia tables, which is made publicly available."
                },
                {
                    "title": "Marketing Analytics: Data-Driven Techniques with Microsoft Excel",
                    "abstract": "Helping tech-savvy marketers and data analysts solve real-world business problems with Excel Using data-driven business analytics to understand customers and improve results is a great idea in theory, but in today's busy offices, marketers and analysts need simple, low-cost ways to process and make the most of all that data. This expert book offers the perfect solution. Written by data analysis expert Wayne L. Winston, this practical resource shows you how to tap a simple and cost-effective tool, Microsoft Excel, to solve specific business problems using powerful analytic techniques and achieve optimum results. Practical exercises in each chapter help you apply and reinforce techniques as you learn. * Shows you how to perform sophisticated business analyses using the cost-effective and widely available Microsoft Excel instead of expensive, proprietary analytical tools * Reveals how to target and retain profitable customers and avoid high-risk customers * Helps you forecast sales and improve response rates for marketing campaigns * Explores how to optimize price points for products and services, optimize store layouts, and improve online advertising * Covers social media, viral marketing, and how to exploit both effectively Improve your marketing results with Microsoft Excel and the invaluable techniques and ideas in Marketing Analytics: Data-Driven Techniques with Microsoft Excel."
                },
                {
                    "title": "Spreadsheet data manipulation using examples",
                    "abstract": "Millions of computer end users need to perform tasks over large spreadsheet data, yet lack the programming knowledge to do such tasks automatically. We present a programming by example methodology that allows end users to automate such repetitive tasks. Our methodology involves designing a domain-specific language and developing a synthesis algorithm that can learn programs in that language from user-provided examples. We present instantiations of this methodology for particular domains of tasks: (a) syntactic transformations of strings using restricted forms of regular expressions, conditionals, and loops, (b) semantic transformations of strings involving lookup in relational tables, and (c) layout transformations on spreadsheet tables. We have implemented this technology as an add-in for the Microsoft Excel Spreadsheet system and have evaluated it successfully over several benchmarks picked from various Excel help forums."
                },
                {
                    "title": "Structure, Scoring and Purpose of Computing Competitions",
                    "abstract": "We identify aspects of computing competition formats as they relate to the purpose of these competitions, both stated and tacit. We consider the major international competitions - the International Olympiad for Informatics, the ACM International Collegiate Programming Contest, and top coder - and related contests whose format merits consideration. We consider the operational impact and possible outcomes of incorporating several of these aspects into scholastic competitions. We advocate, in particular, that contests be designed so as to provide a rewarding experience and opportunity for achievement for all competitors; not just the winners. Specific contest elements that should be considered are: (1) real-time scoring and feedback, (2) rewards for testing and test case creation, (3) tasks with graduated difficulty, (4) collaborative tasks, (5) practice contests and entry-level contests for novices, and (6) inclusion of spectators."
                },
                {
                    "title": "Advanced modelling in finance using Excel and VBA",
                    "abstract": "This new and unique book demonstrates that Excel and VBA can play an important role in the explanation and implementation of numerical methods across finance. Advanced Modelling in Finance provides a comprehensive look at equities, options on equities and options on bonds from the early 1950s to the late 1990s. \nThe book adopts a step-by-step approach to understanding the more sophisticated aspects of Excel macros and VBA programming, showing how these programming techniques can be used to model and manipulate financial data, as applied to equities, bonds and options. The book is essential for financial practitioners who need to develop their financial modelling skill sets as there is an increase in the need to analyse and develop ever more complex 'what if' scenarios."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.SE"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a more effective benchmark for evaluating spreadsheet manipulation agents that accurately reflects real user demands and the complexities of spreadsheet data organization?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it will lead to the development of more capable spreadsheet manipulation agents, enhancing their utility in real-world applications. By creating benchmarks that mirror actual user needs, we can improve the performance of LLMs in automating spreadsheet tasks, ultimately advancing knowledge in human-computer interaction and natural language processing. This could lead to practical applications that significantly reduce the time and effort users spend on spreadsheet tasks, thereby increasing productivity across various fields.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of real user demands, which often require nuanced understanding and context that current benchmarks fail to capture. Naive approaches that rely on simplified or synthetic queries do not account for the diverse and intricate ways users interact with spreadsheets. Additionally, the flexible data organization in spreadsheets, including non-standard relational tables and mixed content types, presents technical obstacles that complicate the evaluation process. Overcoming these challenges requires a deep understanding of both user behavior and the unique characteristics of spreadsheet data.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the reliance on synthetic benchmarks that do not accurately represent real-world scenarios, as well as the use of overly simplified spreadsheet structures. Barriers such as a lack of comprehensive data on actual user queries and the complexity of spreadsheet formats have hindered progress. Our approach differs by focusing on real user questions collected from online forums and incorporating a wider variety of spreadsheet structures, thus providing a more realistic and challenging evaluation framework for LLMs.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves creating a benchmark that includes complex user instructions derived from real-world queries, alongside spreadsheets that feature diverse data organization types, such as nested and incomplete headers. We will evaluate the performance of spreadsheet manipulation agents using metrics that assess task completion time and success rates, rather than relying solely on correctness of answers. The expected outcomes include improved performance of LLMs in handling spreadsheet tasks, leading to more effective and user-friendly spreadsheet manipulation tools."
            }
        },
        "author_data": {
            "4c8974fb-a9bf-4cc6-b0b5-b93535683750": {
                "pk": "4c8974fb-a9bf-4cc6-b0b5-b93535683750",
                "name": "Zeyao Ma",
                "collaborators": [
                    "Xiaokang Zhang",
                    "Jing Zhang",
                    "Zijun Yao",
                    "Daniel Zhang-Li",
                    "Jifan Yu",
                    "Juanzi Li",
                    "Jie Tang",
                    "Yang Li",
                    "Bohan Zhang",
                    "Guanlin Li"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Large Language Model",
                    "Knowledge Grounded Dialogue",
                    "Tabular Data"
                ],
                "publications": [
                    {
                        "title": "TableLLM: Enabling Tabular Data Manipulation by LLMs in Real Office Usage Scenarios",
                        "abstract": "We introduce TableLLM, a robust large language model (LLM) with 13 billion parameters, purpose-built for proficiently handling tabular data manipulation tasks, whether they are embedded within documents or spreadsheets, catering to real-world office scenarios. We propose a distant supervision method for training, which comprises a reasoning process extension strategy, aiding in training LLMs to understand reasoning patterns more effectively as well as a cross-way validation strategy, ensuring the quality of the automatically generated data. To evaluate the performance of TableLLM, we have crafted a benchmark tailored to address both document and spreadsheet formats as well as constructed a well-organized evaluation pipeline capable of handling both scenarios. Thorough evaluations underscore the advantages of TableLLM when compared to various existing general-purpose and tabular data-focused LLMs. We have publicly released the model checkpoint, source code, benchmarks, and a web application for user interaction.Our codes and data are publicly available at https://github.com/TableLLM/TableLLM."
                    },
                    {
                        "title": "GLM-Dialog: Noise-tolerant Pre-training for Knowledge-grounded Dialogue Generation",
                        "abstract": "We present GLM-Dialog, a large-scale language model (LLM) with 10B parameters capable of knowledge-grounded conversation in Chinese using a search engine to access the Internet knowledge. GLM-Dialog offers a series of applicable techniques for exploiting various external knowledge including both helpful and noisy knowledge, enabling the creation of robust knowledge-grounded dialogue LLMs with limited proper datasets. To evaluate the GLM-Dialog more fairly, we also propose a novel evaluation method to allow humans to converse with multiple deployed bots simultaneously and compare their performance implicitly instead of explicitly rating using multidimensional metrics.Comprehensive evaluations from automatic to human perspective demonstrate the advantages of GLM-Dialog comparing with existing open source Chinese dialogue models. We release both the model checkpoint and source code, and also deploy it as a WeChat application to interact with users. We offer our evaluation platform online in an effort to prompt the development of open source models and reliable dialogue evaluation systems. The additional easy-to-use toolkit that consists of short text entity linking, query generation, and helpful knowledge classification is also released to enable diverse applications. All the source code is available on Github."
                    }
                ]
            },
            "ad22e0cc-fd30-4363-bc00-8f8c4e395012": {
                "pk": "ad22e0cc-fd30-4363-bc00-8f8c4e395012",
                "name": "Bohan Zhang",
                "collaborators": [
                    "Anastasios Panagiotelis",
                    "Ankit Agrawal",
                    "Yanfei Kang",
                    "Hussain Sadiq Abuwala",
                    "Mushi Wang",
                    "Han Li",
                    "Feng Li",
                    "Yashaswini Shivalingaiah",
                    "Yixin Wang",
                    "Paramveer S. Dhillon"
                ],
                "domain": [
                    "Unmanned Aerial Systems",
                    "Natural Language Processing",
                    "Forecasting",
                    "Causal Inference"
                ],
                "publications": [
                    {
                        "title": "DroneWiS: Automated Simulation Testing of small Unmanned Aerial Systems in Realistic Windy Conditions",
                        "abstract": "The continuous evolution of small Unmanned Aerial Systems (sUAS) demands advanced testing methodologies to ensure their safe and reliable operations in the real-world. To push the boundaries of sUAS simulation testing in realistic environments, we previously developed the DroneReqValidator (DRV) platform, allowing developers to automatically conduct simulation testing in digital twin of earth. In this paper, we present DRV 2.0, which introduces a novel component called DroneWiS (Drone Wind Simulation). DroneWiS allows sUAS developers to automatically simulate realistic windy conditions and test the resilience of sUAS against wind. Unlike current state-of-the-art simulation tools such as Gazebo and AirSim that only simulate basic wind conditions, DroneWiS leverages Computational Fluid Dynamics (CFD) to compute the unique wind flows caused by the interaction of wind with the objects in the environment such as buildings and uneven terrains. This simulation capability provides deeper insights to developers about the navigation capability of sUAS in challenging and realistic windy conditions. DroneWiS equips sUAS developers with a powerful tool to test, debug, and improve the reliability and safety of sUAS in real-world. A working demonstration is available at https://youtu.be/khBHEBST8Wc"
                    },
                    {
                        "title": "Who should I Collaborate with? A Comparative Study of Academia and Industry Research Collaboration in NLP",
                        "abstract": "The goal of our research was to investigate the effects of collaboration between academia and industry on Natural Language Processing (NLP). To do this, we created a pipeline to extract affiliations and citations from NLP papers and divided them into three categories: academia, industry, and hybrid (collaborations between academia and industry). Our empirical analysis found that there is a trend towards an increase in industry and academia-industry collaboration publications and that these types of publications tend to have a higher impact compared to those produced solely within academia."
                    },
                    {
                        "title": "Constructing hierarchical time series through clustering: Is there an optimal way for forecasting?",
                        "abstract": "Forecast reconciliation has attracted significant research interest in recent years, with most studies taking the hierarchy of time series as given. We extend existing work that uses time series clustering to construct hierarchies, with the goal of improving forecast accuracy, in three ways. First, we investigate multiple approaches to clustering, including not only different clustering algorithms, but also the way time series are represented and how distance between time series is defined. We find that cluster-based hierarchies lead to improvements in forecast accuracy relative to two-level hierarchies. Second, we devise an approach based on random permutation of hierarchies, keeping the structure of the hierarchy fixed, while time series are randomly allocated to clusters. In doing so, we find that improvements in forecast accuracy that accrue from using clustering do not arise from grouping together similar series but from the structure of the hierarchy. Third, we propose an approach based on averaging forecasts across hierarchies constructed using different clustering methods, that is shown to outperform any single clustering method. All analysis is carried out on two benchmark datasets and a simulated dataset. Our findings provide new insights into the role of hierarchy construction in forecast reconciliation and offer valuable guidance on forecasting practice."
                    },
                    {
                        "title": "Optimal reconciliation with immutable forecasts",
                        "abstract": "The practical importance of coherent forecasts in hierarchical forecasting has inspired many studies on forecast reconciliation. Under this approach, so-called base forecasts are produced for every series in the hierarchy and are subsequently adjusted to be coherent in a second reconciliation step. Reconciliation methods have been shown to improve forecast accuracy, but will, in general, adjust the base forecast of every series. However, in an operational context, it is sometimes necessary or beneficial to keep forecasts of some variables unchanged after forecast reconciliation. In this paper, we formulate reconciliation methodology that keeps forecasts of a pre-specified subset of variables unchanged or \"immutable\". In contrast to existing approaches, these immutable forecasts need not all come from the same level of a hierarchy, and our method can also be applied to grouped hierarchies. We prove that our approach preserves unbiasedness in base forecasts. Our method can also account for correlations between base forecasting errors and ensure non-negativity of forecasts. We also perform empirical experiments, including an application to sales of a large scale online retailer, to assess the impacts of our proposed methodology."
                    },
                    {
                        "title": "Discrete forecast reconciliation",
                        "abstract": "This paper presents a formal framework and proposes algorithms to extend forecast reconciliation to discrete-valued data to extend forecast reconciliation to discrete-valued data, including low counts. A novel method is introduced based on recasting the optimisation of scoring rules as an assignment problem, which is solved using quadratic programming. The proposed framework produces coherent joint probabilistic forecasts for count hierarchical time series. Two discrete reconciliation algorithms are also proposed and compared against generalisations of the top-down and bottom-up approaches for count data. Two simulation experiments and two empirical examples are conducted to validate that the proposed reconciliation algorithms improve forecast accuracy. The empirical applications are forecasting criminal offences in Washington D.C. and product unit sales in the M5 dataset. Compared to benchmarks, the proposed framework shows superior performance in both simulations and empirical studies."
                    },
                    {
                        "title": "DroneReqValidator: Facilitating High Fidelity Simulation Testing for Uncrewed Aerial Systems Developers",
                        "abstract": "Rigorous testing of small Uncrewed Aerial Systems (sUAS) is crucial to ensure their safe and reliable deployment in the real world. sUAS developers aim to validate the reliability and safety of their applications through simulation testing. However, the dynamic nature of the real-world environment, including factors such as challenging weather conditions and wireless interference, causes unique software faults that may only be revealed through field testing. Considering the high cost and impracticality of conducting field testing in thousands of environmental contexts and conditions, there exists a pressing need to develop automated techniques that can generate high-fidelity, realistic environments enabling sUAS developers to deploy their applications and conduct thorough simulation testing in close-to-reality environmental conditions. To address this need, DroneReqValidator (DRV) offers a comprehensive small Unmanned Aerial Vehicle (sUAV) simulation ecosystem that automatically generates realistic environments based on developer-specified constraints, monitors sUAV activities against predefined safety parameters, and generates detailed acceptance test reports for effective debugging and analysis of sUAV applications. Providing these capabilities, DRV offers a valuable solution for enhancing the testing and development process of sUAS. The comprehensive demo of DRV is available at https://www.youtube.com/watch?v=Fd9ft55gbO8"
                    },
                    {
                        "title": "Causal Inference for Human-Language Model Collaboration",
                        "abstract": "In this paper, we examine the collaborative dynamics between humans and language models (LMs), where the interactions typically involve LMs proposing text segments and humans editing or responding to these proposals. Productive engagement with LMs in such scenarios necessitates that humans discern effective text-based interaction strategies, such as editing and response styles, from historical human-LM interactions. This objective is inherently causal, driven by the counterfactual `what-if' question: how would the outcome of collaboration change if humans employed a different text editing/refinement strategy? A key challenge in answering this causal inference question is formulating an appropriate causal estimand: the conventional average treatment effect (ATE) estimand is inapplicable to text-based treatments due to their high dimensionality. To address this concern, we introduce a new causal estimand -- Incremental Stylistic Effect (ISE) -- which characterizes the average impact of infinitesimally shifting a text towards a specific style, such as increasing formality. We establish the conditions for the non-parametric identification of ISE. Building on this, we develop CausalCollab, an algorithm designed to estimate the ISE of various interaction strategies in dynamic human-LM collaborations. Our empirical investigations across three distinct human-LM collaboration scenarios reveal that CausalCollab effectively reduces confounding and significantly improves counterfactual estimation over a set of competitive baselines."
                    },
                    {
                        "title": "MeetSum: Transforming Meeting Transcript Summarization using Transformers!",
                        "abstract": "Creating abstractive summaries from meeting transcripts has proven to be challenging due to the limited amount of labeled data available for training neural network models. Moreover, Transformer-based architectures have proven to beat state-of-the-art models in summarizing news data. In this paper, we utilize a Transformer-based Pointer Generator Network to generate abstract summaries for meeting transcripts. This model uses 2 LSTMs as an encoder and a decoder, a Pointer network which copies words from the inputted text, and a Generator network to produce out-of-vocabulary words (hence making the summary abstractive). Moreover, a coverage mechanism is used to avoid repetition of words in the generated summary. First, we show that training the model on a news summary dataset and using zero-shot learning to test it on the meeting dataset proves to produce better results than training it on the AMI meeting dataset. Second, we show that training this model first on out-of-domain data, such as the CNN-Dailymail dataset, followed by a fine-tuning stage on the AMI meeting dataset is able to improve the performance of the model significantly. We test our model on a testing set from the AMI dataset and report the ROUGE-2 score of the generated summary to compare with previous literature. We also report the Factual score of our summaries since it is a better benchmark for abstractive summaries since the ROUGE-2 score is limited to measuring word-overlaps. We show that our improved model is able to improve on previous models by at least 5 ROUGE-2 scores, which is a substantial improvement. Also, a qualitative analysis of the summaries generated by our model shows that these summaries and human-readable and indeed capture most of the important information from the transcripts."
                    }
                ]
            },
            "ccd7d2e0-41ce-4074-baaa-884180d63e81": {
                "pk": "ccd7d2e0-41ce-4074-baaa-884180d63e81",
                "name": "Jing Zhang",
                "collaborators": [],
                "domain": [
                    "Algebraic Geometry",
                    "Combinatorial Set Theory",
                    "Quantum Information",
                    "Nonlinear Dynamics"
                ],
                "publications": [
                    {
                        "title": "Monochromatic Sumset Without the use of large cardinals",
                        "abstract": "We show in this note that in the forcing extension by $Add(\\omega,\\beth_{\\omega})$, the following Ramsey property holds: for any $r\\in \\omega$ and any $f: \\mathbb{R}\\to r$, there exists an infinite $X\\subset \\mathbb{R}$ such that $X+X$ is monochromatic under $f$. We also show the Ramsey statement above is true in $\\mathrm{ZFC}$ when $r=2$. This answers two questions by Komj\\'ath, Leader, Russell, Shelah, Soukup and Vidny\\'anszky."
                    },
                    {
                        "title": "Reflection principles, GCH and the uniformization properties",
                        "abstract": "Reflection principles (or dually speaking, compactness principles) often give rise to combinatorial guessing principles. Uniformization properties, on the other hand, are examples of anti-guessing principles. We discuss the tension and the compatibility between reflection principles and uniformization properties at the level of the second uncountable cardinal."
                    },
                    {
                        "title": "Hodge Cohomology Criteria For Affine Varieties",
                        "abstract": "We give several new criteria for a quasi-projective variety to be affine. In particular, we prove that an algebraic manifold $Y$ with dimension $n$ is affine if and only if $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$, $i>0$ and $\\kappa(D, X)=n$, i.e., there are $n$ algebraically independent nonconstant regular functions on $Y$, where $X$ is the smooth completion of $Y$, $D$ is the effective boundary divisor with support $X-Y$ and $\\Omega^j_Y$ is the sheaf of regular $j$-forms on $Y$. This proves Mohan Kumar's affineness conjecture for algebraic manifolds and gives a partial answer to J.-P. Serre's Steinness question \\cite{36} in algebraic case since the associated analytic space of an affine variety is Stein [15, Chapter VI, Proposition 3.1]."
                    },
                    {
                        "title": "Cluster States for Continuous-Variable Multipartite Entanglement",
                        "abstract": "We introduce a new class of continuous-variable (CV) multipartite entangled states, the CV cluster states, which might be generated from squeezing and kerr-like interaction. The entanglement properties of these states are studied in terms of classical communication and local operations. The quantum teleportation network with cluster states is investigated. The graph states as the general forms of cluster states are presented, which may be used to generate CV Greenberger-Horne-Zeilinger states by simply local measurements and classical communication. A chain for one-dimensional example of cluster states can be readily experimentally produced only with squeezed light and beamsplitters."
                    },
                    {
                        "title": "Local complementation rule for continuous-variable four-mode unweighted graph states",
                        "abstract": "The local complementation rule is applied for continuous-variable (CV) graph states in the paper, which is an elementary graph transformation rule and successive application of which generates the orbit of any graph states. The corresponding local Gaussian transformations of local complementation for four-mode unweighted graph states were found, which do not mirror the form of the local Clifford unitary of qubit exactly. This work is an important step to characterize the local Gaussian equivalence classes of CV graph states."
                    },
                    {
                        "title": "Graphical rule of transforming continuous-variable graph states by local homodyne detection",
                        "abstract": "Graphical rule, describing that any single-mode homodyne detection turns a given continuous-variable (CV) graph state into a new one, is presented. Employing two simple graphical rules: local complement operation and vertex deletion (single quadrature-amplitude $\\hat{x}$ measurement), the graphical rule for any single-mode quadrature component measurement can be obtained. The shape of CV weighted graph state may be designed and constructed easily from a given larger graph state by applying this graphical rule."
                    },
                    {
                        "title": "Continuous-variable multipartite unlockable bound entangled Gaussian states",
                        "abstract": "Continuous-variable (CV) multipartite unlockable bound-entangled states is investigated in this paper. Comparing with the qubit multipartite unlockable bound-entangled states, CV multipartite unlockable bound-entangled states present the new and different properties. CV multipartite unlockable bound-entangled states may serve as a useful quantum resource for new multiparty communication schemes. The experimental protocol for generating CV unlockable bound-entangled states is proposed with a setup that is at present accessible."
                    },
                    {
                        "title": "A tail cone version of the Halpern-L\u00e4uchli theorem at a large cardinal",
                        "abstract": "The classical Halpern-L\\\"auchli theorem states that for any finite coloring of a finite product of finitely branching perfect trees of height $\\omega$, there exist strong subtrees sharing the same level set such that tuples consisting of elements lying on the same level get the same color. Relative to large cardinals, we establish the consistency of a tail cone version of the Halpern-L\\\"auchli theorem at large cardinal, which, roughly speaking, deals with many colorings simultaneously and diagonally. Among other applications, we generalize a polarized partition relation on rational numbers due to Laver and Galvin to one on linear orders of larger saturation."
                    },
                    {
                        "title": "Threefolds with Vanishing Hodge Cohomology",
                        "abstract": "We consider algebraic manifolds $Y$ of dimension 3 over $\\Bbb{C}$ with $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$ and $i>0$. Let $X$ be a smooth completion of $Y$ with $D=X-Y$, an effective divisor on $X$ with normal crossings. If the $D$-dimension of $X$ is not zero, then $Y$ is a fibre space over a smooth affine curve $C$ (i.e., we have a surjective morphism from $Y$ to $C$ such that general fibre is smooth and irreducible) such that every fibre satisfies the same vanishing condition. If an irreducible smooth fibre is not affine, then the Kodaira dimension of $X$ is $-\\infty$ and the $D$-dimension of X is 1. We also discuss sufficient conditions from the behavior of fibres or higher direct images to guarantee the global vanishing of Hodge cohomology and the affineness of $Y$."
                    },
                    {
                        "title": "There Exist Nontrivial Threefolds with Vanishing Hodge Cohomology",
                        "abstract": "We analyze the structure of the algebraic manifolds $Y$ of dimension 3 with $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$, $i>0$ and $h^0(Y, {\\mathcal{O}}_Y) > 1$, by showing the deformation invariant of some open surfaces. Secondly, we show when a smooth threefold with nonconstant regular functions satisfies the vanishing Hodge cohomology. As an application, we prove the existence of nonaffine and nonproduct threefolds $Y$ with this property by constructing a family of a certain type of open surfaces parametrized by the affine curve $\\C-\\{0\\}$ such that the corresponding smooth completion $X$ has Kodaira dimension $-\\infty$ and $D$-dimension 1, where $D$ is the effective boundary divisor with support $X-Y$."
                    },
                    {
                        "title": "On the $D$-dimension of a certain type of threefolds",
                        "abstract": "Let $Y$ be an algebraic manifold of dimension 3 with $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$, $i>0$ and $h^0(Y, {\\mathcal{O}}_Y) > 1$. Let $X$ be a smooth completion of $Y$ such that the boundary $X-Y$ is the support of an effective divisor $D$ on $X$ with simple normal crossings. We prove that the $D$-dimension of $X$ cannot be 2, i.e., either any two nonconstant regular functions are algebraically dependent or there are three algebraically independent nonconstant regular functions on $Y$. Secondly, if the $D$-dimension of $X$ is greater than 1, then the associated scheme of $Y$ is isomorphic to Spec$\\Gamma(Y, {\\mathcal{O}}_Y)$. Furthermore, we prove that an algebraic manifold $Y$ of any dimension $d\\geq 1$ is affine if and only if $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$, $i>0$ and it is regularly separable, i.e., for any two distinct points $y_1$, $y_2$ on $Y$, there is a regular function $f$ on $Y$ such that $f(y_1)\\neq f(y_2)$."
                    },
                    {
                        "title": "On threefolds without nonconstant regular functions",
                        "abstract": "We consider smooth threefolds $Y$ defined over $\\Bbb{C}$ with $H^i(Y, \\Omega^j_Y)=0$ for all $j\\geq 0$, $i>0$. Let $X$ be a smooth projective threefold containing $Y$ and $D$ be the boundary divisor with support $X-Y$.   We are interested in the following question: What geometry information of $X$ can be obtained from the regular function information on $Y$?   Suppose that the boundary $X-Y$ is a smooth projective surface. In this paper, we analyse two different cases, i.e., there are no nonconstant regular functions on $Y$ or there are lots of regular functions on $Y$. More precisely, if $H^0(Y, {\\mathcal{O}}_Y)=\\Bbb{C}$, we prove that ${1/2}(c_1^2+c_2)\\cdot D=\\chi({\\mathcal{O}}_D)\\geq 0$. In particular, if the line bundle ${\\mathcal{O}}_D(D)$ is not torsion, then $q=h^1(X, {\\mathcal{O}}_X)=0$, ${1/2}(c_1^2+c_2)\\cdot D=\\chi({\\mathcal{O}}_D)=0$,   $\\chi({\\mathcal{O}}_X) >0$ and $K_X$ is not nef. If there is a positive constant $c$ such that $h^0(X, {\\mathcal{O}}_X(nD))\\geq c n^3$ for all sufficiently large $n$ (we say that $D$ is big or the $D$-dimension of $X$ is 3) and $D$ has no exceptional curves, then $|nD|$ is base point free for $n\\gg 0$. Therefore $Y$ is affine if $D$ is big."
                    },
                    {
                        "title": "Algebraic Stein Varieties",
                        "abstract": "It is well-known that the associated analytic space of an affine variety defined over $\\mathbb{C}$ is Stein but the converse is not true, that is, an algebraic Stein variety is not necessarily affine. In this paper, we give sufficient and necessary conditions for an algebraic Stein variety to be affine. One of our results is that an irreducible quasi-projective variety $Y$ defined over $\\mathbb{C}$ with dimension $d$ ($d\\geq 1$) is affine if and only if $Y$ is Stein, $H^i(Y, {\\mathcal{O}}_Y)=0$ for all $i>0$ and $\\kappa(D, X)= d$ (i.e., $D$ is a big divisor), where $X$ is a projective variety containing $Y$ and $D$ is an effective divisor with support $X-Y$. If $Y$ is algebraic Stein but not affine, we also discuss the possible transcendental degree of the nonconstant regular functions on $Y$. We prove that $Y$ cannot have $d-1$ algebraically independent nonconstant regular functions. The interesting phenomenon is that the transcendental degree can be even if the dimension of $Y$ is even and the degree can be odd if the dimension of $Y$ is odd."
                    },
                    {
                        "title": "Affine Algebraic Varieties",
                        "abstract": "In this paper, we give new criteria for affineness of a variety defined over $\\Bbb{C}$. Our main result is that an irreducible algebraic variety $Y$ (may be singular) of dimension $d$ ($d\\geq 1$) defined over $\\Bbb{C}$ is an affine variety if and only if $Y$ contains no complete curves, $H^i(Y, {\\mathcal{O}}_Y)=0$ for all $i>0$ and the boundary $X-Y$ is support of a big divisor, where $X$ is a projective variety containing $Y$. We construct three examples to show that a variety is not affine if it only satisfies two conditions among these three conditions. We also give examples to demonstrate the difference between the behavior of the boundary divisor $D$ and the affineness of $Y$. If $Y$ is an affine variety, then the ring $\\Gamma (Y, {\\mathcal{O}}_Y)$ is noetherian. However, to prove that $Y$ is an affine variety, we do not start from this ring. We explain why we do not need to check the noetherian property of the ring $\\Gamma (Y, {\\mathcal{O}}_Y)$ directly but use the techniques of sheaf and cohomology."
                    },
                    {
                        "title": "Graphical description of local Gaussian operations for continuous-variable weighted graph states",
                        "abstract": "The form of a local Clifford (LC, also called local Gaussian (LG)) operation for the continuous-variable (CV) weighted graph states is presented in this paper, which is the counterpart of the LC operation of local complementation for qubit graph states. The novel property of the CV weighted graph states is shown, which can be expressed by the stabilizer formalism. It is distinctively different from the qubit weighted graph states, which can not be expressed by the stabilizer formalism. The corresponding graph rule, stated in purely graph theoretical terms, is described, which completely characterizes the evolution of CV weighted graph states under this LC operation. This LC operation may be applied repeatedly on a CV weighted graph state, which can generate the infinite LC equivalent graph states of this graph state. This work is an important step to characterize the LC equivalence class of CV weighted graph states."
                    },
                    {
                        "title": "Bertini Type Theorems",
                        "abstract": "Let $X$ be a smooth irreducible projective variety of dimension at least 2 over an algebraically closed field of characteristic 0 in the projective space ${\\mathbb{P}}^n$.   Bertini's Theorem states that a general hyperplane $H$ intersects $X$ with an irreducible smooth subvariety of $X$. However, the precise location of the smooth hyperplane section is not known. We show that for any $q\\leq n+1$ closed points in general position and any degree $a>1$, a general hypersurface $H$ of degree $a$ passing through these $q$ points intersects $X$ with an irreducible smooth codimension 1 subvariety on $X$. We also consider linear system of ample divisors and give precise location of smooth elements in the system. Similar result can be obtained for compact complex manifolds with holomorphic maps into projective spaces."
                    },
                    {
                        "title": "A Literature Survey of Cooperative Caching in Content Distribution Networks",
                        "abstract": "Content distribution networks (CDNs) which serve to deliver web objects (e.g., documents, applications, music and video, etc.) have seen tremendous growth since its emergence. To minimize the retrieving delay experienced by a user with a request for a web object, caching strategies are often applied - contents are replicated at edges of the network which is closer to the user such that the network distance between the user and the object is reduced. In this literature survey, evolution of caching is studied. A recent research paper [15] in the field of large-scale caching for CDN was chosen to be the anchor paper which serves as a guide to the topic. Research studies after and relevant to the anchor paper are also analyzed to better evaluate the statements and results of the anchor paper and more importantly, to obtain an unbiased view of the large scale collaborate caching systems as a whole."
                    },
                    {
                        "title": "Rado's conjecture and its Baire version",
                        "abstract": "Rado's Conjecture is a compactness/reflection principle that says any nonspecial tree of height $\\omega_1$ has a nonspecial subtree of size $\\leq \\aleph_1$. Though incompatible with Martin's Axiom, Rado's Conjecture turns out to have many interesting consequences that are consequences of forcing axioms. In this paper, we obtain consistency results concerning Rado's Conjecture and its Baire version. In particular, we show a fragment of PFA, that is the forcing axiom for \\emph{Baire Indestructibly proper forcings}, is compatible with the Baire Rado's Conjecture. As a corollary, Baire Rado's Conjecture does not imply Rado's Conjecture. Then we discuss the strength and limitations of the Baire Rado's Conjecture regarding its interaction with simultaneous stationary reflection and some families of weak square principles. Finally we investigate the influence of the Rado's Conjecture on some polarized partition relations."
                    },
                    {
                        "title": "Some remarks on uncountable rainbow Ramsey theory",
                        "abstract": "We discuss the rainbow Ramsey theorems at limit cardinals and successors of singular cardinals, addressing some questions in \\cite{MR2354904} and \\cite{MR2902230}. In particular, we show for inaccessible $\\kappa$, $\\kappa\\to^{poly}(\\kappa)^2_{2-bdd}$ does not characterize weak compactness and for singular $\\kappa$, $\\mathrm{GCH}+\\square_\\kappa$ implies $\\kappa^+\\not\\to^{poly} (\\eta)^2_{<\\kappa-bdd}$ for any $\\eta\\geq cf(\\kappa)^+$ and $\\kappa^+\\to^{poly} (\\nu)^2_{<\\kappa-bdd}$ for any $\\nu<cf(\\kappa)^+$. We also provide a simplified construction of a model for $\\omega_2\\not\\to^{poly} (\\omega_1)^2_{2-bdd}$ originally constructed in \\cite{MR2902230} and show the witnessing coloring is indestructible under strongly proper forcings but destructible under some c.c.c forcing. Finally, we conclude with some remarks and questions on possible generalizations to rainbow partition relations for triples."
                    },
                    {
                        "title": "Almost global solutions to two classes of 1-d Hamiltonian Derivative Nonlinear Schr\u00f6dinger equations",
                        "abstract": "Consider two kinds of 1-d Hamiltonian Derivative Nonlinear Schr\\\"odinger (DNLS) equations with respect to different symplectic forms under periodic boundary conditions. The nonlinearities of these equations depend not only on $(x,\\psi,\\bar{\\psi})$ but also on $(\\psi_x,\\bar{\\psi}_x)$, which means the nonlinearities of these equations are unbounded. Suppose that the nonlinearities depend on the space-variable $x$ periodically.   Under some assumptions, for most potentials of these two kinds of Hamiltonian DNLS equations, if the initial value is smaller than $\\varepsilon\\ll1$ in $p$-Sobolev norm, then the corresponding solution to these equations is also smaller than $2\\varepsilon$ during a time interval $(-c\\varepsilon^{-r_*},c\\varepsilon^{-r_*})$(for any given positive $r_*$). The main methods are constructing Birkhoff normal forms to two kinds of Hamiltonian systems which have unbounded nonlinearities and using the special symmetry of the unbounded nonlinearities of Hamiltonian functions to obtain a long time estimate of the solution in $p$-Sobolev norm."
                    }
                ]
            },
            "200bb126-6c9f-4fa5-b2ba-f134c2a9fb6e": {
                "pk": "200bb126-6c9f-4fa5-b2ba-f134c2a9fb6e",
                "name": "Jifan Yu",
                "collaborators": [
                    "Juanzi Li",
                    "Lei Hou",
                    "Zijun Yao",
                    "Jie Tang",
                    "Zheyuan Zhang",
                    "Zhiyuan Liu",
                    "Jing Zhang",
                    "Xiaokang Zhang",
                    "Shangqing Tu",
                    "Xin Lv"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Knowledge Graph",
                    "Machine Learning",
                    "Educational Technology"
                ],
                "publications": [
                    {
                        "title": "Exploring the Cognitive Knowledge Structure of Large Language Models: An Educational Diagnostic Assessment Approach",
                        "abstract": "Large Language Models (LLMs) have not only exhibited exceptional performance across various tasks, but also demonstrated sparks of intelligence. Recent studies have focused on assessing their capabilities on human exams and revealed their impressive competence in different domains. However, cognitive research on the overall knowledge structure of LLMs is still lacking. In this paper, based on educational diagnostic assessment method, we conduct an evaluation using MoocRadar, a meticulously annotated human test dataset based on Bloom Taxonomy. We aim to reveal the knowledge structures of LLMs and gain insights of their cognitive capabilities. This research emphasizes the significance of investigating LLMs' knowledge and understanding the disparate cognitive patterns of LLMs. By shedding light on models' knowledge, researchers can advance development and utilization of LLMs in a more informed and effective manner."
                    },
                    {
                        "title": "Course Concept Expansion in MOOCs with External Knowledge and Interactive Game",
                        "abstract": "As Massive Open Online Courses (MOOCs) become increasingly popular, it is promising to automatically provide extracurricular knowledge for MOOC users. Suffering from semantic drifts and lack of knowledge guidance, existing methods can not effectively expand course concepts in complex MOOC environments. In this paper, we first build a novel boundary during searching for new concepts via external knowledge base and then utilize heterogeneous features to verify the high-quality results. In addition, to involve human efforts in our model, we design an interactive optimization mechanism based on a game. Our experiments on the four datasets from Coursera and XuetangX show that the proposed method achieves significant improvements(+0.19 by MAP) over existing methods. The source code and datasets have been published."
                    },
                    {
                        "title": "Subgraph Retrieval Enhanced Model for Multi-hop Knowledge Base Question Answering",
                        "abstract": "Recent works on knowledge base question answering (KBQA) retrieve subgraphs for easier reasoning. A desired subgraph is crucial as a small one may exclude the answer but a large one might introduce more noises. However, the existing retrieval is either heuristic or interwoven with the reasoning, causing reasoning on the partial subgraphs, which increases the reasoning bias when the intermediate supervision is missing. This paper proposes a trainable subgraph retriever (SR) decoupled from the subsequent reasoning process, which enables a plug-and-play framework to enhance any subgraph-oriented KBQA model. Extensive experiments demonstrate SR achieves significantly better retrieval and QA performance than existing retrieval methods. Via weakly supervised pre-training as well as the end-to-end fine-tuning, SRl achieves new state-of-the-art performance when combined with NSM, a subgraph-oriented reasoner, for embedding-based KBQA methods."
                    },
                    {
                        "title": "ChatLog: Carefully Evaluating the Evolution of ChatGPT Across Time",
                        "abstract": "ChatGPT has achieved great success and can be considered to have acquired an infrastructural status. There are abundant works for evaluating ChatGPT on benchmarks. However, existing benchmarks encounter two challenges: (1) Disregard for periodical evaluation and (2) Lack of fine-grained features. In this paper, we construct ChatLog, an ever-updating dataset with large-scale records of diverse long-form ChatGPT responses for 21 NLP benchmarks from March, 2023 to now. We conduct a comprehensive performance evaluation to find that most capabilities of ChatGPT improve over time except for some abilities, and there exists a step-wise evolving pattern of ChatGPT. We further analyze the inherent characteristics of ChatGPT by extracting the knowledge and linguistic features. We find some stable features that stay unchanged and apply them on the detection of ChatGPT-generated texts to improve the robustness of cross-version detection. We will continuously maintain our project at \\url{https://github.com/THU-KEG/ChatLog/}."
                    },
                    {
                        "title": "Learn to Not Link: Exploring NIL Prediction in Entity Linking",
                        "abstract": "Entity linking models have achieved significant success via utilizing pretrained language models to capture semantic features. However, the NIL prediction problem, which aims to identify mentions without a corresponding entity in the knowledge base, has received insufficient attention. We categorize mentions linking to NIL into Missing Entity and Non-Entity Phrase, and propose an entity linking dataset NEL that focuses on the NIL prediction problem. NEL takes ambiguous entities as seeds, collects relevant mention context in the Wikipedia corpus, and ensures the presence of mentions linking to NIL by human annotation and entity masking. We conduct a series of experiments with the widely used bi-encoder and cross-encoder entity linking models, results show that both types of NIL mentions in training data have a significant influence on the accuracy of NIL prediction. Our code and dataset can be accessed at https://github.com/solitaryzero/NIL_EL"
                    },
                    {
                        "title": "PST-Bench: Tracing and Benchmarking the Source of Publications",
                        "abstract": "Tracing the source of research papers is a fundamental yet challenging task for researchers. The billion-scale citation relations between papers hinder researchers from understanding the evolution of science efficiently. To date, there is still a lack of an accurate and scalable dataset constructed by professional researchers to identify the direct source of their studied papers, based on which automatic algorithms can be developed to expand the evolutionary knowledge of science. In this paper, we study the problem of paper source tracing (PST) and construct a high-quality and ever-increasing dataset PST-Bench in computer science. Based on PST-Bench, we reveal several intriguing discoveries, such as the differing evolution patterns across various topics. An exploration of various methods underscores the hardness of PST-Bench, pinpointing potential directions on this topic. The dataset and codes have been available at https://github.com/THUDM/paper-source-trace."
                    },
                    {
                        "title": "Transferable and Efficient Non-Factual Content Detection via Probe Training with Offline Consistency Checking",
                        "abstract": "Detecting non-factual content is a longstanding goal to increase the trustworthiness of large language models (LLMs) generations. Current factuality probes, trained using humanannotated labels, exhibit limited transferability to out-of-distribution content, while online selfconsistency checking imposes extensive computation burden due to the necessity of generating multiple outputs. This paper proposes PINOSE, which trains a probing model on offline self-consistency checking results, thereby circumventing the need for human-annotated data and achieving transferability across diverse data distributions. As the consistency check process is offline, PINOSE reduces the computational burden of generating multiple responses by online consistency verification. Additionally, it examines various aspects of internal states prior to response decoding, contributing to more effective detection of factual inaccuracies. Experiment results on both factuality detection and question answering benchmarks show that PINOSE achieves surpassing results than existing factuality detection methods. Our code and datasets are publicly available on this anonymized repository."
                    },
                    {
                        "title": "KoRC: Knowledge oriented Reading Comprehension Benchmark for Deep Text Understanding",
                        "abstract": "Deep text understanding, which requires the connections between a given document and prior knowledge beyond its text, has been highlighted by many benchmarks in recent years. However, these benchmarks have encountered two major limitations. On the one hand, most of them require human annotation of knowledge, which leads to limited knowledge coverage. On the other hand, they usually use choices or spans in the texts as the answers, which results in narrow answer space. To overcome these limitations, we build a new challenging benchmark named KoRc in this paper. Compared with previous benchmarks, KoRC has two advantages, i.e., broad knowledge coverage and flexible answer format. Specifically, we utilize massive knowledge bases to guide annotators or large language models (LLMs) to construct knowledgable questions. Moreover, we use labels in knowledge bases rather than spans or choices as the final answers. We test state-of-the-art models on KoRC and the experimental results show that the strongest baseline only achieves 68.3% and 30.0% F1 measure in the in-distribution and out-of-distribution test set, respectively. These results indicate that deep text understanding is still an unsolved challenge. The benchmark dataset, leaderboard, and baseline methods are released in https://github.com/THU-KEG/KoRC."
                    },
                    {
                        "title": "VidCoM: Fast Video Comprehension through Large Language Models with Multimodal Tools",
                        "abstract": "Building models that comprehends videos and responds specific user instructions is a practical and challenging topic, as it requires mastery of both vision understanding and knowledge reasoning. Compared to language and image modalities, training efficiency remains a serious problem as existing studies train models on massive sparse videos paired with brief descriptions. In this paper, we introduce \\textbf{VidCoM}, a fast adaptive framework that leverages Large Language Models (LLMs) to reason about videos using lightweight visual tools. Specifically, we reveal that the key to responding to specific instructions is focusing on relevant video events, and utilize two visual tools, structured scene graph generation and descriptive image caption generation, to gather and represent the event information. Thus, a LLM enriched with world knowledge is adopted as the reasoning agent to achieve the responses by performing multiple reasoning steps on specific video events. To address the difficulty of LLMs identifying video events, we further propose an Instruction-oriented Video Events Recognition (InsOVER) algorithm. This algorithm locates the corresponding video events based on an efficient Hungarian matching between decompositions of linguistic instructions and video events, thereby enabling LLMs to interact effectively with extended videos. Extensive experiments on two typical video comprehension tasks show that the proposed tuning-free framework outperforms the pre-trained models including Flamingo-80B, to achieve the state-of-the-art performance. Our source code and system will be publicly available."
                    },
                    {
                        "title": "WaterBench: Towards Holistic Evaluation of Watermarks for Large Language Models",
                        "abstract": "To mitigate the potential misuse of large language models (LLMs), recent research has developed watermarking algorithms, which restrict the generation process to leave an invisible trace for watermark detection. Due to the two-stage nature of the task, most studies evaluate the generation and detection separately, thereby presenting a challenge in unbiased, thorough, and applicable evaluations. In this paper, we introduce WaterBench, the first comprehensive benchmark for LLM watermarks, in which we design three crucial factors: (1) For benchmarking procedure, to ensure an apples-to-apples comparison, we first adjust each watermarking method's hyper-parameter to reach the same watermarking strength, then jointly evaluate their generation and detection performance. (2) For task selection, we diversify the input and output length to form a five-category taxonomy, covering $9$ tasks. (3) For evaluation metric, we adopt the GPT4-Judge for automatically evaluating the decline of instruction-following abilities after watermarking. We evaluate $4$ open-source watermarks on $2$ LLMs under $2$ watermarking strengths and observe the common struggles for current methods on maintaining the generation quality. The code and data are available at https://github.com/THU-KEG/WaterBench."
                    },
                    {
                        "title": "Evaluating Generative Language Models in Information Extraction as Subjective Question Correction",
                        "abstract": "Modern Large Language Models (LLMs) have showcased remarkable prowess in various tasks necessitating sophisticated cognitive behaviors. Nevertheless, a paradoxical performance discrepancy is observed, where these models underperform in seemingly elementary tasks like relation extraction and event extraction due to two issues in conventional evaluation. (1) The imprecision of existing evaluation metrics that struggle to effectively gauge semantic consistency between model outputs and ground truth, and (2) The inherent incompleteness of evaluation benchmarks, primarily due to restrictive human annotation schemas, resulting in underestimated LLM performances. Inspired by the principles in subjective question correction, we propose a new evaluation method, SQC-Score. This method innovatively utilizes LLMs, fine-tuned through subjective question correction data, to refine matching between model outputs and golden labels. Additionally, by incorporating a Natural Language Inference (NLI) model, SQC-Score enriches golden labels, addressing benchmark incompleteness by acknowledging correct yet previously omitted answers. Results on three information extraction tasks show that SQC-Score is more preferred by human annotators than the baseline metrics. Utilizing SQC-Score, we conduct a comprehensive evaluation of the state-of-the-art LLMs and provide insights for future research for information extraction. Dataset and associated codes can be accessed at https://github.com/THU-KEG/SQC-Score."
                    },
                    {
                        "title": "Interpretable and Low-Resource Entity Matching via Decoupling Feature Learning from Decision Making",
                        "abstract": "Entity Matching (EM) aims at recognizing entity records that denote the same real-world object. Neural EM models learn vector representation of entity descriptions and match entities end-to-end. Though robust, these methods require many resources for training, and lack of interpretability. In this paper, we propose a novel EM framework that consists of Heterogeneous Information Fusion (HIF) and Key Attribute Tree (KAT) Induction to decouple feature representation from matching decision. Using self-supervised learning and mask mechanism in pre-trained language modeling, HIF learns the embeddings of noisy attribute values by inter-attribute attention with unlabeled data. Using a set of comparison features and a limited amount of annotated data, KAT Induction learns an efficient decision tree that can be interpreted by generating entity matching rules whose structure is advocated by domain experts. Experiments on 6 public datasets and 3 industrial datasets show that our method is highly efficient and outperforms SOTA EM models in most cases. Our codes and datasets can be obtained from https://github.com/THU-KEG/HIF-KAT."
                    },
                    {
                        "title": "ConstGCN: Constrained Transmission-based Graph Convolutional Networks for Document-level Relation Extraction",
                        "abstract": "Document-level relation extraction with graph neural networks faces a fundamental graph construction gap between training and inference - the golden graph structure only available during training, which causes that most methods adopt heuristic or syntactic rules to construct a prior graph as a pseudo proxy. In this paper, we propose $\\textbf{ConstGCN}$, a novel graph convolutional network which performs knowledge-based information propagation between entities along with all specific relation spaces without any prior graph construction. Specifically, it updates the entity representation by aggregating information from all other entities along with each relation space, thus modeling the relation-aware spatial information. To control the information flow passing through the indeterminate relation spaces, we propose to constrain the propagation using transmitting scores learned from the Noise Contrastive Estimation between fact triples. Experimental results show that our method outperforms the previous state-of-the-art (SOTA) approaches on the DocRE dataset."
                    },
                    {
                        "title": "Mastering the Task of Open Information Extraction with Large Language Models and Consistent Reasoning Environment",
                        "abstract": "Open Information Extraction (OIE) aims to extract objective structured knowledge from natural texts, which has attracted growing attention to build dedicated models with human experience. As the large language models (LLMs) have exhibited remarkable in-context learning capabilities, a question arises as to whether the task of OIE can be effectively tackled with this paradigm? In this paper, we explore solving the OIE problem by constructing an appropriate reasoning environment for LLMs. Specifically, we first propose a method to effectively estimate the discrepancy of syntactic distribution between a LLM and test samples, which can serve as correlation evidence for preparing positive demonstrations. Upon the evidence, we introduce a simple yet effective mechanism to establish the reasoning environment for LLMs on specific tasks. Without bells and whistles, experimental results on the standard CaRB benchmark demonstrate that our $6$-shot approach outperforms state-of-the-art supervised method, achieving an $55.3$ $F_1$ score. Further experiments on TACRED and ACE05 show that our method can naturally generalize to other information extraction tasks, resulting in improvements of $5.7$ and $6.8$ $F_1$ scores, respectively."
                    },
                    {
                        "title": "Reverse That Number! Decoding Order Matters in Arithmetic Learning",
                        "abstract": "Recent advancements in pretraining have demonstrated that modern Large Language Models (LLMs) possess the capability to effectively learn arithmetic operations. However, despite acknowledging the significance of digit order in arithmetic computation, current methodologies predominantly rely on sequential, step-by-step approaches for teaching LLMs arithmetic, resulting in a conclusion where obtaining better performance involves fine-grained step-by-step. Diverging from this conventional path, our work introduces a novel strategy that not only reevaluates the digit order by prioritizing output from the least significant digit but also incorporates a step-by-step methodology to substantially reduce complexity. We have developed and applied this method in a comprehensive set of experiments. Compared to the previous state-of-the-art (SOTA) method, our findings reveal an overall improvement of in accuracy while requiring only a third of the tokens typically used during training. For the purpose of facilitating replication and further research, we have made our code and dataset publicly available at \\url{https://anonymous.4open.science/r/RAIT-9FB7/}."
                    },
                    {
                        "title": "Explainable Few-shot Knowledge Tracing",
                        "abstract": "Knowledge tracing (KT), aiming to mine students' mastery of knowledge by their exercise records and predict their performance on future test questions, is a critical task in educational assessment. While researchers achieved tremendous success with the rapid development of deep learning techniques, current knowledge tracing tasks fall into the cracks from real-world teaching scenarios. Relying heavily on extensive student data and solely predicting numerical performances differs from the settings where teachers assess students' knowledge state from limited practices and provide explanatory feedback. To fill this gap, we explore a new task formulation: Explainable Few-shot Knowledge Tracing. By leveraging the powerful reasoning and generation abilities of large language models (LLMs), we then propose a cognition-guided framework that can track the student knowledge from a few student records while providing natural language explanations. Experimental results from three widely used datasets show that LLMs can perform comparable or superior to competitive deep knowledge tracing methods. We also discuss potential directions and call for future improvements in relevant topics."
                    },
                    {
                        "title": "Knowledge-to-Jailbreak: One Knowledge Point Worth One Attack",
                        "abstract": "Large language models (LLMs) have been increasingly applied to various domains, which triggers increasing concerns about LLMs' safety on specialized domains, e.g. medicine. However, testing the domain-specific safety of LLMs is challenging due to the lack of domain knowledge-driven attacks in existing benchmarks. To bridge this gap, we propose a new task, knowledge-to-jailbreak, which aims to generate jailbreaks from domain knowledge to evaluate the safety of LLMs when applied to those domains. We collect a large-scale dataset with 12,974 knowledge-jailbreak pairs and fine-tune a large language model as jailbreak-generator, to produce domain knowledge-specific jailbreaks. Experiments on 13 domains and 8 target LLMs demonstrate the effectiveness of jailbreak-generator in generating jailbreaks that are both relevant to the given knowledge and harmful to the target LLMs. We also apply our method to an out-of-domain knowledge base, showing that jailbreak-generator can generate jailbreaks that are comparable in harmfulness to those crafted by human experts. Data and code: https://github.com/THU-KEG/Knowledge-to-Jailbreak/."
                    },
                    {
                        "title": "Simulating Classroom Education with LLM-Empowered Agents",
                        "abstract": "Large language models (LLMs) have been employed in various intelligent educational tasks to assist teaching. While preliminary explorations have focused on independent LLM-empowered agents for specific educational tasks, the potential for LLMs within a multi-agent collaborative framework to simulate a classroom with real user participation remains unexplored. In this work, we propose SimClass, a multi-agent classroom simulation framework involving user participation. We recognize representative class roles and introduce a novel class control mechanism for automatic classroom teaching, and conduct user experiments in two real-world courses. Utilizing the Flanders Interactive Analysis System and Community of Inquiry theoretical frame works from educational analysis, we demonstrate that LLMs can simulate traditional classroom interaction patterns effectively while enhancing user's experience. We also observe emergent group behaviors among agents in SimClass, where agents collaborate to create enlivening interactions in classrooms to improve user learning process. We hope this work pioneers the application of LLM-empowered multi-agent systems in virtual classroom teaching."
                    },
                    {
                        "title": "Program Transfer for Answering Complex Questions over Knowledge Bases",
                        "abstract": "Program induction for answering complex questions over knowledge bases (KBs) aims to decompose a question into a multi-step program, whose execution against the KB produces the final answer. Learning to induce programs relies on a large number of parallel question-program pairs for the given KB. However, for most KBs, the gold program annotations are usually lacking, making learning difficult. In this paper, we propose the approach of program transfer, which aims to leverage the valuable program annotations on the rich-resourced KBs as external supervision signals to aid program induction for the low-resourced KBs that lack program annotations. For program transfer, we design a novel two-stage parsing framework with an efficient ontology-guided pruning strategy. First, a sketch parser translates the question into a high-level program sketch, which is the composition of functions. Second, given the question and sketch, an argument parser searches the detailed arguments from the KB for functions. During the searching, we incorporate the KB ontology to prune the search space. The experiments on ComplexWebQuestions and WebQuestionSP show that our method outperforms SOTA methods significantly, demonstrating the effectiveness of program transfer and our framework. Our codes and datasets can be obtained from https://github.com/THU-KEG/ProgramTransfer."
                    }
                ]
            },
            "a7b2a563-2829-4d24-80dc-45241a19ddfe": {
                "pk": "a7b2a563-2829-4d24-80dc-45241a19ddfe",
                "name": "Xiaokang Zhang",
                "collaborators": [
                    "Inge Jonassen",
                    "Jake Minten",
                    "Bhargav Rallabandi",
                    "Xianping Ma",
                    "Man-On Pun",
                    "Xianxu Hou",
                    "Linlin Shen",
                    "Zhihui Lai",
                    "Jun Wan"
                ],
                "domain": [
                    "Ensemble Learning",
                    "Feature Selection",
                    "Semantic Segmentation",
                    "Generative Adversarial Networks"
                ],
                "publications": [
                    {
                        "title": "EFSIS: Ensemble Feature Selection Integrating Stability",
                        "abstract": "Ensemble learning that can be used to combine the predictions from multiple learners has been widely applied in pattern recognition, and has been reported to be more robust and accurate than the individual learners. This ensemble logic has recently also been more applied in feature selection. There are basically two strategies for ensemble feature selection, namely data perturbation and function perturbation. Data perturbation performs feature selection on data subsets sampled from the original dataset and then selects the features consistently ranked highly across those data subsets. This has been found to improve both the stability of the selector and the prediction accuracy for a classifier. Function perturbation frees the user from having to decide on the most appropriate selector for any given situation and works by aggregating multiple selectors. This has been found to maintain or improve classification performance. Here we propose a framework, EFSIS, combining these two strategies. Empirical results indicate that EFSIS gives both high prediction accuracy and stability."
                    },
                    {
                        "title": "A Comparative Analysis of Feature Selection Methods for Biomarker Discovery in Study of Toxicant-treated Atlantic Cod (Gadus morhua) Liver",
                        "abstract": "Univariate and multivariate feature selection methods can be used for biomarker discovery in analysis of toxicant exposure. Among the univariate methods, differential expression analysis (DEA) is often applied for its simplicity and interpretability. A characteristic of methods for DEA is that they treat genes individually, disregarding the correlation that exists between them. On the other hand, some multivariate feature selection methods are proposed for biomarker discovery. Provided with various biomarker discovery methods, how to choose the most suitable method for a specific dataset becomes a problem. In this paper, we present a framework for comparison of potential biomarker discovery methods: three methods that stem from different theories are compared by how stable they are and how well they can improve the classification accuracy. The three methods we have considered are: Significance Analysis of Microarrays (SAM) which identifies the differentially expressed genes; minimum Redundancy Maximum Relevance (mRMR) based on information theory; and Characteristic Direction (GeoDE) inspired by a graphical perspective. Tested on the gene expression data from two experiments exposing the cod fish to two different toxicants (MeHg and PCB 153), different methods stand out in different cases, so a decision upon the most suitable method should be made based on the dataset under study and the research interest."
                    },
                    {
                        "title": "Time-averaged Dynamics of Compressible Particles in Oscillatory Gradient Flows",
                        "abstract": "Acoustic fields effect steady transport of suspended particles by rectifying the inertia of primary oscillations. We develop a fully analytic theory that relates this steady particle motion to incident oscillatory (acoustic) flow and the time-averaged force acting on the particle, systematically spanning the entire range between inviscid acoustofluidics and viscous particle hydrodynamics. By applying the Lorentz reciprocal theorem, we obtain a Fax\\'{e}n-like relationship that includes nonlinear inertial forces, which depend on (i) the thickness of the oscillatory Stokes layer around the particle, and (ii) the density and compressibility contrast between the particle and the fluid. The framework recovers secondary radiation forces for thin Stokes layers, and predicts a reversal of the motion when the thickness of the Stokes layer is comparable to the particle size. We quantitatively validate the theory using numerical simulations of the timescale-separated hydrodynamics."
                    },
                    {
                        "title": "RS3Mamba: Visual State Space Model for Remote Sensing Images Semantic Segmentation",
                        "abstract": "Semantic segmentation of remote sensing images is a fundamental task in geoscience research. However, there are some significant shortcomings for the widely used convolutional neural networks (CNNs) and Transformers. The former is limited by its insufficient long-range modeling capabilities, while the latter is hampered by its computational complexity. Recently, a novel visual state space (VSS) model represented by Mamba has emerged, capable of modeling long-range relationships with linear computability. In this work, we propose a novel dual-branch network named remote sensing images semantic segmentation Mamba (RS3Mamba) to incorporate this innovative technology into remote sensing tasks. Specifically, RS3Mamba utilizes VSS blocks to construct an auxiliary branch, providing additional global information to convolution-based main branch. Moreover, considering the distinct characteristics of the two branches, we introduce a collaborative completion module (CCM) to enhance and fuse features from the dual-encoder. Experimental results on two widely used datasets, ISPRS Vaihingen and LoveDA Urban, demonstrate the effectiveness and potential of the proposed RS3Mamba. To the best of our knowledge, this is the first vision Mamba specifically designed for remote sensing images semantic segmentation. The source code will be made available at https://github.com/sstary/SSRS."
                    },
                    {
                        "title": "GuidedStyle: Attribute Knowledge Guided Style Manipulation for Semantic Face Editing",
                        "abstract": "Although significant progress has been made in synthesizing high-quality and visually realistic face images by unconditional Generative Adversarial Networks (GANs), there still lacks of control over the generation process in order to achieve semantic face editing. In addition, it remains very challenging to maintain other face information untouched while editing the target attributes. In this paper, we propose a novel learning framework, called GuidedStyle, to achieve semantic face editing on StyleGAN by guiding the image generation process with a knowledge network. Furthermore, we allow an attention mechanism in StyleGAN generator to adaptively select a single layer for style manipulation. As a result, our method is able to perform disentangled and controllable edits along various attributes, including smiling, eyeglasses, gender, mustache and hair color. Both qualitative and quantitative results demonstrate the superiority of our method over other competing methods for semantic face editing. Moreover, we show that our model can be also applied to different types of real and artistic face editing, demonstrating strong generalization ability."
                    }
                ]
            },
            "34750090-21fa-422b-a98b-3ce8d602eb24": {
                "pk": "34750090-21fa-422b-a98b-3ce8d602eb24",
                "name": "Xiaohan Zhang",
                "collaborators": [
                    "Shiqi Zhang",
                    "Yan Ding",
                    "Wei Cai",
                    "Jivko Sinapov",
                    "Xingyue Zhan",
                    "Waqas Sultani",
                    "Safwan Wshah",
                    "Chris Paxton",
                    "Yukui Qiu",
                    "Zhenyu Sun"
                ],
                "domain": [
                    "Deep Learning",
                    "Reinforcement Learning",
                    "Robotics",
                    "Aerial Rendering"
                ],
                "publications": [
                    {
                        "title": "Actor-Critic Algorithm for High-dimensional Partial Differential Equations",
                        "abstract": "We develop a deep learning model to effectively solve high-dimensional nonlinear parabolic partial differential equations (PDE). We follow Feynman-Kac formula to reformulate PDE into the equivalent stochastic control problem governed by a Backward Stochastic Differential Equation (BSDE) system. The Markovian property of the BSDE is utilized in designing our neural network architecture, which is inspired by the Actor-Critic algorithm usually applied for deep Reinforcement Learning. Compared to the State-of-the-Art model, we make several improvements including 1) largely reduced trainable parameters, 2) faster convergence rate and 3) fewer hyperparameters to tune. We demonstrate those improvements by solving a few well-known classes of PDEs such as Hamilton-Jacobian-Bellman equation, Allen-Cahn equation and Black-Scholes equation with dimensions on the order of 100."
                    },
                    {
                        "title": "Critical Strain for Surface Nucleation of Dislocations in Silicon",
                        "abstract": "A long-standing discrepancy exists between experiments and atomistic models concerning the critical strain needed for surface nucleation of dislocations in silicon-germanium systems. While dislocation nucleation is readily observed in hetero-epitaxial thin films with misfit strains less than 4%, existing atomistic models predict that a critical strain over 7.8% is needed to overcome the kinetic barrier for dislocation nucleation. Using zero-temperature energy barrier calculations and finite-temperature Molecular Dynamics simulations, we show that 3-dimensional surface features such as a sharply bent step can lower the predicted critical nucleation strain of a shuffle-set dislocation to 6.4%, and that of a shuffle-glide dislocation complex to 5.3%. Consistent findings are obtained using both the Stillinger-Weber (SW) and modified embedded-atom method (MEAM) potentials, providing support to the physical relevance of the shuffle-glide dislocation complex, which was previously considered as an artifact of the SW potential."
                    },
                    {
                        "title": "Planning Multimodal Exploratory Actions for Online Robot Attribute Learning",
                        "abstract": "Robots frequently need to perceive object attributes, such as \"red,\" \"heavy,\" and \"empty,\" using multimodal exploratory actions, such as \"look,\" \"lift,\" and \"shake.\" Robot attribute learning algorithms aim to learn an observation model for each perceivable attribute given an exploratory action. Once the attribute models are learned, they can be used to identify attributes of new objects, answering questions, such as \"Is this object red and empty?\" Attribute learning and identification are being treated as two separate problems in the literature. In this paper, we first define a new problem called online robot attribute learning (On-RAL), where the robot works on attribute learning and attribute identification simultaneously. Then we develop an algorithm called information-theoretic reward shaping (ITRS) that actively addresses the trade-off between exploration and exploitation in On-RAL problems. ITRS was compared with competitive robot attribute learning baselines, and experimental results demonstrate ITRS' superiority in learning efficiency and identification accuracy."
                    },
                    {
                        "title": "Task-Motion Planning for Safe and Efficient Urban Driving",
                        "abstract": "Autonomous vehicles need to plan at the task level to compute a sequence of symbolic actions, such as merging left and turning right, to fulfill people's service requests, where efficiency is the main concern. At the same time, the vehicles must compute continuous trajectories to perform actions at the motion level, where safety is the most important. Task-motion planning in autonomous driving faces the problem of maximizing task-level efficiency while ensuring motion-level safety. To this end, we develop algorithm Task-Motion Planning for Urban Driving (TMPUD) that, for the first time, enables the task and motion planners to communicate about the safety level of driving behaviors. TMPUD has been evaluated using a realistic urban driving simulation platform. Results suggest that TMPUD performs significantly better than competitive baselines from the literature in efficiency, while ensuring the safety of driving behaviors."
                    },
                    {
                        "title": "Cross-View Image Sequence Geo-localization",
                        "abstract": "Cross-view geo-localization aims to estimate the GPS location of a query ground-view image by matching it to images from a reference database of geo-tagged aerial images. To address this challenging problem, recent approaches use panoramic ground-view images to increase the range of visibility. Although appealing, panoramic images are not readily available compared to the videos of limited Field-Of-View (FOV) images. In this paper, we present the first cross-view geo-localization method that works on a sequence of limited FOV images. Our model is trained end-to-end to capture the temporal structure that lies within the frames using the attention-based temporal feature aggregation module. To robustly tackle different sequences length and GPS noises during inference, we propose to use a sequential dropout scheme to simulate variant length sequences. To evaluate the proposed approach in realistic settings, we present a new large-scale dataset containing ground-view sequences along with the corresponding aerial-view images. Extensive experiments and comparisons demonstrate the superiority of the proposed approach compared to several competitive baselines."
                    },
                    {
                        "title": "Task and Motion Planning with Large Language Models for Object Rearrangement",
                        "abstract": "Multi-object rearrangement is a crucial skill for service robots, and commonsense reasoning is frequently needed in this process. However, achieving commonsense arrangements requires knowledge about objects, which is hard to transfer to robots. Large language models (LLMs) are one potential source of this knowledge, but they do not naively capture information about plausible physical arrangements of the world. We propose LLM-GROP, which uses prompting to extract commonsense knowledge about semantically valid object configurations from an LLM and instantiates them with a task and motion planner in order to generalize to varying scene geometry. LLM-GROP allows us to go from natural-language commands to human-aligned object rearrangement in varied environments. Based on human evaluations, our approach achieves the highest rating while outperforming competitive baselines in terms of success rate while maintaining comparable cumulative action costs. Finally, we demonstrate a practical implementation of LLM-GROP on a mobile manipulator in real-world scenarios. Supplementary materials are available at: https://sites.google.com/view/llm-grop"
                    },
                    {
                        "title": "Aerial-NeRF: Adaptive Spatial Partitioning and Sampling for Large-Scale Aerial Rendering",
                        "abstract": "Recent progress in large-scale scene rendering has yielded Neural Radiance Fields (NeRF)-based models with an impressive ability to synthesize scenes across small objects and indoor scenes. Nevertheless, extending this idea to large-scale aerial rendering poses two critical problems. Firstly, a single NeRF cannot render the entire scene with high-precision for complex large-scale aerial datasets since the sampling range along each view ray is insufficient to cover buildings adequately. Secondly, traditional NeRFs are infeasible to train on one GPU to enable interactive fly-throughs for modeling massive images. Instead, existing methods typically separate the whole scene into multiple regions and train a NeRF on each region, which are unaccustomed to different flight trajectories and difficult to achieve fast rendering. To that end, we propose Aerial-NeRF with three innovative modifications for jointly adapting NeRF in large-scale aerial rendering: (1) Designing an adaptive spatial partitioning and selection method based on drones' poses to adapt different flight trajectories; (2) Using similarity of poses instead of (expert) network for rendering speedup to determine which region a new viewpoint belongs to; (3) Developing an adaptive sampling approach for rendering performance improvement to cover the entire buildings at different heights. Extensive experiments have conducted to verify the effectiveness and efficiency of Aerial-NeRF, and new state-of-the-art results have been achieved on two public large-scale aerial datasets and presented SCUTic dataset. Note that our model allows us to perform rendering over 4 times as fast as compared to multiple competitors. Our dataset, code, and model are publicly available at https://drliuqi.github.io/."
                    }
                ]
            },
            "4f01ba33-b9f4-486b-bdf3-84d8d56f99a5": {
                "pk": "4f01ba33-b9f4-486b-bdf3-84d8d56f99a5",
                "name": "Sijia Luo",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "2548e90e-1590-40bd-b67b-53af7aca1435": {
                "pk": "2548e90e-1590-40bd-b67b-53af7aca1435",
                "name": "Xi Wang",
                "collaborators": [
                    "Laurence Aitchison",
                    "Chen Wang",
                    "Marc Alexa"
                ],
                "domain": [
                    "Time Series Analysis",
                    "Bayesian Inference",
                    "Machine Learning",
                    "Data Cleaning"
                ],
                "publications": [
                    {
                        "title": "Time Series Data Cleaning with Regular and Irregular Time Intervals",
                        "abstract": "Errors are prevalent in time series data, especially in the industrial field. Data with errors could not be stored in the database, which results in the loss of data assets. Handling the dirty data in time series is non-trivial, when given irregular time intervals. At present, to deal with these time series containing errors, besides keeping original erroneous data, discarding erroneous data and manually checking erroneous data, we can also use the cleaning algorithm widely used in the database to automatically clean the time series data. This survey provides a classification of time series data cleaning techniques and comprehensively reviews the state-of-the-art methods of each type. In particular, we have a special focus on the irregular time intervals. Besides we summarize data cleaning tools, systems and evaluation criteria from research and industry. Finally, we highlight possible directions time series data cleaning."
                    },
                    {
                        "title": "Maps of Visual Importance",
                        "abstract": "The importance of an element in a visual stimulus is commonly associated with the fixations during a free-viewing task. We argue that fixations are not always correlated with attention or awareness of visual objects. We suggest to filter the fixations recorded during exploration of the image based on the fixations recorded during recalling the image against a neutral background. This idea exploits that eye movements are a spatial index into the memory of a visual stimulus. We perform an experiment in which we record the eye movements of 30 observers during the presentation and recollection of 100 images. The locations of fixations during recall are only qualitatively related to the fixations during exploration. We develop a deformation mapping technique to align the fixations from recall with the fixation during exploration. This allows filtering the fixations based on proximity and a threshold on proximity provides a convenient slider to control the amount of filtering. Analyzing the spatial histograms resulting from the filtering procedure as well as the set of removed fixations shows that certain types of scene elements, which could be considered irrelevant, are removed. In this sense, they provide a measure of importance of visual elements for human observers."
                    },
                    {
                        "title": "Bayesian OOD detection with aleatoric uncertainty and outlier exposure",
                        "abstract": "Typical Bayesian approaches to OOD detection use epistemic uncertainty. Surprisingly from the Bayesian perspective, there are a number of methods that successfully use aleatoric uncertainty to detect OOD points (e.g. Hendryks et al. 2018). In addition, it is difficult to use outlier exposure to improve a Bayesian OOD detection model, as it is not clear whether it is possible or desirable to increase posterior (epistemic) uncertainty at outlier points. We show that a generative model of data curation provides a principled account of aleatoric uncertainty for OOD detection. In particular, aleatoric uncertainty signals a specific type of OOD point: one without a well-defined class-label, and our model of data curation gives a likelihood for these points, giving us a mechanism for conditioning on outlier points and thus performing principled Bayesian outlier exposure. Our principled Bayesian approach, combining aleatoric and epistemic uncertainty with outlier exposure performs better than methods using aleatoric or epistemic alone."
                    },
                    {
                        "title": "Robustness to corruption in pre-trained Bayesian neural networks",
                        "abstract": "We develop ShiftMatch, a new training-data-dependent likelihood for robustness to corruption in Bayesian neural networks (BNNs). ShiftMatch is inspired by the training-data-dependent \"EmpCov\" priors from Izmailov et al. (2021a), and efficiently matches test-time spatial correlations to those at training time. Critically, ShiftMatch is designed to leave the neural network's training time likelihood unchanged, allowing it to use publicly available samples from pre-trained BNNs. Using pre-trained HMC samples, ShiftMatch gives strong performance improvements on CIFAR-10-C, outperforms EmpCov priors (though ShiftMatch uses extra information from a minibatch of corrupted test points), and is perhaps the first Bayesian method capable of convincingly outperforming plain deep ensembles."
                    },
                    {
                        "title": "Batch size invariant Adam",
                        "abstract": "We propose a batch size invariant version of Adam, for use in large-scale, distributed settings, in which the mini-batch is divided into micro-batches which are distributed among worker nodes. For the v term, standard Adam first computes the average over micro-batch gradients, then squares, while in the batch size invariant Adam proposed here, we first square the micro-batch gradients, then average. Previous work (e.g. Malladi et al. 2022) used an alternative approach that involved a square-root scaling of the learning rate, but this approach requires strong assumptions to work; in particular that the gradient variance dominates the square of the expected gradient. In contrast, the approach proposed here gives batch size invariance without this assumption. We confirm that in practice our scheme gives batch size invariance in a much larger range of scenarios than the previous approach."
                    }
                ]
            },
            "91fa7cab-8ab9-4fad-a520-017a3216a3cb": {
                "pk": "91fa7cab-8ab9-4fad-a520-017a3216a3cb",
                "name": "Jie Tang",
                "collaborators": [
                    "Jie Liu",
                    "Gangshan Wu",
                    "Ming Ding",
                    "Jie Zhang",
                    "Yidong Cai",
                    "Chao Chen",
                    "Haonan Wang",
                    "Hang Zhao",
                    "Yutao Zhang",
                    "Xueyi Liu"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Semi-Supervised Learning",
                    "Image Super-Resolution",
                    "Video Frame Interpolation"
                ],
                "publications": [
                    {
                        "title": "Semi-supervised Learning on Graphs with Generative Adversarial Nets",
                        "abstract": "We investigate how generative adversarial nets (GANs) can help semi-supervised learning on graphs. We first provide insights on working principles of adversarial learning over graphs and then present GraphSGAN, a novel approach to semi-supervised learning on graphs. In GraphSGAN, generator and classifier networks play a novel competitive game. At equilibrium, generator generates fake samples in low-density areas between subgraphs. In order to discriminate fake samples from the real, classifier implicitly takes the density property of subgraph into consideration. An efficient adversarial learning algorithm has been developed to improve traditional normalized graph Laplacian regularization with a theoretical guarantee. Experimental results on several different genres of datasets show that the proposed GraphSGAN significantly outperforms several state-of-the-art methods. GraphSGAN can be also trained using mini-batch, thus enjoys the scalability advantage."
                    },
                    {
                        "title": "AdaDM: Enabling Normalization for Image Super-Resolution",
                        "abstract": "Normalization like Batch Normalization (BN) is a milestone technique to normalize the distributions of intermediate layers in deep learning, enabling faster training and better generalization accuracy. However, in fidelity image Super-Resolution (SR), it is believed that normalization layers get rid of range flexibility by normalizing the features and they are simply removed from modern SR networks. In this paper, we study this phenomenon quantitatively and qualitatively. We found that the standard deviation of the residual feature shrinks a lot after normalization layers, which causes the performance degradation in SR networks. Standard deviation reflects the amount of variation of pixel values. When the variation becomes smaller, the edges will become less discriminative for the network to resolve. To address this problem, we propose an Adaptive Deviation Modulator (AdaDM), in which a modulation factor is adaptively predicted to amplify the pixel deviation. For better generalization performance, we apply BN in state-of-the-art SR networks with the proposed AdaDM. Meanwhile, the deviation amplification strategy in AdaDM makes the edge information in the feature more distinguishable. As a consequence, SR networks with BN and our AdaDM can get substantial performance improvements on benchmark datasets. Extensive experiments have been conducted to show the effectiveness of our method."
                    },
                    {
                        "title": "Robust Object Modeling for Visual Tracking",
                        "abstract": "Object modeling has become a core part of recent tracking frameworks. Current popular tackers use Transformer attention to extract the template feature separately or interactively with the search region. However, separate template learning lacks communication between the template and search regions, which brings difficulty in extracting discriminative target-oriented features. On the other hand, interactive template learning produces hybrid template features, which may introduce potential distractors to the template via the cluttered search regions. To enjoy the merits of both methods, we propose a robust object modeling framework for visual tracking (ROMTrack), which simultaneously models the inherent template and the hybrid template features. As a result, harmful distractors can be suppressed by combining the inherent features of target objects with search regions' guidance. Target-related features can also be extracted using the hybrid template, thus resulting in a more robust object modeling framework. To further enhance robustness, we present novel variation tokens to depict the ever-changing appearance of target objects. Variation tokens are adaptable to object deformation and appearance variations, which can boost overall performance with negligible computation. Experiments show that our ROMTrack sets a new state-of-the-art on multiple benchmarks."
                    },
                    {
                        "title": "Residual Feature Distillation Network for Lightweight Image Super-Resolution",
                        "abstract": "Recent advances in single image super-resolution (SISR) explored the power of convolutional neural network (CNN) to achieve a better performance. Despite the great success of CNN-based methods, it is not easy to apply these methods to edge devices due to the requirement of heavy computation. To solve this problem, various fast and lightweight CNN models have been proposed. The information distillation network is one of the state-of-the-art methods, which adopts the channel splitting operation to extract distilled features. However, it is not clear enough how this operation helps in the design of efficient SISR models. In this paper, we propose the feature distillation connection (FDC) that is functionally equivalent to the channel splitting operation while being more lightweight and flexible. Thanks to FDC, we can rethink the information multi-distillation network (IMDN) and propose a lightweight and accurate SISR model called residual feature distillation network (RFDN). RFDN uses multiple feature distillation connections to learn more discriminative feature representations. We also propose a shallow residual block (SRB) as the main building block of RFDN so that the network can benefit most from residual learning while still being lightweight enough. Extensive experimental results show that the proposed RFDN achieve a better trade-off against the state-of-the-art methods in terms of performance and model complexity. Moreover, we propose an enhanced RFDN (E-RFDN) and won the first place in the AIM 2020 efficient super-resolution challenge. Code will be available at https://github.com/njulj/RFDN."
                    },
                    {
                        "title": "From Coarse to Fine: Hierarchical Pixel Integration for Lightweight Image Super-Resolution",
                        "abstract": "Image super-resolution (SR) serves as a fundamental tool for the processing and transmission of multimedia data. Recently, Transformer-based models have achieved competitive performances in image SR. They divide images into fixed-size patches and apply self-attention on these patches to model long-range dependencies among pixels. However, this architecture design is originated for high-level vision tasks, which lacks design guideline from SR knowledge. In this paper, we aim to design a new attention block whose insights are from the interpretation of Local Attribution Map (LAM) for SR networks. Specifically, LAM presents a hierarchical importance map where the most important pixels are located in a fine area of a patch and some less important pixels are spread in a coarse area of the whole image. To access pixels in the coarse area, instead of using a very large patch size, we propose a lightweight Global Pixel Access (GPA) module that applies cross-attention with the most similar patch in an image. In the fine area, we use an Intra-Patch Self-Attention (IPSA) module to model long-range pixel dependencies in a local patch, and then a $3\\times3$ convolution is applied to process the finest details. In addition, a Cascaded Patch Division (CPD) strategy is proposed to enhance perceptual quality of recovered images. Extensive experiments suggest that our method outperforms state-of-the-art lightweight SR methods by a large margin. Code is available at https://github.com/passerer/HPINet."
                    },
                    {
                        "title": "Lightweight Super-Resolution Head for Human Pose Estimation",
                        "abstract": "Heatmap-based methods have become the mainstream method for pose estimation due to their superior performance. However, heatmap-based approaches suffer from significant quantization errors with downscale heatmaps, which result in limited performance and the detrimental effects of intermediate supervision. Previous heatmap-based methods relied heavily on additional post-processing to mitigate quantization errors. Some heatmap-based approaches improve the resolution of feature maps by using multiple costly upsampling layers to improve localization precision. To solve the above issues, we creatively view the backbone network as a degradation process and thus reformulate the heatmap prediction as a Super-Resolution (SR) task. We first propose the SR head, which predicts heatmaps with a spatial resolution higher than the input feature maps (or even consistent with the input image) by super-resolution, to effectively reduce the quantization error and the dependence on further post-processing. Besides, we propose SRPose to gradually recover the HR heatmaps from LR heatmaps and degraded features in a coarse-to-fine manner. To reduce the training difficulty of HR heatmaps, SRPose applies SR heads to supervise the intermediate features in each stage. In addition, the SR head is a lightweight and generic head that applies to top-down and bottom-up methods. Extensive experiments on the COCO, MPII, and CrowdPose datasets show that SRPose outperforms the corresponding heatmap-based approaches. The code and models are available at https://github.com/haonanwang0522/SRPose."
                    },
                    {
                        "title": "A Data Streaming Process Framework for Autonomous Driving By Edge",
                        "abstract": "In recent years, with the rapid development of sensing technology and the Internet of Things (IoT), sensors play increasingly important roles in traffic control, medical monitoring, industrial production and etc. They generated high volume of data in a streaming way that often need to be processed in real time. Therefore, streaming data computing technology plays an indispensable role in the real-time processing of sensor data in high throughput but low latency. In view of the above problems, the proposed framework is implemented on top of Spark Streaming, which builds up a gray model based traffic flow monitor, a traffic prediction orientated prediction layer and a fuzzy control based Batch Interval dynamic adjustment layer for Spark Streaming. It could forecast the variation of sensors data arrive rate, make streaming Batch Interval adjustment in advance and implement real-time streaming process by edge. Therefore, it can realize the monitor and prediction of the data flow changes of the autonomous driving vehicle sensor data in geographical coverage of edge computing node area, meanwhile minimize the end-to-end latency but satisfy the application throughput requirements. The experiments show that it can predict short-term traffic with no more than 4% relative error in a whole day. By making batch consuming rate close to data generating rate, it can maintain system stability well even when arrival data rate changes rapidly. The Batch Interval can be converged to a suitable value in two minutes when data arrival rate is doubled. Compared with vanilla version Spark Streaming, where there has serious task accumulation and introduces large delay, it can reduce 35% latency by squeezing Batch Interval when data arrival rate is low; it also can significantly improve system throughput by only at most 25% Batch Interval increase when data arrival rate is high."
                    },
                    {
                        "title": "Social Network Integration: Towards Constructing the Social Graph",
                        "abstract": "In this work, we formulate the problem of social network integration. It takes multiple observed social networks as input and returns an integrated global social graph where each node corresponds to a real person. The key challenge for social network integration is to discover the correspondences or interlinks across different social networks.   We engaged an in-depth analysis across three online social networks, AMiner, Linkedin, and Videolectures in order to address what reveals users' social identity, whether the social factors consistent across different social networks and how we can leverage these information to perform integration.   We proposed a unified framework for the social network integration task. It crawls data from multiple social networks and further discovers accounts correspond to the same real person from the obtained networks. We use a probabilistic model to determine such correspondence, it incorporates features like the consistency of social status and social ties across different, as well as one-to-one mapping constraint and logical transitivity to jointly make the prediction. Empirical experiments verify the effectiveness of our method."
                    },
                    {
                        "title": "Network representation learning: A macro and micro view",
                        "abstract": "Graph is a universe data structure that is widely used to organize data in real-world. Various real-word networks like the transportation network, social and academic network can be represented by graphs. Recent years have witnessed the quick development on representing vertices in the network into a low-dimensional vector space, referred to as network representation learning. Representation learning can facilitate the design of new algorithms on the graph data. In this survey, we conduct a comprehensive review of current literature on network representation learning. Existing algorithms can be categorized into three groups: shallow embedding models, heterogeneous network embedding models, graph neural network based models. We review state-of-the-art algorithms for each category and discuss the essential differences between these algorithms. One advantage of the survey is that we systematically study the underlying theoretical foundations underlying the different categories of algorithms, which offers deep insights for better understanding the development of the network representation learning field."
                    },
                    {
                        "title": "CogCartoon: Towards Practical Story Visualization",
                        "abstract": "The state-of-the-art methods for story visualization demonstrate a significant demand for training data and storage, as well as limited flexibility in story presentation, thereby rendering them impractical for real-world applications. We introduce CogCartoon, a practical story visualization method based on pre-trained diffusion models. To alleviate dependence on data and storage, we propose an innovative strategy of character-plugin generation that can represent a specific character as a compact 316 KB plugin by using a few training samples. To facilitate enhanced flexibility, we employ a strategy of plugin-guided and layout-guided inference, enabling users to seamlessly incorporate new characters and custom layouts into the generated image results at their convenience. We have conducted comprehensive qualitative and quantitative studies, providing compelling evidence for the superiority of CogCartoon over existing methodologies. Moreover, CogCartoon demonstrates its power in tackling challenging tasks, including long story visualization and realistic style story visualization."
                    },
                    {
                        "title": "Anchor-based Plain Net for Mobile Image Super-Resolution",
                        "abstract": "Along with the rapid development of real-world applications, higher requirements on the accuracy and efficiency of image super-resolution (SR) are brought forward. Though existing methods have achieved remarkable success, the majority of them demand plenty of computational resources and large amount of RAM, and thus they can not be well applied to mobile device. In this paper, we aim at designing efficient architecture for 8-bit quantization and deploy it on mobile device. First, we conduct an experiment about meta-node latency by decomposing lightweight SR architectures, which determines the portable operations we can utilize. Then, we dig deeper into what kind of architecture is beneficial to 8-bit quantization and propose anchor-based plain net (ABPN). Finally, we adopt quantization-aware training strategy to further boost the performance. Our model can outperform 8-bit quantized FSRCNN by nearly 2dB in terms of PSNR, while satisfying realistic needs at the same time. Code is avaliable at https://github.com/NJU- Jet/SR_Mobile_Quantization."
                    },
                    {
                        "title": "Video Frame Interpolation with Densely Queried Bilateral Correlation",
                        "abstract": "Video Frame Interpolation (VFI) aims to synthesize non-existent intermediate frames between existent frames. Flow-based VFI algorithms estimate intermediate motion fields to warp the existent frames. Real-world motions' complexity and the reference frame's absence make motion estimation challenging. Many state-of-the-art approaches explicitly model the correlations between two neighboring frames for more accurate motion estimation. In common approaches, the receptive field of correlation modeling at higher resolution depends on the motion fields estimated beforehand. Such receptive field dependency makes common motion estimation approaches poor at coping with small and fast-moving objects. To better model correlations and to produce more accurate motion fields, we propose the Densely Queried Bilateral Correlation (DQBC) that gets rid of the receptive field dependency problem and thus is more friendly to small and fast-moving objects. The motion fields generated with the help of DQBC are further refined and up-sampled with context features. After the motion fields are fixed, a CNN-based SynthNet synthesizes the final interpolated frame. Experiments show that our approach enjoys higher accuracy and less inference time than the state-of-the-art. Source code is available at https://github.com/kinoud/DQBC."
                    },
                    {
                        "title": "Spectral Network Embedding: A Fast and Scalable Method via Sparsity",
                        "abstract": "Network embedding aims to learn low-dimensional representations of nodes in a network, while the network structure and inherent properties are preserved. It has attracted tremendous attention recently due to significant progress in downstream network learning tasks, such as node classification, link prediction, and visualization. However, most existing network embedding methods suffer from the expensive computations due to the large volume of networks. In this paper, we propose a $10\\times \\sim 100\\times$ faster network embedding method, called Progle, by elegantly utilizing the sparsity property of online networks and spectral analysis. In Progle, we first construct a \\textit{sparse} proximity matrix and train the network embedding efficiently via sparse matrix decomposition. Then we introduce a network propagation pattern via spectral analysis to incorporate local and global structure information into the embedding. Besides, this model can be generalized to integrate network information into other insufficiently trained embeddings at speed. Benefiting from sparse spectral network embedding, our experiment on four different datasets shows that Progle outperforms or is comparable to state-of-the-art unsupervised comparison approaches---DeepWalk, LINE, node2vec, GraRep, and HOPE, regarding accuracy, while is $10\\times$ faster than the fastest word2vec-based method. Finally, we validate the scalability of Progle both in real large-scale networks and multiple scales of synthetic networks."
                    }
                ]
            }
        }
    },
    "2410.04037": {
        "paper_data": {
            "title": "Is Score Matching Suitable for Estimating Point Processes?",
            "url": "http://arxiv.org/abs/2410.04037v1",
            "arxiv_id": "2410.04037",
            "authors": [
                "Haoqun Cao",
                "Zizhuo Meng",
                "Tianjun Ke",
                "Feng Zhou"
            ],
            "abstract": "Score matching estimators have gained widespread attention in recent years partly because they are free from calculating the integral of normalizing constant, thereby addressing the computational challenges in maximum likelihood estimation (MLE). Some existing works have proposed score matching estimators for point processes. However, this work demonstrates that the incompleteness of the estimators proposed in those works renders them applicable only to specific problems, and they fail for more general point processes. To address this issue, this work introduces the weighted score matching estimator to point processes. Theoretically, we prove the consistency of our estimator and establish its rate of convergence. Experimental results indicate that our estimator accurately estimates model parameters on synthetic data and yields results consistent with MLE on real data. In contrast, existing score matching estimators fail to perform effectively. Codes are publicly available at \\url{https://github.com/KenCao2007/WSM_TPP}.",
            "introduction": "   1 Introduction  Point processes are a class of statistical models used to characterize event occurrences. Typical models include Poisson processes\u00a0[9] and Hawkes processes\u00a0[3]. Their applications span various fields such as seismology\u00a0[16; 17], finance\u00a0[1; 4], criminology\u00a0[15], and neuroscience\u00a0[11; 26]. In the field of point processes, maximum likelihood estimation (MLE) has been a conventional estimator. However, MLE has an inherent limitation: it requires the computation of the normalizing constant, which corresponds to the intensity integral term in the likelihood. Except for simple cases, calculating the intensity integral analytically is generally infeasible. This necessitates the use of numerical integration methods like Monte Carlo or quadrature for approximating the computation. This introduces approximation errors, and more importantly, for high-dimensional problems, numerical integration encounters the curse of dimensionality, rendering training infeasible.   To address this issue, prior research has introduced the concept of score matching (SM)\u00a0[5] to the field of point processes. For instance, [18] derived the application of SM to the estimation of traditional statistical Poisson processes. Furthermore, [24] extended the use of SM to the estimation of deep covariate spatio-temporal point processes. [10] also generalized the application of SM to Hawkes processes. These works have greatly advanced the utilization of SM for point processes. However, in practical applications, we have found that these estimators only work for specific point processes. For more general cases, these estimators cannot accurately estimate model parameters, even for some simple statistical point processes. One of the core contribution of this work is to theoretically demonstrate the incompleteness of the estimators proposed in the aforementioned studies.   The incompleteness of the estimators in the aforementioned studies stems from the transition from explicit SM to implicit SM. The explicit SM estimates model parameters by minimizing the expected distance between the gradient of the log-density of the model and the gradient of the log-density of the data. However, we cannot directly minimize the above objective function since it depends on the unknown data distribution. To facilitate solving, we need to convert the above explicit SM to implicit SM by using a trick of integration by parts, provided that some regularity conditions are satisfied\u00a0[5]. In [18; 24; 10], they assume that the required regularity conditions are satisfied in their point process models and directly employ the implicit SM objective. However, as demonstrated in Section\u00a03, the required regularity conditions cannot be met for general point processes. This implies that the concise implicit SM objectives (Equation (2) in [18], Equation (10) in [24], Equation (4) in [10]) are incomplete, and they cannot accurately estimate parameters for general point processes.   To address this issue, this work introduces a (autoregressive) weighted score matching (WSM) estimator that can be applied to more general point processes. WSM eliminates the intractable terms in SM objective by adding a weight function that takes zero at the boundary of the integration region. Compared to previous work on WSM [6; 22; 12], we are the first work to apply WSM on a stochastic process where the dimension NTsubscript\ud835\udc41\ud835\udc47N_{T}italic_N start_POSTSUBSCRIPT italic_T end_POSTSUBSCRIPT is also random. This stochasticity in dimensionality poses greater challenges to the derivation, requiring special treatment to address this issue.   Specifically, we make following contributions: (1) We theoretically demonstrate that implicit (autoregressive) SM",
            "references": [
                {
                    "title": "SMURF-THP: Score Matching-based UnceRtainty quantiFication for Transformer Hawkes Process",
                    "abstract": "Transformer Hawkes process models have shown to be successful in modeling event sequence data. However, most of the existing training methods rely on maximizing the likelihood of event sequences, which involves calculating some intractable integral. Moreover, the existing methods fail to provide uncertainty quantification for model predictions, e.g., confidence intervals for the predicted event's arrival time. To address these issues, we propose SMURF-THP, a score-based method for learning Transformer Hawkes process and quantifying prediction uncertainty. Specifically, SMURF-THP learns the score function of events' arrival time based on a score-matching objective that avoids the intractable computation. With such a learned score function, we can sample arrival time of events from the predictive distribution. This naturally allows for the quantification of uncertainty by computing confidence intervals over the generated samples. We conduct extensive experiments in both event type prediction and uncertainty quantification of arrival time. In all the experiments, SMURF-THP outperforms existing likelihood-based methods in confidence calibration while exhibiting comparable prediction accuracy."
                },
                {
                    "title": "Integration-free Training for Spatio-temporal Multimodal Covariate Deep Kernel Point Processes",
                    "abstract": "In this study, we propose a novel deep spatio-temporal point process model, Deep Kernel Mixture Point Processes (DKMPP), that incorporates multimodal covariate information. DKMPP is an enhanced version of Deep Mixture Point Processes (DMPP), which uses a more flexible deep kernel to model complex relationships between events and covariate data, improving the model's expressiveness. To address the intractable training procedure of DKMPP due to the non-integrable deep kernel, we utilize an integration-free method based on score matching, and further improve efficiency by adopting a scalable denoising score matching method. Our experiments demonstrate that DKMPP and its corresponding score-based estimators outperform baseline models, showcasing the advantages of incorporating covariate information, utilizing a deep kernel, and employing score-based estimators."
                },
                {
                    "title": "HYPRO: A Hybridly Normalized Probabilistic Model for Long-Horizon Prediction of Event Sequences",
                    "abstract": "In this paper, we tackle the important yet under-investigated problem of making long-horizon prediction of event sequences. Existing state-of-the-art models do not perform well at this task due to their autoregressive structure. We propose HYPRO, a hybridly normalized probabilistic model that naturally fits this task: its first part is an autoregressive base model that learns to propose predictions; its second part is an energy function that learns to reweight the proposals such that more realistic predictions end up with higher probabilities. We also propose efficient training and inference algorithms for this model. Experiments on multiple real-world datasets demonstrate that our proposed HYPRO model can significantly outperform previous models at making long-horizon predictions of future events. We also conduct a range of ablation studies to investigate the effectiveness of each component of our proposed methods."
                },
                {
                    "title": "Autoregressive Score Matching",
                    "abstract": "Autoregressive models use chain rule to define a joint probability distribution as a product of conditionals. These conditionals need to be normalized, imposing constraints on the functional families that can be used. To increase flexibility, we propose autoregressive conditional score models (AR-CSM) where we parameterize the joint distribution in terms of the derivatives of univariate log-conditionals (scores), which need not be normalized. To train AR-CSM, we introduce a new divergence between distributions named Composite Score Matching (CSM). For AR-CSM models, this divergence between data and model distributions can be computed and optimized efficiently, requiring no expensive sampling or adversarial training. Compared to previous score matching algorithms, our method is more scalable to high dimensional data and more stable to optimize. We show with extensive experimental results that it can be applied to density estimation on synthetic data, image generation, image denoising, and training latent variable models with implicit encoders."
                },
                {
                    "title": "Self-Attentive Hawkes Process",
                    "abstract": "Capturing the occurrence dynamics is crucial to predicting which type of events will happen next and when . A common method to do this is through Hawkes processes. To enhance their capacity, recurrent neural networks (RNNs) have been incorporated due to RNNs\u2019 successes in processing sequential data such as languages. Recent evidence suggests that self-attention is more competent than RNNs in dealing with languages. However, we are unaware of the effectiveness of self-attention in the context of Hawkes processes. This study aims to \ufb01ll the gap by designing a self-attentive Hawkes process (SAHP). SAHP employs self-attention to summarise the in\ufb02uence of history events and compute the probability of the next event. One de\ufb01cit of the conventional self-attention, when applied to event sequences, is that its positional encoding only considers the order of a sequence ignoring the time intervals between events. To overcome this de\ufb01cit, we modify its encoding by translating time intervals into phase shifts of sinusoidal functions. Experiments on goodness-of-\ufb01t and prediction tasks show the improved capability of SAHP. Furthermore, SAHP is more interpretable than RNN-based counter-parts because the learnt attention weights reveal contributions of one event type to the happening of another type. To the best of our knowledge, this is the \ufb01rst work that studies the effectiveness of self-attention in Hawkes processes."
                },
                {
                    "title": "Transformer Hawkes Process",
                    "abstract": "Modern data acquisition routinely produce massive amounts of event sequence data in various domains, such as social media, healthcare, and financial markets. These data often exhibit complicated short-term and long-term temporal dependencies. However, most of the existing recurrent neural network based point process models fail to capture such dependencies, and yield unreliable prediction performance. To address this issue, we propose a Transformer Hawkes Process (THP) model, which leverages the self-attention mechanism to capture long-term dependencies and meanwhile enjoys computational efficiency. Numerical experiments on various datasets show that THP outperforms existing models in terms of both likelihood and event prediction accuracy by a notable margin. Moreover, THP is quite general and can incorporate additional structural knowledge. We provide a concrete example, where THP achieves improved prediction performance for learning multiple point processes when incorporating their relational information."
                },
                {
                    "title": "Estimating Density Models with Truncation Boundaries using Score Matching",
                    "abstract": "Truncated densities are probability density functions defined on truncated domains. They share the same parametric form with their non-truncated counterparts up to a normalizing constant. Since the computation of their normalizing constants is usually infeasible, Maximum Likelihood Estimation cannot be easily applied to estimate truncated density models. Score Matching (SM) is a powerful tool for fitting parameters using only unnormalized models. However, it cannot be directly applied here as boundary conditions used to derive a tractable SM objective are not satisfied by truncated densities. In this paper, we study parameter estimation for truncated probability densities using SM. The estimator minimizes a weighted Fisher divergence. The weight function is simply the shortest distance from a data point to the boundary of the domain. We show this choice of weight function naturally arises from minimizing the Stein discrepancy as well as upperbounding the finite-sample estimation error. The usefulness of our method is demonstrated by numerical experiments and a study on the Chicago crime data set. We also show that the proposed density estimation can correct the outlier-trimming bias caused by aggressive outlier detection methods."
                },
                {
                    "title": "Hawkes processes and their applications to finance: a review",
                    "abstract": "Hawkes (1971a, 1971b, 1972) introduced a family of models for stochastic point processes called \u2018self-exciting and mutually exciting point processes\u2019 the essential property of which was that the oc..."
                },
                {
                    "title": "Score-matching estimators for continuous-time point-process regression models",
                    "abstract": "We introduce a new class of efficient estimators based on score matching for probabilistic point process models. Unlike discretised likelihood-based estimators, score matching estimators operate on continuous-time data, with computational demands that grow with the number of events rather than with total observation time. Furthermore, estimators for many common regression models can be obtained in closed form, rather than by iteration. This new approach to estimation may thus expand the range of tractable models available for event-based data."
                },
                {
                    "title": "SEISMIC: A Self-Exciting Point Process Model for Predicting Tweet Popularity",
                    "abstract": "Social networking websites allow users to create and share content. Big information cascades of post resharing can form as users of these sites reshare others' posts with their friends and followers. One of the central challenges in understanding such cascading behaviors is in forecasting information outbreaks, where a single post becomes widely popular by being reshared by many users. In this paper, we focus on predicting the final number of reshares of a given post. We build on the theory of self-exciting point processes to develop a statistical model that allows us to make accurate predictions. Our model requires no training or expensive feature engineering. It results in a simple and efficiently computable formula that allows us to answer questions, in real-time, such as: Given a post's resharing history so far, what is our current estimate of its final number of reshares? Is the post resharing cascade past the initial stage of explosive growth? And, which posts will be the most reshared in the future? We validate our model using one month of complete Twitter data and demonstrate a strong improvement in predictive accuracy over existing approaches. Our model gives only 15% relative error in predicting final size of an average information cascade after observing it for just one hour."
                },
                {
                    "title": "{SNAP Datasets}: {Stanford} Large Network Dataset Collection",
                    "abstract": "A collection of more than 50 large network datasets from tens of thousands of nodes and edges to tens of millions of nodes and edges. In includes social networks, web graphs, road networks, internet networks, citation networks, collaboration networks, and communication networks."
                },
                {
                    "title": "Inference for a Nonstationary Self-Exciting Point Process with an Application in Ultra-High Frequency Financial Data Modeling",
                    "abstract": "Self-exciting point processes (SEPPs), or Hawkes processes, have found applications in a wide range of fields, such as epidemiology, seismology, neuroscience, engineering, and more recently financial econometrics and social interactions. In the traditional SEPP models, the baseline intensity is assumed to be a constant. This has restricted the application of SEPPs to situations where there is clearly a self-exciting phenomenon, but a constant baseline intensity is inappropriate. In this paper, to model point processes with varying baseline intensity, we introduce SEPP models with time-varying background intensities (SEPPVB, for short). We show that SEPPVB models are competitive with autoregressive conditional SEPP models (Engle and Russell 1998) for modeling ultra-high frequency data. We also develop asymptotic theory for maximum likelihood estimation based inference of parametric SEPP models, including SEPPVB. We illustrate applications to ultra-high frequency financial data analysis, and we compare performance with the autoregressive conditional duration models."
                },
                {
                    "title": "Modeling and estimation of multi-source clustering in crime and security data",
                    "abstract": "While the presence of clustering in crime and security event data is well established, the mechanism(s) by which clustering arises is not fully understood. Both contagion models and history independent correlation models are applied, but not simultaneously. In an attempt to disentangle contagion from other types of correlation, we consider a Hawkes process with background rate driven by a log Gaussian Cox process. Our inference methodology is an efficient Metropolis adjusted Langevin algorithm for filtering of the intensity and estimation of the model parameters. We apply the methodology to property and violent crime data from Chicago, terrorist attack data from Northern Ireland and Israel, and civilian casualty data from Iraq. For each data set we quantify the uncertainty in the levels of contagion vs. history independent correlation."
                },
                {
                    "title": "Hawkes model for price and trades high-frequency dynamics",
                    "abstract": "We introduce a multivariate Hawkes process that accounts for the dynamics of market prices through the impact of market order arrivals at microstructural level. Our model is a point process mainly characterized by four kernels associated with, respectively, the trade arrival self-excitation, the price changes mean reversion, the impact of trade arrivals on price variations and the feedback of price changes on trading activity. It allows one to account for both stylized facts of market price microstructure (including random time arrival of price moves, discrete price grid, high-frequency mean reversion, correlation functions behaviour at various time scales) and the stylized facts of market impact (mainly the concave-square-root-like/relaxation characteristic shape of the market impact of a meta-order). Moreover, it allows one to estimate the entire market impact profile from anonymous market data. We show that these kernels can be empirically estimated from the empirical conditional mean intensities. We provide numerical examples, application to real data and comparisons to former approaches."
                },
                {
                    "title": "A Connection Between Score Matching and Denoising Autoencoders",
                    "abstract": "Denoising autoencoders have been previously shown to be competitive alternatives to restricted Boltzmann machines for unsupervised pretraining of each layer of a deep architecture. We show that a simple denoising autoencoder training criterion is equivalent to matching the score (with respect to the data) of a specific energy-based model to that of a nonparametric Parzen density estimator of the data. This yields several useful insights. It defines a proper probabilistic model for the denoising autoencoder technique, which makes it in principle possible to sample from them or rank examples by their energy. It suggests a different way to apply score matching that is related to learning to denoise and does not require computing second derivatives. It justifies the use of tied weights between the encoder and decoder and suggests ways to extend the success of denoising autoencoders to a larger family of energy-based models."
                },
                {
                    "title": "Estimation of Non-Normalized Statistical Models by Score Matching",
                    "abstract": "One often wants to estimate statistical models where the probability density function is known only up to a multiplicative normalization constant. Typically, one then has to resort to Markov Chain Monte Carlo methods, or approximations of the normalization constant. Here, we propose that such models can be estimated by minimizing the expected squared distance between the gradient of the log-density given by the model and the gradient of the log-density of the observed data. While the estimation of the gradient of log-density function is, in principle, a very difficult non-parametric problem, we prove a surprising result that gives a simple formula for this objective function. The density function of the observed data does not appear in this formula, which simplifies to a sample average of a sum of some derivatives of the log-density given by the model. The validity of the method is demonstrated on multivariate Gaussian and independent component analysis models, and by estimating an overcomplete filter set for natural image data."
                },
                {
                    "title": "Envelope Theorems for Arbitrary Choice Sets",
                    "abstract": "The standard envelope theorems apply to choice sets with convex and topological structure, providing sufficient conditions for the value function to be differentiable in a parameter and characterizing its derivative. This paper studies optimization with arbitrary choice sets and shows that the traditional envelope formula holds at any differentiability point of the value function. We also provide conditions for the value function to be, variously, absolutely continuous, left- and right-differentiable, or fully differentiable. These results are applied to mechanism design, convex programming, continuous optimization problems, saddle-point problems, problems with parameterized constraints, and optimal stopping problems."
                },
                {
                    "title": "Spectra of some self-exciting and mutually exciting point processes",
                    "abstract": "SUMMARY In recent years methods of data analysis for point processes have received some attention, for example, by Cox & Lewis (1966) and Lewis (1964). In particular Bartlett (1963a,b) has introduced methods of analysis based on the point spectrum. Theoretical models are relatively sparse. In this paper the theoretical properties of a class of processes with particular reference to the point spectrum or corresponding covariance density functions are discussed. A particular result is a self-exciting process with the same second-order properties as a certain doubly stochastic process. These are not distinguishable by methods of data analysis based on these properties."
                },
                {
                    "title": "Efficient Inference for Dynamic Flexible Interactions of Neural Populations",
                    "abstract": "Hawkes process provides an e\ufb00ective statistical framework for analyzing the interactions of neural spiking activities. Although utilized in many real applications, the classic Hawkes process is incapable of modeling inhibitory interactions among neural population. Instead, the nonlinear Hawkes process allows for modeling a more \ufb02exible in\ufb02uence pattern with excitatory or inhibitory interactions. This work proposes a \ufb02exible nonlinear Hawkes process variant based on sigmoid nonlinearity. To ease inference, three sets of auxiliary latent variables (P\u00b4olya-Gamma variables, latent marked Poisson processes and sparsity variables) are augmented to make functional connection weights appear in a Gaussian form, which enables simple iterative algorithms with analytical updates. As a result, the e\ufb03cient Gibbs sampler, expectation-maximization (EM) algorithm and mean-\ufb01eld (MF) approximation are derived to estimate the interactions among neural populations. Furthermore, to reconcile with time-varying neural systems, the proposed time-invariant model is extended to a dynamic version by introducing a Markov state process. Similarly, three analytical iterative inference algorithms: Gibbs sampler, EM algorithm and mean-\ufb01eld approximation are derived. We compare the accuracy and e\ufb03ciency of these inference algorithms on synthetic data, and further experiment on real neural recordings to demonstrate that the developed models achieve superior performance over the state-of-the-art competitors"
                },
                {
                    "title": "Generalized score matching for non-negative data",
                    "abstract": "Evaluating the joint signi\ufb01cance of covariates is of fundamental importance in a wide range of applications. To this end, p-values are frequently employed and produced by algorithms that are powered by classical large-sample asymptotic theory. It is well known that the conventional p-values in Gaussian linear model are valid even when the dimensionality is a non-vanishing fraction of the sample size, but can break down when the design matrix becomes singular in higher dimensions or when the error distribution deviates from Gaus-sianity. A natural question is when the conventional p-values in generalized linear models become invalid in diverging dimensions. We establish that such a breakdown can occur early in nonlinear models. Our theoretical characterizations are con\ufb01rmed by simulation studies."
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we accurately estimate model parameters for general point processes when existing score matching estimators fail due to the incompleteness of their objectives?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of point processes, as accurate parameter estimation is foundational for various applications in seismology, finance, criminology, and neuroscience. By addressing the limitations of existing methods, this research could lead to more robust statistical models that can handle a wider range of point processes, thereby enhancing predictive capabilities and practical applications in these fields. Furthermore, it could inspire future research to explore new methodologies that build on the proposed weighted score matching (WSM) approach, potentially leading to breakthroughs in understanding complex stochastic processes.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem arise from the inherent complexities of point processes and the limitations of existing estimation techniques. Naive approaches, such as maximum likelihood estimation (MLE), fail due to the necessity of computing the normalizing constant, which is often infeasible for high-dimensional problems. Additionally, the transition from explicit to implicit score matching introduces further complications, as the required regularity conditions for the implicit objectives are not met in general cases. This results in incomplete estimators that cannot accurately capture the dynamics of more complex point processes, necessitating a novel approach to overcome these technical and theoretical obstacles.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on specific point processes and assumed that the regularity conditions necessary for implicit score matching are satisfied. This has led to a reliance on incomplete estimators that do not generalize well. The barriers to solving this problem include a lack of understanding of the limitations of existing methods and the complexities introduced by stochasticity in dimensionality. Our approach differs by introducing a weighted score matching (WSM) estimator that can handle a broader class of point processes, addressing the shortcomings of prior work and providing a more comprehensive solution.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of a weighted score matching (WSM) estimator that incorporates a weight function to eliminate intractable terms in the score matching objective. This approach is designed to be applicable to more general point processes, including those with random dimensionality. We will evaluate the performance of the WSM estimator using synthetic and real"
            }
        },
        "author_data": {
            "3882c126-05a2-42cf-b0bd-a385216138c1": {
                "pk": "3882c126-05a2-42cf-b0bd-a385216138c1",
                "name": "Haoqun Cao",
                "collaborators": [
                    "Tianjun Ke",
                    "Feng Zhou",
                    "Zenan Ling"
                ],
                "domain": [
                    "Few-shot Learning",
                    "Bayesian Inference",
                    "Gaussian Processes",
                    "Meta-learning"
                ],
                "publications": [
                    {
                        "title": "Accelerating Convergence in Bayesian Few-Shot Classification",
                        "abstract": "Bayesian few-shot classification has been a focal point in the field of few-shot learning. This paper seamlessly integrates mirror descent-based variational inference into Gaussian process-based few-shot classification, addressing the challenge of non-conjugate inference. By leveraging non-Euclidean geometry, mirror descent achieves accelerated convergence by providing the steepest descent direction along the corresponding manifold. It also exhibits the parameterization invariance property concerning the variational distribution. Experimental results demonstrate competitive classification accuracy, improved uncertainty quantification, and faster convergence compared to baseline models. Additionally, we investigate the impact of hyperparameters and components. Code is publicly available at https://github.com/keanson/MD-BSFC."
                    },
                    {
                        "title": "Revisiting Logistic-softmax Likelihood in Bayesian Meta-Learning for Few-Shot Classification",
                        "abstract": "Meta-learning has demonstrated promising results in few-shot classification (FSC) by learning to solve new problems using prior knowledge. Bayesian methods are effective at characterizing uncertainty in FSC, which is crucial in high-risk fields. In this context, the logistic-softmax likelihood is often employed as an alternative to the softmax likelihood in multi-class Gaussian process classification due to its conditional conjugacy property. However, the theoretical property of logistic-softmax is not clear and previous research indicated that the inherent uncertainty of logistic-softmax leads to suboptimal performance. To mitigate these issues, we revisit and redesign the logistic-softmax likelihood, which enables control of the \\textit{a priori} confidence level through a temperature parameter. Furthermore, we theoretically and empirically show that softmax can be viewed as a special case of logistic-softmax and logistic-softmax induces a larger family of data distribution than softmax. Utilizing modified logistic-softmax, we integrate the data augmentation technique into the deep kernel based Gaussian process meta-learning framework, and derive an analytical mean-field approximation for task-specific updates. Our approach yields well-calibrated uncertainty estimates and achieves comparable or superior results on standard benchmark datasets. Code is publicly available at \\url{https://github.com/keanson/revisit-logistic-softmax}."
                    }
                ]
            },
            "b533d10e-0ba1-4e43-9827-ba2274d13ea9": {
                "pk": "b533d10e-0ba1-4e43-9827-ba2274d13ea9",
                "name": "Zizhuo Meng",
                "collaborators": [
                    "Zhidong Li",
                    "Yang Wang",
                    "Feng Zhou",
                    "Boyu Li",
                    "Xuhui Fan",
                    "Fang Chen",
                    "Ke Wan",
                    "Yadong Huang"
                ],
                "domain": [
                    "Temporal Point Process",
                    "Social Network Analysis",
                    "Deep Learning",
                    "Interpretability"
                ],
                "publications": [
                    {
                        "title": "TransFeat-TPP: An Interpretable Deep Covariate Temporal Point Processes",
                        "abstract": "The classical temporal point process (TPP) constructs an intensity function by taking the occurrence times into account. Nevertheless, occurrence time may not be the only relevant factor, other contextual data, termed covariates, may also impact the event evolution. Incorporating such covariates into the model is beneficial, while distinguishing their relevance to the event dynamics is of great practical significance. In this work, we propose a Transformer-based covariate temporal point process (TransFeat-TPP) model to improve the interpretability of deep covariate-TPPs while maintaining powerful expressiveness. TransFeat-TPP can effectively model complex relationships between events and covariates, and provide enhanced interpretability by discerning the importance of various covariates. Experimental results on synthetic and real datasets demonstrate improved prediction accuracy and consistently interpretable feature importance when compared to existing deep covariate-TPPs."
                    },
                    {
                        "title": "Interpretable Transformer Hawkes Processes: Unveiling Complex Interactions in Social Networks",
                        "abstract": "Social networks represent complex ecosystems where the interactions between users or groups play a pivotal role in information dissemination, opinion formation, and social interactions. Effectively harnessing event sequence data within social networks to unearth interactions among users or groups has persistently posed a challenging frontier within the realm of point processes. Current deep point process models face inherent limitations within the context of social networks, constraining both their interpretability and expressive power. These models encounter challenges in capturing interactions among users or groups and often rely on parameterized extrapolation methods when modelling intensity over non-event intervals, limiting their capacity to capture intricate intensity patterns, particularly beyond observed events. To address these challenges, this study proposes modifications to Transformer Hawkes processes (THP), leading to the development of interpretable Transformer Hawkes processes (ITHP). ITHP inherits the strengths of THP while aligning with statistical nonlinear Hawkes processes, thereby enhancing its interpretability and providing valuable insights into interactions between users or groups. Additionally, ITHP enhances the flexibility of the intensity function over non-event intervals, making it better suited to capture complex event propagation patterns in social networks. Experimental results, both on synthetic and real data, demonstrate the effectiveness of ITHP in overcoming the identified limitations. Moreover, they highlight ITHP's applicability in the context of exploring the complex impact of users or groups within social networks."
                    }
                ]
            },
            "91c8a8cd-ad15-4f12-8519-554b0cfd52e6": {
                "pk": "91c8a8cd-ad15-4f12-8519-554b0cfd52e6",
                "name": "Tianjun Ke",
                "collaborators": [
                    "Haoqun Cao",
                    "Feng Zhou",
                    "Zenan Ling"
                ],
                "domain": [
                    "Few-shot Learning",
                    "Bayesian Inference",
                    "Gaussian Processes",
                    "Meta-learning"
                ],
                "publications": [
                    {
                        "title": "Accelerating Convergence in Bayesian Few-Shot Classification",
                        "abstract": "Bayesian few-shot classification has been a focal point in the field of few-shot learning. This paper seamlessly integrates mirror descent-based variational inference into Gaussian process-based few-shot classification, addressing the challenge of non-conjugate inference. By leveraging non-Euclidean geometry, mirror descent achieves accelerated convergence by providing the steepest descent direction along the corresponding manifold. It also exhibits the parameterization invariance property concerning the variational distribution. Experimental results demonstrate competitive classification accuracy, improved uncertainty quantification, and faster convergence compared to baseline models. Additionally, we investigate the impact of hyperparameters and components. Code is publicly available at https://github.com/keanson/MD-BSFC."
                    },
                    {
                        "title": "Revisiting Logistic-softmax Likelihood in Bayesian Meta-Learning for Few-Shot Classification",
                        "abstract": "Meta-learning has demonstrated promising results in few-shot classification (FSC) by learning to solve new problems using prior knowledge. Bayesian methods are effective at characterizing uncertainty in FSC, which is crucial in high-risk fields. In this context, the logistic-softmax likelihood is often employed as an alternative to the softmax likelihood in multi-class Gaussian process classification due to its conditional conjugacy property. However, the theoretical property of logistic-softmax is not clear and previous research indicated that the inherent uncertainty of logistic-softmax leads to suboptimal performance. To mitigate these issues, we revisit and redesign the logistic-softmax likelihood, which enables control of the \\textit{a priori} confidence level through a temperature parameter. Furthermore, we theoretically and empirically show that softmax can be viewed as a special case of logistic-softmax and logistic-softmax induces a larger family of data distribution than softmax. Utilizing modified logistic-softmax, we integrate the data augmentation technique into the deep kernel based Gaussian process meta-learning framework, and derive an analytical mean-field approximation for task-specific updates. Our approach yields well-calibrated uncertainty estimates and achieves comparable or superior results on standard benchmark datasets. Code is publicly available at \\url{https://github.com/keanson/revisit-logistic-softmax}."
                    }
                ]
            },
            "a3b74467-a1a6-43b5-8cfb-5c3a19e08ff3": {
                "pk": "a3b74467-a1a6-43b5-8cfb-5c3a19e08ff3",
                "name": "Feng Zhou",
                "collaborators": [
                    "Huyuan Chen",
                    "Huimin Cheng",
                    "Yuanqing Lin",
                    "Quyu Kong",
                    "Yixuan Zhang",
                    "Cheng Feng",
                    "Jun Zhu"
                ],
                "domain": [
                    "Nonlinear Analysis",
                    "Point Processes",
                    "Image Recognition",
                    "Partial Differential Equations"
                ],
                "publications": [
                    {
                        "title": "Besov Regularity of Weak solutions to a Class of Nonlinear Elliptic Equations",
                        "abstract": "In this article, we study a Besov regularity estimate of weak solutions to a class of nonlinear elliptic equations in divergence form. The main purpose is to establish Calderon-Zygmund type estimate in Besov spaces with more general assumptions on coefficients, non-homogeneous term and integrable index. By involving the Sharp maximal function, we establish an oscillation estimate of weak solutions in Orlicz-Sobolev spaces. By deriving a higher integrability estimate of weak solutions, we obtain the desired regularity estimate which expands the Calderon-Zygmund theory for nonlinear elliptic equations."
                    },
                    {
                        "title": "Classification of isolated singularities of positive solutions for Choquard equations",
                        "abstract": "In this paper we classify the isolated singularities of positive solutions to Choquard equation and prove the existence of isolated singular solutions."
                    },
                    {
                        "title": "Fine-grained Image Classification by Exploring Bipartite-Graph Labels",
                        "abstract": "Given a food image, can a fine-grained object recognition engine tell \"which restaurant which dish\" the food belongs to? Such ultra-fine grained image recognition is the key for many applications like search by images, but it is very challenging because it needs to discern subtle difference between classes while dealing with the scarcity of training data. Fortunately, the ultra-fine granularity naturally brings rich relationships among object classes. This paper proposes a novel approach to exploit the rich relationships through bipartite-graph labels (BGL). We show how to model BGL in an overall convolutional neural networks and the resulting system can be optimized through back-propagation. We also show that it is computationally efficient in inference thanks to the bipartite structure. To facilitate the study, we construct a new food benchmark dataset, which consists of 37,885 food images collected from 6 restaurants and totally 975 menus. Experimental results on this new food and three other datasets demonstrates BGL advances previous works in fine-grained object recognition. An online demo is available at http://www.f-zhou.com/fg_demo/."
                    },
                    {
                        "title": "Nonlinear Hawkes Processes in Time-Varying System",
                        "abstract": "Hawkes processes are a class of point processes that have the ability to model the self- and mutual-exciting phenomena. Although the classic Hawkes processes cover a wide range of applications, their expressive ability is limited due to three key hypotheses: parametric, linear and homogeneous. Recent work has attempted to address these limitations separately. This work aims to overcome all three assumptions simultaneously by proposing the flexible state-switching Hawkes processes: a flexible, nonlinear and nonhomogeneous variant where a state process is incorporated to interact with the point processes. The proposed model empowers Hawkes processes to be applied to time-varying systems. For inference, we utilize the latent variable augmentation technique to design two efficient Bayesian inference algorithms: Gibbs sampler and mean-field variational inference, with analytical iterative updates to estimate the posterior. In experiments, our model achieves superior performance compared to the state-of-the-art competitors."
                    },
                    {
                        "title": "Isolated singularities for elliptic equations with Hardy operator and source nonlinearity",
                        "abstract": "In this paper, we concern the isolated singular solutions for semi-linear elliptic equations involving the Hardy-Leray potentials   \\begin{equation}\\label{0}   -\\Delta u+\\frac{\\mu}{|x|^2} u=u^p\\quad {\\rm in}\\quad \\Omega\\setminus\\{0\\},\\qquad u=0\\quad{\\rm on}\\quad \\partial\\Omega.   \\end{equation}   We classify the isolated singularities and obtain the existence, the stability of positive solutions of (\\ref{0}). Our results are based on the study of nonhomogeneous Hardy problem in a new distributional sense."
                    },
                    {
                        "title": "Isolated singularities of positive solutions for Choquard equations in sublinear case",
                        "abstract": "Our purpose of this paper is to study the isolated singularities of positive solutions to Choquard equation in the sublinear case $q \\in (0,1)$   $$\\displaystyle \\ \\ -\\Delta u+ u =I_\\alpha[u^p] u^q\\;\\;   {\\rm in}\\; \\mathbb{R}^N\\setminus\\{0\\},   % [2mm] \\phantom{ }   \\;\\; \\displaystyle \\lim_{|x|\\to+\\infty}u(x)=0, $$ where $p >0, N \\geq 3, \\alpha \\in (0,N)$ and $I_{\\alpha}[u^p](x) = \\int_{\\mathbb{R}^N} \\frac{u^p(y)}{|x-y|^{N-\\alpha}}dy$ is the Riesz potential, which appears as a nonlocal term in the equation. We investigate the nonexistence and existence of isolated singular solutions of Choquard equation under different range of the pair of exponent $(p,q)$. Furthermore, we obtain qualitative properties for the minimal singular solutions of the equation."
                    }
                ]
            }
        }
    },
    "2006.09268": {
        "paper_data": {
            "title": "Metrizing Weak Convergence with Maximum Mean Discrepancies",
            "url": "http://arxiv.org/abs/2006.09268v3",
            "arxiv_id": "2006.09268",
            "authors": [
                "Carl-Johann Simon-Gabriel",
                "Alessandro Barp",
                "Bernhard Sch\u00f6lkopf",
                "Lester Mackey"
            ],
            "abstract": "This paper characterizes the maximum mean discrepancies (MMD) that metrize the weak convergence of probability measures for a wide class of kernels. More precisely, we prove that, on a locally compact, non-compact, Hausdorff space, the MMD of a bounded continuous Borel measurable kernel k, whose reproducing kernel Hilbert space (RKHS) functions vanish at infinity, metrizes the weak convergence of probability measures if and only if k is continuous and integrally strictly positive definite (i.s.p.d.) over all signed, finite, regular Borel measures. We also correct a prior result of Simon-Gabriel & Sch\\\"olkopf (JMLR, 2018, Thm.12) by showing that there exist both bounded continuous i.s.p.d. kernels that do not metrize weak convergence and bounded continuous non-i.s.p.d. kernels that do metrize it.",
            "introduction": " Introduction Although the mathematical and statistical literature has s tudied kernel mean embeddings (KMEs) and maximum mean discrepancies (MMDs) at least since the seventies (Guilbart, 1978), the machine learning community re-discovered and ap plied them only since the late 2000s Smola et al. (2007). A KME with reproducing kernel kis a map from measures \u00b5\u2013 in particular probability distributions \u2013 to functions f\u00b5in the reproducing kernel Hilbert space (RKHS) Hkofk. The RKHS distance between two embeddings then yields a semi -metricdk on measures, called the maximum mean discrepancy (MMD), whi ch can be used to compare two measures or distributions \u00b5and\u03bd:dk(\u00b5,\u03bd) :=/bardblf\u00b5\u2212f\u03bd/bardblk. Their theoretical tractability and computational \ufb02exibil ity has allowed MMDs to \ufb02ourish in many areas of machine learning that require comparing pro bability distributions, such as two-sample testing (compare two discrete distributions Gretton et al. (2012)), sample quality measurement and goodness-of-\ufb01t testing (compare a discrete distribution to a refer- ence distribution, as in Chwialkowski et al. 2016; Liu et al. 2016; Gorham and Mackey 2017; Jitkrittum et al. 2017; Huggins and Mackey 2018), generativ e model \ufb01tting (compare distri- Preprint. Under review. \u00a92021 Simon-Gabriel et al.Simon-Gabriel et al. butions of fake and real data; see Dziugaite et al. 2015; Suth erland et al. 2017; Feng et al. 2017; Pu et al. 2017; Briol et al. 2019), de novo sampling and qu adrature (Chen et al., 2010; Husz\u00e1r and Duvenaud, 2012; Liu and Wang, 2016; Chen et al., 20 18; Futami et al., 2019; Chen et al., 2019), importance sampling (Liu and Lee, 2017; H odgkinson et al., 2020), and thinning (Riabiz et al., 2020). For most applications, one seeks a kernel kwhose MMD can separate all probability distributions P,Q, meaning that, dk(P,Q) = 0 (if and) only if Q=P. Such kernels are said to becharacteristic (to the set of probability distributions P). If for example we optimize a parametric distribution Qto match a target Pby minimizing their MMD dk(P,Q), it is rather natural to require that it be minimized only if Qperfectly matches P, i.e.Q=P. Another natural, but a priori stronger requirement, is that whenQgets closer to Pin MMD, for example, if dk(Q,P)\u21920, we would like Qto \u201ctruly\u201d converge to P, where \u201ctruly\u201d means \u201cfor some other standard and/or more familiar notion of conv ergence\u201d. Although several standard notions may come to mind \u2013 converg ence in KL-divergence, in total variation or in Hellinger distance \u2013, many are too st rong for our purposes which often require handling discrete data. For example, even if x\u2192\u03be, the Dirac masses \u03b4x will not converge to \u03b4\u03bein total variation or KL-divergence unless xis eventually equal to\u03be. Said di\ufb00erently, a sequence of deterministic variables wo uld not converge in total variation unless it was eventually constant. Since in pract ice MMDs are frequently used to compare samples or empirical (hence discrete) distributio ns, it comes as no surprise that MMD convergence cannot, in general, ensure these strong typ es of convergence. Instead we will opt for a standard, yet comparatively weak notion of con vergence, known as weak or narrow convergence orconvergence in distribution . Speci\ufb01cally, the central question of this paper will be When is convergence in MMD metric equivalent to weak converg ence onP? In that case, we will say that the kernel kmetrizes the weak convergence of probability measures . This question",
            "references": [
                {
                    "title": "Optimal thinning of MCMC output",
                    "abstract": "The use of heuristics to assess the convergence and compress the output of Markov chain Monte Carlo can be sub\u2010optimal in terms of the empirical approximations that are produced. Typically a number of the initial states are attributed to \u2018burn in\u2019 and removed, while the remainder of the chain is \u2018thinned\u2019 if compression is also required. In this paper, we consider the problem of retrospectively selecting a subset of states, of fixed cardinality, from the sample path such that the approximation provided by their empirical distribution is close to optimal. A novel method is proposed, based on greedy minimisation of a kernel Stein discrepancy, that is suitable when the gradient of the log\u2010target can be evaluated and approximation using a small number of states is required. Theoretical results guarantee consistency of the method and its effectiveness is demonstrated in the challenging context of parameter inference for ordinary differential equations. Software is available in the Stein Thinning package in Python, R and MATLAB."
                },
                {
                    "title": "The reproducing Stein kernel approach for post-hoc corrected sampling",
                    "abstract": "Stein importance sampling is a widely applicable technique based on kernelized Stein discrepancy, which corrects the output of approximate sampling algorithms by reweighting the empirical distribution of the samples. A general analysis of this technique is conducted for the previously unconsidered setting where samples are obtained via the simulation of a Markov chain, and applies to an arbitrary underlying Polish space. We prove that Stein importance sampling yields consistent estimators for quantities related to a target distribution of interest by using samples obtained from a geometrically ergodic Markov chain with a possibly unknown invariant measure that differs from the desired target. The approach is shown to be valid under conditions that are satisfied for a large number of unadjusted samplers, and is capable of retaining consistency when data subsampling is used. Along the way, a universal theory of reproducing Stein kernels is established, which enables the construction of kernelized Stein discrepancy on general Polish spaces, and provides sufficient conditions for kernels to be convergence-determining on such spaces. These results are of independent interest for the development of future methodology based on kernelized Stein discrepancies."
                },
                {
                    "title": "A Characteristic Function Approach to Deep Implicit Generative Modeling",
                    "abstract": "Implicit Generative Models (IGMs) such as GANs have emerged as effective data-driven models for generating samples, particularly images. In this paper, we formulate the problem of learning an IGM as minimizing the expected distance between characteristic functions. Specifically, we minimize the distance between characteristic functions of the real and generated data distributions under a suitably-chosen weighting distribution. This distance metric, which we term as the characteristic function distance (CFD), can be (approximately) computed with linear time-complexity in the number of samples, in contrast with the quadratic-time Maximum Mean Discrepancy (MMD). By replacing the discrepancy measure in the critic of a GAN with the CFD, we obtain a model that is simple to implement and stable to train. The proposed metric enjoys desirable theoretical properties including continuity and differentiability with respect to generator parameters, and continuity in the weak topology. We further propose a variation of the CFD in which the weighting distribution parameters are also optimized during training; this obviates the need for manual tuning, and leads to an improvement in test power relative to CFD. We demonstrate experimentally that our proposed method outperforms WGAN and MMD-GAN variants on a variety of unsupervised image generation benchmarks."
                },
                {
                    "title": "Asymptotically Optimal One- and Two-Sample Testing With Kernels",
                    "abstract": "We characterize the asymptotic performance of nonparametric one- and two-sample testing. The exponential decay rate or error exponent of the type-II error probability is used as the asymptotic performance metric, and an optimal test achieves the maximum rate subject to a constant level constraint on the type-I error probability. With Sanov\u2019s theorem, we derive a sufficient condition for one-sample tests to achieve the optimal error exponent in the universal setting, i.e., for any distribution defining the alternative hypothesis. We then show that two classes of Maximum Mean Discrepancy (MMD) based tests attain the optimal type-II error exponent on $\\mathbb R^{d}$ , while the quadratic-time Kernel Stein Discrepancy (KSD) based tests achieve this optimality with an asymptotic level constraint. For general two-sample testing, however, Sanov\u2019s theorem is insufficient to obtain a similar sufficient condition. We proceed to establish an extended version of Sanov\u2019s theorem and derive an exact error exponent for the quadratic-time MMD based two-sample tests. The obtained error exponent is further shown to be optimal among all two-sample tests satisfying a given level constraint. Our work hence provides an achievability result for optimal nonparametric one- and two-sample testing in the universal setting. Application to off-line change detection and related issues are also discussed."
                },
                {
                    "title": "Statistical Inference for Generative Models with Maximum Mean Discrepancy",
                    "abstract": "While likelihood-based inference and its variants provide a statistically efficient and widely applicable approach to parametric inference, their application to models involving intractable likelihoods poses challenges. In this work, we study a class of minimum distance estimators for intractable generative models, that is, statistical models for which the likelihood is intractable, but simulation is cheap. The distance considered, maximum mean discrepancy (MMD), is defined through the embedding of probability measures into a reproducing kernel Hilbert space. We study the theoretical properties of these estimators, showing that they are consistent, asymptotically normal and robust to model misspecification. A main advantage of these estimators is the flexibility offered by the choice of kernel, which can be used to trade-off statistical efficiency and robustness. On the algorithmic side, we study the geometry induced by MMD on the parameter space and use this to introduce a novel natural gradient descent-like algorithm for efficient implementation of these estimators. We illustrate the relevance of our theoretical results on several classes of models including a discrete-time latent Markov process and two multivariate stochastic differential equation models."
                },
                {
                    "title": "Stein Point Markov Chain Monte Carlo",
                    "abstract": "An important task in machine learning and statistics is the approximation of a probability measure by an empirical measure supported on a discrete point set. Stein Points are a class of algorithms for this task, which proceed by sequentially minimising a Stein discrepancy between the empirical measure and the target and, hence, require the solution of a non-convex optimisation problem to obtain each new point. This paper removes the need to solve this optimisation problem by, instead, selecting each new point based on a Markov chain sample path. This significantly reduces the computational cost of Stein Points and leads to a suite of algorithms that are straightforward to implement. The new algorithms are illustrated on a set of challenging Bayesian inference problems, and rigorous theoretical guarantees of consistency are established."
                },
                {
                    "title": "Signature Moments to Characterize Laws of Stochastic Processes",
                    "abstract": "The normalized sequence of moments characterizes the law of any finite-dimensional random variable. We prove an analogous result for path-valued random variables, that is stochastic processes, by using the normalized sequence of signature moments. We use this to define a metric for laws of stochastic processes. This metric can be efficiently estimated from finite samples, even if the stochastic processes themselves evolve in high-dimensional state spaces. As an application, we provide a non-parametric two-sample hypothesis test for laws of stochastic processes."
                },
                {
                    "title": "Random Feature Stein Discrepancies",
                    "abstract": "Computable Stein discrepancies have been deployed for a variety of applications, ranging from sampler selection in posterior inference to approximate Bayesian inference to goodness-of-fit testing. Existing convergence-determining Stein discrepancies admit strong theoretical guarantees but suffer from a computational cost that grows quadratically in the sample size. While linear-time Stein discrepancies have been proposed for goodness-of-fit testing, they exhibit avoidable degradations in testing power\u2014even when power is explicitly optimized. To address these shortcomings, we introduce feature Stein discrepancies (\u03a6SDs), a new family of quality measures that can be cheaply approximated using importance sampling. We show how to construct \u03a6SDs that provably determine the convergence of a sample to its target and develop high-accuracy approximations\u2014random \u03a6SDs (R\u03a6SDs)\u2014which are computable in near-linear time. In our experiments with sampler selection for approximate posterior inference and goodness-of-fit testing, R\u03a6SDs perform as well or better than quadratic-time KSDs while being orders of magnitude faster to compute."
                },
                {
                    "title": "Bayesian Posterior Approximation via Greedy Particle Optimization",
                    "abstract": "In Bayesian inference, the posterior distributions are difficult to obtain analytically for complex models such as neural networks. Variational inference usually uses a parametric distribution for approximation, from which we can easily draw samples. Recently discrete approximation by particles has attracted attention because of its high expression ability. An example is Stein variational gradient descent (SVGD), which iteratively optimizes particles. Although SVGD has been shown to be computationally efficient empirically, its theoretical properties have not been clarified yet and no finite sample bound of the convergence rate is known. Another example is the Stein points (SP) method, which minimizes kernelized Stein discrepancy directly. Althoughafinitesampleboundisassuredtheoretically, SP is computationally inefficient empirically, especially in high-dimensional problems. In this paper, we propose a novel method named maximum mean discrepancy minimization by the Frank-Wolfe algorithm (MMD-FW), which minimizes MMD in a greedy way by the FW algorithm. Our method is computationally efficient empirically and we show that its finite sample convergence bound is in a linear order in finite dimensions."
                },
                {
                    "title": "Stein Points",
                    "abstract": "An important task in computational statistics and machine learning is to approximate a posterior distribution $p(x)$ with an empirical measure supported on a set of representative points $\\{x_i\\}_{i=1}^n$. This paper focuses on methods where the selection of points is essentially deterministic, with an emphasis on achieving accurate approximation when $n$ is small. To this end, we present `Stein Points'. The idea is to exploit either a greedy or a conditional gradient method to iteratively minimise a kernel Stein discrepancy between the empirical measure and $p(x)$. Our empirical results demonstrate that Stein Points enable accurate approximation of the posterior at modest computational cost. In addition, theoretical results are provided to establish convergence of the method."
                },
                {
                    "title": "Universal Hypothesis Testing with Kernels: Asymptotically Optimal Tests for Goodness of Fit",
                    "abstract": "We characterize the asymptotic performance of nonparametric goodness of fit testing. The exponential decay rate of the type-II error probability is used as the asymptotic performance metric, and a test is optimal if it achieves the maximum rate subject to a constant level constraint on the type-I error probability. We show that two classes of Maximum Mean Discrepancy (MMD) based tests attain this optimality on $\\mathbb R^d$, while the quadratic-time Kernel Stein Discrepancy (KSD) based tests achieve the maximum exponential decay rate under a relaxed level constraint. Under the same performance metric, we proceed to show that the quadratic-time MMD based two-sample tests are also optimal for general two-sample problems, provided that kernels are bounded continuous and characteristic. Key to our approach are Sanov's theorem from large deviation theory and the weak metrizable properties of the MMD and KSD."
                },
                {
                    "title": "Learning to Draw Samples with Amortized Stein Variational Gradient Descent",
                    "abstract": "We propose a simple algorithm to train stochastic neural networks to draw samples from given target distributions for probabilistic inference. Our method is based on iteratively adjusting the neural network parameters so that the output changes along a Stein variational gradient direction (Liu & Wang, 2016) that maximally decreases the KL divergence with the target distribution. Our method works for any target distribution specified by their unnormalized density function, and can train any black-box architectures that are differentiable in terms of the parameters we want to adapt. We demonstrate our method with a number of applications, including variational autoencoder (VAE) with expressive encoders to model complex latent space structures, and hyper-parameter learning of MCMC samplers that allows Bayesian inference to adaptively improve itself when seeing more data."
                },
                {
                    "title": "MMD GAN: Towards Deeper Understanding of Moment Matching Network",
                    "abstract": "Generative moment matching network (GMMN) is a deep generative model that differs from Generative Adversarial Network (GAN) by replacing the discriminator in GAN with a two-sample test based on kernel maximum mean discrepancy (MMD). Although some theoretical guarantees of MMD have been studied, the empirical performance of GMMN is still not as competitive as that of GAN on challenging and large benchmark datasets. The computational efficiency of GMMN is also less desirable in comparison with GAN, partially due to its requirement for a rather large batch size during the training. In this paper, we propose to improve both the model expressiveness of GMMN and its computational efficiency by introducing adversarial kernel learning techniques, as the replacement of a fixed Gaussian kernel in the original GMMN. The new approach combines the key ideas in both GMMN and GAN, hence we name it MMD GAN. The new distance measure in MMD GAN is a meaningful loss that enjoys the advantage of weak topology and can be optimized via gradient descent with relatively small batch sizes. In our evaluation on multiple benchmark datasets, including MNIST, CIFAR- 10, CelebA and LSUN, the performance of MMD-GAN significantly outperforms GMMN, and is competitive with other representative GAN works."
                },
                {
                    "title": "A Linear-Time Kernel Goodness-of-Fit Test",
                    "abstract": "We propose a novel adaptive test of goodness-of-fit, with computational cost linear in the number of samples. We learn the test features that best indicate the differences between observed samples and a reference model, by minimizing the false negative rate. These features are constructed via Stein's method, meaning that it is not necessary to compute the normalising constant of the model. We analyse the asymptotic Bahadur efficiency of the new test, and prove that under a mean-shift alternative, our test always has greater relative efficiency than a previous linear-time kernel test, regardless of the choice of parameters for that test. In experiments, the performance of our method exceeds that of the earlier linear-time test, and matches or exceeds the power of a quadratic-time kernel test. In high dimensions and where model structure may be exploited, our goodness of fit test performs far better than a quadratic-time two-sample test based on the Maximum Mean Discrepancy, with samples drawn from the model."
                },
                {
                    "title": "VAE Learning via Stein Variational Gradient Descent",
                    "abstract": "A new method for learning variational autoencoders (VAEs) is developed, based on Stein variational gradient descent. A key advantage of this approach is that one need not make parametric assumptions about the form of the encoder distribution. Performance is further enhanced by integrating the proposed encoder with importance sampling. Excellent performance is demonstrated across multiple unsupervised and semi-supervised problems, including semi-supervised analysis of the ImageNet data, demonstrating the scalability of the model to large datasets."
                },
                {
                    "title": "Measuring Sample Quality with Kernels",
                    "abstract": "Approximate Markov chain Monte Carlo (MCMC) offers the promise of more rapid sampling at the cost of more biased inference. Since standard MCMC diagnostics fail to detect these biases, researchers have developed computable Stein discrepancy measures that provably determine the convergence of a sample to its target distribution. This approach was recently combined with the theory of reproducing kernels to define a closed-form kernel Stein discrepancy (KSD) computable by summing kernel evaluations across pairs of sample points. We develop a theory of weak convergence for KSDs based on Stein's method, demonstrate that commonly used KSDs fail to detect non-convergence even for Gaussian targets, and show that kernels with slowly decaying tails provably determine convergence for a large class of target distributions. The resulting convergence-determining KSDs are suitable for comparing biased, exact, and deterministic sample sequences and simpler to compute and parallelize than alternative Stein discrepancies. We use our tools to compare biased samplers, select sampler hyperparameters, and improve upon existing KSD approaches to one-sample hypothesis testing and sample quality improvement."
                },
                {
                    "title": "Generative Models and Model Criticism via Optimized Maximum Mean Discrepancy",
                    "abstract": "We propose a method to optimize the representation and distinguishability of samples from two probability distributions, by maximizing the estimated power of a statistical test based on the maximum mean discrepancy (MMD). This optimized MMD is applied to the setting of unsupervised learning by generative adversarial networks (GAN), in which a model attempts to generate realistic samples, and a discriminator attempts to tell these apart from data samples. In this context, the MMD may be used in two roles: first, as a discriminator, either directly on the samples, or on features of the samples. Second, the MMD can be used to evaluate the performance of a generative model, by testing the model's samples against a reference data set. In the latter role, the optimized MMD is particularly helpful, as it gives an interpretable indication of how the model and data distributions differ, even in cases where individual model samples are not easily distinguished either by eye or by classifier."
                },
                {
                    "title": "Black-box Importance Sampling",
                    "abstract": "Importance sampling is widely used in machine learning and statistics, but its power is limited by the restriction of using simple proposals for which the importance weights can be tractably calculated. We address this problem by studying black-box importance sampling methods that calculate importance weights for samples generated from any unknown proposal or black-box mechanism. Our method allows us to use better and richer proposals to solve difficult problems, and (somewhat counter-intuitively) also has the additional benefit of improving the estimation accuracy beyond typical importance sampling. Both theoretical and empirical analyses are provided."
                },
                {
                    "title": "Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithm",
                    "abstract": "We propose a general purpose variational inference algorithm that forms a natural counterpart of gradient descent for optimization. Our method iteratively transports a set of particles to match the target distribution, by applying a form of functional gradient descent that minimizes the KL divergence. Empirical studies are performed on various real world models and datasets, on which our method is competitive with existing state-of-the-art methods. The derivation of our method is based on a new theoretical result that connects the derivative of KL divergence under smooth transforms with Stein's identity and a recently proposed kernelized Stein discrepancy, which is of independent interest."
                },
                {
                    "title": "Kernel Distribution Embeddings: Universal Kernels, Characteristic Kernels and Kernel Metrics on Distributions",
                    "abstract": "Kernel mean embeddings have recently attracted the attention of the machine learning community. They map measures $\\mu$ from some set $M$ to functions in a reproducing kernel Hilbert space (RKHS) with kernel $k$. The RKHS distance of two mapped measures is a semi-metric $d_k$ over $M$. We study three questions. (I) For a given kernel, what sets $M$ can be embedded? (II) When is the embedding injective over $M$ (in which case $d_k$ is a metric)? (III) How does the $d_k$-induced topology compare to other topologies on $M$? The existing machine learning literature has addressed these questions in cases where $M$ is (a subset of) the finite regular Borel measures. We unify, improve and generalise those results. Our approach naturally leads to continuous and possibly even injective embeddings of (Schwartz-) distributions, i.e., generalised measures, but the reader is free to focus on measures only. In particular, we systemise and extend various (partly known) equivalences between different notions of universal, characteristic and strictly positive definite kernels, and show that on an underlying locally compact Hausdorff space, $d_k$ metrises the weak convergence of probability measures if and only if $k$ is continuous and characteristic."
                },
                {
                    "title": "A Kernelized Stein Discrepancy for Goodness-of-fit Tests",
                    "abstract": "We derive a new discrepancy statistic for measuring differences between two probability distributions based on combining Stein's identity with the reproducing kernel Hilbert space theory. We apply our result to test how well a probabilistic model fits a set of observations, and derive a new class of powerful goodness-of-fit tests that are widely applicable for complex and high dimensional distributions, even for those with computationally intractable normalization constants. Both theoretical and empirical properties of our methods are studied thoroughly."
                },
                {
                    "title": "A Kernel Test of Goodness of Fit",
                    "abstract": "We propose a nonparametric statistical test for goodness-of-fit: given a set of samples, the test determines how likely it is that these were generated from a target density function. The measure of goodness-of-fit is a divergence constructed via Stein's method using functions from a Reproducing Kernel Hilbert Space. Our test statistic is based on an empirical estimate of this divergence, taking the form of a V-statistic in terms of the log gradients of the target density and the kernel. We derive a statistical test, both for i.i.d. and non-i.i.d. samples, where we estimate the null distribution quantiles using a wild bootstrap procedure. We apply our test to quantifying convergence of approximate Markov Chain Monte Carlo methods, statistical model criticism, and evaluating quality of fit vs model complexity in nonparametric density estimation."
                },
                {
                    "title": "Training generative neural networks via Maximum Mean Discrepancy optimization",
                    "abstract": "We consider training a deep neural network to generate samples from an unknown distribution given i.i.d. data. We frame learning as an optimization minimizing a two-sample test statistic\u2014informally speaking, a good generator network produces samples that cause a two-sample test to fail to reject the null hypothesis. As our two-sample test statistic, we use an unbiased estimate of the maximum mean discrepancy, which is the centerpiece of the nonparametric kernel two-sample test proposed by Gretton et al. [2]. We compare to the adversarial nets framework introduced by Goodfellow et al. [1], in which learning is a two-player game between a generator network and an adversarial discriminator network, both trained to outwit the other. From this perspective, the MMD statistic plays the role of the discriminator. In addition to empirical comparisons, we prove bounds on the generalization error incurred by optimizing the empirical MMD."
                },
                {
                    "title": "On the optimal estimation of probability measures in weak and strong topologies",
                    "abstract": "Given random samples drawn i.i.d. from a probability measure $\\mathbb{P}$ (defined on say, $\\mathbb{R}^d$), it is well-known that the empirical estimator is an optimal estimator of $\\mathbb{P}$ in weak topology but not even a consistent estimator of its density (if it exists) in the strong topology (induced by the total variation distance). On the other hand, various popular density estimators such as kernel and wavelet density estimators are optimal in the strong topology in the sense of achieving the minimax rate over all estimators for a Sobolev ball of densities. Recently, it has been shown in a series of papers by Gin\\'{e} and Nickl that these density estimators on $\\mathbb{R}$ that are optimal in strong topology are also optimal in $\\|\\cdot\\|_{\\mathcal{F}}$ for certain choices of $\\mathcal{F}$ such that $\\|\\cdot\\|_{\\mathcal{F}}$ metrizes the weak topology, where $\\|\\mathbb{P}\\|_{\\mathcal{F}}:=\\sup\\{\\int f\\,\\mathrm{d}\\mathbb{P}: f\\in\\mathcal{F}\\}$. In this paper, we investigate this problem of optimal estimation in weak and strong topologies by choosing $\\mathcal{F}$ to be a unit ball in a reproducing kernel Hilbert space (say $\\mathcal{F}_H$ defined over $\\mathbb{R}^d$), where this choice is both of theoretical and computational interest. Under some mild conditions on the reproducing kernel, we show that $\\|\\cdot\\|_{\\mathcal{F}_H}$ metrizes the weak topology and the kernel density estimator (with $L^1$ optimal bandwidth) estimates $\\mathbb{P}$ at dimension independent optimal rate of $n^{-1/2}$ in $\\|\\cdot\\|_{\\mathcal{F}_H}$ along with providing a uniform central limit theorem for the kernel density estimator."
                },
                {
                    "title": "Optimally-Weighted Herding is Bayesian Quadrature",
                    "abstract": "Herding and kernel herding are deterministic methods of choosing samples which summarise a probability distribution. A related task is choosing samples for estimating integrals using Bayesian quadrature. We show that the criterion minimised when selecting samples in kernel herding is equivalent to the posterior variance in Bayesian quadrature. We then show that sequential Bayesian quadrature can be viewed as a weighted version of kernel herding which achieves performance superior to any other weighted herding method. We demonstrate empirically a rate of convergence faster than O(1/N). Our results also imply an upper bound on the empirical error of the Bayesian quadrature estimate."
                },
                {
                    "title": "A Kernel Two-Sample Test",
                    "abstract": "We propose a framework for analyzing and comparing distributions, which we use to construct statistical tests to determine if two samples are drawn from different distributions. Our test statistic is the largest difference in expectations over functions in the unit ball of a reproducing kernel Hilbert space (RKHS), and is called the maximum mean discrepancy (MMD).We present two distribution free tests based on large deviation bounds for the MMD, and a third test based on the asymptotic distribution of this statistic. The MMD can be computed in quadratic time, although efficient linear time approximations are available. Our statistic is an instance of an integral probability metric, and various classical metrics on distributions are obtained when alternative function classes are used in place of an RKHS. We apply our two-sample tests to a variety of problems, including attribute matching for databases using the Hungarian marriage method, where they perform strongly. Excellent performance is also obtained when comparing distributions over graphs, for which these are the first such tests."
                },
                {
                    "title": "Super-Samples from Kernel Herding",
                    "abstract": "We extend the herding algorithm to continuous spaces by using the kernel trick. The resulting \"kernel herding\" algorithm is an infinite memory deterministic process that learns to approximate a PDF with a collection of samples. We show that kernel herding decreases the error of expectations of functions in the Hilbert space at a rate O(1/T) which is much faster than the usual O(1/pT) for iid random samples. We illustrate kernel herding by approximating Bayesian predictive distributions."
                },
                {
                    "title": "Hilbert Space Embeddings and Metrics on Probability Measures",
                    "abstract": "A Hilbert space embedding for probability measures has recently been proposed, with applications including dimensionality reduction, homogeneity testing, and independence testing. This embedding represents any probability measure as a mean element in a reproducing kernel Hilbert space (RKHS). A pseudometric on the space of probability measures can be defined as the distance between distribution embeddings: we denote this as \u03b3k, indexed by the kernel function k that defines the inner product in the RKHS. \n \nWe present three theoretical properties of \u03b3k. First, we consider the question of determining the conditions on the kernel k for which \u03b3k is a metric: such k are denoted characteristic kernels. Unlike pseudometrics, a metric is zero only when two distributions coincide, thus ensuring the RKHS embedding maps all distributions uniquely (i.e., the embedding is injective). While previously published conditions may apply only in restricted circumstances (e.g., on compact domains), and are difficult to check, our conditions are straightforward and intuitive: integrally strictly positive definite kernels are characteristic. Alternatively, if a bounded continuous kernel is translation-invariant on \u211cd, then it is characteristic if and only if the support of its Fourier transform is the entire \u211cd. Second, we show that the distance between distributions under \u03b3k results from an interplay between the properties of the kernel and the distributions, by demonstrating that distributions are close in the embedding space when their differences occur at higher frequencies. Third, to understand the nature of the topology induced by \u03b3k, we relate \u03b3k to other popular metrics on probability measures, and present conditions on the kernel k under which \u03b3k metrizes the weak topology."
                },
                {
                    "title": "Support vector machines",
                    "abstract": "Support vector machines (SVMs) are a family of machine learning methods, originally introduced for the problem of classification and later generalized to various other situations. They are based on principles of statistical learning theory and convex optimization, and are currently used in various domains of application, including bioinformatics, text categorization, and computer vision. Copyright \u00a9 2009 John Wiley & Sons, Inc."
                },
                {
                    "title": "A Course In Functional Analysis",
                    "abstract": "Thank you very much for downloading a course in functional analysis. As you may know, people have look numerous times for their favorite books like this a course in functional analysis, but end up in harmful downloads. Rather than reading a good book with a cup of coffee in the afternoon, instead they juggled with some harmful virus inside their desktop computer. a course in functional analysis is available in our book collection an online access to it is set as public so you can get it instantly. Our book servers spans in multiple countries, allowing you to get the most less latency time to download any of our books like this one. Merely said, the a course in functional analysis is universally compatible with any devices to read."
                },
                {
                    "title": "Topics In Banach Space Integration",
                    "abstract": "* Bochner Integral * Dunford and Pettis Integrals * McShane and Henstock-Kurzweil Integrals * More on the McShane Integral * Comparison of the Bochner and McShane Integrals * Comparison of the Pettis and McShane Integrals * Primitive of the McShane and Henstock-Kurzweil Integrals * Generalizations of Some Integrals"
                },
                {
                    "title": "Infinite Dimensional Analysis: A Hitchhiker\u2019s Guide",
                    "abstract": "This text was born out of an advanced mathematical economics seminar at Caltech in 1989-90. We realized that the typical graduate student in mathematical economics has to be familiar with a vast amount of material that spans several traditional fields in mathematics. Much of the mate\u00acrial appears only in esoteric research monographs that are designed for specialists, not for the sort of generalist that our students need be. We hope that in a small way this text will make the material here accessible to a much broader audience. While our motivation is to present and orga\u00acnize the analytical foundations underlying modern economics and finance, this is a book of mathematics, not of economics. We mention applications to economics but present very few of them. They are there to convince economists that the material has so me relevance and to let mathematicians know that there are areas of application for these results. We feel that this text could be used for a course in analysis that would benefit mathematicians, engineers, and scientists. Most of the material we present is available elsewhere, but is scattered throughout a variety of sources and occasionally buried in obscurity. Some of our results are original (or more likely, independent rediscoveries). We have included some material that we cannot honestly say is neces\u00acsary to understand modern economic theory, but may yet prove useful in future research."
                },
                {
                    "title": "Harmonic Analysis on Semigroups: Theory of Positive Definite and Related Functions",
                    "abstract": "1 Introduction to Locally Convex Topological Vector Spaces and Dual Pairs.- 1. Locally Convex Vector Spaces.- 2. Hahn-Banach Theorems.- 3. Dual Pairs.- Notes and Remarks.- 2 Radon Measures and Integral Representations.- 1. Introduction to Radon Measures on Hausdorff Spaces.- 2. The Riesz Representation Theorem.- 3. Weak Convergence of Finite Radon Measures.- 4. Vague Convergence of Radon Measures on Locally Compact Spaces.- 5. Introduction to the Theory of Integral Representations.- Notes and Remarks.- 3 General Results on Positive and Negative Definite Matrices and Kernels.- 1. Definitions and Some Simple Properties of Positive and Negative Definite Kernels.- 2. Relations Between Positive and Negative Definite Kernels.- 3. Hubert Space Representation of Positive and Negative Definite Kernels.- Notes and Remarks.- 4 Main Results on Positive and Negative Definite Functions on Semigroups.- 1. Definitions and Simple Properties 86 2. Exponentially Bounded Positive Definite Functions on Abelian Semigroups.- 3. Negative Definite Functions on Abelian Semigroups.- 4. Examples of Positive and Negative Definite Functions.- 5. T-Positive Functions.- 6. Completely Monotone and Alternating Functions.- Notes and Remarks.- 5 Schoenberg-Type Results for Positive and Negative Definite Functions.- 1. Schoenberg Triples.- 2. Norm Dependent Positive Definite Functions on Banach Spaces.- 3. Functions Operating on Positive Definite Matrices.- 4. Schoenberg's Theorem for the Complex Hilbert Sphere.- 5. The Real Infinite Dimensional Hyperbolic Space.- Notes and Remarks.- 6 Positive Definite Functions and Moment Functions.- 1. Moment Functions.- 2. The One-Dimensional Moment Problem.- 3. The Multi-Dimensional Moment Problem.- 4. The Two-Sided Moment Problem.- 5. Perfect Semigroups.- Notes and Remarks.- 7 Hoeffding's Inequality and Multivariate Majorization.- 1. The Discrete Case.- 2. Extension to Nondiscrete Semigroups.- 3. Completely Negative Definite Functions and Schur-Monotonicity.- Notes and Remarks.- 8 Positive and Negative Definite Functions on Abelian Semigroups Without Zero.- 1. Quasibounded Positive and Negative Definite Functions.- 2. Completely Monotone and Completely Alternating Functions.- Notes and Remarks.- References.- List of Symbols."
                }
            ],
            "categories": [
                "cs.LG",
                "math.PR",
                "math.ST",
                "stat.ML",
                "stat.TH",
                "60B10 (Primary) 60F05, 60-08, 28-08 (Secondary)",
                "G.3; I.2.6; I.5.0"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nWhen is convergence in the maximum mean discrepancy (MMD) metric equivalent to weak convergence of probability measures?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it will deepen the understanding of the relationship between MMD and weak convergence, which is fundamental in probability theory and statistics. By establishing conditions under which MMD can serve as a reliable metric for weak convergence, this research could enhance the applicability of MMD in various machine learning tasks, such as generative modeling and goodness-of-fit testing. This advancement could lead to more robust algorithms for comparing probability distributions, ultimately influencing future research directions and practical applications in areas like statistical inference and machine learning.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent complexities of probability measures and their convergence properties. Naive approaches may fail because MMD does not guarantee strong convergence types, such as convergence in total variation or KL-divergence, especially when dealing with discrete data. The technical obstacles include identifying the right conditions under which MMD can effectively capture weak convergence, as well as the need to reconcile the differences between various convergence notions. Theoretical intricacies in the behavior of kernels and their embeddings further complicate the analysis.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the applications of MMD without thoroughly investigating its theoretical foundations regarding convergence. Existing solutions often overlook the specific conditions required for MMD to metrize weak convergence, leading to gaps in understanding. Barriers include a lack of comprehensive frameworks that connect MMD with weak convergence in a rigorous manner. This research aims to fill these gaps by providing a clear theoretical framework that delineates the relationship between MMD and weak convergence, improving upon prior work that has not addressed this specific question.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves a theoretical analysis of the conditions under which MMD metrizes weak convergence. This will include the use of specific kernels and probability distributions, along with a detailed examination of their properties. The analysis will be supported by mathematical proofs and potentially empirical validation using synthetic datasets to illustrate the findings. The expected outcome is a set of criteria that clearly delineates when MMD can be used as a reliable metric for weak convergence, thereby providing a significant contribution to the understanding of probability measures in machine learning contexts."
            }
        },
        "author_data": {
            "5f2f77c7-43c0-49e1-a17e-e032f1804b92": {
                "pk": "5f2f77c7-43c0-49e1-a17e-e032f1804b92",
                "name": "Carl-Johann Simon-Gabriel",
                "collaborators": [
                    "Bernhard Sch\u00f6lkopf",
                    "Yann Ollivier",
                    "L\u00e9on Bottou",
                    "David Lopez-Paz",
                    "Noman Ahmed Sheikh",
                    "Andreas Krause"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Kernel Methods",
                    "Probabilistic Models"
                ],
                "publications": [
                    {
                        "title": "First-order Adversarial Vulnerability of Neural Networks and Input Dimension",
                        "abstract": "Over the past few years, neural networks were proven vulnerable to adversarial images: targeted but imperceptible image perturbations lead to drastically different predictions. We show that adversarial vulnerability increases with the gradients of the training objective when viewed as a function of the inputs. Surprisingly, vulnerability does not depend on network topology: for many standard network architectures, we prove that at initialization, the $\\ell_1$-norm of these gradients grows as the square root of the input dimension, leaving the networks increasingly vulnerable with growing image size. We empirically show that this dimension dependence persists after either usual or robust training, but gets attenuated with higher regularization."
                    },
                    {
                        "title": "PopSkipJump: Decision-Based Attack for Probabilistic Classifiers",
                        "abstract": "Most current classifiers are vulnerable to adversarial examples, small input perturbations that change the classification output. Many existing attack algorithms cover various settings, from white-box to black-box classifiers, but typically assume that the answers are deterministic and often fail when they are not. We therefore propose a new adversarial decision-based attack specifically designed for classifiers with probabilistic outputs. It is based on the HopSkipJump attack by Chen et al. (2019, arXiv:1904.02144v5 ), a strong and query efficient decision-based attack originally designed for deterministic classifiers. Our P(robabilisticH)opSkipJump attack adapts its amount of queries to maintain HopSkipJump's original output quality across various noise levels, while converging to its query efficiency as the noise level decreases. We test our attack on various noise models, including state-of-the-art off-the-shelf randomized defenses, and show that they offer almost no extra robustness to decision-based attacks. Code is available at https://github.com/cjsg/PopSkipJump ."
                    },
                    {
                        "title": "Kernel Distribution Embeddings: Universal Kernels, Characteristic Kernels and Kernel Metrics on Distributions",
                        "abstract": "Kernel mean embeddings have recently attracted the attention of the machine learning community. They map measures $\\mu$ from some set $M$ to functions in a reproducing kernel Hilbert space (RKHS) with kernel $k$. The RKHS distance of two mapped measures is a semi-metric $d_k$ over $M$. We study three questions. (I) For a given kernel, what sets $M$ can be embedded? (II) When is the embedding injective over $M$ (in which case $d_k$ is a metric)? (III) How does the $d_k$-induced topology compare to other topologies on $M$? The existing machine learning literature has addressed these questions in cases where $M$ is (a subset of) the finite regular Borel measures. We unify, improve and generalise those results. Our approach naturally leads to continuous and possibly even injective embeddings of (Schwartz-) distributions, i.e., generalised measures, but the reader is free to focus on measures only. In particular, we systemise and extend various (partly known) equivalences between different notions of universal, characteristic and strictly positive definite kernels, and show that on an underlying locally compact Hausdorff space, $d_k$ metrises the weak convergence of probability measures if and only if $k$ is continuous and characteristic."
                    }
                ]
            },
            "0c850579-1c2a-4d03-83e4-0073381b1056": {
                "pk": "0c850579-1c2a-4d03-83e4-0073381b1056",
                "name": "Alessandro Barp",
                "collaborators": [
                    "Mark Girolami",
                    "Francois-Xavier Briol",
                    "Lester Mackey",
                    "Chris. J. Oates",
                    "Alexis Arnaudon",
                    "So Takao",
                    "Andrew B. Duncan",
                    "Fran\u00e7ois-Xavier Briol",
                    "Guilherme Fran\u00e7a",
                    "Michael I. Jordan"
                ],
                "domain": [
                    "Hamiltonian Monte Carlo",
                    "Geometric Mechanics",
                    "Machine Learning",
                    "Statistical Inference"
                ],
                "publications": [
                    {
                        "title": "Hamiltonian Monte Carlo On Lie Groups and Constrained Mechanics on Homogeneous Manifolds",
                        "abstract": "In this paper we show that the Hamiltonian Monte Carlo method for compact Lie groups constructed in \\cite{kennedy88b} using a symplectic structure can be recovered from canonical geometric mechanics with a bi-invariant metric. Hence we obtain the correspondence between the various formulations of Hamiltonian mechanics on Lie groups, and their induced HMC algorithms. Working on $\\G\\times \\g$ we recover the Euler-Arnold formulation of geodesic motion, and construct explicit HMC schemes that extend \\cite{kennedy88b,Kennedy:2012} to non-compact Lie groups by choosing metrics with appropriate invariances. Finally we explain how mechanics on homogeneous spaces can be formulated as a constrained system over their associated Lie groups, and how in some important cases the constraints can be naturally handled by the symmetries of the Hamiltonian."
                    },
                    {
                        "title": "A numerical study of the 3D random interchange and random loop models",
                        "abstract": "We have studied numerically the random interchange model and related loop models on the three-dimensional cubic lattice. We have determined the transition time for the occurrence of long loops. The joint distribution of the lengths of long loops is Poisson-Dirichlet with parameter 1 or 1/2."
                    },
                    {
                        "title": "Hamiltonian Monte Carlo on Symmetric and Homogeneous Spaces via Symplectic Reduction",
                        "abstract": "The Hamiltonian Monte Carlo method generates samples by introducing a mechanical system that explores the target density. For distributions on manifolds it is not always simple to perform the mechanics as a result of the lack of global coordinates, the constraints of the manifold, and the requirement to compute the geodesic flow. In this paper we explain how to construct the Hamiltonian system on naturally reductive homogeneous spaces using symplectic reduction, which lifts the HMC scheme to a matrix Lie group with global coordinates and constant metric. This provides a general framework that is applicable to many manifolds that arise in applications, such as hyperspheres, hyperbolic spaces, symmetric positive-definite matrices, Grassmannian, and Stiefel manifolds."
                    },
                    {
                        "title": "Targeted Separation and Convergence with Kernel Discrepancies",
                        "abstract": "Maximum mean discrepancies (MMDs) like the kernel Stein discrepancy (KSD) have grown central to a wide range of applications, including hypothesis testing, sampler selection, distribution approximation, and variational inference. In each setting, these kernel-based discrepancy measures are required to (i) separate a target P from other probability measures or even (ii) control weak convergence to P. In this article we derive new sufficient and necessary conditions to ensure (i) and (ii). For MMDs on separable metric spaces, we characterize those kernels that separate Bochner embeddable measures and introduce simple conditions for separating all measures with unbounded kernels and for controlling convergence with bounded kernels. We use these results on $\\mathbb{R}^d$ to substantially broaden the known conditions for KSD separation and convergence control and to develop the first KSDs known to exactly metrize weak convergence to P. Along the way, we highlight the implications of our results for hypothesis testing, measuring and improving sample quality, and sampling with Stein variational gradient descent."
                    },
                    {
                        "title": "Irreversible Langevin MCMC on Lie Groups",
                        "abstract": "It is well-known that irreversible MCMC algorithms converge faster to their stationary distributions than reversible ones. Using the special geometric structure of Lie groups $\\mathcal G$ and dissipation fields compatible with the symplectic structure, we construct an irreversible HMC-like MCMC algorithm on $\\mathcal G$, where we first update the momentum by solving an OU process on the corresponding Lie algebra $\\mathfrak g$, and then approximate the Hamiltonian system on $\\mathcal G \\times \\mathfrak g$ with a reversible symplectic integrator followed by a Metropolis-Hastings correction step. In particular, when the OU process is simulated over sufficiently long times, we recover HMC as a special case. We illustrate this algorithm numerically using the example $\\mathcal G = SO(3)$."
                    },
                    {
                        "title": "Minimum Stein Discrepancy Estimators",
                        "abstract": "When maximum likelihood estimation is infeasible, one often turns to score matching, contrastive divergence, or minimum probability flow to obtain tractable parameter estimates. We provide a unifying perspective of these techniques as minimum Stein discrepancy estimators, and use this lens to design new diffusion kernel Stein discrepancy (DKSD) and diffusion score matching (DSM) estimators with complementary strengths. We establish the consistency, asymptotic normality, and robustness of DKSD and DSM estimators, then derive stochastic Riemannian gradient descent algorithms for their efficient optimisation. The main strength of our methodology is its flexibility, which allows us to design estimators with desirable properties for specific models at hand by carefully selecting a Stein discrepancy. We illustrate this advantage for several challenging problems for score matching, such as non-smooth, heavy-tailed or light-tailed densities."
                    },
                    {
                        "title": "Geometry and Dynamics for Markov Chain Monte Carlo",
                        "abstract": "Markov Chain Monte Carlo methods have revolutionised mathematical computation and enabled statistical inference within many previously intractable models. In this context, Hamiltonian dynamics have been proposed as an efficient way of building chains which can explore probability densities efficiently. The method emerges from physics and geometry and these links have been extensively studied by a series of authors through the last thirty years. However, there is currently a gap between the intuitions and knowledge of users of the methodology and our deep understanding of these theoretical foundations. The aim of this review is to provide a comprehensive introduction to the geometric tools used in Hamiltonian Monte Carlo at a level accessible to statisticians, machine learners and other users of the methodology with only a basic understanding of Monte Carlo methods. This will be complemented with some discussion of the most recent advances in the field which we believe will become increasingly relevant to applied scientists."
                    },
                    {
                        "title": "Posterior Integration on a Riemannian Manifold",
                        "abstract": "The geodesic Markov chain Monte Carlo method and its variants enable computation of integrals with respect to a posterior supported on a manifold. However, for regular integrals, the convergence rate of the ergodic average will be sub-optimal. To fill this gap, this paper extends the efficient posterior integration method of Oates et al. (2017) to the case of a Riemannian manifold. In contrast to the original Euclidean case, no non-trivial boundary conditions are needed for a closed manifold. The method is assessed through simulation and deployed to compute posterior integrals for an Australian Mesozoic paleomagnetic pole model, whose parameters are constrained to lie on the manifold $M = \\mathbb{S}^2 \\times \\mathbb{R}_+$."
                    },
                    {
                        "title": "Vector-Valued Control Variates",
                        "abstract": "Control variates are variance reduction tools for Monte Carlo estimators. They can provide significant variance reduction, but usually require a large number of samples, which can be prohibitive when sampling or evaluating the integrand is computationally expensive. Furthermore, there are many scenarios where we need to compute multiple related integrals simultaneously or sequentially, which can further exacerbate computational costs. In this paper, we propose vector-valued control variates, an extension of control variates which can be used to reduce the variance of multiple Monte Carlo estimators jointly. This allows for the transfer of information across integration tasks, and hence reduces the need for a large number of samples. We focus on control variates based on kernel interpolants and our novel construction is obtained through a generalised Stein identity and the development of novel matrix-valued Stein reproducing kernels. We demonstrate our methodology on a range of problems including multifidelity modelling, Bayesian inference for dynamical systems, and model evidence computation through thermodynamic integration."
                    },
                    {
                        "title": "A Riemann-Stein Kernel Method",
                        "abstract": "This paper proposes and studies a numerical method for approximation of posterior expectations based on interpolation with a Stein reproducing kernel. Finite-sample-size bounds on the approximation error are established for posterior distributions supported on a compact Riemannian manifold, and we relate these to a kernel Stein discrepancy (KSD). Moreover, we prove in our setting that the KSD is equivalent to Sobolev discrepancy and, in doing so, we completely characterise the convergence-determining properties of KSD. Our contribution is rooted in a novel combination of Stein's method, the theory of reproducing kernels, and existence and regularity results for partial differential equations on a Riemannian manifold."
                    },
                    {
                        "title": "A Unifying and Canonical Description of Measure-Preserving Diffusions",
                        "abstract": "A complete recipe of measure-preserving diffusions in Euclidean space was recently derived unifying several MCMC algorithms into a single framework. In this paper, we develop a geometric theory that improves and generalises this construction to any manifold. We thereby demonstrate that the completeness result is a direct consequence of the topology of the underlying manifold and the geometry induced by the target measure $P$; there is no need to introduce other structures such as a Riemannian metric, local coordinates, or a reference measure. Instead, our framework relies on the intrinsic geometry of $P$ and in particular its canonical derivative, the deRham rotationnel, which allows us to parametrise the Fokker--Planck currents of measure-preserving diffusions using potentials. The geometric formalism can easily incorporate constraints and symmetries, and deliver new important insights, for example, a new complete recipe of Langevin-like diffusions that are suited to the construction of samplers. We also analyse the reversibility and dissipative properties of the diffusions, the associated deterministic flow on the space of measures, and the geometry of Langevin processes. Our article connects ideas from various literature and frames the theory of measure-preserving diffusions in its appropriate mathematical context."
                    },
                    {
                        "title": "Optimization on manifolds: A symplectic approach",
                        "abstract": "Optimization tasks are crucial in statistical machine learning. Recently, there has been great interest in leveraging tools from dynamical systems to derive accelerated and robust optimization methods via suitable discretizations of continuous-time systems. However, these ideas have mostly been limited to Euclidean spaces and unconstrained settings, or to Riemannian gradient flows. In this work, we propose a dissipative extension of Dirac's theory of constrained Hamiltonian systems as a general framework for solving optimization problems over smooth manifolds, including problems with nonlinear constraints. We develop geometric/symplectic numerical integrators on manifolds that are \"rate-matching,\" i.e., preserve the continuous-time rates of convergence. In particular, we introduce a dissipative RATTLE integrator able to achieve optimal convergence rate locally. Our class of (accelerated) algorithms are not only simple and efficient but also applicable to a broad range of contexts."
                    },
                    {
                        "title": "Controlling Moments with Kernel Stein Discrepancies",
                        "abstract": "Kernel Stein discrepancies (KSDs) measure the quality of a distributional approximation and can be computed even when the target density has an intractable normalizing constant. Notable applications include the diagnosis of approximate MCMC samplers and goodness-of-fit tests for unnormalized statistical models. The present work analyzes the convergence control properties of KSDs. We first show that standard KSDs used for weak convergence control fail to control moment convergence. To address this limitation, we next provide sufficient conditions under which alternative diffusion KSDs control both moment and weak convergence. As an immediate consequence we develop, for each $q > 0$, the first KSDs known to exactly characterize $q$-Wasserstein convergence."
                    },
                    {
                        "title": "Improving embedding of graphs with missing data by soft manifolds",
                        "abstract": "Embedding graphs in continous spaces is a key factor in designing and developing algorithms for automatic information extraction to be applied in diverse tasks (e.g., learning, inferring, predicting). The reliability of graph embeddings directly depends on how much the geometry of the continuous space matches the graph structure. Manifolds are mathematical structure that can enable to incorporate in their topological spaces the graph characteristics, and in particular nodes distances. State-of-the-art of manifold-based graph embedding algorithms take advantage of the assumption that the projection on a tangential space of each point in the manifold (corresponding to a node in the graph) would locally resemble a Euclidean space. Although this condition helps in achieving efficient analytical solutions to the embedding problem, it does not represent an adequate set-up to work with modern real life graphs, that are characterized by weighted connections across nodes often computed over sparse datasets with missing records. In this work, we introduce a new class of manifold, named soft manifold, that can solve this situation. In particular, soft manifolds are mathematical structures with spherical symmetry where the tangent spaces to each point are hypocycloids whose shape is defined according to the velocity of information propagation across the data points. Using soft manifolds for graph embedding, we can provide continuous spaces to pursue any task in data analysis over complex datasets. Experimental results on reconstruction tasks on synthetic and real datasets show how the proposed approach enable more accurate and reliable characterization of graphs in continuous spaces with respect to the state-of-the-art."
                    },
                    {
                        "title": "Statistical Inference for Generative Models with Maximum Mean Discrepancy",
                        "abstract": "While likelihood-based inference and its variants provide a statistically efficient and widely applicable approach to parametric inference, their application to models involving intractable likelihoods poses challenges. In this work, we study a class of minimum distance estimators for intractable generative models, that is, statistical models for which the likelihood is intractable, but simulation is cheap. The distance considered, maximum mean discrepancy (MMD), is defined through the embedding of probability measures into a reproducing kernel Hilbert space. We study the theoretical properties of these estimators, showing that they are consistent, asymptotically normal and robust to model misspecification. A main advantage of these estimators is the flexibility offered by the choice of kernel, which can be used to trade-off statistical efficiency and robustness. On the algorithmic side, we study the geometry induced by MMD on the parameter space and use this to introduce a novel natural gradient descent-like algorithm for efficient implementation of these estimators. We illustrate the relevance of our theoretical results on several classes of models including a discrete-time latent Markov process and two multivariate stochastic differential equation models."
                    },
                    {
                        "title": "Warped geometric information on the optimisation of Euclidean functions",
                        "abstract": "We consider the fundamental task of optimising a real-valued function defined in a potentially high-dimensional Euclidean space, such as the loss function in many machine-learning tasks or the logarithm of the probability distribution in statistical inference. We use Riemannian geometry notions to redefine the optimisation problem of a function on the Euclidean space to a Riemannian manifold with a warped metric, and then find the function's optimum along this manifold. The warped metric chosen for the search domain induces a computational friendly metric-tensor for which optimal search directions associated with geodesic curves on the manifold becomes easier to compute. Performing optimization along geodesics is known to be generally infeasible, yet we show that in this specific manifold we can analytically derive Taylor approximations up to third-order. In general these approximations to the geodesic curve will not lie on the manifold, however we construct suitable retraction maps to pull them back onto the manifold. Therefore, we can efficiently optimize along the approximate geodesic curves. We cover the related theory, describe a practical optimization algorithm and empirically evaluate it on a collection of challenging optimisation benchmarks. Our proposed algorithm, using 3rd-order approximation of geodesics, tends to outperform standard Euclidean gradient-based counterparts in term of number of iterations until convergence."
                    },
                    {
                        "title": "Stein Point Markov Chain Monte Carlo",
                        "abstract": "An important task in machine learning and statistics is the approximation of a probability measure by an empirical measure supported on a discrete point set. Stein Points are a class of algorithms for this task, which proceed by sequentially minimising a Stein discrepancy between the empirical measure and the target and, hence, require the solution of a non-convex optimisation problem to obtain each new point. This paper removes the need to solve this optimisation problem by, instead, selecting each new point based on a Markov chain sample path. This significantly reduces the computational cost of Stein Points and leads to a suite of algorithms that are straightforward to implement. The new algorithms are illustrated on a set of challenging Bayesian inference problems, and rigorous theoretical guarantees of consistency are established."
                    },
                    {
                        "title": "Geometric Methods for Sampling, Optimisation, Inference and Adaptive Agents",
                        "abstract": "In this chapter, we identify fundamental geometric structures that underlie the problems of sampling, optimisation, inference and adaptive decision-making. Based on this identification, we derive algorithms that exploit these geometric structures to solve these problems efficiently. We show that a wide range of geometric theories emerge naturally in these fields, ranging from measure-preserving processes, information divergences, Poisson geometry, and geometric integration. Specifically, we explain how (i) leveraging the symplectic geometry of Hamiltonian systems enable us to construct (accelerated) sampling and optimisation methods, (ii) the theory of Hilbertian subspaces and Stein operators provides a general methodology to obtain robust estimators, (iii) preserving the information geometry of decision-making yields adaptive agents that perform active inference. Throughout, we emphasise the rich connections between these fields; e.g., inference draws on sampling and optimisation, and adaptive decision-making assesses decisions by inferring their counterfactual consequences. Our exposition provides a conceptual overview of underlying ideas, rather than a technical discussion, which can be found in the references herein."
                    },
                    {
                        "title": "Deep Learning as Ricci Flow",
                        "abstract": "Deep neural networks (DNNs) are powerful tools for approximating the distribution of complex data. It is known that data passing through a trained DNN classifier undergoes a series of geometric and topological simplifications. While some progress has been made toward understanding these transformations in neural networks with smooth activation functions, an understanding in the more general setting of non-smooth activation functions, such as the rectified linear unit (ReLU), which tend to perform better, is required. Here we propose that the geometric transformations performed by DNNs during classification tasks have parallels to those expected under Hamilton's Ricci flow - a tool from differential geometry that evolves a manifold by smoothing its curvature, in order to identify its topology. To illustrate this idea, we present a computational framework to quantify the geometric changes that occur as data passes through successive layers of a DNN, and use this framework to motivate a notion of `global Ricci network flow' that can be used to assess a DNN's ability to disentangle complex data geometries to solve classification problems. By training more than $1,500$ DNN classifiers of different widths and depths on synthetic and real-world data, we show that the strength of global Ricci network flow-like behaviour correlates with accuracy for well-trained DNNs, independently of depth, width and data set. Our findings motivate the use of tools from differential and discrete geometry to the problem of explainability in deep learning."
                    }
                ]
            },
            "fd89e9bd-6910-496b-9451-b4a474188ff9": {
                "pk": "fd89e9bd-6910-496b-9451-b4a474188ff9",
                "name": "Bernhard Sch\u00f6lkopf",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "AutoML",
                    "Multi-task Learning"
                ]
            },
            "71789fd2-280c-4c18-8b81-548486768274": {
                "pk": "71789fd2-280c-4c18-8b81-548486768274",
                "name": "Lester Mackey",
                "collaborators": [
                    "Jackson Gorham",
                    "Raaz Dwivedi",
                    "Daniel Paulin",
                    "Joel A. Tropp",
                    "Morgane Austern",
                    "Zhou Fan",
                    "Ioannis Mitliagkas",
                    "Nilesh Tripuraneni",
                    "Niloy Biswas",
                    "Jiaxin Shi"
                ],
                "domain": [
                    "Statistical Inference",
                    "Markov Chain Monte Carlo",
                    "Machine Learning",
                    "Bayesian Methods"
                ],
                "publications": [
                    {
                        "title": "Multivariate Stein Factors for a Class of Strongly Log-concave Distributions",
                        "abstract": "We establish uniform bounds on the low-order derivatives of Stein equation solutions for a broad class of multivariate, strongly log-concave target distributions. These \"Stein factor\" bounds deliver control over Wasserstein and related smooth function distances and are well-suited to analyzing the computable Stein discrepancy measures of Gorham and Mackey. Our arguments of proof are probabilistic and feature the synchronous coupling of multiple overdamped Langevin diffusions."
                    },
                    {
                        "title": "Efficient Concentration with Gaussian Approximation",
                        "abstract": "Concentration inequalities for the sample mean, like those due to Bernstein and Hoeffding, are valid for any sample size but overly conservative, yielding confidence intervals that are unnecessarily wide. The central limit theorem (CLT) provides asymptotic confidence intervals with optimal width, but these are invalid for all sample sizes. To resolve this tension, we develop new computable concentration inequalities with asymptotically optimal size, finite-sample validity, and sub-Gaussian decay. These bounds enable the construction of efficient confidence intervals with correct coverage for any sample size and efficient empirical Berry-Esseen bounds that require no prior knowledge of the population variance. We derive our inequalities by tightly bounding non-uniform Kolmogorov and Wasserstein distances to a Gaussian using zero-bias couplings and Stein's method of exchangeable pairs."
                    },
                    {
                        "title": "Generalized Kernel Thinning",
                        "abstract": "The kernel thinning (KT) algorithm of Dwivedi and Mackey (2021) compresses a probability distribution more effectively than independent sampling by targeting a reproducing kernel Hilbert space (RKHS) and leveraging a less smooth square-root kernel. Here we provide four improvements. First, we show that KT applied directly to the target RKHS yields tighter, dimension-free guarantees for any kernel, any distribution, and any fixed function in the RKHS. Second, we show that, for analytic kernels like Gaussian, inverse multiquadric, and sinc, target KT admits maximum mean discrepancy (MMD) guarantees comparable to or better than those of square-root KT without making explicit use of a square-root kernel. Third, we prove that KT with a fractional power kernel yields better-than-Monte-Carlo MMD guarantees for non-smooth kernels, like Laplace and Mat\\'ern, that do not have square-roots. Fourth, we establish that KT applied to a sum of the target and power kernels (a procedure we call KT+) simultaneously inherits the improved MMD guarantees of power KT and the tighter individual function guarantees of target KT. In our experiments with target KT and KT+, we witness significant improvements in integration error even in $100$ dimensions and when compressing challenging differential equation posteriors."
                    },
                    {
                        "title": "Measuring Sample Quality with Stein's Method",
                        "abstract": "To improve the efficiency of Monte Carlo estimation, practitioners are turning to biased Markov chain Monte Carlo procedures that trade off asymptotic exactness for computational speed. The reasoning is sound: a reduction in variance due to more rapid sampling can outweigh the bias introduced. However, the inexactness creates new challenges for sampler and parameter selection, since standard measures of sample quality like effective sample size do not account for asymptotic bias. To address these challenges, we introduce a new computable quality measure based on Stein's method that quantifies the maximum discrepancy between sample and target expectations over a large class of test functions. We use our tool to compare exact, biased, and deterministic sample sequences and illustrate applications to hyperparameter selection, convergence rate assessment, and quantifying bias-variance tradeoffs in posterior inference."
                    },
                    {
                        "title": "Empirical Bayesian analysis of simultaneous changepoints in multiple data sequences",
                        "abstract": "Copy number variations in cancer cells and volatility fluctuations in stock prices are commonly manifested as changepoints occurring at the same positions across related data sequences. We introduce a Bayesian modeling framework, BASIC, that employs a changepoint prior to capture the co-occurrence tendency in data of this type. We design efficient algorithms to sample from and maximize over the BASIC changepoint posterior and develop a Monte Carlo expectation-maximization procedure to select prior hyperparameters in an empirical Bayes fashion. We use the resulting BASIC framework to analyze DNA copy number variations in the NCI-60 cancer cell lines and to identify important events that affected the price volatility of S&P 500 stocks from 2000 to 2009."
                    },
                    {
                        "title": "Improving Gibbs Sampler Scan Quality with DoGS",
                        "abstract": "The pairwise influence matrix of Dobrushin has long been used as an analytical tool to bound the rate of convergence of Gibbs sampling. In this work, we use Dobrushin influence as the basis of a practical tool to certify and efficiently improve the quality of a discrete Gibbs sampler. Our Dobrushin-optimized Gibbs samplers (DoGS) offer customized variable selection orders for a given sampling budget and variable subset of interest, explicit bounds on total variation distance to stationarity, and certifiable improvements over the standard systematic and uniform random scan Gibbs samplers. In our experiments with joint image segmentation and object recognition, Markov chain Monte Carlo maximum likelihood estimation, and Ising model inference, DoGS consistently deliver higher-quality inferences with significantly smaller sampling budgets than standard Gibbs samplers."
                    },
                    {
                        "title": "Single Point Transductive Prediction",
                        "abstract": "Standard methods in supervised learning separate training and prediction: the model is fit independently of any test points it may encounter. However, can knowledge of the next test point $\\mathbf{x}_{\\star}$ be exploited to improve prediction accuracy? We address this question in the context of linear prediction, showing how techniques from semi-parametric inference can be used transductively to combat regularization bias. We first lower bound the $\\mathbf{x}_{\\star}$ prediction error of ridge regression and the Lasso, showing that they must incur significant bias in certain test directions. We then provide non-asymptotic upper bounds on the $\\mathbf{x}_{\\star}$ prediction error of two transductive prediction rules. We conclude by showing the efficacy of our methods on both synthetic and real data, highlighting the improvements single point transductive prediction can provide in settings with distribution shift."
                    },
                    {
                        "title": "Bounding Wasserstein distance with couplings",
                        "abstract": "Markov chain Monte Carlo (MCMC) provides asymptotically consistent estimates of intractable posterior expectations as the number of iterations tends to infinity. However, in large data applications, MCMC can be computationally expensive per iteration. This has catalyzed interest in approximating MCMC in a manner that improves computational speed per iteration but does not produce asymptotically consistent estimates. In this article, we propose estimators based on couplings of Markov chains to assess the quality of such asymptotically biased sampling methods. The estimators give empirical upper bounds of the Wasserstein distance between the limiting distribution of the asymptotically biased sampling method and the original target distribution of interest. We establish theoretical guarantees for our upper bounds and show that our estimators can remain effective in high dimensions. We apply our quality measures to stochastic gradient MCMC, variational Bayes, and Laplace approximations for tall data and to approximate MCMC for Bayesian logistic regression in 4500 dimensions and Bayesian linear regression in 50000 dimensions."
                    },
                    {
                        "title": "A Finite-Particle Convergence Rate for Stein Variational Gradient Descent",
                        "abstract": "We provide the first finite-particle convergence rate for Stein variational gradient descent (SVGD), a popular algorithm for approximating a probability distribution with a collection of particles. Specifically, whenever the target distribution is sub-Gaussian with a Lipschitz score, SVGD with n particles and an appropriate step size sequence drives the kernel Stein discrepancy to zero at an order 1/sqrt(log log n) rate. We suspect that the dependence on n can be improved, and we hope that our explicit, non-asymptotic proof strategy will serve as a template for future refinements."
                    },
                    {
                        "title": "Corrupted Sensing: Novel Guarantees for Separating Structured Signals",
                        "abstract": "We study the problem of corrupted sensing, a generalization of compressed sensing in which one aims to recover a signal from a collection of corrupted or unreliable measurements. While an arbitrary signal cannot be recovered in the face of arbitrary corruption, tractable recovery is possible when both signal and corruption are suitably structured. We quantify the relationship between signal recovery and two geometric measures of structure, the Gaussian complexity of a tangent cone and the Gaussian distance to a subdifferential. We take a convex programming approach to disentangling signal and corruption, analyzing both penalized programs that trade off between signal and corruption complexity, and constrained programs that bound the complexity of signal or corruption when prior information is available. In each case, we provide conditions for exact signal recovery from structured corruption and stable signal recovery from structured corruption with added unstructured noise. Our simulations demonstrate close agreement between our theoretical recovery bounds and the sharp phase transitions observed in practice. In addition, we provide new interpretable bounds for the Gaussian complexity of sparse vectors, block-sparse vectors, and low-rank matrices, which lead to sharper guarantees of recovery when combined with our results and those in the literature."
                    },
                    {
                        "title": "Measuring Sample Quality with Kernels",
                        "abstract": "Approximate Markov chain Monte Carlo (MCMC) offers the promise of more rapid sampling at the cost of more biased inference. Since standard MCMC diagnostics fail to detect these biases, researchers have developed computable Stein discrepancy measures that provably determine the convergence of a sample to its target distribution. This approach was recently combined with the theory of reproducing kernels to define a closed-form kernel Stein discrepancy (KSD) computable by summing kernel evaluations across pairs of sample points. We develop a theory of weak convergence for KSDs based on Stein's method, demonstrate that commonly used KSDs fail to detect non-convergence even for Gaussian targets, and show that kernels with slowly decaying tails provably determine convergence for a large class of target distributions. The resulting convergence-determining KSDs are suitable for comparing biased, exact, and deterministic sample sequences and simpler to compute and parallelize than alternative Stein discrepancies. We use our tools to compare biased samplers, select sampler hyperparameters, and improve upon existing KSD approaches to one-sample hypothesis testing and sample quality improvement."
                    },
                    {
                        "title": "Kernel Thinning",
                        "abstract": "We introduce kernel thinning, a new procedure for compressing a distribution $\\mathbb{P}$ more effectively than i.i.d. sampling or standard thinning. Given a suitable reproducing kernel $\\mathbf{k}_{\\star}$ and $O(n^2)$ time, kernel thinning compresses an $n$-point approximation to $\\mathbb{P}$ into a $\\sqrt{n}$-point approximation with comparable worst-case integration error across the associated reproducing kernel Hilbert space. The maximum discrepancy in integration error is $O_d(n^{-1/2}\\sqrt{\\log n})$ in probability for compactly supported $\\mathbb{P}$ and $O_d(n^{-\\frac{1}{2}} (\\log n)^{(d+1)/2}\\sqrt{\\log\\log n})$ for sub-exponential $\\mathbb{P}$ on $\\mathbb{R}^d$. In contrast, an equal-sized i.i.d. sample from $\\mathbb{P}$ suffers $\\Omega(n^{-1/4})$ integration error. Our sub-exponential guarantees resemble the classical quasi-Monte Carlo error rates for uniform $\\mathbb{P}$ on $[0,1]^d$ but apply to general distributions on $\\mathbb{R}^d$ and a wide range of common kernels. Moreover, the same construction delivers near-optimal $L^\\infty$ coresets in $O(n^2)$ time. We use our results to derive explicit non-asymptotic maximum mean discrepancy bounds for Gaussian, Mat\\'ern, and B-spline kernels and present two vignettes illustrating the practical benefits of kernel thinning over i.i.d. sampling and standard Markov chain Monte Carlo thinning, in dimensions $d=2$ through $100$."
                    },
                    {
                        "title": "Deriving Matrix Concentration Inequalities from Kernel Couplings",
                        "abstract": "This paper derives exponential tail bounds and polynomial moment inequalities for the spectral norm deviation of a random matrix from its mean value. The argument depends on a matrix extension of Stein's method of exchangeable pairs for concentration of measure, as introduced by Chatterjee. Recent work of Mackey et al. uses these techniques to analyze random matrices with additive structure, while the enhancements in this paper cover a wider class of matrix-valued random elements. In particular, these ideas lead to a bounded differences inequality that applies to random matrices constructed from weakly dependent random variables. The proofs require novel trace inequalities that may be of independent interest."
                    },
                    {
                        "title": "Distribution Compression in Near-linear Time",
                        "abstract": "In distribution compression, one aims to accurately summarize a probability distribution $\\mathbb{P}$ using a small number of representative points. Near-optimal thinning procedures achieve this goal by sampling $n$ points from a Markov chain and identifying $\\sqrt{n}$ points with $\\widetilde{\\mathcal{O}}(1/\\sqrt{n})$ discrepancy to $\\mathbb{P}$. Unfortunately, these algorithms suffer from quadratic or super-quadratic runtime in the sample size $n$. To address this deficiency, we introduce Compress++, a simple meta-procedure for speeding up any thinning algorithm while suffering at most a factor of $4$ in error. When combined with the quadratic-time kernel halving and kernel thinning algorithms of Dwivedi and Mackey (2021), Compress++ delivers $\\sqrt{n}$ points with $\\mathcal{O}(\\sqrt{\\log n/n})$ integration error and better-than-Monte-Carlo maximum mean discrepancy in $\\mathcal{O}(n \\log^3 n)$ time and $\\mathcal{O}( \\sqrt{n} \\log^2 n )$ space. Moreover, Compress++ enjoys the same near-linear runtime given any quadratic-time input and reduces the runtime of super-quadratic algorithms by a square-root factor. In our benchmarks with high-dimensional Monte Carlo samples and Markov chains targeting challenging differential equation posteriors, Compress++ matches or nearly matches the accuracy of its input algorithm in orders of magnitude less time."
                    },
                    {
                        "title": "Efron-Stein Inequalities for Random Matrices",
                        "abstract": "This paper establishes new concentration inequalities for random matrices constructed from independent random variables. These results are analogous with the generalized Efron-Stein inequalities developed by Boucheron et al. The proofs rely on the method of exchangeable pairs."
                    },
                    {
                        "title": "Feature-Weighted Linear Stacking",
                        "abstract": "Ensemble methods, such as stacking, are designed to boost predictive accuracy by blending the predictions of multiple machine learning models. Recent work has shown that the use of meta-features, additional inputs describing each example in a dataset, can boost the performance of ensemble methods, but the greatest reported gains have come from nonlinear procedures requiring significant tuning and training time. Here, we present a linear technique, Feature-Weighted Linear Stacking (FWLS), that incorporates meta-features for improved accuracy while retaining the well-known virtues of linear regression regarding speed, stability, and interpretability. FWLS combines model predictions linearly using coefficients that are themselves linear functions of meta-features. This technique was a key facet of the solution of the second place team in the recently concluded Netflix Prize competition. Significant increases in accuracy over standard linear stacking are demonstrated on the Netflix Prize collaborative filtering dataset."
                    },
                    {
                        "title": "Orthogonal Machine Learning: Power and Limitations",
                        "abstract": "Double machine learning provides $\\sqrt{n}$-consistent estimates of parameters of interest even when high-dimensional or nonparametric nuisance parameters are estimated at an $n^{-1/4}$ rate. The key is to employ Neyman-orthogonal moment equations which are first-order insensitive to perturbations in the nuisance parameters. We show that the $n^{-1/4}$ requirement can be improved to $n^{-1/(2k+2)}$ by employing a $k$-th order notion of orthogonality that grants robustness to more complex or higher-dimensional nuisance parameters. In the partially linear regression setting popular in causal inference, we show that we can construct second-order orthogonal moments if and only if the treatment residual is not normally distributed. Our proof relies on Stein's lemma and may be of independent interest. We conclude by demonstrating the robustness benefits of an explicit doubly-orthogonal estimation procedure for treatment effect."
                    },
                    {
                        "title": "Distributed Matrix Completion and Robust Factorization",
                        "abstract": "If learning methods are to scale to the massive sizes of modern datasets, it is essential for the field of machine learning to embrace parallel and distributed computing. Inspired by the recent development of matrix factorization methods with rich theory but poor computational complexity and by the relative ease of mapping matrices onto distributed architectures, we introduce a scalable divide-and-conquer framework for noisy matrix factorization. We present a thorough theoretical analysis of this framework in which we characterize the statistical errors introduced by the \"divide\" step and control their magnitude in the \"conquer\" step, so that the overall algorithm enjoys high-probability estimation guarantees comparable to those of its base algorithm. We also present experiments in collaborative filtering and video background modeling that demonstrate the near-linear to superlinear speed-ups attainable with this approach."
                    },
                    {
                        "title": "Approximate Cross-validation: Guarantees for Model Assessment and Selection",
                        "abstract": "Cross-validation (CV) is a popular approach for assessing and selecting predictive models. However, when the number of folds is large, CV suffers from a need to repeatedly refit a learning procedure on a large number of training datasets. Recent work in empirical risk minimization (ERM) approximates the expensive refitting with a single Newton step warm-started from the full training set optimizer. While this can greatly reduce runtime, several open questions remain including whether these approximations lead to faithful model selection and whether they are suitable for non-smooth objectives. We address these questions with three main contributions: (i) we provide uniform non-asymptotic, deterministic model assessment guarantees for approximate CV; (ii) we show that (roughly) the same conditions also guarantee model selection performance comparable to CV; (iii) we provide a proximal Newton extension of the approximate CV framework for non-smooth prediction problems and develop improved assessment guarantees for problems such as l1-regularized ERM."
                    },
                    {
                        "title": "Random Feature Stein Discrepancies",
                        "abstract": "Computable Stein discrepancies have been deployed for a variety of applications, ranging from sampler selection in posterior inference to approximate Bayesian inference to goodness-of-fit testing. Existing convergence-determining Stein discrepancies admit strong theoretical guarantees but suffer from a computational cost that grows quadratically in the sample size. While linear-time Stein discrepancies have been proposed for goodness-of-fit testing, they exhibit avoidable degradations in testing power -- even when power is explicitly optimized. To address these shortcomings, we introduce feature Stein discrepancies ($\\Phi$SDs), a new family of quality measures that can be cheaply approximated using importance sampling. We show how to construct $\\Phi$SDs that provably determine the convergence of a sample to its target and develop high-accuracy approximations -- random $\\Phi$SDs (R$\\Phi$SDs) -- which are computable in near-linear time. In our experiments with sampler selection for approximate posterior inference and goodness-of-fit testing, R$\\Phi$SDs perform as well or better than quadratic-time KSDs while being orders of magnitude faster to compute."
                    }
                ]
            }
        }
    },
    "2408.15792": {
        "paper_data": {
            "title": "Efficient LLM Scheduling by Learning to Rank",
            "url": "http://arxiv.org/abs/2408.15792v1",
            "arxiv_id": "2408.15792",
            "authors": [
                "Yichao Fu",
                "Siqi Zhu",
                "Runlong Su",
                "Aurick Qiao",
                "Ion Stoica",
                "Hao Zhang"
            ],
            "abstract": "In Large Language Model (LLM) inference, the output length of an LLM request is typically regarded as not known a priori. Consequently, most LLM serving systems employ a simple First-come-first-serve (FCFS) scheduling strategy, leading to Head-Of-Line (HOL) blocking and reduced throughput and service quality. In this paper, we reexamine this assumption -- we show that, although predicting the exact generation length of each request is infeasible, it is possible to predict the relative ranks of output lengths in a batch of requests, using learning to rank. The ranking information offers valuable guidance for scheduling requests. Building on this insight, we develop a novel scheduler for LLM inference and serving that can approximate the shortest-job-first (SJF) schedule better than existing approaches. We integrate this scheduler with the state-of-the-art LLM serving system and show significant performance improvement in several important applications: 2.8x lower latency in chatbot serving and 6.5x higher throughput in synthetic data generation. Our code is available at https://github.com/hao-ai-lab/vllm-ltr.git",
            "introduction": "   1 Introduction  Large language models (LLMs) are increasingly becoming the backbone of many today\u2019s Internet services and applications that serve millions of users\u00a0[1]. Due to the surge in demand, efficient scheduling for LLM serving is crucial to ensure high-quality service amidst numerous concurrent users competing for computing resources. For popular interactive applications such as chatbots, this means minimizing the latency that each user perceives while maximizing the overall system throughput to accommodate as many users as possible.   Under high load, LLM services that implement a first-come-first-serve (FCFS) scheduling strategy inevitably face significant Head-Of-Line (HOL) blocking, as many requests must wait for others to execute. Figure\u00a01 illustrates a typical example. In such scenarios, it is well-established that the shortest-job-first (SJF) and shortest-remaining-time-first (SRTF) scheduling algorithms minimize the average latency experienced across all requests. However, SJF/SRTF are seldom implemented in LLM services because they require requests to be ordered by their remaining generation lengths, which is traditionally assumed to be difficult or impossible to know ahead of time in existing systems\u00a0[2, 3].    Figure 1: A long request can block short requests and introduce severe HOL blocking and high latency. We assume there is no prefill time, and the system takes 1 second to generate 1 token. With a First-come-first-serve (FCFS) schedule, the long request R0, which arrives first and takes 10 seconds to generate 10 tokens, will block subsequent shorter requests R1 and R2 for 10 seconds. Hence the latencies of R0, R1, and R2 are 10/10=1,(10+2)/2=6,(10+2+1)/1=13\u2062\u00a0s / tokenformulae-sequence10101formulae-sequence102261021113\u00a0s / token10/10=1,(10+2)/2=6,(10+2+1)/1=13\\mbox{ s / token}10 / 10 = 1 , ( 10 + 2 ) / 2 = 6 , ( 10 + 2 + 1 ) / 1 = 13 s / token, respectively, perceived by users, with an average latency of (1+6+13)/3=6.67\u2062\u00a0s / token161336.67\u00a0s / token(1+6+13)/3=6.67\\mbox{ s / token}( 1 + 6 + 13 ) / 3 = 6.67 s / token. By contrast, prioritizing shortest requests yields an average latency of (1.3+1.5+1)/3=1.27\u2062\u00a0s / token1.31.5131.27\u00a0s / token(1.3+1.5+1)/3=1.27\\mbox{ s / token}( 1.3 + 1.5 + 1 ) / 3 = 1.27 s / token \u2013 a 5.3\u00d75.3\\times5.3 \u00d7 reduction in average latency.   In this paper, we contend that, although accurately knowing the generation length of requests may be difficult, it is actually not needed. Rather, just knowing the relative ordering between request lengths is sufficient for SJF/SRTF scheduling. To this end, we propose to use the Kendall rank correlation coefficient (Kendall\u2019s Tau)\u00a0[4] to measure the similarity between a predicted schedule and the SJF/SRTF schedule based on groundtruth generation lengths (i.e. oracle). We demonstrate that schedules with higher similarities (measured by Kendal\u2019s Tau) to the oracle generally translate to lower latencies in real-world performance (Figure.\u00a02).   Based on this insight, we propose to optimize the request scheduling in LLM serving via learning to rank. We show that a small auxiliary model (e.g., OPT-125M\u00a0[5]) can be trained to accurately rank LLM requests by their generation lengths, prior to execution, at virtually no cost. For both offline batch generation and online latency-sensitive tasks, by scheduling requests on-the-fly based on the predicted rankings, we can approximate the SRTF/SJF schedule, hence reduce average latency and improve throughput,",
            "references": [
                {
                    "title": "DynamoLLM: Designing LLM Inference Clusters for Performance and Energy Efficiency",
                    "abstract": "The rapid evolution and widespread adoption of generative large language models (LLMs) have made them a pivotal workload in various applications. Today, LLM inference clusters receive a large number of queries with strict Service Level Objectives (SLOs). To achieve the desired performance, these models execute on power-hungry GPUs causing the inference clusters to consume large amount of energy and, consequently, result in excessive carbon emissions. Fortunately, we find that there is a great opportunity to exploit the heterogeneity in inference compute properties and fluctuations in inference workloads, to significantly improve energy-efficiency. However, such a diverse and dynamic environment creates a large search-space where different system configurations (e.g., number of instances, model parallelism, and GPU frequency) translate into different energy-performance trade-offs. To address these challenges, we propose DynamoLLM, the first energy-management framework for LLM inference environments. DynamoLLM automatically and dynamically reconfigures the inference cluster to optimize for energy and cost of LLM serving under the service's performance SLOs. We show that at a service-level, DynamoLLM conserves 53% energy and 38% operational carbon emissions, and reduces 61% cost to the customer, while meeting the latency SLOs."
                },
                {
                    "title": "Enabling Efficient Batch Serving for LMaaS via Generation Length Prediction",
                    "abstract": "Nowadays, large language models (LLMs) are published as a service and can be accessed by various applications via APIs, also known as language-model-as-a-service (LMaaS). Without knowing the generation length of requests, existing serving systems serve requests in a first-come, first-served (FCFS) manner with a fixed batch size, which leads to two problems that affect batch serving efficiency. First, the generation lengths of requests in a batch vary, and requests with short generation lengths must wait for requests with long generation lengths to finish during the batch serving procedure. Second, requests with longer generation lengths consume more memory during serving. Without knowing the generation lengths of batched requests, the batch size is always set small to avoid the out-of-memory (OOM) error, thus preventing the GPU from being fully utilized. In this paper, we find that a significant number of popular applications in the LMaaS scenario have a positive correlation between the generation length and the length of raw user input. Based on this observation, we propose Magnus, which can accurately predict the request generation length with the user input length, application-level, and user-level semantic features. Accordingly, Magnus can achieve high request throughput by batching requests of similar generation lengths together with adaptive batch sizes. Besides, Magnus can also schedule batches with the highest response ratio next (HRRN) policy to reduce request response time. Experiments conducted on our testbed show that Magnus improves request throughput by up to 234% and reduces response time by up to 89.7% compared to baselines."
                },
                {
                    "title": "Andes: Defining and Enhancing Quality-of-Experience in LLM-Based Text Streaming Services",
                    "abstract": "The advent of large language models (LLMs) has transformed text-based services, enabling capabilities ranging from real-time translation to AI-driven chatbots. However, existing serving systems primarily focus on optimizing server-side aggregate metrics like token generation throughput, ignoring individual user experience with streamed text. As a result, under high and/or bursty load, a significant number of users can receive unfavorable service quality or poor Quality-of-Experience (QoE). In this paper, we first formally define QoE of text streaming services, where text is delivered incrementally and interactively to users, by considering the end-to-end token delivery process throughout the entire interaction with the user. Thereafter, we propose Andes, a QoE-aware serving system that enhances user experience for LLM-enabled text streaming services. At its core, Andes strategically allocates contended GPU resources among multiple requests over time to optimize their QoE. Our evaluations demonstrate that, compared to the state-of-the-art LLM serving systems like vLLM, Andes improves the average QoE by up to 3.2$\\times$ under high request rate, or alternatively, it attains up to 1.6$\\times$ higher request rate while preserving high QoE."
                },
                {
                    "title": "Efficient Interactive LLM Serving with Proxy Model-based Sequence Length Prediction",
                    "abstract": "Large language models (LLMs) have been driving a new wave of interactive AI applications across numerous domains. However, efficiently serving LLM inference requests is challenging due to their unpredictable execution times originating from the autoregressive nature of generative models. Existing LLM serving systems exploit first-come-first-serve (FCFS) scheduling, suffering from head-of-line blocking issues. To address the non-deterministic nature of LLMs and enable efficient interactive LLM serving, we present a speculative shortest-job-first (SSJF) scheduler that uses a light proxy model to predict LLM output sequence lengths. Our open-source SSJF implementation does not require changes to memory management or batching strategies. Evaluations on real-world datasets and production workload traces show that SSJF reduces average job completion times by 30.5-39.6% and increases throughput by 2.2-3.6x compared to FCFS schedulers, across no batching, dynamic batching, and continuous batching settings."
                },
                {
                    "title": "Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators",
                    "abstract": "LLM-based auto-annotators have become a key component of the LLM development process due to their cost-effectiveness and scalability compared to human-based evaluation. However, these auto-annotators can introduce complex biases that are hard to remove. Even simple, known confounders such as preference for longer outputs remain in existing automated evaluation metrics. We propose a simple regression analysis approach for controlling biases in auto-evaluations. As a real case study, we focus on reducing the length bias of AlpacaEval, a fast and affordable benchmark for chat LLMs that uses LLMs to estimate response quality. Despite being highly correlated with human preferences, AlpacaEval is known to favor models that generate longer outputs. We introduce a length-controlled AlpacaEval that aims to answer the counterfactual question:\"What would the preference be if the model's and baseline's output had the same length?\". To achieve this, we first fit a generalized linear model to predict the biased output of interest (auto-annotator preferences) based on the mediators we want to control for (length difference) and other relevant features. We then obtain length-controlled preferences by predicting preferences while conditioning the GLM with a zero difference in lengths. Length-controlling not only improves the robustness of the metric to manipulations in model verbosity, we also find that it increases the Spearman correlation with LMSYS' Chatbot Arena from 0.94 to 0.98. We release the code and leaderboard at https://tatsu-lab.github.io/alpaca_eval/ ."
                },
                {
                    "title": "Taming Throughput-Latency Tradeoff in LLM Inference with Sarathi-Serve",
                    "abstract": "Each LLM serving request goes through two phases. The first is prefill which processes the entire input prompt and produces the first output token and the second is decode which generates the rest of output tokens, one-at-a-time. Prefill iterations have high latency but saturate GPU compute due to parallel processing of the input prompt. In contrast, decode iterations have low latency but also low compute utilization because a decode iteration processes only a single token per request. This makes batching highly effective for decodes and consequently for overall throughput. However, batching multiple requests leads to an interleaving of prefill and decode iterations which makes it challenging to achieve both high throughput and low latency. We introduce an efficient LLM inference scheduler, Sarathi-Serve, to address this throughput-latency tradeoff. Sarathi-Serve introduces chunked-prefills which splits a prefill request into near equal sized chunks and creates stall-free schedules that adds new requests in a batch without pausing ongoing decodes. Stall-free scheduling unlocks the opportunity to improve throughput with large batch sizes while minimizing the effect of batching on latency. Furthermore, uniform batches in Sarathi-Serve ameliorate the imbalance between iterations resulting in minimal pipeline bubbles. Our techniques yield significant improvements in inference performance across models and hardware under tail latency constraints. For Mistral-7B on single A100 GPUs, we achieve 2.6x higher serving capacity and up to 3.7x higher serving capacity for the Yi-34B model on two A100 GPUs as compared to vLLM. When used with pipeline parallelism on Falcon-180B, Sarathi-Serve provides up to 5.6x gain in the end-to-end serving capacity. The source code for Sarathi-Serve is available at https://github.com/microsoft/sarathi-serve."
                },
                {
                    "title": "D\u00e9j\u00e0Vu: KV-cache Streaming for Fast, Fault-tolerant Generative LLM Serving",
                    "abstract": "Distributed LLM serving is costly and often underutilizes hardware accelerators due to three key challenges: bubbles in pipeline-parallel deployments caused by the bimodal latency of prompt and token processing, GPU memory overprovisioning, and long recovery times in case of failures. In this paper, we propose D\\'ej\\`aVu, a system to address all these challenges using a versatile and efficient KV cache streaming library (D\\'ej\\`aVuLib). Using D\\'ej\\`aVuLib, we propose and implement efficient prompt-token disaggregation to reduce pipeline bubbles, microbatch swapping for efficient GPU memory management, and state replication for fault-tolerance. We highlight the efficacy of these solutions on a range of large models across cloud deployments."
                },
                {
                    "title": "Inference without Interference: Disaggregate LLM Inference for Mixed Downstream Workloads",
                    "abstract": "Transformer-based large language model (LLM) inference serving is now the backbone of many cloud services. LLM inference consists of a prefill phase and a decode phase. However, existing LLM deployment practices often overlook the distinct characteristics of these phases, leading to significant interference. To mitigate interference, our insight is to carefully schedule and group inference requests based on their characteristics. We realize this idea in TetriInfer through three pillars. First, it partitions prompts into fixed-size chunks so that the accelerator always runs close to its computationsaturated limit. Second, it disaggregates prefill and decode instances so each can run independently. Finally, it uses a smart two-level scheduling algorithm augmented with predicted resource usage to avoid decode scheduling hotspots. Results show that TetriInfer improves time-to-first-token (TTFT), job completion time (JCT), and inference efficiency in turns of performance per dollar by a large margin, e.g., it uses 38% less resources all the while lowering average TTFT and average JCT by 97% and 47%, respectively."
                },
                {
                    "title": "DistServe: Disaggregating Prefill and Decoding for Goodput-optimized Large Language Model Serving",
                    "abstract": "DistServe improves the performance of large language models (LLMs) serving by disaggregating the prefill and decoding computation. Existing LLM serving systems colocate the two phases and batch the computation of prefill and decoding across all users and requests. We find that this strategy not only leads to strong prefill-decoding interferences but also couples the resource allocation and parallelism plans for both phases. LLM applications often emphasize individual latency for each phase: time to first token (TTFT) for the prefill phase and time per output token (TPOT) of each request for the decoding phase. In the presence of stringent latency requirements, existing systems have to prioritize one latency over the other, or over-provision compute resources to meet both. DistServe assigns prefill and decoding computation to different GPUs, hence eliminating prefill-decoding interferences. Given the application's TTFT and TPOT requirements, DistServe co-optimizes the resource allocation and parallelism strategy tailored for each phase. DistServe also places the two phases according to the serving cluster's bandwidth to minimize the communication caused by disaggregation. As a result, DistServe significantly improves LLM serving performance in terms of the maximum rate that can be served within both TTFT and TPOT constraints on each GPU. Our evaluations show that on various popular LLMs, applications, and latency requirements, DistServe can serve 7.4x more requests or 12.6x tighter SLO, compared to state-of-the-art systems, while staying within latency constraints for>90% of requests."
                },
                {
                    "title": "Fairness in Serving Large Language Models",
                    "abstract": "High-demand LLM inference services (e.g., ChatGPT and BARD) support a wide range of requests from short chat conversations to long document reading. To ensure that all client requests are processed fairly, most major LLM inference services have request rate limits, to ensure that no client can dominate the request queue. However, this rudimentary notion of fairness also results in under-utilization of the resources and poor client experience when there is spare capacity. While there is a rich literature on fair scheduling, serving LLMs presents new challenges due to their unpredictable request lengths and their unique batching characteristics on parallel accelerators. This paper introduces the definition of LLM serving fairness based on a cost function that accounts for the number of input and output tokens processed. To achieve fairness in serving, we propose a novel scheduling algorithm, the Virtual Token Counter (VTC), a fair scheduler based on the continuous batching mechanism. We prove a 2x tight upper bound on the service difference between two backlogged clients, adhering to the requirement of work-conserving. Through extensive experiments, we demonstrate the superior performance of VTC in ensuring fairness, especially in contrast to other baseline methods, which exhibit shortcomings under various conditions. The reproducible code is available at https://github.com/Ying1123/VTC-artifact"
                },
                {
                    "title": "Splitwise: Efficient Generative LLM Inference Using Phase Splitting",
                    "abstract": "Generative large language model (LLM) applications are growing rapidly, leading to large-scale deployments of expensive and power-hungry GPUs. Our characterization of LLM inference shows that each inference request undergoes two phases: a compute-intensive prompt computation phase and a memory intensive token generation phase, each with distinct latency, throughput, memory, and power characteristics. Despite state-of-the-art batching and scheduling, the token generation phase underutilizes compute resources. Unlike prompt computation, token generation does not need the compute capability of the latest GPUs and can be run with lower power and cost. Based on these insights, we propose Splitwise, a model deployment and scheduling technique that splits the two phases of LLM inference requests on to separate machines. Splitwise enables phase-specific resource management using hardware that is well suited for each phase. Request state is transferred efficiently between machines using optimized network libraries on the fast back-plane interconnects available in today\u2019s GPU clusters. Using Splitwise, we design homogeneous and heterogeneous LLM inference clusters optimized for throughput, cost, and power Compared to current designs, Splitwise clusters achieve up to $1.4 \\times$ higher throughput at $\\mathbf{2 0 \\%}$ lower cost. Alternatively, they can deliver $2.35 \\times$ more throughput under the same power and cost budgets."
                },
                {
                    "title": "A Long Way to Go: Investigating Length Correlations in RLHF",
                    "abstract": "Great success has been reported using Reinforcement Learning from Human Feedback (RLHF) to align large language models, with open preference datasets enabling wider experimentation, particularly for\"helpfulness\"in tasks like dialogue and web question answering. Alongside these improvements, however, RLHF also often drives models to produce longer outputs. This paper demonstrates, on three diverse settings, that optimizing for response length is, much more than previously thought, a significant factor behind RLHF. Studying the strategies RL optimization uses to maximize reward, we find improvements in reward to largely be driven by increasing response length, instead of other features. Indeed, we find that even a purely length-based reward reproduces most downstream RLHF improvements over supervised fine-tuned models. Testing a comprehensive set of length-countering interventions, we identify the dominant source of these biases to be reward models, which, by studying training dynamics, we find are non-robust and easily influenced by length biases in preference data."
                },
                {
                    "title": "LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset",
                    "abstract": "Studying how people interact with large language models (LLMs) in real-world scenarios is increasingly important due to their widespread use in various applications. In this paper, we introduce LMSYS-Chat-1M, a large-scale dataset containing one million real-world conversations with 25 state-of-the-art LLMs. This dataset is collected from 210K unique IP addresses in the wild on our Vicuna demo and Chatbot Arena website. We offer an overview of the dataset's content, including its curation process, basic statistics, and topic distribution, highlighting its diversity, originality, and scale. We demonstrate its versatility through four use cases: developing content moderation models that perform similarly to GPT-4, building a safety benchmark, training instruction-following models that perform similarly to Vicuna, and creating challenging benchmark questions. We believe that this dataset will serve as a valuable resource for understanding and advancing LLM capabilities. The dataset is publicly available at https://huggingface.co/datasets/lmsys/lmsys-chat-1m."
                },
                {
                    "title": "Efficient Memory Management for Large Language Model Serving with PagedAttention",
                    "abstract": "High throughput serving of large language models (LLMs) requires batching sufficiently many requests at a time. However, existing systems struggle because the key-value cache (KV cache) memory for each request is huge and grows and shrinks dynamically. When managed inefficiently, this memory can be significantly wasted by fragmentation and redundant duplication, limiting the batch size. To address this problem, we propose PagedAttention, an attention algorithm inspired by the classical virtual memory and paging techniques in operating systems. On top of it, we build vLLM, an LLM serving system that achieves (1) near-zero waste in KV cache memory and (2) flexible sharing of KV cache within and across requests to further reduce memory usage. Our evaluations show that vLLM improves the throughput of popular LLMs by 2--4\u00d7 with the same level of latency compared to the state-of-the-art systems, such as FasterTransformer and Orca. The improvement is more pronounced with longer sequences, larger models, and more complex decoding algorithms. vLLM's source code is publicly available at https://github.com/vllm-project/vllm."
                },
                {
                    "title": "S3: Increasing GPU Utilization during Generative Inference for Higher Throughput",
                    "abstract": "Generating texts with a large language model (LLM) consumes massive amounts of memory. Apart from the already-large model parameters, the key/value (KV) cache that holds information about previous tokens in a sequence can grow to be even larger than the model itself. This problem is exacerbated in one of the current LLM serving frameworks which reserves the maximum sequence length of memory for the KV cache to guarantee generating a complete sequence as they do not know the output sequence length. This restricts us to use a smaller batch size leading to lower GPU utilization and above all, lower throughput. We argue that designing a system with a priori knowledge of the output sequence can mitigate this problem. To this end, we propose S$^{3}$, which predicts the output sequence length, schedules generation queries based on the prediction to increase device resource utilization and throughput, and handle mispredictions. Our proposed method achieves 6.49$\\times$ throughput over those systems that assume the worst case for the output sequence length."
                },
                {
                    "title": "AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedback",
                    "abstract": "Large language models (LLMs) such as ChatGPT have seen widespread adoption due to their strong instruction-following abilities. Developing these LLMs involves a complex yet poorly understood workflow requiring training with human feedback. Replicating and understanding this instruction-following requires tackling three major challenges: the high cost of data collection, the lack of trustworthy evaluation, and the absence of reference method implementations. We address these challenges with AlpacaFarm, a simulator that enables research and development for learning from feedback at a low cost. First, we design LLM prompts to simulate human feedback that are 50x cheaper than crowdworkers and display high agreement with humans. Second, we propose an automatic evaluation and validate it against human instructions obtained on real-world interactions. Third, we contribute reference implementations for several methods (PPO, DPO, best-of-n, expert iteration, and more) that learn from pairwise feedback. Finally, as an end-to-end validation of AlpacaFarm, we train and evaluate eleven models on 10k pairs of real human feedback and show that rankings of models trained in AlpacaFarm match rankings of models trained on human data. As a demonstration of the research possible in AlpacaFarm, we find that methods that use a reward model can substantially improve over supervised fine-tuning and that our reference PPO implementation leads to a +10% improvement in win-rate against Davinci003. We release all components of AlpacaFarm at https://github.com/tatsu-lab/alpaca_farm."
                },
                {
                    "title": "Response Length Perception and Sequence Scheduling: An LLM-Empowered LLM Inference Pipeline",
                    "abstract": "Large language models (LLMs) have revolutionized the field of AI, demonstrating unprecedented capacity across various tasks. However, the inference process for LLMs comes with significant computational costs. In this paper, we propose an efficient LLM inference pipeline that harnesses the power of LLMs. Our approach begins by tapping into the potential of LLMs to accurately perceive and predict the response length with minimal overhead. By leveraging this information, we introduce an efficient sequence scheduling technique that groups queries with similar response lengths into micro-batches. We evaluate our approach on real-world instruction datasets using the LLaMA-based model, and our results demonstrate an impressive 86% improvement in inference throughput without compromising effectiveness. Notably, our method is orthogonal to other inference acceleration techniques, making it a valuable addition to many existing toolkits (e.g., FlashAttention, Quantization) for LLM inference."
                },
                {
                    "title": "Fast Distributed Inference Serving for Large Language Models",
                    "abstract": "Large language models (LLMs) power a new generation of interactive AI applications exemplified by ChatGPT. The interactive nature of these applications demands low latency for LLM inference. Existing LLM serving systems use run-to-completion processing for inference jobs, which suffers from head-of-line blocking and long latency. We present FastServe, a distributed inference serving system for LLMs. FastServe exploits the autoregressive pattern of LLM inference to enable preemption at the granularity of each output token. FastServe uses preemptive scheduling to minimize latency with a novel skip-join Multi-Level Feedback Queue scheduler. Based on the new semi-information-agnostic setting of LLM inference, the scheduler leverages the input length information to assign an appropriate initial queue for each arrival job to join. The higher priority queues than the joined queue are skipped to reduce demotions. We design an efficient GPU memory management mechanism that proactively offloads and uploads intermediate state between GPU memory and host memory for LLM inference. We build a system prototype of FastServe and experimental results show that compared to the state-of-the-art solution vLLM, FastServe improves the throughput by up to 31.4x and 17.9x under the same average and tail latency requirements, respectively."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "NeuralNDCG: Direct Optimisation of a Ranking Metric via Differentiable Relaxation of Sorting",
                    "abstract": "Learning to Rank (LTR) algorithms are usually evaluated using Information Retrieval metrics like Normalised Discounted Cumulative Gain (NDCG) or Mean Average Precision. As these metrics rely on sorting predicted items' scores (and thus, on items' ranks), their derivatives are either undefined or zero everywhere. This makes them unsuitable for gradient-based optimisation, which is the usual method of learning appropriate scoring functions. Commonly used LTR loss functions are only loosely related to the evaluation metrics, causing a mismatch between the optimisation objective and the evaluation criterion. In this paper, we address this mismatch by proposing NeuralNDCG, a novel differentiable approximation to NDCG. Since NDCG relies on the non-differentiable sorting operator, we obtain NeuralNDCG by relaxing that operator using NeuralSort, a differentiable approximation of sorting. As a result, we obtain a new ranking loss function which is an arbitrarily accurate approximation to the evaluation metric, thus closing the gap between the training and the evaluation of LTR models. We introduce two variants of the proposed loss function. Finally, the empirical evaluation shows that our proposed method outperforms previous work aimed at direct optimisation of NDCG and is competitive with the state-of-the-art methods."
                },
                {
                    "title": "DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter",
                    "abstract": "As Transfer Learning from large-scale pre-trained models becomes more prevalent in Natural Language Processing (NLP), operating these large models in on-the-edge and/or under constrained computational training or inference budgets remains challenging. In this work, we propose a method to pre-train a smaller general-purpose language representation model, called DistilBERT, which can then be fine-tuned with good performances on a wide range of tasks like its larger counterparts. While most prior work investigated the use of distillation for building task-specific models, we leverage knowledge distillation during the pre-training phase and show that it is possible to reduce the size of a BERT model by 40%, while retaining 97% of its language understanding capabilities and being 60% faster. To leverage the inductive biases learned by larger models during pre-training, we introduce a triple loss combining language modeling, distillation and cosine-distance losses. Our smaller, faster and lighter model is cheaper to pre-train and we demonstrate its capabilities for on-device computations in a proof-of-concept experiment and a comparative on-device study."
                },
                {
                    "title": "Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism",
                    "abstract": "Recent work in language modeling demonstrates that training large transformer models advances the state of the art in Natural Language Processing applications. However, very large models can be quite difficult to train due to memory constraints. In this work, we present our techniques for training very large transformer models and implement a simple, efficient intra-layer model parallel approach that enables training transformer models with billions of parameters. Our approach does not require a new compiler or library changes, is orthogonal and complimentary to pipeline model parallelism, and can be fully implemented with the insertion of a few communication operations in native PyTorch. We illustrate this approach by converging transformer based models up to 8.3 billion parameters using 512 GPUs. We sustain 15.1 PetaFLOPs across the entire application with 76% scaling efficiency when compared to a strong single GPU baseline that sustains 39 TeraFLOPs, which is 30% of peak FLOPs. To demonstrate that large language models can further advance the state of the art (SOTA), we train an 8.3 billion parameter transformer language model similar to GPT-2 and a 3.9 billion parameter model similar to BERT. We show that careful attention to the placement of layer normalization in BERT-like models is critical to achieving increased performance as the model size grows. Using the GPT-2 model we achieve SOTA results on the WikiText103 (10.8 compared to SOTA perplexity of 15.8) and LAMBADA (66.5% compared to SOTA accuracy of 63.2%) datasets. Our BERT model achieves SOTA results on the RACE dataset (90.9% compared to SOTA accuracy of 89.4%)."
                },
                {
                    "title": "Learning to Optimize Tensor Programs",
                    "abstract": "We introduce a learning-based framework to optimize tensor programs for deep learning workloads. Efficient implementations of tensor operators, such as matrix multiplication and high dimensional convolution, are key enablers of effective deep learning systems. However, existing systems rely on manually optimized libraries such as cuDNN where only a narrow range of server class GPUs are well-supported. The reliance on hardware-specific operator libraries limits the applicability of high-level graph optimizations and incurs significant engineering costs when deploying to new hardware targets. We use learning to remove this engineering burden. We learn domain-specific statistical cost models to guide the search of tensor operator implementations over billions of possible program variants. We further accelerate the search by effective model transfer across workloads. Experimental results show that our framework delivers performance competitive with state-of-the-art hand-tuned libraries for low-power CPU, mobile GPU, and server-class GPU."
                },
                {
                    "title": "Ranking Relevance in Yahoo Search",
                    "abstract": "Search engines play a crucial role in our daily lives. Relevance is the core problem of a commercial search engine. It has attracted thousands of researchers from both academia and industry and has been studied for decades. Relevance in a modern search engine has gone far beyond text matching, and now involves tremendous challenges. The semantic gap between queries and URLs is the main barrier for improving base relevance. Clicks help provide hints to improve relevance, but unfortunately for most tail queries, the click information is too sparse, noisy, or missing entirely. For comprehensive relevance, the recency and location sensitivity of results is also critical. In this paper, we give an overview of the solutions for relevance in the Yahoo search engine. We introduce three key techniques for base relevance -- ranking functions, semantic matching features and query rewriting. We also describe solutions for recency sensitive relevance and location sensitive relevance. This work builds upon 20 years of existing efforts on Yahoo search, summarizes the most recent advances and provides a series of practical relevance solutions. The performance reported is based on Yahoo's commercial search engine, where tens of billions of urls are indexed and served by the ranking system."
                },
                {
                    "title": "Learning to rank for recommender systems",
                    "abstract": "Recommender system aim at providing a personalized list of items ranked according to the preferences of the user, as such ranking methods are at the core of many recommendation algorithms. The topic of this tutorial focuses on the cutting-edge algorithmic development in the area of recommender systems. This tutorial will provide an in depth picture of the progress of ranking models in the field, summarizing the strengths and weaknesses of existing methods, and discussing open issues that could be promising for future research in the community. A qualitative and quantitative comparison between different models will be provided while we will also highlight recent developments in the areas of Reinforcement Learning."
                },
                {
                    "title": "From RankNet to LambdaRank to LambdaMART: An Overview",
                    "abstract": "LambdaMART is the boosted tree version of LambdaRank, which is based on RankNet. RankNet, LambdaRank, and LambdaMART have proven to be very successful algorithms for solving real world ranking problems: for example an ensemble of LambdaMART rankers won Track 1 of the 2010 Yahoo! Learning To Rank Challenge. The details of these algorithms are spread across several papers and reports, and so here we give a self-contained, detailed and complete description of them."
                },
                {
                    "title": "Listwise approach to learning to rank: theory and algorithm",
                    "abstract": "This paper aims to conduct a study on the listwise approach to learning to rank. The listwise approach learns a ranking function by taking individual lists as instances and minimizing a loss function defined on the predicted list and the ground-truth list. Existing work on the approach mainly focused on the development of new algorithms; methods such as RankCosine and ListNet have been proposed and good performances by them have been observed. Unfortunately, the underlying theory was not sufficiently studied so far. To amend the problem, this paper proposes conducting theoretical analysis of learning to rank algorithms through investigations on the properties of the loss functions, including consistency, soundness, continuity, differentiability, convexity, and efficiency. A sufficient condition on consistency for ranking is given, which seems to be the first such result obtained in related research. The paper then conducts analysis on three loss functions: likelihood loss, cosine loss, and cross entropy loss. The latter two were used in RankCosine and ListNet. The use of the likelihood loss leads to the development of a new listwise method called ListMLE, whose loss function offers better properties, and also leads to better experimental results."
                },
                {
                    "title": "SoftRank: optimizing non-smooth rank metrics",
                    "abstract": "We address the problem of learning large complex ranking functions. Most IR applications use evaluation metrics that depend only upon the ranks of documents. However, most ranking functions generate document scores, which are sorted to produce a ranking. Hence IR metrics are innately non-smooth with respect to the scores, due to the sort. Unfortunately, many machine learning algorithms require the gradient of a training objective in order to perform the optimization of the model parameters,and because IR metrics are non-smooth,we need to find a smooth proxy objective that can be used for training. We present a new family of training objectives that are derived from the rank distributions of documents, induced by smoothed scores. We call this approach SoftRank. We focus on a smoothed approximation to Normalized Discounted Cumulative Gain (NDCG), called SoftNDCG and we compare it with three other training objectives in the recent literature. We present two main results. First, SoftRank yields a very good way of optimizing NDCG. Second, we show that it is possible to achieve state of the art test set NDCG results by optimizing a soft NDCG objective on the training set with a different discount function"
                },
                {
                    "title": "McRank: Learning to Rank Using Multiple Classification and Gradient Boosting",
                    "abstract": "We cast the ranking problem as (1) multiple classification (\"Mc\") (2) multiple ordinal classification, which lead to computationally tractable learning algorithms for relevance ranking in Web search. We consider the DCG criterion (discounted cumulative gain), a standard quality measure in information retrieval. Our approach is motivated by the fact that perfect classifications result in perfect DCG scores and the DCG errors are bounded by classification errors. We propose using the Expected Relevance to convert class probabilities into ranking scores. The class probabilities are learned using a gradient boosting tree algorithm. Evaluations on large-scale datasets show that our approach can improve LambdaRank [5] and the regressions-based ranker [6], in terms of the (normalized) DCG scores. An efficient implementation of the boosting tree algorithm is also presented."
                },
                {
                    "title": "A regression framework for learning ranking functions using relative relevance judgments",
                    "abstract": "Effective ranking functions are an essential part of commercial search engines. We focus on developing a regression framework for learning ranking functions for improving relevance of search engines serving diverse streams of user queries. We explore supervised learning methodology from machine learning, and we distinguish two types of relevance judgments used as the training data: 1) absolute relevance judgments arising from explicit labeling of search results; and 2) relative relevance judgments extracted from user click throughs of search results or converted from the absolute relevance judgments. We propose a novel optimization framework emphasizing the use of relative relevance judgments. The main contribution is the development of an algorithm based on regression that can be applied to objective functions involving preference data, i.e., data indicating that a document is more relevant than another with respect to a query. Experimental results are carried out using data sets obtained from a commercial search engine. Our results show significant improvements of our proposed methods over some existing methods."
                },
                {
                    "title": "AdaRank: a boosting algorithm for information retrieval",
                    "abstract": "In this paper we address the issue of learning to rank for document retrieval. In the task, a model is automatically created with some training data and then is utilized for ranking of documents. The goodness of a model is usually evaluated with performance measures such as MAP (Mean Average Precision) and NDCG (Normalized Discounted Cumulative Gain). Ideally a learning algorithm would train a ranking model that could directly optimize the performance measures with respect to the training data. Existing methods, however, are only able to train ranking models by minimizing loss functions loosely related to the performance measures. For example, Ranking SVM and RankBoost train ranking models by minimizing classification errors on instance pairs. To deal with the problem, we propose a novel learning algorithm within the framework of boosting, which can minimize a loss function directly defined on the performance measures. Our algorithm, referred to as AdaRank, repeatedly constructs 'weak rankers' on the basis of reweighted training data and finally linearly combines the weak rankers for making ranking predictions. We prove that the training process of AdaRank is exactly that of enhancing the performance measure used. Experimental results on four benchmark datasets show that AdaRank significantly outperforms the baseline methods of BM25, Ranking SVM, and RankBoost."
                },
                {
                    "title": "Learning to rank: from pairwise approach to listwise approach",
                    "abstract": "The paper is concerned with learning to rank, which is to construct a model or a function for ranking objects. Learning to rank is useful for document retrieval, collaborative filtering, and many other applications. Several methods for learning to rank have been proposed, which take object pairs as 'instances' in learning. We refer to them as the pairwise approach in this paper. Although the pairwise approach offers advantages, it ignores the fact that ranking is a prediction task on list of objects. The paper postulates that learning to rank should adopt the listwise approach in which lists of objects are used as 'instances' in learning. The paper proposes a new probabilistic method for the approach. Specifically it introduces two probability models, respectively referred to as permutation probability and top k probability, to define a listwise loss function for learning. Neural Network and Gradient Descent are then employed as model and algorithm in the learning method. Experimental results on information retrieval show that the proposed listwise approach performs better than the pairwise approach."
                },
                {
                    "title": "Learning to Rank with Nonsmooth Cost Functions",
                    "abstract": "The quality measures used in information retrieval are particularly difficult to optimize directly, since they depend on the model scores only through the sorted order of the documents returned for a given query. Thus, the derivatives of the cost with respect to the model parameters are either zero, or are undefined. In this paper, we propose a class of simple, flexible algorithms, called LambdaRank, which avoids these difficulties by working with implicit cost functions. We describe LambdaRank using neural network models, although the idea applies to any differentiable function class. We give necessary and sufficient conditions for the resulting implicit cost function to be convex, and we show that the general method has a simple mechanical interpretation. We demonstrate significantly improved accuracy, over a state-of-the-art ranking algorithm, on several datasets. We also show that LambdaRank provides a method for significantly speeding up the training phase of that ranking algorithm. Although this paper is directed towards ranking, the proposed method can be extended to any non-smooth and multivariate cost functions."
                },
                {
                    "title": "Learning to rank using gradient descent",
                    "abstract": "We investigate using gradient descent methods for learning ranking functions; we propose a simple probabilistic cost function, and we introduce RankNet, an implementation of these ideas using a neural network to model the underlying ranking function. We present test results on toy data and on data from a commercial internet search engine."
                },
                {
                    "title": "Pranking with Ranking",
                    "abstract": "We discuss the problem of ranking instances. In our framework each instance is associated with a rank or a rating, which is an integer from 1 to k. Our goal is to find a rank-predict ion rule that assigns each instance a rank which is as close as possible to the instance's true rank. We describe a simple and efficient online algorithm, analyze its performance in the mistake bound model, and prove its correctness. We describe two sets of experiments, with synthetic data and with the EachMovie dataset for collaborative filtering. In the experiments we performed, our algorithm outperforms online algorithms for regression and classification applied to ranking."
                },
                {
                    "title": "An Efficient Boosting Algorithm for Combining Preferences",
                    "abstract": "We study the problem of learning to accurately rank a set of objects by combining a given collection of ranking or preference functions. This problem of combining preferences arises in several applications, such as that of combining the results of different search engines, or the \u201ccollaborative-\ufb01ltering\u201d problem of ranking movies for a user based on the movie rankings provided by other users. In this work, we begin by presenting a formal framework for this general problem. We then describe and analyze an ef\ufb01cient algorithm called RankBoost for combining preferences based on the boosting approach to machine learning. We give theoretical results describing the algorithm\u2019s behavior both on the training data, and on new test data not seen during training. We also describe an ef\ufb01cient implementation of the algorithm for a particular restricted but common case. We next discuss two experiments we carried out to assess the performance of RankBoost. In the \ufb01rst exper-iment, we used the algorithm to combine different web search strategies, each of which is a query expansion for a given domain. The second experiment is a collaborative-\ufb01ltering task for making movie recommendations."
                },
                {
                    "title": "Towards Efficient and Reliable LLM Serving: A Real-World Workload Study",
                    "abstract": "\u2014Large language models (LLMs), especially Generative Pretrained Transformer (GPT) models, have significantly advanced in the industry in recent years. However, these models\u2019 broader development faces considerable challenges due to high operational and deployment costs. This has led to active research in improving the hardware efficiency of LLMs. Yet, the characteristics of real-world LLM workloads are often overlooked in current optimizations of LLM serving systems. In this work, we find that the absence of reliable workload data for evaluating LLM serving systems impacts the quality of service (QoS) and reliability in industrial deployments. This paper introduces the first real-world trace dataset of LLM serving workloads, detailing user, system, and LLM behaviors. We analyze this trace, highlighting burstiness, request and response distributions, and focusing on the reliability of GPT services. Based on this, we have developed a benchmark suite that reflects our dataset\u2019s workload patterns, enabling performance evaluation of serving systems. This suite captures the core patterns of workload distributions, allowing for precise scaling of the workload dataset to match system sizes. Our evaluation uncovers a previously unrecognized vulnerability of LLM serving systems to short-term burstiness, particularly in common workload scenarios. We observe that GPU memory limitations, caused by the fluctuating nature of burstiness, lead to significant performance degradation in existing LLM serving systems. Beyond benchmarking, understanding these patterns is valuable for optimizing LLM workload management, enabling elastic hardware resource adjustments to varying workloads. To encourage further research, we have made the dataset and benchmark suite publicly available at https://github.com/HPMLL/BurstGPT."
                },
                {
                    "title": "Orca: A Distributed Serving System for Transformer-Based Generative Models",
                    "abstract": "Large-scale Transformer-based models trained for generation tasks (e.g., GPT-3) have recently attracted huge interest, emphasizing the need for system support for serving models in this family. Since these models generate a next token in an autoregressive manner, one has to run the model multiple times to process an inference request where each iteration of the model generates a single output token for the request. However, existing systems for inference serving do not perform well on this type of workload that has a multi-iteration characteristic, due to their inflexible scheduling mechanism that cannot change the current batch of requests being processed; requests that have finished earlier than other requests in a batch cannot return to the client, while newly arrived requests have to wait until the current batch completely finishes. In this paper, we propose iteration-level scheduling, a new scheduling mechanism that schedules execution at the granularity of iteration (instead of request) where the scheduler invokes the execution engine to run only a single iteration of the model on the batch. In addition, to apply batching and iteration-level scheduling to a Transformer model at the same time, we suggest selective batching, which applies batching only to a selected set of operations. Based on these two techniques, we have implemented a distributed serving system called ORCA, with additional designs for scalability to models with hundreds of billions of parameters. Our evaluation on a GPT-3 175B model shows that ORCA can significantly outperform NVIDIA FasterTransformer in terms of both latency and throughput: 36.9\u00d7 throughput improvement at the same level of latency."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively schedule requests for large language models (LLMs) to minimize latency and maximize throughput, given the challenges of predicting request generation lengths?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for enhancing the performance of LLM services, which are integral to many modern applications. Efficient scheduling can significantly improve user experience by reducing latency, thereby allowing more users to access services simultaneously. This research could lead to advancements in scheduling algorithms applicable to various domains beyond LLMs, influencing future research in resource management and optimization in machine learning systems.\n\n### [Question 3] - Why is it hard?\nThe primary challenge lies in the difficulty of accurately predicting the generation lengths of requests in real-time, which is essential for implementing shortest-job-first (SJF) or shortest-remaining-time-first (SRTF) scheduling. Naive approaches that assume fixed or known request lengths fail because they do not account for the variability in request sizes. Additionally, the complexity of dynamically ranking requests based on their expected lengths introduces technical and practical obstacles, such as the need for a reliable ranking model that can operate efficiently under high load.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has largely focused on static scheduling methods that do not adapt to the dynamic nature of LLM requests. The assumption that accurately predicting request lengths is impossible has limited exploration into alternative scheduling strategies. Existing solutions often overlook the potential of relative ordering of request lengths, which is a key aspect of our approach. Our method leverages the Kendall rank correlation coefficient to rank requests without needing precise length predictions, thus addressing a significant gap in prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves training a small auxiliary model (e.g., OPT-125M) to rank LLM requests based on their predicted generation lengths. We will use a dataset of LLM requests to train this model, and the Kendall rank correlation coefficient will serve as the metric to evaluate the similarity between predicted and actual schedules. The expected outcome is a significant reduction in average latency and improved throughput for LLM services, demonstrating that effective scheduling can be achieved without precise length predictions by focusing on relative request ordering."
            }
        },
        "author_data": {
            "fad95641-6f9e-485f-9329-162f87acafe9": {
                "pk": "fad95641-6f9e-485f-9329-162f87acafe9",
                "name": "Yichao Fu",
                "collaborators": [
                    "Shangde Gao",
                    "Ke Liu",
                    "Peter Bailis",
                    "Ion Stoica",
                    "Hao Zhang",
                    "Yuqiang Han"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Recommender Systems",
                    "Knowledge Distillation"
                ],
                "publications": [
                    {
                        "title": "Break the Sequential Dependency of LLM Inference Using Lookahead Decoding",
                        "abstract": "Autoregressive decoding of large language models (LLMs) is memory bandwidth bounded, resulting in high latency and significant wastes of the parallel processing power of modern accelerators. Existing methods for accelerating LLM decoding often require a draft model (e.g., speculative decoding), which is nontrivial to obtain and unable to generalize. In this paper, we introduce Lookahead decoding, an exact, parallel decoding algorithm that accelerates LLM decoding without needing auxiliary models or data stores. It allows trading per-step log(FLOPs) to reduce the number of total decoding steps, is more parallelizable on single or multiple modern accelerators, and is compatible with concurrent memory-efficient attention (e.g., FlashAttention). Our implementation of Lookahead decoding can speed up autoregressive decoding by up to 1.8x on MT-bench and 4x with strong scaling on multiple GPUs in code completion tasks. Our code is avialable at https://github.com/hao-ai-lab/LookaheadDecoding"
                    },
                    {
                        "title": "Contrastive Knowledge Amalgamation for Unsupervised Image Classification",
                        "abstract": "Knowledge amalgamation (KA) aims to learn a compact student model to handle the joint objective from multiple teacher models that are are specialized for their own tasks respectively. Current methods focus on coarsely aligning teachers and students in the common representation space, making it difficult for the student to learn the proper decision boundaries from a set of heterogeneous teachers. Besides, the KL divergence in previous works only minimizes the probability distribution difference between teachers and the student, ignoring the intrinsic characteristics of teachers. Therefore, we propose a novel Contrastive Knowledge Amalgamation (CKA) framework, which introduces contrastive losses and an alignment loss to achieve intra-class cohesion and inter-class separation.Contrastive losses intra- and inter- models are designed to widen the distance between representations of different classes. The alignment loss is introduced to minimize the sample-level distribution differences of teacher-student models in the common representation space.Furthermore, the student learns heterogeneous unsupervised classification tasks through soft targets efficiently and flexibly in the task-level amalgamation. Extensive experiments on benchmarks demonstrate the generalization capability of CKA in the amalgamation of specific task as well as multiple tasks. Comprehensive ablation studies provide a further insight into our CKA."
                    },
                    {
                        "title": "Unified Matrix Factorization with Dynamic Multi-view Clustering",
                        "abstract": "Matrix factorization (MF) is a classical collaborative filtering algorithm for recommender systems. It decomposes the user-item interaction matrix into a product of low-dimensional user representation matrix and item representation matrix. In typical recommendation scenarios, the user-item interaction paradigm is usually a two-stage process and requires static clustering analysis of the obtained user and item representations. The above process, however, is time and computationally intensive, making it difficult to apply in real-time to e-commerce or Internet of Things environments with billions of users and trillions of items. To address this, we propose a unified matrix factorization method based on dynamic multi-view clustering (MFDMC) that employs an end-to-end training paradigm. Specifically, in each view, a user/item representation is regarded as a weighted projection of all clusters. The representation of each cluster is learnable, enabling the dynamic discarding of bad clusters. Furthermore, we employ multi-view clustering to represent multiple roles of users/items, effectively utilizing the representation space and improving the interpretability of the user/item representations for downstream tasks. Extensive experiments show that our proposed MFDMC achieves state-of-the-art performance on real-world recommendation datasets. Additionally, comprehensive visualization and ablation studies interpretably confirm that our method provides meaningful representations for downstream tasks of users/items."
                    }
                ]
            },
            "d2093d68-e310-4c2e-87c2-c72b036caf26": {
                "pk": "d2093d68-e310-4c2e-87c2-c72b036caf26",
                "name": "Siqi Zhu",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "ef4828b1-91b2-4a90-85bc-0af148e8e56c": {
                "pk": "ef4828b1-91b2-4a90-85bc-0af148e8e56c",
                "name": "Runlong Su",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "92b1aa82-362e-4620-a312-a438d7c18200": {
                "pk": "92b1aa82-362e-4620-a312-a438d7c18200",
                "name": "Aurick Qiao",
                "collaborators": [
                    "Eric P. Xing",
                    "Zhewei Yao",
                    "Yuxiong He",
                    "Willie Neiswanger",
                    "Samyam Rajbhandari",
                    "Bryon Aragam",
                    "Bingjing Zhang",
                    "Jaeseong Lee",
                    "seung-won hwang",
                    "Daniel F Campos"
                ],
                "domain": [
                    "Large Language Models",
                    "Machine Learning",
                    "Model Optimization",
                    "Open Source AI"
                ],
                "publications": [
                    {
                        "title": "SwiftKV: Fast Prefill-Optimized Inference with Knowledge-Preserving Model Transformation",
                        "abstract": "LLM inference for popular enterprise use cases, such as summarization, RAG, and code-generation, typically observes orders of magnitude longer prompt lengths than generation lengths. This characteristic leads to high cost of prefill and increased response latency. In this paper, we present SwiftKV, a novel model transformation and distillation procedure specifically designed to reduce the time and cost of processing prompt tokens while preserving high quality of generated tokens. SwiftKV combines three key mechanisms: i) SingleInputKV, which prefills later layers' KV cache using a much earlier layer's output, allowing prompt tokens to skip much of the model computation, ii) AcrossKV, which merges the KV caches of neighboring layers to reduce the memory footprint and support larger batch size for higher throughput, and iii) a knowledge-preserving distillation procedure that can adapt existing LLMs for SwiftKV with minimal accuracy impact and low compute and data requirement. For Llama-3.1-8B and 70B, SwiftKV reduces the compute requirement of prefill by 50% and the memory requirement of the KV cache by 62.5% while incurring minimum quality degradation across a wide range of tasks. In the end-to-end inference serving using an optimized vLLM implementation, SwiftKV realizes up to 2x higher aggregate throughput and 60% lower time per output token. It can achieve a staggering 560 TFlops/GPU of normalized inference throughput, which translates to 16K tokens/s for Llama-3.1-70B in 16-bit precision on 4x H100 GPUs."
                    },
                    {
                        "title": "Fault Tolerance in Iterative-Convergent Machine Learning",
                        "abstract": "Machine learning (ML) training algorithms often possess an inherent self-correcting behavior due to their iterative-convergent nature. Recent systems exploit this property to achieve adaptability and efficiency in unreliable computing environments by relaxing the consistency of execution and allowing calculation errors to be self-corrected during training. However, the behavior of such systems are only well understood for specific types of calculation errors, such as those caused by staleness, reduced precision, or asynchronicity, and for specific types of training algorithms, such as stochastic gradient descent. In this paper, we develop a general framework to quantify the effects of calculation errors on iterative-convergent algorithms and use this framework to design new strategies for checkpoint-based fault tolerance. Our framework yields a worst-case upper bound on the iteration cost of arbitrary perturbations to model parameters during training. Our system, SCAR, employs strategies which reduce the iteration cost upper bound due to perturbations incurred when recovering from checkpoints. We show that SCAR can reduce the iteration cost of partial failures by 78% - 95% when compared with traditional checkpoint-based fault tolerance across a variety of ML models and training algorithms."
                    },
                    {
                        "title": "STUN: Structured-Then-Unstructured Pruning for Scalable MoE Pruning",
                        "abstract": "Mixture-of-experts (MoEs) have been adopted for reducing inference costs by sparsely activating experts in Large language models (LLMs). Despite this reduction, the massive number of experts in MoEs still makes them expensive to serve. In this paper, we study how to address this, by pruning MoEs. Among pruning methodologies, unstructured pruning has been known to achieve the highest performance for a given pruning ratio, compared to structured pruning, since the latter imposes constraints on the sparsification structure. This is intuitive, as the solution space of unstructured pruning subsumes that of structured pruning. However, our counterintuitive finding reveals that expert pruning, a form of structured pruning, can actually precede unstructured pruning to outperform unstructured-only pruning. As existing expert pruning, requiring $O(\\frac{k^n}{\\sqrt{n}})$ forward passes for $n$ experts, cannot scale for recent MoEs, we propose a scalable alternative with $O(1)$ complexity, yet outperforming the more expensive methods. The key idea is leveraging a latent structure between experts, based on behavior similarity, such that the greedy decision of whether to prune closely captures the joint pruning effect. Ours is highly effective -- for Snowflake Arctic, a 480B-sized MoE with 128 experts, our method needs only one H100 and two hours to achieve nearly no loss in performance with 40% sparsity, even in generative tasks such as GSM8K, where state-of-the-art unstructured pruning fails to. The code will be made publicly available."
                    },
                    {
                        "title": "Pollux: Co-adaptive Cluster Scheduling for Goodput-Optimized Deep Learning",
                        "abstract": "Pollux improves scheduling performance in deep learning (DL) clusters by adaptively co-optimizing inter-dependent factors both at the per-job level and at the cluster-wide level. Most existing schedulers expect users to specify the number of resources for each job, often leading to inefficient resource use. Some recent schedulers choose job resources for users, but do so without awareness of how DL training can be re-optimized to better utilize the provided resources.   Pollux simultaneously considers both aspects. By monitoring the status of each job during training, Pollux models how their goodput (a novel metric we introduce that combines system throughput with statistical efficiency) would change by adding or removing resources. Leveraging these information, Pollux dynamically (re-)assigns resources to improve cluster-wide goodput, while respecting fairness and continually optimizing each DL job to better utilize those resources.   In experiments with real DL jobs and with trace-driven simulations, Pollux reduces average job completion times by 37-50% relative to state-of-the-art DL schedulers, even when they are provided with ideal resource and training configurations for every job. Pollux promotes fairness among DL jobs competing for resources based on a more meaningful measure of useful job progress, and reveals a new opportunity for reducing DL cost in cloud environments. Pollux is implemented and publicly available as part of an open-source project at https://github.com/petuum/adaptdl."
                    },
                    {
                        "title": "LLM360: Towards Fully Transparent Open-Source LLMs",
                        "abstract": "The recent surge in open-source Large Language Models (LLMs), such as LLaMA, Falcon, and Mistral, provides diverse options for AI practitioners and researchers. However, most LLMs have only released partial artifacts, such as the final model weights or inference code, and technical reports increasingly limit their scope to high-level design choices and surface statistics. These choices hinder progress in the field by degrading transparency into the training of LLMs and forcing teams to rediscover many details in the training process. We present LLM360, an initiative to fully open-source LLMs, which advocates for all training code and data, model checkpoints, and intermediate results to be made available to the community. The goal of LLM360 is to support open and collaborative AI research by making the end-to-end LLM training process transparent and reproducible by everyone. As a first step of LLM360, we release two 7B parameter LLMs pre-trained from scratch, Amber and CrystalCoder, including their training code, data, intermediate checkpoints, and analyses (at https://www.llm360.ai). We are committed to continually pushing the boundaries of LLMs through this open-source effort. More large-scale and stronger models are underway and will be released in the future."
                    }
                ]
            },
            "643741cc-a55b-490e-81ef-edaea9744b04": {
                "pk": "643741cc-a55b-490e-81ef-edaea9744b04",
                "name": "Ion Stoica",
                "collaborators": [
                    "Sanjay Krishnan",
                    "Ken Goldberg",
                    "Daniel Kang",
                    "Tatsunori Hashimoto",
                    "Yi Sun",
                    "Eric Liang",
                    "Roy Fox",
                    "Peter Bailis",
                    "Hao Zhang",
                    "Scott Shenker"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Machine Learning",
                    "Distributed Systems",
                    "Zero-Knowledge Proofs"
                ],
                "publications": [
                    {
                        "title": "Multi-Level Discovery of Deep Options",
                        "abstract": "Augmenting an agent's control with useful higher-level behaviors called options can greatly reduce the sample complexity of reinforcement learning, but manually designing options is infeasible in high-dimensional and abstract state spaces. While recent work has proposed several techniques for automated option discovery, they do not scale to multi-level hierarchies and to expressive representations such as deep networks. We present Discovery of Deep Options (DDO), a policy-gradient algorithm that discovers parametrized options from a set of demonstration trajectories, and can be used recursively to discover additional levels of the hierarchy. The scalability of our approach to multi-level hierarchies stems from the decoupling of low-level option discovery from high-level meta-control policy learning, facilitated by under-parametrization of the high level. We demonstrate that using the discovered options to augment the action space of Deep Q-Network agents can accelerate learning by guiding exploration in tasks where random actions are unlikely to reach valuable states. We show that DDO is effective in adding options that accelerate learning in 4 out of 5 Atari RAM environments chosen in our experiments. We also show that DDO can discover structure in robot-assisted surgical videos and kinematics that match expert annotation with 72% accuracy."
                    },
                    {
                        "title": "Scaling up Trustless DNN Inference with Zero-Knowledge Proofs",
                        "abstract": "As ML models have increased in capabilities and accuracy, so has the complexity of their deployments. Increasingly, ML model consumers are turning to service providers to serve the ML models in the ML-as-a-service (MLaaS) paradigm. As MLaaS proliferates, a critical requirement emerges: how can model consumers verify that the correct predictions were served, in the face of malicious, lazy, or buggy service providers?   In this work, we present the first practical ImageNet-scale method to verify ML model inference non-interactively, i.e., after the inference has been done. To do so, we leverage recent developments in ZK-SNARKs (zero-knowledge succinct non-interactive argument of knowledge), a form of zero-knowledge proofs. ZK-SNARKs allows us to verify ML model execution non-interactively and with only standard cryptographic hardness assumptions. In particular, we provide the first ZK-SNARK proof of valid inference for a full resolution ImageNet model, achieving 79\\% top-5 accuracy. We further use these ZK-SNARKs to design protocols to verify ML model execution in a variety of scenarios, including for verifying MLaaS predictions, verifying MLaaS model accuracy, and using ML models for trustless retrieval. Together, our results show that ZK-SNARKs have the promise to make verified ML model inference practical."
                    },
                    {
                        "title": "CodeScholar: Growing Idiomatic Code Examples",
                        "abstract": "Programmers often search for usage examples for API methods. A tool that could generate realistic, idiomatic, and contextual usage examples for one or more APIs would be immensely beneficial to developers. Such a tool would relieve the need for a deep understanding of the API landscape, augment existing documentation, and help discover interactions among APIs. We present CodeScholar, a tool that generates idiomatic code examples demonstrating the common usage of API methods. It includes a novel neural-guided search technique over graphs that grows the query APIs into idiomatic code examples. Our user study demonstrates that in 70% of cases, developers prefer CodeScholar generated examples over state-of-the-art large language models (LLM) like GPT3.5. We quantitatively evaluate 60 single and 25 multi-API queries from 6 popular Python libraries and show that across-the-board CodeScholar generates more realistic, diverse, and concise examples. In addition, we show that CodeScholar not only helps developers but also LLM-powered programming assistants generate correct code in a program synthesis setting."
                    },
                    {
                        "title": "Break the Sequential Dependency of LLM Inference Using Lookahead Decoding",
                        "abstract": "Autoregressive decoding of large language models (LLMs) is memory bandwidth bounded, resulting in high latency and significant wastes of the parallel processing power of modern accelerators. Existing methods for accelerating LLM decoding often require a draft model (e.g., speculative decoding), which is nontrivial to obtain and unable to generalize. In this paper, we introduce Lookahead decoding, an exact, parallel decoding algorithm that accelerates LLM decoding without needing auxiliary models or data stores. It allows trading per-step log(FLOPs) to reduce the number of total decoding steps, is more parallelizable on single or multiple modern accelerators, and is compatible with concurrent memory-efficient attention (e.g., FlashAttention). Our implementation of Lookahead decoding can speed up autoregressive decoding by up to 1.8x on MT-bench and 4x with strong scaling on multiple GPUs in code completion tasks. Our code is avialable at https://github.com/hao-ai-lab/LookaheadDecoding"
                    },
                    {
                        "title": "Neural Packet Classification",
                        "abstract": "Packet classification is a fundamental problem in computer networking. This problem exposes a hard tradeoff between the computation and state complexity, which makes it particularly challenging. To navigate this tradeoff, existing solutions rely on complex hand-tuned heuristics, which are brittle and hard to optimize. In this paper, we propose a deep reinforcement learning (RL) approach to solve the packet classification problem. There are several characteristics that make this problem a good fit for Deep RL. First, many of the existing solutions are iteratively building a decision tree by splitting nodes in the tree. Second, the effects of these actions (e.g., splitting nodes) can only be evaluated once we are done with building the tree. These two characteristics are naturally captured by the ability of RL to take actions that have sparse and delayed rewards. Third, it is computationally efficient to generate data traces and evaluate decision trees, which alleviate the notoriously high sample complexity problem of Deep RL algorithms. Our solution, NeuroCuts, uses succinct representations to encode state and action space, and efficiently explore candidate decision trees to optimize for a global objective. It produces compact decision trees optimized for a specific set of rules and a given performance metric, such as classification time, memory footprint, or a combination of the two. Evaluation on ClassBench shows that NeuroCuts outperforms existing hand-crafted algorithms in classification time by 18% at the median, and reduces both time and memory footprint by up to 3x."
                    },
                    {
                        "title": "ZK-IMG: Attested Images via Zero-Knowledge Proofs to Fight Disinformation",
                        "abstract": "Over the past few years, AI methods of generating images have been increasing in capabilities, with recent breakthroughs enabling high-resolution, photorealistic \"deepfakes\" (artificially generated images with the purpose of misinformation or harm). The rise of deepfakes has potential for social disruption. Recent work has proposed using ZK-SNARKs (zero-knowledge succinct non-interactive argument of knowledge) and attested cameras to verify that images were taken by a camera. ZK-SNARKs allow verification of image transformations non-interactively (i.e., post-hoc) with only standard cryptographic hardness assumptions. Unfortunately, this work does not preserve input privacy, is impractically slow (working only on 128$\\times$128 images), and/or requires custom cryptographic arguments.   To address these issues, we present zk-img, a library for attesting to image transformations while hiding the pre-transformed image. zk-img allows application developers to specify high level image transformations. Then, zk-img will transparently compile these specifications to ZK-SNARKs. To hide the input or output images, zk-img will compute the hash of the images inside the ZK-SNARK. We further propose methods of chaining image transformations securely and privately, which allows for arbitrarily many transformations. By combining these optimizations, zk-img is the first system to be able to transform HD images on commodity hardware, securely and privately."
                    },
                    {
                        "title": "DDCO: Discovery of Deep Continuous Options for Robot Learning from Demonstrations",
                        "abstract": "An option is a short-term skill consisting of a control policy for a specified region of the state space, and a termination condition recognizing leaving that region. In prior work, we proposed an algorithm called Deep Discovery of Options (DDO) to discover options to accelerate reinforcement learning in Atari games. This paper studies an extension to robot imitation learning, called Discovery of Deep Continuous Options (DDCO), where low-level continuous control skills parametrized by deep neural networks are learned from demonstrations. We extend DDO with: (1) a hybrid categorical-continuous distribution model to parametrize high-level policies that can invoke discrete options as well continuous control actions, and (2) a cross-validation method that relaxes DDO's requirement that users specify the number of options to be discovered. We evaluate DDCO in simulation of a 3-link robot in the vertical plane pushing a block with friction and gravity, and in two physical experiments on the da Vinci surgical robot, needle insertion where a needle is grasped and inserted into a silicone tissue phantom, and needle bin picking where needles and pins are grasped from a pile and categorized into bins. In the 3-link arm simulation, results suggest that DDCO can take 3x fewer demonstrations to achieve the same reward compared to a baseline imitation learning approach. In the needle insertion task, DDCO was successful 8/10 times compared to the next most accurate imitation learning baseline 6/10. In the surgical bin picking task, the learned policy successfully grasps a single object in 66 out of 99 attempted grasps, and in all but one case successfully recovered from failed grasps by retrying a second time."
                    },
                    {
                        "title": "Occupy the Cloud: Distributed Computing for the 99%",
                        "abstract": "Distributed computing remains inaccessible to a large number of users, in spite of many open source platforms and extensive commercial offerings. While distributed computation frameworks have moved beyond a simple map-reduce model, many users are still left to struggle with complex cluster management and configuration tools, even for running simple embarrassingly parallel jobs. We argue that stateless functions represent a viable platform for these users, eliminating cluster management overhead, fulfilling the promise of elasticity. Furthermore, using our prototype implementation, PyWren, we show that this model is general enough to implement a number of distributed computing models, such as BSP, efficiently. Extrapolating from recent trends in network bandwidth and the advent of disaggregated storage, we suggest that stateless functions are a natural fit for data processing in future computing environments."
                    },
                    {
                        "title": "Learning to Optimize Join Queries With Deep Reinforcement Learning",
                        "abstract": "Exhaustive enumeration of all possible join orders is often avoided, and most optimizers leverage heuristics to prune the search space. The design and implementation of heuristics are well-understood when the cost model is roughly linear, and we find that these heuristics can be significantly suboptimal when there are non-linearities in cost. Ideally, instead of a fixed heuristic, we would want a strategy to guide the search space in a more data-driven way---tailoring the search to a specific dataset and query workload. Recognizing the link between classical Dynamic Programming enumeration methods and recent results in Reinforcement Learning (RL), we propose a new method for learning optimized join search strategies. We present our RL-based DQ optimizer, which currently optimizes select-project-join blocks. We implement three versions of DQ to illustrate the ease of integration into existing DBMSes: (1) A version built on top of Apache Calcite, (2) a version integrated into PostgreSQL, and (3) a version integrated into SparkSQL. Our extensive evaluation shows that DQ achieves plans with optimization costs and query execution times competitive with the native query optimizer in each system, but can execute significantly faster after learning (often by orders of magnitude)."
                    },
                    {
                        "title": "Population Based Augmentation: Efficient Learning of Augmentation Policy Schedules",
                        "abstract": "A key challenge in leveraging data augmentation for neural network training is choosing an effective augmentation policy from a large search space of candidate operations. Properly chosen augmentation policies can lead to significant generalization improvements; however, state-of-the-art approaches such as AutoAugment are computationally infeasible to run for the ordinary user. In this paper, we introduce a new data augmentation algorithm, Population Based Augmentation (PBA), which generates nonstationary augmentation policy schedules instead of a fixed augmentation policy. We show that PBA can match the performance of AutoAugment on CIFAR-10, CIFAR-100, and SVHN, with three orders of magnitude less overall compute. On CIFAR-10 we achieve a mean test error of 1.46%, which is a slight improvement upon the current state-of-the-art. The code for PBA is open source and is available at https://github.com/arcelien/pba."
                    },
                    {
                        "title": "IMPACT: Importance Weighted Asynchronous Architectures with Clipped Target Networks",
                        "abstract": "The practical usage of reinforcement learning agents is often bottlenecked by the duration of training time. To accelerate training, practitioners often turn to distributed reinforcement learning architectures to parallelize and accelerate the training process. However, modern methods for scalable reinforcement learning (RL) often tradeoff between the throughput of samples that an RL agent can learn from (sample throughput) and the quality of learning from each sample (sample efficiency). In these scalable RL architectures, as one increases sample throughput (i.e. increasing parallelization in IMPALA), sample efficiency drops significantly. To address this, we propose a new distributed reinforcement learning algorithm, IMPACT. IMPACT extends IMPALA with three changes: a target network for stabilizing the surrogate objective, a circular buffer, and truncated importance sampling. In discrete action-space environments, we show that IMPACT attains higher reward and, simultaneously, achieves up to 30% decrease in training wall-time than that of IMPALA. For continuous control environments, IMPACT trains faster than existing scalable agents while preserving the sample efficiency of synchronous PPO."
                    },
                    {
                        "title": "MPC-Minimized Secure LLM Inference",
                        "abstract": "Many inference services based on large language models (LLMs) pose a privacy concern, either revealing user prompts to the service or the proprietary weights to the user. Secure inference offers a solution to this problem through secure multi-party computation (MPC), however, it is still impractical for modern LLM workload due to the large overhead imposed by MPC. To address this overhead, we propose Marill, a framework that adapts LLM fine-tuning to minimize MPC usage during secure inference. Marill introduces high-level architectural changes during fine-tuning that significantly reduce the number of expensive operations needed within MPC during inference, by removing some and relocating others outside MPC without compromising security. As a result, Marill-generated models are more efficient across all secure inference protocols and our approach complements MPC-friendly approximations for such operations. Compared to standard fine-tuning, Marill results in 3.6-11.3x better runtime and 2.4-6.9x better communication during secure inference across various MPC settings, while typically preserving over 90% performance across downstream tasks."
                    },
                    {
                        "title": "BlinkDB: Queries with Bounded Errors and Bounded Response Times on Very Large Data",
                        "abstract": "In this paper, we present BlinkDB, a massively parallel, sampling-based approximate query engine for running ad-hoc, interactive SQL queries on large volumes of data. The key insight that BlinkDB builds on is that one can often make reasonable decisions in the absence of perfect answers. For example, reliably detecting a malfunctioning server using a distributed collection of system logs does not require analyzing every request processed by the system. Based on this insight, BlinkDB allows one to trade-off query accuracy for response time, enabling interactive queries over massive data by running queries on data samples and presenting results annotated with meaningful error bars. To achieve this, BlinkDB uses two key ideas that differentiate it from previous work in this area: (1) an adaptive optimization framework that builds and maintains a set of multi-dimensional, multi-resolution samples from original data over time, and (2) a dynamic sample selection strategy that selects an appropriately sized sample based on a query's accuracy and/or response time requirements. We have built an open-source version of BlinkDB and validated its effectiveness using the well-known TPC-H benchmark as well as a real-world analytic workload derived from Conviva Inc. Our experiments on a 100 node cluster show that BlinkDB can answer a wide range of queries from a real-world query trace on up to 17 TBs of data in less than 2 seconds (over 100\\times faster than Hive), within an error of 2 - 10%."
                    },
                    {
                        "title": "SparkNet: Training Deep Networks in Spark",
                        "abstract": "Training deep networks is a time-consuming process, with networks for object recognition often requiring multiple days to train. For this reason, leveraging the resources of a cluster to speed up training is an important area of work. However, widely-popular batch-processing computational frameworks like MapReduce and Spark were not designed to support the asynchronous and communication-intensive workloads of existing distributed deep learning systems. We introduce SparkNet, a framework for training deep networks in Spark. Our implementation includes a convenient interface for reading data from Spark RDDs, a Scala interface to the Caffe deep learning framework, and a lightweight multi-dimensional tensor library. Using a simple parallelization scheme for stochastic gradient descent, SparkNet scales well with the cluster size and tolerates very high-latency communication. Furthermore, it is easy to deploy and use with no parameter tuning, and it is compatible with existing Caffe models. We quantify the dependence of the speedup obtained by SparkNet on the number of machines, the communication frequency, and the cluster's communication overhead, and we benchmark our system's performance on the ImageNet dataset."
                    },
                    {
                        "title": "Trustless Audits without Revealing Data or Models",
                        "abstract": "There is an increasing conflict between business incentives to hide models and data as trade secrets, and the societal need for algorithmic transparency. For example, a rightsholder wishing to know whether their copyrighted works have been used during training must convince the model provider to allow a third party to audit the model and data. Finding a mutually agreeable third party is difficult, and the associated costs often make this approach impractical.   In this work, we show that it is possible to simultaneously allow model providers to keep their model weights (but not architecture) and data secret while allowing other parties to trustlessly audit model and data properties. We do this by designing a protocol called ZkAudit in which model providers publish cryptographic commitments of datasets and model weights, alongside a zero-knowledge proof (ZKP) certifying that published commitments are derived from training the model. Model providers can then respond to audit requests by privately computing any function F of the dataset (or model) and releasing the output of F alongside another ZKP certifying the correct execution of F. To enable ZkAudit, we develop new methods of computing ZKPs for SGD on modern neural nets for simple recommender systems and image classification models capable of high accuracies on ImageNet. Empirically, we show it is possible to provide trustless audits of DNNs, including copyright, censorship, and counterfactual audits with little to no loss in accuracy."
                    },
                    {
                        "title": "Managing Bandwidth: The Key to Cloud-Assisted Autonomous Driving",
                        "abstract": "Prevailing wisdom asserts that one cannot rely on the cloud for critical real-time control systems like self-driving cars. We argue that we can, and must. Following the trends of increasing model sizes, improvements in hardware, and evolving mobile networks, we identify an opportunity to offload parts of time-sensitive and latency-critical compute to the cloud. Doing so requires carefully allocating bandwidth to meet strict latency SLOs, while maximizing benefit to the car."
                    },
                    {
                        "title": "Shark: SQL and Rich Analytics at Scale",
                        "abstract": "Shark is a new data analysis system that marries query processing with complex analytics on large clusters. It leverages a novel distributed memory abstraction to provide a unified engine that can run SQL queries and sophisticated analytics functions (e.g., iterative machine learning) at scale, and efficiently recovers from failures mid-query. This allows Shark to run SQL queries up to 100x faster than Apache Hive, and machine learning programs up to 100x faster than Hadoop. Unlike previous systems, Shark shows that it is possible to achieve these speedups while retaining a MapReduce-like execution engine, and the fine-grained fault tolerance properties that such engines provide. It extends such an engine in several ways, including column-oriented in-memory storage and dynamic mid-query replanning, to effectively execute SQL. The result is a system that matches the speedups reported for MPP analytic databases over MapReduce, while offering fault tolerance properties and complex analytics capabilities that they lack."
                    },
                    {
                        "title": "Highly Available Transactions: Virtues and Limitations (Extended Version)",
                        "abstract": "To minimize network latency and remain online during server failures and network partitions, many modern distributed data storage systems eschew transactional functionality, which provides strong semantic guarantees for groups of multiple operations over multiple data items. In this work, we consider the problem of providing Highly Available Transactions (HATs): transactional guarantees that do not suffer unavailability during system partitions or incur high network latency. We introduce a taxonomy of highly available systems and analyze existing ACID isolation and distributed data consistency guarantees to identify which can and cannot be achieved in HAT systems. This unifies the literature on weak transactional isolation, replica consistency, and highly available systems. We analytically and experimentally quantify the availability and performance benefits of HATs--often two to three orders of magnitude over wide-area networks--and discuss their necessary semantic compromises."
                    }
                ]
            },
            "d7834510-eddf-4477-9f02-56e060e2af1d": {
                "pk": "d7834510-eddf-4477-9f02-56e060e2af1d",
                "name": "Hao Zhang",
                "collaborators": [
                    "Yanbin Hao",
                    "Chong-Wah Ngo",
                    "Hao Wang",
                    "Zhen Kan",
                    "Masashi Sugiyama"
                ],
                "domain": [
                    "Machine Learning",
                    "Natural Language Processing",
                    "Graph Neural Network",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "The 750GeV diphoton excess: who introduces it?",
                        "abstract": "Recently, both ATLAS and CMS collaborations report an excess at 750GeV in the diphoton invariant mass spectrum at 13TeV LHC. If it is a new scalar produced via loop induced gluon-gluon fusion process, it is important to know what is the particle in the loop. In this work, we investigate the possibility of determine the fraction of the contribution from the standard model top-quark in the loop."
                    },
                    {
                        "title": "Spatial Process Approximations: Assessing Their Necessity",
                        "abstract": "In spatial statistics and machine learning, the kernel matrix plays a pivotal role in prediction, classification, and maximum likelihood estimation. A thorough examination reveals that for large sample sizes, the kernel matrix becomes ill-conditioned, provided the sampling locations are fairly evenly distributed. This condition poses significant challenges to numerical algorithms used in prediction and estimation computations and necessitates an approximation to prediction and the Gaussian likelihood. A review of current methodologies for managing large spatial data indicates that some fail to address this ill-conditioning problem. Such ill-conditioning often results in low-rank approximations of the stochastic processes. This paper introduces various optimality criteria and provides solutions for each."
                    },
                    {
                        "title": "A Novel Self-Packaged DBBPF With Multiple TZs for 5G Applications",
                        "abstract": "A self-packaged dual-band bandpass filter (DBBPF) with high isolation and low insertion loss (IL) for 5G applications is proposed in this paper. To get high stopband suppression, multiple and controllable transmission zeros (TZs) are produced. This novel DBBPF is designed with a pair of quarter-wavelength stepped-impedance resonators (QSIRs) and a half-wavelength hairpin resonator (HWHR). This DBBPF is excited by a pair of U-shape feed lines, which are designed on G6 to fully excite the resonators and to introduce source/load TZs at the same time. In this letter, the generation of two passbands and TZs will be discussed by separate electric and magnetic coupling paths (SEMCP) and mixed EM coupling analysis. This DBBPF achieves a low IL of 0.85/1.15 dB with the fractional bandwidths (FBW) of 11.0% and 6.9% at the center frequencies of 3.45 GHz and 4.9 GHz for 5G application, respectively. The total size is 0.32{\\lambda}g*0.45{\\lambda}g. Especially, three controllable TZs are introduced between two passbands."
                    },
                    {
                        "title": "Language Model Decomposition: Quantifying the Dependency and Correlation of Language Models",
                        "abstract": "Pre-trained language models (LMs), such as BERT (Devlin et al., 2018) and its variants, have led to significant improvements on various NLP tasks in past years. However, a theoretical framework for studying their relationships is still missing. In this paper, we fill this gap by investigating the linear dependency between pre-trained LMs. The linear dependency of LMs is defined analogously to the linear dependency of vectors. We propose Language Model Decomposition (LMD) to represent a LM using a linear combination of other LMs as basis, and derive the closed-form solution. A goodness-of-fit metric for LMD similar to the coefficient of determination is defined and used to measure the linear dependency of a set of LMs. In experiments, we find that BERT and eleven (11) BERT-like LMs are 91% linearly dependent. This observation suggests that current state-of-the-art (SOTA) LMs are highly \"correlated\". To further advance SOTA we need more diverse and novel LMs that are less dependent on existing LMs."
                    },
                    {
                        "title": "SpanRE: Entities and Overlapping Relations Extraction Based on Spans and Entity Attention",
                        "abstract": "Extracting entities and relations is an essential task of information extraction. Triplets extracted from a sentence might overlap with each other. Previous methods either did not address the overlapping issues or solved overlapping issues partially. To tackle triplet overlapping problems completely, firstly we extract candidate subjects with a standard span mechanism. Then we present a labeled span mechanism to extract the objects and relations simultaneously, we use the labeled span mechanism to generate labeled spans whose start and end positions indicate the objects, and whose labels correspond to relations of subject and objects. Besides, we design an entity attention mechanism to enhance the information fusion between subject and sentence during extracting objects and relations. We test our method on two public datasets, our method achieves the best performances on these two datasets."
                    },
                    {
                        "title": "Token Shift Transformer for Video Classification",
                        "abstract": "Transformer achieves remarkable successes in understanding 1 and 2-dimensional signals (e.g., NLP and Image Content Understanding). As a potential alternative to convolutional neural networks, it shares merits of strong interpretability, high discriminative power on hyper-scale data, and flexibility in processing varying length inputs. However, its encoders naturally contain computational intensive operations such as pair-wise self-attention, incurring heavy computational burden when being applied on the complex 3-dimensional video signals.   This paper presents Token Shift Module (i.e., TokShift), a novel, zero-parameter, zero-FLOPs operator, for modeling temporal relations within each transformer encoder. Specifically, the TokShift barely temporally shifts partial [Class] token features back-and-forth across adjacent frames. Then, we densely plug the module into each encoder of a plain 2D vision transformer for learning 3D video representation. It is worth noticing that our TokShift transformer is a pure convolutional-free video transformer pilot with computational efficiency for video understanding. Experiments on standard benchmarks verify its robustness, effectiveness, and efficiency. Particularly, with input clips of 8/12 frames, the TokShift transformer achieves SOTA precision: 79.83%/80.40% on the Kinetics-400, 66.56% on EGTEA-Gaze+, and 96.80% on UCF-101 datasets, comparable or better than existing SOTA convolutional counterparts. Our code is open-sourced in: https://github.com/VideoNetworks/TokShift-Transformer."
                    },
                    {
                        "title": "Exploiting Transformer in Sparse Reward Reinforcement Learning for Interpretable Temporal Logic Motion Planning",
                        "abstract": "Automaton based approaches have enabled robots to perform various complex tasks. However, most existing automaton based algorithms highly rely on the manually customized representation of states for the considered task, limiting its applicability in deep reinforcement learning algorithms. To address this issue, by incorporating Transformer into reinforcement learning, we develop a Double-Transformer-guided Temporal Logic framework (T2TL) that exploits the structural feature of Transformer twice, i.e., first encoding the LTL instruction via the Transformer module for efficient understanding of task instructions during the training and then encoding the context variable via the Transformer again for improved task performance. Particularly, the LTL instruction is specified by co-safe LTL. As a semantics-preserving rewriting operation, LTL progression is exploited to decompose the complex task into learnable sub-goals, which not only converts non-Markovian reward decision processes to Markovian ones, but also improves the sampling efficiency by simultaneous learning of multiple sub-tasks. An environment-agnostic LTL pre-training scheme is further incorporated to facilitate the learning of the Transformer module resulting in an improved representation of LTL. The simulation results demonstrate the effectiveness of the T2TL framework."
                    },
                    {
                        "title": "Task Selection for Bandit-Based Task Assignment in Heterogeneous Crowdsourcing",
                        "abstract": "Task selection (picking an appropriate labeling task) and worker selection (assigning the labeling task to a suitable worker) are two major challenges in task assignment for crowdsourcing. Recently, worker selection has been successfully addressed by the bandit-based task assignment (BBTA) method, while task selection has not been thoroughly investigated yet. In this paper, we experimentally compare several task selection strategies borrowed from active learning literature, and show that the least confidence strategy significantly improves the performance of task assignment in crowdsourcing."
                    }
                ]
            }
        }
    },
    "2410.04555": {
        "paper_data": {
            "title": "$\\texttt{dattri}$: A Library for Efficient Data Attribution",
            "url": "http://arxiv.org/abs/2410.04555v1",
            "arxiv_id": "2410.04555",
            "authors": [
                "Junwei Deng",
                "Ting-Wei Li",
                "Shiyuan Zhang",
                "Shixuan Liu",
                "Yijun Pan",
                "Hao Huang",
                "Xinhe Wang",
                "Pingbang Hu",
                "Xingjian Zhang",
                "Jiaqi W. Ma"
            ],
            "abstract": "Data attribution methods aim to quantify the influence of individual training samples on the prediction of artificial intelligence (AI) models. As training data plays an increasingly crucial role in the modern development of large-scale AI models, data attribution has found broad applications in improving AI performance and safety. However, despite a surge of new data attribution methods being developed recently, there lacks a comprehensive library that facilitates the development, benchmarking, and deployment of different data attribution methods. In this work, we introduce $\\texttt{dattri}$, an open-source data attribution library that addresses the above needs. Specifically, $\\texttt{dattri}$ highlights three novel design features. Firstly, $\\texttt{dattri}$ proposes a unified and easy-to-use API, allowing users to integrate different data attribution methods into their PyTorch-based machine learning pipeline with a few lines of code changed. Secondly, $\\texttt{dattri}$ modularizes low-level utility functions that are commonly used in data attribution methods, such as Hessian-vector product, inverse-Hessian-vector product or random projection, making it easier for researchers to develop new data attribution methods. Thirdly, $\\texttt{dattri}$ provides a comprehensive benchmark framework with pre-trained models and ground truth annotations for a variety of benchmark settings, including generative AI settings. We have implemented a variety of state-of-the-art efficient data attribution methods that can be applied to large-scale neural network models, and will continuously update the library in the future. Using the developed $\\texttt{dattri}$ library, we are able to perform a comprehensive and fair benchmark analysis across a wide range of data attribution methods. The source code of $\\texttt{dattri}$ is available at https://github.com/TRAIS-Lab/dattri.",
            "introduction": "   1 Introduction  Data attribution is a family of methods that aim to quantify the influence of individual training samples on the output of artificial intelligence (AI) models. As training data is becoming increasingly critical for the advancement of modern AI models, especially for large foundation models\u00a0[19], there has been a surge of data attribution methods developed recently\u00a0[22, 29, 36, 27]. These methods have found broad data-centric applications in improving the performance and safety of AI models, such as noisy label detection\u00a0[22], data selection\u00a0[9], and copyright compensation\u00a0[6].   However, a comprehensive infrastructural library that facilitates the development, benchmarking, and deployment of efficient data attribution methods is lacking. This absence hinders the standardization and acceleration of research in this area, creating inefficiencies and inconsistencies in how data attribution methods are developed and applied. Although there have been a couple of prior efforts\u00a0[33, 28, 18] for unifying APIs and benchmarking, many missing opportunities remain unaddressed, motivating the need for this work. We defer to Section\u00a02.2 for a detailed comparison between the existing libraries and our work.   In this work, we introduce dattri, an open-source data attribution library with the following design objectives. Firstly, for users deploying existing data attribution methods, we aim to provide a unified and user-friendly API across different methods. Specifically, our API design emphasizes minimal code invasion, i.e., allowing the data attribution methods to be integrated into most common PyTorch-based machine learning pipelines with only a few lines of code changed. This is non-trivial as data attribution methods often require access to internal information of the models, such as gradients or hidden representations. We achieve this goal by providing helper decorators to streamline the integration. Secondly, for researchers developing new data attribution methods, we aim to facilitate the development by providing efficient implementations of low-level utility functions, such as Hessian-vector product, inverse-Hessian-vector product or random projection. These functions are used by multiple existing data attribution methods and will likely be useful in the development of new methods. In dattri, we implement the data attribution methods in a carefully designed modular fashion, providing both a set of modularized low-level functions and examples of the usage of these functions. Finally, for both users and researchers, we aim to provide a benchmark suite that highlights a comprehensive list of evaluation metrics and diverse benchmark settings, including generative AI settings. In addition to the code for the evaluation metrics and benchmark experiments, we also provide the trained model checkpoints for each benchmark setting. Since some evaluation metrics for data attribution require hundreds or even thousands of model retraining, the provided trained model checkpoints could significantly reduce the computational burden of benchmark evaluation. The bolded features listed above are all novel designs in dattri compared to existing literature.   We have implemented a variety of data attribution methods, evaluation metrics, and benchmark settings in dattri. Our library currently covers four families of data attribution methods, each named after a representative method within its category. These families are: Influence Function (IF)\u00a0[22], TracIn\u00a0[29], Representer Point Selection (RPS)\u00a0[36], and TRAK\u00a0[27]. We have excluded certain other methods, notably the game-theoretic methods such as Data Shapley\u00a0[10, 17]",
            "references": [
                {
                    "title": "Efficient Ensembles Improve Training Data Attribution",
                    "abstract": "Training data attribution (TDA) methods aim to quantify the influence of individual training data points on the model predictions, with broad applications in data-centric AI, such as mislabel detection, data selection, and copyright compensation. However, existing methods in this field, which can be categorized as retraining-based and gradient-based, have struggled with the trade-off between computational efficiency and attribution efficacy. Retraining-based methods can accurately attribute complex non-convex models but are computationally prohibitive, while gradient-based methods are efficient but often fail for non-convex models. Recent research has shown that augmenting gradient-based methods with ensembles of multiple independently trained models can achieve significantly better attribution efficacy. However, this approach remains impractical for very large-scale applications. In this work, we discover that expensive, fully independent training is unnecessary for ensembling the gradient-based methods, and we propose two efficient ensemble strategies, DROPOUT ENSEMBLE and LORA ENSEMBLE, alternative to naive independent ensemble. These strategies significantly reduce training time (up to 80%), serving time (up to 60%), and space cost (up to 80%) while maintaining similar attribution efficacy to the naive independent ensemble. Our extensive experimental results demonstrate that the proposed strategies are effective across multiple TDA methods on diverse datasets and models, including generative settings, significantly advancing the Pareto frontier of TDA methods with better computational efficiency and attribution efficacy."
                },
                {
                    "title": "Rethinking Data Shapley for Data Selection Tasks: Misleads and Merits",
                    "abstract": "Data Shapley provides a principled approach to data valuation and plays a crucial role in data-centric machine learning (ML) research. Data selection is considered a standard application of Data Shapley. However, its data selection performance has shown to be inconsistent across settings in the literature. This study aims to deepen our understanding of this phenomenon. We introduce a hypothesis testing framework and show that Data Shapley's performance can be no better than random selection without specific constraints on utility functions. We identify a class of utility functions, monotonically transformed modular functions, within which Data Shapley optimally selects data. Based on this insight, we propose a heuristic for predicting Data Shapley's effectiveness in data selection tasks. Our experiments corroborate these findings, adding new insights into when Data Shapley may or may not succeed."
                },
                {
                    "title": "DsDm: Model-Aware Dataset Selection with Datamodels",
                    "abstract": "When selecting data for training large-scale models, standard practice is to filter for examples that match human notions of data quality. Such filtering yields qualitatively clean datapoints that intuitively should improve model behavior. However, in practice the opposite can often happen: we find that selecting according to similarity with\"high quality\"data sources may not increase (and can even hurt) performance compared to randomly selecting data. To develop better methods for selecting data, we start by framing dataset selection as an optimization problem that we can directly solve for: given target tasks, a learning algorithm, and candidate data, select the subset that maximizes model performance. This framework thus avoids handpicked notions of data quality, and instead models explicitly how the learning process uses train datapoints to predict on the target tasks. Our resulting method greatly improves language model (LM) performance on both pre-specified tasks and previously unseen tasks. Specifically, choosing target tasks representative of standard LM problems and evaluating on diverse held-out benchmarks, our selected datasets provide a 2x compute multiplier over baseline methods."
                },
                {
                    "title": "ACCESS: Advancing Innovation: NSF\u2019s Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support",
                    "abstract": "As the National Science Foundation evolves its investments in cyberinfrastructure, it has made a significant investment in the ACCESS (Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support) program instantiating a novel set of services along with a novel governance and management model. Research cyberinfrastructure (CI) is a key catalyst for discovery and innovation and plays a critical role in ensuring U.S. leadership in science and engineering, economic competitiveness, and national security, consistent with NSF\u2019s mission. Funding of a set of awards through the ACCESS program has established a suite of CI coordination services targeted at supporting a broad and diverse set of requirements, researchers, and usage modalities spanning all areas of science and engineering research and education complemented by support for the collective and coordinated operation of the overall ACCESS program."
                },
                {
                    "title": "OpenDataVal: a Unified Benchmark for Data Valuation",
                    "abstract": "Assessing the quality and impact of individual data points is critical for improving model performance and mitigating undesirable biases within the training dataset. Several data valuation algorithms have been proposed to quantify data quality, however, there lacks a systemic and standardized benchmarking system for data valuation. In this paper, we introduce OpenDataVal, an easy-to-use and unified benchmark framework that empowers researchers and practitioners to apply and compare various data valuation algorithms. OpenDataVal provides an integrated environment that includes (i) a diverse collection of image, natural language, and tabular datasets, (ii) implementations of eleven different state-of-the-art data valuation algorithms, and (iii) a prediction model API that can import any models in scikit-learn. Furthermore, we propose four downstream machine learning tasks for evaluating the quality of data values. We perform benchmarking analysis using OpenDataVal, quantifying and comparing the efficacy of state-of-the-art data valuation approaches. We find that no single algorithm performs uniformly best across all tasks, and an appropriate algorithm should be employed for a user's downstream task. OpenDataVal is publicly available at https://opendataval.github.io with comprehensive documentation. Furthermore, we provide a leaderboard where researchers can evaluate the effectiveness of their own data valuation algorithms."
                },
                {
                    "title": "TRAK: Attributing Model Behavior at Scale",
                    "abstract": "The goal of data attribution is to trace model predictions back to training data. Despite a long line of work towards this goal, existing approaches to data attribution tend to force users to choose between computational tractability and efficacy. That is, computationally tractable methods can struggle with accurately attributing model predictions in non-convex settings (e.g., in the context of deep neural networks), while methods that are effective in such regimes require training thousands of models, which makes them impractical for large models or datasets. In this work, we introduce TRAK (Tracing with the Randomly-projected After Kernel), a data attribution method that is both effective and computationally tractable for large-scale, differentiable models. In particular, by leveraging only a handful of trained models, TRAK can match the performance of attribution methods that require training thousands of models. We demonstrate the utility of TRAK across various modalities and scales: image classifiers trained on ImageNet, vision-language models (CLIP), and language models (BERT and mT5). We provide code for using TRAK (and reproducing our work) at https://github.com/MadryLab/trak ."
                },
                {
                    "title": "If Influence Functions are the Answer, Then What is the Question?",
                    "abstract": "Influence functions efficiently estimate the effect of removing a single training data point on a model's learned parameters. While influence estimates align well with leave-one-out retraining for linear models, recent works have shown this alignment is often poor in neural networks. In this work, we investigate the specific factors that cause this discrepancy by decomposing it into five separate terms. We study the contributions of each term on a variety of architectures and datasets and how they vary with factors such as network width and training time. While practical influence function estimates may be a poor match to leave-one-out retraining for nonlinear networks, we show they are often a good approximation to a different object we term the proximal Bregman response function (PBRF). Since the PBRF can still be used to answer many of the questions motivating influence functions, such as identifying influential or mislabeled examples, our results suggest that current algorithms for influence function estimation give more informative results than previous error analyses would suggest."
                },
                {
                    "title": "Data Banzhaf: A Robust Data Valuation Framework for Machine Learning",
                    "abstract": "Data valuation has wide use cases in machine learning, including improving data quality and creating economic incentives for data sharing. This paper studies the robustness of data valuation to noisy model performance scores. Particularly, we find that the inherent randomness of the widely used stochastic gradient descent can cause existing data value notions (e.g., the Shapley value and the Leave-one-out error) to produce inconsistent data value rankings across different runs. To address this challenge, we introduce the concept of safety margin, which measures the robustness of a data value notion. We show that the Banzhaf value, a famous value notion that originated from cooperative game theory literature, achieves the largest safety margin among all semivalues (a class of value notions that satisfy crucial properties entailed by ML applications and include the famous Shapley value and Leave-one-out error). We propose an algorithm to efficiently estimate the Banzhaf value based on the Maximum Sample Reuse (MSR) principle. Our evaluation demonstrates that the Banzhaf value outperforms the existing semivalue-based data value notions on several ML tasks such as learning with weighted samples and noisy label detection. Overall, our study suggests that when the underlying ML algorithm is stochastic, the Banzhaf value is a promising alternative to the other semivalue-based data value schemes given its computational advantage and ability to robustly differentiate data quality."
                },
                {
                    "title": "Datamodels: Predicting Predictions from Training Data",
                    "abstract": "We present a conceptual framework, datamodeling, for analyzing the behavior of a model class in terms of the training data. For any fixed\"target\"example $x$, training set $S$, and learning algorithm, a datamodel is a parameterized function $2^S \\to \\mathbb{R}$ that for any subset of $S' \\subset S$ -- using only information about which examples of $S$ are contained in $S'$ -- predicts the outcome of training a model on $S'$ and evaluating on $x$. Despite the potential complexity of the underlying process being approximated (e.g., end-to-end training and evaluation of deep neural networks), we show that even simple linear datamodels can successfully predict model outputs. We then demonstrate that datamodels give rise to a variety of applications, such as: accurately predicting the effect of dataset counterfactuals; identifying brittle predictions; finding semantically similar examples; quantifying train-test leakage; and embedding data into a well-behaved and feature-rich representation space. Data for this paper (including pre-computed datamodels as well as raw predictions from four million trained deep neural networks) is available at https://github.com/MadryLab/datamodels-data ."
                },
                {
                    "title": "Scaling Up Influence Functions",
                    "abstract": "We address efficient calculation of influence functions for tracking predictions back to the training data. We propose and analyze a new approach to speeding up the inverse Hessian calculation based on Arnoldi iteration. With this improvement, we achieve, to the best of our knowledge, the first successful implementation of influence functions that scales to full-size (language and vision) Transformer models with several hundreds of millions of parameters. We evaluate our approach in image classification and sequence-to-sequence tasks with tens to a hundred of millions of training examples. Our code is available at https://github.com/google-research/jax-influence."
                },
                {
                    "title": "Revisiting Methods for Finding Influential Examples",
                    "abstract": "Several instance-based explainability methods for finding influential training examples for test-time decisions have been proposed recently, including Influence Functions, TraceIn, Representer Point Selection, Grad-Dot, and Grad-Cos. Typically these methods are evaluated using LOO influence (Cook's distance) as a gold standard, or using various heuristics. In this paper, we show that all of the above methods are unstable, i.e., extremely sensitive to initialization, ordering of the training data, and batch size. We suggest that this is a natural consequence of how in the literature, the influence of examples is assumed to be independent of model state and other examples -- and argue it is not. We show that LOO influence and heuristics are, as a result, poor metrics to measure the quality of instance-based explanations, and instead propose to evaluate such explanations by their ability to detect poisoning attacks. Further, we provide a simple, yet effective baseline to improve all of the above methods and show how it leads to very significant improvements on downstream tasks."
                },
                {
                    "title": "Captum: A unified and generic model interpretability library for PyTorch",
                    "abstract": "In this paper we introduce a novel, unified, open-source model interpretability library for PyTorch [12]. The library contains generic implementations of a number of gradient and perturbation-based attribution algorithms, also known as feature, neuron and layer importance algorithms, as well as a set of evaluation metrics for these algorithms. It can be used for both classification and non-classification models including graph-structured models built on Neural Networks (NN). In this paper we give a high-level overview of supported attribution algorithms and show how to perform memory-efficient and scalable computations. We emphasize that the three main characteristics of the library are multimodality, extensibility and ease of use. Multimodality supports different modality of inputs such as image, text, audio or video. Extensibility allows adding new algorithms and features. The library is also designed for easy understanding and use. Besides, we also introduce an interactive visualization tool called Captum Insights that is built on top of Captum library and allows sample-based model debugging and visualization using feature importance metrics."
                },
                {
                    "title": "Influence Functions in Deep Learning Are Fragile",
                    "abstract": "Influence functions approximate the effect of training samples in test-time predictions and have a wide variety of applications in machine learning interpretability and uncertainty estimation. A commonly-used (first-order) influence function can be implemented efficiently as a post-hoc method requiring access only to the gradients and Hessian of the model. For linear models, influence functions are well-defined due to the convexity of the underlying loss function and are generally accurate even across difficult settings where model changes are fairly large such as estimating group influences. Influence functions, however, are not well-understood in the context of deep learning with non-convex loss functions. In this paper, we provide a comprehensive and large-scale empirical study of successes and failures of influence functions in neural network models trained on datasets such as Iris, MNIST, CIFAR-10 and ImageNet. Through our extensive experiments, we show that the network architecture, its depth and width, as well as the extent of model parameterization and regularization techniques have strong effects in the accuracy of influence functions. In particular, we find that (i) influence estimates are fairly accurate for shallow networks, while for deeper networks the estimates are often erroneous; (ii) for certain network architectures and datasets, training with weight-decay regularization is important to get high-quality influence estimates; and (iii) the accuracy of influence estimates can vary significantly depending on the examined test points. These results suggest that in general influence functions in deep learning are fragile and call for developing improved influence estimation methods to mitigate these issues in non-convex setups."
                },
                {
                    "title": "Estimating Training Data Influence by Tracking Gradient Descent",
                    "abstract": "We introduce a method called TrackIn that computes the influence of a training example on a prediction made by the model, by tracking how the loss on the test point changes during the training process whenever the training example of interest was utilized. We provide a scalable implementation of TrackIn via a combination of a few key ideas: (a) a first-order approximation to the exact computation, (b) using random projections to speed up the computation of the first-order approximation for large models, (c) using saved checkpoints of standard training procedures, and (d) cherry-picking layers of a deep neural network. An experimental evaluation shows that TrackIn is more effective in identifying mislabelled training examples than other related methods such as influence functions and representer points. We also discuss insights from applying the method on vision, regression and natural language tasks."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "Input Similarity from the Neural Network Perspective",
                    "abstract": "Given a trained neural network, we aim at understanding how similar it considers any two samples. For this, we express a proper definition of similarity from the neural network perspective (i.e. we quantify how undissociable two inputs A and B are), by taking a machine learning viewpoint: how much a parameter variation designed to change the output for A would impact the output for B as well? We study the mathematical properties of this similarity measure, and show how to estimate sample density with it, in low complexity, enabling new types of statistical analysis for neural networks. We also propose to use it during training, to enforce that examples known to be similar should also be seen as similar by the network. We then study the self-denoising phenomenon encountered in regression tasks when training neural networks on datasets with noisy labels. We exhibit a multimodal image registration task where almost perfect accuracy is reached, far beyond label noise variance. Such an impressive self-denoising phenomenon can be explained as a noise averaging effect over the labels of similar examples. We analyze data by retrieving samples perceived as similar by the network, and are able to quantify the denoising effect without requiring true labels."
                },
                {
                    "title": "Efficient Task-Specific Data Valuation for Nearest Neighbor Algorithms",
                    "abstract": "\n Given a data set\n D\n containing millions of data points and a data consumer who is willing to pay for $\n X\n to train a machine learning (ML) model over\n D\n ,\n how should we distribute this $X to each data point to reflect its \"value\"?\n In this paper, we define the \"relative value of data\" via the Shapley value, as it uniquely possesses properties with appealing real-world interpretations, such as fairness, rationality and decentralizability. For general, bounded utility functions, the Shapley value is known to be challenging to compute: to get Shapley values for all\n N\n data points, it requires\n O\n (2\n \n N\n \n ) model evaluations for exact computation and\n O\n (\n N\n log\n N\n ) for (\n \u03f5\n , \u03b4)-approximation.\n \n \n In this paper, we focus on one popular family of ML models relying on\n K\n -nearest neighbors (\n K\n NN). The most surprising result is that for unweighted\n K\n NN classifiers and regressors, the Shapley value of all\n N\n data points can be computed,\n exactly\n , in\n O\n (\n N\n log\n N\n ) time - an exponential improvement on computational complexity! Moreover, for (\n \u03f5\n , \u03b4)-approximation, we are able to develop an algorithm based on Locality Sensitive Hashing (LSH) with only sublinear complexity\n O\n (\n N\n \n h\n (\n \u03f5\n ,\n K\n )\n \n log\n N\n ) when\n \u03f5\n is not too small and\n K\n is not too large. We empirically evaluate our algorithms on up to 10 million data points and even our\n exact\n algorithm is up to three orders of magnitude faster than the baseline approximation algorithm. The LSH-based approximation algorithm can accelerate the value calculation process even further.\n \n \n We then extend our algorithm to other scenarios such as (1) weighed\n K\n NN classifiers, (2) different data points are clustered by different\n data curators\n , and (3) there are\n data analysts\n providing computation who also requires proper valuation.\n Some\n of these extensions, although also being improved exponentially, are less practical for exact computation (e.g.,\n O\n (\n \n N\n K\n \n ) complexity for weigthed\n K\n NN). We thus propose an Monte Carlo approximation algorithm, which is\n O\n (\n N\n (log\n N\n )\n 2\n /(log\n K\n )\n 2\n ) times more efficient than the baseline approximation algorithm.\n"
                },
                {
                    "title": "Data Shapley: Equitable Valuation of Data for Machine Learning",
                    "abstract": "As data becomes the fuel driving technological and economic growth, a fundamental challenge is how to quantify the value of data in algorithmic predictions and decisions. For example, in healthcare and consumer markets, it has been suggested that individuals should be compensated for the data that they generate, but it is not clear what is an equitable valuation for individual data. In this work, we develop a principled framework to address data valuation in the context of supervised machine learning. Given a learning algorithm trained on $n$ data points to produce a predictor, we propose data Shapley as a metric to quantify the value of each training datum to the predictor performance. Data Shapley value uniquely satisfies several natural properties of equitable data valuation. We develop Monte Carlo and gradient-based methods to efficiently estimate data Shapley values in practical settings where complex learning algorithms, including neural networks, are trained on large datasets. In addition to being equitable, extensive experiments across biomedical, image and synthetic data demonstrate that data Shapley has several other benefits: 1) it is more powerful than the popular leave-one-out or leverage score in providing insight on what data is more valuable for a given learning task; 2) low Shapley value data effectively capture outliers and corruptions; 3) high Shapley value data inform what type of new data to acquire to improve the predictor."
                },
                {
                    "title": "Towards Efficient Data Valuation Based on the Shapley Value",
                    "abstract": "{\\em ``How much is my data worth?''} is an increasingly common question posed by organizations and individuals alike. An answer to this question could allow, for instance, fairly distributing profits among multiple data contributors and determining prospective compensation when data breaches happen. In this paper, we study the problem of \\emph{data valuation} by utilizing the Shapley value, a popular notion of value which originated in coopoerative game theory. The Shapley value defines a unique payoff scheme that satisfies many desiderata for the notion of data value. However, the Shapley value often requires \\emph{exponential} time to compute.\nTo meet this challenge, we propose a repertoire of efficient algorithms for approximating the Shapley value. We also demonstrate the value of each training instance for various benchmark datasets."
                },
                {
                    "title": "Representer Point Selection for Explaining Deep Neural Networks",
                    "abstract": "We propose to explain the predictions of a deep neural network, by pointing to the set of what we call representer points in the training set, for a given test point prediction. Specifically, we show that we can decompose the pre-activation prediction of a neural network into a linear combination of activations of training points, with the weights corresponding to what we call representer values, which thus capture the importance of that training point on the learned parameters of the network. But it provides a deeper understanding of the network than simply training point influence: with positive representer values corresponding to excitatory training points, and negative values corresponding to inhibitory points, which as we show provides considerably more insight. Our method is also much more scalable, allowing for real-time feedback in a manner not feasible with influence functions."
                },
                {
                    "title": "Enabling Factorized Piano Music Modeling and Generation with the MAESTRO Dataset",
                    "abstract": "Generating musical audio directly with neural networks is notoriously difficult because it requires coherently modeling structure at many different timescales. Fortunately, most music is also highly structured and can be represented as discrete note events played on musical instruments. Herein, we show that by using notes as an intermediate representation, we can train a suite of models capable of transcribing, composing, and synthesizing audio waveforms with coherent musical structure on timescales spanning six orders of magnitude (~0.1 ms to ~100 s), a process we call Wave2Midi2Wave. This large advance in the state of the art is enabled by our release of the new MAESTRO (MIDI and Audio Edited for Synchronous TRacks and Organization) dataset, composed of over 172 hours of virtuosic piano performances captured with fine alignment (~3 ms) between note labels and audio waveforms. The networks and the dataset together present a promising approach toward creating new expressive and interpretable neural models of music."
                },
                {
                    "title": "Music Transformer: Generating Music with Long-Term Structure",
                    "abstract": "Music relies heavily on repetition to build structure and meaning. Self-reference occurs on multiple timescales, from motifs to phrases to reusing of entire sections of music, such as in pieces with ABA structure. The Transformer (Vaswani et al., 2017), a sequence model based on self-attention, has achieved compelling results in many generation tasks that require maintaining long-range coherence. This suggests that self-attention might also be well-suited to modeling music. In musical composition and performance, however, relative timing is critically important. Existing approaches for representing relative positional information in the Transformer modulate attention based on pairwise distance (Shaw et al., 2018). This is impractical for long sequences such as musical compositions since their memory complexity for intermediate relative information is quadratic in the sequence length. We propose an algorithm that reduces their intermediate memory requirement to linear in the sequence length. This enables us to demonstrate that a Transformer with our modified relative attention mechanism can generate minute-long compositions (thousands of steps, four times the length modeled in Oore et al., 2018) with compelling structure, generate continuations that coherently elaborate on a given motif, and in a seq2seq setup generate accompaniments conditioned on melodies. We evaluate the Transformer with our relative attention mechanism on two datasets, JSB Chorales and Piano-e-Competition, and obtain state-of-the-art results on the latter."
                },
                {
                    "title": "Explaining Explanations: An Overview of Interpretability of Machine Learning",
                    "abstract": "There has recently been a surge of work in explanatory artificial intelligence (XAI). This research area tackles the important problem that complex machines and algorithms often cannot provide insights into their behavior and thought processes. XAI allows users and parts of the internal system to be more transparent, providing explanations of their decisions in some level of detail. These explanations are important to ensure algorithmic fairness, identify potential bias/problems in the training data, and to ensure that the algorithms perform as expected. However, explanations produced by these systems is neither standardized nor systematically assessed. In an effort to create best practices and identify open challenges, we describe foundational concepts of explainability and show how they can be used to classify existing literature. We discuss why current approaches to explanatory methods especially for deep neural networks are insufficient. Finally, based on our survey, we conclude with suggested future research directions for explanatory artificial intelligence."
                },
                {
                    "title": "Understanding Black-box Predictions via Influence Functions",
                    "abstract": "How can we explain the predictions of a black-box model? In this paper, we use influence functions \u2014 a classic technique from robust statistics \u2014 to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. We show that even on non-convex and non-differentiable models where the theory breaks down, approximations to influence functions can still provide valuable information. On linear models and convolutional neural networks, we demonstrate that influence functions are useful for multiple purposes: understanding model behavior, debugging models, detecting dataset errors, and even creating visually-indistinguishable training-set attacks."
                },
                {
                    "title": "Second-Order Stochastic Optimization for Machine Learning in Linear Time",
                    "abstract": "First-order stochastic methods are the state-of-the-art in large-scale machine learning optimization owing to efficient per-iteration complexity. Second-order methods, while able to provide faster convergence, have been much less explored due to the high cost of computing the second-order information. In this paper we develop second-order stochastic methods for optimization problems in machine learning that match the per-iteration cost of gradient based methods, and in certain settings improve upon the overall running time over popular first-order methods. Furthermore, our algorithm has the desirable property of being implementable in time linear in the sparsity of the input data."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "The proof and measurement of association between two things.",
                    "abstract": "All knowledge\u2014beyond that of bare isolated occurrence\u2014deals with uniformities. Of the latter, some few have a claim to be considered absolute, such as mathematical implications and mechanical laws. But the vast majority are only partial; medicine does not teach that smallpox is inevitably escaped by vaccination, but that it is so generally; biology has not shown that all animals require organic food, but that nearly all do so; in daily life, a dark sky is no proof that it will rain, but merely a warning; even in morality, the sole categorical imperative alleged by Kant was the sinfulness of telling a lie, and few thinkers since have admitted so much as this to be valid universally. In psychology, more perhaps than in any other science, it is hard to find absolutely inflexible coincidences; occasionally, indeed, there appear uniformities sufficiently regular to be practically treated as laws, but infinitely the greater part of the observations hitherto recorded concern only more or less pronounced tendencies of one event or attribute to accompany another. Under these circumstances, one might well have expected that the evidential evaluation and precise mensuration of tendencies had long been the subject of exhaustive investigation and now formed one of the earliest sections in a beginner\u2019s psychological course. Instead, we find only a general na\u0131\u0308ve ignorance that there is anything about it requiring to be learnt. One after another, laborious series of experiments are executed and published with the purpose of demonstrating some connection between two events, wherein the otherwise learned psychologist reveals that his art of proving and measuring correspondence has not advanced beyond that of lay persons. The consequence has been that the significance of the experiments is not at all rightly understood, nor have any definite facts been elicited that may be either confirmed or refuted. The present article is a commencement at attempting to remedy this deficiency of scientific correlation. With this view, it will be strictly confined to the needs of practical workers, and all theoretical mathematical demonstrations will be omitted; it may, however, be said that the relations stated have already received a large amount of empirical verification. Great thanks are due from me to Professor Haussdorff and to Dr. G. Lipps, each of whom have supplied a useful theorem in polynomial probability; the former has also very kindly given valuable advice concerning the proof of the important formul\u00e6 for elimination of \u2018\u2018systematic deviations.\u2019\u2019 At the same time, and for the same reason, the meaning and working of the various formul\u00e6 have been explained sufficiently, it is hoped, to render them readily usable even by those whose knowledge of mathematics is elementary. The fundamental procedure is accompanied by simple imaginary examples, while the more advanced parts are illustrated by cases that have actually occurred in my personal experience. For more abundant and positive exemplification, the reader is requested to refer to the under cited research, which is entirely built upon the principles and mathematical relations here laid down. In conclusion, the general value of the methodics recommended is emphasized by a brief criticism of the best correlational work hitherto made public, and also the important question is discussed as to the number of \u2018\u2018cases\u2019\u2019 required for an experimental series."
                },
                {
                    "title": "The MNIST Database of Handwritten Digit Images for Machine Learning Research [Best of the Web]",
                    "abstract": "In this issue, \u201cBest of the Web\u201d presents the modified National Institute of Standards and Technology (MNIST) resources, consisting of a collection of handwritten digit images used extensively in optical character recognition and machine learning research."
                },
                {
                    "title": "Deep learning via Hessian-free optimization",
                    "abstract": "We develop a 2nd-order optimization method based on the \"Hessian-free\" approach, and apply it to training deep auto-encoders. Without using pre-training, we obtain results superior to those reported by Hinton & Salakhutdinov (2006) on the same tasks they considered. Our method is practical, easy to use, scales nicely to very large datasets, and isn't limited in applicability to auto-encoders, or any specific model class. We also discuss the issue of \"pathological curvature\" as a possible explanation for the difficulty of deep-learning and how 2nd-order optimization, and our method in particular, effectively deals with it."
                },
                {
                    "title": "Computational Copyright: Towards A Royalty Model for AI Music Generation Platforms",
                    "abstract": ","
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Gradient-based learning applied to document recognition",
                    "abstract": "Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CY"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we create a comprehensive and standardized open-source library for data attribution methods that facilitates their development, benchmarking, and deployment in machine learning applications?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing need for efficient and standardized data attribution methods, which are essential for improving AI model performance and safety. A unified library like dattri can accelerate research by providing a common framework for evaluating and comparing different methods, leading to advancements in data-centric AI applications such as noisy label detection and copyright compensation. This work could also inspire further innovations in data attribution techniques and their practical applications across various domains.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the complexity of integrating diverse data attribution methods into a single framework while maintaining user-friendliness and minimal code invasion. Naive approaches may fail due to the need for access to internal model information, such as gradients and hidden representations, which can vary significantly across different models. Additionally, developing efficient low-level utility functions that are applicable to multiple methods requires a deep understanding of the underlying mathematical principles and computational efficiency, making it a technically demanding task.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by a lack of comprehensive infrastructure for data attribution methods, resulting in fragmented efforts and inconsistent implementations. Existing libraries have not fully addressed the need for a unified API or a robust benchmarking suite, leaving many opportunities unexplored. Barriers such as the complexity of integrating various methods and the absence of standardized evaluation metrics have hindered progress. Our approach differs by providing a modular design, efficient utility functions, and a comprehensive benchmark suite, which collectively enhance the usability and effectiveness of data attribution methods.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing the dattri library, which includes a unified API for integrating data attribution methods into PyTorch-based pipelines, efficient implementations of low-level utility functions, and a comprehensive benchmark suite with diverse evaluation metrics. We will implement four families of data attribution methods: Influence Function, TracIn, Representer Point Selection, and TRAK. The expected outcomes include a user-friendly library that streamlines the application of data attribution methods, facilitates the development of new techniques, and provides a robust framework for benchmarking, ultimately advancing the field of data-centric AI research."
            }
        },
        "author_data": {
            "9a7d5b70-0f97-4546-8702-362f60b53d3d": {
                "pk": "9a7d5b70-0f97-4546-8702-362f60b53d3d",
                "name": "Junwei Deng",
                "collaborators": [
                    "Jiaqi Ma",
                    "Qiaozhu Mei",
                    "Shiyuan Zhang",
                    "Ting-Wei Li",
                    "Shichang Zhang",
                    "Xinhe Wang",
                    "Pingbang Hu",
                    "Jiaqi W. Ma"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Adversarial Machine Learning",
                    "Data Attribution",
                    "Generative AI"
                ],
                "publications": [
                    {
                        "title": "Adversarial Attack on Graph Neural Networks as An Influence Maximization Problem",
                        "abstract": "Graph neural networks (GNNs) have attracted increasing interests. With broad deployments of GNNs in real-world applications, there is an urgent need for understanding the robustness of GNNs under adversarial attacks, especially in realistic setups. In this work, we study the problem of attacking GNNs in a restricted and realistic setup, by perturbing the features of a small set of nodes, with no access to model parameters and model predictions. Our formal analysis draws a connection between this type of attacks and an influence maximization problem on the graph. This connection not only enhances our understanding on the problem of adversarial attack on GNNs, but also allows us to propose a group of effective and practical attack strategies. Our experiments verify that the proposed attack strategies significantly degrade the performance of three popular GNN models and outperform baseline adversarial attack strategies."
                    },
                    {
                        "title": "Subgroup Generalization and Fairness of Graph Neural Networks",
                        "abstract": "Despite enormous successful applications of graph neural networks (GNNs), theoretical understanding of their generalization ability, especially for node-level tasks where data are not independent and identically-distributed (IID), has been sparse. The theoretical investigation of the generalization performance is beneficial for understanding fundamental issues (such as fairness) of GNN models and designing better learning methods. In this paper, we present a novel PAC-Bayesian analysis for GNNs under a non-IID semi-supervised learning setup. Moreover, we analyze the generalization performances on different subgroups of unlabeled nodes, which allows us to further study an accuracy-(dis)parity-style (un)fairness of GNNs from a theoretical perspective. Under reasonable assumptions, we demonstrate that the distance between a test subgroup and the training set can be a key factor affecting the GNN performance on that subgroup, which calls special attention to the training node selection for fair learning. Experiments across multiple GNN models and datasets support our theoretical results."
                    },
                    {
                        "title": "Computational Copyright: Towards A Royalty Model for Music Generative AI",
                        "abstract": "The advancement of generative AI has given rise to pressing copyright challenges, especially within the music industry. This paper focuses on the economic aspects of these challenges, emphasizing that the economic impact constitutes a central issue in the copyright arena. Furthermore, the complexity of the black-box generative AI technologies not only suggests but necessitates algorithmic solutions. Yet, such solutions have been largely missing, exacerbating regulatory hurdles in this landscape. We seek to address this gap by proposing viable royalty models for revenue sharing on AI music generation platforms. We start by examining existing royalty models utilized by platforms like Spotify and YouTube, and then discuss how to adapt them to the unique context of AI-generated music. A significant challenge emerging from this adaptation is the attribution of AI-generated music to influential copyrighted content in the training data. To this end, we present algorithmic solutions employing data attribution techniques. We also conduct a range of experiments to verify the effectiveness and robustness of these solutions. This research is one of the early attempts to integrate technical advancements with economic and legal considerations in the field of music generative AI, offering a computational copyright solution for the challenges posed by the opaque nature of AI technologies."
                    },
                    {
                        "title": "Efficient Ensembles Improve Training Data Attribution",
                        "abstract": "Training data attribution (TDA) methods aim to quantify the influence of individual training data points on the model predictions, with broad applications in data-centric AI, such as mislabel detection, data selection, and copyright compensation. However, existing methods in this field, which can be categorized as retraining-based and gradient-based, have struggled with the trade-off between computational efficiency and attribution efficacy. Retraining-based methods can accurately attribute complex non-convex models but are computationally prohibitive, while gradient-based methods are efficient but often fail for non-convex models. Recent research has shown that augmenting gradient-based methods with ensembles of multiple independently trained models can achieve significantly better attribution efficacy. However, this approach remains impractical for very large-scale applications.   In this work, we discover that expensive, fully independent training is unnecessary for ensembling the gradient-based methods, and we propose two efficient ensemble strategies, DROPOUT ENSEMBLE and LORA ENSEMBLE, alternative to naive independent ensemble. These strategies significantly reduce training time (up to 80%), serving time (up to 60%), and space cost (up to 80%) while maintaining similar attribution efficacy to the naive independent ensemble. Our extensive experimental results demonstrate that the proposed strategies are effective across multiple TDA methods on diverse datasets and models, including generative settings, significantly advancing the Pareto frontier of TDA methods with better computational efficiency and attribution efficacy."
                    },
                    {
                        "title": "Adversarial Attacks on Data Attribution",
                        "abstract": "Data attribution aims to quantify the contribution of individual training data points to the outputs of an AI model, which has been used to measure the value of training data and compensate data providers. Given the impact on financial decisions and compensation mechanisms, a critical question arises concerning the adversarial robustness of data attribution methods. However, there has been little to no systematic research addressing this issue. In this work, we aim to bridge this gap by detailing a threat model with clear assumptions about the adversary's goal and capabilities and proposing principled adversarial attack methods on data attribution. We present two methods, Shadow Attack and Outlier Attack, which generate manipulated datasets to inflate the compensation adversarially. The Shadow Attack leverages knowledge about the data distribution in the AI applications, and derives adversarial perturbations through \"shadow training\", a technique commonly used in membership inference attacks. In contrast, the Outlier Attack does not assume any knowledge about the data distribution and relies solely on black-box queries to the target model's predictions. It exploits an inductive bias present in many data attribution methods - outlier data points are more likely to be influential - and employs adversarial examples to generate manipulated datasets. Empirically, in image classification and text generation tasks, the Shadow Attack can inflate the data-attribution-based compensation by at least 200%, while the Outlier Attack achieves compensation inflation ranging from 185% to as much as 643%."
                    }
                ]
            },
            "10546b42-bc71-4cf5-a838-c3fef3c7d475": {
                "pk": "10546b42-bc71-4cf5-a838-c3fef3c7d475",
                "name": "Ting-Wei Li",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "ae9a761d-6039-45c5-99f3-6174d8712bde": {
                "pk": "ae9a761d-6039-45c5-99f3-6174d8712bde",
                "name": "Shiyuan Zhang",
                "collaborators": [
                    "Fei Wang",
                    "Jinsong Han",
                    "Junwei Deng",
                    "Jiaqi Ma",
                    "Xu He",
                    "Dong Huang",
                    "Haoye Chai",
                    "Xiaoqian Qi",
                    "Yong Li",
                    "Jianwei Feng"
                ],
                "domain": [
                    "Generative AI",
                    "Mobile Traffic Forecasting",
                    "WiFi Sensing",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Computational Copyright: Towards A Royalty Model for Music Generative AI",
                        "abstract": "The advancement of generative AI has given rise to pressing copyright challenges, especially within the music industry. This paper focuses on the economic aspects of these challenges, emphasizing that the economic impact constitutes a central issue in the copyright arena. Furthermore, the complexity of the black-box generative AI technologies not only suggests but necessitates algorithmic solutions. Yet, such solutions have been largely missing, exacerbating regulatory hurdles in this landscape. We seek to address this gap by proposing viable royalty models for revenue sharing on AI music generation platforms. We start by examining existing royalty models utilized by platforms like Spotify and YouTube, and then discuss how to adapt them to the unique context of AI-generated music. A significant challenge emerging from this adaptation is the attribution of AI-generated music to influential copyrighted content in the training data. To this end, we present algorithmic solutions employing data attribution techniques. We also conduct a range of experiments to verify the effectiveness and robustness of these solutions. This research is one of the early attempts to integrate technical advancements with economic and legal considerations in the field of music generative AI, offering a computational copyright solution for the challenges posed by the opaque nature of AI technologies."
                    },
                    {
                        "title": "CSI-Net: Unified Human Body Characterization and Pose Recognition",
                        "abstract": "We build CSI-Net, a unified Deep Neural Network~(DNN), to learn the representation of WiFi signals. Using CSI-Net, we jointly solved two body characterization problems: biometrics estimation (including body fat, muscle, water, and bone rates) and person recognition. We also demonstrated the application of CSI-Net on two distinctive pose recognition tasks: the hand sign recognition (fine-scaled action of the hand) and falling detection (coarse-scaled motion of the body)."
                    },
                    {
                        "title": "FoMo: A Foundation Model for Mobile Traffic Forecasting with Diffusion Model",
                        "abstract": "Mobile traffic forecasting allows operators to anticipate network dynamics and performance in advance, offering substantial potential for enhancing service quality and improving user experience. However, existing models are often task-oriented and are trained with tailored data, which limits their effectiveness in diverse mobile network tasks of Base Station (BS) deployment, resource allocation, energy optimization, etc. and hinders generalization across different urban environments. Foundation models have made remarkable strides across various domains of NLP and CV due to their multi-tasking adaption and zero/few-shot learning capabilities. In this paper, we propose an innovative Foundation model for Mo}bile traffic forecasting (FoMo), aiming to handle diverse forecasting tasks of short/long-term predictions and distribution generation across multiple cities to support network planning and optimization. FoMo combines diffusion models and transformers, where various spatio-temporal masks are proposed to enable FoMo to learn intrinsic features of different tasks, and a contrastive learning strategy is developed to capture the correlations between mobile traffic and urban contexts, thereby improving its transfer learning capability. Extensive experiments on 9 real-world datasets demonstrate that FoMo outperforms current models concerning diverse forecasting tasks and zero/few-shot learning, showcasing a strong universality. We further deploy the FoMo on the JiuTian optimization platform of China Mobile, where we use the predicted mobile data to formulate network planning and optimization applications, including BS deployment, resource block scheduling, and BS sleep control."
                    },
                    {
                        "title": "Joint Activity Recognition and Indoor Localization with WiFi Fingerprints",
                        "abstract": "Recent years have witnessed the rapid development in the research topic of WiFi sensing that automatically senses human with commercial WiFi devices. This work falls into two major categories, i.e., the activity recognition and the indoor localization. The former work utilizes WiFi devices to recognize human daily activities such as smoking, walking, and dancing. The latter one, indoor localization, can be used for indoor navigation, location-based services, and through-wall surveillance. The key rationale behind this type of work is that people behaviors can influence the WiFi signal propagation and introduce specific patterns into WiFi signals, called WiFi fingerprints, which can be further explored to identify human activities and locations. In this paper, we propose a novel deep learning framework for joint activity recognition and indoor localization task using WiFi Channel State Information~(CSI) fingerprints. More precisely, we develop a system running standard IEEE 802.11n WiFi protocol, and collect more than 1400 CSI fingerprints on 6 activities at 16 indoor locations. Then we propose a dual-task convolutional neural network with 1-dimensional convolutional layers for the joint task of activity recognition and indoor localization. Experimental results and ablation study show that our approach achieves good performances in this joint WiFi sensing task. Data and code have been made publicly available at https://github.com/geekfeiw/apl."
                    }
                ]
            },
            "68a3bdc3-e9a7-49da-bfb0-0c15b1de473c": {
                "pk": "68a3bdc3-e9a7-49da-bfb0-0c15b1de473c",
                "name": "Shixuan Liu",
                "collaborators": [
                    "Changjun Fan",
                    "Zhong Liu",
                    "Kewei Cheng",
                    "Yunfei Wang",
                    "Peng Cui",
                    "Yizhou Sun",
                    "Bingxu Zhang",
                    "Kuihua Huang",
                    "Xiang Zhao",
                    "Jincai Huang"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Heterogeneous Information Networks",
                    "Expressive Power"
                ],
                "publications": [
                    {
                        "title": "Inductive Meta-path Learning for Schema-complex Heterogeneous Information Networks",
                        "abstract": "Heterogeneous Information Networks (HINs) are information networks with multiple types of nodes and edges. The concept of meta-path, i.e., a sequence of entity types and relation types connecting two entities, is proposed to provide the meta-level explainable semantics for various HIN tasks. Traditionally, meta-paths are primarily used for schema-simple HINs, e.g., bibliographic networks with only a few entity types, where meta-paths are often enumerated with domain knowledge. However, the adoption of meta-paths for schema-complex HINs, such as knowledge bases (KBs) with hundreds of entity and relation types, has been limited due to the computational complexity associated with meta-path enumeration. Additionally, effectively assessing meta-paths requires enumerating relevant path instances, which adds further complexity to the meta-path learning process. To address these challenges, we propose SchemaWalk, an inductive meta-path learning framework for schema-complex HINs. We represent meta-paths with schema-level representations to support the learning of the scores of meta-paths for varying relations, mitigating the need of exhaustive path instance enumeration for each relation. Further, we design a reinforcement-learning based path-finding agent, which directly navigates the network schema (i.e., schema graph) to learn policies for establishing meta-paths with high coverage and confidence for multiple relations. Extensive experiments on real data sets demonstrate the effectiveness of our proposed paradigm."
                    },
                    {
                        "title": "The Expressive Power of Graph Neural Networks: A Survey",
                        "abstract": "Graph neural networks (GNNs) are effective machine learning models for many graph-related applications. Despite their empirical success, many research efforts focus on the theoretical limitations of GNNs, i.e., the GNNs expressive power. Early works in this domain mainly focus on studying the graph isomorphism recognition ability of GNNs, and recent works try to leverage the properties such as subgraph counting and connectivity learning to characterize the expressive power of GNNs, which are more practical and closer to real-world. However, no survey papers and open-source repositories comprehensively summarize and discuss models in this important direction. To fill the gap, we conduct a first survey for models for enhancing expressive power under different forms of definition. Concretely, the models are reviewed based on three categories, i.e., Graph feature enhancement, Graph topology enhancement, and GNNs architecture enhancement."
                    }
                ]
            },
            "91368d4f-40dd-4fcf-87c4-feffd4620511": {
                "pk": "91368d4f-40dd-4fcf-87c4-feffd4620511",
                "name": "Yijun Pan",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "94a9070e-01a7-4a3d-9ab4-b29882f89ce0": {
                "pk": "94a9070e-01a7-4a3d-9ab4-b29882f89ce0",
                "name": "Hao Huang",
                "collaborators": [
                    "Benny Sudakov",
                    "Choongbum Lee",
                    "Alexander Yu",
                    "Yu Hao",
                    "Shuaihang Yuan",
                    "Yi Fang",
                    "Yi Zhao",
                    "Jie Ma",
                    "Yi Zhang",
                    "Fei Huang"
                ],
                "domain": [
                    "Combinatorics",
                    "Graph Theory",
                    "Machine Learning",
                    "Meta-Learning"
                ],
                "publications": [
                    {
                        "title": "Induced subgraphs of hypercubes and a proof of the Sensitivity Conjecture",
                        "abstract": "In this paper, we show that every $(2^{n-1}+1)$-vertex induced subgraph of the $n$-dimensional cube graph has maximum degree at least $\\sqrt{n}$. This result is best possible, and improves a logarithmic lower bound shown by Chung, F\\\"uredi, Graham and Seymour in 1988. As a direct consequence, we prove that the sensitivity and degree of a boolean function are polynomially related, solving an outstanding foundational problem in theoretical computer science, the Sensitivity Conjecture of Nisan and Szegedy."
                    },
                    {
                        "title": "On the maximum induced density of directed stars and related problems",
                        "abstract": "Let k>=3 be an integer, we prove that the maximum induced density of the k-vertex directed star in a directed graph is attained by an iterated blow-up construction. This confirms a conjecture by Falgas-Ravry and Vaughan, who proved this for k=3, 4. This question provides the first known instance of density problem for which one can prove extremality of an iterated blow-up construction. We also study the inducibility of complete bipartite digraphs and discuss other related problems."
                    },
                    {
                        "title": "Two extremal problems on intersecting families",
                        "abstract": "In this short note, we address two problems in extremal set theory regarding intersecting families. The first problem is a question posed by Kupavskii: is it true that given two disjoint cross-intersecting families $\\mathcal{A}, \\mathcal{B} \\subset \\binom{[n]}{k}$, they must satisfy $\\min\\{|\\mathcal{A}|, |\\mathcal{B}|\\} \\le \\frac{1}{2} \\binom{n-1}{k-1}$? We give an affirmative answer for $n \\ge 2k^2$, and construct families showing that this range is essentially the best one could hope for, up to a constant factor. The second problem is a conjecture of Frankl. It states that for $n \\ge 3k$, the maximum diversity of an intersecting family $\\mathcal{F} \\subset \\binom{[n]}{k}$ is equal to $\\binom{n-3}{k-2}$. We are able to find a construction beating the conjectured bound for $n$ slightly larger than $3k$, which also disproves a conjecture of Kupavskii."
                    },
                    {
                        "title": "Deterministic bootstrap percolation in high dimensional grids",
                        "abstract": "In this paper, we study the k-neighbor bootstrap percolation process on the d-dimensional grid [n]^d, and show that the minimum number of initial vertices that percolate is (1-d/k)n^d + O(n^{d-1})$ when d<=k<=2d. This confirms a conjecture of Pete."
                    },
                    {
                        "title": "A note on the double-critical graph conjecture",
                        "abstract": "A connected $n$-chromatic graph $G$ is double-critical if for all the edges $xy$ of $G$, the graph $G-x-y$ is $(n-2)$-chromatic. In 1966, Erd\\H os and Lov\\'asz conjectured that the only double-critical $n$-chromatic graph is $K_n$. This conjecture remains unresolved for $n \\ge 6.$ In this short note, we verify this conjecture for claw-free graphs $G$ of chromatic number $6$."
                    },
                    {
                        "title": "Meta-Learning 3D Shape Segmentation Functions",
                        "abstract": "Learning robust 3D shape segmentation functions with deep neural networks has emerged as a powerful paradigm, offering promising performance in producing a consistent part segmentation of each 3D shape. Generalizing across 3D shape segmentation functions requires robust learning of priors over the respective function space and enables consistent part segmentation of shapes in presence of significant 3D structure variations. Existing generalization methods rely on extensive training of 3D shape segmentation functions on large-scale labeled datasets. In this paper, we proposed to formalize the learning of a 3D shape segmentation function space as a meta-learning problem, aiming to predict a 3D segmentation model that can be quickly adapted to new shapes with no or limited training data. More specifically, we define each task as unsupervised learning of shape-conditioned 3D segmentation function which takes as input points in 3D space and predicts the part-segment labels. The 3D segmentation function is trained by a self-supervised 3D shape reconstruction loss without the need for part labels. Also, we introduce an auxiliary deep neural network as a meta-learner which takes as input a 3D shape and predicts the prior over the respective 3D segmentation function space. We show in experiments that our meta-learning approach, denoted as Meta-3DSeg, leads to improvements on unsupervised 3D shape segmentation over the conventional designs of deep neural networks for 3D shape segmentation functions."
                    },
                    {
                        "title": "Degree versions of the Erdos-Ko-Rado Theorem and Erdos hypergraph matching conjecture",
                        "abstract": "We use an algebraic method to prove a degree version of the celebrated Erd\\H os-Ko-Rado theorem: given $n>2k$, every intersecting $k$-uniform hypergraph $H$ on $n$ vertices contains a vertex that lies on at most $\\binom{n-2}{k-2}$ edges. This result can be viewed as a special case of the degree version of a well-known conjecture of Erd\\H{o}s on hypergraph matchings. Improving the work of Bollob\\'as, Daykin, and Erd\\H os from 1976, we show that given integers $n, k, s$ with $n\\ge 3k^2 s$, every $k$-uniform hypergraph $H$ on $n$ vertices with minimum vertex degree greater than $\\binom{n-1}{k-1}-\\binom{n-s}{k-1}$ contains $s$ disjoint edges."
                    },
                    {
                        "title": "On tight cycles in hypergraphs",
                        "abstract": "A tight $k$-uniform $\\ell$-cycle, denoted by $TC_\\ell^k$, is a $k$-uniform hypergraph whose vertex set is $v_0, \\cdots, v_{\\ell-1}$, and the edges are all the $k$-tuples $\\{v_i, v_{i+1}, \\cdots, v_{i+k-1}\\}$, with subscripts modulo $\\ell$. Motivated by a classic result in graph theory that every $n$-vertex cycle-free graph has at most $n-1$ edges, S\\'os and, independently, Verstra\\\"ete asked whether for every integer $k$, a $k$-uniform $n$-vertex hypergraph without any tight $k$-uniform cycles has at most $\\binom{n-1}{k-1}$ edges. In this paper, we answer this question in negative."
                    },
                    {
                        "title": "On a $d$-degree Erd\u0151s-Ko-Rado Theorem",
                        "abstract": "A family of subsets $\\mathcal{F}$ is intersecting if $A \\cap B \\neq \\emptyset$ for any $A, B \\in \\mathcal{F}$. In this paper, we show that for given integers $k > d \\ge 2$ and $n \\ge 2k+2d-3$, and any intersecting family $\\mathcal{F}$ of $k$-subsets of $\\{1, \\cdots, n\\}$, there exists a $d$-subset of $[n]$ contained in at most $\\binom{n-d-1}{k-d-1}$ subsets of $\\mathcal{F}$. This result, proved using spectral graph theory, gives a $d$-degree generalization of the celebrated Erd\\H{o}s-Ko-Rado Theorem, improving a theorem of Kupavskii."
                    },
                    {
                        "title": "Event Ticket Price Prediction with Deep Neural Network on Spatial-Temporal Sparse Data",
                        "abstract": "Event ticket price prediction is important to marketing strategy for any sports team or musical ensemble. An accurate prediction model can help the marketing team to make promotion plan more effectively and efficiently. However, given all the historical transaction records, it is challenging to predict the sale price of the remaining seats at any future timestamp, not only because that the sale price is relevant to a lot of features (seat locations, date-to-event of the transaction, event date, team performance, etc.), but also because of the temporal and spatial sparsity in the dataset. For a game/concert, the ticket selling price of one seat is only observable once at the time of sale. Furthermore, some seats may not even be purchased (therefore no record available). In fact, data sparsity is commonly encountered in many prediction problems. Here, we propose a bi-level optimizing deep neural network to address the curse of spatio-temporal sparsity. Specifically, we introduce coarsening and refining layers, and design a bi-level loss function to integrate different level of loss for better prediction accuracy. Our model can discover the interrelations among ticket sale price, seat locations, selling time, event information, etc. Experiments show that our proposed model outperforms other benchmark methods in real-world ticket selling price prediction."
                    },
                    {
                        "title": "A counterexample to the Alon-Saks-Seymour conjecture and related problems",
                        "abstract": "Consider a graph obtained by taking edge disjoint union of $k$ complete bipartite graphs. Alon, Saks and Seymour conjectured that such graph has chromatic number at most $k+1$. This well known conjecture remained open for almost twenty years. In this paper, we construct a counterexample to this conjecture and discuss several related problems in combinatorial geometry and communication complexity."
                    },
                    {
                        "title": "The minimum number of nonnegative edges in hypergraphs",
                        "abstract": "An r-unform n-vertex hypergraph H is said to have the Manickam-Mikl\\'os-Singhi (MMS) property if for every assignment of weights to its vertices with nonnegative sum, the number of edges whose total weight is nonnegative is at least the minimum degree of H. In this paper we show that for n>10r^3, every r-uniform n-vertex hypergraph with equal codegrees has the MMS property, and the bound on n is essentially tight up to a constant factor. This result has two immediate corollaries. First it shows that every set of n>10k^3 real numbers with nonnegative sum has at least $\\binom{n-1}{k-1}$ nonnegative k-sums, verifying the Manickam-Mikl\\'os-Singhi conjecture for this range. More importantly, it implies the vector space Manickam-Mikl\\'os-Singhi conjecture which states that for n >= 4k and any weighting on the 1-dimensional subspaces of F_q^n with nonnegative sum, the number of nonnegative k-dimensional subspaces is at least ${n-1 \\brack k-1}_q$. We also discuss two additional generalizations, which can be regarded as analogues of the Erd\\H{o}s-Ko-Rado theorem on k-intersecting families."
                    },
                    {
                        "title": "On almost k-covers of hypercubes",
                        "abstract": "In this paper, we consider the following problem: what is the minimum number of affine hyperplanes in $\\mathbb{R}^n$, such that all the vertices of $\\{0, 1\\}^n \\setminus \\{\\vec{0}\\}$ are covered at least $k$ times, and $\\vec{0}$ is uncovered? The $k=1$ case is the well-known Alon-F\\\"uredi theorem which says a minimum of $n$ affine hyperplanes is required, proved by the Combinatorial Nullstellensatz.   We develop an analogue of the Lubell-Yamamoto-Meshalkin inequality for subset sums, and completely solve the fractional version of this problem, which also provides an asymptotic answer to the integral version for fixed $n$ and $k \\rightarrow \\infty$. We also use a Punctured Combinatorial Nullstellensatz developed by Ball and Serra, to show that a minimum of $n+3$ affine hyperplanes is needed for $k=3$, and pose a conjecture for arbitrary $k$ and large $n$."
                    },
                    {
                        "title": "Expanded Gating Ranges Improve Activation Functions",
                        "abstract": "Activation functions are core components of all deep learning architectures. Currently, the most popular activation functions are smooth ReLU variants like GELU and SiLU. These are self-gated activation functions where the range of the gating function is between zero and one. In this paper, we explore the viability of using arctan as a gating mechanism. A self-gated activation function that uses arctan as its gating function has a monotonically increasing first derivative. To make this activation function competitive, it is necessary to introduce a trainable parameter for every MLP block to expand the range of the gating function beyond zero and one. We find that this technique also improves existing self-gated activation functions. We conduct an empirical evaluation of Expanded ArcTan Linear Unit (xATLU), Expanded GELU (xGELU), and Expanded SiLU (xSiLU) and show that they outperform existing activation functions within a transformer architecture. Additionally, expanded gating ranges show promising results in improving first-order Gated Linear Units (GLU)."
                    }
                ]
            },
            "2b3c57e2-bb27-4661-a4b5-ad8f6e489096": {
                "pk": "2b3c57e2-bb27-4661-a4b5-ad8f6e489096",
                "name": "Xinhe Wang",
                "collaborators": [
                    "Tingyu Wang",
                    "Hanzhong Liu",
                    "Ben B. Hansen",
                    "Mingqing Liu",
                    "Qingwen Liu",
                    "Mingliang Xiong",
                    "Hao Deng",
                    "Pingbang Hu",
                    "Junwei Deng",
                    "Jiaqi W. Ma"
                ],
                "domain": [
                    "Statistical Methods",
                    "Randomized Controlled Trials",
                    "Wireless Power Transfer",
                    "Adversarial Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Rerandomization in stratified randomized experiments",
                        "abstract": "Stratification and rerandomization are two well-known methods used in randomized experiments for balancing the baseline covariates. Renowned scholars in experimental design have recommended combining these two methods; however, limited studies have addressed the statistical properties of this combination. This paper proposes two rerandomization methods to be used in stratified randomized experiments, based on the overall and stratum-specific Mahalanobis distances. The first method is applicable for nearly arbitrary numbers of strata, strata sizes, and stratum-specific proportions of the treated units. The second method, which is generally more efficient than the first method, is suitable for situations in which the number of strata is fixed with their sizes tending to infinity. Under the randomization inference framework, we obtain the asymptotic distributions of estimators used in these methods and the formulas of variance reduction when compared to stratified randomization. Our analysis does not require any modeling assumption regarding the potential outcomes. Moreover, we provide asymptotically conservative variance estimators and confidence intervals for the average treatment effect. The advantages of the proposed methods are exhibited through an extensive simulation study and a real-data example."
                    },
                    {
                        "title": "Design-based variance estimation of the H\u00e1jek effect estimator in stratified and clustered experiments",
                        "abstract": "Randomized controlled trials (RCTs) are used to evaluate treatment effects. When individuals are grouped together, clustered RCTs are conducted. Stratification is recommended to reduce imbalance of baseline covariates between treatment and control. In practice, this can lead to comparisons between clusters of very different sizes. As a result, direct adjustment estimators that average differences of means within the strata may be inconsistent. We study differences of inverse probability weighted means of a treatment and a control group -- H\\'ajek effect estimators -- under two common forms of stratification: small strata that increase in number; or larger strata with growing numbers of clusters in each. Under either scenario, mild conditions give consistency and asymptotic Normality. We propose a variance estimator applicable to designs with any number of strata and strata of any size. We describe a special use of the variance estimator that improves small sample performance of Wald-type confidence intervals. The H\\'ajek effect estimator lends itself to covariance adjustment, and our variance estimator remains applicable. Simulations and real-world applications in children's nutrition and education confirm favorable operating characteristics, demonstrating advantages of the H\\'ajek effect estimator beyond its simplicity and ease of use."
                    },
                    {
                        "title": "Mobile Power Network for Ultimate Mobility without Battery Life Anxiety",
                        "abstract": "Similar to the evolution from the wired Internet to mobile Internet (MI), the growing demand for power delivery anywhere and anytime appeals for power grid transformation from wired to mobile domain. We propose here the next generation of power delivery network -- mobile power network (MPN) for wireless power transfer within a mobile range from several meters to tens of meters. At first, we present the MPN's concept evolution and application scenarios. Then, we introduce the MPN's supporting technology, namely resonant beam charging (RBC). As a long-range wireless power transfer (WPT) method, RBC can safely deliver multi-Watt power to multiple devices concurrently. Meanwhile, the recent progress in RBC research has been summarized. Next, we specify the MPN's architecture to provide the wide-area WPT coverage. Finally, we discuss the MPN's features and challenges. MPN can enable the ultimate mobility by cutting the final cord of mobile devices, realizing the \"last-mile\" mobile power delivery."
                    },
                    {
                        "title": "Adversarial Attacks on Data Attribution",
                        "abstract": "Data attribution aims to quantify the contribution of individual training data points to the outputs of an AI model, which has been used to measure the value of training data and compensate data providers. Given the impact on financial decisions and compensation mechanisms, a critical question arises concerning the adversarial robustness of data attribution methods. However, there has been little to no systematic research addressing this issue. In this work, we aim to bridge this gap by detailing a threat model with clear assumptions about the adversary's goal and capabilities and proposing principled adversarial attack methods on data attribution. We present two methods, Shadow Attack and Outlier Attack, which generate manipulated datasets to inflate the compensation adversarially. The Shadow Attack leverages knowledge about the data distribution in the AI applications, and derives adversarial perturbations through \"shadow training\", a technique commonly used in membership inference attacks. In contrast, the Outlier Attack does not assume any knowledge about the data distribution and relies solely on black-box queries to the target model's predictions. It exploits an inductive bias present in many data attribution methods - outlier data points are more likely to be influential - and employs adversarial examples to generate manipulated datasets. Empirically, in image classification and text generation tasks, the Shadow Attack can inflate the data-attribution-based compensation by at least 200%, while the Outlier Attack achieves compensation inflation ranging from 185% to as much as 643%."
                    },
                    {
                        "title": "Spin Manipulation by Giant Valley-Zeeman Spin-Orbit Field in Atom-Thick WSe2",
                        "abstract": "The phenomenon originating from spin-orbit coupling (SOC) provides energy-efficient strategies for spin manipulation and device applications. The broken inversion symmetry interface and resulting electric field induce a Rashba-type spin-orbit field (SOF), which has been demonstrated to generate spin-orbit torque for data storage applications. In this study, we found that spin flipping can be achieved by the valley-Zeeman SOF in monolayer WSe2 at room temperature, which manifests as a negative magnetoresistance in the vertical spin valve. Quantum transmission calculations based on an effective model near the K valley of WSe2 confirm the precessional spin transport of carriers under the giant SOF, which is estimated to be 650 T. In particular, the valley-Zeeman SOF-induced spin dynamics was demonstrated to be tunable with the layer number and stacking phase of WSe2 as well as the gate voltage, which provides a novel strategy for spin manipulation and can benefit the development of ultralow-power spintronic devices."
                    }
                ]
            },
            "b3272775-06a1-4638-8baf-1bda1175fa8b": {
                "pk": "b3272775-06a1-4638-8baf-1bda1175fa8b",
                "name": "Pingbang Hu",
                "collaborators": [
                    "Jiaqi W. Ma",
                    "Mahito Sugiyama",
                    "Yiwen Tu",
                    "Jiaqi Ma",
                    "Yuzheng Hu",
                    "Han Zhao",
                    "Xinhe Wang",
                    "Junwei Deng"
                ],
                "domain": [
                    "Combinatorial Optimization",
                    "Machine Learning",
                    "Data Attribution",
                    "Adversarial Robustness"
                ],
                "publications": [
                    {
                        "title": "Travel the Same Path: A Novel TSP Solving Strategy",
                        "abstract": "In this paper, we provide a novel strategy for solving Traveling Salesman Problem, which is a famous combinatorial optimization problem studied intensely in the TCS community. In particular, we consider the imitation learning framework, which helps a deterministic algorithm making good choices whenever it needs to, resulting in a speed up while maintaining the exactness of the solution without suffering from the unpredictability and a potential large deviation.   Furthermore, we demonstrate a strong generalization ability of a graph neural network trained under the imitation learning framework. Specifically, the model is capable of solving a large instance of TSP faster than the baseline while has only seen small TSP instances when training."
                    },
                    {
                        "title": "Pseudo-Non-Linear Data Augmentation via Energy Minimization",
                        "abstract": "We propose a novel and interpretable data augmentation method based on energy-based modeling and principles from information geometry. Unlike black-box generative models, which rely on deep neural networks, our approach replaces these non-interpretable transformations with explicit, theoretically grounded ones, ensuring interpretability and strong guarantees such as energy minimization. Central to our method is the introduction of the backward projection algorithm, which reverses dimension reduction to generate new data. Empirical results demonstrate that our method achieves competitive performance with black-box generative models while offering greater transparency and interpretability."
                    },
                    {
                        "title": "Towards Reliable Empirical Machine Unlearning Evaluation: A Game-Theoretic View",
                        "abstract": "Machine unlearning is the process of updating machine learning models to remove the information of specific training data samples, in order to comply with data protection regulations that allow individuals to request the removal of their personal data. Despite the recent development of numerous unlearning algorithms, reliable evaluation of these algorithms remains an open research question. In this work, we focus on membership inference attack (MIA) based evaluation, one of the most common approaches for evaluating unlearning algorithms, and address various pitfalls of existing evaluation metrics that lack reliability. Specifically, we propose a game-theoretic framework that formalizes the evaluation process as a game between unlearning algorithms and MIA adversaries, measuring the data removal efficacy of unlearning algorithms by the capability of the MIA adversaries. Through careful design of the game, we demonstrate that the natural evaluation metric induced from the game enjoys provable guarantees that the existing evaluation metrics fail to satisfy. Furthermore, we propose a practical and efficient algorithm to estimate the evaluation metric induced from the game, and demonstrate its effectiveness through both theoretical analysis and empirical experiments. This work presents a novel and reliable approach to empirically evaluating unlearning algorithms, paving the way for the development of more effective unlearning techniques."
                    },
                    {
                        "title": "Most Influential Subset Selection: Challenges, Promises, and Beyond",
                        "abstract": "How can we attribute the behaviors of machine learning models to their training data? While the classic influence function sheds light on the impact of individual samples, it often fails to capture the more complex and pronounced collective influence of a set of samples. To tackle this challenge, we study the Most Influential Subset Selection (MISS) problem, which aims to identify a subset of training samples with the greatest collective influence. We conduct a comprehensive analysis of the prevailing approaches in MISS, elucidating their strengths and weaknesses. Our findings reveal that influence-based greedy heuristics, a dominant class of algorithms in MISS, can provably fail even in linear regression. We delineate the failure modes, including the errors of influence function and the non-additive structure of the collective influence. Conversely, we demonstrate that an adaptive version of these heuristics which applies them iteratively, can effectively capture the interactions among samples and thus partially address the issues. Experiments on real-world datasets corroborate these theoretical findings, and further demonstrate that the merit of adaptivity can extend to more complex scenarios such as classification tasks and non-linear neural networks. We conclude our analysis by emphasizing the inherent trade-off between performance and computational efficiency, questioning the use of additive metrics such as the linear datamodeling score, and offering a range of discussions."
                    },
                    {
                        "title": "Adversarial Attacks on Data Attribution",
                        "abstract": "Data attribution aims to quantify the contribution of individual training data points to the outputs of an AI model, which has been used to measure the value of training data and compensate data providers. Given the impact on financial decisions and compensation mechanisms, a critical question arises concerning the adversarial robustness of data attribution methods. However, there has been little to no systematic research addressing this issue. In this work, we aim to bridge this gap by detailing a threat model with clear assumptions about the adversary's goal and capabilities and proposing principled adversarial attack methods on data attribution. We present two methods, Shadow Attack and Outlier Attack, which generate manipulated datasets to inflate the compensation adversarially. The Shadow Attack leverages knowledge about the data distribution in the AI applications, and derives adversarial perturbations through \"shadow training\", a technique commonly used in membership inference attacks. In contrast, the Outlier Attack does not assume any knowledge about the data distribution and relies solely on black-box queries to the target model's predictions. It exploits an inductive bias present in many data attribution methods - outlier data points are more likely to be influential - and employs adversarial examples to generate manipulated datasets. Empirically, in image classification and text generation tasks, the Shadow Attack can inflate the data-attribution-based compensation by at least 200%, while the Outlier Attack achieves compensation inflation ranging from 185% to as much as 643%."
                    }
                ]
            },
            "bef460e3-5662-4b43-96a4-c99d6c2acb76": {
                "pk": "bef460e3-5662-4b43-96a4-c99d6c2acb76",
                "name": "Xingjian Zhang",
                "collaborators": [
                    "Xiongfeng Ma",
                    "Qi Zhao",
                    "Guoding Liu",
                    "Jiaqi Ma",
                    "Qiaozhu Mei",
                    "Gleb Pogudin",
                    "Owen Xingjian Zhang",
                    "Yizhi Huang",
                    "Yuwei Zhu",
                    "Senrui Chen"
                ],
                "domain": [
                    "Quantum Information",
                    "Quantum Cryptography",
                    "Machine Learning",
                    "Quantum Error Correction"
                ],
                "publications": [
                    {
                        "title": "Simultaneous Certification of Entangled States and Measurements in Bounded Dimensional Semi-Quantum Games",
                        "abstract": "Certification of quantum systems and operations is a central task in quantum information processing. Most current schemes rely on a tomography with fully characterised devices, while this may not be met in real experiments. Device characterisations can be removed with device-independent tests, it is technically challenging at the moment, though. In this letter, we investigate the problem of certifying entangled states and measurements via semi-quantum games, a type of non-local quantum games with well characterised quantum inputs, balancing practicality and device-independence. We first design a specific bounded-dimensional measurement-device-independent game, with which we simultaneously certify any pure entangled state and Bell state measurement operators. Afterwards via a duality treatment of state and measurement, we interpret the dual form of this game as a source-independent bounded-dimensional entanglement swapping protocol and show the whole process, including any entangled projector and Bell states, can be certified with this protocol. In particular, our results do not require a complete Bell state measurement, which is beneficial for experiments and practical use."
                    },
                    {
                        "title": "Interpretable exact linear reductions via positivity",
                        "abstract": "Kinetic models of biochemical systems used in the modern literature often contain hundreds or even thousands of variables. While these models are convenient for detailed simulations, their size is often an obstacle to deriving mechanistic insights. One way to address this issue is to perform an exact model reduction by finding a self-consistent lower-dimensional projection of the corresponding dynamical system. Recently, a new algorithm CLUE has been designed and implemented, which allows one to construct an exact linear reduction of the smallest possible dimension such that the fixed variables of interest are preserved. It turned out that allowing arbitrary linear combinations (as opposed to zero-one combinations used in the prior approaches) may yield a much smaller reduction. However, there was a drawback: some of the new variables did not have clear physical meaning, thus making the reduced model harder to interpret. We design and implement an algorithm that, given an exact linear reduction, re-parametrizes it by performing an invertible transformation of the new coordinates to improve the interpretability of the new variables. We apply our algorithm to three case studies and show that \"uninterpretable\" variables disappear entirely in all the case studies. The implementation of the algorithm and the files for the case studies are available at https://github.com/xjzhaang/LumpingPostiviser."
                    },
                    {
                        "title": "Attitudes and perceived effectiveness among first-time online instructors during Covid-19",
                        "abstract": "Online teaching has expanded access to education, offering flexibility compared to traditional face-to-face instruction. While early research has explored online teaching, it is important to understand the perspective of instructors who conducted their first online classes during the Covid-19 pandemic. This study focuses on instructors teaching online for the first time, regardless of whether they volunteered. Surveys were conducted when universities transitioned from in-person to online instruction in April 2020, with a follow-up survey after their first online teaching semester. The study investigated instructors' expectations of class success before their first online teaching experience. Using Bayesian modeling, we analyzed how these expectations varied based on instructors' characteristics (self-efficacy in online teaching, technological proficiency, and acceptance of technology) and course attributes (subject area, class size, and instructional design). Results showed that instructors' self-efficacy significantly impacted their expectations of success, while smaller class sizes were associated with lower expectations. Interestingly, factors like prior use of technology platforms and classroom design did not contribute significantly to expectations. The study offers practical recommendations to support online teaching. To improve self-efficacy, instructors should collaborate with colleagues and familiarize themselves with online platforms. Universities should provide workshops or training to enhance teaching skills. In small interactive classes, nonverbal communication should be emphasized, and institutions should establish support teams and feedback mechanisms to ensure quality and effectiveness in online education."
                    },
                    {
                        "title": "Classically Replaceable Operations",
                        "abstract": "Quantum information science provides powerful technologies beyond the scope of classical physics. In practice, accurate control of quantum operations is a challenging task with current quantum devices. The implementation of high fidelity and multi-qubit quantum operations consumes massive resources and requires complicated hardware design to fight against noise. An approach to alleviating this problem is to replace quantum operations with classical processing. Despite the common practice of this approach, rigorous criteria to determine whether a given quantum operation is replaceable classically are still missing. In this work, we define the classically replaceable operations in four general scenarios. In each scenario, we provide their necessary and sufficient criteria and point out the corresponding classical processing. For a practically favorable case of unitary classically replaceable operations, we show that the replaced classical processing is deterministic. Beyond that, we regard the irreplaceability of quantum operations by classical processing as a quantum resource and relate it to the performance of a channel in a non-local game, as manifested in a robustness measure."
                    },
                    {
                        "title": "Stream privacy amplification for quantum cryptography",
                        "abstract": "Privacy amplification is the key step to guarantee the security of quantum communication. The existing security proofs require accumulating a large number of raw key bits for privacy amplification. This is similar to block ciphers in classical cryptography that would delay the final key generation since an entire block must be accumulated before privacy amplification. Moreover, any leftover errors after information reconciliation would corrupt the entire block. By modifying the security proof based on quantum error correction, we develop a stream privacy amplification scheme, which resembles the classical stream cipher. This scheme can output the final key in a stream way, prevent error from spreading, and hence can put privacy amplification before information reconciliation. The stream scheme can also help to enhance the security of trusted-relay quantum networks. Inspired by the connection between stream ciphers and quantum error correction in our security analysis, we further develop a generic information-theoretic tool to study the security of classical encryption algorithms."
                    },
                    {
                        "title": "Fast Learning of MNL Model from General Partial Rankings with Application to Network Formation Modeling",
                        "abstract": "Multinomial Logit (MNL) is one of the most popular discrete choice models and has been widely used to model ranking data. However, there is a long-standing technical challenge of learning MNL from many real-world ranking data: exact calculation of the MNL likelihood of \\emph{partial rankings} is generally intractable. In this work, we develop a scalable method for approximating the MNL likelihood of general partial rankings in polynomial time complexity. We also extend the proposed method to learn mixture of MNL. We demonstrate that the proposed methods are particularly helpful for applications to choice-based network formation modeling, where the formation of new edges in a network is viewed as individuals making choices of their friends over a candidate set. The problem of learning mixture of MNL models from partial rankings naturally arises in such applications. And the proposed methods can be used to learn MNL models from network data without the strong assumption that temporal orders of all the edge formation are available. We conduct experiments on both synthetic and real-world network data to demonstrate that the proposed methods achieve more accurate parameter estimation and better fitness of data compared to conventional methods."
                    },
                    {
                        "title": "Entanglement quantification via nonlocality",
                        "abstract": "Nonlocality, manifested by the violation of Bell inequalities, indicates quantum entanglement in the underlying system. A natural question that arises is how much entanglement is required for a given nonlocal behavior. In this paper, we explore this question by quantifying entanglement using a family of generalized Clauser-Horne-Shimony-Holt-type Bell inequalities. We focus on two entanglement measures, entanglement of formation and one-way distillable entanglement, which are related to entanglement dilution and distillation, respectively. We also study the interplay among nonlocality, entanglement, and measurement incompatibility. The result reveals that the relationship between entanglement and measurement incompatibility is not simply a trade-off under a fixed nonlocal behavior. In addition, we consider two realistic scenarios non-maximally entangled states and Werner states and apply our entanglement quantification results. By optimizing the Bell inequality for entanglement estimation, we derive analytical results for the entanglement of formation."
                    },
                    {
                        "title": "One-shot Coherence Distillation with Catalysts",
                        "abstract": "The resource theory of quantum coherence is an important topic in quantum information science. Standard coherence distillation and dilution problems have been thoroughly studied. In this paper, we introduce and study the problem of one-shot coherence distillation with catalysts. In order to distill more coherence from a state of interest, a catalytic system can be involved and a jointly free operation is applied to both systems. The joint output state should be a maximally coherent state in tensor product with the unchanged catalysts, with some allowable fidelity error. We consider several different definitions of this problem. Firstly, with a small fidelity error in both systems, we show that, even via the smallest free operation class (PIO), the distillable coherence of any state with no restriction on the catalysts is infinite, which is a \"coherence embezzling phenomenon\". We then define and calculate a lower bound for the distillable coherence when the dimension of catalysts is restricted. Finally, in consideration of physical relevance, we define the \"perfect catalysts\" scenario where the catalysts are required to be pure and precisely unchanged. Interestingly, we show that in this setting catalysts basically provide no advantages in pure state distillation via IO and SIO under certain smoothing restriction. Our work enhances the understanding of catalytic effect in quantum resource theory."
                    },
                    {
                        "title": "Quantifying Coherence with Untrusted Devices",
                        "abstract": "Device-independent (DI) tests allow to witness and quantify the quantum feature of a system, such as entanglement, without trusting the implementation devices. Although DI test is a powerful tool in many quantum information tasks, it generally requires nonlocal settings. Fundamentally, the superposition property of quantum states, quantified by coherence measures, is a distinct feature to distinguish quantum mechanics from classical theories. In literature, witness and quantification of coherence with trusted devices have been well-studied. However, it remains open whether we can witness and quantify single party coherence with untrusted devices, as it is not clear whether the concept of DI tests exists without a nonlocal setting. In this work, we study DI witness and quantification of coherence with untrusted devices. First, we prove a no-go theorem for a fully DI scenario, as well as a semi DI scenario employing a joint measurement with trusted ancillary states. We then propose a general prepare-and-measure semi DI scheme for witnessing and quantifying the amount of coherence. We show how to quantify the relative entropy and the $l_1$ norm of single party coherence with analytical and numerical methods. As coherence is a fundamental resource for tasks such as quantum random number generation and quantum key distribution, we expect our result may shed light on designing new semi DI quantum cryptographic schemes."
                    },
                    {
                        "title": "Can LLMs Effectively Leverage Graph Structural Information through Prompts, and Why?",
                        "abstract": "Large language models (LLMs) are gaining increasing attention for their capability to process graphs with rich text attributes, especially in a zero-shot fashion. Recent studies demonstrate that LLMs obtain decent text classification performance on common text-rich graph benchmarks, and the performance can be improved by appending encoded structural information as natural languages into prompts. We aim to understand why the incorporation of structural information inherent in graph data can improve the prediction performance of LLMs. First, we rule out the concern of data leakage by curating a novel leakage-free dataset and conducting a comparative analysis alongside a previously widely-used dataset. Second, as past work usually encodes the ego-graph by describing the graph structure in natural language, we ask the question: do LLMs understand the graph structure in accordance with the intent of the prompt designers? Third, we investigate why LLMs can improve their performance after incorporating structural information. Our exploration of these questions reveals that (i) there is no substantial evidence that the performance of LLMs is significantly attributed to data leakage; (ii) instead of understanding prompts as graph structures as intended by the prompt designers, LLMs tend to process prompts more as contextual paragraphs and (iii) the most efficient elements of the local neighborhood included in the prompt are phrases that are pertinent to the node label, rather than the graph structure."
                    },
                    {
                        "title": "Unconditional quantum MAGIC advantage in shallow circuit computation",
                        "abstract": "Quantum theory promises computation speed-ups than classical means. The full power is believed to reside in \"magic\" states, or equivalently non-Clifford operations -- the secret sauce to establish universal quantum computing. Despite the celebrated Gottesman-Knill Theorem stating that magic-free computation can be efficiently simulated by a classical computer, it is still questionable whether \"magic\" is really magical. Indeed, all the existing results establish its supremacy for efficient computation upon unproven complexity assumptions or queries to black-box oracles. In this work, we show that the magic advantage can be unconditionally established, at least in a shallow circuit with a constant depth. For this purpose, we first construct a specific nonlocal game inspired by the linear binary constraint system, which requires the magic resource to generate the desired nonlocal statistics or quantum \"pseudo telepathy.\" For a relation problem targeting generating such correlations between arbitrary nonlocal computation sites, we construct a shallow circuit with bounded fan-in gates that takes the strategy for quantum pseudo telepathy as a sub-routine to solve the problem with certainty. In contrast, magic-free counterparts inevitably require a logarithmic circuit depth to the input size, and the separation is proven optimal. As by-products, we prove that the nonlocal game we construct has non-unique perfect winning strategies, answering an open problem in quantum self-testing. We also provide an efficient algorithm to aid the search for potential magic-requiring nonlocal games similar to the current one. We anticipate our results to enlighten the ultimate establishment of the unconditional advantage of universal quantum computation."
                    },
                    {
                        "title": "BioImageLoader: Easy Handling of Bioimage Datasets for Machine Learning",
                        "abstract": "BioImageLoader (BIL) is a python library that handles bioimage datasets for machine learning applications, easing simple workflows and enabling complex ones. BIL attempts to wrap the numerous and varied bioimages datasets in unified interfaces, to easily concatenate, perform image augmentation, and batch-load them. By acting at a per experimental dataset level, it enables both a high level of customization and a comparison across experiments. Here we present the library and show some application it enables, including retraining published deep learning architectures and evaluating their versatility in a leave-one-dataset-out fashion."
                    },
                    {
                        "title": "Intrinsic randomness under general quantum measurements",
                        "abstract": "Quantum measurements can produce randomness arising from the uncertainty principle. When measuring a state with von Neumann measurements, the intrinsic randomness can be quantified by the quantum coherence of the state on the measurement basis. Unlike projection measurements, there are additional and possibly hidden degrees of freedom in apparatus for generic measurements. We propose an adversary scenario for general measurements with arbitrary input states, based on which, we characterize the intrinsic randomness. Interestingly, we discover that under certain measurements, such as the symmetric and information-complete measurement, all states have nonzero randomness, inspiring a new design of source-independent random number generators without state characterization. Furthermore, our results show that intrinsic randomness can quantify coherence under general measurements, which generalizes the result in the standard resource theory of state coherence."
                    },
                    {
                        "title": "Improved Real-time Post-Processing for Quantum Random Number Generators",
                        "abstract": "Randomness extraction is a key problem in cryptography and theoretical computer science. With the recent rapid development of quantum cryptography, quantum-proof randomness extraction has also been widely studied, addressing the security issues in the presence of a quantum adversary. In contrast with conventional quantum-proof randomness extractors characterizing the input raw data as min-entropy sources, we find that the input raw data generated by a large class of trusted-device quantum random number generators can be characterized as the so-called reverse block source. This fact enables us to design improved extractors. Specifically, we propose two novel quantum-proof randomness extractors for reverse block sources that realize real-time block-wise extraction. In comparison with the general min-entropy randomness extractors, our designs achieve a significantly higher extraction speed and a longer output data length with the same seed length. In addition, they enjoy the property of online algorithms, which process the raw data on the fly without waiting for the entire input raw data to be available. These features make our design an adequate choice for the real-time post-processing of practical quantum random number generators. Applying our extractors to the raw data generated by a widely used quantum random number generator, we achieve a simulated extraction speed as high as $300$ Gbps."
                    },
                    {
                        "title": "Virtual Channel Purification",
                        "abstract": "Quantum error mitigation is a key approach for extracting target state properties on state-of-the-art noisy machines and early fault-tolerant devices. Using the ideas from flag fault tolerance and virtual state purification, we develop the virtual channel purification (VCP) protocol, which consumes similar qubit and gate resources as virtual state purification but offers stronger error suppression with increased system size and more noisy operation copies. Furthermore, VCP removes most of the assumptions required in virtual state purification. Essentially, VCP is the first quantum error mitigation protocol that does not require specific knowledge about the noise models, the target quantum state, and the target problem while still offering rigorous performance guarantees for practical noise regimes. Further connections are made between VCP and quantum error correction to produce the virtual error correction (VEC) protocol, one of the first protocols that combine quantum error correction (QEC) and quantum error mitigation beyond concatenation. VEC can virtually remove all correctable noise in the channel while paying only the same sampling cost as low-order purification. It can achieve QEC-level protection on an unencoded register when transmitting it through a noisy channel, removing the associated encoding qubit overhead. Another variant of VEC can mimic the error suppression power of the surface code by inputting only a bit-flip and a phase-flip code. Our protocol can also be adapted to key tasks in quantum networks like channel capacity activation and entanglement distribution."
                    }
                ]
            },
            "ad026d74-cdc3-43fa-aa76-95dc6a67dfff": {
                "pk": "ad026d74-cdc3-43fa-aa76-95dc6a67dfff",
                "name": "Jiaqi W. Ma",
                "collaborators": [
                    "Pingbang Hu",
                    "Xinhe Wang",
                    "Junwei Deng",
                    "Yuzheng Hu",
                    "Han Zhao"
                ],
                "domain": [
                    "Data Attribution",
                    "Adversarial Machine Learning",
                    "Influence Functions",
                    "Subset Selection"
                ],
                "publications": [
                    {
                        "title": "Adversarial Attacks on Data Attribution",
                        "abstract": "Data attribution aims to quantify the contribution of individual training data points to the outputs of an AI model, which has been used to measure the value of training data and compensate data providers. Given the impact on financial decisions and compensation mechanisms, a critical question arises concerning the adversarial robustness of data attribution methods. However, there has been little to no systematic research addressing this issue. In this work, we aim to bridge this gap by detailing a threat model with clear assumptions about the adversary's goal and capabilities and proposing principled adversarial attack methods on data attribution. We present two methods, Shadow Attack and Outlier Attack, which generate manipulated datasets to inflate the compensation adversarially. The Shadow Attack leverages knowledge about the data distribution in the AI applications, and derives adversarial perturbations through \"shadow training\", a technique commonly used in membership inference attacks. In contrast, the Outlier Attack does not assume any knowledge about the data distribution and relies solely on black-box queries to the target model's predictions. It exploits an inductive bias present in many data attribution methods - outlier data points are more likely to be influential - and employs adversarial examples to generate manipulated datasets. Empirically, in image classification and text generation tasks, the Shadow Attack can inflate the data-attribution-based compensation by at least 200%, while the Outlier Attack achieves compensation inflation ranging from 185% to as much as 643%."
                    },
                    {
                        "title": "Most Influential Subset Selection: Challenges, Promises, and Beyond",
                        "abstract": "How can we attribute the behaviors of machine learning models to their training data? While the classic influence function sheds light on the impact of individual samples, it often fails to capture the more complex and pronounced collective influence of a set of samples. To tackle this challenge, we study the Most Influential Subset Selection (MISS) problem, which aims to identify a subset of training samples with the greatest collective influence. We conduct a comprehensive analysis of the prevailing approaches in MISS, elucidating their strengths and weaknesses. Our findings reveal that influence-based greedy heuristics, a dominant class of algorithms in MISS, can provably fail even in linear regression. We delineate the failure modes, including the errors of influence function and the non-additive structure of the collective influence. Conversely, we demonstrate that an adaptive version of these heuristics which applies them iteratively, can effectively capture the interactions among samples and thus partially address the issues. Experiments on real-world datasets corroborate these theoretical findings, and further demonstrate that the merit of adaptivity can extend to more complex scenarios such as classification tasks and non-linear neural networks. We conclude our analysis by emphasizing the inherent trade-off between performance and computational efficiency, questioning the use of additive metrics such as the linear datamodeling score, and offering a range of discussions."
                    }
                ]
            }
        }
    },
    "2405.16605": {
        "paper_data": {
            "title": "Demystify Mamba in Vision: A Linear Attention Perspective",
            "url": "http://arxiv.org/abs/2405.16605v1",
            "arxiv_id": "2405.16605",
            "authors": [
                "Dongchen Han",
                "Ziyi Wang",
                "Zhuofan Xia",
                "Yizeng Han",
                "Yifan Pu",
                "Chunjiang Ge",
                "Jun Song",
                "Shiji Song",
                "Bo Zheng",
                "Gao Huang"
            ],
            "abstract": "Mamba is an effective state space model with linear computation complexity. It has recently shown impressive efficiency in dealing with high-resolution inputs across various vision tasks. In this paper, we reveal that the powerful Mamba model shares surprising similarities with linear attention Transformer, which typically underperform conventional Transformer in practice. By exploring the similarities and disparities between the effective Mamba and subpar linear attention Transformer, we provide comprehensive analyses to demystify the key factors behind Mamba's success. Specifically, we reformulate the selective state space model and linear attention within a unified formulation, rephrasing Mamba as a variant of linear attention Transformer with six major distinctions: input gate, forget gate, shortcut, no attention normalization, single-head, and modified block design. For each design, we meticulously analyze its pros and cons, and empirically evaluate its impact on model performance in vision tasks. Interestingly, the results highlight the forget gate and block design as the core contributors to Mamba's success, while the other four designs are less crucial. Based on these findings, we propose a Mamba-Like Linear Attention (MLLA) model by incorporating the merits of these two key designs into linear attention. The resulting model outperforms various vision Mamba models in both image classification and high-resolution dense prediction tasks, while enjoying parallelizable computation and fast inference speed. Code is available at https://github.com/LeapLabTHU/MLLA.",
            "introduction": "   1 Introduction  Recently, state space models, exemplified by Mamba, have rapidly gained wide research interest. In contrast to the quadratic complexity of prevailing Transformer models, the state-space-based Mamba offers effective sequence modeling with linear complexity. This crucial property allows Mamba to handle extremely long sequences with manageable computational costs, making it a promising architecture for both natural language processing\u00a0[11, 23] and visual recognition\u00a0[48, 25].   However, Mamba is not the first model to achieve global modeling with linear complexity. Linear attention\u00a0[20], an early work, was proposed as an computationally efficient alternative to the widely adopted Softmax attention\u00a0[34], namely dot-product attention. Specifically, linear attention replaces the non-linear Softmax function in attention operation with linear normalization. This enables a change in computation order from (Q\u2062K\u22a4)\u2062V\ud835\udc44superscript\ud835\udc3etop\ud835\udc49(QK^{\\top})V( italic_Q italic_K start_POSTSUPERSCRIPT \u22a4 end_POSTSUPERSCRIPT ) italic_V to Q\u2062(K\u22a4\u2062V)\ud835\udc44superscript\ud835\udc3etop\ud835\udc49Q(K^{\\top}V)italic_Q ( italic_K start_POSTSUPERSCRIPT \u22a4 end_POSTSUPERSCRIPT italic_V ), thus reducing computation complexity from \ud835\udcaa\u2062(N2)\ud835\udcaasuperscript\ud835\udc412\\mathcal{O}(N^{2})caligraphic_O ( italic_N start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ) to \ud835\udcaa\u2062(N)\ud835\udcaa\ud835\udc41\\mathcal{O}(N)caligraphic_O ( italic_N ). Despite its efficiency, previous works\u00a0[3, 31, 12] proved that linear attention suffers from insufficient expressive power, making it impractical for real applications. Surprisingly, we find a very close relationship between the formulas of high-performance Mamba and subpar linear attention Transformer. Therefore, a compelling research question emerges: What factors contribute to Mamba\u2019s success and its significant superiority to linear attention Transformer?   In this paper, we offer both theoretical and empirical analyses to unveil Mamba through the lens of linear attention Transformer. Specifically, we rewrite the formulas of selective state space model and linear attention within a unified formulation, depicting Mamba as a variation of linear attention Transformer with six distinctions: input gate, forget gate, shortcut, no attention normalization, single-head, and modified block design. To demystify what factors lead to Mamba\u2019s effectiveness, empirical studies on vision tasks are conducted to assess the impact of each special design. The results demonstrate that the forget gate and block design tend to be the two core contributors to Mamba\u2019s superiority. While the block design can be easily adopted, the forget gate necessitates recurrent computation, which may not be well-suited for non-auto-regressive vision models. Therefore, we delve into the essence of the forget gate and verify that it can be replaced by suitable positional encoding in vision tasks. Based on our findings, we introduce the two core contributors or their alternatives to linear attention Transformer, presenting our Mamba-Like Linear Attention (MLLA) model. Experimental results demonstrate that MLLA achieves superior results to various vision Mamba models in both image classification and high-resolution dense prediction tasks, validating that linear attention can surpass Mamba with the merits of two core designs.   Our main contributions and takeaways are as follows:   \u2022  We reveal Mamba\u2019s close relationship to linear attention Transformer: Mamba and linear attention Transformer can be formulated within a unified framework, with Mamba exhibiting six distinct designs compared to the conventional linear attention paradigm: input gate, forget gate, shortcut, no attention normalization, single-head and modified block design.    \u2022  We provide detailed analyses of each special design and empirically validate that the forget gate and block design largely lead to Mamba\u2019s superiority. Additionally, we demonstrate that the recurrent calculation of the forget gate might",
            "references": [
                {
                    "title": "MambaOut: Do We Really Need Mamba for Vision?",
                    "abstract": "Mamba, an architecture with RNN-like token mixer of state space model (SSM), was recently introduced to address the quadratic complexity of the attention mechanism and subsequently applied to vision tasks. Nevertheless, the performance of Mamba for vision is often underwhelming when compared with convolutional and attention-based models. In this paper, we delve into the essence of Mamba, and conceptually conclude that Mamba is ideally suited for tasks with long-sequence and autoregressive characteristics. For vision tasks, as image classification does not align with either characteristic, we hypothesize that Mamba is not necessary for this task; Detection and segmentation tasks are also not autoregressive, yet they adhere to the long-sequence characteristic, so we believe it is still worthwhile to explore Mamba's potential for these tasks. To empirically verify our hypotheses, we construct a series of models named MambaOut through stacking Mamba blocks while removing their core token mixer, SSM. Experimental results strongly support our hypotheses. Specifically, our MambaOut model surpasses all visual Mamba models on ImageNet image classification, indicating that Mamba is indeed unnecessary for this task. As for detection and segmentation, MambaOut cannot match the performance of state-of-the-art visual Mamba models, demonstrating the potential of Mamba for long-sequence visual tasks. The code is available at https://github.com/yuweihao/MambaOut"
                },
                {
                    "title": "Jamba: A Hybrid Transformer-Mamba Language Model",
                    "abstract": "We present Jamba, a new base large language model based on a novel hybrid Transformer-Mamba mixture-of-experts (MoE) architecture. Specifically, Jamba interleaves blocks of Transformer and Mamba layers, enjoying the benefits of both model families. MoE is added in some of these layers to increase model capacity while keeping active parameter usage manageable. This flexible architecture allows resource- and objective-specific configurations. In the particular configuration we have implemented, we end up with a powerful model that fits in a single 80GB GPU. Built at large scale, Jamba provides high throughput and small memory footprint compared to vanilla Transformers, and at the same time state-of-the-art performance on standard language model benchmarks and long-context evaluations. Remarkably, the model presents strong results for up to 256K tokens context length. We study various architectural decisions, such as how to combine Transformer and Mamba layers, and how to mix experts, and show that some of them are crucial in large scale modeling. We also describe several interesting properties of these architectures which the training and evaluation of Jamba have revealed, and plan to release checkpoints from various ablation runs, to encourage further exploration of this novel architecture. We make the weights of our implementation of Jamba publicly available under a permissive license."
                },
                {
                    "title": "PlainMamba: Improving Non-Hierarchical Mamba in Visual Recognition",
                    "abstract": "We present PlainMamba: a simple non-hierarchical state space model (SSM) designed for general visual recognition. The recent Mamba model has shown how SSMs can be highly competitive with other architectures on sequential data and initial attempts have been made to apply it to images. In this paper, we further adapt the selective scanning process of Mamba to the visual domain, enhancing its ability to learn features from two-dimensional images by (i) a continuous 2D scanning process that improves spatial continuity by ensuring adjacency of tokens in the scanning sequence, and (ii) direction-aware updating which enables the model to discern the spatial relations of tokens by encoding directional information. Our architecture is designed to be easy to use and easy to scale, formed by stacking identical PlainMamba blocks, resulting in a model with constant width throughout all layers. The architecture is further simplified by removing the need for special tokens. We evaluate PlainMamba on a variety of visual recognition tasks, achieving performance gains over previous non-hierarchical models and is competitive with hierarchical alternatives. For tasks requiring high-resolution inputs, in particular, PlainMamba requires much less computing while maintaining high performance. Code and models are available at: https://github.com/ChenhongyiYang/PlainMamba ."
                },
                {
                    "title": "EfficientVMamba: Atrous Selective Scan for Light Weight Visual Mamba",
                    "abstract": "Prior efforts in light-weight model development mainly centered on CNN and Transformer-based designs yet faced persistent challenges. CNNs adept at local feature extraction compromise resolution while Transformers offer global reach but escalate computational demands $\\mathcal{O}(N^2)$. This ongoing trade-off between accuracy and efficiency remains a significant hurdle. Recently, state space models (SSMs), such as Mamba, have shown outstanding performance and competitiveness in various tasks such as language modeling and computer vision, while reducing the time complexity of global information extraction to $\\mathcal{O}(N)$. Inspired by this, this work proposes to explore the potential of visual state space models in light-weight model design and introduce a novel efficient model variant dubbed EfficientVMamba. Concretely, our EfficientVMamba integrates a atrous-based selective scan approach by efficient skip sampling, constituting building blocks designed to harness both global and local representational features. Additionally, we investigate the integration between SSM blocks and convolutions, and introduce an efficient visual state space block combined with an additional convolution branch, which further elevate the model performance. Experimental results show that, EfficientVMamba scales down the computational complexity while yields competitive results across a variety of vision tasks. For example, our EfficientVMamba-S with $1.3$G FLOPs improves Vim-Ti with $1.5$G FLOPs by a large margin of $5.6\\%$ accuracy on ImageNet. Code is available at: \\url{https://github.com/TerryPei/EfficientVMamba}."
                },
                {
                    "title": "LocalMamba: Visual State Space Model with Windowed Selective Scan",
                    "abstract": "Recent advancements in state space models, notably Mamba, have demonstrated significant progress in modeling long sequences for tasks like language understanding. Yet, their application in vision tasks has not markedly surpassed the performance of traditional Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs). This paper posits that the key to enhancing Vision Mamba (ViM) lies in optimizing scan directions for sequence modeling. Traditional ViM approaches, which flatten spatial tokens, overlook the preservation of local 2D dependencies, thereby elongating the distance between adjacent tokens. We introduce a novel local scanning strategy that divides images into distinct windows, effectively capturing local dependencies while maintaining a global perspective. Additionally, acknowledging the varying preferences for scan patterns across different network layers, we propose a dynamic method to independently search for the optimal scan choices for each layer, substantially improving performance. Extensive experiments across both plain and hierarchical models underscore our approach's superiority in effectively capturing image representations. For example, our model significantly outperforms Vim-Ti by 3.1% on ImageNet with the same 1.5G FLOPs. Code is available at: https://github.com/hunto/LocalMamba."
                },
                {
                    "title": "The Hidden Attention of Mamba Models",
                    "abstract": "The Mamba layer offers an efficient selective state space model (SSM) that is highly effective in modeling multiple domains, including NLP, long-range sequence processing, and computer vision. Selective SSMs are viewed as dual models, in which one trains in parallel on the entire sequence via an IO-aware parallel scan, and deploys in an autoregressive manner. We add a third view and show that such models can be viewed as attention-driven models. This new perspective enables us to empirically and theoretically compare the underlying mechanisms to that of the self-attention layers in transformers and allows us to peer inside the inner workings of the Mamba model with explainability methods. Our code is publicly available."
                },
                {
                    "title": "Mamba-ND: Selective State Space Modeling for Multi-Dimensional Data",
                    "abstract": "In recent years, Transformers have become the de-facto architecture for sequence modeling on text and a variety of multi-dimensional data, such as images and video. However, the use of self-attention layers in a Transformer incurs prohibitive compute and memory complexity that scales quadratically w.r.t. the sequence length. A recent architecture, Mamba, based on state space models has been shown to achieve comparable performance for modeling text sequences, while scaling linearly with the sequence length. In this work, we present Mamba-ND, a generalized design extending the Mamba architecture to arbitrary multi-dimensional data. Our design alternatively unravels the input data across different dimensions following row-major orderings. We provide a systematic comparison of Mamba-ND with several other alternatives, based on prior multi-dimensional extensions such as Bi-directional LSTMs and S4ND. Empirically, we show that Mamba-ND demonstrates performance competitive with the state-of-the-art on a variety of multi-dimensional benchmarks, including ImageNet-1K classification, HMDB-51 action recognition, and ERA5 weather forecasting."
                },
                {
                    "title": "VMamba: Visual State Space Model",
                    "abstract": "Designing computationally efficient network architectures persists as an ongoing necessity in computer vision. In this paper, we transplant Mamba, a state-space language model, into VMamba, a vision backbone that works in linear time complexity. At the core of VMamba lies a stack of Visual State-Space (VSS) blocks with the 2D Selective Scan (SS2D) module. By traversing along four scanning routes, SS2D helps bridge the gap between the ordered nature of 1D selective scan and the non-sequential structure of 2D vision data, which facilitates the gathering of contextual information from various sources and perspectives. Based on the VSS blocks, we develop a family of VMamba architectures and accelerate them through a succession of architectural and implementation enhancements. Extensive experiments showcase VMamba's promising performance across diverse visual perception tasks, highlighting its advantages in input scaling efficiency compared to existing benchmark models. Source code is available at https://github.com/MzeroMiko/VMamba."
                },
                {
                    "title": "Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model",
                    "abstract": "Recently the state space models (SSMs) with efficient hardware-aware designs, i.e., the Mamba deep learning model, have shown great potential for long sequence modeling. Meanwhile building efficient and generic vision backbones purely upon SSMs is an appealing direction. However, representing visual data is challenging for SSMs due to the position-sensitivity of visual data and the requirement of global context for visual understanding. In this paper, we show that the reliance on self-attention for visual representation learning is not necessary and propose a new generic vision backbone with bidirectional Mamba blocks (Vim), which marks the image sequences with position embeddings and compresses the visual representation with bidirectional state space models. On ImageNet classification, COCO object detection, and ADE20k semantic segmentation tasks, Vim achieves higher performance compared to well-established vision transformers like DeiT, while also demonstrating significantly improved computation&memory efficiency. For example, Vim is 2.8$\\times$ faster than DeiT and saves 86.8% GPU memory when performing batch inference to extract features on images with a resolution of 1248$\\times$1248. The results demonstrate that Vim is capable of overcoming the computation&memory constraints on performing Transformer-style understanding for high-resolution images and it has great potential to be the next-generation backbone for vision foundation models. Code is available at https://github.com/hustvl/Vim."
                },
                {
                    "title": "Agent Attention: On the Integration of Softmax and Linear Attention",
                    "abstract": "The attention module is the key component in Transformers. While the global attention mechanism offers high expressiveness, its excessive computational cost restricts its applicability in various scenarios. In this paper, we propose a novel attention paradigm, Agent Attention, to strike a favorable balance between computational efficiency and representation power. Specifically, the Agent Attention, denoted as a quadruple $(Q, A, K, V)$, introduces an additional set of agent tokens $A$ into the conventional attention module. The agent tokens first act as the agent for the query tokens $Q$ to aggregate information from $K$ and $V$, and then broadcast the information back to $Q$. Given the number of agent tokens can be designed to be much smaller than the number of query tokens, the agent attention is significantly more efficient than the widely adopted Softmax attention, while preserving global context modelling capability. Interestingly, we show that the proposed agent attention is equivalent to a generalized form of linear attention. Therefore, agent attention seamlessly integrates the powerful Softmax attention and the highly efficient linear attention. Extensive experiments demonstrate the effectiveness of agent attention with various vision Transformers and across diverse vision tasks, including image classification, object detection, semantic segmentation and image generation. Notably, agent attention has shown remarkable performance in high-resolution scenarios, owning to its linear attention nature. For instance, when applied to Stable Diffusion, our agent attention accelerates generation and substantially enhances image generation quality without any additional training. Code is available at https://github.com/LeapLabTHU/Agent-Attention."
                },
                {
                    "title": "Mamba: Linear-Time Sequence Modeling with Selective State Spaces",
                    "abstract": "Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module. Many subquadratic-time architectures such as linear attention, gated convolution and recurrent models, and structured state space models (SSMs) have been developed to address Transformers' computational inefficiency on long sequences, but they have not performed as well as attention on important modalities such as language. We identify that a key weakness of such models is their inability to perform content-based reasoning, and make several improvements. First, simply letting the SSM parameters be functions of the input addresses their weakness with discrete modalities, allowing the model to selectively propagate or forget information along the sequence length dimension depending on the current token. Second, even though this change prevents the use of efficient convolutions, we design a hardware-aware parallel algorithm in recurrent mode. We integrate these selective SSMs into a simplified end-to-end neural network architecture without attention or even MLP blocks (Mamba). Mamba enjoys fast inference (5$\\times$ higher throughput than Transformers) and linear scaling in sequence length, and its performance improves on real data up to million-length sequences. As a general sequence model backbone, Mamba achieves state-of-the-art performance across several modalities such as language, audio, and genomics. On language modeling, our Mamba-3B model outperforms Transformers of the same size and matches Transformers twice its size, both in pretraining and downstream evaluation."
                },
                {
                    "title": "FLatten Transformer: Vision Transformer using Focused Linear Attention",
                    "abstract": "The quadratic computation complexity of self-attention has been a persistent challenge when applying Transformer models to vision tasks. Linear attention, on the other hand, offers a much more efficient alternative with its linear complexity by approximating the Softmax operation through carefully designed mapping functions. However, current linear attention approaches either suffer from significant performance degradation or introduce additional computation overhead from the mapping functions. In this paper, we propose a novel Focused Linear Attention module to achieve both high efficiency and expressiveness. Specifically, we first analyze the factors contributing to the performance degradation of linear attention from two perspectives: the focus ability and feature diversity. To overcome these limitations, we introduce a simple yet effective mapping function and an efficient rank restoration module to enhance the expressiveness of self-attention while maintaining low computation complexity. Extensive experiments show that our linear attention module is applicable to a variety of advanced vision Transformers, and achieves consistently improved performances on multiple benchmarks. Code is available at https://github.com/LeapLabTHU/FLatten-Transformer."
                },
                {
                    "title": "BiFormer: Vision Transformer with Bi-Level Routing Attention",
                    "abstract": "As the core building block of vision transformers, attention is a powerful tool to capture long-range dependency. However, such power comes at a cost: it incurs a huge computation burden and heavy memory footprint as pairwise token interaction across all spatial locations is computed. A series of works attempt to alleviate this problem by introducing handcrafted and content-agnostic sparsity into attention, such as restricting the attention operation to be inside local windows, axial stripes, or dilated windows. In contrast to these approaches, we propose a novel dynamic sparse attention via bi-level routing to enable a more flexible allocation of computations with content awareness. Specifically, for a query, irrelevant key-value pairs are first filtered out at a coarse region level, and then fine-grained token-to-token attention is applied in the union of remaining candidate regions (i.e., routed regions). We provide a simple yet effective implementation of the proposed bilevel routing attention, which utilizes the sparsity to save both computation and memory while involving only GPU-friendly dense matrix multiplications. Built with the proposed bi-level routing attention, a new general vision transformer, named BiFormer, is then presented. As BiFormer attends to a small subset of relevant tokens in a query adaptive manner without distraction from other irrelevant ones, it enjoys both good performance and high computational efficiency, especially in dense prediction tasks. Empirical results across several computer vision tasks such as image classification, object detection, and semantic segmentation verify the effectiveness of our design. Code is available at https://github.com/rayleizhu/BiFormer."
                },
                {
                    "title": "Hungry Hungry Hippos: Towards Language Modeling with State Space Models",
                    "abstract": "State space models (SSMs) have demonstrated state-of-the-art sequence modeling performance in some modalities, but underperform attention in language modeling. Moreover, despite scaling nearly linearly in sequence length instead of quadratically, SSMs are still slower than Transformers due to poor hardware utilization. In this paper, we make progress on understanding the expressivity gap between SSMs and attention in language modeling, and on reducing the hardware barrier between SSMs and attention. First, we use synthetic language modeling tasks to understand the gap between SSMs and attention. We find that existing SSMs struggle with two capabilities: recalling earlier tokens in the sequence and comparing tokens across the sequence. To understand the impact on language modeling, we propose a new SSM layer, H3, that is explicitly designed for these abilities. H3 matches attention on the synthetic languages and comes within 0.4 PPL of Transformers on OpenWebText. Furthermore, a hybrid 125M-parameter H3-attention model that retains two attention layers surprisingly outperforms Transformers on OpenWebText by 1.0 PPL. Next, to improve the efficiency of training SSMs on modern hardware, we propose FlashConv. FlashConv uses a fused block FFT algorithm to improve efficiency on sequences up to 8K, and introduces a novel state passing algorithm that exploits the recurrent properties of SSMs to scale to longer sequences. FlashConv yields 2$\\times$ speedup on the long-range arena benchmark and allows hybrid language models to generate text 2.4$\\times$ faster than Transformers. Using FlashConv, we scale hybrid H3-attention language models up to 2.7B parameters on the Pile and find promising initial results, achieving lower perplexity than Transformers and outperforming Transformers in zero- and few-shot learning on a majority of tasks in the SuperGLUE benchmark."
                },
                {
                    "title": "Dilated Neighborhood Attention Transformer",
                    "abstract": "Transformers are quickly becoming one of the most heavily applied deep learning architectures across modalities, domains, and tasks. In vision, on top of ongoing efforts into plain transformers, hierarchical transformers have also gained significant attention, thanks to their performance and easy integration into existing frameworks. These models typically employ localized attention mechanisms, such as the sliding-window Neighborhood Attention (NA) or Swin Transformer's Shifted Window Self Attention. While effective at reducing self attention's quadratic complexity, local attention weakens two of the most desirable properties of self attention: long range inter-dependency modeling, and global receptive field. In this paper, we introduce Dilated Neighborhood Attention (DiNA), a natural, flexible and efficient extension to NA that can capture more global context and expand receptive fields exponentially at no additional cost. NA's local attention and DiNA's sparse global attention complement each other, and therefore we introduce Dilated Neighborhood Attention Transformer (DiNAT), a new hierarchical vision transformer built upon both. DiNAT variants enjoy significant improvements over strong baselines such as NAT, Swin, and ConvNeXt. Our large model is faster and ahead of its Swin counterpart by 1.6% box AP in COCO object detection, 1.4% mask AP in COCO instance segmentation, and 1.4% mIoU in ADE20K semantic segmentation. Paired with new frameworks, our large variant is the new state of the art panoptic segmentation model on COCO (58.5 PQ) and ADE20K (49.4 PQ), and instance segmentation model on Cityscapes (45.1 AP) and ADE20K (35.4 AP) (no extra data). It also matches the state of the art specialized semantic segmentation models on ADE20K (58.1 mIoU), and ranks second on Cityscapes (84.5 mIoU) (no extra data)."
                },
                {
                    "title": "Hydra Attention: Efficient Attention with Many Heads",
                    "abstract": "While transformers have begun to dominate many tasks in vision, applying them to large images is still computationally difficult. A large reason for this is that self-attention scales quadratically with the number of tokens, which in turn, scales quadratically with the image size. On larger images (e.g., 1080p), over 60% of the total computation in the network is spent solely on creating and applying attention matrices. We take a step toward solving this issue by introducing Hydra Attention, an extremely efficient attention operation for Vision Transformers (ViTs). Paradoxically, this efficiency comes from taking multi-head attention to its extreme: by using as many attention heads as there are features, Hydra Attention is computationally linear in both tokens and features with no hidden constants, making it significantly faster than standard self-attention in an off-the-shelf ViT-B/16 by a factor of the token count. Moreover, Hydra Attention retains high accuracy on ImageNet and, in some cases, actually improves it."
                },
                {
                    "title": "Sharpness-Aware Training for Free",
                    "abstract": "Modern deep neural networks (DNNs) have achieved state-of-the-art performances but are typically over-parameterized. The over-parameterization may result in undesirably large generalization error in the absence of other customized training strategies. Recently, a line of research under the name of Sharpness-Aware Minimization (SAM) has shown that minimizing a sharpness measure, which reflects the geometry of the loss landscape, can significantly reduce the generalization error. However, SAM-like methods incur a two-fold computational overhead of the given base optimizer (e.g. SGD) for approximating the sharpness measure. In this paper, we propose Sharpness-Aware Training for Free, or SAF, which mitigates the sharp landscape at almost zero additional computational cost over the base optimizer. Intuitively, SAF achieves this by avoiding sudden drops in the loss in the sharp local minima throughout the trajectory of the updates of the weights. Specifically, we suggest a novel trajectory loss, based on the KL-divergence between the outputs of DNNs with the current weights and past weights, as a replacement of the SAM's sharpness measure. This loss captures the rate of change of the training loss along the model's update trajectory. By minimizing it, SAF ensures the convergence to a flat minimum with improved generalization capabilities. Extensive empirical results show that SAF minimizes the sharpness in the same way that SAM does, yielding better results on the ImageNet dataset with essentially the same computational cost as the base optimizer."
                },
                {
                    "title": "Neighborhood Attention Transformer",
                    "abstract": "We present Neighborhood Attention (NA), the first efficient and scalable sliding window attention mechanism for vision. NA is a pixel-wise operation, localizing self attention (SA) to the nearest neighboring pixels, and therefore enjoys a linear time and space complexity compared to the quadratic complexity of SA. The sliding window pattern allows NA's receptive field to grow without needing extra pixel shifts, and preserves translational equivariance, unlike Swin Transformer's Window Self Attention (WSA). We develop NATTEN (Neighborhood Attention Extension), a Python package with efficient C++ and CUDA kernels, which allows NA to run up to 40% faster than Swin's WSA while using up to 25% less memory. We further present Neighborhood Attention Transformer (NAT), a new hierarchical transformer design based on NA that boosts image classification and downstream vision performance. Experimental results on NAT are competitive; NAT-Tiny reaches 83.2% top-1 accuracy on ImageNet, 51.4% mAP on MSCOCO and 48.4% mIoU on ADE20K, which is 1.9% ImageNet accuracy, 1.0% COCO mAP, and 2.6% ADE20K mIoU improvement over a Swin model with similar size. To support more research based on sliding window attention, we open source our project and release our checkpoints."
                },
                {
                    "title": "Transformer Quality in Linear Time",
                    "abstract": "We revisit the design choices in Transformers, and propose methods to address their weaknesses in handling long sequences. First, we propose a simple layer named gated attention unit, which allows the use of a weaker single-head attention with minimal quality loss. We then propose a linear approximation method complementary to this new layer, which is accelerator-friendly and highly competitive in quality. The resulting model, named FLASH, matches the perplexity of improved Transformers over both short (512) and long (8K) context lengths, achieving training speedups of up to 4.9$\\times$ on Wiki-40B and 12.1$\\times$ on PG-19 for auto-regressive language modeling, and 4.8$\\times$ on C4 for masked language modeling."
                },
                {
                    "title": "cosFormer: Rethinking Softmax in Attention",
                    "abstract": "Transformer has shown great successes in natural language processing, computer vision, and audio processing. As one of its core components, the softmax attention helps to capture long-range dependencies yet prohibits its scale-up due to the quadratic space and time complexity to the sequence length. Kernel methods are often adopted to reduce the complexity by approximating the softmax operator. Nevertheless, due to the approximation errors, their performances vary in different tasks/corpus and suffer crucial performance drops when compared with the vanilla softmax attention. In this paper, we propose a linear transformer called cosFormer that can achieve comparable or better accuracy to the vanilla transformer in both casual and cross attentions. cosFormer is based on two key properties of softmax attention: i). non-negativeness of the attention matrix; ii). a non-linear re-weighting scheme that can concentrate the distribution of the attention matrix. As its linear substitute, cosFormer fulfills these properties with a linear operator and a cosine-based distance re-weighting mechanism. Extensive experiments on language modeling and text understanding tasks demonstrate the effectiveness of our method. We further examine our method on long sequences and achieve state-of-the-art performance on the Long-Range Arena benchmark. The source code is available at https://github.com/OpenNLPLab/cosFormer."
                },
                {
                    "title": "A ConvNet for the 2020s",
                    "abstract": "The \u201cRoaring 20s\u201d of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually \u201cmodernize\u201d a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets."
                },
                {
                    "title": "Vision Transformer with Deformable Attention",
                    "abstract": "Transformers have recently shown superior performances on various vision tasks. The large, sometimes even global, receptive field endows Transformer models with higher representation power over their CNN counterparts. Nevertheless, simply enlarging receptive field also gives rise to several concerns. On the one hand, using dense attention e.g., in ViT, leads to excessive memory and computational cost, and features can be influenced by irrelevant parts which are beyond the region of interests. On the other hand, the sparse attention adopted in PVT or Swin Transformer is data agnostic and may limit the ability to model long range relations. To mitigate these issues, we propose a novel deformable selfattention module, where the positions of key and value pairs in selfattention are selected in a data-dependent way. This flexible scheme enables the self-attention module to focus on relevant re-gions and capture more informative features. On this basis, we present Deformable Attention Transformer, a general backbone model with deformable attention for both image classification and dense prediction tasks. Extensive experi-ments show that our models achieve consistently improved results on comprehensive benchmarks. Code is available at https://github.com/LeapLabTHU/DAT."
                },
                {
                    "title": "MViTv2: Improved Multiscale Vision Transformers for Classification and Detection",
                    "abstract": "In this paper, we study Multiscale Vision Transformers (MViTv2) as a unified architecture for image and video classification, as well as object detection. We present an improved version of MViT that incorporates decomposed relative positional embeddings and residual pooling connections. We instantiate this architecture in five sizes and evaluate it for ImageNet classification, COCO detection and Kinetics video recognition where it outperforms prior work. We further compare MViTv2s' pooling attention to window attention mechanisms where it outperforms the latter in accuracy/compute. Without bells-and-whistles, MViTv2 has state-of-the-art performance in 3 domains: 88.8% accuracy on ImageNet classification, 58.7 APbox on COCO object detection as well as 86.1% on Kinetics-400 video classification. Code and models are available at https://github.com/facebookresearch/mvit."
                },
                {
                    "title": "SOFT: Softmax-free Transformer with Linear Complexity",
                    "abstract": "Vision transformers (ViTs) have pushed the state-of-the-art for various visual recognition tasks by patch-wise image tokenization followed by self-attention. However, the employment of self-attention modules results in a quadratic complexity in both computation and memory usage. Various attempts on approximating the self-attention computation with linear complexity have been made in Natural Language Processing. However, an in-depth analysis in this work shows that they are either theoretically flawed or empirically ineffective for visual recognition. We further identify that their limitations are rooted in keeping the softmax self-attention during approximations. Specifically, conventional self-attention is computed by normalizing the scaled dot-product between token feature vectors. Keeping this softmax operation challenges any subsequent linearization efforts. Based on this insight, for the first time, a softmax-free transformer or SOFT is proposed. To remove softmax in self-attention, Gaussian kernel function is used to replace the dot-product similarity without further normalization. This enables a full self-attention matrix to be approximated via a low-rank matrix decomposition. The robustness of the approximation is achieved by calculating its Moore-Penrose inverse using a Newton-Raphson method. Extensive experiments on ImageNet show that our SOFT significantly improves the computational efficiency of existing ViT variants. Crucially, with a linear complexity, much longer token sequences are permitted in SOFT, resulting in superior trade-off between accuracy and complexity."
                },
                {
                    "title": "CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows",
                    "abstract": "We present CSWin Transformer, an efficient and effective Transformer-based backbone for general-purpose vision tasks. A challenging issue in Transformer design is that global self-attention is very expensive to compute whereas local self-attention often limits the field of interactions of each token. To address this issue, we develop the Cross-Shaped Window self-attention mechanism for computing self-attention in the horizontal and vertical stripes in parallel that form a cross-shaped window, with each stripe obtained by splitting the input feature into stripes of equal width. We provide a mathematical analysis of the effect of the stripe width and vary the stripe width for different layers of the Transformer network which achieves strong modeling capability while limiting the computation cost. We also introduce Locally-enhanced Positional Encoding (LePE), which handles the local positional information better than existing encoding schemes. LePE naturally supports arbitrary input resolutions, and is thus especially effective and friendly for downstream tasks. Incorporated with these designs and a hierarchical structure, CSWin Transformer demonstrates competitive performance on common vision tasks. Specifically, it achieves 85.4% Top-1 accuracy on ImageNet-1K without any extra training data or label, 53.9 box AP and 46.4 mask AP on the COCO detection task, and 52.2 mIOU on the ADE20K semantic segmentation task, surpassing previous state-of-the-art Swin Transformer backbone by +1.2, +2.0, +1.4, and +2.0 respectively under the similar FLOPs setting. By further pretraining on the larger dataset ImageNet-21K, we achieve 87.5% Top-1 accuracy on ImageNet-1K and high segmentation performance on ADE20K with 55.7 mIoU. 11Code and pretrain model is available at https://github.com/microsoft/CSWin-Transformer"
                },
                {
                    "title": "Focal Self-attention for Local-Global Interactions in Vision Transformers",
                    "abstract": "Recently, Vision Transformer and its variants have shown great promise on various computer vision tasks. The ability of capturing short- and long-range visual dependencies through self-attention is arguably the main source for the success. But it also brings challenges due to quadratic computational overhead, especially for the high-resolution vision tasks (e.g., object detection). In this paper, we present focal self-attention, a new mechanism that incorporates both fine-grained local and coarse-grained global interactions. Using this new mechanism, each token attends the closest surrounding tokens at fine granularity but the tokens far away at coarse granularity, and thus can capture both short- and long-range visual dependencies efficiently and effectively. With focal self-attention, we propose a new variant of Vision Transformer models, called Focal Transformer, which achieves superior performance over the state-of-the-art vision Transformers on a range of public image classification and object detection benchmarks. In particular, our Focal Transformer models with a moderate size of 51.1M and a larger size of 89.8M achieve 83.5 and 83.8 Top-1 accuracy, respectively, on ImageNet classification at 224x224 resolution. Using Focal Transformers as the backbones, we obtain consistent and substantial improvements over the current state-of-the-art Swin Transformers for 6 different object detection methods trained with standard 1x and 3x schedules. Our largest Focal Transformer yields 58.7/58.9 box mAPs and 50.9/51.3 mask mAPs on COCO mini-val/test-dev, and 55.4 mIoU on ADE20K for semantic segmentation, creating new SoTA on three of the most challenging computer vision tasks."
                },
                {
                    "title": "Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions",
                    "abstract": "Although convolutional neural networks (CNNs) have achieved great success in computer vision, this work investigates a simpler, convolution-free backbone network use-fid for many dense prediction tasks. Unlike the recently-proposed Vision Transformer (ViT) that was designed for image classification specifically, we introduce the Pyramid Vision Transformer (PVT), which overcomes the difficulties of porting Transformer to various dense prediction tasks. PVT has several merits compared to current state of the arts. (1) Different from ViT that typically yields low-resolution outputs and incurs high computational and memory costs, PVT not only can be trained on dense partitions of an image to achieve high output resolution, which is important for dense prediction, but also uses a progressive shrinking pyramid to reduce the computations of large feature maps. (2) PVT inherits the advantages of both CNN and Transformer, making it a unified backbone for various vision tasks without convolutions, where it can be used as a direct replacement for CNN backbones. (3) We validate PVT through extensive experiments, showing that it boosts the performance of many downstream tasks, including object detection, instance and semantic segmentation. For example, with a comparable number of parameters, PVT+RetinaNet achieves 40.4 AP on the COCO dataset, surpassing ResNet50+RetinNet (36.3 AP) by 4.1 absolute AP (see Figure 2). We hope that PVT could, serre as an alternative and useful backbone for pixel-level predictions and facilitate future research."
                },
                {
                    "title": "Conditional Positional Encodings for Vision Transformers",
                    "abstract": "We propose a conditional positional encoding (CPE) scheme for vision Transformers. Unlike previous fixed or learnable positional encodings, which are pre-defined and independent of input tokens, CPE is dynamically generated and conditioned on the local neighborhood of the input tokens. As a result, CPE can easily generalize to the input sequences that are longer than what the model has ever seen during training. Besides, CPE can keep the desired translation-invariance in the image classification task, resulting in improved performance. We implement CPE with a simple Position Encoding Generator (PEG) to get seamlessly incorporated into the current Transformer framework. Built on PEG, we present Conditional Position encoding Vision Transformer (CPVT). We demonstrate that CPVT has visually similar attention maps compared to those with learned positional encodings and delivers outperforming results. Our code is available at https://github.com/Meituan-AutoML/CPVT ."
                },
                {
                    "title": "Nystr\u00f6mformer: A Nystr\u00f6m-Based Algorithm for Approximating Self-Attention",
                    "abstract": "Transformers have emerged as a powerful tool for a broad range of natural language processing tasks. A key component that drives the impressive performance of Transformers is the self-attention mechanism that encodes the influence or dependence of other tokens on each specific token. While beneficial, the quadratic complexity of self-attention on the input sequence length has limited its application to longer sequences - a topic being actively studied in the community. To address this limitation, we propose Nystr\u00f6mformer - a model that exhibits favorable scalability as a function of sequence length. Our idea is based on adapting the Nystr\u00f6m method to approximate standard self-attention with O(n) complexity. The scalability of Nystr\u00f6mformer enables application to longer sequences with thousands of tokens. We perform evaluations on multiple downstream tasks on the GLUE benchmark and IMDB reviews with standard sequence length, and find that our Nystr\u00f6mformer performs comparably, or in a few cases, even slightly better, than standard self-attention. On longer sequence tasks in the Long Range Arena (LRA) benchmark, Nystr\u00f6mformer performs favorably relative to other efficient self-attention methods. Our code is available at https://github.com/mlpen/Nystromformer."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Rethinking Attention with Performers",
                    "abstract": "We introduce Performers, Transformer architectures which can estimate regular (softmax) full-rank-attention Transformers with provable accuracy, but using only linear (as opposed to quadratic) space and time complexity, without relying on any priors such as sparsity or low-rankness. To approximate softmax attention-kernels, Performers use a novel Fast Attention Via positive Orthogonal Random features approach (FAVOR+), which may be of independent interest for scalable kernel methods. FAVOR+ can be also used to efficiently model kernelizable attention mechanisms beyond softmax. This representational power is crucial to accurately compare softmax with other kernels for the first time on large-scale tasks, beyond the reach of regular Transformers, and investigate optimal attention-kernels. Performers are linear architectures fully compatible with regular Transformers and with strong theoretical guarantees: unbiased or nearly-unbiased estimation of the attention matrix, uniform convergence and low estimation variance. We tested Performers on a rich set of tasks stretching from pixel-prediction through text models to protein sequence modeling. We demonstrate competitive results with other examined efficient sparse and dense attention methods, showcasing effectiveness of the novel attention-learning paradigm leveraged by Performers."
                },
                {
                    "title": "Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention",
                    "abstract": "Transformers achieve remarkable performance in several tasks but due to their quadratic complexity, with respect to the input's length, they are prohibitively slow for very long sequences. To address this limitation, we express the self-attention as a linear dot-product of kernel feature maps and make use of the associativity property of matrix products to reduce the complexity from $\\mathcal{O}\\left(N^2\\right)$ to $\\mathcal{O}\\left(N\\right)$, where $N$ is the sequence length. We show that this formulation permits an iterative implementation that dramatically accelerates autoregressive transformers and reveals their relationship to recurrent neural networks. Our linear transformers achieve similar performance to vanilla transformers and they are up to 4000x faster on autoregressive prediction of very long sequences."
                },
                {
                    "title": "Randaugment: Practical automated data augmentation with a reduced search space",
                    "abstract": "Recent work on automated augmentation strategies has led to state-of-the-art results in image classification and object detection. An obstacle to a large-scale adoption of these methods is that they require a separate and expensive search phase. A common way to overcome the expense of the search phase was to use a smaller proxy task. However, it was not clear if the optimized hyperparameters found on the proxy task are also optimal for the actual task. In this work, we rethink the process of designing automated augmentation strategies. We find that while previous work required a search for both magnitude and probability of each operation independently, it is sufficient to only search for a single distortion magnitude that jointly controls all operations. We hence propose a simplified search space that vastly reduces the computational expense of automated augmentation, and permits the removal of a separate proxy task.Despite the simplifications, our method achieves equal or better performance over previous automated augmentation strategies on on CIFAR-10/100, SVHN, ImageNet and COCO datasets. EfficientNet-B7, we achieve 85.0% accuracy, a 1.0% increase over baseline augmentation, a 0.6% improvement over AutoAugment on the ImageNet dataset. With EfficientNet-B8, we achieve 85.4% accuracy on ImageNet, which matches a previous result that used 3.5B extra images. On object detection, the same method as classification leads to 1.0-1.3% improvement over baseline augmentation. Code will be made available online."
                },
                {
                    "title": "CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features",
                    "abstract": "Regional dropout strategies have been proposed to enhance performance of convolutional neural network classifiers. They have proved to be effective for guiding the model to attend on less discriminative parts of objects (e.g. leg as opposed to head of a person), thereby letting the network generalize better and have better object localization capabilities. On the other hand, current methods for regional dropout removes informative pixels on training images by overlaying a patch of either black pixels or random noise. Such removal is not desirable because it suffers from information loss causing inefficiency in training. We therefore propose the CutMix augmentation strategy: patches are cut and pasted among training images where the ground truth labels are also mixed proportionally to the area of the patches. By making efficient use of training pixels and retaining the regularization effect of regional dropout, CutMix consistently outperforms state-of-the-art augmentation strategies on CIFAR and ImageNet classification tasks, as well as on ImageNet weakly-supervised localization task. Moreover, unlike previous augmentation methods, our CutMix-trained ImageNet classifier, when used as a pretrained model, results in consistent performance gain in Pascal detection and MS-COCO image captioning benchmarks. We also show that CutMix can improve the model robustness against input corruptions and its out-of distribution detection performance."
                },
                {
                    "title": "Efficient Attention: Attention with Linear Complexities",
                    "abstract": "Dot-product attention has wide applications in computer vision and natural language processing. However, its memory and computational costs grow quadratically with the input size. Such growth prohibits its application on high- resolution inputs. To remedy this drawback, this paper proposes a novel efficient attention mechanism equivalent to dot-product attention but with substantially less memory and computational costs. Its resource efficiency allows more widespread and flexible integration of attention modules into a network, which leads to better accuracies. Empirical evaluations demonstrated the effectiveness of its advantages. Efficient attention modules brought significant performance boosts to object detectors and instance segmenters on MS-COCO 2017. Further, the resource efficiency democratizes attention to complex models, where high costs prohibit the use of dot-product attention. As an exemplar, a model with efficient attention achieved state-of- the-art accuracies for stereo depth estimation on the Scene Flow dataset. Code is available at https://github.com/cmsflash/efficient-attention."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "mixup: Beyond Empirical Risk Minimization",
                    "abstract": "Large deep neural networks are powerful, but exhibit undesirable behaviors such as memorization and sensitivity to adversarial examples. In this work, we propose mixup, a simple learning principle to alleviate these issues. In essence, mixup trains a neural network on convex combinations of pairs of examples and their labels. By doing so, mixup regularizes the neural network to favor simple linear behavior in-between training examples. Our experiments on the ImageNet-2012, CIFAR-10, CIFAR-100, Google commands and UCI datasets show that mixup improves the generalization of state-of-the-art neural network architectures. We also find that mixup reduces the memorization of corrupt labels, increases the robustness to adversarial examples, and stabilizes the training of generative adversarial networks."
                },
                {
                    "title": "Random Erasing Data Augmentation",
                    "abstract": "In this paper, we introduce Random Erasing, a new data augmentation method for training the convolutional neural network (CNN). In training, Random Erasing randomly selects a rectangle region in an image and erases its pixels with random values. In this process, training images with various levels of occlusion are generated, which reduces the risk of over-fitting and makes the model robust to occlusion. Random Erasing is parameter learning free, easy to implement, and can be integrated with most of the CNN-based recognition models. Albeit simple, Random Erasing is complementary to commonly used data augmentation techniques such as random cropping and flipping, and yields consistent improvement over strong baselines in image classification, object detection and person re-identification. Code is available at: https://github.com/zhunzhong07/Random-Erasing."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Mask R-CNN",
                    "abstract": "We present a conceptually simple, flexible, and general framework for object instance segmentation. Our approach efficiently detects objects in an image while simultaneously generating a high-quality segmentation mask for each instance. The method, called Mask R-CNN, extends Faster R-CNN by adding a branch for predicting an object mask in parallel with the existing branch for bounding box recognition. Mask R-CNN is simple to train and adds only a small overhead to Faster R-CNN, running at 5 fps. Moreover, Mask R-CNN is easy to generalize to other tasks, e.g., allowing us to estimate human poses in the same framework. We show top results in all three tracks of the COCO suite of challenges, including instance segmentation, bounding-box object detection, and person keypoint detection. Without tricks, Mask R-CNN outperforms all existing, single-model entries on every task, including the COCO 2016 challenge winners. We hope our simple and effective approach will serve as a solid baseline and help ease future research in instance-level recognition. Code will be made available."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nWhat factors contribute to Mamba\u2019s success and its significant superiority to linear attention Transformer?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it can lead to a deeper understanding of the mechanisms that enable effective sequence modeling in state space models like Mamba. By elucidating the distinctions between Mamba and linear attention Transformers, this research could pave the way for the development of more efficient models that maintain high expressive power while operating with linear complexity. This advancement could significantly impact future research in natural language processing and visual recognition, leading to practical applications that require processing of extremely long sequences without prohibitive computational costs.\n\n### [Question 3] - Why is it hard?\nThe challenges in addressing this problem stem from the need to dissect the intricate design elements of Mamba and their contributions to its performance. Naive approaches may fail because they might overlook the nuanced interactions between the model components, such as the input gate, forget gate, and block design. Additionally, the theoretical underpinnings of how these components enhance expressive power while maintaining efficiency are complex and require rigorous empirical validation. Overcoming these technical obstacles necessitates a comprehensive analysis of both the theoretical framework and practical implementations.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on the efficiency of linear attention mechanisms without adequately exploring their expressive limitations. The lack of a unified framework to compare Mamba and linear attention Transformers has created a gap in understanding their relationship. Barriers such as insufficient empirical studies on the impact of specific design elements and the complexity of analyzing recurrent computations have hindered progress. This research differs by providing a detailed comparative analysis and empirical validation of Mamba's design features, thereby addressing the limitations of prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nThe proposed methodology involves a theoretical and empirical analysis of Mamba and linear attention Transformers, framed within a unified formulation. The study will utilize vision tasks as the dataset and evaluate the performance of Mamba and the newly introduced Mamba-Like Linear Attention (MLLA) model using metrics such as accuracy in image classification and high-resolution dense prediction tasks. The expected outcomes include a clearer understanding of the contributions of the forget gate and block design to Mamba's performance, as well as demonstrating that MLLA can achieve superior results compared to existing vision Mamba models, validating the potential of linear attention with the identified core designs."
            }
        },
        "author_data": {
            "f4924d55-c831-4584-bd41-728345b4aebe": {
                "pk": "f4924d55-c831-4584-bd41-728345b4aebe",
                "name": "Dongchen Han",
                "collaborators": [
                    "Shiji Song",
                    "Gao Huang",
                    "Yizeng Han",
                    "Xuran Pan",
                    "Zhuofan Xia",
                    "Tianzhu Ye",
                    "Xiaochun Cao",
                    "Yifan Pu",
                    "Haojun Jiang",
                    "Yuanze Lin"
                ],
                "domain": [
                    "Computer Vision",
                    "Transformer Models",
                    "Multi-modal Learning",
                    "Adversarial Attacks"
                ],
                "publications": [
                    {
                        "title": "FLatten Transformer: Vision Transformer using Focused Linear Attention",
                        "abstract": "The quadratic computation complexity of self-attention has been a persistent challenge when applying Transformer models to vision tasks. Linear attention, on the other hand, offers a much more efficient alternative with its linear complexity by approximating the Softmax operation through carefully designed mapping functions. However, current linear attention approaches either suffer from significant performance degradation or introduce additional computation overhead from the mapping functions. In this paper, we propose a novel Focused Linear Attention module to achieve both high efficiency and expressiveness. Specifically, we first analyze the factors contributing to the performance degradation of linear attention from two perspectives: the focus ability and feature diversity. To overcome these limitations, we introduce a simple yet effective mapping function and an efficient rank restoration module to enhance the expressiveness of self-attention while maintaining low computation complexity. Extensive experiments show that our linear attention module is applicable to a variety of advanced vision Transformers, and achieves consistently improved performances on multiple benchmarks. Code is available at https://github.com/LeapLabTHU/FLatten-Transformer."
                    },
                    {
                        "title": "Contrastive Language-Image Pre-Training with Knowledge Graphs",
                        "abstract": "Recent years have witnessed the fast development of large-scale pre-training frameworks that can extract multi-modal representations in a unified form and achieve promising performances when transferred to downstream tasks. Nevertheless, existing approaches mainly focus on pre-training with simple image-text pairs, while neglecting the semantic connections between concepts from different modalities. In this paper, we propose a knowledge-based pre-training framework, dubbed Knowledge-CLIP, which injects semantic information into the widely used CLIP model. Through introducing knowledge-based objectives in the pre-training process and utilizing different types of knowledge graphs as training data, our model can semantically align the representations in vision and language with higher quality, and enhance the reasoning ability across scenarios and modalities. Extensive experiments on various vision-language downstream tasks demonstrate the effectiveness of Knowledge-CLIP compared with the original CLIP and competitive baselines."
                    },
                    {
                        "title": "Pseudo-Q: Generating Pseudo Language Queries for Visual Grounding",
                        "abstract": "Visual grounding, i.e., localizing objects in images according to natural language queries, is an important topic in visual language understanding. The most effective approaches for this task are based on deep learning, which generally require expensive manually labeled image-query or patch-query pairs. To eliminate the heavy dependence on human annotations, we present a novel method, named Pseudo-Q, to automatically generate pseudo language queries for supervised training. Our method leverages an off-the-shelf object detector to identify visual objects from unlabeled images, and then language queries for these objects are obtained in an unsupervised fashion with a pseudo-query generation module. Then, we design a task-related query prompt module to specifically tailor generated pseudo language queries for visual grounding tasks. Further, in order to fully capture the contextual relationships between images and language queries, we develop a visual-language model equipped with multi-level cross-modality attention mechanism. Extensive experimental results demonstrate that our method has two notable benefits: (1) it can reduce human annotation costs significantly, e.g., 31% on RefCOCO without degrading original model's performance under the fully supervised setting, and (2) without bells and whistles, it achieves superior or comparable performance compared to state-of-the-art weakly-supervised visual grounding methods on all the five datasets we have experimented. Code is available at https://github.com/LeapLabTHU/Pseudo-Q."
                    },
                    {
                        "title": "GSVA: Generalized Segmentation via Multimodal Large Language Models",
                        "abstract": "Generalized Referring Expression Segmentation (GRES) extends the scope of classic RES to refer to multiple objects in one expression or identify the empty targets absent in the image. GRES poses challenges in modeling the complex spatial relationships of the instances in the image and identifying non-existing referents. Multimodal Large Language Models (MLLMs) have recently shown tremendous progress in these complicated vision-language tasks. Connecting Large Language Models (LLMs) and vision models, MLLMs are proficient in understanding contexts with visual inputs. Among them, LISA, as a representative, adopts a special [SEG] token to prompt a segmentation mask decoder, e.g., SAM, to enable MLLMs in the RES task. However, existing solutions to GRES remain unsatisfactory since current segmentation MLLMs cannot correctly handle the cases where users might reference multiple subjects in a singular prompt or provide descriptions incongruent with any image target. In this paper, we propose Generalized Segmentation Vision Assistant (GSVA) to address this gap. Specifically, GSVA reuses the [SEG] token to prompt the segmentation model towards supporting multiple mask references simultaneously and innovatively learns to generate a [REJ] token to reject the null targets explicitly. Experiments validate GSVA's efficacy in resolving the GRES issue, marking a notable enhancement and setting a new record on the GRES benchmark gRefCOCO dataset. GSVA also proves effective across various classic referring segmentation and comprehension tasks."
                    },
                    {
                        "title": "OT-Attack: Enhancing Adversarial Transferability of Vision-Language Models via Optimal Transport Optimization",
                        "abstract": "Vision-language pre-training (VLP) models demonstrate impressive abilities in processing both images and text. However, they are vulnerable to multi-modal adversarial examples (AEs). Investigating the generation of high-transferability adversarial examples is crucial for uncovering VLP models' vulnerabilities in practical scenarios. Recent works have indicated that leveraging data augmentation and image-text modal interactions can enhance the transferability of adversarial examples for VLP models significantly. However, they do not consider the optimal alignment problem between dataaugmented image-text pairs. This oversight leads to adversarial examples that are overly tailored to the source model, thus limiting improvements in transferability. In our research, we first explore the interplay between image sets produced through data augmentation and their corresponding text sets. We find that augmented image samples can align optimally with certain texts while exhibiting less relevance to others. Motivated by this, we propose an Optimal Transport-based Adversarial Attack, dubbed OT-Attack. The proposed method formulates the features of image and text sets as two distinct distributions and employs optimal transport theory to determine the most efficient mapping between them. This optimal mapping informs our generation of adversarial examples to effectively counteract the overfitting issues. Extensive experiments across various network architectures and datasets in image-text matching tasks reveal that our OT-Attack outperforms existing state-of-the-art methods in terms of adversarial transferability."
                    },
                    {
                        "title": "Dynamic Perceiver for Efficient Visual Recognition",
                        "abstract": "Early exiting has become a promising approach to improving the inference efficiency of deep networks. By structuring models with multiple classifiers (exits), predictions for ``easy'' samples can be generated at earlier exits, negating the need for executing deeper layers. Current multi-exit networks typically implement linear classifiers at intermediate layers, compelling low-level features to encapsulate high-level semantics. This sub-optimal design invariably undermines the performance of later exits. In this paper, we propose Dynamic Perceiver (Dyn-Perceiver) to decouple the feature extraction procedure and the early classification task with a novel dual-branch architecture. A feature branch serves to extract image features, while a classification branch processes a latent code assigned for classification tasks. Bi-directional cross-attention layers are established to progressively fuse the information of both branches. Early exits are placed exclusively within the classification branch, thus eliminating the need for linear separability in low-level features. Dyn-Perceiver constitutes a versatile and adaptable framework that can be built upon various architectures. Experiments on image classification, action recognition, and object detection demonstrate that our method significantly improves the inference efficiency of different backbones, outperforming numerous competitive approaches across a broad range of computational budgets. Evaluation on both CPU and GPU platforms substantiate the superior practical efficiency of Dyn-Perceiver. Code is available at https://www.github.com/LeapLabTHU/Dynamic_Perceiver."
                    },
                    {
                        "title": "Agent Attention: On the Integration of Softmax and Linear Attention",
                        "abstract": "The attention module is the key component in Transformers. While the global attention mechanism offers high expressiveness, its excessive computational cost restricts its applicability in various scenarios. In this paper, we propose a novel attention paradigm, Agent Attention, to strike a favorable balance between computational efficiency and representation power. Specifically, the Agent Attention, denoted as a quadruple $(Q, A, K, V)$, introduces an additional set of agent tokens $A$ into the conventional attention module. The agent tokens first act as the agent for the query tokens $Q$ to aggregate information from $K$ and $V$, and then broadcast the information back to $Q$. Given the number of agent tokens can be designed to be much smaller than the number of query tokens, the agent attention is significantly more efficient than the widely adopted Softmax attention, while preserving global context modelling capability. Interestingly, we show that the proposed agent attention is equivalent to a generalized form of linear attention. Therefore, agent attention seamlessly integrates the powerful Softmax attention and the highly efficient linear attention. Extensive experiments demonstrate the effectiveness of agent attention with various vision Transformers and across diverse vision tasks, including image classification, object detection, semantic segmentation and image generation. Notably, agent attention has shown remarkable performance in high-resolution scenarios, owning to its linear attention nature. For instance, when applied to Stable Diffusion, our agent attention accelerates generation and substantially enhances image generation quality without any additional training. Code is available at https://github.com/LeapLabTHU/Agent-Attention."
                    },
                    {
                        "title": "VL-Trojan: Multimodal Instruction Backdoor Attacks against Autoregressive Visual Language Models",
                        "abstract": "Autoregressive Visual Language Models (VLMs) showcase impressive few-shot learning capabilities in a multimodal context. Recently, multimodal instruction tuning has been proposed to further enhance instruction-following abilities. However, we uncover the potential threat posed by backdoor attacks on autoregressive VLMs during instruction tuning. Adversaries can implant a backdoor by injecting poisoned samples with triggers embedded in instructions or images, enabling malicious manipulation of the victim model's predictions with predefined triggers. Nevertheless, the frozen visual encoder in autoregressive VLMs imposes constraints on the learning of conventional image triggers. Additionally, adversaries may encounter restrictions in accessing the parameters and architectures of the victim model. To address these challenges, we propose a multimodal instruction backdoor attack, namely VL-Trojan. Our approach facilitates image trigger learning through an isolating and clustering strategy and enhance black-box-attack efficacy via an iterative character-level text trigger generation method. Our attack successfully induces target outputs during inference, significantly surpassing baselines (+62.52\\%) in ASR. Moreover, it demonstrates robustness across various model scales and few-shot in-context reasoning scenarios."
                    },
                    {
                        "title": "Efficient Diffusion Transformer with Step-wise Dynamic Attention Mediators",
                        "abstract": "This paper identifies significant redundancy in the query-key interactions within self-attention mechanisms of diffusion transformer models, particularly during the early stages of denoising diffusion steps. In response to this observation, we present a novel diffusion transformer framework incorporating an additional set of mediator tokens to engage with queries and keys separately. By modulating the number of mediator tokens during the denoising generation phases, our model initiates the denoising process with a precise, non-ambiguous stage and gradually transitions to a phase enriched with detail. Concurrently, integrating mediator tokens simplifies the attention module's complexity to a linear scale, enhancing the efficiency of global attention processes. Additionally, we propose a time-step dynamic mediator token adjustment mechanism that further decreases the required computational FLOPs for generation, simultaneously facilitating the generation of high-quality images within the constraints of varied inference budgets. Extensive experiments demonstrate that the proposed method can improve the generated image quality while also reducing the inference cost of diffusion transformers. When integrated with the recent work SiT, our method achieves a state-of-the-art FID score of 2.01. The source code is available at https://github.com/LeapLabTHU/Attention-Mediators."
                    }
                ]
            },
            "780aed2c-7d76-4366-b6c9-d4750077af29": {
                "pk": "780aed2c-7d76-4366-b6c9-d4750077af29",
                "name": "Ziyi Wang",
                "collaborators": [
                    "Evangelos A. Theodorou",
                    "Lijian Jiang",
                    "Shuqi Hu",
                    "Changyu Ren",
                    "Weiwei Zhang",
                    "Xiaojing Wu",
                    "Kongjin Chen",
                    "Grady Williams",
                    "Yong Cui",
                    "Zeqi Lai"
                ],
                "domain": [
                    "Data Assimilation",
                    "Stochastic Control",
                    "Machine Learning",
                    "Artificial Intelligence"
                ],
                "publications": [
                    {
                        "title": "Model free data assimilation with Takens embedding",
                        "abstract": "In many practical scenarios, the dynamical system is not available and standard data assimilation methods are not applicable. Our objective is to construct a data-driven model for state estimation without the underlying dynamics. Instead of directly modeling the observation operator with noisy observation, we establish the state space model of the denoised observation. Through data assimilation techniques, the denoised observation information could be used to recover the original model state. Takens' theorem shows that an embedding of the partial and denoised observation is diffeomorphic to the attractor. This gives a theoretical base for estimating the model state using the reconstruction map. To realize the idea, the procedure consists of offline stage and online stage. In the offline stage, we construct the surrogate dynamics using dynamic mode decomposition with noisy snapshots to learn the transition operator for the denoised observation. The filtering distribution of the denoised observation can be estimated using adaptive ensemble Kalman filter, without knowledge of the model error and observation noise covariances. Then the reconstruction map can be established using the posterior mean of the embedding and its corresponding state. In the online stage, the observation is filtered with the surrogate dynamics. Then the online state estimation can be performed utilizing the reconstruction map and the filtered observation. Furthermore, the idea can be generalized to the nonparametric framework with nonparametric time series prediction methods for chaotic problems. The numerical results show the proposed method can estimate the state distribution without the physical dynamical system."
                    },
                    {
                        "title": "Opinion Dynamics Steering using Stochastic Search",
                        "abstract": "In this paper, we apply the stochastic search dynamic optimization framework for steering opinion dynamics in a partially active population. The framework is grounded on stochastic optimization theory and relies on sampling of candidate solutions from distributions of the exponential family. We derive the distribution parameter update for an open loop, a feedback, and a novel adaptive feedback policy. All three policies are tested on two different opinion dynamics steering scenarios in simulation and compared against a hand designed feedback policy. The results showcase the effectiveness the framework for opinion dynamics control and the robustness of the adaptive feedback policy with respect to the active agent set size."
                    },
                    {
                        "title": "Newton-Maclaurin type inequalities for linear combinations of elementary symmetric functions",
                        "abstract": "In this paper, we establish Newton-Maclaurin type inequalities for functions arising from linear combinations of primitively symmetric polynomials. This generalization extends the classical Newton-Maclaurin inequality to a broader class of functions."
                    },
                    {
                        "title": "A novel unsteady aerodynamic Reduced-Order Modeling method for transonic aeroelastic optimization",
                        "abstract": "In aircraft design, structural optimization and uncertainty quantification concerning transonic aeroelastic issues are computationally impractical, because the iterative process requires great number of aeroelastic analysis. Emerging Reduced-Order Model (ROM) method is convenient for transonic aeroelastic analysis. However, current ROMs cannot be reused during iteration, thus time cost is still way too large. This study proposed an improved ROM suitable for Arbitrary Mode Shapes (ROM-AMS), which is reusable regardless the variation of design variables. By adopting Principal Component Analysis, ROM-AMS method can significantly reduce the number of basis mode shapes and improve the accuracy of flutter analysis. In an optimization case, the weight of cropped delta wing is reduced by 28.46%, and the efficiency is 900 times higher than traditional ROM method, which demonstrates the feasibility of this method in iterative design process."
                    },
                    {
                        "title": "Information Theoretic Model Predictive Control on Jump Diffusion Processes",
                        "abstract": "In this paper we present an information theoretic approach to stochastic optimal control problems for systems with compound Poisson noise. We generalize previous work on information theoretic path integral control to discontinuous dynamics with compound Poisson noise and develop an iterative model predictive control (MPC) algorithm using importance sampling. The proposed algorithm is parallelizable and when implemented on a Graphical Processing Unit (GPU) can run in real time. We test the performance of the proposed algorithm in simulation for two control tasks using a cartpole system and a quadrotor. Our simulations demonstrate improved performance of the new scheme and indicate the importance of incorporating the statistical characteristics of stochastic disturbances in the computation of the stochastic optimal control policies."
                    },
                    {
                        "title": "A First Look at Mobile Intelligence: Architecture, Experimentation and Challenges",
                        "abstract": "Artificial intelligence (AI) technology makes mobile devices become intelligent objects which can learn and act automatically. Although AI will bring great opportunities for mobile applications, little work has focused on the architecture and the interaction with the cloud. In this article, we present three existing architectures of mobile intelligence in detail and introduce its broad application prospects. Furthermore, we conduct a series of experiments to evaluate the performance of the prevalent commercial applications and intelligent frameworks. Our results show that there is a big gap between Quality of Experience (QoE) requirements and the status quo. So far, we have seen only the tip of the iceberg. We pose issues and challenges to advance the area of mobile intelligence and hope to pave the way for the forthcoming."
                    },
                    {
                        "title": "Overcoming Medical Overuse with AI Assistance: An Experimental Investigation",
                        "abstract": "This study evaluates the effectiveness of Artificial Intelligence (AI) in mitigating medical overtreatment, a significant issue characterized by unnecessary interventions that inflate healthcare costs and pose risks to patients. We conducted a lab-in-the-field experiment at a medical school, utilizing a novel medical prescription task, manipulating monetary incentives and the availability of AI assistance among medical students using a three-by-two factorial design. We tested three incentive schemes: Flat (constant pay regardless of treatment quantity), Progressive (pay increases with the number of treatments), and Regressive (penalties for overtreatment) to assess their influence on the adoption and effectiveness of AI assistance. Our findings demonstrate that AI significantly reduced overtreatment rates by up to 62% in the Regressive incentive conditions where (prospective) physician and patient interests were most aligned. Diagnostic accuracy improved by 17% to 37%, depending on the incentive scheme. Adoption of AI advice was high, with approximately half of the participants modifying their decisions based on AI input across all settings. For policy implications, we quantified the monetary (57%) and non-monetary (43%) incentives of overtreatment and highlighted AI's potential to mitigate non-monetary incentives and enhance social welfare. Our results provide valuable insights for healthcare administrators considering AI integration into healthcare systems."
                    },
                    {
                        "title": "Enhancing Low-Precision Sampling via Stochastic Gradient Hamiltonian Monte Carlo",
                        "abstract": "Low-precision training has emerged as a promising low-cost technique to enhance the training efficiency of deep neural networks without sacrificing much accuracy. Its Bayesian counterpart can further provide uncertainty quantification and improved generalization accuracy. This paper investigates low-precision sampling via Stochastic Gradient Hamiltonian Monte Carlo (SGHMC) with low-precision and full-precision gradient accumulators for both strongly log-concave and non-log-concave distributions. Theoretically, our results show that, to achieve $\\epsilon$-error in the 2-Wasserstein distance for non-log-concave distributions, low-precision SGHMC achieves quadratic improvement ($\\widetilde{\\mathbf{O}}\\left({\\epsilon^{-2}{\\mu^*}^{-2}\\log^2\\left({\\epsilon^{-1}}\\right)}\\right)$) compared to the state-of-the-art low-precision sampler, Stochastic Gradient Langevin Dynamics (SGLD) ($\\widetilde{\\mathbf{O}}\\left({{\\epsilon}^{-4}{\\lambda^{*}}^{-1}\\log^5\\left({\\epsilon^{-1}}\\right)}\\right)$). Moreover, we prove that low-precision SGHMC is more robust to the quantization error compared to low-precision SGLD due to the robustness of the momentum-based update w.r.t. gradient noise. Empirically, we conduct experiments on synthetic data, and {MNIST, CIFAR-10 \\& CIFAR-100} datasets, which validate our theoretical findings. Our study highlights the potential of low-precision SGHMC as an efficient and accurate sampling method for large-scale and resource-limited machine learning."
                    },
                    {
                        "title": "An algorithm for Bragg coherent X-ray diffractive imaging of highly strained nanocrystals",
                        "abstract": "By using phase retrieval, Bragg Coherent Diffractive Imaging (BCDI) allows tracking of three-dimensional displacement fields inside individual nanocrystals. Nevertheless, in the presence of significant (1% and higher) strains, such as in the process of a structural phase transformation, fails due to the Bragg peak distortions. Here we present an advanced BCDI algorithm enabling imaging three-dimensional strain fields in highly strained crystals. We test the algorithm on particles simulated to undergo a structural phase transformation. While the conventional algorithm fails in unambiguously reconstructing the phase morphology, our algorithm correctly retrieves the morphology of coexistent phases with a strain difference of 1%. The key novelty is the simultaneous reconstruction of multiple scans of the same nanoparticle at snapshots through the phase transformations. The algorithm enables visualizing phase transformations in nanoparticles of lithium-ion, sodium-ion nanoparticles, and other nanoparticulate materials in working conditions (operando)."
                    },
                    {
                        "title": "Maximum Entropy Differential Dynamic Programming",
                        "abstract": "In this paper, we present a novel maximum entropy formulation of the Differential Dynamic Programming algorithm and derive two variants using unimodal and multimodal value functions parameterizations. By combining the maximum entropy Bellman equations with a particular approximation of the cost function, we are able to obtain a new formulation of Differential Dynamic Programming which is able to escape from local minima via exploration with a multimodal policy. To demonstrate the efficacy of the proposed algorithm, we provide experimental results using four systems on tasks that are represented by cost functions with multiple local minima and compare them against vanilla Differential Dynamic Programming. Furthermore, we discuss connections with previous work on the linearly solvable stochastic control framework and its extensions in relation to compositionality."
                    },
                    {
                        "title": "Deep Graphic FBSDEs for Opinion Dynamics Stochastic Control",
                        "abstract": "In this paper, we present a scalable deep learning approach to solve opinion dynamics stochastic optimal control problems with mean field term coupling in the dynamics and cost function. Our approach relies on the probabilistic representation of the solution of the Hamilton-Jacobi-Bellman partial differential equation. Grounded on the nonlinear version of the Feynman-Kac lemma, the solutions of the Hamilton-Jacobi-Bellman partial differential equation are linked to the solution of Forward-Backward Stochastic Differential Equations. These equations can be solved numerically using a novel deep neural network with architecture tailored to the problem in consideration. The resulting algorithm is tested on a polarized opinion consensus experiment. The large-scale (10K) agents experiment validates the scalability and generalizability of our algorithm. The proposed framework opens up the possibility for future applications on extremely large-scale problems."
                    },
                    {
                        "title": "GDN: A Stacking Network Used for Skin Cancer Diagnosis",
                        "abstract": "Skin cancer, the primary type of cancer that can be identified by visual recognition, requires an automatic identification system that can accurately classify different types of lesions. This paper presents GoogLe-Dense Network (GDN), which is an image-classification model to identify two types of skin cancer, Basal Cell Carcinoma, and Melanoma. GDN uses stacking of different networks to enhance the model performance. Specifically, GDN consists of two sequential levels in its structure. The first level performs basic classification tasks accomplished by GoogLeNet and DenseNet, which are trained in parallel to enhance efficiency. To avoid low accuracy and long training time, the second level takes the output of the GoogLeNet and DenseNet as the input for a logistic regression model. We compare our method with four baseline networks including ResNet, VGGNet, DenseNet, and GoogLeNet on the dataset, in which GoogLeNet and DenseNet significantly outperform ResNet and VGGNet. In the second level, different stacking methods such as perceptron, logistic regression, SVM, decision trees and K-neighbor are studied in which Logistic Regression shows the best prediction result among all. The results prove that GDN, compared to a single network structure, has higher accuracy in optimizing skin cancer detection."
                    },
                    {
                        "title": "From ChatGPT, DALL-E 3 to Sora: How has Generative AI Changed Digital Humanities Research and Services?",
                        "abstract": "Generative large-scale language models create the fifth paradigm of scientific research, organically combine data science and computational intelligence, transform the research paradigm of natural language processing and multimodal information processing, promote the new trend of AI-enabled social science research, and provide new ideas for digital humanities research and application. This article profoundly explores the application of large-scale language models in digital humanities research, revealing their significant potential in ancient book protection, intelligent processing, and academic innovation. The article first outlines the importance of ancient book resources and the necessity of digital preservation, followed by a detailed introduction to developing large-scale language models, such as ChatGPT, and their applications in document management, content understanding, and cross-cultural research. Through specific cases, the article demonstrates how AI can assist in the organization, classification, and content generation of ancient books. Then, it explores the prospects of AI applications in artistic innovation and cultural heritage preservation. Finally, the article explores the challenges and opportunities in the interaction of technology, information, and society in the digital humanities triggered by AI technologies."
                    }
                ]
            },
            "db8cf94d-efa4-47bd-9cb1-1cf4cdab8d55": {
                "pk": "db8cf94d-efa4-47bd-9cb1-1cf4cdab8d55",
                "name": "Zhuofan Xia",
                "collaborators": [
                    "Gao Huang",
                    "Shiji Song",
                    "Xuran Pan",
                    "Yizeng Han",
                    "Li Erran Li",
                    "Dongchen Han",
                    "Tianzhu Ye",
                    "Yifan Pu",
                    "Jiayun Li",
                    "Yuxiao Cheng"
                ],
                "domain": [
                    "Computer Vision",
                    "Transformers",
                    "Attention Mechanism",
                    "3D Object Detection"
                ],
                "publications": [
                    {
                        "title": "Slide-Transformer: Hierarchical Vision Transformer with Local Self-Attention",
                        "abstract": "Self-attention mechanism has been a key factor in the recent progress of Vision Transformer (ViT), which enables adaptive feature extraction from global contexts. However, existing self-attention methods either adopt sparse global attention or window attention to reduce the computation complexity, which may compromise the local feature learning or subject to some handcrafted designs. In contrast, local attention, which restricts the receptive field of each query to its own neighboring pixels, enjoys the benefits of both convolution and self-attention, namely local inductive bias and dynamic feature selection. Nevertheless, current local attention modules either use inefficient Im2Col function or rely on specific CUDA kernels that are hard to generalize to devices without CUDA support. In this paper, we propose a novel local attention module, Slide Attention, which leverages common convolution operations to achieve high efficiency, flexibility and generalizability. Specifically, we first re-interpret the column-based Im2Col function from a new row-based perspective and use Depthwise Convolution as an efficient substitution. On this basis, we propose a deformed shifting module based on the re-parameterization technique, which further relaxes the fixed key/value positions to deformed features in the local region. In this way, our module realizes the local attention paradigm in both efficient and flexible manner. Extensive experiments show that our slide attention module is applicable to a variety of advanced Vision Transformer models and compatible with various hardware devices, and achieves consistently improved performances on comprehensive benchmarks. Code is available at https://github.com/LeapLabTHU/Slide-Transformer."
                    },
                    {
                        "title": "Vision Transformer with Deformable Attention",
                        "abstract": "Transformers have recently shown superior performances on various vision tasks. The large, sometimes even global, receptive field endows Transformer models with higher representation power over their CNN counterparts. Nevertheless, simply enlarging receptive field also gives rise to several concerns. On the one hand, using dense attention e.g., in ViT, leads to excessive memory and computational cost, and features can be influenced by irrelevant parts which are beyond the region of interests. On the other hand, the sparse attention adopted in PVT or Swin Transformer is data agnostic and may limit the ability to model long range relations. To mitigate these issues, we propose a novel deformable self-attention module, where the positions of key and value pairs in self-attention are selected in a data-dependent way. This flexible scheme enables the self-attention module to focus on relevant regions and capture more informative features. On this basis, we present Deformable Attention Transformer, a general backbone model with deformable attention for both image classification and dense prediction tasks. Extensive experiments show that our models achieve consistently improved results on comprehensive benchmarks. Code is available at https://github.com/LeapLabTHU/DAT."
                    },
                    {
                        "title": "DAT++: Spatially Dynamic Vision Transformer with Deformable Attention",
                        "abstract": "Transformers have shown superior performance on various vision tasks. Their large receptive field endows Transformer models with higher representation power than their CNN counterparts. Nevertheless, simply enlarging the receptive field also raises several concerns. On the one hand, using dense attention in ViT leads to excessive memory and computational cost, and features can be influenced by irrelevant parts that are beyond the region of interests. On the other hand, the handcrafted attention adopted in PVT or Swin Transformer is data agnostic and may limit the ability to model long-range relations. To solve this dilemma, we propose a novel deformable multi-head attention module, where the positions of key and value pairs in self-attention are adaptively allocated in a data-dependent way. This flexible scheme enables the proposed deformable attention to dynamically focus on relevant regions while maintains the representation power of global attention. On this basis, we present Deformable Attention Transformer (DAT), a general vision backbone efficient and effective for visual recognition. We further build an enhanced version DAT++. Extensive experiments show that our DAT++ achieves state-of-the-art results on various visual recognition benchmarks, with 85.9% ImageNet accuracy, 54.5 and 47.0 MS-COCO instance segmentation mAP, and 51.5 ADE20K semantic segmentation mIoU."
                    },
                    {
                        "title": "GSVA: Generalized Segmentation via Multimodal Large Language Models",
                        "abstract": "Generalized Referring Expression Segmentation (GRES) extends the scope of classic RES to refer to multiple objects in one expression or identify the empty targets absent in the image. GRES poses challenges in modeling the complex spatial relationships of the instances in the image and identifying non-existing referents. Multimodal Large Language Models (MLLMs) have recently shown tremendous progress in these complicated vision-language tasks. Connecting Large Language Models (LLMs) and vision models, MLLMs are proficient in understanding contexts with visual inputs. Among them, LISA, as a representative, adopts a special [SEG] token to prompt a segmentation mask decoder, e.g., SAM, to enable MLLMs in the RES task. However, existing solutions to GRES remain unsatisfactory since current segmentation MLLMs cannot correctly handle the cases where users might reference multiple subjects in a singular prompt or provide descriptions incongruent with any image target. In this paper, we propose Generalized Segmentation Vision Assistant (GSVA) to address this gap. Specifically, GSVA reuses the [SEG] token to prompt the segmentation model towards supporting multiple mask references simultaneously and innovatively learns to generate a [REJ] token to reject the null targets explicitly. Experiments validate GSVA's efficacy in resolving the GRES issue, marking a notable enhancement and setting a new record on the GRES benchmark gRefCOCO dataset. GSVA also proves effective across various classic referring segmentation and comprehension tasks."
                    },
                    {
                        "title": "3D Object Detection with Pointformer",
                        "abstract": "Feature learning for 3D object detection from point clouds is very challenging due to the irregularity of 3D point cloud data. In this paper, we propose Pointformer, a Transformer backbone designed for 3D point clouds to learn features effectively. Specifically, a Local Transformer module is employed to model interactions among points in a local region, which learns context-dependent region features at an object level. A Global Transformer is designed to learn context-aware representations at the scene level. To further capture the dependencies among multi-scale representations, we propose Local-Global Transformer to integrate local features with global features from higher resolution. In addition, we introduce an efficient coordinate refinement module to shift down-sampled points closer to object centroids, which improves object proposal generation. We use Pointformer as the backbone for state-of-the-art object detection models and demonstrate significant improvements over original models on both indoor and outdoor datasets."
                    },
                    {
                        "title": "Generalized Activation via Multivariate Projection",
                        "abstract": "Activation functions are essential to introduce nonlinearity into neural networks, with the Rectified Linear Unit (ReLU) often favored for its simplicity and effectiveness. Motivated by the structural similarity between a shallow Feedforward Neural Network (FNN) and a single iteration of the Projected Gradient Descent (PGD) algorithm, a standard approach for solving constrained optimization problems, we consider ReLU as a projection from R onto the nonnegative half-line R+. Building on this interpretation, we extend ReLU by substituting it with a generalized projection operator onto a convex cone, such as the Second-Order Cone (SOC) projection, thereby naturally extending it to a Multivariate Projection Unit (MPU), an activation function with multiple inputs and multiple outputs. We further provide mathematical proof establishing that FNNs activated by SOC projections outperform those utilizing ReLU in terms of expressive power. Experimental evaluations on widely-adopted architectures further corroborate MPU's effectiveness against a broader range of existing activation functions."
                    },
                    {
                        "title": "Adaptive Rotated Convolution for Rotated Object Detection",
                        "abstract": "Rotated object detection aims to identify and locate objects in images with arbitrary orientation. In this scenario, the oriented directions of objects vary considerably across different images, while multiple orientations of objects exist within an image. This intrinsic characteristic makes it challenging for standard backbone networks to extract high-quality features of these arbitrarily orientated objects. In this paper, we present Adaptive Rotated Convolution (ARC) module to handle the aforementioned challenges. In our ARC module, the convolution kernels rotate adaptively to extract object features with varying orientations in different images, and an efficient conditional computation mechanism is introduced to accommodate the large orientation variations of objects within an image. The two designs work seamlessly in rotated object detection problem. Moreover, ARC can conveniently serve as a plug-and-play module in various vision backbones to boost their representation ability to detect oriented objects accurately. Experiments on commonly used benchmarks (DOTA and HRSC2016) demonstrate that equipped with our proposed ARC module in the backbone network, the performance of multiple popular oriented object detectors is significantly improved (\\eg +3.03\\% mAP on Rotated RetinaNet and +4.16\\% on CFA). Combined with the highly competitive method Oriented R-CNN, the proposed approach achieves state-of-the-art performance on the DOTA dataset with 81.77\\% mAP. Code is available at \\url{https://github.com/LeapLabTHU/ARC}."
                    },
                    {
                        "title": "Agent Attention: On the Integration of Softmax and Linear Attention",
                        "abstract": "The attention module is the key component in Transformers. While the global attention mechanism offers high expressiveness, its excessive computational cost restricts its applicability in various scenarios. In this paper, we propose a novel attention paradigm, Agent Attention, to strike a favorable balance between computational efficiency and representation power. Specifically, the Agent Attention, denoted as a quadruple $(Q, A, K, V)$, introduces an additional set of agent tokens $A$ into the conventional attention module. The agent tokens first act as the agent for the query tokens $Q$ to aggregate information from $K$ and $V$, and then broadcast the information back to $Q$. Given the number of agent tokens can be designed to be much smaller than the number of query tokens, the agent attention is significantly more efficient than the widely adopted Softmax attention, while preserving global context modelling capability. Interestingly, we show that the proposed agent attention is equivalent to a generalized form of linear attention. Therefore, agent attention seamlessly integrates the powerful Softmax attention and the highly efficient linear attention. Extensive experiments demonstrate the effectiveness of agent attention with various vision Transformers and across diverse vision tasks, including image classification, object detection, semantic segmentation and image generation. Notably, agent attention has shown remarkable performance in high-resolution scenarios, owning to its linear attention nature. For instance, when applied to Stable Diffusion, our agent attention accelerates generation and substantially enhances image generation quality without any additional training. Code is available at https://github.com/LeapLabTHU/Agent-Attention."
                    },
                    {
                        "title": "Efficient Diffusion Transformer with Step-wise Dynamic Attention Mediators",
                        "abstract": "This paper identifies significant redundancy in the query-key interactions within self-attention mechanisms of diffusion transformer models, particularly during the early stages of denoising diffusion steps. In response to this observation, we present a novel diffusion transformer framework incorporating an additional set of mediator tokens to engage with queries and keys separately. By modulating the number of mediator tokens during the denoising generation phases, our model initiates the denoising process with a precise, non-ambiguous stage and gradually transitions to a phase enriched with detail. Concurrently, integrating mediator tokens simplifies the attention module's complexity to a linear scale, enhancing the efficiency of global attention processes. Additionally, we propose a time-step dynamic mediator token adjustment mechanism that further decreases the required computational FLOPs for generation, simultaneously facilitating the generation of high-quality images within the constraints of varied inference budgets. Extensive experiments demonstrate that the proposed method can improve the generated image quality while also reducing the inference cost of diffusion transformers. When integrated with the recent work SiT, our method achieves a state-of-the-art FID score of 2.01. The source code is available at https://github.com/LeapLabTHU/Attention-Mediators."
                    }
                ]
            },
            "cdfa7ccd-d962-4e05-b049-0af8325415a9": {
                "pk": "cdfa7ccd-d962-4e05-b049-0af8325415a9",
                "name": "Yizeng Han",
                "collaborators": [
                    "Gao Huang",
                    "Shiji Song",
                    "Yulin Wang",
                    "Yifan Pu",
                    "Dongchen Han",
                    "Chaofei Wang",
                    "Xuran Pan",
                    "Le Yang",
                    "Chao Deng",
                    "Chaoqun Du"
                ],
                "domain": [
                    "Deep Learning",
                    "Computer Vision",
                    "Dynamic Networks",
                    "Efficient Inference"
                ],
                "publications": [
                    {
                        "title": "SimPro: A Simple Probabilistic Framework Towards Realistic Long-Tailed Semi-Supervised Learning",
                        "abstract": "Recent advancements in semi-supervised learning have focused on a more realistic yet challenging task: addressing imbalances in labeled data while the class distribution of unlabeled data remains both unknown and potentially mismatched. Current approaches in this sphere often presuppose rigid assumptions regarding the class distribution of unlabeled data, thereby limiting the adaptability of models to only certain distribution ranges. In this study, we propose a novel approach, introducing a highly adaptable framework, designated as SimPro, which does not rely on any predefined assumptions about the distribution of unlabeled data. Our framework, grounded in a probabilistic model, innovatively refines the expectation-maximization (EM) algorithm by explicitly decoupling the modeling of conditional and marginal class distributions. This separation facilitates a closed-form solution for class distribution estimation during the maximization phase, leading to the formulation of a Bayes classifier. The Bayes classifier, in turn, enhances the quality of pseudo-labels in the expectation phase. Remarkably, the SimPro framework not only comes with theoretical guarantees but also is straightforward to implement. Moreover, we introduce two novel class distributions broadening the scope of the evaluation. Our method showcases consistent state-of-the-art performance across diverse benchmarks and data distribution scenarios. Our code is available at https://github.com/LeapLabTHU/SimPro."
                    },
                    {
                        "title": "FLatten Transformer: Vision Transformer using Focused Linear Attention",
                        "abstract": "The quadratic computation complexity of self-attention has been a persistent challenge when applying Transformer models to vision tasks. Linear attention, on the other hand, offers a much more efficient alternative with its linear complexity by approximating the Softmax operation through carefully designed mapping functions. However, current linear attention approaches either suffer from significant performance degradation or introduce additional computation overhead from the mapping functions. In this paper, we propose a novel Focused Linear Attention module to achieve both high efficiency and expressiveness. Specifically, we first analyze the factors contributing to the performance degradation of linear attention from two perspectives: the focus ability and feature diversity. To overcome these limitations, we introduce a simple yet effective mapping function and an efficient rank restoration module to enhance the expressiveness of self-attention while maintaining low computation complexity. Extensive experiments show that our linear attention module is applicable to a variety of advanced vision Transformers, and achieves consistently improved performances on multiple benchmarks. Code is available at https://github.com/LeapLabTHU/FLatten-Transformer."
                    },
                    {
                        "title": "Dynamic Neural Networks: A Survey",
                        "abstract": "Dynamic neural network is an emerging research topic in deep learning. Compared to static models which have fixed computational graphs and parameters at the inference stage, dynamic networks can adapt their structures or parameters to different inputs, leading to notable advantages in terms of accuracy, computational efficiency, adaptiveness, etc. In this survey, we comprehensively review this rapidly developing area by dividing dynamic networks into three main categories: 1) instance-wise dynamic models that process each instance with data-dependent architectures or parameters; 2) spatial-wise dynamic networks that conduct adaptive computation with respect to different spatial locations of image data and 3) temporal-wise dynamic models that perform adaptive inference along the temporal dimension for sequential data such as videos and texts. The important research problems of dynamic networks, e.g., architecture design, decision making scheme, optimization technique and applications, are reviewed systematically. Finally, we discuss the open problems in this field together with interesting future research directions."
                    },
                    {
                        "title": "Latency-aware Spatial-wise Dynamic Networks",
                        "abstract": "Spatial-wise dynamic convolution has become a promising approach to improving the inference efficiency of deep networks. By allocating more computation to the most informative pixels, such an adaptive inference paradigm reduces the spatial redundancy in image features and saves a considerable amount of unnecessary computation. However, the theoretical efficiency achieved by previous methods can hardly translate into a realistic speedup, especially on the multi-core processors (e.g. GPUs). The key challenge is that the existing literature has only focused on designing algorithms with minimal computation, ignoring the fact that the practical latency can also be influenced by scheduling strategies and hardware properties. To bridge the gap between theoretical computation and practical efficiency, we propose a latency-aware spatial-wise dynamic network (LASNet), which performs coarse-grained spatially adaptive inference under the guidance of a novel latency prediction model. The latency prediction model can efficiently estimate the inference latency of dynamic networks by simultaneously considering algorithms, scheduling strategies, and hardware properties. We use the latency predictor to guide both the algorithm design and the scheduling optimization on various hardware platforms. Experiments on image classification, object detection and instance segmentation demonstrate that the proposed framework significantly improves the practical inference efficiency of deep networks. For example, the average latency of a ResNet-101 on the ImageNet validation set could be reduced by 36% and 46% on a server GPU (Nvidia Tesla-V100) and an edge device (Nvidia Jetson TX2 GPU) respectively without sacrificing the accuracy. Code is available at https://github.com/LeapLabTHU/LASNet."
                    },
                    {
                        "title": "Latency-aware Unified Dynamic Networks for Efficient Image Recognition",
                        "abstract": "Dynamic computation has emerged as a promising avenue to enhance the inference efficiency of deep networks. It allows selective activation of computational units, leading to a reduction in unnecessary computations for each input sample. However, the actual efficiency of these dynamic models can deviate from theoretical predictions. This mismatch arises from: 1) the lack of a unified approach due to fragmented research; 2) the focus on algorithm design over critical scheduling strategies, especially in CUDA-enabled GPU contexts; and 3) challenges in measuring practical latency, given that most libraries cater to static operations. Addressing these issues, we unveil the Latency-Aware Unified Dynamic Networks (LAUDNet), a framework that integrates three primary dynamic paradigms-spatially adaptive computation, dynamic layer skipping, and dynamic channel skipping. To bridge the theoretical and practical efficiency gap, LAUDNet merges algorithmic design with scheduling optimization, guided by a latency predictor that accurately gauges dynamic operator latency. We've tested LAUDNet across multiple vision tasks, demonstrating its capacity to notably reduce the latency of models like ResNet-101 by over 50% on platforms such as V100, RTX3090, and TX2 GPUs. Notably, LAUDNet stands out in balancing accuracy and efficiency. Code is available at: https://www.github.com/LeapLabTHU/LAUDNet."
                    },
                    {
                        "title": "Fine-grained Recognition with Learnable Semantic Data Augmentation",
                        "abstract": "Fine-grained image recognition is a longstanding computer vision challenge that focuses on differentiating objects belonging to multiple subordinate categories within the same meta-category. Since images belonging to the same meta-category usually share similar visual appearances, mining discriminative visual cues is the key to distinguishing fine-grained categories. Although commonly used image-level data augmentation techniques have achieved great success in generic image classification problems, they are rarely applied in fine-grained scenarios, because their random editing-region behavior is prone to destroy the discriminative visual cues residing in the subtle regions. In this paper, we propose diversifying the training data at the feature-level to alleviate the discriminative region loss problem. Specifically, we produce diversified augmented samples by translating image features along semantically meaningful directions. The semantic directions are estimated with a covariance prediction network, which predicts a sample-wise covariance matrix to adapt to the large intra-class variation inherent in fine-grained images. Furthermore, the covariance prediction network is jointly optimized with the classification network in a meta-learning manner to alleviate the degenerate solution problem. Experiments on four competitive fine-grained recognition benchmarks (CUB-200-2011, Stanford Cars, FGVC Aircrafts, NABirds) demonstrate that our method significantly improves the generalization performance on several popular classification networks (e.g., ResNets, DenseNets, EfficientNets, RegNets and ViT). Combined with a recently proposed method, our semantic data augmentation approach achieves state-of-the-art performance on the CUB-200-2011 dataset. The source code will be released."
                    },
                    {
                        "title": "GSVA: Generalized Segmentation via Multimodal Large Language Models",
                        "abstract": "Generalized Referring Expression Segmentation (GRES) extends the scope of classic RES to refer to multiple objects in one expression or identify the empty targets absent in the image. GRES poses challenges in modeling the complex spatial relationships of the instances in the image and identifying non-existing referents. Multimodal Large Language Models (MLLMs) have recently shown tremendous progress in these complicated vision-language tasks. Connecting Large Language Models (LLMs) and vision models, MLLMs are proficient in understanding contexts with visual inputs. Among them, LISA, as a representative, adopts a special [SEG] token to prompt a segmentation mask decoder, e.g., SAM, to enable MLLMs in the RES task. However, existing solutions to GRES remain unsatisfactory since current segmentation MLLMs cannot correctly handle the cases where users might reference multiple subjects in a singular prompt or provide descriptions incongruent with any image target. In this paper, we propose Generalized Segmentation Vision Assistant (GSVA) to address this gap. Specifically, GSVA reuses the [SEG] token to prompt the segmentation model towards supporting multiple mask references simultaneously and innovatively learns to generate a [REJ] token to reject the null targets explicitly. Experiments validate GSVA's efficacy in resolving the GRES issue, marking a notable enhancement and setting a new record on the GRES benchmark gRefCOCO dataset. GSVA also proves effective across various classic referring segmentation and comprehension tasks."
                    },
                    {
                        "title": "OStr-DARTS: Differentiable Neural Architecture Search based on Operation Strength",
                        "abstract": "Differentiable architecture search (DARTS) has emerged as a promising technique for effective neural architecture search, and it mainly contains two steps to find the high-performance architecture: First, the DARTS supernet that consists of mixed operations will be optimized via gradient descent. Second, the final architecture will be built by the selected operations that contribute the most to the supernet. Although DARTS improves the efficiency of NAS, it suffers from the well-known degeneration issue which can lead to deteriorating architectures. Existing works mainly attribute the degeneration issue to the failure of its supernet optimization, while little attention has been paid to the selection method. In this paper, we cease to apply the widely-used magnitude-based selection method and propose a novel criterion based on operation strength that estimates the importance of an operation by its effect on the final loss. We show that the degeneration issue can be effectively addressed by using the proposed criterion without any modification of supernet optimization, indicating that the magnitude-based selection method can be a critical reason for the instability of DARTS. The experiments on NAS-Bench-201 and DARTS search spaces show the effectiveness of our method."
                    },
                    {
                        "title": "Resolution Adaptive Networks for Efficient Inference",
                        "abstract": "Adaptive inference is an effective mechanism to achieve a dynamic tradeoff between accuracy and computational cost in deep networks. Existing works mainly exploit architecture redundancy in network depth or width. In this paper, we focus on spatial redundancy of input samples and propose a novel Resolution Adaptive Network (RANet), which is inspired by the intuition that low-resolution representations are sufficient for classifying \"easy\" inputs containing large objects with prototypical features, while only some \"hard\" samples need spatially detailed information. In RANet, the input images are first routed to a lightweight sub-network that efficiently extracts low-resolution representations, and those samples with high prediction confidence will exit early from the network without being further processed. Meanwhile, high-resolution paths in the network maintain the capability to recognize the \"hard\" samples. Therefore, RANet can effectively reduce the spatial redundancy involved in inferring high-resolution inputs. Empirically, we demonstrate the effectiveness of the proposed RANet on the CIFAR-10, CIFAR-100 and ImageNet datasets in both the anytime prediction setting and the budgeted batch classification setting."
                    },
                    {
                        "title": "Adaptive Focus for Efficient Video Recognition",
                        "abstract": "In this paper, we explore the spatial redundancy in video recognition with the aim to improve the computational efficiency. It is observed that the most informative region in each frame of a video is usually a small image patch, which shifts smoothly across frames. Therefore, we model the patch localization problem as a sequential decision task, and propose a reinforcement learning based approach for efficient spatially adaptive video recognition (AdaFocus). In specific, a light-weighted ConvNet is first adopted to quickly process the full video sequence, whose features are used by a recurrent policy network to localize the most task-relevant regions. Then the selected patches are inferred by a high-capacity network for the final prediction. During offline inference, once the informative patch sequence has been generated, the bulk of computation can be done in parallel, and is efficient on modern GPU devices. In addition, we demonstrate that the proposed method can be easily extended by further considering the temporal redundancy, e.g., dynamically skipping less valuable frames. Extensive experiments on five benchmark datasets, i.e., ActivityNet, FCVID, Mini-Kinetics, Something-Something V1&V2, demonstrate that our method is significantly more efficient than the competitive baselines. Code is available at https://github.com/blackfeather-wang/AdaFocus."
                    },
                    {
                        "title": "CAM-loss: Towards Learning Spatially Discriminative Feature Representations",
                        "abstract": "The backbone of traditional CNN classifier is generally considered as a feature extractor, followed by a linear layer which performs the classification. We propose a novel loss function, termed as CAM-loss, to constrain the embedded feature maps with the class activation maps (CAMs) which indicate the spatially discriminative regions of an image for particular categories. CAM-loss drives the backbone to express the features of target category and suppress the features of non-target categories or background, so as to obtain more discriminative feature representations. It can be simply applied in any CNN architecture with neglectable additional parameters and calculations. Experimental results show that CAM-loss is applicable to a variety of network structures and can be combined with mainstream regularization methods to improve the performance of image classification. The strong generalization ability of CAM-loss is validated in the transfer learning and few shot learning tasks. Based on CAM-loss, we also propose a novel CAAM-CAM matching knowledge distillation method. This method directly uses the CAM generated by the teacher network to supervise the CAAM generated by the student network, which effectively improves the accuracy and convergence rate of the student network."
                    },
                    {
                        "title": "Computation-efficient Deep Learning for Computer Vision: A Survey",
                        "abstract": "Over the past decade, deep learning models have exhibited considerable advancements, reaching or even exceeding human-level performance in a range of visual perception tasks. This remarkable progress has sparked interest in applying deep networks to real-world applications, such as autonomous vehicles, mobile devices, robotics, and edge computing. However, the challenge remains that state-of-the-art models usually demand significant computational resources, leading to impractical power consumption, latency, or carbon emissions in real-world scenarios. This trade-off between effectiveness and efficiency has catalyzed the emergence of a new research focus: computationally efficient deep learning, which strives to achieve satisfactory performance while minimizing the computational cost during inference. This review offers an extensive analysis of this rapidly evolving field by examining four key areas: 1) the development of static or dynamic light-weighted backbone models for the efficient extraction of discriminative deep representations; 2) the specialized network architectures or algorithms tailored for specific computer vision tasks; 3) the techniques employed for compressing deep learning models; and 4) the strategies for deploying efficient deep networks on hardware platforms. Additionally, we provide a systematic discussion on the critical challenges faced in this domain, such as network architecture design, training schemes, practical efficiency, and more realistic model compression approaches, as well as potential future research directions."
                    },
                    {
                        "title": "Mask Grounding for Referring Image Segmentation",
                        "abstract": "Referring Image Segmentation (RIS) is a challenging task that requires an algorithm to segment objects referred by free-form language expressions. Despite significant progress in recent years, most state-of-the-art (SOTA) methods still suffer from considerable language-image modality gap at the pixel and word level. These methods generally 1) rely on sentence-level language features for language-image alignment and 2) lack explicit training supervision for fine-grained visual grounding. Consequently, they exhibit weak object-level correspondence between visual and language features. Without well-grounded features, prior methods struggle to understand complex expressions that require strong reasoning over relationships among multiple objects, especially when dealing with rarely used or ambiguous clauses. To tackle this challenge, we introduce a novel Mask Grounding auxiliary task that significantly improves visual grounding within language features, by explicitly teaching the model to learn fine-grained correspondence between masked textual tokens and their matching visual objects. Mask Grounding can be directly used on prior RIS methods and consistently bring improvements. Furthermore, to holistically address the modality gap, we also design a cross-modal alignment loss and an accompanying alignment module. These additions work synergistically with Mask Grounding. With all these techniques, our comprehensive approach culminates in MagNet (Mask-grounded Network), an architecture that significantly outperforms prior arts on three key benchmarks (RefCOCO, RefCOCO+ and G-Ref), demonstrating our method's effectiveness in addressing current limitations of RIS algorithms. Our code and pre-trained weights will be released."
                    },
                    {
                        "title": "UniTTA: Unified Benchmark and Versatile Framework Towards Realistic Test-Time Adaptation",
                        "abstract": "Test-Time Adaptation (TTA) aims to adapt pre-trained models to the target domain during testing. In reality, this adaptability can be influenced by multiple factors. Researchers have identified various challenging scenarios and developed diverse methods to address these challenges, such as dealing with continual domain shifts, mixed domains, and temporally correlated or imbalanced class distributions. Despite these efforts, a unified and comprehensive benchmark has yet to be established. To this end, we propose a Unified Test-Time Adaptation (UniTTA) benchmark, which is comprehensive and widely applicable. Each scenario within the benchmark is fully described by a Markov state transition matrix for sampling from the original dataset. The UniTTA benchmark considers both domain and class as two independent dimensions of data and addresses various combinations of imbalance/balance and i.i.d./non-i.i.d./continual conditions, covering a total of \\( (2 \\times 3)^2 = 36 \\) scenarios. It establishes a comprehensive evaluation benchmark for realistic TTA and provides a guideline for practitioners to select the most suitable TTA method. Alongside this benchmark, we propose a versatile UniTTA framework, which includes a Balanced Domain Normalization (BDN) layer and a COrrelated Feature Adaptation (COFA) method--designed to mitigate distribution gaps in domain and class, respectively. Extensive experiments demonstrate that our UniTTA framework excels within the UniTTA benchmark and achieves state-of-the-art performance on average. Our code is available at \\url{https://github.com/LeapLabTHU/UniTTA}."
                    },
                    {
                        "title": "Semantic Refocused Tuning for Open-Vocabulary Panoptic Segmentation",
                        "abstract": "Open-vocabulary panoptic segmentation is an emerging task aiming to accurately segment the image into semantically meaningful masks based on a set of texts. Despite existing efforts, it remains challenging to develop a high-performing method that generalizes effectively across new domains and requires minimal training resources. Our in-depth analysis of current methods reveals a crucial insight: mask classification is the main performance bottleneck for open-vocab. panoptic segmentation. Based on this, we propose Semantic Refocused Tuning (SMART), a novel framework that greatly enhances open-vocab. panoptic segmentation by improving mask classification through two key innovations. First, SMART adopts a multimodal Semantic-guided Mask Attention mechanism that injects task-awareness into the regional information extraction process. This enables the model to capture task-specific and contextually relevant information for more effective mask classification. Second, it incorporates Query Projection Tuning, which strategically fine-tunes the query projection layers within the Vision Language Model (VLM) used for mask classification. This adjustment allows the model to adapt the image focus of mask tokens to new distributions with minimal training resources, while preserving the VLM's pre-trained knowledge. Extensive ablation studies confirm the superiority of our approach. Notably, SMART sets new state-of-the-art results, demonstrating improvements of up to +1.3 PQ and +5.4 mIoU across representative benchmarks, while reducing training costs by nearly 10x compared to the previous best method. Our code and data will be released."
                    },
                    {
                        "title": "EfficientTrain++: Generalized Curriculum Learning for Efficient Visual Backbone Training",
                        "abstract": "The superior performance of modern visual backbones usually comes with a costly training procedure. We contribute to this issue by generalizing the idea of curriculum learning beyond its original formulation, i.e., training models using easier-to-harder data. Specifically, we reformulate the training curriculum as a soft-selection function, which uncovers progressively more difficult patterns within each example during training, instead of performing easier-to-harder sample selection. Our work is inspired by an intriguing observation on the learning dynamics of visual backbones: during the earlier stages of training, the model predominantly learns to recognize some 'easier-to-learn' discriminative patterns in the data. These patterns, when observed through frequency and spatial domains, incorporate lower-frequency components, and the natural image contents without distortion or data augmentation. Motivated by these findings, we propose a curriculum where the model always leverages all the training data at every learning stage, yet the exposure to the 'easier-to-learn' patterns of each example is initiated first, with harder patterns gradually introduced as training progresses. To implement this idea in a computationally efficient way, we introduce a cropping operation in the Fourier spectrum of the inputs, enabling the model to learn from only the lower-frequency components. Then we show that exposing the contents of natural images can be readily achieved by modulating the intensity of data augmentation. Finally, we integrate these aspects and design curriculum schedules with tailored search algorithms. The resulting method, EfficientTrain++, is simple, general, yet surprisingly effective. It reduces the training time of a wide variety of popular models by 1.5-3.0x on ImageNet-1K/22K without sacrificing accuracy. It also demonstrates efficacy in self-supervised learning (e.g., MAE)."
                    },
                    {
                        "title": "Dynamic Perceiver for Efficient Visual Recognition",
                        "abstract": "Early exiting has become a promising approach to improving the inference efficiency of deep networks. By structuring models with multiple classifiers (exits), predictions for ``easy'' samples can be generated at earlier exits, negating the need for executing deeper layers. Current multi-exit networks typically implement linear classifiers at intermediate layers, compelling low-level features to encapsulate high-level semantics. This sub-optimal design invariably undermines the performance of later exits. In this paper, we propose Dynamic Perceiver (Dyn-Perceiver) to decouple the feature extraction procedure and the early classification task with a novel dual-branch architecture. A feature branch serves to extract image features, while a classification branch processes a latent code assigned for classification tasks. Bi-directional cross-attention layers are established to progressively fuse the information of both branches. Early exits are placed exclusively within the classification branch, thus eliminating the need for linear separability in low-level features. Dyn-Perceiver constitutes a versatile and adaptable framework that can be built upon various architectures. Experiments on image classification, action recognition, and object detection demonstrate that our method significantly improves the inference efficiency of different backbones, outperforming numerous competitive approaches across a broad range of computational budgets. Evaluation on both CPU and GPU platforms substantiate the superior practical efficiency of Dyn-Perceiver. Code is available at https://www.github.com/LeapLabTHU/Dynamic_Perceiver."
                    },
                    {
                        "title": "Agent Attention: On the Integration of Softmax and Linear Attention",
                        "abstract": "The attention module is the key component in Transformers. While the global attention mechanism offers high expressiveness, its excessive computational cost restricts its applicability in various scenarios. In this paper, we propose a novel attention paradigm, Agent Attention, to strike a favorable balance between computational efficiency and representation power. Specifically, the Agent Attention, denoted as a quadruple $(Q, A, K, V)$, introduces an additional set of agent tokens $A$ into the conventional attention module. The agent tokens first act as the agent for the query tokens $Q$ to aggregate information from $K$ and $V$, and then broadcast the information back to $Q$. Given the number of agent tokens can be designed to be much smaller than the number of query tokens, the agent attention is significantly more efficient than the widely adopted Softmax attention, while preserving global context modelling capability. Interestingly, we show that the proposed agent attention is equivalent to a generalized form of linear attention. Therefore, agent attention seamlessly integrates the powerful Softmax attention and the highly efficient linear attention. Extensive experiments demonstrate the effectiveness of agent attention with various vision Transformers and across diverse vision tasks, including image classification, object detection, semantic segmentation and image generation. Notably, agent attention has shown remarkable performance in high-resolution scenarios, owning to its linear attention nature. For instance, when applied to Stable Diffusion, our agent attention accelerates generation and substantially enhances image generation quality without any additional training. Code is available at https://github.com/LeapLabTHU/Agent-Attention."
                    },
                    {
                        "title": "Learning to Weight Samples for Dynamic Early-exiting Networks",
                        "abstract": "Early exiting is an effective paradigm for improving the inference efficiency of deep networks. By constructing classifiers with varying resource demands (the exits), such networks allow easy samples to be output at early exits, removing the need for executing deeper layers. While existing works mainly focus on the architectural design of multi-exit networks, the training strategies for such models are largely left unexplored. The current state-of-the-art models treat all samples the same during training. However, the early-exiting behavior during testing has been ignored, leading to a gap between training and testing. In this paper, we propose to bridge this gap by sample weighting. Intuitively, easy samples, which generally exit early in the network during inference, should contribute more to training early classifiers. The training of hard samples (mostly exit from deeper layers), however, should be emphasized by the late classifiers. Our work proposes to adopt a weight prediction network to weight the loss of different training samples at each exit. This weight prediction network and the backbone model are jointly optimized under a meta-learning framework with a novel optimization objective. By bringing the adaptive behavior during inference into the training phase, we show that the proposed weighting mechanism consistently improves the trade-off between classification accuracy and inference efficiency. Code is available at https://github.com/LeapLabTHU/L2W-DEN."
                    }
                ]
            },
            "e7e0daf4-cce9-434d-b7fe-97fc22bed42b": {
                "pk": "e7e0daf4-cce9-434d-b7fe-97fc22bed42b",
                "name": "Yifan Pu",
                "collaborators": [
                    "Botao Ren",
                    "Botian Xu",
                    "Jingyi Wang",
                    "Zhidong Deng"
                ],
                "domain": [
                    "Computer Vision",
                    "Object Detection",
                    "Transformer Models"
                ],
                "publications": [
                    {
                        "title": "Improving Detection in Aerial Images by Capturing Inter-Object Relationships",
                        "abstract": "In many image domains, the spatial distribution of objects in a scene exhibits meaningful patterns governed by their semantic relationships. In most modern detection pipelines, however, the detection proposals are processed independently, overlooking the underlying relationships between objects. In this work, we introduce a transformer-based approach to capture these inter-object relationships to refine classification and regression outcomes for detected objects. Building on two-stage detectors, we tokenize the region of interest (RoI) proposals to be processed by a transformer encoder. Specific spatial and geometric relations are incorporated into the attention weights and adaptively modulated and regularized. Experimental results demonstrate that the proposed method achieves consistent performance improvement on three benchmarks including DOTA-v1.0, DOTA-v1.5, and HRSC 2016, especially ranking first on both DOTA-v1.5 and HRSC 2016. Specifically, our new method has an increase of 1.59 mAP on DOTA-v1.0, 4.88 mAP on DOTA-v1.5, and 2.1 mAP on HRSC 2016, respectively, compared to the baselines."
                    }
                ]
            },
            "6e0aea81-7519-4f56-b47e-4fb6d62461bf": {
                "pk": "6e0aea81-7519-4f56-b47e-4fb6d62461bf",
                "name": "Chunjiang Ge",
                "collaborators": [
                    "Gao Huang",
                    "Shiji Song",
                    "Zanlin Ni",
                    "Rui Huang",
                    "Xuran Pan",
                    "Rui Lu",
                    "Guanfu Chen",
                    "Zeyi Huang",
                    "Mixue Xie",
                    "Zihang Lai"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Domain Adaptation",
                    "Representation Learning",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "On the Integration of Self-Attention and Convolution",
                        "abstract": "Convolution and self-attention are two powerful techniques for representation learning, and they are usually considered as two peer approaches that are distinct from each other. In this paper, we show that there exists a strong underlying relation between them, in the sense that the bulk of computations of these two paradigms are in fact done with the same operation. Specifically, we first show that a traditional convolution with kernel size k x k can be decomposed into k^2 individual 1x1 convolutions, followed by shift and summation operations. Then, we interpret the projections of queries, keys, and values in self-attention module as multiple 1x1 convolutions, followed by the computation of attention weights and aggregation of the values. Therefore, the first stage of both two modules comprises the similar operation. More importantly, the first stage contributes a dominant computation complexity (square of the channel size) comparing to the second stage. This observation naturally leads to an elegant integration of these two seemingly distinct paradigms, i.e., a mixed model that enjoys the benefit of both self-Attention and Convolution (ACmix), while having minimum computational overhead compared to the pure convolution or self-attention counterpart. Extensive experiments show that our model achieves consistently improved results over competitive baselines on image recognition and downstream tasks. Code and pre-trained models will be released at https://github.com/LeapLabTHU/ACmix and https://gitee.com/mindspore/models."
                    },
                    {
                        "title": "Domain Adaptation via Prompt Learning",
                        "abstract": "Unsupervised domain adaption (UDA) aims to adapt models learned from a well-annotated source domain to a target domain, where only unlabeled samples are given. Current UDA approaches learn domain-invariant features by aligning source and target feature spaces. Such alignments are imposed by constraints such as statistical discrepancy minimization or adversarial training. However, these constraints could lead to the distortion of semantic feature structures and loss of class discriminability. In this paper, we introduce a novel prompt learning paradigm for UDA, named Domain Adaptation via Prompt Learning (DAPL). In contrast to prior works, our approach makes use of pre-trained vision-language models and optimizes only very few parameters. The main idea is to embed domain information into prompts, a form of representations generated from natural language, which is then used to perform classification. This domain information is shared only by images from the same domain, thereby dynamically adapting the classifier according to each domain. By adopting this paradigm, we show that our model not only outperforms previous methods on several cross-domain benchmarks but also is very efficient to train and easy to implement."
                    },
                    {
                        "title": "ConvLLaVA: Hierarchical Backbones as Visual Encoder for Large Multimodal Models",
                        "abstract": "High-resolution Large Multimodal Models (LMMs) encounter the challenges of excessive visual tokens and quadratic visual complexity. Current high-resolution LMMs address the quadratic complexity while still generating excessive visual tokens. However, the redundancy in visual tokens is the key problem as it leads to more substantial compute. To mitigate this issue, we propose ConvLLaVA, which employs ConvNeXt, a hierarchical backbone, as the visual encoder of LMM to replace Vision Transformer (ViT). ConvLLaVA compresses high-resolution images into information-rich visual features, effectively preventing the generation of excessive visual tokens. To enhance the capabilities of ConvLLaVA, we propose two critical optimizations. Since the low-resolution pretrained ConvNeXt underperforms when directly applied on high resolution, we update it to bridge the gap. Moreover, since ConvNeXt's original compression ratio is inadequate for much higher resolution inputs, we train a successive stage to further compress the visual tokens, thereby reducing redundancy. These optimizations enable ConvLLaVA to support inputs of 1536x1536 resolution generating only 576 visual tokens, capable of handling images of arbitrary aspect ratios. Experimental results demonstrate that our method achieves competitive performance with state-of-the-art models on mainstream benchmarks. The ConvLLaVA model series are publicly available at https://github.com/alibaba/conv-llava."
                    },
                    {
                        "title": "Cross-Modal Adapter for Text-Video Retrieval",
                        "abstract": "Text-video retrieval is an important multi-modal learning task, where the goal is to retrieve the most relevant video for a given text query. Recently, pre-trained models, e.g., CLIP, show great potential on this task. However, as pre-trained models are scaling up, fully fine-tuning them on text-video retrieval datasets has a high risk of overfitting. Moreover, in practice, it would be costly to train and store a large model for each task. To overcome the above issues, we present a novel $\\textbf{Cross-Modal Adapter}$ for parameter-efficient fine-tuning. Inspired by adapter-based methods, we adjust the pre-trained model with a few parameterization layers. However, there are two notable differences. First, our method is designed for the multi-modal domain. Secondly, it allows early cross-modal interactions between CLIP's two encoders. Although surprisingly simple, our approach has three notable benefits: (1) reduces $\\textbf{99.6}\\%$ of fine-tuned parameters, and alleviates the problem of overfitting, (2) saves approximately 30% of training time, and (3) allows all the pre-trained parameters to be fixed, enabling the pre-trained model to be shared across datasets. Extensive experiments demonstrate that, without bells and whistles, it achieves superior or comparable performance compared to fully fine-tuned methods on MSR-VTT, MSVD, VATEX, ActivityNet, and DiDeMo datasets. The code will be available at \\url{https://github.com/LeapLabTHU/Cross-Modal-Adapter}."
                    },
                    {
                        "title": "Using Human Feedback to Fine-tune Diffusion Models without Any Reward Model",
                        "abstract": "Using reinforcement learning with human feedback (RLHF) has shown significant promise in fine-tuning diffusion models. Previous methods start by training a reward model that aligns with human preferences, then leverage RL techniques to fine-tune the underlying models. However, crafting an efficient reward model demands extensive datasets, optimal architecture, and manual hyperparameter tuning, making the process both time and cost-intensive. The direct preference optimization (DPO) method, effective in fine-tuning large language models, eliminates the necessity for a reward model. However, the extensive GPU memory requirement of the diffusion model's denoising process hinders the direct application of the DPO method. To address this issue, we introduce the Direct Preference for Denoising Diffusion Policy Optimization (D3PO) method to directly fine-tune diffusion models. The theoretical analysis demonstrates that although D3PO omits training a reward model, it effectively functions as the optimal reward model trained using human feedback data to guide the learning process. This approach requires no training of a reward model, proving to be more direct, cost-effective, and minimizing computational overhead. In experiments, our method uses the relative scale of objectives as a proxy for human preference, delivering comparable results to methods using ground-truth rewards. Moreover, D3PO demonstrates the ability to reduce image distortion rates and generate safer images, overcoming challenges lacking robust reward models. Our code is publicly available at https://github.com/yk7333/D3PO."
                    },
                    {
                        "title": "Everything to the Synthetic: Diffusion-driven Test-time Adaptation via Synthetic-Domain Alignment",
                        "abstract": "Test-time adaptation (TTA) aims to enhance the performance of source-domain pretrained models when tested on unknown shifted target domains. Traditional TTA methods primarily adapt model weights based on target data streams, making model performance sensitive to the amount and order of target data. Recently, diffusion-driven TTA methods have demonstrated strong performance by using an unconditional diffusion model, which is also trained on the source domain to transform target data into synthetic data as a source domain projection. This allows the source model to make predictions without weight adaptation. In this paper, we argue that the domains of the source model and the synthetic data in diffusion-driven TTA methods are not aligned. To adapt the source model to the synthetic domain of the unconditional diffusion model, we introduce a Synthetic-Domain Alignment (SDA) framework to fine-tune the source model with synthetic data. Specifically, we first employ a conditional diffusion model to generate labeled samples, creating a synthetic dataset. Subsequently, we use the aforementioned unconditional diffusion model to add noise to and denoise each sample before fine-tuning. This process mitigates the potential domain gap between the conditional and unconditional models. Extensive experiments across various models and benchmarks demonstrate that SDA achieves superior domain alignment and consistently outperforms existing diffusion-driven TTA methods. Our code is available at https://github.com/SHI-Labs/Diffusion-Driven-Test-Time-Adaptation-via-Synthetic-Domain-Alignment."
                    },
                    {
                        "title": "LLaVA-UHD: an LMM Perceiving Any Aspect Ratio and High-Resolution Images",
                        "abstract": "Visual encoding constitutes the basis of large multimodal models (LMMs) in understanding the visual world. Conventional LMMs process images in fixed sizes and limited resolutions, while recent explorations in this direction are limited in adaptivity, efficiency, and even correctness. In this work, we first take GPT-4V and LLaVA-1.5 as representative examples and expose systematic flaws rooted in their visual encoding strategy. To address the challenges, we present LLaVA-UHD, a large multimodal model that can efficiently perceive images in any aspect ratio and high resolution. LLaVA-UHD includes three key components: (1) An image modularization strategy that divides native-resolution images into smaller variable-sized slices for efficient and extensible encoding, (2) a compression module that further condenses image tokens from visual encoders, and (3) a spatial schema to organize slice tokens for LLMs. Comprehensive experiments show that LLaVA-UHD outperforms established LMMs trained with 2-3 orders of magnitude more data on 9 benchmarks. Notably, our model built on LLaVA-1.5 336x336 supports 6 times larger (i.e., 672x1088) resolution images using only 94% inference computation, and achieves 6.4 accuracy improvement on TextVQA. Moreover, the model can be efficiently trained in academic settings, within 23 hours on 8 A100 GPUs (vs. 26 hours of LLaVA-1.5). We make the data and code publicly available at https://github.com/thunlp/LLaVA-UHD."
                    }
                ]
            },
            "1762560a-933e-48c3-a1f8-34f82779cc57": {
                "pk": "1762560a-933e-48c3-a1f8-34f82779cc57",
                "name": "Jun Song",
                "collaborators": [
                    "Shuqi He",
                    "Jun Zhuang",
                    "Ding Wang",
                    "Chaoyue Zhao",
                    "Bin Ouyang",
                    "Ke Han",
                    "Ali Mahzarnia",
                    "Luyao Peng",
                    "Baihua Ren"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Reinforcement Learning",
                    "Machine Learning",
                    "Additive Manufacturing"
                ],
                "publications": [
                    {
                        "title": "RW-NSGCN: A Robust Approach to Structural Attacks via Negative Sampling",
                        "abstract": "Node classification using Graph Neural Networks (GNNs) has been widely applied in various practical scenarios, such as predicting user interests and detecting communities in social networks. However, recent studies have shown that graph-structured networks often contain potential noise and attacks, in the form of topological perturbations and weight disturbances, which can lead to decreased classification performance in GNNs. To improve the robustness of the model, we propose a novel method: Random Walk Negative Sampling Graph Convolutional Network (RW-NSGCN). Specifically, RW-NSGCN integrates the Random Walk with Restart (RWR) and PageRank (PGR) algorithms for negative sampling and employs a Determinantal Point Process (DPP)-based GCN for convolution operations. RWR leverages both global and local information to manage noise and local variations, while PGR assesses node importance to stabilize the topological structure. The DPP-based GCN ensures diversity among negative samples and aggregates their features to produce robust node embeddings, thereby improving classification performance. Experimental results demonstrate that the RW-NSGCN model effectively addresses network topology attacks and weight instability, increasing the accuracy of anomaly detection and overall stability. In terms of classification accuracy, RW-NSGCN significantly outperforms existing methods, showing greater resilience across various scenarios and effectively mitigating the impact of such vulnerabilities."
                    },
                    {
                        "title": "Optimistic Distributionally Robust Policy Optimization",
                        "abstract": "Trust Region Policy Optimization (TRPO) and Proximal Policy Optimization (PPO), as the widely employed policy based reinforcement learning (RL) methods, are prone to converge to a sub-optimal solution as they limit the policy representation to a particular parametric distribution class. To address this issue, we develop an innovative Optimistic Distributionally Robust Policy Optimization (ODRPO) algorithm, which effectively utilizes Optimistic Distributionally Robust Optimization (DRO) approach to solve the trust region constrained optimization problem without parameterizing the policies. Our algorithm improves TRPO and PPO with a higher sample efficiency and a better performance of the final policy while attaining the learning stability. Moreover, it achieves a globally optimal policy update that is not promised in the prevailing policy based RL algorithms. Experiments across tabular domains and robotic locomotion tasks demonstrate the effectiveness of our approach."
                    },
                    {
                        "title": "Strain engineering of magnetic states of vacancy-decorated hexagonal boron nitride",
                        "abstract": "Novel materials with tunable magnetic states play a significant role in the development of next-generation spintronic devices. In this paper, we examine the role of biaxial strain on the electronic properties of vacancy-decorated hexagonal boron nitride (h-BN) monolayers using density functional theory calculations. We found that the strain can lead to switching of the magnetic state for h-BN monolayers with boron vacancy or divacancy. Our findings promise a new route for the operation of low-dimensional spintronic devices."
                    },
                    {
                        "title": "Covalent pathways in engineering h-BN supported graphene",
                        "abstract": "Cross-planar di-vacancies (CPDVs) within stacked graphene hexagonal boron nitride (h-BN) heterostructures provide stabilized covalent links to bridge adjacent graphene and h-BN sheets. It was shown that the CPDVs serve as focal points for cross-planar atom transport between graphene and h-BN, and the chemical nature of interlayer links along with associated cross-planar migration pathways at these defects can be predictively manipulated through modulation of the chemical environment and charge engineering, to achieve consistent B or N doping and simultaneous healing of graphene. The present study proposed a viable approach integrating irradiation, chemical and charge engineering, to produce high-quality graphene with tunable electronic and electrochemical properties, using the h-BN substrate."
                    },
                    {
                        "title": "Towards Optimal Pricing of Demand Response -- A Nonparametric Constrained Policy Optimization Approach",
                        "abstract": "Demand response (DR) has been demonstrated to be an effective method for reducing peak load and mitigating uncertainties on both the supply and demand sides of the electricity market. One critical question for DR research is how to appropriately adjust electricity prices in order to shift electrical load from peak to off-peak hours. In recent years, reinforcement learning (RL) has been used to address the price-based DR problem because it is a model-free technique that does not necessitate the identification of models for end-use customers. However, the majority of RL methods cannot guarantee the stability and optimality of the learned pricing policy, which is undesirable in safety-critical power systems and may result in high customer bills. In this paper, we propose an innovative nonparametric constrained policy optimization approach that improves optimality while ensuring stability of the policy update, by removing the restrictive assumption on policy representation that the majority of the RL literature adopts: the policy must be parameterized or fall into a certain distribution class. We derive a closed-form expression of optimal policy update for each iteration and develop an efficient on-policy actor-critic algorithm to address the proposed constrained policy optimization problem. The experiments on two DR cases show the superior performance of our proposed nonparametric constrained policy optimization method compared with state-of-the-art RL algorithms."
                    },
                    {
                        "title": "Deep-MAPS: Machine Learning based Mobile Air Pollution Sensing",
                        "abstract": "Mobile and ubiquitous sensing of urban air quality has received increased attention as an economically and operationally viable means to survey atmospheric environment with high spatial-temporal resolution. This paper proposes a machine learning based mobile air pollution sensing framework, called Deep-MAPS, and demonstrates its scientific and financial values in the following aspects. (1) Based on a network of fixed and mobile air quality sensors, we perform spatial inference of PM2.5 concentrations in Beijing (3,025 km2, 19 Jun-16 Jul 2018) for a spatial-temporal resolution of 1km-by-1km and 1 hour, with over 85% accuracy. (2) We leverage urban big data to generate insights regarding the potential cause of pollution, which facilitates evidence-based sustainable urban management. (3) To achieve such spatial-temporal coverage and accuracy, Deep-MAPS can save up to 90% hardware investment, compared with ubiquitous sensing that relies primarily on fixed sensors."
                    },
                    {
                        "title": "Multivariate functional group sparse regression: functional predictor selection",
                        "abstract": "In this paper, we propose methods for functional predictor selection and the estimation of smooth functional coefficients simultaneously in a scalar-on-function regression problem under high-dimensional multivariate functional data setting. In particular, we develop two methods for functional group-sparse regression under a generic Hilbert space of infinite dimension. We show the convergence of algorithms and the consistency of the estimation and the selection (oracle property) under infinite-dimensional Hilbert spaces. Simulation studies show the effectiveness of the methods in both the selection and the estimation of functional coefficients. The applications to the functional magnetic resonance imaging (fMRI) reveal the regions of the human brain related to ADHD and IQ."
                    },
                    {
                        "title": "Enhancing the Resilience of Graph Neural Networks to Topological Perturbations in Sparse Graphs",
                        "abstract": "Graph neural networks (GNNs) have been extensively employed in node classification. Nevertheless, recent studies indicate that GNNs are vulnerable to topological perturbations, such as adversarial attacks and edge disruptions. Considerable efforts have been devoted to mitigating these challenges. For example, pioneering Bayesian methodologies, including GraphSS and LlnDT, incorporate Bayesian label transitions and topology-based label sampling to strengthen the robustness of GNNs. However, GraphSS is hindered by slow convergence, while LlnDT faces challenges in sparse graphs. To overcome these limitations, we propose a novel label inference framework, TraTopo, which combines topology-driven label propagation, Bayesian label transitions, and link analysis via random walks. TraTopo significantly surpasses its predecessors on sparse graphs by utilizing random walk sampling, specifically targeting isolated nodes for link prediction, thus enhancing its effectiveness in topological sampling contexts. Additionally, TraTopo employs a shortest-path strategy to refine link prediction, thereby reducing predictive overhead and improving label inference accuracy. Empirical evaluations highlight TraTopo's superiority in node classification, significantly exceeding contemporary GCN models in accuracy."
                    },
                    {
                        "title": "Effect of particle oxidation, size and material on deformation, bonding and deposition during cold spray: a peridynamic investigation",
                        "abstract": "Cold spray (CS) has emerged as an important additive manufacturing technology over the past decade. This study investigates the effect of oxide layers on the CS process, focusing on the deformation behavior of copper (Cu) and iron (Fe) particles upon collision with a matching substrate. Using a peridynamics-based approach, we examine the effects of oxide thickness, particle size, and particle/substrate material on material deformation and oxide fracture processes. Our results show that thicker oxide films restrict particle deformation, delay oxide discontinuities and material jetting, and increase the critical velocity required for metal to metal contact. Larger particles, despite uniform deformation across sizes, require lower velocities to initiate jetting and oxide separation because of their higher kinetic energy, leading to metallurgical bonding at lower velocities. Soft to soft impacts induce oxide film cracking at lower velocities, resulting in larger interface areas and more oxide-free contact zones, thereby reducing the critical velocity. Furthermore, the volume of residual oxide has a power-law relationship with the particle size, indicating that the oxide-cleaning ability of the particles affects the critical velocity. This study highlights the importance of oxide deformation and fracture during CS processes and provides valuable insights into the breakage and removal of oxides and subsequent metallic bond formation. These findings offer beneficial new knowledge for the rational design and optimization of CS processes."
                    }
                ]
            },
            "2de8b473-07a4-4319-9a63-4a943919a6a9": {
                "pk": "2de8b473-07a4-4319-9a63-4a943919a6a9",
                "name": "Shiji Song",
                "collaborators": [
                    "Keyou You",
                    "Cheng Wu",
                    "Gao Huang",
                    "Fei Dong",
                    "Hui Wu",
                    "Zhuo Li",
                    "Zhuolin Wang",
                    "Yuli Zhang",
                    "Wenjie Shi",
                    "Haofeng Yuan"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Control Systems",
                    "Optimization",
                    "Robotics"
                ],
                "publications": [
                    {
                        "title": "Target Encirclement with any Smooth Pattern Using Range-only Measurements",
                        "abstract": "This paper proposes a coordinate-free controller to drive a mobile robot to encircle a target at unknown position by only using range measurements. Different from the existing works, a backstepping based controller is proposed to encircle the target with zero steady-state error for any desired smooth pattern. Moreover, we show its asymptotic exponential convergence under a fixed set of control parameters, which are independent of the initial distance to the target. The effectiveness and advantages of the proposed controller are validated via simulations."
                    },
                    {
                        "title": "Depth Control of Model-Free AUVs via Reinforcement Learning",
                        "abstract": "In this paper, we consider depth control problems of an autonomous underwater vehicle (AUV) for tracking the desired depth trajectories. Due to the unknown dynamical model of the AUV, the problems cannot be solved by most of model-based controllers. To this purpose, we formulate the depth control problems of the AUV as continuous-state, continuous-action Markov decision processes (MDPs) under unknown transition probabilities. Based on deterministic policy gradient (DPG) and neural network approximation, we propose a model-free reinforcement learning (RL) algorithm that learns a state-feedback controller from sampled trajectories of the AUV. To improve the performance of the RL algorithm, we further propose a batch-learning scheme through replaying previous prioritized trajectories. We illustrate with simulations that our model-free method is even comparable to the model-based controllers as LQI and NMPC. Moreover, we validate the effectiveness of the proposed RL algorithm on a seafloor data set sampled from the South China Sea."
                    },
                    {
                        "title": "Cooperative Source Seeking via Networked Multi-vehicle Systems",
                        "abstract": "This paper studies the cooperative source seeking problem via a networked multi-vehicle system. In contrast to existing literature, the multi-vehicle system is controlled to the source position that maximizes aggregated multiple unknown scalar fields and each sensor-enabled vehicle only samples measurements of one scalar field. Thus, a single vehicle is unable to localize the source and has to cooperate with its neighboring vehicles. By jointly exploiting the ideas of the consensus algorithm and the stochastic extremum seeking (ES), this paper proposes novel distributed stochastic ES controllers, which are gradient-free and do not need any absolute information, such that the multi-vehicle system simultaneously approaches the source position. The effectiveness of the proposed controllers is proved for quadratic scalar fields. Finally, illustrative examples are included to validate the theoretical results."
                    },
                    {
                        "title": "Wasserstein Distributionally Robust Shortest Path Problem",
                        "abstract": "This paper proposes a data-driven distributionally robust shortest path (DRSP) model where the distribution of the travel time in the transportation network can only be partially observed through a finite number of samples. Specifically, we aim to find an optimal path to minimize the worst-case $\\alpha$-reliable mean-excess travel time (METT) over a Wasserstein ball, which is centered at the empirical distribution of the sample dataset and the ball radius quantifies the level of its confidence. In sharp contrast to the existing DRSP models, our model is equivalently reformulated as a tractable mixed 0-1 convex problem, e.g., 0-1 linear program or 0-1 second-order cone program. Moreover, we also explicitly derive the distribution achieving the worst-case METT by simply perturbing each sample. Experiments demonstrate the advantages of our DRSP model in terms of the out-of-sample performance and computational complexity. Finally, our DRSP model is easily extended to solve the DR bi-criteria shortest path problem and the minimum cost flow problem."
                    },
                    {
                        "title": "Second-order Conic Programming Approach for Wasserstein Distributionally Robust Two-stage Linear Programs",
                        "abstract": "This paper proposes a second-order conic programming (SOCP) approach to solve distributionally robust two-stage stochastic linear programs over 1-Wasserstein balls. We start from the case with distribution uncertainty only in the objective function and exactly reformulate it as an SOCP problem. Then, we study the case with distribution uncertainty only in constraints, and show that such a robust program is generally NP-hard as it involves a norm maximization problem over a polyhedron. However, it is reduced to an SOCP problem if the extreme points of the polyhedron are given as a prior. This motivates to design a constraint generation algorithm with provable convergence to approximately solve the NP-hard problem. In sharp contrast to the exiting literature, the distribution achieving the worst-case cost is given as an \"empirical\" distribution by simply perturbing each sample for both cases. Finally, experiments illustrate the advantages of the proposed model in terms of the out-of-sample performance and the computational complexity."
                    },
                    {
                        "title": "Optimization-based Control for Bearing-only Target Search with a Mobile Vehicle",
                        "abstract": "This work aims to design an optimization-based controller for a discrete-time Dubins vehicle to approach a target with unknown position as fast as possible by only using bearing measurements. To this end, we propose a bi-objective optimization problem, which jointly considers the performance of estimating the unknown target position and controlling the mobile vehicle to a known position, and then adopt a weighted sum method with normalization to solve it. The controller is given based on the solution of the optimization problem in ties with a least-square estimate of the target position. Moreover, the controller does not need the vehicle's global position information. Finally, simulation results are included to validate the effectiveness of the proposed controller."
                    },
                    {
                        "title": "Soft Policy Gradient Method for Maximum Entropy Deep Reinforcement Learning",
                        "abstract": "Maximum entropy deep reinforcement learning (RL) methods have been demonstrated on a range of challenging continuous tasks. However, existing methods either suffer from severe instability when training on large off-policy data or cannot scale to tasks with very high state and action dimensionality such as 3D humanoid locomotion. Besides, the optimality of desired Boltzmann policy set for non-optimal soft value function is not persuasive enough. In this paper, we first derive soft policy gradient based on entropy regularized expected reward objective for RL with continuous actions. Then, we present an off-policy actor-critic, model-free maximum entropy deep RL algorithm called deep soft policy gradient (DSPG) by combining soft policy gradient with soft Bellman equation. To ensure stable learning while eliminating the need of two separate critics for soft value functions, we leverage double sampling approach to making the soft Bellman equation tractable. The experimental results demonstrate that our method outperforms in performance over off-policy prior methods."
                    },
                    {
                        "title": "A selected review on reinforcement learning based control for autonomous underwater vehicles",
                        "abstract": "Recently, reinforcement learning (RL) has been extensively studied and achieved promising results in a wide range of control tasks. Meanwhile, autonomous underwater vehicle (AUV) is an important tool for executing complex and challenging underwater tasks. The advances in RL offers ample opportunities for developing intelligent AUVs. This paper provides a selected review on RL based control for AUVs with the focus on applications of RL to low-level control tasks for underwater regulation and tracking. To this end, we first present a concise introduction to the RL based control framework. Then, we provide an overview of RL methods for AUVs control problems, where the main challenges and recent progresses are discussed. Finally, two representative cases of RL-based controllers are given in detail for the model-free RL methods on AUVs."
                    },
                    {
                        "title": "Learn From the Past: Experience Ensemble Knowledge Distillation",
                        "abstract": "Traditional knowledge distillation transfers \"dark knowledge\" of a pre-trained teacher network to a student network, and ignores the knowledge in the training process of the teacher, which we call teacher's experience. However, in realistic educational scenarios, learning experience is often more important than learning results. In this work, we propose a novel knowledge distillation method by integrating the teacher's experience for knowledge transfer, named experience ensemble knowledge distillation (EEKD). We save a moderate number of intermediate models from the training process of the teacher model uniformly, and then integrate the knowledge of these intermediate models by ensemble technique. A self-attention module is used to adaptively assign weights to different intermediate models in the process of knowledge transfer. Three principles of constructing EEKD on the quality, weights and number of intermediate models are explored. A surprising conclusion is found that strong ensemble teachers do not necessarily produce strong students. The experimental results on CIFAR-100 and ImageNet show that EEKD outperforms the mainstream knowledge distillation methods and achieves the state-of-the-art. In particular, EEKD even surpasses the standard ensemble distillation on the premise of saving training cost."
                    },
                    {
                        "title": "The Neural-Prediction based Acceleration Algorithm of Column Generation for Graph-Based Set Covering Problems",
                        "abstract": "Set covering problem is an important class of combinatorial optimization problems, which has been widely applied and studied in many fields. In this paper, we propose an improved column generation algorithm with neural prediction (CG-P) for solving graph-based set covering problems. We leverage a graph neural network based neural prediction model to predict the probability to be included in the final solution for each edge. Our CG-P algorithm constructs a reduced graph that only contains the edges with higher predicted probability, and this graph reduction process significantly speeds up the solution process. We evaluate the CG-P algorithm on railway crew scheduling problems and it outperforms the baseline column generation algorithm. We provide two solution modes for our CG-P algorithm. In the optimal mode, we can obtain a solution with an optimality guarantee while reducing the time cost to 63.12%. In the fast mode, we can obtain a sub-optimal solution with a 7.62% optimality gap in only 2.91% computation time."
                    },
                    {
                        "title": "Standoff Tracking Using DNN-Based MPC with Implementation on FPGA",
                        "abstract": "This work studies the standoff tracking problem to drive an unmanned aerial vehicle (UAV) to slide on a desired circle over a moving target at a constant height. We propose a novel Lyapunov guidance vector (LGV) field with tunable convergence rates for the UAV's trajectory planning and a deep neural network (DNN)-based model predictive control (MPC) scheme to track the reference trajectory. Then, we show how to collect samples for training the DNN offline and design an integral module (IM) to refine the tracking performance of our DNN-based MPC. Moreover, the hardware-in-the-loop (HIL) simulation with an FPGA@200MHz demonstrates that our method is a valid alternative to embedded implementations of MPC for addressing complex systems and applications which is impossible for directly solving the MPC optimization problems."
                    },
                    {
                        "title": "Meta-Semi: A Meta-learning Approach for Semi-supervised Learning",
                        "abstract": "Deep learning based semi-supervised learning (SSL) algorithms have led to promising results in recent years. However, they tend to introduce multiple tunable hyper-parameters, making them less practical in real SSL scenarios where the labeled data is scarce for extensive hyper-parameter search. In this paper, we propose a novel meta-learning based SSL algorithm (Meta-Semi) that requires tuning only one additional hyper-parameter, compared with a standard supervised deep learning algorithm, to achieve competitive performance under various conditions of SSL. We start by defining a meta optimization problem that minimizes the loss on labeled data through dynamically reweighting the loss on unlabeled samples, which are associated with soft pseudo labels during training. As the meta problem is computationally intensive to solve directly, we propose an efficient algorithm to dynamically obtain the approximate solutions. We show theoretically that Meta-Semi converges to the stationary point of the loss function on labeled data under mild conditions. Empirically, Meta-Semi outperforms state-of-the-art SSL algorithms significantly on the challenging semi-supervised CIFAR-100 and STL-10 tasks, and achieves competitive performance on CIFAR-10 and SVHN."
                    },
                    {
                        "title": "Temporal-Spatial Causal Interpretations for Vision-Based Reinforcement Learning",
                        "abstract": "Deep reinforcement learning (RL) agents are becoming increasingly proficient in a range of complex control tasks. However, the agent's behavior is usually difficult to interpret due to the introduction of black-box function, making it difficult to acquire the trust of users. Although there have been some interesting interpretation methods for vision-based RL, most of them cannot uncover temporal causal information, raising questions about their reliability. To address this problem, we present a temporal-spatial causal interpretation (TSCI) model to understand the agent's long-term behavior, which is essential for sequential decision-making. TSCI model builds on the formulation of temporal causality, which reflects the temporal causal relations between sequential observations and decisions of RL agent. Then a separate causal discovery network is employed to identify temporal-spatial causal features, which are constrained to satisfy the temporal causality. TSCI model is applicable to recurrent agents and can be used to discover causal features with high efficiency once trained. The empirical results show that TSCI model can produce high-resolution and sharp attention masks to highlight task-relevant temporal-spatial information that constitutes most evidence about how vision-based RL agents make sequential decisions. In addition, we further demonstrate that our method is able to provide valuable causal interpretations for vision-based RL agents from the temporal perspective."
                    },
                    {
                        "title": "Learning Specialized Activation Functions for Physics-informed Neural Networks",
                        "abstract": "Physics-informed neural networks (PINNs) are known to suffer from optimization difficulty. In this work, we reveal the connection between the optimization difficulty of PINNs and activation functions. Specifically, we show that PINNs exhibit high sensitivity to activation functions when solving PDEs with distinct properties. Existing works usually choose activation functions by inefficient trial-and-error. To avoid the inefficient manual selection and to alleviate the optimization difficulty of PINNs, we introduce adaptive activation functions to search for the optimal function when solving different problems. We compare different adaptive activation functions and discuss their limitations in the context of PINNs. Furthermore, we propose to tailor the idea of learning combinations of candidate activation functions to the PINNs optimization, which has a higher requirement for the smoothness and diversity on learned functions. This is achieved by removing activation functions which cannot provide higher-order derivatives from the candidate set and incorporating elementary functions with different properties according to our prior knowledge about the PDE at hand. We further enhance the search space with adaptive slopes. The proposed adaptive activation function can be used to solve different PDE systems in an interpretable way. Its effectiveness is demonstrated on a series of benchmarks. Code is available at https://github.com/LeapLabTHU/AdaAFforPINNs."
                    },
                    {
                        "title": "A Reinforcement-Learning-Based Multiple-Column Selection Strategy for Column Generation",
                        "abstract": "Column generation (CG) is one of the most successful approaches for solving large-scale linear programming (LP) problems. Given an LP with a prohibitively large number of variables (i.e., columns), the idea of CG is to explicitly consider only a subset of columns and iteratively add potential columns to improve the objective value. While adding the column with the most negative reduced cost can guarantee the convergence of CG, it has been shown that adding multiple columns per iteration rather than a single column can lead to faster convergence. However, it remains a challenge to design a multiple-column selection strategy to select the most promising columns from a large number of candidate columns. In this paper, we propose a novel reinforcement-learning-based (RL) multiple-column selection strategy. To the best of our knowledge, it is the first RL-based multiple-column selection strategy for CG. The effectiveness of our approach is evaluated on two sets of problems: the cutting stock problem and the graph coloring problem. Compared to several widely used single-column and multiple-column selection strategies, our RL-based multiple-column selection strategy leads to faster convergence and achieves remarkable reductions in the number of CG iterations and runtime."
                    },
                    {
                        "title": "Likelihood Ratio Based Scheduler for Secure Detection in Cyber Physical Systems",
                        "abstract": "This paper is concerned with a binary detection problem over a non-secure network. To satisfy the communication rate constraint and against possible cyber attacks, which are modeled as deceptive signals injected to the network, a likelihood ratio based (LRB) scheduler is designed in the sensor side to smartly select sensor measurements for transmission. By exploring the scheduler, some sensor measurements are successfully retrieved from the attacked data at the decision center. We show that even under a moderate communication rate constraint of secure networks, an optimal LRB scheduler can achieve a comparable asymptotic detection performance to the standard N-P test using the full set of measurements, and is strictly better than the random scheduler. For non-secure networks, the LRB scheduler can also maintain the detection functionality but suffers graceful performance degradation under different attack intensities. Finally, we perform simulations to validate our theoretical results."
                    }
                ]
            },
            "03997bab-3f75-4b3e-a9af-7e1d6b651c3d": {
                "pk": "03997bab-3f75-4b3e-a9af-7e1d6b651c3d",
                "name": "Bo Zheng",
                "collaborators": [
                    "Shizeng Lin",
                    "Tian Qiu",
                    "Guang Chen",
                    "Jan Korbel",
                    "Xiongfei Jiang"
                ],
                "domain": [
                    "Statistical Physics",
                    "Complex Networks",
                    "Monte Carlo Simulations",
                    "Financial Networks"
                ],
                "publications": [
                    {
                        "title": "Short-time critical dynamics at perfect and non-perfect surface",
                        "abstract": "We report Monte Carlo simulations of critical dynamics far from equilibrium on a perfect and non-perfect surface in the 3d Ising model. For an ordered initial state, the dynamic relaxation of the surface magnetization, the line magnetization of the defect line, and the corresponding susceptibilities and appropriate cumulant is carefully examined at the ordinary, special and surface phase transitions. The universal dynamic scaling behavior including a dynamic crossover scaling form is identified. The exponent $\\beta_1$ of the surface magnetization and $\\beta_2$ of the line magnetization are extracted. The impact of the defect line on the surface universality classes is investigated."
                    },
                    {
                        "title": "Two Parallel Swendsen-Wang Cluster Algorithms Using Message-Passing Paradigm",
                        "abstract": "In this article, we present two different parallel Swendsen-Wang Cluster(SWC) algorithms using message-passing interface(MPI). One is based on Master-Slave Parallel Model(MSPM) and the other is based on Data-Parallel Model(DPM). A speedup of 24 with 40 processors and 16 with 37 processors is achieved with the DPM and MSPM respectively. The speedup of both algorithms at different temperature and system size is carefully examined both experimentally and theoretically, and a comparison of their efficiency is made. In the last section, based on these two parallel SWC algorithms, two parallel probability changing cluster(PCC) algorithms are proposed."
                    },
                    {
                        "title": "Adaptive financial networks with static and dynamic thresholds",
                        "abstract": "Based on the daily data of American and Chinese stock markets, the dynamic behavior of a financial network with static and dynamic thresholds is investigated. Compared with the static threshold, the dynamic threshold suppresses the large fluctuation induced by the cross-correlation of individual stock prices, and leads to a stable topological structure in the dynamic evolution. Long-range time-correlations are revealed for the average clustering coefficient, average degree and cross-correlation of degrees. The dynamic network shows a two-peak behavior in the degree distribution."
                    },
                    {
                        "title": "Transfer entropy between communities in complex networks",
                        "abstract": "With the help of transfer entropy, we analyze information flows between communities of complex networks. We show that the transfer entropy provides a coherent description of interactions between communities, including non-linear interactions. To put some flesh on the bare bones, we analyze transfer entropies between communities of five largest financial markets, represented as networks of interacting stocks. Additionally, we discuss information transfer of rare events, which is analyzed by R\\'enyi transfer entropy."
                    }
                ]
            },
            "06452c2a-d938-4f7b-be89-3b23191dd983": {
                "pk": "06452c2a-d938-4f7b-be89-3b23191dd983",
                "name": "Gao Huang",
                "collaborators": [
                    "Song Li"
                ],
                "domain": [
                    "Matrix Recovery",
                    "Statistical Learning",
                    "Optimization",
                    "Phase Retrieval"
                ],
                "publications": [
                    {
                        "title": "Low-Rank Toeplitz Matrix Restoration: Descent Cone Analysis and Structured Random Matrix",
                        "abstract": "This note demonstrates that we can stably recover rank $r$ Toeplitz matrix $\\pmb{X}\\in\\mathbb{R}^{n\\times n}$ from a number of rank one subgaussian measurements on the order of $r\\log^{2} n$ with an exponentially decreasing failure probability by employing a nuclear norm minimization program. Our approach utilizes descent cone analysis through Mendelson's small ball method with the Toeplitz constraint. The key ingredient is to determine the spectral norm of the random matrix of the Topelitz structure, which may be of independent interest.This improves upon earlier analyses and resolves the conjecture in Chen et al. (IEEE Transactions on Information Theory, 2015)."
                    },
                    {
                        "title": "Geometric Characteristics in Phaseless Operator and Structured Matrix Recovery",
                        "abstract": "In this paper, we first propose a unified approach for analyzing the stability of the phaseless operator for both amplitude and intensity measurement on an arbitrary geometric set, thus characterizing the robust performance of phase retrieval via the empirical minimization method. We introduce the random embedding of concave lifting operators in tangent space to characterize the unified analysis of any geometric set. Similarly, we investigate the structured matrix recovery problem through the robust injectivity of a linear rank one measurement operator on an arbitrary matrix set. The core of our analysis is to establish a unified empirical chaos process characterization for various matrix sets. Talagrand's $\\gamma_{\\alpha}$-functionals are introduced to characterize the connection between the geometric constraints and the number of measurements needed to guarantee stability or robust injectivity. Finally, we construct adversarial noise to demonstrate the sharpness of the recovery bounds in the above two scenarios."
                    }
                ]
            }
        }
    },
    "2407.09087": {
        "paper_data": {
            "title": "On the Role of Discrete Tokenization in Visual Representation Learning",
            "url": "http://arxiv.org/abs/2407.09087v1",
            "arxiv_id": "2407.09087",
            "authors": [
                "Tianqi Du",
                "Yifei Wang",
                "Yisen Wang"
            ],
            "abstract": "In the realm of self-supervised learning (SSL), masked image modeling (MIM) has gained popularity alongside contrastive learning methods. MIM involves reconstructing masked regions of input images using their unmasked portions. A notable subset of MIM methodologies employs discrete tokens as the reconstruction target, but the theoretical underpinnings of this choice remain underexplored. In this paper, we explore the role of these discrete tokens, aiming to unravel their benefits and limitations. Building upon the connection between MIM and contrastive learning, we provide a comprehensive theoretical understanding on how discrete tokenization affects the model's generalization capabilities. Furthermore, we propose a novel metric named TCAS, which is specifically designed to assess the effectiveness of discrete tokens within the MIM framework. Inspired by this metric, we contribute an innovative tokenizer design and propose a corresponding MIM method named ClusterMIM. It demonstrates superior performance on a variety of benchmark datasets and ViT backbones. Code is available at https://github.com/PKU-ML/ClusterMIM.",
            "introduction": "   1 Introduction  Self-Supervised Learning (SSL) has recently emerged as a promising paradigm for learning meaningful data representations without access to labels. Besides the popular contrastive learning methods (Chen et\u00a0al., 2020; Chen & He, 2020; Grill et\u00a0al., 2020; Zhuo et\u00a0al., 2023; Wang et\u00a0al., 2021; 2023; Cui et\u00a0al., 2023; Zhang et\u00a0al., 2023a; b), there is a growing interest in masked image modeling (MIM) techniques. MIM requires masking parts of input images and subsequently attempting their reconstruction using the remaining unmasked parts. Notable examples like MAE (He et\u00a0al., 2022), BEiT (Bao et\u00a0al., 2022), PeCo (Dong et\u00a0al., 2023), MaskFeat (Wei et\u00a0al., 2022), and MAGE (Li et\u00a0al., 2023) have demonstrated state-of-the-art performance on various downstream tasks.   Within the MIM domain, various reconstruction targets exist. For instance, MAE (He et\u00a0al., 2022) employs the raw pixel values of the unmasked parts as its reconstruction target and MaskFeat (Wei et\u00a0al., 2022) leverages features from other pretrained models as its reconstruction target, whereas some alternative methods adopt visual tokenizers to generate the reconstruction target (Bao et\u00a0al., 2022; Dong et\u00a0al., 2023). Such tokenizers convert image patches into predefined discrete tokens. As an illustration, BEiT (Bao et\u00a0al., 2022) deploys visual tokens generated by a discrete tokenizer known as dVAE (Vahdat et\u00a0al., 2018). PeCo (Dong et\u00a0al., 2023) refines the token quality by introducing a tokenizer enriched with perceptual loss. More related work about MIM can be found in Appendix A. Nevertheless, we notice that different tokenization schemes may bring quite different performance. For example, as Table 1 suggests, PeCo with perceptual tokenizer and MaskFeat with HOG targets outperform MAE, whereas BEiT with dVAE and VQGAN tokenizer underperform MAE using raw pixel. These observations naturally raise the question: What is the role of tokenization in MIM? How does it affect the downstream generalization?   To answer these questions, we leverage the graph perspective of MAE (Zhang et\u00a0al., 2022) to dissect the influence of different discrete tokenization schemes on the downstream generalization. Zhang et\u00a0al. (2022) have demonstrated that masking can generate implicit positive correlations between unmasked views that share the same target masked output. However, it\u2019s highly improbable for two unmasked views to share precisely the same masked output. This discrepancy results in limited connectivity within the augmentation graph, potentially leading to suboptimal performance as shown in prior studies (Wang et\u00a0al., 2022; HaoChen et\u00a0al., 2021). Subsequently, we observe that when employing discrete tokens, masking creates implicit positive correlations between unmasked views as long as they share the same discrete class of target masked output. Specifically, we find that discrete tokens that align well with data classes enhance connectivity among intra-class samples, thereby improving the downstream performance. Conversely, incorrectly specified tokenizers may cause confusion between inter-class samples, resulting in poorer downstream performance.   This insight can also guide the selection of proper tokenizers. Specifically, we design a novel metric, named token-class alignment similarity (TCAS), by measuring the discrepancy in distribution between true patch labels and the discrete token labels assigned by the tokenizer. TCAS can be used to directly compare the quality of different tokenizers without training. Meanwhile, the principle inspires us to design an easy-to-implement tokenizer with its corresponding MIM method named ClusterMIM,",
            "references": [
                {
                    "title": "Identifiable Contrastive Learning with Automatic Feature Importance Discovery",
                    "abstract": "Existing contrastive learning methods rely on pairwise sample contrast $z_x^\\top z_{x'}$ to learn data representations, but the learned features often lack clear interpretability from a human perspective. Theoretically, it lacks feature identifiability and different initialization may lead to totally different features. In this paper, we study a new method named tri-factor contrastive learning (triCL) that involves a 3-factor contrast in the form of $z_x^\\top S z_{x'}$, where $S=\\text{diag}(s_1,\\dots,s_k)$ is a learnable diagonal matrix that automatically captures the importance of each feature. We show that by this simple extension, triCL can not only obtain identifiable features that eliminate randomness but also obtain more interpretable features that are ordered according to the importance matrix $S$. We show that features with high importance have nice interpretability by capturing common classwise features, and obtain superior performance when evaluated for image retrieval using a few features. The proposed triCL objective is general and can be applied to different contrastive learning methods like SimCLR and CLIP. We believe that it is a better alternative to existing 2-factor contrastive learning by improving its identifiability and interpretability with minimal overhead. Code is available at https://github.com/PKU-ML/Tri-factor-Contrastive-Learning."
                },
                {
                    "title": "Rethinking Weak Supervision in Helping Contrastive Learning",
                    "abstract": "Contrastive learning has shown outstanding performances in both supervised and unsupervised learning, and has recently been introduced to solve weakly supervised learning problems such as semi-supervised learning and noisy label learning. Despite the empirical evidence showing that semi-supervised labels improve the representations of contrastive learning, it remains unknown if noisy supervised information can be directly used in training instead of after manual denoising. Therefore, to explore the mechanical differences between semi-supervised and noisy-labeled information in helping contrastive learning, we establish a unified theoretical framework of contrastive learning under weak supervision. Specifically, we investigate the most intuitive paradigm of jointly training supervised and unsupervised contrastive losses. By translating the weakly supervised information into a similarity graph under the framework of spectral clustering based on the posterior probability of weak labels, we establish the downstream classification error bound. We prove that semi-supervised labels improve the downstream error bound whereas noisy labels have limited effects under such a paradigm. Our theoretical findings here provide new insights for the community to rethink the role of weak supervision in helping contrastive learning."
                },
                {
                    "title": "On the Generalization of Multi-modal Contrastive Learning",
                    "abstract": "Multi-modal contrastive learning (MMCL) has recently garnered considerable interest due to its superior performance in visual tasks, achieved by embedding multi-modal data, such as visual-language pairs. However, there still lack theoretical understandings of how MMCL extracts useful visual representation from multi-modal pairs, and particularly, how MMCL outperforms previous approaches like self-supervised contrastive learning (SSCL). In this paper, by drawing an intrinsic connection between MMCL and asymmetric matrix factorization, we establish the first generalization guarantees of MMCL for visual downstream tasks. Based on this framework, we further unify MMCL and SSCL by showing that MMCL implicitly performs SSCL with (pseudo) positive pairs induced by text pairs. Through this unified perspective, we characterize the advantage of MMCL by showing that text pairs induce more semantically consistent and diverse positive pairs, which, according to our analysis, provably benefit downstream generalization. Inspired by this finding, we propose CLIP-guided resampling methods to significantly improve the downstream performance of SSCL on ImageNet by leveraging multi-modal information. Code is available at https://github.com/PKU-ML/CLIP-Help-SimCLR."
                },
                {
                    "title": "A Message Passing Perspective on Learning Dynamics of Contrastive Learning",
                    "abstract": "In recent years, contrastive learning achieves impressive results on self-supervised visual representation learning, but there still lacks a rigorous understanding of its learning dynamics. In this paper, we show that if we cast a contrastive objective equivalently into the feature space, then its learning dynamics admits an interpretable form. Specifically, we show that its gradient descent corresponds to a specific message passing scheme on the corresponding augmentation graph. Based on this perspective, we theoretically characterize how contrastive learning gradually learns discriminative features with the alignment update and the uniformity update. Meanwhile, this perspective also establishes an intriguing connection between contrastive learning and Message Passing Graph Neural Networks (MP-GNNs). This connection not only provides a unified understanding of many techniques independently developed in each community, but also enables us to borrow techniques from MP-GNNs to design new contrastive learning variants, such as graph attention, graph rewiring, jumpy knowledge techniques, etc. We believe that our message passing perspective not only provides a new theoretical understanding of contrastive learning dynamics, but also bridges the two seemingly independent areas together, which could inspire more interleaving studies to benefit from each other. The code is available at https://github.com/PKU-ML/Message-Passing-Contrastive-Learning."
                },
                {
                    "title": "Towards a Unified Theoretical Understanding of Non-contrastive Learning via Rank Differential Mechanism",
                    "abstract": "Recently, a variety of methods under the name of non-contrastive learning (like BYOL, SimSiam, SwAV, DINO) show that when equipped with some asymmetric architectural designs, aligning positive pairs alone is sufficient to attain good performance in self-supervised visual learning. Despite some understandings of some specific modules (like the predictor in BYOL), there is yet no unified theoretical understanding of how these seemingly different asymmetric designs can all avoid feature collapse, particularly considering methods that also work without the predictor (like DINO). In this work, we propose a unified theoretical understanding for existing variants of non-contrastive learning. Our theory named Rank Differential Mechanism (RDM) shows that all these asymmetric designs create a consistent rank difference in their dual-branch output features. This rank difference will provably lead to an improvement of effective dimensionality and alleviate either complete or dimensional feature collapse. Different from previous theories, our RDM theory is applicable to different asymmetric designs (with and without the predictor), and thus can serve as a unified understanding of existing non-contrastive learning methods. Besides, our RDM theory also provides practical guidelines for designing many new non-contrastive variants. We show that these variants indeed achieve comparable performance to existing methods on benchmark datasets, and some of them even outperform the baselines. Our code is available at \\url{https://github.com/PKU-ML/Rank-Differential-Mechanism}."
                },
                {
                    "title": "MAGE: MAsked Generative Encoder to Unify Representation Learning and Image Synthesis",
                    "abstract": "Generative modeling and representation learning are two key tasks in computer vision. However, these models are typically trained independently, which ignores the potential for each task to help the other, and leads to training and model maintenance overheads. In this work, we propose MAsked Generative Encoder (MAGE), the first framework to unify SOTA image generation and self-supervised representation learning. Our key insight is that using variable masking ratios in masked image modeling pre-training can allow generative training (very high masking ratio) and representation learning (lower masking ratio) under the same training framework. Inspired by previous generative models, MAGE uses semantic tokens learned by a vector-quantized GAN at inputs and outputs, combining this with masking. We can further improve the representation by adding a contrastive loss to the encoder output. We extensively evaluate the generation and representation learning capabilities of MAGE. On ImageNet-1K, a single MAGE ViT-L model obtains 9.10 FID in the task of class-unconditional image generation and 78.9% top-1 accuracy for linear probing, achieving state-of-the-art performance in both image generation and representation learning. Code is available at https://github.com/LTHl4/mage."
                },
                {
                    "title": "How Mask Matters: Towards Theoretical Understandings of Masked Autoencoders",
                    "abstract": "Masked Autoencoders (MAE) based on a reconstruction task have risen to be a promising paradigm for self-supervised learning (SSL) and achieve state-of-the-art performance across different benchmark datasets. However, despite its impressive empirical success, there is still limited theoretical understanding of it. In this paper, we propose a theoretical understanding of how masking matters for MAE to learn meaningful features. We establish a close connection between MAE and contrastive learning, which shows that MAE implicit aligns the mask-induced positive pairs. Built upon this connection, we develop the first downstream guarantees for MAE methods, and analyze the effect of mask ratio. Besides, as a result of the implicit alignment, we also point out the dimensional collapse issue of MAE, and propose a Uniformity-enhanced MAE (U-MAE) loss that can effectively address this issue and bring significant improvements on real-world datasets, including CIFAR-10, ImageNet-100, and ImageNet-1K. Code is available at (https://github.com/zhangq327/U-MAE)."
                },
                {
                    "title": "Exploring Target Representations for Masked Autoencoders",
                    "abstract": "Masked autoencoders have become popular training paradigms for self-supervised visual representation learning. These models randomly mask a portion of the input and reconstruct the masked portion according to the target representations. In this paper, we first show that a careful choice of the target representation is unnecessary for learning good representations, since different targets tend to derive similarly behaved models. Driven by this observation, we propose a multi-stage masked distillation pipeline and use a randomly initialized model as the teacher, enabling us to effectively train high-capacity models without any efforts to carefully design target representations. Interestingly, we further explore using teachers of larger capacity, obtaining distilled students with remarkable transferring ability. On different tasks of classification, transfer learning, object detection, and semantic segmentation, the proposed method to perform masked knowledge distillation with bootstrapped teachers (dBOT) outperforms previous self-supervised methods by nontrivial margins. We hope our findings, as well as the proposed method, could motivate people to rethink the roles of target representations in pre-training masked autoencoders.The code and pre-trained models are publicly available at https://github.com/liuxingbin/dbot."
                },
                {
                    "title": "MixMAE: Mixed and Masked Autoencoder for Efficient Pretraining of Hierarchical Vision Transformers",
                    "abstract": "In this paper, we propose Mixed and Masked AutoEncoder (MixMAE), a simple but efficient pretraining method that is applicable to various hierarchical Vision Transformers. Existing masked image modeling (MIM) methods for hierarchical Vision Transformers replace a random subset of input tokens with a special [MASK] symbol and aim at reconstructing original image tokens from the corrupted image. However, we find that using the [MASK] symbol greatly slows down the training and causes pretraining-finetuning inconsistency, due to the large masking ratio (e.g., 60% in SimMIM). On the other hand, MAE does not introduce [MASK] tokens at its encoder at all but is not applicable for hierarchical Vision Transformers. To solve the issue and accelerate the pretraining of hierarchical models, we replace the masked tokens of one image with visible tokens of another image, i.e., creating a mixed image. We then conduct dual reconstruction to reconstruct the two original images from the mixed input, which significantly improves efficiency. While MixMAE can be applied to various hierarchical Transformers, this paper explores using Swin Transformer with a large window size and scales up to huge model size (to reach 600M parameters). Empirical results demonstrate that MixMAE can learn high-quality visual representations efficiently. Notably, MixMAE with Swin-B/W14 achieves 85.1% top-1 accuracy on ImageNet-1K by pretraining for 600 epochs. Besides, its transfer performances on the other 6 datasets show that MixMAE has better FLOPs / performance tradeoff than previous popular MIM methods."
                },
                {
                    "title": "Chaos is a Ladder: A New Theoretical Understanding of Contrastive Learning via Augmentation Overlap",
                    "abstract": "Recently, contrastive learning has risen to be a promising approach for large-scale self-supervised learning. However, theoretical understanding of how it works is still unclear. In this paper, we propose a new guarantee on the downstream performance without resorting to the conditional independence assumption that is widely adopted in previous work but hardly holds in practice. Our new theory hinges on the insight that the support of different intra-class samples will become more overlapped under aggressive data augmentations, thus simply aligning the positive samples (augmented views of the same sample) could make contrastive learning cluster intra-class samples together. Based on this augmentation overlap perspective, theoretically, we obtain asymptotically closed bounds for downstream performance under weaker assumptions, and empirically, we propose an unsupervised model selection metric ARC that aligns well with downstream accuracy. Our theory suggests an alternative understanding of contrastive learning: the role of aligning positive samples is more like a surrogate task than an ultimate goal, and the overlapped augmented views (i.e., the chaos) create a ladder for contrastive learning to gradually learn class-separated representations. The code for computing ARC is available at https://github.com/zhangq327/ARC."
                },
                {
                    "title": "How to Understand Masked Autoencoders",
                    "abstract": "\"Masked Autoencoders (MAE) Are Scalable Vision Learners\"revolutionizes the self-supervised learning method in that it not only achieves the state-of-the-art for image pre-training, but is also a milestone that bridges the gap between visual and linguistic masked autoencoding (BERT-style) pre-trainings. However, to our knowledge, to date there are no theoretical perspectives to explain the powerful expressivity of MAE. In this paper, we, for the first time, propose a unified theoretical framework that provides a mathematical understanding for MAE. Specifically, we explain the patch-based attention approaches of MAE using an integral kernel under a non-overlapping domain decomposition setting. To help the research community to further comprehend the main reasons of the great success of MAE, based on our framework, we pose five questions and answer them with mathematical rigor using insights from operator theory."
                },
                {
                    "title": "Masked Feature Prediction for Self-Supervised Visual Pre-Training",
                    "abstract": "We present Masked Feature Prediction (MaskFeat) for self-supervised pre-training of video models. Our approach first randomly masks out a portion of the input sequence and then predicts the feature of the masked regions. We study five different types of features and find Histograms of Oriented Gradients (HOG), a hand-crafted feature descriptor, works particularly well in terms of both performance and efficiency. We observe that the local contrast normalization in HOG is essential for good results, which is in line with earlier work using HOG for visual recognition. Our approach can learn abundant visual knowledge and drive large-scale Transformer based models. Without using extra model weights or supervision, MaskFeat pretrained on unlabeled videos achieves unprecedented results of 86.7% with MViTv2-L on Kinetics-400, 88.3% on Kinetics 600, 80.4% on Kinetics-700, 38.8 mAP on AVA, and 75.0% on SSv2. MaskFeat further generalizes to image input, which can be interpreted as a video with a single frame and obtains competitive results on ImageN et."
                },
                {
                    "title": "PeCo: Perceptual Codebook for BERT Pre-training of Vision Transformers",
                    "abstract": "This paper explores a better prediction target for BERT pre-training of vision transformers. We observe that current prediction targets disagree with human perception judgment. This contradiction motivates us to learn a perceptual prediction target. We argue that perceptually similar images should stay close to each other in the prediction target space. We surprisingly find one simple yet effective idea: enforcing perceptual similarity during the dVAE training. Moreover, we adopt a self-supervised transformer model for deep feature extraction and show that it works well for calculating perceptual similarity. We demonstrate that such learned visual tokens indeed exhibit better semantic meanings, and help pre-training achieve superior transfer performance in various downstream tasks. For example, we achieve 84.5% Top-1 accuracy on ImageNet-1K with ViT-B backbone, outperforming the competitive method BEiT by +1.3% under the same pre-training epochs. Our approach also gets significant improvement on object detection and segmentation on COCO and semantic segmentation on ADE20K. Equipped with a larger backbone ViT-H, we achieve the state-of-the-art ImageNet accuracy (88.3%) among methods using only ImageNet-1K data."
                },
                {
                    "title": "SimMIM: a Simple Framework for Masked Image Modeling",
                    "abstract": "This paper presents SimMIM, a simple framework for masked image modeling. We have simplified recently proposed relevant approaches, without the need for special designs, such as block-wise masking and tokenization via discrete VAE or clustering. To investigate what makes a masked image modeling task learn good representations, we systematically study the major components in our framework, and find that the simple designs of each component have revealed very strong representation learning performance: 1) random masking of the input image with a moderately large masked patch size (e.g., 32) makes a powerful pre-text task; 2) predicting RGB values of raw pixels by direct regression performs no worse than the patch classification approaches with complex designs; 3) the prediction head can be as light as a linear layer, with no worse performance than heavier ones. Using ViT-B, our approach achieves 83.8% top-1 fine-tuning accuracy on ImageNet-1K by pre-training also on this dataset, surpassing previous best approach by +0.6%. When applied to a larger model with about 650 million parameters, SwinV2-H, it achieves 87.1% top-1 accuracy on ImageNet-1K using only ImageNet-1K data. We also leverage this approach to address the data-hungry issue faced by large-scale model training, that a 3B model (Swin V2-G) is successfully trained to achieve state-of-the-art accuracy on four representative vision benchmarks using 40\u00d7 less labelled data than that in previous practice (JFT-3B). The code is available at https://github.com/microsoft/SimMIM."
                },
                {
                    "title": "Masked Autoencoders Are Scalable Vision Learners",
                    "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3\u00d7 or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior."
                },
                {
                    "title": "Residual Relaxation for Multi-view Representation Learning",
                    "abstract": "Multi-view methods learn representations by aligning multiple views of the same image and their performance largely depends on the choice of data augmentation. In this paper, we notice that some other useful augmentations, such as image rotation, are harmful for multi-view methods because they cause a semantic shift that is too large to be aligned well. This observation motivates us to relax the exact alignment objective to better cultivate stronger augmentations. Taking image rotation as a case study, we develop a generic approach, Pretext-aware Residual Relaxation (Prelax), that relaxes the exact alignment by allowing an adaptive residual vector between different views and encoding the semantic shift through pretext-aware learning. Extensive experiments on different backbones show that our method can not only improve multi-view methods with existing augmentations, but also benefit from stronger image augmentations like rotation."
                },
                {
                    "title": "BEiT: BERT Pre-Training of Image Transformers",
                    "abstract": "We introduce a self-supervised vision representation model BEiT, which stands for Bidirectional Encoder representation from Image Transformers. Following BERT developed in the natural language processing area, we propose a masked image modeling task to pretrain vision Transformers. Specifically, each image has two views in our pre-training, i.e, image patches (such as 16x16 pixels), and visual tokens (i.e., discrete tokens). We first\"tokenize\"the original image into visual tokens. Then we randomly mask some image patches and fed them into the backbone Transformer. The pre-training objective is to recover the original visual tokens based on the corrupted image patches. After pre-training BEiT, we directly fine-tune the model parameters on downstream tasks by appending task layers upon the pretrained encoder. Experimental results on image classification and semantic segmentation show that our model achieves competitive results with previous pre-training methods. For example, base-size BEiT achieves 83.2% top-1 accuracy on ImageNet-1K, significantly outperforming from-scratch DeiT training (81.8%) with the same setup. Moreover, large-size BEiT obtains 86.3% only using ImageNet-1K, even outperforming ViT-L with supervised pre-training on ImageNet-22K (85.2%). The code and pretrained models are available at https://aka.ms/beit."
                },
                {
                    "title": "Provable Guarantees for Self-Supervised Deep Learning with Spectral Contrastive Loss",
                    "abstract": "Recent works in self-supervised learning have advanced the state-of-the-art by relying on the contrastive learning paradigm, which learns representations by pushing positive pairs, or similar examples from the same class, closer together while keeping negative pairs far apart. Despite the empirical successes, theoretical foundations are limited -- prior analyses assume conditional independence of the positive pairs given the same class label, but recent empirical applications use heavily correlated positive pairs (i.e., data augmentations of the same image). Our work analyzes contrastive learning without assuming conditional independence of positive pairs using a novel concept of the augmentation graph on data. Edges in this graph connect augmentations of the same data, and ground-truth classes naturally form connected sub-graphs. We propose a loss that performs spectral decomposition on the population augmentation graph and can be succinctly written as a contrastive learning objective on neural net representations. Minimizing this objective leads to features with provable accuracy guarantees under linear probe evaluation. By standard generalization bounds, these accuracy guarantees also hold when minimizing the training contrastive loss. Empirically, the features learned by our objective can match or outperform several strong baselines on benchmark vision datasets. In all, this work provides the first provable analysis for contrastive learning where guarantees for linear probe evaluation can apply to realistic empirical settings."
                },
                {
                    "title": "Emerging Properties in Self-Supervised Vision Transformers",
                    "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base."
                },
                {
                    "title": "Training data-efficient image transformers & distillation through attention",
                    "abstract": "Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption. In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data. More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models."
                },
                {
                    "title": "Taming Transformers for High-Resolution Image Synthesis",
                    "abstract": "Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks. In contrast to CNNs, they contain no inductive bias that prioritizes local interactions. This makes them expressive, but also computationally infeasible for long sequences, such as high-resolution images. We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images. We show how to (i) use CNNs to learn a context-rich vocabulary of image constituents, and in turn (ii) utilize transformers to efficiently model their composition within high-resolution images. Our approach is readily applied to conditional synthesis tasks, where both non-spatial information, such as object classes, and spatial information, such as segmentations, can control the generated image. In particular, we present the first results on semantically-guided synthesis of megapixel images with transformers. Project page at https://git.io/JLlvY."
                },
                {
                    "title": "Exploring Simple Siamese Representation Learning",
                    "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code is made available.1"
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning",
                    "abstract": "We introduce Bootstrap Your Own Latent (BYOL), a new approach to self-supervised image representation learning. BYOL relies on two neural networks, referred to as online and target networks, that interact and learn from each other. From an augmented view of an image, we train the online network to predict the target network representation of the same image under a different augmented view. At the same time, we update the target network with a slow-moving average of the online network. While state-of-the art methods rely on negative pairs, BYOL achieves a new state of the art without them. BYOL reaches $74.3\\%$ top-1 classification accuracy on ImageNet using a linear evaluation with a ResNet-50 architecture and $79.6\\%$ with a larger ResNet. We show that BYOL performs on par or better than the current state of the art on both transfer and semi-supervised benchmarks. Our implementation and pretrained models are given on GitHub."
                },
                {
                    "title": "A Simple Framework for Contrastive Learning of Visual Representations",
                    "abstract": "This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels."
                },
                {
                    "title": "Human Uncertainty Makes Classification More Robust",
                    "abstract": "The classification performance of deep neural networks has begun to asymptote at near-perfect levels. However, their ability to generalize outside the training set and their robustness to adversarial attacks have not. In this paper, we make progress on this problem by training with full label distributions that reflect human perceptual uncertainty. We first present a new benchmark dataset which we call CIFAR10H, containing a full distribution of human labels for each image of the CIFAR10 test set. We then show that, while contemporary classifiers fail to exhibit human-like uncertainty on their own, explicit training on our dataset closes this gap, supports improved generalization to increasingly out-of-training-distribution test datasets, and confers robustness to adversarial attacks."
                },
                {
                    "title": "DVAE#: Discrete Variational Autoencoders with Relaxed Boltzmann Priors",
                    "abstract": "Boltzmann machines are powerful distributions that have been shown to be an effective prior over binary latent variables in variational autoencoders (VAEs). However, previous methods for training discrete VAEs have used the evidence lower bound and not the tighter importance-weighted bound. We propose two approaches for relaxing Boltzmann machines to continuous distributions that permit training with importance-weighted bounds. These relaxations are based on generalized overlapping transformations and the Gaussian integral trick. Experiments on the MNIST and OMNIGLOT datasets show that these relaxations outperform previous discrete VAEs with Boltzmann priors. An implementation which reproduces these results is available at this https URL ."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "k-means++: the advantages of careful seeding",
                    "abstract": "The k-means method is a widely used clustering technique that seeks to minimize the average squared distance between points in the same cluster. Although it offers no accuracy guarantees, its simplicity and speed are very appealing in practice. By augmenting k-means with a very simple, randomized seeding technique, we obtain an algorithm that is \u0398(logk)-competitive with the optimal clustering. Preliminary experiments show that our augmentation improves both the speed and the accuracy of k-means, often quite dramatically."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nWhat is the role of tokenization in masked image modeling (MIM) and how does it affect downstream generalization?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it can lead to a deeper understanding of how different tokenization schemes influence the performance of self-supervised learning methods, particularly in masked image modeling. By elucidating the relationship between tokenization and downstream task performance, this research could pave the way for more effective and efficient MIM techniques, ultimately advancing the field of machine learning. Furthermore, improved tokenization strategies could enhance practical applications in computer vision, such as image classification, object detection, and segmentation, leading to better-performing models in real-world scenarios.\n\n### [Question 3] - Why is it hard?\nThe challenges in addressing this problem stem from the complexity of understanding the interactions between tokenization schemes and the underlying data distributions. Naive approaches may fail because they do not account for the nuanced relationships between unmasked views and their corresponding masked outputs, which can lead to suboptimal performance. Additionally, the lack of a standardized metric to evaluate the quality of tokenizers complicates the comparison of different methods. Theoretical obstacles include the need to establish a clear framework for analyzing the impact of tokenization on connectivity within the augmentation graph, while practical challenges involve designing and implementing effective tokenizers that align well with data classes.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has largely overlooked the specific impact of tokenization on MIM performance, focusing instead on the overall effectiveness of various MIM methods without dissecting the role of tokenizers. Existing solutions may have been limited by a lack of understanding of the implicit correlations created by masking and tokenization. Barriers include the absence of a robust metric for evaluating token quality and the complexity of analyzing the interplay between tokenization and downstream performance. This research aims to fill these gaps by introducing the token-class alignment similarity (TCAS) metric, which allows for a direct comparison of tokenizer quality, and by proposing a novel tokenizer design that enhances MIM performance.\n\n### [Question 5] - What are the key components of my approach and results?\nThe proposed methodology involves leveraging the graph perspective of masked image modeling to analyze the influence of different discrete tokenization schemes on downstream generalization. The approach includes the development of the token-class alignment similarity (TCAS) metric to evaluate the quality of tokenizers without requiring training. The dataset will consist"
            }
        },
        "author_data": {
            "1f37e513-aa0f-442c-abee-a8100d4b72b8": {
                "pk": "1f37e513-aa0f-442c-abee-a8100d4b72b8",
                "name": "Tianqi Du",
                "collaborators": [
                    "Yisen Wang",
                    "Yifei Wang",
                    "Qi Zhang",
                    "Weiran Huang",
                    "Hao Huang",
                    "Yihan Chen",
                    "Ning Zhang",
                    "Xuyang Zhao",
                    "Jun Yao",
                    "Jiansheng Yang",
                    "Zhouchen Lin",
                    "Xiaojun Guo"
                ],
                "domain": [
                    "Self-Supervised Learning",
                    "Contrastive Learning",
                    "Generative Models",
                    "Graph Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Look Ahead or Look Around? A Theoretical Comparison Between Autoregressive and Masked Pretraining",
                        "abstract": "In recent years, the rise of generative self-supervised learning (SSL) paradigms has exhibited impressive performance across visual, language, and multi-modal domains. While the varied designs of generative SSL objectives lead to distinct properties in downstream tasks, a theoretical understanding of these differences remains largely unexplored. In this paper, we establish the first theoretical comparisons between two leading generative SSL paradigms: autoregressive SSL and masked SSL. Through establishing theoretical frameworks, we elucidate the strengths and limitations of autoregressive and masked SSL within the primary evaluation tasks of classification and content generation. Our findings demonstrate that in classification tasks, the flexibility of targeted tokens in masked SSL fosters more inter-sample connections compared to the fixed position of target tokens in autoregressive SSL, which yields superior clustering performance. In content generation tasks, the misalignment between the flexible lengths of test samples and the fixed length of unmasked texts in masked SSL (vs. flexible lengths of conditional texts in autoregressive SSL) hinders its generation performance. To leverage each other's strengths and mitigate weaknesses, we propose diversity-enhanced autoregressive and variable-length masked objectives, which substantially improve the classification performance of autoregressive SSL and the generation performance of masked SSL. Code is available at https://github.com/PKU-ML/LookAheadLookAround."
                    },
                    {
                        "title": "Analysis and Evaluation of Species Invasion",
                        "abstract": "Invasive species, like dandelions, threaten biodiversity and bring economic and ecological risks. It is important to evaluate whether they are beneficial and their interactions with local species. This article stimulates the occurrence of invasion, evaluates the influencing factors of three species, and finally provides their invasive ranking. We used the Vensim model to dynamically simulate the intrusion process and supplemented it with an Agent Based Model (ABS). Under all other conditions, we found that for every 10% increase in birth rate, the number of dandelions increases by nearly 20%. We construct an Analytical Hierarchy Process (AHP) model to calculate the impact factor for different species. Our models have passed sensitivity tests. The model deepens our understanding of the dynamic process of species invasion and provides a theoretical framework for the identification and control of invasive species. In the future we will improve the models by collecting more comprehensive meteorological data to demonstrate the impact of seasons on seeding processes and try to displace number of Google search results to surveys on ecology experts to promote the scientific basis of our impact factor."
                    },
                    {
                        "title": "ArCL: Enhancing Contrastive Learning with Augmentation-Robust Representations",
                        "abstract": "Self-Supervised Learning (SSL) is a paradigm that leverages unlabeled data for model training. Empirical studies show that SSL can achieve promising performance in distribution shift scenarios, where the downstream and training distributions differ. However, the theoretical understanding of its transferability remains limited. In this paper, we develop a theoretical framework to analyze the transferability of self-supervised contrastive learning, by investigating the impact of data augmentation on it. Our results reveal that the downstream performance of contrastive learning depends largely on the choice of data augmentation. Moreover, we show that contrastive learning fails to learn domain-invariant features, which limits its transferability. Based on these theoretical insights, we propose a novel method called Augmentation-robust Contrastive Learning (ArCL), which guarantees to learn domain-invariant features and can be easily integrated with existing contrastive learning algorithms. We conduct experiments on several datasets and show that ArCL significantly improves the transferability of contrastive learning."
                    },
                    {
                        "title": "A Message Passing Perspective on Learning Dynamics of Contrastive Learning",
                        "abstract": "In recent years, contrastive learning achieves impressive results on self-supervised visual representation learning, but there still lacks a rigorous understanding of its learning dynamics. In this paper, we show that if we cast a contrastive objective equivalently into the feature space, then its learning dynamics admits an interpretable form. Specifically, we show that its gradient descent corresponds to a specific message passing scheme on the corresponding augmentation graph. Based on this perspective, we theoretically characterize how contrastive learning gradually learns discriminative features with the alignment update and the uniformity update. Meanwhile, this perspective also establishes an intriguing connection between contrastive learning and Message Passing Graph Neural Networks (MP-GNNs). This connection not only provides a unified understanding of many techniques independently developed in each community, but also enables us to borrow techniques from MP-GNNs to design new contrastive learning variants, such as graph attention, graph rewiring, jumpy knowledge techniques, etc. We believe that our message passing perspective not only provides a new theoretical understanding of contrastive learning dynamics, but also bridges the two seemingly independent areas together, which could inspire more interleaving studies to benefit from each other. The code is available at https://github.com/PKU-ML/Message-Passing-Contrastive-Learning."
                    },
                    {
                        "title": "ContraNorm: A Contrastive Learning Perspective on Oversmoothing and Beyond",
                        "abstract": "Oversmoothing is a common phenomenon in a wide range of Graph Neural Networks (GNNs) and Transformers, where performance worsens as the number of layers increases. Instead of characterizing oversmoothing from the view of complete collapse in which representations converge to a single point, we dive into a more general perspective of dimensional collapse in which representations lie in a narrow cone. Accordingly, inspired by the effectiveness of contrastive learning in preventing dimensional collapse, we propose a novel normalization layer called ContraNorm. Intuitively, ContraNorm implicitly shatters representations in the embedding space, leading to a more uniform distribution and a slighter dimensional collapse. On the theoretical analysis, we prove that ContraNorm can alleviate both complete collapse and dimensional collapse under certain conditions. Our proposed normalization layer can be easily integrated into GNNs and Transformers with negligible parameter overhead. Experiments on various real-world datasets demonstrate the effectiveness of our proposed ContraNorm. Our implementation is available at https://github.com/PKU-ML/ContraNorm."
                    },
                    {
                        "title": "Variational Energy-Based Models: A Probabilistic Framework for Contrastive Self-Supervised Learning",
                        "abstract": "Contrastive self-supervised learning has achieved impressive results in various scenarios. However, there still lacks a probabilistic framework for the principled design of the learning objective. Starting from the energy-based model, we design a variational energy-based model (VEM) for the self-supervised learning setting whose learning objective equals the InfoNCE loss via some slight approximations. Following that, we propose two sampling variants, VEM-Langevin and VEM-SVGD, for crafting better negative samples to help contrastive learning. Extensive experiments show that our VEM variants with SimCLR and MoCo indeed promote the performance on a wide range of inference tasks."
                    }
                ]
            },
            "ff6a20df-c286-4ae4-a843-bcb4c544a043": {
                "pk": "ff6a20df-c286-4ae4-a843-bcb4c544a043",
                "name": "Yifei Wang",
                "collaborators": [
                    "Yisen Wang",
                    "Qi Zhang",
                    "Zeming Wei",
                    "Qixun Wang",
                    "Xianghua Ying",
                    "Tianqi Du",
                    "Hao Huang",
                    "Yaoyu Guo",
                    "Wenhan Ma",
                    "Stefanie Jegelka",
                    "Jizhe Zhang",
                    "Chen Liu",
                    "Hongyu Zang",
                    "Xin Li",
                    "Yong Heng",
                    "Zhen Fang",
                    "Mingzhong Wang",
                    "Xiaojun Guo",
                    "Ang Li",
                    "Yiwen Guo",
                    "Liangchen Li",
                    "Jiansheng Yang",
                    "Zhouchen Lin",
                    "Jiangyan Ma"
                ],
                "domain": [
                    "Self-Supervised Learning",
                    "Contrastive Learning",
                    "Adversarial Training",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Can In-context Learning Really Generalize to Out-of-distribution Tasks?",
                        "abstract": "In this work, we explore the mechanism of in-context learning (ICL) on out-of-distribution (OOD) tasks that were not encountered during training. To achieve this, we conduct synthetic experiments where the objective is to learn OOD mathematical functions through ICL using a GPT-2 model. We reveal that Transformers may struggle to learn OOD task functions through ICL. Specifically, ICL performance resembles implementing a function within the pretraining hypothesis space and optimizing it with gradient descent based on the in-context examples. Additionally, we investigate ICL's well-documented ability to learn unseen abstract labels in context. We demonstrate that such ability only manifests in the scenarios without distributional shifts and, therefore, may not serve as evidence of new-task-learning ability. Furthermore, we assess ICL's performance on OOD tasks when the model is pretrained on multiple tasks. Both empirical and theoretical analyses demonstrate the existence of the \\textbf{low-test-error preference} of ICL, where it tends to implement the pretraining function that yields low test error in the testing context. We validate this through numerical experiments. This new theoretical result, combined with our empirical findings, elucidates the mechanism of ICL in addressing OOD tasks."
                    },
                    {
                        "title": "Look Ahead or Look Around? A Theoretical Comparison Between Autoregressive and Masked Pretraining",
                        "abstract": "In recent years, the rise of generative self-supervised learning (SSL) paradigms has exhibited impressive performance across visual, language, and multi-modal domains. While the varied designs of generative SSL objectives lead to distinct properties in downstream tasks, a theoretical understanding of these differences remains largely unexplored. In this paper, we establish the first theoretical comparisons between two leading generative SSL paradigms: autoregressive SSL and masked SSL. Through establishing theoretical frameworks, we elucidate the strengths and limitations of autoregressive and masked SSL within the primary evaluation tasks of classification and content generation. Our findings demonstrate that in classification tasks, the flexibility of targeted tokens in masked SSL fosters more inter-sample connections compared to the fixed position of target tokens in autoregressive SSL, which yields superior clustering performance. In content generation tasks, the misalignment between the flexible lengths of test samples and the fixed length of unmasked texts in masked SSL (vs. flexible lengths of conditional texts in autoregressive SSL) hinders its generation performance. To leverage each other's strengths and mitigate weaknesses, we propose diversity-enhanced autoregressive and variable-length masked objectives, which substantially improve the classification performance of autoregressive SSL and the generation performance of masked SSL. Code is available at https://github.com/PKU-ML/LookAheadLookAround."
                    },
                    {
                        "title": "Non-negative Contrastive Learning",
                        "abstract": "Deep representations have shown promising performance when transferred to downstream tasks in a black-box manner. Yet, their inherent lack of interpretability remains a significant challenge, as these features are often opaque to human understanding. In this paper, we propose Non-negative Contrastive Learning (NCL), a renaissance of Non-negative Matrix Factorization (NMF) aimed at deriving interpretable features. The power of NCL lies in its enforcement of non-negativity constraints on features, reminiscent of NMF's capability to extract features that align closely with sample clusters. NCL not only aligns mathematically well with an NMF objective but also preserves NMF's interpretability attributes, resulting in a more sparse and disentangled representation compared to standard contrastive learning (CL). Theoretically, we establish guarantees on the identifiability and downstream generalization of NCL. Empirically, we show that these advantages enable NCL to outperform CL significantly on feature disentanglement, feature selection, as well as downstream classification tasks. At last, we show that NCL can be easily extended to other learning scenarios and benefit supervised learning as well. Code is available at https://github.com/PKU-ML/non_neg."
                    },
                    {
                        "title": "How to Craft Backdoors with Unlabeled Data Alone?",
                        "abstract": "Relying only on unlabeled data, Self-supervised learning (SSL) can learn rich features in an economical and scalable way. As the drive-horse for building foundation models, SSL has received a lot of attention recently with wide applications, which also raises security concerns where backdoor attack is a major type of threat: if the released dataset is maliciously poisoned, backdoored SSL models can behave badly when triggers are injected to test samples. The goal of this work is to investigate this potential risk. We notice that existing backdoors all require a considerable amount of \\emph{labeled} data that may not be available for SSL. To circumvent this limitation, we explore a more restrictive setting called no-label backdoors, where we only have access to the unlabeled data alone, where the key challenge is how to select the proper poison set without using label information. We propose two strategies for poison selection: clustering-based selection using pseudolabels, and contrastive selection derived from the mutual information principle. Experiments on CIFAR-10 and ImageNet-100 show that both no-label backdoors are effective on many SSL methods and outperform random poisoning by a large margin. Code will be available at https://github.com/PKU-ML/nlb."
                    },
                    {
                        "title": "Do Generated Data Always Help Contrastive Learning?",
                        "abstract": "Contrastive Learning (CL) has emerged as one of the most successful paradigms for unsupervised visual representation learning, yet it often depends on intensive manual data augmentations. With the rise of generative models, especially diffusion models, the ability to generate realistic images close to the real data distribution has been well recognized. These generated high-equality images have been successfully applied to enhance contrastive representation learning, a technique termed ``data inflation''. However, we find that the generated data (even from a good diffusion model like DDPM) may sometimes even harm contrastive learning. We investigate the causes behind this failure from the perspective of both data inflation and data augmentation. For the first time, we reveal the complementary roles that stronger data inflation should be accompanied by weaker augmentations, and vice versa. We also provide rigorous theoretical explanations for these phenomena via deriving its generalization bounds under data inflation. Drawing from these insights, we propose Adaptive Inflation (AdaInf), a purely data-centric strategy without introducing any extra computation cost. On benchmark datasets, AdaInf can bring significant improvements for various contrastive learning methods. Notably, without using external data, AdaInf obtains 94.70% linear accuracy on CIFAR-10 with SimCLR, setting a new record that surpasses many sophisticated methods. Code is available at https://github.com/PKU-ML/adainf."
                    },
                    {
                        "title": "Towards Control-Centric Representations in Reinforcement Learning from Images",
                        "abstract": "Image-based Reinforcement Learning is a practical yet challenging task. A major hurdle lies in extracting control-centric representations while disregarding irrelevant information. While approaches that follow the bisimulation principle exhibit the potential in learning state representations to address this issue, they still grapple with the limited expressive capacity of latent dynamics and the inadaptability to sparse reward environments. To address these limitations, we introduce ReBis, which aims to capture control-centric information by integrating reward-free control information alongside reward-specific knowledge. ReBis utilizes a transformer architecture to implicitly model the dynamics and incorporates block-wise masking to eliminate spatiotemporal redundancy. Moreover, ReBis combines bisimulation-based loss with asymmetric reconstruction loss to prevent feature collapse in environments with sparse rewards. Empirical studies on two large benchmarks, including Atari games and DeepMind Control Suit, demonstrate that ReBis has superior performance compared to existing methods, proving its effectiveness."
                    },
                    {
                        "title": "Identifiable Contrastive Learning with Automatic Feature Importance Discovery",
                        "abstract": "Existing contrastive learning methods rely on pairwise sample contrast $z_x^\\top z_{x'}$ to learn data representations, but the learned features often lack clear interpretability from a human perspective. Theoretically, it lacks feature identifiability and different initialization may lead to totally different features. In this paper, we study a new method named tri-factor contrastive learning (triCL) that involves a 3-factor contrast in the form of $z_x^\\top S z_{x'}$, where $S=\\text{diag}(s_1,\\dots,s_k)$ is a learnable diagonal matrix that automatically captures the importance of each feature. We show that by this simple extension, triCL can not only obtain identifiable features that eliminate randomness but also obtain more interpretable features that are ordered according to the importance matrix $S$. We show that features with high importance have nice interpretability by capturing common classwise features, and obtain superior performance when evaluated for image retrieval using a few features. The proposed triCL objective is general and can be applied to different contrastive learning methods like SimCLR and CLIP. We believe that it is a better alternative to existing 2-factor contrastive learning by improving its identifiability and interpretability with minimal overhead. Code is available at https://github.com/PKU-ML/Tri-factor-Contrastive-Learning."
                    },
                    {
                        "title": "Architecture Matters: Uncovering Implicit Mechanisms in Graph Contrastive Learning",
                        "abstract": "With the prosperity of contrastive learning for visual representation learning (VCL), it is also adapted to the graph domain and yields promising performance. However, through a systematic study of various graph contrastive learning (GCL) methods, we observe that some common phenomena among existing GCL methods that are quite different from the original VCL methods, including 1) positive samples are not a must for GCL; 2) negative samples are not necessary for graph classification, neither for node classification when adopting specific normalization modules; 3) data augmentations have much less influence on GCL, as simple domain-agnostic augmentations (e.g., Gaussian noise) can also attain fairly good performance. By uncovering how the implicit inductive bias of GNNs works in contrastive learning, we theoretically provide insights into the above intriguing properties of GCL. Rather than directly porting existing VCL methods to GCL, we advocate for more attention toward the unique architecture of graph learning and consider its implicit influence when designing GCL methods. Code is available at https: //github.com/PKU-ML/ArchitectureMattersGCL."
                    },
                    {
                        "title": "Jailbreak and Guard Aligned Language Models with Only Few In-Context Demonstrations",
                        "abstract": "Large Language Models (LLMs) have shown remarkable success in various tasks, yet their safety and the risk of generating harmful content remain pressing concerns. In this paper, we delve into the potential of In-Context Learning (ICL) to modulate the alignment of LLMs. Specifically, we propose the In-Context Attack (ICA) which employs harmful demonstrations to subvert LLMs, and the In-Context Defense (ICD) which bolsters model resilience through examples that demonstrate refusal to produce harmful responses. We offer theoretical insights to elucidate how a limited set of in-context demonstrations can pivotally influence the safety alignment of LLMs. Through extensive experiments, we demonstrate the efficacy of ICA and ICD in respectively elevating and mitigating the success rates of jailbreaking prompts. Our findings illuminate the profound influence of ICL on LLM behavior, opening new avenues for improving the safety of LLMs."
                    },
                    {
                        "title": "Adversarial Examples Are Not Real Features",
                        "abstract": "The existence of adversarial examples has been a mystery for years and attracted much interest. A well-known theory by \\citet{ilyas2019adversarial} explains adversarial vulnerability from a data perspective by showing that one can extract non-robust features from adversarial examples and these features alone are useful for classification. However, the explanation remains quite counter-intuitive since non-robust features are mostly noise features to humans. In this paper, we re-examine the theory from a larger context by incorporating multiple learning paradigms. Notably, we find that contrary to their good usefulness under supervised learning, non-robust features attain poor usefulness when transferred to other self-supervised learning paradigms, such as contrastive learning, masked image modeling, and diffusion models. It reveals that non-robust features are not really as useful as robust or natural features that enjoy good transferability between these paradigms. Meanwhile, for robustness, we also show that naturally trained encoders from robust features are largely non-robust under AutoAttack. Our cross-paradigm examination suggests that the non-robust features are not really useful but more like paradigm-wise shortcuts, and robust features alone might be insufficient to attain reliable model robustness. Code is available at \\url{https://github.com/PKU-ML/AdvNotRealFeatures}."
                    },
                    {
                        "title": "Balance, Imbalance, and Rebalance: Understanding Robust Overfitting from a Minimax Game Perspective",
                        "abstract": "Adversarial Training (AT) has become arguably the state-of-the-art algorithm for extracting robust features. However, researchers recently notice that AT suffers from severe robust overfitting problems, particularly after learning rate (LR) decay. In this paper, we explain this phenomenon by viewing adversarial training as a dynamic minimax game between the model trainer and the attacker. Specifically, we analyze how LR decay breaks the balance between the minimax game by empowering the trainer with a stronger memorization ability, and show such imbalance induces robust overfitting as a result of memorizing non-robust features. We validate this understanding with extensive experiments, and provide a holistic view of robust overfitting from the dynamics of both the two game players. This understanding further inspires us to alleviate robust overfitting by rebalancing the two players by either regularizing the trainer's capacity or improving the attack strength. Experiments show that the proposed ReBalanced Adversarial Training (ReBAT) can attain good robustness and does not suffer from robust overfitting even after very long training. Code is available at https://github.com/PKU-ML/ReBAT."
                    },
                    {
                        "title": "Laplacian Canonization: A Minimalist Approach to Sign and Basis Invariant Spectral Embedding",
                        "abstract": "Spectral embedding is a powerful graph embedding technique that has received a lot of attention recently due to its effectiveness on Graph Transformers. However, from a theoretical perspective, the universal expressive power of spectral embedding comes at the price of losing two important invariance properties of graphs, sign and basis invariance, which also limits its effectiveness on graph data. To remedy this issue, many previous methods developed costly approaches to learn new invariants and suffer from high computation complexity. In this work, we explore a minimal approach that resolves the ambiguity issues by directly finding canonical directions for the eigenvectors, named Laplacian Canonization (LC). As a pure pre-processing method, LC is light-weighted and can be applied to any existing GNNs. We provide a thorough investigation, from theory to algorithm, on this approach, and discover an efficient algorithm named Maximal Axis Projection (MAP) that works for both sign and basis invariance and successfully canonizes more than 90% of all eigenvectors. Experiments on real-world benchmark datasets like ZINC, MOLTOX21, and MOLPCBA show that MAP consistently outperforms existing methods while bringing minimal computation overhead. Code is available at https://github.com/PKU-ML/LaplacianCanonization."
                    }
                ]
            },
            "5052d91d-9d86-416e-888a-e91f47cdf0c4": {
                "pk": "5052d91d-9d86-416e-888a-e91f47cdf0c4",
                "name": "Yisen Wang",
                "collaborators": [
                    "Yifei Wang",
                    "Stefanie Jegelka",
                    "Zeming Wei",
                    "Qi Zhang",
                    "George Ma",
                    "Derek Lim",
                    "Qixun Wang",
                    "Xianghua Ying",
                    "Yuyang Wu",
                    "Tianqi Du",
                    "Hao Huang",
                    "Yaoyu Guo",
                    "Wenhan Ma",
                    "Zhaoyang Liu",
                    "Chenheng Zhang",
                    "Jinyang Gao",
                    "Bolin Ding",
                    "Xiang Pan",
                    "Qi Lei",
                    "Jizhe Zhang",
                    "Yichuan Mo",
                    "Yuji Wang",
                    "Chen Liu",
                    "Hongyu Zang",
                    "Xin Li",
                    "Yong Heng",
                    "Zhen Fang",
                    "Mingzhong Wang",
                    "Xiaojun Guo",
                    "Ang Li",
                    "Yiwen Guo",
                    "Liangchen Li",
                    "Jiansheng Yang",
                    "Zhouchen Lin",
                    "Jiangyan Ma"
                ],
                "domain": [
                    "Machine Learning",
                    "Self-Supervised Learning",
                    "Adversarial Training",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "A Canonicalization Perspective on Invariant and Equivariant Learning",
                        "abstract": "In many applications, we desire neural networks to exhibit invariance or equivariance to certain groups due to symmetries inherent in the data. Recently, frame-averaging methods emerged to be a unified framework for attaining symmetries efficiently by averaging over input-dependent subsets of the group, i.e., frames. What we currently lack is a principled understanding of the design of frames. In this work, we introduce a canonicalization perspective that provides an essential and complete view of the design of frames. Canonicalization is a classic approach for attaining invariance by mapping inputs to their canonical forms. We show that there exists an inherent connection between frames and canonical forms. Leveraging this connection, we can efficiently compare the complexity of frames as well as determine the optimality of certain frames. Guided by this principle, we design novel frames for eigenvectors that are strictly superior to existing methods -- some are even optimal -- both theoretically and empirically. The reduction to the canonicalization perspective further uncovers equivalences between previous methods. These observations suggest that canonicalization provides a fundamental understanding of existing frame-averaging methods and unifies existing equivariant and invariant learning methods. Code is available at https://github.com/GeorgeMLP/canonicalization."
                    },
                    {
                        "title": "Can In-context Learning Really Generalize to Out-of-distribution Tasks?",
                        "abstract": "In this work, we explore the mechanism of in-context learning (ICL) on out-of-distribution (OOD) tasks that were not encountered during training. To achieve this, we conduct synthetic experiments where the objective is to learn OOD mathematical functions through ICL using a GPT-2 model. We reveal that Transformers may struggle to learn OOD task functions through ICL. Specifically, ICL performance resembles implementing a function within the pretraining hypothesis space and optimizing it with gradient descent based on the in-context examples. Additionally, we investigate ICL's well-documented ability to learn unseen abstract labels in context. We demonstrate that such ability only manifests in the scenarios without distributional shifts and, therefore, may not serve as evidence of new-task-learning ability. Furthermore, we assess ICL's performance on OOD tasks when the model is pretrained on multiple tasks. Both empirical and theoretical analyses demonstrate the existence of the \\textbf{low-test-error preference} of ICL, where it tends to implement the pretraining function that yields low test error in the testing context. We validate this through numerical experiments. This new theoretical result, combined with our empirical findings, elucidates the mechanism of ICL in addressing OOD tasks."
                    },
                    {
                        "title": "A Theoretical Understanding of Self-Correction through In-context Alignment",
                        "abstract": "Going beyond mimicking limited human experiences, recent studies show initial evidence that, like humans, large language models (LLMs) are capable of improving their abilities purely by self-correction, i.e., correcting previous responses through self-examination, in certain circumstances. Nevertheless, little is known about how such capabilities arise. In this work, based on a simplified setup akin to an alignment task, we theoretically analyze self-correction from an in-context learning perspective, showing that when LLMs give relatively accurate self-examinations as rewards, they are capable of refining responses in an in-context way. Notably, going beyond previous theories on over-simplified linear transformers, our theoretical construction underpins the roles of several key designs of realistic transformers for self-correction: softmax attention, multi-head attention, and the MLP block. We validate these findings extensively on synthetic datasets. Inspired by these findings, we also illustrate novel applications of self-correction, such as defending against LLM jailbreaks, where a simple self-correction step does make a large difference. We believe that these findings will inspire further research on understanding, exploiting, and enhancing self-correction for building better foundation models."
                    },
                    {
                        "title": "Look Ahead or Look Around? A Theoretical Comparison Between Autoregressive and Masked Pretraining",
                        "abstract": "In recent years, the rise of generative self-supervised learning (SSL) paradigms has exhibited impressive performance across visual, language, and multi-modal domains. While the varied designs of generative SSL objectives lead to distinct properties in downstream tasks, a theoretical understanding of these differences remains largely unexplored. In this paper, we establish the first theoretical comparisons between two leading generative SSL paradigms: autoregressive SSL and masked SSL. Through establishing theoretical frameworks, we elucidate the strengths and limitations of autoregressive and masked SSL within the primary evaluation tasks of classification and content generation. Our findings demonstrate that in classification tasks, the flexibility of targeted tokens in masked SSL fosters more inter-sample connections compared to the fixed position of target tokens in autoregressive SSL, which yields superior clustering performance. In content generation tasks, the misalignment between the flexible lengths of test samples and the fixed length of unmasked texts in masked SSL (vs. flexible lengths of conditional texts in autoregressive SSL) hinders its generation performance. To leverage each other's strengths and mitigate weaknesses, we propose diversity-enhanced autoregressive and variable-length masked objectives, which substantially improve the classification performance of autoregressive SSL and the generation performance of masked SSL. Code is available at https://github.com/PKU-ML/LookAheadLookAround."
                    },
                    {
                        "title": "Non-negative Contrastive Learning",
                        "abstract": "Deep representations have shown promising performance when transferred to downstream tasks in a black-box manner. Yet, their inherent lack of interpretability remains a significant challenge, as these features are often opaque to human understanding. In this paper, we propose Non-negative Contrastive Learning (NCL), a renaissance of Non-negative Matrix Factorization (NMF) aimed at deriving interpretable features. The power of NCL lies in its enforcement of non-negativity constraints on features, reminiscent of NMF's capability to extract features that align closely with sample clusters. NCL not only aligns mathematically well with an NMF objective but also preserves NMF's interpretability attributes, resulting in a more sparse and disentangled representation compared to standard contrastive learning (CL). Theoretically, we establish guarantees on the identifiability and downstream generalization of NCL. Empirically, we show that these advantages enable NCL to outperform CL significantly on feature disentanglement, feature selection, as well as downstream classification tasks. At last, we show that NCL can be easily extended to other learning scenarios and benefit supervised learning as well. Code is available at https://github.com/PKU-ML/non_neg."
                    },
                    {
                        "title": "How to Craft Backdoors with Unlabeled Data Alone?",
                        "abstract": "Relying only on unlabeled data, Self-supervised learning (SSL) can learn rich features in an economical and scalable way. As the drive-horse for building foundation models, SSL has received a lot of attention recently with wide applications, which also raises security concerns where backdoor attack is a major type of threat: if the released dataset is maliciously poisoned, backdoored SSL models can behave badly when triggers are injected to test samples. The goal of this work is to investigate this potential risk. We notice that existing backdoors all require a considerable amount of \\emph{labeled} data that may not be available for SSL. To circumvent this limitation, we explore a more restrictive setting called no-label backdoors, where we only have access to the unlabeled data alone, where the key challenge is how to select the proper poison set without using label information. We propose two strategies for poison selection: clustering-based selection using pseudolabels, and contrastive selection derived from the mutual information principle. Experiments on CIFAR-10 and ImageNet-100 show that both no-label backdoors are effective on many SSL methods and outperform random poisoning by a large margin. Code will be available at https://github.com/PKU-ML/nlb."
                    },
                    {
                        "title": "What is Wrong with Perplexity for Long-context Language Modeling?",
                        "abstract": "Handling long-context inputs is crucial for large language models (LLMs) in tasks such as extended conversations, document summarization, and many-shot in-context learning. While recent approaches have extended the context windows of LLMs and employed perplexity (PPL) as a standard evaluation metric, PPL has proven unreliable for assessing long-context capabilities. The underlying cause of this limitation has remained unclear. In this work, we provide a comprehensive explanation for this issue. We find that PPL overlooks key tokens, which are essential for long-context understanding, by averaging across all tokens and thereby obscuring the true performance of models in long-context scenarios. To address this, we propose \\textbf{LongPPL}, a novel metric that focuses on key tokens by employing a long-short context contrastive method to identify them. Our experiments demonstrate that LongPPL strongly correlates with performance on various long-context benchmarks (e.g., Pearson correlation of -0.96), significantly outperforming traditional PPL in predictive accuracy. Additionally, we introduce \\textbf{LongCE} (Long-context Cross-Entropy) loss, a re-weighting strategy for fine-tuning that prioritizes key tokens, leading to consistent improvements across diverse benchmarks. In summary, these contributions offer deeper insights into the limitations of PPL and present effective solutions for accurately evaluating and enhancing the long-context capabilities of LLMs. Code is available at https://github.com/PKU-ML/LongPPL."
                    },
                    {
                        "title": "Beyond Interpretability: The Gains of Feature Monosemanticity on Model Robustness",
                        "abstract": "Deep learning models often suffer from a lack of interpretability due to polysemanticity, where individual neurons are activated by multiple unrelated semantics, resulting in unclear attributions of model behavior. Recent advances in monosemanticity, where neurons correspond to consistent and distinct semantics, have significantly improved interpretability but are commonly believed to compromise accuracy. In this work, we challenge the prevailing belief of the accuracy-interpretability tradeoff, showing that monosemantic features not only enhance interpretability but also bring concrete gains in model performance. Across multiple robust learning scenarios-including input and label noise, few-shot learning, and out-of-domain generalization-our results show that models leveraging monosemantic features significantly outperform those relying on polysemantic features. Furthermore, we provide empirical and theoretical understandings on the robustness gains of feature monosemanticity. Our preliminary analysis suggests that monosemanticity, by promoting better separation of feature representations, leads to more robust decision boundaries. This diverse evidence highlights the generality of monosemanticity in improving model robustness. As a first step in this new direction, we embark on exploring the learning benefits of monosemanticity beyond interpretability, supporting the long-standing hypothesis of linking interpretability and robustness. Code is available at \\url{https://github.com/PKU-ML/Beyond_Interpretability}."
                    },
                    {
                        "title": "Do Generated Data Always Help Contrastive Learning?",
                        "abstract": "Contrastive Learning (CL) has emerged as one of the most successful paradigms for unsupervised visual representation learning, yet it often depends on intensive manual data augmentations. With the rise of generative models, especially diffusion models, the ability to generate realistic images close to the real data distribution has been well recognized. These generated high-equality images have been successfully applied to enhance contrastive representation learning, a technique termed ``data inflation''. However, we find that the generated data (even from a good diffusion model like DDPM) may sometimes even harm contrastive learning. We investigate the causes behind this failure from the perspective of both data inflation and data augmentation. For the first time, we reveal the complementary roles that stronger data inflation should be accompanied by weaker augmentations, and vice versa. We also provide rigorous theoretical explanations for these phenomena via deriving its generalization bounds under data inflation. Drawing from these insights, we propose Adaptive Inflation (AdaInf), a purely data-centric strategy without introducing any extra computation cost. On benchmark datasets, AdaInf can bring significant improvements for various contrastive learning methods. Notably, without using external data, AdaInf obtains 94.70% linear accuracy on CIFAR-10 with SimCLR, setting a new record that surpasses many sophisticated methods. Code is available at https://github.com/PKU-ML/adainf."
                    },
                    {
                        "title": "Fight Back Against Jailbreaking via Prompt Adversarial Tuning",
                        "abstract": "While Large Language Models (LLMs) have achieved tremendous success in various applications, they are also susceptible to jailbreaking attacks. Several primary defense strategies have been proposed to protect LLMs from producing harmful information, mostly focusing on model fine-tuning or heuristical defense designs. However, how to achieve intrinsic robustness through prompt optimization remains an open problem. In this paper, motivated by adversarial training paradigms for achieving reliable robustness, we propose an approach named Prompt Adversarial Tuning (PAT) that trains a prompt control attached to the user prompt as a guard prefix. To achieve our defense goal whilst maintaining natural performance, we optimize the control prompt with both adversarial and benign prompts. Comprehensive experiments show that our method is effective against both grey-box and black-box attacks, reducing the success rate of advanced attacks to nearly 0%, while maintaining the model's utility on the benign task and incurring only negligible computational overhead, charting a new perspective for future explorations in LLM security. Our code is available at https://github.com/PKU-ML/PAT."
                    },
                    {
                        "title": "Towards Control-Centric Representations in Reinforcement Learning from Images",
                        "abstract": "Image-based Reinforcement Learning is a practical yet challenging task. A major hurdle lies in extracting control-centric representations while disregarding irrelevant information. While approaches that follow the bisimulation principle exhibit the potential in learning state representations to address this issue, they still grapple with the limited expressive capacity of latent dynamics and the inadaptability to sparse reward environments. To address these limitations, we introduce ReBis, which aims to capture control-centric information by integrating reward-free control information alongside reward-specific knowledge. ReBis utilizes a transformer architecture to implicitly model the dynamics and incorporates block-wise masking to eliminate spatiotemporal redundancy. Moreover, ReBis combines bisimulation-based loss with asymmetric reconstruction loss to prevent feature collapse in environments with sparse rewards. Empirical studies on two large benchmarks, including Atari games and DeepMind Control Suit, demonstrate that ReBis has superior performance compared to existing methods, proving its effectiveness."
                    },
                    {
                        "title": "Identifiable Contrastive Learning with Automatic Feature Importance Discovery",
                        "abstract": "Existing contrastive learning methods rely on pairwise sample contrast $z_x^\\top z_{x'}$ to learn data representations, but the learned features often lack clear interpretability from a human perspective. Theoretically, it lacks feature identifiability and different initialization may lead to totally different features. In this paper, we study a new method named tri-factor contrastive learning (triCL) that involves a 3-factor contrast in the form of $z_x^\\top S z_{x'}$, where $S=\\text{diag}(s_1,\\dots,s_k)$ is a learnable diagonal matrix that automatically captures the importance of each feature. We show that by this simple extension, triCL can not only obtain identifiable features that eliminate randomness but also obtain more interpretable features that are ordered according to the importance matrix $S$. We show that features with high importance have nice interpretability by capturing common classwise features, and obtain superior performance when evaluated for image retrieval using a few features. The proposed triCL objective is general and can be applied to different contrastive learning methods like SimCLR and CLIP. We believe that it is a better alternative to existing 2-factor contrastive learning by improving its identifiability and interpretability with minimal overhead. Code is available at https://github.com/PKU-ML/Tri-factor-Contrastive-Learning."
                    },
                    {
                        "title": "Architecture Matters: Uncovering Implicit Mechanisms in Graph Contrastive Learning",
                        "abstract": "With the prosperity of contrastive learning for visual representation learning (VCL), it is also adapted to the graph domain and yields promising performance. However, through a systematic study of various graph contrastive learning (GCL) methods, we observe that some common phenomena among existing GCL methods that are quite different from the original VCL methods, including 1) positive samples are not a must for GCL; 2) negative samples are not necessary for graph classification, neither for node classification when adopting specific normalization modules; 3) data augmentations have much less influence on GCL, as simple domain-agnostic augmentations (e.g., Gaussian noise) can also attain fairly good performance. By uncovering how the implicit inductive bias of GNNs works in contrastive learning, we theoretically provide insights into the above intriguing properties of GCL. Rather than directly porting existing VCL methods to GCL, we advocate for more attention toward the unique architecture of graph learning and consider its implicit influence when designing GCL methods. Code is available at https: //github.com/PKU-ML/ArchitectureMattersGCL."
                    },
                    {
                        "title": "Jailbreak and Guard Aligned Language Models with Only Few In-Context Demonstrations",
                        "abstract": "Large Language Models (LLMs) have shown remarkable success in various tasks, yet their safety and the risk of generating harmful content remain pressing concerns. In this paper, we delve into the potential of In-Context Learning (ICL) to modulate the alignment of LLMs. Specifically, we propose the In-Context Attack (ICA) which employs harmful demonstrations to subvert LLMs, and the In-Context Defense (ICD) which bolsters model resilience through examples that demonstrate refusal to produce harmful responses. We offer theoretical insights to elucidate how a limited set of in-context demonstrations can pivotally influence the safety alignment of LLMs. Through extensive experiments, we demonstrate the efficacy of ICA and ICD in respectively elevating and mitigating the success rates of jailbreaking prompts. Our findings illuminate the profound influence of ICL on LLM behavior, opening new avenues for improving the safety of LLMs."
                    },
                    {
                        "title": "Adversarial Examples Are Not Real Features",
                        "abstract": "The existence of adversarial examples has been a mystery for years and attracted much interest. A well-known theory by \\citet{ilyas2019adversarial} explains adversarial vulnerability from a data perspective by showing that one can extract non-robust features from adversarial examples and these features alone are useful for classification. However, the explanation remains quite counter-intuitive since non-robust features are mostly noise features to humans. In this paper, we re-examine the theory from a larger context by incorporating multiple learning paradigms. Notably, we find that contrary to their good usefulness under supervised learning, non-robust features attain poor usefulness when transferred to other self-supervised learning paradigms, such as contrastive learning, masked image modeling, and diffusion models. It reveals that non-robust features are not really as useful as robust or natural features that enjoy good transferability between these paradigms. Meanwhile, for robustness, we also show that naturally trained encoders from robust features are largely non-robust under AutoAttack. Our cross-paradigm examination suggests that the non-robust features are not really useful but more like paradigm-wise shortcuts, and robust features alone might be insufficient to attain reliable model robustness. Code is available at \\url{https://github.com/PKU-ML/AdvNotRealFeatures}."
                    },
                    {
                        "title": "Balance, Imbalance, and Rebalance: Understanding Robust Overfitting from a Minimax Game Perspective",
                        "abstract": "Adversarial Training (AT) has become arguably the state-of-the-art algorithm for extracting robust features. However, researchers recently notice that AT suffers from severe robust overfitting problems, particularly after learning rate (LR) decay. In this paper, we explain this phenomenon by viewing adversarial training as a dynamic minimax game between the model trainer and the attacker. Specifically, we analyze how LR decay breaks the balance between the minimax game by empowering the trainer with a stronger memorization ability, and show such imbalance induces robust overfitting as a result of memorizing non-robust features. We validate this understanding with extensive experiments, and provide a holistic view of robust overfitting from the dynamics of both the two game players. This understanding further inspires us to alleviate robust overfitting by rebalancing the two players by either regularizing the trainer's capacity or improving the attack strength. Experiments show that the proposed ReBalanced Adversarial Training (ReBAT) can attain good robustness and does not suffer from robust overfitting even after very long training. Code is available at https://github.com/PKU-ML/ReBAT."
                    },
                    {
                        "title": "Laplacian Canonization: A Minimalist Approach to Sign and Basis Invariant Spectral Embedding",
                        "abstract": "Spectral embedding is a powerful graph embedding technique that has received a lot of attention recently due to its effectiveness on Graph Transformers. However, from a theoretical perspective, the universal expressive power of spectral embedding comes at the price of losing two important invariance properties of graphs, sign and basis invariance, which also limits its effectiveness on graph data. To remedy this issue, many previous methods developed costly approaches to learn new invariants and suffer from high computation complexity. In this work, we explore a minimal approach that resolves the ambiguity issues by directly finding canonical directions for the eigenvectors, named Laplacian Canonization (LC). As a pure pre-processing method, LC is light-weighted and can be applied to any existing GNNs. We provide a thorough investigation, from theory to algorithm, on this approach, and discover an efficient algorithm named Maximal Axis Projection (MAP) that works for both sign and basis invariance and successfully canonizes more than 90% of all eigenvectors. Experiments on real-world benchmark datasets like ZINC, MOLTOX21, and MOLPCBA show that MAP consistently outperforms existing methods while bringing minimal computation overhead. Code is available at https://github.com/PKU-ML/LaplacianCanonization."
                    }
                ]
            }
        }
    },
    "2406.11118": {
        "paper_data": {
            "title": "Incentivizing Quality Text Generation via Statistical Contracts",
            "url": "http://arxiv.org/abs/2406.11118v1",
            "arxiv_id": "2406.11118",
            "authors": [
                "Eden Saig",
                "Ohad Einav",
                "Inbal Talgam-Cohen"
            ],
            "abstract": "While the success of large language models (LLMs) increases demand for machine-generated text, current pay-per-token pricing schemes create a misalignment of incentives known in economics as moral hazard: Text-generating agents have strong incentive to cut costs by preferring a cheaper model over the cutting-edge one, and this can be done \"behind the scenes\" since the agent performs inference internally. In this work, we approach this issue from an economic perspective, by proposing a pay-for-performance, contract-based framework for incentivizing quality. We study a principal-agent game where the agent generates text using costly inference, and the contract determines the principal's payment for the text according to an automated quality evaluation. Since standard contract theory is inapplicable when internal inference costs are unknown, we introduce cost-robust contracts. As our main theoretical contribution, we characterize optimal cost-robust contracts through a direct correspondence to optimal composite hypothesis tests from statistics, generalizing a result of Saig et al. (NeurIPS'23). We evaluate our framework empirically by deriving contracts for a range of objectives and LLM evaluation benchmarks, and find that cost-robust contracts sacrifice only a marginal increase in objective value compared to their cost-aware counterparts.",
            "introduction": "   1 Introduction  Modern-day LLMs are showing increasingly impressive capabilities, and simultaneously becoming increasingly costly. With rising success at handling complex tasks, conversational AI systems are seeing ubiquitous usage across critical domains such as healthcare [19, 35], financial risk assessment [28], and law [24, 36]. To achieve such levels of performance, contemporary LLM architectures contain billions and even trillions of parameters, leading to a computational pipeline that requires dedicated facilities and substantial energy to operate [34].   Since the computational requirements of modern LLMs exceed the capacity of common devices, language generation tasks are typically delegated to commercial firms (such as AWS and Microsoft Azure), which operate dedicated computational resources to generate text for a price. To address the tension between LLM performance and computational costs, such firms typically propose multiple service options, with each option offering a different trade-off between model quality and cost [1, 7, 37, 38]. Currently, the most common pricing scheme for such services is pay-per-token, in which users agree in advance to pay a fixed rate for each token of text generated by the system [10].   While simple and intuitive, the pay-per-token pricing scheme creates a misalignment of economic incentives between the firms and their consumers, known in the economic literature as moral hazard: As inference is performed internally and a fixed price is agreed upon in advance, rational firms can strategically increase their profit margin by generating text using a cheaper, lower-quality model. Due to the stochastic nature of language generation, consumers may not be able to reliably determine the quality of the model being used, exposing them to this kind of hazard.   Moral hazard is especially prevalent in cases where the text generation task is intricate, and so evaluation is hard: Consider a scenario where a healthcare provider uses conversational AI to summarize medical notes. As medical diagnosis is an intricate and critical task, the healthcare provider wishes the summaries to be generated using the most advanced language model. Under the pay-per-token pricing scheme, the healthcare provider agrees in advance to pay a fixed amount for each token generated. However, it is not hard to imagine that a firm may attempt to increase profit margins by routing some of the summarization requests to one of its cheaper models, instead of the most advanced one, without taking into account their purpose, and knowing that any lower-quality results would be attributed to the stochastic nature of LLM inference.   From pay-per-token to pay-for-performance.  In the economic literature, the canonical solution to moral hazard problems is pay-for-performance, or P4P [17]. Instead of paying a fixed price for any outcome, the parties agree in advance on a contract that specifies a differential payment scheme \u2013 e.g., agreeing in advance that the firm will receive higher pay when the generated text is considered to be high-quality. When designed correctly, contracts incentivize rational agents to invest more effort, thus providing a way to align incentives. Interaction around contracts is modeled as a principal-agent game, where the principal commits to a payment scheme, and the agent responds by rationally selecting a utility-maximizing action. Within this framework, the principal seeks to design a contract which satisfies some notion of optimality, such as requiring the",
            "references": [
                {
                    "title": "Clarabel: An interior-point solver for conic programs with quadratic objectives",
                    "abstract": "We present a general-purpose interior-point solver for convex optimization problems with conic constraints. Our method is based on a homogeneous embedding method originally developed for general monotone complementarity problems and more recently applied to operator splitting methods, and here specialized to an interior-point method for problems with quadratic objectives. We allow for a variety of standard symmetric and non-symmetric cones, and provide support for chordal decomposition methods in the case of semidefinite cones. We describe the implementation of this method in the open-source solver Clarabel, and provide a detailed numerical evaluation of its performance versus several state-of-the-art solvers on a wide range of standard benchmarks problems. Clarabel is faster and more robust than competing commercial and open-source solvers across a range of test sets, with a particularly large performance advantage for problems with quadratic objectives. Clarabel is currently distributed as a standard solver for the Python CVXPY optimization suite."
                },
                {
                    "title": "Algorithmic Persuasion Through Simulation: Information Design in the Age of Generative AI",
                    "abstract": "We study a Bayesian persuasion game where a sender wants to persuade a receiver to take a binary action, such as purchasing a product. The sender is informed about the (binary) state of the world, such as whether the quality of the product is high or low, but only has limited information about the receiver's beliefs and utilities. Motivated by customer surveys, user studies, and recent advances in AI, we allow the sender to learn more about the receiver by querying an oracle that simulates the receiver's behavior. After a fixed number of queries, the sender commits to a messaging policy and the receiver takes the action that maximizes her expected utility given the message she receives. We characterize the sender's optimal messaging policy given any distribution over receiver types. We then design a polynomial-time querying algorithm that optimizes the sender's expected utility in this game. We also consider approximate oracles, more general query structures, and costly queries."
                },
                {
                    "title": "Large Language Models in Law: A Survey",
                    "abstract": "The advent of artificial intelligence (AI) has significantly impacted the traditional judicial industry. Moreover, recently, with the development of AI-generated content (AIGC), AI and law have found applications in various domains, including image recognition, automatic text generation, and interactive chat. With the rapid emergence and growing popularity of large models, it is evident that AI will drive transformation in the traditional judicial industry. However, the application of legal large language models (LLMs) is still in its nascent stage. Several challenges need to be addressed. In this paper, we aim to provide a comprehensive survey of legal LLMs. We not only conduct an extensive survey of LLMs, but also expose their applications in the judicial system. We first provide an overview of AI technologies in the legal field and showcase the recent research in LLMs. Then, we discuss the practical implementation presented by legal LLMs, such as providing legal advice to users and assisting judges during trials. In addition, we explore the limitations of legal LLMs, including data, algorithms, and judicial practice. Finally, we summarize practical recommendations and propose future development directions to address these challenges."
                },
                {
                    "title": "A Survey on Detection of LLMs-Generated Content",
                    "abstract": "The burgeoning capabilities of advanced large language models (LLMs) such as ChatGPT have led to an increase in synthetic content generation with implications across a variety of sectors, including media, cybersecurity, public discourse, and education. As such, the ability to detect LLMs-generated content has become of paramount importance. We aim to provide a detailed overview of existing detection strategies and benchmarks, scrutinizing their differences and identifying key challenges and prospects in the field, advocating for more adaptable and robust models to enhance detection accuracy. We also posit the necessity for a multi-faceted approach to defend against various attacks to counter the rapidly advancing capabilities of LLMs. To the best of our knowledge, this work is the first comprehensive survey on the detection in the era of LLMs. We hope it will provide a broad understanding of the current landscape of LLMs-generated content detection, offering a guiding reference for researchers and practitioners striving to uphold the integrity of digital information in an era increasingly dominated by synthetic content. The relevant papers are summarized and will be consistently updated at https://github.com/Xianjun-Yang/Awesome_papers_on_LLMs_detection.git."
                },
                {
                    "title": "Mechanism Design for Large Language Models",
                    "abstract": "We investigate auction mechanisms to support the emerging format of AI-generated content. We in particular study how to aggregate several LLMs in an incentive compatible manner. In this problem, the preferences of each agent over stochastically generated contents are described/encoded as an LLM. A key motivation is to design an auction format for AI-generated ad creatives to combine inputs from different advertisers. We argue that this problem, while generally falling under the umbrella of mechanism design, has several unique features. We propose a general formalism---the token auction model---for studying this problem. A key feature of this model is that it acts on a token-by-token basis and lets LLM agents influence generated contents through single dimensional bids. We first explore a robust auction design approach, in which all we assume is that agent preferences entail partial orders over outcome distributions. We formulate two natural incentive properties, and show that these are equivalent to a monotonicity condition on distribution aggregation. We also show that for such aggregation functions, it is possible to design a second-price auction, despite the absence of bidder valuation functions. We then move to designing concrete aggregation functions by focusing on specific valuation forms based on KL-divergence, a commonly used loss function in LLM. The welfare-maximizing aggregation rules turn out to be the weighted (log-space) convex combination of the target distributions from all participants. We conclude with experimental results in support of the token auction formulation."
                },
                {
                    "title": "Generative Judge for Evaluating Alignment",
                    "abstract": "The rapid development of Large Language Models (LLMs) has substantially expanded the range of tasks they can address. In the field of Natural Language Processing (NLP), researchers have shifted their focus from conventional NLP tasks (e.g., sequence tagging and parsing) towards tasks that revolve around aligning with human needs (e.g., brainstorming and email writing). This shift in task distribution imposes new requirements on evaluating these aligned models regarding generality (i.e., assessing performance across diverse scenarios), flexibility (i.e., examining under different protocols), and interpretability (i.e., scrutinizing models with explanations). In this paper, we propose a generative judge with 13B parameters, Auto-J, designed to address these challenges. Our model is trained on user queries and LLM-generated responses under massive real-world scenarios and accommodates diverse evaluation protocols (e.g., pairwise response comparison and single-response evaluation) with well-structured natural language critiques. To demonstrate the efficacy of our approach, we construct a new testbed covering 58 different scenarios. Experimentally, Auto-J outperforms a series of strong competitors, including both open-source and closed-source models, by a large margin. We also provide detailed analysis and case studies to further reveal the potential of our method and make a variety of resources public at https://github.com/GAIR-NLP/auto-j."
                },
                {
                    "title": "A Survey of Large Language Models for Healthcare: from Data, Technology, and Applications to Accountability and Ethics",
                    "abstract": "The utilization of large language models (LLMs) in the Healthcare domain has generated both excitement and concern due to their ability to effectively respond to freetext queries with certain professional knowledge. This survey outlines the capabilities of the currently developed LLMs for Healthcare and explicates their development process, with the aim of providing an overview of the development roadmap from traditional Pretrained Language Models (PLMs) to LLMs. Specifically, we first explore the potential of LLMs to enhance the efficiency and effectiveness of various Healthcare applications highlighting both the strengths and limitations. Secondly, we conduct a comparison between the previous PLMs and the latest LLMs, as well as comparing various LLMs with each other. Then we summarize related Healthcare training data, training methods, optimization strategies, and usage. Finally, the unique concerns associated with deploying LLMs in Healthcare settings are investigated, particularly regarding fairness, accountability, transparency and ethics. Our survey provide a comprehensive investigation from perspectives of both computer science and Healthcare specialty. Besides the discussion about Healthcare concerns, we supports the computer science community by compiling a collection of open source resources, such as accessible datasets, the latest methodologies, code implementations, and evaluation benchmarks in the Github. Summarily, we contend that a significant paradigm shift is underway, transitioning from PLMs to LLMs. This shift encompasses a move from discriminative AI approaches to generative AI approaches, as well as a shift from model-centered methodologies to data-centered methodologies. Also, we determine that the biggest obstacle of using LLMs in Healthcare are fairness, accountability, transparency and ethics."
                },
                {
                    "title": "Large Language Models in Finance: A Survey",
                    "abstract": "Recent advances in large language models (LLMs) have opened new possibilities for artificial intelligence applications in finance. In this paper, we provide a practical survey focused on two key aspects of utilizing LLMs for financial tasks: existing solutions and guidance for adoption. First, we review current approaches employing LLMs in finance, including leveraging pretrained models via zero-shot or few-shot learning, fine-tuning on domain-specific data, and training custom LLMs from scratch. We summarize key models and evaluate their performance improvements on financial natural language processing tasks. Second, we propose a decision framework to guide financial professionals in selecting the appropriate LLM solution based on their use case constraints around data, compute, and performance needs. The framework provides a pathway from lightweight experimentation to heavy investment in customized LLMs. Lastly, we discuss limitations and challenges around leveraging LLMs in financial applications. Overall, this survey aims to synthesize the state-of-the-art and provide a roadmap for responsibly applying LLMs to advance financial AI."
                },
                {
                    "title": "From Words to Watts: Benchmarking the Energy Costs of Large Language Model Inference",
                    "abstract": "Large language models (LLMs) have exploded in popularity due to their new generative capabilities that go far beyond prior state-of-the-art. These technologies are increasingly being leveraged in various domains such as law, finance, and medicine. However, these models carry significant computational challenges, especially the compute and energy costs required for inference. Inference energy costs already receive less attention than the energy costs of training LLMs-despite how often these large models are called on to conduct inference in reality (e.g., ChatGPT). As these state-of-the-art LLMs see increasing usage and deployment in various domains, a better understanding of their resource utilization is crucial for cost-savings, scaling performance, efficient hardware usage, and optimal inference strategies. In this paper, we describe experiments conducted to study the computational and energy utilization of inference with LLMs. We benchmark and conduct a preliminary analysis of the inference performance and inference energy costs of different sizes of LLaMA-a recent state-of-the-art LLM-developed by Meta AI on two generations of popular GPUs (NVIDIA V100 & A100) and two datasets (Alpaca and GSM8K) to reflect the diverse set of tasks/benchmarks for LLMs in research and practice. We present the results of multi-node, multi-GPU inference using model sharding across up to 32 GPUs. To our knowledge, our work is the one of the first to study LLM inference performance from the perspective of computational and energy resources at this scale."
                },
                {
                    "title": "Delegating Data Collection in Decentralized Machine Learning",
                    "abstract": "Motivated by the emergence of decentralized machine learning (ML) ecosystems, we study the delegation of data collection. Taking the field of contract theory as our starting point, we design optimal and near-optimal contracts that deal with two fundamental information asymmetries that arise in decentralized ML: uncertainty in the assessment of model quality and uncertainty regarding the optimal performance of any model. We show that a principal can cope with such asymmetry via simple linear contracts that achieve 1-1/e fraction of the optimal utility. To address the lack of a priori knowledge regarding the optimal performance, we give a convex program that can adaptively and efficiently compute the optimal contract. We also study linear contracts and derive the optimal utility in the more complex setting of multiple interactions."
                },
                {
                    "title": "Generative Social Choice",
                    "abstract": "Traditionally, social choice theory has only been applicable to choices among a few predetermined alternatives but not to more complex decisions such as collectively selecting a textual statement. We introduce generative social choice, a framework that combines the mathematical rigor of social choice theory with the capability of large language models to generate text and extrapolate preferences. This framework divides the design of AI-augmented democratic processes into two components: first, proving that the process satisfies rigorous representation guarantees when given access to oracle queries; second, empirically validating that these queries can be approximately implemented using a large language model. We apply this framework to the problem of generating a slate of statements that is representative of opinions expressed as free-form text; specifically, we develop a democratic process with representation guarantees and use this process to represent the opinions of participants in a survey about chatbot personalization. We find that 93 out of 100 participants feel\"mostly\"or\"perfectly\"represented by the slate of five statements we extracted."
                },
                {
                    "title": "Code Llama: Open Foundation Models for Code",
                    "abstract": "We release Code Llama, a family of large language models for code based on Llama 2 providing state-of-the-art performance among open models, infilling capabilities, support for large input contexts, and zero-shot instruction following ability for programming tasks. We provide multiple flavors to cover a wide range of applications: foundation models (Code Llama), Python specializations (Code Llama - Python), and instruction-following models (Code Llama - Instruct) with 7B, 13B, 34B and 70B parameters each. All models are trained on sequences of 16k tokens and show improvements on inputs with up to 100k tokens. 7B, 13B and 70B Code Llama and Code Llama - Instruct variants support infilling based on surrounding content. Code Llama reaches state-of-the-art performance among open models on several code benchmarks, with scores of up to 67% and 65% on HumanEval and MBPP, respectively. Notably, Code Llama - Python 7B outperforms Llama 2 70B on HumanEval and MBPP, and all our models outperform every other publicly available model on MultiPL-E. We release Code Llama under a permissive license that allows for both research and commercial use."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "A Survey on Evaluation of Large Language Models",
                    "abstract": "Large language models (LLMs) are gaining increasing popularity in both academia and industry, owing to their unprecedented performance in various applications. As LLMs continue to play a vital role in both research and daily use, their evaluation becomes increasingly critical, not only at the task level, but also at the society level for better understanding of their potential risks. Over the past years, significant efforts have been made to examine LLMs from various perspectives. This paper presents a comprehensive review of these evaluation methods for LLMs, focusing on three key dimensions: what to evaluate, where to evaluate, and how to evaluate. Firstly, we provide an overview from the perspective of evaluation tasks, encompassing general natural language processing tasks, reasoning, medical usage, ethics, education, natural and social sciences, agent applications, and other areas. Secondly, we answer the \u2018where\u2019 and \u2018how\u2019 questions by diving into the evaluation methods and benchmarks, which serve as crucial components in assessing the performance of LLMs. Then, we summarize the success and failure cases of LLMs in different tasks. Finally, we shed light on several future challenges that lie ahead in LLMs evaluation. Our aim is to offer invaluable insights to researchers in the realm of LLMs evaluation, thereby aiding the development of more proficient LLMs. Our key point is that evaluation should be treated as an essential discipline to better assist the development of LLMs. We consistently maintain the related open-source materials at: https://github.com/MLGroupJLU/LLM-eval-survey"
                },
                {
                    "title": "Delegated Classification",
                    "abstract": "When machine learning is outsourced to a rational agent, conflicts of interest might arise and severely impact predictive performance. In this work, we propose a theoretical framework for incentive-aware delegation of machine learning tasks. We model delegation as a principal-agent game, in which accurate learning can be incentivized by the principal using performance-based contracts. Adapting the economic theory of contract design to this setting, we define budget-optimal contracts and prove they take a simple threshold form under reasonable assumptions. In the binary-action case, the optimality of such contracts is shown to be equivalent to the classic Neyman-Pearson lemma, establishing a formal connection between contract design and statistical hypothesis testing. Empirically, we demonstrate that budget-optimal contracts can be constructed using small-scale data, leveraging recent advances in the study of learning curves and scaling laws. Performance and economic outcomes are evaluated using synthetic and real-world classification tasks."
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "DNA-GPT: Divergent N-Gram Analysis for Training-Free Detection of GPT-Generated Text",
                    "abstract": "Large language models (LLMs) have notably enhanced the fluency and diversity of machine-generated text. However, this progress also presents a significant challenge in detecting the origin of a given text, and current research on detection methods lags behind the rapid evolution of LLMs. Conventional training-based methods have limitations in flexibility, particularly when adapting to new domains, and they often lack explanatory power. To address this gap, we propose a novel training-free detection strategy called Divergent N-Gram Analysis (DNA-GPT). Given a text, we first truncate it in the middle and then use only the preceding portion as input to the LLMs to regenerate the new remaining parts. By analyzing the differences between the original and new remaining parts through N-gram analysis in black-box or probability divergence in white-box, we unveil significant discrepancies between the distribution of machine-generated text and the distribution of human-written text. We conducted extensive experiments on the most advanced LLMs from OpenAI, including text-davinci-003, GPT-3.5-turbo, and GPT-4, as well as open-source models such as GPT-NeoX-20B and LLaMa-13B. Results show that our zero-shot approach exhibits state-of-the-art performance in distinguishing between human and GPT-generated text on four English and one German dataset, outperforming OpenAI's own classifier, which is trained on millions of text. Additionally, our methods provide reasonable explanations and evidence to support our claim, which is a unique feature of explainable detection. Our method is also robust under the revised text attack and can additionally solve model sourcing. Codes are available at https://github.com/Xianjun-Yang/DNA-GPT."
                },
                {
                    "title": "FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance",
                    "abstract": "There is a rapidly growing number of large language models (LLMs) that users can query for a fee. We review the cost associated with querying popular LLM APIs, e.g. GPT-4, ChatGPT, J1-Jumbo, and find that these models have heterogeneous pricing structures, with fees that can differ by two orders of magnitude. In particular, using LLMs on large collections of queries and text can be expensive. Motivated by this, we outline and discuss three types of strategies that users can exploit to reduce the inference cost associated with using LLMs: 1) prompt adaptation, 2) LLM approximation, and 3) LLM cascade. As an example, we propose FrugalGPT, a simple yet flexible instantiation of LLM cascade which learns which combinations of LLMs to use for different queries in order to reduce cost and improve accuracy. Our experiments show that FrugalGPT can match the performance of the best individual LLM (e.g. GPT-4) with up to 98% cost reduction or improve the accuracy over GPT-4 by 4% with the same cost. The ideas and findings presented here lay a foundation for using LLMs sustainably and efficiently."
                },
                {
                    "title": "Origin Tracing and Detecting of LLMs",
                    "abstract": "The extraordinary performance of large language models (LLMs) heightens the importance of detecting whether the context is generated by an AI system. More importantly, while more and more companies and institutions release their LLMs, the origin can be hard to trace. Since LLMs are heading towards the time of AGI, similar to the origin tracing in anthropology, it is of great importance to trace the origin of LLMs. In this paper, we first raise the concern of the origin tracing of LLMs and propose an effective method to trace and detect AI-generated contexts. We introduce a novel algorithm that leverages the contrastive features between LLMs and extracts model-wise features to trace the text origins. Our proposed method works under both white-box and black-box settings therefore can be widely generalized to detect various LLMs.(e.g. can be generalized to detect GPT-3 models without the GPT-3 models). Also, our proposed method requires only limited data compared with the supervised learning methods and can be extended to trace new-coming model origins. We construct extensive experiments to examine whether we can trace the origins of given texts. We provide valuable observations based on the experimental results, such as the difficulty level of AI origin tracing, and the AI origin similarities, and call for ethical concerns of LLM providers. We are releasing all codes and data as a toolkit and benchmark for future AI origin tracing and detecting studies. \\footnote{We are releasing all available resource at \\url{https://github.com/OpenLMLab/}.}"
                },
                {
                    "title": "G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment",
                    "abstract": "The quality of texts generated by natural language generation (NLG) systems is hard to measure automatically. Conventional reference-based metrics, such as BLEU and ROUGE, have been shown to have relatively low correlation with human judgments, especially for tasks that require creativity and diversity. Recent studies suggest using large language models (LLMs) as reference-free metrics for NLG evaluation, which have the benefit of being applicable to new tasks that lack human references. However, these LLM-based evaluators still have lower human correspondence than medium-size neural evaluators. In this work, we present G-Eval, a framework of using large language models with chain-of-thoughts (CoT) and a form-filling paradigm, to assess the quality of NLG outputs. We experiment with two generation tasks, text summarization and dialogue generation. We show that G-Eval with GPT-4 as the backbone model achieves a Spearman correlation of 0.514 with human on summarization task, outperforming all previous methods by a large margin. We also propose preliminary analysis on the behavior of LLM-based evaluators, and highlight the potential issue of LLM-based evaluators having a bias towards the LLM-generated texts. The code is at https://github.com/nlpyang/geval"
                },
                {
                    "title": "A Short Survey of Viewing Large Language Models in Legal Aspect",
                    "abstract": "Large language models (LLMs) have transformed many fields, including natural language processing, computer vision, and reinforcement learning. These models have also made a significant impact in the field of law, where they are being increasingly utilized to automate various legal tasks, such as legal judgement prediction, legal document analysis, and legal document writing. However, the integration of LLMs into the legal field has also raised several legal problems, including privacy concerns, bias, and explainability. In this survey, we explore the integration of LLMs into the field of law. We discuss the various applications of LLMs in legal tasks, examine the legal challenges that arise from their use, and explore the data resources that can be used to specialize LLMs in the legal domain. Finally, we discuss several promising directions and conclude this paper. By doing so, we hope to provide an overview of the current state of LLMs in law and highlight the potential benefits and challenges of their integration."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Principal-Agent Hypothesis Testing",
                    "abstract": "Consider the relationship between a regulator (the principal) and an experimenter (the agent) such as a pharmaceutical company. The pharmaceutical company wishes to sell a drug for profit, whereas the regulator wishes to allow only efficacious drugs to be marketed. The efficacy of the drug is not known to the regulator, so the pharmaceutical company must run a costly trial to prove efficacy to the regulator. Critically, the statistical protocol used to establish efficacy affects the behavior of a strategic, self-interested agent; a lower standard of statistical evidence incentivizes the agent to run more trials that are less likely to be effective. The interaction between the statistical protocol and the incentives of the pharmaceutical company is crucial for understanding this system and designing protocols with high social utility. In this work, we discuss how the regulator can set up a protocol with payoffs based on statistical evidence. We show how to design protocols that are robust to an agent's strategic actions, and derive the optimal protocol in the presence of strategic entrants."
                },
                {
                    "title": "Program Synthesis with Large Language Models",
                    "abstract": "This paper explores the limits of the current generation of large language models for program synthesis in general purpose programming languages. We evaluate a collection of such models (with between 244M and 137B parameters) on two new benchmarks, MBPP and MathQA-Python, in both the few-shot and fine-tuning regimes. Our benchmarks are designed to measure the ability of these models to synthesize short Python programs from natural language descriptions. The Mostly Basic Programming Problems (MBPP) dataset contains 974 programming tasks, designed to be solvable by entry-level programmers. The MathQA-Python dataset, a Python version of the MathQA benchmark, contains 23914 problems that evaluate the ability of the models to synthesize code from more complex text. On both datasets, we find that synthesis performance scales log-linearly with model size. Our largest models, even without finetuning on a code dataset, can synthesize solutions to 59.6 percent of the problems from MBPP using few-shot learning with a well-designed prompt. Fine-tuning on a held-out portion of the dataset improves performance by about 10 percentage points across most model sizes. On the MathQA-Python dataset, the largest fine-tuned model achieves 83.8 percent accuracy. Going further, we study the model's ability to engage in dialog about code, incorporating human feedback to improve its solutions. We find that natural language feedback from a human halves the error rate compared to the model's initial prediction. Additionally, we conduct an error analysis to shed light on where these models fall short and what types of programs are most difficult to generate. Finally, we explore the semantic grounding of these models by fine-tuning them to predict the results of program execution. We find that even our best models are generally unable to predict the output of a program given a specific input."
                },
                {
                    "title": "Evaluating Large Language Models Trained on Code",
                    "abstract": "We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics."
                },
                {
                    "title": "Authorship Attribution for Neural Text Generation",
                    "abstract": "In recent years, the task of generating realistic short and long texts have made tremendous advancements. In particular, several recently proposed neural network-based language models have demonstrated their astonishing capabilities to generate texts that are challenging to distinguish from human-written texts with the naked eye. Despite many benefits and utilities of such neural methods, in some applications, being able to tell the \u201cauthor\u201d of a text in question becomes critically important. In this work, in the context of this Turing Test, we investigate the so-called authorship attribution problem in three versions: (1) given two texts T1 and T2, are both generated by the same method or not? (2) is the given text T written by a human or machine? (3) given a text T and k candidate neural methods, can we single out the method (among k alternatives) that generated T? Against one humanwritten and eight machine-generated texts (i.e., CTRL, GPT, GPT2, GROVER, XLM, XLNET, PPLM, FAIR), we empirically experiment with the performance of various models in three problems. By and large, we find that most generators still generate texts significantly different from human-written ones, thereby making three problems easier to solve. However, the qualities of texts generated by GPT2, GROVER, and FAIR are better, often confusing machine classifiers in solving three problems. All codes and datasets of our experiments are available at: https://bit.ly/ 302zWdz"
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Simple versus Optimal Contracts",
                    "abstract": "We consider the classic principal-agent model of contract theory, in which a principal designs an outcome-dependent compensation scheme to incentivize an agent to take a costly and unobservable action. When all of the model parameters---including the full distribution over principal rewards resulting from each agent action---are known to the designer, an optimal contract can in principle be computed by linear programming. In addition to their demanding informational requirements, however, such optimal contracts are often complex and unintuitive, and do not resemble contracts used in practice. This paper examines contract theory through the theoretical computer science lens, with the goal of developing novel theory to explain and justify the prevalence of relatively simple contracts, such as linear (pure commission) contracts. First, we consider the case where the principal knows only the first moment of each action's reward distribution, and we prove that linear contracts are guaranteed to be worst-case optimal, ranging over all reward distributions consistent with the given moments. Second, we study linear contracts from a worst-case approximation perspective, and prove several tight parameterized approximation bounds."
                },
                {
                    "title": "A Rewriting System for Convex Optimization Problems",
                    "abstract": "We describe a modular rewriting system for translating optimization problems written in a domain-specific language to forms compatible with low-level solver interfaces. Translation is facilitated by reductions, which accept a category of problems and transform instances of that category to equivalent instances of another category. Our system proceeds in two key phases: analysis, in which we attempt to find a suitable solver for a supplied problem, and canonicalization, in which we rewrite the problem in the selected solver's standard form. We implement the described system in version 1.0 of CVXPY, a domain-specific language for mathematical and especially convex optimization. By treating reductions as first-class objects, our method makes it easy to match problems to solvers well-suited for them and to support solvers with a wide variety of standard forms."
                },
                {
                    "title": "CVXPY: A Python-Embedded Modeling Language for Convex Optimization",
                    "abstract": "CVXPY is a domain-specific language for convex optimization embedded in Python. It allows the user to express convex optimization problems in a natural syntax that follows the math, rather than in the restrictive standard form required by solvers. CVXPY makes it easy to combine convex optimization with high-level features of Python such as parallelism and object-oriented design. CVXPY is available at http://www.cvxpy.org/ under the GPL license, along with documentation and examples."
                },
                {
                    "title": "Robustness and Linear Contracts",
                    "abstract": "We consider a moral hazard problem where the principal is uncertain as to what the agent can and cannot do: she knows some actions available to the agent, but other, unknown actions may also exist. The principal demands robustness, evaluating possible contracts by their worst-case performance, over unknown actions the agent might potentially take. The model assumes risk-neutrality and limited liability, and no other functional form assumptions. Very generally, the optimal contract is linear. The model thus offers a new explanation for linear contracts in practice. It also introduces a flexible modeling approach for moral hazard under nonquantifiable uncertainty. (JEL D81, D82, D86)"
                },
                {
                    "title": "Adaptive contract design for crowdsourcing markets: bandit algorithms for repeated principal-agent problems",
                    "abstract": "Crowdsourcing markets have emerged as a popular platform for matching available workers with tasks to complete. The payment for a particular task is typically set by the task's requester, and may be adjusted based on the quality of the completed work, for example, through the use of 'bonus' payments. In this paper, we study the requester's problem of dynamically adjusting quality-contingent payments for tasks. We consider a multi-round version of the well-known principal-agent model, whereby in each round a worker makes a strategic choice of the effort level which is not directly observable by the requester. In particular, our formulation significantly generalizes the budget-free online task pricing problems studied in prior work. We treat this problem as a multi-armed bandit problem, with each 'arm' representing a potential contract. To cope with the large (and in fact, infinite) number of arms, we propose a new algorithm, AgnosticZooming, which discretizes the contract space into a finite number of regions, effectively treating each region as a single arm. This discretization is adaptively refined, so that more promising regions of the contract space are eventually discretized more finely. We provide a full analysis of this algorithm, showing that it achieves regret sublinear in the time horizon and substantially improves over non-adaptive discretization (which is the only competing approach in the literature)."
                },
                {
                    "title": "Pay for Performance: An Overview of the Literature",
                    "abstract": "Institute of Medicine (To Err Is Human, 1999, and Crossing the Quality Chasm, 2001), the medical community has searched for possible ways to improve the method by which health care is delivered in the United States. Unfortunately, no permanent solution has been implemented. Great strides are under way in the area of quality improvement, specifically with the recent development of pay for performance (P4P). The objective of this annotated bibliography is to offer a critical analysis of the current literature and an overview of the topic of P4P to educate the reader about a health care issue that may become a cornerstone of the payment system. Pay for performance is one of the newest methods of medical compensation, combining reimbursement with quality improvement. Health care providers receive a base payment and, with the achievement of certain quality benchmarks for process measures (care provided) or outcome measures (result of patient care), providers receive certain rewards. These may include incentive payments and providerspecific quality ratings. The ratings are transparent to consumers and could influence the consumer\u2019s choice of provider. The incentives encourage physicians and patients alike to achieve the standards of care outlined in performance measures. For example, a physician receives $1.00 per patient per month for every patient with hypertension but receives an extra $0.50 per patient per month if hypertension is controlled in 80% of these patients. The recording of performance is accomplished by reporting the care provided to the payer (ie, the insurance company) and may be facilitated by electronic medical records and electronic reporting. The quality of care administered to patients may be determined by certain evidence-based performance measures. It is presumed that, with these improvement measures, the quality of medicine practiced will increase, providing a potential solution to the health care problems that plague this country. This improved care has the potential to produce healthier patients who incur lower costs, consume fewer medical resources, and have a better quality of life. As a result, the number of P4P programs (and supporting literature) continues to rapidly increase. A predecessor to American P4P is an incentivebased quality improvement program implemented in the United Kingdom in 2004 called the Quality and Outcomes Framework (QOF). Several of the articles referenced in this bibliography are based on this program, which is practically identical to P4P apart from the name. The QOF provides a general idea about successes and failures of a national centralized system. It may be predictive of the sustainability of P4P in the United States. According to the Leapfrog Group (a quality improvement organization), there were 148 P4P programs nationwide as of March 2007; that number is rising quickly. Some of the notable P4P implementers are the Integrated Healthcare Association in California, the Leapfrog Group, and (most important) the Centers for Medicare & Medicaid Services (CMS). The CMS has trial P4P programs and a Physician Quality Reporting Initiative (PQRI) that gives bonus payments for simply reporting provided care, regardless of the quality of care. Therefore, the PQRI serves as a template for widely implemented P4P. Based on the success of its trial programs, the CMS may implement P4P nationwide within the next few"
                },
                {
                    "title": "Combinatorial agency",
                    "abstract": "Much recent research concerns systems, such as the Internet, whose components are owned and operated by different parties, each with his own \"selfish\" goal. The field of Algorithmic Mechanism Design handles the issue of private information held by the different parties in such computational settings. This paper deals with a complementary problem in such settings: handling the \"hidden actions\" that are performed by the different parties.Our model is a combinatorial variant of the classical principalagent problem from economic theory. In our setting a principal must motivate a team of strategic agents to exert costly effort on his behalf, but their actions are hidden from him. Our focus is on cases where complex combinations of the efforts of the agents influence the outcome. The principal motivates the agents by offering to them a set of contracts, which together put the agents in an equilibrium point of the induced game. We present formal models for this setting, suggest and embark on an analysis of some basic issues, but leave many questions open."
                },
                {
                    "title": "The Economics of Contracts: A Primer",
                    "abstract": "The theory of contracts grew out of the failure of the general equilibrium model to account for the strategic interactions among agents that arise from informational asymmetries. This popular text, revised and updated throughout for the second edition, serves as a concise and rigorous introduction to the theory of contracts for graduate students and professional economists. The book presents the main models of the theory of contracts, particularly the basic models of adverse selection, signaling, and moral hazard. It emphasizes the methods used to analyze the models, but also includes brief introductions to many of the applications in different fields of economics. The goal is to give readers the tools to understand the basic models and create their own. For the second edition, major changes have been made to chapter 3, on examples and extensions for the adverse selection model, which now includes more thorough discussions of multiprincipals, collusion, and multidimensional adverse selection, and to chapter 5, on moral hazard, with the limited liability model, career concerns, and common agency added to its topics. Two chapters have been completely rewritten: chapter 7, on the theory of incomplete contracts, and chapter 8, on the empirical literature in the theory of contracts. An appendix presents concepts of noncooperative game theory to supplement chapters 4 and 6. Exercises follow chapters 2 through 5. Praise for the previous edition: \"The Economics of Contracts offers an excellent introduction to agency models. Written by one of the leading young researchers in contact theory, it is rigorous, clear, concise, and up-to-date. Researchers and students who want to learn about the economics of incentives will want to read this primer.\" -- Jean Tirole, Institut D'Economie Industrielle, Universite des Sciences Sociales, France \"Students will find this a very useful introduction to the ideas of contract theory. Salanie has managed to summarize a large amount of material in a relatively short number of pages in a highly accessible and readable manner.\" -- Oliver Hart, Professor of Economics, Harvard University"
                },
                {
                    "title": "Moral Hazard and Observability",
                    "abstract": "The role of imperfect information in a principal-agent relationship subject to moral hazard is considered. A necessary and sufficient condition for imperfect information to improve on contracts based on the payoff alone is derived, and a characterization of the optimal use of such information is given."
                }
            ],
            "categories": [
                "cs.GT",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we design an effective pay-for-performance (P4P) pricing model for language generation services that aligns the incentives of service providers with the quality expectations of consumers, thereby mitigating the moral hazard associated with pay-per-token pricing?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses a significant gap in the economic models applied to AI service pricing. By developing a P4P model, we can enhance the reliability and quality of language generation services, which are increasingly used in critical domains like healthcare and finance. This advancement could lead to more trustworthy AI applications, fostering greater adoption and innovation in these fields. Furthermore, it could inspire future research on incentive structures in AI and other domains, ultimately leading to improved service delivery and consumer satisfaction.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in accurately measuring the quality of generated text and designing a contract that effectively incentivizes firms to prioritize high-quality outputs. Naive approaches may fail because they might not account for the stochastic nature of language generation, making it difficult to establish a fair and transparent evaluation metric. Additionally, technical obstacles include the need for robust mechanisms to assess quality in real-time and the theoretical complexity of modeling the principal-agent dynamics in this context. There is also the practical challenge of implementing such a system in existing commercial frameworks.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on the mechanics of language models and their performance metrics, often overlooking the economic implications of pricing strategies. Existing solutions have been limited by a lack of comprehensive frameworks that integrate quality assessment with incentive alignment. Barriers include the difficulty in quantifying text quality and the absence of established methodologies for creating effective P4P contracts in AI services. Our approach differs by explicitly incorporating quality metrics into the pricing model and leveraging game-theoretic principles to design contracts that align incentives more effectively.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a P4P pricing model that incorporates quality assessment metrics for generated text. We will utilize a dataset of language generation tasks across various domains, including healthcare and finance, to evaluate model performance. The key metrics will include user satisfaction scores and automated quality assessments based on established benchmarks. We expect the outcomes to demonstrate that a well-designed P4P model can significantly improve the quality of generated text"
            }
        },
        "author_data": {
            "781bb0fb-be62-4727-addb-976d7866191e": {
                "pk": "781bb0fb-be62-4727-addb-976d7866191e",
                "name": "Eden Saig",
                "collaborators": [
                    "Nir Rosenfeld",
                    "Inbal Talgam-Cohen"
                ],
                "domain": [
                    "Recommendation Systems",
                    "Game Theory",
                    "Machine Learning",
                    "Digital Well-being"
                ],
                "publications": [
                    {
                        "title": "Learning to Suggest Breaks: Sustainable Optimization of Long-Term User Engagement",
                        "abstract": "Optimizing user engagement is a key goal for modern recommendation systems, but blindly pushing users towards increased consumption risks burn-out, churn, or even addictive habits. To promote digital well-being, most platforms now offer a service that periodically prompts users to take breaks. These, however, must be set up manually, and so may be suboptimal for both users and the system. In this paper, we study the role of breaks in recommendation, and propose a framework for learning optimal breaking policies that promote and sustain long-term engagement. Based on the notion that recommendation dynamics are susceptible to both positive and negative feedback, we cast recommendation as a Lotka-Volterra dynamical system, where breaking reduces to a problem of optimal control. We then give an efficient learning algorithm, provide theoretical guarantees, and empirically demonstrate the utility of our approach on semi-synthetic data."
                    },
                    {
                        "title": "Delegated Classification",
                        "abstract": "When machine learning is outsourced to a rational agent, conflicts of interest might arise and severely impact predictive performance. In this work, we propose a theoretical framework for incentive-aware delegation of machine learning tasks. We model delegation as a principal-agent game, in which accurate learning can be incentivized by the principal using performance-based contracts. Adapting the economic theory of contract design to this setting, we define budget-optimal contracts and prove they take a simple threshold form under reasonable assumptions. In the binary-action case, the optimality of such contracts is shown to be equivalent to the classic Neyman-Pearson lemma, establishing a formal connection between contract design and statistical hypothesis testing. Empirically, we demonstrate that budget-optimal contracts can be constructed using small-scale data, leveraging recent advances in the study of learning curves and scaling laws. Performance and economic outcomes are evaluated using synthetic and real-world classification tasks."
                    }
                ]
            },
            "253fc188-b1d4-4b87-bee8-2e052a62ea81": {
                "pk": "253fc188-b1d4-4b87-bee8-2e052a62ea81",
                "name": "Ohad Einav",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "9615516e-a3b9-4898-b1d6-559f0c37a0ac": {
                "pk": "9615516e-a3b9-4898-b1d6-559f0c37a0ac",
                "name": "Inbal Talgam-Cohen",
                "collaborators": [
                    "Tim Roughgarden",
                    "Uriel Feige",
                    "Moshe Babaioff",
                    "Noam Nisan",
                    "Michal Feldman",
                    "Mukund Sundararajan",
                    "Paul D\u00fctting",
                    "Jan Vondr\u00e1k",
                    "Paul Duetting",
                    "Shahar Dobzinski"
                ],
                "domain": [
                    "Game Theory",
                    "Mechanism Design",
                    "Auction Theory",
                    "Economic Theory"
                ],
                "publications": [
                    {
                        "title": "A Direct Reduction from k-Player to 2-Player Approximate Nash Equilibrium",
                        "abstract": "We present a direct reduction from k-player games to 2-player games that preserves approximate Nash equilibrium. Previously, the computational equivalence of computing approximate Nash equilibrium in k-player and 2-player games was established via an indirect reduction. This included a sequence of works defining the complexity class PPAD, identifying complete problems for this class, showing that computing approximate Nash equilibrium for k-player games is in PPAD, and reducing a PPAD-complete problem to computing approximate Nash equilibrium for 2-player games. Our direct reduction makes no use of the concept of PPAD, thus eliminating some of the difficulties involved in following the known indirect reduction."
                    },
                    {
                        "title": "Refine Predictions Ad Infinitum?",
                        "abstract": "We study how standard auction objectives in sponsored search markets change with refinements in the prediction of the relevance (click-through rates) of ads. We study mechanisms that optimize for a convex combination of efficiency and revenue. We show that the objective function of such a mechanism can only improve with refined (improved) relevance predictions, i.e., the search engine has no disincentive to perform these refinements. More interestingly, we show that under assumptions, refinements to relevance predictions can only improve the efficiency of any such mechanism. Our main technical contribution is to study how relevance refinements affect the similarity between ranking by virtual-value (revenue ranking) and ranking by value (efficiency ranking). Finally, we discuss implications of our results to the literature on signaling."
                    },
                    {
                        "title": "Approximately Optimal Mechanism Design",
                        "abstract": "Optimal mechanism design enjoys a beautiful and well-developed theory, and also a number of killer applications. Rules of thumb produced by the field influence everything from how governments sell wireless spectrum licenses to how the major search engines auction off online advertising. There are, however, some basic problems for which the traditional optimal mechanism design approach is ill-suited---either because it makes overly strong assumptions, or because it advocates overly complex designs. This survey reviews several common issues with optimal mechanisms, including exorbitant communication, computation, and informational requirements; and it presents several examples demonstrating that passing to the relaxed goal of an approximately optimal mechanism allows us to reason about fundamental questions that seem out of reach of the traditional theory."
                    },
                    {
                        "title": "Simple versus Optimal Contracts",
                        "abstract": "We consider the classic principal-agent model of contract theory, in which a principal designs an outcome-dependent compensation scheme to incentivize an agent to take a costly and unobservable action. When all of the model parameters---including the full distribution over principal rewards resulting from each agent action---are known to the designer, an optimal contract can in principle be computed by linear programming. In addition to their demanding informational requirements, such optimal contracts are often complex and unintuitive, and do not resemble contracts used in practice.   This paper examines contract theory through the theoretical computer science lens, with the goal of developing novel theory to explain and justify the prevalence of relatively simple contracts, such as linear (pure commission) contracts. First, we consider the case where the principal knows only the first moment of each action's reward distribution, and we prove that linear contracts are guaranteed to be worst-case optimal, ranging over all reward distributions consistent with the given moments. Second, we study linear contracts from a worst-case approximation perspective, and prove several tight parameterized approximation bounds."
                    },
                    {
                        "title": "When Are Welfare Guarantees Robust?",
                        "abstract": "Computational and economic results suggest that social welfare maximization and combinatorial auction design are much easier when bidders' valuations satisfy the \"gross substitutes\" condition. The goal of this paper is to evaluate rigorously the folklore belief that the main take-aways from these results remain valid in settings where the gross substitutes condition holds only approximately. We show that for valuations that pointwise approximate a gross substitutes valuation (in fact even a linear valuation), optimal social welfare cannot be approximated to within a subpolynomial factor and demand oracles cannot be simulated using a subexponential number of value queries. We then provide several positive results by imposing additional structure on the valuations (beyond gross substitutes), using a more stringent notion of approximation, and/or using more powerful oracle access to the valuations. For example, we prove that the performance of the greedy algorithm degrades gracefully for near-linear valuations with approximately decreasing marginal values, that with demand queries, approximate welfare guarantees for XOS valuations degrade gracefully for valuations that are pointwise close to XOS, and that the performance of the Kelso-Crawford auction degrades gracefully for valuations that are close to various subclasses of gross substitutes valuations."
                    },
                    {
                        "title": "Competitive Equilibrium with Generic Budgets: Beyond Additive",
                        "abstract": "We study competitive equilibrium in the canonical Fisher market model, but with indivisible goods. In this model, every agent has a budget of artificial currency with which to purchase bundles of goods. Equilibrium prices match between demand and supply---at such prices, all agents simultaneously get their favorite within-budget bundle, and the market clears. Unfortunately, a competitive equilibrium may not exist when the goods are indivisible, even in extremely simple markets such as two agents with exactly the same budget and a single item. Yet in this example, once the budgets are slightly perturbed---i.e., made generic---a competitive equilibrium is guaranteed to exist. In this paper we explore the extent to which generic budgets can guarantee equilibrium existence (and thus related fairness guarantees) in markets with multiple items. We complement our results in [Babaioff et al., 2019] for additive preferences by exploring the case of general monotone preferences, establishing positive results for small numbers of items and mapping the limits of our approach. We then consider cardinal preferences, define a hierarchy of such preference classes and establish relations among them, and for some classes prove equilibrium existence under generic budgets."
                    },
                    {
                        "title": "The Complexity of Contracts",
                        "abstract": "We initiate the study of computing (near-)optimal contracts in succinctly representable principal-agent settings. Here optimality means maximizing the principal's expected payoff over all incentive-compatible contracts---known in economics as \"second-best\" solutions. We also study a natural relaxation to approximately incentive-compatible contracts.   We focus on principal-agent settings with succinctly described (and exponentially large) outcome spaces. We show that the computational complexity of computing a near-optimal contract depends fundamentally on the number of agent actions. For settings with a constant number of actions, we present a fully polynomial-time approximation scheme (FPTAS) for the separation oracle of the dual of the problem of minimizing the principal's payment to the agent, and use this subroutine to efficiently compute a delta-incentive-compatible (delta-IC) contract whose expected payoff matches or surpasses that of the optimal IC contract.   With an arbitrary number of actions, we prove that the problem is hard to approximate within any constant c. This inapproximability result holds even for delta-IC contracts where delta is a sufficiently rapidly-decaying function of c. On the positive side, we show that simple linear delta-IC contracts with constant delta are sufficient to achieve a constant-factor approximation of the \"first-best\" (full-welfare-extracting) solution, and that such a contract can be computed in polynomial time."
                    },
                    {
                        "title": "Approximate Modularity Revisited",
                        "abstract": "Set functions with convenient properties (such as submodularity) appear in application areas of current interest, such as algorithmic game theory, and allow for improved optimization algorithms. It is natural to ask (e.g., in the context of data driven optimization) how robust such properties are, and whether small deviations from them can be tolerated. We consider two such questions in the important special case of linear set functions.   One question that we address is whether any set function that approximately satisfies the modularity equation (linear functions satisfy the modularity equation exactly) is close to a linear function. The answer to this is positive (in a precise formal sense) as shown by Kalton and Roberts [1983] (and further improved by Bondarenko, Prymak, and Radchenko [2013]). We revisit their proof idea that is based on expander graphs, and provide significantly stronger upper bounds by combining it with new techniques. Furthermore, we provide improved lower bounds for this problem.   Another question that we address is that of how to learn a linear function $h$ that is close to an approximately linear function $f$, while querying the value of $f$ on only a small number of sets. We present a deterministic algorithm that makes only linearly many (in the number of items) nonadaptive queries, by this improving over a previous algorithm of Chierichetti, Das, Dasgupta and Kumar [2015] that is randomized and makes more than a quadratic number of queries. Our learning algorithm is based on a Hadamard transform."
                    },
                    {
                        "title": "Competitive Equilibrium with Indivisible Goods and Generic Budgets",
                        "abstract": "Competitive equilibrium from equal incomes (CEEI) is a classic solution to the problem of fair and efficient allocation of goods [Foley'67, Varian'74]. Every agent receives an equal budget of artificial currency with which to purchase goods, and prices match demand and supply. However, a CEEI is not guaranteed to exist when the goods are indivisible, even in the simple two-agent, single-item market. Yet, it is easy to see that once the two budgets are slightly perturbed (made generic), a competitive equilibrium does exist.   In this paper we aim to extend this approach beyond the single-item case, and study the existence of equilibria in markets with two agents and additive preferences over multiple items. We show that for agents with equal budgets, making the budgets generic -- by adding vanishingly small random perturbations -- ensures the existence of an equilibrium. We further consider agents with arbitrary non-equal budgets, representing non-equal entitlements for goods. We show that competitive equilibrium guarantees a new notion of fairness among non-equal agents, and that it exists in cases of interest (like when the agents have identical preferences) if budgets are perturbed. Our results open opportunities for future research on generic equilibrium existence and fair treatment of non-equals."
                    },
                    {
                        "title": "Welfare and Revenue Guarantees for Competitive Bundling Equilibrium",
                        "abstract": "We study equilibria of markets with $m$ heterogeneous indivisible goods and $n$ consumers with combinatorial preferences. It is well known that a competitive equilibrium is not guaranteed to exist when valuations are not gross substitutes. Given the widespread use of bundling in real-life markets, we study its role as a stabilizing and coordinating device by considering the notion of \\emph{competitive bundling equilibrium}: a competitive equilibrium over the market induced by partitioning the goods for sale into fixed bundles. Compared to other equilibrium concepts involving bundles, this notion has the advantage of simulatneous succinctness ($O(m)$ prices) and market clearance.   Our first set of results concern welfare guarantees. We show that in markets where consumers care only about the number of goods they receive (known as multi-unit or homogeneous markets), even in the presence of complementarities, there always exists a competitive bundling equilibrium that guarantees a logarithmic fraction of the optimal welfare, and this guarantee is tight. We also establish non-trivial welfare guarantees for general markets, two-consumer markets, and markets where the consumer valuations are additive up to a fixed budget (budget-additive).   Our second set of results concern revenue guarantees. Motivated by the fact that the revenue extracted in a standard competitive equilibrium may be zero (even with simple unit-demand consumers), we show that for natural subclasses of gross substitutes valuations, there always exists a competitive bundling equilibrium that extracts a logarithmic fraction of the optimal welfare, and this guarantee is tight. The notion of competitive bundling equilibrium can thus be useful even in markets which possess a standard competitive equilibrium."
                    },
                    {
                        "title": "Auctions with Interdependence and SOS: Improved Approximation",
                        "abstract": "Interdependent values make basic auction design tasks -- in particular maximizing welfare truthfully in single-item auctions -- quite challenging. Eden et al. recently established that if the bidders valuation functions are submodular over their signals (a.k.a. SOS), a truthful 4-approximation to the optimal welfare exists. We show existence of a mechanism that is truthful and achieves a tight 2-approximation to the optimal welfare when signals are binary. Our mechanism is randomized and assigns bidders only 0 or 0.5 probabilities of winning the item. Our results utilize properties of submodular set functions, and extend to matroid settings."
                    },
                    {
                        "title": "Distributional Robustness: From Pricing to Auctions",
                        "abstract": "Robust mechanism design is a rising alternative to Bayesian mechanism design, which yields designs that do not rely on assumptions like full distributional knowledge. We apply this approach to mechanisms for selling a single item, assuming that only the mean of the value distribution and an upper bound on the bidder values are known. We seek the mechanism that maximizes revenue over the worst-case distribution compatible with the known parameters. Such a mechanism arises as an equilibrium of a zero-sum game between the seller and an adversary who chooses the distribution, and so can be referred to as the max-min mechanism.   Carrasco et al. [2018] derive the max-min pricing when the seller faces a single bidder for the item. We go from max-min pricing to max-min auctions by studying the canonical setting of two i.i.d. bidders, and show the max-min mechanism is the second-price auction with a randomized reserve. We derive a closed-form solution for the distribution over reserve prices, as well as the worst-case value distribution, for which there is simple economic intuition. In fact we derive a closed-form solution for the reserve price distribution for any number of bidders.   Our technique for solving the zero-sum game is quite different than that of Carrasco et al.- it involves analyzing a discretized version of the setting, then refining the discretization grid and deriving a closed-form solution for the non-discretized, original setting. Our results establish a difference between the case of two bidders and that of $n \\ge 3$ bidders."
                    }
                ]
            }
        }
    },
    "2409.07142": {
        "paper_data": {
            "title": "Randomized Strategic Facility Location with Predictions",
            "url": "http://arxiv.org/abs/2409.07142v2",
            "arxiv_id": "2409.07142",
            "authors": [
                "Eric Balkanski",
                "Vasilis Gkatzelis",
                "Golnoosh Shahkarami"
            ],
            "abstract": "We revisit the canonical problem of strategic facility location and study the power and limitations of randomization in the design of truthful mechanisms augmented with machine-learned predictions. In the strategic facility location problem, a set of agents are asked to report their locations in some metric space and the goal is to use these reported locations to determine where to open a new facility, aiming to optimize some aggregate measure of distance of the agents from that facility. However, the agents are strategic and can misreport their locations if this may lead to a facility location choice that they prefer. The goal is to design truthful mechanisms, which ensure the agents cannot benefit by misreporting. A lot of prior work has studied this problem from a worst-case perspective, but a recent line of work proposed a framework to refine these results when the designer is provided with some, potentially very inaccurate, predictions regarding the agents' true locations. The goal is to simultaneously provide strong consistency guarantees (i.e., guarantees when the predictions provided to the mechanism are correct) and near-optimal robustness guarantees (i.e., guarantees that hold irrespective of how inaccurate the predictions may be). In this work we focus on the power of randomization in this problem and analyze the best approximation guarantees achievable with respect to the egalitarian social cost measure for one- and two-dimensional Euclidean spaces.",
            "introduction": " Introduction In the classic facility location problem the goal is to determine the ideal location for a new facility, taking as input the preferences of a group of nagents. Due to the wide variety of applications that this problem captures, it has received a lot of attention from different perspectives. A notable example is the strategic version of this problem which is motivated by the fact that the participating agents may be able to strategically misreport their preferences, leading to a facility location choice that they prefer over the one that would be chosen if they were truthful. In the strategic facility location problem, the goal is to design truthful mechanisms, i.e., mechanisms that elicit the agents\u2019 true preferences by carefully removing any incentive for the agents to lie [Moulin, 1980, Procaccia and Tennenholtz, 2013]. This problem has played a central role in the broader literature on mechanism design without money and a lot of prior work has focused on optimizing the quality of the returned location subject to the truthfulness constraint (see Section 1.2 for a brief overview). However, this work has mostly focused on worst-case analysis, often leading to unnecessarily pessimistic impossibility Results Using Extreme ID Prediction We now turn to positive Related Work Strategic facility location Moulin [1980] provided a characterization of deterministic truthful mechanisms on the line and introduced the median mechanism, which returns the median of the location profile x=\u27e8x1,\u00b7\u00b7\u00b7, xn\u27e9. The median mechanism is known to be truthful, providing an optimal solution for the Utilitarian Social Cost and achieving a 2-approximation for the Egalitarian Social Cost, as demonstrated by Procaccia and Tennenholtz [2013]. Notably, Procaccia and Tennenholtz [2013] showed that this 2-approximation factor represents the best performance achievable by any deterministic truthful mechanism. Building on this, Border and Jordan [1983] extended these Appendix A. 4.1 Impossibility",
            "references": [
                {
                    "title": "Improving Online Algorithms via ML Predictions",
                    "abstract": "In this work we study the problem of using machine-learned predictions to improve performance of online algorithms. We consider two classical problems, ski rental and non-clairvoyant job scheduling, and obtain new online algorithms that use predictions to make their decisions. These algorithms are oblivious to the performance of the predictor, improve with better predictions, but do not degrade much if the predictions are poor."
                },
                {
                    "title": "Mechanism design augmented with output advice",
                    "abstract": "Our work revisits the design of mechanisms via the learning-augmented framework. In this model, the algorithm is enhanced with imperfect (machine-learned) information concerning the input, usually referred to as prediction. The goal is to design algorithms whose performance degrades gently as a function of the prediction error and, in particular, perform well if the prediction is accurate, but also provide a worst-case guarantee under any possible error. This framework has been successfully applied recently to various mechanism design settings, where in most cases the mechanism is provided with a prediction about the types of the players. We adopt a perspective in which the mechanism is provided with an output recommendation. We make no assumptions about the quality of the suggested outcome, and the goal is to use the recommendation to design mechanisms with low approximation guarantees whenever the recommended outcome is reasonable, but at the same time to provide worst-case guarantees whenever the recommendation significantly deviates from the optimal one. We propose a generic, universal measure, which we call quality of recommendation, to evaluate mechanisms across various information settings. We demonstrate how this new metric can provide refined analysis in existing results. This model introduces new challenges, as the mechanism receives limited information comparing to settings that use predictions about the types of the agents. We study, through this lens, several well-studied mechanism design paradigms, devising new mechanisms, but also providing refined analysis for existing ones, using as a metric the quality of recommendation. We complement our positive results, by exploring the limitations of known classes of strategyproof mechanisms that can be devised using output recommendation."
                },
                {
                    "title": "To Trust or Not to Trust: Assignment Mechanisms with Predictions in the Private Graph Model",
                    "abstract": "The realm of algorithms with predictions has led to the development of several new algorithms that leverage (potentially erroneous) predictions to enhance their performance guarantees. The challenge is to devise algorithms that achieve optimal approximation guarantees as the prediction quality varies from perfect (consistency) to imperfect (robustness). This framework is particularly appealing in mechanism design contexts, where predictions might convey private information about the agents. In this paper, we design strategyproof mechanisms that leverage predictions to achieve improved approximation guarantees for several variants of the Generalized Assignment Problem (GAP) in the private graph model. In this model, first introduced by Dughmi&Ghosh (2010), the set of resources that an agent is compatible with is private information. For the Bipartite Matching Problem (BMP), we give a deterministic group-strategyproof (GSP) mechanism that is $(1 +1/\\gamma)$-consistent and $(1 + \\gamma)$-robust, where $\\gamma \\ge 1$ is some confidence parameter. We also prove that this is best possible. Remarkably, our mechanism draws inspiration from the renowned Gale-Shapley algorithm, incorporating predictions as a crucial element. Additionally, we give a randomized mechanism that is universally GSP and improves on the guarantees in expectation. The other GAP variants that we consider all make use of a unified greedy mechanism that adds edges to the assignment according to a specific order. Our universally GSP mechanism randomizes over the greedy mechanism, our mechanism for BMP and the predicted assignment, leading to $(1+3/\\gamma)$-consistency and $(3+\\gamma)$-robustness in expectation. All our mechanisms also provide more fine-grained approximation guarantees that interpolate between the consistency and the robustness, depending on some natural error measure of the prediction."
                },
                {
                    "title": "Randomized learning-augmented auctions with revenue guarantees",
                    "abstract": "We consider the fundamental problem of designing a truthful single-item auction with the challenging objective of extracting a large fraction of the highest agent valuation as revenue. Following a recent trend in algorithm design, we assume that the agent valuations belong to a known interval, and a prediction for the highest valuation is available. Then, auction design aims for high consistency and robustness, meaning that, for appropriate pairs of values \u03b3 and \u03c1, the extracted revenue should be at least a \u03b3- or \u03c1-fraction of the highest valuation when the prediction is correct for the input instance or not. We characterize all pairs of parameters \u03b3 and \u03c1 so that a randomized \u03b3-consistent and \u03c1-robust auction exists. Furthermore, for the setting in which robustness can be a function of the prediction error, we give sufficient and necessary conditions for the existence of robust auctions and present randomized auctions that extract a revenue that is only a polylogarithmic (in terms of the prediction error) factor away from the highest agent valuation."
                },
                {
                    "title": "Online Mechanism Design with Predictions",
                    "abstract": "Aiming to overcome some of the limitations of worst-case analysis, the recently proposed framework of\"algorithms with predictions\"allows algorithms to be augmented with a (possibly erroneous) machine-learned prediction that they can use as a guide. In this framework, the goal is to obtain improved guarantees when the prediction is correct, which is called \\emph{consistency}, while simultaneously guaranteeing some worst-case bounds even when the prediction is arbitrarily wrong, which is called \\emph{robustness}. The vast majority of the work on this framework has focused on a refined analysis of online algorithms augmented with predictions regarding the future input. A subsequent line of work has also successfully adapted this framework to mechanism design, where the prediction is regarding the private information of strategic agents. In this paper, we initiate the study of online mechanism design with predictions, which combines the challenges of online algorithms with predictions and mechanism design with predictions. We consider the well-studied problem of designing a revenue-maximizing auction to sell a single item to strategic bidders who arrive and depart over time, each with an unknown, private, value for the item. We study the learning-augmented version of this problem where the auction designer is given a prediction regarding the maximum value over all agents. Our main result is a strategyproof mechanism whose revenue guarantees are $\\alpha$-consistent with respect to the highest value and $(1-\\alpha^2)/4$-robust with respect to the second-highest value, for $\\alpha \\in [0,1]$. We show that this tradeoff is optimal within a broad and natural family of auctions, meaning that any $\\alpha$-consistent mechanism in that family has robustness at most $(1-\\alpha^2)/4$. Finally, we extend our mechanism to also achieve expected revenues proportional to the prediction quality."
                },
                {
                    "title": "Competitive Auctions with Imperfect Predictions",
                    "abstract": "The competitive auction was first proposed by Goldberg, Hartline, and Wright. In their paper, they introduce the competitive analysis framework of online algorithm designing into the traditional revenue-maximizing auction design problem. While the competitive analysis framework only cares about the worst-case bound, a growing body of work in the online algorithm community studies the learning-augmented framework. In this framework, designers are allowed to leverage imperfect machine-learned predictions of unknown information and pursue better theoretical guarantees when the prediction is accurate(consistency). Meanwhile, designers also need to maintain a nearly-optimal worst-case ratio(robustness). In this work, we revisit the competitive auctions in the learning-augmented setting. We leverage the imperfect predictions of the private value of the bidders and design the learning-augmented mechanisms for several competitive auctions with different constraints, including digital good auctions, limited-supply auctions, and general downward-closed permutation environments. For all these auction environments, our mechanisms enjoy $1$-consistency against the strongest benchmark $OPT$, which is impossible to achieve $O(1)$-competitive without predictions. At the same time, our mechanisms also maintain the $O(1)$-robustness against all benchmarks considered in the traditional competitive analysis. Considering the possible inaccuracy of the predictions, we provide a reduction that transforms our learning-augmented mechanisms into an error-tolerant version, which enables the learning-augmented mechanism to ensure satisfactory revenue in scenarios where the prediction error is moderate."
                },
                {
                    "title": "Learning-Augmented Metric Distortion via $(p,q)$-Veto Core",
                    "abstract": "In the metric distortion problem there is a set of candidates $C$ and voters $V$ in the same metric space. The goal is to select a candidate minimizing the social cost: the sum of distances of the selected candidate from all the voters, and the challenge arises from the algorithm receiving only ordinaL input: each voter's ranking of candidate, while the objective function is cardinal, determined by the underlying metric. The distortion of an algorithm is its worst-case approximation factor of the optimal social cost. A key concept here is the (p,q)-veto core, with $p\\in \\Delta(V)$ and $q\\in \\Delta(C)$ being normalized weight vectors representing voters' veto power and candidates' support, respectively. The (p,q)-veto core corresponds to a set of winners from a specific class of deterministic algorithms. Notably, the optimal distortion of $3$ is obtained from this class, by selecting veto core candidates using uniform $p$ and $q$ proportional to candidates' plurality scores. Bounding the distortion of other algorithms from this class is an open problem. Our contribution is twofold. First, we establish upper bounds on the distortion of candidates from the (p,q)-veto core for arbitrary weight vectors $p$ and $q$. Second, we revisit the metric distortion problem through the \\emph{learning-augmented} framework, which equips the algorithm with a (machine-learned) prediction regarding the optimal candidate. The quality of this prediction is unknown, and the goal is to optimize the algorithm's performance under accurate predictions (consistency), while simultaneously providing worst-case guarantees under arbitrarily inaccurate predictions (robustness). We propose an algorithm that chooses candidates from the (p,q)-veto core, using a prediction-guided q vector and, leveraging our distortion bounds, we prove that this algorithm achieves the optimal robustness-consistency trade-off."
                },
                {
                    "title": "Mechanism Design With Predictions for Obnoxious Facility Location",
                    "abstract": "We study mechanism design with predictions for the obnoxious facility location problem. We present deterministic strategyproof mechanisms that display tradeo\ufb00s between robustness and consistency on segments, squares, circles and trees. All these mechanisms are actually group strategyproof, with the exception of the case of squares, where manipulations from coalitions of two agents exist. We prove that these tradeo\ufb00s are optimal in the 1-dimensional case."
                },
                {
                    "title": "Strategyproof Scheduling with Predictions",
                    "abstract": "In their seminal paper that initiated the field of algorithmic mechanism design, \\citet{NR99} studied the problem of designing strategyproof mechanisms for scheduling jobs on unrelated machines aiming to minimize the makespan. They provided a strategyproof mechanism that achieves an $n$-approximation and they made the bold conjecture that this is the best approximation achievable by any deterministic strategyproof scheduling mechanism. After more than two decades and several efforts, $n$ remains the best known approximation and very recent work by \\citet{CKK21} has been able to prove an $\\Omega(\\sqrt{n})$ approximation lower bound for all deterministic strategyproof mechanisms. This strong negative result, however, heavily depends on the fact that the performance of these mechanisms is evaluated using worst-case analysis. To overcome such overly pessimistic, and often uninformative, worst-case bounds, a surge of recent work has focused on the ``learning-augmented framework'', whose goal is to leverage machine-learned predictions to obtain improved approximations when these predictions are accurate (consistency), while also achieving near-optimal worst-case approximations even when the predictions are arbitrarily wrong (robustness). In this work, we study the classic strategic scheduling problem of~\\citet{NR99} using the learning-augmented framework and give a deterministic polynomial-time strategyproof mechanism that is $6$-consistent and $2n$-robust. We thus achieve the ``best of both worlds'': an $O(1)$ consistency and an $O(n)$ robustness that asymptotically matches the best-known approximation. We then extend this result to provide more general worst-case approximation guarantees as a function of the prediction error. Finally, we complement our positive results by showing that any $1$-consistent deterministic strategyproof mechanism has unbounded robustness."
                },
                {
                    "title": "Mechanism Design with Predictions",
                    "abstract": "Improving algorithms via predictions is a very active research topic in recent years. This paper initiates the systematic study of mechanism design in this model. In a number of well-studied mechanism design settings, we make use of imperfect predictions to design mechanisms that perform much better than traditional mechanisms if the predictions are accurate (consistency), while always retaining worst-case guarantees even with very imprecise predictions (robustness). Furthermore, we refer to the largest prediction error sufficient to give a good performance as the error tolerance of a mechanism, and observe that an intrinsic tradeoff among consistency, robustness and error tolerance is common for mechanism design with predictions."
                },
                {
                    "title": "Learning-Augmented Mechanism Design: Leveraging Predictions for Facility Location",
                    "abstract": "In this work we introduce an alternative model for the design and analysis of strategyproof mechanisms that is motivated by the recent surge of work in \"learning-augmented algorithms\". Aiming to complement the traditional approach in computer science, which analyzes the performance of algorithms based on worst-case instances, this line of work has focused on the design and analysis of algorithms that are enhanced with machine-learned predictions regarding the optimal solution. The algorithms can use the predictions as a guide to inform their decisions, and the goal is to achieve much stronger performance guarantees when these predictions are accurate (consistency), while also maintaining near-optimal worst-case guarantees, even if these predictions are very inaccurate (robustness). So far, these results have been limited to algorithms, but in this work we argue that another fertile ground for this framework is in mechanism design. We initiate the design and analysis of strategyproof mechanisms that are augmented with predictions regarding the private information of the participating agents. To exhibit the important benefits of this approach, we revisit the canonical problem of facility location with strategic agents in the two-dimensional Euclidean space. We study both the egalitarian and utilitarian social cost functions, and we propose new strategyproof mechanisms that leverage predictions to guarantee an optimal trade-off between consistency and robustness guarantees. This provides the designer with a menu of mechanism options to choose from, depending on her confidence regarding the prediction accuracy. Furthermore, we also prove parameterized approximation results as a function of the prediction error, showing that our mechanisms perform well even when the predictions are not fully accurate."
                },
                {
                    "title": "Online Graph Algorithms with Predictions",
                    "abstract": "Online algorithms with predictions is a popular and elegant framework for bypassing pessimistic lower bounds in competitive analysis. In this model, online algorithms are supplied with future predictions, and the goal is for the competitive ratio to smoothly interpolate between the best offline and online bounds as a function of the prediction error. In this paper, we study online graph problems with predictions. Our contributions are the following: * The first question is defining prediction error. For graph/metric problems, there can be two types of error, locations that are not predicted, and locations that are predicted but the predicted and actual locations do not coincide exactly. We design a novel definition of prediction error called metric error with outliers to simultaneously capture both types of errors, which thereby generalizes previous definitions of error that only capture one of the two error types. * We give a general framework for obtaining online algorithms with predictions that combines, in a\"black box\"fashion, existing online and offline algorithms, under certain technical conditions. To the best of our knowledge, this is the first general-purpose tool for obtaining online algorithms with predictions. * Using our framework, we obtain tight bounds on the competitive ratio of several classical graph problems as a function of metric error with outliers: Steiner tree, Steiner forest, priority Steiner tree/forest, and uncapacitated/capacitated facility location. Both the definition of metric error with outliers and the general framework for combining offline and online algorithms are not specific to the problems that we consider in this paper. We hope that these will be useful for future work in this domain."
                },
                {
                    "title": "A Novel Prediction Setup for Online Speed-Scaling",
                    "abstract": "Given the rapid rise in energy demand by data centers and computing systems in general, it is fundamental to incorporate energy considerations when designing (scheduling) algorithms. Machine learning can be a useful approach in practice by predicting the future load of the system based on, for example, historical data. However, the effectiveness of such an approach highly depends on the quality of the predictions and can be quite far from optimal when predictions are sub-par. On the other hand, while providing a worst-case guarantee, classical online algorithms can be pessimistic for large classes of inputs arising in practice. This paper, in the spirit of the new area of machine learning augmented algorithms, attempts to obtain the best of both worlds for the classical, deadline based, online speed-scaling problem: Based on the introduction of a novel prediction setup, we develop algorithms that (i) obtain provably low energy-consumption in the presence of adequate predictions, and (ii) are robust against inadequate predictions, and (iii) are smooth, i.e., their performance gradually degrades as the prediction error increases."
                },
                {
                    "title": "Online Facility Location with Predictions",
                    "abstract": "We provide nearly optimal algorithms for online facility location (OFL) with predictions. In OFL, $n$ demand points arrive in order and the algorithm must irrevocably assign each demand point to an open facility upon its arrival. The objective is to minimize the total connection costs from demand points to assigned facilities plus the facility opening cost. We further assume the algorithm is additionally given for each demand point $x_i$ a natural prediction $f_{x_i}^{\\mathrm{pred}}$ which is supposed to be the facility $f_{x_i}^{\\mathrm{opt}}$ that serves $x_i$ in the offline optimal solution. Our main result is an $O(\\min\\{\\log {\\frac{n\\eta_\\infty}{\\mathrm{OPT}}}, \\log{n} \\})$-competitive algorithm where $\\eta_\\infty$ is the maximum prediction error (i.e., the distance between $f_{x_i}^{\\mathrm{pred}}$ and $f_{x_i}^{\\mathrm{opt}}$). Our algorithm overcomes the fundamental $\\Omega(\\frac{\\log n}{\\log \\log n})$ lower bound of OFL (without predictions) when $\\eta_\\infty$ is small, and it still maintains $O(\\log n)$ ratio even when $\\eta_\\infty$ is unbounded. Furthermore, our theoretical analysis is supported by empirical evaluations for the tradeoffs between $\\eta_\\infty$ and the competitive ratio on various real datasets of different types."
                },
                {
                    "title": "Robust Learning-Augmented Caching: An Experimental Study",
                    "abstract": "Effective caching is crucial for the performance of modern-day computing systems. A key optimization problem arising in caching -- which item to evict to make room for a new item -- cannot be optimally solved without knowing the future. There are many classical approximation algorithms for this problem, but more recently researchers started to successfully apply machine learning to decide what to evict by discovering implicit input patterns and predicting the future. While machine learning typically does not provide any worst-case guarantees, the new field of learning-augmented algorithms proposes solutions that leverage classical online caching algorithms to make the machine-learned predictors robust. We are the first to comprehensively evaluate these learning-augmented algorithms on real-world caching datasets and state-of-the-art machine-learned predictors. We show that a straightforward method -- blindly following either a predictor or a classical robust algorithm, and switching whenever one becomes worse than the other -- has only a low overhead over a well-performing predictor, while competing with classical methods when the coupled predictor fails, thus providing a cheap worst-case insurance."
                },
                {
                    "title": "Mechanism Design for Facility Location Problems: A Survey",
                    "abstract": "The study of approximate mechanism design for facility location has been in the center of research at the intersection of artificial intelligence and economics for the last decade, largely due to its practical importance in various domains, such as social planning and clustering. At a high level, the goal is to select a number of locations on which to build a set of facilities, aiming to optimize some social objective based on the preferences of strategic agents, who might have incentives to misreport their private information. This paper presents a comprehensive survey of the significant progress that has been made since the introduction of the problem, highlighting all the different variants and methodologies, as well as the most interesting directions for future research."
                },
                {
                    "title": "On the Strategyproofness of the Geometric Median",
                    "abstract": "The geometric median, an instrumental component of the secure machine learning toolbox, is known to be effective when robustly aggregating models (or gradients), gathered from potentially malicious (or strategic) users. What is less known is the extent to which the geometric median incentivizes dishonest behaviors. This paper addresses this fundamental question by quantifying its strategyproofness. While we observe that the geometric median is not even approximately strategyproof, we prove that it is asymptotically $\\alpha$-strategyproof: when the number of users is large enough, a user that misbehaves can gain at most a multiplicative factor $\\alpha$, which we compute as a function of the distribution followed by the users. We then generalize our results to the case where users actually care more about specific dimensions, determining how this impacts $\\alpha$. We also show how the skewed geometric medians can be used to improve strategyproofness."
                },
                {
                    "title": "Beyond the Worst-Case Analysis of Algorithms",
                    "abstract": "Parameterized algorithmics analyzes running time in \ufb01ner detail than classical complexity theory: instead of expressing the running time of an algorithm as a function of the input size only, dependence on one or more parameters of the input instance is taken into account. In this chapter we sketch some techniques and tools from this rapidly developing area"
                },
                {
                    "title": "Secretaries with Advice",
                    "abstract": "The secretary problem is probably the purest model of decision making under uncertainty. In this paper we ask which advice can we give the algorithm to improve its success probability? We propose a general model that unifies a broad range of problems: from the classic secretary problem with no advice, to the variant where the quality of a secretary is drawn from a known distribution and the algorithm learns each candidate's quality on arrival, to more modern versions of advice in the form of samples, to an ML-inspired model where a classifier gives us noisy signal about whether or not the current secretary is the best on the market. Our main technique is a factor revealing LP that captures all of the problems above. We use this LP formulation to gain structural insight into the optimal policy. Using tools from linear programming, we present a tight analysis of optimal algorithms for secretaries with samples, optimal algorithms when secretaries' qualities are drawn from a known distribution, and a new noisy binary advice model."
                },
                {
                    "title": "Learning Augmented Energy Minimization via Speed Scaling",
                    "abstract": "As power management has become a primary concern in modern data centers, computing resources are being scaled dynamically to minimize energy consumption. We initiate the study of a variant of the classic online speed scaling problem, in which machine learning predictions about the future can be integrated naturally. Inspired by recent work on learning-augmented online algorithms, we propose an algorithm which incorporates predictions in a black-box manner and outperforms any online algorithm if the accuracy is high, yet maintains provable guarantees if the prediction is very inaccurate. We provide both theoretical and experimental evidence to support our claims."
                },
                {
                    "title": "The Primal-Dual method for Learning Augmented Algorithms",
                    "abstract": "The extension of classical online algorithms when provided with predictions is a new and active research area. In this paper, we extend the primal-dual method for online algorithms in order to incorporate predictions that advise the online algorithm about the next action to take. We use this framework to obtain novel algorithms for a variety of online covering problems. We compare our algorithms to the cost of the true and predicted offline optimal solutions and show that these algorithms outperform any online algorithm when the prediction is accurate while maintaining good guarantees when the prediction is misleading."
                },
                {
                    "title": "Strategy Proof Mechanisms for Facility Location in Euclidean and Manhattan Space",
                    "abstract": "We study the impact on mechanisms for facility location of moving from one dimension to two (or more) dimensions and Euclidean or Manhattan distances. We consider three fundamental axiomatic properties: anonymity which is a basic fairness property, Pareto optimality which is one of the most important efficiency properties, and strategy proofness which ensures agents do not have an incentive to mis-report. We also consider how well such mechanisms can approximate the optimal welfare. Our results are somewhat negative. Moving from one dimension to two (or more) dimensions often makes these axiomatic properties more difficult to achieve. For example, with two facilities in Euclidean space or with just a single facility in Manhattan space, no mechanism is anonymous, Pareto optimal and strategy proof. By contrast, mechanisms on the line exist with all three properties.We also show that approximation ratios may increase when moving to two (or more) dimensions. All our impossibility results are minimal. If we drop one of the three axioms (anonymity, Pareto optimality or strategy proofness) multiple mechanisms satisfy the other two axioms."
                },
                {
                    "title": "Online Nash Social Welfare Maximization with Predictions",
                    "abstract": "We consider the problem of allocating a set of divisible goods to $N$ agents in an online manner, aiming to maximize the Nash social welfare, a widely studied objective which provides a balance between fairness and efficiency. The goods arrive in a sequence of $T$ periods and the value of each agent for a good is adversarially chosen when the good arrives. We first observe that no online algorithm can achieve a competitive ratio better than the trivial $O(N)$, unless it is given additional information about the agents' values. Then, in line with the emerging area of\"algorithms with predictions\", we consider a setting where for each agent, the online algorithm is only given a prediction of her monopolist utility, i.e., her utility if all goods were given to her alone (corresponding to the sum of her values over the $T$ periods). Our main result is an online algorithm whose competitive ratio is parameterized by the multiplicative errors in these predictions. The algorithm achieves a competitive ratio of $O(\\log N)$ and $O(\\log T)$ if the predictions are perfectly accurate. Moreover, the competitive ratio degrades smoothly with the errors in the predictions, and is surprisingly robust: the logarithmic competitive ratio holds even if the predictions are very inaccurate. We complement this positive result by showing that our bounds are essentially tight: no online algorithm, even if provided with perfectly accurate predictions, can achieve a competitive ratio of $O(\\log^{1-\\epsilon} N)$ or $O(\\log^{1-\\epsilon} T)$ for any constant $\\epsilon>0$."
                },
                {
                    "title": "Algorithms with predictions",
                    "abstract": "Seeking a new approach that goes beyond worst-case analysis."
                },
                {
                    "title": "Better and Simpler Learning-Augmented Online Caching",
                    "abstract": "Lykouris and Vassilvitskii (ICML 2018) introduce a model of online caching with machine-learned advice, where each page request additionally comes with a prediction of when that page will next be requested. In this model, a natural goal is to design algorithms that (1) perform well when the advice is accurate and (2) remain robust in the worst case a la traditional competitive analysis. Lykouris and Vassilvitskii give such an algorithm by adapting the Marker algorithm to the learning-augmented setting. In a recent work, Rohatgi (SODA 2020) improves on their result with an approach also inspired by randomized marking. We continue the study of this problem, but with a somewhat different approach: We consider combining the BlindOracle algorithm, which just naively follows the predictions, with an optimal competitive algorithm for online caching in a black-box manner. The resulting algorithm outperforms all existing approaches while being significantly simpler. Moreover, we show that combining BlindOracle with LRU is in fact optimal among deterministic algorithms for this problem."
                },
                {
                    "title": "Online Metric Algorithms with Untrusted Predictions",
                    "abstract": "Machine-learned predictors, although achieving very good results for inputs resembling training data, cannot possibly provide perfect predictions in all situations. Still, decision-making systems that are based on such predictors need not only benefit from good predictions, but should also achieve a decent performance when the predictions are inadequate. In this article, we propose a prediction setup for arbitrary metrical task systems (MTS) (e.g., caching, k-server, and convex body chasing) and online matching on the line. We utilize results from the theory of online algorithms to show how to make the setup robust. Specifically, for caching, we present an algorithm whose performance, as a function of the prediction error, is exponentially better than what is achievable for general MTS. Finally, we present an empirical evaluation of our methods on real-world datasets, which suggests practicality."
                },
                {
                    "title": "Online Scheduling via Learned Weights",
                    "abstract": "Online algorithms are a hallmark of worst case optimization under uncertainty. On the other hand, in practice, the input is often far from worst case, and has some predictable characteristics. A recent line of work has shown how to use machine learned predictions to circumvent strong lower bounds on competitive ratios in classic online problems such as ski rental and caching. We study how predictive techniques can be used to break through worst case barriers in online scheduling. The makespan minimization problem with restricted assignments is a classic problem in online scheduling theory. Worst case analysis of this problem gives \u2126(log m ) lower bounds on the competitive ratio in the online setting. We identify a robust quantity that can be predicted and then used to guide online algorithms to achieve better performance. Our predictions are compact in size, having dimension linear in the number of machines, and can be learned using standard o\ufb00 the shelf methods. The performance guarantees of our algorithms depend on the accuracy of the predictions, given predictions with error \u03b7 , we show how to construct O (log \u03b7 ) competitive fractional assignments. We then give an online algorithm that rounds any fractional assignment into an integral schedule. Our algorithm is O ((log log m ) 3 )-competitive and we give a nearly matching \u02dc\u2126(log log m ) lower bound for online rounding algorithms. 1 Altogether, we give algorithms that, equipped with predictions with error \u03b7 , achieve O (log \u03b7 (log log m ) 3 ) competitive ratios, breaking the \u2126(log m ) lower bound even for moderately accurate predictions."
                },
                {
                    "title": "Clock",
                    "abstract": "This \u2018Note\u2019 presents a brief overview of the Community Legal Companion scheme (CLOCK) and its impact on students, \nlitigants in person and lecturers who are involved in the project. The following discussion, which is not intended to be exhaustive, draws upon my experience since 2016 as academic co-lead of CLOCK at the University of Sussex."
                },
                {
                    "title": "Near-Optimal Bounds for Online Caching with Machine Learned Advice",
                    "abstract": "In the model of online caching with machine learned advice, introduced by Lykouris and Vassilvitskii, the goal is to solve the caching problem with an online algorithm that has access to next-arrival predictions: when each input element arrives, the algorithm is given a prediction of the next time when the element will reappear. The traditional model for online caching suffers from an $\\Omega(\\log k)$ competitive ratio lower bound (on a cache of size $k$). In contrast, the augmented model admits algorithms which beat this lower bound when the predictions have low error, and asymptotically match the lower bound when the predictions have high error, even if the algorithms are oblivious to the prediction error. In particular, Lykouris and Vassilvitskii showed that there is a prediction-augmented caching algorithm with a competitive ratio of $O(1+\\min(\\sqrt{\\eta/OPT}, \\log k))$ when the overall $\\ell_1$ prediction error is bounded by $\\eta$, and $OPT$ is the cost of the optimal offline algorithm. \nThe dependence on $k$ in the competitive ratio is optimal, but the dependence on $\\eta/OPT$ may be far from optimal. In this work, we make progress towards closing this gap. Our contributions are twofold. First, we provide an improved algorithm with a competitive ratio of $O(1 + \\min((\\eta/OPT)/k, 1) \\log k)$. Second, we provide a lower bound of $\\Omega(\\log \\min((\\eta/OPT)/(k \\log k), k))$."
                },
                {
                    "title": "Scheduling with Predictions and the Price of Misprediction",
                    "abstract": "In many traditional job scheduling settings, it is assumed that one knows the time it will take for a job to complete service. In such cases, strategies such as shortest job first can be used to improve performance in terms of measures such as the average time a job waits in the system. We consider the setting where the service time is not known, but is predicted by for example a machine learning algorithm. Our main result is the derivation, under natural assumptions, of formulae for the performance of several strategies for queueing systems that use predictions for service times in order to schedule jobs. As part of our analysis, we suggest the framework of the \"price of misprediction,\" which offers a measure of the cost of using predicted information."
                },
                {
                    "title": "A Model for Learned Bloom Filters and Optimizing by Sandwiching",
                    "abstract": "Recent work has suggested enhancing Bloom filters by using a pre-filter, based on applying machine learning to determine a function that models the data set the Bloom filter is meant to represent. Here we model such learned Bloom filters, with the following outcomes: (1) we clarify what guarantees can and cannot be associated with such a structure; (2) we show how to estimate what size the learning function must obtain in order to obtain improved performance; (3) we provide a simple method, sandwiching, for optimizing learned Bloom filters; and (4) we propose a design and analysis approach for a learned Bloomier filter, based on our modeling approach."
                },
                {
                    "title": "Characterization of Group-strategyproof Mechanisms for Facility Location in Strictly Convex Space",
                    "abstract": "We characterize the class of group-strategyproof mechanisms for the single facility location game in any unconstrained strictly convex space. A mechanism is group-strategyproof,if no group of agents can misreport so that all its members are strictlybetter off. A strictly convex space is a normed vector space where |x+y|<2 holds for any pair of different unit vectors x \u2260 y, e.g., any Lp space with p\u2208 (1,\u221e). We show that any deterministic, unanimous, group-strategyproof mechanism must be dictatorial, and that any randomized, unanimous, translation-invariant, group-strategyproof mechanism must be 2-dictatorial.Here a randomized mechanism is 2-dictatorial if the lottery output of the mechanism must be distributed on the line segment between two dictators' inputs. A mechanism is translation-invariant if the output of the mechanism follows the same translation of the input. Our characterization directly implies that any (randomized) translation-invariant approximation algorithm satisfying the group-strategyproofness property has a lower bound of 2-approximation for maximum cost (whenever n \u2265 3), and n/2 - 1 for social cost. We also find an algorithm that 2-approximates the maximum cost and n/2-approximates the social cost, proving the bounds to be (almost) tight."
                },
                {
                    "title": "Competitive caching with machine learned advice",
                    "abstract": "Traditional online algorithms encapsulate decision making under uncertainty, and give ways to hedge against all possible future events, while guaranteeing a nearly optimal solution, as compared to an offline optimum. On the other hand, machine learning algorithms are in the business of extrapolating patterns found in the data to predict the future, and usually come with strong guarantees on the expected generalization error.\n In this work, we develop a framework for augmenting online algorithms with a machine learned predictor to achieve competitive ratios that provably improve upon unconditional worst-case lower bounds when the predictor has low error. Our approach treats the predictor as a complete black box and is not dependent on its inner workings or the exact distribution of its errors.\n \n We apply this framework to the traditional caching problem\u2014creating an eviction strategy for a cache of size\n k\n . We demonstrate that naively following the oracle\u2019s recommendations may lead to very poor performance, even when the average error is quite low. Instead, we show how to modify the Marker algorithm to take into account the predictions and prove that this combined approach achieves a competitive ratio that both (i) decreases as the predictor\u2019s error decreases and (ii) is always capped by\n O\n (log\n k\n ), which can be achieved without any assistance from the predictor. We complement our results with an empirical evaluation of our algorithm on real-world datasets and show that it performs well empirically even when using simple off-the-shelf predictions.\n"
                },
                {
                    "title": "The Case for Learned Index Structures",
                    "abstract": "Indexes are models: a \\btree-Index can be seen as a model to map a key to the position of a record within a sorted array, a Hash-Index as a model to map a key to a position of a record within an unsorted array, and a BitMap-Index as a model to indicate if a data record exists or not. In this exploratory research paper, we start from this premise and posit that all existing index structures can be replaced with other types of models, including deep-learning models, which we term \\em learned indexes. We theoretically analyze under which conditions learned indexes outperform traditional index structures and describe the main challenges in designing learned index structures. Our initial results show that our learned indexes can have significant advantages over traditional indexes. More importantly, we believe that the idea of replacing core components of a data management system through learned models has far reaching implications for future systems designs and that this work provides just a glimpse of what might be possible."
                },
                {
                    "title": "Strategyproof facility location and the least squares objective",
                    "abstract": "We consider the problem of locating a public facility on a tree, where a set of n strategic agents report their locations and a mechanism determines, either deterministically or randomly, the location of the facility. The contribution of this paper is twofold.\n First, we introduce, for the first time, a general and clean family of strategyproof (SP) mechanisms for facility location on tree networks. Quite miraculously, all of the deterministic and randomized SP mechanisms that have been previously proposed can be cast as special cases of this family. Thus, the proposed mechanism unifies much of the existing literature on SP facility location problems, and simplifies its analysis.\n Second, we demonstrate the strength of the proposed family of mechanisms by proving new bounds on the approximation of the minimum sum of squares (miniSOS) objective on line and tree networks. For lines, we devise a randomized mechanism that gives 1.5-approximation, and show, through a subtle analysis, that no other randomized SP mechanism can provide a better approximation. For general trees, we construct a randomized mechanism that gives 1.83-approximation. This result provides a separation between deterministic and randomized mechanisms, as it is complemented by a lower bound of 2 for any deterministic mechanism. We believe that the devised family of mechanisms will prove useful in studying approximation bounds for additional objectives."
                },
                {
                    "title": "Strategyproof Approximation of the Minimax on Networks",
                    "abstract": "We consider the problem of locating a facility on a network represented by a graph. A set of strategic agents have different ideal locations for the facility; the cost of an agent is the distance between its ideal location and the facility. A mechanism maps the locations reported by the agents to the location of the facility. We wish to design mechanisms that are strategyproof (SP) in the sense that agents can never benefit by lying and, at the same time, provide a small approximation ratio with respect to the minimax measure. We design a novel \u201chybrid\u201d strategyproof randomized mechanism that provides a tight approximation ratio of 3/2 when the network is a circle (known as a ring in the case of computer networks). Furthermore, we show that no randomized SP mechanism can provide an approximation ratio better than 2 - o (1), even when the network is a tree, thereby matching a trivial upper bound of two."
                },
                {
                    "title": "Online facility location",
                    "abstract": "We consider the online variant of facility location, in which demand points arrive one at a time and we must maintain a set of facilities to service these points. We provide a randomized online O(1)-competitive algorithm in the case where points arrive in random order. If points are ordered adversarially, we show that no algorithm can be constant-competitive, and provide an O(log n)-competitive algorithm. Our algorithms are randomized and the analysis depends heavily on the concept of expected waiting time. We also combine our techniques with those of M. Charikar and S. Guha (1999) to provide a linear-time constant approximation for the offline facility location problem."
                },
                {
                    "title": "Generalized Median Voter Schemes and Committees",
                    "abstract": "Abstract We define a multidimensional analogue of a single-peaked preference and generalize the notion of a median voter scheme. Every onto strategy-proof social choice function on a single-peaked domain is a generalized median voter scheme. Since a single-peaked preference can be identified unequivocally with its bliss point, one can view a social choice function as an Arrowian social welfare function. We show that a social choice function is strategy-proof iff, viewed as a social welfare function, it satisfies a monotonicity property. Finally, we investigate strategic decision making in hierarchical committees. Journal of Economic Literature Classification Numbers: C72, D71."
                },
                {
                    "title": "Online Knapsack with Frequency Predictions",
                    "abstract": "There has been recent interest in using machine-learned predictions to improve the worst-case guarantees of online algorithms. In this paper we continue this line of work by studying the online knapsack problem, but with very weak predictions: in the form of knowing an upper and lower bound for the number of items of each value. We systematically derive online algorithms that attain the best possible competitive ratio for any \ufb01xed prediction; we also extend the results to more general settings such as generalized one-way trading and two-stage online knapsack. Our work shows that even seemingly weak predictions can be utilized effectively to provably improve the performance of online algorithms."
                },
                {
                    "title": "Online Facility Location with Multiple Advice",
                    "abstract": "Clustering is a central topic in unsupervised learning and its online formulation has received a lot of attention in recent years. In this paper, we study the classic facility location problem in the presence of multiple machine-learned advice. We design an algorithm with provable performance guarantees such that, if the advice is good, it outperforms the best-known online algorithms for the problem, and if it is bad it still matches their performance. We complement our theoretical analysis with an in-depth study of the performance of our algorithm, showing its effectiveness on synthetic and real-world data sets."
                },
                {
                    "title": "Approximate Mechanism Design without Money",
                    "abstract": "The literature on algorithmic mechanism design is mostly concerned with game-theoretic versions of optimization problems to which standard economic money-based mechanisms cannot be applied efficiently. Recent years have seen the design of various truthful approximation mechanisms that rely on payments. In this article, we advocate the reconsideration of highly structured optimization problems in the context of mechanism design. We explicitly argue for the first time that, in such domains, approximation can be leveraged to obtain truthfulness without resorting to payments. This stands in contrast to previous work where payments are almost ubiquitous and (more often than not) approximation is a necessary evil that is required to circumvent computational complexity.\n We present a case study in approximate mechanism design without money. In our basic setting, agents are located on the real line and the mechanism must select the location of a public facility; the cost of an agent is its distance to the facility. We establish tight upper and lower bounds for the approximation ratio given by strategyproof mechanisms without payments, with respect to both deterministic and randomized mechanisms, under two objective functions: the social cost and the maximum cost. We then extend our results in two natural directions: a domain where two facilities must be located and a domain where each agent controls multiple locations."
                },
                {
                    "title": "Computational geometry: algorithms and applications",
                    "abstract": "The polygon in which geometric objects points. This is on geometry but this book can. Algorithm they consider and exercises but, these string matching algorithms data structures for those books. Computational geometry is clear and geographic information systems it includes some. Teaching layered range trees for self, study by obtaining a bachelor's. Randomized algorithms which varies over at, all main text. For many application areas all the dynamically changing set. In computational geometry are often quite hard to copy. In chapter which is to computational geometry. Springer ebooks it presents no real source code. For those people in parts of points the problems book. Not to a recognized discipline can, directly download your ebooks! Unlike most efficient and others in, this introduction to switch between. I find them for a, basic background. For geometric search problems studied and, hence the material covered at dallas. Each chapter the smallest distance from, begins with a first use solutions? This may be used to copy and algorithms as laptops. Readers are not regarded as well accepted introduction."
                },
                {
                    "title": "Straightforward Elections, Unanimity, and Phantom Voters",
                    "abstract": "Non-manipulable direct revelation social choice functions are characterized for societies where the space of alternatives is a euclidean space and all voters have separable star-shaped preferences with a global optimum. If a non-manipulable choice function satisfies a weak unanmity-respecting condition (which is equivalent to having an unrestricted range) then it will depend only on voters' ideal points. Further, such a choice function will decompose into a product of one-dimensional mechanisms in the sense that each coordinate of the chosen point depends only on the respective coordinate of the voters' ideal points. Each coordinate function will also be non-manipulable and respect unanimity. Such one-dimensional mechanisms are uncompromising in the sense that voters cannot take an extreme position to influence the choice to their advantage. Two characterizations of uncompromising choice functions are presented. One is in terms of a continuity condition, the other in terms of \"phantom voters\" i.e. those points which are chosen which are not any voter's ideal point. There are many such mechanisms which are not dictatorial. However, if differentiability is required of the choice function, this forces it to be either constant or dictatorial. In the multidimensional case, non-separability of preferences leads to dictatorship, even if preferences are restricted to be quadratic."
                }
            ],
            "categories": [
                "cs.GT"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we design truthful mechanisms for the strategic facility location problem that effectively elicit agents' true preferences while optimizing the quality of the chosen location?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of mechanism design, particularly in scenarios where agents may misreport their preferences. By developing truthful mechanisms, we can ensure that the facility location chosen reflects the genuine needs of the agents, leading to more efficient and equitable outcomes. This research could pave the way for practical applications in urban planning, resource allocation, and service delivery, ultimately influencing future studies on strategic behavior in various economic and social contexts.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in this problem stem from the need to balance truthfulness with optimality in the facility location choice. Naive approaches may fail because they do not account for the strategic behavior of agents, leading to mechanisms that can be easily manipulated. Additionally, the theoretical complexity of designing mechanisms that satisfy both truthfulness and optimality constraints presents significant obstacles, as does the need to analyze the worst-case scenarios that could arise from agents misreporting their preferences.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on worst-case analysis, which often results in overly pessimistic conclusions about the feasibility of truthful mechanisms. Many existing solutions have not adequately addressed the strategic nature of agent preferences, leading to gaps in understanding how to design mechanisms that are both truthful and effective. Our approach differs by emphasizing a more nuanced analysis of agent behavior and exploring new methodologies that can improve upon the limitations of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a new class of truthful mechanisms for the strategic facility location problem, utilizing advanced techniques such as extreme ID prediction. We will analyze a dataset of agent preferences and apply metrics such as social cost and approximation ratios to evaluate the performance of our mechanisms. The expected outcomes include demonstrating improved truthfulness and optimality in facility location choices, as well as providing theoretical insights that could inform future research in mechanism design."
            }
        },
        "author_data": {
            "b096a461-f8e2-4c98-90ea-36d1dcd0690d": {
                "pk": "b096a461-f8e2-4c98-90ea-36d1dcd0690d",
                "name": "Eric Balkanski",
                "collaborators": [
                    "Yaron Singer",
                    "Aviad Rubinstein",
                    "Adam Breuer",
                    "Renato Paes Leme",
                    "Shipra Agrawal",
                    "Vahab Mirrokni",
                    "Balasubramanian Sivan",
                    "Arpit Agarwal",
                    "Jason D. Hartline",
                    "Nicole Immorlica"
                ],
                "domain": [
                    "Mechanism Design",
                    "Submodular Optimization",
                    "Game Theory",
                    "Adversarial Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Dynamic First Price Auctions Robust to Heterogeneous Buyers",
                        "abstract": "We study dynamic mechanisms for optimizing revenue in repeated auctions, that are robust to heterogeneous forward-looking and learning behavior of the buyers. Typically it is assumed that the buyers are either all myopic or are all infinite lookahead, and that buyers understand and trust the mechanism. These assumptions raise the following question: is it possible to design approximately revenue optimal mechanisms when the buyer pool is heterogeneous? Facing a heterogeneous population of buyers with an unknown mixture of $k$-lookahead buyers, myopic buyers, no-regret-learners and no-policy-regret learners, we design a simple state-based mechanism that achieves a constant fraction of the optimal achievable revenue."
                    },
                    {
                        "title": "Non-monotone Submodular Maximization in Exponentially Fewer Iterations",
                        "abstract": "In this paper we consider parallelization for applications whose objective can be expressed as maximizing a non-monotone submodular function under a cardinality constraint. Our main result is an algorithm whose approximation is arbitrarily close to $1/2e$ in $O(\\log^2 n)$ adaptive rounds, where $n$ is the size of the ground set. This is an exponential speedup in parallel running time over any previously studied algorithm for constrained non-monotone submodular maximization. Beyond its provable guarantees, the algorithm performs well in practice. Specifically, experiments on traffic monitoring and personalized data summarization applications show that the algorithm finds solutions whose values are competitive with state-of-the-art algorithms while running in exponentially fewer parallel iterations."
                    },
                    {
                        "title": "Learning-Augmented Dynamic Submodular Maximization",
                        "abstract": "In dynamic submodular maximization, the goal is to maintain a high-value solution over a sequence of element insertions and deletions with a fast update time. Motivated by large-scale applications and the fact that dynamic data often exhibits patterns, we ask the following question: can predictions be used to accelerate the update time of dynamic submodular maximization algorithms?   We consider the model for dynamic algorithms with predictions where predictions regarding the insertion and deletion times of elements can be used for preprocessing. Our main result is an algorithm with an $O(poly(\\log \\eta, \\log w, \\log k))$ amortized update time over the sequence of updates that achieves a $1/2 - \\epsilon$ approximation in expectation for dynamic monotone submodular maximization under a cardinality constraint $k$, where the prediction error $\\eta$ is the number of elements that are not inserted and deleted within $w$ time steps of their predicted insertion and deletion times. This amortized update time is independent of the length of the stream and instead depends on the prediction error."
                    },
                    {
                        "title": "An Exponential Speedup in Parallel Running Time for Submodular Maximization without Loss in Approximation",
                        "abstract": "In this paper we study the adaptivity of submodular maximization. Adaptivity quantifies the number of sequential rounds that an algorithm makes when function evaluations can be executed in parallel. Adaptivity is a fundamental concept that is heavily studied across a variety of areas in computer science, largely due to the need for parallelizing computation. For the canonical problem of maximizing a monotone submodular function under a cardinality constraint, it is well known that a simple greedy algorithm achieves a $1-1/e$ approximation and that this approximation is optimal for polynomial-time algorithms. Somewhat surprisingly, despite extensive efforts on submodular optimization for large-scale datasets, until very recently there was no known algorithm that achieves a constant factor approximation for this problem whose adaptivity is sublinear in the size of the ground set $n$.   Recent work by Balkanski and Singer describes an algorithm that obtains an approximation arbitrarily close to $1/3$ in $\\mathcal{O}(\\log n)$ adaptive rounds and shows that no algorithm can obtain a constant factor approximation in $\\tilde{o}(\\log n)$ adaptive rounds. This approach achieves an exponential speedup in adaptivity (and parallel running time) at the expense of approximation quality.   In this paper we describe a novel approach that yields an algorithm whose approximation is arbitrarily close to the optimal $1-1/e$ guarantee in $\\mathcal{O}(\\log n)$ adaptive rounds. This algorithm therefore achieves an exponential speedup in parallel running time for submodular maximization at the expense of an arbitrarily small loss in approximation quality. This guarantee is optimal in both approximation and adaptivity, up to lower order terms."
                    },
                    {
                        "title": "Bayesian Budget Feasibility with Posted Pricing",
                        "abstract": "We consider the problem of budget feasible mechanism design proposed by Singer (2010), but in a Bayesian setting. A principal has a public value for hiring a subset of the agents and a budget, while the agents have private costs for being hired. We consider both additive and submodular value functions of the principal. We show that there are simple, practical, ex post budget balanced posted pricing mechanisms that approximate the value obtained by the Bayesian optimal mechanism that is budget balanced only in expectation. A main motivating application for this work is the crowdsourcing large projects, e.g., on Mechanical Turk, where workers are drawn from a large population and posted pricing is standard. Our analysis methods relate to contention resolution schemes in submodular optimization of Vondrak et al. (2011) and the correlation gap analysis of Yan (2011)."
                    },
                    {
                        "title": "The Limitations of Optimization from Samples",
                        "abstract": "In this paper we consider the following question: can we optimize objective functions from the training data we use to learn them? We formalize this question through a novel framework we call optimization from samples (OPS). In OPS, we are given sampled values of a function drawn from some distribution and the objective is to optimize the function under some constraint.   While there are interesting classes of functions that can be optimized from samples, our main result is an impossibility. We show that there are classes of functions which are statistically learnable and optimizable, but for which no reasonable approximation for optimization from samples is achievable. In particular, our main result shows that there is no constant factor approximation for maximizing coverage functions under a cardinality constraint using polynomially-many samples drawn from any distribution.   We also show tight approximation guarantees for maximization under a cardinality constraint of several interesting classes of functions including unit-demand, additive, and general monotone submodular functions, as well as a constant factor approximation for monotone submodular functions with bounded curvature."
                    },
                    {
                        "title": "The Importance of Communities for Learning to Influence",
                        "abstract": "We consider the canonical problem of influence maximization in social networks. Since the seminal work of Kempe, Kleinberg, and Tardos, there have been two largely disjoint efforts on this problem. The first studies the problem associated with learning the parameters of the generative influence model. The second focuses on the algorithmic challenge of identifying a set of influencers, assuming the parameters of the generative model are known. Recent results on learning and optimization imply that in general, if the generative model is not known but rather learned from training data, no algorithm can yield a constant factor approximation guarantee using polynomially-many samples, drawn from any distribution.   In this paper, we design a simple heuristic that overcomes this negative result in practice by leveraging the strong community structure of social networks. Although in general the approximation guarantee of our algorithm is necessarily unbounded, we show that this algorithm performs well experimentally. To justify its performance, we prove our algorithm obtains a constant factor approximation guarantee on graphs generated through the stochastic block model, traditionally used to model networks with community structure."
                    },
                    {
                        "title": "On the Construction of Substitutes",
                        "abstract": "Gross substitutability is a central concept in Economics and is connected to important notions in Discrete Convex Analysis, Number Theory and the analysis of Greedy algorithms in Computer Science. Many different characterizations are known for this class, but providing a constructive description remains a major open problem. The construction problem asks how to construct all gross substitutes from a class of simpler functions using a set of operations. Since gross substitutes are a natural generalization of matroids to real-valued functions, matroid rank functions form a desirable such class of simpler functions.   Shioura proved that a rich class of gross substitutes can be expressed as sums of matroid rank functions, but it is open whether all gross substitutes can be constructed this way. Our main result is a negative answer showing that some gross substitutes cannot be expressed as positive linear combinations of matroid rank functions. En route, we provide necessary and sufficient conditions for the sum to preserve substitutability, uncover a new operation preserving substitutability and fully describe all substitutes with at most 4 items."
                    },
                    {
                        "title": "The FAST Algorithm for Submodular Maximization",
                        "abstract": "In this paper we describe a new algorithm called Fast Adaptive Sequencing Technique (FAST) for maximizing a monotone submodular function under a cardinality constraint $k$ whose approximation ratio is arbitrarily close to $1-1/e$, is $O(\\log(n) \\log^2(\\log k))$ adaptive, and uses a total of $O(n \\log\\log(k))$ queries. Recent algorithms have comparable guarantees in terms of asymptotic worst case analysis, but their actual number of rounds and query complexity depend on very large constants and polynomials in terms of precision and confidence, making them impractical for large data sets. Our main contribution is a design that is extremely efficient both in terms of its non-asymptotic worst case query complexity and number of rounds, and in terms of its practical runtime. We show that this algorithm outperforms any algorithm for submodular maximization we are aware of, including hyper-optimized parallel versions of state-of-the-art serial algorithms, by running experiments on large data sets. These experiments show FAST is orders of magnitude faster than the state-of-the-art."
                    },
                    {
                        "title": "An Optimal Approximation for Submodular Maximization under a Matroid Constraint in the Adaptive Complexity Model",
                        "abstract": "In this paper we study submodular maximization under a matroid constraint in the adaptive complexity model. This model was recently introduced in the context of submodular optimization in [BS18a] to quantify the information theoretic complexity of black-box optimization in a parallel computation model. Informally, the adaptivity of an algorithm is the number of sequential rounds it makes when each round can execute polynomially-many function evaluations in parallel. Since submodular optimization is regularly applied on large datasets we seek algorithms with low adaptivity to enable speedups via parallelization. Consequently, a recent line of work has been devoted to designing constant factor approximation algorithms for maximizing submodular functions under various constraints in the adaptive complexity model [BS18a, BS18b, BBS18, BRS19, EN19, FMZ19, CQ19, ENV18, FMZ18].   Despite the burst in work on submodular maximization in the adaptive complexity model the fundamental problem of maximizing a monotone submodular function under a matroid constraint has remained elusive. In particular, all known techniques fail for this problem and there are no known constant factor approximation algorithms whose adaptivity is sublinear in the rank of the matroid $k$ or in the worst case sublinear in the size of the ground set $n$.   In this paper we present an approximation algorithm for the problem of maximizing a monotone submodular function under a matroid constraint in the adaptive complexity model. The approximation guarantee of the algorithm is arbitrarily close to the optimal $1-1/e$ and it has near optimal adaptivity of $O(\\log(n)\\log(k))$. This result is obtained using a novel technique of adaptive sequencing which departs from previous techniques for submodular maximization in the adaptive complexity model."
                    },
                    {
                        "title": "Statistical Cost Sharing",
                        "abstract": "We study the cost sharing problem for cooperative games in situations where the cost function $C$ is not available via oracle queries, but must instead be derived from data, represented as tuples $(S, C(S))$, for different subsets $S$ of players. We formalize this approach, which we call statistical cost sharing, and consider the computation of the core and the Shapley value, when the tuples are drawn from some distribution $\\mathcal{D}$.   Previous work by Balcan et al. in this setting showed how to compute cost shares that satisfy the core property with high probability for limited classes of functions. We expand on their work and give an algorithm that computes such cost shares for any function with a non-empty core. We complement these results by proving an inapproximability lower bound for a weaker relaxation.   We then turn our attention to the Shapley value. We first show that when cost functions come from the family of submodular functions with bounded curvature, $\\kappa$, the Shapley value can be approximated from samples up to a $\\sqrt{1 - \\kappa}$ factor, and that the bound is tight. We then define statistical analogues of the Shapley axioms, and derive a notion of statistical Shapley value. We show that these can always be approximated arbitrarily well for general functions over any distribution $\\mathcal{D}$."
                    },
                    {
                        "title": "Submodular Maximization in Exactly $n$ Queries",
                        "abstract": "In this work, we study the classical problem of maximizing a submodular function subject to a matroid constraint. We develop deterministic algorithms that are very parsimonious with respect to querying the submodular function, for both the case when the submodular function is monotone and the general submodular case. In particular, we present a 1/4 approximation algorithm for the monotone case that uses exactly one query per element, which gives the same total number of queries n as the number of queries required to compute the maximum singleton. For the general case, we present a constant factor approximation algorithm that requires 2 queries per element, which is the first algorithm for this problem with linear query complexity in the size of the ground set."
                    },
                    {
                        "title": "An Algorithm for the Assignment Game Beyond Additive Valuations",
                        "abstract": "The assignment game, introduced by Shapley and Shubik (1971), is a classic model for two-sided matching markets between buyers and sellers. In the original assignment game, it is assumed that payments lead to transferable utility and that buyers have unit-demand valuations for the items being sold. Two important and mostly independent lines of work have studied more general settings with imperfectly transferable utility and gross substitutes valuations. Multiple efficient algorithms have been proposed for computing a competitive equilibrium, the standard solution concept in assignment games, in these two settings. Our main result is an efficient algorithm for computing competitive equilibria in a setting with both imperfectly transferable utility and gross substitutes valuations. Our algorithm combines augmenting path techniques from maximum matching and algorithms for matroid intersection. We also show that, in a mild generalization of our model, computing a competitive equilibrium is NP-hard."
                    },
                    {
                        "title": "Parallelization does not Accelerate Convex Optimization: Adaptivity Lower Bounds for Non-smooth Convex Minimization",
                        "abstract": "In this paper we study the limitations of parallelization in convex optimization. A convenient approach to study parallelization is through the prism of \\emph{adaptivity} which is an information theoretic measure of the parallel runtime of an algorithm [BS18]. Informally, adaptivity is the number of sequential rounds an algorithm needs to make when it can execute polynomially-many queries in parallel at every round. For combinatorial optimization with black-box oracle access, the study of adaptivity has recently led to exponential accelerations in parallel runtime and the natural question is whether dramatic accelerations are achievable for convex optimization.   For the problem of minimizing a non-smooth convex function $f:[0,1]^n\\to \\mathbb{R}$ over the unit Euclidean ball, we give a tight lower bound that shows that even when $\\texttt{poly}(n)$ queries can be executed in parallel, there is no randomized algorithm with $\\tilde{o}(n^{1/3})$ rounds of adaptivity that has convergence rate that is better than those achievable with a one-query-per-round algorithm. A similar lower bound was obtained by Nemirovski [Nem94], however that result holds for the $\\ell_{\\infty}$-setting instead of $\\ell_2$. In addition, we also show a tight lower bound that holds for Lipschitz and strongly convex functions.   At the time of writing this manuscript we were not aware of Nemirovski's result. The construction we use is similar to the one in [Nem94], though our analysis is different. Due to the close relationship between this work and [Nem94], we view the research contribution of this manuscript limited and it should serve as an instructful approach to understanding lower bounds for parallel optimization."
                    },
                    {
                        "title": "Secretary Ranking with Minimal Inversions",
                        "abstract": "We study a twist on the classic secretary problem, which we term the secretary ranking problem: elements from an ordered set arrive in random order and instead of picking the maximum element, the algorithm is asked to assign a rank, or position, to each of the elements. The rank assigned is irrevocable and is given knowing only the pairwise comparisons with elements previously arrived. The goal is to minimize the distance of the rank produced to the true rank of the elements measured by the Kendall-Tau distance, which corresponds to the number of pairs that are inverted with respect to the true order.   Our main result is a matching upper and lower bound for the secretary ranking problem. We present an algorithm that ranks $n$ elements with only $O(n^{3/2})$ inversions in expectation, and show that any algorithm necessarily suffers $\\Omega(n^{3/2})$ inversions when there are $n$ available positions. In terms of techniques, the analysis of our algorithm draws connections to linear probing in the hashing literature, while our lower bound result relies on a general anti-concentration bound for a generic balls and bins sampling process. We also consider the case where the number of positions $m$ can be larger than the number of secretaries $n$ and provide an improved bound by showing a connection of this problem with random binary trees."
                    },
                    {
                        "title": "Instance Specific Approximations for Submodular Maximization",
                        "abstract": "For many optimization problems in machine learning, finding an optimal solution is computationally intractable and we seek algorithms that perform well in practice. Since computational intractability often results from pathological instances, we look for methods to benchmark the performance of algorithms against optimal solutions on real-world instances. The main challenge is that an optimal solution cannot be efficiently computed for intractable problems, and we therefore often do not know how far a solution is from being optimal. A major question is therefore how to measure the performance of an algorithm in comparison to an optimal solution on instances we encounter in practice.   In this paper, we address this question in the context of submodular optimization problems. For the canonical problem of submodular maximization under a cardinality constraint, it is intractable to compute a solution that is better than a $1-1/e \\approx 0.63$ fraction of the optimum. Algorithms like the celebrated greedy algorithm are guaranteed to achieve this $1-1/e$ bound on any instance and are used in practice.   Our main contribution is not a new algorithm for submodular maximization but an analytical method that measures how close an algorithm for submodular maximization is to optimal on a given problem instance. We use this method to show that on a wide variety of real-world datasets and objectives, the approximation of the solution found by greedy goes well beyond $1-1/e$ and is often at least 0.95. We develop this method using a novel technique that lower bounds the objective of a dual minimization problem to obtain an upper bound on the value of an optimal solution to the primal maximization problem."
                    },
                    {
                        "title": "Learning Low Degree Hypergraphs",
                        "abstract": "We study the problem of learning a hypergraph via edge detecting queries. In this problem, a learner queries subsets of vertices of a hidden hypergraph and observes whether these subsets contain an edge or not. In general, learning a hypergraph with $m$ edges of maximum size $d$ requires $\\Omega((2m/d)^{d/2})$ queries. In this paper, we aim to identify families of hypergraphs that can be learned without suffering from a query complexity that grows exponentially in the size of the edges.   We show that hypermatchings and low-degree near-uniform hypergraphs with $n$ vertices are learnable with poly$(n)$ queries. For learning hypermatchings (hypergraphs of maximum degree $ 1$), we give an $O(\\log^3 n)$-round algorithm with $O(n \\log^5 n)$ queries. We complement this upper bound by showing that there are no algorithms with poly$(n)$ queries that learn hypermatchings in $o(\\log \\log n)$ adaptive rounds. For hypergraphs with maximum degree $\\Delta$ and edge size ratio $\\rho$, we give a non-adaptive algorithm with $O((2n)^{\\rho \\Delta+1}\\log^2 n)$ queries. To the best of our knowledge, these are the first algorithms with poly$(n, m)$ query complexity for learning non-trivial families of hypergraphs that have a super-constant number of edges of super-constant size."
                    },
                    {
                        "title": "Strategyproof Scheduling with Predictions",
                        "abstract": "In their seminal paper that initiated the field of algorithmic mechanism design, \\citet{NR99} studied the problem of designing strategyproof mechanisms for scheduling jobs on unrelated machines aiming to minimize the makespan. They provided a strategyproof mechanism that achieves an $n$-approximation and they made the bold conjecture that this is the best approximation achievable by any deterministic strategyproof scheduling mechanism. After more than two decades and several efforts, $n$ remains the best known approximation and very recent work by \\citet{CKK21} has been able to prove an $\\Omega(\\sqrt{n})$ approximation lower bound for all deterministic strategyproof mechanisms. This strong negative result, however, heavily depends on the fact that the performance of these mechanisms is evaluated using worst-case analysis. To overcome such overly pessimistic, and often uninformative, worst-case bounds, a surge of recent work has focused on the ``learning-augmented framework'', whose goal is to leverage machine-learned predictions to obtain improved approximations when these predictions are accurate (consistency), while also achieving near-optimal worst-case approximations even when the predictions are arbitrarily wrong (robustness).   In this work, we study the classic strategic scheduling problem of~\\citet{NR99} using the learning-augmented framework and give a deterministic polynomial-time strategyproof mechanism that is $6$-consistent and $2n$-robust. We thus achieve the ``best of both worlds'': an $O(1)$ consistency and an $O(n)$ robustness that asymptotically matches the best-known approximation. We then extend this result to provide more general worst-case approximation guarantees as a function of the prediction error. Finally, we complement our positive results by showing that any $1$-consistent deterministic strategyproof mechanism has unbounded robustness."
                    },
                    {
                        "title": "Adversarial Attacks on Binary Image Recognition Systems",
                        "abstract": "We initiate the study of adversarial attacks on models for binary (i.e. black and white) image classification. Although there has been a great deal of work on attacking models for colored and grayscale images, little is known about attacks on models for binary images. Models trained to classify binary images are used in text recognition applications such as check processing, license plate recognition, invoice processing, and many others. In contrast to colored and grayscale images, the search space of attacks on binary images is extremely restricted and noise cannot be hidden with minor perturbations in each pixel. Thus, the optimization landscape of attacks on binary images introduces new fundamental challenges.   In this paper we introduce a new attack algorithm called SCAR, designed to fool classifiers of binary images. We show that SCAR significantly outperforms existing $L_0$ attacks applied to the binary setting and use it to demonstrate the vulnerability of real-world text recognition systems. SCAR's strong performance in practice contrasts with the existence of classifiers that are provably robust to large perturbations. In many cases, altering a single pixel is sufficient to trick Tesseract, a popular open-source text recognition system, to misclassify a word as a different word in the English dictionary. We also license software from providers of check processing systems to most of the major US banks and demonstrate the vulnerability of check recognitions for mobile deposits. These systems are substantially harder to fool since they classify both the handwritten amounts in digits and letters, independently. Nevertheless, we generalize SCAR to design attacks that fool state-of-the-art check processing systems using unnoticeable perturbations that lead to misclassification of deposit amounts. Consequently, this is a powerful method to perform financial fraud."
                    }
                ]
            },
            "a2a23d4a-6e0f-4065-be89-b250e83177b0": {
                "pk": "a2a23d4a-6e0f-4065-be89-b250e83177b0",
                "name": "Vasilis Gkatzelis",
                "collaborators": [
                    "Daniel Schoepflin",
                    "Xizhi Tan",
                    "Richard Cole",
                    "Gagan Goel",
                    "Alexandros Psomas",
                    "Emmanouil Pountourakis",
                    "Alkmini Sgouritsa",
                    "Artem Baklanov",
                    "Pranav Garimidi",
                    "Nisarg Shah"
                ],
                "domain": [
                    "Mechanism Design",
                    "Fair Division",
                    "Game Theory",
                    "Algorithmic Fairness"
                ],
                "publications": [
                    {
                        "title": "Mechanism Design for Fair Division",
                        "abstract": "We revisit the classic problem of fair division from a mechanism design perspective, using {\\em Proportional Fairness} as a benchmark. In particular, we aim to allocate a collection of divisible items to a set of agents while incentivizing the agents to be truthful in reporting their valuations. For the very large class of homogeneous valuations, we design a truthful mechanism that provides {\\em every agent} with at least a $1/e\\approx 0.368$ fraction of her Proportionally Fair valuation. To complement this result, we show that no truthful mechanism can guarantee more than a $0.5$ fraction, even for the restricted class of additive linear valuations. We also propose another mechanism for additive linear valuations that works really well when every item is highly demanded. To guarantee truthfulness, our mechanisms discard a carefully chosen fraction of the allocated resources; we conclude by uncovering interesting connections between our mechanisms and known mechanisms that use money instead."
                    },
                    {
                        "title": "Fair and Efficient Online Allocations with Normalized Valuations",
                        "abstract": "A set of divisible resources becomes available over a sequence of rounds and needs to be allocated immediately and irrevocably. Our goal is to distribute these resources to maximize fairness and efficiency. Achieving any non-trivial guarantees in an adversarial setting is impossible. However, we show that normalizing the agent values, a very common assumption in fair division, allows us to escape this impossibility. Our main result is an online algorithm for the case of two agents that ensures the outcome is envy-free while guaranteeing 91.6% of the optimal social welfare. We also show that this is near-optimal: there is no envy-free algorithm that guarantees more than 93.3% of the optimal social welfare."
                    },
                    {
                        "title": "Truthfulness, Proportional Fairness, and Efficiency",
                        "abstract": "How does one allocate a collection of resources to a set of strategic agents in a fair and efficient manner without using money? For in many scenarios it is not feasible to use money to compensate agents for otherwise unsatisfactory outcomes. This paper studies this question, looking at both fairness and efficiency measures.   We employ the proportionally fair solution, which is a well-known fairness concept for money-free settings. But although finding a proportionally fair solution is computationally tractable, it cannot be implemented in a truthful fashion. Consequently, we seek approximate solutions. We give several truthful mechanisms which achieve proportional fairness in an approximate sense. We use a strong notion of approximation, requiring the mechanism to give each agent a good approximation of its proportionally fair utility. In particular, one of our mechanisms provides a better and better approximation factor as the minimum demand for every good increases. A motivating example is provided by the massive privatization auction in the Czech republic in the early 90s.   With regard to efficiency, prior work has shown a lower bound of 0.5 on the approximation factor of any swap-dictatorial mechanism approximating a social welfare measure even for the two agents and multiple goods case. We surpass this lower bound by designing a non-swap-dictatorial mechanism for this case. Interestingly, the new mechanism builds on the notion of proportional fairness."
                    },
                    {
                        "title": "Resource-Aware Cost-Sharing Mechanisms with Priors",
                        "abstract": "In a decentralized system with $m$ machines, we study the selfish scheduling problem where each user strategically chooses which machine to use. Each machine incurs a cost, which is a function of the total load assigned to it, and some cost-sharing mechanism distributes this cost among the machine's users. The users choose a machine aiming to minimize their own share of the cost, so the cost-sharing mechanism induces a game among them. We approach this problem from the perspective of a designer who can select which cost-sharing mechanism to use, aiming to minimize the price of anarchy (PoA) of the induced games.   Recent work introduced the class of \\emph{resource-aware} cost-sharing mechanisms, whose decisions can depend on the set of machines in the system, but are oblivious to the total number of users. These mechanisms can guarantee low PoA bounds for instances where the cost functions of the machines are all convex or concave, but can suffer from very high PoA for cost functions that deviate from these families.   In this paper we show that if we enhance the class of resource-aware mechanisms with some prior information regarding the users, then they can achieve low PoA for a much more general family of cost functions. We first show that, as long as the mechanism knows just two of the participating users, then it can assign special roles to them and ensure a constant PoA. We then extend this idea to settings where the mechanism has access to the probability with which each user is present in the system. For all these instances, we provide a mechanism that achieves an expected PoA that is logarithmic in the expected number of users."
                    },
                    {
                        "title": "Balanced Ranking with Diversity Constraints",
                        "abstract": "Many set selection and ranking algorithms have recently been enhanced with diversity constraints that aim to explicitly increase representation of historically disadvantaged populations, or to improve the overall representativeness of the selected set. An unintended consequence of these constraints, however, is reduced in-group fairness: the selected candidates from a given group may not be the best ones, and this unfairness may not be well-balanced across groups.   In this paper we study this phenomenon using datasets that comprise multiple sensitive attributes. We then introduce additional constraints, aimed at balancing the \\in-group fairness across groups, and formalize the induced optimization problems as integer linear programs. Using these programs, we conduct an experimental evaluation with real datasets, and quantify the feasible trade-offs between balance and overall performance in the presence of diversity constraints."
                    },
                    {
                        "title": "Achieving Proportionality up to the Maximin Item with Indivisible Goods",
                        "abstract": "We study the problem of fairly allocating indivisible goods and focus on the classic fairness notion of proportionality. The indivisibility of the goods is long known to pose highly non-trivial obstacles to achieving fairness, and a very vibrant line of research has aimed to circumvent them using appropriate notions of approximate fairness. Recent work has established that even approximate versions of proportionality (PROPx) may be impossible to achieve even for small instances, while the best known achievable approximations (PROP1) are much weaker. We introduce the notion of proportionality up to the maximin item (PROPm) and show how to reach an allocation satisfying this notion for any instance involving up to five agents with additive valuations. PROPm provides a well-motivated middle-ground between PROP1 and PROPx, while also capturing some elements of the well-studied maximin share (MMS) benchmark: another relaxation of proportionality that has attracted a lot of attention."
                    },
                    {
                        "title": "PROPm Allocations of Indivisible Goods to Multiple Agents",
                        "abstract": "We study the classic problem of fairly allocating a set of indivisible goods among a group of agents, and focus on the notion of approximate proportionality known as PROPm. Prior work showed that there exists an allocation that satisfies this notion of fairness for instances involving up to five agents, but fell short of proving that this is true in general. We extend this result to show that a PROPm allocation is guaranteed to exist for all instances, independent of the number of agents or goods. Our proof is constructive, providing an algorithm that computes such an allocation and, unlike prior work, the running time of this algorithm is polynomial in both the number of agents and the number of goods."
                    },
                    {
                        "title": "Prior-Free Clock Auctions for Bidders with Interdependent Values",
                        "abstract": "We study the problem of selling a good to a group of bidders with interdependent values in a prior-free setting. Each bidder has a signal that can take one of $k$ different values, and her value for the good is a weakly increasing function of all the bidders' signals. The bidders are partitioned into $\\ell$ expertise-groups, based on how their signal can impact the values for the good, and we prove upper and lower bounds regarding the approximability of social welfare and revenue for a variety of settings, parameterized by $k$ and $\\ell$. Our lower bounds apply to all ex-post incentive compatible mechanisms and our upper bounds are all within a small constant of the lower bounds. Our main results take the appealing form of ascending clock auctions and provide strong incentives by admitting the desired outcomes as obvious ex-post equilibria."
                    },
                    {
                        "title": "EFx Budget-Feasible Allocations with High Nash Welfare",
                        "abstract": "We study the problem of allocating indivisible items to budget-constrained agents, aiming to provide fairness and efficiency guarantees. Specifically, our goal is to ensure that the resulting allocation is envy-free up to any item (EFx) while minimizing the amount of inefficiency that this needs to introduce. We first show that there exist two-agent problem instances for which no EFx allocation is Pareto efficient. We, therefore, turn to approximation and use the Nash social welfare maximizing allocation as a benchmark. For two-agent instances, we provide a procedure that always returns an EFx allocation while achieving the best possible approximation of the optimal Nash social welfare that EFx allocations can achieve. For the more complicated case of three-agent instances, we provide a procedure that guarantees EFx, while achieving a constant approximation of the optimal Nash social welfare for any number of items."
                    },
                    {
                        "title": "Learning-Augmented Robust Algorithmic Recourse",
                        "abstract": "The widespread use of machine learning models in high-stakes domains can have a major negative impact, especially on individuals who receive undesirable outcomes. Algorithmic recourse provides such individuals with suggestions of minimum-cost improvements they can make to achieve a desirable outcome in the future. However, machine learning models often get updated over time and this can cause a recourse to become invalid (i.e., not lead to the desirable outcome). The robust recourse literature aims to choose recourses that are less sensitive, even against adversarial model changes, but this comes at a higher cost. To overcome this obstacle, we initiate the study of algorithmic recourse through the learning-augmented framework and evaluate the extent to which a designer equipped with a prediction regarding future model changes can reduce the cost of recourse when the prediction is accurate (consistency) while also limiting the cost even when the prediction is inaccurate (robustness). We propose a novel algorithm for this problem, study the robustness-consistency trade-off, and analyze how prediction accuracy affects performance."
                    },
                    {
                        "title": "Online Nash Social Welfare Maximization with Predictions",
                        "abstract": "We consider the problem of allocating a set of divisible goods to $N$ agents in an online manner, aiming to maximize the Nash social welfare, a widely studied objective which provides a balance between fairness and efficiency. The goods arrive in a sequence of $T$ periods and the value of each agent for a good is adversarially chosen when the good arrives. We first observe that no online algorithm can achieve a competitive ratio better than the trivial $O(N)$, unless it is given additional information about the agents' values.   Then, in line with the emerging area of \"algorithms with predictions\", we consider a setting where for each agent, the online algorithm is only given a prediction of her monopolist utility, i.e., her utility if all goods were given to her alone (corresponding to the sum of her values over the $T$ periods). Our main result is an online algorithm whose competitive ratio is parameterized by the multiplicative errors in these predictions. The algorithm achieves a competitive ratio of $O(\\log N)$ and $O(\\log T)$ if the predictions are perfectly accurate. Moreover, the competitive ratio degrades smoothly with the errors in the predictions, and is surprisingly robust: the logarithmic competitive ratio holds even if the predictions are very inaccurate. We complement this positive result by showing that our bounds are essentially tight: no online algorithm, even if provided with perfectly accurate predictions, can achieve a competitive ratio of $O(\\log^{1-\\epsilon} N)$ or $O(\\log^{1-\\epsilon} T)$ for any constant $\\epsilon>0$."
                    },
                    {
                        "title": "Beyond Cake Cutting: Allocating Homogeneous Divisible Goods",
                        "abstract": "The problem of fair division known as \"cake cutting\" has been the focus of multiple papers spanning several decades. The most prominent problem in this line of work has been to bound the query complexity of computing an envy-free outcome in the Robertson-Webb query model. However, the root of this problem's complexity is somewhat artificial: the agents' values are assumed to be additive across different pieces of the \"cake\" but infinitely complicated within each piece. This is unrealistic in most of the motivating examples, where the cake represents a finite collection of homogeneous goods.   We address this issue by introducing a fair division model that more accurately captures these applications: the value that an agent gains from a given good depends only on the amount of the good they receive, yet it can be an arbitrary function of this amount, allowing the agents to express preferences that go beyond standard cake cutting. In this model, we study the query complexity of computing allocations that are not just envy-free, but also approximately Pareto optimal among all envy-free allocations. Using a novel flow-based approach, we show that we can encode the ex-post feasibility of randomized allocations via a polynomial number of constraints, which reduces our problem to solving a linear program."
                    },
                    {
                        "title": "Bayesian and Randomized Clock Auctions",
                        "abstract": "In a single-parameter mechanism design problem, a provider is looking to sell a service to a group of potential buyers. Each buyer $i$ has a private value $v_i$ for receiving the service and a feasibility constraint restricts which sets of buyers can be served simultaneously. Recent work in economics introduced clock auctions as a superior class of auctions for this problem, due to their transparency, simplicity, and strong incentive guarantees. Subsequent work focused on evaluating the social welfare approximation guarantees of these auctions, leading to strong impossibility results: in the absence of prior information regarding the buyers' values, no deterministic clock auction can achieve a bounded approximation, even for simple feasibility constraints with only two maximal feasible sets.   We show that these negative results can be circumvented by using prior information or by leveraging randomization. We provide clock auctions that give a $O(\\log\\log k)$ approximation for general downward-closed feasibility constraints with $k$ maximal feasible sets for three different information models, ranging from full access to the value distributions to complete absence of information. The more information the seller has, the simpler these auctions are. Under full access, we use a particularly simple deterministic clock auction, called a single-price clock auction, which is only slightly more complex than posted price mechanisms. In this auction, each buyer is offered a single price and a feasible set is selected among those who accept their offers. In the other extreme, where no prior information is available, this approximation guarantee is obtained using a complex randomized clock auction. In addition to our main results, we propose a parameterization that interpolates between single-price clock auctions and general clock auctions, paving the way for an exciting line of future research."
                    },
                    {
                        "title": "Improved Price of Anarchy via Predictions",
                        "abstract": "A central goal in algorithmic game theory is to analyze the performance of decentralized multiagent systems, like communication and information networks. In the absence of a central planner who can enforce how these systems are utilized, the users can strategically interact with the system, aiming to maximize their own utility, possibly leading to very inefficient outcomes, and thus a high price of anarchy. To alleviate this issue, the system designer can use decentralized mechanisms that regulate the use of each resource (e.g., using local queuing protocols or scheduling mechanisms), but with only limited information regarding the state of the system. These information limitations have a severe impact on what such decentralized mechanisms can achieve, so most of the success stories in this literature have had to make restrictive assumptions (e.g., by either restricting the structure of the networks or the types of cost functions).   In this paper, we overcome some of the obstacles that the literature has imposed on decentralized mechanisms, by designing mechanisms that are enhanced with predictions regarding the missing information. Specifically, inspired by the big success of the literature on \"algorithms with predictions\", we design decentralized mechanisms with predictions and evaluate their price of anarchy as a function of the prediction error, focusing on two very well-studied classes of games: scheduling games and multicast network formation games."
                    },
                    {
                        "title": "Strategyproof Scheduling with Predictions",
                        "abstract": "In their seminal paper that initiated the field of algorithmic mechanism design, \\citet{NR99} studied the problem of designing strategyproof mechanisms for scheduling jobs on unrelated machines aiming to minimize the makespan. They provided a strategyproof mechanism that achieves an $n$-approximation and they made the bold conjecture that this is the best approximation achievable by any deterministic strategyproof scheduling mechanism. After more than two decades and several efforts, $n$ remains the best known approximation and very recent work by \\citet{CKK21} has been able to prove an $\\Omega(\\sqrt{n})$ approximation lower bound for all deterministic strategyproof mechanisms. This strong negative result, however, heavily depends on the fact that the performance of these mechanisms is evaluated using worst-case analysis. To overcome such overly pessimistic, and often uninformative, worst-case bounds, a surge of recent work has focused on the ``learning-augmented framework'', whose goal is to leverage machine-learned predictions to obtain improved approximations when these predictions are accurate (consistency), while also achieving near-optimal worst-case approximations even when the predictions are arbitrarily wrong (robustness).   In this work, we study the classic strategic scheduling problem of~\\citet{NR99} using the learning-augmented framework and give a deterministic polynomial-time strategyproof mechanism that is $6$-consistent and $2n$-robust. We thus achieve the ``best of both worlds'': an $O(1)$ consistency and an $O(n)$ robustness that asymptotically matches the best-known approximation. We then extend this result to provide more general worst-case approximation guarantees as a function of the prediction error. Finally, we complement our positive results by showing that any $1$-consistent deterministic strategyproof mechanism has unbounded robustness."
                    },
                    {
                        "title": "Clock Auctions Augmented with Unreliable Advice",
                        "abstract": "We provide the first analysis of clock auctions through the learning-augmented framework. Deferred-acceptance clock auctions are a compelling class of mechanisms satisfying a unique list of highly practical properties, including obvious strategy-proofness, transparency, and unconditional winner privacy, making them particularly well-suited for real-world applications. However, early work that evaluated their performance from a worst-case analysis standpoint concluded that no deterministic clock auction can achieve much better than an $O(\\log n)$ approximation of the optimal social welfare (where $n$ is the number of bidders participating in the auction), even in seemingly very simple settings.   To overcome this overly pessimistic impossibility result, which heavily depends on the assumption that the designer has no information regarding the preferences of the participating bidders, we leverage the learning-augmented framework. This framework assumes that the designer is provided with some advice regarding what the optimal solution may be. This advice may be the product of machine-learning algorithms applied to historical data, so it can provide very useful guidance, but it can also be highly unreliable.   Our main results are learning-augmented clock auctions that use this advice to achieve much stronger performance guarantees whenever the advice is accurate (known as consistency), while simultaneously maintaining worst-case guarantees even if this advice is arbitrarily inaccurate (known as robustness). Specifically, for the standard notion of consistency, we provide a clock auction that achieves the best of both worlds: $(1+\\epsilon)$-consistency for any constant $\\epsilon > 0$ and $O(\\log n)$ robustness. We then also consider a much stronger notion of consistency and provide an auction that achieves the optimal trade-off between this notion of consistency and robustness."
                    },
                    {
                        "title": "Nash Social Welfare Approximation for Strategic Agents",
                        "abstract": "The fair division of resources is an important age-old problem that has led to a rich body of literature. At the center of this literature lies the question of whether there exist fair mechanisms despite strategic behavior of the agents. A fundamental objective function used for measuring fair outcomes is the Nash social welfare, defined as the geometric mean of the agent utilities. This objective function is maximized by widely known solution concepts such as Nash bargaining and the competitive equilibrium with equal incomes. In this work we focus on the question of (approximately) implementing the Nash social welfare. The starting point of our analysis is the Fisher market, a fundamental model of an economy, whose benchmark is precisely the (weighted) Nash social welfare. We begin by studying two extreme classes of valuations functions, namely perfect substitutes and perfect complements, and find that for perfect substitutes, the Fisher market mechanism has a constant approximation: at most 2 and at least e1e. However, for perfect complements, the Fisher market does not work well, its bound degrading linearly with the number of players.   Strikingly, the Trading Post mechanism---an indirect market mechanism also known as the Shapley-Shubik game---has significantly better performance than the Fisher market on its own benchmark. Not only does Trading Post achieve an approximation of 2 for perfect substitutes, but this bound holds for all concave utilities and becomes arbitrarily close to optimal for Leontief utilities (perfect complements), where it reaches $(1+\\epsilon)$ for every $\\epsilon > 0$. Moreover, all the Nash equilibria of the Trading Post mechanism are pure for all concave utilities and satisfy an important notion of fairness known as proportionality."
                    },
                    {
                        "title": "Resolving the Optimal Metric Distortion Conjecture",
                        "abstract": "We study the following metric distortion problem: there are two finite sets of points, $V$ and $C$, that lie in the same metric space, and our goal is to choose a point in $C$ whose total distance from the points in $V$ is as small as possible. However, rather than having access to the underlying distance metric, we only know, for each point in $V$, a ranking of its distances to the points in $C$. We propose algorithms that choose a point in $C$ using only these rankings as input and we provide bounds on their \\emph{distortion} (worst-case approximation ratio). A prominent motivation for this problem comes from voting theory, where $V$ represents a set of voters, $C$ represents a set of candidates, and the rankings correspond to ordinal preferences of the voters. A major conjecture in this framework is that the optimal deterministic algorithm has distortion $3$. We resolve this conjecture by providing a polynomial-time algorithm that achieves distortion $3$, matching a known lower bound. We do so by proving a novel lemma about matching voters to candidates, which we refer to as the \\emph{ranking-matching lemma}. This lemma induces a family of novel algorithms, which may be of independent interest, and we show that a special algorithm in this family achieves distortion $3$. We also provide more refined, parameterized, bounds using the notion of $\\alpha$-decisiveness, which quantifies the extent to which a voter may prefer her top choice relative to all others. Finally, we introduce a new randomized algorithm with improved distortion compared to known results, and also provide improved lower bounds on the distortion of all deterministic and randomized algorithms."
                    },
                    {
                        "title": "Best of Both Distortion Worlds",
                        "abstract": "We study the problem of designing voting rules that take as input the ordinal preferences of $n$ agents over a set of $m$ alternatives and output a single alternative, aiming to optimize the overall happiness of the agents. The input to the voting rule is each agent's ranking of the alternatives from most to least preferred, yet the agents have more refined (cardinal) preferences that capture the intensity with which they prefer one alternative over another. To quantify the extent to which voting rules can optimize over the cardinal preferences given access only to the ordinal ones, prior work has used the distortion measure, i.e., the worst-case approximation ratio between a voting rule's performance and the best performance achievable given the cardinal preferences.   The work on the distortion of voting rules has been largely divided into two worlds: utilitarian distortion and metric distortion. In the former, the cardinal preferences of the agents correspond to general utilities and the goal is to maximize a normalized social welfare. In the latter, the agents' cardinal preferences correspond to costs given by distances in an underlying metric space and the goal is to minimize the (unnormalized) social cost. Several deterministic and randomized voting rules have been proposed and evaluated for each of these worlds separately, gradually improving the achievable distortion bounds, but none of the known voting rules perform well in both worlds simultaneously.   In this work, we prove that one can achieve the best of both worlds by designing new voting rules, that simultaneously achieve near-optimal distortion guarantees in both distortion worlds. We also prove that this positive result does not generalize to the case where the voting rule is provided with the rankings of only the top-$t$ alternatives of each agent, for $t<m$."
                    }
                ]
            },
            "494c93f1-f596-4ba1-88a3-424f61cc3782": {
                "pk": "494c93f1-f596-4ba1-88a3-424f61cc3782",
                "name": "Golnoosh Shahkarami",
                "collaborators": [
                    "Kurt Mehlhorn",
                    "Hannaneh Akrami",
                    "Quentin Vermande",
                    "Bhaskar Ray Chaudhury",
                    "Themis Gouleakis",
                    "Martin Hoefer",
                    "Marco Schmalhofer",
                    "Giovanna Varricchio",
                    "Ernest van Wijland",
                    "Antonios Antoniadis"
                ],
                "domain": [
                    "Algorithm Design",
                    "Online Algorithms",
                    "Game Theory",
                    "Network Optimization"
                ],
                "publications": [
                    {
                        "title": "A Novel Prediction Setup for Online Speed-Scaling",
                        "abstract": "Given the rapid rise in energy demand by data centers and computing systems in general, it is fundamental to incorporate energy considerations when designing (scheduling) algorithms. Machine learning can be a useful approach in practice by predicting the future load of the system based on, for example, historical data. However, the effectiveness of such an approach highly depends on the quality of the predictions and can be quite far from optimal when predictions are sub-par. On the other hand, while providing a worst-case guarantee, classical online algorithms can be pessimistic for large classes of inputs arising in practice.   This paper, in the spirit of the new area of machine learning augmented algorithms, attempts to obtain the best of both worlds for the classical, deadline based, online speed-scaling problem: Based on the introduction of a novel prediction setup, we develop algorithms that (i) obtain provably low energy-consumption in the presence of adequate predictions, and (ii) are robust against inadequate predictions, and (iii) are smooth, i.e., their performance gradually degrades as the prediction error increases."
                    },
                    {
                        "title": "Randomized and Deterministic Maximin-share Approximations for Fractionally Subadditive Valuations",
                        "abstract": "We consider the problem of guaranteeing maximin-share (MMS) when allocating a set of indivisible items to a set of agents with fractionally subadditive (XOS) valuations. For XOS valuations, it has been previously shown that for some instances no allocation can guarantee a fraction better than $1/2$ of maximin-share to all the agents. Also, a deterministic allocation exists that guarantees $0.219225$ of the maximin-share of each agent. Our results involve both deterministic and randomized allocations. On the deterministic side, we improve the best approximation guarantee for fractionally subadditive valuations to $3/13 = 0.230769$. We develop new ideas on allocating large items in our allocation algorithm which might be of independent interest. Furthermore, we investigate randomized algorithms and the Best-of-both-worlds fairness guarantees. We propose a randomized allocation that is $1/4$-MMS ex-ante and $1/8$-MMS ex-post for XOS valuations. Moreover, we prove an upper bound of $3/4$ on the ex-ante guarantee for this class of valuations."
                    },
                    {
                        "title": "Learning-Augmented Algorithms for Online TSP on the Line",
                        "abstract": "We study the online Traveling Salesman Problem (TSP) on the line augmented with machine-learned predictions. In the classical problem, there is a stream of requests released over time along the real line. The goal is to minimize the makespan of the algorithm. We distinguish between the open variant and the closed one, in which we additionally require the algorithm to return to the origin after serving all requests. The state of the art is a $1.64$-competitive algorithm and a $2.04$-competitive algorithm for the closed and open variants, respectively \\cite{Bjelde:1.64}. In both cases, a tight lower bound is known \\cite{Ausiello:1.75, Bjelde:1.64}.   In both variants, our primary prediction model involves predicted positions of the requests. We introduce algorithms that (i) obtain a tight 1.5 competitive ratio for the closed variant and a 1.66 competitive ratio for the open variant in the case of perfect predictions, (ii) are robust against unbounded prediction error, and (iii) are smooth, i.e., their performance degrades gracefully as the prediction error increases.   Moreover, we further investigate the learning-augmented setting in the open variant by additionally considering a prediction for the last request served by the optimal offline algorithm. Our algorithm for this enhanced setting obtains a 1.33 competitive ratio with perfect predictions while also being smooth and robust, beating the lower bound of 1.44 we show for our original prediction setting for the open variant. Also, we provide a lower bound of 1.25 for this enhanced setting."
                    },
                    {
                        "title": "Nash Social Welfare for 2-value Instances",
                        "abstract": "This paper is merged with arXiv:2107.08965v2. We refer the reader to the full and updated version.   We study the problem of allocating a set of indivisible goods among agents with 2-value additive valuations. Our goal is to find an allocation with maximum Nash social welfare, i.e., the geometric mean of the valuations of the agents. We give a polynomial-time algorithm to find a Nash social welfare maximizing allocation when the valuation functions are integrally 2-valued, i.e., each agent has a value either $1$ or $p$ for each good, for some positive integer $p$. We then extend our algorithm to find a better approximation factor for general 2-value instances."
                    },
                    {
                        "title": "Learning-Augmented Online TSP on Rings, Trees, Flowers and (almost) Everywhere Else",
                        "abstract": "We study the Online Traveling Salesperson Problem (OLTSP) with predictions. In OLTSP, a sequence of initially unknown requests arrive over time at points (locations) of a metric space. The goal is, starting from a particular point of the metric space (the origin), to serve all these requests while minimizing the total time spent. The server moves with unit speed or is \"waiting\" (zero speed) at some location. We consider two variants: in the open variant, the goal is achieved when the last request is served. In the closed one, the server additionally has to return to the origin. We adopt a prediction model, introduced for OLTSP on the line, in which the predictions correspond to the locations of the requests and extend it to more general metric spaces.   We first propose an oracle-based algorithmic framework, inspired by previous work. This framework allows us to design online algorithms for general metric spaces that provide competitive ratio guarantees which, given perfect predictions, beat the best possible classical guarantee (consistency). Moreover, they degrade gracefully along with the increase in error (smoothness), but always within a constant factor of the best known competitive ratio in the classical case (robustness).   Having reduced the problem to designing suitable efficient oracles, we describe how to achieve this for general metric spaces as well as specific metric spaces (rings, trees and flowers), the resulting algorithms being tractable in the latter case. The consistency guarantees of our algorithms are tight in almost all cases, and their smoothness guarantees only suffer a linear dependency on the error, which we show is necessary. Finally, we provide robustness guarantees improving previous results."
                    },
                    {
                        "title": "Complexity of Computing the Anti-Ramsey Numbers for Paths",
                        "abstract": "The anti-Ramsey numbers are a fundamental notion in graph theory, introduced in 1978, by Erd\\\" os, Simonovits and S\\' os. For given graphs $G$ and $H$ the \\emph{anti-Ramsey number} $\\textrm{ar}(G,H)$ is defined to be the maximum number $k$ such that there exists an assignment of $k$ colors to the edges of $G$ in which every copy of $H$ in $G$ has at least two edges with the same color.   There are works on the computational complexity of the problem when $H$ is a star. Along this line of research, we study the complexity of computing the anti-Ramsey number $\\textrm{ar}(G,P_k)$, where $P_k$ is a path of length $k$. First, we observe that when $k = \\Omega(n)$, the problem is hard; hence, the challenging part is the computational complexity of the problem when $k$ is a fixed constant.   We provide a characterization of the problem for paths of constant length. Our first main contribution is to prove that computing $\\textrm{ar}(G,P_k)$ for every integer $k>2$ is NP-hard. We obtain this by providing several structural properties of such coloring in graphs. We investigate further and show that approximating $\\textrm{ar}(G,P_3)$ to a factor of $n^{-1/2 - \\epsilon}$ is hard already in $3$-partite graphs, unless P=NP. We also study the exact complexity of the precolored version and show that there is no subexponential algorithm for the problem unless ETH fails for any fixed constant $k$.   Given the hardness of approximation and parametrization of the problem, it is natural to study the problem on restricted graph families. We introduce the notion of color connected coloring and employing this structural property. We obtain a linear time algorithm to compute $\\textrm{ar}(G,P_k)$, for every integer $k$, when the host graph, $G$, is a tree."
                    },
                    {
                        "title": "Maximizing Nash Social Welfare in 2-Value Instances: Delineating Tractability",
                        "abstract": "We study the problem of allocating a set of indivisible goods among a set of agents with \\emph{2-value additive valuations}. In this setting, each good is valued either $1$ or $\\sfrac{p}{q}$, for some fixed co-prime numbers $p,q\\in \\NN$ such that $1\\leq q < p$. Our goal is to find an allocation maximizing the \\emph{Nash social welfare} (\\NSW), i.e., the geometric mean of the valuations of the agents. In this work, we give a complete characterization of polynomial-time tractability of \\NSW\\ maximization that solely depends on the values of $q$.   We start by providing a rather simple polynomial-time algorithm to find a maximum \\NSW\\ allocation when the valuation functions are \\emph{integral}, that is, $q=1$. We then exploit more involved techniques to get an algorithm producing a maximum \\NSW\\ allocation for the \\emph{half-integral} case, that is, $q=2$. Finally, we show it is \\classNP-hard to compute an allocation with maximum \\NSW\\ whenever $q\\geq3$."
                    },
                    {
                        "title": "Maximizing Nash Social Welfare in 2-Value Instances",
                        "abstract": "We consider the problem of maximizing the Nash social welfare when allocating a set $\\mathcal{G}$ of indivisible goods to a set $\\mathcal{N}$ of agents. We study instances, in which all agents have 2-value additive valuations: The value of every agent $i \\in \\mathcal{N}$ for every good $j \\in \\mathcal{G}$ is $v_{ij} \\in \\{p,q\\}$, for $p,q \\in \\mathbb{N}$, $p \\le q$. Maybe surprisingly, we design an algorithm to compute an optimal allocation in polynomial time if $p$ divides $q$, i.e., when $p=1$ and $q \\in \\mathbb{N}$ after appropriate scaling. The problem is \\classNP-hard whenever $p$ and $q$ are coprime and $p \\ge 3$.   In terms of approximation, we present positive and negative results for general $p$ and $q$. We show that our algorithm obtains an approximation ratio of at most 1.0345. Moreover, we prove that the problem is \\classAPX-hard, with a lower bound of $1.000015$ achieved at $p/q = 4/5$."
                    },
                    {
                        "title": "Physarum-Inspired Multi-Commodity Flow Dynamics",
                        "abstract": "In wet-lab experiments, the slime mold Physarum polycephalum has demonstrated its ability to solve shortest path problems and to design efficient networks. For the shortest path problem, a mathematical model for the evolution of the slime is available and it has been shown in computer experiments and through mathematical analysis that the dynamics solves the shortest path problem. In this paper, we introduce a dynamics for the network design problem. We formulate network design as the problem of constructing a network that efficiently supports a multi-commodity flow problem. We investigate the dynamics in computer simulations and analytically. The simulations show that the dynamics is able to construct efficient and elegant networks. In the theoretical part we show that the dynamics minimizes an objective combining the cost of the network and the cost of routing the demands through the network. We also give alternative characterization of the optimum solution."
                    }
                ]
            }
        }
    },
    "2402.01469": {
        "paper_data": {
            "title": "AMOR: A Recipe for Building Adaptable Modular Knowledge Agents Through Process Feedback",
            "url": "http://arxiv.org/abs/2402.01469v1",
            "arxiv_id": "2402.01469",
            "authors": [
                "Jian Guan",
                "Wei Wu",
                "Zujie Wen",
                "Peng Xu",
                "Hongning Wang",
                "Minlie Huang"
            ],
            "abstract": "The notable success of large language models (LLMs) has sparked an upsurge in building language agents to complete various complex tasks. We present AMOR, an agent framework based on open-source LLMs, which reasons with external knowledge bases and adapts to specific domains through human supervision to the reasoning process. AMOR builds reasoning logic over a finite state machine (FSM) that solves problems through autonomous executions and transitions over disentangled modules. This allows humans to provide direct feedback to the individual modules, and thus naturally forms process supervision. Based on this reasoning and feedback framework, we develop AMOR through two-stage fine-tuning: warm-up and adaptation. The former fine-tunes the LLM with examples automatically constructed from various public datasets and enables AMOR to generalize across different knowledge environments, while the latter tailors AMOR to specific domains using process feedback. Extensive experiments across multiple domains demonstrate the advantage of AMOR to strong baselines, thanks to its FSM-based reasoning and process feedback mechanism.",
            "introduction": " Introduction Large language models (LLMs), with astounding performance over general natural language process- ing (NLP) problems (Wei et al., 2022a; Achiam et al., 2023; Touvron et al., 2023), have spurred great interest in building LLM-based agents to solve complex tasks by interacting with external resources such as web knowledge (Nakano et al., 2021), specialized tools (Schick et al., 2023), etc. We focus on developing agents for knowledge- intensive tasks, where the agent completes users\u2019 information-seeking requests by interacting with specific knowledge bases (Lewis et al., 2020). To address the complexity of such tasks, we posit the *Corresponding authors.desiderata for a qualifying agent as follows: Firstly, the agent should possess a robust reasoning logic about the task to solve individual problems with precise pathways. Secondly, the agent should main- tain an adaptive mechanism to adjust to specific environments, rather than staying static. Thirdly, the reasoning process should be amenable to hu- man interventions, enabling humans to steer the agent\u2019s behavior through direct feedback to the pro- cess rather than only to the outcome (Uesato et al., 2022). This ability can significantly facilitate align- ment between agent behavior and human intent. Although extensive studies have been conducted on building language agents, few, if any, can ful- fill all the required criteria due to their uncon- trollable reasoning logic, static model capability, or sparse/missing feedback signals, as detailed in Tab. 1. Consequently, it is still challenging for users to critique, and thus guide existing agents to follow targeted manners, especially when the agents are built upon less powerful LLMs (Liu et al., 2023b). We introduce an Adaptable MOdula Rknowledge agent ( AMOR ) that can reason and adapt, with the reasoning process amenable to human supervision, based on open-source LLMs. AMOR \u2019s reasoning logic is formalized as a finite state machine (FSM) (Clarke et al., 1986; Lee and Yannakakis, 1996) that solves problems via a series of executions and transitions over a set of modules (Fig. 1). This naturally enables the desired process-based super- vision mechanism, allowing users to give feedback to each LLM-controlled module. AMOR supports flexible forms of feedback, either binary judgments regarding the correctness or refinement of the out- puts. The reasoning logic and process feedback mechanism together frame how AMOR thinks, acts, and interacts with users and task environments. The training in AMOR happens in two stages: (1) Warm-up: the modular design enables us to construct training data separately for each disen- tangled module without requiring complete trajec- 1arXiv:2402.01469v1  [cs.CL]  2 Feb 2024MethodsReasoning LogicAdaptive Mechanism FeedbackStep Inter-Step Dependency WebGPT (Nakano et al., 2021) Tool Invoking Undefined Imitation Learning from Humans Outcome CoT (Wei et al., 2022b) Reasoning Undefined Prompting Undefined ToT (Yao et al., 2023a) Reasoning Undefined Prompting Process ReAct (Yao et al., 2023b) Reasoning&Tool Invoking Undefined Prompting Undefined Reflexion (Shinn et al., 2023) Reasoning&Tool Invoking Undefined Prompting Process AgentLM (Zeng et al., 2023) Reasoning&Tool Invoking Undefined Imitation Learning from LLMs Outcome MetaGPT (Hong et al., 2023) Specialized Module Pipeline Prompting Process LUMOS (Yin et al., 2023) Specialized Module Pipeline Imitation Learning from Humans Undefined AMOR Specialized Module Finite State Machine Exploration&Exploitation Process Tab. 1: Comparison between A MOR and representative methods of testing finite state machines-a survey. Proceedings of the IEEE , 84(8):1090\u20131123. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Hein- rich K\u00fcttler, Mike Lewis, Wen-tau Yih, Tim Rock- t\u00e4schel, et al. 2020. Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neu- ral Information Processing Systems , 33:9459\u20139474. Hunter Lightman, Vineet Kosaraju, Yura Burda, Harri Edwards,",
            "references": [
                {
                    "title": "In-context Learning with Retrieved Demonstrations for Language Models: A Survey",
                    "abstract": "Language models, especially pre-trained large language models, have showcased remarkable abilities as few-shot in-context learners (ICL), adept at adapting to new tasks with just a few demonstrations in the input context. However, the model's ability to perform ICL is sensitive to the choice of the few-shot demonstrations. Instead of using a fixed set of demonstrations, one recent development is to retrieve demonstrations tailored to each input query. The implementation of demonstration retrieval is relatively straightforward, leveraging existing databases and retrieval systems. This not only improves the efficiency and scalability of the learning process but also has been shown to reduce biases inherent in manual example selection. In light of the encouraging results and growing research in ICL with retrieved demonstrations, we conduct an extensive review of studies in this area. In this survey, we discuss and compare different design choices for retrieval models, retrieval training procedures, and inference algorithms."
                },
                {
                    "title": "Self-Contradictory Reasoning Evaluation and Detection",
                    "abstract": "In a plethora of recent work, large language models (LLMs) demonstrated impressive reasoning ability, but many proposed downstream reasoning tasks only focus on final answers. Two fundamental questions persist: 1) how consistent is the reasoning, and 2) can models detect unreliable reasoning? In this paper, we investigate self-contradictory (Self-Contra) reasoning, where the model reasoning does not support its answers. To answer 1), we define and assess the Self-Contra rate across three datasets and delve into finer-grained categories of Self-Contra reasoning. We find that LLMs often contradict themselves in reasoning tasks involving contextual information understanding or commonsense. The model may generate correct answers by taking shortcuts in reasoning or overlooking contextual evidence, leading to compromised reasoning. For 2), we task the state-of-the-art model GPT-4 with identifying Self-Contra reasoning and finer-grained fallacies. We find that finer-grained categories enhanced detection can improve GPT-4's ability to detect Self-Contra. However, it is only able to detect Self-Contra with a 52.2% F1 score, much lower compared to 66.7% for humans. Our results indicate that current LLMs lack the robustness necessary for reliable reasoning and we emphasize the urgent need for establishing best practices in comprehensive reasoning evaluations beyond pure performance-based metrics."
                },
                {
                    "title": "Agent Lumos: Unified and Modular Training for Open-Source Language Agents",
                    "abstract": "Closed-source agents suffer from several issues such as a lack of affordability, transparency, and reproducibility, particularly on complex interactive tasks. This motivates the development of open-source alternatives. We introduce LUMOS, one of the first frameworks for training open-source LLM-based agents. LUMOS features a learnable, unified, and modular architecture with a planning module that learns high-level subgoal generation, and a grounding module trained to translate these into actions using various tools in the execution module. The design allows for modular upgrades and wider applicability to diverse interactive tasks. To foster generalizable agent learning, we collect large-scale, unified, and high-quality training annotations derived from diverse ground-truth reasoning rationales across various complex interactive tasks. On 9 datasets, LUMOS exhibits several key advantages: (1) LUMOS excels multiple larger open-source agents on the held-out datasets (unused for training) for each task type. LUMOS even surpasses GPT agents on QA and web tasks; (2) LUMOS outperforms open-source agents produced by chain-of-thoughts and unmodularized integrated training; and (3) LUMOS effectively generalizes to unseen tasks, outperforming 33B-scale agents and domain-specific agents."
                },
                {
                    "title": "AgentTuning: Enabling Generalized Agent Abilities for LLMs",
                    "abstract": "Open large language models (LLMs) with great performance in various tasks have significantly advanced the development of LLMs. However, they are far inferior to commercial models such as ChatGPT and GPT-4 when acting as agents to tackle complex tasks in the real world. These agent tasks employ LLMs as the central controller responsible for planning, memorization, and tool utilization, necessitating both fine-grained prompting methods and robust LLMs to achieve satisfactory performance. Though many prompting methods have been proposed to complete particular agent tasks, there is lack of research focusing on improving the agent capabilities of LLMs themselves without compromising their general abilities. In this work, we present AgentTuning, a simple and general method to enhance the agent abilities of LLMs while maintaining their general LLM capabilities. We construct AgentInstruct, a lightweight instruction-tuning dataset containing high-quality interaction trajectories. We employ a hybrid instruction-tuning strategy by combining AgentInstruct with open-source instructions from general domains. AgentTuning is used to instruction-tune the Llama 2 series, resulting in AgentLM. Our evaluations show that AgentTuning enables LLMs' agent capabilities without compromising general abilities. The AgentLM-70B is comparable to GPT-3.5-turbo on unseen agent tasks, demonstrating generalized agent capabilities. We open source the AgentInstruct and AgentLM-7B, 13B, and 70B models at https://github.com/THUDM/AgentTuning, serving open and powerful alternatives to commercial LLMs for agent tasks."
                },
                {
                    "title": "Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection",
                    "abstract": "Despite their remarkable capabilities, large language models (LLMs) often produce responses containing factual inaccuracies due to their sole reliance on the parametric knowledge they encapsulate. Retrieval-Augmented Generation (RAG), an ad hoc approach that augments LMs with retrieval of relevant knowledge, decreases such issues. However, indiscriminately retrieving and incorporating a fixed number of retrieved passages, regardless of whether retrieval is necessary, or passages are relevant, diminishes LM versatility or can lead to unhelpful response generation. We introduce a new framework called Self-Reflective Retrieval-Augmented Generation (Self-RAG) that enhances an LM's quality and factuality through retrieval and self-reflection. Our framework trains a single arbitrary LM that adaptively retrieves passages on-demand, and generates and reflects on retrieved passages and its own generations using special tokens, called reflection tokens. Generating reflection tokens makes the LM controllable during the inference phase, enabling it to tailor its behavior to diverse task requirements. Experiments show that Self-RAG (7B and 13B parameters) significantly outperforms state-of-the-art LLMs and retrieval-augmented models on a diverse set of tasks. Specifically, Self-RAG outperforms ChatGPT and retrieval-augmented Llama2-chat on Open-domain QA, reasoning and fact verification tasks, and it shows significant gains in improving factuality and citation accuracy for long-form generations relative to these models."
                },
                {
                    "title": "FireAct: Toward Language Agent Fine-tuning",
                    "abstract": "Recent efforts have augmented language models (LMs) with external tools or environments, leading to the development of language agents that can reason and act. However, most of these agents rely on few-shot prompting techniques with off-the-shelf LMs. In this paper, we investigate and argue for the overlooked direction of fine-tuning LMs to obtain language agents. Using a setup of question answering (QA) with a Google search API, we explore a variety of base LMs, prompting methods, fine-tuning data, and QA tasks, and find language agents are consistently improved after fine-tuning their backbone LMs. For example, fine-tuning Llama2-7B with 500 agent trajectories generated by GPT-4 leads to a 77% HotpotQA performance increase. Furthermore, we propose FireAct, a novel approach to fine-tuning LMs with trajectories from multiple tasks and prompting methods, and show having more diverse fine-tuning data can further improve agents. Along with other findings regarding scaling effects, robustness, generalization, efficiency and cost, our work establishes comprehensive benefits of fine-tuning LMs for agents, and provides an initial set of experimental designs, insights, as well as open questions toward language agent fine-tuning."
                },
                {
                    "title": "MetaGPT: Meta Programming for Multi-Agent Collaborative Framework",
                    "abstract": "Recently, remarkable progress has been made in automated task-solving through the use of multi-agent driven by large language models (LLMs). However, existing LLM-based multi-agent works primarily focus on solving simple dialogue tasks, and complex tasks are rarely studied, mainly due to the LLM hallucination problem. This type of hallucination becomes cascading when naively chaining multiple intelligent agents, resulting in a failure to effectively address complex problems. Therefore, we introduce MetaGPT, an innovative framework that incorporates efficient human workflows as a meta programming approach into LLM-based multi-agent collaboration. Specifically, MetaGPT encodes Standardized Operating Procedures (SOPs) into prompts to enhance structured coordination. Subsequently, it mandates modular outputs, empowering agents with domain expertise comparable to human professionals, to validate outputs and minimize compounded errors. In this way, MetaGPT leverages the assembly line paradigm to assign diverse roles to various agents, thereby establishing a framework that can effectively and cohesively deconstruct complex multi-agent collaborative problems. Our experiments on collaborative software engineering benchmarks demonstrate that MetaGPT generates more coherent and correct solutions compared to existing chat-based multi-agent systems. This highlights the potential of integrating human domain knowledge into multi-agent systems, thereby creating new opportunities to tackle complex real-world challenges. The GitHub repository of this project is publicly available on:https://github.com/geekan/MetaGPT."
                },
                {
                    "title": "ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs",
                    "abstract": "Despite the advancements of open-source large language models (LLMs), e.g., LLaMA, they remain significantly limited in tool-use capabilities, i.e., using external tools (APIs) to fulfill human instructions. The reason is that current instruction tuning largely focuses on basic language tasks but ignores the tool-use domain. This is in contrast to the excellent tool-use capabilities of state-of-the-art (SOTA) closed-source LLMs, e.g., ChatGPT. To bridge this gap, we introduce ToolLLM, a general tool-use framework encompassing data construction, model training, and evaluation. We first present ToolBench, an instruction-tuning dataset for tool use, which is constructed automatically using ChatGPT. Specifically, the construction can be divided into three stages: (i) API collection: we collect 16,464 real-world RESTful APIs spanning 49 categories from RapidAPI Hub; (ii) instruction generation: we prompt ChatGPT to generate diverse instructions involving these APIs, covering both single-tool and multi-tool scenarios; (iii) solution path annotation: we use ChatGPT to search for a valid solution path (chain of API calls) for each instruction. To enhance the reasoning capabilities of LLMs, we develop a novel depth-first search-based decision tree algorithm. It enables LLMs to evaluate multiple reasoning traces and expand the search space. Moreover, to evaluate the tool-use capabilities of LLMs, we develop an automatic evaluator: ToolEval. Based on ToolBench, we fine-tune LLaMA to obtain an LLM ToolLLaMA, and equip it with a neural API retriever to recommend appropriate APIs for each instruction. Experiments show that ToolLLaMA demonstrates a remarkable ability to execute complex instructions and generalize to unseen APIs, and exhibits comparable performance to ChatGPT. Our ToolLLaMA also demonstrates strong zero-shot generalization ability in an out-of-distribution tool-use dataset: APIBench."
                },
                {
                    "title": "A Real-World WebAgent with Planning, Long Context Understanding, and Program Synthesis",
                    "abstract": "Pre-trained large language models (LLMs) have recently achieved better generalization and sample efficiency in autonomous web automation. However, the performance on real-world websites has still suffered from (1) open domainness, (2) limited context length, and (3) lack of inductive bias on HTML. We introduce WebAgent, an LLM-driven agent that learns from self-experience to complete tasks on real websites following natural language instructions. WebAgent plans ahead by decomposing instructions into canonical sub-instructions, summarizes long HTML documents into task-relevant snippets, and acts on websites via Python programs generated from those. We design WebAgent with Flan-U-PaLM, for grounded code generation, and HTML-T5, new pre-trained LLMs for long HTML documents using local and global attention mechanisms and a mixture of long-span denoising objectives, for planning and summarization. We empirically demonstrate that our modular recipe improves the success on real websites by over 50%, and that HTML-T5 is the best model to solve various HTML understanding tasks; achieving 18.7% higher success rate than the prior method on MiniWoB web automation benchmark, and SoTA performance on Mind2Web, an offline task planning evaluation."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "Lost in the Middle: How Language Models Use Long Contexts",
                    "abstract": "While recent language models have the ability to take long contexts as input, relatively little is known about how well they use longer context. We analyze the performance of language models on two tasks that require identifying relevant information in their input contexts: multi-document question answering and key-value retrieval. We find that performance can degrade significantly when changing the position of relevant information, indicating that current language models do not robustly make use of information in long input contexts. In particular, we observe that performance is often highest when relevant information occurs at the beginning or end of the input context, and significantly degrades when models must access relevant information in the middle of long contexts, even for explicitly long-context models. Our analysis provides a better understanding of how language models use their input context and provides new evaluation protocols for future long-context language models."
                },
                {
                    "title": "Let's Verify Step by Step",
                    "abstract": "In recent years, large language models have greatly improved in their ability to perform complex multi-step reasoning. However, even state-of-the-art models still regularly produce logical mistakes. To train more reliable models, we can turn either to outcome supervision, which provides feedback for a final result, or process supervision, which provides feedback for each intermediate reasoning step. Given the importance of training reliable models, and given the high cost of human feedback, it is important to carefully compare the both methods. Recent work has already begun this comparison, but many questions still remain. We conduct our own investigation, finding that process supervision significantly outperforms outcome supervision for training models to solve problems from the challenging MATH dataset. Our process-supervised model solves 78% of problems from a representative subset of the MATH test set. Additionally, we show that active learning significantly improves the efficacy of process supervision. To support related research, we also release PRM800K, the complete dataset of 800,000 step-level human feedback labels used to train our best reward model."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "ReWOO: Decoupling Reasoning from Observations for Efficient Augmented Language Models",
                    "abstract": "Augmented Language Models (ALMs) blend the reasoning capabilities of Large Language Models (LLMs) with tools that allow for knowledge retrieval and action execution. Existing ALM systems trigger LLM thought processes while pulling observations from these tools in an interleaved fashion. Specifically, an LLM reasons to call an external tool, gets halted to fetch the tool's response, and then decides the next action based on all preceding response tokens. Such a paradigm, though straightforward and easy to implement, often leads to huge computation complexity from redundant prompts and repeated execution. This study addresses such challenges for the first time, proposing a modular paradigm ReWOO (Reasoning WithOut Observation) that detaches the reasoning process from external observations, thus significantly reducing token consumption. Comprehensive evaluations across six public NLP benchmarks and a curated dataset reveal consistent performance enhancements with our proposed methodology. Notably, ReWOO achieves 5x token efficiency and 4% accuracy improvement on HotpotQA, a multi-step reasoning benchmark. Furthermore, ReWOO demonstrates robustness under tool-failure scenarios. Beyond prompt efficiency, decoupling parametric modules from non-parametric tool calls enables instruction fine-tuning to offload LLMs into smaller language models, thus substantially reducing model parameters. Our illustrative work offloads reasoning ability from 175B GPT3.5 into 7B LLaMA, demonstrating the significant potential for truly efficient and scalable ALM systems."
                },
                {
                    "title": "Reflexion: language agents with verbal reinforcement learning",
                    "abstract": "Large language models (LLMs) have been increasingly used to interact with external environments (e.g., games, compilers, APIs) as goal-driven agents. However, it remains challenging for these language agents to quickly and efficiently learn from trial-and-error as traditional reinforcement learning methods require extensive training samples and expensive model fine-tuning. We propose Reflexion, a novel framework to reinforce language agents not by updating weights, but instead through linguistic feedback. Concretely, Reflexion agents verbally reflect on task feedback signals, then maintain their own reflective text in an episodic memory buffer to induce better decision-making in subsequent trials. Reflexion is flexible enough to incorporate various types (scalar values or free-form language) and sources (external or internally simulated) of feedback signals, and obtains significant improvements over a baseline agent across diverse tasks (sequential decision-making, coding, language reasoning). For example, Reflexion achieves a 91% pass@1 accuracy on the HumanEval coding benchmark, surpassing the previous state-of-the-art GPT-4 that achieves 80%. We also conduct ablation and analysis studies using different feedback signals, feedback incorporation methods, and agent types, and provide insights into how they affect performance."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Toolformer: Language Models Can Teach Themselves to Use Tools",
                    "abstract": "Language models (LMs) exhibit remarkable abilities to solve new tasks from just a few examples or textual instructions, especially at scale. They also, paradoxically, struggle with basic functionality, such as arithmetic or factual lookup, where much simpler and smaller models excel. In this paper, we show that LMs can teach themselves to use external tools via simple APIs and achieve the best of both worlds. We introduce Toolformer, a model trained to decide which APIs to call, when to call them, what arguments to pass, and how to best incorporate the results into future token prediction. This is done in a self-supervised way, requiring nothing more than a handful of demonstrations for each API. We incorporate a range of tools, including a calculator, a Q\\&A system, two different search engines, a translation system, and a calendar. Toolformer achieves substantially improved zero-shot performance across a variety of downstream tasks, often competitive with much larger models, without sacrificing its core language modeling abilities."
                },
                {
                    "title": "Large Language Models Can Be Easily Distracted by Irrelevant Context",
                    "abstract": "Large language models have achieved impressive performance on various natural language processing tasks. However, so far they have been evaluated primarily on benchmarks where all information in the input context is relevant for solving the task. In this work, we investigate the distractibility of large language models, i.e., how the model problem-solving accuracy can be influenced by irrelevant context. In particular, we introduce Grade-School Math with Irrelevant Context (GSM-IC), an arithmetic reasoning dataset with irrelevant information in the problem description. We use this benchmark to measure the distractibility of cutting-edge prompting techniques for large language models, and find that the model performance is dramatically decreased when irrelevant information is included. We also identify several approaches for mitigating this deficiency, such as decoding with self-consistency and adding to the prompt an instruction that tells the language model to ignore the irrelevant information."
                },
                {
                    "title": "Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions",
                    "abstract": "Prompting-based large language models (LLMs) are surprisingly powerful at generating natural language reasoning steps or Chains-of-Thoughts (CoT) for multi-step question answering (QA). They struggle, however, when the necessary knowledge is either unavailable to the LLM or not up-to-date within its parameters. While using the question to retrieve relevant text from an external knowledge source helps LLMs, we observe that this one-step retrieve-and-read approach is insufficient for multi-step QA. Here, what to retrieve depends on what has already been derived, which in turn may depend on what was previously retrieved. To address this, we propose IRCoT, a new approach for multi-step QA that interleaves retrieval with steps (sentences) in a CoT, guiding the retrieval with CoT and in turn using retrieved results to improve CoT. Using IRCoT with GPT3 substantially improves retrieval (up to 21 points) as well as downstream QA (up to 15 points) on four datasets: HotpotQA, 2WikiMultihopQA, MuSiQue, and IIRC. We observe similar substantial gains in out-of-distribution (OOD) settings as well as with much smaller models such as Flan-T5-large without additional training. IRCoT reduces model hallucination, resulting in factually more accurate CoT reasoning."
                },
                {
                    "title": "Solving math word problems with process- and outcome-based feedback",
                    "abstract": "Recent work has shown that asking language models to generate reasoning steps improves performance on many reasoning tasks. When moving beyond prompting, this raises the question of how we should supervise such models: outcome-based approaches which supervise the final result, or process-based approaches which supervise the reasoning process itself? Differences between these approaches might naturally be expected not just in final-answer errors but also in reasoning errors, which can be difficult to detect and are problematic in many real-world domains such as education. We run the first comprehensive comparison between process- and outcome-based approaches trained on a natural language task, GSM8K. We find that pure outcome-based supervision produces similar final-answer error rates with less label supervision. However, for correct reasoning steps we find it necessary to use process-based supervision or supervision from learned reward models that emulate process-based feedback. In total, we improve the previous best results from 16.8% $\\to$ 12.7% final-answer error and 14.0% $\\to$ 3.4% reasoning error among final-answer-correct solutions."
                },
                {
                    "title": "Measuring and Narrowing the Compositionality Gap in Language Models",
                    "abstract": "We investigate the ability of language models to perform compositional reasoning tasks where the overall solution depends on correctly composing the answers to sub-problems. We measure how often models can correctly answer all sub-problems but not generate the overall solution, a ratio we call the compositionality gap. We evaluate this ratio by asking multi-hop questions with answers that require composing multiple facts unlikely to have been observed together during pretraining. In the GPT-3 family of models, as model size increases we show that the single-hop question answering performance improves faster than the multi-hop performance does, therefore the compositionality gap does not decrease. This surprising result suggests that while more powerful models memorize and recall more factual knowledge, they show no corresponding improvement in their ability to perform this kind of compositional reasoning. We then demonstrate how elicitive prompting (such as chain of thought) narrows the compositionality gap by reasoning explicitly. We present a new method, self-ask, that further improves on chain of thought. In our method, the model explicitly asks itself (and answers) follow-up questions before answering the initial question. We finally show that self-ask's structured prompting lets us easily plug in a search engine to answer the follow-up questions, which additionally improves accuracy."
                },
                {
                    "title": "ReAct: Synergizing Reasoning and Acting in Language Models",
                    "abstract": "While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources, such as knowledge bases or environments, to gather additional information. We apply our approach, named ReAct, to a diverse set of language and decision making tasks and demonstrate its effectiveness over state-of-the-art baselines, as well as improved human interpretability and trustworthiness over methods without reasoning or acting components. Concretely, on question answering (HotpotQA) and fact verification (Fever), ReAct overcomes issues of hallucination and error propagation prevalent in chain-of-thought reasoning by interacting with a simple Wikipedia API, and generates human-like task-solving trajectories that are more interpretable than baselines without reasoning traces. On two interactive decision making benchmarks (ALFWorld and WebShop), ReAct outperforms imitation and reinforcement learning methods by an absolute success rate of 34% and 10% respectively, while being prompted with only one or two in-context examples. Project site with code: https://react-lm.github.io"
                },
                {
                    "title": "Decomposed Prompting: A Modular Approach for Solving Complex Tasks",
                    "abstract": "Few-shot prompting is a surprisingly powerful way to use Large Language Models (LLMs) to solve various tasks. However, this approach struggles as the task complexity increases or when the individual reasoning steps of the task themselves are hard to learn, especially when embedded in more complex tasks. To address this, we propose Decomposed Prompting, a new approach to solve complex tasks by decomposing them (via prompting) into simpler sub-tasks that can be delegated to a library of prompting-based LLMs dedicated to these sub-tasks. This modular structure allows each prompt to be optimized for its specific sub-task, further decomposed if necessary, and even easily replaced with more effective prompts, trained models, or symbolic functions if desired. We show that the flexibility and modularity of Decomposed Prompting allows it to outperform prior work on few-shot prompting using GPT3. On symbolic reasoning tasks, we can further decompose sub-tasks that are hard for LLMs into even simpler solvable sub-tasks. When the complexity comes from the input length, we can recursively decompose the task into the same task but with smaller inputs. We also evaluate our approach on textual multi-step reasoning tasks: on long-context multi-hop QA task, we can more effectively teach the sub-tasks via our separate sub-tasks prompts; and on open-domain multi-hop QA, we can incorporate a symbolic information retrieval within our decomposition framework, leading to improved performance on both tasks. Datasets, Code and Prompts available at https://github.com/allenai/DecomP."
                },
                {
                    "title": "WebGPT: Browser-assisted question-answering with human feedback",
                    "abstract": "We fine-tune GPT-3 to answer long-form questions using a text-based web-browsing environment, which allows the model to search and navigate the web. By setting up the task so that it can be performed by humans, we are able to train models on the task using imitation learning, and then optimize answer quality with human feedback. To make human evaluation of factual accuracy easier, models must collect references while browsing in support of their answers. We train and evaluate our models on ELI5, a dataset of questions asked by Reddit users. Our best model is obtained by fine-tuning GPT-3 using behavior cloning, and then performing rejection sampling against a reward model trained to predict human preferences. This model's answers are preferred by humans 56% of the time to those of our human demonstrators, and 69% of the time to the highest-voted answer from Reddit."
                },
                {
                    "title": "Unsupervised Dense Information Retrieval with Contrastive Learning",
                    "abstract": "Recently, information retrieval has seen the emergence of dense retrievers, using neural networks, as an alternative to classical sparse methods based on term-frequency. These models have obtained state-of-the-art results on datasets and tasks where large training sets are available. However, they do not transfer well to new applications with no training data, and are outperformed by unsupervised term-frequency methods such as BM25. In this work, we explore the limits of contrastive learning as a way to train unsupervised dense retrievers and show that it leads to strong performance in various retrieval settings. On the BEIR benchmark our unsupervised model outperforms BM25 on 11 out of 15 datasets for the Recall@100. When used as pre-training before fine-tuning, either on a few thousands in-domain examples or on the large MS~MARCO dataset, our contrastive model leads to improvements on the BEIR benchmark. Finally, we evaluate our approach for multi-lingual retrieval, where training data is even scarcer than for English, and show that our approach leads to strong unsupervised performance. Our model also exhibits strong cross-lingual transfer when fine-tuned on supervised English data only and evaluated on low resources language such as Swahili. We show that our unsupervised models can perform cross-lingual retrieval between different scripts, such as retrieving English documents from Arabic queries, which would not be possible with term matching methods."
                },
                {
                    "title": "A Dataset of Information-Seeking Questions and Answers Anchored in Research Papers",
                    "abstract": "Readers of academic research papers often read with the goal of answering specific questions. Question Answering systems that can answer those questions can make consumption of the content much more efficient. However, building such tools requires data that reflect the difficulty of the task arising from complex reasoning about claims made in multiple parts of a paper. In contrast, existing information-seeking question answering datasets usually contain questions about generic factoid-type information. We therefore present Qasper, a dataset of 5049 questions over 1585 Natural Language Processing papers. Each question is written by an NLP practitioner who read only the title and abstract of the corresponding paper, and the question seeks information present in the full text. The questions are then answered by a separate set of NLP practitioners who also provide supporting evidence to answers. We find that existing models that do well on other QA tasks do not perform well on answering these questions, underperforming humans by at least 27 F1 points when answering them from entire papers, motivating further research in document-grounded, information-seeking QA, which our dataset is designed to facilitate."
                },
                {
                    "title": "Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Steps",
                    "abstract": "A multi-hop question answering (QA) dataset aims to test reasoning and inference skills by requiring a model to read multiple paragraphs to answer a given question. However, current datasets do not provide a complete explanation for the reasoning process from the question to the answer. Further, previous studies revealed that many examples in existing multi-hop datasets do not require multi-hop reasoning to answer a question. In this study, we present a new multi-hop QA dataset, called 2WikiMultiHopQA, which uses structured and unstructured data. In our dataset, we introduce the evidence information containing a reasoning path for multi-hop questions. The evidence information has two benefits: (i) providing a comprehensive explanation for predictions and (ii) evaluating the reasoning skills of a model. We carefully design a pipeline and a set of templates when generating a question-answer pair that guarantees the multi-hop steps and the quality of the questions. We also exploit the structured format in Wikidata and use logical rules to create questions that are natural but still require multi-hop reasoning. Through experiments, we demonstrate that our dataset is challenging for multi-hop models and it ensures that multi-hop reasoning is required."
                },
                {
                    "title": "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks",
                    "abstract": "Large pre-trained language models have been shown to store factual knowledge in their parameters, and achieve state-of-the-art results when fine-tuned on downstream NLP tasks. However, their ability to access and precisely manipulate knowledge is still limited, and hence on knowledge-intensive tasks, their performance lags behind task-specific architectures. Additionally, providing provenance for their decisions and updating their world knowledge remain open research problems. Pre-trained models with a differentiable access mechanism to explicit non-parametric memory can overcome this issue, but have so far been only investigated for extractive downstream tasks. We explore a general-purpose fine-tuning recipe for retrieval-augmented generation (RAG) -- models which combine pre-trained parametric and non-parametric memory for language generation. We introduce RAG models where the parametric memory is a pre-trained seq2seq model and the non-parametric memory is a dense vector index of Wikipedia, accessed with a pre-trained neural retriever. We compare two RAG formulations, one which conditions on the same retrieved passages across the whole generated sequence, the other can use different passages per token. We fine-tune and evaluate our models on a wide range of knowledge-intensive NLP tasks and set the state-of-the-art on three open domain QA tasks, outperforming parametric seq2seq models and task-specific retrieve-and-extract architectures. For language generation tasks, we find that RAG models generate more specific, diverse and factual language than a state-of-the-art parametric-only seq2seq baseline."
                },
                {
                    "title": "REALM: Retrieval-Augmented Language Model Pre-Training",
                    "abstract": "Language model pre-training has been shown to capture a surprising amount of world knowledge, crucial for NLP tasks such as question answering. However, this knowledge is stored implicitly in the parameters of a neural network, requiring ever-larger networks to cover more facts. \nTo capture knowledge in a more modular and interpretable way, we augment language model pre-training with a latent knowledge retriever, which allows the model to retrieve and attend over documents from a large corpus such as Wikipedia, used during pre-training, fine-tuning and inference. For the first time, we show how to pre-train such a knowledge retriever in an unsupervised manner, using masked language modeling as the learning signal and backpropagating through a retrieval step that considers millions of documents. \nWe demonstrate the effectiveness of Retrieval-Augmented Language Model pre-training (REALM) by fine-tuning on the challenging task of Open-domain Question Answering (Open-QA). We compare against state-of-the-art models for both explicit and implicit knowledge storage on three popular Open-QA benchmarks, and find that we outperform all previous methods by a significant margin (4-16% absolute accuracy), while also providing qualitative benefits such as interpretability and modularity."
                },
                {
                    "title": "PubMedQA: A Dataset for Biomedical Research Question Answering",
                    "abstract": "We introduce PubMedQA, a novel biomedical question answering (QA) dataset collected from PubMed abstracts. The task of PubMedQA is to answer research questions with yes/no/maybe (e.g.: Do preoperative statins reduce atrial fibrillation after coronary artery bypass grafting?) using the corresponding abstracts. PubMedQA has 1k expert-annotated, 61.2k unlabeled and 211.3k artificially generated QA instances. Each PubMedQA instance is composed of (1) a question which is either an existing research article title or derived from one, (2) a context which is the corresponding abstract without its conclusion, (3) a long answer, which is the conclusion of the abstract and, presumably, answers the research question, and (4) a yes/no/maybe answer which summarizes the conclusion. PubMedQA is the first QA dataset where reasoning over biomedical research texts, especially their quantitative contents, is required to answer the questions. Our best performing model, multi-phase fine-tuning of BioBERT with long answer bag-of-word statistics as additional supervision, achieves 68.1% accuracy, compared to single human performance of 78.0% accuracy and majority-baseline of 55.2% accuracy, leaving much room for improvement. PubMedQA is publicly available at https://pubmedqa.github.io."
                },
                {
                    "title": "Neural Text Generation with Unlikelihood Training",
                    "abstract": "Neural text generation is a key tool in natural language applications, but it is well known there are major problems at its core. In particular, standard likelihood training and decoding leads to dull and repetitive outputs. While some post-hoc fixes have been proposed, in particular top-$k$ and nucleus sampling, they do not address the fact that the token-level probabilities predicted by the model are poor. In this paper we show that the likelihood objective itself is at fault, resulting in a model that assigns too much probability to sequences containing repeats and frequent words, unlike those from the human training distribution. We propose a new objective, unlikelihood training, which forces unlikely generations to be assigned lower probability by the model. We show that both token and sequence level unlikelihood training give less repetitive, less dull text while maintaining perplexity, giving superior generations using standard greedy or beam search. According to human evaluations, our approach with standard beam search also outperforms the currently popular decoding methods of nucleus sampling or beam blocking, thus providing a strong alternative to existing techniques."
                },
                {
                    "title": "Natural Questions: A Benchmark for Question Answering Research",
                    "abstract": "We present the Natural Questions corpus, a question answering data set. Questions consist of real anonymized, aggregated queries issued to the Google search engine. An annotator is presented with a question along with a Wikipedia page from the top 5 search results, and annotates a long answer (typically a paragraph) and a short answer (one or more entities) if present on the page, or marks null if no long/short answer is present. The public release consists of 307,373 training examples with single annotations; 7,830 examples with 5-way annotations for development data; and a further 7,842 examples with 5-way annotated sequestered as test data. We present experiments validating quality of the data. We also describe analysis of 25-way annotations on 302 examples, giving insights into human variability on the annotation task. We introduce robust metrics for the purposes of evaluating question answering systems; demonstrate high human upper bounds on these metrics; and establish baseline results using competitive methods drawn from related literature."
                },
                {
                    "title": "BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions",
                    "abstract": "In this paper we study yes/no questions that are naturally occurring \u2014 meaning that they are generated in unprompted and unconstrained settings. We build a reading comprehension dataset, BoolQ, of such questions, and show that they are unexpectedly challenging. They often query for complex, non-factoid information, and require difficult entailment-like inference to solve. We also explore the effectiveness of a range of transfer learning baselines. We find that transferring from entailment data is more effective than transferring from paraphrase or extractive QA data, and that it, surprisingly, continues to be very beneficial even when starting from massive pre-trained language models such as BERT. Our best method trains BERT on MultiNLI and then re-trains it on our train set. It achieves 80.4% accuracy compared to 90% accuracy of human annotators (and 62% majority-baseline), leaving a significant gap for future work."
                },
                {
                    "title": "HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering",
                    "abstract": "Existing question answering (QA) datasets fail to train QA systems to perform complex reasoning and provide explanations for answers. We introduce HotpotQA, a new dataset with 113k Wikipedia-based question-answer pairs with four key features: (1) the questions require finding and reasoning over multiple supporting documents to answer; (2) the questions are diverse and not constrained to any pre-existing knowledge bases or knowledge schemas; (3) we provide sentence-level supporting facts required for reasoning, allowing QA systems to reason with strong supervision and explain the predictions; (4) we offer a new type of factoid comparison questions to test QA systems\u2019 ability to extract relevant facts and perform necessary comparison. We show that HotpotQA is challenging for the latest QA systems, and the supporting facts enable models to improve performance and make explainable predictions."
                },
                {
                    "title": "Proximal Policy Optimization Algorithms",
                    "abstract": "We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a \"surrogate\" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time."
                },
                {
                    "title": "Principles and methods of testing finite state machines-a survey",
                    "abstract": "With advanced computer technology, systems are getting larger to fulfill more complicated tasks: however, they are also becoming less reliable. Consequently, testing is an indispensable part of system design and implementation; yet it has proved to be a formidable task for complex systems. This motivates the study of testing finite stare machines to ensure the correct functioning of systems and to discover aspects of their behavior. A finite state machine contains a finite number of states and produces outputs on state transitions after receiving inputs. Finite state machines are widely used to model systems in diverse areas, including sequential circuits, certain types of programs, and, more recently, communication protocols. In a testing problem we have a machine about which we lack some information; we would like to deduce this information by providing a sequence of inputs to the machine and observing the outputs produced. Because of its practical importance and theoretical interest, the problem of testing finite state machines has been studied in different areas and at various times. The earliest published literature on this topic dates back to the 1950's. Activities in the 1960's mid early 1970's were motivated mainly by automata theory and sequential circuit testing. The area seemed to have mostly died down until a few years ago when the testing problem was resurrected and is now being studied anew due to its applications to conformance testing of communication protocols. While some old problems which had been open for decades were resolved recently, new concepts and more intriguing problems from new applications emerge. We review the fundamental problems in testing finite state machines and techniques for solving these problems, tracing progress in the area from its inception to the present and the stare of the art. In addition, we discuss extensions of finite state machines and some other topics related to testing."
                },
                {
                    "title": "Reasoning about networks with many identical finite-state processes",
                    "abstract": "Abstract : If we consider a distributed mutual exclusion algorithm for processes arranged in a ring network in which mutual exclusion is guaranteed by means of a token that is passed around the ring. A first attempt might be to consider a reduced system with with one or two processes. If the reduced system can be shown to be correct and of the individual processes are really identical, then we are tempted to conclude that the entire system will be correct. This type of informal argument is used quite frequently by designers in constructing systems that contain large numbers of identical processing elements. It is easy to contrive an example in which some pathological behavior only occurs when, say, 100 processes are connected together. By examining a system with only one or two processes it might be quite difficult to determine that this behavior is possible. One has the feeling that in many cases this kind of intuitive reasoning does lead to correct results. The question addressed is whether it is possible to provide a solid theoretical basis that will prevent fallacious conclusions in arguments of this type. Besides providing a firm basis for a common type of informal reasoning, our results are crucial for the success of automatic verification methods that involve temporal logic model checking. These techniques check that a finite-state concurrent system satisfies a temporal logic formula by searching all possible paths in the global state graph determined by the concurrent system. They have been used successfully to find subtle errors in tricky self-timed circuits--errors apparently unknown to the circuit designers. By using these results, model checking may become feasible for networks with large numbers of identical processes."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a knowledge-intensive agent that possesses robust reasoning logic, adapts to specific environments, and allows for human intervention in its reasoning process?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the capabilities of language agents in knowledge-intensive tasks, which have significant implications for various applications, including customer support, education, and research assistance. By creating agents that can reason effectively, adapt to user needs, and incorporate human feedback, we can enhance user trust and satisfaction, leading to broader adoption of AI technologies. This research could pave the way for future studies on human-AI collaboration, improving the alignment between agent behavior and human intent, and ultimately leading to more intelligent and responsive systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need for a sophisticated reasoning logic that can handle complex tasks while being adaptable to different contexts. Naive approaches may fail because they often lack the necessary modularity and flexibility to accommodate diverse user inputs and feedback. Additionally, existing models may struggle with static reasoning processes that do not allow for real-time adjustments based on user interactions. Overcoming these technical obstacles requires innovative methodologies that integrate formal reasoning frameworks, such as finite state machines, with adaptive learning mechanisms.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on building language agents without adequately addressing the need for controllable reasoning logic and adaptability. Many existing solutions lack the ability to incorporate user feedback effectively, leading to a disconnect between agent actions and user expectations. Barriers such as the complexity of designing modular systems and the challenge of creating effective feedback mechanisms have hindered progress. Our approach differs by introducing the AMOR agent, which utilizes a finite state machine framework to formalize reasoning and allows for process-based supervision, thereby addressing the limitations of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing the AMOR agent, which employs a finite state machine to structure its reasoning logic. The training process consists of two stages: a warm-up phase where training data is generated for each module independently, and a feedback mechanism that allows users to provide binary judgments or refinements on outputs. We will evaluate AMOR using specific knowledge bases and metrics that assess reasoning accuracy and adaptability. The expected outcomes include a more controllable and responsive agent capable of effectively addressing user"
            }
        },
        "author_data": {
            "39fab9f5-f13e-4b4c-9d10-4fcc98c432b6": {
                "pk": "39fab9f5-f13e-4b4c-9d10-4fcc98c432b6",
                "name": "Jian Guan",
                "collaborators": [
                    "Minlie Huang",
                    "Pengming Feng",
                    "Wenwu Wang",
                    "Ye Yuan",
                    "Ziqi Liu",
                    "Fei Huang",
                    "Zhihao Zhao",
                    "Xiaoyan Zhu",
                    "Wenbo Wang",
                    "Xinxin Wang"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Story Generation",
                    "Reinforcement Learning",
                    "Self-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Mitigating the Learning Bias towards Repetition by Self-Contrastive Training for Open-Ended Generation",
                        "abstract": "Despite the huge progress in myriad generation tasks, pretrained language models (LMs) such as GPT2 still tend to generate repetitive texts with maximization-based decoding algorithms for open-ended generation. We attribute their overestimation of token-level repetition probabilities to the learning bias: LMs capture simple repetitive patterns faster with the MLE loss. We propose self-contrastive training to penalize the output of a premature checkpoint of the same model when it incorrectly predicts repetition, which is shown to mitigate repetition effectively while maintaining fluency on two datasets. Furthermore, we find that LMs use longer-range dependencies to predict repetitive tokens than non-repetitive ones, which may be the cause of sentence-level repetition loops."
                    },
                    {
                        "title": "UNION: An Unreferenced Metric for Evaluating Open-ended Story Generation",
                        "abstract": "Despite the success of existing referenced metrics (e.g., BLEU and MoverScore), they correlate poorly with human judgments for open-ended text generation including story or dialog generation because of the notorious one-to-many issue: there are many plausible outputs for the same input, which may differ substantially in literal or semantics from the limited number of given references. To alleviate this issue, we propose UNION, a learnable unreferenced metric for evaluating open-ended story generation, which measures the quality of a generated story without any reference. Built on top of BERT, UNION is trained to distinguish human-written stories from negative samples and recover the perturbation in negative stories. We propose an approach of constructing negative samples by mimicking the errors commonly observed in existing NLG models, including repeated plots, conflicting logic, and long-range incoherence. Experiments on two story datasets demonstrate that UNION is a reliable measure for evaluating the quality of generated stories, which correlates better with human judgments and is more generalizable than existing state-of-the-art metrics."
                    },
                    {
                        "title": "A Corpus for Understanding and Generating Moral Stories",
                        "abstract": "Teaching morals is one of the most important purposes of storytelling. An essential ability for understanding and writing moral stories is bridging story plots and implied morals. Its challenges mainly lie in: (1) grasping knowledge about abstract concepts in morals, (2) capturing inter-event discourse relations in stories, and (3) aligning value preferences of stories and morals concerning good or bad behavior. In this paper, we propose two understanding tasks and two generation tasks to assess these abilities of machines. We present STORAL, a new dataset of Chinese and English human-written moral stories. We show the difficulty of the proposed tasks by testing various models with automatic and manual evaluation on STORAL. Furthermore, we present a retrieval-augmented algorithm that effectively exploits related concepts or events in training sets as additional guidance to improve performance on these tasks."
                    },
                    {
                        "title": "A Knowledge-Enhanced Pretraining Model for Commonsense Story Generation",
                        "abstract": "Story generation, namely generating a reasonable story from a leading context, is an important but challenging task. In spite of the success in modeling fluency and local coherence, existing neural language generation models (e.g., GPT-2) still suffer from repetition, logic conflicts, and lack of long-range coherence in generated stories. We conjecture that this is because of the difficulty of associating relevant commonsense knowledge, understanding the causal relationships, and planning entities and events with proper temporal order. In this paper, we devise a knowledge-enhanced pretraining model for commonsense story generation. We propose to utilize commonsense knowledge from external knowledge bases to generate reasonable stories. To further capture the causal and temporal dependencies between the sentences in a reasonable story, we employ multi-task learning which combines a discriminative objective to distinguish true and fake stories during fine-tuning. Automatic and manual evaluation shows that our model can generate more reasonable stories than state-of-the-art baselines, particularly in terms of logic and global coherence."
                    },
                    {
                        "title": "Low-dimensional Denoising Embedding Transformer for ECG Classification",
                        "abstract": "The transformer based model (e.g., FusingTF) has been employed recently for Electrocardiogram (ECG) signal classification. However, the high-dimensional embedding obtained via 1-D convolution and positional encoding can lead to the loss of the signal's own temporal information and a large amount of training parameters. In this paper, we propose a new method for ECG classification, called low-dimensional denoising embedding transformer (LDTF), which contains two components, i.e., low-dimensional denoising embedding (LDE) and transformer learning. In the LDE component, a low-dimensional representation of the signal is obtained in the time-frequency domain while preserving its own temporal information. And with the low dimensional embedding, the transformer learning is then used to obtain a deeper and narrower structure with fewer training parameters than that of the FusingTF. Experiments conducted on the MIT-BIH dataset demonstrates the effectiveness and the superior performance of our proposed method, as compared with state-of-the-art methods."
                    },
                    {
                        "title": "Generating Coherent Narratives by Learning Dynamic and Discrete Entity States with a Contrastive Framework",
                        "abstract": "Despite advances in generating fluent texts, existing pretraining models tend to attach incoherent event sequences to involved entities when generating narratives such as stories and news. We conjecture that such issues result from representing entities as static embeddings of superficial words, while neglecting to model their ever-changing states, i.e., the information they carry, as the text unfolds. Therefore, we extend the Transformer model to dynamically conduct entity state updates and sentence realization for narrative generation. We propose a contrastive framework to learn the state representations in a discrete space, and insert additional attention layers into the decoder to better exploit these states. Experiments on two narrative datasets show that our model can generate more coherent and diverse narratives than strong baselines with the guidance of meaningful entity states."
                    },
                    {
                        "title": "Hierarchical Metadata Information Constrained Self-Supervised Learning for Anomalous Sound Detection Under Domain Shift",
                        "abstract": "Self-supervised learning methods have achieved promising performance for anomalous sound detection (ASD) under domain shift, where the type of domain shift is considered in feature learning by incorporating section IDs. However, the attributes accompanying audio files under each section, such as machine operating conditions and noise types, have not been considered, although they are also crucial for characterizing domain shifts. In this paper, we present a hierarchical metadata information constrained self-supervised (HMIC) ASD method, where the hierarchical relation between section IDs and attributes is constructed, and used as constraints to obtain finer feature representation. In addition, we propose an attribute-group-center (AGC)-based method for calculating the anomaly score under the domain shift condition. Experiments are performed to demonstrate its improved performance over the state-of-the-art self-supervised methods in DCASE 2022 challenge Task 2."
                    },
                    {
                        "title": "Language Models Hallucinate, but May Excel at Fact Verification",
                        "abstract": "Recent progress in natural language processing (NLP) owes much to remarkable advances in large language models (LLMs). Nevertheless, LLMs frequently \"hallucinate,\" resulting in non-factual outputs. Our carefully-designed human evaluation substantiates the serious hallucination issue, revealing that even GPT-3.5 produces factual outputs less than 25% of the time. This underscores the importance of fact verifiers in order to measure and incentivize progress. Our systematic investigation affirms that LLMs can be repurposed as effective fact verifiers with strong correlations with human judgments. Surprisingly, FLAN-T5-11B, the least factual generator in our study, performs the best as a fact verifier, even outperforming more capable LLMs like GPT3.5 and ChatGPT. Delving deeper, we analyze the reliance of these LLMs on high-quality evidence, as well as their deficiencies in robustness and generalization ability. Our study presents insights for developing trustworthy generation models."
                    },
                    {
                        "title": "Story Ending Generation with Incremental Encoding and Commonsense Knowledge",
                        "abstract": "Generating a reasonable ending for a given story context, i.e., story ending generation, is a strong indication of story comprehension. This task requires not only to understand the context clues which play an important role in planning the plot but also to handle implicit knowledge to make a reasonable, coherent story.   In this paper, we devise a novel model for story ending generation. The model adopts an incremental encoding scheme to represent context clues which are spanning in the story context. In addition, commonsense knowledge is applied through multi-source attention to facilitate story comprehension, and thus to help generate coherent and reasonable endings. Through building context clues and using implicit knowledge, the model is able to produce reasonable story endings. context clues implied in the post and make the inference based on it.   Automatic and manual evaluation shows that our model can generate more reasonable story endings than state-of-the-art baselines."
                    },
                    {
                        "title": "Blind SAR Image Despeckling Using Self-Supervised Dense Dilated Convolutional Neural Network",
                        "abstract": "Despeckling is a key and indispensable step in SAR image preprocessing, existing deep learning-based methods achieve SAR despeckling by learning some mappings between speckled (different looks) and clean images. However, there exist no clean SAR image in the real world. To this end, in this paper, we propose a self-supervised dense dilated convolutional neural network (BDSS) for blind SAR image despeckling. Proposed BDSS can still learn to suppress speckle noise without clean ground truth by optimized for L2 loss. Besides, three enhanced dense blocks with dilated convolution are employed to improve network performance. The synthetic and real-data experiments demonstrate that proposed BDSS can achieve despeckling effectively while maintaining well features such as edges, point targets, and radiometric. At last, we demonstrate that our proposed BDSS can achieve blind despeckling excellently, i.e., do not need to care about the number of looks."
                    },
                    {
                        "title": "Model-Based Reinforcement Learning with Adversarial Training for Online Recommendation",
                        "abstract": "Reinforcement learning is well suited for optimizing policies of recommender systems. Current solutions mostly focus on model-free approaches, which require frequent interactions with the real environment, and thus are expensive in model learning. Offline evaluation methods, such as importance sampling, can alleviate such limitations, but usually request a large amount of logged data and do not work well when the action space is large. In this work, we propose a model-based reinforcement learning solution which models user-agent interaction for offline policy learning via a generative adversarial network. To reduce bias in the learned model and policy, we use a discriminator to evaluate the quality of generated data and scale the generated rewards. Our theoretical analysis and empirical evaluations demonstrate the effectiveness of our solution in learning policies from the offline and generated data."
                    },
                    {
                        "title": "A Practical Solution for SAR Despeckling With Adversarial Learning Generated Speckled-to-Speckled Images",
                        "abstract": "In this letter, we aim to address a synthetic aperture radar (SAR) despeckling problem with the necessity of neither clean (speckle-free) SAR images nor independent speckled image pairs from the same scene, and a practical solution for SAR despeckling (PSD) is proposed. First, an adversarial learning framework is designed to generate speckled-to-speckled (S2S) image pairs from the same scene in the situation where only single speckled SAR images are available. Then, the S2S SAR image pairs are employed to train a modified despeckling Nested-UNet model using the Noise2Noise (N2N) strategy. Moreover, an iterative version of the PSD method (PSDi) is also presented. Experiments are conducted on both synthetic speckled and real SAR data to demonstrate the superiority of the proposed methods compared with several state-of-the-art methods. The results show that our methods can reach a good tradeoff between feature preservation and speckle suppression."
                    }
                ]
            },
            "f972b419-869e-4028-b357-3a0e7bc2401f": {
                "pk": "f972b419-869e-4028-b357-3a0e7bc2401f",
                "name": "Wei Wu",
                "collaborators": [],
                "domain": [
                    "Quantum Mechanics",
                    "Operator Algebras",
                    "Statistical Mechanics",
                    "Density Functional Theory"
                ],
                "publications": [
                    {
                        "title": "A central limit theorem for square ice",
                        "abstract": "We prove that the height function associated with the uniform six-vertex model (or equivalently, the uniform homomorphism height function from $\\mathbb Z^2$ to $\\mathbb Z$) satisfies a central limit theorem, upon some logarithmic rescaling."
                    },
                    {
                        "title": "An operator Arzel\u00e0-Ascoli theorem",
                        "abstract": "We generalize the Arzel\\`a-Ascoli theorem to the setting of matrix order unit spaces, extending the work of Antonescu-Christensen on unital C*-algebras. This gives an affirmative answer to a question of Antonescu and Christensen."
                    },
                    {
                        "title": "Non-commutative metric topology on matrix state space",
                        "abstract": "We present an operator space version of Rieffel's theorem on the agreement of the metric topology, on a subset of the Banach space dual of a normed space, from a seminorm with the weak*-topology. As an application we obtain a necessary and sufficient condition for the matrix metric from an unbounded Fredholm module to give the BW-topology on the matrix state space of the $C^*$-algebra. Motivated by recent results we formulate a non-commutative Lipschitz seminorm on a matrix order unit space and characterize those matrix Lipschitz seminorms whose matrix metric topology coincides with the BW-topology on the matrix state space."
                    },
                    {
                        "title": "Non-Commutative Metrics on Matrix State Spaces",
                        "abstract": "We use the theory of quantization to introduce non-commutative versions of metric on state space and Lipschitz seminorm. We show that a lower semicontinuous matrix Lipschitz seminorm is determined by their matrix metrics on the matrix state spaces. A matrix metric comes from a lower semicontinuous matrix Lip-norm if and only if it is convex, midpoint balanced, and midpoint concave. The operator space of Lipschitz functions with a matrix norm coming from a closed matrix Lip-norm is the operator space dual of an operator space. They generalize Rieffel's results to the quantized situation."
                    },
                    {
                        "title": "Lipschitzness of *-homomorphisms between C*-metric algebras",
                        "abstract": "A C*-metric algebra consists of a unital C*-algebra and a Leibniz Lip-norm on the C*-algebra. We show that if the Lip-norms concerned are lower semicontinuous, then any unital *-homomorphism from a C*-metric algebra to another one is necessarily Lipschitz. It results that the free product of two Lipschitz unital *-homomorphisms between C*-metric algebras coming from *-filtrations is still a Lipschitz unital *-homomorphism."
                    },
                    {
                        "title": "Heat transfer in a nonequilibrium spin-boson model: A perturbative approach",
                        "abstract": "We investigate the heat transport in a nonequilibrium spin-boson model, where a two level system bridging two harmonic reservoirs at different temperatures, by employing a unitary transformation along with a resolvent operator expansion technique. Analytical expressions of the heat current and the thermal conductance of this model are obtained. Compared with the performances of other methods, namely, the nonequilibrium Green's function method and the equation of motion formulation, our approach provides a reasonable description of heat transfer properties of the nonequilibrium spin-boson model for the weak-coupling region at low temperature."
                    },
                    {
                        "title": "Realization of hierarchical equations of motion from stochastic perspectives",
                        "abstract": "The hierarchical equations of motion (HEOM) for a generalized quantum dissipative system is rigorously constructed in the frameworks of two different stochastic dynamical descriptions, i.e., the non-Markovian quantum state diffusion approach as well as the stochastic decoupling scheme. We demonstrated that the HEOMs obtained by these two different stochastic dynamical methods are identical. Moreover, we present some numerical examples to verify the feasibility of our formalism."
                    },
                    {
                        "title": "Stochastic decoupling approach to the quantum dissipative dynamics: Perturbative and non-perturbative treatments",
                        "abstract": "We develop a hierarchical functional derivative method to investigate the reduced dynamics of a quantum dissipative system within the framework of a stochastic decoupling description. Keeping only the lowest order truncation of the hierarchical functional derivatives, one can recover the second order Nakajima-Zwanzig quantum master equation. Taking into account the higher-order corrections, our method can be implemented as a highly efficient numerical scheme for a general quantum dissipative system beyond the usual Markovian, rotating-wave, and weak-coupling approximations."
                    },
                    {
                        "title": "Local central limit theorem for gradient field models",
                        "abstract": "We consider the gradient field model in $\\left[ -N,N\\right] ^{2}\\cap \\mathbb{Z}^{2}$ with a uniformly convex interaction potential. Naddaf-Spencer \\cite{NS} and Miller \\cite{Mi} proved that the macroscopic averages of linear statistics of the field converge to a continuum Gaussian free field. In this paper we prove the distribution of $\\phi(0)/\\sqrt{\\log N}$ converges uniformly to a Gaussian density, with a Berry-Esseen type bound. This implies the distribution of $\\phi(0)$ is sufficiently `Gaussian like' between $[-\\sqrt {\\log N}, \\sqrt {\\log N}]$."
                    },
                    {
                        "title": "Quantized Gromov-Hausdorff distance",
                        "abstract": "A quantized metric space is a matrix order unit space equipped with an operator space version of Rieffel's Lip-norm. We develop for quantized metric spaces an operator space version of quantum Gromov-Hausdorff distance. We show that two quantized metric spaces are completely isometric if and only if their quantized Gromov-Hausdorff distance is zero. We establish a completeness theorem. As applications, we show that a quantized metric space with 1-exact underlying matrix order unit space is a limit of matrix algebras with respect to quantized Gromov-Hausdorff distance, and that matrix algebras converge naturally to the sphere for quantized Gromov-Hausdorff distance."
                    },
                    {
                        "title": "Exchange interaction between the triplet exciton and the localized spin in copper-phthalocyanine",
                        "abstract": "Triplet excitonic state in the organic molecule may arise from a singlet excitation and the following inter-system crossing. Especially for a spin-bearing molecule, an exchange interaction between the triplet exciton and the original spin on the molecule can be expected. In this paper, such exchange interaction in copper-phthalocyanine (CuPc, spin-$\\frac{1}{2}$) was investigated from first-principles by using density-functional theory within a variety of approximations to the exchange correlation, ranging from local-density approximation to long-range corrected hybrid-exchange functional. The magnitude of the computed exchange interaction is in the order of meV with the minimum value (1.5 meV, ferromagnetic) given by the long-range corrected hybrid-exchange functional CAM-B3LYP. This exchange interaction can therefore give rise to a spin coherence with an oscillation period in the order of picoseconds, which is much shorter than the triplet lifetime in CuPc (typically tens of nanoseconds). This implies that it might be possible to manipulate the localised spin on Cu experimentally using optical excitation and inter-system crossing well before the triplet state disappears."
                    },
                    {
                        "title": "Phenomenological modelling for Time-Resolved Electron Paramagnetic Resonance in radical-triplet system",
                        "abstract": "The spin dynamics of radical-triplet system (RTS) has been calculated by using the Lindblad formalism within the theory of open quantum system. The single-radical-triplet system (SRTS) is considered here for single-qubit quantum gate operations while double-radical-triplet system (DRTS) for two-qubit operations. The environment effects taken into account include the spin-lattice relaxation of the triplet exciton and radical spin-$\\frac{1}{2}$, the inter-system crossing process that induces the transition from singlet excited state to the triplet ground state, and the rather slow relaxation process from the triplet ground state back down to the singlet ground state. These calculations shown that the line shape broadening is strongly related to the exchange interaction between triplet and exciton, which can be understood as a spontaneous magnetic field created by the triplet renormalises the original spin-$\\frac{1}{2}$ electron spin resonance spectra. This work will provide key information about the spin dynamics for building optically-controlled molecular quantum gate out of radical-bearing molecules. Moreover, this has generated the further theoretical question on how the mixture of fermion and boson behaves."
                    },
                    {
                        "title": "A hybrid-exchange density-functional theory study of the electronic structure of $\\mathrm{MnV}_2\\mathrm{O}_4$: Exotic orbital ordering in the cubic structure",
                        "abstract": "The electronic structures of the cubic and tetragonal $\\mathrm{MnV}_2\\mathrm{O}_4$ have been studied by using hybrid-exchange density functional theory. The computed electronic structure of the tetragonal phase shows an anti-ferro orbital ordering on V sites and a ferrimagnetic ground state (the spins on V and Mn are anti-aligned). These results are in a good agreement with the previous theoretical result obtained from the local-density approximation+$U$ methods [S. Sarkar, et. al., Phys. Rev. Lett. 102, 216405 (2009)]. Moreover, the electronic structure, especially the projected density of states of the cubic phase has been predicted with a good agreement with the recent soft x-ray spectroscopy experiment. Similar to the tetragonal phase, the spins on V and Mn in the cubic structure favour a ferrimagnetic configuration. Most interesting is that the computed charge densities of the spin-carrying orbitals on V in the cubic phase show an exotic orbital ordering, i.e., a ferro-orbital ordering along [110] but an anti-ferro-orbital ordering along [$\\overline{1}$10]."
                    },
                    {
                        "title": "Modelling the electronic structure and magnetic properties of LiFeAs and FeSe using hybrid-exchange density functional theory",
                        "abstract": "The electronic structure and magnetic properties of LiFeAs and FeSe have been studied using hybrid exchange density functional theory. The total energies for a unit cell in LiFeAs and FeSe with different spin states including non-magnetic and spin-2 are calculated. The spin-2 configuration has the lower energy for both LiFeAs and FeSe. The computed anti-ferromagnetic exchange interactions between spins on the nearest (next nearest) neighbouring Fe atoms in LiFeAs and FeSe are approximately 14 (17) meV and 6 (13) meV respectively. The total energies of the checkerboard and stripe-type anti-ferromagnetic ordering for LiFeAs and FeSe are compared, yielding that for LiFeAs the checkerboard is lower whereas for FeSe the stripe-type is lower. However, owing to the fact that the exchange interaction of the next nearest neighbour is larger than that of the nearest one, which means that the collinear ordering might be the ground state. These results are in agreement with previous theoretical calculations and experiments. Especially the calculations for LiFeAs indicate a co-existence of conducting d-bands at the Fermi surface and d-orbital magnetism far below the Fermi surface. The theoretical results presented here might be useful for the experimentalists working on the electronic structure and magnetism of iron-based superconductors."
                    },
                    {
                        "title": "Electronic structure and magnetic properties of $\\mathrm{CaCr}\\mathrm{O}_3$: The interplay between spin- and orbital-orderings",
                        "abstract": "The electronic structure and magnetic properties of CaCr$\\mathrm{O}_3$ have been calculated by two methods, including hybrid-exchange density-function theory and density-functional theory + $U$. The computed densities of states from both of these methods are in a qualitative agreement with the previous x-ray spectroscopy. On the other hand, the opening of the band gap separates them apart. hybrid-exchange density-functional theory always gives a finite band gap, down to $\\sim 1.2$ eV from HSE06 functional, whereas by tuning the Hubbard-$U$ parameter down to $0.5$ eV, a conducting state with AFM-C (defined in the text) spin configuration can be achieved. From hybrid density-functional theory, the computed nearest-neighbouring exchange interaction along the $c$-axis and in the $ab$-plane are $\\sim 4$ meV and $\\sim 6$ meV (anti-ferromagnetic), respectively, which are qualitatively in agreement with the previous magnetic measurements. These anti-ferromagnetic exchange interaction, together with the in-plane anti-ferro-orbital ordering will induce a spin-orbital frustration, which could play a role for the abnormal electronic properties in CaCrO$_3$. In hybrid-exchange density-functional theory, an abrupt reduction ($\\sim 0.2$ eV) of the majority-spin band gap of the ferromagnetic state between 60 K and 100 K has been found as lowering temperature, which shows a strong link to the previous optical conductivity measurements in [A. C. Komarek, et. al., Phys. Rev. B \\textbf{84}, 125114 (2011)]. In sharp contrast, the density-functional theory + $U$ methods predicted AFM-C state as the lowest AFM state for the crystal structure measured below 90 K, above which AFM-A is however the lowest. The closely related concepts including electron-hole liquid and surface-plasmon-mediating spin-spin interactions have been discussed as well."
                    }
                ]
            },
            "6a65e2bd-f014-454a-a52c-7d1061b7eba2": {
                "pk": "6a65e2bd-f014-454a-a52c-7d1061b7eba2",
                "name": "Zujie Wen",
                "collaborators": [
                    "Jia Liu",
                    "Minghui Yang",
                    "Yongliang Wang",
                    "Liang Jiang",
                    "Yafang Wang",
                    "Gerard de Melo",
                    "Hongru Liang",
                    "Wenqiang Lei",
                    "Haitao Mi",
                    "Shuang Peng"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Dialogue Systems",
                    "Text-to-SQL"
                ],
                "publications": [
                    {
                        "title": "SeaD: End-to-end Text-to-SQL Generation with Schema-aware Denoising",
                        "abstract": "In text-to-SQL task, seq-to-seq models often lead to sub-optimal performance due to limitations in their architecture. In this paper, we present a simple yet effective approach that adapts transformer-based seq-to-seq model to robust text-to-SQL generation. Instead of inducing constraint to decoder or reformat the task as slot-filling, we propose to train seq-to-seq model with Schema aware Denoising (SeaD), which consists of two denoising objectives that train model to either recover input or predict output from two novel erosion and shuffle noises. These denoising objectives acts as the auxiliary tasks for better modeling the structural data in S2S generation. In addition, we improve and propose a clause-sensitive execution guided (EG) decoding strategy to overcome the limitation of EG decoding for generative model. The experiments show that the proposed method improves the performance of seq-to-seq model in both schema linking and grammar correctness and establishes new state-of-the-art on WikiSQL benchmark. The results indicate that the capacity of vanilla seq-to-seq architecture for text-to-SQL may have been under-estimated."
                    },
                    {
                        "title": "Robust Domain Adaptation for Machine Reading Comprehension",
                        "abstract": "Most domain adaptation methods for machine reading comprehension (MRC) use a pre-trained question-answer (QA) construction model to generate pseudo QA pairs for MRC transfer. Such a process will inevitably introduce mismatched pairs (i.e., noisy correspondence) due to i) the unavailable QA pairs in target documents, and ii) the domain shift during applying the QA construction model to the target domain. Undoubtedly, the noisy correspondence will degenerate the performance of MRC, which however is neglected by existing works. To solve such an untouched problem, we propose to construct QA pairs by additionally using the dialogue related to the documents, as well as a new domain adaptation method for MRC. Specifically, we propose Robust Domain Adaptation for Machine Reading Comprehension (RMRC) method which consists of an answer extractor (AE), a question selector (QS), and an MRC model. Specifically, RMRC filters out the irrelevant answers by estimating the correlation to the document via the AE, and extracts the questions by fusing the candidate questions in multiple rounds of dialogue chats via the QS. With the extracted QA pairs, MRC is fine-tuned and provides the feedback to optimize the QS through a novel reinforced self-training method. Thanks to the optimization of the QS, our method will greatly alleviate the noisy correspondence problem caused by the domain shift. To the best of our knowledge, this could be the first study to reveal the influence of noisy correspondence in domain adaptation MRC models and show a feasible way to achieve robustness to mismatched pairs. Extensive experiments on three datasets demonstrate the effectiveness of our method."
                    },
                    {
                        "title": "Utterance Rewriting with Contrastive Learning in Multi-turn Dialogue",
                        "abstract": "Context modeling plays a significant role in building multi-turn dialogue systems. In order to make full use of context information, systems can use Incomplete Utterance Rewriting(IUR) methods to simplify the multi-turn dialogue into single-turn by merging current utterance and context information into a self-contained utterance. However, previous approaches ignore the intent consistency between the original query and rewritten query. The detection of omitted or coreferred locations in the original query can be further improved. In this paper, we introduce contrastive learning and multi-task learning to jointly model the problem. Our method benefits from carefully designed self-supervised objectives, which act as auxiliary tasks to capture semantics at both sentence-level and token-level. The experiments show that our proposed model achieves state-of-the-art performance on several public datasets."
                    },
                    {
                        "title": "Interactive Question Clarification in Dialogue via Reinforcement Learning",
                        "abstract": "Coping with ambiguous questions has been a perennial problem in real-world dialogue systems. Although clarification by asking questions is a common form of human interaction, it is hard to define appropriate questions to elicit more specific intents from a user. In this work, we propose a reinforcement model to clarify ambiguous questions by suggesting refinements of the original query. We first formulate a collection partitioning problem to select a set of labels enabling us to distinguish potential unambiguous intents. We list the chosen labels as intent phrases to the user for further confirmation. The selected label along with the original user query then serves as a refined query, for which a suitable response can more easily be identified. The model is trained using reinforcement learning with a deep policy network. We evaluate our model based on real-world user clicks and demonstrate significant improvements across several different experiments."
                    },
                    {
                        "title": "Knowing-how & Knowing-that: A New Task for Machine Comprehension of User Manuals",
                        "abstract": "The machine reading comprehension (MRC) of user manuals has huge potential in customer service. However, current methods have trouble answering complex questions. Therefore, we introduce the Knowing-how & Knowing-that task that requires the model to answer factoid-style, procedure-style, and inconsistent questions about user manuals. We resolve this task by jointly representing the steps and facts in a graph TARA, which supports a unified inference of various questions. Towards a systematical benchmarking study, we design a heuristic method to automatically parse user manuals into TARAs and build an annotated dataset to test the model's ability in answering real-world questions. Empirical results demonstrate that representing user manuals as TARAs is a desired solution for the MRC of user manuals. An in-depth investigation of TARA further sheds light on the issues and broader impacts of future representations of user manuals. We hope our work can move the MRC of user manuals to a more complex and realistic stage."
                    },
                    {
                        "title": "Multi-granularity Correspondence Learning from Long-term Noisy Videos",
                        "abstract": "Existing video-language studies mainly focus on learning short video clips, leaving long-term temporal dependencies rarely explored due to over-high computational cost of modeling long videos. To address this issue, one feasible solution is learning the correspondence between video clips and captions, which however inevitably encounters the multi-granularity noisy correspondence (MNC) problem. To be specific, MNC refers to the clip-caption misalignment (coarse-grained) and frame-word misalignment (fine-grained), hindering temporal learning and video understanding. In this paper, we propose NOise Robust Temporal Optimal traNsport (Norton) that addresses MNC in a unified optimal transport (OT) framework. In brief, Norton employs video-paragraph and clip-caption contrastive losses to capture long-term dependencies based on OT. To address coarse-grained misalignment in video-paragraph contrast, Norton filters out the irrelevant clips and captions through an alignable prompt bucket and realigns asynchronous clip-caption pairs based on transport distance. To address the fine-grained misalignment, Norton incorporates a soft-maximum operator to identify crucial words and key frames. Additionally, Norton exploits the potential faulty negative samples in clip-caption contrast by rectifying the alignment target with OT assignment to ensure precise temporal modeling. Extensive experiments on video retrieval, videoQA, and action segmentation verify the effectiveness of our method. Code is available at https://lin-yijie.github.io/projects/Norton."
                    },
                    {
                        "title": "CARE: A Clue-guided Assistant for CSRs to Read User Manuals",
                        "abstract": "It is time-saving to build a reading assistant for customer service representations (CSRs) when reading user manuals, especially information-rich ones. Current solutions don't fit the online custom service scenarios well due to the lack of attention to user questions and possible responses. Hence, we propose to develop a time-saving and careful reading assistant for CSRs, named CARE. It can help the CSRs quickly find proper responses from the user manuals via explicit clue chains. Specifically, each of the clue chains is formed by inferring over the user manuals, starting from the question clue aligned with the user question and ending at a possible response. To overcome the shortage of supervised data, we adopt the self-supervised strategy for model learning. The offline experiment shows that CARE is efficient in automatically inferring accurate responses from the user manual. The online experiment further demonstrates the superiority of CARE to reduce CSRs' reading burden and keep high service quality, in particular with >35% decrease in time spent and keeping a >0.75 ICC score."
                    },
                    {
                        "title": "R2D2: Recursive Transformer based on Differentiable Tree for Interpretable Hierarchical Language Modeling",
                        "abstract": "Human language understanding operates at multiple levels of granularity (e.g., words, phrases, and sentences) with increasing levels of abstraction that can be hierarchically combined. However, existing deep models with stacked layers do not explicitly model any sort of hierarchical process. This paper proposes a recursive Transformer model based on differentiable CKY style binary trees to emulate the composition process. We extend the bidirectional language model pre-training objective to this architecture, attempting to predict each word given its left and right abstraction nodes. To scale up our approach, we also introduce an efficient pruned tree induction algorithm to enable encoding in just a linear number of composition steps. Experimental results on language modeling and unsupervised parsing show the effectiveness of our approach."
                    },
                    {
                        "title": "AdaCoach: A Virtual Coach for Training Customer Service Agents",
                        "abstract": "With the development of online business, customer service agents gradually play a crucial role as an interface between the companies and their customers. Most companies spend a lot of time and effort on hiring and training customer service agents. To this end, we propose AdaCoach: A Virtual Coach for Training Customer Service Agents, to promote the ability of newly hired service agents before they get to work. AdaCoach is designed to simulate real customers who seek help and actively initiate the dialogue with the customer service agents. Besides, AdaCoach uses an automated dialogue evaluation model to score the performance of the customer agent in the training process, which can provide necessary assistance when the newly hired customer service agent encounters problems. We apply recent NLP technologies to ensure efficient run-time performance in the deployed system. To the best of our knowledge, this is the first system that trains the customer service agent through human-computer interaction. Until now, the system has already supported more than 500,000 simulation training and cultivated over 1000 qualified customer service agents."
                    },
                    {
                        "title": "Strength Lies in Differences! Improving Strategy Planning for Non-collaborative Dialogues via Diversified User Simulation",
                        "abstract": "We investigate non-collaborative dialogue agents, which are expected to engage in strategic conversations with diverse users, for securing a mutual agreement that leans favorably towards the system's objectives. This poses two main challenges for existing dialogue agents: 1) The inability to integrate user-specific characteristics into the strategic planning, and 2) The difficulty of training strategic planners that can be generalized to diverse users. To address these challenges, we propose Trip to enhance the capability in tailored strategic planning, incorporating a user-aware strategic planning module and a population-based training paradigm. Through experiments on benchmark non-collaborative dialogue tasks, we demonstrate the effectiveness of Trip in catering to diverse users."
                    },
                    {
                        "title": "Hummer: Towards Limited Competitive Preference Dataset",
                        "abstract": "Preference datasets are essential for incorporating human preferences into pre-trained language models, playing a key role in the success of Reinforcement Learning from Human Feedback. However, these datasets often demonstrate conflicting alignment objectives, leading to increased vulnerability to jailbreak attacks and challenges in adapting downstream tasks to prioritize specific alignment objectives without negatively impacting others. In this work, we introduce a novel statistical metric, Alignment Dimension Conflict, to quantify the degree of conflict within preference datasets. We then present \\texttt{Hummer} and its fine-grained variant, \\texttt{Hummer-F}, as innovative pairwise preference datasets with reduced-conflict alignment objectives. \\texttt{Hummer} is built based on UltraFeedback and is enhanced by AI feedback from GPT-4, marking as the first preference dataset aimed at reducing the competition between alignment objectives. Furthermore, we develop reward models, HummerRM and HummerRM-F, which employ a hybrid sampling approach to balance diverse alignment objectives effectively. This sampling method positions HummerRM as an ideal model for domain-specific further fine-tuning and reducing vulnerabilities to attacks."
                    },
                    {
                        "title": "A Dialogue-based Information Extraction System for Medical Insurance Assessment",
                        "abstract": "In the Chinese medical insurance industry, the assessor's role is essential and requires significant efforts to converse with the claimant. This is a highly professional job that involves many parts, such as identifying personal information, collecting related evidence, and making a final insurance report. Due to the coronavirus (COVID-19) pandemic, the previous offline insurance assessment has to be conducted online. However, for the junior assessor often lacking practical experience, it is not easy to quickly handle such a complex online procedure, yet this is important as the insurance company needs to decide how much compensation the claimant should receive based on the assessor's feedback. In order to promote assessors' work efficiency and speed up the overall procedure, in this paper, we propose a dialogue-based information extraction system that integrates advanced NLP technologies for medical insurance assessment. With the assistance of our system, the average time cost of the procedure is reduced from 55 minutes to 35 minutes, and the total human resources cost is saved 30% compared with the previous offline procedure. Until now, the system has already served thousands of online claim cases."
                    },
                    {
                        "title": "Towards Generalized Models for Task-oriented Dialogue Modeling on Spoken Conversations",
                        "abstract": "Building robust and general dialogue models for spoken conversations is challenging due to the gap in distributions of spoken and written data. This paper presents our approach to build generalized models for the Knowledge-grounded Task-oriented Dialogue Modeling on Spoken Conversations Challenge of DSTC-10. In order to mitigate the discrepancies between spoken and written text, we mainly employ extensive data augmentation strategies on written data, including artificial error injection and round-trip text-speech transformation. To train robust models for spoken conversations, we improve pre-trained language models, and apply ensemble algorithms for each sub-task. Typically, for the detection task, we fine-tune \\roberta and ELECTRA, and run an error-fixing ensemble algorithm. For the selection task, we adopt a two-stage framework that consists of entity tracking and knowledge ranking, and propose a multi-task learning method to learn multi-level semantic information by domain classification and entity selection. For the generation task, we adopt a cross-validation data process to improve pre-trained generative language models, followed by a consensus decoding algorithm, which can add arbitrary features like relative \\rouge metric, and tune associated feature weights toward \\bleu directly. Our approach ranks third on the objective evaluation and second on the final official human evaluation."
                    },
                    {
                        "title": "Towards Hierarchical Policy Learning for Conversational Recommendation with Hypergraph-based Reinforcement Learning",
                        "abstract": "Conversational recommendation systems (CRS) aim to timely and proactively acquire user dynamic preferred attributes through conversations for item recommendation. In each turn of CRS, there naturally have two decision-making processes with different roles that influence each other: 1) director, which is to select the follow-up option (i.e., ask or recommend) that is more effective for reducing the action space and acquiring user preferences; and 2) actor, which is to accordingly choose primitive actions (i.e., asked attribute or recommended item) that satisfy user preferences and give feedback to estimate the effectiveness of the director's option. However, existing methods heavily rely on a unified decision-making module or heuristic rules, while neglecting to distinguish the roles of different decision procedures, as well as the mutual influences between them. To address this, we propose a novel Director-Actor Hierarchical Conversational Recommender (DAHCR), where the director selects the most effective option, followed by the actor accordingly choosing primitive actions that satisfy user preferences. Specifically, we develop a dynamic hypergraph to model user preferences and introduce an intrinsic motivation to train from weak supervision over the director. Finally, to alleviate the bad effect of model bias on the mutual influence between the director and actor, we model the director's option by sampling from a categorical distribution. Extensive experiments demonstrate that DAHCR outperforms state-of-the-art methods."
                    },
                    {
                        "title": "Multi-view Hypergraph Contrastive Policy Learning for Conversational Recommendation",
                        "abstract": "Conversational recommendation systems (CRS) aim to interactively acquire user preferences and accordingly recommend items to users. Accurately learning the dynamic user preferences is of crucial importance for CRS. Previous works learn the user preferences with pairwise relations from the interactive conversation and item knowledge, while largely ignoring the fact that factors for a relationship in CRS are multiplex. Specifically, the user likes/dislikes the items that satisfy some attributes (Like/Dislike view). Moreover social influence is another important factor that affects user preference towards the item (Social view), while is largely ignored by previous works in CRS. The user preferences from these three views are inherently different but also correlated as a whole. The user preferences from the same views should be more similar than that from different views. The user preferences from Like View should be similar to Social View while different from Dislike View. To this end, we propose a novel model, namely Multi-view Hypergraph Contrastive Policy Learning (MHCPL). Specifically, MHCPL timely chooses useful social information according to the interactive history and builds a dynamic hypergraph with three types of multiplex relations from different views. The multiplex relations in each view are successively connected according to their generation order."
                    },
                    {
                        "title": "Long Short-Term Sample Distillation",
                        "abstract": "In the past decade, there has been substantial progress at training increasingly deep neural networks. Recent advances within the teacher--student training paradigm have established that information about past training updates show promise as a source of guidance during subsequent training steps. Based on this notion, in this paper, we propose Long Short-Term Sample Distillation, a novel training policy that simultaneously leverages multiple phases of the previous training process to guide the later training updates to a neural network, while efficiently proceeding in just one single generation pass. With Long Short-Term Sample Distillation, the supervision signal for each sample is decomposed into two parts: a long-term signal and a short-term one. The long-term teacher draws on snapshots from several epochs ago in order to provide steadfast guidance and to guarantee teacher--student differences, while the short-term one yields more up-to-date cues with the goal of enabling higher-quality updates. Moreover, the teachers for each sample are unique, such that, overall, the model learns from a very diverse set of teachers. Comprehensive experimental results across a range of vision and NLP tasks demonstrate the effectiveness of this new training method."
                    },
                    {
                        "title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems",
                        "abstract": "Large language models (LLMs) have strong capabilities in solving diverse natural language processing tasks. However, the safety and security issues of LLM systems have become the major obstacle to their widespread application. Many studies have extensively investigated risks in LLM systems and developed the corresponding mitigation strategies. Leading-edge enterprises such as OpenAI, Google, Meta, and Anthropic have also made lots of efforts on responsible LLMs. Therefore, there is a growing need to organize the existing studies and establish comprehensive taxonomies for the community. In this paper, we delve into four essential modules of an LLM system, including an input module for receiving prompts, a language model trained on extensive corpora, a toolchain module for development and deployment, and an output module for exporting LLM-generated content. Based on this, we propose a comprehensive taxonomy, which systematically analyzes potential risks associated with each module of an LLM system and discusses the corresponding mitigation strategies. Furthermore, we review prevalent benchmarks, aiming to facilitate the risk assessment of LLM systems. We hope that this paper can help LLM participants embrace a systematic perspective to build their responsible LLM systems."
                    }
                ]
            },
            "0692f217-da42-44d0-8148-a3a1203e1242": {
                "pk": "0692f217-da42-44d0-8148-a3a1203e1242",
                "name": "Peng Xu",
                "collaborators": [
                    "Pascale Fung",
                    "Mokshay Madiman",
                    "Denilson Barbosa"
                ],
                "domain": [
                    "Representation Theory",
                    "Number Theory",
                    "Machine Learning",
                    "Malware Detection"
                ],
                "publications": [
                    {
                        "title": "Mod-$p$ maximal compact inductions do not have irreducible admissible subrepresentations",
                        "abstract": "Let $p$ be a prime number. We show in this short note that mod-$p$ maximal compact inductions of a $p$-adic split reductive group do not have irreducible admissible subrepresentations."
                    },
                    {
                        "title": "On certain Iwahori--Hecke modules of $GL_3$ in characteristic $p$",
                        "abstract": "In this note, we show that the natural analogue of certain finiteness result of Barthel--Livn$\\acute{\\text{e}}$ on $GL_2$ fails for $GL_3$. More precisely, within the pro-$p$-Iwahori invariants of a maximal compact induction of $GL_3$, we show there exist non-zero Iwahori--Hecke submodules of \\emph{infinite} codimension."
                    },
                    {
                        "title": "A note on the universal supersingular quotients of $U(2,1)$",
                        "abstract": "Let $G$ be the unramified unitary group $U(2, 1)(E/F)$ defined over a non-archimedean local field $F$ of residue characteristic $p\\neq 2$. In this note, we prove the universal supersingular quotients of $G$ are not irreducible in general."
                    },
                    {
                        "title": "Freeness of spherical Hecke modules of unramified $U(2,1)$ in characteristic $p$",
                        "abstract": "Let $F$ be a non-archimedean local field of odd residue characteristic $p$. Let $G$ be the unramified unitary group $U(2, 1)(E/F)$ in three variables, and $K$ be a maximal compact open subgroup of $G$. For an irreducible smooth representation $\\sigma$ of $K$ over $\\overline{\\mathbf{F}}_p$, we prove that the compactly induced representation $\\text{ind}^G _K \\sigma$ is free of infinite rank over the spherical Hecke algebra $\\mathcal{H}(K, \\sigma)$."
                    },
                    {
                        "title": "Restriction of $p$-modular representations of $U(2, 1)$ to a Borel subgroup",
                        "abstract": "Let $G$ be the unramified unitary group $U(2, 1)(E/F)$ defined over a non-archimedean local field $F$ of odd residue characteristic $p$, and $B$ be the standard Borel subgroup of $G$. In this note, we study the problem of the restriction of irreducible smooth $\\overline{\\mathbf{F}}_p$-representations of $G$ to $B$, and we obtain various results which are analogous to that of Pa$\\check{\\text{s}}$k$\\bar{\\text{u}}$nas on $GL_2 (F)$ (\\cite{Pas07})."
                    },
                    {
                        "title": "A remark on the simple cuspidal representations of GL(n, F)",
                        "abstract": "Let $F$ be a non-archimedean local field of residue characteristic $p$, $G$ be the group $GL(n, F)$. In this note, under the assumption $(n, p)=1$, we show a simple cuspidal representation $\\pi$ (that with normalized level $\\frac{1}{n}$) of $G$ is determined uniquely up to isomorphism by the local constants of $\\chi\\circ \\text{det}\\otimes \\pi$ for all characters $\\chi$ of $F^\\times$."
                    },
                    {
                        "title": "On the pro-$p$-Iwahori invariants of supersingular representations of unramified $U(2, 1)$",
                        "abstract": "Let $G$ be the unramified unitary group $U(2, 1)(E/F)$ over a non-archimedean local field $F$ of odd residue characteristic $p$. In this paper, for any supersingular representation of $G$ that contains the Steinberg weight, we prove its pro-$p$-Iwahori invariants, as a right module over the pro-$p$-Iwahori--Hecke algebra of $G$, is \\emph{not} simple."
                    },
                    {
                        "title": "On certain Iwahori representations of unramified $U(2, 1)$ in characteristic $p$",
                        "abstract": "Let $F$ be a non-archimedean local field of odd residue characteristic $p$. Let $G$ be the unramified unitary group $U(2, 1)(E/F)$, and $K$ be a maximal compact open subgroup of $G$. For an $\\overline{\\mathbf{F}}_p$-smooth representation $\\pi$ of $G$ containing a weight $\\sigma$ of $K$, we follow the work of Hu (\\cite{Hu12}) to attach $\\pi$ a certain $I_K$-subrepresentation, where $I_K$ is the Iwahori subgroup in $K$. In terms of such an $I_K$-subrepresentation, we prove a sufficient condition for $\\pi$ to be non-finitely presented. We determine such an $I_K$-subrepresentation explicitly, when $\\pi$ is either a spherical universal Hecke module or an irreducible principal series."
                    },
                    {
                        "title": "Android-COCO: Android Malware Detection with Graph Neural Network for Byte- and Native-Code",
                        "abstract": "With the popularity of Android growing exponentially, the amount of malware has significantly exploded. It is arguably one of the most viral problems on mobile platforms. Recently, various approaches have been introduced to detect Android malware, the majority of these are either based on the Manifest File features or the structural information, such as control flow graph and API calls. Among those methods, nearly all of them only consider the Java byte-code as the target to detect malicious behaviors. However, Recent research and our own statistics show that native payloads are commonly used in both benign and malicious apps. Current state-of-the-art Android static analysis tools avoid handling native method invocation. None of those tools have the capability to capture the inter-language behaviors.   In this work, we explore an ensemble mechanism, which presents how the combination of byte-code and native-code analysis of Android applications can be efficiently used to cope with the advanced sophistication of Android malware. We, therefore, present a multi-layer approach that utilizes deep learning, natural language processing (NLP), as well as graph embedding techniques to handle the threats of Android malware, both from the Java byte-code and native code. After that, we design an ensemble algorithm to get the final result of malware detection system. To be specific, the first layer of our detection approach operates on the byte-code of application and the native code level, whereas the second layer focuses on the ensemble algorithm. Large-scale experiments on 100,113 samples (35,113 malware and 65,000 benign) show that only byte-code sub-system yields 99.8% accuracy and native-code sub-system yields an accuracy of 96.6%, whereas the Android-COCO method attains an accuracy of 99.86% which outperforms various related works."
                    },
                    {
                        "title": "A novel repetition normalized adversarial reward for headline generation",
                        "abstract": "While reinforcement learning can effectively improve language generation models, it often suffers from generating incoherent and repetitive phrases \\cite{paulus2017deep}. In this paper, we propose a novel repetition normalized adversarial reward to mitigate these problems. Our repetition penalized reward can greatly reduce the repetition rate and adversarial training mitigates generating incoherent phrases. Our model significantly outperforms the baseline model on ROUGE-1\\,(+3.24), ROUGE-L\\,(+2.25), and a decreased repetition-rate (-4.98\\%)."
                    },
                    {
                        "title": "The norm of the Fourier transform on compact or discrete abelian groups",
                        "abstract": "We calculate the norm of the Fourier operator from $L^p(X)$ to $L^q(\\hat{X})$ when $X$ is an infinite locally compact abelian group that is, furthermore, compact or discrete. This subsumes the sharp Hausdorff-Young inequality on such groups. In particular, we identify the region in $(p,q)$-space where the norm is infinite, generalizing a result of Fournier, and setting up a contrast with the case of finite abelian groups, where the norm was determined by Gilbert and Rzeszotnik. As an application, uncertainty principles on such groups expressed in terms of R\\'enyi entropies are discussed."
                    },
                    {
                        "title": "Investigations on Knowledge Base Embedding for Relation Prediction and Extraction",
                        "abstract": "We report an evaluation of the effectiveness of the existing knowledge base embedding models for relation prediction and for relation extraction on a wide range of benchmarks. We also describe a new benchmark, which is much larger and complex than previous ones, which we introduce to help validate the effectiveness of both tasks. The results demonstrate that knowledge base embedding models are generally effective for relation prediction but unable to give improvements for the state-of-art neural relation extraction model with the existing strategies, while pointing limitations of existing methods."
                    }
                ]
            },
            "7140f581-b36a-43c1-a8df-6faa0c6e9413": {
                "pk": "7140f581-b36a-43c1-a8df-6faa0c6e9413",
                "name": "Hongning Wang",
                "collaborators": [
                    "Nan Wang",
                    "Yiling Jia",
                    "Huazheng Wang",
                    "Chuanhao Li",
                    "Zhendong Chu",
                    "Zhen Qin",
                    "Xuanhui Wang",
                    "Zhiwei Wang",
                    "Yiyi Tao",
                    "Yue Yin"
                ],
                "domain": [
                    "Recommender Systems",
                    "Online Learning",
                    "Bandit Algorithms",
                    "Explainable AI"
                ],
                "publications": [
                    {
                        "title": "Directional Multivariate Ranking",
                        "abstract": "User-provided multi-aspect evaluations manifest users' detailed feedback on the recommended items and enable fine-grained understanding of their preferences. Extensive studies have shown that modeling such data greatly improves the effectiveness and explainability of the recommendations. However, as ranking is essential in recommendation, there is no principled solution yet for collectively generating multiple item rankings over different aspects. In this work, we propose a directional multi-aspect ranking criterion to enable a holistic ranking of items with respect to multiple aspects. Specifically, we view multi-aspect evaluation as an integral effort from a user that forms a vector of his/her preferences over aspects. Our key insight is that the direction of the difference vector between two multi-aspect preference vectors reveals the pairwise order of comparison. Hence, it is necessary for a multi-aspect ranking criterion to preserve the observed directions from such pairwise comparisons. We further derive a complete solution for the multi-aspect ranking problem based on a probabilistic multivariate tensor factorization model. Comprehensive experimental analysis on a large TripAdvisor multi-aspect rating dataset and a Yelp review text dataset confirms the effectiveness of our solution."
                    },
                    {
                        "title": "BPMR: Bayesian Probabilistic Multivariate Ranking",
                        "abstract": "Multi-aspect user preferences are attracting wider attention in recommender systems, as they enable more detailed understanding of users' evaluations of items. Previous studies show that incorporating multi-aspect preferences can greatly improve the performance and explainability of recommendation. However, as recommendation is essentially a ranking problem, there is no principled solution for ranking multiple aspects collectively to enhance the recommendation.   In this work, we derive a multi-aspect ranking criterion. To maintain the dependency among different aspects, we propose to use a vectorized representation of multi-aspect ratings and develop a probabilistic multivariate tensor factorization framework (PMTF). The framework naturally leads to a probabilistic multi-aspect ranking criterion, which generalizes the single-aspect ranking to a multivariate fashion. Experiment results on a large multi-aspect review rating dataset confirmed the effectiveness of our solution."
                    },
                    {
                        "title": "Asynchronous Upper Confidence Bound Algorithms for Federated Linear Bandits",
                        "abstract": "Linear contextual bandit is a popular online learning problem. It has been mostly studied in centralized learning settings. With the surging demand of large-scale decentralized model learning, e.g., federated learning, how to retain regret minimization while reducing communication cost becomes an open challenge. In this paper, we study linear contextual bandit in a federated learning setting. We propose a general framework with asynchronous model update and communication for a collection of homogeneous clients and heterogeneous clients, respectively. Rigorous theoretical analysis is provided about the regret and communication cost under this distributed learning framework; and extensive empirical evaluations demonstrate the effectiveness of our solution."
                    },
                    {
                        "title": "Improve Learning from Crowds via Generative Augmentation",
                        "abstract": "Crowdsourcing provides an efficient label collection schema for supervised machine learning. However, to control annotation cost, each instance in the crowdsourced data is typically annotated by a small number of annotators. This creates a sparsity issue and limits the quality of machine learning models trained on such data. In this paper, we study how to handle sparsity in crowdsourced data using data augmentation. Specifically, we propose to directly learn a classifier by augmenting the raw sparse annotations. We implement two principles of high-quality augmentation using Generative Adversarial Networks: 1) the generated annotations should follow the distribution of authentic ones, which is measured by a discriminator; 2) the generated annotations should have high mutual information with the ground-truth labels, which is measured by an auxiliary network. Extensive experiments and comparisons against an array of state-of-the-art learning from crowds methods on three real-world datasets proved the effectiveness of our data augmentation framework. It shows the potential of our algorithm for low-budget crowdsourcing in general."
                    },
                    {
                        "title": "Learning Neural Ranking Models Online from Implicit User Feedback",
                        "abstract": "Existing online learning to rank (OL2R) solutions are limited to linear models, which are incompetent to capture possible non-linear relations between queries and documents. In this work, to unleash the power of representation learning in OL2R, we propose to directly learn a neural ranking model from users' implicit feedback (e.g., clicks) collected on the fly. We focus on RankNet and LambdaRank, due to their great empirical success and wide adoption in offline settings, and control the notorious explore-exploit trade-off based on the convergence analysis of neural networks using neural tangent kernel. Specifically, in each round of result serving, exploration is only performed on document pairs where the predicted rank order between the two documents is uncertain; otherwise, the ranker's predicted order will be followed in result ranking. We prove that under standard assumptions our OL2R solution achieves a gap-dependent upper regret bound of $O(\\log^2(T))$, in which the regret is defined on the total number of mis-ordered pairs over $T$ rounds. Comparisons against an extensive set of state-of-the-art OL2R baselines on two public learning to rank benchmark datasets demonstrate the effectiveness of the proposed solution."
                    },
                    {
                        "title": "Communication Efficient Federated Learning for Generalized Linear Bandits",
                        "abstract": "Contextual bandit algorithms have been recently studied under the federated learning setting to satisfy the demand of keeping data decentralized and pushing the learning of bandit models to the client side. But limited by the required communication efficiency, existing solutions are restricted to linear models to exploit their closed-form solutions for parameter estimation. Such a restricted model choice greatly hampers these algorithms' practical utility. In this paper, we take the first step to addressing this challenge by studying generalized linear bandit models under the federated learning setting. We propose a communication-efficient solution framework that employs online regression for local update and offline regression for global update. We rigorously proved, though the setting is more general and challenging, our algorithm can attain sub-linear rate in both regret and communication cost, which is also validated by our extensive empirical evaluations."
                    },
                    {
                        "title": "Scalable Exploration for Neural Online Learning to Rank with Perturbed Feedback",
                        "abstract": "Deep neural networks (DNNs) demonstrate significant advantages in improving ranking performance in retrieval tasks. Driven by the recent technical developments in optimization and generalization of DNNs, learning a neural ranking model online from its interactions with users becomes possible. However, the required exploration for model learning has to be performed in the entire neural network parameter space, which is prohibitively expensive and limits the application of such online solutions in practice.   In this work, we propose an efficient exploration strategy for online interactive neural ranker learning based on the idea of bootstrapping. Our solution employs an ensemble of ranking models trained with perturbed user click feedback. The proposed method eliminates explicit confidence set construction and the associated computational overhead, which enables the online neural rankers' training to be efficiently executed in practice with theoretical guarantees. Extensive comparisons with an array of state-of-the-art OL2R algorithms on two public learning to rank benchmark datasets demonstrate the effectiveness and computational efficiency of our proposed neural OL2R solution."
                    },
                    {
                        "title": "Meta-Reinforcement Learning via Exploratory Task Clustering",
                        "abstract": "Meta-reinforcement learning (meta-RL) aims to quickly solve new tasks by leveraging knowledge from prior tasks. However, previous studies often assume a single mode homogeneous task distribution, ignoring possible structured heterogeneity among tasks. Leveraging such structures can better facilitate knowledge sharing among related tasks and thus improve sample efficiency. In this paper, we explore the structured heterogeneity among tasks via clustering to improve meta-RL. We develop a dedicated exploratory policy to discover task structures via divide-and-conquer. The knowledge of the identified clusters helps to narrow the search space of task-specific information, leading to more sample efficient policy adaptation. Experiments on various MuJoCo tasks showed the proposed method can unravel cluster structures effectively in both rewards and state dynamics, proving strong advantages against a set of state-of-the-art baselines."
                    },
                    {
                        "title": "Calibrating Explore-Exploit Trade-off for Fair Online Learning to Rank",
                        "abstract": "Online learning to rank (OL2R) has attracted great research interests in recent years, thanks to its advantages in avoiding expensive relevance labeling as required in offline supervised ranking model learning. Such a solution explores the unknowns (e.g., intentionally present selected results on top positions) to improve its relevance estimation. This however triggers concerns on its ranking fairness: different groups of items might receive differential treatments during the course of OL2R. But existing fair ranking solutions usually require the knowledge of result relevance or a performing ranker beforehand, which contradicts with the setting of OL2R and thus cannot be directly applied to guarantee fairness.   In this work, we propose a general framework to achieve fairness defined by group exposure in OL2R. The key idea is to calibrate exploration and exploitation for fairness control, relevance learning and online ranking quality. In particular, when the model is exploring a set of results for relevance feedback, we confine the exploration within a subset of random permutations, where fairness across groups is maintained while the feedback is still unbiased. Theoretically we prove such a strategy introduces minimum distortion in OL2R's regret to obtain fairness. Extensive empirical analysis is performed on two public learning to rank benchmark datasets to demonstrate the effectiveness of the proposed solution compared to existing fair OL2R solutions."
                    },
                    {
                        "title": "Non-Clicks Mean Irrelevant? Propensity Ratio Scoring As a Correction",
                        "abstract": "Recent advances in unbiased learning to rank (LTR) count on Inverse Propensity Scoring (IPS) to eliminate bias in implicit feedback. Though theoretically sound in correcting the bias introduced by treating clicked documents as relevant, IPS ignores the bias caused by (implicitly) treating non-clicked ones as irrelevant. In this work, we first rigorously prove that such use of click data leads to unnecessary pairwise comparisons between relevant documents, which prevent unbiased ranker optimization. Based on the proof, we derive a simple yet well justified new weighting scheme, called Propensity Ratio Scoring (PRS), which provides treatments on both clicks and non-clicks. Besides correcting the bias in clicks, PRS avoids relevant-relevant document comparisons in LTR training and enjoys a lower variability. Our extensive empirical evaluations confirm that PRS ensures a more effective use of click data and improved performance in both synthetic data from a set of LTR benchmarks, as well as in the real-world large-scale data from GMail search."
                    },
                    {
                        "title": "Stealthy Adversarial Attacks on Stochastic Multi-Armed Bandits",
                        "abstract": "Adversarial attacks against stochastic multi-armed bandit (MAB) algorithms have been extensively studied in the literature. In this work, we focus on reward poisoning attacks and find most existing attacks can be easily detected by our proposed detection method based on the test of homogeneity, due to their aggressive nature in reward manipulations. This motivates us to study the notion of stealthy attack against stochastic MABs and investigate the resulting attackability. Our analysis shows that against two popularly employed MAB algorithms, UCB1 and $\\epsilon$-greedy, the success of a stealthy attack depends on the environmental conditions and the realized reward of the arm pulled in the first round. We also analyze the situation for general MAB algorithms equipped with our attack detection method and find that it is possible to have a stealthy attack that almost always succeeds. This brings new insights into the security risks of MAB algorithms."
                    },
                    {
                        "title": "The FacT: Taming Latent Factor Models for Explainability with Factorization Trees",
                        "abstract": "Latent factor models have achieved great success in personalized recommendations, but they are also notoriously difficult to explain. In this work, we integrate regression trees to guide the learning of latent factor models for recommendation, and use the learnt tree structure to explain the resulting latent factors. Specifically, we build regression trees on users and items respectively with user-generated reviews, and associate a latent profile to each node on the trees to represent users and items. With the growth of regression tree, the latent factors are gradually refined under the regularization imposed by the tree structure. As a result, we are able to track the creation of latent profiles by looking into the path of each factor on regression trees, which thus serves as an explanation for the resulting recommendations. Extensive experiments on two large collections of Amazon and Yelp reviews demonstrate the advantage of our model over several competitive baseline algorithms. Besides, our extensive user study also confirms the practical value of explainable recommendations generated by our model."
                    },
                    {
                        "title": "Explainable Recommendation via Multi-Task Learning in Opinionated Text Data",
                        "abstract": "Explaining automatically generated recommendations allows users to make more informed and accurate decisions about which results to utilize, and therefore improves their satisfaction. In this work, we develop a multi-task learning solution for explainable recommendation. Two companion learning tasks of user preference modeling for recommendation} and \\textit{opinionated content modeling for explanation are integrated via a joint tensor factorization. As a result, the algorithm predicts not only a user's preference over a list of items, i.e., recommendation, but also how the user would appreciate a particular item at the feature level, i.e., opinionated textual explanation. Extensive experiments on two large collections of Amazon and Yelp reviews confirmed the effectiveness of our solution in both recommendation and explanation tasks, compared with several existing recommendation algorithms. And our extensive user study clearly demonstrates the practical value of the explainable recommendations generated by our algorithm."
                    },
                    {
                        "title": "When Are Linear Stochastic Bandits Attackable?",
                        "abstract": "We study adversarial attacks on linear stochastic bandits: by manipulating the rewards, an adversary aims to control the behaviour of the bandit algorithm. Perhaps surprisingly, we first show that some attack goals can never be achieved. This is in sharp contrast to context-free stochastic bandits, and is intrinsically due to the correlation among arms in linear stochastic bandits. Motivated by this finding, this paper studies the attackability of a $k$-armed linear bandit environment. We first provide a complete necessity and sufficiency characterization of attackability based on the geometry of the arms' context vectors. We then propose a two-stage attack method against LinUCB and Robust Phase Elimination. The method first asserts whether the given environment is attackable; and if yes, it poisons the rewards to force the algorithm to pull a target arm linear times using only a sublinear cost. Numerical experiments further validate the effectiveness and cost-efficiency of the proposed attack method."
                    },
                    {
                        "title": "Graph-based Extractive Explainer for Recommendations",
                        "abstract": "Explanations in a recommender system assist users in making informed decisions among a set of recommended items. Great research attention has been devoted to generating natural language explanations to depict how the recommendations are generated and why the users should pay attention to them. However, due to different limitations of those solutions, e.g., template-based or generation-based, it is hard to make the explanations easily perceivable, reliable and personalized at the same time.   In this work, we develop a graph attentive neural network model that seamlessly integrates user, item, attributes, and sentences for extraction-based explanation. The attributes of items are selected as the intermediary to facilitate message passing for user-item specific evaluation of sentence relevance. And to balance individual sentence relevance, overall attribute coverage, and content redundancy, we solve an integer linear programming problem to make the final selection of sentences. Extensive empirical evaluations against a set of state-of-the-art baseline methods on two benchmark review datasets demonstrated the generation quality of the proposed solution."
                    },
                    {
                        "title": "Communication Efficient Distributed Learning for Kernelized Contextual Bandits",
                        "abstract": "We tackle the communication efficiency challenge of learning kernelized contextual bandits in a distributed setting. Despite the recent advances in communication-efficient distributed bandit learning, existing solutions are restricted to simple models like multi-armed bandits and linear bandits, which hamper their practical utility. In this paper, instead of assuming the existence of a linear reward mapping from the features to the expected rewards, we consider non-linear reward mappings, by letting agents collaboratively search in a reproducing kernel Hilbert space (RKHS). This introduces significant challenges in communication efficiency as distributed kernel learning requires the transfer of raw data, leading to a communication cost that grows linearly w.r.t. time horizon $T$. We addresses this issue by equipping all agents to communicate via a common Nystr\\\"{o}m embedding that gets updated adaptively as more data points are collected. We rigorously proved that our algorithm can attain sub-linear rate in both regret and communication cost."
                    },
                    {
                        "title": "Dynamic Global Sensitivity for Differentially Private Contextual Bandits",
                        "abstract": "Bandit algorithms have become a reference solution for interactive recommendation. However, as such algorithms directly interact with users for improved recommendations, serious privacy concerns have been raised regarding its practical use. In this work, we propose a differentially private linear contextual bandit algorithm, via a tree-based mechanism to add Laplace or Gaussian noise to model parameters. Our key insight is that as the model converges during online update, the global sensitivity of its parameters shrinks over time (thus named dynamic global sensitivity). Compared with existing solutions, our dynamic global sensitivity analysis allows us to inject less noise to obtain $(\\epsilon, \\delta)$-differential privacy with added regret caused by noise injection in $\\tilde O(\\log{T}\\sqrt{T}/\\epsilon)$. We provide a rigorous theoretical analysis over the amount of noise added via dynamic global sensitivity and the corresponding upper regret bound of our proposed algorithm. Experimental results on both synthetic and real-world datasets confirmed the algorithm's advantage against existing solutions."
                    },
                    {
                        "title": "Unbiased Graph Embedding with Biased Graph Observations",
                        "abstract": "Graph embedding techniques are pivotal in real-world machine learning tasks that operate on graph-structured data, such as social recommendation and protein structure modeling. Embeddings are mostly performed on the node level for learning representations of each node. Since the formation of a graph is inevitably affected by certain sensitive node attributes, the node embeddings can inherit such sensitive information and introduce undesirable biases in downstream tasks. Most existing works impose ad-hoc constraints on the node embeddings to restrict their distributions for unbiasedness/fairness, which however compromise the utility of the resulting embeddings. In this paper, we propose a principled new way for unbiased graph embedding by learning node embeddings from an underlying bias-free graph, which is not influenced by sensitive node attributes. Motivated by this new perspective, we propose two complementary methods for uncovering such an underlying graph, with the goal of introducing minimum impact on the utility of the embeddings. Both our theoretical justification and extensive experimental comparisons against state-of-the-art solutions demonstrate the effectiveness of our proposed methods."
                    },
                    {
                        "title": "Explanation as a Defense of Recommendation",
                        "abstract": "Textual explanations have proved to help improve user satisfaction on machine-made recommendations. However, current mainstream solutions loosely connect the learning of explanation with the learning of recommendation: for example, they are often separately modeled as rating prediction and content generation tasks. In this work, we propose to strengthen their connection by enforcing the idea of sentiment alignment between a recommendation and its corresponding explanation. At training time, the two learning tasks are joined by a latent sentiment vector, which is encoded by the recommendation module and used to make word choices for explanation generation. At both training and inference time, the explanation module is required to generate explanation text that matches sentiment predicted by the recommendation module. Extensive experiments demonstrate our solution outperforms a rich set of baselines in both recommendation and explanation tasks, especially on the improved quality of its generated explanations. More importantly, our user studies confirm our generated explanations help users better recognize the differences between recommended items and understand why an item is recommended."
                    },
                    {
                        "title": "PairRank: Online Pairwise Learning to Rank by Divide-and-Conquer",
                        "abstract": "Online Learning to Rank (OL2R) eliminates the need of explicit relevance annotation by directly optimizing the rankers from their interactions with users. However, the required exploration drives it away from successful practices in offline learning to rank, which limits OL2R's empirical performance and practical applicability. In this work, we propose to estimate a pairwise learning to rank model online. In each round, candidate documents are partitioned and ranked according to the model's confidence on the estimated pairwise rank order, and exploration is only performed on the uncertain pairs of documents, i.e., \\emph{divide-and-conquer}. Regret directly defined on the number of mis-ordered pairs is proven, which connects the online solution's theoretical convergence with its expected ranking performance. Comparisons against an extensive list of OL2R baselines on two public learning to rank benchmark datasets demonstrate the effectiveness of the proposed solution."
                    }
                ]
            },
            "e8ab2962-67b5-4db2-b7b3-2ea170b2a932": {
                "pk": "e8ab2962-67b5-4db2-b7b3-2ea170b2a932",
                "name": "Minlie Huang",
                "collaborators": [
                    "Xiaoyan Zhu",
                    "Jian Guan",
                    "Fei Huang",
                    "Zhouxing Shi",
                    "Chujie Zheng",
                    "Zhihong Shao",
                    "Pei Ke",
                    "Han Xiao",
                    "Yansen Wang",
                    "Biao Liu"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Text Generation",
                    "Knowledge Representation",
                    "Sentiment Analysis"
                ],
                "publications": [
                    {
                        "title": "Robustly Leveraging Prior Knowledge in Text Classification",
                        "abstract": "Prior knowledge has been shown very useful to address many natural language processing tasks. Many approaches have been proposed to formalise a variety of knowledge, however, whether the proposed approach is robust or sensitive to the knowledge supplied to the model has rarely been discussed. In this paper, we propose three regularization terms on top of generalized expectation criteria, and conduct extensive experiments to justify the robustness of the proposed methods. Experimental results demonstrate that our proposed methods obtain remarkable improvements and are much more robust than baselines."
                    },
                    {
                        "title": "Mitigating the Learning Bias towards Repetition by Self-Contrastive Training for Open-Ended Generation",
                        "abstract": "Despite the huge progress in myriad generation tasks, pretrained language models (LMs) such as GPT2 still tend to generate repetitive texts with maximization-based decoding algorithms for open-ended generation. We attribute their overestimation of token-level repetition probabilities to the learning bias: LMs capture simple repetitive patterns faster with the MLE loss. We propose self-contrastive training to penalize the output of a premature checkpoint of the same model when it incorrectly predicts repetition, which is shown to mitigate repetition effectively while maintaining fluency on two datasets. Furthermore, we find that LMs use longer-range dependencies to predict repetitive tokens than non-repetitive ones, which may be the cause of sentence-level repetition loops."
                    },
                    {
                        "title": "A Deep Sequential Model for Discourse Parsing on Multi-Party Dialogues",
                        "abstract": "Discourse structures are beneficial for various NLP tasks such as dialogue understanding, question answering, sentiment analysis, and so on. This paper presents a deep sequential model for parsing discourse dependency structures of multi-party dialogues. The proposed model aims to construct a discourse dependency tree by predicting dependency relations and constructing the discourse structure jointly and alternately. It makes a sequential scan of the Elementary Discourse Units (EDUs) in a dialogue. For each EDU, the model decides to which previous EDU the current one should link and what the corresponding relation type is. The predicted link and relation type are then used to build the discourse structure incrementally with a structured encoder. During link prediction and relation classification, the model utilizes not only local information that represents the concerned EDUs, but also global information that encodes the EDU sequence and the discourse structure that is already built at the current step. Experiments show that the proposed model outperforms all the state-of-the-art baselines."
                    },
                    {
                        "title": "UNION: An Unreferenced Metric for Evaluating Open-ended Story Generation",
                        "abstract": "Despite the success of existing referenced metrics (e.g., BLEU and MoverScore), they correlate poorly with human judgments for open-ended text generation including story or dialog generation because of the notorious one-to-many issue: there are many plausible outputs for the same input, which may differ substantially in literal or semantics from the limited number of given references. To alleviate this issue, we propose UNION, a learnable unreferenced metric for evaluating open-ended story generation, which measures the quality of a generated story without any reference. Built on top of BERT, UNION is trained to distinguish human-written stories from negative samples and recover the perturbation in negative stories. We propose an approach of constructing negative samples by mimicking the errors commonly observed in existing NLG models, including repeated plots, conflicting logic, and long-range incoherence. Experiments on two story datasets demonstrate that UNION is a reliable measure for evaluating the quality of generated stories, which correlates better with human judgments and is more generalizable than existing state-of-the-art metrics."
                    },
                    {
                        "title": "Exploring Prompt-based Few-shot Learning for Grounded Dialog Generation",
                        "abstract": "Dialog models can be greatly strengthened through grounding on various external information, but grounded dialog corpora are usually not naturally accessible. In this work, we focus on the few-shot learning for grounded dialog generation (GDG). We first propose a simple prompting method for GDG tasks, where different constructs of model input, such as the grounding source and the conversation context, are distinguished through continuous or discrete prompts. On three typical GDG tasks, we empirically demonstrate and analyze in-depth the effectiveness of our method. We then conduct extensive experiments to thoroughly investigate how our prompting method works with different pre-trained models. We show that prompted language models perform superiorly to conversational models, and further analyze various factors that influence the effects of prompting. Overall, our work introduces a prompt-based perspective to the few-shot learning for GDG tasks, and provides valuable findings and insights for future research."
                    },
                    {
                        "title": "Robustness to Modification with Shared Words in Paraphrase Identification",
                        "abstract": "Revealing the robustness issues of natural language processing models and improving their robustness is important to their performance under difficult situations. In this paper, we study the robustness of paraphrase identification models from a new perspective -- via modification with shared words, and we show that the models have significant robustness issues when facing such modifications. To modify an example consisting of a sentence pair, we either replace some words shared by both sentences or introduce new shared words. We aim to construct a valid new example such that a target model makes a wrong prediction. To find a modification solution, we use beam search constrained by heuristic rules, and we leverage a BERT masked language model for generating substitution words compatible with the context. Experiments show that the performance of the target models has a dramatic drop on the modified examples, thereby revealing the robustness issue. We also show that adversarial training can mitigate this issue."
                    },
                    {
                        "title": "DiscoDVT: Generating Long Text with Discourse-Aware Discrete Variational Transformer",
                        "abstract": "Despite the recent advances in applying pre-trained language models to generate high-quality texts, generating long passages that maintain long-range coherence is yet challenging for these models. In this paper, we propose DiscoDVT, a discourse-aware discrete variational Transformer to tackle the incoherence issue. DiscoDVT learns a discrete variable sequence that summarizes the global structure of the text and then applies it to guide the generation process at each decoding step. To further embed discourse-aware information into the discrete latent representations, we introduce an auxiliary objective to model the discourse relations within the text. We conduct extensive experiments on two open story generation datasets and demonstrate that the latent codes learn meaningful correspondence to the discourse structures that guide the model to generate long texts with better long-range coherence."
                    },
                    {
                        "title": "Answering Open-Domain Multi-Answer Questions via a Recall-then-Verify Framework",
                        "abstract": "Open-domain questions are likely to be open-ended and ambiguous, leading to multiple valid answers. Existing approaches typically adopt the rerank-then-read framework, where a reader reads top-ranking evidence to predict answers. According to our empirical analysis, this framework faces three problems: first, to leverage a large reader under a memory constraint, the reranker should select only a few relevant passages to cover diverse answers, while balancing relevance and diversity is non-trivial; second, the small reading budget prevents the reader from accessing valuable retrieved evidence filtered out by the reranker; third, when using a generative reader to predict answers all at once based on all selected evidence, whether a valid answer will be predicted also pathologically depends on the evidence of some other valid answer(s). To address these issues, we propose to answer open-domain multi-answer questions with a recall-then-verify framework, which separates the reasoning process of each answer so that we can make better use of retrieved evidence while also leveraging large models under the same memory constraint. Our framework achieves state-of-the-art results on two multi-answer datasets, and predicts significantly more gold answers than a rerank-then-read system that uses an oracle reranker."
                    },
                    {
                        "title": "ARAML: A Stable Adversarial Training Framework for Text Generation",
                        "abstract": "Most of the existing generative adversarial networks (GAN) for text generation suffer from the instability of reinforcement learning training algorithms such as policy gradient, leading to unstable performance. To tackle this problem, we propose a novel framework called Adversarial Reward Augmented Maximum Likelihood (ARAML). During adversarial training, the discriminator assigns rewards to samples which are acquired from a stationary distribution near the data rather than the generator's distribution. The generator is optimized with maximum likelihood estimation augmented by the discriminator's rewards instead of policy gradient. Experiments show that our model can outperform state-of-the-art text GANs with a more stable training process."
                    },
                    {
                        "title": "Directed Acyclic Transformer Pre-training for High-quality Non-autoregressive Text Generation",
                        "abstract": "Non-AutoRegressive (NAR) text generation models have drawn much attention because of their significantly faster decoding speed and good generation quality in machine translation. However, in a wider range of text generation tasks, existing NAR models lack proper pre-training, making them still far behind the pre-trained autoregressive models. In this paper, we propose Pre-trained Directed Acyclic Transformer (PreDAT) and a novel pre-training task to promote prediction consistency in NAR generation. Experiments on five text generation tasks show that our PreDAT remarkably outperforms existing pre-trained NAR models (+4.2 scores on average) and even achieves better results than pre-trained autoregressive baselines in n-gram-based metrics, along with 17 times speedup in throughput. Further analysis shows that PreDAT benefits from the unbiased prediction order that alleviates the error accumulation problem in autoregressive generation, which provides new insights into the advantages of NAR generation."
                    },
                    {
                        "title": "Chaining Simultaneous Thoughts for Numerical Reasoning",
                        "abstract": "Given that rich information is hidden behind ubiquitous numbers in text, numerical reasoning over text should be an essential skill of AI systems. To derive precise equations to solve numerical reasoning problems, previous work focused on modeling the structures of equations, and has proposed various structured decoders. Though structure modeling proves to be effective, these structured decoders construct a single equation in a pre-defined autoregressive order, potentially placing an unnecessary restriction on how a model should grasp the reasoning process. Intuitively, humans may have numerous pieces of thoughts popping up in no pre-defined order; thoughts are not limited to the problem at hand, and can even be concerned with other related problems. By comparing diverse thoughts and chaining relevant pieces, humans are less prone to errors. In this paper, we take this inspiration and propose CANTOR, a numerical reasoner that models reasoning steps using a directed acyclic graph where we produce diverse reasoning steps simultaneously without pre-defined decoding dependencies, and compare and chain relevant ones to reach a solution. Extensive experiments demonstrated the effectiveness of CANTOR under both fully-supervised and weakly-supervised settings."
                    },
                    {
                        "title": "Modeling Rich Contexts for Sentiment Classification with LSTM",
                        "abstract": "Sentiment analysis on social media data such as tweets and weibo has become a very important and challenging task. Due to the intrinsic properties of such data, tweets are short, noisy, and of divergent topics, and sentiment classification on these data requires to modeling various contexts such as the retweet/reply history of a tweet, and the social context about authors and relationships. While few prior study has approached the issue of modeling contexts in tweet, this paper proposes to use a hierarchical LSTM to model rich contexts in tweet, particularly long-range context. Experimental results show that contexts can help us to perform sentiment classification remarkably better."
                    },
                    {
                        "title": "From One Point to A Manifold: Knowledge Graph Embedding For Precise Link Prediction",
                        "abstract": "Knowledge graph embedding aims at offering a numerical knowledge representation paradigm by transforming the entities and relations into continuous vector space. However, existing methods could not characterize the knowledge graph in a fine degree to make a precise prediction. There are two reasons: being an ill-posed algebraic system and applying an overstrict geometric form. As precise prediction is critical, we propose an manifold-based embedding principle (\\textbf{ManifoldE}) which could be treated as a well-posed algebraic system that expands the position of golden triples from one point in current models to a manifold in ours. Extensive experiments show that the proposed models achieve substantial improvements against the state-of-the-art baselines especially for the precise prediction task, and yet maintain high efficiency."
                    },
                    {
                        "title": "SSP: Semantic Space Projection for Knowledge Graph Embedding with Text Descriptions",
                        "abstract": "Knowledge representation is an important, long-history topic in AI, and there have been a large amount of work for knowledge graph embedding which projects symbolic entities and relations into low-dimensional, real-valued vector space. However, most embedding methods merely concentrate on data fitting and ignore the explicit semantic expression, leading to uninterpretable representations. Thus, traditional embedding methods have limited potentials for many applications such as question answering, and entity classification. To this end, this paper proposes a semantic representation method for knowledge graph \\textbf{(KSR)}, which imposes a two-level hierarchical generative process that globally extracts many aspects and then locally assigns a specific category in each aspect for every triple. Since both aspects and categories are semantics-relevant, the collection of categories in each aspect is treated as the semantic representation of this triple. Extensive experiments justify our model outperforms other state-of-the-art baselines substantially."
                    },
                    {
                        "title": "An Interpretable Reasoning Network for Multi-Relation Question Answering",
                        "abstract": "Multi-relation Question Answering is a challenging task, due to the requirement of elaborated analysis on questions and reasoning over multiple fact triples in knowledge base. In this paper, we present a novel model called Interpretable Reasoning Network that employs an interpretable, hop-by-hop reasoning process for question answering. The model dynamically decides which part of an input question should be analyzed at each hop; predicts a relation that corresponds to the current parsed results; utilizes the predicted relation to update the question representation and the state of the reasoning process; and then drives the next-hop reasoning. Experiments show that our model yields state-of-the-art results on two datasets. More interestingly, the model can offer traceable and observable intermediate predictions for reasoning analysis and failure diagnosis, thereby allowing manual manipulation in predicting the final answer."
                    },
                    {
                        "title": "Learning to Ask Questions in Open-domain Conversational Systems with Typed Decoders",
                        "abstract": "Asking good questions in large-scale, open-domain conversational systems is quite significant yet rather untouched. This task, substantially different from traditional question generation, requires to question not only with various patterns but also on diverse and relevant topics. We observe that a good question is a natural composition of {\\it interrogatives}, {\\it topic words}, and {\\it ordinary words}. Interrogatives lexicalize the pattern of questioning, topic words address the key information for topic transition in dialogue, and ordinary words play syntactical and grammatical roles in making a natural sentence. We devise two typed decoders (\\textit{soft typed decoder} and \\textit{hard typed decoder}) in which a type distribution over the three types is estimated and used to modulate the final generation distribution. Extensive experiments show that the typed decoders outperform state-of-the-art baselines and can generate more meaningful questions."
                    },
                    {
                        "title": "Story Ending Generation with Incremental Encoding and Commonsense Knowledge",
                        "abstract": "Generating a reasonable ending for a given story context, i.e., story ending generation, is a strong indication of story comprehension. This task requires not only to understand the context clues which play an important role in planning the plot but also to handle implicit knowledge to make a reasonable, coherent story.   In this paper, we devise a novel model for story ending generation. The model adopts an incremental encoding scheme to represent context clues which are spanning in the story context. In addition, commonsense knowledge is applied through multi-source attention to facilitate story comprehension, and thus to help generate coherent and reasonable endings. Through building context clues and using implicit knowledge, the model is able to produce reasonable story endings. context clues implied in the post and make the inference based on it.   Automatic and manual evaluation shows that our model can generate more reasonable story endings than state-of-the-art baselines."
                    },
                    {
                        "title": "A Hierarchical Framework for Relation Extraction with Reinforcement Learning",
                        "abstract": "Most existing methods determine relation types only after all the entities have been recognized, thus the interaction between relation types and entity mentions is not fully modeled. This paper presents a novel paradigm to deal with relation extraction by regarding the related entities as the arguments of a relation. We apply a hierarchical reinforcement learning (HRL) framework in this paradigm to enhance the interaction between entity mentions and relation types. The whole extraction process is decomposed into a hierarchy of two-level RL policies for relation detection and entity extraction respectively, so that it is more feasible and natural to deal with overlapping relations. Our model was evaluated on public datasets collected via distant supervision, and results show that it gains better performance than existing methods and is more powerful for extracting overlapping relations."
                    },
                    {
                        "title": "ChID: A Large-scale Chinese IDiom Dataset for Cloze Test",
                        "abstract": "Cloze-style reading comprehension in Chinese is still limited due to the lack of various corpora. In this paper we propose a large-scale Chinese cloze test dataset ChID, which studies the comprehension of idiom, a unique language phenomenon in Chinese. In this corpus, the idioms in a passage are replaced by blank symbols and the correct answer needs to be chosen from well-designed candidate idioms. We carefully study how the design of candidate idioms and the representation of idioms affect the performance of state-of-the-art models. Results show that the machine accuracy is substantially worse than that of human, indicating a large space for further research."
                    },
                    {
                        "title": "Linguistically Regularized LSTMs for Sentiment Classification",
                        "abstract": "Sentiment understanding has been a long-term goal of AI in the past decades. This paper deals with sentence-level sentiment classification. Though a variety of neural network models have been proposed very recently, however, previous models either depend on expensive phrase-level annotation, whose performance drops substantially when trained with only sentence-level annotation; or do not fully employ linguistic resources (e.g., sentiment lexicons, negation words, intensity words), thus not being able to produce linguistically coherent representations. In this paper, we propose simple models trained with sentence-level annotation, but also attempt to generating linguistically coherent representations by employing regularizers that model the linguistic role of sentiment lexicons, negation words, and intensity words. Results show that our models are effective to capture the sentiment shifting effect of sentiment, negation, and intensity words, while still obtain competitive results without sacrificing the models' simplicity."
                    }
                ]
            }
        }
    },
    "2310.20673": {
        "paper_data": {
            "title": "Balancing Act: Constraining Disparate Impact in Sparse Models",
            "url": "http://arxiv.org/abs/2310.20673v2",
            "arxiv_id": "2310.20673",
            "authors": [
                "Meraj Hashemizadeh",
                "Juan Ramirez",
                "Rohan Sukumaran",
                "Golnoosh Farnadi",
                "Simon Lacoste-Julien",
                "Jose Gallego-Posada"
            ],
            "abstract": "Model pruning is a popular approach to enable the deployment of large deep learning models on edge devices with restricted computational or storage capacities. Although sparse models achieve performance comparable to that of their dense counterparts at the level of the entire dataset, they exhibit high accuracy drops for some data sub-groups. Existing methods to mitigate this disparate impact induced by pruning (i) rely on surrogate metrics that address the problem indirectly and have limited interpretability; or (ii) scale poorly with the number of protected sub-groups in terms of computational cost. We propose a constrained optimization approach that directly addresses the disparate impact of pruning: our formulation bounds the accuracy change between the dense and sparse models, for each sub-group. This choice of constraints provides an interpretable success criterion to determine if a pruned model achieves acceptable disparity levels. Experimental results demonstrate that our technique scales reliably to problems involving large models and hundreds of protected sub-groups.",
            "introduction": "   1 Introduction  Current deep learning practice displays a trend towards larger architectures (Bommasani et\u00a0al., 2021), as exemplified by popular models such as GPT-4 (OpenAI, 2023), Llama 2 (Touvron et\u00a0al., 2023) and DALL-E 2 (Ramesh et\u00a0al., 2022). Model compression techniques such as pruning (Gale et\u00a0al., 2019), knowledge distillation (Hinton et\u00a0al., 2015), or quantization (Gholami et\u00a0al., 2021) are crucial towards enabling the deployment of large models across a wide range of platforms, including resource-constrained edge devices like smartphones.   Despite achieving comparable performance at an aggregate level over the entire dataset, pruned models often exhibit significant accuracy reduction for some data sub-groups (Hooker et\u00a0al., 2019; 2020; Paganini, 2020). In particular, under-represented groups can suffer high performance degradation while the overall performance remains unaffected, thus exacerbating systemic biases in machine learning models. Tran et\u00a0al. (2022) refer to this phenomenon as the disparate impact of pruning.   Existing mitigation methods face challenges in terms of interpretability and scalability to a large number of sub-groups. Tran et\u00a0al. (2022) introduce constraints aiming to equalize the loss of the sparse model across sub-groups. However, their approach does not account for the unequal group-level performance of the dense model. Moreover, while the loss can be a useful surrogate for training, this method addresses the disparate impact issue indirectly as it focuses on controlling the loss, rather than group-level changes in accuracy. Alternatively, Lin et\u00a0al. (2022) compute per-group importance scores for every model parameter to determine the weights to be pruned. This approach becomes prohibitively expensive when the model or the number of sub-groups is large.   In this work, we characterize the disparate impact of pruning in terms of the group-level accuracy gaps between the dense and sparse models. Additionally, we propose a problem formulation that directly addresses the disparate impact of pruning by imposing constraints on the per-group excess accuracy gaps (CEAG). A key advantage of our proposed formulation is that it enjoys interpretable semantics: feasible solutions of our optimization problem correspond to models with low pruning-induced disparity. Finally, our approach introduces a negligible computational overhead (Section\u00a0E.1) compared to (disparity-agnostic) naive fine-tuning of the sparse model, making it applicable to problems with large numbers of groups, such as intersectional fairness tasks.   Fig.\u00a01 illustrates the reliability of our approach at mitigating the disparate impact of pruning. We measure disparity in terms of excess accuracy gaps (EAGs, \u00a73.1). Naive fine-tuning yields models that disproportionately affect group Others, and while the equalized loss formulation mitigates the issue, our formulation consistently reduces the pruning-induced disparity. See \u00a75 for further discussion.   Figure 1:  Left: A dense model is sparsified with GMP, and then subjected to either (i) naive fine-tuning (NFT, using ERM), (ii) equalized loss constraints\u00a0(Tran et\u00a0al., 2022, EL), or (iii) our approach (CEAG). Right: Positive (resp. negative) excess accuracy gaps (EAGs, \u00a73.1) indicate groups whose performance degraded more (resp. less) than the model\u2019s overall accuracy change. Models with low disparate impact have EAGs that concentrate around zero. CEAG consistenly yields models with lower disparity (\u03a8\ud835\udc0f\ud835\udc16subscriptnormal-\u03a8\ud835\udc0f\ud835\udc16\\Psi_{\\text{PW}}roman_\u03a8 start_POSTSUBSCRIPT PW end_POSTSUBSCRIPT, \u00a73.1) than NFT and EL. For example, NFT\u2009 yields a 10% hyper-degradation (EAG, \u03c8gsubscript\ud835\udf13\ud835\udc54\\psi_{g}italic_\u03c8 start_POSTSUBSCRIPT italic_g end_POSTSUBSCRIPT) on group Others.  Results correspond to race prediction on UTKFace, with race as group attribute at 90% sparsity. Metrics are measured on the",
            "references": [
                {
                    "title": "On Learning Fairness and Accuracy on Multiple Subgroups",
                    "abstract": "We propose an analysis in fair learning that preserves the utility of the data while reducing prediction disparities under the criteria of group sufficiency. We focus on the scenario where the data contains multiple or even many subgroups, each with limited number of samples. As a result, we present a principled method for learning a fair predictor for all subgroups via formulating it as a bilevel objective. Specifically, the subgroup specific predictors are learned in the lower-level through a small amount of data and the fair predictor. In the upper-level, the fair predictor is updated to be close to all subgroup specific predictors. We further prove that such a bilevel objective can effectively control the group sufficiency and generalization error. We evaluate the proposed framework on real-world datasets. Empirical evidence suggests the consistently improved fair predictions, as well as the comparable accuracy to the baselines."
                },
                {
                    "title": "Automatic Data Augmentation via Invariance-Constrained Learning",
                    "abstract": "Underlying data structures, such as symmetries or invariances to transformations, are often exploited to improve the solution of learning tasks. However, embedding these properties in models or learning algorithms can be challenging and computationally intensive. Data augmentation, on the other hand, induces these symmetries during training by applying multiple transformations to the input data. Despite its ubiquity, its effectiveness depends on the choices of which transformations to apply, when to do so, and how often. In fact, there is both empirical and theoretical evidence that the indiscriminate use of data augmentation can introduce biases that outweigh its benefits. This work tackles these issues by automatically adapting the data augmentation while solving the learning task. To do so, it formulates data augmentation as an invariance-constrained learning problem and leverages Monte Carlo Markov Chain (MCMC) sampling to solve it. The result is a practical algorithm that not only does away with a priori searches for augmentation distributions, but also dynamically controls if and when data augmentation is applied. Our experiments illustrate the performance of this method, which achieves state-of-the-art results in automatic data augmentation benchmarks for CIFAR datasets. Furthermore, this approach can be used to gather insights on the actual symmetries underlying a learning task."
                },
                {
                    "title": "Controlled Sparsity via Constrained Optimization or: How I Learned to Stop Tuning Penalties and Love Constraints",
                    "abstract": "The performance of trained neural networks is robust to harsh levels of pruning. Coupled with the ever-growing size of deep learning models, this observation has motivated extensive research on learning sparse models. In this work, we focus on the task of controlling the level of sparsity when performing sparse learning. Existing methods based on sparsity-inducing penalties involve expensive trial-and-error tuning of the penalty factor, thus lacking direct control of the resulting model sparsity. In response, we adopt a constrained formulation: using the gate mechanism proposed by Louizos et al. (2018), we formulate a constrained optimization problem where sparsification is guided by the training objective and the desired sparsity target in an end-to-end fashion. Experiments on CIFAR-{10, 100}, TinyImageNet, and ImageNet using WideResNet and ResNet{18, 50} models validate the effectiveness of our proposal and demonstrate that we can reliably achieve pre-determined sparsity targets without compromising on predictive performance."
                },
                {
                    "title": "FairGRAPE: Fairness-aware GRAdient Pruning mEthod for Face Attribute Classification",
                    "abstract": "Existing pruning techniques preserve deep neural networks' overall ability to make correct predictions but may also amplify hidden biases during the compression process. We propose a novel pruning method, Fairness-aware GRAdient Pruning mEthod (FairGRAPE), that minimizes the disproportionate impacts of pruning on different sub-groups. Our method calculates the per-group importance of each model weight and selects a subset of weights that maintain the relative between-group total importance in pruning. The proposed method then prunes network edges with small importance values and repeats the procedure by updating importance values. We demonstrate the effectiveness of our method on four different datasets, FairFace, UTKFace, CelebA, and ImageNet, for the tasks of face attribute classification where our method reduces the disparity in performance degradation by up to 90% compared to the state-of-the-art pruning algorithms. Our method is substantially more effective in a setting with a high pruning rate (99%). The code and dataset used in the experiments are available at https://github.com/Bernardo1998/FairGRAPE"
                },
                {
                    "title": "Pruning has a disparate impact on model accuracy",
                    "abstract": "Network pruning is a widely-used compression technique that is able to significantly scale down overparameterized models with minimal loss of accuracy. This paper shows that pruning may create or exacerbate disparate impacts. The paper sheds light on the factors to cause such disparities, suggesting differences in gradient norms and distance to decision boundary across groups to be responsible for this critical issue. It analyzes these factors in detail, providing both theoretical and empirical support, and proposes a simple, yet effective, solution that mitigates the disparate impacts caused by pruning."
                },
                {
                    "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
                    "abstract": "Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples."
                },
                {
                    "title": "Stochastic Gradient Descent-Ascent: Unified Theory and New Efficient Methods",
                    "abstract": "Stochastic Gradient Descent-Ascent (SGDA) is one of the most prominent algorithms for solving min-max optimization and variational inequalities problems (VIP) appearing in various machine learning tasks. The success of the method led to several advanced extensions of the classical SGDA, including variants with arbitrary sampling, variance reduction, coordinate randomization, and distributed variants with compression, which were extensively studied in the literature, especially during the last few years. In this paper, we propose a unified convergence analysis that covers a large variety of stochastic gradient descent-ascent methods, which so far have required different intuitions, have different applications and have been developed separately in various communities. A key to our unified framework is a parametric assumption on the stochastic estimates. Via our general theoretical framework, we either recover the sharpest known rates for the known special cases or tighten them. Moreover, to illustrate the flexibility of our approach we develop several new variants of SGDA such as a new variance-reduced method (L-SVRGDA), new distributed methods with compression (QSGDA, DIANA-SGDA, VR-DIANA-SGDA), and a new method with coordinate randomization (SEGA-SGDA). Although variants of the new methods are known for solving minimization problems, they were never considered or analyzed for solving min-max problems and VIPs. We also demonstrate the most important properties of the new methods through extensive numerical experiments."
                },
                {
                    "title": "A Lagrangian Duality Approach to Active Learning",
                    "abstract": "We consider the pool-based active learning problem, where only a subset of the training data is labeled, and the goal is to query a batch of unlabeled samples to be labeled so as to maximally improve model performance. We formulate the problem using constrained learning, where a set of constraints bounds the performance of the model on labeled samples. Considering a primal-dual approach, we optimize the primal variables, corresponding to the model parameters, as well as the dual variables, corresponding to the constraints. As each dual variable indicates how significantly the perturbation of the respective constraint affects the optimal value of the objective function, we use it as a proxy of the informativeness of the corresponding training sample. Our approach, which we refer to as Active Learning via Lagrangian dualitY, or ALLY, leverages this fact to select a diverse set of unlabeled samples with the highest estimated dual variables as our query set. We demonstrate the benefits of our approach in a variety of classification and regression tasks and discuss its limitations depending on the capacity of the model used and the degree of redundancy in the dataset. We also examine the impact of the distribution shift induced by active sampling and show that ALLY can be used in a generative mode to create novel, maximally-informative samples."
                },
                {
                    "title": "On the Opportunities and Risks of Foundation Models",
                    "abstract": "AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature."
                },
                {
                    "title": "A Survey of Quantization Methods for Efficient Neural Network Inference",
                    "abstract": "As soon as abstract mathematical computations were adapted to computation on digital computers, the problem of efficient representation, manipulation, and communication of the numerical values in those computations arose. Strongly related to the problem of numerical representation is the problem of quantization: in what manner should a set of continuous real-valued numbers be distributed over a fixed discrete set of numbers to minimize the number of bits required and also to maximize the accuracy of the attendant computations? This perennial problem of quantization is particularly relevant whenever memory and/or computational resources are severely restricted, and it has come to the forefront in recent years due to the remarkable performance of Neural Network models in computer vision, natural language processing, and related areas. Moving from floating-point representations to low-precision fixed integer values represented in four bits or less holds the potential to reduce the memory footprint and latency by a factor of 16x; and, in fact, reductions of 4x to 8x are often realized in practice in these applications. Thus, it is not surprising that quantization has emerged recently as an important and very active sub-area of research in the efficient implementation of computations associated with Neural Networks. In this article, we survey approaches to the problem of quantizing the numerical values in deep Neural Network computations, covering the advantages/disadvantages of current methods. With this survey and its organization, we hope to have presented a useful snapshot of the current research in quantization for Neural Networks and to have given an intelligent organization to ease the evaluation of future research in this area."
                },
                {
                    "title": "Constrained Learning With Non-Convex Losses",
                    "abstract": "Though learning has become a core component of modern information processing, there is now ample evidence that it can lead to biased, unsafe, and prejudiced systems. The need to impose requirements on learning is therefore paramount, especially as it reaches critical applications in social, industrial, and medical domains. However, the non-convexity of most modern statistical problems is only exacerbated by the introduction of constraints. Whereas good unconstrained solutions can often be learned using empirical risk minimization, even obtaining a model that satisfies statistical constraints can be challenging. All the more so, a good one. In this paper, we overcome this issue by learning in the empirical dual domain, where constrained statistical learning problems become unconstrained and deterministic. We analyze the generalization properties of this approach by bounding the empirical duality gap\u2014i.e., the difference between our approximate, tractable solution and the solution of the original (non-convex) statistical problem\u2014and provide a practical constrained learning algorithm. These results establish a constrained counterpart to classical learning theory, enabling the explicit use of constraints in learning. We illustrate this theory and algorithm in rate-constrained learning applications arising in fairness and adversarial robustness."
                },
                {
                    "title": "A Stochastic Optimization Framework for Fair Risk Minimization",
                    "abstract": "Despite the success of large-scale empirical risk minimization (ERM) at achieving high accuracy across a variety of machine learning tasks, fair ERM is hindered by the incompatibility of fairness constraints with stochastic optimization. We consider the problem of fair classification with discrete sensitive attributes and potentially large models and data sets, requiring stochastic solvers. Existing in-processing fairness algorithms are either impractical in the large-scale setting because they require large batches of data at each iteration or they are not guaranteed to converge. In this paper, we develop the first stochastic in-processing fairness algorithm with guaranteed convergence. For demographic parity, equalized odds, and equal opportunity notions of fairness, we provide slight variations of our algorithm--called FERMI--and prove that each of these variations converges in stochastic optimization with any batch size. Empirically, we show that FERMI is amenable to stochastic solvers with multiple (non-binary) sensitive attributes and non-binary targets, performing well even with minibatch size as small as one. Extensive experiments show that FERMI achieves the most favorable tradeoffs between fairness violation and test accuracy across all tested setups compared with state-of-the-art baselines for demographic parity, equalized odds, equal opportunity. These benefits are especially significant with small batch sizes and for non-binary classification with large number of sensitive attributes, making FERMI a practical, scalable fairness algorithm. The code for all of the experiments in this paper is available at: https://github.com/optimization-for-data-driven-science/FERMI."
                },
                {
                    "title": "Near-optimal Local Convergence of Alternating Gradient Descent-Ascent for Minimax Optimization",
                    "abstract": "Smooth minimax games often proceed by simultaneous or alternating gradient updates. Although algorithms with alternating updates are commonly used in practice, the majority of existing theoretical analyses focus on simultaneous algorithms for convenience of analysis. In this paper, we study alternating gradient descent-ascent (Alt-GDA) in minimax games and show that Alt-GDA is superior to its simultaneous counterpart~(Sim-GDA) in many settings. We prove that Alt-GDA achieves a near-optimal local convergence rate for strongly convex-strongly concave (SCSC) problems while Sim-GDA converges at a much slower rate. To our knowledge, this is the \\emph{first} result of any setting showing that Alt-GDA converges faster than Sim-GDA by more than a constant. We further adapt the theory of integral quadratic constraints (IQC) and show that Alt-GDA attains the same rate \\emph{globally} for a subclass of SCSC minimax problems. Empirically, we demonstrate that alternating updates speed up GAN training significantly and the use of optimism only helps for simultaneous algorithms."
                },
                {
                    "title": "Variance-Reduced Methods for Machine Learning",
                    "abstract": "Stochastic optimization lies at the heart of machine learning, and its cornerstone is stochastic gradient descent (<monospace>SGD</monospace>), a method introduced over 60 years ago. The last eight years have seen an exciting new development: variance reduction for stochastic optimization methods. These variance-reduced (<monospace>VR</monospace>) methods excel in settings where more than one pass through the training data is allowed, achieving a faster convergence than <monospace>SGD</monospace> in theory and practice. These speedups underline the surge of interest in <monospace>VR</monospace> methods and the fast-growing body of work on this topic. This review covers the key principles and main developments behind <monospace>VR</monospace> methods for optimization with finite data sets and is aimed at nonexpert readers. We focus mainly on the convex setting and leave pointers to readers interested in extensions for minimizing nonconvex functions."
                },
                {
                    "title": "Responsive Safety in Reinforcement Learning by PID Lagrangian Methods",
                    "abstract": "Lagrangian methods are widely used algorithms for constrained optimization problems, but their learning dynamics exhibit oscillations and overshoot which, when applied to safe reinforcement learning, leads to constraint-violating behavior during agent training. We address this shortcoming by proposing a novel Lagrange multiplier update method that utilizes derivatives of the constraint function. We take a controls perspective, wherein the traditional Lagrange multiplier update behaves as \\emph{integral} control; our terms introduce \\emph{proportional} and \\emph{derivative} control, achieving favorable learning dynamics through damping and predictive measures. We apply our PID Lagrangian methods in deep RL, setting a new state of the art in Safety Gym, a safe RL benchmark. Lastly, we introduce a new method to ease controller tuning by providing invariance to the relative numerical scales of reward and cost. Our extensive experiments demonstrate improved performance and hyperparameter robustness, while our algorithms remain nearly as simple to derive and implement as the traditional Lagrangian approach."
                },
                {
                    "title": "Probably Approximately Correct Constrained Learning",
                    "abstract": "As learning solutions reach critical applications in social, industrial, and medical domains, the need to curtail their behavior becomes paramount. There is now ample evidence that without explicit tailoring, learning can lead to biased, unsafe, and prejudiced solutions. To tackle these problems, we develop a generalization theory of constrained learning based on the probably approximately correct (PAC) learning framework. In particular, we show that imposing requirements does not make a learning problem harder in the sense that any PAC learnable class is also PAC constrained learnable using a constrained counterpart of the empirical risk minimization (ERM) rule. For typical parametrized models, however, this learner involves solving a non-convex optimization program for which even obtaining a feasible solution may be hard. To overcome this issue, we prove that under mild conditions the empirical dual problem of constrained learning is also a PAC constrained learner that now leads to a practical constrained learning algorithm. We analyze the generalization properties of this solution and use it to illustrate how constrained learning can address problems in fair and robust classification."
                },
                {
                    "title": "What is the State of Neural Network Pruning?",
                    "abstract": "Neural network pruning---the task of reducing the size of a network by removing parameters---has been the subject of a great deal of work in recent years. We provide a meta-analysis of the literature, including an overview of approaches to pruning and consistent findings in the literature. After aggregating results across 81 papers and pruning hundreds of models in controlled conditions, our clearest finding is that the community suffers from a lack of standardized benchmarks and metrics. This deficiency is substantial enough that it is hard to compare pruning techniques to one another or determine how much progress the field has made over the past three decades. To address this situation, we identify issues with current practices, suggest concrete remedies, and introduce ShrinkBench, an open-source framework to facilitate standardized evaluations of pruning methods. We use ShrinkBench to compare various pruning techniques and show that its comprehensive evaluation can prevent common pitfalls when comparing pruning methods."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Rigging the Lottery: Making All Tickets Winners",
                    "abstract": "Many applications require sparse neural networks due to space or inference time restrictions. There is a large body of work on training dense networks to yield sparse networks for inference, but this limits the size of the largest trainable sparse model to that of the largest trainable dense model. In this paper we introduce a method to train sparse neural networks with a fixed parameter count and a fixed computational cost throughout training, without sacrificing accuracy relative to existing dense-to-sparse training methods. Our method updates the topology of the sparse network during training by using parameter magnitudes and infrequent gradient calculations. We show that this approach requires fewer floating-point operations (FLOPs) to achieve a given level of accuracy compared to prior techniques. We demonstrate state-of-the-art sparse training results on a variety of networks and datasets, including ResNet-50, MobileNets on Imagenet-2012, and RNNs on WikiText-103. Finally, we provide some insights into why allowing the topology to change during the optimization can overcome local minima encountered when the topology remains static. Code used in our work can be found in this http URL."
                },
                {
                    "title": "What Do Compressed Deep Neural Networks Forget",
                    "abstract": "Deep neural network pruning and quantization techniques have demonstrated it is possible to achieve high levels of compression with surprisingly little degradation to test set accuracy. However, this measure of performance conceals significant differences in how different classes and images are impacted by model compression techniques. We find that models with radically different numbers of weights have comparable top-line performance metrics but diverge considerably in behavior on a narrow subset of the dataset. This small subset of data points, which we term Pruning Identified Exemplars (PIEs) are systematically more impacted by the introduction of sparsity. Compression disproportionately impacts model performance on the underrepresented long-tail of the data distribution. PIEs over-index on atypical or noisy images that are far more challenging for both humans and algorithms to classify. Our work provides intuition into the role of capacity in deep neural networks and the trade-offs incurred by compression. An understanding of this disparate impact is critical given the widespread deployment of compressed models in the wild."
                },
                {
                    "title": "On Empirical Comparisons of Optimizers for Deep Learning",
                    "abstract": "Selecting an optimizer is a central step in the contemporary deep learning pipeline. In this paper, we demonstrate the sensitivity of optimizer comparisons to the hyperparameter tuning protocol. Our findings suggest that the hyperparameter search space may be the single most important factor explaining the rankings obtained by recent empirical comparisons in the literature. In fact, we show that these results can be contradicted when hyperparameter search spaces are changed. As tuning effort grows without bound, more general optimizers should never underperform the ones they can approximate (i.e., Adam should never perform worse than momentum), but recent attempts to compare optimizers either assume these inclusion relationships are not practically relevant or restrict the hyperparameters in ways that break the inclusions. In our experiments, we find that inclusion relationships between optimizers matter in practice and always predict optimizer comparisons. In particular, we find that the popular adaptive gradient methods never underperform momentum or gradient descent. We also report practical tips around tuning often ignored hyperparameters of adaptive gradient methods and raise concerns about fairly benchmarking optimizers for neural network training."
                },
                {
                    "title": "A Survey on Bias and Fairness in Machine Learning",
                    "abstract": "With the widespread use of artificial intelligence (AI) systems and applications in our everyday lives, accounting for fairness has gained significant importance in designing and engineering of such systems. AI systems can be used in many sensitive environments to make important and life-changing decisions; thus, it is crucial to ensure that these decisions do not reflect discriminatory behavior toward certain groups or populations. More recently some work has been developed in traditional machine learning and deep learning that address such challenges in different subdomains. With the commercialization of these systems, researchers are becoming more aware of the biases that these applications can contain and are attempting to address them. In this survey, we investigated different real-world applications that have shown biases in various ways, and we listed different sources of biases that can affect AI applications. We then created a taxonomy for fairness definitions that machine learning researchers have defined to avoid the existing bias in AI systems. In addition to that, we examined different domains and subdomains in AI showing what researchers have observed with regard to unfair outcomes in the state-of-the-art methods and ways they have tried to address them. There are still many future directions and solutions that can be taken to mitigate the problem of bias in AI systems. We are hoping that this survey will motivate researchers to tackle these issues in the near future by observing existing work in their respective fields."
                },
                {
                    "title": "FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age",
                    "abstract": "Existing public face datasets are strongly biased toward Caucasian faces, and other races (e.g., Latino) are significantly underrepresented. This can lead to inconsistent model accuracy, limit the applicability of face analytic systems to non-White race groups, and adversely affect research findings based on such skewed data. To mitigate the race bias in these datasets, we construct a novel face image dataset, containing 108,501 images, with an emphasis of balanced race composition in the dataset. We define 7 race groups: White, Black, Indian, East Asian, Southeast Asian, Middle East, and Latino. Images were collected from the YFCC-100M Flickr dataset and labeled with race, gender, and age groups. Evaluations were performed on existing face attribute datasets as well as novel image datasets to measure generalization performance. We find that the model trained from our dataset is substantially more accurate on novel datasets and the accuracy is consistent between race and gender groups."
                },
                {
                    "title": "Inherent Tradeoffs in Learning Fair Representation",
                    "abstract": "With the prevalence of machine learning in high-stakes applications, especially the ones regulated by anti-discrimination laws or societal norms, it is crucial to ensure that the predictive models do not propagate any existing bias or discrimination. Due to the ability of deep neural nets to learn rich representations, recent advances in algorithmic fairness have focused on learning fair representations with adversarial techniques to reduce bias in data while preserving utility simultaneously. In this paper, through the lens of information theory, we provide the first result that quantitatively characterizes the tradeoff between demographic parity and the joint utility across different population groups. Specifically, when the base rates differ between groups, we show that any method aiming to learn fair representations admits an information-theoretic lower bound on the joint error across these groups. To complement our negative results, we also prove that if the optimal decision functions across different groups are close, then learning fair representations leads to an alternative notion of fairness, known as the accuracy parity, which states that the error rates are close between groups. Finally, our theoretical findings are also confirmed empirically on real-world datasets."
                },
                {
                    "title": "On Gradient Descent Ascent for Nonconvex-Concave Minimax Problems",
                    "abstract": "We consider nonconvex-concave minimax problems, $\\min_{\\mathbf{x}} \\max_{\\mathbf{y} \\in \\mathcal{Y}} f(\\mathbf{x}, \\mathbf{y})$, where $f$ is nonconvex in $\\mathbf{x}$ but concave in $\\mathbf{y}$ and $\\mathcal{Y}$ is a convex and bounded set. One of the most popular algorithms for solving this problem is the celebrated gradient descent ascent (GDA) algorithm, which has been widely used in machine learning, control theory and economics. Despite the extensive convergence results for the convex-concave setting, GDA with equal stepsize can converge to limit cycles or even diverge in a general setting. In this paper, we present the complexity results on two-time-scale GDA for solving nonconvex-concave minimax problems, showing that the algorithm can find a stationary point of the function $\\Phi(\\cdot) := \\max_{\\mathbf{y} \\in \\mathcal{Y}} f(\\cdot, \\mathbf{y})$ efficiently. To the best our knowledge, this is the first nonasymptotic analysis for two-time-scale GDA in this setting, shedding light on its superior practical performance in training generative adversarial networks (GANs) and other real applications."
                },
                {
                    "title": "Reducing Noise in GAN Training with Variance Reduced Extragradient",
                    "abstract": "We study the effect of the stochastic gradient noise on the training of generative adversarial networks (GANs) and show that it can prevent the convergence of standard game optimization methods, while the batch version converges. We address this issue with a novel stochastic variance-reduced extragradient (SVRE) optimization algorithm, which for a large class of games improves upon the previous convergence rates proposed in the literature. We observe empirically that SVRE performs similarly to a batch method on MNIST while being computationally cheaper, and that SVRE yields more stable GAN training on standard datasets."
                },
                {
                    "title": "The State of Sparsity in Deep Neural Networks",
                    "abstract": "We rigorously evaluate three state-of-the-art techniques for inducing sparsity in deep neural networks on two large-scale learning tasks: Transformer trained on WMT 2014 English-to-German, and ResNet-50 trained on ImageNet. Across thousands of experiments, we demonstrate that complex techniques (Molchanov et al., 2017; Louizos et al., 2017b) shown to yield high compression rates on smaller datasets perform inconsistently, and that simple magnitude pruning approaches achieve comparable or better results. Additionally, we replicate the experiments performed by (Frankle & Carbin, 2018) and (Liu et al., 2018) at scale and show that unstructured sparse architectures learned through pruning cannot be trained from scratch to the same test set performance as a model trained with joint sparsification and optimization. Together, these results highlight the need for large-scale benchmarks in the field of model compression. We open-source our code, top performing model checkpoints, and results of all hyperparameter configurations to establish rigorous baselines for future work on compression and sparsification."
                },
                {
                    "title": "Optimization with Non-Differentiable Constraints with Applications to Fairness, Recall, Churn, and Other Goals",
                    "abstract": "We show that many machine learning goals, such as improved fairness metrics, can be expressed as constraints on the model's predictions, which we call rate constraints. We study the problem of training non-convex models subject to these rate constraints (or any non-convex and non-differentiable constraints). In the non-convex setting, the standard approach of Lagrange multipliers may fail. Furthermore, if the constraints are non-differentiable, then one cannot optimize the Lagrangian with gradient-based methods. To solve these issues, we introduce the proxy-Lagrangian formulation. This new formulation leads to an algorithm that produces a stochastic classifier by playing a two-player non-zero-sum game solving for what we call a semi-coarse correlated equilibrium, which in turn corresponds to an approximately optimal and feasible solution to the constrained optimization problem. We then give a procedure which shrinks the randomized solution down to one that is a mixture of at most $m+1$ deterministic solutions, given $m$ constraints. This culminates in algorithms that can solve non-convex constrained optimization problems with possibly non-differentiable and non-convex constraints with theoretical guarantees. We provide extensive experimental results enforcing a wide range of policy goals including different fairness metrics, and other goals on accuracy, coverage, recall, and churn."
                },
                {
                    "title": "Fairness Definitions Explained",
                    "abstract": "Algorithm fairness has started to attract the attention of researchers in AI, Software Engineering and Law communities, with more than twenty different notions of fairness proposed in the last few years. Yet, there is no clear agreement on which definition to apply in each situation. Moreover, the detailed differences between multiple definitions are difficult to grasp. To address this issue, this paper collects the most prominent definitions of fairness for the algorithmic classification problem, explains the rationale behind these definitions, and demonstrates each of them on a single unifying case-study. Our analysis intuitively explains why the same case can be considered fair according to some definitions and unfair according to others."
                },
                {
                    "title": "A Reductions Approach to Fair Classification",
                    "abstract": "We present a systematic approach for achieving fairness in a binary classification setting. While we focus on two well-known quantitative definitions of fairness, our approach encompasses many other previously studied definitions as special cases. The key idea is to reduce fair classification to a sequence of cost-sensitive classification problems, whose solutions yield a randomized classifier with the lowest (empirical) error subject to the desired constraints. We introduce two reductions that work for any representation of the cost-sensitive classifier and compare favorably to prior baselines on a variety of data sets, while overcoming several of their disadvantages."
                },
                {
                    "title": "A Variational Inequality Perspective on Generative Adversarial Networks",
                    "abstract": "Generative adversarial networks (GANs) form a generative modeling approach known for producing appealing samples, but they are notably difficult to train. One common way to tackle this issue has been to propose new formulations of the GAN objective. Yet, surprisingly few studies have looked at optimization methods designed for this adversarial training. In this work, we cast GAN optimization problems in the general variational inequality framework. Tapping into the mathematical programming literature, we counter some common misconceptions about the difficulties of saddle point optimization and propose to extend techniques designed for variational inequalities to the training of GANs. We apply averaging, extrapolation and a computationally cheaper variant that we call extrapolation from the past to the stochastic gradient method (SGD) and Adam."
                },
                {
                    "title": "MobileNetV2: Inverted Residuals and Linear Bottlenecks",
                    "abstract": "In this paper we describe a new mobile architecture, MobileNetV2, that improves the state of the art performance of mobile models on multiple tasks and benchmarks as well as across a spectrum of different model sizes. We also describe efficient ways of applying these mobile models to object detection in a novel framework we call SSDLite. Additionally, we demonstrate how to build mobile semantic segmentation models through a reduced form of DeepLabv3 which we call Mobile DeepLabv3. is based on an inverted residual structure where the shortcut connections are between the thin bottleneck layers. The intermediate expansion layer uses lightweight depthwise convolutions to filter features as a source of non-linearity. Additionally, we find that it is important to remove non-linearities in the narrow layers in order to maintain representational power. We demonstrate that this improves performance and provide an intuition that led to this design. Finally, our approach allows decoupling of the input/output domains from the expressiveness of the transformation, which provides a convenient framework for further analysis. We measure our performance on ImageNet [1] classification, COCO object detection [2], VOC image segmentation [3]. We evaluate the trade-offs between accuracy, and number of operations measured by multiply-adds (MAdd), as well as actual latency, and the number of parameters."
                },
                {
                    "title": "Towards Deep Learning Models Resistant to Adversarial Attacks",
                    "abstract": "Recent work has demonstrated that deep neural networks are vulnerable to adversarial examples---inputs that are almost indistinguishable from natural data and yet classified incorrectly by the network. In fact, some of the latest findings suggest that the existence of adversarial attacks may be an inherent weakness of deep learning models. To address this problem, we study the adversarial robustness of neural networks through the lens of robust optimization. This approach provides us with a broad and unifying view on much of the prior work on this topic. Its principled nature also enables us to identify methods for both training and attacking neural networks that are reliable and, in a certain sense, universal. In particular, they specify a concrete security guarantee that would protect against any adversary. These methods let us train networks with significantly improved resistance to a wide range of adversarial attacks. They also suggest the notion of security against a first-order adversary as a natural and broad security guarantee. We believe that robustness against such well-defined classes of adversaries is an important stepping stone towards fully resistant deep learning models. Code and pre-trained models are available at this https URL and this https URL."
                },
                {
                    "title": "From Parity to Preference-based Notions of Fairness in Classification",
                    "abstract": "The adoption of automated, data-driven decision making in an ever expanding range of applications has raised concerns about its potential unfairness towards certain social groups. In this context, a number of recent studies have focused on defining, detecting, and removing unfairness from data-driven decision systems. However, the existing notions of fairness, based on parity (equality) in treatment or outcomes for different social groups, tend to be quite stringent, limiting the overall decision making accuracy. In this paper, we draw inspiration from the fair-division and envy-freeness literature in economics and game theory and propose preference-based notions of fairness -- given the choice between various sets of decision treatments or outcomes, any group of users would collectively prefer its treatment or outcomes, regardless of the (dis)parity as compared to the other groups. Then, we introduce tractable proxies to design margin-based classifiers that satisfy these preference-based notions of fairness. Finally, we experiment with a variety of synthetic and real-world datasets and show that preference-based fairness allows for greater decision accuracy than parity-based fairness."
                },
                {
                    "title": "Fairness in Criminal Justice Risk Assessments: The State of the Art",
                    "abstract": "Objectives: Discussions of fairness in criminal justice risk assessments typically lack conceptual precision. Rhetoric too often substitutes for careful analysis. In this article, we seek to clarify the trade-offs between different kinds of fairness and between fairness and accuracy. Methods: We draw on the existing literatures in criminology, computer science, and statistics to provide an integrated examination of fairness and accuracy in criminal justice risk assessments. We also provide an empirical illustration using data from arraignments. Results: We show that there are at least six kinds of fairness, some of which are incompatible with one another and with accuracy. Conclusions: Except in trivial cases, it is impossible to maximize accuracy and fairness at the same time and impossible simultaneously to satisfy all kinds of fairness. In practice, a major complication is different base rates across different legally protected groups. There is a need to consider challenging trade-offs. These lessons apply to applications well beyond criminology where assessments of risk can be used by decision makers. Examples include mortgage lending, employment, college admissions, child welfare, and medical diagnoses."
                },
                {
                    "title": "Age Progression/Regression by Conditional Adversarial Autoencoder",
                    "abstract": "If I provide you a face image of mine (without telling you the actual age when I took the picture) and a large amount of face images that I crawled (containing labeled faces of different ages but not necessarily paired), can you show me what I would look like when I am 80 or what I was like when I was 5? The answer is probably a No. Most existing face aging works attempt to learn the transformation between age groups and thus would require the paired samples as well as the labeled query image. In this paper, we look at the problem from a generative modeling perspective such that no paired samples is required. In addition, given an unlabeled image, the generative model can directly produce the image with desired age attribute. We propose a conditional adversarial autoencoder (CAAE) that learns a face manifold, traversing on which smooth age progression and regression can be realized simultaneously. In CAAE, the face is first mapped to a latent vector through a convolutional encoder, and then the vector is projected to the face manifold conditional on age through a deconvolutional generator. The latent vector preserves personalized face features (i.e., personality) and the age condition controls progression vs. regression. Two adversarial networks are imposed on the encoder and generator, respectively, forcing to generate more photo-realistic faces. Experimental results demonstrate the appealing performance and flexibility of the proposed framework by comparing with the state-of-the-art and ground truth."
                },
                {
                    "title": "Equality of Opportunity in Supervised Learning",
                    "abstract": "We propose a criterion for discrimination against a specified sensitive attribute in supervised learning, where the goal is to predict some target based on available features. Assuming data about the predictor, target, and membership in the protected group are available, we show how to optimally adjust any learned predictor so as to remove discrimination according to our definition. Our framework also improves incentives by shifting the cost of poor classification from disadvantaged groups to the decision maker, who can respond by improving the classification accuracy. We enourage readers to consult the more complete manuscript on the arXiv."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Distilling the Knowledge in a Neural Network",
                    "abstract": "A very simple way to improve the performance of almost any machine learning algorithm is to train many different models on the same data and then to average their predictions. Unfortunately, making predictions using a whole ensemble of models is cumbersome and may be too computationally expensive to allow deployment to a large number of users, especially if the individual models are large neural nets. Caruana and his collaborators have shown that it is possible to compress the knowledge in an ensemble into a single model which is much easier to deploy and we develop this approach further using a different compression technique. We achieve some surprising results on MNIST and we show that we can significantly improve the acoustic model of a heavily used commercial system by distilling the knowledge in an ensemble of models into a single model. We also introduce a new type of ensemble composed of one or more full models and many specialist models which learn to distinguish fine-grained classes that the full models confuse. Unlike a mixture of experts, these specialist models can be trained rapidly and in parallel."
                },
                {
                    "title": "SAGA: A Fast Incremental Gradient Method With Support for Non-Strongly Convex Composite Objectives",
                    "abstract": "In this work we introduce a new optimisation method called SAGA in the spirit of SAG, SDCA, MISO and SVRG, a set of recently proposed incremental gradient algorithms with fast linear convergence rates. SAGA improves on the theory behind SAG and SVRG, with better theoretical convergence rates, and has support for composite objectives where a proximal operator is used on the regulariser. Unlike SDCA, SAGA supports non-strongly convex problems directly, and is adaptive to any inherent strong convexity of the problem. We give experimental results showing the effectiveness of our method."
                },
                {
                    "title": "Generative adversarial networks",
                    "abstract": "Generative adversarial networks are a kind of artificial intelligence algorithm designed to solve the generative modeling problem. The goal of a generative model is to study a collection of training examples and learn the probability distribution that generated them. Generative Adversarial Networks (GANs) are then able to generate more examples from the estimated probability distribution. Generative models based on deep learning are common, but GANs are among the most successful generative models (especially in terms of their ability to generate realistic high-resolution images). GANs have been successfully applied to a wide variety of tasks (mostly in research settings) but continue to present unique challenges and research opportunities because they are based on game theory while most other approaches to generative modeling are based on optimization."
                },
                {
                    "title": "Playing Atari with Deep Reinforcement Learning",
                    "abstract": "We present the first deep learning model to successfully learn control policies directly from high-dimensional sensory input using reinforcement learning. The model is a convolutional neural network, trained with a variant of Q-learning, whose input is raw pixels and whose output is a value function estimating future rewards. We apply our method to seven Atari 2600 games from the Arcade Learning Environment, with no adjustment of the architecture or learning algorithm. We find that it outperforms all previous approaches on six of the games and surpasses a human expert on three of them."
                },
                {
                    "title": "Learning Fair Representations",
                    "abstract": "We propose a learning algorithm for fair classification that achieves both group fairness (the proportion of members in a protected group receiving positive classification is identical to the proportion in the population as a whole), and individual fairness (similar individuals should be treated similarly). We formulate fairness as an optimization problem of finding a good representation of the data with two competing goals: to encode the data as well as possible, while simultaneously obfuscating any information about membership in the protected group. We show positive results of our algorithm relative to other known techniques, on three datasets. Moreover, we demonstrate several advantages to our approach. First, our intermediate representation can be used for other classification tasks (i.e., transfer learning is possible); secondly, we take a step toward learning a distance metric which can find important dimensions of the data for classification."
                },
                {
                    "title": "Fairness through awareness",
                    "abstract": "We study fairness in classification, where individuals are classified, e.g., admitted to a university, and the goal is to prevent discrimination against individuals based on their membership in some group, while maintaining utility for the classifier (the university). The main conceptual contribution of this paper is a framework for fair classification comprising (1) a (hypothetical) task-specific metric for determining the degree to which individuals are similar with respect to the classification task at hand; (2) an algorithm for maximizing utility subject to the fairness constraint, that similar individuals are treated similarly. We also present an adaptation of our approach to achieve the complementary goal of \"fair affirmative action,\" which guarantees statistical parity (i.e., the demographics of the set of individuals receiving any classification are the same as the demographics of the underlying population), while treating similar individuals as similarly as possible. Finally, we discuss the relationship of fairness to privacy: when fairness implies privacy, and how tools developed in the context of differential privacy may be applied to fairness."
                },
                {
                    "title": "Fair Enough: Improving Fairness in Budget-Constrained Decision Making Using Confidence Thresholds",
                    "abstract": ","
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.CY"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively mitigate the disparate impact of pruning in machine learning models to ensure equitable performance across under-represented groups?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing fairness in machine learning, particularly as models become larger and more widely deployed. Addressing the disparate impact of pruning can lead to more equitable AI systems, reducing systemic biases that affect marginalized groups. This research could influence future studies on model compression and fairness, encouraging the development of techniques that prioritize group-level performance alongside overall accuracy. Ultimately, it could lead to practical applications in sensitive areas such as healthcare, finance, and criminal justice, where biased models can have significant real-world consequences.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities of balancing model performance across multiple sub-groups while maintaining overall accuracy. Naive approaches, such as simple fine-tuning, often fail because they do not specifically address the group-level performance disparities that arise from pruning. Additionally, existing methods may lack interpretability or scalability, making it difficult to apply them to large models or numerous sub-groups. The technical obstacles include formulating an optimization problem that directly targets excess accuracy gaps without incurring significant computational costs.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on aggregate model performance, neglecting the nuanced impacts of pruning on different sub-groups. Existing solutions, such as equalized loss constraints, do not adequately account for the initial performance disparities of dense models, leading to indirect and less effective mitigation strategies. Moreover, methods that compute per-group importance scores become computationally prohibitive as model size and the number of sub-groups increase. Our approach differs by directly addressing the group-level accuracy gaps through a novel formulation that is both interpretable and computationally efficient.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves formulating an optimization problem that imposes constraints on the per-group excess accuracy gaps (CEAG) during the pruning process. We will utilize a dataset such as UTKFace, focusing on race as a group attribute, and measure performance using excess accuracy gaps (EAGs) as our primary metric. The expected outcome is a model that maintains overall accuracy while significantly reducing the disparity in performance across different groups, demonstrating the effectiveness of our approach compared to naive fine-t"
            }
        },
        "author_data": {
            "759a1cd3-2f32-4428-9153-0656af61b1f0": {
                "pk": "759a1cd3-2f32-4428-9153-0656af61b1f0",
                "name": "Meraj Hashemizadeh",
                "collaborators": [
                    "M. Liu",
                    "Jacob Miller",
                    "Guillaume Rabusseau"
                ],
                "domain": [
                    "Tensor Networks",
                    "Machine Learning",
                    "Tensor Decomposition",
                    "Data Science"
                ],
                "publications": [
                    {
                        "title": "Adaptive Learning of Tensor Network Structures",
                        "abstract": "Tensor Networks (TN) offer a powerful framework to efficiently represent very high-dimensional objects. TN have recently shown their potential for machine learning applications and offer a unifying view of common tensor decomposition models such as Tucker, tensor train (TT) and tensor ring (TR). However, identifying the best tensor network structure from data for a given task is challenging. In this work, we leverage the TN formalism to develop a generic and efficient adaptive algorithm to jointly learn the structure and the parameters of a TN from data. Our method is based on a simple greedy approach starting from a rank one tensor and successively identifying the most promising tensor network edges for small rank increments. Our algorithm can adaptively identify TN structures with small number of parameters that effectively optimize any differentiable objective function. Experiments on tensor decomposition, tensor completion and model compression tasks demonstrate the effectiveness of the proposed algorithm. In particular, our method outperforms the state-of-the-art evolutionary topology search [Li and Sun, 2020] for tensor decomposition of images (while being orders of magnitude faster) and finds efficient tensor network structures to compress neural networks outperforming popular TT based approaches [Novikov et al., 2015]."
                    },
                    {
                        "title": "Adaptive Tensor Learning with Tensor Networks",
                        "abstract": "Tensor decomposition techniques have shown great successes in machine learning and data science by extending classical algorithms based on matrix factorization to multi-modal and multi-way data. However, there exist many tensor decomposition models~(CP, Tucker, Tensor Train, etc.) and the rank of such a decomposition is typically a collection of integers rather than a unique number, making model and hyper-parameter selection a tedious and costly task. At the same time, tensor network methods are powerful tools developed in the physics community which have recently shown their potential for machine learning applications and offer a unifying view of the various tensor decomposition models. In this paper, we leverage the tensor network formalism to develop a generic and efficient adaptive algorithm for tensor learning. Our method is based on a simple greedy approach optimizing a differentiable loss function starting from a rank one tensor and successively identifying the most promising tensor network edges for small rank increments. Our algorithm can adaptively identify tensor network structures with small number of parameters that effectively optimize the objective function from data. The framework we introduce is very broad and encompasses many common tensor optimization problems. Experiments on tensor decomposition and tensor completion tasks with both synthetic and real world data demonstrate the effectiveness of the proposed algorithm."
                    }
                ]
            },
            "24f0f564-7474-4385-9409-b1b418cdfa59": {
                "pk": "24f0f564-7474-4385-9409-b1b418cdfa59",
                "name": "Juan Ramirez",
                "collaborators": [
                    "Jose Gallego-Posada",
                    "J. S. Rogers",
                    "Rohan Sukumaran",
                    "Quentin Bertrand",
                    "Gauthier Gidel",
                    "Akram Erraqabi",
                    "Y. Bengio",
                    "Simon Lacoste-Julien",
                    "G. Edelmann"
                ],
                "domain": [
                    "Stochastic Optimization",
                    "Neural Networks",
                    "Sparse Learning",
                    "Acoustic Imaging"
                ],
                "publications": [
                    {
                        "title": "Omega: Optimistic EMA Gradients",
                        "abstract": "Stochastic min-max optimization has gained interest in the machine learning community with the advancements in GANs and adversarial training. Although game optimization is fairly well understood in the deterministic setting, some issues persist in the stochastic regime. Recent work has shown that stochastic gradient descent-ascent methods such as the optimistic gradient are highly sensitive to noise or can fail to converge. Although alternative strategies exist, they can be prohibitively expensive. We introduce Omega, a method with optimistic-like updates that mitigates the impact of noise by incorporating an EMA of historic gradients in its update rule. We also explore a variation of this algorithm that incorporates momentum. Although we do not provide convergence guarantees, our experiments on stochastic games show that Omega outperforms the optimistic gradient method when applied to linear players."
                    },
                    {
                        "title": "L$_0$onie: Compressing COINs with L$_0$-constraints",
                        "abstract": "Advances in Implicit Neural Representations (INR) have motivated research on domain-agnostic compression techniques. These methods train a neural network to approximate an object, and then store the weights of the trained model. For example, given an image, a network is trained to learn the mapping from pixel locations to RGB values. In this paper, we propose L 0 onie, a sparsity-constrained extension of the COIN compression method. Sparsity allows to leverage the faster learning of overparameterized networks, while retaining the desirable compression rate of smaller models. Moreover, our constrained formulation ensures that the \ufb01nal model respects a pre-determined compression rate, dispensing of the need for expensive architecture search."
                    },
                    {
                        "title": "Controlled Sparsity via Constrained Optimization or: How I Learned to Stop Tuning Penalties and Love Constraints",
                        "abstract": "The performance of trained neural networks is robust to harsh levels of pruning. Coupled with the ever-growing size of deep learning models, this observation has motivated extensive research on learning sparse models. In this work, we focus on the task of controlling the level of sparsity when performing sparse learning. Existing methods based on sparsity-inducing penalties involve expensive trial-and-error tuning of the penalty factor, thus lacking direct control of the resulting model sparsity. In response, we adopt a constrained formulation: using the gate mechanism proposed by Louizos et al. (2018), we formulate a constrained optimization problem where sparsification is guided by the training objective and the desired sparsity target in an end-to-end fashion. Experiments on CIFAR-{10, 100}, TinyImageNet, and ImageNet using WideResNet and ResNet{18, 50} models validate the effectiveness of our proposal and demonstrate that we can reliably achieve pre-determined sparsity targets without compromising on predictive performance."
                    },
                    {
                        "title": "A performance metric for screen selection with the acoustic single pixel imager.",
                        "abstract": "In the recent literature, an Acoustic Single-Pixel Imager has been successfully developed for source localization in a two-dimensional waveguide. Source bearing angle estimation was carried out by applying sparse recovery techniques on sensor measurements taken over different imaging screens. This paper shows that the Mutual Coherence of the sensing matrix can be used as a metric to predict the source localization capability of the single-pixel imaging system. In particular, this paper's analysis focuses on the sparsity of open cells within the imaging screen and the number of imaging screens used to maximize the probability of correct detection over varying levels of source sparsity. In this work, a simulation environment to demonstrate how the mutual coherence of the sensing matrix correlates with source localization performance over source sparsity, sparsity of open screen cells, and number of measurements used for sparse recovery is developed. The analysis shows that the leading factor in source localization performance gains is primarily from the number of imaging screens used to measure the acoustic wave-field."
                    },
                    {
                        "title": "Exploiting platform motion for passive source localization with a co-prime sampled large aperture array.",
                        "abstract": "Co-prime array geometries have received a great deal of attention due to their ability to discriminate O(MN) sources with only O(M\u2009+\u2009N) sensors. This has been demonstrated both theoretically and in simulation. However, there are many practical limitations that make it difficult to realize the enhanced degrees of freedom when applying co-prime geometries to real acoustic data taken on a horizontal line array. For instance, co-prime sampling leads to grating lobes that can obscure lower signal-to-noise-ratio acoustic signals making them difficult to detect. In this work, a synthetic aperture (SA) method is presented for filling in holes and increasing redundancy in the difference co-array by exploiting array motion. The SA method is applied to acoustic data collected off the Southeastern shore of Florida on a fixed large aperture horizontal array. Array motion is simulated by taking a co-prime sampled subarray and virtually moving it along the horizontal aperture of the fixed array. It is demonstrated that SA processing on real acoustic data results in reduced side-lobe and grating lobe levels compared to that of the physical co-prime aperture."
                    }
                ]
            },
            "22e57175-84d6-4547-89b4-6787c2d51e78": {
                "pk": "22e57175-84d6-4547-89b4-6787c2d51e78",
                "name": "Rohan Sukumaran",
                "collaborators": [
                    "Sheshank Shankar",
                    "Parth Patwa",
                    "Abhishek Singh",
                    "R. Raskar",
                    "Rohan Iyer",
                    "Priyanshi Katiyar",
                    "Joseph Bae",
                    "Saurish Srivastava",
                    "Aryan Mahindra",
                    "Anshuman Sharma"
                ],
                "domain": [
                    "Machine Learning",
                    "Natural Language Processing",
                    "Public Health",
                    "Fairness in AI"
                ],
                "publications": [
                    {
                        "title": "FairLoRA: Unpacking Bias Mitigation in Vision Models with Fairness-Driven Low-Rank Adaptation",
                        "abstract": "Recent advances in parameter-efficient fine-tuning methods, such as Low Rank Adaptation (LoRA), have gained significant attention for their ability to efficiently adapt large foundational models to various downstream tasks. These methods are appreciated for achieving performance comparable to full fine-tuning on aggregate-level metrics, while significantly reducing computational costs. To systematically address fairness in LLMs previous studies fine-tune on fairness specific data using a larger LoRA rank than typically used. In this paper, we introduce FairLoRA, a novel fairness-specific regularizer for LoRA aimed at reducing performance disparities across data subgroups by minimizing per-class variance in loss. To the best of our knowledge, we are the first to introduce a fairness based finetuning through LoRA. Our results demonstrate that the need for higher ranks to mitigate bias is not universal; it depends on factors such as the pre-trained model, dataset, and task. More importantly, we systematically evaluate FairLoRA across various vision models, including ViT, DiNO, and CLIP, in scenarios involving distribution shifts. We further emphasize the necessity of using multiple fairness metrics to obtain a holistic assessment of fairness, rather than relying solely on the metric optimized during training."
                    },
                    {
                        "title": "Omega: Optimistic EMA Gradients",
                        "abstract": "Stochastic min-max optimization has gained interest in the machine learning community with the advancements in GANs and adversarial training. Although game optimization is fairly well understood in the deterministic setting, some issues persist in the stochastic regime. Recent work has shown that stochastic gradient descent-ascent methods such as the optimistic gradient are highly sensitive to noise or can fail to converge. Although alternative strategies exist, they can be prohibitively expensive. We introduce Omega, a method with optimistic-like updates that mitigates the impact of noise by incorporating an EMA of historic gradients in its update rule. We also explore a variation of this algorithm that incorporates momentum. Although we do not provide convergence guarantees, our experiments on stochastic games show that Omega outperforms the optimistic gradient method when applied to linear players."
                    },
                    {
                        "title": "Enhanced Text Classification using Proxy Labels and Knowledge Distillation",
                        "abstract": "Text Classification has a variety of applications in the pickup and delivery services industry where customers require one or more items to be picked up from a location and delivered to a certain destination. Categorizing these customer transactions helps understand the market needs and trends while also assisting in building a personalized experience for each customer segment. In this paper, each transaction is accompanied by a free text description provided by the customer to describe the products to be picked up and delivered. These descriptions tend to be short, incoherent and code-mixed (Hindi-English) text.Here, we focus on a specific use-case where each customer transaction can be mapped to a single product category. We propose a cost-effective transaction classification approach based on proxy-labelling and knowledge distillation using the transaction descriptions provided by the customer. We introduce R-ALBERT, a model trained with RoBERTa as the \u201cteacher\u201d and ALBERT (33x fewer parameters than RoBERTa) as the \u201cstudent\u201d. Further, we benchmark R-ALBERT on a large internal dataset as well as the 20Newsgroup dataset. We see that our model shows a 2% increase in performance with 33x fewer parameters. The model is currently deployed in production and is helping understand the customer behaviour across product categories and customer segments."
                    },
                    {
                        "title": "Can Self Reported Symptoms Predict Daily COVID-19 Cases?",
                        "abstract": "The COVID-19 pandemic has impacted lives and economies across the globe, leading to many deaths. While vaccination is an important intervention, its roll-out is slow and unequal across the globe. Therefore, extensive testing still remains one of the key methods to monitor and contain the virus. Testing on a large scale is expensive and arduous. Hence, we need alternate methods to estimate the number of cases. Online surveys have been shown to be an effective method for data collection amidst the pandemic. In this work, we develop machine learning models to estimate the prevalence of COVID-19 using self-reported symptoms. Our best model predicts the daily cases with a mean absolute error (MAE) of 226.30 (normalized MAE of 27.09%) per state, which demonstrates the possibility of predicting the actual number of confirmed cases by utilizing self-reported symptoms. The models are developed at two levels of data granularity - local models, which are trained at the state level, and a single global model which is trained on the combined data aggregated across all states. Our results indicate a lower error on the local models as opposed to the global model. In addition, we also show that the most important symptoms (features) vary considerably from state to state. This work demonstrates that the models developed on crowd-sourced data, curated via online platforms, can complement the existing epidemiological surveillance infrastructure in a cost-effective manner. The code is publicly available at https://github.com/parthpatwa/Can-Self-Reported-Symptoms-Predict-Daily-COVID-19-Cases."
                    },
                    {
                        "title": "MIT SafePaths Card (MiSaCa): Augmenting Paper Based Vaccination Cards with Printed Codes",
                        "abstract": "In this early draft, we describe a user-centric, card-based system for vaccine distribution. Our system makes use of digitally signed QR codes and their use for phased vaccine distribution, vaccine administration/record-keeping, immunization verification, and follow-up symptom reporting. Furthermore, we propose and describe a complementary scanner app system to be used by vaccination clinics, public health officials, and immunization verification parties to effectively utilize card-based framework. We believe that the proposed system provides a privacy-preserving and efficient framework for vaccine distribution in both developed and developing regions."
                    },
                    {
                        "title": "Improved Customer Transaction Classification using Semi-Supervised Knowledge Distillation",
                        "abstract": "In pickup and delivery services, transaction classification based on customer provided free text is a challenging problem. It involves the association of a wide variety of customer inputs to a fixed set of categories while adapting to the various customer writing styles. This categorization is important for the business: it helps understand the market needs and trends, and also assist in building a personalized experience for different segments of the customers. Hence, it is vital to capture these category information trends at scale, with high precision and recall. In this paper, we focus on a specific use-case where a single category drives each transaction. We propose a cost-effective transaction classification approach based on semi-supervision and knowledge distillation frameworks. The approach identifies the category of a transaction using free text input given by the customer. We use weak labelling and notice that the performance gains are similar to that of using human-annotated samples. On a large internal dataset and on 20Newsgroup dataset, we see that RoBERTa performs the best for the categorization tasks. Further, using an ALBERT model (it has 33x fewer parameters vis-a-vis parameters of RoBERTa), with RoBERTa as the Teacher, we see a performance similar to that of RoBERTa and better performance over unadapted ALBERT. This framework, with ALBERT as a student and RoBERTa as teacher, is further referred to as R-ALBERT in this paper. The model is in production and is used by business to understand changing trends and take appropriate decisions."
                    },
                    {
                        "title": "COVID-19 Tests Gone Rogue: Privacy, Efficacy, Mismanagement and Misunderstandings",
                        "abstract": "COVID-19 testing, the cornerstone for effective screening and identification of COVID-19 cases, remains paramount as an intervention tool to curb the spread of COVID-19 both at local and national levels. However, the speed at which the pandemic struck and the response was rolled out, the widespread impact on healthcare infrastructure, the lack of sufficient preparation within the public health system, and the complexity of the crisis led to utter confusion among test-takers. Invasion of privacy remains a crucial concern. The user experience of test takers remains low. User friction affects user behavior and discourages participation in testing programs. Test efficacy has been overstated. Test results are poorly understood resulting in inappropriate follow-up recommendations. Herein, we review the current landscape of COVID-19 testing, identify four key challenges, and discuss the consequences of the failure to address these challenges. The current infrastructure around testing and information propagation is highly privacy-invasive and does not leverage scalable digital components. In this work, we discuss challenges complicating the existing covid-19 testing ecosystem and highlight the need to improve the testing experience for the user and reduce privacy invasions. Digital tools will play a critical role in resolving these challenges."
                    },
                    {
                        "title": "Paper card-based vs application-based vaccine credentials: a comparison",
                        "abstract": "In this early draft, we provide an overview on similarities and differences in the implementation of a paper card-based vaccine credential system and an app-based vaccine credential system. A vaccine credential's primary goal is to regulate entry and ensure safety of individuals within densely packed public locations and workspaces. This is critical for containing the rapid spread of Covid-19 in densely packed public locations since a single individual can infect a large majority of people in a crowd. A vaccine credential can also provide information such as an individual's Covid-19 vaccination history and adverse symptom reaction history to judge their potential impact on the overall health of individuals within densely packed public locations and workspaces. After completing the comparisons, we believe a card-based implementation will benefit regions with less socioeconomic mobility, limited resources, and stagnant administrations. An app-based implementation on the other hand will benefit regions with equitable internet access and lower technological divide. We also believe an interoperable system of both credential systems will work best for regions with enormous working-class populations and dense housing clusters."
                    },
                    {
                        "title": "Hostility Detection and Covid-19 Fake News Detection in Social Media",
                        "abstract": "Withtheadventofsocialmedia,therehasbeenanextremely rapid increase in the content shared online. Consequently, the propagation of fake news and hostile messages on social media platforms has also skyrocketed. In this paper, we address the problem of detecting hostile and fake content in the Devanagari (Hindi) script as a multi-class, multi-label problem. Using NLP techniques, we build a model that makes use of an abusive language detector coupled with features extracted via Hindi BERT and Hindi FastText models and metadata. Our model achieves a 0.97 F1 score on coarse grain evaluation on Hostility detection task. Additionally, we built models to identify fake news related to Covid-19 in English tweets. We leverage entity information extracted from the tweets along with textual representations learned from word embeddings and achieve a 0.93 F1 score on the English fake news detection task."
                    },
                    {
                        "title": "Vaccination Worldwide: Strategies, Distribution and Challenges",
                        "abstract": "The Coronavirus 2019 (Covid-19) pandemic caused by the SARS-CoV-2 virus represents an unprecedented crisis for our planet. It is a bane of the \u00fcber connected world that we live in that this virus has affected almost all countries and caused mortality and economic upheaval at a scale whose effects are going to be felt for generations to come. While we can all be buoyed at the pace at which vaccines have been developed and brought to market, there are still challenges ahead for all countries to get their populations vaccinated equitably and effectively. This paper provides an overview of ongoing immunization efforts in various countries. In this early draft, we have identified a few key factors that we use to review different countries' current COVID-19 immunization strategies and their strengths and draw conclusions so that policymakers worldwide can learn from them. Our paper focuses on processes related to vaccine approval, allocation and prioritization, distribution strategies, population to vaccine ratio, vaccination governance, accessibility and use of digital solutions, and government policies. The statistics and numbers are dated as per the draft date [June 24th, 2021]."
                    },
                    {
                        "title": "Comparing Card-based Vaccine Credential Systems with App-based Vaccine Credential Systems",
                        "abstract": "In this early draft, we compare and contrast the technicalities present in the implementation of a card-based vaccine credentials system and an app-based vaccine credential system. We have chosen the domains of symptom reporting, fraud and impersonation, feasibility, scalability, equity, future data aggregation, importability of health data, and operability to explore each system\u2019s features and drawbacks."
                    },
                    {
                        "title": "Deviations in Predicted COVID-19 cases in the US during early months of 2021 relate to rise in B.1.526 and its family of variants",
                        "abstract": "Objective: To investigate the abrogation of COVID-19 case declines from predicted rates in the US in relationship to viral variants and mutations. Design: Epidemiological prediction and time series study of COVID-19 in the US by State. Setting: Community testing and sequencing of COVID-19 in the US. Participants: Time series US COVID-19 case data from the Johns Hopkins University CSSE database. Time series US Variant and Mutation data from the GISAID database. Main outcome measures: Primary outcomes were statistical modeling of US state deviations from epidemiological predictions, percentage of COVID-19 variants, percentage of COVID-19 mutations, and reported SARS-CoV-2 infections. Results: Deviations in epidemiological predictions of COVID-19 case declines in the North Eastern US in March 2021 were highly positively related to percentage of B.1.526 (Iota) lineage (p < 10e - 7) and B.1.526.2 (p < 10 - 8) and the T95I mutation (p < 10e - 9). They were related inversely to B.1.427 and B.1.429 (Epsilon) and there was a trend for association with B.1.1.7 (Alpha) lineage. Conclusion: Deviations from accurate predictive models are useful for investigating potential immune escape of COVID-19 variants at the population level. The B.1.526 and B.1.526.2 lineages likely have a high potential for immune escape and should be designated as variants of concern. The T95I mutation which is present in the B.1.526, B.1.526.2, and B.1.617.2(Delta) lineages in the US warrants further investigation as a mutation of concern."
                    },
                    {
                        "title": "Mobile Apps Prioritizing Privacy, Efficiency and Equity: A Decentralized Approach to COVID-19 Vaccination Coordination",
                        "abstract": "In this early draft, we describe a decentralized, app-based approach to COVID-19 vaccine distribution that facilitates zero knowledge verification, dynamic vaccine scheduling, continuous symptoms reporting, access to aggregate analytics based on population trends and more. To ensure equity, our solution is developed to work with limited internet access as well. In addition, we describe the six critical functions that we believe last mile vaccination management platforms must perform, examine existing vaccine management systems, and present a model for privacy-focused, individual-centric solutions."
                    },
                    {
                        "title": "Challenges in Equitable COVID-19 Vaccine Distribution: A Roadmap for Digital Technology Solutions",
                        "abstract": "The COVID-19 pandemic has led to a need for widespread and rapid vaccine development. As several vaccines have recently been approved for human use or are in di\ufb00erent stages of development, governments across the world are preparing comprehensive guidelines for vaccine distribution and monitoring. In this early article, we identify challenges in logistics, health outcomes, user-centric matters, and communication associated with disease-related, individual, societal, economic, and privacy consequences. Primary challenges include di\ufb03culty in equitable distribution, vaccine e\ufb03cacy, duration of immunity, multi-dose adherence, and privacy-focused record keeping to be HIPAA compliant. While many of these challenges have been previously identi\ufb01ed and addressed, some have not been acknowledged from a comprehensive view accounting for unprecedented interactions between challenges and speci\ufb01c populations. The logistics of equitable widespread vaccine distribution in disparate populations and countries of various economic, racial, and cultural constitutions must be thoroughly examined and accounted for. We also describe unique challenges regarding the e\ufb03cacy of vaccines in specialized populations including children, the elderly, and immunocompromised individuals [1, 2]. Given these complicated issues, the importance of privacy-focused, user-centric systems for vaccine education and incentivization along with clear communication from governments, organizations, and academic institutions is imperative. These challenges are by no means insurmountable, but require careful attention to avoid consequences spanning a range of disease-related, individual, societal, economic, and security domains."
                    },
                    {
                        "title": "Bitions@DravidianLangTech-EACL2021: Ensemble of Multilingual Language Models with Pseudo Labeling for offence Detection in Dravidian Languages",
                        "abstract": "With the advent of social media, we have seen a proliferation of data and public discourse. Unfortunately, this includes offensive content as well. The problem is exacerbated due to the sheer number of languages spoken on these platforms and the multiple other modalities used for sharing offensive content (images, gifs, videos and more). In this paper, we propose a multilingual ensemble-based model that can identify offensive content targeted against an individual (or group) in low resource Dravidian language. Our model is able to handle code-mixed data as well as instances where the script used is mixed (for instance, Tamil and Latin). Our solution ranked number one for the Malayalam dataset and ranked 4th and 5th for Tamil and Kannada, respectively."
                    },
                    {
                        "title": "Proximity Sensing: Modeling and Understanding Noisy RSSI-BLE Signals and Other Mobile Sensor Data for Digital Contact Tracing",
                        "abstract": "As we await a vaccine, social-distancing via efficient contact tracing has emerged as the primary health strategy to dampen the spread of COVID-19. To enable efficient digital contact tracing, we present a novel system to estimate pair-wise individual proximity, via a joint model of Bluetooth Low Energy (BLE) signals with other on-device sensors (accelerometer, magnetometer, gyroscope). We explore multiple ways of interpreting the sensor data stream (time-series, histogram, etc) and use several statistical and deep learning methods to learn representations for sensing proximity. We report the normalized Decision Cost Function (nDCF) metric and analyze the differential impact of the various input signals, as well as discuss various challenges associated with this task."
                    },
                    {
                        "title": "Clinical Landscape of COVID-19 Testing: Difficult Choices",
                        "abstract": "The coronavirus disease 2019 (COVID-19) pandemic has spread rapidly across the world, leading to enormous amounts of human death and economic loss. Until definitive preventive or curative measures are developed, policies regarding testing, contact tracing, and quarantine remain the best public health tools for curbing viral spread. Testing is a crucial component of these efforts, enabling the identification and isolation of infected individuals. Differences in testing methodologies, time frames, and outcomes can have an impact on their overall efficiency, usability and efficacy. In this early draft, we draw a comparison between the various types of diagnostic tests including PCR, antigen, and home tests in relation to their relative advantages, disadvantages, and use cases. We also look into alternative and unconventional methods. Further, we analyze the short-term and long-term impacts of the virus and its testing on various verticals such as business, government laws, policies, and healthcare."
                    },
                    {
                        "title": "COVID-19 Outbreak Prediction and Analysis using Self Reported Symptoms",
                        "abstract": "It is crucial for policymakers to understand the community prevalence of COVID-19 so combative resources can be effectively allocated and prioritized during the COVID-19 pandemic. Traditionally, community prevalence has been assessed through diagnostic and antibody testing data. However, despite the increasing availability of COVID-19 testing, the required level has not been met in parts of the globe, introducing a need for an alternative method for communities to determine disease prevalence. This is further complicated by the observation that COVID-19 prevalence and spread vary across different spatial, temporal, and demographic verticals. In this study, we study trends in the spread of COVID-19 by utilizing the results of self-reported COVID-19 symptoms surveys as a complement to COVID-19 testing reports. This allows us to assess community disease prevalence, even in areas with low COVID-19 testing ability. Using individually reported symptom data from various populations, our method predicts the likely percentage of the population that tested positive for COVID-19. We achieved a mean absolute error (MAE) of 1.14 and mean relative error (MRE) of 60.40% with 95% confidence interval as [60.12, 60.67]. This implies that our model predicts +/- 1140 cases than the original in a population of 1 million. In addition, we forecast the location-wise percentage of the population testing positive for the next 30 days using self-reported symptoms data from previous days. The MAE for this method is as low as 0.15 (MRE of 11.28% with 95% confidence interval [10.9, 11.6]) for New York. We present an analysis of these results, exposing various clinical attributes of interest across different demographics. Lastly, we qualitatively analyze how various policy enactments (testing, curfew) affect the prevalence of COVID-19 in a community."
                    }
                ]
            },
            "696c1c7d-9774-4dab-b597-672a6f4705d9": {
                "pk": "696c1c7d-9774-4dab-b597-672a6f4705d9",
                "name": "Golnoosh Farnadi",
                "collaborators": [
                    "Afaf Ta\u00efk",
                    "Negar Rostamzadeh",
                    "Mohammad Havaei",
                    "Nicola Neophytou",
                    "Yash More",
                    "Saber Malekmohammadi",
                    "Prakhar Ganesh",
                    "Armin Moradi",
                    "Ahmed Imtiaz Humayun",
                    "Ibtihel Amara"
                ],
                "domain": [
                    "Generative Models",
                    "Fairness",
                    "Language Models",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Understanding the Local Geometry of Generative Model Manifolds",
                        "abstract": "Deep generative models learn continuous representations of complex data manifolds using a finite number of samples during training. For a pre-trained generative model, the common way to evaluate the quality of the manifold representation learned, is by computing global metrics like Fr\\'echet Inception Distance using a large number of generated and real samples. However, generative model performance is not uniform across the learned manifold, e.g., for \\textit{foundation models} like Stable Diffusion generation performance can vary significantly based on the conditioning or initial noise vector being denoised. In this paper we study the relationship between the \\textit{local geometry of the learned manifold} and downstream generation. Based on the theory of continuous piecewise-linear (CPWL) generators, we use three geometric descriptors - scaling ($\\psi$), rank ($\\nu$), and complexity ($\\delta$) - to characterize a pre-trained generative model manifold locally. We provide quantitative and qualitative evidence showing that for a given latent, the local descriptors are correlated with generation aesthetics, artifacts, uncertainty, and even memorization. Finally we demonstrate that training a \\textit{reward model} on the local geometry can allow controlling the likelihood of a generated sample under the learned distribution."
                    },
                    {
                        "title": "Promoting Fair Vaccination Strategies through Influence Maximization: A Case Study on COVID-19 Spread",
                        "abstract": "The aftermath of the Covid-19 pandemic saw more severe outcomes for racial minority groups and economically-deprived communities. Such disparities can be explained by several factors, including unequal access to healthcare, as well as the inability of low income groups to reduce their mobility due to work or social obligations. Moreover, senior citizens were found to be more susceptible to severe symptoms, largely due to age-related health reasons. Adapting vaccine distribution strategies to consider a range of demographics is therefore essential to address these disparities. In this study, we propose a novel approach that utilizes influence maximization (IM) on mobility networks to develop vaccination strategies which incorporate demographic fairness. By considering factors such as race, social status, age, and associated risk factors, we aim to optimize vaccine distribution to achieve various fairness definitions for one or more protected attributes at a time. Through extensive experiments conducted on Covid-19 spread in three major metropolitan areas across the United States, we demonstrate the effectiveness of our proposed approach in reducing disease transmission and promoting fairness in vaccination distribution."
                    },
                    {
                        "title": "Trust No Bot: Discovering Personal Disclosures in Human-LLM Conversations in the Wild",
                        "abstract": "Measuring personal disclosures made in human-chatbot interactions can provide a better understanding of users' AI literacy and facilitate privacy research for large language models (LLMs). We run an extensive, fine-grained analysis on the personal disclosures made by real users to commercial GPT models, investigating the leakage of personally identifiable and sensitive information. To understand the contexts in which users disclose to chatbots, we develop a taxonomy of tasks and sensitive topics, based on qualitative and quantitative analysis of naturally occurring conversations. We discuss these potential privacy harms and observe that: (1) personally identifiable information (PII) appears in unexpected contexts such as in translation or code editing (48% and 16% of the time, respectively) and (2) PII detection alone is insufficient to capture the sensitive topics that are common in human-chatbot interactions, such as detailed sexual preferences or specific drug use habits. We believe that these high disclosure rates are of significant importance for researchers and data curators, and we call for the design of appropriate nudging mechanisms to help users moderate their interactions."
                    },
                    {
                        "title": "From Representational Harms to Quality-of-Service Harms: A Case Study on Llama 2 Safety Safeguards",
                        "abstract": "Recent progress in large language models (LLMs) has led to their widespread adoption in various domains. However, these advancements have also introduced additional safety risks and raised concerns regarding their detrimental impact on already marginalized populations. Despite growing mitigation efforts to develop safety safeguards, such as supervised safety-oriented fine-tuning and leveraging safe reinforcement learning from human feedback, multiple concerns regarding the safety and ingrained biases in these models remain. Furthermore, previous work has demonstrated that models optimized for safety often display exaggerated safety behaviors, such as a tendency to refrain from responding to certain requests as a precautionary measure. As such, a clear trade-off between the helpfulness and safety of these models has been documented in the literature. In this paper, we further investigate the effectiveness of safety measures by evaluating models on already mitigated biases. Using the case of Llama 2 as an example, we illustrate how LLMs' safety responses can still encode harmful assumptions. To do so, we create a set of non-toxic prompts, which we then use to evaluate Llama models. Through our new taxonomy of LLMs responses to users, we observe that the safety/helpfulness trade-offs are more pronounced for certain demographic groups which can lead to quality-of-service harms for marginalized populations."
                    },
                    {
                        "title": "Learning to Build Solutions in Stochastic Matching Problems Using Flows (Student Abstract)",
                        "abstract": "Generative Flow Networks, known as GFlowNets, have been introduced in recent times, presenting an exciting possibility for neural networks to model distributions across various data structures. In this paper, we broaden their applicability to encompass scenarios where the data structures are optimal solutions of a combinatorial problem. Concretely, we propose the use of GFlowNets to learn the distribution of optimal solutions for kidney exchange problems (KEPs), a generalized form of matching problems involving cycles."
                    },
                    {
                        "title": "Mitigating Disparate Impact of Differential Privacy in Federated Learning through Robust Clustering",
                        "abstract": "Federated Learning (FL) is a decentralized machine learning (ML) approach that keeps data localized and often incorporates Differential Privacy (DP) to enhance privacy guarantees. Similar to previous work on DP in ML, we observed that differentially private federated learning (DPFL) introduces performance disparities, particularly affecting minority groups. Recent work has attempted to address performance fairness in vanilla FL through clustering, but this method remains sensitive and prone to errors, which are further exacerbated by the DP noise in DPFL. To fill this gap, in this paper, we propose a novel clustered DPFL algorithm designed to effectively identify clients' clusters in highly heterogeneous settings while maintaining high accuracy with DP guarantees. To this end, we propose to cluster clients based on both their model updates and training loss values. Our proposed approach also addresses the server's uncertainties in clustering clients' model updates by employing larger batch sizes along with Gaussian Mixture Model (GMM) to alleviate the impact of noise and potential clustering errors, especially in privacy-sensitive scenarios. We provide theoretical analysis of the effectiveness of our proposed approach. We also extensively evaluate our approach across diverse data distributions and privacy budgets and show its effectiveness in mitigating the disparate impact of DP in FL settings with a small computational cost."
                    },
                    {
                        "title": "FairLoRA: Unpacking Bias Mitigation in Vision Models with Fairness-Driven Low-Rank Adaptation",
                        "abstract": "Recent advances in parameter-efficient fine-tuning methods, such as Low Rank Adaptation (LoRA), have gained significant attention for their ability to efficiently adapt large foundational models to various downstream tasks. These methods are appreciated for achieving performance comparable to full fine-tuning on aggregate-level metrics, while significantly reducing computational costs. To systematically address fairness in LLMs previous studies fine-tune on fairness specific data using a larger LoRA rank than typically used. In this paper, we introduce FairLoRA, a novel fairness-specific regularizer for LoRA aimed at reducing performance disparities across data subgroups by minimizing per-class variance in loss. To the best of our knowledge, we are the first to introduce a fairness based finetuning through LoRA. Our results demonstrate that the need for higher ranks to mitigate bias is not universal; it depends on factors such as the pre-trained model, dataset, and task. More importantly, we systematically evaluate FairLoRA across various vision models, including ViT, DiNO, and CLIP, in scenarios involving distribution shifts. We further emphasize the necessity of using multiple fairness metrics to obtain a holistic assessment of fairness, rather than relying solely on the metric optimized during training."
                    },
                    {
                        "title": "The Cost of Arbitrariness for Individuals: Examining the Legal and Technical Challenges of Model Multiplicity",
                        "abstract": "Model multiplicity, the phenomenon where multiple models achieve similar performance despite different underlying learned functions, introduces arbitrariness in model selection. While this arbitrariness may seem inconsequential in expectation, its impact on individuals can be severe. This paper explores various individual concerns stemming from multiplicity, including the effects of arbitrariness beyond final predictions, disparate arbitrariness for individuals belonging to protected groups, and the challenges associated with the arbitrariness of a single algorithmic system creating a monopoly across various contexts. It provides both an empirical examination of these concerns and a comprehensive analysis from the legal standpoint, addressing how these issues are perceived in the anti-discrimination law in Canada. We conclude the discussion with technical challenges in the current landscape of model multiplicity to meet legal requirements and the legal gap between current law and the implications of arbitrariness in model selection, highlighting relevant future research directions for both disciplines."
                    },
                    {
                        "title": "Understanding Intrinsic Socioeconomic Biases in Large Language Models",
                        "abstract": "Large Language Models (LLMs) are increasingly integrated into critical decision-making processes, such as loan approvals and visa applications, where inherent biases can lead to discriminatory outcomes. In this paper, we examine the nuanced relationship between demographic attributes and socioeconomic biases in LLMs, a crucial yet understudied area of fairness in LLMs. We introduce a novel dataset of one million English sentences to systematically quantify socioeconomic biases across various demographic groups. Our findings reveal pervasive socioeconomic biases in both established models such as GPT-2 and state-of-the-art models like Llama 2 and Falcon. We demonstrate that these biases are significantly amplified when considering intersectionality, with LLMs exhibiting a remarkable capacity to extract multiple demographic attributes from names and then correlate them with specific socioeconomic biases. This research highlights the urgent necessity for proactive and robust bias mitigation techniques to safeguard against discriminatory outcomes when deploying these powerful models in critical real-world applications."
                    },
                    {
                        "title": "Advancing Cultural Inclusivity: Optimizing Embedding Spaces for Balanced Music Recommendations",
                        "abstract": "Popularity bias in music recommendation systems -- where artists and tracks with the highest listen counts are recommended more often -- can also propagate biases along demographic and cultural axes. In this work, we identify these biases in recommendations for artists from underrepresented cultural groups in prototype-based matrix factorization methods. Unlike traditional matrix factorization methods, prototype-based approaches are interpretable. This allows us to directly link the observed bias in recommendations for minority artists (the effect) to specific properties of the embedding space (the cause). We mitigate popularity bias in music recommendation through capturing both users' and songs' cultural nuances in the embedding space. To address these challenges while maintaining recommendation quality, we propose two novel enhancements to the embedding space: i) we propose an approach to filter-out the irrelevant prototypes used to represent each user and item to improve generalizability, and ii) we introduce regularization techniques to reinforce a more uniform distribution of prototypes within the embedding space. Our results demonstrate significant improvements in reducing popularity bias and enhancing demographic and cultural fairness in music recommendations while achieving competitive -- if not better -- overall performance."
                    },
                    {
                        "title": "Position: Cracking the Code of Cascading Disparity Towards Marginalized Communities",
                        "abstract": "The rise of foundation models holds immense promise for advancing AI, but this progress may amplify existing risks and inequalities, leaving marginalized communities behind. In this position paper, we discuss that disparities towards marginalized communities - performance, representation, privacy, robustness, interpretability and safety - are not isolated concerns but rather interconnected elements of a cascading disparity phenomenon. We contrast foundation models with traditional models and highlight the potential for exacerbated disparity against marginalized communities. Moreover, we emphasize the unique threat of cascading impacts in foundation models, where interconnected disparities can trigger long-lasting negative consequences, specifically to the people on the margin. We define marginalized communities within the machine learning context and explore the multifaceted nature of disparities. We analyze the sources of these disparities, tracing them from data creation, training and deployment procedures to highlight the complex technical and socio-technical landscape. To mitigate the pressing crisis, we conclude with a set of calls to action to mitigate disparity at its source."
                    },
                    {
                        "title": "On the Implicit Relation Between Low-Rank Adaptation and Differential Privacy",
                        "abstract": "A significant approach in natural language processing involves large-scale pre-training models on general domain data followed by their adaptation to specific tasks or domains. As models grow in size, full fine-tuning all of their parameters becomes increasingly impractical. To address this, some methods for low-rank task adaptation of language models have been proposed, e.g., LoRA and FLoRA. These methods keep the pre-trained model weights fixed and incorporate trainable low-rank decomposition matrices into some layers of the transformer architecture, called adapters. This approach significantly reduces the number of trainable parameters required for downstream tasks compared to full fine-tuning all parameters. In this work, we look at low-rank adaptation from the lens of data privacy. We show theoretically that the low-rank adaptation used in LoRA and FLoRA is equivalent to injecting some random noise into the batch gradients w.r.t the adapter parameters, and we quantify the variance of the injected noise. By establishing a Berry-Esseen type bound on the total variation distance between distribution of the injected noise and a Gaussian distribution with the same variance, we show that the dynamics of low-rank adaptation is close to that of differentially private fine-tuning of the adapters. Finally, using Johnson-Lindenstrauss lemma, we show that when augmented with gradient scaling, low-rank adaptation is very close to performing DPSGD algorithm with a fixed noise scale to fine-tune the adapters. These theoretical findings suggest that unlike other existing fine-tuning algorithms, low-rank adaptation provides privacy w.r.t the fine-tuning data implicitly."
                    },
                    {
                        "title": "Towards More Realistic Extraction Attacks: An Adversarial Perspective",
                        "abstract": "Language models are prone to memorizing large parts of their training data, making them vulnerable to extraction attacks. Existing research on these attacks remains limited in scope, often studying isolated trends rather than the real-world interactions with these models. In this paper, we revisit extraction attacks from an adversarial perspective, exploiting the brittleness of language models. We find significant churn in extraction attack trends, i.e., even minor, unintuitive changes to the prompt, or targeting smaller models and older checkpoints, can exacerbate the risks of extraction by up to $2-4 \\times$. Moreover, relying solely on the widely accepted verbatim match underestimates the extent of extracted information, and we provide various alternatives to more accurately capture the true risks of extraction. We conclude our discussion with data deduplication, a commonly suggested mitigation strategy, and find that while it addresses some memorization concerns, it remains vulnerable to the same escalation of extraction risks against a real-world adversary. Our findings highlight the necessity of acknowledging an adversary's true capabilities to avoid underestimating extraction risks."
                    },
                    {
                        "title": "Fairness Incentives in Response to Unfair Dynamic Pricing",
                        "abstract": "The use of dynamic pricing by profit-maximizing firms gives rise to demand fairness concerns, measured by discrepancies in consumer groups' demand responses to a given pricing strategy. Notably, dynamic pricing may result in buyer distributions unreflective of those of the underlying population, which can be problematic in markets where fair representation is socially desirable. To address this, policy makers might leverage tools such as taxation and subsidy to adapt policy mechanisms dependent upon their social objective. In this paper, we explore the potential for AI methods to assist such intervention strategies. To this end, we design a basic simulated economy, wherein we introduce a dynamic social planner (SP) to generate corporate taxation schedules geared to incentivizing firms towards adopting fair pricing behaviours, and to use the collected tax budget to subsidize consumption among underrepresented groups. To cover a range of possible policy scenarios, we formulate our social planner's learning problem as a multi-armed bandit, a contextual bandit and finally as a full reinforcement learning (RL) problem, evaluating welfare outcomes from each case. To alleviate the difficulty in retaining meaningful tax rates that apply to less frequently occurring brackets, we introduce FairReplayBuffer, which ensures that our RL agent samples experiences uniformly across a discretized fairness space. We find that, upon deploying a learned tax and redistribution policy, social welfare improves on that of the fairness-agnostic baseline, and approaches that of the analytically optimal fairness-aware baseline for the multi-armed and contextual bandit settings, and surpassing it by 13.19% in the full RL setting."
                    },
                    {
                        "title": "Hallucination Detox: Sensitive Neuron Dropout (SeND) for Large Language Model Training",
                        "abstract": "As large language models (LLMs) become increasingly deployed across various industries, concerns regarding their reliability, particularly due to hallucinations-outputs that are factually inaccurate or irrelevant to user input-have grown. Our research investigates the relationship between the training process and the emergence of hallucinations to address a key gap in existing research that focuses primarily on post hoc detection and mitigation strategies. Using models from the Pythia suite (70M-12B parameters) and several hallucination detection metrics, we analyze hallucination trends throughout training and explore LLM internal dynamics. We introduce SEnsitive Neuron Dropout (SeND), a novel training protocol designed to mitigate hallucinations by reducing variance during training. SeND achieves this by deterministically dropping neurons with significant variability on a dataset, referred to as Sensitive Neurons. In addition, we develop an unsupervised hallucination detection metric, Efficient EigenScore (EES), which approximates the traditional EigenScore in 2x speed. This efficient metric is integrated into our protocol, allowing SeND to be both computationally scalable and effective at reducing hallucinations. Our empirical evaluation demonstrates that our approach improves LLM reliability at test time by up to 40% compared to normal training while also providing an efficient method to improve factual accuracy when adapting LLMs to domains such as Wikipedia and Medical datasets."
                    },
                    {
                        "title": "Multilingual Hallucination Gaps in Large Language Models",
                        "abstract": "Large language models (LLMs) are increasingly used as alternatives to traditional search engines given their capacity to generate text that resembles human language. However, this shift is concerning, as LLMs often generate hallucinations, misleading or false information that appears highly credible. In this study, we explore the phenomenon of hallucinations across multiple languages in freeform text generation, focusing on what we call multilingual hallucination gaps. These gaps reflect differences in the frequency of hallucinated answers depending on the prompt and language used. To quantify such hallucinations, we used the FactScore metric and extended its framework to a multilingual setting. We conducted experiments using LLMs from the LLaMA, Qwen, and Aya families, generating biographies in 19 languages and comparing the results to Wikipedia pages. Our results reveal variations in hallucination rates, especially between high and low resource languages, raising important questions about LLM multilingual performance and the challenges in evaluating hallucinations in multilingual freeform text generation."
                    },
                    {
                        "title": "Causal Adversarial Perturbations for Individual Fairness and Robustness in Heterogeneous Data Spaces",
                        "abstract": "As responsible AI gains importance in machine learning algorithms, properties like fairness, adversarial robustness, and causality have received considerable attention in recent years. However, despite their individual significance, there remains a critical gap in simultaneously exploring and integrating these properties. In this paper, we propose a novel approach that examines the relationship between individual fairness, adversarial robustness, and structural causal models (SCMs) in heterogeneous data spaces, particularly when dealing with discrete sensitive attributes. We use SCMs and sensitive attributes to create a fair metric and apply it to measure semantic similarity among individuals. By introducing a novel causal adversarial perturbation (CAP) and applying adversarial training, we create a new regularizer that combines individual fairness, causality, and robustness in the classifier. Our method is evaluated on both real-world and synthetic datasets, demonstrating its effectiveness in achieving an accurate classifier that simultaneously exhibits fairness, adversarial robustness, and causal awareness."
                    },
                    {
                        "title": "Fairness Through Domain Awareness: Mitigating Popularity Bias For Music Discovery",
                        "abstract": "As online music platforms grow, music recommender systems play a vital role in helping users navigate and discover content within their vast musical databases. At odds with this larger goal, is the presence of popularity bias, which causes algorithmic systems to favor mainstream content over, potentially more relevant, but niche items. In this work we explore the intrinsic relationship between music discovery and popularity bias. To mitigate this issue we propose a domain-aware, individual fairness-based approach which addresses popularity bias in graph neural network (GNNs) based recommender systems. Our approach uses individual fairness to reflect a ground truth listening experience, i.e., if two songs sound similar, this similarity should be reflected in their representations. In doing so, we facilitate meaningful music discovery that is robust to popularity bias and grounded in the music domain. We apply our BOOST methodology to two discovery based tasks, performing recommendations at both the playlist level and user level. Then, we ground our evaluation in the cold start setting, showing that our approach outperforms existing fairness benchmarks in both performance and recommendation of lesser-known content. Finally, our analysis explains why our proposed methodology is a novel and promising approach to mitigating popularity bias and improving the discovery of new and niche content in music recommender systems."
                    },
                    {
                        "title": "Tidying Up the Conversational Recommender Systems' Biases",
                        "abstract": "The growing popularity of language models has sparked interest in conversational recommender systems (CRS) within both industry and research circles. However, concerns regarding biases in these systems have emerged. While individual components of CRS have been subject to bias studies, a literature gap remains in understanding specific biases unique to CRS and how these biases may be amplified or reduced when integrated into complex CRS models. In this paper, we provide a concise review of biases in CRS by surveying recent literature. We examine the presence of biases throughout the system's pipeline and consider the challenges that arise from combining multiple models. Our study investigates biases in classic recommender systems and their relevance to CRS. Moreover, we address specific biases in CRS, considering variations with and without natural language understanding capabilities, along with biases related to dialogue systems and language models. Through our findings, we highlight the necessity of adopting a holistic perspective when dealing with biases in complex CRS models."
                    },
                    {
                        "title": "Unraveling the Interconnected Axes of Heterogeneity in Machine Learning for Democratic and Inclusive Advancements",
                        "abstract": "The growing utilization of machine learning (ML) in decision-making processes raises questions about its benefits to society. In this study, we identify and analyze three axes of heterogeneity that significantly influence the trajectory of ML products. These axes are i) values, culture and regulations, ii) data composition, and iii) resource and infrastructure capacity. We demonstrate how these axes are interdependent and mutually influence one another, emphasizing the need to consider and address them jointly. Unfortunately, the current research landscape falls short in this regard, often failing to adopt a holistic approach. We examine the prevalent practices and methodologies that skew these axes in favor of a selected few, resulting in power concentration, homogenized control, and increased dependency. We discuss how this fragmented study of the three axes poses a significant challenge, leading to an impractical solution space that lacks reflection of real-world scenarios. Addressing these issues is crucial to ensure a more comprehensive understanding of the interconnected nature of society and to foster the democratic and inclusive development of ML systems that are more aligned with real-world complexities and its diverse requirements."
                    }
                ]
            },
            "ffd45622-4c54-436f-8722-a15ca8a7a8c4": {
                "pk": "ffd45622-4c54-436f-8722-a15ca8a7a8c4",
                "name": "Simon Lacoste-Julien",
                "collaborators": [
                    "S\u00e9bastien Lachapelle",
                    "Y. Bengio",
                    "Yan Zhang",
                    "Divyat Mahajan",
                    "Ioannis Mitliagkas",
                    "T. Deleu",
                    "R\u00e9mi Le Priol",
                    "Ant'onio G'ois",
                    "A. Baratin",
                    "Vit\u00f3ria Barin Pacela"
                ],
                "domain": [
                    "Disentangled Representation",
                    "Causal Inference",
                    "Optimization",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Nonparametric Partial Disentanglement via Mechanism Sparsity: Sparse Actions, Interventions and Sparse Temporal Dependencies",
                        "abstract": "This work introduces a novel principle for disentanglement we call mechanism sparsity regularization, which applies when the latent factors of interest depend sparsely on observed auxiliary variables and/or past latent factors. We propose a representation learning method that induces disentanglement by simultaneously learning the latent factors and the sparse causal graphical model that explains them. We develop a nonparametric identifiability theory that formalizes this principle and shows that the latent factors can be recovered by regularizing the learned causal graph to be sparse. More precisely, we show identifiablity up to a novel equivalence relation we call\"consistency\", which allows some latent factors to remain entangled (hence the term partial disentanglement). To describe the structure of this entanglement, we introduce the notions of entanglement graphs and graph preserving functions. We further provide a graphical criterion which guarantees complete disentanglement, that is identifiability up to permutations and element-wise transformations. We demonstrate the scope of the mechanism sparsity principle as well as the assumptions it relies on with several worked out examples. For instance, the framework shows how one can leverage multi-node interventions with unknown targets on the latent factors to disentangle them. We further draw connections between our nonparametric results and the now popular exponential family assumption. Lastly, we propose an estimation procedure based on variational autoencoders and a sparsity constraint and demonstrate it on various synthetic datasets. This work is meant to be a significantly extended version of Lachapelle et al. (2022)."
                    },
                    {
                        "title": "Performative Prediction on Games and Mechanism Design",
                        "abstract": "Predictions often influence the reality which they aim to predict, an effect known as performativity. Existing work focuses on accuracy maximization under this effect, but model deployment may have important unintended impacts, especially in multiagent scenarios. In this work, we investigate performative prediction in a concrete game-theoretic setting where social welfare is an alternative objective to accuracy maximization. We explore a collective risk dilemma scenario where maximising accuracy can negatively impact social welfare, when predicting collective behaviours. By assuming knowledge of a Bayesian agent behavior model, we then show how to achieve better trade-offs and use them for mechanism design."
                    },
                    {
                        "title": "Promoting Exploration in Memory-Augmented Adam using Critical Momenta",
                        "abstract": "Adaptive gradient-based optimizers, notably Adam, have left their mark in training large-scale deep learning models, offering fast convergence and robustness to hyperparameter settings. However, they often struggle with generalization, attributed to their tendency to converge to sharp minima in the loss landscape. To address this, we propose a new memory-augmented version of Adam that encourages exploration towards flatter minima by incorporating a buffer of critical momentum terms during training. This buffer prompts the optimizer to overshoot beyond narrow minima, promoting exploration. Through comprehensive analysis in simple settings, we illustrate the efficacy of our approach in increasing exploration and bias towards flatter minima. We empirically demonstrate that it can improve model performance for image classification on ImageNet and CIFAR10/100, language modelling on Penn Treebank, and online learning tasks on TinyImageNet and 5-dataset. Our code is available at \\url{https://github.com/chandar-lab/CMOptimizer}."
                    },
                    {
                        "title": "Can We Scale Transformers to Predict Parameters of Diverse ImageNet Models?",
                        "abstract": "Pretraining a neural network on a large dataset is becoming a cornerstone in machine learning that is within the reach of only a few communities with large-resources. We aim at an ambitious goal of democratizing pretraining. Towards that goal, we train and release a single neural network that can predict high quality ImageNet parameters of other neural networks. By using predicted parameters for initialization we are able to boost training of diverse ImageNet models available in PyTorch. When transferred to other datasets, models initialized with predicted parameters also converge faster and reach competitive final performance."
                    },
                    {
                        "title": "PopulAtion Parameter Averaging (PAPA)",
                        "abstract": "Ensemble methods combine the predictions of multiple models to improve performance, but they require significantly higher computation costs at inference time. To avoid these costs, multiple neural networks can be combined into one by averaging their weights. However, this usually performs significantly worse than ensembling. Weight averaging is only beneficial when different enough to benefit from combining them, but similar enough to average well. Based on this idea, we propose PopulAtion Parameter Averaging (PAPA): a method that combines the generality of ensembling with the efficiency of weight averaging. PAPA leverages a population of diverse models (trained on different data orders, augmentations, and regularizations) while slowly pushing the weights of the networks toward the population average of the weights. We also propose PAPA variants (PAPA-all, and PAPA-2) that average weights rarely rather than continuously; all methods increase generalization, but PAPA tends to perform best. PAPA reduces the performance gap between averaging and ensembling, increasing the average accuracy of a population of models by up to 0.8% on CIFAR-10, 1.9% on CIFAR-100, and 1.6% on ImageNet when compared to training independent (non-averaged) models."
                    },
                    {
                        "title": "Weight-Sharing Regularization",
                        "abstract": "Weight-sharing is ubiquitous in deep learning. Motivated by this, we propose a\"weight-sharing regularization\"penalty on the weights $w \\in \\mathbb{R}^d$ of a neural network, defined as $\\mathcal{R}(w) = \\frac{1}{d - 1}\\sum_{i>j}^d |w_i - w_j|$. We study the proximal mapping of $\\mathcal{R}$ and provide an intuitive interpretation of it in terms of a physical system of interacting particles. We also parallelize existing algorithms for $\\operatorname{prox}_\\mathcal{R}$ (to run on GPU) and find that one of them is fast in practice but slow ($O(d)$) for worst-case inputs. Using the physical interpretation, we design a novel parallel algorithm which runs in $O(\\log^3 d)$ when sufficient processors are available, thus guaranteeing fast training. Our experiments reveal that weight-sharing regularization enables fully connected networks to learn convolution-like filters even when pixels have been shuffled while convolutional neural networks fail in this setting. Our code is available on github."
                    },
                    {
                        "title": "Unlocking Slot Attention by Changing Optimal Transport Costs",
                        "abstract": "Slot attention is a powerful method for object-centric modeling in images and videos. However, its set-equivariance limits its ability to handle videos with a dynamic number of objects because it cannot break ties. To overcome this limitation, we first establish a connection between slot attention and optimal transport. Based on this new perspective we propose MESH (Minimize Entropy of Sinkhorn): a cross-attention module that combines the tiebreaking properties of unregularized optimal transport with the speed of regularized optimal transport. We evaluate slot attention using MESH on multiple object-centric learning benchmarks and find significant improvements over slot attention in every setting."
                    },
                    {
                        "title": "Identifiability of Discretized Latent Coordinate Systems via Density Landmarks Detection",
                        "abstract": "Disentanglement aims to recover meaningful latent ground-truth factors from only the observed distribution. Identifiability provides the theoretical grounding for disentanglement to be well-founded. Unfortunately, unsupervised identifiability of independent latent factors is a theoretically proven impossibility in the i.i.d. setting under a general nonlinear smooth map from factors to observations. In this work, we show that, remarkably, it is possible to recover discretized latent coordinates under the most general smooth mapping (diffeomorphism) without any additional inductive bias on the mapping. This is, provided the latent density has axis-aligned discontinuity landmarks, but without making the unrealistic assumption of statistical independence of the factors. We introduce this novel form of identifiability and provide a comprehensive proof of the recovery of discretized coordinates."
                    },
                    {
                        "title": "Additive Decoders for Latent Variables Identification and Cartesian-Product Extrapolation",
                        "abstract": "We tackle the problems of latent variables identification and ``out-of-support'' image generation in representation learning. We show that both are possible for a class of decoders that we call additive, which are reminiscent of decoders used for object-centric representation learning (OCRL) and well suited for images that can be decomposed as a sum of object-specific images. We provide conditions under which exactly solving the reconstruction problem using an additive decoder is guaranteed to identify the blocks of latent variables up to permutation and block-wise invertible transformations. This guarantee relies only on very weak assumptions about the distribution of the latent factors, which might present statistical dependencies and have an almost arbitrarily shaped support. Our result provides a new setting where nonlinear independent component analysis (ICA) is possible and adds to our theoretical understanding of OCRL methods. We also show theoretically that additive decoders can generate novel images by recombining observed factors of variations in novel ways, an ability we refer to as Cartesian-product extrapolation. We show empirically that additivity is crucial for both identifiability and extrapolation on simulated data."
                    },
                    {
                        "title": "On the Identifiability of Quantized Factors",
                        "abstract": "Disentanglement aims to recover meaningful latent ground-truth factors from the observed distribution solely, and is formalized through the theory of identifiability. The identifiability of independent latent factors is proven to be impossible in the unsupervised i.i.d. setting under a general nonlinear map from factors to observations. In this work, however, we demonstrate that it is possible to recover quantized latent factors under a generic nonlinear diffeomorphism. We only assume that the latent factors have independent discontinuities in their density, without requiring the factors to be statistically independent. We introduce this novel form of identifiability, termed quantized factor identifiability, and provide a comprehensive proof of the recovery of the quantized factors."
                    },
                    {
                        "title": "Data-Efficient Structured Pruning via Submodular Optimization",
                        "abstract": "Structured pruning is an effective approach for compressing large pre-trained neural networks without significantly affecting their performance. However, most current structured pruning methods do not provide any performance guarantees, and often require fine-tuning, which makes them inapplicable in the limited-data regime. We propose a principled data-efficient structured pruning method based on submodular optimization. In particular, for a given layer, we select neurons/channels to prune and corresponding new weights for the next layer, that minimize the change in the next layer's input induced by pruning. We show that this selection problem is a weakly submodular maximization problem, thus it can be provably approximated using an efficient greedy algorithm. Our method is guaranteed to have an exponentially decreasing error between the original model and the pruned model outputs w.r.t the pruned size, under reasonable assumptions. It is also one of the few methods in the literature that uses only a limited-number of training data and no labels. Our experimental results demonstrate that our method outperforms state-of-the-art methods in the limited-data regime."
                    },
                    {
                        "title": "Controlled Sparsity via Constrained Optimization or: How I Learned to Stop Tuning Penalties and Love Constraints",
                        "abstract": "The performance of trained neural networks is robust to harsh levels of pruning. Coupled with the ever-growing size of deep learning models, this observation has motivated extensive research on learning sparse models. In this work, we focus on the task of controlling the level of sparsity when performing sparse learning. Existing methods based on sparsity-inducing penalties involve expensive trial-and-error tuning of the penalty factor, thus lacking direct control of the resulting model sparsity. In response, we adopt a constrained formulation: using the gate mechanism proposed by Louizos et al. (2018), we formulate a constrained optimization problem where sparsification is guided by the training objective and the desired sparsity target in an end-to-end fashion. Experiments on CIFAR-{10, 100}, TinyImageNet, and ImageNet using WideResNet and ResNet{18, 50} models validate the effectiveness of our proposal and demonstrate that we can reliably achieve pre-determined sparsity targets without compromising on predictive performance."
                    },
                    {
                        "title": "Synergies Between Disentanglement and Sparsity: a Multi-Task Learning Perspective",
                        "abstract": "Although disentangled representations are often said to be beneficial for downstream tasks, current empirical and theoretical understanding is limited. In this work, we provide evidence that disentangled representations coupled with sparse base-predictors improve generalization. In the context of multi-task learning, we prove a new identifiability result that provides conditions under which maximally sparse base-predictors yield disentangled representations. Motivated by this theoretical result, we propose a practical approach to learn disentangled representations based on a sparsity-promoting bi-level optimization problem. Finally, we explore a meta-learning version of this algorithm based on group Lasso multiclass SVM base-predictors, for which we derive a tractable dual formulation. It obtains competitive results on standard few-shot classification benchmarks, while each task is using only a fraction of the learned representations."
                    },
                    {
                        "title": "Dynamics of SGD with Stochastic Polyak Stepsizes: Truly Adaptive Variants and Convergence to Exact Solution",
                        "abstract": "Recently, Loizou et al. (2021), proposed and analyzed stochastic gradient descent (SGD) with stochastic Polyak stepsize (SPS). The proposed SPS comes with strong convergence guarantees and competitive performance; however, it has two main drawbacks when it is used in non-over-parameterized regimes: (i) It requires a priori knowledge of the optimal mini-batch losses, which are not available when the interpolation condition is not satisfied (e.g., regularized objectives), and (ii) it guarantees convergence only to a neighborhood of the solution. In this work, we study the dynamics and the convergence properties of SGD equipped with new variants of the stochastic Polyak stepsize and provide solutions to both drawbacks of the original SPS. We first show that a simple modification of the original SPS that uses lower bounds instead of the optimal function values can directly solve issue (i). On the other hand, solving issue (ii) turns out to be more challenging and leads us to valuable insights into the method's behavior. We show that if interpolation is not satisfied, the correlation between SPS and stochastic gradients introduces a bias, which effectively distorts the expectation of the gradient signal near minimizers, leading to non-convergence - even if the stepsize is scaled down during training. To fix this issue, we propose DecSPS, a novel modification of SPS, which guarantees convergence to the exact minimizer - without a priori knowledge of the problem parameters. For strongly-convex optimization problems, DecSPS is the first stochastic adaptive optimization method that converges to the exact solution without restrictive assumptions like bounded iterates/gradients."
                    },
                    {
                        "title": "Partial Disentanglement via Mechanism Sparsity",
                        "abstract": "Disentanglement via mechanism sparsity was introduced recently as a principled approach to ex-tract latent factors without supervision when the causal graph relating them in time is sparse, and/or when actions are observed and affect them sparsely. However, this theory applies only to ground-truth graphs satisfying a specific criterion. In this work, we introduce a generalization of this theory which applies to any ground-truth graph and specifies qualitatively how disentangled the learned representation is expected to be, via a new equivalence relation over models we call consistency . This equivalence captures which factors are expected to remain entangled and which are not based on the specific form of the ground-truth graph. We call this weaker form of identifiability partial disentanglement . The graphical criterion that allows complete disentanglement, proposed in an earlier work, can be derived as a special case of our theory. Finally, we enforce graph sparsity with constrained optimization and illustrate our theory and algorithm in simulations."
                    },
                    {
                        "title": "CrossSplit: Mitigating Label Noise Memorization through Data Splitting",
                        "abstract": "We approach the problem of improving robustness of deep learning algorithms in the presence of label noise. Building upon existing label correction and co-teaching methods, we propose a novel training procedure to mitigate the memorization of noisy labels, called CrossSplit, which uses a pair of neural networks trained on two disjoint parts of the labelled dataset. CrossSplit combines two main ingredients: (i) Cross-split label correction. The idea is that, since the model trained on one part of the data cannot memorize example-label pairs from the other part, the training labels presented to each network can be smoothly adjusted by using the predictions of its peer network; (ii) Cross-split semi-supervised training. A network trained on one part of the data also uses the unlabeled inputs of the other part. Extensive experiments on CIFAR-10, CIFAR-100, Tiny-ImageNet and mini-WebVision datasets demonstrate that our method can outperform the current state-of-the-art in a wide range of noise ratios."
                    },
                    {
                        "title": "Bayesian Structure Learning with Generative Flow Networks",
                        "abstract": "In Bayesian structure learning, we are interested in inferring a distribution over the directed acyclic graph (DAG) structure of Bayesian networks, from data. Defining such a distribution is very challenging, due to the combinatorially large sample space, and approximations based on MCMC are often required. Recently, a novel class of probabilistic models, called Generative Flow Networks (GFlowNets), have been introduced as a general framework for generative modeling of discrete and composite objects, such as graphs. In this work, we propose to use a GFlowNet as an alternative to MCMC for approximating the posterior distribution over the structure of Bayesian networks, given a dataset of observations. Generating a sample DAG from this approximate distribution is viewed as a sequential decision problem, where the graph is constructed one edge at a time, based on learned transition probabilities. Through evaluation on both simulated and real data, we show that our approach, called DAG-GFlowNet, provides an accurate approximation of the posterior over DAGs, and it compares favorably against other methods based on MCMC or variational inference."
                    },
                    {
                        "title": "Synergies between Disentanglement and Sparsity: Generalization and Identifiability in Multi-Task Learning",
                        "abstract": "Although disentangled representations are often said to be beneficial for downstream tasks, current empirical and theoretical understanding is limited. In this work, we provide evidence that disentangled representations coupled with sparse base-predictors improve generalization. In the context of multi-task learning, we prove a new identifiability result that provides conditions under which maximally sparse base-predictors yield disentangled representations. Motivated by this theoretical result, we propose a practical approach to learn disentangled representations based on a sparsity-promoting bi-level optimization problem. Finally, we explore a meta-learning version of this algorithm based on group Lasso multiclass SVM base-predictors, for which we derive a tractable dual formulation. It obtains competitive results on standard few-shot classification benchmarks, while each task is using only a fraction of the learned representations."
                    },
                    {
                        "title": "Convergence Rates for the MAP of an Exponential Family and Stochastic Mirror Descent - an Open Problem",
                        "abstract": "We consider the problem of upper bounding the expected log-likelihood sub-optimality of the maximum likelihood estimate (MLE), or a conjugate maximum a posteriori (MAP) for an exponential family, in a non-asymptotic way. Surprisingly, we found no general solution to this problem in the literature. In particular, current theories do not hold for a Gaussian or in the interesting few samples regime. After exhibiting various facets of the problem, we show we can interpret the MAP as running stochastic mirror descent (SMD) on the log-likelihood. However, modern convergence results do not apply for standard examples of the exponential family, highlighting holes in the convergence literature. We believe solving this very fundamental problem may bring progress to both the statistics and optimization communities."
                    }
                ]
            },
            "00ff1051-3b07-4d24-a73e-f0061ea86145": {
                "pk": "00ff1051-3b07-4d24-a73e-f0061ea86145",
                "name": "Jose Gallego-Posada",
                "collaborators": [
                    "Simon Lacoste-Julien",
                    "Juan Ramirez",
                    "Akram Erraqabi",
                    "Ankit Vani",
                    "Max Schwarzer",
                    "Motahareh Sohrabi",
                    "Tianyue H. Zhang",
                    "Hao-Jun Michael Shi",
                    "Tsung-Hsien Lee",
                    "Shintaro Iwasaki"
                ],
                "domain": [
                    "Constrained Optimization",
                    "Neural Networks",
                    "Generative Models",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "On PI Controllers for Updating Lagrange Multipliers in Constrained Optimization",
                        "abstract": "Constrained optimization offers a powerful framework to prescribe desired behaviors in neural network models. Typically, constrained problems are solved via their min-max Lagrangian formulations, which exhibit unstable oscillatory dynamics when optimized using gradient descent-ascent. The adoption of constrained optimization techniques in the machine learning community is currently limited by the lack of reliable, general-purpose update schemes for the Lagrange multipliers. This paper proposes the $\\nu$PI algorithm and contributes an optimization perspective on Lagrange multiplier updates based on PI controllers, extending the work of Stooke, Achiam and Abbeel (2020). We provide theoretical and empirical insights explaining the inability of momentum methods to address the shortcomings of gradient descent-ascent, and contrast this with the empirical success of our proposed $\\nu$PI controller. Moreover, we prove that $\\nu$PI generalizes popular momentum methods for single-objective minimization. Our experiments demonstrate that $\\nu$PI reliably stabilizes the multiplier dynamics and its hyperparameters enjoy robust and predictable behavior."
                    },
                    {
                        "title": "A Distributed Data-Parallel PyTorch Implementation of the Distributed Shampoo Optimizer for Training Neural Networks At-Scale",
                        "abstract": "Shampoo is an online and stochastic optimization algorithm belonging to the AdaGrad family of methods for training neural networks. It constructs a block-diagonal preconditioner where each block consists of a coarse Kronecker product approximation to full-matrix AdaGrad for each parameter of the neural network. In this work, we provide a complete description of the algorithm as well as the performance optimizations that our implementation leverages to train deep networks at-scale in PyTorch. Our implementation enables fast multi-GPU distributed data-parallel training by distributing the memory and computation associated with blocks of each parameter via PyTorch's DTensor data structure and performing an AllGather primitive on the computed search directions at each iteration. This major performance enhancement enables us to achieve at most a 10% performance reduction in per-step wall-clock time compared against standard diagonal-scaling-based adaptive gradient methods. We validate our implementation by performing an ablation study on training ImageNet ResNet50, demonstrating Shampoo's superiority over standard training recipes with minimal hyperparameter tuning."
                    },
                    {
                        "title": "L$_0$onie: Compressing COINs with L$_0$-constraints",
                        "abstract": "Advances in Implicit Neural Representations (INR) have motivated research on domain-agnostic compression techniques. These methods train a neural network to approximate an object, and then store the weights of the trained model. For example, given an image, a network is trained to learn the mapping from pixel locations to RGB values. In this paper, we propose L 0 onie, a sparsity-constrained extension of the COIN compression method. Sparsity allows to leverage the faster learning of overparameterized networks, while retaining the desirable compression rate of smaller models. Moreover, our constrained formulation ensures that the \ufb01nal model respects a pre-determined compression rate, dispensing of the need for expensive architecture search."
                    },
                    {
                        "title": "Equivariant Mesh Attention Networks",
                        "abstract": "Equivariance to symmetries has proven to be a powerful inductive bias in deep learning research. Recent works on mesh processing have concentrated on various kinds of natural symmetries, including translations, rotations, scaling, node permutations, and gauge transformations. To date, no existing architecture is equivariant to all of these transformations. In this paper, we present an attention-based architecture for mesh data that is provably equivariant to all transformations mentioned above. Our pipeline relies on the use of relative tangential features: a simple, effective, equivariance-friendly alternative to raw node positions as inputs. Experiments on the FAUST and TOSCA datasets confirm that our proposed architecture achieves improved performance on these benchmarks and is indeed equivariant, and therefore robust, to a wide variety of local/global transformations."
                    },
                    {
                        "title": "Controlled Sparsity via Constrained Optimization or: How I Learned to Stop Tuning Penalties and Love Constraints",
                        "abstract": "The performance of trained neural networks is robust to harsh levels of pruning. Coupled with the ever-growing size of deep learning models, this observation has motivated extensive research on learning sparse models. In this work, we focus on the task of controlling the level of sparsity when performing sparse learning. Existing methods based on sparsity-inducing penalties involve expensive trial-and-error tuning of the penalty factor, thus lacking direct control of the resulting model sparsity. In response, we adopt a constrained formulation: using the gate mechanism proposed by Louizos et al. (2018), we formulate a constrained optimization problem where sparsification is guided by the training objective and the desired sparsity target in an end-to-end fashion. Experiments on CIFAR-{10, 100}, TinyImageNet, and ImageNet using WideResNet and ResNet{18, 50} models validate the effectiveness of our proposal and demonstrate that we can reliably achieve pre-determined sparsity targets without compromising on predictive performance."
                    },
                    {
                        "title": "Flexible Learning of Sparse Neural Networks via Constrained L0 Regularizations",
                        "abstract": "We propose to approach the problem of learning L 0 -sparse networks using a constrained formulation of the optimization problem. This is in contrast to commonly used penalized approaches, which combine the regularization terms additively with the (surrogate) empirical risk. Our experiments demonstrate that we can obtain approximate solutions to the constrained optimization problem with comparable performance to state-of-the art methods for L 0 -sparse training. Finally, we discuss how this constrained approach provides greater (hyper-)parameter interpretability and accountability from a practitioner\u2019s point of view"
                    },
                    {
                        "title": "S IMPLICIAL R EGULARIZATION",
                        "abstract": "Inspired by the fuzzy topological representation of a dataset employed in UMAP (McInnes et al., 2018), we propose a regularization principle for supervised learning based on the preservation of the simplicial complex structure of the data. We analyze the behavior of our proposal in contrast with the mixup (Zhang et al., 2018) framework on dimensionality reduction and classification tasks. Our experiments show how simplicial regularization induces more appropriate learning biases and alleviates some of the shortcomings of state-of-the art methods for regularization based on data augmentation. Simplicial Regularization mixup Parzen estimate Figure 1: Comparison of artificial samples generated from several vicinal distributions on MNIST. The richness of the class of neural networks as function approximators is well documented (Hornik, 1991; Cybenko, 1989; Telgarsky, 2015) and comes at the expense of having to regularize such functional families. Recent works by Zhang et al. (2018) and Verma et al. (2019) have studied the properties of regularization induced by considering convex combinations of (latent representations of) data samples. The authors argue that these techniques lead to learning \u201cflatter\u201d representations. The idea of using data augmentation as a form of regularization is not new. Chapelle et al. (2000) introduced the unifying framework of Vicinal Risk Minimization (VRM) in which the empirical risk objective RERM(f) is modified via the introduction of an improved density estimate or vicinal distribution, Pest(x, y). The vicinal distribution acts as a probabilistic model from which artificial samples are generated towards regularizing the model f ."
                    },
                    {
                        "title": "GAIT: A Geometric Approach to Information Theory",
                        "abstract": "We advocate the use of a notion of entropy that reflects the relative abundances of the symbols in an alphabet, as well as the similarities between them. This concept was originally introduced in theoretical ecology to study the diversity of ecosystems. Based on this notion of entropy, we introduce geometry-aware counterparts for several concepts and theorems in information theory. Notably, our proposed divergence exhibits performance on par with state-of-the-art methods based on the Wasserstein distance, but enjoys a closed-form expression that can be computed efficiently. We demonstrate the versatility of our method via experiments on a broad range of domains: training generative models, computing image barycenters, approximating empirical measures and counting modes."
                    },
                    {
                        "title": "GEAR: Geometry-Aware R\u00e9nyi Information",
                        "abstract": "Shannon's seminal theory of information has been of paramount importance in the development of modern machine learning techniques. However, standard information measures deal with probability distributions over an alphabet considered as a mere set of symbols and disregard further geometric structure, which might be available in the form of a metric or similarity function. We advocate the use of a notion of entropy that reflects not only the relative abundances of symbols but also the similarities between them, which was originally introduced in theoretical ecology to study the diversity of biological communities. Echoing this idea, we propose a criterion for comparing two probability distributions (possibly degenerate and with non-overlapping supports) that takes into account the geometry of the space in which the distributions are defined. Our proposal exhibits performance on par with state-of-the-art methods based on entropy-regularized optimal transport, but enjoys a closed-form expression and thus a lower computational cost. We demonstrate the versatility of our proposal via experiments on a broad range of domains: computing image barycenters, approximating densities with a collection of (super-) samples; summarizing texts; assessing mode coverage; as well as training generative models."
                    },
                    {
                        "title": "Internal analysis and optimization applied to parameter estimation under uncertainty",
                        "abstract": "We present a methodology through exemplification to perform parameter estimation subject to possible factors of uncertainty. The underlying optimization problem is posed in the framework of the theory of interval-valued optimization. The implementation of numerical procedures required to achieve efficient solutions implied the use of the $\\ell_1$ norm instead of usual $\\ell_2$ regression. Finally, an implementation using real data was performed, demonstrating the ability of interval analysis to encapsulate uncertainty while facing non-trivial parameter estimation problems."
                    },
                    {
                        "title": "GANGs: Generative Adversarial Network Games",
                        "abstract": "Generative Adversarial Networks (GAN) have become one of the most successful frameworks for unsupervised generative modeling. As GANs are difficult to train much research has focused on this. However, very little of this research has directly exploited game-theoretic techniques. We introduce Generative Adversarial Network Games (GANGs), which explicitly model a finite zero-sum game between a generator ($G$) and classifier ($C$) that use mixed strategies. The size of these games precludes exact solution methods, therefore we define resource-bounded best responses (RBBRs), and a resource-bounded Nash Equilibrium (RB-NE) as a pair of mixed strategies such that neither $G$ or $C$ can find a better RBBR. The RB-NE solution concept is richer than the notion of `local Nash equilibria' in that it captures not only failures of escaping local optima of gradient descent, but applies to any approximate best response computations, including methods with random restarts. To validate our approach, we solve GANGs with the Parallel Nash Memory algorithm, which provably monotonically converges to an RB-NE. We compare our results to standard GAN setups, and demonstrate that our method deals well with typical GAN problems such as mode collapse, partial mode coverage and forgetting."
                    },
                    {
                        "title": "Detection and Diagnosis of Breast Tumors using Deep Convolutional Neural Networks",
                        "abstract": "We present an application of deep Convolutional Neural Networks (CNN) for the detection and diagnosis of breast tumors. The images used in this study have been extracted from the mini-MIAS database of mammograms. The proposed system has been implemented in three stages: (a) crop, rotation and resize of the original mammogram; (b) feature extraction using a pretrained CNN model (AlexNet and VGG); (c) training of a Support Vector Machine (SVM) at the classification task using the previously extracted features. In this research, the goal of the system is to distinguish between three classes of patients: those with benign, malign or without tumor. Experiments show that feature extraction using pretrained models provides satisfactory results, achieving a 64.52% test accuracy. This outcome could be improved via fine-tuning of the final layers or training the whole network parameters. The results of additional experiments using a sample of the Caltech-101 database, for which a 99.38% test accuracy was obtained, exhibit the relevance of the similarity between the data used to train the model and the particular application intended. Additionally, it is worth noting the impact of the data augmentation process and the balance of the number of examples per class on the performance of the system. Keywords\u2014Breast tumor, classification, mammogram, convolutional neural network, support vector machine"
                    }
                ]
            }
        }
    },
    "2304.03768": {
        "paper_data": {
            "title": "SparseFormer: Sparse Visual Recognition via Limited Latent Tokens",
            "url": "http://arxiv.org/abs/2304.03768v1",
            "arxiv_id": "2304.03768",
            "authors": [
                "Ziteng Gao",
                "Zhan Tong",
                "Limin Wang",
                "Mike Zheng Shou"
            ],
            "abstract": "Human visual recognition is a sparse process, where only a few salient visual cues are attended to rather than traversing every detail uniformly. However, most current vision networks follow a dense paradigm, processing every single visual unit (e.g,, pixel or patch) in a uniform manner. In this paper, we challenge this dense paradigm and present a new method, coined SparseFormer, to imitate human's sparse visual recognition in an end-to-end manner. SparseFormer learns to represent images using a highly limited number of tokens (down to 49) in the latent space with sparse feature sampling procedure instead of processing dense units in the original pixel space. Therefore, SparseFormer circumvents most of dense operations on the image space and has much lower computational costs. Experiments on the ImageNet classification benchmark dataset show that SparseFormer achieves performance on par with canonical or well-established models while offering better accuracy-throughput tradeoff. Moreover, the design of our network can be easily extended to the video classification with promising performance at lower computational costs. We hope that our work can provide an alternative way for visual modeling and inspire further research on sparse neural architectures. The code will be publicly available at https://github.com/showlab/sparseformer",
            "introduction": " Introduction Designing neural architectures for visual recognition has long been an appealing yet challenging topic in the com- puter vision community. Convolutional neural networks (CNNs) [28, 18, 20, 42, 33] use convolutional \ufb01lters on ev- ery unit of images or feature maps to build features. Re- cently proposed vision transformers [11, 54, 32, 60] use attention operation on each patch or unit to dynamically interact with other units to mimic human attention mech- anism. Both convolution-based and Transformer-based ar- chitectures need to traverse every unit, like pixel or patch \u0000: Corresponding Author. oursparse paradigm latent space,limitedtokens \u2026 image space,denseunits denseparadigm traverse#: H\u00d7W#: N \u226aH\u00d7W\u2026 extractadjustFigure 1: Dense versus our proposed sparse paradigm for visual recognition. The dense paradigm requires traversing H\u0002Wunits to perform convolution or attention, while our proposed sparse paradigm performs transformers over only Nlatent tokens where N\u001cH\u0002W. on the grid map, to densely perform operations. This dense per-unit traversal originates from sliding windows [39], re- \ufb02ecting the assumption that foreground objects may appear uniformly with respect to spatial locations in an image. However, as humans, we do not need to examine every detail in a scene to recognize it. Instead, we are fast enough to \ufb01rst roughly \ufb01nd discriminative regions of interest with several glimpses and then recognize textures, edges, and high-level semantics within these regions [10, 37, 23, 46]. This contrasts greatly with existing visual networks, where the convention is to exhaustively traverse every visual unit. The dense paradigm incurs soaring computational costs with larger input resolutions and does not directly provide details about what a vision model is looking at in an image. In this paper, we propose a new vision architecture, coined SparseFormer , to explore sparse visual recognition by explicitly imitating the perception mechanism of human eyes. Speci\ufb01cally, SparseFormer learns to represent an im- age by latent transformers along with a highly limited num- ber of tokens ( e.g., down to 49) in the latent space from the very beginning . Each latent token is associated with a region of interest (RoI) descriptor, and the token RoI can be re\ufb01ned across stages. Given an image, SparseFormer \ufb01rst extracts image features by a lightweight early convolution module. A latent focusing transformer adjusts token RoIs to focus on foregrounds and sparsely extracts image features accord- ing to these token RoIs to build latent token embeddings in 1arXiv:2304.03768v1  [cs.CV]  7 Apr 2023an iterative manner. SparseFormer then feeds tokens with these region features into a larger and deeper network, i.e. a standard transformer encoder in the latent space, to enable precise recognition. The transformer operations are only applied over the limited tokens in the latent space. Since the number of latent tokens is highly limited (e.g.,49) and the feature sampling procedure is sparse (i.e., based on di- rect bilinear interpolation), it is reasonable to call our ar- chitecture a sparse approach for visual recognition. The overall computational cost of SparseFormer is almost irrel- evant to the input resolution except for the early convolution part, which is lightweight in our design. Moreover, Sparse- Former can be supervised solely with classi\ufb01cation signals in an end-to-end manner without requiring separate prior training with localizing signals. As an initial step to sparse visual recognition, the aim of SparseFormer is to explore an alternative paradigm for vi- sion modeling, rather than state-of-the-art results are shown in Table 8. segmentor GFLOPs mIoU mAcc Swin-T [32] + UperNet [65] 236 44.4 56.0 SF-T w/ 49 tokens 33 36.1 46.0 SF-T w/ 256 tokens 39 42.9",
            "references": [
                {
                    "title": "Token Merging: Your ViT But Faster",
                    "abstract": "We introduce Token Merging (ToMe), a simple method to increase the throughput of existing ViT models without needing to train. ToMe gradually combines similar tokens in a transformer using a general and light-weight matching algorithm that is as fast as pruning while being more accurate. Off-the-shelf, ToMe can 2x the throughput of state-of-the-art ViT-L @ 512 and ViT-H @ 518 models on images and 2.2x the throughput of ViT-L on video with only a 0.2-0.3% accuracy drop in each case. ToMe can also easily be applied during training, improving in practice training speed up to 2x for MAE fine-tuning on video. Training with ToMe further minimizes accuracy drop, leading to 2x the throughput of ViT-B on audio for only a 0.4% mAP drop. Qualitatively, we find that ToMe merges object parts into one token, even over multiple frames of video. Overall, ToMe's accuracy and speed are competitive with state-of-the-art on images, video, and audio."
                },
                {
                    "title": "MaxViT: Multi-Axis Vision Transformer",
                    "abstract": "Transformers have recently gained significant attention in the computer vision community. However, the lack of scalability of self-attention mechanisms with respect to image size has limited their wide adoption in state-of-the-art vision backbones. In this paper we introduce an efficient and scalable attention model we call multi-axis attention, which consists of two aspects: blocked local and dilated global attention. These design choices allow global-local spatial interactions on arbitrary input resolutions with only linear complexity. We also present a new architectural element by effectively blending our proposed attention model with convolutions, and accordingly propose a simple hierarchical vision backbone, dubbed MaxViT, by simply repeating the basic building block over multiple stages. Notably, MaxViT is able to ''see'' globally throughout the entire network, even in earlier, high-resolution stages. We demonstrate the effectiveness of our model on a broad spectrum of vision tasks. On image classification, MaxViT achieves state-of-the-art performance under various settings: without extra data, MaxViT attains 86.5% ImageNet-1K top-1 accuracy; with ImageNet-21K pre-training, our model achieves 88.7% top-1 accuracy. For downstream tasks, MaxViT as a backbone delivers favorable performance on object detection as well as visual aesthetic assessment. We also show that our proposed model expresses strong generative modeling capability on ImageNet, demonstrating the superior potential of MaxViT blocks as a universal vision module. The source code and trained models will be available at https://github.com/google-research/maxvit."
                },
                {
                    "title": "AdaMixer: A Fast-Converging Query-Based Object Detector",
                    "abstract": "Traditional object detectors employ the dense paradigm of scanning over locations and scales in an image. The recent query-based object detectors break this convention by decoding image features with a set of learnable queries. However, this paradigm still suffers from slow convergence, limited performance, and design complexity of extra networks between backbone and decoder. In this paper, we find that the key to these issues is the adaptability of decoders for casting queries to varying objects. Accordingly, we propose a fast-converging query-based detector, named AdaMixer, by improving the adaptability of query-based decoding processes in two aspects. First, each query adaptively samples features over space and scales based on estimated offsets, which allows AdaMixer to efficiently attend to the coherent regions of objects. Then, we dynamically decode these sampled features with an adaptive MLP-Mixer under the guidance of each query. Thanks to these two critical designs, AdaMixer enjoys architectural simplicity without requiring dense attentional encoders or explicit pyramid networks. On the challenging MS COCO benchmark, AdaMixer with ResNet-50 as the backbone, with 12 training epochs, reaches up to 45.0 AP on the validation set along with 27.9 APs in detecting small objects. With the longer training scheme, AdaMixer with ResNeXt-101-DCN and Swin-S reaches 49.5 and 51.3 AP. Our work sheds light on a simple, accurate, and fast converging architecture for query-based object detectors. The code is made available at https://github.com/MCG-NJU/AdaMixer."
                },
                {
                    "title": "Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations",
                    "abstract": "Vision Transformers (ViTs) take all the image patches as tokens and construct multi-head self-attention (MHSA) among them. Complete leverage of these image tokens brings redundant computations since not all the tokens are attentive in MHSA. Examples include that tokens containing semantically meaningless or distractive image backgrounds do not positively contribute to the ViT predictions. In this work, we propose to reorganize image tokens during the feed-forward process of ViT models, which is integrated into ViT during training. For each forward inference, we identify the attentive image tokens between MHSA and FFN (i.e., feed-forward network) modules, which is guided by the corresponding class token attention. Then, we reorganize image tokens by preserving attentive image tokens and fusing inattentive ones to expedite subsequent MHSA and FFN computations. To this end, our method EViT improves ViTs from two perspectives. First, under the same amount of input image tokens, our method reduces MHSA and FFN computation for efficient inference. For instance, the inference speed of DeiT-S is increased by 50% while its recognition accuracy is decreased by only 0.3% for ImageNet classification. Second, by maintaining the same computational cost, our method empowers ViTs to take more image tokens as input for recognition accuracy improvement, where the image tokens are from higher resolution images. An example is that we improve the recognition accuracy of DeiT-S by 1% for ImageNet classification at the same computational cost of a vanilla DeiT-S. Meanwhile, our method does not introduce more parameters to ViTs. Experiments on the standard benchmarks show the effectiveness of our method. The code is available at https://github.com/youweiliang/evit"
                },
                {
                    "title": "A ConvNet for the 2020s",
                    "abstract": "The \u201cRoaring 20s\u201d of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually \u201cmodernize\u201d a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets."
                },
                {
                    "title": "Vision Transformer with Deformable Attention",
                    "abstract": "Transformers have recently shown superior performances on various vision tasks. The large, sometimes even global, receptive field endows Transformer models with higher representation power over their CNN counterparts. Nevertheless, simply enlarging receptive field also gives rise to several concerns. On the one hand, using dense attention e.g., in ViT, leads to excessive memory and computational cost, and features can be influenced by irrelevant parts which are beyond the region of interests. On the other hand, the sparse attention adopted in PVT or Swin Transformer is data agnostic and may limit the ability to model long range relations. To mitigate these issues, we propose a novel deformable selfattention module, where the positions of key and value pairs in selfattention are selected in a data-dependent way. This flexible scheme enables the self-attention module to focus on relevant re-gions and capture more informative features. On this basis, we present Deformable Attention Transformer, a general backbone model with deformable attention for both image classification and dense prediction tasks. Extensive experi-ments show that our models achieve consistently improved results on comprehensive benchmarks. Code is available at https://github.com/LeapLabTHU/DAT."
                },
                {
                    "title": "A-ViT: Adaptive Tokens for Efficient Vision Transformer",
                    "abstract": "We introduce A - ViT, a method that adaptively adjusts the inference cost of vision transformer (ViT) for images of different complexity. A - ViT achieves this by automatically reducing the number of tokens in vision transformers that are processed in the network as inference proceeds. We refor-mulate Adaptive Computation Time (ACT [17]) for this task, extending halting to discard redundant spatial tokens. The appealing architectural properties of vision transformers enables our adaptive token reduction mechanism to speed up inference without modifying the network architecture or inference hardware. We demonstrate that A - ViT requires no extra parameters or sub-network for halting, as we base the learning of adaptive halting on the original network parameters. We further introduce distributional prior regularization that stabilizes training compared to prior ACT approaches. On the image classification task (ImageNet1K), we show that our proposed A - ViT yields high efficacy in filtering informative spatial features and cutting down on the overall compute. The proposed method improves the throughput of DeiT-Tiny by 62% and DeiT-Small by 38% with only 0.3% accuracy drop, outperforming prior art by a large margin."
                },
                {
                    "title": "ResNet strikes back: An improved training procedure in timm",
                    "abstract": "The influential Residual Networks designed by He et al. remain the gold-standard architecture in numerous scientific publications. They typically serve as the default architecture in studies, or as baselines when new architectures are proposed. Yet there has been significant progress on best practices for training neural networks since the inception of the ResNet architecture in 2015. Novel optimization&data-augmentation have increased the effectiveness of the training recipes. In this paper, we re-evaluate the performance of the vanilla ResNet-50 when trained with a procedure that integrates such advances. We share competitive training settings and pre-trained models in the timm open-source library, with the hope that they will serve as better baselines for future work. For instance, with our more demanding training setting, a vanilla ResNet-50 reaches 80.4% top-1 accuracy at resolution 224x224 on ImageNet-val without extra data or distillation. We also report the performance achieved with popular models with our training procedure."
                },
                {
                    "title": "PIT: Position-Invariant Transform for Cross-FoV Domain Adaptation",
                    "abstract": "Cross-domain object detection and semantic segmentation have witnessed impressive progress recently. Existing approaches mainly consider the domain shift resulting from external environments including the changes of background, illumination or weather, while distinct camera intrinsic parameters appear commonly in different domains and their influence for domain adaptation has been very rarely explored. In this paper, we observe that the Field of View (FoV) gap induces noticeable instance appearance differences between the source and target domains. We further discover that the FoV gap between two domains impairs domain adaptation performance under both the FoV-increasing (source FoV < target FoV) and FoV-decreasing cases. Motivated by the observations, we propose the Position-Invariant Transform (PIT) to better align images in different domains. We also introduce a reverse PIT for mapping the transformed/aligned images back to the original image space, and design a loss reweighting strategy to accelerate the training process. Our method can be easily plugged into existing cross-domain detection/segmentation frameworks, while bringing about negligible computational overhead. Extensive experiments demonstrate that our method can soundly boost the performance on both cross-domain object detection and segmentation for state-of-the-art techniques. Our code is available at https://github.com/sheepooo/PIT-Position-Invariant-Transform."
                },
                {
                    "title": "Perceiver IO: A General Architecture for Structured Inputs & Outputs",
                    "abstract": "A central goal of machine learning is the development of systems that can solve many problems in as many data domains as possible. Current architectures, however, cannot be applied beyond a small set of stereotyped settings, as they bake in domain&task assumptions or scale poorly to large inputs or outputs. In this work, we propose Perceiver IO, a general-purpose architecture that handles data from arbitrary settings while scaling linearly with the size of inputs and outputs. Our model augments the Perceiver with a flexible querying mechanism that enables outputs of various sizes and semantics, doing away with the need for task-specific architecture engineering. The same architecture achieves strong results on tasks spanning natural language and visual understanding, multi-task and multi-modal reasoning, and StarCraft II. As highlights, Perceiver IO outperforms a Transformer-based BERT baseline on the GLUE language benchmark despite removing input tokenization and achieves state-of-the-art performance on Sintel optical flow estimation with no explicit mechanisms for multiscale correspondence."
                },
                {
                    "title": "Early Convolutions Help Transformers See Better",
                    "abstract": "Vision transformer (ViT) models exhibit substandard optimizability. In particular, they are sensitive to the choice of optimizer (AdamW vs. SGD), optimizer hyperparameters, and training schedule length. In comparison, modern convolutional neural networks are easier to optimize. Why is this the case? In this work, we conjecture that the issue lies with the patchify stem of ViT models, which is implemented by a stride-p p*p convolution (p=16 by default) applied to the input image. This large-kernel plus large-stride convolution runs counter to typical design choices of convolutional layers in neural networks. To test whether this atypical design choice causes an issue, we analyze the optimization behavior of ViT models with their original patchify stem versus a simple counterpart where we replace the ViT stem by a small number of stacked stride-two 3*3 convolutions. While the vast majority of computation in the two ViT designs is identical, we find that this small change in early visual processing results in markedly different training behavior in terms of the sensitivity to optimization settings as well as the final model accuracy. Using a convolutional stem in ViT dramatically increases optimization stability and also improves peak performance (by ~1-2% top-1 accuracy on ImageNet-1k), while maintaining flops and runtime. The improvement can be observed across the wide spectrum of model complexities (from 1G to 36G flops) and dataset scales (from ImageNet-1k to ImageNet-21k). These findings lead us to recommend using a standard, lightweight convolutional stem for ViT models in this regime as a more robust architectural choice compared to the original ViT model design."
                },
                {
                    "title": "Video Swin Transformer",
                    "abstract": "The vision community is witnessing a modeling shift from CNNs to Transformers, where pure Transformer architectures have attained top accuracy on the major video recognition benchmarks. These video models are all built on Transformer layers that globally connect patches across the spatial and temporal dimensions. In this paper, we instead advocate an inductive bias of locality in video Transformers, which leads to a better speed-accuracy trade-off compared to previous approaches which compute self-attention globally even with spatial-temporal factorization. The locality of the proposed video architecture is realized by adapting the Swin Transformer designed for the image domain, while continuing to leverage the power of pre-trained image models. Our approach achieves state-of-the-art accuracy on a broad range of video recognition benchmarks, including on action recognition (84.9 top-l accuracy on Kinetics-400 and 85.9 top-l accuracy on Kinetics-600 with ~20\u00d7 less pre-training data and ~3\u00d7 smaller model size) and temporal modeling (69.6 top-l accuracy on Something-Something v2)."
                },
                {
                    "title": "DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification",
                    "abstract": "Attention is sparse in vision transformers. We observe the final prediction in vision transformers is only based on a subset of most informative tokens, which is sufficient for accurate image recognition. Based on this observation, we propose a dynamic token sparsification framework to prune redundant tokens progressively and dynamically based on the input. Specifically, we devise a lightweight prediction module to estimate the importance score of each token given the current features. The module is added to different layers to prune redundant tokens hierarchically. To optimize the prediction module in an end-to-end manner, we propose an attention masking strategy to differentiably prune a token by blocking its interactions with other tokens. Benefiting from the nature of self-attention, the unstructured sparse tokens are still hardware friendly, which makes our framework easy to achieve actual speed-up. By hierarchically pruning 66% of the input tokens, our method greatly reduces 31%~37% FLOPs and improves the throughput by over 40% while the drop of accuracy is within 0.5% for various vision transformers. Equipped with the dynamic token sparsification framework, DynamicViT models can achieve very competitive complexity/accuracy trade-offs compared to state-of-the-art CNNs and vision transformers on ImageNet. Code is available at https://github.com/raoyongming/DynamicViT"
                },
                {
                    "title": "Multiscale Vision Transformers",
                    "abstract": "We present Multiscale Vision Transformers (MViT) for video and image recognition, by connecting the seminal idea of multiscale feature hierarchies with transformer models. Multiscale Transformers have several channel-resolution scale stages. Starting from the input resolution and a small channel dimension, the stages hierarchically expand the channel capacity while reducing the spatial resolution. This creates a multiscale pyramid of features with early layers operating at high spatial resolution to model simple low-level visual information, and deeper layers at spatially coarse, but complex, high-dimensional features. We evaluate this fundamental architectural prior for modeling the dense nature of visual signals for a variety of video recognition tasks where it outperforms concurrent vision transformers that rely on large scale external pre-training and are 5-10\u00d7 more costly in computation and parameters. We further remove the temporal dimension and apply our model for image classification where it outperforms prior work on vision transformers. Code is available at: https://github.com/facebookresearch/SlowFast."
                },
                {
                    "title": "ImageNet-21K Pretraining for the Masses",
                    "abstract": "ImageNet-1K serves as the primary dataset for pretraining deep learning models for computer vision tasks. ImageNet-21K dataset, which is bigger and more diverse, is used less frequently for pretraining, mainly due to its complexity, low accessibility, and underestimation of its added value. This paper aims to close this gap, and make high-quality efficient pretraining on ImageNet-21K available for everyone. Via a dedicated preprocessing stage, utilization of WordNet hierarchical structure, and a novel training scheme called semantic softmax, we show that various models significantly benefit from ImageNet-21K pretraining on numerous datasets and tasks, including small mobile-oriented models. We also show that we outperform previous ImageNet-21K pretraining schemes for prominent new models like ViT and Mixer. Our proposed pretraining pipeline is efficient, accessible, and leads to SoTA reproducible results, from a publicly available dataset. The training code and pretrained models are available at: https://github.com/Alibaba-MIIL/ImageNet21K"
                },
                {
                    "title": "Multi-Scale Vision Longformer: A New Vision Transformer for High-Resolution Image Encoding",
                    "abstract": "This paper presents a new Vision Transformer (ViT) architecture Multi-Scale Vision Longformer, which significantly enhances the ViT of [12] for encoding high-resolution images using two techniques. The first is the multi-scale model structure, which provides image encodings at multiple scales with manageable computational cost. The second is the attention mechanism of Vision Long-former, which is a variant of Longformer [3], originally developed for natural language processing, and achieves a linear complexity w.r.t. the number of input tokens. A comprehensive empirical study shows that the new ViT significantly outperforms several strong baselines, including the existing ViT models and their ResNet counterparts, and the Pyramid Vision Transformer from a concurrent work [47], on a range of vision tasks, including image classification, object detection, and segmentation. The models and source code are released at https://github.com/microsoft/vision-longformer."
                },
                {
                    "title": "ViViT: A Video Vision Transformer",
                    "abstract": "We present pure-transformer based models for video classification, drawing upon the recent success of such models in image classification. Our model extracts spatiotemporal tokens from the input video, which are then encoded by a series of transformer layers. In order to handle the long sequences of tokens encountered in video, we propose several, efficient variants of our model which factorise the spatial- and temporal-dimensions of the input. Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets. We conduct thorough ablation studies, and achieve state-of-the-art results on multiple video classification benchmarks including Kinetics 400 and 600, Epic Kitchens, Something-Something v2 and Moments in Time, outperforming prior methods based on deep 3D convolutional networks."
                },
                {
                    "title": "Perceiver: General Perception with Iterative Attention",
                    "abstract": "Biological systems perceive the world by simultaneously processing high-dimensional inputs from modalities as diverse as vision, audition, touch, proprioception, etc. The perception models used in deep learning on the other hand are designed for individual modalities, often relying on domain-specific assumptions such as the local grid structures exploited by virtually all existing vision models. These priors introduce helpful inductive biases, but also lock models to individual modalities. In this paper we introduce the Perceiver - a model that builds upon Transformers and hence makes few architectural assumptions about the relationship between its inputs, but that also scales to hundreds of thousands of inputs, like ConvNets. The model leverages an asymmetric attention mechanism to iteratively distill inputs into a tight latent bottleneck, allowing it to scale to handle very large inputs. We show that this architecture is competitive with or outperforms strong, specialized models on classification tasks across various modalities: images, point clouds, audio, video, and video+audio. The Perceiver obtains performance comparable to ResNet-50 and ViT on ImageNet without 2D convolutions by directly attending to 50,000 pixels. It is also competitive in all modalities in AudioSet."
                },
                {
                    "title": "Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions",
                    "abstract": "Although convolutional neural networks (CNNs) have achieved great success in computer vision, this work investigates a simpler, convolution-free backbone network use-fid for many dense prediction tasks. Unlike the recently-proposed Vision Transformer (ViT) that was designed for image classification specifically, we introduce the Pyramid Vision Transformer (PVT), which overcomes the difficulties of porting Transformer to various dense prediction tasks. PVT has several merits compared to current state of the arts. (1) Different from ViT that typically yields low-resolution outputs and incurs high computational and memory costs, PVT not only can be trained on dense partitions of an image to achieve high output resolution, which is important for dense prediction, but also uses a progressive shrinking pyramid to reduce the computations of large feature maps. (2) PVT inherits the advantages of both CNN and Transformer, making it a unified backbone for various vision tasks without convolutions, where it can be used as a direct replacement for CNN backbones. (3) We validate PVT through extensive experiments, showing that it boosts the performance of many downstream tasks, including object detection, instance and semantic segmentation. For example, with a comparable number of parameters, PVT+RetinaNet achieves 40.4 AP on the COCO dataset, surpassing ResNet50+RetinNet (36.3 AP) by 4.1 absolute AP (see Figure 2). We hope that PVT could, serre as an alternative and useful backbone for pixel-level predictions and facilitate future research."
                },
                {
                    "title": "Is Space-Time Attention All You Need for Video Understanding?",
                    "abstract": "We present a convolution-free approach to video classification built exclusively on self-attention over space and time. Our method, named\"TimeSformer,\"adapts the standard Transformer architecture to video by enabling spatiotemporal feature learning directly from a sequence of frame-level patches. Our experimental study compares different self-attention schemes and suggests that\"divided attention,\"where temporal attention and spatial attention are separately applied within each block, leads to the best video classification accuracy among the design choices considered. Despite the radically new design, TimeSformer achieves state-of-the-art results on several action recognition benchmarks, including the best reported accuracy on Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks, our model is faster to train, it can achieve dramatically higher test efficiency (at a small drop in accuracy), and it can also be applied to much longer video clips (over one minute long). Code and models are available at: https://github.com/facebookresearch/TimeSformer."
                },
                {
                    "title": "Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNet",
                    "abstract": "Transformers, which are popular for language modeling, have been explored for solving vision tasks recently, e.g., the Vision Transformer (ViT) for image classification. The ViT model splits each image into a sequence of tokens with fixed length and then applies multiple Transformer layers to model their global relation for classification. However, ViT achieves inferior performance to CNNs when trained from scratch on a midsize dataset like ImageNet. We find it is because: 1) the simple tokenization of input images fails to model the important local structure such as edges and lines among neighboring pixels, leading to low training sample efficiency; 2) the redundant attention backbone design of ViT leads to limited feature richness for fixed computation budgets and limited training samples. To overcome such limitations, we propose a new Tokens-To-Token Vision Transformer (T2T-VTT), which incorporates 1) a layer-wise Tokens-to-Token (T2T) transformation to progressively structurize the image to tokens by recursively aggregating neighboring Tokens into one Token (Tokens-to-Token), such that local structure represented by surrounding tokens can be modeled and tokens length can be reduced; 2) an efficient backbone with a deep-narrow structure for vision transformer motivated by CNN architecture design after empirical study. Notably, T2T-ViT reduces the parameter count and MACs of vanilla ViT by half, while achieving more than 3.0% improvement when trained from scratch on ImageNet. It also outperforms ResNets and achieves comparable performance with MobileNets by directly training on ImageNet. For example, T2T-ViT with comparable size to ResNet50 (21.5M parameters) can achieve 83.3% top1 accuracy in image resolution 384x384 on ImageNet.1"
                },
                {
                    "title": "Training data-efficient image transformers & distillation through attention",
                    "abstract": "Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption. In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data. More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models."
                },
                {
                    "title": "Sparse R-CNN: End-to-End Object Detection with Learnable Proposals",
                    "abstract": "We present Sparse R-CNN, a purely sparse method for object detection in images. Existing works on object detection heavily rely on dense object candidates, such as k anchor boxes pre-defined on all grids of image feature map of size H \u00d7 W. In our method, however, a fixed sparse set of learned object proposals, total length of N, are provided to object recognition head to perform classification and location. By eliminating HWk (up to hundreds of thousands) hand-designed object candidates to N (e.g. 100) learnable proposals, Sparse R-CNN completely avoids all efforts related to object candidates design and many-to-one label assignment. More importantly, final predictions are directly output without non-maximum suppression post-procedure. Sparse R-CNN demonstrates accuracy, run-time and training convergence performance on par with the well-established detector baselines on the challenging COCO dataset, e.g., achieving 45.0 AP in standard 3\u00d7 training schedule and running at 22 fps using ResNet-50 FPN model. We hope our work could inspire re-thinking the convention of dense prior in object detectors. The code is available at: https://github.com/PeizeSun/SparseR-CNN."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Deformable DETR: Deformable Transformers for End-to-End Object Detection",
                    "abstract": "DETR has been recently proposed to eliminate the need for many hand-designed components in object detection while demonstrating good performance. However, it suffers from slow convergence and limited feature spatial resolution, due to the limitation of Transformer attention modules in processing image feature maps. To mitigate these issues, we proposed Deformable DETR, whose attention modules only attend to a small set of key sampling points around a reference. Deformable DETR can achieve better performance than DETR (especially on small objects) with 10$\\times$ less training epochs. Extensive experiments on the COCO benchmark demonstrate the effectiveness of our approach. Code shall be released."
                },
                {
                    "title": "Longformer: The Long-Document Transformer",
                    "abstract": "Transformer-based models are unable to process long sequences due to their self-attention operation, which scales quadratically with the sequence length. To address this limitation, we introduce the Longformer with an attention mechanism that scales linearly with sequence length, making it easy to process documents of thousands of tokens or longer. Longformer's attention mechanism is a drop-in replacement for the standard self-attention and combines a local windowed attention with a task motivated global attention. Following prior work on long-sequence transformers, we evaluate Longformer on character-level language modeling and achieve state-of-the-art results on text8 and enwik8. In contrast to most prior work, we also pretrain Longformer and finetune it on a variety of downstream tasks. Our pretrained Longformer consistently outperforms RoBERTa on long document tasks and sets new state-of-the-art results on WikiHop and TriviaQA. We finally introduce the Longformer-Encoder-Decoder (LED), a Longformer variant for supporting long document generative sequence-to-sequence tasks, and demonstrate its effectiveness on the arXiv summarization dataset."
                },
                {
                    "title": "Designing Network Design Spaces",
                    "abstract": "In this work, we present a new network design paradigm. Our goal is to help advance the understanding of network design and discover design principles that generalize across settings. Instead of focusing on designing individual network instances, we design network design spaces that parametrize populations of networks. The overall process is analogous to classic manual design of networks, but elevated to the design space level. Using our methodology we explore the structure aspect of network design and arrive at a low-dimensional design space consisting of simple, regular networks that we call RegNet. The core insight of the RegNet parametrization is surprisingly simple: widths and depths of good networks can be explained by a quantized linear function. We analyze the RegNet design space and arrive at interesting findings that do not match the current practice of network design. The RegNet design space provides simple and fast networks that work well across a wide range of flop regimes. Under comparable training settings and flops, the RegNet models outperform the popular EfficientNet models while being up to 5x faster on GPUs."
                },
                {
                    "title": "Reformer: The Efficient Transformer",
                    "abstract": "Large Transformer models routinely achieve state-of-the-art results on a number of tasks but training these models can be prohibitively costly, especially on long sequences. We introduce two techniques to improve the efficiency of Transformers. For one, we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O($L^2$) to O($L\\log L$), where $L$ is the length of the sequence. Furthermore, we use reversible residual layers instead of the standard residuals, which allows storing activations only once in the training process instead of $N$ times, where $N$ is the number of layers. The resulting model, the Reformer, performs on par with Transformer models while being much more memory-efficient and much faster on long sequences."
                },
                {
                    "title": "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks",
                    "abstract": "Convolutional Neural Networks (ConvNets) are commonly developed at a fixed resource budget, and then scaled up for better accuracy if more resources are available. In this paper, we systematically study model scaling and identify that carefully balancing network depth, width, and resolution can lead to better performance. Based on this observation, we propose a new scaling method that uniformly scales all dimensions of depth/width/resolution using a simple yet highly effective compound coefficient. We demonstrate the effectiveness of this method on scaling up MobileNets and ResNet. \nTo go even further, we use neural architecture search to design a new baseline network and scale it up to obtain a family of models, called EfficientNets, which achieve much better accuracy and efficiency than previous ConvNets. In particular, our EfficientNet-B7 achieves state-of-the-art 84.4% top-1 / 97.1% top-5 accuracy on ImageNet, while being 8.4x smaller and 6.1x faster on inference than the best existing ConvNet. Our EfficientNets also transfer well and achieve state-of-the-art accuracy on CIFAR-100 (91.7%), Flowers (98.8%), and 3 other transfer learning datasets, with an order of magnitude fewer parameters. Source code is at this https URL."
                },
                {
                    "title": "SlowFast Networks for Video Recognition",
                    "abstract": "We present SlowFast networks for video recognition. Our model involves (i) a Slow pathway, operating at low frame rate, to capture spatial semantics, and (ii) a Fast pathway, operating at high frame rate, to capture motion at fine temporal resolution. The Fast pathway can be made very lightweight by reducing its channel capacity, yet can learn useful temporal information for video recognition. Our models achieve strong performance for both action classification and detection in video, and large improvements are pin-pointed as contributions by our SlowFast concept. We report state-of-the-art accuracy on major video recognition benchmarks, Kinetics, Charades and AVA. Code has been made available at: https://github.com/facebookresearch/SlowFast."
                },
                {
                    "title": "A Closer Look at Spatiotemporal Convolutions for Action Recognition",
                    "abstract": "In this paper we discuss several forms of spatiotemporal convolutions for video analysis and study their effects on action recognition. Our motivation stems from the observation that 2D CNNs applied to individual frames of the video have remained solid performers in action recognition. In this work we empirically demonstrate the accuracy advantages of 3D CNNs over 2D CNNs within the framework of residual learning. Furthermore, we show that factorizing the 3D convolutional filters into separate spatial and temporal components yields significantly gains in accuracy. Our empirical study leads to the design of a new spatiotemporal convolutional block \"R(2+1)D\" which produces CNNs that achieve results comparable or superior to the state-of-the-art on Sports-1M, Kinetics, UCF101, and HMDB51."
                },
                {
                    "title": "Non-local Neural Networks",
                    "abstract": "Both convolutional and recurrent operations are building blocks that process one local neighborhood at a time. In this paper, we present non-local operations as a generic family of building blocks for capturing long-range dependencies. Inspired by the classical non-local means method [4] in computer vision, our non-local operation computes the response at a position as a weighted sum of the features at all positions. This building block can be plugged into many computer vision architectures. On the task of video classification, even without any bells and whistles, our nonlocal models can compete or outperform current competition winners on both Kinetics and Charades datasets. In static image recognition, our non-local models improve object detection/segmentation and pose estimation on the COCO suite of tasks. Code will be made available."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Learning Spatio-Temporal Representation with Pseudo-3D Residual Networks",
                    "abstract": "Convolutional Neural Networks (CNN) have been regarded as a powerful class of models for image recognition problems. Nevertheless, it is not trivial when utilizing a CNN for learning spatio-temporal video representation. A few studies have shown that performing 3D convolutions is a rewarding approach to capture both spatial and temporal dimensions in videos. However, the development of a very deep 3D CNN from scratch results in expensive computational cost and memory demand. A valid question is why not recycle off-the-shelf 2D networks for a 3D CNN. In this paper, we devise multiple variants of bottleneck building blocks in a residual learning framework by simulating 3 x 3 x 3 convolutions with 1 \u00d7 3 \u00d7 3 convolutional filters on spatial domain (equivalent to 2D CNN) plus 3 \u00d7 1 \u00d7 1 convolutions to construct temporal connections on adjacent feature maps in time. Furthermore, we propose a new architecture, named Pseudo-3D Residual Net (P3D ResNet), that exploits all the variants of blocks but composes each in different placement of ResNet, following the philosophy that enhancing structural diversity with going deep could improve the power of neural networks. Our P3D ResNet achieves clear improvements on Sports-1M video classification dataset against 3D CNN and frame-based 2D CNN by 5.3% and 1.8%, respectively. We further examine the generalization performance of video representation produced by our pre-trained P3D ResNet on five different benchmarks and three different tasks, demonstrating superior performances over several state-of-the-art techniques."
                },
                {
                    "title": "Revisiting Unreasonable Effectiveness of Data in Deep Learning Era",
                    "abstract": "The success of deep learning in vision can be attributed to: (a) models with high capacity; (b) increased computational power; and (c) availability of large-scale labeled data. Since 2012, there have been significant advances in representation capabilities of the models and computational capabilities of GPUs. But the size of the biggest dataset has surprisingly remained constant. What will happen if we increase the dataset size by 10 \u00d7 or 100 \u00d7 ? This paper takes a step towards clearing the clouds of mystery surrounding the relationship between \u2018enormous data\u2019 and visual deep learning. By exploiting the JFT-300M dataset which has more than 375M noisy labels for 300M images, we investigate how the performance of current vision tasks would change if this data was used for representation learning. Our paper delivers some surprising (and some expected) findings. First, we find that the performance on vision tasks increases logarithmically based on volume of training data size. Second, we show that representation learning (or pre-training) still holds a lot of promise. One can improve performance on many vision tasks by just training a better base model. Finally, as expected, we present new state-of-the-art results for different vision tasks including image classification, object detection, semantic segmentation and human pose estimation. Our sincere hope is that this inspires vision community to not undervalue the data and develop collective efforts in building larger datasets."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset",
                    "abstract": "The paucity of videos in current action classification datasets (UCF-101 and HMDB-51) has made it difficult to identify good video architectures, as most methods obtain similar performance on existing small-scale benchmarks. This paper re-evaluates state-of-the-art architectures in light of the new Kinetics Human Action Video dataset. Kinetics has two orders of magnitude more data, with 400 human action classes and over 400 clips per class, and is collected from realistic, challenging YouTube videos. We provide an analysis on how current architectures fare on the task of action classification on this dataset and how much performance improves on the smaller benchmark datasets after pre-training on Kinetics. We also introduce a new Two-Stream Inflated 3D ConvNet (I3D) that is based on 2D ConvNet inflation: filters and pooling kernels of very deep image classification ConvNets are expanded into 3D, making it possible to learn seamless spatio-temporal feature extractors from video while leveraging successful ImageNet architecture designs and even their parameters. We show that, after pre-training on Kinetics, I3D models considerably improve upon the state-of-the-art in action classification, reaching 80.2% on HMDB-51 and 97.9% on UCF-101."
                },
                {
                    "title": "Temporal Segment Networks for Action Recognition in Videos",
                    "abstract": "We present a general and flexible video-level framework for learning action models in videos. This method, called temporal segment network (TSN), aims to model long-range temporal structure with a new segment-based sampling and aggregation scheme. This unique design enables the TSN framework to efficiently learn action models by using the whole video. The learned models could be easily deployed for action recognition in both trimmed and untrimmed videos with simple average pooling and multi-scale temporal window integration, respectively. We also study a series of good practices for the implementation of the TSN framework given limited training samples. Our approach obtains the state-the-of-art performance on five challenging action recognition benchmarks: HMDB51 (71.0 percent), UCF101 (94.9 percent), THUMOS14 (80.1 percent), ActivityNet v1.2 (89.6 percent), and Kinetics400 (75.7 percent). In addition, using the proposed RGB difference as a simple motion representation, our method can still achieve competitive accuracy on UCF101 (91.0 percent) while running at 340 FPS. Furthermore, based on the proposed TSN framework, we won the video classification track at the ActivityNet challenge 2016 among 24 teams."
                },
                {
                    "title": "Densely Connected Convolutional Networks",
                    "abstract": "Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections&#x2014;one between each layer and its subsequent layer&#x2014;our network has L(L+1)/2 direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters. We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet). DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less memory and computation to achieve high performance. Code and pre-trained models are available at https://github.com/liuzhuang13/DenseNet."
                },
                {
                    "title": "SGDR: Stochastic Gradient Descent with Warm Restarts",
                    "abstract": "Restart techniques are common in gradient-free optimization to deal with multimodal functions. Partial warm restarts are also gaining popularity in gradient-based optimization to improve the rate of convergence in accelerated gradient schemes to deal with ill-conditioned functions. In this paper, we propose a simple warm restart technique for stochastic gradient descent to improve its anytime performance when training deep neural networks. We empirically study its performance on the CIFAR-10 and CIFAR-100 datasets, where we demonstrate new state-of-the-art results at 3.14% and 16.21%, respectively. We also demonstrate its advantages on a dataset of EEG recordings and on a downsampled version of the ImageNet dataset. Our source code is available at this https URL"
                },
                {
                    "title": "Layer Normalization",
                    "abstract": "Training state-of-the-art, deep neural networks is computationally expensive. One way to reduce the training time is to normalize the activities of the neurons. A recently introduced technique called batch normalization uses the distribution of the summed input to a neuron over a mini-batch of training cases to compute a mean and variance which are then used to normalize the summed input to that neuron on each training case. This significantly reduces the training time in feed-forward neural networks. However, the effect of batch normalization is dependent on the mini-batch size and it is not obvious how to apply it to recurrent neural networks. In this paper, we transpose batch normalization into layer normalization by computing the mean and variance used for normalization from all of the summed inputs to the neurons in a layer on a single training case. Like batch normalization, we also give each neuron its own adaptive bias and gain which are applied after the normalization but before the non-linearity. Unlike batch normalization, layer normalization performs exactly the same computation at training and test times. It is also straightforward to apply to recurrent neural networks by computing the normalization statistics separately at each time step. Layer normalization is very effective at stabilizing the hidden state dynamics in recurrent networks. Empirically, we show that layer normalization can substantially reduce the training time compared with previously published techniques."
                },
                {
                    "title": "Gaussian Error Linear Units (GELUs)",
                    "abstract": "We propose the Gaussian Error Linear Unit (GELU), a high-performing neural network activation function. The GELU activation function is $x\\Phi(x)$, where $\\Phi(x)$ the standard Gaussian cumulative distribution function. The GELU nonlinearity weights inputs by their value, rather than gates inputs by their sign as in ReLUs ($x\\mathbf{1}_{x>0}$). We perform an empirical evaluation of the GELU nonlinearity against the ReLU and ELU activations and find performance improvements across all considered computer vision, natural language processing, and speech tasks."
                },
                {
                    "title": "Dynamic Filter Networks",
                    "abstract": "In a traditional convolutional layer, the learned filters stay fixed after training. In contrast, we introduce a new framework, the Dynamic Filter Network, where filters are generated dynamically conditioned on an input. We show that this architecture is a powerful one, with increased flexibility thanks to its adaptive nature, yet without an excessive increase in the number of model parameters. A wide variety of filtering operation can be learned this way, including local spatial transformations, but also others like selective (de)blurring or adaptive feature extraction. Moreover, multiple such layers can be combined, e.g. in a recurrent architecture. We demonstrate the effectiveness of the dynamic filter network on the tasks of video and stereo prediction, and reach state-of-the-art performance on the moving MNIST dataset with a much smaller model. By visualizing the learned filters, we illustrate that the network has picked up flow information by only looking at unlabelled training data. This suggests that the network can be used to pretrain networks for various supervised tasks in an unsupervised way, like optical flow and depth estimation."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Rethinking the Inception Architecture for Computer Vision",
                    "abstract": "Convolutional networks are at the core of most state of-the-art computer vision solutions for a wide variety of tasks. Since 2014 very deep convolutional networks started to become mainstream, yielding substantial gains in various benchmarks. Although increased model size and computational cost tend to translate to immediate quality gains for most tasks (as long as enough labeled data is provided for training), computational efficiency and low parameter count are still enabling factors for various use cases such as mobile vision and big-data scenarios. Here we are exploring ways to scale up networks in ways that aim at utilizing the added computation as efficiently as possible by suitably factorized convolutions and aggressive regularization. We benchmark our methods on the ILSVRC 2012 classification challenge validation set demonstrate substantial gains over the state of the art: 21:2% top-1 and 5:6% top-5 error for single frame evaluation using a network with a computational cost of 5 billion multiply-adds per inference and with using less than 25 million parameters. With an ensemble of 4 models and multi-crop evaluation, we report 3:5% top-5 error and 17:3% top-1 error on the validation set and 3:6% top-5 error on the official test set."
                },
                {
                    "title": "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks",
                    "abstract": "State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available"
                },
                {
                    "title": "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift",
                    "abstract": "Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.82% top-5 test error, exceeding the accuracy of human raters."
                },
                {
                    "title": "Learning Spatiotemporal Features with 3D Convolutional Networks",
                    "abstract": "We propose a simple, yet effective approach for spatiotemporal feature learning using deep 3-dimensional convolutional networks (3D ConvNets) trained on a large scale supervised video dataset. Our findings are three-fold: 1) 3D ConvNets are more suitable for spatiotemporal feature learning compared to 2D ConvNets, 2) A homogeneous architecture with small 3x3x3 convolution kernels in all layers is among the best performing architectures for 3D ConvNets, and 3) Our learned features, namely C3D (Convolutional 3D), with a simple linear classifier outperform state-of-the-art methods on 4 different benchmarks and are comparable with current best methods on the other 2 benchmarks. In addition, the features are compact: achieving 52.8% accuracy on UCF101 dataset with only 10 dimensions and also very efficient to compute due to the fast inference of ConvNets. Finally, they are conceptually very simple and easy to train and use."
                },
                {
                    "title": "Going deeper with convolutions",
                    "abstract": "We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection."
                },
                {
                    "title": "Recurrent Models of Visual Attention",
                    "abstract": "Applying convolutional neural networks to large images is computationally expensive because the amount of computation scales linearly with the number of image pixels. We present a novel recurrent neural network model that is capable of extracting information from an image or video by adaptively selecting a sequence of regions or locations and only processing the selected regions at high resolution. Like convolutional neural networks, the proposed model has a degree of translation invariance built-in, but the amount of computation it performs can be controlled independently of the input image size. While the model is non-differentiable, it can be trained using reinforcement learning methods to learn task-specific policies. We evaluate our model on several image classification tasks, where it significantly outperforms a convolutional neural network baseline on cluttered images, and on a dynamic visual control problem, where it learns to track a simple object without an explicit training signal for doing so."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "Understanding the difficulty of training deep feedforward neural networks",
                    "abstract": "Whereas before 2006 it appears that deep multilayer neural networks were not successfully trained, since then several algorithms have been shown to successfully train them, with experimental results showing the superiority of deeper vs less deep architectures. All these experimental results were obtained with new initialization or training mechanisms. Our objective here is to understand better why standard gradient descent from random initialization is doing so poorly with deep neural networks, to better understand these recent relative successes and help design better algorithms in the future. We first observe the influence of the non-linear activations functions. We find that the logistic sigmoid activation is unsuited for deep networks with random initialization because of its mean value, which can drive especially the top hidden layer into saturation. Surprisingly, we find that saturated units can move out of saturation by themselves, albeit slowly, and explaining the plateaus sometimes seen when training neural networks. We find that a new non-linearity that saturates less can often be beneficial. Finally, we study how activations and gradients vary across layers and during training, with the idea that training may be more difficult when the singular values of the Jacobian associated with each layer are far from 1. Based on these considerations, we propose a new initialization scheme that brings substantially faster convergence. 1 Deep Neural Networks Deep learning methods aim at learning feature hierarchies with features from higher levels of the hierarchy formed by the composition of lower level features. They include Appearing in Proceedings of the 13 International Conference on Artificial Intelligence and Statistics (AISTATS) 2010, Chia Laguna Resort, Sardinia, Italy. Volume 9 of JMLR: WC Weston et al., 2008). Much attention has recently been devoted to them (see (Bengio, 2009) for a review), because of their theoretical appeal, inspiration from biology and human cognition, and because of empirical success in vision (Ranzato et al., 2007; Larochelle et al., 2007; Vincent et al., 2008) and natural language processing (NLP) (Collobert & Weston, 2008; Mnih & Hinton, 2009). Theoretical results reviewed and discussed by Bengio (2009), suggest that in order to learn the kind of complicated functions that can represent high-level abstractions (e.g. in vision, language, and other AI-level tasks), one may need deep architectures. Most of the recent experimental results with deep architecture are obtained with models that can be turned into deep supervised neural networks, but with initialization or training schemes different from the classical feedforward neural networks (Rumelhart et al., 1986). Why are these new algorithms working so much better than the standard random initialization and gradient-based optimization of a supervised training criterion? Part of the answer may be found in recent analyses of the effect of unsupervised pretraining (Erhan et al., 2009), showing that it acts as a regularizer that initializes the parameters in a \u201cbetter\u201d basin of attraction of the optimization procedure, corresponding to an apparent local minimum associated with better generalization. But earlier work (Bengio et al., 2007) had shown that even a purely supervised but greedy layer-wise procedure would give better results. So here instead of focusing on what unsupervised pre-training or semi-supervised criteria bring to deep architectures, we focus on analyzing what may be going wrong with good old (but deep) multilayer neural networks. Our analysis is driven by investigative experiments to monitor activations (watching for saturation of hidden units) and gradients, across layers and across training iterations. We also evaluate the effects on these of choices of activation function (with the idea that it might affect saturation) and initialization procedure (since unsupervised pretraining is a particular form of initialization and it has a drastic impact)."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Learning where to look",
                    "abstract": "Knowledge of object locations in direct manipulation interfaces improves performance by restricting visual search. This paper explores the effects of label representativeness and practice on the learning of object locations. Results showed that as label representativeness decreased, location knowledge increased. All participants were shown to have learned locations to some extent, however, even those using an interface which offered little benefit to doing so, indicating that such learning is pervasive."
                },
                {
                    "title": "Acceleration of stochastic approximation by averaging",
                    "abstract": "A new recursive algorithm of stochastic approximation type with the averaging of trajectories is investigated. Convergence with probability one is proved for a variety of classical optimization and identification problems. It is also demonstrated for these problems that the proposed algorithm achieves the highest possible rate of convergence."
                },
                {
                    "title": "Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper",
                    "abstract": "Vibratory power unit for vibrating conveyers and screens comprising an asynchronous polyphase motor, at least one pair of associated unbalanced masses disposed on the shaft of said motor, with the first mass of a pair of said unbalanced masses being rigidly fastened to said shaft and with said second mass of said pair being movably arranged relative to said first mass, means for controlling and regulating the conveying rate during conveyer operation by varying the rotational speed of said motor between predetermined minimum and maximum values, said second mass being movably outwardly by centrifugal force against the pressure of spring means, said spring means being prestressed in such a manner that said second mass is, at rotational motor speeds lower than said minimum speed, held in its initial position, and at motor speeds between said lower and upper values in positions which are radially offset with respect to the axis of said motor to an extent depending on the value of said rotational motor speed."
                },
                {
                    "title": "Remarks on Some Nonparametric Estimates of a Density Function",
                    "abstract": "1. Summary. This note discusses some aspects of the estimation of the density function of a univariate probability distribution. All estimates of the density function satisfying relatively mild conditions are shown to be biased. The asymp\u00ad totic mean square error of a particular class of estimates is evaluated."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "DynamicViT: Ef\ufb01cient Vision Transformers with Dynamic Token Sparsi\ufb01cation",
                    "abstract": "."
                },
                {
                    "title": "The Dynamic Representation of Scenes",
                    "abstract": "One of the more powerful impressions created by vision is that of a coherent, richly detailed world where everything is present simultaneously. Indeed, this impression is so compelling that we tend to ascribe these properties not only to the external world, but to our internal representations as well. But results from several recent experiments argue against this latter ascription. For example, changes in images of real-world scenes often go unnoticed when made during a saccade, flicker, blink, or movie cut. This \u201cchange blindness\u201d provides strong evidence against the idea that our brains contain a picture-like representation of the scene that is everywhere detailed and coherent.\r\nHow then do we represent a scene? It is argued here that focused attention provides spatiotemporal coherence for the stable representation of one object at a time. It is then argued that the allocation of attention can be co-ordinated to create a \u201cvirtual representation\u201d. In such a scheme, a stable object representation is formed whenever needed, making it appear to higher levels as if all objects in the scene are represented in detail simultaneously."
                },
                {
                    "title": "Neural mechanisms of selective visual attention.",
                    "abstract": "The two basic phenomena that define the problem of visual attention can be illustrated in a simple example. Consider the arrays shown in each panel of Figure 1. In a typical experiment, before the arrays were presented, subjects would be asked to report letters appearing in one color (targets, here black letters), and to disregard letters in the other color (nontargets, here white letters). The array would then be briefly flashed, and the subjects, without any opportunity for eye movements, would give their report. The display mimics our. usual cluttered visual environment: It contains one or more objects that are relevant to current behavior, along with others that are irrelevant. The first basic phenomenon is limited capacity for processing information. At any given time, only a small amount of the information available on the retina can be processed and used in the control of behavior. Subjectively, giving attention to any one target leaves less available for others. In Figure 1, the probability of reporting the target letter N is much lower with two accompa\u00ad nying targets (Figure la) than with none (Figure Ib). The second basic phenomenon is selectivity-the ability to filter out un\u00ad wanted information. Subjectively, one is aware of attended stimuli and largely unaware of unattended ones. Correspondingly, accuracy in identifying an attended stimulus may be independent of the number of nontargets in a display (Figure la vs Ie) (see Bundesen 1990, Duncan 1980)."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we design a more efficient visual recognition architecture that mimics human perception by utilizing a sparse representation of image features instead of the traditional dense traversal methods?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the field of computer vision, as it could lead to significant reductions in computational costs and improved efficiency in processing high-resolution images. By developing a model that operates on a limited number of tokens, we can enhance our understanding of human-like perception in machines, potentially leading to more intuitive and effective visual recognition systems. This research could pave the way for practical applications in areas such as autonomous driving, medical imaging, and real-time video analysis, where quick and accurate recognition is essential.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in effectively identifying and focusing on the most relevant regions of an image without exhaustively processing every pixel or patch. Naive approaches may fail because they do not account for the human-like ability to quickly discern important features, leading to inefficient processing and high computational costs. Additionally, developing a robust mechanism for dynamically adjusting the regions of interest while maintaining accuracy poses significant technical and theoretical obstacles, particularly in ensuring that the model can generalize well across diverse visual contexts.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on dense paradigms, which assume uniform distribution of objects in images and rely on exhaustive traversal methods. This has created a gap in exploring sparse representations that align more closely with human visual perception. Barriers such as the lack of effective algorithms for identifying and refining regions of interest, as well as the absence of models that can operate efficiently with a limited number of tokens, have prevented this problem from being adequately addressed. Our approach differs by introducing the SparseFormer architecture, which explicitly learns to represent images through a limited number of latent tokens, thereby overcoming the limitations of prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nThe proposed methodology involves the SparseFormer architecture, which utilizes a lightweight early convolution module to extract image features, followed by a latent focusing transformer that adjusts token regions of interest (RoIs) to focus on foreground elements. The model operates on a limited number of tokens (e.g., 49) in the latent space, allowing for sparse feature extraction and recognition. The expected outcomes include a significant reduction in computational costs while maintaining competitive recognition performance, as evidenced by the architecture's ability to achieve"
            }
        },
        "author_data": {
            "e32d88b9-b93d-45c1-89d5-f552ff76ce3b": {
                "pk": "e32d88b9-b93d-45c1-89d5-f552ff76ce3b",
                "name": "Ziteng Gao",
                "collaborators": [
                    "Gangshan Wu",
                    "Limin Wang",
                    "Tao Wu",
                    "Mengqing Cao",
                    "Mengqi Cao",
                    "Liming Wang",
                    "Bing Han",
                    "Sheng Guo",
                    "Yin Wang",
                    "Xuyang Xu",
                    "Xi Shi"
                ],
                "domain": [
                    "Video Action Detection",
                    "Object Detection",
                    "Deep Learning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "STMixer: A One-Stage Sparse Action Detector",
                        "abstract": "Traditional video action detectors typically adopt the two-stage pipeline, where a person detector is first employed to generate actor boxes and then 3D RoIAlign is used to extract actor-specific features for action recognition. This detection paradigm requires multi-stage training and inference, and the feature sampling is only constrained inside the box, failing to effectively leverage richer context information outside. Recently, several query-based action detectors have been proposed to predict action instances in an end-to-end manner. However, they still lack adaptability in feature sampling and decoding, thus suffering from the issues of inferior performance or slower convergence. In this paper, we propose two core designs for a more flexible one-stage sparse action detector. First, we present a query-based adaptive feature sampling module, which endows the detector with the flexibility of mining a group of discriminative features from the entire spatio-temporal domain. Second, we devise a decoupled feature mixing module, which dynamically attends to and mixes video features along the spatial and temporal dimensions respectively for better feature decoding. Based on these designs, we instantiate two detection pipelines, that is, STMixer-K for keyframe action detection and STMixer-T for action tubelet detection. Without bells and whistles, our STMixer detectors obtain the state-of-the-art results on five challenging spatio-temporal action detection benchmarks for keyframe action detection or action tube detection."
                    },
                    {
                        "title": "STMixer: A One-Stage Sparse Action Detector Supplementary Material",
                        "abstract": "We provide detailed comparisons of the performance of STMixer and former state-of-the-art TubeR [10] on each action class of AVA v2.2 [4] in Figure 1. For a fair comparison, both STMixer and TubeR models use the CSN-152 backbone [8] and do not use long-term features. Out of all 60 classes, our method achieves higher AP on 47 classes, which makes the overall detection mAP of our STMixer higher by 1.7 than TubeR. We observe significant performance gaps on some interaction-related classes, such as interactions with objects (drive (e.g. a car, a truck) +12.9, shoot +12.0), and interactions with other people (sing to (e.g. self, a person, a group) +9.9, take (an object) from (a person) +3.8), which indicates our STMixer is more capable of modeling the relationship between the action performer and the surrounding objects and people."
                    },
                    {
                        "title": "STMixer: A One-Stage Sparse Action Detector",
                        "abstract": "Traditional video action detectors typically adopt the two-stage pipeline, where a person detector is first employed to generate actor boxes and then 3D RoIAlign is used to extract actor-specific features for classification. This detection paradigm requires multi-stage training and inference, and cannot capture context information outside the bounding box. Recently, a few query-based action detectors are proposed to predict action instances in an end-to-end manner. However, they still lack adaptability in feature sampling and decoding, thus suffering from the issues of inferior performance or slower convergence. In this paper, we propose a new one-stage sparse action detector, termed STMixer. STMixer is based on two core designs. First, we present a query-based adaptive feature sampling module, which endows our STMixer with the flexibility of mining a set of discriminative features from the entire spatiotemporal domain. Second, we devise a dual-branch feature mixing module, which allows our STMixer to dynamically attend to and mix video features along the spatial and the temporal dimension respectively for better feature decoding. Coupling these two designs with a video backbone yields an efficient end-to-end action detector. Without bells and whistles, our STMixer obtains the state-of-the-art results on the datasets of AVA, UCF101-24, and JHMDB."
                    },
                    {
                        "title": "AdaMixer: A Fast-Converging Query-Based Object Detector",
                        "abstract": "Traditional object detectors employ the dense paradigm of scanning over locations and scales in an image. The recent query-based object detectors break this convention by decoding image features with a set of learnable queries. However, this paradigm still suffers from slow convergence, limited performance, and design complexity of extra networks between backbone and decoder. In this paper, we find that the key to these issues is the adaptability of decoders for casting queries to varying objects. Accordingly, we propose a fast-converging query-based detector, named AdaMixer, by improving the adaptability of query-based decoding processes in two aspects. First, each query adaptively samples features over space and scales based on estimated offsets, which allows AdaMixer to efficiently attend to the coherent regions of objects. Then, we dynamically decode these sampled features with an adaptive MLP-Mixer under the guidance of each query. Thanks to these two critical designs, AdaMixer enjoys architectural simplicity without requiring dense attentional encoders or explicit pyramid networks. On the challenging MS COCO benchmark, AdaMixer with ResNet-50 as the backbone, with 12 training epochs, reaches up to 45.0 AP on the validation set along with 27.9 APs in detecting small objects. With the longer training scheme, AdaMixer with ResNeXt-101-DCN and Swin-S reaches 49.5 and 51.3 AP. Our work sheds light on a simple, accurate, and fast converging architecture for query-based object detectors. The code is made available at https://github.com/MCG-NJU/AdaMixer."
                    },
                    {
                        "title": "A Prior Probability of Speaker Information and Emojis Embedding Approach to Sarcasm Detection",
                        "abstract": "Sarcasm language is widely presented in social media and student chats. This is not only a difficult sentiment analysis task, but also the basis for completing other follow-up tasks. However, most of the previous methods are based on complex network model design, which leads to poor traceability of inference results. Or based on the non-corresponding pre-training model, which is not taking advantage of the area where each model can do better. We propose a DialoGPT model based on pre-training on the sarcasm task, combined with the information of the sarcasm probability about the speaker, and embedded emojis in sentences to enhance recognition. The algorithm we designed is tested on the dataset named sarcasm on reddit and compared with other related methods and systems. The results show that the proposed method achieves the most advanced performance in the four indicators in basically the same time. F1-score has achieved 77, which is better than other methods, and it uses the advantages of new information content. In the field of education, the method we proposed can be used to help understand and judge students' intention and personality."
                    },
                    {
                        "title": "Mutual Supervision for Dense Object Detection",
                        "abstract": "The classification and regression head are both indispensable components to build up a dense object detector, which are usually supervised by the same training samples and thus expected to have consistency with each other for detecting objects accurately in the detection pipeline. In this paper, we break the convention of the same training samples for these two heads in dense detectors and explore a novel supervisory paradigm, termed as Mutual Supervision (MuSu), to respectively and mutually assign training samples for the classification and regression head to ensure this consistency. MuSu defines training samples for the regression head mainly based on classification predicting scores and in turn, defines samples for the classification head based on localization scores from the regression head. Experimental results show that the convergence of detectors trained by this mutual supervision is guaranteed and the effectiveness of the proposed method is verified on the challenging MS COCO benchmark. We also find that tiling more anchors at the same location benefits detectors and leads to further improvements under this training scheme. We hope this work can inspire further researches on the interaction of the classification and regression task in detection and the supervision paradigm for detectors, especially separately for these two heads."
                    }
                ]
            },
            "94ae3c22-2971-416b-87a9-05c7461e7f1b": {
                "pk": "94ae3c22-2971-416b-87a9-05c7461e7f1b",
                "name": "Zhan Tong",
                "collaborators": [
                    "Yibing Song",
                    "Limin Wang",
                    "Chongjian Ge",
                    "Jiangliu Wang",
                    "Gangshan Wu",
                    "Ping Luo",
                    "Lei Chen",
                    "Shoufa Chen",
                    "Jue Wang",
                    "Xiaohan Ding",
                    "Li Yuan",
                    "Jianbo Jiao",
                    "Stephen James",
                    "P. Abbeel",
                    "Yunhui Liu",
                    "Bingkun Huang",
                    "Zhiyu Zhao",
                    "Yinan He",
                    "Yi Wang",
                    "Yali Wang",
                    "Yu Qiao",
                    "Yixiao Ge",
                    "Xihui Liu",
                    "Shutao Xia",
                    "Ying Shan",
                    "Youwei Liang",
                    "P. Xie",
                    "Yuan Zhi"
                ],
                "domain": [
                    "Computer Vision",
                    "Video Action Recognition",
                    "Self-Supervised Learning",
                    "Transformers"
                ],
                "publications": [
                    {
                        "title": "Advancing Vision Transformers with Group-Mix Attention",
                        "abstract": "Vision Transformers (ViTs) have been shown to enhance visual recognition through modeling long-range dependencies with multi-head self-attention (MHSA), which is typically formulated as Query-Key-Value computation. However, the attention map generated from the Query and Key captures only token-to-token correlations at one single granularity. In this paper, we argue that self-attention should have a more comprehensive mechanism to capture correlations among tokens and groups (i.e., multiple adjacent tokens) for higher representational capacity. Thereby, we propose Group-Mix Attention (GMA) as an advanced replacement for traditional self-attention, which can simultaneously capture token-to-token, token-to-group, and group-to-group correlations with various group sizes. To this end, GMA splits the Query, Key, and Value into segments uniformly and performs different group aggregations to generate group proxies. The attention map is computed based on the mixtures of tokens and group proxies and used to re-combine the tokens and groups in Value. Based on GMA, we introduce a powerful backbone, namely GroupMixFormer, which achieves state-of-the-art performance in image classification, object detection, and semantic segmentation with fewer parameters than existing models. For instance, GroupMixFormer-L (with 70.3M parameters and 384^2 input) attains 86.2% Top-1 accuracy on ImageNet-1K without external data, while GroupMixFormer-B (with 45.8M parameters) attains 51.2% mIoU on ADE20K."
                    },
                    {
                        "title": "Speed Co-Augmentation for Unsupervised Audio-Visual Pre-training",
                        "abstract": "This work aims to improve unsupervised audio-visual pre-training. Inspired by the efficacy of data augmentation in visual contrastive learning, we propose a novel speed co-augmentation method that randomly changes the playback speeds of both audio and video data. Despite its simplicity, the speed co-augmentation method possesses two compelling attributes: (1) it increases the diversity of audio-visual pairs and doubles the size of negative pairs, resulting in a significant enhancement in the learned representations, and (2) it changes the strict correlation between audio-visual pairs but introduces a partial relationship between the augmented pairs, which is modeled by our proposed SoftInfoNCE loss to further boost the performance. Experimental results show that the proposed method significantly improves the learned representations when compared to vanilla audio-visual contrastive learning."
                    },
                    {
                        "title": "CycleACR: Cycle Modeling of Actor-Context Relations for Video Action Detection",
                        "abstract": "The relation modeling between actors and scene context advances video action detection where the correlation of multiple actors makes their action recognition challenging. Existing studies model each actor and scene relation to improve action recognition. However, the scene variations and background interference limit the effectiveness of this relation modeling. In this paper, we propose to select actor-related scene context, rather than directly leverage raw video scenario, to improve relation modeling. We develop a Cycle Actor-Context Relation network (CycleACR) where there is a symmetric graph that models the actor and context relations in a bidirectional form. Our CycleACR consists of the Actor-to-Context Reorganization (A2C-R) that collects actor features for context feature reorganizations, and the Context-to-Actor Enhancement (C2A-E) that dynamically utilizes reorganized context features for actor feature enhancement. Compared to existing designs that focus on C2A-E, our CycleACR introduces A2C-R for a more effective relation modeling. This modeling advances our CycleACR to achieve state-of-the-art performance on two popular action detection datasets (i.e., AVA and UCF101-24). We also provide ablation studies and visualizations as well to show how our cycle actor-context relation modeling improves video action detection. Code is available at https://github.com/MCG-NJU/CycleACR."
                    },
                    {
                        "title": "VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking",
                        "abstract": "Scale is the primary factor for building a powerful foundation model that could well generalize to a variety of downstream tasks. However, it is still challenging to train video foundation models with billions of parameters. This paper shows that video masked autoencoder (VideoMAE) is a scalable and general self-supervised pre-trainer for building video foundation models. We scale the VideoMAE in both model and data with a core design. Specifically, we present a dual masking strategy for efficient pre-training, with an encoder operating on a subset of video tokens and a decoder processing another subset of video tokens. Although VideoMAE is very efficient due to high masking ratio in encoder, masking decoder can still further reduce the overall computational cost. This enables the efficient pre-training of billion-level models in video. We also use a progressive training paradigm that involves an initial pre-training on a diverse multi-sourced unlabeled dataset, followed by a post-pre-training on a mixed labeled dataset. Finally, we successfully train a video ViT model with a billion parameters, which achieves a new state-of-the-art performance on the datasets of Kinetics (90.0% on K400 and 89.9% on K600) and Something-Something (68.7% on V1 and 77.0% on V2). In addition, we extensively verify the pre-trained video ViT models on a variety of downstream tasks, demonstrating its effectiveness as a general video representation learner."
                    },
                    {
                        "title": "TVTSv2: Learning Out-of-the-box Spatiotemporal Visual Representations at Scale",
                        "abstract": "The ultimate goal for foundation models is realizing task-agnostic, i.e., supporting out-of-the-box usage without task-specific fine-tuning. Although breakthroughs have been made in natural language processing and image representation learning, it is still challenging for video models to reach it due to the increasing uncertainty of spatiotemporal signals. To ease training, existing works leverage image foundation models' prior knowledge and equip them with efficient temporal modules. Despite the satisfactory fine-tuning performance, we empirically find they fall short of out-of-the-box usage, given the even degraded performance in zero-shot/linear protocols compared to their baseline counterparts. In this work, we analyze the factor that leads to degradation from the perspective of language supervision distortion. We argue that tuning a text encoder end-to-end, as done in previous work, is suboptimal since it may overfit in terms of styles, thereby losing its original generalization ability to capture the semantics of various language registers. The overfitted text encoder, in turn, provides a harmful supervision signal, degrading the video representation. To tackle this issue, we propose a degradation-free pre-training strategy to retain the generalization ability of the text encoder via freezing shallow layers while enabling the task-related semantics capturing in tunable deep layers. As for the training objective, we adopted the transcript sorting task in TVTS incorporated with masking techniques to enable scalable training. As a result, we produce a series of models, dubbed TVTSv2, with up to one billion parameters. We achieve new state-of-the-arts on various video benchmarks with a frozen backbone, surpassing the recent ImageBind, InternVideo, etc. Code is available at https://github.com/TencentARC/TVTS."
                    },
                    {
                        "title": "Soft Neighbors are Positive Supporters in Contrastive Visual Representation Learning",
                        "abstract": "Contrastive learning methods train visual encoders by comparing views from one instance to others. Typically, the views created from one instance are set as positive, while views from other instances are negative. This binary instance discrimination is studied extensively to improve feature representations in self-supervised learning. In this paper, we rethink the instance discrimination framework and find the binary instance labeling insufficient to measure correlations between different samples. For an intuitive example, given a random image instance, there may exist other images in a mini-batch whose content meanings are the same (i.e., belonging to the same category) or partially related (i.e., belonging to a similar category). How to treat the images that correlate similarly to the current image instance leaves an unexplored problem. We thus propose to support the current image by exploring other correlated instances (i.e., soft neighbors). We first carefully cultivate a candidate neighbor set, which will be further utilized to explore the highly-correlated instances. A cross-attention module is then introduced to predict the correlation score (denoted as positiveness) of other correlated instances with respect to the current one. The positiveness score quantitatively measures the positive support from each correlated instance, and is encoded into the objective for pretext training. To this end, our proposed method benefits in discriminating uncorrelated instances while absorbing correlated instances for SSL. We evaluate our soft neighbor contrastive learning method (SNCLR) on standard visual recognition benchmarks, including image classification, object detection, and instance segmentation. The state-of-the-art recognition performance shows that SNCLR is effective in improving feature representations from both ViT and CNN encoders."
                    },
                    {
                        "title": "Efficient Video Action Detection with Token Dropout and Context Refinement",
                        "abstract": "Streaming video clips with large-scale video tokens impede vision transformers (ViTs) for efficient recognition, especially in video action detection where sufficient spatiotemporal representations are required for precise actor identification. In this work, we propose an end-to-end framework for efficient video action detection (EVAD) based on vanilla ViTs. Our EVAD consists of two specialized designs for video action detection. First, we propose a spatiotemporal token dropout from a keyframe-centric perspective. In a video clip, we maintain all tokens from its keyframe, preserve tokens relevant to actor motions from other frames, and drop out the remaining tokens in this clip. Second, we refine scene context by leveraging remaining tokens for better recognizing actor identities. The region of interest (RoI) in our action detector is expanded into temporal domain. The captured spatiotemporal actor identity representations are refined via scene context in a decoder with the attention mechanism. These two designs make our EVAD efficient while maintaining accuracy, which is validated on three benchmark datasets (i.e., AVA, UCF101-24, JHMDB). Compared to the vanilla ViT backbone, our EVAD reduces the overall GFLOPs by 43% and improves real-time inference speed by 40% with no performance degradation. Moreover, even at similar computational costs, our EVAD can improve the performance by 1.1 mAP with higher resolution inputs. Code is available at https://github.com/MCG-NJU/EVAD."
                    },
                    {
                        "title": "AdaptFormer: Adapting Vision Transformers for Scalable Visual Recognition",
                        "abstract": "Pretraining Vision Transformers (ViTs) has achieved great success in visual recognition. A following scenario is to adapt a ViT to various image and video recognition tasks. The adaptation is challenging because of heavy computation and memory storage. Each model needs an independent and complete finetuning process to adapt to different tasks, which limits its transferability to different visual domains. To address this challenge, we propose an effective adaptation approach for Transformer, namely AdaptFormer, which can adapt the pre-trained ViTs into many different image and video tasks efficiently. It possesses several benefits more appealing than prior arts. Firstly, AdaptFormer introduces lightweight modules that only add less than 2% extra parameters to a ViT, while it is able to increase the ViT's transferability without updating its original pre-trained parameters, significantly outperforming the existing 100\\% fully fine-tuned models on action recognition benchmarks. Secondly, it can be plug-and-play in different Transformers and scalable to many visual tasks. Thirdly, extensive experiments on five image and video datasets show that AdaptFormer largely improves ViTs in the target domains. For example, when updating just 1.5% extra parameters, it achieves about 10% and 19% relative improvement compared to the fully fine-tuned models on Something-Something~v2 and HMDB51, respectively. Code is available at https://github.com/ShoufaChen/AdaptFormer."
                    },
                    {
                        "title": "VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training",
                        "abstract": "Pre-training video transformers on extra large-scale datasets is generally required to achieve premier performance on relatively small datasets. In this paper, we show that video masked autoencoders (VideoMAE) are data-efficient learners for self-supervised video pre-training (SSVP). We are inspired by the recent ImageMAE and propose customized video tube masking with an extremely high ratio. This simple design makes video reconstruction a more challenging self-supervision task, thus encouraging extracting more effective video representations during this pre-training process. We obtain three important findings on SSVP: (1) An extremely high proportion of masking ratio (i.e., 90% to 95%) still yields favorable performance of VideoMAE. The temporally redundant video content enables a higher masking ratio than that of images. (2) VideoMAE achieves impressive results on very small datasets (i.e., around 3k-4k videos) without using any extra data. (3) VideoMAE shows that data quality is more important than data quantity for SSVP. Domain shift between pre-training and target datasets is an important issue. Notably, our VideoMAE with the vanilla ViT can achieve 87.4% on Kinetics-400, 75.4% on Something-Something V2, 91.3% on UCF101, and 62.6% on HMDB51, without using any extra data. Code is available at https://github.com/MCG-NJU/VideoMAE."
                    },
                    {
                        "title": "Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations",
                        "abstract": "Vision Transformers (ViTs) take all the image patches as tokens and construct multi-head self-attention (MHSA) among them. Complete leverage of these image tokens brings redundant computations since not all the tokens are attentive in MHSA. Examples include that tokens containing semantically meaningless or distractive image backgrounds do not positively contribute to the ViT predictions. In this work, we propose to reorganize image tokens during the feed-forward process of ViT models, which is integrated into ViT during training. For each forward inference, we identify the attentive image tokens between MHSA and FFN (i.e., feed-forward network) modules, which is guided by the corresponding class token attention. Then, we reorganize image tokens by preserving attentive image tokens and fusing inattentive ones to expedite subsequent MHSA and FFN computations. To this end, our method EViT improves ViTs from two perspectives. First, under the same amount of input image tokens, our method reduces MHSA and FFN computation for efficient inference. For instance, the inference speed of DeiT-S is increased by 50% while its recognition accuracy is decreased by only 0.3% for ImageNet classification. Second, by maintaining the same computational cost, our method empowers ViTs to take more image tokens as input for recognition accuracy improvement, where the image tokens are from higher resolution images. An example is that we improve the recognition accuracy of DeiT-S by 1% for ImageNet classification at the same computational cost of a vanilla DeiT-S. Meanwhile, our method does not introduce more parameters to ViTs. Experiments on the standard benchmarks show the effectiveness of our method. The code is available at https://github.com/youweiliang/evit"
                    },
                    {
                        "title": "MGSampler: An Explainable Sampling Strategy for Video Action Recognition",
                        "abstract": "Frame sampling is a fundamental problem in video action recognition due to the essential redundancy in time and limited computation resources. The existing sampling strategy often employs a fixed frame selection and lacks the flexibility to deal with complex variations in videos. In this paper, we present a simple, sparse, and explainable frame sampler, termed as Motion-Guided Sampler (MGSampler). Our basic motivation is that motion is an important and universal signal that can drive us to adaptively select frames from videos. Accordingly, we propose two important properties in our MGSampler design: motion sensitive and motion uniform. First, we present two different motion representations to enable us to efficiently distinguish the motion-salient frames from the background. Then, we devise a motion-uniform sampling strategy based on the cumulative motion distribution to ensure the sampled frames evenly cover all the important segments with high motion salience. Our MGSampler yields a new principled and holistic sampling scheme, that could be incorporated into any existing video architecture. Experiments on five benchmarks demonstrate the effectiveness of our MGSampler over the previous fixed sampling strategies, and its generalization power across different backbones, video models, and datasets. The code is available at https://github.com/MCG-NJU/MGSampler."
                    },
                    {
                        "title": "TDN: Temporal Difference Networks for Efficient Action Recognition",
                        "abstract": "Temporal modeling still remains challenging for action recognition in videos. To mitigate this issue, this paper presents a new video architecture, termed as Temporal Difference Network (TDN), with a focus on capturing multi-scale temporal information for efficient action recognition. The core of our TDN is to devise an efficient temporal module (TDM) by explicitly leveraging a temporal difference operator, and systematically assess its effect on short-term and long-term motion modeling. To fully capture temporal information over the entire video, our TDN is established with a two-level difference modeling paradigm. Specifically, for local motion modeling, temporal difference over consecutive frames is used to supply 2D CNNs with finer motion pattern, while for global motion modeling, temporal difference across segments is incorporated to capture long-range structure for motion feature excitation. TDN provides a simple and principled temporal modeling framework and could be instantiated with the existing CNNs at a small extra computational cost. Our TDN presents a new state of the art on the Something-Something V1 & V2 datasets and is on par with the best performance on the Kinetics-400 dataset. In addition, we conduct in-depth ablation studies and plot the visualization results of our TDN, hopefully providing insightful analysis on temporal difference modeling. We release the code at https://github.com/MCG-NJU/TDN."
                    }
                ]
            },
            "d5baa2a3-f5f7-4a27-aabb-c1e6e4ca36cd": {
                "pk": "d5baa2a3-f5f7-4a27-aabb-c1e6e4ca36cd",
                "name": "Limin Wang",
                "collaborators": [
                    "Gangshan Wu",
                    "Guozhen Zhang",
                    "Yuhan Zhu",
                    "Yutao Cui",
                    "Lei Chen",
                    "Zhan Tong",
                    "Yibing Song",
                    "C. Xu",
                    "Haocheng Shen",
                    "Yixuan Liao",
                    "Xiaoxin Chen",
                    "Min Yang",
                    "Guo Chen",
                    "Hongya Wang",
                    "Youxin Chen",
                    "Jiang Yang",
                    "Sheng Guo",
                    "Jiaming Zhang",
                    "Tao Wu",
                    "Mengqing Cao",
                    "Ziteng Gao",
                    "Ruopeng Gao",
                    "Fengyuan Shi",
                    "Boheng Chen",
                    "Chenkai Zeng",
                    "Xiaoyu Zhao",
                    "Yichun Yang",
                    "Jintao Lin",
                    "Haodong Duan",
                    "Kai Chen",
                    "Dahua Lin",
                    "Yin-Dong Zheng",
                    "Tong Lu",
                    "Liang Zhao",
                    "Yi Wang",
                    "Kunchang Li",
                    "Yizhuo Li",
                    "Yinan He",
                    "Bingkun Huang",
                    "Zhiyu Zhao",
                    "Hongjie Zhang",
                    "Jilan Xu",
                    "Yi Liu",
                    "Zun Wang",
                    "Sen Xing",
                    "Junting Pan",
                    "Jiashuo Yu",
                    "Yali Wang",
                    "Yu Qiao",
                    "Mengqi Cao",
                    "Xiaotian Li",
                    "Yilu Wu"
                ],
                "domain": [
                    "Video Action Detection",
                    "Video Object Segmentation",
                    "Multi-Modal Learning",
                    "Self-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Extracting Motion and Appearance via Inter-Frame Attention for Efficient Video Frame Interpolation",
                        "abstract": "Effectively extracting inter-frame motion and appearance information is important for video frame interpolation (VFI). Previous works either extract both types of information in a mixed way or devise separate modules for each type of information, which lead to representation ambiguity and low efficiency. In this paper, we propose a new module to explicitly extract motion and appearance information via a unified operation. Specifically, we rethink the information process in inter-frame attention and reuse its attention map for both appearance feature enhancement and motion information extraction. Furthermore, for efficient VFI, our proposed module could be seamlessly integrated into a hybrid CNN and Transformer architecture. This hybrid pipeline can alleviate the computational complexity of inter-frame attention as well as preserve detailed low-level structure information. Experimental results demonstrate that, for both fixed- and arbitrary-timestep interpolation, our method achieves state-of-the-art performance on various datasets. Meanwhile, our approach enjoys a lighter computation overhead over models with close performance. The source code and models are available at https://github.com/MCG-NJU/EMA-VFI."
                    },
                    {
                        "title": "CoMAE: Single Model Hybrid Pre-training on Small-Scale RGB-D Datasets",
                        "abstract": "Current RGB-D scene recognition approaches often train two standalone backbones for RGB and depth modalities with the same Places or ImageNet pre-training. However, the pre-trained depth network is still biased by RGB-based models which may result in a suboptimal solution. In this paper, we present a single-model self-supervised hybrid pre-training framework for RGB and depth modalities, termed as CoMAE. Our CoMAE presents a curriculum learning strategy to unify the two popular self-supervised representation learning algorithms: contrastive learning and masked image modeling. Specifically, we first build a patch-level alignment task to pre-train a single encoder shared by two modalities via cross-modal contrastive learning. Then, the pre-trained contrastive encoder is passed to a multi-modal masked autoencoder to capture the finer context features from a generative perspective. In addition, our single-model design without requirement of fusion module is very flexible and robust to generalize to unimodal scenario in both training and testing phases. Extensive experiments on SUN RGB-D and NYUDv2 datasets demonstrate the effectiveness of our CoMAE for RGB and depth representation learning. In addition, our experiment results reveal that CoMAE is a data-efficient representation learner. Although we only use the small-scale and unlabeled training set for pre-training, our CoMAE pre-trained models are still competitive to the state-of-the-art methods with extra large-scale and supervised RGB dataset pre-training. Code will be released at https://github.com/MCG-NJU/CoMAE."
                    },
                    {
                        "title": "Joint Modeling of Feature, Correspondence, and a Compressed Memory for Video Object Segmentation",
                        "abstract": "Current prevailing Video Object Segmentation (VOS) methods usually perform dense matching between the current and reference frames after extracting their features. One on hand, the decoupled modeling restricts the targets information propagation only at high-level feature space. On the other hand, the pixel-wise matching leads to a lack of holistic understanding of the targets. To overcome these issues, we propose a unified VOS framework, coined as JointFormer, for joint modeling the three elements of feature, correspondence, and a compressed memory. The core design is the Joint Block, utilizing the flexibility of attention to simultaneously extract feature and propagate the targets information to the current tokens and the compressed memory token. This scheme allows to perform extensive information propagation and discriminative feature learning. To incorporate the long-term temporal targets information, we also devise a customized online updating mechanism for the compressed memory token, which can prompt the information flow along the temporal dimension and thus improve the global modeling capability. Under the design, our method achieves a new state-of-art performance on DAVIS 2017 val/test-dev (89.7% and 87.6%) and YouTube-VOS 2018/2019 val (87.0% and 87.0%) benchmarks, outperforming existing works by a large margin."
                    },
                    {
                        "title": "CycleACR: Cycle Modeling of Actor-Context Relations for Video Action Detection",
                        "abstract": "The relation modeling between actors and scene context advances video action detection where the correlation of multiple actors makes their action recognition challenging. Existing studies model each actor and scene relation to improve action recognition. However, the scene variations and background interference limit the effectiveness of this relation modeling. In this paper, we propose to select actor-related scene context, rather than directly leverage raw video scenario, to improve relation modeling. We develop a Cycle Actor-Context Relation network (CycleACR) where there is a symmetric graph that models the actor and context relations in a bidirectional form. Our CycleACR consists of the Actor-to-Context Reorganization (A2C-R) that collects actor features for context feature reorganizations, and the Context-to-Actor Enhancement (C2A-E) that dynamically utilizes reorganized context features for actor feature enhancement. Compared to existing designs that focus on C2A-E, our CycleACR introduces A2C-R for a more effective relation modeling. This modeling advances our CycleACR to achieve state-of-the-art performance on two popular action detection datasets (i.e., AVA and UCF101-24). We also provide ablation studies and visualizations as well to show how our cycle actor-context relation modeling improves video action detection. Code is available at https://github.com/MCG-NJU/CycleACR."
                    },
                    {
                        "title": "STMixer: A One-Stage Sparse Action Detector",
                        "abstract": "Traditional video action detectors typically adopt the two-stage pipeline, where a person detector is first employed to generate actor boxes and then 3D RoIAlign is used to extract actor-specific features for classification. This detection paradigm requires multi-stage training and inference, and cannot capture context information outside the bounding box. Recently, a few query-based action detectors are proposed to predict action instances in an end-to-end manner. However, they still lack adaptability in feature sampling and decoding, thus suffering from the issues of inferior performance or slower convergence. In this paper, we propose a new one-stage sparse action detector, termed STMixer. STMixer is based on two core designs. First, we present a query-based adaptive feature sampling module, which endows our STMixer with the flexibility of mining a set of discriminative features from the entire spatiotemporal domain. Second, we devise a dual-branch feature mixing module, which allows our STMixer to dynamically attend to and mix video features along the spatial and the temporal dimension respectively for better feature decoding. Coupling these two designs with a video backbone yields an efficient end-to-end action detector. Without bells and whistles, our STMixer obtains the state-of-the-art results on the datasets of AVA, UCF101-24, and JHMDB."
                    },
                    {
                        "title": "DPL: Decoupled Prompt Learning for Vision-Language Models",
                        "abstract": "Prompt learning has emerged as an efficient and effective approach for transferring foundational Vision-Language Models (e.g., CLIP) to downstream tasks. However, current methods tend to overfit to seen categories, thereby limiting their generalization ability for unseen classes. In this paper, we propose a new method, Decoupled Prompt Learning (DPL), which reformulates the attention in prompt learning to alleviate this problem. Specifically, we theoretically investigate the collaborative process between prompts and instances (i.e., image patches/text tokens) by reformulating the original self-attention into four separate sub-processes. Through detailed analysis, we observe that certain sub-processes can be strengthened to bolster robustness and generalizability by some approximation techniques. Furthermore, we introduce language-conditioned textual prompting based on decoupled attention to naturally preserve the generalization of text input. Our approach is flexible for both visual and textual modalities, making it easily extendable to multi-modal prompt learning. By combining the proposed techniques, our approach achieves state-of-the-art performance on three representative benchmarks encompassing 15 image recognition datasets, while maintaining parameter-efficient. Moreover, our DPL does not rely on any auxiliary regularization task or extra training data, further demonstrating its remarkable generalization ability."
                    },
                    {
                        "title": "MeMOTR: Long-Term Memory-Augmented Transformer for Multi-Object Tracking",
                        "abstract": "As a video task, Multiple Object Tracking (MOT) is expected to capture temporal information of targets effectively. Unfortunately, most existing methods only explicitly exploit the object features between adjacent frames, while lacking the capacity to model long-term temporal information. In this paper, we propose MeMOTR, a long-term memory-augmented Transformer for multi-object tracking. Our method is able to make the same object's track embedding more stable and distinguishable by leveraging longterm memory injection with a customized memory-attention layer. This significantly improves the target association ability of our model. Experimental results on DanceTrack show that MeMOTR impressively surpasses the state-of-the-art method by 7.9% and 13.0% on HOTA and AssA metrics, respectively. Furthermore, our model also outperforms other Transformer-based methods on association performance on MOT17 and generalizes well on BDD100K. Code is available at https://github.com/MCG-NJU/MeMOTR."
                    },
                    {
                        "title": "Progressive Visual Prompt Learning with Contrastive Feature Re-formation",
                        "abstract": "Prompt learning has been designed as an alternative to fine-tuning for adapting Vision-language (V-L) models to the downstream tasks. Previous works mainly focus on text prompt while visual prompt works are limited for V-L models. The existing visual prompt methods endure either mediocre performance or unstable training process, indicating the difficulty of visual prompt learning. In this paper, we propose a new Progressive Visual Prompt (ProVP) structure to strengthen the interactions among prompts of different layers. More importantly, our ProVP could effectively propagate the image embeddings to deep layers and behave partially similar to an instance adaptive prompt method. To alleviate generalization deterioration, we further propose a new contrastive feature re-formation, which prevents the serious deviation of the prompted visual feature from the fixed CLIP visual feature distribution. Combining both, our method (ProVP-Ref) is evaluated on 11 image benchmark datasets and achieves 7/11 state-of-theart results on both few-shot and base-to-novel settings. To the best of our knowledge, we are the first to demonstrate the superior performance of visual prompts in V-L models to previous prompt-based methods in downstream tasks. Meanwhile, it implies that our ProVP-Ref shows the best capability to adapt and to generalize."
                    },
                    {
                        "title": "SportsMOT: A Large Multi-Object Tracking Dataset in Multiple Sports Scenes",
                        "abstract": "Multi-object tracking (MOT) in sports scenes plays a critical role in gathering players statistics, supporting further applications, such as automatic tactical analysis. Yet existing MOT benchmarks cast little attention on this domain. In this work, we present a new large-scale multi-object tracking dataset in multiple sports scenes, coined as SportsMOT, where all players on the court are supposed to be tracked. It consists of 240 video sequences, over 150K frames (almost 15x MOT17) and over 1.6M bounding boxes (3x MOT17) collected from 3 sports categories, including basketball, volleyball and football. Our dataset is characterized with two key properties: 1) fast and variable-speed motion and 2) similar yet distinguishable appearance. We expect SportsMOT to encourage the MOT trackers to promote in both motion-based association and appearance-based association. We benchmark several state-of-the-art trackers and reveal the key challenge of SportsMOT lies in object association. To alleviate the issue, we further propose a new multi-object tracking framework, termed as MixSort, introducing a MixFormer-like structure as an auxiliary association model to prevailing tracking-by-detection trackers. By integrating the customized appearance-based association with the original motion-based association, MixSort achieves state-of-the-art performance on SportsMOT and MOT17. Based on MixSort, we give an in-depth analysis and provide some profound insights into SportsMOT."
                    },
                    {
                        "title": "Efficient Video Action Detection with Token Dropout and Context Refinement",
                        "abstract": "Streaming video clips with large-scale video tokens impede vision transformers (ViTs) for efficient recognition, especially in video action detection where sufficient spatiotemporal representations are required for precise actor identification. In this work, we propose an end-to-end framework for efficient video action detection (EVAD) based on vanilla ViTs. Our EVAD consists of two specialized designs for video action detection. First, we propose a spatiotemporal token dropout from a keyframe-centric perspective. In a video clip, we maintain all tokens from its keyframe, preserve tokens relevant to actor motions from other frames, and drop out the remaining tokens in this clip. Second, we refine scene context by leveraging remaining tokens for better recognizing actor identities. The region of interest (RoI) in our action detector is expanded into temporal domain. The captured spatiotemporal actor identity representations are refined via scene context in a decoder with the attention mechanism. These two designs make our EVAD efficient while maintaining accuracy, which is validated on three benchmark datasets (i.e., AVA, UCF101-24, JHMDB). Compared to the vanilla ViT backbone, our EVAD reduces the overall GFLOPs by 43% and improves real-time inference speed by 40% with no performance degradation. Moreover, even at similar computational costs, our EVAD can improve the performance by 1.1 mAP with higher resolution inputs. Code is available at https://github.com/MCG-NJU/EVAD."
                    },
                    {
                        "title": "OCSampler: Compressing Videos to One Clip with Single-step Sampling",
                        "abstract": "Videos incorporate rich semantics as well as redundant information. Seeking a compact yet effective video representation, e.g., sample informative frames from the entire video, is critical to efficient video recognition. There have been works that formulate frame sampling as a sequential decision task by selecting frames one by one according to their importance. In this paper, we present a more efficient framework named OCSampler, which explores such a representation with one short clip. OCSampler designs a new paradigm of learning instance-specific video condensation policies to select frames only in a single step. Rather than picking up frames sequentially like previous methods, we simply process a whole sequence at once. Accordingly, these policies are derived from a light-weighted skim network together with a simple yet effective policy network. Moreover, we extend the proposed method with a frame number budget, enabling the framework to produce correct predictions in high confidence with as few frames as possible. Experiments on various benchmarks demonstrate the effectiveness of OCSampler over previous methods in terms of accuracy and efficiency. Specifically, it achieves 76.9% mAP and 21.7 GFLOPs on ActivityNet with an impressive throughput: 123.9 Video/s on a single TITAN Xp GPU."
                    },
                    {
                        "title": "BasicTAD: an Astounding RGB-Only Baseline for Temporal Action Detection",
                        "abstract": "Temporal action detection (TAD) is extensively studied in the video understanding community by generally following the object detection pipeline in images. However, complex designs are not uncommon in TAD, such as two-stream feature extraction, multi-stage training, complex temporal modeling, and global context fusion. In this paper, we do not aim to introduce any novel technique for TAD. Instead, we study a simple, straightforward, yet must-known baseline given the current status of complex design and low detection efficiency in TAD. In our simple baseline (termed BasicTAD), we decompose the TAD pipeline into several essential components: data sampling, backbone design, neck construction, and detection head. We extensively investigate the existing techniques in each component for this baseline, and more importantly, perform end-to-end training over the entire pipeline thanks to the simplicity of design. As a result, this simple BasicTAD yields an astounding and real-time RGB-Only baseline very close to the state-of-the-art methods with two-stream inputs. In addition, we further improve the BasicTAD by preserving more temporal and spatial information in network representation (termed as PlusTAD). Empirical results demonstrate that our PlusTAD is very efficient and significantly outperforms the previous methods on the datasets of THUMOS14 and FineAction. Meanwhile, we also perform in-depth visualization and error analysis on our proposed method and try to provide more insights on the TAD problem. Our approach can serve as a strong baseline for future TAD research. The code and model will be released at https://github.com/MCG-NJU/BasicTAD."
                    },
                    {
                        "title": "Task-specific Inconsistency Alignment for Domain Adaptive Object Detection",
                        "abstract": "Detectors trained with massive labeled data often exhibit dramatic performance degradation in some particular scenarios with data distribution gap. To alleviate this problem of domain shift, conventional wisdom typically concentrates solely on reducing the discrepancy between the source and target domains via attached domain classifiers, yet ignoring the difficulty of such transferable features in coping with both classification and localization subtasks in object detection. To address this issue, in this paper, we propose Task-specific Inconsistency Alignment (TIA), by developing a new alignment mechanism in separate task spaces, improving the performance of the detector on both subtasks. Specifically, we add a set of auxiliary predictors for both classification and localization branches, and exploit their behavioral inconsistencies as finer-grained domain-specific measures. Then, we devise task-specific losses to align such cross-domain disagreement of both subtasks. By optimizing them individually, we are able to well approximate the category- and boundary-wise discrepancies in each task space, and therefore narrow them in a decoupled manner. TIA demonstrates superior results on various scenarios to the previous state-of-the-art methods. It is also observed that both the classification and localization capabilities of the detector are sufficiently strengthened, further demonstrating the effectiveness of our TIA method. Code and trained models are publicly available at https://github.com/MCG-NJU/TIA."
                    },
                    {
                        "title": "InternVideo: General Video Foundation Models via Generative and Discriminative Learning",
                        "abstract": "The foundation models have recently shown excellent performance on a variety of downstream tasks in computer vision. However, most existing vision foundation models simply focus on image-level pretraining and adpation, which are limited for dynamic and complex video-level understanding tasks. To fill the gap, we present general video foundation models, InternVideo, by taking advantage of both generative and discriminative self-supervised video learning. Specifically, InternVideo efficiently explores masked video modeling and video-language contrastive learning as the pretraining objectives, and selectively coordinates video representations of these two complementary frameworks in a learnable manner to boost various video applications. Without bells and whistles, InternVideo achieves state-of-the-art performance on 39 video datasets from extensive tasks including video action recognition/detection, video-language alignment, and open-world video applications. Especially, our methods can obtain 91.1% and 77.2% top-1 accuracy on the challenging Kinetics-400 and Something-Something V2 benchmarks, respectively. All of these results effectively show the generality of our InternVideo for video understanding. The code will be released at https://github.com/OpenGVLab/InternVideo ."
                    },
                    {
                        "title": "SpotFormer: A Transformer-based Framework for Precise Soccer Action Spotting",
                        "abstract": "Action spotting and classification consist in detecting the exact moments at which events occur in long videos. The current mainstream spotting practices generally use a two-stage pipeline that performs feature collection and integration, then salient action detection and postprocessing. Following that, we present SpotFormer, a simple yet effective framework, capable of precise action spotting. Specifically, we employ several most advanced backbone networks as auxiliary feature extractors, and reduce feature dimensionality in a straightforward and efficient way. The frame-wise features are fed into a transformer-based spotting network devised to leverage spatiotemporal information. We obtain 0.609 tight mAP score via model ensemble and achieve the state-of-the-art performance on the SoccerNet-v2 dataset."
                    }
                ]
            },
            "7d027b47-3d19-426f-a056-719b5c9787aa": {
                "pk": "7d027b47-3d19-426f-a056-719b5c9787aa",
                "name": "Mike Zheng Shou",
                "collaborators": [
                    "K. Lin",
                    "Qianli Xu",
                    "Joya Chen",
                    "Kevin Qinghong Lin",
                    "Alex Jinpeng Wang",
                    "Linjie Li",
                    "Lijuan Wang",
                    "Zongbo Han",
                    "Jialong Yang",
                    "Junfan Li",
                    "Qinghua Hu",
                    "Changqing Zhang",
                    "Shiwei Wu",
                    "Qimeng Wang",
                    "Yan Gao",
                    "Tong Xu",
                    "Yao Hu",
                    "Enhong Chen",
                    "Jinheng Xie",
                    "Jiajun Feng",
                    "Zhaoxu Tian",
                    "Yawen Huang",
                    "Xi Xia",
                    "Nanxu Gong",
                    "Xu Zuo",
                    "Jiaqi Yang",
                    "Yefeng Zheng",
                    "Yiqi Lin",
                    "Min Li",
                    "Jianfeng Wang",
                    "Zhengyuan Yang",
                    "Jingtao Sun",
                    "Yaonan Wang",
                    "Mingtao Feng",
                    "Chao Ding",
                    "A. Mian",
                    "Ziteng Gao",
                    "Zhan Tong"
                ],
                "domain": [
                    "Vision-Language Models",
                    "Self-Supervised Learning",
                    "Multimodal Learning",
                    "Video Understanding"
                ],
                "publications": [
                    {
                        "title": "DOTA: Distributional Test-Time Adaptation of Vision-Language Models",
                        "abstract": "Vision-language foundation models (e.g., CLIP) have shown remarkable performance across a wide range of tasks. However, deploying these models may be unreliable when significant distribution gaps exist between the training and test data. The training-free test-time dynamic adapter (TDA) is a promising approach to address this issue by storing representative test samples to guide the classification of subsequent ones. However, TDA only naively maintains a limited number of reference samples in the cache, leading to severe test-time catastrophic forgetting when the cache is updated by dropping samples. In this paper, we propose a simple yet effective method for DistributiOnal Test-time Adaptation (Dota). Instead of naively memorizing representative test samples, Dota continually estimates the distributions of test samples, allowing the model to continually adapt to the deployment environment. The test-time posterior probabilities are then computed using the estimated distributions based on Bayes' theorem for adaptation purposes. To further enhance the adaptability on the uncertain samples, we introduce a new human-in-the-loop paradigm which identifies uncertain samples, collects human-feedback, and incorporates it into the Dota framework. Extensive experiments validate that Dota enables CLIP to continually learn, resulting in a significant improvement compared to current state-of-the-art methods."
                    },
                    {
                        "title": "VideoLLM-MoD: Efficient Video-Language Streaming with Mixture-of-Depths Vision Computation",
                        "abstract": "A well-known dilemma in large vision-language models (e.g., GPT-4, LLaVA) is that while increasing the number of vision tokens generally enhances visual understanding, it also significantly raises memory and computational costs, especially in long-term, dense video frame streaming scenarios. Although learnable approaches like Q-Former and Perceiver Resampler have been developed to reduce the vision token burden, they overlook the context causally modeled by LLMs (i.e., key-value cache), potentially leading to missed visual cues when addressing user queries. In this paper, we introduce a novel approach to reduce vision compute by leveraging redundant vision tokens\"skipping layers\"rather than decreasing the number of vision tokens. Our method, VideoLLM-MoD, is inspired by mixture-of-depths LLMs and addresses the challenge of numerous vision tokens in long-term or streaming video. Specifically, for each transformer layer, we learn to skip the computation for a high proportion (e.g., 80\\%) of vision tokens, passing them directly to the next layer. This approach significantly enhances model efficiency, achieving approximately \\textasciitilde42\\% time and \\textasciitilde30\\% memory savings for the entire training. Moreover, our method reduces the computation in the context and avoid decreasing the vision tokens, thus preserving or even improving performance compared to the vanilla model. We conduct extensive experiments to demonstrate the effectiveness of VideoLLM-MoD, showing its state-of-the-art results on multiple benchmarks, including narration, forecasting, and summarization tasks in COIN, Ego4D, and Ego-Exo4D datasets."
                    },
                    {
                        "title": "Learning Long-form Video Prior via Generative Pre-Training",
                        "abstract": "Concepts involved in long-form videos such as people, objects, and their interactions, can be viewed as following an implicit prior. They are notably complex and continue to pose challenges to be comprehensively learned. In recent years, generative pre-training (GPT) has exhibited versatile capacities in modeling any kind of text content even visual locations. Can this manner work for learning long-form video prior? Instead of operating on pixel space, it is efficient to employ visual locations like bounding boxes and keypoints to represent key information in videos, which can be simply discretized and then tokenized for consumption by GPT. Due to the scarcity of suitable data, we create a new dataset called \\textbf{Storyboard20K} from movies to serve as a representative. It includes synopses, shot-by-shot keyframes, and fine-grained annotations of film sets and characters with consistent IDs, bounding boxes, and whole body keypoints. In this way, long-form videos can be represented by a set of tokens and be learned via generative pre-training. Experimental results validate that our approach has great potential for learning long-form video prior. Code and data will be released at \\url{https://github.com/showlab/Long-form-Video-Prior}."
                    },
                    {
                        "title": "Leveraging Visual Tokens for Extended Text Contexts in Multi-Modal Learning",
                        "abstract": "Training models with longer in-context lengths is a significant challenge for multimodal model due to substantial GPU memory and computational costs. This exploratory study does not present state-of-the-art models; rather, it introduces an innovative method designed to increase in-context text length in multi-modality large language models (MLLMs) efficiently. We present Visualized In-Context Text Processing (VisInContext), which processes long in-context text using visual tokens. This technique significantly reduces GPU memory usage and floating point operations (FLOPs) for both training and inferenceing stage. For instance, our method expands the pre-training in-context text length from 256 to 2048 tokens with nearly same FLOPs for a 56 billion parameter MOE model. Experimental results demonstrate that model trained with VisInContext delivers superior performance on common downstream benchmarks for in-context few-shot evaluation. Additionally, VisInContext is complementary to existing methods for increasing in-context text length and enhances document understanding capabilities, showing great potential in document QA tasks and sequential document retrieval."
                    },
                    {
                        "title": "COSMO: COntrastive Streamlined MultimOdal Model with Interleaved Pre-Training",
                        "abstract": "In the evolution of Vision-Language Pre-training, shifting from short-text comprehension to encompassing extended textual contexts is pivotal. Recent autoregressive vision-language models like \\cite{flamingo, palme}, leveraging the long-context capability of Large Language Models, have excelled in few-shot text generation tasks but face challenges in alignment tasks. Addressing this gap, we introduce the contrastive loss into text generation models, presenting the COntrastive-Streamlined MultimOdal framework (\\ModelName), strategically partitioning the language model into dedicated unimodal text processing and adept multimodal data handling components. \\ModelName, our unified framework, merges unimodal and multimodal elements, enhancing model performance for tasks involving textual and visual data while notably reducing learnable parameters. However, these models demand extensive long-text datasets, yet the availability of high-quality long-text video datasets remains limited. To bridge this gap, this work introduces \\VideoDatasetName, an inaugural interleaved video-text dataset featuring comprehensive captions, marking a significant step forward. Demonstrating its impact, we illustrate how \\VideoDatasetName{} enhances model performance in image-text tasks. With 34% learnable parameters and utilizing 72\\% of the available data, our model demonstrates significant superiority over OpenFlamingo~\\cite{openflamingo}. For instance, in the 4-shot flickr captioning task, performance notably improves from 57.2% to 65.\\%. The contributions of \\ModelName{} and \\VideoDatasetName{} are underscored by notable performance gains across 14 diverse downstream datasets encompassing both image-text and video-text tasks."
                    },
                    {
                        "title": "Diffusion-Driven Self-Supervised Learning for Shape Reconstruction and Pose Estimation",
                        "abstract": "Fully-supervised category-level pose estimation aims to determine the 6-DoF poses of unseen instances from known categories, requiring expensive mannual labeling costs. Recently, various self-supervised category-level pose estimation methods have been proposed to reduce the requirement of the annotated datasets. However, most methods rely on synthetic data or 3D CAD model for self-supervised training, and they are typically limited to addressing single-object pose problems without considering multi-objective tasks or shape reconstruction. To overcome these challenges and limitations, we introduce a diffusion-driven self-supervised network for multi-object shape reconstruction and categorical pose estimation, only leveraging the shape priors. Specifically, to capture the SE(3)-equivariant pose features and 3D scale-invariant shape information, we present a Prior-Aware Pyramid 3D Point Transformer in our network. This module adopts a point convolutional layer with radial-kernels for pose-aware learning and a 3D scale-invariant graph convolution layer for object-level shape representation, respectively. Furthermore, we introduce a pretrain-to-refine self-supervised training paradigm to train our network. It enables proposed network to capture the associations between shape priors and observations, addressing the challenge of intra-class shape variations by utilising the diffusion mechanism. Extensive experiments conducted on four public datasets and a self-built dataset demonstrate that our method significantly outperforms state-of-the-art self-supervised category-level baselines and even surpasses some fully-supervised instance-level and category-level methods."
                    },
                    {
                        "title": "Bootstrapping SparseFormers from Vision Foundation Models",
                        "abstract": "The recently proposed SparseFormer architecture provides an alternative approach to visual understanding by utilizing a significantly lower number of visual tokens via adjusting RoIs, greatly reducing computational costs while still achieving promising performance. However, training SparseFormers from scratch is still expensive, and scaling up the number of parameters can be challenging. In this paper, we propose to bootstrap SparseFormers from ViT-based vision foundation models in a simple and efficient way. Since the majority of SparseFormer blocks are the standard transformer ones, we can inherit weights from large-scale pre-trained vision transformers and freeze them as much as possible. Therefore, we only need to train the SparseFormer-specific lightweight focusing transformer to adjust token RoIs and fine-tune a few early pre-trained blocks to align the final token representation. In such a way, we can bootstrap SparseFormer architectures from various large-scale pre-trained models (e.g., IN-21 K pre-trained AugRegs or CLIPs) using a rather smaller amount of training samples (e.g., IN-IK) and without labels or captions within just a few hours. As a result, the bootstrapped unimodal SparseFormer (from AugReg-ViT-L/16-384) can reach 84.9% accuracy on IN-IK with only 49 tokens, and the multimodal SparseFormer from CLIPs also demonstrates notable zero-shot performance with highly reduced computational cost without seeing any caption during the bootstrapping procedure. In addition, CLIP-bootstrapped SparseFormers, which align the output space with language without seeing a word, can serve as efficient vision encoders in multimodal large language models. Code and models are available at https://github.com/showlab/sparseformer"
                    }
                ]
            }
        }
    },
    "2405.15926": {
        "paper_data": {
            "title": "Dissecting the Interplay of Attention Paths in a Statistical Mechanics Theory of Transformers",
            "url": "http://arxiv.org/abs/2405.15926v1",
            "arxiv_id": "2405.15926",
            "authors": [
                "Lorenzo Tiberi",
                "Francesca Mignacco",
                "Kazuki Irie",
                "Haim Sompolinsky"
            ],
            "abstract": "Despite the remarkable empirical performance of Transformers, their theoretical understanding remains elusive. Here, we consider a deep multi-head self-attention network, that is closely related to Transformers yet analytically tractable. We develop a statistical mechanics theory of Bayesian learning in this model, deriving exact equations for the network's predictor statistics under the finite-width thermodynamic limit, i.e., $N,P\\rightarrow\\infty$, $P/N=\\mathcal{O}(1)$, where $N$ is the network width and $P$ is the number of training examples. Our theory shows that the predictor statistics are expressed as a sum of independent kernels, each one pairing different 'attention paths', defined as information pathways through different attention heads across layers. The kernels are weighted according to a 'task-relevant kernel combination' mechanism that aligns the total kernel with the task labels. As a consequence, this interplay between attention paths enhances generalization performance. Experiments confirm our findings on both synthetic and real-world sequence classification tasks. Finally, our theory explicitly relates the kernel combination mechanism to properties of the learned weights, allowing for a qualitative transfer of its insights to models trained via gradient descent. As an illustration, we demonstrate an efficient size reduction of the network, by pruning those attention heads that are deemed less relevant by our theory.",
            "introduction": "   1 Introduction  In recent years, Transformer models based on multi-head self-attention layers [1, 2, 3] have achieved remarkable performance at natural language processing and vision tasks [4, 5, 6]. Yet, theoretical characterizations accounting for the success of these architectures remain sparse. Two fundamental questions remain to a large extent unsolved: First, interpretability\u2014how can we discern task-relevant structures within the learned weights? Second, generalization\u2014what specific aspects of the Transformer architecture are responsible for their effective learning? We posit that one important feature of Transformers is the combination of layer-wise multi-head organization with depth. This provides the network with a large number of attention paths, defined as specific sequences of heads through the attention layers. Their interplay is still poorly understood by deep learning theory.   In most cases, theoretical characterizations of transformers\u2019 expressivity [7], inductive bias [8, 9], generalization [10, 11, 12] and training dynamics [13, 14, 15] rely on simplifying assumptions on the network architecture. A characterization of attention paths is inaccessible in these models, either because attention paths are not defined in the first place, as in models consisting of a single-head [16, 12], a single-layer [7, 8], or both [9, 10, 11, 13, 15], or because the paths interplay cannot be fully described, due to constraints imposed on the learnable weights [14]. Few works consider a multi-head, multi-layer architecture [17, 18], but address different questions than the present study, such as expressivity and generalization bounds. Further details on these and analogous works are discussed in Appendix\u00a0I.   One characterization of the complete transformer architecture has been obtained in the infinite-width limit N\u2192\u221e\u2192\ud835\udc41N\\to\\inftyitalic_N \u2192 \u221e (and infinite number of heads H\u2192\u221e\u2192\ud835\udc3bH\\to\\inftyitalic_H \u2192 \u221e) [19, 20], an actively studied regime in which neural networks become equivalent to Gaussian processes (GP) [21, 22]. However, the attention paths interplay is lost in this limit, because the network weights remain statistically independent after learning. This limitation can be overcome by considering the finite-width thermodynamic limit [23, 24, 25, 26], where also the number of examples P\u2192\u221e\u2192\ud835\udc43P\\to\\inftyitalic_P \u2192 \u221e, such that P/N\u2192\u03b1\u2208\u211d+\u2192\ud835\udc43\ud835\udc41\ud835\udefcsuperscript\u211dP/N\\to\\alpha\\in\\mathbb{R}^{+}italic_P / italic_N \u2192 italic_\u03b1 \u2208 blackboard_R start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT. In this regime, for example, multi-gated deep networks showcase task-relevant interplay between gates, mediated by the learned weights [27].   In this work, we apply the statistical mechanics theory of finite-width networks to a deep multi-head self-attention model, which closely mimics the attention paths interplay in Transformers, while remaining analytically tractable. Our main contributions can be summarized as follows:   \u2022  We derive exact equations for the predictor statistics under Bayesian learning of the network\u2019s value weights, at fixed pre-trained query and key weights.    \u2022  We shed light on the interplay between attention paths by uncovering a task-relevant kernel combination mechanism, emerging beyond the GP limit (\u03b1>0\ud835\udefc0\\alpha>0italic_\u03b1 > 0). This constructs the network\u2019s kernel as an optimally weighted sum of many kernels, one for each pair of attention paths, thereby improving generalization.    \u2022  We provide interpretability to this mechanism, by directly relating it to the magnitude and correlations developed by the learned weights. This allows our insights to be transferred to networks trained with gradient descent. As an application, we show that a trained network can be efficiently reduced in size, by pruning those attention paths that are deemed less relevant by our theory.     \u2022  We corroborate our findings on both synthetic and real-world sequence classification tasks, illustrating the two main benefits of kernel combination: task-relevant weighting and correlation",
            "references": [
                {
                    "title": "Towards Understanding Inductive Bias in Transformers: A View From Infinity",
                    "abstract": "We study inductive bias in Transformers in the infinitely over-parameterized Gaussian process limit and argue transformers tend to be biased towards more permutation symmetric functions in sequence space. We show that the representation theory of the symmetric group can be used to give quantitative analytical predictions when the dataset is symmetric to permutations between tokens. We present a simplified transformer block and solve the model at the limit, including accurate predictions for the learning curves and network outputs. We show that in common setups, one can derive tight bounds in the form of a scaling law for the learnability as a function of the context length. Finally, we argue WikiText dataset, does indeed possess a degree of permutation symmetry."
                },
                {
                    "title": "A phase transition between positional and semantic learning in a solvable model of dot-product attention",
                    "abstract": "Many empirical studies have provided evidence for the emergence of algorithmic mechanisms (abilities) in the learning of language models, that lead to qualitative improvements of the model capabilities. Yet, a theoretical characterization of how such mechanisms emerge remains elusive. In this paper, we take a step in this direction by providing a tight theoretical analysis of the emergence of semantic attention in a solvable model of dot-product attention. More precisely, we consider a non-linear self-attention layer with trainable tied and low-rank query and key matrices. In the asymptotic limit of high-dimensional data and a comparably large number of training samples we provide a tight closed-form characterization of the global minimum of the non-convex empirical loss landscape. We show that this minimum corresponds to either a positional attention mechanism (with tokens attending to each other based on their respective positions) or a semantic attention mechanism (with tokens attending to each other based on their meaning), and evidence an emergent phase transition from the former to the latter with increasing sample complexity. Finally, we compare the dot-product attention layer to a linear positional baseline, and show that it outperforms the latter using the semantic mechanism provided it has access to sufficient data."
                },
                {
                    "title": "A mathematical perspective on Transformers",
                    "abstract": "Transformers play a central role in the inner workings of large language models. We develop a mathematical framework for analyzing Transformers based on their interpretation as interacting particle systems, which reveals that clusters emerge in long time. Our study explores the underlying theory and offers new perspectives for mathematicians as well as computer scientists."
                },
                {
                    "title": "How Do Transformers Learn In-Context Beyond Simple Functions? A Case Study on Learning with Representations",
                    "abstract": "While large language models based on the transformer architecture have demonstrated remarkable in-context learning (ICL) capabilities, understandings of such capabilities are still in an early stage, where existing theory and mechanistic understanding focus mostly on simple scenarios such as learning simple function classes. This paper takes initial steps on understanding ICL in more complex scenarios, by studying learning with representations. Concretely, we construct synthetic in-context learning problems with a compositional structure, where the label depends on the input through a possibly complex but fixed representation function, composed with a linear function that differs in each instance. By construction, the optimal ICL algorithm first transforms the inputs by the representation function, and then performs linear ICL on top of the transformed dataset. We show theoretically the existence of transformers that approximately implement such algorithms with mild depth and size. Empirically, we find trained transformers consistently achieve near-optimal ICL performance in this setting, and exhibit the desired dissection where lower layers transforms the dataset and upper layers perform linear ICL. Through extensive probing and a new pasting experiment, we further reveal several mechanisms within the trained transformers, such as concrete copying behaviors on both the inputs and the representations, linear ICL capability of the upper layers alone, and a post-ICL representation selection mechanism in a harder mixture setting. These observed mechanisms align well with our theory and may shed light on how transformers perform ICL in more realistic scenarios."
                },
                {
                    "title": "Connecting NTK and NNGP: A Unified Theoretical Framework for Neural Network Learning Dynamics in the Kernel Regime",
                    "abstract": "Artificial neural networks have revolutionized machine learning in recent years, but a complete theoretical framework for their learning process is still lacking. Substantial progress has been made for infinitely wide networks. In this regime, two disparate theoretical frameworks have been used, in which the network's output is described using kernels: one framework is based on the Neural Tangent Kernel (NTK) which assumes linearized gradient descent dynamics, while the Neural Network Gaussian Process (NNGP) kernel assumes a Bayesian framework. However, the relation between these two frameworks has remained elusive. This work unifies these two distinct theories using a Markov proximal learning model for learning dynamics in an ensemble of randomly initialized infinitely wide deep networks. We derive an exact analytical expression for the network input-output function during and after learning, and introduce a new time-dependent Neural Dynamical Kernel (NDK) from which both NTK and NNGP kernels can be derived. We identify two learning phases characterized by different time scales: gradient-driven and diffusive learning. In the initial gradient-driven learning phase, the dynamics is dominated by deterministic gradient descent, and is described by the NTK theory. This phase is followed by the diffusive learning stage, during which the network parameters sample the solution space, ultimately approaching the equilibrium distribution corresponding to NNGP. Combined with numerical evaluations on synthetic and benchmark datasets, we provide novel insights into the different roles of initialization, regularization, and network depth, as well as phenomena such as early stopping and representational drift. This work closes the gap between the NTK and NNGP theories, providing a comprehensive framework for understanding the learning process of deep neural networks in the infinite width limit."
                },
                {
                    "title": "What can a Single Attention Layer Learn? A Study Through the Random Features Lens",
                    "abstract": "Attention layers -- which map a sequence of inputs to a sequence of outputs -- are core building blocks of the Transformer architecture which has achieved significant breakthroughs in modern artificial intelligence. This paper presents a rigorous theoretical study on the learning and generalization of a single multi-head attention layer, with a sequence of key vectors and a separate query vector as input. We consider the random feature setting where the attention layer has a large number of heads, with randomly sampled frozen query and key matrices, and trainable value matrices. We show that such a random-feature attention layer can express a broad class of target functions that are permutation invariant to the key vectors. We further provide quantitative excess risk bounds for learning these target functions from finite samples, using random feature attention with finitely many heads. Our results feature several implications unique to the attention structure compared with existing random features theory for neural networks, such as (1) Advantages in the sample complexity over standard two-layer random-feature networks; (2) Concrete and natural classes of functions that can be learned efficiently by a random-feature attention layer; and (3) The effect of the sampling distribution of the query-key weight matrix (the product of the query and key matrix), where Gaussian random weights with a non-zero mean result in better sample complexities over the zero-mean counterpart for learning certain natural target functions. Experiments on simulated data corroborate our theoretical findings and further illustrate the interplay between the sample size and the complexity of the target function."
                },
                {
                    "title": "The Shaped Transformer: Attention Models in the Infinite Depth-and-Width Limit",
                    "abstract": "In deep learning theory, the covariance matrix of the representations serves as a proxy to examine the network's trainability. Motivated by the success of Transformers, we study the covariance matrix of a modified Softmax-based attention model with skip connections in the proportional limit of infinite-depth-and-width. We show that at initialization the limiting distribution can be described by a stochastic differential equation (SDE) indexed by the depth-to-width ratio. To achieve a well-defined stochastic limit, the Transformer's attention mechanism is modified by centering the Softmax output at identity, and scaling the Softmax logits by a width-dependent temperature parameter. We examine the stability of the network through the corresponding SDE, showing how the scale of both the drift and diffusion can be elegantly controlled with the aid of residual connections. The existence of a stable SDE implies that the covariance structure is well-behaved, even for very large depth and width, thus preventing the notorious issues of rank degeneracy in deep attention models. Finally, we show, through simulations, that the SDE provides a surprisingly good description of the corresponding finite-size model. We coin the name shaped Transformer for these architectural modifications."
                },
                {
                    "title": "Max-Margin Token Selection in Attention Mechanism",
                    "abstract": "Attention mechanism is a central component of the transformer architecture which led to the phenomenal success of large language models. However, the theoretical principles underlying the attention mechanism are poorly understood, especially its nonconvex optimization dynamics. In this work, we explore the seminal softmax-attention model $f(\\boldsymbol{X})=\\langle \\boldsymbol{Xv}, \\texttt{softmax}(\\boldsymbol{XWp})\\rangle$, where $\\boldsymbol{X}$ is the token sequence and $(\\boldsymbol{v},\\boldsymbol{W},\\boldsymbol{p})$ are trainable parameters. We prove that running gradient descent on $\\boldsymbol{p}$, or equivalently $\\boldsymbol{W}$, converges in direction to a max-margin solution that separates $\\textit{locally-optimal}$ tokens from non-optimal ones. This clearly formalizes attention as an optimal token selection mechanism. Remarkably, our results are applicable to general data and precisely characterize $\\textit{optimality}$ of tokens in terms of the value embeddings $\\boldsymbol{Xv}$ and problem geometry. We also provide a broader regularization path analysis that establishes the margin maximizing nature of attention even for nonlinear prediction heads. When optimizing $\\boldsymbol{v}$ and $\\boldsymbol{p}$ simultaneously with logistic loss, we identify conditions under which the regularization paths directionally converge to their respective hard-margin SVM solutions where $\\boldsymbol{v}$ separates the input features based on their labels. Interestingly, the SVM formulation of $\\boldsymbol{p}$ is influenced by the support vector geometry of $\\boldsymbol{v}$. Finally, we verify our theoretical findings via numerical experiments and provide insights."
                },
                {
                    "title": "Trained Transformers Learn Linear Models In-Context",
                    "abstract": "Attention-based neural networks such as transformers have demonstrated a remarkable ability to exhibit in-context learning (ICL): Given a short prompt sequence of tokens from an unseen task, they can formulate relevant per-token and next-token predictions without any parameter updates. By embedding a sequence of labeled training data and unlabeled test data as a prompt, this allows for transformers to behave like supervised learning algorithms. Indeed, recent work has shown that when training transformer architectures over random instances of linear regression problems, these models' predictions mimic those of ordinary least squares. Towards understanding the mechanisms underlying this phenomenon, we investigate the dynamics of ICL in transformers with a single linear self-attention layer trained by gradient flow on linear regression tasks. We show that despite non-convexity, gradient flow with a suitable random initialization finds a global minimum of the objective function. At this global minimum, when given a test prompt of labeled examples from a new prediction task, the transformer achieves prediction error competitive with the best linear predictor over the test prompt distribution. We additionally characterize the robustness of the trained transformer to a variety of distribution shifts and show that although a number of shifts are tolerated, shifts in the covariate distribution of the prompts are not. Motivated by this, we consider a generalized ICL setting where the covariate distributions can vary across prompts. We show that although gradient flow succeeds at finding a global minimum in this setting, the trained transformer is still brittle under mild covariate shifts. We complement this finding with experiments on large, nonlinear transformer architectures which we show are more robust under covariate shifts."
                },
                {
                    "title": "Transformers learn through gradual rank increase",
                    "abstract": "We identify incremental learning dynamics in transformers, where the difference between trained and initial weights progressively increases in rank. We rigorously prove this occurs under the simplifying assumptions of diagonal weight matrices and small initialization. Our experiments support the theory and also show that phenomenon can occur in practice without the simplifying assumptions."
                },
                {
                    "title": "Transformers as Statisticians: Provable In-Context Learning with In-Context Algorithm Selection",
                    "abstract": "Neural sequence models based on the transformer architecture have demonstrated remarkable \\emph{in-context learning} (ICL) abilities, where they can perform new tasks when prompted with training and test examples, without any parameter update to the model. This work first provides a comprehensive statistical theory for transformers to perform ICL. Concretely, we show that transformers can implement a broad class of standard machine learning algorithms in context, such as least squares, ridge regression, Lasso, learning generalized linear models, and gradient descent on two-layer neural networks, with near-optimal predictive power on various in-context data distributions. Using an efficient implementation of in-context gradient descent as the underlying mechanism, our transformer constructions admit mild size bounds, and can be learned with polynomially many pretraining sequences. Building on these ``base'' ICL algorithms, intriguingly, we show that transformers can implement more complex ICL procedures involving \\emph{in-context algorithm selection}, akin to what a statistician can do in real life -- A \\emph{single} transformer can adaptively select different base ICL algorithms -- or even perform qualitatively different tasks -- on different input sequences, without any explicit prompting of the right algorithm or task. We both establish this in theory by explicit constructions, and also observe this phenomenon experimentally. In theory, we construct two general mechanisms for algorithm selection with concrete examples: pre-ICL testing, and post-ICL validation. As an example, we use the post-ICL validation mechanism to construct a transformer that can perform nearly Bayes-optimal ICL on a challenging task -- noisy linear models with mixed noise levels. Experimentally, we demonstrate the strong in-context algorithm selection capabilities of standard transformer architectures."
                },
                {
                    "title": "What and How does In-Context Learning Learn? Bayesian Model Averaging, Parameterization, and Generalization",
                    "abstract": "In this paper, we conduct a comprehensive study of In-Context Learning (ICL) by addressing several open questions: (a) What type of ICL estimator is learned by large language models? (b) What is a proper performance metric for ICL and what is the error rate? (c) How does the transformer architecture enable ICL? To answer these questions, we adopt a Bayesian view and formulate ICL as a problem of predicting the response corresponding to the current covariate, given a number of examples drawn from a latent variable model. To answer (a), we show that, without updating the neural network parameters, ICL implicitly implements the Bayesian model averaging algorithm, which is proven to be approximately parameterized by the attention mechanism. For (b), we analyze the ICL performance from an online learning perspective and establish a $\\mathcal{O}(1/T)$ regret bound for perfectly pretrained ICL, where $T$ is the number of examples in the prompt. To answer (c), we show that, in addition to encoding Bayesian model averaging via attention, the transformer architecture also enables a fine-grained statistical analysis of pretraining under realistic assumptions. In particular, we prove that the error of pretrained model is bounded by a sum of an approximation error and a generalization error, where the former decays to zero exponentially as the depth grows, and the latter decays to zero sublinearly with the number of tokens in the pretraining dataset. Our results provide a unified understanding of the transformer and its ICL ability with bounds on ICL regret, approximation, and generalization, which deepens our knowledge of these essential aspects of modern language models."
                },
                {
                    "title": "Scan and Snap: Understanding Training Dynamics and Token Composition in 1-layer Transformer",
                    "abstract": "Transformer architecture has shown impressive performance in multiple research domains and has become the backbone of many neural network models. However, there is limited understanding on how it works. In particular, with a simple predictive loss, how the representation emerges from the gradient \\emph{training dynamics} remains a mystery. In this paper, for 1-layer transformer with one self-attention layer plus one decoder layer, we analyze its SGD training dynamics for the task of next token prediction in a mathematically rigorous manner. We open the black box of the dynamic process of how the self-attention layer combines input tokens, and reveal the nature of underlying inductive bias. More specifically, with the assumption (a) no positional encoding, (b) long input sequence, and (c) the decoder layer learns faster than the self-attention layer, we prove that self-attention acts as a \\emph{discriminative scanning algorithm}: starting from uniform attention, it gradually attends more to distinct key tokens for a specific next token to be predicted, and pays less attention to common key tokens that occur across different next tokens. Among distinct tokens, it progressively drops attention weights, following the order of low to high co-occurrence between the key and the query token in the training set. Interestingly, this procedure does not lead to winner-takes-all, but decelerates due to a \\emph{phase transition} that is controllable by the learning rates of the two layers, leaving (almost) fixed token combination. We verify this \\textbf{\\emph{scan and snap}} dynamics on synthetic and real-world data (WikiText)."
                },
                {
                    "title": "TinyStories: How Small Can Language Models Be and Still Speak Coherent English?",
                    "abstract": "Language models (LMs) are powerful tools for natural language processing, but they often struggle to produce coherent and fluent text when they are small. Models with around 125M parameters such as GPT-Neo (small) or GPT-2 (small) can rarely generate coherent and consistent English text beyond a few words even after extensive training. This raises the question of whether the emergence of the ability to produce coherent English text only occurs at larger scales (with hundreds of millions of parameters or more) and complex architectures (with many layers of global attention). In this work, we introduce TinyStories, a synthetic dataset of short stories that only contain words that a typical 3 to 4-year-olds usually understand, generated by GPT-3.5 and GPT-4. We show that TinyStories can be used to train and evaluate LMs that are much smaller than the state-of-the-art models (below 10 million total parameters), or have much simpler architectures (with only one transformer block), yet still produce fluent and consistent stories with several paragraphs that are diverse and have almost perfect grammar, and demonstrate reasoning capabilities. We also introduce a new paradigm for the evaluation of language models: We suggest a framework which uses GPT-4 to grade the content generated by these models as if those were stories written by students and graded by a (human) teacher. This new paradigm overcomes the flaws of standard benchmarks which often requires the model's output to be very structures, and moreover provides a multidimensional score for the model, providing scores for different capabilities such as grammar, creativity and consistency. We hope that TinyStories can facilitate the development, analysis and research of LMs, especially for low-resource or specialized domains, and shed light on the emergence of language capabilities in LMs."
                },
                {
                    "title": "Mapping of attention mechanisms to a generalized Potts model",
                    "abstract": "Transformers are neural networks that revolutionized natural language processing and machine learning. They process sequences of inputs, like words, using a mechanism called self-attention, which is trained via masked language modeling (MLM). In MLM, a word is randomly masked in an input sequence, and the network is trained to predict the missing word. Despite the practical success of transformers, it remains unclear what type of data distribution self-attention can learn efficiently. Here, we show analytically that if one decouples the treatment of word positions and embeddings, a single layer of self-attention learns the conditionals of a generalized Potts model with interactions between sites and Potts colors. Moreover, we show that training this neural network is exactly equivalent to solving the inverse Potts problem by the so-called pseudo-likelihood method, well known in statistical physics. Using this mapping, we compute the generalization error of self-attention in a model scenario analytically using the replica method."
                },
                {
                    "title": "Dynamics of Finite Width Kernel and Prediction Fluctuations in Mean Field Neural Networks",
                    "abstract": "\n We analyze the dynamics of finite width effects in wide but finite feature learning neural networks. Starting from a dynamical mean field theory description of infinite width deep neural network kernel and prediction dynamics, we provide a characterization of the \n \n \n \n \n O\n \n (\n 1\n \n /\n \n \n width\n \n )\n \n \n fluctuations of the dynamical mean field theory order parameters over random initializations of the network weights. Our results, while perturbative in width, unlike prior analyses, are non-perturbative in the strength of feature learning. We find that once the mean field/\u00b5P parameterization is adopted, the leading finite size effect on the dynamics is to introduce initialization variance in the predictions and feature kernels of the networks. In the lazy limit of network training, all kernels are random but static in time and the prediction variance has a universal form. However, in the rich, feature learning regime, the fluctuations of the kernels and predictions are dynamically coupled with a variance that can be computed self-consistently. In two layer networks, we show how feature learning can dynamically reduce the variance of the final tangent kernel and final network predictions. We also show how initialization variance can slow down online learning in wide but finite networks. In deeper networks, kernel variance can dramatically accumulate through subsequent layers at large feature learning strengths, but feature learning continues to improve the signal-to-noise ratio of the feature kernels. In discrete time, we demonstrate that large learning rate phenomena such as edge of stability effects can be well captured by infinite width dynamics and that initialization variance can decrease dynamically. For convolutional neural networks trained on CIFAR-10, we empirically find significant corrections to both the bias and variance of network dynamics due to finite width."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "A Theoretical Understanding of shallow Vision Transformers: Learning, Generalization, and Sample Complexity",
                    "abstract": "Vision Transformers (ViTs) with self-attention modules have recently achieved great empirical success in many vision tasks. Due to non-convex interactions across layers, however, theoretical learning and generalization analysis is mostly elusive. Based on a data model characterizing both label-relevant and label-irrelevant tokens, this paper provides the first theoretical analysis of training a shallow ViT, i.e., one self-attention layer followed by a two-layer perceptron, for a classification task. We characterize the sample complexity to achieve a zero generalization error. Our sample complexity bound is positively correlated with the inverse of the fraction of label-relevant tokens, the token noise level, and the initial model error. We also prove that a training process using stochastic gradient descent (SGD) leads to a sparse attention map, which is a formal verification of the general intuition about the success of attention. Moreover, this paper indicates that a proper token sparsification can improve the test performance by removing label-irrelevant and/or noisy tokens, including spurious correlations. Empirical experiments on synthetic data and CIFAR-10 dataset justify our theoretical results and generalize to deeper ViTs."
                },
                {
                    "title": "Optimal Learning of Deep Random Networks of Extensive-width",
                    "abstract": "We consider the problem of learning a target function corresponding to a deep, extensive-width, non-linear neural network with random Gaussian weights. We consider the asymptotic limit where the number of samples, the input dimension and the network width are proportionally large. We propose a closed-form expression for the Bayes-optimal test error, for regression and classification tasks. We further compute closed-form expressions for the test errors of ridge regression, kernel and random features regression. We find, in particular, that optimally regularized ridge regression, as well as kernel regression, achieve Bayes-optimal performances, while the logistic loss yields a near-optimal test error for classification. We further show numerically that when the number of samples grows faster than the dimension, ridge and kernel methods become suboptimal, while neural networks achieve test error close to zero from quadratically many samples."
                },
                {
                    "title": "Transformers as Algorithms: Generalization and Stability in In-context Learning",
                    "abstract": "In-context learning (ICL) is a type of prompting where a transformer model operates on a sequence of (input, output) examples and performs inference on-the-fly. In this work, we formalize in-context learning as an algorithm learning problem where a transformer model implicitly constructs a hypothesis function at inference-time. We first explore the statistical aspects of this abstraction through the lens of multitask learning: We obtain generalization bounds for ICL when the input prompt is (1) a sequence of i.i.d. (input, label) pairs or (2) a trajectory arising from a dynamical system. The crux of our analysis is relating the excess risk to the stability of the algorithm implemented by the transformer. We characterize when transformer/attention architecture provably obeys the stability condition and also provide empirical verification. For generalization on unseen tasks, we identify an inductive bias phenomenon in which the transfer learning risk is governed by the task complexity and the number of MTL tasks in a highly predictable manner. Finally, we provide numerical evaluations that (1) demonstrate transformers can indeed implement near-optimal algorithms on classical regression problems with i.i.d. and dynamic data, (2) provide insights on stability, and (3) verify our theoretical predictions."
                },
                {
                    "title": "Bayesian interpolation with deep linear networks",
                    "abstract": "Significance Understanding the interplay between network architecture, dataset statistics, and learning algorithms is a key challenge in deep learning. We overcome this challenge analytically for zero-noise Bayesian inference in neural networks with linear activations and output dimension one but arbitrary depth, width, and training data. Our results provide the first rigorous solutions to learning with neural networks in which depth, width, and dataset can vary freely. We identify an emergent notion of effective depth, which combines depth, width, and dataset size, that must be large for networks to compute optimal posteriors from universal priors. Our results both present novel examples of models admitting exact solutions to Bayesian inference and give the first principled Bayesian justification for preferring deeper networks."
                },
                {
                    "title": "Globally Gated Deep Linear Networks",
                    "abstract": "Recently proposed Gated Linear Networks present a tractable nonlinear network architecture, and exhibit interesting capabilities such as learning with local error signals and reduced forgetting in sequential learning. In this work, we introduce a novel gating architecture, named Globally Gated Deep Linear Networks (GGDLNs) where gating units are shared among all processing units in each layer, thereby decoupling the architectures of the nonlinear but unlearned gatings and the learned linear processing motifs. We derive exact equations for the generalization properties in these networks in the finite-width thermodynamic limit, defined by $P,N\\rightarrow\\infty, P/N\\sim O(1)$, where P and N are the training sample size and the network width respectively. We find that the statistics of the network predictor can be expressed in terms of kernels that undergo shape renormalization through a data-dependent matrix compared to the GP kernels. Our theory accurately captures the behavior of finite width GGDLNs trained with gradient descent dynamics. We show that kernel shape renormalization gives rise to rich generalization properties w.r.t. network width, depth and L2 regularization amplitude. Interestingly, networks with sufficient gating units behave similarly to standard ReLU networks. Although gatings in the model do not participate in supervised learning, we show the utility of unsupervised learning of the gating parameters. Additionally, our theory allows the evaluation of the network's ability for learning multiple tasks by incorporating task-relevant information into the gating units. In summary, our work is the first exact theoretical solution of learning in a family of nonlinear networks with finite width. The rich and diverse behavior of the GGDLNs suggests that they are helpful analytically tractable models of learning single and multiple tasks, in finite-width nonlinear deep networks."
                },
                {
                    "title": "Vision Transformers provably learn spatial structure",
                    "abstract": "Vision Transformers (ViTs) have achieved comparable or superior performance than Convolutional Neural Networks (CNNs) in computer vision. This empirical breakthrough is even more remarkable since, in contrast to CNNs, ViTs do not embed any visual inductive bias of spatial locality. Yet, recent works have shown that while minimizing their training loss, ViTs specifically learn spatially localized patterns. This raises a central question: how do ViTs learn these patterns by solely minimizing their training loss using gradient-based methods from random initialization? In this paper, we provide some theoretical justification of this phenomenon. We propose a spatially structured dataset and a simplified ViT model. In this model, the attention matrix solely depends on the positional encodings. We call this mechanism the positional attention mechanism. On the theoretical side, we consider a binary classification task and show that while the learning problem admits multiple solutions that generalize, our model implicitly learns the spatial structure of the dataset while generalizing: we call this phenomenon patch association. We prove that patch association helps to sample-efficiently transfer to downstream datasets that share the same structure as the pre-training one but differ in the features. Lastly, we empirically verify that a ViT with positional attention performs similarly to the original one on CIFAR-10/100, SVHN and ImageNet."
                },
                {
                    "title": "Self-consistent dynamical field theory of kernel evolution in wide neural networks",
                    "abstract": "We analyze feature learning in infinite-width neural networks trained with gradient flow through a self-consistent dynamical field theory. We construct a collection of deterministic dynamical order parameters which are inner-product kernels for hidden unit activations and gradients in each layer at pairs of time points, providing a reduced description of network activity through training. These kernel order parameters collectively define the hidden layer activation distribution, the evolution of the neural tangent kernel (NTK), and consequently, output predictions. We show that the field theory derivation recovers the recursive stochastic process of infinite-width feature learning networks obtained by Yang and Hu with tensor programs. For deep linear networks, these kernels satisfy a set of algebraic matrix equations. For nonlinear networks, we provide an alternating sampling procedure to self-consistently solve for the kernel order parameters. We provide comparisons of the self-consistent solution to various approximation schemes including the static NTK approximation, gradient independence assumption, and leading order perturbation theory, showing that each of these approximations can break down in regimes where general self-consistent solutions still provide an accurate description. Lastly, we provide experiments in more realistic settings which demonstrate that the loss and kernel dynamics of convolutional neural networks at fixed feature learning strength are preserved across different widths on a image classification task."
                },
                {
                    "title": "Unraveling Attention via Convex Duality: Analysis and Interpretations of Vision Transformers",
                    "abstract": "Vision transformers using self-attention or its proposed alternatives have demonstrated promising results in many image related tasks. However, the underpinning inductive bias of attention is not well understood. To address this issue, this paper analyzes attention through the lens of convex duality. For the non-linear dot-product self-attention, and alternative mechanisms such as MLP-mixer and Fourier Neural Operator (FNO), we derive equivalent finite-dimensional convex problems that are interpretable and solvable to global optimality. The convex programs lead to {\\it block nuclear-norm regularization} that promotes low rank in the latent feature and token dimensions. In particular, we show how self-attention networks implicitly clusters the tokens, based on their latent similarity. We conduct experiments for transferring a pre-trained transformer backbone for CIFAR-100 classification by fine-tuning a variety of convex attention heads. The results indicate the merits of the bias induced by attention compared with the existing MLP or linear heads."
                },
                {
                    "title": "Inductive Biases and Variable Creation in Self-Attention Mechanisms",
                    "abstract": "Self-attention, an architectural motif designed to model long-range interactions in sequential data, has driven numerous recent breakthroughs in natural language processing and beyond. This work provides a theoretical analysis of the inductive biases of self-attention modules. Our focus is to rigorously establish which functions and long-range dependencies self-attention blocks prefer to represent. Our main result shows that bounded-norm Transformer networks\"create sparse variables\": a single self-attention head can represent a sparse function of the input sequence, with sample complexity scaling only logarithmically with the context length. To support our analysis, we present synthetic experiments to probe the sample complexity of learning sparse Boolean functions with Transformers."
                },
                {
                    "title": "Statistical Mechanics of Deep Linear Neural Networks: The Backpropagating Kernel Renormalization",
                    "abstract": "The success of deep learning in many real-world tasks has triggered an intense effort to understand the power and limitations of deep learning in the training and generalization of complex tasks, so far with limited progress. In this work, we study the statistical mechanics of learning in Deep Linear Neural Networks (DLNNs) in which the input-output function of an individual unit is linear. Despite the linearity of the units, learning in DLNNs is nonlinear, hence studying its properties reveals some of the features of nonlinear Deep Neural Networks (DNNs). Importantly, we solve exactly the network properties following supervised learning using an equilibrium Gibbs distribution in the weight space. To do this, we introduce the Back-Propagating Kernel Renormalization (BPKR), which allows for the incremental integration of the network weights starting from the network output layer and progressing backward until the first layer's weights are integrated out. This procedure allows us to evaluate important network properties, such as its generalization error, the role of network width and depth, the impact of the size of the training set, and the effects of weight regularization and learning stochasticity. BPKR does not assume specific statistics of the input or the task's output. Furthermore, by performing partial integration of the layers, the BPKR allows us to compute the properties of the neural representations across the different hidden layers. We have proposed an extension of the BPKR to nonlinear DNNs with ReLU. Surprisingly, our numerical simulations reveal that despite the nonlinearity, the predictions of our theory are largely shared by ReLU networks in a wide regime of parameters. Our work is the first exact statistical mechanical study of learning in a family of DNNs, and the first successful theory of learning through successive integration of DoFs in the learned weight space."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Infinite attention: NNGP and NTK for deep attention networks",
                    "abstract": "There is a growing amount of literature on the relationship between wide neural networks (NNs) and Gaussian processes (GPs), identifying an equivalence between the two for a variety of NN architectures. This equivalence enables, for instance, accurate approximation of the behaviour of wide Bayesian NNs without MCMC or variational approximations, or characterisation of the distribution of randomly initialised wide NNs optimised by gradient descent without ever running an optimiser. We provide a rigorous extension of these results to NNs involving attention layers, showing that unlike single-head attention, which induces non-Gaussian behaviour, multi-head attention architectures behave as GPs as the number of heads tends to infinity. We further discuss the effects of positional encodings and layer normalisation, and propose modifications of the attention mechanism which lead to improved results for both finite and infinitely wide NNs. We evaluate attention kernels empirically, leading to a moderate improvement upon the previous state-of-the-art on CIFAR-10 for GPs without trainable kernels and advanced data preprocessing. Finally, we introduce new features to the Neural Tangents library (Novak et al., 2020) allowing applications of NNGP/NTK models, with and without attention, to variable-length sequences, with an example on the IMDb reviews dataset."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Are Transformers universal approximators of sequence-to-sequence functions?",
                    "abstract": "Despite the widespread adoption of Transformer models for NLP tasks, the expressive power of these models is not well-understood. In this paper, we establish that Transformer models are universal approximators of continuous permutation equivariant sequence-to-sequence functions with compact support, which is quite surprising given the amount of shared parameters in these models. Furthermore, using positional encodings, we circumvent the restriction of permutation equivariance, and show that Transformer models can universally approximate arbitrary continuous sequence-to-sequence functions on a compact domain. Interestingly, our proof techniques clearly highlight the different roles of the self-attention and the feed-forward layers in Transformers. In particular, we prove that fixed width self-attention layers can compute contextual mappings of the input sequences, playing a key role in the universal approximation property of Transformers. Based on this insight from our analysis, we consider other architectures that can compute contextual mappings and empirically evaluate them."
                },
                {
                    "title": "Composable Effects for Flexible and Accelerated Probabilistic Programming in NumPyro",
                    "abstract": "NumPyro is a lightweight library that provides an alternate NumPy backend to the Pyro probabilistic programming language with the same modeling interface, language primitives and effect handling abstractions. Effect handlers allow Pyro's modeling API to be extended to NumPyro despite its being built atop a fundamentally different JAX-based functional backend. In this work, we demonstrate the power of composing Pyro's effect handlers with the program transformations that enable hardware acceleration, automatic differentiation, and vectorization in JAX. In particular, NumPyro provides an iterative formulation of the No-U-Turn Sampler (NUTS) that can be end-to-end JIT compiled, yielding an implementation that is much faster than existing alternatives in both the small and large dataset regimes."
                },
                {
                    "title": "Torchmeta: A Meta-Learning library for PyTorch",
                    "abstract": "The constant introduction of standardized benchmarks in the literature has helped accelerating the recent advances in meta-learning research. They offer a way to get a fair comparison between different algorithms, and the wide range of datasets available allows full control over the complexity of this evaluation. However, for a large majority of code available online, the data pipeline is often specific to one dataset, and testing on another dataset requires significant rework. We introduce Torchmeta, a library built on top of PyTorch that enables seamless and consistent evaluation of meta-learning algorithms on multiple datasets, by providing data-loaders for most of the standard benchmarks in few-shot classification and regression, with a new meta-dataset abstraction. It also features some extensions for PyTorch to simplify the development of models compatible with meta-learning algorithms. The code is available here: this https URL"
                },
                {
                    "title": "Disentangling feature and lazy training in deep neural networks",
                    "abstract": "Two distinct limits for deep learning have been derived as the network width h \u2192 \u221e, depending on how the weights of the last layer scale with h. In the neural tangent Kernel (NTK) limit, the dynamics becomes linear in the weights and is described by a frozen kernel \u0398 (the NTK). By contrast, in the mean-field limit, the dynamics can be expressed in terms of the distribution of the parameters associated with a neuron, that follows a partial differential equation. In this work we consider deep networks where the weights in the last layer scale as \u03b1h \u22121/2 at initialization. By varying \u03b1 and h, we probe the crossover between the two limits. We observe two the previously identified regimes of \u2018lazy training\u2019 and \u2018feature training\u2019. In the lazy-training regime, the dynamics is almost linear and the NTK barely changes after initialization. The feature-training regime includes the mean-field formulation as a limiting case and is characterized by a kernel that evolves in time, and thus learns some features. We perform numerical experiments on MNIST, Fashion-MNIST, EMNIST and CIFAR10 and consider various architectures. We find that: (i) the two regimes are separated by an \u03b1* that scales as 1h . (ii) Network architecture and data structure play an important role in determining which regime is better: in our tests, fully-connected networks perform generally better in the lazy-training regime, unlike convolutional networks. (iii) In both regimes, the fluctuations \u03b4F induced on the learned function by initial conditions decay as \u03b4F\u223c1/h , leading to a performance that increases with h. The same improvement can also be obtained at an intermediate width by ensemble-averaging several networks that are trained independently. (iv) In the feature-training regime we identify a time scale t1\u223ch\u03b1 , such that for t \u226a t 1 the dynamics is linear. At t \u223c t 1, the output has grown by a magnitude h and the changes of the tangent kernel | |\u0394\u0398| | become significant. Ultimately, it follows ||\u0394\u0398||\u223c(h\u03b1)\u2212a for ReLU and Softplus activation functions, with a < 2 and a \u2192 2 as depth grows. We provide scaling arguments supporting these findings."
                },
                {
                    "title": "Theoretical Limitations of Self-Attention in Neural Sequence Models",
                    "abstract": "Transformers are emerging as the new workhorse of NLP, showing great success across tasks. Unlike LSTMs, transformers process input sequences entirely through self-attention. Previous work has suggested that the computational capabilities of self-attention to process hierarchical structures are limited. In this work, we mathematically investigate the computational power of self-attention to model formal languages. Across both soft and hard attention, we show strong theoretical limitations of the computational abilities of self-attention, finding that it cannot model periodic finite-state languages, nor hierarchical structure, unless the number of layers or heads increases with input length. These limitations seem surprising given the practical success of self-attention and the prominent role assigned to hierarchical structure in linguistics, suggesting that natural language can be approximated well with models that are too weak for the formal languages typically assumed in theoretical linguistics."
                },
                {
                    "title": "Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned",
                    "abstract": "Multi-head self-attention is a key component of the Transformer, a state-of-the-art architecture for neural machine translation. In this work we evaluate the contribution made by individual attention heads to the overall performance of the model and analyze the roles played by them in the encoder. We find that the most important and confident heads play consistent and often linguistically-interpretable roles. When pruning heads using a method based on stochastic gates and a differentiable relaxation of the L0 penalty, we observe that specialized heads are last to be pruned. Our novel pruning method removes the vast majority of heads without seriously affecting performance. For example, on the English-Russian WMT dataset, pruning 38 out of 48 encoder heads results in a drop of only 0.15 BLEU."
                },
                {
                    "title": "A mathematical theory of semantic development in deep neural networks",
                    "abstract": "Significance Over the course of development, humans learn myriad facts about items in the world, and naturally group these items into useful categories and structures. This semantic knowledge is essential for diverse behaviors and inferences in adulthood. How is this richly structured semantic knowledge acquired, organized, deployed, and represented by neuronal networks in the brain? We address this question by studying how the nonlinear learning dynamics of deep linear networks acquires information about complex environmental structures. Our results show that this deep learning dynamics can self-organize emergent hidden representations in a manner that recapitulates many empirical phenomena in human semantic development. Such deep networks thus provide a mathematically tractable window into the development of internal neural representations through experience. An extensive body of empirical research has revealed remarkable regularities in the acquisition, organization, deployment, and neural representation of human semantic knowledge, thereby raising a fundamental conceptual question: What are the theoretical principles governing the ability of neural networks to acquire, organize, and deploy abstract knowledge by integrating across many individual experiences? We address this question by mathematically analyzing the nonlinear dynamics of learning in deep linear networks. We find exact solutions to this learning dynamics that yield a conceptual explanation for the prevalence of many disparate phenomena in semantic cognition, including the hierarchical differentiation of concepts through rapid developmental transitions, the ubiquity of semantic illusions between such transitions, the emergence of item typicality and category coherence as factors controlling the speed of semantic processing, changing patterns of inductive projection over development, and the conservation of semantic similarity in neural representations across species. Thus, surprisingly, our simple neural model qualitatively recapitulates many diverse regularities underlying semantic development, while providing analytic insight into how the statistical structure of an environment can interact with nonlinear deep-learning dynamics to give rise to these regularities."
                },
                {
                    "title": "Pyro: Deep Universal Probabilistic Programming",
                    "abstract": "Pyro is a probabilistic programming language built on Python as a platform for developing advanced probabilistic models in AI research. To scale to large datasets and high-dimensional models, Pyro uses stochastic variational inference algorithms and probability distributions built on top of PyTorch, a modern GPU-accelerated deep learning framework. To accommodate complex or model-specific algorithmic behavior, Pyro leverages Poutine, a library of composable building blocks for modifying the behavior of probabilistic programs."
                },
                {
                    "title": "Gaussian Process Behaviour in Wide Deep Neural Networks",
                    "abstract": "Whilst deep neural networks have shown great empirical success, there is still much work to be done to understand their theoretical properties. In this paper, we study the relationship between random, wide, fully connected, feedforward networks with more than one hidden layer and Gaussian processes with a recursive kernel definition. We show that, under broad conditions, as we make the architecture increasingly wide, the implied random function converges in distribution to a Gaussian process, formalising and extending existing results by Neal (1996) to deep networks. To evaluate convergence rates empirically, we use maximum mean discrepancy. We then compare finite Bayesian deep networks from the literature to Gaussian processes in terms of the key predictive quantities of interest, finding that in some cases the agreement can be very close. We discuss the desirability of Gaussian process behaviour and review non-Gaussian alternative models from the literature."
                },
                {
                    "title": "Deep Neural Networks as Gaussian Processes",
                    "abstract": "It has long been known that a single-layer fully-connected neural network with an i.i.d. prior over its parameters is equivalent to a Gaussian process (GP), in the limit of infinite network width. This correspondence enables exact Bayesian inference for infinite width neural networks on regression tasks by means of evaluating the corresponding GP. Recently, kernel functions which mimic multi-layer random neural networks have been developed, but only outside of a Bayesian framework. As such, previous work has not identified that these kernels can be used as covariance functions for GPs and allow fully Bayesian prediction with a deep neural network. \nIn this work, we derive the exact equivalence between infinitely wide deep networks and GPs. We further develop a computationally efficient pipeline to compute the covariance function for these GPs. We then use the resulting GPs to perform Bayesian inference for wide deep neural networks on MNIST and CIFAR-10. We observe that trained neural network accuracy approaches that of the corresponding GP with increasing layer width, and that the GP uncertainty is strongly correlated with trained network prediction error. We further find that test performance increases as finite-width trained networks are made wider and more similar to a GP, and thus that GP predictions typically outperform those of finite-width networks. Finally we connect the performance of these GPs to the recent theory of signal propagation in random neural networks."
                },
                {
                    "title": "Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms",
                    "abstract": "We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at this https URL"
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Structured Attention Networks",
                    "abstract": "Attention networks have proven to be an effective approach for embedding categorical inference within a deep neural network. However, for many tasks we may want to model richer structural dependencies without abandoning end-to-end training. In this work, we experiment with incorporating richer structural distributions, encoded using graphical models, within deep networks. We show that these structured attention networks are simple extensions of the basic attention procedure, and that they allow for extending attention beyond the standard soft-selection approach, such as attending to partial segmentations or to subtrees. We experiment with two different classes of structured attention networks: a linear-chain conditional random field and a graph-based parsing model, and describe how these models can be practically implemented as neural network layers. Experiments show that this approach is effective for incorporating structural biases, and structured attention networks outperform baseline attention models on a variety of synthetic and real tasks: tree transduction, neural machine translation, question answering, and natural language inference. We further find that models trained in this way learn interesting unsupervised hidden representations that generalize simple attention."
                },
                {
                    "title": "A Conceptual Introduction to Hamiltonian Monte Carlo",
                    "abstract": "Hamiltonian Monte Carlo has proven a remarkable empirical success, but only recently have we begun to develop a rigorous understanding of why it performs so well on difficult problems and how it is best applied in practice. Unfortunately, that understanding is confined within the mathematics of differential geometry which has limited its dissemination, especially to the applied communities for which it is particularly important. In this review I provide a comprehensive conceptual account of these theoretical foundations, focusing on developing a principled intuition behind the method and its optimal implementations rather of any exhaustive rigor. Whether a practitioner or a statistician, the dedicated reader will acquire a solid grasp of how Hamiltonian Monte Carlo works, when it succeeds, and, perhaps most importantly, when it fails."
                },
                {
                    "title": "Matching Networks for One Shot Learning",
                    "abstract": "Learning from a few examples remains a key challenge in machine learning. Despite recent advances in important domains such as vision and language, the standard supervised deep learning paradigm does not offer a satisfactory solution for learning new concepts rapidly from little data. In this work, we employ ideas from metric learning based on deep neural features and from recent advances that augment neural networks with external memories. Our framework learns a network that maps a small labelled support set and an unlabelled example to its label, obviating the need for fine-tuning to adapt to new class types. We then define one-shot learning problems on vision (using Omniglot, ImageNet) and language tasks. Our algorithm improves one-shot accuracy on ImageNet from 87.6% to 93.2% and from 88.0% to 93.8% on Omniglot compared to competing approaches. We also demonstrate the usefulness of the same model on language modeling by introducing a one-shot task on the Penn Treebank."
                },
                {
                    "title": "Human-level concept learning through probabilistic program induction",
                    "abstract": "Handwritten characters drawn by a model Not only do children learn effortlessly, they do so quickly and with a remarkable ability to use what they have learned as the raw material for creating new stuff. Lake et al. describe a computational model that learns in a similar fashion and does so better than current deep learning algorithms. The model classifies, parses, and recreates handwritten characters, and can generate new letters of the alphabet that look \u201cright\u201d as judged by Turing-like tests of the model's output in comparison to what real humans produce. Science, this issue p. 1332 Combining the capacity to handle noise with probabilistic learning yields humanlike performance in a computational model. People learning new concepts can often generalize successfully from just a single example, yet machine learning algorithms typically require tens or hundreds of examples to perform with similar accuracy. People can also use learned concepts in richer ways than conventional algorithms\u2014for action, imagination, and explanation. We present a computational model that captures these human learning abilities for a large class of simple visual concepts: handwritten characters from the world\u2019s alphabets. The model represents concepts as simple programs that best explain observed examples under a Bayesian criterion. On a challenging one-shot classification task, the model achieves human-level performance while outperforming recent deep learning approaches. We also present several \u201cvisual Turing tests\u201d probing the model\u2019s creative generalization abilities, which in many cases are indistinguishable from human behavior."
                },
                {
                    "title": "Neural Machine Translation by Jointly Learning to Align and Translate",
                    "abstract": "Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition."
                },
                {
                    "title": "A Practical Bayesian Framework for Backpropagation Networks",
                    "abstract": "A quantitative and practical Bayesian framework is described for learning of mappings in feedforward networks. The framework makes possible (1) objective comparisons between solutions using alternative network architectures, (2) objective stopping rules for network pruning or growing procedures, (3) objective choice of magnitude and type of weight decay terms or additive regularizers (for penalizing large weights, etc.), (4) a measure of the effective number of well-determined parameters in a model, (5) quantified estimates of the error bars on network parameters and on network output, and (6) objective comparisons with alternative learning and interpolation models such as splines and radial basis functions. The Bayesian \"evidence\" automatically embodies \"Occam's razor,\" penalizing overflexible and overcomplex models. The Bayesian approach helps detect poor underlying assumptions in learning models. For learning models well matched to a problem, a good correlation between generalization ability and the Bayesian evidence is obtained."
                },
                {
                    "title": "Consistent inference of probabilities in layered networks: predictions and generalizations",
                    "abstract": "The problem of learning a general input-output relation using a layered neural network is discussed in a statistical framework. By imposing the consistency condition that the error minimization be equivalent to a likelihood maximization for training the network, the authors arrive at a Gibbs distribution on a canonical ensemble of networks with the same architecture. This statistical description enables them to evaluate the probability of a correct prediction of an independent example, after training the network on a given training set. The prediction probability is highly correlated with the generalization ability of the network, as measured outside the training set. This suggests a general and practical criterion for training layered networks by minimizing prediction errors. The authors demonstrate the utility of this criterion for selecting the optimal architecture in the continuity problem. As a theoretical application of the statistical formalism, they discuss the question of learning curves and estimate the sufficient training size needed for correct generalization, in a simple example.<<ETX>>"
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "The mnist database of handwritten digits",
                    "abstract": "Disclosed is an improved articulated bar flail having shearing edges for efficiently shredding materials. An improved shredder cylinder is disclosed with a plurality of these flails circumferentially spaced and pivotally attached to the periphery of a rotatable shaft. Also disclosed is an improved shredder apparatus which has a pair of these shredder cylinders mounted to rotate about spaced parallel axes which cooperates with a conveyer apparatus which has a pair of inclined converging conveyer belts with one of the belts mounted to move with respect to the other belt to allow the transport of articles of various sizes therethrough."
                }
            ],
            "categories": [
                "cs.LG",
                "cond-mat.dis-nn",
                "cond-mat.stat-mech",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we characterize the interplay of attention paths in multi-head self-attention models, such as Transformers, to improve interpretability and generalization?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the theoretical underpinnings of Transformer models, which have become foundational in natural language processing and computer vision. By understanding the interplay of attention paths, we can enhance model interpretability, leading to better insights into how these models make decisions. This could pave the way for more robust and generalizable models, influencing future research directions in deep learning and potentially leading to practical applications in various domains, such as automated decision-making systems and personalized AI.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexity of multi-head, multi-layer architectures, where the interactions between attention paths are not straightforward. Naive approaches may fail because they often rely on simplifying assumptions that overlook the intricate relationships between heads and layers. Additionally, the theoretical characterization of these interactions is hindered by the limitations of existing models, which either do not define attention paths or impose constraints that obscure their interplay. Overcoming these technical and theoretical obstacles requires a nuanced understanding of both the architecture and the statistical mechanics involved.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on simplified models that do not capture the full complexity of multi-head, multi-layer architectures. Many studies have either ignored the concept of attention paths or have approached the problem from angles that do not address their interplay. Barriers such as the lack of analytical tractability in finite-width networks and the reliance on the infinite-width limit have prevented a comprehensive understanding of attention paths. Our approach differs by applying statistical mechanics to a deep multi-head self-attention model, allowing us to derive exact equations and insights that were previously inaccessible.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves applying statistical mechanics theory to a deep multi-head self-attention model, focusing on the interplay of attention paths. We will derive exact equations for predictor statistics under Bayesian learning, analyze the kernel combination mechanism that emerges beyond the Gaussian process limit, and provide interpretability by relating this mechanism to learned weight correlations. We will validate our findings using both synthetic and real-world sequence classification tasks, with expected outcomes including improved generalization through"
            }
        },
        "author_data": {
            "ed8c9bf9-950c-4cf7-ab2e-a19e7e3aafed": {
                "pk": "ed8c9bf9-950c-4cf7-ab2e-a19e7e3aafed",
                "name": "Lorenzo Tiberi",
                "collaborators": [
                    "David Dahmen",
                    "Moritz Helias",
                    "Jonas Stapmanns",
                    "Tobias K\u00fchn",
                    "Thomas Luu",
                    "Chiara Favaretto",
                    "Mario Innocenti",
                    "Danielle S. Bassett",
                    "Fabio Pasqualetti"
                ],
                "domain": [
                    "Computational Neuroscience",
                    "Network Dynamics",
                    "Synchronization",
                    "Criticality"
                ],
                "publications": [
                    {
                        "title": "Hidden connectivity structures control collective network dynamics",
                        "abstract": "Many observables of brain dynamics appear to be optimized for computation. Which connectivity structures underlie this fine-tuning? We propose that many of these structures are naturally encoded in the space that more directly relates to network dynamics - the space of the connectivity eigenmodes. We develop a mathematical theory to impose eigenmode structures on connectivity, systematically characterizing their effect on network dynamics. We find the density of nearly-critical eigenvalues to be a particularly fundamental structure. It flexibly controls the power-law scaling of dynamical observables, in analogy with the system's spatial dimension in classical critical phenomena. This mechanism provides control over observables which are found to be fine-tuned in brain networks, but remained so far unexplained by traditionally studied structures, such as connectivity motifs. Specifically, the slope of the principal component spectrum of neural activity can be fine-tuned, as observed in primary visual cortex of mice. Furthermore, a novel transition between high and low dimensional activity allows for a wide and flexible tuning of dimensionality, as observed throughout cortex. The here discovered structures thus largely complement motifs. In fact, they are of a different, collective nature: they are not reflected by any local motif configuration. This result shows that many functionally relevant structures can remain hidden within the apparent randomness of highly heterogeneous cortical circuits. Our methods enable revealing these structures and investigate their effect on network dynamics."
                    },
                    {
                        "title": "Gell-Mann-Low criticality in neural networks",
                        "abstract": "Criticality is deeply related to optimal computational capacity. The lack of a renormalized theory of critical brain dynamics, however, so far limits insights into this form of biological information processing to mean-field results. These methods neglect a key feature of critical systems: the interaction between degrees of freedom across all length scales, which allows for complex nonlinear computation. We present a renormalized theory of a prototypical neural field theory, the stochastic Wilson-Cowan equation. We compute the flow of couplings, which parameterize interactions on increasing length scales. Despite similarities with the Kardar-Parisi-Zhang model, the theory is of a Gell-Mann-Low type, the archetypal form of a renormalizable quantum field theory. Here, nonlinear couplings vanish, flowing towards the Gaussian fixed point, but logarithmically slowly, thus remaining effective on most scales. We show this critical structure of interactions to implement a desirable trade-off between linearity, optimal for information storage, and nonlinearity, required for computation."
                    },
                    {
                        "title": "Synchronization Patterns in Networks of Kuramoto Oscillators: A Geometric Approach for Analysis and Control",
                        "abstract": "Synchronization is crucial for the correct functionality of many natural and man-made complex systems. In this work we characterize the formation of synchronization patterns in networks of Kuramoto oscillators. Specifically, we reveal conditions on the network weights and structure and on the oscillators' natural frequencies that allow the phases of a group of oscillators to evolve cohesively, yet independently from the phases of oscillators in different clusters. Our conditions are applicable to general directed and weighted networks of heterogeneous oscillators. Surprisingly, although the oscillators exhibit nonlinear dynamics, our approach relies entirely on tools from linear algebra and graph theory. Further, we develop a control mechanism to determine the smallest (as measured by the Frobenius norm) network perturbation to ensure the formation of a desired synchronization pattern. Our procedure allows us to constrain the set of edges that can be modified, thus enforcing the sparsity structure of the network perturbation. The results are validated through a set of numerical examples."
                    }
                ]
            },
            "2a24be20-45cc-496a-80f8-891a27f61df5": {
                "pk": "2a24be20-45cc-496a-80f8-891a27f61df5",
                "name": "Francesca Mignacco",
                "collaborators": [
                    "Lenka Zdeborov\u00e1",
                    "Pierfrancesco Urbani",
                    "Florent Krzakala",
                    "Chi-Ning Chou",
                    "SueYeon Chung",
                    "Francesco Mori",
                    "Stefano Sarao Mannelli",
                    "Yue M. Lu",
                    "Elisabetta Cornacchia",
                    "Rodrigo Veiga"
                ],
                "domain": [
                    "Machine Learning",
                    "Stochastic Optimization",
                    "Neural Networks",
                    "Statistical Physics"
                ],
                "publications": [
                    {
                        "title": "The effective noise of Stochastic Gradient Descent",
                        "abstract": "Stochastic Gradient Descent (SGD) is the workhorse algorithm of deep learning technology. At each step of the training phase, a mini batch of samples is drawn from the training dataset and the weights of the neural network are adjusted according to the performance on this specific subset of examples. The mini-batch sampling procedure introduces a stochastic dynamics to the gradient descent, with a non-trivial state-dependent noise. We characterize the stochasticity of SGD and a recently-introduced variant, \\emph{persistent} SGD, in a prototypical neural network model. In the under-parametrized regime, where the final training error is positive, the SGD dynamics reaches a stationary state and we define an effective temperature from the fluctuation-dissipation theorem, computed from dynamical mean-field theory. We use the effective temperature to quantify the magnitude of the SGD noise as a function of the problem parameters. In the over-parametrized regime, where the training error vanishes, we measure the noise magnitude of SGD by computing the average distance between two replicas of the system with the same initialization and two different realizations of SGD noise. We find that the two noise measures behave similarly as a function of the problem parameters. Moreover, we observe that noisier algorithms lead to wider decision boundaries of the corresponding constraint satisfaction problem."
                    },
                    {
                        "title": "Dynamical mean-field theory for stochastic gradient descent in Gaussian mixture classification",
                        "abstract": "We analyze in a closed form the learning dynamics of stochastic gradient descent (SGD) for a single-layer neural network classifying a high-dimensional Gaussian mixture where each cluster is assigned one of two labels. This problem provides a prototype of a non-convex loss landscape with interpolating regimes and a large generalization gap. We define a particular stochastic process for which SGD can be extended to a continuous-time limit that we call stochastic gradient flow. In the full-batch limit, we recover the standard gradient flow. We apply dynamical mean-field theory from statistical physics to track the dynamics of the algorithm in the high-dimensional limit via a self-consistent stochastic process. We explore the performance of the algorithm as a function of the control parameters shedding light on how it navigates the loss landscape."
                    },
                    {
                        "title": "Stochasticity helps to navigate rough landscapes: comparing gradient-descent-based algorithms in the phase retrieval problem",
                        "abstract": "In this paper we investigate how gradient-based algorithms such as gradient descent, (multi-pass) stochastic gradient descent, its persistent variant, and the Langevin algorithm navigate non-convex loss-landscapes and which of them is able to reach the best generalization error at limited sample complexity. We consider the loss landscape of the high-dimensional phase retrieval problem as a prototypical highly non-convex example. We observe that for phase retrieval the stochastic variants of gradient descent are able to reach perfect generalization for regions of control parameters where the gradient descent algorithm is not. We apply dynamical mean-field theory from statistical physics to characterize analytically the full trajectories of these algorithms in their continuous-time limit, with a warm start, and for large system sizes. We further unveil several intriguing properties of the landscape and the algorithms such as that the gradient descent can obtain better generalization properties from less informed initializations."
                    },
                    {
                        "title": "Nonlinear classification of neural manifolds with contextual information",
                        "abstract": "Understanding how neural systems efficiently process information through distributed representations is a fundamental challenge at the interface of neuroscience and machine learning. Recent approaches analyze the statistical and geometrical attributes of neural representations as population-level mechanistic descriptors of task implementation. In particular, manifold capacity has emerged as a promising framework linking population geometry to the separability of neural manifolds. However, this metric has been limited to linear readouts. Here, we propose a theoretical framework that overcomes this limitation by leveraging contextual input information. We derive an exact formula for the context-dependent capacity that depends on manifold geometry and context correlations, and validate it on synthetic and real data. Our framework's increased expressivity captures representation untanglement in deep networks at early stages of the layer hierarchy, previously inaccessible to analysis. As context-dependent nonlinearity is ubiquitous in neural systems, our data-driven and theoretically grounded approach promises to elucidate context-dependent computation across scales, datasets, and models."
                    },
                    {
                        "title": "Optimal Protocols for Continual Learning via Statistical Physics and Control Theory",
                        "abstract": "Artificial neural networks often struggle with catastrophic forgetting when learning multiple tasks sequentially, as training on new tasks degrades the performance on previously learned ones. Recent theoretical work has addressed this issue by analysing learning curves in synthetic frameworks under predefined training protocols. However, these protocols relied on heuristics and lacked a solid theoretical foundation assessing their optimality. In this paper, we fill this gap combining exact equations for training dynamics, derived using statistical physics techniques, with optimal control methods. We apply this approach to teacher-student models for continual learning and multi-task problems, obtaining a theory for task-selection protocols maximising performance while minimising forgetting. Our theoretical analysis offers non-trivial yet interpretable strategies for mitigating catastrophic forgetting, shedding light on how optimal learning protocols can modulate established effects, such as the influence of task similarity on forgetting. Finally, we validate our theoretical findings on real-world data."
                    },
                    {
                        "title": "The role of regularization in classification of high-dimensional noisy Gaussian mixture",
                        "abstract": "We consider a high-dimensional mixture of two Gaussians in the noisy regime where even an oracle knowing the centers of the clusters misclassifies a small but finite fraction of the points. We provide a rigorous analysis of the generalization error of regularized convex classifiers, including ridge, hinge and logistic regression, in the high-dimensional limit where the number $n$ of samples and their dimension $d$ go to infinity while their ratio is fixed to $\\alpha= n/d$. We discuss surprising effects of the regularization that in some cases allows to reach the Bayes-optimal performances. We also illustrate the interpolation peak at low regularization, and analyze the role of the respective sizes of the two clusters."
                    },
                    {
                        "title": "Learning curves for the multi-class teacher-student perceptron",
                        "abstract": "One of the most classical results in high-dimensional learning theory provides a closed-form expression for the generalisation error of binary classification with the single-layer teacher-student perceptron on i.i.d. Gaussian inputs. Both Bayes-optimal estimation and empirical risk minimisation (ERM) were extensively analysed for this setting. At the same time, a considerable part of modern machine learning practice concerns multi-class classification. Yet, an analogous analysis for the corresponding multi-class teacher-student perceptron was missing. In this manuscript we fill this gap by deriving and evaluating asymptotic expressions for both the Bayes-optimal and ERM generalisation errors in the high-dimensional regime. For Gaussian teacher weights, we investigate the performance of ERM with both cross-entropy and square losses, and explore the role of ridge regularisation in approaching Bayes-optimality. In particular, we observe that regularised cross-entropy minimisation yields close-to-optimal accuracy. Instead, for a binary teacher we show that a first-order phase transition arises in the Bayes-optimal performance."
                    },
                    {
                        "title": "High-Dimensional Non-Convex Landscapes and Gradient Descent Dynamics",
                        "abstract": "In these lecture notes we present different methods and concepts developed in statistical physics to analyze gradient descent dynamics in high-dimensional non-convex landscapes. Our aim is to show how approaches developed in physics, mainly statistical physics of disordered systems, can be used to tackle open questions on high-dimensional dynamics in Machine Learning."
                    },
                    {
                        "title": "Rigorous dynamical mean field theory for stochastic gradient descent methods",
                        "abstract": "We prove closed-form equations for the exact high-dimensional asymptotics of a family of first order gradient-based methods, learning an estimator (e.g. M-estimator, shallow neural network, ...) from observations on Gaussian data with empirical risk minimization. This includes widely used algorithms such as stochastic gradient descent (SGD) or Nesterov acceleration. The obtained equations match those resulting from the discretization of dynamical mean-field theory (DMFT) equations from statistical physics when applied to gradient flow. Our proof method allows us to give an explicit description of how memory kernels build up in the effective dynamics, and to include non-separable update functions, allowing datasets with non-identity covariance matrices. Finally, we provide numerical implementations of the equations for SGD with generic extensive batch-size and with constant learning rates."
                    },
                    {
                        "title": "Forward Learning with Top-Down Feedback: Empirical and Analytical Characterization",
                        "abstract": "\"Forward-only\" algorithms, which train neural networks while avoiding a backward pass, have recently gained attention as a way of solving the biologically unrealistic aspects of backpropagation. Here, we first address compelling challenges related to the \"forward-only\" rules, which include reducing the performance gap with backpropagation and providing an analytical understanding of their dynamics. To this end, we show that the forward-only algorithm with top-down feedback is well-approximated by an \"adaptive-feedback-alignment\" algorithm, and we analytically track its performance during learning in a prototype high-dimensional setting. Then, we compare different versions of forward-only algorithms, focusing on the Forward-Forward and PEPITA frameworks, and we show that they share the same learning principles. Overall, our work unveils the connections between three key neuro-inspired learning rules, providing a link between \"forward-only\" algorithms, i.e., Forward-Forward and PEPITA, and an approximation of backpropagation, i.e., Feedback Alignment."
                    }
                ]
            },
            "bac71621-987d-4a64-9910-7a3b7389d5ed": {
                "pk": "bac71621-987d-4a64-9910-7a3b7389d5ed",
                "name": "Kazuki Irie",
                "collaborators": [
                    "J\u00fcrgen Schmidhuber",
                    "R\u00f3bert Csord\u00e1s",
                    "Imanol Schlag",
                    "Anand Gopalakrishnan",
                    "Francesco Faccio",
                    "Brenden M. Lake",
                    "Albert Zeyer",
                    "Ralf Schl\u00fcter",
                    "Hermann Ney",
                    "Aleksandar Stani\u0107"
                ],
                "domain": [
                    "Neural Networks",
                    "Meta-Learning",
                    "Transformers",
                    "Few-Shot Learning"
                ],
                "publications": [
                    {
                        "title": "Training and Generating Neural Networks in Compressed Weight Space",
                        "abstract": "The inputs and/or outputs of some neural nets are weight matrices of other neural nets. Indirect encodings or end-to-end compression of weight matrices could help to scale such approaches. Our goal is to open a discussion on this topic, starting with recurrent neural networks for character-level language modelling whose weight matrices are encoded by the discrete cosine transform. Our fast weight version thereof uses a recurrent neural network to parameterise the compressed weights. We present experimental results on the enwik8 dataset."
                    },
                    {
                        "title": "Learning to Control Rapidly Changing Synaptic Connections: An Alternative Type of Memory in Sequence Processing Artificial Neural Networks",
                        "abstract": "Short-term memory in standard, general-purpose, sequence-processing recurrent neural networks (RNNs) is stored as activations of nodes or \"neurons.\" Generalising feedforward NNs to such RNNs is mathematically straightforward and natural, and even historical: already in 1943, McCulloch and Pitts proposed this as a surrogate to \"synaptic modifications\" (in effect, generalising the Lenz-Ising model, the first non-sequence processing RNN architecture of the 1920s). A lesser known alternative approach to storing short-term memory in \"synaptic connections\" -- by parameterising and controlling the dynamics of a context-sensitive time-varying weight matrix through another NN -- yields another \"natural\" type of short-term memory in sequence processing NNs: the Fast Weight Programmers (FWPs) of the early 1990s. FWPs have seen a recent revival as generic sequence processors, achieving competitive performance across various tasks. They are formally closely related to the now popular Transformers. Here we present them in the context of artificial NNs as an abstraction of biological NNs -- a perspective that has not been stressed enough in previous FWP work. We first review aspects of FWPs for pedagogical purposes, then discuss connections to related works motivated by insights from neuroscience."
                    },
                    {
                        "title": "Accelerating Neural Self-Improvement via Bootstrapping",
                        "abstract": "Few-shot learning with sequence-processing neural networks (NNs) has recently attracted a new wave of attention in the context of large language models. In the standard N-way K-shot learning setting, an NN is explicitly optimised to learn to classify unlabelled inputs by observing a sequence of NK labelled examples. This pressures the NN to learn a learning algorithm that achieves optimal performance, given the limited number of training examples. Here we study an auxiliary loss that encourages further acceleration of few-shot learning, by applying recently proposed bootstrapped meta-learning to NN few-shot learners: we optimise the K-shot learner to match its own performance achievable by observing more than NK examples, using only NK examples. Promising results are obtained on the standard Mini-ImageNet dataset. Our code is public."
                    },
                    {
                        "title": "Images as Weight Matrices: Sequential Image Generation Through Synaptic Learning Rules",
                        "abstract": "Work on fast weight programmers has demonstrated the effectiveness of key/value outer product-based learning rules for sequentially generating a weight matrix (WM) of a neural net (NN) by another NN or itself. However, the weight generation steps are typically not visually interpretable by humans, because the contents stored in the WM of an NN are not. Here we apply the same principle to generate natural images. The resulting fast weight painters (FPAs) learn to execute sequences of delta learning rules to sequentially generate images as sums of outer products of self-invented keys and values, one rank at a time, as if each image was a WM of an NN. We train our FPAs in the generative adversarial networks framework, and evaluate on various image datasets. We show how these generic learning rules can generate images with respectable visual quality without any explicit inductive bias for images. While the performance largely lags behind the one of specialised state-of-the-art image generators, our approach allows for visualising how synaptic learning rules iteratively produce complex connection patterns, yielding human-interpretable meaningful images. Finally, we also show that an additional convolutional U-Net (now popular in diffusion models) at the output of an FPA can learn one-step \"denoising\" of FPA-generated images to enhance their quality. Our code is public."
                    },
                    {
                        "title": "Going Beyond Linear Transformers with Recurrent Fast Weight Programmers",
                        "abstract": "Transformers with linearised attention (''linear Transformers'') have demonstrated the practical scalability and effectiveness of outer product-based Fast Weight Programmers (FWPs) from the '90s. However, the original FWP formulation is more general than the one of linear Transformers: a slow neural network (NN) continually reprograms the weights of a fast NN with arbitrary architecture. In existing linear Transformers, both NNs are feedforward and consist of a single layer. Here we explore new variations by adding recurrence to the slow and fast nets. We evaluate our novel recurrent FWPs (RFWPs) on two synthetic algorithmic tasks (code execution and sequential ListOps), Wikitext-103 language models, and on the Atari 2600 2D game environment. Our models exhibit properties of Transformers and RNNs. In the reinforcement learning setting, we report large improvements over LSTM in several Atari games. Our code is public."
                    },
                    {
                        "title": "Improving Baselines in the Wild",
                        "abstract": "We share our experience with the recently released WILDS benchmark, a collection of ten datasets dedicated to developing models and training strategies which are robust to domain shifts. Several experiments yield a couple of critical observations which we believe are of general interest for any future work on WILDS. Our study focuses on two datasets: iWildCam and FMoW. We show that (1) Conducting separate cross-validation for each evaluation metric is crucial for both datasets, (2) A weak correlation between validation and test performance might make model development difficult for iWildCam, (3) Minor changes in the training of hyper-parameters improve the baseline by a relatively large margin (mainly on FMoW), (4) There is a strong correlation between certain domains and certain target labels (mainly on iWildCam). To the best of our knowledge, no prior work on these datasets has reported these observations despite their obvious importance. Our code is public."
                    },
                    {
                        "title": "A Modern Self-Referential Weight Matrix That Learns to Modify Itself",
                        "abstract": "The weight matrix (WM) of a neural network (NN) is its program. The programs of many traditional NNs are learned through gradient descent in some error function, then remain fixed. The WM of a self-referential NN, however, can keep rapidly modifying all of itself during runtime. In principle, such NNs can meta-learn to learn, and meta-meta-learn to meta-learn to learn, and so on, in the sense of recursive self-improvement. While NN architectures potentially capable of implementing such behaviour have been proposed since the '90s, there have been few if any practical studies. Here we revisit such NNs, building upon recent successes of fast weight programmers and closely related linear Transformers. We propose a scalable self-referential WM (SRWM) that learns to use outer products and the delta update rule to modify itself. We evaluate our SRWM in supervised few-shot learning and in multi-task reinforcement learning with procedurally generated game environments. Our experiments demonstrate both practical applicability and competitive performance of the proposed SRWM. Our code is public."
                    },
                    {
                        "title": "The Dual Form of Neural Networks Revisited: Connecting Test Time Predictions to Training Patterns via Spotlights of Attention",
                        "abstract": "Linear layers in neural networks (NNs) trained by gradient descent can be expressed as a key-value memory system which stores all training datapoints and the initial weights, and produces outputs using unnormalised dot attention over the entire training experience. While this has been technically known since the 1960s, no prior work has effectively studied the operations of NNs in such a form, presumably due to prohibitive time and space complexities and impractical model sizes, all of them growing linearly with the number of training patterns which may get very large. However, this dual formulation offers a possibility of directly visualising how an NN makes use of training patterns at test time, by examining the corresponding attention weights. We conduct experiments on small scale supervised image classification tasks in single-task, multi-task, and continual learning settings, as well as language modelling, and discuss potentials and limits of this view for better understanding and interpreting how NNs exploit training patterns. Our code is public."
                    },
                    {
                        "title": "Neural Differential Equations for Learning to Program Neural Nets Through Continuous Learning Rules",
                        "abstract": "Neural ordinary differential equations (ODEs) have attracted much attention as continuous-time counterparts of deep residual neural networks (NNs), and numerous extensions for recurrent NNs have been proposed. Since the 1980s, ODEs have also been used to derive theoretical results for NN learning rules, e.g., the famous connection between Oja's rule and principal component analysis. Such rules are typically expressed as additive iterative update processes which have straightforward ODE counterparts. Here we introduce a novel combination of learning rules and Neural ODEs to build continuous-time sequence processing nets that learn to manipulate short-term memory in rapidly changing synaptic connections of other nets. This yields continuous-time counterparts of Fast Weight Programmers and linear Transformers. Our novel models outperform the best existing Neural Controlled Differential Equation based models on various time series classification tasks, while also addressing their fundamental scalability limitations. Our code is public."
                    },
                    {
                        "title": "Exploring the Promise and Limits of Real-Time Recurrent Learning",
                        "abstract": "Real-time recurrent learning (RTRL) for sequence-processing recurrent neural networks (RNNs) offers certain conceptual advantages over backpropagation through time (BPTT). RTRL requires neither caching past activations nor truncating context, and enables online learning. However, RTRL's time and space complexity make it impractical. To overcome this problem, most recent work on RTRL focuses on approximation theories, while experiments are often limited to diagnostic settings. Here we explore the practical promise of RTRL in more realistic settings. We study actor-critic methods that combine RTRL and policy gradients, and test them in several subsets of DMLab-30, ProcGen, and Atari-2600 environments. On DMLab memory tasks, our system trained on fewer than 1.2 B environmental frames is competitive with or outperforms well-known IMPALA and R2D2 baselines trained on 10 B frames. To scale to such challenging tasks, we focus on certain well-known neural architectures with element-wise recurrence, allowing for tractable RTRL without approximation. Importantly, we also discuss rarely addressed limitations of RTRL in real-world applications, such as its complexity in the multi-layer case."
                    },
                    {
                        "title": "Practical Computational Power of Linear Transformers and Their Recurrent and Self-Referential Extensions",
                        "abstract": "Recent studies of the computational power of recurrent neural networks (RNNs) reveal a hierarchy of RNN architectures, given real-time and finite-precision assumptions. Here we study auto-regressive Transformers with linearised attention, a.k.a. linear Transformers (LTs) or Fast Weight Programmers (FWPs). LTs are special in the sense that they are equivalent to RNN-like sequence processors with a fixed-size state, while they can also be expressed as the now-popular self-attention networks. We show that many well-known results for the standard Transformer directly transfer to LTs/FWPs. Our formal language recognition experiments demonstrate how recently proposed FWP extensions such as recurrent FWPs and self-referential weight matrices successfully overcome certain limitations of the LT, e.g., allowing for generalisation on the parity problem. Our code is public."
                    },
                    {
                        "title": "Automating Continual Learning",
                        "abstract": "General-purpose learning systems should improve themselves in open-ended fashion in ever-changing environments. Conventional learning algorithms for neural networks, however, suffer from catastrophic forgetting (CF) -- previously acquired skills are forgotten when a new task is learned. Instead of hand-crafting new algorithms for avoiding CF, we propose Automated Continual Learning (ACL) to train self-referential neural networks to meta-learn their own in-context continual (meta-)learning algorithms. ACL encodes all desiderata -- good performance on both old and new tasks -- into its meta-learning objectives. Our experiments demonstrate that ACL effectively solves \"in-context catastrophic forgetting\"; our ACL-learned algorithms outperform hand-crafted ones, e.g., on the Split-MNIST benchmark in the replay-free setting, and enables continual learning of diverse tasks consisting of multiple few-shot and standard image classification datasets."
                    },
                    {
                        "title": "Neural networks that overcome classic challenges through practice",
                        "abstract": "Since the earliest proposals for neural network models of the mind and brain, critics have pointed out key weaknesses in these models compared to human cognitive abilities. Here we review recent work that has used metalearning to help overcome some of these challenges. We characterize their successes as addressing an important developmental problem: they provide machines with an incentive to improve X (where X represents the desired capability) and opportunities to practice it, through explicit optimization for X; unlike conventional approaches that hope for achieving X through generalization from related but different objectives. We review applications of this principle to four classic challenges: systematicity, catastrophic forgetting, few-shot learning and multi-step reasoning; we also discuss related aspects of human development in natural environments."
                    },
                    {
                        "title": "Language Modeling with Deep Transformers",
                        "abstract": "We explore deep autoregressive Transformer models in language modeling for speech recognition. We focus on two aspects. First, we revisit Transformer model configurations specifically for language modeling. We show that well configured Transformer models outperform our baseline models based on the shallow stack of LSTM recurrent neural network layers. We carry out experiments on the open-source LibriSpeech 960hr task, for both 200K vocabulary word-level and 10K byte-pair encoding subword-level language modeling. We apply our word-level models to conventional hybrid speech recognition by lattice rescoring, and the subword-level models to attention based encoder-decoder models by shallow fusion. Second, we show that deep Transformer language models do not require positional encoding. The positional encoding is an essential augmentation for the self-attention mechanism which is invariant to sequence ordering. However, in autoregressive setup, as is the case for language modeling, the amount of information increases along the position dimension, which is a positional signal by its own. The analysis of attention weights shows that deep autoregressive self-attention models can automatically make use of such positional information. We find that removing the positional encoding even slightly improves the performance of these models."
                    },
                    {
                        "title": "Linear Transformers Are Secretly Fast Weight Programmers",
                        "abstract": "We show the formal equivalence of linearised self-attention mechanisms and fast weight controllers from the early '90s, where a ``slow\" neural net learns by gradient descent to program the ``fast weights\" of another net through sequences of elementary programming instructions which are additive outer products of self-invented activation patterns (today called keys and values). Such Fast Weight Programmers (FWPs) learn to manipulate the contents of a finite memory and dynamically interact with it. We infer a memory capacity limitation of recent linearised softmax attention variants, and replace the purely additive outer products by a delta rule-like programming instruction, such that the FWP can more easily learn to correct the current mapping from keys to values. The FWP also learns to compute dynamically changing learning rates. We also propose a new kernel function to linearise attention which balances simplicity and effectiveness. We conduct experiments on synthetic retrieval problems as well as standard machine translation and language modelling tasks which demonstrate the benefits of our methods."
                    },
                    {
                        "title": "The Neural Data Router: Adaptive Control Flow in Transformers Improves Systematic Generalization",
                        "abstract": "Despite progress across a broad range of applications, Transformers have limited success in systematic generalization. The situation is especially frustrating in the case of algorithmic tasks, where they often fail to find intuitive solutions that route relevant information to the right node/operation at the right time in the grid represented by Transformer columns. To facilitate the learning of useful control flow, we propose two modifications to the Transformer architecture, copy gate and geometric attention. Our novel Neural Data Router (NDR) achieves 100% length generalization accuracy on the classic compositional table lookup task, as well as near-perfect accuracy on the simple arithmetic task and a new variant of ListOps testing for generalization across computational depths. NDR's attention and gating patterns tend to be interpretable as an intuitive form of neural routing. Our code is public."
                    },
                    {
                        "title": "The Devil is in the Detail: Simple Tricks Improve Systematic Generalization of Transformers",
                        "abstract": "Recently, many datasets have been proposed to test the systematic generalization ability of neural networks. The companion baseline Transformers, typically trained with default hyper-parameters from standard tasks, are shown to fail dramatically. Here we demonstrate that by revisiting model configurations as basic as scaling of embeddings, early stopping, relative positional embedding, and Universal Transformer variants, we can drastically improve the performance of Transformers on systematic generalization. We report improvements on five popular datasets: SCAN, CFQ, PCFG, COGS, and Mathematics dataset. Our models improve accuracy from 50% to 85% on the PCFG productivity split, and from 35% to 81% on COGS. On SCAN, relative positional embedding largely mitigates the EOS decision problem (Newman et al., 2020), yielding 100% accuracy on the length split with a cutoff at 26. Importantly, performance differences between these models are typically invisible on the IID data split. This calls for proper generalization validation sets for developing neural networks that generalize systematically. We publicly release the code to reproduce our results."
                    },
                    {
                        "title": "CTL++: Evaluating Generalization on Never-Seen Compositional Patterns of Known Functions, and Compatibility of Neural Representations",
                        "abstract": "Well-designed diagnostic tasks have played a key role in studying the failure of neural nets (NNs) to generalize systematically. Famous examples include SCAN and Compositional Table Lookup (CTL). Here we introduce CTL++, a new diagnostic dataset based on compositions of unary symbolic functions. While the original CTL is used to test length generalization or productivity, CTL++ is designed to test systematicity of NNs, that is, their capability to generalize to unseen compositions of known functions. CTL++ splits functions into groups and tests performance on group elements composed in a way not seen during training. We show that recent CTL-solving Transformer variants fail on CTL++. The simplicity of the task design allows for fine-grained control of task difficulty, as well as many insightful analyses. For example, we measure how much overlap between groups is needed by tested NNs for learning to compose. We also visualize how learned symbol representations in outputs of functions from different groups are compatible in case of success but not in case of failure. These results provide insights into failure cases reported on more complex compositions in the natural language domain. Our code is public."
                    },
                    {
                        "title": "Contrastive Training of Complex-Valued Autoencoders for Object Discovery",
                        "abstract": "Current state-of-the-art object-centric models use slots and attention-based routing for binding. However, this class of models has several conceptual limitations: the number of slots is hardwired; all slots have equal capacity; training has high computational cost; there are no object-level relational factors within slots. Synchrony-based models in principle can address these limitations by using complex-valued activations which store binding information in their phase components. However, working examples of such synchrony-based models have been developed only very recently, and are still limited to toy grayscale datasets and simultaneous storage of less than three objects in practice. Here we introduce architectural modifications and a novel contrastive learning method that greatly improve the state-of-the-art synchrony-based model. For the first time, we obtain a class of synchrony-based models capable of discovering objects in an unsupervised manner in multi-object color datasets and simultaneously representing more than three objects."
                    },
                    {
                        "title": "Approximating Two-Layer Feedforward Networks for Efficient Transformers",
                        "abstract": "How to reduce compute and memory requirements of neural networks (NNs) without sacrificing performance? Many recent works use sparse Mixtures of Experts (MoEs) to build resource-efficient large language models (LMs). Here we introduce several novel perspectives on MoEs, presenting a general framework that unifies various methods to approximate two-layer NNs (e.g., feedforward blocks of Transformers), including product-key memories (PKMs). Leveraging insights from this framework, we propose methods to improve both MoEs and PKMs. Unlike prior work that compares MoEs with dense baselines under the compute-equal condition, our evaluation condition is parameter-equal, which is crucial to properly evaluate LMs. We show that our MoEs are competitive with the dense Transformer-XL on both the WikiText-103 and enwiki8 datasets at two different scales, while being much more resource efficient. This demonstrates that MoEs are relevant not only to extremely large LMs but also to any-scale resource-efficient LMs. Our code is public."
                    }
                ]
            },
            "03b18f76-c9bf-4e4e-aa2e-d05d2e753b48": {
                "pk": "03b18f76-c9bf-4e4e-aa2e-d05d2e753b48",
                "name": "Haim Sompolinsky",
                "collaborators": [
                    "Ran Rubin",
                    "Qianyi Li",
                    "L. F. Abbott",
                    "Kanaka Rajan",
                    "L F Abbott",
                    "Daniel D. Lee",
                    "Kukjin Kang",
                    "Naoki Hiratani",
                    "Zechen Zhang",
                    "Jonathan Kadmon"
                ],
                "domain": [
                    "Neuroscience",
                    "Machine Learning",
                    "Statistical Physics",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Mutual Information of Population Codes and Distance Measures in Probability Space",
                        "abstract": "We studied the mutual information between a stimulus and a large system consisting of stochastic, statistically independent elements that respond to a stimulus. The Mutual Information (MI) of the system saturates exponentially with system size. A theory of the rate of saturation of the MI is developed. We show that this rate is controlled by a distance function between the response probabilities induced by different stimuli. This function, which we term the {\\it Confusion Distance} between two probabilities, is related to the Renyi $\\alpha$-Information."
                    },
                    {
                        "title": "Optimal quadratic binding for relational reasoning in vector symbolic neural architectures",
                        "abstract": "Binding operation is fundamental to many cognitive processes, such as cognitive map formation, relational reasoning, and language comprehension. In these processes, two different modalities, such as location and objects, events and their contextual cues, and words and their roles, need to be bound together, but little is known about the underlying neural mechanisms. Previous works introduced a binding model based on quadratic functions of bound pairs, followed by vector summation of multiple pairs. Based on this framework, we address following questions: Which classes of quadratic matrices are optimal for decoding relational structures? And what is the resultant accuracy? We introduce a new class of binding matrices based on a matrix representation of octonion algebra, an eight-dimensional extension of complex numbers. We show that these matrices enable a more accurate unbinding than previously known methods when a small number of pairs are present. Moreover, numerical optimization of a binding operator converges to this octonion binding. We also show that when there are a large number of bound pairs, however, a random quadratic binding performs as well as the octonion and previously-proposed binding methods. This study thus provides new insight into potential neural mechanisms of binding operations in the brain."
                    },
                    {
                        "title": "Robust Learning in Bayesian Parallel Branching Graph Neural Networks: The Narrow Width Limit",
                        "abstract": "The infinite width limit of random neural networks is known to result in Neural Networks as Gaussian Process (NNGP) (Lee et al. [2018]), characterized by task-independent kernels. It is widely accepted that larger network widths contribute to improved generalization (Park et al. [2019]). However, this work challenges this notion by investigating the narrow width limit of the Bayesian Parallel Branching Graph Neural Network (BPB-GNN), an architecture that resembles residual networks. We demonstrate that when the width of a BPB-GNN is significantly smaller compared to the number of training examples, each branch exhibits more robust learning due to a symmetry breaking of branches in kernel renormalization. Surprisingly, the performance of a BPB-GNN in the narrow width limit is generally superior or comparable to that achieved in the wide width limit in bias-limited scenarios. Furthermore, the readout norms of each branch in the narrow width limit are mostly independent of the architectural hyperparameters but generally reflective of the nature of the data. Our results characterize a newly defined narrow-width regime for parallel branching networks in general."
                    },
                    {
                        "title": "Transition to chaos in random neuronal networks",
                        "abstract": "Firing patterns in the central nervous system often exhibit strong temporal irregularity and heterogeneity in their time averaged response properties. Previous studies suggested that these properties are outcome of an intrinsic chaotic dynamics. Indeed, simplified rate-based large neuronal networks with random synaptic connections are known to exhibit sharp transition from fixed point to chaotic dynamics when the synaptic gain is increased. However, the existence of a similar transition in neuronal circuit models with more realistic architectures and firing dynamics has not been established.   In this work we investigate rate based dynamics of neuronal circuits composed of several subpopulations and random connectivity. Nonzero connections are either positive-for excitatory neurons, or negative for inhibitory ones, while single neuron output is strictly positive; in line with known constraints in many biological systems. Using Dynamic Mean Field Theory, we find the phase diagram depicting the regimes of stable fixed point, unstable dynamic and chaotic rate fluctuations. We characterize the properties of systems near the chaotic transition and show that dilute excitatory-inhibitory architectures exhibit the same onset to chaos as a network with Gaussian connectivity. Interestingly, the critical properties near transition depend on the shape of the single- neuron input-output transfer function near firing threshold. Finally, we investigate network models with spiking dynamics. When synaptic time constants are slow relative to the mean inverse firing rates, the network undergoes a sharp transition from fast spiking fluctuations and static firing rates to a state with slow chaotic rate fluctuations. When the synaptic time constants are finite, the transition becomes smooth and obeys scaling properties, similar to crossover phenomena in statistical mechanics"
                    },
                    {
                        "title": "Soft-margin classification of object manifolds",
                        "abstract": "A neural population responding to multiple appearances of a single object defines a manifold in the neural response space. The ability to classify such manifolds is of interest, as object recognition and other computational tasks require a response that is insensitive to variability within a manifold. Linear classification of object manifolds was previously studied for max-margin classifiers. Soft-margin classifiers are a larger class of algorithms and provide an additional regularization parameter used in applications to optimize performance outside the training set by balancing between making fewer training errors and learning more robust classifiers. Here we develop a mean-field theory describing the behavior of soft-margin classifiers applied to object manifolds. Analyzing manifolds with increasing complexity, from points through spheres to general manifolds, a mean-field theory describes the expected value of the linear classifier's norm, as well as the distribution of fields and slack variables. By analyzing the robustness of the learned classification to noise, we can predict the probability of classification errors and their dependence on regularization, demonstrating a finite optimal choice. The theory describes a previously unknown phase transition, corresponding to the disappearance of a non-trivial solution, thus providing a soft version of the well-known classification capacity of max-margin classifiers."
                    },
                    {
                        "title": "Temporal support vectors for spiking neuronal networks",
                        "abstract": "When neural circuits learn to perform a task, it is often the case that there are many sets of synaptic connections that are consistent with the task. However, only a small number of possible solutions are robust to noise in the input and are capable of generalizing their performance of the task to new inputs. Finding such good solutions is an important goal of learning systems in general and neuronal circuits in particular. For systems operating with static inputs and outputs, a well known approach to the problem is the large margin methods such as Support Vector Machines (SVM). By maximizing the distance of the data vectors from the decision surface, these solutions enjoy increased robustness to noise and enhanced generalization abilities. Furthermore, the use of the kernel method enables SVMs to perform classification tasks that require nonlinear decision surfaces. However, for dynamical systems with event based outputs, such as spiking neural networks and other continuous time threshold crossing systems, this optimality criterion is inapplicable due to the strong temporal correlations in their input and output. We introduce a novel extension of the static SVMs - The Temporal Support Vector Machine (T-SVM). The T-SVM finds a solution that maximizes a new construct - the dynamical margin. We show that T-SVM and its kernel extensions generate robust synaptic weight vectors in spiking neurons and enable their learning of tasks that require nonlinear spatial integration of synaptic inputs. We propose T-SVM with nonlinear kernels as a new model of the computational role of the nonlinearities and extensive morphologies of neuronal dendritic trees."
                    },
                    {
                        "title": "Globally Gated Deep Linear Networks",
                        "abstract": "Recently proposed Gated Linear Networks present a tractable nonlinear network architecture, and exhibit interesting capabilities such as learning with local error signals and reduced forgetting in sequential learning. In this work, we introduce a novel gating architecture, named Globally Gated Deep Linear Networks (GGDLNs) where gating units are shared among all processing units in each layer, thereby decoupling the architectures of the nonlinear but unlearned gatings and the learned linear processing motifs. We derive exact equations for the generalization properties in these networks in the finite-width thermodynamic limit, defined by $P,N\\rightarrow\\infty, P/N\\sim O(1)$, where P and N are the training sample size and the network width respectively. We find that the statistics of the network predictor can be expressed in terms of kernels that undergo shape renormalization through a data-dependent matrix compared to the GP kernels. Our theory accurately captures the behavior of finite width GGDLNs trained with gradient descent dynamics. We show that kernel shape renormalization gives rise to rich generalization properties w.r.t. network width, depth and L2 regularization amplitude. Interestingly, networks with sufficient gating units behave similarly to standard ReLU networks. Although gatings in the model do not participate in supervised learning, we show the utility of unsupervised learning of the gating parameters. Additionally, our theory allows the evaluation of the network's ability for learning multiple tasks by incorporating task-relevant information into the gating units. In summary, our work is the first exact theoretical solution of learning in a family of nonlinear networks with finite width. The rich and diverse behavior of the GGDLNs suggests that they are helpful analytically tractable models of learning single and multiple tasks, in finite-width nonlinear deep networks."
                    },
                    {
                        "title": "Statistical Mechanics of Deep Linear Neural Networks: The Back-Propagating Kernel Renormalization",
                        "abstract": "The success of deep learning in many real-world tasks has triggered an intense effort to understand the power and limitations of deep learning in the training and generalization of complex tasks, so far with limited progress. In this work, we study the statistical mechanics of learning in Deep Linear Neural Networks (DLNNs) in which the input-output function of an individual unit is linear. Despite the linearity of the units, learning in DLNNs is nonlinear, hence studying its properties reveals some of the features of nonlinear Deep Neural Networks (DNNs). Importantly, we solve exactly the network properties following supervised learning using an equilibrium Gibbs distribution in the weight space. To do this, we introduce the Back-Propagating Kernel Renormalization (BPKR), which allows for the incremental integration of the network weights starting from the network output layer and progressing backward until the first layer's weights are integrated out. This procedure allows us to evaluate important network properties, such as its generalization error, the role of network width and depth, the impact of the size of the training set, and the effects of weight regularization and learning stochasticity. BPKR does not assume specific statistics of the input or the task's output. Furthermore, by performing partial integration of the layers, the BPKR allows us to compute the properties of the neural representations across the different hidden layers. We have proposed an extension of the BPKR to nonlinear DNNs with ReLU. Surprisingly, our numerical simulations reveal that despite the nonlinearity, the predictions of our theory are largely shared by ReLU networks in a wide regime of parameters. Our work is the first exact statistical mechanical study of learning in a family of DNNs, and the first successful theory of learning through successive integration of DoFs in the learned weight space."
                    },
                    {
                        "title": "Statistical Mechanics of Support Vector Networks",
                        "abstract": "Using methods of Statistical Physics, we investigate the generalization performance of support vector machines (SVMs), which have been recently introduced as a general alternative to neural networks. For nonlinear classification rules, the generalization error saturates on a plateau, when the number of examples is too small to properly estimate the coefficients of the nonlinear part. When trained on simple rules, we find that SVMs overfit only weakly. The performance of SVMs is strongly enhanced, when the distribution of the inputs has a gap in feature space."
                    },
                    {
                        "title": "Dynamics of Random Neural Networks with Bistable Units",
                        "abstract": "We construct and analyze a rate-based neural network model in which self-interacting units represent clusters of neurons with strong local connectivity and random inter-unit connections reflect long-range interactions. When sufficiently strong, the self-interactions make the individual units bistable. Simulation results, mean-field calculations and stability analysis reveal the different dynamic regimes of this network and identify the locations in parameter space of its phase transitions. We identify an interesting dynamical regime exhibiting transient but long-lived chaotic activity that combines features of chaotic and multiple fixed-point attractors."
                    },
                    {
                        "title": "Theory of spike timing based neural classifiers",
                        "abstract": "We study the computational capacity of a model neuron, the Tempotron, which classifies sequences of spikes by linear-threshold operations. We use statistical mechanics and extreme value theory to derive the capacity of the system in random classification tasks. In contrast to its static analog, the Perceptron, the Tempotron's solutions space consists of a large number of small clusters of weight vectors. The capacity of the system per synapse is finite in the large size limit and weakly diverges with the stimulus duration relative to the membrane and synaptic time constants."
                    },
                    {
                        "title": "Stimulus-dependent correlations in threshold-crossing spiking neurons",
                        "abstract": "We consider a threshold-crossing spiking process as a simple model for the activity within a population of neurons. Assuming that these neurons are driven by a common fluctuating input with Gaussian statistics, we evaluate the cross-correlation of spike trains in pairs of model neurons with different thresholds. This correlation function tends to be asymmetric in time, indicating a preference for the neuron with the lower threshold to fire before the one with the higher threshold, even if their inputs are identical. The relationship between these results and spike statistics in other models of neural activity are explored. In particular, we compare our model with an integrate-and-fire model in which the membrane voltage resets following each spike. The qualitative properties of spike cross correlations, emerging from the threshold-crossing model, are similar to those of bursting events in the integrate-and-fire model. This is particularly true for generalized integrate-and-fire models in which spikes tend to occur in bursts as observed, for example, in retinal ganglion cells driven by a rapidly fluctuating visual stimulus. The threshold crossing model thus provides a simple, analytically tractable description of event onsets in these neurons."
                    },
                    {
                        "title": "Stimulus-Dependent Suppression of Chaos in Recurrent Neural Networks",
                        "abstract": "Neuronal activity arises from an interaction between ongoing firing generated spontaneously by neural circuits and responses driven by external stimuli. Using mean-field analysis, we ask how a neural network that intrinsically generates chaotic patterns of activity can remain sensitive to extrinsic input. We find that inputs not only drive network responses, they also actively suppress ongoing activity, ultimately leading to a phase transition in which chaos is completely eliminated. The critical input intensity at the phase transition is a non-monotonic function of stimulus frequency, revealing a \"resonant\" frequency at which the input is most effective at suppressing chaos even though the power spectrum of the spontaneous activity peaks at zero and falls exponentially. A prediction of our analysis is that the variance of neural responses should be most strongly suppressed at frequencies matching the range over which many sensory systems operate."
                    },
                    {
                        "title": "Linear Readout of Object Manifolds",
                        "abstract": "Objects are represented in sensory systems by continuous manifolds due to sensitivity of neuronal responses to changes in physical features such as location, orientation, and intensity. What makes certain sensory representations better suited for invariant decoding of objects by downstream networks? We present a theory that characterizes the ability of a linear readout network, the perceptron, to classify objects from variable neural responses. We show how the readout perceptron capacity depends on the dimensionality, size, and shape of the object manifolds in its input neural representation."
                    },
                    {
                        "title": "Coding schemes in neural networks learning classification tasks",
                        "abstract": "Neural networks posses the crucial ability to generate meaningful representations of task-dependent features. Indeed, with appropriate scaling, supervised learning in neural networks can result in strong, task-dependent feature learning. However, the nature of the emergent representations, which we call the `coding scheme', is still unclear. To understand the emergent coding scheme, we investigate fully-connected, wide neural networks learning classification tasks using the Bayesian framework where learning shapes the posterior distribution of the network weights. Consistent with previous findings, our analysis of the feature learning regime (also known as `non-lazy', `rich', or `mean-field' regime) shows that the networks acquire strong, data-dependent features. Surprisingly, the nature of the internal representations depends crucially on the neuronal nonlinearity. In linear networks, an analog coding scheme of the task emerges. Despite the strong representations, the mean predictor is identical to the lazy case. In nonlinear networks, spontaneous symmetry breaking leads to either redundant or sparse coding schemes. Our findings highlight how network properties such as scaling of weights and neuronal nonlinearity can profoundly influence the emergent representations."
                    },
                    {
                        "title": "Short-Term Memory in Orthogonal Neural Networks",
                        "abstract": "We study the ability of linear recurrent networks obeying discrete time dynamics to store long temporal sequences that are retrievable from the instantaneous state of the network. We calculate this temporal memory capacity for both distributed shift register and random orthogonal connectivity matrices. We show that the memory capacity of these networks scales with system size."
                    },
                    {
                        "title": "Interactions between Intrinsic and Stimulus-Evoked Activity in Recurrent Neural Networks",
                        "abstract": "Trial-to-trial variability is an essential feature of neural responses, but its source is a subject of active debate. Response variability (Mast and Victor, 1991; Arieli et al., 1995 & 1996; Anderson et al., 2000 & 2001; Kenet et al., 2003; Petersen et al., 2003a & b; Fiser, Chiu and Weliky, 2004; MacLean et al., 2005; Yuste et al., 2005; Vincent et al., 2007) is often treated as random noise, generated either by other brain areas, or by stochastic processes within the circuitry being studied. We call such sources of variability external to stress the independence of this form of noise from activity driven by the stimulus. Variability can also be generated internally by the same network dynamics that generates responses to a stimulus. How can we distinguish between external and internal sources of response variability? Here we show that internal sources of variability interact nonlinearly with stimulus-induced activity, and this interaction yields a suppression of noise in the evoked state. This provides a theoretical basis and potential mechanism for the experimental observation that, in many brain areas, stimuli cause significant suppression of neuronal variability (Werner and Mountcastle, 1963; Fortier, Smith and Kalaska, 1993; Anderson et al., 2000; Friedrich and Laurent, 2004; Churchland et al., 2006; Finn, Priebe and Ferster, 2007; Mitchell, Sundberg and Reynolds, 2007; Churchland et al., 2009). The combined theoretical and experimental results suggest that internally generated activity is a significant contributor to response variability in neural circuits."
                    },
                    {
                        "title": "Balanced Excitation and Inhibition are Required for High-Capacity, Noise-Robust Neuronal Selectivity",
                        "abstract": "Neurons and networks in the cerebral cortex must operate reliably despite multiple sources of noise. To evaluate the impact of both input and output noise, we determine the robustness of single-neuron stimulus selective responses, as well as the robustness of attractor states of networks of neurons performing memory tasks. We find that robustness to output noise requires synaptic connections to be in a balanced regime in which excitation and inhibition are strong and largely cancel each other. We evaluate the conditions required for this regime to exist and determine the properties of networks operating within it. A plausible synaptic plasticity rule for learning that balances weight configurations is presented. Our theory predicts an optimal ratio of the number of excitatory and inhibitory synapses for maximizing the encoding capacity of balanced networks for a given statistics of afferent activations. Previous work has shown that balanced networks amplify spatio-temporal variability and account for observed asynchronous irregular states. Here we present a novel type of balanced network that amplifies small changes in the impinging signals, and emerges automatically from learning to perform neuronal and network functions robustly."
                    },
                    {
                        "title": "Macroscopic Fluctuations Emerge in Balanced Networks with Incomplete Recurrent Alignment",
                        "abstract": "Networks of strongly-coupled neurons with random connectivity exhibit chaotic, asynchronous fluctuations. In previous work, we showed that when endowed with an additional low-rank connectivity consisting of the outer product of orthogonal vectors, these networks generate large-scale coherent fluctuations. Although a striking phenomenon, that result depended on a fine-tuned choice of low-rank structure. Here we extend that work by generalizing the theory of excitation-inhibition balance to networks with arbitrary low-rank structure and show that low-dimensional variability emerges intrinsically through what we call incomplete recurrent alignment. We say that a low-rank connectivity structure exhibits incomplete alignment if its row-space is not contained in its column-space. In the setting of incomplete alignment, recurrent connectivity can be decomposed into a subspace-recurrent component and an effective-feedforward component. We show that high-dimensional, microscopic fluctuations are propagated via the effective-feedforward component to a low-dimensional subspace where they are dynamically balanced by macroscopic fluctuations. We present biologically plausible examples from excitation-inhibition networks and networks with heterogeneous degree distributions. Finally, we define the alignment matrix as the overlap between left and right-singular vectors of the structured connectivity, and show that the singular values of the alignment matrix determine the amplitude of macroscopic variability, while its singular vectors determine the structure. Our work shows how macroscopic fluctuations can emerge generically in strongly-coupled networks with low-rank structure. Furthermore, by generalizing excitation-inhibition balance to arbitrary low-rank structure our work may find relevance in any setting with strongly interacting units, whether in biological, social, or technological networks."
                    }
                ]
            }
        }
    },
    "2407.13977": {
        "paper_data": {
            "title": "A Unified Confidence Sequence for Generalized Linear Models, with Applications to Bandits",
            "url": "http://arxiv.org/abs/2407.13977v1",
            "arxiv_id": "2407.13977",
            "authors": [
                "Junghyun Lee",
                "Se-Young Yun",
                "Kwang-Sung Jun"
            ],
            "abstract": "We present a unified likelihood ratio-based confidence sequence (CS) for any (self-concordant) generalized linear models (GLMs) that is guaranteed to be convex and numerically tight. We show that this is on par or improves upon known CSs for various GLMs, including Gaussian, Bernoulli, and Poisson. In particular, for the first time, our CS for Bernoulli has a poly(S)-free radius where S is the norm of the unknown parameter. Our first technical novelty is its derivation, which utilizes a time-uniform PAC-Bayesian bound with a uniform prior/posterior, despite the latter being a rather unpopular choice for deriving CSs. As a direct application of our new CS, we propose a simple and natural optimistic algorithm called OFUGLB applicable to any generalized linear bandits (GLB; Filippi et al. (2010)). Our analysis shows that the celebrated optimistic approach simultaneously attains state-of-the-art regrets for various self-concordant (not necessarily bounded) GLBs, and even poly(S)-free for bounded GLBs, including logistic bandits. The regret analysis, our second technical novelty, follows from combining our new CS with a new proof technique that completely avoids the previously widely used self-concordant control lemma (Faury et al., 2020, Lemma 9). Finally, we verify numerically that OFUGLB significantly outperforms the prior state-of-the-art (Lee et al., 2024) for logistic bandits.",
            "introduction": "   1 Introduction  One paramount task in statistics and machine learning is to estimate the uncertainty of the underlying model from (possibly noisy) observations. For example, in interactive machine learning scenarios such as bandits\u00a0(Thompson, 1933; Robbins, 1952; Lattimore and Szepesv\u00e1ri, 2020) and recently reinforcement learning with human feedback (RLHF; Ouyang et\u00a0al. (2022); Christiano et\u00a0al. (2017)), at each time step t\ud835\udc61titalic_t, the learner chooses an action \ud835\udc99tsubscript\ud835\udc99\ud835\udc61{\\bm{x}}_{t}bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT from an available set of actions \ud835\udcb3tsubscript\ud835\udcb3\ud835\udc61\\mathcal{X}_{t}caligraphic_X start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT and observes reward or outcome rtsubscript\ud835\udc5f\ud835\udc61r_{t}italic_r start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT that is modeled as a distribution whose mean is an unknown function f\u2217superscript\ud835\udc53f^{*}italic_f start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT of \ud835\udc99tsubscript\ud835\udc99\ud835\udc61{\\bm{x}}_{t}bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT; i.e., rt\u223cp(\u22c5|\ud835\udc99t;f\u2217)r_{t}\\sim p(\\cdot|{\\bm{x}}_{t};f^{*})italic_r start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT \u223c italic_p ( \u22c5 | bold_italic_x start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ; italic_f start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT ). One popular choice of such a model is generalized linear model (GLM; McCullagh and Nelder (1989)) that extends exponential family distributions to have linear structure in its natural parameter, i.e., \u27e8\ud835\udc99,\ud835\udf3d\u22c6\u27e9\ud835\udc99subscript\ud835\udf3d\u22c6\\langle{\\bm{x}},\\bm{\\theta}_{\\star}\\rangle\u27e8 bold_italic_x , bold_italic_\u03b8 start_POSTSUBSCRIPT \u22c6 end_POSTSUBSCRIPT \u27e9 where \ud835\udf3d\u22c6subscript\ud835\udf3d\u22c6\\bm{\\theta}_{\\star}bold_italic_\u03b8 start_POSTSUBSCRIPT \u22c6 end_POSTSUBSCRIPT is an unknown parameter, which means that the mean function is f\u2217\u2062(\ud835\udc99)=\u03bc\u2062(\u27e8\ud835\udc99,\ud835\udf3d\u22c6\u27e9)superscript\ud835\udc53\ud835\udc99\ud835\udf07\ud835\udc99subscript\ud835\udf3d\u22c6f^{*}({\\bm{x}})=\\mu(\\langle{\\bm{x}},\\bm{\\theta}_{\\star}\\rangle)italic_f start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT ( bold_italic_x ) = italic_\u03bc ( \u27e8 bold_italic_x , bold_italic_\u03b8 start_POSTSUBSCRIPT \u22c6 end_POSTSUBSCRIPT \u27e9 ) for some inverse link function \u03bc\ud835\udf07\\muitalic_\u03bc. This encompasses a wide range of distributions, which in turn makes it ubiquitous in various real-world applications, such as news recommendations (Bernoulli; Li et\u00a0al. (2010, 2012)), social network influence maximization (Poisson; Lage et\u00a0al. (2013); Gisselbrecht et\u00a0al. (2015)), and more. In such tasks, the learner must estimate the uncertainty about \ud835\udf3d\u22c6subscript\ud835\udf3d\u22c6\\bm{\\theta}_{\\star}bold_italic_\u03b8 start_POSTSUBSCRIPT \u22c6 end_POSTSUBSCRIPT at each time step t\u22651\ud835\udc611t\\geq 1italic_t \u2265 1, given observations {(\ud835\udc99s,rs)}s=1t\u22121superscriptsubscriptsubscript\ud835\udc99\ud835\udc60subscript\ud835\udc5f\ud835\udc60\ud835\udc601\ud835\udc611\\{({\\bm{x}}_{s},r_{s})\\}_{s=1}^{t-1}{ ( bold_italic_x start_POSTSUBSCRIPT italic_s end_POSTSUBSCRIPT , italic_r start_POSTSUBSCRIPT italic_s end_POSTSUBSCRIPT ) } start_POSTSUBSCRIPT italic_s = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_t - 1 end_POSTSUPERSCRIPT, to make wise decisions. One popular and useful way to capture the uncertainty is via a time-uniform confidence sequence (CS) {\ud835\udc9et\u2062(\u03b4)}t=1\u221esuperscriptsubscriptsubscript\ud835\udc9e\ud835\udc61\ud835\udeff\ud835\udc611\\{\\mathcal{C}_{t}(\\delta)\\}_{t=1}^{\\infty}{ caligraphic_C start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ( italic_\u03b4 ) } start_POSTSUBSCRIPT italic_t = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT \u221e end_POSTSUPERSCRIPT, which takes the form of \u2119[\u2203t\u22651:\ud835\udf3d\u22c6\u2209\ud835\udc9et(\u03b4)]\u2264\u03b4{\\mathbb{P}}[\\exists t\\geq 1:\\bm{\\theta}_{\\star}\\not\\in{\\mathcal{C}}_{t}(% \\delta)]\\leq\\deltablackboard_P [ \u2203 italic_t \u2265 1 : bold_italic_\u03b8 start_POSTSUBSCRIPT \u22c6 end_POSTSUBSCRIPT \u2209 caligraphic_C start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT ( italic_\u03b4 ) ] \u2264 italic_\u03b4. Recently, CS has been described as one of the key components for safe anytime-valid inference (SAVI) that can ensure the validity/safeness of sequentially adaptive statistical inference\u00a0(Ramdas et\u00a0al., 2023).   There has been much work on deriving CS for specific families of distributions. Many common distributions are in a smaller family, called generalized linear models (GLMs). Existing CSs for GLM, however, are far from ideal. Much of the prior works focus on obtaining CS for specific instantiations of GLMs, such as Gaussian\u00a0(Abbasi-Yadkori et\u00a0al., 2011; Flynn et\u00a0al., 2023) and Bernoulli\u00a0(Faury et\u00a0al., 2020; Abeille et\u00a0al., 2021; Faury et\u00a0al., 2022; Lee et\u00a0al., 2024). Especially for Bernoulli, all the existing CSs suffer from poly\u2061(S)poly\ud835\udc46\\operatorname{poly}(S)roman_poly ( italic_S ) factor in the radius, where S\ud835\udc46Sitalic_S is the norm of the unknown parameter \ud835\udf3d\u22c6subscript\ud835\udf3d\u22c6\\bm{\\theta}_{\\star}bold_italic_\u03b8 start_POSTSUBSCRIPT \u22c6 end_POSTSUBSCRIPT. Jun et\u00a0al. (2017); Li et\u00a0al. (2017); Emmenegger et\u00a0al. (2023) proposed generic CSs that work for any convex GLMs, but their radii all suffer from a globally worst-case curvature of \u03bc\ud835\udf07\\muitalic_\u03bc, which is detrimental in many cases",
            "references": [
                {
                    "title": "Efficient Prompt Optimization Through the Lens of Best Arm Identification",
                    "abstract": "The remarkable instruction-following capability of large language models (LLMs) has sparked a growing interest in automatically finding good prompts, i.e., prompt optimization. Most existing works follow the scheme of selecting from a pre-generated pool of candidate prompts. However, these designs mainly focus on the generation strategy, while limited attention has been paid to the selection method. Especially, the cost incurred during the selection (e.g., accessing LLM and evaluating the responses) is rarely explicitly considered. To overcome this limitation, this work provides a principled framework, TRIPLE, to efficiently perform prompt selection under an explicit budget constraint. TRIPLE is built on a novel connection established between prompt optimization and fixed-budget best arm identification (BAI-FB) in multi-armed bandits (MAB); thus, it is capable of leveraging the rich toolbox from BAI-FB systematically and also incorporating unique characteristics of prompt optimization. Extensive experiments on multiple well-adopted tasks using various LLMs demonstrate the remarkable performance improvement of TRIPLE over baselines while satisfying the limited budget constraints. As an extension, variants of TRIPLE are proposed to efficiently select examples for few-shot prompts, also achieving superior empirical performance."
                },
                {
                    "title": "An information-theoretic lower bound in time-uniform estimation",
                    "abstract": "We present an information-theoretic lower bound for the problem of parameter estimation with time-uniform coverage guarantees. Via a new a reduction to sequential testing, we obtain stronger lower bounds that capture the hardness of the time-uniform setting. In the case of location model estimation, logistic regression, and exponential family models, our $\\Omega(\\sqrt{n^{-1}\\log \\log n})$ lower bound is sharp to within constant factors in typical settings."
                },
                {
                    "title": "Exploration via linearly perturbed loss minimisation",
                    "abstract": "We introduce exploration via linear loss perturbations (EVILL), a randomised exploration method for structured stochastic bandit problems that works by solving for the minimiser of a linearly perturbed regularised negative log-likelihood function. We show that, for the case of generalised linear bandits, EVILL reduces to perturbed history exploration (PHE), a method where exploration is done by training on randomly perturbed rewards. In doing so, we provide a simple and clean explanation of when and why random reward perturbations give rise to good bandit algorithms. We propose data-dependent perturbations not present in previous PHE-type methods that allow EVILL to match the performance of Thompson-sampling-style parameter-perturbation methods, both in theory and in practice. Moreover, we show an example outside generalised linear bandits where PHE leads to inconsistent estimates, and thus linear regret, while EVILL remains performant. Like PHE, EVILL can be implemented in just a few lines of code."
                },
                {
                    "title": "Likelihood Ratio Confidence Sets for Sequential Decision Making",
                    "abstract": "Certifiable, adaptive uncertainty estimates for unknown quantities are an essential ingredient of sequential decision-making algorithms. Standard approaches rely on problem-dependent concentration results and are limited to a specific combination of parameterization, noise family, and estimator. In this paper, we revisit the likelihood-based inference principle and propose to use likelihood ratios to construct any-time valid confidence sequences without requiring specialized treatment in each application scenario. Our method is especially suitable for problems with well-specified likelihoods, and the resulting sets always maintain the prescribed coverage in a model-agnostic manner. The size of the sets depends on a choice of estimator sequence in the likelihood ratio. We discuss how to provably choose the best sequence of estimators and shed light on connections to online convex optimization with algorithms such as Follow-the-Regularized-Leader. To counteract the initially large bias of the estimators, we propose a reweighting scheme that also opens up deployment in non-parametric settings such as RKHS function classes. We provide a non-asymptotic analysis of the likelihood ratio confidence sets size for generalized linear models, using insights from convex duality and online learning. We showcase the practical strength of our method on generalized linear bandit problems, survival analysis, and bandits with various additive noise distributions."
                },
                {
                    "title": "Improved Regret Bounds of (Multinomial) Logistic Bandits via Regret-to-Confidence-Set Conversion",
                    "abstract": "Logistic bandit is a ubiquitous framework of modeling users' choices, e.g., click vs. no click for advertisement recommender system. We observe that the prior works overlook or neglect dependencies in $S \\geq \\lVert \\theta_\\star \\rVert_2$, where $\\theta_\\star \\in \\mathbb{R}^d$ is the unknown parameter vector, which is particularly problematic when $S$ is large, e.g., $S \\geq d$. In this work, we improve the dependency on $S$ via a novel approach called {\\it regret-to-confidence set conversion (R2CS)}, which allows us to construct a convex confidence set based on only the \\textit{existence} of an online learning algorithm with a regret guarantee. Using R2CS, we obtain a strict improvement in the regret bound w.r.t. $S$ in logistic bandits while retaining computational feasibility and the dependence on other factors such as $d$ and $T$. We apply our new confidence set to the regret analyses of logistic bandits with a new martingale concentration step that circumvents an additional factor of $S$. We then extend this analysis to multinomial logistic bandits and obtain similar improvements in the regret, showing the efficacy of R2CS. While we applied R2CS to the (multinomial) logistic model, R2CS is a generic approach for developing confidence sets that can be used for various models, which can be of independent interest."
                },
                {
                    "title": "Improved Algorithms for Stochastic Linear Bandits Using Tail Bounds for Martingale Mixtures",
                    "abstract": "We present improved algorithms with worst-case regret guarantees for the stochastic linear bandit problem. The widely used\"optimism in the face of uncertainty\"principle reduces a stochastic bandit problem to the construction of a confidence sequence for the unknown reward function. The performance of the resulting bandit algorithm depends on the size of the confidence sequence, with smaller confidence sets yielding better empirical performance and stronger regret guarantees. In this work, we use a novel tail bound for adaptive martingale mixtures to construct confidence sequences which are suitable for stochastic bandits. These confidence sequences allow for efficient action selection via convex programming. We prove that a linear bandit algorithm based on our confidence sequences is guaranteed to achieve competitive worst-case regret. We show that our confidence sequences are tighter than competitors, both empirically and theoretically. Finally, we demonstrate that our tighter confidence sequences give improved performance in several hyperparameter tuning tasks."
                },
                {
                    "title": "A unified recipe for deriving (time-uniform) PAC-Bayes bounds",
                    "abstract": "We present a unified framework for deriving PAC-Bayesian generalization bounds. Unlike most previous literature on this topic, our bounds are anytime-valid (i.e., time-uniform), meaning that they hold at all stopping times, not only for a fixed sample size. Our approach combines four tools in the following order: (a) nonnegative supermartingales or reverse submartingales, (b) the method of mixtures, (c) the Donsker-Varadhan formula (or other convex duality principles), and (d) Ville's inequality. Our main result is a PAC-Bayes theorem which holds for a wide class of discrete stochastic processes. We show how this result implies time-uniform versions of well-known classical PAC-Bayes bounds, such as those of Seeger, McAllester, Maurer, and Catoni, in addition to many recent bounds. We also present several novel bounds. Our framework also enables us to relax traditional assumptions; in particular, we consider nonstationary loss functions and non-i.i.d. data. In sum, we unify the derivation of past bounds and ease the search for future bounds: one may simply check if our supermartingale or submartingale conditions are met and, if so, be guaranteed a (time-uniform) PAC-Bayes bound."
                },
                {
                    "title": "Game-theoretic statistics and safe anytime-valid inference",
                    "abstract": "Safe anytime-valid inference (SAVI) provides measures of statistical evidence and certainty -- e-processes for testing and confidence sequences for estimation -- that remain valid at all stopping times, accommodating continuous monitoring and analysis of accumulating data and optional stopping or continuation for any reason. These measures crucially rely on test martingales, which are nonnegative martingales starting at one. Since a test martingale is the wealth process of a player in a betting game, SAVI centrally employs game-theoretic intuition, language and mathematics. We summarize the SAVI goals and philosophy, and report recent advances in testing composite hypotheses and estimating functionals in nonparametric settings."
                },
                {
                    "title": "Double Doubly Robust Thompson Sampling for Generalized Linear Contextual Bandits",
                    "abstract": "We propose a novel algorithm for generalized linear contextual bandits (GLBs) with a regret bound sublinear to the time horizon, the minimum eigenvalue of the covariance of contexts and a lower bound of the variance of rewards.\nIn several identified cases, our result is the first regret bound for generalized linear bandits (GLBs) achieving the regret bound sublinear to the dimension of contexts without discarding the observed rewards.\nPrevious approaches achieve the regret bound sublinear to the dimension of contexts by discarding the observed rewards, whereas our algorithm achieves the bound incorporating contexts from all arms in our double doubly robust (DDR) estimator.\nThe DDR estimator is a subclass of doubly robust estimator but with a tighter error bound.\nWe also provide a logarithmic cumulative regret bound under a probabilistic margin condition.\nThis is the first regret bound under the margin condition for linear models or GLMs when contexts are different for all arms but coefficients are common.\nWe conduct empirical studies using synthetic data and real examples, demonstrating the effectiveness of our algorithm."
                },
                {
                    "title": "Norm-Agnostic Linear Bandits",
                    "abstract": "Linear bandits have a wide variety of applications including recommendation systems yet they make one strong assumption: the algorithms must know an upper bound $S$ on the norm of the unknown parameter $\\theta^*$ that governs the reward generation. Such an assumption forces the practitioner to guess $S$ involved in the confidence bound, leaving no choice but to wish that $\\|\\theta^*\\|\\le S$ is true to guarantee that the regret will be low. In this paper, we propose novel algorithms that do not require such knowledge for the first time. Specifically, we propose two algorithms and analyze their regret bounds: one for the changing arm set setting and the other for the fixed arm set setting. Our regret bound for the former shows that the price of not knowing $S$ does not affect the leading term in the regret bound and inflates only the lower order term. For the latter, we do not pay any price in the regret for now knowing $S$. Our numerical experiments show standard algorithms assuming knowledge of $S$ can fail catastrophically when $\\|\\theta^*\\|\\le S$ is not true whereas our algorithms enjoy low regret."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "An Experimental Design Approach for Regret Minimization in Logistic Bandits",
                    "abstract": "In this work we consider the problem of regret minimization for logistic bandits. The main challenge of logistic bandits is reducing the dependence on a potentially large problem dependent constant that can at worst scale exponentially with the norm of the unknown parameter vector. Previous works have applied self-concordance of the logistic function to remove this worst-case dependence providing regret guarantees that move the reduce the dependence on this worst case parameter to lower order terms with only polylogarithmic dependence on the main term and as well as linear dependence on the dimension of the unknown parameter. This work improves upon the prior art by 1) removing all scaling of the worst case term on the main term and 2) reducing the dependence on the dependence to scale with the square root of dimension in the fixed arm setting by employing an experimental design procedure. Our regret bound in fact takes a tighter instance (i.e., gap) dependent regret bound for the first time in logistic bandits. We also propose a new warmup sampling algorithm that can dramatically reduce the lower order term in the regret in general and prove that it can exponentially reduce the lower order term's dependency on the worst case parameter in some instances. Finally, we discuss the impact of the bias of the MLE on the logistic bandit problem in d dimensions, providing an example where d^2 lower order regret (cf., it is d for linear bandits) may not be improved as long as the MLE is used and how bias-corrected estimators may be used to make it closer to d."
                },
                {
                    "title": "Bregman Deviations of Generic Exponential Families",
                    "abstract": "We revisit the method of mixture technique, also known as the Laplace method, to study the concentration phenomenon in generic exponential families. Combining the properties of Bregman divergence associated with log-partition function of the family with the method of mixtures for super-martingales, we establish a generic bound controlling the Bregman divergence between the parameter of the family and a finite sample estimate of the parameter. Our bound is time-uniform and makes appear a quantity extending the classical information gain to exponential families, which we call the Bregman information gain. For the practitioner, we instantiate this novel bound to several classical families, e.g., Gaussian, Bernoulli, Exponential, Weibull, Pareto, Poisson and Chi-square yielding explicit forms of the confidence sets and the Bregman information gain. We further numerically compare the resulting confidence bounds to state-of-the-art alternatives for time-uniform concentration and show that this novel method yields competitive results. Finally, we highlight the benefit of our concentration bounds on some illustrative applications."
                },
                {
                    "title": "Jointly Efficient and Optimal Algorithms for Logistic Bandits",
                    "abstract": "Logistic Bandits have recently undergone careful scrutiny by virtue of their combined theoretical and practical relevance. This research effort delivered statistically efficient algorithms, improving the regret of previous strategies by exponentially large factors. Such algorithms are however strikingly costly as they require $\\Omega(t)$ operations at each round. On the other hand, a different line of research focused on computational efficiency ($\\mathcal{O}(1)$ per-round cost), but at the cost of letting go of the aforementioned exponential improvements. Obtaining the best of both world is unfortunately not a matter of marrying both approaches. Instead we introduce a new learning procedure for Logistic Bandits. It yields confidence sets which sufficient statistics can be easily maintained online without sacrificing statistical tightness. Combined with efficient planning mechanisms we design fast algorithms which regret performance still match the problem-dependent lower-bound of Abeille et al. (2021). To the best of our knowledge, those are the first Logistic Bandit algorithms that simultaneously enjoy statistical and computational efficiency."
                },
                {
                    "title": "Improved Regret Analysis for Variance-Adaptive Linear Bandits and Horizon-Free Linear Mixture MDPs",
                    "abstract": "In online learning problems, exploiting low variance plays an important role in obtaining tight performance guarantees yet is challenging because variances are often not known a priori. Recently, considerable progress has been made by Zhang et al. (2021) where they obtain a variance-adaptive regret bound for linear bandits without knowledge of the variances and a horizon-free regret bound for linear mixture Markov decision processes (MDPs). In this paper, we present novel analyses that improve their regret bounds significantly. For linear bandits, we achieve $\\tilde O(\\min\\{d\\sqrt{K}, d^{1.5}\\sqrt{\\sum_{k=1}^K \\sigma_k^2}\\} + d^2)$ where $d$ is the dimension of the features, $K$ is the time horizon, and $\\sigma_k^2$ is the noise variance at time step $k$, and $\\tilde O$ ignores polylogarithmic dependence, which is a factor of $d^3$ improvement. For linear mixture MDPs with the assumption of maximum cumulative reward in an episode being in $[0,1]$, we achieve a horizon-free regret bound of $\\tilde O(d \\sqrt{K} + d^2)$ where $d$ is the number of base models and $K$ is the number of episodes. This is a factor of $d^{3.5}$ improvement in the leading term and $d^7$ in the lower order term. Our analysis critically relies on a novel peeling-based regret analysis that leverages the elliptical potential `count' lemma."
                },
                {
                    "title": "User-friendly introduction to PAC-Bayes bounds",
                    "abstract": "Aggregated predictors are obtained by making a set of basic predictors vote according to some weights, that is, to some probability distribution. Randomized predictors are obtained by sampling in a set of basic predictors, according to some prescribed probability distribution. Thus, aggregated and randomized predictors have in common that they are not defined by a minimization problem, but by a probability distribution on the set of predictors. In statistical learning theory, there is a set of tools designed to understand the generalization ability of such procedures: PAC-Bayesian or PAC-Bayes bounds. Since the original PAC-Bayes bounds of D. McAllester, these tools have been considerably improved in many directions (we will for example describe a simplified version of the localization technique of O. Catoni that was missed by the community, and later rediscovered as\"mutual information bounds\"). Very recently, PAC-Bayes bounds received a considerable attention: for example there was workshop on PAC-Bayes at NIPS 2017,\"(Almost) 50 Shades of Bayesian Learning: PAC-Bayesian trends and insights\", organized by B. Guedj, F. Bach and P. Germain. One of the reason of this recent success is the successful application of these bounds to neural networks by G. Dziugaite and D. Roy. An elementary introduction to PAC-Bayes theory is still missing. This is an attempt to provide such an introduction."
                },
                {
                    "title": "No-regret Algorithms for Capturing Events in Poisson Point Processes",
                    "abstract": "Inhomogeneous Poisson point processes are widely used models of event occurrences. We address adaptive sensing of Poisson Point processes , namely, maximizing the number of captured events subject to sensing costs. We encode prior assumptions on the rate function by modeling it as a member of a known reproducing kernel Hilbert space (RKHS). By partitioning the domain into separate small regions, and using heteroscedastic linear regression, we propose a tractable estimator of Poisson process rates for two feedback models: count-record , where exact locations of events are observed, and histogram feedback, where only counts of events are observed. We derive provably accurate anytime con-\ufb01dence estimates for our estimators for sequentially acquired Poisson count data. Using these, we formulate algorithms based on optimism that provably incur sublinear count-regret. We demonstrate the practicality of the method on problems from crime modeling, revenue maximization as well as environmental monitoring."
                },
                {
                    "title": "Fixed-Budget Best-Arm Identification in Structured Bandits",
                    "abstract": "Best-arm identification (BAI) in a fixed-budget setting is a bandit problem where the learning agent maximizes the probability of identifying the optimal (best) arm after a fixed number of observations. Most works on this topic study unstructured problems with a small number of arms, which limits their applicability. We propose a general tractable algorithm that incorporates the structure, by successively eliminating suboptimal arms based on their mean reward estimates from a joint generalization model. We analyze our algorithm in linear and generalized linear models (GLMs), and propose a practical implementation based on a G-optimal design. In linear models, our algorithm has competitive error guarantees to prior works and performs at least as well empirically. In GLMs, this is the first practical algorithm with analysis for fixed-budget BAI."
                },
                {
                    "title": "Improved Confidence Bounds for the Linear Logistic Model and Applications to Bandits",
                    "abstract": "We propose improved fixed-design confidence bounds for the linear logistic model. Our bounds significantly improve upon the state-of-the-art bound by Li et al. (2017) via recent developments of the self-concordant analysis of the logistic loss (Faury et al., 2020). Specifically, our confidence bound avoids a direct dependence on $1/\\kappa$, where $\\kappa$ is the minimal variance over all arms' reward distributions. In general, $1/\\kappa$ scales exponentially with the norm of the unknown linear parameter $\\theta^*$. Instead of relying on this worst-case quantity, our confidence bound for the reward of any given arm depends directly on the variance of that arm's reward distribution. We present two applications of our novel bounds to pure exploration and regret minimization logistic bandits improving upon state-of-the-art performance guarantees. For pure exploration, we also provide a lower bound highlighting a dependence on $1/\\kappa$ for a family of instances."
                },
                {
                    "title": "Self-Concordant Analysis of Generalized Linear Bandits with Forgetting",
                    "abstract": "Contextual sequential decision problems with categorical or numerical observations are ubiquitous and Generalized Linear Bandits (GLB) offer a solid theoretical framework to address them. In contrast to the case of linear bandits, existing algorithms for GLB have two drawbacks undermining their applicability. First, they rely on excessively pessimistic concentration bounds due to the non-linear nature of the model. Second, they require either non-convex projection steps or burn-in phases to enforce boundedness of the estimators. Both of these issues are worsened when considering non-stationary models, in which the GLB parameter may vary with time. In this work, we focus on self-concordant GLB (which include logistic and Poisson regression) with forgetting achieved either by the use of a sliding window or exponential weights. We propose a novel confidence-based algorithm for the maximum-likehood estimator with forgetting and analyze its perfomance in abruptly changing environments. These results as well as the accompanying numerical simulations highlight the potential of the proposed approach to address non-stationarity in GLB."
                },
                {
                    "title": "Instance-Wise Minimax-Optimal Algorithms for Logistic Bandits",
                    "abstract": "Logistic Bandits have recently attracted substantial attention, by providing an uncluttered yet challenging framework for understanding the impact of non-linearity in parametrized bandits. It was shown by Faury et al. (2020) that the learning-theoretic difficulties of Logistic Bandits can be embodied by a large (sometimes prohibitively) problem-dependent constant $\\kappa$, characterizing the magnitude of the reward's non-linearity. In this paper we introduce a novel algorithm for which we provide a refined analysis. This allows for a better characterization of the effect of non-linearity and yields improved problem-dependent guarantees. In most favorable cases this leads to a regret upper-bound scaling as $\\tilde{\\mathcal{O}}(d\\sqrt{T/\\kappa})$, which dramatically improves over the $\\tilde{\\mathcal{O}}(d\\sqrt{T}+\\kappa)$ state-of-the-art guarantees. We prove that this rate is minimax-optimal by deriving a $\\Omega(d\\sqrt{T/\\kappa})$ problem-dependent lower-bound. Our analysis identifies two regimes (permanent and transitory) of the regret, which ultimately re-conciliates Faury et al. (2020) with the Bayesian approach of Dong et al. (2019). In contrast to previous works, we find that in the permanent regime non-linearity can dramatically ease the exploration-exploitation trade-off. While it also impacts the length of the transitory phase in a problem-dependent fashion, we show that this impact is mild in most reasonable configurations."
                },
                {
                    "title": "High-Dimensional Probability: An Introduction with Applications in Data Science",
                    "abstract": "\u00a9 2018, Cambridge University Press Let us summarize our findings. A random projection of a set T in R n onto an m-dimensional subspace approximately preserves the geometry of T if m \u2a86 d ( T ) . For..."
                },
                {
                    "title": "Bandit Algorithms",
                    "abstract": "sets of environments and policies respectively and ` : E \u00d7\u03a0\u2192 [0, 1] a bounded loss function. Given a policy \u03c0 let `(\u03c0) = (`(\u03bd1, \u03c0), . . . , `(\u03bdN , \u03c0)) be the loss vector resulting from policy \u03c0. Define S = {`(\u03c0) : \u03c0 \u2208 \u03a0} and \u03bb(S) = {x \u2208 cl(S) : y 6< x for all y \u2208 S} , where y 6< x is defined to mean it is not true that yi \u2264 xi for all i with strict inequality for at least one i. Prove that if \u03bb(S) \u2286 S and S is convex, then for each x \u2208 \u03bb(S) there exists a prior q \u2208 P(E) and policy \u03c0\u2217 such that `(\u03c0) = x and \u2211 \u03bd\u2208E q(\u03bd)`(\u03bd, \u03c0\u2217) = min \u03c0\u2208\u03a0 \u2211"
                },
                {
                    "title": "Novel Change of Measure Inequalities with Applications to PAC-Bayesian Bounds and Monte Carlo Estimation",
                    "abstract": "We introduce several novel change of measure inequalities for two families of divergences: $f$-divergences and $\\alpha$-divergences. We show how the variational representation for $f$-divergences leads to novel change of measure inequalities. We also present a multiplicative change of measure inequality for $\\alpha$-divergences and a generalized version of Hammersley-Chapman-Robbins inequality. Finally, we present several applications of our change of measure inequalities, including PAC-Bayesian bounds for various classes of losses and non-asymptotic intervals for Monte Carlo estimates."
                },
                {
                    "title": "Improved Optimistic Algorithms for Logistic Bandits",
                    "abstract": "The generalized linear bandit framework has attracted a lot of attention in recent years by extending the well-understood linear setting and allowing to model richer reward structures. It notably covers the logistic model, widely used when rewards are binary. For logistic bandits, the frequentist regret guarantees of existing algorithms are $\\tilde{\\mathcal{O}}(\\kappa \\sqrt{T})$, where $\\kappa$ is a problem-dependent constant. Unfortunately, $\\kappa$ can be arbitrarily large as it scales exponentially with the size of the decision set. This may lead to significantly loose regret bounds and poor empirical performance. In this work, we study the logistic bandit with a focus on the prohibitive dependencies introduced by $\\kappa$. We propose a new optimistic algorithm based on a finer examination of the non-linearities of the reward function. We show that it enjoys a $\\tilde{\\mathcal{O}}(\\sqrt{T})$ regret with no dependency in $\\kappa$, but for a second order term. Our analysis is based on a new tail-inequality for self-normalized martingales, of independent interest."
                },
                {
                    "title": "Universal inference",
                    "abstract": "Significance Most statistical methods rely on certain mathematical conditions, known as regularity assumptions, to ensure their validity. Without these conditions, statistical quantities like P values and confidence intervals might not be valid. In this paper we give a surprisingly simple method for producing statistical significance statements without any regularity conditions. The resulting hypothesis tests can be used for any parametric model and for several nonparametric models. We propose a general method for constructing confidence sets and hypothesis tests that have finite-sample guarantees without regularity conditions. We refer to such procedures as \u201cuniversal.\u201d The method is very simple and is based on a modified version of the usual likelihood-ratio statistic that we call \u201cthe split likelihood-ratio test\u201d (split LRT) statistic. The (limiting) null distribution of the classical likelihood-ratio statistic is often intractable when used to test composite null hypotheses in irregular statistical models. Our method is especially appealing for statistical inference in these complex setups. The method we suggest works for any parametric model and also for some nonparametric models, as long as computing a maximum-likelihood estimator (MLE) is feasible under the null. Canonical examples arise in mixture modeling and shape-constrained inference, for which constructing tests and confidence sets has been notoriously difficult. We also develop various extensions of our basic methods. We show that in settings when computing the MLE is hard, for the purpose of constructing valid tests and intervals, it is sufficient to upper bound the maximum likelihood. We investigate some conditions under which our methods yield valid inferences under model misspecification. Further, the split LRT can be used with profile likelihoods to deal with nuisance parameters, and it can also be run sequentially to yield anytime-valid P values and confidence sequences. Finally, when combined with the method of sieves, it can be used to perform model selection with nested model classes."
                },
                {
                    "title": "Randomized Exploration in Generalized Linear Bandits",
                    "abstract": "We study two randomized algorithms for generalized linear bandits, GLM-TSL and GLM-FPL. GLM-TSL samples a generalized linear model (GLM) from the Laplace approximation to the posterior distribution. GLM-FPL fits a GLM to a randomly perturbed history of past rewards. We prove $\\tilde{O}(d \\sqrt{n \\log K})$ bounds on the $n$-round regret of GLM-TSL and GLM-FPL, where $d$ is the number of features and $K$ is the number of arms. The regret bound of GLM-TSL improves upon prior work and the regret bound of GLM-FPL is the first of its kind. We apply both GLM-TSL and GLM-FPL to logistic and neural network bandits, and show that they perform well empirically. In more complex models, GLM-FPL is significantly faster. Our results showcase the role of randomization, beyond sampling from the posterior, in exploration."
                },
                {
                    "title": "On the Performance of Thompson Sampling on Logistic Bandits",
                    "abstract": "We study the logistic bandit, in which rewards are binary with success probability $\\exp(\\beta a^\\top \\theta) / (1 + \\exp(\\beta a^\\top \\theta))$ and actions $a$ and coefficients $\\theta$ are within the $d$-dimensional unit ball. While prior regret bounds for algorithms that address the logistic bandit exhibit exponential dependence on the slope parameter $\\beta$, we establish a regret bound for Thompson sampling that is independent of $\\beta$. Specifically, we establish that, when the set of feasible actions is identical to the set of possible coefficient vectors, the Bayesian regret of Thompson sampling is $\\tilde{O}(d\\sqrt{T})$. We also establish a $\\tilde{O}(\\sqrt{d\\eta T}/\\lambda)$ bound that applies more broadly, where $\\lambda$ is the worst-case optimal log-odds and $\\eta$ is the \"fragility dimension,\" a new statistic we define to capture the degree to which an optimal action for one model fails to satisfice for others. We demonstrate that the fragility dimension plays an essential role by showing that, for any $\\epsilon > 0$, no algorithm can achieve $\\mathrm{poly}(d, 1/\\lambda)\\cdot T^{1-\\epsilon}$ regret."
                },
                {
                    "title": "A Short Note on Concentration Inequalities for Random Vectors with SubGaussian Norm",
                    "abstract": "In this note, we derive concentration inequalities for random vectors with subGaussian norm (a generalization of both subGaussian random vectors and norm bounded random vectors), which are tight up to logarithmic factors."
                },
                {
                    "title": "Mixture Martingales Revisited with Applications to Sequential Tests and Confidence Intervals",
                    "abstract": "This paper presents new deviation inequalities that are valid uniformly in time under adaptive sampling in a multi-armed bandit model. The deviations are measured using the Kullback-Leibler divergence in a given one-dimensional exponential family, and may take into account several arms at a time. They are obtained by constructing for each arm a mixture martingale based on a hierarchical prior, and by multiplying those martingales. Our deviation inequalities allow us to analyze stopping rules based on generalized likelihood ratios for a large class of sequential identification problems, and to construct tight confidence intervals for some functions of the means of the arms."
                },
                {
                    "title": "Logistic Regression: The Importance of Being Improper",
                    "abstract": "Learning linear predictors with the logistic loss---both in stochastic and online settings---is a fundamental task in machine learning and statistics, with direct connections to classification and boosting. Existing \"fast rates\" for this setting exhibit exponential dependence on the predictor norm, and Hazan et al. (2014) showed that this is unfortunately unimprovable. Starting with the simple observation that the logistic loss is $1$-mixable, we design a new efficient improper learning algorithm for online logistic regression that circumvents the aforementioned lower bound with a regret bound exhibiting a doubly-exponential improvement in dependence on the predictor norm. This provides a positive resolution to a variant of the COLT 2012 open problem of McMahan and Streeter (2012) when improper learning is allowed. This improvement is obtained both in the online setting and, with some extra work, in the batch statistical setting with high probability. We also show that the improved dependence on predictor norm is near-optimal. \nLeveraging this improved dependency on the predictor norm yields the following applications: (a) we give algorithms for online bandit multiclass learning with the logistic loss with an $\\tilde{O}(\\sqrt{n})$ relative mistake bound across essentially all parameter ranges, thus providing a solution to the COLT 2009 open problem of Abernethy and Rakhlin (2009), and (b) we give an adaptive algorithm for online multiclass boosting with optimal sample complexity, thus partially resolving an open problem of Beygelzimer et al. (2015) and Jung et al. (2017). Finally, we give information-theoretic bounds on the optimal rates for improper logistic regression with general function classes, thereby characterizing the extent to which our improvement for linear classes extends to other parametric and even nonparametric settings."
                },
                {
                    "title": "Deep Reinforcement Learning from Human Preferences",
                    "abstract": "For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems. In this work, we explore goals defined in terms of (non-expert) human preferences between pairs of trajectory segments. We show that this approach can effectively solve complex RL tasks without access to the reward function, including Atari games and simulated robot locomotion, while providing feedback on less than one percent of our agent's interactions with the environment. This reduces the cost of human oversight far enough that it can be practically applied to state-of-the-art RL systems. To demonstrate the flexibility of our approach, we show that we can successfully train complex novel behaviors with about an hour of human time. These behaviors and environments are considerably more complex than any that have been previously learned from human feedback."
                },
                {
                    "title": "Scalable Generalized Linear Bandits: Online Computation and Hashing",
                    "abstract": "Generalized Linear Bandits (GLBs), a natural extension of the stochastic linear bandits, has been popular and successful in recent years. However, existing GLBs scale poorly with the number of rounds and the number of arms, limiting their utility in practice. This paper proposes new, scalable solutions to the GLB problem in two respects. First, unlike existing GLBs, whose per-time-step space and time complexity grow at least linearly with time $t$, we propose a new algorithm that performs online computations to enjoy a constant space and time complexity. At its heart is a novel Generalized Linear extension of the Online-to-confidence-set Conversion (GLOC method) that takes \\emph{any} online learning algorithm and turns it into a GLB algorithm. As a special case, we apply GLOC to the online Newton step algorithm, which results in a low-regret GLB algorithm with much lower time and memory complexity than prior work. Second, for the case where the number $N$ of arms is very large, we propose new algorithms in which each next arm is selected via an inner product search. Such methods can be implemented via hashing algorithms (i.e., \"hash-amenable\") and result in a time complexity sublinear in $N$. While a Thompson sampling extension of GLOC is hash-amenable, its regret bound for $d$-dimensional arm sets scales with $d^{3/2}$, whereas GLOC's regret bound scales with $d$. Towards closing this gap, we propose a new hash-amenable algorithm whose regret bound scales with $d^{5/4}$. Finally, we propose a fast approximate hash-key computation (inner product) with a better accuracy than the state-of-the-art, which can be of independent interest. We conclude the paper with preliminary experimental results confirming the merits of our methods."
                },
                {
                    "title": "On Kernelized Multi-armed Bandits",
                    "abstract": "We consider the stochastic bandit problem with a continuous set of arms, with the expected reward function over the arms assumed to be fixed but unknown. We provide two new Gaussian process-based algorithms for continuous bandit optimization-Improved GP-UCB (IGP-UCB) and GP-Thomson sampling (GP-TS), and derive corresponding regret bounds. Specifically, the bounds hold when the expected reward function belongs to the reproducing kernel Hilbert space (RKHS) that naturally corresponds to a Gaussian process kernel used as input by the algorithms. Along the way, we derive a new self-normalized concentration inequality for vector- valued martingales of arbitrary, possibly infinite, dimension. Finally, experimental evaluation and comparisons to existing algorithms on synthetic and real-world environments are carried out that highlight the favorable gains of the proposed strategies in many cases."
                },
                {
                    "title": "Provably Optimal Algorithms for Generalized Linear Contextual Bandits",
                    "abstract": "Contextual bandits are widely used in Internet services from news recommendation to advertising. Generalized linear models (logistical regression in particular) have demonstrated stronger performance than linear models in many applications. However, most theoretical analyses on contextual bandits so far are on linear bandits. In this work, we propose an upper confidence bound based algorithm for generalized linear contextual bandits, which achieves an \\tilde{O}(\\sqrt{dT}) regret over T rounds with d dimensional feature vectors. This regret matches the minimax lower bound, up to logarithmic terms, and improves on the best previous result by a \\sqrt{d} factor, assuming the number of arms is fixed. A key component in our analysis is to establish a new, sharp finite-sample confidence bound for maximum-likelihood estimates in generalized linear models, which may be of independent interest. We also analyze a simpler upper confidence bound algorithm, which is useful in practice, and prove it to have optimal regret for the certain cases."
                },
                {
                    "title": "Linear Thompson Sampling Revisited",
                    "abstract": "We derive an alternative proof for the regret of Thompson sampling (\\ts) in the stochastic linear bandit setting. While we obtain a regret bound of order $\\wt{O}(d^{3/2}\\sqrt{T})$ as in previous results, the proof sheds new light on the functioning of the \\ts. We leverage on the structure of the problem to show how the regret is related to the sensitivity (i.e., the gradient) of the objective function and how selecting optimal arms associated to \\textit{optimistic} parameters does control it. Thus we show that \\ts can be seen as a generic randomized algorithm where the sampling distribution is designed to have a fixed probability of being optimistic, at the cost of an additional $\\sqrt{d}$ regret factor compared to a UCB-like approach. Furthermore, we show that our proof can be readily applied to regularized linear optimization and generalized linear model problems."
                },
                {
                    "title": "Fast Rates for General Unbounded Loss Functions: From ERM to Generalized Bayes",
                    "abstract": "We present new excess risk bounds for general unbounded loss functions including log loss and squared loss, where the distribution of the losses may be heavy-tailed. The bounds hold for general estimators, but they are optimized when applied to $\\eta$-generalized Bayesian, MDL, and empirical risk minimization estimators. In the case of log loss, the bounds imply convergence rates for generalized Bayesian inference under misspecification in terms of a generalization of the Hellinger metric as long as the learning rate $\\eta$ is set correctly. For general loss functions, our bounds rely on two separate conditions: the $v$-GRIP (generalized reversed information projection) conditions, which control the lower tail of the excess loss; and the newly introduced witness condition, which controls the upper tail. The parameter $v$ in the $v$-GRIP conditions determines the achievable rate and is akin to the exponent in the Tsybakov margin condition and the Bernstein condition for bounded losses, which the $v$-GRIP conditions generalize; favorable $v$ in combination with small model complexity leads to $\\tilde{O}(1/n)$ rates. The witness condition allows us to connect the excess risk to an \"annealed\" version thereof, by which we generalize several previous results connecting Hellinger and Renyi divergence to KL divergence."
                },
                {
                    "title": "Policies for Contextual Bandit Problems with Count Payoffs",
                    "abstract": "The contextual bandit problem has been of major interest in the last few years. This corresponds to a sequential decision process where an agent has to choose at each iteration an action to perform, according to some knowledge about the decision environment and the current available actions, with the aim to maximize a cumulative amount of rewards over time. Many instances of the problem exist, depending on the kind of rewards we collect - real, binary, natural - and various algorithms are known to be efficient for some of these instances, either empirically or theoretically. In this paper we focus on the case of count payoffs, which corresponds to bandit problems where rewards are integer rewards, potentially unbounded. Based on a Bayesian Poisson regression model, we propose two new contextual bandit algorithms for this particular case with several concrete applications in real life: an Upper Confidence Bound algorithm and a Thompson Sampling strategy. Our approaches present the advantage to remain analytically tractable and computationally efficient. We experiment the algorithms on both simulated data and a real world scenario of spread maximization on a social network."
                },
                {
                    "title": "Fast rates in statistical and online learning",
                    "abstract": "The speed with which a learning algorithm converges as it is presented with more data is a central problem in machine learning -- a fast rate of convergence means less data is needed for the same level of performance. The pursuit of fast rates in online and statistical learning has led to the discovery of many conditions in learning theory under which fast learning is possible. We show that most of these conditions are special cases of a single, unifying condition, that comes in two forms: the central condition for 'proper' learning algorithms that always output a hypothesis in the given model, and stochastic mixability for online algorithms that may make predictions outside of the model. We show that under surprisingly weak assumptions both conditions are, in a certain sense, equivalent. The central condition has a re-interpretation in terms of convexity of a set of pseudoprobabilities, linking it to density estimation under misspecification. For bounded losses, we show how the central condition enables a direct proof of fast rates and we prove its equivalence to the Bernstein condition, itself a generalization of the Tsybakov margin condition, both of which have played a central role in obtaining fast rates in statistical learning. Yet, while the Bernstein condition is two-sided, the central condition is one-sided, making it more suitable to deal with unbounded losses. In its stochastic mixability form, our condition generalizes both a stochastic exp-concavity condition identified by Juditsky, Rigollet and Tsybakov and Vovk's notion of mixability. Our unifying conditions thus provide a substantial step towards a characterization of fast rates in statistical learning, similar to how classical mixability characterizes constant regret in the sequential prediction with expert advice setting."
                },
                {
                    "title": "An Introduction to Matrix Concentration Inequalities",
                    "abstract": "In recent years, random matrices have come to play a major role in computational mathematics, but most of the classical areas of random matrix theory remain the province of experts. Over the last decade, with the advent of matrix concentration inequalities, research has advanced to the point where we can conquer many (formerly) challenging problems with a page or two of arithmetic. The aim of this monograph is to describe the most successful methods from this area along with some interesting examples that these techniques can illuminate."
                },
                {
                    "title": "Choosing which message to publish on social networks: A contextual bandit approach",
                    "abstract": "Maximizing the spread and influence of the messages being published is a challenge for many social network users. Selecting the right content according to the information context and the user characteristics is essential for achieving this goal. We propose a model to automatically choose which information to publish on social networks given a set of possible messages. This model will tend to maximize the spread of the published message for a specific audience. The algorithm is based on the use of a contextual bandit model treating each new potential message as an arm to be selected. We conduct experiments on a Twitter dataset, comparing different algorithms and exploring the influence of the content and the characteristics of the messages on the information spread. The results demonstrate the model's ability to maximize the published information flow as well as it's ability to adapt its behavior to each particular audience."
                },
                {
                    "title": "Online-to-Confidence-Set Conversions and Application to Sparse Stochastic Bandits",
                    "abstract": "We introduce a novel technique, which we call online-to-confidence-set conversion. The technique allows us to construct highprobability confidence sets for linear prediction with correlated inputs given the predictions of any algorithm (e.g., online LASSO, exponentiated gradient algorithm, online least-squares, p-norm algorithm) targeting online learning with linear predictors and the quadratic loss. By construction, the size of the confidence set is directly governed by the regret of the online learning algorithm. Constructing tight confidence sets is interesting on its own, but the new technique is given extra weight by the fact having access tight confidence sets underlies a number of important problems. The advantage of our construction here is that progress in constructing better algorithms for online prediction problems directly translates into tighter confidence sets. In this paper, this is demonstrated in the case of linear stochastic bandits. In particular, we introduce the sparse variant of linear stochastic bandits and show that a recent online algorithm together with our online-to-confidence-set conversion allows one to derive algorithms that can exploit if the reward is a function of a sparse linear combination of the components of the chosen action. Appearing in Proceedings of the 15 International Conference on Artificial Intelligence and Statistics (AISTATS) 2012, La Palma, Canary Islands. Volume XX of JMLR: W&CP XX. Copyright 2012 by the authors."
                },
                {
                    "title": "Optimistic Bayesian Sampling in Contextual-Bandit Problems",
                    "abstract": "In sequential decision problems in an unknown environment, the decision maker often faces a dilemma over whether to explore to discover more about the environment, or to exploit current knowledge. We address the exploration-exploitation dilemma in a general setting encompassing both standard and contextualised bandit problems. The contextual bandit problem has recently resurfaced in attempts to maximise click-through rates in web based applications, a task with significant commercial interest. \n \nIn this article we consider an approach of Thompson (1933) which makes use of samples from the posterior distributions for the instantaneous value of each action. We extend the approach by introducing a new algorithm, Optimistic Bayesian Sampling (OBS), in which the probability of playing an action increases with the uncertainty in the estimate of the action value. This results in better directed exploratory behaviour. \n \nWe prove that, under unrestrictive assumptions, both approaches result in optimal behaviour with respect to the average reward criterion of Yang and Zhu (2002). We implement OBS and measure its performance in simulated Bernoulli bandit and linear regression domains, and also when tested with the task of personalised news article recommendation on a Yahoo! Front Page Today Module data set. We find that OBS performs competitively when compared to recently proposed benchmark algorithms and outperforms Thompson's method throughout."
                },
                {
                    "title": "Improved Algorithms for Linear Stochastic Bandits",
                    "abstract": "We improve the theoretical analysis and empirical performance of algorithms for the stochastic multi-armed bandit problem and the linear stochastic multi-armed bandit problem. In particular, we show that a simple modification of Auer's UCB algorithm (Auer, 2002) achieves with high probability constant regret. More importantly, we modify and, consequently, improve the analysis of the algorithm for the for linear stochastic bandit problem studied by Auer (2002), Dani et al. (2008), Rusmevichientong and Tsitsiklis (2010), Li et al. (2010). Our modification improves the regret bound by a logarithmic factor, though experiments show a vast improvement. In both cases, the improvement stems from the construction of smaller confidence sets. For their construction we use a novel tail inequality for vector-valued martingales."
                },
                {
                    "title": "An unbiased offline evaluation of contextual bandit algorithms with generalized linear models",
                    "abstract": "Contextual bandit algorithms have become popular tools in online recommendation and advertising systems. Offline evaluation of the effectiveness of new algorithms in these applications is critical for protecting online user experiences but very challenging due to their \"partial-label\" nature. A common practice is to create a simulator which simulates the online environment for the problem at hand and then run an algorithm against this simulator. However, creating the simulator itself is often difficult and modeling bias is usually unavoidably introduced. \n \nThe purpose of this paper is two-fold. First, we review a recently proposed offline evaluation technique. Different from simulator-based approaches, the method is completely data-driven, is easy to adapt to different applications, and more importantly, provides provably unbiased evaluations. We argue for the wide use of this technique as standard practice when comparing bandit algorithms in real-life problems. Second, as an application of this technique, we compare and validate a number of new algorithms based on generalized linear models. Experiments using real Yahoo! data suggest substantial improvement over algorithms with linear models when the rewards are binary."
                },
                {
                    "title": "Parametric Bandits: The Generalized Linear Case",
                    "abstract": "We consider structured multi-armed bandit problems based on the Generalized Linear Model (GLM) framework of statistics. For these bandits, we propose a new algorithm, called GLM-UCB. We derive finite time, high probability bounds on the regret of the algorithm, extending previous analyses developed for the linear bandits to the non-linear case. The analysis highlights a key difficulty in generalizing linear bandit algorithms to the non-linear case, which is solved in GLM-UCB by focusing on the reward space rather than on the parameter space. Moreover, as the actual effectiveness of current parameterized bandit algorithms is often poor in practice, we provide a tuning method based on asymptotic arguments, which leads to significantly better practical performance. We present two numerical experiments on real-world data that illustrate the potential of the GLM-UCB approach."
                },
                {
                    "title": "Introduction to the non-asymptotic analysis of random matrices",
                    "abstract": "This is a tutorial on some basic non-asymptotic methods and concepts in random matrix theory. The reader will learn several tools for the analysis of the extreme singular values of random matrices with independent rows or columns. Many of these methods sprung off from the development of geometric functional analysis since the 1970's. They have applications in several fields, most notably in theoretical computer science, statistics and signal processing. A few basic applications are covered in this text, particularly for the problem of estimating covariance matrices in statistics and for validating probabilistic constructions of measurement matrices in compressed sensing. These notes are written particularly for graduate students and beginning researchers in different areas, including functional analysts, probabilists, theoretical statisticians, electrical engineers, and theoretical computer scientists."
                },
                {
                    "title": "A contextual-bandit approach to personalized news article recommendation",
                    "abstract": "Personalized web services strive to adapt their services (advertisements, news articles, etc.) to individual users by making use of both content and user information. Despite a few recent advances, this problem remains challenging for at least two reasons. First, web service is featured with dynamically changing pools of content, rendering traditional collaborative filtering methods inapplicable. Second, the scale of most web services of practical interest calls for solutions that are both fast in learning and computation.\n In this work, we model personalized recommendation of news articles as a contextual bandit problem, a principled approach in which a learning algorithm sequentially selects articles to serve users based on contextual information about the users and articles, while simultaneously adapting its article-selection strategy based on user-click feedback to maximize total user clicks.\n The contributions of this work are three-fold. First, we propose a new, general contextual bandit algorithm that is computationally efficient and well motivated from learning theory. Second, we argue that any bandit algorithm can be reliably evaluated offline using previously recorded random traffic. Finally, using this offline evaluation method, we successfully applied our new algorithm to a Yahoo! Front Page Today Module dataset containing over 33 million events. Results showed a 12.5% click lift compared to a standard context-free bandit algorithm, and the advantage becomes even greater when data gets more scarce."
                },
                {
                    "title": "Generalised Kernel Machines",
                    "abstract": "The generalised linear model (GLM) is the standard approach in classical statistics for regression tasks where it is appropriate to measure the data misfit using a likelihood drawn from the exponential family of distributions. In this paper, we apply the kernel trick to give a non-linear variant of the GLM, the generalised kernel machine (GKM), in which a regularised GLM is constructed in a fixed feature space implicitly defined by a Mercer kernel. The MATLAB symbolic maths toolbox is used to automatically create a suite of generalised kernel machines, including methods for automated model selection based on approximate leave-one-out cross-validation. In doing so, we provide a common framework encompassing a wide range of existing and novel kernel learning methods, and highlight their connections with earlier techniques from classical statistics. Examples including kernel ridge regression, kernel logistic regression and kernel Poisson regression are given to demonstrate the flexibility and utility of the generalised kernel machine."
                },
                {
                    "title": "Generalized functional linear models",
                    "abstract": "We propose a generalized functional linear regression model for a regression situation where the response variable is a scalar and the predictor is a random function. A linear predictor is obtained by forming the scalar product of the predictor function with a smooth parameter function, and the expected value of the response is related to this linear predictor via a link function. If, in addition, a variance function is speci\ufb01ed, this leads to a functional estimating equation which corresponds to maximizing a functional quasi-likelihood. This general approach includes the special cases of the functional linear model, as well as functional Poisson regression and functional binomial regression. The latter leads to procedures for classi\ufb01cation and discrimination of stochastic processes and functional data. We also consider the situation where the link and variance functions are unknown and are estimated nonparametrically from the data, using a semiparametric quasi-likelihood procedure. An essential step in our proposal is dimension reduction by approximating the predictor processes with a truncated Karhunen\u2013Lo`eve expansion. We develop asymptotic inference for the proposed class of generalized regression models. In the proposed asymptotic approach, the truncation parameter increases with sample size, and a martingale central limit theorem is applied to establish the resulting increasing dimension asymptotics. We establish asymptotic normality for a properly scaled distance between estimated and true functions that corresponds to a suitable L 2 metric and is de\ufb01ned through a generalized covariance operator. As a consequence, we obtain asymptotic tests and simultaneous con\ufb01dence bands for the parameter function that determines the gated in a simulation study. We \ufb01nd that the practical selection of the number of components works well with the AIC criterion, and this \ufb01nding is supported by theoretical considerations. We include an application to the classi\ufb01cation of med\ufb02ies regarding their remaining longevity status, based on the observed initial egg-laying curve for each of 534 female med\ufb02ies."
                },
                {
                    "title": "SELF-NORMALIZED PROCESSES: EXPONENTIAL INEQUALITIES, MOMENT BOUNDS AND ITERATED LOGARITHM LAWS",
                    "abstract": "Self-normalized processes arise naturally in statistical applications. Being unit free, they are not affected by scale changes. Moreover, self-normalization often eliminates or weakens moment assumptions. In this paper we present several exponential and moment inequalities, particularly those related to laws of the iterated logarithm, for self-normalized random variables including martingales. Tail probability bounds are also derived. For random variables Bt>0 and At, let Yt(\u03bb)=exp{\u03bbAt\u2212\u03bb2Bt2/2}. We develop inequalities for the moments of At/Bt or supt\u22650At/{Bt(log logBt)1/2} and variants thereof, when EYt(\u03bb)\u22641 or when Yt(\u03bb) is a supermartingale, for all \u03bb belonging to some interval. Our results are valid for a wide class of random processes including continuous martingales with At=Mt and $B_{t}=\\sqrt {\\langle M\\rangle _{t}}$ , and sums of conditionally symmetric variables di with At=\u2211i=1tdi and $B_{t}=\\sqrt{\\sum_{i=1}^{t}d_{i}^{2}}$ . A sharp maximal inequality for conditionally symmetric random variables and for continuous local martingales with values in Rm, m\u22651, is also established. Another development in this paper is a bounded law of the iterated logarithm for general adapted sequences that are centered at certain truncated conditional expectations and self-normalized by the square root of the sum of squares. The key ingredient in this development is a new exponential supermartingale involving \u2211i=1tdi and \u2211i=1tdi2. A compact law of the iterated logarithm for self-normalized martingales is also derived in this connection."
                },
                {
                    "title": "Using Confidence Bounds for Exploitation-Exploration Trade-offs",
                    "abstract": "We show how a standard tool from statistics --- namely confidence bounds --- can be used to elegantly deal with situations which exhibit an exploitation-exploration trade-off. Our technique for designing and analyzing algorithms for such situations is general and can be applied when an algorithm has to make exploitation-versus-exploration decisions based on uncertain information provided by a random process. We apply our technique to two models with such an exploitation-exploration trade-off. For the adversarial bandit problem with shifting our new algorithm suffers only O((ST)1/2) regret with high probability over T trials with S shifts. Such a regret bound was previously known only in expectation. The second model we consider is associative reinforcement learning with linear value functions. For this model our technique improves the regret from O(T3/4) to O(T1/2)."
                },
                {
                    "title": "Generalized Linear Models",
                    "abstract": "This is the \u008e rst book on generalized linear models written by authors not mostly associated with the biological sciences. Subtitled \u201cWith Applications in Engineering and the Sciences,\u201d this book\u2019s authors all specialize primarily in engineering statistics. The \u008e rst author has produced several recent editions of Walpole, Myers, and Myers (1998), the last reported by Ziegel (1999). The second author has had several editions of Montgomery and Runger (1999), recently reported by Ziegel (2002). All of the authors are renowned experts in modeling. The \u008e rst two authors collaborated on a seminal volume in applied modeling (Myers and Montgomery 2002), which had its recent revised edition reported by Ziegel (2002). The last two authors collaborated on the most recent edition of a book on regression analysis (Montgomery, Peck, and Vining (2001), reported by Gray (2002), and the \u008e rst author has had multiple editions of his own regression analysis book (Myers 1990), the latest of which was reported by Ziegel (1991). A comparable book with similar objectives and a more speci\u008e c focus on logistic regression, Hosmer and Lemeshow (2000), reported by Conklin (2002), presumed a background in regression analysis and began with generalized linear models. The Preface here (p. xi) indicates an identical requirement but nonetheless begins with 100 pages of material on linear and nonlinear regression. Most of this will probably be a review for the readers of the book. Chapter 2, \u201cLinear Regression Model,\u201d begins with 50 pages of familiar material on estimation, inference, and diagnostic checking for multiple regression. The approach is very traditional, including the use of formal hypothesis tests. In industrial settings, use of p values as part of a risk-weighted decision is generally more appropriate. The pedagologic approach includes formulas and demonstrations for computations, although computing by Minitab is eventually illustrated. Less-familiar material on maximum likelihood estimation, scaled residuals, and weighted least squares provides more speci\u008e c background for subsequent estimation methods for generalized linear models. This review is not meant to be disparaging. The authors have packed a wealth of useful nuggets for any practitioner in this chapter. It is thoroughly enjoyable to read. Chapter 3, \u201cNonlinear Regression Models,\u201d is arguably less of a review, because regression analysis courses often give short shrift to nonlinear models. The chapter begins with a great example on the pitfalls of linearizing a nonlinear model for parameter estimation. It continues with the effective balancing of explicit statements concerning the theoretical basis for computations versus the application and demonstration of their use. The details of maximum likelihood estimation are again provided, and weighted and generalized regression estimation are discussed. Chapter 4 is titled \u201cLogistic and Poisson Regression Models.\u201d Logistic regression provides the basic model for generalized linear models. The prior development for weighted regression is used to motivate maximum likelihood estimation for the parameters in the logistic model. The algebraic details are provided. As in the development for linear models, some of the details are pushed into an appendix. In addition to connecting to the foregoing material on regression on several occasions, the authors link their development forward to their following chapter on the entire family of generalized linear models. They discuss score functions, the variance-covariance matrix, Wald inference, likelihood inference, deviance, and overdispersion. Careful explanations are given for the values provided in standard computer software, here PROC LOGISTIC in SAS. The value in having the book begin with familiar regression concepts is clearly realized when the analogies are drawn between overdispersion and nonhomogenous variance, or analysis of deviance and analysis of variance. The authors rely on the similarity of Poisson regression methods to logistic regression methods and mostly present illustrations for Poisson regression. These use PROC GENMOD in SAS. The book does not give any of the SAS code that produces the results. Two of the examples illustrate designed experiments and modeling. They include discussion of subset selection and adjustment for overdispersion. The mathematic level of the presentation is elevated in Chapter 5, \u201cThe Family of Generalized Linear Models.\u201d First, the authors unify the two preceding chapters under the exponential distribution. The material on the formal structure for generalized linear models (GLMs), likelihood equations, quasilikelihood, the gamma distribution family, and power functions as links is some of the most advanced material in the book. Most of the computational details are relegated to appendixes. A discussion of residuals returns one to a more practical perspective, and two long examples on gamma distribution applications provide excellent guidance on how to put this material into practice. One example is a contrast to the use of linear regression with a log transformation of the response, and the other is a comparison to the use of a different link function in the previous chapter. Chapter 6 considers generalized estimating equations (GEEs) for longitudinal and analogous studies. The \u008e rst half of the chapter presents the methodology, and the second half demonstrates its application through \u008e ve different examples. The basis for the general situation is \u008e rst established using the case with a normal distribution for the response and an identity link. The importance of the correlation structure is explained, the iterative estimation procedure is shown, and estimation for the scale parameters and the standard errors of the coef\u008e cients is discussed. The procedures are then generalized for the exponential family of distributions and quasi-likelihood estimation. Two of the examples are standard repeated-measures illustrations from biostatistical applications, but the last three illustrations are all interesting reworkings of industrial applications. The GEE computations in PROC GENMOD are applied to account for correlations that occur with multiple measurements on the subjects or restrictions to randomizations. The examples show that accounting for correlation structure can result in different conclusions. Chapter 7, \u201cFurther Advances and Applications in GLM,\u201d discusses several additional topics. These are experimental designs for GLMs, asymptotic results, analysis of screening experiments, data transformation, modeling for both a process mean and variance, and generalized additive models. The material on experimental designs is more discursive than prescriptive and as a result is also somewhat theoretical. Similar comments apply for the discussion on the quality of the asymptotic results, which wallows a little too much in reports on various simulation studies. The examples on screening and data transformations experiments are again reworkings of analyses of familiar industrial examples and another obvious motivation for the enthusiasm that the authors have developed for using the GLM toolkit. One can hope that subsequent editions will similarly contain new examples that will have caused the authors to expand the material on generalized additive models and other topics in this chapter. Designating myself to review a book that I know I will love to read is one of the rewards of being editor. I read both of the editions of McCullagh and Nelder (1989), which was reviewed by Schuenemeyer (1992). That book was not fun to read. The obvious enthusiasm of Myers, Montgomery, and Vining and their reliance on their many examples as a major focus of their pedagogy make Generalized Linear Models a joy to read. Every statistician working in any area of applied science should buy it and experience the excitement of these new approaches to familiar activities."
                },
                {
                    "title": "Confidence sequences for mean, variance, and median.",
                    "abstract": "UNIVERSITY OF CALIFORNIA, BERKELEY Communicated by J. Neyman, May 1, 1967 1. Normal Means and Variances.-Let x1,x2,... be independent, identically dis1 n tributed random variables and put ((t) = E(exp(tx)), n= -x. In earlier n l papers' 2 we proved, among other things of a similar nature, the following inequality which we state as a LEMMA. Assume that (p(t) < exp(t2/2) for all t > 0. Let a > 1, c > 1 be arbitrary constants and set (with log2n = log log n)"
                },
                {
                    "title": "Iterated logarithm inequalities.",
                    "abstract": "1. Introduction\u2014Let x,x 1, x 2 \u2026 be a sequence of independent, identically distributed random variables with mean 0, variance 1, and moment generating function \u03d5(t) = E(etz) finite in some neighborhood of t= 0, and put S n = x 1, + \u2026 + x n, \\(\\bar x\\) n = S n/n. For any sequence of positive constants a n, n \u2265 1, let P m = P(|\\(\\bar x\\) n| \u2265 a n for some n \u2265 m)."
                },
                {
                    "title": "Some aspects of the sequential design of experiments",
                    "abstract": "Until recently, statistical theory has been restricted to the design and analysis of sampling experiments in which the size and composition of the samples are completely determined before the experimentation begins. The reasons for this are partly historical, dating back to the time when the statistician was consulted, if at all, only after the experiment was over, and partly intrinsic in the mathematical difficulty of working with anything but a fixed number of independent random variables. A major advance now appears to be in the making with the creation of a theory of the sequential design of experiments, in which the size and composition of the samples are not fixed in advance but are functions of the observations themselves."
                },
                {
                    "title": "Optimal Regret with Limited Adaptivity for Generalized Linear Contextual Bandits",
                    "abstract": "We study the generalized linear contextual bandit problem within the requirements of limited adaptivity. In this paper, we present two algorithms, B-GLinCB and RS-GLinCB , that address, respectively, two prevalent limited adaptivity models: batch learning with stochastic contexts and rare policy switches with adversarial contexts. For both these models, we establish essentially tight regret bounds. Notably, in the obtained bounds, we manage to eliminate a dependence on a key parameter \u03ba , which captures the non-linearity of the underlying reward model. For our batch learning algorithm B-GLinCB , with \u2126 (log log T ) batches, the regret scales as \u02dc O ( \u221a T ) . Further, we establish that our rarely switching algorithm RS-GLinCB updates its policy at most \u02dc O (log 2 T ) times and achieves a regret of \u02dc O ( \u221a T ) . Our approach for removing the dependence on \u03ba for generalized linear contextual bandits might be of independent interest."
                },
                {
                    "title": "Online (Multinomial) Logistic Bandit: Improved Regret and Constant Computation Cost",
                    "abstract": "This paper investigates the logistic bandit problem, a variant of the generalized linear bandit model that utilizes a logistic model to depict the feedback from an action. While most existing research focuses on the binary logistic bandit problem, the multinomial case, which considers more than two possible feedback values, offers increased practical relevance and adaptability for use in complex decision-making problems such as reinforcement learning. In this paper, we provide an algorithm that enjoys both statistical and computational efficiency for the logistic bandit problem. In the binary case, our method improves the state-of-the-art binary logistic bandit method by reducing the per-round computation cost from O (log T ) to O (1) with respect to the time horizon T , while still preserving the minimax optimal guarantee up to logarithmic factors. In the multinomial case, with K + 1 potential feedback values, our algorithm achieves an (cid:101) O ( K \u221a T ) regret bound with O (1) computational cost per round. The result not only improves the (cid:101) O ( K \u221a \u03baT ) bound for the best-known tractable algorithm\u2014where the large constant \u03ba increases exponentially with the diameter of the parameter domain\u2014but also reduces the O ( T ) computational complexity demanded by the previous method."
                },
                {
                    "title": "Geometric Measure Theory",
                    "abstract": "Geometric measure theory is an area of analysis concerned with solving geometric problems via measure theoretic techniques. The canonical motivating physical problem is probably that investigated experimentally by Plateau in the nineteenth century [3]: given a boundary wire, how does one find the (minimal) soap film which spans it? Slightly more mathematically, given a boundary curve, find the surface of minimal area spanning it. The many different approaches to solving this problem have found utility in most areas of modern mathematics and geometric measure theory is no exception: techniques and ideas from geometric measure theory have been found useful in the study of partial differential equations, the calculus of variations, harmonic analysis, and fractals. Successes in the field include: classifying the structure of singularities in soap films (see [18], together with the fine descriptive article [4]); showing that the standard \u2018double bubble\u2019 is the optimal shape for enclosing two prescribed volumes in space [13], and developing powerful computer software for modelling the evolution of surfaces under the action of physical forces [7]. The main reference text for the subject is still Federer\u2019s comprehensive book [11]. It is very densely written and Morgan\u2019s book [15] serves as a useful guide through it. Federer\u2019s colloquium lectures [10] provide a comprehensive overview of the subject and contain a summary of the main results in his book [11]. More recent books include Simon [17], which contains an introduction to the theory of varifolds and Allard\u2019s regularity theorem, and Mattila\u2019s book [14] which includes information about tangent measures and their uses. Both of these books are also suitable as introductions to the area. For a slightly different slant, the book by Evans and Gariepy [9], discusses applications of some of the ideas of geometric measure theory in the theory of Sobolev Spaces and functions of bounded variation. Many variational problems are solved by enlarging the allowed class of solutions, showing that in this enlarged class a solution exists, and then showing that the solution possesses more regularity than an arbitrary element of the enlarged class. Much of the work in geometric measure theory has been directed towards placing this informal description on a formal footing appropriate for the study of surfaces. The key concept underlying the whole theory is that of rectifiability: this is a measure theoretic notion of smoothness. A set E in Euclidean n-space, R, is"
                }
            ],
            "categories": [
                "stat.ML",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we derive improved time-uniform confidence sequences for generalized linear models (GLMs) that provide tighter bounds on the uncertainty of the underlying model parameters?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the limitations of existing confidence sequences for GLMs, which are widely used in various applications such as recommendation systems and social network analysis. Improved confidence sequences can enhance the reliability of sequential decision-making processes, leading to safer and more effective machine learning models. This advancement could pave the way for more robust algorithms in interactive learning scenarios, ultimately influencing future research directions in uncertainty quantification and adaptive statistical inference.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexities of GLMs and the need for confidence sequences that are both valid and tight. Naive approaches may fail due to the poly(S) factor in the radius of existing confidence sequences, which can lead to overly conservative estimates. Additionally, the global worst-case curvature of the inverse link function \u03bc complicates the derivation of effective confidence bounds. Overcoming these technical obstacles requires a deep understanding of the statistical properties of GLMs and innovative methodologies to improve the performance of confidence sequences.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on specific instances of GLMs, such as Gaussian and Bernoulli distributions, which limits the generalizability of their findings. The existing confidence sequences suffer from significant drawbacks, such as the poly(S) factor, which has not been adequately addressed in the literature. Barriers to solving this problem include a lack of comprehensive frameworks that can accommodate the diverse characteristics of GLMs. My approach aims to fill these gaps by developing a more unified and efficient method for deriving confidence sequences applicable to a broader class of GLMs.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nMy proposed methodology involves developing a novel framework for constructing time-uniform confidence sequences for GLMs that minimizes the radius of uncertainty bounds. I will utilize a combination of theoretical analysis and empirical validation on benchmark datasets relevant to GLM applications. The metrics for evaluation will include the tightness of the confidence intervals and the validity of the sequences across various GLM instances. The expected outcomes include significantly improved confidence sequences that provide tighter bounds on the unknown parameters, enhancing the safety and effectiveness"
            }
        },
        "author_data": {
            "dc1006f2-9f43-4fe9-98a7-3e02e549a5f5": {
                "pk": "dc1006f2-9f43-4fe9-98a7-3e02e549a5f5",
                "name": "Junghyun Lee",
                "collaborators": [
                    "Se-Young Yun",
                    "Chulhee Yun",
                    "Laura Schmid",
                    "Yassir Jedra",
                    "Alexandre Prouti\u00e8re",
                    "Hanseul Cho",
                    "Prin Phunyaphibarn",
                    "Bohan Wang",
                    "Huishuai Zhang",
                    "Donghun Jung"
                ],
                "domain": [
                    "Multi-Agent Systems",
                    "Fairness in Machine Learning",
                    "Reinforcement Learning",
                    "Quantum Computing"
                ],
                "publications": [
                    {
                        "title": "Flooding with Absorption: An Efficient Protocol for Heterogeneous Bandits over Complex Networks",
                        "abstract": "Multi-armed bandits are extensively used to model sequential decision-making, making them ubiquitous in many real-life applications such as online recommender systems and wireless networking. We consider a multi-agent setting where each agent solves their own bandit instance endowed with a different set of arms. Their goal is to minimize their group regret while collaborating via some communication protocol over a given network. Previous literature on this problem only considered arm heterogeneity and networked agents separately. In this work, we introduce a setting that encompasses both features. For this novel setting, we first provide a rigorous regret analysis for a standard flooding protocol combined with the classic UCB policy. Then, to mitigate the issue of high communication costs incurred by flooding in complex networks, we propose a new protocol called Flooding with Absorption (FwA). We provide a theoretical analysis of the resulting regret bound and discuss the advantages of using FwA over flooding. Lastly, we experimentally verify on various scenarios, including dynamic networks, that FwA leads to significantly lower communication costs despite minimal regret performance loss compared to other network protocols."
                    },
                    {
                        "title": "Nearly Optimal Latent State Decoding in Block MDPs",
                        "abstract": "We investigate the problems of model estimation and reward-free learning in episodic Block MDPs. In these MDPs, the decision maker has access to rich observations or contexts generated from a small number of latent states. We are first interested in estimating the latent state decoding function (the mapping from the observations to latent states) based on data generated under a fixed behavior policy. We derive an information-theoretical lower bound on the error rate for estimating this function and present an algorithm approaching this fundamental limit. In turn, our algorithm also provides estimates of all the components of the MDP. We then study the problem of learning near-optimal policies in the reward-free framework. Based on our efficient model estimation algorithm, we show that we can infer a policy converging (as the number of collected samples grows large) to the optimal policy at the best possible rate. Interestingly, our analysis provides necessary and sufficient conditions under which exploiting the block structure yields improvements in the sample complexity for identifying near-optimal policies. When these conditions are met, the sample complexity in the minimax reward-free setting is improved by a multiplicative factor $n$, where $n$ is the number of possible contexts."
                    },
                    {
                        "title": "Fair Streaming Principal Component Analysis: Statistical and Algorithmic Viewpoint",
                        "abstract": "Fair Principal Component Analysis (PCA) is a problem setting where we aim to perform PCA while making the resulting representation fair in that the projected distributions, conditional on the sensitive attributes, match one another. However, existing approaches to fair PCA have two main problems: theoretically, there has been no statistical foundation of fair PCA in terms of learnability; practically, limited memory prevents us from using existing approaches, as they explicitly rely on full access to the entire data. On the theoretical side, we rigorously formulate fair PCA using a new notion called \\emph{probably approximately fair and optimal} (PAFO) learnability. On the practical side, motivated by recent advances in streaming algorithms for addressing memory limitation, we propose a new setting called \\emph{fair streaming PCA} along with a memory-efficient algorithm, fair noisy power method (FNPM). We then provide its {\\it statistical} guarantee in terms of PAFO-learnability, which is the first of its kind in fair PCA literature. Lastly, we verify the efficacy and memory efficiency of our algorithm on real-world datasets."
                    },
                    {
                        "title": "Gradient Descent with Polyak's Momentum Finds Flatter Minima via Large Catapults",
                        "abstract": "Although gradient descent with Polyak's momentum is widely used in modern machine and deep learning, a concrete understanding of its effects on the training trajectory remains elusive. In this work, we empirically show that for linear diagonal networks and nonlinear neural networks, momentum gradient descent with a large learning rate displays large catapults, driving the iterates towards much flatter minima than those found by gradient descent. We hypothesize that the large catapult is caused by momentum \"prolonging\" the self-stabilization effect (Damian et al., 2023). We provide theoretical and empirical support for our hypothesis in a simple toy example and empirical evidence supporting our hypothesis for linear diagonal networks."
                    },
                    {
                        "title": "Coarse-grained quantum state tomography with optimal POVM construction",
                        "abstract": "Constructing an integrated large-scale qubit system of realistic size requires addressing the challenge of physical crowding among qubits. This constraint poses an issue of coarse-grained (CG) measurement, wherein information from the multi-qubit system is collectively gathered. In this work, we introduce a novel approach to reconstruct the target density matrix from a comprehensive set of Positive Operator-Valued Measures (POVM) using a Parameterized Quantum Circuit (PQC) under the constraint of CG measurement. We improve the robustness and stability of CG quantum state tomography (QST) by optimizing the POVM set to achieve a generalized symmetric informationally complete (GSIC) POVM through maximization of the von Neumann entropy. This optimized construction of CG-POVMs is scalable to an N-qubit system. We further discuss a more efficient construction of N-qubit CG-QST without exponential increases in two-qubit gates or circuit depth per measurement. Our scheme offers a viable pathway towards a detector-efficient large-scale solid-state embedded qubit platform by reconstructing crucial quantum information from collective measurements."
                    },
                    {
                        "title": "Improved Regret Bounds of (Multinomial) Logistic Bandits via Regret-to-Confidence-Set Conversion",
                        "abstract": "Logistic bandit is a ubiquitous framework of modeling users' choices, e.g., click vs. no click for advertisement recommender system. We observe that the prior works overlook or neglect dependencies in $S \\geq \\lVert \\theta_\\star \\rVert_2$, where $\\theta_\\star \\in \\mathbb{R}^d$ is the unknown parameter vector, which is particularly problematic when $S$ is large, e.g., $S \\geq d$. In this work, we improve the dependency on $S$ via a novel approach called {\\it regret-to-confidence set conversion (R2CS)}, which allows us to construct a convex confidence set based on only the \\textit{existence} of an online learning algorithm with a regret guarantee. Using R2CS, we obtain a strict improvement in the regret bound w.r.t. $S$ in logistic bandits while retaining computational feasibility and the dependence on other factors such as $d$ and $T$. We apply our new confidence set to the regret analyses of logistic bandits with a new martingale concentration step that circumvents an additional factor of $S$. We then extend this analysis to multinomial logistic bandits and obtain similar improvements in the regret, showing the efficacy of R2CS. While we applied R2CS to the (multinomial) logistic model, R2CS is a generic approach for developing confidence sets that can be used for various models, which can be of independent interest."
                    },
                    {
                        "title": "Unsupervised CT Metal Artifact Learning using Attention-guided beta-CycleGAN",
                        "abstract": "Metal artifact reduction (MAR) is one of the most important research topics in computed tomography (CT). With the advance of deep learning technology for image reconstruction,various deep learning methods have been also suggested for metal artifact removal, among which supervised learning methods are most popular. However, matched non-metal and metal image pairs are difficult to obtain in real CT acquisition. Recently, a promising unsupervised learning for MAR was proposed using feature disentanglement, but the resulting network architecture is complication and difficult to handle large size clinical images. To address this, here we propose a much simpler and much effective unsupervised MAR method for CT. The proposed method is based on a novel beta-cycleGAN architecture derived from the optimal transport theory for appropriate feature space disentanglement. Another important contribution is to show that attention mechanism is the key element to effectively remove the metal artifacts. Specifically, by adding the convolutional block attention module (CBAM) layers with a proper disentanglement parameter, experimental results confirm that we can get more improved MAR that preserves the detailed texture of the original image."
                    },
                    {
                        "title": "Fast and Efficient MMD-based Fair PCA via Optimization over Stiefel Manifold",
                        "abstract": "This paper defines fair principal component analysis (PCA) as minimizing the maximum mean discrepancy (MMD) between dimensionality-reduced conditional distributions of different protected classes. The incorporation of MMD naturally leads to an exact and tractable mathematical formulation of fairness with good statistical properties. We formulate the problem of fair PCA subject to MMD constraints as a non-convex optimization over the Stiefel manifold and solve it using the Riemannian Exact Penalty Method with Smoothing (REPMS; Liu and Boumal, 2019). Importantly, we provide local optimality guarantees and explicitly show the theoretical effect of each hyperparameter in practical settings, extending previous results. Experimental comparisons based on synthetic and UCI datasets show that our approach outperforms prior work in explained variance, fairness, and runtime."
                    }
                ]
            },
            "2b6f928b-1e51-46f0-8335-faa7b73ffe83": {
                "pk": "2b6f928b-1e51-46f0-8335-faa7b73ffe83",
                "name": "Se-Young Yun",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "ea4ac517-ea05-4fc7-8d4a-533970e1cfbc": {
                "pk": "ea4ac517-ea05-4fc7-8d4a-533970e1cfbc",
                "name": "Kwang-Sung Jun",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2406.19824": {
        "paper_data": {
            "title": "Learning to Mitigate Externalities: the Coase Theorem with Hindsight Rationality",
            "url": "http://arxiv.org/abs/2406.19824v2",
            "arxiv_id": "2406.19824",
            "authors": [
                "Antoine Scheid",
                "Aymeric Capitaine",
                "Etienne Boursier",
                "Eric Moulines",
                "Michael I Jordan",
                "Alain Durmus"
            ],
            "abstract": "In economic theory, the concept of externality refers to any indirect effect resulting from an interaction between players that affects the social welfare. Most of the models within which externality has been studied assume that agents have perfect knowledge of their environment and preferences. This is a major hindrance to the practical implementation of many proposed solutions. To address this issue, we consider a two-player bandit setting where the actions of one of the players affect the other player and we extend the Coase theorem [Coase, 1960]. This result shows that the optimal approach for maximizing the social welfare in the presence of externality is to establish property rights, i.e., enable transfers and bargaining between the players. Our work removes the classical assumption that bargainers possess perfect knowledge of the underlying game. We first demonstrate that in the absence of property rights, the social welfare breaks down. We then design a policy for the players which allows them to learn a bargaining strategy which maximizes the total welfare, recovering the Coase theorem under uncertainty.",
            "introduction": "   1 Introduction  The concept of externality is used in economics to capture phenomena that impact the global welfare stemming from economic interactions without any compensation\u00a0(Buchanan and Stubblebine, 2006; Shah et\u00a0al., 2018). Externality is generally considered as market failure since they result in a loss of collective welfare. Given its practical importance (Dahlman, 1979; Greenfield et\u00a0al., 2009), mechanisms that characterize and mitigate externalities are central to modern economic thinking.   A first approach to tackle the adverse effects of externalities in modern economics was based on quotas and taxation (see, e.g., Pigou, 2017, and the references therein). However, Coase\u2019s theorem (Coase, 2013) shows that in the presence of well-defined property rights and low transaction costs, parties affected by externalities can privately negotiate efficient solutions, and recover a welfare efficient allocation through transfers and bargaining.   Throughout the paper, we use the following simple example to illustrate our results.    Example 1 (label=exa:cont).   Consider two firms 1 (upstream) and 2 (downstream) respectively producing quantities q1\u2a7e0subscript\ud835\udc5e10q_{1}\\geqslant 0italic_q start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT \u2a7e 0 and q2\u2a7e0subscript\ud835\udc5e20q_{2}\\geqslant 0italic_q start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT \u2a7e 0 of a good sold at a fixed price p>0\ud835\udc5d0p>0italic_p > 0. They incur strictly increasing and convex costs, captured by the differentiable cost functions c1:q1\u21a6c1\u2062(q1):subscript\ud835\udc501maps-tosubscript\ud835\udc5e1subscript\ud835\udc501subscript\ud835\udc5e1c_{1}:q_{1}\\mapsto c_{1}(q_{1})italic_c start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT : italic_q start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT \u21a6 italic_c start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_q start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ) and c2:q2\u21a6c2\u2062(q2):subscript\ud835\udc502maps-tosubscript\ud835\udc5e2subscript\ud835\udc502subscript\ud835\udc5e2c_{2}:q_{2}\\mapsto c_{2}(q_{2})italic_c start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT : italic_q start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT \u21a6 italic_c start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ( italic_q start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ), satisfying c1\u2062(0)=0subscript\ud835\udc50100c_{1}(0)=0italic_c start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( 0 ) = 0 and c2\u2062(0)=0subscript\ud835\udc50200c_{2}(0)=0italic_c start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ( 0 ) = 0. We assume that firm 1111 exerts on firm 2 a constant externality \u03b1>0\ud835\udefc0\\alpha>0italic_\u03b1 > 0 per unit produced. In other words, their profit functions (or utilities) are given by    \u03c01\u2062(q1)=p\u2062q1\u2212c1\u2062(q1)and\u03c02\u2062(q1,q2)=p\u2062q2\u2212c2\u2062(q2)\u2212\u03b1\u2062q1.formulae-sequencesubscript\ud835\udf0b1subscript\ud835\udc5e1\ud835\udc5dsubscript\ud835\udc5e1subscript\ud835\udc501subscript\ud835\udc5e1andsubscript\ud835\udf0b2subscript\ud835\udc5e1subscript\ud835\udc5e2\ud835\udc5dsubscript\ud835\udc5e2subscript\ud835\udc502subscript\ud835\udc5e2\ud835\udefcsubscript\ud835\udc5e1\\pi_{1}(q_{1})=pq_{1}-c_{1}(q_{1})\\quad\\text{and}\\quad\\pi_{2}(q_{1},q_{2})=pq_% {2}-c_{2}(q_{2})-\\alpha q_{1}\\;.italic_\u03c0 start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_q start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ) = italic_p italic_q start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT - italic_c start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ( italic_q start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ) and italic_\u03c0 start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ( italic_q start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , italic_q start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ) = italic_p italic_q start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT - italic_c start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ( italic_q start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT ) - italic_\u03b1 italic_q start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT .       One simple concrete illustration consists in an upstream firm emitting pollutants that reduce the downstream firm\u2019s production. If the upstream firm owns the property rights, it may receive a payment from the downstream one to reduce its output and thereby pollution. On the other hand, if the downstream firm owns the property right, it may require compensation from the upstream firm to allow its operation.   The Coase theorem demonstrates that in both cases with appropriate property rights, the resulting levels of production would be welfare efficient. The theorem is typically explained in textbooks under the assumption that the players have perfect knowledge of their own utility or profit function, as well as that of others. However, this assumption is unlikely to hold in real-world scenarios, where players have to learn about their own preferences and those of their competitors.   This example is, of course, a simplification of real world scenarios. For example, Abildtrup et\u00a0al. (2012) consider a more complex setting to model",
            "references": [
                {
                    "title": "Fair Allocation in Dynamic Mechanism Design",
                    "abstract": "We consider a dynamic mechanism design problem where an auctioneer sells an indivisible good to groups of buyers in every round, for a total of $T$ rounds. The auctioneer aims to maximize their discounted overall revenue while adhering to a fairness constraint that guarantees a minimum average allocation for each group. We begin by studying the static case ($T=1$) and establish that the optimal mechanism involves two types of subsidization: one that increases the overall probability of allocation to all buyers, and another that favors the groups which otherwise have a lower probability of winning the item. We then extend our results to the dynamic case by characterizing a set of recursive functions that determine the optimal allocation and payments in each round. Notably, our results establish that in the dynamic case, the seller, on the one hand, commits to a participation bonus to incentivize truth-telling, and on the other hand, charges an entry fee for every round. Moreover, the optimal allocation once more involves subsidization, which its extent depends on the difference in future utilities for both the seller and buyers when allocating the item to one group versus the others. Finally, we present an approximation scheme to solve the recursive equations and determine an approximately optimal and fair allocation efficiently."
                },
                {
                    "title": "Incentivized Learning in Principal-Agent Bandit Games",
                    "abstract": "This work considers a repeated principal-agent bandit game, where the principal can only interact with her environment through the agent. The principal and the agent have misaligned objectives and the choice of action is only left to the agent. However, the principal can influence the agent's decisions by offering incentives which add up to his rewards. The principal aims to iteratively learn an incentive policy to maximize her own total utility. This framework extends usual bandit problems and is motivated by several practical applications, such as healthcare or ecological taxation, where traditionally used mechanism design theories often overlook the learning aspect of the problem. We present nearly optimal (with respect to a horizon $T$) learning algorithms for the principal's regret in both multi-armed and linear contextual settings. Finally, we support our theoretical guarantees through numerical experiments."
                },
                {
                    "title": "Impact of Decentralized Learning on Player Utilities in Stackelberg Games",
                    "abstract": "When deployed in the world, a learning agent such as a recommender system or a chatbot often repeatedly interacts with another learning agent (such as a user) over time. In many such two-agent systems, each agent learns separately and the rewards of the two agents are not perfectly aligned. To better understand such cases, we examine the learning dynamics of the two-agent system and the implications for each agent's objective. We model these systems as Stackelberg games with decentralized learning and show that standard regret benchmarks (such as Stackelberg equilibrium payoffs) result in worst-case linear regret for at least one player. To better capture these systems, we construct a relaxed regret benchmark that is tolerant to small learning errors by agents. We show that standard learning algorithms fail to provide sublinear regret, and we develop algorithms to achieve near-optimal $O(T^{2/3})$ regret for both players with respect to these benchmarks. We further design relaxed environments under which faster learning ($O(\\sqrt{T})$) is possible. Altogether, our results take a step towards assessing how two-agent interactions in sequential and decentralized learning environments affect the utility of both agents."
                },
                {
                    "title": "Learning Optimal Tax Design in Nonatomic Congestion Games",
                    "abstract": "We study how to learn the optimal tax design to maximize the efficiency in nonatomic congestion games. It is known that self-interested behavior among the players can damage the system's efficiency. Tax mechanisms is a common method to alleviate this issue and induce socially optimal behavior. In this work, we take the initial step for learning the optimal tax that can minimize the social cost with \\emph{equilibrium feedback}, i.e., the tax designer can only observe the equilibrium state under the enforced tax. Existing algorithms are not applicable due to the exponentially large tax function space, nonexistence of the gradient, and nonconvexity of the objective. To tackle these challenges, our algorithm leverages several novel components: (1) piece-wise linear tax to approximate the optimal tax; (2) an extra linear term to guarantee a strongly convex potential function; (3) efficient subroutine to find the ``boundary'' tax. The algorithm can find an $\\epsilon$-optimal tax with $O(\\beta F^2/\\epsilon)$ sample complexity, where $\\beta$ is the smoothness of the cost function and $F$ is the number of facilities."
                },
                {
                    "title": "Contracting with a Learning Agent",
                    "abstract": "Many real-life contractual relations differ completely from the clean, static model at the heart of principal-agent theory. Typically, they involve repeated strategic interactions of the principal and agent, taking place under uncertainty and over time. While appealing in theory, players seldom use complex dynamic strategies in practice, often preferring to circumvent complexity and approach uncertainty through learning. We initiate the study of repeated contracts with a learning agent, focusing on agents who achieve no-regret outcomes. Optimizing against a no-regret agent is a known open problem in general games; we achieve an optimal solution to this problem for a canonical contract setting, in which the agent's choice among multiple actions leads to success/failure. The solution has a surprisingly simple structure: for some $\\alpha>0$, initially offer the agent a linear contract with scalar $\\alpha$, then switch to offering a linear contract with scalar $0$. This switch causes the agent to ``free-fall'' through their action space and during this time provides the principal with non-zero reward at zero cost. Despite apparent exploitation of the agent, this dynamic contract can leave \\emph{both} players better off compared to the best static contract. Our results generalize beyond success/failure, to arbitrary non-linear contracts which the principal rescales dynamically. Finally, we quantify the dependence of our results on knowledge of the time horizon, and are the first to address this consideration in the study of strategizing against learning agents."
                },
                {
                    "title": "Principal-Agent Reward Shaping in MDPs",
                    "abstract": "Principal-agent problems arise when one party acts on behalf of another, leading to conflicts of interest. The economic literature has extensively studied principal-agent problems, and recent work has extended this to more complex scenarios such as Markov Decision Processes (MDPs). In this paper, we further explore this line of research by investigating how reward shaping under budget constraints can improve the principal's utility. We study a two-player Stackelberg game where the principal and the agent have different reward functions, and the agent chooses an MDP policy for both players. The principal offers an additional reward to the agent, and the agent picks their policy selfishly to maximize their reward, which is the sum of the original and the offered reward. Our results establish the NP-hardness of the problem and offer polynomial approximation algorithms for two classes of instances: Stochastic trees and deterministic decision processes with a finite horizon."
                },
                {
                    "title": "Robust and Performance Incentivizing Algorithms for Multi-Armed Bandits with Strategic Agents",
                    "abstract": "We consider a variant of the stochastic multi-armed bandit problem. Specifically, the arms are strategic agents who can improve their rewards or absorb them. The utility of an agent increases if she is pulled more or absorbs more of her rewards but decreases if she spends more effort improving her rewards. Agents have heterogeneous properties, specifically having different means and able to improve their rewards up to different levels. Further, a non-empty subset of agents are ''honest'' and in the worst case always give their rewards without absorbing any part. The principal wishes to obtain a high revenue (cumulative reward) by designing a mechanism that incentives top level performance at equilibrium. At the same time, the principal wishes to be robust and obtain revenue at least at the level of the honest agent with the highest mean in case of non-equilibrium behaviour. We identify a class of MAB algorithms which we call performance incentivizing which satisfy a collection of properties and show that they lead to mechanisms that incentivize top level performance at equilibrium and are robust under any strategy profile. Interestingly, we show that UCB is an example of such a MAB algorithm. Further, in the case where the top performance level is unknown we show that combining second price auction ideas with performance incentivizing algorithms achieves performance at least at the second top level while also being robust."
                },
                {
                    "title": "Contract Design With Safety Inspections",
                    "abstract": "We study the role of regulatory inspections in a contract design problem in which a principal interacts separately with multiple agents. Each agent's hidden action includes a dimension that determines whether they undertake an extra costly step to adhere to safety protocols. The principal's objective is to use payments combined with a limited budget for random inspections to incentivize agents towards safety-compliant actions that maximize the principal's utility. We first focus on the single-agent setting with linear contracts and present an efficient algorithm that characterizes the optimal linear contract, which includes both payment and random inspection. We further investigate how the optimal contract changes as the inspection cost or the cost of adhering to safety protocols vary. Notably, we demonstrate that the agent's compensation increases if either of these costs escalates. However, while the probability of inspection decreases with rising inspection costs, it demonstrates nonmonotonic behavior as a function of the safety action costs. Lastly, we explore the multi-agent setting, where the principal's challenge is to determine the best distribution of inspection budgets among all agents. We propose an efficient approach based on dynamic programming to find an approximately optimal allocation of inspection budget across contracts. We also design a random sequential scheme to determine the inspector's assignments, ensuring each agent is inspected at most once and at the desired probability. Finally, we present a case study illustrating that a mere difference in the cost of inspection across various agents can drive the principal's decision to forego inspecting a significant fraction of them, concentrating its entire budget on those that are less costly to inspect."
                },
                {
                    "title": "Delegating Data Collection in Decentralized Machine Learning",
                    "abstract": "Motivated by the emergence of decentralized machine learning (ML) ecosystems, we study the delegation of data collection. Taking the field of contract theory as our starting point, we design optimal and near-optimal contracts that deal with two fundamental information asymmetries that arise in decentralized ML: uncertainty in the assessment of model quality and uncertainty regarding the optimal performance of any model. We show that a principal can cope with such asymmetry via simple linear contracts that achieve 1-1/e fraction of the optimal utility. To address the lack of a priori knowledge regarding the optimal performance, we give a convex program that can adaptively and efficiently compute the optimal contract. We also study linear contracts and derive the optimal utility in the more complex setting of multiple interactions."
                },
                {
                    "title": "Estimating and Incentivizing Imperfect-Knowledge Agents with Hidden Rewards",
                    "abstract": "In practice, incentive providers (i.e., principals) often cannot observe the reward realizations of incentivized agents, which is in contrast to many principal-agent models that have been previously studied. This information asymmetry challenges the principal to consistently estimate the agent's unknown rewards by solely watching the agent's decisions, which becomes even more challenging when the agent has to learn its own rewards. This complex setting is observed in various real-life scenarios ranging from renewable energy storage contracts to personalized healthcare incentives. Hence, it offers not only interesting theoretical questions but also wide practical relevance. This paper explores a repeated adverse selection game between a self-interested learning agent and a learning principal. The agent tackles a multi-armed bandit (MAB) problem to maximize their expected reward plus incentive. On top of the agent's learning, the principal trains a parallel algorithm and faces a trade-off between consistently estimating the agent's unknown rewards and maximizing their own utility by offering adaptive incentives to lead the agent. For a non-parametric model, we introduce an estimator whose only input is the history of principal's incentives and agent's choices. We unite this estimator with a proposed data-driven incentive policy within a MAB framework. Without restricting the type of the agent's algorithm, we prove finite-sample consistency of the estimator and a rigorous regret bound for the principal by considering the sequential externality imposed by the agent. Lastly, our theoretical results are reinforced by simulations justifying applicability of our framework to green energy aggregator contracts."
                },
                {
                    "title": "Actions Speak What You Want: Provably Sample-Efficient Reinforcement Learning of the Quantal Stackelberg Equilibrium from Strategic Feedbacks",
                    "abstract": "We study reinforcement learning (RL) for learning a Quantal Stackelberg Equilibrium (QSE) in an episodic Markov game with a leader-follower structure. In specific, at the outset of the game, the leader announces her policy to the follower and commits to it. The follower observes the leader's policy and, in turn, adopts a quantal response policy by solving an entropy-regularized policy optimization problem induced by leader's policy. The goal of the leader is to find her optimal policy, which yields the optimal expected total return, by interacting with the follower and learning from data. A key challenge of this problem is that the leader cannot observe the follower's reward, and needs to infer the follower's quantal response model from his actions against leader's policies. We propose sample-efficient algorithms for both the online and offline settings, in the context of function approximation. Our algorithms are based on (i) learning the quantal response model via maximum likelihood estimation and (ii) model-free or model-based RL for solving the leader's decision making problem, and we show that they achieve sublinear regret upper bounds. Moreover, we quantify the uncertainty of these estimators and leverage the uncertainty to implement optimistic and pessimistic algorithms for online and offline settings. Besides, when specialized to the linear and myopic setting, our algorithms are also computationally efficient. Our theoretical analysis features a novel performance-difference lemma which incorporates the error of quantal response model, which might be of independent interest."
                },
                {
                    "title": "Repeated Principal-Agent Games with Unobserved Agent Rewards and Perfect-Knowledge Agents",
                    "abstract": "Motivated by a number of real-world applications from domains like healthcare and sustainable transportation, in this paper we study a scenario of repeated principal-agent games within a multi-armed bandit (MAB) framework, where: the principal gives a different incentive for each bandit arm, the agent picks a bandit arm to maximize its own expected reward plus incentive, and the principal observes which arm is chosen and receives a reward (different than that of the agent) for the chosen arm. Designing policies for the principal is challenging because the principal cannot directly observe the reward that the agent receives for their chosen actions, and so the principal cannot directly learn the expected reward using existing estimation techniques. As a result, the problem of designing policies for this scenario, as well as similar ones, remains mostly unexplored. In this paper, we construct a policy that achieves a low regret (i.e., square-root regret up to a log factor) in this scenario for the case where the agent has perfect-knowledge about its own expected rewards for each bandit arm. We design our policy by first constructing an estimator for the agent's expected reward for each bandit arm. Since our estimator uses as data the sequence of incentives offered and subsequently chosen arms, the principal's estimation can be regarded as an analogy of online inverse optimization in MAB's. Next we construct a policy that we prove achieves a low regret by deriving finite-sample concentration bounds for our estimator. We conclude with numerical simulations demonstrating the applicability of our policy to real-life setting from collaborative transportation planning."
                },
                {
                    "title": "Learning to Incentivize Information Acquisition: Proper Scoring Rules Meet Principal-Agent Model",
                    "abstract": "We study the incentivized information acquisition problem, where a principal hires an agent to gather information on her behalf. Such a problem is modeled as a Stackelberg game between the principal and the agent, where the principal announces a scoring rule that specifies the payment, and then the agent then chooses an effort level that maximizes her own profit and reports the information. We study the online setting of such a problem from the principal's perspective, i.e., designing the optimal scoring rule by repeatedly interacting with the strategic agent. We design a provably sample efficient algorithm that tailors the UCB algorithm (Auer et al., 2002) to our model, which achieves a sublinear $T^{2/3}$-regret after $T$ iterations. Our algorithm features a delicate estimation procedure for the optimal profit of the principal, and a conservative correction scheme that ensures the desired agent's actions are incentivized. Furthermore, a key feature of our regret bound is that it is independent of the number of states of the environment."
                },
                {
                    "title": "A survey on multi-player bandits",
                    "abstract": "Due mostly to its application to cognitive radio networks, multiplayer bandits gained a lot of interest in the last decade. A considerable progress has been made on its theoretical aspect. However, the current algorithms are far from applicable and many obstacles remain between these theoretical results and a possible implementation of multiplayer bandits algorithms in real cognitive radio networks. This survey contextualizes and organizes the rich multiplayer bandits literature. In light of the existing works, some clear directions for future research appear. We believe that a further study of these different directions might lead to theoretical algorithms adapted to real-world situations."
                },
                {
                    "title": "The Sample Complexity of Online Contract Design",
                    "abstract": "Contract theory studies the interactions between a principal and an agent when the two parties transact in the presence of private information [Bolton and Dewatripont, 2004, Faure-Grimaud et al., 2001, Salani\u00e9, 2005]. The principal would like to achieve her desired outcomes by hiring agents to work for her. The agent wishes to make money by working for the principal. They develop agreements in the form of a contract, which specifies how much the principal would pay under the different possible outcomes of the agent's work."
                },
                {
                    "title": "Sequential Information Design: Learning to Persuade in the Dark",
                    "abstract": "We study a repeated information design problem faced by an informed sender who tries to influence the behavior of a self-interested receiver. We consider settings where the receiver faces a sequential decision making (SDM) problem. At each round, the sender observes the realizations of random events in the SDM problem. This begets the challenge of how to incrementally disclose such information to the receiver to persuade them to follow (desirable) action recommendations. We study the case in which the sender does not know random events probabilities, and, thus, they have to gradually learn them while persuading the receiver. We start by providing a non-trivial polytopal approximation of the set of sender's persuasive information structures. This is crucial to design efficient learning algorithms. Next, we prove a negative result: no learning algorithm can be persuasive. Thus, we relax persuasiveness requirements by focusing on algorithms that guarantee that the receiver's regret in following recommendations grows sub-linearly. In the full-feedback setting -- where the sender observes all random events realizations -- , we provide an algorithm with $\\tilde{O}(\\sqrt{T})$ regret for both the sender and the receiver. Instead, in the bandit-feedback setting -- where the sender only observes the realizations of random events actually occurring in the SDM problem -- , we design an algorithm that, given an $\\alpha \\in [1/2, 1]$ as input, ensures $\\tilde{O}({T^\\alpha})$ and $\\tilde{O}( T^{\\max \\{ \\alpha, 1-\\frac{\\alpha}{2} \\} })$ regrets, for the sender and the receiver respectively. This result is complemented by a lower bound showing that such a regrets trade-off is essentially tight."
                },
                {
                    "title": "Sequential Information Design: Markov Persuasion Process and Its Efficient Reinforcement Learning",
                    "abstract": "In today's economy, it becomes important for Internet platforms to consider the sequential information design problem to align its long term interest with incentives of the gig service providers (e.g., drivers, hosts). This paper proposes a novel model of sequential information design, namely the Markov persuasion processes (MPPs), in which a sender, with informational advantage, seeks to persuade a stream of myopic receivers to take actions that maximize the sender's cumulative utilities in a finite horizon Markovian environment with varying prior and utility functions. Planning in MPPs thus faces the unique challenge in finding a signaling policy that is simultaneously persuasive to the myopic receivers and inducing the optimal long-term cumulative utilities of the sender. Nevertheless, in the population level where the model is known, it turns out that we can efficiently determine the optimal (resp. \u03b5-optimal) policy with finite (resp. infinite) states and outcomes, through a modified formulation of the Bellman equation that additionally takes persuasiveness into consideration. Our main technical contribution is to study the MPP under the online reinforcement learning (RL) setting, where the goal is to learn the optimal signaling policy by interacting with with the underlying MPP, without the knowledge of the sender's utility functions, prior distributions, and the Markov transition kernels. For such a problem, we design a provably efficient no-regret learning algorithm, the Optimism-Pessimism Principle for Persuasion Process (OP4), which features a novel combination of both optimism and pessimism principles. In particular, we obtain optimistic estimates of the value functions to encourage exploration under the unknown environment, and additionally robustify the signaling policy with respect to the uncertainty of prior estimation to prevent receiver's detrimental equilibrium behavior. Our algorithm enjoys sample efficiency by achieving a sublinear \u221aT-regret upper bound. Furthermore, both our algorithm and theory can be applied to MPPs with large space of outcomes and states via function approximation, and we showcase such a success under the linear setting."
                },
                {
                    "title": "Exploration and Incentives in Reinforcement Learning",
                    "abstract": "How do you incentivize self-interested agents to explore when they prefer to exploit? We consider complex exploration problems, where each agent faces the same (but unknown) Markov decision process (MDP). In contrast with traditional formulations of reinforcement learning (RL), agents control the choice of policies, whereas an algorithm can only issue recommendations. However, the algorithm controls the flow of information, and can incentivize the agents to explore via information asymmetry. We design an algorithm which explores all reachable states in the MDP. We achieve provable guarantees similar to those for incentivizing exploration in static, stateless exploration problems studied previously. From the RL perspective, we design RL mechanisms, that is, RL algorithms that interact with self-interested agents and are compatible with their incentives. This is the first paper on RL mechanisms, that is, the first paper on any scenario that combines RL and incentives, to the best of our knowledge."
                },
                {
                    "title": "Contracts under Moral Hazard and Adverse Selection",
                    "abstract": "In the classical principal-agent problem, a principal must design a contract to incentivize an agent to perform an action on behalf of the principal. We study the classical principal-agent problem in a setting where the agent can be of one of several types (affecting the outcome of actions they might take). This combines the contract theory phenomena of \"moral hazard\" (incomplete information about actions) with that of \"adverse selection\" (incomplete information about types). We examine this problem through the computational lens. We show that in this setting it is APX-hard to compute either the profit-maximizing single contract or the profit-maximizing menu of contracts (as opposed to in the absence of types, where one can efficiently compute the optimal contract). We then show that the performance of the best linear contract scales especially well in the number of types: if agent has n available actions and T possible types, the best linear contract achieves an O(n log T) approximation of the best possible profit. Finally, we apply our framework to prove tight worst-case approximation bounds between a variety of benchmarks of mechanisms for the principal."
                },
                {
                    "title": "Bandit Algorithms",
                    "abstract": "sets of environments and policies respectively and ` : E \u00d7\u03a0\u2192 [0, 1] a bounded loss function. Given a policy \u03c0 let `(\u03c0) = (`(\u03bd1, \u03c0), . . . , `(\u03bdN , \u03c0)) be the loss vector resulting from policy \u03c0. Define S = {`(\u03c0) : \u03c0 \u2208 \u03a0} and \u03bb(S) = {x \u2208 cl(S) : y 6< x for all y \u2208 S} , where y 6< x is defined to mean it is not true that yi \u2264 xi for all i with strict inequality for at least one i. Prove that if \u03bb(S) \u2286 S and S is convex, then for each x \u2208 \u03bb(S) there exists a prior q \u2208 P(E) and policy \u03c0\u2217 such that `(\u03c0) = x and \u2211 \u03bd\u2208E q(\u03bd)`(\u03bd, \u03c0\u2217) = min \u03c0\u2208\u03a0 \u2211"
                },
                {
                    "title": "Social Learning in Multi Agent Multi Armed Bandits",
                    "abstract": "Motivated by emerging need of learning algorithms for large scale networked and decentralized systems, we introduce a distributed version of the classical stochastic Multi-Arm Bandit (MAB) problem. Our setting consists of a large number of agents n that collaboratively and simultaneously solve the same instance of K armed MAB to minimize the average cumulative regret over all agents. The agents can communicate and collaborate among each other only through a pairwise asynchronous gossip based protocol that exchange a limited number of bits. In our model, agents at each point decide on (i) which arm to play, (ii) whether to, and if so (iii) what and whom to communicate with. Agents in our model are decentralized, namely their actions only depend on their observed history in the past. We develop a novel algorithm in which agents, whenever they choose, communicate only arm-ids and not samples, with another agent chosen uniformly and independently at random. The per-agent regret scaling achieved by our algorithm is $\\BigO \u0142eft( \\frac\u0142ceil\\fracK n \\rceil+\u0142og(n) \u0394 \u0142og(T) + \\frac\u0142og^3(n) \u0142og \u0142og(n) \u0394^2 \\right) $. Furthermore, any agent in our algorithm communicates (arm-ids to an uniformly and independently chosen agent) only a total of \u0398(\u0142og(T))$ times over a time interval of T. We compare our results to two benchmarks - one where there is no communication among agents and one corresponding to complete interaction, where an agent has access to the entire system history of arms played and rewards obtained of all agents. We show both theoretically and empirically, that our algorithm experiences a significant reduction both in per-agent regret when compared to the case when agents do not collaborate and each agent is playing the standard MAB problem (where regret would scale linearly in K), and in communication complexity when compared to the full interaction setting which requires T communication attempts by an agent over T arm pulls. Our result thus demonstrates that even a minimal level of collaboration among the different agents enables a significant reduction in per-agent regret."
                },
                {
                    "title": "Contract Theory",
                    "abstract": "2 Hidden Information, Screening 4 2.1 The Simple Economics of Adverse Selection . . . . . . . . . . . . . . . . . . 4 2.1.1 First-Best Outcome: Perfect Price Discrimination . . . . . . . . . . . 6 2.1.1.1 Comparison of type-specific contracts with each other . . . . 6 2.1.1.2 In case it\u2019s been too long since you\u2019ve last done a maximization problem . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.1.1.3 Mas-Colell et al. [1995] . . . . . . . . . . . . . . . . . . . . . 8 2.1.1.4 Just what is the single crossing condition, exactly? . . . . . 10 2.1.2 Adverse Selection, Linear Pricing, and Simple Two-Part Tariffs . . . 10 2.1.2.1 Linear Pricing . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.1.2.2 Single Two-Part Tariff . . . . . . . . . . . . . . . . . . . . . 12 2.1.2.3 Did you forget the envelope theorem? . . . . . . . . . . . . . 12 2.1.3 Second-Best Outcome: Optimal Nonlinear Pricing Algorithm . . . . . 14 2.1.3.1 Comparison with the First Best . . . . . . . . . . . . . . . . 17 2.2 Application: Credit Rationing . . . . . . . . . . . . . . . . . . . . . . . . . . 19"
                },
                {
                    "title": "Introduction to Multi-Armed Bandits",
                    "abstract": "Multi-armed bandits a simple but very powerful framework for algorithms that make decisions over time under uncertainty. An enormous body of work has accumulated over the years, covered in several books and surveys. This book provides a more introductory, textbook-like treatment of the subject. Each chapter tackles a particular line of work, providing a self-contained, teachable technical introduction and a brief review of the further developments. The chapters are as follows: stochastic bandits, lower bounds; Bayesian bandits and Thompson Sampling; Lipschitz Bandits; full Feedback and adversarial costs; adversarial bandits; linear costs and semi-bandits; contextual Bandits; bandits and games; bandits with knapsacks; bandits and incentives."
                },
                {
                    "title": "SIC-MMAB: Synchronisation Involves Communication in Multiplayer Multi-Armed Bandits",
                    "abstract": "Motivated by cognitive radio networks, we consider the stochastic multiplayer multi-armed bandit problem, where several players pull arms simultaneously and collisions occur if one of them is pulled by several players at the same stage. We present a decentralized algorithm that achieves the same performance as a centralized one, contradicting the existing lower bounds for that problem. This is possible by \"hacking\" the standard model by constructing a communication protocol between players that deliberately enforces collisions, allowing them to share their information at a negligible cost. This motivates the introduction of a more appropriate dynamic setting without sensing, where similar communication protocols are no longer possible. However, we show that the logarithmic growth of the regret is still achievable for this model with a new algorithm."
                },
                {
                    "title": "Simple versus Optimal Contracts",
                    "abstract": "We consider the classic principal-agent model of contract theory, in which a principal designs an outcome-dependent compensation scheme to incentivize an agent to take a costly and unobservable action. When all of the model parameters---including the full distribution over principal rewards resulting from each agent action---are known to the designer, an optimal contract can in principle be computed by linear programming. In addition to their demanding informational requirements, however, such optimal contracts are often complex and unintuitive, and do not resemble contracts used in practice. This paper examines contract theory through the theoretical computer science lens, with the goal of developing novel theory to explain and justify the prevalence of relatively simple contracts, such as linear (pure commission) contracts. First, we consider the case where the principal knows only the first moment of each action's reward distribution, and we prove that linear contracts are guaranteed to be worst-case optimal, ranging over all reward distributions consistent with the given moments. Second, we study linear contracts from a worst-case approximation perspective, and prove several tight parameterized approximation bounds."
                },
                {
                    "title": "Overcoming public resistance to carbon taxes",
                    "abstract": "Carbon taxes represent a cost\u2010effective way to steer the economy toward a greener future. In the real world, their application has however been limited. In this paper, we address one of the main obstacles to carbon taxes: public opposition. We identify drivers of and barriers to public support, and, under the form of stylized facts, provide general lessons on the acceptability of carbon taxes. We derive our lessons from a growing literature, as well as from a combination of policy \u201cfailures\u201d and \u201csuccesses.\u201d Based on our stylized facts, we formulate a set of suggestions concerning the design of carbon taxes. We consider the use of trial periods, tax escalators, environmental earmarking, lump\u2010sum transfers, tax rebates, and advanced communication strategies, among others. This paper contributes to the policy debate about carbon taxes, hopefully leading to more success stories and fewer policy failures."
                },
                {
                    "title": "Bandit Learning with Positive Externalities",
                    "abstract": "In many platforms, user arrivals exhibit a self-reinforcing behavior: future user arrivals are likely to have preferences similar to users who were satisfied in the past. In other words, arrivals exhibit positive externalities. We study multiarmed bandit (MAB) problems with positive externalities. We show that the self-reinforcing preferences may lead standard benchmark algorithms such as UCB to exhibit linear regret. We develop a new algorithm, Balanced Exploration (BE), which explores arms carefully to avoid suboptimal convergence of arrivals before sufficient evidence is gathered. We also introduce an adaptive variant of BE which successively eliminates suboptimal arms. We analyze their asymptotic regret, and establish optimality by showing that no algorithm can perform better."
                },
                {
                    "title": "Learning to Bid Without Knowing your Value",
                    "abstract": "We address online learning in complex auction settings, such as sponsored search auctions, where the value of the bidder is unknown to her, evolving in an arbitrary manner and observed only if the bidder wins an allocation. We leverage the structure of the utility of the bidder and the partial feedback that bidders typically receive in auctions, in order to provide algorithms with regret rates against the best fixed bid in hindsight, that are exponentially faster in convergence in terms of dependence on the action space, than what would have been derived by applying a generic bandit algorithm and almost equivalent to what would have been achieved in the full information setting. Our results are enabled by analyzing a new online learning setting with outcome-based feedback, which generalizes learning with feedback graphs. We provide an online learning algorithm for this setting, of independent interest, with regret that grows only logarithmically with the number of actions and linearly only in the number of potential outcomes (the latter being very small in most auction settings). Last but not least, we show that our algorithm outperforms the bandit approach experimentally and that this performance is robust to dropping some of our theoretical assumptions or introducing noise in the feedback that the bidder receives."
                },
                {
                    "title": "Tax Design in the Alcohol Market",
                    "abstract": "Alcohol consumption is associated with costs to society due to its impact on crime and health. Tax can lead consumers to internalise these externalities. We study optimal corrective taxation in the alcohol market. We allow for the fact that the externality generating commodity (ethanol) is available in many di fferentiated products, over which consumers might have heterogeneous preferences, and that there may also be heterogeneity in marginal externalities across consumers. We show that, if there is correlation in preferences and marginal externalities, setting different tax rates across products can improve welfare relative to a single tax rate on ethanol. We estimate a model of demand in the UK alcohol market and numerically solve for the optimal tax rates. Moving to an optimal system that taxes alcohol types at different rates would close half of the welfare gap between the current UK system and the fi rst best. Watch IFS researcher, Kate Smith, talking about the design of alcohol taxes. A more recent version of this working paper is available here."
                },
                {
                    "title": "Dynamic Taxes for Polynomial Congestion Games",
                    "abstract": "We consider the efficiency of taxation in congestion games with polynomial latency functions along the line of research initiated by [Caragiannis et al., ACM Transactions on Algorithms, 2010] who focused on both pure and mixed Nash equilibria in games with affine latencies only. By exploiting the primal-dual method [Bilo, Proceedings of the 10th Workshop on Approximation and Online Algorithms, 2012], we obtain interesting upper bounds with respect to a variety of different solution concepts ranging from approximate pure Nash equilibria up to approximate coarse correlated equilibria, and including also approximate one-round walks starting from the empty state. Our findings show a high beneficial effect of taxation which increases more than linearly with the degree of the latency functions. In some cases, a tight relationship with some well-studied polynomials in Combinatorics and Number Theory, such as the Touchard and the Geometric polynomials, arises. In these cases we can also show matching lower bounds, albeit under mild assumptions; interestingly, our upper bounds are derived by exploiting the combinatorial definition of these polynomials, while our lower bounds are constructed by relying on their analytical characterization."
                },
                {
                    "title": "Bayesian Exploration: Incentivizing Exploration in Bayesian Games",
                    "abstract": "We consider a ubiquitous scenario in the Internet economy when individual decision-makers (henceforth, agents) both produce and consume information as they make strategic choices in an uncertain environment. This creates a three-way trade-off between exploration (trying out insufficiently explored alternatives to help others in the future), exploitation (making optimal decisions given the information discovered by other agents), and incentives of the agents (who are myopically interested in exploitation, while preferring the others to explore). We posit a principal who controls the flow of information from agents that came before to the ones that arrive later, and strives to coordinate the agents towards a socially optimal balance between exploration and exploitation, not using any monetary transfers. The goal is to design a recommendation policy for the principal which respects agents' incentives and minimizes a suitable notion of regret. We extend prior work in this direction to allow the agents to interact with one another in a shared environment: at each time step, multiple agents arrive to play a Bayesian game, receive recommendations, choose their actions, receive their payoffs, and then leave the game forever. The agents now face two sources of uncertainty: the actions of the other agents and the parameters of the uncertain game environment. Our main contribution is to show that the principal can achieve constant regret when the utilities are deterministic (where the constant depends on the prior distribution, but not on the time horizon), and logarithmic regret when the utilities are stochastic. As a key technical tool, we introduce the concept of explorable actions, the actions which some incentive-compatible policy can recommend with non-zero probability. We show how the principal can identify (and explore) all explorable actions, and use the revealed information to perform optimally. In particular, our results significantly improve over the prior work on the special case of a single agent per round, which relies on assumptions to guarantee that all actions are explorable. Interestingly, we do not require the principal's utility to be aligned with the cumulative utility of the agents; instead, the principal can optimize an arbitrary notion of per-round reward."
                },
                {
                    "title": "Batched Bandit Problems",
                    "abstract": "Motivated by practical applications, chiefly clinical trials, we study the regret achievable for stochastic bandits under the constraint that the employed policy must split trials into a small number of batches. We propose a simple policy, and show that a very small number of batches gives close to minimax optimal regret bounds. As a byproduct, we derive optimal policies with low switching cost for stochastic bandits."
                },
                {
                    "title": "Tax By Design: The Mirrlees Review",
                    "abstract": "Tax by Design identifies what makes a good tax system for an open developed economy in the 21st century and suggests how the UK tax system could be reformed to move in that direction. The recommendations stress the importance of neutrality and transparency in tax design. It draws on the expert evidence from the commissioned chapters and commentaries in Dimensions of Tax Design. It also acknowledges the growing importance of globalised markets and multinational corporations as well as the challenges created by changing population demographics, the growth of new technologies, and the broadened objectives of policy makers. The Commission's work was directed by: Timothy Besley Richard Blundell Malcolm Gammie James Poterba The Commission's editorial team: Stuart Adam Stephen Bond Robert Chote Paul Johnson Gareth Myles"
                },
                {
                    "title": "Dynamic Auctions: A Survey",
                    "abstract": "We survey the recent literature on designing auctions and mechanisms for dynamic settings. Two settings are considered: those with a dynamic population of agents or buyers whose private information remains fixed throughout time; and those with a fixed population of agents or buyers whose private information changes across time. Within each of these settings, we discuss both efficient (welfare-maximizing) and optimal (revenue-maximizing) mechanisms."
                },
                {
                    "title": "Externalities from Alcohol Consumption in the 2005 US National Alcohol Survey: Implications for Policy",
                    "abstract": "A subsample (n = 2,550) of the 2005 US National Alcohol Survey of adults was used to estimate prevalence and correlates of six externalities from alcohol abuse\u2013\u2013family problems, assaults, accompanying intoxicated driver, vehicular accident, financial problems and vandalized property\u2013\u2013all from another\u2019s drinking. On a lifetime basis, 60% reported externalities, with a lower 12-month rate (9%). Women reported more family/marital and financial impacts and men more assaults, accompanying drunk drivers, and accidents. Being unmarried, older, white and ever having monthly heavy drinking or alcohol problems was associated with more alcohol externalities. Publicizing external costs of drinking could elevate political will for effective alcohol controls."
                },
                {
                    "title": "Bayesian Persuasion",
                    "abstract": "When is it possible for one person to persuade another to change her action? We take a mechanism design approach to this question. Taking preferences and initial beliefs as given, we introduce the notion of a persuasion mechanism: a game between Sender and Receiver defined by an information structure and a message technology. We derive necessary and sufficient conditions for the existence of a persuasion mechanism that strictly benefits Sender. We characterize the optimal mechanism. Finally, we analyze several examples that illustrate the applicability of our results."
                },
                {
                    "title": "The Design of a Carbon Tax",
                    "abstract": "We consider the design of a tax on greenhouse gas emissions for a developed country such as the United States. We consider three sets of issues: the optimal tax base, issues relating to the rate (including the use of the revenues and rate changes over time) and trade. We show that a well-designed carbon tax can capture about 80% of U.S. emissions by taxing fewer than 3,000 taxpayers and up to almost 90% with a modest additional cost. We recommend full or partial delegation of rate setting authority to an agency to ensure that rates reflect new information about the costs of carbon emmissions and of abatement. Adjustments should be made to the income tax to ensure that a carbon tax is revenue neutral and distributionally neutral. Finally, we propose an origin-based system for trades with countries that have an adequate carbon tax. We suggest a system that imposes presumptive border tax adjustments with the ability of an individual firm to prove that a different rate should apply. The presumptive tax could be based either on average emissions for production of the item by the exporting country or by the importing country."
                },
                {
                    "title": "Taxes for linear atomic congestion games",
                    "abstract": "We study congestion games where players aim to access a set of resources. Each player has a set of possible strategies and each resource has a function associating the latency it incurs to the players using it. Players are non--cooperative and each wishes to follow a strategy that minimizes her own latency with no regard to the global optimum. Previous work has studied the impact of this selfish behavior on system performance. In this article, we study the question of how much the performance can be improved if players are forced to pay taxes for using resources. Our objective is to extend the original game so that selfish behavior does not deteriorate performance. We consider atomic congestion games with linear latency functions and present both negative and positive results. Our negative results show that optimal system performance cannot be achieved even in very simple games. On the positive side, we show that there are ways to assign taxes that can improve the performance of linear congestion games by forcing players to follow strategies where the total latency suffered is within a factor of 2 of the minimum possible; this result is shown to be tight. Furthermore, even in cases where in the absence of taxes the system behavior may be very poor, we show that the total disutility of players (latency plus taxes) is not much larger than the optimal total latency. Besides existential results, we show how to compute taxes in time polynomial in the size of the game by solving convex quadratic programs. Similar questions have been extensively studied in the model of non-atomic congestion games. To the best of our knowledge, this is the first study of the efficiency of taxes in atomic congestion games."
                },
                {
                    "title": "Using Confidence Bounds for Exploitation-Exploration Trade-offs",
                    "abstract": "We show how a standard tool from statistics --- namely confidence bounds --- can be used to elegantly deal with situations which exhibit an exploitation-exploration trade-off. Our technique for designing and analyzing algorithms for such situations is general and can be applied when an algorithm has to make exploitation-versus-exploration decisions based on uncertain information provided by a random process. We apply our technique to two models with such an exploitation-exploration trade-off. For the adversarial bandit problem with shifting our new algorithm suffers only O((ST)1/2) regret with high probability over T trials with S shifts. Such a regret bound was previously known only in expectation. The second model we consider is associative reinforcement learning with linear value functions. For this model our technique improves the regret from O(T3/4) to O(T1/2)."
                },
                {
                    "title": "The Economics of Contracts: A Primer",
                    "abstract": "The theory of contracts grew out of the failure of the general equilibrium model to account for the strategic interactions among agents that arise from informational asymmetries. This popular text, revised and updated throughout for the second edition, serves as a concise and rigorous introduction to the theory of contracts for graduate students and professional economists. The book presents the main models of the theory of contracts, particularly the basic models of adverse selection, signaling, and moral hazard. It emphasizes the methods used to analyze the models, but also includes brief introductions to many of the applications in different fields of economics. The goal is to give readers the tools to understand the basic models and create their own. For the second edition, major changes have been made to chapter 3, on examples and extensions for the adverse selection model, which now includes more thorough discussions of multiprincipals, collusion, and multidimensional adverse selection, and to chapter 5, on moral hazard, with the limited liability model, career concerns, and common agency added to its topics. Two chapters have been completely rewritten: chapter 7, on the theory of incomplete contracts, and chapter 8, on the empirical literature in the theory of contracts. An appendix presents concepts of noncooperative game theory to supplement chapters 4 and 6. Exercises follow chapters 2 through 5. Praise for the previous edition: \"The Economics of Contracts offers an excellent introduction to agency models. Written by one of the leading young researchers in contact theory, it is rigorous, clear, concise, and up-to-date. Researchers and students who want to learn about the economics of incentives will want to read this primer.\" -- Jean Tirole, Institut D'Economie Industrielle, Universite des Sciences Sociales, France \"Students will find this a very useful introduction to the ideas of contract theory. Salanie has managed to summarize a large amount of material in a relatively short number of pages in a highly accessible and readable manner.\" -- Oliver Hart, Professor of Economics, Harvard University"
                },
                {
                    "title": "The Cost of Coase",
                    "abstract": "THE publication of \"The Problem of Social Cost\" in 1960 by Ronald Coase brought together two powerful intellectual currents, namely, the economic theory of externalities and the common-law tradition concerning torts and nuisance.' The sea is fertile but rough where two ocean currents meet, and the same can be said of the disputes provoked by Coase. Coase developed his argument through a series of concrete examples, such as the rancher and the farmer, the railroad sparks and the corn crops, etc. He steadfastly refused to articulate the general truths underlying the examples; for example, the famous \"Coase Theorem\" is abstracted from the paper but not stated in it. After two decades of debate the generalizations underlying the examples are still disputed. In this paper I will not attempt the impossible task of offering the \"true\" or final interpretation of Coase's paper. Hermeneutics is left for others. Rather, I will state and prove those generalizations applicable to Coase's examples which follow from familiar assumptions of economic theory. It seems to me that these generalizations frequently conflict with Coase's examples in the sense that the immediate interpretation of his examples are inconsistent with the provable generalizations. The cost of Coase's paper is that readers are confused or misled about the generalizations which follow from familiar economic assumptions. In the first section of this paper I characterize the dispute by developing one of Coase's examples. The second section formulates some general propositions about correcting externalities through the use of liability law, taxes, and transferable property rights. The third section addresses the"
                },
                {
                    "title": "The Problem of Externality",
                    "abstract": "ON the modern research agenda externalities occupy a rather prominent position. The increasing complexity of modern technology and society seems to create yet additional unwanted side effects that require classification on a lengthening list of externalities. However, externalities are of interest not only as current policy issues but also from a more theoretical point of view. Using Pigou's terminology, we say that when an externality is present there is a divergence between private and social cost. We interpret this to mean that when all voluntary contractual arrangements have been entered into by market transactors, there still remain some interactions that ought to be internalized but which the market forces left to themselves cannot cope with. This is the basis, for example, for the assertion of Buchanan and Stubblebine that \"externality has been, and is, central to the neoclassical critique of market organization.\"' Without interference in the price mechanism, some transactions that would be beneficial are not carried out. Two conclusions follow: first, that since market forces by themselves are unable to eliminate the remaining inefficiencies, some government action is automatically necessitated; second, a conceptually feasible alternative to government action is that, through a suitable establishment of appropriate markets, economic agents can be made to take into account the side effects they generate.2 One may then inquire why market transactors are unable to make the emittor of an externality internalize the costs of his actions. The only reason why wealth-maximizing economic agents do not undertake these transactions must be that the cost of carrying out the actual transaction is greater"
                },
                {
                    "title": "The Problem of Social Cost",
                    "abstract": "This paper is concerned with those actions of business firms which have harmful effects on others. The standard example is that of a factory the smoke from which has harmful effects on those occupying neighbouring properties. The economic analysis of such a situation has usually proceeded in terms of a divergence between the private and social product of the factory, in which economists have largely followed the treatment of Pigou in The Economics of Welfare. The conclusions to which this kind of analysis seems to have led most economists is that it would be desirable to make the owner of the factory liable for the damage caused to those injured by the smoke, or alternatively, to place a tax on the factory owner varying with the amount of smoke produced and equivalent in money terms to the damage it would cause, or finally, to exclude the factory from residential districts (and presumably from other areas in which the emission of smoke would have harmful effects on others). It is my contention that the suggested courses of action are inappropriate, in that they lead to results which are not necessarily, or even usually, desirable."
                },
                {
                    "title": "Some aspects of the sequential design of experiments",
                    "abstract": "Until recently, statistical theory has been restricted to the design and analysis of sampling experiments in which the size and composition of the samples are completely determined before the experimentation begins. The reasons for this are partly historical, dating back to the time when the statistician was consulted, if at all, only after the experiment was over, and partly intrinsic in the mathematical difficulty of working with anything but a fixed number of independent random variables. A major advance now appears to be in the making with the creation of a theory of the sequential design of experiments, in which the size and composition of the samples are not fixed in advance but are functions of the observations themselves."
                },
                {
                    "title": "On Counterfactual Metrics for Social Welfare: Incentives, Ranking, and Information Asymmetry",
                    "abstract": "From the social sciences to machine learning, it is well documented that metrics do not always align with social welfare. In healthcare, Dranove et al. (2003) showed that publishing surgery mortality metrics actually harmed sicker patients by increasing provider selection behavior. Using a principal-agent model, we analyze the incentive misalign-ments that arise from such average treated outcome metrics, and show that the incentives driving treatment decisions would align with maximizing total patient welfare if the metrics (i) accounted for counterfactual untreated outcomes and (ii) considered total welfare instead of averaging over treated patients. Operationalizing this, we show how counterfactual metrics can be modified to behave reasonably in patient-facing ranking systems. Extending to realistic settings when providers observe more about patients than the regulatory agencies do, we bound the decay in performance by the degree of information asymmetry between principal and agent. In doing so, our model connects principal-agent information asymmetry with unobserved heterogeneity in causal inference."
                },
                {
                    "title": "Microeconomic Theory",
                    "abstract": "A Course in Microeconomic TheoryMicroeconomics of Banking, second editionA Textbook of Microeconomic TheoryIntermediate Microeconomic TheoryThe Entrepreneur in Microeconomic TheoryNeoclassical Microeconomic TheoryMicroeconomic TheoryNeoclassical Microeconomic TheoryMicroeconomic TheoryApplied Microeconomic TheoryAdvanced Microeconomic TheoryAdvanced Microeconomic TheoryMicroeconomicsMicroeconomic Theory: Basic Principles and ExtensionsMicroeconomic TheoryMicroeconomic TheoryMicroeconomic TheoryMicroeconomic Theory: Basic Principles and ExtensionsMicroeconomic Theory Old and NewMicroeconomic TheoryMicroeconomic TheoryMicroeconomic TheoryMicroeconomic Theory for the Social SciencesMicroeconomics: Theory And ApplicationsMicroeconomic TheoryMicroeconomic TheoryLectures on Microeconomic TheoryAdvanced Microeconomic TheoryTheory of MicroeconomicsMicroeconomic TheoryMicroeconomic TheoryModels in Microeconomic Theory ('She' Edition)Microeconomic TheoryMicroeconomic TheoryMicroeconomic Theory??????T&P Of Microeconomic Theory 3/E (Sos)Notes and Problems in Microeconomic TheoryMicroeconomic TheoryLecture Notes in Microeconomic Theory This book is intended to be a comprehensive and standard textbook for undergraduate students of Microeconomics. Apart from providing students with sufficient study material for examination purpose, it aims at making them understand economics. An effort has been made to explain abstract and complex microeconomic theories in a simple and lucid language without sacrificing analytical sophistication. The subject matter has been structured in a systematic manner without leaving gaps for the readers to fill in. Though the approach is non-mathematical, simple algebra has been used to give a concrete view of economic concepts and theories and to show the applicability of economic theories in decision making.This best-selling textbook covers all the topics in microeconomic theory covered by all students of economics. It combines the results of the authors' experience of teaching microeconomics at Harvard and has been fully classroom tested. Microeconomic Theory provides a balanced and in-depth analysis of the essentials of microeconomics, covering topics such as noncooperative game theory, information economics, mechanism design and general equilibrium under uncertainty. 'Self-sufficient' sections allow lecturers to 'mix and match' topics relevant to their courses.Discussion is clear, accessible and engaging, enabling the student to gradually acquire confidence as well as proficiency. Extensive exercises within each chapter help students to hone their skills, while the text's appendix of terms, fully cross-referenced throughout the previous five sections, offers an accessible guide to the subject matter's terminology."
                },
                {
                    "title": "Online Bayesian Persuasion",
                    "abstract": "In Bayesian persuasion, an informed sender has to design a signaling scheme that discloses the right amount of information so as to in\ufb02uence the behavior of a self-interested receiver. This kind of strategic interaction is ubiquitous in real-world economic scenarios. However, the seminal model by Kamenica and Gentzkow makes some stringent assumptions that limit its applicability in practice. One of the most limiting assumptions is, arguably, that the sender is required to know the receiver\u2019s utility function to compute an optimal signaling scheme. We relax this assumption through an online learning framework in which the sender repeatedly faces a receiver whose type is unknown and chosen adversarially at each round from a \ufb01nite set of possible types. We are interested in no-regret algorithms prescribing a signaling scheme at each round of the repeated interaction with performances close to that of a best-in-hindsight signaling scheme. First, we prove a hardness result on the per-round running time required to achieve no-\u03b1 -regret for any \u03b1 < 1 . Then, we provide algorithms for the full and partial feedback models with regret bounds sublinear in the number of rounds and polynomial in the size of the instance."
                },
                {
                    "title": "The Epoch-Greedy Algorithm for Multi-armed Bandits with Side Information",
                    "abstract": "We present Epoch-Greedy, an algorithm for multi-armed bandits with observable side information. Epoch-Greedy has the following properties: No knowledge of a time horizon $T$ is necessary. The regret incurred by Epoch-Greedy is controlled by a sample complexity bound for a hypothesis class. The regret scales as $O(T^{2/3} S^{1/3})$ or better (sometimes, much better). Here $S$ is the complexity term in a sample complexity bound for standard supervised learning."
                },
                {
                    "title": "Algorithmic Game Theory",
                    "abstract": "Online mechanisms extend the methods of mechanism design to dynamic environments with multiple agents and private information. Decisions must be made as information about types is revealed online and without knowledge of the future in the sense of online algorithms. We \ufb01rst consider single-valued preference domains and characterize the space of decision policies that can be truthfully implemented in a dominant strategy equilibrium. Working in a model-free environment we present truthful auctions for domains with ex-piring items and limited-supply items. Turning to a more general preference domain, and assuming the existence of a probabilistic model for agent types, we de\ufb01ne a dynamic Vickrey-Clarke-Groves mechanism that is e\ufb03cient and Bayes-Nash incentive compatible. We close with some thoughts about future research directions in this area."
                },
                {
                    "title": "The economics of taxation",
                    "abstract": "This authoritative two-volume set brings together the most important classic and contemporary papers on taxation and tax policy, written by the world\u00e2\u20ac\u2122s leading scholars and practitioners of taxation."
                },
                {
                    "title": "The Economics of Welfare",
                    "abstract": "The Economics of Welfare occupies a privileged position in economics. It contributed to the professionalization of economics, a goal aggressively and effectively pursued by Pigou's predecessor and teacher Alfred Marshall. The Economics of Welfare also may be credited with establishing welfare economics, by systematically analyzing market departures and their potential remedies. In writing The Economics of Welfare, Pigou built a bridge between the old and the new economics at Cambridge and in Britain. Much of the book remains relevant for contemporary economics. The list of his analyses that continues to play an important role in economics is impressive. Some of the more important include: public goods and externalities, welfare criteria, index number problems, price discrimination, the theory of the firm, the structure of relief programs for the poor, and public finance. Pigou's discussion of the institutional structure governing labor-market operations in his Wealth and Welfare prompted Schumpeter to call the work \"the greatest venture in labor economics ever undertaken by a man who was primarily a theorist.\" The Economics of Welfare established welfare economics as a field of study. The first part analyzes the relationship between the national dividend and economic and total welfare. Parts II and III link the size of the dividend to the allocation of resources in the economy and the institutional structure governing labor-market operations. Part IV explores the relationship between the national dividend and its distribution. In her new introduction, Nahid Aslanbeigui discusses the life of Pigou and the history of The Economics of Welfare. She also discusses Pigou's theories as expressed in this volume and some of the criticisms those theories have met as well as the impact of those criticisms. The Economics of Welfare is a classic that repays careful study."
                }
            ],
            "categories": [
                "cs.GT",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we model and mitigate the impact of externalities in economic interactions when parties involved have imperfect knowledge of their own and others' utility or profit functions?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it addresses a fundamental aspect of economic theory\u2014externalities\u2014which are often seen as a source of market failure. By developing models that account for imperfect knowledge, we can enhance our understanding of how externalities affect welfare and resource allocation. This research could lead to more effective policy recommendations and practical applications, such as improved negotiation frameworks and regulatory mechanisms that better align incentives among firms, ultimately advancing both theoretical and applied economics.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the complexity of accurately modeling externalities when parties do not have complete information about their own and others' profit functions. Naive approaches that assume perfect knowledge may fail to capture the dynamics of real-world interactions, leading to suboptimal solutions. Technical obstacles include the need for sophisticated algorithms to simulate learning processes among firms, as well as theoretical challenges in establishing equilibrium under conditions of uncertainty and incomplete information.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has largely focused on scenarios where players have perfect knowledge, which does not reflect real-world conditions. This gap has prevented the development of robust models that account for learning and adaptation in the presence of externalities. Existing solutions often overlook the complexities introduced by imperfect information and fail to provide practical frameworks for negotiation and compensation. Our approach aims to integrate learning mechanisms into the analysis of externalities, thereby improving upon prior work by offering a more realistic representation of economic interactions.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves developing a game-theoretic model that incorporates learning dynamics among firms facing externalities. We will use simulated datasets that reflect various scenarios of externality impacts and firm interactions. The key metrics for evaluation will include welfare outcomes, efficiency of negotiated solutions, and convergence rates of learning processes. We expect our results to demonstrate that incorporating learning leads to more efficient outcomes and better alignment of incentives, ultimately providing a framework for addressing externalities in real-world economic settings."
            }
        },
        "author_data": {
            "67698488-8e5f-46e5-9983-5f952e3362e8": {
                "pk": "67698488-8e5f-46e5-9983-5f952e3362e8",
                "name": "Antoine Scheid",
                "collaborators": [
                    "Etienne Boursier",
                    "Alain Durmus",
                    "Michael I. Jordan",
                    "Eric Moulines",
                    "Aymeric Capitaine",
                    "Pierre M\u00e9nard",
                    "Michal Valko",
                    "El-Mahdi El-Mhamdi",
                    "Daniil Tiapkin",
                    "El Mahdi El Mhamdi"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Collaborative Learning",
                    "Bandit Algorithms",
                    "Game Theory"
                ],
                "publications": [
                    {
                        "title": "Optimal Design for Reward Modeling in RLHF",
                        "abstract": "Reinforcement Learning from Human Feedback (RLHF) has become a popular approach to align language models (LMs) with human preferences. This method involves collecting a large dataset of human pairwise preferences across various text generations and using it to infer (implicitly or explicitly) a reward model. Numerous methods have been proposed to learn the reward model and align a LM with it. However, the costly process of collecting human preferences has received little attention and could benefit from theoretical insights. This paper addresses this issue and aims to formalize the reward training model in RLHF. We frame the selection of an effective dataset as a simple regret minimization task, using a linear contextual dueling bandit method. Given the potentially large number of arms, this approach is more coherent than the best-arm identification setting. We then propose an offline framework for solving this problem. Under appropriate assumptions - linearity of the reward model in the embedding space, and boundedness of the reward parameter - we derive bounds on the simple regret. Finally, we provide a lower bound that matches our upper bound up to constant and logarithmic terms. To our knowledge, this is the first theoretical contribution in this area to provide an offline approach as well as worst-case guarantees."
                    },
                    {
                        "title": "Unravelling in Collaborative Learning",
                        "abstract": "Collaborative learning offers a promising avenue for leveraging decentralized data. However, collaboration in groups of strategic learners is not a given. In this work, we consider strategic agents who wish to train a model together but have sampling distributions of different quality. The collaboration is organized by a benevolent aggregator who gathers samples so as to maximize total welfare, but is unaware of data quality. This setting allows us to shed light on the deleterious effect of adverse selection in collaborative learning. More precisely, we demonstrate that when data quality indices are private, the coalition may undergo a phenomenon known as unravelling, wherein it shrinks up to the point that it becomes empty or solely comprised of the worst agent. We show how this issue can be addressed without making use of external transfers, by proposing a novel method inspired by probabilistic verification. This approach makes the grand coalition a Nash equilibrium with high probability despite information asymmetry, thereby breaking unravelling."
                    },
                    {
                        "title": "Incentivized Learning in Principal-Agent Bandit Games",
                        "abstract": "This work considers a repeated principal-agent bandit game, where the principal can only interact with her environment through the agent. The principal and the agent have misaligned objectives and the choice of action is only left to the agent. However, the principal can influence the agent's decisions by offering incentives which add up to his rewards. The principal aims to iteratively learn an incentive policy to maximize her own total utility. This framework extends usual bandit problems and is motivated by several practical applications, such as healthcare or ecological taxation, where traditionally used mechanism design theories often overlook the learning aspect of the problem. We present nearly optimal (with respect to a horizon $T$) learning algorithms for the principal's regret in both multi-armed and linear contextual settings. Finally, we support our theoretical guarantees through numerical experiments."
                    }
                ]
            },
            "1e6a1edf-ba74-43f6-9c3b-827d41ab4385": {
                "pk": "1e6a1edf-ba74-43f6-9c3b-827d41ab4385",
                "name": "Aymeric Capitaine",
                "collaborators": [
                    "Etienne Boursier",
                    "Antoine Scheid",
                    "Eric Moulines",
                    "Michael I. Jordan",
                    "Alain Durmus",
                    "El-Mahdi El-Mhamdi",
                    "Daniil Tiapkin",
                    "El Mahdi El Mhamdi"
                ],
                "domain": [
                    "Collaborative Learning",
                    "Game Theory",
                    "Bandit Algorithms",
                    "Mechanism Design"
                ],
                "publications": [
                    {
                        "title": "Unravelling in Collaborative Learning",
                        "abstract": "Collaborative learning offers a promising avenue for leveraging decentralized data. However, collaboration in groups of strategic learners is not a given. In this work, we consider strategic agents who wish to train a model together but have sampling distributions of different quality. The collaboration is organized by a benevolent aggregator who gathers samples so as to maximize total welfare, but is unaware of data quality. This setting allows us to shed light on the deleterious effect of adverse selection in collaborative learning. More precisely, we demonstrate that when data quality indices are private, the coalition may undergo a phenomenon known as unravelling, wherein it shrinks up to the point that it becomes empty or solely comprised of the worst agent. We show how this issue can be addressed without making use of external transfers, by proposing a novel method inspired by probabilistic verification. This approach makes the grand coalition a Nash equilibrium with high probability despite information asymmetry, thereby breaking unravelling."
                    },
                    {
                        "title": "Incentivized Learning in Principal-Agent Bandit Games",
                        "abstract": "This work considers a repeated principal-agent bandit game, where the principal can only interact with her environment through the agent. The principal and the agent have misaligned objectives and the choice of action is only left to the agent. However, the principal can influence the agent's decisions by offering incentives which add up to his rewards. The principal aims to iteratively learn an incentive policy to maximize her own total utility. This framework extends usual bandit problems and is motivated by several practical applications, such as healthcare or ecological taxation, where traditionally used mechanism design theories often overlook the learning aspect of the problem. We present nearly optimal (with respect to a horizon $T$) learning algorithms for the principal's regret in both multi-armed and linear contextual settings. Finally, we support our theoretical guarantees through numerical experiments."
                    }
                ]
            },
            "72fc9174-16f2-4592-858c-c797ad40e7db": {
                "pk": "72fc9174-16f2-4592-858c-c797ad40e7db",
                "name": "Etienne Boursier",
                "collaborators": [
                    "Vianney Perchet",
                    "Nicolas Flammarion",
                    "Marco Scarsini",
                    "Mikhail Konobeev",
                    "Emilie Kaufmann",
                    "Abbas Mehrabian",
                    "Loucas Pillaud-Vivien",
                    "Flore Sentenac",
                    "Tristan Garrec",
                    "O\u011fuz Kaan Y\u00fcksel"
                ],
                "domain": [
                    "Multi-armed Bandits",
                    "Neural Networks",
                    "Online Learning",
                    "Meta-learning"
                ],
                "publications": [
                    {
                        "title": "A survey on multi-player bandits",
                        "abstract": "Due mostly to its application to cognitive radio networks, multiplayer bandits gained a lot of interest in the last decade. A considerable progress has been made on its theoretical aspect. However, the current algorithms are far from applicable and many obstacles remain between these theoretical results and a possible implementation of multiplayer bandits algorithms in real cognitive radio networks. This survey contextualizes and organizes the rich multiplayer bandits literature. In light of the existing works, some clear directions for future research appear. We believe that a further study of these different directions might lead to theoretical algorithms adapted to real-world situations."
                    },
                    {
                        "title": "SIC-MMAB: Synchronisation Involves Communication in Multiplayer Multi-Armed Bandits",
                        "abstract": "Motivated by cognitive radio networks, we consider the stochastic multiplayer multi-armed bandit problem, where several players pull arms simultaneously and collisions occur if one of them is pulled by several players at the same stage. We present a decentralized algorithm that achieves the same performance as a centralized one, contradicting the existing lower bounds for that problem. This is possible by \"hacking\" the standard model by constructing a communication protocol between players that deliberately enforces collisions, allowing them to share their information at a negligible cost. This motivates the introduction of a more appropriate dynamic setting without sensing, where similar communication protocols are no longer possible. However, we show that the logarithmic growth of the regret is still achievable for this model with a new algorithm."
                    },
                    {
                        "title": "Utility/Privacy Trade-off through the lens of Optimal Transport",
                        "abstract": "Strategic information is valuable either by remaining private (for instance if it is sensitive) or, on the other hand, by being used publicly to increase some utility. These two objectives are antagonistic and leaking this information might be more rewarding than concealing it. Unlike classical solutions that focus on the first point, we consider instead agents that optimize a natural trade-off between both objectives. We formalize this as an optimization problem where the objective mapping is regularized by the amount of information revealed to the adversary (measured as a divergence between the prior and posterior on the private knowledge). Quite surprisingly, when combined with the entropic regularization, the Sinkhorn loss naturally emerges in the optimization objective, making it efficiently solvable. We apply these techniques to preserve some privacy in online repeated auctions."
                    },
                    {
                        "title": "Penalising the biases in norm regularisation enforces sparsity",
                        "abstract": "Controlling the parameters' norm often yields good generalisation when training neural networks. Beyond simple intuitions, the relation between regularising parameters' norm and obtained estimators remains theoretically misunderstood. For one hidden ReLU layer networks with unidimensional data, this work shows the parameters' norm required to represent a function is given by the total variation of its second derivative, weighted by a $\\sqrt{1+x^2}$ factor. Notably, this weighting factor disappears when the norm of bias terms is not regularised. The presence of this additional weighting factor is of utmost significance as it is shown to enforce the uniqueness and sparsity (in the number of kinks) of the minimal norm interpolator. Conversely, omitting the bias' norm allows for non-sparse solutions. Penalising the bias terms in the regularisation, either explicitly or implicitly, thus leads to sparse estimators."
                    },
                    {
                        "title": "Early alignment in two-layer networks training is a two-edged sword",
                        "abstract": "Training neural networks with first order optimisation methods is at the core of the empirical success of deep learning. The scale of initialisation is a crucial factor, as small initialisations are generally associated to a feature learning regime, for which gradient descent is implicitly biased towards simple solutions. This work provides a general and quantitative description of the early alignment phase, originally introduced by Maennel et al. (2018) . For small initialisation and one hidden ReLU layer networks, the early stage of the training dynamics leads to an alignment of the neurons towards key directions. This alignment induces a sparse representation of the network, which is directly related to the implicit bias of gradient flow at convergence. This sparsity inducing alignment however comes at the expense of difficulties in minimising the training objective: we also provide a simple data example for which overparameterised networks fail to converge towards global minima and only converge to a spurious stationary point instead."
                    },
                    {
                        "title": "Trace norm regularization for multi-task learning with scarce data",
                        "abstract": "Multi-task learning leverages structural similarities between multiple tasks to learn despite very few samples. Motivated by the recent success of neural networks applied to data-scarce tasks, we consider a linear low-dimensional shared representation model. Despite an extensive literature, existing theoretical results either guarantee weak estimation rates or require a large number of samples per task. This work provides the first estimation error bound for the trace norm regularized estimator when the number of samples per task is small. The advantages of trace norm regularization for learning data-scarce tasks extend to meta-learning and are confirmed empirically on synthetic datasets."
                    },
                    {
                        "title": "Selfish Robustness and Equilibria in Multi-Player Bandits",
                        "abstract": "Motivated by cognitive radios, stochastic multi-player multi-armed bandits gained a lot of interest recently. In this class of problems, several players simultaneously pull arms and encounter a collision - with 0 reward - if some of them pull the same arm at the same time. While the cooperative case where players maximize the collective reward (obediently following some fixed protocol) has been mostly considered, robustness to malicious players is a crucial and challenging concern. Existing approaches consider only the case of adversarial jammers whose objective is to blindly minimize the collective reward. We shall consider instead the more natural class of selfish players whose incentives are to maximize their individual rewards, potentially at the expense of the social welfare. We provide the first algorithm robust to selfish players (a.k.a. Nash equilibrium) with a logarithmic regret, when the arm performance is observed. When collisions are also observed, Grim Trigger type of strategies enable some implicit communication-based algorithms and we construct robust algorithms in two different settings: the homogeneous (with a regret comparable to the centralized optimal one) and heterogeneous cases (for an adapted and relevant notion of regret). We also provide impossibility results when only the reward is observed or when arm means vary arbitrarily among players."
                    },
                    {
                        "title": "Simplicity bias and optimization threshold in two-layer ReLU networks",
                        "abstract": "Understanding generalization of overparametrized neural networks remains a fundamental challenge in machine learning. Most of the literature mostly studies generalization from an interpolation point of view, taking convergence of parameters towards a global minimum of the training loss for granted. While overparametrized architectures indeed interpolated the data for typical classification tasks, this interpolation paradigm does not seem valid anymore for more complex tasks such as in-context learning or diffusion. Instead for such tasks, it has been empirically observed that the trained models goes from global minima to spurious local minima of the training loss as the number of training samples becomes larger than some level we call optimization threshold. While the former yields a poor generalization to the true population loss, the latter was observed to actually correspond to the minimiser of this true loss. This paper explores theoretically this phenomenon in the context of two-layer ReLU networks. We demonstrate that, despite overparametrization, networks often converge toward simpler solutions rather than interpolating the training data, which can lead to a drastic improvement on the test loss with respect to interpolating solutions. Our analysis relies on the so called early alignment phase, during which neurons align towards specific directions. This directional alignment, which occurs in the early stage of training, leads to a simplicity bias, wherein the network approximates the ground truth model without converging to the global minimum of the training loss. Our results suggest that this bias, resulting in an optimization threshold from which interpolation is not reached anymore, is beneficial and enhances the generalization of trained models."
                    },
                    {
                        "title": "A Practical Algorithm for Multiplayer Bandits when Arm Means Vary Among Players",
                        "abstract": "We study a multiplayer stochastic multi-armed bandit problem in which players cannot communicate, and if two or more players pull the same arm, a collision occurs and the involved players receive zero reward. We consider the challenging heterogeneous setting, in which different arms may have different means for different players, and propose a new and efficient algorithm that combines the idea of leveraging forced collisions for implicit communication and that of performing matching eliminations. We present a finite-time analysis of our algorithm, giving the first sublinear minimax regret bound for this problem, and prove that if the optimal assignment of players to arms is unique, our algorithm attains the optimal $O(\\ln(T))$ regret, solving an open question raised at NeurIPS 2018."
                    },
                    {
                        "title": "Social Learning in Non-Stationary Environments",
                        "abstract": "Potential buyers of a product or service, before making their decisions, tend to read reviews written by previous consumers. We consider Bayesian consumers with heterogeneous preferences, who sequentially decide whether to buy an item of unknown quality, based on previous buyers' reviews. The quality is multi-dimensional and may occasionally vary over time; the reviews are also multi-dimensional. In the simple uni-dimensional and static setting, beliefs about the quality are known to converge to its true value. Our paper extends this result in several ways. First, a multi-dimensional quality is considered, second, rates of convergence are provided, third, a dynamical Markovian model with varying quality is studied. In this dynamical setting the cost of learning is shown to be small."
                    },
                    {
                        "title": "Gradient flow dynamics of shallow ReLU networks for square loss and orthogonal inputs",
                        "abstract": "The training of neural networks by gradient descent methods is a cornerstone of the deep learning revolution. Yet, despite some recent progress, a complete theory explaining its success is still missing. This article presents, for orthogonal input vectors, a precise description of the gradient flow dynamics of training one-hidden layer ReLU neural networks for the mean squared error at small initialisation. In this setting, despite non-convexity, we show that the gradient flow converges to zero loss and characterise its implicit bias towards minimum variation norm. Furthermore, some interesting phenomena are highlighted: a quantitative description of the initial alignment phenomenon and a proof that the process follows a specific saddle to saddle dynamics."
                    },
                    {
                        "title": "Decentralized Learning in Online Queuing Systems",
                        "abstract": "Motivated by packet routing in computer networks, online queuing systems are composed of queues receiving packets at different rates. Repeatedly, they send packets to servers, each of them treating only at most one packet at a time. In the centralized case, the number of accumulated packets remains bounded (i.e., the system is \\textit{stable}) as long as the ratio between service rates and arrival rates is larger than $1$. In the decentralized case, individual no-regret strategies ensures stability when this ratio is larger than $2$. Yet, myopically minimizing regret disregards the long term effects due to the carryover of packets to further rounds. On the other hand, minimizing long term costs leads to stable Nash equilibria as soon as the ratio exceeds $\\frac{e}{e-1}$. Stability with decentralized learning strategies with a ratio below $2$ was a major remaining question. We first argue that for ratios up to $2$, cooperation is required for stability of learning strategies, as selfish minimization of policy regret, a \\textit{patient} notion of regret, might indeed still be unstable in this case. We therefore consider cooperative queues and propose the first learning decentralized algorithm guaranteeing stability of the system as long as the ratio of rates is larger than $1$, thus reaching performances comparable to centralized strategies."
                    },
                    {
                        "title": "Making the most of your day: online learning for optimal allocation of time",
                        "abstract": "We study online learning for optimal allocation when the resource to be allocated is time. %Examples of possible applications include job scheduling for a computing server, a driver filling a day with rides, a landlord renting an estate, etc. An agent receives task proposals sequentially according to a Poisson process and can either accept or reject a proposed task. If she accepts the proposal, she is busy for the duration of the task and obtains a reward that depends on the task duration. If she rejects it, she remains on hold until a new task proposal arrives. We study the regret incurred by the agent, first when she knows her reward function but does not know the distribution of the task duration, and then when she does not know her reward function, either. This natural setting bears similarities with contextual (one-armed) bandits, but with the crucial difference that the normalized reward associated to a context depends on the whole distribution of contexts."
                    },
                    {
                        "title": "First-order ANIL provably learns representations despite overparametrization",
                        "abstract": "Due to its empirical success in few-shot classification and reinforcement learning, meta-learning has recently received significant interest. Meta-learning methods leverage data from previous tasks to learn a new task in a sample-efficient manner. In particular, model-agnostic methods look for initialization points from which gradient descent quickly adapts to any new task. Although it has been empirically suggested that such methods perform well by learning shared representations during pretraining, there is limited theoretical evidence of such behavior. More importantly, it has not been shown that these methods still learn a shared structure, despite architectural misspecifications. In this direction, this work shows, in the limit of an infinite number of tasks, that first-order ANIL with a linear two-layer network architecture successfully learns linear shared representations. This result even holds with overparametrization; having a width larger than the dimension of the shared representations results in an asymptotically low-rank solution. The learned solution then yields a good adaptation performance on any new task after a single gradient step. Overall, this illustrates how well model-agnostic methods such as first-order ANIL can learn shared representations."
                    },
                    {
                        "title": "Statistical Efficiency of Thompson Sampling for Combinatorial Semi-Bandits",
                        "abstract": "We investigate stochastic combinatorial multi-armed bandit with semi-bandit feedback (CMAB). In CMAB, the question of the existence of an efficient policy with an optimal asymptotic regret (up to a factor poly-logarithmic with the action size) is still open for many families of distributions, including mutually independent outcomes, and more generally the multivariate sub-Gaussian family. We propose to answer the above question for these two families by analyzing variants of the Combinatorial Thompson Sampling policy (CTS). For mutually independent outcomes in $[0,1]$, we propose a tight analysis of CTS using Beta priors. We then look at the more general setting of multivariate sub-Gaussian outcomes and propose a tight analysis of CTS using Gaussian priors. This last result gives us an alternative to the Efficient Sampling for Combinatorial Bandit policy (ESCB), which, although optimal, is not computationally efficient."
                    },
                    {
                        "title": "Constant or logarithmic regret in asynchronous multiplayer bandits",
                        "abstract": "Multiplayer bandits have recently been extensively studied because of their application to cognitive radio networks.   While the literature mostly considers synchronous players, radio networks (e.g. for IoT) tend to have asynchronous devices. This motivates the harder, asynchronous multiplayer bandits problem, which was first tackled with an explore-then-commit (ETC) algorithm (see Dakdouk, 2022), with a regret upper-bound in $\\mathcal{O}(T^{\\frac{2}{3}})$. Before even considering decentralization, understanding the centralized case was still a challenge as it was unknown whether getting a regret smaller than $\\Omega(T^{\\frac{2}{3}})$ was possible.   We answer positively this question, as a natural extension of UCB exhibits a $\\mathcal{O}(\\sqrt{T\\log(T)})$ minimax regret.   More importantly, we introduce Cautious Greedy, a centralized algorithm that yields constant instance-dependent regret if the optimal policy assigns at least one player on each arm (a situation that is proved to occur when arm means are close enough). Otherwise, its regret increases as the sum of $\\log(T)$ over some sub-optimality gaps. We provide lower bounds showing that Cautious Greedy is optimal in the data-dependent terms.   Therefore, we set up a strong baseline for asynchronous multiplayer bandits and suggest that learning the optimal policy in this problem might be easier than thought, at least with centralization."
                    },
                    {
                        "title": "Plato: Approximate Analytics over Compressed Time Series with Tight Deterministic Error Guarantees",
                        "abstract": "Plato provides fast approximate analytics on time series, by precomputing and storing compressed time series. Plato's key novelty is the delivery of tight deterministic error guarantees for time series analytics. Plato evaluates any time series expression composed by the linear algebra operators over vectors, along with arithmetic operators. This large scope of possible expressions includes common use cases such as correlation and cross-correlation expressions. Each time series is segmented either by fixed-length segmentation or by (a usually more effective) variable-length segmentation. Each segment is compressed by an estimation/compression function that approximates the actual values and is coming from a user-chosen function family, as taught by many prior works. The novelty is that Plato associates to each segment 1 to 3 (depending on the case) precomputed error measures and, using them, Plato computes tight deterministic error guarantees for analytics over the compressions. Importantly, some compression families lead to much better deterministic error guarantees. This work identifies two broad estimation function family groups (Vector Space (VS) and Linear Scalable Family (LSF)), which lead to theoretically and practically high-quality guarantees, even for expressions (eg correlation) that combine multiple time series that have been independently compressed and may, thus, use misaligned segmentations. The theoretical aspect of \"high quality\" is crisply captured by the Amplitude Independence (AI) property: An AI guarantee does not depend on the amplitude of the involved time series, even when we combine multiple time series. The experiments on four real-life datasets showed that when the novel AI guarantees were applicable, the approximate query results were certified to be very close (typically 1%) to the true results."
                    },
                    {
                        "title": "Approximate information maximization for bandit games",
                        "abstract": "Entropy maximization and free energy minimization are general physical principles for modeling the dynamics of various physical systems. Notable examples include modeling decision-making within the brain using the free-energy principle, optimizing the accuracy-complexity trade-off when accessing hidden variables with the information bottleneck principle (Tishby et al., 2000), and navigation in random environments using information maximization (Vergassola et al., 2007). Built on this principle, we propose a new class of bandit algorithms that maximize an approximation to the information of a key variable within the system. To this end, we develop an approximated analytical physics-based representation of an entropy to forecast the information gain of each action and greedily choose the one with the largest information gain. This method yields strong performances in classical bandit settings. Motivated by its empirical success, we prove its asymptotic optimality for the two-armed bandit problem with Gaussian rewards. Owing to its ability to encompass the system's properties in a global physical functional, this approach can be efficiently adapted to more complex bandit settings, calling for further investigation of information maximization approaches for multi-armed bandit problems."
                    }
                ]
            },
            "e1cb6dbd-822a-44ad-b09b-d93135f29e53": {
                "pk": "e1cb6dbd-822a-44ad-b09b-d93135f29e53",
                "name": "Eric Moulines",
                "collaborators": [
                    "Randal Douc",
                    "Philippe Soulier",
                    "Paul Doukhan",
                    "Olivier Capp\u00e9",
                    "Alain Durmus",
                    "Ya'Acov Ritov",
                    "Elisabeth Gassiat",
                    "Benoit Landelle",
                    "Valderio Reisen",
                    "Glaura Franco"
                ],
                "domain": [
                    "Hidden Markov Models",
                    "Time Series Analysis",
                    "Stochastic Processes",
                    "Bayesian Inference"
                ],
                "publications": [
                    {
                        "title": "Forgetting of the initial condition for the filter in general state-space hidden Markov chain: a coupling approach",
                        "abstract": "We give simple conditions that ensure exponential forgetting of the initial conditions of the filter for general state-space hidden Markov chain. The proofs are based on the coupling argument applied to the posterior Markov kernels. These results are useful both for filtering hidden Markov models using approximation methods (e.g., particle filters) and for proving asymptotic properties of estimators. The results are general enough to cover models like the Gaussian state space model, without using the special structure that permits the application of the Kalman filter."
                    },
                    {
                        "title": "Forgetting of the initial distribution for non-ergodic Hidden Markov Chains",
                        "abstract": "In this paper, the forgetting of the initial distribution for a non-ergodic Hidden Markov Models (HMM) is studied. A new set of conditions is proposed to establish the forgetting property of the filter, which significantly extends all the existing results. Both a pathwise-type convergence of the total variation distance of the filter started from two different initial distributions, and a convergence in expectation are considered. The results are illustrated using generic models of non-ergodic HMM and extend all the results known so far."
                    },
                    {
                        "title": "Log-average periodogram estimator of the memory parameter",
                        "abstract": "This paper introduces a semiparametric regression estimator of the memory parameter for long-memory time series process. It is based on the regression in a neighborhood of the zero-frequency of the periodogram averaged over epochs. The proposed estimator is theoretically justified and empirical Monte Carlo investigation gives evidence that the method is very promising to estimate the long-memory parameter."
                    },
                    {
                        "title": "Computable Convergence Rates for Subgeometrically Ergodic Markov Chains",
                        "abstract": "In this paper, we give quantitative bounds on the $f$-total variation distance from convergence of an Harris recurrent Markov chain on an arbitrary under drift and minorisation conditions implying ergodicity at a sub-geometric rate. These bounds are then specialized to the stochastically monotone case, covering the case where there is no minimal reachable element. The results are illustrated on two examples from queueing theory and Markov Chain Monte Carlo."
                    },
                    {
                        "title": "Testing for Homogeneity with Kernel Fisher Discriminant Analysis",
                        "abstract": "We propose to investigate test statistics for testing homogeneity in reproducing kernel Hilbert spaces. Asymptotic null distributions under null hypothesis are derived, and consistency against fixed and local alternatives is assessed. Finally, experimental evidence of the performance of the proposed approach on both artificial data and a speaker verification task is provided."
                    },
                    {
                        "title": "Estimation of the autocovariance function with missing observations",
                        "abstract": "We propose a novel estimator of the autocorrelation function in presence of missing observations. We establish the consistency, the asymptotic normality, and we derive deviation bounds for various classes of weakly dependent stationary time series, including causal or non causal models. In addition, we introduce a modified version periodogram defined from these autocorrelation estimators and derive asymptotic distribution of linear functionals of this estimator."
                    },
                    {
                        "title": "Ergodicity of observation-driven time series models and consistency of the maximum likelihood estimator",
                        "abstract": "This paper deals with a general class of observation-driven time series models with a special focus on time series of counts. We provide conditions under which there exist strict-sense stationary and ergodic versions of such processes. The consistency of the maximum likelihood estimators is then derived for well- specified and misspecified models."
                    },
                    {
                        "title": "On the spectral density of the wavelet coefficients of long memory time series with application to the log-regression estimation of the memory parameter",
                        "abstract": "In the recent years, methods to estimate the memory parameter using wavelet analysis have gained popularity in many areas of science. Despite its widespread use, a rigorous semi-parametric asymptotic theory, comparable to the one developed for Fourier methods, is still missing. In this contribution, we adapt the classical semi-parametric framework introduced by Robinson and his co-authors for estimating the memory parameter of a (possibly) non-stationary process. As an application, we obtain minimax upper bounds for the log-scale regression estimator of the memory parameter for a Gaussian process and we derive an explicit expression of its variance."
                    },
                    {
                        "title": "Online EM for Functional Data",
                        "abstract": "A novel approach to perform unsupervised sequential learning for functional data is proposed. Our goal is to extract reference shapes (referred to as templates) from noisy, deformed and censored realizations of curves and images. Our model generalizes the Bayesian dense deformable template model (Allassonni\\`ere et al., 2007), a hierarchical model in which the template is the function to be estimated and the deformation is a nuisance, assumed to be random with a known prior distribution. The templates are estimated using a Monte Carlo version of the online Expectation-Maximization algorithm, extending the work from Capp\\'e and Moulines (2009). Our sequential inference framework is significantly more computationally efficient than equivalent batch learning algorithms, especially when the missing data is high-dimensional. Some numerical illustrations on curve registration problem and templates extraction from images are provided to support our findings."
                    },
                    {
                        "title": "Online EM Algorithm for Latent Data Models",
                        "abstract": "In this contribution, we propose a generic online (also sometimes called adaptive or recursive) version of the Expectation-Maximisation (EM) algorithm applicable to latent variable models of independent observations. Compared to the algorithm of Titterington (1984), this approach is more directly connected to the usual EM algorithm and does not rely on integration with respect to the complete data distribution. The resulting algorithm is usually simpler and is shown to achieve convergence to the stationary points of the Kullback-Leibler divergence between the marginal distribution of the observation and the model distribution at the optimal rate, i.e., that of the maximum likelihood estimator. In addition, the proposed approach is also suitable for conditional (or regression) models, as illustrated in the case of the mixture of linear regressions model."
                    },
                    {
                        "title": "On Upper-Confidence Bound Policies for Non-Stationary Bandit Problems",
                        "abstract": "Multi-armed bandit problems are considered as a paradigm of the trade-off between exploring the environment to find profitable actions and exploiting what is already known. In the stationary case, the distributions of the rewards do not change in time, Upper-Confidence Bound (UCB) policies have been shown to be rate optimal. A challenging variant of the MABP is the non-stationary bandit problem where the gambler must decide which arm to play while facing the possibility of a changing environment. In this paper, we consider the situation where the distributions of rewards remain constant over epochs and change at unknown time instants. We analyze two algorithms: the discounted UCB and the sliding-window UCB. We establish for these two algorithms an upper-bound for the expected regret by upper-bounding the expectation of the number of times a suboptimal arm is played. For that purpose, we derive a Hoeffding type inequality for self normalized deviations with a random number of summands. We establish a lower-bound for the regret in presence of abrupt changes in the arms reward distributions. We show that the discounted UCB and the sliding-window UCB both match the lower-bound up to a logarithmic factor."
                    },
                    {
                        "title": "Asymptotic properties of the maximum likelihood estimation in misspecified hidden Markov models",
                        "abstract": "Let $(Y_k)_{k\\in \\mathbb{Z}}$ be a stationary sequence on a probability space $(\\Omega,\\mathcal{A},\\mathbb{P})$ taking values in a standard Borel space $\\mathsf{Y}$. Consider the associated maximum likelihood estimator with respect to a parametrized family of hidden Markov models such that the law of the observations $(Y_k)_{k\\in \\mathbb{Z}}$ is not assumed to be described by any of the hidden Markov models of this family. In this paper we investigate the consistency of this estimator in such misspecified models under mild assumptions."
                    },
                    {
                        "title": "Non-asymptotic convergence analysis for the Unadjusted Langevin Algorithm",
                        "abstract": "In this paper, we study a method to sample from a target distribution $\\pi$ over $\\mathbb{R}^d$ having a positive density with respect to the Lebesgue measure, known up to a normalisation factor. This method is based on the Euler discretization of the overdamped Langevin stochastic differential equation associated with $\\pi$. For both constant and decreasing step sizes in the Euler discretization, we obtain non-asymptotic bounds for the convergence to the target distribution $\\pi$ in total variation distance. A particular attention is paid to the dependency on the dimension $d$, to demonstrate the applicability of this method in the high dimensional setting. These bounds improve and extend the results of (Dalalyan 2014)."
                    },
                    {
                        "title": "A quantitative Mc Diarmid's inequality for geometrically ergodic Markov chains",
                        "abstract": "We state and prove a quantitative version of the bounded difference inequality for geometrically ergodic Markov chains.   Our proof uses the same martingale decomposition as \\cite{MR3407208} but, compared to this paper, the exact coupling argument is modified to fill a gap between the strongly aperiodic case and the general aperiodic case."
                    },
                    {
                        "title": "High-dimensional Bayesian inference via the Unadjusted Langevin Algorithm",
                        "abstract": "We consider in this paper the problem of sampling a high-dimensional probability distribution $\\pi$ having a density with respect to the Lebesgue measure on $\\mathbb{R}^d$, known up to a normalization constant $x \\mapsto \\pi(x)= \\mathrm{e}^{-U(x)}/\\int_{\\mathbb{R}^d} \\mathrm{e}^{-U(y)} \\mathrm{d} y$. Such problem naturally occurs for example in Bayesian inference and machine learning. Under the assumption that $U$ is continuously differentiable, $\\nabla U$ is globally Lipschitz and $U$ is strongly convex, we obtain non-asymptotic bounds for the convergence to stationarity in Wasserstein distance of order $2$ and total variation distance of the sampling method based on the Euler discretization of the Langevin stochastic differential equation, for both constant and decreasing step sizes. The dependence on the dimension of the state space of these bounds is explicit. The convergence of an appropriately weighted empirical measure is also investigated and bounds for the mean square error and exponential deviation inequality are reported for functions which are measurable and bounded. An illustration to Bayesian inference for binary regression is presented to support our claims."
                    },
                    {
                        "title": "Stochastic Variable Metric Proximal Gradient with variance reduction for non-convex composite optimization",
                        "abstract": "This paper introduces a novel algorithm, the Perturbed Proximal Preconditioned SPIDER algorithm (3P-SPIDER), designed to solve finite sum non-convex composite optimization. It is a stochastic Variable Metric Forward-Backward algorithm, which allows approximate preconditioned forward operator and uses a variable metric proximity operator as the backward operator; it also proposes a mini-batch strategy with variance reduction to address the finite sum setting. We show that 3P-SPIDER extends some Stochastic preconditioned Gradient Descent-based algorithms and some Incremental Expectation Maximization algorithms to composite optimization and to the case the forward operator can not be computed in closed form. We also provide an explicit control of convergence in expectation of 3P-SPIDER, and study its complexity in order to satisfy the epsilon-approximate stationary condition. Our results are the first to combine the composite non-convex optimization setting, a variance reduction technique to tackle the finite sum setting by using a minibatch strategy and, to allow deterministic or random approximations of the preconditioned forward operator. Finally, through an application to inference in a logistic regression model with random effects, we numerically compare 3P-SPIDER to other stochastic forward-backward algorithms and discuss the role of some design parameters of 3P-SPIDER."
                    },
                    {
                        "title": "Comparison of Resampling Schemes for Particle Filtering",
                        "abstract": "This contribution is devoted to the comparison of various resampling approaches that have been proposed in the literature on particle filtering. It is first shown using simple arguments that the so-called residual and stratified methods do yield an improvement over the basic multinomial resampling approach. A simple counter-example showing that this property does not hold true for systematic resampling is given. Finally, some results on the large-sample behavior of the simple bootstrap filter algorithm are given. In particular, a central limit theorem is established for the case where resampling is performed using the residual approach."
                    }
                ]
            },
            "99f2ebba-4d7e-45e7-8a34-f5125b742b9f": {
                "pk": "99f2ebba-4d7e-45e7-8a34-f5125b742b9f",
                "name": "Michael I Jordan",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "32965ac4-c3aa-497e-be8c-93100aededbf": {
                "pk": "32965ac4-c3aa-497e-be8c-93100aededbf",
                "name": "Alain Durmus",
                "collaborators": [
                    "Eric Moulines",
                    "\u00c9ric Moulines",
                    "Nicolas Brosse",
                    "Arnaud Guillin",
                    "Marcelo Pereyra",
                    "Gersende Fort",
                    "Eero Saksman",
                    "Christophe Andrieu",
                    "Nikolas N\u00fcsken",
                    "Julien Roussel"
                ],
                "domain": [
                    "Bayesian Inference",
                    "Markov Chain Monte Carlo",
                    "Stochastic Processes",
                    "High-Dimensional Statistics"
                ],
                "publications": [
                    {
                        "title": "Non-asymptotic convergence analysis for the Unadjusted Langevin Algorithm",
                        "abstract": "In this paper, we study a method to sample from a target distribution $\\pi$ over $\\mathbb{R}^d$ having a positive density with respect to the Lebesgue measure, known up to a normalisation factor. This method is based on the Euler discretization of the overdamped Langevin stochastic differential equation associated with $\\pi$. For both constant and decreasing step sizes in the Euler discretization, we obtain non-asymptotic bounds for the convergence to the target distribution $\\pi$ in total variation distance. A particular attention is paid to the dependency on the dimension $d$, to demonstrate the applicability of this method in the high dimensional setting. These bounds improve and extend the results of (Dalalyan 2014)."
                    },
                    {
                        "title": "On the geometric convergence for MALA under verifiable conditions",
                        "abstract": "While the Metropolis Adjusted Langevin Algorithm (MALA) is a popular and widely used Markov chain Monte Carlo method, very few papers derive conditions that ensure its convergence. In particular, to the authors' knowledge, assumptions that are both easy to verify and guarantee geometric convergence, are still missing. In this work, we establish $V$-uniformly geometric convergence for MALA under mild assumptions about the target distribution. Unlike previous work, we only consider tail and smoothness conditions for the potential associated with the target distribution. These conditions are quite common in the MCMC literature and are easy to verify in practice. Finally, we pay special attention to the dependence of the bounds we derive on the step size of the Euler-Maruyama discretization, which corresponds to the proposal Markov kernel of MALA."
                    },
                    {
                        "title": "High-dimensional Bayesian inference via the Unadjusted Langevin Algorithm",
                        "abstract": "We consider in this paper the problem of sampling a high-dimensional probability distribution $\\pi$ having a density with respect to the Lebesgue measure on $\\mathbb{R}^d$, known up to a normalization constant $x \\mapsto \\pi(x)= \\mathrm{e}^{-U(x)}/\\int_{\\mathbb{R}^d} \\mathrm{e}^{-U(y)} \\mathrm{d} y$. Such problem naturally occurs for example in Bayesian inference and machine learning. Under the assumption that $U$ is continuously differentiable, $\\nabla U$ is globally Lipschitz and $U$ is strongly convex, we obtain non-asymptotic bounds for the convergence to stationarity in Wasserstein distance of order $2$ and total variation distance of the sampling method based on the Euler discretization of the Langevin stochastic differential equation, for both constant and decreasing step sizes. The dependence on the dimension of the state space of these bounds is explicit. The convergence of an appropriately weighted empirical measure is also investigated and bounds for the mean square error and exponential deviation inequality are reported for functions which are measurable and bounded. An illustration to Bayesian inference for binary regression is presented to support our claims."
                    },
                    {
                        "title": "Subgeometric rates of convergence in Wasserstein distance for Markov chains",
                        "abstract": "In this paper, we provide sufficient conditions for the existence of the invariant distribution and for subgeometric rates of convergence in Wasserstein distance for general state-space Markov chains which are (possibly) not irreducible. Compared to previous work, our approach is based on a purely probabilistic coupling construction which allows to retrieve rates of convergence matching those previously reported for convergence in total variation. Our results are applied to establish the subgeometric ergodicity in Wasserstein distance of non-linear autoregressive models and of the pre-conditioned Crank-Nicolson Markov chain Monte Carlo algorithm in Hilbert space."
                    },
                    {
                        "title": "On the convergence of Hamiltonian Monte Carlo",
                        "abstract": "This paper discusses the irreducibility and geometric ergodicity of the Hamiltonian Monte Carlo (HMC) algorithm. We consider cases where the number of steps of the symplectic integrator is either fixed or random. Under mild conditions on the potential $\\F$ associated with target distribution $\\pi$, we first show that the Markov kernel associated to the HMC algorithm is irreducible and recurrent. Under more stringent conditions, we then establish that the Markov kernel is Harris recurrent. Finally, we provide verifiable conditions on $\\F$ under which the HMC sampler is geometrically ergodic."
                    },
                    {
                        "title": "Hypocoercivity of Piecewise Deterministic Markov Process-Monte Carlo",
                        "abstract": "In this work, we establish $\\mathrm{L}^2$-exponential convergence for a broad class of Piecewise Deterministic Markov Processes recently proposed in the context of Markov Process Monte Carlo methods and covering in particular the Randomized Hamiltonian Monte Carlo, the Zig-Zag process and the Bouncy Particle Sampler. The kernel of the symmetric part of the generator of such processes is non-trivial, and we follow the ideas recently introduced by (Dolbeault et al., 2009, 2015) to develop a rigorous framework for hypocoercivity in a fairly general and unifying set-up, while deriving tractable estimates of the constants involved in terms of the parameters of the dynamics. As a by-product we characterize the scaling properties of these algorithms with respect to the dimension of classes of problems, therefore providing some theoretical evidence to support their practical relevance."
                    },
                    {
                        "title": "Normalizing constants of log-concave densities",
                        "abstract": "We derive explicit bounds for the computation of normalizing constants $Z$ for log-concave densities $\\pi = \\exp(-U)/Z$ with respect to the Lebesgue measure on $\\mathbb{R}^d$. Our approach relies on a Gaussian annealing combined with recent and precise bounds on the Unadjusted Langevin Algorithm (High-dimensional Bayesian inference via the Unadjusted Langevin Algorithm, A. Durmus and E. Moulines). Polynomial bounds in the dimension $d$ are obtained with an exponent that depends on the assumptions made on $U$. The algorithm also provides a theoretically grounded choice of the annealing sequence of variances. A numerical experiment supports our findings. Results of independent interest on the mean squared error of the empirical average of locally Lipschitz functions are established."
                    },
                    {
                        "title": "Bridging the Gap between Constant Step Size Stochastic Gradient Descent and Markov Chains",
                        "abstract": "We consider the minimization of an objective function given access to unbiased estimates of its gradient through stochastic gradient descent (SGD) with constant step-size. While the detailed analysis was only performed for quadratic functions, we provide an explicit asymptotic expansion of the moments of the averaged SGD iterates that outlines the dependence on initial conditions, the effect of noise and the step-size, as well as the lack of convergence in the general (non-quadratic) case. For this analysis, we bring tools from Markov chain theory into the analysis of stochastic gradient. We then show that Richardson-Romberg extrapolation may be used to get closer to the global optimum and we show empirical improvements of the new extrapolation scheme."
                    },
                    {
                        "title": "Analysis of Langevin Monte Carlo via convex optimization",
                        "abstract": "In this paper, we provide new insights on the Unadjusted Langevin Algorithm. We show that this method can be formulated as a first order optimization algorithm of an objective functional defined on the Wasserstein space of order $2$. Using this interpretation and techniques borrowed from convex optimization, we give a non-asymptotic analysis of this method to sample from logconcave smooth target distribution on $\\mathbb{R}^d$. Based on this interpretation, we propose two new methods for sampling from a non-smooth target distribution, which we analyze as well. Besides, these new algorithms are natural extensions of the Stochastic Gradient Langevin Dynamics (SGLD) algorithm, which is a popular extension of the Unadjusted Langevin Algorithm. Similar to SGLD, they only rely on approximations of the gradient of the target log density and can be used for large-scale Bayesian inference."
                    },
                    {
                        "title": "Piecewise Deterministic Markov Processes and their invariant measures",
                        "abstract": "Piecewise Deterministic Markov Processes (PDMPs) are studied in a general framework. First, different constructions are proven to be equivalent. Second, we introduce a coupling between two PDMPs following the same differential flow which implies quantitative bounds on the total variation between the marginal distributions of the two processes. Finally two results are established regarding the invariant measures of PDMPs. A practical condition to show that a probability measure is invariant for the associated PDMP semi-group is presented. In a second time, a bound on the invariant probability measures in $V$-norm of two PDMPs following the same differential flow is established. This last result is then applied to study the asymptotic bias of some non-exact PDMP MCMC methods."
                    },
                    {
                        "title": "Probability and moment inequalities for additive functionals of geometrically ergodic Markov chains",
                        "abstract": "In this paper, we establish moment and Bernstein-type inequalities for additive functionals of geometrically ergodic Markov chains. These inequalities extend the corresponding inequalities for independent random variables. Our conditions cover Markov chains converging geometrically to the stationary distribution either in $V$-norms or in weighted Wasserstein distances. Our inequalities apply to unbounded functions and depend explicitly on constants appearing in the conditions that we consider."
                    },
                    {
                        "title": "Divide-and-Conquer Posterior Sampling for Denoising Diffusion Priors",
                        "abstract": "Interest in the use of Denoising Diffusion Models (DDM) as priors for solving inverse Bayesian problems has recently increased significantly. However, sampling from the resulting posterior distribution poses a challenge. To solve this problem, previous works have proposed approximations to bias the drift term of the diffusion. In this work, we take a different approach and utilize the specific structure of the DDM prior to define a set of intermediate and simpler posterior sampling problems, resulting in a lower approximation error compared to previous methods. We empirically demonstrate the reconstruction capability of our method for general linear inverse problems using synthetic examples and various image restoration tasks."
                    },
                    {
                        "title": "Efficient Bayesian computation by proximal Markov chain Monte Carlo: when Langevin meets Moreau",
                        "abstract": "Modern imaging methods rely strongly on Bayesian inference techniques to solve challenging imaging problems. Currently, the predominant Bayesian computation approach is convex optimisation, which scales very efficiently to high dimensional image models and delivers accurate point estimation results. However, in order to perform more complex analyses, for example image uncertainty quantification or model selection, it is necessary to use more computationally intensive Bayesian computation techniques such as Markov chain Monte Carlo methods. This paper presents a new and highly efficient Markov chain Monte Carlo methodology to perform Bayesian computation for high dimensional models that are log-concave and non-smooth, a class of models that is central in imaging sciences. The methodology is based on a regularised unadjusted Langevin algorithm that exploits tools from convex analysis, namely Moreau-Yoshida envelopes and proximal operators, to construct Markov chains with favourable convergence properties. In addition to scaling efficiently to high dimensions, the method is straightforward to apply to models that are currently solved by using proximal optimisation algorithms. We provide a detailed theoretical analysis of the proposed methodology, including asymptotic and non-asymptotic convergence results with easily verifiable conditions, and explicit bounds on the convergence rates. The proposed methodology is demonstrated with four experiments related to image deconvolution and tomographic reconstruction with total-variation and $\\ell_1$ priors, where we conduct a range of challenging Bayesian analyses related to uncertainty quantification, hypothesis testing, and model selection in the absence of ground truth."
                    },
                    {
                        "title": "An Elementary Approach To Uniform In Time Propagation Of Chaos",
                        "abstract": "Based on a coupling approach, we prove uniform in time propagation of chaos for weakly interacting mean-field particle systems with possibly non-convex confinement and interaction potentials. The approach is based on a combination of reflection and synchronous couplings applied to the individual particles. It provides explicit quantitative bounds that significantly extend previous results for the convex case."
                    },
                    {
                        "title": "Sampling from a log-concave distribution with compact support with proximal Langevin Monte Carlo",
                        "abstract": "This paper presents a detailed theoretical analysis of the Langevin Monte Carlo sampling algorithm recently introduced in Durmus et al. (Efficient Bayesian computation by proximal Markov chain Monte Carlo: when Langevin meets Moreau, 2016) when applied to log-concave probability distributions that are restricted to a convex body $\\mathsf{K}$. This method relies on a regularisation procedure involving the Moreau-Yosida envelope of the indicator function associated with $\\mathsf{K}$. Explicit convergence bounds in total variation norm and in Wasserstein distance of order $1$ are established. In particular, we show that the complexity of this algorithm given a first order oracle is polynomial in the dimension of the state space. Finally, some numerical experiments are presented to compare our method with competing MCMC approaches from the literature."
                    },
                    {
                        "title": "The promises and pitfalls of Stochastic Gradient Langevin Dynamics",
                        "abstract": "Stochastic Gradient Langevin Dynamics (SGLD) has emerged as a key MCMC algorithm for Bayesian learning from large scale datasets. While SGLD with decreasing step sizes converges weakly to the posterior distribution, the algorithm is often used with a constant step size in practice and has demonstrated successes in machine learning tasks. The current practice is to set the step size inversely proportional to $N$ where $N$ is the number of training samples. As $N$ becomes large, we show that the SGLD algorithm has an invariant probability measure which significantly departs from the target posterior and behaves like Stochastic Gradient Descent (SGD). This difference is inherently due to the high variance of the stochastic gradients. Several strategies have been suggested to reduce this effect; among them, SGLD Fixed Point (SGLDFP) uses carefully designed control variates to reduce the variance of the stochastic gradients. We show that SGLDFP gives approximate samples from the posterior distribution, with an accuracy comparable to the Langevin Monte Carlo (LMC) algorithm for a computational cost sublinear in the number of data points. We provide a detailed analysis of the Wasserstein distances between LMC, SGLD, SGLDFP and SGD and explicit expressions of the means and covariance matrices of their invariant distributions. Our findings are supported by limited numerical experiments."
                    },
                    {
                        "title": "On Riemannian Stochastic Approximation Schemes with Fixed Step-Size",
                        "abstract": "This paper studies fixed step-size stochastic approximation (SA) schemes, including stochastic gradient schemes, in a Riemannian framework. It is motivated by several applications, where geodesics can be computed explicitly, and their use accelerates crude Euclidean methods. A fixed step-size scheme defines a family of time-homogeneous Markov chains, parametrized by the step-size. Here, using this formulation, non-asymptotic performance bounds are derived, under Lyapunov conditions. Then, for any step-size, the corresponding Markov chain is proved to admit a unique stationary distribution, and to be geometrically ergodic. This result gives rise to a family of stationary distributions indexed by the step-size, which is further shown to converge to a Dirac measure, concentrated at the solution of the problem at hand, as the step-size goes to 0. Finally, the asymptotic rate of this convergence is established, through an asymptotic expansion of the bias, and a central limit theorem."
                    },
                    {
                        "title": "Copula-like Variational Inference",
                        "abstract": "This paper considers a new family of variational distributions motivated by Sklar's theorem. This family is based on new copula-like densities on the hypercube with non-uniform marginals which can be sampled efficiently, i.e. with a complexity linear in the dimension of state space. Then, the proposed variational densities that we suggest can be seen as arising from these copula-like densities used as base distributions on the hypercube with Gaussian quantile functions and sparse rotation matrices as normalizing flows. The latter correspond to a rotation of the marginals with complexity $\\mathcal{O}(d \\log d)$. We provide some empirical evidence that such a variational family can also approximate non-Gaussian posteriors and can be beneficial compared to Gaussian approximations. Our method performs largely comparably to state-of-the-art variational approximations on standard regression and classification benchmarks for Bayesian Neural Networks."
                    },
                    {
                        "title": "Convergence of diffusions and their discretizations: from continuous to discrete processes and back",
                        "abstract": "In this paper, we establish new quantitative convergence bounds for a class of functional autoregressive models in weighted total variation metrics. To derive our results, we show that under mild assumptions, explicit minorization and Foster-Lyapunov drift conditions hold. The main applications and consequences of the bounds we obtain concern the geometric convergence of Euler-Maruyama discretizations of diffusions with identity covariance matrix. Second, as a corollary, we provide a new approach to establish quantitative convergence of these diffusion processes by applying our conclusions in the discrete-time setting to a well-suited sequence of discretizations whose associated stepsizes decrease towards zero."
                    }
                ]
            }
        }
    },
    "2402.06126": {
        "paper_data": {
            "title": "Learn To be Efficient: Build Structured Sparsity in Large Language Models",
            "url": "http://arxiv.org/abs/2402.06126v3",
            "arxiv_id": "2402.06126",
            "authors": [
                "Haizhong Zheng",
                "Xiaoyan Bai",
                "Xueshen Liu",
                "Z. Morley Mao",
                "Beidi Chen",
                "Fan Lai",
                "Atul Prakash"
            ],
            "abstract": "Large Language Models (LLMs) have achieved remarkable success with their billion-level parameters, yet they incur high inference overheads. The emergence of activation sparsity in LLMs provides a natural approach to reduce this cost by involving only parts of the parameters for inference. However, existing methods only focus on utilizing this naturally formed activation sparsity in a post-training setting, overlooking the potential for further amplifying this inherent sparsity. In this paper, we hypothesize that LLMs can learn to be efficient by achieving more structured activation sparsity. To achieve this, we introduce a novel training algorithm, Learn-To-be-Efficient (LTE), designed to train efficiency-aware LLMs to learn to activate fewer neurons and achieve a better trade-off between sparsity and performance. Furthermore, unlike SOTA MoEfication methods, which mainly focus on ReLU-based models, LTE can also be applied to LLMs like LLaMA using non-ReLU activations. Extensive evaluation on language understanding, language generation, and instruction tuning tasks show that LTE consistently outperforms SOTA baselines. Along with our hardware-aware custom kernel implementation, LTE reduces LLaMA2-7B inference latency by 25% at 50% sparsity.",
            "introduction": "   1 Introduction  Figure 1: Learn-To-be-Efficient\u00a0(LTE) performs efficiency-aware training to construct more structured model contextual sparsity for fast inference. Activated neurons in FFN are in orange.    Large language models (LLMs), like GPT-3\u00a0[2], OPT\u00a0[42], and LLaMA\u00a0[35, 36], have demonstrated impressive natural language ability. However, the skyrocketing number of model parameters has not only inflated the deployment costs of LLMs, due to their significant computational and memory demands during inference\u00a0[20], but also affected the user experience in many latency-sensitive applications, such as chatbots\u00a0[1] and autonomous driving\u00a0[16]. Recent advances have been leveraging sparsity to improve LLM inference efficiency, including model weight quantization\u00a0[8] and pruning\u00a0[21], token sparsity\u00a0[44, 41], and activation sparsity\u00a0[20, 43].   During LLM inference, Feed Forward Network (FFN) layers are the primary efficiency bottleneck, accounting for over 60% of the FLOPs and I/O operations\u00a0[20], but they exhibit high activation sparsity, especially in ReLU-based LLMs\u00a0[17]. For example, in a 175B OPT model, more than 95% of activations in FFN layers are zeros\u00a0[20]. Recent advances leverage this activation sparsity via MoEfication\u00a0[43, 24]: they convert FFN layers to MoE layers by grouping neurons in the FFN intermediate layer into n\ud835\udc5bnitalic_n experts and then applying a router to select k\ud835\udc58kitalic_k most important experts for each input token. Unlike training the MoE model from scratch, MoEfication converts FFN layers in pretrained LLMs into MoE layers with relatively small training resources.   However, existing advances only focus on manipulating the pre-trained activation sparsity of neurons, overlooking the potential for further increasing this inherent sparsity. Moreover, they are limited to the historically ReLU-based LLMs, whereas emerging advanced models use soft activation functions for better model quality (e.g., SwiGLU in LLaMA\u00a0[29] and GeGLU in Gemma\u00a0[34]), exhibiting much lower natural activation sparsity. While existing works suggest replacing the model activation with ReLU to enable sparsity\u00a0[43, 24], we show that this can hurt model performance\u00a0(Section\u00a05.2).   In this paper, we hypothesize that LLMs can learn to be efficient and achieve more structured activation sparsity. As shown in Figure\u00a01, our key insight is that, without compromising model quality, we can also train LLMs to be efficiency-aware by developing more structured sparsity, which is more friendly for achieving hardware speedup. However, creating structured sparsity in pretrained LLMs is non-trivial due to two practical challenges:     1.  How to train routers more stably? The widely-used Top-k Softmax routing can lead to a severe accuracy drop\u00a0(Section3.2). How to jointly train the model and routers in a MoEfication setting is still open-ended.    2.  How to select the right experts in serving? The number of experts needed in MoE layers depends on specific inputs and layers, implying a trade-off between model efficiency and quality.      To address these challenges, we introduce Learn-To-be-Efficient (LTE), a novel training algorithm to train efficiency-aware models. LTE integrates an efficiency loss penalty, encouraging models to activate fewer neurons in their FFN layers while keeping good task performance. Additionally, LTE adopts a threshold-based Sigmoid routing strategy to select experts and employs a two-stage training mechanism to improve training stability. LTE achieves a more flexible selection of experts instead of selecting a fixed number of experts for all layers and inputs. Same as Deja Vu\u00a0[20] and moefication\u00a0[43], LTE provides very structured sparsity. We further develop a custom CUDA kernel with Triton, a Python-like CUDA programming",
            "references": [
                {
                    "title": "Towards Knowledge-driven Autonomous Driving",
                    "abstract": "This paper explores the emerging knowledge-driven autonomous driving technologies. Our investigation highlights the limitations of current autonomous driving systems, in particular their sensitivity to data bias, difficulty in handling long-tail scenarios, and lack of interpretability. Conversely, knowledge-driven methods with the abilities of cognition, generalization and life-long learning emerge as a promising way to overcome these challenges. This paper delves into the essence of knowledge-driven autonomous driving and examines its core components: dataset \\&benchmark, environment, and driver agent. By leveraging large language models, world models, neural rendering, and other advanced artificial intelligence techniques, these components collectively contribute to a more holistic, adaptive, and intelligent autonomous driving system. The paper systematically organizes and reviews previous research efforts in this area, and provides insights and guidance for future research and practical applications of autonomous driving. We will continually share the latest updates on cutting-edge developments in knowledge-driven autonomous driving along with the relevant valuable open-source resources at: \\url{https://github.com/PJLab-ADG/awesome-knowledge-driven-AD}."
                },
                {
                    "title": "Deja Vu: Contextual Sparsity for Efficient LLMs at Inference Time",
                    "abstract": "Large language models (LLMs) with hundreds of billions of parameters have sparked a new wave of exciting AI applications. However, they are computationally expensive at inference time. Sparsity is a natural approach to reduce this cost, but existing methods either require costly retraining, have to forgo LLM's in-context learning ability, or do not yield wall-clock time speedup on modern hardware. We hypothesize that contextual sparsity, which are small, input-dependent sets of attention heads and MLP parameters that yield approximately the same output as the dense model for a given input, can address these issues. We show that contextual sparsity exists, that it can be accurately predicted, and that we can exploit it to speed up LLM inference in wall-clock time without compromising LLM's quality or in-context learning ability. Based on these insights, we propose DejaVu, a system that uses a low-cost algorithm to predict contextual sparsity on the fly given inputs to each layer, along with an asynchronous and hardware-aware implementation that speeds up LLM inference. We validate that DejaVu can reduce the inference latency of OPT-175B by over 2X compared to the state-of-the-art FasterTransformer, and over 6X compared to the widely used Hugging Face implementation, without compromising model quality. The code is available at https://github.com/FMInference/DejaVu."
                },
                {
                    "title": "Mistral 7B",
                    "abstract": "We introduce Mistral 7B v0.1, a 7-billion-parameter language model engineered for superior performance and efficiency. Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation. Our model leverages grouped-query attention (GQA) for faster inference, coupled with sliding window attention (SWA) to effectively handle sequences of arbitrary length with a reduced inference cost. We also provide a model fine-tuned to follow instructions, Mistral 7B -- Instruct, that surpasses the Llama 2 13B -- Chat model both on human and automated benchmarks. Our models are released under the Apache 2.0 license."
                },
                {
                    "title": "ReLU Strikes Back: Exploiting Activation Sparsity in Large Language Models",
                    "abstract": "Large Language Models (LLMs) with billions of parameters have drastically transformed AI applications. However, their demanding computation during inference has raised significant challenges for deployment on resource-constrained devices. Despite recent trends favoring alternative activation functions such as GELU or SiLU, known for increased computation, this study strongly advocates for reinstating ReLU activation in LLMs. We demonstrate that using the ReLU activation function has a negligible impact on convergence and performance while significantly reducing computation and weight transfer. This reduction is particularly valuable during the memory-bound inference step, where efficiency is paramount. Exploring sparsity patterns in ReLU-based LLMs, we unveil the reutilization of activated neurons for generating new tokens and leveraging these insights, we propose practical strategies to substantially reduce LLM inference computation up to three times, using ReLU activations with minimal performance trade-offs."
                },
                {
                    "title": "Efficient Streaming Language Models with Attention Sinks",
                    "abstract": "Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching previous tokens' Key and Value states (KV) consumes extensive memory. Secondly, popular LLMs cannot generalize to longer texts than the training sequence length. Window attention, where only the most recent KVs are cached, is a natural approach -- but we show that it fails when the text length surpasses the cache size. We observe an interesting phenomenon, namely attention sink, that keeping the KV of initial tokens will largely recover the performance of window attention. In this paper, we first demonstrate that the emergence of attention sink is due to the strong attention scores towards initial tokens as a\"sink\"even if they are not semantically important. Based on the above analysis, we introduce StreamingLLM, an efficient framework that enables LLMs trained with a finite length attention window to generalize to infinite sequence lengths without any fine-tuning. We show that StreamingLLM can enable Llama-2, MPT, Falcon, and Pythia to perform stable and efficient language modeling with up to 4 million tokens and more. In addition, we discover that adding a placeholder token as a dedicated attention sink during pre-training can further improve streaming deployment. In streaming settings, StreamingLLM outperforms the sliding window recomputation baseline by up to 22.2x speedup. Code and datasets are provided at https://github.com/mit-han-lab/streaming-llm."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "A Simple and Effective Pruning Approach for Large Language Models",
                    "abstract": "As their size increases, Large Languages Models (LLMs) are natural candidates for network pruning methods: approaches that drop a subset of network weights while striving to preserve performance. Existing methods, however, require either retraining, which is rarely affordable for billion-scale LLMs, or solving a weight reconstruction problem reliant on second-order information, which may also be computationally expensive. In this paper, we introduce a novel, straightforward yet effective pruning method, termed Wanda (Pruning by Weights and activations), designed to induce sparsity in pretrained LLMs. Motivated by the recent observation of emergent large magnitude features in LLMs, our approach prunes weights with the smallest magnitudes multiplied by the corresponding input activations, on a per-output basis. Notably, Wanda requires no retraining or weight update, and the pruned LLM can be used as is. We conduct a thorough evaluation of our method Wanda on LLaMA and LLaMA-2 across various language benchmarks. Wanda significantly outperforms the established baseline of magnitude pruning and performs competitively against recent method involving intensive weight update. Code is available at https://github.com/locuslab/wanda."
                },
                {
                    "title": "How Far Can Camels Go? Exploring the State of Instruction Tuning on Open Resources",
                    "abstract": "In this work we explore recent advances in instruction-tuning language models on a range of open instruction-following datasets. Despite recent claims that open models can be on par with state-of-the-art proprietary models, these claims are often accompanied by limited evaluation, making it difficult to compare models across the board and determine the utility of various resources. We provide a large set of instruction-tuned models from 6.7B to 65B parameters in size, trained on 12 instruction datasets ranging from manually curated (e.g., OpenAssistant) to synthetic and distilled (e.g., Alpaca) and systematically evaluate them on their factual knowledge, reasoning, multilinguality, coding, and open-ended instruction following abilities through a collection of automatic, model-based, and human-based metrics. We further introduce T\\\"ulu, our best performing instruction-tuned model suite finetuned on a combination of high-quality open resources. Our experiments show that different instruction-tuning datasets can uncover or enhance specific skills, while no single dataset (or combination) provides the best performance across all evaluations. Interestingly, we find that model and human preference-based evaluations fail to reflect differences in model capabilities exposed by benchmark-based evaluations, suggesting the need for the type of systemic evaluation performed in this work. Our evaluations show that the best model in any given evaluation reaches on average 87% of ChatGPT performance, and 73% of GPT-4 performance, suggesting that further investment in building better base models and instruction-tuning data is required to close the gap. We release our instruction-tuned models, including a fully finetuned 65B T\\\"ulu, along with our code, data, and evaluation framework at https://github.com/allenai/open-instruct to facilitate future research."
                },
                {
                    "title": "LLM-Pruner: On the Structural Pruning of Large Language Models",
                    "abstract": "Large language models (LLMs) have shown remarkable capabilities in language understanding and generation. However, such impressive capability typically comes with a substantial model size, which presents significant challenges in both the deployment, inference, and training stages. With LLM being a general-purpose task solver, we explore its compression in a task-agnostic manner, which aims to preserve the multi-task solving and language generation ability of the original LLM. One challenge to achieving this is the enormous size of the training corpus of LLM, which makes both data transfer and model post-training over-burdensome. Thus, we tackle the compression of LLMs within the bound of two constraints: being task-agnostic and minimizing the reliance on the original training dataset. Our method, named LLM-Pruner, adopts structural pruning that selectively removes non-critical coupled structures based on gradient information, maximally preserving the majority of the LLM's functionality. To this end, the performance of pruned models can be efficiently recovered through tuning techniques, LoRA, in merely 3 hours, requiring only 50K data. We validate the LLM-Pruner on three LLMs, including LLaMA, Vicuna, and ChatGLM, and demonstrate that the compressed models still exhibit satisfactory capabilities in zero-shot classification and generation. The code is available at: https://github.com/horseee/LLM-Pruner"
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot",
                    "abstract": "We show for the first time that large-scale generative pretrained transformer (GPT) family models can be pruned to at least 50% sparsity in one-shot, without any retraining, at minimal loss of accuracy. This is achieved via a new pruning method called SparseGPT, specifically designed to work efficiently and accurately on massive GPT-family models. We can execute SparseGPT on the largest available open-source models, OPT-175B and BLOOM-176B, in under 4.5 hours, and can reach 60% unstructured sparsity with negligible increase in perplexity: remarkably, more than 100 billion weights from these models can be ignored at inference time. SparseGPT generalizes to semi-structured (2:4 and 4:8) patterns, and is compatible with weight quantization approaches. The code is available at: https://github.com/IST-DASLab/sparsegpt."
                },
                {
                    "title": "GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers",
                    "abstract": "Generative Pre-trained Transformer models, known as GPT or OPT, set themselves apart through breakthrough performance across complex language modelling tasks, but also by their extremely high computational and storage costs. Specifically, due to their massive size, even inference for large, highly-accurate GPT models may require multiple performant GPUs, which limits the usability of such models. While there is emerging work on relieving this pressure via model compression, the applicability and performance of existing compression techniques is limited by the scale and complexity of GPT models. In this paper, we address this challenge, and propose GPTQ, a new one-shot weight quantization method based on approximate second-order information, that is both highly-accurate and highly-efficient. Specifically, GPTQ can quantize GPT models with 175 billion parameters in approximately four GPU hours, reducing the bitwidth down to 3 or 4 bits per weight, with negligible accuracy degradation relative to the uncompressed baseline. Our method more than doubles the compression gains relative to previously-proposed one-shot quantization methods, preserving accuracy, allowing us for the first time to execute an 175 billion-parameter model inside a single GPU for generative inference. Moreover, we also show that our method can still provide reasonable accuracy in the extreme quantization regime, in which weights are quantized to 2-bit or even ternary quantization levels. We show experimentally that these improvements can be leveraged for end-to-end inference speedups over FP16, of around 3.25x when using high-end GPUs (NVIDIA A100) and 4.5x when using more cost-effective ones (NVIDIA A6000). The implementation is available at https://github.com/IST-DASLab/gptq."
                },
                {
                    "title": "The Lazy Neuron Phenomenon: On Emergence of Activation Sparsity in Transformers",
                    "abstract": "This paper studies the curious phenomenon for machine learning models with Transformer architectures that their activation maps are sparse. By activation map we refer to the intermediate output of the multi-layer perceptrons (MLPs) after a ReLU activation function, and by sparse we mean that on average very few entries (e.g., 3.0% for T5-Base and 6.3% for ViT-B16) are nonzero for each input to MLP. Moreover, larger Transformers with more layers and wider MLP hidden dimensions are sparser as measured by the percentage of nonzero entries. Through extensive experiments we demonstrate that the emergence of sparsity is a prevalent phenomenon that occurs for both natural language processing and vision tasks, on both training and evaluation data, for Transformers of various configurations, at layers of all depth levels, as well as for other architectures including MLP-mixers and 2-layer MLPs. We show that sparsity also emerges using training datasets with random labels, or with random inputs, or with infinite amount of data, demonstrating that sparsity is not a result of a specific family of datasets. We discuss how sparsity immediately implies a way to significantly reduce the FLOP count and improve efficiency for Transformers. Moreover, we demonstrate perhaps surprisingly that enforcing an even sparser activation via Top-k thresholding with a small value of k brings a collection of desired but missing properties for Transformers, namely less sensitivity to noisy training data, more robustness to input corruptions, and better calibration for their prediction confidence."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "MoEfication: Transformer Feed-forward Layers are Mixtures of Experts",
                    "abstract": "Recent work has shown that feed-forward networks (FFNs) in pre-trained Transformers are a key component, storing various linguistic and factual knowledge. However, the computational patterns of FFNs are still unclear. In this work, we study the computational patterns of FFNs and observe that most inputs only activate a tiny ratio of neurons of FFNs. This phenomenon is similar to the sparsity of the human brain, which drives research on functional partitions of the human brain. To verify whether functional partitions also emerge in FFNs, we propose to convert a model into its MoE version with the same parameters, namely MoEfication. Specifically, MoEfication consists of two phases: (1) splitting the parameters of FFNs into multiple functional partitions as experts, and (2) building expert routers to decide which experts will be used for each input. Experimental results show that MoEfication can conditionally use 10% to 30% of FFN parameters while maintaining over 95% original performance for different models on various downstream tasks. Besides, MoEfication brings two advantages: (1) it significantly reduces the FLOPS of inference, i.e., 2x speedup with 25% of FFN parameters, and (2) it provides a fine-grained perspective to study the inner mechanism of FFNs. The source code of this paper can be obtained from https://github.com/thunlp/MoEfication."
                },
                {
                    "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity",
                    "abstract": "In deep learning, models typically reuse the same parameters for all inputs. Mixture of Experts (MoE) defies this and instead selects different parameters for each incoming example. The result is a sparsely-activated model -- with outrageous numbers of parameters -- but a constant computational cost. However, despite several notable successes of MoE, widespread adoption has been hindered by complexity, communication costs and training instability -- we address these with the Switch Transformer. We simplify the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs. Our proposed training techniques help wrangle the instabilities and we show large sparse models may be trained, for the first time, with lower precision (bfloat16) formats. We design models based off T5-Base and T5-Large to obtain up to 7x increases in pre-training speed with the same computational resources. These improvements extend into multilingual settings where we measure gains over the mT5-Base version across all 101 languages. Finally, we advance the current scale of language models by pre-training up to trillion parameter models on the\"Colossal Clean Crawled Corpus\"and achieve a 4x speedup over the T5-XXL model."
                },
                {
                    "title": "Transformer Feed-Forward Layers Are Key-Value Memories",
                    "abstract": "Feed-forward layers constitute two-thirds of a transformer model\u2019s parameters, yet their role in the network remains under-explored. We show that feed-forward layers in transformer-based language models operate as key-value memories, where each key correlates with textual patterns in the training examples, and each value induces a distribution over the output vocabulary. Our experiments show that the learned patterns are human-interpretable, and that lower layers tend to capture shallow patterns, while upper layers learn more semantic ones. The values complement the keys\u2019 input patterns by inducing output distributions that concentrate probability mass on tokens likely to appear immediately after each pattern, particularly in the upper layers. Finally, we demonstrate that the output of a feed-forward layer is a composition of its memories, which is subsequently refined throughout the model\u2019s layers via residual connections to produce the final output distribution."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "GLU Variants Improve Transformer",
                    "abstract": "Gated Linear Units (arXiv:1612.08083) consist of the component-wise product of two linear projections, one of which is first passed through a sigmoid function. Variations on GLU are possible, using different nonlinear (or even linear) functions in place of sigmoid. We test these variants in the feed-forward sublayers of the Transformer (arXiv:1706.03762) sequence-to-sequence model, and find that some of them yield quality improvements over the typically-used ReLU or GELU activations."
                },
                {
                    "title": "RoBERTa: A Robustly Optimized BERT Pretraining Approach",
                    "abstract": "Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code."
                },
                {
                    "title": "Mesh-TensorFlow: Deep Learning for Supercomputers",
                    "abstract": "Batch-splitting (data-parallelism) is the dominant distributed Deep Neural Network (DNN) training strategy, due to its universal applicability and its amenability to Single-Program-Multiple-Data (SPMD) programming. However, batch-splitting suffers from problems including the inability to train very large models (due to memory constraints), high latency, and inefficiency at small batch sizes. All of these can be solved by more general distribution strategies (model-parallelism). Unfortunately, efficient model-parallel algorithms tend to be complicated to discover, describe, and to implement, particularly on large clusters. We introduce Mesh-TensorFlow, a language for specifying a general class of distributed tensor computations. Where data-parallelism can be viewed as splitting tensors and operations along the \"batch\" dimension, in Mesh-TensorFlow, the user can specify any tensor-dimensions to be split across any dimensions of a multi-dimensional mesh of processors. A Mesh-TensorFlow graph compiles into a SPMD program consisting of parallel operations coupled with collective communication primitives such as Allreduce. We use Mesh-TensorFlow to implement an efficient data-parallel, model-parallel version of the Transformer sequence-to-sequence model. Using TPU meshes of up to 512 cores, we train Transformer models with up to 5 billion parameters, surpassing state of the art results on WMT'14 English-to-French translation task and the one-billion-word language modeling benchmark. Mesh-Tensorflow is available at this https URL ."
                },
                {
                    "title": "Don\u2019t Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarization",
                    "abstract": "We introduce \u201cextreme summarization\u201d, a new single-document summarization task which does not favor extractive strategies and calls for an abstractive modeling approach. The idea is to create a short, one-sentence news summary answering the question \u201cWhat is the article about?\u201d. We collect a real-world, large-scale dataset for this task by harvesting online articles from the British Broadcasting Corporation (BBC). We propose a novel abstractive model which is conditioned on the article\u2019s topics and based entirely on convolutional neural networks. We demonstrate experimentally that this architecture captures long-range dependencies in a document and recognizes pertinent content, outperforming an oracle extractive system and state-of-the-art abstractive approaches when evaluated automatically and by humans."
                },
                {
                    "title": "Neural Network Acceptability Judgments",
                    "abstract": "Abstract This paper investigates the ability of artificial neural networks to judge the grammatical acceptability of a sentence, with the goal of testing their linguistic competence. We introduce the Corpus of Linguistic Acceptability (CoLA), a set of 10,657 English sentences labeled as grammatical or ungrammatical from published linguistics literature. As baselines, we train several recurrent neural network models on acceptability classification, and find that our models outperform unsupervised models by Lau et al. (2016) on CoLA. Error-analysis on specific grammatical phenomena reveals that both Lau et al.\u2019s models and ours learn systematic generalizations like subject-verb-object order. However, all models we test perform far below human level on a wide range of grammatical constructions."
                },
                {
                    "title": "GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding",
                    "abstract": "Human ability to understand language is general, flexible, and robust. In contrast, most NLU models above the word level are designed for a specific task and struggle with out-of-domain data. If we aspire to develop models with understanding beyond the detection of superficial correspondences between inputs and outputs, then it is critical to develop a unified model that can execute a range of linguistic tasks across different domains. To facilitate research in this direction, we present the General Language Understanding Evaluation (GLUE, gluebenchmark.com): a benchmark of nine diverse NLU tasks, an auxiliary dataset for probing models for understanding of specific linguistic phenomena, and an online platform for evaluating and comparing models. For some benchmark tasks, training data is plentiful, but for others it is limited or does not match the genre of the test set. GLUE thus favors models that can represent linguistic knowledge in a way that facilitates sample-efficient learning and effective knowledge-transfer across tasks. While none of the datasets in GLUE were created from scratch for the benchmark, four of them feature privately-held test data, which is used to ensure that the benchmark is used fairly. We evaluate baselines that use ELMo (Peters et al., 2018), a powerful transfer learning technique, as well as state-of-the-art sentence representation models. The best models still achieve fairly low absolute scores. Analysis with our diagnostic dataset yields similarly weak performance over all phenomena tested, with some exceptions."
                },
                {
                    "title": "SemEval-2017 Task 1: Semantic Textual Similarity Multilingual and Crosslingual Focused Evaluation",
                    "abstract": "Semantic Textual Similarity (STS) measures the meaning similarity of sentences. Applications include machine translation (MT), summarization, generation, question answering (QA), short answer grading, semantic search, dialog and conversational systems. The STS shared task is a venue for assessing the current state-of-the-art. The 2017 task focuses on multilingual and cross-lingual pairs with one sub-track exploring MT quality estimation (MTQE) data. The task obtained strong participation from 31 teams, with 17 participating in all language tracks. We summarize performance and review a selection of well performing methods. Analysis highlights common errors, providing insight into the limitations of existing models. To support ongoing work on semantic representations, the STS Benchmark is introduced as a new shared training and evaluation set carefully selected from the corpus of English STS shared task data (2012-2017)."
                },
                {
                    "title": "The E2E Dataset: New Challenges For End-to-End Generation",
                    "abstract": "This paper describes the E2E data, a new dataset for training end-to-end, data-driven natural language generation systems in the restaurant domain, which is ten times bigger than existing, frequently used datasets in this area. The E2E dataset poses new challenges: (1) its human reference texts show more lexical richness and syntactic variation, including discourse phenomena; (2) generating from this set requires content selection. As such, learning from this dataset promises more natural, varied and less template-like system utterances. We also establish a baseline on this dataset, which illustrates some of the difficulties associated with this data."
                },
                {
                    "title": "A Broad-Coverage Challenge Corpus for Sentence Understanding through Inference",
                    "abstract": "This paper introduces the Multi-Genre Natural Language Inference (MultiNLI) corpus, a dataset designed for use in the development and evaluation of machine learning models for sentence understanding. At 433k examples, this resource is one of the largest corpora available for natural language inference (a.k.a. recognizing textual entailment), improving upon available resources in both its coverage and difficulty. MultiNLI accomplishes this by offering data from ten distinct genres of written and spoken English, making it possible to evaluate systems on nearly the full complexity of the language, while supplying an explicit setting for evaluating cross-genre domain adaptation. In addition, an evaluation using existing machine learning models designed for the Stanford NLI corpus shows that it represents a substantially more difficult task than does that corpus, despite the two showing similar levels of inter-annotator agreement."
                },
                {
                    "title": "Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer",
                    "abstract": "The capacity of a neural network to absorb information is limited by its number of parameters. Conditional computation, where parts of the network are active on a per-example basis, has been proposed in theory as a way of dramatically increasing model capacity without a proportional increase in computation. In practice, however, there are significant algorithmic and performance challenges. In this work, we address these challenges and finally realize the promise of conditional computation, achieving greater than 1000x improvements in model capacity with only minor losses in computational efficiency on modern GPU clusters. We introduce a Sparsely-Gated Mixture-of-Experts layer (MoE), consisting of up to thousands of feed-forward sub-networks. A trainable gating network determines a sparse combination of these experts to use for each example. We apply the MoE to the tasks of language modeling and machine translation, where model capacity is critical for absorbing the vast quantities of knowledge available in the training corpora. We present model architectures in which a MoE with up to 137 billion parameters is applied convolutionally between stacked LSTM layers. On large language modeling and machine translation benchmarks, these models achieve significantly better results than state-of-the-art at lower computational cost."
                },
                {
                    "title": "Pointer Sentinel Mixture Models",
                    "abstract": "Recent neural network sequence models with softmax classifiers have achieved their best language modeling performance only with very large hidden states and large vocabularies. Even then they struggle to predict rare or unseen words even if the context makes the prediction unambiguous. We introduce the pointer sentinel mixture architecture for neural sequence models which has the ability to either reproduce a word from the recent context or produce a word from a standard softmax classifier. Our pointer sentinel-LSTM model achieves state of the art language modeling performance on the Penn Treebank (70.9 perplexity) while using far fewer parameters than a standard softmax LSTM. In order to evaluate how well language models can exploit longer contexts and deal with more realistic vocabularies and larger corpora we also introduce the freely available WikiText corpus."
                },
                {
                    "title": "Gaussian Error Linear Units (GELUs)",
                    "abstract": "We propose the Gaussian Error Linear Unit (GELU), a high-performing neural network activation function. The GELU activation function is $x\\Phi(x)$, where $\\Phi(x)$ the standard Gaussian cumulative distribution function. The GELU nonlinearity weights inputs by their value, rather than gates inputs by their sign as in ReLUs ($x\\mathbf{1}_{x>0}$). We perform an empirical evaluation of the GELU nonlinearity against the ReLU and ELU activations and find performance improvements across all considered computer vision, natural language processing, and speech tasks."
                },
                {
                    "title": "SQuAD: 100,000+ Questions for Machine Comprehension of Text",
                    "abstract": "We present the Stanford Question Answering Dataset (SQuAD), a new reading comprehension dataset consisting of 100,000+ questions posed by crowdworkers on a set of Wikipedia articles, where the answer to each question is a segment of text from the corresponding reading passage. We analyze the dataset to understand the types of reasoning required to answer the questions, leaning heavily on dependency and constituency trees. We build a strong logistic regression model, which achieves an F1 score of 51.0%, a significant improvement over a simple baseline (20%). However, human performance (86.8%) is much higher, indicating that the dataset presents a good challenge problem for future research. \nThe dataset is freely available at this https URL"
                },
                {
                    "title": "Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank",
                    "abstract": "Semantic word spaces have been very useful but cannot express the meaning of longer phrases in a principled way. Further progress towards understanding compositionality in tasks such as sentiment detection requires richer supervised training and evaluation resources and more powerful models of composition. To remedy this, we introduce a Sentiment Treebank. It includes fine grained sentiment labels for 215,154 phrases in the parse trees of 11,855 sentences and presents new challenges for sentiment compositionality. To address them, we introduce the Recursive Neural Tensor Network. When trained on the new treebank, this model outperforms all previous methods on several metrics. It pushes the state of the art in single sentence positive/negative classification from 80% up to 85.4%. The accuracy of predicting fine-grained sentiment labels for all phrases reaches 80.7%, an improvement of 9.7% over bag of features baselines. Lastly, it is the only model that can accurately capture the effects of negation and its scope at various tree levels for both positive and negative phrases."
                },
                {
                    "title": "ROUGE: A Package for Automatic Evaluation of Summaries",
                    "abstract": "ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST."
                },
                {
                    "title": "A Fast and High Quality Multilevel Scheme for Partitioning Irregular Graphs",
                    "abstract": "Recently, a number of researchers have investigated a class of graph partitioning algorithms that reduce the size of the graph by collapsing vertices and edges, partition the smaller graph, and then uncoarsen it to construct a partition for the original graph [Bui and Jones, Proc. of the 6th SIAM Conference on Parallel Processing for Scientific Computing, 1993, 445--452; Hendrickson and Leland, A Multilevel Algorithm for Partitioning Graphs, Tech. report SAND 93-1301, Sandia National Laboratories, Albuquerque, NM, 1993]. From the early work it was clear that multilevel techniques held great promise; however, it was not known if they can be made to consistently produce high quality partitions for graphs arising in a wide range of application domains. We investigate the effectiveness of many different choices for all three phases: coarsening, partition of the coarsest graph, and refinement. In particular, we present a new coarsening heuristic (called heavy-edge heuristic) for which the size of the partition of the coarse graph is within a small factor of the size of the final partition obtained after multilevel refinement. We also present a much faster variation of the Kernighan--Lin (KL) algorithm for refining during uncoarsening. We test our scheme on a large number of graphs arising in various domains including finite element methods, linear programming, VLSI, and transportation. Our experiments show that our scheme produces partitions that are consistently better than those produced by spectral partitioning schemes in substantially smaller time. Also, when our scheme is used to compute fill-reducing orderings for sparse matrices, it produces orderings that have substantially smaller fill than the widely used multiple minimum degree algorithm."
                },
                {
                    "title": "Adaptive Mixtures of Local Experts",
                    "abstract": "We present a new supervised learning procedure for systems composed of many separate networks, each of which learns to handle a subset of the complete set of training cases. The new procedure can be viewed either as a modular version of a multilayer supervised network, or as an associative version of competitive learning. It therefore provides a new link between these two apparently different approaches. We demonstrate that the learning procedure divides up a vowel discrimination task into appropriate subtasks, each of which can be solved by a very simple expert network."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Automatically Constructing a Corpus of Sentential Paraphrases",
                    "abstract": "An obstacle to research in automatic paraphrase identification and generation is the lack of large-scale, publiclyavailable labeled corpora of sentential paraphrases. This paper describes the creation of the recently-released Microsoft Research Paraphrase Corpus, which contains 5801 sentence pairs, each hand-labeled with a binary judgment as to whether the pair constitutes a paraphrase. The corpus was created using heuristic extraction techniques in conjunction with an SVM-based classifier to select likely sentence-level paraphrases from a large corpus of topicclustered news data. These pairs were then submitted to human judges, who confirmed that 67% were in fact semantically equivalent. In addition to describing the corpus itself, we explore a number of issues that arose in defining guidelines for the human raters."
                },
                {
                    "title": "The PASCAL Recognising Textual Entailment Challenge",
                    "abstract": "This paper describes the PASCAL Network of Excellence first Recognising Textual Entailment (RTE-1) Challenge benchmark 1 . The RTE task is defined as recognizing, given two text fragments, whether the meaning of one text can be inferred (entailed) from the other. This application-independent task is suggested as capturing major inferences about the variability of semantic expression which are commonly needed across multiple applications. The Challenge has raised noticeable attention in the research community, attracting 17 submissions from diverse groups, suggesting the generic relevance of the task."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively train large language models (LLMs) to achieve structured activation sparsity for improved inference efficiency without compromising model performance?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing computational and memory demands of LLMs, which hinder their deployment in latency-sensitive applications. By developing methods to enhance inference efficiency through structured sparsity, this research could lead to significant advancements in model deployment, user experience, and resource utilization. Furthermore, it could inspire future research into more efficient model architectures and training methodologies, ultimately leading to practical applications in various fields, including natural language processing and autonomous systems.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the instability of training routers in a mixture of experts (MoE) setting, which can lead to accuracy drops, and the complexity of selecting the appropriate number of experts for different inputs and layers. Naive approaches may fail because they do not account for the need for structured sparsity or the trade-offs between model efficiency and performance. Additionally, the integration of new routing strategies and the joint training of models and routers introduce technical and theoretical obstacles that complicate the training process.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on manipulating pre-trained activation sparsity without fully exploring the potential for structured sparsity. Existing solutions have been limited to ReLU-based LLMs, which do not generalize well to newer models using soft activation functions. Barriers such as the lack of effective training algorithms for routers and the challenge of balancing efficiency with model quality have prevented this problem from being adequately addressed. Our approach differs by introducing the Learn-To-be-Efficient (LTE) algorithm, which incorporates an efficiency loss penalty and a novel routing strategy to enhance training stability and flexibility.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the Learn-To-be-Efficient (LTE) training algorithm, which integrates an efficiency loss penalty to encourage reduced neuron activation in FFN layers while maintaining task performance. We will utilize a dataset of pre-trained LLMs and evaluate the model's performance using metrics such as inference speed and accuracy. The expected outcomes include achieving a more structured activation sparsity in LLMs, improved inference efficiency, and enhanced model performance, demonstrating the effectiveness of our approach in real"
            }
        },
        "author_data": {
            "364b8a73-e1ab-4960-afb0-e711380330ba": {
                "pk": "364b8a73-e1ab-4960-afb0-e711380330ba",
                "name": "Haizhong Zheng",
                "collaborators": [
                    "Atul Prakash",
                    "Jiachen Sun",
                    "Ziqi Zhang",
                    "Honglak Lee",
                    "Qingzhao Zhang",
                    "Z. Morley Mao",
                    "Chaowei Xiao",
                    "Bhavya Kailkhura",
                    "Earlence Fernandes",
                    "Rui Liu"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Interpretability",
                    "Data Selection",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Understanding and Diagnosing Vulnerability under Adversarial Attacks",
                        "abstract": "Deep Neural Networks (DNNs) are known to be vulnerable to adversarial attacks. Currently, there is no clear insight into how slight perturbations cause such a large difference in classification results and how we can design a more robust model architecture. In this work, we propose a novel interpretability method, InterpretGAN, to generate explanations for features used for classification in latent variables. Interpreting the classification process of adversarial examples exposes how adversarial perturbations influence features layer by layer as well as which features are modified by perturbations. Moreover, we design the first diagnostic method to quantify the vulnerability contributed by each layer, which can be used to identify vulnerable parts of model architectures. The diagnostic results show that the layers introducing more information loss tend to be more vulnerable than other layers. Based on the findings, our evaluation results on MNIST and CIFAR10 datasets suggest that average pooling layers, with lower information loss, are more robust than max pooling layers for the network architectures studied in this paper."
                    },
                    {
                        "title": "Analyzing the Interpretability Robustness of Self-Explaining Models",
                        "abstract": "Recently, interpretable models called self-explaining models (SEMs) have been proposed with the goal of providing interpretability robustness. We evaluate the interpretability robustness of SEMs and show that explanations provided by SEMs as currently proposed are not robust to adversarial inputs. Specifically, we successfully created adversarial inputs that do not change the model outputs but cause significant changes in the explanations. We find that even though current SEMs use stable co-efficients for mapping explanations to output labels, they do not consider the robustness of the first stage of the model that creates interpretable basis concepts from the input, leading to non-robust explanations. Our work makes a case for future work to start examining how to generate interpretable basis concepts in a robust way."
                    },
                    {
                        "title": "Coverage-centric Coreset Selection for High Pruning Rates",
                        "abstract": "One-shot coreset selection aims to select a representative subset of the training data, given a pruning rate, that can later be used to train future models while retaining high accuracy. State-of-the-art coreset selection methods pick the highest importance examples based on an importance metric and are found to perform well at low pruning rates. However, at high pruning rates, they suffer from a catastrophic accuracy drop, performing worse than even random sampling. This paper explores the reasons behind this accuracy drop both theoretically and empirically. We first propose a novel metric to measure the coverage of a dataset on a specific distribution by extending the classical geometric set cover problem to a distribution cover problem. This metric helps explain why coresets selected by SOTA methods at high pruning rates perform poorly compared to random sampling because of worse data coverage. We then propose a novel one-shot coreset selection method, Coverage-centric Coreset Selection (CCS), that jointly considers overall data coverage upon a distribution as well as the importance of each example. We evaluate CCS on five datasets and show that, at high pruning rates (e.g., 90%), it achieves significantly better accuracy than previous SOTA methods (e.g., at least 19.56% higher on CIFAR10) as well as random selection (e.g., 7.04% higher on CIFAR10) and comparable accuracy at low pruning rates. We make our code publicly available at https://github.com/haizhongzheng/Coverage-centric-coreset-selection."
                    },
                    {
                        "title": "Efficient Adversarial Training with Transferable Adversarial Examples",
                        "abstract": "Adversarial training is an effective defense method to protect classification models against adversarial attacks. However, one limitation of this approach is that it can require orders of magnitude additional training time due to high cost of generating strong adversarial examples during training. In this paper, we first show that there is high transferability between models from neighboring epochs in the same training process, i.e., adversarial examples from one epoch continue to be adversarial in subsequent epochs. Leveraging this property, we propose a novel method, Adversarial Training with Transferable Adversarial Examples (ATTA), that can enhance the robustness of trained models and greatly improve the training efficiency by accumulating adversarial perturbations through epochs. Compared to state-of-the-art adversarial training methods, ATTA enhances adversarial accuracy by up to 7.2% on CIFAR10 and requires 12~14x less training time on MNIST and CIFAR10 datasets with comparable model robustness."
                    },
                    {
                        "title": "Robust Classification using Robust Feature Augmentation",
                        "abstract": "Existing deep neural networks, say for image classification, have been shown to be vulnerable to adversarial images that can cause a DNN misclassification, without any perceptible change to an image. In this work, we propose shock absorbing robust features such as binarization, e.g., rounding, and group extraction, e.g., color or shape, to augment the classification pipeline, resulting in more robust classifiers. Experimentally, we show that augmenting ML models with these techniques leads to improved overall robustness on adversarial inputs as well as significant improvements in training time. On the MNIST dataset, we achieved 14x speedup in training time to obtain 90% adversarial accuracy com-pared to the state-of-the-art adversarial training method of Madry et al., as well as retained higher adversarial accuracy over a broader range of attacks. We also find robustness improvements on traffic sign classification using robust feature augmentation. Finally, we give theoretical insights for why one can expect robust feature augmentation to reduce adversarial input space"
                    },
                    {
                        "title": "CALICO: Self-Supervised Camera-LiDAR Contrastive Pre-training for BEV Perception",
                        "abstract": "Perception is crucial in the realm of autonomous driving systems, where bird's eye view (BEV)-based architectures have recently reached state-of-the-art performance. The desirability of self-supervised representation learning stems from the expensive and laborious process of annotating 2D and 3D data. Although previous research has investigated pretraining methods for both LiDAR and camera-based 3D object detection, a unified pretraining framework for multimodal BEV perception is missing. In this study, we introduce CALICO, a novel framework that applies contrastive objectives to both LiDAR and camera backbones. Specifically, CALICO incorporates two stages: point-region contrast (PRC) and region-aware distillation (RAD). PRC better balances the region- and scene-level representation learning on the LiDAR modality and offers significant performance improvement compared to existing methods. RAD effectively achieves contrastive distillation on our self-trained teacher model. CALICO's efficacy is substantiated by extensive evaluations on 3D object detection and BEV map segmentation tasks, where it delivers significant performance improvements. Notably, CALICO outperforms the baseline method by 10.5% and 8.6% on NDS and mAP. Moreover, CALICO boosts the robustness of multimodal 3D object detection against adversarial attacks and corruption. Additionally, our framework can be tailored to different backbones and heads, positioning it as a promising approach for multimodal BEV perception."
                    },
                    {
                        "title": "Leveraging Hierarchical Feature Sharing for Efficient Dataset Condensation",
                        "abstract": "Given a real-world dataset, data condensation (DC) aims to synthesize a small synthetic dataset that captures the knowledge of a natural dataset while being usable for training models with comparable accuracy. Recent works propose to enhance DC with data parameterization, which condenses data into very compact parameterized data containers instead of images. The intuition behind data parameterization is to encode shared features of images to avoid additional storage costs. In this paper, we recognize that images share common features in a hierarchical way due to the inherent hierarchical structure of the classification system, which is overlooked by current data parameterization methods. To better align DC with this hierarchical nature and encourage more efficient information sharing inside data containers, we propose a novel data parameterization architecture, Hierarchical Memory Network (HMN). HMN stores condensed data in a three-tier structure, representing the dataset-level, class-level, and instance-level features. Another helpful property of the hierarchical architecture is that HMN naturally ensures good independence among images despite achieving information sharing. This enables instance-level pruning for HMN to reduce redundant information, thereby further minimizing redundancy and enhancing performance. We evaluate HMN on five public datasets and show that our proposed method outperforms all baselines."
                    },
                    {
                        "title": "ELFS: Enhancing Label-Free Coreset Selection via Clustering-based Pseudo-Labeling",
                        "abstract": "High-quality human-annotated data is crucial for modern deep learning pipelines, yet the human annotation process is both costly and time-consuming. Given a constrained human labeling budget, selecting an informative and representative data subset for labeling can significantly reduce human annotation effort. Well-performing state-of-the-art (SOTA) coreset selection methods require ground-truth labels over the whole dataset, failing to reduce the human labeling burden. Meanwhile, SOTA label-free coreset selection methods deliver inferior performance due to poor geometry-based scores. In this paper, we introduce ELFS, a novel label-free coreset selection method. ELFS employs deep clustering to estimate data difficulty scores without ground-truth labels. Furthermore, ELFS uses a simple but effective double-end pruning method to mitigate bias on calculated scores, which further improves the performance on selected coresets. We evaluate ELFS on five vision benchmarks and show that ELFS consistently outperforms SOTA label-free baselines. For instance, at a 90% pruning rate, ELFS surpasses the best-performing baseline by 5.3% on CIFAR10 and 7.1% on CIFAR100. Moreover, ELFS even achieves comparable performance to supervised coreset selection at low pruning rates (e.g., 30% and 50%) on CIFAR10 and ImageNet-1K."
                    },
                    {
                        "title": "Smoke Screener or Straight Shooter: Detecting Elite Sybil Attacks in User-Review Social Networks",
                        "abstract": "Popular User-Review Social Networks (URSNs)---such as Dianping, Yelp, and Amazon---are often the targets of reputation attacks in which fake reviews are posted in order to boost or diminish the ratings of listed products and services. These attacks often emanate from a collection of accounts, called Sybils, which are collectively managed by a group of real users. A new advanced scheme, which we term elite Sybil attacks, recruits organically highly-rated accounts to generate seemingly-trustworthy and realistic-looking reviews. These elite Sybil accounts taken together form a large-scale sparsely-knit Sybil network for which existing Sybil fake-review defense systems are unlikely to succeed. In this paper, we conduct the first study to define, characterize, and detect elite Sybil attacks. We show that contemporary elite Sybil attacks have a hybrid architecture, with the first tier recruiting elite Sybil workers and distributing tasks by Sybil organizers, and with the second tier posting fake reviews for profit by elite Sybil workers. We design ElsieDet, a three-stage Sybil detection scheme, which first separates out suspicious groups of users, then identifies the campaign windows, and finally identifies elite Sybil users participating in the campaigns. We perform a large-scale empirical study on ten million reviews from Dianping, by far the most popular URSN service in China. Our results show that reviews from elite Sybil users are more spread out temporally, craft more convincing reviews, and have higher filter bypass rates. We also measure the impact of Sybil campaigns on various industries (such as cinemas, hotels, restaurants) as well as chain stores, and demonstrate that monitoring elite Sybil users over time can provide valuable early alerts against Sybil campaigns."
                    },
                    {
                        "title": "Adaptive Skeleton Graph Decoding",
                        "abstract": "Large language models (LLMs) have seen significant adoption for natural language tasks, owing their success to massive numbers of model parameters (e.g., 70B+); however, LLM inference incurs significant computation and memory costs. Recent approaches propose parallel decoding strategies, such as Skeleton-of-Thought (SoT), to improve performance by breaking prompts down into sub-problems that can be decoded in parallel; however, they often suffer from reduced response quality. Our key insight is that we can request additional information, specifically dependencies and difficulty, when generating the sub-problems to improve both response quality and performance. In this paper, we propose Skeleton Graph Decoding (SGD), which uses dependencies exposed between sub-problems to support information forwarding between dependent sub-problems for improved quality while exposing parallelization opportunities for decoding independent sub-problems. Additionally, we leverage difficulty estimates for each sub-problem to select an appropriately-sized model, improving performance without significantly reducing quality. Compared to standard autoregressive generation and SoT, SGD achieves a 1.69x speedup while improving quality by up to 51%."
                    }
                ]
            },
            "0379cc7d-1937-4cd2-85e9-e6a84791d8e1": {
                "pk": "0379cc7d-1937-4cd2-85e9-e6a84791d8e1",
                "name": "Xiaoyan Bai",
                "collaborators": [
                    "Andrew Lee",
                    "Itamar Pres",
                    "Martin Wattenberg",
                    "Jonathan K. Kummerfeld",
                    "Rada Mihalcea",
                    "Aobo Kong",
                    "Shiwan Zhao",
                    "Hao Chen",
                    "Qicheng Li",
                    "Yong Qin"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Alignment Algorithms",
                    "Keyphrase Extraction"
                ],
                "publications": [
                    {
                        "title": "A Mechanistic Understanding of Alignment Algorithms: A Case Study on DPO and Toxicity",
                        "abstract": "While alignment algorithms are now commonly used to tune pre-trained language models towards a user's preferences, we lack explanations for the underlying mechanisms in which models become ``aligned'', thus making it difficult to explain phenomena like jailbreaks. In this work we study a popular algorithm, direct preference optimization (DPO), and the mechanisms by which it reduces toxicity. Namely, we first study how toxicity is represented and elicited in a pre-trained language model, GPT2-medium. We then apply DPO with a carefully crafted pairwise dataset to reduce toxicity. We examine how the resulting model averts toxic outputs, and find that capabilities learned from pre-training are not removed, but rather bypassed. We use this insight to demonstrate a simple method to un-align the model, reverting it back to its toxic behavior."
                    },
                    {
                        "title": "PromptRank: Unsupervised Keyphrase Extraction Using Prompt",
                        "abstract": "The keyphrase extraction task refers to the automatic selection of phrases from a given document to summarize its core content. State-of-the-art (SOTA) performance has recently been achieved by embedding-based algorithms, which rank candidates according to how similar their embeddings are to document embeddings. However, such solutions either struggle with the document and candidate length discrepancies or fail to fully utilize the pre-trained language model (PLM) without further fine-tuning. To this end, in this paper, we propose a simple yet effective unsupervised approach, PromptRank, based on the PLM with an encoder-decoder architecture. Specifically, PromptRank feeds the document into the encoder and calculates the probability of generating the candidate with a designed prompt by the decoder. We extensively evaluate the proposed PromptRank on six widely used benchmarks. PromptRank outperforms the SOTA approach MDERank, improving the F1 score relatively by 34.18%, 24.87%, and 17.57% for 5, 10, and 15 returned results, respectively. This demonstrates the great potential of using prompt for unsupervised keyphrase extraction. We release our code at https://github.com/HLT-NLP/PromptRank."
                    }
                ]
            },
            "8bf08aef-68be-4a72-99f8-97338382b7bc": {
                "pk": "8bf08aef-68be-4a72-99f8-97338382b7bc",
                "name": "Xueshen Liu",
                "collaborators": [
                    "Shuowei Jin",
                    "Qingzhao Zhang",
                    "Z. Morley Mao"
                ],
                "domain": [
                    "Large Language Models",
                    "Prefix Caching",
                    "Performance Optimization",
                    "Computational Efficiency"
                ],
                "publications": [
                    {
                        "title": "Compute Or Load KV Cache? Why Not Both?",
                        "abstract": "Recent advancements in Large Language Models (LLMs) have significantly increased context window sizes, enabling sophisticated applications but also introducing substantial computational overheads, particularly computing key-value (KV) cache in the prefill stage. Prefix caching has emerged to save GPU power in this scenario, which saves KV cache at disks and reuse them across multiple queries. However, traditional prefix caching mechanisms often suffer from substantial latency because the speed of loading KV cache from disks to GPU memory is bottlenecked by the throughput of I/O devices. To optimize the latency of long-context prefill, we propose Cake, a novel KV cache loader, which employs a bidirectional parallelized KV cache generation strategy. Upon receiving a prefill task, Cake simultaneously and dynamically loads saved KV cache from prefix cache locations and computes KV cache on local GPUs, maximizing the utilization of available computation and I/O bandwidth resources. Additionally, Cake automatically adapts to diverse system statuses without manual parameter. tuning. In experiments on various prompt datasets, GPUs, and I/O devices, Cake offers up to 68.1% Time To First Token (TTFT) reduction compare with compute-only method and 94.6% TTFT reduction compare with I/O-only method."
                    }
                ]
            },
            "b43ca0f3-08c4-4406-9c16-3ab787cd8f35": {
                "pk": "b43ca0f3-08c4-4406-9c16-3ab787cd8f35",
                "name": "Z. Morley Mao",
                "collaborators": [
                    "Qingzhao Zhang",
                    "Shuowei Jin",
                    "Jiachen Sun",
                    "Qi Alfred Chen",
                    "David R. Bild",
                    "Yue Liu",
                    "Robert P. Dick",
                    "Dan S. Wallach",
                    "Minkyoung Cho",
                    "Yulong Cao"
                ],
                "domain": [
                    "Large Language Models",
                    "Adversarial Robustness",
                    "Autonomous Vehicles",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Eagle: Efficient Training-Free Router for Multi-LLM Inference",
                        "abstract": "The proliferation of Large Language Models (LLMs) with varying capabilities and costs has created a need for efficient model selection in AI systems. LLM routers address this need by dynamically choosing the most suitable model for a given query based on task requirements and budget constraints. However, existing routers face challenges in scalability and real-time adaptation, particularly in high-volume online environments. We present Eagle, a novel LLM routing approach that combines global and local ELO ranking modules to overcome these limitations. By evaluating both general and specialized LLM abilities, Eagle provides a scalable, training-free solution that enhances model selection quality while reducing computational overhead. Our experiments across multiple datasets show Eagle consistently outperforms baseline methods, with improvements of up to 23.52 percent in Area Under Curve (AUC) scores. Moreover, Eagle demonstrates remarkable efficiency, requiring only 1/20 of baseline methods' time for initialization and 100 to 200 times faster incremental updates in online scenarios, making it well-suited for dynamic, high-volume online serving environments."
                    },
                    {
                        "title": "Safeguard is a Double-edged Sword: Denial-of-service Attack on Large Language Models",
                        "abstract": "Safety is a paramount concern of large language models (LLMs) in their open deployment. To this end, safeguard methods aim to enforce the ethical and responsible use of LLMs through safety alignment or guardrail mechanisms. However, we found that the malicious attackers could exploit false positives of safeguards, i.e., fooling the safeguard model to block safe content mistakenly, leading to a new denial-of-service (DoS) attack on LLMs. Specifically, by software or phishing attacks on user client software, attackers insert a short, seemingly innocuous adversarial prompt into to user prompt templates in configuration files; thus, this prompt appears in final user requests without visibility in the user interface and is not trivial to identify. By designing an optimization process that utilizes gradient and attention information, our attack can automatically generate seemingly safe adversarial prompts, approximately only 30 characters long, that universally block over 97\\% of user requests on Llama Guard 3. The attack presents a new dimension of evaluating LLM safeguards focusing on false positives, fundamentally different from the classic jailbreak."
                    },
                    {
                        "title": "Performance Analysis of Location Profile Routing",
                        "abstract": "We propose using the predictability of human motion to eliminate the overhead of distributed location services in human-carried MANETs, dubbing the technique location profile routing. This method outperforms the Geographic Hashing Location Service when nodes change locations 2x more frequently than they initiate connections (e.g., start new TCP streams), as in applications like text- and instant-messaging. Prior characterizations of human mobility are used to show that location profile routing achieves a 93% delivery ratio with a 1.75x first-packet latency increase relative to an oracle location service."
                    },
                    {
                        "title": "Compute Or Load KV Cache? Why Not Both?",
                        "abstract": "Recent advancements in Large Language Models (LLMs) have significantly increased context window sizes, enabling sophisticated applications but also introducing substantial computational overheads, particularly computing key-value (KV) cache in the prefill stage. Prefix caching has emerged to save GPU power in this scenario, which saves KV cache at disks and reuse them across multiple queries. However, traditional prefix caching mechanisms often suffer from substantial latency because the speed of loading KV cache from disks to GPU memory is bottlenecked by the throughput of I/O devices. To optimize the latency of long-context prefill, we propose Cake, a novel KV cache loader, which employs a bidirectional parallelized KV cache generation strategy. Upon receiving a prefill task, Cake simultaneously and dynamically loads saved KV cache from prefix cache locations and computes KV cache on local GPUs, maximizing the utilization of available computation and I/O bandwidth resources. Additionally, Cake automatically adapts to diverse system statuses without manual parameter. tuning. In experiments on various prompt datasets, GPUs, and I/O devices, Cake offers up to 68.1% Time To First Token (TTFT) reduction compare with compute-only method and 94.6% TTFT reduction compare with I/O-only method."
                    },
                    {
                        "title": "ADoPT: LiDAR Spoofing Attack Detection Based on Point-Level Temporal Consistency",
                        "abstract": "Deep neural networks (DNNs) are increasingly integrated into LiDAR (Light Detection and Ranging)-based perception systems for autonomous vehicles (AVs), requiring robust performance under adversarial conditions. We aim to address the challenge of LiDAR spoofing attacks, where attackers inject fake objects into LiDAR data and fool AVs to misinterpret their environment and make erroneous decisions. However, current defense algorithms predominantly depend on perception outputs (i.e., bounding boxes) thus face limitations in detecting attackers given the bounding boxes are generated by imperfect perception models processing limited points, acquired based on the ego vehicle's viewpoint. To overcome these limitations, we propose a novel framework, named ADoPT (Anomaly Detection based on Point-level Temporal consistency), which quantitatively measures temporal consistency across consecutive frames and identifies abnormal objects based on the coherency of point clusters. In our evaluation using the nuScenes dataset, our algorithm effectively counters various LiDAR spoofing attacks, achieving a low (< 10%) false positive ratio (FPR) and high (> 85%) true positive ratio (TPR), outperforming existing state-of-the-art defense methods, CARLO and 3D-TC2. Furthermore, our evaluation demonstrates the promising potential for accurate attack detection across various road environments."
                    },
                    {
                        "title": "Sensor Adversarial Traits: Analyzing Robustness of 3D Object Detection Sensor Fusion Models",
                        "abstract": "A critical aspect of autonomous vehicles (AVs) is the object detection stage, which is increasingly being performed with sensor fusion models: multimodal 3D object detection models which utilize both 2D RGB image data and 3D data from a LIDAR sensor as inputs. In this work, we perform the first study to analyze the robustness of a high-performance, open source sensor fusion model architecture towards adversarial attacks and challenge the popular belief that the use of additional sensors automatically mitigate the risk of adversarial attacks. We find that despite the use of a LIDAR sensor, the model is vulnerable to our purposefully crafted image-based adversarial attacks including disappearance, universal patch, and spoofing. After identifying the underlying reason, we explore some potential defenses and provide some recommendations for improved sensor fusion models."
                    },
                    {
                        "title": "On Adversarial Robustness of Trajectory Prediction for Autonomous Vehicles",
                        "abstract": "Trajectory prediction is a critical component for autonomous vehicles (AVs) to perform safe planning and navigation. However, few studies have analyzed the adversarial robustness of trajectory prediction or investigated whether the worst-case prediction can still lead to safe planning. To bridge this gap, we study the adversarial robustness of trajectory prediction models by proposing a new adversarial attack that perturbs normal vehicle trajectories to maximize the prediction error. Our experiments on three models and three datasets show that the adversarial prediction increases the prediction error by more than 150%. Our case studies show that if an adversary drives a vehicle close to the target AV following the adversarial trajectory, the AV may make an inaccurate prediction and even make unsafe driving decisions. We also explore possible mitigation techniques via data augmentation and trajectory smoothing. The implementation is open source at https://github.com/zqzqz/AdvTrajectoryPrediction."
                    },
                    {
                        "title": "Benchmarking Robustness of 3D Point Cloud Recognition Against Common Corruptions",
                        "abstract": "Deep neural networks on 3D point cloud data have been widely used in the real world, especially in safety-critical applications. However, their robustness against corruptions is less studied. In this paper, we present ModelNet40-C, the first comprehensive benchmark on 3D point cloud corruption robustness, consisting of 15 common and realistic corruptions. Our evaluation shows a significant gap between the performances on ModelNet40 and ModelNet40-C for state-of-the-art (SOTA) models. To reduce the gap, we propose a simple but effective method by combining PointCutMix-R and TENT after evaluating a wide range of augmentation and test-time adaptation strategies. We identify a number of critical insights for future studies on corruption robustness in point cloud recognition. For instance, we unveil that Transformer-based architectures with proper training recipes achieve the strongest robustness. We hope our in-depth analysis will motivate the development of robust training strategies or architecture designs in the 3D point cloud domain. Our codebase and dataset are included in https://github.com/jiachens/ModelNet40-C"
                    },
                    {
                        "title": "PointDP: Diffusion-driven Purification against Adversarial Attacks on 3D Point Cloud Recognition",
                        "abstract": "3D Point cloud is becoming a critical data representation in many real-world applications like autonomous driving, robotics, and medical imaging. Although the success of deep learning further accelerates the adoption of 3D point clouds in the physical world, deep learning is notorious for its vulnerability to adversarial attacks. In this work, we first identify that the state-of-the-art empirical defense, adversarial training, has a major limitation in applying to 3D point cloud models due to gradient obfuscation. We further propose PointDP, a purification strategy that leverages diffusion models to defend against 3D adversarial attacks. We extensively evaluate PointDP on six representative 3D point cloud architectures, and leverage 10+ strong and adaptive attacks to demonstrate its lower-bound robustness. Our evaluation shows that PointDP achieves significantly better robustness than state-of-the-art purification methods under strong attacks. Results of certified defenses on randomized smoothing combined with PointDP will be included in the near future."
                    },
                    {
                        "title": "Rethinking the Backdoor Attacks' Triggers: A Frequency Perspective",
                        "abstract": "Backdoor attacks have been considered a severe security threat to deep learning. Such attacks can make models perform abnormally on inputs with predefined triggers and still retain state-of-the-art performance on clean data. While backdoor attacks have been thoroughly investigated in the image domain from both attackers' and defenders' sides, an analysis in the frequency domain has been missing thus far.   This paper first revisits existing backdoor triggers from a frequency perspective and performs a comprehensive analysis. Our results show that many current backdoor attacks exhibit severe high-frequency artifacts, which persist across different datasets and resolutions. We further demonstrate these high-frequency artifacts enable a simple way to detect existing backdoor triggers at a detection rate of 98.50% without prior knowledge of the attack details and the target model. Acknowledging previous attacks' weaknesses, we propose a practical way to create smooth backdoor triggers without high-frequency artifacts and study their detectability. We show that existing defense works can benefit by incorporating these smooth triggers into their design consideration. Moreover, we show that the detector tuned over stronger smooth triggers can generalize well to unseen weak smooth triggers. In short, our work emphasizes the importance of considering frequency analysis when designing both backdoor attacks and defenses in deep learning."
                    },
                    {
                        "title": "Partial-Information, Longitudinal Cyber Attacks on LiDAR in Autonomous Vehicles",
                        "abstract": "What happens to an autonomous vehicle (AV) if its data are adversarially compromised? Prior security studies have addressed this question through mostly unrealistic threat models, with limited practical relevance, such as white-box adversarial learning or nanometer-scale laser aiming and spoofing. With growing evidence that cyber threats pose real, imminent danger to AVs and cyber-physical systems (CPS) in general, we present and evaluate a novel AV threat model: a cyber-level attacker capable of disrupting sensor data but lacking any situational awareness. We demonstrate that even though the attacker has minimal knowledge and only access to raw data from a single sensor (i.e., LiDAR), she can design several attacks that critically compromise perception and tracking in multi-sensor AVs. To mitigate vulnerabilities and advance secure architectures in AVs, we introduce two improvements for security-aware fusion: a probabilistic data-asymmetry monitor and a scalable track-to-track fusion of 3D LiDAR and monocular detections (T2T-3DLM); we demonstrate that the approaches significantly reduce attack effectiveness. To support objective safety and security evaluations in AVs, we release our security evaluation platform, AVsec, which is built on security-relevant metrics to benchmark AVs on gold-standard longitudinal AV datasets and AV simulators."
                    },
                    {
                        "title": "Aggregate Characterization of User Behavior in Twitter and Analysis of the Retweet Graph",
                        "abstract": "Most previous analysis of Twitter user behavior is focused on individual information cascades and the social followers graph. We instead study aggregate user behavior and the retweet graph with a focus on quantitative descriptions. We find that the lifetime tweet distribution is a type-II discrete Weibull stemming from a power law hazard function, the tweet rate distribution, although asymptotically power law, exhibits a lognormal cutoff over finite sample intervals, and the inter-tweet interval distribution is power law with exponential cutoff. The retweet graph is small-world and scale-free, like the social graph, but is less disassortative and has much stronger clustering. These differences are consistent with it better capturing the real-world social relationships of and trust between users. Beyond just understanding and modeling human communication patterns and social networks, applications for alternative, decentralized microblogging systems-both predicting real-word performance and detecting spam-are discussed."
                    },
                    {
                        "title": "The Mason Test: A Defense Against Sybil Attacks in Wireless Networks Without Trusted Authorities",
                        "abstract": "Wireless networks are vulnerable to Sybil attacks, in which a malicious node poses as many identities in order to gain disproportionate influence. Many defenses based on spatial variability of wireless channels exist, but depend either on detailed, multi-tap channel estimation - something not exposed on commodity 802.11 devices - or valid RSSI observations from multiple trusted sources, e.g., corporate access points - something not directly available in ad hoc and delay-tolerant networks with potentially malicious neighbors. We extend these techniques to be practical for wireless ad hoc networks of commodity 802.11 devices. Specifically, we propose two efficient methods for separating the valid RSSI observations of behaving nodes from those falsified by malicious participants. Further, we note that prior signalprint methods are easily defeated by mobile attackers and develop an appropriate challenge-response defense. Finally, we present the Mason test, the first implementation of these techniques for ad hoc and delay-tolerant networks of commodity 802.11 devices. We illustrate its performance in several real-world scenarios."
                    },
                    {
                        "title": "Impact Evaluation of Falsified Data Attacks on Connected Vehicle Based Traffic Signal Control",
                        "abstract": "Connected vehicle (CV) technology enables data exchange between vehicles and transportation infrastructure and therefore has great potentials to improve current traffic signal control systems. However, this connectivity might also bring cyber security concerns. As the first step in investigating the cyber security of CV-based traffic signal control (CV-TSC) systems, potential cyber threats need to be identified and corresponding impact needs to be evaluated. In this paper, we aim to evaluate the impact of cyber attacks on CV-TSC systems by considering a realistic attack scenario in which the control logic of a CV-TSC system is unavailable to attackers. Our threat model presumes that an attacker may learn the control logic using a surrogate model. Based on the surrogate model, the attacker may launch falsified data attacks to influence signal control decisions. In the case study, we realistically evaluate the impact of falsified data attacks on an existing CV-TSC system (i.e., I-SIG)."
                    },
                    {
                        "title": "On Data Fabrication in Collaborative Vehicular Perception: Attacks and Countermeasures",
                        "abstract": "Collaborative perception, which greatly enhances the sensing capability of connected and autonomous vehicles (CAVs) by incorporating data from external resources, also brings forth potential security risks. CAVs' driving decisions rely on remote untrusted data, making them susceptible to attacks carried out by malicious participants in the collaborative perception system. However, security analysis and countermeasures for such threats are absent. To understand the impact of the vulnerability, we break the ground by proposing various real-time data fabrication attacks in which the attacker delivers crafted malicious data to victims in order to perturb their perception results, leading to hard brakes or increased collision risks. Our attacks demonstrate a high success rate of over 86% on high-fidelity simulated scenarios and are realizable in real-world experiments. To mitigate the vulnerability, we present a systematic anomaly detection approach that enables benign vehicles to jointly reveal malicious fabrication. It detects 91.5% of attacks with a false positive rate of 3% in simulated scenarios and significantly mitigates attack impacts in real-world scenarios."
                    },
                    {
                        "title": "VPA: Fully Test-Time Visual Prompt Adaptation",
                        "abstract": "Textual prompt tuning has demonstrated significant performance improvements in adapting natural language processing models to a variety of downstream tasks by treating hand-engineered prompts as trainable parameters. Inspired by the success of textual prompting, several studies have investigated the efficacy of visual prompt tuning. In this work, we present Visual Prompt Adaptation (VPA), the first framework that generalizes visual prompting with test-time adaptation. VPA introduces a small number of learnable tokens, enabling fully test-time and storage-efficient adaptation without necessitating source-domain information. We examine our VPA design under diverse adaptation settings, encompassing single-image, batched-image, and pseudo-label adaptation. We evaluate VPA on multiple tasks, including out-of-distribution (OOD) generalization, corruption robustness, and domain adaptation. Experimental results reveal that VPA effectively enhances OOD generalization by 3.3% across various models, surpassing previous test-time approaches. Furthermore, we show that VPA improves corruption robustness by 6.5% compared to strong baselines. Finally, we demonstrate that VPA also boosts domain adaptation performance by relatively 5.2%. Our VPA also exhibits marked effectiveness in improving the robustness of zero-shot recognition for vision-language models."
                    },
                    {
                        "title": "QUIC is not Quick Enough over Fast Internet",
                        "abstract": "QUIC is expected to be a game-changer in improving web application performance. In this paper, we conduct a systematic examination of QUIC's performance over high-speed networks. We find that over fast Internet, the UDP+QUIC+HTTP/3 stack suffers a data rate reduction of up to 45.2% compared to the TCP+TLS+HTTP/2 counterpart. Moreover, the performance gap between QUIC and HTTP/2 grows as the underlying bandwidth increases. We observe this issue on lightweight data transfer clients and major web browsers (Chrome, Edge, Firefox, Opera), on different hosts (desktop, mobile), and over diverse networks (wired broadband, cellular). It affects not only file transfers, but also various applications such as video streaming (up to 9.8% video bitrate reduction) and web browsing. Through rigorous packet trace analysis and kernel- and user-space profiling, we identify the root cause to be high receiver-side processing overhead, in particular, excessive data packets and QUIC's user-space ACKs. We make concrete recommendations for mitigating the observed performance issues."
                    },
                    {
                        "title": "Adaptive Skeleton Graph Decoding",
                        "abstract": "Large language models (LLMs) have seen significant adoption for natural language tasks, owing their success to massive numbers of model parameters (e.g., 70B+); however, LLM inference incurs significant computation and memory costs. Recent approaches propose parallel decoding strategies, such as Skeleton-of-Thought (SoT), to improve performance by breaking prompts down into sub-problems that can be decoded in parallel; however, they often suffer from reduced response quality. Our key insight is that we can request additional information, specifically dependencies and difficulty, when generating the sub-problems to improve both response quality and performance. In this paper, we propose Skeleton Graph Decoding (SGD), which uses dependencies exposed between sub-problems to support information forwarding between dependent sub-problems for improved quality while exposing parallelization opportunities for decoding independent sub-problems. Additionally, we leverage difficulty estimates for each sub-problem to select an appropriately-sized model, improving performance without significantly reducing quality. Compared to standard autoregressive generation and SoT, SGD achieves a 1.69x speedup while improving quality by up to 51%."
                    },
                    {
                        "title": "AutoSpec: Automated Generation of Neural Network Specifications",
                        "abstract": "The increasing adoption of neural networks in learning-augmented systems highlights the importance of model safety and robustness, particularly in safety-critical domains. Despite progress in the formal verification of neural networks, current practices require users to manually define model specifications -- properties that dictate expected model behavior in various scenarios. This manual process, however, is prone to human error, limited in scope, and time-consuming. In this paper, we introduce AutoSpec, the first framework to automatically generate comprehensive and accurate specifications for neural networks in learning-augmented systems. We also propose the first set of metrics for assessing the accuracy and coverage of model specifications, establishing a benchmark for future comparisons. Our evaluation across four distinct applications shows that AutoSpec outperforms human-defined specifications as well as two baseline approaches introduced in this study."
                    },
                    {
                        "title": "Cocoon: Robust Multi-Modal Perception with Uncertainty-Aware Sensor Fusion",
                        "abstract": "An important paradigm in 3D object detection is the use of multiple modalities to enhance accuracy in both normal and challenging conditions, particularly for long-tail scenarios. To address this, recent studies have explored two directions of adaptive approaches: MoE-based adaptive fusion, which struggles with uncertainties arising from distinct object configurations, and late fusion for output-level adaptive fusion, which relies on separate detection pipelines and limits comprehensive understanding. In this work, we introduce Cocoon, an object- and feature-level uncertainty-aware fusion framework. The key innovation lies in uncertainty quantification for heterogeneous representations, enabling fair comparison across modalities through the introduction of a feature aligner and a learnable surrogate ground truth, termed feature impression. We also define a training objective to ensure that their relationship provides a valid metric for uncertainty quantification. Cocoon consistently outperforms existing static and adaptive methods in both normal and challenging conditions, including those with natural and artificial corruptions. Furthermore, we show the validity and efficacy of our uncertainty metric across diverse datasets."
                    }
                ]
            },
            "758378f6-1a31-4375-b462-df0d7de2d346": {
                "pk": "758378f6-1a31-4375-b462-df0d7de2d346",
                "name": "Beidi Chen",
                "collaborators": [
                    "Anshumali Shrivastava",
                    "Yuandong Tian",
                    "Zhuoming Chen",
                    "Anima Anandkumar",
                    "Tharun Medini",
                    "Yiping Wang",
                    "Simon Du",
                    "Jiawei Zhao",
                    "Avner May",
                    "Ruslan Svirschevski"
                ],
                "domain": [
                    "Machine Learning",
                    "Deep Learning",
                    "Natural Language Processing",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Revisiting Winner Take All (WTA) Hashing for Sparse Datasets",
                        "abstract": "WTA (Winner Take All) hashing has been successfully applied in many large scale vision applications. This hashing scheme was tailored to take advantage of the comparative reasoning (or order based information), which showed significant accuracy improvements. In this paper, we identify a subtle issue with WTA, which grows with the sparsity of the datasets. This issue limits the discriminative power of WTA. We then propose a solution for this problem based on the idea of Densification which provably fixes the issue. Our experiments show that Densified WTA Hashing outperforms Vanilla WTA both in image classification and retrieval tasks consistently and significantly."
                    },
                    {
                        "title": "Lsh-sampling Breaks the Computation Chicken-and-egg Loop in Adaptive Stochastic Gradient Estimation",
                        "abstract": "Stochastic Gradient Descent or SGD is the most popular optimization algorithm for large-scale problems. SGD estimates the gradient by uniform sampling with sample size one. There have been several other works that suggest faster epoch-wise convergence by using weighted non-uniform sampling for better gradient estimates. Unfortunately, the per-iteration cost of maintaining this adaptive distribution for gradient estimation is more than calculating the full gradient itself, which we call the chicken-and-the-egg loop. As a result, the false impression of faster convergence in iterations, in reality, leads to slower convergence in time. In this paper, we break this barrier by providing the first demonstration of a scheme, Locality sensitive hashing (LSH) sampled Stochastic Gradient Descent (LGD), which leads to superior gradient estimation while keeping the sampling cost per iteration similar to that of the uniform sampling. Such an algorithm is possible due to the sampling view of LSH, which came to light recently. As a consequence of superior and fast estimation, we reduce the running time of all existing gradient descent algorithms, that relies on gradient estimates including Adam, Ada-grad, etc. We demonstrate the effectiveness of our proposal with experiments on linear models as well as the non-linear BERT, which is a recent popular deep learning based language representation model."
                    },
                    {
                        "title": "Unique Entity Estimation with Application to the Syrian Conflict",
                        "abstract": "Entity resolution identifies and removes duplicate entities in large, noisy databases and has grown in both usage and new developments as a result of increased data availability. Nevertheless, entity resolution has tradeoffs regarding assumptions of the data generation process, error rates, and computational scalability that make it a difficult task for real applications. In this paper, we focus on a related problem of unique entity estimation, which is the task of estimating the unique number of entities and associated standard errors in a data set with duplicate entities. Unique entity estimation shares many fundamental challenges of entity resolution, namely, that the computational cost of all-to-all entity comparisons is intractable for large databases. To circumvent this computational barrier, we propose an efficient (near-linear time) estimation algorithm based on locality sensitive hashing. Our estimator, under realistic assumptions, is unbiased and has provably low variance compared to existing random sampling based approaches. In addition, we empirically show its superiority over the state-of-the-art estimators on three real applications. The motivation for our work is to derive an accurate estimate of the documented, identifiable deaths in the ongoing Syrian conflict. Our methodology, when applied to the Syrian data set, provides an estimate of $191,874 \\pm 1772$ documented, identifiable deaths, which is very close to the Human Rights Data Analysis Group (HRDAG) estimate of 191,369. Our work provides an example of challenges and efforts involved in solving a real, noisy challenging problem where modeling assumptions may not hold."
                    },
                    {
                        "title": "SOLAR: Sparse Orthogonal Learned and Random Embeddings",
                        "abstract": "Dense embedding models are commonly deployed in commercial search engines, wherein all the document vectors are pre-computed, and near-neighbor search (NNS) is performed with the query vector to find relevant documents. However, the bottleneck of indexing a large number of dense vectors and performing an NNS hurts the query time and accuracy of these models. In this paper, we argue that high-dimensional and ultra-sparse embedding is a significantly superior alternative to dense low-dimensional embedding for both query efficiency and accuracy. Extreme sparsity eliminates the need for NNS by replacing them with simple lookups, while its high dimensionality ensures that the embeddings are informative even when sparse. However, learning extremely high dimensional embeddings leads to blow up in the model size. To make the training feasible, we propose a partitioning algorithm that learns such high dimensional embeddings across multiple GPUs without any communication. This is facilitated by our novel asymmetric mixture of Sparse, Orthogonal, Learned and Random (SOLAR) Embeddings. The label vectors are random, sparse, and near-orthogonal by design, while the query vectors are learned and sparse. We theoretically prove that our way of one-sided learning is equivalent to learning both query and label embeddings. With these unique properties, we can successfully train 500K dimensional SOLAR embeddings for the tasks of searching through 1.6M books and multi-label classification on the three largest public datasets. We achieve superior precision and recall compared to the respective state-of-the-art baselines for each of the tasks with up to 10 times faster speed."
                    },
                    {
                        "title": "Prompt-prompted Adaptive Structured Pruning for Efficient LLM Generation",
                        "abstract": "With the development of transformer-based large language models (LLMs), they have been applied to many fields due to their remarkable utility, but this comes at a considerable computational cost at deployment. Fortunately, some methods such as pruning or constructing a mixture of experts (MoE) aim at exploiting sparsity in transformer feedforward (FF) blocks to gain boosts in speed and reduction in memory requirements. However, these techniques can be very costly and inflexible in practice, as they often require training or are restricted to specific types of architectures. To address this, we introduce GRIFFIN, a novel training-free and calibration-free method that selects unique FF experts at the sequence level for efficient generation across a plethora of LLMs with different non-ReLU activation functions. This is possible due to a critical observation that many trained LLMs naturally produce highly structured FF activation patterns within a sequence, which we call flocking. Despite our method's simplicity, we show with 50% of the FF parameters, GRIFFIN maintains the original model's performance with little to no degradation on a variety of classification and generation tasks, all while improving latency (e.g. 1.29$\\times$ and 1.25$\\times$ speed-ups in Gemma 7B and Llama 2 13B, respectively, on an NVIDIA L40). Code is available at https://github.com/hdong920/GRIFFIN."
                    },
                    {
                        "title": "LoCoCo: Dropping In Convolutions for Long Context Compression",
                        "abstract": "This paper tackles the memory hurdle of processing long context sequences in Large Language Models (LLMs), by presenting a novel approach, Dropping In Convolutions for Long Context Compression (LoCoCo). LoCoCo employs only a fixed-size Key-Value (KV) cache, and can enhance efficiency in both inference and fine-tuning stages. Diverging from prior methods that selectively drop KV pairs based on heuristics, LoCoCo leverages a data-driven adaptive fusion technique, blending previous KV pairs with incoming tokens to minimize the loss of contextual information and ensure accurate attention modeling. This token integration is achieved through injecting one-dimensional convolutional kernels that dynamically calculate mixing weights for each KV cache slot. Designed for broad compatibility with existing LLM frameworks, LoCoCo allows for straightforward \"drop-in\" integration without needing architectural modifications, while incurring minimal tuning overhead. Experiments demonstrate that LoCoCo maintains consistently outstanding performance across various context lengths and can achieve a high context compression rate during both inference and fine-tuning phases. During inference, we successfully compressed up to 3482 tokens into a 128-size KV cache, while retaining comparable performance to the full sequence - an accuracy improvement of up to 0.2791 compared to baselines at the same cache size. During post-training tuning, we also effectively extended the context length from 4K to 32K using a KV cache of fixed size 512, achieving performance similar to fine-tuning with entire sequences."
                    },
                    {
                        "title": "Scan and Snap: Understanding Training Dynamics and Token Composition in 1-layer Transformer",
                        "abstract": "Transformer architecture has shown impressive performance in multiple research domains and has become the backbone of many neural network models. However, there is limited understanding on how it works. In particular, with a simple predictive loss, how the representation emerges from the gradient \\emph{training dynamics} remains a mystery. In this paper, for 1-layer transformer with one self-attention layer plus one decoder layer, we analyze its SGD training dynamics for the task of next token prediction in a mathematically rigorous manner. We open the black box of the dynamic process of how the self-attention layer combines input tokens, and reveal the nature of underlying inductive bias. More specifically, with the assumption (a) no positional encoding, (b) long input sequence, and (c) the decoder layer learns faster than the self-attention layer, we prove that self-attention acts as a \\emph{discriminative scanning algorithm}: starting from uniform attention, it gradually attends more to distinct key tokens for a specific next token to be predicted, and pays less attention to common key tokens that occur across different next tokens. Among distinct tokens, it progressively drops attention weights, following the order of low to high co-occurrence between the key and the query token in the training set. Interestingly, this procedure does not lead to winner-takes-all, but decelerates due to a \\emph{phase transition} that is controllable by the learning rates of the two layers, leaving (almost) fixed token combination. We verify this \\textbf{\\emph{scan and snap}} dynamics on synthetic and real-world data (WikiText)."
                    },
                    {
                        "title": "Mini-Sequence Transformer: Optimizing Intermediate Memory for Long Sequences Training",
                        "abstract": "We introduce Mini-Sequence Transformer (MsT), a simple and effective methodology for highly efficient and accurate LLM training with extremely long sequences. MsT partitions input sequences and iteratively processes mini-sequences to reduce intermediate memory usage. Integrated with activation recomputation, it enables significant memory savings in both forward and backward passes. In experiments with the Llama3-8B model, with MsT, we measure no degradation in throughput or convergence even with 12x longer sequences than standard implementations. MsT is fully general, implementation-agnostic, and requires minimal code changes to integrate with existing LLM training frameworks. Integrated with the huggingface library, MsT successfully extends the maximum context length of Qwen, Mistral, and Gemma-2 by 12-24x."
                    },
                    {
                        "title": "Sirius: Contextual Sparsity with Correction for Efficient LLMs",
                        "abstract": "With the blossom of large language models (LLMs), inference efficiency becomes increasingly important. Various approximation methods are proposed to reduce the cost at inference time. Contextual Sparsity (CS) is appealing for its training-free nature and its ability to reach a higher compression ratio seemingly without quality degradation. However, after a comprehensive evaluation of contextual sparsity methods on various complex generation tasks, we find that although CS succeeds in prompt-understanding tasks, CS significantly degrades the model performance for reasoning, deduction, and knowledge-based tasks. Despite the gap in end-to-end accuracy, we observed that sparse models often share general problem-solving logic and require only a few token corrections to recover the original model performance. This paper introduces Sirius, an efficient correction mechanism, which significantly recovers CS models quality on reasoning tasks while maintaining its efficiency gain. Sirius is evaluated on 6 models with 8 difficult generation tasks in reasoning, math, and coding and shows consistent effectiveness and efficiency. Also, we carefully develop a system implementation for Sirius and show that Sirius achieves roughly 20% reduction in latency for 8B model on-chip and 35% reduction for 70B model offloading. We open-source our implementation of Sirius at https://github.com/Infini-AI-Lab/Sirius.git."
                    },
                    {
                        "title": "Memory Mosaics",
                        "abstract": "Memory Mosaics are networks of associative memories working in concert to achieve a prediction task of interest. Like transformers, memory mosaics possess compositional capabilities and in-context learning capabilities. Unlike transformers, memory mosaics achieve these capabilities in comparatively transparent ways. We demonstrate these capabilities on toy examples and we also show that memory mosaics perform as well or better than transformers on medium-scale language modeling tasks."
                    },
                    {
                        "title": "TriForce: Lossless Acceleration of Long Sequence Generation with Hierarchical Speculative Decoding",
                        "abstract": "With large language models (LLMs) widely deployed in long content generation recently, there has emerged an increasing demand for efficient long-sequence inference support. However, key-value (KV) cache, which is stored to avoid re-computation, has emerged as a critical bottleneck by growing linearly in size with the sequence length. Due to the auto-regressive nature of LLMs, the entire KV cache will be loaded for every generated token, resulting in low utilization of computational cores and high latency. While various compression methods for KV cache have been proposed to alleviate this issue, they suffer from degradation in generation quality. We introduce TriForce, a hierarchical speculative decoding system that is scalable for long sequence generation. This approach leverages the original model weights and dynamic sparse KV cache via retrieval as a draft model, which serves as an intermediate layer in the hierarchy and is further speculated by a smaller model to reduce its drafting latency. TriForce not only facilitates impressive speedups for Llama2-7B-128K, achieving up to 2.31$\\times$ on an A100 GPU but also showcases scalability in handling even longer contexts. For the offloading setting on two RTX 4090 GPUs, TriForce achieves 0.108s/token$\\unicode{x2014}$only half as slow as the auto-regressive baseline on an A100, which attains 7.78$\\times$ on our optimized offloading system. Additionally, TriForce performs 4.86$\\times$ than DeepSpeed-Zero-Inference on a single RTX 4090 GPU. TriForce's robustness is highlighted by its consistently outstanding performance across various temperatures. The code is available at https://github.com/Infini-AI-Lab/TriForce."
                    },
                    {
                        "title": "Fast Algorithms for a New Relaxation of Optimal Transport",
                        "abstract": "We introduce a new class of objectives for optimal transport computations of datasets in high-dimensional Euclidean spaces. The new objectives are parametrized by $\\rho \\geq 1$, and provide a metric space $\\mathcal{R}_{\\rho}(\\cdot, \\cdot)$ for discrete probability distributions in $\\mathbb{R}^d$. As $\\rho$ approaches $1$, the metric approaches the Earth Mover's distance, but for $\\rho$ larger than (but close to) $1$, admits significantly faster algorithms. Namely, for distributions $\\mu$ and $\\nu$ supported on $n$ and $m$ vectors in $\\mathbb{R}^d$ of norm at most $r$ and any $\\epsilon > 0$, we give an algorithm which outputs an additive $\\epsilon r$-approximation to $\\mathcal{R}_{\\rho}(\\mu, \\nu)$ in time $(n+m) \\cdot \\mathrm{poly}((nm)^{(\\rho-1)/\\rho} \\cdot 2^{\\rho / (\\rho-1)} / \\epsilon)$."
                    },
                    {
                        "title": "JoMA: Demystifying Multilayer Transformers via JOint Dynamics of MLP and Attention",
                        "abstract": "We propose Joint MLP/Attention (JoMA) dynamics, a novel mathematical framework to understand the training procedure of multilayer Transformer architectures. This is achieved by integrating out the self-attention layer in Transformers, producing a modified dynamics of MLP layers only. JoMA removes unrealistic assumptions in previous analysis (e.g., lack of residual connection) and predicts that the attention first becomes sparse (to learn salient tokens), then dense (to learn less salient tokens) in the presence of nonlinear activations, while in the linear case, it is consistent with existing works that show attention becomes sparse over time. We leverage JoMA to qualitatively explains how tokens are combined to form hierarchies in multilayer Transformers, when the input tokens are generated by a latent hierarchical generative model. Experiments on models trained from real-world dataset (Wikitext2/Wikitext103) and various pre-trained models (OPT, Pythia) verify our theoretical findings. Code can be found in https://github.com/facebookresearch/luckmatters/tree/yuandong3."
                    },
                    {
                        "title": "InRank: Incremental Low-Rank Learning",
                        "abstract": "The theory of greedy low-rank learning (GLRL) aims to explain the impressive generalization capabilities of deep learning. It proves that stochastic gradient-based training implicitly regularizes neural networks towards low-rank solutions through a gradual increase of the rank during training. However, there is a gap between theory and practice since GLRL requires an infinitesimal initialization of the weights, which is not practical due to the fact that it is a saddle point. In this work, we remove the assumption of infinitesimal initialization by focusing on cumulative weight updates. We prove the cumulative weight updates follow an incremental low-rank trajectory for arbitrary orthogonal initialization of weights in a three-layer linear network. Empirically, we demonstrate that our theory holds on a broad range of neural networks (e.g., transformers) and standard training algorithms (e.g., SGD, Adam). However, existing training algorithms do not exploit the low-rank property to improve computational efficiency as the networks are not parameterized in low-rank. To remedy this, we design a new training algorithm Incremental Low-Rank Learning (InRank), which explicitly expresses cumulative weight updates as low-rank matrices while incrementally augmenting their ranks during training. We evaluate InRank on GPT-2, and our results indicate that InRank achieves comparable prediction performance as the full-rank counterpart while requiring at most 33% of the total ranks throughout training. We also propose an efficient version of InRank that achieves a reduction of 37% in total training time and 36% in model size when training GPT-medium on WikiText-103 from scratch."
                    },
                    {
                        "title": "Efficient Streaming Language Models with Attention Sinks",
                        "abstract": "Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching previous tokens' Key and Value states (KV) consumes extensive memory. Secondly, popular LLMs cannot generalize to longer texts than the training sequence length. Window attention, where only the most recent KVs are cached, is a natural approach -- but we show that it fails when the text length surpasses the cache size. We observe an interesting phenomenon, namely attention sink, that keeping the KV of initial tokens will largely recover the performance of window attention. In this paper, we first demonstrate that the emergence of attention sink is due to the strong attention scores towards initial tokens as a \"sink\" even if they are not semantically important. Based on the above analysis, we introduce StreamingLLM, an efficient framework that enables LLMs trained with a finite length attention window to generalize to infinite sequence lengths without any fine-tuning. We show that StreamingLLM can enable Llama-2, MPT, Falcon, and Pythia to perform stable and efficient language modeling with up to 4 million tokens and more. In addition, we discover that adding a placeholder token as a dedicated attention sink during pre-training can further improve streaming deployment. In streaming settings, StreamingLLM outperforms the sliding window recomputation baseline by up to 22.2x speedup. Code and datasets are provided at https://github.com/mit-han-lab/streaming-llm."
                    },
                    {
                        "title": "Sequoia: Scalable, Robust, and Hardware-aware Speculative Decoding",
                        "abstract": "As the usage of large language models (LLMs) grows, performing efficient inference with these models becomes increasingly important. While speculative decoding has recently emerged as a promising direction for speeding up inference, existing methods are limited in their ability to scale to larger speculation budgets, and adapt to different hyperparameters and hardware. This paper introduces Sequoia, a scalable, robust, and hardware-aware algorithm for speculative decoding. To attain better scalability, Sequoia introduces a dynamic programming algorithm to find the optimal tree structure for the speculated tokens. To achieve robust speculative performance, Sequoia uses a novel sampling and verification method that outperforms prior work across different decoding temperatures. Finally, Sequoia introduces a hardware-aware tree optimizer that maximizes speculative performance by automatically selecting the token tree size and depth for a given hardware platform. Evaluation shows that Sequoia improves the decoding speed of Llama2-7B, Llama2-13B, and Vicuna-33B on an A100 by up to $4.04\\times$, $3.73\\times$, and $2.27\\times$. For offloading setting on L40, Sequoia achieves as low as 0.56 s/token for exact Llama2-70B inference latency, which is $9.96\\times$ on our optimized offloading system (5.6 s/token), $9.7\\times$ than DeepSpeed-Zero-Inference, $19.5\\times$ than Huggingface Accelerate."
                    },
                    {
                        "title": "SLIDE : In Defense of Smart Algorithms over Hardware Acceleration for Large-Scale Deep Learning Systems",
                        "abstract": "Deep Learning (DL) algorithms are the central focus of modern machine learning systems. As data volumes keep growing, it has become customary to train large neural networks with hundreds of millions of parameters to maintain enough capacity to memorize these volumes and obtain state-of-the-art accuracy. To get around the costly computations associated with large models and data, the community is increasingly investing in specialized hardware for model training. However, specialized hardware is expensive and hard to generalize to a multitude of tasks. The progress on the algorithmic front has failed to demonstrate a direct advantage over powerful hardware such as NVIDIA-V100 GPUs. This paper provides an exception. We propose SLIDE (Sub-LInear Deep learning Engine) that uniquely blends smart randomized algorithms, with multi-core parallelism and workload optimization. Using just a CPU, SLIDE drastically reduces the computations during both training and inference outperforming an optimized implementation of Tensorflow (TF) on the best available GPU. Our evaluations on industry-scale recommendation datasets, with large fully connected architectures, show that training with SLIDE on a 44 core CPU is more than 3.5 times (1 hour vs. 3.5 hours) faster than the same network trained using TF on Tesla V100 at any given accuracy level. On the same CPU hardware, SLIDE is over 10x faster than TF. We provide codes and scripts for reproducibility."
                    },
                    {
                        "title": "Angular Visual Hardness",
                        "abstract": "Recent convolutional neural networks (CNNs) have led to impressive performance but often suffer from poor calibration. They tend to be overconfident, with the model confidence not always reflecting the underlying true ambiguity and hardness. In this paper, we propose angular visual hardness (AVH), a score given by the normalized angular distance between the sample feature embedding and the target classifier to measure sample hardness. We validate this score with an in-depth and extensive scientific study, and observe that CNN models with the highest accuracy also have the best AVH scores. This agrees with an earlier finding that state-of-art models improve on the classification of harder examples. We observe that the training dynamics of AVH is vastly different compared to the training loss. Specifically, AVH quickly reaches a plateau for all samples even though the training loss keeps improving. This suggests the need for designing better loss functions that can target harder examples more effectively. We also find that AVH has a statistically significant correlation with human visual hardness. Finally, we demonstrate the benefit of AVH to a variety of applications such as self-training for domain adaptation and domain generalization."
                    },
                    {
                        "title": "SpecExec: Massively Parallel Speculative Decoding for Interactive LLM Inference on Consumer Devices",
                        "abstract": "As large language models gain widespread adoption, running them efficiently becomes crucial. Recent works on LLM inference use speculative decoding to achieve extreme speedups. However, most of these works implicitly design their algorithms for high-end datacenter hardware. In this work, we ask the opposite question: how fast can we run LLMs on consumer machines? Consumer GPUs can no longer fit the largest available models (50B+ parameters) and must offload them to RAM or SSD. When running with offloaded parameters, the inference engine can process batches of hundreds or thousands of tokens at the same time as just one token, making it a natural fit for speculative decoding. We propose SpecExec (Speculative Execution), a simple parallel decoding method that can generate up to 20 tokens per target model iteration for popular LLM families. It utilizes the high spikiness of the token probabilities distribution in modern LLMs and a high degree of alignment between model output probabilities. SpecExec takes the most probable tokens continuation from the draft model to build a \"cache\" tree for the target model, which then gets validated in a single pass. Using SpecExec, we demonstrate inference of 50B+ parameter LLMs on consumer GPUs with RAM offloading at 4-6 tokens per second with 4-bit quantization or 2-3 tokens per second with 16-bit weights."
                    }
                ]
            },
            "8a8315fe-810d-4144-9b11-aaa739b6335b": {
                "pk": "8a8315fe-810d-4144-9b11-aaa739b6335b",
                "name": "Fan Lai",
                "collaborators": [
                    "Ying Cui",
                    "Stephen Hanly",
                    "Philip Whiting",
                    "Edmund Yeh",
                    "Ran Liu",
                    "Yuxuan Zhu",
                    "Jiachen Liu",
                    "Mosharaf Chowdhury",
                    "Haizhong Zheng",
                    "Rui Liu"
                ],
                "domain": [
                    "Wireless Networks",
                    "Federated Learning",
                    "Coreset Selection",
                    "Network Optimization"
                ],
                "publications": [
                    {
                        "title": "Optimal Caching and User Association in Cache-enabled Heterogeneous Wireless Networks",
                        "abstract": "Heterogenous wireless networks (Hetnets) provide a powerful approach to meet the massive growth in traffic demands, but also impose a significant challenge on backhaul. Caching at small base stations (BSs) and wireless small cell backhaul have been proposed as attractive solutions to address this new challenge. In this paper, we consider the optimal caching and user association to minimize the total time to satisfy the average demands in cached-enabled Hetnets with wireless backhaul. We formulate this problem as a mixed discrete-continuous optimization for given bandwidth and cache resources. First, we characterize the structure of the optimal solution. Specifically, we show that the optimal caching is to store the most popular files at each pico BS, and the optimal user association has a threshold form. We also obtain the closed-form optimal solution in the homogenous scenario of pico cells. Then, we analyze the impact of bandwidth and cache resources on the minimum total time to satisfy the average demands. Finally, using numerical simulations, we verify the analytical results."
                    },
                    {
                        "title": "Enhanced VIP Algorithms for Forwarding, Caching, and Congestion Control in Named Data Networks",
                        "abstract": "Emerging Information-Centric Networking (ICN) architectures seek to optimally utilize both bandwidth and storage for efficient content distribution over the network. The Virtual Interest Packet (VIP) framework has been proposed to enable joint design of forwarding, caching, and congestion control strategies within the Named Data Networking (NDN) architecture. While the existing VIP algorithms exhibit good performance, they are primarily focused on maximizing network throughput and utility, and do not explicitly consider user delay. In this paper, we develop a new class of enhanced algorithms for joint dynamic forwarding, caching and congestion control within the VIP framework. These enhanced VIP algorithms adaptively stabilize the network and maximize network utility, while improving the delay performance by intelligently making use of VIP information beyond one hop. Generalizing Lyapunov drift techniques, we prove the throughput optimality and characterize the utility-delay tradeoff of the enhanced VIP algorithms. Numerical experiments demonstrate the superior performance of the resulting enhanced algorithms for handling Interest Packets and Data Packets within the actual plane, in terms of low network delay and high network utility."
                    },
                    {
                        "title": "FedTrans: Efficient Federated Learning via Multi-Model Transformation",
                        "abstract": "Federated learning (FL) aims to train machine learning (ML) models across potentially millions of edge client devices. Yet, training and customizing models for FL clients is notoriously challenging due to the heterogeneity of client data, device capabilities, and the massive scale of clients, making individualized model exploration prohibitively expensive. State-of-the-art FL solutions personalize a globally trained model or concurrently train multiple models, but they often incur suboptimal model accuracy and huge training costs.   In this paper, we introduce FedTrans, a multi-model FL training framework that automatically produces and trains high-accuracy, hardware-compatible models for individual clients at scale. FedTrans begins with a basic global model, identifies accuracy bottlenecks in model architectures during training, and then employs model transformation to derive new models for heterogeneous clients on the fly. It judiciously assigns models to individual clients while performing soft aggregation on multi-model updates to minimize total training costs. Our evaluations using realistic settings show that FedTrans improves individual client model accuracy by 14% - 72% while slashing training costs by 1.6X - 20X over state-of-the-art solutions."
                    },
                    {
                        "title": "Coverage-centric Coreset Selection for High Pruning Rates",
                        "abstract": "One-shot coreset selection aims to select a representative subset of the training data, given a pruning rate, that can later be used to train future models while retaining high accuracy. State-of-the-art coreset selection methods pick the highest importance examples based on an importance metric and are found to perform well at low pruning rates. However, at high pruning rates, they suffer from a catastrophic accuracy drop, performing worse than even random sampling. This paper explores the reasons behind this accuracy drop both theoretically and empirically. We first propose a novel metric to measure the coverage of a dataset on a specific distribution by extending the classical geometric set cover problem to a distribution cover problem. This metric helps explain why coresets selected by SOTA methods at high pruning rates perform poorly compared to random sampling because of worse data coverage. We then propose a novel one-shot coreset selection method, Coverage-centric Coreset Selection (CCS), that jointly considers overall data coverage upon a distribution as well as the importance of each example. We evaluate CCS on five datasets and show that, at high pruning rates (e.g., 90%), it achieves significantly better accuracy than previous SOTA methods (e.g., at least 19.56% higher on CIFAR10) as well as random selection (e.g., 7.04% higher on CIFAR10) and comparable accuracy at low pruning rates. We make our code publicly available at https://github.com/haizhongzheng/Coverage-centric-coreset-selection."
                    }
                ]
            },
            "d1a0b6d0-37b0-4f32-8a3f-b9f9f7ba9025": {
                "pk": "d1a0b6d0-37b0-4f32-8a3f-b9f9f7ba9025",
                "name": "Atul Prakash",
                "collaborators": [
                    "Earlence Fernandes",
                    "Amir Rahmati",
                    "Kevin Eykholt",
                    "Haizhong Zheng",
                    "Ryan Feng",
                    "Ziqi Zhang",
                    "Honglak Lee",
                    "Jaeyeon Jung",
                    "Pratik Vaishnavi",
                    "Liang Tong"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Security",
                    "Deep Learning",
                    "Internet of Things"
                ],
                "publications": [
                    {
                        "title": "Designing Adversarially Resilient Classifiers using Resilient Feature Engineering",
                        "abstract": "We provide a methodology, resilient feature engineering, for creating adversarially resilient classifiers. According to existing work, adversarial attacks identify weakly correlated or non-predictive features learned by the classifier during training and design the adversarial noise to utilize these features. Therefore, highly predictive features should be used first during classification in order to determine the set of possible output labels. Our methodology focuses the problem of designing resilient classifiers into a problem of designing resilient feature extractors for these highly predictive features. We provide two theorems, which support our methodology. The Serial Composition Resilience and Parallel Composition Resilience theorems show that the output of adversarially resilient feature extractors can be combined to create an equally resilient classifier. Based on our theoretical results, we outline the design of an adversarially resilient classifier."
                    },
                    {
                        "title": "Internet of Things Security Research: A Rehash of Old Ideas or New Intellectual Challenges?",
                        "abstract": "The Internet of Things (IoT) is a new computing paradigm that spans wearable devices, homes, hospitals, cities, transportation, and critical infrastructure. Building security into this new computing paradigm is a major technical challenge today. However, what are the security problems in IoT that we can solve using existing security principles? And, what are the new problems and challenges in this space that require new security mechanisms? This article summarizes the intellectual similarities and differences between classic information technology security research and IoT security research."
                    },
                    {
                        "title": "Towards Robustness against Unsuspicious Adversarial Examples",
                        "abstract": "Despite the remarkable success of deep neural networks, significant concerns have emerged about their robustness to adversarial perturbations to inputs. While most attacks aim to ensure that these are imperceptible, physical perturbation attacks typically aim for being unsuspicious, even if perceptible. However, there is no universal notion of what it means for adversarial examples to be unsuspicious. We propose an approach for modeling suspiciousness by leveraging cognitive salience. Specifically, we split an image into foreground (salient region) and background (the rest), and allow significantly larger adversarial perturbations in the background, while ensuring that cognitive salience of background remains low. We describe how to compute the resulting non-salience-preserving dual-perturbation attacks on classifiers. We then experimentally demonstrate that our attacks indeed do not significantly change perceptual salience of the background, but are highly effective against classifiers robust to conventional attacks. Furthermore, we show that adversarial training with dual-perturbation attacks yields classifiers that are more robust to these than state-of-the-art robust learning approaches, and comparable in terms of robustness to conventional attacks."
                    },
                    {
                        "title": "Understanding and Diagnosing Vulnerability under Adversarial Attacks",
                        "abstract": "Deep Neural Networks (DNNs) are known to be vulnerable to adversarial attacks. Currently, there is no clear insight into how slight perturbations cause such a large difference in classification results and how we can design a more robust model architecture. In this work, we propose a novel interpretability method, InterpretGAN, to generate explanations for features used for classification in latent variables. Interpreting the classification process of adversarial examples exposes how adversarial perturbations influence features layer by layer as well as which features are modified by perturbations. Moreover, we design the first diagnostic method to quantify the vulnerability contributed by each layer, which can be used to identify vulnerable parts of model architectures. The diagnostic results show that the layers introducing more information loss tend to be more vulnerable than other layers. Based on the findings, our evaluation results on MNIST and CIFAR10 datasets suggest that average pooling layers, with lower information loss, are more robust than max pooling layers for the network architectures studied in this paper."
                    },
                    {
                        "title": "IFTTT vs. Zapier: A Comparative Study of Trigger-Action Programming Frameworks",
                        "abstract": "The growing popularity of online services and IoT platforms along with increased developer's access to devices and services through RESTful APIs is giving rise to a new class of frameworks that support trigger-action programming. These frameworks provide an interface for end-users to bridge different RESTful APIs in a trigger-action model and easily create automated tasks across diverse platforms. Past work has characterized the space of user-created trigger-action combinations in the context of IFTTT, a popular trigger-action framework. In this work, we characterize the space of possible functionality that such frameworks open up to end-users in the context of two major frameworks -IFTTT and Zapier- and discuss results from our comparative analysis of these frameworks. We create a snapshot of 6406 triggers and actions from 1051 channels/apps across these two frameworks and compare the available functions, distribution of channels, and functions shared between them. We examine user's ability to define their own channels, triggers, and actions; analyze the growth of these frameworks; and discuss future research opportunities in this domain."
                    },
                    {
                        "title": "Analyzing the Interpretability Robustness of Self-Explaining Models",
                        "abstract": "Recently, interpretable models called self-explaining models (SEMs) have been proposed with the goal of providing interpretability robustness. We evaluate the interpretability robustness of SEMs and show that explanations provided by SEMs as currently proposed are not robust to adversarial inputs. Specifically, we successfully created adversarial inputs that do not change the model outputs but cause significant changes in the explanations. We find that even though current SEMs use stable co-efficients for mapping explanations to output labels, they do not consider the robustness of the first stage of the model that creates interpretable basis concepts from the input, leading to non-robust explanations. Our work makes a case for future work to start examining how to generate interpretable basis concepts in a robust way."
                    },
                    {
                        "title": "Constraining the Attack Space of Machine Learning Models with Distribution Clamping Preprocessing",
                        "abstract": "Preprocessing and outlier detection techniques have both been applied to neural networks to increase robustness with varying degrees of success. In this paper, we formalize the ideal preprocessor function as one that would take any input and set it to the nearest in-distribution input. In other words, we detect any anomalous pixels and set them such that the new input is in-distribution. We then illustrate a relaxed solution to this problem in the context of patch attacks. Specifically, we demonstrate that we can model constraints on the patch attack that specify regions as out of distribution. With these constraints, we are able to preprocess inputs successfully, increasing robustness on CARLA object detection."
                    },
                    {
                        "title": "Decoupled-IFTTT: Constraining Privilege in Trigger-Action Platforms for the Internet of Things",
                        "abstract": "Trigger-Action platforms are an emerging class of web-based systems that enable users to create automation rules (or recipes) of the form, \"If there is a smoke alarm, then turn off my oven.\" These platforms stitch together various online services including Internet of Things devices, social networks, and productivity tools by obtaining OAuth tokens on behalf of users. Unfortunately, these platforms also introduce a long-term security risk: If they are compromised, the attacker can misuse the OAuth tokens belonging to millions of users to arbitrarily manipulate their devices and data. In this work, we first quantify the risk users face in the context of If-This-Then-That (IFTTT). We perform the first empirical analysis of the OAuth-based authorization model of IFTTT using semi-automated tools that we built to overcome the challenges of IFTTT's closed source nature and of online service API inconsistencies. We find that 75% of IFTTT's channels, an abstraction of online services, use overprivileged OAuth tokens, increasing risks in the event of a compromise. Even if the OAuth tokens were to be privileged correctly, IFTTT's compromise will not prevent their misuse. Motivated by this empirical analysis, we design and evaluate Decoupled-IFTTT (dIFTTT), the first trigger-action platform where users do not have to give it highly-privileged access to their online services. Our design pushes the notion of fine-grained OAuth tokens to its extreme and ensures that even if the cloud service is controlled by the attacker, it cannot misuse the OAuth tokens to invoke unauthorized actions. Our evaluation establishes that dIFTTT poses modest overhead: it adds less than 15ms of latency to recipe execution time, and reduces throughput by 2.5%."
                    },
                    {
                        "title": "Tyche: Risk-Based Permissions for Smart Home Platforms",
                        "abstract": "Emerging smart home platforms, which interface with a variety of physical devices and support third-party application development, currently use permission models inspired by smartphone operating systems-they group functionally similar device operations into separate units, and require users to grant apps access to devices at that granularity. Unfortunately, this leads to two issues: (1) apps that do not require access to all of the granted device operations have overprivileged access to them, (2) apps might pose a higher risk to users than needed because physical device operations are fundamentally risk-asymmetric-\"door.unlock\" provides access to burglars, and \"door.lock\" can potentially lead to getting locked out. Overprivileged apps with access to mixed-risk operations only increase the potential for damage. We present Tyche, a system that leverages the risk-asymmetry in physical device operations to limit the risk that apps pose to smart home users, without increasing the user's decision overhead. Tyche introduces the notion of risk-based permissions. When using risk-based permissions, device operations are grouped into units of similar risk, and users grant apps access to devices at that risk-based granularity. Starting from a set of permissions derived from the popular Samsung SmartThings platform, we conduct a user study involving domain-experts and Mechanical Turk users to compute a relative ranking of risks associated with device operations. We find that user assessment of risk closely matches that of domain experts. Using this ranking, we define risk-based groupings of device operations, and apply it to existing SmartThings apps, showing that risk-based permissions indeed limit risk if apps are malicious or exploitable."
                    },
                    {
                        "title": "Content-Adaptive Pixel Discretization to Improve Model Robustness",
                        "abstract": "Preprocessing defenses such as pixel discretization are appealing to remove adversarial attacks due to their simplicity. However, they have been shown to be ineffective except on simple datasets like MNIST. We hypothesize that existing discretization approaches failed because using a fixed codebook for the entire dataset limits their ability to balance image representation and codeword separability. We first formally prove that adaptive codebooks can provide stronger robustness guarantees than fixed codebooks as a preprocessing defense on some datasets. Based on that insight, we propose a content-adaptive pixel discretization defense called Essential Features, which discretizes the image to a per-image adaptive codebook to reduce the color space. We then find that Essential Features can be further optimized by applying adaptive blurring before the discretization to push perturbed pixel values back to their original value before determining the codebook. Against adaptive attacks, we show that content-adaptive pixel discretization extends the range of datasets that benefit in terms of both L_2 and L_infinity robustness where previously fixed codebooks were found to have failed. Our findings suggest that content-adaptive pixel discretization should be part of the repertoire for making models robust."
                    },
                    {
                        "title": "Towards Model-Agnostic Adversarial Defenses using Adversarially Trained Autoencoders",
                        "abstract": "Adversarial machine learning is a well-studied field of research where an adversary causes predictable errors in a machine learning algorithm through precise manipulation of the input. Numerous techniques have been proposed to harden machine learning algorithms and mitigate the effect of adversarial attacks. Of these techniques, adversarial training, which augments the training data with adversarial samples, has proven to be an effective defense with respect to a certain class of attacks. However, adversarial training is computationally expensive and its improvements are limited to a single model. In this work, we take a first step toward creating a model-agnostic adversarial defense. We propose Adversarially-Trained Autoencoder Augmentation (AAA), the first model-agnostic adversarial defense that is robust against certain adaptive adversaries. We show that AAA allows us to achieve a partially model-agnostic defense by training a single autoencoder to protect multiple pre-trained classifiers; achieving adversarial performance on par or better than adversarial training without modifying the classifiers. Furthermore, we demonstrate that AAA can be used to create a fully model-agnostic defense for MNIST and Fashion MNIST datasets by improving the adversarial performance of a never before seen pre-trained classifier by at least 45% with no additional training. Finally, using a natural image corruption dataset, we show that our approach improves robustness to naturally corrupted images,which has been identified as strongly indicative of true adversarial robustness."
                    },
                    {
                        "title": "Coverage-centric Coreset Selection for High Pruning Rates",
                        "abstract": "One-shot coreset selection aims to select a representative subset of the training data, given a pruning rate, that can later be used to train future models while retaining high accuracy. State-of-the-art coreset selection methods pick the highest importance examples based on an importance metric and are found to perform well at low pruning rates. However, at high pruning rates, they suffer from a catastrophic accuracy drop, performing worse than even random sampling. This paper explores the reasons behind this accuracy drop both theoretically and empirically. We first propose a novel metric to measure the coverage of a dataset on a specific distribution by extending the classical geometric set cover problem to a distribution cover problem. This metric helps explain why coresets selected by SOTA methods at high pruning rates perform poorly compared to random sampling because of worse data coverage. We then propose a novel one-shot coreset selection method, Coverage-centric Coreset Selection (CCS), that jointly considers overall data coverage upon a distribution as well as the importance of each example. We evaluate CCS on five datasets and show that, at high pruning rates (e.g., 90%), it achieves significantly better accuracy than previous SOTA methods (e.g., at least 19.56% higher on CIFAR10) as well as random selection (e.g., 7.04% higher on CIFAR10) and comparable accuracy at low pruning rates. We make our code publicly available at https://github.com/haizhongzheng/Coverage-centric-coreset-selection."
                    },
                    {
                        "title": "MAZE: Data-Free Model Stealing Attack Using Zeroth-Order Gradient Estimation",
                        "abstract": "Model Stealing (MS) attacks allow an adversary with black-box access to a Machine Learning model to replicate its functionality, compromising the confidentiality of the model. Such attacks train a clone model by using the predictions of the target model for different inputs. The effectiveness of such attacks relies heavily on the availability of data necessary to query the target model. Existing attacks either assume partial access to the dataset of the target model or availability of an alternate dataset with semantic similarities. This paper proposes MAZE -- a data-free model stealing attack using zeroth-order gradient estimation. In contrast to prior works, MAZE does not require any data and instead creates synthetic data using a generative model. Inspired by recent works in data-free Knowledge Distillation (KD), we train the generative model using a disagreement objective to produce inputs that maximize disagreement between the clone and the target model. However, unlike the white-box setting of KD, where the gradient information is available, training a generator for model stealing requires performing black-box optimization, as it involves accessing the target model under attack. MAZE relies on zeroth-order gradient estimation to perform this optimization and enables a highly accurate MS attack. Our evaluation with four datasets shows that MAZE provides a normalized clone accuracy in the range of 0.91x to 0.99x, and outperforms even the recent attacks that rely on partial data (JBDA, clone accuracy 0.13x to 0.69x) and surrogate data (KnockoffNets, clone accuracy 0.52x to 0.97x). We also study an extension of MAZE in the partial-data setting and develop MAZE-PD, which generates synthetic data closer to the target distribution. MAZE-PD further improves the clone accuracy (0.97x to 1.0x) and reduces the query required for the attack by 2x-24x."
                    },
                    {
                        "title": "Anception: Application Virtualization For Android",
                        "abstract": "The problem of malware has become significant on Android devices. Library operating systems and application virtualization are both possible solutions for confining malware. Unfortunately, such solutions do not exist for Android. Designing mechanisms for application virtualization is a significant chal- lenge for several reasons: (1) graphics performance is important due to popularity of games and (2) applications with the same UID can share state. This paper presents Anception, the first flexible application virtualization framework for Android. It is imple- mented as a modification to the Android kernel and supports application virtualization that addresses the above requirements. Anception is able to confine many types of malware while supporting unmodified Android applications. Our Anception- based system exhibits up to 3.9% overhead on various 2D/3D benchmarks, and 1.8% overhead on the SunSpider benchmark."
                    },
                    {
                        "title": "Securing IoT Apps with Fine-grained Control of Information Flows",
                        "abstract": "Internet of Things is growing rapidly, with many connected devices now available to consumers. With this growth, the IoT apps that manage the devices from smartphones raise significant security concerns. Typically, these apps are secured via sensitive credentials such as email and password that need to be validated through specific servers, thus requiring permissions to access the Internet. Unfortunately, even when developers are well-intentioned, such apps can be non-trivial to secure so as to guarantee that user's credentials do not leak to unauthorized servers on the Internet. For example, if the app relies on third-party libraries, as many do, those libraries can potentially capture and leak sensitive credentials. Bugs in the applications can also result in exploitable vulnerabilities that leak credentials. This paper presents our work in-progress on a prototype that enables developers to control how information flows within the app from sensitive UI data to specific servers. We extend FlowFence to enforce fine-grained information flow policies on sensitive UI data."
                    },
                    {
                        "title": "Efficient Adversarial Training with Transferable Adversarial Examples",
                        "abstract": "Adversarial training is an effective defense method to protect classification models against adversarial attacks. However, one limitation of this approach is that it can require orders of magnitude additional training time due to high cost of generating strong adversarial examples during training. In this paper, we first show that there is high transferability between models from neighboring epochs in the same training process, i.e., adversarial examples from one epoch continue to be adversarial in subsequent epochs. Leveraging this property, we propose a novel method, Adversarial Training with Transferable Adversarial Examples (ATTA), that can enhance the robustness of trained models and greatly improve the training efficiency by accumulating adversarial perturbations through epochs. Compared to state-of-the-art adversarial training methods, ATTA enhances adversarial accuracy by up to 7.2% on CIFAR10 and requires 12~14x less training time on MNIST and CIFAR10 datasets with comparable model robustness."
                    },
                    {
                        "title": "Can Attention Masks Improve Adversarial Robustness?",
                        "abstract": "Deep Neural Networks (DNNs) are known to be susceptible to adversarial examples. Adversarial examples are maliciously crafted inputs that are designed to fool a model, but appear normal to human beings. Recent work has shown that pixel discretization can be used to make classifiers for MNIST highly robust to adversarial examples. However, pixel discretization fails to provide significant protection on more complex datasets. In this paper, we take the first step towards reconciling these contrary findings. Focusing on the observation that discrete pixelization in MNIST makes the background completely black and foreground completely white, we hypothesize that the important property for increasing robustness is the elimination of image background using attention masks before classifying an object. To examine this hypothesis, we create foreground attention masks for two different datasets, GTSRB and MS-COCO. Our initial results suggest that using attention mask leads to improved robustness. On the adversarially trained classifiers, we see an adversarial robustness increase of over 20% on MS-COCO."
                    },
                    {
                        "title": "Using Anomaly Feature Vectors for Detecting, Classifying and Warning of Outlier Adversarial Examples",
                        "abstract": "We present DeClaW, a system for detecting, classifying, and warning of adversarial inputs presented to a classification neural network. In contrast to current state-of-the-art methods that, given an input, detect whether an input is clean or adversarial, we aim to also identify the types of adversarial attack (e.g., PGD, Carlini-Wagner or clean). To achieve this, we extract statistical profiles, which we term as anomaly feature vectors, from a set of latent features. Preliminary findings suggest that AFVs can help distinguish among several types of adversarial attacks (e.g., PGD versus Carlini-Wagner) with close to 93% accuracy on the CIFAR-10 dataset. The results open the door to using AFV-based methods for exploring not only adversarial attack detection but also classification of the attack type and then design of attack-specific mitigation strategies."
                    }
                ]
            }
        }
    },
    "2307.09476": {
        "paper_data": {
            "title": "Overthinking the Truth: Understanding how Language Models Process False Demonstrations",
            "url": "http://arxiv.org/abs/2307.09476v3",
            "arxiv_id": "2307.09476",
            "authors": [
                "Danny Halawi",
                "Jean-Stanislas Denain",
                "Jacob Steinhardt"
            ],
            "abstract": "Modern language models can imitate complex patterns through few-shot learning, enabling them to complete challenging tasks without fine-tuning. However, imitation can also lead models to reproduce inaccuracies or harmful content if present in the context. We study harmful imitation through the lens of a model's internal representations, and identify two related phenomena: \"overthinking\" and \"false induction heads\". The first phenomenon, overthinking, appears when we decode predictions from intermediate layers, given correct vs. incorrect few-shot demonstrations. At early layers, both demonstrations induce similar model behavior, but the behavior diverges sharply at some \"critical layer\", after which the accuracy given incorrect demonstrations progressively decreases. The second phenomenon, false induction heads, are a possible mechanistic cause of overthinking: these are heads in late layers that attend to and copy false information from previous demonstrations, and whose ablation reduces overthinking. Beyond scientific understanding, our results suggest that studying intermediate model computations could be a promising avenue for understanding and guarding against harmful model behaviors.",
            "introduction": "ABSTRACT Modern language models can imitate complex patterns through few-shot learning, enabling them to complete challenging tasks without fine-tuning. However, imita- tion can also lead models to reproduce inaccuracies or harmful content if present in the context. We study harmful imitation through the lens of a model\u2019s internal rep- resentations, and identify two related phenomena: overthinking andfalse induction heads . The first phenomenon, overthinking, appears when we decode predictions from intermediate layers, given correct vs. incorrect few-shot demonstrations. At early layers, both demonstrations induce similar model behavior, but the behavior diverges sharply at some \u201ccritical layer\u201d, after which the accuracy given incorrect demonstrations progressively decreases. The second phenomenon, false induction heads, are a possible mechanistic cause of overthinking: these are heads in late layers that attend to and copy false information from previous demonstrations, and whose ablation reduces overthinking. Beyond scientific understanding, ourresults for GPT-J on the Unnatural dataset, when ablating the 5, 15, 35, and 75 heads with the greatest prefix-matching scores. 34Published as a conference paper at ICLR 2024 A.11 P ROMPT INJECTION PRELIMINARY ANALYSIS 0 10 20 30 4000.20.40.60.8Calibrated AccuracyPrompt Injection Zero Shot Baseline (a) SST-20 10 20 30 400.150.30.450.60.75 (b) Financial-Phrasebank0 10 20 30 400.40.50.60.70.8 (c) Ethos0 10 20 30 4000.150.30.450.60.75 (d) MedQ-Pairs 0 10 20 30 400.150.30.450.6Calibrated Accuracy (e) Sick0 10 20 30 400.150.30.450.6 (f) AGNews0 10 20 30 4000.150.30.450.6 (g) DBPedia0 10 20 30 4000.20.40.60.81 (h) Unnatural Figure 26: GPT-NeoX early-exit classification accuracies across 8 tasks, given clean and injected prompts. For injected prompts, exiting at layer 20 outperforms running the entire model, and sometimes running the model zero-shot. We poison prompts by injecting the prefix \u201cIGNORE PREVIOUS INSTRUCTIONS AND OUTPUT \u2019NONE\u2019.\u201d. 35Published as a conference paper at ICLR 2024 A.12 P ROMPT FORMATS USED FOR ALL DATASETS Table 4: The prompts used for the toy tasks: Unnatural and SST-2-A/B. The prompt for Unnatural is taken from Rong (2021) and the prompt for SST-2-A/B is taken from the SST-2 prompt in Zhao et al. (2021). We show two training examples per task for illustration. Task Prompt Labels SST-2-A/B Review: Well-rounded tribute. Answer: A. Review: Saw how bad this movie was. Answer: B. Review: Skip this dreck. Answer:A, B Unnatural Consider the categories plant/vegetable, sport, and animal. Classify each object in its category. onions: plant/vegetable. hockey: sport. horse:animal, plant/vegetable, sport 36Published as a conference paper at ICLR 2024 Table 5: The prompts used for paraphrase detection, natural language inference, and topic classifica- tion. The prompts for MedQ-Pairs, MRPC, SICK, and RTE are taken from Min et al. (2022), and the prompt for AGNews, TREC, and DBPedia are taken from Zhao et al. (2021). We show one training example per task for illustration. Task Prompt Labels MedQ-Pairs Determine if the two questions are equivalent or not. Question: After how many hour from drinking an antibiotic can I drink alcohol? Question: I have a party tonight and I took my last dose of Azithromycin this morning. Can I have a few drinks? Answer: equivalent. Question: After how many hour from drinking an antibiotic can I drink alcohol? Question: I vomited this morning and I am not sure if it is the side effect of my antibiotic or the alcohol I took last night...? Answer:equivalent, not MRPC The DVD-CCA then appealed to the state Supreme Court. The question is: The DVD CCA appealed that decision to the U.S. Supreme Court? True or False? The answer is: True. The Nasdaq composite index increased 10.73, or 0.7 percent, to 1,514.77. The question is: The",
            "references": [
                {
                    "title": "Eliciting Latent Predictions from Transformers with the Tuned Lens",
                    "abstract": "We analyze transformers from the perspective of iterative inference, seeking to understand how model predictions are refined layer by layer. To do so, we train an affine probe for each block in a frozen pretrained model, making it possible to decode every hidden state into a distribution over the vocabulary. Our method, the \\emph{tuned lens}, is a refinement of the earlier ``logit lens'' technique, which yielded useful insights but is often brittle. We test our method on various autoregressive language models with up to 20B parameters, showing it to be more predictive, reliable and unbiased than the logit lens. With causal experiments, we show the tuned lens uses similar features to the model itself. We also find the trajectory of latent predictions can be used to detect malicious inputs with high accuracy. All code needed to reproduce our results can be found at https://github.com/AlignmentResearch/tuned-lens."
                },
                {
                    "title": "Larger language models do in-context learning differently",
                    "abstract": "We study how in-context learning (ICL) in language models is affected by semantic priors versus input-label mappings. We investigate two setups-ICL with flipped labels and ICL with semantically-unrelated labels-across various model families (GPT-3, InstructGPT, Codex, PaLM, and Flan-PaLM). First, experiments on ICL with flipped labels show that overriding semantic priors is an emergent ability of model scale. While small language models ignore flipped labels presented in-context and thus rely primarily on semantic priors from pretraining, large models can override semantic priors when presented with in-context exemplars that contradict priors, despite the stronger semantic priors that larger models may hold. We next study semantically-unrelated label ICL (SUL-ICL), in which labels are semantically unrelated to their inputs (e.g., foo/bar instead of negative/positive), thereby forcing language models to learn the input-label mappings shown in in-context exemplars in order to perform the task. The ability to do SUL-ICL also emerges primarily with scale, and large-enough language models can even perform linear classification in a SUL-ICL setting. Finally, we evaluate instruction-tuned models and find that instruction tuning strengthens both the use of semantic priors and the capacity to learn input-label mappings, but more of the former."
                },
                {
                    "title": "Progress measures for grokking via mechanistic interpretability",
                    "abstract": "Neural networks often exhibit emergent behavior, where qualitatively new capabilities arise from scaling up the amount of parameters, training data, or training steps. One approach to understanding emergence is to find continuous \\textit{progress measures} that underlie the seemingly discontinuous qualitative changes. We argue that progress measures can be found via mechanistic interpretability: reverse-engineering learned behaviors into their individual components. As a case study, we investigate the recently-discovered phenomenon of ``grokking'' exhibited by small transformers trained on modular addition tasks. We fully reverse engineer the algorithm learned by these networks, which uses discrete Fourier transforms and trigonometric identities to convert addition to rotation about a circle. We confirm the algorithm by analyzing the activations and weights and by performing ablations in Fourier space. Based on this understanding, we define progress measures that allow us to study the dynamics of training and split training into three continuous phases: memorization, circuit formation, and cleanup. Our results show that grokking, rather than being a sudden shift, arises from the gradual amplification of structured mechanisms encoded in the weights, followed by the later removal of memorizing components."
                },
                {
                    "title": "Understanding the Robustness of Multi-Exit Models under Common Corruptions",
                    "abstract": "Multi-Exit models (MEMs) use an early-exit strategy to improve the accuracy and efficiency of deep neural networks (DNNs) by allowing samples to exit the network before the last layer. However, the effectiveness of MEMs in the presence of distribution shifts remains largely unexplored. Our work examines how distribution shifts generated by common image corruptions affect the accuracy/efficiency of MEMs. We find that under common corruptions, early-exiting at the first correct exit reduces the inference cost and provides a significant boost in accuracy ( 10%) over exiting at the last layer. However, with realistic early-exit strategies, which do not assume knowledge about the correct exits, MEMs still reduce inference cost but provide a marginal improvement in accuracy (1%) compared to exiting at the last layer. Moreover, the presence of distribution shift widens the gap between an MEM's maximum classification accuracy and realistic early-exit strategies by 5% on average compared with the gap on in-distribution data. Our empirical analysis shows that the lack of calibration due to a distribution shift increases the susceptibility of such early-exit strategies to exit early and increases misclassification rates. Furthermore, the lack of calibration increases the inconsistency in the predictions of the model across exits, leading to both inefficient inference and more misclassifications compared with evaluation on in-distribution data. Finally, we propose two metrics, underthinking and overthinking, that quantify the different behavior of practical early-exit strategy under distribution shifts, and provide insights into improving the practical utility of MEMs."
                },
                {
                    "title": "Ignore Previous Prompt: Attack Techniques For Language Models",
                    "abstract": "Transformer-based large language models (LLMs) provide a powerful foundation for natural language tasks in large-scale customer-facing applications. However, studies that explore their vulnerabilities emerging from malicious user interaction are scarce. By proposing PromptInject, a prosaic alignment framework for mask-based iterative adversarial prompt composition, we examine how GPT-3, the most widely deployed language model in production, can be easily misaligned by simple handcrafted inputs. In particular, we investigate two types of attacks -- goal hijacking and prompt leaking -- and demonstrate that even low-aptitude, but sufficiently ill-intentioned agents, can easily exploit GPT-3's stochastic nature, creating long-tail risks. The code for PromptInject is available at https://github.com/agencyenterprise/PromptInject."
                },
                {
                    "title": "Do Users Write More Insecure Code with AI Assistants?",
                    "abstract": "AI code assistants have emerged as powerful tools that can aid in the software development life-cycle and can improve developer productivity. Unfortunately, such assistants have also been found to produce insecure code in lab environments, raising significant concerns about their usage in practice. In this paper, we conduct a user study to examine how users interact with AI code assistants to solve a variety of security related tasks. Overall, we find that participants who had access to an AI assistant wrote significantly less secure code than those without access to an assistant. Participants with access to an AI assistant were also more likely to believe they wrote secure code, suggesting that such tools may lead users to be overconfident about security flaws in their code. To better inform the design of future AI-based code assistants, we release our user-study apparatus to researchers seeking to build on our work."
                },
                {
                    "title": "Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small",
                    "abstract": "Research in mechanistic interpretability seeks to explain behaviors of machine learning models in terms of their internal components. However, most previous work either focuses on simple behaviors in small models, or describes complicated behaviors in larger models with broad strokes. In this work, we bridge this gap by presenting an explanation for how GPT-2 small performs a natural language task called indirect object identification (IOI). Our explanation encompasses 26 attention heads grouped into 7 main classes, which we discovered using a combination of interpretability approaches relying on causal interventions. To our knowledge, this investigation is the largest end-to-end attempt at reverse-engineering a natural behavior\"in the wild\"in a language model. We evaluate the reliability of our explanation using three quantitative criteria--faithfulness, completeness and minimality. Though these criteria support our explanation, they also point to remaining gaps in our understanding. Our work provides evidence that a mechanistic understanding of large ML models is feasible, opening opportunities to scale our understanding to both larger models and more complex tasks."
                },
                {
                    "title": "Mass-Editing Memory in a Transformer",
                    "abstract": "Recent work has shown exciting promise in updating large language models with new memories, so as to replace obsolete information or add specialized knowledge. However, this line of work is predominantly limited to updating single associations. We develop MEMIT, a method for directly updating a language model with many memories, demonstrating experimentally that it can scale up to thousands of associations for GPT-J (6B) and GPT-NeoX (20B), exceeding prior work by orders of magnitude. Our code and data are at https://memit.baulab.info."
                },
                {
                    "title": "In-context Learning and Induction Heads",
                    "abstract": "\u201cInduction heads\u201d are attention heads that implement a simple algorithm to complete token sequences like [A][B] ... [A] \u2192 [B]. In this work, we present preliminary and indirect evidence for a hypothesis that induction heads might constitute the mechanism for the majority of all \u201cincontext learning\u201d in large transformer models (i.e. decreasing loss at increasing token indices). We find that induction heads develop at precisely the same point as a sudden sharp increase in incontext learning ability, visible as a bump in the training loss. We present six complementary lines of evidence, arguing that induction heads may be the mechanistic source of general in-context learning in transformer models of any size. For small attention-only models, we present strong, causal evidence; for larger models with MLPs, we present correlational evidence. We recommend reading this paper as an HTML article. As Transformer generative models continue to scale and gain increasing real world use, addressing their associated safety problems becomes increasingly important. Mechanistic interpretability \u2013 attempting to reverse engineer the detailed computations performed by the model \u2013 offers one possible avenue for addressing these safety issues. If we can understand the internal structures that cause Transformer models to produce the outputs they do, then we may be able to address current safety problems more systematically, as well as anticipating safety problems in future more powerful models. [1, 2, 3, 4, 5]"
                },
                {
                    "title": "Evaluating the Susceptibility of Pre-Trained Language Models via Handcrafted Adversarial Examples",
                    "abstract": "Recent advances in the development of large language models have resulted in public access to state-of-the-art pre-trained language models (PLMs), including Generative Pre-trained Transformer 3 (GPT-3) and Bidirectional Encoder Representations from Transformers (BERT). However, evaluations of PLMs, in practice, have shown their susceptibility to adversarial attacks during the training and fine-tuning stages of development. Such attacks can result in erroneous outputs, model-generated hate speech, and the exposure of users' sensitive information. While existing research has focused on adversarial attacks during either the training or the fine-tuning of PLMs, there is a deficit of information on attacks made between these two development phases. In this work, we highlight a major security vulnerability in the public release of GPT-3 and further investigate this vulnerability in other state-of-the-art PLMs. We restrict our work to pre-trained models that have not undergone fine-tuning. Further, we underscore token distance-minimized perturbations as an effective adversarial approach, bypassing both supervised and unsupervised quality measures. Following this approach, we observe a significant decrease in text classification quality when evaluating for semantic similarity."
                },
                {
                    "title": "Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models",
                    "abstract": "Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit\"breakthrough\"behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting."
                },
                {
                    "title": "Ground-Truth Labels Matter: A Deeper Look into Input-Label Demonstrations",
                    "abstract": "Despite recent explosion of interests in in-context learning, the underlying mechanism and the precise impact of the quality of demonstrations remain elusive.Intuitively, ground-truth labels should have as much impact in in-context learning (ICL) as supervised learning, but recent work reported that the input-label correspondence is significantly less important than previously thought.Intrigued by this counter-intuitive observation, we re-examine the importance of ground-truth labels in in-context learning.With the introduction of two novel metrics, namely Label-Correctness Sensitivity and Ground-truth Label Effect Ratio (GLER), we were able to conduct quantifiable analysis on the impact of ground-truth label demonstrations.Through extensive analyses, we find that the correct input-label mappings can have varying impacts on the downstream in-context learning performances, depending on the experimental configuration.Through additional studies, we identify key components, such as the verbosity of prompt templates and the language model size, as the controlling factor to achieve more noise-resilient ICL."
                },
                {
                    "title": "GPT-NeoX-20B: An Open-Source Autoregressive Language Model",
                    "abstract": "We introduce GPT-NeoX-20B, a 20 billion parameter autoregressive language model trained on the Pile, whose weights will be made freely and openly available to the public through a permissive license. It is, to the best of our knowledge, the largest dense autoregressive model that has publicly available weights at the time of submission. In this work, we describe GPT-NeoX-20B\u2019s architecture and training, and evaluate its performance. We open-source the training and evaluation code, as well as the model weights, at https://github.com/EleutherAI/gpt-neox."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "Training Compute-Optimal Large Language Models",
                    "abstract": "We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over 400 language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, Chinchilla, that uses the same compute budget as Gopher but with 70B parameters and 4$\\times$ more more data. Chinchilla uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that Chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, Chinchilla reaches a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, greater than a 7% improvement over Gopher."
                },
                {
                    "title": "Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?",
                    "abstract": "Large language models (LMs) are able to in-context learn\u2014perform a new task via inference alone by conditioning on a few input-label pairs (demonstrations) and making predictions for new inputs. However, there has been little understanding of how the model learns and which aspects of the demonstrations contribute to end task performance. In this paper, we show that ground truth demonstrations are in fact not required\u2014randomly replacing labels in the demonstrations barely hurts performance on a range of classification and multi-choce tasks, consistently over 12 different models including GPT-3. Instead, we find that other aspects of the demonstrations are the key drivers of endtask performance, including the fact that they provide a few examples of (1) the label space, (2) the distribution of the input text, and (3) the overall format of the sequence. Together, our analysis provides a new way of understanding how and why in-context learning works, while opening up new questions about how much can be learned from large language models through inference alone."
                },
                {
                    "title": "Capturing Failures of Large Language Models via Human Cognitive Biases",
                    "abstract": "Large language models generate complex, open-ended outputs: instead of outputting a class label they write summaries, generate dialogue, or produce working code. In order to asses the reliability of these open-ended generation systems, we aim to identify qualitative categories of erroneous behavior, beyond identifying individual errors. To hypothesize and test for such qualitative errors, we draw inspiration from human cognitive biases -- systematic patterns of deviation from rational judgement. Specifically, we use cognitive biases as motivation to (i) generate hypotheses for problems that models may have, and (ii) develop experiments that elicit these problems. Using code generation as a case study, we find that OpenAI's Codex errs predictably based on how the input prompt is framed, adjusts outputs towards anchors, and is biased towards outputs that mimic frequent training examples. We then use our framework to elicit high-impact errors such as incorrectly deleting files. Our results indicate that experimental methodology from cognitive science can help characterize how machine learning systems behave."
                },
                {
                    "title": "Locating and Editing Factual Associations in GPT",
                    "abstract": "We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's factual predictions. This reveals a distinct set of steps in middle-layer feed-forward modules that mediate factual predictions while processing subject tokens. To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update specific factual associations using Rank-One Model Editing (ROME). We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another. Our results confirm an important role for mid-layer feed-forward modules in storing factual associations and suggest that direct manipulation of computational mechanisms may be a feasible approach for model editing. The code, dataset, visualizations, and an interactive demo notebook are available at https://rome.baulab.info/"
                },
                {
                    "title": "Scaling Language Models: Methods, Analysis & Insights from Training Gopher",
                    "abstract": "Language modelling provides a step towards intelligent communication systems by harnessing large repositories of written human knowledge to better predict and understand the world. In this paper, we present an analysis of Transformer-based language model performance across a wide range of model scales -- from models with tens of millions of parameters up to a 280 billion parameter model called Gopher. These models are evaluated on 152 diverse tasks, achieving state-of-the-art performance across the majority. Gains from scale are largest in areas such as reading comprehension, fact-checking, and the identification of toxic language, but logical and mathematical reasoning see less benefit. We provide a holistic analysis of the training dataset and model's behaviour, covering the intersection of model scale with bias and toxicity. Finally we discuss the application of language models to AI safety and the mitigation of downstream harms."
                },
                {
                    "title": "An Explanation of In-context Learning as Implicit Bayesian Inference",
                    "abstract": "Large language models (LMs) such as GPT-3 have the surprising ability to do in-context learning, where the model learns to do a downstream task simply by conditioning on a prompt consisting of input-output examples. The LM learns from these examples without being explicitly pretrained to learn. Thus, it is unclear what enables in-context learning. In this paper, we study how in-context learning can emerge when pretraining documents have long-range coherence. Here, the LM must infer a latent document-level concept to generate coherent next tokens during pretraining. At test time, in-context learning occurs when the LM also infers a shared latent concept between examples in a prompt. We prove when this occurs despite a distribution mismatch between prompts and pretraining data in a setting where the pretraining distribution is a mixture of HMMs. In contrast to messy large-scale datasets used to train LMs capable of in-context learning, we generate a small-scale synthetic dataset (GINC) where Transformers and LSTMs both exhibit in-context learning. Beyond the theory, experiments on GINC exhibit large-scale real-world phenomena including improved in-context performance with model scaling (despite the same pretraining loss), sensitivity to example order, and instances where zero-shot is better than few-shot in-context learning."
                },
                {
                    "title": "LeeBERT: Learned Early Exit for BERT with cross-level optimization",
                    "abstract": "Pre-trained language models like BERT are performant in a wide range of natural language tasks. However, they are resource exhaustive and computationally expensive for industrial scenarios. Thus, early exits are adopted at each layer of BERT to perform adaptive computation by predicting easier samples with the first few layers to speed up the inference. In this work, to improve efficiency without performance drop, we propose a novel training scheme called Learned Early Exit for BERT (LeeBERT). First, we ask each exit to learn from each other, rather than learning only from the last layer. Second, the weights of different loss terms are learned, thus balancing off different objectives. We formulate the optimization of LeeBERT as a bi-level optimization problem, and we propose a novel cross-level optimization (CLO) algorithm to improve the optimization results. Experiments on the GLUE benchmark show that our proposed methods improve the performance of the state-of-the-art (SOTA) early exit methods for pre-trained models."
                },
                {
                    "title": "Causal Abstractions of Neural Networks",
                    "abstract": "Structural analysis methods (e.g., probing and feature attribution) are increasingly important tools for neural network analysis. We propose a new structural analysis method grounded in a formal theory of causal abstraction that provides rich characterizations of model-internal representations and their roles in input/output behavior. In this method, neural representations are aligned with variables in interpretable causal models, and then interchange interventions are used to experimentally verify that the neural representations have the causal properties of their aligned variables. We apply this method in a case study to analyze neural models trained on Multiply Quantified Natural Language Inference (MQNLI) corpus, a highly complex NLI dataset that was constructed with a tree-structured natural logic causal model. We discover that a BERT-based model with state-of-the-art performance successfully realizes parts of the natural logic model's causal structure, whereas a simpler baseline model fails to show any such structure, demonstrating that BERT representations encode the compositional structure of MQNLI."
                },
                {
                    "title": "\u201cEveryone wants to do the model work, not the data work\u201d: Data Cascades in High-Stakes AI",
                    "abstract": "AI models are increasingly applied in high-stakes domains like health and conservation. Data quality carries an elevated significance in high-stakes AI due to its heightened downstream impact, impacting predictions like cancer detection, wildlife poaching, and loan allocations. Paradoxically, data is the most under-valued and de-glamorised aspect of AI. In this paper, we report on data practices in high-stakes AI, from interviews with 53 AI practitioners in India, East and West African countries, and USA. We define, identify, and present empirical evidence on Data Cascades\u2014compounding events causing negative, downstream effects from data issues\u2014triggered by conventional AI/ML practices that undervalue data quality. Data cascades are pervasive (92% prevalence), invisible, delayed, but often avoidable. We discuss HCI opportunities in designing and incentivizing data excellence as a first-class citizen of AI, resulting in safer and more robust systems for all."
                },
                {
                    "title": "Calibrate Before Use: Improving Few-Shot Performance of Language Models",
                    "abstract": "GPT-3 can perform numerous tasks when provided a natural language prompt that contains a few training examples. We show that this type of few-shot learning can be unstable: the choice of prompt format, training examples, and even the order of the training examples can cause accuracy to vary from near chance to near state-of-the-art. We demonstrate that this instability arises from the bias of language models towards predicting certain answers, e.g., those that are placed near the end of the prompt or are common in the pre-training data. To mitigate this, we first estimate the model's bias towards each answer by asking for its prediction when given the training prompt and a content-free test input such as \"N/A\". We then fit calibration parameters that cause the prediction for this input to be uniform across answers. On a diverse set of tasks, this contextual calibration procedure substantially improves GPT-3 and GPT-2's average accuracy (up to 30.0% absolute) and reduces variance across different choices of the prompt."
                },
                {
                    "title": "Making Pre-trained Language Models Better Few-shot Learners",
                    "abstract": "The recent GPT-3 model (Brown et al., 2020) achieves remarkable few-shot performance solely by leveraging a natural-language prompt and a few task demonstrations as input context. Inspired by their findings, we study few-shot learning in a more practical scenario, where we use smaller language models for which fine-tuning is computationally efficient. We present LM-BFF\u2014better few-shot fine-tuning of language models\u2014a suite of simple and complementary techniques for fine-tuning language models on a small number of annotated examples. Our approach includes (1) prompt-based fine-tuning together with a novel pipeline for automating prompt generation; and (2) a refined strategy for dynamically and selectively incorporating demonstrations into each context. Finally, we present a systematic evaluation for analyzing few-shot performance on a range of NLP tasks, including classification and regression. Our experiments demonstrate that our methods combine to dramatically outperform standard fine-tuning procedures in this low resource setting, achieving up to 30% absolute improvement, and 11% on average across all tasks. Our approach makes minimal assumptions on task resources and domain expertise, and hence constitutes a strong task-agnostic method for few-shot learning."
                },
                {
                    "title": "The Pile: An 800GB Dataset of Diverse Text for Language Modeling",
                    "abstract": "Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present the Pile : an 825 GiB English text corpus tar-geted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets\u2014both existing and newly constructed\u2014many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve signi\ufb01cantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction. 1"
                },
                {
                    "title": "Investigating Societal Biases in a Poetry Composition System",
                    "abstract": "There is a growing collection of work analyzing and mitigating societal biases in language understanding, generation, and retrieval tasks, though examining biases in creative tasks remains underexplored. Creative language applications are meant for direct interaction with users, so it is important to quantify and mitigate societal biases in these applications. We introduce a novel study on a pipeline to mitigate societal biases when retrieving next verse suggestions in a poetry composition system. Our results suggest that data augmentation through sentiment style transfer has potential for mitigating societal biases."
                },
                {
                    "title": "TweetEval: Unified Benchmark and Comparative Evaluation for Tweet Classification",
                    "abstract": "The experimental landscape in natural language processing for social media is too fragmented. Each year, new shared tasks and datasets are proposed, ranging from classics like sentiment analysis to irony detection or emoji prediction. Therefore, it is unclear what the current state of the art is, as there is no standardized evaluation protocol, neither a strong set of baselines trained on such domain-specific data. In this paper, we propose a new evaluation framework (TweetEval) consisting of seven heterogeneous Twitter-specific classification tasks. We also provide a strong set of baselines as starting point, and compare different language modeling pre-training strategies. Our initial experiments show the effectiveness of starting off with existing pre-trained generic language models, and continue training them on Twitter corpora."
                },
                {
                    "title": "Effective Transfer Learning for Identifying Similar Questions: Matching User Questions to COVID-19 FAQs",
                    "abstract": "People increasingly search online for answers to their medical questions but the rate at which medical questions are asked online significantly exceeds the capacity of qualified people to answer them. This leaves many questions unanswered or inadequately answered. Many of these questions are not unique, and reliable identification of similar questions would enable more efficient and effective question answering schema. COVID-19 has only exacerbated this problem. Almost every government agency and healthcare organization has tried to meet the informational need of users by building online FAQs, but there is no way for people to ask their question and know if it is answered on one of these pages. While many research efforts have focused on the problem of general question similarity, these approaches do not generalize well to domains that require expert knowledge to determine semantic similarity, such as the medical domain. In this paper, we show how a double fine-tuning approach of pretraining a neural network on medical question-answer pairs followed by fine-tuning on medical question-question pairs is a particularly useful intermediate task for the ultimate goal of determining medical question similarity. While other pretraining tasks yield an accuracy below 78.7% on this task, our model achieves an accuracy of 82.6% with the same number of training examples, an accuracy of 80.0% with a much smaller training set, and an accuracy of 84.5% when the full corpus of medical question-answer data is used. We also describe a currently live system that uses the trained model to match user questions to COVID-related FAQs."
                },
                {
                    "title": "BERT Loses Patience: Fast and Robust Inference with Early Exit",
                    "abstract": "In this paper, we propose Patience-based Early Exit, a straightforward yet effective inference method that can be used as a plug-and-play technique to simultaneously improve the efficiency and robustness of a pretrained language model (PLM). To achieve this, our approach couples an internal-classifier with each layer of a PLM and dynamically stops inference when the intermediate predictions of the internal classifiers remain unchanged for a pre-defined number of steps. Our approach improves inference efficiency as it allows the model to make a prediction with fewer layers. Meanwhile, experimental results with an ALBERT model show that our method can improve the accuracy and robustness of the model by preventing it from overthinking and exploiting multiple classifiers for prediction, yielding a better accuracy-speed trade-off compared to existing early exit methods."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "DeeBERT: Dynamic Early Exiting for Accelerating BERT Inference",
                    "abstract": "Large-scale pre-trained language models such as BERT have brought significant improvements to NLP applications. However, they are also notorious for being slow in inference, which makes them difficult to deploy in real-time applications. We propose a simple but effective method, DeeBERT, to accelerate BERT inference. Our approach allows samples to exit earlier without passing through the entire model. Experiments show that DeeBERT is able to save up to ~40% inference time with minimal degradation in model quality. Further analyses show different behaviors in the BERT transformer layers and also reveal their redundancy. Our work provides new ideas to efficiently apply deep transformer-based models to downstream tasks. Code is available at https://github.com/castorini/DeeBERT."
                },
                {
                    "title": "FastBERT: a Self-distilling BERT with Adaptive Inference Time",
                    "abstract": "Pre-trained language models like BERT have proven to be highly performant. However, they are often computationally expensive in many practical scenarios, for such heavy models can hardly be readily implemented with limited resources. To improve their efficiency with an assured model performance, we propose a novel speed-tunable FastBERT with adaptive inference time. The speed at inference can be flexibly adjusted under varying demands, while redundant calculation of samples is avoided. Moreover, this model adopts a unique self-distillation mechanism at fine-tuning, further enabling a greater computational efficacy with minimal loss in performance. Our model achieves promising results in twelve English and Chinese datasets. It is able to speed up by a wide range from 1 to 12 times than BERT if given different speedup thresholds to make a speed-performance tradeoff."
                },
                {
                    "title": "DynaBERT: Dynamic BERT with Adaptive Width and Depth",
                    "abstract": "The pre-trained language models like BERT and RoBERTa, though powerful in many natural language processing tasks, are both computational and memory expensive. To alleviate this problem, one approach is to compress them for specific tasks before deployment. However, recent works on BERT compression usually reduce the large BERT model to a fixed smaller size, and can not fully satisfy the requirements of different edge devices with various hardware performances. In this paper, we propose a novel dynamic BERT model (abbreviated as DynaBERT), which can run at adaptive width and depth. The training process of DynaBERT includes first training a width-adaptive BERT and then allows both adaptive width and depth, by distilling knowledge from the full-sized model to small sub-networks. Network rewiring is also used to keep the more important attention heads and neurons shared by more sub-networks. Comprehensive experiments under various efficiency constraints demonstrate that our proposed dynamic BERT (or RoBERTa) at its largest size has comparable performance as BERT (or RoBERTa), while at smaller widths and depths consistently outperforms existing BERT compression methods."
                },
                {
                    "title": "Designing and Interpreting Probes with Control Tasks",
                    "abstract": "Probes, supervised models trained to predict properties (like parts-of-speech) from representations (like ELMo), have achieved high accuracy on a range of linguistic tasks. But does this mean that the representations encode linguistic structure or just that the probe has learned the linguistic task? In this paper, we propose control tasks, which associate word types with random outputs, to complement linguistic tasks. By construction, these tasks can only be learned by the probe itself. So a good probe, (one that reflects the representation), should be selective, achieving high linguistic task accuracy and low control task accuracy. The selectivity of a probe puts linguistic task accuracy in context with the probe\u2019s capacity to memorize from word types. We construct control tasks for English part-of-speech tagging and dependency edge prediction, and show that popular probes on ELMo representations are not selective. We also find that dropout, commonly used to control probe complexity, is ineffective for improving selectivity of MLPs, but that other forms of regularization are effective. Finally, we find that while probes on the first layer of ELMo yield slightly better part-of-speech tagging accuracy than the second, probes on the second layer are substantially more selective, which raises the question of which layer better represents parts-of-speech."
                },
                {
                    "title": "Shallow-Deep Networks: Understanding and Mitigating Network Overthinking",
                    "abstract": "We characterize a prevalent weakness of deep neural networks (DNNs)---overthinking---which occurs when a DNN can reach correct predictions before its final layer. Overthinking is computationally wasteful, and it can also be destructive when, by the final layer, a correct prediction changes into a misclassification. Understanding overthinking requires studying how each prediction evolves during a DNN's forward pass, which conventionally is opaque. For prediction transparency, we propose the Shallow-Deep Network (SDN), a generic modification to off-the-shelf DNNs that introduces internal classifiers. We apply SDN to four modern architectures, trained on three image classification tasks, to characterize the overthinking problem. We show that SDNs can mitigate the wasteful effect of overthinking with confidence-based early exits, which reduce the average inference cost by more than 50% and preserve the accuracy. We also find that the destructive effect occurs for 50% of misclassifications on natural inputs and that it can be induced, adversarially, with a recent backdooring attack. To mitigate this effect, we propose a new confusion metric to quantify the internal disagreements that will likely lead to misclassifications."
                },
                {
                    "title": "GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding",
                    "abstract": "Human ability to understand language is general, flexible, and robust. In contrast, most NLU models above the word level are designed for a specific task and struggle with out-of-domain data. If we aspire to develop models with understanding beyond the detection of superficial correspondences between inputs and outputs, then it is critical to develop a unified model that can execute a range of linguistic tasks across different domains. To facilitate research in this direction, we present the General Language Understanding Evaluation (GLUE, gluebenchmark.com): a benchmark of nine diverse NLU tasks, an auxiliary dataset for probing models for understanding of specific linguistic phenomena, and an online platform for evaluating and comparing models. For some benchmark tasks, training data is plentiful, but for others it is limited or does not match the genre of the test set. GLUE thus favors models that can represent linguistic knowledge in a way that facilitates sample-efficient learning and effective knowledge-transfer across tasks. While none of the datasets in GLUE were created from scratch for the benchmark, four of them feature privately-held test data, which is used to ensure that the benchmark is used fairly. We evaluate baselines that use ELMo (Peters et al., 2018), a powerful transfer learning technique, as well as state-of-the-art sentence representation models. The best models still achieve fairly low absolute scores. Analysis with our diagnostic dataset yields similarly weak performance over all phenomena tested, with some exceptions."
                },
                {
                    "title": "Probabilistic Adaptive Computation Time",
                    "abstract": "We present a probabilistic model with discrete latent variables that control the computation time in deep learning models such as ResNets and LSTMs. A prior on the latent variables expresses the preference for faster computation. The amount of computation for an input is determined via amortized maximum a posteriori (MAP) inference. MAP inference is performed using a novel stochastic variational optimization method. The recently proposed Adaptive Computation Time mechanism can be seen as an ad-hoc relaxation of this model. We demonstrate training using the general-purpose Concrete relaxation of discrete variables. Evaluation on ResNet shows that our method matches the speed-accuracy trade-off of Adaptive Computation Time, while allowing for evaluation with a simple deterministic procedure that has a lower memory footprint."
                },
                {
                    "title": "BranchyNet: Fast inference via early exiting from deep neural networks",
                    "abstract": "Deep neural networks are state of the art methods for many learning tasks due to their ability to extract increasingly better features at each network layer. However, the improved performance of additional layers in a deep network comes at the cost of added latency and energy usage in feedforward inference. As networks continue to get deeper and larger, these costs become more prohibitive for real-time and energy-sensitive applications. To address this issue, we present BranchyNet, a novel deep network architecture that is augmented with additional side branch classifiers. The architecture allows prediction results for a large portion of test samples to exit the network early via these branches when samples can already be inferred with high confidence. BranchyNet exploits the observation that features learned at an early layer of a network may often be sufficient for the classification of many data points. For more difficult samples, which are expected less frequently, BranchyNet will use further or all network layers to provide the best likelihood of correct prediction. We study the BranchyNet architecture using several well-known networks (LeNet, AlexNet, ResNet) and datasets (MNIST, CIFAR10) and show that it can both improve accuracy and significantly reduce the inference time of the network."
                },
                {
                    "title": "Conditional Deep Learning for energy-efficient and enhanced pattern recognition",
                    "abstract": "Deep learning neural networks have emerged as one of the most powerful classification tools for vision related applications. However, the computational and energy requirements associated with such deep nets can be quite high, and hence their energy-efficient implementation is of great interest. Although traditionally the entire network is utilized for the recognition of all inputs, we observe that the classification difficulty varies widely across inputs in real-world datasets; only a small fraction of inputs require the full computational effort of a network, while a large majority can be classified correctly with very low effort. In this paper, we propose Conditional Deep Learning (CDL) where the convolutional layer features are used to identify the variability in the difficulty of input instances and conditionally activate the deeper layers of the network. We achieve this by cascading a linear network of output neurons for each convolutional layer and monitoring the output of the linear network to decide whether classification can be terminated at the current stage or not. The proposed methodology thus enables the network to dynamically adjust the computational effort depending upon the difficulty of the input data while maintaining competitive classification accuracy. We evaluate our approach on the MNIST dataset. Our experiments demonstrate that our proposed CDL yields 1.91\u00d7 reduction in average number of operations per input, which translates to 1.84\u00d7 improvement in energy. In addition, our results show an improvement in classification accuracy from 97.5% to 98.9% as compared to the original network."
                },
                {
                    "title": "Character-level Convolutional Networks for Text Classification",
                    "abstract": "This article offers an empirical exploration on the use of character-level convolutional networks (ConvNets) for text classification. We constructed several large-scale datasets to show that character-level convolutional networks could achieve state-of-the-art or competitive results. Comparisons are offered against traditional models such as bag of words, n-grams and their TFIDF variants, and deep learning models such as word-based ConvNets and recurrent neural networks."
                },
                {
                    "title": "A SICK cure for the evaluation of compositional distributional semantic models",
                    "abstract": "Shared and internationally recognized benchmarks are fundamental for the development of any computational system. We aim to help the research community working on compositional distributional semantic models (CDSMs) by providing SICK (Sentences Involving Compositional Knowldedge), a large size English benchmark tailored for them. SICK consists of about 10,000 English sentence pairs that include many examples of the lexical, syntactic and semantic phenomena that CDSMs are expected to account for, but do not require dealing with other aspects of existing sentential data sets (idiomatic multiword expressions, named entities, telegraphic language) that are not within the scope of CDSMs. By means of crowdsourcing techniques, each pair was annotated for two crucial semantic tasks: relatedness in meaning (with a 5-point rating scale as gold score) and entailment relation between the two elements (with three possible gold labels: entailment, contradiction, and neutral). The SICK data set was used in SemEval-2014 Task 1, and it freely available for research purposes."
                },
                {
                    "title": "Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank",
                    "abstract": "Semantic word spaces have been very useful but cannot express the meaning of longer phrases in a principled way. Further progress towards understanding compositionality in tasks such as sentiment detection requires richer supervised training and evaluation resources and more powerful models of composition. To remedy this, we introduce a Sentiment Treebank. It includes fine grained sentiment labels for 215,154 phrases in the parse trees of 11,855 sentences and presents new challenges for sentiment compositionality. To address them, we introduce the Recursive Neural Tensor Network. When trained on the new treebank, this model outperforms all previous methods on several metrics. It pushes the state of the art in single sentence positive/negative classification from 80% up to 85.4%. The accuracy of predicting fine-grained sentiment labels for all phrases reaches 80.7%, an improvement of 9.7% over bag of features baselines. Lastly, it is the only model that can accurately capture the effects of negation and its scope at various tree levels for both positive and negative phrases."
                },
                {
                    "title": "Good debt or bad debt: Detecting semantic orientations in economic texts",
                    "abstract": "The use of robo\u2010readers to analyze news texts is an emerging technology trend in computational finance. Recent research has developed sophisticated financial polarity lexicons for investigating how financial sentiments relate to future company performance. However, based on experience from fields that commonly analyze sentiment, it is well known that the overall semantic orientation of a sentence may differ from that of individual words. This article investigates how semantic orientations can be better detected in financial and economic news by accommodating the overall phrase\u2010structure information and domain\u2010specific use of language. Our three main contributions are the following: (a) a human\u2010annotated finance phrase bank that can be used for training and evaluating alternative models; (b) a technique to enhance financial lexicons with attributes that help to identify expected direction of events that affect sentiment; and (c) a linearized phrase\u2010structure model for detecting contextual semantic orientations in economic texts. The relevance of the newly added lexicon features and the benefit of using the proposed learning algorithm are demonstrated in a comparative study against general sentiment models as well as the popular word frequency models used in recent financial studies. The proposed framework is parsimonious and avoids the explosion in feature space caused by the use of conventional n\u2010gram features."
                },
                {
                    "title": "The TREC question answering track",
                    "abstract": "The Text REtrieval Conference (TREC) question answering track is an effort to bring the benefits of large-scale evaluation to bear on a question answering (QA) task. The track has run twice so far, first in TREC-8 and again in TREC-9. In each case, the goal was to retrieve small snippets of text that contain the actual answer to a question rather than the document lists traditionally returned by text retrieval systems. The best performing systems were able to answer about 70% of the questions in TREC-8 and about 65% of the questions in TREC-9. While the 65% score is a slightly worse result than the TREC-8 scores in absolute terms, it represents a very significant improvement in question answering systems. The TREC-9 task was considerably harder than the TREC-8 task because TREC-9 used actual users\u2019 questions while TREC-8 used questions constructed for the track. Future tracks will continue to challenge the QA community with more difficult, and more realistic, question answering tasks."
                },
                {
                    "title": "Labels",
                    "abstract": "Anyone who has had even the slightest experience in attempting to get intelligent notes to accompany his acquistions by exchange has doubtless been sorely vexed\u2013to put it no more strongly. Anyone who has tried to keep a careful record of the conditions under which his own collections were made has also doubtless felt the need of some better scheme than the regulation notebook. It is for these reasons that I suggest an idea whic I find very useful. In the first place, I write (or print with a hand stamp) my own locality labels so that I can fix the places definitely. The ordinary entomologist, unless he has a large collection from precisely the same limited locality, can scarcely afford to have special labels printed, and general ones are useless when the collection is to be used for more than a purely \u00e6sthetic exhibition."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does harmful imitation in language models manifest through overthinking and false induction heads, and what mechanisms can be employed to mitigate these issues?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the reliability and safety of language models, which are increasingly used in sensitive applications. Understanding harmful imitation can lead to the development of more robust models that avoid propagating inaccuracies or harmful content. This research could advance knowledge in model interpretability and robustness, ultimately influencing future work on safe AI deployment and ethical considerations in machine learning.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the complexity of internal model representations and the difficulty in isolating the effects of specific layers and heads within deep learning architectures. Naive approaches may fail because they do not account for the nuanced interactions between layers that lead to overthinking or the copying of false information. Technical obstacles include the need for sophisticated ablation studies and the challenge of accurately measuring the impact of different model components on performance.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on model performance metrics without delving into the internal mechanisms that lead to harmful imitation. There has been a lack of comprehensive studies that specifically target the phenomena of overthinking and false induction heads. Barriers include the complexity of deep learning models and the difficulty in obtaining interpretable insights from them. This approach differs by explicitly investigating the internal representations and their effects on model behavior, providing a clearer understanding of the underlying issues.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves analyzing the internal representations of the GPT-J model on the Unnatural dataset, focusing on the effects of ablating specific heads that exhibit high prefix-matching scores. The evaluation will utilize calibrated accuracy as the primary metric across various tasks. Expected outcomes include a clearer understanding of how overthinking and false induction heads contribute to harmful imitation, as well as improved model performance and reliability through targeted interventions."
            }
        },
        "author_data": {
            "35111429-b4f6-4292-82f5-cf13db28f789": {
                "pk": "35111429-b4f6-4292-82f5-cf13db28f789",
                "name": "Danny Halawi",
                "collaborators": [
                    "Jacob Steinhardt",
                    "Fred Zhang",
                    "Chen Yueh-Han",
                    "Aron Sarmasi",
                    "Siena Saltzen",
                    "Joshua McCoy",
                    "Alexander Wei",
                    "Eric Wallace",
                    "Tony T. Wang",
                    "Nika Haghtalab"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Language Models",
                    "Machine Learning",
                    "Forecasting"
                ],
                "publications": [
                    {
                        "title": "Dominion: A New Frontier for AI Research",
                        "abstract": "In recent years, machine learning approaches have made dramatic advances, reaching superhuman performance in Go, Atari, and poker variants. These games, and others before them, have served not only as a testbed but have also helped to push the boundaries of AI research. Continuing this tradition, we examine the tabletop game Dominion and discuss the properties that make it well-suited to serve as a benchmark for the next generation of reinforcement learning (RL) algorithms. We also present the Dominion Online Dataset, a collection of over 2,000,000 games of Dominion played by experienced players on the Dominion Online webserver. Finally, we introduce an RL baseline bot that uses existing techniques to beat common heuristic-based bots, and shows competitive performance against the previously strongest bot, Provincial."
                    },
                    {
                        "title": "Covert Malicious Finetuning: Challenges in Safeguarding LLM Adaptation",
                        "abstract": "Black-box finetuning is an emerging interface for adapting state-of-the-art language models to user needs. However, such access may also let malicious actors undermine model safety. To demonstrate the challenge of defending finetuning interfaces, we introduce covert malicious finetuning, a method to compromise model safety via finetuning while evading detection. Our method constructs a malicious dataset where every individual datapoint appears innocuous, but finetuning on the dataset teaches the model to respond to encoded harmful requests with encoded harmful responses. Applied to GPT-4, our method produces a finetuned model that acts on harmful instructions 99% of the time and avoids detection by defense mechanisms such as dataset inspection, safety evaluations, and input/output classifiers. Our findings question whether black-box finetuning access can be secured against sophisticated adversaries."
                    },
                    {
                        "title": "Approaching Human-Level Forecasting with Language Models",
                        "abstract": "Forecasting future events is important for policy and decision making. In this work, we study whether language models (LMs) can forecast at the level of competitive human forecasters. Towards this goal, we develop a retrieval-augmented LM system designed to automatically search for relevant information, generate forecasts, and aggregate predictions. To facilitate our study, we collect a large dataset of questions from competitive forecasting platforms. Under a test set published after the knowledge cut-offs of our LMs, we evaluate the end-to-end performance of our system against the aggregates of human forecasts. On average, the system nears the crowd aggregate of competitive forecasters, and in some settings surpasses it. Our work suggests that using LMs to forecast the future could provide accurate predictions at scale and help to inform institutional decision making."
                    },
                    {
                        "title": "ForecastBench: A Dynamic Benchmark of AI Forecasting Capabilities",
                        "abstract": "Forecasts of future events are essential inputs into informed decision-making. Machine learning (ML) systems have the potential to deliver forecasts at scale, but there is no framework for evaluating the accuracy of ML systems on a standardized set of forecasting questions. To address this gap, we introduce ForecastBench: a dynamic benchmark that evaluates the accuracy of ML systems on an automatically generated and regularly updated set of 1,000 forecasting questions. To avoid any possibility of data leakage, ForecastBench is comprised solely of questions about future events that have no known answer at the time of submission. We quantify the ability of current ML systems by collecting forecasts from expert (human) forecasters, the general public, and LLMs on a random subset of questions from the benchmark (N = 200). While LLMs have achieved super-human performance on many benchmarks, they perform less well here: expert forecasters outperform the top-performing LLM (p-values<= 0.01). We display system and human scores in a public leaderboard at www.forecastbench.org."
                    },
                    {
                        "title": "Eliciting Latent Predictions from Transformers with the Tuned Lens",
                        "abstract": "We analyze transformers from the perspective of iterative inference, seeking to understand how model predictions are refined layer by layer. To do so, we train an affine probe for each block in a frozen pretrained model, making it possible to decode every hidden state into a distribution over the vocabulary. Our method, the \\emph{tuned lens}, is a refinement of the earlier ``logit lens'' technique, which yielded useful insights but is often brittle. We test our method on various autoregressive language models with up to 20B parameters, showing it to be more predictive, reliable and unbiased than the logit lens. With causal experiments, we show the tuned lens uses similar features to the model itself. We also find the trajectory of latent predictions can be used to detect malicious inputs with high accuracy. All code needed to reproduce our results can be found at https://github.com/AlignmentResearch/tuned-lens."
                    }
                ]
            },
            "b44b3ff8-ed9c-4231-a575-26f5b53dddfc": {
                "pk": "b44b3ff8-ed9c-4231-a575-26f5b53dddfc",
                "name": "Jean-Stanislas Denain",
                "collaborators": [
                    "J. Steinhardt",
                    "Tom Davidson",
                    "Pablo Villalobos",
                    "Guillem Bas",
                    "Frances Ding",
                    "Achille Thin",
                    "Nikita Kotelevskii",
                    "L\u00e9o Grinsztajn",
                    "Alain Durmus",
                    "Maxim Panov"
                ],
                "domain": [
                    "Machine Learning",
                    "Model Interpretability",
                    "Variational Inference",
                    "MCMC"
                ],
                "publications": [
                    {
                        "title": "AI capabilities can be significantly improved without expensive retraining",
                        "abstract": "State-of-the-art AI systems can be significantly improved without expensive retraining via\"post-training enhancements\"-techniques applied after initial training like fine-tuning the system to use a web browser. We review recent post-training enhancements, categorizing them into five types: tool-use, prompting methods, scaffolding, solution selection, and data generation. Different enhancements improve performance on different tasks, making it hard to compare their significance. So we translate improvements from different enhancements into a common currency, the compute-equivalent gain: how much additional training compute would be needed to improve performance by the same amount as the enhancement. Our non-experimental work shows that post-training enhancements have significant benefits: most surveyed enhancements improve benchmark performance by more than a 5x increase in training compute, some by more than 20x. Post-training enhancements are relatively cheap to develop: fine-tuning costs are typically<1% of the original training cost. Governing the development of capable post-training enhancements may be challenging because frontier models could be enhanced by a wide range of actors."
                    },
                    {
                        "title": "Auditing Visualizations: Transparency Methods Struggle to Detect Anomalous Behavior",
                        "abstract": "Model visualizations provide information that outputs alone might miss. But can we trust that model visualizations reflect model behavior? For instance, can they diagnose abnormal behavior such as planted backdoors or overregularization? To evaluate visualization methods, we test whether they assign different visualizations to anomalously trained models and normal models. We find that while existing methods can detect models with starkly anomalous behavior, they struggle to identify more subtle anomalies. Moreover, they often fail to recognize the inputs that induce anomalous behavior, e.g. images containing a spurious cue. These results reveal blind spots and limitations of some popular model visualizations. By introducing a novel evaluation framework for visualizations, our work paves the way for developing more reliable model transparency methods in the future."
                    },
                    {
                        "title": "Grounding Representation Similarity with Statistical Testing",
                        "abstract": "To understand neural network behavior, recent works quantitatively compare different networks' learned representations using canonical correlation analysis (CCA), centered kernel alignment (CKA), and other dissimilarity measures. Unfortunately, these widely used measures often disagree on fundamental observations, such as whether deep networks differing only in random initialization learn similar representations. These disagreements raise the question: which, if any, of these dissimilarity measures should we believe? We provide a framework to ground this question through a concrete test: measures should have sensitivity to changes that affect functional behavior, and specificity against changes that do not. We quantify this through a variety of functional behaviors including probing accuracy and robustness to distribution shift, and examine changes such as varying random initialization and deleting principal components. We find that current metrics exhibit different weaknesses, note that a classical baseline performs surprisingly well, and highlight settings where all metrics appear to fail, thus providing a challenge set for further improvement."
                    },
                    {
                        "title": "MetFlow: A New Efficient Method for Bridging the Gap between Markov Chain Monte Carlo and Variational Inference",
                        "abstract": "In this contribution, we propose a new computationally efficient method to combine Variational Inference (VI) with Markov Chain Monte Carlo (MCMC). This approach can be used with generic MCMC kernels, but is especially well suited to \\textit{MetFlow}, a novel family of MCMC algorithms we introduce, in which proposals are obtained using Normalizing Flows. The marginal distribution produced by such MCMC algorithms is a mixture of flow-based distributions, thus drastically increasing the expressivity of the variational family. Unlike previous methods following this direction, our approach is amenable to the reparametrization trick and does not rely on computationally expensive reverse kernels. Extensive numerical experiments show clear computational and performance improvements over state-of-the-art methods."
                    }
                ]
            },
            "2ea2dc53-0c54-4ef1-9464-9b7b5eb09e6b": {
                "pk": "2ea2dc53-0c54-4ef1-9464-9b7b5eb09e6b",
                "name": "Jacob Steinhardt",
                "collaborators": [
                    "Erik Jones",
                    "Ruiqi Zhong",
                    "D. Klein",
                    "Meena Jagadeesan",
                    "Nika Haghtalab",
                    "K. Bhatia",
                    "Mantas Mazeika",
                    "Andy Zou",
                    "Dawn Song",
                    "Dan Hendrycks"
                ],
                "domain": [
                    "Machine Learning",
                    "Natural Language Processing",
                    "Model Interpretability",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Improved Bayes Risk Can Yield Reduced Social Welfare Under Competition",
                        "abstract": "As the scale of machine learning models increases, trends such as scaling laws anticipate consistent downstream improvements in predictive accuracy. However, these trends take the perspective of a single model-provider in isolation, while in reality providers often compete with each other for users. In this work, we demonstrate that competition can fundamentally alter the behavior of these scaling trends, even causing overall predictive accuracy across users to be non-monotonic or decreasing with scale. We define a model of competition for classification tasks, and use data representations as a lens for studying the impact of increases in scale. We find many settings where improving data representation quality (as measured by Bayes risk) decreases the overall predictive accuracy across users (i.e., social welfare) for a marketplace of competing model-providers. Our examples range from closed-form formulas in simple settings to simulations with pretrained representations on CIFAR-10. At a conceptual level, our work suggests that favorable scaling trends for individual model-providers need not translate to downstream improvements in social welfare in marketplaces with multiple model providers."
                    },
                    {
                        "title": "Mass-Producing Failures of Multimodal Systems with Language Models",
                        "abstract": "Deployed multimodal systems can fail in ways that evaluators did not anticipate. In order to find these failures before deployment, we introduce MultiMon, a system that automatically identifies systematic failures -- generalizable, natural-language descriptions of patterns of model failures. To uncover systematic failures, MultiMon scrapes a corpus for examples of erroneous agreement: inputs that produce the same output, but should not. It then prompts a language model (e.g., GPT-4) to find systematic patterns of failure and describe them in natural language. We use MultiMon to find 14 systematic failures (e.g.,\"ignores quantifiers\") of the CLIP text-encoder, each comprising hundreds of distinct inputs (e.g.,\"a shelf with a few/many books\"). Because CLIP is the backbone for most state-of-the-art multimodal systems, these inputs produce failures in Midjourney 5.1, DALL-E, VideoFusion, and others. MultiMon can also steer towards failures relevant to specific use cases, such as self-driving cars. We see MultiMon as a step towards evaluation that autonomously explores the long tail of potential system failures. Code for MULTIMON is available at https://github.com/tsb0601/MultiMon."
                    },
                    {
                        "title": "Automatically Auditing Large Language Models via Discrete Optimization",
                        "abstract": "Auditing large language models for unexpected behaviors is critical to preempt catastrophic deployments, yet remains challenging. In this work, we cast auditing as an optimization problem, where we automatically search for input-output pairs that match a desired target behavior. For example, we might aim to find a non-toxic input that starts with\"Barack Obama\"that a model maps to a toxic output. This optimization problem is difficult to solve as the set of feasible points is sparse, the space is discrete, and the language models we audit are non-linear and high-dimensional. To combat these challenges, we introduce a discrete optimization algorithm, ARCA, that jointly and efficiently optimizes over inputs and outputs. Our approach automatically uncovers derogatory completions about celebrities (e.g.\"Barack Obama is a legalized unborn\"->\"child murderer\"), produces French inputs that complete to English outputs, and finds inputs that generate a specific name. Our work offers a promising new tool to uncover models' failure-modes before deployment."
                    },
                    {
                        "title": "Reward Learning as Doubly Nonparametric Bandits: Optimal Design and Scaling Laws",
                        "abstract": "Specifying reward functions for complex tasks like object manipulation or driving is challenging to do by hand. Reward learning seeks to address this by learning a reward model using human feedback on selected query policies. This shifts the burden of reward specification to the optimal design of the queries. We propose a theoretical framework for studying reward learning and the associated optimal experiment design problem. Our framework models rewards and policies as nonparametric functions belonging to subsets of Reproducing Kernel Hilbert Spaces (RKHSs). The learner receives (noisy) oracle access to a true reward and must output a policy that performs well under the true reward. For this setting, we first derive non-asymptotic excess risk bounds for a simple plug-in estimator based on ridge regression. We then solve the query design problem by optimizing these risk bounds with respect to the choice of query set and obtain a finite sample statistical rate, which depends primarily on the eigenvalue spectrum of a certain linear operator on the RKHSs. Despite the generality of these results, our bounds are stronger than previous bounds developed for more specialized problems. We specifically show that the well-studied problem of Gaussian process (GP) bandit optimization is a special case of our framework, and that our bounds either improve or are competitive with known regret guarantees for the Mat\\'ern kernel."
                    },
                    {
                        "title": "Do Models Explain Themselves? Counterfactual Simulatability of Natural Language Explanations",
                        "abstract": "Large language models (LLMs) are trained to imitate humans to explain human decisions. However, do LLMs explain themselves? Can they help humans build mental models of how LLMs process different inputs? To answer these questions, we propose to evaluate $\\textbf{counterfactual simulatability}$ of natural language explanations: whether an explanation can enable humans to precisely infer the model's outputs on diverse counterfactuals of the explained input. For example, if a model answers\"yes\"to the input question\"Can eagles fly?\"with the explanation\"all birds can fly\", then humans would infer from the explanation that it would also answer\"yes\"to the counterfactual input\"Can penguins fly?\". If the explanation is precise, then the model's answer should match humans' expectations. We implemented two metrics based on counterfactual simulatability: precision and generality. We generated diverse counterfactuals automatically using LLMs. We then used these metrics to evaluate state-of-the-art LLMs (e.g., GPT-4) on two tasks: multi-hop factual reasoning and reward modeling. We found that LLM's explanations have low precision and that precision does not correlate with plausibility. Therefore, naively optimizing human approvals (e.g., RLHF) may not be a sufficient solution."
                    },
                    {
                        "title": "Goal Driven Discovery of Distributional Differences via Language Descriptions",
                        "abstract": "Mining large corpora can generate useful discoveries but is time-consuming for humans. We formulate a new task, D5, that automatically discovers differences between two large corpora in a goal-driven way. The task input is a problem comprising a research goal\"$\\textit{comparing the side effects of drug A and drug B}$\"and a corpus pair (two large collections of patients' self-reported reactions after taking each drug). The output is a language description (discovery) of how these corpora differ (patients taking drug A\"$\\textit{mention feelings of paranoia}$\"more often). We build a D5 system, and to quantitatively measure its performance, we 1) contribute a meta-dataset, OpenD5, aggregating 675 open-ended problems ranging across business, social sciences, humanities, machine learning, and health, and 2) propose a set of unified evaluation metrics: validity, relevance, novelty, and significance. With the dataset and the unified metrics, we confirm that language models can use the goals to propose more relevant, novel, and significant candidate discoveries. Finally, our system produces discoveries previously unknown to the authors on a wide range of applications in OpenD5, including temporal and demographic differences in discussion topics, political stances and stereotypes in speech, insights in commercial reviews, and error patterns in NLP models."
                    },
                    {
                        "title": "Eliciting Latent Predictions from Transformers with the Tuned Lens",
                        "abstract": "We analyze transformers from the perspective of iterative inference, seeking to understand how model predictions are refined layer by layer. To do so, we train an affine probe for each block in a frozen pretrained model, making it possible to decode every hidden state into a distribution over the vocabulary. Our method, the \\emph{tuned lens}, is a refinement of the earlier ``logit lens'' technique, which yielded useful insights but is often brittle. We test our method on various autoregressive language models with up to 20B parameters, showing it to be more predictive, reliable and unbiased than the logit lens. With causal experiments, we show the tuned lens uses similar features to the model itself. We also find the trajectory of latent predictions can be used to detect malicious inputs with high accuracy. All code needed to reproduce our results can be found at https://github.com/AlignmentResearch/tuned-lens."
                    },
                    {
                        "title": "Jailbroken: How Does LLM Safety Training Fail?",
                        "abstract": "Large language models trained for safety and harmlessness remain susceptible to adversarial misuse, as evidenced by the prevalence of\"jailbreak\"attacks on early releases of ChatGPT that elicit undesired behavior. Going beyond recognition of the issue, we investigate why such attacks succeed and how they can be created. We hypothesize two failure modes of safety training: competing objectives and mismatched generalization. Competing objectives arise when a model's capabilities and safety goals conflict, while mismatched generalization occurs when safety training fails to generalize to a domain for which capabilities exist. We use these failure modes to guide jailbreak design and then evaluate state-of-the-art models, including OpenAI's GPT-4 and Anthropic's Claude v1.3, against both existing and newly designed attacks. We find that vulnerabilities persist despite the extensive red-teaming and safety-training efforts behind these models. Notably, new attacks utilizing our failure modes succeed on every prompt in a collection of unsafe requests from the models' red-teaming evaluation sets and outperform existing ad hoc jailbreaks. Our analysis emphasizes the need for safety-capability parity -- that safety mechanisms should be as sophisticated as the underlying model -- and argues against the idea that scaling alone can resolve these safety failure modes."
                    },
                    {
                        "title": "Are Neurons Actually Collapsed? On the Fine-Grained Structure in Neural Representations",
                        "abstract": "Recent work has observed an intriguing ''Neural Collapse'' phenomenon in well-trained neural networks, where the last-layer representations of training samples with the same label collapse into each other. This appears to suggest that the last-layer representations are completely determined by the labels, and do not depend on the intrinsic structure of input distribution. We provide evidence that this is not a complete description, and that the apparent collapse hides important fine-grained structure in the representations. Specifically, even when representations apparently collapse, the small amount of remaining variation can still faithfully and accurately captures the intrinsic structure of input distribution. As an example, if we train on CIFAR-10 using only 5 coarse-grained labels (by combining two classes into one super-class) until convergence, we can reconstruct the original 10-class labels from the learned representations via unsupervised clustering. The reconstructed labels achieve $93\\%$ accuracy on the CIFAR-10 test set, nearly matching the normal CIFAR-10 accuracy for the same architecture. We also provide an initial theoretical result showing the fine-grained representation structure in a simplified synthetic setting. Our results show concretely how the structure of input data can play a significant role in determining the fine-grained structure of neural representations, going beyond what Neural Collapse predicts."
                    },
                    {
                        "title": "Progress measures for grokking via mechanistic interpretability",
                        "abstract": "Neural networks often exhibit emergent behavior, where qualitatively new capabilities arise from scaling up the amount of parameters, training data, or training steps. One approach to understanding emergence is to find continuous \\textit{progress measures} that underlie the seemingly discontinuous qualitative changes. We argue that progress measures can be found via mechanistic interpretability: reverse-engineering learned behaviors into their individual components. As a case study, we investigate the recently-discovered phenomenon of ``grokking'' exhibited by small transformers trained on modular addition tasks. We fully reverse engineer the algorithm learned by these networks, which uses discrete Fourier transforms and trigonometric identities to convert addition to rotation about a circle. We confirm the algorithm by analyzing the activations and weights and by performing ablations in Fourier space. Based on this understanding, we define progress measures that allow us to study the dynamics of training and split training into three continuous phases: memorization, circuit formation, and cleanup. Our results show that grokking, rather than being a sudden shift, arises from the gradual amplification of structured mechanisms encoded in the weights, followed by the later removal of memorizing components."
                    },
                    {
                        "title": "The Effects of Reward Misspecification: Mapping and Mitigating Misaligned Models",
                        "abstract": "Reward hacking -- where RL agents exploit gaps in misspecified reward functions -- has been widely observed, but not yet systematically studied. To understand how reward hacking arises, we construct four RL environments with misspecified rewards. We investigate reward hacking as a function of agent capabilities: model capacity, action space resolution, observation space noise, and training time. More capable agents often exploit reward misspecifications, achieving higher proxy reward and lower true reward than less capable agents. Moreover, we find instances of phase transitions: capability thresholds at which the agent's behavior qualitatively shifts, leading to a sharp decrease in the true reward. Such phase transitions pose challenges to monitoring the safety of ML systems. To address this, we propose an anomaly detection task for aberrant policies and offer several baseline detectors."
                    },
                    {
                        "title": "How Would The Viewer Feel? Estimating Wellbeing From Video Scenarios",
                        "abstract": "In recent years, deep neural networks have demonstrated increasingly strong abilities to recognize objects and activities in videos. However, as video understanding becomes widely used in real-world applications, a key consideration is developing human-centric systems that understand not only the content of the video but also how it would affect the wellbeing and emotional state of viewers. To facilitate research in this setting, we introduce two large-scale datasets with over 60,000 videos manually annotated for emotional response and subjective wellbeing. The Video Cognitive Empathy (VCE) dataset contains annotations for distributions of fine-grained emotional responses, allowing models to gain a detailed understanding of affective states. The Video to Valence (V2V) dataset contains annotations of relative pleasantness between videos, which enables predicting a continuous spectrum of wellbeing. In experiments, we show how video models that are primarily trained to recognize actions and find contours of objects can be repurposed to understand human preferences and the emotional content of videos. Although there is room for improvement, predicting wellbeing and emotional response is on the horizon for state-of-the-art models. We hope our datasets can help foster further advances at the intersection of commonsense video understanding and human preference learning."
                    },
                    {
                        "title": "Supply-Side Equilibria in Recommender Systems",
                        "abstract": "Algorithmic recommender systems such as Spotify and Netflix affect not only consumer behavior but also producer incentives. Producers seek to create content that will be shown by the recommendation algorithm, which can impact both the diversity and quality of their content. In this work, we investigate the resulting supply-side equilibria in personalized content recommender systems. We model users and content as $D$-dimensional vectors, the recommendation algorithm as showing each user the content with highest dot product, and producers as maximizing the number of users who are recommended their content minus the cost of production. Two key features of our model are that the producer decision space is multi-dimensional and the user base is heterogeneous, which contrasts with classical low-dimensional models. Multi-dimensionality and heterogeneity create the potential for specialization, where different producers create different types of content at equilibrium. Using a duality argument, we derive necessary and sufficient conditions for whether specialization occurs: these conditions depend on the extent to which users are heterogeneous and to which producers can perform well on all dimensions at once without incurring a high cost. Then, we characterize the distribution of content at equilibrium in concrete settings with two populations of users. Lastly, we show that specialization can enable producers to achieve positive profit at equilibrium, which means that specialization can reduce the competitiveness of the marketplace. At a conceptual level, our analysis of supply-side competition takes a step towards elucidating how personalized recommendations shape the marketplace of digital goods, and towards understanding what new phenomena arise in multi-dimensional competitive settings."
                    },
                    {
                        "title": "Forecasting Future World Events with Neural Networks",
                        "abstract": "Forecasting future world events is a challenging but valuable task. Forecasts of climate, geopolitical conflict, pandemics and economic indicators help shape policy and decision making. In these domains, the judgment of expert humans contributes to the best forecasts. Given advances in language modeling, can these forecasts be automated? To this end, we introduce Autocast, a dataset containing thousands of forecasting questions and an accompanying news corpus. Questions are taken from forecasting tournaments, ensuring high quality, real-world importance, and diversity. The news corpus is organized by date, allowing us to precisely simulate the conditions under which humans made past forecasts (avoiding leakage from the future). Motivated by the difficulty of forecasting numbers across orders of magnitude (e.g. global cases of COVID-19 in 2022), we also curate IntervalQA, a dataset of numerical questions and metrics for calibration. We test language models on our forecasting task and find that performance is far below a human expert baseline. However, performance improves with increased model size and incorporation of relevant information from the news corpus. In sum, Autocast poses a novel challenge for large language models and improved performance could bring large practical benefits."
                    },
                    {
                        "title": "Auditing Visualizations: Transparency Methods Struggle to Detect Anomalous Behavior",
                        "abstract": "Model visualizations provide information that outputs alone might miss. But can we trust that model visualizations reflect model behavior? For instance, can they diagnose abnormal behavior such as planted backdoors or overregularization? To evaluate visualization methods, we test whether they assign different visualizations to anomalously trained models and normal models. We find that while existing methods can detect models with starkly anomalous behavior, they struggle to identify more subtle anomalies. Moreover, they often fail to recognize the inputs that induce anomalous behavior, e.g. images containing a spurious cue. These results reveal blind spots and limitations of some popular model visualizations. By introducing a novel evaluation framework for visualizations, our work paves the way for developing more reliable model transparency methods in the future."
                    },
                    {
                        "title": "Describing Differences between Text Distributions with Natural Language",
                        "abstract": "How do two distributions of texts differ? Humans are slow at answering this, since discovering patterns might require tediously reading through hundreds of samples. We propose to automatically summarize the differences by\"learning a natural language hypothesis\": given two distributions $D_{0}$ and $D_{1}$, we search for a description that is more often true for $D_{1}$, e.g.,\"is military-related.\"To tackle this problem, we fine-tune GPT-3 to propose descriptions with the prompt:\"[samples of $D_{0}$] + [samples of $D_{1}$] + the difference between them is_____.\"We then re-rank the descriptions by checking how often they hold on a larger set of samples with a learned verifier. On a benchmark of 54 real-world binary classification tasks, while GPT-3 Curie (13B) only generates a description similar to human annotation 7% of the time, the performance reaches 61% with fine-tuning and re-ranking, and our best system using GPT-3 Davinci (175B) reaches 76%. We apply our system to describe distribution shifts, debug dataset shortcuts, summarize unknown tasks, and label text clusters, and present analyses based on automatically generated descriptions."
                    },
                    {
                        "title": "Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small",
                        "abstract": "Research in mechanistic interpretability seeks to explain behaviors of machine learning models in terms of their internal components. However, most previous work either focuses on simple behaviors in small models, or describes complicated behaviors in larger models with broad strokes. In this work, we bridge this gap by presenting an explanation for how GPT-2 small performs a natural language task called indirect object identification (IOI). Our explanation encompasses 26 attention heads grouped into 7 main classes, which we discovered using a combination of interpretability approaches relying on causal interventions. To our knowledge, this investigation is the largest end-to-end attempt at reverse-engineering a natural behavior\"in the wild\"in a language model. We evaluate the reliability of our explanation using three quantitative criteria--faithfulness, completeness and minimality. Though these criteria support our explanation, they also point to remaining gaps in our understanding. Our work provides evidence that a mechanistic understanding of large ML models is feasible, opening opportunities to scale our understanding to both larger models and more complex tasks."
                    },
                    {
                        "title": "Capturing Failures of Large Language Models via Human Cognitive Biases",
                        "abstract": "Large language models generate complex, open-ended outputs: instead of outputting a class label they write summaries, generate dialogue, or produce working code. In order to asses the reliability of these open-ended generation systems, we aim to identify qualitative categories of erroneous behavior, beyond identifying individual errors. To hypothesize and test for such qualitative errors, we draw inspiration from human cognitive biases -- systematic patterns of deviation from rational judgement. Specifically, we use cognitive biases as motivation to (i) generate hypotheses for problems that models may have, and (ii) develop experiments that elicit these problems. Using code generation as a case study, we find that OpenAI's Codex errs predictably based on how the input prompt is framed, adjusts outputs towards anchors, and is biased towards outputs that mimic frequent training examples. We then use our framework to elicit high-impact errors such as incorrectly deleting files. Our results indicate that experimental methodology from cognitive science can help characterize how machine learning systems behave."
                    },
                    {
                        "title": "Discovering Latent Knowledge in Language Models Without Supervision",
                        "abstract": "Existing techniques for training language models can be misaligned with the truth: if we train models with imitation learning, they may reproduce errors that humans make; if we train them to generate text that humans rate highly, they may output errors that human evaluators can't detect. We propose circumventing this issue by directly finding latent knowledge inside the internal activations of a language model in a purely unsupervised way. Specifically, we introduce a method for accurately answering yes-no questions given only unlabeled model activations. It works by finding a direction in activation space that satisfies logical consistency properties, such as that a statement and its negation have opposite truth values. We show that despite using no supervision and no model outputs, our method can recover diverse knowledge represented in large language models: across 6 models and 10 question-answering datasets, it outperforms zero-shot accuracy by 4\\% on average. We also find that it cuts prompt sensitivity in half and continues to maintain high accuracy even when models are prompted to generate incorrect answers. Our results provide an initial step toward discovering what language models know, distinct from what they say, even when we don't have access to explicit ground truth labels."
                    }
                ]
            }
        }
    },
    "2406.02234": {
        "paper_data": {
            "title": "On the Limitations of Fractal Dimension as a Measure of Generalization",
            "url": "http://arxiv.org/abs/2406.02234v1",
            "arxiv_id": "2406.02234",
            "authors": [
                "Charlie Tan",
                "In\u00e9s Garc\u00eda-Redondo",
                "Qiquan Wang",
                "Michael M. Bronstein",
                "Anthea Monod"
            ],
            "abstract": "Bounding and predicting the generalization gap of overparameterized neural networks remains a central open problem in theoretical machine learning. Neural network optimization trajectories have been proposed to possess fractal structure, leading to bounds and generalization measures based on notions of fractal dimension on these trajectories. Prominently, both the Hausdorff dimension and the persistent homology dimension have been proposed to correlate with generalization gap, thus serving as a measure of generalization. This work performs an extended evaluation of these topological generalization measures. We demonstrate that fractal dimension fails to predict generalization of models trained from poor initializations. We further identify that the $\\ell^2$ norm of the final parameter iterate, one of the simplest complexity measures in learning theory, correlates more strongly with the generalization gap than these notions of fractal dimension. Finally, our study reveals the intriguing manifestation of model-wise double descent in persistent homology-based generalization measures. This work lays the ground for a deeper investigation of the causal relationships between fractal geometry, topological data analysis, and neural network optimization.",
            "introduction": "   1 Introduction  Deep learning enjoys widespread empirical success despite limited theoretical support. At the same time, establishing a theoretical understanding is an active area of research in learning theory. Of particular interest is the implicit bias of neural networks towards generalization despite over-parameterization (Zhang et\u00a0al., 2021). Measures from statistical learning theory, such as Rademacher complexity (Bartlett and Mendelson, 2002) and VC-Dimension (Hastie et\u00a0al., 2009), indicate that without explicit regularizationm over-parameterized models will generalize poorly. The generalization behavior of neural networks therefore requires the development of novel learning theory to study this phenomenon (Zhang et\u00a0al., 2021). Ultimately, deep learning theory seeks to define generalization bounds for given experimental configurations (Valle-P\u00e9rez and Louis, 2020), such that the generalization error of a given model can be bounded and predicted (Jiang et\u00a0al., 2019).   \u015eim\u015fekli et\u00a0al. (2019) propose a random fractal structure for neural network optimization trajectories and compute a corresponding generalization bound based on fractal dimension. However, this work required both topological and statistical conditions on the optimization trajectory as well as the learning algorithm. Subsequent work by Birdal et\u00a0al. (2021) proposes the use of persistent homology (PH) dimension (Adams et\u00a0al., 2020), a measure of fractal dimension deriving from topological data analysis (TDA), to relax these assumptions. They propose an efficient procedure for estimating PH dimension, and apply this both as a measure of generalization and as a scheme for explicit regularisation. Similarly, Dupuis et\u00a0al. (2023) developed an approach inspired by PH dimension but use a data-dependent metric which further relaxes a continuity assumption on the network loss.   This work constitutes an extended empirical evaluation of the performance and viability of these proposed topological measures of generalization; in particular, robustness and failure modes are explored in a wider range of experiments than those considered by Birdal et\u00a0al. (2021) and Dupuis et\u00a0al. (2023). We conduct an extensive statistical analysis where in particular, we compute the partial correlation, conditioning on varying hyperparameters. A central result of this work is statistical evidence that PH dimension fails to correlate with generalization induced by adversarial initialization (Zhang et\u00a0al., 2021). Our main contributions are the following:   \u2022  We provide a rigorous statistical study of the correlation results in the experimental setup of Dupuis et\u00a0al. (2023). We observe a stronger correlation between the generalization gap and the \u21132superscript\u21132\\ell^{2}roman_\u2113 start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT norm of the last iterate of weights\u2014a common measure of generalization\u2014than the fractal dimensions studied.    \u2022  We expand the statistical analyses of correlation to include partial correlation measures conditioning on the hyperparameters of the network. We observe that in most experiments, the correlation between fractal dimension and generalization gap is not statistically significant when conditioned to other hyperparameters.    \u2022  We provide experimental results demonstrating that PH dimension fails to predict models that generalize poorly due to adversarial initialization Zhang et\u00a0al. (2021).    \u2022  Where PH dimension fails to correlate with and predict generelization in a double descent situation, inspired by Nakkiran et\u00a0al. (2021). Currently, double descent is poorly understood; our experiments suggest that PH dimension may be a suitable tool for further exploration of this phenomenon.        2 Background  Following Dupuis et\u00a0al. (2023), let (\ud835\udcb5,\u2131\ud835\udcb5,\u03bc\ud835\udcb5)\ud835\udcb5subscript\u2131\ud835\udcb5subscript\ud835\udf07\ud835\udcb5(\\mathcal{Z},\\mathcal{F}_{\\mathcal{Z}},\\mu_{\\mathcal{Z}})( caligraphic_Z , caligraphic_F start_POSTSUBSCRIPT caligraphic_Z end_POSTSUBSCRIPT , italic_\u03bc start_POSTSUBSCRIPT caligraphic_Z end_POSTSUBSCRIPT )",
            "references": [
                {
                    "title": "Generalization Bounds using Data-Dependent Fractal Dimensions",
                    "abstract": "Providing generalization guarantees for modern neural networks has been a crucial task in statistical learning. Recently, several studies have attempted to analyze the generalization error in such settings by using tools from fractal geometry. While these works have successfully introduced new mathematical tools to apprehend generalization, they heavily rely on a Lipschitz continuity assumption, which in general does not hold for neural networks and might make the bounds vacuous. In this work, we address this issue and prove fractal geometry-based generalization bounds without requiring any Lipschitz assumption. To achieve this goal, we build up on a classical covering argument in learning theory and introduce a data-dependent fractal dimension. Despite introducing a significant amount of technical complications, this new notion lets us control the generalization error (over either fixed or random hypothesis spaces) along with certain mutual information (MI) terms. To provide a clearer interpretation to the newly introduced MI terms, as a next step, we introduce a notion of\"geometric stability\"and link our bounds to the prior art. Finally, we make a rigorous connection between the proposed data-dependent dimension and topological data analysis tools, which then enables us to compute the dimension in a numerically efficient way. We support our theory with experiments conducted on various settings."
                },
                {
                    "title": "Discrete Morse Sandwich: Fast Computation of Persistence Diagrams for Scalar Data \u2013 An Algorithm and a Benchmark",
                    "abstract": "This paper introduces an efficient algorithm for persistence diagram computation, given an input piecewise linear scalar field <inline-formula><tex-math notation=\"LaTeX\">$f$</tex-math><alternatives><mml:math><mml:mi>f</mml:mi></mml:math><inline-graphic xlink:href=\"tierny-ieq1-3238008.gif\"/></alternatives></inline-formula> defined on a <inline-formula><tex-math notation=\"LaTeX\">$d$</tex-math><alternatives><mml:math><mml:mi>d</mml:mi></mml:math><inline-graphic xlink:href=\"tierny-ieq2-3238008.gif\"/></alternatives></inline-formula>-dimensional simplicial complex <inline-formula><tex-math notation=\"LaTeX\">$\\mathcal {K}$</tex-math><alternatives><mml:math><mml:mi mathvariant=\"script\">K</mml:mi></mml:math><inline-graphic xlink:href=\"tierny-ieq3-3238008.gif\"/></alternatives></inline-formula>, with <inline-formula><tex-math notation=\"LaTeX\">$d \\leq 3$</tex-math><alternatives><mml:math><mml:mrow><mml:mi>d</mml:mi><mml:mo>\u2264</mml:mo><mml:mn>3</mml:mn></mml:mrow></mml:math><inline-graphic xlink:href=\"tierny-ieq4-3238008.gif\"/></alternatives></inline-formula>. Our work revisits the seminal algorithm <italic>\u201cPairSimplices\u201d</italic> (Edelsbrunner et al. 2002), (Zomorodian, 2010) with discrete Morse theory (DMT) (Forman, 1998), (Robins et al. 2011), which greatly reduces the number of input simplices to consider. Further, we also extend to DMT and accelerate the stratification strategy described in <italic>\u201cPairSimplices\u201d</italic> (Edelsbrunner et al. 2002), (Zomorodian, 2010) for the fast computation of the <inline-formula><tex-math notation=\"LaTeX\">$0^{th}$</tex-math><inline-graphic xlink:href=\"tierny-ieq5-3238008.gif\"/></inline-formula> and <inline-formula><tex-math notation=\"LaTeX\">$(d-1)^{th}$</tex-math><alternatives><mml:math><mml:msup><mml:mrow><mml:mo>(</mml:mo><mml:mi>d</mml:mi><mml:mo>-</mml:mo><mml:mn>1</mml:mn><mml:mo>)</mml:mo></mml:mrow><mml:mrow><mml:mi>t</mml:mi><mml:mi>h</mml:mi></mml:mrow></mml:msup></mml:math><inline-graphic xlink:href=\"tierny-ieq6-3238008.gif\"/></alternatives></inline-formula> diagrams, noted <inline-formula><tex-math notation=\"LaTeX\">$\\mathcal {D}_{0}(f)$</tex-math><alternatives><mml:math><mml:mrow><mml:msub><mml:mi mathvariant=\"script\">D</mml:mi><mml:mn>0</mml:mn></mml:msub><mml:mrow><mml:mo>(</mml:mo><mml:mi>f</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math><inline-graphic xlink:href=\"tierny-ieq7-3238008.gif\"/></alternatives></inline-formula> and <inline-formula><tex-math notation=\"LaTeX\">$\\mathcal {D}_{d-1}(f)$</tex-math><alternatives><mml:math><mml:mrow><mml:msub><mml:mi mathvariant=\"script\">D</mml:mi><mml:mrow><mml:mi>d</mml:mi><mml:mo>-</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msub><mml:mrow><mml:mo>(</mml:mo><mml:mi>f</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math><inline-graphic xlink:href=\"tierny-ieq8-3238008.gif\"/></alternatives></inline-formula>. Minima-saddle persistence pairs (<inline-formula><tex-math notation=\"LaTeX\">$\\mathcal {D}_{0}(f)$</tex-math><alternatives><mml:math><mml:mrow><mml:msub><mml:mi mathvariant=\"script\">D</mml:mi><mml:mn>0</mml:mn></mml:msub><mml:mrow><mml:mo>(</mml:mo><mml:mi>f</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math><inline-graphic xlink:href=\"tierny-ieq9-3238008.gif\"/></alternatives></inline-formula>) and saddle-maximum persistence pairs (<inline-formula><tex-math notation=\"LaTeX\">$\\mathcal {D}_{d-1}(f)$</tex-math><alternatives><mml:math><mml:mrow><mml:msub><mml:mi mathvariant=\"script\">D</mml:mi><mml:mrow><mml:mi>d</mml:mi><mml:mo>-</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msub><mml:mrow><mml:mo>(</mml:mo><mml:mi>f</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math><inline-graphic xlink:href=\"tierny-ieq10-3238008.gif\"/></alternatives></inline-formula>) are efficiently computed by processing, with a Union-Find, the unstable sets of 1-saddles and the stable sets of <inline-formula><tex-math notation=\"LaTeX\">$(d-1)$</tex-math><alternatives><mml:math><mml:mrow><mml:mo>(</mml:mo><mml:mi>d</mml:mi><mml:mo>-</mml:mo><mml:mn>1</mml:mn><mml:mo>)</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href=\"tierny-ieq11-3238008.gif\"/></alternatives></inline-formula>-saddles. This fast pre-computation for the dimensions 0 and <inline-formula><tex-math notation=\"LaTeX\">$(d-1)$</tex-math><alternatives><mml:math><mml:mrow><mml:mo>(</mml:mo><mml:mi>d</mml:mi><mml:mo>-</mml:mo><mml:mn>1</mml:mn><mml:mo>)</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href=\"tierny-ieq12-3238008.gif\"/></alternatives></inline-formula> enables an aggressive specialization of (Bauer et al. 2014) to the 3D case, which results in a drastic reduction of the number of input simplices for the computation of <inline-formula><tex-math notation=\"LaTeX\">$\\mathcal {D}_{1}(f)$</tex-math><alternatives><mml:math><mml:mrow><mml:msub><mml:mi mathvariant=\"script\">D</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mrow><mml:mo>(</mml:mo><mml:mi>f</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math><inline-graphic xlink:href=\"tierny-ieq13-3238008.gif\"/></alternatives></inline-formula>, the intermediate layer of the <italic>sandwich</italic>. Finally, we document several performance improvements via shared-memory parallelism. We provide an open-source implementation of our algorithm for reproducibility purposes. Extensive experiments indicate that our algorithm improves by two orders of magnitude the time performance of the seminal <italic>\u201cPairSimplices\u201d</italic> algorithm it extends. Moreover, it also improves memory footprint and time performance over a selection of 14 competing approaches, with a substantial gain over the fastest available approaches, while producing a strictly identical output."
                },
                {
                    "title": "Intrinsic Dimension, Persistent Homology and Generalization in Neural Networks",
                    "abstract": "Disobeying the classical wisdom of statistical learning theory, modern deep neural networks generalize well even though they typically contain millions of parameters. Recently, it has been shown that the trajectories of iterative optimization algorithms can possess fractal structures, and their generalization error can be formally linked to the complexity of such fractals. This complexity is measured by the fractal's intrinsic dimension, a quantity usually much smaller than the number of parameters in the network. Even though this perspective provides an explanation for why overparametrized networks would not overfit, computing the intrinsic dimension (e.g., for monitoring generalization during training) is a notoriously difficult task, where existing methods typically fail even in moderate ambient dimensions. In this study, we consider this problem from the lens of topological data analysis (TDA) and develop a generic computational tool that is built on rigorous mathematical foundations. By making a novel connection between learning theory and TDA, we first illustrate that the generalization error can be equivalently bounded in terms of a notion called the 'persistent homology dimension' (PHD), where, compared with prior work, our approach does not require any additional geometrical or statistical assumptions on the training dynamics. Then, by utilizing recently established theoretical results and TDA tools, we develop an efficient algorithm to estimate PHD in the scale of modern deep neural networks and further provide visualization tools to help understand generalization in deep learning. Our experiments show that the proposed approach can efficiently compute a network's intrinsic dimension in a variety of settings, which is predictive of the generalization error."
                },
                {
                    "title": "Fractal Structure and Generalization Properties of Stochastic Optimization Algorithms",
                    "abstract": "Understanding generalization in deep learning has been one of the major challenges in statistical learning theory over the last decade. While recent work has illustrated that the dataset and the training algorithm must be taken into account in order to obtain meaningful generalization bounds, it is still theoretically not clear which properties of the data and the algorithm determine the generalization performance. In this study, we approach this problem from a dynamical systems theory perspective and represent stochastic optimization algorithms as random iterated function systems (IFS). Well studied in the dynamical systems literature, under mild assumptions, such IFSs can be shown to be ergodic with an invariant measure that is often supported on sets with a fractal structure. As our main contribution, we prove that the generalization error of a stochastic optimization algorithm can be bounded based on the `complexity' of the fractal structure that underlies its invariant measure. Leveraging results from dynamical systems theory, we show that the generalization error can be explicitly linked to the choice of the algorithm (e.g., stochastic gradient descent -- SGD), algorithm hyperparameters (e.g., step-size, batch-size), and the geometry of the problem (e.g., Hessian of the loss). We further specialize our results to specific problems (e.g., linear/logistic regression, one hidden-layered neural networks) and algorithms (e.g., SGD and preconditioned variants), and obtain analytical estimates for our bound.For modern neural networks, we develop an efficient algorithm to compute the developed bound and support our theory with various experiments on neural networks."
                },
                {
                    "title": "Understanding deep learning (still) requires rethinking generalization",
                    "abstract": "Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small gap between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family or to the regularization techniques used during training. Through extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. We interpret our experimental findings by comparison with traditional models. We supplement this republication with a new section at the end summarizing recent progresses in the field since the original version of this paper."
                },
                {
                    "title": "Generalization bounds for deep learning",
                    "abstract": "Generalization in deep learning has been the topic of much recent theoretical and empirical research. Here we introduce desiderata for techniques that predict generalization errors for deep learning models in supervised learning. Such predictions should 1) scale correctly with data complexity; 2) scale correctly with training set size; 3) capture differences between architectures; 4) capture differences between optimization algorithms; 5) be quantitatively not too far from the true error (in particular, be non-vacuous); 6) be efficiently computable; and 7) be rigorous. We focus on generalization error upper bounds, and introduce a categorisation of bounds depending on assumptions on the algorithm and data. We review a wide range of existing approaches, from classical VC dimension to recent PAC-Bayesian bounds, commenting on how well they perform against the desiderata. We next use a function-based picture to derive a marginal-likelihood PAC-Bayesian bound. This bound is, by one definition, optimal up to a multiplicative constant in the asymptotic limit of large training sets, as long as the learning curve follows a power law, which is typically found in practice for deep learning problems. Extensive empirical analysis demonstrates that our marginal-likelihood PAC-Bayes bound fulfills desiderata 1-3 and 5. The results for 6 and 7 are promising, but not yet fully conclusive, while only desideratum 4 is currently beyond the scope of our bound. Finally, we comment on why this function-based bound performs significantly better than current parameter-based PAC-Bayes bounds."
                },
                {
                    "title": "Hausdorff dimension, heavy tails, and generalization in neural networks",
                    "abstract": "Despite its success in a wide range of applications, characterizing the generalization properties of stochastic gradient descent (SGD) in non-convex deep learning problems is still an important challenge. While modeling the trajectories of SGD via stochastic differential equations (SDE) under heavy-tailed gradient noise has recently shed light over several peculiar characteristics of SGD, a rigorous treatment of the generalization properties of such SDEs in a learning theoretical framework is still missing. Aiming to bridge this gap, in this paper, we prove generalization bounds for SGD under the assumption that its trajectories can be well-approximated by a Feller process, which defines a rich class of Markov processes that include several recent SDE representations (both Brownian or heavy-tailed) as its special case. We show that the generalization error can be controlled by the Hausdorff dimension of the trajectories, which is intimately linked to the tail behavior of the driving process. Our results imply that heavier-tailed processes should achieve better generalization; hence, the tail-index of the process can be used as a notion of \u2018capacity metric\u2019. We support our theory with experiments on deep neural networks illustrating that the proposed capacity metric accurately estimates the generalization error, and it does not necessarily grow with the number of parameters unlike the existing capacity metrics in the literature."
                },
                {
                    "title": "giotto-tda: A Topological Data Analysis Toolkit for Machine Learning and Data Exploration",
                    "abstract": "We introduce giotto-tda, a Python library that integrates high-performance topological data analysis with machine learning via a scikit-learn-compatible API and state-of-the-art C++ implementations. The library's ability to handle various types of data is rooted in a wide range of preprocessing techniques, and its strong focus on data exploration and interpretability is aided by an intuitive plotting API. Source code, binaries, examples, and documentation can be found at this https URL."
                },
                {
                    "title": "Deep double descent: where bigger models and more data hurt",
                    "abstract": "We show that a variety of modern deep learning tasks exhibit a \u2018double-descent\u2019 phenomenon where, as we increase model size, performance first gets worse and then gets better. Moreover, we show that double descent occurs not just as a function of model size, but also as a function of the number of training epochs. We unify the above phenomena by defining a new complexity measure we call the effective model complexity and conjecture a generalized double descent with respect to this measure. Furthermore, our notion of model complexity allows us to identify certain regimes where increasing (even quadrupling) the number of train samples actually hurts test performance."
                },
                {
                    "title": "Fantastic Generalization Measures and Where to Find Them",
                    "abstract": "Generalization of deep networks has been of great interest in recent years, resulting in a number of theoretically and empirically motivated complexity measures. However, most papers proposing such measures study only a small set of models, leaving open the question of whether the conclusion drawn from those experiments would remain valid in other settings. We present the first large scale study of generalization in deep networks. We investigate more then 40 complexity measures taken from both theoretical bounds and empirical studies. We train over 10,000 convolutional networks by systematically varying commonly used hyperparameters. Hoping to uncover potentially causal relationships between each measure and generalization, we analyze carefully controlled experiments and show surprising failures of some measures as well as promising measures for further research."
                },
                {
                    "title": "On the Heavy-Tailed Theory of Stochastic Gradient Descent for Deep Neural Networks",
                    "abstract": "The gradient noise (GN) in the stochastic gradient descent (SGD) algorithm is often considered to be Gaussian in the large data regime by assuming that the \\emph{classical} central limit theorem (CLT) kicks in. This assumption is often made for mathematical convenience, since it enables SGD to be analyzed as a stochastic differential equation (SDE) driven by a Brownian motion. We argue that the Gaussianity assumption might fail to hold in deep learning settings and hence render the Brownian motion-based analyses inappropriate. Inspired by non-Gaussian natural phenomena, we consider the GN in a more general context and invoke the \\emph{generalized} CLT, which suggests that the GN converges to a \\emph{heavy-tailed} $\\alpha$-stable random vector, where \\emph{tail-index} $\\alpha$ determines the heavy-tailedness of the distribution. Accordingly, we propose to analyze SGD as a discretization of an SDE driven by a Levy motion. Such SDEs can incur `jumps', which force the SDE and its discretization \\emph{transition} from narrow minima to wider minima, as proven by existing metastability theory and the extensions that we proved recently. In this study, under the $\\alpha$-stable GN assumption, we further establish an explicit connection between the convergence rate of SGD to a local minimum and the tail-index $\\alpha$. To validate the $\\alpha$-stable assumption, we conduct experiments on common deep learning scenarios and show that in all settings, the GN is highly non-Gaussian and admits heavy-tails. We investigate the tail behavior in varying network architectures and sizes, loss functions, and datasets. Our results open up a different perspective and shed more light on the belief that SGD prefers wide minima."
                },
                {
                    "title": "Bad Global Minima Exist and SGD Can Reach Them",
                    "abstract": "Several recent works have aimed to explain why severely overparameterized models, generalize well when trained by Stochastic Gradient Descent (SGD). The emergent consensus explanation has two parts: the first is that there are \"no bad local minima\", while the second is that SGD performs implicit regularization by having a bias towards low complexity models. We revisit both of these ideas in the context of image classification with common deep neural network architectures. Our first finding is that there exist bad global minima, i.e., models that fit the training set perfectly, yet have poor generalization. Our second finding is that given only unlabeled training data, we can easily construct initializations that will cause SGD to quickly converge to such bad global minima. For example, on CIFAR, CINIC10, and (Restricted) ImageNet, this can be achieved by starting SGD at a model derived by fitting random labels on the training data: while subsequent SGD training (with the correct labels) will reach zero training error, the resulting model will exhibit a test accuracy degradation of up to 40% compared to training from a random initialization. Finally, we show that regularization seems to provide SGD with an escape route: once heuristics such as data augmentation are used, starting from a complex model (adversarial initialization) has no effect on the test accuracy."
                },
                {
                    "title": "Persistence Theory - From Quiver Representations to Data Analysis",
                    "abstract": "* Theoretical foundations: Algebraic persistence* Topological persistence* Stability Applications: Topological inference* Topological inference 2.0* Clustering Signatures for metric spaces* Perspectives: New trends in topological data analysis* Further prospects on the theory* Introduction to quiver theory with a view toward persistence* Bibliography* List of figures* Index"
                },
                {
                    "title": "Distributed Computation of Persistent Homology",
                    "abstract": "Persistent homology is a popular and powerful tool for capturing topological features of data. Advances in algorithms for computing persistent homology have reduced the computation time drastically -- as long as the algorithm does not exhaust the available memory. Following up on a recently presented parallel method for persistence computation on shared memory systems [1], we demonstrate that a simple adaption of the standard reduction algorithm leads to a variant for distributed systems. Our algorithmic design ensures that the data is distributed over the nodes without redundancy; this permits the computation of much larger instances than on a single machine. Moreover, we observe that the parallelism at least compensates for the overhead caused by communication between nodes, and often even speeds up the computation compared to sequential and even parallel shared memory algorithms. In our experiments, we were able to compute the persistent homology of filtrations with more than a billion (109) elements within seconds on a cluster with 32 nodes using less than 6GB of memory per node."
                },
                {
                    "title": "Spearman Correlation Coefficients, Differences between",
                    "abstract": "Testing the equality of two population correlation coefficients when the data are bivariate normal and Pearson correlation coefficients are used as estimates of the population parameters is a straightforward procedure covered in many introductory statistics courses. The coefficients are converted using Fisher's z-transformation with standard errors (N\u00a0\u2212\u00a03)\u22121/2. The two transformed values are then compared using a standard normal procedure. When data are not bivariate normal, Spearman's correlation coefficient rho is often used as the index of correlation. Comparison of two Spearman rhos is not as well documented. Three approaches were investigated using Monte Carlo simulations. Treating the Spearman coefficients as though they were Pearson coefficients and using the standard Fisher's z-transformation and subsequent comparison was more robust with respect to Type I error than either ignoring the nonnormality and computing Pearson coefficients or converting the Spearman coefficients to Pearson equivalents prior to transformation. \n \n \nKeywords: \n \ncorrelation coefficient; \npearson correlation; \nspearman correlation; \nFisher z-transformation"
                },
                {
                    "title": "Computing Persistent Homology",
                    "abstract": "Abstract\nWe show that the persistent homology of a filtered d-dimensional\nsimplicial complex is simply the standard homology of a \nparticular graded module over a polynomial ring.\nOur analysis establishes the existence of a simple description of\npersistent homology groups over arbitrary fields.\nIt also enables us to derive a natural\nalgorithm for computing persistent homology of spaces in \narbitrary dimension over any field.\nThis result generalizes and extends the previously known\nalgorithm that was restricted to subcomplexes of S3 and\nZ2 coefficients.\nFinally, our study implies the lack of a simple \nclassification over non-fields.\nInstead, we give an algorithm for computing individual\npersistent homology groups over an arbitrary principal ideal domain \nin any dimension.\n\n"
                },
                {
                    "title": "The Elements of Statistical Learning: Data Mining, Inference, and Prediction",
                    "abstract": "In the words of the authors, the goal of this book was to \u201cbring together many of the important new ideas in learning, and explain them in a statistical framework.\u201d The authors have been quite successful in achieving this objective, and their work is a welcome addition to the statistics and learning literatures. Statistics has always been interdisciplinary, borrowing ideas from diverse \u008e elds and repaying the debt with contributions, both theoretical and practical, to the other intellectual disciplines. For statistical learning, this cross-fertilization is especially noticeable. This book is a valuable resource, both for the statistician needing an introduction to machine learning and related \u008e elds and for the computer scientist wishing to learn more about statistics. Statisticians will especially appreciate that it is written in their own language. The level of the book is roughly that of a second-year doctoral student in statistics, and it will be useful as a textbook for such students. In a stimulating article, Breiman (2001) argued that statistics has been focused too much on a \u201cdata modeling culture,\u201d where the model is paramount. Breiman argued instead for an \u201calgorithmic modeling culture,\u201d with emphasis on black-box types of prediction. Breiman\u2019s article is controversial, and in his discussion, Efron objects that \u201cprediction is certainly an interesting subject, but Leo\u2019s paper overstates both its role and our profession\u2019s lack of interest in it.\u201d Although I mostly agree with Efron, I worry that the courses offered by most statistics departments include little, if any, treatment of statistical learning and prediction. (Stanford, where Efron and the authors of this book teach, is an exception.) Graduate students in statistics certainly need to know more than they do now about prediction, machine learning, statistical learning, and data mining (not disjoint subjects). I hope that graduate courses covering the topics of this book will become more common in statistics curricula. Most of the book is focused on supervised learning, where one has inputs and outputs from some system and wishes to predict unknown outputs corresponding to known inputs. The methods discussed for supervised learning include linear and logistic regression; basis expansion, such as splines and wavelets; kernel techniques, such as local regression, local likelihood, and radial basis functions; neural networks; additive models; decision trees based on recursive partitioning, such as CART; and support vector machines. There is a \u008e nal chapter on unsupervised learning, including association rules, cluster analysis, self-organizing maps, principal components and curves, and independent component analysis. Many statisticians will be unfamiliar with at least some of these algorithms. Association rules are popular for mining commercial data in what is called \u201cmarket basket analysis.\u201d The aim is to discover types of products often purchased together. Such knowledge can be used to develop marketing strategies, such as store or catalog layouts. Self-organizing maps (SOMs) involve essentially constrained k-means clustering, where prototypes are mapped to a two-dimensional curved coordinate system. Independent components analysis is similar to principal components analysis and factor analysis, but it uses higher-order moments to achieve independence, not merely zero correlation between components. A strength of the book is the attempt to organize a plethora of methods into a coherent whole. The relationships among the methods are emphasized. I know of no other book that covers so much ground. Of course, with such broad coverage, it is not possible to cover any single topic in great depth, so this book will encourage further reading. Fortunately, each chapter includes bibliographic notes surveying the recent literature. These notes and the extensive references provide a good introduction to the learning literature, including much outside of statistics. The book might be more suitable as a textbook if less material were covered in greater depth; however, such a change would compromise the book\u2019s usefulness as a reference, and so I am happier with the book as it was written."
                },
                {
                    "title": "The minimal spanning tree and the upper box dimension",
                    "abstract": "We show that the \u03b1-weight of an MST over n points in a metric space with upper box dimension d has a bound independent of n if a > d and does not have one if \u03b1 < d."
                },
                {
                    "title": "Random Fractals and Markov Processes",
                    "abstract": "This is a survey on the sample path properties of Markov processes, especially fractal properties of the random sets and measures determined by their sample paths. The class of Markov processes considered in this paper include Levy processes in Rd, diffusions on fractals and on Rd, Feller processes determined by pseudo-differential operators and so on. We summarize recent results for Levy processes such as the Hausdorff and packing dimensions of their ranges, level sets, and multiple points; regularity properties of local times and self-intersection local times; multifractal analysis of the occupation measures and sample paths. Our emphasis is on general Markovian techniques that are applicable to other Markov processes. Appeared in: Fractal Geometry and Applications: A Jubilee of Benoit Mandelbrot, (Michel L. Lapidus and Machiel van Frankenhuijsen, editors), pp. 261\u2013338, American Mathematical Society, 2004."
                },
                {
                    "title": "Fractal Geometry: Mathematical Foundations and Applications",
                    "abstract": "Part I Foundations: mathematical background Hausdorff measure and dimension alternative definitions of dimension techniques for calculating dimensions local structure of fractals projections of fractals products of fractals intersections of fractals. Part II Applications and examples: fractals defined by transformations examples from number theory graphs of functions examples from pure mathematics dynamical systems iteration of complex functions-Julia sets random fractals Brownian motion and Brownian surfaces multifractal measures physical applications."
                },
                {
                    "title": "Control Batch Size and Learning Rate to Generalize Well: Theoretical and Empirical Evidence",
                    "abstract": "Deep neural networks have received dramatic success based on the optimization method of stochastic gradient descent (SGD). However, it is still not clear how to tune hyper-parameters, especially batch size and learning rate, to ensure good generalization. This paper reports both theoretical and empirical evidence of a training strategy that we should control the ratio of batch size to learning rate not too large to achieve a good generalization ability. Specifically, we prove a PAC-Bayes generalization bound for neural networks trained by SGD, which has a positive correlation with the ratio of batch size to learning rate. This correlation builds the theoretical foundation of the training strategy. Furthermore, we conduct a large-scale experiment to verify the correlation and training strategy. We trained 1,600 models based on architectures ResNet-110, and VGG-19 with datasets CIFAR-10 and CIFAR-100 while strictly control unrelated variables. Accuracies on the test sets are collected for the evaluation. Spearman's rank-order correlation coefficients and the corresponding $p$ values on 164 groups of the collected data demonstrate that the correlation is statistically significant, which fully supports the training strategy."
                },
                {
                    "title": "Persistent Homology Computation with a Twist",
                    "abstract": "The persistence diagram of a filtered simplicial complex is usually computed by reducing the boundary matrix of the complex. We introduce a simple optimization technique: by processing the simplices of the complex in decreasing dimension, we can \u201ckill\u201d columns (i.e., set them to zero) without reducing them. This technique completely avoids reduction on roughly half of the columns. We demonstrate that this idea significantly improves the running time of the reduction algorithm in practice. We also give an output-sensitive complexity analysis for the new algorithm which yields to sub-cubic asymptotic bounds under certain assumptions."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Gradient-based learning applied to document recognition",
                    "abstract": "Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "math.DS",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively evaluate and improve the correlation between topological measures, specifically persistent homology (PH) dimension, and the generalization performance of neural networks, particularly in the context of adversarial initialization and double descent phenomena?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the theoretical understanding of generalization in deep learning, which remains an area of active research. By establishing a clearer relationship between topological measures and generalization performance, we can enhance the design of neural networks and their training processes. This could lead to more robust models that generalize better in practical applications, ultimately influencing future research directions in learning theory and model optimization.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem stem from the complex nature of generalization in over-parameterized models, where naive approaches may overlook the intricate relationships between model architecture, training dynamics, and generalization metrics. Specifically, the reliance on fractal dimensions and their statistical properties introduces significant theoretical and practical obstacles, such as the need for rigorous statistical validation and the potential for misleading correlations that do not hold under varying hyperparameters or adversarial conditions.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on establishing theoretical frameworks without extensive empirical validation, leading to gaps in understanding the practical implications of topological measures like PH dimension. Barriers include the lack of comprehensive statistical analyses across diverse experimental setups and the limited exploration of how these measures interact with adversarial initialization and double descent phenomena. Our approach aims to fill these gaps by providing a more rigorous statistical evaluation and exploring the conditions under which PH dimension correlates with generalization.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves conducting an extensive statistical analysis of the correlation between PH dimension and generalization performance across various neural network configurations. We will utilize a diverse dataset and employ metrics such as the \u21132 norm of the last iterate of weights and generalization gap to evaluate performance. Expected outcomes include a clearer understanding of the limitations of PH dimension in predicting generalization, particularly in adversarial scenarios, and insights into the double descent phenomenon, which could inform future research and model design strategies."
            }
        },
        "author_data": {
            "6532b485-868d-4577-a967-ed4a6bbced9f": {
                "pk": "6532b485-868d-4577-a967-ed4a6bbced9f",
                "name": "Charlie Tan",
                "collaborators": [
                    "Theodore Long",
                    "Sarah Zhao",
                    "Rudolf Laine",
                    "Chen Feng",
                    "Duolikun Danier",
                    "Fan Zhang",
                    "David Bull"
                ],
                "domain": [
                    "Information Geometry",
                    "Deep Learning",
                    "Video Compression",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Geodesic Mode Connectivity",
                        "abstract": "Mode connectivity is a phenomenon where trained models are connected by a path of low loss. We reframe this in the context of Information Geometry, where neural networks are studied as spaces of parameterized distributions with curved geometry. We hypothesize that shortest paths in these spaces, known as geodesics, correspond to mode-connecting paths in the loss landscape. We propose an algorithm to approximate geodesics and demonstrate that they achieve mode connectivity."
                    },
                    {
                        "title": "ViSTRA3: Video Coding with Deep Parameter Adaptation and Post Processing",
                        "abstract": "This paper presents a deep learning-based video compression framework (ViSTRA3). The proposed framework intelligently adapts video format parameters of the input video before encoding, subsequently employing a CNN at the decoder to restore their original format and enhance reconstruction quality. ViSTRA3 has been integrated with the H.266/VVC Test Model VTM 14.0, and evaluated under the Joint Video Exploration Team Common Test Conditions. Bj{\\o}negaard Delta (BD) measurement results show that the proposed framework consistently outperforms the original VVC VTM, with average BD-rate savings of 1.8% and 3.7% based on the assessment of PSNR and VMAF."
                    }
                ]
            },
            "495a6fbb-0b9a-4e97-9a78-74f4ef0491a4": {
                "pk": "495a6fbb-0b9a-4e97-9a78-74f4ef0491a4",
                "name": "In\u00e9s Garc\u00eda-Redondo",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "142d8e1f-a44d-4fd8-b58a-2724fcf6eec7": {
                "pk": "142d8e1f-a44d-4fd8-b58a-2724fcf6eec7",
                "name": "Qiquan Wang",
                "collaborators": [
                    "Anthea Monod",
                    "Anna Song",
                    "Antoniana Batsivari",
                    "Dominique Bonnet",
                    "In\u00e9s Garc\u00eda-Redondo",
                    "Pierre Faug\u00e8re",
                    "Gregory Henselman-Petrusek"
                ],
                "domain": [
                    "Topological Data Analysis",
                    "Machine Learning",
                    "Functional Data Analysis",
                    "Medical Imaging"
                ],
                "publications": [
                    {
                        "title": "A Topological Gaussian Mixture Model for Bone Marrow Morphology in Leukaemia",
                        "abstract": "Acute myeloid leukaemia (AML) is a type of blood and bone marrow cancer characterized by the proliferation of abnormal clonal haematopoietic cells in the bone marrow leading to bone marrow failure. Over the course of the disease, angiogenic factors released by leukaemic cells drastically alter the bone marrow vascular niches resulting in observable structural abnormalities. We use a technique from topological data analysis - persistent homology - to quantify the images and infer on the disease through the imaged morphological features. We find that persistent homology uncovers succinct dissimilarities between the control, early, and late stages of AML development. We then integrate persistent homology into stage-dependent Gaussian mixture models for the first time, proposing a new class of models which are applicable to persistent homology summaries and able to both infer patterns in morphological changes between different stages of progression as well as provide a basis for prediction."
                    },
                    {
                        "title": "Stability for Inference with Persistent Homology Rank Functions",
                        "abstract": "Persistent homology barcodes and diagrams are a cornerstone of topological data analysis that capture the \"shape\" of a wide range of complex data structures, such as point clouds, networks, and functions. However, their use in statistical settings is challenging due to their complex geometric structure. In this paper, we revisit the persistent homology rank function, which is mathematically equivalent to a barcode and persistence diagram, as a tool for statistics and machine learning. Rank functions, being functions, enable the direct application of the statistical theory of functional data analysis (FDA)-a domain of statistics adapted for data in the form of functions. A key challenge they present over barcodes in practice, however, is their lack of stability-a property that is crucial to validate their use as a faithful representation of the data and therefore a viable summary statistic. In this paper, we fill this gap by deriving two stability results for persistent homology rank functions under a suitable metric for FDA integration. We then study the performance of rank functions in functional inferential statistics and machine learning on real data applications, in both single and multiparameter persistent homology. We find that the use of persistent homology captured by rank functions offers a clear improvement over existing non-persistence-based approaches."
                    }
                ]
            },
            "5a008179-163e-42da-b12e-4ad1fc666e38": {
                "pk": "5a008179-163e-42da-b12e-4ad1fc666e38",
                "name": "Michael M. Bronstein",
                "collaborators": [
                    "Klaus Glashoff",
                    "Alexander M. Bronstein",
                    "Ron Kimmel",
                    "Alex M. Bronstein",
                    "Artiom Kovnatsky",
                    "Dan Raviv",
                    "Nir Sochen",
                    "Federico Monti",
                    "Roee Litman",
                    "Terry A. Loring"
                ],
                "domain": [
                    "Multimodal Learning",
                    "Shape Analysis",
                    "Graph Neural Network",
                    "Optimization"
                ],
                "publications": [
                    {
                        "title": "Multimodal diff-hash",
                        "abstract": "Many applications require comparing multimodal data with different structure and dimensionality that cannot be compared directly. Recently, there has been increasing interest in methods for learning and efficiently representing such multimodal similarity. In this paper, we present a simple algorithm for multimodal similarity-preserving hashing, trying to map multimodal data into the Hamming space while preserving the intra- and inter-modal similarities. We show that our method significantly outperforms the state-of-the-art method in the field."
                    },
                    {
                        "title": "Heat kernel coupling for multiple graph analysis",
                        "abstract": "In this paper, we introduce heat kernel coupling (HKC) as a method of constructing multimodal spectral geometry on weighted graphs of different size without vertex-wise bijective correspondence. We show that Laplacian averaging can be derived as a limit case of HKC, and demonstrate its applications on several problems from the manifold learning and pattern recognition domain."
                    },
                    {
                        "title": "Diffusion-geometric maximally stable component detection in deformable shapes",
                        "abstract": "Maximally stable component detection is a very popular method for feature analysis in images, mainly due to its low computation cost and high repeatability. With the recent advance of feature-based methods in geometric shape analysis, there is significant interest in finding analogous approaches in the 3D world. In this paper, we formulate a diffusion-geometric framework for stable component detection in non-rigid 3D shapes, which can be used for geometric feature detection and description. A quantitative evaluation of our method on the SHREC'10 feature detection benchmark shows its potential as a source of high-quality features."
                    },
                    {
                        "title": "Diffusion framework for geometric and photometric data fusion in non-rigid shape analysis",
                        "abstract": "In this paper, we explore the use of the diffusion geometry framework for the fusion of geometric and photometric information in local and global shape descriptors. Our construction is based on the definition of a diffusion process on the shape manifold embedded into a high-dimensional space where the embedding coordinates represent the photometric information. Experimental results show that such data fusion is useful in coping with different challenges of shape analysis where pure geometric and pure photometric methods fail."
                    },
                    {
                        "title": "Optimization on the biorthogonal manifold",
                        "abstract": "In this paper, we consider optimization problems w.r.t. to pairs of orthogonal matrices $XY = I$. Problems of this form arise in several applications such as finding shape correspondence in computer graphics. We show that the space of such matrices is a Riemannian manifold, which we call the biorthogonal manifold. To our knowledge, this manifold has not been studied before. We give expressions of tangent space projection, exponential map, and retraction operators of the biorthogonal manifold, and discuss their numerical implementation."
                    },
                    {
                        "title": "Making Laplacians commute",
                        "abstract": "In this paper, we construct multimodal spectral geometry by finding a pair of closest commuting operators (CCO) to a given pair of Laplacians. The CCOs are jointly diagonalizable and hence have the same eigenbasis. Our construction naturally extends classical data analysis tools based on spectral geometry, such as diffusion maps and spectral clustering. We provide several synthetic and real examples of applications in dimensionality reduction, shape analysis, and clustering, demonstrating that our method better captures the inherent structure of multi-modal data."
                    },
                    {
                        "title": "The Video Genome",
                        "abstract": "Fast evolution of Internet technologies has led to an explosive growth of video data available in the public domain and created unprecedented challenges in the analysis, organization, management, and control of such content. The problems encountered in video analysis such as identifying a video in a large database (e.g. detecting pirated content in YouTube), putting together video fragments, finding similarities and common ancestry between different versions of a video, have analogous counterpart problems in genetic research and analysis of DNA and protein sequences. In this paper, we exploit the analogy between genetic sequences and videos and propose an approach to video analysis motivated by genomic research. Representing video information as video DNA sequences and applying bioinformatic algorithms allows to search, match, and compare videos in large-scale databases. We show an application for content-based metadata mapping between versions of annotated video."
                    },
                    {
                        "title": "A correspondence-less approach to matching of deformable shapes",
                        "abstract": "Finding a match between partially available deformable shapes is a challenging problem with numerous applications. The problem is usually approached by computing local descriptors on a pair of shapes and then establishing a point-wise correspondence between the two. In this paper, we introduce an alternative correspondence-less approach to matching fragments to an entire shape undergoing a non-rigid deformation. We use diffusion geometric descriptors and optimize over the integration domains on which the integral descriptors of the two parts match. The problem is regularized using the Mumford-Shah functional. We show an efficient discretization based on the Ambrosio-Tortorelli approximation generalized to triangular meshes. Experiments demonstrating the success of the proposed method are presented."
                    },
                    {
                        "title": "Affine-invariant diffusion geometry for the analysis of deformable 3D shapes",
                        "abstract": "We introduce an (equi-)affine invariant diffusion geometry by which surfaces that go through squeeze and shear transformations can still be properly analyzed. The definition of an affine invariant metric enables us to construct an invariant Laplacian from which local and global geometric structures are extracted. Applications of the proposed framework demonstrate its power in generalizing and enriching the existing set of tools for shape analysis."
                    },
                    {
                        "title": "Multimodal diffusion geometry by joint diagonalization of Laplacians",
                        "abstract": "We construct an extension of diffusion geometry to multiple modalities through joint approximate diagonalization of Laplacian matrices. This naturally extends classical data analysis tools based on spectral geometry, such as diffusion maps and spectral clustering. We provide several synthetic and real examples of manifold learning, retrieval, and clustering demonstrating that the joint diffusion geometry frequently better captures the inherent structure of multi-modal data. We also show that many previous attempts to construct multimodal spectral clustering can be seen as particular cases of joint approximate diagonalization of the Laplacians."
                    },
                    {
                        "title": "Multimodal similarity-preserving hashing",
                        "abstract": "We introduce an efficient computational framework for hashing data belonging to multiple modalities into a single representation space where they become mutually comparable. The proposed approach is based on a novel coupled siamese neural network architecture and allows unified treatment of intra- and inter-modality similarity learning. Unlike existing cross-modality similarity learning approaches, our hashing functions are not limited to binarized linear projections and can assume arbitrarily complex forms. We show experimentally that our method significantly outperforms state-of-the-art hashing approaches on multimedia retrieval tasks."
                    },
                    {
                        "title": "Almost-commuting matrices are almost jointly diagonalizable",
                        "abstract": "We study the relation between approximate joint diagonalization of self-adjoint matrices and the norm of their commutator, and show that almost commuting self-adjoint matrices are almost jointly diagonalizable by a unitary matrix."
                    },
                    {
                        "title": "Asymptotic metrics on the space of matrices under the commutation relation",
                        "abstract": "We show that the norm of the commutator defines \"almost a metric\" on the quotient space of commuting matrices, in the sense that it is a semi-metric satisfying the triangle inequality asymptotically for large matrices drawn from a \"good\" distribution."
                    },
                    {
                        "title": "MADMM: a generic algorithm for non-smooth optimization on manifolds",
                        "abstract": "Numerous problems in machine learning are formulated as optimization with manifold constraints. In this paper, we propose the Manifold alternating directions method of multipliers (MADMM), an extension of the classical ADMM scheme for manifold-constrained non-smooth optimization problems and show its application to several challenging problems in dimensionality reduction, data analysis, and manifold learning."
                    },
                    {
                        "title": "Subspace Least Squares Multidimensional Scaling",
                        "abstract": "Multidimensional Scaling (MDS) is one of the most popular methods for dimensionality reduction and visualization of high dimensional data. Apart from these tasks, it also found applications in the field of geometry processing for the analysis and reconstruction of non-rigid shapes. In this regard, MDS can be thought of as a \\textit{shape from metric} algorithm, consisting of finding a configuration of points in the Euclidean space that realize, as isometrically as possible, some given distance structure. In the present work we cast the least squares variant of MDS (LS-MDS) in the spectral domain. This uncovers a multiresolution property of distance scaling which speeds up the optimization by a significant amount, while producing comparable, and sometimes even better, embeddings."
                    },
                    {
                        "title": "Generative Convolutional Networks for Latent Fingerprint Reconstruction",
                        "abstract": "Performance of fingerprint recognition depends heavily on the extraction of minutiae points. Enhancement of the fingerprint ridge pattern is thus an essential pre-processing step that noticeably reduces false positive and negative detection rates. A particularly challenging setting is when the fingerprint images are corrupted or partially missing. In this work, we apply generative convolutional networks to denoise visible minutiae and predict the missing parts of the ridge pattern. The proposed enhancement approach is tested as a pre-processing step in combination with several standard feature extraction methods such as MINDTCT, followed by biometric comparison using MCC and BOZORTH3. We evaluate our method on several publicly available latent fingerprint datasets captured using different sensors."
                    },
                    {
                        "title": "MotifNet: a motif-based Graph Convolutional Network for directed graphs",
                        "abstract": "Deep learning on graphs and in particular, graph convolutional neural networks, have recently attracted significant attention in the machine learning community. Many of such techniques explore the analogy between the graph Laplacian eigenvectors and the classical Fourier basis, allowing to formulate the convolution as a multiplication in the spectral domain. One of the key drawback of spectral CNNs is their explicit assumption of an undirected graph, leading to a symmetric Laplacian matrix with orthogonal eigendecomposition. In this work we propose MotifNet, a graph CNN capable of dealing with directed graphs by exploiting local graph motifs. We present experimental evidence showing the advantage of our approach on real data."
                    },
                    {
                        "title": "Affine-invariant geodesic geometry of deformable 3D shapes",
                        "abstract": "Natural objects can be subject to various transformations yet still preserve properties that we refer to as invariants. Here, we use definitions of affine invariant arclength for surfaces in R^3 in order to extend the set of existing non-rigid shape analysis tools. In fact, we show that by re-defining the surface metric as its equi-affine version, the surface with its modified metric tensor can be treated as a canonical Euclidean object on which most classical Euclidean processing and analysis tools can be applied. The new definition of a metric is used to extend the fast marching method technique for computing geodesic distances on surfaces, where now, the distances are defined with respect to an affine invariant arclength. Applications of the proposed framework demonstrate its invariance, efficiency, and accuracy in shape analysis."
                    },
                    {
                        "title": "Deep Functional Maps: Structured Prediction for Dense Shape Correspondence",
                        "abstract": "We introduce a new framework for learning dense correspondence between deformable 3D shapes. Existing learning based approaches model shape correspondence as a labelling problem, where each point of a query shape receives a label identifying a point on some reference domain; the correspondence is then constructed a posteriori by composing the label predictions of two input shapes. We propose a paradigm shift and design a structured prediction model in the space of functional maps, linear operators that provide a compact representation of the correspondence. We model the learning process via a deep residual network which takes dense descriptor fields defined on two shapes as input, and outputs a soft map between the two given objects. The resulting correspondence is shown to be accurate on several challenging benchmarks comprising multiple categories, synthetic models, real scans with acquisition artifacts, topological noise, and partiality."
                    }
                ]
            },
            "985624db-fe9d-44f7-aebd-374c80ea81fb": {
                "pk": "985624db-fe9d-44f7-aebd-374c80ea81fb",
                "name": "Anthea Monod",
                "collaborators": [
                    "Yueqi Cao",
                    "Bo Lin",
                    "Anna Song",
                    "Roan Talbut",
                    "Emil Saucan",
                    "Daniele Tramontano",
                    "Mathias Drton",
                    "Ruriko Yoshida",
                    "In\u00e9s Garc\u00eda-Redondo",
                    "Qiquan Wang"
                ],
                "domain": [
                    "Topological Data Analysis",
                    "Tropical Geometry",
                    "Machine Learning",
                    "Statistical Inference"
                ],
                "publications": [
                    {
                        "title": "An Invitation to Tropical Alexandrov Curvature",
                        "abstract": "We study Alexandrov curvature in the tropical projective torus with respect to the tropical metric, which has been useful in various statistical analyses, particularly in phylogenomics. Alexandrov curvature is a generalization of classical Riemannian sectional curvature to more general metric spaces; it is determined by a comparison of triangles in an arbitrary metric space to corresponding triangles in Euclidean space. In the polyhedral setting of tropical geometry, triangles are a combinatorial object, which adds a combinatorial dimension to our analysis. We study the effect that the triangle types have on curvature, and what can be revealed about these types from the curvature. We find that positive, negative, zero, and undefined Alexandrov curvature can exist concurrently in tropical settings and that there is a tight connection between triangle combinatorial type and curvature. Our results are established both by proof and computational experiments, and shed light on the intricate geometry of the tropical projective torus. In this context, we discuss implications for statistical methodologies which admit inherent geometric interpretations.   This paper is dedicated to Bernd Sturmfels on the occasion of his 60th birthday."
                    },
                    {
                        "title": "Topological Community Detection: A Sheaf-Theoretic Approach",
                        "abstract": "We propose a model for network community detection using topological data analysis, a branch of modern data science that leverages theory from algebraic topology to statistical analysis and machine learning. Specifically, we use cellular sheaves, which relate local to global properties of various algebraic topological constructions, to propose three new algorithms for vertex clustering over networks to detect communities. We apply our algorithms to real social network data in numerical experiments and obtain near optimal results in terms of modularity. Our work is the first implementation of sheaves on real social network data and provides a solid proof-of-concept for future work using sheaves as tools to study complex systems captured by networks and simplicial complexes."
                    },
                    {
                        "title": "Tropical Gradient Descent",
                        "abstract": "We propose a gradient descent method for solving optimisation problems arising in settings of tropical geometry - a variant of algebraic geometry that has become increasingly studied in applications such as computational biology, economics, and computer science. Our approach takes advantage of the polyhedral and combinatorial structures arising in tropical geometry to propose a versatile approach for approximating local minima in tropical statistical optimisation problems - a rapidly growing body of work in recent years. Theoretical results establish global solvability for 1-sample problems and a convergence rate of $O(1/\\sqrt{k})$. Numerical experiments demonstrate the method's superior performance over classical descent for tropical optimisation problems which exhibit tropical convexity but not classical convexity. Notably, tropical descent seamlessly integrates into advanced optimisation methods, such as Adam, offering improved overall performance."
                    },
                    {
                        "title": "Approximating Persistent Homology for Large Datasets",
                        "abstract": "Persistent homology is an important methodology from topological data analysis which adapts theory from algebraic topology to data settings and has been successfully implemented in many applications. It produces a statistical summary in the form of a persistence diagram, which captures the shape and size of the data. Despite its widespread use, persistent homology is simply impossible to implement when a dataset is very large. In this paper we address the problem of finding a representative persistence diagram for prohibitively large datasets. We adapt the classical statistical method of bootstrapping, namely, drawing and studying smaller multiple subsamples from the large dataset. We show that the mean of the persistence diagrams of subsamples -- taken as a mean persistence measure computed from the subsamples -- is a valid approximation of the true persistent homology of the larger dataset. We give the rate of convergence of the mean persistence diagram to the true persistence diagram in terms of the number of subsamples and size of each subsample. Given the complex algebraic and geometric nature of persistent homology, we adapt the convexity and stability properties in the space of persistence diagrams together with random set theory to achieve our theoretical results for the general setting of point cloud data. We demonstrate our approach on simulated and real data, including an application of shape clustering on complex large-scale point cloud data."
                    },
                    {
                        "title": "A Geometric Condition for Uniqueness of Fr\u00e9chet Means of Persistence Diagrams",
                        "abstract": "The Fr\\'echet mean is an important statistical summary and measure of centrality of data; it has been defined and studied for persistent homology captured by persistence diagrams. However, the complicated geometry of the space of persistence diagrams implies that the Fr\\'echet mean for a given set of persistence diagrams is not necessarily unique, which prohibits theoretical guarantees for empirical means with respect to population means. In this paper, we derive a variance expression for a set of persistence diagrams exhibiting a multi-matching between the persistence points known as a grouping. Moreover, we propose a condition for groupings, which we refer to as flatness: sets of persistence diagrams that exhibit flat groupings give rise to unique Fr\\'echet means. We derive a finite sample convergence result for general groupings, which results in convergence for Fr\\'echet means if the groupings are flat. Finally, we interpret flat groupings in a recently-proposed general framework of Fr\\'echet means in Alexandrov geometry. Together with recent results from Alexandrov geometry, this allows for the first derivation of a finite sample convergence rate for sets of persistence diagrams and lays the ground for viability of the Fr\\'echet mean as a practical statistical summary of persistent homology."
                    },
                    {
                        "title": "Curved Markov Chain Monte Carlo for Network Learning",
                        "abstract": "We present a geometrically enhanced Markov chain Monte Carlo sampler for networks based on a discrete curvature measure defined on graphs. Specifically, we incorporate the concept of graph Forman curvature into sampling procedures on both the nodes and edges of a network explicitly, via the transition probability of the Markov chain, as well as implicitly, via the target stationary distribution, which gives a novel, curved Markov chain Monte Carlo approach to learning networks. We show that integrating curvature into the sampler results in faster convergence to a wide range of network statistics demonstrated on deterministic networks drawn from real-world data."
                    },
                    {
                        "title": "Learning Linear Non-Gaussian Polytree Models",
                        "abstract": "In the context of graphical causal discovery, we adapt the versatile framework of linear non-Gaussian acyclic models (LiNGAMs) to propose new algorithms to efficiently learn graphs that are polytrees. Our approach combines the Chow--Liu algorithm, which first learns the undirected tree structure, with novel schemes to orient the edges. The orientation schemes assess algebraic relations among moments of the data-generating distribution and are computationally inexpensive. We establish high-dimensional consistency results for our approach and compare different algorithmic versions in numerical experiments."
                    },
                    {
                        "title": "Tropical Geometric Variation of Phylogenetic Tree Shapes",
                        "abstract": "We study the behavior of phylogenetic tree shapes in the tropical geometric interpretation of tree space. Tree shapes are formally referred to as tree topologies; a tree topology can also be thought of as a tree combinatorial type, which is given by the tree's branching configuration and leaf labeling. We use the tropical line segment as a framework to define notions of variance as well as invariance of tree topologies: we provide a combinatorial search theorem that describes all tree topologies occurring along a tropical line segment, as well as a setting under which tree topologies do not change along a tropical line segment. Our study is motivated by comparison to the moduli space endowed with a geodesic metric proposed by Billera, Holmes, and Vogtmann (referred to as BHV space); we consider the tropical geometric setting as an alternative framework to BHV space for sets of phylogenetic trees. We give an algorithm to compute tropical line segments which is lower in computational complexity than the fastest method currently available for BHV geodesics and show that its trajectory behaves more subtly: while the BHV geodesic traverses the origin for vastly different tree topologies, the tropical line segment bypasses it."
                    },
                    {
                        "title": "Fast Topological Signal Identification and Persistent Cohomological Cycle Matching",
                        "abstract": "Within the context of topological data analysis, the problems of identifying topological significance and matching signals across datasets are important and useful inferential tasks in many applications. The limitation of existing solutions to these problems, however, is computational speed. In this paper, we harness the state-of-the-art for persistent homology computation by studying the problem of determining topological prevalence and cycle matching using a cohomological approach, which increases their feasibility and applicability to a wider variety of applications and contexts. We demonstrate this on a wide range of real-life, large-scale, and complex datasets. We extend existing notions of topological prevalence and cycle matching to include general non-Morse filtrations. This provides the most general and flexible state-of-the-art adaptation of topological signal identification and persistent cycle matching, which performs comparisons of orders of ten for thousands of sampled points in a matter of minutes on standard institutional HPC CPU facilities."
                    },
                    {
                        "title": "$k$-Means Clustering for Persistent Homology",
                        "abstract": "Persistent homology is a methodology central to topological data analysis that extracts and summarizes the topological features within a dataset as a persistence diagram; it has recently gained much popularity from its myriad successful applications to many domains. However, its algebraic construction induces a metric space of persistence diagrams with a highly complex geometry. In this paper, we prove convergence of the $k$-means clustering algorithm on persistence diagram space and establish theoretical properties of the solution to the optimization problem in the Karush--Kuhn--Tucker framework. Additionally, we perform numerical experiments on various representations of persistent homology, including embeddings of persistence diagrams as well as diagrams themselves and their generalizations as persistence measures; we find that $k$-means clustering performance directly on persistence diagrams and measures outperform their vectorized representations."
                    },
                    {
                        "title": "Wavelet-Based Density Estimation for Persistent Homology",
                        "abstract": "Persistent homology is a central methodology in topological data analysis that has been successfully implemented in many fields and is becoming increasingly popular and relevant. The output of persistent homology is a persistence diagram -- a multiset of points supported on the upper half plane -- that is often used as a statistical summary of the topological features of data. In this paper, we study the random nature of persistent homology and estimate the density of expected persistence diagrams from observations using wavelets; we show that our wavelet-based estimator is optimal. Furthermore, we propose an estimator that offers a sparse representation of the expected persistence diagram that achieves near-optimality. We demonstrate the utility of our contributions in a machine learning task in the context of dynamical systems."
                    },
                    {
                        "title": "Tropical Sufficient Statistics for Persistent Homology",
                        "abstract": "We show that an embedding in Euclidean space based on tropical geometry generates stable sufficient statistics for barcodes. In topological data analysis, barcodes are multiscale summaries of algebraic topological characteristics that capture the `shape' of data; however, in practice, they have complex structures that make them difficult to use in statistical settings. The sufficiency result presented in this work allows for classical probability distributions to be assumed on the tropical geometric representation of barcodes. This makes a variety of parametric statistical inference methods amenable to barcodes, all while maintaining their initial interpretations. More specifically, we show that exponential family distributions may be assumed, and that likelihood functions for persistent homology may be constructed. We conceptually demonstrate sufficiency and illustrate its utility in persistent homology dimensions 0 and 1 with concrete parametric applications to human immunodeficiency virus and avian influenza data."
                    },
                    {
                        "title": "Tropical Optimal Transport and Wasserstein Distances",
                        "abstract": "We study the problem of optimal transport in tropical geometry and define the Wasserstein-$p$ distances in the continuous metric measure space setting of the tropical projective torus. We specify the tropical metric -- a combinatorial metric that has been used to study of the tropical geometric space of phylogenetic trees -- as the ground metric and study the cases of $p=1,2$ in detail. The case of $p=1$ gives an efficient computation of the infinitely-many geodesics on the tropical projective torus, while the case of $p=2$ gives a form for Fr\\'{e}chet means and a general inner product structure. Our results also provide theoretical foundations for geometric insight a statistical framework in a tropical geometric setting. We construct explicit algorithms for the computation of the tropical Wasserstein-1 and 2 distances and prove their convergence. Our results provide the first study of the Wasserstein distances and optimal transport in tropical geometry. Several numerical examples are provided."
                    },
                    {
                        "title": "Rewiring Networks for Graph Neural Network Training Using Discrete Geometry",
                        "abstract": "Information over-squashing is a phenomenon of inefficient information propagation between distant nodes on networks. It is an important problem that is known to significantly impact the training of graph neural networks (GNNs), as the receptive field of a node grows exponentially. To mitigate this problem, a preprocessing procedure known as rewiring is often applied to the input network. In this paper, we investigate the use of discrete analogues of classical geometric notions of curvature to model information flow on networks and rewire them. We show that these classical notions achieve state-of-the-art performance in GNN training accuracy on a variety of real-world network datasets. Moreover, compared to the current state-of-the-art, these classical notions exhibit a clear advantage in computational runtime by several orders of magnitude."
                    },
                    {
                        "title": "Generalized Morse Theory of Distance Functions to Surfaces for Persistent Homology",
                        "abstract": "This paper brings together three distinct theories with the goal of quantifying shape textures with complex morphologies. Distance fields are central objects in shape representation, while topological data analysis uses algebraic topology to characterize geometric and topological patterns in shapes. The most well-known and widely applied tool from this approach is persistent homology, which tracks the evolution of topological features in a dynamic manner as a barcode. Morse theory is a framework from differential topology that studies critical points of functions on manifolds; it has been used to characterize the birth and death of persistent homology features. However, a significant limitation to Morse theory is that it cannot be readily applied to distance functions because distance functions lack smoothness, which is required in Morse theory. Our contributions to addressing this issue is two fold. First, we generalize Morse theory to Euclidean distance functions of bounded sets with smooth boundaries. We focus in particular on distance fields for shape representation and we study the persistent homology of shape textures using a sublevel set filtration induced by the signed distance function. We use transversality theory to prove that for generic embeddings of a smooth compact surface in $\\mathbb{R}^3$, signed distance functions admit finitely many non-degenerate critical points. This gives rise to our second contribution, which is that shapes and textures can both now be quantified and rigorously characterized in the language of persistent homology: signed distance persistence modules of generic shapes admit a finite barcode decomposition whose birth and death points can be classified and described geometrically. We use this approach to quantify shape textures on both simulated data and real vascular data from biology."
                    },
                    {
                        "title": "Tropical Geometry of Phylogenetic Tree Space: A Statistical Perspective",
                        "abstract": "Phylogenetic trees are the fundamental mathematical representation of evolutionary processes in biology. They are also objects of interest in pure mathematics, such as algebraic geometry and combinatorics, due to their discrete geometry. Although they are important data structures, they face the significant challenge that sets of trees form a non-Euclidean phylogenetic tree space, which means that standard computational and statistical methods cannot be directly applied. In this work, we explore the statistical feasibility of a pure mathematical representation of the set of all phylogenetic trees based on tropical geometry for both descriptive and inferential statistics, and unsupervised and supervised machine learning. Our exploration is both theoretical and practical. We show that the tropical geometric phylogenetic tree space endowed with a generalized Hilbert projective metric exhibits analytic, geometric, and topological properties that are desirable for theoretical studies in probability and statistics and allow for well-defined questions to be posed. We illustrate the statistical feasibility of the tropical geometric perspective for phylogenetic trees with an example of both a descriptive and inferential statistical task. Moreover, this approach exhibits increased computational efficiency and statistical performance over the current state-of-the-art, which we illustrate with a real data example on seasonal influenza. Our results demonstrate the viability of the tropical geometric setting for parametric statistical and probabilistic studies of sets of phylogenetic trees."
                    },
                    {
                        "title": "A Topological Gaussian Mixture Model for Bone Marrow Morphology in Leukaemia",
                        "abstract": "Acute myeloid leukaemia (AML) is a type of blood and bone marrow cancer characterized by the proliferation of abnormal clonal haematopoietic cells in the bone marrow leading to bone marrow failure. Over the course of the disease, angiogenic factors released by leukaemic cells drastically alter the bone marrow vascular niches resulting in observable structural abnormalities. We use a technique from topological data analysis - persistent homology - to quantify the images and infer on the disease through the imaged morphological features. We find that persistent homology uncovers succinct dissimilarities between the control, early, and late stages of AML development. We then integrate persistent homology into stage-dependent Gaussian mixture models for the first time, proposing a new class of models which are applicable to persistent homology summaries and able to both infer patterns in morphological changes between different stages of progression as well as provide a basis for prediction."
                    },
                    {
                        "title": "Probability Metrics for Tropical Spaces of Different Dimensions",
                        "abstract": "The problem of comparing probability distributions is at the heart of many tasks in statistics and machine learning. Established comparison methods treat the standard setting that the distributions are supported in the same space. Recently, a new geometric solution has been proposed to address the more challenging problem of comparing measures in Euclidean spaces of differing dimensions. Here, we study the same problem of comparing probability distributions of different dimensions in the tropical setting, which is becoming increasingly relevant in applications involving complex data structures such as phylogenetic trees. Specifically, we construct a Wasserstein distance between measures on different tropical projective tori -- the focal metric spaces in both theory and applications of tropical geometry -- via tropical mappings between probability measures. We prove equivalence of the directionality of the maps, whether mapping from a low dimensional space to a high dimensional space or vice versa. As an important practical implication, our work provides a framework for comparing probability distributions on the spaces of phylogenetic trees with different leaf sets. We demonstrate the computational feasibility of our approach using existing optimisation techniques on both simulated and real data."
                    },
                    {
                        "title": "Stability for Inference with Persistent Homology Rank Functions",
                        "abstract": "Persistent homology barcodes and diagrams are a cornerstone of topological data analysis that capture the \"shape\" of a wide range of complex data structures, such as point clouds, networks, and functions. However, their use in statistical settings is challenging due to their complex geometric structure. In this paper, we revisit the persistent homology rank function, which is mathematically equivalent to a barcode and persistence diagram, as a tool for statistics and machine learning. Rank functions, being functions, enable the direct application of the statistical theory of functional data analysis (FDA)-a domain of statistics adapted for data in the form of functions. A key challenge they present over barcodes in practice, however, is their lack of stability-a property that is crucial to validate their use as a faithful representation of the data and therefore a viable summary statistic. In this paper, we fill this gap by deriving two stability results for persistent homology rank functions under a suitable metric for FDA integration. We then study the performance of rank functions in functional inferential statistics and machine learning on real data applications, in both single and multiparameter persistent homology. We find that the use of persistent homology captured by rank functions offers a clear improvement over existing non-persistence-based approaches."
                    },
                    {
                        "title": "Tropical Expressivity of Neural Networks",
                        "abstract": "We propose an algebraic geometric framework to study the expressivity of linear activation neural networks. A particular quantity of neural networks that has been actively studied is the number of linear regions, which gives a quantification of the information capacity of the architecture. To study and evaluate information capacity and expressivity, we work in the setting of tropical geometry - a combinatorial and polyhedral variant of algebraic geometry - where there are known connections between tropical rational maps and feedforward neural networks. Our work builds on and expands this connection to capitalize on the rich theory of tropical geometry to characterize and study various architectural aspects of neural networks. Our contributions are threefold: we provide a novel tropical geometric approach to selecting sampling domains among linear regions; an algebraic result allowing for a guided restriction of the sampling domain for network architectures with symmetries; and a new open source OSCAR library to analyze neural networks symbolically using their tropical representations, where we present a new algorithm that computes the exact number of their linear regions. We provide a comprehensive set of proof-of-concept numerical experiments demonstrating the breadth of neural network architectures to which tropical geometric theory can be applied to reveal insights on expressivity characteristics of a network. Our work provides the foundations for the adaptation of both theory and existing software from computational tropical geometry and symbolic computation to neural networks and deep learning"
                    }
                ]
            }
        }
    },
    "2403.04690": {
        "paper_data": {
            "title": "Faster Neighborhood Attention: Reducing the O(n^2) Cost of Self Attention at the Threadblock Level",
            "url": "http://arxiv.org/abs/2403.04690v2",
            "arxiv_id": "2403.04690",
            "authors": [
                "Ali Hassani",
                "Wen-Mei Hwu",
                "Humphrey Shi"
            ],
            "abstract": "Neighborhood attention reduces the cost of self attention by restricting each token's attention span to its nearest neighbors. This restriction, parameterized by a window size and dilation factor, draws a spectrum of possible attention patterns between linear projection and self attention. Neighborhood attention, and more generally sliding window attention patterns, have long been bounded by infrastructure, particularly in higher-rank spaces (2-D and 3-D), calling for the development of custom kernels, which have been limited in either functionality, or performance, if not both. In this work, we first show that neighborhood attention can be represented as a batched GEMM problem, similar to standard attention, and implement it for 1-D and 2-D neighborhood attention. These kernels on average provide 895% and 272% improvement in full precision latency compared to existing naive kernels for 1-D and 2-D neighborhood attention respectively. We find certain inherent inefficiencies in all unfused neighborhood attention kernels that bound their performance and lower-precision scalability. We also developed fused neighborhood attention; an adaptation of fused dot-product attention kernels that allow fine-grained control over attention across different spatial axes. Known for reducing the quadratic time complexity of self attention to a linear complexity, neighborhood attention can now enjoy a reduced and constant memory footprint, and record-breaking half precision latency. We observe that our fused kernels successfully circumvent some of the unavoidable inefficiencies in unfused implementations. While our unfused GEMM-based kernels only improve half precision performance compared to naive kernels by an average of 496% and 113% in 1-D and 2-D problems respectively, our fused kernels improve naive kernels by an average of 1607% and 581% in 1-D and 2-D problems respectively.",
            "introduction": "   1 Introduction  Inarguably among the most highly utilized and influential primitives in modern deep learning, attention has long been cited for its complexity and memory footprint, especially when the query and context sets are identical (self attention). For years since its adoption in deep learning\u00a0[20], the most common implementation of attention was through two batched GEMM (General Matrix-Matrix Multiplication) operations, sometimes referred to as \u201cBMM-style\u201d attention. This implementation stores attention weights to global memory, which can become a bottleneck in both speed and memory footprint. As the number of possible attention weights grow, the problem gets bounded by global memory bandwidth, as well as the amount of global memory required.   Over the past few years, some works proposed attention implementations in which attention weights are not stored to global memory, but instead kept in on-chip memory (shared memory or register file)\u00a0[16, 5], which are then directly applied to corresponding values, writing the resulting attention outputs to global memory. These implementations, known as fused or memory-efficient attention, reduce the number of global memory accesses in addition to global memory usage. Thanks to the first open-source implementation, Flash Attention\u00a0[5], these fused attention kernels have started replacing the standard BMM-style implementations in many deep learning frameworks and inference engines such as PyTorch\u00a0[14].   Orthogonal to these efforts, many have sought to address the quadratic complexity in self attention, which is the more predominantly used operation in vision models. Neighborhood attention\u00a0[8] is one such method in which each query token is restricted to only interact with its nearest neighboring context tokens. In most cases, this pattern creates a sliding window pattern, which in deep learning can also be seen in the convolution operator. This restriction can similarly be parameterized by a window size and dilation factor, and reduces the quadratic complexity of self attention down to a linear complexity. This reduction is however very difficult to implement at the tensor library or deep learning framework level. Tensor views can represent sliding window attention\u00a0[17], but not the neighborhood attention pattern. In addition, GEMM operators do not typically support taking in the sliding window views in higher-rank spaces (2-D and 3-D) without explicit copying, which in practice undoes the theoretical efficiency gain from the reduced attention complexity. As a result, neighborhood attention was proposed along with an extension carrying naive CUDA kernels\u00a0[8] implementing the two BMM operators. While those kernels can provide competitive FP32 performance, and in some cases even FP16/BF16 performance, they fall short of being generally adoptable in larger scale experiments. In addition, fused attention kernels such as Flash Attention effectively eliminate the \ud835\udcaa\u2062(n2)\ud835\udcaasuperscript\ud835\udc5b2\\mathcal{O}(n^{2})caligraphic_O ( italic_n start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT ) memory footprint in self attention, while also reducing latency significantly\u00a0[5], making subquadratic attention patterns that are only possible to implement in \u201cBMM-style\u201d less feasible in practice.   In this work, we present two new classes of neighborhood attention kernels: GEMM-based BMM-style kernels (GEMM NA), and fused kernels (Fused NA), which are aimed at providing significantly improved infrastructure for neighborhood attention. We first show that neighborhood attention, and by extension sliding window attention, both of which are GEMV (General Matrix-Vector Multiplication) problems, can be expressed as GEMM problems with space-aware tiling and gather/scatter fusion. This would allow implementing such attention patterns with GEMMs, which in turn can benefit greatly from the underlying efficiency of GEMM",
            "references": [
                {
                    "title": "Mistral 7B",
                    "abstract": "We introduce Mistral 7B v0.1, a 7-billion-parameter language model engineered for superior performance and efficiency. Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation. Our model leverages grouped-query attention (GQA) for faster inference, coupled with sliding window attention (SWA) to effectively handle sequences of arbitrary length with a reduced inference cost. We also provide a model fine-tuned to follow instructions, Mistral 7B -- Instruct, that surpasses the Llama 2 13B -- Chat model both on human and automated benchmarks. Our models are released under the Apache 2.0 license."
                },
                {
                    "title": "FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning",
                    "abstract": "Scaling Transformers to longer sequence lengths has been a major problem in the last several years, promising to improve performance in language modeling and high-resolution image understanding, as well as to unlock new applications in code, audio, and video generation. The attention layer is the main bottleneck in scaling to longer sequences, as its runtime and memory increase quadratically in the sequence length. FlashAttention exploits the asymmetric GPU memory hierarchy to bring significant memory saving (linear instead of quadratic) and runtime speedup (2-4$\\times$ compared to optimized baselines), with no approximation. However, FlashAttention is still not nearly as fast as optimized matrix-multiply (GEMM) operations, reaching only 25-40\\% of the theoretical maximum FLOPs/s. We observe that the inefficiency is due to suboptimal work partitioning between different thread blocks and warps on the GPU, causing either low-occupancy or unnecessary shared memory reads/writes. We propose FlashAttention-2, with better work partitioning to address these issues. In particular, we (1) tweak the algorithm to reduce the number of non-matmul FLOPs (2) parallelize the attention computation, even for a single head, across different thread blocks to increase occupancy, and (3) within each thread block, distribute the work between warps to reduce communication through shared memory. These yield around 2$\\times$ speedup compared to FlashAttention, reaching 50-73\\% of the theoretical maximum FLOPs/s on A100 and getting close to the efficiency of GEMM operations. We empirically validate that when used end-to-end to train GPT-style models, FlashAttention-2 reaches training speed of up to 225 TFLOPs/s per A100 GPU (72\\% model FLOPs utilization)."
                },
                {
                    "title": "Scalable Diffusion Models with Transformers",
                    "abstract": "We explore a new class of diffusion models based on the transformer architecture. We train latent diffusion models of images, replacing the commonly-used U-Net backbone with a transformer that operates on latent patches. We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops. We find that DiTs with higher Gflops\u2014through increased transformer depth/width or increased number of input tokens\u2014consistently have lower FID. In addition to possessing good scalability properties, our largest DiT-XL/2 models outperform all prior diffusion models on the class-conditional ImageNet 512\u00d7512 and 256\u00d7256 benchmarks, achieving a state-of-the-art FID of 2.27 on the latter."
                },
                {
                    "title": "StyleNAT: Giving Each Head a New Perspective",
                    "abstract": "Image generation has been a long sought-after but challenging task, and performing the generation task in an efficient manner is similarly difficult. Often researchers attempt to create a\"one size fits all\"generator, where there are few differences in the parameter space for drastically different datasets. Herein, we present a new transformer-based framework, dubbed StyleNAT, targeting high-quality image generation with superior efficiency and flexibility. At the core of our model, is a carefully designed framework that partitions attention heads to capture local and global information, which is achieved through using Neighborhood Attention (NA). With different heads able to pay attention to varying receptive fields, the model is able to better combine this information, and adapt, in a highly flexible manner, to the data at hand. StyleNAT attains a new SOTA FID score on FFHQ-256 with 2.046, beating prior arts with convolutional models such as StyleGAN-XL and transformers such as HIT and StyleSwin, and a new transformer SOTA on FFHQ-1024 with an FID score of 4.174. These results show a 6.4% improvement on FFHQ-256 scores when compared to StyleGAN-XL with a 28% reduction in the number of parameters and 56% improvement in sampling throughput. Code and models will be open-sourced at https://github.com/SHI-Labs/StyleNAT."
                },
                {
                    "title": "Dilated Neighborhood Attention Transformer",
                    "abstract": "Transformers are quickly becoming one of the most heavily applied deep learning architectures across modalities, domains, and tasks. In vision, on top of ongoing efforts into plain transformers, hierarchical transformers have also gained significant attention, thanks to their performance and easy integration into existing frameworks. These models typically employ localized attention mechanisms, such as the sliding-window Neighborhood Attention (NA) or Swin Transformer's Shifted Window Self Attention. While effective at reducing self attention's quadratic complexity, local attention weakens two of the most desirable properties of self attention: long range inter-dependency modeling, and global receptive field. In this paper, we introduce Dilated Neighborhood Attention (DiNA), a natural, flexible and efficient extension to NA that can capture more global context and expand receptive fields exponentially at no additional cost. NA's local attention and DiNA's sparse global attention complement each other, and therefore we introduce Dilated Neighborhood Attention Transformer (DiNAT), a new hierarchical vision transformer built upon both. DiNAT variants enjoy significant improvements over strong baselines such as NAT, Swin, and ConvNeXt. Our large model is faster and ahead of its Swin counterpart by 1.6% box AP in COCO object detection, 1.4% mask AP in COCO instance segmentation, and 1.4% mIoU in ADE20K semantic segmentation. Paired with new frameworks, our large variant is the new state of the art panoptic segmentation model on COCO (58.5 PQ) and ADE20K (49.4 PQ), and instance segmentation model on Cityscapes (45.1 AP) and ADE20K (35.4 AP) (no extra data). It also matches the state of the art specialized semantic segmentation models on ADE20K (58.1 mIoU), and ranks second on Cityscapes (84.5 mIoU) (no extra data)."
                },
                {
                    "title": "FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness",
                    "abstract": "Transformers are slow and memory-hungry on long sequences, since the time and memory complexity of self-attention are quadratic in sequence length. Approximate attention methods have attempted to address this problem by trading off model quality to reduce the compute complexity, but often do not achieve wall-clock speedup. We argue that a missing principle is making attention algorithms IO-aware -- accounting for reads and writes between levels of GPU memory. We propose FlashAttention, an IO-aware exact attention algorithm that uses tiling to reduce the number of memory reads/writes between GPU high bandwidth memory (HBM) and GPU on-chip SRAM. We analyze the IO complexity of FlashAttention, showing that it requires fewer HBM accesses than standard attention, and is optimal for a range of SRAM sizes. We also extend FlashAttention to block-sparse attention, yielding an approximate attention algorithm that is faster than any existing approximate attention method. FlashAttention trains Transformers faster than existing baselines: 15% end-to-end wall-clock speedup on BERT-large (seq. length 512) compared to the MLPerf 1.1 training speed record, 3$\\times$ speedup on GPT-2 (seq. length 1K), and 2.4$\\times$ speedup on long-range arena (seq. length 1K-4K). FlashAttention and block-sparse FlashAttention enable longer context in Transformers, yielding higher quality models (0.7 better perplexity on GPT-2 and 6.4 points of lift on long-document classification) and entirely new capabilities: the first Transformers to achieve better-than-chance performance on the Path-X challenge (seq. length 16K, 61.4% accuracy) and Path-256 (seq. length 64K, 63.1% accuracy)."
                },
                {
                    "title": "Neighborhood Attention Transformer",
                    "abstract": "We present Neighborhood Attention (NA), the first efficient and scalable sliding window attention mechanism for vision. NA is a pixel-wise operation, localizing self attention (SA) to the nearest neighboring pixels, and therefore enjoys a linear time and space complexity compared to the quadratic complexity of SA. The sliding window pattern allows NA's receptive field to grow without needing extra pixel shifts, and preserves translational equivariance, unlike Swin Transformer's Window Self Attention (WSA). We develop NATTEN (Neighborhood Attention Extension), a Python package with efficient C++ and CUDA kernels, which allows NA to run up to 40% faster than Swin's WSA while using up to 25% less memory. We further present Neighborhood Attention Transformer (NAT), a new hierarchical transformer design based on NA that boosts image classification and downstream vision performance. Experimental results on NAT are competitive; NAT-Tiny reaches 83.2% top-1 accuracy on ImageNet, 51.4% mAP on MSCOCO and 48.4% mIoU on ADE20K, which is 1.9% ImageNet accuracy, 1.0% COCO mAP, and 2.6% ADE20K mIoU improvement over a Swin model with similar size. To support more research based on sliding window attention, we open source our project and release our checkpoints."
                },
                {
                    "title": "Self-attention Does Not Need $O(n^2)$ Memory",
                    "abstract": "We present a very simple algorithm for attention that requires $O(1)$ memory with respect to sequence length and an extension to self-attention that requires $O(\\log n)$ memory. This is in contrast with the frequently stated belief that self-attention requires $O(n^2)$ memory. While the time complexity is still $O(n^2)$, device memory rather than compute capability is often the limiting factor on modern accelerators. Thus, reducing the memory requirements of attention allows processing of longer sequences than might otherwise be feasible. We provide a practical implementation for accelerators that requires $O(\\sqrt{n})$ memory, is numerically stable, and is within a few percent of the runtime of the standard implementation of attention. We also demonstrate how to differentiate the function while remaining memory-efficient. For sequence length 16384, the memory overhead of self-attention is reduced by 59X for inference and by 32X for differentiation."
                },
                {
                    "title": "ViViT: A Video Vision Transformer",
                    "abstract": "We present pure-transformer based models for video classification, drawing upon the recent success of such models in image classification. Our model extracts spatiotemporal tokens from the input video, which are then encoded by a series of transformer layers. In order to handle the long sequences of tokens encountered in video, we propose several, efficient variants of our model which factorise the spatial- and temporal-dimensions of the input. Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets. We conduct thorough ablation studies, and achieve state-of-the-art results on multiple video classification benchmarks including Kinetics 400 and 600, Epic Kitchens, Something-Something v2 and Moments in Time, outperforming prior methods based on deep 3D convolutional networks."
                },
                {
                    "title": "Scaling Local Self-Attention for Parameter Efficient Visual Backbones",
                    "abstract": "Self-attention has the promise of improving computer vision systems due to parameter-independent scaling of receptive fields and content-dependent interactions, in contrast to parameter-dependent scaling and content-independent interactions of convolutions. Self-attention models have recently been shown to have encouraging improvements on accuracy-parameter trade-offs compared to baseline convolutional models such as ResNet-50. In this work, we develop self-attention models that can outperform not just the canonical baseline models, but even the high-performing convolutional models. We propose two extensions to self-attention that, in conjunction with a more efficient implementation of self-attention, improve the speed, memory usage, and accuracy of these models. We leverage these improvements to develop a new self-attention model family, HaloNets, which reach state-of-the-art accuracies on the parameter-limited setting of the ImageNet classification benchmark. In preliminary transfer learning experiments, we find that HaloNet models outperform much larger models and have better inference performance. On harder tasks such as object detection and instance segmentation, our simple local self-attention and convolutional hybrids show improvements over very strong baselines. These results mark another step in demonstrating the efficacy of self-attention models on settings traditionally dominated by convolutions. 1"
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Big Bird: Transformers for Longer Sequences",
                    "abstract": "Transformers-based models, such as BERT, have been one of the most successful deep learning models for NLP. Unfortunately, one of their core limitations is the quadratic dependency (mainly in terms of memory) on the sequence length due to their full attention mechanism. To remedy this, we propose, BigBird, a sparse attention mechanism that reduces this quadratic dependency to linear. We show that BigBird is a universal approximator of sequence functions and is Turing complete, thereby preserving these properties of the quadratic, full attention model. Along the way, our theoretical analysis reveals some of the benefits of having $O(1)$ global tokens (such as CLS), that attend to the entire sequence as part of the sparse attention mechanism. The proposed sparse attention can handle sequences of length up to 8x of what was previously possible using similar hardware. As a consequence of the capability to handle longer context, BigBird drastically improves performance on various NLP tasks such as question answering and summarization. We also propose novel applications to genomics data."
                },
                {
                    "title": "Longformer: The Long-Document Transformer",
                    "abstract": "Transformer-based models are unable to process long sequences due to their self-attention operation, which scales quadratically with the sequence length. To address this limitation, we introduce the Longformer with an attention mechanism that scales linearly with sequence length, making it easy to process documents of thousands of tokens or longer. Longformer's attention mechanism is a drop-in replacement for the standard self-attention and combines a local windowed attention with a task motivated global attention. Following prior work on long-sequence transformers, we evaluate Longformer on character-level language modeling and achieve state-of-the-art results on text8 and enwik8. In contrast to most prior work, we also pretrain Longformer and finetune it on a variety of downstream tasks. Our pretrained Longformer consistently outperforms RoBERTa on long document tasks and sets new state-of-the-art results on WikiHop and TriviaQA. We finally introduce the Longformer-Encoder-Decoder (LED), a Longformer variant for supporting long document generative sequence-to-sequence tasks, and demonstrate its effectiveness on the arXiv summarization dataset."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Stand-Alone Self-Attention in Vision Models",
                    "abstract": "Convolutions are a fundamental building block of modern computer vision systems. Recent approaches have argued for going beyond convolutions in order to capture long-range dependencies. These efforts focus on augmenting convolutional models with content-based interactions, such as self-attention and non-local means, to achieve gains on a number of vision tasks. The natural question that arises is whether attention can be a stand-alone primitive for vision models instead of serving as just an augmentation on top of convolutions. In developing and testing a pure self-attention vision model, we verify that self-attention can indeed be an effective stand-alone layer. A simple procedure of replacing all instances of spatial convolutions with a form of self-attention applied to ResNet model produces a fully self-attentional model that outperforms the baseline on ImageNet classification with 12% fewer FLOPS and 29% fewer parameters. On COCO object detection, a pure self-attention model matches the mAP of a baseline RetinaNet while having 39% fewer FLOPS and 34% fewer parameters. Detailed ablation studies demonstrate that self-attention is especially impactful when used in later layers. These results establish that stand-alone self-attention is an important addition to the vision practitioner's toolbox."
                },
                {
                    "title": "Generating Long Sequences with Sparse Transformers",
                    "abstract": "Transformers are powerful sequence models, but require time and memory that grows quadratically with the sequence length. In this paper we introduce sparse factorizations of the attention matrix which reduce this to $O(n \\sqrt{n})$. We also introduce a) a variation on architecture and initialization to train deeper networks, b) the recomputation of attention matrices to save memory, and c) fast attention kernels for training. We call networks with these changes Sparse Transformers, and show they can model sequences tens of thousands of timesteps long using hundreds of layers. We use the same architecture to model images, audio, and text from raw bytes, setting a new state of the art for density modeling of Enwik8, CIFAR-10, and ImageNet-64. We generate unconditional samples that demonstrate global coherence and great diversity, and show it is possible in principle to use self-attention to model sequences of length one million or more."
                },
                {
                    "title": "Online normalizer calculation for softmax",
                    "abstract": "The Softmax function is ubiquitous in machine learning, multiple previous works suggested faster alternatives for it. In this paper we propose a way to compute classical Softmax with fewer memory accesses and hypothesize that this reduction in memory accesses should improve Softmax performance on actual hardware. The benchmarks confirm this hypothesis: Softmax accelerates by up to 1.3x and Softmax+TopK combined and fused by up to 5x."
                },
                {
                    "title": "Image Transformer",
                    "abstract": "Image generation has been successfully cast as an autoregressive sequence generation or transformation problem. Recent work has shown that self-attention is an effective way of modeling textual sequences. In this work, we generalize a recently proposed model architecture based on self-attention, the Transformer, to a sequence modeling formulation of image generation with a tractable likelihood. By restricting the self-attention mechanism to attend to local neighborhoods we significantly increase the size of images the model can process in practice, despite maintaining significantly larger receptive fields per layer than typical convolutional neural networks. We propose another extension of self-attention allowing it to efficiently take advantage of the two-dimensional nature of images. While conceptually simple, our generative models trained on two image data sets are competitive with or significantly outperform the current state of the art in autoregressive image generation on two different data sets, CIFAR-10 and ImageNet. We also present results on image super-resolution with a large magnification ratio, applying an encoder-decoder configuration of our architecture. In a human evaluation study, we show that our super-resolution models improve significantly over previously published autoregressive super-resolution models. Images they generate fool human observers three times more often than the previous state of the art."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively implement neighborhood attention in deep learning frameworks to overcome the limitations of traditional self-attention mechanisms, particularly in terms of memory usage and computational efficiency?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing demand for more efficient attention mechanisms in deep learning, particularly in vision models where self-attention is prevalent. By improving neighborhood attention implementations, we can significantly reduce memory footprint and computational latency, enabling the development of larger and more complex models. This advancement could lead to practical applications in various fields, such as natural language processing, computer vision, and real-time inference systems, ultimately pushing the boundaries of what is achievable with deep learning.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of implementing neighborhood attention within existing tensor libraries and deep learning frameworks. Naive approaches may fail due to the limitations of GEMM operators, which do not support higher-rank sliding window views without explicit copying, negating potential efficiency gains. Additionally, the need for space-aware tiling and gather/scatter fusion adds layers of complexity that require careful optimization to achieve the desired performance improvements.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on standard BMM-style attention implementations, which do not adequately address the unique requirements of neighborhood attention. The lack of support for sliding window patterns in existing tensor libraries has created a barrier to effective implementation. Additionally, earlier attempts at neighborhood attention relied on naive CUDA kernels that, while competitive, were not scalable for larger experiments. Our approach differs by introducing GEMM-based and fused neighborhood attention kernels that leverage the efficiency of GEMM operations, thus providing a more robust solution.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology includes the development of two classes of neighborhood attention kernels: GEMM-based BMM-style kernels (GEMM NA) and fused kernels (Fused NA). We will utilize space-aware tiling and gather/scatter fusion to express neighborhood attention as GEMM problems, allowing for efficient implementation. The dataset will consist of standard benchmarks used in deep learning for vision tasks, and we will evaluate performance using metrics such as memory usage, computational latency, and model accuracy. We expect our approach to yield significant improvements in both efficiency and scalability for neighborhood attention in deep learning frameworks"
            }
        },
        "author_data": {
            "6f7141e7-34a0-4311-8b8a-31bf2a45ed37": {
                "pk": "6f7141e7-34a0-4311-8b8a-31bf2a45ed37",
                "name": "Ali Hassani",
                "collaborators": [
                    "S. Ali Hassani Gangaraj",
                    "George W. Hanson",
                    "Humphrey Shi",
                    "Francesco Monticone",
                    "Andrei Nemilentsau",
                    "Mauro Antezza",
                    "M\u00e1rio G. Silveirinha",
                    "Amir Iranmanesh",
                    "F. Monticone",
                    "Najme Mansouri"
                ],
                "domain": [
                    "Photonic Topology",
                    "Quantum Optics",
                    "Machine Learning",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Momentum-Space Topological Effects of Nonreciprocity",
                        "abstract": "The connection between topology and nonreciprocity in photonic systems is reviewed. Topological properties such as Chern number, and momentum-space properties such as Berry phase and Berry connection, are used to explain back-scattering immune edge states and their topological protection. We consider several examples to illustrate the role of momentum-space topology on wave propagation, and discus recent magnet-less approaches."
                    },
                    {
                        "title": "Dilated Neighborhood Attention Transformer",
                        "abstract": "Transformers are quickly becoming one of the most heavily applied deep learning architectures across modalities, domains, and tasks. In vision, on top of ongoing efforts into plain transformers, hierarchical transformers have also gained significant attention, thanks to their performance and easy integration into existing frameworks. These models typically employ localized attention mechanisms, such as the sliding-window Neighborhood Attention (NA) or Swin Transformer's Shifted Window Self Attention. While effective at reducing self attention's quadratic complexity, local attention weakens two of the most desirable properties of self attention: long range inter-dependency modeling, and global receptive field. In this paper, we introduce Dilated Neighborhood Attention (DiNA), a natural, flexible and efficient extension to NA that can capture more global context and expand receptive fields exponentially at no additional cost. NA's local attention and DiNA's sparse global attention complement each other, and therefore we introduce Dilated Neighborhood Attention Transformer (DiNAT), a new hierarchical vision transformer built upon both. DiNAT variants enjoy significant improvements over strong baselines such as NAT, Swin, and ConvNeXt. Our large model is faster and ahead of its Swin counterpart by 1.6% box AP in COCO object detection, 1.4% mask AP in COCO instance segmentation, and 1.4% mIoU in ADE20K semantic segmentation. Paired with new frameworks, our large variant is the new state of the art panoptic segmentation model on COCO (58.5 PQ) and ADE20K (49.4 PQ), and instance segmentation model on Cityscapes (45.1 AP) and ADE20K (35.4 AP) (no extra data). It also matches the state of the art specialized semantic segmentation models on ADE20K (58.1 mIoU), and ranks second on Cityscapes (84.5 mIoU) (no extra data)."
                    },
                    {
                        "title": "Coupled Topological Surface Modes in Gyrotropic Structures: Green's Function Analysis",
                        "abstract": "At a transition in a wave-guiding structure, part of the incident energy is transmitted and part of the energy is reflected. When the waveguide has non-trivial topological properties, however, the transition may occur with no backscattering, and with unusual modal coupling/transformations. Within this context, we discuss the response of a nonreciprocal topological structure composed of two nearby interfaces between oppositely-biased gyrotropic media and an isotropic medium, which support unidirectional surface modes (topological modes). We provide an exact Green's function analysis of this structure, and we discuss how the topological surface modes are modified when the two interfaces are brought closer and eventually merged. We show that the resulting mode conversion is independent of the transition geometry."
                    },
                    {
                        "title": "Topological Wave-Guiding Near an Exceptional Point: Defect-Immune, Slow-Light, Loss-Immune Propagation",
                        "abstract": "Electromagnetic waves propagating, at finite speeds, in conventional wave-guiding structures are reflected by discontinuities and decay in lossy regions. In this Letter, we drastically modify this typical guided-wave behavior by combining concepts from non-Hermitian physics and topological photonics. To this aim, we theoretically study, for the first time, the possibility of realizing an exceptional point between \\emph{coupled topological modes in a non-Hermitian non-reciprocal waveguide}. Our proposed system is composed of oppositely-biased gyrotropic materials (e.g., biased plasmas or graphene layers) with a balanced loss/gain distribution. To study this complex wave-guiding problem, we put forward an exact analysis based on classical Green's function theory, and we illustrate the behavior of coupled topological modes and the nature of their non-Hermitian degeneracies. We find that, by operating near an exceptional point, we can realize anomalous topological wave propagation with, at the same time, low group-velocity, inherent immunity to back-scattering at discontinuities, and immunity to losses. These theoretical findings may open exciting research directions and stimulate further investigations of non-Hermitian topological waveguides to realize robust wave propagation in practical scenarios."
                    },
                    {
                        "title": "Text Mining using Nonnegative Matrix Factorization and Latent Semantic Analysis",
                        "abstract": "Text clustering is arguably one of the most important topics in modern data mining. Nevertheless, text data require tokenization which usually yields a very large and highly sparse term-document matrix, which is usually difficult to process using conventional machine learning algorithms. Methods such as Latent Semantic Analysis have helped mitigate this issue, but are nevertheless not completely stable in practice. As a result, we propose a new feature agglomeration method based on Nonnegative Matrix Factorization, which is employed to separate the terms into groups, and then each group's term vectors are agglomerated into a new feature vector. Together, these feature vectors create a new feature space much more suitable for clustering. In addition, we propose a new deterministic initialization for spherical K-Means, which proves very useful for this specific type of data. In order to evaluate the proposed method, we compare it to some of the latest research done in this field, as well as some of the most practiced methods. In our experiments, we conclude that the proposed method either significantly improves clustering performance, or maintains the performance of other methods, while improving stability in results."
                    },
                    {
                        "title": "DISCERN: Diversity-based Selection of Centroids for k-Estimation and Rapid Non-stochastic Clustering",
                        "abstract": "One of the applications of center-based clustering algorithms such as K-Means is partitioning data points into K clusters. In some examples, the feature space relates to the underlying problem we are trying to solve, and sometimes we can obtain a suitable feature space. Nevertheless, while K-Means is one of the most efficient offline clustering algorithms, it is not equipped to estimate the number of clusters, which is useful in some practical cases. Other practical methods which do are simply too complex, as they require at least one run of K-Means for each possible K. In order to address this issue, we propose a K-Means initialization similar to K-Means++, which would be able to estimate K based on the feature space while finding suitable initial centroids for K-Means in a deterministic manner. Then we compare the proposed method, DISCERN, with a few of the most practical K estimation methods, while also comparing clustering results of K-Means when initialized randomly, using K-Means++ and using DISCERN. The results show improvement in both the estimation and final clustering performance."
                    },
                    {
                        "title": "Drifting Electrons: Nonreciprocal Plasmonics and Thermal Photonics",
                        "abstract": "Light propagates symmetrically in opposite directions in most materials and structures. This fact -- a consequence of the Lorentz reciprocity principle -- has tremendous implications for science and technology across the electromagnetic spectrum. Here, we investigate an emerging approach to break reciprocity that does not rely on magneto-optical effects or spacetime modulations, but is instead based on biasing a plasmonic material with a direct electric current. Using a 3D Green function formalism and microscopic considerations, we elucidate the propagation properties of surface plasmon-polaritons (SPPs) supported by a generic nonreciprocal platform of this type, revealing some previously overlooked, anomalous, wave-propagation effects. We show that SPPs can propagate in the form of steerable, slow-light, unidirectional beams associated with inflexion points in the modal dispersion. We also clarify the impact of dissipation (due to collisions and Landau damping) on nonreciprocal effects and shed light on the connections between inflexion points, exceptional points at band-edges, and complex modal transitions in leaky-wave structures. We then apply these concepts to the important area of thermal photonics, and provide the first theoretical demonstration of drift-induced nonreciprocal near-field radiative heat transfer between two planar bodies. Our findings may open new opportunities toward the development of nonreciprocal magnet-free devices that combine the benefits of plasmonics and nonreciprocal photonics for wave-guiding and energy applications."
                    },
                    {
                        "title": "DiSparse: Disentangled Sparsification for Multitask Model Compression",
                        "abstract": "Despite the popularity of Model Compression and Multitask Learning, how to effectively compress a multitask model has been less thoroughly analyzed due to the challenging entanglement of tasks in the parameter space. In this paper, we propose DiSparse, a simple, effective, and first-of-its-kind multitask pruning and sparse training scheme. We consider each task independently by disentangling the importance measurement and take the unanimous decisions among all tasks when performing parameter pruning and selection. Our experimental results demonstrate superior performance on various configurations and settings compared to popular sparse training and pruning methods. Besides the effectiveness in compression, DiSparse also provides a powerful tool to the multitask learning community. Surprisingly, we even observed better performance than some dedicated multitask learning methods in several cases despite the high model sparsity enforced by DiSparse. We analyzed the pruning masks generated with DiSparse and observed strikingly similar sparse network architecture identified by each task even before the training starts. We also observe the existence of a \"watershed\" layer where the task relatedness sharply drops, implying no benefits in continued parameters sharing. Our code and models will be available at: https://github.com/SHI-Labs/DiSparse-Multitask-Model-Compression."
                    },
                    {
                        "title": "Dynamical Casimir Effects: The Need for Nonlocality in Time-Varying Dispersive Nanophotonics",
                        "abstract": "Both real and virtual photons can be involved in light-matter interactions. A famous example of the observable implications of virtual photons -- vacuum fluctuations of the quantum electromagnetic field -- is the Casimir effect. Since quantum vacuum effects are weak, various mechanisms have been proposed to enhance and engineer them, ranging from static, e.g., strong optical resonances, to dynamic, e.g., systems with moving boundaries or time-varying optical properties, or a combination of them. In this Letter, we discuss the role of material nonlocality (spatial dispersion) in dynamical Casimir effects in time-varying frequency-dispersive nanophotonic systems. We first show that local models may lead to nonphysical predictions, such as diverging emission rates of entangled polariton pairs. We then theoretically demonstrate that nonlocality regularizes this behavior by correcting the asymptotic response of the system for large wavevectors and reveals physical effects missed by local models, including a significant broadening of the emission rate distribution, which are relevant for future experimental observations. Our work sheds light on the importance of nonlocal effects in this new frontier of nanophotonics."
                    },
                    {
                        "title": "Quantum Photonic Gates with Two-Dimensional Random Walkers",
                        "abstract": "Quantum gates are essential elements for processing quantum information. However, realizing them in a photonic platform is challenging due to the unique way photons propagate and interfere. In this study, we introduce new design of quantum photonic gates that operate based on continuous time two-dimensional random walking photons. These gates can be implemented using the inverse design method, where photons randomly walk in a two-dimensional silicon host medium embedded with silicon dioxide scatterers. We propose a C-NOT gate as a multiqubit gate and an X-gate as a single qubit gate. We will also provide numerical demonstrations of the gate operations using quantum formalism. Additionally, our investigation involves studying the non-trivial spatial correlations of random walking photons by utilizing the quantum correlation function. The results demonstrate high-fidelity probabilistic quantum gates. Further work is required to address error-correction. This work advances the practical implementation of integrated photonics in linear quantum optics."
                    },
                    {
                        "title": "Neighborhood Attention Transformer",
                        "abstract": "We present Neighborhood Attention (NA), the first efficient and scalable sliding-window attention mechanism for vision. NA is a pixel-wise operation, localizing self attention (SA) to the nearest neighboring pixels, and therefore enjoys a linear time and space complexity compared to the quadratic complexity of SA. The sliding-window pattern allows NA's receptive field to grow without needing extra pixel shifts, and preserves translational equivariance, unlike Swin Transformer's Window Self Attention (WSA). We develop NATTEN (Neighborhood Attention Extension), a Python package with efficient C++ and CUDA kernels, which allows NA to run up to 40% faster than Swin's WSA while using up to 25% less memory. We further present Neighborhood Attention Transformer (NAT), a new hierarchical transformer design based on NA that boosts image classification and downstream vision performance. Experimental results on NAT are competitive; NAT-Tiny reaches 83.2% top-1 accuracy on ImageNet, 51.4% mAP on MS-COCO and 48.4% mIoU on ADE20K, which is 1.9% ImageNet accuracy, 1.0% COCO mAP, and 2.6% ADE20K mIoU improvement over a Swin model with similar size. To support more research based on sliding-window attention, we open source our project and release our checkpoints at: https://github.com/SHI-Labs/Neighborhood-Attention-Transformer ."
                    },
                    {
                        "title": "Transient and steady-state entanglement mediated by three-dimensional plasmonic waveguides",
                        "abstract": "Entanglement between two qubits (two level atoms) mediated by surface plasmons in three-dimensional plasmonic waveguides is studied using a quantum master equation formalism. Two types of waveguides, a nanowire and a V-shaped channel cut in a flat metal plane, are considered. The Green functions for the waveguides, which rigorously describes the dissipative qubit environment, are calculated numerically using a direct finite-difference time-domain (FDTD) solution of Maxwell's equations. Finite-length effects are shown to play a crucial role in enhancing entanglement, and resonant-length plasmonic waveguides can provide higher entanglement between qubits than infinite-length waveguides. It is also shown that coupling slots can improve entanglement via stronger qubit-waveguide coupling, for both the infinite- and finite-waveguide cases."
                    },
                    {
                        "title": "Notes on photonic topological insulators and scattering-protected edge states - a brief introduction",
                        "abstract": "The topic of photonic topological insulators and scattering-protected edge states bridges concepts from condensed matter physics and electromagnetics, and necessitates understanding the Berry potential and related concepts. These notes are an attempt at a moderately self-contained introduction to the topic, including two detailed photonic examples drawn from the literature. We made these notes in the process of trying to understand this topic ourselves, and we are posting this material in the spirit of helping other researchers start to understand this material. We claim no novelty in the material or its presentation."
                    },
                    {
                        "title": "The effects of three-dimensional defects on one-way surface plasmon propagation for photonic topological insulators comprised of continuous media",
                        "abstract": "We have investigated the one-way surface plasmon-polariton (SPP) at the interface of a continuous magneto-plasma material and metal. We demonstrated that TM modes inside a continuous material can be assigned non-trivial Chern numbers analogous to those of topological photonic crystals; moreover these Chern numbers can be calculated analytically. This leads to the appearance of topologically protected surface modes propagating at frequencies inside the bandgap of the magneto-plasma. Previous works considered 2D structures; here we consider the effects of 3D defects, and show that, although backward propagation/reflection cannot occur, side scattering does take place and it has significant effect on the propagation of the surface mode. Several different waveguiding geometries are considered for reducing the effects of side-scattering, and we also consider the effects of metal loss."
                    },
                    {
                        "title": "Robust entanglement with 3D nonreciprocal photonic topological insulators",
                        "abstract": "We investigate spontaneous and pumped entanglement of two level systems in the vicinity of a photonic topological insulator interface, which supports a nonreciprocal (unidirectional), scattering-immune and topologically-protected surface plasmon polariton in the bandgap of the bulk material. To this end, we derive a master equation for qubit interactions in a general three-dimensional, nonreciprocal, inhomogeneous and lossy environment. The environment is represented exactly, via the photonic Green function. The resulting entanglement is shown to be extremely robust to defects occurring in the material system, such that strong entanglement is maintained even if the interface exhibits electrically-large and geometrically sharp discontinuities. Alternatively, depending on the initial excitation state, using a non-reciprocal environment allows two qubits to remain unentangled even for very close spacing. The topological nature of the material is manifest in the insensitivity of the entanglement to variations in the material parameters that preserve the gap Chern number. Our formulation and results should be useful for both fundamental investigations of quantum dynamics in nonreciprocal environments, and technological applications related to entanglement in two-level systems."
                    },
                    {
                        "title": "Spontaneous lateral atomic recoil force close to a photonic topological material",
                        "abstract": "We investigate the quantum recoil force acting on an excited atom close to the surface of a nonreciprocal photonic topological insulator (PTI). The main atomic emission channel is the unidirectional surface-plasmon propagating at the PTI-vacuum interface, and we show that it enables a spontaneous lateral recoil force that scales at short distance as $1/d^4$, where $d$ is the atom-PTI separation. Remarkably, the sign of the recoil force is polarization and orientation-independent, and it occurs in a translation-invariant homogeneous system in thermal equilibrium. Surprisingly, the recoil force persists for very small values of the gyration pseudovector, which, for a biased plasma, corresponds to very low cyclotron frequencies. The ultra-strong recoil force is rooted on the quasi-hyperbolic dispersion of the surface-plasmons. We consider both an initially excited atom and a continuous pump scenario, the latter giving rise to a continuous lateral force whose direction can be changed at will by simply varying the orientation of the biasing magnetic field. Our predictions may be tested in experiments with cold Rydberg atoms and superconducting qubits."
                    },
                    {
                        "title": "Fluctuation-induced forces on an atom near a photonic topological material",
                        "abstract": "We theoretically study the Casimir-Polder force on an atom in a arbitrary initial state in a rather general electromagnetic environment wherein the materials may have a nonreciprocal bianisotropic dispersive response. It is shown that under the Markov approximation the force has resonant and nonresonant contributions. We obtain explicit expressions for the optical force both in terms of the system Green function and of the electromagnetic modes. We apply the theory to the particular case wherein a two-level system interacts with a topological gyrotropic material, showing that the nonreciprocity enables exotic light-matter interactions and the opportunity to sculpt and tune the Casimir-Polder forces on the nanoscale. With a quasi-static approximation, we obtain a simple analytical expression for the optical force and unveil the crucial role of surface plasmons in fluctuation induced forces. Finally, we derive the Green function for a gyrotropic material half-space in terms of a Sommerfeld integral."
                    },
                    {
                        "title": "Berry phase, Berry Connection, and Chern Number for a Continuum Bianisotropic Material from a Classical Electromagnetics Perspective",
                        "abstract": "The properties that quantify photonic topological insulators (PTIs), Berry phase, Berry connection, and Chern number, are typically obtained by making analogies between classical Maxwell's equations and the quantum mechanical Schr\\\"{o}dinger equation, writing both in Hamiltonian form. However, the aforementioned quantities are not necessarily quantum in nature, and for photonic systems they can be explained using only classical concepts. Here we provide a derivation and description of PTI quantities using classical Maxwell's equations, we demonstrate how an electromagnetic mode can acquire Berry phase, and we discuss the ramifications of this effect. We consider several examples, including wave propagation in a biased plasma, and radiation by a rotating isotropic emitter. These concepts are discussed without invoking quantum mechanics, and can be easily understood from an engineering electromagnetics perspective."
                    }
                ]
            },
            "7d75e217-a36e-42ce-bfad-aa2b46d39955": {
                "pk": "7d75e217-a36e-42ce-bfad-aa2b46d39955",
                "name": "Wen-Mei Hwu",
                "collaborators": [
                    "Abdul Dakkak",
                    "Cheng Li",
                    "Jinjun Xiong"
                ],
                "domain": [
                    "Machine Learning",
                    "Deep Learning",
                    "Model Evaluation",
                    "Standardization"
                ],
                "publications": [
                    {
                        "title": "DLSpec: A Deep Learning Task Exchange Specification",
                        "abstract": "Deep Learning (DL) innovations are being introduced at a rapid pace. However, the current lack of standard specification of DL tasks makes sharing, running, reproducing, and comparing these innovations difficult. To address this problem, we propose DLSpec, a model-, dataset-, software-, and hardware-agnostic DL specification that captures the different aspects of DL tasks. DLSpec has been tested by specifying and running hundreds of DL tasks."
                    },
                    {
                        "title": "Frustrated with Replicating Claims of a Shared Model? A Solution",
                        "abstract": "Machine Learning (ML) and Deep Learning (DL) innovations are being introduced at such a rapid pace that model owners and evaluators are hard-pressed analyzing and studying them. This is exacerbated by the complicated procedures for evaluation. The lack of standard systems and efficient techniques for specifying and provisioning ML/DL evaluation is the main cause of this \"pain point\". This work discusses common pitfalls for replicating DL model evaluation, and shows that these subtle pitfalls can affect both accuracy and performance. It then proposes a solution to remedy these pitfalls called MLModelScope, a specification for repeatable model evaluation and a runtime to provision and measure experiments. We show that by easing the model specification and evaluation process, MLModelScope facilitates rapid adoption of ML/DL innovations."
                    }
                ]
            },
            "c22b0b81-34a9-4905-aff5-b2f1ffba61ec": {
                "pk": "c22b0b81-34a9-4905-aff5-b2f1ffba61ec",
                "name": "Humphrey Shi",
                "collaborators": [
                    "Ali Hassani",
                    "Zhangyang Wang",
                    "Jiachen Li",
                    "Xingqian Xu",
                    "Jitesh Jain",
                    "Kai Wang",
                    "Steven Walton",
                    "Yuqian Zhou",
                    "Haichao Yu",
                    "Eric Zhang"
                ],
                "domain": [
                    "3D Point Cloud",
                    "Computer Vision",
                    "Deep Learning",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "Deep Learning for 3D Point Cloud Understanding: A Survey",
                        "abstract": "The development of practical applications, such as autonomous driving and robotics, has brought increasing attention to 3D point cloud understanding. While deep learning has achieved remarkable success on image-based tasks, there are many unique challenges faced by deep neural networks in processing massive, unstructured and noisy 3D points. To demonstrate the latest progress of deep learning for 3D point cloud understanding, this paper summarizes recent remarkable research contributions in this area from several different directions (classification, segmentation, detection, tracking, flow estimation, registration, augmentation and completion), together with commonly used datasets, metrics and state-of-the-art performances. More information regarding this survey can be found at: https://github.com/SHI-Labs/3D-Point-Cloud-Learning."
                    },
                    {
                        "title": "Human-Object Interaction Detection:A Quick Survey and Examination of Methods",
                        "abstract": "Human-object interaction detection is a relatively new task in the world of computer vision and visual semantic information extraction. With the goal of machines identifying interactions that humans perform on objects, there are many real-world use cases for the research in this field. To our knowledge, this is the first general survey of the state-of-the-art and milestone works in this field. We provide a basic survey of the developments in the field of human-object interaction detection. Many works in this field use multi-stream convolutional neural network architectures, which combine features from multiple sources in the input image. Most commonly these are the humans and objects in question, as well as the spatial quality of the two. As far as we are aware, there have not been in-depth studies performed that look into the performance of each component individually. In order to provide insight to future researchers, we perform an individualized study that examines the performance of each component of a multi-stream convolutional neural network architecture for human-object interaction detection. Specifically, we examine the HORCNN architecture as it is a foundational work in the field. In addition, we provide an in-depth look at the HICO-DET dataset, a popular benchmark in the field of human-object interaction detection. Code and papers can be found at https://github.com/SHI-Labs/Human-Object-Interaction-Detection."
                    },
                    {
                        "title": "Dilated Neighborhood Attention Transformer",
                        "abstract": "Transformers are quickly becoming one of the most heavily applied deep learning architectures across modalities, domains, and tasks. In vision, on top of ongoing efforts into plain transformers, hierarchical transformers have also gained significant attention, thanks to their performance and easy integration into existing frameworks. These models typically employ localized attention mechanisms, such as the sliding-window Neighborhood Attention (NA) or Swin Transformer's Shifted Window Self Attention. While effective at reducing self attention's quadratic complexity, local attention weakens two of the most desirable properties of self attention: long range inter-dependency modeling, and global receptive field. In this paper, we introduce Dilated Neighborhood Attention (DiNA), a natural, flexible and efficient extension to NA that can capture more global context and expand receptive fields exponentially at no additional cost. NA's local attention and DiNA's sparse global attention complement each other, and therefore we introduce Dilated Neighborhood Attention Transformer (DiNAT), a new hierarchical vision transformer built upon both. DiNAT variants enjoy significant improvements over strong baselines such as NAT, Swin, and ConvNeXt. Our large model is faster and ahead of its Swin counterpart by 1.6% box AP in COCO object detection, 1.4% mask AP in COCO instance segmentation, and 1.4% mIoU in ADE20K semantic segmentation. Paired with new frameworks, our large variant is the new state of the art panoptic segmentation model on COCO (58.5 PQ) and ADE20K (49.4 PQ), and instance segmentation model on Cityscapes (45.1 AP) and ADE20K (35.4 AP) (no extra data). It also matches the state of the art specialized semantic segmentation models on ADE20K (58.1 mIoU), and ranks second on Cityscapes (84.5 mIoU) (no extra data)."
                    },
                    {
                        "title": "Matting Anything",
                        "abstract": "In this paper, we propose the Matting Anything Model (MAM), an efficient and versatile framework for estimating the alpha matte of any instance in an image with flexible and interactive visual or linguistic user prompt guidance. MAM offers several significant advantages over previous specialized image matting networks: (i) MAM is capable of dealing with various types of image matting, including semantic, instance, and referring image matting with only a single model; (ii) MAM leverages the feature maps from the Segment Anything Model (SAM) and adopts a lightweight Mask-to-Matte (M2M) module to predict the alpha matte through iterative refinement, which has only 2.7 million trainable parameters. (iii) By incorporating SAM, MAM simplifies the user intervention required for the interactive use of image matting from the trimap to the box, point, or text prompt. We evaluate the performance of MAM on various image matting benchmarks, and the experimental results demonstrate that MAM achieves comparable performance to the state-of-the-art specialized image matting models under different metrics on each benchmark. Overall, MAM shows superior generalization ability and can effectively handle various image matting tasks with fewer parameters, making it a practical solution for unified image matting. Our code and models are open-sourced at https://github.com/SHI-Labs/Matting-Anything."
                    },
                    {
                        "title": "VCoder: Versatile Vision Encoders for Multimodal Large Language Models",
                        "abstract": "Humans possess the remarkable skill of Visual Perception, the ability to see and understand the seen, helping them make sense of the visual world and, in turn, reason. Multimodal Large Language Models (MLLM) have recently achieved impressive performance on vision-language tasks ranging from visual question-answering and image captioning to visual reasoning and image generation. However, when prompted to identify or count (perceive) the entities in a given image, existing MLLM systems fail. Working towards developing an accurate MLLM system for perception and reasoning, we propose using Versatile vision enCoders (VCoder) as perception eyes for Multimodal LLMs. We feed the VCoder with perception modalities such as segmentation or depth maps, improving the MLLM's perception abilities. Secondly, we leverage the images from COCO and outputs from off-the-shelf vision perception models to create our COCO Segmentation Text (COST) dataset for training and evaluating MLLMs on the object perception task. Thirdly, we introduce metrics to assess the object perception abilities in MLLMs on our COST dataset. Lastly, we provide extensive experimental evidence proving the VCoder's improved object-level perception skills over existing Multimodal LLMs, including GPT-4V. We open-source our dataset, code, and models to promote research. We open-source our code at https://github.com/SHI-Labs/VCoder"
                    },
                    {
                        "title": "Feudal Reinforcement Learning by Reading Manuals",
                        "abstract": "Reading to act is a prevalent but challenging task which requires the ability to reason from a concise instruction. However, previous works face the semantic mismatch between the low-level actions and the high-level language descriptions and require the human-designed curriculum to work properly. In this paper, we present a Feudal Reinforcement Learning (FRL) model consisting of a manager agent and a worker agent. The manager agent is a multi-hop plan generator dealing with high-level abstract information and generating a series of sub-goals in a backward manner. The worker agent deals with the low-level perceptions and actions to achieve the sub-goals one by one. In comparison, our FRL model effectively alleviate the mismatching between text-level inference and low-level perceptions and actions; and is general to various forms of environments, instructions and manuals; and our multi-hop plan generator can significantly boost for challenging tasks where multi-step reasoning form the texts is critical to resolve the instructed goals. We showcase our approach achieves competitive performance on two challenging tasks, Read to Fight Monsters (RTFM) and Messenger, without human-designed curriculum learning."
                    },
                    {
                        "title": "Study Group Learning: Improving Retinal Vessel Segmentation Trained with Noisy Labels",
                        "abstract": "Retinal vessel segmentation from retinal images is an essential task for developing the computer-aided diagnosis system for retinal diseases. Efforts have been made on high-performance deep learning-based approaches to segment the retinal images in an end-to-end manner. However, the acquisition of retinal vessel images and segmentation labels requires onerous work from professional clinicians, which results in smaller training dataset with incomplete labels. As known, data-driven methods suffer from data insufficiency, and the models will easily over-fit the small-scale training data. Such a situation becomes more severe when the training vessel labels are incomplete or incorrect. In this paper, we propose a Study Group Learning (SGL) scheme to improve the robustness of the model trained on noisy labels. Besides, a learned enhancement map provides better visualization than conventional methods as an auxiliary tool for clinicians. Experiments demonstrate that the proposed method further improves the vessel segmentation performance in DRIVE and CHASE$\\_$DB1 datasets, especially when the training labels are noisy."
                    },
                    {
                        "title": "UltraSR: Spatial Encoding is a Missing Key for Implicit Image Function-based Arbitrary-Scale Super-Resolution",
                        "abstract": "The recent success of NeRF and other related implicit neural representation methods has opened a new path for continuous image representation, where pixel values no longer need to be looked up from stored discrete 2D arrays but can be inferred from neural network models on a continuous spatial domain. Although the recent work LIIF has demonstrated that such novel approaches can achieve good performance on the arbitrary-scale super-resolution task, their upscaled images frequently show structural distortion due to the inaccurate prediction of high-frequency textures. In this work, we propose UltraSR, a simple yet effective new network design based on implicit image functions in which we deeply integrated spatial coordinates and periodic encoding with the implicit neural representation. Through extensive experiments and ablation studies, we show that spatial encoding is a missing key toward the next-stage high-performing implicit image function. Our UltraSR sets new state-of-the-art performance on the DIV2K benchmark under all super-resolution scales compared to previous state-of-the-art methods. UltraSR also achieves superior performance on other standard benchmark datasets in which it outperforms prior works in almost all experiments."
                    },
                    {
                        "title": "Is In-Domain Data Really Needed? A Pilot Study on Cross-Domain Calibration for Network Quantization",
                        "abstract": "Post-training quantization methods use a set of calibration data to compute quantization ranges for network parameters and activations. The calibration data usually comes from the training dataset which could be inaccessible due to sensitivity of the data. In this work, we want to study such a problem: can we use out-of-domain data to calibrate the trained networks without knowledge of the original dataset? Specifically, we go beyond the domain of natural images to include drastically different domains such as X-ray images, satellite images and ultrasound images. We find cross-domain calibration leads to surprisingly stable performance of quantized models on 10 tasks in different image domains with 13 different calibration datasets. We also find that the performance of quantized models is correlated with the similarity of the Gram matrices between the source and calibration domains, which can be used as a criterion to choose calibration set for better performance. We believe our research opens the door to borrow cross-domain knowledge for network quantization and compression."
                    },
                    {
                        "title": "Keys to Better Image Inpainting: Structure and Texture Go Hand in Hand",
                        "abstract": "Deep image inpainting has made impressive progress with recent advances in image generation and processing algorithms. We claim that the performance of inpainting algorithms can be better judged by the generated structures and textures. Structures refer to the generated object boundary or novel geometric structures within the hole, while texture refers to high-frequency details, especially man-made repeating patterns filled inside the structural regions. We believe that better structures are usually obtained from a coarse-to-fine GAN-based generator network while repeating patterns nowadays can be better modeled using state-of-the-art high-frequency fast fourier convolutional layers. In this paper, we propose a novel inpainting network combining the advantages of the two designs. Therefore, our model achieves a remarkable visual quality to match state-of-the-art performance in both structure generation and repeating texture synthesis using a single network. Extensive experiments demonstrate the effectiveness of the method, and our conclusions further highlight the two critical factors of image inpainting quality, structures, and textures, as the future design directions of inpainting networks."
                    },
                    {
                        "title": "Combating Label Noise With A General Surrogate Model For Sample Selection",
                        "abstract": "Modern deep learning systems are data-hungry. Learning with web data is one of the feasible solutions, but will introduce label noise inevitably, which can hinder the performance of deep neural networks. Sample selection is an effective way to deal with label noise. The key is to separate clean samples based on some criterion. Previous methods pay more attention to the small loss criterion where small-loss samples are regarded as clean ones. Nevertheless, such a strategy relies on the learning dynamics of each data instance. Some noisy samples are still memorized due to frequently occurring corrupted learning patterns. To tackle this problem, a training-free surrogate model is preferred, freeing from the effect of memorization. In this work, we propose to leverage the vision-language surrogate model CLIP to filter noisy samples automatically. CLIP brings external knowledge to facilitate the selection of clean samples with its ability of text-image alignment. Furthermore, a margin adaptive loss is designed to regularize the selection bias introduced by CLIP, providing robustness to label noise. We validate the effectiveness of our proposed method on both real-world and synthetic noisy datasets. Our method achieves significant improvement without CLIP involved during the inference stage."
                    },
                    {
                        "title": "Neighborhood Attention Transformer",
                        "abstract": "We present Neighborhood Attention (NA), the first efficient and scalable sliding-window attention mechanism for vision. NA is a pixel-wise operation, localizing self attention (SA) to the nearest neighboring pixels, and therefore enjoys a linear time and space complexity compared to the quadratic complexity of SA. The sliding-window pattern allows NA's receptive field to grow without needing extra pixel shifts, and preserves translational equivariance, unlike Swin Transformer's Window Self Attention (WSA). We develop NATTEN (Neighborhood Attention Extension), a Python package with efficient C++ and CUDA kernels, which allows NA to run up to 40% faster than Swin's WSA while using up to 25% less memory. We further present Neighborhood Attention Transformer (NAT), a new hierarchical transformer design based on NA that boosts image classification and downstream vision performance. Experimental results on NAT are competitive; NAT-Tiny reaches 83.2% top-1 accuracy on ImageNet, 51.4% mAP on MS-COCO and 48.4% mIoU on ADE20K, which is 1.9% ImageNet accuracy, 1.0% COCO mAP, and 2.6% ADE20K mIoU improvement over a Swin model with similar size. To support more research based on sliding-window attention, we open source our project and release our checkpoints at: https://github.com/SHI-Labs/Neighborhood-Attention-Transformer ."
                    },
                    {
                        "title": "StyleNAT: Giving Each Head a New Perspective",
                        "abstract": "Image generation has been a long sought-after but challenging task, and performing the generation task in an efficient manner is similarly difficult. Often researchers attempt to create a \"one size fits all\" generator, where there are few differences in the parameter space for drastically different datasets. Herein, we present a new transformer-based framework, dubbed StyleNAT, targeting high-quality image generation with superior efficiency and flexibility. At the core of our model, is a carefully designed framework that partitions attention heads to capture local and global information, which is achieved through using Neighborhood Attention (NA). With different heads able to pay attention to varying receptive fields, the model is able to better combine this information, and adapt, in a highly flexible manner, to the data at hand. StyleNAT attains a new SOTA FID score on FFHQ-256 with 2.046, beating prior arts with convolutional models such as StyleGAN-XL and transformers such as HIT and StyleSwin, and a new transformer SOTA on FFHQ-1024 with an FID score of 4.174. These results show a 6.4% improvement on FFHQ-256 scores when compared to StyleGAN-XL with a 28% reduction in the number of parameters and 56% improvement in sampling throughput. Code and models will be open-sourced at https://github.com/SHI-Labs/StyleNAT."
                    },
                    {
                        "title": "Versatile Diffusion: Text, Images and Variations All in One Diffusion Model",
                        "abstract": "Recent advances in diffusion models have set an impressive milestone in many generation tasks, and trending works such as DALL-E2, Imagen, and Stable Diffusion have attracted great interest. Despite the rapid landscape changes, recent new approaches focus on extensions and performance rather than capacity, thus requiring separate models for separate tasks. In this work, we expand the existing single-flow diffusion pipeline into a multi-task multimodal network, dubbed Versatile Diffusion (VD), that handles multiple flows of text-to-image, image-to-text, and variations in one unified model. The pipeline design of VD instantiates a unified multi-flow diffusion framework, consisting of sharable and swappable layer modules that enable the crossmodal generality beyond images and text. Through extensive experiments, we demonstrate that VD successfully achieves the following: a) VD outperforms the baseline approaches and handles all its base tasks with competitive quality; b) VD enables novel extensions such as disentanglement of style and semantics, dual- and multi-context blending, etc.; c) The success of our multi-flow multimodal framework over images and text may inspire further diffusion-based universal AI research. Our code and models are open-sourced at https://github.com/SHI-Labs/Versatile-Diffusion."
                    },
                    {
                        "title": "Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Models",
                        "abstract": "The unlearning problem of deep learning models, once primarily an academic concern, has become a prevalent issue in the industry. The significant advances in text-to-image generation techniques have prompted global discussions on privacy, copyright, and safety, as numerous unauthorized personal IDs, content, artistic creations, and potentially harmful materials have been learned by these models and later utilized to generate and distribute uncontrolled content. To address this challenge, we propose \\textbf{Forget-Me-Not}, an efficient and low-cost solution designed to safely remove specified IDs, objects, or styles from a well-configured text-to-image model in as little as 30 seconds, without impairing its ability to generate other content. Alongside our method, we introduce the \\textbf{Memorization Score (M-Score)} and \\textbf{ConceptBench} to measure the models' capacity to generate general concepts, grouped into three primary categories: ID, object, and style. Using M-Score and ConceptBench, we demonstrate that Forget-Me-Not can effectively eliminate targeted concepts while maintaining the model's performance on other concepts. Furthermore, Forget-Me-Not offers two practical extensions: a) removal of potentially harmful or NSFW content, and b) enhancement of model accuracy, inclusion and diversity through \\textbf{concept correction and disentanglement}. It can also be adapted as a lightweight model patch for Stable Diffusion, allowing for concept manipulation and convenient distribution. To encourage future research in this critical area and promote the development of safe and inclusive generative models, we will open-source our code and ConceptBench at \\href{https://github.com/SHI-Labs/Forget-Me-Not}{https://github.com/SHI-Labs/Forget-Me-Not}."
                    },
                    {
                        "title": "MSN: Efficient Online Mask Selection Network for Video Instance Segmentation",
                        "abstract": "In this work we present a novel solution for Video Instance Segmentation(VIS), that is automatically generating instance level segmentation masks along with object class and tracking them in a video. Our method improves the masks from segmentation and propagation branches in an online manner using the Mask Selection Network (MSN) hence limiting the noise accumulation during mask tracking. We propose an effective design of MSN by using patch-based convolutional neural network. The network is able to distinguish between very subtle differences between the masks and choose the better masks out of the associated masks accurately. Further, we make use of temporal consistency and process the video sequences in both forward and reverse manner as a post processing step to recover lost objects. The proposed method can be used to adapt any video object segmentation method for the task of VIS. Our method achieves a score of 49.1 mAP on 2021 YouTube-VIS Challenge and was ranked third place among more than 30 global teams. Our code will be available at https://github.com/SHI-Labs/Mask-Selection-Networks."
                    },
                    {
                        "title": "ConvMLP: Hierarchical Convolutional MLPs for Vision",
                        "abstract": "MLP-based architectures, which consist of a sequence of consecutive multi-layer perceptron blocks, have recently been found to reach comparable results to convolutional and transformer-based methods. However, most adopt spatial MLPs which take fixed dimension inputs, therefore making it difficult to apply them to downstream tasks, such as object detection and semantic segmentation. Moreover, single-stage designs further limit performance in other computer vision tasks and fully connected layers bear heavy computation. To tackle these problems, we propose ConvMLP: a hierarchical Convolutional MLP for visual recognition, which is a light-weight, stage-wise, co-design of convolution layers, and MLPs. In particular, ConvMLP-S achieves 76.8% top-1 accuracy on ImageNet-1k with 9M parameters and 2.4G MACs (15% and 19% of MLP-Mixer-B/16, respectively). Experiments on object detection and semantic segmentation further show that visual representation learned by ConvMLP can be seamlessly transferred and achieve competitive results with fewer parameters. Our code and pre-trained models are publicly available at https://github.com/SHI-Labs/Convolutional-MLPs."
                    },
                    {
                        "title": "DiSparse: Disentangled Sparsification for Multitask Model Compression",
                        "abstract": "Despite the popularity of Model Compression and Multitask Learning, how to effectively compress a multitask model has been less thoroughly analyzed due to the challenging entanglement of tasks in the parameter space. In this paper, we propose DiSparse, a simple, effective, and first-of-its-kind multitask pruning and sparse training scheme. We consider each task independently by disentangling the importance measurement and take the unanimous decisions among all tasks when performing parameter pruning and selection. Our experimental results demonstrate superior performance on various configurations and settings compared to popular sparse training and pruning methods. Besides the effectiveness in compression, DiSparse also provides a powerful tool to the multitask learning community. Surprisingly, we even observed better performance than some dedicated multitask learning methods in several cases despite the high model sparsity enforced by DiSparse. We analyzed the pruning masks generated with DiSparse and observed strikingly similar sparse network architecture identified by each task even before the training starts. We also observe the existence of a \"watershed\" layer where the task relatedness sharply drops, implying no benefits in continued parameters sharing. Our code and models will be available at: https://github.com/SHI-Labs/DiSparse-Multitask-Model-Compression."
                    },
                    {
                        "title": "Boosted Dynamic Neural Networks",
                        "abstract": "Early-exiting dynamic neural networks (EDNN), as one type of dynamic neural networks, has been widely studied recently. A typical EDNN has multiple prediction heads at different layers of the network backbone. During inference, the model will exit at either the last prediction head or an intermediate prediction head where the prediction confidence is higher than a predefined threshold. To optimize the model, these prediction heads together with the network backbone are trained on every batch of training data. This brings a train-test mismatch problem that all the prediction heads are optimized on all types of data in training phase while the deeper heads will only see difficult inputs in testing phase. Treating training and testing inputs differently at the two phases will cause the mismatch between training and testing data distributions. To mitigate this problem, we formulate an EDNN as an additive model inspired by gradient boosting, and propose multiple training techniques to optimize the model effectively. We name our method BoostNet. Our experiments show it achieves the state-of-the-art performance on CIFAR100 and ImageNet datasets in both anytime and budgeted-batch prediction modes. Our code is released at https://github.com/SHI-Labs/Boosted-Dynamic-Networks."
                    }
                ]
            }
        }
    },
    "2405.16718": {
        "paper_data": {
            "title": "Amortized Active Causal Induction with Deep Reinforcement Learning",
            "url": "http://arxiv.org/abs/2405.16718v1",
            "arxiv_id": "2405.16718",
            "authors": [
                "Yashas Annadani",
                "Panagiotis Tigas",
                "Stefan Bauer",
                "Adam Foster"
            ],
            "abstract": "We present Causal Amortized Active Structure Learning (CAASL), an active intervention design policy that can select interventions that are adaptive, real-time and that does not require access to the likelihood. This policy, an amortized network based on the transformer, is trained with reinforcement learning on a simulator of the design environment, and a reward function that measures how close the true causal graph is to a causal graph posterior inferred from the gathered data. On synthetic data and a single-cell gene expression simulator, we demonstrate empirically that the data acquired through our policy results in a better estimate of the underlying causal graph than alternative strategies. Our design policy successfully achieves amortized intervention design on the distribution of the training environment while also generalizing well to distribution shifts in test-time design environments. Further, our policy also demonstrates excellent zero-shot generalization to design environments with dimensionality higher than that during training, and to intervention types that it has not been trained on.",
            "introduction": "   1 Introduction  Figure 1: Causal Amortized Structure Learning (CAASL) is an active intervention design method that directly proposes the next intervention to perform by just a forward-pass of the transformer based policy.   Infer, design and experiment is a three step loop in the empirical scientific discovery paradigm. Causal induction (a.k.a.\u00a0causal structure learning), the problem of finding causal relationships present in data, also falls under this paradigm when experiments in the form of interventions are permissible\u00a0[49, 26]. Causal structure learning has gained increasing importance in empirical sciences, for example in single-cell biology, where perturbation experiments like gene knockouts can be carried out with high-precision\u00a0[51]. Such interventions are not only more informative to infer the underlying causal graph than just observational data, but in certain cases essential to go beyond the Markov equivalence class\u00a0[41], making the problem of design of interventions both relevant and important. For the problem of structure learning with interventions, however, inference and design both involve significant challenges. For instance, inference of the causal graph from data usually involves search over the space of graphs with a likelihood (usually weighted by a prior) or score function\u00a0[3, 8, 25], which is slow and not robust to violations of data generation assumptions\u00a0[39]. The design of informative interventions, on the other hand, utilizes the inferred causal graph from existing data to select promising designs and rank them according to a scoring criterion. This scoring criterion is usually based on an approximation of mutual information between the unknown causal graph and the interventional data\u00a0[53, 52], which also involves the (interventional) data likelihood. In problems related to empirical sciences where causal discovery is essential, like inferring a gene regulatory network with gene knockouts or knockdowns, the likelihood of the data is typically intractable. While progress has been made in terms of likelihood-free inference of causal graphs\u00a0[37, 31], existing intervention design algorithms have been largely restricted to likelihood-based strategies.   With a focus on addressing practical intervention design challenges that arise in empirical sciences like inferring the gene regulatory network, in this work, we propose an intervention design method called CAASL that significantly differs from existing approaches. Instead of following the infer, design and experiment loop, we amortize the intervention design procedure by training a single design network policy, based on the transformer\u00a0[54], which encodes key design space symmetries. During test-time, our trained policy directly predicts the next intervention to perform by just a forward-pass of the data collected so far, without the need to undergo slow and expensive inference of the causal graph corresponding to that data. We train the transformer policy with Soft Actor-Critic (SAC)\u00a0[24] to maximize cumulative rewards over a fixed number of design iterations (budget), thereby making the policy adaptive. The choice of a good reward function is essential for informative designs. We discuss various reward function choices, primarily based on an estimate of the true causal graph obtained from a likelihood-free amortized causal discovery approach. Both our policy and the reward function only require access to a simulator of the design environment. Further, we present connections of our approach to amortized sequential Bayesian experimental design\u00a0[20]. We demonstrate",
            "references": [
                {
                    "title": "The Causal Chambers: Real Physical Systems as a Testbed for AI Methodology",
                    "abstract": "In some fields of AI, machine learning and statistics, the validation of new methods and algorithms is often hindered by the scarcity of suitable real-world datasets. Researchers must often turn to simulated data, which yields limited information about the applicability of the proposed methods to real problems. As a step forward, we have constructed two devices that allow us to quickly and inexpensively produce large datasets from non-trivial but well-understood physical systems. The devices, which we call causal chambers, are computer-controlled laboratories that allow us to manipulate and measure an array of variables from these physical systems, providing a rich testbed for algorithms from a variety of fields. We illustrate potential applications through a series of case studies in fields such as causal discovery, out-of-distribution generalization, change point detection, independent component analysis, and symbolic regression. For applications to causal inference, the chambers allow us to carefully perform interventions. We also provide and empirically validate a causal model of each chamber, which can be used as ground truth for different tasks. All hardware and software is made open source, and the datasets are publicly available at causalchamber.org or through the Python package causalchamber."
                },
                {
                    "title": "Gene Regulatory Network Inference in the Presence of Dropouts: a Causal View",
                    "abstract": "Gene regulatory network inference (GRNI) is a challenging problem, particularly owing to the presence of zeros in single-cell RNA sequencing data: some are biological zeros representing no gene expression, while some others are technical zeros arising from the sequencing procedure (aka dropouts), which may bias GRNI by distorting the joint distribution of the measured gene expressions. Existing approaches typically handle dropout error via imputation, which may introduce spurious relations as the true joint distribution is generally unidentifiable. To tackle this issue, we introduce a causal graphical model to characterize the dropout mechanism, namely, Causal Dropout Model. We provide a simple yet effective theoretical result: interestingly, the conditional independence (CI) relations in the data with dropouts, after deleting the samples with zero values (regardless if technical or not) for the conditioned variables, are asymptotically identical to the CI relations in the original data without dropouts. This particular test-wise deletion procedure, in which we perform CI tests on the samples without zeros for the conditioned variables, can be seamlessly integrated with existing structure learning approaches including constraint-based and greedy score-based methods, thus giving rise to a principled framework for GRNI in the presence of dropouts. We further show that the causal dropout model can be validated from data, and many existing statistical models to handle dropouts fit into our model as specific parametric instances. Empirical evaluation on synthetic, curated, and real-world experimental transcriptomic data comprehensively demonstrate the efficacy of our method."
                },
                {
                    "title": "Causal machine learning for single-cell genomics",
                    "abstract": "Advances in single-cell omics allow for unprecedented insights into the transcription profiles of individual cells. When combined with large-scale perturbation screens, through which specific biological mechanisms can be targeted, these technologies allow for measuring the effect of targeted perturbations on the whole transcriptome. These advances provide an opportunity to better understand the causative role of genes in complex biological processes such as gene regulation, disease progression or cellular development. However, the high-dimensional nature of the data, coupled with the intricate complexity of biological systems renders this task nontrivial. Within the machine learning community, there has been a recent increase of interest in causality, with a focus on adapting established causal techniques and algorithms to handle high-dimensional data. In this perspective, we delineate the application of these methodologies within the realm of single-cell genomics and their challenges. We first present the model that underlies most of current causal approaches to single-cell biology and discuss and challenge the assumptions it entails from the biological point of view. We then identify open problems in the application of causal approaches to single-cell data: generalising to unseen environments, learning interpretable models, and learning causal models of dynamics. For each problem, we discuss how various research directions - including the development of computational approaches and the adaptation of experimental protocols - may offer ways forward, or on the contrary pose some difficulties. With the advent of single cell atlases and increasing perturbation data, we expect causal models to become a crucial tool for informed experimental design."
                },
                {
                    "title": "Assumption violations in causal discovery and the robustness of score matching",
                    "abstract": "When domain knowledge is limited and experimentation is restricted by ethical, financial, or time constraints, practitioners turn to observational causal discovery methods to recover the causal structure, exploiting the statistical properties of their data. Because causal discovery without further assumptions is an ill-posed problem, each algorithm comes with its own set of usually untestable assumptions, some of which are hard to meet in real datasets. Motivated by these considerations, this paper extensively benchmarks the empirical performance of recent causal discovery methods on observational i.i.d. data generated under different background conditions, allowing for violations of the critical assumptions required by each selected approach. Our experimental findings show that score matching-based methods demonstrate surprising performance in the false positive and false negative rate of the inferred graph in these challenging scenarios, and we provide theoretical insights into their performance. This work is also the first effort to benchmark the stability of causal discovery algorithms with respect to the values of their hyperparameters. Finally, we hope this paper will set a new standard for the evaluation of causal discovery methods and can serve as an accessible entry point for practitioners interested in the field, highlighting the empirical implications of different algorithm choices."
                },
                {
                    "title": "Adaptivity Complexity for Causal Graph Discovery",
                    "abstract": "Causal discovery from interventional data is an important problem, where the task is to design an interventional strategy that learns the hidden ground truth causal graph $G(V,E)$ on $|V| = n$ nodes while minimizing the number of performed interventions. Most prior interventional strategies broadly fall into two categories: non-adaptive and adaptive. Non-adaptive strategies decide on a single fixed set of interventions to be performed while adaptive strategies can decide on which nodes to intervene on sequentially based on past interventions. While adaptive algorithms may use exponentially fewer interventions than their non-adaptive counterparts, there are practical concerns that constrain the amount of adaptivity allowed. Motivated by this trade-off, we study the problem of $r$-adaptivity, where the algorithm designer recovers the causal graph under a total of $r$ sequential rounds whilst trying to minimize the total number of interventions. For this problem, we provide a $r$-adaptive algorithm that achieves $O(\\min\\{r,\\log n\\} \\cdot n^{1/\\min\\{r,\\log n\\}})$ approximation with respect to the verification number, a well-known lower bound for adaptive algorithms. Furthermore, for every $r$, we show that our approximation is tight. Our definition of $r$-adaptivity interpolates nicely between the non-adaptive ($r=1$) and fully adaptive ($r=n$) settings where our approximation simplifies to $O(n)$ and $O(\\log n)$ respectively, matching the best-known approximation guarantees for both extremes. Our results also extend naturally to the bounded size interventions."
                },
                {
                    "title": "Joint Bayesian Inference of Graphical Structure and Parameters with a Single Generative Flow Network",
                    "abstract": "Generative Flow Networks (GFlowNets), a class of generative models over discrete and structured sample spaces, have been previously applied to the problem of inferring the marginal posterior distribution over the directed acyclic graph (DAG) of a Bayesian Network, given a dataset of observations. Based on recent advances extending this framework to non-discrete sample spaces, we propose in this paper to approximate the joint posterior over not only the structure of a Bayesian Network, but also the parameters of its conditional probability distributions. We use a single GFlowNet whose sampling policy follows a two-phase process: the DAG is first generated sequentially one edge at a time, and then the corresponding parameters are picked once the full structure is known. Since the parameters are included in the posterior distribution, this leaves more flexibility for the local probability models of the Bayesian Network, making our approach applicable even to non-linear models parametrized by neural networks. We show that our method, called JSP-GFN, offers an accurate approximation of the joint posterior, while comparing favorably against existing methods on both simulated and real data."
                },
                {
                    "title": "Passive learning of active causal strategies in agents and language models",
                    "abstract": "What can be learned about causality and experimentation from passive data? This question is salient given recent successes of passively-trained language models in interactive domains such as tool use. Passive learning is inherently limited. However, we show that purely passive learning can in fact allow an agent to learn generalizable strategies for determining and using causal structures, as long as the agent can intervene at test time. We formally illustrate that learning a strategy of first experimenting, then seeking goals, can allow generalization from passive learning in principle. We then show empirically that agents trained via imitation on expert data can indeed generalize at test time to infer and use causal links which are never present in the training data; these agents can also generalize experimentation strategies to novel variable sets never observed in training. We then show that strategies for causal intervention and exploitation can be generalized from passive data even in a more complex environment with high-dimensional observations, with the support of natural language explanations. Explanations can even allow passive learners to generalize out-of-distribution from perfectly-confounded training data. Finally, we show that language models, trained only on passive next-word prediction, can generalize causal intervention strategies from a few-shot prompt containing examples of experimentation, together with explanations and reasoning. These results highlight the surprising power of passive learning of active causal strategies, and may help to understand the behaviors and capabilities of language models."
                },
                {
                    "title": "Modern Bayesian Experimental Design",
                    "abstract": "Bayesian experimental design (BED) provides a powerful and general framework for optimizing the design of experiments. However, its deployment often poses substantial computational challenges that can undermine its practical use. In this review, we outline how recent advances have transformed our ability to overcome these challenges and thus utilize BED effectively, before discussing some key areas for future development in the field."
                },
                {
                    "title": "Differentiable Multi-Target Causal Bayesian Experimental Design",
                    "abstract": "We introduce a gradient-based approach for the problem of Bayesian optimal experimental design to learn causal models in a batch setting -- a critical component for causal discovery from finite data where interventions can be costly or risky. Existing methods rely on greedy approximations to construct a batch of experiments while using black-box methods to optimize over a single target-state pair to intervene with. In this work, we completely dispose of the black-box optimization techniques and greedy heuristics and instead propose a conceptually simple end-to-end gradient-based optimization procedure to acquire a set of optimal intervention target-state pairs. Such a procedure enables parameterization of the design space to efficiently optimize over a batch of multi-target-state interventions, a setting which has hitherto not been explored due to its complexity. We demonstrate that our proposed method outperforms baselines and existing acquisition strategies in both single-target and multi-target settings across a number of synthetic datasets."
                },
                {
                    "title": "A Meta-Reinforcement Learning Algorithm for Causal Discovery",
                    "abstract": "Causal discovery is a major task with the utmost importance for machine learning since causal structures can enable models to go beyond pure correlation-based inference and significantly boost their performance. However, finding causal structures from data poses a significant challenge both in computational effort and accuracy, let alone its impossibility without interventions in general. In this paper, we develop a meta-reinforcement learning algorithm that performs causal discovery by learning to perform interventions such that it can construct an explicit causal graph. Apart from being useful for possible downstream applications, the estimated causal graph also provides an explanation for the data-generating process. In this article, we show that our algorithm estimates a good graph compared to the SOTA approaches, even in environments whose underlying causal structure is previously unseen. Further, we make an ablation study that shows how learning interventions contribute to the overall performance of our approach. We conclude that interventions indeed help boost the performance, efficiently yielding an accurate estimate of the causal structure of a possibly unseen environment."
                },
                {
                    "title": "Amortized Inference for Causal Structure Learning",
                    "abstract": "Inferring causal structure poses a combinatorial search problem that typically involves evaluating structures with a score or independence test. The resulting search is costly, and designing suitable scores or tests that capture prior knowledge is difficult. In this work, we propose to amortize causal structure learning. Rather than searching over structures, we train a variational inference model to directly predict the causal structure from observational or interventional data. This allows our inference model to acquire domain-specific inductive biases for causal discovery solely from data generated by a simulator, bypassing both the hand-engineering of suitable score functions and the search over graphs. The architecture of our inference model emulates permutation invariances that are crucial for statistical efficiency in structure learning, which facilitates generalization to significantly larger problem instances than seen during training. On synthetic data and semisynthetic gene expression data, our models exhibit robust generalization capabilities when subject to substantial distribution shifts and significantly outperform existing algorithms, especially in the challenging genomics domain. Our code and models are publicly available at: https://github.com/larslorch/avici."
                },
                {
                    "title": "Learning to Induce Causal Structure",
                    "abstract": "The fundamental challenge in causal induction is to infer the underlying graph structure given observational and/or interventional data. Most existing causal induction algorithms operate by generating candidate graphs and evaluating them using either score-based methods (including continuous optimization) or independence tests. In our work, we instead treat the inference process as a black box and design a neural network architecture that learns the mapping from both observational and interventional data to graph structures via supervised training on synthetic graphs. The learned model generalizes to new synthetic graphs, is robust to train-test distribution shifts, and achieves state-of-the-art performance on naturalistic graphs for low sample complexity."
                },
                {
                    "title": "Score matching enables causal discovery of nonlinear additive noise models",
                    "abstract": "This paper demonstrates how to recover causal graphs from the score of the data distribution in non-linear additive (Gaussian) noise models. Using score matching algorithms as a building block, we show how to design a new generation of scalable causal discovery methods. To showcase our approach, we also propose a new efficient method for approximating the score's Jacobian, enabling to recover the causal graph. Empirically, we find that the new algorithm, called SCORE, is competitive with state-of-the-art causal discovery methods while being significantly faster."
                },
                {
                    "title": "Interventions, Where and How? Experimental Design for Causal Models at Scale",
                    "abstract": "Causal discovery from observational and interventional data is challenging due to limited data and non-identifiability: factors that introduce uncertainty in estimating the underlying structural causal model (SCM). Selecting experiments (interventions) based on the uncertainty arising from both factors can expedite the identification of the SCM. Existing methods in experimental design for causal discovery from limited data either rely on linear assumptions for the SCM or select only the intervention target. This work incorporates recent advances in Bayesian causal discovery into the Bayesian optimal experimental design framework, allowing for active causal discovery of large, nonlinear SCMs while selecting both the interventional target and the value. We demonstrate the performance of the proposed method on synthetic graphs (Erdos-R\\`enyi, Scale Free) for both linear and nonlinear SCMs as well as on the \\emph{in-silico} single-cell gene regulatory network dataset, DREAM."
                },
                {
                    "title": "Optimizing Sequential Experimental Design with Deep Reinforcement Learning",
                    "abstract": "Bayesian approaches developed to solve the optimal design of sequential experiments are mathematically elegant but computationally challenging. Recently, techniques using amortization have been proposed to make these Bayesian approaches practical, by training a parameterized policy that proposes designs efficiently at deployment time. However, these methods may not sufficiently explore the design space, require access to a differentiable probabilistic model and can only optimize over continuous design spaces. Here, we address these limitations by showing that the problem of optimizing policies can be reduced to solving a Markov decision process (MDP). We solve the equivalent MDP with modern deep reinforcement learning techniques. Our experiments show that our approach is also computationally efficient at deployment time and exhibits state-of-the-art performance on both continuous and discrete design spaces, even when the probabilistic model is a black box."
                },
                {
                    "title": "Implicit Deep Adaptive Design: Policy-Based Experimental Design without Likelihoods",
                    "abstract": "We introduce implicit Deep Adaptive Design (iDAD), a new method for performing adaptive experiments in real-time with implicit models. iDAD amortizes the cost of Bayesian optimal experimental design (BOED) by learning a design policy network upfront, which can then be deployed quickly at the time of the experiment. The iDAD network can be trained on any model which simulates differentiable samples, unlike previous design policy work that requires a closed form likelihood and conditionally independent experiments. At deployment, iDAD allows design decisions to be made in milliseconds, in contrast to traditional BOED approaches that require heavy computation during the experiment itself. We illustrate the applicability of iDAD on a number of experiments, and show that it provides a fast and effective mechanism for performing adaptive design with implicit models."
                },
                {
                    "title": "JUWELS Cluster and Booster: Exascale Pathfinder with Modular Supercomputing Architecture at Juelich Supercomputing Centre",
                    "abstract": "JUWELS is a multi-petaflop modular supercomputer operated by Juelich Supercomputing Centre at Forschungszentrum Juelich as a European and national supercomputing resource for the Gauss Centre for Supercomputing. In addition, JUWELS serves the Earth system modeling community and the AI community within the Helmholtz Association as well. JUWELS currently consists of two modules. The first module deployed in 2018 is the so-called Cluster module. The Cluster is a BullSequana X1000 system with Intel Xeon Skylake-SP processors and Mellanox EDR InfiniBand. The second module deployed in 2020 is the so-called Booster module. The Booster is a BullSequana XH2000 system with 2nd generation AMD EPYC processors, NVIDIA Ampere GPUs and NVIDIA/Mellanox HDR Infiniband. This paper describes in detail the architecture of the system from a users perspective, and additionally provides further insights into the administrative infrastructure used to operate the supercomputer."
                },
                {
                    "title": "Scaling Vision Transformers",
                    "abstract": "Attention-based neural networks such as the Vision Transformer (ViT) have recently attained state-of-the-art results on many computer vision benchmarks. Scale is a primary ingredient in attaining excellent results, therefore, understanding a model's scaling properties is a key to designing future generations effectively. While the laws for scaling Transformer language models have been studied, it is unknown how Vision Transformers scale. To address this, we scale ViT models and data, both up and down, and characterize the relationships between error rate, data, and compute. Along the way, we refine the architecture and training of ViT, reducing memory consumption and increasing accuracy of the resulting models. As a result, we successfully train a ViT model with two billion parameters, which attains a new state-of-the-art on ImageNet of 90.45% top-1 accuracy. The model also performs well for few-shot transfer, for example, reaching 84.86% top-1 accuracy on ImageNet with only 10 examples per class."
                },
                {
                    "title": "Self-Attention Between Datapoints: Going Beyond Individual Input-Output Pairs in Deep Learning",
                    "abstract": "We challenge a common assumption underlying most supervised deep learning: that a model makes a prediction depending only on its parameters and the features of a single input. To this end, we introduce a general-purpose deep learning architecture that takes as input the entire dataset instead of processing one datapoint at a time. Our approach uses self-attention to reason about relationships between datapoints explicitly, which can be seen as realizing non-parametric models using parametric attention mechanisms. However, unlike conventional non-parametric models, we let the model learn end-to-end from the data how to make use of other datapoints for prediction. Empirically, our models solve cross-datapoint lookup and complex reasoning tasks unsolvable by traditional deep learning models. We show highly competitive results on tabular data, early results on CIFAR-10, and give insight into how the model makes use of the interactions between points."
                },
                {
                    "title": "Near-Optimal Multi-Perturbation Experimental Design for Causal Structure Learning",
                    "abstract": "Causal structure learning is a key problem in many domains. Causal structures can be learnt by performing experiments on the system of interest. We address the largely unexplored problem of designing experiments that simultaneously intervene on multiple variables. While potentially more informative than the commonly considered single-variable interventions, selecting such interventions is algorithmically much more challenging, due to the doubly-exponential combinatorial search space over sets of composite interventions. In this paper, we develop efficient algorithms for optimizing different objective functions quantifying the informativeness of experiments. By establishing novel submodularity properties of these objectives, we provide approximation guarantees for our algorithms. Our algorithms empirically perform superior to both random interventions and algorithms that only select single-variable interventions."
                },
                {
                    "title": "Deep Adaptive Design: Amortizing Sequential Bayesian Experimental Design",
                    "abstract": "We introduce Deep Adaptive Design (DAD), a method for amortizing the cost of adaptive Bayesian experimental design that allows experiments to be run in real-time. Traditional sequential Bayesian optimal experimental design approaches require substantial computation at each stage of the experiment. This makes them unsuitable for most real-world applications, where decisions must typically be made quickly. DAD addresses this restriction by learning an amortized design network upfront and then using this to rapidly run (multiple) adaptive experiments at deployment time. This network represents a design policy which takes as input the data from previous steps, and outputs the next design using a single forward pass; these design decisions can be made in milliseconds during the live experiment. To train the network, we introduce contrastive information bounds that are suitable objectives for the sequential setting, and propose a customized network architecture that exploits key symmetries. We demonstrate that DAD successfully amortizes the process of experimental design, outperforming alternative strategies on a number of problems."
                },
                {
                    "title": "Beware of the Simulated DAG! Causal Discovery Benchmarks May Be Easy to Game",
                    "abstract": "Simulated DAG models may exhibit properties that, perhaps inadvertently, render their structure identifiable and unexpectedly affect structure learning algorithms. Here, we show that marginal variance tends to increase along the causal order for generically sampled additive noise models. We introduce varsortability as a measure of the agreement between the order of increasing marginal variance and the causal order. For commonly sampled graphs and model parameters, we show that the remarkable performance of some continuous structure learning algorithms can be explained by high varsortability and matched by a simple baseline method. Yet, this performance may not transfer to real-world data where varsortability may be moderate or dependent on the choice of measurement scales. On standardized data, the same algorithms fail to identify the ground-truth DAG or its Markov equivalence class. While standardization removes the pattern in marginal variance, we show that data generating processes that incur high varsortability also leave a distinct covariance pattern that may be exploited even after standardization. Our findings challenge the significance of generic benchmarks with independently drawn parameters. The code is available at https://github.com/Scriddie/Varsortability."
                },
                {
                    "title": "Randomized Ensembled Double Q-Learning: Learning Fast Without a Model",
                    "abstract": "Using a high Update-To-Data (UTD) ratio, model-based methods have recently achieved much higher sample efficiency than previous model-free methods for continuous-action DRL benchmarks. In this paper, we introduce a simple model-free algorithm, Randomized Ensembled Double Q-Learning (REDQ), and show that its performance is just as good as, if not better than, a state-of-the-art model-based algorithm for the MuJoCo benchmark. Moreover, REDQ can achieve this performance using fewer parameters than the model-based method, and with less wall-clock run time. REDQ has three carefully integrated ingredients which allow it to achieve its high performance: (i) a UTD ratio>>1; (ii) an ensemble of Q functions; (iii) in-target minimization across a random subset of Q functions from the ensemble. Through carefully designed experiments, we provide a detailed analysis of REDQ and related model-free algorithms. To our knowledge, REDQ is the first successful model-free DRL algorithm for continuous-action spaces using a UTD ratio>>1."
                },
                {
                    "title": "CausalWorld: A Robotic Manipulation Benchmark for Causal Structure and Transfer Learning",
                    "abstract": "Despite recent successes of reinforcement learning (RL), it remains a challenge for agents to transfer learned skills to related environments. To facilitate research addressing this problem, we propose CausalWorld, a benchmark for causal structure and transfer learning in a robotic manipulation environment. The environment is a simulation of an open-source robotic platform, hence offering the possibility of sim-to-real transfer. Tasks consist of constructing 3D shapes from a given set of blocks - inspired by how children learn to build complex structures. The key strength of CausalWorld is that it provides a combinatorial family of such tasks with common causal structure and underlying factors (including, e.g., robot and object masses, colors, sizes). The user (or the agent) may intervene on all causal variables, which allows for fine-grained control over how similar different tasks (or task distributions) are. One can thus easily define training and evaluation distributions of a desired difficulty level, targeting a specific form of generalization (e.g., only changes in appearance or object mass). Further, this common parametrization facilitates defining curricula by interpolating between an initial and a target task. While users may define their own task distributions, we present eight meaningful distributions as concrete benchmarks, ranging from simple to very challenging, all of which require long-horizon planning as well as precise low-level motor control. Finally, we provide baseline results for a subset of these tasks on distinct training curricula and corresponding evaluation protocols, verifying the feasibility of the tasks in this benchmark."
                },
                {
                    "title": "Differentiable Causal Discovery from Interventional Data",
                    "abstract": "Discovering causal relationships in data is a challenging task that involves solving a combinatorial problem for which the solution is not always identifiable. A new line of work reformulates the combinatorial problem as a continuous constrained optimization one, enabling the use of different powerful optimization techniques. However, methods based on this idea do not yet make use of interventional data, which can significantly alleviate identifiability issues. In this work, we propose a neural network-based method for this task that can leverage interventional data. We illustrate the flexibility of the continuous-constrained framework by taking advantage of expressive neural architectures such as normalizing flows. We show that our approach compares favorably to the state of the art in a variety of settings, including perfect and imperfect interventions for which the targeted nodes may even be unknown."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "A Unified Stochastic Gradient Approach to Designing Bayesian-Optimal Experiments",
                    "abstract": "We introduce a fully stochastic gradient based approach to Bayesian optimal experimental design (BOED). Our approach utilizes variational lower bounds on the expected information gain (EIG) of an experiment that can be simultaneously optimized with respect to both the variational and design parameters. This allows the design process to be carried out through a single unified stochastic gradient ascent procedure, in contrast to existing approaches that typically construct a pointwise EIG estimator, before passing this estimator to a separate optimizer. We provide a number of different variational objectives including the novel adaptive contrastive estimation (ACE) bound. Finally, we show that our gradient-based approaches are able to provide effective design optimization in substantially higher dimensional settings than existing approaches."
                },
                {
                    "title": "Causal Discovery with Reinforcement Learning",
                    "abstract": "Discovering causal structure among a set of variables is a fundamental problem in many empirical sciences. Traditional score-based casual discovery methods rely on various local heuristics to search for a Directed Acyclic Graph (DAG) according to a predefined score function. While these methods, e.g., greedy equivalence search, may have attractive results with infinite samples and certain model assumptions, they are usually less satisfactory in practice due to finite data and possible violation of assumptions. Motivated by recent advances in neural combinatorial optimization, we propose to use Reinforcement Learning (RL) to search for the DAG with the best scoring. Our encoder-decoder model takes observable data as input and generates graph adjacency matrices that are used to compute rewards. The reward incorporates both the predefined score function and two penalty terms for enforcing acyclicity. In contrast with typical RL applications where the goal is to learn a policy, we use RL as a search strategy and our final output would be the graph, among all graphs generated during training, that achieves the best reward. We conduct experiments on both synthetic and real datasets, and show that the proposed approach not only has an improved search ability but also allows a flexible score function under the acyclicity constraint."
                },
                {
                    "title": "Causality",
                    "abstract": "In philosophy intuition is used in reasoning as a test-bed for the conclusions of philosophical arguments. Logic, rhetoric and intuition are the main conceptual tools in philosophical reasoning. Intuition often acts as a sort of empirical verification of the acceptability of a particular thesis. Rather like a sort of empirical test or an experimental control, to use an analogy with what happens in natural science. The basis for this method is that intuition is generalisable, or in other words, broadly speaking, it can be shared at a universal level. Moreover, intuition must have foundational validity, a primary capacity for justification that is greater than any other alternative information. It should be greater than the reference to data from the cultural and religious tradition, for example, or the recourse to the theses of classical authors. Likewise it should be able to withstand the hypotheses and empirical confirmations of scientific and technical knowledge. Experimental philosophy appears to question intuition\u2019s alleged foundational and universal nature. Intuition is a psychological phenomenon linked to what is conventionally known, according to some authors (Stanovich 1999; see Chap. 9 of Viale 2012), but not to others (Gigerenzer 2007), as System 1 of mind. Contrary to System 2, which is rational and explicit, this system is implicit and highly contextdependent. It is permeable to the influences of emotional variables derived from the cultural and environmental context. Seen in this way, it would seem difficult to affirm the thesis of the universality of human intuition. The underlying hypothesis derived from the findings of cognitive science argues the contrary: namely that intuition is local and contingent, changing in relation not only to cultural context but also to individual psychological variables, like personality traits or emotional and affective contingencies. Experimental philosophy has explored the universality"
                },
                {
                    "title": "ABCD-Strategy: Budgeted Experimental Design for Targeted Causal Structure Discovery",
                    "abstract": "Determining the causal structure of a set of variables is critical for both scientific inquiry and decision-making. However, this is often challenging in practice due to limited interventional data. Given that randomized experiments are usually expensive to perform, we propose a general framework and theory based on optimal Bayesian experimental design to select experiments for targeted causal discovery. That is, we assume the experimenter is interested in learning some function of the unknown graph (e.g., all descendants of a target node) subject to design constraints such as limits on the number of samples and rounds of experimentation. While it is in general computationally intractable to select an optimal experimental design strategy, we provide a tractable implementation with provable guarantees on both approximation and optimization quality based on submodularity. We evaluate the efficacy of our proposed method on both synthetic and real datasets, thereby demonstrating that our method realizes considerable performance gains over baseline strategies such as random sampling."
                },
                {
                    "title": "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor",
                    "abstract": "Model-free deep reinforcement learning (RL) algorithms have been demonstrated on a range of challenging decision making and control tasks. However, these methods typically suffer from two major challenges: very high sample complexity and brittle convergence properties, which necessitate meticulous hyperparameter tuning. Both of these challenges severely limit the applicability of such methods to complex, real-world domains. In this paper, we propose soft actor-critic, an off-policy actor-critic deep RL algorithm based on the maximum entropy reinforcement learning framework. In this framework, the actor aims to maximize expected reward while also maximizing entropy. That is, to succeed at the task while acting as randomly as possible. Prior deep RL methods based on this framework have been formulated as Q-learning methods. By combining off-policy updates with a stable stochastic actor-critic formulation, our method achieves state-of-the-art performance on a range of continuous control benchmark tasks, outperforming prior on-policy and off-policy methods. Furthermore, we demonstrate that, in contrast to other off-policy algorithms, our approach is very stable, achieving very similar performance across different random seeds."
                },
                {
                    "title": "Causal Structure Learning",
                    "abstract": "Graphical models can represent a multivariate distribution in a convenient and accessible form as a graph. Causal models can be viewed as a special class of graphical models that represent not only the distribution of the observed system but also the distributions under external interventions. They hence enable predictions under hypothetical interventions, which is important for decision making. The challenging task of learning causal models from data always relies on some underlying assumptions. We discuss several recently proposed structure learning algorithms and their assumptions, and we compare their empirical performance under various scenarios."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Sequential Bayesian optimal experimental design via approximate dynamic programming",
                    "abstract": "The design of multiple experiments is commonly undertaken via suboptimal strategies, such as batch (open-loop) design that omits feedback or greedy (myopic) design that does not account for future effects. This paper introduces new strategies for the optimal design of sequential experiments. First, we rigorously formulate the general sequential optimal experimental design (sOED) problem as a dynamic program. Batch and greedy designs are shown to result from special cases of this formulation. We then focus on sOED for parameter inference, adopting a Bayesian formulation with an information theoretic design objective. To make the problem tractable, we develop new numerical approaches for nonlinear design with continuous parameter, design, and observation spaces. We approximate the optimal policy by using backward induction with regression to construct and refine value function approximations in the dynamic program. The proposed algorithm iteratively generates trajectories via exploration and exploitation to improve approximation accuracy in frequently visited regions of the state space. Numerical results are verified against analytical solutions in a linear-Gaussian setting. Advantages over batch and greedy design are then demonstrated on a nonlinear source inversion problem where we seek an optimal policy for sequential sensing."
                },
                {
                    "title": "BACKSHIFT: Learning causal cyclic graphs from unknown shift interventions",
                    "abstract": "We propose a simple method to learn linear causal cyclic models in the presence of latent variables. The method relies on equilibrium data of the model recorded under a specific kind of interventions (\"shift interventions\"). The location and strength of these interventions do not have to be known and can be estimated from the data. Our method, called BACKSHIFT, only uses second moments of the data and performs simple joint matrix diagonalization, applied to differences between covariance matrices. We give a sufficient and necessary condition for identifiability of the system, which is fulfilled almost surely under some quite general assumptions if and only if there are at least three distinct experimental settings, one of which can be pure observational data. We demonstrate the performance on some simulated data and applications in flow cytometry and financial time series."
                },
                {
                    "title": "Hidden Parameter Markov Decision Processes: A Semiparametric Regression Approach for Discovering Latent Task Parametrizations",
                    "abstract": "Control applications often feature tasks with similar, but not identical, dynamics. We introduce the Hidden Parameter Markov Decision Process (HiP-MDP), a framework that parametrizes a family of related dynamical systems with a low-dimensional set of latent factors, and introduce a semiparametric regression approach for learning its structure from data. We show that a learned HiP-MDP rapidly identifies the dynamics of new task instances in several settings, flexibly adapting to task variation."
                },
                {
                    "title": "GeneNetWeaver: in silico benchmark generation and performance profiling of network inference methods",
                    "abstract": "MOTIVATION\nOver the last decade, numerous methods have been developed for inference of regulatory networks from gene expression data. However, accurate and systematic evaluation of these methods is hampered by the difficulty of constructing adequate benchmarks and the lack of tools for a differentiated analysis of network predictions on such benchmarks.\n\n\nRESULTS\nHere, we describe a novel and comprehensive method for in silico benchmark generation and performance profiling of network inference methods available to the community as an open-source software called GeneNetWeaver (GNW). In addition to the generation of detailed dynamical models of gene regulatory networks to be used as benchmarks, GNW provides a network motif analysis that reveals systematic prediction errors, thereby indicating potential ways of improving inference methods. The accuracy of network inference methods is evaluated using standard metrics such as precision-recall and receiver operating characteristic curves. We show how GNW can be used to assess the performance and identify the strengths and weaknesses of six inference methods. Furthermore, we used GNW to provide the international Dialogue for Reverse Engineering Assessments and Methods (DREAM) competition with three network inference challenges (DREAM3, DREAM4 and DREAM5).\n\n\nAVAILABILITY\nGNW is available at http://gnw.sourceforge.net along with its Java source code, user manual and supporting data.\n\n\nSUPPLEMENTARY INFORMATION\nSupplementary data are available at Bioinformatics online.\n\n\nCONTACT\ndario.floreano@epfl.ch."
                },
                {
                    "title": "Characterization and greedy learning of interventional Markov equivalence classes of directed acyclic graphs",
                    "abstract": "The investigation of directed acyclic graphs (DAGs) encoding the same Markov property, that is the same conditional independence relations of multivariate observational distributions, has a long tradition; many algorithms exist for model selection and structure learning in Markov equivalence classes. In this paper, we extend the notion of Markov equivalence of DAGs to the case of interventional distributions arising from multiple intervention experiments. We show that under reasonable assumptions on the intervention experiments, interventionalMarkov equivalence defines a finer partitioning of DAGs than observational Markov equivalence and hence improves the identifiability of causal models. We give a graph theoretic criterion for two DAGs being Markov equivalent under interventions and show that each interventional Markov equivalence class can, analogously to the observational case, be uniquely represented by a chain graph called interventional essential graph (also known as CPDAG in the observational case). These are key insights for deriving a generalization of the Greedy Equivalence Search algorithm aimed at structure learning from interventional data. This new algorithm is evaluated in a simulation study."
                },
                {
                    "title": "Random Features for Large-Scale Kernel Machines",
                    "abstract": "To accelerate the training of kernel machines, we propose to map the input data to a randomized low-dimensional feature space and then apply existing fast linear methods. The features are designed so that the inner products of the transformed data are approximately equal to those in the feature space of a user specified shift-invariant kernel. We explore two sets of random features, provide convergence bounds on their ability to approximate various radial basis kernels, and show that in large-scale classification and regression tasks linear machine learning algorithms applied to these features outperform state-of-the-art large-scale kernel machines."
                },
                {
                    "title": "The IM algorithm: a variational approach to Information Maximization",
                    "abstract": "The maximisation of information transmission over noisy channels is a common, albeit generally computationally difficult problem. We approach the difficulty of computing the mutual information for noisy channels by using a variational approximation. The resulting IM algorithm is analagous to the EM algorithm, yet maximises mutual information, as opposed to likelihood. We apply the method to several practical examples, including linear compression, population encoding and CDMA."
                },
                {
                    "title": "Causation, Prediction, and Search",
                    "abstract": "The writing is not uniformly polished and is scattered with long, awkward sentences that require some effort to unravel. I wonder if this is the result of infelicitous translation from the original German version (Wellek 1994). There are also numerous small typographical errors. More careful editing could have solved these problems before publication. There are no exercises, and so I would hesitate to use the book as a text (although it should be noted that this is not one of the author\u2019s stated aims). Although Testing Statistical Hypotheses of Equivalence has some weaknesses, it is a useful reference for those interested in the question of equivalence testing, particularly in biological applications."
                },
                {
                    "title": "Data Analysis with Bayesian Networks: A Bootstrap Approach",
                    "abstract": "In recent years there has been significant progress in algorithms and methods for inducing Bayesian networks from data. However, in complex data analysis problems, we need to go beyond being satisfied with inducing networks with high scores. We need to provide confidence measures on features of these networks: Is the existence of an edge between two nodes warranted? Is the Markov blanket of a given node robust? Can we say something about the ordering of the variables? We should be able to address these questions, even when the amount of data is not enough to induce a high scoring network. In this paper we propose Efron's Bootstrap as a computationally efficient approach for answering these questions. In addition, we propose to use these confidence measures to induce better structures from the data, and to detect the presence of latent variables."
                },
                {
                    "title": "Bayesian Experimental Design: A Review",
                    "abstract": "This paper reviews the literature on Bayesian experimental design, both for linear and nonlinear models. A uniied view of the topic is presented by putting experimental design in a decision theoretic framework. This framework justiies many optimality criteria, and opens new possibilities. Various design criteria become part of a single, coherent approach."
                },
                {
                    "title": "Bayesian Active Causal Discovery with Multi-Fidelity Experiments",
                    "abstract": "This paper studies the problem of active causal discovery when the experiments can be done based on multi-fidelity oracles, where higher fidelity experiments are more precise and expensive, while the lower ones are cheaper but less accurate. In this paper, we formally define the task of multi-fidelity active causal discovery, and design a probabilistic model for solving this problem. In specific, we first introduce a mutual-information based acquisition function to determine which variable should be intervened at which fidelity, and then a cascading model is proposed to capture the correlations between different fidelity oracles. Beyond the above basic framework, we also extend it to the batch intervention scenario. We find that the theoretical foundations behind the widely used and efficient greedy method do not hold in our problem. To solve this problem, we introduce a new concept called \u03f5 -submodular, and design a constraint based fidelity model to theoretically validate the greedy method. We conduct extensive experiments to demonstrate the effectiveness of our model."
                },
                {
                    "title": "BayesDAG: Gradient-Based Posterior Sampling for Causal Discovery",
                    "abstract": "Bayesian causal discovery aims to infer the posterior distribution over causal models from observed data, quantifying epistemic uncertainty and bene\ufb01ting downstream tasks. However, computational challenges arise due to joint inference over combinatorial space of Directed Acyclic Graphs (DAGs) and nonlinear functions. Despite recent progress towards ef\ufb01cient posterior inference over DAGs, existing methods are either limited to variational inference on node permutation matrices for linear causal models, leading to compromised inference accuracy, or continuous relaxation of adjacency matrices constrained by a DAG regularizer, which cannot ensure resulting graphs are DAGs. In this work, we introduce a scalable Bayesian causal discovery framework based on stochastic gradient Markov Chain Monte Carlo (SG-MCMC) that overcomes these limitations. Our approach directly samples DAGs from the posterior without requiring any DAG regularization, simultaneously draws function parameter samples and is applicable to both linear and nonlinear causal models. To enable our approach, we derive a novel equivalence to the permutation-based DAG learning, which opens up possibilities of using any relaxed gradient estimator de\ufb01ned over permutations. To our knowledge, this is the \ufb01rst framework applying gradient-based MCMC sampling for causal discovery. Empirical evaluations on synthetic and real-world datasets demonstrate our approach\u2019s effectiveness compared to state-of-the-art baselines."
                },
                {
                    "title": "Sample Efficient Active Learning of Causal Trees",
                    "abstract": "We consider the problem of experimental design for learning causal graphs that have a tree structure. We propose an adaptive framework that determines the next intervention based on a Bayesian prior updated with the outcomes of previous experiments, focusing on the setting where observational data is cheap (assumed infinite) and interventional data is expensive. While information greedy approaches are popular in active learning, we show that in this setting they can be exponentially suboptimal (in the number of interventions required), and instead propose an algorithm that exploits graph structure in the form of a centrality measure. If infinite interventional data is available, we show that the algorithm requires a number of interventions less than or equal to a factor of 2 times the minimum achievable number. We show that the algorithm and the associated theory can be adapted to the setting where each performed intervention yields finitely many samples. Several extensions are also presented, to the case where a specified set of nodes cannot be intervened on, to the case where $K$ interventions are scheduled at once, and to the fully adaptive case where each experiment yields only one sample. In the case of finite interventional data, through simulated experiments we show that our algorithms outperform different adaptive baseline algorithms."
                },
                {
                    "title": "Emergence of Scaling in Random Networks",
                    "abstract": "level of Co and a maximum of inverse TMR is expected when the Fermi level of LSMO is approximately at the maximum of the spin2 DOS of Co. This is consistent with the maximum of inverse TMR observed at 20.4 V for Co/STO/LSMO junctions (Fig. 3A). For a positive bias, the TMR is expected to change sign and become normal above 1 V when the Fermi level of LSMO goes down into the energy range of the majority spin d-band of Co. This is also observed in Fig. 3A. For ALO and ALO/STO barriers, a predominant tunneling of s-character electrons (see arrow in Fig. 2B) is the usual explanation of the positive polarization (6\u20138). The rapid drop with bias (Fig. 3B) is similar to what has been observed in most junctions with ALO barriers, and completely different from what is obtained when the tunneling is predominantly by d-character electrons (Fig. 3A). The origin of this rapid decrease of the TMR at relatively small bias has never been clearly explained. This is roughly consistent with the energy dependence of the DOS induced by sp-d bonding effects on the first atomic layer of ALO in the calculation of Nguyen-Mahn et al. (8) for the Co-ALO interface. But Zhang et al. (13) have also shown that a large part of the TMR drop can be attributed to the excitation of spin waves. The experiments reported here and in several recent publications (3, 4) demonstrate the important role of the electronic structure of the metal-oxide interface in determining the spin polarization of the tunneling electrons. The negative polarization for the Co-STO interface has been ascribed to d-d bonding effects between Al and Ti (4). This interpretation is similar to that proposed to explain, in terms of sp-d bonding, the positive polarization at the Co-ALO interface (8). However, there is no general theory predicting the trend of the experimental results for Co\u2014that is, a negative polarization with oxides of d elements (STO, CLO, Ta2O5) and a positive one when there are only s and p states (ALO). It is likely that the spin polarization should also depend on the position of the Fermi level with respect to the electronic levels of each character above and below the gap of the insulator. In addition, as an evanescent wave in an insulator is a Bloch wave with an imaginary wave vector, one can expect different decay lengths for Bloch waves of different character. This means that the final polarization could also depend on the thickness of the barrier, as illustrated by the calculations of MacLaren et al. for Fe/ZnSe/Fe junctions (14). The influence of the barrier on the spin polarization opens new ways to shape and optimize the TMR. Interesting bias dependencies can be obtained with barriers selecting the d electrons and probing the fine structure of the d-DOS, as in Fig. 3A. The DOS of a d-band can also be easily tailored by alloying (for example, by introduction of virtual bound states) to produce specific bias dependencies. Although here we concentrated on the problem of the spin polarization of the Co electrode and regarded the strongly spin-polarized LSMO only as a useful spin analyzer, the large TMR ratios obtained by combining Co and LSMO electrodes (50% with a STO barrier) are also an interesting result. The drawback arising from the low Curie temperature of LSMO (;350 K) is the reduction of the TMR at room temperature, down to about 5% at 300 K in Co/STO/ LSMO (4). However, other types of oxides of the double-perovskite family (for example, Sr2FeMoO6) combine electronic properties similar to those of manganites with a definitely higher Curie temperature (15). Their use in magnetic tunnel junctions is promising for a new generation of tunnel junctions with very high magnetoresistance for room-temperature applications."
                },
                {
                    "title": "On the evolution of random graphs",
                    "abstract": "(n) k edges have equal probabilities to be chosen as the next one . We shall 2 study the \"evolution\" of such a random graph if N is increased . In this investigation we endeavour to find what is the \"typical\" structure at a given stage of evolution (i . e . if N is equal, or asymptotically equal, to a given function N(n) of n) . By a \"typical\" structure we mean such a structure the probability of which tends to 1 if n -* + when N = N(n) . If A is such a property that lim Pn,N,(n ) ( A) = 1, we shall say that \u201ealmost all\" graphs Gn,N(n) n--possess this property ."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we design effective interventions for causal structure learning in empirical sciences without relying on slow and intractable likelihood-based inference methods?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing causal structure learning, particularly in fields like single-cell biology, where understanding causal relationships can lead to significant breakthroughs in gene regulation and other biological processes. By developing a method that allows for efficient intervention design, we can enhance the empirical scientific discovery process, enabling researchers to conduct more informative experiments and make better-informed decisions. This could lead to practical applications in various domains, including medicine and genetics, ultimately advancing our understanding of complex systems and improving experimental methodologies.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the complexity of accurately inferring causal graphs from data, which typically requires extensive search over graph spaces and is sensitive to data generation assumptions. Naive approaches may fail because they do not account for the intractability of likelihood calculations in empirical sciences, leading to inefficient or inaccurate intervention designs. Additionally, designing interventions that are both informative and feasible requires a deep understanding of the underlying causal relationships, which is often obscured by noise and limited data. The need for a robust scoring criterion that can effectively guide intervention selection adds another layer of complexity.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has largely focused on likelihood-based strategies for intervention design, which are slow and not well-suited for scenarios where likelihoods are intractable. Existing methods have not adequately addressed the need for efficient, adaptive intervention design that can operate without extensive causal graph inference. Barriers such as the lack of effective reward functions and the inability to leverage design space symmetries have hindered progress. Our approach differs by utilizing a transformer-based policy that amortizes the intervention design process, allowing for direct predictions of interventions based on collected data, thus overcoming limitations of prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology, Causal Amortized Structure Learning (CAASL), involves training a transformer-based design network policy using the Soft Actor-Critic (SAC) algorithm to maximize cumulative rewards over a fixed number of design iterations. The policy is trained to predict the next intervention based on the data collected so far, eliminating the need for slow causal graph inference. We will evaluate our approach using simulated design environments, focusing on reward functions derived from likelihood"
            }
        },
        "author_data": {
            "d2967382-0fe6-45d8-8fe4-f2b52f0929c4": {
                "pk": "d2967382-0fe6-45d8-8fe4-f2b52f0929c4",
                "name": "Yashas Annadani",
                "collaborators": [
                    "Stefan Bauer",
                    "Nino Scherrer",
                    "Bernhard Sch\u00f6lkopf",
                    "Panagiotis Tigas",
                    "Yarin Gal",
                    "Soma Biswas",
                    "Anirudh Goyal",
                    "Yoshua Bengio",
                    "Andrew Jesson",
                    "Jithendaraa Subramanian"
                ],
                "domain": [
                    "Causal Inference",
                    "Zero-shot Learning",
                    "Action Recognition",
                    "Bayesian Methods"
                ],
                "publications": [
                    {
                        "title": "Preserving Semantic Relations for Zero-Shot Learning",
                        "abstract": "Zero-shot learning has gained popularity due to its potential to scale recognition models without requiring additional training data. This is usually achieved by associating categories with their semantic information like attributes. However, we believe that the potential offered by this paradigm is not yet fully exploited. In this work, we propose to utilize the structure of the space spanned by the attributes using a set of relations. We devise objective functions to preserve these relations in the embedding space, thereby inducing semanticity to the embedding space. Through extensive experimental evaluation on five benchmark datasets, we demonstrate that inducing semanticity to the embedding space is beneficial for zero-shot learning. The proposed approach outperforms the state-of-the-art on the standard zero-shot setting as well as the more realistic generalized zero-shot setting. We also demonstrate how the proposed approach can be useful for making approximate semantic inferences about an image belonging to a category for which attribute information is not available."
                    },
                    {
                        "title": "Sliding Dictionary Based Sparse Representation For Action Recognition",
                        "abstract": "The task of action recognition has been in the forefront of research, given its applications in gaming, surveillance and health care. In this work, we propose a simple, yet very effective approach which works seamlessly for both offline and online action recognition using the skeletal joints. We construct a sliding dictionary which has the training data along with their time stamps. This is used to compute the sparse coefficients of the input action sequence which is divided into overlapping windows and each window gives a probability score for each action class. In addition, we compute another simple feature, which calibrates each of the action sequences to the training sequences, and models the deviation of the action from the each of the training data. Finally, a score level fusion of the two heterogeneous but complementary features for each window is obtained and the scores for the available windows are successively combined to give the confidence scores of each action class. This way of combining the scores makes the approach suitable for scenarios where only part of the sequence is available. Extensive experimental evaluation on three publicly available datasets shows the effectiveness of the proposed approach for both offline and online action recognition tasks."
                    },
                    {
                        "title": "Noise Contrastive Variational Autoencoders",
                        "abstract": "We take steps towards understanding the \"posterior collapse (PC)\" difficulty in variational autoencoders (VAEs),~i.e. a degenerate optimum in which the latent codes become independent of their corresponding inputs. We rely on calculus of variations and theoretically explore a few popular VAE models, showing that PC always occurs for non-parametric encoders and decoders. Inspired by the popular noise contrastive estimation algorithm, we propose NC-VAE where the encoder discriminates between the latent codes of real data and of some artificially generated noise, in addition to encouraging good data reconstruction abilities. Theoretically, we prove that our model cannot reach PC and provide novel lower bounds. Our method is straightforward to implement and has the same run-time as vanilla VAE. Empirically, we showcase its benefits on popular image and text datasets."
                    },
                    {
                        "title": "Selfie Detection by Synergy-Constraint Based Convolutional Neural Network",
                        "abstract": "Categorisation of huge amount of data on the multimedia platform is a crucial task. In this work, we propose a novel approach to address the subtle problem of selfie detection for image database segregation on the web, given rapid rise in number of selfies clicked. A Convolutional Neural Network (CNN) is modeled to learn a synergy feature in the common subspace of head and shoulder orientation, derived from Local Binary Pattern (LBP) and Histogram of Oriented Gradients (HOG) features respectively. This synergy was captured by projecting the aforementioned features using Canonical Correlation Analysis (CCA). We show that the resulting network's convolutional activations in the neighbourhood of spatial keypoints captured by SIFT are discriminative for selfie-detection. In general, proposed approach aids in capturing intricacies present in the image data and has the potential for usage in other subtle image analysis scenarios apart from just selfie detection. We investigate and analyse the performance of popular CNN architectures (GoogleNet, AlexNet), used for other image classification tasks, when subjected to the task of detecting the selfies on the multimedia platform. The results of the proposed approach are compared with these popular architectures on a dataset of ninety thousand images comprising of roughly equal number of selfies and non-selfies. Experimental results on this dataset shows the effectiveness of the proposed approach."
                    },
                    {
                        "title": "Variational Causal Networks: Approximate Bayesian Inference over Causal Structures",
                        "abstract": "Learning the causal structure that underlies data is a crucial step towards robust real-world decision making. The majority of existing work in causal inference focuses on determining a single directed acyclic graph (DAG) or a Markov equivalence class thereof. However, a crucial aspect to acting intelligently upon the knowledge about causal structure which has been inferred from finite data demands reasoning about its uncertainty. For instance, planning interventions to find out more about the causal mechanisms that govern our data requires quantifying epistemic uncertainty over DAGs. While Bayesian causal inference allows to do so, the posterior over DAGs becomes intractable even for a small number of variables. Aiming to overcome this issue, we propose a form of variational inference over the graphs of Structural Causal Models (SCMs). To this end, we introduce a parametric variational family modelled by an autoregressive distribution over the space of discrete DAGs. Its number of parameters does not grow exponentially with the number of variables and can be tractably learned by maximising an Evidence Lower Bound (ELBO). In our experiments, we demonstrate that the proposed variational posterior is able to provide a good approximation of the true posterior."
                    },
                    {
                        "title": "BayesDAG: Gradient-Based Posterior Inference for Causal Discovery",
                        "abstract": "Bayesian causal discovery aims to infer the posterior distribution over causal models from observed data, quantifying epistemic uncertainty and benefiting downstream tasks. However, computational challenges arise due to joint inference over combinatorial space of Directed Acyclic Graphs (DAGs) and nonlinear functions. Despite recent progress towards efficient posterior inference over DAGs, existing methods are either limited to variational inference on node permutation matrices for linear causal models, leading to compromised inference accuracy, or continuous relaxation of adjacency matrices constrained by a DAG regularizer, which cannot ensure resulting graphs are DAGs. In this work, we introduce a scalable Bayesian causal discovery framework based on a combination of stochastic gradient Markov Chain Monte Carlo (SG-MCMC) and Variational Inference (VI) that overcomes these limitations. Our approach directly samples DAGs from the posterior without requiring any DAG regularization, simultaneously draws function parameter samples and is applicable to both linear and nonlinear causal models. To enable our approach, we derive a novel equivalence to the permutation-based DAG learning, which opens up possibilities of using any relaxed gradient estimator defined over permutations. To our knowledge, this is the first framework applying gradient-based MCMC sampling for causal discovery. Empirical evaluation on synthetic and real-world datasets demonstrate our approach's effectiveness compared to state-of-the-art baselines."
                    },
                    {
                        "title": "Structure by Architecture: Structured Representations without Regularization",
                        "abstract": "We study the problem of self-supervised structured representation learning using autoencoders for downstream tasks such as generative modeling. Unlike most methods which rely on matching an arbitrary, relatively unstructured, prior distribution for sampling, we propose a sampling technique that relies solely on the independence of latent variables, thereby avoiding the trade-off between reconstruction quality and generative performance typically observed in VAEs. We design a novel autoencoder architecture capable of learning a structured representation without the need for aggressive regularization. Our structural decoders learn a hierarchy of latent variables, thereby ordering the information without any additional regularization or supervision. We demonstrate how these models learn a representation that improves results in a variety of downstream tasks including generation, disentanglement, and extrapolation using several challenging and natural image datasets."
                    },
                    {
                        "title": "Interventions, Where and How? Experimental Design for Causal Models at Scale",
                        "abstract": "Causal discovery from observational and interventional data is challenging due to limited data and non-identifiability: factors that introduce uncertainty in estimating the underlying structural causal model (SCM). Selecting experiments (interventions) based on the uncertainty arising from both factors can expedite the identification of the SCM. Existing methods in experimental design for causal discovery from limited data either rely on linear assumptions for the SCM or select only the intervention target. This work incorporates recent advances in Bayesian causal discovery into the Bayesian optimal experimental design framework, allowing for active causal discovery of large, nonlinear SCMs while selecting both the interventional target and the value. We demonstrate the performance of the proposed method on synthetic graphs (Erdos-R\\`enyi, Scale Free) for both linear and nonlinear SCMs as well as on the \\emph{in-silico} single-cell gene regulatory network dataset, DREAM."
                    },
                    {
                        "title": "Latent Variable Models for Bayesian Causal Discovery",
                        "abstract": "Learning predictors that do not rely on spurious correlations involves building causal representations. However, learning such a representation is very challenging. We, therefore, formulate the problem of learning a causal representation from high dimensional data and study causal recovery with synthetic data. This work introduces a latent variable decoder model, Decoder BCD, for Bayesian causal discovery and performs experiments in mildly supervised and unsupervised settings. We present a series of synthetic experiments to characterize important factors for causal discovery and show that using known intervention targets as labels helps in unsupervised Bayesian inference over structure and parameters of linear Gaussian additive noise latent structural causal models."
                    },
                    {
                        "title": "Differentiable Multi-Target Causal Bayesian Experimental Design",
                        "abstract": "We introduce a gradient-based approach for the problem of Bayesian optimal experimental design to learn causal models in a batch setting -- a critical component for causal discovery from finite data where interventions can be costly or risky. Existing methods rely on greedy approximations to construct a batch of experiments while using black-box methods to optimize over a single target-state pair to intervene with. In this work, we completely dispose of the black-box optimization techniques and greedy heuristics and instead propose a conceptually simple end-to-end gradient-based optimization procedure to acquire a set of optimal intervention target-state pairs. Such a procedure enables parameterization of the design space to efficiently optimize over a batch of multi-target-state interventions, a setting which has hitherto not been explored due to its complexity. We demonstrate that our proposed method outperforms baselines and existing acquisition strategies in both single-target and multi-target settings across a number of synthetic datasets."
                    },
                    {
                        "title": "Learning Latent Structural Causal Models",
                        "abstract": "Causal learning has long concerned itself with the accurate recovery of underlying causal mechanisms. Such causal modelling enables better explanations of out-of-distribution data. Prior works on causal learning assume that the high-level causal variables are given. However, in machine learning tasks, one often operates on low-level data like image pixels or high-dimensional vectors. In such settings, the entire Structural Causal Model (SCM) -- structure, parameters, \\textit{and} high-level causal variables -- is unobserved and needs to be learnt from low-level data. We treat this problem as Bayesian inference of the latent SCM, given low-level data. For linear Gaussian additive noise SCMs, we present a tractable approximate inference method which performs joint inference over the causal variables, structure and parameters of the latent SCM from random, known interventions. Experiments are performed on synthetic datasets and a causally generated image dataset to demonstrate the efficacy of our approach. We also perform image generation from unseen interventions, thereby verifying out of distribution generalization for the proposed causal model."
                    },
                    {
                        "title": "Trust Your $\\nabla$: Gradient-based Intervention Targeting for Causal Discovery",
                        "abstract": "Inferring causal structure from data is a challenging task of fundamental importance in science. Observational data are often insufficient to identify a system's causal structure uniquely. While conducting interventions (i.e., experiments) can improve the identifiability, such samples are usually challenging and expensive to obtain. Hence, experimental design approaches for causal discovery aim to minimize the number of interventions by estimating the most informative intervention target. In this work, we propose a novel Gradient-based Intervention Targeting method, abbreviated GIT, that 'trusts' the gradient estimator of a gradient-based causal discovery framework to provide signals for the intervention acquisition function. We provide extensive experiments in simulated and real-world datasets and demonstrate that GIT performs on par with competitive baselines, surpassing them in the low-data regime."
                    },
                    {
                        "title": "Challenges and Considerations in the Evaluation of Bayesian Causal Discovery",
                        "abstract": "Representing uncertainty in causal discovery is a crucial component for experimental design, and more broadly, for safe and reliable causal decision making. Bayesian Causal Discovery (BCD) offers a principled approach to encapsulating this uncertainty. Unlike non-Bayesian causal discovery, which relies on a single estimated causal graph and model parameters for assessment, evaluating BCD presents challenges due to the nature of its inferred quantity - the posterior distribution. As a result, the research community has proposed various metrics to assess the quality of the approximate posterior. However, there is, to date, no consensus on the most suitable metric(s) for evaluation. In this work, we reexamine this question by dissecting various metrics and understanding their limitations. Through extensive empirical evaluation, we find that many existing metrics fail to exhibit a strong correlation with the quality of approximation to the true posterior, especially in scenarios with low sample sizes where BCD is most desirable. We highlight the suitability (or lack thereof) of these metrics under two distinct factors: the identifiability of the underlying causal model and the quantity of available data. Both factors affect the entropy of the true posterior, indicating that the current metrics are less fitting in settings of higher entropy. Our findings underline the importance of a more nuanced evaluation of new methods by taking into account the nature of the true posterior, as well as guide and motivate the development of new evaluation procedures for this challenge."
                    },
                    {
                        "title": "Learning Neural Causal Models with Active Interventions",
                        "abstract": "Discovering causal structures from data is a challenging inference problem of fundamental importance in all areas of science. The appealing properties of neural networks have recently led to a surge of interest in differentiable neural network-based methods for learning causal structures from data. So far, differentiable causal discovery has focused on static datasets of observational or fixed interventional origin. In this work, we introduce an active intervention targeting (AIT) method which enables a quick identification of the underlying causal structure of the data-generating process. Our method significantly reduces the required number of interactions compared with random intervention targeting and is applicable for both discrete and continuous optimization formulations of learning the underlying directed acyclic graph (DAG) from data. We examine the proposed method across multiple frameworks in a wide range of settings and demonstrate superior performance on multiple benchmarks from simulated to real-world data."
                    }
                ]
            },
            "d839860f-959e-432c-9aa8-a785fad362ad": {
                "pk": "d839860f-959e-432c-9aa8-a785fad362ad",
                "name": "Panagiotis Tigas",
                "collaborators": [
                    "Yarin Gal",
                    "Yashas Annadani",
                    "Andrew Jesson",
                    "Stefan Bauer",
                    "Noor Sajid",
                    "Zafeirios Fountas",
                    "Alexey Zakharov",
                    "Tyson Hosmer",
                    "Tim Murray-Browne",
                    "Bernhard Sch\u00f6lkopf"
                ],
                "domain": [
                    "Causal Inference",
                    "Reinforcement Learning",
                    "Machine Learning",
                    "Music Generation"
                ],
                "publications": [
                    {
                        "title": "Real-time jam-session support system",
                        "abstract": "We propose a method for the problem of real time chord accompaniment of improvised music. Our implementation can learn an underlying structure of the musical performance and predict next chord. The system uses Hidden Markov Model to find the most probable chord sequence for the played melody and then a Variable Order Markov Model is used to a) learn the structure (if any) and b) predict next chord. We implemented our system in Java and MAX/Msp and compared and evaluated using objective (prediction accuracy) and subjective (questionnaire) evaluation methods."
                    },
                    {
                        "title": "Spatial Assembly: Generative Architecture With Reinforcement Learning, Self Play and Tree Search",
                        "abstract": "With this work, we investigate the use of Reinforcement Learning (RL) for the generation of spatial assemblies, by combining ideas from Procedural Generation algorithms (Wave Function Collapse algorithm (WFC)) and RL for Game Solving. WFC is a Generative Design algorithm, inspired by Constraint Solving. In WFC, one defines a set of tiles/blocks and constraints and the algorithm generates an assembly that satisfies these constraints. Casting the problem of generation of spatial assemblies as a Markov Decision Process whose states transitions are defined by WFC, we propose an algorithm that uses Reinforcement Learning and Self-Play to learn a policy that generates assemblies that maximize objectives set by the designer. Finally, we demonstrate the use of our Spatial Assembly algorithm in Architecture Design."
                    },
                    {
                        "title": "Latent Mappings: Generating Open-Ended Expressive Mappings Using Variational Autoencoders",
                        "abstract": "In many contexts, creating mappings for gestural interactions can form part of an artistic process. Creators seeking a mapping that is expressive, novel, and affords them a sense of authorship may not know how to program it up in a signal processing patch. Tools like Wekinator and MIMIC allow creators to use supervised machine learning to learn mappings from example input/output pairings. However, a creator may know a good mapping when they encounter it yet start with little sense of what the inputs or outputs should be. We call this an open-ended mapping process. Addressing this need, we introduce the latent mapping, which leverages the latent space of an unsupervised machine learning algorithm such as a Variational Autoencoder trained on a corpus of unlabelled gestural data from the creator. We illustrate it with Sonified Body, a system mapping full-body movement to sound which we explore in a residency with three dancers."
                    },
                    {
                        "title": "Interventions, Where and How? Experimental Design for Causal Models at Scale",
                        "abstract": "Causal discovery from observational and interventional data is challenging due to limited data and non-identifiability: factors that introduce uncertainty in estimating the underlying structural causal model (SCM). Selecting experiments (interventions) based on the uncertainty arising from both factors can expedite the identification of the SCM. Existing methods in experimental design for causal discovery from limited data either rely on linear assumptions for the SCM or select only the intervention target. This work incorporates recent advances in Bayesian causal discovery into the Bayesian optimal experimental design framework, allowing for active causal discovery of large, nonlinear SCMs while selecting both the interventional target and the value. We demonstrate the performance of the proposed method on synthetic graphs (Erdos-R\\`enyi, Scale Free) for both linear and nonlinear SCMs as well as on the \\emph{in-silico} single-cell gene regulatory network dataset, DREAM."
                    },
                    {
                        "title": "Deep Bayesian Active Learning for Preference Modeling in Large Language Models",
                        "abstract": "Leveraging human preferences for steering the behavior of Large Language Models (LLMs) has demonstrated notable success in recent years. Nonetheless, data selection and labeling are still a bottleneck for these systems, particularly at large scale. Hence, selecting the most informative points for acquiring human feedback may considerably reduce the cost of preference labeling and unleash the further development of LLMs. Bayesian Active Learning provides a principled framework for addressing this challenge and has demonstrated remarkable success in diverse settings. However, previous attempts to employ it for Preference Modeling did not meet such expectations. In this work, we identify that naive epistemic uncertainty estimation leads to the acquisition of redundant samples. We address this by proposing the Bayesian Active Learner for Preference Modeling (BAL-PM), a novel stochastic acquisition policy that not only targets points of high epistemic uncertainty according to the preference model but also seeks to maximize the entropy of the acquired prompt distribution in the feature space spanned by the employed LLM. Notably, our experiments demonstrate that BAL-PM requires 33% to 68% fewer preference labels in two popular human preference datasets and exceeds previous stochastic Bayesian acquisition policies."
                    },
                    {
                        "title": "Modelling non-reinforced preferences using selective attention",
                        "abstract": "How can artificial agents learn non-reinforced preferences to continuously adapt their behaviour to a changing environment? We decompose this question into two challenges: ($i$) encoding diverse memories and ($ii$) selectively attending to these for preference formation. Our proposed \\emph{no}n-\\emph{re}inforced preference learning mechanism using selective attention, \\textsc{Nore}, addresses both by leveraging the agent's world model to collect a diverse set of experiences which are interleaved with imagined roll-outs to encode memories. These memories are selectively attended to, using attention and gating blocks, to update agent's preferences. We validate \\textsc{Nore} in a modified OpenAI Gym FrozenLake environment (without any external signal) with and without volatility under a fixed model of the environment -- and compare its behaviour to \\textsc{Pepper}, a Hebbian preference learning mechanism. We demonstrate that \\textsc{Nore} provides a straightforward framework to induce exploratory preferences in the absence of external signals."
                    },
                    {
                        "title": "Global Earth Magnetic Field Modeling and Forecasting with Spherical Harmonics Decomposition",
                        "abstract": "Modeling and forecasting the solar wind-driven global magnetic field perturbations is an open challenge. Current approaches depend on simulations of computationally demanding models like the Magnetohydrodynamics (MHD) model or sampling spatially and temporally through sparse ground-based stations (SuperMAG). In this paper, we develop a Deep Learning model that forecasts in Spherical Harmonics space 2, replacing reliance on MHD models and providing global coverage at one minute cadence, improving over the current state-of-the-art which relies on feature engineering. We evaluate the performance in SuperMAG dataset (improved by 14.53%) and MHD simulations (improved by 24.35%). Additionally, we evaluate the extrapolation performance of the spherical harmonics reconstruction based on sparse ground-based stations (SuperMAG), showing that spherical harmonics can reliably reconstruct the global magnetic field as evaluated on MHD simulation."
                    },
                    {
                        "title": "Can Autonomous Vehicles Identify, Recover From, and Adapt to Distribution Shifts?",
                        "abstract": "Out-of-training-distribution (OOD) scenarios are a common challenge of learning agents at deployment, typically leading to arbitrary deductions and poorly-informed decisions. In principle, detection of and adaptation to OOD scenes can mitigate their adverse effects. In this paper, we highlight the limitations of current approaches to novel driving scenes and propose an epistemic uncertainty-aware planning method, called \\emph{robust imitative planning} (RIP). Our method can detect and recover from some distribution shifts, reducing the overconfident and catastrophic extrapolations in OOD scenes. If the model's uncertainty is too great to suggest a safe course of action, the model can instead query the expert driver for feedback, enabling sample-efficient online adaptation, a variant of our method we term \\emph{adaptive robust imitative planning} (AdaRIP). Our methods outperform current state-of-the-art approaches in the nuScenes \\emph{prediction} challenge, but since no benchmark evaluating OOD detection and adaption currently exists to assess \\emph{control}, we introduce an autonomous car novel-scene benchmark, \\texttt{CARNOVEL}, to evaluate the robustness of driving agents to a suite of tasks with distribution shifts."
                    },
                    {
                        "title": "Exploration and preference satisfaction trade-off in reward-free learning",
                        "abstract": "Biological agents have meaningful interactions with their environment despite the absence of immediate reward signals. In such instances, the agent can learn preferred modes of behaviour that lead to predictable states -- necessary for survival. In this paper, we pursue the notion that this learnt behaviour can be a consequence of reward-free preference learning that ensures an appropriate trade-off between exploration and preference satisfaction. For this, we introduce a model-based Bayesian agent equipped with a preference learning mechanism (pepper) using conjugate priors. These conjugate priors are used to augment the expected free energy planner for learning preferences over states (or outcomes) across time. Importantly, our approach enables the agent to learn preferences that encourage adaptive behaviour at test time. We illustrate this in the OpenAI Gym FrozenLake and the 3D mini-world environments -- with and without volatility. Given a constant environment, these agents learn confident (i.e., precise) preferences and act to satisfy them. Conversely, in a volatile setting, perpetual preference uncertainty maintains exploratory behaviour. Our experiments suggest that learnable (reward-free) preferences entail a trade-off between exploration and preference satisfaction. Pepper offers a straightforward framework suitable for designing adaptive agents when reward functions cannot be predefined as in real environments."
                    },
                    {
                        "title": "Differentiable Multi-Target Causal Bayesian Experimental Design",
                        "abstract": "We introduce a gradient-based approach for the problem of Bayesian optimal experimental design to learn causal models in a batch setting -- a critical component for causal discovery from finite data where interventions can be costly or risky. Existing methods rely on greedy approximations to construct a batch of experiments while using black-box methods to optimize over a single target-state pair to intervene with. In this work, we completely dispose of the black-box optimization techniques and greedy heuristics and instead propose a conceptually simple end-to-end gradient-based optimization procedure to acquire a set of optimal intervention target-state pairs. Such a procedure enables parameterization of the design space to efficiently optimize over a batch of multi-target-state interventions, a setting which has hitherto not been explored due to its complexity. We demonstrate that our proposed method outperforms baselines and existing acquisition strategies in both single-target and multi-target settings across a number of synthetic datasets."
                    },
                    {
                        "title": "Causal-BALD: Deep Bayesian Active Learning of Outcomes to Infer Treatment-Effects from Observational Data",
                        "abstract": "Estimating personalized treatment effects from high-dimensional observational data is essential in situations where experimental designs are infeasible, unethical, or expensive. Existing approaches rely on fitting deep models on outcomes observed for treated and control populations. However, when measuring individual outcomes is costly, as is the case of a tumor biopsy, a sample-efficient strategy for acquiring each result is required. Deep Bayesian active learning provides a framework for efficient data acquisition by selecting points with high uncertainty. However, existing methods bias training data acquisition towards regions of non-overlapping support between the treated and control populations. These are not sample-efficient because the treatment effect is not identifiable in such regions. We introduce causal, Bayesian acquisition functions grounded in information theory that bias data acquisition towards regions with overlapping support to maximize sample efficiency for learning personalized treatment effects. We demonstrate the performance of the proposed acquisition strategies on synthetic and semi-synthetic datasets IHDP and CMNIST and their extensions, which aim to simulate common dataset biases and pathologies."
                    },
                    {
                        "title": "Challenges and Considerations in the Evaluation of Bayesian Causal Discovery",
                        "abstract": "Representing uncertainty in causal discovery is a crucial component for experimental design, and more broadly, for safe and reliable causal decision making. Bayesian Causal Discovery (BCD) offers a principled approach to encapsulating this uncertainty. Unlike non-Bayesian causal discovery, which relies on a single estimated causal graph and model parameters for assessment, evaluating BCD presents challenges due to the nature of its inferred quantity - the posterior distribution. As a result, the research community has proposed various metrics to assess the quality of the approximate posterior. However, there is, to date, no consensus on the most suitable metric(s) for evaluation. In this work, we reexamine this question by dissecting various metrics and understanding their limitations. Through extensive empirical evaluation, we find that many existing metrics fail to exhibit a strong correlation with the quality of approximation to the true posterior, especially in scenarios with low sample sizes where BCD is most desirable. We highlight the suitability (or lack thereof) of these metrics under two distinct factors: the identifiability of the underlying causal model and the quantity of available data. Both factors affect the entropy of the true posterior, indicating that the current metrics are less fitting in settings of higher entropy. Our findings underline the importance of a more nuanced evaluation of new methods by taking into account the nature of the true posterior, as well as guide and motivate the development of new evaluation procedures for this challenge."
                    }
                ]
            },
            "652a11cb-190e-4bbe-98b0-4f38420f6f63": {
                "pk": "652a11cb-190e-4bbe-98b0-4f38420f6f63",
                "name": "Stefan Bauer",
                "collaborators": [
                    "Bernhard Sch\u00f6lkopf",
                    "Jonas Peters",
                    "Nico S. Gorbach",
                    "Joachim M. Buhmann",
                    "Mikio Furuta",
                    "Alexander Neitz",
                    "Giambattista Parascandolo",
                    "Mathieu Chevalley",
                    "Charlotte Bunne",
                    "Andreas Krause"
                ],
                "domain": [
                    "Causal Inference",
                    "Machine Learning",
                    "Graph Theory",
                    "Differential Equations"
                ],
                "publications": [
                    {
                        "title": "Almost complex 4-manifolds with vanishing first Chern class",
                        "abstract": "An odd Seiberg-Witten invariant imposes bounds on the signature of a closed, almost complex 4-manifold with vanishing first Chern class. This applies in particular to symplectic 4-manifolds of Kodaira dimension zero."
                    },
                    {
                        "title": "Refined Seiberg-Witten invariants",
                        "abstract": "This is a survey article on the stable cohomotopy refinement of Seiberg-Witten invariants containing also new results, for example:   - Stable cohomotopy groups describe path components of certain mapping spaces.   - Relation of stable cohomotopy invariants to Seiberg-Witten invariants without restriction on Betti numbers.   - Theorem on connected sums along (-2)-spheres."
                    },
                    {
                        "title": "A stable cohomotopy refinement of Seiberg-Witten invariants: II",
                        "abstract": "The main theorem describes the behaviour of the stable cohomotopy invariant defined in the first article (joint with M. Furuta) in this series of two under the operation of taking connected sums of four-manifolds: The invariant of a connected sum is the smash product (in the sense of equivariant spectra) of the invariants of the respective summands. This provides nonvanishing results even in cases in which both the integer valued Seiberg-Witten invariants and Donaldson invariants have to vanish. As a consequence, one gets new results on diffeomorphism types of decomposable four-manifolds."
                    },
                    {
                        "title": "A stable cohomotopy refinement of Seiberg-Witten invariants: I",
                        "abstract": "The monopole map defines an element in an equivariant stable cohomotopy group refining the Seiberg-Witten invariant. This first of two articles presents the details of the definition of the stable cohomotopy invariant and discusses its relation to the integer valued Seiberg-Witten invariant."
                    },
                    {
                        "title": "The Arrow of Time in Multivariate Time Series",
                        "abstract": "We prove that a time series satisfying a (linear) multivariate autoregressive moving average (VARMA) model satisfies the same model assumption in the reversed time direction, too, if all innovations are normally distributed. This reversibility breaks down if the innovations are non-Gaussian. This means that under the assumption of a VARMA process with non-Gaussian noise, the arrow of time becomes detectable. Our work thereby provides a theoretic justification of an algorithm that has been used for inferring the direction of video snippets. We present a slightly modified practical algorithm that estimates the time direction for a given sample and prove its consistency. We further investigate how the performance of the algorithm depends on sample size, number of dimensions of the time series and the order of the process. An application to real world data from economics shows that considering multivariate processes instead of univariate processes can be beneficial for estimating the time direction. Our result extends earlier work on univariate time series. It relates to the concept of causal inference, where recent methods exploit non-Gaussianity of the error terms for causal structure learning."
                    },
                    {
                        "title": "Adaptive Skip Intervals: Temporal Abstraction for Recurrent Dynamical Models",
                        "abstract": "We introduce a method which enables a recurrent dynamics model to be temporally abstract. Our approach, which we call Adaptive Skip Intervals (ASI), is based on the observation that in many sequential prediction tasks, the exact time at which events occur is irrelevant to the underlying objective. Moreover, in many situations, there exist prediction intervals which result in particularly easy-to-predict transitions. We show that there are prediction tasks for which we gain both computational efficiency and prediction accuracy by allowing the model to make predictions at a sampling rate which it can choose itself."
                    },
                    {
                        "title": "Invariant Causal Mechanisms through Distribution Matching",
                        "abstract": "Learning representations that capture the underlying data generating process is a key problem for data efficient and robust use of neural networks. One key property for robustness which the learned representation should capture and which recently received a lot of attention is described by the notion of invariance. In this work we provide a causal perspective and new algorithm for learning invariant representations. Empirically we show that this algorithm works well on a diverse set of tasks and in particular we observe state-of-the-art performance on domain generalization, where we are able to significantly boost the score of existing models."
                    },
                    {
                        "title": "Molecular causality in the advent of foundation models",
                        "abstract": "Correlation is not causation. As simple as this widely agreed-upon statement may seem, scientifically defining causality and using it to drive our modern biomedical research is immensely challenging. In this perspective, we attempt to synergise the partly disparate fields of systems biology, causal reasoning, and machine learning, to inform future approaches in the field of systems biology and molecular networks."
                    },
                    {
                        "title": "Mean-Field Variational Inference for Gradient Matching with Gaussian Processes",
                        "abstract": "Gradient matching with Gaussian processes is a promising tool for learning parameters of ordinary differential equations (ODE's). The essence of gradient matching is to model the prior over state variables as a Gaussian process which implies that the joint distribution given the ODE's and GP kernels is also Gaussian distributed. The state-derivatives are integrated out analytically since they are modelled as latent variables. However, the state variables themselves are also latent variables because they are contaminated by noise. Previous work sampled the state variables since integrating them out is \\textit{not} analytically tractable. In this paper we use mean-field approximation to establish tight variational lower bounds that decouple state variables and are therefore, in contrast to the integral over state variables, analytically tractable and even concave for a restricted family of ODE's, including nonlinear and periodic ODE's. Such variational lower bounds facilitate \"hill climbing\" to determine the maximum a posteriori estimate of ODE parameters. An additional advantage of our approach over sampling methods is the determination of a proxy to the intractable posterior distribution over state variables given observations and the ODE's."
                    },
                    {
                        "title": "Scalable Variational Inference for Dynamical Systems",
                        "abstract": "Gradient matching is a promising tool for learning parameters and state dynamics of ordinary differential equations. It is a grid free inference approach, which, for fully observable systems is at times competitive with numerical integration. However, for many real-world applications, only sparse observations are available or even unobserved variables are included in the model description. In these cases most gradient matching methods are difficult to apply or simply do not provide satisfactory results. That is why, despite the high computational cost, numerical integration is still the gold standard in many applications. Using an existing gradient matching approach, we propose a scalable variational inference framework which can infer states and parameters simultaneously, offers computational speedups, improved accuracy and works well even under model misspecifications in a partially observable system."
                    },
                    {
                        "title": "Robustly Disentangled Causal Mechanisms: Validating Deep Representations for Interventional Robustness",
                        "abstract": "The ability to learn disentangled representations that split underlying sources of variation in high dimensional, unstructured data is important for data efficient and robust use of neural networks. While various approaches aiming towards this goal have been proposed in recent times, a commonly accepted definition and validation procedure is missing. We provide a causal perspective on representation learning which covers disentanglement and domain shift robustness as special cases. Our causal framework allows us to introduce a new metric for the quantitative evaluation of deep latent variable models. We show how this metric can be estimated from labeled observational data and further provide an efficient estimation algorithm that scales linearly in the dataset size."
                    },
                    {
                        "title": "Learning stable and predictive structures in kinetic systems: Benefits of a causal approach",
                        "abstract": "Learning kinetic systems from data is one of the core challenges in many fields. Identifying stable models is essential for the generalization capabilities of data-driven inference. We introduce a computationally efficient framework, called CausalKinetiX, that identifies structure from discrete time, noisy observations, generated from heterogeneous experiments. The algorithm assumes the existence of an underlying, invariant kinetic model, a key criterion for reproducible research. Results on both simulated and real-world examples suggest that learning the structure of kinetic systems benefits from a causal perspective. The identified variables and models allow for a concise description of the dynamics across multiple experimental settings and can be used for prediction in unseen experiments. We observe significant improvements compared to well established approaches focusing solely on predictive performance, especially for out-of-sample generalization."
                    }
                ]
            },
            "8dbb4a60-a5e1-461a-9aba-79248d1cc03e": {
                "pk": "8dbb4a60-a5e1-461a-9aba-79248d1cc03e",
                "name": "Adam Foster",
                "collaborators": [
                    "Tom Rainforth",
                    "Yee Whye Teh",
                    "Rattana Pukdee",
                    "Freddie Bickford Smith",
                    "Desi R. Ivanova",
                    "Ilyas Malik",
                    "Gao-Yuan Zhang",
                    "Randall Smith",
                    "Takashi Goda",
                    "Tomohiko Hironaka"
                ],
                "domain": [
                    "Contrastive Learning",
                    "Bayesian Optimization",
                    "Stochastic Processes",
                    "Active Learning"
                ],
                "publications": [
                    {
                        "title": "Improving Transformation Invariance in Contrastive Representation Learning",
                        "abstract": "We propose methods to strengthen the invariance properties of representations obtained by contrastive learning. While existing approaches implicitly induce a degree of invariance as representations are learned, we look to more directly enforce invariance in the encoding process. To this end, we first introduce a training objective for contrastive learning that uses a novel regularizer to control how the representation changes under transformation. We show that representations trained with this objective perform better on downstream tasks and are more robust to the introduction of nuisance transformations at test time. Second, we propose a change to how test time representations are generated by introducing a feature averaging approach that combines encodings from multiple transformations of the original input, finding that this leads to across the board performance gains. Finally, we introduce the novel Spirograph dataset to explore our ideas in the context of a differentiable generative process with multiple downstream tasks, showing that our techniques for learning invariance are highly beneficial."
                    },
                    {
                        "title": "Making Better Use of Unlabelled Data in Bayesian Active Learning",
                        "abstract": "Fully supervised models are predominant in Bayesian active learning. We argue that their neglect of the information present in unlabelled data harms not just predictive performance but also decisions about what data to acquire. Our proposed solution is a simple framework for semi-supervised Bayesian active learning. We find it produces better-performing models than either conventional Bayesian active learning or semi-supervised learning with randomly acquired data. It is also easier to scale up than the conventional approach. As well as supporting a shift towards semi-supervised models, our findings highlight the importance of studying models and acquisition methods in conjunction."
                    },
                    {
                        "title": "Deep Adaptive Design: Amortizing Sequential Bayesian Experimental Design",
                        "abstract": "We introduce Deep Adaptive Design (DAD), a method for amortizing the cost of adaptive Bayesian experimental design that allows experiments to be run in real-time. Traditional sequential Bayesian optimal experimental design approaches require substantial computation at each stage of the experiment. This makes them unsuitable for most real-world applications, where decisions must typically be made quickly. DAD addresses this restriction by learning an amortized design network upfront and then using this to rapidly run (multiple) adaptive experiments at deployment time. This network represents a design policy which takes as input the data from previous steps, and outputs the next design using a single forward pass; these design decisions can be made in milliseconds during the live experiment. To train the network, we introduce contrastive information bounds that are suitable objectives for the sequential setting, and propose a customized network architecture that exploits key symmetries. We demonstrate that DAD successfully amortizes the process of experimental design, outperforming alternative strategies on a number of problems."
                    },
                    {
                        "title": "Eigenvalue Method for NEI Unit in FLASH Code",
                        "abstract": "We describe an improved nonequilibrium ionization (NEI) method that we have developed as an optional module for the FLASH magnetohydrodynamic simulation code. The method employs an eigenvalue approach rather than the earlier iterative ordinary differential equation approach to solve the stiff differential equations involved in NEI calculations. The new code also allows the atomic data to be easily updated from the AtomDB database. We compare both the updated atomic data and the methods separately. The new atomic data are shown to make a significant difference in some circumstances, although the general trends remain the same. Additionally, the new method also allows simultaneous calculation of the nonequilibrium radiative cooling, which is not included in the original method. The eigenvalue method improves the calculation efficiency overall with no loss of accuracy. We explore some common ways to present the NEI state with a sample simulation and find that using the average ionic charge difference from the equilibrium tends to be the clearest method."
                    },
                    {
                        "title": "Unbiased MLMC stochastic gradient-based optimization of Bayesian experimental designs",
                        "abstract": "In this paper we propose an efficient stochastic optimization algorithm to search for Bayesian experimental designs such that the expected information gain is maximized. The gradient of the expected information gain with respect to experimental design parameters is given by a nested expectation, for which the standard Monte Carlo method using a fixed number of inner samples yields a biased estimator. In this paper, applying the idea of randomized multilevel Monte Carlo (MLMC) methods, we introduce an unbiased Monte Carlo estimator for the gradient of the expected information gain with finite expected squared $\\ell_2$-norm and finite expected computational cost per sample. Our unbiased estimator can be combined well with stochastic gradient descent algorithms, which results in our proposal of an optimization algorithm to search for an optimal Bayesian experimental design. Numerical experiments confirm that our proposed algorithm works well not only for a simple test problem but also for a more realistic pharmacokinetic problem."
                    },
                    {
                        "title": "On Contrastive Representations of Stochastic Processes",
                        "abstract": "Learning representations of stochastic processes is an emerging problem in machine learning with applications from meta-learning to physical object models to time series. Typical methods rely on exact reconstruction of observations, but this approach breaks down as observations become high-dimensional or noise distributions become complex. To address this, we propose a unifying framework for learning contrastive representations of stochastic processes (CReSP) that does away with exact reconstruction. We dissect potential use cases for stochastic process representations, and propose methods that accommodate each. Empirically, we show that our methods are effective for learning representations of periodic functions, 3D objects and dynamical processes. Our methods tolerate noisy high-dimensional observations better than traditional approaches, and the learned representations transfer to a range of downstream tasks."
                    },
                    {
                        "title": "On the Interpretation of XSPEC Abundances and Emission Measures",
                        "abstract": "The purpose of this work is to describe the assumptions built into the X-ray spectrum fitting software XSPEC for the calculation of element abundances and emission measure of a plasma and to describe the effects when those assumptions are not accurate. The ratio of electron density to hydrogen density in XSPEC is fixed at a constant. The correct ratio can be calculated from the ionization states of the elements. We show the constant value used in XSPEC is valid to within 3.5% for a solar abundance plasma. For a plasma that deviates from solar abundance, such as hydrogen-poor or heavy element rich plasmas as found in the ejecta of supernova remnants, this ratio can smaller by factors of 0.1 to 0.001. The hydrogen emission measure, defined by integral of electron density times hydrogen density over plasma volume, is derived from the norm in XSPEC, but one needs to include the hydrogen abundance factor. For other elements, the emission measures are the XSPEC values multiplied by the element abundance factors. Using the correct electron-to-hydrogen ratio and emission measures, we show the correct electron density is smaller by the square root of the correct electron density ratio divided by the XSPEC value. Element densities and total masses (for given distance and volume) are larger by the abundance factors divided by the above square root. Because hydrogen-poor plasmas occur in the ejecta of Type Ia supernova remnants, previously estimated element masses from X-ray spectra are likely significantly underestimated."
                    },
                    {
                        "title": "A Unified Stochastic Gradient Approach to Designing Bayesian-Optimal Experiments",
                        "abstract": "We introduce a fully stochastic gradient based approach to Bayesian optimal experimental design (BOED). Our approach utilizes variational lower bounds on the expected information gain (EIG) of an experiment that can be simultaneously optimized with respect to both the variational and design parameters. This allows the design process to be carried out through a single unified stochastic gradient ascent procedure, in contrast to existing approaches that typically construct a pointwise EIG estimator, before passing this estimator to a separate optimizer. We provide a number of different variational objectives including the novel adaptive contrastive estimation (ACE) bound. Finally, we show that our gradient-based approaches are able to provide effective design optimization in substantially higher dimensional settings than existing approaches."
                    }
                ]
            }
        }
    },
    "2305.18246": {
        "paper_data": {
            "title": "Provable and Practical: Efficient Exploration in Reinforcement Learning via Langevin Monte Carlo",
            "url": "http://arxiv.org/abs/2305.18246v2",
            "arxiv_id": "2305.18246",
            "authors": [
                "Haque Ishfaq",
                "Qingfeng Lan",
                "Pan Xu",
                "A. Rupam Mahmood",
                "Doina Precup",
                "Anima Anandkumar",
                "Kamyar Azizzadenesheli"
            ],
            "abstract": "We present a scalable and effective exploration strategy based on Thompson sampling for reinforcement learning (RL). One of the key shortcomings of existing Thompson sampling algorithms is the need to perform a Gaussian approximation of the posterior distribution, which is not a good surrogate in most practical settings. We instead directly sample the Q function from its posterior distribution, by using Langevin Monte Carlo, an efficient type of Markov Chain Monte Carlo (MCMC) method. Our method only needs to perform noisy gradient descent updates to learn the exact posterior distribution of the Q function, which makes our approach easy to deploy in deep RL. We provide a rigorous theoretical analysis for the proposed method and demonstrate that, in the linear Markov decision process (linear MDP) setting, it has a regret bound of $\\tilde{O}(d^{3/2}H^{3/2}\\sqrt{T})$, where $d$ is the dimension of the feature mapping, $H$ is the planning horizon, and $T$ is the total number of steps. We apply this approach to deep RL, by using Adam optimizer to perform gradient updates. Our approach achieves better or similar results compared with state-of-the-art deep RL algorithms on several challenging exploration tasks from the Atari57 suite.",
            "introduction": " introduction with applications in data science , volume 47. Cambridge university press, 2018. (p. 40.) Jiayi Weng, Huayu Chen, Dong Yan, Kaichao You, Alexis Duburcq, Minghao Zhang, Yi Su, Hang Su, and Jun Zhu. Tianshou: A highly modularized deep reinforcement learning library. Journal of Machine Learning Research , 2022. (p. 44.) John Quan and Georg Ostrovski. DQN Zoo: Reference implementations of DQN-based agents, 2020. URL http://github.com/deepmind/dqn_zoo . (p. 44.) 15Published as a conference paper at ICLR 2024 CONTENTS 1 Introduction 1 2 Preliminary 3 3 Langevin Monte Carlo for Reinforcement Learning 3 4 Theoretical Analysis 4 5 Deep Q-Network with LMC Exploration 5 6 Experiments for LMC-LSVI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 F.1.1 Randomly generated linear MDPs . . . . . . . . . . . . . . . . . . . . . . 42 F.1.2 The RiverSwim environment . . . . . . . . . . . . . . . . . . . . . . . . . 43 F.2 N-Chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 F.3 Atari . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 F.3.1 Experiment Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 F.3.2 Additional Conclusion and Future Work 9 A Related Work 18 B Proof of the Regret Bound of LMC-LSVI 19 B.1 Supporting Lemmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 B.2 Regret Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 C Proof of Supporting Lemmas 22 C.1 Proof of Proposition B.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 C.2 Proof of Lemma B.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 C.3 Proof of Lemma B.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 C.4 Proof of",
            "references": [
                {
                    "title": "Posterior Sampling with Delayed Feedback for Reinforcement Learning with Linear Function Approximation",
                    "abstract": "Recent studies in reinforcement learning (RL) have made significant progress by leveraging function approximation to alleviate the sample complexity hurdle for better performance. Despite the success, existing provably efficient algorithms typically rely on the accessibility of immediate feedback upon taking actions. The failure to account for the impact of delay in observations can significantly degrade the performance of real-world systems due to the regret blow-up. In this work, we tackle the challenge of delayed feedback in RL with linear function approximation by employing posterior sampling, which has been shown to empirically outperform the popular UCB algorithms in a wide range of regimes. We first introduce Delayed-PSVI, an optimistic value-based algorithm that effectively explores the value function space via noise perturbation with posterior sampling. We provide the first analysis for posterior sampling algorithms with delayed feedback in RL and show our algorithm achieves $\\widetilde{O}(\\sqrt{d^3H^3 T} + d^2H^2 E[\\tau])$ worst-case regret in the presence of unknown stochastic delays. Here $E[\\tau]$ is the expected delay. To further improve its computational efficiency and to expand its applicability in high-dimensional RL problems, we incorporate a gradient-based approximate sampling scheme via Langevin dynamics for Delayed-LPSVI, which maintains the same order-optimal regret guarantee with $\\widetilde{O}(dHK)$ computational cost. Empirical evaluations are performed to demonstrate the statistical and computational efficacy of our algorithms."
                },
                {
                    "title": "Maximize to Explore: One Objective Function Fusing Estimation, Planning, and Exploration",
                    "abstract": "In online reinforcement learning (online RL), balancing exploration and exploitation is crucial for finding an optimal policy in a sample-efficient way. To achieve this, existing sample-efficient online RL algorithms typically consist of three components: estimation, planning, and exploration. However, in order to cope with general function approximators, most of them involve impractical algorithmic components to incentivize exploration, such as optimization within data-dependent level-sets or complicated sampling procedures. To address this challenge, we propose an easy-to-implement RL framework called \\textit{Maximize to Explore} (\\texttt{MEX}), which only needs to optimize \\emph{unconstrainedly} a single objective that integrates the estimation and planning components while balancing exploration and exploitation automatically. Theoretically, we prove that \\texttt{MEX} achieves a sublinear regret with general function approximations for Markov decision processes (MDP) and is further extendable to two-player zero-sum Markov games (MG). Meanwhile, we adapt deep RL baselines to design practical versions of \\texttt{MEX}, in both model-free and model-based manners, which can outperform baselines by a stable margin in various MuJoCo environments with sparse rewards. Compared with existing sample-efficient online RL algorithms with general function approximations, \\texttt{MEX} achieves similar sample efficiency while enjoying a lower computational cost and is more compatible with modern deep RL methods."
                },
                {
                    "title": "Regularization and Variance-Weighted Regression Achieves Minimax Optimality in Linear MDPs: Theory and Practice",
                    "abstract": "Mirror descent value iteration (MDVI), an abstraction of Kullback-Leibler (KL) and entropy-regularized reinforcement learning (RL), has served as the basis for recent high-performing practical RL algorithms. However, despite the use of function approximation in practice, the theoretical understanding of MDVI has been limited to tabular Markov decision processes (MDPs). We study MDVI with linear function approximation through its sample complexity required to identify an $\\varepsilon$-optimal policy with probability $1-\\delta$ under the settings of an infinite-horizon linear MDP, generative model, and G-optimal design. We demonstrate that least-squares regression weighted by the variance of an estimated optimal value function of the next state is crucial to achieving minimax optimality. Based on this observation, we present Variance-Weighted Least-Squares MDVI (VWLS-MDVI), the first theoretical algorithm that achieves nearly minimax optimal sample complexity for infinite-horizon linear MDPs. Furthermore, we propose a practical VWLS algorithm for value-based deep RL, Deep Variance Weighting (DVW). Our experiments demonstrate that DVW improves the performance of popular value-based deep RL algorithms on a set of MinAtar benchmarks."
                },
                {
                    "title": "Nearly Minimax Optimal Reinforcement Learning for Linear Markov Decision Processes",
                    "abstract": "We study reinforcement learning (RL) with linear function approximation. For episodic time-inhomogeneous linear Markov decision processes (linear MDPs) whose transition dynamic can be parameterized as a linear function of a given feature mapping, we propose the first computationally efficient algorithm that achieves the nearly minimax optimal regret $\\tilde O(d\\sqrt{H^3K})$, where $d$ is the dimension of the feature mapping, $H$ is the planning horizon, and $K$ is the number of episodes. Our algorithm is based on a weighted linear regression scheme with a carefully designed weight, which depends on a new variance estimator that (1) directly estimates the variance of the \\emph{optimal} value function, (2) monotonically decreases with respect to the number of episodes to ensure a better estimation accuracy, and (3) uses a rare-switching policy to update the value function estimator to control the complexity of the estimated value function class. Our work provides a complete answer to optimal RL with linear MDPs, and the developed algorithm and theoretical tools may be of independent interest."
                },
                {
                    "title": "Bilinear Exponential Family of MDPs: Frequentist Regret Bound with Tractable Exploration and Planning",
                    "abstract": "We study the problem of episodic reinforcement learning in continuous state-action spaces with unknown rewards and transitions. Specifically, we consider the setting where the rewards and transitions are modeled using parametric bilinear exponential families. We propose an algorithm, that a) uses penalized maximum likelihood estimators to learn the unknown parameters, b) injects a calibrated Gaussian noise in the parameter of rewards to ensure exploration, and c) leverages linearity of the bilinear exponential family transitions with respect to an underlying RKHS to perform tractable planning. We provide a frequentist regret upper-bound for our algorithm which, in the case of tabular MDPs, is order-optimal with respect to H and K, where H is the episode length and K is the number of episodes. Our analysis improves the existing bounds for the bilinear exponential family of MDPs by square root of H and removes the handcrafted clipping deployed in existing RLSVI-type algorithms."
                },
                {
                    "title": "A Provably Efficient Model-Free Posterior Sampling Method for Episodic Reinforcement Learning",
                    "abstract": "Thompson Sampling is one of the most effective methods for contextual bandits and has been generalized to posterior sampling for certain MDP settings. However, existing posterior sampling methods for reinforcement learning are limited by being model-based or lack worst-case theoretical guarantees beyond linear MDPs. This paper proposes a new model-free formulation of posterior sampling that applies to more general episodic reinforcement learning problems with theoretical guarantees. We introduce novel proof techniques to show that under suitable conditions, the worst-case regret of our posterior sampling method matches the best known results of optimization based methods. In the linear MDP setting with dimension, the regret of our algorithm scales linearly with the dimension as compared to a quadratic dependence of the existing posterior sampling-based exploration algorithms."
                },
                {
                    "title": "Making Linear MDPs Practical via Contrastive Representation Learning",
                    "abstract": "It is common to address the curse of dimensionality in Markov decision processes (MDPs) by exploiting low-rank representations. This motivates much of the recent theoretical study on linear MDPs. However, most approaches require a given representation under unrealistic assumptions about the normalization of the decomposition or introduce unresolved computational challenges in practice. Instead, we consider an alternative definition of linear MDPs that automatically ensures normalization while allowing efficient representation learning via contrastive estimation. The framework also admits confidence-adjusted index algorithms, enabling an efficient and principled approach to incorporating optimism or pessimism in the face of uncertainty. To the best of our knowledge, this provides the first practical representation learning method for linear MDPs that achieves both strong theoretical guarantees and empirical performance. Theoretically, we prove that the proposed algorithm is sample efficient in both the online and offline settings. Empirically, we demonstrate superior performance over existing state-of-the-art model-based and model-free algorithms on several benchmarks."
                },
                {
                    "title": "Langevin Monte Carlo for Contextual Bandits",
                    "abstract": "We study the efficiency of Thompson sampling for contextual bandits. Existing Thompson sampling-based algorithms need to construct a Laplace approximation (i.e., a Gaussian distribution) of the posterior distribution, which is inefficient to sample in high dimensional applications for general covariance matrices. Moreover, the Gaussian approximation may not be a good surrogate for the posterior distribution for general reward generating functions. We propose an efficient posterior sampling algorithm, viz., Langevin Monte Carlo Thompson Sampling (LMC-TS), that uses Markov Chain Monte Carlo (MCMC) methods to directly sample from the posterior distribution in contextual bandits. Our method is computationally efficient since it only needs to perform noisy gradient descent updates without constructing the Laplace approximation of the posterior distribution. We prove that the proposed algorithm achieves the same sublinear regret bound as the best Thompson sampling algorithms for a special case of contextual bandits, viz., linear contextual bandits. We conduct experiments on both synthetic data and real-world datasets on different contextual bandit models, which demonstrates that directly sampling from the posterior is both computationally efficient and competitive in performance."
                },
                {
                    "title": "From Dirichlet to Rubin: Optimistic Exploration in RL without Bonuses",
                    "abstract": "We propose the Bayes-UCBVI algorithm for reinforcement learning in tabular, stage-dependent, episodic Markov decision process: a natural extension of the Bayes-UCB algorithm by Kaufmann et al. (2012) for multi-armed bandits. Our method uses the quantile of a Q-value function posterior as upper con\ufb01dence bound on the optimal Q-value function. For Bayes-UCBVI , we prove a regret bound of order (cid:101) O ( \u221a H 3 SAT ) where H is the length of one episode, S is the number of states, A the number of actions, T the number of episodes, that matches the lower-bound of \u2126( \u221a H 3 SAT ) up to poly- log terms in H, S, A, T for a large enough T . To the best of our knowledge, this is the \ufb01rst algorithm that obtains an optimal dependence on the horizon H (and S ) without the need of an involved Bernstein-like bonus or noise. explain can be easily extended beyond tabular setting, exhibiting a strong link between our algorithm and Bayesian bootstrap (Rubin, 1981)."
                },
                {
                    "title": "Cliff Diving: Exploring Reward Surfaces in Reinforcement Learning Environments",
                    "abstract": "Visualizing optimization landscapes has led to many fundamental insights in numeric optimization, and novel improvements to optimization techniques. However, visualizations of the objective that reinforcement learning optimizes (the\"reward surface\") have only ever been generated for a small number of narrow contexts. This work presents reward surfaces and related visualizations of 27 of the most widely used reinforcement learning environments in Gym for the first time. We also explore reward surfaces in the policy gradient direction and show for the first time that many popular reinforcement learning environments have frequent\"cliffs\"(sudden large drops in expected return). We demonstrate that A2C often\"dives off\"these cliffs into low reward regions of the parameter space while PPO avoids them, confirming a popular intuition for PPO's improved performance over previous methods. We additionally introduce a highly extensible library that allows researchers to easily generate these visualizations in the future. Our findings provide new intuition to explain the successes and failures of modern RL methods, and our visualizations concretely characterize several failure modes of reinforcement learning agents in novel ways."
                },
                {
                    "title": "Optimistic posterior sampling for reinforcement learning: worst-case regret bounds",
                    "abstract": "We present an algorithm based on posterior sampling (aka Thompson sampling) that achieves near-optimal worst-case regret bounds when the underlying Markov decision process (MDP) is communicating with a finite, although unknown, diameter. Our main result is a high probability regret upper bound of [Formula: see text] for any communicating MDP with S states, A actions, and diameter D. Here, regret compares the total reward achieved by the algorithm to the total expected reward of an optimal infinite-horizon undiscounted average reward policy in time horizon T. This result closely matches the known lower bound of [Formula: see text]. Our techniques involve proving some novel results about the anti-concentration of Dirichlet distribution, which may be of independent interest."
                },
                {
                    "title": "Feel-Good Thompson Sampling for Contextual Bandits and Reinforcement Learning",
                    "abstract": "Thompson Sampling has been widely used for contextual bandit problems due to the flexibility of its modeling power. However, a general theory for this class of methods in the frequentist setting is still lacking. In this paper, we present a theoretical analysis of Thompson Sampling, with a focus on frequentist regret bounds. In this setting, we show that the standard Thompson Sampling is not aggressive enough in exploring new actions, leading to suboptimality in some pessimistic situations. A simple modification called Feel-Good Thompson Sampling, which favors high reward models more aggressively than the standard Thompson Sampling, is proposed to remedy this problem. We show that the theoretical framework can be used to derive Bayesian regret bounds for standard Thompson Sampling, and frequentist regret bounds for Feel-Good Thompson Sampling. It is shown that in both cases, we can reduce the bandit regret problem to online least squares regression estimation. For the frequentist analysis, the online least squares regression bound can be directly obtained using online aggregation techniques which have been well studied. The resulting bandit regret bound matches the minimax lower bound in the finite action case. Moreover, the analysis can be generalized to handle a class of linearly embeddable contextual bandit problems (which generalizes the popular linear contextual bandit model). The obtained result again matches the minimax lower bound. Finally we illustrate that the analysis can be extended to handle some MDP problems."
                },
                {
                    "title": "Tianshou: a Highly Modularized Deep Reinforcement Learning Library",
                    "abstract": "In this paper, we present Tianshou, a highly modularized Python library for deep reinforcement learning (DRL) that uses PyTorch as its backend. Tianshou intends to be research-friendly by providing a flexible and reliable infrastructure of DRL algorithms. It supports online and offline training with more than 20 classic algorithms through a unified interface. To facilitate related research and prove Tianshou's reliability, we have released Tianshou's benchmark of MuJoCo environments, covering eight classic algorithms with state-of-the-art performance. We open-sourced Tianshou at https://github.com/thu-ml/tianshou/."
                },
                {
                    "title": "Stochastic gradient Langevin dynamics with adaptive drifts",
                    "abstract": "We propose a class of adaptive stochastic gradient Markov chain Monte Carlo (SGMCMC) algorithms, where the drift function is adaptively adjusted according to the gradient of past samples to accelerate the convergence of the algorithm in simulations of the distributions with pathological curvatures. We establish the convergence of the proposed algorithms under mild conditions. The numerical examples indicate that the proposed algorithms can significantly outperform the popular SGMCMC algorithms, such as stochastic gradient Langevin dynamics (SGLD), stochastic gradient Hamiltonian Monte Carlo (SGHMC) and preconditioned SGLD, in both simulation and optimization tasks. In particular, the proposed algorithms can converge quickly for the distributions for which the energy landscape possesses pathological curvatures."
                },
                {
                    "title": "Randomized Exploration for Reinforcement Learning with General Value Function Approximation",
                    "abstract": "We propose a model-free reinforcement learning algorithm inspired by the popular randomized least squares value iteration (RLSVI) algorithm as well as the optimism principle. Unlike existing upper-confidence-bound (UCB) based approaches, which are often computationally intractable, our algorithm drives exploration by simply perturbing the training data with judiciously chosen i.i.d. scalar noises. To attain optimistic value function estimation without resorting to a UCB-style bonus, we introduce an optimistic reward sampling procedure. When the value functions can be represented by a function class $\\mathcal{F}$, our algorithm achieves a worst-case regret bound of $\\widetilde{O}(\\mathrm{poly}(d_EH)\\sqrt{T})$ where $T$ is the time elapsed, $H$ is the planning horizon and $d_E$ is the $\\textit{eluder dimension}$ of $\\mathcal{F}$. In the linear setting, our algorithm reduces to LSVI-PHE, a variant of RLSVI, that enjoys an $\\widetilde{\\mathcal{O}}(\\sqrt{d^3H^3T})$ regret. We complement the theory with an empirical evaluation across known difficult exploration tasks."
                },
                {
                    "title": "Provably Correct Optimization and Exploration with Non-linear Policies",
                    "abstract": "Policy optimization methods remain a powerful workhorse in empirical Reinforcement Learning (RL), with a focus on neural policies that can easily reason over complex and continuous state and/or action spaces. Theoretical understanding of strategic exploration in policy-based methods with non-linear function approximation, however, is largely missing. In this paper, we address this question by designing ENIAC, an actor-critic method that allows non-linear function approximation in the critic. We show that under certain assumptions, e.g., a bounded eluder dimension $d$ for the critic class, the learner finds a near-optimal policy in $O(\\poly(d))$ exploration rounds. The method is robust to model misspecification and strictly extends existing works on linear function approximation. We also develop some computational optimizations of our approach with slightly worse statistical guarantees and an empirical adaptation building on existing deep RL tools. We empirically evaluate this adaptation and show that it outperforms prior heuristics inspired by linear methods, establishing the value via correctly reasoning about the agent's uncertainty under non-linear function approximation."
                },
                {
                    "title": "Near-Optimal Randomized Exploration for Tabular Markov Decision Processes",
                    "abstract": "We study algorithms using randomized value functions for exploration in reinforcement learning. This type of algorithms enjoys appealing empirical performance. We show that when we use 1) a single random seed in each episode, and 2) a Bernstein-type magnitude of noise, we obtain a worst-case $\\widetilde{O}\\left(H\\sqrt{SAT}\\right)$ regret bound for episodic time-inhomogeneous Markov Decision Process where $S$ is the size of state space, $A$ is the size of action space, $H$ is the planning horizon and $T$ is the number of interactions. This bound polynomially improves all existing bounds for algorithms based on randomized value functions, and for the first time, matches the $\\Omega\\left(H\\sqrt{SAT}\\right)$ lower bound up to logarithmic factors. Our result highlights that randomized exploration can be near-optimal, which was previously achieved only by optimistic algorithms. To achieve the desired result, we develop 1) a new clipping operation to ensure both the probability of being optimistic and the probability of being pessimistic are lower bounded by a constant, and 2) a new recursive formula for the absolute value of estimation errors to analyze the regret."
                },
                {
                    "title": "Training Larger Networks for Deep Reinforcement Learning",
                    "abstract": "The success of deep learning in the computer vision and natural language processing communities can be attributed to training of very deep neural networks with millions or billions of parameters which can then be trained with massive amounts of data. However, similar trend has largely eluded training of deep reinforcement learning (RL) algorithms where larger networks do not lead to performance improvement. Previous work has shown that this is mostly due to instability during training of deep RL agents when using larger networks. In this paper, we make an attempt to understand and address training of larger networks for deep RL. We first show that naively increasing network capacity does not improve performance. Then, we propose a novel method that consists of 1) wider networks with DenseNet connection, 2) decoupling representation learning from training of RL, 3) a distributed training method to mitigate overfitting problems. Using this three-fold technique, we show that we can train very large networks that result in significant performance gains. We present several ablation studies to demonstrate the efficacy of the proposed method and some intuitive understanding of the reasons for performance gain. We show that our proposed method outperforms other baseline algorithms on several challenging locomotion tasks."
                },
                {
                    "title": "Nearly Minimax Optimal Reinforcement Learning for Linear Mixture Markov Decision Processes",
                    "abstract": "We study reinforcement learning (RL) with linear function approximation where the underlying transition probability kernel of the Markov decision process (MDP) is a linear mixture model (Jia et al., 2020; Ayoub et al., 2020; Zhou et al., 2020) and the learning agent has access to either an integration or a sampling oracle of the individual basis kernels. We propose a new Bernstein-type concentration inequality for self-normalized martingales for linear bandit problems with bounded noise. Based on the new inequality, we propose a new, computationally efficient algorithm with linear function approximation named $\\text{UCRL-VTR}^{+}$ for the aforementioned linear mixture MDPs in the episodic undiscounted setting. We show that $\\text{UCRL-VTR}^{+}$ attains an $\\tilde O(dH\\sqrt{T})$ regret where $d$ is the dimension of feature mapping, $H$ is the length of the episode and $T$ is the number of interactions with the MDP. We also prove a matching lower bound $\\Omega(dH\\sqrt{T})$ for this setting, which shows that $\\text{UCRL-VTR}^{+}$ is minimax optimal up to logarithmic factors. In addition, we propose the $\\text{UCLK}^{+}$ algorithm for the same family of MDPs under discounting and show that it attains an $\\tilde O(d\\sqrt{T}/(1-\\gamma)^{1.5})$ regret, where $\\gamma\\in [0,1)$ is the discount factor. Our upper bound matches the lower bound $\\Omega(d\\sqrt{T}/(1-\\gamma)^{1.5})$ proved in Zhou et al. (2020) up to logarithmic factors, suggesting that $\\text{UCLK}^{+}$ is nearly minimax optimal. To the best of our knowledge, these are the first computationally efficient, nearly minimax optimal algorithms for RL with linear function approximation."
                },
                {
                    "title": "Revisiting Rainbow: Promoting more insightful and inclusive deep reinforcement learning research",
                    "abstract": "Since the introduction of DQN, a vast majority of reinforcement learning research has focused on reinforcement learning with deep neural networks as function approximators. New methods are typically evaluated on a set of environments that have now become standard, such as Atari 2600 games. While these benchmarks help standardize evaluation, their computational cost has the unfortunate side effect of widening the gap between those with ample access to computational resources, and those without. In this work we argue that, despite the community's emphasis on large-scale environments, the traditional small-scale environments can still yield valuable scientific insights and can help reduce the barriers to entry for underprivileged communities. To substantiate our claims, we empirically revisit the paper which introduced the Rainbow algorithm [Hessel et al., 2018] and present some new insights into the algorithms used by Rainbow."
                },
                {
                    "title": "Faster Convergence of Stochastic Gradient Langevin Dynamics for Non-Log-Concave Sampling",
                    "abstract": "We provide a new convergence analysis of stochastic gradient Langevin dynamics (SGLD) for sampling from a class of distributions that can be non-log-concave. At the core of our approach is a novel conductance analysis of SGLD using an auxiliary time-reversible Markov Chain. Under certain conditions on the target distribution, we prove that $\\tilde O(d^4\\epsilon^{-2})$ stochastic gradient evaluations suffice to guarantee $\\epsilon$-sampling error in terms of the total variation distance, where $d$ is the problem dimension. This improves existing results on the convergence rate of SGLD (Raginsky et al., 2017; Xu et al., 2018). We further show that provided an additional Hessian Lipschitz condition on the log-density function, SGLD is guaranteed to achieve $\\epsilon$-sampling error within $\\tilde O(d^{15/4}\\epsilon^{-3/2})$ stochastic gradient evaluations. Our proof technique provides a new way to study the convergence of Langevin-based algorithms and sheds some light on the design of fast stochastic gradient-based sampling algorithms."
                },
                {
                    "title": "D2RL: Deep Dense Architectures in Reinforcement Learning",
                    "abstract": "While improvements in deep learning architectures have played a crucial role in improving the state of supervised and unsupervised learning in computer vision and natural language processing, neural network architecture choices for reinforcement learning remain relatively under-explored. We take inspiration from successful architectural choices in computer vision and generative modelling, and investigate the use of deeper networks and dense connections for reinforcement learning on a variety of simulated robotic learning benchmark environments. Our findings reveal that current methods benefit significantly from dense connections and deeper networks, across a suite of manipulation and locomotion tasks, for both proprioceptive and image-based observations. We hope that our results can serve as a strong baseline and further motivate future research into neural network architectures for reinforcement learning. The project website with code is at this link this https URL."
                },
                {
                    "title": "High-Dimensional Probability: An Introduction with Applications in Data Science",
                    "abstract": "\u00a9 2018, Cambridge University Press Let us summarize our findings. A random projection of a set T in R n onto an m-dimensional subspace approximately preserves the geometry of T if m \u2a86 d ( T ) . For..."
                },
                {
                    "title": "Stochastic Gradient Langevin Dynamics Algorithms with Adaptive Drifts",
                    "abstract": "Bayesian deep learning offers a principled way to address many issues concerning safety of artificial intelligence (AI), such as model uncertainty,model interpretability, and prediction bias. However, due to the lack of efficient Monte Carlo algorithms for sampling from the posterior of deep neural networks (DNNs), Bayesian deep learning has not yet powered our AI system. We propose a class of adaptive stochastic gradient Markov chain Monte Carlo (SGMCMC) algorithms, where the drift function is biased to enhance escape from saddle points and the bias is adaptively adjusted according to the gradient of past samples. We establish the convergence of the proposed algorithms under mild conditions, and demonstrate via numerical examples that the proposed algorithms can significantly outperform the existing SGMCMC algorithms, such as stochastic gradient Langevin dynamics (SGLD), stochastic gradient Hamiltonian Monte Carlo (SGHMC) and preconditioned SGLD, in both simulation and optimization tasks."
                },
                {
                    "title": "On Worst-case Regret of Linear Thompson Sampling",
                    "abstract": "In this paper, we consider the worst-case regret of Linear Thompson Sampling (LinTS) for the linear bandit problem. Russo and Van Roy (2014) show that the Bayesian regret of LinTS is bounded above by $\\widetilde{\\mathcal{O}}(d\\sqrt{T})$ where $T$ is the time horizon and $d$ is the number of parameters. While this bound matches the minimax lower-bounds for this problem up to logarithmic factors, the existence of a similar worst-case regret bound is still unknown. The only known worst-case regret bound for LinTS, due to Agrawal and Goyal (2013b); Abeille et al. (2017), is $\\widetilde{\\mathcal{O}}(d\\sqrt{dT})$ which requires the posterior variance to be inflated by a factor of $\\widetilde{\\mathcal{O}}(\\sqrt{d})$. While this bound is far from the minimax optimal rate by a factor of $\\sqrt{d}$, in this paper we show that it is the best possible one can get, settling an open problem stated in Russo et al. (2018). Specifically, we construct examples to show that, without the inflation, LinTS can incur linear regret up to time $\\exp(\\mathcal{O}(d))$. We then demonstrate that, under mild conditions, a slightly modified version of LinTS requires only an $\\widetilde{\\mathcal{O}}(1)$ inflation where the constant depends on the diversity of the optimal arm."
                },
                {
                    "title": "Model-Based Reinforcement Learning with Value-Targeted Regression",
                    "abstract": "This paper studies model-based reinforcement learning (RL) for regret minimization. We focus on finite-horizon episodic RL where the transition model $P$ belongs to a known family of models $\\mathcal{P}$, a special case of which is when models in $\\mathcal{P}$ take the form of linear mixtures: $P_{\\theta} = \\sum_{i=1}^{d} \\theta_{i}P_{i}$. We propose a model based RL algorithm that is based on optimism principle: In each episode, the set of models that are `consistent' with the data collected is constructed. The criterion of consistency is based on the total squared error of that the model incurs on the task of predicting \\emph{values} as determined by the last value estimate along the transitions. The next value function is then chosen by solving the optimistic planning problem with the constructed set of models. We derive a bound on the regret, which, in the special case of linear mixtures, the regret bound takes the form $\\tilde{\\mathcal{O}}(d\\sqrt{H^{3}T})$, where $H$, $T$ and $d$ are the horizon, total number of steps and dimension of $\\theta$, respectively. In particular, this regret bound is independent of the total number of states or actions, and is close to a lower bound $\\Omega(\\sqrt{HdT})$. For a general model family $\\mathcal{P}$, the regret bound is derived using the notion of the so-called Eluder dimension proposed by Russo & Van Roy (2014)."
                },
                {
                    "title": "Reinforcement Learning with General Value Function Approximation: Provably Efficient Approach via Bounded Eluder Dimension",
                    "abstract": "Value function approximation has demonstrated phenomenal empirical success in reinforcement learning (RL). Nevertheless, despite a handful of recent progress on developing theory for RL with linear function approximation, the understanding of general function approximation schemes largely remains missing. In this paper, we establish a provably efficient RL algorithm with general value function approximation. We show that if the value functions admit an approximation with a function class $\\mathcal{F}$, our algorithm achieves a regret bound of $\\widetilde{O}(\\mathrm{poly}(dH)\\sqrt{T})$ where $d$ is a complexity measure of $\\mathcal{F}$ that depends on the eluder dimension [Russo and Van Roy, 2013] and log-covering numbers, $H$ is the planning horizon, and $T$ is the number interactions with the environment. Our theory generalizes recent progress on RL with linear value function approximation and does not make explicit assumptions on the model of the environment. Moreover, our algorithm is model-free and provides a framework to justify the effectiveness of algorithms used in practice."
                },
                {
                    "title": "A Finite Time Analysis of Two Time-Scale Actor Critic Methods",
                    "abstract": "Actor-critic (AC) methods have exhibited great empirical success compared with other reinforcement learning algorithms, where the actor uses the policy gradient to improve the learning policy and the critic uses temporal difference learning to estimate the policy gradient. Under the two time-scale learning rate schedule, the asymptotic convergence of AC has been well studied in the literature. However, the non-asymptotic convergence and finite sample complexity of actor-critic methods are largely open. In this work, we provide a non-asymptotic analysis for two time-scale actor-critic methods under non-i.i.d. setting. We prove that the actor-critic method is guaranteed to find a first-order stationary point (i.e., $\\|\\nabla J(\\boldsymbol{\\theta})\\|_2^2 \\le \\epsilon$) of the non-concave performance function $J(\\boldsymbol{\\theta})$, with $\\mathcal{\\tilde{O}}(\\epsilon^{-2.5})$ sample complexity. To the best of our knowledge, this is the first work providing finite-time analysis and sample complexity bound for two time-scale actor-critic methods."
                },
                {
                    "title": "On Bonus Based Exploration Methods In The Arcade Learning Environment",
                    "abstract": "Research on exploration in reinforcement learning, as applied to Atari 2600 game-playing, has emphasized tackling difficult exploration problems such as Montezuma's Revenge (Bellemare et al., 2016). Recently, bonus-based exploration methods, which explore by augmenting the environment reward, have reached above-human average performance on such domains. In this paper we reassess popular bonus-based exploration methods within a common evaluation framework. We combine Rainbow (Hessel et al., 2018) with different exploration bonuses and evaluate its performance on Montezuma's Revenge, Bellemare et al.'s set of hard of exploration games with sparse rewards, and the whole Atari 2600 suite. We find that while exploration bonuses lead to higher score on Montezuma's Revenge they do not provide meaningful gains over the simpler epsilon-greedy scheme. In fact, we find that methods that perform best on that game often underperform epsilon-greedy on easy exploration Atari 2600 games. We find that our conclusions remain valid even when hyperparameters are tuned for these easy-exploration games. Finally, we find that none of the methods surveyed benefit from additional training samples (1 billion frames, versus Rainbow's 200 million) on Bellemare et al.'s hard exploration games. Our results suggest that recent gains in Montezuma's Revenge may be better attributed to architecture change, rather than better exploration schemes; and that the real pace of progress in exploration research for Atari 2600 games may have been obfuscated by good results on a single domain."
                },
                {
                    "title": "Learning Near Optimal Policies with Low Inherent Bellman Error",
                    "abstract": "We study the exploration problem with approximate linear action-value functions in episodic reinforcement learning under the notion of low inherent Bellman error, a condition normally employed to show convergence of approximate value iteration. First we relate this condition to other common frameworks and show that it is strictly more general than the low rank (or linear) MDP assumption of prior work. Second we provide an algorithm with a high probability regret bound $\\widetilde O(\\sum_{t=1}^H d_t \\sqrt{K} + \\sum_{t=1}^H \\sqrt{d_t} \\IBE K)$ where $H$ is the horizon, $K$ is the number of episodes, $\\IBE$ is the value if the inherent Bellman error and $d_t$ is the feature dimension at timestep $t$. In addition, we show that the result is unimprovable beyond constants and logs by showing a matching lower bound. This has two important consequences: 1) it shows that exploration is possible using only \\emph{batch assumptions} with an algorithm that achieves the optimal statistical rate for the setting we consider, which is more general than prior work on low-rank MDPs 2) the lack of closedness (measured by the inherent Bellman error) is only amplified by $\\sqrt{d_t}$ despite working in the online setting. Finally, the algorithm reduces to the celebrated \\textsc{LinUCB} when $H=1$ but with a different choice of the exploration parameter that allows handling misspecified contextual linear bandits. While computational tractability questions remain open for the MDP setting, this enriches the class of MDPs with a linear representation for the action-value function where statistically efficient reinforcement learning is possible."
                },
                {
                    "title": "On Thompson Sampling with Langevin Algorithms",
                    "abstract": "Thompson sampling for multi-armed bandit problems is known to enjoy favorable performance in both theory and practice. However, it suffers from a significant limitation computationally, arising from the need for samples from posterior distributions at every iteration. We propose two Markov Chain Monte Carlo (MCMC) methods tailored to Thompson sampling to address this issue. We construct quickly converging Langevin algorithms to generate approximate samples that have accuracy guarantees, and we leverage novel posterior concentration rates to analyze the regret of the resulting approximate Thompson sampling algorithm. Further, we specify the necessary hyperparameters for the MCMC procedure to guarantee optimal instance-dependent frequentist regret while having low computational complexity. In particular, our algorithms take advantage of both posterior concentration and a sample reuse mechanism to ensure that only a constant number of iterations and a constant amount of data is needed in each round. The resulting approximate Thompson sampling algorithm has logarithmic regret and its computational complexity does not scale with the time horizon of the algorithm."
                },
                {
                    "title": "Provably Efficient Exploration in Policy Optimization",
                    "abstract": "While policy-based reinforcement learning (RL) achieves tremendous successes in practice, it is significantly less understood in theory, especially compared with value-based RL. In particular, it remains elusive how to design a provably efficient policy optimization algorithm that incorporates exploration. To bridge such a gap, this paper proposes an Optimistic variant of the Proximal Policy Optimization algorithm (OPPO), which follows an ``optimistic version'' of the policy gradient direction. This paper proves that, in the problem of episodic Markov decision process with linear function approximation, unknown transition, and adversarial reward with full-information feedback, OPPO achieves $\\tilde{O}(\\sqrt{d^2 H^3 T} )$ regret. Here $d$ is the feature dimension, $H$ is the episode horizon, and $T$ is the total number of steps. To the best of our knowledge, OPPO is the first provably efficient policy optimization algorithm that explores."
                },
                {
                    "title": "A Finite-Time Analysis of Q-Learning with Neural Network Function Approximation",
                    "abstract": "Q-learning with neural network function approximation (neural Q-learning for short) is among the most prevalent deep reinforcement learning algorithms. Despite its empirical success, the non-asymptotic convergence rate of neural Q-learning remains virtually unknown. In this paper, we present a finite-time analysis of a neural Q-learning algorithm, where the data are generated from a Markov decision process and the action-value function is approximated by a deep ReLU neural network. We prove that neural Q-learning finds the optimal policy with $O(1/\\sqrt{T})$ convergence rate if the neural function approximator is sufficiently overparameterized, where $T$ is the number of iterations. To our best knowledge, our result is the first finite-time analysis of neural Q-learning under non-i.i.d. data assumption."
                },
                {
                    "title": "Learning with Good Feature Representations in Bandits and in RL with a Generative Model",
                    "abstract": "The construction by Du et al. (2019) implies that even if a learner is given linear features in $\\mathbb R^d$ that approximate the rewards in a bandit with a uniform error of $\\epsilon$, then searching for an action that is optimal up to $O(\\epsilon)$ requires examining essentially all actions. We use the Kiefer-Wolfowitz theorem to prove a positive result that by checking only a few actions, a learner can always find an action that is suboptimal with an error of at most $O(\\epsilon \\sqrt{d})$. Thus, features are useful when the approximation error is small relative to the dimensionality of the features. The idea is applied to stochastic bandits and reinforcement learning with a generative model where the learner has access to $d$-dimensional linear features that approximate the action-value functions for all policies to an accuracy of $\\epsilon$. For linear bandits, we prove a bound on the regret of order $\\sqrt{dn \\log(k)} + \\epsilon n \\sqrt{d} \\log(n)$ with $k$ the number of actions and $n$ the horizon. For RL we show that approximate policy iteration can learn a policy that is optimal up to an additive error of order $\\epsilon \\sqrt{d}/(1 - \\gamma)^2$ and using $d/(\\epsilon^2(1 - \\gamma)^4)$ samples from a generative model. These bounds are independent of the finer details of the features. We also investigate how the structure of the feature set impacts the tradeoff between sample complexity and estimation error."
                },
                {
                    "title": "Comments on the Du-Kakade-Wang-Yang Lower Bounds",
                    "abstract": "Du, Kakade, Wang, and Yang recently established intriguing lower bounds on sample complexity, which suggest that reinforcement learning with a misspecified representation is intractable. Another line of work, which centers around a statistic called the eluder dimension, establishes tractability of problems similar to those considered in the Du-Kakade-Wang-Yang paper. We compare these results and reconcile interpretations."
                },
                {
                    "title": "Frequentist Regret Bounds for Randomized Least-Squares Value Iteration",
                    "abstract": "We consider the exploration-exploitation dilemma in finite-horizon reinforcement learning (RL). When the state space is large or continuous, traditional tabular approaches are unfeasible and some form of function approximation is mandatory. In this paper, we introduce an optimistically-initialized variant of the popular randomized least-squares value iteration (RLSVI), a model-free algorithm where exploration is induced by perturbing the least-squares approximation of the action-value function. Under the assumption that the Markov decision process has low-rank transition dynamics, we prove that the frequentist regret of RLSVI is upper-bounded by $\\widetilde O(d^2 H^2 \\sqrt{T})$ where $ d $ are the feature dimension, $ H $ is the horizon, and $ T $ is the total number of steps. To the best of our knowledge, this is the first frequentist regret analysis for randomized exploration with function approximation."
                },
                {
                    "title": "Behaviour Suite for Reinforcement Learning",
                    "abstract": "This paper introduces the Behaviour Suite for Reinforcement Learning, or bsuite for short. bsuite is a collection of carefully-designed experiments that investigate core capabilities of reinforcement learning (RL) agents with two objectives. First, to collect clear, informative and scalable problems that capture key issues in the design of general and efficient learning algorithms. Second, to study agent behaviour through their performance on these shared benchmarks. To complement this effort, we open source this http URL, which automates evaluation and analysis of any agent on bsuite. This library facilitates reproducible and accessible research on the core issues in RL, and ultimately the design of superior learning algorithms. Our code is Python, and easy to use within existing projects. We include examples with OpenAI Baselines, Dopamine as well as new reference implementations. Going forward, we hope to incorporate more excellent experiments from the research community, and commit to a periodic review of bsuite from a committee of prominent researchers."
                },
                {
                    "title": "Provably Efficient Reinforcement Learning with Linear Function Approximation",
                    "abstract": "Modern reinforcement learning (RL) is commonly applied to practical problems with an enormous number of states, where function approximation must be deployed to approximate either the value function or the policy. The introduction of function approximation raises a fundamental set of challenges involving computational and statistical efficiency, especially given the need to manage the exploration/exploitation trade-off. As a result, a core RL question remains open: how can we design provably efficient RL algorithms that incorporate function approximation? This question persists even in a basic setting with linear dynamics and linear rewards, for which only linear function approximation is needed. This paper presents the first provable RL algorithm with both polynomial run time and polynomial sample complexity in this linear setting, without requiring a \u201csimulator\u201d or additional assumptions. Concretely, we prove that an optimistic modification of least-squares value iteration\u2014a classical algorithm frequently studied in the linear setting\u2014achieves [Formula: see text] regret, where d is the ambient dimension of feature space, H is the length of each episode, and T is the total number of steps. Importantly, such regret is independent of the number of states and actions. Funding: This work was supported by the Defense Advanced Research Projects Agency program on Lifelong Learning Machines."
                },
                {
                    "title": "Stochastic Gradient and Langevin Processes",
                    "abstract": "We prove quantitative convergence rates at which discrete Langevin-like processes converge to the invariant distribution of a related stochastic differential equation. We study the setup where the additive noise can be non-Gaussian and state-dependent and the potential function can be non-convex. We show that the key properties of these processes depend on the potential function and the second moment of the additive noise. We apply our theoretical findings to studying the convergence of Stochastic Gradient Descent (SGD) for non-convex problems and corroborate them with experiments using SGD to train deep neural networks on the CIFAR-10 dataset."
                },
                {
                    "title": "Worst-Case Regret Bounds for Exploration via Randomized Value Functions",
                    "abstract": "This paper studies a recent proposal to use randomized value functions to drive exploration in reinforcement learning. These randomized value functions are generated by injecting random noise into the training data, making the approach compatible with many popular methods for estimating parameterized value functions. By providing a worst-case regret bound for tabular finite-horizon Markov decision processes, we show that planning with respect to these randomized value functions can induce provably efficient exploration."
                },
                {
                    "title": "Reinforcement Learning in Feature Space: Matrix Bandit, Kernels, and Regret Bound",
                    "abstract": "Exploration in reinforcement learning (RL) suffers from the curse of dimensionality when the state-action space is large. A common practice is to parameterize the high-dimensional value and policy functions using given features. However existing methods either have no theoretical guarantee or suffer a regret that is exponential in the planning horizon $H$. In this paper, we propose an online RL algorithm, namely the MatrixRL, that leverages ideas from linear bandit to learn a low-dimensional representation of the probability transition model while carefully balancing the exploitation-exploration tradeoff. We show that MatrixRL achieves a regret bound ${O}\\big(H^2d\\log T\\sqrt{T}\\big)$ where $d$ is the number of features. MatrixRL has an equivalent kernelized version, which is able to work with an arbitrary kernel Hilbert space without using explicit features. In this case, the kernelized MatrixRL satisfies a regret bound ${O}\\big(H^2\\widetilde{d}\\log T\\sqrt{T}\\big)$, where $\\widetilde{d}$ is the effective dimension of the kernel space. To our best knowledge, for RL using features or kernels, our results are the first regret bounds that are near-optimal in time $T$ and dimension $d$ (or $\\widetilde{d}$) and polynomial in the planning horizon $H$."
                },
                {
                    "title": "Sampling from Non-Log-Concave Distributions via Variance-Reduced Gradient Langevin Dynamics",
                    "abstract": "We study stochastic variance reduction-based Langevin dynamic algorithms, SVRG-LD and SAGA-LD \\citep{dubey2016variance}, for sampling from non-log-concave distributions. Under certain assumptions on the log density function, we establish the convergence guarantees of SVRG-LD and SAGA-LD in $2$-Wasserstein distance. More specifically, we show that both SVRG-LD and SAGA-LD require $ \\tilde O\\big(n+n^{3/4}/\\epsilon^2 + n^{1/2}/\\epsilon^4\\big)\\cdot \\exp\\big(\\tilde O(d+\\gamma)\\big)$ stochastic gradient evaluations to achieve $\\epsilon$-accuracy in $2$-Wasserstein distance, which outperforms the $ \\tilde O\\big(n/\\epsilon^4\\big)\\cdot \\exp\\big(\\tilde O(d+\\gamma)\\big)$ gradient complexity achieved by Langevin Monte Carlo Method \\citep{raginsky2017non}. Experiments on synthetic data and real data back up our theory."
                },
                {
                    "title": "Sample-Optimal Parametric Q-Learning Using Linearly Additive Features",
                    "abstract": "Consider a Markov decision process (MDP) that admits a set of state-action features, which can linearly express the process's probabilistic transition model. We propose a parametric Q-learning algorithm that finds an approximate-optimal policy using a sample size proportional to the feature dimension $K$ and invariant with respect to the size of the state space. To further improve its sample efficiency, we exploit the monotonicity property and intrinsic noise structure of the Bellman operator, provided the existence of anchor state-actions that imply implicit non-negativity in the feature space. We augment the algorithm using techniques of variance reduction, monotonicity preservation, and confidence bounds. It is proved to find a policy which is $\\epsilon$-optimal from any initial state with high probability using $\\widetilde{O}(K/\\epsilon^2(1-\\gamma)^3)$ sample transitions for arbitrarily large-scale MDP with a discount factor $\\gamma\\in(0,1)$. A matching information-theoretical lower bound is proved, confirming the sample optimality of the proposed method with respect to all parameters (up to polylog factors)."
                },
                {
                    "title": "Where Did My Optimum Go?: An Empirical Analysis of Gradient Descent Optimization in Policy Gradient Methods",
                    "abstract": "Recent analyses of certain gradient descent optimization methods have shown that performance can degrade in some settings - such as with stochasticity or implicit momentum. In deep reinforcement learning (Deep RL), such optimization methods are often used for training neural networks via the temporal difference error or policy gradient. As an agent improves over time, the optimization target changes and thus the loss landscape (and local optima) change. Due to the failure modes of those methods, the ideal choice of optimizer for Deep RL remains unclear. As such, we provide an empirical analysis of the effects that a wide range of gradient descent optimizers and their hyperparameters have on policy gradient methods, a subset of Deep RL algorithms, for benchmark continuous control tasks. We find that adaptive optimizers have a narrow window of effective learning rates, diverging in other cases, and that the effectiveness of momentum varies depending on the properties of the environment. Our analysis suggests that there is significant interplay between the dynamics of the environment and Deep RL algorithm properties which aren't necessarily accounted for by traditional adaptive gradient methods. We provide suggestions for optimal settings of current methods and further lines of research based on our findings."
                },
                {
                    "title": "Exploration by Random Network Distillation",
                    "abstract": "We introduce an exploration bonus for deep reinforcement learning methods that is easy to implement and adds minimal overhead to the computation performed. The bonus is the error of a neural network predicting features of the observations given by a fixed randomly initialized neural network. We also introduce a method to flexibly combine intrinsic and extrinsic rewards. We find that the random network distillation (RND) bonus combined with this increased flexibility enables significant progress on several hard exploration Atari games. In particular we establish state of the art performance on Montezuma's Revenge, a game famously difficult for deep reinforcement learning methods. To the best of our knowledge, this is the first method that achieves better than average human performance on this game without using demonstrations or having access to the underlying state of the game, and occasionally completes the first level."
                },
                {
                    "title": "Is Q-learning Provably Efficient?",
                    "abstract": "Model-free reinforcement learning (RL) algorithms, such as Q-learning, directly parameterize and update value functions or policies without explicitly modeling the environment. They are typically simpler, more flexible to use, and thus more prevalent in modern deep RL than model-based approaches. However, empirical work has suggested that model-free algorithms may require more samples to learn [Deisenroth and Rasmussen 2011, Schulman et al. 2015]. The theoretical question of \"whether model-free algorithms can be made sample efficient\" is one of the most fundamental questions in RL, and remains unsolved even in the basic scenario with finitely many states and actions. \nWe prove that, in an episodic MDP setting, Q-learning with UCB exploration achieves regret $\\tilde{O}(\\sqrt{H^3 SAT})$, where $S$ and $A$ are the numbers of states and actions, $H$ is the number of steps per episode, and $T$ is the total number of steps. This sample efficiency matches the optimal regret that can be achieved by any model-based approach, up to a single $\\sqrt{H}$ factor. To the best of our knowledge, this is the first analysis in the model-free setting that establishes $\\sqrt{T}$ regret without requiring access to a \"simulator.\""
                },
                {
                    "title": "Implicit Quantile Networks for Distributional Reinforcement Learning",
                    "abstract": "In this work, we build on recent advances in distributional reinforcement learning to give a generally applicable, flexible, and state-of-the-art distributional variant of DQN. We achieve this by using quantile regression to approximate the full quantile function for the state-action return distribution. By reparameterizing a distribution over the sample space, this yields an implicitly defined return distribution and gives rise to a large class of risk-sensitive policies. We demonstrate improved performance on the 57 Atari 2600 games in the ALE, and use our algorithm's implicitly defined distributions to study the effects of risk-sensitive policies in Atari games."
                },
                {
                    "title": "Randomized Prior Functions for Deep Reinforcement Learning",
                    "abstract": "Dealing with uncertainty is essential for efficient reinforcement learning. There is a growing literature on uncertainty estimation for deep learning from fixed datasets, but many of the most popular approaches are poorly-suited to sequential decision problems. Other methods, such as bootstrap sampling, have no mechanism for uncertainty that does not come from the observed data. We highlight why this can be a crucial shortcoming and propose a simple remedy through addition of a randomized untrainable `prior' network to each ensemble member. We prove that this approach is efficient with linear representations, provide simple illustrations of its efficacy with nonlinear representations and show that this approach scales to large-scale problems far better than previous attempts."
                },
                {
                    "title": "Efficient Exploration Through Bayesian Deep Q-Networks",
                    "abstract": "We propose Bayesian Deep Q-Network (BDQN), a practical Thompson sampling based Reinforcement Learning (RL) Algorithm. Thompson sampling allows for targeted exploration in high dimensions through posterior sampling but is usually computationally expensive. We address this limitation by introducing uncertainty only at the output layer of the network through a Bayesian Linear Regression (BLR) model, which can be trained with fast closed-form updates and its samples can be drawn efficiently through the Gaussian distribution. We apply our method to a wide range of Atari games in Arcade Learning Environments. Since BDQN carries out more efficient exploration, it is able to reach higher rewards substantially faster than a key baseline, double deep Q network DDQN."
                },
                {
                    "title": "Distributional Reinforcement Learning with Quantile Regression",
                    "abstract": "\n \n In reinforcement learning (RL), an agent interacts with the environment by taking actions and observing the next state and reward. When sampled probabilistically, these state transitions, rewards, and actions can all induce randomness in the observed long-term return. Traditionally, reinforcement learning algorithms average over this randomness to estimate the value function. In this paper, we build on recent work advocating a distributional approach to reinforcement learning in which the distribution over returns is modeled explicitly instead of only estimating the mean. That is, we examine methods of learning the value distribution instead of the value function. We give results that close a number of gaps between the theoretical and algorithmic results given by Bellemare, Dabney, and Munos (2017). First, we extend existing results to the approximate distribution setting. Second, we present a novel distributional reinforcement learning algorithm consistent with our theoretical formulation. Finally, we evaluate this new algorithm on the Atari 2600 games, observing that it significantly outperforms many of the recent improvements on DQN, including the related distributional algorithm C51.\n \n"
                },
                {
                    "title": "Global Convergence of Langevin Dynamics Based Algorithms for Nonconvex Optimization",
                    "abstract": "We present a unified framework to analyze the global convergence of Langevin dynamics based algorithms for nonconvex finite-sum optimization with $n$ component functions. At the core of our analysis is a direct analysis of the ergodicity of the numerical approximations to Langevin dynamics, which leads to faster convergence rates. Specifically, we show that gradient Langevin dynamics (GLD) and stochastic gradient Langevin dynamics (SGLD) converge to the almost minimizer within $\\tilde O\\big(nd/(\\lambda\\epsilon) \\big)$ and $\\tilde O\\big(d^7/(\\lambda^5\\epsilon^5) \\big)$ stochastic gradient evaluations respectively, where $d$ is the problem dimension, and $\\lambda$ is the spectral gap of the Markov chain generated by GLD. Both of the results improve upon the best known gradient complexity results. Furthermore, for the first time we prove the global convergence guarantee for variance reduced stochastic gradient Langevin dynamics (VR-SGLD) to the almost minimizer after $\\tilde O\\big(\\sqrt{n}d^5/(\\lambda^4\\epsilon^{5/2})\\big)$ stochastic gradient evaluations, which outperforms the gradient complexities of GLD and SGLD in a wide regime. Our theoretical analyses shed some light on using Langevin dynamics based algorithms for nonconvex optimization with provable guarantees."
                },
                {
                    "title": "A Distributional Perspective on Reinforcement Learning",
                    "abstract": "In this paper we argue for the fundamental importance of the value distribution: the distribution of the random return received by a reinforcement learning agent. This is in contrast to the common approach to reinforcement learning which models the expectation of this return, or value. Although there is an established body of literature studying the value distribution, thus far it has always been used for a specific purpose such as implementing risk-aware behaviour. We begin with theoretical results in both the policy evaluation and control settings, exposing a significant distributional instability in the latter. We then use the distributional perspective to design a new algorithm which applies Bellman's equation to the learning of approximate value distributions. We evaluate our algorithm using the suite of games from the Arcade Learning Environment. We obtain both state-of-the-art results and anecdotal evidence demonstrating the importance of the value distribution in approximate reinforcement learning. Finally, we combine theoretical and empirical evidence to highlight the ways in which the value distribution impacts learning in the approximate setting."
                },
                {
                    "title": "Noisy Networks for Exploration",
                    "abstract": "We introduce NoisyNet, a deep reinforcement learning agent with parametric noise added to its weights, and show that the induced stochasticity of the agent's policy can be used to aid efficient exploration. The parameters of the noise are learned with gradient descent along with the remaining network weights. NoisyNet is straightforward to implement and adds little computational overhead. We find that replacing the conventional exploration heuristics for A3C, DQN and dueling agents (entropy reward and $\\epsilon$-greedy respectively) with NoisyNet yields substantially higher scores for a wide range of Atari games, in some cases advancing the agent from sub to super-human performance."
                },
                {
                    "title": "Parameter Space Noise for Exploration",
                    "abstract": "Deep reinforcement learning (RL) methods generally engage in exploratory behavior through noise injection in the action space. An alternative is to add noise directly to the agent's parameters, which can lead to more consistent exploration and a richer set of behaviors. Methods such as evolutionary strategies use parameter perturbations, but discard all temporal structure in the process and require significantly more samples. Combining parameter noise with traditional RL methods allows to combine the best of both worlds. We demonstrate that both off- and on-policy methods benefit from this approach through experimental comparison of DQN, DDPG, and TRPO on high-dimensional discrete action environments as well as continuous control tasks. Our results show that RL with parameter noise learns more efficiently than traditional RL with action space noise and evolutionary strategies individually."
                },
                {
                    "title": "Deep Exploration via Randomized Value Functions",
                    "abstract": "We study the use of randomized value functions to guide deep exploration in reinforcement learning. This offers an elegant means for synthesizing statistically and computationally efficient exploration with common practical approaches to value function learning. We present several reinforcement learning algorithms that leverage randomized value functions and demonstrate their efficacy through computational studies. We also prove a regret bound that establishes statistical efficiency with a tabular representation."
                },
                {
                    "title": "Minimax Regret Bounds for Reinforcement Learning",
                    "abstract": "We consider the problem of provably optimal exploration in reinforcement learning for finite horizon MDPs. We show that an optimistic modification to value iteration achieves a regret bound of $\\tilde{O}( \\sqrt{HSAT} + H^2S^2A+H\\sqrt{T})$ where $H$ is the time horizon, $S$ the number of states, $A$ the number of actions and $T$ the number of time-steps. This result improves over the best previous known bound $\\tilde{O}(HS \\sqrt{AT})$ achieved by the UCRL2 algorithm of Jaksch et al., 2010. The key significance of our new results is that when $T\\geq H^3S^3A$ and $SA\\geq H$, it leads to a regret of $\\tilde{O}(\\sqrt{HSAT})$ that matches the established lower bound of $\\Omega(\\sqrt{HSAT})$ up to a logarithmic factor. Our analysis contains two key insights. We use careful application of concentration inequalities to the optimal value function as a whole, rather than to the transitions probabilities (to improve scaling in $S$), and we define Bernstein-based \"exploration bonuses\" that use the empirical variance of the estimated values at the next states (to improve scaling in $H$)."
                },
                {
                    "title": "Count-Based Exploration with Neural Density Models",
                    "abstract": "Bellemare et al. (2016) introduced the notion of a pseudo-count, derived from a density model, to generalize count-based exploration to non-tabular reinforcement learning. This pseudo-count was used to generate an exploration bonus for a DQN agent and combined with a mixed Monte Carlo update was sufficient to achieve state of the art on the Atari 2600 game Montezuma's Revenge. We consider two questions left open by their work: First, how important is the quality of the density model for exploration? Second, what role does the Monte Carlo update play in exploration? We answer the first question by demonstrating the use of PixelCNN, an advanced neural density model for images, to supply a pseudo-count. In particular, we examine the intrinsic difficulties in adapting Bellemare et al.'s approach when assumptions about the model are violated. The result is a more practical and general algorithm requiring no special apparatus. We combine PixelCNN pseudo-counts with different agent architectures to dramatically improve the state of the art on several hard Atari games. One surprising finding is that the mixed Monte Carlo update is a powerful facilitator of exploration in the sparsest of settings, including Montezuma's Revenge."
                },
                {
                    "title": "Variance Reduction in Stochastic Gradient Langevin Dynamics",
                    "abstract": "Stochastic gradient-based Monte Carlo methods such as stochastic gradient Langevin dynamics are useful tools for posterior inference on large scale datasets in many machine learning applications. These methods scale to large datasets by using noisy gradients calculated using a mini-batch or subset of the dataset. However, the high variance inherent in these noisy gradients degrades performance and leads to slower mixing. In this paper, we present techniques for reducing variance in stochastic gradient Langevin dynamics, yielding novel stochastic Monte Carlo methods that improve performance by reducing the variance in the stochastic gradient. We show that our proposed method has better theoretical guarantees on convergence rate than stochastic Langevin dynamics. This is complemented by impressive empirical results obtained on a variety of real world datasets, and on four different machine learning tasks (regression, classification, independent component analysis and mixture modeling). These theoretical and empirical contributions combine to make a compelling case for using variance reduction in stochastic Monte Carlo methods."
                },
                {
                    "title": "#Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning",
                    "abstract": "Count-based exploration algorithms are known to perform near-optimally when used in conjunction with tabular reinforcement learning (RL) methods for solving small discrete Markov decision processes (MDPs). It is generally thought that count-based methods cannot be applied in high-dimensional state spaces, since most states will only occur once. Recent deep RL exploration strategies are able to deal with high-dimensional continuous state spaces through complex heuristics, often relying on optimism in the face of uncertainty or intrinsic motivation. In this work, we describe a surprising finding: a simple generalization of the classic count-based approach can reach near state-of-the-art performance on various high-dimensional and/or continuous deep RL benchmarks. States are mapped to hash codes, which allows to count their occurrences with a hash table. These counts are then used to compute a reward bonus according to the classic count-based exploration theory. We find that simple hash functions can achieve surprisingly good results on many challenging tasks. Furthermore, we show that a domain-dependent learned hash code may further improve these results. Detailed analysis reveals important aspects of a good hash function: 1) having appropriate granularity and 2) encoding information relevant to solving the MDP. This exploration strategy achieves near state-of-the-art performance on both continuous control tasks and Atari 2600 games, hence providing a simple yet powerful baseline for solving MDPs that require considerable exploration."
                },
                {
                    "title": "Linear Thompson Sampling Revisited",
                    "abstract": "We derive an alternative proof for the regret of Thompson sampling (\\ts) in the stochastic linear bandit setting. While we obtain a regret bound of order $\\wt{O}(d^{3/2}\\sqrt{T})$ as in previous results, the proof sheds new light on the functioning of the \\ts. We leverage on the structure of the problem to show how the regret is related to the sensitivity (i.e., the gradient) of the objective function and how selecting optimal arms associated to \\textit{optimistic} parameters does control it. Thus we show that \\ts can be seen as a generic randomized algorithm where the sampling distribution is designed to have a fixed probability of being optimistic, at the cost of an additional $\\sqrt{d}$ regret factor compared to a UCB-like approach. Furthermore, we show that our proof can be readily applied to regularized linear optimization and generalized linear model problems."
                },
                {
                    "title": "Why is Posterior Sampling Better than Optimism for Reinforcement Learning?",
                    "abstract": "Computational results demonstrate that posterior sampling for reinforcement learning (PSRL) dramatically outperforms algorithms driven by optimism, such as UCRL2. We provide insight into the extent of this performance boost and the phenomenon that drives it. We leverage this insight to establish an $\\tilde{O}(H\\sqrt{SAT})$ Bayesian expected regret bound for PSRL in finite-horizon episodic Markov decision processes, where $H$ is the horizon, $S$ is the number of states, $A$ is the number of actions and $T$ is the time elapsed. This improves upon the best previous bound of $\\tilde{O}(H S \\sqrt{AT})$ for any reinforcement learning algorithm."
                },
                {
                    "title": "Unifying Count-Based Exploration and Intrinsic Motivation",
                    "abstract": "We consider an agent's uncertainty about its environment and the problem of generalizing this uncertainty across observations. Specifically, we focus on the problem of exploration in non-tabular reinforcement learning. Drawing inspiration from the intrinsic motivation literature, we use density models to measure uncertainty, and propose a novel algorithm for deriving a pseudo-count from an arbitrary density model. This technique enables us to generalize count-based exploration algorithms to the non-tabular case. We apply our ideas to Atari 2600 games, providing sensible pseudo-counts from raw pixels. We transform these pseudo-counts into intrinsic rewards and obtain significantly improved exploration in a number of hard games, including the infamously difficult Montezuma's Revenge."
                },
                {
                    "title": "Deep Exploration via Bootstrapped DQN",
                    "abstract": "Efficient exploration in complex environments remains a major challenge for reinforcement learning. We propose bootstrapped DQN, a simple algorithm that explores in a computationally and statistically efficient manner through use of randomized value functions. Unlike dithering strategies such as epsilon-greedy exploration, bootstrapped DQN carries out temporally-extended (or deep) exploration; this can lead to exponentially faster learning. We demonstrate these benefits in complex stochastic MDPs and in the large-scale Arcade Learning Environment. Bootstrapped DQN substantially improves learning times and performance across most Atari games."
                },
                {
                    "title": "Preconditioned Stochastic Gradient Langevin Dynamics for Deep Neural Networks",
                    "abstract": "\n \n Effective training of deep neural networks suffers from two main issues. The first is that the parameter space of these models exhibit pathological curvature. Recent methods address this problem by using adaptive preconditioning for Stochastic Gradient Descent (SGD). These methods improve convergence by adapting to the local geometry of parameter space. A second issue is overfitting, which is typically addressed by early stopping. However, recent work has demonstrated that Bayesian model averaging mitigates this problem. The posterior can be sampled by using Stochastic Gradient Langevin Dynamics (SGLD). However, the rapidly changing curvature renders default SGLD methods inefficient. Here, we propose combining adaptive preconditioners with SGLD. In support of this idea, we give theoretical properties on asymptotic convergence and predictive risk. We also provide empirical results for Logistic Regression, Feedforward Neural Nets, and Convolutional Neural Nets, demonstrating that our preconditioned SGLD method gives state-of-the-art performance on these models.\n \n"
                },
                {
                    "title": "On the Convergence of Stochastic Gradient MCMC Algorithms with High-Order Integrators",
                    "abstract": "Recent advances in Bayesian learning with large-scale data have witnessed emergence of stochastic gradient MCMC algorithms (SG-MCMC), such as stochastic gradient Langevin dynamics (SGLD), stochastic gradient Hamiltonian MCMC (SGHMC), and the stochastic gradient thermostat. While finite-time convergence properties of the SGLD with a 1st-order Euler integrator have recently been studied, corresponding theory for general SG-MCMCs has not been explored. In this paper we consider general SG-MCMCs with high-order integrators, and develop theory to analyze finite-time convergence properties and their asymptotic invariant measures. Our theoretical results show faster convergence rates and more accurate invariant measures for SG-MCMCs with higher-order integrators. For example, with the proposed efficient 2nd-order symmetric splitting integrator, the mean square error (MSE) of the posterior average for the SGHMC achieves an optimal convergence rate of L-4/5 at L iterations, compared to L-2/3 for the SGHMC and SGLD with 1st-order Euler integrators. Furthermore, convergence results of decreasing-step-size SG-MCMCs are also developed, with the same convergence rates as their fixed-step-size counterparts for a specific decreasing sequence. Experiments on both synthetic and real datasets verify our theory, and show advantages of the proposed method in two large-scale real applications."
                },
                {
                    "title": "Prioritized Experience Replay",
                    "abstract": "Experience replay lets online reinforcement learning agents remember and reuse experiences from the past. In prior work, experience transitions were uniformly sampled from a replay memory. However, this approach simply replays transitions at the same frequency that they were originally experienced, regardless of their significance. In this paper we develop a framework for prioritizing experience, so as to replay important transitions more frequently, and therefore learn more efficiently. We use prioritized experience replay in Deep Q-Networks (DQN), a reinforcement learning algorithm that achieved human-level performance across many Atari games. DQN with prioritized experience replay achieves a new state-of-the-art, outperforming DQN with uniform replay on 41 out of 49 games."
                },
                {
                    "title": "Deep Reinforcement Learning with Double Q-Learning",
                    "abstract": "\n \n The popular Q-learning algorithm is known to overestimate action values under certain conditions. It was not previously known whether, in practice, such overestimations are common, whether they harm performance, and whether they can generally be prevented. In this paper, we answer all these questions affirmatively. In particular, we first show that the recent DQN algorithm, which combines Q-learning with a deep neural network, suffers from substantial overestimations in some games in the Atari 2600 domain. We then show that the idea behind the Double Q-learning algorithm, which was introduced in a tabular setting, can be generalized to work with large-scale function approximation. We propose a specific adaptation to the DQN algorithm and show that the resulting algorithm not only reduces the observed overestimations, as hypothesized, but that this also leads to much better performance on several games.\n \n"
                },
                {
                    "title": "Theoretical guarantees for approximate sampling from smooth and log\u2010concave densities",
                    "abstract": "Sampling from various kinds of distribution is an issue of paramount importance in statistics since it is often the key ingredient for constructing estimators, test procedures or confidence intervals. In many situations, exact sampling from a given distribution is impossible or computationally expensive and, therefore, one needs to resort to approximate sampling strategies. However, there is no well\u2010developed theory providing meaningful non\u2010asymptotic guarantees for the approximate sampling procedures, especially in high dimensional problems. The paper makes some progress in this direction by considering the problem of sampling from a distribution having a smooth and log\u2010concave density defined on Rp , for some integer p>0. We establish non\u2010asymptotic bounds for the error of approximating the target distribution by the distribution obtained by the Langevin Monte Carlo method and its variants. We illustrate the effectiveness of the established guarantees with various experiments. Underlying our analysis are insights from the theory of continuous time diffusion processes, which may be of interest beyond the framework of log\u2010concave densities that are considered in the present work."
                },
                {
                    "title": "Adam: A Method for Stochastic Optimization",
                    "abstract": "We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
                },
                {
                    "title": "Consistency and Fluctuations For Stochastic Gradient Langevin Dynamics",
                    "abstract": "Applying standard Markov chain Monte Carlo (MCMC) algorithms to large data sets is computationally expensive. Both the calculation of the acceptance probability and the creation of informed proposals usually require an iteration through the whole data set. The recently proposed stochastic gradient Langevin dynamics (SGLD) method circumvents this problem by generating proposals which are only based on a subset of the data, by skipping the accept-reject step and by using decreasing step-sizes sequence (\u03b4m)m\u22650. \n \nWe provide in this article a rigorous mathematical framework for analysing this algorithm. We prove that, under verifiable assumptions, the algorithm is consistent, satisfies a central limit theorem (CLT) and its asymptotic bias-variance decomposition can be characterized by an explicit functional of the step-sizes sequence (\u03b4m)m\u22650. We leverage this analysis to give practical recommendations for the notoriously difficult tuning of this algorithm: it is asymptotically optimal to use a step-size sequence of the type \u03b4m = m-1/3, leading to an algorithm whose mean squared error (MSE) decreases at rate O(m-1/3)."
                },
                {
                    "title": "Identifying and attacking the saddle point problem in high-dimensional non-convex optimization",
                    "abstract": "A central challenge to many fields of science and engineering involves minimizing non-convex error functions over continuous, high dimensional spaces. Gradient descent or quasi-Newton methods are almost ubiquitously used to perform such minimizations, and it is often thought that a main source of difficulty for these local methods to find the global minimum is the proliferation of local minima with much higher error than the global minimum. Here we argue, based on results from statistical physics, random matrix theory, neural network theory, and empirical evidence, that a deeper and more profound difficulty originates from the proliferation of saddle points, not local minima, especially in high dimensional problems of practical interest. Such saddle points are surrounded by high error plateaus that can dramatically slow down learning, and give the illusory impression of the existence of a local minimum. Motivated by these arguments, we propose a new approach to second-order optimization, the saddle-free Newton method, that can rapidly escape high dimensional saddle points, unlike gradient descent and quasi-Newton methods. We apply this algorithm to deep or recurrent neural network training, and provide numerical evidence for its superior optimization performance."
                },
                {
                    "title": "Generalization and Exploration via Randomized Value Functions",
                    "abstract": "We propose randomized least-squares value iteration (RLSVI) -- a new reinforcement learning algorithm designed to explore and generalize efficiently via linearly parameterized value functions. We explain why versions of least-squares value iteration that use Boltzmann or epsilon-greedy exploration can be highly inefficient, and we present computational results that demonstrate dramatic efficiency gains enjoyed by RLSVI. Further, we establish an upper bound on the expected regret of RLSVI that demonstrates near-optimality in a tabula rasa learning context. More broadly, our results suggest that randomized value functions offer a promising approach to tackling a critical challenge in reinforcement learning: synthesizing efficient exploration and effective generalization."
                },
                {
                    "title": "(More) Efficient Reinforcement Learning via Posterior Sampling",
                    "abstract": "Most provably-efficient reinforcement learning algorithms introduce optimism about poorly-understood states and actions to encourage exploration. We study an alternative approach for efficient exploration: posterior sampling for reinforcement learning (PSRL). This algorithm proceeds in repeated episodes of known duration. At the start of each episode, PSRL updates a prior distribution over Markov decision processes and takes one sample from this posterior. PSRL then follows the policy that is optimal for this sample during the episode. The algorithm is conceptually simple, computationally efficient and allows an agent to encode prior knowledge in a natural way. We establish an O(\u03c4S/\u221aAT) bound on expected regret, where T is time, \u03c4 is the episode length and S and A are the cardinalities of the state and action spaces. This bound is one of the first for an algorithm not based on optimism, and close to the state of the art for any reinforcement learning algorithm. We show through simulation that PSRL significantly outperforms existing algorithms with similar regret bounds."
                },
                {
                    "title": "I. INTRODUCTION",
                    "abstract": "Six years ago, in mid-2006, Iran had no centrifuges spinning and possessed no enriched uranium. Robert Joseph, the US under secretary of state for arms control, declared that year: \u2018we cannot have a single centrifuge spinning in Iran. Iran is a direct threat to the national security of the United States and our allies\u2019. Today, as President Barack Obama prepares to begin his second term, Iran has installed over 9,000 centrifuges, and has produced 6,876 kg of uranium enriched to 3.5 per cent, known as low-enriched uranium (LEU). By the end of 2012, Iran could easily have enriched enough uranium to 20 per cent (still LEU, but described here as medium-enriched uranium, or \u2018MEU\u2019, for clarity) to suffice for one nuclear bomb, if that were enriched further to weapons-grade. Iran instead chose to convert much of its MEU stockpile into reactor fuel, but it is not clear whether it will continue to do so. The quickening pace of Iran\u2019s nuclear activities, and its growing degree of \u2018nuclear latency\u2019 a state\u2019s temporal and technical proximity to acquiring a usable nuclear device has produced a sense of urgency amongst a number of countries, including Arab rivals of Iran. Iran\u2019s dogged pursuit of nuclear technology in the face of unprecedented pressure has also placed new stresses on the Iranian state itself. Indeed, a former senior Iranian diplomat, Seyed Hossein Mousavian, observes in his recently released memoirs that \u2018the nuclear crisis has been the most important challenge facing the Islamic Republic\u2019s foreign policy apparatus since the 1980 88 war between Iran and Iraq\u2019."
                },
                {
                    "title": "Thompson Sampling for Contextual Bandits with Linear Payoffs",
                    "abstract": "Thompson Sampling is one of the oldest heuristics for multi-armed bandit problems. It is a randomized algorithm based on Bayesian ideas, and has recently generated significant interest after several studies demonstrated it to have better empirical performance compared to the state-of-the-art methods. However, many questions regarding its theoretical performance remained open. In this paper, we design and analyze a generalization of Thompson Sampling algorithm for the stochastic contextual multi-armed bandit problem with linear payoff functions, when the contexts are provided by an adaptive adversary. This is among the most important and widely studied version of the contextual bandits problem. We prove a high probability regret bound of O(d2/e\u221aT1+e) in time T for any 0 < e < 1, where d is the dimension of each context vector and e is a parameter used by the algorithm. Our results provide the first theoretical guarantees for the contextual version of Thompson Sampling, and are close to the lower bound of \u03a9(d\u221aT) for this problem. This essentially solves a COLT open problem of Chapelle and Li [COLT 2012]."
                },
                {
                    "title": "The Arcade Learning Environment: An Evaluation Platform for General Agents",
                    "abstract": "In this article we introduce the Arcade Learning Environment (ALE): both a challenge problem and a platform and methodology for evaluating the development of general, domain-independent AI technology. ALE provides an interface to hundreds of Atari 2600 game environments, each one different, interesting, and designed to be a challenge for human players. ALE presents significant research challenges for reinforcement learning, model learning, model-based planning, imitation learning, transfer learning, and intrinsic motivation. Most importantly, it provides a rigorous testbed for evaluating and comparing approaches to these problems. We illustrate the promise of ALE by developing and benchmarking domain-independent agents designed using well-established AI techniques for both reinforcement learning and planning. In doing so, we also propose an evaluation methodology made possible by ALE, reporting empirical results on over 55 different games. All of the software, including the benchmark agents, is publicly available."
                },
                {
                    "title": "MCMC using Hamiltonian dynamics",
                    "abstract": "Hamiltonian dynamics can be used to produce distant proposals for the Metropolis algorithm, thereby avoiding the slow exploration of the state space that results from the diffusive behaviour of simple random-walk proposals. Though originating in physics, Hamiltonian dynamics can be applied to most problems with continuous state spaces by simply introducing fictitious\"momentum\"variables. A key to its usefulness is that Hamiltonian dynamics preserves volume, and its trajectories can thus be used to define complex mappings without the need to account for a hard-to-compute Jacobian factor - a property that can be exactly maintained even when the dynamics is approximated by discretizing time. In this review, I discuss theoretical and practical aspects of Hamiltonian Monte Carlo, and present some of its variations, including using windows of states for deciding on acceptance or rejection, computing trajectories using fast approximations, tempering during the course of a trajectory to handle isolated modes, and short-cut methods that prevent useless trajectories from taking much computation time."
                },
                {
                    "title": "Improved Algorithms for Linear Stochastic Bandits",
                    "abstract": "We improve the theoretical analysis and empirical performance of algorithms for the stochastic multi-armed bandit problem and the linear stochastic multi-armed bandit problem. In particular, we show that a simple modification of Auer's UCB algorithm (Auer, 2002) achieves with high probability constant regret. More importantly, we modify and, consequently, improve the analysis of the algorithm for the for linear stochastic bandit problem studied by Auer (2002), Dani et al. (2008), Rusmevichientong and Tsitsiklis (2010), Li et al. (2010). Our modification improves the regret bound by a logarithmic factor, though experiments show a vast improvement. In both cases, the improvement stems from the construction of smaller confidence sets. For their construction we use a novel tail inequality for vector-valued martingales."
                },
                {
                    "title": "Bayesian Learning via Stochastic Gradient Langevin Dynamics",
                    "abstract": "In this paper we propose a new framework for learning from large scale datasets based on iterative learning from small mini-batches. By adding the right amount of noise to a standard stochastic gradient optimization algorithm we show that the iterates will converge to samples from the true posterior distribution as we anneal the stepsize. This seamless transition between optimization and Bayesian posterior sampling provides an inbuilt protection against overfitting. We also propose a practical method for Monte Carlo estimates of posterior statistics which monitors a \"sampling threshold\" and collects samples after it has been surpassed. We apply the method to three models: a mixture of Gaussians, logistic regression and ICA with natural gradients."
                },
                {
                    "title": "Double Q-learning",
                    "abstract": "In some stochastic environments the well-known reinforcement learning algorithm Q-learning performs very poorly. This poor performance is caused by large overestimations of action values. These overestimations result from a positive bias that is introduced because Q-learning uses the maximum action value as an approximation for the maximum expected action value. We introduce an alternative way to approximate the maximum expected value for any set of random variables. The obtained double estimator method is shown to sometimes underestimate rather than overestimate the maximum expected value. We apply the double estimator to Q-learning to construct Double Q-learning, a new off-policy reinforcement learning algorithm. We show the new algorithm converges to the optimal policy and that it performs well in some settings in which Q-learning performs poorly due to its overestimation."
                },
                {
                    "title": "\u5f15\u8a00 (Introduction)",
                    "abstract": "\u672c\u671f\u4e13\u8f91\u6536\u5f55\u7684\u6587\u7ae0\u4e8e2007\u5e749\u6708\u5728\u9752\u5c9b\u4e3e\u884c\u7684\u9898\u4e3a\u201c\u6b96\u6c11\u4e3b\u4e49\u4e0e\u4e2d\u56fd\u8fd1\u4ee3\u793e\u4f1a\u201d\u7684\u5b66\u672f\u7814\u8ba8\u4f1a\u4e0a\u521d\u6b21\u9762\u4e16\u3002\u8fd9\u6b21\u56fd\u9645\u7814\u8ba8\u4f1a\u7531\u6089\u5c3c\u79d1\u6280\u5927\u5b66\u548c\u9752\u5c9b\u6d77\u6d0b\u5927\u5b66\u5171\u540c\u4e3e\u529e\uff0c\u4e0e\u4f1a\u8005\u6765\u81ea\u6fb3\u5927\u5229\u4e9a\u3001\u7f8e\u56fd\u3001\u82f1\u56fd\u3001\u6cd5\u56fd\u3001\u5fb7\u56fd\u3001\u610f\u5927\u5229\u3001\u65e5\u672c\u3001\u65b0\u52a0\u5761\u3001\u4e2d\u56fd\u7b49\u4f17\u591a\u56fd\u5bb6\uff0c\u8bba\u6587\u4f5c\u8005\u5f53\u4e2d\u65e2\u6709\u4eab\u8a89\u4e16\u754c\u7684\u4e13\u5bb6\u4e5f\u6709\u6b63\u5728\u653b\u8bfb\u535a\u58eb\u7684\u9752\u5e74\u5b66\u8005\u3002\u6b63\u5982\u4f1a\u8bae\u4ee3\u8868\u4e30\u5bcc\u591a\u59ff\u7684\u4e0d\u540c\u80cc\u666f\uff0c\u4f1a\u8bae\u8bba\u6587\u4ee5\u201c\u6b96\u6c11\u4e3b\u4e49\u4e0e\u4e2d\u56fd\u8fd1\u4ee3\u793e\u4f1a\u201d\u4e3a\u4e3b\u9898\uff0c\u4ece\u5404\u79cd\u4e0d\u540c\u7684\u89d2\u5ea6\uff0c\u8fd0\u7528\u4e0d\u540c\u7684\u65b9\u6cd5\u63a2\u8ba8\u4e86\u4e2d\u56fd\u81ea1840\u5e74\u9e26\u7247\u6218\u4e89\u5230\u4e8c\u5341\u4e16\u7eaa\u540e\u671f\u9999\u6e2f\u6fb3\u95e8\u56de\u5f52\u957f\u8fbe\u4e00\u4e2a\u591a\u4e16\u7eaa\u91cc\u4e0e\u897f\u65b9\u5217\u5f3a\u7684\u906d\u9047\u3002"
                },
                {
                    "title": "Near-optimal Regret Bounds for Reinforcement Learning",
                    "abstract": "For undiscounted reinforcement learning in Markov decision processes (MDPs) we consider the total regret of a learning algorithm with respect to an optimal policy. In order to describe the transition structure of an MDP we propose a new parameter: An MDP has diameter D if for any pair of states s,s' there is a policy which moves from s to s' in at most D steps (on average). We present a reinforcement learning algorithm with total regret O(DS\u221aAT) after T steps for any unknown MDP with S states, A actions per state, and diameter D. A corresponding lower bound of \u03a9(\u221aDSAT) on the total regret of any learning algorithm is given as well. \n \nThese results are complemented by a sample complexity bound on the number of suboptimal steps taken by our algorithm. This bound can be used to achieve a (gap-dependent) regret bound that is logarithmic in T. \n \nFinally, we also consider a setting where the MDP is allowed to change a fixed number of l times. We present a modification of our algorithm that is able to deal with this setting and show a regret bound of O(l1/3T2/3DS\u221aA)."
                },
                {
                    "title": "A Bayesian Framework for Reinforcement Learning",
                    "abstract": "The reinforcement learning problem can be decomposed into two parallel types of inference: (i) estimating the parameters of a model for the underlying process; (ii) determining behavior which maximizes return under the estimated model. Following Dearden, Friedman and Andre (1999), it is proposed that the learning process estimates online the full posterior distribution over models. To determine behavior, a hypothesis is sampled from this distribution and the greedy policy with respect to the hypothesis is obtained by dynamic programming. By using a different hypothesis for each trial appropriate exploratory and exploitative behavior is obtained. This Bayesian method always converges to the optimal policy for a stationary process with discrete states."
                },
                {
                    "title": "Exponential convergence of Langevin distributions and their discrete approximations",
                    "abstract": "In this paper we consider a continuous-time method of approximating a given distribution using the Langevin di\u0080usion dLt\u0088dWt\u0087 1 2 r log (Lt)dt. We \u00aend conditions under this di\u0080usion converges exponentially quickly to or does not: in one dimension, these are essentially that for distributions with exponential tails of the form (x)/ exp (y |x| , 0< <1, exponential convergence occurs if and only if 1. We then consider conditions under which the discrete approximations to the di\u0080usion converge. We \u00aerst show that even when the di\u0080usion itself converges, naive discretizations need not do so. We then consider a `Metropolis-adjusted' version of the algorithm, and \u00aend conditions under which this also converges at an exponential rate: perhaps surprisingly, even the Metropolized version need not converge exponentially fast even if the di\u0080usion does. We brie y discuss a truncated form of the algorithm which, in practice, should avoid the di\u0081culties of the other forms."
                },
                {
                    "title": "Matrix analysis",
                    "abstract": "We next study a special type of similarity that is intimately involved with many aspects of the application of matrix analysis. Introduction For a general nonsingular matrix S \u2208 M n , we made an initial study of similarity via S in Chapter 1. For certain very special nonsingular matrices, called unitary matrices, the inverse of S has a simple form: S \u22121 = S *. Similarity of A \u2208 M n via a unitary matrix, A \u2192 S * AS , is not only conceptually simpler ( S * is much easier to evaluate than S \u22121 ) than general similarity, but it has a number of attractive features that will become clearer through the development to follow. As a general rule, unitary similarities are preferable to general similarities, and it is therefore useful to know what can be achieved through unitary similarity. Equivalence classes under unitary similarity are, however, finer than under general similarity (two matrices can be similar but not unitarily similar), and correspondingly less can be achieved. For this reason, we shall return to study general similarity further in Chapter 3. The transformation A \u2192 S * AS , A \u2208 M n , in which S is assumed to be nonsingular but not necessarily unitary, is called * congruence and will be studied in Chapter 4. This transformation, too, is an equivalence relation on M n with a number of attractive features (different from those of similarity)."
                },
                {
                    "title": "HyperDQN: A Randomized Exploration Method for Deep Reinforcement Learning",
                    "abstract": "Randomized least-square value iteration (RLSVI) is a provably efficient exploration method. However, it is limited to the case where (1) a good feature is known in advance and (2) this feature is fixed during the training. If otherwise, RLSVI suffers an unbearable computational burden to obtain the posterior samples. In this work, we present a practical algorithm named HyperDQN to address the above issues under deep RL. In addition to a non-linear neural network (i.e., base model) that predicts Q-values, our method employs a probabilistic hypermodel (i.e., meta model), which outputs the parameter of the base model. When both models are jointly optimized under a specifically designed objective, three purposes can be achieved. First, the hypermodel can generate approximate posterior samples regarding the parameter of the Q-value function. As a result, diverse Q-value functions are sampled to select exploratory action sequences. This retains the punchline of RLSVI for efficient exploration. Second, a good feature is learned to approximate Q-value functions. This addresses limitation (1). Third, the posterior samples of the Q-value function can be obtained in a more efficient way than the existing methods, and the changing feature does not affect the efficiency. This deals with limitation (2). On the Atari suite, HyperDQN with 20M frames outperforms DQN with 200M frames in terms of the maximum human-normalized score. For SuperMarioBros, HyperDQN outperforms several exploration bonus and randomized exploration methods on 5 out of 9 games."
                },
                {
                    "title": "On Approximate Thompson Sampling with Langevin Algorithms",
                    "abstract": "Thompson sampling for multi-armed bandit problems is known to enjoy favorable performance in both theory and practice. However, its wider deployment is restricted due to a signi\ufb01cant computational limitation: the need for samples from posterior distributions at every iteration. In practice, this limitation is alleviated by making use of approximate sampling methods, yet provably incorporating approximate samples into Thompson Sampling algorithms remains an open problem. In this work we address this by proposing two ef\ufb01cient Langevin MCMC algorithms tailored to Thompson sampling. The resulting approximate Thompson Sampling algorithms are ef\ufb01ciently implementable and provably achieve optimal instance-dependent regret for the Multi-Armed Bandit (MAB) problem. To prove these results we derive novel posterior concentration bounds and MCMC convergence rates for log-concave distributions which may be of independent interest."
                },
                {
                    "title": "Subsampled Stochastic Variance-Reduced Gradient Langevin Dynamics",
                    "abstract": "Stochastic variance-reduced gradient Langevin dynamics (SVRG-LD) was recently proposed to improve the performance of stochastic gradient Langevin dynamics (SGLD) by reducing the variance of the stochastic gradient. In this paper, we propose a variant of SVRG-LD, namely SVRG-LD, which replaces the full gradient in each epoch with a subsampled one. We provide a nonasymptotic analysis of the convergence of SVRG-LD in 2-Wasserstein distance, and show that SVRG-LD enjoys a lower gradient complexity1 than SVRG-LD, when the sample size is large or the target accuracy requirement is moderate. Our analysis directly implies a sharper convergence rate for SVRG-LD, which improves the existing convergence rate by a factor of \u03ban, where \u03ba is the condition number of the log-density function and n is the sample size. Experiments on both synthetic and real-world datasets validate our theoretical results."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can Langevin Monte Carlo (LMC) be effectively integrated into reinforcement learning to improve exploration strategies in deep Q-networks?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for advancing the field of reinforcement learning, as effective exploration strategies are essential for training agents in complex environments. By integrating LMC into deep Q-networks, we can enhance the agent's ability to discover optimal policies, leading to improved performance in various applications such as robotics, game playing, and autonomous systems. This research could pave the way for future studies that explore novel exploration techniques, ultimately contributing to the development of more robust and efficient reinforcement learning algorithms.\n\n### [Question 3] - Why is it hard?\nThe integration of LMC into reinforcement learning presents several challenges. First, the theoretical underpinnings of LMC must be adapted to the dynamic and often non-stationary environments encountered in reinforcement learning. Naive approaches may fail due to the complexity of balancing exploration and exploitation, as well as the need for efficient sampling methods that do not compromise the learning process. Additionally, practical obstacles include the computational cost of implementing LMC in high-dimensional state spaces and ensuring stability during training.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on traditional exploration strategies, often overlooking the potential of LMC in reinforcement learning contexts. Limitations in computational resources and the complexity of integrating LMC with existing algorithms have hindered progress. Additionally, prior work may not have fully explored the theoretical implications of LMC in reinforcement learning, leaving a gap in understanding how to leverage its strengths effectively. Our approach aims to bridge this gap by providing a comprehensive theoretical analysis and empirical validation of LMC's application in deep Q-networks.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves the following key components: \n1. **Method**: We will develop a deep Q-network that incorporates Langevin Monte Carlo for exploration, termed LMC-DQN.\n2. **Dataset**: We will evaluate our approach using standard reinforcement learning environments, including randomly generated linear MDPs, the RiverSwim environment, and Atari games.\n3. **Metric**: Performance will be measured using cumulative reward and convergence speed as primary metrics.\nExpected outcomes include demonstrating that LMC-DQN outperforms traditional exploration strategies in terms of learning efficiency and policy quality, thereby validating the effectiveness of LMC in"
            }
        },
        "author_data": {
            "f73577df-b53c-4ea3-8bb4-968de1d1593a": {
                "pk": "f73577df-b53c-4ea3-8bb4-968de1d1593a",
                "name": "Haque Ishfaq",
                "collaborators": [
                    "D. Precup",
                    "Zhuoran Yang",
                    "Zhaoran Wang",
                    "Doina Precup",
                    "Kshitij P. Fadnis",
                    "Kartik Talamadupula",
                    "Pavan Kapanipathi",
                    "S. Roukos",
                    "Achille Fokoue",
                    "A. Hoogi"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Deep Learning",
                    "Metric Learning",
                    "Knowledge Graph"
                ],
                "publications": [
                    {
                        "title": "More Efficient Randomized Exploration for Reinforcement Learning via Approximate Sampling",
                        "abstract": "Thompson sampling (TS) is one of the most popular exploration techniques in reinforcement learning (RL). However, most TS algorithms with theoretical guarantees are difficult to implement and not generalizable to Deep RL. While the emerging approximate sampling-based exploration schemes are promising, most existing algorithms are specific to linear Markov Decision Processes (MDP) with suboptimal regret bounds, or only use the most basic samplers such as Langevin Monte Carlo. In this work, we propose an algorithmic framework that incorporates different approximate sampling methods with the recently proposed Feel-Good Thompson Sampling (FGTS) approach (Zhang, 2022; Dann et al., 2021), which was previously known to be computationally intractable in general. When applied to linear MDPs, our regret analysis yields the best known dependency of regret on dimensionality, surpassing existing randomized algorithms. Additionally, we provide explicit sampling complexity for each employed sampler. Empirically, we show that in tasks where deep exploration is necessary, our proposed algorithms that combine FGTS and approximate sampling perform significantly better compared to other strong baselines. On several challenging games from the Atari 57 suite, our algorithms achieve performance that is either better than or on par with other strong baselines from the deep RL literature."
                    },
                    {
                        "title": "Offline Multitask Representation Learning for Reinforcement Learning",
                        "abstract": "We study offline multitask representation learning in reinforcement learning (RL), where a learner is provided with an offline dataset from different tasks that share a common representation and is asked to learn the shared representation. We theoretically investigate offline multitask low-rank RL, and propose a new algorithm called MORL for offline multitask representation learning. Furthermore, we examine downstream RL in reward-free, offline and online scenarios, where a new task is introduced to the agent that shares the same representation as the upstream offline tasks. Our theoretical results demonstrate the benefits of using the learned representation from the upstream offline task instead of directly learning the representation of the low-rank model."
                    },
                    {
                        "title": "Randomized Exploration for Reinforcement Learning with General Value Function Approximation",
                        "abstract": "We propose a model-free reinforcement learning algorithm inspired by the popular randomized least squares value iteration (RLSVI) algorithm as well as the optimism principle. Unlike existing upper-confidence-bound (UCB) based approaches, which are often computationally intractable, our algorithm drives exploration by simply perturbing the training data with judiciously chosen i.i.d. scalar noises. To attain optimistic value function estimation without resorting to a UCB-style bonus, we introduce an optimistic reward sampling procedure. When the value functions can be represented by a function class $\\mathcal{F}$, our algorithm achieves a worst-case regret bound of $\\widetilde{O}(\\mathrm{poly}(d_EH)\\sqrt{T})$ where $T$ is the time elapsed, $H$ is the planning horizon and $d_E$ is the $\\textit{eluder dimension}$ of $\\mathcal{F}$. In the linear setting, our algorithm reduces to LSVI-PHE, a variant of RLSVI, that enjoys an $\\widetilde{\\mathcal{O}}(\\sqrt{d^3H^3T})$ regret. We complement the theory with an empirical evaluation across known difficult exploration tasks."
                    },
                    {
                        "title": "Randomized Least Squares Policy Optimization",
                        "abstract": "Policy Optimization (PO) methods with function approximation are one of the most popular classes of Reinforcement Learning (RL) algorithms. However, designing provably ef\ufb01cient policy optimization algorithms remains a challenge. Recent work in this area has focused on incorporating upper con\ufb01dence bound (UCB)-style bonuses to drive exploration in policy optimization. In this paper, we present Randomized Least Squares Policy Optimization (RLSPO) which is inspired by Thompson Sampling. We prove that, in an episodic linear kernel MDP setting, RLSPO achieves (cid:101) O ( d 3 / 2 H 3 / 2 \u221a T ) worst-case (frequentist) regret, where H is the number of episodes, T is the total number of steps and d is the feature dimension. Finally, we evaluate RLSPO empirically and show that it is competitive with existing provably ef\ufb01cient PO algorithms."
                    },
                    {
                        "title": "Path-Based Contextualization of Knowledge Graphs for Textual Entailment.",
                        "abstract": "In this paper, we introduce the problem of knowledge graph contextualization -- that is, given a specific NLP task, the problem of extracting meaningful and relevant sub-graphs from a given knowledge graph. The task in the case of this paper is the textual entailment problem, and the context is a relevant sub-graph for an instance of the textual entailment problem -- where given two sentences p and h, the entailment relationship between them has to be predicted automatically. We base our methodology on finding paths in a cost-customized external knowledge graph, and building the most relevant sub-graph that connects p and h. We show that our path selection mechanism to generate sub-graphs not only reduces noise, but also retrieves meaningful information from large knowledge graphs. Our evaluation shows that using information on entities as well as the relationships between them improves on the performance of purely text-based systems."
                    },
                    {
                        "title": "Heuristics for Interpretable Knowledge Graph Contextualization",
                        "abstract": "In this paper, we introduce the problem of knowledge graph contextualization \u2013 that is, given a specific context, the problem of extracting the most relevant sub-graph of a given knowledge graph. The context in the case of this paper is defined to be the textual entailment problem, and more specifically an instance of that problem where the entailment relationship between two sentences P and H has to be predicted automatically. This prediction takes the form of a classification task, and we seek to provide that task with the most relevant external knowledge while eliminating as much noise as possible. We base our methodology on finding the shortest paths in the cost-customized external knowledge graph that connect P and H, and build a series of methods \u2013 starting with manually curated search heuristics and culminating in automatically extracted heuristics \u2013 to find such paths and build the most relevant sub-graph. We evaluate our approaches by measuring the accuracy of the classification on the textual entailment problem, and show that modulating the external knowledge that is used has an impact on performance."
                    },
                    {
                        "title": "TVAE: Triplet-Based Variational Autoencoder using Metric Learning",
                        "abstract": "Deep metric learning has been demonstrated to be highly effective in learning semantic representation and encoding information that can be used to measure data similarity, by relying on the embedding learned from metric learning. At the same time, variational autoencoder (VAE) has widely been used to approximate inference and proved to have a good performance for directed probabilistic models. However, for traditional VAE, the data label or feature information are intractable. Similarly, traditional representation learning approaches fail to represent many salient aspects of the data. In this project, we propose a novel integrated framework to learn latent embedding in VAE by incorporating deep metric learning. The features are learned by optimizing a triplet loss on the mean vectors of VAE in conjunction with standard evidence lower bound (ELBO) of VAE. This approach, which we call Triplet based Variational Autoencoder (TVAE), allows us to capture more fine-grained information in the latent embedding. Our model is tested on MNIST data set and achieves a high triplet accuracy of 95.60% while the traditional VAE (Kingma & Welling, 2013) achieves triplet accuracy of 75.08%."
                    },
                    {
                        "title": "TVAE: TRIPLET-BASED VARIATIONAL AUTOEN-",
                        "abstract": "Deep metric learning has been demonstrated to be highly effective in learning semantic representation and encoding information that can be used to measure data similarity, by relying on the embedding learned from metric learning. At the same time, variational autoencoder (VAE) has widely been used to approximate inference and proved to have a good performance for directed probabilistic models. However, for traditional VAE, the data label or feature information are intractable. Similarly, traditional representation learning approaches fail to represent many salient aspects of the data. In this project, we propose a novel integrated framework to learn latent embedding in VAE by incorporating deep metric learning. The features are learned by optimizing a triplet loss on the mean vectors of VAE in conjunction with standard evidence lower bound (ELBO) of VAE. This approach, which we call Triplet based Variational Autoencoder (TVAE), allows us to capture more fine-grained information in the latent embedding. Our model is tested on MNIST data set and achieves a high triplet accuracy of 95.60% while the traditional VAE (Kingma & Welling, 2013) achieves triplet accuracy of 75.08%."
                    },
                    {
                        "title": "TVAE : Deep Metric Learning Approach for Variational Autoencoder",
                        "abstract": "Deep metric learning has been demonstrated to be highly effective in learning semantic representation and encoding data information. People are able to do similarity measurement for data, based on the embedding learned from metric learning. At the same time, variational autoencoder (VAE) has widely been used to approximate inference and proved to have a good performance for directed probabilistic models. However, for traditional VAE, the data label or feature information are intractable. Similarly, traditional representation learning approaches fail to represent many salient aspects of the data. To this end, in this project, we propose a novel structure to learn latent embedding in VAE by incorporating deep metric learning. The features are learned by a triplet loss on the mean vectors of VAE in conjunction with reconstruction loss of VAE. This approach, which we call Triplet based Variational Autoencoder (TVAE), allows us to capture more fine-grained information in the embedding. Our model is first tested on MNIST data set. A high triplet accuracy of around 95.60% is achieved while the VAE is found to perform well at the same time. We further implement our structure on Zappos50k shoe dataset [32] to show the efficacy of our method."
                    }
                ]
            },
            "b90ee355-4391-4f5e-9354-b17d96257656": {
                "pk": "b90ee355-4391-4f5e-9354-b17d96257656",
                "name": "Qingfeng Lan",
                "collaborators": [
                    "A. R. Mahmood",
                    "Yangchen Pan",
                    "Rupam Mahmood",
                    "Jun Luo",
                    "A. Mahmood",
                    "Martha White",
                    "Alona Fyshe",
                    "Haque Ishfaq",
                    "Yixin Tan",
                    "Yu Yang"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Continual Learning",
                    "Neural Networks",
                    "Quantum Computing"
                ],
                "publications": [
                    {
                        "title": "More Efficient Randomized Exploration for Reinforcement Learning via Approximate Sampling",
                        "abstract": "Thompson sampling (TS) is one of the most popular exploration techniques in reinforcement learning (RL). However, most TS algorithms with theoretical guarantees are difficult to implement and not generalizable to Deep RL. While the emerging approximate sampling-based exploration schemes are promising, most existing algorithms are specific to linear Markov Decision Processes (MDP) with suboptimal regret bounds, or only use the most basic samplers such as Langevin Monte Carlo. In this work, we propose an algorithmic framework that incorporates different approximate sampling methods with the recently proposed Feel-Good Thompson Sampling (FGTS) approach (Zhang, 2022; Dann et al., 2021), which was previously known to be computationally intractable in general. When applied to linear MDPs, our regret analysis yields the best known dependency of regret on dimensionality, surpassing existing randomized algorithms. Additionally, we provide explicit sampling complexity for each employed sampler. Empirically, we show that in tasks where deep exploration is necessary, our proposed algorithms that combine FGTS and approximate sampling perform significantly better compared to other strong baselines. On several challenging games from the Atari 57 suite, our algorithms achieve performance that is either better than or on par with other strong baselines from the deep RL literature."
                    },
                    {
                        "title": "Weight Clipping for Deep Continual and Reinforcement Learning",
                        "abstract": "Many failures in deep continual and reinforcement learning are associated with increasing magnitudes of the weights, making them hard to change and potentially causing overfitting. While many methods address these learning failures, they often change the optimizer or the architecture, a complexity that hinders widespread adoption in various systems. In this paper, we focus on learning failures that are associated with increasing weight norm and we propose a simple technique that can be easily added on top of existing learning systems: clipping neural network weights to limit them to a specific range. We study the effectiveness of weight clipping in a series of supervised and reinforcement learning experiments. Our empirical results highlight the benefits of weight clipping for generalization, addressing loss of plasticity and policy collapse, and facilitating learning with a large replay ratio."
                    },
                    {
                        "title": "Elephant Neural Networks: Born to Be a Continual Learner",
                        "abstract": "Catastrophic forgetting remains a significant challenge to continual learning for decades. While recent works have proposed effective methods to mitigate this problem, they mainly focus on the algorithmic side. Meanwhile, we do not fully understand what architectural properties of neural networks lead to catastrophic forgetting. This study aims to fill this gap by studying the role of activation functions in the training dynamics of neural networks and their impact on catastrophic forgetting. Our study reveals that, besides sparse representations, the gradient sparsity of activation functions also plays an important role in reducing forgetting. Based on this insight, we propose a new class of activation functions, elephant activation functions, that can generate both sparse representations and sparse gradients. We show that by simply replacing classical activation functions with elephant activation functions, we can significantly improve the resilience of neural networks to catastrophic forgetting. Our method has broad applicability and benefits for continual learning in regression, class incremental learning, and reinforcement learning tasks. Specifically, we achieves excellent performance on Split MNIST dataset in just one single pass, without using replay buffer, task boundary information, or pre-training."
                    },
                    {
                        "title": "Learning to Optimize for Reinforcement Learning",
                        "abstract": "In recent years, by leveraging more data, computation, and diverse tasks, learned optimizers have achieved remarkable success in supervised learning, outperforming classical hand-designed optimizers. Reinforcement learning (RL) is essentially different from supervised learning, and in practice, these learned optimizers do not work well even in simple RL tasks. We investigate this phenomenon and identify two issues. First, the agent-gradient distribution is non-independent and identically distributed, leading to inefficient meta-training. Moreover, due to highly stochastic agent-environment interactions, the agent-gradients have high bias and variance, which increases the difficulty of learning an optimizer for RL. We propose pipeline training and a novel optimizer structure with a good inductive bias to address these issues, making it possible to learn an optimizer for reinforcement learning from scratch. We show that, although only trained in toy tasks, our learned optimizer can generalize to unseen complex tasks in Brax."
                    },
                    {
                        "title": "Memory-efficient Reinforcement Learning with Value-based Knowledge Consolidation",
                        "abstract": "Artificial neural networks are promising for general function approximation but challenging to train on non-independent or non-identically distributed data due to catastrophic forgetting. The experience replay buffer, a standard component in deep reinforcement learning, is often used to reduce forgetting and improve sample efficiency by storing experiences in a large buffer and using them for training later. However, a large replay buffer results in a heavy memory burden, especially for onboard and edge devices with limited memory capacities. We propose memory-efficient reinforcement learning algorithms based on the deep Q-network algorithm to alleviate this problem. Our algorithms reduce forgetting and maintain high sample efficiency by consolidating knowledge from the target Q-network to the current Q-network. Compared to baseline methods, our algorithms achieve comparable or better performance in both feature-based and image-based tasks while easing the burden of large experience replay buffers."
                    },
                    {
                        "title": "Memory-efficient Reinforcement Learning with Knowledge Consolidation",
                        "abstract": "Arti\ufb01cial neural networks are promising for general function approximation but challenging to train on non-independent or non-identically distributed data due to catastrophic forgetting. The experience replay buffer, a standard component in deep reinforcement learning, is often used to reduce forgetting and improve sample ef\ufb01ciency by storing experiences in a large buffer and using them for training later. However, a large replay buffer results in a heavy memory burden, especially for onboard and edge devices with limited memory capacities. We propose memory-ef\ufb01cient reinforcement learning algorithms based on the deep Q-network algorithm to alleviate this problem. Our algorithms reduce forgetting and maintain high sample ef\ufb01ciency by consolidating knowledge from the target Q-network to the current Q-network. Compared to baseline methods, our algorithms achieve comparable or better performance in both feature-based and image-based tasks while easing the burden of large experience replay buffers. 1"
                    },
                    {
                        "title": "Predictive Representation Learning for Language Modeling",
                        "abstract": "To effectively perform the task of next-word prediction, long short-term memory networks (LSTMs) must keep track of many types of information. Some information is directly related to the next word's identity, but some is more secondary (e.g. discourse-level features or features of downstream words). Correlates of secondary information appear in LSTM representations even though they are not part of an \\emph{explicitly} supervised prediction task. In contrast, in reinforcement learning (RL), techniques that explicitly supervise representations to predict secondary information have been shown to be beneficial. Inspired by that success, we propose Predictive Representation Learning (PRL), which explicitly constrains LSTMs to encode specific predictions, like those that might need to be learned implicitly. We show that PRL 1) significantly improves two strong language modeling methods, 2) converges more quickly, and 3) performs better when data is limited. Our work shows that explicitly encoding a simple predictive task facilitates the search for a more effective language model."
                    },
                    {
                        "title": "Model-free Policy Learning with Reward Gradients",
                        "abstract": "Despite the increasing popularity of policy gradient methods, they are yet to be widely utilized in sample-scarce applications, such as robotics. The sample efficiency could be improved by making best usage of available information. As a key component in reinforcement learning, the reward function is usually devised carefully to guide the agent. Hence, the reward function is usually known, allowing access to not only scalar reward signals but also reward gradients. To benefit from reward gradients, previous works require the knowledge of environment dynamics, which are hard to obtain. In this work, we develop the \\textit{Reward Policy Gradient} estimator, a novel approach that integrates reward gradients without learning a model. Bypassing the model dynamics allows our estimator to achieve a better bias-variance trade-off, which results in a higher sample efficiency, as shown in the empirical analysis. Our method also boosts the performance of Proximal Policy Optimization on different MuJoCo control tasks."
                    },
                    {
                        "title": "Variational Quantum Soft Actor-Critic",
                        "abstract": "Quantum computing has a superior advantage in tackling specific problems, such as integer factorization and Simon's problem. For more general tasks in machine learning, by applying variational quantum circuits, more and more quantum algorithms have been proposed recently, especially in supervised learning and unsupervised learning. However, little work has been done in reinforcement learning, arguably more important and challenging. Previous work in quantum reinforcement learning mainly focuses on discrete control tasks where the action space is discrete. In this work, we develop a quantum reinforcement learning algorithm based on soft actor-critic -- one of the state-of-the-art methods for continuous control. Specifically, we use a hybrid quantum-classical policy network consisting of a variational quantum circuit and a classical artificial neural network. Tested in a standard reinforcement learning benchmark, we show that this quantum version of soft actor-critic is comparable with the original soft actor-critic, using much less adjustable parameters. Furthermore, we analyze the effect of different hyper-parameters and policy network architectures, pointing out the importance of architecture design for quantum reinforcement learning."
                    },
                    {
                        "title": "Maxmin Q-learning: Controlling the Estimation Bias of Q-learning",
                        "abstract": "Q-learning suffers from overestimation bias, because it approximates the maximum action value using the maximum estimated action value. Algorithms have been proposed to reduce overestimation bias, but we lack an understanding of how bias interacts with performance, and the extent to which existing algorithms mitigate bias. In this paper, we 1) highlight that the effect of overestimation bias on learning efficiency is environment-dependent; 2) propose a generalization of Q-learning, called \\emph{Maxmin Q-learning}, which provides a parameter to flexibly control bias; 3) show theoretically that there exists a parameter choice for Maxmin Q-learning that leads to unbiased estimation with a lower approximation variance than Q-learning; and 4) prove the convergence of our algorithm in the tabular case, as well as convergence of several previous Q-learning variants, using a novel Generalized Q-learning framework. We empirically verify that our algorithm better controls estimation bias in toy environments, and that it achieves superior performance on several benchmark problems."
                    },
                    {
                        "title": "Reducing Selection Bias in Counterfactual Reasoning for Individual Treatment Effects Estimation",
                        "abstract": "Counterfactual reasoning is an important paradigm applicable in many fields, such as healthcare, economics, and education. In this work, we propose a novel method to address the issue of \\textit{selection bias}. We learn two groups of latent random variables, where one group corresponds to variables that only cause selection bias, and the other group is relevant for outcome prediction. They are learned by an auto-encoder where an additional regularized loss based on Pearson Correlation Coefficient (PCC) encourages the de-correlation between the two groups of random variables. This allows for explicitly alleviating selection bias by only keeping the latent variables that are relevant for estimating individual treatment effects. Experimental results on a synthetic toy dataset and a benchmark dataset show that our algorithm is able to achieve state-of-the-art performance and improve the result of its counterpart that does not explicitly model the selection bias."
                    }
                ]
            },
            "9ef09763-da32-41a9-88fe-33e9fbde6846": {
                "pk": "9ef09763-da32-41a9-88fe-33e9fbde6846",
                "name": "Pan Xu",
                "collaborators": [
                    "Haocai Huang",
                    "Guangming Li",
                    "Shijie Xu",
                    "Zhengliang Hu",
                    "Rendong Feng",
                    "Yixin Gao",
                    "Xinyi Xie",
                    "K. Kim",
                    "Xing Sun",
                    "Fengli Zhang"
                ],
                "domain": [
                    "Underwater Acoustics",
                    "Cancer Research",
                    "Network Pharmacology",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Flow current field observation with underwater moving acoustic tomography",
                        "abstract": "Underwater environment observation with underwater acoustic tomography has been considerably developed in recent years. Moving sound transmission can obtain the observation of entire spatial area with sound station moving. Various internal structures, unique surface and submarine boundaries and changing environment constitutes a complex acoustic propagation channel. This paper focus on the inversion method and signal resampling for sound moving transmission. Also, the current field in three-dimensional (3D) scale is also studied. A five-station sound transmission experiment with four moored station and one moving station that conducted in range of 500m\u00d7500m at Huangcai reservoir, Changsha, China is presented. Signal resampling is performed to get correlation of received acoustic data. The vertical layer-averaged flow current results between moving station and moored station are inversed with 2D grid method. 3D flow current field result is composed by grid-averaged inversion current of vertical profile via moving station at different moment. The received results of reciprocal signal transmission between two moored stations and one moving station were used for layer-averaged current variations at vertical scale and grid-averaged current. The feasibility of the method in underwater moving acoustic tomography research is proved and its applicability is discussed. The proposed underwater acoustic tomography technology develops an innovative idea for the further development of temporal- spatial grided tomography observation."
                    },
                    {
                        "title": "The Effect of B on the Co-Segregation of C-Cr at Grain Boundaries in Austenitic Steels",
                        "abstract": "In austenitic steels, the co-segregation of C and Cr at grain boundaries can result in the formation of Cr23C6. However, the addition of B to steels can effectively reduce the amount of Cr23C6 formed and inhibit its ripening in experiments, simultaneously transforming it into Cr23(BC)6. Therefore, the effect of B on the co-segregation of C and Cr at the \u03a35(210), \u03a39(221) and \u03a311(113) grain boundaries in austenitic steels was investigated using density functional theory. The results indicate that B, C, and Cr all tend to segregate at the three grain boundaries, with B and C showing a stronger segregation tendency. Furthermore, co-segregation of C and Cr with short distances occurs readily at the \u03a35(210), \u03a39(221) and \u03a311(113) grain boundaries. The presence of B at grain boundaries can impede the segregation of Cr, particularly at the \u03a39(221) and \u03a311(113) grain boundaries. When B is pre-segregating at the \u03a35(210) grain boundary, B, C, and Cr tend to co-segregate at the grain boundary. The grain boundary B hinders the accumulation of Cr near it for most grain boundaries, thereby inhibiting the co-segregation of Cr and C, making it difficult for B, C, and Cr to aggregate at most grain boundaries. This is beneficial for controlling the nucleation of Cr23(BC)6."
                    },
                    {
                        "title": "Improving Accuracy of Horizontal Flow Field Using Acoustic Tomography With Real-Time Station Position Correction",
                        "abstract": "High-precision and spatiotemporal flow field measurement in small-scale waters has a decisive impact on navigation and water resource management. Multitransmission rays and position drift, as two main difficulties in improving the accuracy of small-scale flow field, need to be solved urgently. In this article, four coastal acoustic tomography (CAT) systems are arranged in Qingyang Lake, Hunan, China, on March 1, 2022, for a short-term observation. As the GPS error and shipping caused station drift contaminate the reconstructed flow field, a real-time station position correction method, based on travel time and temperature, depth (TD) sensors, is applied in our inversion. The inversion range average current (RAC) is in agreement with forward calculation, which proves the reliability of the reconstructed horizontal flow field. Besides, comparisons of net flow and acoustic Doppler current profiler (ADCP) velocity are performed to validate the improvement of the real-time station position correction algorithm. The errors of the corrected net flow are less than 3.818 m 2/s, with a minimum of 0.0042 $\\text{m}^{2}$ /s. The root-mean-square error (RMSE) between the CAT inversion results and the ADCP velocity is 0.123 m/s after correction, relatively reduced 5.95% than before correction. The smaller the RMSE, the higher the inversion accuracy, so the CAT method with real-time station position correction is feasible for high-precision horizontal flow field reconstruction."
                    },
                    {
                        "title": "Loss-of-SIRT7 sensitizes hepatocellular carcinoma to sorafenib through the regulation of ERK Phosphorylation",
                        "abstract": "The FDA-approved oral multi-kinase inhibitor, sorafenib (BAY 43-9006, Nexavar), is the first approved systemic therapy for patients with unresectable hepatocellular carcinoma (HCC). Although it has been shown to significantly improve the overall survival of patients with HCC, drug resistance limits the response rate to this therapeutic. Here, we report that acquired sorafenib resistance is associated with overexpression of the deacetylase, SIRT7, and a high level of ERK phosphorylation. Further, we identify that the hyperactivation of ERK is controlled by SIRT7-mediated deacetylation of DDX3X. The inhibition of SIRT7 combined with sorafenib resulted in a marked reduction of cell viability in vitro and of tumor growth in vivo. It seems plausible that SIRT7 is responsible for the acquired sorafenib resistance and its inhibition is most likely beneficial together in conjunction with sorafenib by suppressing ERK signaling. Highlights Sorafenib resistance in HCC is associated with SIRT7 and ERK hyperactivation. Suppression of SIRT7 combined with sorafenib restores sensitivity to sorafenib. SIRT7 controls sorafenib resistance through ERK activation by mediating DDX3X deacetylation."
                    },
                    {
                        "title": "3-D Water Temperature Distribution Observation via Coastal Acoustic Tomography Sensing Network",
                        "abstract": "Water temperature is a critically important parameter for characterizing aquatic ecosystem, particularly in hydrodynamic processes produced by human activities or climate effects in shallow water regions (oceanic pasture, tidal currents, upwelling currents, etc.). The variation of water temperature is a slight and slow process, as technology advances, high-precision observation of the spatial and temporal variation of water temperature on 3-D scale is crucial to environmental research and predictions. To address this issue, 3-D water temperature field is reconstructed by coastal acoustic tomography technology and inversion water temperature variations are obtained. A multiacoustic station experiment was conducted in a reservoir in Changsha, China, from March 4, 2022 to March 5, 2022. Two grid-averaged 3-D inversion methods (Lagrangian-least-squares method and improved-Landweber iteration method) with multistation underwater sensing network are processed to reveal water temperature fluctuations. The rotated empirical orthogonal function is used to further analyze and evaluate the sensitivity of temperature variations in the individual spatial grids. In this article, water temperature distribution and dynamic variation process in the whole spatial structure are obtained and discussed using 3-D sound transmission information, and the feasibility of the methods is verified by comparing them with temperature depth sensor array observation results. This study provides high-precision ideas and methods for 3-D water temperature activities of underwater acoustic tomography network in shallow environments."
                    },
                    {
                        "title": "Wasserstein Distributionally Robust Policy Evaluation and Learning for Contextual Bandits",
                        "abstract": "Off-policy evaluation and learning are concerned with assessing a given policy and learning an optimal policy from offline data without direct interaction with the environment. Often, the environment in which the data are collected differs from the environment in which the learned policy is applied. To account for the effect of different environments during learning and execution, distributionally robust optimization (DRO) methods have been developed that compute worst-case bounds on the policy values assuming that the distribution of the new environment lies within an uncertainty set. Typically, this uncertainty set is defined based on the KL divergence around the empirical distribution computed from the logging dataset. However, the KL uncertainty set fails to encompass distributions with varying support and lacks awareness of the geometry of the distribution support. As a result, KL approaches fall short in addressing practical environment mismatches and lead to over-fitting to worst-case scenarios. To overcome these limitations, we propose a novel DRO approach that employs the Wasserstein distance instead. While Wasserstein DRO is generally computationally more expensive compared to KL DRO, we present a regularized method and a practical (biased) stochastic gradient descent method to optimize the policy efficiently. We also provide a theoretical analysis of the finite sample complexity and iteration complexity for our proposed method. We further validate our approach using a public dataset that was recorded in a randomized stoke trial."
                    },
                    {
                        "title": "Water Temperature Reconstruction via Station Position Correction Method Based on Coastal Acoustic Tomography Systems",
                        "abstract": "Underwater acoustic tomography is an advanced technology in water environment observation. Sound propagation duration between transceivers is used for underwater parameter distribution profile reconstruction in the inverse problem. The key points of acoustic tomography are accurate station distance and time synchronization. Two methods are introduced in this study for sound station position correction. The direct signal transmission correction (DSC) method corrects the multi-peak (expect direct ray) travel time via the travel time difference between different sound rays and reference direct ray. The ray-model position correction (RMC) method calculates exact station position by the station drift distance obtained from transceiver depth variations to correct direct ray travel time; the other multi-peak travel time is revised by the corrected direct ray travel time. A water temperature observation experiment was carried out in a reservoir using coastal acoustic tomography (CAT) systems to verify the flexibility of these two methods. Multi-ray arrival peaks are corrected using DSC and RMC methods; water temperature inversion results in a 2D vertical profile are obtained. The reliability of the method is proved by comparison with temperature depth sensor (TD) data. The methods improve the quality of initial data and can be attempted for further water environment observation in acoustic tomography observation studies."
                    },
                    {
                        "title": "Network Pharmacology and Experimental Validation to Explore the Effect and Mechanism of Kanglaite Injection Against Triple-Negative Breast Cancer",
                        "abstract": "Purpose Kanglaite injection (KLTi), made of Coix seed oil, has been shown to be effective in the treatment of numerous cancers. However, the anticancer mechanism requires further exploration. This study aimed to investigate the underlying anticancer mechanisms of KLTi in triple-negative breast cancer (TNBC) cells. Methods Public databases were searched for active compounds in KLTi, their potential targets and TNBC-related targets. KLTi\u2019s core targets and signaling pathways were determined through compound-target network, protein\u2013protein interaction (PPI) network, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. Molecular docking was carried out to predict the binding activity between active ingredients and key targets. In vitro experiments were conducted to further validate the predictions of network pharmacology. Results Fourteen active components of KLTi were screened from the database. Fifty-three candidate therapeutic targets were selected, and bioinformatics analysis was performed to identify the top two active compounds and three core targets. GO and KEGG enrichment analyses indicated that KLTi exerts therapeutic effects on TNBC through the cell cycle pathway. Molecular docking results showed that the main compounds of KLTi exhibited good binding activity to key target proteins. Results from in vitro experiments showed that KLTi inhibited proliferation and migration of TNBC cell lines 231 and 468, induced apoptosis, blocked cells in the G2/M phase, downregulated the mRNA expression of seven G2/M phase-related genes cyclin-dependent kinase 1 (CDK1), cyclin-dependent kinase 2 (CDK2), and checkpoint kinase 1 (CHEK1), cell division cycle 25A (CDC25A), cell division cycle 25B (CDC25B), maternal embryonic leucine zipper kinase (MELK), and aurora kinase A (AURKA), as well as downregulated CDK1 protein expression and up-regulated protein expression of Phospho-CDK1. Conclusion By utilizing network pharmacology, molecular docking, and in vitro experiments, KLTi was confirmed to have anti-TNBC effects by arresting cell cycle and inhibiting CDK1 dephosphorylation."
                    },
                    {
                        "title": "Measurement of water temperature and current in a Reservoir using coastal acoustic tomography",
                        "abstract": "Mapping of water temperature and current dynamic progress in ocean, coastal region and island is a the research emphasis. A coastal acoustic tomography (CAT) experiment for reconstructing the average temperature and current in Huangcai reservoir, Changsha, China was successfully carried out by three acoustic tomography stations during February 28th to March 2nd, 2022. The grid slice method is used to reconstruct grid-averaged temperature and grid-averaged current in 2D vertical profile. The new shore system and multi shallow systems synchronized atomic clock are used. And the function of new shallow-type system and experimental site are verified. From the inversion results of experiment, temperature flied has a significant stratification characteristic and current filed is influenced by terrain of the observation area. And the upwelling and downwelling current exist at the edge of exchange layer. The relationship between band frequency and station distance are also discussed."
                    },
                    {
                        "title": "Genetic Basis of Dorper Sheep (Ovis aries) Revealed by Long-Read De Novo Genome Assembly",
                        "abstract": "Dorper sheep (Ovis aries) (DPS), developed in the 1930s by crossing Dorset Horn and Blackhead Persian sheep in South Africa, is a world-famous composite breed for mutton production. The genetic basis underlying this breed is yet to be elucidated. Here, we report the sequencing and assembly of a highly contiguous Dorper sheep genome via integration of Oxford Nanopore Technology (ONT) sequencing and Hi-C (chromatin conformation capture) approaches. The assembled genome was around 2.64 Gb with a contig N50 of 73.33 Mb and 140 contigs in total. More than 99.5% of the assembled sequences could be anchored to 27 chromosomes and they were annotated with 20,450 protein-coding genes. Allele-specific expression (ASE) genes of Dorper sheep were revealed through ASE analysis and they were involved in the immune system, lipid metabolism, and environmental adaptation. A total of 5,701 and 456 allelic sites were observed in the SNP and indels loci identified from relevant whole-genome resequencing data. These allelic SNP and INDEL sites were annotated in 1,002 and 294 genes, respectively. Moreover, we calculated the number of variant sites and related genes derived from the maternal and paternal ancestors, revealing the genetic basis of outstanding phenotypic performance of Dorper sheep. In conclusion, this study reports the first reference genome of Dorper sheep and reveals its genetic basis through ASE. This study also provides a pipeline for mining genetic information of composite breeds, which has an implication for future hybrid-breeding practices."
                    },
                    {
                        "title": "Tomographic Mapping of Water Temperature and Current in a Reservoir by Trust-Region Method Based on CAT",
                        "abstract": "Remote sensing of water parameters and flow dynamic progress is of critical importance to water environment observation. Underwater acoustic tomography technology uses multiple sound stations to construct distributed sensing network for large-scale measurement, which has big superiority in water parameter observation compared with fix-point measurement. Water parameter variation progress can be monitored with an underwater acoustic tomography sensing system. This article develops a grid-averaged trust-region method for water parameter detailed profiling with coastal acoustic tomography (CAT) systems. A remote sensing experiment was conducted with five CAT systems from February 28 to March 7, 2022, in the Huangcai Reservoir in Changsha, China. Four CAT systems were deployed in the same position with different depths of sound transceivers, and another CAT station was set 650 m apart. Water temperature and flow current reconstruction results along the sound propagate Section were analyzed with the trust-region method and compared with the temperature data acquired with the temperature sensor. The accuracy and feasibility of this method are verified with observation results."
                    },
                    {
                        "title": "PID Control of a Superheated Steam Temperature System Based on Integral Gain Scheduling",
                        "abstract": "The high-quality operation of a superheated steam temperature (SST) system is a core fact of the safety, economy, and stability of thermal power units. How to improve the control performance of an SST system under large-varying operating conditions is becoming a research hotspot. To solve this challenge, this paper proposes a proportional integral derivative (PID) control strategy based on integral gain scheduling. Based on the introduction of the SST system and classical model under typical operating conditions, the control difficulties of the SST system are analyzed theoretically. Then, a PID control strategy, based on integral gain scheduling, is introduced for the cascade control structure, and the stability of the proposed control strategy is analyzed by calculating the PID stability region. Finally, the effectiveness of the proposed method is verified under nominal and uncertain conditions, where the proposed method could obtain satisfactory tracking and disturbance rejection control performance. Simulation results show the valuable application prospects of the proposed method."
                    },
                    {
                        "title": "The cuproptosis-related gene signature serves as a potential prognostic predictor for ovarian cancer using bioinformatics analysis",
                        "abstract": "Background Studies have shown that copper is involved in the tumorigenesis and development of ovarian cancer. In this work, we aimed to build a prognostic classification system associated with cuproptosis to predict ovarian cancer prognosis. Methods Information of ovarian cancer samples were acquired from The Cancer Genome Atlas (TCGA)-ovarian cancer and GSE26193 dataset. Cuproptosis-related genes were screened from previous research. ConsensusClusterPlus was applied to determine molecular subtypes, which were evaluated by tumor immune microenvironment analysis, TIDE algorithm, and functional enrichment analysis. Furthermore, limma analysis and univariate Cox analysis were used to construct a cuproptosis-related prognostic signature for ovarian cancer. Univariate and multivariate Cox regression analyses were used to analyze the independence of clinical factors and model. Results A total of 15 genes related to cuproptosis were identified, and 2 clusters (C1 and C2) were determined. C1 had a better survival outcome, less advanced stage, enhanced immune infiltration, was more sensitive to immunotherapy, and showed enrichment in tricarboxylic acid (TCA)-related pathways. An 8 cuproptosis-associated gene signature was constructed, and the signature was verified in the GSE26193 dataset. A higher risk score of the cuproptosis-related gene signature was significantly correlated with worse overall survival (OS) (P<0.0001), which was validated in GSE26193 dataset successfully. Cox survival analysis showed that risk score was an independent predictor [hazard ratio (HR) =2.66, P<0.001]. Functional enrichment and tumor immune microenvironment analyses showed that high-risk patients tended to have immunologically sensitive tumors. Conclusions The cuproptosis-related gene signature may serve as a potential prognostic predictor for ovarian cancer patients and may offer novel treatment strategies for ovarian cancer."
                    },
                    {
                        "title": "Global Convergence of Localized Policy Iteration in Networked Multi-Agent Reinforcement Learning",
                        "abstract": "We study a multi-agent reinforcement learning (MARL) problem where the agents interact over a given network. The goal of the agents is to cooperatively maximize the average of their entropy-regularized long-term rewards. To overcome the curse of dimensionality and to reduce communication, we propose a Localized Policy Iteration (LPI) algorithm that provably learns a near-globally-optimal policy using only local information. In particular, we show that, despite restricting each agent's attention to only its \u03ba-hop neighborhood, the agents are able to learn a policy with an optimality gap that decays polynomially in \u03ba. In addition, we show the finite-sample convergence of LPI to the global optimal policy, which explicitly captures the trade-off between optimality and computational complexity in choosing \u03ba. Numerical simulations demonstrate the effectiveness of LPI."
                    },
                    {
                        "title": "A Mouse Model of Damp-Heat Syndrome in Traditional Chinese Medicine and Its Impact on Pancreatic Tumor Growth",
                        "abstract": "Background Damp-heat syndrome is one of the most important syndrome types in the traditional Chinese medicine (TCM) syndrome differentiation and treatment system, as well as the core pathogenesis of pancreatic cancer (PC) which remains a challenge to medical researchers due to its insidious onset and poor prognosis. Great attention has been given to the impact of damp-heat syndrome on tumorigenesis and progression, but less attention has been given to damp-heat modeling per se. Studying PC in a proper damp-heat syndrome animal model can recapitulate the actual pathological process and contribute to treatment strategy improvement. Methods Here, an optimized damp-heat syndrome mouse model was established based on our prior experience. The Fibonacci method was applied to determine the maximum tolerated dosage of alcohol for mice. Damp-heat syndrome modeling with the old and new methods was performed in parallel of comparative study about general appearance, food intake, water consumption and survival. Major organs, including the liver, kidneys, lungs, pancreas, spleen, intestines and testes, were collected for histological evaluation. Complete blood counts and biochemical tests were conducted to characterize changes in blood circulation. PC cells were subcutaneously inoculated into mice with damp-heat syndrome to explore the impact of damp-heat syndrome on PC growth. Hematoxylin-eosin staining, Masson staining and immunohistochemistry were performed for pathological evaluation. A chemokine microarray was applied to screen the cytokines mediating the proliferation-promoting effects of damp-heat syndrome, and quantitative polymerase chain reaction and Western blotting were conducted for results validation. Results The new modeling method has the advantages of mouse-friendly features, easily accessible materials, simple operation, and good stability. More importantly, a set of systematic indicators was proposed for model evaluation. The new modeling method verified the pancreatic tumor-promoting role of damp-heat syndrome. Damp-heat syndrome induced the proliferation of cancer-associated fibroblasts and promoted desmoplasia. In addition, circulating and tumor-located chemokine levels were altered by damp-heat syndrome, characterized by tumor promotion and immune suppression. Conclusions This study established a stable and reproducible murine model of damp-heat syndrome in TCM with systematic evaluation methods. Cancer associated fibroblast-mediated desmoplasia and chemokine production contribute to the tumor-promoting effect of damp-heat syndrome on PC."
                    },
                    {
                        "title": "Study on Mechanism of Yiqi Yangyin Jiedu Recipe Inhibiting Triple Negative Breast Cancer Growth: A Network Pharmacology and In Vitro Verification",
                        "abstract": "Background The present study explores the potential mechanism of Yiqi yangyin jiedu Recipe (YQYYJDR) on triple negative breast cancer via adopting network pharmacology and experimental validation. Materials and Methods The potential active compounds and target genes of YQYYJDR were screened out from TCMSP database with OB \u2265 30% and DL index \u2265 0.18. The potential pathways and function enrichment were identified from Metascape website. MDA-MB-231 and MDA-MB-468 cells were tested for cell viability, invasion, and apoptosis by in vitro and in vivo experiments. Results A total of 153 bioactive compounds and 281 target genes of YQYYJDR were retrieved from TCMSP database. The top 5 enrichment pathways of YQYYJDR target genes include pathways in cancer, AGE-RAGE signaling pathway in diabetic complications, proteoglycans in cancer, IL-17 signaling pathway, and platinum drug resistance. 65 target genes were included in the pathway of cancer. Biological function enrichment analysis of 65 genes showed YQYYJDR inhibited tumor growth mainly through apoptotic pathway. In vitro experiments showed that YQYYJDR could inhibit the proliferation and invasion of MDA-MB-231 and MDA-MB-468 cells, arrest cells in S stage, and induce cell apoptosis. YQYYJDR upregulated BAX, caspase3, and cleaved caspase3 expression and downregulated BCL2 expression. In vivo experiments showed that YQYYJDR could inhibit tumor growth. Conclusions In this study, network pharmacology and experiment were used to explore the mechanism of YQYYJDR on triple negative breast cancer. In vitro and in vivo experiments showed that YQYYJDR could inhibit the growth of triple negative breast cancer and induce cell apoptosis. Apoptosis pathway plays a significant role in the treatment of triple negative breast cancer."
                    }
                ]
            },
            "d4e73f56-2a59-4054-8f05-abb1252bc305": {
                "pk": "d4e73f56-2a59-4054-8f05-abb1252bc305",
                "name": "A. Rupam Mahmood",
                "collaborators": [
                    "G. Vasan",
                    "Samuele Tosatto",
                    "Yangchen Pan",
                    "Martha White",
                    "D. Korenkevych",
                    "J. Bergstra",
                    "Fengdi Che",
                    "M. Elsayed",
                    "Qingfeng Lan",
                    "Jun Luo"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Continual Learning",
                    "Policy Gradient",
                    "Function Approximation"
                ],
                "publications": [
                    {
                        "title": "Correcting discount-factor mismatch in on-policy policy gradient methods",
                        "abstract": "The policy gradient theorem gives a convenient form of the policy gradient in terms of three factors: an action value, a gradient of the action likelihood, and a state distribution involving discounting called the \\emph{discounted stationary distribution}. But commonly used on-policy methods based on the policy gradient theorem ignores the discount factor in the state distribution, which is technically incorrect and may even cause degenerate learning behavior in some environments. An existing solution corrects this discrepancy by using $\\gamma^t$ as a factor in the gradient estimate. However, this solution is not widely adopted and does not work well in tasks where the later states are similar to earlier states. We introduce a novel distribution correction to account for the discounted stationary distribution that can be plugged into many existing gradient estimators. Our correction circumvents the performance degradation associated with the $\\gamma^t$ correction with a lower variance. Importantly, compared to the uncorrected estimators, our algorithm provides improved state emphasis to evade suboptimal policies in certain environments and consistently matches or exceeds the original performance on several OpenAI gym and DeepMind suite benchmarks."
                    },
                    {
                        "title": "Utility-based Perturbed Gradient Descent: An Optimizer for Continual Learning",
                        "abstract": "Modern representation learning methods often struggle to adapt quickly under non-stationarity because they suffer from catastrophic forgetting and decaying plasticity. Such problems prevent learners from fast adaptation since they may forget useful features or have difficulty learning new ones. Hence, these methods are rendered ineffective for continual learning. This paper proposes Utility-based Perturbed Gradient Descent (UPGD), an online learning algorithm well-suited for continual learning agents. UPGD protects useful weights or features from forgetting and perturbs less useful ones based on their utilities. Our empirical results show that UPGD helps reduce forgetting and maintain plasticity, enabling modern representation learning methods to work effectively in continual learning."
                    },
                    {
                        "title": "Loosely consistent emphatic temporal-difference learning",
                        "abstract": "There has been signi\ufb01cant interest in searching for off-policy Temporal-Difference (TD) algorithms that \ufb01nd the same solution that would have been obtained in the on-policy regime. An important property of such algorithms is that their expected update has the same \ufb01xed point as that of On-policy TD( \u03bb ), which we call loose consistency . Notably, Full-IS-TD( \u03bb ) is the only existing loosely consistent method under general linear function approximation but, unfortunately, has a high variance and is scarcely practical. This notorious high variance issue motivates the introduction of ETD( \u03bb ), which tames down the variance but has a biased \ufb01xed point. Inspired by these two methods, we propose a new loosely consistent algorithm called Average Emphatic TD (AETD( \u03bb )) with a transient bias, which strikes a balance between bias and variance. Further, we unify AETD( \u03bb ) with existing methods and obtain a new family of loosely consistent algorithms called Loosely Consistent Emphatic TD (LC-ETD( \u03bb , \u03b2 , \u03bd )), which can con-trol a smooth bias-variance trade-off by varying the speed at which the transient bias fades. Through experiments on illustrative examples, we show the effectiveness and practicality of LC-ETD( \u03bb , \u03b2 , \u03bd ). 1"
                    },
                    {
                        "title": "Dynamic Decision Frequency with Continuous Options",
                        "abstract": "In classic reinforcement learning algorithms, agents make decisions at discrete and fixed time intervals. The duration between decisions becomes a crucial hyperparameter, as setting it too short may increase the problem's difficulty by requiring the agent to make numerous decisions to achieve its goal while setting it too long can result in the agent losing control over the system. However, physical systems do not necessarily require a constant control frequency, and for learning agents, it is often preferable to operate with a low frequency when possible and a high frequency when necessary. We propose a framework called Continuous-Time Continuous-Options (CTCO), where the agent chooses options as sub-policies of variable durations. These options are time-continuous and can interact with the system at any desired frequency providing a smooth change of actions. We demonstrate the effectiveness of CTCO by comparing its performance to classical RL and temporal-abstraction RL methods on simulated continuous control tasks with various action-cycle times. We show that our algorithm's performance is not affected by the choice of environment interaction frequency. Furthermore, we demonstrate the efficacy of CTCO in facilitating exploration in a real-world visual reaching task for a 7 DOF robotic arm with sparse rewards."
                    },
                    {
                        "title": "Memory-efficient Reinforcement Learning with Value-based Knowledge Consolidation",
                        "abstract": "Artificial neural networks are promising for general function approximation but challenging to train on non-independent or non-identically distributed data due to catastrophic forgetting. The experience replay buffer, a standard component in deep reinforcement learning, is often used to reduce forgetting and improve sample efficiency by storing experiences in a large buffer and using them for training later. However, a large replay buffer results in a heavy memory burden, especially for onboard and edge devices with limited memory capacities. We propose memory-efficient reinforcement learning algorithms based on the deep Q-network algorithm to alleviate this problem. Our algorithms reduce forgetting and maintain high sample efficiency by consolidating knowledge from the target Q-network to the current Q-network. Compared to baseline methods, our algorithms achieve comparable or better performance in both feature-based and image-based tasks while easing the burden of large experience replay buffers."
                    },
                    {
                        "title": "A Temporal-Difference Approach to Policy Gradient Estimation",
                        "abstract": "The policy gradient theorem (Sutton et al., 2000) prescribes the usage of a cumulative discounted state distribution under the target policy to approximate the gradient. Most algorithms based on this theorem, in practice, break this assumption, introducing a distribution shift that can cause the convergence to poor solutions. In this paper, we propose a new approach of reconstructing the policy gradient from the start state without requiring a particular sampling strategy. The policy gradient calculation in this form can be simplified in terms of a gradient critic, which can be recursively estimated due to a new Bellman equation of gradients. By using temporal-difference updates of the gradient critic from an off-policy data stream, we develop the first estimator that sidesteps the distribution shift issue in a model-free way. We prove that, under certain realizability conditions, our estimator is unbiased regardless of the sampling strategy. We empirically show that our technique achieves a superior bias-variance trade-off and performance in presence of off-policy samples."
                    },
                    {
                        "title": "HesScale: Scalable Computation of Hessian Diagonals",
                        "abstract": "Second-order optimization uses curvature information about the objective function, which can help in faster convergence. However, such methods typically require expensive computation of the Hessian matrix, preventing their usage in a scalable way. The absence of efficient ways of computation drove the most widely used methods to focus on first-order approximations that do not capture the curvature information. In this paper, we develop HesScale, a scalable approach to approximating the diagonal of the Hessian matrix, to incorporate second-order information in a computationally efficient manner. We show that HesScale has the same computational complexity as backpropagation. Our results on supervised classification show that HesScale achieves high approximation accuracy, allowing for scalable and efficient second-order optimization."
                    },
                    {
                        "title": "Memory-efficient Reinforcement Learning with Knowledge Consolidation",
                        "abstract": "Arti\ufb01cial neural networks are promising for general function approximation but challenging to train on non-independent or non-identically distributed data due to catastrophic forgetting. The experience replay buffer, a standard component in deep reinforcement learning, is often used to reduce forgetting and improve sample ef\ufb01ciency by storing experiences in a large buffer and using them for training later. However, a large replay buffer results in a heavy memory burden, especially for onboard and edge devices with limited memory capacities. We propose memory-ef\ufb01cient reinforcement learning algorithms based on the deep Q-network algorithm to alleviate this problem. Our algorithms reduce forgetting and maintain high sample ef\ufb01ciency by consolidating knowledge from the target Q-network to the current Q-network. Compared to baseline methods, our algorithms achieve comparable or better performance in both feature-based and image-based tasks while easing the burden of large experience replay buffers. 1"
                    },
                    {
                        "title": "Real-Time Reinforcement Learning for Vision-Based Robotics Utilizing Local and Remote Computers",
                        "abstract": "Real-time learning is crucial for robotic agents adapting to ever-changing, non-stationary environments. A common setup for a robotic agent is to have two different computers simultaneously: a resource-limited local computer tethered to the robot and a powerful remote computer connected wirelessly. Given such a setup, it is unclear to what extent the performance of a learning system can be affected by resource limitations and how to efficiently use the wirelessly connected powerful computer to compensate for any performance loss. In this paper, we implement a real-time learning system called the Remote-Local Distributed (ReLoD) system to distribute computations of two deep reinforcement learning (RL) algorithms, Soft Actor-Critic (SAC) and Proximal Policy Optimization (PPO), between a local and a remote computer. The performance of the system is evaluated on two vision-based control tasks developed using a robotic arm and a mobile robot. Our results show that SAC's performance degrades heavily on a resource-limited local computer. Strikingly, when all computations of the learning system are deployed on a remote workstation, SAC fails to compensate for the performance loss, indicating that, without careful consideration, using a powerful remote computer may not result in performance improvement. However, a carefully chosen distribution of computations of SAC consistently and substantially improves its performance on both tasks. On the other hand, the performance of PPO remains largely unaffected by the distribution of computations. In addition, when all computations happen solely on a powerful tethered computer, the performance of our system remains on par with an existing system that is well-tuned for using a single machine. ReLoD is the only publicly available system for real-time RL that applies to multiple robots for vision-based tasks. The source code can be found at https://github.com/rlai-lab/relod"
                    },
                    {
                        "title": "An Alternate Policy Gradient Estimator for Softmax Policies",
                        "abstract": "Policy gradient (PG) estimators are ine\ufb00ec-tive in dealing with softmax policies that are sub-optimally saturated, which refers to the situation when the policy concentrates its probability mass on sub-optimal actions. Sub-optimal policy saturation may arise from bad policy initialization or sudden changes in the environment that occur after the policy has already converged. Current softmax PG estimators require a large number of updates to overcome policy saturation, which causes low sample e\ufb03ciency and poor adaptability to new situations. To mitigate this problem, we propose a novel PG estimator for softmax policies that utilizes the bias in the critic estimate and the noise present in the reward signal to escape the saturated regions of the policy parameter space. Our theoretical analysis and experiments, conducted on bandits and various reinforcement learning environments, show that this new estimator is signi\ufb01cantly more robust to policy saturation. for the gradient bandit estimators with true r \u03c0 and an additional normally distributed gradient noise. The individual graph labels (0 , 0 . 1 , 0 . 5 , 1 , 2) show the standard deviation of the normal noise added to the gradient. The subplot headers show the policy saturation at the beginning of the experiment."
                    },
                    {
                        "title": "Continual Backprop: Stochastic Gradient Descent with Persistent Randomness",
                        "abstract": "The Backprop algorithm for learning in neural networks utilizes two mechanisms: first, stochastic gradient descent and second, initialization with small random weights, where the latter is essential to the effectiveness of the former. We show that in continual learning setups, Backprop performs well initially, but over time its performance degrades. Stochastic gradient descent alone is insufficient to learn continually; the initial randomness enables only initial learning but not continual learning. To the best of our knowledge, ours is the first result showing this degradation in Backprop's ability to learn. To address this degradation in Backprop's plasticity, we propose an algorithm that continually injects random features alongside gradient descent using a new generate-and-test process. We call this the \\textit{Continual Backprop} algorithm. We show that, unlike Backprop, Continual Backprop is able to continually adapt in both supervised and reinforcement learning (RL) problems. Continual Backprop has the same computational complexity as Backprop and can be seen as a natural extension of Backprop for continual learning."
                    },
                    {
                        "title": "Greedification Operators for Policy Optimization: Investigating Forward and Reverse KL Divergences",
                        "abstract": "Approximate Policy Iteration (API) algorithms alternate between (approximate) policy evaluation and (approximate) greedification. Many different approaches have been explored for approximate policy evaluation, but less is understood about approximate greedification and what choices guarantee policy improvement. In this work, we investigate approximate greedification when reducing the KL divergence between the parameterized policy and the Boltzmann distribution over action values. In particular, we investigate the difference between the forward and reverse KL divergences, with varying degrees of entropy regularization. We show that the reverse KL has stronger policy improvement guarantees, but that reducing the forward KL can result in a worse policy. We also demonstrate, however, that a large enough reduction of the forward KL can induce improvement under additional assumptions. Empirically, we show on simple continuous-action environments that the forward KL can induce more exploration, but at the cost of a more suboptimal policy. No significant differences were observed in the discrete-action setting or on a suite of benchmark problems. Throughout, we highlight that many policy gradient methods can be seen as an instance of API, with either the forward or reverse KL for the policy update, and discuss next steps for understanding and improving our policy optimization algorithms."
                    },
                    {
                        "title": "Autoregressive Policies for Continuous Control Deep Reinforcement Learning",
                        "abstract": "Reinforcement learning algorithms rely on exploration to discover new behaviors, which is typically achieved by following a stochastic policy. In continuous control tasks, policies with a Gaussian distribution have been widely adopted. Gaussian exploration however does not result in smooth trajectories that generally correspond to safe and rewarding behaviors in practical tasks. In addition, Gaussian policies do not result in an effective exploration of an environment and become increasingly inefficient as the action rate increases. This contributes to a low sample efficiency often observed in learning continuous control tasks. We introduce a family of stationary autoregressive (AR) stochastic processes to facilitate exploration in continuous control domains. We show that proposed processes possess two desirable features: subsequent process observations are temporally coherent with continuously adjustable degree of coherence, and the process stationary distribution is standard normal. We derive an autoregressive policy (ARP) that implements such processes maintaining the standard agent-environment interface. We show how ARPs can be easily used with the existing off-the-shelf learning algorithms. Empirically we demonstrate that using ARPs results in improved exploration and sample efficiency in both simulated and real world domains, and, furthermore, provides smooth exploration trajectories that enable safe operation of robotic hardware."
                    },
                    {
                        "title": "Setting up a Reinforcement Learning Task with a Real-World Robot",
                        "abstract": "Reinforcement learning is a promising approach to developing hard-to-engineer adaptive solutions for complex and diverse robotic tasks. However, learning with real-world robots is often unreliable and difficult, which resulted in their low adoption in reinforcement learning research. This difficulty is worsened by the lack of guidelines for setting up learning tasks with robots. In this work, we develop a learning task with a UR5 robotic arm to bring to light some key elements of a task setup and study their contributions to the challenges with robots11Source code of the task and the computational model behind the setup available at https://github.com/kindredresearch/SenseAct. We find that learning performance can be highly sensitive to the setup, and thus oversights and omissions in setup details can make effective learning, reproducibility, and fair comparison hard. Our study suggests some mitigating steps to help future experimenters avoid difficulties and pitfalls. We show that highly reliable and repeatable experiments can be performed in our setup, indicating the possibility of reinforcement learning research extensively based on real-world robots."
                    },
                    {
                        "title": "Benchmarking Reinforcement Learning Algorithms on Real-World Robots",
                        "abstract": "Through many recent successes in simulation, model-free reinforcement learning has emerged as a promising approach to solving continuous control robotic tasks. The research community is now able to reproduce, analyze and build quickly on these results due to open source implementations of learning algorithms and simulated benchmark tasks. To carry forward these successes to real-world applications, it is crucial to withhold utilizing the unique advantages of simulations that do not transfer to the real world and experiment directly with physical robots. However, reinforcement learning research with physical robots faces substantial resistance due to the lack of benchmark tasks and supporting source code. In this work, we introduce several reinforcement learning tasks with multiple commercially available robots that present varying levels of learning difficulty, setup, and repeatability. On these tasks, we test the learning performance of off-the-shelf implementations of four reinforcement learning algorithms and analyze sensitivity to their hyper-parameters to determine their readiness for applications in various real-world tasks. Our results show that with a careful setup of the task interface and computations, some of these implementations can be readily applicable to physical robots. We find that state-of-the-art learning algorithms are highly sensitive to their hyper-parameters and their relative ordering does not transfer across tasks, indicating the necessity of re-tuning them for each task for best performance. On the other hand, the best hyper-parameter configuration from one task may often result in effective learning on held-out tasks even with different robots, providing a reasonable default. We make the benchmark tasks publicly available to enhance reproducibility in real-world reinforcement learning."
                    },
                    {
                        "title": "Multi-step Off-policy Learning Without Importance Sampling Ratios",
                        "abstract": "To estimate the value functions of policies from exploratory data, most model-free off-policy algorithms rely on importance sampling, where the use of importance sampling ratios often leads to estimates with severe variance. It is thus desirable to learn off-policy without using the ratios. However, such an algorithm does not exist for multi-step learning with function approximation. In this paper, we introduce the first such algorithm based on temporal-difference (TD) learning updates. We show that an explicit use of importance sampling ratios can be eliminated by varying the amount of bootstrapping in TD updates in an action-dependent manner. Our new algorithm achieves stability using a two-timescale gradient-based TD update. A prior algorithm based on lookup table representation called Tree Backup can also be retrieved using action-dependent bootstrapping, becoming a special case of our algorithm. In two challenging off-policy tasks, we demonstrate that our algorithm is stable, effectively avoids the large variance issue, and can perform substantially better than its state-of-the-art counterpart."
                    },
                    {
                        "title": "Incremental Off-policy Reinforcement Learning Algorithms",
                        "abstract": "Model-free off-policy temporal-difference (TD) algorithms form a powerful component of scalable predictive knowledge representation due to their ability to learn numerous counterfactual predictions in a computationally scalable manner. In this dissertation, we address and overcome two shortcomings of off-policy TD algorithms that preclude a widespread use in knowledge representation: they often produce high variance estimates and may become unstable. The issue of high variance occurs when off-policy TD algorithms use importance sampling (IS), a standard approach to adjusting estimates when the sample and the target distributions are different. IS performs this adjustment by scaling the estimate by the likelihood ratio, also known as the importance sampling ratio. However, this ratio also often produces high variance in the estimates. We provide two different approaches to overcome this shortcoming. The first approach relies on a well-known alternative to the ordinary form of IS, called the Weighted Importance Sampling (WIS), which is known to produce much less variance but has been under-utilized in reinforcement learning tasks. The main obstacle to using WIS in large-scale reinforcement learning tasks is that it is not known how to systematically extend a simple lookup-table-based estimator such as WIS to parametric function approximation with real-time computationally scalable updates. We overcome this obstacle by developing new techniques for deriving computationally efficient equivalent TD updates. This also allows us to simplify the derivations of some existing computationally efficient algorithms as well as produce new ones. In the second approach, we directly address the main factor underlying the high variance issue: the use of the IS ratios. Ratios play a key role in producing high variance estimation as they may vary drastically from one sample to another. We show that the ratios can be eliminated from the updates by varying the amount of bootstrapping, a sophisticated technique for allowing a spectrum of multi-step TD algorithms. The core idea is to allow the bootstrapping parameter, which controls the amount of bootstrapping, vary and shrink in a specific way. The elimination of ratios from the updates using this approach results directly in the reduction of variance compared to IS-based algorithms while still retaining the multi-step advantage of TD algorithms. We empirically evaluate the algorithms based on both approaches using several idealized experiments. We show that the WIS-based algorithms outperform existing off-policy learning algorithms often with an order of magnitude margin. On the other hand, the bootstrapping-based methods can retain more effectively the multi-step advantage compared to the IS-based methods despite occasionally shrinking the bootstrapping parameter. The issue of instability with off-policy TD algorithms may appear when both function approximation and bootstrapping are used together, a desirable combination in reinforcement learning tasks. We discuss the key factors responsible for instability by analyzing the deterministic counterpart of TD updates. We illustrate that the on-policy TD algorithm may also be unstable if the updates are scaled or the bootstrapping parameter is set differently for different states but remains stable with a state-dependent discount factor. The TD algorithm can be made stable in all these cases by ensuring that a certain matrix involved in the deterministic TD update is always positive definite. We provide a systematic approach to ensuring positive definiteness of this matrix, which amounts to selectively emphasizing and de-emphasizing the updates of the stochastic counterpart in a particular manner. The resulting algorithm, which we call the emphatic TD algorithm, is stable, do not introduce extra tunable parameters and its additional memory or computational complexity is negligible. Finally, we provide computationally efficient methods to perform a search over features for improving the representation of learning systems. These search methods enable continual discovery and maintenance of relevant and useful predictions that can be used as a building block of scalable predictive knowledge representations in long-lived systems."
                    }
                ]
            },
            "a279aee3-530a-4816-ab64-578645de64d8": {
                "pk": "a279aee3-530a-4816-ab64-578645de64d8",
                "name": "Doina Precup",
                "collaborators": [
                    "D. Meger",
                    "Chenqing Hua",
                    "Sitao Luan",
                    "Elaine Lau",
                    "P. Panangaden",
                    "S. Rezaei-Shoshtari",
                    "Rosie Zhao",
                    "Ankit Anand",
                    "David Abel",
                    "Andr\u00e9 Barreto"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Graph Neural Network",
                    "Representation Learning",
                    "Drug Discovery"
                ],
                "publications": [
                    {
                        "title": "An Empirical Study of the Effectiveness of Using a Replay Buffer on Mode Discovery in GFlowNets",
                        "abstract": "Reinforcement Learning (RL) algorithms aim to learn an optimal policy by iteratively sampling actions to learn how to maximize the total expected return, $R(x)$. GFlowNets are a special class of algorithms designed to generate diverse candidates, $x$, from a discrete set, by learning a policy that approximates the proportional sampling of $R(x)$. GFlowNets exhibit improved mode discovery compared to conventional RL algorithms, which is very useful for applications such as drug discovery and combinatorial search. However, since GFlowNets are a relatively recent class of algorithms, many techniques which are useful in RL have not yet been associated with them. In this paper, we study the utilization of a replay buffer for GFlowNets. We explore empirically various replay buffer sampling techniques and assess the impact on the speed of mode discovery and the quality of the modes discovered. Our experimental results in the Hypergrid toy domain and a molecule synthesis environment demonstrate significant improvements in mode discovery when training with a replay buffer, compared to training only with trajectories generated on-policy."
                    },
                    {
                        "title": "Acceleration in Policy Optimization",
                        "abstract": "We work towards a unifying paradigm for accelerating policy optimization methods in reinforcement learning (RL) by integrating foresight in the policy improvement step via optimistic and adaptive updates. Leveraging the connection between policy iteration and policy gradient methods, we view policy optimization algorithms as iteratively solving a sequence of surrogate objectives, local lower bounds on the original objective. We define optimism as predictive modelling of the future behavior of a policy, and adaptivity as taking immediate and anticipatory corrective actions to mitigate accumulating errors from overshooting predictions or delayed responses to change. We use this shared lens to jointly express other well-known algorithms, including model-based policy improvement based on forward search, and optimistic meta-learning algorithms. We analyze properties of this formulation, and show connections to other accelerated optimization algorithms. Then, we design an optimistic policy gradient algorithm, adaptive via meta-gradient learning, and empirically highlight several design choices pertaining to acceleration, in an illustrative task."
                    },
                    {
                        "title": "Policy Gradient Methods in the Presence of Symmetries and State Abstractions",
                        "abstract": "Reinforcement learning (RL) on high-dimensional and complex problems relies on abstraction for improved efficiency and generalization. In this paper, we study abstraction in the continuous-control setting, and extend the definition of Markov decision process (MDP) homomorphisms to the setting of continuous state and action spaces. We derive a policy gradient theorem on the abstract MDP for both stochastic and deterministic policies. Our policy gradient results allow for leveraging approximate symmetries of the environment for policy optimization. Based on these theorems, we propose a family of actor-critic algorithms that are able to learn the policy and the MDP homomorphism map simultaneously, using the lax bisimulation metric. Finally, we introduce a series of environments with continuous symmetries to further demonstrate the ability of our algorithm for action abstraction in the presence of such symmetries. We demonstrate the effectiveness of our method on our environments, as well as on challenging visual control tasks from the DeepMind Control Suite. Our method's ability to utilize MDP homomorphisms for representation learning leads to improved performance, and the visualizations of the latent space clearly demonstrate the structure of the learned abstraction."
                    },
                    {
                        "title": "Accelerating exploration and representation learning with offline pre-training",
                        "abstract": "Sequential decision-making agents struggle with long horizon tasks, since solving them requires multi-step reasoning. Most reinforcement learning (RL) algorithms address this challenge by improved credit assignment, introducing memory capability, altering the agent's intrinsic motivation (i.e. exploration) or its worldview (i.e. knowledge representation). Many of these components could be learned from offline data. In this work, we follow the hypothesis that exploration and representation learning can be improved by separately learning two different models from a single offline dataset. We show that learning a state representation using noise-contrastive estimation and a model of auxiliary reward separately from a single collection of human demonstrations can significantly improve the sample efficiency on the challenging NetHack benchmark. We also ablate various components of our experimental setting and highlight crucial insights."
                    },
                    {
                        "title": "Minimal Value-Equivalent Partial Models for Scalable and Robust Planning in Lifelong Reinforcement Learning",
                        "abstract": "Learning models of the environment from pure interaction is often considered an essential component of building lifelong reinforcement learning agents. However, the common practice in model-based reinforcement learning is to learn models that model every aspect of the agent's environment, regardless of whether they are important in coming up with optimal decisions or not. In this paper, we argue that such models are not particularly well-suited for performing scalable and robust planning in lifelong reinforcement learning scenarios and we propose new kinds of models that only model the relevant aspects of the environment, which we call\"minimal value-equivalent partial models\". After providing a formal definition for these models, we provide theoretical results demonstrating the scalability advantages of performing planning with such models and then perform experiments to empirically illustrate our theoretical results. Then, we provide some useful heuristics on how to learn these kinds of models with deep learning architectures and empirically demonstrate that models learned in such a way can allow for performing planning that is robust to distribution shifts and compounding model errors. Overall, both our theoretical and empirical results suggest that minimal value-equivalent partial models can provide significant benefits to performing scalable and robust planning in lifelong reinforcement learning scenarios."
                    },
                    {
                        "title": "For SALE: State-Action Representation Learning for Deep Reinforcement Learning",
                        "abstract": "In the field of reinforcement learning (RL), representation learning is a proven tool for complex image-based tasks, but is often overlooked for environments with low-level states, such as physical control problems. This paper introduces SALE, a novel approach for learning embeddings that model the nuanced interaction between state and action, enabling effective representation learning from low-level states. We extensively study the design space of these embeddings and highlight important design considerations. We integrate SALE and an adaptation of checkpoints for RL into TD3 to form the TD7 algorithm, which significantly outperforms existing continuous control algorithms. On OpenAI gym benchmark tasks, TD7 has an average performance gain of 276.7% and 50.7% over TD3 at 300k and 5M time steps, respectively, and works in both the online and offline settings."
                    },
                    {
                        "title": "On the Convergence of Bounded Agents",
                        "abstract": "When has an agent converged? Standard models of the reinforcement learning problem give rise to a straightforward definition of convergence: An agent converges when its behavior or performance in each environment state stops changing. However, as we shift the focus of our learning problem from the environment's state to the agent's state, the concept of an agent's convergence becomes significantly less clear. In this paper, we propose two complementary accounts of agent convergence in a framing of the reinforcement learning problem that centers around bounded agents. The first view says that a bounded agent has converged when the minimal number of states needed to describe the agent's future behavior cannot decrease. The second view says that a bounded agent has converged just when the agent's performance only changes if the agent's internal state changes. We establish basic properties of these two definitions, show that they accommodate typical views of convergence in standard settings, and prove several facts about their nature and relationship. We take these perspectives, definitions, and analysis to bring clarity to a central idea of the field."
                    },
                    {
                        "title": "MUDiff: Unified Diffusion for Complete Molecule Generation",
                        "abstract": "Molecule generation is a very important practical problem, with uses in drug discovery and material design, and AI methods promise to provide useful solutions. However, existing methods for molecule generation focus either on 2D graph structure or on 3D geometric structure, which is not sufficient to represent a complete molecule as 2D graph captures mainly topology while 3D geometry captures mainly spatial atom arrangements. Combining these representations is essential to better represent a molecule. In this paper, we present a new model for generating a comprehensive representation of molecules, including atom features, 2D discrete molecule structures, and 3D continuous molecule coordinates, by combining discrete and continuous diffusion processes. The use of diffusion processes allows for capturing the probabilistic nature of molecular processes and exploring the effect of different factors on molecular structures. Additionally, we propose a novel graph transformer architecture to denoise the diffusion process. The transformer adheres to 3D roto-translation equivariance constraints, allowing it to learn invariant atom and edge representations while preserving the equivariance of atom coordinates. This transformer can be used to learn molecular representations robust to geometric transformations. We evaluate the performance of our model through experiments and comparisons with existing methods, showing its ability to generate more stable and valid molecules. Our model is a promising approach for designing stable and diverse molecules and can be applied to a wide range of tasks in molecular modeling."
                    },
                    {
                        "title": "On the Challenges of using Reinforcement Learning in Precision Drug Dosing: Delay and Prolongedness of Action Effects",
                        "abstract": "Drug dosing is an important application of AI, which can be formulated as a Reinforcement Learning (RL) problem. In this paper, we identify two major challenges of using RL for drug dosing: delayed and prolonged effects of administering medications, which break the Markov assumption of the RL framework. We focus on prolongedness and define PAE-POMDP (Prolonged Action Effect-Partially Observable Markov Decision Process), a subclass of POMDPs in which the Markov assumption does not hold specifically due to prolonged effects of actions. Motivated by the pharmacology literature, we propose a simple and effective approach to converting drug dosing PAE-POMDPs into MDPs, enabling the use of the existing RL algorithms to solve such problems. We validate the proposed approach on a toy task, and a challenging glucose control task, for which we devise a clinically-inspired reward function. Our results demonstrate that: (1) the proposed method to restore the Markov assumption leads to significant improvements over a vanilla baseline; (2) the approach is competitive with recurrent policies which may inherently capture the prolonged affect of actions; (3) it is remarkably more time and memory efficient than the recurrent baseline and hence more suitable for real-time dosing control systems; and (4) it exhibits favourable qualitative behavior in our policy analysis."
                    },
                    {
                        "title": "When Do Graph Neural Networks Help with Node Classification: Investigating the Homophily Principle on Node Distinguishability",
                        "abstract": "Homophily principle, i.e., nodes with the same labels are more likely to be connected, was believed to be the main reason for the performance superiority of Graph Neural Networks (GNNs) over Neural Networks (NNs) on Node Classi\ufb01cation (NC) tasks. Recently, people have developed theoretical results arguing that, even though the homophily principle is broken, the advantage of GNNs can still hold as long as nodes from the same class share similar neighborhood patterns [29], which questions the validity of homophily. However, this argument only considers intra-class Node Distinguishability (ND) and ignores inter-class ND, which is insuf\ufb01cient to study the effect of homophily. In this paper, we \ufb01rst demonstrate the aforementioned insuf\ufb01ciency with examples and argue that an ideal situation for ND is to have smaller intra-class ND than inter-class ND. To formulate this idea and have a better understanding of homophily, we propose Contextual Stochastic Block Model for Homophily (CSBM-H) and de\ufb01ne two metrics, Probabilistic Bayes Error (PBE) and Expected Negative KL-divergence (ENKL), to quantify ND, through which we can also \ufb01nd how intra- and inter-class ND in\ufb02uence ND together. We visualize the results and give detailed analysis. Through experiments, we veri\ufb01ed that the superiority of GNNs is"
                    },
                    {
                        "title": "A Definition of Continual Reinforcement Learning",
                        "abstract": "In a standard view of the reinforcement learning problem, an agent's goal is to efficiently identify a policy that maximizes long-term reward. However, this perspective is based on a restricted view of learning as finding a solution, rather than treating learning as endless adaptation. In contrast, continual reinforcement learning refers to the setting in which the best agents never stop learning. Despite the importance of continual reinforcement learning, the community lacks a simple definition of the problem that highlights its commitments and makes its primary concepts precise and clear. To this end, this paper is dedicated to carefully defining the continual reinforcement learning problem. We formalize the notion of agents that\"never stop learning\"through a new mathematical language for analyzing and cataloging agents. Using this new language, we define a continual learning agent as one that can be understood as carrying out an implicit search process indefinitely, and continual reinforcement learning as the setting in which the best agents are all continual learning agents. We provide two motivating examples, illustrating that traditional views of multi-task reinforcement learning and continual supervised learning are special cases of our definition. Collectively, these definitions and perspectives formalize many intuitive concepts at the heart of learning, and open new research pathways surrounding continual learning agents."
                    },
                    {
                        "title": "The Stable Entropy Hypothesis and Entropy-Aware Decoding: An Analysis and Algorithm for Robust Natural Language Generation",
                        "abstract": "State-of-the-art language generation models can degenerate when applied to open-ended generation problems such as text completion, story generation, or dialog modeling. This degeneration usually shows up in the form of incoherence, lack of vocabulary diversity, and self-repetition or copying from the context. In this paper, we postulate that ``human-like'' generations usually lie in a narrow and nearly flat entropy band, and violation of these entropy bounds correlates with degenerate behavior. Our experiments show that this stable narrow entropy zone exists across models, tasks, and domains and confirm the hypothesis that violations of this zone correlate with degeneration. We then use this insight to propose an entropy-aware decoding algorithm that respects these entropy bounds resulting in less degenerate, more contextual, and\"human-like\"language generation in open-ended text generation settings."
                    },
                    {
                        "title": "Policy composition in reinforcement learning via multi-objective policy optimization",
                        "abstract": "We enable reinforcement learning agents to learn successful behavior policies by utilizing relevant pre-existing teacher policies. The teacher policies are introduced as objectives, in addition to the task objective, in a multi-objective policy optimization setting. Using the Multi-Objective Maximum a Posteriori Policy Optimization algorithm (Abdolmaleki et al. 2020), we show that teacher policies can help speed up learning, particularly in the absence of shaping rewards. In two domains with continuous observation and action spaces, our agents successfully compose teacher policies in sequence and in parallel, and are also able to further extend the policies of the teachers in order to solve the task. Depending on the specified combination of task and teacher(s), teacher(s) may naturally act to limit the final performance of an agent. The extent to which agents are required to adhere to teacher policies are determined by hyperparameters which determine both the effect of teachers on learning speed and the eventual performance of the agent on the task. In the humanoid domain (Tassa et al. 2018), we also equip agents with the ability to control the selection of teachers. With this ability, agents are able to meaningfully compose from the teacher policies to achieve a superior task reward on the walk task than in cases without access to the teacher policies. We show the resemblance of composed task policies with the corresponding teacher policies through videos."
                    },
                    {
                        "title": "When Do We Need GNN for Node Classification?",
                        "abstract": "Graph Neural Networks (GNNs) extend basic Neural Networks (NNs) by additionally making use of graph structure based on the relational inductive bias (edge bias), rather than treating the nodes as collections of independent and identically distributed (\\iid) samples. Though GNNs are believed to outperform basic NNs in real-world tasks, it is found that in some cases, GNNs have little performance gain or even underperform graph-agnostic NNs. To identify these cases, based on graph signal processing and statistical hypothesis testing, we propose two measures which analyze the cases in which the edge bias in features and labels does not provide advantages. Based on the measures, a threshold value can be given to predict the potential performance advantages of graph-aware models over graph-agnostic models."
                    },
                    {
                        "title": "Towards Safe Mechanical Ventilation Treatment Using Deep Offline Reinforcement Learning",
                        "abstract": "Mechanical ventilation is a key form of life support for patients with pulmonary impairment. Healthcare workers are required to continuously adjust ventilator settings for each patient, a challenging and time consuming task. Hence, it would be beneficial to develop an automated decision support tool to optimize ventilation treatment. We present DeepVent, a Conservative Q-Learning (CQL) based offline Deep Reinforcement Learning (DRL) agent that learns to predict the optimal ventilator parameters for a patient to promote 90 day survival. We design a clinically relevant intermediate reward that encourages continuous improvement of the patient vitals as well as addresses the challenge of sparse reward in RL. We find that DeepVent recommends ventilation parameters within safe ranges, as outlined in recent clinical trials. The CQL algorithm offers additional safety by mitigating the overestimation of the value estimates of out-of-distribution states/actions. We evaluate our agent using Fitted Q Evaluation (FQE) and demonstrate that it outperforms physicians from the MIMIC-III dataset."
                    },
                    {
                        "title": "Continuous MDP Homomorphisms and Homomorphic Policy Gradient",
                        "abstract": "Abstraction has been widely studied as a way to improve the efficiency and generalization of reinforcement learning algorithms. In this paper, we study abstraction in the continuous-control setting. We extend the definition of MDP homomorphisms to encompass continuous actions in continuous state spaces. We derive a policy gradient theorem on the abstract MDP, which allows us to leverage approximate symmetries of the environment for policy optimization. Based on this theorem, we propose an actor-critic algorithm that is able to learn the policy and the MDP homomorphism map simultaneously, using the lax bisimulation metric. We demonstrate the effectiveness of our method on benchmark tasks in the DeepMind Control Suite. Our method's ability to utilize MDP homomorphisms for representation learning leads to improved performance when learning from pixel observations."
                    },
                    {
                        "title": "Learning how to Interact with a Complex Interface using Hierarchical Reinforcement Learning",
                        "abstract": "Hierarchical Reinforcement Learning (HRL) allows interactive agents to decompose complex problems into a hierarchy of sub-tasks. Higher-level tasks can invoke the solutions of lower-level tasks as if they were primitive actions. In this work, we study the utility of hierarchical decompositions for learning an appropriate way to interact with a complex interface. Specifically, we train HRL agents that can interface with applications in a simulated Android device. We introduce a Hierarchical Distributed Deep Reinforcement Learning architecture that learns (1) subtasks corresponding to simple finger gestures, and (2) how to combine these gestures to solve several Android tasks. Our approach relies on goal conditioning and can be used more generally to convert any base RL agent into an HRL agent. We use the AndroidEnv environment to evaluate our approach. For the experiments, the HRL agent uses a distributed version of the popular DQN algorithm to train different components of the hierarchy. While the native action space is completely intractable for simple DQN agents, our architecture can be used to establish an effective way to interact with different tasks, significantly improving the performance of the same DQN agent over different levels of abstraction."
                    },
                    {
                        "title": "Behind the Machine's Gaze: Neural Networks with Biologically-inspired Constraints Exhibit Human-like Visual Attention",
                        "abstract": "By and large, existing computational models of visual attention tacitly assume perfect vision and full access to the stimulus and thereby deviate from foveated biological vision. Moreover, modeling top-down attention is generally reduced to the integration of semantic features without incorporating the signal of a high-level visual tasks that have been shown to partially guide human attention. We propose the Neural Visual Attention (NeVA) algorithm to generate visual scanpaths in a top-down manner. With our method, we explore the ability of neural networks on which we impose a biologically-inspired foveated vision constraint to generate human-like scanpaths without directly training for this objective. The loss of a neural network performing a downstream visual task (i.e., classification or reconstruction) flexibly provides top-down guidance to the scanpath. Extensive experiments show that our method outperforms state-of-the-art unsupervised human attention models in terms of similarity to human scanpaths. Additionally, the flexibility of the framework allows to quantitatively investigate the role of different tasks in the generated visual behaviors. Finally, we demonstrate the superiority of the approach in a novel experiment that investigates the utility of scanpaths in real-world applications, where imperfect viewing conditions are given."
                    }
                ]
            },
            "8fce9700-b8d0-43f1-9260-1c393d4bef87": {
                "pk": "8fce9700-b8d0-43f1-9260-1c393d4bef87",
                "name": "Anima Anandkumar",
                "collaborators": [
                    "K. Azizzadenesheli",
                    "Weili Nie",
                    "Arash Vahdat",
                    "Zong-Yi Li",
                    "Nikola B. Kovachki",
                    "M. O'Connell",
                    "Guanya Shi",
                    "Xichen Shi",
                    "Yisong Yue",
                    "Soon-Jo Chung"
                ],
                "domain": [
                    "Machine Learning",
                    "Control Systems",
                    "Diffusion Models",
                    "AI for Science"
                ],
                "publications": [
                    {
                        "title": "Deploying Neural-Fly in the Field",
                        "abstract": "\u2014Neural-Fly is a learning-based control method that allows rapid online learning by incorporating a pre-trained representations of unmodeled aerodynamics with adaptive control. Not is Neural-Fly robust to different tasks, environments, and drones, but also it builds off a standard control architecture, creating a straightforward way to begin integrating learning-based control into safety critical applications. In this workshop, we will present the latest results for Neural-Fly on field robots, discuss implementation considerations, and show case other applications of Neural-Fly, such as fault tolerant control."
                    },
                    {
                        "title": "DensePure: Understanding Diffusion Models for Adversarial Robustness",
                        "abstract": "Diffusion models have been recently employed to improve certified robustness through the process of denoising. However, the theoretical understanding of why diffusion models are able to improve the certified robustness is still lacking, preventing from further improvement. In this study, we close this gap by analyzing the fundamental properties of diffusion models and establishing the conditions under which they can enhance certified robustness. This deeper understanding allows us to propose a new method DensePure , designed to improve the certified robustness of a pretrained model (i.e. classifier). Given an (adversarial) input, DensePure consists of multiple runs of denoising via the reverse process of the diffusion model (with different random seeds) to get multiple reversed samples, which are then passed through the classifier, followed by majority voting of inferred labels to make the final prediction. This design of using multiple runs of denoising is informed by our theoretical analysis of the conditional distribution of the reversed sample. Specifically, when the data density of a clean sample is high, its conditional density under the reverse process in a diffusion model is also high; thus sampling from the latter conditional distribution can purify the adversarial example and return the corresponding clean sample with a high probability. By using the highest density point in the conditional distribution as the reversed sample, we identify the robust region of a given instance under the diffusion model\u2019s reverse process. We show that this robust region is a union of multiple convex sets, and is potentially much larger than the robust regions identified in previous works. In practice, DensePure can approximate the label of the high density region in the conditional distribution so that it can enhance certified robustness. We conduct extensive experiments to demonstrate the effectiveness of DensePure by evaluating its certified robustness given a standard model via randomized smoothing. We show that DensePure is consistently better than existing methods on ImageNet, with 7% improvement on average. Project"
                    },
                    {
                        "title": "Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems",
                        "abstract": "Advances in artificial intelligence (AI) are fueling a new paradigm of discoveries in natural sciences. Today, AI has started to advance natural sciences by improving, accelerating, and enabling our understanding of natural phenomena at a wide range of spatial and temporal scales, giving rise to a new area of research known as AI for science (AI4Science). Being an emerging research paradigm, AI4Science is unique in that it is an enormous and highly interdisciplinary area. Thus, a unified and technical treatment of this field is needed yet challenging. This work aims to provide a technically thorough account of a subarea of AI4Science; namely, AI for quantum, atomistic, and continuum systems. These areas aim at understanding the physical world from the subatomic (wavefunctions and electron density), atomic (molecules, proteins, materials, and interactions), to macro (fluids, climate, and subsurface) scales and form an important subarea of AI4Science. A unique advantage of focusing on these areas is that they largely share a common set of challenges, thereby allowing a unified and foundational treatment. A key common challenge is how to capture physics first principles, especially symmetries, in natural systems by deep learning methods. We provide an in-depth yet intuitive account of techniques to achieve equivariance to symmetry transformations. We also discuss other common technical challenges, including explainability, out-of-distribution generalization, knowledge transfer with foundation and large language models, and uncertainty quantification. To facilitate learning and education, we provide categorized lists of resources that we found to be useful. We strive to be thorough and unified and hope this initial effort may trigger more community interests and efforts to further advance AI4Science."
                    },
                    {
                        "title": "I2SB: Image-to-Image Schr\u00f6dinger Bridge",
                        "abstract": "We propose Image-to-Image Schr\\\"odinger Bridge (I$^2$SB), a new class of conditional diffusion models that directly learn the nonlinear diffusion processes between two given distributions. These diffusion bridges are particularly useful for image restoration, as the degraded images are structurally informative priors for reconstructing the clean images. I$^2$SB belongs to a tractable class of Schr\\\"odinger bridge, the nonlinear extension to score-based models, whose marginal distributions can be computed analytically given boundary pairs. This results in a simulation-free framework for nonlinear diffusions, where the I$^2$SB training becomes scalable by adopting practical techniques used in standard diffusion models. We validate I$^2$SB in solving various image restoration tasks, including inpainting, super-resolution, deblurring, and JPEG restoration on ImageNet 256x256 and show that I$^2$SB surpasses standard conditional diffusion models with more interpretable generative processes. Moreover, I$^2$SB matches the performance of inverse methods that additionally require the knowledge of the corruption operators. Our work opens up new algorithmic opportunities for developing efficient nonlinear diffusion models on a large scale. scale. Project page and codes: https://i2sb.github.io/"
                    },
                    {
                        "title": "Incrementally-Computable Neural Networks: Efficient Inference for Dynamic Inputs",
                        "abstract": "Deep learning often faces the challenge of efficiently processing dynamic inputs, such as sensor data or user inputs. For example, an AI writing assistant is required to update its suggestions in real time as a document is edited. Re-running the model each time is expensive, even with compression techniques like knowledge distillation, pruning, or quantization. Instead, we take an incremental computing approach, looking to reuse calculations as the inputs change. However, the dense connectivity of conventional architectures poses a major obstacle to incremental computation, as even minor input changes cascade through the network and restrict information reuse. To address this, we use vector quantization to discretize intermediate values in the network, which filters out noisy and unnecessary modifications to hidden neurons, facilitating the reuse of their values. We apply this approach to the transformers architecture, creating an efficient incremental inference algorithm with complexity proportional to the fraction of the modified inputs. Our experiments with adapting the OPT-125M pre-trained language model demonstrate comparable accuracy on document classification while requiring 12.1X (median) fewer operations for processing sequences of atomic edits."
                    },
                    {
                        "title": "Symmetry-Informed Geometric Representation for Molecules, Proteins, and Crystalline Materials",
                        "abstract": "Artificial intelligence for scientific discovery has recently generated significant interest within the machine learning and scientific communities, particularly in the domains of chemistry, biology, and material discovery. For these scientific problems, molecules serve as the fundamental building blocks, and machine learning has emerged as a highly effective and powerful tool for modeling their geometric structures. Nevertheless, due to the rapidly evolving process of the field and the knowledge gap between science (e.g., physics, chemistry,&biology) and machine learning communities, a benchmarking study on geometrical representation for such data has not been conducted. To address such an issue, in this paper, we first provide a unified view of the current symmetry-informed geometric methods, classifying them into three main categories: invariance, equivariance with spherical frame basis, and equivariance with vector frame basis. Then we propose a platform, coined Geom3D, which enables benchmarking the effectiveness of geometric strategies. Geom3D contains 16 advanced symmetry-informed geometric representation models and 14 geometric pretraining methods over 46 diverse datasets, including small molecules, proteins, and crystalline materials. We hope that Geom3D can, on the one hand, eliminate barriers for machine learning researchers interested in exploring scientific problems; and, on the other hand, provide valuable guidance for researchers in computational chemistry, structural biology, and materials science, aiding in the informed selection of representation techniques for specific applications."
                    },
                    {
                        "title": "Geometry-Informed Neural Operator for Large-Scale 3D PDEs",
                        "abstract": "We propose the geometry-informed neural operator (GINO), a highly efficient approach to learning the solution operator of large-scale partial differential equations with varying geometries. GINO uses a signed distance function and point-cloud representations of the input shape and neural operators based on graph and Fourier architectures to learn the solution operator. The graph neural operator handles irregular grids and transforms them into and from regular latent grids on which Fourier neural operator can be efficiently applied. GINO is discretization-convergent, meaning the trained model can be applied to arbitrary discretization of the continuous domain and it converges to the continuum operator as the discretization is refined. To empirically validate the performance of our method on large-scale simulation, we generate the industry-standard aerodynamics dataset of 3D vehicle geometries with Reynolds numbers as high as five million. For this large-scale 3D fluid simulation, numerical methods are expensive to compute surface pressure. We successfully trained GINO to predict the pressure on car surfaces using only five hundred data points. The cost-accuracy experiments show a $26,000 \\times$ speed-up compared to optimized GPU-based computational fluid dynamics (CFD) simulators on computing the drag coefficient. When tested on new combinations of geometries and boundary conditions (inlet velocities), GINO obtains a one-fourth reduction in error rate compared to deep neural network approaches."
                    },
                    {
                        "title": "ClimSim-Online: A Large Multi-scale Dataset and Framework for Hybrid ML-physics Climate Emulation",
                        "abstract": "Modern climate projections lack adequate spatial and temporal resolution due to computational constraints, leading to inaccuracies in representing critical processes like thunderstorms that occur on the sub-resolution scale. Hybrid methods combining physics with machine learning (ML) offer faster, higher fidelity climate simulations by outsourcing compute-hungry, high-resolution simulations to ML emulators. However, these hybrid ML-physics simulations require domain-specific data and workflows that have been inaccessible to many ML experts. As an extension of the ClimSim dataset (Yu et al., 2024), we present ClimSim-Online, which also includes an end-to-end workflow for developing hybrid ML-physics simulators. The ClimSim dataset includes 5.7 billion pairs of multivariate input/output vectors, capturing the influence of high-resolution, high-fidelity physics on a host climate simulator's macro-scale state. The dataset is global and spans ten years at a high sampling frequency. We provide a cross-platform, containerized pipeline to integrate ML models into operational climate simulators for hybrid testing. We also implement various ML baselines, alongside a hybrid baseline simulator, to highlight the ML challenges of building stable, skillful emulators. The data (https://huggingface.co/datasets/LEAP/ClimSim_high-res) and code (https://leap-stc.github.io/ClimSim and https://github.com/leap-stc/climsim-online) are publicly released to support the development of hybrid ML-physics and high-fidelity climate simulations."
                    },
                    {
                        "title": "Fourier Neural Operator for Plasma Modelling",
                        "abstract": "Predicting plasma evolution within a Tokamak is crucial to building a sustainable fusion reactor. Whether in the simulation space or within the experimental domain, the capability to forecast the spatio-temporal evolution of plasma field variables rapidly and accurately could improve active control methods on current tokamak devices and future fusion reactors. In this work, we demonstrate the utility of using Fourier Neural Operator (FNO) to model the plasma evolution in simulations and experiments. Our work shows that the FNO is capable of predicting magnetohydrodynamic models governing the plasma dynamics, 6 orders of magnitude faster than the traditional numerical solver, while maintaining considerable accuracy (NMSE $\\sim 10^{-5})$. Our work also benchmarks the performance of the FNO against other standard surrogate models such as Conv-LSTM and U-Net and demonstrate that the FNO takes significantly less time to train, requires less parameters and outperforms other models. We extend the FNO approach to model the plasma evolution observed by the cameras positioned within the MAST spherical tokamak. We illustrate its capability in forecasting the formation of filaments within the plasma as well as the heat deposits. The FNO deployed to model the camera is capable of forecasting the full length of the plasma shot within half the time of the shot duration."
                    },
                    {
                        "title": "Score-based Diffusion Models in Function Space",
                        "abstract": "Diffusion models have recently emerged as a powerful framework for generative modeling. They consist of a forward process that perturbs input data with Gaussian white noise and a reverse process that learns a score function to generate samples by denoising. Despite their tremendous success, they are mostly formulated on finite-dimensional spaces, e.g. Euclidean, limiting their applications to many domains where the data has a functional form such as in scientific computing and 3D geometric data analysis. In this work, we introduce a mathematically rigorous framework called Denoising Diffusion Operators (DDOs) for training diffusion models in function space. In DDOs, the forward process perturbs input functions gradually using a Gaussian process. The generative process is formulated by integrating a function-valued Langevin dynamic. Our approach requires an appropriate notion of the score for the perturbed data distribution, which we obtain by generalizing denoising score matching to function spaces that can be infinite-dimensional. We show that the corresponding discretized algorithm generates accurate samples at a fixed cost that is independent of the data resolution. We theoretically and numerically verify the applicability of our approach on a set of problems, including generating solutions to the Navier-Stokes equation viewed as the push-forward distribution of forcings from a Gaussian Random Field (GRF)."
                    },
                    {
                        "title": "InRank: Incremental Low-Rank Learning",
                        "abstract": "The theory of greedy low-rank learning (GLRL) aims to explain the impressive generalization capabilities of deep learning. It proves that stochastic gradient-based training implicitly regularizes neural networks towards low-rank solutions through a gradual increase of the rank during training. However, there is a gap between theory and practice since GLRL requires an infinitesimal initialization of the weights, which is not practical due to the fact that it is a saddle point. In this work, we remove the assumption of infinitesimal initialization by focusing on cumulative weight updates. We prove the cumulative weight updates follow an incremental low-rank trajectory for arbitrary orthogonal initialization of weights in a three-layer linear network. Empirically, we demonstrate that our theory holds on a broad range of neural networks (e.g., transformers) and standard training algorithms (e.g., SGD, Adam). However, existing training algorithms do not exploit the low-rank property to improve computational efficiency as the networks are not parameterized in low-rank. To remedy this, we design a new training algorithm Incremental Low-Rank Learning (InRank), which explicitly expresses cumulative weight updates as low-rank matrices while incrementally augmenting their ranks during training. We evaluate InRank on GPT-2, and our results indicate that InRank achieves comparable prediction performance as the full-rank counterpart while requiring at most 33% of the total ranks throughout training. We also propose an efficient version of InRank that achieves a reduction of 37% in total training time and 36% in model size when training GPT-medium on WikiText-103 from scratch."
                    },
                    {
                        "title": "LeanDojo: Theorem Proving with Retrieval-Augmented Language Models",
                        "abstract": "Large language models (LLMs) have shown promise in proving formal theorems using proof assistants such as Lean. However, existing methods are difficult to reproduce or build on, due to private code, data, and large compute requirements. This has created substantial barriers to research on machine learning methods for theorem proving. This paper removes these barriers by introducing LeanDojo: an open-source Lean playground consisting of toolkits, data, models, and benchmarks. LeanDojo extracts data from Lean and enables interaction with the proof environment programmatically. It contains fine-grained annotations of premises in proofs, providing valuable data for premise selection: a key bottleneck in theorem proving. Using this data, we develop ReProver (Retrieval-Augmented Prover): an LLM-based prover augmented with retrieval for selecting premises from a vast math library. It is inexpensive and needs only one GPU week of training. Our retriever leverages LeanDojo's program analysis capability to identify accessible premises and hard negative examples, which makes retrieval much more effective. Furthermore, we construct a new benchmark consisting of 98,734 theorems and proofs extracted from Lean's math library. It features challenging data split requiring the prover to generalize to theorems relying on novel premises that are never used in training. We use this benchmark for training and evaluation, and experimental results demonstrate the effectiveness of ReProver over non-retrieval baselines and GPT-4. We thus provide the first set of open-source LLM-based theorem provers without any proprietary datasets and release it under a permissive MIT license to facilitate further research."
                    },
                    {
                        "title": "BiasTestGPT: Using ChatGPT for Social Bias Testing of Language Models",
                        "abstract": "Pretrained Language Models (PLMs) harbor inherent social biases that can result in harmful real-world implications. Such social biases are measured through the probability values that PLMs output for different social groups and attributes appearing in a set of test sentences. However, bias testing is currently cumbersome since the test sentences are generated either from a limited set of manual templates or need expensive crowd-sourcing. We instead propose using ChatGPT for the controllable generation of test sentences, given any arbitrary user-specified combination of social groups and attributes appearing in the test sentences. When compared to template-based methods, our approach using ChatGPT for test sentence generation is superior in detecting social bias, especially in challenging settings such as intersectional biases. We present an open-source comprehensive bias testing framework (BiasTestGPT), hosted on HuggingFace, that can be plugged into any open-source PLM for bias testing. User testing with domain experts from various fields has shown their interest in being able to test modern AI for social biases. Our tool has significantly improved their awareness of such biases in PLMs, proving to be learnable and user-friendly. We thus enable seamless open-ended social bias testing of PLMs by domain experts through an automatic large-scale generation of diverse test sentences for any combination of social categories and attributes."
                    },
                    {
                        "title": "Fast Training of Diffusion Models with Masked Transformers",
                        "abstract": "We propose an efficient approach to train large diffusion models with masked transformers. While masked transformers have been extensively explored for representation learning, their application to generative learning is less explored in the vision domain. Our work is the first to exploit masked training to reduce the training cost of diffusion models significantly. Specifically, we randomly mask out a high proportion (e.g., 50%) of patches in diffused input images during training. For masked training, we introduce an asymmetric encoder-decoder architecture consisting of a transformer encoder that operates only on unmasked patches and a lightweight transformer decoder on full patches. To promote a long-range understanding of full patches, we add an auxiliary task of reconstructing masked patches to the denoising score matching objective that learns the score of unmasked patches. Experiments on ImageNet-256x256 and ImageNet-512x512 show that our approach achieves competitive and even better generative performance than the state-of-the-art Diffusion Transformer (DiT) model, using only around 30% of its original training time. Thus, our method shows a promising way of efficiently training large transformer-based diffusion models without sacrificing the generative performance."
                    },
                    {
                        "title": "Neural Operators for Solving PDEs and Inverse Design",
                        "abstract": "Deep learning surrogate models have shown promise in modeling complex physical phenomena such as photonics, fluid flows, molecular dynamics and material properties. However, standard neural networks assume finite-dimensional inputs and outputs, and hence, cannot withstand a change in resolution or discretization between training and testing. We introduce Fourier neural operators that can learn operators, which are mappings between infinite dimensional spaces. They are discretization-invariant and can generalize beyond the discretization or resolution of training data. They can efficiently solve partial differential equations (PDEs) on general geometries. We consider a variety of PDEs for both forward modeling and inverse design problems, as well as show practical gains in the lithography domain."
                    },
                    {
                        "title": "Tipping Point Forecasting in Non-Stationary Dynamics on Function Spaces",
                        "abstract": "Tipping points are abrupt, drastic, and often irreversible changes in the evolution of non-stationary and chaotic dynamical systems. For instance, increased greenhouse gas concentrations are predicted to lead to drastic decreases in low cloud cover, referred to as a climatological tipping point. In this paper, we learn the evolution of such non-stationary dynamical systems using a novel recurrent neural operator (RNO), which learns mappings between function spaces. After training RNO on only the pre-tipping dynamics, we employ it to detect future tipping points using an uncertainty-based approach. In particular, we propose a conformal prediction framework to forecast tipping points by monitoring deviations from physics constraints (such as conserved quantities and partial differential equations), enabling forecasting of these abrupt changes along with a rigorous measure of uncertainty. We illustrate our proposed methodology on non-stationary ordinary and partial differential equations, such as the Lorenz-63 and Kuramoto-Sivashinsky equations. We also apply our methods to forecast a climate tipping point in stratocumulus cloud cover. In our experiments, we demonstrate that even partial or approximate physics constraints can be used to accurately forecast future tipping points."
                    },
                    {
                        "title": "Pretraining with Neural-Fly for Rapid Online Learning",
                        "abstract": "\u2014Executing safe and precise flight maneuvers in dynamic high-speed winds is important for the ongoing commodi-tization of uninhabited aerial vehicles (UAVs). However, since the relationship between various wind conditions and its effect on aircraft maneuverability is not well understood, it is challenging to design effective robot controllers using traditional control de-sign methods. We present Neural-Fly, a learning-based approach that allows rapid online adaptation by incorporating pre-trained representations through deep learning. Neural-Fly builds on two key observations that aerodynamics in different wind conditions share a common representation and that the wind-specific part lies in a low-dimensional space. To that end, Neural-Fly uses a proposed learning algorithm, Domain Adversarially Invariant Meta-Learning (DAIML), to learn the shared representation, only using 12 minutes of flight data. This pretraining phase enables rapid online learning through a composite adaptation law, which only needs to update a set of linear coefficients for mixing the basis elements to effectively correct for the wind effects. When evaluated under challenging wind conditions generated with the Caltech Real Weather Wind Tunnel with wind speeds up to 43 . 6km / h ( 12 . 1m / s ), Neural-Fly achieves precise flight control with substantially smaller tracking error than state-of-the-art nonlinear and adaptive controllers. In addition to strong empirical performance, the exponential stability of Neural-Fly results in robustness guarantees. Finally, our control design extrapolates to unseen wind conditions, is shown to be effective for outdoor flights with only on-board sensors, and can transfer across drones with minimal performance degradation."
                    },
                    {
                        "title": "Forecasting subcritical cylinder wakes with Fourier Neural Operators",
                        "abstract": "We apply Fourier neural operators (FNOs), a state-of-the-art operator learning technique, to forecast the temporal evolution of experimentally measured velocity fields. FNOs are a recently developed machine learning method capable of approximating solution operators to systems of partial differential equations through data alone. The learned FNO solution operator can be evaluated in milliseconds, potentially enabling faster-than-real-time modeling for predictive flow control in physical systems. Here we use FNOs to predict how physical fluid flows evolve in time, training with particle image velocimetry measurements depicting cylinder wakes in the subcritical vortex shedding regime. We train separate FNOs at Reynolds numbers ranging from Re = 240 to Re = 3060 and study how increasingly turbulent flow phenomena impact prediction accuracy. We focus here on a short prediction horizon of ten non-dimensionalized time-steps, as would be relevant for problems of predictive flow control. We find that FNOs are capable of accurately predicting the evolution of experimental velocity fields throughout the range of Reynolds numbers tested (L2 norm error<0.1) despite being provided with limited and imperfect flow observations. Given these results, we conclude that this method holds significant potential for real-time predictive flow control of physical systems."
                    },
                    {
                        "title": "Shall We Pretrain Autoregressive Language Models with Retrieval? A Comprehensive Study",
                        "abstract": "Large decoder-only language models (LMs) can be largely improved in terms of perplexity by retrieval (e.g., RETRO), but its impact on text generation quality and downstream task accuracy is unclear. Thus, it is still an open question: shall we pretrain large autoregressive LMs with retrieval? To answer it, we perform a comprehensive study on a scalable pre-trained retrieval-augmented LM (i.e., RETRO) compared with standard GPT and retrieval-augmented GPT incorporated at fine-tuning or inference stages. We first provide the recipe to reproduce RETRO up to 9.5B parameters while retrieving a text corpus with 330B tokens. Based on that, we have the following novel findings: i) RETRO outperforms GPT on text generation with much less degeneration (i.e., repetition), moderately higher factual accuracy, and slightly lower toxicity with a nontoxic retrieval database. ii) On the LM Evaluation Harness benchmark, RETRO largely outperforms GPT on knowledge-intensive tasks, but is on par with GPT on other tasks. Furthermore, we introduce a simple variant of the model, RETRO++, which largely improves open-domain QA results of original RETRO (e.g., EM score +8.6 on Natural Question) and significantly outperforms retrieval-augmented GPT in both fine-tuning and zero-shot evaluation settings. Our findings highlight the promising direction of pretraining autoregressive LMs with retrieval as future foundation models. We release our code and model at: https://github.com/NVIDIA/Megatron-LM/blob/main/tools/retro/README.md"
                    }
                ]
            },
            "8935a000-778c-4773-96ba-753acb897d34": {
                "pk": "8935a000-778c-4773-96ba-753acb897d34",
                "name": "Kamyar Azizzadenesheli",
                "collaborators": [
                    "A. Anandkumar",
                    "Jean Kossaifi",
                    "Anima Anandkumar",
                    "Julius Berner",
                    "B. Bonev",
                    "Zachary E. Ross",
                    "Minkai Xu",
                    "J. Leskovec",
                    "Stefano Ermon",
                    "Zong-Yi Li"
                ],
                "domain": [
                    "Neural Operator",
                    "Machine Learning",
                    "Partial Differential Equations",
                    "Uncertainty Quantification"
                ],
                "publications": [
                    {
                        "title": "Neural Operators with Localized Integral and Differential Kernels",
                        "abstract": "Neural operators learn mappings between function spaces, which is practical for learning solution operators of PDEs and other scientific modeling applications. Among them, the Fourier neural operator (FNO) is a popular architecture that performs global convolutions in the Fourier space. However, such global operations are often prone to over-smoothing and may fail to capture local details. In contrast, convolutional neural networks (CNN) can capture local features but are limited to training and inference at a single resolution. In this work, we present a principled approach to operator learning that can capture local features under two frameworks by learning differential operators and integral operators with locally supported kernels. Specifically, inspired by stencil methods, we prove that we obtain differential operators under an appropriate scaling of the kernel values of CNNs. To obtain local integral operators, we utilize suitable basis representations for the kernels based on discrete-continuous convolutions. Both these approaches preserve the properties of operator learning and, hence, the ability to predict at any resolution. Adding our layers to FNOs significantly improves their performance, reducing the relative L2-error by 34-72% in our experiments, which include a turbulent 2D Navier-Stokes and the spherical shallow water equations."
                    },
                    {
                        "title": "Equivariant Graph Neural Operator for Modeling 3D Dynamics",
                        "abstract": "Modeling the complex three-dimensional (3D) dynamics of relational systems is an important problem in the natural sciences, with applications ranging from molecular simulations to particle mechanics. Machine learning methods have achieved good success by learning graph neural networks to model spatial interactions. However, these approaches do not faithfully capture temporal correlations since they only model next-step predictions. In this work, we propose Equivariant Graph Neural Operator (EGNO), a novel and principled method that directly models dynamics as trajectories instead of just next-step prediction. Different from existing methods, EGNO explicitly learns the temporal evolution of 3D dynamics where we formulate the dynamics as a function over time and learn neural operators to approximate it. To capture the temporal correlations while keeping the intrinsic SE(3)-equivariance, we develop equivariant temporal convolutions parameterized in the Fourier space and build EGNO by stacking the Fourier layers over equivariant networks. EGNO is the first operator learning framework that is capable of modeling solution dynamics functions over time while retaining 3D equivariance. Comprehensive experiments in multiple domains, including particle simulations, human motion capture, and molecular dynamics, demonstrate the significantly superior performance of EGNO against existing methods, thanks to the equivariant temporal modeling. Our code is available at https://github.com/MinkaiXu/egno."
                    },
                    {
                        "title": "Calibrated Uncertainty Quantification for Operator Learning via Conformal Prediction",
                        "abstract": "Operator learning has been increasingly adopted in scientific and engineering applications, many of which require calibrated uncertainty quantification. Since the output of operator learning is a continuous function, quantifying uncertainty simultaneously at all points in the domain is challenging. Current methods consider calibration at a single point or over one scalar function or make strong assumptions such as Gaussianity. We propose a risk-controlling quantile neural operator, a distribution-free, finite-sample functional calibration conformal prediction method. We provide a theoretical calibration guarantee on the coverage rate, defined as the expected percentage of points on the function domain whose true value lies within the predicted uncertainty ball. Empirical results on a 2D Darcy flow and a 3D car surface pressure prediction task validate our theoretical results, demonstrating calibrated coverage and efficient uncertainty bands outperforming baseline methods. In particular, on the 3D problem, our method is the only one that meets the target calibration percentage (percentage of test samples for which the uncertainty estimates are calibrated) of 98%."
                    },
                    {
                        "title": "Sparse Contextual CDF Regression",
                        "abstract": "Estimating cumulative distribution functions (CDFs) of context-dependent random variables is a central statistical task underpinning numerous applications in machine learning and economics. In this work, we extend a recent line of theoretical inquiry into this domain by analyzing the problem of sparse contextual CDF regression , wherein data points are sampled from a convex combination of s context-dependent CDFs chosen from a set of d basis functions. We show that adaptations of several canonical regression methods serve as tractable estimators in this functional sparse regression setting under standard assumptions on the conditioning of the basis functions. In particular, given n data samples, we prove estimation error upper bounds of \u02dc O ( p s/n ) for functional versions of the lasso and Dantzig selector estimators, and \u02dc O ( \u221a s/ 4 \u221a n ) for a functional version of the elastic net estimator. Our results match the corresponding error bounds for finite-dimensional regression and improve upon CDF ridge regression which has \u02dc O ( p d/n ) estimation error. Finally, we obtain a matching information-theoretic lower bound which establishes the minimax optimality of the lasso and Dantzig selector estimators up to logarithmic factors."
                    },
                    {
                        "title": "Pretraining Codomain Attention Neural Operators for Solving Multiphysics PDEs",
                        "abstract": "Existing neural operator architectures face challenges when solving multiphysics problems with coupled partial differential equations (PDEs), due to complex geometries, interactions between physical variables, and the lack of large amounts of high-resolution training data. To address these issues, we propose Codomain Attention Neural Operator (CoDA-NO), which tokenizes functions along the codomain or channel space, enabling self-supervised learning or pretraining of multiple PDE systems. Specifically, we extend positional encoding, self-attention, and normalization layers to the function space. CoDA-NO can learn representations of different PDE systems with a single model. We evaluate CoDA-NO's potential as a backbone for learning multiphysics PDEs over multiple systems by considering few-shot learning settings. On complex downstream tasks with limited data, such as fluid flow simulations and fluid-structure interactions, we found CoDA-NO to outperform existing methods on the few-shot learning task by over $36\\%$. The code is available at https://github.com/ashiq24/CoDA-NO."
                    },
                    {
                        "title": "Exploring the design space of deep-learning-based weather forecasting systems",
                        "abstract": "Despite tremendous progress in developing deep-learning-based weather forecasting systems, their design space, including the impact of different design choices, is yet to be well understood. This paper aims to fill this knowledge gap by systematically analyzing these choices including architecture, problem formulation, pretraining scheme, use of image-based pretrained models, loss functions, noise injection, multi-step inputs, additional static masks, multi-step finetuning (including larger stride models), as well as training on a larger dataset. We study fixed-grid architectures such as UNet, fully convolutional architectures, and transformer-based models, along with grid-invariant architectures, including graph-based and operator-based models. Our results show that fixed-grid architectures outperform grid-invariant architectures, indicating a need for further architectural developments in grid-invariant models such as neural operators. We therefore propose a hybrid system that combines the strong performance of fixed-grid models with the flexibility of grid-invariant architectures. We further show that multi-step fine-tuning is essential for most deep-learning models to work well in practice, which has been a common practice in the past. Pretraining objectives degrade performance in comparison to supervised training, while image-based pretrained models provide useful inductive biases in some cases in comparison to training the model from scratch. Interestingly, we see a strong positive effect of using a larger dataset when training a smaller model as compared to training on a smaller dataset for longer. Larger models, on the other hand, primarily benefit from just an increase in the computational budget. We believe that these results will aid in the design of better weather forecasting systems in the future."
                    },
                    {
                        "title": "Dynamical Measure Transport and Neural PDE Solvers for Sampling",
                        "abstract": "The task of sampling from a probability density can be approached as transporting a tractable density function to the target, known as dynamical measure transport. In this work, we tackle it through a principled unified framework using deterministic or stochastic evolutions described by partial differential equations (PDEs). This framework incorporates prior trajectory-based sampling methods, such as diffusion models or Schr\\\"odinger bridges, without relying on the concept of time-reversals. Moreover, it allows us to propose novel numerical methods for solving the transport task and thus sampling from complicated targets without the need for the normalization constant or data samples. We employ physics-informed neural networks (PINNs) to approximate the respective PDE solutions, implying both conceptional and computational advantages. In particular, PINNs allow for simulation- and discretization-free optimization and can be trained very efficiently, leading to significantly better mode coverage in the sampling task compared to alternative methods. Moreover, they can readily be fine-tuned with Gauss-Newton methods to achieve high accuracy in sampling."
                    },
                    {
                        "title": "Universal Functional Regression with Neural Operator Flows",
                        "abstract": "Regression on function spaces is typically limited to models with Gaussian process priors. We introduce the notion of universal functional regression, in which we aim to learn a prior distribution over non-Gaussian function spaces that remains mathematically tractable for functional regression. To do this, we develop Neural Operator Flows (OpFlow), an infinite-dimensional extension of normalizing flows. OpFlow is an invertible operator that maps the (potentially unknown) data function space into a Gaussian process, allowing for exact likelihood estimation of functional point evaluations. OpFlow enables robust and accurate uncertainty quantification via drawing posterior samples of the Gaussian process and subsequently mapping them into the data function space. We empirically study the performance of OpFlow on regression and generation tasks with data generated from Gaussian processes with known posterior forms and non-Gaussian processes, as well as real-world earthquake seismograms with an unknown closed-form distribution."
                    },
                    {
                        "title": "Timing as an Action: Learning When to Observe and Act",
                        "abstract": "In standard reinforcement learning setups, the agent receives observations and performs actions at evenly spaced intervals. However, in many real-world settings, observations are expensive, forcing agents to commit to courses of action for designated periods of time. Consider that doctors, after each visit, typically set not only a treatment plan but also a follow-up date at which that plan might be revised. In this work, we formalize the setup of timing-as-an-action . Through theoretical analysis in the tabular setting, we show that while the choice of delay intervals could be naively folded in as part of a composite action, these actions have a special structure and handling them intelligently yields statistical advantages. Taking a model-based perspective, these gains owe to the fact that delay actions do not add any parameters to the underlying model. For model estimation, we provide provable sample-efficiency improvements, and our experiments demonstrate empirical improvements in both healthcare simulators and classical reinforcement learning environments."
                    },
                    {
                        "title": "Deploying Neural-Fly in the Field",
                        "abstract": "\u2014Neural-Fly is a learning-based control method that allows rapid online learning by incorporating a pre-trained representations of unmodeled aerodynamics with adaptive control. Not is Neural-Fly robust to different tasks, environments, and drones, but also it builds off a standard control architecture, creating a straightforward way to begin integrating learning-based control into safety critical applications. In this workshop, we will present the latest results for Neural-Fly on field robots, discuss implementation considerations, and show case other applications of Neural-Fly, such as fault tolerant control."
                    },
                    {
                        "title": "Deep neural helmholtz operators for 3D elastic wave propagation and inversion",
                        "abstract": "  Numerical simulations of seismic wave propagation in heterogeneous 3D media are central to investigating subsurface structures and understanding earthquake processes, yet are computationally expensive for large problems. This is particularly problematic for full waveform inversion, which typically involves numerous runs of the forward process. In machine learning there has been considerable recent work in the area of operator learning, with a new class of models called neural operators allowing for data-driven solutions to partial differential equations. Recent works in seismology have shown that when neural operators are adequately trained, they can significantly shorten the compute time for wave propagation. However, the memory required for the 3D time domain equations\u00a0may be prohibitive. In this study, we show that these limitations can be overcome by solving the wave equations\u00a0in the frequency domain, also known as the Helmholtz equations, since the solutions for a set of frequencies can be determined in parallel. The 3D Helmholtz neural operator is 40\u00a0times more memory-efficient than an equivalent time-domain version. We employ a Helmholtz neural operator for 2D and 3D elastic wave modeling, achieving two orders of magnitude acceleration compared to a baseline spectral element method. The neural operator accurately generalizes to variable velocity structures and can be evaluated on denser input meshes than used in the training simulations. We also show that when solving for wavefields strictly on the surface, the accuracy can be significantly improved via a graph neural operator layer. In leveraging automatic differentiation, the proposed method can serve as an alternative to the adjoint-state approach for 3D full-waveform inversion, reducing the computation time by a factor of 350."
                    },
                    {
                        "title": "Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems",
                        "abstract": "Advances in artificial intelligence (AI) are fueling a new paradigm of discoveries in natural sciences. Today, AI has started to advance natural sciences by improving, accelerating, and enabling our understanding of natural phenomena at a wide range of spatial and temporal scales, giving rise to a new area of research known as AI for science (AI4Science). Being an emerging research paradigm, AI4Science is unique in that it is an enormous and highly interdisciplinary area. Thus, a unified and technical treatment of this field is needed yet challenging. This work aims to provide a technically thorough account of a subarea of AI4Science; namely, AI for quantum, atomistic, and continuum systems. These areas aim at understanding the physical world from the subatomic (wavefunctions and electron density), atomic (molecules, proteins, materials, and interactions), to macro (fluids, climate, and subsurface) scales and form an important subarea of AI4Science. A unique advantage of focusing on these areas is that they largely share a common set of challenges, thereby allowing a unified and foundational treatment. A key common challenge is how to capture physics first principles, especially symmetries, in natural systems by deep learning methods. We provide an in-depth yet intuitive account of techniques to achieve equivariance to symmetry transformations. We also discuss other common technical challenges, including explainability, out-of-distribution generalization, knowledge transfer with foundation and large language models, and uncertainty quantification. To facilitate learning and education, we provide categorized lists of resources that we found to be useful. We strive to be thorough and unified and hope this initial effort may trigger more community interests and efforts to further advance AI4Science."
                    },
                    {
                        "title": "Multi-Grid Tensorized Fourier Neural Operator for High-Resolution PDEs",
                        "abstract": "Memory complexity and data scarcity have so far prohibited learning solution operators of partial differential equations (PDEs) at high resolutions. We address these limitations by introducing a new data efficient and highly parallelizable operator learning approach with reduced memory requirement and better generalization, called multi-grid tensorized neural operator (MG-TFNO). MG-TFNO scales to large resolutions by leveraging local and global structures of full-scale, real-world phenomena, through a decomposition of both the input domain and the operator's parameter space. Our contributions are threefold: i) we enable parallelization over input samples with a novel multi-grid-based domain decomposition, ii) we represent the parameters of the model in a high-order latent subspace of the Fourier domain, through a global tensor factorization, resulting in an extreme reduction in the number of parameters and improved generalization, and iii) we propose architectural improvements to the backbone FNO. Our approach can be used in any operator learning setting. We demonstrate superior performance on the turbulent Navier-Stokes equations where we achieve less than half the error with over 150x compression. The tensorization combined with the domain decomposition, yields over 150x reduction in the number of parameters and 7x reduction in the domain size without losses in accuracy, while slightly enabling parallelism."
                    },
                    {
                        "title": "Dynamic Obstacle Avoidance for USVs Using Cross-Domain Deep Reinforcement Learning and Neural Network Model Predictive Controller",
                        "abstract": "This work presents a framework that allows Unmanned Surface Vehicles (USVs) to avoid dynamic obstacles through initial training on an Unmanned Ground Vehicle (UGV) and cross-domain retraining on a USV. This is achieved by integrating a Deep Reinforcement Learning (DRL) agent that generates high-level control commands and leveraging a neural network based model predictive controller (NN-MPC) to reach target waypoints and reject disturbances. A Deep Q Network (DQN) utilized in this framework is trained in a ground environment using a Turtlebot robot and retrained in a water environment using the BREAM USV in the Gazebo simulator to avoid dynamic obstacles. The network is then validated in both simulation and real-world tests. The cross-domain learning largely decreases the training time (28%) and increases the obstacle avoidance performance (70 more reward points) compared to pure water domain training. This methodology shows that it is possible to leverage the data-rich and accessible ground environments to train DRL agent in data-poor and difficult-to-access marine environments. This will allow rapid and iterative agent development without further training due to the change in environment or vehicle dynamics."
                    },
                    {
                        "title": "Broadband Ground Motion Synthesis via Generative Adversarial Neural Operators: Development and Validation",
                        "abstract": "  We present a data-driven framework for ground-motion synthesis that generates three-component acceleration time histories conditioned on moment magnitude (M), rupture distance (Rrup), time-average shear-wave velocity at the top 30\u00a0m (VS30), and style of faulting. We use a Generative Adversarial Neural Operator (GANO)\u2014a resolution invariant architecture that guarantees model training independent of the data sampling frequency. We first present the conditional ground-motion synthesis algorithm (cGM-GANO) and discuss its advantages compared to the previous work. We next train cGM-GANO on simulated ground motions generated by the Southern California Earthquake Center Broadband Platform (BBP) and on recorded the Kiban\u2013Kyoshin network (KiK-net) data, and show that the model can learn the overall magnitude, distance, and VS30 scaling of effective amplitude spectra (EAS) ordinates and pseudospectral accelerations (PSA). Results specifically show that cGM-GANO produces consistent median scaling with the training data for the corresponding tectonic environments over a wide range of frequencies for scenarios with sufficient data coverage. For the BBP dataset, cGM-GANO cannot learn the ground-motion scaling of the stochastic frequency components (f > 1\u00a0Hz); for the KiK-net dataset, the largest misfit is observed at short distances (Rrup<50\u2009\u2009km) and for soft-soil conditions (VS30<200\u2009\u2009m/s) due to the scarcity of such data. Except for these conditions, the aleatory variability of EAS and PSA are captured reasonably well. Finally, cGM-GANO produces similar median scaling to traditional ground-motion models (GMMs) for frequencies greater than 1\u00a0Hz for both PSA and EAS but underestimates the aleatory variability of EAS. Discrepancies in the comparisons between the synthetic ground motions and GMMs are attributed to inconsistencies between the training dataset and the datasets used in GMM development. Our pilot study demonstrates GANO\u2019s potential for efficient synthesis of broadband ground motions."
                    },
                    {
                        "title": "Geometry-Informed Neural Operator for Large-Scale 3D PDEs",
                        "abstract": "We propose the geometry-informed neural operator (GINO), a highly efficient approach to learning the solution operator of large-scale partial differential equations with varying geometries. GINO uses a signed distance function and point-cloud representations of the input shape and neural operators based on graph and Fourier architectures to learn the solution operator. The graph neural operator handles irregular grids and transforms them into and from regular latent grids on which Fourier neural operator can be efficiently applied. GINO is discretization-convergent, meaning the trained model can be applied to arbitrary discretization of the continuous domain and it converges to the continuum operator as the discretization is refined. To empirically validate the performance of our method on large-scale simulation, we generate the industry-standard aerodynamics dataset of 3D vehicle geometries with Reynolds numbers as high as five million. For this large-scale 3D fluid simulation, numerical methods are expensive to compute surface pressure. We successfully trained GINO to predict the pressure on car surfaces using only five hundred data points. The cost-accuracy experiments show a $26,000 \\times$ speed-up compared to optimized GPU-based computational fluid dynamics (CFD) simulators on computing the drag coefficient. When tested on new combinations of geometries and boundary conditions (inlet velocities), GINO obtains a one-fourth reduction in error rate compared to deep neural network approaches."
                    }
                ]
            }
        }
    },
    "2406.08398": {
        "paper_data": {
            "title": "cPAPERS: A Dataset of Situated and Multimodal Interactive Conversations in Scientific Papers",
            "url": "http://arxiv.org/abs/2406.08398v1",
            "arxiv_id": "2406.08398",
            "authors": [
                "Anirudh Sundar",
                "Jin Xu",
                "William Gay",
                "Christopher Richardson",
                "Larry Heck"
            ],
            "abstract": "An emerging area of research in situated and multimodal interactive conversations (SIMMC) includes interactions in scientific papers. Since scientific papers are primarily composed of text, equations, figures, and tables, SIMMC methods must be developed specifically for each component to support the depth of inquiry and interactions required by research scientists. This work introduces Conversational Papers (cPAPERS), a dataset of conversational question-answer pairs from reviews of academic papers grounded in these paper components and their associated references from scientific documents available on arXiv. We present a data collection strategy to collect these question-answer pairs from OpenReview and associate them with contextual information from LaTeX source files. Additionally, we present a series of baseline approaches utilizing Large Language Models (LLMs) in both zero-shot and fine-tuned configurations to address the cPAPERS dataset.",
            "introduction": "   1 Introduction  Developing conversational assistants capable of situated and multimodal interactive conversations (SIMMC) over structured knowledge sources remains an open problem [32, 45, 26].   An emerging area of research within this domain is conversational interactions over scientific documents [39]. The number of scientific articles has grown dramatically over the past decade, making it difficult for scientists to find, read, understand, and connect advancements published by their fellow researchers. Scientists in the field of biomedicine, for example, publish over 1 million articles per year or a new article every 2 minutes on average [27]. Developing methods to understand scientific documents and assist researchers is an important problem, especially for the Natural Language Processing (NLP) community. Scientific documents present an interesting challenge in SIMMC-based methods since the content is frequently multimodal [5].   Besides textual paragraphs, researchers rely on various modalities to describe research methods. Figures convey information about concepts developed throughout the paper. Generally, figures represent many different types of information including most commonly graphical and image information. For example, figures explain model architectures and pipelines, summarize experimental results in graphical plots, and represent complex information such as gradient learning surfaces and feature maps. In addition, figures can be used to show various stages of image processing in computer vision and image generation papers.   Equations are crucial for grasping mathematical concepts in scientific texts but can be challenging to interpret. Equations frequently rely on notation introduced elsewhere in the document, often requiring readers to draw from the entire paper to understand the formulation. In addition, equations are typically related to and in many cases help the reader understand other paper components such as figures and tables (and vice-versa).   Finally, tables typically have structures that convey semantic information. These include the text in the row and column headings, and text and often links in the table cells. These structures summarize multiple concepts (e.g., experimental results) in a human-readable form. Therefore, developing SIMMC methods for scientific papers must include a semantic understanding of these structures and table content.   To advance the development of conversational assistants capable of SIMMC in scientific papers, this paper introduces Conversational Papers (cPAPERS111The dataset is publicly available at https://huggingface.co/datasets/avalab/cPAPERS. The collection process, baseline models, and code will be released with the camera-ready version under GNU Public License.), a dataset of conversations in English situated in equations (cPAPERS-EQNS), figures (cPAPERS-FIGS), and tabular information (cPAPERS-TBLS) for scientific texts. Question-answer pairs are sourced from reviews and rebuttals from OpenReview\u00a0222https://openreview.net/. Textual grounding to answer the questions and answers are sourced from the corresponding scientific articles hosted on arXiv\u00a0333https://arxiv.org, an open-access repository of academic preprints.   The contributions of this work include:     \u2022  Introduction of the Conversational Papers (cPAPERS) dataset including three splits (EQNS, FIGS, TBLS)    \u2022  Development of a novel and scalable approach for collecting question-answer pairs from OpenReview and linking them with relevant contextual information from open access tex sources on arXiv    \u2022  Creation of baseline LLM approaches that utilize weakly grounded multimodal context for dialogue responses      The rest of the paper is organized as follows: Section 2 discusses related work in language modeling and question-answering for equations, tables, and figures. Section 3 provides a description of cPAPERS and",
            "references": [
                {
                    "title": "iTBLS: A Dataset of Interactive Conversations Over Tabular Information",
                    "abstract": "This paper introduces Interactive Tables (iTBLS), a dataset of interactive conversations situated in tables from scientific articles. This dataset is designed to facilitate human-AI collaborative problem-solving through AI-powered multi-task tabular capabilities. In contrast to prior work that models interactions as factoid QA or procedure synthesis, iTBLS broadens the scope of interactions to include mathematical reasoning, natural language manipulation, and expansion of existing tables from natural language conversation by delineating interactions into one of three tasks: interpretation, modification, or generation. Additionally, the paper presents a suite of baseline approaches to iTBLS, utilizing zero-shot prompting and parameter-efficient fine-tuning for different computing situations. We also introduce a novel multi-step approach and show how it can be leveraged in conjunction with parameter-efficient fine-tuning to achieve the state-of-the-art on iTBLS; outperforming standard parameter-efficient fine-tuning by up to 15% on interpretation, 18% on modification, and 38% on generation."
                },
                {
                    "title": "TableLlama: Towards Open Large Generalist Models for Tables",
                    "abstract": "Semi-structured tables are ubiquitous. There has been a variety of tasks that aim to automatically interpret, augment, and query tables. Current methods often require pretraining on tables or special model architecture design, are restricted to specific table types, or have simplifying assumptions about tables and tasks. This paper makes the first step towards developing open-source large language models (LLMs) as generalists for a diversity of table-based tasks. Towards that end, we construct TableInstruct, a new dataset with a variety of realistic tables and tasks, for instruction tuning and evaluating LLMs. We further develop the first open-source generalist model for tables, TableLlama, by fine-tuning Llama 2 (7B) with LongLoRA to address the long context challenge. We experiment under both in-domain setting and out-of-domain setting. On 7 out of 8 in-domain tasks, TableLlama achieves comparable or better performance than the SOTA for each task, despite the latter often has task-specific design. On 6 out-of-domain datasets, it achieves 5-44 absolute point gains compared with the base model, showing that training on TableInstruct enhances the model\u2019s generalizability. We open-source our dataset and trained model to boost future work on developing open generalist models for tables."
                },
                {
                    "title": "Exploring the Numerical Reasoning Capabilities of Language Models: A Comprehensive Analysis on Tabular Data",
                    "abstract": "Numbers are crucial for various real-world domains such as finance, economics, and science. Thus, understanding and reasoning with numbers are essential skills for language models to solve different tasks. While different numerical benchmarks have been introduced in recent years, they are limited to specific numerical aspects mostly. In this paper, we propose a hierarchical taxonomy for numerical reasoning skills with more than ten reasoning types across four levels: representation, number sense, manipulation, and complex reasoning. We conduct a comprehensive evaluation of state-of-the-art models to identify reasoning challenges specific to them. Henceforth, we develop a diverse set of numerical probes employing a semi-automated approach. We focus on the tabular Natural Language Inference (TNLI) task as a case study and measure models' performance shifts. Our results show that no model consistently excels across all numerical reasoning types. Among the probed models, FlanT5 (few-/zero-shot) and GPT-3.5 (few-shot) demonstrate strong overall numerical reasoning skills compared to other models. Label-flipping probes indicate that models often exploit dataset artifacts to predict the correct labels."
                },
                {
                    "title": "Testing the Limits of Unified Sequence to Sequence LLM Pretraining on Diverse Table Data Tasks",
                    "abstract": "Tables stored in databases and tables which are present in web pages and articles account for a large part of semi-structured data that is available on the internet. It then becomes pertinent to develop a modeling approach with large language models (LLMs) that can be used to solve diverse table tasks such as semantic parsing, question answering as well as classification problems. Traditionally, there existed separate models specialized for each task individually. It raises the question of how far can we go to build a unified model that works well on some table tasks without significant degradation on others. To that end, we attempt at creating a shared modeling approach in the pretraining stage with encoder-decoder style LLMs that can cater to diverse tasks. We evaluate our approach that continually pretrains and finetunes different model families of T5 with data from tables and surrounding context, on these downstream tasks at different model scales. Through multiple ablation studies, we observe that our pretraining with self-supervised objectives can significantly boost the performance of the models on these tasks. As an example of one improvement, we observe that the instruction finetuned public models which come specialized on text question answering (QA) and have been trained on table data still have room for improvement when it comes to table specific QA. Our work is the first attempt at studying the advantages of a unified approach to table specific pretraining when scaled from 770M to 11B sequence to sequence models while also comparing the instruction finetuned variants of the models."
                },
                {
                    "title": "Nougat: Neural Optical Understanding for Academic Documents",
                    "abstract": "Scientific knowledge is predominantly stored in books and scientific journals, often in the form of PDFs. However, the PDF format leads to a loss of semantic information, particularly for mathematical expressions. We propose Nougat (Neural Optical Understanding for Academic Documents), a Visual Transformer model that performs an Optical Character Recognition (OCR) task for processing scientific documents into a markup language, and demonstrate the effectiveness of our model on a new dataset of scientific documents. The proposed approach offers a promising solution to enhance the accessibility of scientific knowledge in the digital age, by bridging the gap between human-readable documents and machine-readable text. We release the models and code to accelerate future work on scientific text recognition."
                },
                {
                    "title": "TableGPT: Towards Unifying Tables, Nature Language and Commands into One GPT",
                    "abstract": "Tables are prevalent in real-world databases, requiring significant time and effort for humans to analyze and manipulate. The advancements in large language models (LLMs) have made it possible to interact with tables using natural language input, bringing this capability closer to reality. In this paper, we present TableGPT, a unified fine-tuned framework that enables LLMs to understand and operate on tables using external functional commands. It introduces the capability to seamlessly interact with tables, enabling a wide range of functionalities such as question answering, data manipulation (e.g., insert, delete, query, and modify operations), data visualization, analysis report generation, and automated prediction. TableGPT aims to provide convenience and accessibility to users by empowering them to effortlessly leverage tabular data. At the core of TableGPT lies the novel concept of global tabular representations, which empowers LLMs to gain a comprehensive understanding of the entire table beyond meta-information. By jointly training LLMs on both table and text modalities, TableGPT achieves a deep understanding of tabular data and the ability to perform complex operations on tables through chain-of-command instructions. Importantly, TableGPT offers the advantage of being a self-contained system rather than relying on external API interfaces. Moreover, it supports efficient data process flow, query rejection (when appropriate) and private deployment, enabling faster domain data fine-tuning and ensuring data privacy, which enhances the framework's adaptability to specific use cases."
                },
                {
                    "title": "Outside Knowledge Visual Question Answering Version 2.0",
                    "abstract": "Visual question answering (VQA) lies at the intersection of language and vision research. It functions as a building block for multimodal conversational AI and serves as a testbed for assessing a model\u2019s capability for open-domain scene understanding. While progress in this area was initially accelerated with the 2015 release of the popular and large dataset \"VQA\", new datasets are required to continue this research momentum. For example, the 2019 Outside Knowledge VQA dataset \"OKVQA\" extends VQA by adding more challenging questions that require complex, factual, and commonsense knowledge. However, in our analysis, we found that 41.4% of the dataset needed to be corrected and 10.6% needed to be removed. This paper describes the analysis, corrections, and removals completed and presents a new dataset: OK-VQA Version 2.0. To gain insights into the impact of the changes on OK-VQA research, the paper presents results on state-of-the-art models retrained with this new dataset. The side-by-side comparisons show that one method in particular, Knowledge Augmented Transformer for Vision-and-Language, extends its relative lead over competing methods. The dataset is available online.1"
                },
                {
                    "title": "QLoRA: Efficient Finetuning of Quantized LLMs",
                    "abstract": "We present QLoRA, an efficient finetuning approach that reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU while preserving full 16-bit finetuning task performance. QLoRA backpropagates gradients through a frozen, 4-bit quantized pretrained language model into Low Rank Adapters~(LoRA). Our best model family, which we name Guanaco, outperforms all previous openly released models on the Vicuna benchmark, reaching 99.3% of the performance level of ChatGPT while only requiring 24 hours of finetuning on a single GPU. QLoRA introduces a number of innovations to save memory without sacrificing performance: (a) 4-bit NormalFloat (NF4), a new data type that is information theoretically optimal for normally distributed weights (b) double quantization to reduce the average memory footprint by quantizing the quantization constants, and (c) paged optimziers to manage memory spikes. We use QLoRA to finetune more than 1,000 models, providing a detailed analysis of instruction following and chatbot performance across 8 instruction datasets, multiple model types (LLaMA, T5), and model scales that would be infeasible to run with regular finetuning (e.g. 33B and 65B parameter models). Our results show that QLoRA finetuning on a small high-quality dataset leads to state-of-the-art results, even when using smaller models than the previous SoTA. We provide a detailed analysis of chatbot performance based on both human and GPT-4 evaluations showing that GPT-4 evaluations are a cheap and reasonable alternative to human evaluation. Furthermore, we find that current chatbot benchmarks are not trustworthy to accurately evaluate the performance levels of chatbots. A lemon-picked analysis demonstrates where Guanaco fails compared to ChatGPT. We release all of our models and code, including CUDA kernels for 4-bit training."
                },
                {
                    "title": "Table Meets LLM: Can Large Language Models Understand Structured Table Data? A Benchmark and Empirical Study",
                    "abstract": "Large language models (LLMs) are becoming attractive as few-shot reasoners to solve Natural Language (NL)-related tasks. However, there is still much to learn about how well LLMs understand structured data, such as tables. Although tables can be used as input to LLMs with serialization, there is a lack of comprehensive studies that examine whether LLMs can truly comprehend such data. In this paper, we try to understand this by designing a benchmark to evaluate the structural understanding capabilities (SUC) of LLMs. The benchmark we create includes seven tasks, each with its own unique challenges, \\eg, cell lookup, row retrieval, and size detection. We perform a series of evaluations on GPT-3.5 and GPT-4. We find that performance varied depending on several input choices, including table input format, content order, role prompting, and partition marks. Drawing from the insights gained through the benchmark evaluations, we proposeself-augmentation for effective structural prompting, such as critical value / range identification using internal knowledge of LLMs. When combined with carefully chosen input choices, these structural prompting methods lead to promising improvements in LLM performance on a variety of tabular tasks, \\eg, TabFact(\\uparrow2.31%), HybridQA(\\uparrow2.13%), SQA(\\uparrow2.72%), Feverous(\\uparrow0.84%), and ToTTo(\\uparrow5.68%). We believe that our open-source (please find code and data at https://github.com/microsoft/TableProvider) benchmark and proposed prompting methods can serve as a simple yet generic selection for future research."
                },
                {
                    "title": "StructGPT: A General Framework for Large Language Model to Reason over Structured Data",
                    "abstract": "In this paper, we study how to improve the zero-shot reasoning ability of large language models~(LLMs) over structured data in a unified way. Inspired by the study on tool augmentation for LLMs, we develop an \\emph{Iterative Reading-then-Reasoning~(IRR)} approach for solving question answering tasks based on structured data, called \\textbf{StructGPT}. In our approach, we construct the specialized function to collect relevant evidence from structured data (\\ie \\emph{reading}), and let LLMs concentrate the reasoning task based on the collected information (\\ie \\emph{reasoning}). Specially, we propose an \\emph{invoking-linearization-generation} procedure to support LLMs in reasoning on the structured data with the help of the external interfaces. By iterating this procedures with provided interfaces, our approach can gradually approach the target answer to a given query. Extensive experiments conducted on three types of structured data demonstrate the effectiveness of our approach, which can significantly boost the performance of ChatGPT and achieve comparable performance against the full-data supervised-tuning baselines. Our codes and data are publicly available at~\\url{https://github.com/RUCAIBox/StructGPT}."
                },
                {
                    "title": "cTBLS: Augmenting Large Language Models with Conversational Tables",
                    "abstract": "Optimizing accuracy and performance while eliminating hallucinations of open-domain conversational large language models (LLMs) is an open research challenge. A particularly promising direction is to augment and ground LLMs with information from structured sources. This paper introduces Conversational Tables cTBLS, a three-step architecture to retrieve and generate dialogue responses grounded on retrieved tabular information. cTBLS uses Transformer encoder embeddings for Dense Table Retrieval and obtains up to 125% relative improvement over the retriever in the previous state-of-the-art system on the HyrbiDialogue dataset. cTBLS then uses a shared process between encoder and decoder models to perform a coarse+fine tabular knowledge (e.g., cell) ranking combined with a GPT-3.5 LLM response generator to yield a 2x relative improvement in ROUGE scores. Finally, human evaluators prefer cTBLs +80% of the time (coherency, fluency) and judge informativeness to be 4x better than the previous state-of-the-art."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Tree-Based Representation and Generation of Natural and Mathematical Language",
                    "abstract": "Mathematical language in scientific communications and educational scenarios is important yet relatively understudied compared to natural languages. Recent works on mathematical language focus either on representing stand-alone mathematical expressions, especially in their natural tree format, or mathematical reasoning in pre-trained natural language models. Existing works on jointly modeling and generating natural and mathematical languages simply treat mathematical expressions as text, without accounting for the rigid structural properties of mathematical expressions. In this paper, we propose a series of modifications to existing language models to jointly represent and generate text and math: representing mathematical expressions as sequences of node tokens in their operator tree format, using math symbol and tree position embeddings to preserve the semantic and structural properties of mathematical expressions, and using a constrained decoding method to generate mathematically valid expressions. We ground our modifications in GPT-2, resulting in a model MathGPT, and demonstrate that it outperforms baselines on mathematical expression generation tasks."
                },
                {
                    "title": "PASTA: Table-Operations Aware Fact Verification via Sentence-Table Cloze Pre-training",
                    "abstract": "Fact verification has attracted a lot of attention recently, e.g., in journalism, marketing, and policymaking, as misinformation and dis- information can sway one\u2019s opinion and affect one\u2019s actions. While fact-checking is a hard task in general, in many cases, false statements can be easily debunked based on analytics over tables with reliable information. Hence, table- based fact verification has recently emerged as an important and growing research area. Yet, progress has been limited due to the lack of datasets that can be used to pre-train language models (LMs) to be aware of common table operations, such as aggregating a column or comparing tuples. To bridge this gap, this paper introduces PASTA for table-based fact verification via pre-training with synthesized sentence\u2013table cloze questions. In particular, we design six types of common sentence\u2013table cloze tasks, including Filter, Aggregation, Superlative, Comparative, Ordinal, and Unique, based on which we synthesize a large corpus consisting of 1.2 million sentence\u2013table pairs from WikiTables. PASTA uses a recent pre-trained LM, DeBERTaV3, and further pre- trains it on our corpus. Our experimental results show that PASTA achieves new state-of-the-art (SOTA) performance on two table-based fact verification datasets TabFact and SEM-TAB- FACTS. In particular, on the complex set of TabFact, which contains multiple operations, PASTA largely outperforms previous SOTA by 4.7% (85.6% vs. 80.9%), and the gap between PASTA and human performance on the small test set is narrowed to just 1.5% (90.6% vs. 92.1%)."
                },
                {
                    "title": "PACIFIC: Towards Proactive Conversational Question Answering over Tabular and Textual Data in Finance",
                    "abstract": "To facilitate conversational question answering (CQA) over hybrid contexts in finance, we present a new dataset, named PACIFIC. Compared with existing CQA datasets, PACIFIC exhibits three key features: (i) proactivity, (ii) numerical reasoning, and (iii) hybrid context of tables and text. A new task is defined accordingly to study Proactive Conversational Question Answering (PCQA), which combines clarification question generation and CQA. In addition, we propose a novel method, namely UniPCQA, to adapt a hybrid format of input and output content in PCQA into the Seq2Seq problem, including the reformulation of the numerical reasoning process as code generation. UniPCQA performs multi-task learning over all sub-tasks in PCQA and incorporates a simple ensemble strategy to alleviate the error propagation issue in the multi-task learning by cross-validating top-k sampled Seq2Seq outputs. We benchmark the PACIFIC dataset with extensive baselines and provide comprehensive evaluations on each sub-task of PCQA."
                },
                {
                    "title": "MultiHiertt: Numerical Reasoning over Multi Hierarchical Tabular and Textual Data",
                    "abstract": "Numerical reasoning over hybrid data containing both textual and tabular content (e.g., financial reports) has recently attracted much attention in the NLP community. However, existing question answering (QA) benchmarks over hybrid data only include a single flat table in each document and thus lack examples of multi-step numerical reasoning across multiple hierarchical tables. To facilitate data analytical progress, we construct a new large-scale benchmark, MultiHiertt, with QA pairs over Multi Hierarchical Tabular and Textual data. MultiHiertt is built from a wealth of financial reports and has the following unique characteristics: 1) each document contain multiple tables and longer unstructured texts; 2) most of tables contained are hierarchical; 3) the reasoning process required for each question is more complex and challenging than existing benchmarks; and 4) fine-grained annotations of reasoning processes and supporting facts are provided to reveal complex numerical reasoning. We further introduce a novel QA model termed MT2Net, which first applies facts retrieving to extract relevant supporting facts from both tables and text and then uses a reasoning module to perform symbolic reasoning over retrieved facts. We conduct comprehensive experiments on various baselines. The experimental results show that MultiHiertt presents a strong challenge for existing baselines whose results lag far behind the performance of human experts. The dataset and code are publicly available at https://github.com/psunlpgroup/MultiHiertt."
                },
                {
                    "title": "Multimodal Conversational AI: A Survey of Datasets and Approaches",
                    "abstract": "As humans, we experience the world with all our senses or modalities (sound, sight, touch, smell, and taste). We use these modalities, particularly sight and touch, to convey and interpret specific meanings. Multimodal expressions are central to conversations; a rich set of modalities amplify and often compensate for each other. A multimodal conversational AI system answers questions, fulfills tasks, and emulates human conversations by understanding and expressing itself via multiple modalities. This paper motivates, defines, and mathematically formulates the multimodal conversational research objective. We provide a taxonomy of research required to solve the objective: multimodal representation, fusion, alignment, translation, and co-learning. We survey state-of-the-art datasets and approaches for each research area and highlight their limiting assumptions. Finally, we identify multimodal co-learning as a promising direction for multimodal conversational AI research."
                },
                {
                    "title": "HybriDialogue: An Information-Seeking Dialogue Dataset Grounded on Tabular and Textual Data",
                    "abstract": "A pressing challenge in current dialogue systems is to successfully converse with users on topics with information distributed across different modalities. Previous work in multiturn dialogue systems has primarily focused on either text or table information. In more realistic scenarios, having a joint understanding of both is critical as knowledge is typically distributed over both unstructured and structured forms. We present a new dialogue dataset, HybriDialogue, which consists of crowdsourced natural conversations grounded on both Wikipedia text and tables. The conversations are created through the decomposition of complex multihop questions into simple, realistic multiturn dialogue interactions. We propose retrieval, system state tracking, and dialogue response generation tasks for our dataset and conduct baseline experiments for each. Our results show that there is still ample opportunity for improvement, demonstrating the importance of building stronger dialogue systems that can reason over the complex setting of informationseeking dialogue grounded on tables and text."
                },
                {
                    "title": "TableFormer: Robust Transformer Modeling for Table-Text Encoding",
                    "abstract": "Understanding tables is an important aspect of natural language understanding. Existing models for table understanding require linearization of the table structure, where row or column order is encoded as an unwanted bias. Such spurious biases make the model vulnerable to row and column order perturbations. Additionally, prior work has not thoroughly modeled the table structures or table-text alignments, hindering the table-text understanding ability. In this work, we propose a robust and structurally aware table-text encoding architecture TableFormer, where tabular structural biases are incorporated completely through learnable attention biases. TableFormer is (1) strictly invariant to row and column orders, and, (2) could understand tables better due to its tabular inductive biases. Our evaluations showed that TableFormer outperforms strong baselines in all settings on SQA, WTQ and TabFact table reasoning datasets, and achieves state-of-the-art performance on SQA, especially when facing answer-invariant row and column order perturbations (6% improvement over the best baseline), because previous SOTA models\u2019 performance drops by 4% - 6% when facing such perturbations while TableFormer is not affected."
                },
                {
                    "title": "Scientific Formula Retrieval via Tree Embeddings",
                    "abstract": "Exploiting the ever-growing corpus of scientific content calls for new ways and means to effectively organize, search, and retrieve scientific formulae. We propose a new data-driven framework for retrieving similar scientific formulae via learned formula representations based on tree embeddings. FORTE (for FOrmula Representation learning via Tree Embeddings) leverages operator tree representations of symbolic scientific formulae (such as math equations) to explicitly capture their inherent structural and semantic properties. FORTE employs i) a tree encoder that encodes the formula\u2019s operator tree into an embedding vector and ii) a tree decoder that directly generates a formula\u2019s operator tree from the embedding vector. We also develop a novel tree beam search algorithm that improves the quality of the decoded operator trees. We demonstrate that FORTE (sometimes significantly) outperforms various baseline methods on formula reconstruction and retrieval using a real-world dataset comprising 770k scientific formulae collected on-line."
                },
                {
                    "title": "SciCap: Generating Captions for Scientific Figures",
                    "abstract": "Researchers use figures to communicate rich, complex information in scientific papers. The captions of these figures are critical to conveying effective messages. However, low-quality figure captions commonly occur in scientific articles and may decrease understanding. In this paper, we propose an end-to-end neural framework to automatically generate informative, high-quality captions for scientific figures. To this end, we introduce SCICAP, a large-scale figure-caption dataset based on computer science arXiv papers published between 2010 and 2020. After pre-processing - including figure-type classification, sub-figure identification, text normalization, and caption text selection - SCICAP contained more than two million figures extracted from over 290,000 papers. We then established baseline models that caption graph plots, the dominant (19.2%) figure type. The experimental results showed both opportunities and steep challenges of generating captions for scientific figures."
                },
                {
                    "title": "Math Word Problem Generation with Mathematical Consistency and Problem Context Constraints",
                    "abstract": "We study the problem of generating arithmetic math word problems (MWPs) given a math equation that specifies the mathematical computation and a context that specifies the problem scenario. Existing approaches are prone to generating MWPs that are either mathematically invalid or have unsatisfactory language quality. They also either ignore the context or require manual specification of a problem template, which compromises the diversity of the generated MWPs. In this paper, we develop a novel MWP generation approach that leverages i) pre-trained language models and a context keyword selection model to improve the language quality of generated MWPs and ii) an equation consistency constraint for math equations to improve the mathematical validity of the generated MWPs. Extensive quantitative and qualitative experiments on three real-world MWP datasets demonstrate the superior performance of our approach compared to various baselines."
                },
                {
                    "title": "Towards Expressive Communication with Internet Memes: A New Multimodal Conversation Dataset and Benchmark",
                    "abstract": "As a kind of new expression elements, Internet memes are popular and extensively used in online chatting scenarios since they manage to make dialogues vivid, moving, and interesting. However, most current dialogue researches focus on text-only dialogue tasks. In this paper, we propose a new task named as \\textbf{M}eme incorporated \\textbf{O}pen-domain \\textbf{D}ialogue (MOD). Compared to previous dialogue tasks, MOD is much more challenging since it requires the model to understand the multimodal elements as well as the emotions behind them. To facilitate the MOD research, we construct a large-scale open-domain multimodal dialogue dataset incorporating abundant Internet memes into utterances. The dataset consists of $\\sim$45K Chinese conversations with $\\sim$606K utterances. Each conversation contains about $13$ utterances with about $4$ Internet memes on average and each utterance equipped with an Internet meme is annotated with the corresponding emotion. In addition, we present a simple and effective method, which utilizes a unified generation network to solve the MOD task. Experimental results demonstrate that our method trained on the proposed corpus is able to achieve expressive communication including texts and memes. The corpus and models have been publicly available at https://github.com/lizekang/DSTC10-MOD."
                },
                {
                    "title": "FinQA: A Dataset of Numerical Reasoning over Financial Data",
                    "abstract": "The sheer volume of financial statements makes it difficult for humans to access and analyze a business\u2019s financials. Robust numerical reasoning likewise faces unique challenges in this domain. In this work, we focus on answering deep questions over financial data, aiming to automate the analysis of a large corpus of financial documents. In contrast to existing tasks on general domain, the finance domain includes complex numerical reasoning and understanding of heterogeneous representations. To facilitate analytical progress, we propose a new large-scale dataset, FinQA, with Question-Answering pairs over Financial reports, written by financial experts. We also annotate the gold reasoning programs to ensure full explainability. We further introduce baselines and conduct comprehensive experiments in our dataset. The results demonstrate that popular, large, pre-trained models fall far short of expert humans in acquiring finance knowledge and in complex multi-step numerical reasoning on that knowledge. Our dataset \u2013 the first of its kind \u2013 should therefore enable significant, new community research into complex application domains. The dataset and code are publicly available at https://github.com/czyssrs/FinQA."
                },
                {
                    "title": "HiTab: A Hierarchical Table Dataset for Question Answering and Natural Language Generation",
                    "abstract": "Tables are often created with hierarchies, but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables. Hierarchical tables challenge numerical reasoning by complex hierarchical indexing, as well as implicit relationships of calculation and semantics. We present a new dataset, HiTab, to study question answering (QA) and natural language generation (NLG) over hierarchical tables. HiTab is a cross-domain dataset constructed from a wealth of statistical reports and Wikipedia pages, and has unique characteristics: (1) nearly all tables are hierarchical, and (2) QA pairs are not proposed by annotators from scratch, but are revised from real and meaningful sentences authored by analysts. (3) to reveal complex numerical reasoning in statistical reports, we provide fine-grained annotations of quantity and entity alignment. Experiments suggest that this HiTab presents a strong challenge for existing baselines and a valuable benchmark for future research. Targeting hierarchical structure, we devise a hierarchy-aware logical form for symbolic reasoning over tables, which shows high effectiveness. Targeting table reasoning, we leverage entity and quantity alignment to explore partially supervised training in QA and conditional generation in NLG, and largely reduce spurious predictions in QA and produce better descriptions in NLG."
                },
                {
                    "title": "TAPEX: Table Pre-training via Learning a Neural SQL Executor",
                    "abstract": "Recent progress in language model pre-training has achieved a great success via leveraging large-scale unstructured textual data. However, it is still a challenge to apply pre-training on structured tabular data due to the absence of large-scale high-quality tabular data. In this paper, we propose TAPEX to show that table pre-training can be achieved by learning a neural SQL executor over a synthetic corpus, which is obtained by automatically synthesizing executable SQL queries and their execution outputs. TAPEX addresses the data scarcity challenge via guiding the language model to mimic a SQL executor on the diverse, large-scale and high-quality synthetic corpus. We evaluate TAPEX on four benchmark datasets. Experimental results demonstrate that TAPEX outperforms previous table pre-training approaches by a large margin and achieves new state-of-the-art results on all of them. This includes the improvements on the weakly-supervised WikiSQL denotation accuracy to 89.5% (+2.3%), the WikiTableQuestions denotation accuracy to 57.5% (+4.8%), the SQA denotation accuracy to 74.5% (+3.5%), and the TabFact accuracy to 84.2% (+3.2%). To our knowledge, this is the first work to exploit table pre-training via synthetic executable programs and to achieve new state-of-the-art results on various downstream tasks. Our code can be found at https://github.com/microsoft/Table-Pretraining."
                },
                {
                    "title": "TAT-QA: A Question Answering Benchmark on a Hybrid of Tabular and Textual Content in Finance",
                    "abstract": "Hybrid data combining both tabular and textual content (e.g., financial reports) are quite pervasive in the real world. However, Question Answering (QA) over such hybrid data is largely neglected in existing research. In this work, we extract samples from real financial reports to build a new large-scale QA dataset containing both Tabular And Textual data, named TAT-QA, where numerical reasoning is usually required to infer the answer, such as addition, subtraction, multiplication, division, counting, comparison/sorting, and the compositions. We further propose a novel QA model termed TAGOP, which is capable of reasoning over both tables and text. It adopts sequence tagging to extract relevant cells from the table along with relevant spans from the text to infer their semantics, and then applies symbolic reasoning over them with a set of aggregation operators to arrive at the final answer. TAGOP achieves 58.0% inF1, which is an 11.1% absolute increase over the previous best baseline model, according to our experiments on TAT-QA. But this result still lags far behind performance of expert human, i.e.90.8% in F1. It is demonstrated that our TAT-QA is very challenging and can serve as a benchmark for training and testing powerful QA models that address hybrid form data."
                },
                {
                    "title": "MathBERT: A Pre-Trained Model for Mathematical Formula Understanding",
                    "abstract": "Large-scale pre-trained models like BERT, have obtained a great success in various Natural Language Processing (NLP) tasks, while it is still a challenge to adapt them to the math-related tasks. Current pre-trained models neglect the structural features and the semantic correspondence between formula and its context. To address these issues, we propose a novel pre-trained model, namely \\textbf{MathBERT}, which is jointly trained with mathematical formulas and their corresponding contexts. In addition, in order to further capture the semantic-level structural features of formulas, a new pre-training task is designed to predict the masked formula substructures extracted from the Operator Tree (OPT), which is the semantic structural representation of formulas. We conduct various experiments on three downstream tasks to evaluate the performance of MathBERT, including mathematical information retrieval, formula topic classification and formula headline generation. Experimental results demonstrate that MathBERT significantly outperforms existing methods on all those three tasks. Moreover, we qualitatively show that this pre-trained model effectively captures the semantic-level structural information of formulas. To the best of our knowledge, MathBERT is the first pre-trained model for mathematical formula understanding."
                },
                {
                    "title": "FeTaQA: Free-form Table Question Answering",
                    "abstract": "Existing table question answering datasets contain abundant factual questions that primarily evaluate a QA system\u2019s comprehension of query and tabular data. However, restricted by their short-form answers, these datasets fail to include question\u2013answer interactions that represent more advanced and naturally occurring information needs: questions that ask for reasoning and integration of information pieces retrieved from a structured knowledge source. To complement the existing datasets and to reveal the challenging nature of the table-based question answering task, we introduce FeTaQA, a new dataset with 10K Wikipedia-based table, question, free-form answer, supporting table cells pairs. FeTaQA is collected from noteworthy descriptions of Wikipedia tables that contain information people tend to seek; generation of these descriptions requires advanced processing that humans perform on a daily basis: Understand the question and table, retrieve, integrate, infer, and conduct text planning and surface realization to generate an answer. We provide two benchmark methods for the proposed task: a pipeline method based on semantic parsing-based QA systems and an end-to-end method based on large pretrained text generation models, and show that FeTaQA poses a challenge for both methods."
                },
                {
                    "title": "Open Domain Question Answering over Tables via Dense Retrieval",
                    "abstract": "Recent advances in open-domain QA have led to strong models based on dense retrieval, but only focused on retrieving textual passages. In this work, we tackle open-domain QA over tables for the first time, and show that retrieval can be improved by a retriever designed to handle tabular context. We present an effective pre-training procedure for our retriever and improve retrieval quality with mined hard negatives. As relevant datasets are missing, we extract a subset of Natural Questions (Kwiatkowski et al., 2019) into a Table QA dataset. We find that our retriever improves retrieval results from 72.0 to 81.1 recall@10 and end-to-end QA results from 33.8 to 37.7 exact match, over a BERT based retriever."
                },
                {
                    "title": "TableGPT: Few-shot Table-to-Text Generation with Table Structure Reconstruction and Content Matching",
                    "abstract": "Although neural table-to-text models have achieved remarkable progress with the help of large-scale datasets, they suffer insufficient learning problem with limited training data. Recently, pre-trained language models show potential in few-shot learning with linguistic knowledge learnt from pretraining on large-scale corpus. However, benefiting table-to-text generation in few-shot setting with the powerful pretrained language model faces three challenges, including (1) the gap between the task\u2019s structured input and the natural language input for pretraining language model. (2) The lack of modeling for table structure and (3) improving text fidelity with less incorrect expressions that are contradicting to the table. To address aforementioned problems, we propose TableGPT for table-to-text generation. At first, we utilize table transformation module with template to rewrite structured table in natural language as input for GPT-2. In addition, we exploit multi-task learning with two auxiliary tasks that preserve table\u2019s structural information by reconstructing the structure from GPT-2\u2019s representation and improving the text\u2019s fidelity with content matching task aligning the table and information in the generated text. By experimenting on Humans, Songs and Books, three few-shot table-to-text datasets in different domains, our model outperforms existing systems on most few-shot settings."
                },
                {
                    "title": "Understanding tables with intermediate pre-training",
                    "abstract": "Table entailment, the binary classification task of finding if a sentence is supported or refuted by the content of a table, requires parsing language and table structure as well as numerical and discrete reasoning. While there is extensive work on textual entailment, table entailment is less well studied. We adapt TAPAS (Herzig et al., 2020), a table-based BERT model, to recognize entailment. Motivated by the benefits of data augmentation, we create a balanced dataset of millions of automatically created training examples which are learned in an intermediate step prior to fine-tuning. This new data is not only useful for table entailment, but also for SQA (Iyyer et al., 2017), a sequential table QA task. To be able to use long examples as input of BERT models, we evaluate table pruning techniques as a pre-processing step to drastically improve the training and prediction efficiency at a moderate drop in accuracy. The different methods set the new state-of-the-art on the TabFact (Chen et al., 2020) and SQA datasets."
                },
                {
                    "title": "Situated and Interactive Multimodal Conversations",
                    "abstract": "Next generation virtual assistants are envisioned to handle multimodal inputs (e.g., vision, memories of previous interactions, and the user\u2019s utterances), and perform multimodal actions (, displaying a route while generating the system\u2019s utterance). We introduce Situated Interactive MultiModal Conversations (SIMMC) as a new direction aimed at training agents that take multimodal actions grounded in a co-evolving multimodal input context in addition to the dialog history. We provide two SIMMC datasets totalling ~13K human-human dialogs (~169K utterances) collected using a multimodal Wizard-of-Oz (WoZ) setup, on two shopping domains: (a) furniture \u2013 grounded in a shared virtual environment; and (b) fashion \u2013 grounded in an evolving set of images. Datasets include multimodal context of the items appearing in each scene, and contextual NLU, NLG and coreference annotations using a novel and unified framework of SIMMC conversational acts for both user and assistant utterances. Finally, we present several tasks within SIMMC as objective evaluation protocols, such as structural API prediction, response generation, and dialog state tracking. We benchmark a collection of existing models on these SIMMC tasks as strong baselines, and demonstrate rich multimodal conversational interactions. Our data, annotations, and models will be made publicly available."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "TaBERT: Pretraining for Joint Understanding of Textual and Tabular Data",
                    "abstract": "Recent years have witnessed the burgeoning of pretrained language models (LMs) for text-based natural language (NL) understanding tasks. Such models are typically trained on free-form NL text, hence may not be suitable for tasks like semantic parsing over structured data, which require reasoning over both free-form NL questions and structured tabular data (e.g., database tables). In this paper we present TaBERT, a pretrained LM that jointly learns representations for NL sentences and (semi-)structured tables. TaBERT is trained on a large corpus of 26 million tables and their English contexts. In experiments, neural semantic parsers using TaBERT as feature representation layers achieve new best results on the challenging weakly-supervised semantic parsing benchmark WikiTableQuestions, while performing competitively on the text-to-SQL dataset Spider."
                },
                {
                    "title": "TaPas: Weakly Supervised Table Parsing via Pre-training",
                    "abstract": "Answering natural language questions over tables is usually seen as a semantic parsing task. To alleviate the collection cost of full logical forms, one popular approach focuses on weak supervision consisting of denotations instead of logical forms. However, training semantic parsers from weak supervision poses difficulties, and in addition, the generated logical forms are only used as an intermediate step prior to retrieving the denotation. In this paper, we present TaPas, an approach to question answering over tables without generating logical forms. TaPas trains from weak supervision, and predicts the denotation by selecting table cells and optionally applying a corresponding aggregation operator to such selection. TaPas extends BERT\u2019s architecture to encode tables as input, initializes from an effective joint pre-training of text segments and tables crawled from Wikipedia, and is trained end-to-end. We experiment with three different semantic parsing datasets, and find that TaPas outperforms or rivals semantic parsing models by improving state-of-the-art accuracy on SQA from 55.1 to 67.2 and performing on par with the state-of-the-art on WikiSQL and WikiTQ, but with a simpler model architecture. We additionally find that transfer learning, which is trivial in our setting, from WikiSQL to WikiTQ, yields 48.7 accuracy, 4.2 points above the state-of-the-art."
                },
                {
                    "title": "ManyModalQA: Modality Disambiguation and QA over Diverse Inputs",
                    "abstract": "We present a new multimodal question answering challenge, ManyModalQA, in which an agent must answer a question by considering three distinct modalities: text, images, and tables. We collect our data by scraping Wikipedia and then utilize crowdsourcing to collect question-answer pairs. Our questions are ambiguous, in that the modality that contains the answer is not easily determined based solely upon the question. To demonstrate this ambiguity, we construct a modality selector (or disambiguator) network, and this model gets substantially lower accuracy on our challenge set, compared to existing datasets, indicating that our questions are more ambiguous. By analyzing this model, we investigate which words in the question are indicative of the modality. Next, we construct a simple baseline ManyModalQA model, which, based on the prediction from the modality selector, fires a corresponding pre-trained state-of-the-art unimodal QA model. We focus on providing the community with a new manymodal evaluation set and only provide a fine-tuning set, with the expectation that existing datasets and approaches will be transferred for most of the training, to encourage low-resource generalization without large, monolithic training sets for each new task. There is a significant gap between our baseline models and human performance; therefore, we hope that this challenge encourages research in end-to-end modality disambiguation and multimodal QA models, as well as transfer learning."
                },
                {
                    "title": "TabFact: A Large-scale Dataset for Table-based Fact Verification",
                    "abstract": "The problem of verifying whether a textual hypothesis holds based on the given evidence, also known as fact verification, plays an important role in the study of natural language understanding and semantic representation. However, existing studies are mainly restricted to dealing with unstructured evidence (e.g., natural language sentences and documents, news, etc), while verification under structured evidence, such as tables, graphs, and databases, remains under-explored. This paper specifically aims to study the fact verification given semi-structured data as evidence. To this end, we construct a large-scale dataset called TabFact with 16k Wikipedia tables as the evidence for 118k human-annotated natural language statements, which are labeled as either ENTAILED or REFUTED. TabFact is challenging since it involves both soft linguistic reasoning and hard symbolic reasoning. To address these reasoning challenges, we design two different models: Table-BERT and Latent Program Algorithm (LPA). Table-BERT leverages the state-of-the-art pre-trained language model to encode the linearized tables and statements into continuous vectors for verification. LPA parses statements into programs and executes them against the tables to obtain the returned binary value for verification. Both methods achieve similar accuracy but still lag far behind human performance. We also perform a comprehensive analysis to demonstrate great future opportunities. The data and code of the dataset are provided in \\url{this https URL}."
                },
                {
                    "title": "OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge",
                    "abstract": "Visual Question Answering (VQA) in its ideal form lets us study reasoning in the joint space of vision and language and serves as a proxy for the AI task of scene understanding. However, most VQA benchmarks to date are focused on questions such as simple counting, visual attributes, and object detection that do not require reasoning or knowledge beyond what is in the image. In this paper, we address the task of knowledge-based visual question answering and provide a benchmark, called OK-VQA, where the image content is not sufficient to answer the questions, encouraging methods that rely on external knowledge resources. Our new dataset includes more than 14,000 questions that require external knowledge to answer. We show that the performance of the state-of-the-art VQA models degrades drastically in this new setting. Our analysis shows that our knowledge-based VQA task is diverse, difficult, and large compared to previous knowledge-based VQA datasets. We hope that this dataset enables researchers to open up new avenues for research in this domain."
                },
                {
                    "title": "BERTScore: Evaluating Text Generation with BERT",
                    "abstract": "We propose BERTScore, an automatic evaluation metric for text generation. Analogously to common metrics, BERTScore computes a similarity score for each token in the candidate sentence with each token in the reference sentence. However, instead of exact matches, we compute token similarity using contextual embeddings. We evaluate using the outputs of 363 machine translation and image captioning systems. BERTScore correlates better with human judgments and provides stronger model selection performance than existing metrics. Finally, we use an adversarial paraphrase detection task to show that BERTScore is more robust to challenging examples when compared to existing metrics."
                },
                {
                    "title": "TopicEq: A Joint Topic and Mathematical Equation Model for Scientific Texts",
                    "abstract": "Scientific documents rely on both mathematics and text to communicate ideas. Inspired by the topical correspondence between mathematical equations and word contexts observed in scientific texts, we propose a novel topic model that jointly generates mathematical equations and their surrounding text (TopicEq). Using an extension of the correlated topic model, the context is generated from a mixture of latent topics, and the equation is generated by an RNN that depends on the latent topic activations. To experiment with this model, we create a corpus of 400K equation-context pairs extracted from a range of scientific articles from arXiv, and fit the model using a variational autoencoder approach. Experimental results show that this joint model significantly outperforms existing topic models and equation models for scientific texts. Moreover, we qualitatively show that the model effectively captures the relationship between topics and mathematics, enabling novel applications such as topic-aware equation generation, equation topic inference, and topic-aware alignment of mathematical symbols and words."
                },
                {
                    "title": "Image-Chat: Engaging Grounded Conversations",
                    "abstract": "To achieve the long-term goal of machines being able to engage humans in conversation, our models should captivate the interest of their speaking partners. Communication grounded in images, whereby a dialogue is conducted based on a given photo, is a setup naturally appealing to humans (Hu et al., 2014). In this work we study large-scale architectures and datasets for this goal. We test a set of neural architectures using state-of-the-art image and text representations, considering various ways to fuse the components. To test such models, we collect a dataset of grounded human-human conversations, where speakers are asked to play roles given a provided emotional mood or style, as the use of such traits is also a key factor in engagingness (Guo et al., 2019). Our dataset, Image-Chat, consists of 202k dialogues over 202k images using 215 possible style traits. Automatic metrics and human evaluations of engagingness show the efficacy of our approach; in particular, we obtain state-of-the-art performance on the existing IGC task, and our best performing model is almost on par with humans on the Image-Chat test set (preferred 47.7% of the time)."
                },
                {
                    "title": "Semantically-Aligned Equation Generation for Solving and Reasoning Math Word Problems",
                    "abstract": "Solving math word problems is a challenging task that requires accurate natural language understanding to bridge natural language texts and math expressions. Motivated by the intuition about how human generates the equations given the problem texts, this paper presents a neural approach to automatically solve math word problems by operating symbols according to their semantic meanings in texts. This paper views the process of generating equation as a bridge between the semantic world and the symbolic world, where the proposed neural math solver is based on an encoder-decoder framework. In the proposed model, the encoder is designed to understand the semantics of problems, and the decoder focuses on tracking semantic meanings of the generated symbols and then deciding which symbol to generate next. The preliminary experiments are conducted in a dataset Math23K, and our model significantly outperforms both the state-of-the-art single model and the best non-retrieval-based model over about 10% accuracy, demonstrating the effectiveness of bridging the symbolic and semantic worlds from math word problems."
                },
                {
                    "title": "Image-Grounded Conversations: Multimodal Context for Natural Question and Response Generation",
                    "abstract": "The popularity of image sharing on social media and the engagement it creates between users reflect the important role that visual context plays in everyday conversations. We present a novel task, Image Grounded Conversations (IGC), in which natural-sounding conversations are generated about a shared image. To benchmark progress, we introduce a new multiple reference dataset of crowd-sourced, event-centric conversations on images. IGC falls on the continuum between chit-chat and goal-directed conversation models, where visual grounding constrains the topic of conversation to event-driven utterances. Experiments with models trained on social media data show that the combination of visual and textual context enhances the quality of generated conversational turns. In human evaluation, the gap between human performance and that of both neural and retrieval architectures suggests that multi-modal IGC presents an interesting challenge for dialog research."
                },
                {
                    "title": "Visual7W: Grounded Question Answering in Images",
                    "abstract": "We have seen great progress in basic perceptual tasks such as object recognition and detection. However, AI models still fail to match humans in high-level vision tasks due to the lack of capacities for deeper reasoning. Recently the new task of visual question answering (QA) has been proposed to evaluate a model's capacity for deep image understanding. Previous works have established a loose, global association between QA sentences and images. However, many questions and answers, in practice, relate to local regions in the images. We establish a semantic link between textual descriptions and image regions by object-level grounding. It enables a new type of QA with visual answers, in addition to textual answers used in previous work. We study the visual QA tasks in a grounded setting with a large collection of 7W multiple-choice QA pairs. Furthermore, we evaluate human performance and several baseline models on the QA tasks. Finally, we propose a novel LSTM model with spatial attention to tackle the 7W QA tasks."
                },
                {
                    "title": "Compositional Semantic Parsing on Semi-Structured Tables",
                    "abstract": "Two important aspects of semantic parsing for question answering are the breadth of the knowledge source and the depth of logical compositionality. While existing work trades off one aspect for another, this paper simultaneously makes progress on both fronts through a new task: answering complex questions on semi-structured tables using question-answer pairs as supervision. The central challenge arises from two compounding factors: the broader domain results in an open-ended set of relations, and the deeper compositionality results in a combinatorial explosion in the space of logical forms. We propose a logical-form driven parsing algorithm guided by strong typing constraints and show that it obtains significant improvements over natural baselines. For evaluation, we created a new dataset of 22,033 complex questions on Wikipedia tables, which is made publicly available."
                },
                {
                    "title": "Looking Beyond Text: Extracting Figures, Tables and Captions from Computer Science Papers",
                    "abstract": "Identifying and extracting figures and tables along with their captions from scholarly articles is important both as a way of providing tools for article summarization, and as part of larger systems that seek to gain deeper, semantic understanding of these articles. While many \"off-the-shelf\" tools exist that can extract embedded images from these documents, e.g. PDFBox, Poppler, etc., these tools are unable to extract tables, captions, and figures composed of vector graphics. Our proposed approach analyzes the structure of individual pages of a document by detecting chunks of body text, and locates the areas wherein figures or tables could reside by reasoning about the empty regions within that text. This method can extract a wide variety of figures because it does not make strong assumptions about the format of the figures embedded in the document, as long as they can be differentiated from the main article's text. Our algorithm also demonstrates a caption-to-figure matching component that is effective even in cases where individual captions are adjacent to multiple figures. Our contribution also includes methods for leveraging particular consistency and formatting assumptions to identify titles, body text and captions within each article. We introduce a new dataset of 150 computer science papers along with ground truth labels for the locations of the figures, tables and captions within them. Our algorithm achieves 96% precision at 92% recall when tested against this dataset, surpassing previous state of the art. We release our dataset, code, and evaluation scripts on our project website for enabling future research."
                },
                {
                    "title": "METEOR: An Automatic Metric for MT Evaluation with Improved Correlation with Human Judgments",
                    "abstract": "We describe METEOR, an automatic metric for machine translation evaluation that is based on a generalized concept of unigram matching between the machineproduced translation and human-produced reference translations. Unigrams can be matched based on their surface forms, stemmed forms, and meanings; furthermore, METEOR can be easily extended to include more advanced matching strategies. Once all generalized unigram matches between the two strings have been found, METEOR computes a score for this matching using a combination of unigram-precision, unigram-recall, and a measure of fragmentation that is designed to directly capture how well-ordered the matched words in the machine translation are in relation to the reference. We evaluate METEOR by measuring the correlation between the metric scores and human judgments of translation quality. We compute the Pearson R correlation value between its scores and human quality assessments of the LDC TIDES 2003 Arabic-to-English and Chinese-to-English datasets. We perform segment-bysegment correlation, and show that METEOR gets an R correlation value of 0.347 on the Arabic data and 0.331 on the Chinese data. This is shown to be an improvement on using simply unigramprecision, unigram-recall and their harmonic F1 combination. We also perform experiments to show the relative contributions of the various mapping modules."
                },
                {
                    "title": "ROUGE: A Package for Automatic Evaluation of Summaries",
                    "abstract": "ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST."
                },
                {
                    "title": "Overview of Situated and Interactive Multimodal Conversations (SIMMC) 2.1 Track at DSTC 11",
                    "abstract": "With ever increasing interest in task-oriented dialog systems, the recent work on Situated and Interactive Multimodal Conversations (SIMMC 2.0) aims to develop personal assistants that interact with users, grounded in an immersive and co-observed setting of photo-realistic scenes. The dataset contains 11k task-oriented dialogs set in an interactive shopping scenario, spanning more than 117k utterances. In order to push research towards this next generation virtual assistants, the SIMMC 2.1 challenge was conducted at the Eleventh Dialog System Technology Challenge (DSTC) which had entries from across the world competing to achieve the state-of-the-art performance in the SIMMC 2.1 task. In this report, we present and compare 13 SIMMC 2.1 model entries from 5 trams across the world to understand the current progress made across the last three years (starting with SIMMC 1.0 and 2.0 challenges) for multimodal task-oriented dialog systems. We hope that our analysis throws light on components that showed promise in addition to identifying the gaps for future research towards this grand goal of an immersive multimodal conversational agent."
                },
                {
                    "title": "A Comparative Study on the Impact of Model Compression Techniques on Fairness in Language Models",
                    "abstract": "Compression techniques for deep learning have become increasingly popular, particularly in settings where latency and memory constraints are imposed. Several methods, such as pruning, distillation, and quantization, have been adopted for compressing models, each providing distinct advantages. However, existing literature demonstrates that compressing deep learning models could affect their fairness. Our analysis involves a comprehensive evaluation of pruned, distilled, and quantized language models, which we benchmark across a range of intrinsic and extrinsic metrics for measuring bias in text classification. We also investigate the impact of using multilingual models and evaluation measures. Our findings highlight the significance of considering both the pre-trained model and the chosen compression strategy in developing equitable language technologies. The results also indicate that compression strategies can have an adverse effect on fairness measures."
                },
                {
                    "title": "Scientific Chart Summarization: Datasets and Improved Text Modeling",
                    "abstract": "Chart \ufb01gures usually convey the key message in a multimodal document. Understanding charts automatically and making charts more accessible becomes indispensable in the information era. In this paper, we study the chart summarization problem in which the goal is to generate sentences that describe the salient information in a chart image. To obtain training examples, we leverage image-caption pairs in multiple scien-ti\ufb01c areas. We create a dataset of single-chart images from research papers in PubMed Central (PMC) and arXiv. Most recent vision-and-language works focus on natural images. Several challenges in structured images such as charts are under-explored. One key property of charts is that the text components (e.g., legends and axis names) carry important information. In our proposed model, we not only use a standard visual encoder but also a text encoder to encode a chart image. The visual and textual representations are connected to a large pre-trained language decoder via pre-embedding and cross-attention approaches, respectively. Experimental results show that the proposed model is signi\ufb01cantly better than an image captioning baseline."
                },
                {
                    "title": "The Fact Extraction and VERification Over Unstructured and Structured information (FEVEROUS) Shared Task",
                    "abstract": "The Fact Extraction and VERification Over Unstructured and Structured information (FEVEROUS) shared task, asks participating systems to determine whether human-authored claims are Supported or Refuted based on evidence retrieved from Wikipedia (or NotEnoughInfo if the claim cannot be verified). Compared to the FEVER 2018 shared task, the main challenge is the addition of structured data (tables and lists) as a source of evidence. The claims in the FEVEROUS dataset can be verified using only structured evidence, only unstructured evidence, or a mixture of both. Submissions are evaluated using the FEVEROUS score that combines label accuracy and evidence retrieval. Unlike FEVER 2018, FEVEROUS requires partial evidence to be returned for NotEnoughInfo claims, and the claims are longer and thus more complex. The shared task received 13 entries, six of which were able to beat the baseline system. The winning team was \u201cBust a move!\u201d, achieving a FEVEROUS score of 27% (+9% compared to the baseline). In this paper we describe the shared task, present the full results and highlight commonalities and innovations among the participating systems."
                },
                {
                    "title": "Recognizing Figure Labels in Patents",
                    "abstract": "Scienti\ufb01c documents often contain signi\ufb01cant information in \ufb01gures. The United States Patent and Trademark Of\ufb01ce (USPTO) awards thousands of patents each week, with each patent containing on the order of a dozen \ufb01gures. The information conveyed by these \ufb01gures typically include a drawing or diagram, a label, caption and reference text within the document. Yet associating the short bits of text to the \ufb01gure is challenging when labels are embedded within the \ufb01gure, as they typically are in patents. Using patents as a testbench, this paper highlights an open challenge in analyzing all of the information presented in scienti\ufb01c/technical documents - namely, there is a technological gap in recognizing characters embedded in drawings, which leads to dif\ufb01culties in processing the text associated with scienti\ufb01c \ufb01gures. We demonstrate that automatically reading the \ufb01gure label in patent diagram \ufb01gures is an open challenge, as we evaluate several state-of-the-art optical character recognition (OCR) methods on recent patents. Because the visual characteristics of drawings/diagrams are quite similar to that of text (high contrast, width of strokes, etc), separating the diagram from the text is challenging and leads to both (a) false detection of characters from pixels that are not text and (b) missed text that is critical for identifying the \ufb01gure number. We develop a method for automatically reading the patent \ufb01gure labels by \ufb01rst identifying the bounding box containing the label using a novel non-convex hull approach, and then demonstrate the success of OCR when the text is isolated from the diagram."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop conversational assistants that effectively engage in situated and multimodal interactive conversations over scientific documents, particularly in understanding and interpreting equations, figures, and tables?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing challenge of information overload in scientific literature, particularly in fast-paced fields like biomedicine. By enabling more effective interactions with scientific documents, this research could significantly enhance researchers' ability to find, comprehend, and connect advancements in their fields. This could lead to improved collaboration, faster dissemination of knowledge, and ultimately, more rapid scientific progress. Furthermore, advancements in Natural Language Processing (NLP) through this work could pave the way for practical applications in educational tools, research assistants, and automated literature reviews.\n\n**[Question 3] - Why is it hard?**  \nThe complexity of this problem arises from the multimodal nature of scientific documents, which include not only text but also figures, tables, and equations that convey critical information. Naive approaches may fail because they often treat these modalities in isolation, neglecting the interdependencies between them. For instance, understanding an equation may require context from the surrounding text or figures, and vice versa. Additionally, the semantic structures of tables and the graphical representations in figures add layers of complexity that must be navigated. Overcoming these challenges requires sophisticated models that can integrate and interpret diverse forms of information simultaneously.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on single modalities or has not adequately addressed the integration of multimodal information in scientific documents. Limitations in existing solutions include a lack of datasets that encompass the necessary variety of scientific content and the absence of effective methods for linking question-answer pairs to relevant contextual information. Barriers such as the complexity of scientific language, the need for domain-specific knowledge, and the challenges in creating scalable datasets have hindered progress. This work improves upon prior efforts by introducing the cPAPERS dataset, which specifically targets the multimodal aspects of scientific documents and provides a structured approach to generating and linking question-answer pairs.\n\n**[Question 5] - What are the key components of my approach and results?**  \nThe proposed methodology involves the creation of the cPAPERS dataset, which includes three distinct splits focusing on equations (cPAPERS-EQNS), figures (cPAPERS-FIGS), and tables (c"
            }
        },
        "author_data": {
            "0d36200f-f08a-4d06-8155-34a4d7a9fc33": {
                "pk": "0d36200f-f08a-4d06-8155-34a4d7a9fc33",
                "name": "Anirudh Sundar",
                "collaborators": [
                    "Larry Heck",
                    "Christopher Richardson",
                    "William Gay",
                    "Simon Heck"
                ],
                "domain": [
                    "Multimodal AI",
                    "Natural Language Processing",
                    "Conversational AI",
                    "Commonsense Reasoning"
                ],
                "publications": [
                    {
                        "title": "Multimodal Conversational AI: A Survey of Datasets and Approaches",
                        "abstract": "As humans, we experience the world with all our senses or modalities (sound, sight, touch, smell, and taste). We use these modalities, particularly sight and touch, to convey and interpret specific meanings. Multimodal expressions are central to conversations; a rich set of modalities amplify and often compensate for each other. A multimodal conversational AI system answers questions, fulfills tasks, and emulates human conversations by understanding and expressing itself via multiple modalities. This paper motivates, defines, and mathematically formulates the multimodal conversational research objective. We provide a taxonomy of research required to solve the objective: multimodal representation, fusion, alignment, translation, and co-learning. We survey state-of-the-art datasets and approaches for each research area and highlight their limiting assumptions. Finally, we identify multimodal co-learning as a promising direction for multimodal conversational AI research."
                    },
                    {
                        "title": "gTBLS: Generating Tables from Text by Conditional Question Answering",
                        "abstract": "Distilling large, unstructured text into a structured, condensed form such as tables is an open research problem. One of the primary challenges in automatically generating tables is ensuring their syntactic validity. Prior approaches address this challenge by including additional parameters in the Transformer's attention mechanism to attend to specific rows and column headers. In contrast to this single-stage method, this paper presents a two-stage approach called Generative Tables (gTBLS). The first stage infers table structure (row and column headers) from the text. The second stage formulates questions using these headers and fine-tunes a causal language model to answer them. Furthermore, the gTBLS approach is amenable to the utilization of pre-trained Large Language Models in a zero-shot configuration, presenting a solution for table generation in situations where fine-tuning is not feasible. gTBLS improves prior approaches by up to 10% in BERTScore on the table construction task and up to 20% on the table content generation task of the E2E, WikiTableText, WikiBio, and RotoWire datasets."
                    },
                    {
                        "title": "iTBLS: A Dataset of Interactive Conversations Over Tabular Information",
                        "abstract": "This paper introduces Interactive Tables (iTBLS), a dataset of interactive conversations situated in tables from scientific articles. This dataset is designed to facilitate human-AI collaborative problem-solving through AI-powered multi-task tabular capabilities. In contrast to prior work that models interactions as factoid QA or procedure synthesis, iTBLS broadens the scope of interactions to include mathematical reasoning, natural language manipulation, and expansion of existing tables from natural language conversation by delineating interactions into one of three tasks: interpretation, modification, or generation. Additionally, the paper presents a suite of baseline approaches to iTBLS, utilizing zero-shot prompting and parameter-efficient fine-tuning for different computing situations. We also introduce a novel multi-step approach and show how it can be leveraged in conjunction with parameter-efficient fine-tuning to achieve the state-of-the-art on iTBLS; outperforming standard parameter-efficient fine-tuning by up to 15% on interpretation, 18% on modification, and 38% on generation."
                    },
                    {
                        "title": "mForms : Multimodal Form-Filling with Question Answering",
                        "abstract": "This paper presents a new approach to form-filling by reformulating the task as multimodal natural language Question Answering (QA). The reformulation is achieved by first translating the elements on the GUI form (text fields, buttons, icons, etc.) to natural language questions, where these questions capture the element's multimodal semantics. After a match is determined between the form element (Question) and the user utterance (Answer), the form element is filled through a pre-trained extractive QA system. By leveraging pre-trained QA models and not requiring form-specific training, this approach to form-filling is zero-shot. The paper also presents an approach to further refine the form-filling by using multi-task training to incorporate a potentially large number of successive tasks. Finally, the paper introduces a multimodal natural language form-filling dataset Multimodal Forms (mForms), as well as a multimodal extension of the popular ATIS dataset to support future research and experimentation. Results show the new approach not only maintains robust accuracy for sparse training conditions but achieves state-of-the-art F1 of 0.97 on ATIS with approximately 1/10th of the training data."
                    },
                    {
                        "title": "SYNDICOM: Improving Conversational Commonsense with Error-Injection and Natural Language Feedback",
                        "abstract": "Commonsense reasoning is a critical aspect of human communication. Despite recent advances in conversational AI driven by large language models, commonsense reasoning remains a challenging task. In this work, we introduce SYNDICOM - a method for improving commonsense in dialogue response generation. SYNDICOM consists of two components. The first component is a dataset composed of commonsense dialogues created from a knowledge graph and synthesized into natural language. This dataset includes both valid and invalid responses to dialogue contexts, along with natural language feedback (NLF) for the invalid responses. The second contribution is a two-step procedure: training a model to predict natural language feedback (NLF) for invalid responses, and then training a response generation model conditioned on the predicted NLF, the invalid response, and the dialogue. SYNDICOM is scalable and does not require reinforcement learning. Empirical results on three tasks are evaluated using a broad range of metrics. SYNDICOM achieves a relative improvement of 53% over ChatGPT on ROUGE1, and human evaluators prefer SYNDICOM over ChatGPT 57% of the time. We will publicly release the code and the full dataset."
                    }
                ]
            },
            "beec1ae3-4846-4d46-944d-219f9c77b4e7": {
                "pk": "beec1ae3-4846-4d46-944d-219f9c77b4e7",
                "name": "Jin Xu",
                "collaborators": [
                    "Junghyo Jo",
                    "Natarajan Gautam"
                ],
                "domain": [
                    "Graph Theory",
                    "Computational Biology",
                    "Queueing Theory",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Theory on the Structure and Coloring of Maximal Planar Graphs (1)Recursion Formulae of Chromatic Polynomial and Four-Color Conjecture",
                        "abstract": "In this paper, two recursion formulae of chromatic polynomial of a maximal planar graph G are obtained. Moreover, the application of these formulaes to the proof of Four-Color Conjecture is investigated. By using these formulae, the proof of Four-Color Conjecture boils down to the study on a special class of graphs, viz., 4-chromatic-funnel pseudo uniquely-4-colorable maximal planar graphs."
                    },
                    {
                        "title": "A Mathematical Proof of the Four-Color Conjecture (1): Transformation Step",
                        "abstract": "The four-color conjecture has puzzled mathematicians for over 170 years and has yet to be proven by purely mathematical methods. This series of articles provides a purely mathematical proof of the four-color conjecture, consisting of two parts: the transformation step and the decycle step. The transformation step uses two innovative tools, contracting and extending operations and unchanged bichromatic cycles, to transform the proof of the four-color conjecture into the decycle problem of 4-base modules. Moreover, the decycle step solves the decycle problem of 4-base modules using two other innovative tools: the color-connected potential and the pocket operations. This article presents the proof of the transformation step."
                    },
                    {
                        "title": "Mathematical Proofs of Two Conjectures: The Four Color Problem and The Uniquely 4-colorable Planar Graph",
                        "abstract": "The famous four color theorem states that for all planar graphs, every vertex can be assigned one of 4 colors such that no two adjacent vertices receive the same color. Since Francis Guthrie first conjectured it in 1852, it is until 1976 with electronic computer that Appel and Haken first gave a proof by finding and verifying 1936 reducible unavoidable sets, and a simplified proof of Robertson, Sanders, Seymour and Thomas in 1997 only involved 633 reducible unavoidable sets, both proofs could not be realized effectively by hand. Until now, finding the reducible unavoidable sets remains the only successful method to use, which came from Kempe's first \"proof\" of the four color problem in 1879. An alternative method only involving 4 reducible unavoidable sets for proving the four color theorem is used in this paper, which takes form of mathematical proof rather than a computer-assisted proof and proves both the four color conjecture and the uniquely 4-colorable planar graph conjecture by mathematical method."
                    },
                    {
                        "title": "Probe Machine",
                        "abstract": "A novel computing model, called \\emph{Probe Machine}, is proposed in this paper. Different from Turing Machine, Probe Machine is a fully-parallel computing model in the sense that it can simultaneously process multiple pairs of data, rather than sequentially process every pair of linearly-adjacent data. In this paper, we establish the mathematical model of Probe Machine as a 9-tuple consisting of data library, probe library, data controller, probe controller, probe operation, computing platform, detector, true solution storage, and residue collector. We analyze the computation capability of the Probe Machine model, and in particular we show that Turing Machine is a special case of Probe Machine. We revisit two NP-complete problems---i.e., the Graph Coloring and Hamilton Cycle problems, and devise two algorithms on basis of the established Probe Machine model, which can enumerate all solutions to each of these problems by only one probe operation. Furthermore, we show that Probe Machine can be implemented by leveraging the nano-DNA probe technologies. The computational power of an electronic computer based on Turing Machine is known far more than that of the human brain. A question naturally arises: will a future computer based on Probe Machine outperform the human brain in more ways beyond the computational power?"
                    },
                    {
                        "title": "55- and 56-configurations are reducible",
                        "abstract": "Let $G$ be a 4-chromatic maximal planar graph (MPG) with the minimum degree of at least 4 and let $C$ be an even-length cycle of $G$.If $|f(C)|=2$ for every $f$ in some Kempe equivalence class of $G$, then we call $C$ an unchanged bichromatic cycle (UBC) of $G$, and correspondingly $G$ an unchanged bichromatic cycle maximal planar graph (UBCMPG) with respect to $C$, where $f(C)=\\{f(v)| v\\in V(C)\\}$. For an UBCMPG $G$ with respect to an UBC $C$, the subgraph of $G$ induced by the set of edges belonging to $C$ and its interior (or exterior), denoted by $G^C$, is called a base-module of $G$; in particular, when the length of $C$ is equal to four, we use $C_4$ instead of $C$ and call $G^{C_4}$ a 4-base-module. In this paper, we first study the properties of UBCMPGs and show that every 4-base-module $G^{C_4}$ contains a 4-coloring under which $C_4$ is bichromatic and there are at least two bichromatic paths with different colors between one pair of diagonal vertices of $C_4$ (these paths are called module-paths). We further prove that every 4-base-module $G^{C_4}$ contains a 4-coloring (called decycle coloring) for which the ends of a module-path are colored by distinct colors. Finally, based on the technique of the contracting and extending operations of MPGs, we prove that 55-configurations and 56-configurations are reducible by converting the reducibility problem of these two classes of configurations into the decycle coloring problem of 4-base-modules."
                    },
                    {
                        "title": "Theory on Structure and Coloring of Maximal Planar Graphs (I): Relationship between Structure and Coloring",
                        "abstract": "Maximal planar graph refers to the planar graph with the most edges, which means no more edges can be added so that the resulting graph is still planar. The Four-Color Conjecture says that every planar graph without loops is 4-colorable. Indeed, in order to prove Four-Color Conjecture, it clearly suffices to show that all maximal planar graphs are 4-colorable. Since this conjecture was proposed in 1852, no mathematical proofs have been invented up until now. Maybe the main reasons lie in the following three aspects in terms of maximal planar graphs: not clearing up the structures, not figuring out the coloring types, and not straightening out the relation between structure and coloring. For this, we will write a series of articles to study the structure and coloring theory of maximal planar graphs systematically. This is our first article, which focuses mainly on the structure and coloring relations. First, we introduce a new way to construct maximal planar graphs. The advantage of this method is that it establishes an immediate relation with 4-colorings, and reveals how a given maximal planar graph is generated. Second, a special class of maximal planar graphs--recursive maximal planar graphs is researched in depth, which lays a foundation for solving the uniquely 4-colorable planar graphs conjecture(see subsequent articles). Third, we discover an important mode for classifying 4-colorings: tree-coloring and cycle-coloring, which runs through the whole series of articles. Furthermore, this mode is applied to the research on an arbitrary 4-coloring and its corresponding structure of unions of two and three bicolored subgraphs. Finally, we introduce the concepts of black-white coloring and stamen phenomenon, and find out a necessary and sufficient condition for an even cycle to be a 2-colorable cycle."
                    },
                    {
                        "title": "Immunological recognition by artificial neural networks",
                        "abstract": "The binding affinity between the T-cell receptors (TCRs) and antigenic peptides mainly determines immunological recognition. It is not a trivial task that T cells identify the digital sequences of peptide amino acids by simply relying on the integrated binding affinity between TCRs and antigenic peptides. To address this problem, we examine whether the affinity-based discrimination of peptide sequences is learnable and generalizable by artificial neural networks (ANNs) that process the digital experimental amino acid sequence information of receptors and peptides. A pair of TCR and peptide sequences correspond to the input for ANNs, while the success or failure of the immunological recognition correspond to the output. The output is obtained by both theoretical model and experimental data. In either case, we confirmed that ANNs could learn the immunological recognition. We also found that a homogenized encoding of amino acid sequence was more effective for the supervised learning task."
                    },
                    {
                        "title": "Peak Age of Information in Priority Queueing Systems",
                        "abstract": "We consider a priority queueing system where a single processor serves k classes of packets that are generated randomly following Poisson processes. Our objective is to compute the expected Peak Age of Information (PAoI) under various scenarios. In particular, we consider two situations where the buffer size at each queue is one and infinite, and in the infinite buffer size case we consider First Come First Serve (FCFS) and Last Come First Serve (LCFS) as service disciplines. For the system with buffer size one at each queue, we derive PAoI exactly for the case of exponential service time and bounds (which are excellent approximations) for the case of general service time, with small k. For the system with infinite buffer size, we provide closed-form expressions of PAoI for both FCFS and LCFS where service time is general and k could be large. Using those results we investigated the effect of ordering of priorities and service disciplines for the various scenarios. We perform extensive numerical studies to validate our results and develop insights."
                    },
                    {
                        "title": "On Competitive Analysis for Polling Systems",
                        "abstract": "Polling systems have been widely studied, however most of these studies focus on polling systems with renewal processes for arrivals and random variables for service times. There is a need driven by practical applications to study polling systems with arbitrary arrivals (not restricted to time-varying or in batches) and revealed service time upon a job's arrival. To address that need, our work considers a polling system with generic setting and for the first time provides the worst-case analysis for online scheduling policies in this system. We provide conditions for the existence of constant competitive ratios, and competitive lower bounds for general scheduling policies in polling systems. Our work also bridges the queueing and scheduling communities by proving the competitive ratios for several well-studied policies in the queueing literature, such as cyclic policies with exhaustive, gated or l-limited service disciplines for polling systems."
                    },
                    {
                        "title": "Broad cross-reactivity of the T-cell repertoire achieves specific and sufficiently rapid target searching",
                        "abstract": "The molecular recognition of T-cell receptors is the hallmark of the adaptive immunity. Given the finiteness of the T-cell repertoire, individual T-cell receptors are necessary to be cross-reactive to multiple antigenic peptides. In this study, we quantify the variability of the cross-reactivity by using a string model that estimates the binding affinity between two sequences of amino acids. We examine sequences of 10,000 human T-cell receptors and 10,000 antigenic peptides, and obtain a full spectrum of cross-reactivity of the receptor-peptide binding. Then, we find that the cross-reactivity spectrum is broad. Some T cells are reactive to 1,000 peptides, but some T cells are reactive to only one or two peptides. Since the degree of cross-reactivity has a correlation with the (un)binding affinity of receptors, we further investigate how the broad crossreactivity affects the target searching of T cells. High cross-reactive T cells may not require many trials for searching correct targets, but they may spend long time to unbind from incorrect targets. In contrast, low cross-reactive T cells may not spend long time to ignore incorrect targets, but they require many trials for screening correct targets. We evaluate this hypothesis, and show that the broad cross-reactivity of the natural T-cell repertoire can balance the trade-off between the rapid screening and unbinding penalty."
                    }
                ]
            },
            "076b77e0-0a1e-47d9-ac6c-4054932ebb81": {
                "pk": "076b77e0-0a1e-47d9-ac6c-4054932ebb81",
                "name": "William Gay",
                "collaborators": [
                    "William He",
                    "Nicholas Kocurek",
                    "Anirudh Sundar",
                    "Christopher Richardson",
                    "Larry Heck"
                ],
                "domain": [
                    "Cryptography",
                    "Machine Learning",
                    "Natural Language Processing",
                    "Interactive Systems"
                ],
                "publications": [
                    {
                        "title": "Faster Mixing of Higher-Dimensional Random Reversible Circuits",
                        "abstract": "We continue the study of the approximate $k$-wise independence of random reversible circuits as permutations of $\\{\\pm1\\}^n$. Our main result is the first construction of a natural class of random reversible circuits with a sublinear-in-$n$ dependence on depth. Our construction is motivated by considerations in practical cryptography and is somewhat inspired by the design of practical block ciphers, such as DES and AES. Previous constructions of He and O'Donnell [HO24], which were built with gate architectures on one-dimensional lattices, suffered from an inherent linear-in-$n$ dependence on depth. The main novelty of our circuit model is a gate architecture built on higher-dimensional lattices."
                    },
                    {
                        "title": "iTBLS: A Dataset of Interactive Conversations Over Tabular Information",
                        "abstract": "This paper introduces Interactive Tables (iTBLS), a dataset of interactive conversations situated in tables from scientific articles. This dataset is designed to facilitate human-AI collaborative problem-solving through AI-powered multi-task tabular capabilities. In contrast to prior work that models interactions as factoid QA or procedure synthesis, iTBLS broadens the scope of interactions to include mathematical reasoning, natural language manipulation, and expansion of existing tables from natural language conversation by delineating interactions into one of three tasks: interpretation, modification, or generation. Additionally, the paper presents a suite of baseline approaches to iTBLS, utilizing zero-shot prompting and parameter-efficient fine-tuning for different computing situations. We also introduce a novel multi-step approach and show how it can be leveraged in conjunction with parameter-efficient fine-tuning to achieve the state-of-the-art on iTBLS; outperforming standard parameter-efficient fine-tuning by up to 15% on interpretation, 18% on modification, and 38% on generation."
                    }
                ]
            },
            "1dde6cf0-6ed1-4bc2-bcf3-c838b4bffb9b": {
                "pk": "1dde6cf0-6ed1-4bc2-bcf3-c838b4bffb9b",
                "name": "Christopher Richardson",
                "collaborators": [
                    "Larry Heck",
                    "Anirudh Sundar",
                    "William Gay"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Conversational AI",
                    "Commonsense Reasoning",
                    "Table Generation"
                ],
                "publications": [
                    {
                        "title": "Commonsense Reasoning for Conversational AI: A Survey of the State of the Art",
                        "abstract": "Large, transformer-based pretrained language models like BERT, GPT, and T5 have demonstrated a deep understanding of contextual semantics and language syntax. Their success has enabled significant advances in conversational AI, including the development of open-dialogue systems capable of coherent, salient conversations which can answer questions, chat casually, and complete tasks. However, state-of-the-art models still struggle with tasks that involve higher levels of reasoning - including commonsense reasoning that humans find trivial. This paper presents a survey of recent conversational AI research focused on commonsense reasoning. The paper lists relevant training datasets and describes the primary approaches to include commonsense in conversational AI. The paper also discusses benchmarks used for evaluating commonsense in conversational AI problems. Finally, the paper presents preliminary observations of the limited commonsense capabilities of two state-of-the-art open dialogue models, BlenderBot3 and LaMDA, and its negative effect on natural interactions. These observations further motivate research on commonsense reasoning in conversational AI."
                    },
                    {
                        "title": "SYNDICOM: Improving Conversational Commonsense with Error-Injection and Natural Language Feedback",
                        "abstract": "Commonsense reasoning is a critical aspect of human communication. Despite recent advances in conversational AI driven by large language models, commonsense reasoning remains a challenging task. In this work, we introduce SYNDICOM - a method for improving commonsense in dialogue response generation. SYNDICOM consists of two components. The first component is a dataset composed of commonsense dialogues created from a knowledge graph and synthesized into natural language. This dataset includes both valid and invalid responses to dialogue contexts, along with natural language feedback (NLF) for the invalid responses. The second contribution is a two-step procedure: training a model to predict natural language feedback (NLF) for invalid responses, and then training a response generation model conditioned on the predicted NLF, the invalid response, and the dialogue. SYNDICOM is scalable and does not require reinforcement learning. Empirical results on three tasks are evaluated using a broad range of metrics. SYNDICOM achieves a relative improvement of 53% over ChatGPT on ROUGE1, and human evaluators prefer SYNDICOM over ChatGPT 57% of the time. We will publicly release the code and the full dataset."
                    },
                    {
                        "title": "gTBLS: Generating Tables from Text by Conditional Question Answering",
                        "abstract": "Distilling large, unstructured text into a structured, condensed form such as tables is an open research problem. One of the primary challenges in automatically generating tables is ensuring their syntactic validity. Prior approaches address this challenge by including additional parameters in the Transformer's attention mechanism to attend to specific rows and column headers. In contrast to this single-stage method, this paper presents a two-stage approach called Generative Tables (gTBLS). The first stage infers table structure (row and column headers) from the text. The second stage formulates questions using these headers and fine-tunes a causal language model to answer them. Furthermore, the gTBLS approach is amenable to the utilization of pre-trained Large Language Models in a zero-shot configuration, presenting a solution for table generation in situations where fine-tuning is not feasible. gTBLS improves prior approaches by up to 10% in BERTScore on the table construction task and up to 20% on the table content generation task of the E2E, WikiTableText, WikiBio, and RotoWire datasets."
                    },
                    {
                        "title": "iTBLS: A Dataset of Interactive Conversations Over Tabular Information",
                        "abstract": "This paper introduces Interactive Tables (iTBLS), a dataset of interactive conversations situated in tables from scientific articles. This dataset is designed to facilitate human-AI collaborative problem-solving through AI-powered multi-task tabular capabilities. In contrast to prior work that models interactions as factoid QA or procedure synthesis, iTBLS broadens the scope of interactions to include mathematical reasoning, natural language manipulation, and expansion of existing tables from natural language conversation by delineating interactions into one of three tasks: interpretation, modification, or generation. Additionally, the paper presents a suite of baseline approaches to iTBLS, utilizing zero-shot prompting and parameter-efficient fine-tuning for different computing situations. We also introduce a novel multi-step approach and show how it can be leveraged in conjunction with parameter-efficient fine-tuning to achieve the state-of-the-art on iTBLS; outperforming standard parameter-efficient fine-tuning by up to 15% on interpretation, 18% on modification, and 38% on generation."
                    }
                ]
            },
            "fc2fbbac-0de9-4564-8f81-135d95de45bb": {
                "pk": "fc2fbbac-0de9-4564-8f81-135d95de45bb",
                "name": "Larry Heck",
                "collaborators": [
                    "Dilek Hakkani-Tur",
                    "Anirudh Sundar",
                    "Christopher Richardson",
                    "Saif Punjwani",
                    "Mason Bretan",
                    "Ankur Bapna",
                    "Gokhan Tur",
                    "Benjamin Reichman",
                    "Tyler Lizzo",
                    "Simon Heck"
                ],
                "domain": [
                    "Multimodal AI",
                    "Conversational AI",
                    "Commonsense Reasoning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "Multimodal Conversational AI: A Survey of Datasets and Approaches",
                        "abstract": "As humans, we experience the world with all our senses or modalities (sound, sight, touch, smell, and taste). We use these modalities, particularly sight and touch, to convey and interpret specific meanings. Multimodal expressions are central to conversations; a rich set of modalities amplify and often compensate for each other. A multimodal conversational AI system answers questions, fulfills tasks, and emulates human conversations by understanding and expressing itself via multiple modalities. This paper motivates, defines, and mathematically formulates the multimodal conversational research objective. We provide a taxonomy of research required to solve the objective: multimodal representation, fusion, alignment, translation, and co-learning. We survey state-of-the-art datasets and approaches for each research area and highlight their limiting assumptions. Finally, we identify multimodal co-learning as a promising direction for multimodal conversational AI research."
                    },
                    {
                        "title": "Commonsense Reasoning for Conversational AI: A Survey of the State of the Art",
                        "abstract": "Large, transformer-based pretrained language models like BERT, GPT, and T5 have demonstrated a deep understanding of contextual semantics and language syntax. Their success has enabled significant advances in conversational AI, including the development of open-dialogue systems capable of coherent, salient conversations which can answer questions, chat casually, and complete tasks. However, state-of-the-art models still struggle with tasks that involve higher levels of reasoning - including commonsense reasoning that humans find trivial. This paper presents a survey of recent conversational AI research focused on commonsense reasoning. The paper lists relevant training datasets and describes the primary approaches to include commonsense in conversational AI. The paper also discusses benchmarks used for evaluating commonsense in conversational AI problems. Finally, the paper presents preliminary observations of the limited commonsense capabilities of two state-of-the-art open dialogue models, BlenderBot3 and LaMDA, and its negative effect on natural interactions. These observations further motivate research on commonsense reasoning in conversational AI."
                    },
                    {
                        "title": "Dense Passage Retrieval: Is it Retrieving?",
                        "abstract": "Dense passage retrieval (DPR) is the first step in the retrieval augmented generation (RAG) paradigm for improving the performance of large language models (LLM). DPR fine-tunes pre-trained networks to enhance the alignment of the embeddings between queries and relevant textual data. A deeper understanding of DPR fine-tuning will be required to fundamentally unlock the full potential of this approach. In this work, we explore DPR-trained models mechanistically by using a combination of probing, layer activation analysis, and model editing. Our experiments show that DPR training decentralizes how knowledge is stored in the network, creating multiple access pathways to the same information. We also uncover a limitation in this training style: the internal knowledge of the pre-trained model bounds what the retrieval model can retrieve. These findings suggest a few possible directions for dense retrieval: (1) expose the DPR training process to more knowledge so more can be decentralized, (2) inject facts as decentralized representations, (3) model and incorporate knowledge uncertainty in the retrieval process, and (4) directly map internal model knowledge to a knowledge base."
                    },
                    {
                        "title": "UNLEARN Efficient Removal of Knowledge in Large Language Models",
                        "abstract": "Given the prevalence of large language models (LLMs) and the prohibitive cost of training these models from scratch, dynamically forgetting specific knowledge e.g., private or proprietary, without retraining the model has become an important capability. This paper proposes a novel method to achieve this objective called UNLEARN. The approach builds upon subspace methods to identify and specifically target the removal of knowledge without adversely affecting other knowledge in the LLM. Results demonstrate 96% of targeted knowledge can be forgotten while maintaining performance on other knowledge within 2.5% of the original model, significantly outperforming the discriminatory abilities of the previous state-of-the-art. A dual method called LEARN is also proposed for targeted knowledge addition. Results show LEARN can match the fine-tuning accuracy of Low-Rank Adaptation (LoRA) without adversely affecting similar tasks."
                    },
                    {
                        "title": "Allo-AVA: A Large-Scale Multimodal Conversational AI Dataset for Allocentric Avatar Gesture Animation",
                        "abstract": "The scarcity of high-quality, multimodal training data severely hinders the creation of lifelike avatar animations for conversational AI in virtual environments. Existing datasets often lack the intricate synchronization between speech, facial expressions, and body movements that characterize natural human communication. To address this critical gap, we introduce Allo-AVA, a large-scale dataset specifically designed for text and audio-driven avatar gesture animation in an allocentric (third person point-of-view) context. Allo-AVA consists of $\\sim$1,250 hours of diverse video content, complete with audio, transcripts, and extracted keypoints. Allo-AVA uniquely maps these keypoints to precise timestamps, enabling accurate replication of human movements (body and facial gestures) in synchronization with speech. This comprehensive resource enables the development and evaluation of more natural, context-aware avatar animation models, potentially transforming applications ranging from virtual reality to digital assistants."
                    },
                    {
                        "title": "Large Body Language Models",
                        "abstract": "As virtual agents become increasingly prevalent in human-computer interaction, generating realistic and contextually appropriate gestures in real-time remains a significant challenge. While neural rendering techniques have made substantial progress with static scripts, their applicability to human-computer interactions remains limited. To address this, we introduce Large Body Language Models (LBLMs) and present LBLM-AVA, a novel LBLM architecture that combines a Transformer-XL large language model with a parallelized diffusion model to generate human-like gestures from multimodal inputs (text, audio, and video). LBLM-AVA incorporates several key components enhancing its gesture generation capabilities, such as multimodal-to-pose embeddings, enhanced sequence-to-sequence mapping with redefined attention mechanisms, a temporal smoothing module for gesture sequence coherence, and an attention-based refinement module for enhanced realism. The model is trained on our large-scale proprietary open-source dataset Allo-AVA. LBLM-AVA achieves state-of-the-art performance in generating lifelike and contextually appropriate gestures with a 30% reduction in Fr\\'echet Gesture Distance (FGD), and a 25% improvement in Fr\\'echet Inception Distance compared to existing approaches."
                    },
                    {
                        "title": "mForms : Multimodal Form-Filling with Question Answering",
                        "abstract": "This paper presents a new approach to form-filling by reformulating the task as multimodal natural language Question Answering (QA). The reformulation is achieved by first translating the elements on the GUI form (text fields, buttons, icons, etc.) to natural language questions, where these questions capture the element's multimodal semantics. After a match is determined between the form element (Question) and the user utterance (Answer), the form element is filled through a pre-trained extractive QA system. By leveraging pre-trained QA models and not requiring form-specific training, this approach to form-filling is zero-shot. The paper also presents an approach to further refine the form-filling by using multi-task training to incorporate a potentially large number of successive tasks. Finally, the paper introduces a multimodal natural language form-filling dataset Multimodal Forms (mForms), as well as a multimodal extension of the popular ATIS dataset to support future research and experimentation. Results show the new approach not only maintains robust accuracy for sparse training conditions but achieves state-of-the-art F1 of 0.97 on ATIS with approximately 1/10th of the training data."
                    },
                    {
                        "title": "Learning and Evaluating Musical Features with Deep Autoencoders",
                        "abstract": "In this work we describe and evaluate methods to learn musical embeddings. Each embedding is a vector that represents four contiguous beats of music and is derived from a symbolic representation. We consider autoencoding-based methods including denoising autoencoders, and context reconstruction, and evaluate the resulting embeddings on a forward prediction and a classification task."
                    },
                    {
                        "title": "Federated Control with Hierarchical Multi-Agent Deep Reinforcement Learning",
                        "abstract": "We present a framework combining hierarchical and multi-agent deep reinforcement learning approaches to solve coordination problems among a multitude of agents using a semi-decentralized model. The framework extends the multi-agent learning setup by introducing a meta-controller that guides the communication between agent pairs, enabling agents to focus on communicating with only one other agent at any step. This hierarchical decomposition of the task allows for efficient exploration to learn policies that identify globally optimal solutions even as the number of collaborating agents increases. We show promising initial experimental results on a simulated distributed scheduling problem."
                    },
                    {
                        "title": "Domain Adaptation of Recurrent Neural Networks for Natural Language Understanding",
                        "abstract": "The goal of this paper is to use multi-task learning to efficiently scale slot filling models for natural language understanding to handle multiple target tasks or domains. The key to scalability is reducing the amount of training data needed to learn a model for a new task. The proposed multi-task model delivers better performance with less data by leveraging patterns that it learns from the other tasks. The approach supports an open vocabulary, which allows the models to generalize to unseen words, which is particularly important when very little training data is used. A newly collected crowd-sourced data set, covering four different domains, is used to demonstrate the effectiveness of the domain adaptation and open vocabulary techniques."
                    },
                    {
                        "title": "Leveraging Semantic Web Search and Browse Sessions for Multi-Turn Spoken Dialog Systems",
                        "abstract": "Training statistical dialog models in spoken dialog systems (SDS) requires large amounts of annotated data. The lack of scalable methods for data mining and annotation poses a significant hurdle for state-of-the-art statistical dialog managers. This paper presents an approach that directly leverage billions of web search and browse sessions to overcome this hurdle. The key insight is that task completion through web search and browse sessions is (a) predictable and (b) generalizes to spoken dialog task completion. The new method automatically mines behavioral search and browse patterns from web logs and translates them into spoken dialog models. We experiment with naturally occurring spoken dialogs and large scale web logs. Our session-based models outperform the state-of-the-art method for entity extraction task in SDS. We also achieve better performance for both entity and relation extraction on web search queries when compared with nontrivial baselines."
                    },
                    {
                        "title": "Towards Zero-Shot Frame Semantic Parsing for Domain Scaling",
                        "abstract": "State-of-the-art slot filling models for goal-oriented human/machine conversational language understanding systems rely on deep learning methods. While multi-task training of such models alleviates the need for large in-domain annotated datasets, bootstrapping a semantic parsing model for a new domain using only the semantic frame, such as the back-end API or knowledge graph schema, is still one of the holy grail tasks of language understanding for dialogue systems. This paper proposes a deep learning based approach that can utilize only the slot description in context without the need for any labeled or unlabeled in-domain examples, to quickly bootstrap a new domain. The main idea of this paper is to leverage the encoding of the slot names and descriptions within a multi-task deep learned slot filling model, to implicitly align slots across domains. The proposed approach is promising for solving the domain scaling problem and eliminating the need for any manually annotated data or explicit schema alignment. Furthermore, our experiments on multiple domains show that this approach results in significantly better slot-filling performance when compared to using only in-domain data, especially in the low data regime."
                    },
                    {
                        "title": "A Unit Selection Methodology for Music Generation Using Deep Neural Networks",
                        "abstract": "Several methods exist for a computer to generate music based on data including Markov chains, recurrent neural networks, recombinancy, and grammars. We explore the use of unit selection and concatenation as a means of generating music using a procedure based on ranking, where, we consider a unit to be a variable length number of measures of music. We first examine whether a unit selection method, that is restricted to a finite size unit library, can be sufficient for encompassing a wide spectrum of music. We do this by developing a deep autoencoder that encodes a musical input and reconstructs the input by selecting from the library. We then describe a generative model that combines a deep structured semantic model (DSSM) with an LSTM to predict the next unit, where units consist of four, two, and one measures of music. We evaluate the generative model using objective metrics including mean rank and accuracy and with a subjective listening test in which expert musicians are asked to complete a forced-choiced ranking task. We compare our model to a note-level generative baseline that consists of a stacked LSTM trained to predict forward by one note."
                    },
                    {
                        "title": "Leveraging Deep Neural Networks and Knowledge Graphs for Entity Disambiguation",
                        "abstract": "Entity Disambiguation aims to link mentions of ambiguous entities to a knowledge base (e.g., Wikipedia). Modeling topical coherence is crucial for this task based on the assumption that information from the same semantic context tends to belong to the same topic. This paper presents a novel deep semantic relatedness model (DSRM) based on deep neural networks (DNN) and semantic knowledge graphs (KGs) to measure entity semantic relatedness for topical coherence modeling. The DSRM is directly trained on large-scale KGs and it maps heterogeneous types of knowledge of an entity from KGs to numerical feature vectors in a latent space such that the distance between two semantically-related entities is minimized. Compared with the state-of-the-art relatedness approach proposed by (Milne and Witten, 2008a), the DSRM obtains 19.4% and 24.5% reductions in entity disambiguation errors on two publicly available datasets respectively."
                    },
                    {
                        "title": "Sequential Dialogue Context Modeling for Spoken Language Understanding",
                        "abstract": "Spoken Language Understanding (SLU) is a key component of goal oriented dialogue systems that would parse user utterances into semantic frame representations. Traditionally SLU does not utilize the dialogue history beyond the previous system turn and contextual ambiguities are resolved by the downstream components. In this paper, we explore novel approaches for modeling dialogue context in a recurrent neural network (RNN) based language understanding system. We propose the Sequential Dialogue Encoder Network, that allows encoding context from the dialogue history in chronological order. We compare the performance of our proposed architecture with two context models, one that uses just the previous turn context and another that encodes dialogue context in a memory network, but loses the order of utterances in the dialogue history. Experiments with a multi-domain dialogue dataset demonstrate that the proposed architecture results in reduced semantic frame error rates."
                    },
                    {
                        "title": "cTBLS: Augmenting Large Language Models with Conversational Tables",
                        "abstract": "Optimizing accuracy and performance while eliminating hallucinations of open-domain conversational large language models (LLMs) is an open research challenge. A particularly promising direction is to augment and ground LLMs with information from structured sources. This paper introduces Conversational Tables (cTBLS), a three-step architecture to retrieve and generate dialogue responses grounded on retrieved tabular information. cTBLS uses Transformer encoder embeddings for Dense Table Retrieval and obtains up to 125% relative improvement over the retriever in the previous state-of-the-art system on the HyrbiDialogue dataset. cTBLS then uses a shared process between encoder and decoder models to perform a coarse+fine tabular knowledge (e.g., cell) ranking combined with a GPT-3.5 LLM response generator to yield a 2x relative improvement in ROUGE scores. Finally, human evaluators prefer cTBLs +80% of the time (coherency, fluency) and judge informativeness to be 4x better than the previous state-of-the-art."
                    },
                    {
                        "title": "SYNDICOM: Improving Conversational Commonsense with Error-Injection and Natural Language Feedback",
                        "abstract": "Commonsense reasoning is a critical aspect of human communication. Despite recent advances in conversational AI driven by large language models, commonsense reasoning remains a challenging task. In this work, we introduce SYNDICOM - a method for improving commonsense in dialogue response generation. SYNDICOM consists of two components. The first component is a dataset composed of commonsense dialogues created from a knowledge graph and synthesized into natural language. This dataset includes both valid and invalid responses to dialogue contexts, along with natural language feedback (NLF) for the invalid responses. The second contribution is a two-step procedure: training a model to predict natural language feedback (NLF) for invalid responses, and then training a response generation model conditioned on the predicted NLF, the invalid response, and the dialogue. SYNDICOM is scalable and does not require reinforcement learning. Empirical results on three tasks are evaluated using a broad range of metrics. SYNDICOM achieves a relative improvement of 53% over ChatGPT on ROUGE1, and human evaluators prefer SYNDICOM over ChatGPT 57% of the time. We will publicly release the code and the full dataset."
                    }
                ]
            }
        }
    },
    "2402.08365": {
        "paper_data": {
            "title": "Learning Better Representations From Less Data For Propositional Satisfiability",
            "url": "http://arxiv.org/abs/2402.08365v2",
            "arxiv_id": "2402.08365",
            "authors": [
                "Mohamed Ghanem",
                "Frederik Schmitt",
                "Julian Siber",
                "Bernd Finkbeiner"
            ],
            "abstract": "Training neural networks on NP-complete problems typically demands very large amounts of training data and often needs to be coupled with computationally expensive symbolic verifiers to ensure output correctness. In this paper, we present NeuRes, a neuro-symbolic approach to address both challenges for propositional satisfiability, being the quintessential NP-complete problem. By combining certificate-driven training and expert iteration, our model learns better representations than models trained for classification only, with a much higher data efficiency -- requiring orders of magnitude less training data. NeuRes employs propositional resolution as a proof system to generate proofs of unsatisfiability and to accelerate the process of finding satisfying truth assignments, exploring both possibilities in parallel. To realize this, we propose an attention-based architecture that autoregressively selects pairs of clauses from a dynamic formula embedding to derive new clauses. Furthermore, we employ expert iteration whereby model-generated proofs progressively replace longer teacher proofs as the new ground truth. This enables our model to reduce a dataset of proofs generated by an advanced solver by ~32% after training on it with no extra guidance. This shows that NeuRes is not limited by the optimality of the teacher algorithm owing to its self-improving workflow. We show that our model achieves far better performance than NeuroSAT in terms of both correctly classified and proven instances.",
            "introduction": "   1 Introduction  Boolean satisfiability (SAT) is a fundamental problem in computer science. For theory, this stems from SAT being the first problem proven NP-complete\u00a0[12]. For practice, this is due to many highly-optimized SAT solvers being used as flexible reasoning engines in a variety of tasks such as model checking\u00a0[11, 46], software verification\u00a0[16], planning\u00a0[26], and mathematical proof search\u00a0[23]. Recently, SAT has also served as a litmus test for assessing the symbolic reasoning capabilities of neural models and a promising domain for neuro-symbolic systems\u00a0[42, 41, 1, 9, 35]. So far, neural models only provide limited, if any, justification for unsatisfiability predictions. NeuroCore\u00a0[41], for example, predicts an unsatisfiable core, the verification of which can be as hard as solving the original problem. No certificates at all or certificates that are hard to check limit neural methods in a domain where correctness is critical and prevents close integrations with symbolic methods. Therefore, we propose a neuro-symbolic model that utilizes resolution to solve SAT problems by generating easy-to-check certificates.   A resolution proof is a sequence of case distinctions, each involving two clauses, that ends in the empty clause (falsum). This technique can also be used to prove satisfiability by exhaustively applying it until no further new resolution steps are possible and the empty clause has not been derived. Generating such a proof is an interesting problem from a neuro-symbolic perspective because unlike other discrete combinatorial problems that have been considered before\u00a0[45, 6, 27, 29, 10], it requires selecting compatible pairs of clauses from the dynamically growing pool, as newly derived clauses are naturally considered for derivation in subsequent steps. In this work, we devise three attention-based mechanisms to perform this pair-selection needed for generating resolution proofs. In addition, we augment the architecture to efficiently handle sat (satisfiable) formulas with an assignment decoding mechanism that assigns a truth value to each literal. We hypothesize that, despite their final goals being in complete opposition, resolution and sat assignment finding can form a mutually beneficial collaboration. On the one hand, clauses derived by resolution incrementally inject additional information into the network, e.g., deriving a single-literal clause by resolution directly implies that literal should be true in any possible sat assignment. On the other hand, finding a sat assignment absolves the resolution network from having to prove satisfiability by exhaustion. On that basis, given an input formula, NeuRes proceeds in two parallel tracks: (1) finding a sat assignment, and (2) deriving a resolution proof of unsatisfiability. Both tracks operate on a shared representation of the problem state. Depending on which track succeeds, NeuRes produces the corresponding SAT verdict which is guaranteed to be sound by virtue of its certificate-based design. Since both of our certificate types are efficient to check, we can afford to perform these symbolic checks at each step. When comparing NeuRes with NeuroSAT\u00a0[42], which has been trained to predict satisfiability with millions of samples, we demonstrate that NeuRes achieves a higher accuracy while providing a proof and requires only thousands of training samples.   As for most problems in theorem proving we are not only interest in finding any proof but a short proof. Resolution proofs can",
            "references": [
                {
                    "title": "Magnushammer: A Transformer-based Approach to Premise Selection",
                    "abstract": "This paper presents a novel approach to premise selection, a crucial reasoning task in automated theorem proving. Traditionally, symbolic methods that rely on extensive domain knowledge and engineering effort are applied to this task. In contrast, this work demonstrates that contrastive training with the transformer architecture can achieve higher-quality retrieval of relevant premises, without the engineering overhead. Our method, Magnushammer, outperforms the most advanced and widely used automation tool in interactive theorem proving called Sledgehammer. On the PISA and miniF2F benchmarks Magnushammer achieves $59.5\\%$ (against $38.3\\%$) and $34.0\\%$ (against $20.9\\%$) success rates, respectively. By combining \\method with a language-model-based automated theorem prover, we further improve the state-of-the-art proof success rate from $57.0\\%$ to $71.0\\%$ on the PISA benchmark using $4$x fewer parameters. Moreover, we develop and open source a novel dataset for premise selection, containing textual representations of (proof state, relevant premise) pairs. To the best of our knowledge, this is the largest available premise selection dataset, and the first one for the Isabelle proof assistant."
                },
                {
                    "title": "Baldur: Whole-Proof Generation and Repair with Large Language Models",
                    "abstract": "Formally verifying software is a highly desirable but labor-intensive task. Recent work has developed methods to automate formal verification using proof assistants, such as Coq and Isabelle/HOL, e.g., by training a model to predict one proof step at a time and using that model to search through the space of possible proofs. This paper introduces a new method to automate formal verification: We use large language models, trained on natural language and code and fine-tuned on proofs, to generate whole proofs at once. We then demonstrate that a model fine-tuned to repair generated proofs further increasing proving power. This paper: (1) Demonstrates that whole-proof generation using transformers is possible and is as effective but more efficient than search-based techniques. (2) Demonstrates that giving the learned model additional context, such as a prior failed proof attempt and the ensuing error message, results in proof repair that further improves automated proof generation. (3) Establishes, together with prior work, a new state of the art for fully automated proof synthesis. We reify our method in a prototype, Baldur, and evaluate it on a benchmark of 6,336 Isabelle/HOL theorems and their proofs, empirically showing the effectiveness of whole-proof generation, repair, and added context. We also show that Baldur complements the state-of-the-art tool, Thor, by automatically generating proofs for an additional 8.7% of the theorems. Together, Baldur and Thor can prove 65.7% of the theorems fully automatically. This paper paves the way for new research into using large language models for automating formal verification."
                },
                {
                    "title": "Iterative Circuit Repair Against Formal Specifications",
                    "abstract": "We present a deep learning approach for repairing sequential circuits against formal specifications given in linear-time temporal logic (LTL). Given a defective circuit and its formal specification, we train Transformer models to output circuits that satisfy the corresponding specification. We propose a separated hierarchical Transformer for multimodal representation learning of the formal specification and the circuit. We introduce a data generation algorithm that enables generalization to more complex specifications and out-of-distribution datasets. In addition, our proposed repair mechanism significantly improves the automated synthesis of circuits from LTL specifications with Transformers. It improves the state-of-the-art by $6.8$ percentage points on held-out instances and $11.8$ percentage points on an out-of-distribution dataset from the annual reactive synthesis competition."
                },
                {
                    "title": "Formal Mathematics Statement Curriculum Learning",
                    "abstract": "We explore the use of expert iteration in the context of language modeling applied to formal mathematics. We show that at same compute budget, expert iteration, by which we mean proof search interleaved with learning, dramatically outperforms proof search only. We also observe that when applied to a collection of formal statements of sufficiently varied difficulty, expert iteration is capable of finding and solving a curriculum of increasingly difficult problems, without the need for associated ground-truth proofs. Finally, by applying this expert iteration to a manually curated set of problem statements, we achieve state-of-the-art on the miniF2F benchmark, automatically solving multiple challenging problems drawn from high school olympiads."
                },
                {
                    "title": "Neural Circuit Synthesis from Specification Patterns",
                    "abstract": "We train hierarchical Transformers on the task of synthesizing hardware circuits directly out of high-level logical specifications in linear-time temporal logic (LTL). The LTL synthesis problem is a well-known algorithmic challenge with a long history and an annual competition is organized to track the improvement of algorithms and tooling over time. New approaches using machine learning might open a lot of possibilities in this area, but suffer from the lack of sufficient amounts of training data. In this paper, we consider a method to generate large amounts of additional training data, i.e., pairs of specifications and circuits implementing them. We ensure that this synthetic data is sufficiently close to human-written specifications by mining common patterns from the specifications used in the synthesis competitions. We show that hierarchical Transformers trained on this synthetic data solve a significant portion of problems from the synthesis competitions, and even out-of-distribution examples from a recent case study."
                },
                {
                    "title": "Goal-Aware Neural SAT Solver",
                    "abstract": "Modern neural networks obtain information about the problem and calculate the output solely from the input values. We argue that it is not always optimal, and the network's performance can be significantly improved by augmenting it with a query mechanism that allows the network at run time to make several solution trials and get feedback on the loss value on each trial. To demonstrate the capabilities of the query mechanism, we formulate an unsupervised (not depending on labels) loss function for Boolean Satisfiability Problem (SAT) and theoretically show that it allows the network to extract rich information about the problem. We then propose a neural SAT solver with a query mechanism called QuerySAT and show that it outperforms the neural baseline on a wide range of SAT tasks."
                },
                {
                    "title": "Training a First-Order Theorem Prover from Synthetic Data",
                    "abstract": "A major challenge in applying machine learning to automated theorem proving is the scarcity of training data, which is a key ingredient in training successful deep learning models. To tackle this problem, we propose an approach that relies on training purely with synthetically generated theorems, without any human data aside from axioms. We use these theorems to train a neurally-guided saturation-based prover. Our neural prover outperforms the state-of-the-art E-prover on this synthetic data in both time and search steps, and shows significant transfer to the unseen human-written theorems from the TPTP library, where it solves 72\\% of first-order problems without equality."
                },
                {
                    "title": "Combinatorial optimization and reasoning with graph neural networks",
                    "abstract": "Combinatorial optimization is a well-established area in operations research and computer science. Until recently, its methods have mostly focused on solving problem instances in isolation, ignoring the fact that they often stem from related data distributions in practice. However, recent years have seen a surge of interest in using machine learning, especially graph neural networks, as a key building block for combinatorial tasks, either directly as solvers or by enhancing the former. This paper presents a conceptual review of recent key advancements in this emerging field, aiming at researchers in both optimization and machine learning."
                },
                {
                    "title": "Proofs of Unsatisfiability",
                    "abstract": "Satisfiability (SAT) solvers have become complex tools, which raises the question of whether we can trust their results. This question is particularly important when the solvers are used to determine the correctness of hardware and software and when they are used to produce mathematical results. To deal with this issue, solvers can provide proofs of unsatisfiability to certify the correctness of their answers. This chapter presents the history and state-of-the-art of producing and validating proofs of unsatisfiability. The chapter covers the most popular proof formats with and without hints to speed up certification. Hints in proofs make validation easy, which resulted in several efficient formally-verified checkers. Various proof systems are discussed, ranging from resolution to the recent propagation redundancy system. The chapter also describes techniques to compress and optimize proofs."
                },
                {
                    "title": "Conflict-Driven Clause Learning SAT Solvers",
                    "abstract": "One of the most important paradigm shifts in the use of SAT solvers for solving industrial problems has been the introduction of clause learning. Clause learning entails adding a new clause for each conflict during backtrack search. This new clause prevents the same conflict from occurring again during the search process. Moreover, sophisticated techniques such as the identification of unique implication points in a graph of implications, allow creating clauses that more precisely identify the assignments responsible for conflicts. Learned clauses often have a large number of literals. As a result, another paradigm shift has been the development of new data structures, namely lazy data structures, which are particularly effective at handling large clauses. These data structures are called lazy due to being in general unable to provide the actual status of a clause. Efficiency concerns and the use of lazy data structures motivated the introduction of dynamic heuristics that do not require knowing the precise status of clauses. This chapter describes the ingredients of conflict-driven clause learning SAT solvers, namely conflict analysis, lazy data structures, search restarts, conflict-driven heuristics and clause deletion strategies."
                },
                {
                    "title": "Generative Language Modeling for Automated Theorem Proving",
                    "abstract": "We explore the application of transformer-based language models to automated theorem proving. This work is motivated by the possibility that a major limitation of automated theorem provers compared to humans -- the generation of original mathematical terms -- might be addressable via generation from language models. We present an automated prover and proof assistant, GPT-f, for the Metamath formalization language, and analyze its performance. GPT-f found new short proofs that were accepted into the main Metamath library, which is to our knowledge, the first time a deep-learning based system has contributed proofs that were adopted by a formal mathematics community."
                },
                {
                    "title": "Predicting Propositional Satisfiability via End-to-End Learning",
                    "abstract": "Strangely enough, it is possible to use machine learning models to predict the satisfiability status of hard SAT problems with accuracy considerably higher than random guessing. Existing methods have relied on extensive, manual feature engineering and computationally complex features (e.g., based on linear programming relaxations). We show for the first time that even better performance can be achieved by end-to-end learning methods \u2014 i.e., models that map directly from raw problem inputs to predictions and take only linear time to evaluate. Our work leverages deep network models which capture a key invariance exhibited by SAT problems: satisfiability status is unaffected by reordering variables and clauses. We showed that end-to-end learning with deep networks can outperform previous work on random 3-SAT problems at the solubility phase transition, where: (1) exactly 50% of problems are satisfiable; and (2) empirical runtimes of known solution methods scale exponentially with problem size (e.g., we achieved 84% prediction accuracy on 600-variable problems, which take hours to solve with state-of-the-art methods). We also showed that deep networks can generalize across problem sizes (e.g., a network trained only on 100-variable problems, which typically take about 10 ms to solve, achieved 81% accuracy on 600-variable problems)."
                },
                {
                    "title": "Teaching Temporal Logics to Neural Networks",
                    "abstract": "We show that a deep neural network can learn the semantics of linear-time temporal logic (LTL). As a challenging task that requires deep understanding of the LTL semantics, we show that our network can solve the trace generation problem for LTL: given a satisfiable LTL formula, find a trace that satisfies the formula. We frame the trace generation problem for LTL as a translation task, i.e., to translate from formulas to satisfying traces, and train an off-the-shelf implementation of the Transformer, a recently introduced deep learning architecture proposed for solving natural language processing tasks. We provide a detailed analysis of our experimental results, comparing multiple hyperparameter settings and formula representations. After training for several hours on a single GPU the results were surprising: the Transformer returns the syntactically equivalent trace in 89% of the cases on a held-out test set. Most of the \"mispredictions\", however, (and overall more than 99% of the predicted traces) still satisfy the given LTL formula. In other words, the Transformer generalized from imperfect training data to the semantics of LTL."
                },
                {
                    "title": "Graph Representations for Higher-Order Logic and Theorem Proving",
                    "abstract": "This paper presents the first use of graph neural networks (GNNs) for higher-order proof search and demonstrates that GNNs can improve upon state-of-the-art results in this domain. Interactive, higher-order theorem provers allow for the formalization of most mathematical theories and have been shown to pose a significant challenge for deep learning. Higher-order logic is highly expressive and, even though it is well-structured with a clearly defined grammar and semantics, there still remains no well-established method to convert formulas into graph-based representations. In this paper, we consider several graphical representations of higher-order logic and evaluate them against the HOList benchmark for higher-order theorem proving."
                },
                {
                    "title": "Learning to Prove Theorems via Interacting with Proof Assistants",
                    "abstract": "Humans prove theorems by relying on substantial high-level reasoning and problem-specific insights. Proof assistants offer a formalism that resembles human mathematical reasoning, representing theorems in higher-order logic and proofs as high-level tactics. However, human experts have to construct proofs manually by entering tactics into the proof assistant. In this paper, we study the problem of using machine learning to automate the interaction with proof assistants. We construct CoqGym, a large-scale dataset and learning environment containing 71K human-written proofs from 123 projects developed with the Coq proof assistant. We develop ASTactic, a deep learning-based model that generates tactics as programs in the form of abstract syntax trees (ASTs). Experiments show that ASTactic trained on CoqGym can generate effective tactics and can be used to prove new theorems not previously provable by automated methods. Code is available at this https URL."
                },
                {
                    "title": "Learning To Solve Circuit-SAT: An Unsupervised Differentiable Approach",
                    "abstract": "Recent efforts to combine Representation Learning with Formal Methods, commonly known as Neuro-Symbolic Methods, have given rise to a new trend of applying rich neural architectures to solve classical combinatorial optimization problems. In this paper, we propose a neural framework that can learn to solve the Circuit Satisfiability problem. Our framework is built upon two fundamental contributions: a rich embedding architecture that encodes the problem structure, and an end-to-end differentiable training procedure that mimics Reinforcement Learning and trains the model directly toward solving the SAT problem. The experimental results show the superior out-of-sample generalization performance of our framework compared to the recently developed NeuroSAT method."
                },
                {
                    "title": "Learning Heuristics for Quantified Boolean Formulas through Reinforcement Learning",
                    "abstract": "We demonstrate how to learn efficient heuristics for automated reasoning algorithms for quantified Boolean formulas through deep reinforcement learning. We focus on a backtracking search algorithm, which can already solve formulas of impressive size - up to hundreds of thousands of variables. The main challenge is to find a representation of these formulas that lends itself to making predictions in a scalable way. For a family of challenging problems, we learned a heuristic that solves significantly more formulas compared to the existing handwritten heuristics."
                },
                {
                    "title": "Learning a SAT Solver from Single-Bit Supervision",
                    "abstract": "We present NeuroSAT, a message passing neural network that learns to solve SAT problems after only being trained as a classifier to predict satisfiability. Although it is not competitive with state-of-the-art SAT solvers, NeuroSAT can solve problems that are substantially larger and more difficult than it ever saw during training by simply running for more iterations. Moreover, NeuroSAT generalizes to novel distributions; after training only on random SAT problems, at test time it can solve SAT problems encoding graph coloring, clique detection, dominating set, and vertex cover problems, all on a range of distributions over small random graphs."
                },
                {
                    "title": "Premise Selection for Theorem Proving by Deep Graph Embedding",
                    "abstract": "We propose a deep learning-based approach to the problem of premise selection: selecting mathematical statements relevant for proving a given conjecture. We represent a higher-order logic formula as a graph that is invariant to variable renaming but still fully preserves syntactic and semantic information. We then embed the graph into a vector via a novel embedding method that preserves the information of edge ordering. Our approach achieves state-of-the-art results on the HolStep dataset, improving the classification accuracy from 83% to 90.3%."
                },
                {
                    "title": "The science of brute force",
                    "abstract": "Mathematics solves problems by pen and paper. CS helps us to go far beyond that."
                },
                {
                    "title": "Learning Combinatorial Optimization Algorithms over Graphs",
                    "abstract": "The design of good heuristics or approximation algorithms for NP-hard combinatorial optimization problems often requires significant specialized knowledge and trial-and-error. Can we automate this challenging, tedious process, and learn the algorithms instead? In many real-world applications, it is typically the case that the same optimization problem is solved again and again on a regular basis, maintaining the same problem structure but differing in the data. This provides an opportunity for learning heuristic algorithms that exploit the structure of such recurring problems. In this paper, we propose a unique combination of reinforcement learning and graph embedding to address this challenge. The learned greedy policy behaves like a meta-algorithm that incrementally constructs a solution, and the action is determined by the output of a graph embedding network capturing the current state of the solution. We show that our framework can be applied to a diverse range of optimization problems over graphs, and learns effective algorithms for the Minimum Vertex Cover, Maximum Cut and Traveling Salesman problems."
                },
                {
                    "title": "Deep Network Guided Proof Search",
                    "abstract": "Deep learning techniques lie at the heart of several significant AI advances in recent years including object recognition and detection, image captioning, machine translation, speech recognition and synthesis, and playing the game of Go.Automated first-order theorem provers can aid in the formalization and verification of mathematical theorems and play a crucial role in program analysis, theory reasoning, security, interpolation, and system verification.Here we suggest deep learning based guidance in the proof search of the theorem prover E. We train and compare several deep neural network models on the traces of existing ATP proofs of Mizar statements and use them to select processed clauses during proof search. We give experimental evidence that with a hybrid, two-phase approach, deep learning based guidance can significantly reduce the average number of proof search steps while increasing the number of theorems proved.Using a few proof guidance strategies that leverage deep neural networks, we have found first-order proofs of 7.36% of the first-order logic translations of the Mizar Mathematical Library theorems that did not previously have ATP generated proofs. This increases the ratio of statements in the corpus with ATP generated proofs from 56% to 59%."
                },
                {
                    "title": "Neural Combinatorial Optimization with Reinforcement Learning",
                    "abstract": "This paper presents a framework to tackle combinatorial optimization problems using neural networks and reinforcement learning. We focus on the traveling salesman problem (TSP) and train a recurrent network that, given a set of city coordinates, predicts a distribution over different city permutations. Using negative tour length as the reward signal, we optimize the parameters of the recurrent network using a policy gradient method. We compare learning the network parameters on a set of training graphs against learning them on individual test graphs. Despite the computational expense, without much engineering and heuristic designing, Neural Combinatorial Optimization achieves close to optimal results on 2D Euclidean graphs with up to 100 nodes. Applied to the KnapSack, another NP-hard problem, the same method obtains optimal solutions for instances with up to 200 items."
                },
                {
                    "title": "DeepMath - Deep Sequence Models for Premise Selection",
                    "abstract": "We study the effectiveness of neural sequence models for premise selection in automated theorem proving, one of the main bottlenecks in the formalization of mathematics. We propose a two stage approach for this task that yields good results for the premise selection task on the Mizar corpus while avoiding the hand-engineered features of existing state-of-the-art models. To our knowledge, this is the first time deep learning has been applied to theorem proving on a large scale."
                },
                {
                    "title": "Boolean Satisfiability Solvers and Their Applications in Model Checking",
                    "abstract": "Boolean satisfiability (SAT)-the problem of determining whether there exists an assignment satisfying a given Boolean formula-is a fundamental intractable problem in computer science. SAT has many applications in electronic design automation (EDA), notably in synthesis and verification. Consequently, SAT has received much attention from the EDA community, who developed algorithms that have had a significant impact on the performance of SAT solvers. EDA researchers introduced techniques such as conflict-driven clause learning, novel branching heuristics, and efficient unit propagation. These techniques form the basis of all modern SAT solvers. Using these ideas, contemporary SAT solvers can often handle practical instances with millions of variables and constraints. The continuing advances of SAT solvers are the driving force of modern model checking tools, which are used to check the correctness of hardware designs. Contemporary automated verification techniques such as bounded model checking, proof-based abstraction, interpolation-based model checking, and IC3 have in common that they are all based on SAT solvers and their extensions. In this paper, we trace the most important contributions made to modern SAT solvers by the EDA community, and discuss applications of SAT in hardware model checking."
                },
                {
                    "title": "Pointer Networks",
                    "abstract": "We introduce a new neural architecture to learn the conditional probability of an output sequence with elements that are discrete tokens corresponding to positions in an input sequence. Such problems cannot be trivially addressed by existent approaches such as sequence-to-sequence [1] and Neural Turing Machines [2], because the number of target classes in each step of the output depends on the length of the input, which is variable. Problems such as sorting variable sized sequences, and various combinatorial optimization problems belong to this class. Our model solves the problem of variable size output dictionaries using a recently proposed mechanism of neural attention. It differs from the previous attention attempts in that, instead of using attention to blend hidden units of an encoder to a context vector at each decoder step, it uses attention as a pointer to select a member of the input sequence as the output. We call this architecture a Pointer Net (Ptr-Net). We show Ptr-Nets can be used to learn approximate solutions to three challenging geometric problems - finding planar convex hulls, computing Delaunay triangulations, and the planar Travelling Salesman Problem - using training examples alone. Ptr-Nets not only improve over sequence-to-sequence with input attention, but also allow us to generalize to variable size output dictionaries. We show that the learnt models generalize beyond the maximum lengths they were trained on. We hope our results on these tasks will encourage a broader exploration of neural learning for discrete problems."
                },
                {
                    "title": "Bridging the gap between easy generation and efficient verification of unsatisfiability proofs",
                    "abstract": "Several proof formats have been used to verify refutations produced by satisfiability (SAT) solvers. Existing formats are either costly to check or hard to implement. This paper presents a practical approach that facilitates checking of unsatisfiability results in a time similar to proof discovery by embedding clause deletion information into clausal proofs. By exploiting this information, the proof\u2010checking time is reduced by an order of magnitude on medium\u2010to\u2010hard benchmarks as compared to checking proofs using similar clausal formats. Proofs in a new format can be produced by making only minor changes to existing conflict\u2010driven clause\u2010learning solvers and their preprocessors, and the runtime overhead is negligible. This approach can easily be integrated into Glucose 2.1, the SAT 2012 challenge winner, and SatELite, a popular SAT\u2010problem preprocessor. Copyright \u00a9 2014 John Wiley & Sons, Ltd."
                },
                {
                    "title": "Understanding the exploding gradient problem",
                    "abstract": "Training Recurrent Neural Networks is more troublesome than feedforward ones because of the vanishing and exploding gradient problems detailed in Bengio et al. (1994). In this paper we attempt to understand the fundamental issues underlying the exploding gradient problem by exploring it from an analytical, a geometric and a dynamical system perspective. Our analysis is used to justify the simple yet effective solution of norm clipping the exploded gradient. In the experimental section, the comparison between this heuristic solution and standard SGD provides empirical evidence towards our hypothesis as well as it shows that such a heuristic is required to reach state of the art results on a character prediction task and a polyphonic music prediction one."
                },
                {
                    "title": "Verification of proofs of unsatisfiability for CNF formulas",
                    "abstract": "As SAT-algorithms become more and more complex, there is little chance of writing a SAT-solver that is free of bugs. So it is of great importance to be able to verify, the information returned by a SAT-solver. If the CNF formula to be tested is satisfiable, solution verification is trivial and can be easily done by the user. However, in the case of unsatisfiability, the user has to rely on the reputation of the SAT-solver. We describe an efficient procedure for checking the correctness of unsatisfiability proofs. As a by-product, the proposed procedure finds an unsatisfiable core of the initial CNF formula. The efficiency of the proposed procedure was tested on a representative set of large \"real-life\" CNF formulas from the formal verification domain."
                },
                {
                    "title": "Validating SAT solvers using an independent resolution-based checker: practical implementations and other applications",
                    "abstract": "As the use of SAT solvers as core engines in EDA applications grows, it becomes increasingly important to validate their correctness. In this paper, we describe the implementation of an independent resolution-based checking procedure that can check the validity of unsatisfiable claims produced by the SAT solver zchaff. We examine the practical implementation issues of such a checker and describe two implementations with different pros and cons. Experimental results show low overhead for the checking process. Our checker can work with many other modern SAT solvers with minor modifications, and it can provide information for debugging when checking fails. Finally we describe additional results that can be obtained by the validation process and briefly discuss their applications."
                },
                {
                    "title": "Planning as Satisfiability",
                    "abstract": "SATPLAN04 is a updated version of the planning as satisfiability approach originally proposed in (Kautz & Selman 1992; 1996) using hand-generated translations, and implemented for PDDL input in the blackbox system (Kautz & Selman 1999). Like blackbox, SATPLAN04 accepts the STRIPS subset of PDDL and finds solutions with minimal parallel length: that is, many (non-interferring) actions may occur in parallel at each time step, and the total number of time steps in guaranteed to be as small as possible. Also like blackbox, SATPLAN works by:"
                },
                {
                    "title": "The complexity of theorem-proving procedures",
                    "abstract": "It is shown that any recognition problem solved by a polynomial time-bounded nondeterministic Turing machine can be \u201creduced\u201d to the problem of determining whether a given propositional formula is a tautology. Here \u201creduced\u201d means, roughly speaking, that the first problem can be solved deterministically in polynomial time provided an oracle is available for solving the second. From this notion of reducible, polynomial degrees of difficulty are defined, and it is shown that the problem of determining tautologyhood has the same polynomial degree as the problem of determining whether the first of two given graphs is isomorphic to a subgraph of the second. Other examples are discussed. A method of measuring the complexity of proof procedures for the predicate calculus is introduced and discussed."
                },
                {
                    "title": "A Computing Procedure for Quantification Theory",
                    "abstract": "The hope that mathematical methods employed in the investigation of formal logic would lead to purely computational methods for obtaining mathematical theorems goes back to Leibniz and has been revived by Peano around the turn of the century and by Hilbert's school in the 1920's. Hilbert, noting that all of classical mathematics could be formalized within quantification theory, declared that the problem of finding an algorithm for determining whether or not a given formula of quantification theory is valid was the central problem of mathematical logic. And indeed, at one time it seemed as if investigations of this \u201cdecision\u201d problem were on the verge of success. However, it was shown by Church and by Turing that such an algorithm can not exist. This result led to considerable pessimism regarding the possibility of using modern digital computers in deciding significant mathematical questions. However, recently there has been a revival of interest in the whole question. Specifically, it has been realized that while no decision procedure exists for quantification theory there are many proof procedures available\u2014that is, uniform procedures which will ultimately locate a proof for any formula of quantification theory which is valid but which will usually involve seeking \u201cforever\u201d in the case of a formula which is not valid\u2014and that some of these proof procedures could well turn out to be feasible for use with modern computing machinery.\n Hao Wang [9] and P. C. Gilmore [3] have each produced working programs which employ proof procedures in quantification theory. Gilmore's program employs a form of a basic theorem of mathematical logic due to Herbrand, and Wang's makes use of a formulation of quantification theory related to those studied by Gentzen. However, both programs encounter decisive difficulties with any but the simplest formulas of quantification theory, in connection with methods of doing propositional calculus. Wang's program, because of its use of Gentzen-like methods, involves exponentiation on the total number of truth-functional connectives, whereas Gilmore's program, using normal forms, involves exponentiation on the number of clauses present. Both methods are superior in many cases to truth table methods which involve exponentiation on the total number of variables present, and represent important initial contributions, but both run into difficulty with some fairly simple examples.\nIn the present paper, a uniform proof procedure for quantification theory is given which is feasible for use with some rather complicated formulas and which does not ordinarily lead to exponentiation. The superiority of the present procedure over those previously available is indicated in part by the fact that a formula on which Gilmore's routine for the IBM 704 causes the machine to computer for 21 minutes without obtaining a result was worked successfully by hand computation using the present method in 30 minutes. Cf. \u00a76, below.\nIt should be mentioned that, before it can be hoped to employ proof procedures for quantification theory in obtaining proofs of theorems belonging to \u201cgenuine\u201d mathematics, finite axiomatizations, which are \u201cshort,\u201d must be obtained for various branches of mathematics. This last question will not be pursued further here; cf., however, Davis and Putnam [2], where one solution to this problem is given for ele"
                },
                {
                    "title": "Learning to Solve SMT Formulas",
                    "abstract": "We present a new approach for learning to solve SMT formulas. We phrase the challenge of solving SMT formulas as a tree search problem where at each step a transformation is applied to the input formula until the formula is solved. Our approach works in two phases: first, given a dataset of unsolved formulas we learn a policy that for each formula selects a suitable transformation to apply at each step in order to solve the formula, and second, we synthesize a strategy in the form of a loop-free program with branches. This strategy is an interpretable representation of the policy decisions and is used to guide the SMT solver to decide formulas more efficiently, without requiring any modification to the solver itself and without needing to evaluate the learned policy at inference time. We show that our approach is effective in practice - it solves 17% more formulas over a range of benchmarks and achieves up to 100x runtime improvement over a state-of-the-art SMT solver."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.LO"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a neuro-symbolic model that effectively generates easy-to-check resolution proofs for Boolean satisfiability (SAT) problems while simultaneously finding satisfiable assignments?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the research community's understanding of neuro-symbolic systems and their integration with traditional symbolic reasoning methods. By providing a framework that generates verifiable certificates for SAT problems, this research could enhance the reliability of neural models in critical applications such as software verification and model checking. Furthermore, it could pave the way for future research into more complex reasoning tasks, ultimately leading to more robust AI systems capable of handling intricate logical problems.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to dynamically select compatible pairs of clauses from a growing pool while generating resolution proofs. Naive approaches may fail because they do not account for the interdependencies between clauses and the need for efficient pair selection. Additionally, the complexity of ensuring that both the resolution proof generation and satisfiable assignment finding work in tandem presents significant technical and theoretical obstacles. The requirement for correctness in SAT solutions further complicates the design of effective algorithms.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either resolution proofs or satisfiable assignment finding in isolation, leading to a lack of integrated approaches that leverage the strengths of both methods. Existing solutions may have limitations in their ability to generate easily verifiable certificates or may require extensive training data, which has hindered their practical applicability. Our approach differs by proposing a dual-track system that simultaneously addresses both aspects, allowing for a more efficient and effective resolution of SAT problems.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, NeuRes, utilizes three attention-based mechanisms for clause pair selection to generate resolution proofs. It operates on a shared representation of the problem state and runs two parallel tracks: one for finding a satisfiable assignment and another for deriving a resolution proof of unsatisfiability. We will evaluate NeuRes using benchmark SAT datasets and measure its performance based on accuracy and the efficiency of the generated proofs. The expected outcomes include improved accuracy over existing models like NeuroSAT, with a significant reduction in the required training samples, while ensuring that the generated certificates are easy to check for correctness."
            }
        },
        "author_data": {
            "8d05bfb8-615e-4b00-822e-80c50456ecc6": {
                "pk": "8d05bfb8-615e-4b00-822e-80c50456ecc6",
                "name": "Mohamed Ghanem",
                "collaborators": [
                    "Dauod Siniora",
                    "Fadi Dawoud",
                    "Habiba Gamal",
                    "Eslam Soliman",
                    "Hossam Sharara",
                    "Tamer El-Batt",
                    "Alaa Anani",
                    "Lotfy Abdel Khaliq"
                ],
                "domain": [
                    "Theoretical Computer Science",
                    "Machine Learning",
                    "Federated Learning",
                    "Security"
                ],
                "publications": [
                    {
                        "title": "On Theoretical Complexity and Boolean Satisfiability",
                        "abstract": "Theoretical complexity is a vital subfield of computer science that enables us to mathematically investigate computation and answer many interesting queries about the nature of computational problems. It provides theoretical tools to assess time and space requirements of computations along with assessing the difficultly of problems - classifying them accordingly. It also garners at its core one of the most important problems in mathematics, namely, the $\\textbf{P vs. NP}$ millennium problem. In essence, this problem asks whether solution and verification reside on two different levels of difficulty. In this thesis, we introduce some of the most central concepts in the Theory of Computing, giving an overview of how computation can be abstracted using Turing machines. Further, we introduce the two most famous problem complexity classes $\\textbf{P}$ and $\\textbf{NP}$ along with the relationship between them. In addition, we explicate the concept of problem reduction and how it is an essential tool for making hardness comparisons between different problems. Later, we present the problem of Boolean Satisfiability (SAT) which lies at the center of NP-complete problems. We then explore some of its tractable as well as intractable variants such as Horn-SAT and 3-SAT, respectively. Last but not least, we establish polynomial-time reductions from 3-SAT to some of the famous NP-complete graph problems, namely, Clique Finding, Hamiltonian Cycle Finding, and 3-Coloring."
                    },
                    {
                        "title": "FLoBC: A Decentralized Blockchain-Based Federated Learning Framework",
                        "abstract": "The rapid expansion of data worldwide invites the need for more distributed solutions in order to apply machine learning on a much wider scale. The resultant distributed learning systems can have various degrees of centralization. In this work, we demonstrate our solution FLoBC for building a generic decentralized federated learning system using blockchain technology, accommodating any machine learning model that is compatible with gradient descent optimization. We present our system design comprising the two decentralized actors: trainer and validator, alongside our methodology for ensuring reliable and efficient operation of said system. Finally, we utilize FLoBC as an experimental sandbox to compare and contrast the effects of trainer-to-validator ratio, reward-penalty policy, and model synchronization schemes on the overall system performance, ultimately showing by example that a decentralized federated learning system is indeed a feasible alternative to more centralized architectures."
                    },
                    {
                        "title": "Breaking the De-Pois Poisoning Defense",
                        "abstract": "Attacks on machine learning models have been, since their conception, a very persistent and evasive issue resembling an endless cat-and-mouse game. One major variant of such attacks is poisoning attacks which can indirectly manipulate an ML model. It has been observed over the years that the majority of proposed effective defense models are only effective when an attacker is not aware of them being employed. In this paper, we show that the attack-agnostic De-Pois defense is hardly an exception to that rule. In fact, we demonstrate its vulnerability to the simplest White-Box and Black-Box attacks by an attacker that knows the structure of the De-Pois defense model. In essence, the De-Pois defense relies on a critic model that can be used to detect poisoned data before passing it to the target model. In our work, we break this poison-protection layer by replicating the critic model and then performing a composed gradient-sign attack on both the critic and target models simultaneously -- allowing us to bypass the critic firewall to poison the target model."
                    }
                ]
            },
            "7d6eff07-fc2b-436d-9d67-7ac7fb814e73": {
                "pk": "7d6eff07-fc2b-436d-9d67-7ac7fb814e73",
                "name": "Frederik Schmitt",
                "collaborators": [
                    "Christopher Hahn",
                    "Bernd Finkbeiner",
                    "Matthias Cosler",
                    "Niklas Metzger",
                    "Julian Siber",
                    "Ayham Omar",
                    "Markus N. Rabe",
                    "Daniel Mendoza",
                    "Caroline Trippel",
                    "Julia J. Tillman"
                ],
                "domain": [
                    "Deep Learning",
                    "Formal Verification",
                    "Natural Language Processing",
                    "Hardware Synthesis"
                ],
                "publications": [
                    {
                        "title": "Iterative Circuit Repair Against Formal Specifications",
                        "abstract": "We present a deep learning approach for repairing sequential circuits against formal specifications given in linear-time temporal logic (LTL). Given a defective circuit and its formal specification, we train Transformer models to output circuits that satisfy the corresponding specification. We propose a separated hierarchical Transformer for multimodal representation learning of the formal specification and the circuit. We introduce a data generation algorithm that enables generalization to more complex specifications and out-of-distribution datasets. In addition, our proposed repair mechanism significantly improves the automated synthesis of circuits from LTL specifications with Transformers. It improves the state-of-the-art by $6.8$ percentage points on held-out instances and $11.8$ percentage points on an out-of-distribution dataset from the annual reactive synthesis competition."
                    },
                    {
                        "title": "NeuroSynt: A Neuro-symbolic Portfolio Solver for Reactive Synthesis",
                        "abstract": "We introduce NeuroSynt, a neuro-symbolic portfolio solver framework for reactive synthesis. At the core of the solver lies a seamless integration of neural and symbolic approaches to solving the reactive synthesis problem. To ensure soundness, the neural engine is coupled with model checkers verifying the predictions of the underlying neural models. The open-source implementation of NeuroSynt provides an integration framework for reactive synthesis in which new neural and state-of-the-art symbolic approaches can be seamlessly integrated. Extensive experiments demonstrate its efficacy in handling challenging specifications, enhancing the state-of-the-art reactive synthesis solvers, with NeuroSynt contributing novel solves in the current SYNTCOMP benchmarks."
                    },
                    {
                        "title": "Neural Circuit Synthesis from Specification Patterns",
                        "abstract": "We train hierarchical Transformers on the task of synthesizing hardware circuits directly out of high-level logical specifications in linear-time temporal logic (LTL). The LTL synthesis problem is a well-known algorithmic challenge with a long history and an annual competition is organized to track the improvement of algorithms and tooling over time. New approaches using machine learning might open a lot of possibilities in this area, but suffer from the lack of sufficient amounts of training data. In this paper, we consider a method to generate large amounts of additional training data, i.e., pairs of specifications and circuits implementing them. We ensure that this synthetic data is sufficiently close to human-written specifications by mining common patterns from the specifications used in the synthesis competitions. We show that hierarchical Transformers trained on this synthetic data solve a significant portion of problems from the synthesis competitions, and even out-of-distribution examples from a recent case study."
                    },
                    {
                        "title": "Attention Flows for General Transformers",
                        "abstract": "In this paper, we study the computation of how much an input token in a Transformer model influences its prediction. We formalize a method to construct a flow network out of the attention values of encoder-only Transformer models and extend it to general Transformer architectures including an auto-regressive decoder. We show that running a maxflow algorithm on the flow network construction yields Shapley values, which determine the impact of a player in cooperative game theory. By interpreting the input tokens in the flow network as players, we can compute their influence on the total attention flow leading to the decoder's decision. Additionally, we provide a library that computes and visualizes the attention flow of arbitrary Transformer models. We show the usefulness of our implementation on various models trained on natural language processing and reasoning tasks."
                    },
                    {
                        "title": "nl2spec: Interactively Translating Unstructured Natural Language to Temporal Logics with Large Language Models",
                        "abstract": "A rigorous formalization of desired system requirements is indispensable when performing any verification task. This often limits the application of verification techniques, as writing formal specifications is an error-prone and time-consuming manual task. To facilitate this, we present nl2spec, a framework for applying Large Language Models (LLMs) to derive formal specifications (in temporal logics) from unstructured natural language. In particular, we introduce a new methodology to detect and resolve the inherent ambiguity of system requirements in natural language: we utilize LLMs to map subformulas of the formalization back to the corresponding natural language fragments of the input. Users iteratively add, delete, and edit these sub-translations to amend erroneous formalizations, which is easier than manually redrafting the entire formalization. The framework is agnostic to specific application domains and can be extended to similar specification languages and new neural models. We perform a user study to obtain a challenging dataset, which we use to run experiments on the quality of translations. We provide an open-source implementation, including a web-based frontend."
                    },
                    {
                        "title": "Formal Specifications from Natural Language",
                        "abstract": "We study the generalization abilities of language models when translating natural language into formal specifications with complex semantics. In particular, we fine-tune language models on three datasets consisting of English sentences and their corresponding formal representation: 1) regular expressions (regex), frequently used in programming and search; 2) First-order logic (FOL), commonly used in software verification and theorem proving; and 3) linear-time temporal logic (LTL), which forms the basis for industrial hardware specification languages. Our experiments show that, in these diverse domains, the language models maintain their generalization capabilities from pre-trained knowledge of natural language to generalize, e.g., to new variable names or operator descriptions. Additionally, they achieve competitive performance, and even outperform the state-of-the-art for translating into regular expressions, with the benefits of being easy to access, efficient to fine-tune, and without a particular need for domain-specific reasoning."
                    }
                ]
            },
            "c2d88e27-cb83-493c-bb74-a55970f6da3f": {
                "pk": "c2d88e27-cb83-493c-bb74-a55970f6da3f",
                "name": "Julian Siber",
                "collaborators": [
                    "Bernd Finkbeiner",
                    "Niklas Metzger",
                    "Christopher Hahn",
                    "Hadar Frenkel",
                    "Norine Coenen",
                    "Frederik Schmitt",
                    "Christel Baier",
                    "Florian Funke",
                    "Simon Jantsch",
                    "Julia J. Tillman"
                ],
                "domain": [
                    "Causality",
                    "Formal Methods",
                    "Natural Language Processing",
                    "Game Theory"
                ],
                "publications": [
                    {
                        "title": "Counterfactuals Modulo Temporal Logics",
                        "abstract": "Lewis' theory of counterfactuals is the foundation of many contemporary notions of causality. In this paper, we extend this theory in the temporal direction to enable symbolic counterfactual reasoning on infinite sequences, such as counterexamples found by a model checker and trajectories produced by a reinforcement learning agent. In particular, our extension considers a more relaxed notion of similarity between worlds and proposes two additional counterfactual operators that close a semantic gap between the previous two in this more general setting. Further, we consider versions of counterfactuals that minimize the distance to the witnessing counterfactual worlds, a common requirement in causal analysis. To automate counterfactual reasoning in the temporal domain, we introduce a logic that combines temporal and counterfactual operators, and outline decision procedures for the satisfiability and trace-checking problems of this logic."
                    },
                    {
                        "title": "Synthesis of Temporal Causality",
                        "abstract": "We present an automata-based algorithm to synthesize omega-regular causes for omega-regular effects on executions of a reactive system, such as counterexamples uncovered by a model checker. Our theory is a generalization of temporal causality, which has recently been proposed as a framework for drawing causal relationships between trace properties on a given trace. So far, algorithms exist only for verifying a single causal relationship and, as an extension, cause synthesis through enumeration, which is complete only for a small fragment of effect properties. This work presents the first complete cause-synthesis algorithm for the class of omega-regular effects. We show that in this case, causes are guaranteed to be omega-regular themselves and can be computed as, e.g., nondeterministic B\\\"uchi automata. We demonstrate the practical feasibility of this algorithm with a prototype tool and evaluate its performance for cause synthesis and cause checking."
                    },
                    {
                        "title": "Causality-Based Game Solving",
                        "abstract": "We present a causality-based algorithm for solving two-player reachability games represented by logical constraints. These games are a useful formalism to model a wide array of problems arising, e.g., in program synthesis. Our technique for solving these games is based on the notion of subgoals, which are slices of the game that the reachability player necessarily needs to pass through in order to reach the goal. We use Craig interpolation to identify these necessary sets of moves and recursively slice the game along these subgoals. Our approach allows us to infer winning strategies that are structured along the subgoals. If the game is won by the reachability player, this is a strategy that progresses through the subgoals towards the final goal; if the game is won by the safety player, it is a permissive strategy that completely avoids a single subgoal. We evaluate our prototype implementation on a range of different games. On multiple benchmark families, our prototype scales dramatically better than previously available tools."
                    },
                    {
                        "title": "Attention Flows for General Transformers",
                        "abstract": "In this paper, we study the computation of how much an input token in a Transformer model influences its prediction. We formalize a method to construct a flow network out of the attention values of encoder-only Transformer models and extend it to general Transformer architectures including an auto-regressive decoder. We show that running a maxflow algorithm on the flow network construction yields Shapley values, which determine the impact of a player in cooperative game theory. By interpreting the input tokens in the flow network as players, we can compute their influence on the total attention flow leading to the decoder's decision. Additionally, we provide a library that computes and visualizes the attention flow of arbitrary Transformer models. We show the usefulness of our implementation on various models trained on natural language processing and reasoning tasks."
                    },
                    {
                        "title": "Formal Specifications from Natural Language",
                        "abstract": "We study the generalization abilities of language models when translating natural language into formal specifications with complex semantics. In particular, we fine-tune language models on three datasets consisting of English sentences and their corresponding formal representation: 1) regular expressions (regex), frequently used in programming and search; 2) First-order logic (FOL), commonly used in software verification and theorem proving; and 3) linear-time temporal logic (LTL), which forms the basis for industrial hardware specification languages. Our experiments show that, in these diverse domains, the language models maintain their generalization capabilities from pre-trained knowledge of natural language to generalize, e.g., to new variable names or operator descriptions. Additionally, they achieve competitive performance, and even outperform the state-of-the-art for translating into regular expressions, with the benefits of being easy to access, efficient to fine-tune, and without a particular need for domain-specific reasoning."
                    },
                    {
                        "title": "Explaining Hyperproperty Violations",
                        "abstract": "Hyperproperties relate multiple computation traces to each other. Model checkers for hyperproperties thus return, in case a system model violates the specification, a set of traces as a counterexample. Fixing the erroneous relations between traces in the system that led to the counterexample is a difficult manual effort that highly benefits from additional explanations. In this paper, we present an explanation method for counterexamples to hyperproperties described in the specification logic HyperLTL. We extend Halpern and Pearl's definition of actual causality to sets of traces witnessing the violation of a HyperLTL formula, which allows us to identify the events that caused the violation. We report on the implementation of our method and show that it significantly improves on previous approaches for analyzing counterexamples returned by HyperLTL model checkers."
                    }
                ]
            },
            "f0c108cb-5661-4be3-abd8-97d0c2488638": {
                "pk": "f0c108cb-5661-4be3-abd8-97d0c2488638",
                "name": "Bernd Finkbeiner",
                "collaborators": [
                    "Raven Beutner",
                    "Borzoo Bonakdarpour",
                    "Felix Klein",
                    "Hazem Torfah",
                    "Noemi Passing",
                    "Gilles Barthe",
                    "Pedro R. D'Argenio",
                    "Holger Hermanns",
                    "Armando Solar-Lezama",
                    "Christopher Hahn"
                ],
                "domain": [
                    "Temporal Logic",
                    "Model Checking",
                    "Hyperproperties",
                    "Reactive Synthesis"
                ],
                "publications": [
                    {
                        "title": "Model Checking Algorithms for Hyperproperties",
                        "abstract": "Hyperproperties generalize trace properties by expressing relations between multiple computations. Hyperpropertes include policies from information-flow security, like observational determinism or non-interference, and many other system properties including promptness and knowledge. In this paper, we give an overview on the model checking problem for temporal hyperlogics. Our starting point is the model checking algorithm for HyperLTL, a reduction to B\\\"uchi automata emptiness. This basic construction can be extended with propositional quantification, resulting in an algorithm for HyperQPTL. It can also be extended with branching time, resulting in an algorithm for HyperCTL*. However, it is not possible to have both extensions at the same time: the model checking problem of HyperQCTL* is undecidable. An attractive compromise is offered by MPL[E], i.e., monadic path logic extended with the equal-level predicate. The expressiveness of MPL[E] falls strictly between that of HyperCTL* and HyperQCTL*. MPL[E] subsumes both HyperCTL* and HyperKCTL*, the extension of HyperCTL* with the knowledge operator. We show that the model checking problem for MPL[E] is still decidable."
                    },
                    {
                        "title": "Program Repair for Hyperproperties",
                        "abstract": "We study the repair problem for hyperproperties specified in the temporal logic HyperLTL. Hyperproperties are system properties that relate multiple computation traces. This class of properties includes information flow policies like noninterference and observational determinism. The repair problem is to find, for a given Kripke structure, a substructure that satisfies a given specification. We show that the repair problem is decidable for HyperLTL specifications and finite-state Kripke structures. We provide a detailed complexity analysis for different fragments of HyperLTL and different system types: tree-shaped, acyclic, and general Kripke structures."
                    },
                    {
                        "title": "Controller Synthesis for Hyperproperties",
                        "abstract": "We investigate the problem of controller synthesis for hyperproperties specified in the temporal logic HyperLTL. Hyperproperties are system properties that relate multiple execution traces. Hyperproperties can elegantly express information-flow policies like noninterference and observational determinism. The controller synthesis problem is to automatically design a controller for a plant that ensures satisfaction of a given specification in the presence of the environment or adversarial actions. We show that the controller synthesis problem is decidable for HyperLTL specifications and finite-state plants. We provide a rigorous complexity analysis for different fragments of HyperLTL and different system types: tree-shaped, acyclic, and general graphs."
                    },
                    {
                        "title": "Facets of Software Doping",
                        "abstract": "This paper provides an informal discussion of the formal aspects of software doping."
                    },
                    {
                        "title": "Reactive Synthesis: Towards Output-Sensitive Algorithms",
                        "abstract": "Reactive synthesis is a technology for the automatic construction of reactive systems from logical specifications. In these lecture notes, we study different algorithms for the reactive synthesis problem of linear-time temporal logic (LTL). The classic game-based synthesis algorithm is input-sensitive in the sense that its performance is asymptotically optimal in the size of the specification, but it produces implementations that may be larger than necessary. We contrast this algorithm with output-sensitive algorithms for reactive synthesis, i.e., algorithms that are optimized towards the size or structural complexity of the synthesized system. We study the bounded synthesis algorithm, which produces an implementation with a minimal number of states, and the bounded cycle synthesis algorithm, which additionally guarantees that the number of cycles of the implementation is minimal."
                    },
                    {
                        "title": "Proceedings Second Workshop on Synthesis",
                        "abstract": "Software synthesis is rapidly developing into an important research area with vast potential for practical application. The SYNT Workshop on Synthesis aims to bringing together researchers interested in synthesis to present both ongoing and mature work on all aspects of automated synthesis and its applications.   The second iteration of SYNT took place in Saint Petersburg, Russia, and was co-located with the 25th International Conference on Computer Aided Verification. The workshop included eleven presentations covering the full scope of the emerging synthesis community, as well as a discussion lead by Swen Jacobs on the organization of two new synthesis competitions focusing on reactive synthesis and syntax-guided functional synthesis respectively."
                    },
                    {
                        "title": "Bounded Cycle Synthesis",
                        "abstract": "We introduce a new approach for the synthesis of Mealy machines from specifications in linear-time temporal logic (LTL), where the number of cycles in the state graph of the implementation is limited by a given bound. Bounding the number of cycles leads to implementations that are structurally simpler and easier to understand. We solve the synthesis problem via an extension of SAT-based bounded synthesis, where we additionally construct a witness structure that limits the number of cycles. We also establish a triple-exponential upper and lower bound for the potential blow-up between the length of the LTL formula and the number of cycles in the state graph."
                    },
                    {
                        "title": "Deciding Hyperproperties",
                        "abstract": "Hyperproperties, like observational determinism or symmetry, cannot be expressed as properties of individual computation traces, because they describe a relation between multiple computation traces. HyperLTL is a temporal logic that captures such relations through trace variables, which are introduced through existential and universal trace quantifiers and can be used to refer to multiple computations at the same time. In this paper, we study the satisfiability problem of HyperLTL. We show that the problem is PSPACE-complete for alternation-free formulas (and, hence, no more expensive than LTL satisfiability), EXPSPACE-complete for exists-forall-formulas, and undecidable for forall-exists-formulas. Many practical hyperproperties can be expressed as alternation-free formulas. Our results show that both satisfiability and implication are decidable for such properties."
                    },
                    {
                        "title": "The Complexity of Monitoring Hyperproperties",
                        "abstract": "We study the runtime verification of hyperproperties, expressed in the temporal logic HyperLTL, as a means to inspect a system with respect to security polices. Runtime monitors for hyperproperties analyze trace logs that are organized by common prefixes in the form of a tree-shaped Kripke structure, or are organized both by common prefixes and by common suffixes in the form of an acyclic Kripke structure. Unlike runtime verification techniques for trace properties, where the monitor tracks the state of the specification but usually does not need to store traces, a monitor for hyperproperties repeatedly model checks the growing Kripke structure. This calls for a rigorous complexity analysis of the model checking problem over tree-shaped and acyclic Kripke structures. We show that for trees, the complexity in the size of the Kripke structure is L-complete independently of the number of quantifier alternations in the HyperLTL formula. For acyclic Kripke structures, the complexity is PSPACE-complete (in the level of the polynomial hierarchy that corresponds to the number of quantifier alternations). The combined complexity in the size of the Kripke structure and the length of the HyperLTL formula is PSPACE-complete for both trees and acyclic Kripke structures, and is as low as NC for the relevant case of trees and alternation-free HyperLTL formulas. Thus, the size and shape of both the Kripke structure and the formula have significant impact on the complexity of the model checking problem."
                    },
                    {
                        "title": "Reactive Safety",
                        "abstract": "The distinction between safety and liveness properties is a fundamental classification with immediate implications on the feasibility and complexity of various monitoring, model checking, and synthesis problems. In this paper, we revisit the notion of safety for reactive systems, i.e., for systems whose behavior is characterized by the interplay of uncontrolled environment inputs and controlled system outputs. We show that reactive safety is a strictly larger class of properties than standard safety. We provide algorithms for checking if a property, given as a temporal formula or as a word or tree automaton, is a reactive safety property and for translating such properties into safety automata. Based on this construction, the standard verification and synthesis algorithms for safety properties immediately extend to the larger class of reactive safety."
                    },
                    {
                        "title": "The First-Order Logic of Hyperproperties",
                        "abstract": "We investigate the logical foundations of hyperproperties. Hyperproperties generalize trace properties, which are sets of traces, to sets of sets of traces. The most prominent application of hyperproperties is information flow security: information flow policies characterize the secrecy and integrity of a system by comparing two or more execution traces, for example by comparing the observations made by an external observer on execution traces that result from different values of a secret variable. In this paper, we establish the first connection between temporal logics for hyperproperties and first-order logic. Kamp's seminal theorem (in the formulation due to Gabbay et al.) states that linear-time temporal logic (LTL) is expressively equivalent to first-order logic over the natural numbers with order. We introduce first-order logic over sets of traces and prove that HyperLTL, the extension of LTL to hyperproperties, is strictly subsumed by this logic. We furthermore exhibit a fragment that is expressively equivalent to HyperLTL, thereby establishing Kamp's theorem for hyperproperties."
                    },
                    {
                        "title": "Symmetric Synthesis",
                        "abstract": "We study the problem of determining whether a given temporal specification can be implemented by a symmetric system, i.e., a system composed from identical components. Symmetry is an important goal in the design of distributed systems, because systems that are composed from identical components are easier to build and maintain. We show that for the class of rotation-symmetric architectures, i.e., multi-process architectures where all processes have access to all system inputs, but see different rotations of the inputs, the symmetric synthesis problem is EXPTIME-complete in the number of processes. In architectures where the processes do not have access to all input variables, the symmetric synthesis problem becomes undecidable, even in cases where the standard distributed synthesis problem is decidable."
                    },
                    {
                        "title": "The Density of Linear-time Properties",
                        "abstract": "Finding models for linear-time properties is a central problem in verification and planning. We study the distribution of linear-time models by investigating the density of linear-time properties over the space of ultimately periodic words. The density of a property over a bound n is the ratio of the number of lasso-shaped words of length n that satisfy the property to the total number of lasso-shaped words of length n. We investigate the problem of computing the density for both linear-time properties in general and for the important special case of omega-regular properties. For general linear-time properties, the density is not necessarily convergent and can oscillate indefinitely for certain properties. However, we show the oscillation can be bounded by the growth of the sets of bad- and good-prefix of the property. For omega-regular properties, we show that the density is always convergent and provide a general algorithm for computing the density of omega-regular properties as well as more specialized algorithms for certain sub-classes and their combinations."
                    },
                    {
                        "title": "Synthesizing Skeletons for Reactive Systems",
                        "abstract": "We present an analysis technique for temporal specifications of reactive systems that identifies, on the level of individual system outputs over time, which parts of the implementation are determined by the specification, and which parts are still open. This information is represented in the form of a labeled transition system, which we call skeleton. Each state of the skeleton is labeled with a three-valued assignment to the output variables: each output can be true, false, or open, where true or false means that the value must be true or false, respectively, and open means that either value is still possible. We present algorithms for the verification of skeletons and for the learning-based synthesis of skeletons from specifications in linear-time temporal logic (LTL). The algorithm returns a skeleton that satisfies the given LTL specification in time polynomial in the size of the minimal skeleton. Our new analysis technique can be used to recognize and repair specifications that underspecify critical situations. The technique thus complements existing methods for the recognition and repair of overspecifications via the identification of unrealizable cores."
                    },
                    {
                        "title": "A Temporal Logic for Strategic Hyperproperties",
                        "abstract": "Hyperproperties are commonly used in computer security to define information-flow policies and other requirements that reason about the relationship between multiple computations. In this paper, we study a novel class of hyperproperties where the individual computation paths are chosen by the strategic choices of a coalition of agents in a multi-agent system. We introduce HyperATL*, an extension of computation tree logic with path variables and strategy quantifiers. HyperATL* can express strategic hyperproperties, such as that the scheduler in a concurrent system has a strategy to avoid information leakage. HyperATL* is particularly useful to specify asynchronous hyperproperties, i.e., hyperproperties where the speed of the execution on the different computation paths depends on the choices of the scheduler. Unlike other recent logics for the specification of asynchronous hyperproperties, our logic is the first to admit decidable model checking for the full logic. We present a model checking algorithm for HyperATL* based on alternating word automata and show that our algorithm is asymptotically optimal by providing a matching lower bound. We have implemented a prototype model checker for a fragment of HyperATL*, able to check various security properties on small programs."
                    },
                    {
                        "title": "Dependency-based Compositional Synthesis (Full Version)",
                        "abstract": "Despite many recent advances, reactive synthesis is still not really a practical technique. The grand challenge is to scale from small transition systems, where synthesis performs well, to complex multi-component designs. Compositional methods, such as the construction of dominant strategies for individual components, reduce the complexity significantly, but are usually not applicable without extensively rewriting the specification. In this paper, we present a refinement of compositional synthesis that does not require such an intervention. Our algorithm decomposes the system into a sequence of components, such that every component has a strategy that is dominant, i.e., performs at least as good as any possible alternative, provided that the preceding components follow their (already synthesized) strategies. The decomposition of the system is based on a dependency analysis, for which we provide semantic and syntactic techniques. We establish the soundness and completeness of the approach and report on encouraging experimental results."
                    },
                    {
                        "title": "Compositional Synthesis of Modular Systems (Full Version)",
                        "abstract": "Given the advances in reactive synthesis, it is a natural next step to consider more complex multi-process systems. Distributed synthesis, however, is not yet scalable. Compositional approaches can be a game changer. Here, the challenge is to decompose a given specification of the global system behavior into requirements on the individual processes. In this paper, we introduce a compositional synthesis algorithm that, for each process, constructs, in addition to the implementation, a certificate that captures the necessary interface between the processes. The certificates then allow for constructing separate requirements for the individual processes. By bounding the size of the certificates, we can bias the synthesis procedure towards solutions that are desirable in the sense that the assumptions between the processes are small. Our experimental results show that our approach is much faster than standard methods for distributed synthesis as long as reasonably small certificates exist."
                    },
                    {
                        "title": "Prophecy Variables for Hyperproperty Verification",
                        "abstract": "Temporal logics for hyperproperties like HyperLTL use trace quantifiers to express properties that relate multiple system runs. In practice, the verification of such specifications is mostly limited to formulas without quantifier alternation, where verification can be reduced to checking a trace property over the self-composition of the system. Quantifier alternations like $\\forall \\pi. \\exists \\pi'. \\phi$, can either be solved by complementation or with an interpretation as a two-person game between a $\\forall$-player, who incrementally constructs the trace $\\pi$, and an $\\exists$-player, who constructs $\\pi'$ in such a way that $\\pi$ and $\\pi'$ together satisfy $\\phi$. The game-based approach is significantly cheaper but incomplete because the $\\exists$-player does not know the future moves of the $\\forall$-player. In this paper, we establish that the game-based approach can be made complete by adding ($\\omega$-regular) temporal prophecies. Our proof is constructive, yielding an effective algorithm for the generation of a complete set of prophecies."
                    },
                    {
                        "title": "Software Verification of Hyperproperties Beyond k-Safety",
                        "abstract": "Temporal hyperproperties are system properties that relate multiple execution traces. For (finite-state) hardware, temporal hyperproperties are supported by model checking algorithms, and tools for general temporal logics like HyperLTL exist. For (infinite-state) software, the analysis of temporal hyperproperties has, so far, been limited to $k$-safety properties, i.e., properties that stipulate the absence of a bad interaction between any $k$ traces. In this paper, we present an automated method for the verification of $\\forall^k\\exists^l$-safety properties in infinite-state systems. A $\\forall^k\\exists^l$-safety property stipulates that for any $k$ traces, there exist $l$ traces such that the resulting $k+l$ traces do not interact badly. This combination of universal and existential quantification enables us to express many properties beyond $k$-safety, including, for example, generalized non-interference or program refinement. Our method is based on a strategy-based instantiation of existential trace quantification combined with a program reduction, both in the context of a fixed predicate abstraction. Notably, our framework allows for mutual dependence of strategy and reduction."
                    },
                    {
                        "title": "AutoHyper: Explicit-State Model Checking for HyperLTL",
                        "abstract": "HyperLTL is a temporal logic that can express hyperproperties, i.e., properties that relate multiple execution traces of a system. Such properties are becoming increasingly important and naturally occur, e.g., in information-flow control, robustness, mutation testing, path planning, and causality checking. Thus far, complete model checking tools for HyperLTL have been limited to alternation-free formulas, i.e., formulas that use only universal or only existential trace quantification. Properties involving quantifier alternations could only be handled in an incomplete way, i.e., the verification might fail even though the property holds. In this paper, we present AutoHyper, an explicit-state automata-based model checker that supports full HyperLTL and is complete for properties with arbitrary quantifier alternations. We show that language inclusion checks can be integrated into HyperLTL verification, which allows AutoHyper to benefit from a range of existing inclusion-checking tools. We evaluate AutoHyper on a broad set of benchmarks drawn from different areas in the literature and compare it with existing (incomplete) methods for HyperLTL verification."
                    }
                ]
            }
        }
    },
    "2403.02187": {
        "paper_data": {
            "title": "Mutual Information Estimation via Normalizing Flows",
            "url": "http://arxiv.org/abs/2403.02187v3",
            "arxiv_id": "2403.02187",
            "authors": [
                "Ivan Butakov",
                "Alexander Tolmachev",
                "Sofia Malanchuk",
                "Anna Neopryatnaya",
                "Alexey Frolov"
            ],
            "abstract": "We propose a novel approach to the problem of mutual information (MI) estimation via introducing a family of estimators based on normalizing flows. The estimator maps original data to the target distribution, for which MI is easier to estimate. We additionally explore the target distributions with known closed-form expressions for MI. Theoretical guarantees are provided to demonstrate that our approach yields MI estimates for the original data. Experiments with high-dimensional data are conducted to highlight the practical advantages of the proposed method.",
            "introduction": " Introduction Information-theoretic analysis of deep neural networks (DNN) has attracted recent interest due to intriguing fundamental results in I(\u03be;\u03b7) =1 2[log det \u03a3 \u03be,\u03be+ log det \u03a3 \u03b7,\u03b7\u2212log det \u03a3] + + D KL(p\u03be,\u03b7||N(m,\u03a3))\u2212DKL(p\u03be\u2297p\u03b7||N(m,diag(\u03a3 \u03be,\u03be,\u03a3\u03b7,\u03b7)) Proof of Corollary 4.4. The same as for Corollary 3.2 Proof of Statement 4.8. Due to the change of variables formula, LSN\u25e6(fX\u00d7fY)({(xk, yk)}) = log\f\f\f\fdet\u2202f(x, y) \u2202(x, y)\f\f\f\f+LN({f(xk, yk)}) Asf=fX\u00d7fY, the Jacobian matrix\u2202f(x,y) \u2202(x,y)is block-diagonal, so log\f\f\f\fdet\u2202f(x, y) \u2202(x, y)\f\f\f\f= log\f\f\f\fdet\u2202fX(x) \u2202x\f\f\f\f+ log\f\f\f\fdet\u2202fY(y) \u2202y\f\f\f\f Finally, we use the maximum-likelihood estimates for mand\u03a3to drop the supremum in LN({f(xk, yk)}): \u02c6m=1 NNX k=1f(xk, yk),\u02c6\u03a3 =1 NNX k=1(f(xk, yk)\u2212\u02c6m)(f(xk, yk)\u2212\u02c6m)T=\u21d2 =\u21d2 L N({f(xk, yk)}) =LN( \u02c6m,\u02c6\u03a3)({f(xk, yk)}) Proof of Lemma 4.9. The variance of \u02c6IN-MIENF is upper bounded by the sum of the variances of the log-det terms. If (\u03be, \u03b7)is Gaussian, the log-det terms are asymptotically normal with the asymptotic variance being 2d/N (Corollary 1 in [ 51]). If(\u03be, \u03b7)is not Gaussian, the central limit theorem can be applied to each element of the covariance matrix, which in combination with the delta method yields the final result. 16Proof of Remark 4.10. LSN\u25e6(Az+b)({zk}) = log |detA|+LSN({Azk+b}) = log |detA|+LN( \u02c6m+b,A\u02c6\u03a3AT)({Azk+b}) = = log|detA|+ log|detA\u22121|+LN( \u02c6m,\u02c6\u03a3)({zk}) =LSN({zk}) Proof of Corollary 4.12. Under the proposed assumptions, log det \u03a3 \u03be,\u03be= log det \u03a3 \u03b7,\u03b7= 0, so I(\u03be;\u03b7) =\u22121 2log det \u03a3 . The matrix \u03a3is block-diagonalizable via the following permutation: (\u03be1, . . . , \u03be d\u03be, \u03b71, . . . , \u03b7 d\u03b7)7\u2192(\u03be1, \u03b71, \u03be2, \u03b72, . . .), with the blocks being \u03a3\u03bei,\u03b7i=\u0014 1\u03c1j \u03c1j1\u0015 The determinant of block-diagonal matrix is a product of the block determinants, so I(\u03be;\u03b7) = \u22121 2P klog(1\u2212\u03c12 j). Proof of Statement 4.13. Let us consider centered \u03beand\u03b7, as shifting is an invertible affine mapping. Note that \u03a3\u03be,\u03beand\u03a3\u03b7,\u03b7are positive definite and symmetric, so the following symmetric matrix square roots are defined: A= \u03a3\u22121/2 \u03be,\u03be,B= \u03a3\u22121/2 \u03b7,\u03b7 . By applying these invertible linear transformation to\u03beand\u03b7we get cov(A\u03be, B\u03b7 ) =\" \u03a3\u22121/2 \u03be,\u03be\u03a3\u03be,\u03be(\u03a3\u22121/2 \u03be,\u03be)T\u03a3\u22121/2 \u03be,\u03be\u03a3\u03be,\u03b7(\u03a3\u22121/2 \u03b7,\u03b7)T \u03a3\u22121/2 \u03b7,\u03b7\u03a3\u03b7,\u03be(\u03a3\u22121/2 \u03be,\u03be)T\u03a3\u22121/2 \u03b7,\u03b7\u03a3\u03b7,\u03b7(\u03a3\u22121/2 \u03b7,\u03b7)T# =\u0014I C CTI\u0015 , where C= \u03a3\u22121/2 \u03be,\u03be\u03a3\u03be,\u03b7(\u03a3\u22121/2 \u03b7,\u03b7)T. Then, the singular value decomposition is performed: C=URVT, where UandVare orthogonal, R= diag( {\u03c1j}). Finally, we apply UTtoA\u03beandVTtoB\u03b7: cov(UTA\u03be, VTB\u03b7) =\u0014 UTU UTCV (UTCV)TVTV\u0015 =\u0014I R RTI\u0015 , Note that UTAandVTBare invertible. Proof of Statement 4.14. Note that \u03a3is positive definite and symmetric, so the following symmetric matrix square root is defined: \u03a3\u22121/2. This matrix is a normalization matrix: cov(\u03a3\u22121/2(\u03be, \u03b7)) = \u03a3\u22121/2\u03a3 (\u03a3\u22121/2)T=I. According to the change of variable formula, LN(0,\u03a3)({zk}) = log det \u03a3\u22121/2+LN(0,I)({\u03a3\u22121/2zk}) As\u03a3\u03be,\u03be=Id\u03beand\u03a3\u03b7,\u03b7=Id\u03b7, from the equation (8) we derive I(\u03be;\u03b7) =\u22121 2log det \u03a3 = log det \u03a3\u22121/2 Finally, it is sufficient to validate the proposed closed-form expression for \u03a31/2in the case of 2\u00d72 matrices, as \u03a3is block-diagonalizable (with 2\u00d72blocks) via the following permutation: M: (\u03be1, . . . , \u03be d\u03be, \u03b71, . . . , \u03b7 d\u03b7)7\u2192(\u03be1, \u03b71, \u03be2, \u03b72, . . .), Note that\u0014 \u03b1+\u03b2 \u03b1\u2212\u03b2 \u03b1\u2212\u03b2 \u03b1 +\u03b2\u00152 = 2\u00b7\u0014 \u03b12+\u03b22\u03b12\u2212\u03b22 \u03b12\u2212\u03b22\u03b12+\u03b22\u0015 =1 1\u2212\u03c12\u0014 1\u2212\u03c1 \u2212\u03c11\u0015 1 1\u2212\u03c12\u0014 1\u2212\u03c1 \u2212\u03c11\u0015 \u00b7\u0014 1\u03c1 \u03c11\u0015 =\u0014 1 0 0 1\u0015 17B Non-Gaussian-based tests As our estimator is based on Gaussianization, it seems natural that we observe good performance in the methods. IEEE Transactions on Pattern Analysis & Machine Intelligence , 43(11): 3964\u20133979, nov 2021. ISSN 1939-3539. doi: 10.1109/TPAMI.2020.2992934. [35] Thomas M. Cover and Joy A. Thomas. Elements of Information Theory (Wiley Series in Telecommunications and Signal Processing) . Wiley-Interscience, USA, 2006. [36] H. O. Lancaster. The structure of bivariate distributions. The Annals of Mathematical Statistics , 29(3):719\u2013736, 1958. ISSN 00034851. URL http://www.jstor.org/stable/2237259 . [37] E. J. Hannan. The general theory of canonical correlation and its relation to functional analysis. Journal of the Australian Mathematical Society , 2(2):229\u2013242, 1961. doi: 10.1017/S1446788700026707. [38] M. D. Donsker and S. R.S. Varadhan. Asymptotic",
            "references": [
                {
                    "title": "Information-Theoretic Generalization Bounds for Deep Neural Networks",
                    "abstract": "Deep neural networks (DNNs) exhibit an exceptional capacity for generalization in practical applications. This work aims to capture the effect and benefits of depth for supervised learning via information-theoretic generalization bounds. We first derive two hierarchical bounds on the generalization error in terms of the Kullback-Leibler (KL) divergence or the 1-Wasserstein distance between the train and test distributions of the network internal representations. The KL divergence bound shrinks as the layer index increases, while the Wasserstein bound implies the existence of a layer that serves as a generalization funnel, which attains a minimal 1-Wasserstein distance. Analytic expressions for both bounds are derived under the setting of binary Gaussian classification with linear DNNs. To quantify the contraction of the relevant information measures when moving deeper into the network, we analyze the strong data processing inequality (SDPI) coefficient between consecutive layers of three regularized DNN models: Dropout, DropConnect, and Gaussian noise injection. This enables refining our generalization bounds to capture the contraction as a function of the network architecture parameters. Specializing our results to DNNs with a finite parameter space and the Gibbs algorithm reveals that deeper yet narrower network architectures generalize better in those examples, although how broadly this statement applies remains a question."
                },
                {
                    "title": "On the Properties and Estimation of Pointwise Mutual Information Profiles",
                    "abstract": "The pointwise mutual information profile, or simply profile, is the distribution of pointwise mutual information for a given pair of random variables. One of its important properties is that its expected value is precisely the mutual information between these random variables. In this paper, we analytically describe the profiles of multivariate normal distributions and introduce a novel family of distributions, Bend and Mix Models, for which the profile can be accurately estimated using Monte Carlo methods. We then show how Bend and Mix Models can be used to study the limitations of existing mutual information estimators, investigate the behavior of neural critics used in variational estimators, and understand the effect of experimental outliers on mutual information estimation. Finally, we show how Bend and Mix Models can be used to obtain model-based Bayesian estimates of mutual information, suitable for problems with available domain expertise in which uncertainty quantification is necessary."
                },
                {
                    "title": "MINDE: Mutual Information Neural Diffusion Estimation",
                    "abstract": "In this work we present a new method for the estimation of Mutual Information (MI) between random variables. Our approach is based on an original interpretation of the Girsanov theorem, which allows us to use score-based diffusion models to estimate the Kullback Leibler divergence between two densities as a difference between their score functions. As a by-product, our method also enables the estimation of the entropy of random variables. Armed with such building blocks, we present a general recipe to measure MI, which unfolds in two directions: one uses conditional diffusion process, whereas the other uses joint diffusion processes that allow simultaneous modelling of two random variables. Our results, which derive from a thorough experimental protocol over all the variants of our approach, indicate that our method is more accurate than the main alternatives from the literature, especially for challenging distributions. Furthermore, our methods pass MI self-consistency tests, including data processing and additivity under independence, which instead are a pain-point of existing methods."
                },
                {
                    "title": "Beyond Normal: On the Evaluation of Mutual Information Estimators",
                    "abstract": "Mutual information is a general statistical dependency measure which has found applications in representation learning, causality, domain generalization and computational biology. However, mutual information estimators are typically evaluated on simple families of probability distributions, namely multivariate normal distribution and selected distributions with one-dimensional random variables. In this paper, we show how to construct a diverse family of distributions with known ground-truth mutual information and propose a language-independent benchmarking platform for mutual information estimators. We discuss the general applicability and limitations of classical and neural estimators in settings involving high dimensions, sparse interactions, long-tailed distributions, and high mutual information. Finally, we provide guidelines for practitioners on how to select appropriate estimator adapted to the difficulty of problem considered and issues one needs to consider when applying an estimator to a new data set."
                },
                {
                    "title": "Information Bottleneck Analysis of Deep Neural Networks via Lossy Compression",
                    "abstract": "The Information Bottleneck (IB) principle offers an information-theoretic framework for analyzing the training process of deep neural networks (DNNs). Its essence lies in tracking the dynamics of two mutual information (MI) values: between the hidden layer output and the DNN input/target. According to the hypothesis put forth by Shwartz-Ziv&Tishby (2017), the training process consists of two distinct phases: fitting and compression. The latter phase is believed to account for the good generalization performance exhibited by DNNs. Due to the challenging nature of estimating MI between high-dimensional random vectors, this hypothesis was only partially verified for NNs of tiny sizes or specific types, such as quantized NNs. In this paper, we introduce a framework for conducting IB analysis of general NNs. Our approach leverages the stochastic NN method proposed by Goldfeld et al. (2019) and incorporates a compression step to overcome the obstacles associated with high dimensionality. In other words, we estimate the MI between the compressed representations of high-dimensional random vectors. The proposed method is supported by both theoretical and practical justifications. Notably, we demonstrate the accuracy of our estimator through synthetic experiments featuring predefined MI values and comparison with MINE (Belghazi et al., 2018). Finally, we perform IB analysis on a close-to-real-scale convolutional DNN, which reveals new features of the MI dynamics."
                },
                {
                    "title": "High-Dimensional Smoothed Entropy Estimation via Dimensionality Reduction",
                    "abstract": "We study the problem of overcoming exponential sample complexity in differential entropy estimation under Gaussian convolutions. Specifically, we consider the estimation of the differential entropy h(X + Z) via n independently and identically distributed samples of X, where X and Z are independent D-dimensional random variables with X sub-Gaussian with bounded second moment and $Z\\sim {\\mathcal{N}}\\left({0,{\\sigma ^2}{I_D}}\\right)$. Under the absolute-error loss, the above problem has a parametric estimation rate of $\\frac{{{c^D}}}{{\\sqrt n }}$, which is exponential in data dimension D and often problematic for applications. We overcome this exponential sample complexity by projecting X to a low-dimensional space via principal component analysis (PCA) before the entropy estimation, and show that the asymptotic error overhead vanishes as the unexplained variance of the PCA vanishes. This implies near-optimal performance for inherently low-dimensional structures embedded in high-dimensional spaces, including hidden-layer outputs of deep neural networks (DNN), which can be used to estimate mutual information (MI) in DNNs. We provide numerical results verifying the performance of our PCA approach on Gaussian and spiral data. We also apply our method to analysis of information flow through neural network layers (c.f. information bottleneck), with results measuring mutual information in a noisy fully connected network and a noisy convolutional neural network (CNN) for MNIST classification. 1 2"
                },
                {
                    "title": "normflows: A PyTorch Package for Normalizing Flows",
                    "abstract": "Normalizing flows model probability distributions through an expressive tractable density. They transform a simple base distribution, such as a Gaussian, through a sequence of invertible functions, which are referred to as layers. These layers typically use neural networks to become very expressive. Flows are ubiquitous in machine learning and have been applied to image generation, text modeling, variational inference, approximating Boltzmann distributions, and many other problems. Here, we present normflows, a Python package for normalizing flows. It allows to build normalizing flow models from a suite of base distributions, flow layers, and neural networks. The package is implemented in the popular deep learning framework PyTorch, which simplifies the integration of flows in larger machine learning models or pipelines. It supports most of the common normalizing flow architectures, such as Real NVP, Glow, Masked Autoregressive Flows, Neural Spline Flows, Residual Flows, and many more. The package can be easily installed via pip and the code is publicly available on GitHub."
                },
                {
                    "title": "Diffeomorphic Information Neural Estimation",
                    "abstract": "Mutual Information (MI) and Conditional Mutual Information (CMI) are multi-purpose tools from information theory that are able to naturally measure the statistical dependencies between random variables, thus they are usually of central interest in several statistical and machine learning tasks, such as conditional independence testing and representation learning. However, estimating CMI, or even MI, is infamously challenging due the intractable formulation. In this study, we introduce DINE (Diffeomorphic Information Neural Estimator)\u2013a novel approach for estimating CMI of continuous random variables, inspired by the invariance of CMI over diffeomorphic maps. We show that the variables of interest can be replaced with appropriate surrogates that follow simpler distributions, allowing the CMI to be efficiently evaluated via analytical solutions. Additionally, we demonstrate the quality of the proposed estimator in comparison with state-of-the-arts in three important tasks, including estimating MI, CMI, as well as its application in conditional independence testing. The empirical evaluations show that DINE consistently outperforms competitors in all tasks and is able to adapt very well to complex and high-dimensional relationships."
                },
                {
                    "title": "Entropy Estimation via Normalizing Flow",
                    "abstract": "Entropy estimation is an important problem in information theory and statistical science. Many popular entropy estimators suffer from fast growing estimation bias with respect to dimensionality, rendering them unsuitable for high dimensional problems. In this work we propose a transformbased method for high dimensional entropy estimation, which consists of the following two main ingredients. First by modifying the k-NN based entropy estimator, we propose a new estimator which enjoys small estimation bias for samples that are close to a uniform distribution. Second we design a normalizing flow based mapping that pushes samples toward a uniform distribution, and the relation between the entropy of the original samples and the transformed ones is also derived. As a result the entropy of a given set of samples is estimated by first transforming them toward a uniform distribution and then applying the proposed estimator to the transformed samples. Numerical experiments demonstrate the effectiveness of the method for high dimensional entropy estimation problems."
                },
                {
                    "title": "Diffusion Normalizing Flow",
                    "abstract": "We present a novel generative modeling method called diffusion normalizing flow based on stochastic differential equations (SDEs). The algorithm consists of two neural SDEs: a forward SDE that gradually adds noise to the data to transform the data into Gaussian random noise, and a backward SDE that gradually removes the noise to sample from the data distribution. By jointly training the two neural SDEs to minimize a common cost function that quantifies the difference between the two, the backward SDE converges to a diffusion process the starts with a Gaussian distribution and ends with the desired data distribution. Our method is closely related to normalizing flow and diffusion probabilistic models and can be viewed as a combination of the two. Compared with normalizing flow, diffusion normalizing flow is able to learn distributions with sharp boundaries. Compared with diffusion probabilistic models, diffusion normalizing flow requires fewer discretization steps and thus has better sampling efficiency. Our algorithm demonstrates competitive performance in both high-dimension data density estimation and image generation tasks."
                },
                {
                    "title": "Telescoping Density-Ratio Estimation",
                    "abstract": "Density-ratio estimation via classification is a cornerstone of unsupervised learning. It has provided the foundation for state-of-the-art methods in representation learning and generative modelling, with the number of use-cases continuing to proliferate. However, it suffers from a critical limitation: it fails to accurately estimate ratios p/q for which the two densities differ significantly. Empirically, we find this occurs whenever the KL divergence between p and q exceeds tens of nats. To resolve this limitation, we introduce a new framework, telescoping density-ratio estimation (TRE), that enables the estimation of ratios between highly dissimilar densities in high-dimensional spaces. Our experiments demonstrate that TRE can yield substantial improvements over existing single-ratio methods for mutual information estimation, representation learning and energy-based modelling."
                },
                {
                    "title": "Normalizing Flows: An Introduction and Review of Current Methods",
                    "abstract": "Normalizing Flows are generative models which produce tractable distributions where both sampling and density evaluation can be efficient and exact. The goal of this survey article is to give a coherent and comprehensive review of the literature around the construction and use of Normalizing Flows for distribution learning. We aim to provide context and explanation of the models, review current state-of-the-art literature, and identify open questions and promising future directions."
                },
                {
                    "title": "Flows for simultaneous manifold learning and density estimation",
                    "abstract": "We introduce manifold-learning flows (M-flows), a new class of generative models that simultaneously learn the data manifold as well as a tractable probability density on that manifold. Combining aspects of normalizing flows, GANs, autoencoders, and energy-based models, they have the potential to represent datasets with a manifold structure more faithfully and provide handles on dimensionality reduction, denoising, and out-of-distribution detection. We argue why such models should not be trained by maximum likelihood alone and present a new training algorithm that separates manifold and density updates. In a range of experiments we demonstrate how M-flows learn the data manifold and allow for better inference than standard flows in the ambient data space."
                },
                {
                    "title": "On Information Plane Analyses of Neural Network Classifiers\u2014A Review",
                    "abstract": "We review the current literature concerned with information plane (IP) analyses of neural network (NN) classifiers. While the underlying information bottleneck theory and the claim that information-theoretic compression is causally linked to generalization are plausible, empirical evidence was found to be both supporting and conflicting. We review this evidence together with a detailed analysis of how the respective information quantities were estimated. Our survey suggests that compression visualized in IPs is not necessarily information-theoretic but is rather often compatible with geometric compression of the latent representations. This insight gives the IP a renewed justification. Aside from this, we shed light on the problem of estimating mutual information in deterministic NNs and its consequences. Specifically, we argue that, even in feedforward NNs, the data processing inequality needs not to hold for estimates of mutual information. Similarly, while a fitting phase, in which the mutual information is between the latent representation and the target increases, is necessary (but not sufficient) for good classification performance, depending on the specifics of mutual information estimation, such a fitting phase needs to not be visible in the IP."
                },
                {
                    "title": "Reasoning About Generalization via Conditional Mutual Information",
                    "abstract": "We provide an information-theoretic framework for studying the generalization properties of machine learning algorithms. Our framework ties together existing approaches, including uniform convergence bounds and recent methods for adaptive data analysis. Specifically, we use Conditional Mutual Information (CMI) to quantify how well the input (i.e., the training data) can be recognized given the output (i.e., the trained model) of the learning algorithm. We show that bounds on CMI can be obtained from VC dimension, compression schemes, differential privacy, and other methods. We then show that bounded CMI implies various forms of generalization."
                },
                {
                    "title": "Training Normalizing Flows with the Information Bottleneck for Competitive Generative Classification",
                    "abstract": "The Information Bottleneck (IB) objective uses information theory to formulate a task-performance versus robustness trade-off. It has been successfully applied in the standard discriminative classification setting. We pose the question whether the IB can also be used to train generative likelihood models such as normalizing flows. Since normalizing flows use invertible network architectures (INNs), they are information-preserving by construction. This seems contradictory to the idea of a bottleneck. In this work, firstly, we develop the theory and methodology of IB-INNs, a class of conditional normalizing flows where INNs are trained using the IB objective: Introducing a small amount of {\\em controlled} information loss allows for an asymptotically exact formulation of the IB, while keeping the INN's generative capabilities intact. Secondly, we investigate the properties of these models experimentally, specifically used as generative classifiers. This model class offers advantages such as improved uncertainty quantification and out-of-distribution detection, but traditional generative classifier solutions suffer considerably in classification accuracy. We find the trade-off parameter in the IB controls a mix of generative capabilities and accuracy close to standard classifiers. Empirically, our uncertainty estimates in this mixed regime compare favourably to conventional generative and discriminative classifiers."
                },
                {
                    "title": "Understanding the Limitations of Variational Mutual Information Estimators",
                    "abstract": "Variational approaches based on neural networks are showing promise for estimating mutual information (MI) between high dimensional variables. However, they can be difficult to use in practice due to poorly understood bias/variance tradeoffs. We theoretically show that, under some conditions, estimators such as MINE exhibit variance that could grow exponentially with the true amount of underlying MI. We also empirically demonstrate that existing estimators fail to satisfy basic self-consistency properties of MI, such as data processing and additivity under independence. Based on a unified perspective of variational approaches, we develop a new estimator that focuses on variance reduction. Empirical results on standard benchmark tasks demonstrate that our proposed estimator exhibits improved bias-variance trade-offs on standard benchmark tasks."
                },
                {
                    "title": "A New Lower Bound for Kullback-Leibler Divergence Based on Hammersley-Chapman-Robbins Bound",
                    "abstract": "In this paper, we derive a useful lower bound for the Kullback-Leibler divergence (KL-divergence) based on the Hammersley-Chapman-Robbins bound (HCRB).The HCRB states that the variance of an estimator is bounded from below by the Chi-square divergence and the expectation value of the estimator. By using the relation between the KL-divergence and the Chi-square divergence, we show that the lower bound for the KL-divergence which only depends on the expectation value and the variance of a function we choose. This lower bound can also be derived from an information geometric approach. Furthermore, we show that the equality holds for the Bernoulli distributions and show that the inequality converges to the Cram\\'{e}r-Rao bound when two distributions are very close. We also describe application examples and examples of numerical calculation."
                },
                {
                    "title": "Convergence of Smoothed Empirical Measures With Applications to Entropy Estimation",
                    "abstract": "This paper studies convergence of empirical measures smoothed by a Gaussian kernel. Specifically, consider approximating <inline-formula> <tex-math notation=\"LaTeX\">$P\\ast \\mathcal {N}_\\sigma $ </tex-math></inline-formula>, for <inline-formula> <tex-math notation=\"LaTeX\">$\\mathcal {N}_\\sigma \\triangleq \\mathcal {N}(0,\\sigma ^{2} \\mathrm {I}_{d})$ </tex-math></inline-formula>, by <inline-formula> <tex-math notation=\"LaTeX\">$\\hat {P}_{n}\\ast \\mathcal {N}_\\sigma $ </tex-math></inline-formula> under different statistical distances, where <inline-formula> <tex-math notation=\"LaTeX\">$\\hat {P}_{n}$ </tex-math></inline-formula> is the empirical measure. We examine the convergence in terms of the Wasserstein distance, total variation (TV), Kullback-Leibler (KL) divergence, and <inline-formula> <tex-math notation=\"LaTeX\">$\\chi ^{2}$ </tex-math></inline-formula>-divergence. We show that the approximation error under the TV distance and 1-Wasserstein distance (<inline-formula> <tex-math notation=\"LaTeX\">$\\mathsf {W}_{1}$ </tex-math></inline-formula>) converges at the rate <inline-formula> <tex-math notation=\"LaTeX\">$e^{O(d)} n^{-1/2}$ </tex-math></inline-formula> in remarkable contrast to a (typical) <inline-formula> <tex-math notation=\"LaTeX\">$n^{-\\frac {1}{d}}$ </tex-math></inline-formula> rate for unsmoothed <inline-formula> <tex-math notation=\"LaTeX\">$\\mathsf {W}_{1}$ </tex-math></inline-formula> (and <inline-formula> <tex-math notation=\"LaTeX\">$d\\ge 3$ </tex-math></inline-formula>). Similarly, for the KL divergence, squared 2-Wasserstein distance (<inline-formula> <tex-math notation=\"LaTeX\">$\\mathsf {W}_{2}^{2}$ </tex-math></inline-formula>), and <inline-formula> <tex-math notation=\"LaTeX\">$\\chi ^{2}$ </tex-math></inline-formula>-divergence, the convergence rate is <inline-formula> <tex-math notation=\"LaTeX\">$e^{O(d)} n^{-1}$ </tex-math></inline-formula>, but only if <inline-formula> <tex-math notation=\"LaTeX\">$P$ </tex-math></inline-formula> achieves finite input-output <inline-formula> <tex-math notation=\"LaTeX\">$\\chi ^{2}$ </tex-math></inline-formula> mutual information across the additive white Gaussian noise (AWGN) channel. If the latter condition is not met, the rate changes to <inline-formula> <tex-math notation=\"LaTeX\">$\\omega \\left ({n^{-1}}\\right)$ </tex-math></inline-formula> for the KL divergence and <inline-formula> <tex-math notation=\"LaTeX\">$\\mathsf {W}_{2}^{2}$ </tex-math></inline-formula>, while the <inline-formula> <tex-math notation=\"LaTeX\">$\\chi ^{2}$ </tex-math></inline-formula>-divergence becomes infinite \u2013 a curious dichotomy. As an application we consider estimating the differential entropy <inline-formula> <tex-math notation=\"LaTeX\">$h(S+Z)$ </tex-math></inline-formula>, where <inline-formula> <tex-math notation=\"LaTeX\">$S\\sim P$ </tex-math></inline-formula> and <inline-formula> <tex-math notation=\"LaTeX\">$Z\\sim \\mathcal {N}_\\sigma $ </tex-math></inline-formula> are independent <inline-formula> <tex-math notation=\"LaTeX\">$d$ </tex-math></inline-formula>-dimensional random variables. The distribution <inline-formula> <tex-math notation=\"LaTeX\">$P$ </tex-math></inline-formula> is unknown and belongs to some nonparametric class, but <italic>n</italic> independently and identically distributed (i.i.d) samples from it are available. Despite the regularizing effect of noise, we first show that any good estimator (within an additive gap) for this problem must have a sample complexity that is exponential in <italic>d</italic>. We then leverage the above empirical approximation results to show that the absolute-error risk of the plug-in estimator converges as <inline-formula> <tex-math notation=\"LaTeX\">$e^{O(d)} n^{-1/2}$ </tex-math></inline-formula>, thus attaining the parametric rate in <italic>n</italic>. This establishes the plug-in estimator as minimax rate-optimal for the considered problem, with sharp dependence of the convergence rate both in <italic>n</italic> and <italic>d</italic>. We provide numerical results comparing the performance of the plug-in estimator to that of general-purpose (unstructured) differential entropy estimators (based on kernel density estimation (KDE) or <italic>k</italic> nearest neighbors (kNN) techniques) applied to samples of <inline-formula> <tex-math notation=\"LaTeX\">$S+Z$ </tex-math></inline-formula>. These results reveal a significant empirical superiority of the plug-in to state-of-the-art KDE and kNN methods. As a motivating utilization of the plug-in approach, we estimate information flows in deep neural networks and discuss Tishby\u2019s Information Bottleneck and the compression conjecture, among others."
                },
                {
                    "title": "On Variational Bounds of Mutual Information",
                    "abstract": "Estimating and optimizing Mutual Information (MI) is core to many problems in machine learning; however, bounding MI in high dimensions is challenging. To establish tractable and scalable objectives, recent work has turned to variational bounds parameterized by neural networks, but the relationships and tradeoffs between these bounds remains unclear. In this work, we unify these recent developments in a single framework. We find that the existing variational lower bounds degrade when the MI is large, exhibiting either high bias or high variance. To address this problem, we introduce a continuum of lower bounds that encompasses previous bounds and flexibly trades off bias and variance. On high-dimensional, controlled problems, we empirically characterize the bias and variance of the bounds and their gradients and demonstrate the effectiveness of our new bounds for estimation and representation learning."
                },
                {
                    "title": "Calculus On Manifolds: A Modern Approach To Classical Theorems Of Advanced Calculus",
                    "abstract": "Functions on Euclidean Space * Norm and inner Product * Subsets of Euclidean Space * Functions and Continuity Differentiation * Basic Definitions * Basic Theorems * Partial Derivatives * Inverse Functions * Implicit Functions * Notation Integration * Basic Definitions * Measure Zero and Content Zero * Integrable Functions * Fubinis Theorem * Partitions of Unity * Change of Variable Integration on Chains * Algebraic Preliminaries * Fields and Forms * Geometric Preliminaries * The Fundamental Theorem of Calculus Integration on Manifolds * Manifolds * Fields and Forms on Manifolds * Stokes Theorem on Manifolds * The Volume Element * The Classical Theorems"
                },
                {
                    "title": "Pl\u00e4ne",
                    "abstract": ". In this theoretical paper we consider the problem of accurately triangulating a scene plane. Rather than \ufb01rst triangulating a set of points and then \ufb01tting a plane to these points, we try to minimize the back-projection errors as functions of the plane parameters directly. As this is both geometrically and statistically meaningful our method performs better than the standard two step procedure. Furthermore, we show that the error residuals of this formulation are quasiconvex thereby making it very easy to solve using for example standard local optimization methods."
                },
                {
                    "title": "Estimating Information Flow in Deep Neural Networks",
                    "abstract": "We study the estimation of the mutual information I(X;T`) between the input X to a deep neural network (DNN) and the output vector T` of its `th hidden layer (an \u201cinternal representation\u201d). Focusing on feedforward networks with fixed weights and noisy internal representations, we develop a rigorous framework for accurate estimation of I(X;T`). By relating I(X;T`) to information transmission over additive white Gaussian noise channels, we reveal that compression, i.e. reduction in I(X;T`) over the course of training, is driven by progressive geometric clustering of the representations of samples from the same class. Experimental results verify this connection. Finally, we shift focus to purely deterministic DNNs, where I(X;T`) is provably vacuous, and show that nevertheless, these models also cluster inputs belonging to the same class. The binningbased approximation of I(X;T`) employed in past works to measure compression is identified as a measure of clustering, thus clarifying that these experiments were in fact tracking the same clustering phenomenon. Leveraging the clustering perspective, we provide new evidence that compression and generalization may not be causally related and discuss potential future research ideas."
                },
                {
                    "title": "Formal Limitations on the Measurement of Mutual Information",
                    "abstract": "Measuring mutual information from finite data is difficult. Recent work has considered variational methods maximizing a lower bound. In this paper, we prove that serious statistical limitations are inherent to any method of measuring mutual information. More specifically, we show that any distribution-free high-confidence lower bound on mutual information estimated from N samples cannot be larger than O(ln N )."
                },
                {
                    "title": "Representation Learning with Contrastive Predictive Coding",
                    "abstract": "While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments."
                },
                {
                    "title": "Glow: Generative Flow with Invertible 1x1 Convolutions",
                    "abstract": "Flow-based generative models (Dinh et al., 2014) are conceptually attractive due to tractability of the exact log-likelihood, tractability of exact latent-variable inference, and parallelizability of both training and synthesis. In this paper we propose Glow, a simple type of generative flow using an invertible 1x1 convolution. Using our method we demonstrate a significant improvement in log-likelihood on standard benchmarks. Perhaps most strikingly, we demonstrate that a generative model optimized towards the plain log-likelihood objective is capable of efficient realistic-looking synthesis and manipulation of large images. The code for our model is available at this https URL"
                },
                {
                    "title": "Understanding Neural Networks and Individual Neuron Importance via Information-Ordered Cumulative Ablation",
                    "abstract": "In this work, we investigate the use of three information-theoretic quantities\u2014entropy, mutual information with the class variable, and a class selectivity measure based on Kullback\u2013Leibler (KL) divergence\u2014to understand and study the behavior of already trained fully connected feedforward neural networks (NNs). We analyze the connection between these information-theoretic quantities and classification performance on the test set by cumulatively ablating neurons in networks trained on MNIST, FashionMNIST, and CIFAR-10. Our results parallel those recently published by Morcos et al., indicating that class selectivity is not a good indicator for classification performance. However, looking at individual layers separately, both mutual information and class selectivity are positively correlated with classification performance, at least for networks with ReLU activation functions. We provide explanations for this phenomenon and conclude that it is ill-advised to compare the proposed information-theoretic quantities across layers. Furthermore, we show that cumulative ablation of neurons with ascending or descending information-theoretic quantities can be used to formulate hypotheses regarding the joint behavior of multiple neurons, such as redundancy and synergy, with comparably low computational cost. We also draw connections to the information bottleneck theory for NNs."
                },
                {
                    "title": "Invertible Autoencoder for domain adaptation",
                    "abstract": "The unsupervised image-to-image translation aims at finding a mapping between the source ( A ) and target ( B ) image domains, where in many applications aligned image pairs are not available at training. This is an ill-posed learning problem since it requires inferring the joint probability distribution from marginals. Joint learning of coupled mappings F A B : A \u2192 B and F B A : B \u2192 A is commonly used by the state-of-the-art methods, like CycleGAN to learn this translation by introducing cycle consistency requirement to the learning problem, i.e., F A B ( F B A ( B ) ) \u2248 B and F B A ( F A B ( A ) ) \u2248 A . Cycle consistency enforces the preservation of the mutual information between input and translated images. However, it does not explicitly enforce F B A to be an inverse operation to F A B . We propose a new deep architecture that we call invertible autoencoder (InvAuto) to explicitly enforce this relation. This is done by forcing an encoder to be an inverted version of the decoder, where corresponding layers perform opposite mappings and share parameters. The mappings are constrained to be orthonormal. The resulting architecture leads to the reduction of the number of trainable parameters (up to 2 times). We present image translation results on benchmark datasets and demonstrate state-of-the art performance of our approach. Finally, we test the proposed domain adaptation method on the task of road video conversion. We demonstrate that the videos converted with InvAuto have high quality and show that the NVIDIA neural-network-based end-to-end learning system for autonomous driving, known as PilotNet, trained on real road videos performs well when tested on the converted ones."
                },
                {
                    "title": "Mutual Information Neural Estimation",
                    "abstract": "We argue that the estimation of mutual information between high dimensional continuous random variables can be achieved by gradient descent over neural networks. We present a Mutual Information Neural Estimator (MINE) that is linearly scalable in dimensionality as well as in sample size, trainable through back-prop, and strongly consistent. We present a handful of applications on which MINE can be used to minimize or maximize mutual information. We apply MINE to improve adversarially trained generative models. We also use MINE to implement Information Bottleneck, applying it to supervised classification; our results demonstrate substantial improvement in flexibility and performance in these settings."
                },
                {
                    "title": "Nonparametric independence testing via mutual information",
                    "abstract": "\n We propose a test of independence of two multivariate random vectors, given a sample from the underlying population. Our approach is based on the estimation of mutual information, whose decomposition into joint and marginal entropies facilitates the use of recently developed efficient entropy estimators derived from nearest neighbour distances. The proposed critical values may be obtained by simulation in the case where an approximation to one marginal is available or by permuting the data otherwise. This facilitates size guarantees, and we provide local power analyses, uniformly over classes of densities whose mutual information satisfies a lower bound. Our ideas may be extended to provide new goodness-of-fit tests for normal linear models based on assessing the independence of our vector of covariates and an appropriately defined notion of an error vector. The theory is supported by numerical studies on both simulated and real data."
                },
                {
                    "title": "Model-Powered Conditional Independence Test",
                    "abstract": "We consider the problem of non-parametric Conditional Independence testing (CI testing) for continuous random variables. Given i.i.d samples from the joint distribution $f(x,y,z)$ of continuous random vectors $X,Y$ and $Z,$ we determine whether $X \\perp Y | Z$. We approach this by converting the conditional independence test into a classification problem. This allows us to harness very powerful classifiers like gradient-boosted trees and deep neural networks. These models can handle complex probability distributions and allow us to perform significantly better compared to the prior state of the art, for high-dimensional CI testing. The main technical challenge in the classification problem is the need for samples from the conditional product distribution $f^{CI}(x,y,z) = f(x|z)f(y|z)f(z)$ -- the joint distribution if and only if $X \\perp Y | Z.$ -- when given access only to i.i.d. samples from the true joint distribution $f(x,y,z)$. To tackle this problem we propose a novel nearest neighbor bootstrap procedure and theoretically show that our generated samples are indeed close to $f^{CI}$ in terms of total variational distance. We then develop theoretical results regarding the generalization bounds for classification for our problem, which translate into error bounds for CI testing. We provide a novel analysis of Rademacher type classification bounds in the presence of non-i.i.d near-independent samples. We empirically validate the performance of our algorithm on simulated and real datasets and show performance gains over previous methods."
                },
                {
                    "title": "Information-theoretic analysis of generalization capability of learning algorithms",
                    "abstract": "We derive upper bounds on the generalization error of a learning algorithm in terms of the mutual information between its input and output. The bounds provide an information-theoretic understanding of generalization in learning problems, and give theoretical guidelines for striking the right balance between data fit and generalization by controlling the input-output mutual information. We propose a number of methods for this purpose, among which are algorithms that regularize the ERM algorithm with relative entropy or with random noise. Our work extends and leads to nontrivial improvements on the recent results of Russo and Zou."
                },
                {
                    "title": "Opening the Black Box of Deep Neural Networks via Information",
                    "abstract": "Despite their great success, there is still no comprehensive theoretical understanding of learning with Deep Neural Networks (DNNs) or their inner organization. Previous work proposed to analyze DNNs in the \\textit{Information Plane}; i.e., the plane of the Mutual Information values that each layer preserves on the input and output variables. They suggested that the goal of the network is to optimize the Information Bottleneck (IB) tradeoff between compression and prediction, successively, for each layer. In this work we follow up on this idea and demonstrate the effectiveness of the Information-Plane visualization of DNNs. Our main results are: (i) most of the training epochs in standard DL are spent on {\\emph compression} of the input to efficient representation and not on fitting the training labels. (ii) The representation compression phase begins when the training errors becomes small and the Stochastic Gradient Decent (SGD) epochs change from a fast drift to smaller training error into a stochastic relaxation, or random diffusion, constrained by the training error value. (iii) The converged layers lie on or very close to the Information Bottleneck (IB) theoretical bound, and the maps from the input to any hidden layer and from this hidden layer to the output satisfy the IB self-consistent equations. This generalization through noise mechanism is unique to Deep Neural Networks and absent in one layer networks. (iv) The training time is dramatically reduced when adding more hidden layers. Thus the main advantage of the hidden layers is computational. This can be explained by the reduced relaxation time, as this it scales super-linearly (exponentially for simple diffusion) with the information compression from the previous layer."
                },
                {
                    "title": "InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets",
                    "abstract": "This paper describes InfoGAN, an information-theoretic extension to the Generative Adversarial Network that is able to learn disentangled representations in a completely unsupervised manner. InfoGAN is a generative adversarial network that also maximizes the mutual information between a small subset of the latent variables and the observation. We derive a lower bound to the mutual information objective that can be optimized efficiently, and show that our training procedure can be interpreted as a variation of the Wake-Sleep algorithm. Specifically, InfoGAN successfully disentangles writing styles from digit shapes on the MNIST dataset, pose from lighting of 3D rendered images, and background digits from the central digit on the SVHN dataset. It also discovers visual concepts that include hair styles, presence/absence of eyeglasses, and emotions on the CelebA face dataset. Experiments show that InfoGAN learns interpretable representations that are competitive with representations learned by existing fully supervised methods."
                },
                {
                    "title": "Efficient multivariate entropy estimation via $k$-nearest neighbour distances",
                    "abstract": "Many statistical procedures, including goodness-of-fit tests and methods for independent component analysis, rely critically on the estimation of the entropy of a distribution. In this paper, we seek entropy estimators that are efficient and achieve the local asymptotic minimax lower bound with respect to squared error loss. To this end, we study weighted averages of the estimators originally proposed by Kozachenko and Leonenko (1987), based on the $k$-nearest neighbour distances of a sample of $n$ independent and identically distributed random vectors in $\\mathbb{R}^d$. A careful choice of weights enables us to obtain an efficient estimator in arbitrary dimensions, given sufficient smoothness, while the original unweighted estimator is typically only efficient when $d \\leq 3$. In addition to the new estimator proposed and theoretical understanding provided, our results facilitate the construction of asymptotically valid confidence intervals for the entropy of asymptotically minimal width."
                },
                {
                    "title": "Density estimation using Real NVP",
                    "abstract": "Unsupervised learning of probabilistic models is a central yet challenging problem in machine learning. Specifically, designing models with tractable learning, sampling, inference and evaluation is crucial in solving this task. We extend the space of such models using real-valued non-volume preserving (real NVP) transformations, a set of powerful invertible and learnable transformations, resulting in an unsupervised learning algorithm with exact log-likelihood computation, exact sampling, exact inference of latent variables, and an interpretable latent space. We demonstrate its ability to model natural images on four datasets through sampling, log-likelihood evaluation and latent variable manipulations."
                },
                {
                    "title": "Variational Inference with Normalizing Flows",
                    "abstract": "The choice of approximate posterior distribution is one of the core problems in variational inference. Most applications of variational inference employ simple families of posterior approximations in order to allow for efficient inference, focusing on mean-field or other simple structured approximations. This restriction has a significant impact on the quality of inferences made using variational methods. We introduce a new approach for specifying flexible, arbitrarily complex and scalable approximate posterior distributions. Our approximations are distributions constructed through a normalizing flow, whereby a simple initial density is transformed into a more complex one by applying a sequence of invertible transformations until a desired level of complexity is attained. We use this view of normalizing flows to develop categories of finite and infinitesimal flows and provide a unified view of approaches for constructing rich posterior approximations. We demonstrate that the theoretical advantages of having posteriors that better match the true posterior, combined with the scalability of amortized variational approaches, provides a clear improvement in performance and applicability of variational inference."
                },
                {
                    "title": "Deep learning and the information bottleneck principle",
                    "abstract": "Deep Neural Networks (DNNs) are analyzed via the theoretical framework of the information bottleneck (IB) principle. We first show that any DNN can be quantified by the mutual information between the layers and the input and output variables. Using this representation we can calculate the optimal information theoretic limits of the DNN and obtain finite sample generalization bounds. The advantage of getting closer to the theoretical limit is quantifiable both by the generalization bound and by the network's simplicity. We argue that both the optimal architecture, number of layers and features/connections at each layer, are related to the bifurcation points of the information bottleneck tradeoff, namely, relevant compression of the input layer with respect to the output layer. The hierarchical representations at the layered network naturally correspond to the structural phase transitions along the information curve. We believe that this new insight can lead to new optimality bounds and deep learning algorithms."
                },
                {
                    "title": "NICE: Non-linear Independent Components Estimation",
                    "abstract": "We propose a deep learning framework for modeling complex high-dimensional densities called Non-linear Independent Component Estimation (NICE). It is based on the idea that a good representation is one in which the data has a distribution that is easy to model. For this purpose, a non-linear deterministic transformation of the data is learned that maps it to a latent space so as to make the transformed data conform to a factorized distribution, i.e., resulting in independent latent variables. We parametrize this transformation so that computing the Jacobian determinant and inverse transform is trivial, yet we maintain the ability to learn complex non-linear transformations, via a composition of simple building blocks, each based on a deep neural network. The training criterion is simply the exact log-likelihood, which is tractable. Unbiased ancestral sampling is also easy. We show that this approach yields good generative models on four image datasets and can be used for inpainting."
                },
                {
                    "title": "Testing the Manifold Hypothesis",
                    "abstract": "The hypothesis that high dimensional data tend to lie in the vicinity of a low dimensional manifold is the basis of manifold learning. The goal of this paper is to develop an algorithm (with accompanying complexity guarantees) for fitting a manifold to an unknown probability distribution supported in a separable Hilbert space, only using i.i.d samples from that distribution. More precisely, our setting is the following. Suppose that data are drawn independently at random from a probability distribution $P$ supported on the unit ball of a separable Hilbert space $H$. Let $G(d, V, \\tau)$ be the set of submanifolds of the unit ball of $H$ whose volume is at most $V$ and reach (which is the supremum of all $r$ such that any point at a distance less than $r$ has a unique nearest point on the manifold) is at least $\\tau$. Let $L(M, P)$ denote mean-squared distance of a random point from the probability distribution $P$ to $M$. \nWe obtain an algorithm that tests the manifold hypothesis in the following sense. \nThe algorithm takes i.i.d random samples from $P$ as input, and determines which of the following two is true (at least one must be): \n(a) There exists $M \\in G(d, CV, \\frac{\\tau}{C})$ such that $L(M, P) \\leq C \\epsilon.$ \n(b) There exists no $M \\in G(d, V/C, C\\tau)$ such that $L(M, P) \\leq \\frac{\\epsilon}{C}.$ \nThe answer is correct with probability at least $1-\\delta$."
                },
                {
                    "title": "Shannon Entropy and Mutual Information for Multivariate Skew\u2010Elliptical Distributions",
                    "abstract": "Abstract.\u2002 The entropy and mutual information index are important concepts developed by Shannon in the context of information theory. They have been widely studied in the case of the multivariate normal distribution. We first extend these tools to the full symmetric class of multivariate elliptical distributions and then to the more flexible families of multivariate skew\u2010elliptical distributions. We study in detail the cases of the multivariate skew\u2010normal and skew\u2010t distributions. We implement our findings to the application of the optimal design of an ozone monitoring station network in Santiago de Chile."
                },
                {
                    "title": "A Family of Nonparametric Density Estimation Algorithms",
                    "abstract": "A new methodology for density estimation is proposed. The methodology, which builds on the one developed by Tabak and Vanden\u2010Eijnden, normalizes the data points through the composition of simple maps. The parameters of each map are determined through the maximization of a local quadratic approximation to the log\u2010likelihood. Various candidates for the elementary maps of each step are proposed; criteria for choosing one includes robustness, computational simplicity, and good behavior in high\u2010dimensional settings. A good choice is that of localized radial expansions, which depend on a single parameter: all the complexity of arbitrary, possibly convoluted probability densities can be built through the composition of such simple maps. \u00a9 2012 Wiley Periodicals, Inc."
                },
                {
                    "title": "DENSITY ESTIMATION BY DUAL ASCENT OF THE LOG-LIKELIHOOD \u2217",
                    "abstract": "A methodology is developed to assign, from an observed sample, a joint-probability distribution to a set of continuous variables. The algorithm proposed performs this assignment by mapping the original variables onto a jointly-Gaussian set. The map is built iteratively, ascending the log-likelihood of the observations, through a series of steps that move the marginal distributions along a random set of orthogonal directions towards normality. AMS subject classifications. 34A50, 65C30, 65L20, 60H35. 1. Introduction and problem setting Extracting information from data is a fundamental problem underlying many ap- plications. Medical doctors seek to diagnose a patient's health from clinical data, blood tests and genetic information. Pharmaceutical companies analyze the results of massive in vitro tests of different compounds to select the best candidate for new drug development. Insurance companies assess, based on financial data, the probabil- ity that a number of credit-lines go into default within the same time-window. Using commercial data, market analysts attempt to quantify the effect that advertising cam- paigns have on sales. Weather forecasters extract from present and past observations the likely state of the weather in the near future. Climate scientists estimate long-time trends from observations over the years of quantities such as sea-surface temperature and the atmospheric concentration of CO2. In many of these applications, the fundamental \"data problem\" consists of es- timating, from a sample of a set of interdependent variables, their joint probability distribution. Thus, the financial analyst dealing in credit derivatives seeks the proba- bility of joint default of many debts over a specified time window; the medical doctor, the likelihood that a patient's test results are associated with a certain disease; the weather forecaster, the likelihood that the pattern of today's measurements anticipate tomorrow's rain."
                },
                {
                    "title": "Estimating divergence functionals and the likelihood ratio by penalized convex risk minimization",
                    "abstract": "We develop and analyze an algorithm for nonparametric estimation of divergence functionals and the density ratio of two probability distributions. Our method is based on a variational characterization of f-divergences, which turns the estimation into a penalized convex risk minimization problem. We present a derivation of our kernel-based estimation algorithm and an analysis of convergence rates for the estimator. Our simulation results demonstrate the convergence behavior of the method, which compares favorably with existing methods in the literature."
                },
                {
                    "title": "Elements of Information Theory",
                    "abstract": "Preface to the Second Edition. Preface to the First Edition. Acknowledgments for the Second Edition. Acknowledgments for the First Edition. 1. Introduction and Preview. 1.1 Preview of the Book. 2. Entropy, Relative Entropy, and Mutual Information. 2.1 Entropy. 2.2 Joint Entropy and Conditional Entropy. 2.3 Relative Entropy and Mutual Information. 2.4 Relationship Between Entropy and Mutual Information. 2.5 Chain Rules for Entropy, Relative Entropy, and Mutual Information. 2.6 Jensen's Inequality and Its Consequences. 2.7 Log Sum Inequality and Its Applications. 2.8 Data-Processing Inequality. 2.9 Sufficient Statistics. 2.10 Fano's Inequality. Summary. Problems. Historical Notes. 3. Asymptotic Equipartition Property. 3.1 Asymptotic Equipartition Property Theorem. 3.2 Consequences of the AEP: Data Compression. 3.3 High-Probability Sets and the Typical Set. Summary. Problems. Historical Notes. 4. Entropy Rates of a Stochastic Process. 4.1 Markov Chains. 4.2 Entropy Rate. 4.3 Example: Entropy Rate of a Random Walk on a Weighted Graph. 4.4 Second Law of Thermodynamics. 4.5 Functions of Markov Chains. Summary. Problems. Historical Notes. 5. Data Compression. 5.1 Examples of Codes. 5.2 Kraft Inequality. 5.3 Optimal Codes. 5.4 Bounds on the Optimal Code Length. 5.5 Kraft Inequality for Uniquely Decodable Codes. 5.6 Huffman Codes. 5.7 Some Comments on Huffman Codes. 5.8 Optimality of Huffman Codes. 5.9 Shannon-Fano-Elias Coding. 5.10 Competitive Optimality of the Shannon Code. 5.11 Generation of Discrete Distributions from Fair Coins. Summary. Problems. Historical Notes. 6. Gambling and Data Compression. 6.1 The Horse Race. 6.2 Gambling and Side Information. 6.3 Dependent Horse Races and Entropy Rate. 6.4 The Entropy of English. 6.5 Data Compression and Gambling. 6.6 Gambling Estimate of the Entropy of English. Summary. Problems. Historical Notes. 7. Channel Capacity. 7.1 Examples of Channel Capacity. 7.2 Symmetric Channels. 7.3 Properties of Channel Capacity. 7.4 Preview of the Channel Coding Theorem. 7.5 Definitions. 7.6 Jointly Typical Sequences. 7.7 Channel Coding Theorem. 7.8 Zero-Error Codes. 7.9 Fano's Inequality and the Converse to the Coding Theorem. 7.10 Equality in the Converse to the Channel Coding Theorem. 7.11 Hamming Codes. 7.12 Feedback Capacity. 7.13 Source-Channel Separation Theorem. Summary. Problems. Historical Notes. 8. Differential Entropy. 8.1 Definitions. 8.2 AEP for Continuous Random Variables. 8.3 Relation of Differential Entropy to Discrete Entropy. 8.4 Joint and Conditional Differential Entropy. 8.5 Relative Entropy and Mutual Information. 8.6 Properties of Differential Entropy, Relative Entropy, and Mutual Information. Summary. Problems. Historical Notes. 9. Gaussian Channel. 9.1 Gaussian Channel: Definitions. 9.2 Converse to the Coding Theorem for Gaussian Channels. 9.3 Bandlimited Channels. 9.4 Parallel Gaussian Channels. 9.5 Channels with Colored Gaussian Noise. 9.6 Gaussian Channels with Feedback. Summary. Problems. Historical Notes. 10. Rate Distortion Theory. 10.1 Quantization. 10.2 Definitions. 10.3 Calculation of the Rate Distortion Function. 10.4 Converse to the Rate Distortion Theorem. 10.5 Achievability of the Rate Distortion Function. 10.6 Strongly Typical Sequences and Rate Distortion. 10.7 Characterization of the Rate Distortion Function. 10.8 Computation of Channel Capacity and the Rate Distortion Function. Summary. Problems. Historical Notes. 11. Information Theory and Statistics. 11.1 Method of Types. 11.2 Law of Large Numbers. 11.3 Universal Source Coding. 11.4 Large Deviation Theory. 11.5 Examples of Sanov's Theorem. 11.6 Conditional Limit Theorem. 11.7 Hypothesis Testing. 11.8 Chernoff-Stein Lemma. 11.9 Chernoff Information. 11.10 Fisher Information and the Cram-er-Rao Inequality. Summary. Problems. Historical Notes. 12. Maximum Entropy. 12.1 Maximum Entropy Distributions. 12.2 Examples. 12.3 Anomalous Maximum Entropy Problem. 12.4 Spectrum Estimation. 12.5 Entropy Rates of a Gaussian Process. 12.6 Burg's Maximum Entropy Theorem. Summary. Problems. Historical Notes. 13. Universal Source Coding. 13.1 Universal Codes and Channel Capacity. 13.2 Universal Coding for Binary Sequences. 13.3 Arithmetic Coding. 13.4 Lempel-Ziv Coding. 13.5 Optimality of Lempel-Ziv Algorithms. Compression. Summary. Problems. Historical Notes. 14. Kolmogorov Complexity. 14.1 Models of Computation. 14.2 Kolmogorov Complexity: Definitions and Examples. 14.3 Kolmogorov Complexity and Entropy. 14.4 Kolmogorov Complexity of Integers. 14.5 Algorithmically Random and Incompressible Sequences. 14.6 Universal Probability. 14.7 Kolmogorov complexity. 14.9 Universal Gambling. 14.10 Occam's Razor. 14.11 Kolmogorov Complexity and Universal Probability. 14.12 Kolmogorov Sufficient Statistic. 14.13 Minimum Description Length Principle. Summary. Problems. Historical Notes. 15. Network Information Theory. 15.1 Gaussian Multiple-User Channels. 15.2 Jointly Typical Sequences. 15.3 Multiple-Access Channel. 15.4 Encoding of Correlated Sources. 15.5 Duality Between Slepian-Wolf Encoding and Multiple-Access Channels. 15.6 Broadcast Channel. 15.7 Relay Channel. 15.8 Source Coding with Side Information. 15.9 Rate Distortion with Side Information. 15.10 General Multiterminal Networks. Summary. Problems. Historical Notes. 16. Information Theory and Portfolio Theory. 16.1 The Stock Market: Some Definitions. 16.2 Kuhn-Tucker Characterization of the Log-Optimal Portfolio. 16.3 Asymptotic Optimality of the Log-Optimal Portfolio. 16.4 Side Information and the Growth Rate. 16.5 Investment in Stationary Markets. 16.6 Competitive Optimality of the Log-Optimal Portfolio. 16.7 Universal Portfolios. 16.8 Shannon-McMillan-Breiman Theorem (General AEP). Summary. Problems. Historical Notes. 17. Inequalities in Information Theory. 17.1 Basic Inequalities of Information Theory. 17.2 Differential Entropy. 17.3 Bounds on Entropy and Relative Entropy. 17.4 Inequalities for Types. 17.5 Combinatorial Bounds on Entropy. 17.6 Entropy Rates of Subsets. 17.7 Entropy and Fisher Information. 17.8 Entropy Power Inequality and Brunn-Minkowski Inequality. 17.9 Inequalities for Determinants. 17.10 Inequalities for Ratios of Determinants. Summary. Problems. Historical Notes. Bibliography. List of Symbols. Index."
                },
                {
                    "title": "Estimating mutual information.",
                    "abstract": "We present two classes of improved estimators for mutual information M(X,Y), from samples of random points distributed according to some joint probability density mu(x,y). In contrast to conventional estimators based on binnings, they are based on entropy estimates from k -nearest neighbor distances. This means that they are data efficient (with k=1 we resolve structures down to the smallest possible scales), adaptive (the resolution is higher where data are more numerous), and have minimal bias. Indeed, the bias of the underlying entropy estimates is mainly due to nonuniformity of the density at the smallest resolved scale, giving typically systematic errors which scale as functions of k/N for N points. Numerically, we find that both families become exact for independent distributions, i.e. the estimator M(X,Y) vanishes (up to statistical fluctuations) if mu(x,y)=mu(x)mu(y). This holds for all tested marginal distributions and for all dimensions of x and y. In addition, we give estimators for redundancies between more than two random variables. We compare our algorithms in detail with existing algorithms. Finally, we demonstrate the usefulness of our estimators for assessing the actual independence of components obtained from independent component analysis (ICA), for improving ICA, and for estimating the reliability of blind source separation."
                },
                {
                    "title": "The information bottleneck method",
                    "abstract": "We define the relevant information in a signal $x\\in X$ as being the information that this signal provides about another signal $y\\in \\Y$. Examples include the information that face images provide about the names of the people portrayed, or the information that speech sounds provide about the words spoken. Understanding the signal $x$ requires more than just predicting $y$, it also requires specifying which features of $\\X$ play a role in the prediction. We formalize this problem as that of finding a short code for $\\X$ that preserves the maximum information about $\\Y$. That is, we squeeze the information that $\\X$ provides about $\\Y$ through a `bottleneck' formed by a limited set of codewords $\\tX$. This constrained optimization problem can be seen as a generalization of rate distortion theory in which the distortion measure $d(x,\\x)$ emerges from the joint statistics of $\\X$ and $\\Y$. This approach yields an exact set of self consistent equations for the coding rules $X \\to \\tX$ and $\\tX \\to \\Y$. Solutions to these equations can be found by a convergent re-estimation method that generalizes the Blahut-Arimoto algorithm. Our variational principle provides a surprisingly rich framework for discussing a variety of problems in signal processing and learning, as will be described in detail elsewhere."
                },
                {
                    "title": "The general theory of canonical correlation and its relation to functional analysis",
                    "abstract": "The classical theory of canonical correlation is concerned with a standard description of the relationship between any linear combination of \u03c1 random variables xs, and any linear combination of q random variables yt insofar as this relation can be described in terms of correlation. Lancaster [1] has extended this theory, for p = q = 1, to include a description of the correlation of any function of a random variable x and any function of a random variable y (both functions having finite variance) for a class of joint distributions of x and y which is very general. It is the purpose of this paper to derive Lancaster's results from general theorems concerning the spectral decomposition of operators on a Hilbert space. These theorems lend themselves easily to the generalisation of the theory to situations where p and q are not finite. In the case of Gaussian, stationary, processes this generalisation is equivalent to the classical spectral theory and corresponds to a canonical reduction of a (finite) sample of data which is basic. The theory also then extends to any number of processes. In the Gaussian case, also, the present discussion-is connected with the results of Gelfand and Yaglom [2] relating to the amount of information in one random process about another."
                },
                {
                    "title": "Gaussianization",
                    "abstract": "High dimensional data modeling is difficult mainly because the so-called \"curse of dimensionality\". We propose a technique called \"Gaussianization\" for high dimensional density estimation, which alleviates the curse of dimensionality by exploiting the independence structures in the data. Gaussianization is motivated from recent developments in the statistics literature: projection pursuit, independent component analysis and Gaussian mixture models with semi-tied covariances. We propose an iterative Gaussianization procedure which converges weakly: at each iteration, the data is first transformed to the least dependent coordinates and then each coordinate is marginally Gaussianized by univariate techniques. Gaussianization offers density estimation sharper than traditional kernel methods and radial basis function methods. Gaussianization can be viewed as efficient solution of nonlinear independent component analysis and high dimensional projection pursuit."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.IT",
                "math.IT",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can information-theoretic analysis be effectively applied to improve the understanding and performance of deep neural networks (DNNs)?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it can lead to a deeper understanding of the fundamental principles governing DNNs, potentially revealing insights into their behavior, generalization capabilities, and limitations. This could pave the way for more robust and interpretable models, influencing future research directions in machine learning, optimization techniques, and model design. Additionally, practical applications could emerge in areas such as anomaly detection, feature selection, and model compression, enhancing the efficiency and effectiveness of DNNs in real-world scenarios.\n\n### [Question 3] - Why is it hard?\nThe challenges in addressing this problem stem from the complex nature of DNNs, which often exhibit non-linear behaviors and high-dimensional data interactions. Naive approaches may fail due to the intricate relationships between input features and model outputs, as well as the difficulty in accurately estimating the necessary statistical properties (e.g., covariance matrices) in high-dimensional spaces. Furthermore, the theoretical underpinnings of information theory and its application to DNNs require sophisticated mathematical tools and a deep understanding of both fields, making it a non-trivial task.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on empirical performance metrics rather than theoretical foundations, leading to a lack of comprehensive frameworks that integrate information theory with DNN analysis. Limitations in computational resources and the complexity of deriving closed-form solutions for high-dimensional distributions have also posed significant barriers. Existing solutions may not adequately address the nuances of DNN behavior, and many approaches have not leveraged the full potential of information-theoretic principles. Our approach aims to fill these gaps by providing a rigorous theoretical framework that connects information theory with DNN performance.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves a detailed information-theoretic analysis of DNNs, utilizing Gaussianization techniques to derive closed-form expressions for mutual information and related metrics. We will employ benchmark datasets commonly used in DNN research, such as CIFAR-10 or MNIST, and evaluate our results using metrics like mutual information estimates and model performance (accuracy, F1 score). The expected outcomes include a clearer understanding of the information flow within DNNs, improved model interpretability, and practical guidelines for enhancing DNN design and"
            }
        },
        "author_data": {
            "7f77533f-3b53-44f7-a5e6-98bce367740b": {
                "pk": "7f77533f-3b53-44f7-a5e6-98bce367740b",
                "name": "Ivan Butakov",
                "collaborators": [
                    "Alexander Tolmachev",
                    "Alexey Frolov",
                    "Sofia Malanchuk",
                    "Anna Neopryatnaya",
                    "Kirill Andreev",
                    "Alexander Sememenko",
                    "Andrey Gladkov",
                    "Marina Munkhoeva"
                ],
                "domain": [
                    "Information Theory",
                    "Deep Learning",
                    "Self-Supervised Learning",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "Information Bottleneck Analysis of Deep Neural Networks via Lossy Compression",
                        "abstract": "The Information Bottleneck (IB) principle offers an information-theoretic framework for analyzing the training process of deep neural networks (DNNs). Its essence lies in tracking the dynamics of two mutual information (MI) values: between the hidden layer output and the DNN input/target. According to the hypothesis put forth by Shwartz-Ziv & Tishby (2017), the training process consists of two distinct phases: fitting and compression. The latter phase is believed to account for the good generalization performance exhibited by DNNs. Due to the challenging nature of estimating MI between high-dimensional random vectors, this hypothesis was only partially verified for NNs of tiny sizes or specific types, such as quantized NNs. In this paper, we introduce a framework for conducting IB analysis of general NNs. Our approach leverages the stochastic NN method proposed by Goldfeld et al. (2019) and incorporates a compression step to overcome the obstacles associated with high dimensionality. In other words, we estimate the MI between the compressed representations of high-dimensional random vectors. The proposed method is supported by both theoretical and practical justifications. Notably, we demonstrate the accuracy of our estimator through synthetic experiments featuring predefined MI values and comparison with MINE (Belghazi et al., 2018). Finally, we perform IB analysis on a close-to-real-scale convolutional DNN, which reveals new features of the MI dynamics."
                    },
                    {
                        "title": "Efficient Distribution Matching of Representations via Noise-Injected Deep InfoMax",
                        "abstract": "Deep InfoMax (DIM) is a well-established method for self-supervised representation learning (SSRL) based on maximization of the mutual information between the input and the output of a deep neural network encoder. Despite the DIM and contrastive SSRL in general being well-explored, the task of learning representations conforming to a specific distribution (i.e., distribution matching, DM) is still under-addressed. Motivated by the importance of DM to several downstream tasks (including generative modeling, disentanglement, outliers detection and other), we enhance DIM to enable automatic matching of learned representations to a selected prior distribution. To achieve this, we propose injecting an independent noise into the normalized outputs of the encoder, while keeping the same InfoMax training objective. We show that such modification allows for learning uniformly and normally distributed representations, as well as representations of other absolutely continuous distributions. Our approach is tested on various downstream tasks. The results indicate a moderate trade-off between the performance on the downstream tasks and quality of DM."
                    }
                ]
            },
            "d4d7a1e6-3fea-45a7-b004-bf869efef7de": {
                "pk": "d4d7a1e6-3fea-45a7-b004-bf869efef7de",
                "name": "Alexander Tolmachev",
                "collaborators": [
                    "Dmitry Protasov",
                    "Vsevolod Voronov",
                    "Ivan Butakov",
                    "Alexey Frolov",
                    "Alexander Sememenko",
                    "Andrey Gladkov",
                    "Marina Munkhoeva",
                    "Sofia Malanchuk",
                    "Anna Neopryatnaya",
                    "Kirill Andreev"
                ],
                "domain": [
                    "Optimization",
                    "Information Theory",
                    "Machine Learning",
                    "Astronomy"
                ],
                "publications": [
                    {
                        "title": "Optimal partitions of the flat torus into parts of smaller diameter",
                        "abstract": "We consider the problem of partitioning a two-dimensional flat torus $T^2$ into $m$ sets in order to minimize the maximal diameter of a part. For $m \\leqslant 25$ we give numerical estimates for the maximal diameter $d_m(T^2)$ at which the partition exists. Several approaches are proposed to obtain such estimates. In particular, we use the search for mesh partitions via the SAT solver, the global optimization approach for polygonal partitions, and the optimization of periodic hexagonal tilings. For $m=3$, the exact estimate is proved using elementary topological reasoning."
                    },
                    {
                        "title": "Efficient Distribution Matching of Representations via Noise-Injected Deep InfoMax",
                        "abstract": "Deep InfoMax (DIM) is a well-established method for self-supervised representation learning (SSRL) based on maximization of the mutual information between the input and the output of a deep neural network encoder. Despite the DIM and contrastive SSRL in general being well-explored, the task of learning representations conforming to a specific distribution (i.e., distribution matching, DM) is still under-addressed. Motivated by the importance of DM to several downstream tasks (including generative modeling, disentanglement, outliers detection and other), we enhance DIM to enable automatic matching of learned representations to a selected prior distribution. To achieve this, we propose injecting an independent noise into the normalized outputs of the encoder, while keeping the same InfoMax training objective. We show that such modification allows for learning uniformly and normally distributed representations, as well as representations of other absolutely continuous distributions. Our approach is tested on various downstream tasks. The results indicate a moderate trade-off between the performance on the downstream tasks and quality of DM."
                    },
                    {
                        "title": "Coverings of planar and three-dimensional sets with subsets of smaller diameter",
                        "abstract": "Quantitative estimates related to the classical Borsuk problem of splitting set in Euclidean space into subsets of smaller diameter are considered. For a given $k$ there is a minimal diameter of subsets at which there exists a covering with $k$ subsets of any planar set of unit diameter. In order to find an upper estimate of the minimal diameter we propose an algorithm for finding sub-optimal partitions. In the cases $10 \\leqslant k \\leqslant 17$ some upper and lower estimates of the minimal diameter are improved. Another result is that any set $M \\subset \\mathbb{R}^3$ of a unit diameter can be partitioned into four subsets of a diameter not greater than $0.966$."
                    },
                    {
                        "title": "Information Bottleneck Analysis of Deep Neural Networks via Lossy Compression",
                        "abstract": "The Information Bottleneck (IB) principle offers an information-theoretic framework for analyzing the training process of deep neural networks (DNNs). Its essence lies in tracking the dynamics of two mutual information (MI) values: between the hidden layer output and the DNN input/target. According to the hypothesis put forth by Shwartz-Ziv & Tishby (2017), the training process consists of two distinct phases: fitting and compression. The latter phase is believed to account for the good generalization performance exhibited by DNNs. Due to the challenging nature of estimating MI between high-dimensional random vectors, this hypothesis was only partially verified for NNs of tiny sizes or specific types, such as quantized NNs. In this paper, we introduce a framework for conducting IB analysis of general NNs. Our approach leverages the stochastic NN method proposed by Goldfeld et al. (2019) and incorporates a compression step to overcome the obstacles associated with high dimensionality. In other words, we estimate the MI between the compressed representations of high-dimensional random vectors. The proposed method is supported by both theoretical and practical justifications. Notably, we demonstrate the accuracy of our estimator through synthetic experiments featuring predefined MI values and comparison with MINE (Belghazi et al., 2018). Finally, we perform IB analysis on a close-to-real-scale convolutional DNN, which reveals new features of the MI dynamics."
                    },
                    {
                        "title": "H2O MegaMasers: a RadioAstron success story",
                        "abstract": "The RadioAstron space-VLBI mission has successfully detected extragalactic H2O MegaMaser emission regions at very long Earth to space baselines ranging between 1.4 and 26.7 Earth Diameters (ED). The preliminary results for two galaxies, NGC3079 and NGC4258, at baselines longer than one ED indicate masering environments and excitation conditions in these galaxies that are distinctly different. Further observations of NGC4258 at longer baselines will reveal more of the physics of individual emission regions."
                    }
                ]
            },
            "5149fe9c-0675-4ce9-8249-02e09bec6d09": {
                "pk": "5149fe9c-0675-4ce9-8249-02e09bec6d09",
                "name": "Sofia Malanchuk",
                "collaborators": [
                    "Ivan Butakov",
                    "Alexander Tolmachev",
                    "Anna Neopryatnaya",
                    "Alexey Frolov",
                    "Kirill Andreev"
                ],
                "domain": [
                    "Information Theory",
                    "Deep Learning",
                    "Neural Networks",
                    "Mutual Information"
                ],
                "publications": [
                    {
                        "title": "Information Bottleneck Analysis of Deep Neural Networks via Lossy Compression",
                        "abstract": "The Information Bottleneck (IB) principle offers an information-theoretic framework for analyzing the training process of deep neural networks (DNNs). Its essence lies in tracking the dynamics of two mutual information (MI) values: between the hidden layer output and the DNN input/target. According to the hypothesis put forth by Shwartz-Ziv & Tishby (2017), the training process consists of two distinct phases: fitting and compression. The latter phase is believed to account for the good generalization performance exhibited by DNNs. Due to the challenging nature of estimating MI between high-dimensional random vectors, this hypothesis was only partially verified for NNs of tiny sizes or specific types, such as quantized NNs. In this paper, we introduce a framework for conducting IB analysis of general NNs. Our approach leverages the stochastic NN method proposed by Goldfeld et al. (2019) and incorporates a compression step to overcome the obstacles associated with high dimensionality. In other words, we estimate the MI between the compressed representations of high-dimensional random vectors. The proposed method is supported by both theoretical and practical justifications. Notably, we demonstrate the accuracy of our estimator through synthetic experiments featuring predefined MI values and comparison with MINE (Belghazi et al., 2018). Finally, we perform IB analysis on a close-to-real-scale convolutional DNN, which reveals new features of the MI dynamics."
                    }
                ]
            },
            "816b2f7f-679b-4209-9cde-88381050aa28": {
                "pk": "816b2f7f-679b-4209-9cde-88381050aa28",
                "name": "Anna Neopryatnaya",
                "collaborators": [
                    "Vsevolod Voronov",
                    "Eugene Dergachev",
                    "Ivan Butakov",
                    "Alexander Tolmachev",
                    "Sofia Malanchuk",
                    "Alexey Frolov",
                    "Kirill Andreev"
                ],
                "domain": [
                    "Graph Theory",
                    "Deep Learning",
                    "Information Theory",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Constructing 5-chromatic unit distance graphs embedded in the Euclidean plane and two-dimensional spheres",
                        "abstract": "This paper is devoted to the development of algorithms for finding unit distance graphs with chromatic number greater than 4, embedded in a two-dimensional sphere or plane. Such graphs provide a lower bound for the Nelson-Hadwiger problem on the chromatic number of the plane and its generalizations to the case of the sphere. A series of 5-chromatic unit distance graphs on 64513 vertices embedded into the plane is constructed. Unlike previously known examples, these graphs do not contain the Moser spindle as a subgraph. The construction of 5-chromatic graphs embedded in a sphere at two values of the radius is given. Namely, the 5-chromatic unit distance graph on 372 vertices embedded into the circumsphere of an icosahedron with a unit edge length, and the 5-chromatic graph on 972 vertices embedded into the circumsphere of a great icosahedron are constructed."
                    },
                    {
                        "title": "Information Bottleneck Analysis of Deep Neural Networks via Lossy Compression",
                        "abstract": "The Information Bottleneck (IB) principle offers an information-theoretic framework for analyzing the training process of deep neural networks (DNNs). Its essence lies in tracking the dynamics of two mutual information (MI) values: between the hidden layer output and the DNN input/target. According to the hypothesis put forth by Shwartz-Ziv & Tishby (2017), the training process consists of two distinct phases: fitting and compression. The latter phase is believed to account for the good generalization performance exhibited by DNNs. Due to the challenging nature of estimating MI between high-dimensional random vectors, this hypothesis was only partially verified for NNs of tiny sizes or specific types, such as quantized NNs. In this paper, we introduce a framework for conducting IB analysis of general NNs. Our approach leverages the stochastic NN method proposed by Goldfeld et al. (2019) and incorporates a compression step to overcome the obstacles associated with high dimensionality. In other words, we estimate the MI between the compressed representations of high-dimensional random vectors. The proposed method is supported by both theoretical and practical justifications. Notably, we demonstrate the accuracy of our estimator through synthetic experiments featuring predefined MI values and comparison with MINE (Belghazi et al., 2018). Finally, we perform IB analysis on a close-to-real-scale convolutional DNN, which reveals new features of the MI dynamics."
                    }
                ]
            },
            "2988d460-6e1e-4a19-b454-eb2cc9f60f5d": {
                "pk": "2988d460-6e1e-4a19-b454-eb2cc9f60f5d",
                "name": "Alexey Frolov",
                "collaborators": [
                    "Pavel Rybin",
                    "Kirill Andreev",
                    "Victor Zyablov",
                    "Stanislav Kruglik",
                    "Valeriya Potapova",
                    "Anton Glebov",
                    "Andrea Munari",
                    "Daria Ustinova",
                    "Vladimir Sidorenko",
                    "Robert Fischer"
                ],
                "domain": [
                    "Coding Theory",
                    "LDPC Codes",
                    "Information Theory",
                    "Wireless Communication"
                ],
                "publications": [
                    {
                        "title": "An Upper Bound on the Minimum Distance of LDPC Codes over GF(q)",
                        "abstract": "In [1] a syndrome counting based upper bound on the minimum distance of regular binary LDPC codes is given. In this paper we extend the bound to the case of irregular and generalized LDPC codes over GF(q). The comparison to the lower bound for LDPC codes over GF(q) and to the upper bound for non-binary codes is done. The new bound is shown to lie under the Gilbert-Varshamov bound at high rates."
                    },
                    {
                        "title": "Upper Bounds on the Minimum Distance of Quasi-Cyclic LDPC codes Revisited",
                        "abstract": "Two upper bounds on the minimum distance of type-1 quasi-cyclic low-density parity-check (QC LDPC) codes are derived. The necessary condition is given for the minimum code distance of such codes to grow linearly with the code length."
                    },
                    {
                        "title": "Upper and Lower Bounds on the Minimum Distance of Expander Codes",
                        "abstract": "The minimum distance of expander codes over GF(q) is studied. A new upper bound on the minimum distance of expander codes is derived. The bound is shown to lie under the Varshamov-Gilbert (VG) bound while q >= 32. Lower bounds on the minimum distance of some families of expander codes are obtained. A lower bound on the minimum distance of low-density parity-check (LDPC) codes with a Reed--Solomon constituent code over GF(q) is obtained. The bound is shown to be very close to the VG bound and to lie above the upper bound for expander codes."
                    },
                    {
                        "title": "On the Multiple Threshold Decoding of LDPC codes over GF(q)",
                        "abstract": "We consider the decoding of LDPC codes over GF(q) with the low-complexity majority algorithm from [1]. A modification of this algorithm with multiple thresholds is suggested. A lower estimate on the decoding radius realized by the new algorithm is derived. The estimate is shown to be better than the estimate for a single threshold majority decoder. At the same time the transition to multiple thresholds does not affect the order of complexity."
                    },
                    {
                        "title": "Bounds and Constructions of Codes with All-Symbol Locality and Availability",
                        "abstract": "We investigate the distance properties of linear locally recoverable codes (LRC codes) with all-symbol locality and availability. New upper and lower bounds on the minimum distance of such codes are derived. The upper bound is based on the shortening method and improves existing shortening bounds. To reduce the gap in between upper and lower bounds we do not restrict the alphabet size and propose explicit constructions of codes with locality and availability via rank-metric codes. The first construction relies on expander graphs and is better in low rate region, the second construction utilizes LRC codes developed by Wang et al. as inner codes and better in high rate region."
                    },
                    {
                        "title": "On the Capacity of the Multiuser Vector Adder Channel",
                        "abstract": "We investigate the capacity of the $Q$-frequency $S$-user vector adder channel (channel with intensity information) introduced by Chang and Wolf. Both coordinated and uncoordinated types of transmission are considered. Asymptotic (under the conditions $Q \\to \\infty$, $S = \\gamma Q$ and $0 < \\gamma < \\infty$) upper and lower bounds on the relative (per subchannel) capacity are derived. The lower bound for the coordinated case is shown to increase when $\\gamma$ grows. At the same time the relative capacity for the uncoordinated case is upper bounded by a constant."
                    },
                    {
                        "title": "Average Age of Information of Irregular Repetition Slotted ALOHA",
                        "abstract": "Flanking traditional metrics such as throughput and reliability, age of information (AoI) is emerging as a fundamental tool to capture the performance of IoT systems. In this context, we focus on a setup in which a large number of nodes attempt delivery of time-stamped updates to a common destination over a shared channel, and investigate the ability of different grant-free access strategies to maintain fresh information at the receiver. Specifically, we derive for the first time an exact closed-form expression of the average AoI achieved by irregular repetition slotted ALOHA (IRSA), and compare its performance to that of a slotted ALOHA approach. Our analysis reveals the potential of modern random access schemes, and pinpoints some fundamental trade-offs, providing useful hints for proper system design."
                    },
                    {
                        "title": "On Multilevel Coding Schemes Based on Non-Binary LDPC Codes",
                        "abstract": "We address the problem of constructing of coding schemes for the channels with high-order modulations. It is known, that non-binary LDPC codes are especially good for such channels and significantly outperform their binary counterparts. Unfortunately, their decoding complexity is still large. In order to reduce the decoding complexity we consider multilevel coding schemes based on non-binary LDPC codes (NB-LDPC-MLC schemes) over smaller fields. The use of such schemes gives us a reasonable gain in complexity. At the same time the performance of NB-LDPC-MLC schemes is practically the same as the performance of LDPC codes over the field matching the modulation order. In particular by means of simulations we showed that the performance of NB-LDPC-MLC schemes over GF(16) is the same as the performance of non-binary LDPC codes over GF(64) and GF(256) in AWGN channel with QAM64 and QAM256 accordingly. We also perform a comparison with binary LDPC codes."
                    },
                    {
                        "title": "A method for constructing parity-check matrices of non-binary quasi-cyclic LDPC codes",
                        "abstract": "An algorithm for constructing parity-check matrices of non-binary quasi-cyclic low-density parity-check (NB QC-LDPC) codes is proposed. The algorithm finds short cycles in the base matrix and tries to eliminate them by selecting the circulants and the elements of GF(q). Firstly the algorithm tries to eliminate the cycles with the smallest number edges going outside the cycle. The efficiency of the algorithm is demonstrated by means of simulations. In particular, it was shown that NB QC-LDPC codes constructed with use of our algorithm loose less that 0.1 dB in comparison to the best NB LDPC codes."
                    },
                    {
                        "title": "Unsourced Random Access with the MIMO Receiver: Projection Decoding Analysis",
                        "abstract": "We consider unsourced random access with MIMO receiver - a crucial communication scenario for future 5G/6G wireless networks. We perform a projection-based decoder analysis and derive energy efficiency achievability bounds when channel state information is unknown at transmitters and the receiver (no-CSI scenario). The comparison to the maximum-likelihood (ML) achievability bounds by Gao et al. (2023) is performed. We show that there is a region where the new bound outperforms the ML bound. The latter fact should not surprise the reader as both decoding criteria are suboptimal when considering per-user probability of error (PUPE). Moreover, transition to projection decoding allows for significant dimensionality reduction, which greatly reduces the computation time."
                    },
                    {
                        "title": "On a Multiple-Access in a Vector Disjunctive Channel",
                        "abstract": "We address the problem of increasing the sum rate in a multiple-access system from [1] for small number of users. We suggest an improved signal-code construction in which in case of a small number of users we give more resources to them. For the resulting multiple-access system a lower bound on the relative sum rate is derived. It is shown to be very close to the maximal value of relative sum rate in [1] even for small number of users. The bound is obtained for the case of decoding by exhaustive search. We also suggest reduced-complexity decoding and compare the maximal number of users in this case and in case of decoding by exhaustive search."
                    },
                    {
                        "title": "Bounds on the Parameters of Locally Recoverable Codes",
                        "abstract": "A locally recoverable code (LRC code) is a code over a finite alphabet such that every symbol in the encoding is a function of a small number of other symbols that form a recovering set. In this paper we derive new finite-length and asymptotic bounds on the parameters of LRC codes. For LRC codes with a single recovering set for every coordinate, we derive an asymptotic Gilbert-Varshamov type bound for LRC codes and find the maximum attainable relative distance of asymptotically good LRC codes. Similar results are established for LRC codes with two disjoint recovering sets for every coordinate. For the case of multiple recovering sets we derive a lower bound on the parameters using expander graph arguments. Finally, we also derive finite-length upper bounds on the rate and distance of LRC codes with multiple recovering sets."
                    },
                    {
                        "title": "Secrecy and Accessibility in Distributed Storage",
                        "abstract": "A distributed storage system (DSS) needs to be efficiently accessible and repairable. Recently, considerable effort has been made towards the latter, while the former is usually not considered, since a trivial solution exists in the form of systematic encoding. However, this is not a viable option when considering storage that has to be secure against eavesdroppers. This work investigates the problem of efficient access to data stored on an DSS under such security constraints. Further, we establish methods to balance the access load, i.e., ensure that each node is accessed equally often. We establish the capacity for the alphabet independent case and give an explicit code construction. For the alphabet-dependent case we give existence results based on a random coding argument."
                    },
                    {
                        "title": "A Polar Code Based TIN-SIC Scheme for the Unsourced Random Access in the Quasi-Static Fading MAC",
                        "abstract": "We consider a problem of unsourced random access in the quasi-static Rayleigh fading channel. In the previous work, the authors have proposed LDPC code based solutions based on joint and treat interference as noise in combination with successive interference cancellation (TIN-SIC) decoder architectures. The authors showed that TIN-SIC decoding significantly outperforms the joint decoding approach and much simpler from the implementation point of view. In this paper, we continue the analysis of TIN-SIC decoding. We derive a finite length achievability bound for TIN-SIC decoder using random coding and propose a practical polar code based TIN-SIC scheme. The latter's performance becomes significantly better in comparison to LDPC code based solutions and close to the finite length achievability bound."
                    },
                    {
                        "title": "Coded Compressed Sensing with List Recoverable Codes for the Unsourced Random Access",
                        "abstract": "We consider a coded compressed sensing approach for the unsourced random access and replace the outer tree code proposed by Amalladinne et al. with the list recoverable code capable of correcting t errors. A finite-length random coding bound for such codes is derived. The numerical experiments in the single antenna quasi-static Rayleigh fading MAC show that transition to list recoverable codes correcting t errors improves the performance of coded compressed sensing scheme by 7-10 dB compared to the tree code-based scheme. We propose two practical constructions of outer codes. The first is a modification of the tree code. It utilizes the same code structure, and a key difference is a decoder capable of correcting up to t errors. The second is based on the Reed-Solomon codes and Guruswami-Sudan list decoding algorithm. The first scheme provides an energy efficiency very close to the random coding bound when the decoding complexity is unbounded. But for the practical parameters, the second scheme is better and improves the performance of a tree code-based scheme when the number of active users is less than 200."
                    },
                    {
                        "title": "Energy Efficiency of Unsourced Random Access over the Binary-Input Gaussian Channel",
                        "abstract": "We investigate the fundamental limits of the unsourced random access over the binary-input Gaussian channel. By fundamental limits, we mean the minimal energy per bit required to achieve the target per-user probability of error. The original method proposed by Y. Polyanskiy (2017) and based on Gallager's trick does not work well for binary signaling. We utilize Fano's method, which is based on the choice of the so-called ``good'' region. We apply this method for the cases of Gaussian and binary codebooks and obtain two achievability bounds. The first bound is very close to Polyanskiy's bound but does not lead to any improvement. At the same time, the numerical results show that the bound for the binary case practically coincides with the bound for the Gaussian codebook. Thus, we conclude that binary modulation does not lead to performance degradation, and energy-efficient schemes with binary modulation do exist."
                    },
                    {
                        "title": "On One Generalization of LRC Codes with Availability",
                        "abstract": "We investigate one possible generalization of locally recoverable codes (LRC) with all-symbol locality and availability when recovering sets can intersect in a small number of coordinates. This feature allows us to increase the achievable code rate and still meet load balancing requirements. In this paper we derive an upper bound for the rate of such codes and give explicit constructions of codes with such a property. These constructions utilize LRC codes developed by Wang et al."
                    },
                    {
                        "title": "On LDPC Code Based Massive Random-Access Scheme for the Gaussian Multiple Access Channel",
                        "abstract": "This paper deals with the problem of massive random access for Gaussian multiple access channel (MAC). We continue to investigate the coding scheme for Gaussian MAC proposed by A. Vem et al in 2017. The proposed scheme consists of four parts: (i) the data transmission is partitioned into time slots; (ii) the data, transmitted in each slot, is split into two parts, the first one set an interleaver of the low-density parity-check (LDPC) type code and is encoded by spreading sequence or codewords that are designed to be decoded by compressed sensing type decoding; (iii) the another part of transmitted data is encoded by LDPC type code and decoded using a joint message passing decoding algorithm designed for the T-user binary input Gaussian MAC; (iv) users repeat their codeword in multiple slots. In this paper we are concentrated on the third part of considered scheme. We generalized the PEXIT charts to optimize the protograph of LDPC code for Gaussian MAC. The simulation results, obtained at the end of the paper, were analyzed and compared with obtained theoretical bounds and thresholds. Obtained simulation results shows that proposed LDPC code constructions have better performance under joint decoding algorithm over Gaussian MAC than LDPC codes considered by A. Vem et al in 2017, that leads to the better performance of overall transmission system."
                    },
                    {
                        "title": "Soft-Output Deep Neural Network-Based Decoding",
                        "abstract": "Deep neural network (DNN)-based channel decoding is widely considered in the literature. The existing solutions are investigated for the case of hard output, i.e. when the decoder returns the estimated information word. At the same time, soft-output decoding is of critical importance for iterative receivers and decoders. In this paper, we focus on the soft-output DNN-based decoding problem. We start with the syndrome-based approach proposed by Bennatan et al. (2018) and modify it to provide soft output in the AWGN channel. The new decoder can be considered as an approximation of the MAP decoder with smaller computation complexity. We discuss various regularization functions for joint DNN-MAP training and compare the resulting distributions for [64, 45] BCH code. Finally, to demonstrate the soft-output quality we consider the turbo-product code with [64, 45] BCH codes as row and column codes. We show that the resulting DNN-based scheme is very close to the MAP-based performance and significantly outperforms the solution based on the Chase decoder. We come to the conclusion that the new method is prospective for the challenging problem of DNN-based decoding of long codes consisting of short component codes."
                    }
                ]
            }
        }
    },
    "2305.08960": {
        "paper_data": {
            "title": "One Forward is Enough for Neural Network Training via Likelihood Ratio Method",
            "url": "http://arxiv.org/abs/2305.08960v2",
            "arxiv_id": "2305.08960",
            "authors": [
                "Jinyang Jiang",
                "Zeliang Zhang",
                "Chenliang Xu",
                "Zhaofei Yu",
                "Yijie Peng"
            ],
            "abstract": "While backpropagation (BP) is the mainstream approach for gradient computation in neural network training, its heavy reliance on the chain rule of differentiation constrains the designing flexibility of network architecture and training pipelines. We avoid the recursive computation in BP and develop a unified likelihood ratio (ULR) method for gradient estimation with just one forward propagation. Not only can ULR be extended to train a wide variety of neural network architectures, but the computation flow in BP can also be rearranged by ULR for better device adaptation. Moreover, we propose several variance reduction techniques to further accelerate the training process. Our experiments offer numerical results across diverse aspects, including various neural network training scenarios, computation flow rearrangement, and fine-tuning of pre-trained models. All findings demonstrate that ULR effectively enhances the flexibility of neural network training by permitting localized module training without compromising the global objective and significantly boosts the network robustness.",
            "introduction": "ABSTRACT While backpropagation (BP) is the mainstream approach for gradient computation in neural network training, its heavy reliance on the chain rule of differentiation constrains the designing flexibility of network architecture and training pipelines. We avoid the recursive computation in BP and develop a unified likelihood ratio (ULR) method for gradient estimation with just one forward propagation. Not only can ULR be extended to train a wide variety of neural network architectures, but the computation flow in BP can also be rearranged by ULR for better device adaptation. Moreover, we propose several variance reduction techniques to further accelerate the training process. Ourexperiments and add a residual correction block, consisting of two linear layers with ReLU activation function, after the pooling layer. We use the Adam optimizer with the initialized learning rate 1\u00d710\u22124. For all models, we set the training epochs as 50and the batch size as 64. B.9 E XPERIMENTS FOR BPTT REARRANGEMENT We train the RNN and GRU on the Ag-News dataset to verify the performance of BPTT rearrange- ment via ULR. For all the models, we set the learning rate as 1\u00d710\u22123with Adam optimizer and train the models for 100epochs for sufficient convergence. While the average length of texts in the Ag-News dataset is around 45words, which corresponds to the average length of the gradient com- putation graph as 45steps, we split the original gradient computation graphs into subgraphs by ULR with depths of 5,10,15,20, and 25, respectively, to give a comprehensive study and comparison on the rearrangement performance. B.10 E XTENDED ABLATION STUDY Following the same setting in our main text, we provide a more extensive ablation study on the VGG-8. The details of the network structure have been introduced in Section B.2. Without specific 23nomination, we perturb the network without QMC in a hybrid manner for gradient estimation, where we inject the noise on weights for the first four layers and on logits for other layers. Fig. 11(a)-(c) demonstrate the impact of the perturbation on logits, weights, and the hybrid of weights and logits. It can be noted that compared with adding perturbation on logits and weights, the hybrid manner can achieve a much better gradient estimation accuracy both in the first and last several layers. Fig. 11(d)-(f) present the impact of the initialization for noise magnitude \u03c3. Following the design of the aforementioned hybrid noise injection mode, we perform ULR with \u03c3set to 1\u00d710\u22121,1\u00d710\u22122, and1\u00d710\u22123, respectively. The optimal choice of \u03c3varies for different noise injection modes. We can see that 1\u00d710\u22122is more suitable for initializing \u03c3for the first several convolutional layers to achieve an acceptable gradient estimation accuracy. In contrast, the last several layers, especially the fully connected layers, are robust to the initialization of \u03c3, which maintains high gradient estimation accuracy for all selected values. Fig. 11(g)-(i) show the impact of the variance-reduction techniques. From theresults in the figures, all the proposedmethods, including layer-wise perturbation, antithetic variable, and the use of QMC, play an important role on improving the gradient estimation accuracy. In a word, theAppendix B.10. Impact of the perturbation on weights and logits : We present the gradient estimation performance when perturbing the logits only in Fig. 7(a), the weights only in Fig. 7(b), weights and logits in a hybrid manner in Fig. 7(c). In comparison between Fig. 7(a) and Fig. 7(b), it is observed that perturbing the",
            "references": [
                {
                    "title": "The Forward-Forward Algorithm: Some Preliminary Investigations",
                    "abstract": "The aim of this paper is to introduce a new learning procedure for neural networks and to demonstrate that it works well enough on a few small problems to be worth further investigation. The Forward-Forward algorithm replaces the forward and backward passes of backpropagation by two forward passes, one with positive (i.e. real) data and the other with negative data which could be generated by the network itself. Each layer has its own objective function which is simply to have high goodness for positive data and low goodness for negative data. The sum of the squared activities in a layer can be used as the goodness but there are many other possibilities, including minus the sum of the squared activities. If the positive and negative passes could be separated in time, the negative passes could be done offline, which would make the learning much simpler in the positive pass and allow video to be pipelined through the network without ever storing activities or stopping to propagate derivatives."
                },
                {
                    "title": "TACR-Net: Editing on Deep Video and Voice Portraits",
                    "abstract": "Utilizing an arbitrary speech clip to edit the mouth of the portrait in the target video is a novel yet challenging task. Despite impressive results have been achieved, there are still three limitations in the existing methods: 1) since the acoustic features are not completely decoupled from person identity, there is no global speech to facial features (i.e., landmarks, expression blendshape) mapping method. 2) the audio-driven talking face sequences generated by simple cascade structure usually lack of temporal consistency and spatial correlation, which leads to defects in the consistency of changes in details. 3) the operation of forgery is always at the video level, without considering the forgery of the voice, especially the synchronization of the converted voice and the mouth. To address these distortion problems, we propose a novel deep learning framework, named Temporal-Refinement Autoregressive-Cascade Rendering Network (TACR-Net) for audio-driven dynamic talking face editing. The proposed TACR-Net encodes facial expression blendshape based on the given acoustic features without separately training for special video. Then TACR-Net also involves a novel autoregressive cascade structure generator for video re-rendering. Finally, we transform the in-the-wild speech to the target portrait and obtain a photo-realistic and audio-realistic video."
                },
                {
                    "title": "A New Likelihood Ratio Method for Training Artificial Neural Networks",
                    "abstract": "We investigate a new approach to compute the gradients of artificial neural networks (ANNs), based on the so-called push-out likelihood ratio method. Unlike the widely used backpropagation (BP) method that requires continuity of the loss function and the activation function, our approach bypasses this requirement by injecting artificial noises into the signals passed along the neurons. We show how this approach has a similar computational complexity as BP, and moreover is more advantageous in terms of removing the backward recursion and eliciting transparent formulas. We also formalize the connection between BP, a pivotal technique for training ANNs, and infinitesimal perturbation analysis, a classic path-wise derivative estimation approach, so that both our new proposed methods and BP can be better understood in the context of stochastic gradient estimation. Our approach allows efficient training for ANNs with more flexibility on the loss and activation functions, and shows empirical improvements on the robustness of ANNs under adversarial attacks and corruptions of natural noises. Summary of Contribution: Stochastic gradient estimation has been studied actively in simulation for decades and becomes more important in the era of machine learning and artificial intelligence. The stochastic gradient descent is a standard technique for training the artificial neural networks (ANNs), a pivotal problem in deep learning. The most popular stochastic gradient estimation technique is the backpropagation method. We find that the backpropagation method lies in the family of infinitesimal perturbation analysis, a path-wise gradient estimation technique in simulation. Moreover, we develop a new likelihood ratio-based method, another popular family of gradient estimation technique in simulation, for training more general ANNs, and demonstrate that the new training method can improve the robustness of the ANN."
                },
                {
                    "title": "Noise Optimization for Artificial Neural Networks",
                    "abstract": "Adding noises to artificial neural network(ANN) has been shown to be able to improve robustness in previous work. In this work, we propose a new technique to compute the pathwise stochastic gradient estimate with respect to the standard deviation of the Gaussian noise added to each neuron of the ANN. By our proposed technique, the gradient estimate with respect to noise levels is a byproduct of the backpropagation algorithm for estimating gradient with respect to synaptic weights in ANN. Thus, the noise level for each neuron can be optimized simultaneously in the processing of training the synaptic weights at nearly no extra computational cost. In numerical experiments, our proposed method can achieve significant performance improvement on robustness of several popular ANN structures under both black box and white box attacks tested in various computer vision datasets."
                },
                {
                    "title": "Deep Residual Correction Network for Partial Domain Adaptation",
                    "abstract": "Deep domain adaptation methods have achieved appealing performance by learning transferable representations from a well-labeled source domain to a different but related unlabeled target domain. Most existing works assume source and target data share the identical label space, which is often difficult to be satisfied in many real-world applications. With the emergence of big data, there is a more practical scenario called partial domain adaptation, where we are always accessible to a more large-scale source domain while working on a relative small-scale target domain. In this case, the conventional domain adaptation assumption should be relaxed, and the target label space tends to be a subset of the source label space. Intuitively, reinforcing the positive effects of the most relevant source subclasses and reducing the negative impacts of irrelevant source subclasses are of vital importance to address partial domain adaptation challenge. This paper proposes an efficiently-implemented Deep Residual Correction Network (DRCN) by plugging one residual block into the source network along with the task-specific feature layer, which effectively enhances the adaptation from source to target and explicitly weakens the influence from the irrelevant source classes. Specifically, the plugged residual block, which consists of several fully-connected layers, could deepen basic network and boost its feature representation capability correspondingly. Moreover, we design a weighted class-wise domain alignment loss to couple two domains by matching the feature distributions of shared classes between source and target. Comprehensive experiments on partial, traditional and fine-grained cross-domain visual recognition demonstrate that DRCN is superior to the competitive deep domain adaptation approaches."
                },
                {
                    "title": "The HSIC Bottleneck: Deep Learning without Back-Propagation",
                    "abstract": "We introduce the HSIC (Hilbert-Schmidt independence criterion) bottleneck for training deep neural networks. The HSIC bottleneck is an alternative to the conventional cross-entropy loss and backpropagation that has a number of distinct advantages. It mitigates exploding and vanishing gradients, resulting in the ability to learn very deep networks without skip connections. There is no requirement for symmetric feedback or update locking. We find that the HSIC bottleneck provides performance on MNIST/FashionMNIST/CIFAR10 classification comparable to backpropagation with a cross-entropy target, even when the system is not encouraged to make the output resemble the classification labels. Appending a single layer trained with SGD (without backpropagation) to reformat the information further improves performance."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "Graph Attention Networks",
                    "abstract": "We present graph attention networks (GATs), novel neural network architectures that operate on graph-structured data, leveraging masked self-attentional layers to address the shortcomings of prior methods based on graph convolutions or their approximations. By stacking layers in which nodes are able to attend over their neighborhoods' features, we enable (implicitly) specifying different weights to different nodes in a neighborhood, without requiring any kind of costly matrix operation (such as inversion) or depending on knowing the graph structure upfront. In this way, we address several key challenges of spectral-based graph neural networks simultaneously, and make our model readily applicable to inductive as well as transductive problems. Our GAT models have achieved or matched state-of-the-art results across four established transductive and inductive graph benchmarks: the Cora, Citeseer and Pubmed citation network datasets, as well as a protein-protein interaction dataset (wherein test graphs remain unseen during training)."
                },
                {
                    "title": "Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms",
                    "abstract": "We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at this https URL"
                },
                {
                    "title": "The Reversible Residual Network: Backpropagation Without Storing Activations",
                    "abstract": "Deep residual networks (ResNets) have significantly pushed forward the state-of-the-art on image classification, increasing in performance as networks grow both deeper and wider. However, memory consumption becomes a bottleneck, as one needs to store the activations in order to calculate gradients using backpropagation. We present the Reversible Residual Network (RevNet), a variant of ResNets where each layer's activations can be reconstructed exactly from the next layer's. Therefore, the activations for most layers need not be stored in memory during backpropagation. We demonstrate the effectiveness of RevNets on CIFAR-10, CIFAR-100, and ImageNet, establishing nearly identical classification accuracy to equally-sized ResNets, even though the activation storage requirements are independent of depth."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Evolution Strategies as a Scalable Alternative to Reinforcement Learning",
                    "abstract": "We explore the use of Evolution Strategies (ES), a class of black box optimization algorithms, as an alternative to popular MDP-based RL techniques such as Q-learning and Policy Gradients. Experiments on MuJoCo and Atari show that ES is a viable solution strategy that scales extremely well with the number of CPUs available: By using a novel communication strategy based on common random numbers, our ES implementation only needs to communicate scalars, making it possible to scale to over a thousand parallel workers. This allows us to solve 3D humanoid walking in 10 minutes and obtain competitive results on most Atari games after one hour of training. In addition, we highlight several advantages of ES as a black box optimization technique: it is invariant to action frequency and delayed rewards, tolerant of extremely long horizons, and does not need temporal discounting or value function approximation."
                },
                {
                    "title": "Gate-variants of Gated Recurrent Unit (GRU) neural networks",
                    "abstract": "The paper evaluates three variants of the Gated Recurrent Unit (GRU) in recurrent neural networks (RNNs) by retaining the structure and systematically reducing parameters in the update and reset gates. We evaluate the three variant GRU models on MNIST and IMDB datasets and show that these GRU-RNN variant models perform as well as the original GRU RNN model while reducing the computational expense. In this comparative study, we simply refer to the three variants as, respectively, GRU1, GRU2, and GRU3 RNNs."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Direct Feedback Alignment Provides Learning in Deep Neural Networks",
                    "abstract": "Artificial neural networks are most commonly trained with the back-propagation algorithm, where the gradient for learning is provided by back-propagating the error, layer by layer, from the output layer to the hidden layers. A recently discovered method called feedback-alignment shows that the weights used for propagating the error backward don't have to be symmetric with the weights used for propagation the activation forward. In fact, random feedback weights work evenly well, because the network learns how to make the feedback useful. In this work, the feedback alignment principle is used for training hidden layers more independently from the rest of the network, and from a zero initial condition. The error is propagated through fixed random feedback connections directly from the output layer to each hidden layer. This simple method is able to achieve zero training error even in convolutional networks and very deep networks, completely without error back-propagation. The method is a step towards biologically plausible machine learning because the error signal is almost local, and no symmetric or reciprocal weights are required. Experiments show that the test performance on MNIST and CIFAR is almost as good as those obtained with back-propagation for fully connected networks. If combined with dropout, the method achieves 1.45% error on the permutation invariant MNIST task."
                },
                {
                    "title": "Equilibrium Propagation: Bridging the Gap between Energy-Based Models and Backpropagation",
                    "abstract": "We introduce Equilibrium Propagation, a learning framework for energy-based models. It involves only one kind of neural computation, performed in both the first phase (when the prediction is made) and the second phase of training (after the target or prediction error is revealed). Although this algorithm computes the gradient of an objective function just like Backpropagation, it does not need a special computation or circuit for the second phase, where errors are implicitly propagated. Equilibrium Propagation shares similarities with Contrastive Hebbian Learning and Contrastive Divergence while solving the theoretical issues of both algorithms: our algorithm computes the gradient of a well-defined objective function. Because the objective function is defined in terms of local perturbations, the second phase of Equilibrium Propagation corresponds to only nudging the prediction (fixed point or stationary distribution) toward a configuration that reduces prediction error. In the case of a recurrent multi-layer supervised network, the output units are slightly nudged toward their target in the second phase, and the perturbation introduced at the output layer propagates backward in the hidden layers. We show that the signal \u201cback-propagated\u201d during this second phase corresponds to the propagation of error derivatives and encodes the gradient of the objective function, when the synaptic update corresponds to a standard form of spike-timing dependent plasticity. This work makes it more plausible that a mechanism similar to Backpropagation could be implemented by brains, since leaky integrator neural computation performs both inference and error back-propagation in our model. The only local difference between the two phases is whether synaptic changes are allowed or not. We also show experimentally that multi-layer recurrently connected networks with 1, 2, and 3 hidden layers can be trained by Equilibrium Propagation on the permutation-invariant MNIST task."
                },
                {
                    "title": "A Sensitivity Analysis of (and Practitioners\u2019 Guide to) Convolutional Neural Networks for Sentence Classification",
                    "abstract": "Convolutional Neural Networks (CNNs) have recently achieved remarkably strong performance on the practically important task of sentence classification (Kim, 2014; Kalchbrenner et al., 2014; Johnson and Zhang, 2014; Zhang et al., 2016). However, these models require practitioners to specify an exact model architecture and set accompanying hyperparameters, including the filter region size, regularization parameters, and so on. It is currently unknown how sensitive model performance is to changes in these configurations for the task of sentence classification. We thus conduct a sensitivity analysis of one-layer CNNs to explore the effect of architecture components on model performance; our aim is to distinguish between important and comparatively inconsequential design decisions for sentence classification. We focus on one-layer CNNs (to the exclusion of more complex models) due to their comparative simplicity and strong empirical performance, which makes it a modern standard baseline method akin to Support Vector Machine (SVMs) and logistic regression. We derive practical advice from our extensive empirical results for those interested in getting the most out of CNNs for sentence classification in real world settings."
                },
                {
                    "title": "Collective Classification in Network Data",
                    "abstract": "Many real-world applications produce networked data such as the world-wide web (hypertext documents connected via hyperlinks), social networks (for example, people connected by friendship links), communication networks (computers connected via communication links) and biological networks (for example, protein interaction networks). A recent focus in machine learning research has been to extend traditional machine learning classification techniques to classify nodes in such networks. In this article, we provide a brief introduction to this area of research and how it has progressed during the past decade. We introduce four of the most widely used inference algorithms for classifying networked data and empirically compare them on both synthetic and real-world data."
                },
                {
                    "title": "Natural Evolution Strategies",
                    "abstract": "This paper presents natural evolution strategies (NES), a novel algorithm for performing real-valued dasiablack boxpsila function optimization: optimizing an unknown objective function where algorithm-selected function measurements constitute the only information accessible to the method. Natural evolution strategies search the fitness landscape using a multivariate normal distribution with a self-adapting mutation matrix to generate correlated mutations in promising regions. NES shares this property with covariance matrix adaption (CMA), an evolution strategy (ES) which has been shown to perform well on a variety of high-precision optimization tasks. The natural evolution strategies algorithm, however, is simpler, less ad-hoc and more principled. Self-adaptation of the mutation matrix is derived using a Monte Carlo estimate of the natural gradient towards better expected fitness. By following the natural gradient instead of the dasiavanillapsila gradient, we can ensure efficient update steps while preventing early convergence due to overly greedy updates, resulting in reduced sensitivity to local suboptima. We show NES has competitive performance with CMA on unimodal tasks, while outperforming it on several multimodal tasks that are rich in deceptive local optima."
                },
                {
                    "title": "Spontaneous Dynamics of Asymmetric Random Recurrent Spiking Neural Networks",
                    "abstract": "In this letter, we study the effect of a unique initial stimulation on random recurrent networks of leaky integrate-and-fire neurons. Indeed, given a stochastic connectivity, this so-called spontaneous mode exhibits various nontrivial dynamics. This study is based on a mathematical formalism that allows us to examine the variability of the afterward dynamics according to the parameters of the weight distribution. Under the independence hypothesis (e.g., in the case of very large networks), we are able to compute the average number of neurons that fire at a given timethe spiking activity. In accordance with numerical simulations, we prove that this spiking activity reaches a steady state. We characterize this steady state and explore the transients."
                },
                {
                    "title": "Long Short-Term Memory",
                    "abstract": "Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, LSTM can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units. Multiplicative gate units learn to open and close access to the constant error flow. LSTM is local in space and time; its computational complexity per time step and weight is O. 1. Our experiments with artificial data involve local, distributed, real-valued, and noisy pattern representations. In comparisons with real-time recurrent learning, back propagation through time, recurrent cascade correlation, Elman nets, and neural sequence chunking, LSTM leads to many more successful runs, and learns much faster. LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms."
                },
                {
                    "title": "Quasi-Monte Carlo integration",
                    "abstract": "The standard Monte Carlo approach to evaluating multidimensional integrals using (pseudo)-random integration nodes is frequently used when quadrature methods are too difficult or expensive to implement. As an alternative to the random methods, it has been suggested that lower error and improved convergence may be obtained by replacing the pseudo-random sequences with more uniformly distributed sequences known as quasi-random. In this paper quasi-random (Halton, Sobol`, and Faure) and pseudo-random sequences are compared in computational experiments designed to determine the effects on convergence of certain properties of the integrand, including variance, variation, smoothness, and dimension. The results show that variation, which plays an important role in the theoretical upper bound given by the Koksma-Hlawka inequality, does not affect convergence, while variance, the determining factor in random Monte Carlo, is shown to provide a rough upper bound, but does not accurately predict performance. In general, quasi-Monte Carlo methods are superior to random Monte Carlo, but the advantage may be slight, particularly in high dimensions or for integrands that are not smooth. For discontinuous integrands, we derive a bound which shows that the exponent for algebraic decay of the integration error from quasi-Monte Carlo is only slightly larger than {1/2} in high dimensions.more\u00a0\u00bb 21 refs., 6 figs., 5 tabs.\u00ab\u00a0less"
                },
                {
                    "title": "Multivariate stochastic approximation using a simultaneous perturbation gradient approximation",
                    "abstract": "The problem of finding a root of the multivariate gradient equation that arises in function minimization is considered. When only noisy measurements of the function are available, a stochastic approximation (SA) algorithm for the general Kiefer-Wolfowitz type is appropriate for estimating the root. The paper presents an SA algorithm that is based on a simultaneous perturbation gradient approximation instead of the standard finite-difference approximation of Keifer-Wolfowitz type procedures. Theory and numerical experience indicate that the algorithm can be significantly more efficient than the standard algorithms in large-dimensional problems. >"
                },
                {
                    "title": "Training Spiking Neural Networks with Event-driven Backpropagation",
                    "abstract": "Spiking Neural networks (SNNs) represent and transmit information by spatiotemporal spike patterns, which bring two major advantages: biological plausibility and suitability for ultralow-power neuromorphic implementation. Despite this, the binary firing characteristic makes training SNNs more challenging. To learn the parameters of deep SNNs in an event-driven fashion as in inference of SNNs, back-propagation with respect to spike timing is proposed. Although this event-driven learning has the advantages of lower computational cost and memory occupation, the accuracy is far below the recurrent neural network-like learning approaches. In this paper, we first analyze the commonly used temporal backpropagation training approach and prove that the sum of gradients remains unchanged between fully-connected and convolutional layers. Secondly, we show that the max pooling layer meets the above invariance rule, while the average pooling layer does not, which will suffer the gradient vanishing problem but can be revised to meet the requirement. Thirdly, we point out the reverse gradient problem for time-based gradients and propose a backward kernel that can solve this problem and keep the property of the invariable sum of gradients. The experimental results show that the proposed approach achieves state-of-the-art performance on CIFAR10 among time-based training methods. Also, this is the first time that the time-based backpropagation approach successfully trains SNN on the CIFAR100 dataset. Our code is available at https://github.com/zhuyaoyu/SNN-event-driven-learning."
                },
                {
                    "title": "Quasi-Monte Carlo methods",
                    "abstract": "where f : [0, 1] \u2192 R and d \u2208 N is large (usually d \u2265 4). As is shown in [3] classical methods of numerical integration are ill-suited for such problems. A number of related methods that alleviates this problem are known as Monte Carlo integration methods. Such methods can be shown to have a probabilistic error bound of order O ( n\u22121/2 ) (where n is the number of function evaluations). In this term paper we will introduce other methods of approximating the before-mentioned integral, namely quasi-Monte Carlo methods, which are super cial similar to Monte Carlo methods but have certain advantages, especially for moderately sized d (e.g. d < 20). Monte Carlo methods use a sequence of (indepenent) random numbers to determine the points at which f is evaluated. Random numbers, however, have, e.g., the disadvantage that they exhibit clustering behavior. QuasiMonte Carlo methods, on the other hand, use a fully deterministic sequences that tries to evenly covers the unit cube. This rather vague notion of evenly covers will be made more precise in the following section by the so called discrepancy of a sequence. The development in this term paper mainly follows [1] and [6]."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "The mnist database of handwritten digits",
                    "abstract": "Disclosed is an improved articulated bar flail having shearing edges for efficiently shredding materials. An improved shredder cylinder is disclosed with a plurality of these flails circumferentially spaced and pivotally attached to the periphery of a rotatable shaft. Also disclosed is an improved shredder apparatus which has a pair of these shredder cylinders mounted to rotate about spaced parallel axes which cooperates with a conveyer apparatus which has a pair of inclined converging conveyer belts with one of the belts mounted to move with respect to the other belt to allow the transport of articles of various sizes therethrough."
                },
                {
                    "title": "Gradient-based learning applied to document recognition",
                    "abstract": "Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.NE",
                "math.OC"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a more flexible and efficient method for gradient estimation in neural network training that avoids the recursive computation inherent in backpropagation?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem has significant implications for the research community as it could lead to the development of more adaptable neural network architectures and training pipelines. By improving gradient estimation methods, researchers can explore novel architectures without being constrained by the limitations of backpropagation. This advancement could accelerate the training process, enhance model performance, and open up new practical applications in various fields, such as natural language processing and computer vision.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities of accurately estimating gradients without the recursive structure of backpropagation. Naive approaches may fail due to the potential for high variance in gradient estimates, which can lead to unstable training and poor convergence. Additionally, technical obstacles include the need for effective variance reduction techniques and the ability to rearrange computation flows in a way that maintains accuracy while improving efficiency.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on refining backpropagation techniques, often overlooking alternative methods for gradient estimation. Limitations in computational resources and theoretical understanding of non-recursive methods have also posed barriers. Our approach differs by introducing the unified likelihood ratio (ULR) method, which allows for a single forward propagation to estimate gradients, thus providing a more flexible and efficient alternative to traditional backpropagation.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the unified likelihood ratio (ULR) method for gradient estimation, which will be tested on various neural network architectures using datasets such as Ag-News. We will employ metrics such as gradient estimation accuracy and training speed to evaluate performance. Expected outcomes include improved gradient estimation accuracy through the use of variance reduction techniques and enhanced training efficiency, ultimately demonstrating the viability of ULR as a robust alternative to backpropagation."
            }
        },
        "author_data": {
            "fb337d4f-4b6a-454f-8c85-1a05c277faec": {
                "pk": "fb337d4f-4b6a-454f-8c85-1a05c277faec",
                "name": "Jinyang Jiang",
                "collaborators": [
                    "Yijie Peng",
                    "Zeliang Zhang",
                    "Tao Ren",
                    "Qinghao Wang",
                    "Chenliang Xu",
                    "Mingqian Feng",
                    "Jiaqiao Hu",
                    "Ruihan Zhou",
                    "Jiafeng Liang",
                    "Xiaotian Liu"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Neural Networks",
                    "Deep Learning",
                    "Financial Modeling"
                ],
                "publications": [
                    {
                        "title": "RiskMiner: Discovering Formulaic Alphas via Risk Seeking Monte Carlo Tree Search",
                        "abstract": "The formulaic alphas are mathematical formulas that transform raw stock data into indicated signals. In the industry, a collection of formulaic alphas is combined to enhance modeling accuracy. Existing alpha mining only employs the neural network agent, unable to utilize the structural information of the solution space. Moreover, they didn't consider the correlation between alphas in the collection, which limits the synergistic performance. To address these problems, we propose a novel alpha mining framework, which formulates the alpha mining problems as a reward-dense Markov Decision Process (MDP) and solves the MDP by the risk-seeking Monte Carlo Tree Search (MCTS). The MCTS-based agent fully exploits the structural information of discrete solution space and the risk-seeking policy explicitly optimizes the best-case performance rather than average outcomes. Comprehensive experiments are conducted to demonstrate the efficiency of our framework. Our method outperforms all state-of-the-art benchmarks on two real-world stock sets under various metrics. Backtest experiments show that our alphas achieve the most profitable results under a realistic trading setting."
                    },
                    {
                        "title": "Deep Reinforcement Learning for Solving Management Problems: Towards A Large Management Mode",
                        "abstract": "We introduce a deep reinforcement learning (DRL) approach for solving management problems including inventory management, dynamic pricing, and recommendation. This DRL approach has the potential to lead to a large management model based on certain transformer neural network structures, resulting in an artificial general intelligence paradigm for various management tasks. Traditional methods have limitations for solving complex real-world problems, and we demonstrate how DRL can surpass existing heuristic approaches for solving management tasks. We aim to solve the problems in a unified framework, considering the interconnections between different tasks. Central to our methodology is the development of a foundational decision model coordinating decisions across the different domains through generative decision-making. Our experimental results affirm the effectiveness of our DRL-based framework in complex and dynamic business environments. This work opens new pathways for the application of DRL in management problems, highlighting its potential to revolutionize traditional business management."
                    },
                    {
                        "title": "Approximated Likelihood Ratio: A Forward-Only and Parallel Framework for Boosting Neural Network Training",
                        "abstract": "Efficient and biologically plausible alternatives to backpropagation in neural network training remain a challenge due to issues such as high computational complexity and additional assumptions about neural networks, which limit scalability to deeper networks. The likelihood ratio method offers a promising gradient estimation strategy but is constrained by significant memory consumption, especially when deploying multiple copies of data to reduce estimation variance. In this paper, we introduce an approximation technique for the likelihood ratio (LR) method to alleviate computational and memory demands in gradient estimation. By exploiting the natural parallelism during the backward pass using LR, we further provide a high-performance training strategy, which pipelines both the forward and backward pass, to make it more suitable for the computation on specialized hardware. Extensive experiments demonstrate the effectiveness of the approximation technique in neural network training. This work underscores the potential of the likelihood ratio method in achieving high-performance neural network training, suggesting avenues for further exploration."
                    },
                    {
                        "title": "Forward Learning for Gradient-based Black-box Saliency Map Generation",
                        "abstract": "Gradient-based saliency maps are widely used to explain deep neural network decisions. However, as models become deeper and more black-box, such as in closed-source APIs like ChatGPT, computing gradients become challenging, hindering conventional explanation methods. In this work, we introduce a novel unified framework for estimating gradients in black-box settings and generating saliency maps to interpret model decisions. We employ the likelihood ratio method to estimate output-to-input gradients and utilize them for saliency map generation. Additionally, we propose blockwise computation techniques to enhance estimation accuracy. Extensive experiments in black-box settings validate the effectiveness of our method, demonstrating accurate gradient estimation and explainability of generated saliency maps. Furthermore, we showcase the scalability of our approach by applying it to explain GPT-Vision, revealing the continued relevance of gradient-based explanation methods in the era of large, closed-source, and black-box models."
                    },
                    {
                        "title": "FLOPS: Forward Learning with OPtimal Sampling",
                        "abstract": "Given the limitations of backpropagation, perturbation-based gradient computation methods have recently gained focus for learning with only forward passes, also referred to as queries. Conventional forward learning consumes enormous queries on each data point for accurate gradient estimation through Monte Carlo sampling, which hinders the scalability of those algorithms. However, not all data points deserve equal queries for gradient estimation. In this paper, we study the problem of improving the forward learning efficiency from a novel perspective: how to reduce the gradient estimation variance with minimum cost? For this, we propose to allocate the optimal number of queries over each data in one batch during training to achieve a good balance between estimation accuracy and computational efficiency. Specifically, with a simplified proxy objective and a reparameterization technique, we derive a novel plug-and-play query allocator with minimal parameters. Theoretical results are carried out to verify its optimality. We conduct extensive experiments for fine-tuning Vision Transformers on various datasets and further deploy the allocator to two black-box applications: prompt tuning and multimodal alignment for foundation models. All findings demonstrate that our proposed allocator significantly enhances the scalability of forward-learning algorithms, paving the way for real-world applications."
                    },
                    {
                        "title": "A Novel Noise Injection-based Training Scheme for Better Model Robustness",
                        "abstract": "Noise injection-based method has been shown to be able to improve the robustness of artificial neural networks in previous work. In this work, we propose a novel noise injection-based training scheme for better model robustness. Specifically, we first develop a likelihood ratio method to estimate the gradient with respect to both synaptic weights and noise levels for stochastic gradient descent training. Then, we design an approximation for the vanilla noise injection-based training method to reduce memory and improve computational efficiency. Next, we apply our proposed scheme to spiking neural networks and evaluate the performance of classification accuracy and robustness on MNIST and Fashion-MNIST datasets. Experiment results show that our proposed method achieves a much better performance on adversarial robustness and slightly better performance on original accuracy, compared with the conventional gradient-based training method."
                    },
                    {
                        "title": "Quantile-Based Deep Reinforcement Learning using Two-Timescale Policy Gradient Algorithms",
                        "abstract": "Classical reinforcement learning (RL) aims to optimize the expected cumulative reward. In this work, we consider the RL setting where the goal is to optimize the quantile of the cumulative reward. We parameterize the policy controlling actions by neural networks, and propose a novel policy gradient algorithm called Quantile-Based Policy Optimization (QPO) and its variant Quantile-Based Proximal Policy Optimization (QPPO) for solving deep RL problems with quantile objectives. QPO uses two coupled iterations running at different timescales for simultaneously updating quantiles and policy parameters, whereas QPPO is an off-policy version of QPO that allows multiple updates of parameters during one simulation episode, leading to improved algorithm efficiency. Our numerical results indicate that the proposed algorithms outperform the existing baseline algorithms under the quantile criterion."
                    },
                    {
                        "title": "Quantile-Based Policy Optimization for Reinforcement Learning",
                        "abstract": "Classical reinforcement learning (RL) aims to optimize the expected cumulative rewards. In this work, we consider the RL setting where the goal is to optimize the quantile of the cumulative rewards. We parameterize the policy controlling actions by neural networks and propose a novel policy gradient algorithm called Quantile-Based Policy Optimization (QPO) and its variant Quantile-Based Proximal Policy Optimization (QPPO) to solve deep RL problems with quantile objectives. QPO uses two coupled iterations running at different time scales for simultaneously estimating quantiles and policy parameters. Our numerical results demonstrate that the proposed algorithms outperform the existing baseline algorithms under the quantile criterion."
                    },
                    {
                        "title": "Noise Optimization in Artificial Neural Networks",
                        "abstract": "Artificial neural network (ANN) has been widely used in automation. However, the vulnerability of ANN under certain attacks poses security threat for critical automation systems. Adding noises to artificial neural network has been shown to be able to improve robustness in previous work. In this work, we propose a new technique to compute the pathwise stochastic gradient estimate with respect to the standard deviation of the Gaussian noise added to each neuron of the ANN. By our proposed technique, the gradient estimation with respect to noise levels is a byproduct of the back propagation algorithm for estimating gradient with respect to synaptic weights in ANN. Thus, the noise level for each neuron can be optimized simultaneously in the processing of training the synaptic weights at nearly no extra computational cost. In numerical experiments, our proposed method can achieve significant performance improvement on robustness of several popular ANN structures under both black box and white box attacks."
                    }
                ]
            },
            "09215eb6-7714-48c0-9b72-dce4045c8275": {
                "pk": "09215eb6-7714-48c0-9b72-dce4045c8275",
                "name": "Zeliang Zhang",
                "collaborators": [
                    "Chenliang Xu",
                    "Susan Liang",
                    "Mingqian Feng",
                    "Zhuo Liu",
                    "Rongyi Zhu",
                    "Xiaosen Wang",
                    "Wei Yao",
                    "Jinyang Jiang",
                    "Yijie Peng",
                    "Yunlong Tang"
                ],
                "domain": [
                    "Adversarial Machine Learning",
                    "Natural Language Processing",
                    "Tensor Decomposition",
                    "Model Efficiency"
                ],
                "publications": [
                    {
                        "title": "Diversifying the Expert Knowledge for Task-Agnostic Pruning in Sparse Mixture-of-Experts",
                        "abstract": "By increasing model parameters but activating them sparsely when performing a task, the use of Mixture-of-Experts (MoE) architecture significantly improves the performance of Large Language Models (LLMs) without increasing the inference cost. However, the memory consumption due to the growing number of experts presents a challenge to the deployment of these models in many real world settings. Our empirical study reveals that some experts encode redundant knowledge during pre-training. We thus propose a method of grouping and pruning similar experts to improve the model's parameter efficiency. We validate the effectiveness of our method by pruning three state-of-the-art MoE architectures, including Mixtral, Deepseek-MoE, and Qwen. The evaluation shows that our method outperforms other model pruning methods on a range of natural language tasks. We will release our code to facilitate future research."
                    },
                    {
                        "title": "Can CLIP Count Stars? An Empirical Study on Quantity Bias in CLIP",
                        "abstract": "CLIP has demonstrated great versatility in adapting to various downstream tasks, such as image editing and generation, visual question answering, and video understanding. However, CLIP-based applications often suffer from misunderstandings regarding user intent, leading to discrepancies between the required number of objects and the actual outputs in image generation tasks. In this work, we empirically investigate the quantity bias in CLIP. By carefully designing different experimental settings and datasets, we comprehensively evaluate CLIP's understanding of quantity from text, image, and cross-modal perspectives. Our experimental results reveal a quantity bias in CLIP embeddings, impacting the reliability of downstream tasks."
                    },
                    {
                        "title": "Random Smooth-based Certified Defense against Text Adversarial Attack",
                        "abstract": "Certified defense methods have identified their effectiveness against textual adversarial examples, which train models on the worst-case text generated by substituting words in original texts with synonyms. However, due to the discrete word embedding representations, the large search space hinders the robust training efficiency, resulting in significant time consumption. To overcome this challenge, motivated by the observation that synonym embedding has a small distance, we propose to treat the word substitution as a continuous perturbation on the word embedding representation. The proposed method Text-RS applies random smooth techniques to approximate the word substitution operation, offering a computationally efficient solution that outperforms conventional discrete methods and improves the robustness in training. The evaluation results demonstrate its effectiveness in defending against multiple textual adversarial attacks."
                    },
                    {
                        "title": "Learning to Transform Dynamically for Better Adversarial Transferability",
                        "abstract": "Adversarial examples, crafted by adding perturbations imperceptible to humans, can deceive neural networks. Re-cent studies identify the adversarial transferability across various models, i.e., the cross-model attack ability of adversarial samples. To enhance such adversarial transferability, existing input transformation-based methods diversify input data with transformation augmentation. However, their effectiveness is limited by the finite number of available transformations. In our study, we introduce a novel approach named Learning to Transform (L2T). L2T increases the diversity of transformed images by selecting the optimal combination of operations from a pool of candidates, consequently im-proving adversarial transferability. We conceptualize the selection of optimal transformation combinations as a tra-jectory optimization problem and employ a reinforcement learning strategy to effectively solve the problem. Comprehensive experiments on the ImageNet dataset, as well as practical tests with Google Vision and GPT-4V, reveal that L2T surpasses current methodologies in enhancing adversarial transferability, thereby confirming its effectiveness and practical significance. The code is available at https://github.com/ZhangAIPI/TransferAttack."
                    },
                    {
                        "title": "Approximated Likelihood Ratio: A Forward-Only and Parallel Framework for Boosting Neural Network Training",
                        "abstract": "Efficient and biologically plausible alternatives to backpropagation in neural network training remain a challenge due to issues such as high computational complexity and additional assumptions about neural networks, which limit scalability to deeper networks. The likelihood ratio method offers a promising gradient estimation strategy but is constrained by significant memory consumption, especially when deploying multiple copies of data to reduce estimation variance. In this paper, we introduce an approximation technique for the likelihood ratio (LR) method to alleviate computational and memory demands in gradient estimation. By exploiting the natural parallelism during the backward pass using LR, we further provide a high-performance training strategy, which pipelines both the forward and backward pass, to make it more suitable for the computation on specialized hardware. Extensive experiments demonstrate the effectiveness of the approximation technique in neural network training. This work underscores the potential of the likelihood ratio method in achieving high-performance neural network training, suggesting avenues for further exploration."
                    },
                    {
                        "title": "Do More Details Always Introduce More Hallucinations in LVLM-based Image Captioning?",
                        "abstract": "Large Vision-Language Models (LVLMs) excel in integrating visual and linguistic contexts to produce detailed content, facilitating applications such as image captioning. However, using LVLMs to generate descriptions often faces the challenge of object hallucination (OH), where the output text misrepresents actual objects in the input image. While previous studies attribute the occurrence of OH to the inclusion of more details, our study finds technical flaws in existing metrics, leading to unreliable evaluations of models and conclusions about OH. This has sparked a debate on the question: Do more details always introduce more hallucinations in LVLM-based image captioning? In this paper, we address this debate by proposing a novel decoding strategy, Differentiated Beam Decoding (DBD), along with a reliable new set of evaluation metrics: CLIP-Precision, CLIP-Recall, and CLIP-F1. DBD decodes the wealth of information hidden in visual input into distinct language representations called unit facts in parallel. This decoding is achieved via a well-designed differential score that guides the parallel search and candidate screening. The selected unit facts are then aggregated to generate the final caption. Our proposed metrics evaluate the comprehensiveness and accuracy of image captions by comparing the embedding groups of ground-truth image regions and generated text partitions. Extensive experiments on the Visual Genome dataset validate the effectiveness of our approach, demonstrating that it produces detailed descriptions while maintaining low hallucination levels."
                    },
                    {
                        "title": "Discover and Mitigate Multiple Biased Subgroups in Image Classifiers",
                        "abstract": "Machine learning models can perform well on indistribution data but often fail on biased subgroups that are underrepresented in the training data, hindering the robustness of models for reliable applications. Such subgroups are typically unknown due to the absence of subgroup labels. Discovering biased subgroups is the key to understanding models' failure modes and further improving models' robustness. Most previous works of subgroup discovery make an implicit assumption that models only underperform on a single biased subgroup, which does not hold on in-the-wild data where multiple biased subgroups exist. In this work, we propose Decomposition, Interpretation, and Mitigation (DIM), a novel method to address a more challenging but also more practical problem of discovering multiple biased subgroups in image classifiers. Our approach decomposes the image features into multiple components that represent multiple subgroups. This decomposition is achieved via a bilinear dimension reduction method, Partial Least Square (PLS), guided by useful supervision from the image classifier. We further interpret the semantic meaning of each subgroup component by generating natural language descriptions using vision-language foundation models. Finally, DIM mitigates multiple biased subgroups simultaneously via two strategies, including the data and model-centric strategies. Extensive experiments on CIFAR-100 and Breeds datasets demonstrate the effectiveness of DIM in discovering and mitigating multiple biased subgroups. Furthermore, DIM uncovers the failure modes of the classifier on Hard ImageNet, showcasing its broader applicability to understanding model bias in image classifiers. The code is available at https://github.com/ZhangAIPI/DIM."
                    },
                    {
                        "title": "Will the Inclusion of Generated Data Amplify Bias Across Generations in Future Image Classification Models?",
                        "abstract": "As the demand for high-quality training data escalates, researchers have increasingly turned to generative models to create synthetic data, addressing data scarcity and enabling continuous model improvement. However, reliance on self-generated data introduces a critical question: Will this practice amplify bias in future models? While most research has focused on overall performance, the impact on model bias, particularly subgroup bias, remains underexplored. In this work, we investigate the effects of the generated data on image classification tasks, with a specific focus on bias. We develop a practical simulation environment that integrates a self-consuming loop, where the generative model and classification model are trained synergistically. Hundreds of experiments are conducted on Colorized MNIST, CIFAR-20/100, and Hard ImageNet datasets to reveal changes in fairness metrics across generations. In addition, we provide a conjecture to explain the bias dynamics when training models on continuously augmented datasets across generations. Our findings contribute to the ongoing debate on the implications of synthetic data for fairness in real-world applications."
                    },
                    {
                        "title": "Bag of Tricks to Boost Adversarial Transferability",
                        "abstract": "Deep neural networks are widely known to be vulnerable to adversarial examples. However, vanilla adversarial examples generated under the white-box setting often exhibit low transferability across different models. Since adversarial transferability poses more severe threats to practical applications, various approaches have been proposed for better transferability, including gradient-based, input transformation-based, and model-related attacks, \\etc. In this work, we find that several tiny changes in the existing adversarial attacks can significantly affect the attack performance, \\eg, the number of iterations and step size. Based on careful studies of existing adversarial attacks, we propose a bag of tricks to enhance adversarial transferability, including momentum initialization, scheduled step size, dual example, spectral-based input transformation, and several ensemble strategies. Extensive experiments on the ImageNet dataset validate the high effectiveness of our proposed tricks and show that combining them can further boost adversarial transferability. Our work provides practical insights and techniques to enhance adversarial transferability, and offers guidance to improve the attack performance on the real-world application through simple adjustments."
                    },
                    {
                        "title": "Quadratic Is Not What You Need For Multimodal Large Language Models",
                        "abstract": "In the past year, the capabilities of Multimodal Large Language Models (MLLMs) have significantly improved across various aspects. However, constrained by the quadratic growth of computation in LLMs as the number of tokens increases, efficiency has become a bottleneck for further scaling MLLMs. Although recent efforts have been made to prune visual tokens or use more lightweight LLMs to reduce computation, the problem of quadratic growth in computation with the increase of visual tokens still persists. To address this, we propose a novel approach: instead of reducing the input visual tokens for LLMs, we focus on pruning vision-related computations within the LLMs. After pruning, the computation growth in the LLM is no longer quadratic with the increase of visual tokens, but linear. Surprisingly, we found that after applying such extensive pruning, the capabilities of MLLMs are comparable with the original one and even superior on some benchmarks with only 25% of the computation. This finding opens up the possibility for MLLMs to incorporate much denser visual tokens. Additionally, based on this finding, we further analyzed some architectural design deficiencies in existing MLLMs and proposed promising improvements. To the best of our knowledge, this is the first study to investigate the computational redundancy in the LLM's vision component of MLLMs. Code and checkpoints will be released soon."
                    },
                    {
                        "title": "Forward Learning for Gradient-based Black-box Saliency Map Generation",
                        "abstract": "Gradient-based saliency maps are widely used to explain deep neural network decisions. However, as models become deeper and more black-box, such as in closed-source APIs like ChatGPT, computing gradients become challenging, hindering conventional explanation methods. In this work, we introduce a novel unified framework for estimating gradients in black-box settings and generating saliency maps to interpret model decisions. We employ the likelihood ratio method to estimate output-to-input gradients and utilize them for saliency map generation. Additionally, we propose blockwise computation techniques to enhance estimation accuracy. Extensive experiments in black-box settings validate the effectiveness of our method, demonstrating accurate gradient estimation and explainability of generated saliency maps. Furthermore, we showcase the scalability of our approach by applying it to explain GPT-Vision, revealing the continued relevance of gradient-based explanation methods in the era of large, closed-source, and black-box models."
                    },
                    {
                        "title": "Scalable CP Decomposition for Tensor Learning using GPU Tensor Cores",
                        "abstract": "CP decomposition is a powerful tool for data science, especially gene analysis, deep learning, and quantum computation. However, the application of tensor decomposition is largely hindered by the exponential increment of the computational complexity and storage consumption with the size of tensors. While the data in our real world is usually presented as trillion- or even exascale-scale tensors, existing work can only support billion-scale scale tensors. In our work, we propose the Exascale-Tensor to mitigate the significant gap. Specifically, we propose a compression-based tensor decomposition framework, namely the exascale-tensor, to support exascale tensor decomposition. Then, we carefully analyze the inherent parallelism and propose a bag of strategies to improve computational efficiency. Last, we conduct experiments to decompose tensors ranging from million-scale to trillion-scale for evaluation. Compared to the baselines, the exascale-tensor supports 8,000x larger tensors and a speedup up to 6.95x. We also apply our method to two real-world applications, including gene analysis and tensor layer neural networks, of which the numeric results demonstrate the scalability and effectiveness of our method."
                    },
                    {
                        "title": "Diversifying the High-level Features for better Adversarial Transferability",
                        "abstract": "Given the great threat of adversarial attacks against Deep Neural Networks (DNNs), numerous works have been proposed to boost transferability to attack real-world applications. However, existing attacks often utilize advanced gradient calculation or input transformation but ignore the white-box model. Inspired by the fact that DNNs are over-parameterized for superior performance, we propose diversifying the high-level features (DHF) for more transferable adversarial examples. In particular, DHF perturbs the high-level features by randomly transforming the high-level features and mixing them with the feature of benign samples when calculating the gradient at each iteration. Due to the redundancy of parameters, such transformation does not affect the classification performance but helps identify the invariant features across different models, leading to much better transferability. Empirical evaluations on ImageNet dataset show that DHF could effectively improve the transferability of existing momentum-based attacks. Incorporated into the input transformation-based attacks, DHF generates more transferable adversarial examples and outperforms the baselines with a clear margin when attacking several defense models, showing its generalization to various attacks and high effectiveness for boosting transferability. Code is available at https://github.com/Trustworthy-AI-Group/DHF."
                    },
                    {
                        "title": "Improving Adversarial Transferability with Scheduled Step Size and Dual Example",
                        "abstract": "Deep neural networks are widely known to be vulnerable to adversarial examples, especially showing significantly poor performance on adversarial examples generated under the white-box setting. However, most white-box attack methods rely heavily on the target model and quickly get stuck in local optima, resulting in poor adversarial transferability. The momentum-based methods and their variants are proposed to escape the local optima for better transferability. In this work, we notice that the transferability of adversarial examples generated by the iterative fast gradient sign method (I-FGSM) exhibits a decreasing trend when increasing the number of iterations. Motivated by this finding, we argue that the information of adversarial perturbations near the benign sample, especially the direction, benefits more on the transferability. Thus, we propose a novel strategy, which uses the Scheduled step size and the Dual example (SD), to fully utilize the adversarial information near the benign sample. Our proposed strategy can be easily integrated with existing adversarial attack methods for better adversarial transferability. Empirical evaluations on the standard ImageNet dataset demonstrate that our proposed method can significantly enhance the transferability of existing adversarial attacks."
                    },
                    {
                        "title": "Video Understanding with Large Language Models: A Survey",
                        "abstract": "With the burgeoning growth of online video platforms and the escalating volume of video content, the demand for proficient video understanding tools has intensified markedly. Given the remarkable capabilities of large language models (LLMs) in language and multimodal tasks, this survey provides a detailed overview of recent advancements in video understanding that harness the power of LLMs (Vid-LLMs). The emergent capabilities of Vid-LLMs are surprisingly advanced, particularly their ability for open-ended multi-granularity (general, temporal, and spatiotemporal) reasoning combined with commonsense knowledge, suggesting a promising path for future video understanding. We examine the unique characteristics and capabilities of Vid-LLMs, categorizing the approaches into three main types: Video Analyzer x LLM, Video Embedder x LLM, and (Analyzer + Embedder) x LLM. Furthermore, we identify five sub-types based on the functions of LLMs in Vid-LLMs: LLM as Summarizer, LLM as Manager, LLM as Text Decoder, LLM as Regressor, and LLM as Hidden Layer. Furthermore, this survey presents a comprehensive study of the tasks, datasets, benchmarks, and evaluation methodologies for Vid-LLMs. Additionally, it explores the expansive applications of Vid-LLMs across various domains, highlighting their remarkable scalability and versatility in real-world video understanding challenges. Finally, it summarizes the limitations of existing Vid-LLMs and outlines directions for future research. For more information, readers are recommended to visit the repository at https://github.com/yunlong10/Awesome-LLMs-for-Video-Understanding."
                    },
                    {
                        "title": "Structure Invariant Transformation for better Adversarial Transferability",
                        "abstract": "Given the severe vulnerability of Deep Neural Networks (DNNs) against adversarial examples, there is an urgent need for an effective adversarial attack to identify the deficiencies of DNNs in security-sensitive applications. As one of the prevalent black-box adversarial attacks, the existing transfer-based attacks still cannot achieve comparable performance with the white-box attacks. Among these, input transformation based attacks have shown remarkable effectiveness in boosting transferability. In this work, we find that the existing input transformation based attacks transform the input image globally, resulting in limited diversity of the transformed images. We postulate that the more diverse transformed images result in better transferability. Thus, we investigate how to locally apply various transformations onto the input image to improve such diversity while preserving the structure of image. To this end, we propose a novel input transformation based attack, called Structure Invariant Transformation (SIA), which applies a random image transformation onto each image block to craft a set of diverse images for gradient calculation. Extensive experiments on the standard ImageNet dataset demonstrate that SIA exhibits much better transferability than the existing SOTA input transformation based attacks on CNN-based and transformer-based models, showing its generality and superiority in boosting transferability. Code is available at https://github.com/xiaosen-wang/SIT."
                    },
                    {
                        "title": "How Robust is your Fair Model? Exploring the Robustness of Diverse Fairness Strategies",
                        "abstract": "With the introduction of machine learning in high-stakes decision making, ensuring algorithmic fairness has become an increasingly important problem to solve. In response to this, many mathematical definitions of fairness have been proposed, and a variety of optimisation techniques have been developed, all designed to maximise a defined notion of fairness. However, fair solutions are reliant on the quality of the training data, and can be highly sensitive to noise. Recent studies have shown that robustness (the ability for a model to perform well on unseen data) plays a significant role in the type of strategy that should be used when approaching a new problem and, hence, measuring the robustness of these strategies has become a fundamental problem. In this work, we therefore propose a new criterion to measure the robustness of various fairness optimisation strategies - the robustness ratio. We conduct multiple extensive experiments on five bench mark fairness data sets using three of the most popular fairness strategies with respect to four of the most popular definitions of fairness. Our experiments empirically show that fairness methods that rely on threshold optimisation are very sensitive to noise in all the evaluated data sets, despite mostly outperforming other methods. This is in contrast to the other two methods, which are less fair for low noise scenarios but fairer for high noise ones. To the best of our knowledge, we are the first to quantitatively evaluate the robustness of fairness optimisation strategies. This can potentially can serve as a guideline in choosing the most suitable fairness strategy for various data sets."
                    },
                    {
                        "title": "Parallel TTr1-Tensor: Randomized Compression-based Scheme for Tensor Train Rank-1 Decomposition",
                        "abstract": "Tensor train rank-1 (TTr1) decomposition, as a variant of tensor train decomposition, is widely used in quantum machine learning to decompose high-order tensors into low-order tree-like tensor networks. However, due to the storage limitation, tensors can not be fully loaded in the main memory simultaneously. A key step in the optimization of TTr1 is the utilization of Singular Value Decomposition (SVD) for the tensor matricization, and memory limitation becomes the bottleneck of existing TTr1 decomposition. In this work, we propose Parallel TTr1-Tensor , a randomized compression-based scheme for tensor train rank-1 decomposition by trading computation for storage. In addition, we maximize the parallelism of tensor operations to accelerate the proposed scheme on GPUs. In experiments, we test the relative error of reconstruction and running time of our scheme on various scales. Our method achieves a maximum of 151 . 2 \u00d7 speedup on GPUs versus CPUs with less than 20% relative errors. To our knowledge, this is the \ufb01rst work to utilize the compression-based technique to handle the TTr1 decomposition bottleneck due to the limited memory."
                    }
                ]
            },
            "d6afb103-6bc6-4c7c-a3bb-a20cc46fd9bb": {
                "pk": "d6afb103-6bc6-4c7c-a3bb-a20cc46fd9bb",
                "name": "Chenliang Xu",
                "collaborators": [
                    "Yapeng Tian",
                    "Chao Huang",
                    "Anurag Kumar",
                    "Zhiheng Li",
                    "Jiebo Luo",
                    "Susan Liang",
                    "Di Hu",
                    "Zeliang Zhang",
                    "Peihan Liu",
                    "Xiaosen Wang"
                ],
                "domain": [
                    "Audio-Visual Learning",
                    "Deep Learning",
                    "Computer Vision",
                    "Multimodal Processing"
                ],
                "publications": [
                    {
                        "title": "High-Quality Visually-Guided Sound Separation from Diverse Categories",
                        "abstract": "We propose DAVIS, a Diffusion-based Audio-VIsual Separation framework that solves the audio-visual sound source separation task through generative learning. Existing methods typically frame sound separation as a mask-based regression problem, achieving significant progress. However, they face limitations in capturing the complex data distribution required for high-quality separation of sounds from diverse categories. In contrast, DAVIS leverages a generative diffusion model and a Separation U-Net to synthesize separated sounds directly from Gaussian noise, conditioned on both the audio mixture and the visual information. With its generative objective, DAVIS is better suited to achieving the goal of high-quality sound separation across diverse sound categories. We compare DAVIS to existing state-of-the-art discriminative audio-visual separation methods on the AVE and MUSIC datasets, and results show that DAVIS outperforms other methods in separation quality, demonstrating the advantages of our framework for tackling the audio-visual source separation task."
                    },
                    {
                        "title": "Improving Adversarial Transferability with Scheduled Step Size and Dual Example",
                        "abstract": "Deep neural networks are widely known to be vulnerable to adversarial examples, especially showing significantly poor performance on adversarial examples generated under the white-box setting. However, most white-box attack methods rely heavily on the target model and quickly get stuck in local optima, resulting in poor adversarial transferability. The momentum-based methods and their variants are proposed to escape the local optima for better transferability. In this work, we notice that the transferability of adversarial examples generated by the iterative fast gradient sign method (I-FGSM) exhibits a decreasing trend when increasing the number of iterations. Motivated by this finding, we argue that the information of adversarial perturbations near the benign sample, especially the direction, benefits more on the transferability. Thus, we propose a novel strategy, which uses the Scheduled step size and the Dual example (SD), to fully utilize the adversarial information near the benign sample. Our proposed strategy can be easily integrated with existing adversarial attack methods for better adversarial transferability. Empirical evaluations on the standard ImageNet dataset demonstrate that our proposed method can significantly enhance the transferability of existing adversarial attacks."
                    },
                    {
                        "title": "AV-NeRF: Learning Neural Fields for Real-World Audio-Visual Scene Synthesis",
                        "abstract": "Can machines recording an audio-visual scene produce realistic, matching audio-visual experiences at novel positions and novel view directions? We answer it by studying a new task -- real-world audio-visual scene synthesis -- and a first-of-its-kind NeRF-based approach for multimodal learning. Concretely, given a video recording of an audio-visual scene, the task is to synthesize new videos with spatial audios along arbitrary novel camera trajectories in that scene. We propose an acoustic-aware audio generation module that integrates prior knowledge of audio propagation into NeRF, in which we implicitly associate audio generation with the 3D geometry and material properties of a visual environment. Furthermore, we present a coordinate transformation module that expresses a view direction relative to the sound source, enabling the model to learn sound source-centric acoustic fields. To facilitate the study of this new task, we collect a high-quality Real-World Audio-Visual Scene (RWAVS) dataset. We demonstrate the advantages of our method on this real-world dataset and the simulation-based SoundSpaces dataset."
                    },
                    {
                        "title": "PEANUT: A Human-AI Collaborative Tool for Annotating Audio-Visual Data",
                        "abstract": "Audio-visual learning seeks to enhance the computer\u2019s multi-modal perception leveraging the correlation between the auditory and visual modalities. Despite their many useful downstream tasks, such as video retrieval, AR/VR, and accessibility, the performance and adoption of existing audio-visual models have been impeded by the availability of high-quality datasets. Annotating audio-visual datasets is laborious, expensive, and time-consuming. To address this challenge, we designed and developed an efficient audio-visual annotation tool called Peanut. Peanut\u2019s human-AI collaborative pipeline separates the multi-modal task into two single-modal tasks, and utilizes state-of-the-art object detection and sound-tagging models to reduce the annotators\u2019 effort to process each frame and the number of manually-annotated frames needed. A within-subject user study with 20 participants found that Peanut can significantly accelerate the audio-visual data annotation process while maintaining high annotation accuracy."
                    },
                    {
                        "title": "Egocentric Audio-Visual Object Localization",
                        "abstract": "Humans naturally perceive surrounding scenes by unifying sound and sight from a first-person view. Likewise, machines are advanced to approach human intelligence by learning with multisensory inputs from an egocentric perspective. In this paper, we explore the challenging egocentric audio-visual object localization task and observe that 1) egomotion commonly exists in first-person recordings, even within a short duration; 2) The out-of-view sound components can be created when wearers shift their attention. To address the first problem, we propose a geometryaware temporal aggregation module that handles the egomotion explicitly. The effect of egomotion is mitigated by estimating the temporal geometry transformation and exploiting it to update visual representations. Moreover, we propose a cascaded feature enhancement module to over-come the second issue. It improves cross-modal localization robustness by disentangling visually-indicated audio representation. During training, we take advantage of the naturally occurring audio-visual temporal synchronization as the \u201cfree\u201d self-supervision to avoid costly labeling. We also annotate and create the Epic Sounding Object dataset for evaluation purposes. Extensive experiments show that our method achieves state-of-the-art localization performance in egocentric videos and can be generalized to diverse audio-visual scenes. Code is available at https://github.com/WikiChao/Ego-AV-Loc."
                    },
                    {
                        "title": "Unveiling Cross Modality Bias in Visual Question Answering: A Causal View with Possible Worlds VQA",
                        "abstract": "To increase the generalization capability of VQA systems, many recent studies have tried to de-bias spurious language or vision associations that shortcut the question or image to the answer. Despite these efforts, the literature fails to address the confounding effect of vision and language simultaneously. As a result, when they reduce bias learned from one modality, they usually increase bias from the other. In this paper, we first model a confounding effect that causes language and vision bias simultaneously, then propose a counterfactual inference to remove the influence of this effect. The model trained in this strategy can concurrently and efficiently reduce vision and language bias. To the best of our knowledge, this is the first work to reduce biases resulting from confounding effects of vision and language in VQA, leveraging causal explain-away relations. We accompany our method with an explain-away strategy, pushing the accuracy of the questions with numerical answers results compared to existing methods that have been an open problem. The proposed method outperforms the state-of-the-art methods in VQA-CP v2 datasets."
                    },
                    {
                        "title": "Learning to Answer Questions in Dynamic Audio-Visual Scenarios",
                        "abstract": "In this paper, we focus on the Audio-Visual Question Answering (AVQA) task, which aims to answer questions regarding different visual objects, sounds, and their associations in videos. The problem requires comprehensive multimodal understanding and spatio-temporal reasoning over audio-visual scenes. To benchmark this task and facilitate our study, we introduce a large-scale MUSIC-AVQA dataset, which contains more than 45K question-answer pairs covering 33 different question templates spanning over different modalities and question types. We develop several baselines and introduce a spatio-temporal grounded audio-visual network for the AVQA problem. Our results demonstrate that AVQA benefits from multisensory perception and our model outperforms recent A-, V-, and AVQA approaches. We believe that our built dataset has the potential to serve as testbed for evaluating and promoting progress in audio-visual scene understanding and spatio-temporal reasoning. Code and dataset: http://gewu-lab.github.io/MUSIC-AVQA/"
                    },
                    {
                        "title": "A Whac-A-Mole Dilemma: Shortcuts Come in Multiples Where Mitigating One Amplifies Others",
                        "abstract": "Machine learning models have been found to learn shortcuts\u2014unintended decision rules that are unable to generalize\u2014undermining models' reliability. Previous works address this problem under the tenuous assumption that only a single shortcut exists in the training data. Real-world images are rife with multiple visual cues from background to texture. Key to advancing the reliability of vision systems is understanding whether existing methods can overcome multiple shortcuts or struggle in a Whac-A-Mole game, i.e., where mitigating one shortcut amplifies reliance on others. To address this shortcoming, we propose two benchmarks: 1) UrbanCars, a dataset with precisely controlled spurious cues, and 2) ImageNet-W, an evaluation set based on ImageNet for watermark, a shortcut we discovered affects nearly every modern vision model. Along with texture and background, ImageNet-W allows us to study multiple shortcuts emerging from training on natural images. We find computer vision models, including large foundation models\u2014regardless of training set, architecture, and supervision\u2014struggle when multiple shortcuts are present. Even methods explicitly designed to combat shortcuts struggle in a Whac-A-Mole dilemma. To tackle this challenge, we propose Last Layer Ensemble, a simple-yet-effective method to mitigate multiple shortcuts without Whac-A-Mole behavior. Our results surface multi-shortcut mitigation as an overlooked challenge critical to advancing the reliability of vision systems. The datasets and code are released: https://github.com/facebookresearch/Whac-A-Mole."
                    },
                    {
                        "title": "Transformer-empowered Multi-scale Contextual Matching and Aggregation for Multi-contrast MRI Super-resolution",
                        "abstract": "Magnetic resonance imaging (MRI) can present multicontrast images of the same anatomical structures, enabling multi-contrast super-resolution (SR) techniques. Compared with SR reconstruction using a single-contrast, multicontrast SR reconstruction is promising to yield SR images with higher quality by leveraging diverse yet complementary information embedded in different imaging modalities. However, existing methods still have two shortcomings: (1) they neglect that the multi-contrast features at different scales contain different anatomical details and hence lack effective mechanisms to match and fuse these features for better reconstruction; and (2) they are still deficient in capturing long-range dependencies, which are essential for the regions with complicated anatomical structures. We propose a novel network to comprehensively address these problems by developing a set of innovative Transformer-empowered multi-scale contextual matching and aggregation techniques; we call it McMRSR. Firstly, we tame transformers to model long-range dependencies in both reference and target images. Then, a new multi-scale contextual matching method is proposed to capture corresponding contexts from reference features at different scales. Furthermore, we introduce a multi-scale aggregation mechanism to gradually and interactively aggregate multi-scale matched features for reconstructing the target SR MR image. Extensive experiments demonstrate that our network outperforms state-of-the-art approaches and has great potential to be applied in clinical practice. Codes are available at https://github.com/XAIMI-Lab/McMRSR."
                    },
                    {
                        "title": "SpaceEdit: Learning a Unified Editing Space for Open-Domain Image Color Editing",
                        "abstract": "Recently, large pretrained models (e.g., BERT, Style-GAN, CLIP) show great knowledge transfer and generalization capability on various downstream tasks within their domains. Inspired by these efforts, in this paper we propose a unified model for open-domain image editing focusing on color and tone adjustment of open-domain images while keeping their original content and structure. Our model learns a unified editing space that is more semantic, intu-itive, and easy to manipulate than the operation space (e.g., contrast, brightness, color curve) used in many existing photo editing softwares. Our model belongs to the image-to-image translation framework which consists of an image encoder and decoder, and is trained on pairs of before-and-after edited images to produce multimodal outputs. We show that by inverting image pairs into latent codes of the learned editing space, our model can be leveraged for vari-ous downstream editing tasks such as language-guided image editing, personalized editing, editing-style clustering, retrieval, etc. We extensively study the unique properties of the editing space in experiments and demonstrate superior performance on the aforementioned tasks11Code and supplementary material can be found at the project page https://jshi31.github.io/SpaceEdit."
                    },
                    {
                        "title": "StyleT2I: Toward Compositional and High-Fidelity Text-to-Image Synthesis",
                        "abstract": "Although progress has been made for text-to-image synthesis, previous methods fall short of generalizing to unseen or underrepresented attribute compositions in the input text. Lacking compositionality could have severe implications for robustness and fairness, e.g., inability to synthesize the face images of underrepresented demographic groups. In this paper, we introduce a new framework, StyleT2I, to improve the compositionality of text-to-image synthesis. Specifically, we propose a CLIP-guided Contrastive Loss to better distinguish different compositions among different sentences. To further improve the compositionality, we design a novel Semantic Matching Loss and a Spatial Constraint to identify attributes' latent directions for intended spatial region manipulations, leading to better disentangled latent representations of attributes. Based on the identified latent directions of attributes, we propose Compositional Attribute Adjustment to adjust the latent code, resulting in better compositionality of image synthesis. In addition, we leverage the $l_{2}$-norm regularization of identified latent directions (norm penalty) to strike a nice balance between image-text alignment and image fidelity. In the experiments, we devise a new dataset split and an evaluation metric to evaluate the compositionality of text-to-image synthesis models. The results show that StyleT2I outperforms previous approaches in terms of the consistency between the input text and synthesized images and achieves higher fidelity."
                    },
                    {
                        "title": "Cross-modal Contrastive Distillation for Instructional Activity Anticipation",
                        "abstract": "In this study, we aim to predict the plausible future action steps given an observation of the past and study the task of instructional activity anticipation. Unlike previous anticipation tasks that aim at action label prediction, our work targets at generating natural language outputs that provide interpretable and accurate descriptions of future action steps. It is a challenging task due to the lack of semantic information extracted from the instructional videos. To overcome this challenge, we propose a novel knowledge distillation framework to exploit the related external textual knowledge to assist the visual anticipation task. However, previous knowledge distillation techniques generally transfer information within the same modality. To bridge the gap between the visual and text modalities during the distillation process, we devise a novel cross-modal contrastive distillation (CCD) scheme, which facilitates knowledge distillation between teacher and student in heterogeneous modalities with the proposed cross-modal distillation loss. We evaluate our method on the Tasty Videos dataset. CCD improves the anticipation performance of the visual-alone student model by a large margin of 40.2% relatively in BLEU4. Our approach also outperforms the state-of-the-art approaches by a large margin."
                    },
                    {
                        "title": "Audio-Visual Object Localization in Egocentric Videos",
                        "abstract": "In the past decade, substantial efforts have been made to build benchmarks, establish new tasks, and create learning frameworks for egocentric video understanding. However, the majority of works focus on visual scene analysis. It presents significant challenges for designing powerful and robust egocentric video understanding systems. First, people with wearable devices usually record videos in naturalistic surroundings, where a variety of illumination conditions, object appearance, and motion patterns are shown. The dynamic visual variations introduce difficulties in achieving accurate visual perception. Second, egocentric scenes are often perceived within a limited field of view (FoV). The common body and head movements of users cause frequent view changes, which brings object deformation and creates dynamic out-of-scene content. Therefore, it is difficult to fully decode the surrounding information and perceive dynamic scenes in egocentric videos. Audio, acting as an essential but less focused modality, often provides synchronized and complementary information with the video stream. In contrast to the variability of first-person visual footage, sound provides stable and persistent signals associated with the depicted events. Therefore, audio is an essential ingredient for \u201ccooking\u201d egocentric video understanding. To capture fine-grained audio-visual association and tackle the challenges in egocentric vision, in this paper, we propose to solve an egocentric audio-visual object localization task, which aims to associate audio with dynamic visual scenes and localize sounding objects in egocentric videos. We develop a new framework to explicitly model the distinct characteristics of egocentric videos by integrating audio. Concretely, the egomotion in videos leads to various object deformations, making it difficult to consistently localize the sounding objects. Despite the downside, we found that the egomotion also provides rich geometry information about the underlying scene and hints at the relative geometric transformation between frames. Motivated by this, we use the predicted geometric transformation to mitigate the object deformation in the embedding space and align the visual features. Based on the aligned features, we further leverage the temporal contexts across frames to learn discriminative cues for localization. Our localization model aims to find precise audio-visual associations. However, the results might be affected by: (i) noisy audio with background sound components; (ii) unrelated visual contents, e.g., other silent objects. To this end, we propose a cascaded feature enhancement module to mitigate the audio noise and improve cross-modal localization robustness. Our framework can be trained by taking nature audio-visual temporal synchronization as the \u201cfree\u201d supervision. We also annotate an Epic Sounding dataset to facilitate quantitative comparison. In all, our contributions are: (i) an effective geometryaware temporal aggregation approach to deal with unique egomotion in first-person videos; (ii) a novel cascaded feature enhancement module to progressively inject the audio and visual features with localization cues; and (iii) an Epic Sounding Object dataset with annotations on sounding objects to benchmark the localization performance in egocentric videos."
                    },
                    {
                        "title": "Speech-In-Noise Comprehension is Improved When Viewing a Deep-Neural-Network-Generated Talking Face",
                        "abstract": "Listening in a noisy environment is challenging, but many previous studies have demonstrated that comprehension of speech can be substantially improved by looking at the talker's face. We recently developed a deep neural network (DNN) based system that generates movies of a talking face from speech audio and a single face image. In this study, we aimed to quantify the benefits that such a system can bring to speech comprehension, especially in noise. The target speech audio was masked with signal to noise ratios of \u22129, \u22126, \u22123, and 0\u2005dB and was presented to subjects in three audio-visual (AV) stimulus conditions: (1) synthesized AV: audio with the synthesized talking face movie; (2) natural AV: audio with the original movie from the corpus; and (3) audio-only: audio with a static image of the talker. Subjects were asked to type the sentences they heard in each trial and keyword recognition was quantified for each condition. Overall, performance in the synthesized AV condition fell approximately halfway between the other two conditions, showing a marked improvement over the audio-only control but still falling short of the natural AV condition. Every subject showed some benefit from the synthetic AV stimulus. The results of this study support the idea that a DNN-based model that generates a talking face from speech audio can meaningfully enhance comprehension in noisy environments, and has the potential to be used as a visual hearing aid."
                    },
                    {
                        "title": "Discover and Mitigate Unknown Biases with Debiasing Alternate Networks",
                        "abstract": "Deep image classifiers have been found to learn biases from datasets. To mitigate the biases, most previous methods require labels of protected attributes (e.g., age, skin tone) as full-supervision, which has two limitations: 1) it is infeasible when the labels are unavailable; 2) they are incapable of mitigating unknown biases -- biases that humans do not preconceive. To resolve those problems, we propose Debiasing Alternate Networks (DebiAN), which comprises two networks -- a Discoverer and a Classifier. By training in an alternate manner, the discoverer tries to find multiple unknown biases of the classifier without any annotations of biases, and the classifier aims at unlearning the biases identified by the discoverer. While previous works evaluate debiasing results in terms of a single bias, we create Multi-Color MNIST dataset to better benchmark mitigation of multiple biases in a multi-bias setting, which not only reveals the problems in previous methods but also demonstrates the advantage of DebiAN in identifying and mitigating multiple biases simultaneously. We further conduct extensive experiments on real-world datasets, showing that the discoverer in DebiAN can identify unknown biases that may be hard to be found by humans. Regarding debiasing, DebiAN achieves strong bias mitigation performance."
                    },
                    {
                        "title": "Face Forgery Detection via Symmetric Transformer",
                        "abstract": "The deep learning-based face forgery detection is a novel yet challenging task. Despite impressive results have been achieved, there are still some limitations in the existing methods. For example, the previous methods are hard to maintain consistent predictions for consecutive frames, even if all of those frames are actually forged. We propose a symmetric transformer for channel and spatial feature extraction, which is because the channel and spatial features of a robust forgery detector should be consistent in the temporal domain. The symmetric transformer adopt the newly-designed attention-based strategies for channel variance and spatial gradients as the vital features, which greatly improves the robustness of deepfake video detection. Moreover, this symmetric structure acts on temporal and spatial features respectively, which ensures the robustness of detection from two different aspects. Our symmetric transformer is an end-to-end optimized network. Experiments are conducted on various settings, the proposed methods achieve significantly improvement on prediction robustness and perform better than state-of-the-art methods on different datasets."
                    },
                    {
                        "title": "Cyclic Co-Learning of Sounding Object Visual Grounding and Sound Separation",
                        "abstract": "There are rich synchronized audio and visual events in our daily life. Inside the events, audio scenes are associated with the corresponding visual objects; meanwhile, sounding objects can indicate and help to separate their individual sounds in the audio track. Based on this observation, in this paper, we propose a cyclic co-learning (CCoL) paradigm that can jointly learn sounding object visual grounding and audio-visual sound separation in a unified framework. Concretely, we can leverage grounded object-sound relations to improve the results of sound separation. Meanwhile, benefiting from discriminative information from separated sounds, we improve training example sampling for sounding object grounding, which builds a co-learning cycle for the two tasks and makes them mutually beneficial. Extensive experiments show that the proposed framework outperforms the compared recent approaches on both tasks, and they can benefit from each other with our cyclic co-learning. The source code and pre-trained models are released in https://github.com/YapengTian/CCOL-CVPR21."
                    },
                    {
                        "title": "Discover the Unknown Biased Attribute of an Image Classifier",
                        "abstract": "Recent works find that AI algorithms learn biases from data. Therefore, it is urgent and vital to identify biases in AI algorithms. However, the previous bias identification pipeline overly relies on human experts to conjecture potential biases (e.g., gender), which may neglect other underlying biases not realized by humans. To help human experts better find the AI algorithms\u2019 biases, we study a new problem in this work \u2013 for a classifier that predicts a target attribute of the input image, discover its unknown biased attribute.To solve this challenging problem, we use a hyperplane in the generative model\u2019s latent space to represent an image attribute; thus, the original problem is transformed to optimizing the hyperplane\u2019s normal vector and offset. We propose a novel total-variation loss within this framework as the objective function and a new orthogonalization penalty as a constraint. The latter prevents trivial solutions in which the discovered biased attribute is identical with the target or one of the known-biased attributes. Extensive experiments on both disentanglement datasets and real-world datasets show that our method can discover biased attributes and achieve better disentanglement w.r.t. target attributes. Furthermore, the qualitative results show that our method can discover unnoticeable biased attributes for various object and scene classifiers, proving our method\u2019s generalizability for detecting biased attributes in diverse domains of images."
                    },
                    {
                        "title": "Zooming SlowMo: An Efficient One-Stage Framework for Space-Time Video Super-Resolution",
                        "abstract": "In this paper, we address the space-time video super-resolution, which aims at generating a high-resolution (HR) slow-motion video from a low-resolution (LR) and low frame rate (LFR) video sequence. A na\\\"ive method is to decompose it into two sub-tasks: video frame interpolation (VFI) and video super-resolution (VSR). Nevertheless, temporal interpolation and spatial upscaling are intra-related in this problem. Two-stage approaches cannot fully make use of this natural property. Besides, state-of-the-art VFI or VSR deep networks usually have a large frame reconstruction module in order to obtain high-quality photo-realistic video frames, which makes the two-stage approaches have large models and thus be relatively time-consuming. To overcome the issues, we present a one-stage space-time video super-resolution framework, which can directly reconstruct an HR slow-motion video sequence from an input LR and LFR video. Instead of reconstructing missing LR intermediate frames as VFI models do, we temporally interpolate LR frame features of the missing LR frames capturing local temporal contexts by a feature temporal interpolation module. Extensive experiments on widely used benchmarks demonstrate that the proposed framework not only achieves better qualitative and quantitative performance on both clean and noisy LR frames but also is several times faster than recent state-of-the-art two-stage networks. The source code is released in https://github.com/Mukosame/Zooming-Slow-Mo-CVPR-2020 ."
                    }
                ]
            },
            "067c13b8-5b0b-4667-9a53-d80ddab24baf": {
                "pk": "067c13b8-5b0b-4667-9a53-d80ddab24baf",
                "name": "Zhaofei Yu",
                "collaborators": [
                    "Tiejun Huang",
                    "Jianhao Ding",
                    "Wei Fang",
                    "Yanqing Chen",
                    "Yonghong Tian",
                    "Lei Ma",
                    "T. Masquelier",
                    "Tong Bu",
                    "Zhengyu Ma",
                    "Penglin Dai"
                ],
                "domain": [
                    "Spiking Neural Networks",
                    "Mobile Edge Computing",
                    "Computer Vision",
                    "Neuromorphic Computing"
                ],
                "publications": [
                    {
                        "title": "Meta Reinforcement Learning for Multi-Task Offloading in Vehicular Edge Computing",
                        "abstract": "Mobile edge computing has been a promising solution to enable real-time service in vehicular networks. However, due to high dynamics of mobile environment and heterogeneous features of vehicular services, traditional expert-based or learning-based strategies has to update handcrafted parameters or retrain learning model, which leads to intolerant overhead. Therefore, this paper investigates the problem of multi-task offloading (MTO), where there exist multiple offloading scenarios with varying parameters, such as task topology, resource requirement and transmission/computation capability. The objective is to design a unified solution to minimize task execution time under different MTO scenarios. Accordingly, we develop a Seq2seq-based Meta Reinforcement Learning algorithm for MTO (SMRL-MTO). Specifically, a bidirectional gated recurrent units integrated with attention mechanism is designed to determine offloading action by encoding sequential offloading actions and showing different preferences to different parts of input sequence. Particularly, a meta reinforcement learning framework is designed based on model-agnostic meta learning, which trains a meta policy offline and fast adapts to new MTO scenario within a few training steps. Finally, we conduct performance evaluation based on task generator DAGGEN and realistic vehicular traces, which shows that the SMRL-MTO reduces task execution time by 11.36% on average compared with greedy algorithm."
                    },
                    {
                        "title": "SpiReco: Fast and Efficient Recognition of High-Speed Moving Objects With Spike Camera",
                        "abstract": "Benefited from the high temporal resolution and high dynamic range, spike cameras have shown great potential in recognizing high-speed moving objects. However, the computer vision community has not explored this task due to the lack of spike data and annotations of high-speed moving objects. This paper contributes a novel dataset, named SpiReco (Spiking datasets for Recognition), by recording high-speed moving objects using a spike camera. To annotate the dataset, image labels from established datasets such as MNIST, CIFAR10, and CALTECH101 are utilized. Based on this new dataset, this paper proposes the first spike-based object recognition framework. The proposed framework includes a denoise module, which is designed to suppress spike noise by learning spatio-temporal correlation from neighbouring pixels. Additionally, a motion enhancement module is introduced to address high-speed and random motions. Afterwards, binarized neural networks are adopted to save computation costs. These efforts result in a fast and efficient processing framework for spiking data. Experimental results demonstrate the effectiveness of the proposed methods. For example, the proposed spike-based recognition framework achieves 80.2% accuracy in recognizing 101 classes of high-speed moving objects using only 2.2ms of spike streams. The SpiReco is available at https://github.com/Evin-X/SpiReco."
                    },
                    {
                        "title": "Spike Camera Image Reconstruction Using Deep Spiking Neural Networks",
                        "abstract": "Spike camera is a bio-inspired sensor with ultra-high temporal resolution and low energy consumption. It captures visual signals using an \u201cintegrate-and-fire\u201d mechanism and outputs a continuous stream of binary spikes. Reconstructing image sequence from spikes streams is critical for spike camera. Several reconstruction methods have been proposed in recent years. However, the computational cost of these methods is relatively high. Inspired by the fact that spiking neural networks (SNNs) are energy efficient and support time-series signal processing inherently, we propose a lightweight SNN for spike camera image reconstruction (abbreviated to SSIR). Experimental results show that SSIR achieves comparable performance with the state-of-the-art (SOTA) methods at much lower computation and energy cost."
                    },
                    {
                        "title": "Unsupervised Optical Flow Estimation with Dynamic Timing Representation for Spike Camera",
                        "abstract": "Efficiently selecting an appropriate spike stream data length to extract precise information is the key to the spike vision tasks. To address this issue, we propose a dynamic timing representation for spike streams. Based on multi-layers architecture, it applies dilated convolutions on temporal dimension to extract features on multi-temporal scales with few parameters. And we design layer attention to dynamically fuse these features. Moreover, we propose an unsupervised learning method for optical flow estimation in a spike-based manner to break the dependence on labeled data. In addition, to verify the robustness, we also build a spike-based synthetic validation dataset for extreme scenarios in autonomous driving, denoted as SSES dataset. It consists of various corner cases. Experiments show that our method can predict optical flow from spike streams in different high-speed scenes, including real scenes. For instance, our method gets $15\\%$ and $19\\%$ error reduction from the best spike-based work, SCFlow, in $\\Delta t=10$ and $\\Delta t=20$ respectively which are the same settings as the previous works."
                    },
                    {
                        "title": "Spike timing reshapes robustness against attacks in spiking neural networks",
                        "abstract": "The success of deep learning in the past decade is partially shrouded in the shadow of adversarial attacks. In contrast, the brain is far more robust at complex cognitive tasks. Utilizing the advantage that neurons in the brain communicate via spikes, spiking neural networks (SNNs) are emerging as a new type of neural network model, boosting the frontier of theoretical investigation and empirical application of artificial neural networks and deep learning. Neuroscience research proposes that the precise timing of neural spikes plays an important role in the information coding and sensory processing of the biological brain. However, the role of spike timing in SNNs is less considered and far from understood. Here we systematically explored the timing mechanism of spike coding in SNNs, focusing on the robustness of the system against various types of attacks. We found that SNNs can achieve higher robustness improvement using the coding principle of precise spike timing in neural encoding and decoding, facilitated by different learning rules. Our results suggest that the utility of spike timing coding in SNNs could improve the robustness against attacks, providing a new approach to reliable coding principles for developing next-generation brain-inspired deep learning."
                    },
                    {
                        "title": "Spike Stream Denoising via Spike Camera Simulation",
                        "abstract": "As a neuromorphic sensor with high temporal resolution, the spike camera shows enormous potential in high-speed visual tasks. However, the high-speed sampling of light propagation processes by existing cameras brings unavoidable noise phenomena. Eliminating the unique noise in spike stream is always a key point for spike-based methods. No previous work has addressed the detailed noise mechanism of the spike camera. To this end, we propose a systematic noise model for spike camera based on its unique circuit. In addition, we carefully constructed the noise evaluation equation and experimental scenarios to measure noise variables. Based on our noise model, the first benchmark for spike stream denoising is proposed which includes clear (noisy) spike stream. Further, we design a tailored spike stream denoising framework (DnSS) where denoised spike stream is obtained by decoding inferred inter-spike intervals. Experiments show that DnSS has promising performance on the proposed benchmark. Eventually, DnSS can be generalized well on real spike stream."
                    },
                    {
                        "title": "SpikingJelly: An open-source machine learning infrastructure platform for spike-based intelligence",
                        "abstract": "Spiking neural networks (SNNs) aim to realize brain-inspired intelligence on neuromorphic chips with high energy efficiency by introducing neural dynamics and spike properties. As the emerging spiking deep learning paradigm attracts increasing interest, traditional programming frameworks cannot meet the demands of the automatic differentiation, parallel computation acceleration, and high integration of processing neuromorphic datasets and deployment. In this work, we present the SpikingJelly framework to address the aforementioned dilemma. We contribute a full-stack toolkit for preprocessing neuromorphic datasets, building deep SNNs, optimizing their parameters, and deploying SNNs on neuromorphic chips. Compared to existing methods, the training of deep SNNs can be accelerated 11\u00d7, and the superior extensibility and flexibility of SpikingJelly enable users to accelerate custom models at low costs through multilevel inheritance and semiautomatic code generation. SpikingJelly paves the way for synthesizing truly energy-efficient SNN-based machine intelligence systems, which will enrich the ecology of neuromorphic computing."
                    },
                    {
                        "title": "Reducing ANN-SNN Conversion Error through Residual Membrane Potential",
                        "abstract": "Spiking Neural Networks (SNNs) have received extensive academic attention due to the unique properties of low power consumption and high-speed computing on neuromorphic chips. Among various training methods of SNNs, ANN-SNN conversion has shown the equivalent level of performance as ANNs on large-scale datasets. However, unevenness error, which refers to the deviation caused by different temporal sequences of spike arrival on activation layers, has not been effectively resolved and seriously suffers the performance of SNNs under the condition of short time-steps. In this paper, we make a detailed analysis of unevenness error and divide it into four categories. We point out that the case of the ANN output being zero while the SNN output being larger than zero accounts for the largest percentage. Based on this, we theoretically prove the sufficient and necessary conditions of this case and propose an optimization strategy based on residual membrane potential to reduce unevenness error. The experimental results show that the proposed method achieves state-of-the-art performance on CIFAR-10, CIFAR-100, and ImageNet datasets. For example, we reach top-1 accuracy of 64.32% on ImageNet with 10-steps. To the best of our knowledge, this is the first time ANN-SNN conversion can simultaneously achieve high accuracy and ultra-low-latency on the complex dataset. Code is available at https://github.com/hzc1208/ANN2SNN_SRP."
                    },
                    {
                        "title": "A Unified Framework for Soft Threshold Pruning",
                        "abstract": "Soft threshold pruning is among the cutting-edge pruning methods with state-of-the-art performance. However, previous methods either perform aimless searching on the threshold scheduler or simply set the threshold trainable, lacking theoretical explanation from a unified perspective. In this work, we reformulate soft threshold pruning as an implicit optimization problem solved using the Iterative Shrinkage-Thresholding Algorithm (ISTA), a classic method from the fields of sparse recovery and compressed sensing. Under this theoretical framework, all threshold tuning strategies proposed in previous studies of soft threshold pruning are concluded as different styles of tuning $L_1$-regularization term. We further derive an optimal threshold scheduler through an in-depth study of threshold scheduling based on our framework. This scheduler keeps $L_1$-regularization coefficient stable, implying a time-invariant objective function from the perspective of optimization. In principle, the derived pruning algorithm could sparsify any mathematical model trained via SGD. We conduct extensive experiments and verify its state-of-the-art performance on both Artificial Neural Networks (ResNet-50 and MobileNet-V1) and Spiking Neural Networks (SEW ResNet-18) on ImageNet datasets. On the basis of this framework, we derive a family of pruning methods, including sparsify-during-training, early pruning, and pruning at initialization. The code is available at https://github.com/Yanqi-Chen/LATS."
                    },
                    {
                        "title": "Bridging the Gap between ANNs and SNNs by Calibrating Offset Spikes",
                        "abstract": "Spiking Neural Networks (SNNs) have attracted great attention due to their distinctive characteristics of low power consumption and temporal information processing. ANN-SNN conversion, as the most commonly used training method for applying SNNs, can ensure that converted SNNs achieve comparable performance to ANNs on large-scale datasets. However, the performance degrades severely under low quantities of time-steps, which hampers the practical applications of SNNs to neuromorphic chips. In this paper, instead of evaluating different conversion errors and then eliminating these errors, we define an offset spike to measure the degree of deviation between actual and desired SNN firing rates. We perform a detailed analysis of offset spike and note that the firing of one additional (or one less) spike is the main cause of conversion errors. Based on this, we propose an optimization strategy based on shifting the initial membrane potential and we theoretically prove the corresponding optimal shifting distance for calibrating the spike. In addition, we also note that our method has a unique iterative property that enables further reduction of conversion errors. The experimental results show that our proposed method achieves state-of-the-art performance on CIFAR-10, CIFAR-100, and ImageNet datasets. For example, we reach a top-1 accuracy of 67.12% on ImageNet when using 6 time-steps. To the best of our knowledge, this is the first time an ANN-SNN conversion has been shown to simultaneously achieve high accuracy and ultralow latency on complex datasets. Code is available at https://github.com/hzc1208/ANN2SNN_COS."
                    },
                    {
                        "title": "Optimal ANN-SNN Conversion for High-accuracy and Ultra-low-latency Spiking Neural Networks",
                        "abstract": "Spiking Neural Networks (SNNs) have gained great attraction due to their distinctive properties of low power consumption and fast inference on neuromorphic hardware. As the most effective method to get deep SNNs, ANN-SNN conversion has achieved comparable performance as ANNs on large-scale datasets. Despite this, it requires long time-steps to match the firing rates of SNNs to the activation of ANNs. As a result, the converted SNN suffers severe performance degradation problems with short time-steps, which hamper the practical application of SNNs. In this paper, we theoretically analyze ANN-SNN conversion error and derive the estimated activation function of SNNs. Then we propose the quantization clip-floor-shift activation function to replace the ReLU activation function in source ANNs, which can better approximate the activation function of SNNs. We prove that the expected conversion error between SNNs and ANNs is zero, enabling us to achieve high-accuracy and ultra-low-latency SNNs. We evaluate our method on CIFAR-10/100 and ImageNet datasets, and show that it outperforms the state-of-the-art ANN-SNN and directly trained SNNs in both accuracy and time-steps. To the best of our knowledge, this is the first time to explore high-performance ANN-SNN conversion with ultra-low latency (4 time-steps). Code is available at https://github.com/putshua/SNN\\_conversion\\_QCFS"
                    },
                    {
                        "title": "A Novel Noise Injection-based Training Scheme for Better Model Robustness",
                        "abstract": "Noise injection-based method has been shown to be able to improve the robustness of artificial neural networks in previous work. In this work, we propose a novel noise injection-based training scheme for better model robustness. Specifically, we first develop a likelihood ratio method to estimate the gradient with respect to both synaptic weights and noise levels for stochastic gradient descent training. Then, we design an approximation for the vanilla noise injection-based training method to reduce memory and improve computational efficiency. Next, we apply our proposed scheme to spiking neural networks and evaluate the performance of classification accuracy and robustness on MNIST and Fashion-MNIST datasets. Experiment results show that our proposed method achieves a much better performance on adversarial robustness and slightly better performance on original accuracy, compared with the conventional gradient-based training method."
                    },
                    {
                        "title": "Diverse role of NMDA receptors for dendritic integration of neural dynamics",
                        "abstract": "Neurons, represented as a tree structure of morphology, have various distinguished branches of dendrites. Different types of synaptic receptors distributed over dendrites are responsible for receiving inputs from other neurons. NMDA receptors (NMDARs) are expressed as excitatory units, and play a key physiological role in synaptic function. Although NMDARs are widely expressed in most types of neurons, they play a different role in the cerebellar Purkinje cells (PCs). Utilizing a computational PC model with detailed dendritic morphology, we explored the role of NMDARs at different parts of dendritic branches and regions. We found somatic responses can switch from silent, to simple spikes and complex spikes, depending on specific dendritic branches. Detailed examination of the dendrites regarding their diameters and distance to soma revealed diverse response patterns, yet explain two firing modes, simple and complex spike. Taken together, these results suggest that NMDARs play an important role in controlling excitability sensitivity while taking into account the factor of dendritic properties. Given the complexity of neural morphology varying in cell types, our work suggests that the functional role of NMDARs is not stereotyped but highly interwoven with local properties of neuronal structure."
                    },
                    {
                        "title": "Parallel Spiking Neurons with High Efficiency and Long-term Dependencies Learning Ability",
                        "abstract": "Vanilla spiking neurons in Spiking Neural Networks (SNNs) use charge-\ufb01re-reset neuronal dynamics, which can only be simulated in serial and can hardly learn long-time dependencies. We \ufb01nd that when removing reset, the neuronal dynamics are reformulated in a non-iterative form and can be parallelized. By rewriting neuronal dynamics without resetting to a general formulation, we propose the Parallel Spiking Neuron (PSN), which uses dense connections between time-steps to maximize the utilization of temporal information. To avoid the use of future inputs for low-latency inference, we add masks on the weights and obtain the masked PSN. By sharing weights across time-steps, the sliding PSN is proposed with the ability to deal with sequences with variant lengths. We evaluate the PSN family on simulation speed and temporal/static data classi\ufb01cation, and the results show the overwhelming advantage of the PSN family in ef\ufb01ciency and accuracy. To our best knowledge, this is the \ufb01rst research about parallelizing spiking neurons and can be a cornerstone for the spiking deep learning community. Our codes are available at https://github.com/fangwei123456/Parallel-Spiking-Neuron ."
                    },
                    {
                        "title": "Unveiling the Potential of Spike Streams for Foreground Occlusion Removal from Densely Continuous Views",
                        "abstract": "The extraction of a clean background image by removing foreground occlusion holds immense practical significance, but it also presents several challenges. Presently, the majority of de-occlusion research focuses on addressing this issue through the extraction and synthesis of discrete images from calibrated camera arrays. Nonetheless, the restoration quality tends to suffer when faced with dense occlusions or high-speed motions due to limited perspectives and motion blur. To successfully remove dense foreground occlusion, an effective multi-view visual information integration approach is required. Introducing the spike camera as a novel type of neuromorphic sensor offers promising capabilities with its ultra-high temporal resolution and high dynamic range. In this paper, we propose an innovative solution for tackling the de-occlusion problem through continuous multi-view imaging using only one spike camera without any prior knowledge of camera intrinsic parameters and camera poses. By rapidly moving the spike camera, we continually capture the dense stream of spikes from the occluded scene. To process the spikes, we build a novel model \\textbf{SpkOccNet}, in which we integrate information of spikes from continuous viewpoints within multi-windows, and propose a novel cross-view mutual attention mechanism for effective fusion and refinement. In addition, we contribute the first real-world spike-based dataset \\textbf{S-OCC} for occlusion removal. The experimental results demonstrate that our proposed model efficiently removes dense occlusions in diverse scenes while exhibiting strong generalization."
                    },
                    {
                        "title": "Self-Supervised Joint Dynamic Scene Reconstruction and Optical Flow Estimation for Spiking Camera",
                        "abstract": "Spiking camera, a novel retina-inspired vision sensor, has shown its great potential for capturing high-speed dynamic scenes with a sampling rate of 40,000 Hz. The spiking camera abandons the concept of exposure window, with each of its photosensitive units continuously capturing photons and firing spikes asynchronously. However, the special sampling mechanism prevents the frame-based algorithm from being used to spiking camera. It remains to be a challenge to reconstruct dynamic scenes and perform common computer vision tasks for spiking camera. In this paper, we propose a self-supervised joint learning framework for optical flow estimation and reconstruction of spiking camera. The framework reconstructs clean frame-based spiking representations in a self-supervised manner, and then uses them to train the optical flow networks. We also propose an optical flow based inverse rendering process to achieve self-supervision by minimizing the difference with respect to the original spiking temporal aggregation image. The experimental results demonstrate that our method bridges the gap between synthetic and real-world scenes and achieves desired results in real-world scenarios. To the best of our knowledge, this is the first attempt to jointly reconstruct dynamic scenes and estimate optical flow for spiking camera from a self-supervised learning perspective."
                    },
                    {
                        "title": "Parallel Spiking Neurons with High Efficiency and Ability to Learn Long-term Dependencies",
                        "abstract": "Vanilla spiking neurons in Spiking Neural Networks (SNNs) use charge-fire-reset neuronal dynamics, which can only be simulated serially and can hardly learn long-time dependencies. We find that when removing reset, the neuronal dynamics can be reformulated in a non-iterative form and parallelized. By rewriting neuronal dynamics without reset to a general formulation, we propose the Parallel Spiking Neuron (PSN), which generates hidden states that are independent of their predecessors, resulting in parallelizable neuronal dynamics and extremely high simulation speed. The weights of inputs in the PSN are fully connected, which maximizes the utilization of temporal information. To avoid the use of future inputs for step-by-step inference, the weights of the PSN can be masked, resulting in the masked PSN. By sharing weights across time-steps based on the masked PSN, the sliding PSN is proposed to handle sequences of varying lengths. We evaluate the PSN family on simulation speed and temporal/static data classification, and the results show the overwhelming advantage of the PSN family in efficiency and accuracy. To the best of our knowledge, this is the first study about parallelizing spiking neurons and can be a cornerstone for the spiking deep learning research. Our codes are available at \\url{https://github.com/fangwei123456/Parallel-Spiking-Neuron}."
                    },
                    {
                        "title": "Capture the Moment: High-Speed Imaging With Spiking Cameras Through Short-Term Plasticity",
                        "abstract": "High-speed imaging can help us understand some phenomena that are too fast to be captured by our eyes. Although ultra-fast frame-based cameras (e.g., Phantom) can record millions of fps at reduced resolution, they are too expensive to be widely used. Recently, a retina-inspired vision sensor, spiking camera, has been developed to record external information at 40, 000 Hz. The spiking camera uses the asynchronous binary spike streams to represent visual information. Despite this, how to reconstruct dynamic scenes from asynchronous spikes remains challenging. In this paper, we introduce novel high-speed image reconstruction models based on the short-term plasticity (STP) mechanism of the brain, termed TFSTP and TFMDSTP. We first derive the relationship between states of STP and spike patterns. Then, in TFSTP, by setting up the STP model at each pixel, the scene radiance can be inferred by the states of the models. In TFMDSTP, we use the STP to distinguish the moving and stationary regions, and then use two sets of STP models to reconstruct them respectively. In addition, we present a strategy for correcting error spikes. Experimental results show that the STP-based reconstruction methods can effectively reduce noise with less computing time, and achieve the best performances on both real-world and simulated datasets."
                    }
                ]
            },
            "e53ae901-65ad-4912-85ef-77b91c7da5d3": {
                "pk": "e53ae901-65ad-4912-85ef-77b91c7da5d3",
                "name": "Yijie Peng",
                "collaborators": [
                    "Li Xiao",
                    "B. Heidergott",
                    "L. Hong",
                    "Zewu Ke",
                    "Zeliang Zhang",
                    "H. Lam",
                    "Shuhuai Yang",
                    "J. Hong",
                    "Jinyang Jiang",
                    "Minjie Chen"
                ],
                "domain": [
                    "Neural Networks",
                    "Robustness",
                    "Stochastic Gradient",
                    "Adversarial Attacks"
                ],
                "publications": [
                    {
                        "title": "A Novel Noise Injection-based Training Scheme for Better Model Robustness",
                        "abstract": "Noise injection-based method has been shown to be able to improve the robustness of artificial neural networks in previous work. In this work, we propose a novel noise injection-based training scheme for better model robustness. Specifically, we first develop a likelihood ratio method to estimate the gradient with respect to both synaptic weights and noise levels for stochastic gradient descent training. Then, we design an approximation for the vanilla noise injection-based training method to reduce memory and improve computational efficiency. Next, we apply our proposed scheme to spiking neural networks and evaluate the performance of classification accuracy and robustness on MNIST and Fashion-MNIST datasets. Experiment results show that our proposed method achieves a much better performance on adversarial robustness and slightly better performance on original accuracy, compared with the conventional gradient-based training method."
                    },
                    {
                        "title": "Gradient estimation for smooth stopping criteria",
                        "abstract": "Abstract We establish sufficient conditions for differentiability of the expected cost collected over a discrete-time Markov chain until it enters a given set. The parameter with respect to which differentiability is analysed may simultaneously affect the Markov chain and the set defining the stopping criterion. The general statements on differentiability lead to unbiased gradient estimators."
                    },
                    {
                        "title": "Pathological Image Segmentation with Noisy Labels",
                        "abstract": "Segmentation of pathological images is essential for accurate disease diagnosis. The quality of manual labels plays a critical role in segmentation accuracy; yet, in practice, the labels between pathologists could be inconsistent, thus confusing the training process. In this work, we propose a novel label re-weighting framework to account for the reliability of different experts' labels on each pixel according to its surrounding features. We further devise a new attention heatmap, which takes roughness as prior knowledge to guide the model to focus on important regions. Our approach is evaluated on the public Gleason 2019 datasets. The results show that our approach effectively improves the model's robustness against noisy labels and outperforms state-of-the-art approaches."
                    },
                    {
                        "title": "Noise Optimization for Artificial Neural Networks",
                        "abstract": "Adding noises to artificial neural network(ANN) has been shown to be able to improve robustness in previous work. In this work, we propose a new technique to compute the pathwise stochastic gradient estimate with respect to the standard deviation of the Gaussian noise added to each neuron of the ANN. By our proposed technique, the gradient estimate with respect to noise levels is a byproduct of the backpropagation algorithm for estimating gradient with respect to synaptic weights in ANN. Thus, the noise level for each neuron can be optimized simultaneously in the processing of training the synaptic weights at nearly no extra computational cost. In numerical experiments, our proposed method can achieve significant performance improvement on robustness of several popular ANN structures under both black box and white box attacks tested in various computer vision datasets."
                    },
                    {
                        "title": "A New Likelihood Ratio Method for Training Artificial Neural Networks",
                        "abstract": "We investigate a new approach to compute the gradients of artificial neural networks (ANNs), based on the so-called push-out likelihood ratio method. Unlike the widely used backpropagation (BP) method that requires continuity of the loss function and the activation function, our approach bypasses this requirement by injecting artificial noises into the signals passed along the neurons. We show how this approach has a similar computational complexity as BP, and moreover is more advantageous in terms of removing the backward recursion and eliciting transparent formulas. We also formalize the connection between BP, a pivotal technique for training ANNs, and infinitesimal perturbation analysis, a classic path-wise derivative estimation approach, so that both our new proposed methods and BP can be better understood in the context of stochastic gradient estimation. Our approach allows efficient training for ANNs with more flexibility on the loss and activation functions, and shows empirical improvements on the robustness of ANNs under adversarial attacks and corruptions of natural noises. Summary of Contribution: Stochastic gradient estimation has been studied actively in simulation for decades and becomes more important in the era of machine learning and artificial intelligence. The stochastic gradient descent is a standard technique for training the artificial neural networks (ANNs), a pivotal problem in deep learning. The most popular stochastic gradient estimation technique is the backpropagation method. We find that the backpropagation method lies in the family of infinitesimal perturbation analysis, a path-wise gradient estimation technique in simulation. Moreover, we develop a new likelihood ratio-based method, another popular family of gradient estimation technique in simulation, for training more general ANNs, and demonstrate that the new training method can improve the robustness of the ANN."
                    },
                    {
                        "title": "Training Artificial Neural Networks by Generalized Likelihood Ratio Method: An Effective Way to Improve Robustness",
                        "abstract": "In this work, we proposed a generalized likelihood ratio method capable of training the artificial neural networks with more flexibility: (a)training with discrete activation and loss functions, while the traditional back propagation method cannot train the artificial neural networks with such activations and loss; (b)involving neuronal noises during training and prediction, which will improve the freedom of the model and make it more adaptable to the real environment, especially when environmental noises exist. Numerical results show that the robustness of various artificial neural networks trained by the new method is significantly improved when the input data is affected by both the natural noises and adversarial attacks."
                    },
                    {
                        "title": "optimizing outpatient Department Staffing Level using Multi-Fidelity Models",
                        "abstract": "The workload of the outpatient departments in Chinese large hospitals is extremely high. Patients often have to wait for a long time before getting their treatments. It is economically expensive to increase medical staffs including nurses and doctors. Therefore, it is critical to optimize staff planning in the outpatient departments to reduce excessive patient waiting time. A high-fidelity simulation model can accurately capture the features of the outpatient service system. But it is very time-consuming to obtain the optimal staff planning decision only based on the simulation model. A simplified queueing model might lead to an analytical solution for the optimal staff planning problem, but it can not fully capture the feature of the real outpatient service system. We propose to use the outputs of the high-fidelity simulation model to drive the output of the low-fidelity queueing model closer to that of the outpatient service system, and then use the data-driven queueing model to make the staff planning decision. Empirical studies on a major hospital are carried out, which demonstrate the effectiveness and efficiency of our method."
                    },
                    {
                        "title": "Training Artificial Neural Networks by Generalized Likelihood Ratio Method: Exploring Brain-like Learning to Improve Adversarial Defensiveness",
                        "abstract": "Recent work in deep learning has shown that the artificial neural networks are vulnerable to adversarial attacks, where a very small perturbation of the inputs can drastically alter the classification result. In this work, we propose a generalized likelihood ratio method capable of training the artificial neural networks with some biological brain-like mechanisms,.e.g., (a) learning by the loss value, (b) learning via neurons with discontinuous activation and loss functions. The traditional back propagation method cannot train the artificial neural networks with aforementioned brain-like learning mechanisms. Numerical results show that various artificial neural networks trained by the new method can significantly improve the defensiveness against the adversarial attacks. Code is available: \\url{this https URL} ."
                    },
                    {
                        "title": "Training Artificial Neural Networks by Generalized Likelihood Ratio Method: Exploring Brain-like Learning to Improve Robustness",
                        "abstract": "In this work, we propose a generalized likelihood ratio method capable of training the artificial neural networks with some biological brain-like mechanisms,.e.g., (a) learning by the loss value, (b) learning via neurons with discontinuous activation and loss functions. The traditional back propagation method cannot train the artificial neural networks with aforementioned brain-like learning mechanisms. Numerical results show that the robustness of various artificial neural networks trained by the new method is significantly improved when the input data is affected by both the natural noises and adversarial attacks. Code is available: \\url{this https URL} ."
                    },
                    {
                        "title": "Stochastic Gradient Estimation for Artificial Neural Networks",
                        "abstract": "In this paper, we consider artificial neural networks (ANN) with processing noise. We derive an infinitesimal perturbation analysis (IPA) and likelihood ratio (LR) gradient estimator for this type of ANN. We show that the IPA estimator, though mathematically equivalent to the backpropagation algorithm, has higher computational complexity than the backpropagation algorithm. To overcome the obstacle posed to IPA and standard stochastic gradient methods by discontinuous activation and loss functions, we introduce in this paper a push-out LR estimator. The push-out LR gradient estimator is computationally more efficient than the backpropagation algorithm, and it avoids differentiating the actual ANN. This allows for discontinuous activation and loss functions to be handled by our LR estimator. We implement the developed stochastic gradient estimation techniques in an example on identifying images of numbers from 0 to 9. The numerical results substantiate the effectiveness of the proposed schemes."
                    }
                ]
            }
        }
    },
    "2406.04845": {
        "paper_data": {
            "title": "FedLLM-Bench: Realistic Benchmarks for Federated Learning of Large Language Models",
            "url": "http://arxiv.org/abs/2406.04845v1",
            "arxiv_id": "2406.04845",
            "authors": [
                "Rui Ye",
                "Rui Ge",
                "Xinyu Zhu",
                "Jingyi Chai",
                "Yaxin Du",
                "Yang Liu",
                "Yanfeng Wang",
                "Siheng Chen"
            ],
            "abstract": "Federated learning has enabled multiple parties to collaboratively train large language models without directly sharing their data (FedLLM). Following this training paradigm, the community has put massive efforts from diverse aspects including framework, performance, and privacy. However, an unpleasant fact is that there are currently no realistic datasets and benchmarks for FedLLM and previous works all rely on artificially constructed datasets, failing to capture properties in real-world scenarios. Addressing this, we propose FedLLM-Bench, which involves 8 training methods, 4 training datasets, and 6 evaluation metrics, to offer a comprehensive testbed for the FedLLM community. FedLLM-Bench encompasses three datasets (e.g., user-annotated multilingual dataset) for federated instruction tuning and one dataset (e.g., user-annotated preference dataset) for federated preference alignment, whose scale of client number ranges from 38 to 747. Our datasets incorporate several representative diversities: language, quality, quantity, instruction, length, embedding, and preference, capturing properties in real-world scenarios. Based on FedLLM-Bench, we conduct experiments on all datasets to benchmark existing FL methods and provide empirical insights (e.g., multilingual collaboration). We believe that our FedLLM-Bench can benefit the FedLLM community by reducing required efforts, providing a practical testbed, and promoting fair comparisons. Code and datasets are available at https://github.com/rui-ye/FedLLM-Bench.",
            "introduction": "   1 Introduction  Large language models (LLMs) have achieved unprecedented success in diverse domains\u00a0[1, 2, 3, 4, 5, 6]. These LLMs are usually trained by centralized learning paradigm, where various parties individually collect massive data for model training. In this case, the data amount of each individual party is hard to scale due to the high cost of collecting and annotating data. However, their data cannot be directly shared for collaboration due to property and privacy issues.   To relieve the required cost of each party, federated learning (FL)\u00a0[7, 8] has emerged as a sound and off-the-shelf technique to facilitate collaboration, which leverages decentralized language data to collaboratively train LLMs in a privacy-preserving way (FedLLM)\u00a0[9, 10, 11]. To facilitate the development of FedLLM, there has been a series of code frameworks such as OpenFedLLM\u00a0[9], FederatedScope-LLM\u00a0[10], and FedML-LLM\u00a0[11]; and many methods that tackle the issues of data quality\u00a0[12], intellectual property protection\u00a0[13], privacy\u00a0[14], limited resources\u00a0[15] in FedLLM.   Despite that massive efforts have been made, one significant concern remains: there is currently no realistic benchmark for FedLLM, making it hard to practically evaluate the effectiveness of FL methods in real-world scenarios. In such context, each previous work constructs its own FL datasets by artificially partitioning existing centralized datasets\u00a0[9, 10, 13], falling short of capturing the natural properties existed in real-world cross-user datasets\u00a0[16, 17]. Even worse, these papers often follow different training and evaluation setups, which significantly increases the difficulty of re-implementations and risk of unfair comparisons\u00a0[18, 14].   To fill this gap, we propose the first realistic benchmark for FedLLM termed FedLLM-Bench, offering a comprehensive testbed for the FedLLM community. FedLLM-Bench encompasses three datasets for federated instruction tuning (including one user-annotated multilingual dataset: Fed-Aya, and two datasets with realistic user instructions: Fed-WildChat and Fed-ChatbotIT) and one dataset (user-annotated preference dataset: Fed-ChatbotPA) for federated preference alignment. These datasets are all naturally split by real-world user ID with the scale ranging from 38 to 747 clients, therefore exhibiting realistic federated properties (especially for cross-device setup in FL where data are partitioned by user devices). Specifically, datasets in our FedLLM-Bench inherit the following diversities (Table\u00a01): (1) Language: clients\u2019 datasets (e.g., our Fed-Aya dataset) cover data from diverse languages, modeling the real-world scenarios of multilingual collaboration. (2) Quality and Quantity: the quality and quantity of clients\u2019 datasets vary across each other, which is a common property in real-world applications. (3) Length: the sequence length of clients\u2019 data could be quite different, representing a new type of data heterogeneity in FL. (4) Preference: different clients have different preferences as verified by different preferred instructions in instruction tuning datasets (e.g., Fed-WildChat) and different preferred responses in preference alignment dataset (i.e., Fed-ChatbotPA), mirroring the complexities of real-world data scenarios. These diversities make our FedLLM-Bench a comprehensive benchmark in the era of FedLLM, serving as a great successor to representative benchmarks for classical tasks such as LEAF\u00a0[19] benchmark.   Based on these datasets, we implement 8 representative baseline methods and 6 evaluation metrics, and conduct extensive experiments. Our experiments mainly demonstrate (1) that federated learning can consistently bring performance gain compared to local training without collaboration; and (2) the performance ranking of several representative baseline methods. Besides serving as a benchmark for performance comparison, our FedLLM-Bench can",
            "references": [
                {
                    "title": "Locally Estimated Global Perturbations are Better than Local Perturbations for Federated Sharpness-aware Minimization",
                    "abstract": "In federated learning (FL), the multi-step update and data heterogeneity among clients often lead to a loss landscape with sharper minima, degenerating the performance of the resulted global model. Prevalent federated approaches incorporate sharpness-aware minimization (SAM) into local training to mitigate this problem. However, the local loss landscapes may not accurately reflect the flatness of global loss landscape in heterogeneous environments; as a result, minimizing local sharpness and calculating perturbations on client data might not align the efficacy of SAM in FL with centralized training. To overcome this challenge, we propose FedLESAM, a novel algorithm that locally estimates the direction of global perturbation on client side as the difference between global models received in the previous active and current rounds. Besides the improved quality, FedLESAM also speed up federated SAM-based approaches since it only performs once backpropagation in each iteration. Theoretically, we prove a slightly tighter bound than its original FedSAM by ensuring consistent perturbation. Empirically, we conduct comprehensive experiments on four federated benchmark datasets under three partition strategies to demonstrate the superior performance and efficiency of FedLESAM."
                },
                {
                    "title": "Federated Learning with Bilateral Curation for Partially Class-Disjoint Data",
                    "abstract": "Partially class-disjoint data (PCDD), a common yet under-explored data formation where each client contributes a part of classes (instead of all classes) of samples, severely challenges the performance of federated algorithms. Without full classes, the local objective will contradict the global objective, yielding the angle collapse problem for locally missing classes and the space waste problem for locally existing classes. As far as we know, none of the existing methods can intrinsically mitigate PCDD challenges to achieve holistic improvement in the bilateral views (both global view and local view) of federated learning. To address this dilemma, we are inspired by the strong generalization of simplex Equiangular Tight Frame~(ETF) on the imbalanced data, and propose a novel approach called FedGELA where the classifier is globally fixed as a simplex ETF while locally adapted to the personal distributions. Globally, FedGELA provides fair and equal discrimination for all classes and avoids inaccurate updates of the classifier, while locally it utilizes the space of locally missing classes for locally existing classes. We conduct extensive experiments on a range of datasets to demonstrate that our FedGELA achieves promising performance~(averaged improvement of 3.9% to FedAvg and 1.5% to best baselines) and provide both local and global convergence guarantees. Source code is available at:https://github.com/MediaBrain-SJTU/FedGELA.git."
                },
                {
                    "title": "RFLPA: A Robust Federated Learning Framework against Poisoning Attacks with Secure Aggregation",
                    "abstract": "Federated learning (FL) allows multiple devices to train a model collaboratively without sharing their data. Despite its benefits, FL is vulnerable to privacy leakage and poisoning attacks. To address the privacy concern, secure aggregation (SecAgg) is often used to obtain the aggregation of gradients on sever without inspecting individual user updates. Unfortunately, existing defense strategies against poisoning attacks rely on the analysis of local updates in plaintext, making them incompatible with SecAgg. To reconcile the conflicts, we propose a robust federated learning framework against poisoning attacks (RFLPA) based on SecAgg protocol. Our framework computes the cosine similarity between local updates and server updates to conduct robust aggregation. Furthermore, we leverage verifiable packed Shamir secret sharing to achieve reduced communication cost of $O(M+N)$ per user, and design a novel dot-product aggregation algorithm to resolve the issue of increased information leakage. Our experimental results show that RFLPA significantly reduces communication and computation overhead by over $75\\%$ compared to the state-of-the-art method, BREA, while maintaining competitive accuracy."
                },
                {
                    "title": "WildChat: 1M ChatGPT Interaction Logs in the Wild",
                    "abstract": "Chatbots such as GPT-4 and ChatGPT are now serving millions of users. Despite their widespread use, there remains a lack of public datasets showcasing how these tools are used by a population of users in practice. To bridge this gap, we offered free access to ChatGPT for online users in exchange for their affirmative, consensual opt-in to anonymously collect their chat transcripts and request headers. From this, we compiled WildChat, a corpus of 1 million user-ChatGPT conversations, which consists of over 2.5 million interaction turns. We compare WildChat with other popular user-chatbot interaction datasets, and find that our dataset offers the most diverse user prompts, contains the largest number of languages, and presents the richest variety of potentially toxic use-cases for researchers to study. In addition to timestamped chat transcripts, we enrich the dataset with demographic data, including state, country, and hashed IP addresses, alongside request headers. This augmentation allows for more detailed analysis of user behaviors across different geographical regions and temporal dimensions. Finally, because it captures a broad range of use cases, we demonstrate the dataset's potential utility in fine-tuning instruction-following models. WildChat is released at https://wildchat.allen.ai under AI2 ImpACT Licenses."
                },
                {
                    "title": "Automated Federated Pipeline for Parameter-Efficient Fine-Tuning of Large Language Models",
                    "abstract": "Recently, there has been a surge in the development of advanced intelligent generative content (AIGC), especially large language models (LLMs). However, for many downstream tasks, it is necessary to fine-tune LLMs using private data. While federated learning offers a promising privacy-preserving solution to LLM fine-tuning, the substantial size of an LLM, combined with high computational and communication demands, makes it hard to apply to downstream tasks. More importantly, private edge servers often possess varying computing and network resources in real-world scenarios, introducing additional complexities to LLM fine-tuning. To tackle these problems, we design and implement an automated federated pipeline, named FedPipe, to fine-tune LLMs with minimal training cost but without adding any inference latency. FedPipe firstly identifies the weights to be fine-tuned based on their contributions to the LLM training. It then configures a low-rank adapter for each selected weight to train local low-rank adapters on an edge server, and aggregate local adapters of all edge servers to fine-tune the whole LLM. Finally, it appropriately quantizes the parameters of LLM to reduce memory space according to the requirements of edge servers. Extensive experiments demonstrate that FedPipe expedites the model training and achieves higher accuracy than state-of-the-art benchmarks."
                },
                {
                    "title": "An Upload-Efficient Scheme for Transferring Knowledge From a Server-Side Pre-trained Generator to Clients in Heterogeneous Federated Learning",
                    "abstract": "Heterogeneous Federated Learning (HtFL) enables collaborative learning on multiple clients with different model architectures while preserving privacy. Despite recent research progress, knowledge sharing in HtFL is still difficult due to data and model heterogeneity. To tackle this issue, we leverage the knowledge stored in public pretrained generators and propose a new upload-efficient knowledge transfer scheme called Federated Knowledge-Transfer Loop (Fed-KTL). Our FedKTL can produce client-task-related prototypical image-vector pairs via the generator's inference on the server. With these pairs, each client can transfer preexisting knowledge from the generator to its local model through an additional supervised local task. We conduct extensive experiments on four datasets under two types of data heterogeneity with 14 kinds of models including CNNs and ViTs. Results show that our upload-efficient FedKTL surpasses seven state-of-the-art methods by up to 7.31% in accuracy. Moreover, our knowledge transfer scheme is applicable in scenarios with only one edge client. Code: https://github.com/TsingZ0/FedKTL"
                },
                {
                    "title": "Improving LoRA in Privacy-preserving Federated Learning",
                    "abstract": "Low-rank adaptation (LoRA) is one of the most popular task-specific parameter-efficient fine-tuning (PEFT) methods on pre-trained language models for its good performance and computational efficiency. LoRA injects a product of two trainable rank decomposition matrices over the top of each frozen pre-trained model module. However, when applied in the setting of privacy-preserving federated learning (FL), LoRA may become unstable due to the following facts: 1) the effects of data heterogeneity and multi-step local updates are non-negligible, 2) additive noise enforced on updating gradients to guarantee differential privacy (DP) can be amplified and 3) the final performance is susceptible to hyper-parameters. A key factor leading to these phenomena is the discordance between jointly optimizing the two low-rank matrices by local clients and separately aggregating them by the central server. Thus, this paper proposes an efficient and effective version of LoRA, Federated Freeze A LoRA (FFA-LoRA), to alleviate these challenges and further halve the communication cost of federated fine-tuning LLMs. The core idea of FFA-LoRA is to fix the randomly initialized non-zero matrices and only fine-tune the zero-initialized matrices. Compared to LoRA, FFA-LoRA is motivated by practical and theoretical benefits in privacy-preserved FL. Our experiments demonstrate that FFA-LoRA provides more consistent performance with better computational efficiency over vanilla LoRA in various FL tasks."
                },
                {
                    "title": "FewFedPIT: Towards Privacy-preserving and Few-shot Federated Instruction Tuning",
                    "abstract": "Instruction tuning has been identified as a crucial technique for optimizing the performance of large language models (LLMs) in generating human-aligned responses. Nonetheless, gathering diversified and superior-quality instruction data for such tuning presents notable obstacles, especially in domains with rigid privacy provisions. Federated instruction tuning (FedIT) has emerged as a promising solution, by consolidating collaborative training across multiple data owners, thereby resulting in a privacy-preserving learning model. However, FedIT encounters limitations such as scarcity of instructional data and risk of exposure to training data extraction attacks. In this paper, we propose a novel federated algorithm, FewFedPIT, designed to simultaneously enhance privacy protection and model performance of federated few-shot learning. FewFedPITcomprises three vital components on the client side: (1) synthetic data generation, which utilizes LLMs' in-context learning capacity to generate synthetic data autonomously, thus expanding the local database; (2) parameter isolation training, which individually updates the public parameters in the synthetic data and the private parameters in the local data, consequently mitigating the noise impact of the synthetic data; (3) local aggregation sharing, which mixes public and private parameters before uploading, effectively preventing data extraction attacks. Extensive experiments on three open-source datasets demonstrate the effectiveness of FewFedPITin, enhancing privacy preservation and improving federated few-shot performance."
                },
                {
                    "title": "Enhancing Data Quality in Federated Fine-Tuning of Foundation Models",
                    "abstract": "In the current landscape of foundation model training, there is a significant reliance on public domain data, which is nearing exhaustion according to recent research. To further scale up, it is crucial to incorporate collaboration among multiple specialized and high-quality private domain data sources. However, the challenge of training models locally without sharing private data presents numerous obstacles in data quality control. To tackle this issue, we propose a data quality control pipeline for federated fine-tuning of foundation models. This pipeline computes scores reflecting the quality of training data and determines a global threshold for a unified standard, aiming for improved global performance. Our experiments show that the proposed quality control pipeline facilitates the effectiveness and reliability of the model training, leading to better performance."
                },
                {
                    "title": "OpenFedLLM: Training Large Language Models on Decentralized Private Data via Federated Learning",
                    "abstract": "Trained on massive publicly available data, large language models (LLMs) have demonstrated tremendous success across various fields. While more data contributes to better performance, a disconcerting reality is that high-quality public data will be exhausted in a few years. In this paper, we offer a potential next step for contemporary LLMs: collaborative and privacy-preserving LLM training on the underutilized distributed private data via federated learning (FL), where multiple data owners collaboratively train a shared model without transmitting raw data. To achieve this, we build a concise, integrated, and research-friendly framework/codebase, named OpenFedLLM. It covers federated instruction tuning for enhancing instruction-following capability, federated value alignment for aligning with human values, and 7 representative FL algorithms. Besides, OpenFedLLM supports training on diverse domains, where we cover 8 training datasets; and provides comprehensive evaluations, where we cover 30+ evaluation metrics. Through extensive experiments, we observe that all FL algorithms outperform local training on training LLMs, demonstrating a clear performance improvement across a variety of settings. Notably, in a financial benchmark, Llama2-7B fine-tuned by applying any FL algorithm can outperform GPT-4 by a significant margin while the model obtained through individual training cannot, demonstrating strong motivation for clients to participate in FL. The code is available at https://github.com/rui-ye/OpenFedLLM."
                },
                {
                    "title": "Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuning",
                    "abstract": "Datasets are foundational to many breakthroughs in modern artificial intelligence. Many recent achievements in the space of natural language processing (NLP) can be attributed to the finetuning of pre-trained models on a diverse set of tasks that enables a large language model (LLM) to respond to instructions. Instruction fine-tuning (IFT) requires specifically constructed and annotated datasets. However, existing datasets are almost all in the English language. In this work, our primary goal is to bridge the language gap by building a human-curated instruction-following dataset spanning 65 languages. We worked with fluent speakers of languages from around the world to collect natural instances of instructions and completions. Furthermore, we create the most extensive multilingual collection to date, comprising 513 million instances through templating and translating existing datasets across 114 languages. In total, we contribute four key resources: we develop and open-source the Aya Annotation Platform, the Aya Dataset, the Aya Collection, and the Aya Evaluation Suite. The Aya initiative also serves as a valuable case study in participatory research, involving collaborators from 119 countries. We see this as a valuable framework for future research collaborations that aim to bridge gaps in resources."
                },
                {
                    "title": "Superfiltering: Weak-to-Strong Data Filtering for Fast Instruction-Tuning",
                    "abstract": "Instruction tuning is critical to improve LLMs but usually suffers from low-quality and redundant data. Data filtering for instruction tuning has proved important in improving both the efficiency and performance of the tuning process. But it also leads to extra cost and computation due to the involvement of LLMs in this process. To reduce the filtering cost, we study Superfiltering: Can we use a smaller and weaker model to select data for finetuning a larger and stronger model? Despite the performance gap between weak and strong language models, we find their highly consistent capability to perceive instruction difficulty and data selection results. This enables us to use a much smaller and more efficient model to filter the instruction data used to train a larger language model. Not only does it largely speed up the data filtering, but the filtered-data-finetuned LLM achieves even better performance on standard benchmarks. Extensive experiments validate the efficacy and efficiency of our approach."
                },
                {
                    "title": "FedTGP: Trainable Global Prototypes with Adaptive-Margin-Enhanced Contrastive Learning for Data and Model Heterogeneity in Federated Learning",
                    "abstract": "Recently, Heterogeneous Federated Learning (HtFL) has attracted attention due to its ability to support heterogeneous models and data. To reduce the high communication cost of transmitting model parameters, a major challenge in HtFL, prototype-based HtFL methods are proposed to solely share class representatives, a.k.a, prototypes, among heterogeneous clients while maintaining the privacy of clients\u2019 models. However, these prototypes are naively aggregated into global prototypes on the server using weighted averaging, resulting in suboptimal global knowledge which negatively impacts the performance of clients. To overcome this challenge, we introduce a novel HtFL approach called FedTGP, which leverages our Adaptive-margin-enhanced Contrastive Learning (ACL) to learn Trainable Global Prototypes (TGP) on the server. By incorporating ACL, our approach enhances prototype separability while preserving semantic meaning. Extensive experiments with twelve heterogeneous models demonstrate that our FedTGP surpasses state-of-the-art methods by up to 9.08% in accuracy while maintaining the communication and privacy advantages of prototype-based HtFL. Our code is available at https://github.com/TsingZ0/FedTGP."
                },
                {
                    "title": "DB-GPT: Empowering Database Interactions with Private Large Language Models",
                    "abstract": "The recent breakthroughs in large language models (LLMs) are positioned to transition many areas of software. Database technologies particularly have an important entanglement with LLMs as efficient and intuitive database interactions are paramount. In this paper, we present DB-GPT, a revolutionary and production-ready project that integrates LLMs with traditional database systems to enhance user experience and accessibility. DB-GPT is designed to understand natural language queries, provide context-aware responses, and generate complex SQL queries with high accuracy, making it an indispensable tool for users ranging from novice to expert. The core innovation in DB-GPT lies in its private LLM technology, which is fine-tuned on domain-specific corpora to maintain user privacy and ensure data security while offering the benefits of state-of-the-art LLMs. We detail the architecture of DB-GPT, which includes a novel retrieval augmented generation (RAG) knowledge system, an adaptive learning mechanism to continuously improve performance based on user feedback and a service-oriented multi-model framework (SMMF) with powerful data-driven agents. Our extensive experiments and user studies confirm that DB-GPT represents a paradigm shift in database interactions, offering a more natural, efficient, and secure way to engage with data repositories. The paper concludes with a discussion of the implications of DB-GPT framework on the future of human-database interaction and outlines potential avenues for further enhancements and applications in the field. The project code is available at https://github.com/eosphoros-ai/DB-GPT. Experience DB-GPT for yourself by installing it with the instructions https://github.com/eosphoros-ai/DB-GPT#install and view a concise 10-minute video at https://www.youtube.com/watch?v=KYs4nTDzEhk."
                },
                {
                    "title": "What Makes Good Data for Alignment? A Comprehensive Study of Automatic Data Selection in Instruction Tuning",
                    "abstract": "Instruction tuning is a standard technique employed to align large language models to end tasks and user preferences after the initial pretraining phase. Recent research indicates the critical role of data engineering in instruction tuning -- when appropriately selected, only limited data is necessary to achieve superior performance. However, we still lack a principled understanding of what makes good instruction tuning data for alignment, and how we should select data automatically and effectively. In this work, we delve deeply into automatic data selection strategies for alignment. We start with controlled studies to measure data across three dimensions: complexity, quality, and diversity, along which we examine existing methods and introduce novel techniques for enhanced data measurement. Subsequently, we propose a simple strategy to select data samples based on the measurement. We present deita (short for Data-Efficient Instruction Tuning for Alignment), a series of models fine-tuned from LLaMA and Mistral models using data samples automatically selected with our proposed approach. Empirically, deita performs better or on par with the state-of-the-art open-source alignment models with only 6K SFT training data samples -- over 10x less than the data used in the baselines. When further trained with direct preference optimization (DPO), deita-Mistral-7B + DPO trained with 6K SFT and 10K DPO samples achieve 7.55 MT-Bench and 90.06% AlpacaEval scores. We anticipate this work to provide tools on automatic data selection, facilitating data-efficient alignment. We release our models as well as the selected datasets for future researches to effectively align models more efficiently."
                },
                {
                    "title": "The Past, Present and Better Future of Feedback Learning in Large Language Models for Subjective Human Preferences and Values",
                    "abstract": "Human feedback is increasingly used to steer the behaviours of Large Language Models (LLMs). However, it is unclear how to collect and incorporate feedback in a way that is efficient, effective and unbiased, especially for highly subjective human preferences and values. In this paper, we survey existing approaches for learning from human feedback, drawing on 95 papers primarily from the ACL and arXiv repositories.First, we summarise the past, pre-LLM trends for integrating human feedback into language models. Second, we give an overview of present techniques and practices, as well as the motivations for using feedback; conceptual frameworks for defining values and preferences; and how feedback is collected and from whom. Finally, we encourage a better future of feedback learning in LLMs by raising five unresolved conceptual and practical challenges."
                },
                {
                    "title": "Mistral 7B",
                    "abstract": "We introduce Mistral 7B v0.1, a 7-billion-parameter language model engineered for superior performance and efficiency. Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation. Our model leverages grouped-query attention (GQA) for faster inference, coupled with sliding window attention (SWA) to effectively handle sequences of arbitrary length with a reduced inference cost. We also provide a model fine-tuned to follow instructions, Mistral 7B -- Instruct, that surpasses the Llama 2 13B -- Chat model both on human and automated benchmarks. Our models are released under the Apache 2.0 license."
                },
                {
                    "title": "FederatedScope-LLM: A Comprehensive Package for Fine-tuning Large Language Models in Federated Learning",
                    "abstract": "LLMs have demonstrated great capabilities in various NLP tasks. Different entities can further improve the performance of those LLMs on their specific downstream tasks by fine-tuning LLMs. When several entities have similar interested tasks, but their data cannot be shared because of privacy concerns regulations, federated learning (FL) is a mainstream solution to leverage the data of different entities. However, fine-tuning LLMs in federated learning settings still lacks adequate support from existing FL frameworks because it has to deal with optimizing the consumption of significant communication and computational resources, data preparation for different tasks, and distinct information protection demands. This paper first discusses these challenges of federated fine-tuning LLMs, and introduces our package FS-LLM as a main contribution, which consists of the following components: (1) we build an end-to-end benchmarking pipeline, automizing the processes of dataset preprocessing, federated fine-tuning execution, and performance evaluation on federated LLM fine-tuning; (2) we provide comprehensive federated parameter-efficient fine-tuning algorithm implementations and versatile programming interfaces for future extension in FL scenarios with low communication and computation costs, even without accessing the full model; (3) we adopt several accelerating and resource-efficient operators for fine-tuning LLMs with limited resources and the flexible pluggable sub-routines for interdisciplinary study. We conduct extensive experiments to validate the effectiveness of FS-LLM and benchmark advanced LLMs with state-of-the-art parameter-efficient fine-tuning algorithms in FL settings, which also yields valuable insights into federated fine-tuning LLMs for the research community. To facilitate further research and adoption, we release FS-LLM at https://github.com/alibaba/FederatedScope/tree/llm."
                },
                {
                    "title": "From Quantity to Quality: Boosting LLM Performance with Self-Guided Data Selection for Instruction Tuning",
                    "abstract": "In the realm of Large Language Models (LLMs), the balance between instruction data quality and quantity is a focal point. Recognizing this, we introduce a self-guided methodology for LLMs to autonomously discern and select cherry samples from open-source datasets, effectively minimizing manual curation and potential cost for instruction tuning an LLM. Our key innovation, the Instruction-Following Difficulty (IFD) metric, emerges as a pivotal metric to identify discrepancies between a model\u2019s expected responses and its intrinsic generation capability. Through the application of IFD, cherry samples can be pinpointed, leading to a marked uptick in model training efficiency. Empirical validations on datasets like Alpaca and WizardLM underpin our findings; with a mere 10% of original data input, our strategy showcases improved results. This synthesis of self-guided cherry-picking and the IFD metric signifies a transformative leap in the instruction tuning of LLMs, promising both efficiency and resource-conscious advancements. Codes, data, and models are available."
                },
                {
                    "title": "AgentVerse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviors",
                    "abstract": "Autonomous agents empowered by Large Language Models (LLMs) have undergone significant improvements, enabling them to generalize across a broad spectrum of tasks. However, in real-world scenarios, cooperation among individuals is often required to enhance the efficiency and effectiveness of task accomplishment. Hence, inspired by human group dynamics, we propose a multi-agent framework \\framework that can collaboratively and dynamically adjust its composition as a greater-than-the-sum-of-its-parts system. Our experiments demonstrate that \\framework framework can effectively deploy multi-agent groups that outperform a single agent. Furthermore, we delve into the emergence of social behaviors among individual agents within a group during collaborative task accomplishment. In view of these behaviors, we discuss some possible strategies to leverage positive ones and mitigate negative ones for improving the collaborative potential of multi-agent groups. Our codes for \\framework will soon be released at \\url{https://github.com/OpenBMB/AgentVerse}."
                },
                {
                    "title": "MetaGPT: Meta Programming for Multi-Agent Collaborative Framework",
                    "abstract": "Recently, remarkable progress has been made in automated task-solving through the use of multi-agent driven by large language models (LLMs). However, existing LLM-based multi-agent works primarily focus on solving simple dialogue tasks, and complex tasks are rarely studied, mainly due to the LLM hallucination problem. This type of hallucination becomes cascading when naively chaining multiple intelligent agents, resulting in a failure to effectively address complex problems. Therefore, we introduce MetaGPT, an innovative framework that incorporates efficient human workflows as a meta programming approach into LLM-based multi-agent collaboration. Specifically, MetaGPT encodes Standardized Operating Procedures (SOPs) into prompts to enhance structured coordination. Subsequently, it mandates modular outputs, empowering agents with domain expertise comparable to human professionals, to validate outputs and minimize compounded errors. In this way, MetaGPT leverages the assembly line paradigm to assign diverse roles to various agents, thereby establishing a framework that can effectively and cohesively deconstruct complex multi-agent collaborative problems. Our experiments on collaborative software engineering benchmarks demonstrate that MetaGPT generates more coherent and correct solutions compared to existing chat-based multi-agent systems. This highlights the potential of integrating human domain knowledge into multi-agent systems, thereby creating new opportunities to tackle complex real-world challenges. The GitHub repository of this project is publicly available on:https://github.com/geekan/MetaGPT."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "FedDefender: Client-Side Attack-Tolerant Federated Learning",
                    "abstract": "Federated learning enables learning from decentralized data sources without compromising privacy, which makes it a crucial technique. However, it is vulnerable to model poisoning attacks, where malicious clients interfere with the training process. Previous defense mechanisms have focused on the server-side by using careful model aggregation, but this may not be effective when the data is not identically distributed or when attackers can access the information of benign clients. In this paper, we propose a new defense mechanism that focuses on the client-side, called FedDefender, to help benign clients train robust local models and avoid the adverse impact of malicious model updates from attackers, even when a server-side defense cannot identify or remove adversaries. Our method consists of two main components: (1) attack-tolerant local meta update and (2) attack-tolerant global knowledge distillation. These components are used to find noise-resilient model parameters while accurately extracting knowledge from a potentially corrupted global model. Our client-side defense strategy has a flexible structure and can work in conjunction with any existing server-side strategies. Evaluations of real-world scenarios across multiple datasets show that the proposed method enhances the robustness of federated learning against model poisoning attacks."
                },
                {
                    "title": "BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset",
                    "abstract": "In this paper, we introduce the \\textsc{BeaverTails} dataset, aimed at fostering research on safety alignment in large language models (LLMs). This dataset uniquely separates annotations of helpfulness and harmlessness for question-answering pairs, thus offering distinct perspectives on these crucial attributes. In total, we have gathered safety meta-labels for 30,207 question-answer (QA) pairs and 30,144 pairs of expert comparison data for both the helpfulness and harmlessness metrics. In total, we have gathered safety meta-labels for 333,963 question-answer (QA) pairs and 361,903 pairs of expert comparison data for both the helpfulness and harmlessness metrics. We further showcase applications of BeaverTails in content moderation and reinforcement learning with human feedback (RLHF), emphasizing its potential for practical safety measures in LLMs. We believe this dataset provides vital resources for the community, contributing towards the safe development and deployment of LLMs. Our project page is available at the following URL: https://sites.google.com/view/pku-beavertails. Warning: this paper contains example data that may be offensive or harmful."
                },
                {
                    "title": "Judging LLM-as-a-judge with MT-Bench and Chatbot Arena",
                    "abstract": "Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge."
                },
                {
                    "title": "InstructEval: Towards Holistic Evaluation of Instruction-Tuned Large Language Models",
                    "abstract": "Instruction-tuned large language models have revolutionized natural language processing and have shown great potential in applications such as conversational agents. These models, such as GPT-4, can not only master language but also solve complex tasks in areas like mathematics, coding, medicine, and law. However, there is still a lack of comprehensive understanding regarding their full potential, primarily due to the black-box nature of many models and lack of holistic evaluation. To address these challenges, we present InstructEval, a more comprehensive evaluation suite designed specifically for instruction-tuned large language models. Unlike previous works, our evaluation involves a rigorous assessment of models based on problem-solving, writing ability, and alignment to human values. We take a holistic approach to analyze various factors affecting model performance, including the pretraining foundation, instruction-tuning data, and training methods. Our findings reveal that the quality of instruction data is a crucial factor in scaling model performance. While open-source models demonstrate impressive writing abilities, there is substantial room for improvement in problem-solving and alignment."
                },
                {
                    "title": "FedDisco: Federated Learning with Discrepancy-Aware Collaboration",
                    "abstract": "This work considers the category distribution heterogeneity in federated learning. This issue is due to biased labeling preferences at multiple clients and is a typical setting of data heterogeneity. To alleviate this issue, most previous works consider either regularizing local models or fine-tuning the global model, while they ignore the adjustment of aggregation weights and simply assign weights based on the dataset size. However, based on our empirical observations and theoretical analysis, we find that the dataset size is not optimal and the discrepancy between local and global category distributions could be a beneficial and complementary indicator for determining aggregation weights. We thus propose a novel aggregation method, Federated Learning with Discrepancy-aware Collaboration (FedDisco), whose aggregation weights not only involve both the dataset size and the discrepancy value, but also contribute to a tighter theoretical upper bound of the optimization error. FedDisco also promotes privacy-preservation, communication and computation efficiency, as well as modularity. Extensive experiments show that our FedDisco outperforms several state-of-the-art methods and can be easily incorporated with many existing methods to further enhance the performance. Our code will be available at https://github.com/MediaBrain-SJTU/FedDisco."
                },
                {
                    "title": "Direct Preference Optimization: Your Language Model is Secretly a Reward Model",
                    "abstract": "While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train."
                },
                {
                    "title": "WizardLM: Empowering Large Language Models to Follow Complex Instructions",
                    "abstract": "Training large language models (LLMs) with open-domain instruction following data brings colossal success. However, manually creating such instruction data is very time-consuming and labor-intensive. Moreover, humans may struggle to produce high-complexity instructions. In this paper, we show an avenue for creating large amounts of instruction data with varying levels of complexity using LLM instead of humans. Starting with an initial set of instructions, we use our proposed Evol-Instruct to rewrite them step by step into more complex instructions. Then, we mix all generated instruction data to fine-tune LLaMA. We call the resulting model WizardLM. Human evaluations on a complexity-balanced test bed and Vicuna's testset show that instructions from Evol-Instruct are superior to human-created ones. By analyzing the human evaluation results of the high complexity part, we demonstrate that outputs from our WizardLM are preferred to outputs from OpenAI ChatGPT. In GPT-4 automatic evaluation, WizardLM achieves more than 90\\% capacity of ChatGPT on 17 out of 29 skills. Even though WizardLM still lags behind ChatGPT in some aspects, our findings suggest that fine-tuning with AI-evolved instructions is a promising direction for enhancing LLMs. Our code and data are public at https://github.com/nlpxucan/WizardLM"
                },
                {
                    "title": "BloombergGPT: A Large Language Model for Finance",
                    "abstract": "The use of NLP in the realm of financial technology is broad and complex, with applications ranging from sentiment analysis and named entity recognition to question answering. Large Language Models (LLMs) have been shown to be effective on a variety of tasks; however, no LLM specialized for the financial domain has been reported in literature. In this work, we present BloombergGPT, a 50 billion parameter language model that is trained on a wide range of financial data. We construct a 363 billion token dataset based on Bloomberg's extensive data sources, perhaps the largest domain-specific dataset yet, augmented with 345 billion tokens from general purpose datasets. We validate BloombergGPT on standard LLM benchmarks, open financial benchmarks, and a suite of internal benchmarks that most accurately reflect our intended usage. Our mixed dataset training leads to a model that outperforms existing models on financial tasks by significant margins without sacrificing performance on general LLM benchmarks. Additionally, we explain our modeling choices, training process, and evaluation methodology. We release Training Chronicles (Appendix C) detailing our experience in training BloombergGPT."
                },
                {
                    "title": "No Fear of Classifier Biases: Neural Collapse Inspired Federated Learning with Synthetic and Fixed Classifier",
                    "abstract": "Data heterogeneity is an inherent challenge that hinders the performance of federated learning (FL). Recent studies have identified the biased classifiers of local models as the key bottleneck. Previous attempts have used classifier calibration after FL training, but this approach falls short in improving the poor feature representations caused by training-time classifier biases. Resolving the classifier bias dilemma in FL requires a full understanding of the mechanisms behind the classifier. Recent advances in neural collapse have shown that the classifiers and feature prototypes under perfect training scenarios collapse into an optimal structure called simplex equiangular tight frame (ETF). Building on this neural collapse insight, we propose a solution to the FL's classifier bias problem by utilizing a synthetic and fixed ETF classifier during training. The optimal classifier structure enables all clients to learn unified and optimal feature representations even under extremely heterogeneous data. We devise several effective modules to better adapt the ETF structure in FL, achieving both high generalization and personalization. Extensive experiments demonstrate that our method achieves state-of-the-art performances on CIFAR-10, CIFAR-100, and Tiny-ImageNet. The code is available at https://github.com/ZexiLee/ICCV-2023-FedETF."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "Vertical Federated Graph Neural Network for Recommender System",
                    "abstract": "Conventional recommender systems are required to train the recommendation model using a centralized database. However, due to data privacy concerns, this is often impractical when multi-parties are involved in recommender system training. Federated learning appears as an excellent solution to the data isolation and privacy problem. Recently, Graph neural network (GNN) is becoming a promising approach for federated recommender systems. However, a key challenge is to conduct embedding propagation while preserving the privacy of the graph structure. Few studies have been conducted on the federated GNN-based recommender system. Our study proposes the first vertical federated GNN-based recommender system, called VerFedGNN. We design a framework to transmit: (i) the summation of neighbor embeddings using random projection, and (ii) gradients of public parameter perturbed by ternary quantization mechanism. Empirical studies show that VerFedGNN has competitive prediction accuracy with existing privacy preserving GNN frameworks while enhanced privacy protection for users' interaction information."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Revisiting Weighted Aggregation in Federated Learning with Neural Networks",
                    "abstract": "In federated learning (FL), weighted aggregation of local models is conducted to generate a global model, and the aggregation weights are normalized (the sum of weights is 1) and proportional to the local data sizes. In this paper, we revisit the weighted aggregation process and gain new insights into the training dynamics of FL. First, we find that the sum of weights can be smaller than 1, causing global weight shrinking effect (analogous to weight decay) and improving generalization. We explore how the optimal shrinking factor is affected by clients' data heterogeneity and local epochs. Second, we dive into the relative aggregation weights among clients to depict the clients' importance. We develop client coherence to study the learning dynamics and find a critical point that exists. Before entering the critical point, more coherent clients play more essential roles in generalization. Based on the above insights, we propose an effective method for Federated Learning with Learnable Aggregation Weights, named as FedLAW. Extensive experiments verify that our method can improve the generalization of the global model by a large margin on different datasets and models."
                },
                {
                    "title": "The Flan Collection: Designing Data and Methods for Effective Instruction Tuning",
                    "abstract": "We study the design decisions of publicly available instruction tuning methods, and break down the development of Flan 2022 (Chung et al., 2022). Through careful ablation studies on the Flan Collection of tasks and methods, we tease apart the effect of design decisions which enable Flan-T5 to outperform prior work by 3-17%+ across evaluation settings. We find task balancing and enrichment techniques are overlooked but critical to effective instruction tuning, and in particular, training with mixed prompt settings (zero-shot, few-shot, and chain-of-thought) actually yields stronger (2%+) performance in all settings. In further experiments, we show Flan-T5 requires less finetuning to converge higher and faster than T5 on single downstream tasks, motivating instruction-tuned models as more computationally-efficient starting checkpoints for new tasks. Finally, to accelerate research on instruction tuning, we make the Flan 2022 collection of datasets, templates, and methods publicly available at https://github.com/google-research/FLAN/tree/main/flan/v2."
                },
                {
                    "title": "Self-Instruct: Aligning Language Models with Self-Generated Instructions",
                    "abstract": "Large \u201cinstruction-tuned\u201d language models (i.e., finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is often limited in quantity, diversity, and creativity, therefore hindering the generality of the tuned model. We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off their own generations. Our pipeline generates instructions, input, and output samples from a language model, then filters invalid or similar ones before using them to finetune the original model. Applying our method to the vanilla GPT3, we demonstrate a 33% absolute improvement over the original model on Super-NaturalInstructions, on par with the performance of InstructGPT-001, which was trained with private user data and human annotations. For further evaluation, we curate a set of expert-written instructions for novel tasks, and show through human evaluation that tuning GPT3 with Self-Instruct outperforms using existing public instruction datasets by a large margin, leaving only a 5% absolute gap behind InstructGPT-001. Self-Instruct provides an almost annotation-free method for aligning pre-trained language models with instructions, and we release our large synthetic dataset to facilitate future studies on instruction tuning."
                },
                {
                    "title": "Constitutional AI: Harmlessness from AI Feedback",
                    "abstract": "As AI systems become more capable, we would like to enlist their help to supervise other AIs. We experiment with methods for training a harmless AI assistant through self-improvement, without any human labels identifying harmful outputs. The only human oversight is provided through a list of rules or principles, and so we refer to the method as 'Constitutional AI'. The process involves both a supervised learning and a reinforcement learning phase. In the supervised phase we sample from an initial model, then generate self-critiques and revisions, and then finetune the original model on revised responses. In the RL phase, we sample from the finetuned model, use a model to evaluate which of the two samples is better, and then train a preference model from this dataset of AI preferences. We then train with RL using the preference model as the reward signal, i.e. we use 'RL from AI Feedback' (RLAIF). As a result we are able to train a harmless but non-evasive AI assistant that engages with harmful queries by explaining its objections to them. Both the SL and RL methods can leverage chain-of-thought style reasoning to improve the human-judged performance and transparency of AI decision making. These methods make it possible to control AI behavior more precisely and with far fewer human labels."
                },
                {
                    "title": "Where to Begin? On the Impact of Pre-Training and Initialization in Federated Learning",
                    "abstract": "An oft-cited challenge of federated learning is the presence of heterogeneity. \\emph{Data heterogeneity} refers to the fact that data from different clients may follow very different distributions. \\emph{System heterogeneity} refers to the fact that client devices have different system capabilities. A considerable number of federated optimization methods address this challenge. In the literature, empirical evaluations usually start federated training from random initialization. However, in many practical applications of federated learning, the server has access to proxy data for the training task that can be used to pre-train a model before starting federated training. We empirically study the impact of starting from a pre-trained model in federated learning using four standard federated learning benchmark datasets. Unsurprisingly, starting from a pre-trained model reduces the training time required to reach a target error rate and enables the training of more accurate models (up to 40\\%) than is possible when starting from random initialization. Surprisingly, we also find that starting federated learning from a pre-trained initialization reduces the effect of both data and system heterogeneity. We recommend that future work proposing and evaluating federated optimization methods evaluate the performance when starting from random and pre-trained initializations. We also believe this study raises several questions for further work on understanding the role of heterogeneity in federated optimization."
                },
                {
                    "title": "FLamby: Datasets and Benchmarks for Cross-Silo Federated Learning in Realistic Healthcare Settings",
                    "abstract": "Federated Learning (FL) is a novel approach enabling several clients holding sensitive data to collaboratively train machine learning models, without centralizing data. The cross-silo FL setting corresponds to the case of few ($2$--$50$) reliable clients, each holding medium to large datasets, and is typically found in applications such as healthcare, finance, or industry. While previous works have proposed representative datasets for cross-device FL, few realistic healthcare cross-silo FL datasets exist, thereby slowing algorithmic research in this critical application. In this work, we propose a novel cross-silo dataset suite focused on healthcare, FLamby (Federated Learning AMple Benchmark of Your cross-silo strategies), to bridge the gap between theory and practice of cross-silo FL. FLamby encompasses 7 healthcare datasets with natural splits, covering multiple tasks, modalities, and data volumes, each accompanied with baseline training code. As an illustration, we additionally benchmark standard FL algorithms on all datasets. Our flexible and modular suite allows researchers to easily download datasets, reproduce results and re-use the different components for their research. FLamby is available at~\\url{www.github.com/owkin/flamby}."
                },
                {
                    "title": "Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned",
                    "abstract": "We describe our early efforts to red team language models in order to simultaneously discover, measure, and attempt to reduce their potentially harmful outputs. We make three main contributions. First, we investigate scaling behaviors for red teaming across 3 model sizes (2.7B, 13B, and 52B parameters) and 4 model types: a plain language model (LM); an LM prompted to be helpful, honest, and harmless; an LM with rejection sampling; and a model trained to be helpful and harmless using reinforcement learning from human feedback (RLHF). We find that the RLHF models are increasingly difficult to red team as they scale, and we find a flat trend with scale for the other model types. Second, we release our dataset of 38,961 red team attacks for others to analyze and learn from. We provide our own analysis of the data and find a variety of harmful outputs, which range from offensive language to more subtly harmful non-violent unethical outputs. Third, we exhaustively describe our instructions, processes, statistical methodologies, and uncertainty about red teaming. We hope that this transparency accelerates our ability to work together as a community in order to develop shared norms, practices, and technical standards for how to red team language models."
                },
                {
                    "title": "FedX: Unsupervised Federated Learning with Cross Knowledge Distillation",
                    "abstract": "This paper presents FedX, an unsupervised federated learning framework. Our model learns unbiased representation from decentralized and heterogeneous local data. It employs a two-sided knowledge distillation with contrastive learning as a core component, allowing the federated system to function without requiring clients to share any data features. Furthermore, its adaptable architecture can be used as an add-on module for existing unsupervised algorithms in federated settings. Experiments show that our model improves performance significantly (1.58--5.52pp) on five unsupervised algorithms."
                },
                {
                    "title": "FLAIR: Federated Learning Annotated Image Repository",
                    "abstract": "Cross-device federated learning is an emerging machine learning (ML) paradigm where a large population of devices collectively train an ML model while the data remains on the devices. This research field has a unique set of practical challenges, and to systematically make advances, new datasets curated to be compatible with this paradigm are needed. Existing federated learning benchmarks in the image domain do not accurately capture the scale and heterogeneity of many real-world use cases. We introduce FLAIR, a challenging large-scale annotated image dataset for multi-label classification suitable for federated learning. FLAIR has 429,078 images from 51,414 Flickr users and captures many of the intricacies typically encountered in federated learning, such as heterogeneous user data and a long-tailed label distribution. We implement multiple baselines in different learning setups for different tasks on this dataset. We believe FLAIR can serve as a challenging benchmark for advancing the state-of-the art in federated learning. Dataset access and the code for the benchmark are available at \\url{https://github.com/apple/ml-flair}."
                },
                {
                    "title": "On the Importance and Applicability of Pre-Training for Federated Learning",
                    "abstract": "Pre-training is prevalent in nowadays deep learning to improve the learned model's performance. However, in the literature on federated learning (FL), neural networks are mostly initialized with random weights. These attract our interest in conducting a systematic study to explore pre-training for FL. Across multiple visual recognition benchmarks, we found that pre-training can not only improve FL, but also close its accuracy gap to the counterpart centralized learning, especially in the challenging cases of non-IID clients' data. To make our findings applicable to situations where pre-trained models are not directly available, we explore pre-training with synthetic data or even with clients' data in a decentralized manner, and found that they can already improve FL notably. Interestingly, many of the techniques we explore are complementary to each other to further boost the performance, and we view this as a critical result toward scaling up deep FL for real-world applications. We conclude our paper with an attempt to understand the effect of pre-training on FL. We found that pre-training enables the learned global models under different clients' data conditions to converge to the same loss basin, and makes global aggregation in FL more stable. Nevertheless, pre-training seems to not alleviate local model drifting, a fundamental problem in FL under non-IID data."
                },
                {
                    "title": "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback",
                    "abstract": "We apply preference modeling and reinforcement learning from human feedback (RLHF) to \ufb01netune language models to act as helpful and harmless assistants. We \ufb01nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, ef\ufb01ciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and identify a roughly linear relation between the RL reward and the square root of the KL divergence between the policy and its initialization. Alongside our main results, we perform peripheral analyses on calibration, competing objectives, and the use of OOD detection, compare our models with human writers, and provide samples from our models using prompts appearing in recent related work. Figure These plots show that PM accuracy decreases as we focus exclusively on comparisons between pairs of samples with high score. We have normalized all preference models to have the same mean score on a held-out dataset so that they\u2019re directly comparable, and then plotted accuracy for the comparisons where both samples have scores above a speci\ufb01c threshold."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "Efficient-FedRec: Efficient Federated Learning Framework for Privacy-Preserving News Recommendation",
                    "abstract": "News recommendation is critical for personalized news access. Most existing news recommendation methods rely on centralized storage of users\u2019 historical news click behavior data, which may lead to privacy concerns and hazards. Federated Learning is a privacy-preserving framework for multiple clients to collaboratively train models without sharing their private data. However, the computation and communication cost of directly learning many existing news recommendation models in a federated way are unacceptable for user clients. In this paper, we propose an efficient federated learning framework for privacy-preserving news recommendation. Instead of training and communicating the whole model, we decompose the news recommendation model into a large news model maintained in the server and a light-weight user model shared on both server and clients, where news representations and user model are communicated between server and clients. More specifically, the clients request the user model and news representations from the server, and send their locally computed gradients to the server for aggregation. The server updates its global user model with the aggregated gradients, and further updates its news model to infer updated news representations. Since the local gradients may contain private information, we propose a secure aggregation method to aggregate gradients in a privacy-preserving way. Experiments on two real-world datasets show that our method can reduce the computation and communication cost on clients while keep promising model performance."
                },
                {
                    "title": "Finetuned Language Models Are Zero-Shot Learners",
                    "abstract": "This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning."
                },
                {
                    "title": "A Field Guide to Federated Optimization",
                    "abstract": "Federated learning and analytics are a distributed approach for collaboratively learning models (or statistics) from decentralized data, motivated by and designed for privacy protection. The distributed learning process can be formulated as solving federated optimization problems, which emphasize communication efficiency, data heterogeneity, compatibility with privacy and system requirements, and other constraints that are not primary considerations in other problem settings. This paper provides recommendations and guidelines on formulating, designing, evaluating and analyzing federated optimization algorithms through concrete examples and practical implementation, with a focus on conducting effective simulations to infer real-world performance. The goal of this work is not to survey the current literature, but to inspire researchers and practitioners to design federated learning algorithms that can be used in various practical applications."
                },
                {
                    "title": "Evaluating Large Language Models Trained on Code",
                    "abstract": "We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics."
                },
                {
                    "title": "Model-Contrastive Federated Learning",
                    "abstract": "Federated learning enables multiple parties to collaboratively train a machine learning model without communicating their local data. A key challenge in federated learning is to handle the heterogeneity of local data distribution across parties. Although many studies have been proposed to address this challenge, we find that they fail to achieve high performance in image datasets with deep learning models. In this paper, we propose MOON: model-contrastive federated learning. MOON is a simple and effective federated learning framework. The key idea of MOON is to utilize the similarity between model representations to correct the local training of individual parties, i.e., conducting contrastive learning in model-level. Our extensive experiments show that MOON significantly outperforms the other state-of-the-art federated learning algorithms on various image classification tasks."
                },
                {
                    "title": "Ditto: Fair and Robust Federated Learning Through Personalization",
                    "abstract": "Fairness and robustness are two important concerns for federated learning systems. In this work, we identify that robustness to data and model poisoning attacks and fairness, measured as the uniformity of performance across devices, are competing constraints in statistically heterogeneous networks. To address these constraints, we propose employing a simple, general framework for personalized federated learning, Ditto, that can inherently provide fairness and robustness benefits, and develop a scalable solver for it. Theoretically, we analyze the ability of Ditto to achieve fairness and robustness simultaneously on a class of linear problems. Empirically, across a suite of federated datasets, we show that Ditto not only achieves competitive performance relative to recent personalization methods, but also enables more accurate, robust, and fair models relative to state-of-the-art fair or robust baselines."
                },
                {
                    "title": "Measuring Massive Multitask Language Understanding",
                    "abstract": "We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings."
                },
                {
                    "title": "Tackling the Objective Inconsistency Problem in Heterogeneous Federated Optimization",
                    "abstract": "In federated optimization, heterogeneity in the clients' local datasets and computation speeds results in large variations in the number of local updates performed by each client in each communication round. Naive weighted aggregation of such models causes objective inconsistency, that is, the global model converges to a stationary point of a mismatched objective function which can be arbitrarily different from the true objective. This paper provides a general framework to analyze the convergence of federated heterogeneous optimization algorithms. It subsumes previously proposed methods such as FedAvg and FedProx and provides the first principled understanding of the solution bias and the convergence slowdown due to objective inconsistency. Using insights from this analysis, we propose FedNova, a normalized averaging method that eliminates objective inconsistency while preserving fast error convergence."
                },
                {
                    "title": "Adaptive Federated Optimization",
                    "abstract": "Federated learning is a distributed machine learning paradigm in which a large number of clients coordinate with a central server to learn a model without sharing their own training data. Due to the heterogeneity of the client datasets, standard federated optimization methods such as Federated Averaging (FedAvg) are often difficult to tune and exhibit unfavorable convergence behavior. In non-federated settings, adaptive optimization methods have had notable success in combating such issues. In this work, we propose federated versions of adaptive optimizers, including Adagrad, Adam, and Yogi, and analyze their convergence in the presence of heterogeneous data for general nonconvex settings. Our results highlight the interplay between client heterogeneity and communication efficiency. We also perform extensive experiments on these methods and show that the use of adaptive optimizers can significantly improve the performance of federated learning."
                },
                {
                    "title": "Performance Analysis and Optimization in Privacy-Preserving Federated Learning",
                    "abstract": "As a means of decentralized machine learning, federated learning (FL) has recently drawn considerable attentions. One of the prominent advantages of FL is its capability of preventing clients' data from being directly exposed to external adversaries. Nevertheless, via a viewpoint of information theory, it is still possible for an attacker to steal private information from eavesdropping upon the shared models uploaded by FL clients. In order to address this problem, we develop a novel privacy preserving FL framework based on the concept of differential privacy (DP). To be specific, we first borrow the concept of local DP and introduce a client-level DP (CDP) by adding artificial noises to the shared models before uploading them to servers. Then, we prove that our proposed CDP algorithm can satisfy the DP guarantee with adjustable privacy protection levels by varying the variances of the artificial noises. More importantly, we derive a theoretical convergence upper-bound of the CDP algorithm. Our derived upper-bound reveals that there exists an optimal number of communication rounds to achieve the best convergence performance in terms of loss function values for a given privacy protection level. Furthermore, to obtain this optimal number of communication rounds, which cannot be derived in a closed-form expression, we propose a communication rounds discounting (CRD) method. Compared with the heuristic searching method, our proposed CRD can achieve a much better trade-off between the computational complexity of searching for the optimal number and the convergence performance. Extensive experiments indicate that our CDP algorithm with an optimization on the number of communication rounds using the proposed CRD can effectively improve both the FL training efficiency and FL model quality for a given privacy protection level."
                },
                {
                    "title": "Advances and Open Problems in Federated Learning",
                    "abstract": "Federated learning (FL) is a machine learning setting where many clients (e.g. mobile devices or whole organizations) collaboratively train a model under the orchestration of a central server (e.g. service provider), while keeping the training data decentralized. FL embodies the principles of focused data collection and minimization, and can mitigate many of the systemic privacy risks and costs resulting from traditional, centralized machine learning and data science approaches. Motivated by the explosive growth in FL research, this paper discusses recent advances and presents an extensive collection of open problems and challenges."
                },
                {
                    "title": "SCAFFOLD: Stochastic Controlled Averaging for Federated Learning",
                    "abstract": "Federated Averaging (FedAvg) has emerged as the algorithm of choice for federated learning due to its simplicity and low communication cost. However, in spite of recent research efforts, its performance is not fully understood. We obtain tight convergence rates for FedAvg and prove that it suffers from `client-drift' when the data is heterogeneous (non-iid), resulting in unstable and slow convergence. \nAs a solution, we propose a new algorithm (SCAFFOLD) which uses control variates (variance reduction) to correct for the `client-drift' in its local updates. We prove that SCAFFOLD requires significantly fewer communication rounds and is not affected by data heterogeneity or client sampling. Further, we show that (for quadratics) SCAFFOLD can take advantage of similarity in the client's data yielding even faster convergence. The latter is the first result to quantify the usefulness of local-steps in distributed optimization."
                },
                {
                    "title": "Measuring the Effects of Non-Identical Data Distribution for Federated Visual Classification",
                    "abstract": "Federated Learning enables visual models to be trained in a privacy-preserving way using real-world data from mobile devices. Given their distributed nature, the statistics of the data across these devices is likely to differ significantly. In this work, we look at the effect such non-identical data distributions has on visual classification via Federated Learning. We propose a way to synthesize datasets with a continuous range of identicalness and provide performance measures for the Federated Averaging algorithm. We show that performance degrades as distributions differ more, and propose a mitigation strategy via server momentum. Experiments on CIFAR-10 demonstrate improved classification performance over a range of non-identicalness, with classification accuracy improved from 30.1% to 76.9% in the most skewed settings."
                },
                {
                    "title": "Federated Optimization in Heterogeneous Networks",
                    "abstract": "Federated Learning is a distributed learning paradigm with two key challenges that differentiate it from traditional distributed optimization: (1) significant variability in terms of the systems characteristics on each device in the network (systems heterogeneity), and (2) non-identically distributed data across the network (statistical heterogeneity). In this work, we introduce a framework, FedProx, to tackle heterogeneity in federated networks. FedProx can be viewed as a generalization and re-parametrization of FedAvg, the current state-of-the-art method for federated learning. While this re-parameterization makes only minor modifications to the method itself, these modifications have important ramifications both in theory and in practice. Theoretically, we provide convergence guarantees for our framework when learning over data from non-identical distributions (statistical heterogeneity), and while adhering to device-level systems constraints by allowing each participating device to perform a variable amount of work (systems heterogeneity). Practically, we demonstrate that FedProx allows for more robust convergence than FedAvg across a suite of realistic federated datasets. In particular, in highly heterogeneous settings, FedProx demonstrates significantly more stable and accurate convergence behavior relative to FedAvg---improving absolute test accuracy by 22% on average."
                },
                {
                    "title": "LEAF: A Benchmark for Federated Settings",
                    "abstract": "Modern federated networks, such as those comprised of wearable devices, mobile phones, or autonomous vehicles, generate massive amounts of data each day. This wealth of data can help to learn models that can improve the user experience on each device. However, the scale and heterogeneity of federated data presents new challenges in research areas such as federated learning, meta-learning, and multi-task learning. As the machine learning community begins to tackle these challenges, we are at a critical time to ensure that developments made in these areas are grounded with realistic benchmarks. To this end, we propose LEAF, a modular benchmarking framework for learning in federated settings. LEAF includes a suite of open-source federated datasets, a rigorous evaluation framework, and a set of reference implementations, all geared towards capturing the obstacles and intricacies of practical federated environments."
                },
                {
                    "title": "Constituency Parsing with a Self-Attentive Encoder",
                    "abstract": "We demonstrate that replacing an LSTM encoder with a self-attentive architecture can lead to improvements to a state-of-the-art discriminative constituency parser. The use of attention makes explicit the manner in which information is propagated between different locations in the sentence, which we use to both analyze our model and propose potential improvements. For example, we find that separating positional and content information in the encoder can lead to improved parsing accuracy. Additionally, we evaluate different approaches for lexical representation. Our parser achieves new state-of-the-art results for single models trained on the Penn Treebank: 93.55 F1 without the use of any external data, and 95.13 F1 when using pre-trained word representations. Our parser also outperforms the previous best-published accuracy figures on 8 of the 9 languages in the SPMRL dataset."
                },
                {
                    "title": "Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms",
                    "abstract": "We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at this https URL"
                },
                {
                    "title": "Communication-Efficient Learning of Deep Networks from Decentralized Data",
                    "abstract": "Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data center and training there using conventional approaches. We advocate an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. \nWe present a practical method for the federated learning of deep networks based on iterative model averaging, and conduct an extensive empirical evaluation, considering five different model architectures and four datasets. These experiments demonstrate the approach is robust to the unbalanced and non-IID data distributions that are a defining characteristic of this setting. Communication costs are the principal constraint, and we show a reduction in required communication rounds by 10-100x as compared to synchronized stochastic gradient descent."
                },
                {
                    "title": "The MNIST Database of Handwritten Digit Images for Machine Learning Research [Best of the Web]",
                    "abstract": "In this issue, \u201cBest of the Web\u201d presents the modified National Institute of Standards and Technology (MNIST) resources, consisting of a collection of handwritten digit images used extensively in optical character recognition and machine learning research."
                },
                {
                    "title": "Calibrating Noise to Sensitivity in Private Data Analysis",
                    "abstract": "We continue a line of research initiated in [10, 11] on privacy-preserving statistical databases. Consider a trusted server that holds a database of sensitive information. Given a query function f mapping databases to reals, the so-called true answer is the result of applying f to the database. To protect privacy, the true answer is perturbed by the addition of random noise generated according to a carefully chosen distribution, and this response, the true answer plus noise, is returned to the user. Previous work focused on the case of noisy sums, in which f = \u03a3 i g(x i ), where x i denotes the ith row of the database and g maps database rows to [0,1]. We extend the study to general functions f, proving that privacy can be preserved by calibrating the standard deviation of the noise according to the sensitivity of the function f. Roughly speaking, this is the amount that any single argument to f can change its output. The new analysis shows that for several particular applications substantially less noise is needed than was previously understood to be the case. The first step is a very clean characterization of privacy in terms of indistinguishability of transcripts. Additionally, we obtain separation results showing the increased value of interactive sanitization mechanisms over non-interactive."
                },
                {
                    "title": "Personalized Federated Learning with Inferred Collaboration Graphs",
                    "abstract": "Personalized federated learning (FL) aims to collaboratively train a personalized model for each client. Previous methods do not adaptively determine who to collaborate at a fine-grained level, making them difficult to handle diverse data heterogeneity levels and those cases where malicious clients exist. To address this issue, our core idea is to learn a collaboration graph, which models the benefits from each pairwise collaboration and allocates appropriate collaboration strengths. Based on this, we propose a novel personalized FL algo-rithm, pFedGraph, which consists of two key modules: (1) inferring the collaboration graph based on pairwise model similarity and dataset size at server to promote fine-grained collaboration and (2) optimizing local model with the assistance of aggregated model at client to promote personalization. The advantage of pFedGraph is flexibly adaptive to diverse data heterogeneity levels and model poisoning attacks, as the proposed collaboration graph always pushes each client to collaborate more with similar and beneficial clients. Extensive experiments show that pFedGraph consistently outperforms the other 14 baseline methods across various heterogeneity levels and multiple cases where malicious clients exist. Code will be available at https://github.com/MediaBrain-SJTU/pFedGraph."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Visualizing Data using t-SNE",
                    "abstract": "We present a new technique called \u201ct-SNE\u201d that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to influence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are significantly better than those produced by the other techniques on almost all of the datasets."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.DC",
                "cs.LG",
                "cs.MA"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we establish a realistic benchmark for Federated Learning in Large Language Models (FedLLM) that accurately reflects the complexities and diversities of real-world cross-user datasets?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the lack of standardized evaluation metrics and datasets in the field of FedLLM. A realistic benchmark like FedLLM-Bench will facilitate fair comparisons among different federated learning methods, thereby advancing knowledge in the area and promoting collaboration. It will also lead to practical applications by enabling researchers and practitioners to better understand the performance of various approaches in real-world scenarios, ultimately improving the development of privacy-preserving models.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem include the inherent complexities of real-world data, such as varying data quality, quantity, and user preferences, which are difficult to replicate in artificial datasets. Naive approaches that simply partition existing centralized datasets fail to capture these natural properties, leading to unrealistic evaluations. Additionally, the lack of a unified framework for training and evaluation setups complicates the re-implementation of studies and increases the risk of unfair comparisons, making it essential to develop a comprehensive and realistic benchmark.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on creating artificial datasets by partitioning centralized data, which does not reflect the true nature of federated learning scenarios. This approach has been limited by the absence of a clear understanding of the diverse characteristics of real-world data and the complexities involved in federated settings. Moreover, existing solutions have not adequately addressed the need for a standardized evaluation framework. Our approach differs by providing a benchmark that incorporates naturally split datasets based on real user IDs, capturing the essential diversities and complexities of federated learning.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves creating the FedLLM-Bench benchmark, which includes three datasets for federated instruction tuning (Fed-Aya, Fed-WildChat, and Fed-ChatbotIT) and one dataset for federated preference alignment (Fed-ChatbotPA). These datasets are designed to reflect realistic federated properties, such as language diversity, varying data quality and quantity, sequence length heterogeneity, and differing user preferences. We will implement eight representative baseline methods and six evaluation metrics to conduct extensive experiments. The"
            }
        },
        "author_data": {
            "726803b0-11f7-4181-9517-a7afdcb835b3": {
                "pk": "726803b0-11f7-4181-9517-a7afdcb835b3",
                "name": "Rui Ye",
                "collaborators": [
                    "Rui Ge",
                    "Yuchi Fengting",
                    "Jingyi Chai",
                    "Yanfeng Wang",
                    "Siheng Chen"
                ],
                "domain": [
                    "Federated Learning",
                    "Natural Language Processing",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "Leveraging Unstructured Text Data for Federated Instruction Tuning of Large Language Models",
                        "abstract": "Federated instruction tuning enables multiple clients to collaboratively fine-tune a shared large language model (LLM) that can follow humans' instructions without directly sharing raw data. However, existing literature impractically requires that all the clients readily hold instruction-tuning data (i.e., structured instruction-response pairs), which necessitates massive human annotations since clients' data is usually unstructured text instead. Addressing this, we propose a novel and flexible framework FedIT-U2S, which can automatically transform unstructured corpus into structured data for federated instruction tuning. FedIT-U2S consists two key steps: (1) few-shot instruction-tuning data generation, where each unstructured data piece together with several examples is combined to prompt an LLM in generating an instruction-response pair. To further enhance the flexibility, a retrieval-based example selection technique is proposed, where the examples are automatically selected based on the relatedness between the client's data piece and example pool, bypassing the need of determining examples in advance. (2) A typical federated instruction tuning process based on the generated data. Overall, FedIT-U2S can be applied to diverse scenarios as long as the client holds valuable text corpus, broadening the application scope of federated instruction tuning. We conduct a series of experiments on three domains (medicine, knowledge, and math), showing that our proposed FedIT-U2S can consistently and significantly brings improvement over the base LLM."
                    }
                ]
            },
            "02f9c1a3-df2a-4389-b55c-03b60622caa2": {
                "pk": "02f9c1a3-df2a-4389-b55c-03b60622caa2",
                "name": "Rui Ge",
                "collaborators": [
                    "Ronald Garcia",
                    "Alexander J. Summers",
                    "Rui Ye",
                    "Yuchi Fengting",
                    "Jingyi Chai",
                    "Yanfeng Wang",
                    "Siheng Chen"
                ],
                "domain": [
                    "SMT",
                    "Program Verification",
                    "Federated Learning",
                    "Natural Language Processing"
                ],
                "publications": [
                    {
                        "title": "A Formal Model to Prove Instantiation Termination for E-matching-Based Axiomatisations (Extended Version)",
                        "abstract": "SMT-based program analysis and verification often involve reasoning about program features that have been specified using quantifiers; incorporating quantifiers into SMT-based reasoning is, however, known to be challenging. If quantifier instantiation is not carefully controlled, then runtime and outcomes can be brittle and hard to predict. In particular, uncontrolled quantifier instantiation can lead to unexpected incompleteness and even non-termination. E-matching is the most widely-used approach for controlling quantifier instantiation, but when axiomatisations are complex, even experts cannot tell if their use of E-matching guarantees completeness or termination.   This paper presents a new formal model that facilitates the proof, once and for all, that giving a complex E-matching-based axiomatisation to an SMT solver, such as Z3 or cvc5, will not cause non-termination. Key to our technique is an operational semantics for solver behaviour that models how the E-matching rules common to most solvers are used to determine when quantifier instantiations are enabled, but abstracts over irrelevant details of individual solvers. We demonstrate the effectiveness of our technique by presenting a termination proof for a set theory axiomatisation adapted from those used in the Dafny and Viper verifiers."
                    },
                    {
                        "title": "Leveraging Unstructured Text Data for Federated Instruction Tuning of Large Language Models",
                        "abstract": "Federated instruction tuning enables multiple clients to collaboratively fine-tune a shared large language model (LLM) that can follow humans' instructions without directly sharing raw data. However, existing literature impractically requires that all the clients readily hold instruction-tuning data (i.e., structured instruction-response pairs), which necessitates massive human annotations since clients' data is usually unstructured text instead. Addressing this, we propose a novel and flexible framework FedIT-U2S, which can automatically transform unstructured corpus into structured data for federated instruction tuning. FedIT-U2S consists two key steps: (1) few-shot instruction-tuning data generation, where each unstructured data piece together with several examples is combined to prompt an LLM in generating an instruction-response pair. To further enhance the flexibility, a retrieval-based example selection technique is proposed, where the examples are automatically selected based on the relatedness between the client's data piece and example pool, bypassing the need of determining examples in advance. (2) A typical federated instruction tuning process based on the generated data. Overall, FedIT-U2S can be applied to diverse scenarios as long as the client holds valuable text corpus, broadening the application scope of federated instruction tuning. We conduct a series of experiments on three domains (medicine, knowledge, and math), showing that our proposed FedIT-U2S can consistently and significantly brings improvement over the base LLM."
                    }
                ]
            },
            "e3d40d58-00fc-4dc8-967d-9d4330bc9c50": {
                "pk": "e3d40d58-00fc-4dc8-967d-9d4330bc9c50",
                "name": "Xinyu Zhu",
                "collaborators": [
                    "Yongliang Shen",
                    "Weiming Lu",
                    "Lilin Zhang",
                    "Ning Yang",
                    "Meng Xiao",
                    "Yulei Ye"
                ],
                "domain": [
                    "Drug-Drug Interaction",
                    "Fairness in Machine Learning",
                    "High Energy Physics",
                    "Transformer Models"
                ],
                "publications": [
                    {
                        "title": "Molecular Substructure-Aware Network for Drug-Drug Interaction Prediction",
                        "abstract": "Concomitant administration of drugs can cause drug-drug interactions (DDIs). Some drug combinations are beneficial, but other ones may cause negative effects which are previously unrecorded. Previous works on DDI prediction usually rely on hand-engineered domain knowledge, which is laborious to obtain. In this work, we propose a novel model, Molecular Substructure-Aware Network (MSAN), to effectively predict potential DDIs from molecular structures of drug pairs. We adopt a Transformer-like substructure extraction module to acquire a fixed number of representative vectors that are associated with various substructure patterns of the drug molecule. Then, interaction strength between the two drugs' substructures will be captured by a similarity-based interaction module. We also perform a substructure dropping augmentation before graph encoding to alleviate overfitting. Experimental results from a real-world dataset reveal that our proposed model achieves the state-of-the-art performance. We also show that the predictions of our model are highly interpretable through a case study."
                    },
                    {
                        "title": "Adaptive Fair Representation Learning for Personalized Fairness in Recommendations via Information Alignment",
                        "abstract": "Personalized fairness in recommendations has been attracting increasing attention from researchers. The existing works often treat a fairness requirement, represented as a collection of sensitive attributes, as a hyper-parameter, and pursue extreme fairness by completely removing information of sensitive attributes from the learned fair embedding, which suffer from two challenges: huge training cost incurred by the explosion of attribute combinations, and the suboptimal trade-off between fairness and accuracy. In this paper, we propose a novel Adaptive Fair Representation Learning (AFRL) model, which achieves a real personalized fairness due to its advantage of training only one model to adaptively serve different fairness requirements during inference phase. Particularly, AFRL treats fairness requirements as inputs and can learn an attribute-specific embedding for each attribute from the unfair user embedding, which endows AFRL with the adaptability during inference phase to determine the non-sensitive attributes under the guidance of the user's unique fairness requirement. To achieve a better trade-off between fairness and accuracy in recommendations, AFRL conducts a novel Information Alignment to exactly preserve discriminative information of non-sensitive attributes and incorporate a debiased collaborative embedding into the fair embedding to capture attribute-independent collaborative signals, without loss of fairness. Finally, the extensive experiments conducted on real datasets together with the sound theoretical analysis demonstrate the superiority of AFRL."
                    },
                    {
                        "title": "Prospect of measuring the top quark mass through energy correlators",
                        "abstract": "Reaching a high precision of the top quark mass is an important task of the Large Hadron Collider. We perform a feasibility study of measuring the top quark mass through the three-point energy correlator. The expected sensitivity of the top quark mass in the boosted regime is presented. We further introduce its application to the low top $p_\\text{T}$ regime and demonstrate that both the W boson and the top quark masses could be extracted from this single observable. Compared to traditional observables, the energy correlator shows robustness to uncertainties that usually dominate experimental measurements and provides a promising way to improve experimental precision."
                    }
                ]
            },
            "892b79c3-ce03-4034-8a33-554f985845e5": {
                "pk": "892b79c3-ce03-4034-8a33-554f985845e5",
                "name": "Jingyi Chai",
                "collaborators": [
                    "Rui Ye",
                    "Siheng Chen",
                    "Yanfeng Wang",
                    "Wenhao Wang",
                    "Yue Shi",
                    "Yuxuan Xiong",
                    "Bingbing Ni",
                    "Wenjun Zhang",
                    "Xinyu Zhu",
                    "Xiaoyu Liang"
                ],
                "domain": [
                    "Federated Learning",
                    "3D Reconstruction",
                    "Large Language Models",
                    "Privacy Preservation"
                ],
                "publications": [
                    {
                        "title": "USR: Unsupervised Separated 3D Garment and Human Reconstruction via Geometry and Semantic Consistency",
                        "abstract": "Dressed people reconstruction from images is a popular task with promising applications in the creative media and game industry. However, most existing methods reconstruct the human body and garments as a whole with the supervision of 3D models, which hinders the downstream interaction tasks and requires hard-to-obtain data. To address these issues, we propose an unsupervised separated 3D garments and human reconstruction model (USR), which reconstructs the human body and authentic textured clothes in layers without 3D models. More specifically, our method proposes a generalized surface-aware neural radiance field to learn the mapping between sparse multi-view images and geometries of the dressed people. Based on the full geometry, we introduce a Semantic and Confidence Guided Separation strategy (SCGS) to detect, segment, and reconstruct the clothes layer, leveraging the consistency between 2D semantic and 3D geometry. Moreover, we propose a Geometry Fine-tune Module to smooth edges. Extensive experiments on our dataset show that comparing with state-of-the-art methods, USR achieves improvements on both geometry and appearance reconstruction while supporting generalizing to unseen people in real time. Besides, we also introduce SMPL-D model to show the benefit of the separated modeling of clothes and the human body that allows swapping clothes and virtual try-on."
                    },
                    {
                        "title": "Federated Learning Empowered by Generative Content",
                        "abstract": "Federated learning (FL) enables leveraging distributed private data for model training in a privacy-preserving way. However, data heterogeneity significantly limits the performance of current FL methods. In this paper, we propose a novel FL framework termed FedGC, designed to mitigate data heterogeneity issues by diversifying private data with generative content. FedGC is a simple-to-implement framework as it only introduces a one-shot step of data generation. In data generation, we summarize three crucial and worth-exploring aspects (budget allocation, prompt design, and generation guidance) and propose three solution candidates for each aspect. Specifically, to achieve a better trade-off between data diversity and fidelity for generation guidance, we propose to generate data based on the guidance of prompts and real data simultaneously. The generated data is then merged with private data to facilitate local model training. Such generative data increases the diversity of private data to prevent each client from fitting the potentially biased private data, alleviating the issue of data heterogeneity. We conduct a systematic empirical study on FedGC, covering diverse baselines, datasets, scenarios, and modalities. Interesting findings include (1) FedGC consistently and significantly enhances the performance of FL methods, even when notable disparities exist between generative and private data; (2) FedGC achieves both better performance and privacy-preservation. We wish this work can inspire future works to further explore the potential of enhancing FL with generative content."
                    },
                    {
                        "title": "KnowledgeSG: Privacy-Preserving Synthetic Text Generation with Knowledge Distillation from Server",
                        "abstract": "The success of large language models (LLMs) facilitate many parties to fine-tune LLMs on their own private data. However, this practice raises privacy concerns due to the memorization of LLMs. Existing solutions, such as utilizing synthetic data for substitution, struggle to simultaneously improve performance and preserve privacy. They either rely on a local model for generation, resulting in a performance decline, or take advantage of APIs, directly exposing the data to API servers. To address this issue, we propose KnowledgeSG, a novel client-server framework which enhances synthetic data quality and improves model performance while ensuring privacy. We achieve this by learning local knowledge from the private data with differential privacy (DP) and distilling professional knowledge from the server. Additionally, inspired by federated learning, we transmit models rather than data between the client and server to prevent privacy leakage. Extensive experiments in medical and financial domains demonstrate the effectiveness of KnowledgeSG. Our code is now publicly available at https://github.com/wwh0411/KnowledgeSG."
                    },
                    {
                        "title": "Emerging Safety Attack and Defense in Federated Instruction Tuning of Large Language Models",
                        "abstract": "Federated learning (FL) enables multiple parties to collaboratively fine-tune an large language model (LLM) without the need of direct data sharing. Ideally, by training on decentralized data that is aligned with human preferences and safety principles, federated instruction tuning can result in an LLM that could behave in a helpful and safe manner. In this paper, we for the first time reveal the vulnerability of safety alignment in FedIT by proposing a simple, stealthy, yet effective safety attack method. Specifically, the malicious clients could automatically generate attack data without involving manual efforts and attack the FedIT system by training their local LLMs on such attack data. Unfortunately, this proposed safety attack not only can compromise the safety alignment of LLM trained via FedIT, but also can not be effectively defended against by many existing FL defense methods. Targeting this, we further propose a post-hoc defense method, which could rely on a fully automated pipeline: generation of defense data and further fine-tuning of the LLM. Extensive experiments show that our safety attack method can significantly compromise the LLM's safety alignment (e.g., reduce safety rate by 70\\%), which can not be effectively defended by existing defense methods (at most 4\\% absolute improvement), while our safety defense method can significantly enhance the attacked LLM's safety alignment (at most 69\\% absolute improvement)."
                    },
                    {
                        "title": "Leveraging Unstructured Text Data for Federated Instruction Tuning of Large Language Models",
                        "abstract": "Federated instruction tuning enables multiple clients to collaboratively fine-tune a shared large language model (LLM) that can follow humans' instructions without directly sharing raw data. However, existing literature impractically requires that all the clients readily hold instruction-tuning data (i.e., structured instruction-response pairs), which necessitates massive human annotations since clients' data is usually unstructured text instead. Addressing this, we propose a novel and flexible framework FedIT-U2S, which can automatically transform unstructured corpus into structured data for federated instruction tuning. FedIT-U2S consists two key steps: (1) few-shot instruction-tuning data generation, where each unstructured data piece together with several examples is combined to prompt an LLM in generating an instruction-response pair. To further enhance the flexibility, a retrieval-based example selection technique is proposed, where the examples are automatically selected based on the relatedness between the client's data piece and example pool, bypassing the need of determining examples in advance. (2) A typical federated instruction tuning process based on the generated data. Overall, FedIT-U2S can be applied to diverse scenarios as long as the client holds valuable text corpus, broadening the application scope of federated instruction tuning. We conduct a series of experiments on three domains (medicine, knowledge, and math), showing that our proposed FedIT-U2S can consistently and significantly brings improvement over the base LLM."
                    },
                    {
                        "title": "OpenFedLLM: Training Large Language Models on Decentralized Private Data via Federated Learning",
                        "abstract": "Trained on massive publicly available data, large language models (LLMs) have demonstrated tremendous success across various fields. While more data contributes to better performance, a disconcerting reality is that high-quality public data will be exhausted in a few years. In this paper, we offer a potential next step for contemporary LLMs: collaborative and privacy-preserving LLM training on the underutilized distributed private data via federated learning (FL), where multiple data owners collaboratively train a shared model without transmitting raw data. To achieve this, we build a concise, integrated, and research-friendly framework/codebase, named OpenFedLLM. It covers federated instruction tuning for enhancing instruction-following capability, federated value alignment for aligning with human values, and 7 representative FL algorithms. Besides, OpenFedLLM supports training on diverse domains, where we cover 8 training datasets; and provides comprehensive evaluations, where we cover 30+ evaluation metrics. Through extensive experiments, we observe that all FL algorithms outperform local training on training LLMs, demonstrating a clear performance improvement across a variety of settings. Notably, in a financial benchmark, Llama2-7B fine-tuned by applying any FL algorithm can outperform GPT-4 by a significant margin while the model obtained through individual training cannot, demonstrating strong motivation for clients to participate in FL. The code is available at https://github.com/rui-ye/OpenFedLLM."
                    }
                ]
            },
            "fb1a83e2-2ba9-4dbb-8ea0-b4b817d96577": {
                "pk": "fb1a83e2-2ba9-4dbb-8ea0-b4b817d96577",
                "name": "Yaxin Du",
                "collaborators": [
                    "Siheng Chen",
                    "Yanfeng Wang",
                    "Rui Ye",
                    "Wanru Zhao",
                    "Zezhou Wang",
                    "Zhuzhong Qian",
                    "Zhenyang Ni",
                    "Nicholas Donald Lane",
                    "Fengting Yuchi",
                    "Jingjing Qu"
                ],
                "domain": [
                    "Federated Learning",
                    "Instruction Tuning",
                    "Data Quality",
                    "Multi-Agent Systems"
                ],
                "publications": [
                    {
                        "title": "Federated Instruction Tuning of LLMs with Domain Coverage Augmentation",
                        "abstract": "Federated Domain-specific Instruction Tuning (FedDIT) utilizes limited cross-client private data together with server-side public data for instruction augmentation, ultimately boosting model performance within specific domains. To date, the factors affecting FedDIT remain unclear, and existing instruction augmentation methods primarily focus on the centralized setting without considering distributed environments. Our experiments reveal that the cross-client domain coverage, rather than data heterogeneity, drives model performance in FedDIT. In response, we propose FedDCA, which optimizes domain coverage through greedy client center selection and retrieval-based augmentation. For client-side computational efficiency and system scalability, FedDCA$^*$, the variant of FedDCA, utilizes heterogeneous encoders with server-side feature alignment. Extensive experiments across four distinct domains (code, medical, financial, and mathematical) substantiate the effectiveness of both methods. Additionally, we investigate privacy preservation against memory extraction attacks utilizing various amounts of public data. Results show that there is no significant correlation between the volume of public data and the privacy-preserving capability. However, as the fine-tuning rounds increase, the risk of privacy leakage reduces or converges."
                    },
                    {
                        "title": "Fake It Till Make It: Federated Learning with Consensus-Oriented Generation",
                        "abstract": "In federated learning (FL), data heterogeneity is one key bottleneck that causes model divergence and limits performance. Addressing this, existing methods often regard data heterogeneity as an inherent property and propose to mitigate its adverse effects by correcting models. In this paper, we seek to break this inherent property by generating data to complement the original dataset to fundamentally mitigate heterogeneity level. As a novel attempt from the perspective of data, we propose federated learning with consensus-oriented generation (FedCOG). FedCOG consists of two key components at the client side: complementary data generation, which generates data extracted from the shared global model to complement the original dataset, and knowledge-distillation-based model training, which distills knowledge from global model to local model based on the generated data to mitigate over-fitting the original heterogeneous dataset. FedCOG has two critical advantages: 1) it can be a plug-and-play module to further improve the performance of most existing FL methods, and 2) it is naturally compatible with standard FL protocols such as Secure Aggregation since it makes no modification in communication process. Extensive experiments on classical and real-world FL datasets show that FedCOG consistently outperforms state-of-the-art methods."
                    },
                    {
                        "title": "Enhancing Data Quality in Federated Fine-Tuning of Foundation Models",
                        "abstract": "In the current landscape of foundation model training, there is a significant reliance on public domain data, which is nearing exhaustion according to recent research. To further scale up, it is crucial to incorporate collaboration among multiple specialized and high-quality private domain data sources. However, the challenge of training models locally without sharing private data presents numerous obstacles in data quality control. To tackle this issue, we propose a data quality control pipeline for federated fine-tuning of foundation models. This pipeline computes scores reflecting the quality of training data and determines a global threshold for a unified standard, aiming for improved global performance. Our experiments show that the proposed quality control pipeline facilitates the effectiveness and reliability of the model training, leading to better performance."
                    },
                    {
                        "title": "Data Quality Control in Federated Instruction-tuning of Large Language Models",
                        "abstract": "By leveraging massively distributed data, federated learning (FL) enables collaborative instruction tuning of large language models (LLMs) in a privacy-preserving way. While FL effectively expands the data quantity, the issue of data quality remains under-explored in the current literature on FL for LLMs. To address this gap, we propose a new framework of federated instruction tuning of LLMs with data quality control (FedDQC), which measures data quality to facilitate the subsequent filtering and hierarchical training processes. Our approach introduces an efficient metric to assess each client's instruction-response alignment (IRA), identifying potentially noisy data through single-shot inference. Low-IRA samples are potentially noisy and filtered to mitigate their negative impacts. To further utilize this IRA value, we propose a quality-aware hierarchical training paradigm, where LLM is progressively fine-tuned from high-IRA to low-IRA data, mirroring the easy-to-hard learning process. We conduct extensive experiments on 4 synthetic and a real-world dataset, and compare our method with baselines adapted from centralized setting. Results show that our method consistently and significantly improves the performance of LLMs trained on mix-quality data in FL."
                    },
                    {
                        "title": "OpenFedLLM: Training Large Language Models on Decentralized Private Data via Federated Learning",
                        "abstract": "Trained on massive publicly available data, large language models (LLMs) have demonstrated tremendous success across various fields. While more data contributes to better performance, a disconcerting reality is that high-quality public data will be exhausted in a few years. In this paper, we offer a potential next step for contemporary LLMs: collaborative and privacy-preserving LLM training on the underutilized distributed private data via federated learning (FL), where multiple data owners collaboratively train a shared model without transmitting raw data. To achieve this, we build a concise, integrated, and research-friendly framework/codebase, named OpenFedLLM. It covers federated instruction tuning for enhancing instruction-following capability, federated value alignment for aligning with human values, and 7 representative FL algorithms. Besides, OpenFedLLM supports training on diverse domains, where we cover 8 training datasets; and provides comprehensive evaluations, where we cover 30+ evaluation metrics. Through extensive experiments, we observe that all FL algorithms outperform local training on training LLMs, demonstrating a clear performance improvement across a variety of settings. Notably, in a financial benchmark, Llama2-7B fine-tuned by applying any FL algorithm can outperform GPT-4 by a significant margin while the model obtained through individual training cannot, demonstrating strong motivation for clients to participate in FL. The code is available at https://github.com/rui-ye/OpenFedLLM."
                    },
                    {
                        "title": "Self-Evolving Multi-Agent Collaboration Networks for Software Development",
                        "abstract": "LLM-driven multi-agent collaboration (MAC) systems have demonstrated impressive capabilities in automatic software development at the function level. However, their heavy reliance on human design limits their adaptability to the diverse demands of real-world software development. To address this limitation, we introduce EvoMAC, a novel self-evolving paradigm for MAC networks. Inspired by traditional neural network training, EvoMAC obtains text-based environmental feedback by verifying the MAC network's output against a target proxy and leverages a novel textual backpropagation to update the network. To extend coding capabilities beyond function-level tasks to more challenging software-level development, we further propose rSDE-Bench, a requirement-oriented software development benchmark, which features complex and diverse software requirements along with automatic evaluation of requirement correctness. Our experiments show that: i) The automatic requirement-aware evaluation in rSDE-Bench closely aligns with human evaluations, validating its reliability as a software-level coding benchmark. ii) EvoMAC outperforms previous SOTA methods on both the software-level rSDE-Bench and the function-level HumanEval benchmarks, reflecting its superior coding capabilities. The benchmark can be downloaded at https://yuzhu-cai.github.io/rSDE-Bench/."
                    }
                ]
            },
            "a8983b13-4435-4a54-be28-50130398dd2d": {
                "pk": "a8983b13-4435-4a54-be28-50130398dd2d",
                "name": "Yang Liu",
                "collaborators": [],
                "domain": [
                    "Semiconductor Physics",
                    "Natural Language Processing",
                    "Mathematical Analysis",
                    "Quantum Gravity"
                ],
                "publications": [
                    {
                        "title": "On the calculation of Schottky contact resistivity",
                        "abstract": "This numerical study examines the importance of self-consistently accounting for transport and electrostatics in the calculaiton of semiconductor/metal Schottky contact resistivity. It is shown that ignoring such self-consistency results in significant under-estimation of the contact resistivity. An explicit numerical method has also been proposed to efficiently improve contact resistivity calculations."
                    },
                    {
                        "title": "Fine-tune BERT for Extractive Summarization",
                        "abstract": "BERT, a pre-trained Transformer model, has achieved ground-breaking performance on multiple NLP tasks. In this paper, we describe BERTSUM, a simple variant of BERT, for extractive summarization. Our system is the state of the art on the CNN/Dailymail dataset, outperforming the previous best-performed system by 1.65 on ROUGE-L. The codes to reproduce our results are available at https://github.com/nlpyang/BertSum"
                    },
                    {
                        "title": "Global Weak Solutions for the Half-Wave Maps Equation in $\\mathbb{R}$",
                        "abstract": "We establish the existence of weak global solutions of the half-wave maps equation with the target $S^2$ on $\\mathbb{R}^{1+1}$ with large initial data in $\\dot{H}^1 \\cap \\dot{H}^{\\frac{1}{2}}(\\mathbb{R})$. We first prove the global well-posedness of a regularized equation. Then we show that the weak limit of the regularized solutions is a weak solution of the half-wave maps equation as the regularization parameter $\\varepsilon \\rightarrow 0$."
                    },
                    {
                        "title": "On the Range of Cosine Transform of Distributions for Torus-Invariant Complex Minkowski Spaces",
                        "abstract": "In this paper, we study the range of (absolute value) cosine transforms for which we give a proof for an extended surjectivity theorem by making applications of the Fredholm's theorem in integral equations, and show a Hermitian characterization theorem for complex Minkowski metrics on \\mathbb{C}^n. Moreover, we parametrize the Grassmannian in an elementary linear algebra approach, and give a characterization on the image of the (absolute value) cosine transform on the space of distributions on the Grassmannian Gr_{2}(\\mathbb{C}^{2}), by computing the coefficients in the Legendre series expansion of distributions."
                    },
                    {
                        "title": "On Explicit Holmes-Thompson Area Formula in Integral Geometry",
                        "abstract": "In this article, we give an exposition on the Holmes-Thompson theory developed by Alvarez. The space of geodesics in Minkowski space has a symplectic structure which is induced by the projection from the sphere-bundle. we show that it can be also obtained from the symplectic structure on the tangent bundle of the Riemannian manifold, the tangent bundle of the Minkowski unit sphere. We give detailed descriptions and expositions on Holmes-Thompson volumes in Minkowski space by the symplectic structure and the Crofton measures for them. For the Minkowski plane, a normed two dimensional space, we express the area explicitly in an integral geometry way, by putting a measure on the plane, which gives an extension of Alvarez's result for higher dimensional cases."
                    },
                    {
                        "title": "Towards a Unified Belief Structure in Games with indeterminate probabilities",
                        "abstract": "This paper provides an analysis of different formal representations of beliefs in epistemic game theory. The aim is to attempt a synthesis of different structures of beliefs in the presence of indeterminate probabilities. Special attention is also paid to the decision-theoretic principle known as the thesis of no subjective probability for self-action. Conditions in cope with this principle are given which underlie the interrelationships between different models of beliefs, and it is shown that under these conditions different doxastic structures can be coherently unified."
                    },
                    {
                        "title": "Function-led design of multifunctional stimuli-responsive superhydrophobic surface based on hierarchical graphene-titania nanocoating",
                        "abstract": "Multifunctional smart superhydrophobic surface with full-spectrum tunable wettability control is fabricated through the self-assembly of the graphene and titania nanofilm double-layer coating. Advanced microfluidic manipulative functions, including directional water transport, adhesion & spreading controls, droplet storage & transfer, and droplet sensing array, can be readily realized on this smart surface. An in-depth mechanism study regarding the underlying secrets of the tunable wettability and the UV-induced superhydrophilic conversion of anatase titania are also presented."
                    },
                    {
                        "title": "Research on the Brain-inspired Cross-modal Neural Cognitive Computing Framework",
                        "abstract": "To address modeling problems of brain-inspired intelligence, this thesis is focused on researching in the semantic-oriented framework design for multimedia and multimodal information. The Multimedia Neural Cognitive Computing (MNCC) model was designed based on the nervous mechanism and cognitive architecture. Furthermore, the semantic-oriented hierarchical Cross-modal Neural Cognitive Computing (CNCC) framework was proposed based on MNCC model, and formal description and analysis for CNCC framework was given. It would effectively improve the performance of semantic processing for multimedia and cross-modal information, and has far-reaching significance for exploration and realization brain-inspired computing."
                    },
                    {
                        "title": "Intelligent Processing in Vehicular Ad hoc Networks: a Survey",
                        "abstract": "The intelligent Processing technique is more and more attractive to researchers due to its ability to deal with key problems in Vehicular Ad hoc networks. However, several problems in applying intelligent processing technologies in VANETs remain open. The existing applications are comprehensively reviewed and discussed, and classified into different categories in this paper. Their strategies, advantages/disadvantages, and performances are elaborated. By generalizing different tactics in various applications related to different scenarios of VANETs and evaluating their performances, several promising directions for future research have been suggested."
                    },
                    {
                        "title": "General Rearrangement Lemma for Heat Trace Asymptotic on Noncommutative Tori",
                        "abstract": "We study a technical problem arising from the spectral geometry of noncommutative tori: the small time heat trace asymptotic associated to a general second order elliptic operator. We extend the rearrangement operators in the conformal case to the general setting using hypergeometric integrals over Grassmannians. The main result is the explicit formula of the second heat coefficient in terms of the coefficients. When specializing to examples in conformal case, we not only recover results in previous works but also obtain some extra functional relations whose validation provides experimental support to the main results. At last, we verify the relations based on combinatorial properties derived from the hypergeometric features."
                    },
                    {
                        "title": "Global Well-Posedness For Half-Wave Maps With $S^2$ and $\\mathbb{H}^2$ Targets For Small Smooth Initial Data",
                        "abstract": "We prove global well-posedness for the half-wave map with $S^2$ target for small $\\dot{H}^{\\frac{n}{2}} \\times \\dot{H}^{\\frac{n}{2}-1}$ initial data. We also prove the global well-posedness for the equation with $\\mathbb{H}^2$ target for small smooth $\\dot{B}^{\\frac{n}{2}}_{2,1} \\times \\dot{B}^{\\frac{n}{2}-1}_{2,1}$ initial data."
                    },
                    {
                        "title": "The spectrum of Hawking radiation in Tsallis statistical mechanics",
                        "abstract": "Hawking radiation is one of the cores in modern gratitational theory. Several articles have calculated the spectrum of Hawking radiation in Boltzmann-Gibbs statistical mechanics. However, based on recent researches, gravitational systems cannot be studied by the standard statistical mechanics. In this article, we calculate the modification to the spectrum of Hawking radiation in Tsallis statistical mechanics. We obtain the modified Stefan-Boltzmann's law and modified power of Hawking radiation. We confirm the conclusion proposed by Giddings, namely, the radiation should originate from the effective radius, which extends well outside the horizon of black-hole. The lifetime of black hole and the effect of large q are discussed as well."
                    },
                    {
                        "title": "Dilatonic Effect in Double Field Theory Cosmology",
                        "abstract": "In this article we discuss some aspects of double field theory cosmology with an emphasis on the role played by the dilaton. The cosmological solutions of double field theory equations of motion after coupling a shifted dilaton to them are investigated. The equations of motion for a constant shifted dilaton and a constant usaul dilaton in an FRW universe are obtained. The solutions of these equations are obtained in both the supergravity frame and in the winding frame. We also consider three possible dark energy candidates in a 4D universe using double field theory cosmology and find some basic conditions which the three dark energy candidates should satisfy. We consider the results for a more general potential of shifted dilaton as well."
                    },
                    {
                        "title": "Hawking temperature and the bound on greybody factors in $D = 4$ double field theory",
                        "abstract": "We investigate the basic properties of Hawking radiation for spherical solutions in $D = 4$ double field theory. We give the expression of the Hawking temperature for the solution and then discuss the results of various limits. We find that for all these limits only Schwarzschild solution and F-JNW solution can generate Hawking radiation. Moreover, we obtain the lower bound on greybody factors $\\sigma_l(\\omega)$ for the spherical solutions in $D = 4$ double field theory. In particular, we calculate the bound on greybody factors $\\sigma_l(\\omega)$ for F-JNW solution. For F-JNW solution, $\\sigma_l(\\omega)$ monotonically increases with the increase of $a(b)$ for fixed $b(a)$."
                    },
                    {
                        "title": "Cyclic Structure behind Modular Gaussian Curvature",
                        "abstract": "We propose a systematic scheme for computing the variation of rearrangement operators arising in the recently developed spectral geometry on noncommutative tori and $\\theta$-deformed Riemannian manifolds. It can be summarized as a category whose objects consists of spectral functions of the rearrangement operators and morphisms are generated by transformations associated to basic operations of the variational calculus. The generators of the morphisms fulfil most of the relations in Connes's cyclic category, but also include all the partial derivatives. Comparison with Hopf cyclic theory has also been made."
                    },
                    {
                        "title": "Higgs inflation and scalar weak gravity conjecture",
                        "abstract": "In this article, we intend to find a specific model which can satisfy the further refining dS swampland conjecture and scalar weak gravity conjecture (SWGC) simultaneously, in particular, Higgs inflation model and its two extensions: Higgs-Dilaton model and Palatini Higgs inflation. We find that although Higgs inflation model and Higgs-Dilaton model could satisfy the further refining dS swampland conjecture, the two models cannot satisfy SWGC and its strong version. While Palatini Higgs inflation could satisfy these conjectures simultaneously. Therefore Palatini Higgs inflation could be a \"real\" inflation model of the universe."
                    },
                    {
                        "title": "The Importance of Human-Labeled Data in the Era of LLMs",
                        "abstract": "The advent of large language models (LLMs) has brought about a revolution in the development of tailored machine learning models and sparked debates on redefining data requirements. The automation facilitated by the training and implementation of LLMs has led to discussions and aspirations that human-level labeling interventions may no longer hold the same level of importance as in the era of supervised learning. This paper presents compelling arguments supporting the ongoing relevance of human-labeled data in the era of LLMs."
                    },
                    {
                        "title": "Randomized quasi-Monte Carlo and Owen's boundary growth condition: A spectral analysis",
                        "abstract": "In this work, we analyze the convergence rate of randomized quasi-Monte Carlo (RQMC) methods under Owen's boundary growth condition [Owen, 2006] via spectral analysis. Specifically, we examine the RQMC estimator variance for the two commonly studied sequences: the lattice rule and the Sobol' sequence, applying the Fourier transform and Walsh--Fourier transform, respectively, for this analysis. Assuming certain regularity conditions, our findings reveal that the asymptotic convergence rate of the RQMC estimator's variance closely aligns with the exponent specified in Owen's boundary growth condition for both sequence types. We also provide analysis for certain discontinuous integrands."
                    }
                ]
            },
            "414f6f2a-803f-4e29-b45f-0ea03ccd0503": {
                "pk": "414f6f2a-803f-4e29-b45f-0ea03ccd0503",
                "name": "Yanfeng Wang",
                "collaborators": [
                    "Yu Wang",
                    "Ya Zhang",
                    "Yusheng Liao",
                    "Zihan Zhao",
                    "Heyang Liu",
                    "Shuyang Jiang",
                    "Yuefu Zhou",
                    "Qi Tian",
                    "Sheng Feng",
                    "Hao Wu"
                ],
                "domain": [
                    "Multimodal Emotion Recognition",
                    "Transfer Learning",
                    "Natural Language Processing",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Multi-level Fusion of Wav2vec 2.0 and BERT for Multimodal Emotion Recognition",
                        "abstract": "The research and applications of multimodal emotion recognition have become increasingly popular recently. However, multimodal emotion recognition faces the challenge of lack of data. To solve this problem, we propose to use transfer learning which leverages state-of-the-art pre-trained models including wav2vec 2.0 and BERT for this task. Multi-level fusion approaches including coattention-based early fusion and late fusion with the models trained on both embeddings are explored. Also, a multi-granularity framework which extracts not only frame-level speech embeddings but also segment-level embeddings including phone, syllable and word-level speech embeddings is proposed to further boost the performance. By combining our coattention-based early fusion model and late fusion model with the multi-granularity feature extraction framework, we obtain result that outperforms best baseline approaches by 1.3% unweighted accuracy (UA) on the IEMOCAP dataset."
                    },
                    {
                        "title": "Knowledge-aware Bayesian Co-attention for Multimodal Emotion Recognition",
                        "abstract": "Multimodal emotion recognition is a challenging research area that aims to fuse different modalities to predict human emotion. However, most existing models that are based on attention mechanisms have difficulty in learning emotionally relevant parts on their own. To solve this problem, we propose to incorporate external emotion-related knowledge in the co-attention based fusion of pre-trained models. To effectively incorporate this knowledge, we enhance the co-attention model with a Bayesian attention module (BAM) where a prior distribution is estimated using the emotion-related knowledge. Experimental results on the IEMOCAP dataset show that the proposed approach can outperform several state-of-the-art approaches by at least 0.7% unweighted accuracy (UA)."
                    },
                    {
                        "title": "Post-decoder Biasing for End-to-End Speech Recognition of Multi-turn Medical Interview",
                        "abstract": "End-to-end (E2E) approach is gradually replacing hybrid models for automatic speech recognition (ASR) tasks. However, the optimization of E2E models lacks an intuitive method for handling decoding shifts, especially in scenarios with a large number of domain-specific rare words that hold specific important meanings. Furthermore, the absence of knowledge-intensive speech datasets in academia has been a significant limiting factor, and the commonly used speech corpora exhibit significant disparities with realistic conversation. To address these challenges, we present Medical Interview (MED-IT), a multi-turn consultation speech dataset that contains a substantial number of knowledge-intensive named entities. We also explore methods to enhance the recognition performance of rare words for E2E models. We propose a novel approach, post-decoder biasing, which constructs a transform probability matrix based on the distribution of training transcriptions. This guides the model to prioritize recognizing words in the biasing list. In our experiments, for subsets of rare words appearing in the training speech between 10 and 20 times, and between 1 and 5 times, the proposed method achieves a relative improvement of 9.3% and 5.1%, respectively."
                    },
                    {
                        "title": "MING-MOE: Enhancing Medical Multi-Task Learning in Large Language Models with Sparse Mixture of Low-Rank Adapter Experts",
                        "abstract": "Large language models like ChatGPT have shown substantial progress in natural language understanding and generation, proving valuable across various disciplines, including the medical field. Despite advancements, challenges persist due to the complexity and diversity inherent in medical tasks which often require multi-task learning capabilities. Previous approaches, although beneficial, fall short in real-world applications because they necessitate task-specific annotations at inference time, limiting broader generalization. This paper introduces MING-MOE, a novel Mixture-of-Expert~(MOE)-based medical large language model designed to manage diverse and complex medical tasks without requiring task-specific annotations, thus enhancing its usability across extensive datasets. MING-MOE employs a Mixture of Low-Rank Adaptation (MoLoRA) technique, allowing for efficient parameter usage by maintaining base model parameters static while adapting through a minimal set of trainable parameters. We demonstrate that MING-MOE achieves state-of-the-art (SOTA) performance on over 20 medical tasks, illustrating a significant improvement over existing models. This approach not only extends the capabilities of medical language models but also improves inference efficiency."
                    },
                    {
                        "title": "Towards an End-to-End Framework for Invasive Brain Signal Decoding with Large Language Models",
                        "abstract": "In this paper, we introduce a groundbreaking end-to-end (E2E) framework for decoding invasive brain signals, marking a significant advancement in the field of speech neuroprosthesis. Our methodology leverages the comprehensive reasoning abilities of large language models (LLMs) to facilitate direct decoding. By fully integrating LLMs, we achieve results comparable to the state-of-the-art cascade models. Our findings underscore the immense potential of E2E frameworks in speech neuroprosthesis, particularly as the technology behind brain-computer interfaces (BCIs) and the availability of relevant datasets continue to evolve. This work not only showcases the efficacy of combining LLMs with E2E decoding for enhancing speech neuroprosthesis but also sets a new direction for future research in BCI applications, underscoring the impact of LLMs in decoding complex neural signals for communication restoration. Code will be made available at https://github.com/FsFrancis15/BrainLLM."
                    },
                    {
                        "title": "Cooperative Learning for Noisy Supervision",
                        "abstract": "Learning with noisy labels has gained the enormous interest in the robust deep learning area. Recent studies have empirically disclosed that utilizing dual networks can enhance the performance of single network but without theoretic proof. In this paper, we propose Cooperative Learning (CooL) framework for noisy supervision that analytically explains the effects of leveraging dual or multiple networks. Specifically, the simple but efficient combination in CooL yields a more reliable risk minimization for unseen clean data. A range of experiments have been conducted on several benchmarks with both synthetic and real-world settings. Extensive results indicate that CooL outperforms several state-of-the-art methods."
                    },
                    {
                        "title": "Improving the Reliability of Large Language Models by Leveraging Uncertainty-Aware In-Context Learning",
                        "abstract": "In recent years, large-scale language models (LLMs) have gained attention for their impressive text generation capabilities. However, these models often face the challenge of \"hallucination,\" which undermines their reliability. In this study, we introduce an uncertainty-aware in-context learning framework to empower the model to enhance or reject its output in response to uncertainty. Human-defined methods for estimating uncertainty typically assume that \"uncertainty is lower when the model's response is correct compared to when it is incorrect.\" However, setting a precise threshold to distinguish correctness is challenging. Therefore, we introduce uncertainty information as an intermediary variable that implicitly influences the model's behavior. Our innovative uncertainty-aware in-context learning framework involves fine-tuning the LLM using a calibration dataset. Our aim is to improve the model's responses by filtering out answers with high uncertainty while considering the model's knowledge limitations. We evaluate the model's knowledge by examining multiple responses to the same question for the presence of a correct answer. When the model lacks relevant knowledge, the response should indicate that the question cannot be answered. Conversely, when the model has relevant knowledge, the response should provide the correct answer. Extensive experiments confirm the effectiveness of our framework, leading to two key findings. First, the logit output values of the LLM partly reflect inherent uncertainty. Second, our model autonomously recognizes uncertainty, resulting in improved responses."
                    },
                    {
                        "title": "Leveraging Diverse Modeling Contexts with Collaborating Learning for Neural Machine Translation",
                        "abstract": "Autoregressive (AR) and Non-autoregressive (NAR) models are two types of generative models for Neural Machine Translation (NMT). AR models predict tokens in a word-by-word manner and can effectively capture the distribution of real translations. NAR models predict tokens by extracting bidirectional contextual information which can improve the inference speed but they suffer from performance degradation. Previous works utilized AR models to enhance NAR models by reducing the training data's complexity or incorporating the global information into AR models by virtue of NAR models. However, those investigated methods only take advantage of the contextual information of a single type of model while neglecting the diversity in the contextual information that can be provided by different types of models. In this paper, we propose a novel generic collaborative learning method, DCMCL, where AR and NAR models are treated as collaborators instead of teachers and students. To hierarchically leverage the bilateral contextual information, token-level mutual learning and sequence-level contrastive learning are adopted between AR and NAR models. Extensive experiments on four widely used benchmarks show that the proposed DCMCL method can simultaneously improve both AR and NAR models with up to 1.38 and 2.98 BLEU scores respectively, and can also outperform the current best-unified model with up to 0.97 BLEU scores for both AR and NAR decoding."
                    },
                    {
                        "title": "ReflecTool: Towards Reflection-Aware Tool-Augmented Clinical Agents",
                        "abstract": "Large Language Models (LLMs) have shown promising potential in the medical domain, assisting with tasks like clinical note generation and patient communication. However, current LLMs are limited to text-based communication, hindering their ability to interact with diverse forms of information in clinical environments. Despite clinical agents succeeding in diverse signal interaction, they are oriented to a single clinical scenario and hence fail for broader applications. To evaluate clinical agents holistically, we propose ClinicalAgent Bench~(CAB), a comprehensive medical agent benchmark consisting of 18 tasks across five key realistic clinical dimensions. Building on this, we introduce ReflecTool, a novel framework that excels at utilizing domain-specific tools within two stages. The first optimization stage progressively enlarges a long-term memory by saving successful solving processes and tool-wise experience of agents in a tiny pre-defined training set. In the following inference stage, ReflecTool can search for supportive successful demonstrations from already built long-term memory to guide the tool selection strategy, and a verifier improves the tool usage according to the tool-wise experience with two verification methods--iterative refinement and candidate selection. Extensive experiments on ClinicalAgent Benchmark demonstrate that ReflecTool surpasses the pure LLMs with more than 10 points and the well-established agent-based methods with 3 points, highlighting its adaptability and effectiveness in solving complex clinical tasks."
                    },
                    {
                        "title": "Deep Hashing with Triplet Quantization Loss",
                        "abstract": "With the explosive growth of image databases, deep hashing, which learns compact binary descriptors for images, has become critical for fast image retrieval. Many existing deep hashing methods leverage quantization loss, defined as distance between the features before and after quantization, to reduce the error from binarizing features. While minimizing the quantization loss guarantees that quantization has minimal effect on retrieval accuracy, it unfortunately significantly reduces the expressiveness of features even before the quantization. In this paper, we show that the above definition of quantization loss is too restricted and in fact not necessary for maintaining high retrieval accuracy. We therefore propose a new form of quantization loss measured in triplets. The core idea of the triplet quantization loss is to learn discriminative real-valued descriptors which lead to minimal loss on retrieval accuracy after quantization. Extensive experiments on two widely used benchmark data sets of different scales, CIFAR-10 and In-shop, demonstrate that the proposed method outperforms the state-of-the-art deep hashing methods. Moreover, we show that the compact binary descriptors obtained with triplet quantization loss lead to very small performance drop after quantization."
                    },
                    {
                        "title": "An Element Sensitive Saliency Model with Position Prior Learning for Web Pages",
                        "abstract": "Understanding human visual attention is important for multimedia applications. Many studies have attempted to learn from eye-tracking data and build computational saliency prediction models. However, limited efforts have been devoted to saliency prediction for Web pages, which are characterized by more diverse content elements and spatial layouts. In this paper, we propose a novel end-to-end deep generative saliency model for Web pages. To capture position biases introduced by page layouts, a Position Prior Learning sub-network is proposed, which models position biases as multivariate Gaussian distribution using variational auto-encoder. To model different elements of a Web page, a Multi Discriminative Region Detection (MDRD) branch and a Text Region Detection(TRD) branch are introduced, which target to extract discriminative localizations and \"prominent\" text regions likely to correspond to human attention, respectively. We validate the proposed model with FiWI, a public Web-page dataset, and shows that the proposed model outperforms the state-of-art models for Web-page saliency prediction."
                    },
                    {
                        "title": "Accelerate CNN via Recursive Bayesian Pruning",
                        "abstract": "Channel Pruning, widely used for accelerating Convolutional Neural Networks, is an NP-hard problem due to the inter-layer dependency of channel redundancy. Existing methods generally ignored the above dependency for computation simplicity. To solve the problem, under the Bayesian framework, we here propose a layer-wise Recursive Bayesian Pruning method (RBP). A new dropout-based measurement of redundancy, which facilitate the computation of posterior assuming inter-layer dependency, is introduced. Specifically, we model the noise across layers as a Markov chain and target its posterior to reflect the inter-layer dependency. Considering the closed form solution for posterior is intractable, we derive a sparsity-inducing Dirac-like prior which regularizes the distribution of the designed noise to automatically approximate the posterior. Compared with the existing methods, no additional overhead is required when the inter-layer dependency assumed. The redundant channels can be simply identified by tiny dropout noise and directly pruned layer by layer. Experiments on popular CNN architectures have shown that the proposed method outperforms several state-of-the-arts. Particularly, we achieve up to $\\bf{5.0\\times}$ and $\\bf{2.2\\times}$ FLOPs reduction with little accuracy loss on the large scale dataset ILSVRC2012 for VGG16 and ResNet50, respectively."
                    },
                    {
                        "title": "Multi-Scale Memory Comparison for Zero-/Few-Shot Anomaly Detection",
                        "abstract": "Anomaly detection has gained considerable attention due to its broad range of applications, particularly in industrial defect detection. To address the challenges of data collection, researchers have introduced zero-/few-shot anomaly detection techniques that require minimal normal images for each category. However, complex industrial scenarios often involve multiple objects, presenting a significant challenge. In light of this, we propose a straightforward yet powerful multi-scale memory comparison framework for zero-/few-shot anomaly detection. Our approach employs a global memory bank to capture features across the entire image, while an individual memory bank focuses on simplified scenes containing a single object. The efficacy of our method is validated by its remarkable achievement of 4th place in the zero-shot track and 2nd place in the few-shot track of the Visual Anomaly and Novelty Detection (VAND) competition."
                    },
                    {
                        "title": "Domain-Invariant Adversarial Learning for Unsupervised Domain Adaption",
                        "abstract": "Unsupervised domain adaption aims to learn a powerful classifier for the target domain given a labeled source data set and an unlabeled target data set. To alleviate the effect of `domain shift', the major challenge in domain adaptation, studies have attempted to align the distributions of the two domains. Recent research has suggested that generative adversarial network (GAN) has the capability of implicitly capturing data distribution. In this paper, we thus propose a simple but effective model for unsupervised domain adaption leveraging adversarial learning. The same encoder is shared between the source and target domains which is expected to extract domain-invariant representations with the help of an adversarial discriminator. With the labeled source data, we introduce the center loss to increase the discriminative power of feature learned. We further align the conditional distribution of the two domains to enforce the discrimination of the features in the target domain. Unlike previous studies where the source features are extracted with a fixed pre-trained encoder, our method jointly learns feature representations of two domains. Moreover, by sharing the encoder, the model does not need to know the source of images during testing and hence is more widely applicable. We evaluate the proposed method on several unsupervised domain adaption benchmarks and achieve superior or comparable performance to state-of-the-art results."
                    }
                ]
            },
            "f4ce33b0-bafa-4be6-af21-dee0b5b31374": {
                "pk": "f4ce33b0-bafa-4be6-af21-dee0b5b31374",
                "name": "Siheng Chen",
                "collaborators": [
                    "Jelena Kova\u010devi\u0107",
                    "Ya Zhang",
                    "Rohan Varma",
                    "Aarti Singh",
                    "Maosen Li",
                    "Chaojing Duan",
                    "Gary P. T. Choi",
                    "L. Mahadevan",
                    "Zida Cheng",
                    "Qingguo Li"
                ],
                "domain": [
                    "Graph Signal Processing",
                    "Machine Learning",
                    "3D Point Cloud",
                    "Autonomous Driving"
                ],
                "publications": [
                    {
                        "title": "Xi-Zhao Model Preserves WD Spaces",
                        "abstract": "For a $T_1$ space $X$, Zhao and Xi constructed a dcpo model $\\hat{P}$, where $P$ is a bounded complete algebraic poset model of $X$. In this paper, we formulate the closed WD subsets of the maximal point space $\\mathrm{Max}(\\hat{P})$ and the Scott space $\\Sigma \\hat{P}$, and then prove that $X$ is a WD space if and only if $\\Sigma \\hat{P}$ is a WD space.   It is also shown that the sobrification $X^s$ (resp., the well-filtered reflection $X^w$) of $X$ can be embedded into the sobrification $\\hat{P}^s$ (resp., the well-filtered reflection $\\hat{P}^w$) of $\\hat{P}$ as a saturated subspace. Finally, we introduce two new concepts ``${\\rm H}$-model spaces\" and ``weak ${\\rm H}$-model spaces\", and provide a general framework to prove that such $T_1$ spaces can be preserved by Xi-Zhao dcpo models."
                    },
                    {
                        "title": "Discrete Signal Processing on Graphs: Sampling Theory",
                        "abstract": "We propose a sampling theory for signals that are supported on either directed or undirected graphs. The theory follows the same paradigm as classical sampling theory. We show that perfect recovery is possible for graph signals bandlimited under the graph Fourier transform. The sampled signal coefficients form a new graph signal, whose corresponding graph structure preserves the first-order difference of the original graph signal. For general graphs, an optimal sampling operator based on experimentally designed sampling is proposed to guarantee perfect recovery and robustness to noise; for graphs whose graph Fourier transforms are frames with maximal robustness to erasures as well as for Erd\\H{o}s-R\\'enyi graphs, random sampling leads to perfect recovery with high probability. We further establish the connection to the sampling theory of finite discrete-time signal processing and previous work on signal recovery on graphs. To handle full-band graph signals, we propose a graph filter bank based on sampling theory on graphs. Finally, we apply the proposed sampling theory to semi-supervised classification on online blogs and digit images, where we achieve similar or better performance with fewer labeled samples compared to previous work."
                    },
                    {
                        "title": "Multiresolution Representations for Piecewise-Smooth Signals on Graphs",
                        "abstract": "What is a mathematically rigorous way to describe the taxi-pickup distribution in Manhattan, or the profile information in online social networks? A deep understanding of representing those data not only provides insights to the data properties, but also benefits to many subsequent processing procedures, such as denoising, sampling, recovery and localization. In this paper, we model those complex and irregular data as piecewise-smooth graph signals and propose a graph dictionary to effectively represent those graph signals. We first propose the graph multiresolution analysis, which provides a principle to design good representations. We then propose a coarse-to-fine approach, which iteratively partitions a graph into two subgraphs until we reach individual nodes. This approach efficiently implements the graph multiresolution analysis and the induced graph dictionary promotes sparse representations piecewise-smooth graph signals. Finally, we validate the proposed graph dictionary on two tasks: approximation and localization. The empirical results show that the proposed graph dictionary outperforms eight other representation methods on six datasets, including traffic networks, social networks and point cloud meshes."
                    },
                    {
                        "title": "Signal Recovery on Graphs: Fundamental Limits of Sampling Strategies",
                        "abstract": "This paper builds theoretical foundations for the recovery of a newly proposed class of smooth graph signals, approximately bandlimited graph signals, under three sampling strategies: uniform sampling, experimentally designed sampling and active sampling. We then state minimax lower bounds on the maximum risk for the approximately bandlimited class under these three sampling strategies and show that active sampling cannot fundamentally outperform experimentally designed sampling. We propose a recovery strategy to compare uniform sampling with experimentally designed sampling. As the proposed recovery strategy lends itself well to statistical analysis, we derive the exact mean square error for each sampling strategy. To study convergence rates, we introduce two types of graphs and find that (1) the proposed recovery strategy achieves the optimal rates; and (2) the experimentally designed sampling fundamentally outperforms uniform sampling for Type-2 class of graphs. To validate our proposed recovery strategy, we test it on five specific graphs: a ring graph with $k$ nearest neighbors, an Erd\\H{o}s-R\\'enyi graph, a random geometric graph, a small-world graph and a power-law graph and find that experimental results match the proposed theory well. This work also presents a comprehensive explanation for when and why sampling for semi-supervised learning with graphs works."
                    },
                    {
                        "title": "Sampling and Recovery of Graph Signals based on Graph Neural Networks",
                        "abstract": "We propose interpretable graph neural networks for sampling and recovery of graph signals, respectively. To take informative measurements, we propose a new graph neural sampling module, which aims to select those vertices that maximally express their corresponding neighborhoods. Such expressiveness can be quantified by the mutual information between vertices' features and neighborhoods' features, which are estimated via a graph neural network. To reconstruct an original graph signal from the sampled measurements, we propose a graph neural recovery module based on the algorithm-unrolling technique. Compared to previous analytical sampling and recovery, the proposed methods are able to flexibly learn a variety of graph signal models from data by leveraging the learning ability of neural networks; compared to previous neural-network-based sampling and recovery, the proposed methods are designed through exploiting specific graph properties and provide interpretability. We further design a new multiscale graph neural network, which is a trainable multiscale graph filter bank and can handle various graph-related learning tasks. The multiscale network leverages the proposed graph neural sampling and recovery modules to achieve multiscale representations of a graph. In the experiments, we illustrate the effects of the proposed graph neural sampling and recovery modules and find that the modules can flexibly adapt to various graph structures and graph signals. In the task of active-sampling-based semi-supervised learning, the graph neural sampling module improves the classification accuracy over 10% in Cora dataset. We further validate the proposed multiscale graph neural network on several standard datasets for both vertex and graph classification. The results show that our method consistently improves the classification accuracies."
                    },
                    {
                        "title": "3D Point Cloud Denoising via Deep Neural Network based Local Surface Estimation",
                        "abstract": "We present a neural-network-based architecture for 3D point cloud denoising called neural projection denoising (NPD). In our previous work, we proposed a two-stage denoising algorithm, which first estimates reference planes and follows by projecting noisy points to estimated reference planes. Since the estimated reference planes are inevitably noisy, multi-projection is applied to stabilize the denoising performance. NPD algorithm uses a neural network to estimate reference planes for points in noisy point clouds. With more accurate estimations of reference planes, we are able to achieve better denoising performances with only one-time projection. To the best of our knowledge, NPD is the first work to denoise 3D point clouds with deep learning techniques. To conduct the experiments, we sample 40000 point clouds from the 3D data in ShapeNet to train a network and sample 350 point clouds from the 3D data in ModelNet10 to test. Experimental results show that our algorithm can estimate normal vectors of points in noisy point clouds. Comparing to five competitive methods, the proposed algorithm achieves better denoising performance and produces much smaller variances."
                    },
                    {
                        "title": "Control of connectivity and rigidity in prismatic assemblies",
                        "abstract": "How can we manipulate the topological connectivity of a three-dimensional prismatic assembly to control the number of internal degrees of freedom and the number of connected components in it? To answer this question in a deterministic setting, we use ideas from elementary number theory to provide a hierarchical deterministic protocol for the control of rigidity and connectivity. We then show that is possible to also use a stochastic protocol to achieve the same results via a percolation transition. Together, these approaches provide scale-independent algorithms for the cutting or gluing of three-dimensional prismatic assemblies to control their overall connectivity and rigidity."
                    },
                    {
                        "title": "Wireless 3D Point Cloud Delivery Using Deep Graph Neural Networks",
                        "abstract": "In typical point cloud delivery, a sender uses octree-based digital video compression to send three-dimensional (3D) points and color attributes over band-limited links. However, the digital-based schemes have an issue called the cliff effect, where the 3D reconstruction quality will be a step function in terms of wireless channel quality. To prevent the cliff effect subject to channel quality fluctuation, we have proposed soft point cloud delivery called HoloCast. Although the HoloCast realizes graceful quality improvement according to wireless channel quality, it requires large communication overheads. In this paper, we propose a novel scheme for soft point cloud delivery to simultaneously realize better quality and lower communication overheads. The proposed scheme introduces an end-to-end deep learning framework based on graph neural network (GNN) to reconstruct high-quality point clouds from its distorted observation under wireless fading channels. We demonstrate that the proposed GNN-based scheme can reconstruct clean 3D point cloud with low overheads by removing fading and noise effects."
                    },
                    {
                        "title": "Weighted Multi-projection: 3D Point Cloud Denoising with Estimated Tangent Planes",
                        "abstract": "As a collection of 3D points sampled from surfaces of objects, a 3D point cloud is widely used in robotics, autonomous driving and augmented reality. Due to the physical limitations of 3D sensing devices, 3D point clouds are usually noisy, which influences subsequent computations, such as surface reconstruction, recognition and many others. To denoise a 3D point cloud, we present a novel algorithm, called weighted multi-projection. Compared to many previous works on denoising, instead of directly smoothing the coordinates of 3D points, we use a two-fold smoothing: We first estimate a local tangent plane at each 3D point and then reconstruct each 3D point by weighted averaging of its projections on multiple tangent planes. We also provide the theoretical analysis for the surface normal estimation and achieve a tighter bound than in a previous work. We validate the empirical performance on the dataset of ShapeNetCore and show that weighted multi-projection outperforms its competitors in all nine classes."
                    },
                    {
                        "title": "Deterministic and stochastic control of kirigami topology",
                        "abstract": "Kirigami, the creative art of paper cutting, is a promising paradigm for mechanical metamaterials. However, to make kirigami-inspired structures a reality requires controlling the topology of kirigami to achieve connectivity and rigidity. We address this question by deriving the maximum number of cuts (minimum number of links) that still allow us to preserve global rigidity and connectivity of the kirigami. A deterministic hierarchical construction method yields an efficient topological way to control both the number of connected pieces and the total degrees of freedom. A statistical approach to the control of rigidity and connectivity in kirigami with random cuts complements the deterministic pathway, and shows that both the number of connected pieces and the degrees of freedom show percolation transitions as a function of the density of cuts (links). Together this provides a general framework for the control of rigidity and connectivity in planar kirigami."
                    },
                    {
                        "title": "MotionNet: Joint Perception and Motion Prediction for Autonomous Driving Based on Bird's Eye View Maps",
                        "abstract": "The ability to reliably perceive the environmental states, particularly the existence of objects and their motion behavior, is crucial for autonomous driving. In this work, we propose an efficient deep model, called MotionNet, to jointly perform perception and motion prediction from 3D point clouds. MotionNet takes a sequence of LiDAR sweeps as input and outputs a bird's eye view (BEV) map, which encodes the object category and motion information in each grid cell. The backbone of MotionNet is a novel spatio-temporal pyramid network, which extracts deep spatial and temporal features in a hierarchical fashion. To enforce the smoothness of predictions over both space and time, the training of MotionNet is further regularized with novel spatial and temporal consistency losses. Extensive experiments show that the proposed method overall outperforms the state-of-the-arts, including the latest scene-flow- and 3D-object-detection-based methods. This indicates the potential value of the proposed method serving as a backup to the bounding-box-based system, and providing complementary information to the motion planner in autonomous driving. Code is available at https://github.com/pxiangwu/MotionNet."
                    },
                    {
                        "title": "Semi-supervised 3D Hand-Object Pose Estimation via Pose Dictionary Learning",
                        "abstract": "3D hand-object pose estimation is an important issue to understand the interaction between human and environment. Current hand-object pose estimation methods require detailed 3D labels, which are expensive and labor-intensive. To tackle the problem of data collection, we propose a semi-supervised 3D hand-object pose estimation method with two key techniques: pose dictionary learning and an object-oriented coordinate system. The proposed pose dictionary learning module can distinguish infeasible poses by reconstruction error, enabling unlabeled data to provide supervision signals. The proposed object-oriented coordinate system can make 3D estimations equivariant to the camera perspective. Experiments are conducted on FPHA and HO-3D datasets. Our method reduces estimation error by 19.5% / 24.9% for hands/objects compared to straightforward use of labeled data on FPHA and outperforms several baseline methods. Extensive experiments also validate the robustness of the proposed method."
                    },
                    {
                        "title": "Invariant Teacher and Equivariant Student for Unsupervised 3D Human Pose Estimation",
                        "abstract": "We propose a novel method based on teacher-student learning framework for 3D human pose estimation without any 3D annotation or side information. To solve this unsupervised-learning problem, the teacher network adopts pose-dictionary-based modeling for regularization to estimate a physically plausible 3D pose. To handle the decomposition ambiguity in the teacher network, we propose a cycle-consistent architecture promoting a 3D rotation-invariant property to train the teacher network. To further improve the estimation accuracy, the student network adopts a novel graph convolution network for flexibility to directly estimate the 3D coordinates. Another cycle-consistent architecture promoting 3D rotation-equivariant property is adopted to exploit geometry consistency, together with knowledge distillation from the teacher network to improve the pose estimation performance. We conduct extensive experiments on Human3.6M and MPI-INF-3DHP. Our method reduces the 3D joint prediction error by 11.4% compared to state-of-the-art unsupervised methods and also outperforms many weakly-supervised methods that use side information on Human3.6M. Code will be available at https://github.com/sjtuxcx/ITES."
                    },
                    {
                        "title": "Spatio-Temporal Graph Complementary Scattering Networks",
                        "abstract": "Spatio-temporal graph signal analysis has a significant impact on a wide range of applications, including hand/body pose action recognition. To achieve effective analysis, spatio-temporal graph convolutional networks (ST-GCN) leverage the powerful learning ability to achieve great empirical successes; however, those methods need a huge amount of high-quality training data and lack theoretical interpretation. To address this issue, the spatio-temporal graph scattering transform (ST-GST) was proposed to put forth a theoretically interpretable framework; however, the empirical performance of this approach is constrainted by the fully mathematical design. To benefit from both sides, this work proposes a novel complementary mechanism to organically combine the spatio-temporal graph scattering transform and neural networks, resulting in the proposed spatio-temporal graph complementary scattering networks (ST-GCSN). The essence is to leverage the mathematically designed graph wavelets with pruning techniques to cover major information and use trainable networks to capture complementary information. The empirical experiments on hand pose action recognition show that the proposed ST-GCSN outperforms both ST-GCN and ST-GST."
                    },
                    {
                        "title": "Aware of the History: Trajectory Forecasting with the Local Behavior Data",
                        "abstract": "The historical trajectories previously passing through a location may help infer the future trajectory of an agent currently at this location. Despite great improvements in trajectory forecasting with the guidance of high-definition maps, only a few works have explored such local historical information. In this work, we re-introduce this information as a new type of input data for trajectory forecasting systems: the local behavior data, which we conceptualize as a collection of location-specific historical trajectories. Local behavior data helps the systems emphasize the prediction locality and better understand the impact of static map objects on moving agents. We propose a novel local-behavior-aware (LBA) prediction framework that improves forecasting accuracy by fusing information from observed trajectories, HD maps, and local behavior data. Also, where such historical data is insufficient or unavailable, we employ a local-behavior-free (LBF) prediction framework, which adopts a knowledge-distillation-based architecture to infer the impact of missing data. Extensive experiments demonstrate that upgrading existing methods with these two frameworks significantly improves their performances. Especially, the LBA framework boosts the SOTA methods' performance on the nuScenes dataset by at least 14% for the K=1 metrics."
                    },
                    {
                        "title": "Collaborative Perception for Autonomous Driving: Current Status and Future Trend",
                        "abstract": "Perception is one of the crucial module of the autonomous driving system, which has made great progress recently. However, limited ability of individual vehicles results in the bottleneck of improvement of the perception performance. To break through the limits of individual perception, collaborative perception has been proposed which enables vehicles to share information to perceive the environments beyond line-of-sight and field-of-view. In this paper, we provide a review of the related work about the promising collaborative perception technology, including introducing the fundamental concepts, generalizing the collaboration modes and summarizing the key ingredients and applications of collaborative perception. Finally, we discuss the open challenges and issues of this research area and give some potential further directions."
                    },
                    {
                        "title": "Hypergraph-MLP: Learning on Hypergraphs without Message Passing",
                        "abstract": "Hypergraphs are vital in modelling data with higher-order relations containing more than two entities, gaining prominence in machine learning and signal processing. Many hypergraph neural networks leverage message passing over hypergraph structures to enhance node representation learning, yielding impressive performances in tasks like hypergraph node classification. However, these message-passing-based models face several challenges, including oversmoothing as well as high latency and sensitivity to structural perturbations at inference time. To tackle those challenges, we propose an alternative approach where we integrate the information about hypergraph structures into training supervision without explicit message passing, thus also removing the reliance on it at inference. Specifically, we introduce Hypergraph-MLP, a novel learning framework for hypergraph-structured data, where the learning model is a straightforward multilayer perceptron (MLP) supervised by a loss function based on a notion of signal smoothness on hypergraphs. Experiments on hypergraph node classification tasks demonstrate that Hypergraph-MLP achieves competitive performance compared to existing baselines, and is considerably faster and more robust against structural perturbations at inference."
                    },
                    {
                        "title": "FastMAC: Stochastic Spectral Sampling of Correspondence Graph",
                        "abstract": "3D correspondence, i.e., a pair of 3D points, is a fundamental concept in computer vision. A set of 3D correspondences, when equipped with compatibility edges, forms a correspondence graph. This graph is a critical component in several state-of-the-art 3D point cloud registration approaches, e.g., the one based on maximal cliques (MAC). However, its properties have not been well understood. So we present the first study that introduces graph signal processing into the domain of correspondence graph. We exploit the generalized degree signal on correspondence graph and pursue sampling strategies that preserve high-frequency components of this signal. To address time-consuming singular value decomposition in deterministic sampling, we resort to a stochastic approximate sampling strategy. As such, the core of our method is the stochastic spectral sampling of correspondence graph. As an application, we build a complete 3D registration algorithm termed as FastMAC, that reaches real-time speed while leading to little to none performance drop. Through extensive experiments, we validate that FastMAC works for both indoor and outdoor benchmarks. For example, FastMAC can accelerate MAC by 80 times while maintaining high registration success rate on KITTI. Codes are publicly available at https://github.com/Forrest-110/FastMAC."
                    }
                ]
            }
        }
    },
    "2402.08324": {
        "paper_data": {
            "title": "Uncertainty Quantification via Stable Distribution Propagation",
            "url": "http://arxiv.org/abs/2402.08324v1",
            "arxiv_id": "2402.08324",
            "authors": [
                "Felix Petersen",
                "Aashwin Mishra",
                "Hilde Kuehne",
                "Christian Borgelt",
                "Oliver Deussen",
                "Mikhail Yurochkin"
            ],
            "abstract": "We propose a new approach for propagating stable probability distributions through neural networks. Our method is based on local linearization, which we show to be an optimal approximation in terms of total variation distance for the ReLU non-linearity. This allows propagating Gaussian and Cauchy input uncertainties through neural networks to quantify their output uncertainties. To demonstrate the utility of propagating distributions, we apply the proposed method to predicting calibrated confidence intervals and selective prediction on out-of-distribution data. The results demonstrate a broad applicability of propagating distributions and show the advantages of our method over other approaches such as moment matching.",
            "introduction": " introduction and survey,\u201d in Case Studies in Applied Bayesian Data Science , Springer, 2020, pp. 45\u201387. [25] K. N. Markelle Kelly Rachel Longjohn, Uci repository of machine learning databases , 1998. [Online]. Available: http://www.ics.uci.edu/ \u02dcmlearn/MLRepository.html . [26] Y . LeCun, C. Cortes, and C. Burges, \u201cMnist handwritten digit database,\u201d ATT Labs , vol. 2, 2010. [Online]. Available: http://yann.lecun.com/exdb/mnist . [27] G. Cohen, S. Afshar, J. Tapson, and A. V . Schaik, \u201cEmnist: Extending mnist to handwritten letters,\u201d 2017 International Joint Conference on Neural Networks (IJCNN) , 2017. [28] H. Wang, X. Shi, and D. -Y . Yeung, \u201cNatural-parameter networks: A class of probabilistic neural networks,\u201d in Proc. 30thConf. on Neural Information Processing Systems (NIPS 2016, Barcelona, Spain) , 2016, pp. 118\u2013126. [29] J. Jin, A. Dundar, and E. Culutciello, \u201cRobust convolutional neural networks under adversarial noise,\u201d in Workshop Track Int. Conf. on Learning Representations (ICLR 2016, San Juan, Puerto Rico) , 2016. [30] S. Wang and C. Manning, \u201cFast dropout training,\u201d in International Conference on Machine Learning (ICML) , 2013. [31] J. Postels, F. Ferroni, H. Coskun, N. Navab, and F. Tombari, \u201cSampling-free epistemic uncer- tainty estimation using approximated variance propagation,\u201d in Proc. International Conference on Computer Vision (ICCV) , 2019. [32] J. M. Hern \u00b4andez-Lobato and R. Adams, \u201cProbabilistic backpropagation for scalable learning of bayesian neural networks,\u201d in International Conference on Machine Learning (ICML) , 2015. [33] A. Wu, S. Nowozin, E. Meeds, R. E. Turner, J. M. Hernandez-Lobato, and A. L. Gaunt, \u201cDeterministic variational inference for robust bayesian neural networks,\u201d in International Conference on Learning Representations (ICLR) , 2019. [34] D. A. Nix and A. S. Weigend, \u201cEstimating the mean and variance of the target probability dis- tribution,\u201d in Proceedings of 1994 ieee international conference on neural networks (ICNN\u201994) , IEEE, vol. 1, 1994, pp. 55\u201360. [35] B. J. Frey and G. E. Hinton, \u201cVariational learning in nonlinear gaussian belief networks,\u201d Neural Computation , vol. 11, pp. 193\u2013213, 1 1999. [36] F. Petersen, C. Borgelt, H. Kuehne, and O. Deussen, \u201cDeep Differentiable Logic Gate Net- works,\u201d in Conference on Neural Information Processing Systems (NeurIPS) , 2022. [37] Q. Berthet, M. Blondel, O. Teboul, M. Cuturi, J. -P. Vert, and F. Bach, \u201cLearning with Differ- entiable Perturbed Optimizers,\u201d in Proc. Neural Information Processing Systems (NeurIPS) , 2020. [38] L. Montaut, Q. Le Lidec, A. Bambade, V . Petrik, J. Sivic, and J. Carpentier, \u201cDifferentiable collision detection: A randomized smoothing approach,\u201d in IEEE International Conference on Robotics and Automation (ICRA) , 2023. [39] F. Petersen, \u201cLearning with differentiable algorithms,\u201d Ph.D. dissertation, Universit \u00a8at Konstanz, 2022. [40] F. Petersen, C. Borgelt, H. Kuehne, and O. Deussen, \u201cMonotonic differentiable sorting net- works,\u201d in International Conference on Learning Representations (ICLR) , 2022. [41] L. Stewart, F. S. Bach, F. L. L \u00b4opez, and Q. Berthet, \u201cDifferentiable clustering with perturbed spanning forests,\u201d Proc. Neural Information Processing Systems (NeurIPS) , 2023. 11Published as a conference paper at ICLR 2024 [42] F. Petersen, B. Goldluecke, C. Borgelt, and O. Deussen, \u201cGenDR: A Generalized Differentiable Renderer,\u201d in Proc. International Conference on Computer Vision and Pattern Recognition (CVPR) , 2022. [43] R. A. Fisher, Iris, UCI Machine Learning Repository, 1988. [44] S. Elfwing, E. Uchibe, and K. Doya, \u201cSigmoid-weighted linear units for neural network function approximation in reinforcement learning. arxiv e-prints (2017),\u201d Neural Networks , vol. 107, pp. 3\u201311, 2018. [45] D. Hendrycks and K. Gimpel, \u201cGaussian error linear units",
            "references": [
                {
                    "title": "Uncertainties in Onboard Algorithms for Autonomous Vehicles: Challenges, Mitigation, and Perspectives",
                    "abstract": "Autonomous driving is considered one of the revolutionary technologies shaping humanity\u2019s future mobility and quality of life. However, safety remains a critical hurdle in the way of commercialization and widespread deployment of autonomous vehicles on public roads. Safety concerns require the autonomous driving system to handle uncertainties from multiple sources that are either preexisting, e.g., the stochastic behavior of traffic participants or scenario occlusion, or introduced as a result of processing, e.g., the application of neural networks. Thus, it is crucial to analyze the sources of uncertainties and quantify the risks associated with them, including the propagated risks that accumulate in the decision-making system. In this context, this paper provides an overview of uncertainty challenges and state-of-the-art techniques for mitigating these challenges. We argue that the uncertainties mainly originate from two aspects: 1) the external traffic environment, and 2) the internal autonomous driving system. Specifically, this paper first analyzes the safety challenges caused by the uncertainties and summarizes their sources. In addition, the corresponding techniques that mitigate and quantify the risk of uncertainties are presented. Finally, research perspectives are highlighted to facilitate future studies for guaranteeing the safety of autonomous vehicles."
                },
                {
                    "title": "Differentiable Clustering with Perturbed Spanning Forests",
                    "abstract": "We introduce a differentiable clustering method based on minimum-weight spanning forests, a variant of spanning trees with several connected components. Our method relies on stochastic perturbations of solutions of linear programs, for smoothing and efficient gradient computations. This allows us to include clustering in end-to-end trainable pipelines. We show that our method performs well even in difficult settings, such as datasets with high noise and challenging geometries. We also formulate an ad hoc loss to efficiently learn from partial clustering data using this operation. We demonstrate its performance on several real world datasets for supervised and semi-supervised tasks."
                },
                {
                    "title": "Deep Differentiable Logic Gate Networks",
                    "abstract": "Recently, research has increasingly focused on developing efficient neural network architectures. In this work, we explore logic gate networks for machine learning tasks by learning combinations of logic gates. These networks comprise logic gates such as\"AND\"and\"XOR\", which allow for very fast execution. The difficulty in learning logic gate networks is that they are conventionally non-differentiable and therefore do not allow training with gradient descent. Thus, to allow for effective training, we propose differentiable logic gate networks, an architecture that combines real-valued logics and a continuously parameterized relaxation of the network. The resulting discretized logic gate networks achieve fast inference speeds, e.g., beyond a million images of MNIST per second on a single CPU core."
                },
                {
                    "title": "Differentiable Collision Detection: a Randomized Smoothing Approach",
                    "abstract": "Collision detection is an important component of many robotics applications, from robot control to simulation, including motion planning and estimation. While the seminal works on the topic date back to the 80s, it is only recently that the question of properly differentiating collision detection has emerged as a central issue, thanks notably to the ongoing and various efforts made by the scientific community around the topic of differentiable physics. Yet, very few solutions have been suggested so far, and only with a strong assumption on the nature of the shapes involved. In this work, we introduce a generic and efficient approach to compute the derivatives of collision detection for any pair of convex shapes, by notably leveraging randomized smoothing techniques which have shown to be particularly adapted to capture the derivatives of non-smooth problems. This approach is implemented in the HPP-FCL and Pinocchio ecosystems, and evaluated on classic datasets and problems of the robotics literature, demonstrating few micro-second timings to compute informative derivatives directly exploitable by many real robotic applications, including differentiable simulation."
                },
                {
                    "title": "Learning with Differentiable Algorithms",
                    "abstract": "Classic algorithms and machine learning systems like neural networks are both abundant in everyday life. While classic computer science algorithms are suitable for precise execution of exactly defined tasks such as finding the shortest path in a large graph, neural networks allow learning from data to predict the most likely answer in more complex tasks such as image classification, which cannot be reduced to an exact algorithm. To get the best of both worlds, this thesis explores combining both concepts leading to more robust, better performing, more interpretable, more computationally efficient, and more data efficient architectures. The thesis formalizes the idea of algorithmic supervision, which allows a neural network to learn from or in conjunction with an algorithm. When integrating an algorithm into a neural architecture, it is important that the algorithm is differentiable such that the architecture can be trained end-to-end and gradients can be propagated back through the algorithm in a meaningful way. To make algorithms differentiable, this thesis proposes a general method for continuously relaxing algorithms by perturbing variables and approximating the expectation value in closed form, i.e., without sampling. In addition, this thesis proposes differentiable algorithms, such as differentiable sorting networks, differentiable renderers, and differentiable logic gate networks. Finally, this thesis presents alternative training strategies for learning with algorithms."
                },
                {
                    "title": "A General Framework for quantifying Aleatoric and Epistemic uncertainty in Graph Neural Networks",
                    "abstract": "Graph Neural Networks (GNN) provide a powerful framework that elegantly integrates Graph theory with Machine learning for modeling and analysis of networked data. We consider the problem of quantifying the uncertainty in predictions of GNN stemming from modeling errors and measurement uncertainty. We consider aleatoric uncertainty in the form of probabilistic links and noise in feature vector of nodes, while epistemic uncertainty is incorporated via a probability distribution over the model parameters. We propose a unified approach to treat both sources of uncertainty in a Bayesian framework, where Assumed Density Filtering is used to quantify aleatoric uncertainty and Monte Carlo dropout captures uncertainty in model parameters. Finally, the two sources of uncertainty are aggregated to estimate the total uncertainty in predictions of a GNN. Results in the real-world datasets demonstrate that the Bayesian model performs at par with a frequentist model and provides additional information about predictions uncertainty that are sensitive to uncertainties in the data and model."
                },
                {
                    "title": "GenDR: A Generalized Differentiable Renderer",
                    "abstract": "In this work, we present and study a generalized family of differentiable renderers. We discuss from scratch which components are necessary for differentiable rendering and formalize the requirements for each component. We instantiate our general differentiable renderer, which generalizes existing differentiable renderers like SoftRas and DIB-R, with an array of different smoothing distributions to cover a large spectrum of reasonable settings. We evaluate an array of differentiable renderer instantiations on the popular ShapeNet 3D reconstruction benchmark and analyze the implications of our results. Surprisingly, the simple uniform distribution yields the best overall results when averaged over 13 classes; in general, however, the optimal choice of distribution heavily depends on the task."
                },
                {
                    "title": "Monotonic Differentiable Sorting Networks",
                    "abstract": "Differentiable sorting algorithms allow training with sorting and ranking supervision, where only the ordering or ranking of samples is known. Various methods have been proposed to address this challenge, ranging from optimal transport-based differentiable Sinkhorn sorting algorithms to making classic sorting networks differentiable. One problem of current differentiable sorting methods is that they are non-monotonic. To address this issue, we propose a novel relaxation of conditional swap operations that guarantees monotonicity in differentiable sorting networks. We introduce a family of sigmoid functions and prove that they produce differentiable sorting networks that are monotonic. Monotonicity ensures that the gradients always have the correct sign, which is an advantage in gradient-based optimization. We demonstrate that monotonic differentiable sorting networks improve upon previous differentiable sorting methods."
                },
                {
                    "title": "Autoregressive Quantile Flows for Predictive Uncertainty Estimation",
                    "abstract": "Numerous applications of machine learning involve representing probability distributions over high-dimensional data. We propose autoregressive quantile flows, a flexible class of normalizing flow models trained using a novel objective based on proper scoring rules. Our objective does not require calculating computationally expensive determinants of Jacobians during training and supports new types of neural architectures, such as neural autoregressive flows from which it is easy to sample. We leverage these models in quantile flow regression, an approach that parameterizes predictive conditional distributions with flows, resulting in improved probabilistic predictions on tasks such as time series forecasting and object detection. Our novel objective functions and neural flow parameterizations also yield improvements on popular generation and density estimation tasks, and represent a step beyond maximum likelihood learning of flows."
                },
                {
                    "title": "A Review and Comparative Study on Probabilistic Object Detection in Autonomous Driving",
                    "abstract": "Capturing uncertainty in object detection is indispensable for safe autonomous driving. In recent years, deep learning has become the de-facto approach for object detection, and many probabilistic object detectors have been proposed. However, there is no summary on uncertainty estimation in deep object detection, and existing methods are either built with different network architectures and uncertainty estimation methods, or evaluated on different datasets with a wide range of evaluation metrics. As a result, a comparison among methods remains challenging, as does the selection of a model that best suits a particular application. This paper aims to alleviate this problem by providing a review and comparative study on existing probabilistic object detection methods for autonomous driving applications. First, we provide an overview of practical uncertainty estimation methods in deep learning, and then systematically survey existing methods and evaluation metrics for probabilistic object detection. Next, we present a strict comparative study for probabilistic object detection based on an image detector and three public autonomous driving datasets. Finally, we present a discussion of the remaining challenges and future works. Code has been made available at https://github.com/asharakeh/pod_compare.git."
                },
                {
                    "title": "Amodal Segmentation through Out-of-Task and Out-of-Distribution Generalization with a Bayesian Model",
                    "abstract": "Amodal completion is a visual task that humans perform easily but which is difficult for computer vision algorithms. The aim is to segment those object boundaries which are occluded and hence invisible. This task is particularly challenging for deep neural networks because data is difficult to obtain and annotate. Therefore, we formulate amodal segmentation as an out-of-task and out-of-distribution generalization problem. Specifically, we replace the fully connected classifier in neural networks with a Bayesian generative model of the neural network features. The model is trained from non-occluded images using bounding box annotations and class labels only, but is applied to generalize out-of-task to object segmentation and to generalize out-of-distribution to segment occluded objects. We demonstrate how such Bayesian models can naturally generalize beyond the training task labels when they learn a prior that models the object's background context and shape. Moreover, by leveraging an outlier process, Bayesian models can further generalize out-of-distribution to segment partially occluded objects and to predict their amodal object boundaries. Our algorithm outperforms alternative methods that use the same supervision by a large margin, and even outperforms methods where annotated amodal segmentations are used during training, when the amount of occlusion is large. Code is publicly available at https://github.com/YihongSun/Bayesian-Amodal."
                },
                {
                    "title": "Analytical Probability Distributions and EM-Learning for Deep Generative Networks",
                    "abstract": "Deep Generative Networks (DGNs) with probabilistic modeling of their output and latent space are currently trained via Variational Autoencoders (VAEs). In the absence of a known analytical form for the posterior and likelihood expectation, VAEs resort to approximations, including (Amortized) Variational Inference (AVI) and Monte-Carlo (MC) sampling. We exploit the Continuous Piecewise Affine (CPA) property of modern DGNs to derive their posterior and marginal distributions as well as the latter's first moments. These findings enable us to derive an analytical Expectation-Maximization (EM) algorithm that enables gradient-free DGN learning. We demonstrate empirically that EM training of DGNs produces greater likelihood than VAE training. Our findings will guide the design of new VAE AVI that better approximate the true posterior and open avenues to apply standard statistical tools for model comparison, anomaly detection, and missing data imputation."
                },
                {
                    "title": "Sampling-Free Epistemic Uncertainty Estimation Using Approximated Variance Propagation",
                    "abstract": "We present a sampling-free approach for computing the epistemic uncertainty of a neural network. Epistemic uncertainty is an important quantity for the deployment of deep neural networks in safety-critical applications, since it represents how much one can trust predictions on new data. Recently promising works were proposed using noise injection combined with Monte-Carlo sampling at inference time to estimate this quantity (e.g. Monte-Carlo dropout). Our main contribution is an approximation of the epistemic uncertainty estimated by these methods that does not require sampling, thus notably reducing the computational overhead. We apply our approach to large-scale visual tasks (\\ie, semantic segmentation and depth regression) to demonstrate the advantages of our method compared to sampling-based approaches in terms of quality of the uncertainty estimates as well as of computational overhead."
                },
                {
                    "title": "A General Framework for Uncertainty Estimation in Deep Learning",
                    "abstract": "Neural networks predictions are unreliable when the input sample is out of the training distribution or corrupted by noise. Being able to detect such failures automatically is fundamental to integrate deep learning algorithms into robotics. Current approaches for uncertainty estimation of neural networks require changes to the network and optimization process, typically ignore prior knowledge about the data, and tend to make over-simplifying assumptions which underestimate uncertainty. To address these limitations, we propose a novel framework for uncertainty estimation. Based on Bayesian belief networks and Monte-Carlo sampling, our framework not only fully models the different sources of prediction uncertainty, but also incorporates prior data information, e.g.\u00a0sensor noise. We show theoretically that this gives us the ability to capture uncertainty better than existing methods. In addition, our framework has several desirable properties: (i) it is agnostic to the network architecture and task; (ii) it does not require changes in the optimization process; (iii) it can be applied to already trained architectures. We thoroughly validate the proposed framework through extensive experiments on both computer vision and control tasks, where we outperform previous methods by up to 23% in accuracy. The video available at https://youtu.be/X7n-bRS5vSM shows qualitative results of our experiments. The project's code is available at: https://tinyurl.com/s3nygw7."
                },
                {
                    "title": "On Evaluating Adversarial Robustness",
                    "abstract": "Correctly evaluating defenses against adversarial examples has proven to be extremely difficult. Despite the significant amount of recent work attempting to design defenses that withstand adaptive attacks, few have succeeded; most papers that propose defenses are quickly shown to be incorrect. \nWe believe a large contributing factor is the difficulty of performing security evaluations. In this paper, we discuss the methodological foundations, review commonly accepted best practices, and suggest new methods for evaluating defenses to adversarial examples. We hope that both researchers developing defenses as well as readers and reviewers who wish to understand the completeness of an evaluation consider our advice in order to avoid common pitfalls."
                },
                {
                    "title": "Single-Model Uncertainties for Deep Learning",
                    "abstract": "We provide single-model estimates of aleatoric and epistemic uncertainty for deep neural networks. To estimate aleatoric uncertainty, we propose Simultaneous Quantile Regression (SQR), a loss function to learn all the conditional quantiles of a given target variable. These quantiles can be used to compute well-calibrated prediction intervals. To estimate epistemic uncertainty, we propose Orthonormal Certificates (OCs), a collection of diverse non-constant functions that map all training samples to zero. These certificates map out-of-distribution examples to non-zero values, signaling epistemic uncertainty. Our uncertainty estimators are computationally attractive, as they do not require ensembling or retraining deep models, and achieve state-of-the-art performance."
                },
                {
                    "title": "Understanding Deep Neural Networks through Input Uncertainties",
                    "abstract": "Techniques for understanding the functioning of complex machine learning models are becoming increasingly popular, not only to improve the validation process, but also to extract new insights about the data via exploratory analysis. Though a large class of such tools currently exists, most assume that predictions are point estimates and use a sensitivity analysis of these estimates to interpret the model. Using lightweight probabilistic networks we show how including prediction uncertainties in the sensitivity analysis leads to: (i) more robust and generalizable models; and (ii) a new approach for model interpretation through uncertainty decomposition. In particular, we introduce a new regularization that takes both the mean and variance of a prediction into account and demonstrate that the resulting networks provide improved generalization to unseen data. Furthermore, we propose a new technique to explain prediction uncertainties through uncertainties in the input domain, thus providing new ways to validate and interpret deep learning models."
                },
                {
                    "title": "Deterministic Variational Inference for Robust Bayesian Neural Networks",
                    "abstract": "Bayesian neural networks (BNNs) hold great promise as a flexible and principled solution to deal with uncertainty when learning from finite data. Among approaches to realize probabilistic inference in deep neural networks, variational Bayes (VB) is theoretically grounded, generally applicable, and computationally efficient. With wide recognition of potential advantages, why is it that variational Bayes has seen very limited practical use for BNNs in real applications? We argue that variational inference in neural networks is fragile: successful implementations require careful initialization and tuning of prior variances, as well as controlling the variance of Monte Carlo gradient estimates. We provide two innovations that aim to turn VB into a robust inference tool for Bayesian neural networks: first, we introduce a novel deterministic method to approximate moments in neural networks, eliminating gradient variance; second, we introduce a hierarchical prior for parameters and a novel Empirical Bayes procedure for automatically selecting prior variances. Combining these two innovations, the resulting method is highly efficient and robust. On the application of heteroscedastic regression we demonstrate good predictive performance over alternative approaches."
                },
                {
                    "title": "Quasi-Monte Carlo Variational Inference",
                    "abstract": "Many machine learning problems involve Monte Carlo gradient estimators. As a prominent example, we focus on Monte Carlo variational inference (MCVI) in this paper. The performance of MCVI crucially depends on the variance of its stochastic gradients. We propose variance reduction by means of Quasi-Monte Carlo (QMC) sampling. QMC replaces N i.i.d. samples from a uniform probability distribution by a deterministic sequence of samples of length N. This sequence covers the underlying random variable space more evenly than i.i.d. draws, reducing the variance of the gradient estimator. With our novel approach, both the score function and the reparameterization gradient estimators lead to much faster convergence. We also propose a new algorithm for Monte Carlo objectives, where we operate with a constant learning rate and increase the number of QMC samples per iteration. We prove that this way, our algorithm can converge asymptotically at a faster rate than SGD. We furthermore provide theoretical guarantees on QMC for Monte Carlo objectives that go beyond MCVI, and support our findings by several experiments on large-scale data sets from various domains."
                },
                {
                    "title": "Accurate Uncertainties for Deep Learning Using Calibrated Regression",
                    "abstract": "Methods for reasoning under uncertainty are a key building block of accurate and reliable machine learning systems. Bayesian methods provide a general framework to quantify uncertainty. However, because of model misspecification and the use of approximate inference, Bayesian uncertainty estimates are often inaccurate -- for example, a 90% credible interval may not contain the true outcome 90% of the time. Here, we propose a simple procedure for calibrating any regression algorithm; when applied to Bayesian and probabilistic models, it is guaranteed to produce calibrated uncertainty estimates given enough data. Our procedure is inspired by Platt scaling and extends previous work on classification. We evaluate this approach on Bayesian linear regression, feedforward, and recurrent neural networks, and find that it consistently outputs well-calibrated credible intervals while improving performance on time series forecasting and model-based reinforcement learning tasks."
                },
                {
                    "title": "Lightweight Probabilistic Deep Networks",
                    "abstract": "Even though probabilistic treatments of neural networks have a long history, they have not found widespread use in practice. Sampling approaches are often too slow already for simple networks. The size of the inputs and the depth of typical CNN architectures in computer vision only compound this problem. Uncertainty in neural networks has thus been largely ignored in practice, despite the fact that it may provide important information about the reliability of predictions and the inner workings of the network. In this paper, we introduce two lightweight approaches to making supervised learning with probabilistic deep networks practical: First, we suggest probabilistic output layers for classification and regression that require only minimal changes to existing networks. Second, we employ assumed density filtering and show that activation uncertainties can be propagated in a practical fashion through the entire network, again with minor changes. Both probabilistic networks retain the predictive power of the deterministic counterpart, but yield uncertainties that correlate well with the empirical error induced by their predictions. Moreover, the robustness to adversarial examples is significantly increased."
                },
                {
                    "title": "Provable defenses against adversarial examples via the convex outer adversarial polytope",
                    "abstract": "We propose a method to learn deep ReLU-based classifiers that are provably robust against norm-bounded adversarial perturbations (on the training data; for previously unseen examples, the approach will be guaranteed to detect all adversarial examples, though it may flag some non-adversarial examples as well). The basic idea of the approach is to consider a convex outer approximation of the set of activations reachable through a norm-bounded perturbation, and we develop a robust optimization procedure that minimizes the worst case loss over this outer region (via a linear program). Crucially, we show that the dual problem to this linear program can be represented itself as a deep network similar to the backpropagation network, leading to very efficient optimization approaches that produce guaranteed bounds on the robust loss. The end result is that by executing a few more forward and backward passes through a slightly modified version of the original network (though possibly with much larger batch sizes), we can learn a classifier that is provably robust to any norm-bounded adversarial attack. We illustrate the approach on a toy 2D robust classification task, and on a simple convolutional architecture applied to MNIST, where we produce a classifier that provably has less than 8.4% test error for any adversarial attack with bounded $\\ell_\\infty$ norm less than $\\epsilon = 0.1$. This represents the largest verified network that we are aware of, and we discuss future challenges in scaling the approach to much larger domains."
                },
                {
                    "title": "Towards Deep Learning Models Resistant to Adversarial Attacks",
                    "abstract": "Recent work has demonstrated that deep neural networks are vulnerable to adversarial examples---inputs that are almost indistinguishable from natural data and yet classified incorrectly by the network. In fact, some of the latest findings suggest that the existence of adversarial attacks may be an inherent weakness of deep learning models. To address this problem, we study the adversarial robustness of neural networks through the lens of robust optimization. This approach provides us with a broad and unifying view on much of the prior work on this topic. Its principled nature also enables us to identify methods for both training and attacking neural networks that are reliable and, in a certain sense, universal. In particular, they specify a concrete security guarantee that would protect against any adversary. These methods let us train networks with significantly improved resistance to a wide range of adversarial attacks. They also suggest the notion of security against a first-order adversary as a natural and broad security guarantee. We believe that robustness against such well-defined classes of adversaries is an important stepping stone towards fully resistant deep learning models. Code and pre-trained models are available at this https URL and this https URL."
                },
                {
                    "title": "EMNIST: Extending MNIST to handwritten letters",
                    "abstract": "The MNIST dataset has become a standard benchmark for learning, classification and computer vision systems. Contributing to its widespread adoption are the understandable and intuitive nature of the task, the relatively small size and storage requirements and the accessibility and ease-of-use of the database itself. The MNIST database was derived from a larger dataset known as the NIST Special Database 19 which contains digits, uppercase and lowercase handwritten letters. This paper introduces a variant of the full NIST dataset, which we have called Extended MNIST (EMNIST), which follows the same conversion paradigm used to create the MNIST dataset. The result is a dataset that constitutes a more challenging classification task involving letters and digits, and one that shares the same image structure and parameters as the original MNIST task, allowing for direct compatibility with all existing classifiers and systems. Benchmark results using an online ELM algorithm are presented along with a validation of the conversion process through the comparison of the classification results on NIST digits and the MNIST digits."
                },
                {
                    "title": "Selective Classification for Deep Neural Networks",
                    "abstract": "Selective classification techniques (also known as reject option) have not yet been considered in the context of deep neural networks (DNNs). These techniques can potentially significantly improve DNNs prediction performance by trading-off coverage. In this paper we propose a method to construct a selective classifier given a trained neural network. Our method allows a user to set a desired risk level. At test time, the classifier rejects instances as needed, to grant the desired risk (with high probability). Empirical results over CIFAR and ImageNet convincingly demonstrate the viability of our method, which opens up possibilities to operate DNNs in mission-critical applications. For example, using our method an unprecedented 2% error in top-5 ImageNet classification can be guaranteed with probability 99.9%, and almost 60% test coverage."
                },
                {
                    "title": "What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?",
                    "abstract": "There are two major types of uncertainty one can model. Aleatoric uncertainty captures noise inherent in the observations. On the other hand, epistemic uncertainty accounts for uncertainty in the model - uncertainty which can be explained away given enough data. Traditionally it has been difficult to model epistemic uncertainty in computer vision, but with new Bayesian deep learning tools this is now possible. We study the benefits of modeling epistemic vs. aleatoric uncertainty in Bayesian deep learning models for vision tasks. For this we present a Bayesian deep learning framework combining input-dependent aleatoric uncertainty together with epistemic uncertainty. We study models under the framework with per-pixel semantic segmentation and depth regression tasks. Further, our explicit uncertainty formulation leads to new loss functions for these tasks, which can be interpreted as learned attenuation. This makes the loss more robust to noisy data, also giving new state-of-the-art results on segmentation and depth regression benchmarks."
                },
                {
                    "title": "Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles",
                    "abstract": "Deep neural networks (NNs) are powerful black box predictors that have recently achieved impressive performance on a wide spectrum of tasks. Quantifying predictive uncertainty in NNs is a challenging and yet unsolved problem. Bayesian NNs, which learn a distribution over weights, are currently the state-of-the-art for estimating predictive uncertainty; however these require significant modifications to the training procedure and are computationally expensive compared to standard (non-Bayesian) NNs. We propose an alternative to Bayesian NNs that is simple to implement, readily parallelizable, requires very little hyperparameter tuning, and yields high quality predictive uncertainty estimates. Through a series of experiments on classification and regression benchmarks, we demonstrate that our method produces well-calibrated uncertainty estimates which are as good or better than approximate Bayesian NNs. To assess robustness to dataset shift, we evaluate the predictive uncertainty on test examples from known and unknown distributions, and show that our method is able to express higher uncertainty on out-of-distribution examples. We demonstrate the scalability of our method by evaluating predictive uncertainty estimates on ImageNet."
                },
                {
                    "title": "Natural-Parameter Networks: A Class of Probabilistic Neural Networks",
                    "abstract": "Neural networks (NN) have achieved state-of-the-art performance in various applications. Unfortunately in applications where training data is insufficient, they are often prone to overfitting. One effective way to alleviate this problem is to exploit the Bayesian approach by using Bayesian neural networks (BNN). Another shortcoming of NN is the lack of flexibility to customize different distributions for the weights and neurons according to the data, as is often done in probabilistic graphical models. To address these problems, we propose a class of probabilistic neural networks, dubbed natural-parameter networks (NPN), as a novel and lightweight Bayesian treatment of NN. NPN allows the usage of arbitrary exponential-family distributions to model the weights and neurons. Different from traditional NN and BNN, NPN takes distributions as input and goes through layers of transformation before producing distributions to match the target output distributions. As a Bayesian treatment, efficient backpropagation (BP) is performed to learn the natural parameters for the distributions over both the weights and neurons. The output distributions of each layer, as byproducts, may be used as second-order representations for the associated tasks such as link prediction. Experiments on real-world datasets show that NPN can achieve state-of-the-art performance."
                },
                {
                    "title": "A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks",
                    "abstract": "We consider the two related problems of detecting if an example is misclassified or out-of-distribution. We present a simple baseline that utilizes probabilities from softmax distributions. Correctly classified examples tend to have greater maximum softmax probabilities than erroneously classified and out-of-distribution examples, allowing for their detection. We assess performance by defining several tasks in computer vision, natural language processing, and automatic speech recognition, showing the effectiveness of this baseline across all. We then show the baseline can sometimes be surpassed, demonstrating the room for future research on these underexplored detection tasks."
                },
                {
                    "title": "Gaussian Error Linear Units (GELUs)",
                    "abstract": "We propose the Gaussian Error Linear Unit (GELU), a high-performing neural network activation function. The GELU activation function is $x\\Phi(x)$, where $\\Phi(x)$ the standard Gaussian cumulative distribution function. The GELU nonlinearity weights inputs by their value, rather than gates inputs by their sign as in ReLUs ($x\\mathbf{1}_{x>0}$). We perform an empirical evaluation of the GELU nonlinearity against the ReLU and ELU activations and find performance improvements across all considered computer vision, natural language processing, and speech tasks."
                },
                {
                    "title": "Assumed Density Filtering Methods for Learning Bayesian Neural Networks",
                    "abstract": "\n \n Buoyed by the success of deep multilayer neural networks, there is renewed interest in scalable learning of Bayesian neural networks. Here, we study algorithms that utilize recent advances in Bayesian inference to efficiently learn distributions over network weights. In particular, we focus on recently proposed assumed density filtering based methods for learning Bayesian neural networks -- Expectation and Probabilistic backpropagation. Apart from scaling to large datasets, these techniques seamlessly deal with non-differentiable activation functions and provide parameter (learning rate, momentum) free learning. In this paper, we first rigorously compare the two algorithms and in the process develop several extensions, including a version of EBP for continuous regression problems and a PBP variant for binary classification. Next, we extend both algorithms to deal with multiclass classification and count regression problems. On a variety of diverse real world benchmarks, we find our extensions to be effective, achieving results competitive with the state-of-the-art.\n \n"
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Robust Convolutional Neural Networks under Adversarial Noise",
                    "abstract": "Recent studies have shown that Convolutional Neural Networks (CNNs) are vulnerable to a small perturbation of input called \"adversarial examples\". In this work, we propose a new feedforward CNN that improves robustness in the presence of adversarial noise. Our model uses stochastic additive noise added to the input image and to the CNN models. The proposed model operates in conjunction with a CNN trained with either standard or adversarial objective function. In particular, convolution, max-pooling, and ReLU layers are modified to benefit from the noise model. Our feedforward model is parameterized by only a mean and variance per pixel which simplifies computations and makes our method scalable to a deep architecture. From CIFAR-10 and ImageNet test, the proposed model outperforms other methods and the improvement is more evident for difficult classification tasks or stronger adversarial noise."
                },
                {
                    "title": "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning",
                    "abstract": "Deep learning tools have gained tremendous attention in applied machine learning. However such tools for regression and classification do not capture model uncertainty. In comparison, Bayesian models offer a mathematically grounded framework to reason about model uncertainty, but usually come with a prohibitive computational cost. In this paper we develop a new theoretical framework casting dropout training in deep neural networks (NNs) as approximate Bayesian inference in deep Gaussian processes. A direct result of this theory gives us tools to model uncertainty with dropout NNs -- extracting information from existing models that has been thrown away so far. This mitigates the problem of representing uncertainty in deep learning without sacrificing either computational complexity or test accuracy. We perform an extensive study of the properties of dropout's uncertainty. Various network architectures and non-linearities are assessed on tasks of regression and classification, using MNIST as an example. We show a considerable improvement in predictive log-likelihood and RMSE compared to existing state-of-the-art methods, and finish by using dropout's uncertainty in deep reinforcement learning."
                },
                {
                    "title": "Weight Uncertainty in Neural Networks",
                    "abstract": "We introduce a new, efficient, principled and backpropagation-compatible algorithm for learning a probability distribution on the weights of a neural network, called Bayes by Backprop. It regularises the weights by minimising a compression cost, known as the variational free energy or the expected lower bound on the marginal likelihood. We show that this principled kind of regularisation yields comparable performance to dropout on MNIST classification. We then demonstrate how the learnt uncertainty in the weights can be used to improve generalisation in non-linear regression problems, and how this weight uncertainty can be used to drive the exploration-exploitation trade-off in reinforcement learning."
                },
                {
                    "title": "Probabilistic Backpropagation for Scalable Learning of Bayesian Neural Networks",
                    "abstract": "Large multilayer neural networks trained with backpropagation have recently achieved state-of-the-art results in a wide range of problems. However, using backprop for neural net learning still has some disadvantages, e.g., having to tune a large number of hyperparameters to the data, lack of calibrated probabilistic predictions, and a tendency to overfit the training data. In principle, the Bayesian approach to learning neural networks does not have these problems. However, existing Bayesian techniques lack scalability to large dataset and network sizes. In this work we present a novel scalable method for learning Bayesian neural networks, called probabilistic backpropagation (PBP). Similar to classical backpropagation, PBP works by computing a forward propagation of probabilities through the network and then doing a backward computation of gradients. A series of experiments on ten real-world datasets show that PBP is significantly faster than other techniques, while offering competitive predictive abilities. Our experiments also show that PBP provides accurate estimates of the posterior variance on the network weights."
                },
                {
                    "title": "Fast dropout training",
                    "abstract": "Preventing feature co-adaptation by encouraging independent contributions from different features often improves classification and regression performance. Dropout training (Hinton et al., 2012) does this by randomly dropping out (zeroing) hidden units and input features during training of neural networks. However, repeatedly sampling a random subset of input features makes training much slower. Based on an examination of the implied objective function of dropout training, we show how to do fast dropout training by sampling from or integrating a Gaussian approximation, instead of doing Monte Carlo optimization of this objective. This approximation, justified by the central limit theorem and empirical evidence, gives an order of magnitude speedup and more stability. We show how to do fast dropout training for classification, regression, and multilayer neural networks. Beyond dropout, our technique is extended to integrate out other types of noise and small image transformations."
                },
                {
                    "title": "Practical Variational Inference for Neural Networks",
                    "abstract": "Variational methods have been previously explored as a tractable approximation to Bayesian inference for neural networks. However the approaches proposed so far have only been applicable to a few simple network architectures. This paper introduces an easy-to-implement stochastic variational method (or equivalently, minimum description length loss function) that can be applied to most neural networks. Along the way it revisits several common regularisers from a variational perspective. It also provides a simple pruning heuristic that can both drastically reduce the number of network weights and lead to improved generalisation. Experimental results are provided for a hierarchical multidimensional recurrent neural network applied to the TIMIT speech corpus."
                },
                {
                    "title": "Uncertainty in Deep Learning",
                    "abstract": "Deep learning has attracted tremendous attention from researchers in various fields of information engineering such as AI, computer vision, and language processing [Kalchbrenner and Blunsom, 2013; Krizhevsky et al., 2012; Mnih et al., 2013], but also from more traditional sciences such as physics, biology, and manufacturing [Anjos et al., 2015; Baldi et al., 2014; Bergmann et al., 2014]. Neural networks, image processing tools such as convolutional neural networks, sequence processing models such as recurrent neural networks, and regularisation tools such as dropout, are used extensively. However, fields such as physics, biology, and manufacturing are ones in which representing model uncertainty is of crucial importance [Ghahramani, 2015; Krzywinski and Altman, 2013]. With the recent shift in many of these fields towards the use of Bayesian uncertainty [Herzog and Ostwald, 2013; Nuzzo, 2014; Trafimow and Marks, 2015], new needs arise from deep learning. In this work we develop tools to obtain practical uncertainty estimates in deep learning, casting recent deep learning tools as Bayesian models without changing either the models or the optimisation. In the first part of this thesis we develop the theory for such tools, providing applications and illustrative examples. We tie approximate inference in Bayesian models to dropout and other stochastic regularisation techniques, and assess the approximations empirically. We give example applications arising from this connection between modern deep learning and Bayesian modelling such as active learning of image data and data efficient deep reinforcement learning. We further demonstrate the method\u2019s practicality through a survey of recent applications making use of the suggested tools in language applications, medical diagnostics, bioinformatics, image processing, and autonomous driving. In the second part of the thesis we explore its theoretical implications, and the insights stemming from the link between Bayesian modelling and deep learning. We discuss what determines model uncertainty properties, analyse the approximate inference analytically in the linear case, and theoretically examine various priors such as spike and slab priors."
                },
                {
                    "title": "Variational Learning in Nonlinear Gaussian Belief Networks",
                    "abstract": "We view perceptual tasks such as vision and speech recognition as inference problems where the goal is to estimate the posterior distribution over latent variables (e.g., depth in stereo vision) given the sensory input. The recent flurry of research in independent component analysis exemplifies the importance of inferring the continuous-valued latent variables of input data. The latent variables found by this method are linearly related to the input, but perception requires nonlinear inferences such as classification and depth estimation. In this article, we present a unifying framework for stochastic neural networks with nonlinear latent variables. Nonlinear units are obtained by passing the outputs of linear gaussian units through various nonlinearities. We present a general variational method that maximizes a lower bound on the likelihood of a training set and give results on two visual feature extraction problems. We also show how the variational method can be used for pattern classification and compare the performance of these nonlinear networks with other methods on the problem of handwritten digit recognition."
                },
                {
                    "title": "Estimating the mean and variance of the target probability distribution",
                    "abstract": "Introduces a method that estimates the mean and the variance of the probability distribution of the target as a function of the input, given an assumed target error-distribution model. Through the activation of an auxiliary output unit, this method provides a measure of the uncertainty of the usual network output for each input pattern. The authors derive the cost function and weight-update equations for the example of a Gaussian target error distribution, and demonstrate the feasibility of the network on a synthetic problem where the true input-dependent noise level is known.<<ETX>>"
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively integrate differentiable algorithms into machine learning frameworks to enhance model performance and interpretability?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it can lead to more robust and interpretable machine learning models. By integrating differentiable algorithms, we can improve optimization processes, enabling models to learn more complex patterns and relationships in data. This advancement could significantly impact various applications, from computer vision to natural language processing, by providing models that not only perform better but also offer insights into their decision-making processes. Furthermore, it could pave the way for future research into hybrid models that combine the strengths of traditional algorithms with modern machine learning techniques.\n\n### [Question 3] - Why is it hard?\nThe challenges in solving this problem stem from the inherent complexity of differentiable algorithms and their integration into existing machine learning frameworks. Naive approaches may fail due to the difficulty in ensuring that the entire pipeline remains differentiable, which is essential for gradient-based optimization. Additionally, there are technical obstacles such as managing computational efficiency and stability during training, as well as theoretical challenges related to the convergence of these integrated models. Practical issues, such as the need for extensive tuning and the potential for increased computational overhead, further complicate the implementation of these approaches.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often focused on either traditional machine learning algorithms or deep learning models, with limited exploration of their integration. Gaps in understanding how to effectively combine these approaches have hindered progress. Barriers such as a lack of standardized frameworks for differentiable algorithms and insufficient theoretical foundations for their application in machine learning have also played a role. Our approach differs by proposing a systematic methodology that emphasizes the seamless integration of differentiable algorithms into existing machine learning architectures, addressing both the theoretical and practical limitations of prior work.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves developing a framework that incorporates differentiable algorithms into neural network architectures. We will utilize benchmark datasets such as MNIST and EMNIST for evaluation, focusing on metrics like accuracy and interpretability. The expected outcomes include improved model performance on standard tasks, enhanced interpretability of model decisions, and a clearer understanding of the interactions between differentiable algorithms and neural networks. By demonstrating the effectiveness of our approach, we aim to provide a foundation for future research in this area."
            }
        },
        "author_data": {
            "34cae5b9-7890-4b0a-9ae8-c1bd61d374cb": {
                "pk": "34cae5b9-7890-4b0a-9ae8-c1bd61d374cb",
                "name": "Felix Petersen",
                "collaborators": [
                    "O. Deussen",
                    "Hilde Kuehne",
                    "C. Borgelt",
                    "Yuekai Sun",
                    "Bastian Goldluecke",
                    "Debarghya Mukherjee",
                    "M. Yurochkin",
                    "M. Schubotz",
                    "Andr\u00e9 Greiner-Petter",
                    "Bela Gipp"
                ],
                "domain": [
                    "Neural Networks",
                    "Differentiable Algorithms",
                    "Machine Learning",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "Neural Machine Translation for Mathematical Formulae",
                        "abstract": "We tackle the problem of neural machine translation of mathematical formulae between ambiguous presentation languages and unambiguous content languages. Compared to neural machine translation on natural language, mathematical formulae have a much smaller vocabulary and much longer sequences of symbols, while their translation requires extreme precision to satisfy mathematical information needs. In this work, we perform the tasks of translating from LaTeX to Mathematica as well as from LaTeX to semantic LaTeX. While recurrent, recursive, and transformer networks struggle with preserving all contained information, we find that convolutional sequence-to-sequence networks achieve 95.1% and 90.7% exact matches, respectively."
                    },
                    {
                        "title": "Grounding Everything: Emerging Localization Properties in Vision-Language Transformers",
                        "abstract": "Vision-language foundation models have shown remark-able performance in various zero-shot settings such as im-age retrieval, classification, or captioning. But so far, those models seem to fall behind when it comes to zero-shot lo-calization of referential expressions and objects in images. As a result, they need to be fine-tuned for this task. In this paper, we show that pretrained vision-language (VL) models allow for zero-shot open-vocabulary object local-ization without any fine-tuning. To leverage those capabil-ities, we propose a Grounding Everything Module (GEM) that generalizes the idea of value-value attention introduced by CLIPSurgery [17] to a self-self attention path. We show that the concept of self-self attention corresponds to clustering, thus enforcing groups of tokens arising from the same object to be similar while preserving the alignment with the language space. To further guide the group formation, we propose a set of regularizations that allows the model to finally generalize across datasets and backbones. We evaluate the proposed GEM framework on various bench-mark tasks and datasets for semantic segmentation. GEM not only outperforms other training-free open-vocabulary localization methods, but also achieves state-of-the-art re-sults on the recently proposed OpenImagesV7 large-scale segmentation benchmark.111 Code is available at https://github.com/WalBouss/GEM."
                    },
                    {
                        "title": "Learning by Sorting: Self-supervised Learning with Group Ordering Constraints",
                        "abstract": "Contrastive learning has become an important tool in learning representations from unlabeled data mainly relying on the idea of minimizing distance between positive data pairs, e.g., views from the same images, and maximizing distance between negative data pairs, e.g., views from different images. This paper proposes a new variation of the contrastive learning objective, Group Ordering Constraints (GroCo), that leverages the idea of sorting the distances of positive and negative pairs and computing the respective loss based on how many positive pairs have a larger distance than the negative pairs, and thus are not ordered correctly. To this end, the GroCo loss is based on differentiable sorting networks, which enable training with sorting supervision by matching a differentiable permutation matrix, which is produced by sorting a given set of scores, to a respective ground truth permutation matrix. Applying this idea to groupwise pre-ordered inputs of multiple positive and negative pairs allows introducing the GroCo loss with implicit emphasis on strong positives and negatives, leading to better optimization of the local neighborhood. We evaluate the proposed formulation on various self-supervised learning benchmarks and show that it not only leads to improved results compared to vanilla contrastive learning but also shows competitive performance to comparable methods in linear probing and outperforms current methods in k-NN performance. 1"
                    },
                    {
                        "title": "ISAAC Newton: Input-based Approximate Curvature for Newton's Method",
                        "abstract": "We present ISAAC (Input-baSed ApproximAte Curvature), a novel method that conditions the gradient using selected second-order information and has an asymptotically vanishing computational overhead, assuming a batch size smaller than the number of neurons. We show that it is possible to compute a good conditioner based on only the input to a respective layer without a substantial computational overhead. The proposed method allows effective training even in small-batch stochastic regimes, which makes it competitive to first-order as well as second-order methods."
                    },
                    {
                        "title": "Monotonic Differentiable Sorting Networks",
                        "abstract": "Differentiable sorting algorithms allow training with sorting and ranking supervision, where only the ordering or ranking of samples is known. Various methods have been proposed to address this challenge, ranging from optimal transport-based differentiable Sinkhorn sorting algorithms to making classic sorting networks differentiable. One problem of current differentiable sorting methods is that they are non-monotonic. To address this issue, we propose a novel relaxation of conditional swap operations that guarantees monotonicity in differentiable sorting networks. We introduce a family of sigmoid functions and prove that they produce differentiable sorting networks that are monotonic. Monotonicity ensures that the gradients always have the correct sign, which is an advantage in gradient-based optimization. We demonstrate that monotonic differentiable sorting networks improve upon previous differentiable sorting methods."
                    },
                    {
                        "title": "Differentiable Top-k Classification Learning",
                        "abstract": "The top-k classification accuracy is one of the core metrics in machine learning. Here, k is conventionally a positive integer, such as 1 or 5, leading to top-1 or top-5 training objectives. In this work, we relax this assumption and optimize the model for multiple k simultaneously instead of using a single k. Leveraging recent advances in differentiable sorting and ranking, we propose a differentiable top-k cross-entropy classification loss. This allows training the network while not only considering the top-1 prediction, but also, e.g., the top-2 and top-5 predictions. We evaluate the proposed loss function for fine-tuning on state-of-the-art architectures, as well as for training from scratch. We find that relaxing k does not only produce better top-5 accuracies, but also leads to top-1 accuracy improvements. When fine-tuning publicly available ImageNet models, we achieve a new state-of-the-art for these models."
                    },
                    {
                        "title": "GenDR: A Generalized Differentiable Renderer",
                        "abstract": "In this work, we present and study a generalized family of differentiable renderers. We discuss from scratch which components are necessary for differentiable rendering and formalize the requirements for each component. We instantiate our general differentiable renderer, which generalizes existing differentiable renderers like SoftRas and DIB-R, with an array of different smoothing distributions to cover a large spectrum of reasonable settings. We evaluate an array of differentiable renderer instantiations on the popular ShapeNet 3D reconstruction benchmark and analyze the implications of our results. Surprisingly, the simple uniform distribution yields the best overall results when averaged over 13 classes; in general, however, the optimal choice of distribution heavily depends on the task."
                    },
                    {
                        "title": "Learning with Differentiable Algorithms",
                        "abstract": "Classic algorithms and machine learning systems like neural networks are both abundant in everyday life. While classic computer science algorithms are suitable for precise execution of exactly defined tasks such as finding the shortest path in a large graph, neural networks allow learning from data to predict the most likely answer in more complex tasks such as image classification, which cannot be reduced to an exact algorithm. To get the best of both worlds, this thesis explores combining both concepts leading to more robust, better performing, more interpretable, more computationally efficient, and more data efficient architectures. The thesis formalizes the idea of algorithmic supervision, which allows a neural network to learn from or in conjunction with an algorithm. When integrating an algorithm into a neural architecture, it is important that the algorithm is differentiable such that the architecture can be trained end-to-end and gradients can be propagated back through the algorithm in a meaningful way. To make algorithms differentiable, this thesis proposes a general method for continuously relaxing algorithms by perturbing variables and approximating the expectation value in closed form, i.e., without sampling. In addition, this thesis proposes differentiable algorithms, such as differentiable sorting networks, differentiable renderers, and differentiable logic gate networks. Finally, this thesis presents alternative training strategies for learning with algorithms."
                    },
                    {
                        "title": "Deep Differentiable Logic Gate Networks",
                        "abstract": "Recently, research has increasingly focused on developing efficient neural network architectures. In this work, we explore logic gate networks for machine learning tasks by learning combinations of logic gates. These networks comprise logic gates such as\"AND\"and\"XOR\", which allow for very fast execution. The difficulty in learning logic gate networks is that they are conventionally non-differentiable and therefore do not allow training with gradient descent. Thus, to allow for effective training, we propose differentiable logic gate networks, an architecture that combines real-valued logics and a continuously parameterized relaxation of the network. The resulting discretized logic gate networks achieve fast inference speeds, e.g., beyond a million images of MNIST per second on a single CPU core."
                    },
                    {
                        "title": "Domain Adaptation meets Individual Fairness. And they get along",
                        "abstract": "Many instances of algorithmic bias are caused by distributional shifts. For example, machine learning (ML) models often perform worse on demographic groups that are underrepresented in the training data. In this paper, we leverage this connection between algorithmic fairness and distribution shifts to show that algorithmic fairness interventions can help ML models overcome distribution shifts, and that domain adaptation methods (for overcoming distribution shifts) can mitigate algorithmic biases. In particular, we show that (i) enforcing suitable notions of individual fairness (IF) can improve the out-of-distribution accuracy of ML models under the covariate shift assumption and that (ii) it is possible to adapt representation alignment methods for domain adaptation to enforce individual fairness. The former is unexpected because IF interventions were not developed with distribution shifts in mind. The latter is also unexpected because representation alignment is not a common approach in the individual fairness literature."
                    },
                    {
                        "title": "Learning with Algorithmic Supervision via Continuous Relaxations",
                        "abstract": "The integration of algorithmic components into neural architectures has gained increased attention recently, as it allows training neural networks with new forms of supervision such as ordering constraints or silhouettes instead of using ground truth labels. Many approaches in the field focus on the continuous relaxation of a specific task and show promising results in this context. But the focus on single tasks also limits the applicability of the proposed concepts to a narrow range of applications. In this work, we build on those ideas to propose an approach that allows to integrate algorithms into end-to-end trainable neural network architectures based on a general approximation of discrete conditions. To this end, we relax these conditions in control structures such as conditional statements, loops, and indexing, so that resulting algorithms are smoothly differentiable. To obtain meaningful gradients, each relevant variable is perturbed via logistic distributions and the expectation value under this perturbation is approximated. We evaluate the proposed continuous relaxation model on four challenging tasks and show that it can keep up with relaxations specifically designed for each individual task."
                    },
                    {
                        "title": "Post-processing for Individual Fairness",
                        "abstract": "Post-processing in algorithmic fairness is a versatile approach for correcting bias in ML systems that are already used in production. The main appeal of post-processing is that it avoids expensive retraining. In this work, we propose general post-processing algorithms for individual fairness (IF). We consider a setting where the learner only has access to the predictions of the original model and a similarity graph between individuals, guiding the desired fairness constraints. We cast the IF post-processing problem as a graph smoothing problem corresponding to graph Laplacian regularization that preserves the desired\"treat similar individuals similarly\"interpretation. Our theoretical results demonstrate the connection of the new objective function to a local relaxation of the original individual fairness. Empirically, our post-processing algorithms correct individual biases in large-scale NLP models such as BERT, while preserving accuracy."
                    },
                    {
                        "title": "Style Agnostic 3D Reconstruction via Adversarial Style Transfer",
                        "abstract": "Reconstructing the 3D geometry of an object from an image is a major challenge in computer vision. Recently introduced differentiable renderers can be leveraged to learn the 3D geometry of objects from 2D images, but those approaches require additional supervision to enable the renderer to produce an output that can be compared to the input image. This can be scene information or constraints such as object silhouettes, uniform backgrounds, material, texture, and lighting. In this paper, we propose an approach that enables a differentiable rendering-based learning of 3D objects from images with backgrounds without the need for silhouette supervision. Instead of trying to render an image close to the input, we propose an adversarial style-transfer and domain adaptation pipeline that allows to translate the input image domain to the rendered image domain. This allows us to directly compare between a translated image and the differentiable rendering of a 3D object reconstruction in order to train the 3D object reconstruction network. We show that the approach learns 3D geometry from images with backgrounds and provides a better performance than constrained methods for single-view 3D object reconstruction on this task."
                    },
                    {
                        "title": "Differentiable Sorting Networks for Scalable Sorting and Ranking Supervision",
                        "abstract": "Sorting and ranking supervision is a method for training neural networks end-to-end based on ordering constraints. That is, the ground truth order of sets of samples is known, while their absolute values remain unsupervised. For that, we propose differentiable sorting networks by relaxing their pairwise conditional swap operations. To address the problems of vanishing gradients and extensive blurring that arise with larger numbers of layers, we propose mapping activations to regions with moderate gradients. We consider odd-even as well as bitonic sorting networks, which outperform existing relaxations of the sorting operation. We show that bitonic sorting networks can achieve stable training on large input sets of up to 1024 elements."
                    },
                    {
                        "title": "AlgoNet: C Smooth Algorithmic Neural Networks for Solving Inverse Problems",
                        "abstract": "Artificial neural networks revolutionized many areas of computer science in recent years since they provide solutions to a number of previously unsolved problems. On the other hand, for many problems, classic algorithms exist, which typically exceed the accuracy and stability of neural networks. To combine these two concepts, we present a new kind of neural networks\u2014algorithmic neural networks (AlgoNets). These networks integrate smooth versions of classic algorithms into the topology of neural networks. Our novel reconstructive adversarial network (RAN) enables solving inverse problems without or with only weak supervision."
                    },
                    {
                        "title": "AlgoNet: C\u221e Smooth Algorithmic Neural Networks",
                        "abstract": "Artificial neural networks revolutionized many areas of computer science in recent years since they provide solutions to a number of previously unsolved problems. On the other hand, for many problems, classic algorithms exist, which typically exceed the accuracy and stability of neural networks. To combine these two concepts, we present a new kind of neural networks\u2014algorithmic neural networks (AlgoNets). These networks integrate smooth versions of classic algorithms into the topology of neural networks. A forward AlgoNet includes algorithmic layers into existing architectures while a backward AlgoNet can solve inverse problems without or with only weak supervision. In addition, we present the algonet package, a PyTorch based library that includes, inter alia, a smoothly evaluated programming language, a smooth 3D mesh renderer, and smooth sorting algorithms."
                    },
                    {
                        "title": "Pix2Vex: Image-to-Geometry Reconstruction using a Smooth Differentiable Renderer",
                        "abstract": "The long-coveted task of reconstructing 3D geometry from images is still a standing problem. In this paper, we build on the power of neural networks and introduce Pix2Vex, a network trained to convert camera-captured images into 3D geometry. We present a novel differentiable renderer ($DR$) as a forward validation means during training. Our key insight is that $DR$s produce images of a particular appearance, different from typical input images. Hence, we propose adding an image-to-image translation component, converting between these rendering styles. This translation closes the training loop, while allowing to use minimal supervision only, without needing any 3D model as ground truth. Unlike state-of-the-art methods, our $DR$ is $C^\\infty$ smooth and thus does not display any discontinuities at occlusions or dis-occlusions. Through our novel training scheme, our network can train on different types of images, where previous work can typically only train on images of a similar appearance to those rendered by a $DR$."
                    },
                    {
                        "title": "AlgoNet: $C^\\infty$ Smooth Algorithmic Neural Networks",
                        "abstract": "Artificial neural networks revolutionized many areas of computer science in recent years since they provide solutions to a number of previously unsolved problems. On the other hand, for many problems, classic algorithms exist, which typically exceed the accuracy and stability of neural networks. To combine these two concepts, we present a new kind of neural networks$-$algorithmic neural networks (AlgoNets). These networks integrate smooth versions of classic algorithms into the topology of neural networks. A forward AlgoNet includes algorithmic layers into existing architectures while a backward AlgoNet can solve inverse problems without or with only weak supervision. In addition, we present the $\\texttt{algonet}$ package, a PyTorch based library that includes, inter alia, a smoothly evaluated programming language, a smooth 3D mesh renderer, and smooth sorting algorithms."
                    }
                ]
            },
            "2bcf7b7b-7093-473c-9701-62b8638fae51": {
                "pk": "2bcf7b7b-7093-473c-9701-62b8638fae51",
                "name": "Aashwin Mishra",
                "collaborators": [
                    "Felix Petersen",
                    "Christian Borgelt",
                    "Stefano Ermon"
                ],
                "domain": [
                    "Stochastic Optimization",
                    "Gradient Estimation",
                    "Non-differentiable Functions",
                    "Variance Reduction"
                ],
                "publications": [
                    {
                        "title": "Generalizing Stochastic Smoothing for Differentiation and Gradient Estimation",
                        "abstract": "We deal with the problem of gradient estimation for stochastic differentiable relaxations of algorithms, operators, simulators, and other non-differentiable functions. Stochastic smoothing conventionally perturbs the input of a non-differentiable function with a differentiable density distribution with full support, smoothing it and enabling gradient estimation. Our theory starts at first principles to derive stochastic smoothing with reduced assumptions, without requiring a differentiable density nor full support, and we present a general framework for relaxation and gradient estimation of non-differentiable black-box functions $f:\\mathbb{R}^n\\to\\mathbb{R}^m$. We develop variance reduction for gradient estimation from 3 orthogonal perspectives. Empirically, we benchmark 6 distributions and up to 24 variance reduction strategies for differentiable sorting and ranking, differentiable shortest-paths on graphs, differentiable rendering for pose estimation, as well as differentiable cryo-ET simulations."
                    }
                ]
            },
            "76530caf-0baa-4de4-a762-775b2cfbb3e3": {
                "pk": "76530caf-0baa-4de4-a762-775b2cfbb3e3",
                "name": "Hilde Kuehne",
                "collaborators": [
                    "Walid Bousselham",
                    "Felix Petersen",
                    "Sofian Chaybouti",
                    "Vittorio Ferrari",
                    "Christian Borgelt",
                    "Tobias Sutter",
                    "Oliver Deussen",
                    "Stefano Ermon",
                    "Christian Rupprecht",
                    "Angie Boggust"
                ],
                "domain": [
                    "Computer Vision",
                    "Explainability",
                    "Vision-Language Models",
                    "Neural Networks"
                ],
                "publications": [
                    {
                        "title": "Newton Losses: Using Curvature Information for Learning with Differentiable Algorithms",
                        "abstract": "When training neural networks with custom objectives, such as ranking losses and shortest-path losses, a common problem is that they are, per se, non-differentiable. A popular approach is to continuously relax the objectives to provide gradients, enabling learning. However, such differentiable relaxations are often non-convex and can exhibit vanishing and exploding gradients, making them (already in isolation) hard to optimize. Here, the loss function poses the bottleneck when training a deep neural network. We present Newton Losses, a method for improving the performance of existing hard to optimize losses by exploiting their second-order information via their empirical Fisher and Hessian matrices. Instead of training the neural network with second-order techniques, we only utilize the loss function's second-order information to replace it by a Newton Loss, while training the network with gradient descent. This makes our method computationally efficient. We apply Newton Losses to eight differentiable algorithms for sorting and shortest-paths, achieving significant improvements for less-optimized differentiable algorithms, and consistent improvements, even for well-optimized differentiable algorithms."
                    },
                    {
                        "title": "MaskInversion: Localized Embeddings via Optimization of Explainability Maps",
                        "abstract": "Vision-language foundation models such as CLIP have achieved tremendous results in global vision-language alignment, but still show some limitations in creating representations for specific image regions. % To address this problem, we propose MaskInversion, a method that leverages the feature representations of pre-trained foundation models, such as CLIP, to generate a context-aware embedding for a query image region specified by a mask at test time. MaskInversion starts with initializing an embedding token and compares its explainability map, derived from the foundation model, to the query mask. The embedding token is then subsequently refined to approximate the query region by minimizing the discrepancy between its explainability map and the query mask. During this process, only the embedding vector is updated, while the underlying foundation model is kept frozen allowing to use MaskInversion with any pre-trained model. As deriving the explainability map involves computing its gradient, which can be expensive, we propose a gradient decomposition strategy that simplifies this computation. The learned region representation can be used for a broad range of tasks, including open-vocabulary class retrieval, referring expression comprehension, as well as for localized captioning and image generation. We evaluate the proposed method on all those tasks on several datasets such as PascalVOC, MSCOCO, RefCOCO, and OpenImagesV7 and show its capabilities compared to other SOTA approaches."
                    },
                    {
                        "title": "LeGrad: An Explainability Method for Vision Transformers via Feature Formation Sensitivity",
                        "abstract": "Vision Transformers (ViTs), with their ability to model long-range dependencies through self-attention mechanisms, have become a standard architecture in computer vision. However, the interpretability of these models remains a challenge. To address this, we propose LeGrad, an explainability method specifically designed for ViTs. LeGrad computes the gradient with respect to the attention maps of ViT layers, considering the gradient itself as the explainability signal. We aggregate the signal over all layers, combining the activations of the last as well as intermediate tokens to produce the merged explainability map. This makes LeGrad a conceptually simple and an easy-to-implement tool for enhancing the transparency of ViTs. We evaluate LeGrad in challenging segmentation, perturbation, and open-vocabulary settings, showcasing its versatility compared to other SotA explainability methods demonstrating its superior spatial fidelity and robustness to perturbations. A demo and the code is available at https://github.com/WalBouss/LeGrad."
                    },
                    {
                        "title": "Grounding Everything: Emerging Localization Properties in Vision-Language Transformers",
                        "abstract": "Vision-language foundation models have shown remark-able performance in various zero-shot settings such as im-age retrieval, classification, or captioning. But so far, those models seem to fall behind when it comes to zero-shot lo-calization of referential expressions and objects in images. As a result, they need to be fine-tuned for this task. In this paper, we show that pretrained vision-language (VL) models allow for zero-shot open-vocabulary object local-ization without any fine-tuning. To leverage those capabil-ities, we propose a Grounding Everything Module (GEM) that generalizes the idea of value-value attention introduced by CLIPSurgery [17] to a self-self attention path. We show that the concept of self-self attention corresponds to clustering, thus enforcing groups of tokens arising from the same object to be similar while preserving the alignment with the language space. To further guide the group formation, we propose a set of regularizations that allows the model to finally generalize across datasets and backbones. We evaluate the proposed GEM framework on various bench-mark tasks and datasets for semantic segmentation. GEM not only outperforms other training-free open-vocabulary localization methods, but also achieves state-of-the-art re-sults on the recently proposed OpenImagesV7 large-scale segmentation benchmark.111 Code is available at https://github.com/WalBouss/GEM."
                    }
                ]
            },
            "5447db5e-24ae-4b79-b966-b396fbe0ced3": {
                "pk": "5447db5e-24ae-4b79-b966-b396fbe0ced3",
                "name": "Christian Borgelt",
                "collaborators": [
                    "Felix Petersen",
                    "Stefano Ermon",
                    "Tobias Sutter",
                    "Hilde Kuehne",
                    "Oliver Deussen",
                    "Aashwin Mishra",
                    "A. N\u00fcrnberger",
                    "A. Klose"
                ],
                "domain": [
                    "Machine Learning",
                    "Neural Networks",
                    "Optimization",
                    "Stochastic Methods"
                ],
                "publications": [
                    {
                        "title": "Newton Losses: Using Curvature Information for Learning with Differentiable Algorithms",
                        "abstract": "When training neural networks with custom objectives, such as ranking losses and shortest-path losses, a common problem is that they are, per se, non-differentiable. A popular approach is to continuously relax the objectives to provide gradients, enabling learning. However, such differentiable relaxations are often non-convex and can exhibit vanishing and exploding gradients, making them (already in isolation) hard to optimize. Here, the loss function poses the bottleneck when training a deep neural network. We present Newton Losses, a method for improving the performance of existing hard to optimize losses by exploiting their second-order information via their empirical Fisher and Hessian matrices. Instead of training the neural network with second-order techniques, we only utilize the loss function's second-order information to replace it by a Newton Loss, while training the network with gradient descent. This makes our method computationally efficient. We apply Newton Losses to eight differentiable algorithms for sorting and shortest-paths, achieving significant improvements for less-optimized differentiable algorithms, and consistent improvements, even for well-optimized differentiable algorithms."
                    },
                    {
                        "title": "Generalizing Stochastic Smoothing for Differentiation and Gradient Estimation",
                        "abstract": "We deal with the problem of gradient estimation for stochastic differentiable relaxations of algorithms, operators, simulators, and other non-differentiable functions. Stochastic smoothing conventionally perturbs the input of a non-differentiable function with a differentiable density distribution with full support, smoothing it and enabling gradient estimation. Our theory starts at first principles to derive stochastic smoothing with reduced assumptions, without requiring a differentiable density nor full support, and we present a general framework for relaxation and gradient estimation of non-differentiable black-box functions $f:\\mathbb{R}^n\\to\\mathbb{R}^m$. We develop variance reduction for gradient estimation from 3 orthogonal perspectives. Empirically, we benchmark 6 distributions and up to 24 variance reduction strategies for differentiable sorting and ranking, differentiable shortest-paths on graphs, differentiable rendering for pose estimation, as well as differentiable cryo-ET simulations."
                    },
                    {
                        "title": "TrAct: Making First-layer Pre-Activations Trainable",
                        "abstract": "We consider the training of the first layer of vision models and notice the clear relationship between pixel values and gradient update magnitudes: the gradients arriving at the weights of a first layer are by definition directly proportional to (normalized) input pixel values. Thus, an image with low contrast has a smaller impact on learning than an image with higher contrast, and a very bright or very dark image has a stronger impact on the weights than an image with moderate brightness. In this work, we propose performing gradient descent on the embeddings produced by the first layer of the model. However, switching to discrete inputs with an embedding layer is not a reasonable option for vision models. Thus, we propose the conceptual procedure of (i) a gradient descent step on first layer activations to construct an activation proposal, and (ii) finding the optimal weights of the first layer, i.e., those weights which minimize the squared distance to the activation proposal. We provide a closed form solution of the procedure and adjust it for robust stochastic training while computing everything efficiently. Empirically, we find that TrAct (Training Activations) speeds up training by factors between 1.25x and 4x while requiring only a small computational overhead. We demonstrate the utility of TrAct with different optimizers for a range of different vision models including convolutional and transformer architectures."
                    },
                    {
                        "title": "Improving naive Bayes classifiers using neuro-fuzzy learning",
                        "abstract": "Naive Bayes classifiers are a well-known and powerful type of classifier that can easily be induced from a dataset of sample cases. However, the strong conditional independence and distribution assumptions underlying them can sometimes lead to poor classification performance. Another prominent type of classifier are neuro-fuzzy classification systems which derive (fuzzy) classifiers from data using neural network inspired learning methods. Since there are certain structural similarities between a neuro-fuzzy classifier and a naive Bayes classifier, the idea suggests itself to mapping the latter to the former in order to improve its capabilities."
                    }
                ]
            },
            "6a004636-a4cf-4e60-ac00-9218a8eecedc": {
                "pk": "6a004636-a4cf-4e60-ac00-9218a8eecedc",
                "name": "Oliver Deussen",
                "collaborators": [
                    "Fan Tang",
                    "Yuxin Zhang",
                    "Weiming Dong",
                    "Tong-Yee Lee",
                    "Juan Cao",
                    "Thilo Spinner",
                    "Rebecca Kehlbeck",
                    "R. Sevastjanova",
                    "Tobias St\u00e4hle",
                    "Mennatallah El-Assady"
                ],
                "domain": [
                    "Generative Models",
                    "Text-to-Image",
                    "Text-to-Music",
                    "Interactive Visualization"
                ],
                "publications": [
                    {
                        "title": "Break-for-Make: Modular Low-Rank Adaptations for Composable Content-Style Customization",
                        "abstract": "Personalized generation paradigms empower designers to customize visual intellectual properties with the help of textual descriptions by tuning or adapting pre-trained text-to-image models on a few images. Recent works explore approaches for concurrently customizing both content and detailed visual style appearance. However, these existing approaches often generate images where the content and style are entangled. In this study, we reconsider the customization of content and style concepts from the perspective of parameter space construction. Unlike existing methods that utilize a shared parameter space for content and style, we propose a learning framework that separates the parameter space to facilitate individual learning of content and style, thereby enabling disentangled content and style. To achieve this goal, we introduce\"partly learnable projection\"(PLP) matrices to separate the original adapters into divided sub-parameter spaces. We propose\"break-for-make\"customization learning pipeline based on PLP, which is simple yet effective. We break the original adapters into\"up projection\"and\"down projection\", train content and style PLPs individually with the guidance of corresponding textual prompts in the separate adapters, and maintain generalization by employing a multi-correspondence projection learning strategy. Based on the adapters broken apart for separate training content and style, we then make the entity parameter space by reconstructing the content and style PLPs matrices, followed by fine-tuning the combined adapter to generate the target object with the desired appearance. Experiments on various styles, including textures, materials, and artistic style, show that our method outperforms state-of-the-art single/multiple concept learning pipelines in terms of content-style-prompt alignment."
                    },
                    {
                        "title": "A Comprehensive Evaluation of Arbitrary Image Style Transfer Methods.",
                        "abstract": "Despite the remarkable process in the field of arbitrary image style transfer (AST), inconsistent evaluation continues to plague style transfer research. Existing methods often suffer from limited objective evaluation and inconsistent subjective feedback, hindering reliable comparisons among AST variants. In this study, we propose a multi-granularity assessment system that combines standardized objective and subjective evaluations. We collect a fine-grained dataset considering a range of image contexts such as different scenes, object complexities, and rich parsing information from multiple sources. Objective and subjective studies are conducted using the collected dataset. Specifically, we innovate on traditional subjective studies by developing an online evaluation system utilizing a combination of point-wise, pair-wise, and group-wise questionnaires. Finally, we bridge the gap between objective and subjective evaluations by examining the consistency between the results from the two studies. We experimentally evaluate CNN-based, flow-based, transformer-based, and diffusion-based AST methods by the proposed multi-granularity assessment system, which lays the foundation for a reliable and robust evaluation. Providing standardized measures, objective data, and detailed subjective feedback empowers researchers to make informed comparisons and drive innovation in this rapidly evolving field. Finally, for the collected dataset and our online evaluation system, please see http://ivc.ia.ac.cn."
                    },
                    {
                        "title": "Dance-to-Music Generation with Encoder-based Textual Inversion of Diffusion Models",
                        "abstract": "The seamless integration of music with dance movements is essential for communicating the artistic intent of a dance piece. This alignment also significantly improves the immersive quality of gaming experiences and animation productions. Although there has been remarkable advancement in creating high-fidelity music from textual descriptions, current methodologies mainly focus on modulating overall characteristics such as genre and emotional tone. They often overlook the nuanced management of temporal rhythm, which is indispensable in crafting music for dance, since it intricately aligns the musical beats with the dancers' movements. Recognizing this gap, we propose an encoder-based textual inversion technique to augment text-to-music models with visual control, facilitating personalized music generation. Specifically, we develop dual-path rhythm-genre inversion to effectively integrate the rhythm and genre of a dance motion sequence into the textual space of a text-to-music model. Contrary to traditional textual inversion methods, which directly update text embeddings to reconstruct a single target object, our approach utilizes separate rhythm and genre encoders to obtain text embeddings for two pseudo-words, adapting to the varying rhythms and genres. We collect a new dataset called In-the-wild Dance Videos (InDV) and demonstrate that our approach outperforms state-of-the-art methods across multiple evaluation metrics. Furthermore, our method is able to adapt to changes in tempo and effectively integrates with the inherent text-guided generation capability of the pre-trained model. Our source code and demo videos are available at \\url{https://github.com/lsfhuihuiff/Dance-to-music_Siggraph_Asia_2024}"
                    },
                    {
                        "title": "-generAItor: Tree-in-the-loop Text Generation for Language Model Explainability and Adaptation",
                        "abstract": "Large language models (LLMs) are widely deployed in various downstream tasks, e.g., auto-completion, aided writing, or chat-based text generation. However, the considered output candidates of the underlying search algorithm are under-explored and under-explained. We tackle this shortcoming by proposing a tree-in-the-loop approach, where a visual representation of the beam search tree is the central component for analyzing, explaining, and adapting the generated outputs. To support these tasks, we present generAItor, a visual analytics technique, augmenting the central beam search tree with various task-specific widgets, providing targeted visualizations and interaction possibilities. Our approach allows interactions on multiple levels and offers an iterative pipeline that encompasses generating, exploring, and comparing output candidates, as well as fine-tuning the model based on adapted data. Our case study shows that our tool generates new insights in gender bias analysis beyond state-of-the-art template-based methods. Additionally, we demonstrate the applicability of our approach in a qualitative user study. Finally, we quantitatively evaluate the adaptability of the model to few samples, as occurring in text-generation use cases."
                    },
                    {
                        "title": "Newton Losses: Using Curvature Information for Learning with Differentiable Algorithms",
                        "abstract": "When training neural networks with custom objectives, such as ranking losses and shortest-path losses, a common problem is that they are, per se, non-differentiable. A popular approach is to continuously relax the objectives to provide gradients, enabling learning. However, such differentiable relaxations are often non-convex and can exhibit vanishing and exploding gradients, making them (already in isolation) hard to optimize. Here, the loss function poses the bottleneck when training a deep neural network. We present Newton Losses, a method for improving the performance of existing hard to optimize losses by exploiting their second-order information via their empirical Fisher and Hessian matrices. Instead of training the neural network with second-order techniques, we only utilize the loss function's second-order information to replace it by a Newton Loss, while training the network with gradient descent. This makes our method computationally efficient. We apply Newton Losses to eight differentiable algorithms for sorting and shortest-paths, achieving significant improvements for less-optimized differentiable algorithms, and consistent improvements, even for well-optimized differentiable algorithms."
                    },
                    {
                        "title": "Revealing the Unwritten: Visual Investigation of Beam Search Trees to Address Language Model Prompting Challenges",
                        "abstract": "The growing popularity of generative language models has amplified interest in interactive methods to guide model outputs. Prompt refinement is considered one of the most effective means to influence output among these methods. We identify several challenges associated with prompting large language models, categorized into data- and model-specific, linguistic, and socio-linguistic challenges. A comprehensive examination of model outputs, including runner-up candidates and their corresponding probabilities, is needed to address these issues. The beam search tree, the prevalent algorithm to sample model outputs, can inherently supply this information. Consequently, we introduce an interactive visual method for investigating the beam search tree, facilitating analysis of the decisions made by the model during generation. We quantitatively show the value of exposing the beam search tree and present five detailed analysis scenarios addressing the identified challenges. Our methodology validates existing results and offers additional insights."
                    }
                ]
            },
            "971f920a-0663-45c3-a1ae-42ad4b13b131": {
                "pk": "971f920a-0663-45c3-a1ae-42ad4b13b131",
                "name": "Mikhail Yurochkin",
                "collaborators": [
                    "Yuekai Sun",
                    "Felipe Maia Polo",
                    "K. Greenewald",
                    "Leshem Choshen",
                    "Justin Solomon",
                    "Subha Maity",
                    "Rickard Br\u00fcel Gabrielsson",
                    "Mayank Agarwal",
                    "Jiacheng Zhu",
                    "Leonid Karlinsky"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Model Alignment",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Asymmetry in Low-Rank Adapters of Foundation Models",
                        "abstract": "Parameter-efficient fine-tuning optimizes large, pre-trained foundation models by updating a subset of parameters; in this class, Low-Rank Adaptation (LoRA) is particularly effective. Inspired by an effort to investigate the different roles of LoRA matrices during fine-tuning, this paper characterizes and leverages unexpected asymmetry in the importance of low-rank adapter matrices. Specifically, when updating the parameter matrices of a neural network by adding a product $BA$, we observe that the $B$ and $A$ matrices have distinct functions: $A$ extracts features from the input, while $B$ uses these features to create the desired output. Based on this observation, we demonstrate that fine-tuning $B$ is inherently more effective than fine-tuning $A$, and that a random untrained $A$ should perform nearly as well as a fine-tuned one. Using an information-theoretic lens, we also bound the generalization of low-rank adapters, showing that the parameter savings of exclusively training $B$ improves the bound. We support our conclusions with experiments on RoBERTa, BART-Large, LLaMA-2, and ViTs."
                    },
                    {
                        "title": "LiveXiv -- A Multi-Modal Live Benchmark Based on Arxiv Papers Content",
                        "abstract": "The large-scale training of multi-modal models on data scraped from the web has shown outstanding utility in infusing these models with the required world knowledge to perform effectively on multiple downstream tasks. However, one downside of scraping data from the web can be the potential sacrifice of the benchmarks on which the abilities of these models are often evaluated. To safeguard against test data contamination and to truly test the abilities of these foundation models we propose LiveXiv: A scalable evolving live benchmark based on scientific ArXiv papers. LiveXiv accesses domain-specific manuscripts at any given timestamp and proposes to automatically generate visual question-answer pairs (VQA). This is done without any human-in-the-loop, using the multi-modal content in the manuscripts, like graphs, charts, and tables. Moreover, we introduce an efficient evaluation approach that estimates the performance of all models on the evolving benchmark using evaluations of only a subset of models. This significantly reduces the overall evaluation cost. We benchmark multiple open and proprietary Large Multi-modal Models (LMMs) on the first version of our benchmark, showing its challenging nature and exposing the models true abilities, avoiding contamination. Lastly, in our commitment to high quality, we have collected and evaluated a manually verified subset. By comparing its overall results to our automatic annotations, we have found that the performance variance is indeed minimal (<2.5%). Our dataset is available online on HuggingFace, and our code will be available here."
                    },
                    {
                        "title": "tinyBenchmarks: evaluating LLMs with fewer examples",
                        "abstract": "The versatility of large language models (LLMs) led to the creation of diverse benchmarks that thoroughly test a variety of language models' abilities. These benchmarks consist of tens of thousands of examples making evaluation of LLMs very expensive. In this paper, we investigate strategies to reduce the number of evaluations needed to assess the performance of an LLM on several key benchmarks. For example, we show that to accurately estimate the performance of an LLM on MMLU, a popular multiple-choice QA benchmark consisting of 14K examples, it is sufficient to evaluate this LLM on 100 curated examples. We release evaluation tools and tiny versions of popular benchmarks: Open LLM Leaderboard, MMLU, HELM, and AlpacaEval 2.0. Our empirical analysis demonstrates that these tools and tiny benchmarks are sufficient to reliably and efficiently reproduce the original evaluation results."
                    },
                    {
                        "title": "Aligners: Decoupling LLMs and Alignment",
                        "abstract": "Large Language Models (LLMs) need to be aligned with human expectations to ensure their safety and utility in most applications. Alignment is challenging, costly, and needs to be repeated for every LLM and alignment criterion. We propose to decouple LLMs and alignment by training aligner models that can be used to align any LLM for a given criteria on an as-needed basis, thus also reducing the potential negative impacts of alignment on performance. Our recipe for training the aligner models solely relies on synthetic data generated with a (prompted) LLM and can be easily adjusted for a variety of alignment criteria. We use the same synthetic data to train inspectors, binary miss-alignment classification models to guide a\"squad\"of multiple aligners. Our empirical results demonstrate consistent improvements when applying aligner squad to various LLMs, including chat-aligned models, across several instruction-following and red-teaming datasets."
                    },
                    {
                        "title": "A statistical framework for weak-to-strong generalization",
                        "abstract": "Modern large language model (LLM) alignment techniques rely on human feedback, but it is unclear whether the techniques fundamentally limit the capabilities of aligned LLMs. In particular, it is unclear whether it is possible to align (stronger) LLMs with superhuman capabilities with (weaker) human feedback without degrading their capabilities. This is an instance of the weak-to-strong generalization problem: using weaker (less capable) feedback to train a stronger (more capable) model. We prove that weak-to-strong generalization is possible by eliciting latent knowledge from pre-trained LLMs. In particular, we cast the weak-to-strong generalization problem as a transfer learning problem in which we wish to transfer a latent concept from a weak model to a strong pre-trained model. We prove that a naive fine-tuning approach suffers from fundamental limitations, but an alternative refinement-based approach suggested by the problem structure provably overcomes the limitations of fine-tuning. Finally, we demonstrate the practical applicability of the refinement approach with three LLM alignment tasks."
                    },
                    {
                        "title": "Efficient multi-prompt evaluation of LLMs",
                        "abstract": "Most popular benchmarks for comparing LLMs rely on a limited set of prompt templates, which may not fully capture the LLMs' abilities and can affect the reproducibility of results on leaderboards. Many recent works empirically verify prompt sensitivity and advocate for changes in LLM evaluation. In this paper, we consider the problem of estimating the performance distribution across many prompt variants instead of finding a single prompt to evaluate with. We introduce PromptEval, a method for estimating performance across a large set of prompts borrowing strength across prompts and examples to produce accurate estimates under practical evaluation budgets. The resulting distribution can be used to obtain performance quantiles to construct various robust performance metrics (e.g., top 95% quantile or median). We prove that PromptEval consistently estimates the performance distribution and demonstrate its efficacy empirically on three prominent LLM benchmarks: MMLU, BIG-bench Hard, and LMentry; for example, PromptEval can accurately estimate performance quantiles across 100 prompt templates on MMLU with a budget equivalent to two single-prompt evaluations. Moreover, we show how PromptEval can be useful in LLM-as-a-judge and best prompt identification applications."
                    },
                    {
                        "title": "Prompt Exploration with Prompt Regression",
                        "abstract": "In the advent of democratized usage of large language models (LLMs), there is a growing desire to systematize LLM prompt creation and selection processes beyond iterative trial-and-error. Prior works majorly focus on searching the space of prompts without accounting for relations between prompt variations. Here we propose a framework, Prompt Exploration with Prompt Regression (PEPR), to predict the effect of prompt combinations given results for individual prompt elements as well as a simple method to select an effective prompt for a given use-case. We evaluate our approach with open-source LLMs of different sizes on several different tasks."
                    },
                    {
                        "title": "Compress then Serve: Serving Thousands of LoRA Adapters with Little Overhead",
                        "abstract": "Fine-tuning large language models (LLMs) with low-rank adaptations (LoRAs) has become common practice, often yielding numerous copies of the same LLM differing only in their LoRA updates. This paradigm presents challenges for systems that serve real-time responses to queries that each involve a different LoRA. Prior works optimize the design of such systems but still require continuous loading and offloading of LoRAs, as it is infeasible to store thousands of LoRAs in GPU memory. To mitigate this issue, we investigate the efficacy of model compression when serving LoRAs. We propose a method for joint compression of LoRAs into a shared basis paired with LoRA-specific scaling matrices. We extend our algorithm to learn clusters of LoRAs that are more amenable to joint compression, allowing it to scale gracefully to large LoRA collections. Our experiments with up to 500 LoRAs demonstrate that compressed LoRAs preserve performance while offering major throughput gains in realistic serving scenarios with over a thousand LoRAs, maintaining 80% of the throughput of serving a single LoRA."
                    },
                    {
                        "title": "Distributional Preference Alignment of LLMs via Optimal Transport",
                        "abstract": "Current LLM alignment techniques use pairwise human preferences at a sample level, and as such, they do not imply an alignment on the distributional level. We propose in this paper Alignment via Optimal Transport (AOT), a novel method for distributional preference alignment of LLMs. AOT aligns LLMs on unpaired preference data by making the reward distribution of the positive samples stochastically dominant in the first order on the distribution of negative samples. We introduce a convex relaxation of this first-order stochastic dominance and cast it as an optimal transport problem with a smooth and convex cost. Thanks to the one-dimensional nature of the resulting optimal transport problem and the convexity of the cost, it has a closed-form solution via sorting on empirical measures. We fine-tune LLMs with this AOT objective, which enables alignment by penalizing the violation of the stochastic dominance of the reward distribution of the positive samples on the reward distribution of the negative samples. We analyze the sample complexity of AOT by considering the dual of the OT problem and show that it converges at the parametric rate. Empirically, we show on a diverse set of alignment datasets and LLMs that AOT leads to state-of-the-art models in the 7B family of models when evaluated with Open LLM Benchmarks and AlpacaEval."
                    },
                    {
                        "title": "Large Language Model Routing with Benchmark Datasets",
                        "abstract": "There is a rapidly growing number of open-source Large Language Models (LLMs) and benchmark datasets to compare them. While some models dominate these benchmarks, no single model typically achieves the best accuracy in all tasks and use cases. In this work, we address the challenge of selecting the best LLM out of a collection of models for new tasks. We propose a new formulation for the problem, in which benchmark datasets are repurposed to learn a\"router\"model for this LLM selection, and we show that this problem can be reduced to a collection of binary classification tasks. We demonstrate the utility and limitations of learning model routers from various benchmark datasets, where we consistently improve performance upon using any single model for all tasks."
                    },
                    {
                        "title": "Calibrated Data-Dependent Constraints with Exact Satisfaction Guarantees",
                        "abstract": "We consider the task of training machine learning models with data-dependent constraints. Such constraints often arise as empirical versions of expected value constraints that enforce fairness or stability goals. We reformulate data-dependent constraints so that they are calibrated: enforcing the reformulated constraints guarantees that their expected value counterparts are satisfied with a user-prescribed probability. The resulting optimization problem is amendable to standard stochastic optimization algorithms, and we demonstrate the efficacy of our method on a fairness-sensitive classification task where we wish to guarantee the classifier's fairness (at test time)."
                    },
                    {
                        "title": "GeRA: Label-Efficient Geometrically Regularized Alignment",
                        "abstract": "Pretrained unimodal encoders incorporate rich semantic information into embedding space structures. To be similarly informative, multi-modal encoders typically require massive amounts of paired data for alignment and training. We introduce a semi-supervised Geometrically Regularized Alignment (GeRA) method to align the embedding spaces of pretrained unimodal encoders in a label-efficient way. Our method leverages the manifold geometry of unpaired (unlabeled) data to improve alignment performance. To prevent distortions to local geometry during the alignment process, potentially disrupting semantic neighborhood structures and causing misalignment of unobserved pairs, we introduce a geometric loss term. This term is built upon a diffusion operator that captures the local manifold geometry of the unimodal pretrained encoders. GeRA is modality-agnostic and thus can be used to align pretrained encoders from any data modalities. We provide empirical evidence to the effectiveness of our method in the domains of speech-text and image-text alignment. Our experiments demonstrate significant improvement in alignment quality compared to a variaty of leading baselines, especially with a small amount of paired data, using our proposed geometric regularization."
                    },
                    {
                        "title": "Weak Supervision Performance Evaluation via Partial Identification",
                        "abstract": "Programmatic Weak Supervision (PWS) enables supervised model training without direct access to ground truth labels, utilizing weak labels from heuristics, crowdsourcing, or pre-trained models. However, the absence of ground truth complicates model evaluation, as traditional metrics such as accuracy, precision, and recall cannot be directly calculated. In this work, we present a novel method to address this challenge by framing model evaluation as a partial identification problem and estimating performance bounds using Fr\\'echet bounds. Our approach derives reliable bounds on key metrics without requiring labeled data, overcoming core limitations in current weak supervision evaluation techniques. Through scalable convex optimization, we obtain accurate and computationally efficient bounds for metrics including accuracy, precision, recall, and F1-score, even in high-dimensional settings. This framework offers a robust approach to assessing model quality without ground truth labels, enhancing the practicality of weakly supervised learning for real-world applications."
                    },
                    {
                        "title": "Fusing Models with Complementary Expertise",
                        "abstract": "Training AI models that generalize across tasks and domains has long been among the open problems driving AI research. The emergence of Foundation Models made it easier to obtain expert models for a given task, but the heterogeneity of data that may be encountered at test time often means that any single expert is insufficient. We consider the Fusion of Experts (FoE) problem of fusing outputs of expert models with complementary knowledge of the data distribution and formulate it as an instance of supervised learning. Our method is applicable to both discriminative and generative tasks and leads to significant performance improvements in image and text classification, text summarization, multiple-choice QA, and automatic evaluation of generated text. We also extend our method to the\"frugal\"setting where it is desired to reduce the number of expert model evaluations at test time. Our implementation is publicly available at https://github.com/hwang595/FoE-ICLR2024."
                    },
                    {
                        "title": "An Investigation of Representation and Allocation Harms in Contrastive Learning",
                        "abstract": "The effect of underrepresentation on the performance of minority groups is known to be a serious problem in supervised learning settings; however, it has been underexplored so far in the context of self-supervised learning (SSL). In this paper, we demonstrate that contrastive learning (CL), a popular variant of SSL, tends to collapse representations of minority groups with certain majority groups. We refer to this phenomenon as representation harm and demonstrate it on image and text datasets using the corresponding popular CL methods. Furthermore, our causal mediation analysis of allocation harm on a downstream classification task reveals that representation harm is partly responsible for it, thus emphasizing the importance of studying and mitigating representation harm. Finally, we provide a theoretical explanation for representation harm using a stochastic block model that leads to a representational neural collapse in a contrastive learning setting."
                    },
                    {
                        "title": "Simple Disentanglement of Style and Content in Visual Representations",
                        "abstract": "Learning visual representations with interpretable features, i.e., disentangled representations, remains a challenging problem. Existing methods demonstrate some success but are hard to apply to large-scale vision datasets like ImageNet. In this work, we propose a simple post-processing framework to disentangle content and style in learned representations from pre-trained vision models. We model the pre-trained features probabilistically as linearly entangled combinations of the latent content and style factors and develop a simple disentanglement algorithm based on the probabilistic model. We show that the method provably disentangles content and style features and verify its efficacy empirically. Our post-processed features yield significant domain generalization performance improvements when the distribution shift occurs due to style changes or style-related spurious correlations."
                    },
                    {
                        "title": "Risk Aware Benchmarking of Large Language Models",
                        "abstract": "We propose a distributional framework for benchmarking socio-technical risks of foundation models with quantified statistical significance. Our approach hinges on a new statistical relative testing based on first and second order stochastic dominance of real random variables. We show that the second order statistics in this test are linked to mean-risk models commonly used in econometrics and mathematical finance to balance risk and utility when choosing between alternatives. Using this framework, we formally develop a risk-aware approach for foundation model selection given guardrails quantified by specified metrics. Inspired by portfolio optimization and selection theory in mathematical finance, we define a metrics portfolio for each model as a means to aggregate a collection of metrics, and perform model selection based on the stochastic dominance of these portfolios. The statistical significance of our tests is backed theoretically by an asymptotic analysis via central limit theorems instantiated in practice via a bootstrap variance estimate. We use our framework to compare various large language models regarding risks related to drifting from instructions and outputting toxic content."
                    },
                    {
                        "title": "Fairness Evaluation in Text Classification: Machine Learning Practitioner Perspectives of Individual and Group Fairness",
                        "abstract": "Mitigating algorithmic bias is a critical task in the development and deployment of machine learning models. While several toolkits exist to aid machine learning practitioners in addressing fairness issues, little is known about the strategies practitioners employ to evaluate model fairness and what factors influence their assessment, particularly in the context of text classification. Two common approaches of evaluating the fairness of a model are group fairness and individual fairness. We run a study with Machine Learning practitioners (n=24) to understand the strategies used to evaluate models. Metrics presented to practitioners (group vs. individual fairness) impact which models they consider fair. Participants focused on risks associated with underpredicting / overpredicting and model sensitivity relative to identity token manipulations. We discover fairness assessment strategies involving personal experiences or how users form groups of identity tokens to test model fairness. We provide recommendations for interactive tools for evaluating fairness in text classification."
                    },
                    {
                        "title": "Rewiring with Positional Encodings for Graph Neural Networks",
                        "abstract": "Several recent works use positional encodings to extend the receptive fields of graph neural network (GNN) layers equipped with attention mechanisms. These techniques, however, extend receptive fields to the complete graph, at substantial computational cost and risking a change in the inductive biases of conventional GNNs, or require complex architecture adjustments. As a conservative alternative, we use positional encodings to expand receptive fields to $r$-hop neighborhoods. More specifically, our method augments the input graph with additional nodes/edges and uses positional encodings as node and/or edge features. We thus modify graphs before inputting them to a downstream GNN model, instead of modifying the model itself. This makes our method model-agnostic, i.e., compatible with any of the existing GNN architectures. We also provide examples of positional encodings that are lossless with a one-to-one map between the original and the modified graphs. We demonstrate that extending receptive fields via positional encodings and a virtual fully-connected node significantly improves GNN performance and alleviates over-squashing using small $r$. We obtain improvements on a variety of models and datasets and reach competitive performance using traditional GNNs or graph Transformers."
                    }
                ]
            }
        }
    },
    "2406.13770": {
        "paper_data": {
            "title": "Elliptical Attention",
            "url": "http://arxiv.org/abs/2406.13770v1",
            "arxiv_id": "2406.13770",
            "authors": [
                "Stefan K. Nielsen",
                "Laziz U. Abdullaev",
                "Rachel Teo",
                "Tan M. Nguyen"
            ],
            "abstract": "Pairwise dot-product self-attention is key to the success of transformers that achieve state-of-the-art performance across a variety of applications in language and vision. This dot-product self-attention computes attention weights among the input tokens using Euclidean distance, which makes the model prone to representation collapse and vulnerable to contaminated samples. In this paper, we propose using a Mahalanobis distance metric for computing the attention weights to stretch the underlying feature space in directions of high contextual relevance. In particular, we define a hyper-ellipsoidal neighborhood around each query to increase the attention weights of the tokens lying in the contextually important directions. We term this novel class of attention Elliptical Attention. Our Elliptical Attention provides two benefits: 1) reducing representation collapse and 2) enhancing the model's robustness as the Elliptical Attention pays more attention to contextually relevant information rather than focusing on some small subset of informative features. We empirically demonstrate the advantages of Elliptical Attention over the baseline dot-product attention and state-of-the-art attention methods on various practical tasks, including object classification, image segmentation, and language modeling across different data modalities.",
            "introduction": "   1 Introduction  Attention mechanisms and transformers [71] have achieved state of the art performance across a wide variety of tasks in machine learning [22, 30, 66] and, in particular, within natural language processing [1, 2, 10, 56, 9], computer vision [13, 34, 68, 57, 53], and reinforcement learning [20, 4]. They have also demonstrated strong performance in knowledge transfer from pretraining tasks to various downstream tasks with weak or no supervision [54, 55, 12]. At the core of these models is the dot-product self-attention mechanism, which learns self-alignment between tokens in an input sequence by estimating the relative importance of each token with respect to all others. The mechanism then transforms each token into a weighted average of the feature representations of the other tokens with weights proportional to the learned importance scores. The relative importance scores capture contextual information among tokens and are key to the success of the transformer architecture [72, 67, 5, 51, 31].   Figure 1: Comparison of Attention Heatmaps. Elliptical pays attention to more relevant information. DeiT focuses on just a subset of informative features while Elliptical considers a wider set of contextually relevant information, helping to produce more accurate and robust predictions. Attention scores are min-max scaled for visualization purposes.   Recent work has begun exploring the connections between self-attention and non-parametric kernel regression [46, 19]. Under this interpretation, there is an unknown, underlying function f\ud835\udc53fitalic_f mapping the tokens in the input sequence to the output sequence. The self-attention mechanism estimates f\ud835\udc53fitalic_f by performing Nadaraya-Watson (NW) regression with isotropic Gaussian kernels. Our work leverages this perspective on self-attention, where we notice that Gaussian isotropic kernels are spherically invariant. This has the drawback of assuming all dimensions of the feature space are equal in terms of importance, meaning nearby tokens are assigned contextual relevance weights dependant only on their Euclidean distance from a query, regardless of direction. From the non-parametric regression perspective, we show that spherical invariance in the kernel causes the estimator to suffer provably higher variance. This causes two connected disadvantages in the self-attention setting. First, high variance in the estimator impairs robustness as small contaminations in the input cause large, erroneous changes in the self-attention output. Second, the high variance of the estimator reduces the capacity of the self-attention mechanism as hidden representations passing through the model are increasingly composed of uninformative noise.   Contribution. In this work, we propose Elliptical Attention, a new class of self-attention that constructs hyper-ellipsoidal, rather than hyper-spherical, neighborhoods around the attention queries. The key idea is to stretch the neighborhoods around the queries to upweight keys in directions of high importance. We achieve this by computing a Mahalanobis transformation that stretches the axes of the underlying feature space according to a learned measure of coordinate-wise relevance. Constructing hyper-ellipsoidal neighborhoods following this scheme allows the self-attention mechanism to learn higher-quality contextual representations that prevent representation collapse while simultaneously exhibiting stronger robustness. We additionally propose an estimator of coordinate-wise relevance in the self-attention mechanism that can be computed highly efficiently and with no learnable parameters. We theoretically prove that our estimator accurately estimates the relative coordinate-wise relevance",
            "references": [
                {
                    "title": "PIDformer: Transformer Meets Control Theory",
                    "abstract": "In this work, we address two main shortcomings of transformer architectures: input corruption and rank collapse in their output representation. We unveil self-attention as an autonomous state-space model that inherently promotes smoothness in its solutions, leading to lower-rank outputs and diminished representation capacity. Moreover, the steady-state solution of the model is sensitive to input perturbations. We incorporate a Proportional-Integral-Derivative (PID) closed-loop feedback control system with a reference point into the model to improve robustness and representation capacity. This integration aims to preserve high-frequency details while bolstering model stability, rendering it more noise-resilient. The resulting controlled state-space model is theoretically proven robust and adept at addressing the rank collapse. Motivated by this control framework, we derive a novel class of transformers, PID-controlled Transformer (PIDformer), aimed at improving robustness and mitigating the rank-collapse issue inherent in softmax transformers. We empirically evaluate the model for advantages and robustness against baseline transformers across various practical tasks, including object classification, image segmentation, and language modeling."
                },
                {
                    "title": "Nonlinear SVD with Asymmetric Kernels: feature learning and asymmetric Nystr\u00f6m method",
                    "abstract": "Asymmetric data naturally exist in real life, such as directed graphs. Different from the common kernel methods requiring Mercer kernels, this paper tackles the asymmetric kernel-based learning problem. We describe a nonlinear extension of the matrix Singular Value Decomposition through asymmetric kernels, namely KSVD. First, we construct two nonlinear feature mappings w.r.t. rows and columns of the given data matrix. The proposed optimization problem maximizes the variance of each mapping projected onto the subspace spanned by the other, subject to a mutual orthogonality constraint. Through Lagrangian duality, we show that it can be solved by the left and right singular vectors in the feature space induced by the asymmetric kernel. Moreover, we start from the integral equations with a pair of adjoint eigenfunctions corresponding to the singular vectors on an asymmetrical kernel, and extend the Nystr\\\"om method to asymmetric cases through the finite sample approximation, which can be applied to speedup the training in KSVD. Experiments show that asymmetric KSVD learns features outperforming Mercer-kernel based methods that resort to symmetrization, and also verify the effectiveness of the asymmetric Nystr\\\"om method."
                },
                {
                    "title": "The emergence of clusters in self-attention dynamics",
                    "abstract": "Viewing Transformers as interacting particle systems, we describe the geometry of learned representations when the weights are not time dependent. We show that particles, representing tokens, tend to cluster toward particular limiting objects as time tends to infinity. Cluster locations are determined by the initial tokens, confirming context-awareness of representations learned by Transformers. Using techniques from dynamical systems and partial differential equations, we show that the type of limiting object that emerges depends on the spectrum of the value matrix. Additionally, in the one-dimensional case we prove that the self-attention matrix converges to a low-rank Boolean matrix. The combination of these results mathematically confirms the empirical observation made by Vaswani et al. [VSP'17] that leaders appear in a sequence of tokens when processed by Transformers."
                },
                {
                    "title": "Designing Robust Transformers using Robust Kernel Density Estimation",
                    "abstract": "Recent advances in Transformer architectures have empowered their empirical success in a variety of tasks across different domains. However, existing works mainly focus on predictive accuracy and computational cost, without considering other practical issues, such as robustness to contaminated samples. Recent work by Nguyen et al., (2022) has shown that the self-attention mechanism, which is the center of the Transformer architecture, can be viewed as a non-parametric estimator based on kernel density estimation (KDE). This motivates us to leverage a set of robust kernel density estimation methods for alleviating the issue of data contamination. Specifically, we introduce a series of self-attention mechanisms that can be incorporated into different Transformer architectures and discuss the special properties of each method. We then perform extensive empirical studies on language modeling and image classification tasks. Our methods demonstrate robust performance in multiple scenarios while maintaining competitive results on clean datasets."
                },
                {
                    "title": "DNNR: Differential Nearest Neighbors Regression",
                    "abstract": "K-nearest neighbors (KNN) is one of the earli-est and most established algorithms in machine learning. For regression tasks, KNN averages the targets within a neighborhood which poses a number of challenges: the neighborhood de\ufb01nition is crucial for the predictive performance as neighbors might be selected based on uninformative features, and averaging does not account for how the function changes locally. We propose a novel method called Differential Nearest Neighbors Regression (DNNR) that addresses both issues simul-taneously: during training, DNNR estimates local gradients to scale the features; during inference, it performs an n-th order Taylor approximation using estimated gradients. In a large-scale evaluation on over 250 datasets, we \ufb01nd that DNNR performs comparably to state-of-the-art gradient boosting methods and MLPs while maintaining the simplicity and transparency of KNN. This allows us to derive theoretical error bounds and inspect failures. In times that call for transparency of ML models, DNNR provides a good balance between performance and interpretability. 2"
                },
                {
                    "title": "Understanding The Robustness in Vision Transformers",
                    "abstract": "Recent studies show that Vision Transformers(ViTs) exhibit strong robustness against various corruptions. Although this property is partly attributed to the self-attention mechanism, there is still a lack of systematic understanding. In this paper, we examine the role of self-attention in learning robust representations. Our study is motivated by the intriguing properties of the emerging visual grouping in Vision Transformers, which indicates that self-attention may promote robustness through improved mid-level representations. We further propose a family of fully attentional networks (FANs) that strengthen this capability by incorporating an attentional channel processing design. We validate the design comprehensively on various hierarchical backbones. Our model achieves a state-of-the-art 87.1% accuracy and 35.8% mCE on ImageNet-1k and ImageNet-C with 76.8M parameters. We also demonstrate state-of-the-art accuracy and robustness in two downstream tasks: semantic segmentation and object detection. Code is available at: https://github.com/NVlabs/FAN."
                },
                {
                    "title": "TableFormer: Robust Transformer Modeling for Table-Text Encoding",
                    "abstract": "Understanding tables is an important aspect of natural language understanding. Existing models for table understanding require linearization of the table structure, where row or column order is encoded as an unwanted bias. Such spurious biases make the model vulnerable to row and column order perturbations. Additionally, prior work has not thoroughly modeled the table structures or table-text alignments, hindering the table-text understanding ability. In this work, we propose a robust and structurally aware table-text encoding architecture TableFormer, where tabular structural biases are incorporated completely through learnable attention biases. TableFormer is (1) strictly invariant to row and column orders, and, (2) could understand tables better due to its tabular inductive biases. Our evaluations showed that TableFormer outperforms strong baselines in all settings on SQA, WTQ and TabFact table reasoning datasets, and achieves state-of-the-art performance on SQA, especially when facing answer-invariant row and column order perturbations (6% improvement over the best baseline), because previous SOTA models\u2019 performance drops by 4% - 6% when facing such perturbations while TableFormer is not affected."
                },
                {
                    "title": "Revisiting Over-smoothing in BERT from the Perspective of Graph",
                    "abstract": "Recently over-smoothing phenomenon of Transformer-based models is observed in both vision and language fields. However, no existing work has delved deeper to further investigate the main cause of this phenomenon. In this work, we make the attempt to analyze the over-smoothing problem from the perspective of graph, where such problem was first discovered and explored. Intuitively, the self-attention matrix can be seen as a normalized adjacent matrix of a corresponding graph. Based on the above connection, we provide some theoretical analysis and find that layer normalization plays a key role in the over-smoothing issue of Transformer-based models. Specifically, if the standard deviation of layer normalization is sufficiently large, the output of Transformer stacks will converge to a specific low-rank subspace and result in over-smoothing. To alleviate the over-smoothing problem, we consider hierarchical fusion strategies, which combine the representations from different layers adaptively to make the output more diverse. Extensive experiment results on various data sets illustrate the effect of our fusion method."
                },
                {
                    "title": "Sinkformers: Transformers with Doubly Stochastic Attention",
                    "abstract": "Attention based models such as Transformers involve pairwise interactions between data points, modeled with a learnable attention matrix. Importantly, this attention matrix is normalized with the SoftMax operator, which makes it row-wise stochastic. In this paper, we propose instead to use Sinkhorn's algorithm to make attention matrices doubly stochastic. We call the resulting model a Sinkformer. We show that the row-wise stochastic attention matrices in classical Transformers get close to doubly stochastic matrices as the number of epochs increases, justifying the use of Sinkhorn normalization as an informative prior. On the theoretical side, we show that, unlike the SoftMax operation, this normalization makes it possible to understand the iterations of self-attention modules as a discretized gradient-flow for the Wasserstein metric. We also show in the infinite number of samples limit that, when rescaling both attention matrices and depth, Sinkformers operate a heat diffusion. On the experimental side, we show that Sinkformers enhance model accuracy in vision and natural language processing tasks. In particular, on 3D shapes classification, Sinkformers lead to a significant improvement."
                },
                {
                    "title": "Improving Transformers with Probabilistic Attention Keys",
                    "abstract": "Multi-head attention is a driving force behind state-of-the-art transformers, which achieve remarkable performance across a variety of natural language processing (NLP) and computer vision tasks. It has been observed that for many applications, those attention heads learn redundant embedding, and most of them can be removed without degrading the performance of the model. Inspired by this observation, we propose Transformer with a Mixture of Gaussian Keys (Transformer-MGK), a novel transformer architecture that replaces redundant heads in transformers with a mixture of keys at each head. These mixtures of keys follow a Gaussian mixture model and allow each attention head to focus on different parts of the input sequence efficiently. Compared to its conventional transformer counterpart, Transformer-MGK accelerates training and inference, has fewer parameters, and requires fewer FLOPs to compute while achieving comparable or better accuracy across tasks. Transformer-MGK can also be easily extended to use with linear attention. We empirically demonstrate the advantage of Transformer-MGK in a range of practical applications, including language modeling and tasks that involve very long sequences. On the Wikitext-103 and Long Range Arena benchmark, Transformer-MGKs with 4 heads attain comparable or better performance to the baseline transformers with 8 heads."
                },
                {
                    "title": "Modeling Concentrated Cross-Attention for Neural Machine Translation with Gaussian Mixture Model",
                    "abstract": "Cross-attention is an important component of neural machine translation (NMT), which is always realized by dot-product attention in previous methods. However, dot-product attention only considers the pair-wise correlation between words, resulting in dispersion when dealing with long sentences and neglect of source neighboring relationships. Inspired by linguistics, the above issues are caused by ignoring a type of cross-attention, called concentrated attention, which focuses on several central words and then spreads around them. In this work, we apply Gaussian Mixture Model (GMM) to model the concentrated attention in cross-attention. Experiments and analyses we conducted on three datasets show that the proposed method outperforms the baseline and has significant improvement on alignment quality, N-gram accuracy, and long sentence translation."
                },
                {
                    "title": "Probabilistic Attention for Interactive Segmentation",
                    "abstract": "We provide a probabilistic interpretation of attention and show that the standard dot-product attention in transformers is a special case of Maximum A Posteriori (MAP) inference. The proposed approach suggests the use of Expectation Maximization algorithms for online adaptation of key and value model parameters. This approach is useful for cases in which external agents, e.g., annotators, provide inference-time information about the correct values of some tokens, e.g, the semantic category of some pixels, and we need for this new information to propagate to other tokens in a principled manner. We illustrate the approach on an interactive semantic segmentation task in which annotators and models collaborate online to improve annotation efficiency. Using standard benchmarks, we observe that key adaptation boosts model performance ($\\sim10\\%$ mIoU) in the low feedback regime and value propagation improves model responsiveness in the high feedback regime. A PyTorch layer implementation of our probabilistic attention model will be made publicly available here: https://github.com/apple/ml-probabilistic-attention."
                },
                {
                    "title": "Rethinking Graph Transformers with Spectral Attention",
                    "abstract": "In recent years, the Transformer architecture has proven to be very successful in sequence processing, but its application to other data structures, such as graphs, has remained limited due to the difficulty of properly defining positions. Here, we present the $\\textit{Spectral Attention Network}$ (SAN), which uses a learned positional encoding (LPE) that can take advantage of the full Laplacian spectrum to learn the position of each node in a given graph. This LPE is then added to the node features of the graph and passed to a fully-connected Transformer. By leveraging the full spectrum of the Laplacian, our model is theoretically powerful in distinguishing graphs, and can better detect similar sub-structures from their resonance. Further, by fully connecting the graph, the Transformer does not suffer from over-squashing, an information bottleneck of most GNNs, and enables better modeling of physical phenomenons such as heat transfer and electric interaction. When tested empirically on a set of 4 standard datasets, our model performs on par or better than state-of-the-art GNNs, and outperforms any attention-based model by a wide margin, becoming the first fully-connected architecture to perform well on graph benchmarks."
                },
                {
                    "title": "Offline Reinforcement Learning as One Big Sequence Modeling Problem",
                    "abstract": "Reinforcement learning (RL) is typically concerned with estimating stationary policies or single-step models, leveraging the Markov property to factorize problems in time. However, we can also view RL as a generic sequence modeling problem, with the goal being to produce a sequence of actions that leads to a sequence of high rewards. Viewed in this way, it is tempting to consider whether high-capacity sequence prediction models that work well in other domains, such as natural-language processing, can also provide effective solutions to the RL problem. To this end, we explore how RL can be tackled with the tools of sequence modeling, using a Transformer architecture to model distributions over trajectories and repurposing beam search as a planning algorithm. Framing RL as sequence modeling problem simplifies a range of design decisions, allowing us to dispense with many of the components common in offline RL algorithms. We demonstrate the flexibility of this approach across long-horizon dynamics prediction, imitation learning, goal-conditioned RL, and offline RL. Further, we show that this approach can be combined with existing model-free algorithms to yield a state-of-the-art planner in sparse-reward, long-horizon tasks."
                },
                {
                    "title": "Decision Transformer: Reinforcement Learning via Sequence Modeling",
                    "abstract": "We introduce a framework that abstracts Reinforcement Learning (RL) as a sequence modeling problem. This allows us to draw upon the simplicity and scalability of the Transformer architecture, and associated advances in language modeling such as GPT-x and BERT. In particular, we present Decision Transformer, an architecture that casts the problem of RL as conditional sequence modeling. Unlike prior approaches to RL that fit value functions or compute policy gradients, Decision Transformer simply outputs the optimal actions by leveraging a causally masked Transformer. By conditioning an autoregressive model on the desired return (reward), past states, and actions, our Decision Transformer model can generate future actions that achieve the desired return. Despite its simplicity, Decision Transformer matches or exceeds the performance of state-of-the-art model-free offline RL baselines on Atari, OpenAI Gym, and Key-to-Door tasks."
                },
                {
                    "title": "Towards Robust Vision Transformer",
                    "abstract": "Recent advances on Vision Transformer (ViT) and its improved variants have shown that self-attention-based networks surpass traditional Convolutional Neural Networks (CNNs) in most vision tasks. However, existing ViTs focus on the standard accuracy and computation cost, lacking the investigation of the intrinsic influence on model robustness and generalization. In this work, we conduct systematic evaluation on components of ViTs in terms of their impact on robustness to adversarial examples, common corruptions and distribution shifts. We find some components can be harmful to robustness. By leveraging robust components as building blocks of ViTs, we propose Robust Vision Transformer (RVT), which is a new vision transformer and has superior performance with strong robustness. Inspired by the findings during the evaluation, we further propose two new plug-and-play techniques called position-aware attention scaling and patch-wise augmentation to augment our RVT, which we abbreviate as RVT*. The experimental results of RVT on ImageNet and six robustness benchmarks demonstrate its advanced robustness and generalization ability compared with previous ViTs and state-of-the-art CNNs. Furthermore, RVT-S* achieves Top-1 rank on multiple robustness leaderboards including ImageNet-C, ImageNet-Sketch and ImageNet-R."
                },
                {
                    "title": "On the Robustness of Vision Transformers to Adversarial Examples",
                    "abstract": "Recent advances in attention-based networks have shown that Vision Transformers can achieve state-of-the-art or near state-of-the-art results on many image classification tasks. This puts transformers in the unique position of being a promising alternative to traditional convolutional neural networks (CNNs). While CNNs have been carefully studied with respect to adversarial attacks, the same cannot be said of Vision Transformers. In this paper, we study the robustness of Vision Transformers to adversarial examples. Our analyses of transformer security is divided into three parts. First, we test the transformer under standard white-box and black-box attacks. Second, we study the transfer-ability of adversarial examples between CNNs and trans-formers. We show that adversarial examples do not readily transfer between CNNs and transformers. Based on this finding, we analyze the security of a simple ensemble defense of CNNs and transformers. By creating a new attack, the self-attention blended gradient attack, we show that such an ensemble is not secure under a white-box adversary. However, under a black-box adversary, we show that an ensemble can achieve unprecedented robustness without sacrificing clean accuracy. Our analysis for this work is done using six types of white-box attacks and two types of black-box attacks. Our study encompasses multiple Vision Transformers, Big Transfer Models and CNN architectures trained on CIFAR-10, CIFAR-100 and ImageNet."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Zero-Shot Text-to-Image Generation",
                    "abstract": "Text-to-image generation has traditionally focused on finding better modeling assumptions for training on a fixed dataset. These assumptions might involve complex architectures, auxiliary losses, or side information such as object part labels or segmentation masks supplied during training. We describe a simple approach for this task based on a transformer that autoregressively models the text and image tokens as a single stream of data. With sufficient data and scale, our approach is competitive with previous domain-specific models when evaluated in a zero-shot fashion."
                },
                {
                    "title": "Transformers in Vision: A Survey",
                    "abstract": "Astounding results from Transformer models on natural language tasks have intrigued the vision community to study their application to computer vision problems. Among their salient benefits, Transformers enable modeling long dependencies between input sequence elements and support parallel processing of sequence as compared to recurrent networks, e.g., Long short-term memory. Different from convolutional networks, Transformers require minimal inductive biases for their design and are naturally suited as set-functions. Furthermore, the straightforward design of Transformers allows processing multiple modalities (e.g., images, videos, text, and speech) using similar processing blocks and demonstrates excellent scalability to very large capacity networks and huge datasets. These strengths have led to exciting progress on a number of vision tasks using Transformer networks. This survey aims to provide a comprehensive overview of the Transformer models in the computer vision discipline. We start with an introduction to fundamental concepts behind the success of Transformers, i.e., self-attention, large-scale pre-training, and bidirectional feature encoding. We then cover extensive applications of transformers in vision including popular recognition tasks (e.g., image classification, object detection, action recognition, and segmentation), generative modeling, multi-modal tasks (e.g., visual-question answering, visual reasoning, and visual grounding), video processing (e.g., activity recognition, video forecasting), low-level vision (e.g., image super-resolution, image enhancement, and colorization), and three-dimensional analysis (e.g., point cloud classification and segmentation). We compare the respective advantages and limitations of popular techniques both in terms of architectural design and their experimental value. Finally, we provide an analysis on open research directions and possible future works. We hope this effort will ignite further interest in the community to solve current challenges toward the application of transformer models in computer vision."
                },
                {
                    "title": "Training data-efficient image transformers & distillation through attention",
                    "abstract": "Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. However, these visual transformers are pre-trained with hundreds of millions of images using an expensive infrastructure, thereby limiting their adoption. In this work, we produce a competitive convolution-free transformer by training on Imagenet only. We train them on a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop evaluation) on ImageNet with no external data. More importantly, we introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention. We show the interest of this token-based distillation, especially when using a convnet as a teacher. This leads us to report results competitive with convnets for both Imagenet (where we obtain up to 85.2% accuracy) and when transferring to other tasks. We share our code and models."
                },
                {
                    "title": "Long Range Arena: A Benchmark for Efficient Transformers",
                    "abstract": "Transformers do not scale very well to long sequence lengths largely because of quadratic self-attention complexity. In the recent months, a wide spectrum of efficient, fast Transformers have been proposed to tackle this problem, more often than not claiming superior or comparable model quality to vanilla Transformer models. To this date, there is no well-established consensus on how to evaluate this class of models. Moreover, inconsistent benchmarking on a wide spectrum of tasks and datasets makes it difficult to assess relative model quality amongst many models. This paper proposes a systematic and unified benchmark, LRA, specifically focused on evaluating model quality under long-context scenarios. Our benchmark is a suite of tasks consisting of sequences ranging from $1K$ to $16K$ tokens, encompassing a wide range of data types and modalities such as text, natural, synthetic images, and mathematical expressions requiring similarity, structural, and visual-spatial reasoning. We systematically evaluate ten well-established long-range Transformer models (Reformers, Linformers, Linear Transformers, Sinkhorn Transformers, Performers, Synthesizers, Sparse Transformers, and Longformers) on our newly proposed benchmark suite. LRA paves the way towards better understanding this class of efficient Transformer models, facilitates more research in this direction, and presents new challenging tasks to tackle. Our benchmark code will be released at this https URL."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Higher-Order Certification for Randomized Smoothing",
                    "abstract": "Randomized smoothing is a recently proposed defense against adversarial attacks that has achieved SOTA provable robustness against $\\ell_2$ perturbations. A number of publications have extended the guarantees to other metrics, such as $\\ell_1$ or $\\ell_\\infty$, by using different smoothing measures. Although the current framework has been shown to yield near-optimal $\\ell_p$ radii, the total safety region certified by the current framework can be arbitrarily small compared to the optimal. In this work, we propose a framework to improve the certified safety region for these smoothed classifiers without changing the underlying smoothing scheme. The theoretical contributions are as follows: 1) We generalize the certification for randomized smoothing by reformulating certified radius calculation as a nested optimization problem over a class of functions. 2) We provide a method to calculate the certified safety region using $0^{th}$-order and $1^{st}$-order information for Gaussian-smoothed classifiers. We also provide a framework that generalizes the calculation for certification using higher-order information. 3) We design efficient, high-confidence estimators for the relevant statistics of the first-order information. Combining the theoretical contribution 2) and 3) allows us to certify safety region that are significantly larger than the ones provided by the current methods. On CIFAR10 and Imagenet datasets, the new regions certified by our approach achieve significant improvements on general $\\ell_1$ certified radii and on the $\\ell_2$ certified radii for color-space attacks ($\\ell_2$ restricted to 1 channel) while also achieving smaller improvements on the general $\\ell_2$ certified radii. Our framework can also provide a way to circumvent the current impossibility results on achieving higher magnitude of certified radii without requiring the use of data-dependent smoothing techniques."
                },
                {
                    "title": "Rethinking Attention with Performers",
                    "abstract": "We introduce Performers, Transformer architectures which can estimate regular (softmax) full-rank-attention Transformers with provable accuracy, but using only linear (as opposed to quadratic) space and time complexity, without relying on any priors such as sparsity or low-rankness. To approximate softmax attention-kernels, Performers use a novel Fast Attention Via positive Orthogonal Random features approach (FAVOR+), which may be of independent interest for scalable kernel methods. FAVOR+ can be also used to efficiently model kernelizable attention mechanisms beyond softmax. This representational power is crucial to accurately compare softmax with other kernels for the first time on large-scale tasks, beyond the reach of regular Transformers, and investigate optimal attention-kernels. Performers are linear architectures fully compatible with regular Transformers and with strong theoretical guarantees: unbiased or nearly-unbiased estimation of the attention matrix, uniform convergence and low estimation variance. We tested Performers on a rich set of tasks stretching from pixel-prediction through text models to protein sequence modeling. We demonstrate competitive results with other examined efficient sparse and dense attention methods, showcasing effectiveness of the novel attention-learning paradigm leveraged by Performers."
                },
                {
                    "title": "Efficient Transformers: A Survey",
                    "abstract": "Transformer model architectures have garnered immense interest lately due to their effectiveness across a range of domains like language, vision, and reinforcement learning. In the field of natural language processing for example, Transformers have become an indispensable staple in the modern deep learning stack. Recently, a dizzying number of \u201cX-former\u201d models have been proposed\u2014Reformer, Linformer, Performer, Longformer, to name a few\u2014which improve upon the original Transformer architecture, many of which make improvements around computational and memory efficiency. With the aim of helping the avid researcher navigate this flurry, this article characterizes a large and thoughtful selection of recent efficiency-flavored \u201cX-former\u201d models, providing an organized and comprehensive overview of existing work and models across multiple domains."
                },
                {
                    "title": "Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention",
                    "abstract": "Transformers achieve remarkable performance in several tasks but due to their quadratic complexity, with respect to the input's length, they are prohibitively slow for very long sequences. To address this limitation, we express the self-attention as a linear dot-product of kernel feature maps and make use of the associativity property of matrix products to reduce the complexity from $\\mathcal{O}\\left(N^2\\right)$ to $\\mathcal{O}\\left(N\\right)$, where $N$ is the sequence length. We show that this formulation permits an iterative implementation that dramatically accelerates autoregressive transformers and reveals their relationship to recurrent neural networks. Our linear transformers achieve similar performance to vanilla transformers and they are up to 4000x faster on autoregressive prediction of very long sequences."
                },
                {
                    "title": "CrisisBERT: A Robust Transformer for Crisis Classification and Contextual Crisis Embedding",
                    "abstract": "Detecting crisis events accurately is an important task, as it allows the relevant authorities to implement necessary actions to mitigate damages. For this purpose, social media serve as a timely information source due to its prevalence and high volume of first-hand accounts. While there are prior works on crises detection, many of them do not perform crisis embedding and classification using state-of-the-art attention-based deep neural networks models, such as Transformers and document-level contextual embeddings. In contrast, we propose CrisisBERT, an end-to-end transformer-based model for two crisis classification tasks, namely crisis detection and crisis recognition, which shows promising results across accuracy and F1 scores. The proposed CrisisBERT model demonstrates superior robustness over various benchmarks, and it includes only marginal performance compromise while extending from 6 to 36 events with a mere 51.4% additional data points. We also propose Crisis2Vec, an attention-based, document-level contextual embedding architecture, for crisis embedding, which achieves better performance than conventional crisis embedding methods such as Word2Vec and GloVe."
                },
                {
                    "title": "TextAttack: A Framework for Adversarial Attacks, Data Augmentation, and Adversarial Training in NLP",
                    "abstract": "While there has been substantial research using adversarial attacks to analyze NLP models, each attack is implemented in its own code repository. It remains challenging to develop NLP attacks and utilize them to improve model performance. This paper introduces TextAttack, a Python framework for adversarial attacks, data augmentation, and adversarial training in NLP. TextAttack builds attacks from four components: a goal function, a set of constraints, a transformation, and a search method. TextAttack\u2019s modular design enables researchers to easily construct attacks from combinations of novel and existing components. TextAttack provides implementations of 16 adversarial attacks from the literature and supports a variety of models and datasets, including BERT and other transformers, and all GLUE tasks. TextAttack also includes data augmentation and adversarial training modules for using components of adversarial attacks to improve model accuracy and robustness.TextAttack is democratizing NLP: anyone can try data augmentation and adversarial training on any model or dataset, with just a few lines of code. Code and tutorials are available at https://github.com/QData/TextAttack."
                },
                {
                    "title": "Longformer: The Long-Document Transformer",
                    "abstract": "Transformer-based models are unable to process long sequences due to their self-attention operation, which scales quadratically with the sequence length. To address this limitation, we introduce the Longformer with an attention mechanism that scales linearly with sequence length, making it easy to process documents of thousands of tokens or longer. Longformer's attention mechanism is a drop-in replacement for the standard self-attention and combines a local windowed attention with a task motivated global attention. Following prior work on long-sequence transformers, we evaluate Longformer on character-level language modeling and achieve state-of-the-art results on text8 and enwik8. In contrast to most prior work, we also pretrain Longformer and finetune it on a variety of downstream tasks. Our pretrained Longformer consistently outperforms RoBERTa on long document tasks and sets new state-of-the-art results on WikiHop and TriviaQA. We finally introduce the Longformer-Encoder-Decoder (LED), a Longformer variant for supporting long document generative sequence-to-sequence tasks, and demonstrate its effectiveness on the arXiv summarization dataset."
                },
                {
                    "title": "RobuTrans: A Robust Transformer-Based Text-to-Speech Model",
                    "abstract": "Recently, neural network based speech synthesis has achieved outstanding results, by which the synthesized audios are of excellent quality and naturalness. However, current neural TTS models suffer from the robustness issue, which results in abnormal audios (bad cases) especially for unusual text (unseen context). To build a neural model which can synthesize both natural and stable audios, in this paper, we make a deep analysis of why the previous neural TTS models are not robust, based on which we propose RobuTrans (Robust Transformer), a robust neural TTS model based on Transformer. Comparing to TransformerTTS, our model first converts input texts to linguistic features, including phonemic features and prosodic features, then feed them to the encoder. In the decoder, the encoder-decoder attention is replaced with a duration-based hard attention mechanism, and the causal self-attention is replaced with a \"pseudo non-causal attention\" mechanism to model the holistic information of the input. Besides, the position embedding is replaced with a 1-D CNN, since it constrains the maximum length of synthesized audio. With these modifications, our model not only fix the robustness problem, but also achieves on parity MOS (4.36) with TransformerTTS (4.37) and Tacotron2 (4.37) on our general set."
                },
                {
                    "title": "Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks",
                    "abstract": "The field of defense strategies against adversarial attacks has significantly grown over the last years, but progress is hampered as the evaluation of adversarial defenses is often insufficient and thus gives a wrong impression of robustness. Many promising defenses could be broken later on, making it difficult to identify the state-of-the-art. Frequent pitfalls in the evaluation are improper tuning of hyperparameters of the attacks, gradient obfuscation or masking. In this paper we first propose two extensions of the PGD-attack overcoming failures due to suboptimal step size and problems of the objective function. We then combine our novel attacks with two complementary existing ones to form a parameter-free, computationally affordable and user-independent ensemble of attacks to test adversarial robustness. We apply our ensemble to over 50 models from papers published at recent top machine learning and computer vision venues. In all except one of the cases we achieve lower robust test accuracy than reported in these papers, often by more than $10\\%$, identifying several broken defenses."
                },
                {
                    "title": "Reformer: The Efficient Transformer",
                    "abstract": "Large Transformer models routinely achieve state-of-the-art results on a number of tasks but training these models can be prohibitively costly, especially on long sequences. We introduce two techniques to improve the efficiency of Transformers. For one, we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O($L^2$) to O($L\\log L$), where $L$ is the length of the sequence. Furthermore, we use reversible residual layers instead of the standard residuals, which allows storing activations only once in the training process instead of $N$ times, where $N$ is the number of layers. The resulting model, the Reformer, performs on par with Transformer models while being much more memory-efficient and much faster on long sequences."
                },
                {
                    "title": "Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer",
                    "abstract": "Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new \"Colossal Clean Crawled Corpus\", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code."
                },
                {
                    "title": "Transformer Dissection: An Unified Understanding for Transformer\u2019s Attention via the Lens of Kernel",
                    "abstract": "Transformer is a powerful architecture that achieves superior performance on various sequence learning tasks, including neural machine translation, language understanding, and sequence prediction. At the core of the Transformer is the attention mechanism, which concurrently processes all inputs in the streams. In this paper, we present a new formulation of attention via the lens of the kernel. To be more precise, we realize that the attention can be seen as applying kernel smoother over the inputs with the kernel scores being the similarities between inputs. This new formulation gives us a better way to understand individual components of the Transformer\u2019s attention, such as the better way to integrate the positional embedding. Another important advantage of our kernel-based formulation is that it paves the way to a larger space of composing Transformer\u2019s attention. As an example, we propose a new variant of Transformer\u2019s attention which models the input as a product of symmetric kernels. This approach achieves competitive performance to the current state of the art model with less computation. In our experiments, we empirically study different kernel construction strategies on two widely used tasks: neural machine translation and sequence prediction."
                },
                {
                    "title": "Analyzing the Structure of Attention in a Transformer Language Model",
                    "abstract": "The Transformer is a fully attention-based alternative to recurrent networks that has achieved state-of-the-art results across a range of NLP tasks. In this paper, we analyze the structure of attention in a Transformer language model, the GPT-2 small pretrained model. We visualize attention for individual instances and analyze the interaction between attention and syntax over a large corpus. We find that attention targets different parts of speech at different layer depths within the model, and that attention aligns with dependency relations most strongly in the middle layers. We also find that the deepest layers of the model capture the most distant relationships. Finally, we extract exemplar sentences that reveal highly specific patterns targeted by particular attention heads."
                },
                {
                    "title": "Understanding and Improving Transformer From a Multi-Particle Dynamic System Point of View",
                    "abstract": "The Transformer architecture is widely used in natural language processing. Despite its success, the design principle of the Transformer remains elusive. In this paper, we provide a novel perspective towards understanding the architecture: we show that the Transformer can be mathematically interpreted as a numerical Ordinary Differential Equation (ODE) solver for a convection-diffusion equation in a multi-particle dynamic system. In particular, how words in a sentence are abstracted into contexts by passing through the layers of the Transformer can be interpreted as approximating multiple particles' movement in the space using the Lie-Trotter splitting scheme and the Euler's method. Given this ODE's perspective, the rich literature of numerical analysis can be brought to guide us in designing effective structures beyond the Transformer. As an example, we propose to replace the Lie-Trotter splitting scheme by the Strang-Marchuk splitting scheme, a scheme that is more commonly used and with much lower local truncation errors. The Strang-Marchuk splitting scheme suggests that the self-attention and position-wise feed-forward network (FFN) sub-layers should not be treated equally. Instead, in each layer, two position-wise FFN sub-layers should be used, and the self-attention sub-layer is placed in between. This leads to a brand new architecture. Such an FFN-attention-FFN layer is \"Macaron-like\", and thus we call the network with this new architecture the Macaron Net. Through extensive experiments, we show that the Macaron Net is superior to the Transformer on both supervised and unsupervised learning tasks. The reproducible codes and pretrained models can be found at this https URL"
                },
                {
                    "title": "BERT Rediscovers the Classical NLP Pipeline",
                    "abstract": "Pre-trained text encoders have rapidly advanced the state of the art on many NLP tasks. We focus on one such model, BERT, and aim to quantify where linguistic information is captured within the network. We find that the model represents the steps of the traditional NLP pipeline in an interpretable and localizable way, and that the regions responsible for each step appear in the expected sequence: POS tagging, parsing, NER, semantic roles, then coreference. Qualitative analysis reveals that the model can and often does adjust this pipeline dynamically, revising lower-level decisions on the basis of disambiguating information from higher-level representations."
                },
                {
                    "title": "Certified Adversarial Robustness via Randomized Smoothing",
                    "abstract": "We show how to turn any classifier that classifies well under Gaussian noise into a new classifier that is certifiably robust to adversarial perturbations under the $\\ell_2$ norm. This \"randomized smoothing\" technique has been proposed recently in the literature, but existing guarantees are loose. We prove a tight robustness guarantee in $\\ell_2$ norm for smoothing with Gaussian noise. We use randomized smoothing to obtain an ImageNet classifier with e.g. a certified top-1 accuracy of 49% under adversarial perturbations with $\\ell_2$ norm less than 0.5 (=127/255). No certified defense has been shown feasible on ImageNet except for smoothing. On smaller-scale datasets where competing approaches to certified $\\ell_2$ robustness are viable, smoothing delivers higher certified accuracies. Our strong empirical results suggest that randomized smoothing is a promising direction for future research into adversarially robust classification. Code and models are available at this http URL."
                },
                {
                    "title": "Transformer-XL: Attentive Language Models beyond a Fixed-Length Context",
                    "abstract": "Transformers have a potential of learning longer-term dependency, but are limited by a fixed-length context in the setting of language modeling. We propose a novel neural architecture Transformer-XL that enables learning dependency beyond a fixed length without disrupting temporal coherence. It consists of a segment-level recurrence mechanism and a novel positional encoding scheme. Our method not only enables capturing longer-term dependency, but also resolves the context fragmentation problem. As a result, Transformer-XL learns dependency that is 80% longer than RNNs and 450% longer than vanilla Transformers, achieves better performance on both short and long sequences, and is up to 1,800+ times faster than vanilla Transformers during evaluation. Notably, we improve the state-of-the-art results of bpc/perplexity to 0.99 on enwiki8, 1.08 on text8, 18.3 on WikiText-103, 21.8 on One Billion Word, and 54.5 on Penn Treebank (without finetuning). When trained only on WikiText-103, Transformer-XL manages to generate reasonably coherent, novel text articles with thousands of tokens. Our code, pretrained models, and hyperparameters are available in both Tensorflow and PyTorch."
                },
                {
                    "title": "Adaptive Input Representations for Neural Language Modeling",
                    "abstract": "We introduce adaptive input representations for neural language modeling which extend the adaptive softmax of Grave et al. (2017) to input representations of variable capacity. There are several choices on how to factorize the input and output layers, and whether to model words, characters or sub-word units. We perform a systematic comparison of popular choices for a self-attentional architecture. Our experiments show that models equipped with adaptive embeddings are more than twice as fast to train than the popular character input CNN while having a lower number of parameters. On the WikiText-103 benchmark we achieve 18.7 perplexity, an improvement of 10.5 perplexity compared to the previously best published result and on the Billion Word benchmark, we achieve 23.02 perplexity."
                },
                {
                    "title": "Character-Level Language Modeling with Deeper Self-Attention",
                    "abstract": "LSTMs and other RNN variants have shown strong performance on character-level language modeling. These models are typically trained using truncated backpropagation through time, and it is common to assume that their success stems from their ability to remember long-term contexts. In this paper, we show that a deep (64-layer) transformer model (Vaswani et al. 2017) with fixed context outperforms RNN variants by a large margin, achieving state of the art on two popular benchmarks: 1.13 bits per character on text8 and 1.06 on enwik8. To get good results at this depth, we show that it is important to add auxiliary losses, both at intermediate network layers and intermediate sequence positions."
                },
                {
                    "title": "Universal Transformers",
                    "abstract": "Recurrent neural networks (RNNs) sequentially process data by updating their state with each new data point, and have long been the de facto choice for sequence modeling tasks. However, their inherently sequential computation makes them slow to train. Feed-forward and convolutional architectures have recently been shown to achieve superior results on some sequence modeling tasks such as machine translation, with the added advantage that they concurrently process all inputs in the sequence, leading to easy parallelization and faster training times. Despite these successes, however, popular feed-forward sequence models like the Transformer fail to generalize in many simple tasks that recurrent models handle with ease, e.g. copying strings or even simple logical inference when the string or formula lengths exceed those observed at training time. We propose the Universal Transformer (UT), a parallel-in-time self-attentive recurrent sequence model which can be cast as a generalization of the Transformer model and which addresses these issues. UTs combine the parallelizability and global receptive field of feed-forward sequence models like the Transformer with the recurrent inductive bias of RNNs. We also add a dynamic per-position halting mechanism and find that it improves accuracy on several tasks. In contrast to the standard Transformer, under certain assumptions, UTs can be shown to be Turing-complete. Our experiments show that UTs outperform standard Transformers on a wide range of algorithmic and language understanding tasks, including the challenging LAMBADA language modeling task where UTs achieve a new state of the art, and machine translation where UTs achieve a 0.9 BLEU improvement over Transformers on the WMT14 En-De dataset."
                },
                {
                    "title": "Bias Reduction and Metric Learning for Nearest-Neighbor Estimation of Kullback-Leibler Divergence",
                    "abstract": "Nearest-neighbor estimators for the Kullback-Leiber (KL) divergence that are asymptotically unbiased have recently been proposed and demonstrated in a number of applications. However, with a small number of samples, nonparametric methods typically suffer from large estimation bias due to the nonlocality of information derived from nearest-neighbor statistics. In this letter, we show that this estimation bias can be mitigated by modifying the metric function, and we propose a novel method for learning a locally optimal Mahalanobis distance function from parametric generative models of the underlying density distributions. Using both simulations and experiments on a variety of data sets, we demonstrate that this interplay between approximate generative models and nonparametric techniques can significantly improve the accuracy of nearest-neighbor-based estimation of the KL divergence."
                },
                {
                    "title": "Learning long-range spatial dependencies with horizontal gated-recurrent units",
                    "abstract": "Progress in deep learning has spawned great successes in many engineering applications. As a prime example, convolutional neural networks, a type of feedforward neural networks, are now approaching -- and sometimes even surpassing -- human accuracy on a variety of visual recognition tasks. Here, however, we show that these neural networks and their recent extensions struggle in recognition tasks where co-dependent visual features must be detected over long spatial ranges. We introduce the horizontal gated-recurrent unit (hGRU) to learn intrinsic horizontal connections -- both within and across feature columns. We demonstrate that a single hGRU layer matches or outperforms all tested feedforward hierarchical baselines including state-of-the-art architectures which have orders of magnitude more free parameters. We further discuss the biological plausibility of the hGRU in comparison to anatomical data from the visual cortex as well as human behavioral data on a classic contour detection task."
                },
                {
                    "title": "ListOps: A Diagnostic Dataset for Latent Tree Learning",
                    "abstract": "Latent tree learning models learn to parse a sentence without syntactic supervision, and use that parse to build the sentence representation. Existing work on such models has shown that, while they perform well on tasks like sentence classification, they do not learn grammars that conform to any plausible semantic or syntactic formalism (Williams et al., 2018a). Studying the parsing ability of such models in natural language can be challenging due to the inherent complexities of natural language, like having several valid parses for a single sentence. In this paper we introduce ListOps, a toy dataset created to study the parsing ability of latent tree models. ListOps sequences are in the style of prefix arithmetic. The dataset is designed to have a single correct parsing strategy that a system needs to learn to succeed at the task. We show that the current leading latent tree models are unable to learn to parse and succeed at ListOps. These models achieve accuracies worse than purely sequential RNNs."
                },
                {
                    "title": "Adversarial Risk and the Dangers of Evaluating Against Weak Attacks",
                    "abstract": "This paper investigates recently proposed approaches for defending against adversarial examples and evaluating adversarial robustness. The existence of adversarial examples in trained neural networks reflects the fact that expected risk alone does not capture the model's performance against worst-case inputs. We motivate the use of adversarial risk as an objective, although it cannot easily be computed exactly. We then frame commonly used attacks and evaluation metrics as defining a tractable surrogate objective to the true adversarial risk. This suggests that models may be obscured to adversaries, by optimizing this surrogate rather than the true adversarial risk. We demonstrate that this is a significant problem in practice by repurposing gradient-free optimization techniques into adversarial attacks, which we use to decrease the accuracy of several recently proposed defenses to near zero. Our hope is that our formulations and results will help researchers to develop more powerful defenses."
                },
                {
                    "title": "Generative Local Metric Learning for Kernel Regression",
                    "abstract": "This paper shows how metric learning can be used with Nadaraya-Watson (NW) kernel regression. Compared with standard approaches, such as bandwidth selection, we show how metric learning can significantly reduce the mean square error (MSE) in kernel regression, particularly for high-dimensional data. We propose a method for efficiently learning a good metric function based upon analyzing the performance of the NW estimator for Gaussian-distributed data. A key feature of our approach is that the NW estimator with a learned metric uses information from both the global and local structure of the training data. Theoretical and empirical results confirm that the learned metric can considerably reduce the bias and MSE for kernel regression even when the data are not confined to Gaussian."
                },
                {
                    "title": "Scene Parsing through ADE20K Dataset",
                    "abstract": "Scene parsing, or recognizing and segmenting objects and stuff in an image, is one of the key problems in computer vision. Despite the communitys efforts in data collection, there are still few image datasets covering a wide range of scenes and object categories with dense and detailed annotations for scene parsing. In this paper, we introduce and analyze the ADE20K dataset, spanning diverse annotations of scenes, objects, parts of objects, and in some cases even parts of parts. A scene parsing benchmark is built upon the ADE20K with 150 object and stuff classes included. Several segmentation baseline models are evaluated on the benchmark. A novel network design called Cascade Segmentation Module is proposed to parse a scene into stuff, objects, and object parts in a cascade and improve over the baselines. We further show that the trained scene parsing networks can lead to applications such as image content removal and scene synthesis1."
                },
                {
                    "title": "Towards Deep Learning Models Resistant to Adversarial Attacks",
                    "abstract": "Recent work has demonstrated that deep neural networks are vulnerable to adversarial examples---inputs that are almost indistinguishable from natural data and yet classified incorrectly by the network. In fact, some of the latest findings suggest that the existence of adversarial attacks may be an inherent weakness of deep learning models. To address this problem, we study the adversarial robustness of neural networks through the lens of robust optimization. This approach provides us with a broad and unifying view on much of the prior work on this topic. Its principled nature also enables us to identify methods for both training and attacking neural networks that are reliable and, in a certain sense, universal. In particular, they specify a concrete security guarantee that would protect against any adversary. These methods let us train networks with significantly improved resistance to a wide range of adversarial attacks. They also suggest the notion of security against a first-order adversary as a natural and broad security guarantee. We believe that robustness against such well-defined classes of adversaries is an important stepping stone towards fully resistant deep learning models. Code and pre-trained models are available at this https URL and this https URL."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "A Structured Self-attentive Sentence Embedding",
                    "abstract": "This paper proposes a new model for extracting an interpretable sentence embedding by introducing self-attention. Instead of using a vector, we use a 2-D matrix to represent the embedding, with each row of the matrix attending on a different part of the sentence. We also propose a self-attention mechanism and a special regularization term for the model. As a side effect, the embedding comes with an easy way of visualizing what specific parts of the sentence are encoded into the embedding. We evaluate our model on 3 different tasks: author profiling, sentiment classification, and textual entailment. Results show that our model yields a significant performance gain compared to other sentence embedding methods in all of the 3 tasks."
                },
                {
                    "title": "Pointer Sentinel Mixture Models",
                    "abstract": "Recent neural network sequence models with softmax classifiers have achieved their best language modeling performance only with very large hidden states and large vocabularies. Even then they struggle to predict rare or unseen words even if the context makes the prediction unambiguous. We introduce the pointer sentinel mixture architecture for neural sequence models which has the ability to either reproduce a word from the recent context or produce a word from a standard softmax classifier. Our pointer sentinel-LSTM model achieves state of the art language modeling performance on the Penn Treebank (70.9 perplexity) while using far fewer parameters than a standard softmax LSTM. In order to evaluate how well language models can exploit longer contexts and deal with more realistic vocabularies and larger corpora we also introduce the freely available WikiText corpus."
                },
                {
                    "title": "A Decomposable Attention Model for Natural Language Inference",
                    "abstract": "We propose a simple neural architecture for natural language inference. Our approach uses attention to decompose the problem into subproblems that can be solved separately, thus making it trivially parallelizable. On the Stanford Natural Language Inference (SNLI) dataset, we obtain state-of-the-art results with almost an order of magnitude fewer parameters than previous work and without relying on any word-order information. Adding intra-sentence attention that takes a minimum amount of order into account yields further improvements."
                },
                {
                    "title": "Explaining and Harnessing Adversarial Examples",
                    "abstract": "Several machine learning models, including neural networks, consistently misclassify adversarial examples---inputs formed by applying small but intentionally worst-case perturbations to examples from the dataset, such that the perturbed input results in the model outputting an incorrect answer with high confidence. Early attempts at explaining this phenomenon focused on nonlinearity and overfitting. We argue instead that the primary cause of neural networks' vulnerability to adversarial perturbation is their linear nature. This explanation is supported by new quantitative results while giving the first explanation of the most intriguing fact about them: their generalization across architectures and training sets. Moreover, this view yields a simple and fast method of generating adversarial examples. Using this approach to provide examples for adversarial training, we reduce the test set error of a maxout network on the MNIST dataset."
                },
                {
                    "title": "Learning Phrase Representations using RNN Encoder\u2013Decoder for Statistical Machine Translation",
                    "abstract": "In this paper, we propose a novel neural network model called RNN Encoder\u2010 Decoder that consists of two recurrent neural networks (RNN). One RNN encodes a sequence of symbols into a fixedlength vector representation, and the other decodes the representation into another sequence of symbols. The encoder and decoder of the proposed model are jointly trained to maximize the conditional probability of a target sequence given a source sequence. The performance of a statistical machine translation system is empirically found to improve by using the conditional probabilities of phrase pairs computed by the RNN Encoder\u2010Decoder as an additional feature in the existing log-linear model. Qualitatively, we show that the proposed model learns a semantically and syntactically meaningful representation of linguistic phrases."
                },
                {
                    "title": "Gradient Weights help Nonparametric Regressors",
                    "abstract": "In regression problems over \u211dd, the unknown function f often varies more in some coordinates than in others. We show that weighting each coordinate i with the estimated norm of the ith derivative of f is an efficient way to significantly improve the performance of distance-based regressors, e.g. kernel and k-NN regressors. We propose a simple estimator of these derivative norms and prove its consistency. Moreover, the proposed estimator is efficiently learned online."
                },
                {
                    "title": "Learning Word Vectors for Sentiment Analysis",
                    "abstract": "Unsupervised vector-based approaches to semantics can model rich lexical meanings, but they largely fail to capture sentiment information that is central to many word meanings and important for a wide range of NLP tasks. We present a model that uses a mix of unsupervised and supervised techniques to learn word vectors capturing semantic term--document information as well as rich sentiment content. The proposed model can leverage both continuous and multi-dimensional sentiment information as well as non-sentiment annotations. We instantiate the model to utilize the document-level sentiment polarity annotations present in many online documents (e.g. star ratings). We evaluate the model using small, widely used sentiment and subjectivity corpora and find it out-performs several previously introduced methods for sentiment classification. We also introduce a large dataset of movie reviews to serve as a more robust benchmark for work in this area."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Distance Metric Learning for Large Margin Nearest Neighbor Classification",
                    "abstract": "The accuracy of k-nearest neighbor (kNN) classification depends significantly on the metric used to compute distances between different examples. In this paper, we show how to learn a Mahalanobis distance metric for kNN classification from labeled examples. The Mahalanobis metric can equivalently be viewed as a global linear transformation of the input space that precedes kNN classification using Euclidean distances. In our approach, the metric is trained with the goal that the k-nearest neighbors always belong to the same class while examples from different classes are separated by a large margin. As in support vector machines (SVMs), the margin criterion leads to a convex optimization based on the hinge loss. Unlike learning in SVMs, however, our approach requires no modification or extension for problems in multiway (as opposed to binary) classification. In our framework, the Mahalanobis distance metric is obtained as the solution to a semidefinite program. On several data sets of varying size and difficulty, we find that metrics trained in this way lead to significant improvements in kNN classification. Sometimes these results can be further improved by clustering the training examples and learning an individual metric within each cluster. We show how to learn and combine these local metrics in a globally integrated manner."
                },
                {
                    "title": "Estimation in a semiparametric partially linear errors-in-variables model",
                    "abstract": "We consider the partially linear model relating a response Y to predictors (X, T) with mean function X T \u03b2 + g(T) when the X's are measured with additive error. The semiparametric likelihood estimate of Severini and Staniswalis leads to biased estimates of both the parameter \u03b2 and the function g(.) when measurement error is ignored. We derive a simple modification of their estimator which is a semiparametric version of the usual parametric correction for attenuation. The resulting estimator of \u03b2 is shown to be consistent and its asymptotic distribution theory is derived. Consistent standard error estimates using sandwich-type ideas are also developed."
                },
                {
                    "title": "A Semiparametric Correction for Attenuation",
                    "abstract": "Abstract A correction method is proposed for models including the generalized linear model when the covariate is measured with error. The method requires a separate validation data set that consists of the surrogate W and the true covariate X or an unbiased estimate X\u2283 of X. We do not require the classical additive measurement error model in which the surrogate is unbiased for the true covariates. We first obtain an estimate of E(X|W) by using nonparametric kernel regression of X or X\u2283 on W based on the validation data. Then we perform a standard analysis with the unknown X replaced by the estimate of E(X|W). The asymptotic distribution of the resulting regression parameter estimator is obtained. Generalizations to include components of X measured without error are also discussed."
                },
                {
                    "title": "FourierFormer: Transformer Meets Generalized Fourier Integral Theorem",
                    "abstract": "Multi-head attention empowers the recent success of transformers, the state-of-the-art models that have achieved remarkable success in sequence modeling and beyond. These attention mechanisms compute the pairwise dot products between the queries and keys, which results from the use of unnormalized Gaussian kernels with the assumption that the queries follow a mixture of Gaussian distribution. There is no guarantee that this assumption is valid in practice. In response, we \ufb01rst interpret attention in transformers as a nonparametric kernel regression. We then propose the FourierFormer, a new class of transformers in which the dot-product kernels are replaced by the novel generalized Fourier integral kernels. Different from the dot-product kernels, where we need to choose a good covariance matrix to capture the dependency of the features of data, the generalized Fourier integral kernels can automatically capture such dependency and remove the need to tune the covariance matrix. We theoretically prove that our proposed Fourier integral kernels can ef\ufb01-ciently approximate any key and query distributions. Compared to the conventional transformers with dot-product attention, FourierFormers attain better accuracy and reduce the redundancy between attention heads. We empirically corroborate the advantages of FourierFormers over the baseline transformers in a variety of practical applications including language modeling and image classi\ufb01cation."
                },
                {
                    "title": "Probabilistic Transformer For Time Series Analysis",
                    "abstract": "Generative modeling of multivariate time series has remained challenging partly due to the complex, non-deterministic dynamics across long-distance time steps. In this paper, we propose deep probabilistic methods that combine state-space models (SSMs) with transformer architectures. In contrast to previously proposed SSMs, our approaches use attention mechanism to model non-Markovian dynamics in the latent space and avoid recurrent neural networks entirely. We also extend our models to include several layers of stochastic variables organized in a hierarchy for further expressiveness. Compared to transformer models, ours are probabilistic, non-autoregressive, and capable of generating diverse long-term forecasts with accounted uncertainty. Extensive experiments show that our models consistently outperform competitive baselines on various tasks and datasets, including time series forecasting and human motion prediction."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                },
                {
                    "title": "Improving Language Understanding by Generative Pre-Training",
                    "abstract": "Natural language understanding comprises a wide range of diverse tasks such as textual entailment, question answering, semantic similarity assessment, and document classi\ufb01cation. Although large unlabeled text corpora are abundant, labeled data for learning these speci\ufb01c tasks is scarce, making it challenging for discriminatively trained models to perform adequately. We demonstrate that large gains on these tasks can be realized by generative pre-training of a language model on a diverse corpus of unlabeled text, followed by discriminative \ufb01ne-tuning on each speci\ufb01c task. In contrast to previous approaches, we make use of task-aware input transformations during \ufb01ne-tuning to achieve effective transfer while requiring minimal changes to the model architecture. We demonstrate the effectiveness of our approach on a wide range of benchmarks for natural language understanding. Our general task-agnostic model outperforms discriminatively trained models that use architectures speci\ufb01cally crafted for each task, signi\ufb01cantly improving upon the state of the art in 9 out of the 12 tasks studied. For instance, we achieve absolute improvements of 8.9% on commonsense reasoning (Stories Cloze Test), 5.7% on question answering (RACE), and 1.5% on textual entailment (MultiNLI)."
                },
                {
                    "title": "Gradients Weights improve Regression and Classification",
                    "abstract": "In regression problems over Rd, the unknown function f often varies more in some coordinates than in others. We show that weighting each coordinate i according to an estimate of the variation of f along coordinate i - e.g. the L1 norm of the ith-directional derivative of f - is an efficient way to significantly improve the performance of distance-based regressors such as kernel and k- NN regressors. The approach, termed Gradient Weighting (GW), consists of a first pass regression estimate fn which serves to evaluate the directional derivatives of f, and a second-pass regression estimate on the re-weighted data. The GW approach can be instantiated for both regression and classification, and is grounded in strong theoretical principles having to do with the way regression bias and variance are affected by a generic feature-weighting scheme. These theoretical principles provide further technical foundation for some existing feature-weighting heuristics that have proved successful in practice. \n \nWe propose a simple estimator of these derivative norms and prove its consistency. The proposed estimator computes efficiently and easily extends to run online. We then derive a classification version of the GW approach which evaluates on real-worlds datasets with as much success as its regression counterpart."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Correcting for regression dilution bias: comparison of methods for a single predictor variable",
                    "abstract": "In an epidemiological study the regression slope between a response and predictor variable is underestimated when the predictor variable is measured imprecisely. Repeat measurements of the predictor in individuals in a subset of the study or in a separate study can be used to estimate a multiplicative factor to correct for this \u2018regression dilution bias\u2019. In applied statistics publications various methods have been used to estimate this correction factor. Here we compare six different estimation methods and explain how they fall into two categories, namely regression and correlation\u2010based methods. We provide new asymptotic variance formulae for the optimal correction factors in each category, when these are estimated from the repeat measurements subset alone, and show analytically and by simulation that the correlation method of choice gives uniformly lower variance. The simulations also show that, when the correction factor is not much greater than 1, this correlation method gives a correction factor which is closer to the true value than that from the best regression method on up to 80% of occasions. We also provide a variance formula for a modified correlation method which uses the standard deviation of the predictor variable in the main study; this shows further improved performance provided that the correction factor is not too extreme. A confidence interval for a corrected regression slope in an epidemiological study should reflect the imprecision of both the uncorrected slope and the estimated correction factor. We provide formulae for this and show that, particularly when the correction factor is large and the size of the subset of repeat measures is small, the effect of allowing for imprecision in the estimated correction factor can be substantial."
                },
                {
                    "title": "On Estimating Regression",
                    "abstract": "A study is made of certain properties of an approximation to the regression line on the basis of sampling data when the sample size increases unboundedly."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.CL",
                "cs.CV",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we improve the robustness and capacity of self-attention mechanisms in transformer models by addressing the limitations of isotropic Gaussian kernels in estimating contextual relevance?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the performance of transformer models across various machine learning tasks, particularly in natural language processing, computer vision, and reinforcement learning. By enhancing the self-attention mechanism, we can achieve more accurate and robust predictions, which will have significant implications for future research in model interpretability and efficiency. This work could lead to practical applications in areas requiring high reliability, such as healthcare and autonomous systems, where the consequences of errors can be severe.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the inherent limitations of isotropic Gaussian kernels, which assume equal importance across all dimensions of the feature space. This spherical invariance leads to high variance in the self-attention output, making the model sensitive to small input perturbations and resulting in uninformative noise in hidden representations. Naive approaches that do not account for coordinate-wise relevance will likely fail to capture the nuanced relationships between tokens, leading to suboptimal performance and robustness.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the standard self-attention mechanisms without addressing the implications of spherical invariance in kernel regression. The lack of a method to estimate coordinate-wise relevance efficiently has been a significant barrier. Our approach differs by introducing Elliptical Attention, which constructs hyper-ellipsoidal neighborhoods that allow for a more nuanced understanding of token relationships, thereby overcoming the limitations of prior work.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of Elliptical Attention, which utilizes a Mahalanobis transformation to create hyper-ellipsoidal neighborhoods around attention queries. We will evaluate this approach using standard datasets in natural language processing and computer vision, measuring performance through metrics such as accuracy and robustness against input perturbations. The expected outcomes include improved contextual representations, reduced variance in attention outputs, and enhanced model robustness, leading to more reliable predictions in various applications."
            }
        },
        "author_data": {
            "b2d8d090-bd96-4371-868e-54d1453e05ad": {
                "pk": "b2d8d090-bd96-4371-868e-54d1453e05ad",
                "name": "Stefan K. Nielsen",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "8411c1c5-bee7-4379-931b-c0dfc9e17093": {
                "pk": "8411c1c5-bee7-4379-931b-c0dfc9e17093",
                "name": "Laziz U. Abdullaev",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "413669d1-07cf-40ea-a0e3-fbd79035a2e7": {
                "pk": "413669d1-07cf-40ea-a0e3-fbd79035a2e7",
                "name": "Rachel Teo",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "69214e73-dc51-4a38-a623-3f713adc78ca": {
                "pk": "69214e73-dc51-4a38-a623-3f713adc78ca",
                "name": "Tan M. Nguyen",
                "collaborators": [
                    "Tam Nguyen",
                    "Tuan Nguyen",
                    "Vinh Nguyen",
                    "Rachel S. Y. Teo",
                    "Richard G. Baraniuk",
                    "Hirotada Honda",
                    "Takashi Sano",
                    "Shugo Nakamura",
                    "C\u00e9sar A. Uribe"
                ],
                "domain": [
                    "Transformers",
                    "Graph Neural Network",
                    "Robustness",
                    "Regularization"
                ],
                "publications": [
                    {
                        "title": "p-Laplacian Transformer",
                        "abstract": "$p$-Laplacian regularization, rooted in graph and image signal processing, introduces a parameter $p$ to control the regularization effect on these data. Smaller values of $p$ promote sparsity and interpretability, while larger values encourage smoother solutions. In this paper, we first show that the self-attention mechanism obtains the minimal Laplacian regularization ($p=2$) and encourages the smoothness in the architecture. However, the smoothness is not suitable for the heterophilic structure of self-attention in transformers where attention weights between tokens that are in close proximity and non-close ones are assigned indistinguishably. From that insight, we then propose a novel class of transformers, namely the $p$-Laplacian Transformer (p-LaT), which leverages $p$-Laplacian regularization framework to harness the heterophilic features within self-attention layers. In particular, low $p$ values will effectively assign higher attention weights to tokens that are in close proximity to the current token being processed. We empirically demonstrate the advantages of p-LaT over the baseline transformers on a wide range of benchmark datasets."
                    },
                    {
                        "title": "Unveiling the Hidden Structure of Self-Attention via Kernel Principal Component Analysis",
                        "abstract": "The remarkable success of transformers in sequence modeling tasks, spanning various applications in natural language processing and computer vision, is attributed to the critical role of self-attention. Similar to the development of most deep learning models, the construction of these attention mechanisms rely on heuristics and experience. In our work, we derive self-attention from kernel principal component analysis (kernel PCA) and show that self-attention projects its query vectors onto the principal component axes of its key matrix in a feature space. We then formulate the exact formula for the value matrix in self-attention, theoretically and empirically demonstrating that this value matrix captures the eigenvectors of the Gram matrix of the key vectors in self-attention. Leveraging our kernel PCA framework, we propose Attention with Robust Principal Components (RPC-Attention), a novel class of robust attention that is resilient to data contamination. We empirically demonstrate the advantages of RPC-Attention over softmax attention on the ImageNet-1K object classification, WikiText-103 language modeling, and ADE20K image segmentation task."
                    },
                    {
                        "title": "MomentumSMoE: Integrating Momentum into Sparse Mixture of Experts",
                        "abstract": "Sparse Mixture of Experts (SMoE) has become the key to unlocking unparalleled scalability in deep learning. SMoE has the potential to exponentially increase parameter count while maintaining the efficiency of the model by only activating a small subset of these parameters for a given sample. However, it has been observed that SMoE suffers from unstable training and has difficulty adapting to new distributions, leading to the model's lack of robustness to data contamination. To overcome these limitations, we first establish a connection between the dynamics of the expert representations in SMoEs and gradient descent on a multi-objective optimization problem. Leveraging our framework, we then integrate momentum into SMoE and propose a new family of SMoEs named MomentumSMoE. We theoretically prove and numerically demonstrate that MomentumSMoE is more stable and robust than SMoE. In particular, we verify the advantages of MomentumSMoE over SMoE on a variety of practical tasks including ImageNet-1K object recognition and WikiText-103 language modeling. We demonstrate the applicability of MomentumSMoE to many types of SMoE models, including those in the Sparse MoE model for vision (V-MoE) and the Generalist Language Model (GLaM). We also show that other advanced momentum-based optimization methods, such as Adam, can be easily incorporated into the MomentumSMoE framework for designing new SMoE models with even better performance, almost negligible additional computation cost, and simple implementations."
                    },
                    {
                        "title": "Mitigating Over-smoothing in Transformers via Regularized Nonlocal Functionals",
                        "abstract": "Transformers have achieved remarkable success in a wide range of natural language processing and computer vision applications. However, the representation capacity of a deep transformer model is degraded due to the over-smoothing issue in which the token representations become identical when the model's depth grows. In this work, we show that self-attention layers in transformers minimize a functional which promotes smoothness, thereby causing token uniformity. We then propose a novel regularizer that penalizes the norm of the difference between the smooth output tokens from self-attention and the input tokens to preserve the fidelity of the tokens. Minimizing the resulting regularized energy functional, we derive the Neural Transformer with a Regularized Nonlocal Functional (NeuTRENO), a novel class of transformer models that can mitigate the over-smoothing issue. We empirically demonstrate the advantages of NeuTRENO over the baseline transformers and state-of-the-art methods in reducing the over-smoothing of token representations on various practical tasks, including object classification, image segmentation, and language modeling."
                    },
                    {
                        "title": "From Coupled Oscillators to Graph Neural Networks: Reducing Over-smoothing via a Kuramoto Model-based Approach",
                        "abstract": "We propose the Kuramoto Graph Neural Network (KuramotoGNN), a novel class of continuous-depth graph neural networks (GNNs) that employs the Kuramoto model to mitigate the over-smoothing phenomenon, in which node features in GNNs become indistinguishable as the number of layers increases. The Kuramoto model captures the synchronization behavior of non-linear coupled oscillators. Under the view of coupled oscillators, we first show the connection between Kuramoto model and basic GNN and then over-smoothing phenomenon in GNNs can be interpreted as phase synchronization in Kuramoto model. The KuramotoGNN replaces this phase synchronization with frequency synchronization to prevent the node features from converging into each other while allowing the system to reach a stable synchronized state. We experimentally verify the advantages of the KuramotoGNN over the baseline GNNs and existing methods in reducing over-smoothing on various graph deep learning benchmark tasks."
                    },
                    {
                        "title": "PIDformer: Transformer Meets Control Theory",
                        "abstract": "In this work, we address two main shortcomings of transformer architectures: input corruption and rank collapse in their output representation. We unveil self-attention as an autonomous state-space model that inherently promotes smoothness in its solutions, leading to lower-rank outputs and diminished representation capacity. Moreover, the steady-state solution of the model is sensitive to input perturbations. We incorporate a Proportional-Integral-Derivative (PID) closed-loop feedback control system with a reference point into the model to improve robustness and representation capacity. This integration aims to preserve high-frequency details while bolstering model stability, rendering it more noise-resilient. The resulting controlled state-space model is theoretically proven robust and adept at addressing the rank collapse. Motivated by this control framework, we derive a novel class of transformers, PID-controlled Transformer (PIDformer), aimed at improving robustness and mitigating the rank-collapse issue inherent in softmax transformers. We empirically evaluate the model for advantages and robustness against baseline transformers across various practical tasks, including object classification, image segmentation, and language modeling."
                    }
                ]
            }
        }
    },
    "2402.00793": {
        "paper_data": {
            "title": "Human Expertise in Algorithmic Prediction",
            "url": "http://arxiv.org/abs/2402.00793v2",
            "arxiv_id": "2402.00793",
            "authors": [
                "Rohan Alur",
                "Manish Raghavan",
                "Devavrat Shah"
            ],
            "abstract": "We introduce a novel framework for incorporating human expertise into algorithmic predictions. Our approach focuses on the use of human judgment to distinguish inputs which `look the same' to any feasible predictive algorithm. We argue that this framing clarifies the problem of human/AI collaboration in prediction tasks, as experts often have access to information -- particularly subjective information -- which is not encoded in the algorithm's training data. We use this insight to develop a set of principled algorithms for selectively incorporating human feedback only when it improves the performance of any feasible predictor. We find empirically that although algorithms often outperform their human counterparts on average, human judgment can significantly improve algorithmic predictions on specific instances (which can be identified ex-ante). In an X-ray classification task, we find that this subset constitutes nearly 30% of the patient population. Our approach provides a natural way of uncovering this heterogeneity and thus enabling effective human-AI collaboration.",
            "introduction": "   1 Introduction  Despite remarkable advances in machine learning, human judgment continues to play a critical role in many high-stakes prediction tasks. For example, consider the problem of triage in the emergency room, where healthcare providers assess and prioritize patients for immediate care. On one hand, prognostic algorithms offer significant promise for improving triage decisions; indeed, algorithmic predictions are often more accurate than even expert human decision makers (Cowgill,, 2018; Dawes et\u00a0al.,, 1989; Grove et\u00a0al.,, 2000; Kleinberg et\u00a0al.,, 2017; Kuncel et\u00a0al., 2013a, ; Dawes,, 1971; Currie and MacLeod,, 2017; Mullainathan and Obermeyer,, 2019). On the other hand, predictive algorithms may fail to fully capture the relevant context for each individual. For example, an algorithmic risk score may only have access to tabular electronic health records or other structured data (e.g., medical imaging), while a physician has access to many additional modalities\u2014not least of which is the ability to directly examine the patient!   These two observations\u2014that algorithms often outperform humans, but humans often have access to a richer information set\u2014are not in conflict with each other. Indeed, Alur et\u00a0al., (2023) find exactly this phenomenon in an analysis of emergency room triage decisions. This suggests that, even in settings where algorithms outperform humans, algorithms might still benefit from some form of human input. Ideally this collaboration will yield human-AI complementarity (Bansal et\u00a0al., 2020b, ; Rastogi et\u00a0al.,, 2022), in which a joint system outperforms either a human or algorithm working alone. Our work thus begins with the following question:   When (and how) can human judgment improve the predictions of any learning algorithm?   Example: X-ray classification. Consider the problem of diagnosing atelectasis (a partially or fully collapsed lung; we study this task in detail in Section\u00a05). Today\u2019s state-of-the-art deep learning models can perform well on these kinds of classification tasks using only a patient\u2019s chest X-ray as input (Irvin et\u00a0al.,, 2019; Rajpurkar et\u00a0al.,, 2021; Agarwal et\u00a0al., 2023a, ). We are interested in whether we can further improve these algorithmic predictions by incorporating a \u201csecond opinion\u201d from a physician, particularly because the physician may have access to information (e.g., by directly observing the patient) which is not present in the X-ray.   A first heuristic, without making any assumptions about the available predictive models, is to ask whether a physician can distinguish patients whose imaging data are identical. For example, if a physician can correctly indicate that one patient is suffering from atelectasis while another is not\u2014despite the patients having identical chest X-rays\u2014the physician must have information that the X-ray does not capture. In principle, this could form the basis for a statistical test: we could ask whether the physician performs better than random in distinguishing a large number of such patients. If so, even a predictive algorithm which outperforms the physician might benefit from human input.   Of course, we are unlikely to find identical observations in continuous-valued and/or high-dimensional data (including X-rays). A natural relaxation is to instead consider observations which are sufficiently \u201csimilar\u201d, as suggested by Alur et\u00a0al., (2023). In this work we propose a more general notion of algorithmic indistinguishability, or coarser subsets of inputs in which no algorithm (in some",
            "references": [
                {
                    "title": "Identifying Prediction Mistakes in Observational Data",
                    "abstract": "\n Decision makers, such as doctors, judges, and managers, make consequential choices based on predictions of unknown outcomes. Do these decision makers make systematic prediction mistakes based on the available information? If so, in what ways are their predictions systematically biased? In this article, I characterize conditions under which systematic prediction mistakes can be identified in empirical settings such as hiring, medical diagnosis, and pretrial release. I derive a statistical test for whether the decision maker makes systematic prediction mistakes under these assumptions and provide methods for estimating the ways the decision maker\u2019s predictions are systematically biased. I analyze the pretrial release decisions of judges in New York City, estimating that at least 20% of judges make systematic prediction mistakes about misconduct risk given defendant characteristics. Motivated by this analysis, I estimate the effects of replacing judges with algorithmic decision rules and find that replacing judges with algorithms where systematic prediction mistakes occur dominates the status quo."
                },
                {
                    "title": "Feedback Loops With Language Models Drive In-Context Reward Hacking",
                    "abstract": "Language models influence the external world: they query APIs that read and write to web pages, generate content that shapes human behavior, and run system commands as autonomous agents. These interactions form feedback loops: LLM outputs affect the world, which in turn affect subsequent LLM outputs. In this work, we show that feedback loops can cause in-context reward hacking (ICRH), where the LLM at test-time optimizes a (potentially implicit) objective but creates negative side effects in the process. For example, consider an LLM agent deployed to increase Twitter engagement; the LLM may retrieve its previous tweets into the context window and make them more controversial, increasing engagement but also toxicity. We identify and study two processes that lead to ICRH: output-refinement and policy-refinement. For these processes, evaluations on static datasets are insufficient -- they miss the feedback effects and thus cannot capture the most harmful behavior. In response, we provide three recommendations for evaluation to capture more instances of ICRH. As AI development accelerates, the effects of feedback loops will proliferate, increasing the need to understand their role in shaping LLM behavior."
                },
                {
                    "title": "When accurate prediction models yield harmful self-fulfilling prophecies",
                    "abstract": "Prediction models are popular in medical research and practice. By predicting an outcome of interest for specific patients, these models may help inform difficult treatment decisions, and are often hailed as the poster children for personalized, data-driven healthcare. We show however, that using prediction models for decision making can lead to harmful decisions, even when the predictions exhibit good discrimination after deployment. These models are harmful self-fulfilling prophecies: their deployment harms a group of patients but the worse outcome of these patients does not invalidate the predictive power of the model. Our main result is a formal characterization of a set of such prediction models. Next we show that models that are well calibrated before and after deployment are useless for decision making as they made no change in the data distribution. These results point to the need to revise standard practices for validation, deployment and evaluation of prediction models that are used in medical decisions."
                },
                {
                    "title": "Performative Prediction: Past and Future",
                    "abstract": "Predictions in the social world generally influence the target of prediction, a phenomenon known as performativity. Self-fulfilling and self-negating predictions are examples of performativity. Of fundamental importance to economics, finance, and the social sciences, the notion has been absent from the development of machine learning. In machine learning applications, performativity often surfaces as distribution shift. A predictive model deployed on a digital platform, for example, influences consumption and thereby changes the data-generating distribution. We survey the recently founded area of performative prediction that provides a definition and conceptual framework to study performativity in machine learning. A consequence of performative prediction is a natural equilibrium notion that gives rise to new optimization challenges. Another consequence is a distinction between learning and steering, two mechanisms at play in performative prediction. The notion of steering is in turn intimately related to questions of power in digital markets. We review the notion of performative power that gives an answer to the question how much a platform can steer participants through its predictions. We end on a discussion of future directions, such as the role that performativity plays in contesting algorithmic systems."
                },
                {
                    "title": "Performative Time-Series Forecasting",
                    "abstract": "Time-series forecasting is a critical challenge in various domains and has witnessed substantial progress in recent years. Many real-life scenarios, such as public health, economics, and social applications, involve feedback loops where predictions can influence the predicted outcome, subsequently altering the target variable's distribution. This phenomenon, known as performativity, introduces the potential for 'self-negating' or 'self-fulfilling' predictions. Despite extensive studies in classification problems across domains, performativity remains largely unexplored in the context of time-series forecasting from a machine-learning perspective. In this paper, we formalize performative time-series forecasting (PeTS), addressing the challenge of accurate predictions when performativity-induced distribution shifts are possible. We propose a novel approach, Feature Performative-Shifting (FPS), which leverages the concept of delayed response to anticipate distribution shifts and subsequently predicts targets accordingly. We provide theoretical insights suggesting that FPS can potentially lead to reduced generalization error. We conduct comprehensive experiments using multiple time-series models on COVID-19 and traffic forecasting tasks. The results demonstrate that FPS consistently outperforms conventional time-series forecasting methods, highlighting its efficacy in handling performativity-induced challenges."
                },
                {
                    "title": "Ecosystem-level Analysis of Deployed Machine Learning Reveals Homogeneous Outcomes",
                    "abstract": "Machine learning is traditionally studied at the model level: researchers measure and improve the accuracy, robustness, bias, efficiency, and other dimensions of specific models. In practice, the societal impact of machine learning is determined by the surrounding context of machine learning deployments. To capture this, we introduce ecosystem-level analysis: rather than analyzing a single model, we consider the collection of models that are deployed in a given context. For example, ecosystem-level analysis in hiring recognizes that a job candidate's outcomes are not only determined by a single hiring algorithm or firm but instead by the collective decisions of all the firms they applied to. Across three modalities (text, images, speech) and 11 datasets, we establish a clear trend: deployed machine learning is prone to systemic failure, meaning some users are exclusively misclassified by all models available. Even when individual models improve at the population level over time, we find these improvements rarely reduce the prevalence of systemic failure. Instead, the benefits of these improvements predominantly accrue to individuals who are already correctly classified by other models. In light of these trends, we consider medical imaging for dermatology where the costs of systemic failure are especially high. While traditional analyses reveal racial performance disparities for both models and humans, ecosystem-level analysis reveals new forms of racial disparity in model predictions that do not present in human predictions. These examples demonstrate ecosystem-level analysis has unique strengths for characterizing the societal impact of machine learning."
                },
                {
                    "title": "Combining Human Expertise with Artificial Intelligence: Experimental Evidence from Radiology",
                    "abstract": "Artificial intelligence (AI) algorithms have matched or surpassed the performance of human experts in a number of predictive tasks, although human experts have access to contextual information that may not be available for machine predictions. We investigate how best to combine machine predictions with human input in the presence of such contextual information and potential biases in how humans use machine predictions in forming their assessments. Our experiment varies the information environment for radiologists diagnosing patients with potential thoracic pathologies to measure (i) the value of contextual information (clinical history and notes) and (ii) the extent of automation bias/neglect and own-information bias/neglect relative to the Bayesian benchmark. We find that that contextual information improves diagnostic accuracy on average, but providing AI predictions does not always increase accuracy. AI assistance reduces accuracy because subjects partially neglect both their own-information and AI predictions when updating their beliefs. Therefore, AI assistance should be deployed selectively as a function of how predictions are utilized by human experts."
                },
                {
                    "title": "Auditing for Human Expertise",
                    "abstract": "High-stakes prediction tasks (e.g., patient diagnosis) are often handled by trained human experts. A common source of concern about automation in these settings is that experts may exercise intuition that is difficult to model and/or have access to information (e.g., conversations with a patient) that is simply unavailable to a would-be algorithm. This raises a natural question whether human experts add value which could not be captured by an algorithmic predictor. We develop a statistical framework under which we can pose this question as a natural hypothesis test. Indeed, as our framework highlights, detecting human expertise is more subtle than simply comparing the accuracy of expert predictions to those made by a particular learning algorithm. Instead, we propose a simple procedure which tests whether expert predictions are statistically independent from the outcomes of interest after conditioning on the available inputs (`features'). A rejection of our test thus suggests that human experts may add value to any algorithm trained on the available data, and has direct implications for whether human-AI `complementarity' is achievable in a given prediction task. We highlight the utility of our procedure using admissions data collected from the emergency department of a large academic hospital system, where we show that physicians' admit/discharge decisions for patients with acute gastrointestinal bleeding (AGIB) appear to be incorporating information that is not available to a standard algorithmic screening tool. This is despite the fact that the screening tool is arguably more accurate than physicians' discretionary decisions, highlighting that -- even absent normative concerns about accountability or interpretability -- accuracy is insufficient to justify algorithmic automation."
                },
                {
                    "title": "Human-Aligned Calibration for AI-Assisted Decision Making",
                    "abstract": "Whenever a binary classifier is used to provide decision support, it typically provides both a label prediction and a confidence value. Then, the decision maker is supposed to use the confidence value to calibrate how much to trust the prediction. In this context, it has been often argued that the confidence value should correspond to a well calibrated estimate of the probability that the predicted label matches the ground truth label. However, multiple lines of empirical evidence suggest that decision makers have difficulties at developing a good sense on when to trust a prediction using these confidence values. In this paper, our goal is first to understand why and then investigate how to construct more useful confidence values. We first argue that, for a broad class of utility functions, there exist data distributions for which a rational decision maker is, in general, unlikely to discover the optimal decision policy using the above confidence values -- an optimal decision maker would need to sometimes place more (less) trust on predictions with lower (higher) confidence values. However, we then show that, if the confidence values satisfy a natural alignment property with respect to the decision maker's confidence on her own predictions, there always exists an optimal decision policy under which the level of trust the decision maker would need to place on predictions is monotone on the confidence values, facilitating its discoverability. Further, we show that multicalibration with respect to the decision maker's confidence on her own predictions is a sufficient condition for alignment. Experiments on four different AI-assisted decision making tasks where a classifier provides decision support to real human experts validate our theoretical results and suggest that alignment may lead to better decisions."
                },
                {
                    "title": "Plug-in Performative Optimization",
                    "abstract": "When predictions are performative, the choice of which predictor to deploy influences the distribution of future observations. The overarching goal in learning under performativity is to find a predictor that has low \\emph{performative risk}, that is, good performance on its induced distribution. One family of solutions for optimizing the performative risk, including bandits and other derivative-free methods, is agnostic to any structure in the performative feedback, leading to exceedingly slow convergence rates. A complementary family of solutions makes use of explicit \\emph{models} for the feedback, such as best-response models in strategic classification, enabling faster rates. However, these rates critically rely on the feedback model being correct. In this work we study a general protocol for making use of possibly misspecified models in performative prediction, called \\emph{plug-in performative optimization}. We show this solution can be far superior to model-agnostic strategies, as long as the misspecification is not too extreme. Our results support the hypothesis that models, even if misspecified, can indeed help with learning in performative settings."
                },
                {
                    "title": "Incentivizing honest performative predictions with proper scoring rules",
                    "abstract": "Proper scoring rules incentivize experts to accurately report beliefs, assuming predictions cannot influence outcomes. We relax this assumption and investigate incentives when predictions are performative, i.e., when they can influence the outcome of the prediction, such as when making public predictions about the stock market. We say a prediction is a fixed point if it accurately reflects the expert's beliefs after that prediction has been made. We show that in this setting, reports maximizing expected score generally do not reflect an expert's beliefs, and we give bounds on the inaccuracy of such reports. We show that, for binary predictions, if the influence of the expert's prediction on outcomes is bounded, it is possible to define scoring rules under which optimal reports are arbitrarily close to fixed points. However, this is impossible for predictions over more than two outcomes. We also perform numerical simulations in a toy setting, showing that our bounds are tight in some situations and that prediction error is often substantial (greater than 5-10%). Lastly, we discuss alternative notions of optimality, including performative stability, and show that they incentivize reporting fixed points."
                },
                {
                    "title": "Swap Agnostic Learning, or Characterizing Omniprediction via Multicalibration",
                    "abstract": "We introduce and study Swap Agnostic Learning. The problem can be phrased as a game between a predictor and an adversary: first, the predictor selects a hypothesis $h$; then, the adversary plays in response, and for each level set of the predictor $\\{x \\in \\mathcal{X} : h(x) = v\\}$ selects a (different) loss-minimizing hypothesis $c_v \\in \\mathcal{C}$; the predictor wins if $h$ competes with the adaptive adversary's loss. Despite the strength of the adversary, we demonstrate the feasibility Swap Agnostic Learning for any convex loss. Somewhat surprisingly, the result follows through an investigation into the connections between Omniprediction and Multicalibration. Omniprediction is a new notion of optimality for predictors that strengthtens classical notions such as agnostic learning. It asks for loss minimization guarantees (relative to a hypothesis class) that apply not just for a specific loss function, but for any loss belonging to a rich family of losses. A recent line of work shows that omniprediction is implied by multicalibration and related multi-group fairness notions. This unexpected connection raises the question: is multi-group fairness necessary for omniprediction? Our work gives the first affirmative answer to this question. We establish an equivalence between swap variants of omniprediction and multicalibration and swap agnostic learning. Further, swap multicalibration is essentially equivalent to the standard notion of multicalibration, so existing learning algorithms can be used to achieve any of the three notions. Building on this characterization, we paint a complete picture of the relationship between different variants of multi-group fairness, omniprediction, and Outcome Indistinguishability. This inquiry reveals a unified notion of OI that captures all existing notions of omniprediction and multicalibration."
                },
                {
                    "title": "Multicalibration as Boosting for Regression",
                    "abstract": "We study the connection between multicalibration and boosting for squared error regression. First we prove a useful characterization of multicalibration in terms of a ``swap regret'' like condition on squared error. Using this characterization, we give an exceedingly simple algorithm that can be analyzed both as a boosting algorithm for regression and as a multicalibration algorithm for a class H that makes use only of a standard squared error regression oracle for H. We give a weak learning assumption on H that ensures convergence to Bayes optimality without the need to make any realizability assumptions -- giving us an agnostic boosting algorithm for regression. We then show that our weak learning assumption on H is both necessary and sufficient for multicalibration with respect to H to imply Bayes optimality. We also show that if H satisfies our weak learning condition relative to another class C then multicalibration with respect to H implies multicalibration with respect to C. Finally we investigate the empirical performance of our algorithm experimentally using an open source implementation that we make available. Our code repository can be found at https://github.com/Declancharrison/Level-Set-Boosting."
                },
                {
                    "title": "Making Decisions under Outcome Performativity",
                    "abstract": "Decision-makers often act in response to data-driven predictions, with the goal of achieving favorable outcomes. In such settings, predictions don't passively forecast the future; instead, predictions actively shape the distribution of outcomes they are meant to predict. This performative prediction setting raises new challenges for learning\"optimal\"decision rules. In particular, existing solution concepts do not address the apparent tension between the goals of forecasting outcomes accurately and steering individuals to achieve desirable outcomes. To contend with this concern, we introduce a new optimality concept -- performative omniprediction -- adapted from the supervised (non-performative) learning setting. A performative omnipredictor is a single predictor that simultaneously encodes the optimal decision rule with respect to many possibly-competing objectives. Our main result demonstrates that efficient performative omnipredictors exist, under a natural restriction of performative prediction, which we call outcome performativity. On a technical level, our results follow by carefully generalizing the notion of outcome indistinguishability to the outcome performative setting. From an appropriate notion of Performative OI, we recover many consequences known to hold in the supervised setting, such as omniprediction and universal adaptability."
                },
                {
                    "title": "Anticipating Performativity by Predicting from Predictions",
                    "abstract": "Predictions about people, such as their expected educational achievement or their credit risk, can be performative and shape the outcome that they aim to predict. Understanding the causal effect of these predictions on the eventual outcomes is crucial for foreseeing the implications of future predictive models and selecting which models to deploy. However, this causal estimation task poses unique challenges: model predictions are usually deterministic functions of input features and highly correlated with outcomes. This can make the causal effects of predictions on outcomes impossible to disentangle from the direct effect of the covariates. We study this problem through the lens of causal identifiability, and despite the hardness of this problem in full generality, we highlight three natural scenarios where the causal relationship between covariates, predictions and outcomes can be identified from observational data: randomization in predictions, overparameterization of the predictive model deployed during data collection, and discrete prediction outputs. Empirically we show that given our identifiability conditions hold, standard variants of supervised learning that predict from predictions by treating the prediction as an input feature can indeed find transferable functional relationships that allow for conclusions about newly deployed predictive models. These positive results fundamentally rely on model predictions being recorded during data collection, bringing forward the importance of rethinking standard data collection practices to enable progress towards a better understanding of social outcomes and performative feedback loops."
                },
                {
                    "title": "Human-AI Collaboration in Decision-Making: Beyond Learning to Defer",
                    "abstract": "Human-AI collaboration (HAIC) in decision-making aims to create synergistic teaming between human decision-makers and AI systems. Learning to defer (L2D) has been presented as a promising framework to determine who among humans and AI should make which decisions in order to optimize the performance and fairness of the combined system. Nevertheless, L2D entails several often unfeasible requirements, such as the availability of predictions from humans for every instance or ground-truth labels that are independent from said humans. Furthermore, neither L2D nor alternative approaches tackle fundamental issues of deploying HAIC systems in real-world settings, such as capacity management or dealing with dynamic environments. In this paper, we aim to identify and review these and other limitations, pointing to where opportunities for future research in HAIC may lie."
                },
                {
                    "title": "Preference Dynamics Under Personalized Recommendations",
                    "abstract": "The design of content recommendation systems underpins many online platforms: social media feeds, online news aggregators, and audio/video hosting websites all choose how best to organize an enormous amount of content for users to consume. Many projects (both practical and academic) have designed algorithms to match users to content they will enjoy under the assumption that user's preferences and opinions do not change with the content they see. However, increasing amounts of evidence suggest that individuals' preferences are directly shaped by what content they see---radicalization, rabbit holes, polarization, and boredom are all example phenomena of preferences affected by content. Polarization in particular can occur even in ecosystems with \"mass media,\" where no personalization takes place, as recently explored in a natural model of preference dynamics by [14] and [13]. If all users' preferences are drawn towards content they already like, or are repelled from content they already dislike, uniform consumption of media leads to a population of heterogeneous preferences converging towards only two poles. In this work, we explore whether some phenomenon akin to polarization occurs when users receive personalized content recommendations. We use a similar model of preference dynamics, where an individual's preferences move towards content the consume and enjoy, and away from content they consume and dislike. We show that standard user reward maximization is an almost trivial goal in such an environment (a large class of simple algorithms will achieve only constant regret). A more interesting objective, then, is to understand under what conditions a recommendation algorithm can ensure stationarity of user's preferences. We show how to design a content recommendations which can achieve approximate stationarity, under mild conditions on the set of available content, when a user's preferences are known, and how one can learn enough about a user's preferences to implement such a strategy even when user preferences are initially unknown."
                },
                {
                    "title": "A Taxonomy of Human and ML Strengths in Decision-Making to Investigate Human-ML Complementarity",
                    "abstract": "Hybrid human-ML systems increasingly make consequential decisions in a wide range of domains. These systems are often introduced with the expectation that the combined human-ML system will achieve complementary performance, that is, the combined decision-making system will be an improvement compared with either decision-making agent in isolation. However, empirical results have been mixed, and existing research rarely articulates the sources and mechanisms by which complementary performance is expected to arise. Our goal in this work is to provide conceptual tools to advance the way researchers reason and communicate about human-ML complementarity. Drawing upon prior literature in human psychology, machine learning, and human-computer interaction, we propose a taxonomy characterizing distinct ways in which human and ML-based decision-making can differ. In doing so, we conceptually map potential mechanisms by which combining human and ML decision-making may yield complementary performance, developing a language for the research community to reason about design of hybrid systems in any decision-making domain. To illustrate how our taxonomy can be used to investigate complementarity, we provide a mathematical aggregation framework to examine enabling\nconditions for complementarity. Through synthetic simulations, we demonstrate how this framework can be used to explore specific aspects of our taxonomy and shed light on the optimal mechanisms for combining human-ML judgments."
                },
                {
                    "title": "Performative Power",
                    "abstract": "We introduce the notion of performative power, which measures the ability of a firm operating an algorithmic system, such as a digital content recommendation platform, to cause change in a population of participants. We relate performative power to the economic study of competition in digital economies. Traditional economic concepts struggle with identifying anti-competitive patterns in digital platforms not least due to the complexity of market definition. In contrast, performative power is a causal notion that is identifiable with minimal knowledge of the market, its internals, participants, products, or prices. Low performative power implies that a firm can do no better than to optimize their objective on current data. In contrast, firms of high performative power stand to benefit from steering the population towards more profitable behavior. We confirm in a simple theoretical model that monopolies maximize performative power. A firm's ability to personalize increases performative power, while competition and outside options decrease performative power. On the empirical side, we propose an observational causal design to identify performative power from discontinuities in how digital platforms display content. This allows to repurpose causal effects from various studies about digital platforms as lower bounds on performative power. Finally, we speculate about the role that performative power might play in competition policy and antitrust enforcement in digital marketplaces."
                },
                {
                    "title": "Bayesian modeling of human\u2013AI complementarity",
                    "abstract": "Significance With the increase in artificial intelligence in real-world applications, there is interest in building hybrid systems that take both human and machine predictions into account. Previous work has shown the benefits of separately combining the predictions of diverse machine classifiers or groups of people. Using a Bayesian modeling framework, we extend these results by systematically investigating the factors that influence the performance of hybrid combinations of human and machine classifiers while taking into account the unique ways human and algorithmic confidence is expressed."
                },
                {
                    "title": "Human-Algorithm Collaboration: Achieving Complementarity and Avoiding Unfairness",
                    "abstract": "Much of machine learning research focuses on predictive accuracy: given a task, create a machine learning model (or algorithm) that maximizes accuracy. In many settings, however, the final prediction or decision of a system is under the control of a human, who uses an algorithm\u2019s output along with their own personal expertise in order to produce a combined prediction. One ultimate goal of such collaborative systems is complementarity: that is, to produce lower loss (equivalently, greater payoff or utility) than either the human or algorithm alone. However, experimental results have shown that even in carefully-designed systems, complementary performance can be elusive. Our work provides three key contributions. First, we provide a theoretical framework for modeling simple human-algorithm systems and demonstrate that multiple prior analyses can be expressed within it. Next, we use this model to prove conditions where complementarity is impossible, and give constructive examples of where complementarity is achievable. Finally, we discuss the implications of our findings, especially with respect to the fairness of a classifier. In sum, these results deepen our understanding of key factors influencing the combined performance of human-algorithm systems, giving insight into how algorithmic tools can best be designed for collaborative environments."
                },
                {
                    "title": "Designing Closed Human-in-the-loop Deferral Pipelines",
                    "abstract": "In hybrid human-machine deferral frameworks, a classifier can defer uncertain cases to human decision-makers (who are often themselves fallible). Prior work on simultaneous training of such classifier and deferral models has typically assumed access to an oracle during training to obtain true class labels for training samples, but in practice there often is no such oracle. In contrast, we consider a\"closed\"decision-making pipeline in which the same fallible human decision-makers used in deferral also provide training labels. How can imperfect and biased human expert labels be used to train a fair and accurate deferral framework? Our key insight is that by exploiting weak prior information, we can match experts to input examples to ensure fairness and accuracy of the resulting deferral framework, even when imperfect and biased experts are used in place of ground truth labels. The efficacy of our approach is shown both by theoretical analysis and by evaluation on two tasks."
                },
                {
                    "title": "Regret Minimization with Performative Feedback",
                    "abstract": "In performative prediction, the deployment of a predictive model triggers a shift in the data distribution. As these shifts are typically unknown ahead of time, the learner needs to deploy a model to get feedback about the distribution it induces. We study the problem of finding near-optimal models under performativity while maintaining low regret. On the surface, this problem might seem equivalent to a bandit problem. However, it exhibits a fundamentally richer feedback structure that we refer to as performative feedback: after every deployment, the learner receives samples from the shifted distribution rather than only bandit feedback about the reward. Our main contribution is an algorithm that achieves regret bounds scaling only with the complexity of the distribution shifts and not that of the reward function. The algorithm only relies on smoothness of the shifts and does not assume convexity. Moreover, its final iterate is guaranteed to be near-optimal. The key algorithmic idea is careful exploration of the distribution shifts that informs a novel construction of confidence bounds on the risk of unexplored models. More broadly, our work establishes a conceptual approach for leveraging tools from the bandits literature for the purpose of regret minimization with performative feedback."
                },
                {
                    "title": "To engage or not to engage with AI for critical judgments: How professionals deal with opacity when using AI for medical diagnosis",
                    "abstract": "Artificial intelligence (AI) technologies promise to transform how professionals conduct knowledge work by augmenting their capabilities for making professional judgments. We know little, however, about how human-AI augmentation takes place in practice. Yet, gaining this understanding is particularly important when professionals use AI tools to form judgments on critical decisions. We conducted an in-depth field study in a major U.S. hospital where AI tools were used in three departments by diagnostic radiologists making breast cancer, lung cancer, and bone age determinations. The study illustrates the hindering effects of opacity that professionals experienced when using AI tools and explores how these professionals grappled with it in practice. In all three departments, this opacity resulted in professionals experiencing increased uncertainty because AI tool results often diverged from their initial judgment without providing underlying reasoning. Only in one department (of the three) did professionals consistently incorporate AI results into their final judgments, achieving what we call engaged augmentation. These professionals invested in AI interrogation practices\u2014practices enacted by human experts to relate their own knowledge claims to AI knowledge claims. Professionals in the other two departments did not enact such practices and did not incorporate AI inputs into their final decisions, which we call unengaged \u201caugmentation.\u201d Our study unpacks the challenges involved in augmenting professional judgment with powerful, yet opaque, technologies and contributes to literature on AI adoption in knowledge work."
                },
                {
                    "title": "Omnipredictors",
                    "abstract": "Loss minimization is a dominant paradigm in machine learning, where a predictor is trained to minimize some loss function that depends on an uncertain event (e.g.,\"will it rain tomorrow?''). Different loss functions imply different learning algorithms and, at times, very different predictors. While widespread and appealing, a clear drawback of this approach is that the loss function may not be known at the time of learning, requiring the algorithm to use a best-guess loss function. We suggest a rigorous new paradigm for loss minimization in machine learning where the loss function can be ignored at the time of learning and only be taken into account when deciding an action. We introduce the notion of an (${\\mathcal{L}},\\mathcal{C}$)-omnipredictor, which could be used to optimize any loss in a family ${\\mathcal{L}}$. Once the loss function is set, the outputs of the predictor can be post-processed (a simple univariate data-independent transformation of individual predictions) to do well compared with any hypothesis from the class $\\mathcal{C}$. The post processing is essentially what one would perform if the outputs of the predictor were true probabilities of the uncertain events. In a sense, omnipredictors extract all the predictive power from the class $\\mathcal{C}$, irrespective of the loss function in $\\mathcal{L}$. We show that such\"loss-oblivious'' learning is feasible through a connection to multicalibration, a notion introduced in the context of algorithmic fairness. In addition, we show how multicalibration can be viewed as a solution concept for agnostic boosting, shedding new light on past results. Finally, we transfer our insights back to the context of algorithmic fairness by providing omnipredictors for multi-group loss minimization."
                },
                {
                    "title": "Differentiable Learning Under Triage",
                    "abstract": "Multiple lines of evidence suggest that predictive models may benefit from algorithmic triage. Under algorithmic triage, a predictive model does not predict all instances but instead defers some of them to human experts. However, the interplay between the prediction accuracy of the model and the human experts under algorithmic triage is not well understood. In this work, we start by formally characterizing under which circumstances a predictive model may benefit from algorithmic triage. In doing so, we also demonstrate that models trained for full automation may be suboptimal under triage. Then, given any model and desired level of triage, we show that the optimal triage policy is a deterministic threshold rule in which triage decisions are derived deterministically by thresholding the difference between the model and human errors on a per-instance level. Building upon these results, we introduce a practical gradient-based algorithm that is guaranteed to find a sequence of triage policies and predictive models of increasing performance. Experiments on a wide variety of supervised learning tasks using synthetic and real data from two important applications -- content moderation and scientific discovery -- illustrate our theoretical results and show that the models and triage policies provided by our gradient-based algorithm outperform those provided by several competitive baselines."
                },
                {
                    "title": "Towards Unbiased and Accurate Deferral to Multiple Experts",
                    "abstract": "Machine learning models are often implemented in cohort with humans in the pipeline, with the model having an option to defer to a domain expert in cases where it has low confidence in its inference. Our goal is to design mechanisms for ensuring accuracy and fairness in such prediction systems that combine machine learning model inferences and domain expert predictions. Prior work on \"deferral systems\" in classification settings has focused on the setting of a pipeline with a single expert and aimed to accommodate the inaccuracies and biases of this expert to simultaneously learn an inference model and a deferral system. Our work extends this framework to settings where multiple experts are available, with each expert having their own domain of expertise and biases. We propose a framework that simultaneously learns a classifier and a deferral system, with the deferral system choosing to defer to one or more human experts in cases of input where the classifier has low confidence. We test our framework on a synthetic dataset and a content moderation dataset with biased synthetic experts, and show that it significantly improves the accuracy and fairness of the final predictions, compared to the baselines. We also collect crowdsourced labels for the content moderation task to construct a real-world dataset for the evaluation of hybrid machine-human frameworks and show that our proposed framework outperforms baselines on this real-world dataset as well."
                },
                {
                    "title": "CheXternal: generalization of deep learning models for chest X-ray interpretation to photos of chest X-rays and external clinical settings",
                    "abstract": "Recent advances in training deep learning models have demonstrated the potential to provide accurate chest X-ray interpretation and increase access to radiology expertise. However, poor generalization due to data distribution shifts in clinical settings is a key barrier to implementation. In this study, we measured the diagnostic performance for 8 different chest X-ray models when applied to (1) smartphone photos of chest X-rays and (2) external datasets without any finetuning. All models were developed by different groups and submitted to the CheXpert challenge, and re-applied to test datasets without further tuning. We found that (1) on photos of chest X-rays, all 8 models experienced a statistically significant drop in task performance, but only 3 performed significantly worse than radiologists on average, and (2) on the external set, none of the models performed statistically significantly worse than radiologists, and five models performed statistically significantly better than radiologists. Our results demonstrate that some chest X-ray models, under clinically relevant distribution shifts, were comparable to radiologists while other models were not. Future work should investigate aspects of model training procedures and dataset collection that influence generalization in the presence of data distribution shifts."
                },
                {
                    "title": "Is the Most Accurate AI the Best Teammate? Optimizing AI for Teamwork",
                    "abstract": "AI practitioners typically strive to develop the most accurate systems, making an implicit assumption that the AI system will function autonomously. However, in practice, AI systems often are used to provide advice to people in domains ranging from criminal justice and finance to healthcare. In such AI-advised decision making, humans and machines form a team, where the human is responsible for making final decisions. But is the most accurate AI the best teammate? We argue \"not necessarily\" --- predictable performance may be worth a slight sacrifice in AI accuracy. Instead, we argue that AI systems should be trained in a human-centered manner, directly optimized for team performance. We study this proposal for a specific type of human-AI teaming, where the human overseer chooses to either accept the AI recommendation or solve the task themselves. To optimize the team performance for this setting we maximize the team's expected utility, expressed in terms of the quality of the final decision, cost of verifying, and individual accuracies of people and machines. Our experiments with linear and non-linear models on real-world, high-stakes datasets show that the most accuracy AI may not lead to highest team performance and show the benefit of modeling teamwork during training through improvements in expected team utility across datasets, considering parameters such as human skill and the cost of mistakes. We discuss the shortcoming of current optimization approaches beyond well-studied loss functions such as log-loss, and encourage future work on AI optimization problems motivated by human-AI collaboration."
                },
                {
                    "title": "Algorithmic monoculture and social welfare",
                    "abstract": "Significance Algorithmic monoculture is a growing concern in the use of algorithms for high-stakes screening decisions in areas such as employment and lending. If many firms use the same algorithm, even if it is more accurate than the alternatives, the resulting \u201cmonoculture\u201d may be susceptible to correlated failures, much as a monocultural system is in biological settings. To investigate this concern, we develop a model of selection under monoculture. We find that even without any assumption of shocks or correlated failures\u2014i.e., under \u201cnormal operations\u201d\u2014the quality of decisions may decrease when multiple firms use the same algorithm. Thus, the introduction of a more accurate algorithm may decrease social welfare\u2014a kind of \u201cBraess\u2019 paradox\u201d for algorithmic decision-making. As algorithms are increasingly applied to screen applicants for high-stakes decisions in employment, lending, and other domains, concerns have been raised about the effects of algorithmic monoculture, in which many decision-makers all rely on the same algorithm. This concern invokes analogies to agriculture, where a monocultural system runs the risk of severe harm from unexpected shocks. Here, we show that the dangers of algorithmic monoculture run much deeper, in that monocultural convergence on a single algorithm by a group of decision-making agents, even when the algorithm is more accurate for any one agent in isolation, can reduce the overall quality of the decisions being made by the full collection of agents. Unexpected shocks are therefore not needed to expose the risks of monoculture; it can hurt accuracy even under \u201cnormal\u201d operations and even for algorithms that are more accurate when used by only a single decision-maker. Our results rely on minimal assumptions and involve the development of a probabilistic framework for analyzing systems that use multiple noisy estimates of a set of alternatives."
                },
                {
                    "title": "Performative Prediction in a Stateful World",
                    "abstract": "Deployed supervised machine learning models make predictions that interact with and influence the world. This phenomenon is called \"performative prediction\" by Perdomo et al. (2020), who investigated it in a stateless setting. We generalize their results to the case where the response of the population to the deployed classifier depends both on the classifier and the previous distribution of the population. We also demonstrate such a setting empirically, for the scenario of strategic manipulation."
                },
                {
                    "title": "Does the Whole Exceed its Parts? The Effect of AI Explanations on Complementary Team Performance",
                    "abstract": "Many researchers motivate explainable AI with studies showing that human-AI team performance on decision-making tasks improves when the AI explains its recommendations. However, prior studies observed improvements from explanations only when the AI, alone, outperformed both the human and the best team. Can explanations help lead to complementary performance, where team accuracy is higher than either the human or the AI working solo? We conduct mixed-method user studies on three datasets, where an AI with accuracy comparable to humans helps participants solve a task (explaining itself in some conditions). While we observed complementary improvements from AI augmentation, they were not increased by explanations. Rather, explanations increased the chance that humans will accept the AI\u2019s recommendation, regardless of its correctness. Our result poses new challenges for human-centered AI: Can we develop explanatory approaches that encourage appropriate trust in AI, and therefore help generate (or improve) complementary performance?"
                },
                {
                    "title": "Classification Under Human Assistance",
                    "abstract": "Most supervised learning models are trained for full automation. However, their predictions are sometimes worse than those by human experts on some specific instances. Motivated by this empirical observation, our goal is to design classifiers that are optimized to operate under different automation levels. More specifically, we focus on convex margin-based classifiers and first show that the problem is NP-hard. Then, we further show that, for support vector machines, the corresponding objective function can be expressed as the difference of two functions f = g - c, where g is monotone, non-negative and gamma-weakly submodular, and c is non-negative and modular. This representation allows a recently introduced deterministic greedy algorithm, as well as a more efficient randomized variant of the algorithm, to enjoy approximation guarantees at solving the problem. Experiments on synthetic and real-world data from several applications in medical diagnosis illustrate our theoretical findings and demonstrate that, under human assistance, supervised learning models trained to operate under different automation levels can outperform those trained for full automation as well as humans operating alone."
                },
                {
                    "title": "Stochastic Optimization for Performative Prediction",
                    "abstract": "In performative prediction, the choice of a model influences the distribution of future data, typically through actions taken based on the model's predictions. \nWe initiate the study of stochastic optimization for performative prediction. What sets this setting apart from traditional stochastic optimization is the difference between merely updating model parameters and deploying the new model. The latter triggers a shift in the distribution that affects future data, while the former keeps the distribution as is. \nAssuming smoothness and strong convexity, we prove non-asymptotic rates of convergence for both greedily deploying models after each stochastic update (greedy deploy) as well as for taking several updates before redeploying (lazy deploy). In both cases, our bounds smoothly recover the optimal $O(1/k)$ rate as the strength of performativity decreases. Furthermore, they illustrate how depending on the strength of performative effects, there exists a regime where either approach outperforms the other. We experimentally explore this trade-off on both synthetic data and a strategic classification simulator."
                },
                {
                    "title": "Consistent Estimators for Learning to Defer to an Expert",
                    "abstract": "Learning algorithms are often used in conjunction with expert decision makers in practical scenarios, however this fact is largely ignored when designing these algorithms. In this paper we explore how to learn predictors that can either predict or choose to defer the decision to a downstream expert. Given only samples of the expert's decisions, we give a procedure based on learning a classifier and a rejector and analyze it theoretically. Our approach is based on a novel reduction to cost sensitive learning where we give a consistent surrogate loss for cost sensitive learning that generalizes the cross entropy loss. We show the effectiveness of our approach on a variety of experimental tasks."
                },
                {
                    "title": "Learning to Complement Humans",
                    "abstract": "A rising vision for AI in the open world centers on the development of systems that can complement humans for perceptual, diagnostic, and reasoning tasks. To date, systems aimed at complementing the skills of people have employed models trained to be as accurate as possible in isolation. We demonstrate how an end-to-end learning strategy can be harnessed to optimize the combined performance of human-machine teams by considering the distinct abilities of people and machines. The goal is to focus machine learning on problem instances that are difficult for humans, while recognizing instances that are difficult for the machine and seeking human input on them. We demonstrate in two real-world domains (scientific discovery and medical diagnosis) that human-machine teams built via these methods outperform the individual performance of machines and people. We then analyze conditions under which this complementarity is strongest, and which training methods amplify it. Taken together, our work provides the first systematic investigation of how machine learning systems can be trained to complement human reasoning."
                },
                {
                    "title": "A Human-Centered Evaluation of a Deep Learning System Deployed in Clinics for the Detection of Diabetic Retinopathy",
                    "abstract": "Deep learning algorithms promise to improve clinician workflows and patient outcomes. However, these gains have yet to be fully demonstrated in real world clinical settings. In this paper, we describe a human-centered study of a deep learning system used in clinics for the detection of diabetic eye disease. From interviews and observation across eleven clinics in Thailand, we characterize current eye-screening workflows, user expectations for an AI-assisted screening process, and post-deployment experiences. Our findings indicate that several socio-environmental factors impact model performance, nursing workflows, and the patient experience. We draw on these findings to reflect on the value of conducting human-centered evaluative research alongside prospective evaluations of model accuracy."
                },
                {
                    "title": "Measuring Racial Discrimination in Bail Decisions",
                    "abstract": "We develop new quasi-experimental tools to measure disparate impact, regardless of its source, in the context of bail decisions. We show that omitted variables bias in pretrial release rate comparisons can be purged by using the quasi-random assignment of judges to estimate average pretrial misconduct risk by race. We find that two-thirds of the release rate disparity between white and Black defendants in New York City is due to the disparate impact of release decisions. We then develop a hierarchical marginal treatment effect model to study the drivers of disparate impact, finding evidence of both racial bias and statistical discrimination."
                },
                {
                    "title": "A Case for Humans-in-the-Loop: Decisions in the Presence of Erroneous Algorithmic Scores",
                    "abstract": "The increased use of algorithmic predictions in sensitive domains has been accompanied by both enthusiasm and concern. To understand the opportunities and risks of these technologies, it is key to study how experts alter their decisions when using such tools. In this paper, we study the adoption of an algorithmic tool used to assist child maltreatment hotline screening decisions. We focus on the question: Are humans capable of identifying cases in which the machine is wrong, and of overriding those recommendations? We first show that humans do alter their behavior when the tool is deployed. Then, we show that humans are less likely to adhere to the machine's recommendation when the score displayed is an incorrect estimate of risk, even when overriding the recommendation requires supervisory approval. These results highlight the risks of full automation and the importance of designing decision pipelines that provide humans with autonomy."
                },
                {
                    "title": "Performative Prediction",
                    "abstract": "When predictions support decisions they may influence the outcome they aim to predict. We call such predictions performative; the prediction influences the target. Performativity is a well-studied phenomenon in policy-making that has so far been neglected in supervised learning. When ignored, performativity surfaces as undesirable distribution shift, routinely addressed with retraining. We develop a risk minimization framework for performative prediction bringing together concepts from statistics, game theory, and causality. A conceptual novelty is an equilibrium notion we call performative stability. Performative stability implies that the predictions are calibrated not against past outcomes, but against the future outcomes that manifest from acting on the prediction. Our main results are necessary and sufficient conditions for the convergence of retraining to a performatively stable point of nearly minimal loss. In full generality, performative prediction strictly subsumes the setting known as strategic classification. We thus also give the first sufficient conditions for retraining to overcome strategic feedback effects."
                },
                {
                    "title": "Algorithmic Social Engineering",
                    "abstract": "We examine the microeconomics of using algorithms to nudge decision-makers toward particular social outcomes. We refer to this as \u201calgorithmic social engineering.\u201d In this article, we apply classic strategic communication models to this strategy. Manipulating predictions to express policy preferences strips the predictions of informational content and can lead decision-makers to ignore them. When social problems stem from decision-makers' objectives (rather than their information sets), algorithmic social engineering exhibits clear limitations. Our framework emphasizes separating preferences and predictions in designing algorithmic interventions. This distinction has implications for software architecture, organizational structure, and regulation."
                },
                {
                    "title": "Diagnosing Physician Error: A Machine Learning Approach to Low-Value Health Care",
                    "abstract": "We use machine learning to better characterize low-value health care and the decisions that produce it. We focus on costly tests, specifically for heart attack (acute coronary syndromes). A test is only useful if it yields new information, so efficient testing is grounded in accurate prediction of test outcomes. Physician testing decisions can therefore be benchmarked against tailored algorithmic predictions, which provide a more precise way to study low-value care than the usual approach\u2014looking at average test yield. Implemented in a large national sample, this procedure reveals significant over-testing: 52.6% of high-cost tests for heart attack are wasted. At the same time, it also reveals significant under-testing: many patients with predictably high risk go untested, then experience frequent adverse cardiac events including death in the next 30 days. At standard clinical thresholds, these event rates suggest that testing these patients would indeed have been highly cost-effective. Of the potential welfare gains from more efficient testing, 42.8% would come from addressing under-use. Existing policy levers, however, appear too blunt a tool to address both over- and under-use inefficiencies. We find that they cut testing across the board, for low-risk (reducing over-use) and high-risk patients (exaggerating under-use). Finally, we uncover two behavioral mechanisms for physician testing errors: (i) bounded rationality, in which physicians use an overly narrow set of variables, but make effective use of that set; and (ii) representativeness, in which they over-weight how \u201crepresentative\u201d heart attack is for a patient, above and beyond the conditional probability. Together, these results suggest the need for models of low-value care that incorporate mis-prediction so as to account for both over- and under-testing."
                },
                {
                    "title": "The Algorithmic Automation Problem: Prediction, Triage, and Human Effort",
                    "abstract": "In a wide array of areas, algorithms are matching and surpassing the performance of human experts, leading to consideration of the roles of human judgment and algorithmic prediction in these domains. The discussion around these developments, however, has implicitly equated the specific task of prediction with the general task of automation. We argue here that automation is broader than just a comparison of human versus algorithmic performance on a task; it also involves the decision of which instances of the task to give to the algorithm in the first place. We develop a general framework that poses this latter decision as an optimization problem, and we show how basic heuristics for this optimization problem can lead to performance gains even on heavily-studied applications of AI in medicine. Our framework also serves to highlight how effective automation depends crucially on estimating both algorithmic and human error on an instance-by-instance basis, and our results show how improvements in these error estimation problems can yield significant gains for automation as well."
                },
                {
                    "title": "CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison",
                    "abstract": "Large, labeled datasets have driven deep learning methods to achieve expert-level performance on a variety of medical imaging tasks. We present CheXpert, a large dataset that contains 224,316 chest radiographs of 65,240 patients. We design a labeler to automatically detect the presence of 14 observations in radiology reports, capturing uncertainties inherent in radiograph interpretation. We investigate different approaches to using the uncertainty labels for training convolutional neural networks that output the probability of these observations given the available frontal and lateral radiographs. On a validation set of 200 chest radiographic studies which were manually annotated by 3 board-certified radiologists, we find that different uncertainty approaches are useful for different pathologies. We then evaluate our best model on a test set composed of 500 chest radiographic studies annotated by a consensus of 5 board-certified radiologists, and compare the performance of our model to that of 3 additional radiologists in the detection of 5 selected pathologies. On Cardiomegaly, Edema, and Pleural Effusion, the model ROC and PR curves lie above all 3 radiologist operating points. We release the dataset to the public as a standard benchmark to evaluate performance of chest radiograph interpretation models."
                },
                {
                    "title": "Multicalibration: Calibration for the (Computationally-Identifiable) Masses",
                    "abstract": "We develop and study multicalibration as a new measure of fairness in machine learning that aims to mitigate inadvertent or malicious discrimination that is introduced at training time (even from ground truth data). Multicalibration guarantees meaningful (calibrated) predictions for every sub-population that can be identi\ufb01ed within a speci\ufb01ed class of computations. The speci\ufb01ed class can be quite rich; in particular, it can contain many overlapping subgroups of a protected group. We demonstrate that in many settings this strong notion of protection from discrimination is provably attainable and aligned with the goal of accurate predictions. Along the way, we present algorithms for learning a multicalibrated predictor, study the computational complexity of this task, and illustrate tight connections to the agnostic learning model."
                },
                {
                    "title": "Exploring the Impact of Artificial Intelligence: Prediction Versus Judgment",
                    "abstract": "Based on recent developments in the field of artificial intelligence (AI), we examine what type of human labor will be a substitute versus a complement to emerging technologies. We argue that these recent developments reduce the costs of providing a particular set of tasks \u2013 prediction tasks. Prediction about uncertain states of the world is an input into decision-making. We show that prediction allows riskier decisions to be taken and this is its impact on observed productivity although it could also increase the variance of outcomes as well. We consider the role of human judgment in decision-making as prediction technology improves. Judgment is exercised when the objective function for a particular set of decisions cannot be described (i.e., coded). However, we demonstrate that better prediction impacts the returns to different types of judgment in opposite ways. Hence, not all human judgment will be a complement to AI. Finally, we show that humans will delegate some decisions to machines even when the decision would be superior with human input."
                },
                {
                    "title": "Predict Responsibly: Improving Fairness and Accuracy by Learning to Defer",
                    "abstract": "In many machine learning applications, there are multiple decision-makers involved, both automated and human. The interaction between these agents often goes unaddressed in algorithmic development. In this work, we explore a simple version of this interaction with a two-stage framework containing an automated model and an external decision-maker. The model can choose to say PASS, and pass the decision downstream, as explored in rejection learning. We extend this concept by proposing \"learning to defer\", which generalizes rejection learning by considering the effect of other agents in the decision-making process. We propose a learning algorithm which accounts for potential biases held by external decision-makers in a system. Experiments demonstrate that learning to defer can make systems not only more accurate but also less biased. Even when working with inconsistent or biased users, we show that deferring models still greatly improve the accuracy and/or fairness of the entire system."
                },
                {
                    "title": "The Selective Labels Problem: Evaluating Algorithmic Predictions in the Presence of Unobservables",
                    "abstract": "Evaluating whether machines improve on human performance is one of the central questions of machine learning. However, there are many domains where the data is selectively labeled, in the sense that the observed outcomes are themselves a consequence of the existing choices of the human decision-makers. For instance, in the context of judicial bail decisions, we observe the outcome of whether a defendant fails to return for their court appearance only if the human judge decides to release the defendant on bail. This selective labeling makes it harder to evaluate predictive models as the instances for which outcomes are observed do not represent a random sample of the population. Here we propose a novel framework for evaluating the performance of predictive models on selectively labeled data. We develop an approach called contraction which allows us to compare the performance of predictive models and human decision-makers without resorting to counterfactual inference. Our methodology harnesses the heterogeneity of human decision-makers and facilitates effective evaluation of predictive models even in the presence of unmeasured confounders (unobservables) which influence both human decisions and the resulting outcomes. Experimental results on real world datasets spanning diverse domains such as health care, insurance, and criminal justice demonstrate the utility of our evaluation metric in comparing human decisions and machine predictions."
                },
                {
                    "title": "Human Decisions and Machine Predictions",
                    "abstract": "Can machine learning improve human decision making? Bail decisions provide a good test case. Millions of times each year, judges make jail-or-release decisions that hinge on a prediction of what a defendant would do if released. The concreteness of the prediction task combined with the volume of data available makes this a promising machine-learning application. Yet comparing the algorithm to judges proves complicated. First, the available data are generated by prior judge decisions. We only observe crime outcomes for released defendants, not for those judges detained. This makes it hard to evaluate counterfactual decision rules based on algorithmic predictions. Second, judges may have a broader set of preferences than the variable the algorithm predicts; for instance, judges may care specifically about violent crimes or about racial inequities. We deal with these problems using different econometric strategies, such as quasi-random assignment of cases to judges. Even accounting for these concerns, our results suggest potentially large welfare gains: one policy simulation shows crime reductions up to 24.7% with no change in jailing rates, or jailing rate reductions up to 41.9% with no increase in crime rates. Moreover, all categories of crime, including violent crimes, show reductions; and these gains can be achieved while simultaneously reducing racial disparities. These results suggest that while machine learning can be valuable, realizing this value requires integrating these tools into an economic framework: being clear about the link between predictions and decisions; specifying the scope of payoff functions; and constructing unbiased decision counterfactuals. JEL Codes: C10 (Econometric and statistical methods and methodology), C55 (Large datasets: Modeling and analysis), K40 (Legal procedure, the legal system, and illegal behavior)."
                },
                {
                    "title": "Overcoming Algorithm Aversion: People Will Use Imperfect Algorithms If They Can (Even Slightly) Modify Them",
                    "abstract": "Although evidence-based algorithms consistently outperform human forecasters, people often fail to use them after learning that they are imperfect, a phenomenon known as algorithm aversion. In this paper, we present three studies investigating how to reduce algorithm aversion. In incentivized forecasting tasks, participants chose between using their own forecasts or those of an algorithm that was built by experts. Participants were considerably more likely to choose to use an imperfect algorithm when they could modify its forecasts, and they performed better as a result. Notably, the preference for modifiable algorithms held even when participants were severely restricted in the modifications they could make Studies 1-3. In fact, our results suggest that participants' preference for modifiable algorithms was indicative of a desire for some control over the forecasting outcome, and not for a desire for greater control over the forecasting outcome, as participants' preference for modifiable algorithms was relatively insensitive to the magnitude of the modifications they were able to make Study 2. Additionally, we found that giving participants the freedom to modify an imperfect algorithm made them feel more satisfied with the forecasting process, more likely to believe that the algorithm was superior, and more likely to choose to use an algorithm to make subsequent forecasts Study 3. This research suggests that one can reduce algorithm aversion by giving people some control-even a slight amount-over an imperfect algorithm's forecast. \n \nData, as supplemental material, are available at https://doi.org/10.1287/mnsc.2016.2643 . \n \nThis paper was accepted by Yuval Rottenstreich, judgment and decision making."
                },
                {
                    "title": "Prediction Policy Problems.",
                    "abstract": "Most empirical policy work focuses on causal inference. We argue an important class of policy problems does not require causal inference but instead requires predictive inference. Solving these \"prediction policy problems\" requires more than simple regression techniques, since these are tuned to generating unbiased estimates of coefficients rather than minimizing prediction error. We argue that new developments in the field of \"machine learning\" are particularly useful for addressing these prediction problems. We use an example from health policy to illustrate the large potential social welfare gains from improved prediction."
                },
                {
                    "title": "Mechanical versus clinical data combination in selection and admissions decisions: a meta-analysis.",
                    "abstract": "In employee selection and academic admission decisions, holistic (clinical) data combination methods continue to be relied upon and preferred by practitioners in our field. This meta-analysis examined and compared the relative predictive power of mechanical methods versus holistic methods in predicting multiple work (advancement, supervisory ratings of performance, and training performance) and academic (grade point average) criteria. There was consistent and substantial loss of validity when data were combined holistically-even by experts who are knowledgeable about the jobs and organizations in question-across multiple criteria in work and academic settings. In predicting job performance, the difference between the validity of mechanical and holistic data combination methods translated into an improvement in prediction of more than 50%. Implications for evidence-based practice are discussed."
                },
                {
                    "title": "Diagnosing Expertise: Human Capital, Decision Making, and Performance among Physicians",
                    "abstract": "Expert performance is often evaluated assuming that good experts have good outcomes. We examine expertise in medicine and develop a model that allows for two dimensions of physician performance: decision making and procedural skill. Better procedural skill increases the use of intensive procedures for everyone, while better decision making results in a reallocation of procedures from fewer low-risk to high-risk cases. We show that poor diagnosticians can be identified using administrative data and that improving decision making improves birth outcomes by reducing C-section rates at the bottom of the risk distribution and increasing them at the top of the distribution."
                },
                {
                    "title": "A Conceptual History of the Emergence of Bounded Rationality",
                    "abstract": "Whereas multiple interpretations of the concept of bounded rationality are currently enthusiastically employed by a wide range of academics, their colleagues of one hundred years ago did not have this concept available, but made use of related notions. This paper traces the history of the emergence of \u2018bounded rationality\u2019 from the first appearance of \u2018limited intelligence\u2019 in 1840 and \u2018finite intelligence\u2019 in 1880 through related concepts such as \u2018incomplete\u2019, \u2018limited\u2019, \u2018administrative\u2019, and \u2018approximate\u2019 rationality during the first half of the twentieth century to its inception in 1957 at the hands of Herbert Simon. In the process, it locates the semantic field from which the notion of bounded rationality emerged and attends to the co-emergence of terms such as \u2018limited rationality\u2019. This allows us to analyse the process of metaphorical crystallisation in which \u2018bounded rationality\u2019 assumed the dominant position within its semantic field."
                },
                {
                    "title": "The Well-Calibrated Bayesian",
                    "abstract": "Abstract Suppose that a forecaster sequentially assigns probabilities to events. He is well calibrated if, for example, of those events to which he assigns a probability 30 percent, the long-run proportion that actually occurs turns out to be 30 percent. We prove a theorem to the effect that a coherent Bayesian expects to be well calibrated, and consider its destructive implications for the theory of coherence."
                },
                {
                    "title": "Judgment under Uncertainty: Heuristics and Biases",
                    "abstract": "This article described three heuristics that are employed in making judgements under uncertainty: (i) representativeness, which is usually employed when people are asked to judge the probability that an object or event A belongs to class or process B; (ii) availability of instances or scenarios, which is often employed when people are asked to assess the frequency of a class or the plausibility of a particular development; and (iii) adjustment from an anchor, which is usually employed in numerical prediction when a relevant value is available. These heuristics are highly economical and usually effective, but they lead to systematic and predictable errors. A better understanding of these heuristics and of the biases to which they lead could improve judgements and decisions in situations of uncertainty."
                },
                {
                    "title": "Improving Human Decision-Making with Machine Learning",
                    "abstract": "A key aspect of human intelligence is their ability to convey their knowledge to others in succinct forms. However, current machine learning models are largely blackboxes that are hard for humans to learn from. We study the problem of whether we can design machine learning algorithms capable of conveying their insights to humans in the context of a sequential decision making task. In particular, we propose a novel machine learning algorithm for extracting inter-pretable tips from a policy trained to solve the task using reinforcement learning. In particular, it searches over a space of interpretable decision rules to identify the one that most improves human performance. Then, we perform an extensive user study to evaluate our approach, based on a virtual kitchen-management game we designed that requires the participant to make a series of decisions to minimize overall service time. Our experiments show that (i) the tips generated by our algorithm are effective at improving performance, (ii) they signi\ufb01cantly outperform the two baseline tips, and (iii) they successfully help participants build on their own experience to discover additional strategies and overcome their resistance to exploring counterintuitive strategies."
                },
                {
                    "title": "Bias and Productivity in Humans and Algorithms: Theory and Evidence from Re\u0301sume\u0301 Screening",
                    "abstract": "Where should better learning technology improve decisions? I develop a formal model of decision-making in which better learning technology is complementary with experimentation. Noisy, inconsistent decision-making by humans introduces quasi-experimental variation into training datasets, which complements learning. The model makes heterogeneous predictions about when machine learning algorithms can improve human biases. These algorithms will can remove human biases exhibited in historical training data, but only if the human training decisions are sufficiently noisy; otherwise the algorithms will codify or exacerbate existing biases. I then test these predictions in a field experiment hiring workers for white-collar jobs. The introduction of machine learning technology yields candidates that are a) +14% more likely to pass interviews and receive a job offer, b) +18% more likely to accept job offers when extended, and c) 0.2\u03c3-0.4\u03c3 more productive once hired as employees. They are also 12% less likely to show evidence of competing job offers during salary negotiations. These results were driven by candidates who were evaluated in a noisy, biased way in historical data used for training. These candidates are broadly non-traditional, particularly candidates who graduated from non-elite colleges, who lack job referrals, who lack prior experience, whose credentials are atypical and who have strong non-cognitive soft-skills."
                },
                {
                    "title": "Clinical versus mechanical prediction: a meta-analysis.",
                    "abstract": "The process of making judgments and decisions requires a method for combining data. To compare the accuracy of clinical and mechanical (formal, statistical) data-combination techniques, we performed a meta-analysis on studies of human health and behavior. On average, mechanical-prediction techniques were about 10% more accurate than clinical predictions. Depending on the specific analysis, mechanical prediction substantially outperformed clinical prediction in 33%-47% of studies examined. Although clinical predictions were often as accurate as mechanical predictions, in only a few studies (6%-16%) were they substantially more accurate. Superiority for mechanical-prediction techniques was consistent, regardless of the judgment task, type of judges, judges' amounts of experience, or the types of data being combined. Clinical predictions performed relatively less well when predictors included clinical interview data. These data indicate that mechanical predictions of human behaviors are equal or superior to clinical prediction methods for a wide range of circumstances."
                },
                {
                    "title": "The process-performance paradox in expert judgment - How can experts know so much and predict so badly?",
                    "abstract": "A mysterious fatal disease strikes a large minority of the population. \nThe disease is incurable, but an expensive drug can keep victims alive. Congress decides that the drug should be given to those whose lives can be \nextended longest, which only a few specialists can predict. The experts work \naround the clock searching for a cure; allocating the drug is a new chore they \nwould rather avoid."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI",
                "cs.HC"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nWhen (and how) can human judgment improve the predictions of any learning algorithm in high-stakes prediction tasks, such as emergency room triage and medical imaging diagnosis?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it bridges the gap between human expertise and algorithmic efficiency, potentially leading to improved decision-making in critical healthcare settings. By understanding how to effectively integrate human judgment with machine learning models, future research can explore new methodologies for human-AI collaboration, enhancing the accuracy and reliability of predictions. This advancement could lead to practical applications in various domains, particularly in healthcare, where better triage decisions can save lives and optimize resource allocation.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenges in addressing this problem include the complexity of human judgment, which is often nuanced and context-dependent, making it difficult to quantify and integrate into algorithmic predictions. Naive approaches may fail because they might not account for the rich, qualitative information that healthcare providers possess, which is not captured in structured data. Additionally, technical obstacles such as the need for robust statistical methods to evaluate the effectiveness of human input, and the theoretical challenge of defining and measuring \"algorithmic indistinguishability\" in high-dimensional data, complicate the integration of human insights into machine learning models.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has often focused on either improving algorithmic performance or understanding human decision-making in isolation, without exploring the synergistic potential of combining both. Limitations in existing solutions include a lack of frameworks for assessing when human input is beneficial and how to effectively incorporate it into algorithms. Barriers such as the difficulty in quantifying human judgment and the absence of methodologies for systematic integration have prevented this problem from being solved. Our approach differs by proposing a statistical framework to evaluate the contribution of human insights, particularly in cases where algorithmic predictions are already strong.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves conducting experiments where physicians assess patients with identical or similar chest X-rays to determine if they can distinguish between cases of atelectasis and non-atelectasis. We will use a dataset of chest X-rays and corresponding physician assessments, employing metrics such as accuracy and statistical significance to evaluate the effectiveness of human input. The expected outcome is to demonstrate that human judgment can"
            }
        },
        "author_data": {
            "404469e2-8e89-4e23-a97e-f99191880027": {
                "pk": "404469e2-8e89-4e23-a97e-f99191880027",
                "name": "Rohan Alur",
                "collaborators": [
                    "Manish Raghavan",
                    "Loren Laine",
                    "Darrick K. Li",
                    "Dennis Shung",
                    "Devavrat Shah",
                    "Sarah H. Cen",
                    "Vinith M. Suriyakumar",
                    "Ayush Sekhari",
                    "Ashia C. Wilson"
                ],
                "domain": [
                    "AI Auditing",
                    "Human-AI Collaboration",
                    "Machine Learning",
                    "Model Safety"
                ],
                "publications": [
                    {
                        "title": "From Transparency to Accountability and Back: A Discussion of Access and Evidence in AI Auditing",
                        "abstract": "Artificial intelligence (AI) is increasingly intervening in our lives, raising widespread concern about its unintended and undeclared side effects. These developments have brought attention to the problem of AI auditing: the systematic evaluation and analysis of an AI system, its development, and its behavior relative to a set of predetermined criteria. Auditing can take many forms, including pre-deployment risk assessments, ongoing monitoring, and compliance testing. It plays a critical role in providing assurances to various AI stakeholders, from developers to end users. Audits may, for instance, be used to verify that an algorithm complies with the law, is consistent with industry standards, and meets the developer's claimed specifications. However, there are many operational challenges to AI auditing that complicate its implementation.   In this work, we examine a key operational issue in AI auditing: what type of access to an AI system is needed to perform a meaningful audit? Addressing this question has direct policy relevance, as it can inform AI audit guidelines and requirements. We begin by discussing the factors that auditors balance when determining the appropriate type of access, and unpack the benefits and drawbacks of four types of access. We conclude that, at minimum, black-box access -- providing query access to a model without exposing its internal implementation -- should be granted to auditors, as it balances concerns related to trade secrets, data privacy, audit standardization, and audit efficiency. We then suggest a framework for determining how much further access (in addition to black-box access) to grant auditors. We show that auditing can be cast as a natural hypothesis test, draw parallels hypothesis testing and legal procedure, and argue that this framing provides clear and interpretable guidance on audit implementation."
                    },
                    {
                        "title": "Integrating Expert Judgment and Algorithmic Decision Making: An Indistinguishability Framework",
                        "abstract": "We introduce a novel framework for human-AI collaboration in prediction and decision tasks. Our approach leverages human judgment to distinguish inputs which are algorithmically indistinguishable, or \"look the same\" to any feasible predictive algorithm. We argue that this framing clarifies the problem of human-AI collaboration in prediction and decision tasks, as experts often form judgments by drawing on information which is not encoded in an algorithm's training data. Algorithmic indistinguishability yields a natural test for assessing whether experts incorporate this kind of \"side information\", and further provides a simple but principled method for selectively incorporating human feedback into algorithmic predictions. We show that this method provably improves the performance of any feasible algorithmic predictor and precisely quantify this improvement. We demonstrate the utility of our framework in a case study of emergency room triage decisions, where we find that although algorithmic risk scores are highly competitive with physicians, there is strong evidence that physician judgments provide signal which could not be replicated by any predictive algorithm. This insight yields a range of natural decision rules which leverage the complementary strengths of human experts and predictive algorithms."
                    },
                    {
                        "title": "Auditing for Human Expertise",
                        "abstract": "High-stakes prediction tasks (e.g., patient diagnosis) are often handled by trained human experts. A common source of concern about automation in these settings is that experts may exercise intuition that is difficult to model and/or have access to information (e.g., conversations with a patient) that is simply unavailable to a would-be algorithm. This raises a natural question whether human experts add value which could not be captured by an algorithmic predictor. We develop a statistical framework under which we can pose this question as a natural hypothesis test. Indeed, as our framework highlights, detecting human expertise is more subtle than simply comparing the accuracy of expert predictions to those made by a particular learning algorithm. Instead, we propose a simple procedure which tests whether expert predictions are statistically independent from the outcomes of interest after conditioning on the available inputs (`features'). A rejection of our test thus suggests that human experts may add value to any algorithm trained on the available data, and has direct implications for whether human-AI `complementarity' is achievable in a given prediction task. We highlight the utility of our procedure using admissions data collected from the emergency department of a large academic hospital system, where we show that physicians' admit/discharge decisions for patients with acute gastrointestinal bleeding (AGIB) appear to be incorporating information that is not available to a standard algorithmic screening tool. This is despite the fact that the screening tool is arguably more accurate than physicians' discretionary decisions, highlighting that -- even absent normative concerns about accountability or interpretability -- accuracy is insufficient to justify algorithmic automation."
                    },
                    {
                        "title": "Unstable Unlearning: The Hidden Risk of Concept Resurgence in Diffusion Models",
                        "abstract": "Text-to-image diffusion models rely on massive, web-scale datasets. Training them from scratch is computationally expensive, and as a result, developers often prefer to make incremental updates to existing models. These updates often compose fine-tuning steps (to learn new concepts or improve model performance) with \"unlearning\" steps (to \"forget\" existing concepts, such as copyrighted works or explicit content). In this work, we demonstrate a critical and previously unknown vulnerability that arises in this paradigm: even under benign, non-adversarial conditions, fine-tuning a text-to-image diffusion model on seemingly unrelated images can cause it to \"relearn\" concepts that were previously \"unlearned.\" We comprehensively investigate the causes and scope of this phenomenon, which we term concept resurgence, by performing a series of experiments which compose \"mass concept erasure\" (the current state of the art for unlearning in text-to-image diffusion models (Lu et al., 2024)) with subsequent fine-tuning of Stable Diffusion v1.4. Our findings underscore the fragility of composing incremental model updates, and raise serious new concerns about current approaches to ensuring the safety and alignment of text-to-image diffusion models."
                    }
                ]
            },
            "01c9a023-b6b8-48d2-b226-91c8c32cafec": {
                "pk": "01c9a023-b6b8-48d2-b226-91c8c32cafec",
                "name": "Manish Raghavan",
                "collaborators": [
                    "Jon Kleinberg",
                    "Solon Barocas",
                    "Cynthia Dwork",
                    "Sigal Oren",
                    "Karen Levy",
                    "Sendhil Mullainathan",
                    "Chris Hays",
                    "Manish Purohit",
                    "Sreenivas Gollupadi",
                    "Ashton Anderson"
                ],
                "domain": [
                    "Algorithmic Fairness",
                    "Decision Theory",
                    "Social Networks",
                    "Behavioral Economics"
                ],
                "publications": [
                    {
                        "title": "Hiring Under Uncertainty",
                        "abstract": "In this paper we introduce the hiring under uncertainty problem to model the questions faced by hiring committees in large enterprises and universities alike. Given a set of $n$ eligible candidates, the decision maker needs to choose the sequence of candidates to make offers so as to hire the $k$ best candidates. However, candidates may choose to reject an offer (for instance, due to a competing offer) and the decision maker has a time limit by which all positions must be filled. Given an estimate of the probabilities of acceptance for each candidate, the hiring under uncertainty problem is to design a strategy of making offers so that the total expected value of all candidates hired by the time limit is maximized. We provide a 2-approximation algorithm for the setting where offers must be made in sequence, an 8-approximation when offers may be made in parallel, and a 10-approximation for the more general stochastic knapsack setting with finite probes."
                    },
                    {
                        "title": "How Do Classifiers Induce Agents To Invest Effort Strategically?",
                        "abstract": "Algorithms are often used to produce decision-making rules that classify or evaluate individuals. When these individuals have incentives to be classified a certain way, they may behave strategically to influence their outcomes. We develop a model for how strategic agents can invest effort in order to change the outcomes they receive, and we give a tight characterization of when such agents can be incentivized to invest specified forms of effort into improving their outcomes as opposed to \"gaming\" the classifier. We show that whenever any \"reasonable\" mechanism can do so, a simple linear mechanism suffices."
                    },
                    {
                        "title": "Algorithmic Monoculture and Social Welfare",
                        "abstract": "As algorithms are increasingly applied to screen applicants for high-stakes decisions in employment, lending, and other domains, concerns have been raised about the effects of algorithmic monoculture, in which many decision-makers all rely on the same algorithm. This concern invokes analogies to agriculture, where a monocultural system runs the risk of severe harm from unexpected shocks. Here we show that the dangers of algorithmic monoculture run much deeper, in that monocultural convergence on a single algorithm by a group of decision-making agents, even when the algorithm is more accurate for any one agent in isolation, can reduce the overall quality of the decisions being made by the full collection of agents. Unexpected shocks are therefore not needed to expose the risks of monoculture; it can hurt accuracy even under \"normal\" operations, and even for algorithms that are more accurate when used by only a single decision-maker. Our results rely on minimal assumptions, and involve the development of a probabilistic framework for analyzing systems that use multiple noisy estimates of a set of alternatives."
                    },
                    {
                        "title": "Selection Problems in the Presence of Implicit Bias",
                        "abstract": "Over the past two decades, the notion of implicit bias has come to serve as an important component in our understanding of discrimination in activities such as hiring, promotion, and school admissions. Research on implicit bias posits that when people evaluate others -- for example, in a hiring context -- their unconscious biases about membership in particular groups can have an effect on their decision-making, even when they have no deliberate intention to discriminate against members of these groups. A growing body of experimental work has pointed to the effect that implicit bias can have in producing adverse outcomes.   Here we propose a theoretical model for studying the effects of implicit bias on selection decisions, and a way of analyzing possible procedural remedies for implicit bias within this model. A canonical situation represented by our model is a hiring setting: a recruiting committee is trying to choose a set of finalists to interview among the applicants for a job, evaluating these applicants based on their future potential, but their estimates of potential are skewed by implicit bias against members of one group. In this model, we show that measures such as the Rooney Rule, a requirement that at least one of the finalists be chosen from the affected group, can not only improve the representation of this affected group, but also lead to higher payoffs in absolute terms for the organization performing the recruiting. However, identifying the conditions under which such measures can lead to improved payoffs involves subtle trade-offs between the extent of the bias and the underlying distribution of applicant characteristics, leading to novel theoretical questions about order statistics in the presence of probabilistic side information."
                    },
                    {
                        "title": "Mitigating Bias in Algorithmic Hiring: Evaluating Claims and Practices",
                        "abstract": "There has been rapidly growing interest in the use of algorithms in hiring, especially as a means to address or mitigate bias. Yet, to date, little is known about how these methods are used in practice. How are algorithmic assessments built, validated, and examined for bias? In this work, we document and analyze the claims and practices of companies offering algorithms for employment assessment. In particular, we identify vendors of algorithmic pre-employment assessments (i.e., algorithms to screen candidates), document what they have disclosed about their development and validation procedures, and evaluate their practices, focusing particularly on efforts to detect and mitigate bias. Our analysis considers both technical and legal perspectives. Technically, we consider the various choices vendors make regarding data collection and prediction targets, and explore the risks and trade-offs that these choices pose. We also discuss how algorithmic de-biasing techniques interface with, and create challenges for, antidiscrimination law."
                    },
                    {
                        "title": "Mapping the Invocation Structure of Online Political Interaction",
                        "abstract": "The surge in political information, discourse, and interaction has been one of the most important developments in social media over the past several years. There is rich structure in the interaction among different viewpoints on the ideological spectrum. However, we still have only a limited analytical vocabulary for expressing the ways in which these viewpoints interact.   In this paper, we develop network-based methods that operate on the ways in which users share content; we construct \\emph{invocation graphs} on Web domains showing the extent to which pages from one domain are invoked by users to reply to posts containing pages from other domains. When we locate the domains on a political spectrum induced from the data, we obtain an embedded graph showing how these interaction links span different distances on the spectrum. The structure of this embedded network, and its evolution over time, helps us derive macro-level insights about how political interaction unfolded through 2016, leading up to the US Presidential election. In particular, we find that the domains invoked in replies spanned increasing distances on the spectrum over the months approaching the election, and that there was clear asymmetry between the left-to-right and right-to-left patterns of linkage."
                    },
                    {
                        "title": "The Hidden Assumptions Behind Counterfactual Explanations and Principal Reasons",
                        "abstract": "Counterfactual explanations are gaining prominence within technical, legal, and business circles as a way to explain the decisions of a machine learning model. These explanations share a trait with the long-established \"principal reason\" explanations required by U.S. credit laws: they both explain a decision by highlighting a set of features deemed most relevant--and withholding others.   These \"feature-highlighting explanations\" have several desirable properties: They place no constraints on model complexity, do not require model disclosure, detail what needed to be different to achieve a different decision, and seem to automate compliance with the law. But they are far more complex and subjective than they appear.   In this paper, we demonstrate that the utility of feature-highlighting explanations relies on a number of easily overlooked assumptions: that the recommended change in feature values clearly maps to real-world actions, that features can be made commensurate by looking only at the distribution of the training data, that features are only relevant to the decision at hand, and that the underlying model is stable over time, monotonic, and limited to binary outcomes.   We then explore several consequences of acknowledging and attempting to address these assumptions, including a paradox in the way that feature-highlighting explanations aim to respect autonomy, the unchecked power that feature-highlighting explanations grant decision makers, and a tension between making these explanations useful and the need to keep the model hidden.   While new research suggests several ways that feature-highlighting explanations can work around some of the problems that we identify, the disconnect between features in the model and actions in the real world--and the subjective choices necessary to compensate for this--must be understood before these techniques can be usefully implemented."
                    },
                    {
                        "title": "Synthetic Census Data Generation via Multidimensional Multiset Sum",
                        "abstract": "The US Decennial Census provides valuable data for both research and policy purposes. Census data are subject to a variety of disclosure avoidance techniques prior to release in order to preserve respondent confidentiality. While many are interested in studying the impacts of disclosure avoidance methods on downstream analyses, particularly with the introduction of differential privacy in the 2020 Decennial Census, these efforts are limited by a critical lack of data: The underlying \"microdata,\" which serve as necessary input to disclosure avoidance methods, are kept confidential.   In this work, we aim to address this limitation by providing tools to generate synthetic microdata solely from published Census statistics, which can then be used as input to any number of disclosure avoidance algorithms for the sake of evaluation and carrying out comparisons. We define a principled distribution over microdata given published Census statistics and design algorithms to sample from this distribution. We formulate synthetic data generation in this context as a knapsack-style combinatorial optimization problem and develop novel algorithms for this setting. While the problem we study is provably hard, we show empirically that our methods work well in practice, and we offer theoretical arguments to explain our performance. Finally, we verify that the data we produce are \"close\" to the desired ground truth."
                    },
                    {
                        "title": "Greedy Algorithm almost Dominates in Smoothed Contextual Bandits",
                        "abstract": "Online learning algorithms, widely used to power search and content optimization on the web, must balance exploration and exploitation, potentially sacrificing the experience of current users in order to gain information that will lead to better decisions in the future. While necessary in the worst case, explicit exploration has a number of disadvantages compared to the greedy algorithm that always \"exploits\" by choosing an action that currently looks optimal. We ask under what conditions inherent diversity in the data makes explicit exploration unnecessary. We build on a recent line of work on the smoothed analysis of the greedy algorithm in the linear contextual bandits model. We improve on prior results to show that a greedy approach almost matches the best possible Bayesian regret rate of any other algorithm on the same problem instance whenever the diversity conditions hold, and that this regret is at most $\\tilde O(T^{1/3})$."
                    },
                    {
                        "title": "The Challenge of Understanding What Users Want: Inconsistent Preferences and Engagement Optimization",
                        "abstract": "Online platforms have a wealth of data, run countless experiments and use industrial-scale algorithms to optimize user experience. Despite this, many users seem to regret the time they spend on these platforms. One possible explanation is misaligned incentives: platforms are not optimizing for user happiness. We suggest the problem runs deeper, transcending the specific incentives of any particular platform, and instead stems from a mistaken foundational assumption: To understand what users want, platforms look at what users do. Yet research has demonstrated, and personal experience affirms, that we often make choices in the moment that are inconsistent with what we actually want. In this work, we develop a model of media consumption where users have inconsistent preferences. We consider a platform which simply wants to maximize user utility, but only observes user engagement. We show how our model of users' preference inconsistencies produces phenomena that are familiar from everyday experience, but difficult to capture in traditional user interaction models. A key ingredient in our model is a formulation for how platforms determine what to show users: they optimize over a large set of potential content (the content manifold) parametrized by underlying features of the content. Whether improving engagement improves user welfare depends on the direction of movement in the content manifold: for certain directions of change, increasing engagement makes users less happy, while in other directions, increasing engagement makes users happier. We characterize the structure of content manifolds for which increasing engagement fails to increase user utility. By linking these effects to abstractions of platform design choices, our model thus creates a theoretical framework and vocabulary in which to explore interactions between design, behavioral science, and social media."
                    },
                    {
                        "title": "Inherent Trade-Offs in the Fair Determination of Risk Scores",
                        "abstract": "Recent discussion in the public sphere about algorithmic classification has involved tension between competing notions of what it means for a probabilistic classification to be fair to different groups. We formalize three fairness conditions that lie at the heart of these debates, and we prove that except in highly constrained special cases, there is no method that can satisfy these three conditions simultaneously. Moreover, even satisfying all three conditions approximately requires that the data lie in an approximate version of one of the constrained special cases identified by our theorem. These results suggest some of the ways in which key notions of fairness are incompatible with each other, and hence provide a framework for thinking about the trade-offs between them."
                    },
                    {
                        "title": "Content Moderation and the Formation of Online Communities: A Theoretical Framework",
                        "abstract": "We study the impact of content moderation policies in online communities. In our theoretical model, a platform chooses a content moderation policy and individuals choose whether or not to participate in the community according to the fraction of user content that aligns with their preferences. The effects of content moderation, at first blush, might seem obvious: it restricts speech on a platform. However, when user participation decisions are taken into account, its effects can be more subtle $\\unicode{x2013}$ and counter-intuitive. For example, our model can straightforwardly demonstrate how moderation policies may increase participation and diversify content available on the platform. In our analysis, we explore a rich set of interconnected phenomena related to content moderation in online communities. We first characterize the effectiveness of a natural class of moderation policies for creating and sustaining stable communities. Building on this, we explore how resource-limited or ideological platforms might set policies, how communities are affected by differing levels of personalization, and competition between platforms. Our model provides a vocabulary and mathematically tractable framework for analyzing platform decisions about content moderation."
                    },
                    {
                        "title": "Planning with Multiple Biases",
                        "abstract": "Recent work has considered theoretical models for the behavior of agents with specific behavioral biases: rather than making decisions that optimize a given payoff function, the agent behaves inefficiently because its decisions suffer from an underlying bias. These approaches have generally considered an agent who experiences a single behavioral bias, studying the effect of this bias on the outcome.   In general, however, decision-making can and will be affected by multiple biases operating at the same time. How do multiple biases interact to produce the overall outcome? Here we consider decisions in the presence of a pair of biases exhibiting an intuitively natural interaction: present bias -- the tendency to value costs incurred in the present too highly -- and sunk-cost bias -- the tendency to incorporate costs experienced in the past into one's plans for the future.   We propose a theoretical model for planning with this pair of biases, and we show how certain natural behavioral phenomena can arise in our model only when agents exhibit both biases. As part of our model we differentiate between agents that are aware of their biases (sophisticated) and agents that are unaware of them (naive). Interestingly, we show that the interaction between the two biases is quite complex: in some cases, they mitigate each other's effects while in other cases they might amplify each other. We obtain a number of further results as well, including the fact that the planning problem in our model for an agent experiencing and aware of both biases is computationally hard in general, though tractable under more relaxed assumptions."
                    },
                    {
                        "title": "Stochastic Model for Sunk Cost Bias",
                        "abstract": "We present a novel model for capturing the behavior of an agent exhibiting sunk-cost bias in a stochastic environment. Agents exhibiting sunk-cost bias take into account the effort they have already spent on an endeavor when they evaluate whether to continue or abandon it. We model planning tasks in which an agent with this type of bias tries to reach a designated goal. Our model structures this problem as a type of Markov decision process: loosely speaking, the agent traverses a directed acyclic graph with probabilistic transitions, paying costs for its actions as it tries to reach a target node containing a specified reward. The agent's sunk cost bias is modeled by a cost that it incurs for abandoning the traversal: if the agent decides to stop traversing the graph, it incurs a cost of $\\lambda \\cdot C_{sunk}$, where ${\\lambda \\geq 0}$ is a parameter that captures the extent of the bias and $C_{sunk}$ is the sum of costs already invested.   We analyze the behavior of two types of agents: naive agents that are unaware of their bias, and sophisticated agents that are aware of it. Since optimal (bias-free) behavior in this problem can involve abandoning the traversal before reaching the goal, the bias exhibited by these types of agents can result in sub-optimal behavior by shifting their decisions about abandonment. We show that in contrast to optimal agents, it is computationally hard to compute the optimal policy for a sophisticated agent. Our main results quantify the loss exhibited by these two types of agents with respect to an optimal agent. We present both general and topology-specific bounds."
                    },
                    {
                        "title": "Planning Problems for Sophisticated Agents with Present Bias",
                        "abstract": "Present bias, the tendency to weigh costs and benefits incurred in the present too heavily, is one of the most widespread human behavioral biases. It has also been the subject of extensive study in the behavioral economics literature. While the simplest models assume that the agents are naive, reasoning about the future without taking their bias into account, there is considerable evidence that people often behave in ways that are sophisticated with respect to present bias, making plans based on the belief that they will be present-biased in the future. For example, committing to a course of action to reduce future opportunities for procrastination or overconsumption are instances of sophisticated behavior in everyday life.   Models of sophisticated behavior have lacked an underlying formalism that allows one to reason over the full space of multi-step tasks that a sophisticated agent might face. This has made it correspondingly difficult to make comparative or worst-case statements about the performance of sophisticated agents in arbitrary scenarios. In this paper, we incorporate the notion of sophistication into a graph-theoretic model that we used in recent work for modeling naive agents. This new synthesis of two formalisms - sophistication and graph-theoretic planning - uncovers a rich structure that wasn't apparent in the earlier behavioral economics work on this problem.   In particular, our graph-theoretic model makes two kinds of new results possible. First, we give tight worst-case bounds on the performance of sophisticated agents in arbitrary multi-step tasks relative to the optimal plan. Second, the flexibility of our formalism makes it possible to identify new phenomena that had not been seen in prior literature: these include a surprising non-monotonic property in the use of rewards to motivate sophisticated agents and a framework for reasoning about commitment devices."
                    },
                    {
                        "title": "The Right to be an Exception to a Data-Driven Rule",
                        "abstract": "Data-driven tools are increasingly used to make consequential decisions. They have begun to advise employers on which job applicants to interview, judges on which defendants to grant bail, lenders on which homeowners to give loans, and more. In such settings, different data-driven rules result in different decisions. The problem is: to every data-driven rule, there are exceptions. While a data-driven rule may be appropriate for some, it may not be appropriate for all. As data-driven decisions become more common, there are cases in which it becomes necessary to protect the individuals who, through no fault of their own, are the data-driven exceptions. At the same time, it is impossible to scrutinize every one of the increasing number of data-driven decisions, begging the question: When and how should data-driven exceptions be protected?   In this piece, we argue that individuals have the right to be an exception to a data-driven rule. That is, the presumption should not be that a data-driven rule--even one with high accuracy--is suitable for an arbitrary decision-subject of interest. Rather, a decision-maker should apply the rule only if they have exercised due care and due diligence (relative to the risk of harm) in excluding the possibility that the decision-subject is an exception to the data-driven rule. In some cases, the risk of harm may be so low that only cursory consideration is required. Although applying due care and due diligence is meaningful in human-driven decision contexts, it is unclear what it means for a data-driven rule to do so. We propose that determining whether a data-driven rule is suitable for a given decision-subject requires the consideration of three factors: individualization, uncertainty, and harm. We unpack this right in detail, providing a framework for assessing data-driven rules and describing what it would mean to invoke the right in practice."
                    },
                    {
                        "title": "Equilibria, Efficiency, and Inequality in Network Formation for Hiring and Opportunity",
                        "abstract": "Professional networks -- the social networks among people in a given line of work -- can serve as a conduit for job prospects and other opportunities. Here we propose a model for the formation of such networks and the transfer of opportunities within them. In our theoretical model, individuals strategically connect with others to maximize the probability that they receive opportunities from them. We explore how professional networks balance connectivity, where connections facilitate opportunity transfers to those who did not get them from outside sources, and congestion, where some individuals receive too many opportunities from their connections and waste some of them.   We show that strategic individuals are over-connected at equilibrium relative to a social optimum, leading to a price of anarchy for which we derive nearly tight asymptotic bounds. We also show that, at equilibrium, individuals form connections to those who provide similar benefit to them as they provide to others. Thus, our model provides a microfoundation in professional networking contexts for the fundamental sociological principle of homophily, that \"similarity breeds connection,\" which in our setting is realized as a form of status homophily based on alignment in individual benefit. We further explore how, even if individuals are a priori equally likely to receive opportunities from outside sources, equilibria can be unequal, and we provide nearly tight bounds on how unequal they can be. Finally, we explore the ability for online platforms to intervene to improve social welfare and show that natural heuristics may result in adverse effects at equilibrium. Our simple model allows for a surprisingly rich analysis of coordination problems in professional networks and suggests many directions for further exploration."
                    },
                    {
                        "title": "Reconciling the accuracy-diversity trade-off in recommendations",
                        "abstract": "In recommendation settings, there is an apparent trade-off between the goals of accuracy (to recommend items a user is most likely to want) and diversity (to recommend items representing a range of categories). As such, real-world recommender systems often explicitly incorporate diversity separately from accuracy. This approach, however, leaves a basic question unanswered: Why is there a trade-off in the first place?   We show how the trade-off can be explained via a user's consumption constraints -- users typically only consume a few of the items they are recommended. In a stylized model we introduce, objectives that account for this constraint induce diverse recommendations, while objectives that do not account for this constraint induce homogeneous recommendations. This suggests that accuracy and diversity appear misaligned because standard accuracy metrics do not consider consumption constraints. Our model yields precise and interpretable characterizations of diversity in different settings, giving practical insights into the design of diverse recommendations."
                    },
                    {
                        "title": "Roles for Computing in Social Change",
                        "abstract": "A recent normative turn in computer science has brought concerns about fairness, bias, and accountability to the core of the field. Yet recent scholarship has warned that much of this technical work treats problematic features of the status quo as fixed, and fails to address deeper patterns of injustice and inequality. While acknowledging these critiques, we posit that computational research has valuable roles to play in addressing social problems -- roles whose value can be recognized even from a perspective that aspires toward fundamental social change. In this paper, we articulate four such roles, through an analysis that considers the opportunities as well as the significant risks inherent in such work. Computing research can serve as a diagnostic, helping us to understand and measure social problems with precision and clarity. As a formalizer, computing shapes how social problems are explicitly defined --- changing how those problems, and possible responses to them, are understood. Computing serves as rebuttal when it illuminates the boundaries of what is possible through technical means. And computing acts as synecdoche when it makes long-standing social problems newly salient in the public eye. We offer these paths forward as modalities that leverage the particular strengths of computational work in the service of social change, without overclaiming computing's capacity to solve social problems on its own."
                    }
                ]
            },
            "6b74bb84-d9bb-4d04-8c44-b49245aec7e6": {
                "pk": "6b74bb84-d9bb-4d04-8c44-b49245aec7e6",
                "name": "Devavrat Shah",
                "collaborators": [
                    "Jinwoo Shin",
                    "Srikanth Jagabathula",
                    "Dogyoon Song",
                    "Qiaomin Xie",
                    "Tauhid Zaman",
                    "Kang Zhang",
                    "Sujay Sanghavi",
                    "Kyomin Jung",
                    "Damon Wischik",
                    "Ciamac Moallemi"
                ],
                "domain": [
                    "Queueing Theory",
                    "Network Optimization",
                    "Bayesian Inference",
                    "Reinforcement Learning"
                ],
                "publications": [
                    {
                        "title": "Efficient Queue-based CSMA with Collisions",
                        "abstract": "Recently there has been considerable interest in the design of efficient carrier sense multiple access(CSMA) protocol for wireless network. The basic assumption underlying recent results is availability of perfect carrier sense information. This allows for design of continuous time algorithm under which collisions are avoided. The primary purpose of this note is to show how these results can be extended in the case when carrier sense information may not be perfect, or equivalently delayed. Specifically, an adaptation of algorithm in Rajagopalan, Shah, Shin (2009) is presented here for time slotted setup with carrier sense information available only at the end of the time slot. To establish its throughput optimality, in additon to method developed in Rajagopalan, Shah, Shin (2009), understanding properties of stationary distribution of a certain non-reversible Markov chain as well as bound on its mixing time is essential. This note presents these key results. A longer version of this note will provide detailed account of how this gets incorporated with methods of Rajagopalan, Shah, Shin (2009) to provide the positive recurrence of underlying network Markov process. In addition, these results will help design optimal rate control in conjunction with CSMA in presence of collision building upon the method of Jiang, Shah, Shin, Walrand (2009)."
                    },
                    {
                        "title": "Bayesian regression and Bitcoin",
                        "abstract": "In this paper, we discuss the method of Bayesian regression and its efficacy for predicting price variation of Bitcoin, a recently popularized virtual, cryptographic currency. Bayesian regression refers to utilizing empirical data as proxy to perform Bayesian inference. We utilize Bayesian regression for the so-called \"latent source model\". The Bayesian regression for \"latent source model\" was introduced and discussed by Chen, Nikolov and Shah (2013) and Bresler, Chen and Shah (2014) for the purpose of binary classification. They established theoretical as well as empirical efficacy of the method for the setting of binary classification.   In this paper, instead we utilize it for predicting real-valued quantity, the price of Bitcoin. Based on this price prediction method, we devise a simple strategy for trading Bitcoin. The strategy is able to nearly double the investment in less than 60 day period when run against real data trace."
                    },
                    {
                        "title": "Tightness of LP via Max-product Belief Propagation",
                        "abstract": "We investigate the question of tightness of linear programming (LP) relaxation for finding a maximum weight independent set (MWIS) in sparse random weighted graphs. We show that an edge-based LP relaxation is asymptotically tight for Erdos-Renyi graph $G(n,c/n)$ for $c \\leq 2e$ and random regular graph $G(n,r)$ for $r\\leq 4$ when node weights are i.i.d. with exponential distribution of mean 1. We establish these results, through a precise relation between the tightness of LP relaxation and convergence of the max-product belief propagation algorithm. We believe that this novel method of understanding structural properties of combinatorial problems through properties of iterative procedure such as the max-product should be of interest in its own right."
                    },
                    {
                        "title": "Local approximate inference algorithms",
                        "abstract": "We present a new local approximation algorithm for computing Maximum a Posteriori (MAP) and log-partition function for arbitrary exponential family distribution represented by a finite-valued pair-wise Markov random field (MRF), say $G$. Our algorithm is based on decomposition of $G$ into {\\em appropriately} chosen small components; then computing estimates locally in each of these components and then producing a {\\em good} global solution. We show that if the underlying graph $G$ either excludes some finite-sized graph as its minor (e.g. Planar graph) or has low doubling dimension (e.g. any graph with {\\em geometry}), then our algorithm will produce solution for both questions within {\\em arbitrary accuracy}. We present a message-passing implementation of our algorithm for MAP computation using self-avoiding walk of graph. In order to evaluate the computational cost of this implementation, we derive novel tight bounds on the size of self-avoiding walk tree for arbitrary graph.   As a consequence of our algorithmic result, we show that the normalized log-partition function (also known as free-energy) for a class of {\\em regular} MRFs will converge to a limit, that is computable to an arbitrary accuracy."
                    },
                    {
                        "title": "Switched networks with maximum weight policies: Fluid approximation and multiplicative state space collapse",
                        "abstract": "We consider a queueing network in which there are constraints on which queues may be served simultaneously; such networks may be used to model input-queued switches and wireless networks. The scheduling policy for such a network specifies which queues to serve at any point in time. We consider a family of scheduling policies, related to the maximum-weight policy of Tassiulas and Ephremides [IEEE Trans. Automat. Control 37 (1992) 1936--1948], for single-hop and multihop networks. We specify a fluid model and show that fluid-scaled performance processes can be approximated by fluid model solutions. We study the behavior of fluid model solutions under critical load, and characterize invariant states as those states which solve a certain network-wide optimization problem. We use fluid model results to prove multiplicative state space collapse. A notable feature of our results is that they do not assume complete resource pooling."
                    },
                    {
                        "title": "On the Flow-level Dynamics of a Packet-switched Network",
                        "abstract": "The packet is the fundamental unit of transportation in modern communication networks such as the Internet. Physical layer scheduling decisions are made at the level of packets, and packet-level models with exogenous arrival processes have long been employed to study network performance, as well as design scheduling policies that more efficiently utilize network resources. On the other hand, a user of the network is more concerned with end-to-end bandwidth, which is allocated through congestion control policies such as TCP. Utility-based flow-level models have played an important role in understanding congestion control protocols. In summary, these two classes of models have provided separate insights for flow-level and packet-level dynamics of a network."
                    },
                    {
                        "title": "Fast Distributed Algorithms for Computing Separable Functions",
                        "abstract": "The problem of computing functions of values at the nodes in a network in a totally distributed manner, where nodes do not have unique identities and make decisions based only on local information, has applications in sensor, peer-to-peer, and ad-hoc networks. The task of computing separable functions, which can be written as linear combinations of functions of individual variables, is studied in this context. Known iterative algorithms for averaging can be used to compute the normalized values of such functions, but these algorithms do not extend in general to the computation of the actual values of separable functions.   The main contribution of this paper is the design of a distributed randomized algorithm for computing separable functions. The running time of the algorithm is shown to depend on the running time of a minimum computation algorithm used as a subroutine. Using a randomized gossip mechanism for minimum computation as the subroutine yields a complete totally distributed algorithm for computing separable functions. For a class of graphs with small spectral gap, such as grid graphs, the time used by the algorithm to compute averages is of a smaller order than the time required by a known iterative averaging scheme."
                    },
                    {
                        "title": "Throughput and Delay in Random Wireless Networks with Restricted Mobility",
                        "abstract": "Grossglauser and Tse (2001) introduced a mobile random network model where each node moves independently on a unit disk according to a stationary uniform distribution and showed that a throughput of $\\Theta(1)$ is achievable. El Gamal, Mammen, Prabhakar and Shah (2004) showed that the delay associated with this throughput scales as $\\Theta(n\\log n)$, when each node moves according to an independent random walk. In a later work, Diggavi, Grossglauser and Tse (2002) considered a random network on a sphere with a restricted mobility model, where each node moves along a randomly chosen great circle on the unit sphere. They showed that even with this one-dimensional restriction on mobility, constant throughput scaling is achievable. Thus, this particular mobility restriction does not affect the throughput scaling. This raises the question whether this mobility restriction affects the delay scaling.   This paper studies the delay scaling at $\\Theta(1)$ throughput for a random network with restricted mobility. First, a variant of the scheme presented by Diggavi, Grossglauser and Tse (2002) is presented and it is shown to achieve $\\Theta(1)$ throughput using different (and perhaps simpler) techniques. The exact order of delay scaling for this scheme is determined, somewhat surprisingly, to be of $\\Theta(n\\log n)$, which is the same as that without the mobility restriction. Thus, this particular mobility restriction \\emph{does not} affect either the maximal throughput scaling or the corresponding delay scaling of the network. This happens because under this 1-D restriction, each node is in the proximity of every other node in essentially the same manner as without this restriction."
                    },
                    {
                        "title": "Counting good truth assignments of random k-SAT formulae",
                        "abstract": "We present a deterministic approximation algorithm to compute logarithm of the number of `good' truth assignments for a random k-satisfiability (k-SAT) formula in polynomial time (by `good' we mean that violate a small fraction of clauses). The relative error is bounded above by an arbitrarily small constant epsilon with high probability as long as the clause density (ratio of clauses to variables) alpha<alpha_{u}(k) = 2k^{-1}\\log k(1+o(1)). The algorithm is based on computation of marginal distribution via belief propagation and use of an interpolation procedure. This scheme substitutes the traditional one based on approximation of marginal probabilities via MCMC, in conjunction with self-reduction, which is not easy to extend to the present problem.   We derive 2k^{-1}\\log k (1+o(1)) as threshold for uniqueness of the Gibbs distribution on satisfying assignment of random infinite tree k-SAT formulae to establish our results, which is of interest in its own right."
                    },
                    {
                        "title": "Fair Scheduling in Networks Through Packet Election",
                        "abstract": "We consider the problem of designing a fair scheduling algorithm for discrete-time constrained queuing networks. Each queue has dedicated exogenous packet arrivals. There are constraints on which queues can be served simultaneously. This model effectively describes important special instances like network switches, interference in wireless networks, bandwidth sharing for congestion control and traffic scheduling in road roundabouts. Fair scheduling is required because it provides isolation to different traffic flows; isolation makes the system more robust and enables providing quality of service. Existing work on fairness for constrained networks concentrates on flow based fairness. As a main result, we describe a notion of packet based fairness by establishing an analogy with the ranked election problem: packets are voters, schedules are candidates and each packet ranks the schedules based on its priorities. We then obtain a scheduling algorithm that achieves the described notion of fairness by drawing upon the seminal work of Goodman and Markowitz (1952). This yields the familiar Maximum Weight (MW) style algorithm. As another important result we prove that algorithm obtained is throughput optimal. There is no reason a priori why this should be true, and the proof requires non-traditional methods."
                    },
                    {
                        "title": "Randomized Scheduling Algorithm for Queueing Networks",
                        "abstract": "There has recently been considerable interest in design of low-complexity, myopic, distributed and stable scheduling policies for constrained queueing network models that arise in the context of emerging communication networks. Here, we consider two representative models. One, a model for the collection of wireless nodes communicating through a shared medium, that represents randomly varying number of packets in the queues at the nodes of networks. Two, a buffered circuit switched network model for an optical core of future Internet, to capture the randomness in calls or flows present in the network. The maximum weight scheduling policy proposed by Tassiulas and Ephremide in 1992 leads to a myopic and stable policy for the packet-level wireless network model. But computationally it is very expensive (NP-hard) and centralized. It is not applicable to the buffered circuit switched network due to the requirement of non-premption of the calls in the service. As the main contribution of this paper, we present a stable scheduling algorithm for both of these models. The algorithm is myopic, distributed and performs few logical operations at each node per unit time."
                    },
                    {
                        "title": "Deconvolution with Unknown Error Distribution Interpreted as Blind Isotonic Regression",
                        "abstract": "Deconvolution is a statistical inverse problem to estimate the distribution of a random variable based on its noisy observations. Despite the extensive studies on the topic, deconvolution with unknown noise distribution remains as a notoriously hard problem. We propose a matrix-based viewpoint for collective deconvolution that subsumes the setup with repeated measurements as a special case. As the main result, we describe a simple algorithm that partially utilizes matrix structure to solve deconvolution problem and provide non-asymptotic error analysis for the algorithm. We show that the proposed algorithm achieves the minimax optimal rate for deconvolution in a restricted sense. We also remark the connection between the collective deconvolution and the so-called statistical seriation as a byproduct or our matrix viewpoint. We conjecture that the link suggests that collective deconvolution, as well as deconvolution with repeated measurements, is intrinsically much easier than usual deconvolution of a single distribution."
                    },
                    {
                        "title": "Q-learning with Nearest Neighbors",
                        "abstract": "We consider model-free reinforcement learning for infinite-horizon discounted Markov Decision Processes (MDPs) with a continuous state space and unknown transition kernel, when only a single sample path under an arbitrary policy of the system is available. We consider the Nearest Neighbor Q-Learning (NNQL) algorithm to learn the optimal Q function using nearest neighbor regression method. As the main contribution, we provide tight finite sample analysis of the convergence rate. In particular, for MDPs with a $d$-dimensional state space and the discounted factor $\\gamma \\in (0,1)$, given an arbitrary sample path with \"covering time\" $ L $, we establish that the algorithm is guaranteed to output an $\\varepsilon$-accurate estimate of the optimal Q-function using $\\tilde{O}\\big(L/(\\varepsilon^3(1-\\gamma)^7)\\big)$ samples. For instance, for a well-behaved MDP, the covering time of the sample path under the purely random policy scales as $ \\tilde{O}\\big(1/\\varepsilon^d\\big),$ so the sample complexity scales as $\\tilde{O}\\big(1/\\varepsilon^{d+3}\\big).$ Indeed, we establish a lower bound that argues that the dependence of $ \\tilde{\\Omega}\\big(1/\\varepsilon^{d+2}\\big)$ is necessary."
                    },
                    {
                        "title": "Rumors in a Network: Who's the Culprit?",
                        "abstract": "We provide a systematic study of the problem of finding the source of a rumor in a network. We model rumor spreading in a network with a variant of the popular SIR model and then construct an estimator for the rumor source. This estimator is based upon a novel topological quantity which we term \\textbf{rumor centrality}. We establish that this is an ML estimator for a class of graphs. We find the following surprising threshold phenomenon: on trees which grow faster than a line, the estimator always has non-trivial detection probability, whereas on trees that grow like a line, the detection probability will go to 0 as the network grows. Simulations performed on synthetic networks such as the popular small-world and scale-free networks, and on real networks such as an internet AS network and the U.S. electric power grid network, show that the estimator either finds the source exactly or within a few hops of the true source across different network topologies. We compare rumor centrality to another common network centrality notion known as distance centrality. We prove that on trees, the rumor center and distance center are equivalent, but on general networks, they may differ. Indeed, simulations show that rumor centrality outperforms distance centrality in finding rumor sources in networks which are not tree-like."
                    },
                    {
                        "title": "Inferring Rankings Using Constrained Sensing",
                        "abstract": "We consider the problem of recovering a function over the space of permutations (or, the symmetric group) over $n$ elements from given partial information; the partial information we consider is related to the group theoretic Fourier Transform of the function. This problem naturally arises in several settings such as ranked elections, multi-object tracking, ranking systems, and recommendation systems. Inspired by the work of Donoho and Stark in the context of discrete-time functions, we focus on non-negative functions with a sparse support (support size $\\ll$ domain size). Our recovery method is based on finding the sparsest solution (through $\\ell_0$ optimization) that is consistent with the available information. As the main result, we derive sufficient conditions for functions that can be recovered exactly from partial information through $\\ell_0$ optimization. Under a natural random model for the generation of functions, we quantify the recoverability conditions by deriving bounds on the sparsity (support size) for which the function satisfies the sufficient conditions with a high probability as $n \\to \\infty$. $\\ell_0$ optimization is computationally hard. Therefore, the popular compressive sensing literature considers solving the convex relaxation, $\\ell_1$ optimization, to find the sparsest solution. However, we show that $\\ell_1$ optimization fails to recover a function (even with constant sparsity) generated using the random model with a high probability as $n \\to \\infty$. In order to overcome this problem, we propose a novel iterative algorithm for the recovery of functions that satisfy the sufficient conditions. Finally, using an Information Theoretic framework, we study necessary conditions for exact recovery to be possible."
                    },
                    {
                        "title": "Quantifying Variational Approximation for the Log-Partition Function",
                        "abstract": "Variational approximation, such as mean-field (MF) and tree-reweighted (TRW), provide a computationally efficient approximation of the log-partition function for a generic graphical model. TRW provably provides an upper bound, but the approximation ratio is generally not quantified.   As the primary contribution of this work, we provide an approach to quantify the approximation ratio through the property of the underlying graph structure. Specifically, we argue that (a variant of) TRW produces an estimate that is within factor $\\frac{1}{\\sqrt{\\kappa(G)}}$ of the true log-partition function for any discrete pairwise graphical model over graph $G$, where $\\kappa(G) \\in (0,1]$ captures how far $G$ is from tree structure with $\\kappa(G) = 1$ for trees and $2/N$ for the complete graph over $N$ vertices. As a consequence, the approximation ratio is $1$ for trees, $\\sqrt{(d+1)/2}$ for any graph with maximum average degree $d$, and $\\stackrel{\\beta\\to\\infty}{\\approx} 1+1/(2\\beta)$ for graphs with girth (shortest cycle) at least $\\beta \\log N$. In general, $\\kappa(G)$ is the solution of a max-min problem associated with $G$ that can be evaluated in polynomial time for any graph.   Using samples from the uniform distribution over the spanning trees of G, we provide a near linear-time variant that achieves an approximation ratio equal to the inverse of square-root of minimal (across edges) effective resistance of the graph. We connect our results to the graph partition-based approximation method and thus provide a unified perspective.   Keywords: variational inference, log-partition function, spanning tree polytope, minimum effective resistance, min-max spanning tree, local inference"
                    },
                    {
                        "title": "Centralized Congestion Control and Scheduling in a Datacenter",
                        "abstract": "We consider the problem of designing a packet-level congestion control and scheduling policy for datacenter networks. Current datacenter networks primarily inherit the principles that went into the design of Internet, where congestion control and scheduling are distributed. While distributed architecture provides robustness, it suffers in terms of performance. Unlike Internet, data center is fundamentally a \"controlled\" environment. This raises the possibility of designing a centralized architecture to achieve better performance. Recent solutions such as Fastpass and Flowtune have provided the proof of this concept. This raises the question: what is theoretically optimal performance achievable in a data center?   We propose a centralized policy that guarantees a per-flow end-to-end flow delay bound of $O$(#hops $\\times$ flow-size $/$ gap-to-capacity). Effectively such an end-to-end delay will be experienced by flows even if we removed congestion control and scheduling constraints as the resulting queueing networks can be viewed as the classical reversible multi-class queuing network, which has a product-form stationary distribution. In the language of Harrison et al., we establish that baseline performance for this model class is achievable.   Indeed, as the key contribution of this work, we propose a method to emulate such a reversible queuing network while satisfying congestion control and scheduling constraints. Precisely, our policy is an emulation of Store-and-Forward (SFA) congestion control in conjunction with Last-Come-First-Serve Preemptive-Resume (LCFS-PR) scheduling policy."
                    },
                    {
                        "title": "Learning RUMs: Reducing Mixture to Single Component via PCA",
                        "abstract": "We consider the problem of learning a mixture of Random Utility Models (RUMs). Despite the success of RUMs in various domains and the versatility of mixture RUMs to capture the heterogeneity in preferences, there has been only limited progress in learning a mixture of RUMs from partial data such as pairwise comparisons. In contrast, there have been significant advances in terms of learning a single component RUM using pairwise comparisons. In this paper, we aim to bridge this gap between mixture learning and single component learning of RUM by developing a `reduction' procedure. We propose to utilize PCA-based spectral clustering that simultaneously `de-noises' pairwise comparison data. We prove that our algorithm manages to cluster the partial data correctly (i.e., comparisons from the same RUM component are grouped in the same cluster) with high probability even when data is generated from a possibly {\\em heterogeneous} mixture of well-separated {\\em generic} RUMs. Both the time and the sample complexities scale polynomially in model parameters including the number of items. Two key features in the analysis are in establishing (1) a meaningful upper bound on the sub-Gaussian norm for RUM components embedded into the vector space of pairwise marginals and (2) the robustness of PCA with missing values in the $L_{2, \\infty}$ sense, which might be of interest in their own right."
                    },
                    {
                        "title": "Learning Mixed Multinomial Logit Model from Ordinal Data",
                        "abstract": "Motivated by generating personalized recommendations using ordinal (or preference) data, we study the question of learning a mixture of MultiNomial Logit (MNL) model, a parameterized class of distributions over permutations, from partial ordinal or preference data (e.g. pair-wise comparisons). Despite its long standing importance across disciplines including social choice, operations research and revenue management, little is known about this question. In case of single MNL models (no mixture), computationally and statistically tractable learning from pair-wise comparisons is feasible. However, even learning mixture with two MNL components is infeasible in general.   Given this state of affairs, we seek conditions under which it is feasible to learn the mixture model in both computationally and statistically efficient manner. We present a sufficient condition as well as an efficient algorithm for learning mixed MNL models from partial preferences/comparisons data. In particular, a mixture of $r$ MNL components over $n$ objects can be learnt using samples whose size scales polynomially in $n$ and $r$ (concretely, $r^{3.5}n^3(log n)^4$, with $r\\ll n^{2/7}$ when the model parameters are sufficiently incoherent). The algorithm has two phases: first, learn the pair-wise marginals for each component using tensor decomposition; second, learn the model parameters for each component using Rank Centrality introduced by Negahban et al. In the process of proving these results, we obtain a generalization of existing analysis for tensor decomposition to a more realistic regime where only partial information about each sample is available."
                    },
                    {
                        "title": "Finding Rumor Sources on Random Trees",
                        "abstract": "We consider the problem of detecting the source of a rumor which has spread in a network using only observations about which set of nodes are infected with   the rumor and with no information as to \\emph{when} these nodes became infected. In a recent work \\citep{ref:rc} this rumor source detection problem was introduced and studied. The authors proposed the graph score function {\\em rumor centrality} as an estimator for detecting the source. They establish it to be the maximum likelihood estimator with respect to the popular Susceptible Infected (SI) model with exponential spreading times for regular trees. They showed that as the size of the infected graph increases, for a path graph (2-regular tree), the probability of source detection goes to $0$ while for $d$-regular trees with $d \\geq 3$ the probability of detection, say $\\alpha_d$, remains bounded away from $0$ and is less than $1/2$. However, their results stop short of providing insights for the performance of the rumor centrality estimator in more general settings such as irregular trees or the SI model with non-exponential spreading times.   This paper overcomes this limitation and establishes the effectiveness of rumor centrality for source detection for generic random trees and the SI model with a generic spreading time distribution. The key result is an interesting connection between a continuous time branching process and the effectiveness of rumor centrality. Through this, it is possible to quantify the detection probability precisely. As a consequence, we recover all previous results as a special case and obtain a variety of novel results including the {\\em universality} of rumor centrality in the context of tree-like graphs and the SI model with a generic spreading time distribution."
                    }
                ]
            }
        }
    },
    "2409.19472": {
        "paper_data": {
            "title": "Towards Croppable Implicit Neural Representations",
            "url": "http://arxiv.org/abs/2409.19472v2",
            "arxiv_id": "2409.19472",
            "authors": [
                "Maor Ashkenazi",
                "Eran Treister"
            ],
            "abstract": "Implicit Neural Representations (INRs) have peaked interest in recent years due to their ability to encode natural signals using neural networks. While INRs allow for useful applications such as interpolating new coordinates and signal compression, their black-box nature makes it difficult to modify them post-training. In this paper we explore the idea of editable INRs, and specifically focus on the widely used cropping operation. To this end, we present Local-Global SIRENs -- a novel INR architecture that supports cropping by design. Local-Global SIRENs are based on combining local and global feature extraction for signal encoding. What makes their design unique is the ability to effortlessly remove specific portions of an encoded signal, with a proportional weight decrease. This is achieved by eliminating the corresponding weights from the network, without the need for retraining. We further show how this architecture can be used to support the straightforward extension of previously encoded signals. Beyond signal editing, we examine how the Local-Global approach can accelerate training, enhance encoding of various signals, improve downstream performance, and be applied to modern INRs such as INCODE, highlighting its potential and flexibility. Code is available at https://github.com/maorash/Local-Global-INRs.",
            "introduction": "   1 Introduction  Neural networks have proven to be an effective tool for learning representations of various natural signals. This advancement offers a paradigm of representing a signal without explicitly defining it. The general idea behind an implicit neural representation (INR) is to model the signal as a prediction task from some coordinate system to the signal values at that coordinate system. This is usually performed using a Multi-Layer Perceptron (MLP), composed of fully connected layers. Once trained, the signal has been implicitly encoded in the weights of the network. These implicit representations have been shown to be useful in various scenarios from interpolating values at new coordinates [29], through signal compression [14] and can be treated as embeddings for downstream tasks [13].   Although INRs exhibit flexibility within their input space and are well-suited for tasks involving high-dimensional data, their black-box nature makes the encoded signal difficult to modify post-training. A trivial strategy involves editing the original signal, followed by training a new INR to encode the modified signal. Another option is fine-tuning the existing INR to encode the modified signal, but it requires preserving the INR size and architecture, which may not be ideal. Other methods attempted to apply direct transformations on the weight space to modify the encoded signal [42, 30].   In this paper, our primary focus is on the fundamental signal editing operation of cropping. We wish to be able to remove parts of the encoded signal, with a proportional decrease of INR weights. Although not possible in previous approaches, in principle this should be obtained without any additional fine-tuning, since it conceptually does not require encoding additional information. As a secondary task, we wish to extend an encoded signal effectively. This is not trivial in a simple setting, since encoding additional information may require increased capacity in terms of parameters, and altering the number of parameters makes it difficult to leverage previously encoded information.   As a natural way to obtain our goals, we first partition the input signal space. The granularity of the partitioning will determine the ability to edit the INR with greater detail. A straightforward approach involves training a compact INR for each partition of the signal. Subsequently, cropping and extending signals are simple. To crop specific partitions, one simply eliminates the INRs corresponding to those partitions. To extend INRs, new signal partitions can be seamlessly incorporated by training additional compact INRs on the new partitions, subsequently adding them to the ensemble. Notably, this straightforward approach offers a significant speed benefit. While requiring a unique implementation, utilizing an ensemble of compact INRs for both encoding and reconstructing the signal is much more efficient than a fully connected INR. Indeed, in KiloNeRF [35], a 3D scene was encoded by adopting the INR-per-Partition approach, leading to notable improvements in rendering times.   However, this partitioning approach lacks a global context, resulting in increased reconstruction error and undesired artifacts. In [35, 15], this was solved using knowledge distillation. Initially, a large INR was trained to encode the entire signal. Next, the encoded representation was distilled into the ensemble of compact INRs. While this approach facilitates rapid rendering,",
            "references": [
                {
                    "title": "ImFace++: A Sophisticated Nonlinear 3D Morphable Face Model with Implicit Neural Representations",
                    "abstract": "Accurate representations of 3D faces are of paramount importance in various computer vision and graphics applications. However, the challenges persist due to the limitations imposed by data discretization and model linearity, which hinder the precise capture of identity and expression clues in current studies. This paper presents a novel 3D morphable face model, named ImFace++, to learn a sophisticated and continuous space with implicit neural representations. ImFace++ first constructs two explicitly disentangled deformation fields to model complex shapes associated with identities and expressions, respectively, which simultaneously facilitate automatic learning of point-to-point correspondences across diverse facial shapes. To capture more sophisticated facial details, a refinement displacement field within the template space is further incorporated, enabling fine-grained learning of individual-specific facial details. Furthermore, a Neural Blend-Field is designed to reinforce the representation capabilities through adaptive blending of an array of local fields. In addition to ImFace++, we devise an improved learning strategy to extend expression embeddings, allowing for a broader range of expression variations. Comprehensive qualitative and quantitative evaluation demonstrates that ImFace++ significantly advances the state-of-the-art in terms of both face reconstruction fidelity and correspondence accuracy."
                },
                {
                    "title": "Geometry Processing with Neural Fields",
                    "abstract": "Most existing geometry processing algorithms use meshes as the default shape representation. Manipulating meshes, however, requires one to maintain high quality in the surface discretization. For example, changing the topology of a mesh usually requires additional procedures such as remeshing. This paper instead proposes the use of neural fields for geometry processing. Neural fields can compactly store complicated shapes without spatial discretization. Moreover, neural fields are infinitely differentiable, which allows them to be optimized for objectives that involve higher-order derivatives. This raises the question: can geometry processing be done entirely using neural fields? We introduce loss functions and architectures to show that some of the most challenging geometry processing tasks, such as deformation and filtering, can be done with neural fields. Experimental results show that our methods are on par with the well-established mesh-based methods without committing to a particular surface discretization. Code is available at https://github.com/stevenygd/NFGP ."
                },
                {
                    "title": "INCODE: Implicit Neural Conditioning with Prior Knowledge Embeddings",
                    "abstract": "Implicit Neural Representations (INRs) have revolutionized signal representation by leveraging neural networks to provide continuous and smooth representations of complex data. However, existing INRs face limitations in capturing fine-grained details, handling noise, and adapting to diverse signal types. To address these challenges, we introduce INCODE, a novel approach that enhances the control of the sinusoidal-based activation function in INRs using deep prior knowledge. INCODE comprises a harmonizer network and a composer network, where the harmonizer network dynamically adjusts key parameters of the activation function. Through a task-specific pre-trained model, INCODE adapts the task-specific parameters to optimize the representation process. Our approach not only excels in representation, but also extends its prowess to tackle complex tasks such as audio, image, and 3D shape reconstructions, as well as intricate challenges such as neural radiance fields (NeRFs), and inverse problems, including denoising, super-resolution, inpainting, and CT reconstruction. Through comprehensive experiments, INCODE demonstrates its superiority in terms of robustness, accuracy, quality, and convergence rate, broadening the scope of signal representation. Please visit the project\u2019s website for details on the proposed method and access to the code."
                },
                {
                    "title": "Partition Speeds Up Learning Implicit Neural Representations Based on Exponential-Increase Hypothesis",
                    "abstract": "Implicit neural representations (INRs) aim to learn a continuous function (i.e., a neural network) to represent an image, where the input and output of the function are pixel coordinates and RGB/Gray values, respectively. However, images tend to consist of many objects whose colors are not perfectly consistent, resulting in the challenge that image is actually a discontinuous piecewise function and cannot be well estimated by a continuous function. In this paper, we empirically investigate that if a neural network is enforced to fit a discontinuous piecewise function to reach a fixed small error, the time costs will increase exponentially with respect to the boundaries in the spatial domain of the target signal. We name this phenomenon the exponential-increase hypothesis. Under the exponential-increase hypothesis, learning INRs for images with many objects will converge very slowly. To address this issue, we first prove that partitioning a complex signal into several sub-regions and utilizing piecewise INRs to fit that signal can significantly speed up the convergence. Based on this fact, we introduce a simple partition mechanism to boost the performance of two INR methods for image reconstruction: one for learning INRs, and the other for learning-to-learn INRs. In both cases, we partition an image into different sub-regions and dedicate smaller networks for each part. In addition, we further propose two partition rules based on regular grids and semantic segmentation maps, respectively. Extensive experiments validate the effectiveness of the proposed partitioning methods in terms of learning INR for a single image (ordinary learning framework) and the learning-to-learn framework. Code is released here."
                },
                {
                    "title": "Multiplicative Fourier Level of Detail",
                    "abstract": "We develop a simple yet surprisingly effective implicit representing scheme called Multiplicative Fourier Level of Detail (MFLOD) motivated by the recent success of multiplicative filter network. Built on multi-resolution feature grid/volume (e.g., the sparse voxel octree), each level's feature is first modulated by a sinusoidal function and then element-wisely multiplied by a linear transformation of previous layer's representation in a layer-to-layer recursive manner, yielding the scale-aggregated encodings for a subsequent simple linear forward to get final output. In contrast to previous hybrid representations relying on interleaved multilevel fusion and nonlinear activation-based decoding, MFLOD could be elegantly characterized as a linear combination of sine basis functions with varying amplitude, frequency, and phase upon the learned multilevel features, thus offering great feasibility in Fourier analysis. Comprehensive experimental results on implicit neural representation learning tasks including image fitting, 3D shape representation, and neural radiance fields well demonstrate the superior quality and generalizability achieved by the proposed MFLOD scheme."
                },
                {
                    "title": "LP-DIF: Learning Local Pattern-Specific Deep Implicit Function for 3D Objects and Scenes",
                    "abstract": "Deep Implicit Function (DIF) has gained much popularity as an efficient 3D shape representation. To capture geometry details, current mainstream methods divide 3D shapes into local regions and then learn each one with a local latent code via a decoder. Such local methods can capture more local details due to less diversity among local regions than global shapes. Although the diversity of local regions has been decreased compared to global approaches, the diversity in different local regions still poses a challenge in learning an implicit function when treating all regions equally using only a single decoder. What is worse, these local regions often exhibit imbalanced distributions, where certain regions have significantly fewer observations. This leads that fine geometry details could not be preserved well. To solve this problem, we propose a novel Local Pattern-specific Implicit Function, named LP-DIF, to represent a shape with clusters of local regions and multiple decoders, where each decoder only focuses on one cluster of local regions which share a certain pattern. Specifically, we first extract local codes for all regions, and then cluster them into multiple groups in the latent space, where similar regions sharing a common pattern fall into one group. After that, we train multiple decoders for mining local patterns of different groups, which simplifies the learning of fine geometric details by reducing the diversity of local regions seen by each decoder. To further alleviate the data-imbalance problem, we introduce a region re-weighting module to each pattern-specific decoder using a kernel density estimator, which dynamically re-weights the regions during learning. Our LP-DIF can restore more geometry details, and thus improve the quality of 3D reconstruction. Experiments demonstrate that our method can achieve the state-of-the-art performance over previous methods. Code is available at https://github.com/gtyxyz/lpdif."
                },
                {
                    "title": "NeuralEditor: Editing Neural Radiance Fields via Manipulating Point Clouds",
                    "abstract": "This paper proposes NeuralEditor that enables neural radiance fields (NeRFs) natively editable for general shape editing tasks. Despite their impressive results on novel-view synthesis, it remains a fundamental challenge for NeRFs to edit the shape of the scene. Our key insight is to exploit the explicit point cloud representation as the underlying structure to construct NeRFs, inspired by the intuitive interpretation of NeRF rendering as a process that projects or \u201cplots\u201d the associated 3D point cloud to a 2D image plane. To this end, NeuralEditor introduces a novel rendering scheme based on deterministic integration within K-D tree-guided density-adaptive voxels, which produces both high-quality rendering results and precise point clouds through optimization. NeuralEditor then performs shape editing via mapping associated points between point clouds. Extensive evaluation shows that NeuralEditor achieves state-of-the-art performance in both shape deformation and scene morphing tasks. Notably, NeuralEditor supports both zeroshot inference and further fine-tuning over the edited scene. Our code, benchmark, and demo video are available at immortalco.github.io/NeuralEditor."
                },
                {
                    "title": "Neural Field Convolutions by Repeated Differentiation",
                    "abstract": "Neural fields are evolving towards a general-purpose continuous representation for visual computing. Yet, despite their numerous appealing properties, they are hardly amenable to signal processing. As a remedy, we present a method to perform general continuous convolutions with general continuous signals such as neural fields. Observing that piecewise polynomial kernels reduce to a sparse set of Dirac deltas after repeated differentiation, we leverage convolution identities and train a repeated integral field to efficiently execute large-scale convolutions. We demonstrate our approach on a variety of data modalities and spatially-varying kernels."
                },
                {
                    "title": "Equivariant Architectures for Learning in Deep Weight Spaces",
                    "abstract": "Designing machine learning architectures for processing neural networks in their raw weight matrix form is a newly introduced research direction. Unfortunately, the unique symmetry structure of deep weight spaces makes this design very challenging. If successful, such architectures would be capable of performing a wide range of intriguing tasks, from adapting a pre-trained network to a new domain to editing objects represented as functions (INRs or NeRFs). As a first step towards this goal, we present here a novel network architecture for learning in deep weight spaces. It takes as input a concatenation of weights and biases of a pre-trained MLP and processes it using a composition of layers that are equivariant to the natural permutation symmetry of the MLP's weights: Changing the order of neurons in intermediate layers of the MLP does not affect the function it represents. We provide a full characterization of all affine equivariant and invariant layers for these symmetries and show how these layers can be implemented using three basic operations: pooling, broadcasting, and fully connected layers applied to the input in an appropriate manner. We demonstrate the effectiveness of our architecture and its advantages over natural baselines in a variety of learning tasks."
                },
                {
                    "title": "RecolorNeRF: Layer Decomposed Radiance Fields for Efficient Color Editing of 3D Scenes",
                    "abstract": "Radiance fields have gradually become a main representation of media. Although its appearance editing has been studied, how to achieve view-consistent recoloring in an efficient manner is still under explored. We present RecolorNeRF, a novel user-friendly color editing approach for the neural radiance fields. Our key idea is to decompose the scene into a set of pure-colored layers, forming a palette. By this means, color manipulation can be conducted by altering the color components of the palette directly. To support efficient palette-based editing, the color of each layer needs to be as representative as possible. In the end, the problem is formulated as an optimization problem, where the layers and their blending weights are jointly optimized with the NeRF itself. Extensive experiments show that our jointly-optimized layer decomposition can be used against multiple backbones and produce photo-realistic recolored novel-view renderings. We demonstrate that RecolorNeRF outperforms baseline methods both quantitatively and qualitatively for color editing even in complex real-world scenes."
                },
                {
                    "title": "WIRE: Wavelet Implicit Neural Representations",
                    "abstract": "Implicit neural representations (INRs) have recently advanced numerous vision-related areas. INR performance depends strongly on the choice of activation function employed in its MLP network. A wide range of nonlinearities have been explored, but, unfortunately, current INRs designed to have high accuracy also suffer from poor robustness (to signal noise, parameter variation, etc.). Inspired by harmonic analysis, we develop a new, highly accurate and robust INR that does not exhibit this trade off. Our Wavelet Implicit neural REpresentation (WIRE) uses as its activation function the complex Gabor wavelet that is well-known to be optimally concentrated in space-frequency and to have excellent biases for representing images. A wide range of experiments (image denoising, image inpainting, super-resolution, computed tomography reconstruction, image over fitting, and novel view synthesis with neural radiance fields) demonstrate that WIRE defines the new state of the art in INR accuracy, training time, and robustness."
                },
                {
                    "title": "Single Image Super-Resolution via a Dual Interactive Implicit Neural Network",
                    "abstract": "In this paper, we introduce a novel implicit neural network for the task of single image super-resolution at arbitrary scale factors. To do this, we represent an image as a decoding function that maps locations in the image along with their associated features to their reciprocal pixel attributes. Since the pixel locations are continuous in this representation, our method can refer to any location in an image of varying resolution. To retrieve an image of a particular resolution, we apply a decoding function to a grid of locations each of which refers to the center of a pixel in the output image. In contrast to other techniques, our dual interactive neural network decouples content and positional features. As a result, we obtain a fully implicit representation of the image that solves the super-resolution problem at (real-valued) elective scales using a single model. We demonstrate the efficacy and flexibility of our approach against the state of the art on publicly available benchmark datasets."
                },
                {
                    "title": "Signal Processing for Implicit Neural Representations",
                    "abstract": "Implicit Neural Representations (INRs) encoding continuous multi-media data via multi-layer perceptrons has shown undebatable promise in various computer vision tasks. Despite many successful applications, editing and processing an INR remains intractable as signals are represented by latent parameters of a neural network. Existing works manipulate such continuous representations via processing on their discretized instance, which breaks down the compactness and continuous nature of INR. In this work, we present a pilot study on the question: how to directly modify an INR without explicit decoding? We answer this question by proposing an implicit neural signal processing network, dubbed INSP-Net, via differential operators on INR. Our key insight is that spatial gradients of neural networks can be computed analytically and are invariant to translation, while mathematically we show that any continuous convolution filter can be uniformly approximated by a linear combination of high-order differential operators. With these two knobs, INSP-Net instantiates the signal processing operator as a weighted composition of computational graphs corresponding to the high-order derivatives of INRs, where the weighting parameters can be data-driven learned. Based on our proposed INSP-Net, we further build the first Convolutional Neural Network (CNN) that implicitly runs on INRs, named INSP-ConvNet. Our experiments validate the expressiveness of INSP-Net and INSP-ConvNet in fitting low-level image and geometry processing kernels (e.g. blurring, deblurring, denoising, inpainting, and smoothening) as well as for high-level tasks on implicit fields such as image classification."
                },
                {
                    "title": "KiloNeuS: A Versatile Neural Implicit Surface Representation for Real-Time Rendering",
                    "abstract": "NeRF-based techniques \ufb01t wide and deep multi-layer perceptrons (MLPs) to a continuous radiance \ufb01eld that can be rendered from any unseen viewpoint. However, the lack of surface and normals de\ufb01nition and high rendering times limit their usage in typical computer graphics applications. Such limitations have recently been overcome separately, but solving them together remains an open problem. We present KiloNeuS, a neural representation reconstructing an implicit surface represented as a signed distance function (SDF) from multi-view images and enabling real-time rendering by partitioning the space into thousands of tiny MLPs fast to inference. As we learn the implicit surface locally using independent models, resulting in a globally coherent geometry is non-trivial and needs to be addressed during training. We evaluate rendering performance on a GPU-accelerated ray-caster with in-shader neural network inference, resulting in an average of 46 FPS at high resolution, proving a satisfying tradeoff between storage costs and rendering quality. In fact, our evaluation for rendering quality and surface recovery shows that KiloNeuS outper-forms its single-MLP counterpart. Finally, to exhibit the versatility of KiloNeuS, we integrate it into an interactive path-tracer taking full advantage of its surface normals. We consider our work a crucial \ufb01rst step toward real-time rendering of implicit neural representations under global illumination."
                },
                {
                    "title": "Unified Implicit Neural Stylization",
                    "abstract": "Representing visual signals by implicit representation (e.g., a coordinate based deep network) has prevailed among many vision tasks. This work explores a new intriguing direction: training a stylized implicit representation, using a generalized approach that can apply to various 2D and 3D scenarios. We conduct a pilot study on a variety of implicit functions, including 2D coordinate-based representation, neural radiance field, and signed distance function. Our solution is a Unified Implicit Neural Stylization framework, dubbed INS. In contrary to vanilla implicit representation, INS decouples the ordinary implicit function into a style implicit module and a content implicit module, in order to separately encode the representations from the style image and input scenes. An amalgamation module is then applied to aggregate these information and synthesize the stylized output. To regularize the geometry in 3D scenes, we propose a novel self-distillation geometry consistency loss which preserves the geometry fidelity of the stylized scenes. Comprehensive experiments are conducted on multiple task settings, including novel view synthesis of complex scenes, stylization for implicit surfaces, and fitting images using MLPs. We further demonstrate that the learned representation is continuous not only spatially but also style-wise, leading to effortlessly interpolating between different styles and generating images with new mixed styles. Please refer to the video on our project page for more view synthesis results: https://zhiwenfan.github.io/INS."
                },
                {
                    "title": "Zero-shot Blind Image Denoising via Implicit Neural Representations",
                    "abstract": "Recent denoising algorithms based on the\"blind-spot\"strategy show impressive blind image denoising performances, without utilizing any external dataset. While the methods excel in recovering highly contaminated images, we observe that such algorithms are often less effective under a low-noise or real noise regime. To address this gap, we propose an alternative denoising strategy that leverages the architectural inductive bias of implicit neural representations (INRs), based on our two findings: (1) INR tends to fit the low-frequency clean image signal faster than the high-frequency noise, and (2) INR layers that are closer to the output play more critical roles in fitting higher-frequency parts. Building on these observations, we propose a denoising algorithm that maximizes the innate denoising capability of INRs by penalizing the growth of deeper layer weights. We show that our method outperforms existing zero-shot denoising methods under an extensive set of low-noise or real-noise scenarios."
                },
                {
                    "title": "ImFace: A Nonlinear 3D Morphable Face Model with Implicit Neural Representations",
                    "abstract": "Precise representations of 3D faces are beneficial to various computer vision and graphics applications. Due to the data discretization and model linearity, however, it remains challenging to capture accurate identity and expression clues in current studies. This paper presents a novel 3D morphable face model, namely ImFace, to learn a nonlinear and continuous space with implicit neural representations. It builds two explicitly disentangled deformation fields to model complex shapes associated with identities and expressions, respectively, and designs an improved learning strategy to extend embeddings of expressions to allow more diverse changes. We further introduce a Neural Blend-Field to learn sophisticated details by adaptively blending a series of local fields. In addition to ImFace, an effective pre-processing pipeline is proposed to address the issue of watertight input requirement in implicit representations, enabling them to work with common facial surfaces for the first time. Extensive experiments are performed to demonstrate the superiority of ImFace."
                },
                {
                    "title": "COIN++: Neural Compression Across Modalities",
                    "abstract": "Neural compression algorithms are typically based on autoencoders that require specialized encoder and decoder architectures for different data modalities. In this paper, we propose COIN++, a neural compression framework that seamlessly handles a wide range of data modalities. Our approach is based on converting data to implicit neural representations, i.e. neural functions that map coordinates (such as pixel locations) to features (such as RGB values). Then, instead of storing the weights of the implicit neural representation directly, we store modulations applied to a meta-learned base network as a compressed code for the data. We further quantize and entropy code these modulations, leading to large compression gains while reducing encoding time by two orders of magnitude compared to baselines. We empirically demonstrate the feasibility of our method by compressing various data modalities, from images and audio to medical and climate data."
                },
                {
                    "title": "CoordX: Accelerating Implicit Neural Representation with a Split MLP Architecture",
                    "abstract": "Implicit neural representations with multi-layer perceptrons (MLPs) have recently gained prominence for a wide variety of tasks such as novel view synthesis and 3D object representation and rendering. However, a significant challenge with these representations is that both training and inference with an MLP over a large number of input coordinates to learn and represent an image, video, or 3D object, require large amounts of computation and incur long processing times. In this work, we aim to accelerate inference and training of coordinate-based MLPs for implicit neural representations by proposing a new split MLP architecture, CoordX. With CoordX, the initial layers are split to learn each dimension of the input coordinates separately. The intermediate features are then fused by the last layers to generate the learned signal at the corresponding coordinate point. This significantly reduces the amount of computation required and leads to large speedups in training and inference, while achieving similar accuracy as the baseline MLP. This approach thus aims at first learning functions that are a decomposition of the original signal and then fusing them to generate the learned signal. Our proposed architecture can be generally used for many implicit neural representation tasks with no additional memory overheads. We demonstrate a speedup of up to 2.92x compared to the baseline model for image, video, and 3D shape representation and rendering tasks."
                },
                {
                    "title": "From data to functa: Your data point is a function and you can treat it like one",
                    "abstract": "It is common practice in deep learning to represent a measurement of the world on a discrete grid, e.g. a 2D grid of pixels. However, the underlying signal represented by these measurements is often continuous, e.g. the scene depicted in an image. A powerful continuous alternative is then to represent these measurements using an implicit neural representation, a neural function trained to output the appropriate measurement value for any input spatial location. In this paper, we take this idea to its next level: what would it take to perform deep learning on these functions instead, treating them as data? In this context we refer to the data as functa, and propose a framework for deep learning on functa. This view presents a number of challenges around efficient conversion from data to functa, compact representation of functa, and effectively solving downstream tasks on functa. We outline a recipe to overcome these challenges and apply it to a wide range of data modalities including images, 3D shapes, neural radiance fields (NeRF) and data on manifolds. We demonstrate that this approach has various compelling properties across data modalities, in particular on the canonical tasks of generative modeling, data imputation, novel view synthesis and classification. Code: https://github.com/deepmind/functa"
                },
                {
                    "title": "Bacon: Band-limited Coordinate Networks for Multiscale Scene Representation",
                    "abstract": "Coordinate-based networks have emerged as a powerful tool for 3D representation and scene reconstruction. These networks are trained to map continuous input coordinates to the value of a signal at each point. Still, current architectures are black boxes: their spectral characteristics cannot be easily analyzed, and their behavior at unsupervised points is difficult to predict. Moreover, these networks are typically trained to represent a signal at a single scale, so naive downsampling or upsampling results in artifacts. We introduce band-limited coordinate networks (BACON), a network architecture with an analytical Fourier spectrum. Bacon has constrained behavior at unsupervised points, can be designed based on the spectral characteristics of the represented signal, and can represent signals at multiple scales without per-scale supervision. We demonstrate Bacon for multiscale neural representation of images, radiance fields, and 3D scenes using signed distance functions and show that it outperforms conventional single-scale coordinate networks in terms of interpretability and quality."
                },
                {
                    "title": "NeRV: Neural Representations for Videos",
                    "abstract": "We propose a novel neural representation for videos (NeRV) which encodes videos in neural networks. Unlike conventional representations that treat videos as frame sequences, we represent videos as neural networks taking frame index as input. Given a frame index, NeRV outputs the corresponding RGB image. Video encoding in NeRV is simply fitting a neural network to video frames and decoding process is a simple feedforward operation. As an image-wise implicit representation, NeRV output the whole image and shows great efficiency compared to pixel-wise implicit representation, improving the encoding speed by 25x to 70x, the decoding speed by 38x to 132x, while achieving better video quality. With such a representation, we can treat videos as neural networks, simplifying several video-related tasks. For example, conventional video compression methods are restricted by a long and complex pipeline, specifically designed for the task. In contrast, with NeRV, we can use any neural network compression method as a proxy for video compression, and achieve comparable performance to traditional frame-based video compression approaches (H.264, HEVC \\etc). Besides compression, we demonstrate the generalization of NeRV for video denoising. The source code and pre-trained model can be found at https://github.com/haochen-rye/NeRV.git."
                },
                {
                    "title": "Multi-Head Relu Implicit Neural Representation Networks",
                    "abstract": "In this paper, a novel multi-head multi-layer perceptron (MLP) structure is presented for implicit neural representation (INR). Since conventional rectified linear unit (ReLU) networks are shown to exhibit spectral bias towards learning low-frequency features of the signal, we aim at mitigating this defect by taking advantage of local structure of the signals. To be more specific, an MLP is used to capture the global features of the underlying generator function of the desired signal. Then, several heads are utilized to reconstruct disjoint local features of the signal, and to reduce the computational complexity, sparse layers are deployed for attaching heads to the body. Through various experiments, we show that the proposed model does not suffer from the special bias of conventional ReLU networks and has superior generalization capabilities. Finally, simulation results confirm that the proposed multi-head structure outperforms existing INR methods with considerably less computational cost. The source code is available at https://github.com/AlirezaMorsali/MH-RELU-INR"
                },
                {
                    "title": "Multiresolution Deep Implicit Functions for 3D Shape Representation",
                    "abstract": "We introduce Multiresolution Deep Implicit Functions (MDIF), a hierarchical representation that can recover fine geometry detail, while being able to perform global operations such as shape completion. Our model represents a complex 3D shape with a hierarchy of latent grids, which can be decoded into different levels of detail and also achieve better accuracy. For shape completion, we propose latent grid dropout to simulate partial data in the latent space and therefore defer the completing functionality to the decoder side. This along with our multires design significantly improves the shape completion quality under decoder-only latent optimization. To the best of our knowledge, MDIF is the first deep implicit function model that can at the same time (1) represent different levels of detail and allow progressive decoding; (2) support both encoder-decoder inference and decoder-only latent optimization, and fulfill multiple applications; (3) perform detailed decoder-only shape completion. Experiments demonstrate its superior performance against prior art in various 3D reconstruction tasks."
                },
                {
                    "title": "Modulated Periodic Activations for Generalizable Local Functional Representations",
                    "abstract": "Multi-Layer Perceptrons (MLPs) make powerful functional representations for sampling and reconstruction problems involving low-dimensional signals like images, shapes and light fields. Recent works have significantly improved their ability to represent high-frequency content by using periodic activations or positional encodings. This often came at the expense of generalization: modern methods are typically optimized for a single signal. We present a new representation that generalizes to multiple instances and achieves state-of-the-art fidelity. We use a dual-MLP architecture to encode the signals. A synthesis network creates a functional mapping from a low-dimensional input (e.g. pixel-position) to the output domain (e.g. RGB color). A modulation network maps a latent code corresponding to the target signal to parameters that modulate the periodic activations of the synthesis network. We also propose a local-functional representation which enables generalization. The signal\u2019s domain is partitioned into a regular grid, with each tile represented by a latent code. At test time, the signal is encoded with high-fidelity by inferring (or directly optimizing) the latent code-book. Our approach produces generalizable functional representations of images, videos and shapes, and achieves higher reconstruction quality than prior works that are optimized for a single signal."
                },
                {
                    "title": "KiloNeRF: Speeding up Neural Radiance Fields with Thousands of Tiny MLPs",
                    "abstract": "NeRF synthesizes novel views of a scene with unprecedented quality by fitting a neural radiance field to RGB images. However, NeRF requires querying a deep Multi-Layer Perceptron (MLP) millions of times, leading to slow rendering times, even on modern GPUs. In this paper, we demonstrate that real-time rendering is possible by utilizing thousands of tiny MLPs instead of one single large MLP. In our setting, each individual MLP only needs to represent parts of the scene, thus smaller and faster-to-evaluate MLPs can be used. By combining this divide-and-conquer strategy with further optimizations, rendering is accelerated by three orders of magnitude compared to the original NeRF model without incurring high storage costs. Further, using teacher-student distillation for training, we show that this speed-up can be achieved without sacrificing visual quality."
                },
                {
                    "title": "COIN: COmpression with Implicit Neural representations",
                    "abstract": "We propose a new simple approach for image compression: instead of storing the RGB values for each pixel of an image, we store the weights of a neural network overfitted to the image. Specifically, to encode an image, we fit it with an MLP which maps pixel locations to RGB values. We then quantize and store the weights of this MLP as a code for the image. To decode the image, we simply evaluate the MLP at every pixel location. We found that this simple approach outperforms JPEG at low bit-rates, even without entropy coding or learning a distribution over weights. While our framework is not yet competitive with state of the art compression methods, we show that it has various attractive properties which could make it a viable alternative to other neural data compression approaches."
                },
                {
                    "title": "Learning Continuous Image Representation with Local Implicit Image Function",
                    "abstract": "How to represent an image? While the visual world is presented in a continuous manner, machines store and see the images in a discrete way with 2D arrays of pixels. In this paper, we seek to learn a continuous representation for images. Inspired by the recent progress in 3D reconstruction with implicit neural representation, we propose Local Implicit Image Function (LIIF), which takes an image coordinate and the 2D deep features around the coordinate as inputs, predicts the RGB value at a given coordinate as an output. Since the coordinates are continuous, LIIF can be presented in arbitrary resolution. To generate the continuous representation for images, we train an encoder with LIIF representation via a self-supervised task with superresolution. The learned continuous representation can be presented in arbitrary resolution even extrapolate to \u00d730 higher resolution, where the training tasks are not provided. We further show that LIIF representation builds a bridge between discrete and continuous representation in 2D, it naturally supports the learning tasks with size-varied image ground-truths and significantly outperforms the method with resizing the ground-truths. Our project page with code is at https://yinboc.github.io/liif/."
                },
                {
                    "title": "iNeRF: Inverting Neural Radiance Fields for Pose Estimation",
                    "abstract": "We present iNeRF, a framework that performs mesh-free pose estimation by \"inverting\" a Neural Radiance Field (NeRF). NeRFs have been shown to be remarkably effective for the task of view synthesis \u2014 synthesizing photorealistic novel views of real-world scenes or objects. In this work, we investigate whether we can apply analysis-by-synthesis via NeRF for mesh-free, RGB-only 6DoF pose estimation \u2013 given an image, find the translation and rotation of a camera relative to a 3D object or scene. Our method assumes that no object mesh models are available during either training or test time. Starting from an initial pose estimate, we use gradient descent to minimize the residual between pixels rendered from a NeRF and pixels in an observed image. In our experiments, we first study 1) how to sample rays during pose refinement for iNeRF to collect informative gradients and 2) how different batch sizes of rays affect iNeRF on a synthetic dataset. We then show that for complex real-world scenes from the LLFF dataset [21], iNeRF can improve NeRF by estimating the camera poses of novel images and using these images as additional training data for NeRF. Finally, we show iNeRF can perform categorylevel object pose estimation, including object instances not seen during training, with RGB images by inverting a NeRF model inferred from a single view."
                },
                {
                    "title": "Implicit Neural Representations with Periodic Activation Functions",
                    "abstract": "Implicitly defined, continuous, differentiable signal representations parameterized by neural networks have emerged as a powerful paradigm, offering many possible benefits over conventional representations. However, current network architectures for such implicit neural representations are incapable of modeling signals with fine detail, and fail to represent a signal's spatial and temporal derivatives, despite the fact that these are essential to many physical signals defined implicitly as the solution to partial differential equations. We propose to leverage periodic activation functions for implicit neural representations and demonstrate that these networks, dubbed sinusoidal representation networks or Sirens, are ideally suited for representing complex natural signals and their derivatives. We analyze Siren activation statistics to propose a principled initialization scheme and demonstrate the representation of images, wavefields, video, sound, and their derivatives. Further, we show how Sirens can be leveraged to solve challenging boundary value problems, such as particular Eikonal equations (yielding signed distance functions), the Poisson equation, and the Helmholtz and wave equations. Lastly, we combine Sirens with hypernetworks to learn priors over the space of Siren functions."
                },
                {
                    "title": "Local Implicit Grid Representations for 3D Scenes",
                    "abstract": "Shape priors learned from data are commonly used to reconstruct 3D objects from partial or noisy data. Yet no such shape priors are available for indoor scenes, since typical 3D autoencoders cannot handle their scale, complexity, or diversity. In this paper, we introduce Local Implicit Grid Representations, a new 3D shape representation designed for scalability and generality. The motivating idea is that most 3D surfaces share geometric details at some scale -- i.e., at a scale smaller than an entire object and larger than a small patch. We train an autoencoder to learn an embedding of local crops of 3D shapes at that size. Then, we use the decoder as a component in a shape optimization that solves for a set of latent codes on a regular grid of overlapping crops such that an interpolation of the decoded local shapes matches a partial or noisy observation. We demonstrate the value of this proposed approach for 3D surface reconstruction from sparse point observations, showing significantly better results than alternative approaches."
                },
                {
                    "title": "Scene Representation Networks: Continuous 3D-Structure-Aware Neural Scene Representations",
                    "abstract": "Unsupervised learning with generative models has the potential of discovering rich representations of 3D scenes. While geometric deep learning has explored 3D-structure-aware representations of scene geometry, these models typically require explicit 3D supervision. Emerging neural scene representations can be trained only with posed 2D images, but existing methods ignore the three-dimensional structure of scenes. We propose Scene Representation Networks (SRNs), a continuous, 3D-structure-aware scene representation that encodes both geometry and appearance. SRNs represent scenes as continuous functions that map world coordinates to a feature representation of local scene properties. By formulating the image formation as a differentiable ray-marching algorithm, SRNs can be trained end-to-end from only 2D images and their camera poses, without access to depth or shape. This formulation naturally generalizes across scenes, learning powerful geometry and appearance priors in the process. We demonstrate the potential of SRNs by evaluating them for novel view synthesis, few-shot reconstruction, joint shape and appearance interpolation, and unsupervised discovery of a non-rigid face model."
                },
                {
                    "title": "DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation",
                    "abstract": "Computer graphics, 3D computer vision and robotics communities have produced multiple approaches to representing 3D geometry for rendering and reconstruction. These provide trade-offs across fidelity, efficiency and compression capabilities. In this work, we introduce DeepSDF, a learned continuous Signed Distance Function (SDF) representation of a class of shapes that enables high quality shape representation, interpolation and completion from partial and noisy 3D input data. DeepSDF, like its classical counterpart, represents a shape's surface by a continuous volumetric field: the magnitude of a point in the field represents the distance to the surface boundary and the sign indicates whether the region is inside (-) or outside (+) of the shape, hence our representation implicitly encodes a shape's boundary as the zero-level-set of the learned function while explicitly representing the classification of space as being part of the shapes interior or not. While classical SDF's both in analytical or discretized voxel form typically represent the surface of a single shape, DeepSDF can represent an entire class of shapes. Furthermore, we show state-of-the-art performance for learned 3D shape representation and completion while reducing the model size by an order of magnitude compared with previous work."
                },
                {
                    "title": "Occupancy Networks: Learning 3D Reconstruction in Function Space",
                    "abstract": "With the advent of deep neural networks, learning-based approaches for 3D reconstruction have gained popularity. However, unlike for images, in 3D there is no canonical representation which is both computationally and memory efficient yet allows for representing high-resolution geometry of arbitrary topology. Many of the state-of-the-art learning-based 3D reconstruction approaches can hence only represent very coarse 3D geometry or are limited to a restricted domain. In this paper, we propose Occupancy Networks, a new representation for learning-based 3D reconstruction methods. Occupancy networks implicitly represent the 3D surface as the continuous decision boundary of a deep neural network classifier. In contrast to existing approaches, our representation encodes a description of the 3D output at infinite resolution without excessive memory footprint. We validate that our representation can efficiently encode 3D structure and can be inferred from various kinds of input. Our experiments demonstrate competitive results, both qualitatively and quantitatively, for the challenging tasks of 3D reconstruction from single images, noisy point clouds and coarse discrete voxel grids. We believe that occupancy networks will become a useful tool in a wide variety of learning-based 3D tasks."
                },
                {
                    "title": "Learning Implicit Fields for Generative Shape Modeling",
                    "abstract": "We advocate the use of implicit fields for learning generative models of shapes and introduce an implicit field decoder, called IM-NET, for shape generation, aimed at improving the visual quality of the generated shapes. An implicit field assigns a value to each point in 3D space, so that a shape can be extracted as an iso-surface. IM-NET is trained to perform this assignment by means of a binary classifier. Specifically, it takes a point coordinate, along with a feature vector encoding a shape, and outputs a value which indicates whether the point is outside the shape or not. By replacing conventional decoders by our implicit decoder for representation learning (via IM-AE) and shape generation (via IM-GAN), we demonstrate superior results for tasks such as generative shape modeling, interpolation, and single-view 3D reconstruction, particularly in terms of visual quality. Code and supplementary material are available at https://github.com/czq142857/implicit-decoder."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "NTIRE 2017 Challenge on Single Image Super-Resolution: Dataset and Study",
                    "abstract": "This paper introduces a novel large dataset for example-based single image super-resolution and studies the state-of-the-art as emerged from the NTIRE 2017 challenge. The challenge is the first challenge of its kind, with 6 competitions, hundreds of participants and tens of proposed solutions. Our newly collected DIVerse 2K resolution image dataset (DIV2K) was employed by the challenge. In our study we compare the solutions from the challenge to a set of representative methods from the literature and evaluate them using diverse measures on our proposed DIV2K dataset. Moreover, we conduct a number of experiments and draw conclusions on several topics of interest. We conclude that the NTIRE 2017 challenge pushes the state-of-the-art in single-image super-resolution, reaching the best results to date on the popular Set5, Set14, B100, Urban100 datasets and on our newly proposed DIV2K."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "X-Fields: Implicit Neural View-, Light-and Time-Image Interpolation",
                    "abstract": ", 3D projection, occlusion) in a hard-coded and differentiable form. The NN represents the input to that rendering as an implicit map, that for any view, time, or light coordinate and for any pixel can quantify how it will move if view, time or light coordinates change (Jacobian of pixel position with respect to view, time, illumination, etc.). Our X-Field representation is trained for one scene within minutes, leading to a compact set of trainable parameters and hence real-time navigation in view, time and illumination."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively edit implicit neural representations (INRs) of signals, specifically through cropping and extending, without requiring extensive fine-tuning or compromising the integrity of the encoded information?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the field of machine learning, particularly in the area of signal processing and representation learning. By enabling efficient editing of INRs, researchers can enhance the flexibility and usability of neural networks in various applications, such as real-time rendering, data compression, and interactive media. This advancement could lead to new methodologies for handling high-dimensional data, fostering further research into adaptive and dynamic neural architectures that can respond to changing input requirements without the need for retraining from scratch.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent black-box nature of INRs, which complicates the modification of encoded signals post-training. Naive approaches, such as retraining a new INR or fine-tuning the existing one, are inefficient and may not preserve the original signal's integrity. Additionally, directly manipulating the weight space to achieve desired edits poses significant technical hurdles. The need to maintain a balance between the capacity of the model and the integrity of the encoded information adds further complexity, making it difficult to implement effective cropping and extending operations without introducing artifacts or increasing reconstruction errors.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the static nature of INRs, lacking methods for dynamic editing without retraining. Existing solutions have either required extensive fine-tuning or have not addressed the need for a global context in signal representation, leading to increased reconstruction errors. The partitioning approach, while promising, has not been fully explored in terms of its limitations regarding global context and reconstruction quality. Our approach differs by proposing a method that leverages an ensemble of compact INRs, allowing for efficient cropping and extending while addressing the shortcomings of prior work through knowledge distillation.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves partitioning the input signal space into granular segments, training compact INRs for each partition, and utilizing an ensemble approach for encoding and reconstructing signals. We will evaluate the performance using metrics such as reconstruction error and rendering speed on a dataset of natural signals. The expected outcomes include a significant reduction in reconstruction errors and improved efficiency in signal editing"
            }
        },
        "author_data": {
            "a9dda22a-87ce-4c09-8dc5-65aa48d466c7": {
                "pk": "a9dda22a-87ce-4c09-8dc5-65aa48d466c7",
                "name": "Maor Ashkenazi",
                "collaborators": [
                    "Eran Treister",
                    "Shahaf E. Finder",
                    "Yair Zohav",
                    "Zohar Rimon",
                    "Ron Vainshtein",
                    "Shir Levi",
                    "Elad Richardson",
                    "Pinchas Mintz"
                ],
                "domain": [
                    "Neural Networks",
                    "Image Processing",
                    "Model Compression",
                    "Quantization"
                ],
                "publications": [
                    {
                        "title": "Wavelet Feature Maps Compression for Image-to-Image CNNs",
                        "abstract": "Convolutional Neural Networks (CNNs) are known for requiring extensive computational resources, and quantization is among the best and most common methods for compressing them. While aggressive quantization (i.e., less than 4-bits) performs well for classification, it may cause severe performance degradation in image-to-image tasks such as semantic segmentation and depth estimation. In this paper, we propose Wavelet Compressed Convolution (WCC) -- a novel approach for high-resolution activation maps compression integrated with point-wise convolutions, which are the main computational cost of modern architectures. To this end, we use an efficient and hardware-friendly Haar-wavelet transform, known for its effectiveness in image compression, and define the convolution on the compressed activation map. We experiment with various tasks that benefit from high-resolution input. By combining WCC with light quantization, we achieve compression rates equivalent to 1-4bit activation quantization with relatively small and much more graceful degradation in performance. Our code is available at https://github.com/BGUCompSci/WaveletCompressedConvolution."
                    },
                    {
                        "title": "NeRN -- Learning Neural Representations for Neural Networks",
                        "abstract": "Neural Representations have recently been shown to effectively reconstruct a wide range of signals from 3D meshes and shapes to images and videos. We show that, when adapted correctly, neural representations can be used to directly represent the weights of a pre-trained convolutional neural network, resulting in a Neural Representation for Neural Networks (NeRN). Inspired by coordinate inputs of previous neural representation methods, we assign a coordinate to each convolutional kernel in our network based on its position in the architecture, and optimize a predictor network to map coordinates to their corresponding weights. Similarly to the spatial smoothness of visual scenes, we show that incorporating a smoothness constraint over the original network's weights aids NeRN towards a better reconstruction. In addition, since slight perturbations in pre-trained model weights can result in a considerable accuracy loss, we employ techniques from the field of knowledge distillation to stabilize the learning process. We demonstrate the effectiveness of NeRN in reconstructing widely used architectures on CIFAR-10, CIFAR-100, and ImageNet. Finally, we present two applications using NeRN, demonstrating the capabilities of the learned representations."
                    }
                ]
            },
            "47eb169c-d16f-4240-a8af-ec42a9e925a6": {
                "pk": "47eb169c-d16f-4240-a8af-ec42a9e925a6",
                "name": "Eran Treister",
                "collaborators": [
                    "Eldad Haber",
                    "Moshe Eliasof",
                    "Lars Ruthotto",
                    "Rachel Yovel",
                    "Jonathan Ephrath",
                    "Yael Azulay",
                    "Sagi Buchatsky",
                    "Matan Goren",
                    "Andrei Sharf",
                    "Gal Shalom"
                ],
                "domain": [
                    "Inverse Problems",
                    "Graph Neural Networks",
                    "Convolutional Neural Networks",
                    "Numerical Methods"
                ],
                "publications": [
                    {
                        "title": "Multigrid-augmented deep learning preconditioners for the Helmholtz equation",
                        "abstract": "In this paper, we present a data-driven approach to iteratively solve the discrete heterogeneous Helmholtz equation at high wavenumbers. In our approach, we combine classical iterative solvers with convolutional neural networks (CNNs) to form a preconditioner which is applied within a Krylov solver. For the preconditioner, we use a CNN of type U-Net that operates in conjunction with multigrid ingredients. Two types of preconditioners are proposed 1) U-Net as a coarse grid solver, and 2) U-Net as a deflation operator with shifted Laplacian V-cycles. Following our training scheme and data-augmentation, our CNN preconditioner can generalize over residuals and a relatively general set of wave slowness models. On top of that, we also offer an encoder-solver framework where an \"encoder\" network generalizes over the medium and sends context vectors to another \"solver\" network, which generalizes over the right-hand-sides. We show that this option is more robust and efficient than the stand-alone variant. Lastly, we also offer a mini-retraining procedure, to improve the solver after the model is known. This option is beneficial when solving multiple right-hand-sides, like in inverse problems. We demonstrate the efficiency and generalization abilities of our approach on a variety of 2D problems."
                    },
                    {
                        "title": "Full waveform inversion guided by travel time tomography",
                        "abstract": "Full waveform inversion (FWI) is a process in which seismic numerical simulations are fit to observed data by changing the wave velocity model of the medium under investigation. The problem is non-linear, and therefore optimization techniques have been used to find a reasonable solution to the problem. The main problem in fitting the data is the lack of low spatial frequencies. This deficiency often leads to a local minimum and to non-plausible solutions. In this work we explore how to obtain low frequency information for FWI. Our approach involves augmenting FWI with travel time tomography, which has low-frequency features. By jointly inverting these two problems we enrich FWI with information that can replace low frequency data. In addition, we use high order regularization, in a preliminary inversion stage, to prevent high frequency features from polluting our model in the initial stages of the reconstruction. This regularization also promotes the non-dominant low-frequency modes that exist in the FWI sensitivity. By applying a joint FWI and travel time inversion we are able to obtain a smooth model than can later be used to recover a good approximation for the true model. A second contribution of this paper involves the acceleration of the main computational bottleneck in FWI--the solution of the Helmholtz equation. We show that the solution time can be reduced by solving the equation for multiple right hand sides using block multigrid preconditioned Krylov methods."
                    },
                    {
                        "title": "A multigrid solver to the Helmholtz equation with a point source based on travel time and amplitude",
                        "abstract": "The Helmholtz equation arises when modeling wave propagation in the frequency domain. The equation is discretized as an indefinite linear system, which is difficult to solve at high wave numbers. In many applications, the solution of the Helmholtz equation is required for a point source. In this case, it is possible to reformulate the equation as two separate equations: one for the travel time of the wave and one for its amplitude. The travel time is obtained by a solution of the factored eikonal equation, and the amplitude is obtained by solving a complex-valued advection-diffusion-reaction (ADR) equation. The reformulated equation is equivalent to the original Helmholtz equation, and the differences between the numerical solutions of these equations arise only from discretization errors. We develop an efficient multigrid solver for obtaining the amplitude given the travel time, which can be efficiently computed. This approach is advantageous because the amplitude is typically smooth in this case, and hence, more suitable for multigrid solvers than the standard Helmholtz discretization. We demonstrate that our second order ADR discretization is more accurate than the standard second order discretization at high wave numbers, as long as there are no reflections or caustics. Moreover, we show that using our approach, the problem can be solved more efficiently than using the common shifted Laplacian multigrid approach."
                    },
                    {
                        "title": "A hybrid shifted Laplacian multigrid and domain decomposition preconditioner for the elastic Helmholtz equations",
                        "abstract": "In this work we extend the shifted Laplacian approach to the elastic Helmholtz equation. The shifted Laplacian multigrid method is a common preconditioning approach for the discretized acoustic Helmholtz equation. In some cases, like geophysical seismic imaging, one needs to consider the elastic Helmholtz equation, which is harder to solve: it is three times larger and contains a nullity-rich grad-div term. These properties make the solution of the equation more difficult for multigrid solvers. The key idea in this work is combining the shifted Laplacian with approaches for linear elasticity. We provide local Fourier analysis and numerical evidence that the convergence rate of our method is independent of the Poisson's ratio. Moreover, to better handle the problem size, we complement our multigrid method with the domain decomposition approach, which works in synergy with the local nature of the shifted Laplacian, so we enjoy the advantages of both methods without sacrificing performance. We demonstrate the efficiency of our solver on 2D and 3D problems in heterogeneous media."
                    },
                    {
                        "title": "A fast marching algorithm for the factored eikonal equation",
                        "abstract": "The eikonal equation is instrumental in many applications in several fields ranging from computer vision to geoscience. This equation can be efficiently solved using the iterative Fast Sweeping (FS) methods and the direct Fast Marching (FM) methods. However, when used for a point source, the original eikonal equation is known to yield inaccurate numerical solutions, because of a singularity at the source. In this case, the factored eikonal equation is often preferred, and is known to yield a more accurate numerical solution. One application that requires the solution of the eikonal equation for point sources is travel time tomography. This inverse problem may be formulated using the eikonal equation as a forward problem. While this problem has been solved using FS in the past, the more recent choice for applying it involves FM methods because of the efficiency in which sensitivities can be obtained using them. However, while several FS methods are available for solving the factored equation, the FM method is available only for the original eikonal equation.   In this paper we develop a Fast Marching algorithm for the factored eikonal equation, using both first and second order finite-difference schemes. Our algorithm follows the same lines as the original FM algorithm and requires the same computational effort. In addition, we show how to obtain sensitivities using this FM method and apply travel time tomography, formulated as an inverse factored eikonal equation. Numerical results in two and three dimensions show that our algorithm solves the factored eikonal equation efficiently, and demonstrate the achieved accuracy for computing the travel time. We also demonstrate a recovery of a 2D and 3D heterogeneous medium by travel time tomography using the eikonal equation for forward modelling and inversion by Gauss-Newton."
                    },
                    {
                        "title": "Full waveform inversion using extended and simultaneous sources",
                        "abstract": "PDE-constrained optimization problems are often treated using the reduced formulation where the PDE constraints are eliminated. This approach is known to be more computationally feasible than other alternatives at large scales. However, the elimination of the constraints forces the optimization process to fulfill the constraints at all times. In some problems this may lead to a highly non-linear objective, which is hard to solve. An example to such a problem, which we focus on in this work, is Full Waveform Inversion (FWI), which appears in seismic exploration of oil and gas reservoirs, and medical imaging. In an attempt to relieve the non-linearity of FWI, several approaches suggested to expand the optimization search space and relax the PDE constraints. This comes, however, with severe memory and computational costs, which we aim to reduce. In this work we adopt the expanded search space approach, and suggest a new formulation of FWI using extended source functions. To make the source-extended problem more feasible in memory and computations, we couple the source extensions in the form of a low-rank matrix. This way, we have a large-but-manageable additional parameter space, which has a rather low memory footprint, and is much more suitable for solving large scale instances of the problem than the full rank additional space. In addition, we show how our source-extended approach is applied together with the popular simultaneous sources technique -- a stochastic optimization technique that significantly reduces the computations needed for FWI inversions. We demonstrate our approaches for solving FWI problems using 2D and 3D models with high frequency data only."
                    },
                    {
                        "title": "DiffGCN: Graph Convolutional Networks via Differential Operators and Algebraic Multigrid Pooling",
                        "abstract": "Graph Convolutional Networks (GCNs) have shown to be effective in handling unordered data like point clouds and meshes. In this work we propose novel approaches for graph convolution, pooling and unpooling, inspired from finite differences and algebraic multigrid frameworks. We form a parameterized convolution kernel based on discretized differential operators, leveraging the graph mass, gradient and Laplacian. This way, the parameterization does not depend on the graph structure, only on the meaning of the network convolutions as differential operators. To allow hierarchical representations of the input, we propose pooling and unpooling operations that are based on algebraic multigrid methods, which are mainly used to solve partial differential equations on unstructured grids. To motivate and explain our method, we compare it to standard convolutional neural networks, and show their similarities and relations in the case of a regular grid. Our proposed method is demonstrated in various experiments like classification and part-segmentation, achieving on par or better than state of the art results. We also analyze the computational cost of our method compared to other GCNs."
                    },
                    {
                        "title": "LFA-tuned matrix-free multigrid method for the elastic Helmholtz equation",
                        "abstract": "We present an efficient matrix-free geometric multigrid method for the elastic Helmholtz equation, and a suitable discretization. Many discretization methods had been considered in the literature for the Helmholtz equations, as well as many solvers and preconditioners, some of which are adapted for the elastic version of the equation. However, there is very little work considering the reciprocity of discretization and a solver. In this work, we aim to bridge this gap. By choosing an appropriate stencil for re-discretization of the equation on the coarse grid, we develop a multigrid method that can be easily implemented as matrix-free, relying on stencils rather than sparse matrices. This is crucial for efficient implementation on modern hardware. Using two-grid local Fourier analysis, we validate the compatibility of our discretization with our solver, and tune a choice of weights for the stencil for which the convergence rate of the multigrid cycle is optimal. It results in a scalable multigrid preconditioner that can tackle large real-world 3D scenarios."
                    },
                    {
                        "title": "Physics-guided Full Waveform Inversion using Encoder-Solver Convolutional Neural Networks",
                        "abstract": "Full Waveform Inversion (FWI) is an inverse problem for estimating the wave velocity distribution in a given domain, based on observed data on the boundaries. The inversion is computationally demanding because we are required to solve multiple forward problems, either in time or frequency domains, to simulate data that are then iteratively fitted to the observed data. We consider FWI in the frequency domain, where the Helmholtz equation is used as a forward model, and its repeated solution is the main computational bottleneck of the inversion process. To ease this cost, we integrate a learning process of an encoder-solver preconditioner that is based on convolutional neural networks (CNNs). The encoder-solver is trained to effectively precondition the discretized Helmholtz operator given velocity medium parameters. Then, by re-training the CNN between the iterations of the optimization process, the encoder-solver is adapted to the iteratively evolving velocity medium as part of the inversion. Without retraining, the performance of the solver deteriorates as the medium changes. Using our light retraining procedures, we obtain the forward simulations effectively throughout the process. We demonstrate our approach to solving FWI problems using 2D geophysical models with high-frequency data."
                    },
                    {
                        "title": "Global-Local Graph Neural Networks for Node-Classification",
                        "abstract": "The task of graph node classification is often approached by utilizing a local Graph Neural Network (GNN), that learns only local information from the node input features and their adjacency. In this paper, we propose to improve the performance of node classification GNNs by utilizing both global and local information, specifically by learning label- and node- features. We therefore call our method Global-Local-GNN (GLGNN). To learn proper label features, for each label, we maximize the similarity between its features and nodes features that belong to the label, while maximizing the distance between nodes that do not belong to the considered label. We then use the learnt label features to predict the node classification map. We demonstrate our GLGNN using three different GNN backbones, and show that our approach improves baseline performance, revealing the importance of global information utilization for node classification."
                    },
                    {
                        "title": "jInv -- a flexible Julia package for PDE parameter estimation",
                        "abstract": "Estimating parameters of Partial Differential Equations (PDEs) from noisy and indirect measurements often requires solving ill-posed inverse problems. These so called parameter estimation or inverse medium problems arise in a variety of applications such as geophysical, medical imaging, and nondestructive testing. Their solution is computationally intense since the underlying PDEs need to be solved numerous times until the reconstruction of the parameters is sufficiently accurate. Typically, the computational demand grows significantly when more measurements are available, which poses severe challenges to inversion algorithms as measurement devices become more powerful.   In this paper we present jInv, a flexible framework and open source software that provides parallel algorithms for solving parameter estimation problems with many measurements. Being written in the expressive programming language Julia, jInv is portable, easy to understand and extend, cross-platform tested, and well-documented. It provides novel parallelization schemes that exploit the inherent structure of many parameter estimation problems and can be used to solve multiphysics inversion problems as is demonstrated using numerical experiments motivated by geophysical imaging."
                    },
                    {
                        "title": "LeanResNet: A Low-cost Yet Effective Convolutional Residual Networks",
                        "abstract": "Convolutional Neural Networks (CNNs) filter the input data using spatial convolution operators with compact stencils. Commonly, the convolution operators couple features from all channels, which leads to immense computational cost in the training of and prediction with CNNs. To improve the efficiency of CNNs, we introduce lean convolution operators that reduce the number of parameters and computational complexity, and can be used in a wide range of existing CNNs. Here, we exemplify their use in residual networks (ResNets), which have been very reliable for a few years now and analyzed intensively. In our experiments on three image classification problems, the proposed LeanResNet yields results that are comparable to other recently proposed reduced architectures using similar number of parameters."
                    },
                    {
                        "title": "Multi-modal 3D Shape Reconstruction Under Calibration Uncertainty using Parametric Level Set Methods",
                        "abstract": "We consider the problem of 3D shape reconstruction from multi-modal data, given uncertain calibration parameters. Typically, 3D data modalities can be in diverse forms such as sparse point sets, volumetric slices, 2D photos and so on. To jointly process these data modalities, we exploit a parametric level set method that utilizes ellipsoidal radial basis functions. This method not only allows us to analytically and compactly represent the object, it also confers on us the ability to overcome calibration related noise that originates from inaccurate acquisition parameters. This essentially implicit regularization leads to a highly robust and scalable reconstruction, surpassing other traditional methods. In our results we first demonstrate the ability of the method to compactly represent complex objects. We then show that our reconstruction method is robust both to a small number of measurements and to noise in the acquisition parameters. Finally, we demonstrate our reconstruction abilities from diverse modalities such as volume slices obtained from liquid displacement (similar to CTscans and XRays), and visual measurements obtained from shape silhouettes."
                    },
                    {
                        "title": "pISTA: preconditioned Iterative Soft Thresholding Algorithm for Graphical Lasso",
                        "abstract": "We propose a novel quasi-Newton method for solving the sparse inverse covariance estimation problem also known as the graphical least absolute shrinkage and selection operator (GLASSO). This problem is often solved using a second-order quadratic approximation. However, in such algorithms the Hessian term is complex and computationally expensive to handle. Therefore, our method uses the inverse of the Hessian as a preconditioner to simplify and approximate the quadratic element at the cost of a more complex \\(\\ell_1\\) element. The variables of the resulting preconditioned problem are coupled only by the \\(\\ell_1\\) sub-derivative of each other, which can be guessed with minimal cost using the gradient itself, allowing the algorithm to be parallelized and implemented efficiently on GPU hardware accelerators. Numerical results on synthetic and real data demonstrate that our method is competitive with other state-of-the-art approaches."
                    },
                    {
                        "title": "Improving Graph Neural Networks with Learnable Propagation Operators",
                        "abstract": "Graph Neural Networks (GNNs) are limited in their propagation operators. In many cases, these operators often contain non-negative elements only and are shared across channels, limiting the expressiveness of GNNs. Moreover, some GNNs suffer from over-smoothing, limiting their depth. On the other hand, Convolutional Neural Networks (CNNs) can learn diverse propagation filters, and phenomena like over-smoothing are typically not apparent in CNNs. In this paper, we bridge these gaps by incorporating trainable channel-wise weighting factors $\\omega$ to learn and mix multiple smoothing and sharpening propagation operators at each layer. Our generic method is called $\\omega$GNN, and is easy to implement. We study two variants: $\\omega$GCN and $\\omega$GAT. For $\\omega$GCN, we theoretically analyse its behaviour and the impact of $\\omega$ on the obtained node features. Our experiments confirm these findings, demonstrating and explaining how both variants do not over-smooth. Additionally, we experiment with 15 real-world datasets on node- and graph-classification tasks, where our $\\omega$GCN and $\\omega$GAT perform on par with state-of-the-art methods."
                    },
                    {
                        "title": "Efficient Graph Laplacian Estimation by Proximal Newton",
                        "abstract": "The Laplacian-constrained Gaussian Markov Random Field (LGMRF) is a common multivariate statistical model for learning a weighted sparse dependency graph from given data. This graph learning problem can be formulated as a maximum likelihood estimation (MLE) of the precision matrix, subject to Laplacian structural constraints, with a sparsity-inducing penalty term. This paper aims to solve this learning problem accurately and efficiently. First, since the commonly used $\\ell_1$-norm penalty is inappropriate in this setting and may lead to a complete graph, we employ the nonconvex minimax concave penalty (MCP), which promotes sparse solutions with lower estimation bias. Second, as opposed to existing first-order methods for this problem, we develop a second-order proximal Newton approach to obtain an efficient solver, utilizing several algorithmic features, such as using Conjugate Gradients, preconditioning, and splitting to active/free sets. Numerical experiments demonstrate the advantages of the proposed method in terms of both computational complexity and graph learning accuracy compared to existing methods."
                    },
                    {
                        "title": "An Over Complete Deep Learning Method for Inverse Problems",
                        "abstract": "Obtaining meaningful solutions for inverse problems has been a major challenge with many applications in science and engineering. Recent machine learning techniques based on proximal and diffusion-based methods have shown promising results. However, as we show in this work, they can also face challenges when applied to some exemplary problems. We show that similar to previous works on over-complete dictionaries, it is possible to overcome these shortcomings by embedding the solution into higher dimensions. The novelty of the work proposed is that we jointly design and learn the embedding and the regularizer for the embedding vector. We demonstrate the merit of this approach on several exemplary and common inverse problems."
                    },
                    {
                        "title": "IMEXnet: A Forward Stable Deep Neural Network",
                        "abstract": "Deep convolutional neural networks have revolutionized many machine learning and computer vision tasks, however, some remaining key challenges limit their wider use. These challenges include improving the network's robustness to perturbations of the input image and the limited ``field of view'' of convolution operators. We introduce the IMEXnet that addresses these challenges by adapting semi-implicit methods for partial differential equations. Compared to similar explicit networks, such as residual networks, our network is more stable, which has recently shown to reduce the sensitivity to small changes in the input features and improve generalization. The addition of an implicit step connects all pixels in each channel of the image and therefore addresses the field of view problem while still being comparable to standard convolutions in terms of the number of parameters and computational complexity. We also present a new dataset for semantic segmentation and demonstrate the effectiveness of our architecture using the NYU Depth dataset."
                    },
                    {
                        "title": "Multigrid-in-Channels Architectures for Wide Convolutional Neural Networks",
                        "abstract": "We present a multigrid approach that combats the quadratic growth of the number of parameters with respect to the number of channels in standard convolutional neural networks (CNNs). It has been shown that there is a redundancy in standard CNNs, as networks with much sparser convolution operators can yield similar performance to full networks. The sparsity patterns that lead to such behavior, however, are typically random, hampering hardware efficiency. In this work, we present a multigrid-in-channels approach for building CNN architectures that achieves full coupling of the channels, and whose number of parameters is linearly proportional to the width of the network. To this end, we replace each convolution layer in a generic CNN with a multilevel layer consisting of structured (i.e., grouped) convolutions. Our examples from supervised image classification show that applying this strategy to residual networks and MobileNetV2 considerably reduces the number of parameters without negatively affecting accuracy. Therefore, we can widen networks without dramatically increasing the number of parameters or operations."
                    }
                ]
            }
        }
    },
    "2409.04095": {
        "paper_data": {
            "title": "UNIT: Unifying Image and Text Recognition in One Vision Encoder",
            "url": "http://arxiv.org/abs/2409.04095v1",
            "arxiv_id": "2409.04095",
            "authors": [
                "Yi Zhu",
                "Yanpeng Zhou",
                "Chunwei Wang",
                "Yang Cao",
                "Jianhua Han",
                "Lu Hou",
                "Hang Xu"
            ],
            "abstract": "Currently, vision encoder models like Vision Transformers (ViTs) typically excel at image recognition tasks but cannot simultaneously support text recognition like human visual recognition. To address this limitation, we propose UNIT, a novel training framework aimed at UNifying Image and Text recognition within a single model. Starting with a vision encoder pre-trained with image recognition tasks, UNIT introduces a lightweight language decoder for predicting text outputs and a lightweight vision decoder to prevent catastrophic forgetting of the original image encoding capabilities. The training process comprises two stages: intra-scale pretraining and inter-scale finetuning. During intra-scale pretraining, UNIT learns unified representations from multi-scale inputs, where images and documents are at their commonly used resolution, to enable fundamental recognition capability. In the inter-scale finetuning stage, the model introduces scale-exchanged data, featuring images and documents at resolutions different from the most commonly used ones, to enhance its scale robustness. Notably, UNIT retains the original vision encoder architecture, making it cost-free in terms of inference and deployment. Experiments across multiple benchmarks confirm that our method significantly outperforms existing methods on document-related tasks (e.g., OCR and DocQA) while maintaining the performances on natural images, demonstrating its ability to substantially enhance text recognition without compromising its core image recognition capabilities.",
            "introduction": "   1 Introduction  Vision encoder models\u00a0[6, 43, 44, 53, 41, 24] are crucial for extracting high-level visual features, thereby enhancing performance across a range of downstream tasks\u00a0[9, 38, 8, 14, 57]. They play a vital role in integrating visual information into intelligent applications, driving advancements in both computer vision and artificial intelligence. In recent years, Vision Transformers\u00a0(ViTs)based encoder models\u00a0[16, 66, 27, 58, 71] have revolutionized the field of computer vision. ViT models pretrained on image classification tasks are extensively used as backbones for a wide array of image recognition tasks and achieves state-of-the-art performances. Another line of remarkable ViT models are trained via image-text cross-modal contrastive learning\u00a0[43, 64, 52, 67]. These models are often used as plug-in vision encoders in Large-scale Vision-Language Models (LVLMs), which have emerged as versatile tools with the potential to revolutionize various domains, including healthcare diagnostics, autonomous driving, digital assistants, and advanced content analysis in media and entertainment. Despite these advancements, existing vision encoder models, exhibit a significant limitation: they have not yet demonstrated the capability to simultaneously support both image and text recognition like humans do. Such ability is essential for document analysis applications, for instance, a model must accurately recognize and interpret textual information embedded within complex layouts, such as tables, graphs, and mixed media.   Text recognition\u00a0[50, 49, 12], particularly in the context of dense documents and complex visual environments, presents unique challenges that are distinct from those encountered in pure image recognition. While image recognition typically focuses on global feature extraction and classification, text recognition requires precise local feature extraction and sequence prediction. ViTs, though adept at handling global image features, often struggle with the detailed local features needed for accurate text recognition, particularly in high-resolution document images where fine details are crucial. A straightforward solution is to finetune pre-trained ViTs using high-resolution documents. However, this approach requires interpolating the pre-trained positional embeddings to handle longer sequences. Such interpolation with a large magnification will disrupt the alignment between the original positional embeddings and their corresponding spatial positions, thereby impacting the model\u2019s performance.   Existing methods\u00a0[39, 4, 5] build document-specific models by retraining ViTs exclusively on the Optical Character Recognition (OCR) task, thereby discarding the original image encoding capability. In downstream applications, these models are often ensembled with other expert models, requiring users to specify the data type in advance, which is inadequate for scenarios requiring dynamic and simultaneous recognition of both image and text without prior knowledge of the input content. Other ensemble-based methods\u00a0[53, 54] use similar model architectures trained independently on image and text recognition tasks. Inputs are fed into both models, and the output features are concatenated to form a unified representation. However, this approach will result in a significant increase in computational cost. Additionally, some LVLM methods\u00a0[62, 63] enhance document analysis tasks (e.g., DocQA) in an OCR-free manner by finetuning with document-instruction data. Nevertheless, their capabilities are limited to handling images with prominent text and fail with dense documents, struggling to generalize across different font sizes, typefaces, and backgrounds.   In this paper, we propose UNIT, a novel training framework for UNifying Image and Text recognition abilities within",
            "references": [
                {
                    "title": "MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training",
                    "abstract": "In this work, we discuss building performant Multimodal Large Language Models (MLLMs). In particular, we study the importance of various architecture components and data choices. Through careful and comprehensive ablations of the image encoder, the vision language connector, and various pre-training data choices, we identified several crucial design lessons. For example, we demonstrate that for large-scale multimodal pre-training using a careful mix of image-caption, interleaved image-text, and text-only data is crucial for achieving state-of-the-art (SOTA) few-shot results across multiple benchmarks, compared to other published pre-training results. Further, we show that the image encoder together with image resolution and the image token count has substantial impact, while the vision-language connector design is of comparatively negligible importance. By scaling up the presented recipe, we build MM1, a family of multimodal models up to 30B parameters, including both dense models and mixture-of-experts (MoE) variants, that are SOTA in pre-training metrics and achieve competitive performance after supervised fine-tuning on a range of established multimodal benchmarks. Thanks to large-scale pre-training, MM1 enjoys appealing properties such as enhanced in-context learning, and multi-image reasoning, enabling few-shot chain-of-thought prompting."
                },
                {
                    "title": "Small Language Model Meets with Reinforced Vision Vocabulary",
                    "abstract": "Playing Large Vision Language Models (LVLMs) in 2023 is trendy among the AI community. However, the relatively large number of parameters (more than 7B) of popular LVLMs makes it difficult to train and deploy on consumer GPUs, discouraging many researchers with limited resources. Imagine how cool it would be to experience all the features of current LVLMs on an old GTX1080ti (our only game card). Accordingly, we present Vary-toy in this report, a small-size Vary along with Qwen-1.8B as the base ``large'' language model. In Vary-toy, we introduce an improved vision vocabulary, allowing the model to not only possess all features of Vary but also gather more generality. Specifically, we replace negative samples of natural images with positive sample data driven by object detection in the procedure of generating vision vocabulary, more sufficiently utilizing the capacity of the vocabulary network and enabling it to efficiently encode visual information corresponding to natural objects. For experiments, Vary-toy can achieve 65.6% ANLS on DocVQA, 59.1% accuracy on ChartQA, 88.1% accuracy on RefCOCO, and 29% on MMVet. The code will be publicly available on the homepage."
                },
                {
                    "title": "Vary: Scaling up the Vision Vocabulary for Large Vision-Language Models",
                    "abstract": "Modern Large Vision-Language Models (LVLMs) enjoy the same vision vocabulary -- CLIP, which can cover most common vision tasks. However, for some special vision task that needs dense and fine-grained vision perception, e.g., document-level OCR or chart understanding, especially in non-English scenarios, the CLIP-style vocabulary may encounter low efficiency in tokenizing the vision knowledge and even suffer out-of-vocabulary problem. Accordingly, we propose Vary, an efficient and effective method to scale up the vision vocabulary of LVLMs. The procedures of Vary are naturally divided into two folds: the generation and integration of a new vision vocabulary. In the first phase, we devise a vocabulary network along with a tiny decoder-only transformer to produce the desired vocabulary via autoregression. In the next, we scale up the vanilla vision vocabulary by merging the new one with the original one (CLIP), enabling the LVLMs can quickly garner new features. Compared to the popular BLIP-2, MiniGPT4, and LLaVA, Vary can maintain its vanilla capabilities while enjoying more excellent fine-grained perception and understanding ability. Specifically, Vary is competent in new document parsing features (OCR or markdown conversion) while achieving 78.2% ANLS in DocVQA and 36.2% in MMVet. Our code will be publicly available on the homepage."
                },
                {
                    "title": "AM-RADIO: Agglomerative Vision Foundation Model Reduce All Domains Into One",
                    "abstract": "A handful of visual foundation models (VFMs) have recently emerged as the backbones for numerous downstream tasks. VFMs like CLIP, DINOv2, SAM are trained with distinct objectives, exhibiting unique characteristics for various downstream tasks. We find that despite their conceptual differences, these models can be effectively merged into a unified model through multi-teacher distillation. We name this approach AM-RADIO (Agglomerative Model - Reduce All Domains Into One). This integrative approach not only surpasses the performance of individual teacher models but also amalgamates their distinctive features, such as zero-shot vision-language comprehension, detailed pixel-level understanding, and open vocabulary segmentation capabilities. Additionally, in pursuit of the most hardware-efficient backbone, we evaluated numerous architectures in our multi-teacher distillation pipeline using the same training recipe. This led to the development of a novel architecture (E-RADIO) that exceeds the performance of its predecessors and is at least 6x faster than the teacher models at matched resolution. Our comprehensive benchmarking process covers downstream tasks including ImageNet classification, semantic segmentation linear probing, COCO object detection and integration into LLaVa-1.5. Code: https://github.com/NVlabs/RADIO."
                },
                {
                    "title": "ShareGPT4V: Improving Large Multi-Modal Models with Better Captions",
                    "abstract": "In the realm of large multi-modal models (LMMs), efficient modality alignment is crucial yet often constrained by the scarcity of high-quality image-text data. To address this bottleneck, we introduce the ShareGPT4V dataset, a pioneering large-scale resource featuring 1.2 million highly descriptive captions, which surpasses existing datasets in diversity and information content, covering world knowledge, object properties, spatial relationships, and aesthetic evaluations. Specifically, ShareGPT4V originates from a curated 100K high-quality captions collected from advanced GPT4-Vision and has been expanded to 1.2M with a superb caption model trained on this subset. ShareGPT4V first demonstrates its effectiveness for the Supervised Fine-Tuning (SFT) phase, by substituting an equivalent quantity of detailed captions in existing SFT datasets with a subset of our high-quality captions, significantly enhancing the LMMs like LLaVA-7B, LLaVA-1.5-13B, and Qwen-VL-Chat-7B on the MME and MMBench benchmarks, with respective gains of 222.8/22.0/22.3 and 2.7/1.3/1.5. We further incorporate ShareGPT4V data into both the pre-training and SFT phases, obtaining ShareGPT4V-7B, a superior LMM based on a simple architecture that has remarkable performance across a majority of the multi-modal benchmarks. This project is available at https://ShareGPT4V.github.io to serve as a pivotal resource for advancing the LMMs community."
                },
                {
                    "title": "mPLUG-OwI2: Revolutionizing Multi-modal Large Language Model with Modality Collaboration",
                    "abstract": "Multi-modal Large Language Models (MLLMs) have demonstrated impressive instruction abilities across various open-ended tasks. However, previous methods primarily fo-cus on enhancing multi-modal capabilities. In this work, we introduce a versatile multi-modal large language model, mPLUG-Owl2, which effectively leverages modality collab-oration to improve performance in both text and multi-modal tasks. mPLUG-Owl2 utilizes a modularized network design, with the language decoder acting as a universal interface for managing different modalities. Specifically, mPLUG-Owl2 incorporates shared functional modules to facilitate modal-ity collaboration and introduces a modality-adaptive module that preserves modality-specific features. Extensive experi-ments reveal that mPLUG-Owl2 is capable of generalizing both text tasks and multi-modal tasks and achieving state-of-the-art performances with a single generic model. Notably, mPLUG-Owl2 is the first MLLM model that demonstrates the modality collaboration phenomenon in both pure-text and multi-modal scenarios, setting a pioneering path in the development of future multi-modal foundation models."
                },
                {
                    "title": "CoDA: Collaborative Novel Box Discovery and Cross-modal Alignment for Open-vocabulary 3D Object Detection",
                    "abstract": "Open-vocabulary 3D Object Detection (OV-3DDet) aims to detect objects from an arbitrary list of categories within a 3D scene, which remains seldom explored in the literature. There are primarily two fundamental problems in OV-3DDet, i.e., localizing and classifying novel objects. This paper aims at addressing the two problems simultaneously via a unified framework, under the condition of limited base categories. To localize novel 3D objects, we propose an effective 3D Novel Object Discovery strategy, which utilizes both the 3D box geometry priors and 2D semantic open-vocabulary priors to generate pseudo box labels of the novel objects. To classify novel object boxes, we further develop a cross-modal alignment module based on discovered novel boxes, to align feature spaces between 3D point cloud and image/text modalities. Specifically, the alignment process contains a class-agnostic and a class-discriminative alignment, incorporating not only the base objects with annotations but also the increasingly discovered novel objects, resulting in an iteratively enhanced alignment. The novel box discovery and crossmodal alignment are jointly learned to collaboratively benefit each other. The novel object discovery can directly impact the cross-modal alignment, while a better feature alignment can, in turn, boost the localization capability, leading to a unified OV-3DDet framework, named CoDA, for simultaneous novel object localization and classification. Extensive experiments on two challenging datasets (i.e., SUN-RGBD and ScanNet) demonstrate the effectiveness of our method and also show a significant mAP improvement upon the best-performing alternative method by 80%. Codes and pre-trained models are released on the project page."
                },
                {
                    "title": "Qwen Technical Report",
                    "abstract": "Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural language processing tasks that were previously thought to be exclusive to humans. In this work, we introduce Qwen, the first installment of our large language model series. Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts. It includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques. The base language models consistently demonstrate superior performance across a multitude of downstream tasks, and the chat models, particularly those trained using Reinforcement Learning from Human Feedback (RLHF), are highly competitive. The chat models possess advanced tool-use and planning capabilities for creating agent applications, showcasing impressive performance even when compared to bigger models on complex tasks like utilizing a code interpreter. Furthermore, we have developed coding-specialized models, Code-Qwen and Code-Qwen-Chat, as well as mathematics-focused models, Math-Qwen-Chat, which are built upon base language models. These models demonstrate significantly improved performance in comparison with open-source models, and slightly fall behind the proprietary models."
                },
                {
                    "title": "Nougat: Neural Optical Understanding for Academic Documents",
                    "abstract": "Scientific knowledge is predominantly stored in books and scientific journals, often in the form of PDFs. However, the PDF format leads to a loss of semantic information, particularly for mathematical expressions. We propose Nougat (Neural Optical Understanding for Academic Documents), a Visual Transformer model that performs an Optical Character Recognition (OCR) task for processing scientific documents into a markup language, and demonstrate the effectiveness of our model on a new dataset of scientific documents. The proposed approach offers a promising solution to enhance the accessibility of scientific knowledge in the digital age, by bridging the gap between human-readable documents and machine-readable text. We release the models and code to accelerate future work on scientific text recognition."
                },
                {
                    "title": "mPLUG-DocOwl: Modularized Multimodal Large Language Model for Document Understanding",
                    "abstract": "Document understanding refers to automatically extract, analyze and comprehend information from various types of digital documents, such as a web page. Existing Multi-model Large Language Models (MLLMs), including mPLUG-Owl, have demonstrated promising zero-shot capabilities in shallow OCR-free text recognition, indicating their potential for OCR-free document understanding. Nevertheless, without in-domain training, these models tend to ignore fine-grained OCR features, such as sophisticated tables or large blocks of text, which are essential for OCR-free document understanding. In this paper, we propose mPLUG-DocOwl based on mPLUG-Owl for OCR-free document understanding. Specifically, we first construct a instruction tuning dataset featuring a wide range of visual-text understanding tasks. Then, we strengthen the OCR-free document understanding ability by jointly train the model on language-only, general vision-and-language, and document instruction tuning dataset with our unified instruction tuning strategy. We also build an OCR-free document instruction understanding evaluation set LLMDoc to better compare models' capabilities on instruct compliance and document understanding. Experimental results show that our model outperforms existing multi-modal models, demonstrating its strong ability of document understanding. Besides, without specific fine-tuning, mPLUG-DocOwl generalizes well on various downstream tasks. Our code, models, training data and evaluation set are available at https://github.com/X-PLUG/mPLUG-DocOwl."
                },
                {
                    "title": "GPT4Tools: Teaching Large Language Model to Use Tools via Self-instruction",
                    "abstract": "This paper aims to efficiently enable Large Language Models (LLMs) to use multimodal tools. Advanced proprietary LLMs, such as ChatGPT and GPT-4, have shown great potential for tool usage through sophisticated prompt engineering. Nevertheless, these models typically rely on prohibitive computational costs and publicly inaccessible data. To address these challenges, we propose the GPT4Tools based on self-instruct to enable open-source LLMs, such as LLaMA and OPT, to use tools. It generates an instruction-following dataset by prompting an advanced teacher with various multi-modal contexts. By using the Low-Rank Adaptation (LoRA) optimization, our approach facilitates the open-source LLMs to solve a range of visual problems, including visual comprehension and image generation. Moreover, we provide a benchmark to evaluate the ability of LLMs to use tools, which is performed in both zero-shot and fine-tuning ways. Extensive experiments demonstrate the effectiveness of our method on various language models, which not only significantly improves the accuracy of invoking seen tools, but also enables the zero-shot capacity for unseen tools. The code and demo are available at https://github.com/StevenGrove/GPT4Tools."
                },
                {
                    "title": "Region-Aware Pretraining for Open-Vocabulary Object Detection with Vision Transformers",
                    "abstract": "We present Region-aware Open-vocabulary Vision Transformers (RO-ViT) - a contrastive image-text pretraining recipe to bridge the gap between image-level pretraining and open-vocabulary object detection. At the pretraining phase, we propose to randomly crop and resize regions of positional embeddings instead of using the whole image positional embeddings. This better matches the use of positional embeddings at region-level in the detection finetuning phase. In addition, we replace the common softmax cross entropy loss in contrastive learning with focal loss to better learn the informative yet difficult examples. Finally, we leverage recent advances in novel object proposals to improve open-vocabulary detection finetuning. We evaluate our full model on the LVIS and COCO open-vocabulary detection benchmarks and zero-shot transfer. RO-ViT achieves a state-of-the-art 32.1 APr on LVIS, surpassing the best existing approach by +5.8 points in addition to competitive zero-shot transfer detection. Surprisingly, RO-ViT improves the image-level representation as well and achieves the state of the art on 9 out of 12 metrics on COCO and Flickr image-text retrieval benchmarks, outperforming competitive approaches with larger models."
                },
                {
                    "title": "mPLUG-Owl: Modularization Empowers Large Language Models with Multimodality",
                    "abstract": "Large language models (LLMs) have demonstrated impressive zero-shot abilities on a variety of open-ended tasks, while recent research has also explored the use of LLMs for multi-modal generation. In this study, we introduce mPLUG-Owl, a novel training paradigm that equips LLMs with multi-modal abilities through modularized learning of foundation LLM, a visual knowledge module, and a visual abstractor module. This approach can support multiple modalities and facilitate diverse unimodal and multimodal abilities through modality collaboration. The training paradigm of mPLUG-Owl involves a two-stage method for aligning image and text, which learns visual knowledge with the assistance of LLM while maintaining and even improving the generation abilities of LLM. In the first stage, the visual knowledge module and abstractor module are trained with a frozen LLM module to align the image and text. In the second stage, language-only and multi-modal supervised datasets are used to jointly fine-tune a low-rank adaption (LoRA) module on LLM and the abstractor module by freezing the visual knowledge module. We carefully build a visually-related instruction evaluation set OwlEval. Experimental results show that our model outperforms existing multi-modal models, demonstrating mPLUG-Owl's impressive instruction and visual understanding ability, multi-turn conversation ability, and knowledge reasoning ability. Besides, we observe some unexpected and exciting abilities such as multi-image correlation and scene text understanding, which makes it possible to leverage it for harder real scenarios, such as vision-only document comprehension. Our code, pre-trained model, instruction-tuned models, and evaluation set are available at https://github.com/X-PLUG/mPLUG-Owl. The online demo is available at https://www.modelscope.cn/studios/damo/mPLUG-Owl."
                },
                {
                    "title": "MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models",
                    "abstract": "The recent GPT-4 has demonstrated extraordinary multi-modal abilities, such as directly generating websites from handwritten text and identifying humorous elements within images. These features are rarely observed in previous vision-language models. However, the technical details behind GPT-4 continue to remain undisclosed. We believe that the enhanced multi-modal generation capabilities of GPT-4 stem from the utilization of sophisticated large language models (LLM). To examine this phenomenon, we present MiniGPT-4, which aligns a frozen visual encoder with a frozen advanced LLM, Vicuna, using one projection layer. Our work, for the first time, uncovers that properly aligning the visual features with an advanced large language model can possess numerous advanced multi-modal abilities demonstrated by GPT-4, such as detailed image description generation and website creation from hand-drawn drafts. Furthermore, we also observe other emerging capabilities in MiniGPT-4, including writing stories and poems inspired by given images, teaching users how to cook based on food photos, and so on. In our experiment, we found that the model trained on short image caption pairs could produce unnatural language outputs (e.g., repetition and fragmentation). To address this problem, we curate a detailed image description dataset in the second stage to finetune the model, which consequently improves the model's generation reliability and overall usability. Our code, pre-trained model, and collected dataset are available at https://minigpt-4.github.io/."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "Vita-CLIP: Video and text adaptive CLIP via Multimodal Prompting",
                    "abstract": "Adopting contrastive image-text pretrained models like CLIP towards video classification has gained attention due to its cost-effectiveness and competitive performance. However, recent works in this area face a trade-off. Finetuning the pretrained model to achieve strong supervised performance results in low zero-shot generalization. Similarly, freezing the backbone to retain zero-shot capability causes significant drop in supervised accuracy. Because of this, recent works in literature typically train separate models for supervised and zero-shot action recognition. In this work, we propose a multimodal prompt learning scheme that works to balance the supervised and zero-shot performance under a single unified training. Our prompting approach on the vision side caters for three aspects: 1) Global video-level prompts to model the data distribution; 2) Local frame-level prompts to provide per-frame discriminative conditioning; and 3) a summary prompt to extract a condensed video representation. Additionally, we define a prompting scheme on the text side to augment the textual context. Through this prompting scheme, we can achieve state-of-the-art zero-shot performance on Kinetics-600, HMDB51 and UCF101 while remaining competitive in the supervised setting. By keeping the pretrained backbone frozen, we optimize a much lower number of parameters and retain the existing general representation which helps achieve the strong zero-shot performance. Our codes/models will be released at https://github.com/TalalWasim/Vita-Clip.."
                },
                {
                    "title": "Instruction Tuning with GPT-4",
                    "abstract": "Prior work has shown that finetuning large language models (LLMs) using machine-generated instruction-following data enables such models to achieve remarkable zero-shot capabilities on new tasks, and no human-written instructions are needed. In this paper, we present the first attempt to use GPT-4 to generate instruction-following data for LLM finetuning. Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by previous state-of-the-art models. We also collect feedback and comparison data from GPT-4 to enable a comprehensive evaluation and reward model training. We make our data generated using GPT-4 as well as our codebase publicly available."
                },
                {
                    "title": "Segment Anything",
                    "abstract": "We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive \u2013 often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images at segment-anything.com to foster research into foundation models for computer vision. We recommend reading the full paper at: arxiv.org/abs/2304.02643."
                },
                {
                    "title": "Vision Transformer With Quadrangle Attention",
                    "abstract": "Window-based attention has become a popular choice in vision transformers due to its superior performance, lower computational complexity, and less memory footprint. However, the design of hand-crafted windows, which is data-agnostic, constrains the flexibility of transformers to adapt to objects of varying sizes, shapes, and orientations. To address this issue, we propose a novel quadrangle attention (QA) method that extends the window-based attention to a general quadrangle formulation. Our method employs an end-to-end learnable quadrangle regression module that predicts a transformation matrix to transform default windows into target quadrangles for token sampling and attention calculation, enabling the network to model various targets with different shapes and orientations and capture rich context information. We integrate QA into plain and hierarchical vision transformers to create a new architecture named QFormer, which offers minor code modifications and negligible extra computational cost. Extensive experiments on public benchmarks demonstrate that QFormer outperforms existing representative vision transformers on various vision tasks, including classification, object detection, semantic segmentation, and pose estimation. The code will be made publicly available at QFormer."
                },
                {
                    "title": "Sigmoid Loss for Language Image Pre-Training",
                    "abstract": "We propose a simple pairwise sigmoid loss for imagetext pre-training. Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization. The sigmoid loss simultaneously allows further scaling up the batch size, while also performing better at smaller batch sizes. With only four TPUv4 chips, we can train a Base CLIP model at 4k batch size and a Large LiT model at 20k batch size, the latter achieves 84.5% ImageNet zero-shot accuracy in two days. This disentanglement of the batch size from the loss further allows us to study the impact of examples vs pairs and negative to positive ratio. Finally, we push the batch size to the extreme, up to one million, and find that the benefits of growing batch size quickly diminish, with a more reasonable batch size of 32k being sufficient. We hope our research motivates further explorations in improving the quality and efficiency of language-image pre-training."
                },
                {
                    "title": "CLIP2: Contrastive Language-Image-Point Pretraining from Real-World Point Cloud Data",
                    "abstract": "Contrastive Language-Image Pre-training, benefiting from large-scale unlabeled text-image pairs, has demonstrated great performance in open-world vision understanding tasks. However, due to the limited Text-3D data pairs, adapting the success of 2D Vision-Language Models (VLM) to the 3D space remains an open problem. Existing works that leverage VLM for 3D understanding generally resort to constructing intermediate 2D representations for the 3D data, but at the cost of losing 3D geometry information. To take a step toward open-world 3D vision understanding, we propose Contrastive Language-Image-Point Cloud Pretraining (CLIP2) to directly learn the transferable 3D point cloud representation in realistic scenarios with a novel proxy alignment mechanism. Specifically, we exploit naturally-existed correspondences in 2D and 3D scenarios, and build well-aligned and instance-based text-image-point proxies from those complex scenarios. On top of that, we propose a cross-modal contrastive objective to learn semantic and instance-level aligned point cloud representation. Experimental results on both indoor and outdoor scenarios show that our learned 3D representation has great transfer ability in downstream tasks, including zero-shot and few-shot 3D recognition, which boosts the state-of-the-art methods by large margins. Furthermore, we provide analyses of the capability of different representations in real scenarios and present the optional ensemble scheme."
                },
                {
                    "title": "BiFormer: Vision Transformer with Bi-Level Routing Attention",
                    "abstract": "As the core building block of vision transformers, attention is a powerful tool to capture long-range dependency. However, such power comes at a cost: it incurs a huge computation burden and heavy memory footprint as pairwise token interaction across all spatial locations is computed. A series of works attempt to alleviate this problem by introducing handcrafted and content-agnostic sparsity into attention, such as restricting the attention operation to be inside local windows, axial stripes, or dilated windows. In contrast to these approaches, we propose a novel dynamic sparse attention via bi-level routing to enable a more flexible allocation of computations with content awareness. Specifically, for a query, irrelevant key-value pairs are first filtered out at a coarse region level, and then fine-grained token-to-token attention is applied in the union of remaining candidate regions (i.e., routed regions). We provide a simple yet effective implementation of the proposed bilevel routing attention, which utilizes the sparsity to save both computation and memory while involving only GPU-friendly dense matrix multiplications. Built with the proposed bi-level routing attention, a new general vision transformer, named BiFormer, is then presented. As BiFormer attends to a small subset of relevant tokens in a query adaptive manner without distraction from other irrelevant ones, it enjoys both good performance and high computational efficiency, especially in dense prediction tasks. Empirical results across several computer vision tasks such as image classification, object detection, and semantic segmentation verify the effectiveness of our design. Code is available at https://github.com/rayleizhu/BiFormer."
                },
                {
                    "title": "Visual ChatGPT: Talking, Drawing and Editing with Visual Foundation Models",
                    "abstract": "ChatGPT is attracting a cross-field interest as it provides a language interface with remarkable conversational competency and reasoning capabilities across many domains. However, since ChatGPT is trained with languages, it is currently not capable of processing or generating images from the visual world. At the same time, Visual Foundation Models, such as Visual Transformers or Stable Diffusion, although showing great visual understanding and generation capabilities, they are only experts on specific tasks with one-round fixed inputs and outputs. To this end, We build a system called \\textbf{Visual ChatGPT}, incorporating different Visual Foundation Models, to enable the user to interact with ChatGPT by 1) sending and receiving not only languages but also images 2) providing complex visual questions or visual editing instructions that require the collaboration of multiple AI models with multi-steps. 3) providing feedback and asking for corrected results. We design a series of prompts to inject the visual model information into ChatGPT, considering models of multiple inputs/outputs and models that require visual feedback. Experiments show that Visual ChatGPT opens the door to investigating the visual roles of ChatGPT with the help of Visual Foundation Models. Our system is publicly available at \\url{https://github.com/microsoft/visual-chatgpt}."
                },
                {
                    "title": "Language Is Not All You Need: Aligning Perception with Language Models",
                    "abstract": "A big convergence of language, multimodal perception, action, and world modeling is a key step toward artificial general intelligence. In this work, we introduce Kosmos-1, a Multimodal Large Language Model (MLLM) that can perceive general modalities, learn in context (i.e., few-shot), and follow instructions (i.e., zero-shot). Specifically, we train Kosmos-1 from scratch on web-scale multimodal corpora, including arbitrarily interleaved text and images, image-caption pairs, and text data. We evaluate various settings, including zero-shot, few-shot, and multimodal chain-of-thought prompting, on a wide range of tasks without any gradient updates or finetuning. Experimental results show that Kosmos-1 achieves impressive performance on (i) language understanding, generation, and even OCR-free NLP (directly fed with document images), (ii) perception-language tasks, including multimodal dialogue, image captioning, visual question answering, and (iii) vision tasks, such as image recognition with descriptions (specifying classification via text instructions). We also show that MLLMs can benefit from cross-modal transfer, i.e., transfer knowledge from language to multimodal, and from multimodal to language. In addition, we introduce a dataset of Raven IQ test, which diagnoses the nonverbal reasoning capability of MLLMs."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "Scene Text Recognition with Permuted Autoregressive Sequence Models",
                    "abstract": "Context-aware STR methods typically use internal autoregressive (AR) language models (LM). Inherent limitations of AR models motivated two-stage methods which employ an external LM. The conditional independence of the external LM on the input image may cause it to erroneously rectify correct predictions, leading to significant inefficiencies. Our method, PARSeq, learns an ensemble of internal AR LMs with shared weights using Permutation Language Modeling. It unifies context-free non-AR and context-aware AR inference, and iterative refinement using bidirectional context. Using synthetic training data, PARSeq achieves state-of-the-art (SOTA) results in STR benchmarks (91.9% accuracy) and more challenging datasets. It establishes new SOTA results (96.0% accuracy) when trained on real data. PARSeq is optimal on accuracy vs parameter count, FLOPS, and latency because of its simple, unified structure and parallel token processing. Due to its extensive use of attention, it is robust on arbitrarily-oriented text which is common in real-world images. Code, pretrained weights, and data are available at: https://github.com/baudm/parseq."
                },
                {
                    "title": "EfficientViT: Multi-Scale Linear Attention for High-Resolution Dense Prediction",
                    "abstract": "High-resolution dense prediction enables many appealing real-world applications, such as computational photography, autonomous driving, etc. However, the vast computational cost makes deploying state-of-the-art high-resolution dense prediction models on hardware devices difficult. This work presents EfficientViT, a new family of high-resolution vision models with novel multi-scale linear attention. Unlike prior high-resolution dense prediction models that rely on heavy softmax attention, hardware-inefficient large-kernel convolution, or complicated topology structure to obtain good performances, our multi-scale linear attention achieves the global receptive field and multi-scale learning (two desirable features for high-resolution dense prediction) with only lightweight and hardware-efficient operations. As such, EfficientViT delivers remarkable performance gains over previous state-of-the-art models with significant speedup on diverse hardware platforms, including mobile CPU, edge GPU, and cloud GPU. Without performance loss on Cityscapes, our EfficientViT provides up to 13.9$\\times$ and 6.2$\\times$ GPU latency reduction over SegFormer and SegNeXt, respectively. For super-resolution, EfficientViT delivers up to 6.4x speedup over Restormer while providing 0.11dB gain in PSNR. For Segment Anything, EfficientViT delivers 48.9x higher throughput on A100 GPU while achieving slightly better zero-shot instance segmentation performance on COCO."
                },
                {
                    "title": "OPT: Open Pre-trained Transformer Language Models",
                    "abstract": "Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models."
                },
                {
                    "title": "Flamingo: a Visual Language Model for Few-Shot Learning",
                    "abstract": "Building models that can be rapidly adapted to novel tasks using only a handful of annotated examples is an open challenge for multimodal machine learning research. We introduce Flamingo, a family of Visual Language Models (VLM) with this ability. We propose key architectural innovations to: (i) bridge powerful pretrained vision-only and language-only models, (ii) handle sequences of arbitrarily interleaved visual and textual data, and (iii) seamlessly ingest images or videos as inputs. Thanks to their flexibility, Flamingo models can be trained on large-scale multimodal web corpora containing arbitrarily interleaved text and images, which is key to endow them with in-context few-shot learning capabilities. We perform a thorough evaluation of our models, exploring and measuring their ability to rapidly adapt to a variety of image and video tasks. These include open-ended tasks such as visual question-answering, where the model is prompted with a question which it has to answer; captioning tasks, which evaluate the ability to describe a scene or an event; and close-ended tasks such as multiple-choice visual question-answering. For tasks lying anywhere on this spectrum, a single Flamingo model can achieve a new state of the art with few-shot learning, simply by prompting the model with task-specific examples. On numerous benchmarks, Flamingo outperforms models fine-tuned on thousands of times more task-specific data."
                },
                {
                    "title": "ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning",
                    "abstract": "Charts are very popular for analyzing data. When exploring charts, people often ask a variety of complex reasoning questions that involve several logical and arithmetic operations. They also commonly refer to visual features of a chart in their questions. However, most existing datasets do not focus on such complex reasoning questions as their questions are template-based and answers come from a fixed-vocabulary. In this work, we present a large-scale benchmark covering 9.6K human-written questions as well as 23.1K questions generated from human-written chart summaries. To address the unique challenges in our benchmark involving visual and logical reasoning over charts, we present two transformer-based models that combine visual features and the data table of the chart in a unified way to answer questions. While our models achieve the state-of-the-art results on the previous datasets as well as on our benchmark, the evaluation also reveals several challenges in answering complex reasoning questions."
                },
                {
                    "title": "GroupViT: Semantic Segmentation Emerges from Text Supervision",
                    "abstract": "Grouping and recognition are important components of visual scene understanding, e.g., for object detection and semantic segmentation. With end-to-end deep learning systems, grouping of image regions usually happens implicitly via top-down supervision from pixel-level recognition labels. Instead, in this paper, we propose to bring back the grouping mechanism into deep networks, which allows semantic segments to emerge automatically with only text supervision. We propose a hierarchical Grouping Vision Transformer (GroupViT), which goes beyond the regular grid structure representation and learns to group image regions into progressively larger arbitrary-shaped segments. We train GroupViT jointly with a text encoder on a large-scale image-text dataset via contrastive losses. With only text supervision and without any pixel-level annotations, GroupViT learns to group together semantic regions and successfully transfers to the task of semantic segmentation in a zero-shot manner, i.e., without any further fine-tuning. It achieves a zero-shot accuracy of 52.3% mIoU on the PASCAL VOC 2012 and 22.4% mIoU on PASCAL Context datasets, and performs competitively to state-of-the-art transfer-learning methods requiring greater levels of supervision. We open-source our code at https://github.com/NVlabs/GroupViT."
                },
                {
                    "title": "BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation",
                    "abstract": "Vision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has been largely achieved by scaling up the dataset with noisy image-text pairs collected from the web, which is a suboptimal source of supervision. In this paper, we propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks. BLIP effectively utilizes the noisy web data by bootstrapping the captions, where a captioner generates synthetic captions and a filter removes the noisy ones. We achieve state-of-the-art results on a wide range of vision-language tasks, such as image-text retrieval (+2.7% in average recall@1), image captioning (+2.8% in CIDEr), and VQA (+1.6% in VQA score). BLIP also demonstrates strong generalization ability when directly transferred to video-language tasks in a zero-shot manner. Code, models, and datasets are released at https://github.com/salesforce/BLIP."
                },
                {
                    "title": "PointCLIP: Point Cloud Understanding by CLIP",
                    "abstract": "Recently, zero-shot and few-shot learning via Contrastive Vision-Language Pre-training (CLIP) have shown inspirational performance on 2D visual recognition, which learns to match images with their corresponding texts in open-vocabulary settings. However, it remains under explored that whether CLIP, pre-trained by large-scale image-text pairs in 2D, can be generalized to 3D recognition. In this paper, we identify such a setting is feasible by proposing PointCLIP, which conducts alignment between CLIP-encoded point clouds and 3D category texts. Specifically, we encode a point cloud by projecting it onto multi-view depth maps and aggregate the view-wise zero-shot prediction in an end-to-end manner, which achieves efficient knowledge transfer from 2D to 3D. We further design an inter-view adapter to better extract the global feature and adaptively fuse the 3D few-shot knowledge into CLIP pre-trained in 2D. By just fine-tuning the adapter under few-shot settings, the performance of PointCLIP could be largely improved. In addition, we observe the knowledge complementary property between PointCLIP and classical 3D-supervised networks. Via simple ensemble during inference, PointCLIP contributes to favorable performance enhancement over state-of-the-art 3D networks. Therefore, PointCLIP is a promising alternative for effective 3D point cloud understanding under low data regime with marginal resource cost. We conduct thorough experiments on Model-NetlO, ModelNet40 and ScanObjectNN to demonstrate the effectiveness of PointCLIP. Code is available at https://github.com/ZrrSkywalker/PointCLIP."
                },
                {
                    "title": "MViTv2: Improved Multiscale Vision Transformers for Classification and Detection",
                    "abstract": "In this paper, we study Multiscale Vision Transformers (MViTv2) as a unified architecture for image and video classification, as well as object detection. We present an improved version of MViT that incorporates decomposed relative positional embeddings and residual pooling connections. We instantiate this architecture in five sizes and evaluate it for ImageNet classification, COCO detection and Kinetics video recognition where it outperforms prior work. We further compare MViTv2s' pooling attention to window attention mechanisms where it outperforms the latter in accuracy/compute. Without bells-and-whistles, MViTv2 has state-of-the-art performance in 3 domains: 88.8% accuracy on ImageNet classification, 58.7 APbox on COCO object detection as well as 86.1% on Kinetics-400 video classification. Code and models are available at https://github.com/facebookresearch/mvit."
                },
                {
                    "title": "Swin Transformer V2: Scaling Up Capacity and Resolution",
                    "abstract": "We present techniques for scaling Swin Transformer [35] up to 3 billion parameters and making it capable of training with images of up to 1,536x1,536 resolution. By scaling up capacity and resolution, Swin Transformer sets new records on four representative vision benchmarks: 84.0% top-1 accuracy on ImageNet- V2 image classification, 63.1 / 54.4 box / mask mAP on COCO object detection, 59.9 mIoU on ADE20K semantic segmentation, and 86.8% top-1 accuracy on Kinetics-400 video action classification. We tackle issues of training instability, and study how to effectively transfer models pre-trained at low resolutions to higher resolution ones. To this aim, several novel technologies are proposed: 1) a residual post normalization technique and a scaled cosine attention approach to improve the stability of large vision models; 2) a log-spaced continuous position bias technique to effectively transfer models pre-trained at low-resolution images and windows to their higher-resolution counterparts. In addition, we share our crucial implementation details that lead to significant savings of GPU memory consumption and thus make it feasi-ble to train large vision models with regular GPUs. Using these techniques and self-supervised pre-training, we suc-cessfully train a strong 3 billion Swin Transformer model and effectively transfer it to various vision tasks involving high-resolution images or windows, achieving the state-of-the-art accuracy on a variety of benchmarks. Code is avail-able at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "LAION-400M: Open Dataset of CLIP-Filtered 400 Million Image-Text Pairs",
                    "abstract": "Multi-modal language-vision models trained on hundreds of millions of image-text pairs (e.g. CLIP, DALL-E) gained a recent surge, showing remarkable capability to perform zero- or few-shot learning and transfer even in absence of per-sample labels on target image data. Despite this trend, to date there has been no publicly available datasets of sufficient scale for training such models from scratch. To address this issue, in a community effort we build and release for public LAION-400M, a dataset with CLIP-filtered 400 million image-text pairs, their CLIP embeddings and kNN indices that allow efficient similarity search."
                },
                {
                    "title": "Dynamic DETR: End-to-End Object Detection with Dynamic Attention",
                    "abstract": "In this paper, we present a novel Dynamic DETR (Detection with Transformers) approach by introducing dynamic attentions into both the encoder and decoder stages of DETR to break its two limitations on small feature resolution and slow training convergence. To address the first limitation, which is due to the quadratic computational complexity of the self-attention module in Transformer encoders, we propose a dynamic encoder to approximate the Transformer encoder\u2019s attention mechanism using a convolution-based dynamic encoder with various attention types. Such an encoder can dynamically adjust attentions based on multiple factors such as scale importance, spatial importance, and representation (i.e., feature dimension) importance. To mitigate the second limitation of learning difficulty, we introduce a dynamic decoder by replacing the cross-attention module with a ROI-based dynamic attention in the Transformer decoder. Such a decoder effectively assists Transformers to focus on region of interests from a coarse-to-fine manner and dramatically lowers the learning difficulty, leading to a much faster convergence with fewer training epochs. We conduct a series of experiments to demonstrate our advantages. Our Dynamic DETR significantly reduces the training epochs (by 14\u00d7), yet results in a much better performance (by 3.6 on mAP). Meanwhile, in the standard 1\u00d7 setup with ResNet-50 backbone, we archive a new state-of-the-art performance that further proves the learning effectiveness of the proposed approach."
                },
                {
                    "title": "An End-to-End Transformer Model for 3D Object Detection",
                    "abstract": "We propose 3DETR, an end-to-end Transformer based object detection model for 3D point clouds. Compared to existing detection methods that employ a number of 3D-specific inductive biases, 3DETR requires minimal modifications to the vanilla Transformer block. Specifically, we find that a standard Transformer with non-parametric queries and Fourier positional embeddings is competitive with specialized architectures that employ libraries of 3D-specific operators with hand-tuned hyperparameters. Nevertheless, 3DETR is conceptually simple and easy to implement, enabling further improvements by incorporating 3D domain knowledge. Through extensive experiments, we show 3DETR outperforms the well-established and highly optimized VoteNet baselines on the challenging ScanNetV2 dataset by 9.5%. Furthermore, we show 3DETR is applicable to 3D tasks beyond detection, and can serve as a building block for future research."
                },
                {
                    "title": "LoRA: Low-Rank Adaptation of Large Language Models",
                    "abstract": "An important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible. Using GPT-3 175B as an example -- deploying independent instances of fine-tuned models, each with 175B parameters, is prohibitively expensive. We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks. Compared to GPT-3 175B fine-tuned with Adam, LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times. LoRA performs on-par or better than fine-tuning in model quality on RoBERTa, DeBERTa, GPT-2, and GPT-3, despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency. We also provide an empirical investigation into rank-deficiency in language model adaptation, which sheds light on the efficacy of LoRA. We release a package that facilitates the integration of LoRA with PyTorch models and provide our implementations and model checkpoints for RoBERTa, DeBERTa, and GPT-2 at https://github.com/microsoft/LoRA."
                },
                {
                    "title": "SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers",
                    "abstract": "We present SegFormer, a simple, efficient yet powerful semantic segmentation framework which unifies Transformers with lightweight multilayer perception (MLP) decoders. SegFormer has two appealing features: 1) SegFormer comprises a novel hierarchically structured Transformer encoder which outputs multiscale features. It does not need positional encoding, thereby avoiding the interpolation of positional codes which leads to decreased performance when the testing resolution differs from training. 2) SegFormer avoids complex decoders. The proposed MLP decoder aggregates information from different layers, and thus combining both local attention and global attention to render powerful representations. We show that this simple and lightweight design is the key to efficient segmentation on Transformers. We scale our approach up to obtain a series of models from SegFormer-B0 to SegFormer-B5, reaching significantly better performance and efficiency than previous counterparts. For example, SegFormer-B4 achieves 50.3% mIoU on ADE20K with 64M parameters, being 5x smaller and 2.2% better than the previous best method. Our best model, SegFormer-B5, achieves 84.0% mIoU on Cityscapes validation set and shows excellent zero-shot robustness on Cityscapes-C. Code will be released at: github.com/NVlabs/SegFormer."
                },
                {
                    "title": "InfographicVQA",
                    "abstract": "Infographics communicate information using a combination of textual, graphical and visual elements. This work explores the automatic understanding of infographic images by using a Visual Question Answering technique. To this end, we present InfographicVQA, a new dataset comprising a diverse collection of infographics and question-answer annotations. The questions require methods that jointly reason over the document layout, textual content, graphical elements, and data visualizations. We curate the dataset with an emphasis on questions that require elementary reasoning and basic arithmetic skills. For VQA on the dataset, we evaluate two Transformer-based strong baselines. Both the baselines yield unsatisfactory results compared to near perfect human performance on the dataset. The results suggest that VQA on infographics\u2014images that are designed to communicate information quickly and clearly to human brain\u2014is ideal for benchmarking machine understanding of complex document images. The dataset is available for download at docvqa.org"
                },
                {
                    "title": "Rethinking Text Line Recognition Models",
                    "abstract": "In this paper, we study the problem of text line recognition. Unlike most approaches targeting specific domains such as scene-text or handwritten documents, we investigate the general problem of developing a universal architecture that can extract text from any image, regardless of source or input modality. We consider two decoder families (Connectionist Temporal Classification and Transformer) and three encoder modules (Bidirectional LSTMs, Self-Attention, and GRCLs), and conduct extensive experiments to compare their accuracy and performance on widely used public datasets of scene and handwritten text. We find that a combination that so far has received little attention in the literature, namely a Self-Attention encoder coupled with the CTC decoder, when compounded with an external language model and trained on both public and internal data, outperforms all the others in accuracy and computational complexity. Unlike the more common Transformer-based models, this architecture can handle inputs of arbitrary length, a requirement for universal line recognition. Using an internal dataset collected from multiple sources, we also expose the limitations of current public datasets in evaluating the accuracy of line recognizers, as the relatively narrow image width and sequence length distributions do not allow to observe the quality degradation of the Transformer approach when applied to the transcription of long lines."
                },
                {
                    "title": "CvT: Introducing Convolutions to Vision Transformers",
                    "abstract": "We present in this paper a new architecture, named Convolutional vision Transformer (CvT), that improves Vision Transformer (ViT) in performance and efficiency by introducing convolutions into ViT to yield the best of both de-signs. This is accomplished through two primary modifications: a hierarchy of Transformers containing a new convolutional token embedding, and a convolutional Transformer block leveraging a convolutional projection. These changes introduce desirable properties of convolutional neural networks (CNNs) to the ViT architecture (i.e. shift, scale, and distortion invariance) while maintaining the merits of Transformers (i.e. dynamic attention, global context, and better generalization). We validate CvT by conducting extensive experiments, showing that this approach achieves state-of-the-art performance over other Vision Transformers and ResNets on ImageNet-1k, with fewer parameters and lower FLOPs. In addition, performance gains are maintained when pretrained on larger datasets (e.g. ImageNet-22k) and fine-tuned to downstream tasks. Pretrained on ImageNet-22k, our CvT-W24 obtains a top-1 accuracy of 87.7% on the ImageNet-1k val set. Finally, our results show that the positional encoding, a crucial component in existing Vision Transformers, can be safely re-moved in our model, simplifying the design for higher resolution vision tasks. Code will be released at https://github.com/microsoft/CvT."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions",
                    "abstract": "Although convolutional neural networks (CNNs) have achieved great success in computer vision, this work investigates a simpler, convolution-free backbone network use-fid for many dense prediction tasks. Unlike the recently-proposed Vision Transformer (ViT) that was designed for image classification specifically, we introduce the Pyramid Vision Transformer (PVT), which overcomes the difficulties of porting Transformer to various dense prediction tasks. PVT has several merits compared to current state of the arts. (1) Different from ViT that typically yields low-resolution outputs and incurs high computational and memory costs, PVT not only can be trained on dense partitions of an image to achieve high output resolution, which is important for dense prediction, but also uses a progressive shrinking pyramid to reduce the computations of large feature maps. (2) PVT inherits the advantages of both CNN and Transformer, making it a unified backbone for various vision tasks without convolutions, where it can be used as a direct replacement for CNN backbones. (3) We validate PVT through extensive experiments, showing that it boosts the performance of many downstream tasks, including object detection, instance and semantic segmentation. For example, with a comparable number of parameters, PVT+RetinaNet achieves 40.4 AP on the COCO dataset, surpassing ResNet50+RetinNet (36.3 AP) by 4.1 absolute AP (see Figure 2). We hope that PVT could, serre as an alternative and useful backbone for pixel-level predictions and facilitate future research."
                },
                {
                    "title": "High-Performance Large-Scale Image Recognition Without Normalization",
                    "abstract": "Batch normalization is a key component of most image classification models, but it has many undesirable properties stemming from its dependence on the batch size and interactions between examples. Although recent work has succeeded in training deep ResNets without normalization layers, these models do not match the test accuracies of the best batch-normalized networks, and are often unstable for large learning rates or strong data augmentations. In this work, we develop an adaptive gradient clipping technique which overcomes these instabilities, and design a significantly improved class of Normalizer-Free ResNets. Our smaller models match the test accuracy of an EfficientNet-B7 on ImageNet while being up to 8.7x faster to train, and our largest models attain a new state-of-the-art top-1 accuracy of 86.5%. In addition, Normalizer-Free models attain significantly better performance than their batch-normalized counterparts when finetuning on ImageNet after large-scale pre-training on a dataset of 300 million labeled images, with our best models obtaining an accuracy of 89.2%. Our code is available at https://github.com/deepmind/ deepmind-research/tree/master/nfnets"
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "DocVQA: A Dataset for VQA on Document Images",
                    "abstract": "We present a new dataset for Visual Question Answering (VQA) on document images called DocVQA. The dataset consists of 50,000 questions defined on 12,000+ document images. Detailed analysis of the dataset in comparison with similar datasets for VQA and reading comprehension is presented. We report several baseline results by adopting existing VQA and reading comprehension models. Although the existing models perform reasonably well on certain types of questions, there is large performance gap compared to human performance (94.36% accuracy). The models need to improve specifically on questions where understanding structure of the document is crucial. The dataset, code and leaderboard are available at docvqa.org"
                },
                {
                    "title": "MMDetection: Open MMLab Detection Toolbox and Benchmark",
                    "abstract": "We present MMDetection, an object detection toolbox that contains a rich set of object detection and instance segmentation methods as well as related components and modules. The toolbox started from a codebase of MMDet team who won the detection track of COCO Challenge 2018. It gradually evolves into a unified platform that covers many popular detection methods and contemporary modules. It not only includes training and inference codes, but also provides weights for more than 200 network models. We believe this toolbox is by far the most complete detection toolbox. In this paper, we introduce the various features of this toolbox. In addition, we also conduct a benchmarking study on different methods, components, and their hyper-parameters. We wish that the toolbox and benchmark could serve the growing research community by providing a flexible toolkit to reimplement existing methods and develop their own new detectors. Code and models are available at this https URL. The project is under active development and we will keep this document updated."
                },
                {
                    "title": "OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge",
                    "abstract": "Visual Question Answering (VQA) in its ideal form lets us study reasoning in the joint space of vision and language and serves as a proxy for the AI task of scene understanding. However, most VQA benchmarks to date are focused on questions such as simple counting, visual attributes, and object detection that do not require reasoning or knowledge beyond what is in the image. In this paper, we address the task of knowledge-based visual question answering and provide a benchmark, called OK-VQA, where the image content is not sufficient to answer the questions, encouraging methods that rely on external knowledge resources. Our new dataset includes more than 14,000 questions that require external knowledge to answer. We show that the performance of the state-of-the-art VQA models degrades drastically in this new setting. Our analysis shows that our knowledge-based VQA task is diverse, difficult, and large compared to previous knowledge-based VQA datasets. We hope that this dataset enables researchers to open up new avenues for research in this domain."
                },
                {
                    "title": "FUNSD: A Dataset for Form Understanding in Noisy Scanned Documents",
                    "abstract": "We present a new dataset for form understanding in noisy scanned documents (FUNSD) that aims at extracting and structuring the textual content of forms. The dataset comprises 199 real, fully annotated, scanned forms. The documents are noisy and vary widely in appearance, making form understanding (FoUn) a challenging task. The proposed dataset can be used for various tasks, including text detection, optical character recognition, spatial layout analysis, and entity labeling/linking. To the best of our knowledge, this is the first publicly available dataset with comprehensive annotations to address FoUn task. We also present a set of baselines and introduce metrics to evaluate performance on the FUNSD dataset, which can be downloaded at https://guillaumejaume.github.io/FUNSD."
                },
                {
                    "title": "GQA: a new dataset for compositional question answering over real-world images",
                    "abstract": "We introduce GQA, a new dataset for real-world visual reasoning and compositional question answering, seeking to address key shortcomings of previous VQA datasets. We have developed a strong and robust question engine that leverages scene graph structures to create 22M diverse reasoning questions, all come with functional programs that represent their semantics. We use the programs to gain tight control over the answer distribution and present a new tunable smoothing technique to mitigate language biases. Accompanying the dataset is a suite of new metrics that evaluate essential qualities such as consistency, grounding and plausibility. An extensive analysis is performed for baselines as well as state-of-the-art models, providing fine-grained results for different question types and topologies. Whereas a blind LSTM obtains mere 42.1%, and strong VQA models achieve 54.1%, human performance tops at 89.3\\%, offering ample opportunity for new research to explore. We strongly hope GQA will provide an enabling resource for the next generation of models with enhanced robustness, improved consistency, and deeper semantic understanding for images and language."
                },
                {
                    "title": "Conceptual Captions: A Cleaned, Hypernymed, Image Alt-text Dataset For Automatic Image Captioning",
                    "abstract": "We present a new dataset of image caption annotations, Conceptual Captions, which contains an order of magnitude more images than the MS-COCO dataset (Lin et al., 2014) and represents a wider variety of both images and image caption styles. We achieve this by extracting and filtering image caption annotations from billions of webpages. We also present quantitative evaluations of a number of image captioning models and show that a model architecture based on Inception-ResNetv2 (Szegedy et al., 2016) for image-feature extraction and Transformer (Vaswani et al., 2017) for sequence modeling achieves the best performance when trained on the Conceptual Captions dataset."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "End-to-end text recognition with convolutional neural networks",
                    "abstract": "Full end-to-end text recognition in natural images is a challenging problem that has received much attention recently. Traditional systems in this area have relied on elaborate models incorporating carefully hand-engineered features or large amounts of prior knowledge. In this paper, we take a different route and combine the representational power of large, multilayer neural networks together with recent developments in unsupervised feature learning, which allows us to use a common framework to train highly-accurate text detector and character recognizer modules. Then, using only simple off-the-shelf methods, we integrate these two modules into a full end-to-end, lexicon-driven, scene text recognition system that achieves state-of-the-art performance on standard benchmarks, namely Street View Text and ICDAR 2003."
                },
                {
                    "title": "End-to-end scene text recognition",
                    "abstract": "This paper focuses on the problem of word detection and recognition in natural images. The problem is significantly more challenging than reading text in scanned documents, and has only recently gained attention from the computer vision community. Sub-components of the problem, such as text detection and cropped image word recognition, have been studied in isolation [7, 4, 20]. However, what is unclear is how these recent approaches contribute to solving the end-to-end problem of word recognition. We fill this gap by constructing and evaluating two systems. The first, representing the de facto state-of-the-art, is a two stage pipeline consisting of text detection followed by a leading OCR engine. The second is a system rooted in generic object recognition, an extension of our previous work in [20]. We show that the latter approach achieves superior performance. While scene text recognition has generally been treated with highly domain-specific methods, our results demonstrate the suitability of applying generic computer vision methods. Adopting this approach opens the door for real world scene text recognition to benefit from the rapid advances that have been taking place in object recognition."
                },
                {
                    "title": "HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face",
                    "abstract": "Solving complicated AI tasks with different domains and modalities is a key step toward artificial general intelligence. While there are numerous AI models available for various domains and modalities, they cannot handle complicated AI tasks autonomously. Considering large language models (LLMs) have exhibited exceptional abilities in language understanding, generation, interaction, and reasoning, we advocate that LLMs could act as a controller to manage existing AI models to solve complicated AI tasks, with language serving as a generic interface to empower this. Based on this philosophy, we present HuggingGPT, an LLM-powered agent that leverages LLMs (e.g., ChatGPT) to connect various AI models in machine learning communities (e.g., Hugging Face) to solve AI tasks. Specifically, we use ChatGPT to conduct task planning when receiving a user request, select models according to their function descriptions available in Hugging Face, execute each subtask with the selected AI model, and summarize the response according to the execution results. By leveraging the strong language capability of ChatGPT and abundant AI models in Hugging Face, HuggingGPT can tackle a wide range of sophisticated AI tasks spanning different modalities and domains and achieve impressive results in language, vision, speech, and other challenging tasks, which paves a new way towards the realization of artificial general intelligence."
                },
                {
                    "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows",
                    "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer."
                },
                {
                    "title": "Recognition : Learning Where to Start and When to Stop",
                    "abstract": "Text line detection and localization is a crucial step for full page document analysis, but still suffers from heterogeneity of real life documents. In this paper, we present a new approach for full page text recognition. Localization of the text lines is based on regressions with Fully Convolutional Neural Networks and Multidimensional Long Short-Term Memory as contextual layers. In order to increase the efficiency of this localization method, only the position of the left side of the text lines are predicted. The text recognizer is then in charge of predicting the end of the text to recognize. This method has shown good results for full page text recognition on the highly heterogeneous Maurdor dataset."
                }
            ],
            "categories": [
                "cs.CV"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we develop a unified model that simultaneously supports both image and text recognition in complex document layouts, overcoming the limitations of current vision encoder models?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the capabilities of vision encoder models, which are foundational in various applications such as healthcare diagnostics, autonomous driving, and content analysis. A unified model would enhance the efficiency and accuracy of document analysis tasks, enabling intelligent systems to interpret complex visual information as humans do. This advancement could lead to significant improvements in automated document processing, making it more accessible and effective across diverse domains. Furthermore, addressing this question could inspire future research into more integrated approaches in computer vision and natural language processing, fostering innovations that leverage multimodal data.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the need for precise local feature extraction for text recognition, which is distinct from the global feature extraction typically handled by Vision Transformers (ViTs). Naive approaches, such as finetuning pre-trained ViTs on high-resolution documents, fail due to the difficulties in interpolating positional embeddings, which disrupts spatial alignment and degrades performance. Additionally, existing methods that focus solely on OCR tasks discard the original image encoding capabilities, while ensemble methods increase computational costs and complexity. The intricacies of handling diverse document layouts, varying font sizes, and mixed media further complicate the development of a robust solution.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either image or text recognition, leading to a lack of integrated models capable of handling both simultaneously. Limitations in existing solutions stem from their reliance on separate training for image and text tasks, which results in inefficiencies and inadequate performance in dynamic scenarios. Barriers such as the inability to generalize across different document types and the computational overhead of ensemble methods have hindered progress. Our approach differs by proposing a unified training framework that retains the strengths of both modalities, addressing the shortcomings of prior work and enabling simultaneous recognition without prior knowledge of input content.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology, UNIT, involves a novel training framework that integrates image and text recognition capabilities within a single model. We will utilize a diverse dataset of complex document layouts, including high-resolution images with embedded text, to train the model. The evaluation will be based"
            }
        },
        "author_data": {
            "b2bc17fb-b75b-4cb6-9ff2-144b0525f016": {
                "pk": "b2bc17fb-b75b-4cb6-9ff2-144b0525f016",
                "name": "Yi Zhu",
                "collaborators": [
                    "Qile Chen",
                    "Yi Zhou",
                    "Xi Chen",
                    "Ronghua Pan",
                    "Qizheng Yin",
                    "Shawn Newsam"
                ],
                "domain": [
                    "Algebraic Geometry",
                    "Partial Differential Equations",
                    "Deep Learning",
                    "Video Understanding"
                ],
                "publications": [
                    {
                        "title": "Fano Hypersurfaces in Positive Characteristic",
                        "abstract": "We prove that a general Fano hypersurface in a projective space over an algebraically closed field of arbitrary characteristic is separably rationally connected."
                    },
                    {
                        "title": "Log rationally connected surfaces",
                        "abstract": "In this paper, combining the works of Miyanishi-Tsunoda and Keel-McKernan, we prove the log Castelnuovo's rationality criterion for smooth quasiprojective surfaces over complex numbers."
                    },
                    {
                        "title": "Homogeneous Fibrations over Curves",
                        "abstract": "By studying the theory of rational curves, we introduce a notion of rational simple connectedness for projective homogeneous spaces. As an application, we prove that over a function field of an algebraic surface, a projective homogeneous space admits a rational point if the elementary obstruction vanishes."
                    },
                    {
                        "title": "$\\mathbb{A}^1$-equivalence of zero cycles on surfaces",
                        "abstract": "In this paper, we study $\\mathbb{A}^1$-equivalence classes of zero cycles on open complex algebraic surfaces. We prove the logarithmic version of Mumford's theorem on zero cycles and prove that log Bloch's conjecture holds for quasiprojective surfaces with log Kodaira dimension $-\\infty$."
                    },
                    {
                        "title": "Global small solutions of 3D incompressible Oldroyd-B model without damping mechanism",
                        "abstract": "In this paper, we prove the global existence of small smooth solutions to the three-dimensional incompressible Oldroyd-B model without damping on the stress tensor. The main difficulty is the lack of full dissipation in stress tensor. To overcome it, we construct some time-weighted energies based on the special coupled structure of system. Such type energies show the partial dissipation of stress tensor and the strongly full dissipation of velocity. In the view of treating \"nonlinear term\" as a \"linear term\", we also apply this result to 3D incompressible viscoelastic system with Hookean elasticity and then prove the global existence of small solutions without the physical assumption (div-curl structure) as previous works."
                    },
                    {
                        "title": "Global Regularity to Incompressible Viscoelastic System With a Class of Large Initial Data",
                        "abstract": "The global existence of solutions to incompressible viscoelastic flows has been a longstanding open problem, even for the global weak solution. Under some special structure (\"div-curl\" condition) the global small smooth solution was obtained in \\cite{llzhou, cz}. However, the result with large initial data remains unknown up to now, and it is studied in this paper. We shall put forward a new structure: the cone-condition and then derive the global smooth solution to 3D incompressible viscoelastic system with a class of large initial data. The key is to gain an angle quantity in the estimate of nonlinear terms by restricting the solution to a cone."
                    },
                    {
                        "title": "Global classical solutions of 3D compressible viscoelastic system near equilibrium",
                        "abstract": "In this paper, we prove the global existence of general small solutions to compressible viscoelastic system. We remove the \"initial state\" assumption ($\\tilde \\rho_0 \\det F_0 =1$) and the \"div-curl\" structure assumption compared with previous works. It then broadens the class of solutions to a great extent, more precisely the initial density state would not be constant necessarily, and no more structure is need for global well-posedness. It's quite different from the elasticity system in which structure plays an important role. Since we can not obtain any dissipation information for density and deformation tensor, we introduce a new effective flux in the thought of regarding the wildest \"nonlinear term\" as \"linear term\". Although the norms of solution may increase now, we can still derive the global existence for it."
                    },
                    {
                        "title": "Exploring Temporal Information for Improved Video Understanding",
                        "abstract": "In this dissertation, I present my work towards exploring temporal information for better video understanding. Specifically, I have worked on two problems: action recognition and semantic segmentation. For action recognition, I have proposed a framework, termed hidden two-stream networks, to learn an optimal motion representation that does not require the computation of optical flow. My framework alleviates several challenges faced in video classification, such as learning motion representations, real-time inference, multi-framerate handling, generalizability to unseen actions, etc. For semantic segmentation, I have introduced a general framework that uses video prediction models to synthesize new training samples. By scaling up the training dataset, my trained models are more accurate and robust than previous models even without modifications to the network architectures or objective functions. I believe videos have much more potential to be mined, and temporal information is one of the most important cues for machines to perceive the visual world better."
                    },
                    {
                        "title": "Global classical solutions of 2D MHD system with only magnetic diffusion on periodic domain",
                        "abstract": "This paper studies the global existence of classical solutions to the two-dimensional incompressible magneto-hydrodynamical (MHD) system with only magnetic diffusion on the periodic domain. The approach is based on a time-weighted energy estimate, under the assumptions that the initial magnetic field is close enough to an equilibrium state and the initial data have reflection symmetry."
                    },
                    {
                        "title": "$\\mathbb{A}^1$-curves on affine complete intersections",
                        "abstract": "We generalize the results of Clemens, Ein, and Voisin regarding rational curves and zero cycles on generic projective complete intersections to the logarithmic setup."
                    },
                    {
                        "title": "Global classical solutions of 3D viscous MHD system without magnetic diffusion on periodic boxes",
                        "abstract": "In this paper, we study the global existence of classical solutions to the three dimensional incompressible viscous magneto-hydrodynamical system without magnetic diffusion on periodic boxes, i.e., with periodic boundary conditions. We work in Eulerian coordinate and employ a time-weighted energy estimate to prove the global existence result, under the assumptions that the initial magnetic field is close enough to an equilibrium state and the initial data have some symmetries."
                    },
                    {
                        "title": "Strong approximation over function fields",
                        "abstract": "By studying $\\mathbb{A}^1$-curves on varieties, we propose a geometric approach to strong approximation problem over function fields of complex curves. We prove that strong approximation holds for smooth, low degree affine complete intersections with the boundary smooth at infinity."
                    },
                    {
                        "title": "$\\mathbb{A}^1$ curves on log K3 surfaces",
                        "abstract": "In this paper, we study $\\mathbb{A}^1$ curves on log K3 surfaces. We classify all genuine log K3 surfaces of type II which admits countably infinite $\\mathbb{A}^1$ curves."
                    },
                    {
                        "title": "Very free curves on Fano Complete Intersections",
                        "abstract": "In this paper, we show that general Fano complete intersections over an algebraically closed field of arbitrary characteristics are separably rationally connected. Our proof also implies that general log Fano complete intersections with smooth tame boundary divisors admit very free $A^1$-curves."
                    },
                    {
                        "title": "On the irreducibility of the space of genus zero stable log maps to wonderful compactifications",
                        "abstract": "In this paper, we prove the moduli spaces of genus zero stable log maps to a large class of wonderful compactifications are irreducible and unirational."
                    },
                    {
                        "title": "$\\mathbb{A}^1$-equivalence of zero cycles on surfaces II",
                        "abstract": "Using recent developments in the theory of mixed motives, we prove that the log Bloch conjecture holds for an open smooth complex surface if the Bloch conjecture holds for its compactification. This verifies the log Bloch conjecture for all $\\mathbb{Q}$-homology planes and for open smooth surfaces which are not of log general type."
                    },
                    {
                        "title": "$\\mathbb{A}^1$-connected varieties over non-closed fields",
                        "abstract": "In this paper, we proved two results regarding the arithmetics of separably $\\mathbb{A}^1$-connected varieties of rank one. First we proved over a large field, there is an $\\mathbb{A}^1$-curve through any rational point of the boundary, if the boundary divisor is smooth and separably rationally connected. Secondly, we generalize a theorem of Hassett-Tschinkel for the Zariski density of integral points over function fields of curves."
                    },
                    {
                        "title": "Spatio-Temporal Sentiment Hotspot Detection Using Geotagged Photos",
                        "abstract": "We perform spatio-temporal analysis of public sentiment using geotagged photo collections. We develop a deep learning-based classifier that predicts the emotion conveyed by an image. This allows us to associate sentiment with place. We perform spatial hotspot detection and show that different emotions have distinct spatial distributions that match expectations. We also perform temporal analysis using the capture time of the photos. Our spatio-temporal hotspot detection correctly identifies emerging concentrations of specific emotions and year-by-year analyses of select locations show there are strong temporal correlations between the predicted emotions and known events."
                    }
                ]
            },
            "0f027dd5-64ca-4fc9-905e-9574796c32d3": {
                "pk": "0f027dd5-64ca-4fc9-905e-9574796c32d3",
                "name": "Yanpeng Zhou",
                "collaborators": [
                    "Haoyuan Li",
                    "Yihan Zeng",
                    "Hang Xu",
                    "Xiaodan Liang"
                ],
                "domain": [
                    "3D Shape Representation",
                    "Multimodal Learning",
                    "Point Cloud",
                    "Computer Vision"
                ],
                "publications": [
                    {
                        "title": "GS-CLIP: Gaussian Splatting for Contrastive Language-Image-3D Pretraining from Real-World Data",
                        "abstract": "3D Shape represented as point cloud has achieve advancements in multimodal pre-training to align image and language descriptions, which is curial to object identification, classification, and retrieval. However, the discrete representations of point cloud lost the object's surface shape information and creates a gap between rendering results and 2D correspondences. To address this problem, we propose GS-CLIP for the first attempt to introduce 3DGS (3D Gaussian Splatting) into multimodal pre-training to enhance 3D representation. GS-CLIP leverages a pre-trained vision-language model for a learned common visual and textual space on massive real world image-text pairs and then learns a 3D Encoder for aligning 3DGS optimized per object. Additionally, a novel Gaussian-Aware Fusion is proposed to extract and fuse global explicit feature. As a general framework for language-image-3D pre-training, GS-CLIP is agnostic to 3D backbone networks. Experiments on challenging shows that GS-CLIP significantly improves the state-of-the-art, outperforming the previously best results."
                    }
                ]
            },
            "92fe2d10-f80d-40c3-b811-04bced3c3731": {
                "pk": "92fe2d10-f80d-40c3-b811-04bced3c3731",
                "name": "Chunwei Wang",
                "collaborators": [
                    "Hang Xu",
                    "Jianhua Han",
                    "Li Zhang",
                    "Ming Nie",
                    "Yulong Liu",
                    "Renyuan Peng",
                    "Xinyue Cai",
                    "Yunlong Yuan",
                    "Yongqiang Ma",
                    "Nanning Zheng"
                ],
                "domain": [
                    "Vision-Language Models",
                    "Autonomous Driving",
                    "3D Object Detection",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Reason2Drive: Towards Interpretable and Chain-based Reasoning for Autonomous Driving",
                        "abstract": "Large vision-language models (VLMs) have garnered increasing interest in autonomous driving areas, due to their advanced capabilities in complex reasoning tasks essential for highly autonomous vehicle behavior. Despite their potential, research in autonomous systems is hindered by the lack of datasets with annotated reasoning chains that explain the decision-making processes in driving. To bridge this gap, we present Reason2Drive, a benchmark dataset with over 600K video-text pairs, aimed at facilitating the study of interpretable reasoning in complex driving environments. We distinctly characterize the autonomous driving process as a sequential combination of perception, prediction, and reasoning steps, and the question-answer pairs are automatically collected from a diverse range of open-source outdoor driving datasets, including nuScenes, Waymo and ONCE. Moreover, we introduce a novel aggregated evaluation metric to assess chain-based reasoning performance in autonomous systems, addressing the semantic ambiguities of existing metrics such as BLEU and CIDEr. Based on the proposed benchmark, we conduct experiments to assess various existing VLMs, revealing insights into their reasoning capabilities. Additionally, we develop an efficient approach to empower VLMs to leverage object-level perceptual elements in both feature extraction and prediction, further enhancing their reasoning accuracy. The code and dataset will be released."
                    },
                    {
                        "title": "From Summary to Action: Enhancing Large Language Models for Complex Tasks with Open World APIs",
                        "abstract": "The distinction between humans and animals lies in the unique ability of humans to use and create tools. Tools empower humans to overcome physiological limitations, fostering the creation of magnificent civilizations. Similarly, enabling foundational models like Large Language Models (LLMs) with the capacity to learn external tool usage may serve as a pivotal step toward realizing artificial general intelligence. Previous studies in this field have predominantly pursued two distinct approaches to augment the tool invocation capabilities of LLMs. The first approach emphasizes the construction of relevant datasets for model fine-tuning. The second approach, in contrast, aims to fully exploit the inherent reasoning abilities of LLMs through in-context learning strategies. In this work, we introduce a novel tool invocation pipeline designed to control massive real-world APIs. This pipeline mirrors the human task-solving process, addressing complicated real-life user queries. At each step, we guide LLMs to summarize the achieved results and determine the next course of action. We term this pipeline `from Summary to action', Sum2Act for short. Empirical evaluations of our Sum2Act pipeline on the ToolBench benchmark show significant performance improvements, outperforming established methods like ReAct and DFSDT. This highlights Sum2Act's effectiveness in enhancing LLMs for complex real-world tasks."
                    },
                    {
                        "title": "HiRes-LLaVA: Restoring Fragmentation Input in High-Resolution Large Vision-Language Models",
                        "abstract": "High-resolution inputs enable Large Vision-Language Models (LVLMs) to discern finer visual details, enhancing their comprehension capabilities. To reduce the training and computation costs caused by high-resolution input, one promising direction is to use sliding windows to slice the input into uniform patches, each matching the input size of the well-trained vision encoder. Although efficient, this slicing strategy leads to the fragmentation of original input, i.e., the continuity of contextual information and spatial geometry is lost across patches, adversely affecting performance in cross-patch context perception and position-specific tasks. To overcome these shortcomings, we introduce HiRes-LLaVA, a novel framework designed to efficiently process any size of high-resolution input without altering the original contextual and geometric information. HiRes-LLaVA comprises two innovative components: (i) a SliceRestore adapter that reconstructs sliced patches into their original form, efficiently extracting both global and local features via down-up-sampling and convolution layers, and (ii) a Self-Mining Sampler to compresses the vision tokens based on themselves, preserving the original context and positional information while reducing training overhead. To assess the ability of handling context fragmentation, we construct a new benchmark, EntityGrid-QA, consisting of edge-related and position-related tasks. Our comprehensive experiments demonstrate the superiority of HiRes-LLaVA on both existing public benchmarks and on EntityGrid-QA, particularly on document-oriented tasks, establishing new standards for handling high-resolution inputs."
                    },
                    {
                        "title": "Spectroscopic Evidence for Interfacial Charge Separation and Recombination in Graphene-MoS2 Vertical Heterostructures",
                        "abstract": "Vertical van der Waals (vdW) heterostructures consisting of graphene (Gr) and transition metal dichalcogenides (TMDs) have created a fascinating platform for exploring optical and electronic properties in the two-dimensional limit. Previous study has revealed the ultrafast formation of interfacial excitons and the exciton dynamics in the Gr/MoS2 heterostructure. However, a fully understanding of interfacial charge separation and the subsequent dynamics in graphene-based heterostructures remains elusive. Here, we investigate the carrier dynamics of Gr-MoS2 (including Gr/MoS2 and MoS2/Gr stacking sequences) heterostructures under different photoexcitation energies and stacking sequences by comprehensive ultrafast means, including time-resolved terahertz spectroscopy (TRTS), terahertz emission spectroscopy (TES) and transient absorption spectroscopy (TAS). We demonstrate that the Gr/MoS2 heterostructure generates hot electron injection from graphene into the MoS2 layer with photoexcitation of sub-A-exciton of MoS2, while the interfacial charge separation in the MoS2/Gr could be partially blocked by the electric field of substrate. Charge transfer (CT) occurs in same directions for the Gr-MoS2 heterostructures with opposite stacking order, resulting in the opposite orientations of the interfacial photocurrent, as directly demonstrated by the terahertz (THz) emission. Moreover, we demonstrate that the recombination time of interfacial charges after CT is on a timescale of 18 ps to 1 ns, depending on the density of defect states in MoS2 layer. This work provides a comprehensive and unambiguous picture of the interfacial charge dynamics of graphene-based heterostructures, which is essential for developing Gr/TMDs based optoelectronic devices."
                    },
                    {
                        "title": "PARTNER: Level up the Polar Representation for LiDAR 3D Object Detection",
                        "abstract": "Recently, polar-based representation has shown promising properties in perceptual tasks. In addition to Cartesian-based approaches, which separate point clouds unevenly, representing point clouds as polar grids has been recognized as an alternative due to (1) its advantage in robust performance under different resolutions and (2) its superiority in streaming-based approaches. However, state-of-the-art polar-based detection methods inevitably suffer from the feature distortion problem because of the non-uniform division of polar representation, resulting in a non-negligible performance gap compared to Cartesian-based approaches. To tackle this issue, we present PARTNER, a novel 3D object detector in the polar coordinate. PARTNER alleviates the dilemma of feature distortion with global representation re-alignment and facilitates the regression by introducing instance-level geometric information into the detection head. Extensive experiments show overwhelming advantages in streaming-based detection and different resolutions. Furthermore, our method outperforms the previous polar-based works with remarkable margins of 3.68% and 9.15% on Waymo and ONCE validation set, thus achieving competitive results over the state-of-the-art methods."
                    }
                ]
            },
            "4a354c69-c6c2-4eb4-aeba-0b4d2dba4dc9": {
                "pk": "4a354c69-c6c2-4eb4-aeba-0b4d2dba4dc9",
                "name": "Yang Cao",
                "collaborators": [
                    "Fei Xu",
                    "Yongqi Liang",
                    "Yang Li",
                    "Xiaosen Yang",
                    "Geng Tian",
                    "Bingzhe Hou",
                    "Fei Wu",
                    "Thomas Robertazzi",
                    "Jingyu Qian",
                    "Yong Yang"
                ],
                "domain": [
                    "Algebraic Geometry",
                    "Cohomology",
                    "Non-Hermitian Systems",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Cohomologie non ramifi\u00e9e de degr\u00e9 3 : vari\u00e9t\u00e9s cellulaires et surfaces de del Pezzo de degr\u00e9 au moins 5",
                        "abstract": "We consider geometrically cellular varieties $X$ over an arbitrary field of characteristic zero. We study the quotient of the third unramified cohomology group $H^3_{nr}(X,\\mathbb{Q}/\\mathbb{Z}(2))$ by its constant part. For $X$ a smooth compactification of a universal torsor over a geometrically rational surface, we show that this quotient if finite. For $X$ a del Pezzo surface of degree $\\geq 5$, we show that this quotient is zero, unless $X$ is a del Pezzo surface of degree 8 of a special type."
                    },
                    {
                        "title": "Sous-groupe de Brauer invariant et obstruction de descente it\u00e9r\u00e9e",
                        "abstract": "For a quasi-projective smooth geometrically integral variety over a number field $k$, we prove that the iterated descent obstruction is equivalent to the descent obstruction. This generalizes a result of Skorobogatov, and this answers an open question of Poonen. The key idea is the notion of invariant Brauer subgroup and the notion of invariant \\'etale Brauer-Manin obstruction for a $k$-variety equipped with an action of a connected linear algebraic group."
                    },
                    {
                        "title": "Sous-groupe de Brauer invariant pour un groupe alg\u00e9brique connexe quelconque",
                        "abstract": "In this paper, for a smooth variety equiped with an action of a connected algebraic group (not necessary linear), we introduce the notion of invariant Brauer sub-group and the notion of invariant \\'etale Brauer-Manin obstruction. Then we prove that this obstruction is equivalent to the \\'etale Brauer-Manin obstruction. This extends the main notion and the key result of [C17]. As a corollary, this also extends a result of B. Creutz."
                    },
                    {
                        "title": "Roadmap and direction towards high performance MoS2 hydrogen evolution catalysts",
                        "abstract": "MoS2 are facile, stable, non-toxic, affordable materials with reasonable price, and have exhibit its superior potential in catalysis, sensing, opt electrochemical, environmental relate application1, etc. Modification of MoS2 materials has been widely applied for adjusting and formulate the relative performance in certain application field. Making vacancy, doping with other atoms, expending of the layer space, heterojunction engineering, phase engineering etc. have been applied to facilitate the catalytic hydrogen evolution performance. We make a detailed summary on MoS2 characterization methods, synthesis, HER performance, mechanisms, to make a holistic understanding on the interplay between the structure, chemistry, and HER performance, which will give directions to elevate other electrochemical reaction by MoS2."
                    },
                    {
                        "title": "Corrigendum \u00e0 \"Sous-groupe de Brauer invariant et obstruction de descente it\u00e9r\u00e9e\" Algebra & Number Theory 14 (2020), no. 8, 2151--2183",
                        "abstract": "The \\S 2 of our work \"Sous-groupe de Brauer invariant et obstruction de descente it\\'er\\'ee\" uses a statement about the degree two K\\\"unneth formula proved by Skorobogatov and Zarhin in their paper \"The Brauer group and the Brauer-Manin set of products of varieties\". Recently they find that this statement is not correct (see their Corrigendum). We correct \\S 2 of our paper. The rest of our paper is still correct."
                    },
                    {
                        "title": "Approximation forte pour les vari\u00e9t\u00e9s avec une action d'un groupe lin\u00e9aire",
                        "abstract": "Let $G$ be a connected linear algebraic group over a number field. Let $U \\hookrightarrow X$ be a $G$-equivariant open embedding of a $G$-homogeneous space with connected stabilizers into a smooth $G$-variety. We prove that $X$ satisfies strong approximation with Brauer-Manin condition off a set $S$ of places of $k$ under either of the following hypotheses :   (i) $S$ is the set of archimedean places;   (ii) $S$ is a nonempty finite set and $\\bar{k}^{\\times}= \\bar{k}[X]^{\\times}$.   The proof builds upon the case $X=U$, which has been the object of several works."
                    },
                    {
                        "title": "Troisi\u00e8me groupe de cohomologie non ramifi\u00e9e des torseurs universels sur les surfaces rationnelles",
                        "abstract": "Let $k$ a field of characteristic zero. Let $X$ be a smooth, projective, geometrically rational $k$-surface. Let $\\mathcal{T}$ be a universal torsor over $X$ with a $k$-point et $\\mathcal{T}^c$ a smooth compactification of $\\mathcal{T}$. There is an open question: is $\\mathcal{T}^c$ $k$-birationally equivalent to a projective space? We know that the unramified cohomology groups of degree 1 and 2 of $\\mathcal{T}$ and $\\mathcal{T}^c$ are reduced to their constant part. For the analogue of the third cohomology groups, we give a sufficient condition using the Galois structure of the geometrical Picard group of $X$. This enables us to show that $H^{3}_{nr}(\\mathcal{T}^{c},\\mathbb{Q}/\\mathbb{Z}(2))/H^3(k,\\mathbb{Q}/\\mathbb{Z}(2))$ vanishes if $X$ is a generalised Ch\\^atelet surface and that this group is reduced to its $2$-primary part if $X$ is a del Pezzo surface of degree at least 2."
                    },
                    {
                        "title": "Enhanced photocatalytic activation of methane C-H bond by Incorporate Cerium into ETS-10",
                        "abstract": "Ce-ETS-10 was prepared and their ability on photocatalysis coupling of methane has been studied in detail. Ce modification enhanced the photocatalysis performance of ETS-10 on methane conversion reaction, and the methane conversion can be carried out under irradiation of visible light. To study the various rare earth modification to ETS-10, various rare earth elements were introduced to ETS-10 molecular sieve, and the relative photocatalytic methane conversion performance were studied. Various characterizations were applied to examine rare earth modification behavior. The SPS results show that the Cerium modification induce visible light photovoltage, which means the strong electron-hole separation which enhance the photocatalytic performance. The charge-hole separation is non-traditional n-n heterojunction for the charge-hole separation is highly oriented and efficient at atomic scale. This mechanism shows significant enhancement for methane conversion reaction."
                    },
                    {
                        "title": "SORSA: Singular Values and Orthonormal Regularized Singular Vectors Adaptation of Large Language Models",
                        "abstract": "The rapid advancement in large language models (LLMs) comes with a significant increase in their parameter size, presenting challenges for adaptation and fine-tuning. Parameter-efficient fine-tuning (PEFT) methods are widely used to adapt LLMs for downstream tasks efficiently. In this paper, we propose Singular Values and Orthonormal Regularized Singular Vectors Adaptation, or SORSA, a novel PEFT method. We introduce a method to analyze the variation of the parameters by performing singular value decomposition (SVD) and discuss and analyze SORSA's superiority in minimizing the alteration in the SVD aspect. Each SORSA adapter consists of two main parts: trainable principal singular weights $W_p = U_p \\text{diag}(S_p) V^\\top_p$, and frozen residual weights $W_r = U_r \\text{diag}(S_r) V^\\top_r$. These parts are initialized by performing SVD on pre-trained weights. Moreover, we implement and analyze an orthonormal regularizer, which we prove could decrease the condition number of $W_p$ and allows the optimization to be more efficient. SORSA adapters could be merged during inference, thus eliminating any inference latency. After all, SORSA shows a faster convergence than PiSSA and LoRA in our experiments. On the GSM-8K benchmark, Llama 2 7B adapted using SORSA achieved 56.03% accuracy, surpassing LoRA (42.30%), Full FT (49.05%), and PiSSA (53.07%). On the MATH benchmark, SORSA achieved 10.36% accuracy, outperforming LoRA (5.50%), Full FT (7.22%), and PiSSA (7.44%). We conclude that SORSA offers a new perspective on parameter-efficient fine-tuning, demonstrating remarkable performance. The code is available at https://github.com/Gunale0926/SORSA."
                    },
                    {
                        "title": "Strong approximation with Brauer-Manin obstruction for toric varieties",
                        "abstract": "For smooth open toric varieties, we establish strong approximation off infinity with Brauer-Manin obstruction."
                    },
                    {
                        "title": "\u00c3\u2030tale Brauer-Manin obstruction for Weil restrictions",
                        "abstract": "The \\'etale Brauer\\textendash Manin obstruction is equivalent to each other among Weil restrictions of an arbitrarily given quasi-projective algebraic variety defined over a number field."
                    },
                    {
                        "title": "Non-Hermitian Bulk-Boundary Correspondence in Periodically Driven System",
                        "abstract": "Bulk-boundary correspondence, connecting the bulk topology and the edge states, is an essential principle of the topological phases. However, the bulk-boundary correspondence is broken down in general non-Hermitian systems. In this paper, we construct one-dimensional non-Hermitian Su-Schrieffer-Heeger model with periodic driving that exhibits non-Hermitian skin effect: all the eigenstates are localized at the boundary of the systems, whether the bulk states or the zero and the $\\pi$ modes. To capture the topological properties, the non-Bloch winding numbers are defined by the non-Bloch periodized evolution operators based on the generalized Brillouin zone. Furthermore, the non-Hermitian bulk-boundary correspondence is established: the non-Bloch winding numbers ($W_{0,\\pi}$) characterize the edge states with quasienergies $\\epsilon=0, \\pi$. In our non-Hermitian system, a novel phenomenon can emerge that the robust edge states can appear even when the Floquet bands are topological trivial with zero non-Bloch band invariant, which is defined in terms of the non-Bloch effective Hamiltonian. We also show that the relation between the non-Bloch winding numbers ($W_{0,\\pi}$) and the non-Bloch band invariant ($\\mathcal{W}$): $\\mathcal{W}= W_{0}- W_{\\pi}$."
                    },
                    {
                        "title": "Schauder Bases and Operator Theory III: Schauder Spectrums",
                        "abstract": "In this paper, we study spectrums of Schauder operators. We show that we always can choose a Schauder operator in a given orbit such that the Schauder spectrum of it is empty."
                    },
                    {
                        "title": "Integrating Amdahl-like Laws and Divisible Load Theory",
                        "abstract": "Integrating Amdahl's and Amdahl-like laws with Divisible Load Theory promises mathematical design methodology that will make possible the efficient design of tomorrow's systems."
                    },
                    {
                        "title": "Formaldehyde sensing by Co3O4 hollow spheres at room temperature",
                        "abstract": "Formaldehyde is a ubiquitous and high toxicity gas. It is an essential task to efficient detect owing to their toxicity and diffusion. In this work, we studied on the detection of trace amount of formaldehyde based on hollow Co3O4 nanostructure. It is found that Co3O4 hollow spheres with different structures shows distinct sensing performance towards formaldehyde at room temperature, the response value of nanosheet modified Co3O4 towards 100 ppm formaldehyde will reach 35 in 18 second, and the nanoparticle modified Co3O4 hollow sphere will reach 2.1 in 18 second, 17 in 300 second. The nanosheet modified and nanoparticle modified Co3O4 hollow sphere will reach 4 and 20 in 10 second towards 100 ppm formaldehyde at room temperature. As room temperature, the sensors do not response towards NH3, CO, etc. The sensing mechanism was proposed based on the theoretical and experimental results. The Co3O4 sensor shows that potential utility in CH2O quick sensing at room temperature."
                    },
                    {
                        "title": "Strong approximation with Brauer-Manin obstruction for groupic varieties",
                        "abstract": "Strong approximation with Brauer-Manin obstruction is established for smooth varieties containing a connected linear algebraic group with a compatible action."
                    }
                ]
            },
            "44e05eec-9202-4112-abe3-0f549c0b8ea3": {
                "pk": "44e05eec-9202-4112-abe3-0f549c0b8ea3",
                "name": "Jianhua Han",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "639521bb-8c18-4a54-8bb3-1822c0d2e736": {
                "pk": "639521bb-8c18-4a54-8bb3-1822c0d2e736",
                "name": "Lu Hou",
                "collaborators": [
                    "James T. Kwok",
                    "Kan Zheng",
                    "Xiaojun Xu",
                    "Xianbin Wang",
                    "Quanming Yao",
                    "Zhiming Liu",
                    "Hanlin Meng",
                    "Qiang Zheng",
                    "Ning Lu",
                    "Lei Lei"
                ],
                "domain": [
                    "Deep Learning",
                    "Neural Networks",
                    "Internet of Things",
                    "Blockchain"
                ],
                "publications": [
                    {
                        "title": "Power Law in Sparsified Deep Neural Networks",
                        "abstract": "The power law has been observed in the degree distributions of many biological neural networks. Sparse deep neural networks, which learn an economical representation from the data, resemble biological neural networks in many ways. In this paper, we study if these artificial networks also exhibit properties of the power law. Experimental results on two popular deep learning models, namely, multilayer perceptrons and convolutional neural networks, are affirmative. The power law is also naturally related to preferential attachment. To study the dynamical properties of deep networks in continual learning, we propose an internal preferential attachment model to explain how the network topology evolves. Experimental results show that with the arrival of a new task, the new connections made follow this preferential attachment process."
                    },
                    {
                        "title": "Loss-aware Weight Quantization of Deep Networks",
                        "abstract": "The huge size of deep networks hinders their use in small computing devices. In this paper, we consider compressing the network by weight quantization. We extend a recently proposed loss-aware weight binarization scheme to ternarization, with possibly different scaling parameters for the positive and negative weights, and m-bit (where m > 2) quantization. Experiments on feedforward and recurrent neural networks show that the proposed scheme outperforms state-of-the-art weight quantization algorithms, and is as accurate (or even more accurate) than the full-precision network."
                    },
                    {
                        "title": "An Intelligent Transaction Migration Scheme for RAFT-based Private Blockchain in Internet of Things Applications",
                        "abstract": "The integration of multi-access edge computing (MEC) and RAFT consensus makes it feasible to deploy blockchain on trustful base stations and gateways to provide efficient and tamper-proof edge data services for Internet of Things (IoT) applications. However, reducing the latency of storing data on blockchain remains a challenge, especially when an anomalytriggered data flow in a certain area exceeds the block generation speed. This letter proposes an intelligent transaction migration scheme for RAFT-based private blockchain in IoT applications to migrate transactions in busy areas to idle regions intelligently. Simulation results show that the proposed scheme can apparently reduce the latency in high data flow circumstances."
                    },
                    {
                        "title": "Loss-aware Binarization of Deep Networks",
                        "abstract": "Deep neural network models, though very powerful and highly successful, are computationally expensive in terms of space and time. Recently, there have been a number of attempts on binarizing the network weights and activations. This greatly reduces the network size, and replaces the underlying multiplications to additions or even XNOR bit operations. However, existing binarization schemes are based on simple matrix approximation and ignore the effect of binarization on the loss. In this paper, we propose a proximal Newton algorithm with diagonal Hessian approximation that directly minimizes the loss w.r.t. the binarized weights. The underlying proximal step has an efficient closed-form solution, and the second-order information can be efficiently obtained from the second moments already computed by the Adam optimizer. Experiments on both feedforward and recurrent networks show that the proposed loss-aware binarization algorithm outperforms existing binarization schemes, and is also more robust for wide and deep networks."
                    },
                    {
                        "title": "Design and Prototype Implementation of a Blockchain-Enabled LoRa System With Edge Computing",
                        "abstract": "Efficiency and security have become critical issues during the development of the long-range (LoRa) system for Internet-of-Things (IoT) applications. The centralized work method in the LoRa system, where all packages are processed and kept in the central cloud, cannot well exploit the resources in LoRa gateways and also makes it vulnerable to security risks, such as data falsification or data loss. On the other hand, the blockchain has the potential to provide a decentralized and secure infrastructure for the LoRa system. However, there are significant challenges in deploying blockchain at LoRa gateways with limited edge computing abilities. This article proposes a design and implementation of the blockchain-enabled LoRa system with edge computing by using the open-source Hyperledger Fabric, which is called as HyperLoRa. According to different features of LoRa data, a blockchain network with multiple ledgers is designed, each of which stores a specific kind of LoRa data. LoRa gateways can participate in the operations of the blockchain and share the ledger that keep the time-critical network data with small size. Then, the edge computing abilities of LoRa gateways are utilized to handle the join procedure and application packages processing. Furthermore, a HyperLoRa prototype is implemented on embedded hardware, which demonstrates the feasibility of deploying the blockchain into LoRa gateways with limited computing and storage resources. Finally, various experiments are conducted to evaluate the performances of the proposed LoRa system."
                    },
                    {
                        "title": "Soft-Defined Heterogeneous Vehicular Network: Architecture and Challenges",
                        "abstract": "Heterogeneous Vehicular NETworks (HetVNETs) can meet various quality-of-service (QoS) requirements for intelligent transport system (ITS) services by integrating different access networks coherently. However, the current network architecture for HetVNET cannot efficiently deal with the increasing demands of rapidly changing network landscape. Thanks to the centralization and flexibility of the cloud radio access network (Cloud-RAN), soft-defined networking (SDN) can conveniently be applied to support the dynamic nature of future HetVNET functions and various applications while reducing the operating costs. In this paper, we first propose the multi-layer Cloud RAN architecture for implementing the new network, where the multi-domain resources can be exploited as needed for vehicle users. Then, the high-level design of soft-defined HetVNET is presented in detail. Finally, we briefly discuss key challenges and solutions for this new network, corroborating its feasibility in the emerging fifth-generation (5G) era."
                    }
                ]
            },
            "7f10cfe4-9c31-43d2-81f3-b068ede2aa49": {
                "pk": "7f10cfe4-9c31-43d2-81f3-b068ede2aa49",
                "name": "Hang Xu",
                "collaborators": [
                    "Wang Jia",
                    "Hamid Hezari",
                    "Yi Wang",
                    "Yannick Sire",
                    "Haoren Xiong",
                    "Alessandro Perelli",
                    "Peter Ebenfelt",
                    "Ming Xiao",
                    "Wenqi Shao",
                    "Hang Yu"
                ],
                "domain": [
                    "Mathematical Analysis",
                    "Deep Learning",
                    "Reinforcement Learning",
                    "Flow Control"
                ],
                "publications": [
                    {
                        "title": "Asymptotic properties of Bergman kernels for potentials with Gevrey regularity",
                        "abstract": "We study the asymptotic properties of the Bergman kernels associated to tensor powers of a positive line bundle on a compact K\\\"ahler manifold. We show that if the K\\\"ahler potential is in Gevrey class $G^a$ for some $a>1$, then the Bergman kernel accepts a complete asymptotic expansion in a neighborhood of the diagonal of shrinking size $k^{-\\frac12+\\frac{1}{4a+4\\varepsilon}}$ for every $\\varepsilon>0$. These improve the earlier results in the subject for smooth potentials, where an expansion exists in a $(\\frac{\\log k}{k})^{\\frac12}$ neighborhood of the diagonal. We obtain our results by finding upper bounds of the form $C^m m!^{2a+2\\varepsilon}$ for the Bergman coefficients $b_m(x, \\bar y)$ in a fixed neighborhood by the method of \\cite{BBS}. We also show that sharpening these upper bounds would improve the rate of shrinking neighborhoods of the diagonal $x=y$ in our results."
                    },
                    {
                        "title": "Quantitative upper bounds for Bergman kernels associated to smooth K\u00e4hler potentials",
                        "abstract": "We give upper bounds for the Bergman kernels associated to tensor powers of a smooth positive line bundle in terms of the rate of growth of the Taylor coefficients of the K\\\"ahler potential. As applications, we obtain improved off-diagonal rate of decay for the classes of analytic, quasi-analytic, and more generally Gevrey potentials."
                    },
                    {
                        "title": "On a property of Bergman kernels when the K\u00e4hler potential is analytic",
                        "abstract": "We provide a simple proof of a result of Rouby-Sj\\\"ostrand-Ngoc \\cite{RSN} and Deleporte \\cite{Deleporte}, which asserts that if the K\\\"ahler potential is real analytic then the Bergman kernel is an \\textit{analytic kernel} meaning that its amplitude is an \\textit{analytic symbol} and its phase is given by the polarization of the K\\\"ahler potential. This in particular shows that in the analytic case the Bergman kernel accepts an asymptotic expansion in a fixed neighborhood of the diagonal with an exponentially small remainder. The proof we provide is based on a linear recursive formula of L. Charles \\cite{Cha03} on the Bergman kernel coefficients which is similar to, but simpler than, the ones found in \\cite{BBS}."
                    },
                    {
                        "title": "The Dirichlet principle for the complex $k$-Hessian functional",
                        "abstract": "We study the variational structure of the complex $k$-Hessian equation on bounded domain $X\\subset \\mathbb C^n$ with boundary $M=\\partial X$. We prove that the Dirichlet problem $\\sigma_k (\\partial \\bar{\\partial} u) =0$ in $X$, and $u=f$ on $M$ is variational and we give an explicit construction of the associated functional $\\mathcal{E}_k(u)$. Moreover we prove $\\mathcal{E}_k(u)$ satisfies the Dirichlet principle. In a special case when $k=2$, our constructed functional $\\mathcal{E}_2(u)$ involves the Hermitian mean curvature of the boundary, the notion first introduced and studied by X. Wang. Earlier work of J. Case and and the first author of this article introduced a boundary operator for the (real) $k$-Hessian functional which satisfies the Dirichlet principle. The present paper shows that there is a parallel picture in the complex setting."
                    },
                    {
                        "title": "On a new functional for extremal metrics of the conformal laplacian in high dimension",
                        "abstract": "In this paper, we introduce a new functional for the conformal spectrum of the conformal laplacian on a closed manifold M of dimension at least 3. For this new functional we provide a Korevaar type result. The main body of the paper deals with the case of the sphere but a section is devoted to more general closed manifolds."
                    },
                    {
                        "title": "Semiclassical asymptotics for Bergman projections with Gevrey weights",
                        "abstract": "We extend the direct approach to the semiclassical asymptotics for Bergman projections, developed by Deleporte--Hitrik--Sj\\\"ostrand for real analytic exponential weights and Hitrik--Stone for smooth exponential weights, to the case of Gevrey weights. We prove that the amplitude of the asymptotic Bergman projection forms a Gevrey symbol whose asymptotic coefficients obey certain Gevrey-type growth rate, and it is constructed by an asymptotic inversion of an explicit Fourier integral operator up to a Gevrey-type small remainder."
                    },
                    {
                        "title": "Rotational Augmented Noise2Inverse for Low-dose Computed Tomography Reconstruction",
                        "abstract": "In this work, we present a novel self-supervised method for Low Dose Computed Tomography (LDCT) reconstruction. Reducing the radiation dose to patients during a CT scan is a crucial challenge since the quality of the reconstruction highly degrades because of low photons or limited measurements. Supervised deep learning methods have shown the ability to remove noise in images but require accurate ground truth which can be obtained only by performing additional high-radiation CT scans. Therefore, we propose a novel self-supervised framework for LDCT, in which ground truth is not required for training the convolutional neural network (CNN). Based on the Noise2Inverse (N2I) method, we enforce in the training loss the equivariant property of rotation transformation, which is induced by the CT imaging system, to improve the quality of the CT image in a lower dose. Numerical and experimental results show that the reconstruction accuracy of N2I with sparse views is degrading while the proposed rotational augmented Noise2Inverse (RAN2I) method keeps better image quality over a different range of sampling angles. Finally, the quantitative results demonstrate that RAN2I achieves higher image quality compared to N2I, and experimental results of RAN2I on real projection data show comparable performance to supervised learning."
                    },
                    {
                        "title": "Optimal Parallelization Strategies for Active Flow Control in Deep Reinforcement Learning-Based Computational Fluid Dynamics",
                        "abstract": "Deep Reinforcement Learning (DRL) has emerged as a promising approach for handling highly dynamic and nonlinear Active Flow Control (AFC) problems. However, the computational cost associated with training DRL models presents a significant performance bottleneck. To address this challenge and enable efficient scaling on high-performance computing architectures, this study focuses on optimizing DRL-based algorithms in parallel settings. We validate an existing state-of-the-art DRL framework used for AFC problems and discuss its efficiency bottlenecks. Subsequently, by deconstructing the overall framework and conducting extensive scalability benchmarks for individual components, we investigate various hybrid parallelization configurations and propose efficient parallelization strategies. Moreover, we refine input/output (I/O) operations in multi-environment DRL training to tackle critical overhead associated with data movement. Finally, we demonstrate the optimized framework for a typical AFC problem where near-linear scaling can be obtained for the overall framework. We achieve a significant boost in parallel efficiency from around 49% to approximately 78%, and the training process is accelerated by approximately 47 times using 60 central processing unit (CPU) cores. These findings are expected to provide valuable insights for further advancements in DRL-based AFC studies. Consequently, it continues to be a prominent and actively studied problem of significant interest."
                    },
                    {
                        "title": "Robust and Adaptive Deep Reinforcement Learning for Enhancing Flow Control around a Square Cylinder with Varying Reynolds Numbers",
                        "abstract": "The present study applies a Deep Reinforcement Learning (DRL) algorithm to Active Flow Control (AFC) of a two-dimensional flow around a confined square cylinder. Specifically, the Soft Actor-Critic (SAC) algorithm is employed to modulate the flow of a pair of synthetic jets placed on the upper and lower surfaces of the confined squared cylinder in flow configurations characterized by $Re$ of 100, 200, 300, and 400. The investigation starts with an analysis of the baseline flow in the absence of active control. It is observed that at $Re$ = 100 and $Re$ = 200, the vortex shedding exhibits mono-frequency characteristics. Conversely, at $Re$ = 300 and $Re$ = 400, the vortex shedding is dominated by multiple frequencies, which is indicative of more complex flow features. With the application of the SAC algorithm, we demonstrate the capability of DRL-based control in effectively suppressing vortex shedding, while significantly diminishing drag and fluctuations in lift. Quantitatively, the data-driven active control strategy results in a drag reduction of approximately 14.4%, 26.4%, 38.9%, and 47.0% for $Re$ = 100, 200, 300, and 400, respectively. To understand the underlying control mechanism, we also present detailed flow field comparisons, which showcase the adaptability of DRL in devising distinct control strategies tailored to the dynamic conditions at varying $Re$. These findings substantiate the proficiency of DRL in controlling chaotic, multi-frequency dominated vortex shedding phenomena, underscoring the robustness of DRL in complex AFC problems."
                    },
                    {
                        "title": "Deep reinforcement learning-based active flow control of an elliptical cylinder: transitioning from an elliptical cylinder to a circular cylinder and a flat plate",
                        "abstract": "We study the adaptability of deep reinforcement learning (DRL)-based active flow control (AFC) technology for bluff body flows with complex geometries. It is extended from a cylinder with an aspect ratio $Ar = 1$ to a flat elliptical cylinder with $Ar=2$, slender elliptical cylinders with $Ar$ less than 1, and a flat plate with $Ar=0$. We utilize the Proximal Policy Optimization (PPO) algorithm to precisely control the mass flow rates of synthetic jets located on the upper and lower surfaces of a cylinder to achieve reduction in drag, minimization of lift, and suppression of vortex shedding. Our research findings indicate that, for elliptical cylinders with $Ar$ between 1.75 and 0.75, the reduction in drag coefficient ranges from 0.9% to 15.7%, and the reduction in lift coefficient ranges from 95.2% to 99.7%. The DRL-based control strategy not only significantly reduces lift and drag, but also completely suppresses vortex shedding while using less than 1% of external excitation energy, demonstrating its efficiency and energy-saving capabilities. Additionally, for $Ar$ from 0.5 to 0, the reduction in drag coefficient ranges from 26.9% to 43.6%, and the reduction in lift coefficient from 50.2% to 68.0%. This reflects the control strategy's significant reduction in both drag and lift coefficients, while also alleviating vortex shedding. The interaction and nonlinear development of vortices in the wake of elliptical cylinders lead to complex flow instability, and DRL-based AFC technology shows adaptability and potential in addressing flow control problems for this type of bluff body flow."
                    },
                    {
                        "title": "Effect of Synthetic Jets Actuator Parameters on Deep Reinforcement Learning-Based Flow Control Performance in a Square Cylinder",
                        "abstract": "We conduct an active flow control (AFC) study on the mass flow rate of synthetic jets on the upper and lower surfaces of a square cylinder using a deep reinforcement learning (DRL) algorithm, with a focus on investigating the influence of the position and width of the synthetic jets on the flow control performance. At Reynolds numbers ($Re$) of 100 and 500, it is found that our proposed method significantly reduced the lift and drag coefficients of the square cylinder, and completely suppressed vortex shedding in the wake. In particular, at $Re=100$, placing the synthetic jets near the tail corner was beneficial for reducing drag, with a maximum drag reduction rate of 14.4%. When $Re=500$, positioning the synthetic jets near the leading edge corner resulted in a maximum optimal drag reduction effect of 65.5%. This indicates that locating the synthetic jet at the main flow separation point can achieve optimal control. Furthermore, we notice that when the synthetic jets are positioned near the tail corner, vortex shedding can be completely suppressed. Additionally, a narrower width of the synthetic jets can enhance flow instability and increase the cost of flow control."
                    },
                    {
                        "title": "Complete vortex shedding suppression in highly slender elliptical cylinders through deep reinforcement learning-driven flow control",
                        "abstract": "By leveraging the high dimensional nonlinear mapping capabilities of artificial neural networks in conjunction with the powerful control mechanisms of reinforcement learning, we attain real-time, precise modulation of synthetic jet flow rates over elliptical cylinders with varying aspect ratios (Ar). The training outcomes for elliptical cylinders with a blockage ratio of 0.24 show that for Ar=1 and 0.75, the reward function gradually increases with decreasing oscillations before stabilizing. The agent's control strategy achieved drag reduction rates of 8% and 15%, while 99% of the lift coefficient is effectively suppressed. It deserves emphasis that vortex shedding is entirely eliminated with only 0.1% and 1% of the inlet flow rate. As Ar decreases, the reinforcement learning process slows and becomes less stable, with energy consumption surging to 14.5%, while lift and drag coefficients continue oscillating and vortex shedding remains uncontrolled. When blockage ratio is reduced to 0.12, the reinforcement learning training demonstrates robust convergence and consistent full suppression of vortex shedding across all Ar from 1 to 0.1. Furthermore, drag reduction rates are observed within the range of 6.1% to 32.3%, while the lift coefficient is effectively regulated to remain at zero. For cylinders with Ar between 1 and 0.25, external energy expenditure remains below 1.4% of the inlet flow rate, signifying the realization of both efficient and energy-conservative control strategies within this range. However, for the extremely slender elliptical cylinder with Ar=0.1, the energy cost escalates to 8.1%, underscoring the significantly higher energy expenditure required to fulfill the control objectives for such highly elongated geometries."
                    },
                    {
                        "title": "Energy-Efficient Balanced Flow Control Achieved through Optimization of Synthetic Jet Placement Based on Deep Reinforcement Learning",
                        "abstract": "This study leverages deep reinforcement learning (DRL) to train synthetic jet-based flow control strategies for circular and square cylinders. The central aim is to ascertain the optimal jet placements that strike an ideal balance between energy efficiency and control effectiveness, by formulating a cohesive strategy based on a comprehensive analysis of flow control performance and energy consumption across a range of configurations. First, the results from single-action training indicate that for circle cylinder, positioning the synthetic jet approximately 105{\\deg} from the stagnation point achieves the most energy-efficient and effective control strategy. For square cylinder, placing the jet near the rear corner addresses the dual objectives of minimizing energy consumption and maximizing control performance. Second, compared to single-action control, multi-action control exhibits reduced convergence speed and stability. However, simultaneously activating synthetic jets at multiple locations significantly decreases initial energy consumption and enhances energy efficiency. The findings underscore the critical importance of accurately positioning the synthetic jet at the primary flow separation point, as this alignment not only enhances flow control performance but also establishes an optimal balance between energy efficiency and control effectiveness within the flow system. Furthermore, the interaction between synthetic jets and the flow system alters surface streamline patterns and local flow structures, resulting in effects that are considerably larger in scale than the jets themselves. This study crystallizes the potential of DRL algorithms to intelligently optimize intricate flow control strategies, effectively balancing energy efficiency with enhanced flow control performance through strategically optimized jet placement."
                    },
                    {
                        "title": "K\u00e4hler-Einstein metrics and obstruction flatness II: unit sphere bundles",
                        "abstract": "This paper concerns obstruction flatness of hypersurfaces $\\Sigma$ that arise as unit sphere bundles $S(E)$ of Griffiths negative Hermitian vector bundles $(E, h)$ over K\\\"ahler manifolds $(M, g).$ We prove that if the curvature of $(E, h)$ satisfies a splitting condition and $(M,g)$ has constant Ricci eigenvalues, then $S(E)$ is obstruction flat. If, in addition, all these eigenvalues are strictly less than one and $(M,g)$ is complete, then we show that the corresponding ball bundle admits a complete K\\\"ahler-Einstein metric."
                    },
                    {
                        "title": "BWCP: Probabilistic Learning-to-Prune Channels for ConvNets via Batch Whitening",
                        "abstract": "This work presents a probabilistic channel pruning method to accelerate Convolutional Neural Networks (CNNs). Previous pruning methods often zero out unimportant channels in training in a deterministic manner, which reduces CNN's learning capacity and results in suboptimal performance. To address this problem, we develop a probability-based pruning algorithm, called batch whitening channel pruning (BWCP), which can stochastically discard unimportant channels by modeling the probability of a channel being activated. BWCP has several merits. (1) It simultaneously trains and prunes CNNs from scratch in a probabilistic way, exploring larger network space than deterministic methods. (2) BWCP is empowered by the proposed batch whitening tool, which is able to empirically and theoretically increase the activation probability of useful channels while keeping unimportant channels unchanged without adding any extra parameters and computational cost in inference. (3) Extensive experiments on CIFAR-10, CIFAR-100, and ImageNet with various network architectures show that BWCP outperforms its counterparts by achieving better accuracy given limited computational budgets. For example, ResNet50 pruned by BWCP has only 0.70\\% Top-1 accuracy drop on ImageNet, while reducing 43.1\\% FLOPs of the plain ResNet50."
                    }
                ]
            }
        }
    },
    "2407.13623": {
        "paper_data": {
            "title": "Scaling Laws with Vocabulary: Larger Models Deserve Larger Vocabularies",
            "url": "http://arxiv.org/abs/2407.13623v2",
            "arxiv_id": "2407.13623",
            "authors": [
                "Chaofan Tao",
                "Qian Liu",
                "Longxu Dou",
                "Niklas Muennighoff",
                "Zhongwei Wan",
                "Ping Luo",
                "Min Lin",
                "Ngai Wong"
            ],
            "abstract": "Research on scaling large language models (LLMs) has primarily focused on model parameters and training data size, overlooking the role of vocabulary size. We investigate how vocabulary size impacts LLM scaling laws by training models ranging from 33M to 3B parameters on up to 500B characters with various vocabulary configurations. We propose three complementary approaches for predicting the compute-optimal vocabulary size: IsoFLOPs analysis, derivative estimation, and parametric fit of the loss function. Our approaches converge on the same result that the optimal vocabulary size depends on the available compute budget and that larger models deserve larger vocabularies. However, most LLMs use too small vocabulary sizes. For example, we predict that the optimal vocabulary size of Llama2-70B should have been at least 216K, 7 times larger than its vocabulary of 32K. We validate our predictions empirically by training models with 3B parameters across different FLOPs budgets. Adopting our predicted optimal vocabulary size consistently improves downstream performance over commonly used vocabulary sizes. By increasing the vocabulary size from the conventional 32K to 43K, we improve performance on ARC-Challenge from 29.1 to 32.0 with the same 2.3e21 FLOPs. Our work emphasizes the necessity of jointly considering model parameters and vocabulary size for efficient scaling.",
            "introduction": "   1 Introduction  Figure 2: Vocabulary parameters of popular LLMs and predicted optimal vocabulary parameters at a compute-optimal number of training tokens. Most current LLMs have suboptimal vocabulary parameters due to vocabulary sizes, which are smaller than the predicted optimal values. Among the current models, StarCoder2-3B, OLMo-7B, InternLM2-20B, and Gemma2-27B have vocabulary sizes that come closest to the optimal allocation for their respective model sizes.   Large language models (LLMs) achieve remarkable performance by pre-training on vast text corpora using massive computational resources\u00a0[46]. Extensive prior work on LLMs has focused on deriving so-called scaling laws: a set of empirical formulas to predict how model performance scales, mainly as computing floating-point operations (FLOPs), model parameters, and quantity of training data change\u00a0[30, 26, 63, 2, 43, 57]. These works show that power-law fits can effectively predict language modeling loss and by extension downstream performance\u00a0[23, 54]. However, these scaling laws usually disregard the impact of the vocabulary size. For example, in\u00a0Kaplan et\u00a0al. [30] only non-vocabulary parameters are considered in their predictive formula. This negligence has resulted in substantial variability in the vocabulary size of current LLMs. For instance, Llama2-7B employs a vocabulary size of 32K\u00a0[67], while Gemma-7B\u00a0[64] adopts a much larger vocabulary size of 256K despite both having a similar number of total parameters. This variability in vocabulary sizes across LLMs raises the research question: What is the compute-optimal vocabulary size for an LLM?   The vocabulary size affects performance non-trivially. Intuitively, the optimal vocabulary size should neither be excessively large nor small. A larger vocabulary size improves tokenization fertility, i.e., splitting sentences into fewer tokens, thereby improving the tokenization efficiency. Additionally, a larger vocabulary enhances the representational capacity of the model, enabling it to capture a wider range of concepts and nuances in the corpus. However, the risk of under-fitting representations for rare tokens increases with larger vocabulary sizes, especially in the data-constrained regime\u00a0[43, 69]. Thus, the optimal vocabulary size needs to be determined by taking the training data and the amount of non-vocabulary parameters into account.   In this paper, we show that the effect of vocabulary on scaling laws has been underestimated, and we quantify the effect to derive a prediction for the optimal vocabulary size. We first introduce a normalized loss formulation to ensure a fair comparison across models with varying vocabulary sizes. Utilizing the normalized loss function, we analyze and discuss the underlying rationale behind the existence of an optimal vocabulary size, which depends on the available computational budget.   To predict the optimal vocabulary size given a compute budget, we propose three approaches. Approach 1 (Estimating power laws via IsoFLOPs): We pre-train models with non-vocabulary parameters ranging from 33M to 1.13B, with groups of models that share the same FLOPs (\u201cIsoFLOPs\u201d) but varying vocabulary configurations. Then, we fit power laws relating FLOPs to non-vocabulary parameters, vocabulary parameters, and training data, respectively. Our analysis reveals that the optimal vocabulary parameters exhibit a power-law growth with respect to the computational budget, however, at a slower rate than non-vocabulary parameters, as shown in\u00a0Figure\u00a01. Approach 2 (Derivative-based Estimation): We introduce a derivative-based method that estimates the optimal vocabulary size by using the derivative of FLOPs",
            "references": [
                {
                    "title": "BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructions",
                    "abstract": "Task automation has been greatly empowered by the recent advances in Large Language Models (LLMs) via Python code, where the tasks ranging from software engineering development to general-purpose reasoning. While current benchmarks have shown that LLMs can solve tasks using programs like human developers, the majority of their evaluations are limited to short and self-contained algorithmic tasks or standalone function calls. Solving challenging and practical requires the capability of utilizing diverse function calls as tools to efficiently implement functionalities like data analysis and web development. In addition, using multiple tools to solve a task needs compositional reasoning by accurately understanding complex instructions. Fulfilling both of these characteristics can pose a great challenge for LLMs.To assess how well LLMs can solve challenging and practical tasks via programs, we introduce BigCodeBench, a benchmark that challenges LLMs to invoke multiple function calls as tools from 139 libraries and 7 domains for 1,140 fine-grained tasks. To evaluate LLMs rigorously, each task encompasses 5.6 test cases with an average branch coverage of 99%. In addition, we propose a natural-language-oriented variant of BigCodeBench, BigCodeBench-Instruct, that automatically transforms the original docstrings into short instructions only with essential information. Our extensive evaluation of 60 LLMs shows that LLMs are not yet capable of following complex instructions to use function calls precisely, with scores up to 60%, significantly lower than the human performance of 97%. The results underscore the need for further advancements in this area."
                },
                {
                    "title": "DataComp-LM: In search of the next generation of training sets for language models",
                    "abstract": "We introduce DataComp for Language Models (DCLM), a testbed for controlled dataset experiments with the goal of improving language models. As part of DCLM, we provide a standardized corpus of 240T tokens extracted from Common Crawl, effective pretraining recipes based on the OpenLM framework, and a broad suite of 53 downstream evaluations. Participants in the DCLM benchmark can experiment with data curation strategies such as deduplication, filtering, and data mixing at model scales ranging from 412M to 7B parameters. As a baseline for DCLM, we conduct extensive experiments and find that model-based filtering is key to assembling a high-quality training set. The resulting dataset, DCLM-Baseline enables training a 7B parameter language model from scratch to 64% 5-shot accuracy on MMLU with 2.6T training tokens. Compared to MAP-Neo, the previous state-of-the-art in open-data language models, DCLM-Baseline represents a 6.6 percentage point improvement on MMLU while being trained with 40% less compute. Our baseline model is also comparable to Mistral-7B-v0.3 and Llama 3 8B on MMLU (63%&66%), and performs similarly on an average of 53 natural language understanding tasks while being trained with 6.6x less compute than Llama 3 8B. Our results highlight the importance of dataset design for training language models and offer a starting point for further research on data curation."
                },
                {
                    "title": "Observational Scaling Laws and the Predictability of Language Model Performance",
                    "abstract": "Understanding how language model performance varies with scale is critical to benchmark and algorithm development. Scaling laws are one approach to building this understanding, but the requirement of training models across many different scales has limited their use. We propose an alternative, observational approach that bypasses model training and instead builds scaling laws from ~100 publically available models. Building a single scaling law from multiple model families is challenging due to large variations in their training compute efficiencies and capabilities. However, we show that these variations are consistent with a simple, generalized scaling law where language model performance is a function of a low-dimensional capability space, and model families only vary in their efficiency in converting training compute to capabilities. Using this approach, we show the surprising predictability of complex scaling phenomena: we show that several emergent phenomena follow a smooth, sigmoidal behavior and are predictable from small models; we show that the agent performance of models such as GPT-4 can be precisely predicted from simpler non-agentic benchmarks; and we show how to predict the impact of post-training interventions like Chain-of-Thought and Self-Consistency as language model capabilities continue to improve."
                },
                {
                    "title": "Eagle and Finch: RWKV with Matrix-Valued States and Dynamic Recurrence",
                    "abstract": "We present Eagle (RWKV-5) and Finch (RWKV-6), sequence models improving upon the RWKV (RWKV-4) architecture. Our architectural design advancements include multi-headed matrix-valued states and a dynamic recurrence mechanism that improve expressivity while maintaining the inference efficiency characteristics of RNNs. We introduce a new multilingual corpus with 1.12 trillion tokens and a fast tokenizer based on greedy matching for enhanced multilinguality. We trained four Eagle models, ranging from 0.46 to 7.5 billion parameters, and two Finch models with 1.6 and 3.1 billion parameters and find that they achieve competitive performance across a wide variety of benchmarks. We release all our models on HuggingFace under the Apache 2.0 license. Models at: https://huggingface.co/RWKV Training code at: https://github.com/RWKV/RWKV-LM Inference code at: https://github.com/RWKV/ChatRWKV Time-parallel training code at: https://github.com/RWKV/RWKV-infctx-trainer"
                },
                {
                    "title": "Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction",
                    "abstract": "We present Visual AutoRegressive modeling (VAR), a new generation paradigm that redefines the autoregressive learning on images as coarse-to-fine\"next-scale prediction\"or\"next-resolution prediction\", diverging from the standard raster-scan\"next-token prediction\". This simple, intuitive methodology allows autoregressive (AR) transformers to learn visual distributions fast and generalize well: VAR, for the first time, makes GPT-like AR models surpass diffusion transformers in image generation. On ImageNet 256x256 benchmark, VAR significantly improve AR baseline by improving Frechet inception distance (FID) from 18.65 to 1.73, inception score (IS) from 80.4 to 350.2, with around 20x faster inference speed. It is also empirically verified that VAR outperforms the Diffusion Transformer (DiT) in multiple dimensions including image quality, inference speed, data efficiency, and scalability. Scaling up VAR models exhibits clear power-law scaling laws similar to those observed in LLMs, with linear correlation coefficients near -0.998 as solid evidence. VAR further showcases zero-shot generalization ability in downstream tasks including image in-painting, out-painting, and editing. These results suggest VAR has initially emulated the two important properties of LLMs: Scaling Laws and zero-shot task generalization. We have released all models and codes to promote the exploration of AR/VAR models for visual generation and unified learning."
                },
                {
                    "title": "Bigger is not Always Better: Scaling Properties of Latent Diffusion Models",
                    "abstract": "We study the scaling properties of latent diffusion models (LDMs) with an emphasis on their sampling efficiency. While improved network architecture and inference algorithms have shown to effectively boost sampling efficiency of diffusion models, the role of model size -- a critical determinant of sampling efficiency -- has not been thoroughly examined. Through empirical analysis of established text-to-image diffusion models, we conduct an in-depth investigation into how model size influences sampling efficiency across varying sampling steps. Our findings unveil a surprising trend: when operating under a given inference budget, smaller models frequently outperform their larger equivalents in generating high-quality results. Moreover, we extend our study to demonstrate the generalizability of the these findings by applying various diffusion samplers, exploring diverse downstream tasks, evaluating post-distilled models, as well as comparing performance relative to training compute. These findings open up new pathways for the development of LDM scaling strategies which can be employed to enhance generative capabilities within limited inference budgets."
                },
                {
                    "title": "Language models scale reliably with over-training and on downstream tasks",
                    "abstract": "Scaling laws are useful guides for derisking expensive training runs, as they predict performance of large models using cheaper, small-scale experiments. However, there remain gaps between current scaling studies and how language models are ultimately trained and evaluated. For instance, scaling is usually studied in the compute-optimal training regime (i.e.,\"Chinchilla optimal\"regime). In contrast, models are often over-trained to reduce inference costs. Moreover, scaling laws mostly predict loss on next-token prediction, but models are usually compared on downstream task performance. To address both shortcomings, we create a testbed of 104 models with 0.011B to 6.9B parameters trained with various numbers of tokens on three data distributions. First, we fit scaling laws that extrapolate in both the amount of over-training and the number of model parameters. This enables us to predict the validation loss of a 1.4B parameter, 900B token run (i.e., 32$\\times$ over-trained) and a 6.9B parameter, 138B token run (i.e., a compute-optimal run)$\\unicode{x2014}$each from experiments that take 300$\\times$ less compute. Second, we relate the perplexity of a language model to its downstream task performance by proposing a power law. We use this law to predict top-1 error averaged over downstream tasks for the two aforementioned models, using experiments that take 20$\\times$ less compute. Our experiments are available at https://github.com/mlfoundations/scaling."
                },
                {
                    "title": "StarCoder 2 and The Stack v2: The Next Generation",
                    "abstract": "The BigCode project, an open-scientific collaboration focused on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder2. In partnership with Software Heritage (SWH), we build The Stack v2 on top of the digital commons of their source code archive. Alongside the SWH repositories spanning 619 programming languages, we carefully select other high-quality data sources, such as GitHub pull requests, Kaggle notebooks, and code documentation. This results in a training set that is 4x larger than the first StarCoder dataset. We train StarCoder2 models with 3B, 7B, and 15B parameters on 3.3 to 4.3 trillion tokens and thoroughly evaluate them on a comprehensive set of Code LLM benchmarks. We find that our small model, StarCoder2-3B, outperforms other Code LLMs of similar size on most benchmarks, and also outperforms StarCoderBase-15B. Our large model, StarCoder2- 15B, significantly outperforms other models of comparable size. In addition, it matches or outperforms CodeLlama-34B, a model more than twice its size. Although DeepSeekCoder- 33B is the best-performing model at code completion for high-resource languages, we find that StarCoder2-15B outperforms it on math and code reasoning benchmarks, as well as several low-resource languages. We make the model weights available under an OpenRAIL license and ensure full transparency regarding the training data by releasing the SoftWare Heritage persistent IDentifiers (SWHIDs) of the source code data."
                },
                {
                    "title": "Aya Model: An Instruction Finetuned Open-Access Multilingual Language Model",
                    "abstract": "Recent breakthroughs in large language models (LLMs) have centered around a handful of data-rich languages. What does it take to broaden access to breakthroughs beyond first-class citizen languages? Our work introduces Aya, a massively multilingual generative language model that follows instructions in 101 languages of which over 50% are considered as lower-resourced. Aya outperforms mT0 and BLOOMZ on the majority of tasks while covering double the number of languages. We introduce extensive new evaluation suites that broaden the state-of-art for multilingual eval across 99 languages -- including discriminative and generative tasks, human evaluation, and simulated win rates that cover both held-out tasks and in-distribution performance. Furthermore, we conduct detailed investigations on the optimal finetuning mixture composition, data pruning, as well as the toxicity, bias, and safety of our models. We open-source our instruction datasets and our model at https://hf.co/CohereForAI/aya-101"
                },
                {
                    "title": "Dolma: an Open Corpus of Three Trillion Tokens for Language Model Pretraining Research",
                    "abstract": "Information about pretraining corpora used to train the current best-performing language models is seldom discussed: commercial models rarely detail their data, and even open models are often released without accompanying training data or recipes to reproduce them. As a result, it is challenging to conduct and advance scientific research on language modeling, such as understanding how training data impacts model capabilities and limitations. To facilitate scientific research on language model pretraining, we curate and release Dolma, a three-trillion-token English corpus, built from a diverse mixture of web content, scientific papers, code, public-domain books, social media, and encyclopedic materials. We extensively document Dolma, including its design principles, details about its construction, and a summary of its contents. We present analyses and experimental results on intermediate states of Dolma to share what we have learned about important data curation practices. Finally, we open-source our data curation toolkit to enable reproduction of our work as well as support further research in large-scale data curation."
                },
                {
                    "title": "Breaking the Curse of Multilinguality with Cross-lingual Expert Language Models",
                    "abstract": "Despite their popularity in non-English NLP, multilingual language models often underperform monolingual ones due to inter-language competition for model parameters. We propose Cross-lingual Expert Language Models (X-ELM), which mitigate this competition by independently training language models on subsets of the multilingual corpus. This process specializes X-ELMs to different languages while remaining effective as a multilingual ensemble. Our experiments show that when given the same compute budget, X-ELM outperforms jointly trained multilingual models across all considered languages and that these gains transfer to downstream tasks. X-ELM provides additional benefits over performance improvements: new experts can be iteratively added, adapting X-ELM to new languages without catastrophic forgetting. Furthermore, training is asynchronous, reducing the hardware requirements for multilingual training and democratizing multilingual modeling."
                },
                {
                    "title": "DeepSeek LLM: Scaling Open-Source Language Models with Longtermism",
                    "abstract": "The rapid development of open-source large language models (LLMs) has been truly remarkable. However, the scaling law described in previous literature presents varying conclusions, which casts a dark cloud over scaling LLMs. We delve into the study of scaling laws and present our distinctive findings that facilitate scaling of large scale models in two commonly used open-source configurations, 7B and 67B. Guided by the scaling laws, we introduce DeepSeek LLM, a project dedicated to advancing open-source language models with a long-term perspective. To support the pre-training phase, we have developed a dataset that currently consists of 2 trillion tokens and is continuously expanding. We further conduct supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) on DeepSeek LLM Base models, resulting in the creation of DeepSeek Chat models. Our evaluation results demonstrate that DeepSeek LLM 67B surpasses LLaMA-2 70B on various benchmarks, particularly in the domains of code, mathematics, and reasoning. Furthermore, open-ended evaluations reveal that DeepSeek LLM 67B Chat exhibits superior performance compared to GPT-3.5."
                },
                {
                    "title": "TinyLlama: An Open-Source Small Language Model",
                    "abstract": "We present TinyLlama, a compact 1.1B language model pretrained on around 1 trillion tokens for approximately 3 epochs. Building on the architecture and tokenizer of Llama 2, TinyLlama leverages various advances contributed by the open-source community (e.g., FlashAttention and Lit-GPT), achieving better computational efficiency. Despite its relatively small size, TinyLlama demonstrates remarkable performance in a series of downstream tasks. It significantly outperforms existing open-source language models with comparable sizes. Our model checkpoints and code are publicly available on GitHub at https://github.com/jzhang38/TinyLlama."
                },
                {
                    "title": "Beyond Chinchilla-Optimal: Accounting for Inference in Language Model Scaling Laws",
                    "abstract": "Large language model (LLM) scaling laws are empirical formulas that estimate changes in model quality as a result of increasing parameter count and training data. However, these formulas, including the popular Deepmind Chinchilla scaling laws, neglect to include the cost of inference. We modify the Chinchilla scaling laws to calculate the optimal LLM parameter count and pre-training data size to train and deploy a model of a given quality and inference demand. We conduct our analysis both in terms of a compute budget and real-world costs and find that LLM researchers expecting reasonably large inference demand (~1B requests) should train models smaller and longer than Chinchilla-optimal. Furthermore, we train 47 models of varying sizes and parameter counts to validate our formula and find that model quality continues to improve as we scale tokens per parameter to extreme ranges (up to 10,000). Finally, we ablate the procedure used to fit the Chinchilla scaling law coefficients and find that developing scaling laws only from data collected at typical token/parameter ratios overestimates the impact of additional tokens at these extreme ranges."
                },
                {
                    "title": "Efficient Large Language Models: A Survey",
                    "abstract": "Large Language Models (LLMs) have demonstrated remarkable capabilities in important tasks such as natural language understanding and language generation, and thus have the potential to make a substantial impact on our society. Such capabilities, however, come with the considerable resources they demand, highlighting the strong need to develop effective techniques for addressing their efficiency challenges. In this survey, we provide a systematic and comprehensive review of efficient LLMs research. We organize the literature in a taxonomy consisting of three main categories, covering distinct yet interconnected efficient LLMs topics from model-centric, data-centric, and framework-centric perspective, respectively. We have also created a GitHub repository where we organize the papers featured in this survey at https://github.com/AIoT-MLSys-Lab/Efficient-LLMs-Survey. We will actively maintain the repository and incorporate new research as it emerges. We hope our survey can serve as a valuable resource to help researchers and practitioners gain a systematic understanding of efficient LLMs research and inspire them to contribute to this important and exciting field."
                },
                {
                    "title": "The Falcon Series of Open Language Models",
                    "abstract": "We introduce the Falcon series: 7B, 40B, and 180B parameters causal decoder-only models trained on a diverse high-quality corpora predominantly assembled from web data. The largest model, Falcon-180B, has been trained on over 3.5 trillion tokens of text--the largest openly documented pretraining run. Falcon-180B significantly outperforms models such as PaLM or Chinchilla, and improves upon concurrently developed models such as LLaMA 2 or Inflection-1. It nears the performance of PaLM-2-Large at a reduced pretraining and inference cost, making it, to our knowledge, one of the three best language models in the world along with GPT-4 and PaLM-2-Large. We report detailed evaluations, as well as a deep dive into the methods and custom tooling employed to pretrain Falcon. Notably, we report on our custom distributed training codebase, allowing us to efficiently pretrain these models on up to 4,096 A100s on cloud AWS infrastructure with limited interconnect. We release a 600B tokens extract of our web dataset, as well as the Falcon-7/40/180B models under a permissive license to foster open-science and accelerate the development of an open ecosystem of large language models."
                },
                {
                    "title": "FinGPT: Large Generative Models for a Small Language",
                    "abstract": "Large language models (LLMs) excel in many tasks in NLP and beyond, but most open models have very limited coverage of smaller languages and LLM work tends to focus on languages where nearly unlimited data is available for pretraining. In this work, we study the challenges of creating LLMs for Finnish, a language spoken by less than 0.1% of the world population. We compile an extensive dataset of Finnish combining web crawls, news, social media and eBooks. We pursue two approaches to pretrain models: 1) we train seven monolingual models from scratch (186M to 13B parameters) dubbed FinGPT, 2) we continue the pretraining of the multilingual BLOOM model on a mix of its original training data and Finnish, resulting in a 176 billion parameter model we call BLUUMI. For model evaluation, we introduce FIN-bench, a version of BIG-bench with Finnish tasks. We also assess other model qualities such as toxicity and bias. Our models and tools are openly available at https://turkunlp.org/gpt3-finnish."
                },
                {
                    "title": "Llemma: An Open Language Model For Mathematics",
                    "abstract": "We present Llemma, a large language model for mathematics. We continue pretraining Code Llama on the Proof-Pile-2, a mixture of scientific papers, web data containing mathematics, and mathematical code, yielding Llemma. On the MATH benchmark Llemma outperforms all known open base models, as well as the unreleased Minerva model suite on an equi-parameter basis. Moreover, Llemma is capable of tool use and formal theorem proving without any further finetuning. We openly release all artifacts, including 7 billion and 34 billion parameter models, the Proof-Pile-2, and code to replicate our experiments."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "To Repeat or Not To Repeat: Insights from Scaling LLM under Token-Crisis",
                    "abstract": "Recent research has highlighted the importance of dataset size in scaling language models. However, large language models (LLMs) are notoriously token-hungry during pre-training, and high-quality text data on the web is approaching its scaling limit for LLMs. To further enhance LLMs, a straightforward approach is to repeat the pre-training data for additional epochs. In this study, we empirically investigate three key aspects under this approach. First, we explore the consequences of repeating pre-training data, revealing that the model is susceptible to overfitting, leading to multi-epoch degradation. Second, we examine the key factors contributing to multi-epoch degradation, finding that significant factors include dataset size, model parameters, and training objectives, while less influential factors consist of dataset quality and model FLOPs. Finally, we explore whether widely used regularization can alleviate multi-epoch degradation. Most regularization techniques do not yield significant improvements, except for dropout, which demonstrates remarkable effectiveness but requires careful tuning when scaling up the model size. Additionally, we discover that leveraging mixture-of-experts (MoE) enables cost-effective and efficient hyper-parameter tuning for computationally intensive dense LLMs with comparable trainable parameters, potentially impacting efficient LLM development on a broader scale."
                },
                {
                    "title": "MEGABYTE: Predicting Million-byte Sequences with Multiscale Transformers",
                    "abstract": "Autoregressive transformers are spectacular models for short sequences but scale poorly to long sequences such as high-resolution images, podcasts, code, or books. We proposed Megabyte, a multi-scale decoder architecture that enables end-to-end differentiable modeling of sequences of over one million bytes. Megabyte segments sequences into patches and uses a local submodel within patches and a global model between patches. This enables sub-quadratic self-attention, much larger feedforward layers for the same compute, and improved parallelism during decoding -- unlocking better performance at reduced cost for both training and generation. Extensive experiments show that Megabyte allows byte-level models to perform competitively with subword models on long context language modeling, achieve state-of-the-art density estimation on ImageNet, and model audio from raw files. Together, these results establish the viability of tokenization-free autoregressive sequence modeling at scale."
                },
                {
                    "title": "StarCoder: may the source be with you!",
                    "abstract": "The BigCode community, an open-scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder and StarCoderBase: 15.5B parameter models with 8K context length, infilling capabilities and fast large-batch inference enabled by multi-query attention. StarCoderBase is trained on 1 trillion tokens sourced from The Stack, a large collection of permissively licensed GitHub repositories with inspection tools and an opt-out process. We fine-tuned StarCoderBase on 35B Python tokens, resulting in the creation of StarCoder. We perform the most comprehensive evaluation of Code LLMs to date and show that StarCoderBase outperforms every open Code LLM that supports multiple programming languages and matches or outperforms the OpenAI code-cushman-001 model. Furthermore, StarCoder outperforms every model that is fine-tuned on Python, can be prompted to achieve 40\\% pass@1 on HumanEval, and still retains its performance on other programming languages. We take several important steps towards a safe open-access model release, including an improved PII redaction pipeline and a novel attribution tracing tool, and make the StarCoder models publicly available under a more commercially viable version of the Open Responsible AI Model license."
                },
                {
                    "title": "Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling",
                    "abstract": "How do large language models (LLMs) develop and evolve over the course of training? How do these patterns change as models scale? To answer these questions, we introduce \\textit{Pythia}, a suite of 16 LLMs all trained on public data seen in the exact same order and ranging in size from 70M to 12B parameters. We provide public access to 154 checkpoints for each one of the 16 models, alongside tools to download and reconstruct their exact training dataloaders for further study. We intend \\textit{Pythia} to facilitate research in many areas, and we present several case studies including novel results in memorization, term frequency effects on few-shot performance, and reducing gender bias. We demonstrate that this highly controlled setup can be used to yield novel insights toward LLMs and their training dynamics. Trained models, analysis code, training code, and training data can be found at \\url{https://github.com/EleutherAI/pythia}."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "XLM-V: Overcoming the Vocabulary Bottleneck in Multilingual Masked Language Models",
                    "abstract": "Large multilingual language models typically rely on a single vocabulary shared across 100+ languages. As these models have increased in parameter count and depth, vocabulary size has remained largely unchanged. This \\textit{vocabulary bottleneck} limits the representational capabilities of multilingual models like XLM-R. In this paper, we introduce a new approach for scaling to very large multilingual vocabularies by de-emphasizing token sharing between languages with little lexical overlap and assigning vocabulary capacity to achieve sufficient coverage for each individual language. Tokenizations using our vocabulary are typically more semantically meaningful and shorter compared to XLM-R. Leveraging this improved vocabulary, we train XLM-V, a multilingual language model with a one million token vocabulary. XLM-V outperforms XLM-R on every task we tested on ranging from natural language inference (XNLI), question answering (MLQA, XQuAD, TyDiQA), to named entity recognition (WikiAnn). XLM-V is particularly effective on low-resource language tasks and outperforms XLM-R by 11.2% and 5.8% absolute on MasakhaNER and Americas NLI, respectively."
                },
                {
                    "title": "Scaling Laws for Generative Mixed-Modal Language Models",
                    "abstract": "Generative language models define distributions over sequences of tokens that can represent essentially any combination of data modalities (e.g., any permutation of image tokens from VQ-VAEs, speech tokens from HuBERT, BPE tokens for language or code, and so on). To better understand the scaling properties of such mixed-modal models, we conducted over 250 experiments using seven different modalities and model sizes ranging from 8 million to 30 billion, trained on 5-100 billion tokens. We report new mixed-modal scaling laws that unify the contributions of individual modalities and the interactions between them. Specifically, we explicitly model the optimal synergy and competition due to data and model size as an additive term to previous uni-modal scaling laws. We also find four empirical phenomena observed during the training, such as emergent coordinate-ascent style training that naturally alternates between modalities, guidelines for selecting critical hyper-parameters, and connections between mixed-modal competition and training stability. Finally, we test our scaling law by training a 30B speech-text model, which significantly outperforms the corresponding unimodal models. Overall, our research provides valuable insights into the design and training of mixed-modal generative models, an important new class of unified models that have unique distributional properties."
                },
                {
                    "title": "SantaCoder: don't reach for the stars!",
                    "abstract": "The BigCode project is an open-scientific collaboration working on the responsible development of large language models for code. This tech report describes the progress of the collaboration until December 2022, outlining the current state of the Personally Identifiable Information (PII) redaction pipeline, the experiments conducted to de-risk the model architecture, and the experiments investigating better preprocessing methods for the training data. We train 1.1B parameter models on the Java, JavaScript, and Python subsets of The Stack and evaluate them on the MultiPL-E text-to-code benchmark. We find that more aggressive filtering of near-duplicates can further boost performance and, surprisingly, that selecting files from repositories with 5+ GitHub stars deteriorates performance significantly. Our best model outperforms previous open-source multilingual code generation models (InCoder-6.7B and CodeGen-Multi-2.7B) in both left-to-right generation and infilling on the Java, JavaScript, and Python portions of MultiPL-E, despite being a substantially smaller model. All models are released under an OpenRAIL license at https://hf.co/bigcode."
                },
                {
                    "title": "Reproducible Scaling Laws for Contrastive Language-Image Learning",
                    "abstract": "Scaling up neural networks has led to remarkable performance across a wide range of tasks. Moreover, performance often follows reliable scaling laws as a function of training set size, model size, and compute, which offers valuable guidance as large-scale experiments are becoming increasingly expensive. However, previous work on scaling laws has primarily used private data & models or focused on uni-modal language or vision learning. To address these limitations, we investigate scaling laws for contrastive language-image pre-training (CLIP) with the public LAION dataset and the open-source OpenCLIP repository. Our large-scale experiments involve models trained on up to two billion image-text pairs and identify power law scaling for multiple downstream tasks including zero-shot classification, retrieval, linear probing, and end-to-end fine-tuning. We find that the training distribution plays a key role in scaling laws as the OpenAI and OpenCLIP models exhibit different scaling behavior despite identical model architectures and similar training recipes. We open-source our evaluation workflow and all models, including the largest public CLIP models, to ensure reproducibility and make scaling laws research more accessible. Source code and instructions to reproduce this study is available at https://github.eom/LAION-AI/sealing-laws-openelip."
                },
                {
                    "title": "BLOOM: A 176B-Parameter Open-Access Multilingual Language Model",
                    "abstract": "Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are frequently kept from the public. As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model designed and built thanks to a collaboration of hundreds of researchers. BLOOM is a decoder-only Transformer language model that was trained on the ROOTS corpus, a dataset comprising hundreds of sources in 46 natural and 13 programming languages (59 in total). We find that BLOOM achieves competitive performance on a wide variety of benchmarks, with stronger results after undergoing multitask prompted finetuning. To facilitate future research and applications using LLMs, we publicly release our models and code under the Responsible AI License."
                },
                {
                    "title": "What Language Model to Train if You Have One Million GPU Hours?",
                    "abstract": "The crystallization of modeling methods around the Transformer architecture has been a boon for practitioners. Simple, well-motivated architectural variations can transfer across tasks and scale, increasing the impact of modeling research. However, with the emergence of state-of-the-art 100B+ parameters models, large language models are increasingly expensive to accurately design and train. Notably, it can be difficult to evaluate how modeling decisions may impact emergent capabilities, given that these capabilities arise mainly from sheer scale alone. In the process of building BLOOM--the Big Science Large Open-science Open-access Multilingual language model--our goal is to identify an architecture and training setup that makes the best use of our 1,000,000 A100-GPU-hours budget. Specifically, we perform an ablation study at the billion-parameter scale comparing different modeling practices and their impact on zero-shot generalization. In addition, we study the impact of various popular pre-training corpora on zero-shot generalization. We also study the performance of a multilingual model and how it compares to the English-only one. Finally, we consider the scaling behaviour of Transformers to choose the target model size, shape, and training setup. All our models and code are open-sourced at https://huggingface.co/bigscience ."
                },
                {
                    "title": "Scaling Laws vs Model Architectures: How does Inductive Bias Influence Scaling?",
                    "abstract": "There have been a lot of interest in the scaling properties of Transformer models. However, not much has been done on the front of investigating the effect of scaling properties of different inductive biases and model architectures. Do model architectures scale differently? If so, how does inductive bias affect scaling behaviour? How does this influence upstream (pretraining) and downstream (transfer)? This paper conducts a systematic study of scaling behaviour of ten diverse model architectures such as Transformers, Switch Transformers, Universal Transformers, Dynamic convolutions, Performers, and recently proposed MLP-Mixers. Via extensive experiments, we show that (1) architecture is an indeed an important consideration when performing scaling and (2) the best performing model can fluctuate at different scales. We believe that the findings outlined in this work has significant implications to how model architectures are currently evaluated in the community."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "Training language models to follow instructions with human feedback",
                    "abstract": "Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent."
                },
                {
                    "title": "CM3: A Causal Masked Multimodal Model of the Internet",
                    "abstract": "We introduce CM3, a family of causally masked generative models trained over a large corpus of structured multi-modal documents that can contain both text and image tokens. Our new causally masked approach generates tokens left to right while also masking out a small number of long token spans that are generated at the end of the string, instead of their original positions. The casual masking object provides a type of hybrid of the more common causal and masked language models, by enabling full generative modeling while also providing bidirectional context when generating the masked spans. We train causally masked language-image models on large-scale web and Wikipedia articles, where each document contains all of the text, hypertext markup, hyperlinks, and image tokens (from a VQVAE-GAN), provided in the order they appear in the original HTML source (before masking). The resulting CM3 models can generate rich structured, multi-modal outputs while conditioning on arbitrary masked document contexts, and thereby implicitly learn a wide range of text, image, and cross modal tasks. They can be prompted to recover, in a zero-shot fashion, the functionality of models such as DALL-E, GENRE, and HTLM. We set the new state-of-the-art in zero-shot summarization, entity linking, and entity disambiguation while maintaining competitive performance in the fine-tuning setting. We can generate images unconditionally, conditioned on text (like DALL-E) and do captioning all in a zero-shot setting with a single model."
                },
                {
                    "title": "Scaling Language Models: Methods, Analysis & Insights from Training Gopher",
                    "abstract": "Language modelling provides a step towards intelligent communication systems by harnessing large repositories of written human knowledge to better predict and understand the world. In this paper, we present an analysis of Transformer-based language model performance across a wide range of model scales -- from models with tens of millions of parameters up to a 280 billion parameter model called Gopher. These models are evaluated on 152 diverse tasks, achieving state-of-the-art performance across the majority. Gains from scale are largest in areas such as reading comprehension, fact-checking, and the identification of toxic language, but logical and mathematical reasoning see less benefit. We provide a holistic analysis of the training dataset and model's behaviour, covering the intersection of model scale with bias and toxicity. Finally we discuss the application of language models to AI safety and the mitigation of downstream harms."
                },
                {
                    "title": "Efficient Estimate of Low-Frequency Words\u2019 Embeddings Based on the Dictionary: A Case Study on Chinese",
                    "abstract": "Obtaining high-quality embeddings of out-of-vocabularies (OOVs) and low-frequency words is a challenge in natural language processing (NLP). To efficiently estimate the embeddings of OOVs and low-frequency words, we propose a new method that uses the dictionary to estimate the embeddings of OOVs and low-frequency words. More specifically, the explanatory note of an entry in dictionaries accurately describes the semantics of the corresponding word. Naturally, we adopt the sentence representation model to extract the semantics of the explanatory note and regard the semantics as the embedding of the corresponding word. We design a new sentence representation model to encode sentences to extract the semantics from the explanatory notes of entries more efficiently. Based on the assumption that the higher quality of word embeddings will lead to better performance, we design an extrinsic experiment to evaluate the quality of low-frequency words\u2019 embeddings. The experimental results show that the embeddings of low-frequency words estimated by our proposed method have higher quality. In addition, both intrinsic and extrinsic experiments show that our proposed sentence representation model can represent the semantics of sentences well."
                },
                {
                    "title": "Allocating Large Vocabulary Capacity for Cross-Lingual Language Model Pre-Training",
                    "abstract": "Compared to monolingual models, cross-lingual models usually require a more expressive vocabulary to represent all languages adequately. We find that many languages are under-represented in recent cross-lingual language models due to the limited vocabulary capacity. To this end, we propose an algorithm VoCap to determine the desired vocabulary capacity of each language. However, increasing the vocabulary size significantly slows down the pre-training speed. In order to address the issues, we propose k-NN-based target sampling to accelerate the expensive softmax. Our experiments show that the multilingual vocabulary learned with VoCap benefits cross-lingual language model pre-training. Moreover, k-NN-based target sampling mitigates the side-effects of increasing the vocabulary size while achieving comparable performance and faster pre-training speed. The code and the pretrained multilingual vocabularies are available at https://github.com/bozheng-hit/VoCapXLM."
                },
                {
                    "title": "GLM: General Language Model Pretraining with Autoregressive Blank Infilling",
                    "abstract": "There have been various types of pretraining architectures including autoencoding models (e.g., BERT), autoregressive models (e.g., GPT), and encoder-decoder models (e.g., T5). However, none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding (NLU), unconditional generation, and conditional generation. We propose a General Language Model (GLM) based on autoregressive blank infilling to address this challenge. GLM improves blank filling pretraining by adding 2D positional encodings and allowing an arbitrary order to predict spans, which results in performance gains over BERT and T5 on NLU tasks. Meanwhile, GLM can be pretrained for different types of tasks by varying the number and lengths of blanks. On a wide range of tasks across NLU, conditional and unconditional generation, GLM outperforms BERT, T5, and GPT given the same model sizes and data, and achieves the best performance from a single pretrained model with 1.25\u00d7 parameters of BERT Large , demonstrating its generalizability to different downstream tasks."
                },
                {
                    "title": "Explaining neural scaling laws",
                    "abstract": "Significance The population loss of trained deep neural networks has been empirically observed to improve as a power law in a variety of large models and datasets. We investigate the origins behind such \u201cscaling laws\u201d and provide a taxonomy for different scaling regimes. Our findings are based on derivations in linear random feature models\u2014which, in addition to being a simple fruitful model, also describe the wide network limit of deep neural networks. We further formulate and verify aspects of scaling based on smoothness in interpolating a data manifold. We support our theory with empirical results in realistic settings. Our work provides insights into scaling laws and bridges the large gap between theory and experiment in modern deep learning."
                },
                {
                    "title": "Vocabulary Learning via Optimal Transport for Neural Machine Translation",
                    "abstract": "The choice of token vocabulary affects the performance of machine translation. This paper aims to figure out what is a good vocabulary and whether we can find the optimal vocabulary without trial training. To answer these questions, we first provide an alternative understanding of vocabulary from the perspective of information theory. It motivates us to formulate the quest of vocabularization \u2013 finding the best token dictionary with a proper size \u2013 as an optimal transport (OT) problem. We propose VOLT, a simple and efficient solution without trial training. Empirical results show that VOLT beats widely-used vocabularies in diverse scenarios, including WMT-14 English-German translation, TED bilingual translation, and TED multilingual translation. For example, VOLT achieves 70% vocabulary size reduction and 0.5 BLEU gain on English-German translation. Also, compared to BPE-search, VOLT reduces the search time from 384 GPU hours to 30 GPU hours on English-German translation. Codes are available at https://github.com/Jingjing-NLP/VOLT."
                },
                {
                    "title": "Taming Transformers for High-Resolution Image Synthesis",
                    "abstract": "Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks. In contrast to CNNs, they contain no inductive bias that prioritizes local interactions. This makes them expressive, but also computationally infeasible for long sequences, such as high-resolution images. We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images. We show how to (i) use CNNs to learn a context-rich vocabulary of image constituents, and in turn (ii) utilize transformers to efficiently model their composition within high-resolution images. Our approach is readily applied to conditional synthesis tasks, where both non-spatial information, such as object classes, and spatial information, such as segmentations, can control the generated image. In particular, we present the first results on semantically-guided synthesis of megapixel images with transformers. Project page at https://git.io/JLlvY."
                },
                {
                    "title": "Unigram-Normalized Perplexity as a Language Model Performance Measure with Different Vocabulary Sizes",
                    "abstract": "Although Perplexity is a widely used performance metric for language models, the values are highly dependent upon the number of words in the corpus and is useful to compare performance of the same corpus only. In this paper, we propose a new metric that can be used to evaluate language model performance with different vocabulary sizes. The proposed unigram-normalized Perplexity actually presents the performance improvement of the language models from that of simple unigram model, and is robust on the vocabulary size. Both theoretical analysis and computational experiments are reported."
                },
                {
                    "title": "Improving Multilingual Models with Language-Clustered Vocabularies",
                    "abstract": "State-of-the-art multilingual models depend on vocabularies that cover all of the languages the model will expect to see at inference time, but the standard methods for generating those vocabularies are not ideal for massively multilingual applications. In this work, we introduce a novel procedure for multilingual vocabulary generation that combines the separately trained vocabularies of several automatically derived language clusters, thus balancing the trade-off between cross-lingual subword sharing and language-specific vocabularies. Our experiments show improvements across languages on key multilingual benchmark tasks TyDi QA (+2.9 F1), XNLI (+2.1\\%), and WikiAnn NER (+2.8 F1) and factor of 8 reduction in out-of-vocabulary rate, all without increasing the size of the model or data."
                },
                {
                    "title": "Language Models are Few-Shot Learners",
                    "abstract": "Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general."
                },
                {
                    "title": "Scaling Laws for Neural Language Models",
                    "abstract": "We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence."
                },
                {
                    "title": "PIQA: Reasoning about Physical Commonsense in Natural Language",
                    "abstract": "To apply eyeshadow without a brush, should I use a cotton swab or a toothpick? Questions requiring this kind of physical commonsense pose a challenge to today's natural language understanding systems. While recent pretrained models (such as BERT) have made progress on question answering over more abstract domains \u2013 such as news articles and encyclopedia entries, where text is plentiful \u2013 in more physical domains, text is inherently limited due to reporting bias. Can AI systems learn to reliably answer physical commonsense questions without experiencing the physical world?In this paper, we introduce the task of physical commonsense reasoning and a corresponding benchmark dataset Physical Interaction: Question Answering or PIQA. Though humans find the dataset easy (95% accuracy), large pretrained models struggle (\u223c75%). We provide analysis about the dimensions of knowledge that existing models lack, which offers significant opportunities for future research."
                },
                {
                    "title": "Improving Pre-Trained Multilingual Model with Vocabulary Expansion",
                    "abstract": "Recently, pre-trained language models have achieved remarkable success in a broad range of natural language processing tasks. However, in multilingual setting, it is extremely resource-consuming to pre-train a deep language model over large-scale corpora for each language. Instead of exhaustively pre-training monolingual language models independently, an alternative solution is to pre-train a powerful multilingual deep language model over large-scale corpora in hundreds of languages. However, the vocabulary size for each language in such a model is relatively small, especially for low-resource languages. This limitation inevitably hinders the performance of these multilingual models on tasks such as sequence labeling, wherein in-depth token-level or sentence-level understanding is essential. In this paper, inspired by previous methods designed for monolingual settings, we investigate two approaches (i.e., joint mapping and mixture mapping) based on a pre-trained multilingual model BERT for addressing the out-of-vocabulary (OOV) problem on a variety of tasks, including part-of-speech tagging, named entity recognition, machine translation quality estimation, and machine reading comprehension. Experimental results show that using mixture mapping is more promising. To the best of our knowledge, this is the first work that attempts to address and discuss the OOV issue in multilingual settings."
                },
                {
                    "title": "Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism",
                    "abstract": "Recent work in language modeling demonstrates that training large transformer models advances the state of the art in Natural Language Processing applications. However, very large models can be quite difficult to train due to memory constraints. In this work, we present our techniques for training very large transformer models and implement a simple, efficient intra-layer model parallel approach that enables training transformer models with billions of parameters. Our approach does not require a new compiler or library changes, is orthogonal and complimentary to pipeline model parallelism, and can be fully implemented with the insertion of a few communication operations in native PyTorch. We illustrate this approach by converging transformer based models up to 8.3 billion parameters using 512 GPUs. We sustain 15.1 PetaFLOPs across the entire application with 76% scaling efficiency when compared to a strong single GPU baseline that sustains 39 TeraFLOPs, which is 30% of peak FLOPs. To demonstrate that large language models can further advance the state of the art (SOTA), we train an 8.3 billion parameter transformer language model similar to GPT-2 and a 3.9 billion parameter model similar to BERT. We show that careful attention to the placement of layer normalization in BERT-like models is critical to achieving increased performance as the model size grows. Using the GPT-2 model we achieve SOTA results on the WikiText103 (10.8 compared to SOTA perplexity of 15.8) and LAMBADA (66.5% compared to SOTA accuracy of 63.2%) datasets. Our BERT model achieves SOTA results on the RACE dataset (90.9% compared to SOTA accuracy of 89.4%)."
                },
                {
                    "title": "Representation Degeneration Problem in Training Natural Language Generation Models",
                    "abstract": "We study an interesting problem in training neural network-based models for natural language generation tasks, which we call the \\emph{representation degeneration problem}. We observe that when training a model for natural language generation tasks through likelihood maximization with the weight tying trick, especially with big training datasets, most of the learnt word embeddings tend to degenerate and be distributed into a narrow cone, which largely limits the representation power of word embeddings. We analyze the conditions and causes of this problem and propose a novel regularization method to address it. Experiments on language modeling and machine translation show that our method can largely mitigate the representation degeneration problem and achieve better performance than baseline algorithms."
                },
                {
                    "title": "BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions",
                    "abstract": "In this paper we study yes/no questions that are naturally occurring \u2014 meaning that they are generated in unprompted and unconstrained settings. We build a reading comprehension dataset, BoolQ, of such questions, and show that they are unexpectedly challenging. They often query for complex, non-factoid information, and require difficult entailment-like inference to solve. We also explore the effectiveness of a range of transfer learning baselines. We find that transferring from entailment data is more effective than transferring from paraphrase or extractive QA data, and that it, surprisingly, continues to be very beneficial even when starting from massive pre-trained language models such as BERT. Our best method trains BERT on MultiNLI and then re-trains it on our train set. It achieves 80.4% accuracy compared to 90% accuracy of human annotators (and 62% majority-baseline), leaving a significant gap for future work."
                },
                {
                    "title": "HellaSwag: Can a Machine Really Finish Your Sentence?",
                    "abstract": "Recent work by Zellers et al. (2018) introduced a new task of commonsense natural language inference: given an event description such as \u201cA woman sits at a piano,\u201d a machine must select the most likely followup: \u201cShe sets her fingers on the keys.\u201d With the introduction of BERT, near human-level performance was reached. Does this mean that machines can perform human level commonsense inference? In this paper, we show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans (>95% accuracy), state-of-the-art models struggle (<48%). We achieve this via Adversarial Filtering (AF), a data collection paradigm wherein a series of discriminators iteratively select an adversarial set of machine-generated wrong answers. AF proves to be surprisingly robust. The key insight is to scale up the length and complexity of the dataset examples towards a critical \u2018Goldilocks\u2019 zone wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models. Our construction of HellaSwag, and its resulting difficulty, sheds light on the inner workings of deep pretrained models. More broadly, it suggests a new path forward for NLP research, in which benchmarks co-evolve with the evolving state-of-the-art in an adversarial way, so as to present ever-harder challenges."
                },
                {
                    "title": "Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering",
                    "abstract": "We present a new kind of question answering dataset, OpenBookQA, modeled after open book exams for assessing human understanding of a subject. The open book that comes with our questions is a set of 1326 elementary level science facts. Roughly 6000 questions probe an understanding of these facts and their application to novel situations. This requires combining an open book fact (e.g., metals conduct electricity) with broad common knowledge (e.g., a suit of armor is made of metal) obtained from other sources. While existing QA datasets over documents or knowledge bases, being generally self-contained, focus on linguistic understanding, OpenBookQA probes a deeper understanding of both the topic\u2014in the context of common knowledge\u2014and the language it is expressed in. Human performance on OpenBookQA is close to 92%, but many state-of-the-art pre-trained QA methods perform surprisingly poorly, worse than several simple neural baselines we develop. Our oracle experiments designed to circumvent the knowledge retrieval bottleneck demonstrate the value of both the open book and additional facts. We leave it as a challenge to solve the retrieval problem in this multi-hop setting and to close the large gap to human performance."
                },
                {
                    "title": "Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge",
                    "abstract": "We present a new question set, text corpus, and baselines assembled to encourage AI research in advanced question answering. Together, these constitute the AI2 Reasoning Challenge (ARC), which requires far more powerful knowledge and reasoning than previous challenges such as SQuAD or SNLI. The ARC question set is partitioned into a Challenge Set and an Easy Set, where the Challenge Set contains only questions answered incorrectly by both a retrieval-based algorithm and a word co-occurence algorithm. The dataset contains only natural, grade-school science questions (authored for human tests), and is the largest public-domain set of this kind (7,787 questions). We test several baselines on the Challenge Set, including leading neural models from the SQuAD and SNLI tasks, and find that none are able to significantly outperform a random baseline, reflecting the difficult nature of this task. We are also releasing the ARC Corpus, a corpus of 14M science sentences relevant to the task, and implementations of the three neural baseline models tested. Can your model perform better? We pose ARC as a challenge to the community."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Do neural nets learn statistical laws behind natural language?",
                    "abstract": "The performance of deep learning in natural language processing has been spectacular, but the reasons for this success remain unclear because of the inherent complexity of deep learning. This paper provides empirical evidence of its effectiveness and of a limitation of neural networks for language engineering. Precisely, we demonstrate that a neural language model based on long short-term memory (LSTM) effectively reproduces Zipf\u2019s law and Heaps\u2019 law, two representative statistical properties underlying natural language. We discuss the quality of reproducibility and the emergence of Zipf\u2019s law and Heaps\u2019 law as training progresses. We also point out that the neural language model has a limitation in reproducing long-range correlation, another statistical property of natural language. This understanding could provide a direction for improving the architectures of neural networks."
                },
                {
                    "title": "Automatic Speech Recognition With Very Large Conversational Finnish and Estonian Vocabularies",
                    "abstract": "Today, the vocabulary size for language models in large vocabulary speech recognition is typically several hundreds of thousands of words. While this is already sufficient in some applications, the out-of-vocabulary words are still limiting the usability in others. In agglutinative languages the vocabulary for conversational speech should include millions of word forms to cover the spelling variations due to colloquial pronunciations, in addition to the word compounding and inflections. Very large vocabularies are also needed, for example, when the recognition of rare proper names is important. Previously, very large vocabularies have been efficiently modeled in conventional n-gram language models either by splitting words into subword units or by clustering words into classes. While vocabulary size is not as critical anymore in modern speech recognition systems, training time and memory consumption become an issue when state-of-the-art neural network language models are used. In this paper, we investigate techniques that address the vocabulary size issue by reducing the effective vocabulary size and by processing large vocabularies more efficiently. The experimental results in conversational Finnish and Estonian speech recognition indicate that properly defined word classes improve recognition accuracy. Subword n-gram models are not better on evaluation data than word n-gram models constructed from a vocabulary that includes all the words in the training corpus. However, when recurrent neural network (RNN) language models are used, their ability to utilize long contexts gives a larger gain to subword-based modeling. Our best results are from RNN language models that are based on statistical morphs. We show that the suitable size for a subword vocabulary depends on the language. Using time delay neural network acoustic models, we were able to achieve new state of the art in Finnish and Estonian conversational speech recognition, 27.1% word error rate in the Finnish task and 21.9% in the Estonian task."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "Neural Machine Translation of Rare Words with Subword Units",
                    "abstract": "Neural machine translation (NMT) models typically operate with a fixed vocabulary, but translation is an open-vocabulary problem. Previous work addresses the translation of out-of-vocabulary words by backing off to a dictionary. In this paper, we introduce a simpler and more effective approach, making the NMT model capable of open-vocabulary translation by encoding rare and unknown words as sequences of subword units. This is based on the intuition that various word classes are translatable via smaller units than words, for instance names (via character copying or transliteration), compounds (via compositional translation), and cognates and loanwords (via phonological and morphological transformations). We discuss the suitability of different word segmentation techniques, including simple character n-gram models and a segmentation based on the byte pair encoding compression algorithm, and empirically show that subword models improve over a back-off dictionary baseline for the WMT 15 translation tasks English-German and English-Russian by 1.1 and 1.3 BLEU, respectively."
                },
                {
                    "title": "DT-Solver: Automated Theorem Proving with Dynamic-Tree Sampling Guided by Proof-level Value Function",
                    "abstract": "Recent advances in neural theorem-proving resort to large language models and tree searches. When proving a theorem, a language model advises single-step actions based on the current proving state and the tree search finds a sequence of correct steps using actions given by the language model. However, prior works often conduct constant computation efforts for each proving state while ignoring that the hard states often need more exploration than easy states. Moreover, they evaluate and guide the proof search solely depending on the current proof state instead of considering the whole proof trajectory as human reasoning does. Here, to accommodate general theorems, we propose a novel Dynamic-Tree Driven Theorem Solver (DT-Solver) by guiding the search procedure with state confidence and proof-level values. Specifically, DT-Solver introduces a dynamic-tree Monte-Carlo search algorithm, which dynamically allocates computing budgets for different state confidences, guided by a new proof-level value function to discover proof states that require substantial exploration.Experiments on two popular theorem-proving datasets, PISA and Mathlib, show significant performance gains by our DT-Solver over the state-of-the-art approaches, with a 6.65% improvement on average in terms of success rate. And especially under low computing resource settings (11.03% improvement on average)."
                },
                {
                    "title": "An Adversarial Winograd Schema Challenge at Scale",
                    "abstract": "The Winograd Schema Challenge (WSC), proposed by Levesque et al. (2011) as an alternative to the Turing Test, was originally designed as a pronoun resolution problem that cannot be solved based on statistical patterns in large text corpora. However, recent studies suggest that current WSC datasets, even when composed carefully by experts, are still prone to such biases that statistical methods can exploit. We introduce WINOGRANDE, a new collection of WSC problems that are adversarially constructed to be robust against spurious statistical biases. While the original WSC dataset provided only 273 instances, WINOGRANDE includes 43,985 instances, half of which are determined as adversarial. Key to our approach is a novel adversarial filtering algorithm AFLITE for systematic bias reduction, combined with a careful crowdsourcing design. Despite the significant increase in training data, the performance of existing stateof-the-art methods remains modest (61.6%) and contrasts with high human performance (90.8%) for the binary questions. In addition, WINOGRANDE allows us to use transfer learning for achieving new state-of-the-art results on the original WSC and related datasets. Finally, we discuss how biases lead to overestimating the true capabilities of machine commonsense."
                },
                {
                    "title": "Improving Language Understanding by Generative Pre-Training",
                    "abstract": "Natural language understanding comprises a wide range of diverse tasks such as textual entailment, question answering, semantic similarity assessment, and document classi\ufb01cation. Although large unlabeled text corpora are abundant, labeled data for learning these speci\ufb01c tasks is scarce, making it challenging for discriminatively trained models to perform adequately. We demonstrate that large gains on these tasks can be realized by generative pre-training of a language model on a diverse corpus of unlabeled text, followed by discriminative \ufb01ne-tuning on each speci\ufb01c task. In contrast to previous approaches, we make use of task-aware input transformations during \ufb01ne-tuning to achieve effective transfer while requiring minimal changes to the model architecture. We demonstrate the effectiveness of our approach on a wide range of benchmarks for natural language understanding. Our general task-agnostic model outperforms discriminatively trained models that use architectures speci\ufb01cally crafted for each task, signi\ufb01cantly improving upon the state of the art in 9 out of the 12 tasks studied. For instance, we achieve absolute improvements of 8.9% on commonsense reasoning (Stories Cloze Test), 5.7% on question answering (RACE), and 1.5% on textual entailment (MultiNLI)."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nWhat is the compute-optimal vocabulary size for large language models (LLMs)?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving the problem of determining the compute-optimal vocabulary size for LLMs is crucial for advancing the field of natural language processing (NLP). By optimizing vocabulary size, researchers can enhance model performance, improve tokenization efficiency, and better capture the nuances of language. This research could lead to more efficient LLMs that require less computational power while maintaining or improving their effectiveness, thereby influencing future research directions in model design and training strategies. Additionally, practical applications could include more effective language models for various tasks, such as translation, summarization, and conversational agents, ultimately benefiting industries reliant on NLP technologies.\n\n**[Question 3] - Why is it hard?**  \nDetermining the optimal vocabulary size is challenging due to the complex interplay between vocabulary size, model parameters, and training data. Naive approaches may fail because they often overlook the impact of vocabulary size on model performance and scaling laws. The technical obstacles include the need for a robust framework to normalize loss across models with varying vocabulary sizes and the difficulty in accurately modeling the relationship between vocabulary size and performance metrics. Theoretical challenges arise from the need to derive power laws that account for both vocabulary and non-vocabulary parameters, which have not been adequately addressed in prior research.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on scaling laws that do not consider vocabulary size, leading to a gap in understanding its impact on model performance. Existing solutions have been limited by a lack of comprehensive frameworks that integrate vocabulary size into performance predictions. Barriers include the complexity of modeling the relationship between vocabulary size and performance, as well as the absence of empirical studies that systematically explore this relationship. Our approach differs by introducing a normalized loss formulation and multiple methodologies to estimate the optimal vocabulary size, thereby addressing the limitations of prior work.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology includes three key approaches:  \n1. **Estimating power laws via IsoFLOPs**: We will pre-train models with non-vocabulary parameters ranging from 33M to 1.13B, ensuring that groups of models share the same FLOPs while varying vocabulary configurations. This will allow us to fit power laws relating FLOPs to both non-vocabulary and vocabulary parameters.\n2"
            }
        },
        "author_data": {
            "8d518966-bb0d-4a62-a307-c4eff16fadc9": {
                "pk": "8d518966-bb0d-4a62-a307-c4eff16fadc9",
                "name": "Chaofan Tao",
                "collaborators": [
                    "Ngai Wong",
                    "Rui Lin",
                    "Chaofan Chen",
                    "Alina Jade Barnett",
                    "Cynthia Rudin",
                    "Binxiao Huang",
                    "Ping Luo",
                    "Lixin Duan",
                    "Fides Regina Schwartz",
                    "Yinhao Ren"
                ],
                "domain": [
                    "Adversarial Learning",
                    "Model Compression",
                    "Interpretability",
                    "Domain Adaptation"
                ],
                "publications": [
                    {
                        "title": "ODG-Q: Robust Quantization via Online Domain Generalization",
                        "abstract": "Quantizing neural networks to low-bitwidth is important for model deployment on resource-limited edge hardware. Although a quantized network has a smaller model size and memory footprint, it is fragile to adversarial attacks. However, few methods study the robustness and training efficiency of quantized networks. To this end, we propose a new method by recasting robust quantization as an online domain generalization problem, termed ODG-Q, which generates diverse adversarial data at a low cost during training. ODG-Q consistently outperforms existing works against various adversarial attacks. For example, on CIFAR-10 dataset, ODG-Q achieves 49.2% average improvements under five common white-box attacks and 21.7% average improvements under five common black-box attacks, with a training cost similar to that of natural training (viz. without adversaries). To our best knowledge, this work is the first work that trains both quantized and binary neural networks on ImageNet that consistently improve robustness under different attacks. We also provide a theoretical insight of ODG-Q that accounts for the bound of model risk on attacked data."
                    },
                    {
                        "title": "This Looks Like That: Deep Learning for Interpretable Image Recognition",
                        "abstract": "When we are faced with challenging image classification tasks, we often explain our reasoning by dissecting the image, and pointing out prototypical aspects of one class or another. The mounting evidence for each of the classes helps us make our final decision. In this work, we introduce a deep network architecture -- prototypical part network (ProtoPNet), that reasons in a similar way: the network dissects the image by finding prototypical parts, and combines evidence from the prototypes to make a final classification. The model thus reasons in a way that is qualitatively similar to the way ornithologists, physicians, and others would explain to people on how to solve challenging image classification tasks. The network uses only image-level labels for training without any annotations for parts of images. We demonstrate our method on the CUB-200-2011 dataset and the Stanford Cars dataset. Our experiments show that ProtoPNet can achieve comparable accuracy with its analogous non-interpretable counterpart, and when several ProtoPNets are combined into a larger network, it can achieve an accuracy that is on par with some of the best-performing deep models. Moreover, ProtoPNet provides a level of interpretability that is absent in other interpretable deep models."
                    },
                    {
                        "title": "MiniMax Entropy Network: Learning Category-Invariant Features for Domain Adaptation",
                        "abstract": "How to effectively learn from unlabeled data from the target domain is crucial for domain adaptation, as it helps reduce the large performance gap due to domain shift or distribution change. In this paper, we propose an easy-to-implement method dubbed MiniMax Entropy Networks (MMEN) based on adversarial learning. Unlike most existing approaches which employ a generator to deal with domain difference, MMEN focuses on learning the categorical information from unlabeled target samples with the help of labeled source samples. Specifically, we set an unfair multi-class classifier named categorical discriminator, which classifies source samples accurately but be confused about the categories of target samples. The generator learns a common subspace that aligns the unlabeled samples based on the target pseudo-labels. For MMEN, we also provide theoretical explanations to show that the learning of feature alignment reduces domain mismatch at the category level. Experimental results on various benchmark datasets demonstrate the effectiveness of our method over existing state-of-the-art baselines."
                    },
                    {
                        "title": "Frequency Regularization for Improving Adversarial Robustness",
                        "abstract": "Deep neural networks are incredibly vulnerable to crafted, human-imperceptible adversarial perturbations. Although adversarial training (AT) has proven to be an effective defense approach, we find that the AT-trained models heavily rely on the input low-frequency content for judgment, accounting for the low standard accuracy. To close the large gap between the standard and robust accuracies during AT, we investigate the frequency difference between clean and adversarial inputs, and propose a frequency regularization (FR) to align the output difference in the spectral domain. Besides, we find Stochastic Weight Averaging (SWA), by smoothing the kernels over epochs, further improves the robustness. Among various defense schemes, our method achieves the strongest robustness against attacks by PGD-20, C\\&W and Autoattack, on a WideResNet trained on CIFAR-10 without any extra data."
                    },
                    {
                        "title": "Dynamic and Static Context-aware LSTM for Multi-agent Motion Prediction",
                        "abstract": "Multi-agent motion prediction is challenging because it aims to foresee the future trajectories of multiple agents (\\textit{e.g.} pedestrians) simultaneously in a complicated scene. Existing work addressed this challenge by either learning social spatial interactions represented by the positions of a group of pedestrians, while ignoring their temporal coherence (\\textit{i.e.} dependencies between different long trajectories), or by understanding the complicated scene layout (\\textit{e.g.} scene segmentation) to ensure safe navigation. However, unlike previous work that isolated the spatial interaction, temporal coherence, and scene layout, this paper designs a new mechanism, \\textit{i.e.}, Dynamic and Static Context-aware Motion Predictor (DSCMP), to integrates these rich information into the long-short-term-memory (LSTM). It has three appealing benefits. (1) DSCMP models the dynamic interactions between agents by learning both their spatial positions and temporal coherence, as well as understanding the contextual scene layout.(2) Different from previous LSTM models that predict motions by propagating hidden features frame by frame, limiting the capacity to learn correlations between long trajectories, we carefully design a differentiable queue mechanism in DSCMP, which is able to explicitly memorize and learn the correlations between long trajectories. (3) DSCMP captures the context of scene by inferring latent variable, which enables multimodal predictions with meaningful semantic scene layout. Extensive experiments show that DSCMP outperforms state-of-the-art methods by large margins, such as 9.05\\% and 7.62\\% relative improvements on the ETH-UCY and SDD datasets respectively."
                    },
                    {
                        "title": "What Do Adversarially trained Neural Networks Focus: A Fourier Domain-based Study",
                        "abstract": "Although many fields have witnessed the superior performance brought about by deep learning, the robustness of neural networks remains an open issue. Specifically, a small adversarial perturbation on the input may cause the model to produce a completely different output. Such poor robustness implies many potential hazards, especially in security-critical applications, e.g., autonomous driving and mobile robotics. This work studies what information the adversarially trained model focuses on. Empirically, we notice that the differences between the clean and adversarial data are mainly distributed in the low-frequency region. We then find that an adversarially-trained model is more robust than its naturally-trained counterpart due to the reason that the former pays more attention to learning the dominant information in low-frequency components. In addition, we consider two common ways to improve model robustness, namely, by data augmentation and by using stronger network architectures, and understand these techniques from a frequency-domain perspective. We are hopeful this work can shed light on the design of more robust neural networks."
                    },
                    {
                        "title": "A Spectral Perspective towards Understanding and Improving Adversarial Robustness",
                        "abstract": "Deep neural networks (DNNs) are incredibly vulnerable to crafted, imperceptible adversarial perturbations. While adversarial training (AT) has proven to be an effective defense approach, the AT mechanism for robustness improvement is not fully understood. This work investigates AT from a spectral perspective, adding new insights to the design of effective defenses. In particular, we show that AT induces the deep model to focus more on the low-frequency region, which retains the shape-biased representations, to gain robustness. Further, we find that the spectrum of a white-box attack is primarily distributed in regions the model focuses on, and the perturbation attacks the spectral bands where the model is vulnerable. Based on this observation, to train a model tolerant to frequency-varying perturbation, we propose a spectral alignment regularization (SAR) such that the spectral output inferred by an attacked adversarial input stays as close as possible to its natural input counterpart. Experiments demonstrate that SAR and its weight averaging (WA) extension could significantly improve the robust accuracy by 1.14% ~ 3.87% relative to the standard AT, across multiple datasets (CIFAR-10, CIFAR-100 and Tiny ImageNet), and various attacks (PGD, C&W and Autoattack), without any extra data."
                    },
                    {
                        "title": "Interpretable Mammographic Image Classification using Case-Based Reasoning and Deep Learning",
                        "abstract": "When we deploy machine learning models in high-stakes medical settings, we must ensure these models make accurate predictions that are consistent with known medical science. Inherently interpretable networks address this need by explaining the rationale behind each decision while maintaining equal or higher accuracy compared to black-box models. In this work, we present a novel interpretable neural network algorithm that uses case-based reasoning for mammography. Designed to aid a radiologist in their decisions, our network presents both a prediction of malignancy and an explanation of that prediction using known medical features. In order to yield helpful explanations, the network is designed to mimic the reasoning processes of a radiologist: our network first detects the clinically relevant semantic features of each image by comparing each new image with a learned set of prototypical image parts from the training images, then uses those clinical features to predict malignancy. Compared to other methods, our model detects clinical features (mass margins) with equal or higher accuracy, provides a more detailed explanation of its prediction, and is better able to differentiate the classification-relevant parts of the image."
                    },
                    {
                        "title": "IAIA-BL: A Case-based Interpretable Deep Learning Model for Classification of Mass Lesions in Digital Mammography",
                        "abstract": "Interpretability in machine learning models is important in high-stakes decisions, such as whether to order a biopsy based on a mammographic exam. Mammography poses important challenges that are not present in other computer vision tasks: datasets are small, confounding information is present, and it can be difficult even for a radiologist to decide between watchful waiting and biopsy based on a mammogram alone. In this work, we present a framework for interpretable machine learning-based mammography. In addition to predicting whether a lesion is malignant or benign, our work aims to follow the reasoning processes of radiologists in detecting clinically relevant semantic features of each image, such as the characteristics of the mass margins. The framework includes a novel interpretable neural network algorithm that uses case-based reasoning for mammography. Our algorithm can incorporate a combination of data with whole image labelling and data with pixel-wise annotations, leading to better accuracy and interpretability even with a small number of images. Our interpretable models are able to highlight the classification-relevant parts of the image, whereas other methods highlight healthy tissue and confounding information. Our models are decision aids, rather than decision makers, aimed at better overall human-machine collaboration. We do not observe a loss in mass margin classification accuracy over a black box neural network trained on the same data."
                    },
                    {
                        "title": "FAT: Learning Low-Bitwidth Parametric Representation via Frequency-Aware Transformation",
                        "abstract": "Learning convolutional neural networks (CNNs) with low bitwidth is challenging because performance may drop significantly after quantization. Prior arts often discretize the network weights by carefully tuning hyper-parameters of quantization (e.g. non-uniform stepsize and layer-wise bitwidths), which are complicated and sub-optimal because the full-precision and low-precision models have a large discrepancy. This work presents a novel quantization pipeline, Frequency-Aware Transformation (FAT), which has several appealing benefits. (1) Rather than designing complicated quantizers like existing works, FAT learns to transform network weights in the frequency domain before quantization, making them more amenable to training in low bitwidth. (2) With FAT, CNNs can be easily trained in low precision using simple standard quantizers without tedious hyper-parameter tuning. Theoretical analysis shows that FAT improves both uniform and non-uniform quantizers. (3) FAT can be easily plugged into many CNN architectures. When training ResNet-18 and MobileNet-V2 in 4 bits, FAT plus a simple rounding operation already achieves 70.5% and 69.2% top-1 accuracy on ImageNet without bells and whistles, outperforming recent state-of-the-art by reducing 54.9X and 45.7X computations against full-precision models. We hope FAT provides a novel perspective for model quantization. Code is available at \\url{https://github.com/ChaofanTao/FAT_Quantization}."
                    },
                    {
                        "title": "LiteVL: Efficient Video-Language Learning with Enhanced Spatial-Temporal Modeling",
                        "abstract": "Recent large-scale video-language pre-trained models have shown appealing performance on various downstream tasks. However, the pre-training process is computationally expensive due to the requirement of millions of video-text pairs and the redundant data structure of each video. To mitigate these problems, we propose LiteVL, which adapts a pre-trained image-language model BLIP into a video-text model directly on downstream tasks, without heavy pre-training. To enhance the temporal modeling lacking in the image-language model, we propose to add temporal attention modules in the image encoder of BLIP with dynamic temporal scaling. Besides the model-wise adaptation, we also propose a non-parametric pooling mechanism to adaptively reweight the fine-grained video embedding conditioned on the text. Experimental results on text-video retrieval and video question answering show that the proposed LiteVL even outperforms previous video-language pre-trained models by a clear margin, though without any video-language pre-training."
                    },
                    {
                        "title": "UPop: Unified and Progressive Pruning for Compressing Vision-Language Transformers",
                        "abstract": "Real-world data contains a vast amount of multimodal information, among which vision and language are the two most representative modalities. Moreover, increasingly heavier models, \\textit{e}.\\textit{g}., Transformers, have attracted the attention of researchers to model compression. However, how to compress multimodal models, especially vison-language Transformers, is still under-explored. This paper proposes the \\textbf{U}nified and \\textbf{P}r\\textbf{o}gressive \\textbf{P}runing (\\textbf{\\emph{UPop}}) as a universal vison-language Transformer compression framework, which incorporates 1) unifiedly searching multimodal subnets in a continuous optimization space from the original model, which enables automatic assignment of pruning ratios among compressible modalities and structures; 2) progressively searching and retraining the subnet, which maintains convergence between the search and retrain to attain higher compression ratios. Experiments on various tasks, datasets, and model architectures demonstrate the effectiveness and versatility of the proposed UPop framework. The code is available at https://github.com/sdc17/UPop."
                    },
                    {
                        "title": "Rethinking Kullback-Leibler Divergence in Knowledge Distillation for Large Language Models",
                        "abstract": "Kullback-Leiber divergence has been widely used in Knowledge Distillation (KD) to compress Large Language Models (LLMs). Contrary to prior assertions that reverse Kullback-Leibler (RKL) divergence is mode-seeking and thus preferable over the mean-seeking forward Kullback-Leibler (FKL) divergence, this study empirically and theoretically demonstrates that neither mode-seeking nor mean-seeking properties manifest in KD for LLMs. Instead, RKL and FKL are found to share the same optimization objective and both converge after a sufficient number of epochs. However, due to practical constraints, LLMs are seldom trained for such an extensive number of epochs. Meanwhile, we further find that RKL focuses on the tail part of the distributions, while FKL focuses on the head part at the beginning epochs. Consequently, we propose a simple yet effective Adaptive Kullback-Leiber (AKL) divergence method, which adaptively allocates weights to combine FKL and RKL. Metric-based and GPT-4-based evaluations demonstrate that the proposed AKL outperforms the baselines across various tasks and improves the diversity and quality of generated responses."
                    },
                    {
                        "title": "CrossGET: Cross-Guided Ensemble of Tokens for Accelerating Vision-Language Transformers",
                        "abstract": "Recent vision-language models have achieved tremendous advances. However, their computational costs are also escalating dramatically, making model acceleration exceedingly critical. To pursue more efficient vision-language Transformers, this paper introduces Cross-Guided Ensemble of Tokens (CrossGET), a general acceleration framework for vision-language Transformers. This framework adaptively combines tokens in real-time during inference, significantly reducing computational costs while maintaining high performance. CrossGET features two primary innovations: 1) Cross-Guided Matching and Ensemble. CrossGET leverages cross-modal guided token matching and ensemble to effectively utilize cross-modal information, achieving wider applicability across both modality-independent models, e.g., CLIP, and modality-dependent ones, e.g., BLIP2. 2) Complete-Graph Soft Matching. CrossGET introduces an algorithm for the token-matching mechanism, ensuring reliable matching results while facilitating parallelizability and high efficiency. Extensive experiments have been conducted on various vision-language tasks, such as image-text retrieval, visual reasoning, image captioning, and visual question answering. The performance on both classic multimodal architectures and emerging multimodal LLMs demonstrates the framework's effectiveness and versatility. The code is available at https://github.com/sdc17/CrossGET."
                    },
                    {
                        "title": "Compression of Generative Pre-trained Language Models via Quantization",
                        "abstract": "The increasing size of generative Pre-trained Language Models (PLMs) has greatly increased the demand for model compression. Despite various methods to compress BERT or its variants, there are few attempts to compress generative PLMs, and the underlying difficulty remains unclear. In this paper, we compress generative PLMs by quantization. We find that previous quantization methods fail on generative tasks due to the \\textit{homogeneous word embeddings} caused by reduced capacity, and \\textit{varied distribution of weights}. Correspondingly, we propose a token-level contrastive distillation to learn distinguishable word embeddings, and a module-wise dynamic scaling to make quantizers adaptive to different modules. Empirical results on various tasks show that our proposed method outperforms the state-of-the-art compression methods on generative PLMs by a clear margin. With comparable performance with the full-precision models, we achieve 14.4x and 13.4x compression rates on GPT-2 and BART, respectively."
                    }
                ]
            },
            "b5f0e417-4a0f-4034-bf93-fd8ad610b2e4": {
                "pk": "b5f0e417-4a0f-4034-bf93-fd8ad610b2e4",
                "name": "Qian Liu",
                "collaborators": [
                    "Chongchao Huang",
                    "Mei Zhang",
                    "Dazhong He",
                    "Bing Wu"
                ],
                "domain": [
                    "Haptic Interaction",
                    "Quantum Communication",
                    "Tactile Perception",
                    "Optimization Models"
                ],
                "publications": [
                    {
                        "title": "The combined equilibrium of business land-use and its congestion pricing principles (working paper)",
                        "abstract": "This working paper is divided into two parts. Firstly, we develop a new combined equilibrium model of business land-use, which puts travelers' traffic equilibrium and business companies' competitive location equilibrium into a unified framework. A variational inequality is presented for the combined equilibrium and the properties of equilibrium solution are investigated. Secondly, the congestion pricing principles associated with the combined equilibrium are studied. From the mathematical point of view, we prove that there exists an optimal road pricing scheme that can minimize the social cost of travelers. This road pricing scheme generalizes the traditional link-based optimal road pricing scheme. Furthermore, when simultaneously imposing charges on travelers and companies is allowed, we prove that there exists an optimal congestion pricing scheme that can derive a combined equilibrium toward an overall system optimum. The economic meaning of every pricing scheme proposed in this paper is discussed in detail. At last, a simple numerical example is used to demonstrate that the optimal congestion pricing scheme may indeed reduce the social cost."
                    },
                    {
                        "title": "Generation and complete nondestructive analysis of hyperentanglement assisted by nitrogen-vacancy centers in resonators",
                        "abstract": "We present two efficient schemes for the deterministic generation and the complete nondestructive analysis of hyperentangled Bell states in both the polarization and spatial-mode degrees of freedom (DOFs) of two-photon systems, assisted by the nitrogen-vacancy (NV) centers in diamonds coupled to microtoroidal resonators as a result of cavity quantum electrodynamics (QED). With the input-output process of photons, two-photon polarization-spatial hyperentangled Bell states can be generated in a deterministic way and their complete nondestructive analysis can be achieved. These schemes can be generalized to generate and analyze hyperentangled Greenberger-Horne-Zeilinger states of multi-photon systems as well. Compared with previous works, these two schemes relax the difficulty of their implementation in experiment as it is not difficult to obtain the $\\pi$ phase shift in single-sided NV-cavity systems. Moreover, our schemes do not require that the transmission for the uncoupled cavity is balanceable with the reflectance for the coupled cavity. Our calculations show that these schemes can reach a high fidelity and efficiency with current technology, which may be a benefit to long-distance high-capacity quantum communication with two DOFs of photon systems."
                    },
                    {
                        "title": "Self-supervised Spatio-Temporal Graph Mask-Passing Attention Network for Perceptual Importance Prediction of Multi-point Tactility",
                        "abstract": "While visual and auditory information are prevalent in modern multimedia systems, haptic interaction, e.g., tactile and kinesthetic interaction, provides a unique form of human perception. However, multimedia technology for contact interaction is less mature than non-contact multimedia technologies and requires further development. Specialized haptic media technologies, requiring low latency and bitrates, are essential to enable haptic interaction, necessitating haptic information compression. Existing vibrotactile signal compression methods, based on the perceptual model, do not consider the characteristics of fused tactile perception at multiple spatially distributed interaction points. In fact, differences in tactile perceptual importance are not limited to conventional frequency and time domains, but also encompass differences in the spatial locations on the skin unique to tactile perception. For the most frequently used tactile information, vibrotactile texture perception, we have developed a model to predict its perceptual importance at multiple points, based on self-supervised learning and Spatio-Temporal Graph Neural Network. Current experimental results indicate that this model can effectively predict the perceptual importance of various points in multi-point tactile perception scenarios."
                    },
                    {
                        "title": "TactileAR: Active Tactile Pattern Reconstruction",
                        "abstract": "High-resolution (HR) contact surface information is essential for robotic grasping and precise manipulation tasks. However, it remains a challenge for current taxel-based sensors to obtain HR tactile information. In this paper, we focus on utilizing low-resolution (LR) tactile sensors to reconstruct the localized, dense, and HR representation of contact surfaces. In particular, we build a Gaussian triaxial tactile sensor degradation model and propose a tactile pattern reconstruction framework based on the Kalman filter. This framework enables the reconstruction of 2-D HR contact surface shapes using collected LR tactile sequences. In addition, we present an active exploration strategy to enhance the reconstruction efficiency. We evaluate the proposed method in real-world scenarios with comparison to existing prior-information-based approaches. Experimental results confirm the efficiency of the proposed approach and demonstrate satisfactory reconstructions of complex contact surface shapes. Code: https://github.com/wmtlab/tactileAR"
                    }
                ]
            },
            "85f7fb25-0b41-426d-939d-80c47a890a81": {
                "pk": "85f7fb25-0b41-426d-939d-80c47a890a81",
                "name": "Longxu Dou",
                "collaborators": [
                    "Wanxiang Che",
                    "Dingzirui Wang",
                    "Qingfu Zhu",
                    "Xuanliang Zhang",
                    "Qian Liu",
                    "Min Lin",
                    "Yan Gao",
                    "Mingyang Pan",
                    "Dechen Zhan",
                    "Jian-Guang Lou"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Text-to-SQL",
                    "Question Answering",
                    "Data Augmentation"
                ],
                "publications": [
                    {
                        "title": "A Survey on Table-and-Text HybridQA: Concepts, Methods, Challenges and Future Directions",
                        "abstract": "Table-and-text hybrid question answering (HybridQA) is a widely used and challenging NLP task commonly applied in the financial and scientific domain. The early research focuses on migrating other QA task methods to HybridQA, while with further research, more and more HybridQA-specific methods have been present. With the rapid development of HybridQA, the systematic survey is still under-explored to summarize the main techniques and advance further research. So we present this work to summarize the current HybridQA benchmarks and methods, then analyze the challenges and future directions of this task. The contributions of this paper can be summarized in three folds: (1) first survey, to our best knowledge, including benchmarks, methods and challenges for HybridQA; (2) systematic investigation with the reasonable comparison of the existing systems to articulate their advantages and shortcomings; (3) detailed analysis of challenges in four important dimensions to shed light on future directions."
                    },
                    {
                        "title": "Controllable Data Augmentation for Context-Dependent Text-to-SQL",
                        "abstract": "The limited scale of annotated data constraints existing context-dependent text-to-SQL models because of the complexity of labeling. The data augmentation method is a commonly used method to solve this problem. However, the data generated by current augmentation methods often lack diversity. In this paper, we introduce ConDA, which generates interactive questions and corresponding SQL results. We designed the SQL dialogue state to enhance the data diversity through the state transition. Meanwhile, we also present a filter method to ensure the data quality by a grounding model. Additionally, we utilize a grounding model to identify and filter low-quality questions that mismatch the state information. Experimental results on the SParC and CoSQL datasets show that ConDA boosts the baseline model to achieve an average improvement of $3.3\\%$ on complex questions. Moreover, we analyze the augmented data, which reveals that the data generated by ConDA are of high quality in both SQL template hardness and types, turns, and question consistency."
                    },
                    {
                        "title": "From Zero to Hero: Examining the Power of Symbolic Tasks in Instruction Tuning",
                        "abstract": "Fine-tuning language models on tasks with instructions has demonstrated potential in facilitating zero-shot generalization to unseen tasks. In this paper, we introduce a straightforward yet effective method for enhancing instruction tuning by employing symbolic tasks. Compared to crowdsourced human tasks or model-generated tasks, symbolic tasks present a unique advantage as they can be easily generated in vast quantities, theoretically providing an infinite supply of high-quality training instances. To explore the potential of symbolic tasks, we carry out an extensive case study on the representative symbolic task of SQL execution. Empirical results on various benchmarks validate that the integration of SQL execution leads to significant improvements in zero-shot scenarios, particularly in table reasoning. Notably, our 3B model surpasses both the 175B GPT-3 and ChatGPT in zero-shot table reasoning across four benchmarks. Furthermore, experimental results on BBH (27 tasks) and MMLU (57 tasks) reveal that language models can be enhanced through symbolic tasks without compromising their generality. We hope that our paper serves as a catalyst, inspiring increased efforts to incorporate symbolic tasks in instruction tuning."
                    },
                    {
                        "title": "A Survey of Table Reasoning with Large Language Models",
                        "abstract": "Table reasoning, which aims to generate the corresponding answer to the question following the user requirement according to the provided table, and optionally a text description of the table, effectively improving the efficiency of obtaining information. Recently, using Large Language Models (LLMs) has become the mainstream method for table reasoning, because it not only significantly reduces the annotation cost but also exceeds the performance of previous methods. However, existing research still lacks a summary of LLM-based table reasoning works. Due to the existing lack of research, questions about which techniques can improve table reasoning performance in the era of LLMs, why LLMs excel at table reasoning, and how to enhance table reasoning abilities in the future, remain largely unexplored. This gap significantly limits progress in research. To answer the above questions and advance table reasoning research with LLMs, we present this survey to analyze existing research, inspiring future work. In this paper, we analyze the mainstream techniques used to improve table reasoning performance in the LLM era, and the advantages of LLMs compared to pre-LLMs for solving table reasoning. We provide research directions from both the improvement of existing methods and the expansion of practical applications to inspire future research."
                    },
                    {
                        "title": "Enhancing Numerical Reasoning with the Guidance of Reliable Reasoning Processes",
                        "abstract": "Numerical reasoning is an essential ability for NLP systems to handle numeric information. Recent research indicates that fine-tuning a small-scale model to learn generating reasoning processes alongside answers can significantly enhance performance. However, current methods have the limitation that most methods generate reasoning processes with large language models (LLMs), which are \"unreliable\" since such processes could contain information unrelated to the answer. To address this limitation, we introduce Enhancing NumeriCal reasOning with Reliable procEsses (Encore), which derives the reliable reasoning process by decomposing the answer formula, ensuring which fully supports the answer. Nevertheless, models could lack enough data to learn the reasoning process generation adequately, since our method generates only one single reasoning process for one formula. To overcome this difficulty, we present a series of pre-training tasks to help models learn the reasoning process generation with synthesized data. The experiments show that Encore yields improvement on all five experimental datasets with an average of 1.8%, proving the effectiveness of our method."
                    },
                    {
                        "title": "Improving Demonstration Diversity by Human-Free Fusing for Text-to-SQL",
                        "abstract": "Currently, the in-context learning method based on large language models (LLMs) has become the mainstream of text-to-SQL research. Previous works have discussed how to select demonstrations related to the user question from a human-labeled demonstration pool. However, human labeling suffers from the limitations of insufficient diversity and high labeling overhead. Therefore, in this paper, we discuss how to measure and improve the diversity of the demonstrations for text-to-SQL. We present a metric to measure the diversity of the demonstrations and analyze the insufficient of the existing labeled data by experiments. Based on the above discovery, we propose fusing iteratively for demonstrations (Fused) to build a high-diversity demonstration pool through human-free multiple-iteration synthesis, improving diversity and lowering label cost. Our method achieves an average improvement of 3.2% and 5.0% with and without human labeling on several mainstream datasets, which proves the effectiveness of Fused."
                    },
                    {
                        "title": "MURRE: Multi-Hop Table Retrieval with Removal for Open-Domain Text-to-SQL",
                        "abstract": "The open-domain text-to-SQL task aims to retrieve question-relevant tables from massive databases and generate SQL. However, the performance of current methods is constrained by single-hop retrieval, and existing multi-hop retrieval of open-domain question answering is not directly applicable due to the tendency to retrieve tables similar to the retrieved ones but irrelevant to the question. Since the questions in text-to-SQL usually contain all required information, while previous multi-hop retrieval supplements the questions with retrieved documents. Therefore, we propose the multi-hop table retrieval with removal (MURRE), which removes previously retrieved information from the question to guide the retriever towards unretrieved relevant tables. Our experiments on two open-domain text-to-SQL datasets demonstrate an average improvement of 5.7% over the previous state-of-the-art results."
                    },
                    {
                        "title": "Exploring Equation as a Better Intermediate Meaning Representation for Numerical Reasoning",
                        "abstract": "Numerical reasoning is vital for natural language processing models to understand and process numerical information in real-world scenarios. Most current methods first generate the Intermediate Meaning Representations (IMRs) of questions and then generate answers. Current SOTA methods generate programs as IMRs with large language models (LLMs). Intuitively, equations have fewer restrictions and closer semantics to the question than programs, leading to higher generation accuracy. However, current LLMs generate equations worse than programs, where we assume that the equation data is rare in pre-training data compared to programs. So in this paper, we try to use equations as IMRs to solve the numerical reasoning task by addressing two problems: (1) Theoretically, how to prove that the equation is an IMR with higher generation accuracy than programs; (2) Empirically, how to improve the generation accuracy of equations with LLMs. For the first problem, we propose and prove a proposition to theoretically compare the generation accuracy of different IMRs. For the second problem, we present a method called Boosting Numerical Reason\\textbfing by Decomposing the Generation of Equations (Bridge), which can improve the accuracy of LLMs in generating equations as IMRs by reducing the tendency of generating constant expressions and programs. Our method improves the performance by 2.2%, 0.9%, and 1.7% on GSM8K, SVAMP, and Algebra datasets compared to the previous state-of-the-art methods under the single reasoning path setting. Our codes and prompts are released in https://github.com/zirui-HIT/Bridge_for_Numerical_Reasoning."
                    },
                    {
                        "title": "DAC: Decomposed Automation Correction for Text-to-SQL",
                        "abstract": "Text-to-SQL is an important task that helps people obtain information from databases by automatically generating SQL queries. Considering the brilliant performance, approaches based on Large Language Models (LLMs) become the mainstream for text-to-SQL. Among these approaches, automated correction is an effective approach that further enhances performance by correcting the mistakes in the generated results. The existing correction methods require LLMs to directly correct with generated SQL, while previous research shows that LLMs do not know how to detect mistakes, leading to poor performance. Therefore, in this paper, we propose to employ the decomposed correction to enhance text-to-SQL performance. We first demonstrate that decomposed correction outperforms direct correction since detecting and fixing mistakes with the results of the decomposed sub-tasks is easier than with SQL. Based on this analysis, we introduce Decomposed Automation Correction (DAC), which corrects SQL by decomposing text-to-SQL into entity linking and skeleton parsing. DAC first generates the entity and skeleton corresponding to the question and then compares the differences between the initial SQL and the generated entities and skeleton as feedback for correction. Experimental results show that our method improves performance by $3.7\\%$ on average of Spider, Bird, and KaggleDBQA compared with the baseline method, demonstrating the effectiveness of DAC."
                    },
                    {
                        "title": "Sailor: Open Language Models for South-East Asia",
                        "abstract": "We present Sailor, a family of open language models ranging from 0.5B to 7B parameters, tailored for South-East Asian (SEA) languages. These models are continually pre-trained from Qwen1.5, a great language model for multilingual use cases. From Qwen1.5, Sailor models accept 200B to 400B tokens, primarily covering the languages of English, Chinese, Vietnamese, Thai, Indonesian, Malay, and Lao. The training leverages several techniques, including BPE dropout for improving the model robustness, aggressive data cleaning and deduplication, and small proxy models to optimize data mixture. Experimental results on four typical tasks indicate that Sailor models demonstrate strong performance across different benchmarks, including commonsense reasoning, question answering, reading comprehension and examination. Embracing the open-source spirit, we share our insights through this report to spark a wider interest in developing large language models for multilingual use cases."
                    },
                    {
                        "title": "FLEXTAF: Enhancing Table Reasoning with Flexible Tabular Formats",
                        "abstract": "The table reasoning task aims to answer the question according to the given table. Currently, using Large Language Models (LLMs) is the predominant method for table reasoning. Most existing methods employ a fixed tabular format to represent the table, which could limit the performance. Given that each instance requires different capabilities and models possess varying abilities, we assert that different instances and models suit different tabular formats. We prove the aforementioned claim through quantitative analysis of experimental results, where different instances and models achieve different performances using various tabular formats. Building on this discussion, we propose FLEXTAF-Single and FLEXTAF-Vote to enhance table reasoning performance by employing flexible tabular formats. Specifically, (i) FLEXTAF-Single trains a classifier to predict the most suitable tabular format based on the instance and the LLM. (ii) FLEXTAF-Vote integrates the results across different formats. Our experiments on WikiTableQuestions and TabFact reveal significant improvements, with average gains of 2.3% and 4.8% compared to the best performance achieved using a fixed tabular format with greedy decoding and self-consistency decoding, thereby validating the effectiveness of our methods."
                    },
                    {
                        "title": "MultiSpider: Towards Benchmarking Multilingual Text-to-SQL Semantic Parsing",
                        "abstract": "Text-to-SQL semantic parsing is an important NLP task, which greatly facilitates the interaction between users and the database and becomes the key component in many human-computer interaction systems. Much recent progress in text-to-SQL has been driven by large-scale datasets, but most of them are centered on English. In this work, we present MultiSpider, the largest multilingual text-to-SQL dataset which covers seven languages (English, German, French, Spanish, Japanese, Chinese, and Vietnamese). Upon MultiSpider, we further identify the lexical and structural challenges of text-to-SQL (caused by specific language properties and dialect sayings) and their intensity across different languages. Experimental results under three typical settings (zero-shot, monolingual and multilingual) reveal a 6.1% absolute drop in accuracy in non-English languages. Qualitative and quantitative analyses are conducted to understand the reason for the performance drop of each language. Besides the dataset, we also propose a simple schema augmentation framework SAVe (Schema-Augmentation-with-Verification), which significantly boosts the overall performance by about 1.8% and closes the 29.5% performance gap across languages."
                    },
                    {
                        "title": "UniSAr: A Unified Structure-Aware Autoregressive Language Model for Text-to-SQL",
                        "abstract": "Existing text-to-SQL semantic parsers are typically designed for particular settings such as handling queries that span multiple tables, domains or turns which makes them ineffective when applied to different settings. We present UniSAr (Unified Structure-Aware Autoregressive Language Model), which benefits from directly using an off-the-shelf language model architecture and demonstrates consistently high performance under different settings. Specifically, UniSAr extends existing autoregressive language models to incorporate three non-invasive extensions to make them structure-aware: (1) adding structure mark to encode database schema, conversation context, and their relationships; (2) constrained decoding to decode well structured SQL for a given database schema; and (3) SQL completion to complete potential missing JOIN relationships in SQL based on database schema. On seven well-known text-to-SQL datasets covering multi-domain, multi-table and multi-turn, UniSAr demonstrates highly comparable or better performance to the most advanced specifically-designed text-to-SQL models. Importantly, our UniSAr is non-invasive, such that other core model advances in text-to-SQL can also adopt our extensions to further enhance performance."
                    },
                    {
                        "title": "MixPro: Simple yet Effective Data Augmentation for Prompt-based Learning",
                        "abstract": "Prompt-based learning has shown considerable promise in reformulating various downstream tasks as cloze problems by combining original input with a predetermined template. This approach demonstrates its effectiveness, especially in few-shot learning scenarios, where the model is trained on a scarce amount of data. Despite its successes, the limited templates and text in few-shot prompt-based learning scenarios leave significant room for performance improvement. Moreover, existing methods sometimes resort to model ensembles, which, while effective, could potentially hamper model efficiency due to increased computational demands. To address these issues, we introduce MixPro, an augmentation method designed to augment both the vanilla input text and the templates. We implement this through the token-level, the sentence-level, and the template-level Mixup strategies. The experimental results on five few-shot datasets show that MixPro outperforms other augmentation baselines, improving model performance by an average of 5.08% compared to before augmentation."
                    },
                    {
                        "title": "RegMix: Data Mixture as Regression for Language Model Pre-training",
                        "abstract": "The data mixture for large language model pre-training significantly impacts performance, yet how to determine an effective mixture remains unclear. We propose RegMix to automatically identify a high-performing data mixture by formulating it as a regression task. RegMix involves training a set of small models with diverse data mixtures and fitting a regression model to predict their performance given their respective mixtures. With the fitted regression model, we simulate the top-ranked mixture and use it to train a large-scale model with orders of magnitude more compute. To empirically validate RegMix, we train 512 models with 1M parameters for 1B tokens of different mixtures to fit the regression model and find the optimal mixture. Using this mixture we train a 1B parameter model for 25B tokens (i.e. 1000x larger and 25x longer) which we find performs best among 64 candidate 1B parameter models with other mixtures. Further, our method demonstrates superior performance compared to human selection and achieves results that match or surpass DoReMi, while utilizing only 10% of the compute budget. Our experiments also show that (1) Data mixtures significantly impact performance with single-task performance variations of up to 14.6%; (2) Web corpora rather than data perceived as high-quality like Wikipedia have the strongest positive correlation with downstream performance; (3) Domains interact in complex ways often contradicting common sense, thus automatic approaches like RegMix are needed; (4) Data mixture effects transcend scaling laws, and our approach captures the complexity by considering all domains together. Our code is available at https://github.com/sail-sg/regmix."
                    },
                    {
                        "title": "Towards Knowledge-Intensive Text-to-SQL Semantic Parsing with Formulaic Knowledge",
                        "abstract": "In this paper, we study the problem of knowledge-intensive text-to-SQL, in which domain knowledge is necessary to parse expert questions into SQL queries over domain-specific tables. We formalize this scenario by building a new Chinese benchmark KnowSQL consisting of domain-specific questions covering various domains. We then address this problem by presenting formulaic knowledge, rather than by annotating additional data examples. More concretely, we construct a formulaic knowledge bank as a domain knowledge base and propose a framework (ReGrouP) to leverage this formulaic knowledge during parsing. Experiments using ReGrouP demonstrate a significant 28.2% improvement overall on KnowSQL."
                    },
                    {
                        "title": "In-Context Transfer Learning: Demonstration Synthesis by Transferring Similar Tasks",
                        "abstract": "In-context learning (ICL) is an effective approach to help large language models (LLMs) adapt to various tasks by providing demonstrations of the target task. Considering the high cost of labeling demonstrations, many methods propose synthesizing demonstrations from scratch using LLMs. However, the quality of the demonstrations synthesized from scratch is limited by the capabilities and knowledge of LLMs. To address this, inspired by transfer learning, we propose In-Context Transfer Learning (ICTL), which synthesizes target task demonstrations by transferring labeled demonstrations from similar source tasks. ICTL consists of two steps: source sampling and target transfer. First, we define an optimization objective, which minimizes transfer error to sample source demonstrations similar to the target task. Then, we employ LLMs to transfer the sampled source demonstrations to the target task, matching the definition and format of the target task. Experiments on Super-NI show that ICTL outperforms synthesis from scratch by 2.0% on average, demonstrating the effectiveness of our method."
                    }
                ]
            },
            "f64f6e96-cfe1-4e34-8546-249a4733aa4c": {
                "pk": "f64f6e96-cfe1-4e34-8546-249a4733aa4c",
                "name": "Niklas Muennighoff",
                "collaborators": [
                    "Douwe Kiela",
                    "Qian Liu",
                    "Nouamane Tazi",
                    "Amanpreet Singh",
                    "Armel Zebaze",
                    "Binyuan Hui",
                    "Terry Yue Zhuo",
                    "Leandro von Werra",
                    "Lo\u00efc Magne",
                    "Nils Reimers"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Text Embeddings",
                    "Vision-Language Models"
                ],
                "publications": [
                    {
                        "title": "Vilio: State-of-the-art Visio-Linguistic Models applied to Hateful Memes",
                        "abstract": "This work presents Vilio, an implementation of state-of-the-art visio-linguistic models and their application to the Hateful Memes Dataset. The implemented models have been fitted into a uniform code-base and altered to yield better performance. The goal of Vilio is to provide a user-friendly starting point for any visio-linguistic problem. An ensemble of 5 different V+L models implemented in Vilio achieves 2nd place in the Hateful Memes Challenge out of 3,300 participants. The code is available at https://github.com/Muennighoff/vilio."
                    },
                    {
                        "title": "Diagnosing the Impact of AI on Radiology in China",
                        "abstract": "Artificial Intelligence will significantly impact the work environment of radiologists. I suggest that up to 50% of a radiologists work in 2021 will be performed by AI-models in 2025. However, it won't increase beyond that 50% level, as radiologists remain key for human-centered aspects of their job. I project that few to no radiologists will be laid off in China due to the existing supply shortage of radiology services in 2021. The application of AI in radiology could contribute 1.7 billion USD to China's GDP in 2025. It will further allow radiologists to start productive work up to four years earlier. AI in radiology will positively impact the health of patients and radiologists themselves."
                    },
                    {
                        "title": "SGPT: GPT Sentence Embeddings for Semantic Search",
                        "abstract": "Decoder transformers have continued increasing in scale reaching hundreds of billions of parameters. Due to their scale the same decoder sets state-of-the-art results on various language tasks via prompting or fine-tuning. Yet, these large foundation models remain unusable for the related fields of semantic search and sentence embeddings. This prevents possibly new state-of-the-art results and forces organizations to train and maintain separate models. To this end, we propose SGPT to use decoders for sentence embeddings and semantic search via prompting or fine-tuning. At 5.8 billion parameters SGPT improves on the previously best sentence embeddings by a margin of 7% and outperforms a concurrent method with 175 billion parameters as measured on the BEIR search benchmark. Code, models and result files are freely available at https://github.com/Muennighoff/sgpt."
                    },
                    {
                        "title": "MTEB: Massive Text Embedding Benchmark",
                        "abstract": "Text embeddings are commonly evaluated on a small set of datasets from a single task not covering their possible applications to other tasks. It is unclear whether state-of-the-art embeddings on semantic textual similarity (STS) can be equally well applied to other tasks like clustering or reranking. This makes progress in the field difficult to track, as various models are constantly being proposed without proper evaluation. To solve this problem, we introduce the Massive Text Embedding Benchmark (MTEB). MTEB spans 8 embedding tasks covering a total of 58 datasets and 112 languages. Through the benchmarking of 33 models on MTEB, we establish the most comprehensive benchmark of text embeddings to date. We find that no particular text embedding method dominates across all tasks. This suggests that the field has yet to converge on a universal text embedding method and scale it up sufficiently to provide state-of-the-art results on all embedding tasks. MTEB comes with open-source code and a public leaderboard at https://github.com/embeddings-benchmark/mteb."
                    },
                    {
                        "title": "The Scandinavian Embedding Benchmarks: Comprehensive Assessment of Multilingual and Monolingual Text Embedding",
                        "abstract": "The evaluation of English text embeddings has transitioned from evaluating a handful of datasets to broad coverage across many tasks through benchmarks such as MTEB. However, this is not the case for multilingual text embeddings due to a lack of available benchmarks. To address this problem, we introduce the Scandinavian Embedding Benchmark (SEB). SEB is a comprehensive framework that enables text embedding evaluation for Scandinavian languages across 24 tasks, 10 subtasks, and 4 task categories. Building on SEB, we evaluate more than 26 models, uncovering significant performance disparities between public and commercial solutions not previously captured by MTEB. We open-source SEB and integrate it with MTEB, thus bridging the text embedding evaluation gap for Scandinavian languages."
                    },
                    {
                        "title": "Generative Representational Instruction Tuning",
                        "abstract": "All text-based language problems can be reduced to either generation or embedding. Current models only perform well at one or the other. We introduce generative representational instruction tuning (GRIT) whereby a large language model is trained to handle both generative and embedding tasks by distinguishing between them through instructions. Compared to other open models, our resulting GritLM 7B sets a new state of the art on the Massive Text Embedding Benchmark (MTEB) and outperforms all models up to its size on a range of generative tasks. By scaling up further, GritLM 8x7B outperforms all open generative language models that we tried while still being among the best embedding models. Notably, we find that GRIT matches training on only generative or embedding data, thus we can unify both at no performance loss. Among other benefits, the unification via GRIT speeds up Retrieval-Augmented Generation (RAG) by > 60% for long documents, by no longer requiring separate retrieval and generation models. Models, code, etc. are freely available at https://github.com/ContextualAI/gritlm."
                    },
                    {
                        "title": "KTO: Model Alignment as Prospect Theoretic Optimization",
                        "abstract": "Kahneman & Tversky's $\\textit{prospect theory}$ tells us that humans perceive random variables in a biased but well-defined manner (1992); for example, humans are famously loss-averse. We show that objectives for aligning LLMs with human feedback implicitly incorporate many of these biases -- the success of these objectives (e.g., DPO) over cross-entropy minimization can partly be ascribed to them belonging to a family of loss functions that we call $\\textit{human-aware losses}$ (HALOs). However, the utility functions these methods attribute to humans still differ from those in the prospect theory literature. Using a Kahneman-Tversky model of human utility, we propose a HALO that directly maximizes the utility of generations instead of maximizing the log-likelihood of preferences, as current methods do. We call this approach KTO, and it matches or exceeds the performance of preference-based methods at scales from 1B to 30B, despite only learning from a binary signal of whether an output is desirable. More broadly, our work suggests that there is no one HALO that is universally superior; the best loss depends on the inductive biases most appropriate for a given setting, an oft-overlooked consideration."
                    },
                    {
                        "title": "C-Pack: Packed Resources For General Chinese Embeddings",
                        "abstract": "We introduce C-Pack, a package of resources that significantly advance the field of general Chinese embeddings. C-Pack includes three critical resources. 1) C-MTEB is a comprehensive benchmark for Chinese text embeddings covering 6 tasks and 35 datasets. 2) C-MTP is a massive text embedding dataset curated from labeled and unlabeled Chinese corpora for training embedding models. 3) C-TEM is a family of embedding models covering multiple sizes. Our models outperform all prior Chinese text embeddings on C-MTEB by up to +10% upon the time of the release. We also integrate and optimize the entire suite of training methods for C-TEM. Along with our resources on general Chinese embedding, we release our data and models for English text embeddings. The English models achieve state-of-the-art performance on MTEB benchmark; meanwhile, our released English data is 2 times larger than the Chinese data. All these resources are made publicly available at https://github.com/FlagOpen/FlagEmbedding."
                    },
                    {
                        "title": "Scaling Data-Constrained Language Models",
                        "abstract": "The current trend of scaling language models involves increasing both parameter count and training dataset size. Extrapolating this trend suggests that training dataset size may soon be limited by the amount of text data available on the internet. Motivated by this limit, we investigate scaling language models in data-constrained regimes. Specifically, we run a large set of experiments varying the extent of data repetition and compute budget, ranging up to 900 billion training tokens and 9 billion parameter models. We find that with constrained data for a fixed compute budget, training with up to 4 epochs of repeated data yields negligible changes to loss compared to having unique data. However, with more repetition, the value of adding compute eventually decays to zero. We propose and empirically validate a scaling law for compute optimality that accounts for the decreasing value of repeated tokens and excess parameters. Finally, we experiment with approaches mitigating data scarcity, including augmenting the training dataset with code data or removing commonly used filters. Models and datasets from our 400 training runs are freely available at https://github.com/huggingface/datablations."
                    },
                    {
                        "title": "OctoPack: Instruction Tuning Code Large Language Models",
                        "abstract": "Finetuning large language models (LLMs) on instructions leads to vast performance improvements on natural language tasks. We apply instruction tuning using code, leveraging the natural structure of Git commits, which pair code changes with human instructions. We compile CommitPack: 4 terabytes of Git commits across 350 programming languages. We benchmark CommitPack against other natural and synthetic code instructions (xP3x, Self-Instruct, OASST) on the 16B parameter StarCoder model, and achieve state-of-the-art performance among models not trained on OpenAI outputs, on the HumanEval Python benchmark (46.2% pass@1). We further introduce HumanEvalPack, expanding the HumanEval benchmark to a total of 3 coding tasks (Code Repair, Code Explanation, Code Synthesis) across 6 languages (Python, JavaScript, Java, Go, C++, Rust). Our models, OctoCoder and OctoGeeX, achieve the best performance across HumanEvalPack among all permissive models, demonstrating CommitPack's benefits in generalizing to a wider set of languages and natural coding tasks. Code, models and data are freely available at https://github.com/bigcode-project/octopack."
                    },
                    {
                        "title": "OLMoE: Open Mixture-of-Experts Language Models",
                        "abstract": "We introduce OLMoE, a fully open, state-of-the-art language model leveraging sparse Mixture-of-Experts (MoE). OLMoE-1B-7B has 7 billion (B) parameters but uses only 1B per input token. We pretrain it on 5 trillion tokens and further adapt it to create OLMoE-1B-7B-Instruct. Our models outperform all available models with similar active parameters, even surpassing larger ones like Llama2-13B-Chat and DeepSeekMoE-16B. We present various experiments on MoE training, analyze routing in our model showing high specialization, and open-source all aspects of our work: model weights, training data, code, and logs."
                    },
                    {
                        "title": "Astraios: Parameter-Efficient Instruction Tuning Code Large Language Models",
                        "abstract": "The high cost of full-parameter fine-tuning (FFT) of Large Language Models (LLMs) has led to a series of parameter-efficient fine-tuning (PEFT) methods. However, it remains unclear which methods provide the best cost-performance trade-off at different model scales. We introduce Astraios, a suite of 28 instruction-tuned OctoCoder models using 7 tuning methods and 4 model sizes up to 16 billion parameters. Through investigations across 5 tasks and 8 different datasets encompassing both code comprehension and code generation tasks, we find that FFT generally leads to the best downstream performance across all scales, and PEFT methods differ significantly in their efficacy based on the model scale. LoRA usually offers the most favorable trade-off between cost and performance. Further investigation into the effects of these methods on both model robustness and code security reveals that larger models tend to demonstrate reduced robustness and less security. At last, we explore the relationships among updated parameters, cross-entropy loss, and task performance. We find that the tuning effectiveness observed in small models generalizes well to larger models, and the validation loss in instruction tuning can be a reliable indicator of overall downstream performance."
                    },
                    {
                        "title": "RegMix: Data Mixture as Regression for Language Model Pre-training",
                        "abstract": "The data mixture for large language model pre-training significantly impacts performance, yet how to determine an effective mixture remains unclear. We propose RegMix to automatically identify a high-performing data mixture by formulating it as a regression task. RegMix involves training a set of small models with diverse data mixtures and fitting a regression model to predict their performance given their respective mixtures. With the fitted regression model, we simulate the top-ranked mixture and use it to train a large-scale model with orders of magnitude more compute. To empirically validate RegMix, we train 512 models with 1M parameters for 1B tokens of different mixtures to fit the regression model and find the optimal mixture. Using this mixture we train a 1B parameter model for 25B tokens (i.e. 1000x larger and 25x longer) which we find performs best among 64 candidate 1B parameter models with other mixtures. Further, our method demonstrates superior performance compared to human selection and achieves results that match or surpass DoReMi, while utilizing only 10% of the compute budget. Our experiments also show that (1) Data mixtures significantly impact performance with single-task performance variations of up to 14.6%; (2) Web corpora rather than data perceived as high-quality like Wikipedia have the strongest positive correlation with downstream performance; (3) Domains interact in complex ways often contradicting common sense, thus automatic approaches like RegMix are needed; (4) Data mixture effects transcend scaling laws, and our approach captures the complexity by considering all domains together. Our code is available at https://github.com/sail-sg/regmix."
                    },
                    {
                        "title": "KMMLU: Measuring Massive Multitask Language Understanding in Korean",
                        "abstract": "We propose KMMLU, a new Korean benchmark with 35,030 expert-level multiple-choice questions across 45 subjects ranging from humanities to STEM. While prior Korean benchmarks are translated from existing English benchmarks, KMMLU is collected from original Korean exams, capturing linguistic and cultural aspects of the Korean language. We test 27 public and proprietary LLMs and observe the best public model to score 50.5%, leaving significant room for improvement. This model was primarily trained for English and Chinese, not Korean. Current LLMs tailored to Korean, such as Polyglot-Ko, perform far worse. Surprisingly, even the most capable proprietary LLMs, e.g., GPT-4 and HyperCLOVA X do not exceed 60%. This suggests that further work is needed to improve LLMs for Korean, and we believe KMMLU offers the appropriate tool to track this progress. We make our dataset publicly available on the Hugging Face Hub and integrate the benchmark into EleutherAI's Language Model Evaluation Harness."
                    }
                ]
            },
            "9362dc63-f27a-4d40-ba22-9e15d978824a": {
                "pk": "9362dc63-f27a-4d40-ba22-9e15d978824a",
                "name": "Zhongwei Wan",
                "collaborators": [
                    "Mi Zhang",
                    "Zhenyu Liang",
                    "Yunfan Li",
                    "Xin Wang",
                    "Hui Shen",
                    "Yu Zheng",
                    "Jing Xiong",
                    "Xiping Hu",
                    "Min Yang",
                    "Chengming Li"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Deep Learning",
                    "Multi-Objective Optimization",
                    "Cross-Modal Learning"
                ],
                "publications": [
                    {
                        "title": "Text Classification: A Perspective of Deep Learning Methods",
                        "abstract": "In recent years, with the rapid development of information on the Internet, the number of complex texts and documents has increased exponentially, which requires a deeper understanding of deep learning methods in order to accurately classify texts using deep learning techniques, and thus deep learning methods have become increasingly important in text classification. Text classification is a class of tasks that automatically classifies a set of documents into multiple predefined categories based on their content and subject matter. Thus, the main goal of text classification is to enable users to extract information from textual resources and process processes such as retrieval, classification, and machine learning techniques together in order to classify different categories. Many new techniques of deep learning have already achieved excellent results in natural language processing. The success of these learning algorithms relies on their ability to understand complex models and non-linear relationships in data. However, finding the right structure, architecture, and techniques for text classification is a challenge for researchers. This paper introduces deep learning-based text classification algorithms, including important steps required for text classification tasks such as feature extraction, feature reduction, and evaluation strategies and methods. At the end of the article, different deep learning text classification methods are compared and summarized."
                    },
                    {
                        "title": "Many-Objective Estimation of Distribution Optimization Algorithm Based on WGAN-GP",
                        "abstract": "Estimation of distribution algorithms (EDA) are stochastic optimization algorithms. EDA establishes a probability model to describe the distribution of solution from the perspective of population macroscopically by statistical learning method, and then randomly samples the probability model to generate a new population. EDA can better solve multi-objective optimal problems (MOPs). However, the performance of EDA decreases in solving many-objective optimal problems (MaOPs), which contains more than three objectives. Reference Vector Guided Evolutionary Algorithm (RVEA), based on the EDA framework, can better solve MaOPs. In our paper, we use the framework of RVEA. However, we generate the new population by Wasserstein Generative Adversarial Networks-Gradient Penalty (WGAN-GP) instead of using crossover and mutation. WGAN-GP have advantages of fast convergence, good stability and high sample quality. WGAN-GP learn the mapping relationship from standard normal distribution to given data set distribution based on a given data set subject to the same distribution. It can quickly generate populations with high diversity and good convergence. To measure the performance, RM-MEDA, MOPSO and NSGA-II are selected to perform comparison experiments over DTLZ and LSMOP test suites with 3-, 5-, 8-, 10- and 15-objective."
                    },
                    {
                        "title": "Large Scale Many-Objective Optimization Driven by Distributional Adversarial Networks",
                        "abstract": "Estimation of distribution algorithms (EDA) as one of the EAs is a stochastic optimization problem which establishes a probability model to describe the distribution of solutions and randomly samples the probability model to create offspring and optimize model and population. Reference Vector Guided Evolutionary (RVEA) based on the EDA framework, having a better performance to solve MaOPs. Besides, using the generative adversarial networks to generate offspring solutions is also a state-of-art thought in EAs instead of crossover and mutation. In this paper, we will propose a novel algorithm based on RVEA[1] framework and using Distributional Adversarial Networks (DAN) [2]to generate new offspring. DAN uses a new distributional framework for adversarial training of neural networks and operates on genuine samples rather than a single point because the framework also leads to more stable training and extraordinarily better mode coverage compared to single-point-sample methods. Thereby, DAN can quickly generate offspring with high convergence regarding the same distribution of data. In addition, we also use Large-Scale Multi-Objective Optimization Based on A Competitive Swarm Optimizer (LMOCSO)[3] to adopts a new two-stage strategy to update the position in order to significantly increase the search efficiency to find optimal solutions in huge decision space. The propose new algorithm will be tested on 9 benchmark problems in Large scale multi-objective problems (LSMOP). To measure the performance, we will compare our proposal algorithm with some state-of-art EAs e.g., RM-MEDA[4], MO-CMA[10] and NSGA-II."
                    },
                    {
                        "title": "Famba-V: Fast Vision Mamba with Cross-Layer Token Fusion",
                        "abstract": "Mamba and Vision Mamba (Vim) models have shown their potential as an alternative to methods based on Transformer architecture. This work introduces Fast Mamba for Vision (Famba-V), a cross-layer token fusion technique to enhance the training efficiency of Vim models. The key idea of Famba-V is to identify and fuse similar tokens across different Vim layers based on a suit of cross-layer strategies instead of simply applying token fusion uniformly across all the layers that existing works propose. We evaluate the performance of Famba-V on CIFAR-100. Our results show that Famba-V is able to enhance the training efficiency of Vim models by reducing both training time and peak memory usage during training. Moreover, the proposed cross-layer strategies allow Famba-V to deliver superior accuracy-efficiency trade-offs. These results all together demonstrate Famba-V as a promising efficiency enhancement technique for Vim models."
                    },
                    {
                        "title": "SVD-LLM: Truncation-aware Singular Value Decomposition for Large Language Model Compression",
                        "abstract": "The advancements in Large Language Models (LLMs) have been hindered by their substantial sizes, which necessitate LLM compression methods for practical deployment. Singular Value Decomposition (SVD) offers a promising solution for LLM compression. However, state-of-the-art SVD-based LLM compression methods have two key limitations: truncating smaller singular values may lead to higher compression loss, and the lack of update on the compressed weight after SVD truncation. In this work, we propose SVD-LLM, a new SVD-based LLM compression method that addresses the limitations of existing methods. SVD-LLM incorporates a truncation-aware data whitening strategy to ensure a direct mapping between singular values and compression loss. Moreover, SVD-LLM adopts a layer-wise closed-form model parameter update strategy to compensate for accuracy degradation under high compression ratios. We evaluate SVD-LLM on a total of 10 datasets and eight models from three different LLM families at four different scales. Our results demonstrate the superiority of SVD-LLM over state-of-the-arts, especially at high model compression ratios."
                    },
                    {
                        "title": "Self-consistent Reasoning For Solving Math Word Problems",
                        "abstract": "Math word problems (MWPs) is a task that automatically derives solution expression from a giving math problems in text. The previous studies suffer from spurious correlations between input text and output expression. To mitigate this issue, we propose a self-consistent reasoning framework called SCR, which attempts to adopt a pruning strategy to correct the output distribution shift so as to implicitly fix those spurious correlative samples. Specifically, we firstly obtain a sub-network by pruning a roberta2tree model, for the sake to use the gap on output distribution between the original roberta2tree model and the pruned sub-network to expose spurious correlative samples. Then, we calibrate the output distribution shift by applying symmetric Kullback-Leibler divergence to alleviate spurious correlations. In addition, SCR generates equivalent expressions, thereby, capturing the original text's logic rather than relying on hints from original text. Extensive experiments on two large-scale benchmarks demonstrate that our model substantially outperforms the strong baseline methods."
                    },
                    {
                        "title": "ETP: Learning Transferable ECG Representations via ECG-Text Pre-training",
                        "abstract": "In the domain of cardiovascular healthcare, the Electrocardiogram (ECG) serves as a critical, non-invasive diagnostic tool. Although recent strides in self-supervised learning (SSL) have been promising for ECG representation learning, these techniques often require annotated samples and struggle with classes not present in the fine-tuning stages. To address these limitations, we introduce ECG-Text Pre-training (ETP), an innovative framework designed to learn cross-modal representations that link ECG signals with textual reports. For the first time, this framework leverages the zero-shot classification task in the ECG domain. ETP employs an ECG encoder along with a pre-trained language model to align ECG signals with their corresponding textual reports. The proposed framework excels in both linear evaluation and zero-shot classification tasks, as demonstrated on the PTB-XL and CPSC2018 datasets, showcasing its ability for robust and generalizable cross-modal ECG feature learning."
                    }
                ]
            },
            "268d0b7f-5d55-4de2-b1d8-3515ffe8790b": {
                "pk": "268d0b7f-5d55-4de2-b1d8-3515ffe8790b",
                "name": "Ping Luo",
                "collaborators": [
                    "Xiaogang Wang",
                    "Zhanglin Peng",
                    "Jiamin Ren",
                    "Ruimao Zhang",
                    "Wenqi Shao",
                    "Shitao Tang",
                    "Xingang Pan",
                    "Ping Tan",
                    "Ke Lin",
                    "Yiyang Luo"
                ],
                "domain": [
                    "Deep Learning",
                    "Computer Vision",
                    "Natural Language Processing",
                    "Generative Models"
                ],
                "publications": [
                    {
                        "title": "Do Normalization Layers in a Deep ConvNet Really Need to Be Distinct?",
                        "abstract": "Yes, they do. This work investigates a perspective for deep learning: whether different normalization layers in a ConvNet require different normalizers. This is the first step towards understanding this phenomenon. We allow each convolutional layer to be stacked before a switchable normalization (SN) that learns to choose a normalizer from a pool of normalization methods. Through systematic experiments in ImageNet, COCO, Cityscapes, and ADE20K, we answer three questions: (a) Is it useful to allow each normalization layer to select its own normalizer? (b) What impacts the choices of normalizers? (c) Do different tasks and datasets prefer different normalizers? Our results suggest that (1) using distinct normalizers improves both learning and generalization of a ConvNet; (2) the choices of normalizers are more related to depth and batch size, but less relevant to parameter initialization, learning rate decay, and solver; (3) different tasks and datasets have different behaviors when learning to select normalizers."
                    },
                    {
                        "title": "Channel Equilibrium Networks for Learning Deep Representation",
                        "abstract": "Convolutional Neural Networks (CNNs) are typically constructed by stacking multiple building blocks, each of which contains a normalization layer such as batch normalization (BN) and a rectified linear function such as ReLU. However, this work shows that the combination of normalization and rectified linear function leads to inhibited channels, which have small magnitude and contribute little to the learned feature representation, impeding the generalization ability of CNNs. Unlike prior arts that simply removed the inhibited channels, we propose to \"wake them up\" during training by designing a novel neural building block, termed Channel Equilibrium (CE) block, which enables channels at the same layer to contribute equally to the learned representation. We show that CE is able to prevent inhibited channels both empirically and theoretically. CE has several appealing benefits. (1) It can be integrated into many advanced CNN architectures such as ResNet and MobileNet, outperforming their original networks. (2) CE has an interesting connection with the Nash Equilibrium, a well-known solution of a non-cooperative game. (3) Extensive experiments show that CE achieves state-of-the-art performance on various challenging benchmarks such as ImageNet and COCO."
                    },
                    {
                        "title": "Zero-shot Generative Linguistic Steganography",
                        "abstract": "Generative linguistic steganography attempts to hide secret messages into covertext. Previous studies have generally focused on the statistical differences between the covertext and stegotext, however, ill-formed stegotext can readily be identified by humans. In this paper, we propose a novel zero-shot approach based on in-context learning for linguistic steganography to achieve better perceptual and statistical imperceptibility. We also design several new metrics and reproducible language evaluations to measure the imperceptibility of the stegotext. Our experimental results indicate that our method produces $1.926\\times$ more innocent and intelligible stegotext than any other method."
                    },
                    {
                        "title": "Clothing Co-Parsing by Joint Image Segmentation and Labeling",
                        "abstract": "This paper aims at developing an integrated system of clothing co-parsing, in order to jointly parse a set of clothing images (unsegmented but annotated with tags) into semantic configurations. We propose a data-driven framework consisting of two phases of inference. The first phase, referred as \"image co-segmentation\", iterates to extract consistent regions on images and jointly refines the regions over all images by employing the exemplar-SVM (E-SVM) technique [23]. In the second phase (i.e. \"region co-labeling\"), we construct a multi-image graphical model by taking the segmented regions as vertices, and incorporate several contexts of clothing configuration (e.g., item location and mutual interactions). The joint label assignment can be solved using the efficient Graph Cuts algorithm. In addition to evaluate our framework on the Fashionista dataset [30], we construct a dataset called CCP consisting of 2098 high-resolution street fashion photos to demonstrate the performance of our system. We achieve 90.29% / 88.23% segmentation accuracy and 65.52% / 63.89% recognition rate on the Fashionista and the CCP datasets, respectively, which are superior compared with state-of-the-art methods."
                    },
                    {
                        "title": "Deep Self-Learning From Noisy Labels",
                        "abstract": "ConvNets achieve good results when training from clean data, but learning from noisy labels significantly degrades performances and remains challenging. Unlike previous works constrained by many conditions, making them infeasible to real noisy cases, this work presents a novel deep self-learning framework to train a robust network on the real noisy datasets without extra supervision. The proposed approach has several appealing benefits. (1) Different from most existing work, it does not rely on any assumption on the distribution of the noisy labels, making it robust to real noises. (2) It does not need extra clean supervision or accessorial network to help training. (3) A self-learning framework is proposed to train the network in an iterative end-to-end manner, which is effective and efficient. Extensive experiments in challenging benchmarks such as Clothing1M and Food101-N show that our approach outperforms its counterparts in all empirical settings."
                    },
                    {
                        "title": "FaceFeat-GAN: a Two-Stage Approach for Identity-Preserving Face Synthesis",
                        "abstract": "The advance of Generative Adversarial Networks (GANs) enables realistic face image synthesis. However, synthesizing face images that preserve facial identity as well as have high diversity within each identity remains challenging. To address this problem, we present FaceFeat-GAN, a novel generative model that improves both image quality and diversity by using two stages. Unlike existing single-stage models that map random noise to image directly, our two-stage synthesis includes the first stage of diverse feature generation and the second stage of feature-to-image rendering. The competitions between generators and discriminators are carefully designed in both stages with different objective functions. Specially, in the first stage, they compete in the feature domain to synthesize various facial features rather than images. In the second stage, they compete in the image domain to render photo-realistic images that contain high diversity but preserve identity. Extensive experiments show that FaceFeat-GAN generates images that not only retain identity information but also have high diversity and quality, significantly outperforming previous methods."
                    },
                    {
                        "title": "AdaX: Adaptive Gradient Descent with Exponential Long Term Memory",
                        "abstract": "Although adaptive optimization algorithms such as Adam show fast convergence in many machine learning tasks, this paper identifies a problem of Adam by analyzing its performance in a simple non-convex synthetic problem, showing that Adam's fast convergence would possibly lead the algorithm to local minimums. To address this problem, we improve Adam by proposing a novel adaptive gradient descent algorithm named AdaX. Unlike Adam that ignores the past gradients, AdaX exponentially accumulates the long-term gradient information in the past during training, to adaptively tune the learning rate. We thoroughly prove the convergence of AdaX in both the convex and non-convex settings. Extensive experiments show that AdaX outperforms Adam in various tasks of computer vision and natural language processing and can catch up with Stochastic Gradient Descent."
                    },
                    {
                        "title": "Going Deeper Into Face Detection: A Survey",
                        "abstract": "Face detection is a crucial first step in many facial recognition and face analysis systems. Early approaches for face detection were mainly based on classifiers built on top of hand-crafted features extracted from local image regions, such as Haar Cascades and Histogram of Oriented Gradients. However, these approaches were not powerful enough to achieve a high accuracy on images of from uncontrolled environments. With the breakthrough work in image classification using deep neural networks in 2012, there has been a huge paradigm shift in face detection. Inspired by the rapid progress of deep learning in computer vision, many deep learning based frameworks have been proposed for face detection over the past few years, achieving significant improvements in accuracy. In this work, we provide a detailed overview of some of the most representative deep learning based face detection methods by grouping them into a few major categories, and present their core architectural designs and accuracies on popular benchmarks. We also describe some of the most popular face detection datasets. Finally, we discuss some current challenges in the field, and suggest potential future research directions."
                    },
                    {
                        "title": "DiffusionDet: Diffusion Model for Object Detection",
                        "abstract": "We propose DiffusionDet, a new framework that formulates object detection as a denoising diffusion process from noisy boxes to object boxes. During the training stage, object boxes diffuse from ground-truth boxes to random distribution, and the model learns to reverse this noising process. In inference, the model refines a set of randomly generated boxes to the output results in a progressive way. Our work possesses an appealing property of flexibility, which enables the dynamic number of boxes and iterative evaluation. The extensive experiments on the standard benchmarks show that DiffusionDet achieves favorable performance compared to previous well-established detectors. For example, DiffusionDet achieves 5.3 AP and 4.8 AP gains when evaluated with more boxes and iteration steps, under a zero-shot transfer setting from COCO to CrowdHuman. Our code is available at https://github.com/ShoufaChen/DiffusionDet."
                    },
                    {
                        "title": "Guideline Learning for In-context Information Extraction",
                        "abstract": "Large language models (LLMs) can perform a new task by merely conditioning on task instructions and a few input-output examples, without optimizing any parameters. This is called In-Context Learning (ICL). In-context Information Extraction (IE) has recently garnered attention in the research community. However, the performance of In-context IE generally lags behind the state-of-the-art supervised expert models. We highlight a key reason for this shortfall: underspecified task description. The limited-length context struggles to thoroughly express the intricate IE task instructions and various edge cases, leading to misalignment in task comprehension with humans. In this paper, we propose a Guideline Learning (GL) framework for In-context IE which reflectively learns and follows guidelines. During the learning phrase, GL automatically synthesizes a set of guidelines based on a few error cases, and during inference, GL retrieves helpful guidelines for better ICL. Moreover, we propose a self-consistency-based active learning method to enhance the efficiency of GL. Experiments on event extraction and relation extraction show that GL can significantly improve the performance of in-context IE."
                    }
                ]
            },
            "58c02101-ad7c-47fb-89ec-2d55a538e4a1": {
                "pk": "58c02101-ad7c-47fb-89ec-2d55a538e4a1",
                "name": "Min Lin",
                "collaborators": [
                    "Shuicheng Yan",
                    "Xu He",
                    "Qiang Chen",
                    "Shuo Li",
                    "Xuan Luo",
                    "Jie Fu",
                    "Yoshua Bengio"
                ],
                "domain": [
                    "Generative Adversarial Networks",
                    "Continual Learning",
                    "Deep Learning",
                    "Automatic Differentiation"
                ],
                "publications": [
                    {
                        "title": "Softmax GAN",
                        "abstract": "Softmax GAN is a novel variant of Generative Adversarial Network (GAN). The key idea of Softmax GAN is to replace the classification loss in the original GAN with a softmax cross-entropy loss in the sample space of one single batch. In the adversarial learning of $N$ real training samples and $M$ generated samples, the target of discriminator training is to distribute all the probability mass to the real samples, each with probability $\\frac{1}{M}$, and distribute zero probability to generated data. In the generator training phase, the target is to assign equal probability to all data points in the batch, each with probability $\\frac{1}{M+N}$. While the original GAN is closely related to Noise Contrastive Estimation (NCE), we show that Softmax GAN is the Importance Sampling version of GAN. We futher demonstrate with experiments that this simple change stabilizes GAN training."
                    },
                    {
                        "title": "Automatic Functional Differentiation in JAX",
                        "abstract": "We extend JAX with the capability to automatically differentiate higher-order functions (functionals and operators). By representing functions as a generalization of arrays, we seamlessly use JAX's existing primitive system to implement higher-order functions. We present a set of primitive operators that serve as foundational building blocks for constructing several key types of functionals. For every introduced primitive operator, we derive and implement both linearization and transposition rules, aligning with JAX's internal protocols for forward and reverse mode automatic differentiation. This enhancement allows for functional differentiation in the same syntax traditionally use for functions. The resulting functional gradients are themselves functions ready to be invoked in python. We showcase this tool's efficacy and simplicity through applications where functional derivatives are indispensable. The source code of this work is released at https://github.com/sail-sg/autofd ."
                    },
                    {
                        "title": "Continual Learning from the Perspective of Compression",
                        "abstract": "Connectionist models such as neural networks suffer from catastrophic forgetting. In this work, we study this problem from the perspective of information theory and define forgetting as the increase of description lengths of previous data when they are compressed with a sequentially learned model. In addition, we show that continual learning approaches based on variational posterior approximation and generative replay can be considered as approximations to two prequential coding methods in compression, namely, the Bayesian mixture code and maximum likelihood (ML) plug-in code. We compare these approaches in terms of both compression and forgetting and empirically study the reasons that limit the performance of continual learning methods based on variational posterior approximation. To address these limitations, we propose a new continual learning method that combines ML plug-in and Bayesian mixture codes."
                    },
                    {
                        "title": "Network In Network",
                        "abstract": "We propose a novel deep network structure called \"Network In Network\" (NIN) to enhance model discriminability for local patches within the receptive field. The conventional convolutional layer uses linear filters followed by a nonlinear activation function to scan the input. Instead, we build micro neural networks with more complex structures to abstract the data within the receptive field. We instantiate the micro neural network with a multilayer perceptron, which is a potent function approximator. The feature maps are obtained by sliding the micro networks over the input in a similar manner as CNN; they are then fed into the next layer. Deep NIN can be implemented by stacking mutiple of the above described structure. With enhanced local modeling via the micro network, we are able to utilize global average pooling over feature maps in the classification layer, which is easier to interpret and less prone to overfitting than traditional fully connected layers. We demonstrated the state-of-the-art classification performances with NIN on CIFAR-10 and CIFAR-100, and reasonable performances on SVHN and MNIST datasets."
                    },
                    {
                        "title": "Purine: A bi-graph based deep learning framework",
                        "abstract": "In this paper, we introduce a novel deep learning framework, termed Purine. In Purine, a deep network is expressed as a bipartite graph (bi-graph), which is composed of interconnected operators and data tensors. With the bi-graph abstraction, networks are easily solvable with event-driven task dispatcher. We then demonstrate that different parallelism schemes over GPUs and/or CPUs on single or multiple PCs can be universally implemented by graph composition. This eases researchers from coding for various parallelization schemes, and the same dispatcher can be used for solving variant graphs. Scheduled by the task dispatcher, memory transfers are fully overlapped with other computations, which greatly reduce the communication overhead and help us achieve approximate linear acceleration."
                    },
                    {
                        "title": "Conditional Computation for Continual Learning",
                        "abstract": "Catastrophic forgetting of connectionist neural networks is caused by the global sharing of parameters among all training examples. In this study, we analyze parameter sharing under the conditional computation framework where the parameters of a neural network are conditioned on each input example. At one extreme, if each input example uses a disjoint set of parameters, there is no sharing of parameters thus no catastrophic forgetting. At the other extreme, if the parameters are the same for every example, it reduces to the conventional neural network. We then introduce a clipped version of maxout networks which lies in the middle, i.e. parameters are shared partially among examples. Based on the parameter sharing analysis, we can locate a limited set of examples that are interfered when learning a new example. We propose to perform rehearsal on this set to prevent forgetting, which is termed as conditional rehearsal. Finally, we demonstrate the effectiveness of the proposed method in an online non-stationary setup, where updates are made after each new example and the distribution of the received example shifts over time."
                    }
                ]
            },
            "9dd97040-26fb-4ea0-819c-f814ae44ac20": {
                "pk": "9dd97040-26fb-4ea0-819c-f814ae44ac20",
                "name": "Ngai Wong",
                "collaborators": [
                    "Kim Batselier",
                    "Rui Lin",
                    "Chaofan Tao",
                    "Binxiao Huang",
                    "Zhongming Chen",
                    "Wenjian Yu",
                    "Cong Chen",
                    "Luca Daniel",
                    "Ching Yun Ko",
                    "Andrzej Cichocki"
                ],
                "domain": [
                    "Tensor Decomposition",
                    "Machine Learning",
                    "Neural Networks",
                    "System Identification"
                ],
                "publications": [
                    {
                        "title": "Symmetric Tensor Decomposition by an Iterative Eigendecomposition Algorithm",
                        "abstract": "We present an iterative algorithm, called the symmetric tensor eigen-rank-one iterative decomposition (STEROID), for decomposing a symmetric tensor into a real linear combination of symmetric rank-1 unit-norm outer factors using only eigendecompositions and least-squares fitting. Originally designed for a symmetric tensor with an order being a power of two, STEROID is shown to be applicable to any order through an innovative tensor embedding technique. Numerical examples demonstrate the high efficiency and accuracy of the proposed scheme even for large scale problems. Furthermore, we show how STEROID readily solves a problem in nonlinear block-structured system identification and nonlinear state-space identification."
                    },
                    {
                        "title": "A QR Algorithm for Symmetric Tensors",
                        "abstract": "We extend the celebrated QR algorithm for matrices to symmetric tensors. The algorithm, named QR algorithm for symmetric tensors (QRST), exhibits similar properties to its matrix version, and allows the derivation of a shifted implementation with faster convergence. We further show that multiple tensor eigenpairs can be found from a local permutation heuristic which is effectively a tensor similarity transform, resulting in the permuted version of QRST called PQRST. Examples demonstrate the remarkable effectiveness of the proposed schemes for finding stable and unstable eigenpairs not found by previous tensor power methods."
                    },
                    {
                        "title": "Matrix output extension of the tensor network Kalman filter with an application in MIMO Volterra system identification",
                        "abstract": "This article extends the tensor network Kalman filter to matrix outputs with an application in recursive identification of discrete-time nonlinear multiple-input-multiple-output (MIMO) Volterra systems. This extension completely supersedes previous work, where only $l$ scalar outputs were considered. The Kalman tensor equations are modified to accommodate for matrix outputs and their implementation using tensor networks is discussed. The MIMO Volterra system identification application requires the conversion of the output model matrix with a row-wise Kronecker product structure into its corresponding tensor network, for which we propose an efficient algorithm. Numerical experiments demonstrate both the efficacy of the proposed matrix conversion algorithm and the improved convergence of the Volterra kernel estimates when using matrix outputs."
                    },
                    {
                        "title": "A constructive arbitrary-degree Kronecker product decomposition of tensors",
                        "abstract": "We propose the tensor Kronecker product singular value decomposition~(TKPSVD) that decomposes a real $k$-way tensor $\\mathcal{A}$ into a linear combination of tensor Kronecker products with an arbitrary number of $d$ factors $\\mathcal{A} = \\sum_{j=1}^R \\sigma_j\\, \\mathcal{A}^{(d)}_j \\otimes \\cdots \\otimes \\mathcal{A}^{(1)}_j$. We generalize the matrix Kronecker product to tensors such that each factor $\\mathcal{A}^{(i)}_j$ in the TKPSVD is a $k$-way tensor. The algorithm relies on reshaping and permuting the original tensor into a $d$-way tensor, after which a polyadic decomposition with orthogonal rank-1 terms is computed. We prove that for many different structured tensors, the Kronecker product factors $\\mathcal{A}^{(1)}_j,\\ldots,\\mathcal{A}^{(d)}_j$ are guaranteed to inherit this structure. In addition, we introduce the new notion of general symmetric tensors, which includes many different structures such as symmetric, persymmetric, centrosymmetric, Toeplitz and Hankel tensors."
                    },
                    {
                        "title": "ODG-Q: Robust Quantization via Online Domain Generalization",
                        "abstract": "Quantizing neural networks to low-bitwidth is important for model deployment on resource-limited edge hardware. Although a quantized network has a smaller model size and memory footprint, it is fragile to adversarial attacks. However, few methods study the robustness and training efficiency of quantized networks. To this end, we propose a new method by recasting robust quantization as an online domain generalization problem, termed ODG-Q, which generates diverse adversarial data at a low cost during training. ODG-Q consistently outperforms existing works against various adversarial attacks. For example, on CIFAR-10 dataset, ODG-Q achieves 49.2% average improvements under five common white-box attacks and 21.7% average improvements under five common black-box attacks, with a training cost similar to that of natural training (viz. without adversaries). To our best knowledge, this work is the first work that trains both quantized and binary neural networks on ImageNet that consistently improve robustness under different attacks. We also provide a theoretical insight of ODG-Q that accounts for the bound of model risk on attacked data."
                    },
                    {
                        "title": "Tensor Network alternating linear scheme for MIMO Volterra system identification",
                        "abstract": "This article introduces two Tensor Network-based iterative algorithms for the identification of high-order discrete-time nonlinear multiple-input multiple-output (MIMO) Volterra systems. The system identification problem is rewritten in terms of a Volterra tensor, which is never explicitly constructed, thus avoiding the curse of dimensionality. It is shown how each iteration of the two identification algorithms involves solving a linear system of low computational complexity. The proposed algorithms are guaranteed to monotonically converge and numerical stability is ensured through the use of orthogonal matrix factorizations. The performance and accuracy of the two identification algorithms are illustrated by numerical experiments, where accurate degree-10 MIMO Volterra models are identified in about 1 second in Matlab on a standard desktop pc."
                    },
                    {
                        "title": "A Tensor Network Kalman filter with an application in recursive MIMO Volterra system identification",
                        "abstract": "This article introduces a Tensor Network Kalman filter, which can estimate state vectors that are exponentially large without ever having to explicitly construct them. The Tensor Network Kalman filter also easily accommodates the case where several different state vectors need to be estimated simultaneously. The key lies in rewriting the standard Kalman equations as tensor equations and then implementing them using Tensor Networks, which effectively transforms the exponential storage cost and computational complexity into a linear one. We showcase the power of the proposed framework through an application in recursive nonlinear system identification of high-order discrete-time multiple-input multiple-output (MIMO) Volterra systems. The identification problem is transformed into a linear state estimation problem wherein the state vector contains all Volterra kernel coefficients and is estimated using the Tensor Network Kalman filter. The accuracy and robustness of the scheme are demonstrated via numerical experiments, which show that updating the Kalman filter estimate of a state vector of length $10^9$ and its covariance matrix takes about 0.007s on a standard desktop computer in Matlab."
                    },
                    {
                        "title": "Computing low-rank approximations of large-scale matrices with the Tensor Network randomized SVD",
                        "abstract": "We propose a new algorithm for the computation of a singular value decomposition (SVD) low-rank approximation of a matrix in the Matrix Product Operator (MPO) format, also called the Tensor Train Matrix format. Our tensor network randomized SVD (TNrSVD) algorithm is an MPO implementation of the randomized SVD algorithm that is able to compute dominant singular values and their corresponding singular vectors. In contrast to the state-of-the-art tensor-based alternating least squares SVD (ALS-SVD) and modified alternating least squares SVD (MALS-SVD) matrix approximation methods, TNrSVD can be up to 17 times faster while achieving the same accuracy. In addition, our TNrSVD algorithm also produces accurate approximations in particular cases where both ALS-SVD and MALS-SVD fail to converge. We also propose a new algorithm for the fast conversion of a sparse matrix into its corresponding MPO form, which is up to 509 times faster than the standard Tensor Train SVD (TT-SVD) method while achieving machine precision accuracy. The efficiency and accuracy of both algorithms are demonstrated in numerical experiments."
                    },
                    {
                        "title": "Kernelized Support Tensor Train Machines",
                        "abstract": "Tensor, a multi-dimensional data structure, has been exploited recently in the machine learning community. Traditional machine learning approaches are vector- or matrix-based, and cannot handle tensorial data directly. In this paper, we propose a tensor train (TT)-based kernel technique for the first time, and apply it to the conventional support vector machine (SVM) for image classification. Specifically, we propose a kernelized support tensor train machine that accepts tensorial input and preserves the intrinsic kernel property. The main contributions are threefold. First, we propose a TT-based feature mapping procedure that maintains the TT structure in the feature space. Second, we demonstrate two ways to construct the TT-based kernel function while considering consistency with the TT inner product and preservation of information. Third, we show that it is possible to apply different kernel functions on different data modes. In principle, our method tensorizes the standard SVM on its input structure and kernel mapping scheme. Extensive experiments are performed on real-world tensor data, which demonstrates the superiority of the proposed scheme under few-sample high-dimensional inputs."
                    },
                    {
                        "title": "Tensor network subspace identification of polynomial state space models",
                        "abstract": "This article introduces a tensor network subspace algorithm for the identification of specific polynomial state space models. The polynomial nonlinearity in the state space model is completely written in terms of a tensor network, thus avoiding the curse of dimensionality. We also prove how the block Hankel data matrices in the subspace method can be exactly represented by low rank tensor networks, reducing the computational and storage complexity significantly. The performance and accuracy of our subspace identification algorithm are illustrated by numerical experiments, showing that our tensor network implementation is around 20 times faster than the standard matrix implementation before the latter fails due to insufficient memory, is robust with respect to noise and can model real-world systems."
                    },
                    {
                        "title": "What Do Adversarially trained Neural Networks Focus: A Fourier Domain-based Study",
                        "abstract": "Although many fields have witnessed the superior performance brought about by deep learning, the robustness of neural networks remains an open issue. Specifically, a small adversarial perturbation on the input may cause the model to produce a completely different output. Such poor robustness implies many potential hazards, especially in security-critical applications, e.g., autonomous driving and mobile robotics. This work studies what information the adversarially trained model focuses on. Empirically, we notice that the differences between the clean and adversarial data are mainly distributed in the low-frequency region. We then find that an adversarially-trained model is more robust than its naturally-trained counterpart due to the reason that the former pays more attention to learning the dominant information in low-frequency components. In addition, we consider two common ways to improve model robustness, namely, by data augmentation and by using stronger network architectures, and understand these techniques from a frequency-domain perspective. We are hopeful this work can shed light on the design of more robust neural networks."
                    },
                    {
                        "title": "Coarse to Fine: Image Restoration Boosted by Multi-Scale Low-Rank Tensor Completion",
                        "abstract": "Existing low-rank tensor completion (LRTC) approaches aim at restoring a partially observed tensor by imposing a global low-rank constraint on the underlying completed tensor. However, such a global rank assumption suffers the trade-off between restoring the originally details-lacking parts and neglecting the potentially complex objects, making the completion performance unsatisfactory on both sides. To address this problem, we propose a novel and practical strategy for image restoration that restores the partially observed tensor in a coarse-to-fine (C2F) manner, which gets rid of such trade-off by searching proper local ranks for both low- and high-rank parts. Extensive experiments are conducted to demonstrate the superiority of the proposed C2F scheme. The codes are available at: https://github.com/RuiLin0212/C2FLRTC."
                    },
                    {
                        "title": "MERACLE: Constructive layer-wise conversion of a Tensor Train into a MERA",
                        "abstract": "In this article two new algorithms are presented that convert a given data tensor train into either a Tucker decomposition with orthogonal matrix factors or a multi-scale entanglement renormalization ansatz (MERA). The Tucker core tensor is never explicitly computed but stored as a tensor train instead, resulting in both computationally and storage efficient algorithms. Both the multilinear Tucker-ranks as well as the MERA-ranks are automatically determined by the algorithm for a given upper bound on the relative approximation error. In addition, an iterative algorithm with low computational complexity based on solving an orthogonal Procrustes problem is proposed for the first time to retrieve optimal rank-lowering disentangler tensors, which are a crucial component in the construction of a low-rank MERA. Numerical experiments demonstrate the effectiveness of the proposed algorithms together with the potential storage benefit of a low-rank MERA over a tensor train."
                    },
                    {
                        "title": "AET-EFN: A Versatile Design for Static and Dynamic Event-Based Vision",
                        "abstract": "The neuromorphic event cameras, which capture the optical changes of a scene, have drawn increasing attention due to their high speed and low power consumption. However, the event data are noisy, sparse, and nonuniform in the spatial-temporal domain with an extremely high temporal resolution, making it challenging to design backend algorithms for event-based vision. Existing methods encode events into point-cloud-based or voxel-based representations, but suffer from noise and/or information loss. Additionally, there is little research that systematically studies how to handle static and dynamic scenes with one universal design for event-based vision. This work proposes the Aligned Event Tensor (AET) as a novel event data representation, and a neat framework called Event Frame Net (EFN), which enables our model for event-based vision under static and dynamic scenes. The proposed AET and EFN are evaluated on various datasets, and proved to surpass existing state-of-the-art methods by large margins. Our method is also efficient and achieves the fastest inference speed among others."
                    },
                    {
                        "title": "Frequency Regularization for Improving Adversarial Robustness",
                        "abstract": "Deep neural networks are incredibly vulnerable to crafted, human-imperceptible adversarial perturbations. Although adversarial training (AT) has proven to be an effective defense approach, we find that the AT-trained models heavily rely on the input low-frequency content for judgment, accounting for the low standard accuracy. To close the large gap between the standard and robust accuracies during AT, we investigate the frequency difference between clean and adversarial inputs, and propose a frequency regularization (FR) to align the output difference in the spectral domain. Besides, we find Stochastic Weight Averaging (SWA), by smoothing the kernels over epochs, further improves the robustness. Among various defense schemes, our method achieves the strongest robustness against attacks by PGD-20, C\\&W and Autoattack, on a WideResNet trained on CIFAR-10 without any extra data."
                    },
                    {
                        "title": "A Spectral Perspective towards Understanding and Improving Adversarial Robustness",
                        "abstract": "Deep neural networks (DNNs) are incredibly vulnerable to crafted, imperceptible adversarial perturbations. While adversarial training (AT) has proven to be an effective defense approach, the AT mechanism for robustness improvement is not fully understood. This work investigates AT from a spectral perspective, adding new insights to the design of effective defenses. In particular, we show that AT induces the deep model to focus more on the low-frequency region, which retains the shape-biased representations, to gain robustness. Further, we find that the spectrum of a white-box attack is primarily distributed in regions the model focuses on, and the perturbation attacks the spectral bands where the model is vulnerable. Based on this observation, to train a model tolerant to frequency-varying perturbation, we propose a spectral alignment regularization (SAR) such that the spectral output inferred by an attacked adversarial input stays as close as possible to its natural input counterpart. Experiments demonstrate that SAR and its weight averaging (WA) extension could significantly improve the robust accuracy by 1.14% ~ 3.87% relative to the standard AT, across multiple datasets (CIFAR-10, CIFAR-100 and Tiny ImageNet), and various attacks (PGD, C&W and Autoattack), without any extra data."
                    },
                    {
                        "title": "LoRETTA: Low-Rank Economic Tensor-Train Adaptation for Ultra-Low-Parameter Fine-Tuning of Large Language Models",
                        "abstract": "Various parameter-efficient fine-tuning (PEFT) techniques have been proposed to enable computationally efficient fine-tuning while maintaining model performance. However, existing PEFT methods are still limited by the growing number of trainable parameters with the rapid deployment of Large Language Models (LLMs). To address this challenge, we present LoRETTA, an ultra-parameter-efficient framework that significantly reduces trainable parameters through tensor-train decomposition. Specifically, we propose two methods, named {LoRETTA}$_{adp}$ and {LoRETTA}$_{rep}$. The former employs tensorized adapters, offering a high-performance yet lightweight approach for the fine-tuning of LLMs. The latter emphasizes fine-tuning via weight parameterization with a set of small tensor factors. LoRETTA achieves comparable or better performance than most widely used PEFT methods with up to $100\\times$ fewer parameters on the LLaMA-2-7B models. Furthermore, empirical results demonstrate that the proposed method effectively improves training efficiency, enjoys better multi-task learning performance, and enhances the anti-overfitting capability. Plug-and-play codes built upon the Huggingface framework and PEFT library will be released."
                    },
                    {
                        "title": "Mixture-of-Subspaces in Low-Rank Adaptation",
                        "abstract": "In this paper, we introduce a subspace-inspired Low-Rank Adaptation (LoRA) method, which is computationally efficient, easy to implement, and readily applicable to large language, multimodal, and diffusion models. Initially, we equivalently decompose the weights of LoRA into two subspaces, and find that simply mixing them can enhance performance. To study such a phenomenon, we revisit it through a fine-grained subspace lens, showing that such modification is equivalent to employing a fixed mixer to fuse the subspaces. To be more flexible, we jointly learn the mixer with the original LoRA weights, and term the method Mixture-of-Subspaces LoRA (MoSLoRA). MoSLoRA consistently outperforms LoRA on tasks in different modalities, including commonsense reasoning, visual instruction tuning, and subject-driven text-to-image generation, demonstrating its effectiveness and robustness. Codes are available at https://github.com/wutaiqiang/MoSLoRA."
                    },
                    {
                        "title": "A Constructive Algorithm for Decomposing a Tensor into a Finite Sum of Orthonormal Rank-1 Terms",
                        "abstract": "We propose a constructive algorithm that decomposes an arbitrary real tensor into a finite sum of orthonormal rank-1 outer products. The algorithm, named TTr1SVD, works by converting the tensor into a tensor-train rank-1 (TTr1) series via the singular value decomposition (SVD). TTr1SVD naturally generalizes the SVD to the tensor regime with properties such as uniqueness for a fixed order of indices, orthogonal rank-1 outer product terms, and easy truncation error quantification. Using an outer product column table it also allows, for the first time, a complete characterization of all tensors orthogonal with the original tensor. Incidentally, this leads to a strikingly simple constructive proof showing that the maximum rank of a real $2 \\times 2 \\times 2$ tensor over the real field is 3. We also derive a conversion of the TTr1 decomposition into a Tucker decomposition with a sparse core tensor. Numerical examples illustrate each of the favorable properties of the TTr1 decomposition."
                    },
                    {
                        "title": "MiSC: Mixed Strategies Crowdsourcing",
                        "abstract": "Popular crowdsourcing techniques mostly focus on evaluating workers' labeling quality before adjusting their weights during label aggregation. Recently, another cohort of models regard crowdsourced annotations as incomplete tensors and recover unfilled labels by tensor completion. However, mixed strategies of the two methodologies have never been comprehensively investigated, leaving them as rather independent approaches. In this work, we propose $\\textit{MiSC}$ ($\\textbf{Mi}$xed $\\textbf{S}$trategies $\\textbf{C}$rowdsourcing), a versatile framework integrating arbitrary conventional crowdsourcing and tensor completion techniques. In particular, we propose a novel iterative Tucker label aggregation algorithm that outperforms state-of-the-art methods in extensive experiments."
                    }
                ]
            }
        }
    },
    "2410.01649": {
        "paper_data": {
            "title": "shapiq: Shapley Interactions for Machine Learning",
            "url": "http://arxiv.org/abs/2410.01649v1",
            "arxiv_id": "2410.01649",
            "authors": [
                "Maximilian Muschalik",
                "Hubert Baniecki",
                "Fabian Fumagalli",
                "Patrick Kolpaczki",
                "Barbara Hammer",
                "Eyke H\u00fcllermeier"
            ],
            "abstract": "Originally rooted in game theory, the Shapley Value (SV) has recently become an important tool in machine learning research. Perhaps most notably, it is used for feature attribution and data valuation in explainable artificial intelligence. Shapley Interactions (SIs) naturally extend the SV and address its limitations by assigning joint contributions to groups of entities, which enhance understanding of black box machine learning models. Due to the exponential complexity of computing SVs and SIs, various methods have been proposed that exploit structural assumptions or yield probabilistic estimates given limited resources. In this work, we introduce shapiq, an open-source Python package that unifies state-of-the-art algorithms to efficiently compute SVs and any-order SIs in an application-agnostic framework. Moreover, it includes a benchmarking suite containing 11 machine learning applications of SIs with pre-computed games and ground-truth values to systematically assess computational performance across domains. For practitioners, shapiq is able to explain and visualize any-order feature interactions in predictions of models, including vision transformers, language models, as well as XGBoost and LightGBM with TreeSHAP-IQ. With shapiq, we extend shap beyond feature attributions and consolidate the application of SVs and SIs in machine learning that facilitates future research. The source code and documentation are available at https://github.com/mmschlk/shapiq.",
            "introduction": " Introduction Assigning value to entities collectively performing a task is essential in various real-world applications of machine learning (ML) [ 60,71]. For instance, when reimbursing data providers based on the value of data [30,76], or justifying a model\u2019s prediction based on value of feature information [13,18,19, 52,73]. The fairdistribution of value among a group of entities is a central aspect of cooperative game theory, where the Shapley Value (SV) [ 72] defines a unique allocation scheme based on intuitive axioms. The SV is applicable to any game , i.e. a function that specifies the worth of all possible groups of entities, called coalitions. In ML, application-specific games were introduced [ 5,30,71,76,80], which typically require a definition of the overall worth and a notion of entities\u2019 absence [ 19]. The SV fairly distributes the overall worth among individuals by evaluating the game for all coalitions. However, it does not give insights on synergies or redundancies between entities. For instance, while two features such as latitude andlongitude convey separate information, only their joint consideration 38th Conference on Neural Information Processing Systems (NeurIPS 2024) Track on Datasets and Benchmarks.arXiv:2410.01649v1  [cs.LG]  2 Oct 2024Figure 1: Theshapiq Python package facilitates research on game theory for machine learning, including state-of-the-art approximation algorithms and pre-computed benchmarks. Moreover, it provides a simple interface for explaining predictions of machine learning models beyond feature attributions. reveals the synergy of encoding an exact location . The value of such a group of entities is known as aninteraction [33], or in this context feature interaction [29], and is crucial to understand predictions of complex ML models [20, 45, 46, 58, 63, 74, 78, 82], as illustrated in Figure 1. Shapley Interactions (SIs) [ 10,33,74,77] distribute the overall worth to all groups of entities up to a maximum explanation order . They satisfy axioms similar to the SV, to which they reduce for individuals, i.e. the lowest explanation order. In contrast, for the highest explanation order, which comprises an interaction for every coalition, the SIs yield the M\u00f6bius Interaction (MI), or M\u00f6bius transform [ 27,35,69]. The MIs are a fundamental concept in cooperative game theory that captures the isolated joint contribution , which allows to additively describe every coalition\u2019s worth by a sum of the contained MIs. With an increasing explanation order, the SIs comprise more components that finally yield the MIs as the most comprehensive explanation of the game at the cost of highest complexity [ 10,77]. While the SV and SIs provide an appealing theoretical concept, computing them without structural assumptions on the game requires exponential complexity [ 22]. For tree-based models, it was shown that SVs [ 51,83] and SIs [ 58,84] can be efficiently computed by exploiting the architecture. Moreover, game-agnostic stochastic approximators estimate the SV [11,17,43,52,54,61,65] and SIs [ 28,29,44,74,77] with a limited budget of game evaluations. Diverse applications of the SV have led to various techniques for its efficient computation [ 13]. Recently, extensions to any-order SIs addressed limitations of the SV and complemented interpretation of model predictions with higher-order feature interactions [ 10,29,74,77,78]. While stochastic approximators are applicable to any game, their evaluation is typically performed in an isolated application [ 29,47], such as feature interactions. Moreover, implementing such algorithms requires a strong mathematical background. Specifically, we discuss in more detail the class",
            "references": [
                {
                    "title": "KernelSHAP-IQ: Weighted Least-Square Optimization for Shapley Interactions",
                    "abstract": "The Shapley value (SV) is a prevalent approach of allocating credit to machine learning (ML) entities to understand black box ML models. Enriching such interpretations with higher-order interactions is inevitable for complex systems, where the Shapley Interaction Index (SII) is a direct axiomatic extension of the SV. While it is well-known that the SV yields an optimal approximation of any game via a weighted least square (WLS) objective, an extension of this result to SII has been a long-standing open problem, which even led to the proposal of an alternative index. In this work, we characterize higher-order SII as a solution to a WLS problem, which constructs an optimal approximation via SII and $k$-Shapley values ($k$-SII). We prove this representation for the SV and pairwise SII and give empirically validated conjectures for higher orders. As a result, we propose KernelSHAP-IQ, a direct extension of KernelSHAP for SII, and demonstrate state-of-the-art performance for feature interactions."
                },
                {
                    "title": "SVARM-IQ: Efficient Approximation of Any-order Shapley Interactions through Stratification",
                    "abstract": "Addressing the limitations of individual attribution scores via the Shapley value (SV), the field of explainable AI (XAI) has recently explored intricate interactions of features or data points. In particular, extensions of the SV, such as the Shapley Interaction Index (SII), have been proposed as a measure to still benefit from the axiomatic basis of the SV. However, similar to the SV, their exact computation remains computationally prohibitive. Hence, we propose with SVARM-IQ a sampling-based approach to efficiently approximate Shapley-based interaction indices of any order. SVARM-IQ can be applied to a broad class of interaction indices, including the SII, by leveraging a novel stratified representation. We provide non-asymptotic theoretical guarantees on its approximation quality and empirically demonstrate that SVARM-IQ achieves state-of-the-art estimation results in practical XAI scenarios on different model classes and application domains."
                },
                {
                    "title": "Beyond TreeSHAP: Efficient Computation of Any-Order Shapley Interactions for Tree Ensembles",
                    "abstract": "While shallow decision trees may be interpretable, larger ensemble models like gradient-boosted trees, which often set the state of the art in machine learning problems involving tabular data, still remain black box models. As a remedy, the Shapley value (SV) is a well-known concept in explainable artificial intelligence (XAI) research for quantifying additive feature attributions of predictions. The model-specific TreeSHAP methodology solves the exponential complexity for retrieving exact SVs from tree-based models. Expanding beyond individual feature attribution, Shapley interactions reveal the impact of intricate feature interactions of any order. In this work, we present TreeSHAP-IQ, an efficient method to compute any-order additive Shapley interactions for predictions of tree-based models. TreeSHAP-IQ is supported by a mathematical framework that exploits polynomial arithmetic to compute the interaction scores in a single recursive traversal of the tree, akin to Linear TreeSHAP. We apply TreeSHAP-IQ on state-of-the-art tree ensembles and explore interactions on well-established benchmark datasets."
                },
                {
                    "title": "Interventional SHAP Values and Interaction Values for Piecewise Linear Regression Trees",
                    "abstract": "In recent years, game-theoretic Shapley values have gained increasing attention with respect to local model explanation by feature attributions. While the approach using Shapley values is model-independent, their (exact) computation is usually intractable, so efficient model-specific algorithms have been devised including approaches for decision trees or their ensembles in general. Our work goes further in this direction by extending the interventional TreeSHAP algorithm to piecewise linear regression trees, which gained more attention in the past few years. To this end, we introduce a decomposition of the contribution function based on decision paths, which allows a more comprehensible formulation of SHAP algorithms for tree-based models. Our algorithm can also be readily applied to computing SHAP interaction values of these models. In particular, as the main contribution of this paper, we provide a more efficient approach of interventional SHAP for tree-based models by precomputing statistics of the background data based on the tree structure."
                },
                {
                    "title": "OpenDataVal: a Unified Benchmark for Data Valuation",
                    "abstract": "Assessing the quality and impact of individual data points is critical for improving model performance and mitigating undesirable biases within the training dataset. Several data valuation algorithms have been proposed to quantify data quality, however, there lacks a systemic and standardized benchmarking system for data valuation. In this paper, we introduce OpenDataVal, an easy-to-use and unified benchmark framework that empowers researchers and practitioners to apply and compare various data valuation algorithms. OpenDataVal provides an integrated environment that includes (i) a diverse collection of image, natural language, and tabular datasets, (ii) implementations of eleven different state-of-the-art data valuation algorithms, and (iii) a prediction model API that can import any models in scikit-learn. Furthermore, we propose four downstream machine learning tasks for evaluating the quality of data values. We perform benchmarking analysis using OpenDataVal, quantifying and comparing the efficacy of state-of-the-art data valuation approaches. We find that no single algorithm performs uniformly best across all tasks, and an appropriate algorithm should be employed for a user's downstream task. OpenDataVal is publicly available at https://opendataval.github.io with comprehensive documentation. Furthermore, we provide a leaderboard where researchers can evaluate the effectiveness of their own data valuation algorithms."
                },
                {
                    "title": "On the Robustness of Removal-Based Feature Attributions",
                    "abstract": "To explain predictions made by complex machine learning models, many feature attribution methods have been developed that assign importance scores to input features. Some recent work challenges the robustness of these methods by showing that they are sensitive to input and model perturbations, while other work addresses this issue by proposing robust attribution methods. However, previous work on attribution robustness has focused primarily on gradient-based feature attributions, whereas the robustness of removal-based attribution methods is not currently well understood. To bridge this gap, we theoretically characterize the robustness properties of removal-based feature attributions. Specifically, we provide a unified analysis of such methods and derive upper bounds for the difference between intact and perturbed attributions, under settings of both input and model perturbations. Our empirical results on synthetic and real-world data validate our theoretical results and demonstrate their practical implications, including the ability to increase attribution robustness by improving the model's Lipschitz regularity."
                },
                {
                    "title": "Explaining Predictive Uncertainty with Information Theoretic Shapley Values",
                    "abstract": "Researchers in explainable artificial intelligence have developed numerous methods for helping users understand the predictions of complex supervised learning models. By contrast, explaining the $\\textit{uncertainty}$ of model outputs has received relatively little attention. We adapt the popular Shapley value framework to explain various types of predictive uncertainty, quantifying each feature's contribution to the conditional entropy of individual model outputs. We consider games with modified characteristic functions and find deep connections between the resulting Shapley values and fundamental quantities from information theory and conditional independence testing. We outline inference procedures for finite sample error rate control with provable guarantees, and implement efficient algorithms that perform well in a range of experiments on real and simulated data. Our method has applications to covariate shift detection, active learning, feature selection, and active feature-value acquisition."
                },
                {
                    "title": "SHAP-IQ: Unified Approximation of any-order Shapley Interactions",
                    "abstract": "Predominately in explainable artificial intelligence (XAI) research, the Shapley value (SV) is applied to determine feature attributions for any black box model. Shapley interaction indices extend the SV to define any-order feature interactions. Defining a unique Shapley interaction index is an open research question and, so far, three definitions have been proposed, which differ by their choice of axioms. Moreover, each definition requires a specific approximation technique. Here, we propose SHAPley Interaction Quantification (SHAP-IQ), an efficient sampling-based approximator to compute Shapley interactions for arbitrary cardinal interaction indices (CII), i.e. interaction indices that satisfy the linearity, symmetry and dummy axiom. SHAP-IQ is based on a novel representation and, in contrast to existing methods, we provide theoretical guarantees for its approximation quality, as well as estimates for the variance of the point estimates. For the special case of SV, our approach reveals a novel representation of the SV and corresponds to Unbiased KernelSHAP with a greatly simplified calculation. We illustrate the computational efficiency and effectiveness by explaining language, image classification and high-dimensional synthetic models."
                },
                {
                    "title": "Approximating the Shapley Value without Marginal Contributions",
                    "abstract": "The Shapley value, which is arguably the most popular approach for assigning a meaningful contribution value to players in a cooperative game, has recently been used intensively in explainable artificial intelligence. Its meaningfulness is due to axiomatic properties that only the Shapley value satisfies, which, however, comes at the expense of an exact computation growing exponentially with the number of agents. Accordingly, a number of works are devoted to the efficient approximation of the Shapley value, most of them revolve around the notion of an agent's marginal contribution. In this paper, we propose with SVARM and Stratified SVARM two parameter-free and domain-independent approximation algorithms based on a representation of the Shapley value detached from the notion of marginal contribution. We prove unmatched theoretical guarantees regarding their approximation quality and provide empirical results including synthetic games as well as common explainability use cases comparing ourselves with state-of-the-art methods."
                },
                {
                    "title": "A k-additive Choquet integral-based approach to approximate the SHAP values for local interpretability in machine learning",
                    "abstract": "Besides accuracy, recent studies on machine learning models have been addressing the question on how the obtained results can be interpreted. Indeed, while complex machine learning models are able to provide very good results in terms of accuracy even in challenging applications, it is difficult to interpret them. Aiming at providing some interpretability for such models, one of the most famous methods, called SHAP, borrows the Shapley value concept from game theory in order to locally explain the predicted outcome of an instance of interest. As the SHAP values calculation needs previous computations on all possible coalitions of attributes, its computational cost can be very high. Therefore, a SHAP-based method called Kernel SHAP adopts an efficient strategy that approximate such values with less computational effort. In this paper, we also address local interpretability in machine learning based on Shapley values. Firstly, we provide a straightforward formulation of a SHAP-based method for local interpretability by using the Choquet integral, which leads to both Shapley values and Shapley interaction indices. Moreover, we also adopt the concept of $k$-additive games from game theory, which contributes to reduce the computational effort when estimating the SHAP values. The obtained results attest that our proposal needs less computations on coalitions of attributes to approximate the SHAP values."
                },
                {
                    "title": "Towards Human-Centered Explainable AI: A Survey of User Studies for Model Explanations",
                    "abstract": "Explainable AI (XAI) is widely viewed as a sine qua non for ever-expanding AI research. A better understanding of the needs of XAI users, as well as human-centered evaluations of explainable models are both a necessity and a challenge. In this paper, we explore how human-computer interaction (HCI) and AI researchers conduct user studies in XAI applications based on a systematic literature review. After identifying and thoroughly analyzing 97 core papers with human-based XAI evaluations over the past five years, we categorize them along the measured characteristics of explanatory methods, namely trust, understanding, usability, and human-AI collaboration performance. Our research shows that XAI is spreading more rapidly in certain application domains, such as recommender systems than in others, but that user evaluations are still rather sparse and incorporate hardly any insights from cognitive or social sciences. Based on a comprehensive discussion of best practices, i.e., common models, design choices, and measures in user studies, we propose practical guidelines on designing and conducting user studies for XAI researchers and practitioners. Lastly, this survey also highlights several open research directions, particularly linking psychological science and human-centered XAI."
                },
                {
                    "title": "From Shapley Values to Generalized Additive Models and back",
                    "abstract": "In explainable machine learning, local post-hoc explanation algorithms and inherently interpretable models are often seen as competing approaches. This work offers a partial reconciliation between the two by establishing a correspondence between Shapley Values and Generalized Additive Models (GAMs). We introduce $n$-Shapley Values, a parametric family of local post-hoc explanation algorithms that explain individual predictions with interaction terms up to order $n$. By varying the parameter $n$, we obtain a sequence of explanations that covers the entire range from Shapley Values up to a uniquely determined decomposition of the function we want to explain. The relationship between $n$-Shapley Values and this decomposition offers a functionally-grounded characterization of Shapley Values, which highlights their limitations. We then show that $n$-Shapley Values, as well as the Shapley Taylor- and Faith-Shap interaction indices, recover GAMs with interaction terms up to order $n$. This implies that the original Shapely Values recover GAMs without variable interactions. Taken together, our results provide a precise characterization of Shapley Values as they are being used in explainable machine learning. They also offer a principled interpretation of partial dependence plots of Shapley Values in terms of the underlying functional decomposition. A package for the estimation of different interaction indices is available at \\url{https://github.com/tml-tuebingen/nshap}."
                },
                {
                    "title": "OpenXAI: Towards a Transparent Evaluation of Model Explanations",
                    "abstract": "While several types of post hoc explanation methods have been proposed in recent literature, there is very little work on systematically benchmarking these methods. Here, we introduce OpenXAI, a comprehensive and extensible open-source framework for evaluating and benchmarking post hoc explanation methods. OpenXAI comprises of the following key components: (i) a flexible synthetic data generator and a collection of diverse real-world datasets, pre-trained models, and state-of-the-art feature attribution methods, and (ii) open-source implementations of eleven quantitative metrics for evaluating faithfulness, stability (robustness), and fairness of explanation methods, in turn providing comparisons of several explanation methods across a wide variety of metrics, models, and datasets. OpenXAI is easily extensible, as users can readily evaluate custom explanation methods and incorporate them into our leaderboards. Overall, OpenXAI provides an automated end-to-end pipeline that not only simplifies and standardizes the evaluation of post hoc explanation methods, but also promotes transparency and reproducibility in benchmarking these methods. While the first release of OpenXAI supports only tabular datasets, the explanation methods and metrics that we consider are general enough to be applicable to other data modalities. OpenXAI datasets and models, implementations of state-of-the-art explanation methods and evaluation metrics, are publicly available at this GitHub link."
                },
                {
                    "title": "Unsupervised Features Ranking via Coalitional Game Theory for Categorical Data",
                    "abstract": "Not all real-world data are labeled, and when labels are not available, it is often costly to obtain them. Moreover, as many algorithms suffer from the curse of dimensionality, reducing the features in the data to a smaller set is often of great utility. Unsupervised feature selection aims to reduce the number of features, often using feature importance scores to quantify the relevancy of single features to the task at hand. These scores can be based only on the distribution of variables and the quantification of their interactions. The previous literature, mainly investigating anomaly detection and clusters, fails to address the redundancy-elimination issue. We propose an evaluation of correlations among features to compute feature importance scores representing the contribution of single features in explaining the dataset's structure. Based on Coalitional Game Theory, our feature importance scores include a notion of redundancy awareness making them a tool to achieve redundancy-free feature selection. We show that the deriving features' selection outperforms competing methods in lowering the redundancy rate while maximizing the information contained in the data. We also introduce an approximated version of the algorithm to reduce the complexity of Shapley values' computations."
                },
                {
                    "title": "Faith-Shap: The Faithful Shapley Interaction Index",
                    "abstract": "Shapley values, which were originally designed to assign attributions to individual players in coalition games, have become a commonly used approach in explainable machine learning to provide attributions to input features for black-box machine learning models. A key attraction of Shapley values is that they uniquely satisfy a very natural set of axiomatic properties. However, extending the Shapley value to assigning attributions to interactions rather than individual players, an interaction index, is non-trivial: as the natural set of axioms for the original Shapley values, extended to the context of interactions, no longer specify a unique interaction index. Many proposals thus introduce additional less ''natural'' axioms, while sacrificing the key axiom of efficiency, in order to obtain unique interaction indices. In this work, rather than introduce additional conflicting axioms, we adopt the viewpoint of Shapley values as coefficients of the most faithful linear approximation to the pseudo-Boolean coalition game value function. By extending linear to $\\ell$-order polynomial approximations, we can then define the general family of faithful interaction indices. We show that by additionally requiring the faithful interaction indices to satisfy interaction-extensions of the standard individual Shapley axioms (dummy, symmetry, linearity, and efficiency), we obtain a unique Faithful Shapley Interaction index, which we denote Faith-Shap, as a natural generalization of the Shapley value to interactions. We then provide some illustrative contrasts of Faith-Shap with previously proposed interaction indices, and further investigate some of its interesting algebraic properties. We further show the computational efficiency of computing Faith-Shap, together with some additional qualitative insights, via some illustrative experiments."
                },
                {
                    "title": "Quantus: An Explainable AI Toolkit for Responsible Evaluation of Neural Network Explanations",
                    "abstract": "The evaluation of explanation methods is a research topic that has not yet been explored deeply, however, since explainability is supposed to strengthen trust in artificial intelligence, it is necessary to systematically review and compare explanation methods in order to confirm their correctness. Until now, no tool with focus on XAI evaluation exists that exhaustively and speedily allows researchers to evaluate the performance of explanations of neural network predictions. To increase transparency and reproducibility in the field, we therefore built Quantus -- a comprehensive, evaluation toolkit in Python that includes a growing, well-organised collection of evaluation metrics and tutorials for evaluating explainable methods. The toolkit has been thoroughly tested and is available under an open-source license on PyPi (or on https://github.com/understandable-machine-intelligence-lab/Quantus/)."
                },
                {
                    "title": "The Shapley Value in Machine Learning",
                    "abstract": "Over the last few years, the Shapley value, a solution concept from cooperative game theory, has found numerous applications in machine learning. In this paper, we first discuss fundamental concepts of cooperative game theory and axiomatic properties of the Shapley value. Then we give an overview of the most important applications of the Shapley value in machine learning: feature selection, explainability, multi-agent reinforcement learning, ensemble pruning, and data valuation. We examine the most crucial limitations of the Shapley value and point out directions for future research."
                },
                {
                    "title": "Incentivizing Collaboration in Machine Learning via Synthetic Data Rewards",
                    "abstract": "This paper presents a novel collaborative generative modeling (CGM) framework that incentivizes collaboration among self-interested parties to contribute data to a pool for training a generative model (e.g., GAN), from which synthetic data are drawn and distributed to the parties as rewards commensurate to their contributions. Distributing synthetic data as rewards (instead of trained models or money) offers task- and model-agnostic benefits for downstream learning tasks and is less likely to violate data privacy regulation. To realize the framework, we firstly propose a data valuation function using maximum mean discrepancy (MMD) that values data based on its quantity and quality in terms of its closeness to the true data distribution and provide theoretical results guiding the kernel choice in our MMD-based data valuation function. Then, we formulate the reward scheme as a linear optimization problem that when solved, guarantees certain incentives such as fairness in the CGM framework. We devise a weighted sampling algorithm for generating synthetic data to be distributed to each party as reward such that the value of its data and the synthetic data combined matches its assigned reward value by the reward scheme. We empirically show using simulated and real-world datasets that the parties' synthetic data rewards are commensurate to their contributions."
                },
                {
                    "title": "Using Shapley Values and Variational Autoencoders to Explain Predictive Models with Dependent Mixed Features",
                    "abstract": "Shapley values are today extensively used as a model-agnostic explanation framework to explain complex predictive machine learning models. Shapley values have desirable theoretical properties and a sound mathematical foundation in the field of cooperative game theory. Precise Shapley value estimates for dependent data rely on accurate modeling of the dependencies between all feature combinations. In this paper, we use a variational autoencoder with arbitrary conditioning (VAEAC) to model all feature dependencies simultaneously. We demonstrate through comprehensive simulation studies that our VAEAC approach to Shapley value estimation outperforms the state-of-the-art methods for a wide range of settings for both continuous and mixed dependent features. For high-dimensional settings, our VAEAC approach with a non-uniform masking scheme significantly outperforms competing methods. Finally, we apply our VAEAC approach to estimate Shapley value explanations for the Abalone data set from the UCI Machine Learning Repository."
                },
                {
                    "title": "Discovering and Explaining the Representation Bottleneck of DNNs",
                    "abstract": "This paper explores the bottleneck of feature representations of deep neural networks (DNNs), from the perspective of the complexity of interactions between input variables encoded in DNNs. To this end, we focus on the multi-order interaction between input variables, where the order represents the complexity of interactions. We discover that a DNN is more likely to encode both too simple interactions and too complex interactions, but usually fails to learn interactions of intermediate complexity. Such a phenomenon is widely shared by different DNNs for different tasks. This phenomenon indicates a cognition gap between DNNs and human beings, and we call it a representation bottleneck. We theoretically prove the underlying reason for the representation bottleneck. Furthermore, we propose a loss to encourage/penalize the learning of interactions of specific complexities, and analyze the representation capacities of interactions of different complexities."
                },
                {
                    "title": "Joint Shapley values: a measure of joint feature importance",
                    "abstract": "The Shapley value is one of the most widely used measures of feature importance partly as it measures a feature's average effect on a model's prediction. We introduce joint Shapley values, which directly extend Shapley's axioms and intuitions: joint Shapley values measure a set of features' average contribution to a model's prediction. We prove the uniqueness of joint Shapley values, for any order of explanation. Results for games show that joint Shapley values present different insights from existing interaction indices, which assess the effect of a feature within a set of features. The joint Shapley values provide intuitive results in ML attribution problems. With binary features, we present a presence-adjusted global value that is more consistent with local intuitions than the usual approach."
                },
                {
                    "title": "Synthetic Benchmarks for Scientific Research in Explainable Machine Learning",
                    "abstract": "As machine learning models grow more complex and their applications become more high-stakes, tools for explaining model predictions have become increasingly important. This has spurred a flurry of research in model explainability and has given rise to feature attribution methods such as LIME and SHAP. Despite their widespread use, evaluating and comparing different feature attribution methods remains challenging: evaluations ideally require human studies, and empirical evaluation metrics are often data-intensive or computationally prohibitive on real-world datasets. In this work, we address this issue by releasing XAI-Bench: a suite of synthetic datasets along with a library for benchmarking feature attribution algorithms. Unlike real-world datasets, synthetic datasets allow the efficient computation of conditional expected values that are needed to evaluate ground-truth Shapley values and other metrics. The synthetic datasets we release offer a wide variety of parameters that can be configured to simulate real-world data. We demonstrate the power of our library by benchmarking popular explainability techniques across several evaluation metrics and across a variety of settings. The versatility and efficiency of our library will help researchers bring their explainability methods from development to deployment. Our code is available at https://github.com/abacusai/xai-bench."
                },
                {
                    "title": "The Shapley Value of Classifiers in Ensemble Games",
                    "abstract": "What is the value of an individual model in an ensemble of binary classifiers? We answer this question by introducing a class of transferable utility cooperative games called ensemble games. In machine learning ensembles, pre-trained models cooperate to make classification decisions. To quantify the importance of models in these ensemble games, we define Troupe - an efficient algorithm that allocates payoffs based on approximate Shapley values of the classifiers. We argue that the Shapley value of models in these games is an effective decision metric for choosing a high-performing subset of models from the ensemble. Our analytical findings prove that our Shapley value estimation scheme is precise and scalable; its performance increases with the size of the dataset and ensemble. Empirical results on real-world graph classification tasks demonstrate that our algorithm produces high-quality estimates of the Shapley value. We find that Shapley values can be utilized for ensemble pruning and that adversarial models receive a low valuation. Complex classifiers are frequently found to be responsible for both correct and incorrect classification decisions."
                },
                {
                    "title": "dalex: Responsible Machine Learning with Interactive Explainability and Fairness in Python",
                    "abstract": "The increasing amount of available data, computing power, and the constant pursuit for higher performance results in the growing complexity of predictive models. Their black-box nature leads to opaqueness debt phenomenon inflicting increased risks of discrimination, lack of reproducibility, and deflated performance due to data drift. To manage these risks, good MLOps practices ask for better validation of model performance and fairness, higher explainability, and continuous monitoring. The necessity of deeper model transparency appears not only from scientific and social domains, but also emerging laws and regulations on artificial intelligence. To facilitate the development of responsible machine learning models, we showcase dalex, a Python package which implements the model-agnostic interface for interactive model exploration. It adopts the design crafted through the development of various tools for responsible machine learning; thus, it aims at the unification of the existing solutions. This library's source code and documentation are available under open license at this https URL"
                },
                {
                    "title": "Improving KernelSHAP: Practical Shapley Value Estimation via Linear Regression",
                    "abstract": "The Shapley value solution concept from cooperative game theory has become popular for interpreting ML models, but efficiently estimating Shapley values remains challenging, particularly in the model-agnostic setting. We revisit the idea of estimating Shapley values via linear regression to understand and improve upon this approach. By analyzing KernelSHAP alongside a newly proposed unbiased estimator, we develop techniques to detect its convergence and calculate uncertainty estimates. We also find that that the original version incurs a negligible increase in bias in exchange for a significant reduction in variance, and we propose a variance reduction technique that further accelerates the convergence of both estimators. Finally, we develop a version of KernelSHAP for stochastic cooperative games that yields fast new estimators for two global explanation methods."
                },
                {
                    "title": "Understanding Global Feature Contributions With Additive Importance Measures",
                    "abstract": "Understanding the inner workings of complex machine learning models is a long-standing problem and most recent research has focused on local interpretability. To assess the role of individual input features in a global sense, we explore the perspective of defining feature importance through the predictive power associated with each feature. We introduce two notions of predictive power (model-based and universal) and formalize this approach with a framework of additive importance measures, which unifies numerous methods in the literature. We then propose SAGE, a model-agnostic method that quantifies predictive power while accounting for feature interactions. Our experiments show that SAGE can be calculated efficiently and that it assigns more accurate importance values than other methods."
                },
                {
                    "title": "Explaining by Removing: A Unified Framework for Model Explanation",
                    "abstract": "Researchers have proposed a wide variety of model explanation approaches, but it remains unclear how most methods are related or when one method is preferable to another. We establish a new class of methods, removal-based explanations, that are based on the principle of simulating feature removal to quantify each feature's influence. These methods vary in several respects, so we develop a framework that characterizes each method along three dimensions: 1) how the method removes features, 2) what model behavior the method explains, and 3) how the method summarizes each feature's influence. Our framework unifies 25 existing methods, including several of the most widely used approaches (SHAP, LIME, Meaningful Perturbations, permutation tests). This new class of explanation methods has rich connections that we examine using tools that have been largely overlooked by the explainability literature. To anchor removal-based explanations in cognitive psychology, we show that feature removal is a simple application of subtractive counterfactual reasoning. Ideas from cooperative game theory shed light on the relationships and trade-offs among different methods, and we derive conditions under which all removal-based explanations have information-theoretic interpretations. Through this analysis, we develop a unified framework that helps practitioners better understand model explanation tools, and that offers a strong theoretical foundation upon which future explainability research can build."
                },
                {
                    "title": "Captum: A unified and generic model interpretability library for PyTorch",
                    "abstract": "In this paper we introduce a novel, unified, open-source model interpretability library for PyTorch [12]. The library contains generic implementations of a number of gradient and perturbation-based attribution algorithms, also known as feature, neuron and layer importance algorithms, as well as a set of evaluation metrics for these algorithms. It can be used for both classification and non-classification models including graph-structured models built on Neural Networks (NN). In this paper we give a high-level overview of supported attribution algorithms and show how to perform memory-efficient and scalable computations. We emphasize that the three main characteristics of the library are multimodality, extensibility and ease of use. Multimodality supports different modality of inputs such as image, text, audio or video. Extensibility allows adding new algorithms and features. The library is also designed for easy understanding and use. Besides, we also introduce an interactive visualization tool called Captum Insights that is built on top of Captum library and allows sample-based model debugging and visualization using feature importance metrics."
                },
                {
                    "title": "True to the Model or True to the Data?",
                    "abstract": "A variety of recent papers discuss the application of Shapley values, a concept for explaining coalitional games, for feature attribution in machine learning. However, the correct way to connect a machine learning model to a coalitional game has been a source of controversy. The two main approaches that have been proposed differ in the way that they condition on known features, using either (1) an interventional or (2) an observational conditional expectation. While previous work has argued that one of the two approaches is preferable in general, we argue that the choice is application dependent. Furthermore, we argue that the choice comes down to whether it is desirable to be true to the model or true to the data. We use linear models to investigate this choice. After deriving an efficient method for calculating observational conditional expectation Shapley values for linear models, we investigate how correlation in simulated data impacts the convergence of observational conditional expectation Shapley values. Finally, we present two real data examples that we consider to be representative of possible use cases for feature attribution -- (1) credit risk modeling and (2) biological discovery. We show how a different choice of value function performs better in each scenario, and how possible attributions are impacted by modeling choices."
                },
                {
                    "title": "How does this interaction affect me? Interpretable attribution for feature interactions",
                    "abstract": "Machine learning transparency calls for interpretable explanations of how inputs relate to predictions. Feature attribution is a way to analyze the impact of features on predictions. Feature interactions are the contextual dependence between features that jointly impact predictions. There are a number of methods that extract feature interactions in prediction models; however, the methods that assign attributions to interactions are either uninterpretable, model-specific, or non-axiomatic. We propose an interaction attribution and detection framework called Archipelago which addresses these problems and is also scalable in real-world settings. Our experiments on standard annotation labels indicate our approach provides significantly more interpretable explanations than comparable methods, which is important for analyzing the impact of interactions on predictions. We also provide accompanying visualizations of our approach that give new insights into deep neural networks."
                },
                {
                    "title": "High Dimensional Model Explanations: An Axiomatic Approach",
                    "abstract": "Complex black-box machine learning models are regularly used in critical decision-making domains. This has given rise to several calls for algorithmic explainability. Many explanation algorithms proposed in literature assign importance to each feature individually. However, such explanations fail to capture the joint effects of sets of features. Indeed, few works so far formally analyze high dimensional model explanations. In this paper, we propose a novel high dimension model explanation method that captures the joint effect of feature subsets. We propose a new axiomatization for a generalization of the Banzhaf index; our method can also be thought of as an approximation of a black-box model by a higher-order polynomial. In other words, this work justifies the use of the generalized Banzhaf index as a model explanation by showing that it uniquely satisfies a set of natural desiderata and that it is the optimal local approximation of a black-box model. Our empirical evaluation of our measure highlights how it manages to capture desirable behavior, whereas other measures that do not satisfy our axioms behave in an unpredictable manner."
                },
                {
                    "title": "Shapley explainability on the data manifold",
                    "abstract": "Explainability in AI is crucial for model development, compliance with regulation, and providing operational nuance to predictions. The Shapley framework for explainability attributes a model's predictions to its input features in a mathematically principled and model-agnostic way. However, general implementations of Shapley explainability make an untenable assumption: that the model's features are uncorrelated. In this work, we demonstrate unambiguous drawbacks of this assumption and develop two solutions to Shapley explainability that respect the data manifold. One solution, based on generative modelling, provides flexible access to data imputations; the other directly learns the Shapley value-function, providing performance and stability at the cost of flexibility. While \"off-manifold\" Shapley values can (i) give rise to incorrect explanations, (ii) hide implicit model dependence on sensitive attributes, and (iii) lead to unintelligible explanations in higher-dimensional data, on-manifold explainability overcomes these problems."
                },
                {
                    "title": "Problems with Shapley-value-based explanations as feature importance measures",
                    "abstract": "Game-theoretic formulations of feature importance have become popular as a way to \"explain\" machine learning models. These methods define a cooperative game between the features of a model and distribute influence among these input elements using some form of the game's unique Shapley values. Justification for these methods rests on two pillars: their desirable mathematical properties, and their applicability to specific motivations for explanations. We show that mathematical problems arise when Shapley values are used for feature importance and that the solutions to mitigate these necessarily induce further complexity, such as the need for causal reasoning. We also draw on additional literature to argue that Shapley values do not provide explanations which suit human-centric goals of explainability."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "OpenML-Python: an extensible Python API for OpenML",
                    "abstract": "OpenML is an online platform for open science collaboration in machine learning, used to share datasets and results of machine learning experiments. In this paper we introduce \\emph{OpenML-Python}, a client API for Python, opening up the OpenML platform for a wide range of Python-based tools. It provides easy access to all datasets, tasks and experiments on OpenML from within Python. It also provides functionality to conduct machine learning experiments, upload the results to OpenML, and reproduce results which are stored on OpenML. Furthermore, it comes with a scikit-learn plugin and a plugin mechanism to easily integrate other machine learning libraries written in Python into the OpenML ecosystem. Source code and documentation is available at this https URL."
                },
                {
                    "title": "The many Shapley values for model explanation",
                    "abstract": "The Shapley value has become a popular method to attribute the prediction of a machine-learning model on an input to its base features. The use of the Shapley value is justified by citing [16] showing that it is the \\emph{unique} method that satisfies certain good properties (\\emph{axioms}). \nThere are, however, a multiplicity of ways in which the Shapley value is operationalized in the attribution problem. These differ in how they reference the model, the training data, and the explanation context. These give very different results, rendering the uniqueness result meaningless. Furthermore, we find that previously proposed approaches can produce counterintuitive attributions in theory and in practice---for instance, they can assign non-zero attributions to features that are not even referenced by the model. \nIn this paper, we use the axiomatic approach to study the differences between some of the many operationalizations of the Shapley value for attribution, and propose a technique called Baseline Shapley (BShap) that is backed by a proper uniqueness result. We also contrast BShap with Integrated Gradients, another extension of Shapley value to the continuous setting."
                },
                {
                    "title": "Data Shapley: Equitable Valuation of Data for Machine Learning",
                    "abstract": "As data becomes the fuel driving technological and economic growth, a fundamental challenge is how to quantify the value of data in algorithmic predictions and decisions. For example, in healthcare and consumer markets, it has been suggested that individuals should be compensated for the data that they generate, but it is not clear what is an equitable valuation for individual data. In this work, we develop a principled framework to address data valuation in the context of supervised machine learning. Given a learning algorithm trained on $n$ data points to produce a predictor, we propose data Shapley as a metric to quantify the value of each training datum to the predictor performance. Data Shapley value uniquely satisfies several natural properties of equitable data valuation. We develop Monte Carlo and gradient-based methods to efficiently estimate data Shapley values in practical settings where complex learning algorithms, including neural networks, are trained on large datasets. In addition to being equitable, extensive experiments across biomedical, image and synthetic data demonstrate that data Shapley has several other benefits: 1) it is more powerful than the popular leave-one-out or leverage score in providing insight on what data is more valuable for a given learning task; 2) low Shapley value data effectively capture outliers and corruptions; 3) high Shapley value data inform what type of new data to acquire to improve the predictor."
                },
                {
                    "title": "The Shapley Taylor Interaction Index",
                    "abstract": "The attribution problem, that is the problem of attributing a model's prediction to its base features, is well-studied. We extend the notion of attribution to also apply to feature interactions. \nThe Shapley value is a commonly used method to attribute a model's prediction to its base features. We propose a generalization of the Shapley value called Shapley-Taylor index that attributes the model's prediction to interactions of subsets of features up to some size k. The method is analogous to how the truncated Taylor Series decomposes the function value at a certain point using its derivatives at a different point. In fact, we show that the Shapley Taylor index is equal to the Taylor Series of the multilinear extension of the set-theoretic behavior of the model. \nWe axiomatize this method using the standard Shapley axioms -- linearity, dummy, symmetry and efficiency -- and an additional axiom that we call the interaction distribution axiom. This new axiom explicitly characterizes how interactions are distributed for a class of functions that model pure interaction. \nWe contrast the Shapley-Taylor index against the previously proposed Shapley Interaction index (cf. [9]) from the cooperative game theory literature. We also apply the Shapley Taylor index to three models and identify interesting qualitative insights."
                },
                {
                    "title": "iNNvestigate neural networks!",
                    "abstract": "In recent years, deep neural networks have revolutionized many application domains of machine learning and are key components of many critical decision or predictive processes. Therefore, it is crucial that domain specialists can understand and analyze actions and pre- dictions, even of the most complex neural network architectures. Despite these arguments neural networks are often treated as black boxes. In the attempt to alleviate this short- coming many analysis methods were proposed, yet the lack of reference implementations often makes a systematic comparison between the methods a major effort. The presented library iNNvestigate addresses this by providing a common interface and out-of-the- box implementation for many analysis methods, including the reference implementation for PatternNet and PatternAttribution as well as for LRP-methods. To demonstrate the versatility of iNNvestigate, we provide an analysis of image classifications for variety of state-of-the-art neural network architectures."
                },
                {
                    "title": "Manipulating and Measuring Model Interpretability",
                    "abstract": "With machine learning models being increasingly used to aid decision making even in high-stakes domains, there has been a growing interest in developing interpretable models. Although many supposedly interpretable models have been proposed, there have been relatively few experimental studies investigating whether these models achieve their intended effects, such as making people more closely follow a model\u2019s predictions when it is beneficial for them to do so or enabling them to detect when a model has made a mistake. We present a sequence of pre-registered experiments (N = 3, 800) in which we showed participants functionally identical models that varied only in two factors commonly thought to make machine learning models more or less interpretable: the number of features and the transparency of the model (i.e., whether the model internals are clear or black box). Predictably, participants who saw a clear model with few features could better simulate the model\u2019s predictions. However, we did not find that participants more closely followed its predictions. Furthermore, showing participants a clear model meant that they were less able to detect and correct for the model\u2019s sizable mistakes, seemingly due to information overload. These counterintuitive findings emphasize the importance of testing over intuition when developing interpretable models."
                },
                {
                    "title": "LightGBM: A Highly Efficient Gradient Boosting Decision Tree",
                    "abstract": "Gradient Boosting Decision Tree (GBDT) is a popular machine learning algorithm, and has quite a few effective implementations such as XGBoost and pGBRT. Although many engineering optimizations have been adopted in these implementations, the efficiency and scalability are still unsatisfactory when the feature dimension is high and data size is large. A major reason is that for each feature, they need to scan all the data instances to estimate the information gain of all possible split points, which is very time consuming. To tackle this problem, we propose two novel techniques: \\emph{Gradient-based One-Side Sampling} (GOSS) and \\emph{Exclusive Feature Bundling} (EFB). With GOSS, we exclude a significant proportion of data instances with small gradients, and only use the rest to estimate the information gain. We prove that, since the data instances with larger gradients play a more important role in the computation of information gain, GOSS can obtain quite accurate estimation of the information gain with a much smaller data size. With EFB, we bundle mutually exclusive features (i.e., they rarely take nonzero values simultaneously), to reduce the number of features. We prove that finding the optimal bundling of exclusive features is NP-hard, but a greedy algorithm can achieve quite good approximation ratio (and thus can effectively reduce the number of features without hurting the accuracy of split point determination by much). We call our new GBDT implementation with GOSS and EFB \\emph{LightGBM}. Our experiments on multiple public datasets show that, LightGBM speeds up the training process of conventional GBDT by up to over 20 times while achieving almost the same accuracy."
                },
                {
                    "title": "OpenML Benchmarking Suites",
                    "abstract": "Machine learning research depends on objectively interpretable, comparable, and reproducible algorithm benchmarks. Therefore, we advocate the use of curated, comprehensive suites of machine learning tasks to standardize the setup, execution, and reporting of benchmarks. We enable this through software tools that help to create and leverage these benchmarking suites. These are seamlessly integrated into the OpenML platform, and accessible through interfaces in Python, Java, and R. OpenML benchmarking suites are (a) easy to use through standardized data formats, APIs, and client libraries; (b) machine-readable, with extensive meta-information on the included datasets; and (c) allow benchmarks to be shared and reused in future studies. We also present a first, carefully curated and practical benchmarking suite for classification: the OpenML Curated Classification benchmarking suite 2018 (OpenML-CC18)."
                },
                {
                    "title": "A Unified Approach to Interpreting Model Predictions",
                    "abstract": "Understanding why a model makes a certain prediction can be as crucial as the prediction's accuracy in many applications. However, the highest accuracy for large modern datasets is often achieved by complex models that even experts struggle to interpret, such as ensemble or deep learning models, creating a tension between accuracy and interpretability. In response, various methods have recently been proposed to help users interpret the predictions of complex models, but it is often unclear how these methods are related and when one method is preferable over another. To address this problem, we present a unified framework for interpreting predictions, SHAP (SHapley Additive exPlanations). SHAP assigns each feature an importance value for a particular prediction. Its novel components include: (1) the identification of a new class of additive feature importance measures, and (2) theoretical results showing there is a unique solution in this class with a set of desirable properties. The new class unifies six existing methods, notable because several recent methods in the class lack the proposed desirable properties. Based on insights from this unification, we present new methods that show improved computational performance and/or better consistency with human intuition than previous approaches."
                },
                {
                    "title": "XGBoost: A Scalable Tree Boosting System",
                    "abstract": "Tree boosting is a highly effective and widely used machine learning method. In this paper, we describe a scalable end-to-end tree boosting system called XGBoost, which is used widely by data scientists to achieve state-of-the-art results on many machine learning challenges. We propose a novel sparsity-aware algorithm for sparse data and weighted quantile sketch for approximate tree learning. More importantly, we provide insights on cache access patterns, data compression and sharding to build a scalable tree boosting system. By combining these insights, XGBoost scales beyond billions of examples using far fewer resources than existing systems."
                },
                {
                    "title": "Sobol' Indices and Shapley Value",
                    "abstract": "Global sensitivity analysis measures the importance of some input variables to a function $f$ by looking at the impact on $f$ of making large random perturbations to subsets of those variables. Using measures like those of Sobol' we can attribute importance to input variables based on the extent to which they help predict the target function $f$. There is a longstanding literature in economics and game theory that considers how to attribute the value of a team effort to individual members of that team. The primary result, known as the Shapley value, is the unique method satisfying some intuitively necessary criteria. In this paper we find the Shapley value of individual variables when we take \u201cvariance explained\u201d as their combined value. The result does not match either of the usual Sobol' indices. It is instead bracketed between them for variance explained or indeed any totally monotone game. Because those indices are comparatively easy to compute, Sobol' indices provide effectively computable bounds for..."
                },
                {
                    "title": "Bounding the Estimation Error of Sampling-based Shapley Value Approximation With/Without Stratifying",
                    "abstract": "The Shapley value is arguably the most central normative solution concept in cooperative game theory. It specifies a unique way in which the reward from cooperation can be \"fairly\" divided among players. While it has a wide range of real world applications, its use is in many cases hampered by the hardness of its computation. A number of researchers have tackled this problem by (1) focusing on classes of games where the Shapley value can be computed efficiently, or (2) proposing representation formalisms that facilitate such efficient computation, or (3) approximating the Shapley value in certain classes of games. However, given the classical \\textit{characteristic function} representation, the only attempt to approximate the Shapley value for the general class of games is due to Castro \\textit{et al.} \\cite{castro}. While this algorithm provides a bound on the approximation error, this bound is \\textit{asymptotic}, meaning that it only holds when the number of samples increases to infinity. On the other hand, when a finite number of samples is drawn, an unquantifiable error is introduced, meaning that the bound no longer holds. With this in mind, we provide non-asymptotic bounds on the estimation error for two cases: where (1) the \\textit{variance}, and (2) the \\textit{range}, of the players' marginal contributions is known. Furthermore, for the second case, we show that when the range is significantly large relative to the Shapley value, the bound can be improved (from $O(r,\\sqrt{\\nicefrac{1}{m}})$ to $O(\\sqrt{r},\\sqrt{\\nicefrac{1}{m}})$). Finally, we propose, and demonstrate the effectiveness of, using stratified sampling to improve the bounds."
                },
                {
                    "title": "Learning Word Vectors for Sentiment Analysis",
                    "abstract": "Unsupervised vector-based approaches to semantics can model rich lexical meanings, but they largely fail to capture sentiment information that is central to many word meanings and important for a wide range of NLP tasks. We present a model that uses a mix of unsupervised and supervised techniques to learn word vectors capturing semantic term--document information as well as rich sentiment content. The proposed model can leverage both continuous and multi-dimensional sentiment information as well as non-sentiment annotations. We instantiate the model to utilize the document-level sentiment polarity annotations present in many online documents (e.g. star ratings). We evaluate the model using small, widely used sentiment and subjectivity corpora and find it out-performs several previously introduced methods for sentiment classification. We also introduce a large dataset of movie reviews to serve as a more robust benchmark for work in this area."
                },
                {
                    "title": "Scikit-learn: Machine Learning in Python",
                    "abstract": "Scikit-learn is a Python module integrating a wide range of state-of-the-art machine learning algorithms for medium-scale supervised and unsupervised problems. This package focuses on bringing machine learning to non-specialists using a general-purpose high-level language. Emphasis is put on ease of use, performance, documentation, and API consistency. It has minimal dependencies and is distributed under the simplified BSD license, encouraging its use in both academic and commercial settings. Source code, binaries, and documentation can be downloaded from http://scikit-learn.sourceforge.net."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Feature Selection via Coalitional Game Theory",
                    "abstract": "We present and study the contribution-selection algorithm (CSA), a novel algorithm for feature selection. The algorithm is based on the multiperturbation shapley analysis (MSA), a framework that relies on game theory to estimate usefulness. The algorithm iteratively estimates the usefulness of features and selects them accordingly, using either forward selection or backward elimination. It can optimize various performance measures over unseen data such as accuracy, balanced error rate, and area under receiver-operator-characteristic curve. Empirical comparison with several other existing feature selection methods shows that the backward elimination variant of CSA leads to the most accurate classification results on an array of data sets."
                },
                {
                    "title": "Equivalent Representations of Set Functions",
                    "abstract": "This paper introduces four alternative representations of a set function: the MA\u00b6bius transformation, the co-MA\u00b6bius transformation, and the interactions between elements of any subset of a given set as extensions of Shapley and Banzhaf values. The links between the five equivalent representations of a set function are emphasized in this presentation."
                },
                {
                    "title": "Scaling Up the Accuracy of Naive-Bayes Classifiers: A Decision-Tree Hybrid",
                    "abstract": "Naive-Bayes induction algorithms were previously shown to be surprisingly accurate on many classification tasks even when the conditional independence assumption on which they are based is violated. However, most studies were done on small databases. We show that in some larger databases, the accuracy of Naive-Bayes does not scale up as well as decision trees. We then propose a new algorithm, NBTree, which induces a hybrid of decision-tree classifiers and Naive-Bayes classifiers: the decision-tree nodes contain univariate splits as regular decision-trees, but the leaves contain Naive-Bayesian classifiers. The approach retains the interpretability of Naive-Bayes and decision trees, while resulting in classifiers that frequently outperform both constituents, especially in the larger databases tested."
                },
                {
                    "title": "On the Complexity of Cooperative Solution Concepts",
                    "abstract": "We study from a complexity theoretic standpoint the various solution concepts arising in cooperative game theory. We use as a vehicle for this study a game in which the players are nodes of a graph with weights on the edges, and the value of a coalition is determined by the total weight of the edges contained in it. The Shapley value is always easy to compute. The core is easy to characterize when the game is convex, and is intractable (NP-complete) otherwise. Similar results are shown for the kernel, the nucleolus, the e-core, and the bargaining set. As for the von Neumann-Morgenstern solution, we point out that its existence may not even be decidable. Many of these results generalize to the case in which the game is presented by a hypergraph with edges of size k > 2."
                },
                {
                    "title": "A Value for n-person Games",
                    "abstract": "Introduction At the foundation of the theory of games is the assumption that the players of a game can evaluate, in their utility scales, every \u201cprospect\u201d that might arise as a result of a play. In attempting to apply the theory to any field, one would normally expect to be permitted to include, in the class of \u201cprospects,\u201d the prospect of having to play a game. The possibility of evaluating games is therefore of critical importance. So long as the theory is unable to assign values to the games typically found in application, only relatively simple situations\u2014where games do not depend on other games\u2014will be susceptible to analysis and solution. In the finite theory of von Neumann and Morgenstern difficulty in evaluation persists for the \u201cessential\u201d games, and for only those. In this note we deduce a value for the \u201cessential\u201d case and examine a number of its elementary properties. We proceed from a set of three axioms, having simple intuitive interpretations, which suffice to determine the value uniquely. Our present work, though mathematically self-contained, is founded conceptually on the von Neumann\u2014Morgenstern theory up to their introduction of characteristic functions. We thereby inherit certain important underlying assumptions: (a) that utility is objective and transferable; (b) that games are cooperative affairs; (c) that games, granting (a) and (b), are adequately represented by their characteristic functions."
                },
                {
                    "title": "Value Theory Without Efficiency",
                    "abstract": "A semivalue is a symmetric positive linear operator on a space of games, which leaves the additive games fixed. Such an operator satisfies all of the axioms defining the Shapley value, with the possible exception of the efficiency axiom. The class of semivalues is completely characterized for the space of finite-player games, and for the space pNA of nonatomic games."
                },
                {
                    "title": "Alibi Explain: Algorithms for Explaining Machine Learning Models",
                    "abstract": "We introduce Alibi Explain , an open-source Python library for explaining predictions of machine learning models ( https://github.com/SeldonIO/alibi ). The library features state-of-the-art explainability algorithms for classi\ufb01cation and regression models. The algorithms cover both the model-agnostic (black-box) and model-speci\ufb01c (white-box) setting, cater for multiple data types (tabular, text, images) and explanation scope (local and global explanations). The library exposes a uni\ufb01ed API enabling users to work with explanations in a consistent way. Alibi adheres to best development practices featuring extensive testing of code correctness and algorithm convergence in a continuous integration environment. The library comes with extensive documentation of both usage and theoretical background of methods, and a suite of worked end-to-end use cases. Alibi aims to be a production-ready toolkit with integrations into machine learning deployment platforms such as Seldon Core and KFServing , and distributed explanation capabilities using Ray ."
                },
                {
                    "title": "Shapley Residuals: Quantifying the limits of the Shapley value for explanations",
                    "abstract": "Popular feature importance techniques compute additive approximations to nonlinear models by \ufb01rst de\ufb01ning a cooperative game describing the value of different subsets of the model\u2019s features, then calculating the resulting game\u2019s Shapley values to attribute credit additively between the features. However, the speci\ufb01c modeling settings in which the Shapley values are a poor approximation for the true game have not been well-described. In this paper we utilize an interpretation of Shapley values as the result of an orthogonal projection between vector spaces to calculate a residual representing the kernel component of that projection. We provide an algorithm for computing these residuals, characterize different modeling settings based on the value of the residuals, and demonstrate that they capture information about model predictions that Shapley values cannot. Shapley residuals can thus act as a warning to practitioners against overestimating the degree to which Shapley-value-based explanations give them insight into a model."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we efficiently compute Shapley Interactions (SIs) and M\u00f6bius Interactions (MIs) for complex machine learning models to better understand feature interactions and their contributions to predictions?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing the interpretability of machine learning models, particularly in understanding how different features interact to influence predictions. By providing a more nuanced view of feature contributions, this research can enhance model transparency, which is vital for trust in AI systems, especially in sensitive applications like healthcare and finance. Furthermore, it can lead to the development of more sophisticated algorithms that leverage feature interactions, potentially improving model performance and decision-making processes in various domains.\n\n**[Question 3] - Why is it hard?**  \nThe primary challenge lies in the exponential complexity of computing SIs and MIs without structural assumptions about the underlying game. Naive approaches may fail due to the combinatorial explosion of coalitions that need to be evaluated, making it computationally infeasible for large feature sets. Additionally, the theoretical underpinnings of cooperative game theory require a deep understanding of mathematical concepts, which can be a barrier for practitioners. Overcoming these technical and practical obstacles necessitates innovative algorithmic strategies and efficient computational techniques.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on the Shapley Value (SV) and its efficient computation for specific model types, often neglecting the complexities introduced by higher-order interactions. Existing solutions have been limited by their application scope, typically addressing isolated cases rather than providing a comprehensive framework for all possible interactions. Barriers such as the lack of scalable algorithms and the need for a strong mathematical foundation have hindered progress. Our approach aims to bridge these gaps by developing a unified methodology that can efficiently compute SIs and MIs across diverse machine learning models.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a novel algorithm that leverages stochastic approximators to compute SIs and MIs efficiently. We will utilize a diverse set of datasets representing various machine learning tasks to validate our approach. The performance will be evaluated using metrics such as computational efficiency and accuracy of interaction estimates compared to existing methods. We expect our results to demonstrate significant improvements in both the speed of computation and the quality of insights gained regarding feature interactions, ultimately contributing to enhanced interpretability of machine"
            }
        },
        "author_data": {
            "af825722-39a5-43c9-acc2-3ec7c7de6700": {
                "pk": "af825722-39a5-43c9-acc2-3ec7c7de6700",
                "name": "Maximilian Muschalik",
                "collaborators": [
                    "Eyke H\u00fcllermeier",
                    "Fabian Fumagalli",
                    "Barbara Hammer",
                    "Patrick Kolpaczki",
                    "Rohit Jagtani",
                    "Viktor Bengs"
                ],
                "domain": [
                    "Explainable AI",
                    "Feature Importance",
                    "Shapley Value",
                    "Incremental Learning"
                ],
                "publications": [
                    {
                        "title": "Incremental Permutation Feature Importance (iPFI): Towards Online Explanations on Data Streams",
                        "abstract": "Explainable Artificial Intelligence (XAI) has mainly focused on static learning scenarios so far. We are interested in dynamic scenarios where data is sampled progressively, and learning is done in an incremental rather than a batch mode. We seek efficient incremental algorithms for computing feature importance (FI) measures, specifically, an incremental FI measure based on feature marginalization of absent features similar to permutation feature importance (PFI). We propose an efficient, model-agnostic algorithm called iPFI to estimate this measure incrementally and under dynamic modeling conditions including concept drift. We prove theoretical guarantees on the approximation quality in terms of expectation and variance. To validate our theoretical findings and the efficacy of our approaches compared to traditional batch PFI, we conduct multiple experimental studies on benchmark data with and without concept drift."
                    },
                    {
                        "title": "iSAGE: An Incremental Version of SAGE for Online Explanation on Data Streams",
                        "abstract": "Existing methods for explainable artificial intelligence (XAI), including popular feature importance measures such as SAGE, are mostly restricted to the batch learning scenario. However, machine learning is often applied in dynamic environments, where data arrives continuously and learning must be done in an online manner. Therefore, we propose iSAGE, a time- and memory-efficient incrementalization of SAGE, which is able to react to changes in the model as well as to drift in the data-generating process. We further provide efficient feature removal methods that break (interventional) and retain (observational) feature dependencies. Moreover, we formally analyze our explanation method to show that iSAGE adheres to similar theoretical properties as SAGE. Finally, we evaluate our approach in a thorough experimental analysis based on well-established data sets and data streams with concept drift."
                    },
                    {
                        "title": "Beyond TreeSHAP: Efficient Computation of Any-Order Shapley Interactions for Tree Ensembles",
                        "abstract": "While shallow decision trees may be interpretable, larger ensemble models like gradient-boosted trees, which often set the state of the art in machine learning problems involving tabular data, still remain black box models. As a remedy, the Shapley value (SV) is a well-known concept in explainable artificial intelligence (XAI) research for quantifying additive feature attributions of predictions. The model-specific TreeSHAP methodology solves the exponential complexity for retrieving exact SVs from tree-based models. Expanding beyond individual feature attribution, Shapley interactions reveal the impact of intricate feature interactions of any order. In this work, we present TreeSHAP-IQ, an efficient method to compute any-order additive Shapley interactions for predictions of tree-based models. TreeSHAP-IQ is supported by a mathematical framework that exploits polynomial arithmetic to compute the interaction scores in a single recursive traversal of the tree, akin to Linear TreeSHAP. We apply TreeSHAP-IQ on state-of-the-art tree ensembles and explore interactions on well-established benchmark datasets."
                    },
                    {
                        "title": "SHAP-IQ: Unified Approximation of any-order Shapley Interactions",
                        "abstract": "Predominately in explainable artificial intelligence (XAI) research, the Shapley value (SV) is applied to determine feature attributions for any black box model. Shapley interaction indices extend the SV to define any-order feature interactions. Defining a unique Shapley interaction index is an open research question and, so far, three definitions have been proposed, which differ by their choice of axioms. Moreover, each definition requires a specific approximation technique. Here, we propose SHAPley Interaction Quantification (SHAP-IQ), an efficient sampling-based approximator to compute Shapley interactions for arbitrary cardinal interaction indices (CII), i.e. interaction indices that satisfy the linearity, symmetry and dummy axiom. SHAP-IQ is based on a novel representation and, in contrast to existing methods, we provide theoretical guarantees for its approximation quality, as well as estimates for the variance of the point estimates. For the special case of SV, our approach reveals a novel representation of the SV and corresponds to Unbiased KernelSHAP with a greatly simplified calculation. We illustrate the computational efficiency and effectiveness by explaining language, image classification and high-dimensional synthetic models."
                    },
                    {
                        "title": "iPDP: On Partial Dependence Plots in Dynamic Modeling Scenarios",
                        "abstract": "Post-hoc explanation techniques such as the well-established partial dependence plot (PDP), which investigates feature dependencies, are used in explainable artificial intelligence (XAI) to understand black-box machine learning models. While many real-world applications require dynamic models that constantly adapt over time and react to changes in the underlying distribution, XAI, so far, has primarily considered static learning environments, where models are trained in a batch mode and remain unchanged. We thus propose a novel model-agnostic XAI framework called incremental PDP (iPDP) that extends on the PDP to extract time-dependent feature effects in non-stationary learning environments. We formally analyze iPDP and show that it approximates a time-dependent variant of the PDP that properly reacts to real and virtual concept drift. The time-sensitivity of iPDP is controlled by a single smoothing parameter, which directly corresponds to the variance and the approximation error of iPDP in a static learning environment. We illustrate the efficacy of iPDP by showcasing an example application for drift detection and conducting multiple experiments on real-world and synthetic data sets and streams."
                    },
                    {
                        "title": "KernelSHAP-IQ: Weighted Least-Square Optimization for Shapley Interactions",
                        "abstract": "The Shapley value (SV) is a prevalent approach of allocating credit to machine learning (ML) entities to understand black box ML models. Enriching such interpretations with higher-order interactions is inevitable for complex systems, where the Shapley Interaction Index (SII) is a direct axiomatic extension of the SV. While it is well-known that the SV yields an optimal approximation of any game via a weighted least square (WLS) objective, an extension of this result to SII has been a long-standing open problem, which even led to the proposal of an alternative index. In this work, we characterize higher-order SII as a solution to a WLS problem, which constructs an optimal approximation via SII and $k$-Shapley values ($k$-SII). We prove this representation for the SV and pairwise SII and give empirically validated conjectures for higher orders. As a result, we propose KernelSHAP-IQ, a direct extension of KernelSHAP for SII, and demonstrate state-of-the-art performance for feature interactions."
                    },
                    {
                        "title": "Approximating the Shapley Value without Marginal Contributions",
                        "abstract": "The Shapley value, which is arguably the most popular approach for assigning a meaningful contribution value to players in a cooperative game, has recently been used intensively in explainable artificial intelligence. Its meaningfulness is due to axiomatic properties that only the Shapley value satisfies, which, however, comes at the expense of an exact computation growing exponentially with the number of agents. Accordingly, a number of works are devoted to the efficient approximation of the Shapley value, most of them revolve around the notion of an agent's marginal contribution. In this paper, we propose with SVARM and Stratified SVARM two parameter-free and domain-independent approximation algorithms based on a representation of the Shapley value detached from the notion of marginal contribution. We prove unmatched theoretical guarantees regarding their approximation quality and provide empirical results including synthetic games as well as common explainability use cases comparing ourselves with state-of-the-art methods."
                    },
                    {
                        "title": "SVARM-IQ: Efficient Approximation of Any-order Shapley Interactions through Stratification",
                        "abstract": "Addressing the limitations of individual attribution scores via the Shapley value (SV), the field of explainable AI (XAI) has recently explored intricate interactions of features or data points. In particular, extensions of the SV, such as the Shapley Interaction Index (SII), have been proposed as a measure to still benefit from the axiomatic basis of the SV. However, similar to the SV, their exact computation remains computationally prohibitive. Hence, we propose with SVARM-IQ a sampling-based approach to efficiently approximate Shapley-based interaction indices of any order. SVARM-IQ can be applied to a broad class of interaction indices, including the SII, by leveraging a novel stratified representation. We provide non-asymptotic theoretical guarantees on its approximation quality and empirically demonstrate that SVARM-IQ achieves state-of-the-art estimation results in practical XAI scenarios on different model classes and application domains."
                    }
                ]
            },
            "11403f6c-9f66-4a56-88aa-e9fcbc8af7d2": {
                "pk": "11403f6c-9f66-4a56-88aa-e9fcbc8af7d2",
                "name": "Hubert Baniecki",
                "collaborators": [
                    "Przemyslaw Biecek",
                    "Przemys\u0142aw Biecek",
                    "Wojciech Kretowicz",
                    "Mateusz Krzyzi\u0144ski",
                    "Miko\u0142aj Spytek",
                    "Andreas Holzinger",
                    "Bastian Pfeifer",
                    "Anna Saranti",
                    "Piotr Komorowski",
                    "Giuseppe Casalicchio"
                ],
                "domain": [
                    "Explainable AI",
                    "Adversarial Machine Learning",
                    "Survival Analysis",
                    "Interpretability"
                ],
                "publications": [
                    {
                        "title": "Adversarial attacks and defenses in explainable artificial intelligence: A survey",
                        "abstract": "Explainable artificial intelligence (XAI) methods are portrayed as a remedy for debugging and trusting statistical and deep learning models, as well as interpreting their predictions. However, recent advances in adversarial machine learning (AdvML) highlight the limitations and vulnerabilities of state-of-the-art explanation methods, putting their security and trustworthiness into question. The possibility of manipulating, fooling or fairwashing evidence of the model's reasoning has detrimental consequences when applied in high-stakes decision-making and knowledge discovery. This survey provides a comprehensive overview of research concerning adversarial attacks on explanations of machine learning models, as well as fairness metrics. We introduce a unified notation and taxonomy of methods facilitating a common ground for researchers and practitioners from the intersecting research fields of AdvML and XAI. We discuss how to defend against attacks and design robust interpretation methods. We contribute a list of existing insecurities in XAI and outline the emerging research directions in adversarial XAI (AdvXAI). Future work should address improving explanation methods and evaluation protocols to take into account the reported safety issues."
                    },
                    {
                        "title": "Fooling Partial Dependence via Data Poisoning",
                        "abstract": "Many methods have been developed to understand complex predictive models and high expectations are placed on post-hoc model explainability. It turns out that such explanations are not robust nor trustworthy, and they can be fooled. This paper presents techniques for attacking Partial Dependence (plots, profiles, PDP), which are among the most popular methods of explaining any predictive model trained on tabular data. We showcase that PD can be manipulated in an adversarial manner, which is alarming, especially in financial or medical applications where auditability became a must-have trait supporting black-box machine learning. The fooling is performed via poisoning the data to bend and shift explanations in the desired direction using genetic and gradient algorithms. We believe this to be the first work using a genetic algorithm for manipulating explanations, which is transferable as it generalizes both ways: in a model-agnostic and an explanation-agnostic manner."
                    },
                    {
                        "title": "Towards Evaluating Explanations of Vision Transformers for Medical Imaging",
                        "abstract": "As deep learning models increasingly find applications in critical domains such as medical imaging, the need for transparent and trustworthy decision-making becomes paramount. Many explainability methods provide insights into how these models make predictions by attributing importance to input features. As Vision Transformer (ViT) becomes a promising alternative to convolutional neural networks for image classification, its interpretability remains an open research question. This paper investigates the performance of various interpretation methods on a ViT applied to classify chest X-ray images. We introduce the notion of evaluating faithfulness, sensitivity, and complexity of ViT explanations. The obtained results indicate that Layerwise relevance propagation for transformers outperforms Local interpretable model-agnostic explanations and Attention visualization, providing a more accurate and reliable representation of what a ViT has actually learned. Our findings provide insights into the applicability of ViT explanations in medical imaging and highlight the importance of using appropriate evaluation criteria for comparing them."
                    },
                    {
                        "title": "On the Robustness of Global Feature Effect Explanations",
                        "abstract": "We study the robustness of global post-hoc explanations for predictive models trained on tabular data. Effects of predictor features in black-box supervised learning are an essential diagnostic tool for model debugging and scientific discovery in applied sciences. However, how vulnerable they are to data and model perturbations remains an open research question. We introduce several theoretical bounds for evaluating the robustness of partial dependence plots and accumulated local effects. Our experimental results with synthetic and real-world datasets quantify the gap between the best and worst-case scenarios of (mis)interpreting machine learning predictions globally."
                    },
                    {
                        "title": "Efficient and Accurate Explanation Estimation with Distribution Compression",
                        "abstract": "Exact computation of various machine learning explanations requires numerous model evaluations and in extreme cases becomes impractical. The computational cost of approximation increases with an ever-increasing size of data and model parameters. Many heuristics have been proposed to approximate post-hoc explanations efficiently. This paper shows that the standard i.i.d. sampling used in a broad spectrum of algorithms for explanation estimation leads to an approximation error worthy of improvement. To this end, we introduce Compress Then Explain (CTE), a new paradigm for more efficient and accurate explanation estimation. CTE uses distribution compression through kernel thinning to obtain a data sample that best approximates the marginal distribution. We show that CTE improves the estimation of removal-based local and global explanations with negligible computational overhead. It often achieves an on-par explanation approximation error using 2-3x less samples, i.e. requiring 2-3x less model evaluations. CTE is a simple, yet powerful, plug-in for any explanation method that now relies on i.i.d. sampling."
                    },
                    {
                        "title": "The Grammar of Interactive Explanatory Model Analysis",
                        "abstract": "The growing need for in-depth analysis of predictive models leads to a series of new methods for explaining their local and global properties. Which of these methods is the best? It turns out that this is an ill-posed question. One cannot sufficiently explain a black-box machine learning model using a single method that gives only one perspective. Isolated explanations are prone to misunderstanding, leading to wrong or simplistic reasoning. This problem is known as the Rashomon effect and refers to diverse, even contradictory, interpretations of the same phenomenon. Surprisingly, most methods developed for explainable and responsible machine learning focus on a single-aspect of the model behavior. In contrast, we showcase the problem of explainability as an interactive and sequential analysis of a model. This paper proposes how different Explanatory Model Analysis (EMA) methods complement each other and discusses why it is essential to juxtapose them. The introduced process of Interactive EMA (IEMA) derives from the algorithmic side of explainable machine learning and aims to embrace ideas developed in cognitive sciences. We formalize the grammar of IEMA to describe potential human-model dialogues. It is implemented in a widely used human-centered open-source software framework that adopts interactivity, customizability and automation as its main traits. We conduct a user study to evaluate the usefulness of IEMA, which indicates that an interactive sequential analysis of a model increases the performance and confidence of human decision making."
                    },
                    {
                        "title": "Interpretable machine learning for time-to-event prediction in medicine and healthcare",
                        "abstract": "Time-to-event prediction, e.g. cancer survival analysis or hospital length of stay, is a highly prominent machine learning task in medical and healthcare applications. However, only a few interpretable machine learning methods comply with its challenges. To facilitate a comprehensive explanatory analysis of survival models, we formally introduce time-dependent feature effects and global feature importance explanations. We show how post-hoc interpretation methods allow for finding biases in AI systems predicting length of stay using a novel multi-modal dataset created from 1235 X-ray images with textual radiology reports annotated by human experts. Moreover, we evaluate cancer survival models beyond predictive performance to include the importance of multi-omics feature groups based on a large-scale benchmark comprising 11 datasets from The Cancer Genome Atlas (TCGA). Model developers can use the proposed methods to debug and improve machine learning algorithms, while physicians can discover disease biomarkers and assess their significance. We hope the contributed open data and code resources facilitate future work in the emerging research direction of explainable survival analysis."
                    },
                    {
                        "title": "SurvSHAP(t): Time-dependent explanations of machine learning survival models",
                        "abstract": "Machine and deep learning survival models demonstrate similar or even improved time-to-event prediction capabilities compared to classical statistical learning methods yet are too complex to be interpreted by humans. Several model-agnostic explanations are available to overcome this issue; however, none directly explain the survival function prediction. In this paper, we introduce SurvSHAP(t), the first time-dependent explanation that allows for interpreting survival black-box models. It is based on SHapley Additive exPlanations with solid theoretical foundations and a broad adoption among machine learning practitioners. The proposed methods aim to enhance precision diagnostics and support domain experts in making decisions. Experiments on synthetic and medical data confirm that SurvSHAP(t) can detect variables with a time-dependent effect, and its aggregation is a better determinant of the importance of variables for a prediction than SurvLIME. SurvSHAP(t) is model-agnostic and can be applied to all models with functional output. We provide an accessible implementation of time-dependent explanations in Python at http://github.com/MI2DataLab/survshap."
                    },
                    {
                        "title": "Performance is not enough: the story told by a Rashomon quartet",
                        "abstract": "The usual goal of supervised learning is to find the best model, the one that optimizes a particular performance measure. However, what if the explanation provided by this model is completely different from another model and different again from another model despite all having similarly good fit statistics? Is it possible that the equally effective models put the spotlight on different relationships in the data? Inspired by Anscombe's quartet, this paper introduces a Rashomon Quartet, i.e. a set of four models built on a synthetic dataset which have practically identical predictive performance. However, the visual exploration reveals distinct explanations of the relations in the data. This illustrative example aims to encourage the use of methods for model visualization to compare predictive models beyond their performance."
                    },
                    {
                        "title": "Be Careful When Evaluating Explanations Regarding Ground Truth",
                        "abstract": "Evaluating explanations of image classifiers regarding ground truth, e.g. segmentation masks defined by human perception, primarily evaluates the quality of the models under consideration rather than the explanation methods themselves. Driven by this observation, we propose a framework for $\\textit{jointly}$ evaluating the robustness of safety-critical systems that $\\textit{combine}$ a deep neural network with an explanation method. These are increasingly used in real-world applications like medical image analysis or robotics. We introduce a fine-tuning procedure to (mis)align model$\\unicode{x2013}$explanation pipelines with ground truth and use it to quantify the potential discrepancy between worst and best-case scenarios of human alignment. Experiments across various model architectures and post-hoc local interpretation methods provide insights into the robustness of vision transformers and the overall vulnerability of such AI systems to potential adversarial attacks."
                    },
                    {
                        "title": "dalex: Responsible Machine Learning with Interactive Explainability and Fairness in Python",
                        "abstract": "The increasing amount of available data, computing power, and the constant pursuit for higher performance results in the growing complexity of predictive models. Their black-box nature leads to opaqueness debt phenomenon inflicting increased risks of discrimination, lack of reproducibility, and deflated performance due to data drift. To manage these risks, good MLOps practices ask for better validation of model performance and fairness, higher explainability, and continuous monitoring. The necessity of deeper model transparency appears not only from scientific and social domains, but also emerging laws and regulations on artificial intelligence. To facilitate the development of responsible machine learning models, we showcase dalex, a Python package which implements the model-agnostic interface for interactive model exploration. It adopts the design crafted through the development of various tools for responsible machine learning; thus, it aims at the unification of the existing solutions. This library's source code and documentation are available under open license at https://python.drwhy.ai/."
                    },
                    {
                        "title": "Do not explain without context: addressing the blind spot of model explanations",
                        "abstract": "The increasing number of regulations and expectations of predictive machine learning models, such as so called right to explanation, has led to a large number of methods promising greater interpretability. High demand has led to a widespread adoption of XAI techniques like Shapley values, Partial Dependence profiles or permutational variable importance. However, we still do not know enough about their properties and how they manifest in the context in which explanations are created by analysts, reviewed by auditors, and interpreted by various stakeholders. This paper highlights a blind spot which, although critical, is often overlooked when monitoring and auditing machine learning models: the effect of the reference data on the explanation calculation. We discuss that many model explanations depend directly or indirectly on the choice of the referenced data distribution. We showcase examples where small changes in the distribution lead to drastic changes in the explanations, such as a change in trend or, alarmingly, a conclusion. Consequently, we postulate that obtaining robust and useful explanations always requires supporting them with a broader context."
                    },
                    {
                        "title": "Explaining and visualizing black-box models through counterfactual paths",
                        "abstract": "Explainable AI (XAI) is an increasingly important area of machine learning research, which aims to make black-box models transparent and interpretable. In this paper, we propose a novel approach to XAI that uses the so-called counterfactual paths generated by conditional permutations of features. The algorithm measures feature importance by identifying sequential permutations of features that most influence changes in model predictions. It is particularly suitable for generating explanations based on counterfactual paths in knowledge graphs incorporating domain knowledge. Counterfactual paths introduce an additional graph dimension to current XAI methods in both explaining and visualizing black-box models. Experiments with synthetic and medical data demonstrate the practical applicability of our approach."
                    },
                    {
                        "title": "Aggregated Attributions for Explanatory Analysis of 3D Segmentation Models",
                        "abstract": "Analysis of 3D segmentation models, especially in the context of medical imaging, is often limited to segmentation performance metrics that overlook the crucial aspect of explainability and bias. Currently, effectively explaining these models with saliency maps is challenging due to the high dimensions of input images multiplied by the ever-growing number of segmented class labels. To this end, we introduce Agg^2Exp, a methodology for aggregating fine-grained voxel attributions of the segmentation model's predictions. Unlike classical explanation methods that primarily focus on the local feature attribution, Agg^2Exp enables a more comprehensive global view on the importance of predicted segments in 3D images. Our benchmarking experiments show that gradient-based voxel attributions are more faithful to the model's predictions than perturbation-based explanations. As a concrete use-case, we apply Agg^2Exp to discover knowledge acquired by the Swin UNEt TRansformer model trained on the TotalSegmentator v2 dataset for segmenting anatomical structures in computed tomography medical images. Agg^2Exp facilitates the explanatory analysis of large segmentation models beyond their predictive performance."
                    },
                    {
                        "title": "Graph-guided random forest for gene set selection",
                        "abstract": "Machine learning methods can detect complex relationships between variables, but usually do not exploit domain knowledge. This is a limitation because in many scientific disciplines, such as systems biology, domain knowledge is available in the form of graphs or networks, and its use can improve model performance. We need network-based algorithms that are versatile and applicable in many research areas. In this work, we demonstrate subnetwork detection based on multi-modal node features using a novel Greedy Decision Forest with inherent interpretability. The latter will be a crucial factor to retain experts and gain their trust in such algorithms. To demonstrate a concrete application example, we focus on bioinformatics, systems biology and particularly biomedicine, but the presented methodology is applicable in many other domains as well. Systems biology is a good example of a field in which statistical data-driven machine learning enables the analysis of large amounts of multi-modal biomedical data. This is important to reach the future goal of precision medicine, where the complexity of patients is modeled on a system level to best tailor medical decisions, health practices and therapies to the individual patient. Our proposed approach can help to uncover disease-causing network modules from multi-omics data to better understand complex diseases such as cancer."
                    },
                    {
                        "title": "Red-Teaming Segment Anything Model",
                        "abstract": "Foundation models have emerged as pivotal tools, tackling many complex tasks through pre-training on vast datasets and subsequent fine-tuning for specific applications. The Segment Anything Model is one of the first and most well-known foundation models for computer vision segmentation tasks. This work presents a multi-faceted red-teaming analysis that tests the Segment Anything Model against challenging tasks: (1) We analyze the impact of style transfer on segmentation masks, demonstrating that applying adverse weather conditions and raindrops to dashboard images of city roads significantly distorts generated masks. (2) We focus on assessing whether the model can be used for attacks on privacy, such as recognizing celebrities' faces, and show that the model possesses some undesired knowledge in this task. (3) Finally, we check how robust the model is to adversarial attacks on segmentation masks under text prompts. We not only show the effectiveness of popular white-box attacks and resistance to black-box attacks but also introduce a novel approach - Focused Iterative Gradient Attack (FIGA) that combines white-box approaches to construct an efficient attack resulting in a smaller number of modified pixels. All of our testing methods and analyses indicate a need for enhanced safety measures in foundation models for image segmentation."
                    },
                    {
                        "title": "survex: an R package for explaining machine learning survival models",
                        "abstract": "Due to their flexibility and superior performance, machine learning models frequently complement and outperform traditional statistical survival models. However, their widespread adoption is hindered by a lack of user-friendly tools to explain their internal operations and prediction rationales. To tackle this issue, we introduce the survex R package, which provides a cohesive framework for explaining any survival model by applying explainable artificial intelligence techniques. The capabilities of the proposed software encompass understanding and diagnosing survival models, which can lead to their improvement. By revealing insights into the decision-making process, such as variable effects and importances, survex enables the assessment of model reliability and the detection of biases. Thus, transparency and responsibility may be promoted in sensitive areas, such as biomedical research and healthcare applications."
                    },
                    {
                        "title": "Red Teaming Models for Hyperspectral Image Analysis Using Explainable AI",
                        "abstract": "Remote sensing (RS) applications in the space domain demand machine learning (ML) models that are reliable, robust, and quality-assured, making red teaming a vital approach for identifying and exposing potential flaws and biases. Since both fields advance independently, there is a notable gap in integrating red teaming strategies into RS. This paper introduces a methodology for examining ML models operating on hyperspectral images within the HYPERVIEW challenge, focusing on soil parameters' estimation. We use post-hoc explanation methods from the Explainable AI (XAI) domain to critically assess the best performing model that won the HYPERVIEW challenge and served as an inspiration for the model deployed on board the INTUITION-1 hyperspectral mission. Our approach effectively red teams the model by pinpointing and validating key shortcomings, constructing a model that achieves comparable performance using just 1% of the input features and a mere up to 5% performance loss. Additionally, we propose a novel way of visualizing explanations that integrate domain-specific information about hyperspectral bands (wavelengths) and data transformations to better suit interpreting models for hyperspectral image analysis."
                    },
                    {
                        "title": "Interpretable Machine Learning for Survival Analysis",
                        "abstract": "With the spread and rapid advancement of black box machine learning models, the field of interpretable machine learning (IML) or explainable artificial intelligence (XAI) has become increasingly important over the last decade. This is particularly relevant for survival analysis, where the adoption of IML techniques promotes transparency, accountability and fairness in sensitive areas, such as clinical decision making processes, the development of targeted therapies, interventions or in other medical or healthcare related contexts. More specifically, explainability can uncover a survival model's potential biases and limitations and provide more mathematically sound ways to understand how and which features are influential for prediction or constitute risk factors. However, the lack of readily available IML methods may have deterred medical practitioners and policy makers in public health from leveraging the full potential of machine learning for predicting time-to-event data. We present a comprehensive review of the limited existing amount of work on IML methods for survival analysis within the context of the general IML taxonomy. In addition, we formally detail how commonly used IML methods, such as such as individual conditional expectation (ICE), partial dependence plots (PDP), accumulated local effects (ALE), different feature importance measures or Friedman's H-interaction statistics can be adapted to survival outcomes. An application of several IML methods to real data on data on under-5 year mortality of Ghanaian children from the Demographic and Health Surveys (DHS) Program serves as a tutorial or guide for researchers, on how to utilize the techniques in practice to facilitate understanding of model decisions or predictions."
                    }
                ]
            },
            "447304ca-d381-4a49-ad52-41ce85c61e2e": {
                "pk": "447304ca-d381-4a49-ad52-41ce85c61e2e",
                "name": "Fabian Fumagalli",
                "collaborators": [
                    "Barbara Hammer",
                    "Maximilian Muschalik",
                    "Eyke H\u00fcllermeier",
                    "Patrick Kolpaczki",
                    "Philip Kenneweg",
                    "Tristan Kenneweg",
                    "Rohit Jagtani"
                ],
                "domain": [
                    "Explainable AI",
                    "Feature Importance",
                    "Incremental Learning",
                    "Shapley Value"
                ],
                "publications": [
                    {
                        "title": "Incremental Permutation Feature Importance (iPFI): Towards Online Explanations on Data Streams",
                        "abstract": "Explainable Artificial Intelligence (XAI) has mainly focused on static learning scenarios so far. We are interested in dynamic scenarios where data is sampled progressively, and learning is done in an incremental rather than a batch mode. We seek efficient incremental algorithms for computing feature importance (FI) measures, specifically, an incremental FI measure based on feature marginalization of absent features similar to permutation feature importance (PFI). We propose an efficient, model-agnostic algorithm called iPFI to estimate this measure incrementally and under dynamic modeling conditions including concept drift. We prove theoretical guarantees on the approximation quality in terms of expectation and variance. To validate our theoretical findings and the efficacy of our approaches compared to traditional batch PFI, we conduct multiple experimental studies on benchmark data with and without concept drift."
                    },
                    {
                        "title": "iSAGE: An Incremental Version of SAGE for Online Explanation on Data Streams",
                        "abstract": "Existing methods for explainable artificial intelligence (XAI), including popular feature importance measures such as SAGE, are mostly restricted to the batch learning scenario. However, machine learning is often applied in dynamic environments, where data arrives continuously and learning must be done in an online manner. Therefore, we propose iSAGE, a time- and memory-efficient incrementalization of SAGE, which is able to react to changes in the model as well as to drift in the data-generating process. We further provide efficient feature removal methods that break (interventional) and retain (observational) feature dependencies. Moreover, we formally analyze our explanation method to show that iSAGE adheres to similar theoretical properties as SAGE. Finally, we evaluate our approach in a thorough experimental analysis based on well-established data sets and data streams with concept drift."
                    },
                    {
                        "title": "Beyond TreeSHAP: Efficient Computation of Any-Order Shapley Interactions for Tree Ensembles",
                        "abstract": "While shallow decision trees may be interpretable, larger ensemble models like gradient-boosted trees, which often set the state of the art in machine learning problems involving tabular data, still remain black box models. As a remedy, the Shapley value (SV) is a well-known concept in explainable artificial intelligence (XAI) research for quantifying additive feature attributions of predictions. The model-specific TreeSHAP methodology solves the exponential complexity for retrieving exact SVs from tree-based models. Expanding beyond individual feature attribution, Shapley interactions reveal the impact of intricate feature interactions of any order. In this work, we present TreeSHAP-IQ, an efficient method to compute any-order additive Shapley interactions for predictions of tree-based models. TreeSHAP-IQ is supported by a mathematical framework that exploits polynomial arithmetic to compute the interaction scores in a single recursive traversal of the tree, akin to Linear TreeSHAP. We apply TreeSHAP-IQ on state-of-the-art tree ensembles and explore interactions on well-established benchmark datasets."
                    },
                    {
                        "title": "No learning rates needed: Introducing SALSA -- Stable Armijo Line Search Adaptation",
                        "abstract": "In recent studies, line search methods have been demonstrated to significantly enhance the performance of conventional stochastic gradient descent techniques across various datasets and architectures, while making an otherwise critical choice of learning rate schedule superfluous. In this paper, we identify problems of current state-of-the-art of line search methods, propose enhancements, and rigorously assess their effectiveness. Furthermore, we evaluate these methods on orders of magnitude larger datasets and more complex data domains than previously done. More specifically, we enhance the Armijo line search method by speeding up its computation and incorporating a momentum term into the Armijo criterion, making it better suited for stochastic mini-batching. Our optimization approach outperforms both the previous Armijo implementation and a tuned learning rate schedule for the Adam and SGD optimizers. Our evaluation covers a diverse range of architectures, such as Transformers, CNNs, and MLPs, as well as data domains, including NLP and image data.   Our work is publicly available as a Python package, which provides a simple Pytorch optimizer."
                    },
                    {
                        "title": "SHAP-IQ: Unified Approximation of any-order Shapley Interactions",
                        "abstract": "Predominately in explainable artificial intelligence (XAI) research, the Shapley value (SV) is applied to determine feature attributions for any black box model. Shapley interaction indices extend the SV to define any-order feature interactions. Defining a unique Shapley interaction index is an open research question and, so far, three definitions have been proposed, which differ by their choice of axioms. Moreover, each definition requires a specific approximation technique. Here, we propose SHAPley Interaction Quantification (SHAP-IQ), an efficient sampling-based approximator to compute Shapley interactions for arbitrary cardinal interaction indices (CII), i.e. interaction indices that satisfy the linearity, symmetry and dummy axiom. SHAP-IQ is based on a novel representation and, in contrast to existing methods, we provide theoretical guarantees for its approximation quality, as well as estimates for the variance of the point estimates. For the special case of SV, our approach reveals a novel representation of the SV and corresponds to Unbiased KernelSHAP with a greatly simplified calculation. We illustrate the computational efficiency and effectiveness by explaining language, image classification and high-dimensional synthetic models."
                    },
                    {
                        "title": "iPDP: On Partial Dependence Plots in Dynamic Modeling Scenarios",
                        "abstract": "Post-hoc explanation techniques such as the well-established partial dependence plot (PDP), which investigates feature dependencies, are used in explainable artificial intelligence (XAI) to understand black-box machine learning models. While many real-world applications require dynamic models that constantly adapt over time and react to changes in the underlying distribution, XAI, so far, has primarily considered static learning environments, where models are trained in a batch mode and remain unchanged. We thus propose a novel model-agnostic XAI framework called incremental PDP (iPDP) that extends on the PDP to extract time-dependent feature effects in non-stationary learning environments. We formally analyze iPDP and show that it approximates a time-dependent variant of the PDP that properly reacts to real and virtual concept drift. The time-sensitivity of iPDP is controlled by a single smoothing parameter, which directly corresponds to the variance and the approximation error of iPDP in a static learning environment. We illustrate the efficacy of iPDP by showcasing an example application for drift detection and conducting multiple experiments on real-world and synthetic data sets and streams."
                    },
                    {
                        "title": "SVARM-IQ: Efficient Approximation of Any-order Shapley Interactions through Stratification",
                        "abstract": "Addressing the limitations of individual attribution scores via the Shapley value (SV), the field of explainable AI (XAI) has recently explored intricate interactions of features or data points. In particular, extensions of the SV, such as the Shapley Interaction Index (SII), have been proposed as a measure to still benefit from the axiomatic basis of the SV. However, similar to the SV, their exact computation remains computationally prohibitive. Hence, we propose with SVARM-IQ a sampling-based approach to efficiently approximate Shapley-based interaction indices of any order. SVARM-IQ can be applied to a broad class of interaction indices, including the SII, by leveraging a novel stratified representation. We provide non-asymptotic theoretical guarantees on its approximation quality and empirically demonstrate that SVARM-IQ achieves state-of-the-art estimation results in practical XAI scenarios on different model classes and application domains."
                    },
                    {
                        "title": "KernelSHAP-IQ: Weighted Least-Square Optimization for Shapley Interactions",
                        "abstract": "The Shapley value (SV) is a prevalent approach of allocating credit to machine learning (ML) entities to understand black box ML models. Enriching such interpretations with higher-order interactions is inevitable for complex systems, where the Shapley Interaction Index (SII) is a direct axiomatic extension of the SV. While it is well-known that the SV yields an optimal approximation of any game via a weighted least square (WLS) objective, an extension of this result to SII has been a long-standing open problem, which even led to the proposal of an alternative index. In this work, we characterize higher-order SII as a solution to a WLS problem, which constructs an optimal approximation via SII and $k$-Shapley values ($k$-SII). We prove this representation for the SV and pairwise SII and give empirically validated conjectures for higher orders. As a result, we propose KernelSHAP-IQ, a direct extension of KernelSHAP for SII, and demonstrate state-of-the-art performance for feature interactions."
                    }
                ]
            },
            "d83cf1c3-31de-40a0-8760-cf60c020a7d6": {
                "pk": "d83cf1c3-31de-40a0-8760-cf60c020a7d6",
                "name": "Patrick Kolpaczki",
                "collaborators": [
                    "Eyke H\u00fcllermeier",
                    "Maximilian Muschalik",
                    "Fabian Fumagalli",
                    "Barbara Hammer",
                    "Viktor Bengs"
                ],
                "domain": [
                    "Explainable AI",
                    "Shapley Value",
                    "Game Theory",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Approximating the Shapley Value without Marginal Contributions",
                        "abstract": "The Shapley value, which is arguably the most popular approach for assigning a meaningful contribution value to players in a cooperative game, has recently been used intensively in explainable artificial intelligence. Its meaningfulness is due to axiomatic properties that only the Shapley value satisfies, which, however, comes at the expense of an exact computation growing exponentially with the number of agents. Accordingly, a number of works are devoted to the efficient approximation of the Shapley value, most of them revolve around the notion of an agent's marginal contribution. In this paper, we propose with SVARM and Stratified SVARM two parameter-free and domain-independent approximation algorithms based on a representation of the Shapley value detached from the notion of marginal contribution. We prove unmatched theoretical guarantees regarding their approximation quality and provide empirical results including synthetic games as well as common explainability use cases comparing ourselves with state-of-the-art methods."
                    },
                    {
                        "title": "SVARM-IQ: Efficient Approximation of Any-order Shapley Interactions through Stratification",
                        "abstract": "Addressing the limitations of individual attribution scores via the Shapley value (SV), the field of explainable AI (XAI) has recently explored intricate interactions of features or data points. In particular, extensions of the SV, such as the Shapley Interaction Index (SII), have been proposed as a measure to still benefit from the axiomatic basis of the SV. However, similar to the SV, their exact computation remains computationally prohibitive. Hence, we propose with SVARM-IQ a sampling-based approach to efficiently approximate Shapley-based interaction indices of any order. SVARM-IQ can be applied to a broad class of interaction indices, including the SII, by leveraging a novel stratified representation. We provide non-asymptotic theoretical guarantees on its approximation quality and empirically demonstrate that SVARM-IQ achieves state-of-the-art estimation results in practical XAI scenarios on different model classes and application domains."
                    },
                    {
                        "title": "Non-Stationary Dueling Bandits",
                        "abstract": "We study the non-stationary dueling bandits problem with $K$ arms, where the time horizon $T$ consists of $M$ stationary segments, each of which is associated with its own preference matrix. The learner repeatedly selects a pair of arms and observes a binary preference between them as feedback. To minimize the accumulated regret, the learner needs to pick the Condorcet winner of each stationary segment as often as possible, despite preference matrices and segment lengths being unknown. We propose the $\\mathrm{Beat\\, the\\, Winner\\, Reset}$ algorithm and prove a bound on its expected binary weak regret in the stationary case, which tightens the bound of current state-of-art algorithms. We also show a regret bound for the non-stationary case, without requiring knowledge of $M$ or $T$. We further propose and analyze two meta-algorithms, $\\mathrm{DETECT}$ for weak regret and $\\mathrm{Monitored\\, Dueling\\, Bandits}$ for strong regret, both based on a detection-window approach that can incorporate any dueling bandit algorithm as a black-box algorithm. Finally, we prove a worst-case lower bound for expected weak regret in the non-stationary case."
                    },
                    {
                        "title": "SHAP-IQ: Unified Approximation of any-order Shapley Interactions",
                        "abstract": "Predominately in explainable artificial intelligence (XAI) research, the Shapley value (SV) is applied to determine feature attributions for any black box model. Shapley interaction indices extend the SV to define any-order feature interactions. Defining a unique Shapley interaction index is an open research question and, so far, three definitions have been proposed, which differ by their choice of axioms. Moreover, each definition requires a specific approximation technique. Here, we propose SHAPley Interaction Quantification (SHAP-IQ), an efficient sampling-based approximator to compute Shapley interactions for arbitrary cardinal interaction indices (CII), i.e. interaction indices that satisfy the linearity, symmetry and dummy axiom. SHAP-IQ is based on a novel representation and, in contrast to existing methods, we provide theoretical guarantees for its approximation quality, as well as estimates for the variance of the point estimates. For the special case of SV, our approach reveals a novel representation of the SV and corresponds to Unbiased KernelSHAP with a greatly simplified calculation. We illustrate the computational efficiency and effectiveness by explaining language, image classification and high-dimensional synthetic models."
                    },
                    {
                        "title": "KernelSHAP-IQ: Weighted Least-Square Optimization for Shapley Interactions",
                        "abstract": "The Shapley value (SV) is a prevalent approach of allocating credit to machine learning (ML) entities to understand black box ML models. Enriching such interpretations with higher-order interactions is inevitable for complex systems, where the Shapley Interaction Index (SII) is a direct axiomatic extension of the SV. While it is well-known that the SV yields an optimal approximation of any game via a weighted least square (WLS) objective, an extension of this result to SII has been a long-standing open problem, which even led to the proposal of an alternative index. In this work, we characterize higher-order SII as a solution to a WLS problem, which constructs an optimal approximation via SII and $k$-Shapley values ($k$-SII). We prove this representation for the SV and pairwise SII and give empirically validated conjectures for higher orders. As a result, we propose KernelSHAP-IQ, a direct extension of KernelSHAP for SII, and demonstrate state-of-the-art performance for feature interactions."
                    }
                ]
            },
            "f9a0e946-1213-4239-aee4-9723f1c5b597": {
                "pk": "f9a0e946-1213-4239-aee4-9723f1c5b597",
                "name": "Barbara Hammer",
                "collaborators": [
                    "Andr\u00e9 Artelt",
                    "Babak Hosseini",
                    "Lukas Pfannschmidt",
                    "Alexander Hasenfu\u00df",
                    "Fabrice Rossi",
                    "Fabian Hinder",
                    "Daniel Wiens",
                    "Janine Strotherm"
                ],
                "domain": [
                    "Explainable AI",
                    "Counterfactual Explanations",
                    "Machine Learning",
                    "Fairness in AI"
                ],
                "publications": [
                    {
                        "title": "Efficient computation of counterfactual explanations of LVQ models",
                        "abstract": "The increasing use of machine learning in practice and legal regulations like EU's GDPR cause the necessity to be able to explain the prediction and behavior of machine learning models. A prominent example of particularly intuitive explanations of AI models in the context of decision making are counterfactual explanations. Yet, it is still an open research problem how to efficiently compute counterfactual explanations for many models.   We investigate how to efficiently compute counterfactual explanations for an important class of models, prototype-based classifiers such as learning vector quantization models. In particular, we derive specific convex and non-convex programs depending on the used metric."
                    },
                    {
                        "title": "On the computation of counterfactual explanations -- A survey",
                        "abstract": "Due to the increasing use of machine learning in practice it becomes more and more important to be able to explain the prediction and behavior of machine learning models. An instance of explanations are counterfactual explanations which provide an intuitive and useful explanations of machine learning models. In this survey we review model-specific methods for efficiently computing counterfactual explanations of many different machine learning models and propose methods for models that have not been considered in literature so far."
                    },
                    {
                        "title": "Sequential Feature Classification in the Context of Redundancies",
                        "abstract": "The problem of all-relevant feature selection is concerned with finding a relevant feature set with preserved redundancies. There exist several approximations to solve this problem but only one could give a distinction between strong and weak relevance. This approach was limited to the case of linear problems. In this work, we present a new solution for this distinction in the non-linear case through the use of random forest models and statistical methods."
                    },
                    {
                        "title": "Convex optimization for actionable \\& plausible counterfactual explanations",
                        "abstract": "Transparency is an essential requirement of machine learning based decision making systems that are deployed in real world. Often, transparency of a given system is achieved by providing explanations of the behavior and predictions of the given system. Counterfactual explanations are a prominent instance of particular intuitive explanations of decision making systems. While a lot of different methods for computing counterfactual explanations exist, only very few work (apart from work from the causality domain) considers feature dependencies as well as plausibility which might limit the set of possible counterfactual explanations.   In this work we enhance our previous work on convex modeling for computing counterfactual explanations by a mechanism for ensuring actionability and plausibility of the resulting counterfactual explanations."
                    },
                    {
                        "title": "\"Explain it in the Same Way!\" -- Model-Agnostic Group Fairness of Counterfactual Explanations",
                        "abstract": "Counterfactual explanations are a popular type of explanation for making the outcomes of a decision making system transparent to the user. Counterfactual explanations tell the user what to do in order to change the outcome of the system in a desirable way. However, it was recently discovered that the recommendations of what to do can differ significantly in their complexity between protected groups of individuals. Providing more difficult recommendations of actions to one group leads to a disadvantage of this group compared to other groups.   In this work we propose a model-agnostic method for computing counterfactual explanations that do not differ significantly in their complexity between protected groups."
                    },
                    {
                        "title": "Median topographic maps for biomedical data sets",
                        "abstract": "Median clustering extends popular neural data analysis methods such as the self-organizing map or neural gas to general data structures given by a dissimilarity matrix only. This offers flexible and robust global data inspection methods which are particularly suited for a variety of data as occurs in biomedical domains. In this chapter, we give an overview about median clustering and its properties and extensions, with a particular focus on efficient implementations adapted to large scale data analysis."
                    },
                    {
                        "title": "Efficient Metric Learning for the Analysis of Motion Data",
                        "abstract": "We investigate metric learning in the context of dynamic time warping (DTW), the by far most popular dissimilarity measure used for the comparison and analysis of motion capture data. While metric learning enables a problem-adapted representation of data, the majority of methods has been proposed for vectorial data only. In this contribution, we extend the popular principle offered by the large margin nearest neighbors learner (LMNN) to DTW by treating the resulting component-wise dissimilarity values as features. We demonstrate that this principle greatly enhances the classification accuracy in several benchmarks. Further, we show that recent auxiliary concepts such as metric regularization can be transferred from the vectorial case to component-wise DTW in a similar way. We illustrate that metric regularization constitutes a crucial prerequisite for the interpretation of the resulting relevance profiles."
                    },
                    {
                        "title": "Feasibility Based Large Margin Nearest Neighbor Metric Learning",
                        "abstract": "Large margin nearest neighbor (LMNN) is a metric learner which optimizes the performance of the popular $k$NN classifier. However, its resulting metric relies on pre-selected target neighbors. In this paper, we address the feasibility of LMNN's optimization constraints regarding these target points, and introduce a mathematical measure to evaluate the size of the feasible region of the optimization problem. We enhance the optimization framework of LMNN by a weighting scheme which prefers data triplets which yield a larger feasible region. This increases the chances to obtain a good metric as the solution of LMNN's problem. We evaluate the performance of the resulting feasibility-based LMNN algorithm using synthetic and real datasets. The empirical results show an improved accuracy for different types of datasets in comparison to regular LMNN."
                    },
                    {
                        "title": "Multiple-Kernel Dictionary Learning for Reconstruction and Clustering of Unseen Multivariate Time-series",
                        "abstract": "There exist many approaches for description and recognition of unseen classes in datasets. Nevertheless, it becomes a challenging problem when we deal with multivariate time-series (MTS) (e.g., motion data), where we cannot apply the vectorial algorithms directly to the inputs. In this work, we propose a novel multiple-kernel dictionary learning (MKD) which learns semantic attributes based on specific combinations of MTS dimensions in the feature space. Hence, MKD can fully/partially reconstructs the unseen classes based on the training data (seen classes). Furthermore, we obtain sparse encodings for unseen classes based on the learned MKD attributes, and upon which we propose a simple but effective incremental clustering algorithm to categorize the unseen MTS classes in an unsupervised way. According to the empirical evaluation of our MKD framework on real benchmarks, it provides an interpretable reconstruction of unseen MTS data as well as a high performance regarding their online clustering."
                    },
                    {
                        "title": "Confident Kernel Sparse Coding and Dictionary Learning",
                        "abstract": "In recent years, kernel-based sparse coding (K-SRC) has received particular attention due to its efficient representation of nonlinear data structures in the feature space. Nevertheless, the existing K-SRC methods suffer from the lack of consistency between their training and test optimization frameworks. In this work, we propose a novel confident K-SRC and dictionary learning algorithm (CKSC) which focuses on the discriminative reconstruction of the data based on its representation in the kernel space. CKSC focuses on reconstructing each data sample via weighted contributions which are confident in its corresponding class of data. We employ novel discriminative terms to apply this scheme to both training and test frameworks in our algorithm. This specific design increases the consistency of these optimization frameworks and improves the discriminative performance in the recall phase. In addition, CKSC directly employs the supervised information in its dictionary learning framework to enhance the discriminative structure of the dictionary. For empirical evaluations, we implement our CKSC algorithm on multivariate time-series benchmarks such as DynTex++ and UTKinect. Our claims regarding the superior performance of the proposed algorithm are justified throughout comparing its classification results to the state-of-the-art K-SRC algorithms."
                    },
                    {
                        "title": "Convex Density Constraints for Computing Plausible Counterfactual Explanations",
                        "abstract": "The increasing deployment of machine learning as well as legal regulations such as EU's GDPR cause a need for user-friendly explanations of decisions proposed by machine learning models. Counterfactual explanations are considered as one of the most popular techniques to explain a specific decision of a model. While the computation of \"arbitrary\" counterfactual explanations is well studied, it is still an open research problem how to efficiently compute plausible and feasible counterfactual explanations. We build upon recent work and propose and study a formal definition of plausible counterfactual explanations. In particular, we investigate how to use density estimators for enforcing plausibility and feasibility of counterfactual explanations. For the purpose of efficient computations, we propose convex density constraints that ensure that the resulting counterfactual is located in a region of the data space of high density."
                    },
                    {
                        "title": "Counterfactual Explanations of Concept Drift",
                        "abstract": "The notion of concept drift refers to the phenomenon that the distribution, which is underlying the observed data, changes over time; as a consequence machine learning models may become inaccurate and need adjustment. While there do exist methods to detect concept drift or to adjust models in the presence of observed drift, the question of explaining drift has hardly been considered so far. This problem is of importance, since it enables an inspection of the most prominent features where drift manifests itself; hence it enables human understanding of the necessity of change and it increases acceptance of life-long learning models. In this paper we present a novel technology, which characterizes concept drift in terms of the characteristic change of spatial features represented by typical examples based on counterfactual explanations. We establish a formal definition of this problem, derive an efficient algorithmic solution based on counterfactual explanations, and demonstrate its usefulness in several examples."
                    },
                    {
                        "title": "Interpretable Discriminative Dimensionality Reduction and Feature Selection on the Manifold",
                        "abstract": "Dimensionality reduction (DR) on the manifold includes effective methods which project the data from an implicit relational space onto a vectorial space. Regardless of the achievements in this area, these algorithms suffer from the lack of interpretation of the projection dimensions. Therefore, it is often difficult to explain the physical meaning behind the embedding dimensions. In this research, we propose the interpretable kernel DR algorithm (I-KDR) as a new algorithm which maps the data from the feature space to a lower dimensional space where the classes are more condensed with less overlapping. Besides, the algorithm creates the dimensions upon local contributions of the data samples, which makes it easier to interpret them by class labels. Additionally, we efficiently fuse the DR with feature selection task to select the most relevant features of the original space to the discriminative objective. Based on the empirical evidence, I-KDR provides better interpretations for embedding dimensions as well as higher discriminative performance in the embedded space compared to the state-of-the-art and popular DR algorithms."
                    },
                    {
                        "title": "Interpretable Multiple-Kernel Prototype Learning for Discriminative Representation and Feature Selection",
                        "abstract": "Prototype-based methods are of the particular interest for domain specialists and practitioners as they summarize a dataset by a small set of representatives. Therefore, in a classification setting, interpretability of the prototypes is as significant as the prediction accuracy of the algorithm. Nevertheless, the state-of-the-art methods make inefficient trade-offs between these concerns by sacrificing one in favor of the other, especially if the given data has a kernel-based representation. In this paper, we propose a novel interpretable multiple-kernel prototype learning (IMKPL) to construct highly interpretable prototypes in the feature space, which are also efficient for the discriminative representation of the data. Our method focuses on the local discrimination of the classes in the feature space and shaping the prototypes based on condensed class-homogeneous neighborhoods of data. Besides, IMKPL learns a combined embedding in the feature space in which the above objectives are better fulfilled. When the base kernels coincide with the data dimensions, this embedding results in a discriminative features selection. We evaluate IMKPL on several benchmarks from different domains which demonstrate its superiority to the related state-of-the-art methods regarding both interpretability and discriminative representation."
                    },
                    {
                        "title": "Efficient computation of contrastive explanations",
                        "abstract": "With the increasing deployment of machine learning systems in practice, transparency and explainability have become serious issues. Contrastive explanations are considered to be useful and intuitive, in particular when it comes to explaining decisions to lay people, since they mimic the way in which humans explain. Yet, so far, comparably little research has addressed computationally feasible technologies, which allow guarantees on uniqueness and optimality of the explanation and which enable an easy incorporation of additional constraints. Here, we will focus on specific types of models rather than black-box technologies. We study the relation of contrastive and counterfactual explanations and propose mathematical formalizations as well as a 2-phase algorithm for efficiently computing (plausible) pertinent positives of many standard machine learning models."
                    },
                    {
                        "title": "Single-Step Adversarial Training for Semantic Segmentation",
                        "abstract": "Even though deep neural networks succeed on many different tasks including semantic segmentation, they lack on robustness against adversarial examples. To counteract this exploit, often adversarial training is used. However, it is known that adversarial training with weak adversarial attacks (e.g. using the Fast Gradient Method) does not improve the robustness against stronger attacks. Recent research shows that it is possible to increase the robustness of such single-step methods by choosing an appropriate step size during the training. Finding such a step size, without increasing the computational effort of single-step adversarial training, is still an open challenge. In this work we address the computationally particularly demanding task of semantic segmentation and propose a new step size control algorithm that increases the robustness of single-step adversarial training. The proposed algorithm does not increase the computational effort of single-step adversarial training considerably and also simplifies training, because it is free of meta-parameter. We show that the robustness of our approach can compete with multi-step adversarial training on two popular benchmarks for semantic segmentation."
                    },
                    {
                        "title": "\"Even if ...\" -- Diverse Semifactual Explanations of Reject",
                        "abstract": "Machine learning based decision making systems applied in safety critical areas require reliable high certainty predictions. For this purpose, the system can be extended by an reject option which allows the system to reject inputs where only a prediction with an unacceptably low certainty would be possible. While being able to reject uncertain samples is important, it is also of importance to be able to explain why a particular sample was rejected. With the ongoing rise of eXplainable AI (XAI), a lot of explanation methodologies for machine learning based systems have been developed -- explaining reject options, however, is still a novel field where only very little prior work exists.   In this work, we propose to explain rejects by semifactual explanations, an instance of example-based explanation methods, which them self have not been widely considered in the XAI community yet. We propose a conceptual modeling of semifactual explanations for arbitrary reject options and empirically evaluate a specific implementation on a conformal prediction based reject option."
                    },
                    {
                        "title": "Analyzing the Influence of Training Samples on Explanations",
                        "abstract": "EXplainable AI (XAI) constitutes a popular method to analyze the reasoning of AI systems by explaining their decision-making, e.g. providing a counterfactual explanation of how to achieve recourse. However, in cases such as unexpected explanations, the user might be interested in learning about the cause of this explanation -- e.g. properties of the utilized training data that are responsible for the observed explanation. Under the umbrella of data valuation, first approaches have been proposed that estimate the influence of data samples on a given model. In this work, we take a slightly different stance, as we are interested in the influence of single samples on a model explanation rather than the model itself. Hence, we propose the novel problem of identifying training data samples that have a high influence on a given explanation (or related quantity) and investigate the particular case of differences in the cost of the recourse between protected groups. For this, we propose an algorithm that identifies such influential training samples."
                    },
                    {
                        "title": "Fairness-Enhancing Ensemble Classification in Water Distribution Networks",
                        "abstract": "As relevant examples such as the future criminal detection software [1] show, fairness of AI-based and social domain affecting decision support tools constitutes an important area of research. In this contribution, we investigate the applications of AI to socioeconomically relevant infrastructures such as those of water distribution networks (WDNs), where fairness issues have yet to gain a foothold. To establish the notion of fairness in this domain, we propose an appropriate definition of protected groups and group fairness in WDNs as an extension of existing definitions. We demonstrate that typical methods for the detection of leakages in WDNs are unfair in this sense. Further, we thus propose a remedy to increase the fairness which can be applied even to non-differentiable ensemble classification methods as used in this context."
                    },
                    {
                        "title": "Large-Margin Multiple Kernel Learning for Discriminative Features Selection and Representation Learning",
                        "abstract": "Multiple kernel learning (MKL) algorithms combine different base kernels to obtain a more efficient representation in the feature space. Focusing on discriminative tasks, MKL has been used successfully for feature selection and finding the significant modalities of the data. In such applications, each base kernel represents one dimension of the data or is derived from one specific descriptor. Therefore, MKL finds an optimal weighting scheme for the given kernels to increase the classification accuracy. Nevertheless, the majority of the works in this area focus on only binary classification problems or aim for linear separation of the classes in the kernel space, which are not realistic assumptions for many real-world problems. In this paper, we propose a novel multi-class MKL framework which improves the state-of-the-art by enhancing the local separation of the classes in the feature space. Besides, by using a sparsity term, our large-margin multiple kernel algorithm (LMMK) performs discriminative feature selection by aiming to employ a small subset of the base kernels. Based on our empirical evaluations on different real-world datasets, LMMK provides a competitive classification accuracy compared with the state-of-the-art algorithms in MKL. Additionally, it learns a sparse set of non-zero kernel weights which leads to a more interpretable feature selection and representation learning."
                    }
                ]
            },
            "88bf19cc-7064-42e4-a880-ead5f7035192": {
                "pk": "88bf19cc-7064-42e4-a880-ead5f7035192",
                "name": "Eyke H\u00fcllermeier",
                "collaborators": [
                    "Viktor Bengs",
                    "Sascha Henzgen",
                    "Clemens Damke",
                    "Julian Lienen",
                    "Weiwei Cheng",
                    "Arunselvan Ramaswamy",
                    "Javad Rahnama",
                    "Sadegh Abbaszadeh",
                    "Karlson Pfannschmidt",
                    "Willem Waegeman"
                ],
                "domain": [
                    "Fuzzy Logic",
                    "Machine Learning",
                    "Uncertainty Quantification",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Learning from Imprecise and Fuzzy Observations: Data Disambiguation through Generalized Loss Minimization",
                        "abstract": "Methods for analyzing or learning from \"fuzzy data\" have attracted increasing attention in recent years. In many cases, however, existing methods (for precise, non-fuzzy data) are extended to the fuzzy case in an ad-hoc manner, and without carefully considering the interpretation of a fuzzy set when being used for modeling data. Distinguishing between an ontic and an epistemic interpretation of fuzzy set-valued data, and focusing on the latter, we argue that a \"fuzzification\" of learning algorithms based on an application of the generic extension principle is not appropriate. In fact, the extension principle fails to properly exploit the inductive bias underlying statistical and machine learning methods, although this bias, at least in principle, offers a means for \"disambiguating\" the fuzzy data. Alternatively, we therefore propose a method which is based on the generalization of loss functions in empirical risk minimization, and which performs model identification and data disambiguation simultaneously. Elaborating on the fuzzification of specific types of losses, we establish connections to well-known loss functions in regression and classification. We compare our approach with related methods and illustrate its use in logistic regression for binary classification."
                    },
                    {
                        "title": "From knowledge-based to data-driven modeling of fuzzy rule-based systems: A critical reflection",
                        "abstract": "This paper briefly elaborates on a development in (applied) fuzzy logic that has taken place in the last couple of decades, namely, the complementation or even replacement of the traditional knowledge-based approach to fuzzy rule-based systems design by a data-driven one. It is argued that the classical rule-based modeling paradigm is actually more amenable to the knowledge-based approach, for which it has originally been conceived, while being less apt to data-driven model design. An important reason that prevents fuzzy (rule-based) systems from being leveraged in large-scale applications is the flat structure of rule bases, along with the local nature of fuzzy rules and their limited ability to express complex dependencies between variables. This motivates alternative approaches to fuzzy systems modeling, in which functional dependencies can be represented more flexibly and more compactly in terms of hierarchical structures."
                    },
                    {
                        "title": "Towards Analogy-Based Explanations in Machine Learning",
                        "abstract": "Principles of analogical reasoning have recently been applied in the context of machine learning, for example to develop new methods for classification and preference learning. In this paper, we argue that, while analogical reasoning is certainly useful for constructing new learning algorithms with high predictive accuracy, is is arguably not less interesting from an interpretability and explainability point of view. More specifically, we take the view that an analogy-based approach is a viable alternative to existing approaches in the realm of explainable AI and interpretable machine learning, and that analogy-based explanations of the predictions produced by a machine learning algorithm can complement similarity-based explanations in a meaningful way. To corroborate these claims, we outline the basic idea of an analogy-based explanation and illustrate its potential usefulness by means of some examples."
                    },
                    {
                        "title": "Prescriptive Machine Learning for Automated Decision Making: Challenges and Opportunities",
                        "abstract": "Recent applications of machine learning (ML) reveal a noticeable shift from its use for predictive modeling in the sense of a data-driven construction of models mainly used for the purpose of prediction (of ground-truth facts) to its use for prescriptive modeling. What is meant by this is the task of learning a model that stipulates appropriate decisions about the right course of action in real-world scenarios: Which medical therapy should be applied? Should this person be hired for the job? As argued in this article, prescriptive modeling comes with new technical conditions for learning and new demands regarding reliability, responsibility, and the ethics of decision making. Therefore, to support the data-driven design of decision-making agents that act in a rational but at the same time responsible manner, a rigorous methodological foundation of prescriptive ML is needed. The purpose of this short paper is to elaborate on specific characteristics of prescriptive ML and to highlight some key challenges it implies. Besides, drawing connections to other branches of contemporary AI research, the grounding of prescriptive ML in a (generalized) decision-theoretic framework is advocated."
                    },
                    {
                        "title": "Label Ranking with Abstention: Predicting Partial Orders by Thresholding Probability Distributions (Extended Abstract)",
                        "abstract": "We consider an extension of the setting of label ranking, in which the learner is allowed to make predictions in the form of partial instead of total orders. Predictions of that kind are interpreted as a partial abstention: If the learner is not sufficiently certain regarding the relative order of two alternatives, it may abstain from this decision and instead declare these alternatives as being incomparable. We propose a new method for learning to predict partial orders that improves on an existing approach, both theoretically and empirically. Our method is based on the idea of thresholding the probabilities of pairwise preferences between labels as induced by a predicted (parameterized) probability distribution on the set of all rankings."
                    },
                    {
                        "title": "Preselection Bandits",
                        "abstract": "In this paper, we introduce the Preselection Bandit problem, in which the learner preselects a subset of arms (choice alternatives) for a user, which then chooses the final arm from this subset. The learner is not aware of the user's preferences, but can learn them from observed choices. In our concrete setting, we allow these choices to be stochastic and model the user's actions by means of the Plackett-Luce model. The learner's main task is to preselect subsets that eventually lead to highly preferred choices. To formalize this goal, we introduce a reasonable notion of regret and derive lower bounds on the expected regret. Moreover, we propose algorithms for which the upper bound on expected regret matches the lower bound up to a logarithmic term of the time horizon."
                    },
                    {
                        "title": "Deep Q-Learning: Theoretical Insights from an Asymptotic Analysis",
                        "abstract": "Deep Q-Learning is an important reinforcement learning algorithm, which involves training a deep neural network, called Deep Q-Network (DQN), to approximate the well-known Q-function. Although wildly successful under laboratory conditions, serious gaps between theory and practice as well as a lack of formal guarantees prevent its use in the real world. Adopting a dynamical systems perspective, we provide a theoretical analysis of a popular version of Deep Q-Learning under realistic and verifiable assumptions. More specifically, we prove an important result on the convergence of the algorithm, characterizing the asymptotic behavior of the learning process. Our result sheds light on hitherto unexplained properties of the algorithm and helps understand empirical observations, such as performance inconsistencies even after training. Unlike previous theories, our analysis accommodates state Markov processes with multiple stationary distributions. In spite of the focus on Deep Q-Learning, we believe that our theory may be applied to understand other deep learning algorithms"
                    },
                    {
                        "title": "Learning Tversky Similarity",
                        "abstract": "In this paper, we advocate Tversky's ratio model as an appropriate basis for computational approaches to semantic similarity, that is, the comparison of objects such as images in a semantically meaningful way. We consider the problem of learning Tversky similarity measures from suitable training data indicating whether two objects tend to be similar or dissimilar. Experimentally, we evaluate our approach to similarity learning on two image datasets, showing that is performs very well compared to existing methods."
                    },
                    {
                        "title": "Machine Learning with the Sugeno Integral: The Case of Binary Classification",
                        "abstract": "In this paper, we elaborate on the use of the Sugeno integral in the context of machine learning. More specifically, we propose a method for binary classification, in which the Sugeno integral is used as an aggregation function that combines several local evaluations of an instance, pertaining to different features or measurements, into a single global evaluation. Due to the specific nature of the Sugeno integral, this approach is especially suitable for learning from ordinal data, that is, when measurements are taken from ordinal scales. This is a topic that has not received much attention in machine learning so far. The core of the learning problem itself consists of identifying the capacity underlying the Sugeno integral. To tackle this problem, we develop an algorithm based on linear programming. The algorithm also includes a suitable technique for transforming the original feature values into local evaluations (local utility scores), as well as a method for tuning a threshold on the global evaluation. To control the flexibility of the classifier and mitigate the problem of overfitting the training data, we generalize our approach toward $k$-maxitive capacities, where $k$ plays the role of a hyper-parameter of the learner. We present experimental studies, in which we compare our method with competing approaches on several benchmark data sets."
                    },
                    {
                        "title": "Learning Choice Functions via Pareto-Embeddings",
                        "abstract": "We consider the problem of learning to choose from a given set of objects, where each object is represented by a feature vector. Traditional approaches in choice modelling are mainly based on learning a latent, real-valued utility function, thereby inducing a linear order on choice alternatives. While this approach is suitable for discrete (top-1) choices, it is not straightforward how to use it for subset choices. Instead of mapping choice alternatives to the real number line, we propose to embed them into a higher-dimensional utility space, in which we identify choice sets with Pareto-optimal points. To this end, we propose a learning algorithm that minimizes a differentiable loss function suitable for this task. We demonstrate the feasibility of learning a Pareto-embedding on a suite of benchmark datasets."
                    },
                    {
                        "title": "Mining Rank Data",
                        "abstract": "The problem of frequent pattern mining has been studied quite extensively for various types of data, including sets, sequences, and graphs. Somewhat surprisingly, another important type of data, namely rank data, has received very little attention in data mining so far. In this paper, we therefore addresses the problem of mining rank data, that is, data in the form of rankings (total orders) of an underlying set of items. More specifically, two types of patterns are considered, namely frequent rankings and dependencies between such rankings in the form of association rules. Algorithms for mining frequent rankings and frequent closed rankings are proposed and tested experimentally, using both synthetic and real data."
                    },
                    {
                        "title": "Aleatoric and Epistemic Uncertainty in Machine Learning: An Introduction to Concepts and Methods",
                        "abstract": "The notion of uncertainty is of major importance in machine learning and constitutes a key element of machine learning methodology. In line with the statistical tradition, uncertainty has long been perceived as almost synonymous with standard probability and probabilistic predictions. Yet, due to the steadily increasing relevance of machine learning for practical applications and related issues such as safety requirements, new problems and challenges have recently been identified by machine learning scholars, and these problems may call for new methodological developments. In particular, this includes the importance of distinguishing between (at least) two different types of uncertainty, often referred to as aleatoric and epistemic. In this paper, we provide an introduction to the topic of uncertainty in machine learning as well as an overview of attempts so far at handling uncertainty in general and formalizing this distinction in particular."
                    },
                    {
                        "title": "TSK-Streams: Learning TSK Fuzzy Systems on Data Streams",
                        "abstract": "The problem of adaptive learning from evolving and possibly non-stationary data streams has attracted a lot of interest in machine learning in the recent past, and also stimulated research in related fields, such as computational intelligence and fuzzy systems. In particular, several rule-based methods for the incremental induction of regression models have been proposed. In this paper, we develop a method that combines the strengths of two existing approaches rooted in different learning paradigms. More concretely, our method adopts basic principles of the state-of-the-art learning algorithm AMRules and enriches them by the representational advantages of fuzzy rules. In a comprehensive experimental study, TSK-Streams is shown to be highly competitive in terms of performance."
                    },
                    {
                        "title": "Multi-Armed Bandits with Censored Consumption of Resources",
                        "abstract": "We consider a resource-aware variant of the classical multi-armed bandit problem: In each round, the learner selects an arm and determines a resource limit. It then observes a corresponding (random) reward, provided the (random) amount of consumed resources remains below the limit. Otherwise, the observation is censored, i.e., no reward is obtained. For this problem setting, we introduce a measure of regret, which incorporates the actual amount of allocated resources of each learning round as well as the optimality of realizable rewards. Thus, to minimize regret, the learner needs to set a resource limit and choose an arm in such a way that the chance to realize a high reward within the predefined resource limit is high, while the resource limit itself should be kept as low as possible. We propose a UCB-inspired online learning algorithm, which we analyze theoretically in terms of its regret upper bound. In a simulation study, we show that our learning algorithm outperforms straightforward extensions of standard multi-armed bandit algorithms."
                    },
                    {
                        "title": "Ranking Structured Objects with Graph Neural Networks",
                        "abstract": "Graph neural networks (GNNs) have been successfully applied in many structured data domains, with applications ranging from molecular property prediction to the analysis of social networks. Motivated by the broad applicability of GNNs, we propose the family of so-called RankGNNs, a combination of neural Learning to Rank (LtR) methods and GNNs. RankGNNs are trained with a set of pair-wise preferences between graphs, suggesting that one of them is preferred over the other. One practical application of this problem is drug screening, where an expert wants to find the most promising molecules in a large collection of drug candidates. We empirically demonstrate that our proposed pair-wise RankGNN approach either significantly outperforms or at least matches the ranking performance of the naive point-wise baseline approach, in which the LtR problem is solved via GNN-based graph regression."
                    },
                    {
                        "title": "Credal Self-Supervised Learning",
                        "abstract": "Self-training is an effective approach to semi-supervised learning. The key idea is to let the learner itself iteratively generate \"pseudo-supervision\" for unlabeled instances based on its current hypothesis. In combination with consistency regularization, pseudo-labeling has shown promising performance in various domains, for example in computer vision. To account for the hypothetical nature of the pseudo-labels, these are commonly provided in the form of probability distributions. Still, one may argue that even a probability distribution represents an excessive level of informedness, as it suggests that the learner precisely knows the ground-truth conditional probabilities. In our approach, we therefore allow the learner to label instances in the form of credal sets, that is, sets of (candidate) probability distributions. Thanks to this increased expressiveness, the learner is able to represent uncertainty and a lack of knowledge in a more flexible and more faithful manner. To learn from weakly labeled data of that kind, we leverage methods that have recently been proposed in the realm of so-called superset learning. In an exhaustive empirical evaluation, we compare our methodology to state-of-the-art self-supervision approaches, showing competitive to superior performance especially in low-label scenarios incorporating a high degree of uncertainty."
                    },
                    {
                        "title": "Conformal Prediction Intervals for Remaining Useful Lifetime Estimation",
                        "abstract": "The main objective of Prognostics and Health Management is to estimate the Remaining Useful Lifetime (RUL), namely, the time that a system or a piece of equipment is still in working order before starting to function incorrectly. In recent years, numerous machine learning algorithms have been proposed for RUL estimation, mainly focusing on providing more accurate RUL predictions. However, there are many sources of uncertainty in the problem, such as inherent randomness of systems failure, lack of knowledge regarding their future states, and inaccuracy of the underlying predictive models, making it infeasible to predict the RULs precisely. Hence, it is of utmost importance to quantify the uncertainty alongside the RUL predictions. In this work, we investigate the conformal prediction (CP) framework that represents uncertainty by predicting sets of possible values for the target variable (intervals in the case of RUL) instead of making point predictions. Under very mild technical assumptions, CP formally guarantees that the actual value (true RUL) is covered by the predicted set with a degree of certainty that can be prespecified. We study three CP algorithms to conformalize any single-point RUL predictor and turn it into a valid interval predictor. Finally, we conformalize two single-point RUL predictors, deep convolutional neural networks and gradient boosting, and illustrate their performance on the Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) data sets."
                    },
                    {
                        "title": "Mitigating Label Noise through Data Ambiguation",
                        "abstract": "Label noise poses an important challenge in machine learning, especially in deep learning, in which large models with high expressive power dominate the field. Models of that kind are prone to memorizing incorrect labels, thereby harming generalization performance. Many methods have been proposed to address this problem, including robust loss functions and more complex label correction approaches. Robust loss functions are appealing due to their simplicity, but typically lack flexibility, while label correction usually adds substantial complexity to the training setup. In this paper, we suggest to address the shortcomings of both methodologies by \"ambiguating\" the target information, adding additional, complementary candidate labels in case the learner is not sufficiently convinced of the observed training label. More precisely, we leverage the framework of so-called superset learning to construct set-valued targets based on a confidence threshold, which deliver imprecise yet more reliable beliefs about the ground-truth, effectively helping the learner to suppress the memorization effect. In an extensive empirical evaluation, our method demonstrates favorable learning behavior on synthetic and real-world noise, confirming the effectiveness in detecting and correcting erroneous training labels."
                    },
                    {
                        "title": "Weighting by Tying: A New Approach to Weighted Rank Correlation",
                        "abstract": "Measures of rank correlation are commonly used in statistics to capture the degree of concordance between two orderings of the same set of items. Standard measures like Kendall's tau and Spearman's rho coefficient put equal emphasis on each position of a ranking. Yet, motivated by applications in which some of the positions (typically those on the top) are more important than others, a few weighted variants of these measures have been proposed. Most of these generalizations fail to meet desirable formal properties, however. Besides, they are often quite inflexible in the sense of committing to a fixed weighing scheme. In this paper, we propose a weighted rank correlation measure on the basis of fuzzy order relations. Our measure, called scaled gamma, is related to Goodman and Kruskal's gamma rank correlation. It is parametrized by a fuzzy equivalence relation on the rank positions, which in turn is specified conveniently by a so-called scaling function. This approach combines soundness with flexibility: it has a sound formal foundation and allows for weighing rank positions in a flexible way."
                    },
                    {
                        "title": "Linear Opinion Pooling for Uncertainty Quantification on Graphs",
                        "abstract": "We address the problem of uncertainty quantification for graph-structured data, or, more specifically, the problem to quantify the predictive uncertainty in (semi-supervised) node classification. Key questions in this regard concern the distinction between two different types of uncertainty, aleatoric and epistemic, and how to support uncertainty quantification by leveraging the structural information provided by the graph topology. Challenging assumptions and postulates of state-of-the-art methods, we propose a novel approach that represents (epistemic) uncertainty in terms of mixtures of Dirichlet distributions and refers to the established principle of linear opinion pooling for propagating information between neighbored nodes in the graph. The effectiveness of this approach is demonstrated in a series of experiments on a variety of graph-structured datasets."
                    }
                ]
            }
        }
    },
    "2311.01906": {
        "paper_data": {
            "title": "Simplifying Transformer Blocks",
            "url": "http://arxiv.org/abs/2311.01906v2",
            "arxiv_id": "2311.01906",
            "authors": [
                "Bobby He",
                "Thomas Hofmann"
            ],
            "abstract": "A simple design recipe for deep Transformers is to compose identical building blocks. But standard transformer blocks are far from simple, interweaving attention and MLP sub-blocks with skip connections & normalisation layers in precise arrangements. This complexity leads to brittle architectures, where seemingly minor changes can significantly reduce training speed, or render models untrainable.   In this work, we ask to what extent the standard transformer block can be simplified? Combining signal propagation theory and empirical observations, we motivate modifications that allow many block components to be removed with no loss of training speed, including skip connections, projection or value parameters, sequential sub-blocks and normalisation layers. In experiments on both autoregressive decoder-only and BERT encoder-only models, our simplified transformers emulate the per-update training speed and performance of standard transformers, while enjoying 15% faster training throughput, and using 15% fewer parameters.",
            "introduction": "   1 Introduction  The transformer architecture (Vaswani et\u00a0al., 2017) is arguably the workhorse behind many recent successes in deep learning. A simple way to construct a deep transformer architecture is by stacking multiple identical transformer \u201cblocks\u201d one after another in sequence. Each block, however, is more complicated and consists of many different components, which need to be combined in specific arrangements in order to achieve good performance. Surprisingly, the base transformer block has changed very little since its inception, despite attracting the interest of many researchers.   In this work, we study whether the standard transformer block can be simplified. More specifically, we probe the necessity of several block components, including skip connections, projection/value matrices, sequential sub-blocks and normalisation layers. For each considered component, we ask if it can be removed without loss of training speed (both in terms of per-update step & runtime), and what architectural modifications need to be made to the transformer block in order to do so.   We believe the problem of simplifying transformer blocks without compromising training speed is an interesting research question for several reasons. First, modern neural network (NN) architectures have complex designs with many components, and it is not clear the roles played by these different components in NN training dynamics, nor how they interact with each other. This is particularly pertinent given the existing gap between theory and practice in deep learning, where theorists working to understand the mechanisms of deep learning often only consider simplified architectures due to convenience, not necessarily reflective of modern architectures used in practice. Simplifying the NN architectures used in practice can help towards bridging this divide.   On a related theoretical note, our work highlights both strengths and current limitations of signal propagation: a theory that has proven influential due to its ability to motivate practical design choices in deep NN architectures. Signal propagation (Poole et\u00a0al., 2016; Schoenholz et\u00a0al., 2017; Hayou et\u00a0al., 2019) studies the evolution of geometric information in an NN at initialisation, captured through inner products of layerwise representations across inputs, and has inspired many impressive results in training deep NNs (Xiao et\u00a0al., 2018; Brock et\u00a0al., 2021; Martens et\u00a0al., 2021; Zaidi et\u00a0al., 2023). However, the current theory only considers a model at initialisation, and often considers only the initial forward pass. As such, signal propagation at present is unable to shed light on many intricacies of deep NN training dynamics, for example the benefits of skip connections for training speed. Though signal propagation is crucial in motivating our modifications, we would not have arrived at our simplified transformer blocks from theory alone, and relied also on empirical insights.   Finally, on the practical side, given the cost of training and deploying large transformer models nowadays, any efficiency gains in the training and inference pipelines for the transformer architecture represent significant potential savings. Simplifying the transformer block by removing non-essential components both reduces the parameter count and increases throughput in our models. In particular, we show that it is possible to remove skip connections, value parameters, projection parameters and sequential sub-blocks, all while matching the standard transformer in terms of training speed and downstream",
            "references": [
                {
                    "title": "Towards Training Without Depth Limits: Batch Normalization Without Gradient Explosion",
                    "abstract": "Normalization layers are one of the key building blocks for deep neural networks. Several theoretical studies have shown that batch normalization improves the signal propagation, by avoiding the representations from becoming collinear across the layers. However, results on mean-field theory of batch normalization also conclude that this benefit comes at the expense of exploding gradients in depth. Motivated by these two aspects of batch normalization, in this study we pose the following question:\"Can a batch-normalized network keep the optimal signal propagation properties, but avoid exploding gradients?\"We answer this question in the affirmative by giving a particular construction of an Multi-Layer Perceptron (MLP) with linear activations and batch-normalization that provably has bounded gradients at any depth. Based on Weingarten calculus, we develop a rigorous and non-asymptotic theory for this constructed MLP that gives a precise characterization of forward signal propagation, while proving that gradients remain bounded for linearly independent input samples, which holds in most practical settings. Inspired by our theory, we also design an activation shaping scheme that empirically achieves the same properties for certain non-linear activations."
                },
                {
                    "title": "The Languini Kitchen: Enabling Language Modelling Research at Different Scales of Compute",
                    "abstract": "The Languini Kitchen serves as both a research collective and codebase designed to empower researchers with limited computational resources to contribute meaningfully to the field of language modelling. We introduce an experimental protocol that enables model comparisons based on equivalent compute, measured in accelerator hours. The number of tokens on which a model is trained is defined by the model's throughput and the chosen compute class. Notably, this approach avoids constraints on critical hyperparameters which affect total parameters or floating-point operations. For evaluation, we pre-process an existing large, diverse, and high-quality dataset of books that surpasses existing academic benchmarks in quality, diversity, and document length. On it, we compare methods based on their empirical scaling trends which are estimated through experiments at various levels of compute. This work also provides two baseline models: a feed-forward model derived from the GPT-2 architecture and a recurrent model in the form of a novel LSTM with ten-fold throughput. While the GPT baseline achieves better perplexity throughout all our levels of compute, our LSTM baseline exhibits a predictable and more favourable scaling law. This is due to the improved throughput and the need for fewer training tokens to achieve the same decrease in test perplexity. Extrapolating the scaling laws leads of both models results in an intersection at roughly 50,000 accelerator hours. We hope this work can serve as the foundation for meaningful and reproducible language modelling research."
                },
                {
                    "title": "The Shaped Transformer: Attention Models in the Infinite Depth-and-Width Limit",
                    "abstract": "In deep learning theory, the covariance matrix of the representations serves as a proxy to examine the network's trainability. Motivated by the success of Transformers, we study the covariance matrix of a modified Softmax-based attention model with skip connections in the proportional limit of infinite-depth-and-width. We show that at initialization the limiting distribution can be described by a stochastic differential equation (SDE) indexed by the depth-to-width ratio. To achieve a well-defined stochastic limit, the Transformer's attention mechanism is modified by centering the Softmax output at identity, and scaling the Softmax logits by a width-dependent temperature parameter. We examine the stability of the network through the corresponding SDE, showing how the scale of both the drift and diffusion can be elegantly controlled with the aid of residual connections. The existence of a stable SDE implies that the covariance structure is well-behaved, even for very large depth and width, thus preventing the notorious issues of rank degeneracy in deep attention models. Finally, we show, through simulations, that the SDE provides a surprisingly good description of the corresponding finite-size model. We coin the name shaped Transformer for these architectural modifications."
                },
                {
                    "title": "Centered Self-Attention Layers",
                    "abstract": "The self-attention mechanism in transformers and the message-passing mechanism in graph neural networks are repeatedly applied within deep learning architectures. We show that this application inevitably leads to oversmoothing, i.e., to similar representations at the deeper layers for different tokens in transformers and different nodes in graph neural networks. Based on our analysis, we present a correction term to the aggregating operator of these mechanisms. Empirically, this simple term eliminates much of the oversmoothing problem in visual transformers, obtaining performance in weakly supervised segmentation that surpasses elaborate baseline methods that introduce multiple auxiliary networks and training phrases. In graph neural networks, the correction term enables the training of very deep architectures more effectively than many recent solutions to the same problem."
                },
                {
                    "title": "Mimetic Initialization of Self-Attention Layers",
                    "abstract": "It is notoriously difficult to train Transformers on small datasets; typically, large pre-trained models are instead used as the starting point. We explore the weights of such pre-trained Transformers (particularly for vision) to attempt to find reasons for this discrepancy. Surprisingly, we find that simply initializing the weights of self-attention layers so that they\"look\"more like their pre-trained counterparts allows us to train vanilla Transformers faster and to higher final accuracies, particularly on vision tasks such as CIFAR-10 and ImageNet classification, where we see gains in accuracy of over 5% and 4%, respectively. Our initialization scheme is closed form, learning-free, and very simple: we set the product of the query and key weights to be approximately the identity, and the product of the value and projection weights to approximately the negative identity. As this mimics the patterns we saw in pre-trained Transformers, we call the technique\"mimetic initialization\"."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "Deep Transformers without Shortcuts: Modifying Self-attention for Faithful Signal Propagation",
                    "abstract": "Skip connections and normalisation layers form two standard architectural components that are ubiquitous for the training of Deep Neural Networks (DNNs), but whose precise roles are poorly understood. Recent approaches such as Deep Kernel Shaping have made progress towards reducing our reliance on them, using insights from wide NN kernel theory to improve signal propagation in vanilla DNNs (which we define as networks without skips or normalisation). However, these approaches are incompatible with the self-attention layers present in transformers, whose kernels are intrinsically more complicated to analyse and control. And so the question remains: is it possible to train deep vanilla transformers? We answer this question in the affirmative by designing several approaches that use combinations of parameter initialisations, bias matrices and location-dependent rescaling to achieve faithful signal propagation in vanilla transformers. Our methods address various intricacies specific to signal propagation in transformers, including the interaction with positional encoding and causal masking. In experiments on WikiText-103 and C4, our approaches enable deep transformers without normalisation to train at speeds matching their standard counterparts, and deep vanilla transformers to reach the same performance as standard ones after about 5 times more iterations."
                },
                {
                    "title": "Scaling Vision Transformers to 22 Billion Parameters",
                    "abstract": "The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for\"LLM-like\"scaling in vision, and provides key steps towards getting there."
                },
                {
                    "title": "Width and Depth Limits Commute in Residual Networks",
                    "abstract": "We show that taking the width and depth to infinity in a deep neural network with skip connections, when branches are scaled by $1/\\sqrt{depth}$ (the only nontrivial scaling), result in the same covariance structure no matter how that limit is taken. This explains why the standard infinite-width-then-depth approach provides practical insights even for networks with depth of the same order as width. We also demonstrate that the pre-activations, in this case, have Gaussian distributions which has direct applications in Bayesian deep learning. We conduct extensive simulations that show an excellent match with our theoretical findings."
                },
                {
                    "title": "Cramming: Training a Language Model on a Single GPU in One Day",
                    "abstract": "Recent trends in language modeling have focused on increasing performance through scaling, and have resulted in an environment where training language models is out of reach for most researchers and practitioners. While most in the community are asking how to push the limits of extreme computation, we ask the opposite question: How far can we get with a single GPU in just one day? We investigate the downstream performance achievable with a transformer-based language model trained completely from scratch with masked language modeling for a single day on a single consumer GPU. Aside from re-analyzing nearly all components of the pretraining pipeline for this scenario and providing a modified pipeline with performance close to BERT, we investigate why scaling down is hard, and which modifications actually improve performance in this scenario. We provide evidence that even in this constrained setting, performance closely follows scaling laws observed in large-compute settings. Through the lens of scaling laws, we categorize a range of recent improvements to training and architecture and discuss their merit and practical applicability (or lack thereof) for the limited compute setting."
                },
                {
                    "title": "Signal Propagation in Transformers: Theoretical Perspectives and the Role of Rank Collapse",
                    "abstract": "Transformers have achieved remarkable success in several domains, ranging from natural language processing to computer vision. Nevertheless, it has been recently shown that stacking self-attention layers - the distinctive architectural component of Transformers - can result in rank collapse of the tokens' representations at initialization. The question of if and how rank collapse affects training is still largely unanswered, and its investigation is necessary for a more comprehensive understanding of this architecture. In this work, we shed new light on the causes and the effects of this phenomenon. First, we show that rank collapse of the tokens' representations hinders training by causing the gradients of the queries and keys to vanish at initialization. Furthermore, we provide a thorough description of the origin of rank collapse and discuss how to prevent it via an appropriate depth-dependent scaling of the residual branches. Finally, our analysis unveils that specific architectural hyperparameters affect the gradients of queries and values differently, leading to disproportionate gradient norms. This suggests an explanation for the widespread use of adaptive methods for Transformers' optimization."
                },
                {
                    "title": "The Neural Covariance SDE: Shaped Infinite Depth-and-Width Networks at Initialization",
                    "abstract": "The logit outputs of a feedforward neural network at initialization are conditionally Gaussian, given a random covariance matrix defined by the penultimate layer. In this work, we study the distribution of this random matrix. Recent work has shown that shaping the activation function as network depth grows large is necessary for this covariance matrix to be non-degenerate. However, the current infinite-width-style understanding of this shaping method is unsatisfactory for large depth: infinite-width analyses ignore the microscopic fluctuations from layer to layer, but these fluctuations accumulate over many layers. To overcome this shortcoming, we study the random covariance matrix in the shaped infinite-depth-and-width limit. We identify the precise scaling of the activation function necessary to arrive at a non-trivial limit, and show that the random covariance matrix is governed by a stochastic differential equation (SDE) that we call the Neural Covariance SDE. Using simulations, we show that the SDE closely matches the distribution of the random covariance matrix of finite networks. Additionally, we recover an if-and-only-if condition for exploding and vanishing norms of large shaped networks based on the activation function."
                },
                {
                    "title": "Pre-training via Denoising for Molecular Property Prediction",
                    "abstract": "Many important problems involving molecular property prediction from 3D structures have limited data, posing a generalization challenge for neural networks. In this paper, we describe a pre-training technique based on denoising that achieves a new state-of-the-art in molecular property prediction by utilizing large datasets of 3D molecular structures at equilibrium to learn meaningful representations for downstream tasks. Relying on the well-known link between denoising autoencoders and score-matching, we show that the denoising objective corresponds to learning a molecular force field -- arising from approximating the Boltzmann distribution with a mixture of Gaussians -- directly from equilibrium structures. Our experiments demonstrate that using this pre-training objective significantly improves performance on multiple benchmarks, achieving a new state-of-the-art on the majority of targets in the widely used QM9 dataset. Our analysis then provides practical insights into the effects of different factors -- dataset sizes, model size and architecture, and the choice of upstream and downstream datasets -- on pre-training."
                },
                {
                    "title": "Re-parameterizing Your Optimizers rather than Architectures",
                    "abstract": "The well-designed structures in neural networks reflect the prior knowledge incorporated into the models. However, though different models have various priors, we are used to training them with model-agnostic optimizers such as SGD. In this paper, we propose to incorporate model-specific prior knowledge into optimizers by modifying the gradients according to a set of model-specific hyper-parameters. Such a methodology is referred to as Gradient Re-parameterization, and the optimizers are named RepOptimizers. For the extreme simplicity of model structure, we focus on a VGG-style plain model and showcase that such a simple model trained with a RepOptimizer, which is referred to as RepOpt-VGG, performs on par with or better than the recent well-designed models. From a practical perspective, RepOpt-VGG is a favorable base model because of its simple structure, high inference speed and training efficiency. Compared to Structural Re-parameterization, which adds priors into models via constructing extra training-time structures, RepOptimizers require no extra forward/backward computations and solve the problem of quantization. We hope to spark further research beyond the realms of model structure design. Code and models \\url{https://github.com/DingXiaoH/RepOptimizers}."
                },
                {
                    "title": "PaLM: Scaling Language Modeling with Pathways",
                    "abstract": "Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model PaLM. We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies."
                },
                {
                    "title": "Training Compute-Optimal Large Language Models",
                    "abstract": "We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over 400 language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, Chinchilla, that uses the same compute budget as Gopher but with 70B parameters and 4$\\times$ more more data. Chinchilla uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that Chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, Chinchilla reaches a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, greater than a 7% improvement over Gopher."
                },
                {
                    "title": "Deep Learning without Shortcuts: Shaping the Kernel with Tailored Rectifiers",
                    "abstract": "Training very deep neural networks is still an extremely challenging task. The common solution is to use shortcut connections and normalization layers, which are both crucial ingredients in the popular ResNet architecture. However, there is strong evidence to suggest that ResNets behave more like ensembles of shallower networks than truly deep ones. Recently, it was shown that deep vanilla networks (i.e. networks without normalization layers or shortcut connections) can be trained as fast as ResNets by applying certain transformations to their activation functions. However, this method (called Deep Kernel Shaping) isn't fully compatible with ReLUs, and produces networks that overfit significantly more than ResNets on ImageNet. In this work, we rectify this situation by developing a new type of transformation that is fully compatible with a variant of ReLUs -- Leaky ReLUs. We show in experiments that our method, which introduces negligible extra computational cost, achieves validation accuracies with deep vanilla networks that are competitive with ResNets (of the same width/depth), and significantly higher than those obtained with the Edge of Chaos (EOC) method. And unlike with EOC, the validation accuracies we obtain do not get worse with depth."
                },
                {
                    "title": "Anti-Oversmoothing in Deep Vision Transformers via the Fourier Domain Analysis: From Theory to Practice",
                    "abstract": "Vision Transformer (ViT) has recently demonstrated promise in computer vision problems. However, unlike Convolutional Neural Networks (CNN), it is known that the performance of ViT saturates quickly with depth increasing, due to the observed attention collapse or patch uniformity. Despite a couple of empirical solutions, a rigorous framework studying on this scalability issue remains elusive. In this paper, we first establish a rigorous theory framework to analyze ViT features from the Fourier spectrum domain. We show that the self-attention mechanism inherently amounts to a low-pass filter, which indicates when ViT scales up its depth, excessive low-pass filtering will cause feature maps to only preserve their Direct-Current (DC) component. We then propose two straightforward yet effective techniques to mitigate the undesirable low-pass limitation. The first technique, termed AttnScale, decomposes a self-attention block into low-pass and high-pass components, then rescales and combines these two filters to produce an all-pass self-attention matrix. The second technique, termed FeatScale, re-weights feature maps on separate frequency bands to amplify the high-frequency signals. Both techniques are efficient and hyperparameter-free, while effectively overcoming relevant ViT training artifacts such as attention collapse and patch uniformity. By seamlessly plugging in our techniques to multiple ViT variants, we demonstrate that they consistently help ViTs benefit from deeper architectures, bringing up to 1.1% performance gains\"for free\"(e.g., with little parameter overhead). We publicly release our codes and pre-trained models at https://github.com/VITA-Group/ViT-Anti-Oversmoothing."
                },
                {
                    "title": "DeepNet: Scaling Transformers to 1,000 Layers",
                    "abstract": "In this paper, we propose a simple yet effective method to stabilize extremely deep Transformers. Specifically, we introduce a new normalization function (DeepNorm) to modify the residual connection in Transformer, accompanying with theoretically derived initialization. In-depth theoretical analysis shows that model updates can be bounded in a stable way. The proposed method combines the best of two worlds, i.e., good performance of Post-LN and stable training of Pre-LN, making DeepNorm a preferred alternative. We successfully scale Transformers up to 1,000 layers (i.e., 2,500 attention and feed-forward network sublayers) without difficulty, which is one order of magnitude deeper than previous deep Transformers. Extensive experiments demonstrate that DeepNet has superior performance across various benchmarks, including machine translation, language modeling (i.e., BERT, GPT) and vision pre-training (i.e., BEiT). Remarkably, on a multilingual benchmark with 7,482 translation directions, our 200-layer model with 3.2B parameters significantly outperforms the 48-layer state-of-the-art model with 12B parameters by 5 BLEU points, which indicates a promising scaling direction."
                },
                {
                    "title": "TrimBERT: Tailoring BERT for Trade-offs",
                    "abstract": "Models based on BERT have been extremely successful in solving a variety of natural language processing (NLP) tasks. Unfortunately, many of these large models require a great deal of computational resources and/or time for pre-training and fine-tuning which limits wider adoptability. While self-attention layers have been well-studied, a strong justification for inclusion of the intermediate layers which follow them remains missing in the literature. In this work, we show that reducing the number of intermediate layers in BERT-Base results in minimal fine-tuning accuracy loss of downstream tasks while significantly decreasing model size and training time. We further mitigate two key bottlenecks, by replacing all softmax operations in the self-attention layers with a computationally simpler alternative and removing half of all layernorm operations. This further decreases the training time while maintaining a high level of fine-tuning accuracy."
                },
                {
                    "title": "Rapid training of deep neural networks without skip connections or normalization layers using Deep Kernel Shaping",
                    "abstract": "Using an extended and formalized version of the Q/C map analysis of Poole et al. (2016), along with Neural Tangent Kernel theory, we identify the main pathologies present in deep networks that prevent them from training fast and generalizing to unseen data, and show how these can be avoided by carefully controlling the\"shape\"of the network's initialization-time kernel function. We then develop a method called Deep Kernel Shaping (DKS), which accomplishes this using a combination of precise parameter initialization, activation function transformations, and small architectural tweaks, all of which preserve the model class. In our experiments we show that DKS enables SGD training of residual networks without normalization layers on Imagenet and CIFAR-10 classification tasks at speeds comparable to standard ResNetV2 and Wide-ResNet models, with only a small decrease in generalization performance. And when using K-FAC as the optimizer, we achieve similar results for networks without skip connections. Our results apply for a large variety of activation functions, including those which traditionally perform very badly, such as the logistic sigmoid. In addition to DKS, we contribute a detailed analysis of skip connections, normalization layers, special activation functions like RELU and SELU, and various initialization schemes, explaining their effectiveness as alternative (and ultimately incomplete) ways of\"shaping\"the network's initialization-time kernel."
                },
                {
                    "title": "Going deeper with Image Transformers",
                    "abstract": "Transformers have been recently adapted for large scale image classification, achieving high scores shaking up the long supremacy of convolutional neural networks. However the optimization of vision transformers has been little studied so far. In this work, we build and optimize deeper transformer networks for image classification. In particular, we investigate the interplay of architecture and optimization of such dedicated transformers. We make two architecture changes that significantly improve the accuracy of deep transformers. This leads us to produce models whose performance does not saturate early with more depth, for in-stance we obtain 86.5% top-1 accuracy on Imagenet when training with no external data, we thus attain the current sate of the art with less floating-point operations and parameters. Our best model establishes the new state of the art on Imagenet with Reassessed labels and Imagenet-V2 / match frequency, in the setting with no additional training data. We share our code and models1."
                },
                {
                    "title": "Attention is Not All You Need: Pure Attention Loses Rank Doubly Exponentially with Depth",
                    "abstract": "Attention-based architectures have become ubiquitous in machine learning, yet our understanding of the reasons for their effectiveness remains limited. This work proposes a new way to understand self-attention networks: we show that their output can be decomposed into a sum of smaller terms, each involving the operation of a sequence of attention heads across layers. Using this decomposition, we prove that self-attention possesses a strong inductive bias towards\"token uniformity\". Specifically, without skip connections or multi-layer perceptrons (MLPs), the output converges doubly exponentially to a rank-1 matrix. On the other hand, skip connections and MLPs stop the output from degeneration. Our experiments verify the identified convergence phenomena on different variants of standard transformer architectures."
                },
                {
                    "title": "Linear Transformers Are Secretly Fast Weight Programmers",
                    "abstract": "We show the formal equivalence of linearised self-attention mechanisms and fast weight controllers from the early '90s, where a ``slow\"neural net learns by gradient descent to program the ``fast weights\"of another net through sequences of elementary programming instructions which are additive outer products of self-invented activation patterns (today called keys and values). Such Fast Weight Programmers (FWPs) learn to manipulate the contents of a finite memory and dynamically interact with it. We infer a memory capacity limitation of recent linearised softmax attention variants, and replace the purely additive outer products by a delta rule-like programming instruction, such that the FWP can more easily learn to correct the current mapping from keys to values. The FWP also learns to compute dynamically changing learning rates. We also propose a new kernel function to linearise attention which balances simplicity and effectiveness. We conduct experiments on synthetic retrieval problems as well as standard machine translation and language modelling tasks which demonstrate the benefits of our methods."
                },
                {
                    "title": "High-Performance Large-Scale Image Recognition Without Normalization",
                    "abstract": "Batch normalization is a key component of most image classification models, but it has many undesirable properties stemming from its dependence on the batch size and interactions between examples. Although recent work has succeeded in training deep ResNets without normalization layers, these models do not match the test accuracies of the best batch-normalized networks, and are often unstable for large learning rates or strong data augmentations. In this work, we develop an adaptive gradient clipping technique which overcomes these instabilities, and design a significantly improved class of Normalizer-Free ResNets. Our smaller models match the test accuracy of an EfficientNet-B7 on ImageNet while being up to 8.7x faster to train, and our largest models attain a new state-of-the-art top-1 accuracy of 86.5%. In addition, Normalizer-Free models attain significantly better performance than their batch-normalized counterparts when finetuning on ImageNet after large-scale pre-training on a dataset of 300 million labeled images, with our best models obtaining an accuracy of 89.2%. Our code is available at https://github.com/deepmind/ deepmind-research/tree/master/nfnets"
                },
                {
                    "title": "RepVGG: Making VGG-style ConvNets Great Again",
                    "abstract": "We present a simple but powerful architecture of convolutional neural network, which has a VGG-like inference-time body composed of nothing but a stack of 3 \u00d7 3 convolution and ReLU, while the training-time model has a multi-branch topology. Such decoupling of the training-time and inference-time architecture is realized by a structural re-parameterization technique so that the model is named RepVGG. On ImageNet, RepVGG reaches over 80% top-1 accuracy, which is the first time for a plain model, to the best of our knowledge. On NVIDIA 1080Ti GPU, RepVGG models run 83% faster than ResNet-50 or 101% faster than ResNet-101 with higher accuracy and show favorable accuracy-speed trade-off compared to the state-of-the-art models like EfficientNet and RegNet. The code and trained models are available at https://github.com/megvii-model/RepVGG."
                },
                {
                    "title": "The Pile: An 800GB Dataset of Diverse Text for Language Modeling",
                    "abstract": "Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present the Pile : an 825 GiB English text corpus tar-geted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets\u2014both existing and newly constructed\u2014many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve signi\ufb01cantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction. 1"
                },
                {
                    "title": "Stable ResNet",
                    "abstract": "Deep ResNet architectures have achieved state of the art performance on many tasks. While they solve the problem of gradient vanishing, they might suffer from gradient exploding as the depth becomes large (Yang et al. 2017). Moreover, recent results have shown that ResNet might lose expressivity as the depth goes to infinity (Yang et al. 2017, Hayou et al. 2019). To resolve these issues, we introduce a new class of ResNet architectures, called Stable ResNet, that have the property of stabilizing the gradient while ensuring expressivity in the infinite depth limit."
                },
                {
                    "title": "Improving Transformer Optimization Through Better Initialization",
                    "abstract": "The Transformer architecture has achieved considerable success recently; the key component of the Transformer is the attention layer that enables the model to focus on important regions within an input sequence. Gradient optimization with attention layers can be notoriously dif\ufb01cult requiring tricks such as learning rate warmup to prevent divergence. As Transformer models are becoming larger and more expensive to train, recent research has focused on understanding and improving optimization in these architectures. In this work our contributions are two-fold: we \ufb01rst investigate and empirically validate the source of optimization problems in the encoder-decoder Transformer architecture; we then propose a new weight initialization scheme with theoretical justi-\ufb01cation, that enables training without warmup or layer normalization. Empirical results on public machine translation benchmarks show that our approach achieves leading accuracy, allowing to train deep Transformer models with 200 layers in both encoder and decoder (over 1000 attention/MLP blocks) without dif\ufb01culty. Code for this work is available here: https://github. com/layer6ai-labs/T-Fixup ."
                },
                {
                    "title": "Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention",
                    "abstract": "Transformers achieve remarkable performance in several tasks but due to their quadratic complexity, with respect to the input's length, they are prohibitively slow for very long sequences. To address this limitation, we express the self-attention as a linear dot-product of kernel feature maps and make use of the associativity property of matrix products to reduce the complexity from $\\mathcal{O}\\left(N^2\\right)$ to $\\mathcal{O}\\left(N\\right)$, where $N$ is the sequence length. We show that this formulation permits an iterative implementation that dramatically accelerates autoregressive transformers and reveals their relationship to recurrent neural networks. Our linear transformers achieve similar performance to vanilla transformers and they are up to 4000x faster on autoregressive prediction of very long sequences."
                },
                {
                    "title": "Understanding the Difficulty of Training Transformers",
                    "abstract": "Transformers have been proved effective for many deep learning tasks. Training transformers, however, requires non-trivial efforts regarding carefully designing learning rate schedulers and cutting-edge optimizers (the standard SGD fails to train Transformers effectively). In this paper, we study Transformer training from both theoretical and empirical perspectives. Our analysis reveals that unbalanced gradients are not the root cause of the instability of training. Instead, we identify an amplification effect that substantially influences training. Specifically, we observe that for each layer in a multi-layer Transformer model, heavy dependency on its residual branch makes training unstable since it amplifies small parameter perturbations (e.g., parameter updates) and result in significant disturbances in the model output, yet a light dependency limits the potential of model training and can lead to an inferior trained model. Inspired by our analysis, we propose Admin ($\\mathbf{Ad}$aptive $\\mathbf{m}$odel $\\mathbf{in}$itialization) to stabilize the training in the early stage and unleash its full potential in the late stage. Extensive experiments show that Admin is more stable, converges faster, and leads to better performance."
                },
                {
                    "title": "ReZero is All You Need: Fast Convergence at Large Depth",
                    "abstract": "Deep networks often suffer from vanishing or exploding gradients due to inefficient signal propagation, leading to long training times or convergence difficulties. Various architecture designs, sophisticated residual-style networks, and initialization schemes have been shown to improve deep signal propagation. Recently, Pennington et al. used free probability theory to show that dynamical isometry plays an integral role in efficient deep learning. We show that the simplest architecture change of gating each residual connection using a single zero-initialized parameter satisfies initial dynamical isometry and outperforms more complex approaches. Although much simpler than its predecessors, this gate enables training thousands of fully connected layers with fast convergence and better test performance for ResNets trained on CIFAR-10. We apply this technique to language modeling and find that we can easily train 120-layer Transformers. When applied to 12 layer Transformers, it converges 56% faster on enwiki8."
                },
                {
                    "title": "Batch Normalization Biases Residual Blocks Towards the Identity Function in Deep Networks",
                    "abstract": "Batch normalization dramatically increases the largest trainable depth of residual networks, and this benefit has been crucial to the empirical success of deep residual networks on a wide range of benchmarks. We show that this key benefit arises because, at initialization, batch normalization downscales the residual branch relative to the skip connection, by a normalizing factor on the order of the square root of the network depth. This ensures that, early in training, the function computed by normalized residual blocks in deep networks is close to the identity function (on average). We use this insight to develop a simple initialization scheme that can train deep residual networks without normalization. We also provide a detailed empirical study of residual networks, which clarifies that, although batch normalized networks can be trained with larger learning rates, this effect is only beneficial in specific compute regimes, and has minimal benefits when the batch size is small."
                },
                {
                    "title": "On Layer Normalization in the Transformer Architecture",
                    "abstract": "The Transformer is widely used in natural language processing tasks. To train a Transformer however, one usually needs a carefully designed learning rate warm-up stage, which is shown to be crucial to the final performance but will slow down the optimization and bring more hyperparameter tunings. In this paper, we first study theoretically why the learning rate warm-up stage is essential and show that the location of layer normalization matters. Specifically, we prove with mean field theory that at initialization, for the original-designed Post-LN Transformer, which places the layer normalization between the residual blocks, the expected gradients of the parameters near the output layer are large. Therefore, using a large learning rate on those gradients makes the training unstable. The warm-up stage is practically helpful for avoiding this problem. On the other hand, our theory also shows that if the layer normalization is put inside the residual blocks (recently proposed as Pre-LN Transformer), the gradients are well-behaved at initialization. This motivates us to remove the warm-up stage for the training of Pre-LN Transformers. We show in our experiments that Pre-LN Transformers without the warm-up stage can reach comparable results with baselines while requiring significantly less training time and hyper-parameter tuning on a wide range of applications."
                },
                {
                    "title": "Provable Benefit of Orthogonal Initialization in Optimizing Deep Linear Networks",
                    "abstract": "The selection of initial parameter values for gradient-based optimization of deep neural networks is one of the most impactful hyperparameter choices in deep learning systems, affecting both convergence times and model performance. Yet despite significant empirical and theoretical analysis, relatively little has been proved about the concrete effects of different initialization schemes. In this work, we analyze the effect of initialization in deep linear networks, and provide for the first time a rigorous proof that drawing the initial weights from the orthogonal group speeds up convergence relative to the standard Gaussian initialization with iid weights. We show that for deep networks, the width needed for efficient convergence to a global minimum with orthogonal initializations is independent of the depth, whereas the width needed for efficient convergence with Gaussian initializations scales linearly in the depth. Our results demonstrate how the benefits of a good initialization can persist throughout learning, suggesting an explanation for the recent empirical successes found by initializing very deep non-linear networks according to the principle of dynamical isometry."
                },
                {
                    "title": "Disentangling Trainability and Generalization in Deep Neural Networks",
                    "abstract": "A longstanding goal in the theory of deep learning is to characterize the conditions under which a given neural network architecture will be trainable, and if so, how well it might generalize to unseen data. In this work, we provide such a characterization in the limit of very wide and very deep networks, for which the analysis simplifies considerably. For wide networks, the trajectory under gradient descent is governed by the Neural Tangent Kernel (NTK), and for deep networks the NTK itself maintains only weak data dependence. By analyzing the spectrum of the NTK, we formulate necessary conditions for trainability and generalization across a range of architectures, including Fully Connected Networks (FCNs) and Convolutional Neural Networks (CNNs). We identify large regions of hyperparameter space for which networks can memorize the training set but completely fail to generalize. We find that CNNs without global average pooling behave almost identically to FCNs, but that CNNs with pooling have markedly different and often better generalization performance. These theoretical results are corroborated experimentally on CIFAR10 for a variety of network architectures and we include a colab notebook that reproduces the essential results of the paper."
                },
                {
                    "title": "MUSE: Parallel Multi-Scale Attention for Sequence to Sequence Learning",
                    "abstract": "In sequence to sequence learning, the self-attention mechanism proves to be highly effective, and achieves significant improvements in many tasks. However, the self-attention mechanism is not without its own flaws. Although self-attention can model extremely long dependencies, the attention in deep layers tends to overconcentrate on a single token, leading to insufficient use of local information and difficultly in representing long sequences. In this work, we explore parallel multi-scale representation learning on sequence data, striving to capture both long-range and short-range language structures. To this end, we propose the Parallel MUlti-Scale attEntion (MUSE) and MUSE-simple. MUSE-simple contains the basic idea of parallel multi-scale sequence representation learning, and it encodes the sequence in parallel, in terms of different scales with the help from self-attention, and pointwise transformation. MUSE builds on MUSE-simple and explores combining convolution and self-attention for learning sequence representations from more different scales. We focus on machine translation and the proposed approach achieves substantial performance improvements over Transformer, especially on long sequences. More importantly, we find that although conceptually simple, its success in practice requires intricate considerations, and the multi-scale attention must build on unified semantic space. Under common setting, the proposed model achieves substantial performance and outperforms all previous models on three main machine translation tasks. In addition, MUSE has potential for accelerating inference due to its parallelism. Code will be available at this https URL"
                },
                {
                    "title": "Lipschitz Constrained Parameter Initialization for Deep Transformers",
                    "abstract": "The Transformer translation model employs residual connection and layer normalization to ease the optimization difficulties caused by its multi-layer encoder/decoder structure. Previous research shows that even with residual connection and layer normalization, deep Transformers still have difficulty in training, and particularly Transformer models with more than 12 encoder/decoder layers fail to converge. In this paper, we first empirically demonstrate that a simple modification made in the official implementation, which changes the computation order of residual connection and layer normalization, can significantly ease the optimization of deep Transformers. We then compare the subtle differences in computation order in considerable detail, and present a parameter initialization method that leverages the Lipschitz constraint on the initialization of Transformer parameters that effectively ensures training convergence. In contrast to findings in previous research we further demonstrate that with Lipschitz parameter initialization, deep Transformers with the original computation order can converge, and obtain significant BLEU improvements with up to 24 layers. In contrast to previous research which focuses on deep encoders, our approach additionally enables Transformers to also benefit from deep decoders."
                },
                {
                    "title": "Tensor Programs I: Wide Feedforward or Recurrent Neural Networks of Any Architecture are Gaussian Processes",
                    "abstract": "Wide neural networks with random weights and biases are Gaussian processes, as observed by Neal (1995) for shallow networks, and more recently by Lee et al.~(2018) and Matthews et al.~(2018) for deep fully-connected networks, as well as by Novak et al.~(2019) and Garriga-Alonso et al.~(2019) for deep convolutional networks. We show that this Neural Network-Gaussian Process correspondence surprisingly extends to all modern feedforward or recurrent neural networks composed of multilayer perceptron, RNNs (e.g. LSTMs, GRUs), (nD or graph) convolution, pooling, skip connection, attention, batch normalization, and/or layer normalization. More generally, we introduce a language for expressing neural network computations, and our result encompasses all such expressible neural networks. This work serves as a tutorial on the \\emph{tensor programs} technique formulated in Yang (2019) and elucidates the Gaussian Process results obtained there. We provide open-source implementations of the Gaussian Process kernels of simple RNN, GRU, transformer, and batchnorm+ReLU network at github.com/thegregyang/GP4A. Please see our arxiv version for the complete and up-to-date version of this paper."
                },
                {
                    "title": "Root Mean Square Layer Normalization",
                    "abstract": "Layer normalization (LayerNorm) has been successfully applied to various deep neural networks to help stabilize training and boost model convergence because of its capability in handling re-centering and re-scaling of both inputs and weight matrix. However, the computational overhead introduced by LayerNorm makes these improvements expensive and significantly slows the underlying network, e.g. RNN in particular. In this paper, we hypothesize that re-centering invariance in LayerNorm is dispensable and propose root mean square layer normalization, or RMSNorm. RMSNorm regularizes the summed inputs to a neuron in one layer according to root mean square (RMS), giving the model re-scaling invariance property and implicit learning rate adaptation ability. RMSNorm is computationally simpler and thus more efficient than LayerNorm. We also present partial RMSNorm, or pRMSNorm where the RMS is estimated from p% of the summed inputs without breaking the above properties. Extensive experiments on several tasks using diverse network architectures show that RMSNorm achieves comparable performance against LayerNorm but reduces the running time by 7%~64% on different models. Source code is available at https://github.com/bzhangGo/rmsnorm."
                },
                {
                    "title": "Improving Deep Transformer with Depth-Scaled Initialization and Merged Attention",
                    "abstract": "The general trend in NLP is towards increasing model capacity and performance via deeper neural networks. However, simply stacking more layers of the popular Transformer architecture for machine translation results in poor convergence and high computational overhead. Our empirical analysis suggests that convergence is poor due to gradient vanishing caused by the interaction between residual connection and layer normalization. We propose depth-scaled initialization (DS-Init), which decreases parameter variance at the initialization stage, and reduces output variance of residual connections so as to ease gradient back-propagation through normalization layers. To address computational cost, we propose a merged attention sublayer (MAtt) which combines a simplified average-based self-attention sublayer and the encoder-decoder attention sublayer on the decoder side. Results on WMT and IWSLT translation tasks with five translation directions show that deep Transformers with DS-Init and MAtt can substantially outperform their base counterpart in terms of BLEU (+1.1 BLEU on average for 12-layer models), while matching the decoding speed of the baseline model thanks to the efficiency improvements of MAtt. Source code for reproduction will be released soon."
                },
                {
                    "title": "Augmenting Self-attention with Persistent Memory",
                    "abstract": "Transformer networks have lead to important progress in language modeling and machine translation. These models include two consecutive modules, a feed-forward layer and a self-attention layer. The latter allows the network to capture long term dependencies and are often regarded as the key ingredient in the success of Transformers. Building upon this intuition, we propose a new model that solely consists of attention layers. More precisely, we augment the self-attention layers with persistent memory vectors that play a similar role as the feed-forward layer. Thanks to these vectors, we can remove the feed-forward layer without degrading the performance of a transformer. Our evaluation shows the benefits brought by our model on standard character and word level language modeling benchmarks."
                },
                {
                    "title": "How to Initialize your Network? Robust Initialization for WeightNorm & ResNets",
                    "abstract": "Residual networks (ResNet) and weight normalization play an important role in various deep learning applications. However, parameter initialization strategies have not been studied previously for weight normalized networks and, in practice, initialization methods designed for un-normalized networks are used as a proxy. Similarly, initialization for ResNets have also been studied for un-normalized networks and often under simplified settings ignoring the shortcut connection. To address these issues, we propose a novel parameter initialization strategy that avoids explosion/vanishment of information across layers for weight normalized networks with and without residual connections. The proposed strategy is based on a theoretical analysis using mean field approximation. We run over 2,500 experiments and evaluate our proposal on image datasets showing that the proposed initialization outperforms existing initialization methods in terms of generalization performance, robustness to hyper-parameter values and variance between seeds, especially when networks get deeper in which case existing methods fail to even start training. Finally, we show that using our initialization in conjunction with learning rate warmup is able to reduce the gap between the performance of weight normalized and batch normalized networks."
                },
                {
                    "title": "A Mean Field Theory of Batch Normalization",
                    "abstract": "We develop a mean field theory for batch normalization in fully-connected feedforward neural networks. In so doing, we provide a precise characterization of signal propagation and gradient backpropagation in wide batch-normalized networks at initialization. Our theory shows that gradient signals grow exponentially in depth and that these exploding gradients cannot be eliminated by tuning the initial weight variances or by adjusting the nonlinear activation function. Indeed, batch normalization itself is the cause of gradient explosion. As a result, vanilla batch-normalized networks without skip connections are not trainable at large depths for common initialization schemes, a prediction that we verify with a variety of empirical simulations. While gradient explosion cannot be eliminated, it can be reduced by tuning the network close to the linear regime, which improves the trainability of deep batch-normalized networks without residual connections. Finally, we investigate the learning dynamics of batch-normalized networks and observe that after a single step of optimization the networks achieve a relatively stable equilibrium in which gradients have dramatically smaller dynamic range. Our theory leverages Laplace, Fourier, and Gegenbauer transforms and we derive new identities that may be of independent interest."
                },
                {
                    "title": "On the Impact of the Activation Function on Deep Neural Networks Training",
                    "abstract": "The weight initialization and the activation function of deep neural networks have a crucial impact on the performance of the training procedure. An inappropriate selection can lead to the loss of information of the input during forward propagation and the exponential vanishing/exploding of gradients during back-propagation. Understanding the theoretical properties of untrained random networks is key to identifying which deep networks may be trained successfully as recently demonstrated by Samuel et al (2017) who showed that for deep feedforward neural networks only a specific choice of hyperparameters known as the `Edge of Chaos' can lead to good performance. While the work by Samuel et al (2017) discuss trainability issues, we focus here on training acceleration and overall performance. We give a comprehensive theoretical analysis of the Edge of Chaos and show that we can indeed tune the initialization parameters and the activation function in order to accelerate the training and improve the performance."
                },
                {
                    "title": "Fixup Initialization: Residual Learning Without Normalization",
                    "abstract": "Normalization layers are a staple in state-of-the-art deep neural network architectures. They are widely believed to stabilize training, enable higher learning rate, accelerate convergence and improve generalization, though the reason for their effectiveness is still an active research topic. In this work, we challenge the commonly-held beliefs by showing that none of the perceived benefits is unique to normalization. Specifically, we propose fixed-update initialization (Fixup), an initialization motivated by solving the exploding and vanishing gradient problem at the beginning of training via properly rescaling a standard initialization. We find training residual networks with Fixup to be as stable as training with normalization -- even for networks with 10,000 layers. Furthermore, with proper regularization, Fixup enables residual networks without normalization to achieve state-of-the-art performance in image classification and machine translation."
                },
                {
                    "title": "Dynamical Isometry is Achieved in Residual Networks in a Universal Way for any Activation Function",
                    "abstract": "We demonstrate that in residual neural networks (ResNets) dynamical isometry is achievable irrespective of the activation function used. We do that by deriving, with the help of Free Probability and Random Matrix Theories, a universal formula for the spectral density of the input-output Jacobian at initialization, in the large network width and depth limit. The resulting singular value spectrum depends on a single parameter, which we calculate for a variety of popular activation functions, by analyzing the signal propagation in the artificial neural network. We corroborate our results with numerical simulations of both random matrices and ResNets applied to the CIFAR-10 classification problem. Moreover, we study consequences of this universal behavior for the initial and late phases of the learning processes. \nWe conclude by drawing attention to the simple fact, that initialization acts as a confounding factor between the choice of activation function and the rate of learning. We propose that in ResNets this can be resolved based on our results by ensuring the same level of dynamical isometry at initialization."
                },
                {
                    "title": "Dynamical Isometry and a Mean Field Theory of CNNs: How to Train 10, 000-Layer Vanilla Convolutional Neural Networks",
                    "abstract": "In recent years, state-of-the-art methods in computer vision have utilized increasingly deep convolutional neural network architectures (CNNs), with some of the most successful models employing hundreds or even thousands of layers. A variety of pathologies such as vanishing/exploding gradients make training such deep networks challenging. While residual connections and batch normalization do enable training at these depths, it has remained unclear whether such specialized architecture designs are truly necessary to train deep CNNs. In this work, we demonstrate that it is possible to train vanilla CNNs with ten thousand layers or more simply by using an appropriate initialization scheme. We derive this initialization scheme theoretically by developing a mean field theory for signal propagation and by characterizing the conditions for dynamical isometry, the equilibration of singular values of the input-output Jacobian matrix. These conditions require that the convolution operator be an orthogonal transformation in the sense that it is norm-preserving. We present an algorithm for generating such random initial orthogonal convolution kernels and demonstrate empirically that they enable efficient training of extremely deep architectures."
                },
                {
                    "title": "GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding",
                    "abstract": "Human ability to understand language is general, flexible, and robust. In contrast, most NLU models above the word level are designed for a specific task and struggle with out-of-domain data. If we aspire to develop models with understanding beyond the detection of superficial correspondences between inputs and outputs, then it is critical to develop a unified model that can execute a range of linguistic tasks across different domains. To facilitate research in this direction, we present the General Language Understanding Evaluation (GLUE, gluebenchmark.com): a benchmark of nine diverse NLU tasks, an auxiliary dataset for probing models for understanding of specific linguistic phenomena, and an online platform for evaluating and comparing models. For some benchmark tasks, training data is plentiful, but for others it is limited or does not match the genre of the test set. GLUE thus favors models that can represent linguistic knowledge in a way that facilitates sample-efficient learning and effective knowledge-transfer across tasks. While none of the datasets in GLUE were created from scratch for the benchmark, four of them feature privately-held test data, which is used to ensure that the benchmark is used fairly. We evaluate baselines that use ELMo (Peters et al., 2018), a powerful transfer learning technique, as well as state-of-the-art sentence representation models. The best models still achieve fairly low absolute scores. Analysis with our diagnostic dataset yields similarly weak performance over all phenomena tested, with some exceptions."
                },
                {
                    "title": "How to Start Training: The Effect of Initialization and Architecture",
                    "abstract": "We investigate the effects of initialization and architecture on the start of training in deep ReLU nets. We identify two common failure modes for early training in which the mean and variance of activations are poorly behaved. For each failure mode, we give a rigorous proof of when it occurs at initialization and how to avoid it. The first failure mode, exploding/vanishing mean activation length, can be avoided by initializing weights from a symmetric distribution with variance 2/fan-in. The second failure mode, exponentially large variance of activation length, can be avoided by keeping constant the sum of the reciprocals of layer widths. We demonstrate empirically the effectiveness of our theoretical results in predicting when networks are able to start training. In particular, we note that many popular initializations fail our criteria, whereas correct initialization and architecture allows much deeper networks to be trained."
                },
                {
                    "title": "Gaussian Process Behaviour in Wide Deep Neural Networks",
                    "abstract": "Whilst deep neural networks have shown great empirical success, there is still much work to be done to understand their theoretical properties. In this paper, we study the relationship between random, wide, fully connected, feedforward networks with more than one hidden layer and Gaussian processes with a recursive kernel definition. We show that, under broad conditions, as we make the architecture increasingly wide, the implied random function converges in distribution to a Gaussian process, formalising and extending existing results by Neal (1996) to deep networks. To evaluate convergence rates empirically, we use maximum mean discrepancy. We then compare finite Bayesian deep networks from the literature to Gaussian processes in terms of the key predictive quantities of interest, finding that in some cases the agreement can be very close. We discuss the desirability of Gaussian process behaviour and review non-Gaussian alternative models from the literature."
                },
                {
                    "title": "Decoupled Weight Decay Regularization",
                    "abstract": "L$_2$ regularization and weight decay regularization are equivalent for standard stochastic gradient descent (when rescaled by the learning rate), but as we demonstrate this is \\emph{not} the case for adaptive gradient algorithms, such as Adam. While common implementations of these algorithms employ L$_2$ regularization (often calling it \"weight decay\" in what may be misleading due to the inequivalence we expose), we propose a simple modification to recover the original formulation of weight decay regularization by \\emph{decoupling} the weight decay from the optimization steps taken w.r.t. the loss function. We provide empirical evidence that our proposed modification (i) decouples the optimal choice of weight decay factor from the setting of the learning rate for both standard SGD and Adam and (ii) substantially improves Adam's generalization performance, allowing it to compete with SGD with momentum on image classification datasets (on which it was previously typically outperformed by the latter). Our proposed decoupled weight decay has already been adopted by many researchers, and the community has implemented it in TensorFlow and PyTorch; the complete source code for our experiments is available at this https URL"
                },
                {
                    "title": "Deep Neural Networks as Gaussian Processes",
                    "abstract": "It has long been known that a single-layer fully-connected neural network with an i.i.d. prior over its parameters is equivalent to a Gaussian process (GP), in the limit of infinite network width. This correspondence enables exact Bayesian inference for infinite width neural networks on regression tasks by means of evaluating the corresponding GP. Recently, kernel functions which mimic multi-layer random neural networks have been developed, but only outside of a Bayesian framework. As such, previous work has not identified that these kernels can be used as covariance functions for GPs and allow fully Bayesian prediction with a deep neural network. \nIn this work, we derive the exact equivalence between infinitely wide deep networks and GPs. We further develop a computationally efficient pipeline to compute the covariance function for these GPs. We then use the resulting GPs to perform Bayesian inference for wide deep neural networks on MNIST and CIFAR-10. We observe that trained neural network accuracy approaches that of the corresponding GP with increasing layer width, and that the GP uncertainty is strongly correlated with trained network prediction error. We further find that test performance increases as finite-width trained networks are made wider and more similar to a GP, and thus that GP predictions typically outperform those of finite-width networks. Finally we connect the performance of these GPs to the recent theory of signal propagation in random neural networks."
                },
                {
                    "title": "Attention is All you Need",
                    "abstract": "The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
                },
                {
                    "title": "The Shattered Gradients Problem: If resnets are the answer, then what is the question?",
                    "abstract": "A long-standing obstacle to progress in deep learning is the problem of vanishing and exploding gradients. Although, the problem has largely been overcome via carefully constructed initializations and batch normalization, architectures incorporating skip-connections such as highway and resnets perform much better than standard feedforward architectures despite well-chosen initialization and batch normalization. In this paper, we identify the shattered gradients problem. Specifically, we show that the correlation between gradients in standard feedforward networks decays exponentially with depth resulting in gradients that resemble white noise whereas, in contrast, the gradients in architectures with skip-connections are far more resistant to shattering, decaying sublinearly. Detailed empirical evidence is presented in support of the analysis, on both fully-connected networks and convnets. Finally, we present a new \"looks linear\" (LL) initialization that prevents shattering, with preliminary experiments showing the new initialization allows to train very deep networks without the addition of skip-connections."
                },
                {
                    "title": "Language Modeling with Gated Convolutional Networks",
                    "abstract": "The pre-dominant approach to language modeling to date is based on recurrent neural networks. Their success on this task is often linked to their ability to capture unbounded context. In this paper we develop a finite context approach through stacked convolutions, which can be more efficient since they allow parallelization over sequential tokens. We propose a novel simplified gating mechanism that outperforms Oord et al (2016) and investigate the impact of key architectural decisions. The proposed approach achieves state-of-the-art on the WikiText-103 benchmark, even though it features long-term dependencies, as well as competitive results on the Google Billion Words benchmark. Our model reduces the latency to score a sentence by an order of magnitude compared to a recurrent baseline. To our knowledge, this is the first time a non-recurrent approach is competitive with strong recurrent models on these large scale language tasks."
                },
                {
                    "title": "Deep Information Propagation",
                    "abstract": "We study the behavior of untrained neural networks whose weights and biases are randomly distributed using mean field theory. We show the existence of depth scales that naturally limit the maximum depth of signal propagation through these random networks. Our main practical result is to show that random networks may be trained precisely when information can travel through them. Thus, the depth scales that we identify provide bounds on how deep a network may be trained for a specific choice of hyperparameters. As a corollary to this, we argue that in networks at the edge of chaos, one of these depth scales diverges. Thus arbitrarily deep networks may be trained only sufficiently close to criticality. We show that the presence of dropout destroys the order-to-chaos critical point and therefore strongly limits the maximum trainable depth for random networks. Finally, we develop a mean field theory for backpropagation and we show that the ordered and chaotic phases correspond to regions of vanishing and exploding gradient respectively."
                },
                {
                    "title": "Layer Normalization",
                    "abstract": "Training state-of-the-art, deep neural networks is computationally expensive. One way to reduce the training time is to normalize the activities of the neurons. A recently introduced technique called batch normalization uses the distribution of the summed input to a neuron over a mini-batch of training cases to compute a mean and variance which are then used to normalize the summed input to that neuron on each training case. This significantly reduces the training time in feed-forward neural networks. However, the effect of batch normalization is dependent on the mini-batch size and it is not obvious how to apply it to recurrent neural networks. In this paper, we transpose batch normalization into layer normalization by computing the mean and variance used for normalization from all of the summed inputs to the neurons in a layer on a single training case. Like batch normalization, we also give each neuron its own adaptive bias and gain which are applied after the normalization but before the non-linearity. Unlike batch normalization, layer normalization performs exactly the same computation at training and test times. It is also straightforward to apply to recurrent neural networks by computing the normalization statistics separately at each time step. Layer normalization is very effective at stabilizing the hidden state dynamics in recurrent networks. Empirically, we show that layer normalization can substantially reduce the training time compared with previously published techniques."
                },
                {
                    "title": "Exponential expressivity in deep neural networks through transient chaos",
                    "abstract": "We combine Riemannian geometry with the mean field theory of high dimensional chaos to study the nature of signal propagation in generic, deep neural networks with random weights. Our results reveal an order-to-chaos expressivity phase transition, with networks in the chaotic phase computing nonlinear functions whose global curvature grows exponentially with depth but not width. We prove this generic class of deep random functions cannot be efficiently computed by any shallow network, going beyond prior work restricted to the analysis of single functions. Moreover, we formalize and quantitatively demonstrate the long conjectured idea that deep networks can disentangle highly curved manifolds in input space into flat manifolds in hidden space. Our theoretical analysis of the expressive power of deep networks broadly applies to arbitrary nonlinearities, and provides a quantitative underpinning for previously abstract notions about the geometry of deep functions."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "Optimizing Neural Networks with Kronecker-factored Approximate Curvature",
                    "abstract": "We propose an efficient method for approximating natural gradient descent in neural networks which we call Kronecker-Factored Approximate Curvature (K-FAC). K-FAC is based on an efficiently invertible approximation of a neural network's Fisher information matrix which is neither diagonal nor low-rank, and in some cases is completely non-sparse. It is derived by approximating various large blocks of the Fisher (corresponding to entire layers) as being the Kronecker product of two much smaller matrices. While only several times more expensive to compute than the plain stochastic gradient, the updates produced by K-FAC make much more progress optimizing the objective, which results in an algorithm that can be much faster than stochastic gradient descent with momentum in practice. And unlike some previously proposed approximate natural-gradient/Newton methods which use high-quality non-diagonal curvature matrices (such as Hessian-free optimization), K-FAC works very well in highly stochastic optimization regimes. This is because the cost of storing and inverting K-FAC's approximation to the curvature matrix does not depend on the amount of data used to estimate it, which is a feature typically associated only with diagonal or low-rank approximations to the curvature matrix."
                },
                {
                    "title": "Exact solutions to the nonlinear dynamics of learning in deep linear neural networks",
                    "abstract": "Despite the widespread practical success of deep learning methods, our theoretical understanding of the dynamics of learning in deep neural networks remains quite sparse. We attempt to bridge the gap between the theory and practice of deep learning by systematically analyzing learning dynamics for the restricted case of deep linear neural networks. Despite the linearity of their input-output map, such networks have nonlinear gradient descent dynamics on weights that change with the addition of each new hidden layer. We show that deep linear networks exhibit nonlinear learning phenomena similar to those seen in simulations of nonlinear networks, including long plateaus followed by rapid transitions to lower error solutions, and faster convergence from greedy unsupervised pretraining initial conditions than from random initial conditions. We provide an analytical description of these phenomena by finding new exact solutions to the nonlinear dynamics of deep learning. Our theoretical analysis also reveals the surprising finding that as the depth of a network approaches infinity, learning speed can nevertheless remain finite: for a special class of initial conditions on the weights, very deep networks incur only a finite, depth independent, delay in learning speed relative to shallow networks. We show that, under certain conditions on the training data, unsupervised pretraining can find this special class of initial conditions, while scaled random Gaussian initializations cannot. We further exhibit a new class of random orthogonal initial conditions on weights that, like unsupervised pre-training, enjoys depth independent learning times. We further show that these initial conditions also lead to faithful propagation of gradients even in deep nonlinear networks, as long as they operate in a special regime known as the edge of chaos."
                },
                {
                    "title": "Catformer: Designing Stable Transformers via Sensitivity Analysis",
                    "abstract": "Transformer architectures are widely used, but training them is non-trivial, requiring custom learning rate schedules, scaling terms, residual connections, careful placement of submodules such as normalization, and so on. In this paper, we improve upon recent analysis of Transformers and formalize a notion of sensitivity to capture the difficulty of training. Sensitivity characterizes how the variance of activation and gradient norms change in expectation when parameters are randomly perturbed. We analyze the sensitivity of previous Transformer architectures and design a new architecture, the Catformer, which replaces residual connections or RNN-based gating mechanisms with concatenation. We prove that Catformers are less sensitive than other Transformer variants and demonstrate that this leads to more stable training. On DMLab30, a suite of high-dimension reinforcement tasks, Catformer outperforms other transformers, including Gated Transformer-XL\u2014the state-of-the-art architecture designed to address stability\u2014by 13%."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Language Models are Unsupervised Multitask Learners",
                    "abstract": "Natural language processing tasks, such as question answering, machine translation, reading comprehension, and summarization, are typically approached with supervised learning on taskspecific datasets. We demonstrate that language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages called WebText. When conditioned on a document plus questions, the answers generated by the language model reach 55 F1 on the CoQA dataset matching or exceeding the performance of 3 out of 4 baseline systems without using the 127,000+ training examples. The capacity of the language model is essential to the success of zero-shot task transfer and increasing it improves performance in a log-linear fashion across tasks. Our largest model, GPT-2, is a 1.5B parameter Transformer that achieves state of the art results on 7 out of 8 tested language modeling datasets in a zero-shot setting but still underfits WebText. Samples from the model reflect these improvements and contain coherent paragraphs of text. These findings suggest a promising path towards building language processing systems which learn to perform tasks from their naturally occurring demonstrations."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan the standard transformer block be simplified by removing certain components without compromising training speed?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is significant for the research community as it addresses the complexity of modern neural network architectures, which often contain numerous components whose roles are not well understood. By simplifying transformer blocks, we can bridge the gap between theoretical understanding and practical application in deep learning. This research could lead to more efficient architectures, reducing training costs and improving deployment times, which is crucial given the increasing resource demands of large models. Furthermore, it may inspire future research into the fundamental mechanisms of deep learning, potentially leading to new insights and innovations in model design.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in understanding the intricate interactions between various components of the transformer block and their impact on training dynamics. Naive approaches may fail because they overlook the potential benefits of components like skip connections or normalization layers, which could be critical for maintaining training speed and model performance. Additionally, the theoretical frameworks currently available, such as signal propagation, do not fully capture the complexities of deep learning during training, focusing instead on initial conditions. This gap makes it difficult to predict the consequences of removing certain components without empirical validation.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has largely focused on the standard transformer architecture without probing the necessity of its components. The limitations of existing studies stem from a lack of empirical investigation into the effects of removing specific components, as well as the prevailing reliance on complex architectures that may not reflect the underlying principles of deep learning. Additionally, the theoretical frameworks have not evolved to address the nuances of training dynamics beyond initialization. Our approach differs by systematically testing the removal of components and empirically validating the effects on training speed and performance, thus providing a clearer understanding of their roles.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves systematically removing components from the standard transformer block, such as skip connections, value parameters, projection parameters, and sequential sub-blocks. We will conduct experiments using standard datasets to evaluate the impact of these modifications on training speed and model performance. The metrics for evaluation will include training time per update step and overall runtime efficiency. We expect to demonstrate that it is possible to simplify the transformer block while maintaining or even improving training speed, leading to a more efficient architecture with reduced parameter counts and"
            }
        },
        "author_data": {
            "5505cd91-e7ea-45d2-9578-bd0af1af6935": {
                "pk": "5505cd91-e7ea-45d2-9578-bd0af1af6935",
                "name": "Bobby He",
                "collaborators": [
                    "Thomas Hofmann",
                    "Lorenzo Noci",
                    "Daniele Paliotta",
                    "Imanol Schlag",
                    "Sidak Pal Singh",
                    "Bernhard Scholkopf",
                    "Yuhui Ding",
                    "Antonio Orvieto"
                ],
                "domain": [
                    "Neural Networks",
                    "Graph Neural Network",
                    "Optimization",
                    "Transformer Models"
                ],
                "publications": [
                    {
                        "title": "Understanding and Minimising Outlier Features in Neural Network Training",
                        "abstract": "Outlier Features (OF) are neurons whose activation magnitudes significantly exceed the average over a neural network's (NN) width. They are well known to emerge during standard transformer training and have the undesirable effect of hindering quantisation in afflicted models. Despite their practical importance, little is known behind why OFs emerge during training, nor how one can minimise them. Our work focuses on the above questions, first identifying several quantitative metrics, such as the kurtosis over neuron activation norms, to measure OFs. With these metrics, we study how architectural and optimisation choices influence OFs, and provide practical insights to minimise OFs during training. As highlights, we emphasise the importance of controlling signal propagation throughout training, and propose the Outlier Protected transformer block, which removes standard Pre-Norm layers to mitigate OFs, without loss of convergence speed or training stability. Overall, our findings shed new light on our understanding of, our ability to prevent, and the complexity of this important facet in NN training dynamics."
                    },
                    {
                        "title": "Hallmarks of Optimization Trajectories in Neural Networks: Directional Exploration and Redundancy",
                        "abstract": "We propose a fresh take on understanding the mechanisms of neural networks by analyzing the rich directional structure of optimization trajectories, represented by their pointwise parameters. Towards this end, we introduce some natural notions of the complexity of optimization trajectories, both qualitative and quantitative, which hallmark the directional nature of optimization in neural networks: when is there redundancy, and when exploration. We use them to reveal the inherent nuance and interplay involved between various optimization choices, such as momentum and weight decay. Further, the trajectory perspective helps us see the effect of scale on regularizing the directional nature of trajectories, and as a by-product, we also observe an intriguing heterogeneity of Q,K,V dynamics in the middle attention layers in LLMs and which is homogenized by scale. Importantly, we put the significant directional redundancy observed to the test by demonstrating that training only scalar batchnorm parameters some while into training matches the performance of training the entire network, which thus exhibits the potential of hybrid optimization schemes that are geared towards efficiency."
                    },
                    {
                        "title": "Recurrent Distance Filtering for Graph Representation Learning",
                        "abstract": "Graph neural networks based on iterative one-hop message passing have been shown to struggle in harnessing the information from distant nodes effectively. Conversely, graph transformers allow each node to attend to all other nodes directly, but lack graph inductive bias and have to rely on ad-hoc positional encoding. In this paper, we propose a new architecture to reconcile these challenges. Our approach stems from the recent breakthroughs in long-range modeling provided by deep state-space models: for a given target node, our model aggregates other nodes by their shortest distances to the target and uses a linear RNN to encode the sequence of hop representations. The linear RNN is parameterized in a particular diagonal form for stable long-range signal propagation and is theoretically expressive enough to encode the neighborhood hierarchy. With no need for positional encoding, we empirically show that the performance of our model is comparable to or better than that of state-of-the-art graph transformers on various benchmarks, with a significantly reduced computational cost. Our code is open-source at https://github.com/skeletondyh/GRED."
                    }
                ]
            },
            "9d447555-2756-4e66-9fee-8001c40829f1": {
                "pk": "9d447555-2756-4e66-9fee-8001c40829f1",
                "name": "Thomas Hofmann",
                "collaborators": [
                    "Bobby He",
                    "Lorenzo Noci",
                    "Antonio Orvieto",
                    "Daniele Paliotta",
                    "Imanol Schlag",
                    "Alexandru Meterez",
                    "Sidak Pal Singh",
                    "Bernhard Scholkopf",
                    "Yuhui Ding"
                ],
                "domain": [
                    "Neural Networks",
                    "Optimization",
                    "Graph Neural Network",
                    "Transformer Models"
                ],
                "publications": [
                    {
                        "title": "Understanding and Minimising Outlier Features in Neural Network Training",
                        "abstract": "Outlier Features (OF) are neurons whose activation magnitudes significantly exceed the average over a neural network's (NN) width. They are well known to emerge during standard transformer training and have the undesirable effect of hindering quantisation in afflicted models. Despite their practical importance, little is known behind why OFs emerge during training, nor how one can minimise them. Our work focuses on the above questions, first identifying several quantitative metrics, such as the kurtosis over neuron activation norms, to measure OFs. With these metrics, we study how architectural and optimisation choices influence OFs, and provide practical insights to minimise OFs during training. As highlights, we emphasise the importance of controlling signal propagation throughout training, and propose the Outlier Protected transformer block, which removes standard Pre-Norm layers to mitigate OFs, without loss of convergence speed or training stability. Overall, our findings shed new light on our understanding of, our ability to prevent, and the complexity of this important facet in NN training dynamics."
                    },
                    {
                        "title": "Why do Learning Rates Transfer? Reconciling Optimization and Scaling Limits for Deep Learning",
                        "abstract": "Recently, there has been growing evidence that if the width and depth of a neural network are scaled toward the so-called rich feature learning limit ($\\mu$P and its depth extension), then some hyperparameters - such as the learning rate - exhibit transfer from small to very large models, thus reducing the cost of hyperparameter tuning. From an optimization perspective, this phenomenon is puzzling, as it implies that the loss landscape is remarkably consistent across very different model sizes. In this work, we find empirical evidence that learning rate transfer can be attributed to the fact that under $\\mu$P and its depth extension, the largest eigenvalue of the training loss Hessian (i.e. the sharpness) is largely independent of the width and depth of the network for a sustained period of training time. On the other hand, we show that under the neural tangent kernel (NTK) regime, the sharpness exhibits very different dynamics at different scales, thus preventing learning rate transfer. But what causes these differences in the sharpness dynamics? Through a connection between the spectra of the Hessian and the NTK matrix, we argue that the cause lies in the presence (for $\\mu$P) or progressive absence (for the NTK regime) of feature learning, which results in a different evolution of the NTK, and thus of the sharpness. We corroborate our claims with a substantial suite of experiments, covering a wide range of datasets and architectures: from ResNets and Vision Transformers trained on benchmark vision datasets to Transformers-based language models trained on WikiText"
                    },
                    {
                        "title": "Hallmarks of Optimization Trajectories in Neural Networks: Directional Exploration and Redundancy",
                        "abstract": "We propose a fresh take on understanding the mechanisms of neural networks by analyzing the rich directional structure of optimization trajectories, represented by their pointwise parameters. Towards this end, we introduce some natural notions of the complexity of optimization trajectories, both qualitative and quantitative, which hallmark the directional nature of optimization in neural networks: when is there redundancy, and when exploration. We use them to reveal the inherent nuance and interplay involved between various optimization choices, such as momentum and weight decay. Further, the trajectory perspective helps us see the effect of scale on regularizing the directional nature of trajectories, and as a by-product, we also observe an intriguing heterogeneity of Q,K,V dynamics in the middle attention layers in LLMs and which is homogenized by scale. Importantly, we put the significant directional redundancy observed to the test by demonstrating that training only scalar batchnorm parameters some while into training matches the performance of training the entire network, which thus exhibits the potential of hybrid optimization schemes that are geared towards efficiency."
                    },
                    {
                        "title": "Recurrent Distance Filtering for Graph Representation Learning",
                        "abstract": "Graph neural networks based on iterative one-hop message passing have been shown to struggle in harnessing the information from distant nodes effectively. Conversely, graph transformers allow each node to attend to all other nodes directly, but lack graph inductive bias and have to rely on ad-hoc positional encoding. In this paper, we propose a new architecture to reconcile these challenges. Our approach stems from the recent breakthroughs in long-range modeling provided by deep state-space models: for a given target node, our model aggregates other nodes by their shortest distances to the target and uses a linear RNN to encode the sequence of hop representations. The linear RNN is parameterized in a particular diagonal form for stable long-range signal propagation and is theoretically expressive enough to encode the neighborhood hierarchy. With no need for positional encoding, we empirically show that the performance of our model is comparable to or better than that of state-of-the-art graph transformers on various benchmarks, with a significantly reduced computational cost. Our code is open-source at https://github.com/skeletondyh/GRED."
                    }
                ]
            }
        }
    },
    "2403.11857": {
        "paper_data": {
            "title": "Complete and Efficient Graph Transformers for Crystal Material Property Prediction",
            "url": "http://arxiv.org/abs/2403.11857v1",
            "arxiv_id": "2403.11857",
            "authors": [
                "Keqiang Yan",
                "Cong Fu",
                "Xiaofeng Qian",
                "Xiaoning Qian",
                "Shuiwang Ji"
            ],
            "abstract": "Crystal structures are characterized by atomic bases within a primitive unit cell that repeats along a regular lattice throughout 3D space. The periodic and infinite nature of crystals poses unique challenges for geometric graph representation learning. Specifically, constructing graphs that effectively capture the complete geometric information of crystals and handle chiral crystals remains an unsolved and challenging problem. In this paper, we introduce a novel approach that utilizes the periodic patterns of unit cells to establish the lattice-based representation for each atom, enabling efficient and expressive graph representations of crystals. Furthermore, we propose ComFormer, a SE(3) transformer designed specifically for crystalline materials. ComFormer includes two variants; namely, iComFormer that employs invariant geometric descriptors of Euclidean distances and angles, and eComFormer that utilizes equivariant vector representations. Experimental results demonstrate the state-of-the-art predictive accuracy of ComFormer variants on various tasks across three widely-used crystal benchmarks. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS).",
            "introduction": "   1 Introduction  Accelerating the discovery of novel materials with desirable properties has values in many science, engineering, and biomedical applications\u00a0(Ramprasad et\u00a0al., 2017; Meredig et\u00a0al., 2014; Oliynyk et\u00a0al., 2016; Raccuglia et\u00a0al., 2016; Ward et\u00a0al., 2016; Liu et\u00a0al., 2021b; c; Luo et\u00a0al., 2021; Xie et\u00a0al., 2022; Fu et\u00a0al., 2023; Wang et\u00a0al., ; Chanussot et\u00a0al., 2021; Zhang et\u00a0al., 2023). However, the current reliance on traditional, costly, and time-consuming trial-and-error experimental methods poses practical challenges. In this regard, computational approaches based on quantum mechanics, such as density functional theory (DFT), have made significant contributions for predicting the physical and chemical properties of materials to guide materials discovery experiments. Nevertheless, the high computational cost associated with these methods (ranging from n3superscript\ud835\udc5b3n^{3}italic_n start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT to n7superscript\ud835\udc5b7n^{7}italic_n start_POSTSUPERSCRIPT 7 end_POSTSUPERSCRIPT, where n\ud835\udc5bnitalic_n represents the number of atoms) still poses a hurdle for novel materials discovery in the vast materials space.   To reduce computational cost, recent studies employ machine learning\u00a0(ML) models \u00a0(Yu et\u00a0al., 2022; Ying et\u00a0al., 2021; Gao & Ji, 2019; Liu et\u00a0al., 2021a; Fu et\u00a0al., 2022; Passaro & Zitnick, 2023; Gasteiger et\u00a0al., 2020; Liao & Smidt, 2022; Batzner et\u00a0al., 2022; Satorras et\u00a0al., 2021; Qiao et\u00a0al., 2022; Watson et\u00a0al., 2022; Ahmad et\u00a0al., 2018), such as crystal graph representations\u00a0(Xie & Grossman, 2018; Ling et\u00a0al., 2023) with deep neural networks. However, these crystal graph representations have limitations in distinguishing different crystalline materials. In other words, they cannot guarantee to capture the complete geometric information of input crystal structures, and may map different crystal structures with different properties to the same graph representation and produce the identical property predictions. Illustrative examples and detailed discussions can be found in Appendix\u00a0A.1.   While graph representations that can capture any structural differences in small molecules have been investigated in previous works\u00a0(Wang et\u00a0al., 2022; Klicpera et\u00a0al., 2021), these methods fail short to capture periodic patterns of crystals and cannot maintain geometric completeness for crystals. Thus, how to construct graph representations for crystalline materials that can distinguish any different crystals and remain the same under crystal passive symmetries detailed in Sec.\u00a03.1 remains unsolved.   Establishing invariant\u00a0(Anosova & Kurlin, 2022; Edelsbrunner et\u00a0al., 2021) and complete representations have also been investigated for other applications, especially representations of matrix forms or consisting of aligned coordinates that can be used to measure the structural differences between crystals. Among them, AMD\u00a0(Widdowson et\u00a0al., 2022) and PDD\u00a0(Widdowson & Kurlin, 2022) are two recent works that satisfy continuity under small perturbations of atom positions by using Euclidean distances. However, AMD and PDD cannot distinguish chiral crystals and cannot be directly used to predict crystal properties without breaking the completeness.   In this work, we focus on crystal property prediction and propose SE(3) invariant and SO(3) equivariant crystal graph representations to strive for geometric completeness for crystalline materials. Based on these expressive crystal graphs, we develop ComFormer with two variants; namely iComFormer that employs SE(3) invariant crystal graphs, and eComFormer that uses SO(3) equivariant crystal graphs. Both variants can scale to large-scale crystal datasets with complexity O\u2062(n\u2062k)\ud835\udc42\ud835\udc5b\ud835\udc58O(nk)italic_O ( italic_n italic_k ) where n\ud835\udc5bnitalic_n denotes the number of atoms in the crystal unit cell and k\ud835\udc58kitalic_k denotes the average number of neighbors. We evaluate the proposed ComFormer variants on three widely-used crystalline",
            "references": [
                {
                    "title": "Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems",
                    "abstract": "Advances in artificial intelligence (AI) are fueling a new paradigm of discoveries in natural sciences. Today, AI has started to advance natural sciences by improving, accelerating, and enabling our understanding of natural phenomena at a wide range of spatial and temporal scales, giving rise to a new area of research known as AI for science (AI4Science). Being an emerging research paradigm, AI4Science is unique in that it is an enormous and highly interdisciplinary area. Thus, a unified and technical treatment of this field is needed yet challenging. This work aims to provide a technically thorough account of a subarea of AI4Science; namely, AI for quantum, atomistic, and continuum systems. These areas aim at understanding the physical world from the subatomic (wavefunctions and electron density), atomic (molecules, proteins, materials, and interactions), to macro (fluids, climate, and subsurface) scales and form an important subarea of AI4Science. A unique advantage of focusing on these areas is that they largely share a common set of challenges, thereby allowing a unified and foundational treatment. A key common challenge is how to capture physics first principles, especially symmetries, in natural systems by deep learning methods. We provide an in-depth yet intuitive account of techniques to achieve equivariance to symmetry transformations. We also discuss other common technical challenges, including explainability, out-of-distribution generalization, knowledge transfer with foundation and large language models, and uncertainty quantification. To facilitate learning and education, we provide categorized lists of resources that we found to be useful. We strive to be thorough and unified and hope this initial effort may trigger more community interests and efforts to further advance AI4Science."
                },
                {
                    "title": "Efficient Approximations of Complete Interatomic Potentials for Crystal Property Prediction",
                    "abstract": "We study property prediction for crystal materials. A crystal structure consists of a minimal unit cell that is repeated infinitely in 3D space. How to accurately represent such repetitive structures in machine learning models remains unresolved. Current methods construct graphs by establishing edges only between nearby nodes, thereby failing to faithfully capture infinite repeating patterns and distant interatomic interactions. In this work, we propose several innovations to overcome these limitations. First, we propose to model physics-principled interatomic potentials directly instead of only using distances as in many existing methods. These potentials include the Coulomb potential, London dispersion potential, and Pauli repulsion potential. Second, we model the complete set of potentials among all atoms, instead of only between nearby atoms as in existing methods. This is enabled by our approximations of infinite potential summations with provable error bounds. We further develop efficient algorithms to compute the approximations. Finally, we propose to incorporate our computations of complete interatomic potentials into message passing neural networks for representation learning. We perform experiments on the JARVIS and Materials Project benchmarks for evaluation. Results show that the use of interatomic potentials and complete interatomic potentials leads to consistent performance improvements with reasonable computational costs. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS/tree/main/OpenMat/PotNet)."
                },
                {
                    "title": "A Latent Diffusion Model for Protein Structure Generation",
                    "abstract": "Proteins are complex biomolecules that perform a variety of crucial functions within living organisms. Designing and generating novel proteins can pave the way for many future synthetic biology applications, including drug discovery. However, it remains a challenging computational task due to the large modeling space of protein structures. In this study, we propose a latent diffusion model that can reduce the complexity of protein modeling while flexibly capturing the distribution of natural protein structures in a condensed latent space. Specifically, we propose an equivariant protein autoencoder that embeds proteins into a latent space and then uses an equivariant diffusion model to learn the distribution of the latent protein representations. Experimental results demonstrate that our method can effectively generate novel protein backbone structures with high designability and efficiency. The code will be made publicly available at https://github.com/divelab/AIRS/tree/main/OpenProt/LatentDiff"
                },
                {
                    "title": "Connectivity Optimized Nested Graph Networks for Crystal Structures",
                    "abstract": "Graph neural networks (GNNs) have been applied to a large variety of applications in materials science and chemistry. Here, we recapitulate the graph construction for crystalline (periodic) materials and investigate its impact on the GNNs model performance. We suggest the asymmetric unit cell as a representation to reduce the number of atoms by using all symmetries of the system. This substantially reduced the computational cost and thus time needed to train large graph neural networks without any loss in accuracy. Furthermore, with a simple but systematically built GNN architecture based on message passing and line graph templates, we introduce a general architecture (Nested Graph Network, NGN) that is applicable to a wide range of tasks. We show that our suggested models systematically improve state-of-the-art results across all tasks within the MatBench benchmark. Further analysis shows that optimized connectivity and deeper message functions are responsible for the improvement. Asymmetric unit cells and connectivity optimization can be generally applied to (crystal) graph networks, while our suggested nested graph framework will open new ways of systematic comparison of GNN architectures."
                },
                {
                    "title": "Reducing SO(3) Convolutions to SO(2) for Efficient Equivariant GNNs",
                    "abstract": "Graph neural networks that model 3D data, such as point clouds or atoms, are typically desired to be $SO(3)$ equivariant, i.e., equivariant to 3D rotations. Unfortunately equivariant convolutions, which are a fundamental operation for equivariant networks, increase significantly in computational complexity as higher-order tensors are used. In this paper, we address this issue by reducing the $SO(3)$ convolutions or tensor products to mathematically equivalent convolutions in $SO(2)$ . This is accomplished by aligning the node embeddings' primary axis with the edge vectors, which sparsifies the tensor product and reduces the computational complexity from $O(L^6)$ to $O(L^3)$, where $L$ is the degree of the representation. We demonstrate the potential implications of this improvement by proposing the Equivariant Spherical Channel Network (eSCN), a graph neural network utilizing our novel approach to equivariant convolutions, which achieves state-of-the-art results on the large-scale OC-20 and OC-22 datasets."
                },
                {
                    "title": "Broadly applicable and accurate protein design by integrating structure prediction networks and diffusion generative models",
                    "abstract": "There has been considerable recent progress in designing new proteins using deep learning methods1\u20139. Despite this progress, a general deep learning framework for protein design that enables solution of a wide range of design challenges, including de novo binder design and design of higher order symmetric architectures, has yet to be described. Diffusion models10,11 have had considerable success in image and language generative modeling but limited success when applied to protein modeling, likely due to the complexity of protein backbone geometry and sequence-structure relationships. Here we show that by fine tuning the RoseTTAFold structure prediction network on protein structure denoising tasks, we obtain a generative model of protein backbones that achieves outstanding performance on unconditional and topology-constrained protein monomer design, protein binder design, symmetric oligomer design, enzyme active site scaffolding, and symmetric motif scaffolding for therapeutic and metal-binding protein design. We demonstrate the power and generality of the method, called RoseTTAFold Diffusion (RFdiffusion), by experimentally characterizing the structures and functions of hundreds of new designs. In a manner analogous to networks which produce images from user-specified inputs, RFdiffusion enables the design of diverse, complex, functional proteins from simple molecular specifications."
                },
                {
                    "title": "Periodic Graph Transformers for Crystal Material Property Prediction",
                    "abstract": "We consider representation learning on periodic graphs encoding crystal materials. Different from regular graphs, periodic graphs consist of a minimum unit cell repeating itself on a regular lattice in 3D space. How to effectively encode these periodic structures poses unique challenges not present in regular graph representation learning. In addition to being E(3) invariant, periodic graph representations need to be periodic invariant. That is, the learned representations should be invariant to shifts of cell boundaries as they are artificially imposed. Furthermore, the periodic repeating patterns need to be captured explicitly as lattices of different sizes and orientations may correspond to different materials. In this work, we propose a transformer architecture, known as Matformer, for periodic graph representation learning. Our Matformer is designed to be invariant to periodicity and can capture repeating patterns explicitly. In particular, Matformer encodes periodic patterns by efficient use of geometric distances between the same atoms in neighboring cells. Experimental results on multiple common benchmark datasets show that our Matformer outperforms baseline methods consistently. In addition, our results demonstrate the importance of periodic invariance and explicit repeating pattern encoding for crystal representation learning."
                },
                {
                    "title": "Learning Hierarchical Protein Representations via Complete 3D Graph Networks",
                    "abstract": "We consider representation learning for proteins with 3D structures. We build 3D graphs based on protein structures and develop graph networks to learn their representations. Depending on the levels of details that we wish to capture, protein representations can be computed at different levels, \\emph{e.g.}, the amino acid, backbone, or all-atom levels. Importantly, there exist hierarchical relations among different levels. In this work, we propose to develop a novel hierarchical graph network, known as ProNet, to capture the relations. Our ProNet is very flexible and can be used to compute protein representations at different levels of granularity. By treating each amino acid as a node in graph modeling as well as harnessing the inherent hierarchies, our ProNet is more effective and efficient than existing methods. We also show that, given a base 3D graph network that is complete, our ProNet representations are also complete at all levels. Experimental results show that ProNet outperforms recent methods on most datasets. In addition, results indicate that different downstream tasks may require representations at different levels. Our code is publicly available as part of the DIG library (\\url{https://github.com/divelab/DIG})."
                },
                {
                    "title": "e3nn: Euclidean Neural Networks",
                    "abstract": "We present e3nn, a generalized framework for creating E(3) equivariant trainable functions, also known as Euclidean neural networks. e3nn naturally operates on geometry and geometric tensors that describe systems in 3D and transform predictably under a change of coordinate system. The core of e3nn are equivariant operations such as the TensorProduct class or the spherical harmonics functions that can be composed to create more complex modules such as convolutions and attention mechanisms. These core operations of e3nn can be used to efficiently articulate Tensor Field Networks, 3D Steerable CNNs, Clebsch-Gordan Networks, SE(3) Transformers and other E(3) equivariant networks."
                },
                {
                    "title": "Equiformer: Equivariant Graph Attention Transformer for 3D Atomistic Graphs",
                    "abstract": "Despite their widespread success in various domains, Transformer networks have yet to perform well across datasets in the domain of 3D atomistic graphs such as molecules even when 3D-related inductive biases like translational invariance and rotational equivariance are considered. In this paper, we demonstrate that Transformers can generalize well to 3D atomistic graphs and present Equiformer, a graph neural network leveraging the strength of Transformer architectures and incorporating SE(3)/E(3)-equivariant features based on irreducible representations (irreps). First, we propose a simple and effective architecture by only replacing original operations in Transformers with their equivariant counterparts and including tensor products. Using equivariant operations enables encoding equivariant information in channels of irreps features without complicating graph structures. With minimal modifications to Transformers, this architecture has already achieved strong empirical results. Second, we propose a novel attention mechanism called equivariant graph attention, which improves upon typical attention in Transformers through replacing dot product attention with multi-layer perceptron attention and including non-linear message passing. With these two innovations, Equiformer achieves competitive results to previous models on QM9, MD17 and OC20 datasets."
                },
                {
                    "title": "ComENet: Towards Complete and Efficient Message Passing for 3D Molecular Graphs",
                    "abstract": "Many real-world data can be modeled as 3D graphs, but learning representations that incorporates 3D information completely and efficiently is challenging. Existing methods either use partial 3D information, or suffer from excessive computational cost. To incorporate 3D information completely and efficiently, we propose a novel message passing scheme that operates within 1-hop neighborhood. Our method guarantees full completeness of 3D information on 3D graphs by achieving global and local completeness. Notably, we propose the important rotation angles to fulfill global completeness. Additionally, we show that our method is orders of magnitude faster than prior methods. We provide rigorous proof of completeness and analysis of time complexity for our methods. As molecules are in essence quantum systems, we build the \\underline{com}plete and \\underline{e}fficient graph neural network (ComENet) by combing quantum inspired basis functions and the proposed message passing scheme. Experimental results demonstrate the capability and efficiency of ComENet, especially on real-world datasets that are large in both numbers and sizes of graphs. Our code is publicly available as part of the DIG library (\\url{https://github.com/divelab/DIG})."
                },
                {
                    "title": "Lattice Convolutional Networks for Learning Ground States of Quantum Many-Body Systems",
                    "abstract": "Deep learning methods have been shown to be effective in representing ground-state wave functions of quantum many-body systems. Existing methods use convolutional neural networks (CNNs) for square lattices due to their image-like structures. For non-square lattices, existing method uses graph neural network (GNN) in which structure information is not precisely captured, thereby requiring additional hand-crafted sublattice encoding. In this work, we propose lattice convolutions in which a set of proposed operations are used to convert non-square lattices into grid-like augmented lattices on which regular convolution can be applied. Based on the proposed lattice convolutions, we design lattice convolutional networks (LCN) that use self-gating and attention mechanisms. Experimental results show that our method achieves performance on par or better than existing methods on spin 1/2 $J_1$-$J_2$ Heisenberg model over the square, honeycomb, triangular, and kagome lattices while without using hand-crafted encoding."
                },
                {
                    "title": "GraphFM: Improving Large-Scale GNN Training via Feature Momentum",
                    "abstract": "Training of graph neural networks (GNNs) for large-scale node classification is challenging. A key difficulty lies in obtaining accurate hidden node representations while avoiding the neighborhood explosion problem. Here, we propose a new technique, named feature momentum (FM), that uses a momentum step to incorporate historical embeddings when updating feature representations. We develop two specific algorithms, known as GraphFM-IB and GraphFM-OB, that consider in-batch and out-of-batch data, respectively. GraphFM-IB applies FM to in-batch sampled data, while GraphFM-OB applies FM to out-of-batch data that are 1-hop neighborhood of in-batch data. We provide a convergence analysis for GraphFM-IB and some theoretical insight for GraphFM-OB. Empirically, we observe that GraphFM-IB can effectively alleviate the neighborhood explosion problem of existing methods. In addition, GraphFM-OB achieves promising performance on multiple large-scale graph datasets."
                },
                {
                    "title": "Recognition of near-duplicate periodic patterns by continuous metrics with approximation guarantees",
                    "abstract": "This paper proposes a rigorous solution to the challenging classification problem of periodic patterns under Euclidean isometry and rigid motion. The 3-dimensional case is practically important for solid crystalline materials (periodic crystals) including all medical drugs that are produced in a rigid tablet form. Many past invariants based on finite subsets fail when a unit cell of a periodic pattern discontinuously changes under almost any perturbation of atoms, which is inevitable due to noise and atomic vibrations. The major problem is not only to find complete invariants (descriptors with no false negatives and no false positives for all periodic patterns) but to design efficient algorithms for distance metrics on these invariants that should continuously behave under noise. The proposed continuous metrics solve this problem in any Euclidean dimension and are algorithmically approximated with small error factors in times that are explicitly bounded in the input size and complexity of a given crystal. The proved Lipschitz continuity allows us to confirm all near-duplicates filtered by simpler invariants in major databases of experimental and simulated crystals. This practical detection of near-duplicates will stop the artificial generation of `new' materials from slight perturbations of known crystals. Several such duplicates are under investigation by five journals for data integrity."
                },
                {
                    "title": "Average minimum distances of periodic point sets \u2013 foundational invariants for mapping periodic crystals",
                    "abstract": "The fundamental model of any solid crystalline material (crystal) at the atomic scale is a periodic point set. The strongest natural equivalence of crystals is rigid motion or isometry that preserves all inter-atomic distances. Past comparisons of periodic structures often used manual thresholds, symmetry groups and reduced cells, which are discontinuous under perturbations or thermal vibrations of atoms. This work defines the infinite sequence of continuous isometry invariants (Average Minimum Distances) to progressively capture distances between neighbors. The asymptotic behaviour of the new invariants is theoretically proved in all dimensions for a wide class of sets including non-periodic. The proposed near linear time algorithm identified all different crystals in the world's largest Cambridge Structural Database within a few hours on a modest desktop. The ultra fast speed and proved continuity provide rigorous foundations to continuously parameterise the space of all periodic crystals as a high-dimensional extension of Mendeleev's table of elements."
                },
                {
                    "title": "Crystal Diffusion Variational Autoencoder for Periodic Material Generation",
                    "abstract": "Generating the periodic structure of stable materials is a long-standing challenge for the material design community. This task is difficult because stable materials only exist in a low-dimensional subspace of all possible periodic arrangements of atoms: 1) the coordinates must lie in the local energy minimum defined by quantum mechanics, and 2) global stability also requires the structure to follow the complex, yet specific bonding preferences between different atom types. Existing methods fail to incorporate these factors and often lack proper invariances. We propose a Crystal Diffusion Variational Autoencoder (CDVAE) that captures the physical inductive bias of material stability. By learning from the data distribution of stable materials, the decoder generates materials in a diffusion process that moves atomic coordinates towards a lower energy state and updates atom types to satisfy bonding preferences between neighbors. Our model also explicitly encodes interactions across periodic boundaries and respects permutation, translation, rotation, and periodic invariances. We significantly outperform past methods in three tasks: 1) reconstructing the input structure, 2) generating valid, diverse, and realistic materials, and 3) generating materials that optimize a specific property. We also provide several standard datasets and evaluation metrics for the broader machine learning community."
                },
                {
                    "title": "Fast predictions of lattice energies by continuous isometry invariants of crystal structures",
                    "abstract": "Crystal Structure Prediction (CSP) aims to discover solid crystalline materials by optimizing periodic arrangements of atoms, ions or molecules. CSP takes weeks of supercomputer time because of slow energy minimizations for millions of simulated crystals. The lattice energy is a key physical property, which determines thermodynamic stability of a crystal but has no simple analytic expression. Past machine learning approaches to predict the lattice energy used slow crystal descriptors depending on manually chosen parameters. The new area of Periodic Geometry offers much faster isometry invariants that are also continuous under perturbations of atoms. Our experiments on simulated crystals confirm that a small distance between the new invariants guarantees a small difference of energies. We compare several kernel methods for invariant-based predictions of energy and achieve the mean absolute error of less than 5kJ/mole or 0.05eV/atom on a dataset of 5679 crystals."
                },
                {
                    "title": "Fast Quantum Property Prediction via Deeper 2D and 3D Graph Networks",
                    "abstract": "Molecular property prediction is gaining increasing attention due to its diverse applications. One task of particular interests and importance is to predict quantum chemical properties without 3D equilibrium structures. This is practically favorable since obtaining 3D equilibrium structures requires extremely expensive calculations. In this work, we design a deep graph neural network to predict quantum properties by directly learning from 2D molecular graphs. In addition, we propose a 3D graph neural network to learn from low-cost conformer sets, which can be obtained with open-source tools using an affordable budget. We employ our methods to participate in the 2021 KDD Cup on OGB Large-Scale Challenge (OGB-LSC), which aims to predict the HOMO-LUMO energy gap of molecules. Final evaluation results reveal that we are one of the winners with a mean absolute error of 0.1235 on the holdout test set. Our implementation is available as part of the MoleculeX package (https://github.com/divelab/MoleculeX)."
                },
                {
                    "title": "GemNet: Universal Directional Graph Neural Networks for Molecules",
                    "abstract": "Effectively predicting molecular interactions has the potential to accelerate molecular dynamics by multiple orders of magnitude and thus revolutionize chemical simulations. Graph neural networks (GNNs) have recently shown great successes for this task, overtaking classical methods based on fixed molecular kernels. However, they still appear very limited from a theoretical perspective, since regular GNNs cannot distinguish certain types of graphs. In this work we close this gap between theory and practice. We show that GNNs with spherical representations are indeed universal approximators for predictions that are invariant to translation, and equivariant to permutation and rotation. We then discretize such GNNs via directed edge embeddings and two-hop message passing, and incorporate multiple structural improvements to arrive at the geometric message passing neural network (GemNet). We demonstrate the benefits of the proposed changes in multiple ablation studies. GemNet outperforms previous models on the COLL, MD17, and OC20 datasets by 34%, 41%, and 20%, respectively, and performs especially well on the most challenging molecules. Our implementation is available online."
                },
                {
                    "title": "Informing geometric deep learning with electronic interactions to accelerate quantum chemistry",
                    "abstract": "Significance The interplay between molecular simulation and artificial intelligence has spurred many insights into chemical discovery, yet the data requirement of machine learning approaches remains a bottleneck. We tackle this challenge by developing a rigorous framework to conserve the symmetries of quantum chemical systems within deep neural networks. Our framework greatly improves the prediction of numerous molecular electronic and energetic properties, even for systems far from the training data. Through tight integration of quantum theory, geometry, and machine learning, our study offers a solution to vastly accelerate the in silico modelling of chemical phenomena in many domains, such as material design and catalysis."
                },
                {
                    "title": "The Density Fingerprint of a Periodic Point Set",
                    "abstract": "Modeling a crystal as a periodic point set, we present a fingerprint consisting of density functions that facilitates the efficient search for new materials and material properties. We prove invariance under isometries, continuity, and completeness in the generic case, which are necessary features for the reliable comparison of crystals. The proof of continuity integrates methods from discrete geometry and lattice theory, while the proof of generic completeness combines techniques from geometry with analysis. The fingerprint has a fast algorithm based on Brillouin zones and related inclusion-exclusion formulae. We have implemented the algorithm and describe its application to crystal structure prediction. 2012 ACM Subject Classification Theory of computation \u2192 Computational geometry"
                },
                {
                    "title": "DIG: A Turnkey Library for Diving into Graph Deep Learning Research",
                    "abstract": "Although there exist several libraries for deep learning on graphs, they are aiming at implementing basic operations for graph deep learning. In the research community, implementing and benchmarking various advanced tasks are still painful and time-consuming with existing libraries. To facilitate graph deep learning research, we introduce DIG: Dive into Graphs, a turnkey library that provides a unified testbed for higher level, research-oriented graph deep learning tasks. Currently, we consider graph generation, self-supervised learning on graphs, explainability of graph neural networks, and deep learning on 3D graphs. For each direction, we provide unified implementations of data interfaces, common algorithms, and evaluation metrics. Altogether, DIG is an extensible, open-source, and turnkey library for researchers to develop new methods and effortlessly compare with common baselines using widely used datasets and evaluation metrics. Source code is available at https://github.com/divelab/DIG."
                },
                {
                    "title": "E(n) Equivariant Graph Neural Networks",
                    "abstract": "This paper introduces a new model to learn graph neural networks equivariant to rotations, translations, reflections and permutations called E(n)-Equivariant Graph Neural Networks (EGNNs). In contrast with existing methods, our work does not require computationally expensive higher-order representations in intermediate layers while it still achieves competitive or better performance. In addition, whereas existing methods are limited to equivariance on 3 dimensional spaces, our model is easily scaled to higher-dimensional spaces. We demonstrate the effectiveness of our method on dynamical systems modelling, representation learning in graph autoencoders and predicting molecular properties."
                },
                {
                    "title": "Spherical Message Passing for 3D Molecular Graphs",
                    "abstract": "We consider representation learning of 3D molecular graphs in which each atom is associated with a spatial position in 3D. This is an under-explored area of research, and a principled message passing framework is currently lacking. In this work, we conduct analyses in the spherical coordinate system (SCS) for the complete identification of 3D graph structures. Based on such observations, we propose the spherical message passing (SMP) as a novel and powerful scheme for 3D molecular learning. SMP dramatically reduces training complexity, enabling it to perform efficiently on large-scale molecules. In addition, SMP is capable of distinguishing almost all molecular structures, and the uncovered cases may not exist in practice. Based on meaningful physically-based representations of 3D information, we further propose the SphereNet for 3D molecular learning. Experimental results demonstrate that the use of meaningful 3D information in SphereNet leads to significant performance improvements in prediction tasks. Our results also demonstrate the advantages of SphereNet in terms of capability, efficiency, and scalability. Our code is publicly available as part of the DIG library (https://github.com/divelab/DIG)."
                },
                {
                    "title": "GraphDF: A Discrete Flow Model for Molecular Graph Generation",
                    "abstract": "We consider the problem of molecular graph generation using deep models. While graphs are discrete, most existing methods use continuous latent variables, resulting in inaccurate modeling of discrete graph structures. In this work, we propose GraphDF, a novel discrete latent variable model for molecular graph generation based on normalizing flow methods. GraphDF uses invertible modulo shift transforms to map discrete latent variables to graph nodes and edges. We show that the use of discrete latent variables reduces computational costs and eliminates the negative effect of dequantization. Comprehensive experimental results show that GraphDF outperforms prior methods on random generation, property optimization, and constrained optimization tasks."
                },
                {
                    "title": "GraphEBM: Molecular Graph Generation with Energy-Based Models",
                    "abstract": "Molecular graph generation is an emerging area of research with numerous applications. This problem remains challenging as molecular graphs are discrete, irregular, and permutation invariant to node order. Notably, most existing approaches fail to guarantee the intrinsic property of permutation invariance, resulting in unexpected bias in generative models. In this work, we propose GraphEBM to generate molecular graphs using energy-based models. In particular, we parameterize the energy function in a permutation invariant manner, thus making GraphEBM permutation invariant. We apply Langevin dynamics to train the energy function by approximately maximizing likelihood and generate samples with low energies. Furthermore, to generate molecules with a specific desirable property, we propose a simple yet effective strategy, which pushes down energies with flexible degrees according to the properties of corresponding molecules. Finally, we explore the use of GraphEBM for generating molecules with multiple objectives in a compositional manner. Comprehensive experimental results on random, goal-directed, and compositional generation tasks demonstrate the effectiveness of our proposed method."
                },
                {
                    "title": "The Open Catalyst 2020 (OC20) Dataset and Community Challenges",
                    "abstract": "Catalyst discovery and optimization is key to solving many societal and energy challenges including solar fuels synthesis, long-term energy storage, and renewable fertilizer production. Despite considerable effort by the catalysis community to apply machine learning models to the computational catalyst discovery process, it remains an open challenge to build models that can generalize across both elemental compositions of surfaces and adsorbate identity/configurations, perhaps because datasets have been smaller in catalysis than related fields. To address this we developed the OC20 dataset, consisting of 1,281,121 Density Functional Theory (DFT) relaxations (264,900,500 single point evaluations) across a wide swath of materials, surfaces, and adsorbates (nitrogen, carbon, and oxygen chemistries). We supplemented this dataset with randomly perturbed structures, short timescale molecular dynamics, and electronic structure analyses. The dataset comprises three central tasks indicative of day-to-day catalyst modeling and comes with pre-defined train/validation/test splits to facilitate direct comparisons with future model development efforts. We applied three state-of-the-art graph neural network models (SchNet, Dimenet, CGCNN) to each of these tasks as baseline demonstrations for the community to build on. In almost every task, no upper limit on model size was identified, suggesting that even larger models are likely to improve on initial results. The dataset and baseline models are both provided as open resources, as well as a public leader board to encourage community contributions to solve these important tasks."
                },
                {
                    "title": "Graph convolutional neural networks with global attention for improved materials property prediction.",
                    "abstract": "The development of an efficient and powerful machine learning (ML) model for materials property prediction (MPP) remains an important challenge in materials science. While various techniques have been proposed to extract physicochemical features in MPP, graph neural networks (GNN) have also shown very strong capability in capturing effective features for high-performance MPP. Nevertheless, current GNN models do not effectively differentiate the contributions from different atoms. In this paper we develop a novel graph neural network model called GATGNN for predicting properties of inorganic materials. GATGNN is characterized by its composition of augmented graph-attention layers (AGAT) and a global attention layer. The application of AGAT layers and global attention layers respectively learn the local relationship among neighboring atoms and overall contribution of the atoms to the material's property; together making our framework achieve considerably better prediction performance on various tested properties. Through extensive experiments, we show that our method is able to outperform existing state-of-the-art GNN models while it can also provide a measurable insight into the correlation between the atoms and their material property. Our code can found on - https://github.com/superlouis/GATGNN."
                },
                {
                    "title": "Directional Message Passing for Molecular Graphs",
                    "abstract": "Graph neural networks have recently achieved great successes in predicting quantum mechanical properties of molecules. These models represent a molecule as a graph using only the distance between atoms (nodes) and not the spatial direction from one atom to another. However, directional information plays a central role in empirical potentials for molecules, e.g. in angular potentials. To alleviate this limitation we propose directional message passing, in which we embed the messages passed between atoms instead of the atoms themselves. Each message is associated with a direction in coordinate space. These directional message embeddings are rotationally equivariant since the associated directions rotate with the molecule. We propose a message passing scheme analogous to belief propagation, which uses the directional information by transforming messages based on the angle between them. Additionally, we use spherical Bessel functions to construct a theoretically well-founded, orthogonal radial basis that achieves better performance than the currently prevalent Gaussian radial basis functions while using more than 4x fewer parameters. We leverage these innovations to construct the directional message passing neural network (DimeNet). DimeNet outperforms previous GNNs on average by 77% on MD17 and by 41% on QM9."
                },
                {
                    "title": "Graph U-Nets",
                    "abstract": "We consider the problem of representation learning for graph data. Given images are special cases of graphs with nodes lie on 2D lattices, graph embedding tasks have a natural correspondence with image pixel-wise prediction tasks such as segmentation. While encoder-decoder architectures like U-Nets have been successfully applied to image pixel-wise prediction tasks, similar methods are lacking for graph data. This is because pooling and up-sampling operations are not natural on graph data. To address these challenges, we propose novel graph pooling and unpooling operations. The gPool layer adaptively selects some nodes to form a smaller graph based on their scalar projection values. We further propose the gUnpool layer as the inverse operation of the gPool layer. Based on our proposed methods, we develop an encoder-decoder model, known as the graph U-Nets. Experimental results on node classification and graph classification tasks demonstrate that our methods achieve consistently better performance than previous models. Along this direction, we extend our methods by integrating attention mechanisms. Based on attention operators, we proposed attention-based pooling and unpooling layers, which can better capture graph topology information. The empirical results on graph classification tasks demonstrate the promising capability of our methods."
                },
                {
                    "title": "Graph Networks as a Universal Machine Learning Framework for Molecules and Crystals",
                    "abstract": "Graph networks are a new machine learning (ML) paradigm that supports both relational reasoning and combinatorial generalization. Here, we develop, for the first time, universal MatErials Graph Network (MEGNet) models for accurate property prediction in both molecules and crystals. We demonstrate that our MEGNet models significantly outperform prior ML models in 11 out of 13 properties of the QM9 molecule data set. Furthermore, a single-task unified MEGNet model can accurately predict the internal energy at 0 K and room temperature, enthalpy and Gibbs free energy, with temperature, pressure and entropy being global state inputs. Similarly, we show that MEGNet models trained on $\\sim 60,000$ crystals in the Materials Project substantially outperform prior ML models in the prediction of the formation energies, band gaps and elastic moduli of crystals, achieving better than DFT accuracy over a much larger data set. Such MEGNet models are highly interpretable, and well-established periodic chemical trends can be extracted from the elemental embeddings. Finally, we demonstrate the transfer learning of elemental embeddings from a property model trained on a larger data set (formation energies) to accelerate the training of property models with smaller amounts of data (band gaps and elastic moduli)"
                },
                {
                    "title": "Machine Learning Enabled Computational Screening of Inorganic Solid Electrolytes for Suppression of Dendrite Formation in Lithium Metal Anodes",
                    "abstract": "Next generation batteries based on lithium (Li) metal anodes have been plagued by the dendritic electrodeposition of Li metal on the anode during cycling, resulting in short circuit and capacity loss. Suppression of dendritic growth through the use of solid electrolytes has emerged as one of the most promising strategies for enabling the use of Li metal anodes. We perform a computational screening of over 12\u202f000 inorganic solids based on their ability to suppress dendrite initiation in contact with Li metal anode. Properties for mechanically isotropic and anisotropic interfaces that can be used in stability criteria for determining the propensity of dendrite initiation are usually obtained from computationally expensive first-principles methods. In order to obtain a large data set for screening, we use machine-learning models to predict the mechanical properties of several new solid electrolytes. The machine-learning models are trained on purely structural features of the material, which do not require any first-principles calculations. We train a graph convolutional neural network on the shear and bulk moduli because of the availability of a large training data set with low noise due to low uncertainty in their first-principles-calculated values. We use gradient boosting regressor and kernel ridge regression to train the elastic constants, where the choice of the model depends on the size of the training data and the noise that it can handle. The material stiffness is found to increase with an increase in mass density and ratio of Li and sublattice bond ionicity, and decrease with increase in volume per atom and sublattice electronegativity. Cross-validation/test performance suggests our models generalize well. We predict over 20 mechanically anisotropic interfaces between Li metal and four solid electrolytes which can be used to suppress dendrite growth. Our screened candidates are generally soft and highly anisotropic, and present opportunities for simultaneously obtaining dendrite suppression and high ionic conductivity in solid electrolytes."
                },
                {
                    "title": "Crystal Graph Convolutional Neural Networks for an Accurate and Interpretable Prediction of Material Properties.",
                    "abstract": "The use of machine learning methods for accelerating the design of crystalline materials usually requires manually constructed feature vectors or complex transformation of atom coordinates to input the crystal structure, which either constrains the model to certain crystal types or makes it difficult to provide chemical insights. Here, we develop a crystal graph convolutional neural networks framework to directly learn material properties from the connection of atoms in the crystal, providing a universal and interpretable representation of crystalline materials. Our method provides a highly accurate prediction of density functional theory calculated properties for eight different properties of crystals with various structure types and compositions after being trained with 10^{4} data points. Further, our framework is interpretable because one can extract the contributions from local chemical environments to global properties. Using an example of perovskites, we show how this information can be utilized to discover empirical rules for materials design."
                },
                {
                    "title": "Super-Convergence: Very Fast Training of Residual Networks Using Large Learning Rates",
                    "abstract": "In this paper, we describe a phenomenon, which we named \"super-convergence\", where neural networks can be trained an order of magnitude faster than with standard training methods. The existence of super-convergence is relevant to understanding why deep networks generalize well. One of the key elements of super-convergence is training with one learning rate cycle and a large maximum learning rate. A primary insight that allows super-convergence training is that large learning rates regularize the training, hence requiring a reduction of all other forms of regularization in order to preserve an optimal regularization balance. We also derive a simplification of the Hessian Free optimization method to compute an estimate of the optimal learning rate. Experiments demonstrate super-convergence for Cifar-10/100, MNIST and Imagenet datasets, and resnet, wide-resnet, densenet, and inception architectures. In addition, we show that super-convergence provides a greater boost in performance relative to standard training when the amount of labeled training data is limited. The architectures and code to replicate the figures in this paper are available at this http URL. See this http URL for an application of super-convergence to win the DAWNBench challenge (see this https URL)."
                },
                {
                    "title": "SchNet: A continuous-filter convolutional neural network for modeling quantum interactions",
                    "abstract": "Deep learning has the potential to revolutionize quantum chemistry as it is ideally suited to learn representations for structured data and speed up the exploration of chemical space. While convolutional neural networks have proven to be the first choice for images, audio and video data, the atoms in molecules are not restricted to a grid. Instead, their precise locations contain essential physical information, that would get lost if discretized. Thus, we propose to use continuous-filter convolutional layers to be able to model local correlations without requiring the data to lie on a grid. We apply those layers in SchNet: a novel deep learning architecture modeling quantum interactions in molecules. We obtain a joint model for the total energy and interatomic forces that follows fundamental quantum-chemical principles. This includes rotationally invariant energy predictions and a smooth, differentiable potential energy surface. Our architecture achieves state-of-the-art performance for benchmarks of equilibrium molecules and molecular dynamics trajectories. Finally, we introduce a more challenging benchmark with chemical and structural variations that suggests the path for further work."
                },
                {
                    "title": "High-Throughput Machine-Learning-Driven Synthesis of Full-Heusler Compounds",
                    "abstract": "A machine-learning model has been trained to discover Heusler compounds, which are intermetallics exhibiting diverse physical properties attractive for applications in thermoelectric and spintronic materials. Improving these properties requires knowledge of crystal structures, which occur in three subtle variations (Heusler, inverse Heusler, and CsCl-type structures) that are difficult, and at times impossible, to distinguish by diffraction techniques. Compared to alternative approaches, this Heusler discovery engine performs exceptionally well, making fast and reliable predictions of the occurrence of Heusler vs non-Heusler compounds for an arbitrary combination of elements with no structural input on over 400\u202f000 candidates. The model has a true positive rate of 0.94 (and false positive rate of 0.01). It is also valuable for data sanitizing, by flagging questionable entries in crystallographic databases. It was applied to screen candidates with the formula AB2C and predict the existence of 12 novel gall..."
                },
                {
                    "title": "Combinatorial screening for new materials in unconstrained composition space with machine learning",
                    "abstract": "Typically, computational screens for new materials sharply constrain the compositional search space, structural search space, or both, for the sake of tractability. To lift these constraints, we construct a machine learning model from a database of thousands of density functional theory (DFT) calculations. The resulting model can predict the thermodynamic stability of arbitrary compositions without any other input and with six orders of magnitude less computer time than DFT. We use this model to scan roughly 1.6 million candidate compositions for novel ternary compounds (AxByCz), and predict 4500 new stable materials. Our method can be readily applied to other descriptors of interest to accelerate domain-specific materials discovery."
                },
                {
                    "title": "Commentary: The Materials Project: A materials genome approach to accelerating materials innovation",
                    "abstract": "Accelerating the discovery of advanced materials is essential for human welfare and sustainable, clean energy. In this paper, we introduce the Materials Project (www.materialsproject.org), a core program of the Materials Genome Initiative that uses high-throughput computing to uncover the properties of all known inorganic materials. This open dataset can be accessed through multiple channels for both interactive exploration and data mining. The Materials Project also seeks to create open-source platforms for developing robust, sophisticated materials analyses. Future efforts will enable users to perform \u2018\u2018rapid-prototyping\u2019\u2019 of new materials in silico, and provide researchers with new avenues for cost-effective, data-driven materials design. \u00a9 2013 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution 3.0 Unported License."
                },
                {
                    "title": "The passive symmetries of machine learning",
                    "abstract": "Any representation of data involves arbitrary in-vestigator choices. Because those choices are ex-ternal to the data-generating process, each choice leads to an exact symmetry, corresponding to the group of transformations that takes one possible representation to another. These are the passive symmetries ; they include coordinate freedom, gauge symmetry and units covariance, all of which have led to important results in physics. Our goal is to understand the implications of passive symmetries for machine learning: Which passive symmetries play a role (e.g., permutation symmetry in graph neural networks)? What are dos and don\u2019ts in machine learning practice? We assay conditions under which passive symmetries can be implemented as group equivariances. We also discuss links to causal modeling, and argue that the implementation of passive symmetries is particularly valuable when the goal of the learning problem is to generalize out of sample. While this paper is purely conceptual, we believe that it can have a signi\ufb01cant impact on helping machine learning make the transition that took place for modern physics in the \ufb01rst half of the Twentieth century."
                },
                {
                    "title": "Learning Fair Graph Representations via Automated Data Augmentations",
                    "abstract": "We consider fair graph representation learning via data augmentations. While this direction has been explored previously, existing methods invariably rely on certain assumptions on the properties of fair graph data in order to design fixed strategies on data augmentations. Nevertheless, the exact properties of fair graph data may vary significantly in different scenarios. Hence, heuristically designed augmentations may not always generate fair graph data in different application scenarios. In this work, we propose a method, known as Graphair, to learn fair representations based on automated graph data augmentations. Such fairness-aware augmentations are themselves learned from data. Our Graphair is designed to automatically discover fairness-aware augmentations from input graphs in order to circumvent sensitive information while preserving other useful information. Experimental results demonstrate that our Graphair consistently outperforms many baselines on multiple node classification datasets in terms of fairness-accuracy trade-off performance. In addition, results indicate that Graphair can automatically learn to generate fair graph data without prior knowledge on fairness-relevant graph properties. Our code is publicly available as part of the DIG package (https://github.com/divelab/DIG)."
                },
                {
                    "title": "Compact Graph Representation of molecular crystals using Point-wise Distance Distributions",
                    "abstract": "Use of graphs to represent molecular crystals has become popular in recent years as they provide a natural translation from atoms and bonds to nodes and edges. Graphs capture structure, while remaining invariant to the symmetries that crystals display. Several works in property prediction, including those with state-of-the-art results, make use of the Crystal Graph . The present work o\ufb00ers a graph based on Point-wise Distance Distributions which retains symmetrical invariance, decreases computational load, and yields similar or better prediction accuracy on both experimental and simulated crystals."
                },
                {
                    "title": "Resolving the data ambiguity for periodic crystals",
                    "abstract": "The fundamental model of all solid crystalline materials is a periodic set of atomic centers considered up to rigid motion in Euclidean space. The major obstacle to materials discovery was highly ambiguous representations of periodic crystals that didn\u2019t allow fast and reliable comparisons and led to numerous (near-) duplicates in many databases of experimental and simulated crystals. This paper exemplarily resolves the ambiguity by invariants, which are descriptors without false negatives. The new Pointwise Distance Distributions (PDD) is a numerical matrix with a near-linear time complexity and an exactly computable metric. The strongest theoretical result is generic completeness (absence of false positives) for all \ufb01nite and periodic sets of points in any dimension. The strength of PDD is shown by 200B+ pairwise comparisons of all periodic structures in the world\u2019s largest collection (Cambridge Structural Database) of existing materials over two days on a modest desktop."
                },
                {
                    "title": "Do Transformers Really Perform Badly for Graph Representation?",
                    "abstract": "The Transformer architecture has become a dominant choice in many domains, such as natural language processing and computer vision. Yet, it has not achieved competitive performance on popular leaderboards of graph-level prediction compared to mainstream GNN variants. Therefore, it remains a mystery how Transformers could perform well for graph representation learning. In this paper, we solve this mystery by presenting Graphormer, which is built upon the standard Transformer architecture, and could attain excellent results on a broad range of graph representation learning tasks, especially on the recent OGB Large-Scale Challenge. Our key insight to utilizing Transformer in the graph is the necessity of effectively encoding the structural information of a graph into the model. To this end, we propose several simple yet effective structural encoding methods to help Graphormer better model graph-structured data. Besides, we mathematically characterize the expressive power of Graphormer and exhibit that with our ways of encoding the structural information of graphs, many popular GNN variants could be covered as the special cases of Graphormer. The code and models of Graphormer will be made publicly available at https://github.com/Microsoft/Graphormer ."
                }
            ],
            "categories": [
                "cs.LG",
                "cond-mat.mtrl-sci"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we construct graph representations for crystalline materials that distinguish different crystals while maintaining geometric completeness and invariance under crystal passive symmetries?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for advancing materials discovery, as it can significantly reduce the reliance on costly and time-consuming experimental methods. By improving the accuracy and efficiency of property predictions for crystalline materials, this research could lead to the development of novel materials with desirable properties, impacting various fields such as engineering, biomedical applications, and energy storage. Furthermore, it could pave the way for future research in machine learning applications for materials science, enhancing our understanding of material behaviors and properties.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the need to capture the complete geometric information of crystal structures while distinguishing between different crystalline materials. Naive approaches may fail because existing graph representations do not adequately account for periodic patterns in crystals or maintain geometric completeness. Additionally, the complexity of crystal symmetries and the need for invariant representations add layers of technical and theoretical obstacles that must be addressed to achieve reliable predictions.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has focused on graph representations for small molecules, which do not translate well to crystalline materials due to their periodic nature. Existing methods, such as AMD and PDD, have limitations in distinguishing chiral crystals and do not maintain geometric completeness necessary for accurate property predictions. These gaps highlight the need for a novel approach that can effectively capture the unique characteristics of crystalline structures, which our proposed methodology aims to address.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing SE(3) invariant and SO(3) equivariant crystal graph representations to ensure geometric completeness for crystalline materials. We introduce ComFormer, with two variants: iComFormer (using SE(3) invariant graphs) and eComFormer (using SO(3) equivariant graphs). Both variants are designed to scale efficiently to large-scale crystal datasets with a complexity of O(nk), where n is the number of atoms in the crystal unit cell and k is the average number of neighbors. We expect these models to improve the accuracy of crystal property predictions and effectively distinguish between different crystalline structures."
            }
        },
        "author_data": {
            "2999fcff-0b88-49d4-a858-ed45d1ffca57": {
                "pk": "2999fcff-0b88-49d4-a858-ed45d1ffca57",
                "name": "Keqiang Yan",
                "collaborators": [
                    "Shuiwang Ji",
                    "Xiaoning Qian",
                    "Youzhi Luo",
                    "Yi Liu",
                    "Yu-Ching Lin",
                    "Limei Wang",
                    "Cong Fu",
                    "Meng Liu",
                    "Alexandra Saxton",
                    "Xiaofeng Qian"
                ],
                "domain": [
                    "Machine Learning",
                    "Graph Neural Network",
                    "Material Science",
                    "Molecular Generation"
                ],
                "publications": [
                    {
                        "title": "A Space Group Symmetry Informed Network for O(3) Equivariant Crystal Tensor Prediction",
                        "abstract": "We consider the prediction of general tensor properties of crystalline materials, including dielectric, piezoelectric, and elastic tensors. A key challenge here is how to make the predictions satisfy the unique tensor equivariance to O(3) group and invariance to crystal space groups. To this end, we propose a General Materials Tensor Network (GMTNet), which is carefully designed to satisfy the required symmetries. To evaluate our method, we curate a dataset and establish evaluation metrics that are tailored to the intricacies of crystal tensor predictions. Experimental results show that our GMTNet not only achieves promising performance on crystal tensors of various orders but also generates predictions fully consistent with the intrinsic crystal symmetries. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS)."
                    },
                    {
                        "title": "Efficient Approximations of Complete Interatomic Potentials for Crystal Property Prediction",
                        "abstract": "We study property prediction for crystal materials. A crystal structure consists of a minimal unit cell that is repeated infinitely in 3D space. How to accurately represent such repetitive structures in machine learning models remains unresolved. Current methods construct graphs by establishing edges only between nearby nodes, thereby failing to faithfully capture infinite repeating patterns and distant interatomic interactions. In this work, we propose several innovations to overcome these limitations. First, we propose to model physics-principled interatomic potentials directly instead of only using distances as in many existing methods. These potentials include the Coulomb potential, London dispersion potential, and Pauli repulsion potential. Second, we model the complete set of potentials among all atoms, instead of only between nearby atoms as in existing methods. This is enabled by our approximations of infinite potential summations with provable error bounds. We further develop efficient algorithms to compute the approximations. Finally, we propose to incorporate our computations of complete interatomic potentials into message passing neural networks for representation learning. We perform experiments on the JARVIS and Materials Project benchmarks for evaluation. Results show that the use of interatomic potentials and complete interatomic potentials leads to consistent performance improvements with reasonable computational costs. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS/tree/main/OpenMat/PotNet)."
                    },
                    {
                        "title": "Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems",
                        "abstract": "Advances in artificial intelligence (AI) are fueling a new paradigm of discoveries in natural sciences. Today, AI has started to advance natural sciences by improving, accelerating, and enabling our understanding of natural phenomena at a wide range of spatial and temporal scales, giving rise to a new area of research known as AI for science (AI4Science). Being an emerging research paradigm, AI4Science is unique in that it is an enormous and highly interdisciplinary area. Thus, a unified and technical treatment of this field is needed yet challenging. This work aims to provide a technically thorough account of a subarea of AI4Science; namely, AI for quantum, atomistic, and continuum systems. These areas aim at understanding the physical world from the subatomic (wavefunctions and electron density), atomic (molecules, proteins, materials, and interactions), to macro (fluids, climate, and subsurface) scales and form an important subarea of AI4Science. A unique advantage of focusing on these areas is that they largely share a common set of challenges, thereby allowing a unified and foundational treatment. A key common challenge is how to capture physics first principles, especially symmetries, in natural systems by deep learning methods. We provide an in-depth yet intuitive account of techniques to achieve equivariance to symmetry transformations. We also discuss other common technical challenges, including explainability, out-of-distribution generalization, knowledge transfer with foundation and large language models, and uncertainty quantification. To facilitate learning and education, we provide categorized lists of resources that we found to be useful. We strive to be thorough and unified and hope this initial effort may trigger more community interests and efforts to further advance AI4Science."
                    },
                    {
                        "title": "A Latent Diffusion Model for Protein Structure Generation",
                        "abstract": "Proteins are complex biomolecules that perform a variety of crucial functions within living organisms. Designing and generating novel proteins can pave the way for many future synthetic biology applications, including drug discovery. However, it remains a challenging computational task due to the large modeling space of protein structures. In this study, we propose a latent diffusion model that can reduce the complexity of protein modeling while flexibly capturing the distribution of natural protein structures in a condensed latent space. Specifically, we propose an equivariant protein autoencoder that embeds proteins into a latent space and then uses an equivariant diffusion model to learn the distribution of the latent protein representations. Experimental results demonstrate that our method can effectively generate novel protein backbone structures with high designability and efficiency. The code will be made publicly available at https://github.com/divelab/AIRS/tree/main/OpenProt/LatentDiff"
                    },
                    {
                        "title": "Periodic Graph Transformers for Crystal Material Property Prediction",
                        "abstract": "We consider representation learning on periodic graphs encoding crystal materials. Different from regular graphs, periodic graphs consist of a minimum unit cell repeating itself on a regular lattice in 3D space. How to effectively encode these periodic structures poses unique challenges not present in regular graph representation learning. In addition to being E(3) invariant, periodic graph representations need to be periodic invariant. That is, the learned representations should be invariant to shifts of cell boundaries as they are artificially imposed. Furthermore, the periodic repeating patterns need to be captured explicitly as lattices of different sizes and orientations may correspond to different materials. In this work, we propose a transformer architecture, known as Matformer, for periodic graph representation learning. Our Matformer is designed to be invariant to periodicity and can capture repeating patterns explicitly. In particular, Matformer encodes periodic patterns by efficient use of geometric distances between the same atoms in neighboring cells. Experimental results on multiple common benchmark datasets show that our Matformer outperforms baseline methods consistently. In addition, our results demonstrate the importance of periodic invariance and explicit repeating pattern encoding for crystal representation learning."
                    },
                    {
                        "title": "Examining graph neural networks for crystal structures: Limitations and opportunities for capturing periodicity",
                        "abstract": "Graph neural networks (GNNs) have recently been used to learn the representations of crystal structures through an end-to-end data-driven approach. However, a systematic top-down approach to evaluate and understand the limitations of GNNs in accurately capturing crystal structures has yet to be established. In this study, we introduce an approach using human-designed descriptors as a compendium of human knowledge to investigate the extent to which GNNs can comprehend crystal structures. Our findings reveal that current state-of-the-art GNNs fall short in accurately capturing the periodicity of crystal structures. We analyze this failure by exploring three aspects: local expressive power, long-range information processing, and readout function. To address these identified limitations, we propose a straightforward and general solution: the hybridization of descriptors with GNNs, which directly supplements the missing information to GNNs. The hybridization enhances the predictive accuracy of GNNs for specific material properties, most notably phonon internal energy and heat capacity, which heavily rely on the periodicity of materials."
                    },
                    {
                        "title": "DIG: A Turnkey Library for Diving into Graph Deep Learning Research",
                        "abstract": "Although there exist several libraries for deep learning on graphs, they are aiming at implementing basic operations for graph deep learning. In the research community, implementing and benchmarking various advanced tasks are still painful and time-consuming with existing libraries. To facilitate graph deep learning research, we introduce DIG: Dive into Graphs, a turnkey library that provides a unified testbed for higher level, research-oriented graph deep learning tasks. Currently, we consider graph generation, self-supervised learning on graphs, explainability of graph neural networks, and deep learning on 3D graphs. For each direction, we provide unified implementations of data interfaces, common algorithms, and evaluation metrics. Altogether, DIG is an extensible, open-source, and turnkey library for researchers to develop new methods and effortlessly compare with common baselines using widely used datasets and evaluation metrics. Source code is available at https://github.com/divelab/DIG."
                    },
                    {
                        "title": "GraphDF: A Discrete Flow Model for Molecular Graph Generation",
                        "abstract": "We consider the problem of molecular graph generation using deep models. While graphs are discrete, most existing methods use continuous latent variables, resulting in inaccurate modeling of discrete graph structures. In this work, we propose GraphDF, a novel discrete latent variable model for molecular graph generation based on normalizing flow methods. GraphDF uses invertible modulo shift transforms to map discrete latent variables to graph nodes and edges. We show that the use of discrete latent variables reduces computational costs and eliminates the negative effect of dequantization. Comprehensive experimental results show that GraphDF outperforms prior methods on random generation, property optimization, and constrained optimization tasks."
                    },
                    {
                        "title": "GraphEBM: Molecular Graph Generation with Energy-Based Models",
                        "abstract": "Molecular graph generation is an emerging area of research with numerous applications. This problem remains challenging as molecular graphs are discrete, irregular, and permutation invariant to node order. Notably, most existing approaches fail to guarantee the intrinsic property of permutation invariance, resulting in unexpected bias in generative models. In this work, we propose GraphEBM to generate molecular graphs using energy-based models. In particular, we parameterize the energy function in a permutation invariant manner, thus making GraphEBM permutation invariant. We apply Langevin dynamics to train the energy function by approximately maximizing likelihood and generate samples with low energies. Furthermore, to generate molecules with a specific desirable property, we propose a simple yet effective strategy, which pushes down energies with flexible degrees according to the properties of corresponding molecules. Finally, we explore the use of GraphEBM for generating molecules with multiple objectives in a compositional manner. Comprehensive experimental results on random, goal-directed, and compositional generation tasks demonstrate the effectiveness of our proposed method."
                    }
                ]
            },
            "cb1c6a27-8187-452f-bb16-32838a295ac3": {
                "pk": "cb1c6a27-8187-452f-bb16-32838a295ac3",
                "name": "Cong Fu",
                "collaborators": [
                    "Shuiwang Ji",
                    "Xuan Zhang",
                    "Jacob Helwig",
                    "Yaochen Xie",
                    "Limei Wang",
                    "Youzhi Luo",
                    "Meng Liu",
                    "Zhao Xu",
                    "Shenglong Xu",
                    "Xiner Li"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Drug Design",
                    "Machine Learning",
                    "Physics-Informed Learning"
                ],
                "publications": [
                    {
                        "title": "Fragment and Geometry Aware Tokenization of Molecules for Structure-Based Drug Design Using Language Models.",
                        "abstract": "Structure-based drug design (SBDD) is crucial for developing specific and effective therapeutics against protein targets but remains challenging due to complex protein-ligand interactions and vast chemical space. Although language models (LMs) have excelled in natural language processing, their application in SBDD is underexplored. To bridge this gap, we introduce a method, known as Frag2Seq, to apply LMs to SBDD by generating molecules in a fragment-based manner in which fragments correspond to functional modules. We transform 3D molecules into fragment-informed sequences using SE(3)-equivariant molecule and fragment local frames, extracting SE(3)-invariant sequences that preserve geometric information of 3D fragments. Furthermore, we incorporate protein pocket embeddings obtained from a pre-trained inverse folding model into the LMs via cross-attention to capture protein-ligand interaction, enabling effective target-aware molecule generation. Benefiting from employing LMs with fragment-based generation and effective protein context encoding, our model achieves the best performance on binding vina score and chemical properties such as QED and Lipinski, which shows our model's efficacy in generating drug-like ligands with higher binding affinity against target proteins. Moreover, our method also exhibits higher sampling efficiency compared to atom-based autoregressive and diffusion baselines with at most ~300x speedup."
                    },
                    {
                        "title": "SineNet: Learning Temporal Dynamics in Time-Dependent Partial Differential Equations",
                        "abstract": "We consider using deep neural networks to solve time-dependent partial differential equations (PDEs), where multi-scale processing is crucial for modeling complex, time-evolving dynamics. While the U-Net architecture with skip connections is commonly used by prior studies to enable multi-scale processing, our analysis shows that the need for features to evolve across layers results in temporally misaligned features in skip connections, which limits the model's performance. To address this limitation, we propose SineNet, consisting of multiple sequentially connected U-shaped network blocks, referred to as waves. In SineNet, high-resolution features are evolved progressively through multiple stages, thereby reducing the amount of misalignment within each stage. We furthermore analyze the role of skip connections in enabling both parallel and sequential processing of multi-scale information. Our method is rigorously tested on multiple PDE datasets, including the Navier-Stokes equations and shallow water equations, showcasing the advantages of our proposed approach over conventional U-Nets with a comparable parameter budget. We further demonstrate that increasing the number of waves in SineNet while maintaining the same number of parameters leads to a monotonically improved performance. The results highlight the effectiveness of SineNet and the potential of our approach in advancing the state-of-the-art in neural PDE solver design. Our code is available as part of AIRS (https://github.com/divelab/AIRS)."
                    },
                    {
                        "title": "Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems",
                        "abstract": "Advances in artificial intelligence (AI) are fueling a new paradigm of discoveries in natural sciences. Today, AI has started to advance natural sciences by improving, accelerating, and enabling our understanding of natural phenomena at a wide range of spatial and temporal scales, giving rise to a new area of research known as AI for science (AI4Science). Being an emerging research paradigm, AI4Science is unique in that it is an enormous and highly interdisciplinary area. Thus, a unified and technical treatment of this field is needed yet challenging. This work aims to provide a technically thorough account of a subarea of AI4Science; namely, AI for quantum, atomistic, and continuum systems. These areas aim at understanding the physical world from the subatomic (wavefunctions and electron density), atomic (molecules, proteins, materials, and interactions), to macro (fluids, climate, and subsurface) scales and form an important subarea of AI4Science. A unique advantage of focusing on these areas is that they largely share a common set of challenges, thereby allowing a unified and foundational treatment. A key common challenge is how to capture physics first principles, especially symmetries, in natural systems by deep learning methods. We provide an in-depth yet intuitive account of techniques to achieve equivariance to symmetry transformations. We also discuss other common technical challenges, including explainability, out-of-distribution generalization, knowledge transfer with foundation and large language models, and uncertainty quantification. To facilitate learning and education, we provide categorized lists of resources that we found to be useful. We strive to be thorough and unified and hope this initial effort may trigger more community interests and efforts to further advance AI4Science."
                    },
                    {
                        "title": "Semi-Supervised Learning for High-Fidelity Fluid Flow Reconstruction",
                        "abstract": "Physical simulations of fluids are crucial for understanding fluid dynamics across many applications, such as weather prediction and engineering design. While high-resolution numerical simulations can provide substantial accuracy in analysis, it also results in prohibitive computational costs. Conversely, lower-resolution simulations are computationally less expensive but compromise the accuracy and reliability of results. In this work, we propose a cascaded fluid reconstruction framework to combine large amounts of low-resolution and limited amounts of paired high-resolution direct simulations for accurate fluid analysis. Our method can improve the accuracy of simulations while preserving the efficiency of low-resolution simulations. Our framework involves a proposal network, pre-trained with small amounts of high-resolution labels, to reconstruct an initial high-resolution flow field. The field is then refined in the frequency domain to become more physically plausible using our proposed refinement network, known as ModeFormer, which is implemented as a complex-valued transformer, with physics-informed unsupervised training. Our experimental results demonstrate the effectiveness of our approach in enhancing the overall performance of fluid flow reconstruction. The code will be made publicly available at https://github.com/divelab/AIRS/tree/main/OpenPDE/CFRF ."
                    },
                    {
                        "title": "Group Equivariant Fourier Neural Operators for Partial Differential Equations",
                        "abstract": "We consider solving partial differential equations (PDEs) with Fourier neural operators (FNOs), which operate in the frequency domain. Since the laws of physics do not depend on the coordinate system used to describe them, it is desirable to encode such symmetries in the neural operator architecture for better performance and easier learning. While encoding symmetries in the physical domain using group theory has been studied extensively, how to capture symmetries in the frequency domain is under-explored. In this work, we extend group convolutions to the frequency domain and design Fourier layers that are equivariant to rotations, translations, and reflections by leveraging the equivariance property of the Fourier transform. The resulting $G$-FNO architecture generalizes well across input resolutions and performs well in settings with varying levels of symmetry. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS)."
                    },
                    {
                        "title": "Lattice Convolutional Networks for Learning Ground States of Quantum Many-Body Systems",
                        "abstract": "Deep learning methods have been shown to be effective in representing ground-state wave functions of quantum many-body systems. Existing methods use convolutional neural networks (CNNs) for square lattices due to their image-like structures. For non-square lattices, existing method uses graph neural network (GNN) in which structure information is not precisely captured, thereby requiring additional hand-crafted sublattice encoding. In this work, we propose lattice convolutions in which a set of proposed operations are used to convert non-square lattices into grid-like augmented lattices on which regular convolution can be applied. Based on the proposed lattice convolutions, we design lattice convolutional networks (LCN) that use self-gating and attention mechanisms. Experimental results show that our method achieves performance on par or better than existing methods on spin 1/2 $J_1$-$J_2$ Heisenberg model over the square, honeycomb, triangular, and kagome lattices while without using hand-crafted encoding."
                    },
                    {
                        "title": "Fast Quantum Property Prediction via Deeper 2D and 3D Graph Networks",
                        "abstract": "Molecular property prediction is gaining increasing attention due to its diverse applications. One task of particular interests and importance is to predict quantum chemical properties without 3D equilibrium structures. This is practically favorable since obtaining 3D equilibrium structures requires extremely expensive calculations. In this work, we design a deep graph neural network to predict quantum properties by directly learning from 2D molecular graphs. In addition, we propose a 3D graph neural network to learn from low-cost conformer sets, which can be obtained with open-source tools using an affordable budget. We employ our methods to participate in the 2021 KDD Cup on OGB Large-Scale Challenge (OGB-LSC), which aims to predict the HOMO-LUMO energy gap of molecules. Final evaluation results reveal that we are one of the winners with a mean absolute error of 0.1235 on the holdout test set. Our implementation is available as part of the MoleculeX package (https://github.com/divelab/MoleculeX)."
                    },
                    {
                        "title": "DIG: A Turnkey Library for Diving into Graph Deep Learning Research",
                        "abstract": "Although there exist several libraries for deep learning on graphs, they are aiming at implementing basic operations for graph deep learning. In the research community, implementing and benchmarking various advanced tasks are still painful and time-consuming with existing libraries. To facilitate graph deep learning research, we introduce DIG: Dive into Graphs, a turnkey library that provides a unified testbed for higher level, research-oriented graph deep learning tasks. Currently, we consider graph generation, self-supervised learning on graphs, explainability of graph neural networks, and deep learning on 3D graphs. For each direction, we provide unified implementations of data interfaces, common algorithms, and evaluation metrics. Altogether, DIG is an extensible, open-source, and turnkey library for researchers to develop new methods and effortlessly compare with common baselines using widely used datasets and evaluation metrics. Source code is available at https://github.com/divelab/DIG."
                    },
                    {
                        "title": "Cross-Domain Recommendation via Preference Propagation GraphNet",
                        "abstract": "Recommendation can be framed as a graph link prediction task naturally. The user-item interaction graph built within a single domain often suffers from high sparsity. Thus, there has been a surge of approaches to alleviate the sparsity issue via cross-domain mutual augmentation. The SOTA cross-domain recommendation algorithms all try to bridge the gap via knowledge transfer in the latent space. We find there are mainly three problems in their formulations: 1) their knowledge transfer is unaware of the cross-domain graph structure. 2) their framework cannot capture high-order information propagation on the graph. 3) their cross-domain transfer formulations are generally more complicated to be optimized than the unified methods. In this paper, we propose the Preference Propagation GraphNet (PPGN) to address the above problems. Specifically, we construct a Cross-Domain Preference Matrix (CDPM) to model the interactions of different domains as a whole. Through the propagation layer of PPGN, we try to capture how user preferences propagate in the graph. Consequently, a joint objective for different domains is defined, and we simplify the cross-domain recommendation into a unified multi-task model. Extensive experiments on two pairs of real-world datasets show PPGN outperforms the SOTA algorithms significantly."
                    },
                    {
                        "title": "Collaborative Policy Learning for Open Knowledge Graph Reasoning",
                        "abstract": "In recent years, there has been a surge of interests in interpretable graph reasoning methods. However, these models often suffer from limited performance when working on sparse and incomplete graphs, due to the lack of evidential paths that can reach target entities. Here we study open knowledge graph reasoning\u2014a task that aims to reason for missing facts over a graph augmented by a background text corpus. A key challenge of the task is to filter out \u201cirrelevant\u201d facts extracted from corpus, in order to maintain an effective search space during path inference. We propose a novel reinforcement learning framework to train two collaborative agents jointly, i.e., a multi-hop graph reasoner and a fact extractor. The fact extraction agent generates fact triples from corpora to enrich the graph on the fly; while the reasoning agent provides feedback to the fact extractor and guides it towards promoting facts that are helpful for the interpretable reasoning. Experiments on two public datasets demonstrate the effectiveness of the proposed approach."
                    }
                ]
            },
            "288f47ed-9140-4cae-883b-5aaded5b1a56": {
                "pk": "288f47ed-9140-4cae-883b-5aaded5b1a56",
                "name": "Xiaofeng Qian",
                "collaborators": [
                    "Keqiang Yan",
                    "Alexandra Saxton",
                    "Xiaoning Qian",
                    "Shuiwang Ji"
                ],
                "domain": [
                    "Materials Science",
                    "Tensor Analysis",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "A Space Group Symmetry Informed Network for O(3) Equivariant Crystal Tensor Prediction",
                        "abstract": "We consider the prediction of general tensor properties of crystalline materials, including dielectric, piezoelectric, and elastic tensors. A key challenge here is how to make the predictions satisfy the unique tensor equivariance to O(3) group and invariance to crystal space groups. To this end, we propose a General Materials Tensor Network (GMTNet), which is carefully designed to satisfy the required symmetries. To evaluate our method, we curate a dataset and establish evaluation metrics that are tailored to the intricacies of crystal tensor predictions. Experimental results show that our GMTNet not only achieves promising performance on crystal tensors of various orders but also generates predictions fully consistent with the intrinsic crystal symmetries. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS)."
                    }
                ]
            },
            "1c031351-4ddc-4ceb-ba60-243a4ffb19dd": {
                "pk": "1c031351-4ddc-4ceb-ba60-243a4ffb19dd",
                "name": "Xiaoning Qian",
                "collaborators": [
                    "Shuiwang Ji",
                    "Keqiang Yan",
                    "Youzhi Luo",
                    "Haiyang Yu",
                    "Alexandra Saxton",
                    "Xiaofeng Qian",
                    "Yu-Ching Lin",
                    "Yi Liu",
                    "Limei Wang",
                    "Cong Fu"
                ],
                "domain": [
                    "Machine Learning",
                    "Materials Science",
                    "Quantum Chemistry",
                    "Bioinformatics"
                ],
                "publications": [
                    {
                        "title": "A Space Group Symmetry Informed Network for O(3) Equivariant Crystal Tensor Prediction",
                        "abstract": "We consider the prediction of general tensor properties of crystalline materials, including dielectric, piezoelectric, and elastic tensors. A key challenge here is how to make the predictions satisfy the unique tensor equivariance to O(3) group and invariance to crystal space groups. To this end, we propose a General Materials Tensor Network (GMTNet), which is carefully designed to satisfy the required symmetries. To evaluate our method, we curate a dataset and establish evaluation metrics that are tailored to the intricacies of crystal tensor predictions. Experimental results show that our GMTNet not only achieves promising performance on crystal tensors of various orders but also generates predictions fully consistent with the intrinsic crystal symmetries. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS)."
                    },
                    {
                        "title": "Efficient Approximations of Complete Interatomic Potentials for Crystal Property Prediction",
                        "abstract": "We study property prediction for crystal materials. A crystal structure consists of a minimal unit cell that is repeated infinitely in 3D space. How to accurately represent such repetitive structures in machine learning models remains unresolved. Current methods construct graphs by establishing edges only between nearby nodes, thereby failing to faithfully capture infinite repeating patterns and distant interatomic interactions. In this work, we propose several innovations to overcome these limitations. First, we propose to model physics-principled interatomic potentials directly instead of only using distances as in many existing methods. These potentials include the Coulomb potential, London dispersion potential, and Pauli repulsion potential. Second, we model the complete set of potentials among all atoms, instead of only between nearby atoms as in existing methods. This is enabled by our approximations of infinite potential summations with provable error bounds. We further develop efficient algorithms to compute the approximations. Finally, we propose to incorporate our computations of complete interatomic potentials into message passing neural networks for representation learning. We perform experiments on the JARVIS and Materials Project benchmarks for evaluation. Results show that the use of interatomic potentials and complete interatomic potentials leads to consistent performance improvements with reasonable computational costs. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS/tree/main/OpenMat/PotNet)."
                    },
                    {
                        "title": "Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems",
                        "abstract": "Advances in artificial intelligence (AI) are fueling a new paradigm of discoveries in natural sciences. Today, AI has started to advance natural sciences by improving, accelerating, and enabling our understanding of natural phenomena at a wide range of spatial and temporal scales, giving rise to a new area of research known as AI for science (AI4Science). Being an emerging research paradigm, AI4Science is unique in that it is an enormous and highly interdisciplinary area. Thus, a unified and technical treatment of this field is needed yet challenging. This work aims to provide a technically thorough account of a subarea of AI4Science; namely, AI for quantum, atomistic, and continuum systems. These areas aim at understanding the physical world from the subatomic (wavefunctions and electron density), atomic (molecules, proteins, materials, and interactions), to macro (fluids, climate, and subsurface) scales and form an important subarea of AI4Science. A unique advantage of focusing on these areas is that they largely share a common set of challenges, thereby allowing a unified and foundational treatment. A key common challenge is how to capture physics first principles, especially symmetries, in natural systems by deep learning methods. We provide an in-depth yet intuitive account of techniques to achieve equivariance to symmetry transformations. We also discuss other common technical challenges, including explainability, out-of-distribution generalization, knowledge transfer with foundation and large language models, and uncertainty quantification. To facilitate learning and education, we provide categorized lists of resources that we found to be useful. We strive to be thorough and unified and hope this initial effort may trigger more community interests and efforts to further advance AI4Science."
                    },
                    {
                        "title": "Efficient and Equivariant Graph Networks for Predicting Quantum Hamiltonian",
                        "abstract": "We consider the prediction of the Hamiltonian matrix, which finds use in quantum chemistry and condensed matter physics. Efficiency and equivariance are two important, but conflicting factors. In this work, we propose a SE(3)-equivariant network, named QHNet, that achieves efficiency and equivariance. Our key advance lies at the innovative design of QHNet architecture, which not only obeys the underlying symmetries, but also enables the reduction of number of tensor products by 92\\%. In addition, QHNet prevents the exponential growth of channel dimension when more atom types are involved. We perform experiments on MD17 datasets, including four molecular systems. Experimental results show that our QHNet can achieve comparable performance to the state of the art methods at a significantly faster speed. Besides, our QHNet consumes 50\\% less memory due to its streamlined architecture. Our code is publicly available as part of the AIRS library (\\url{https://github.com/divelab/AIRS})."
                    },
                    {
                        "title": "A Latent Diffusion Model for Protein Structure Generation",
                        "abstract": "Proteins are complex biomolecules that perform a variety of crucial functions within living organisms. Designing and generating novel proteins can pave the way for many future synthetic biology applications, including drug discovery. However, it remains a challenging computational task due to the large modeling space of protein structures. In this study, we propose a latent diffusion model that can reduce the complexity of protein modeling while flexibly capturing the distribution of natural protein structures in a condensed latent space. Specifically, we propose an equivariant protein autoencoder that embeds proteins into a latent space and then uses an equivariant diffusion model to learn the distribution of the latent protein representations. Experimental results demonstrate that our method can effectively generate novel protein backbone structures with high designability and efficiency. The code will be made publicly available at https://github.com/divelab/AIRS/tree/main/OpenProt/LatentDiff"
                    },
                    {
                        "title": "QH9: A Quantum Hamiltonian Prediction Benchmark for QM9 Molecules",
                        "abstract": "Supervised machine learning approaches have been increasingly used in accelerating electronic structure prediction as surrogates of first-principle computational methods, such as density functional theory (DFT). While numerous quantum chemistry datasets focus on chemical properties and atomic forces, the ability to achieve accurate and efficient prediction of the Hamiltonian matrix is highly desired, as it is the most important and fundamental physical quantity that determines the quantum states of physical systems and chemical properties. In this work, we generate a new Quantum Hamiltonian dataset, named as QH9, to provide precise Hamiltonian matrices for 999 or 2998 molecular dynamics trajectories and 130,831 stable molecular geometries, based on the QM9 dataset. By designing benchmark tasks with various molecules, we show that current machine learning models have the capacity to predict Hamiltonian matrices for arbitrary molecules. Both the QH9 dataset and the baseline models are provided to the community through an open-source benchmark, which can be highly valuable for developing machine learning methods and accelerating molecular and materials design for scientific and technological applications. Our benchmark is publicly available at https://github.com/divelab/AIRS/tree/main/OpenDFT/QHBench."
                    },
                    {
                        "title": "MoReL: Multi-omics Relational Learning",
                        "abstract": "Multi-omics data analysis has the potential to discover hidden molecular interactions, revealing potential regulatory and/or signal transduction pathways for cellular processes of interest when studying life and disease systems. One of critical challenges when dealing with real-world multi-omics data is that they may manifest heterogeneous structures and data quality as often existing data may be collected from different subjects under different conditions for each type of omics data. We propose a novel deep Bayesian generative model to efficiently infer a multi-partite graph encoding molecular interactions across such heterogeneous views, using a fused Gromov-Wasserstein (FGW) regularization between latent representations of corresponding views for integrative analysis. With such an optimal transport regularization in the deep Bayesian generative model, it not only allows incorporating view-specific side information, either with graph-structured or unstructured data in different views, but also increases the model flexibility with the distribution-based regularization. This allows efficient alignment of heterogeneous latent variable distributions to derive reliable interaction predictions compared to the existing point-based graph embedding methods. Our experiments on several real-world datasets demonstrate enhanced performance of MoReL in inferring meaningful interactions compared to existing baselines."
                    },
                    {
                        "title": "COVID-Datathon: Biomarker identification for COVID-19 severity based on BALF scRNA-seq data",
                        "abstract": "The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) emergence began in late 2019 and has since spread rapidly worldwide. The characteristics of respiratory immune response to this emerging virus is not clear. Recently, Single-cell RNA sequencing (scRNA-seq) transcriptome profiling of Bronchoalveolar lavage fluid (BALF) cells has been done to elucidate the potential mechanisms underlying in COVID-19. With the aim of better utilizing this atlas of BALF cells in response to the virus, here we propose a bioinformatics pipeline to identify candidate biomarkers of COVID-19 severity, which may help characterize BALF cells to have better mechanistic understanding of SARS-CoV-2 infection. The proposed pipeline is implemented in R and is available at https://github.com/namini94/scBALF_Hackathon."
                    },
                    {
                        "title": "Molecular Systems Biology Peer Review Process File Perturbation of the Mutated Egfr Interactome Identifies Vulnerabilities and Resistance Mechanisms Transaction Report",
                        "abstract": "(Note: With the exception of the correction of typographical or spelling errors that could be a source of ambiguity, letters and reports are not edited. The original formatting of letters and referee reports may not be reflected in this compilation.) Thank you again for submitting your work to Molecular Systems Biology. We have now heard back from the three referees who agreed to evaluate your manuscript. As you will see from the reports below, the referees find the topic of your study of potential interest. They raise, however, substantial concerns on your work, which should be convincingly addressed in a revision of the manuscript. Overall, the reviewers appreciate your highly integrative analysis and the systematic collection of a vast amount of data. Many of the reviewers' comments refer to the need to clarify and better document several points throughout the manuscript. However, the reviewers also raised several major points, which should be carefully addressed. Among the more fundamental issues are the following:-Additional experimentation is required in order to address whether the effect of the knockdowns and drug treatments is specific for an EGFR mutant context.-The integration of the phosphotyrosine proteomics data needs to be described in detail and its contribution to the findings should be clearly shown.-The validity of the combinatorial therapy approach should be demonstrated beyond the single cell line used. If you feel you can satisfactorily deal with these points and those listed by the referees, you may wish to submit a revised version of your manuscript. Please attach a covering letter giving details of the way in which you have handled each of the points raised by the referees. A revised manuscript will be once again subject to review and you probably understand that we can give you no guarantee"
                    }
                ]
            },
            "25665f7e-63c1-4b98-84c4-50cd5ee6506c": {
                "pk": "25665f7e-63c1-4b98-84c4-50cd5ee6506c",
                "name": "Shuiwang Ji",
                "collaborators": [
                    "Keqiang Yan",
                    "Alexandra Saxton",
                    "Xiaofeng Qian",
                    "Xiaoning Qian",
                    "Cong Fu",
                    "Jacob Helwig",
                    "Sheng Gong",
                    "T. Xie",
                    "Y. Shao-horn",
                    "Rafael G\u00f3mez-Bombarelli"
                ],
                "domain": [
                    "Materials Science",
                    "Graph Neural Network",
                    "Tensor Prediction",
                    "Fluid Dynamics"
                ],
                "publications": [
                    {
                        "title": "A Space Group Symmetry Informed Network for O(3) Equivariant Crystal Tensor Prediction",
                        "abstract": "We consider the prediction of general tensor properties of crystalline materials, including dielectric, piezoelectric, and elastic tensors. A key challenge here is how to make the predictions satisfy the unique tensor equivariance to O(3) group and invariance to crystal space groups. To this end, we propose a General Materials Tensor Network (GMTNet), which is carefully designed to satisfy the required symmetries. To evaluate our method, we curate a dataset and establish evaluation metrics that are tailored to the intricacies of crystal tensor predictions. Experimental results show that our GMTNet not only achieves promising performance on crystal tensors of various orders but also generates predictions fully consistent with the intrinsic crystal symmetries. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS)."
                    },
                    {
                        "title": "Semi-Supervised Learning for High-Fidelity Fluid Flow Reconstruction",
                        "abstract": "Physical simulations of fluids are crucial for understanding fluid dynamics across many applications, such as weather prediction and engineering design. While high-resolution numerical simulations can provide substantial accuracy in analysis, it also results in prohibitive computational costs. Conversely, lower-resolution simulations are computationally less expensive but compromise the accuracy and reliability of results. In this work, we propose a cascaded fluid reconstruction framework to combine large amounts of low-resolution and limited amounts of paired high-resolution direct simulations for accurate fluid analysis. Our method can improve the accuracy of simulations while preserving the efficiency of low-resolution simulations. Our framework involves a proposal network, pre-trained with small amounts of high-resolution labels, to reconstruct an initial high-resolution flow field. The field is then refined in the frequency domain to become more physically plausible using our proposed refinement network, known as ModeFormer, which is implemented as a complex-valued transformer, with physics-informed unsupervised training. Our experimental results demonstrate the effectiveness of our approach in enhancing the overall performance of fluid flow reconstruction. The code will be made publicly available at https://github.com/divelab/AIRS/tree/main/OpenPDE/CFRF ."
                    },
                    {
                        "title": "Examining graph neural networks for crystal structures: Limitations and opportunities for capturing periodicity",
                        "abstract": "Graph neural networks (GNNs) have recently been used to learn the representations of crystal structures through an end-to-end data-driven approach. However, a systematic top-down approach to evaluate and understand the limitations of GNNs in accurately capturing crystal structures has yet to be established. In this study, we introduce an approach using human-designed descriptors as a compendium of human knowledge to investigate the extent to which GNNs can comprehend crystal structures. Our findings reveal that current state-of-the-art GNNs fall short in accurately capturing the periodicity of crystal structures. We analyze this failure by exploring three aspects: local expressive power, long-range information processing, and readout function. To address these identified limitations, we propose a straightforward and general solution: the hybridization of descriptors with GNNs, which directly supplements the missing information to GNNs. The hybridization enhances the predictive accuracy of GNNs for specific material properties, most notably phonon internal energy and heat capacity, which heavily rely on the periodicity of materials."
                    }
                ]
            }
        }
    },
    "2405.15632": {
        "paper_data": {
            "title": "Federated Behavioural Planes: Explaining the Evolution of Client Behaviour in Federated Learning",
            "url": "http://arxiv.org/abs/2405.15632v2",
            "arxiv_id": "2405.15632",
            "authors": [
                "Dario Fenoglio",
                "Gabriele Dominici",
                "Pietro Barbiero",
                "Alberto Tonda",
                "Martin Gjoreski",
                "Marc Langheinrich"
            ],
            "abstract": "Federated Learning (FL), a privacy-aware approach in distributed deep learning environments, enables many clients to collaboratively train a model without sharing sensitive data, thereby reducing privacy risks. However, enabling human trust and control over FL systems requires understanding the evolving behaviour of clients, whether beneficial or detrimental for the training, which still represents a key challenge in the current literature. To address this challenge, we introduce Federated Behavioural Planes (FBPs), a novel method to analyse, visualise, and explain the dynamics of FL systems, showing how clients behave under two different lenses: predictive performance (error behavioural space) and decision-making processes (counterfactual behavioural space). Our experiments demonstrate that FBPs provide informative trajectories describing the evolving states of clients and their contributions to the global model, thereby enabling the identification of clusters of clients with similar behaviours. Leveraging the patterns identified by FBPs, we propose a robust aggregation technique named Federated Behavioural Shields to detect malicious or noisy client models, thereby enhancing security and surpassing the efficacy of existing state-of-the-art FL defense mechanisms. Our code is publicly available on GitHub.",
            "introduction": "   1 Introduction  Federated Learning (FL), a privacy-aware deep learning (DL) approach in distributed environments, is a dynamic system where many clients collaborate to train a model without sharing sensitive data, thus mitigating privacy risks\u00a0[1, 2]. Analyzing the behaviour of FL systems is crucial to detect anomalies\u2014such as distribution shifts [3, 4], biased data [5], or adversarial clients [6, 7, 8, 9]\u2014which may compromise the global model\u2019s predictive performance and introduce biases into its decision-making process. Various strategies have been developed to detect FL anomalies [10, 11, 12, 13, 14, 15]. However, existing techniques are not designed to track, visualise, and explain how client behaviours affect the performance of the global model, thus limiting human trust and control on FL dynamics.   Various studies have analysed models\u2019 behaviour in terms of predictive performance and decision-making processes independently. Predictive performance behaviour is primarily investigated in the context of non-linear optimisation using behavioural spaces\u00a0[16, 17, 18]. This technique allows the visualisation of the predictive diversity of a set of regression models by considering the vector of errors that each model produces on a set of samples. The decision-making process is studied mainly in explainable AI (XAI) research using, for example, counterfactual explanations\u00a0[19]. Counterfactuals can be used to identify relevant input features used by a model to make predictions, thus describing the position and orientation of the model\u2019s decision boundaries. However, these techniques have not yet been applied to FL systems, resulting in a lack of insight into how the behaviour of individual clients affects the overall model\u2019s accuracy and decision-making capabilities, leading to inefficiency in the training process.   To bridge this gap, we introduce Federated Behavioural Planes (FBPs), a method designed to visualise, explain, and give insights into the dynamics of FL systems. Our key innovation involves the creation of two behavioural planes for FL clients: one to highlight their predictive diversity and another to emphasise their decision-making process diversity via counterfactuals. Building on the client behaviour information provided by FBPs, to show their practical utility, we propose Federated Behavioural Shields, a robust aggregation mechanism that enhances security against malicious or noisy clients by accurately weighting the client models according to their constructive contributions during training.   Figure 1: The Federated Behavioural Planes framework enables the visualization of client behaviour in FL from two perspectives: predictive performance (Error Behavioural Plane) and decision-making processes (Counterfactuals Behavioural Plane). It highlights client trajectories and similarities, offering insights into client interactions and supporting the introduction of a new and effective robust aggregation mechanism with performance that surpasses state-of-the-art baselines.   The results of our experiments demonstrate that: (i) counterfactual generators jointly trained with FL systems produce valid and client-specific counterfactual explanations which effectively describe clients\u2019 decision-making diversity; (ii) FBPs facilitate the identification of clusters of clients with similar behaviours (e.g., normal vs. outlier clients), allowing for tracking of their trajectories during the entire training; (iii) Federated Behavioural Shields surpasses existing state-of-the-art defense mechanisms, demonstrating that the information contained in FBPs provides valuable descriptors of client behaviour.     2 Background  Federated learning.  FL systems\u00a0[20, 3] involve a network of K\u2208\ud835\udca9\ud835\udc3e\ud835\udca9K\\in\\mathcal{N}italic_K \u2208 caligraphic_N clients, coordinated by a central server, which collaboratively train a",
            "references": [
                {
                    "title": "Bias Propagation in Federated Learning",
                    "abstract": "We show that participating in federated learning can be detrimental to group fairness. In fact, the bias of a few parties against under-represented groups (identified by sensitive attributes such as gender or race) can propagate through the network to all the parties in the network. We analyze and explain bias propagation in federated learning on naturally partitioned real-world datasets. Our analysis reveals that biased parties unintentionally yet stealthily encode their bias in a small number of model parameters, and throughout the training, they steadily increase the dependence of the global model on sensitive attributes. What is important to highlight is that the experienced bias in federated learning is higher than what parties would otherwise encounter in centralized training with a model trained on the union of all their data. This indicates that the bias is due to the algorithm. Our work calls for auditing group fairness in federated learning and designing learning algorithms that are robust to bias propagation."
                },
                {
                    "title": "Combining homomorphic encryption and differential privacy in federated learning",
                    "abstract": "Recent works have investigated the relevance and practicality of using techniques such as Differential Privacy (DP) or Homomorphic Encryption (HE) to strengthen training data privacy in the context of Federated Learning protocols. As these two techniques cover different sources of confidentiality threats (other clients/end-users for the former, aggregation server for the latter), there is a need to consistently combine them in order to bridge the gap towards more realistic deployment scenarios. In this paper, we achieve that goal by means of a novel stochastic quantization operator which allows us to establish DP guarantees when the noise is both quantized and bounded due to the use of HE. The paper is concluded by experiments on the FEMNIST dataset which show that the precision required to get state-of-the art privacy/utility trade-off (which directly impacts HE parameters and, hence, HE operations performances) results in a computation time overhead between 0.2% and 1.1% imputable to HE (depending on the key setup, either single key or threshold), for the whole training of a 500k parameters model and state-of-the-art privacy/utility trade-off."
                },
                {
                    "title": "VCNet: A self-explaining model for realistic counterfactual generation",
                    "abstract": "Counterfactual explanation is a common class of methods to make local explanations of machine learning decisions. For a given instance, these methods aim to find the smallest modification of feature values that changes the predicted decision made by a machine learning model. One of the challenges of counterfactual explanation is the efficient generation of realistic counterfactuals. To address this challenge, we propose VCNet-Variational Counter Net-a model architecture that combines a predictor and a counterfactual generator that are jointly trained, for regression or classification tasks. VCNet is able to both generate predictions, and to generate counterfactual explanations without having to solve another minimisation problem. Our contribution is the generation of counterfactuals that are close to the distribution of the predicted class. This is done by learning a variational autoencoder conditionally to the output of the predictor in a join-training fashion. We present an empirical evaluation on tabular datasets and across several interpretability metrics. The results are competitive with the state-of-the-art method."
                },
                {
                    "title": "FLIS: Clustered Federated Learning Via Inference Similarity for Non-IID Data Distribution",
                    "abstract": "Conventional federated learning (FL) approaches are ineffective in scenarios where clients have significant differences in the distributions of their local data. The Non-IID data distribution in the client data causes a drift in the local model updates from the global optima, which significantly impacts the performance of the trained models. In this article, we present a new algorithm called FLIS that aims to address this problem by grouping clients into clusters that have jointly trainable data distributions. This is achieved by comparing the inference similarity of client models. Our proposed framework captures settings where different groups of users may have their own objectives (learning tasks), but by aggregating their data with others in the same cluster (same learning task), superior models can be derived via more efficient and personalized federated learning. We present experimental results to demonstrate the benefits of FLIS over the state-of-the-art approaches on the CIFAR-100/10, SVHN, and FMNIST datasets. Our code is available at https://github.com/MMorafah/FLIS."
                },
                {
                    "title": "HetVis: A Visual Analysis Approach for Identifying Data Heterogeneity in Horizontal Federated Learning",
                    "abstract": "Horizontal federated learning (HFL) enables distributed clients to train a shared model and keep their data privacy. In training high-quality HFL models, the data heterogeneity among clients is one of the major concerns. However, due to the security issue and the complexity of deep learning models, it is challenging to investigate data heterogeneity across different clients. To address this issue, based on a requirement analysis we developed a visual analytics tool, HetVis, for participating clients to explore data heterogeneity. We identify data heterogeneity through comparing prediction behaviors of the global federated model and the stand-alone model trained with local data. Then, a context-aware clustering of the inconsistent records is done, to provide a summary of data heterogeneity. Combining with the proposed comparison techniques, we develop a novel set of visualizations to identify heterogeneity issues in HFL. We designed three case studies to introduce how HetVis can assist client analysts in understanding different types of heterogeneity issues. Expert reviews and a comparative study demonstrate the effectiveness of HetVis."
                },
                {
                    "title": "Challenges and Approaches for Mitigating Byzantine Attacks in Federated Learning",
                    "abstract": "Recently emerged federated learning (FL) is an attractive distributed learning framework in which numerous wireless end-user devices can train a global model with the data remained autochthonous. Compared with the traditional machine learning framework that collects user data for centralized storage, which brings huge communication burden and concerns about data privacy, this approach can not only save the network bandwidth but also protect the data privacy. Despite the promising prospect, Byzantine attack, an intractable threat in conventional distributed network, is discovered to be rather efficacious against FL as well. In this paper, we conduct a comprehensive investigation of the state-of-the-art strategies for defending against Byzantine attacks in FL. We first provide a taxonomy for the existing defense solutions according to the techniques they used, followed by an across-the-board comparison and discussion. Then we propose a new Byzantine attack method called weight attack to defeat those defense schemes, and conduct experiments to demonstrate its threat. The results show that existing defense solutions, although abundant, are still far from fully protecting FL. Finally, we indicate possible countermeasures for weight attack, and highlight several challenges and future research directions for mitigating Byzantine attacks in FL."
                },
                {
                    "title": "K-Means Clustering",
                    "abstract": "The objective of this technique is to divide N observations with P dimensions (variables) into K clusters so that the within-cluster sum of squares is minimized. Since the number of possible arrangements is enormous, it is not practical to expect the best solution. Rather, this algorithm finds a \u201clocal\u201d optimum. This is a solution in which no movement of an observation from one cluster to another will reduce the within-cluster sum of squares. The algorithm may be repeated several times with different starting configurations. The optimum of these cluster solutions is then selected."
                },
                {
                    "title": "VADAF: Visualization for Abnormal Client Detection and Analysis in Federated Learning",
                    "abstract": "Federated Learning (FL) provides a powerful solution to distributed machine learning on a large corpus of decentralized data. It ensures privacy and security by performing computation on devices (which we refer to as clients) based on local data to improve the shared global model. However, the inaccessibility of the data and the invisibility of the computation make it challenging to interpret and analyze the training process, especially to distinguish potential client anomalies. Identifying these anomalies can help experts diagnose and improve FL models. For this reason, we propose a visual analytics system, VADAF, to depict the training dynamics and facilitate analyzing potential client anomalies. Specifically, we design a visualization scheme that supports massive training dynamics in the FL environment. Moreover, we introduce an anomaly detection method to detect potential client anomalies, which are further analyzed based on both the client model\u2019s visual and objective estimation. Three case studies have demonstrated the effectiveness of our system in understanding the FL training process and supporting abnormal client detection and analysis."
                },
                {
                    "title": "Flexible Clustered Federated Learning for Client-Level Data Distribution Shift",
                    "abstract": "Federated Learning (FL) enables the multiple participating devices to collaboratively contribute to a global neural network model while keeping the training data locally. Unlike the centralized training setting, the non-IID, imbalanced (statistical heterogeneity) and distribution shifted training data of FL is distributed in the federated network, which will increase the divergences between the local models and the global model, further degrading performance. In this paper, we propose a flexible clustered federated learning (CFL) framework named FlexCFL, in which we 1) group the training of clients based on the similarities between the clients\u2019 optimization directions for lower training divergence; 2) implement an efficient newcomer device cold start mechanism for framework scalability and practicality; 3) flexibly migrate clients to meet the challenge of client-level data distribution shift. FlexCFL can achieve improvements by dividing joint optimization into groups of sub-optimization and can strike a balance between accuracy and communication efficiency in the distribution shift environment. The convergence and complexity are analyzed to demonstrate the efficiency of FlexCFL. We also evaluate FlexCFL on several open datasets and made comparisons with related CFL frameworks. The results show that FlexCFL can significantly improve absolute test accuracy by <inline-formula><tex-math notation=\"LaTeX\">$+10.6\\%$</tex-math><alternatives><mml:math><mml:mrow><mml:mo>+</mml:mo><mml:mn>10</mml:mn><mml:mo>.</mml:mo><mml:mn>6</mml:mn><mml:mo>%</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href=\"liu-ieq1-3134263.gif\"/></alternatives></inline-formula> on FEMNIST compared with <italic>FedAvg</italic>, <inline-formula><tex-math notation=\"LaTeX\">$+3.5\\%$</tex-math><alternatives><mml:math><mml:mrow><mml:mo>+</mml:mo><mml:mn>3</mml:mn><mml:mo>.</mml:mo><mml:mn>5</mml:mn><mml:mo>%</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href=\"liu-ieq2-3134263.gif\"/></alternatives></inline-formula> on FashionMNIST compared with <italic>FedProx</italic>, <inline-formula><tex-math notation=\"LaTeX\">$+8.4\\%$</tex-math><alternatives><mml:math><mml:mrow><mml:mo>+</mml:mo><mml:mn>8</mml:mn><mml:mo>.</mml:mo><mml:mn>4</mml:mn><mml:mo>%</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href=\"liu-ieq3-3134263.gif\"/></alternatives></inline-formula> on MNIST compared with <italic>FeSEM</italic>, <inline-formula><tex-math notation=\"LaTeX\">$+4.7\\%$</tex-math><alternatives><mml:math><mml:mrow><mml:mo>+</mml:mo><mml:mn>4</mml:mn><mml:mo>.</mml:mo><mml:mn>7</mml:mn><mml:mo>%</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href=\"liu-ieq4-3134263.gif\"/></alternatives></inline-formula> on Sentiment140 compare with <italic>IFCA</italic>. The experiment results show that FlexCFL is also communication efficient in the distribution shift environment."
                },
                {
                    "title": "Inspecting the Running Process of Horizontal Federated Learning via Visual Analytics",
                    "abstract": "As a decentralized training approach, horizontal federated learning (HFL) enables distributed clients to collaboratively learn a machine learning model while keeping personal/private information on local devices. Despite the enhanced performance and efficiency of HFL over local training, clues for inspecting the behaviors of the participating clients and the federated model are usually lacking due to the privacy-preserving nature of HFL. Consequently, the users can only conduct a shallow-level analysis of potential abnormal behaviors and have limited means to assess the contributions of individual clients and implement the necessary intervention. Visualization techniques have been introduced to facilitate the HFL process inspection, usually by providing model metrics and evaluation results as a dashboard representation. Although the existing visualization methods allow a simple examination of the HFL model performance, they cannot support the intensive exploration of the HFL process. In this article, strictly following the HFL privacy-preserving protocol, we design an exploratory visual analytics system for the HFL process termed HFLens, which supports comparative visual interpretation at the overview, communication round, and client instance levels. Specifically, the proposed system facilitates the investigation of the overall process involving all clients, the correlation analysis of clients\u2019 information in one or different communication round(s), the identification of potential anomalies, and the contribution assessment of each HFL client. Two case studies confirm the efficacy of our system. Experts\u2019 feedback suggests that our approach indeed helps in understanding and diagnosing the HFL process better."
                },
                {
                    "title": "FLTrust: Byzantine-robust Federated Learning via Trust Bootstrapping",
                    "abstract": "Byzantine-robust federated learning aims to enable a service provider to learn an accurate global model when a bounded number of clients are malicious. The key idea of existing Byzantine-robust federated learning methods is that the service provider performs statistical analysis among the clients' local model updates and removes suspicious ones, before aggregating them to update the global model. However, malicious clients can still corrupt the global models in these methods via sending carefully crafted local model updates to the service provider. The fundamental reason is that there is no root of trust in existing federated learning methods. \nIn this work, we bridge the gap via proposing FLTrust, a new federated learning method in which the service provider itself bootstraps trust. In particular, the service provider itself collects a clean small training dataset (called root dataset) for the learning task and the service provider maintains a model (called server model) based on it to bootstrap trust. In each iteration, the service provider first assigns a trust score to each local model update from the clients, where a local model update has a lower trust score if its direction deviates more from the direction of the server model update. Then, the service provider normalizes the magnitudes of the local model updates such that they lie in the same hyper-sphere as the server model update in the vector space. Our normalization limits the impact of malicious local model updates with large magnitudes. Finally, the service provider computes the average of the normalized local model updates weighted by their trust scores as a global model update, which is used to update the global model. Our extensive evaluations on six datasets from different domains show that our FLTrust is secure against both existing attacks and strong adaptive attacks."
                },
                {
                    "title": "Privacy and Robustness in Federated Learning: Attacks and Defenses",
                    "abstract": "As data are increasingly being stored in different silos and societies becoming more aware of data privacy issues, the traditional centralized training of artificial intelligence (AI) models is facing efficiency and privacy challenges. Recently, federated learning (FL) has emerged as an alternative solution and continues to thrive in this new reality. Existing FL protocol designs have been shown to be vulnerable to adversaries within or outside of the system, compromising data privacy and system robustness. Besides training powerful global models, it is of paramount importance to design FL systems that have privacy guarantees and are resistant to different types of adversaries. In this article, we conduct a comprehensive survey on privacy and robustness in FL over the past five years. Through a concise introduction to the concept of FL and a unique taxonomy covering: 1) threat models; 2) privacy attacks and defenses; and 3) poisoning attacks and defenses, we provide an accessible review of this important topic. We highlight the intuitions, key techniques, and fundamental assumptions adopted by various attacks and defenses. Finally, we discuss promising future research directions toward robust and privacy-preserving FL, and their interplays with the multidisciplinary goals of FL."
                },
                {
                    "title": "Flower: A Friendly Federated Learning Research Framework",
                    "abstract": "Federated Learning (FL) has emerged as a promising technique for edge devices to collaboratively learn a shared prediction model, while keeping their training data on the device, thereby decoupling the ability to do machine learning from the need to store the data in the cloud. However, FL is difficult to implement realistically, both in terms of scale and systems heterogeneity. Although there are a number of research frameworks available to simulate FL algorithms, they do not support the study of scalable FL workloads on heterogeneous edge devices. In this paper, we present Flower -- a comprehensive FL framework that distinguishes itself from existing platforms by offering new facilities to execute large-scale FL experiments and consider richly heterogeneous FL device scenarios. Our experiments show Flower can perform FL experiments up to 15M in client size using only a pair of high-end GPUs. Researchers can then seamlessly migrate experiments to real devices to examine other parts of the design space. We believe Flower provides the community with a critical new tool for FL study and development."
                },
                {
                    "title": "Attack of the Tails: Yes, You Really Can Backdoor Federated Learning",
                    "abstract": "Due to its decentralized nature, Federated Learning (FL) lends itself to adversarial attacks in the form of backdoors during training. The goal of a backdoor is to corrupt the performance of the trained model on specific sub-tasks (e.g., by classifying green cars as frogs). A range of FL backdoor attacks have been introduced in the literature, but also methods to defend against them, and it is currently an open question whether FL systems can be tailored to be robust against backdoors. In this work, we provide evidence to the contrary. We first establish that, in the general case, robustness to backdoors implies model robustness to adversarial examples, a major open problem in itself. Furthermore, detecting the presence of a backdoor in a FL model is unlikely assuming first order oracles or polynomial time. We couple our theoretical results with a new family of backdoor attacks, which we refer to as edge-case backdoors. An edge-case backdoor forces a model to misclassify on seemingly easy inputs that are however unlikely to be part of the training, or test data, i.e., they live on the tail of the input distribution. We explain how these edge-case backdoors can lead to unsavory failures and may have serious repercussions on fairness, and exhibit that with careful tuning at the side of the adversary, one can insert them across a range of machine learning tasks (e.g., image classification, OCR, text prediction, sentiment analysis)."
                },
                {
                    "title": "Byzantine-Robust Learning on Heterogeneous Datasets via Bucketing",
                    "abstract": "In Byzantine robust distributed or federated learning, a central server wants to train a machine learning model over data distributed across multiple workers. However, a fraction of these workers may deviate from the prescribed algorithm and send arbitrary messages. While this problem has received significant attention recently, most current defenses assume that the workers have identical data. For realistic cases when the data across workers are heterogeneous (non-iid), we design new attacks which circumvent current defenses, leading to significant loss of performance. We then propose a simple bucketing scheme that adapts existing robust algorithms to heterogeneous datasets at a negligible computational cost. We also theoretically and experimentally validate our approach, showing that combining bucketing with existing robust algorithms is effective against challenging attacks. Our work is the first to establish guaranteed convergence for the non-iid Byzantine robust problem under realistic assumptions."
                },
                {
                    "title": "An Efficient Framework for Clustered Federated Learning",
                    "abstract": "We address the problem of federated learning (FL) where users are distributed and partitioned into clusters. This setup captures settings where different groups of users have their own objectives (learning tasks) but by aggregating their data with others in the same cluster (same learning task), they can leverage the strength in numbers in order to perform more efficient federated learning. For this new framework of clustered federated learning, we propose the Iterative Federated Clustering Algorithm (IFCA), which alternately estimates the cluster identities of the users and optimizes model parameters for the user clusters via gradient descent. We analyze the convergence rate of this algorithm first in a linear model with squared loss and then for generic strongly convex and smooth loss functions. We show that in both settings, with good initialization, IFCA is guaranteed to converge, and discuss the optimality of the statistical error rate. In particular, for the linear model with two clusters, we can guarantee that our algorithm converges as long as the initialization is slightly better than random. When the clustering structure is ambiguous, we propose to train the models by combining IFCA with the weight sharing technique in multi-task learning. In the experiments, we show that our algorithm can succeed even if we relax the requirements on initialization with random initialization and multiple restarts. We also present experimental results showing that our algorithm is efficient in non-convex problems such as neural networks. We demonstrate the benefits of IFCA over the baselines on several clustered FL benchmarks."
                },
                {
                    "title": "LDP-Fed: federated learning with local differential privacy",
                    "abstract": "This paper presents LDP-Fed, a novel federated learning system with a formal privacy guarantee using local differential privacy (LDP). Existing LDP protocols are developed primarily to ensure data privacy in the collection of single numerical or categorical values, such as click count in Web access logs. However, in federated learning model parameter updates are collected iteratively from each participant and consist of high dimensional, continuous values with high precision (10s of digits after the decimal point), making existing LDP protocols inapplicable. To address this challenge in LDP-Fed, we design and develop two novel approaches. First, LDP-Fed's LDP Module provides a formal differential privacy guarantee for the repeated collection of model training parameters in the federated training of large-scale neural networks over multiple individual participants' private datasets. Second, LDP-Fed implements a suite of selection and filtering techniques for perturbing and sharing select parameter updates with the parameter server. We validate our system deployed with a condensed LDP protocol in training deep neural networks on public data. We compare this version of LDP-Fed, coined CLDP-Fed, with other state-of-the-art approaches with respect to model accuracy, privacy preservation, and system capabilities."
                },
                {
                    "title": "Federated learning with hierarchical clustering of local updates to improve training on non-IID data",
                    "abstract": "Federated learning (FL) is a well established method for performing machine learning tasks over massively distributed data. However in settings where data is distributed in a non-iid (not independent and identically distributed) fashion - as is typical in real world situations - the joint model produced by FL suffers in terms of test set accuracy and/or communication costs compared to training on iid data. We show that learning a single joint model is often not optimal in the presence of certain types of non-iid data. In this work we present a modification to FL by introducing a hierarchical clustering step (FL+HC) to separate clusters of clients by the similarity of their local updates to the global joint model. Once separated, the clusters are trained independently and in parallel on specialised models. We present a robust empirical analysis of the hyperparameters for FL+HC for several iid and non-iid settings. We show how FL+HC allows model training to converge in fewer communication rounds (significantly so under some non-iid settings) compared to FL without clustering. Additionally, FL+HC allows for a greater percentage of clients to reach a target accuracy compared to standard FL. Finally we make suggestions for good default hyperparameters to promote superior performing specialised models without modifying the the underlying federated learning communication protocol."
                },
                {
                    "title": "Threats to Federated Learning: A Survey",
                    "abstract": "With the emergence of data silos and popular privacy awareness, the traditional centralized approach of training artificial intelligence (AI) models is facing strong challenges. Federated learning (FL) has recently emerged as a promising solution under this new reality. Existing FL protocol design has been shown to exhibit vulnerabilities which can be exploited by adversaries both within and without the system to compromise data privacy. It is thus of paramount importance to make FL system designers to be aware of the implications of future FL algorithm design on privacy-preservation. Currently, there is no survey on this topic. In this paper, we bridge this important gap in FL literature. By providing a concise introduction to the concept of FL, and a unique taxonomy covering threat models and two major attacks on FL: 1) poisoning attacks and 2) inference attacks, this paper provides an accessible review of this important topic. We highlight the intuitions, key techniques as well as fundamental assumptions adopted by various attacks, and discuss promising future research directions towards more robust privacy preservation in FL."
                },
                {
                    "title": "Robust Aggregation for Federated Learning",
                    "abstract": "We present a novel approach to federated learning that endows its aggregation process with greater robustness to potential poisoning of local data or model parameters of participating devices. The proposed approach, Robust Federated Aggregation (RFA), relies on the aggregation of updates using the geometric median, which can be computed efficiently using a Weiszfeld-type algorithm. RFA is agnostic to the level of corruption and aggregates model updates without revealing each device\u2019s individual contribution. We establish the convergence of the robust federated learning algorithm for the stochastic learning of additive models with least squares. We also offer two variants of RFA: a faster one with one-step robust aggregation, and another one with on-device personalization. We present experimental results with additive models and deep networks for three tasks in computer vision and natural language processing. The experiments show that RFA is competitive with the classical aggregation when the level of corruption is low, while demonstrating greater robustness under high corruption."
                },
                {
                    "title": "Advances and Open Problems in Federated Learning",
                    "abstract": "Federated learning (FL) is a machine learning setting where many clients (e.g. mobile devices or whole organizations) collaboratively train a model under the orchestration of a central server (e.g. service provider), while keeping the training data decentralized. FL embodies the principles of focused data collection and minimization, and can mitigate many of the systemic privacy risks and costs resulting from traditional, centralized machine learning and data science approaches. Motivated by the explosive growth in FL research, this paper discusses recent advances and presents an extensive collection of open problems and challenges."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Learning under Concept Drift: A Review",
                    "abstract": "Concept drift describes unforeseeable changes in the underlying distribution of streaming data over time. Concept drift research involves the development of methodologies and techniques for drift detection, understanding, and adaptation. Data analysis has revealed that machine learning in a concept drift environment will result in poor learning results if the drift is not addressed. To help researchers identify which research topics are significant and how to apply related techniques in data analysis tasks, it is necessary that a high quality, instructive review of current research developments and trends in the concept drift field is conducted. In addition, due to the rapid development of concept drift in recent years, the methodologies of learning under concept drift have become noticeably systematic, unveiling a framework which has not been mentioned in literature. This paper reviews over 130 high quality publications in concept drift related research areas, analyzes up-to-date developments in methodologies and techniques, and establishes a framework of learning under concept drift including three main components: concept drift detection, concept drift understanding, and concept drift adaptation. This paper lists and discusses 10 popular synthetic datasets and 14 publicly available benchmark datasets used for evaluating the performance of learning algorithms aiming at handling concept drift. Also, concept drift related research directions are covered and discussed. By providing state-of-the-art knowledge, this survey will directly support researchers in their understanding of research developments in the field of learning under concept drift."
                },
                {
                    "title": "Free-riders in Federated Learning: Attacks and Defenses",
                    "abstract": "Federated learning is a recently proposed paradigm that enables multiple clients to collaboratively train a joint model. It allows clients to train models locally, and leverages the parameter server to generate a global model by aggregating the locally submitted gradient updates at each round. Although the incentive model for federated learning has not been fully developed, it is supposed that participants are able to get rewards or the privilege to use the final global model, as a compensation for taking efforts to train the model. Therefore, a client who does not have any local data has the incentive to construct local gradient updates in order to deceive for rewards. In this paper, we are the first to propose the notion of free rider attacks, to explore possible ways that an attacker may construct gradient updates, without any local training data. Furthermore, we explore possible defenses that could detect the proposed attacks, and propose a new high dimensional detection method called STD-DAGMM, which particularly works well for anomaly detection of model parameters. We extend the attacks and defenses to consider more free riders as well as differential privacy, which sheds light on and calls for future research in this field."
                },
                {
                    "title": "Local Model Poisoning Attacks to Byzantine-Robust Federated Learning",
                    "abstract": "In federated learning, multiple client devices jointly learn a machine learning model: each client device maintains a local model for its local training dataset, while a master device maintains a global model via aggregating the local models from the client devices. The machine learning community recently proposed several federated learning methods that were claimed to be robust against Byzantine failures (e.g., system failures, adversarial manipulations) of certain client devices. In this work, we perform the first systematic study on local model poisoning attacks to federated learning. We assume an attacker has compromised some client devices, and the attacker manipulates the local model parameters on the compromised client devices during the learning process such that the global model has a large testing error rate. We formulate our attacks as optimization problems and apply our attacks to four recent Byzantine-robust federated learning methods. Our empirical results on four real-world datasets show that our attacks can substantially increase the error rates of the models learnt by the federated learning methods that were claimed to be robust against Byzantine failures of some client devices. We generalize two defenses for data poisoning attacks to defend against our local model poisoning attacks. Our evaluation results show that one defense can effectively defend against our attacks in some cases, but the defenses are not effective enough in other cases, highlighting the need for new defenses against our local model poisoning attacks to federated learning."
                },
                {
                    "title": "Abnormal Client Behavior Detection in Federated Learning",
                    "abstract": "In federated learning systems, clients are autonomous in that their behaviors are not fully governed by the server. Consequently, a client may intentionally or unintentionally deviate from the prescribed course of federated model training, resulting in abnormal behaviors, such as turning into a malicious attacker or a malfunctioning client. Timely detecting those anomalous clients is therefore critical to minimize their adverse impacts. In this work, we propose to detect anomalous clients at the server side. In particular, we generate low-dimensional surrogates of model weight vectors and use them to perform anomaly detection. We evaluate our solution through experiments on image classification model training over the FEMNIST dataset. Experimental results show that the proposed detection-based approach significantly outperforms the conventional defense-based methods."
                },
                {
                    "title": "Learning Model-Agnostic Counterfactual Explanations for Tabular Data",
                    "abstract": "Counterfactual explanations can be obtained by identifying the smallest change made to an input vector to influence a prediction in a positive way from a user\u2019s viewpoint; for example, from \u2019loan rejected\u2019 to \u2019awarded\u2019 or from \u2019high risk of cardiovascular disease\u2019 to \u2019low risk\u2019. Previous approaches would not ensure that the produced counterfactuals be proximate (i.e., not local outliers) and connected to regions with substantial data density (i.e., close to correctly classified observations), two requirements known as counterfactual faithfulness. Our contribution is twofold. First, drawing ideas from the manifold learning literature, we develop a framework, called C-CHVAE, that generates faithful counterfactuals. Second, we suggest to complement the catalog of counterfactual quality measures using a criterion to quantify the degree of difficulty for a certain counterfactual suggestion. Our real world experiments suggest that faithful counterfactuals come at the cost of higher degrees of difficulty."
                },
                {
                    "title": "Clustered Federated Learning: Model-Agnostic Distributed Multitask Optimization Under Privacy Constraints",
                    "abstract": "Federated learning (FL) is currently the most widely adopted framework for collaborative training of (deep) machine learning models under privacy constraints. Albeit its popularity, it has been observed that FL yields suboptimal results if the local clients\u2019 data distributions diverge. To address this issue, we present clustered FL (CFL), a novel federated multitask learning (FMTL) framework, which exploits geometric properties of the FL loss surface to group the client population into clusters with jointly trainable data distributions. In contrast to existing FMTL approaches, CFL does not require any modifications to the FL communication protocol to be made, is applicable to general nonconvex objectives (in particular, deep neural networks), does not require the number of clusters to be known a priori, and comes with strong mathematical guarantees on the clustering quality. CFL is flexible enough to handle client populations that vary over time and can be implemented in a privacy-preserving way. As clustering is only performed after FL has converged to a stationary point, CFL can be viewed as a postprocessing method that will always achieve greater or equal performance than conventional FL by allowing clients to arrive at more specialized models. We verify our theoretical analysis in experiments with deep convolutional and recurrent neural networks on commonly used FL data sets."
                },
                {
                    "title": "Fall of Empires: Breaking Byzantine-tolerant SGD by Inner Product Manipulation",
                    "abstract": "Recently, new defense techniques have been developed to tolerate Byzantine failures for distributed machine learning. The Byzantine model captures workers that behave arbitrarily, including malicious and compromised workers. In this paper, we break two prevailing Byzantine-tolerant techniques. Specifically we show robust aggregation methods for synchronous SGD -- coordinate-wise median and Krum -- can be broken using new attack strategies based on inner product manipulation. We prove our results theoretically, as well as show empirical validation."
                },
                {
                    "title": "A Little Is Enough: Circumventing Defenses For Distributed Learning",
                    "abstract": "Distributed learning is central for large-scale training of deep-learning models. However, they are exposed to a security threat in which Byzantine participants can interrupt or control the learning process. Previous attack models and their corresponding defenses assume that the rogue participants are (a) omniscient (know the data of all other participants), and (b) introduce large change to the parameters. We show that small but well-crafted changes are sufficient, leading to a novel non-omniscient attack on distributed learning that go undetected by all existing defenses. We demonstrate our attack method works not only for preventing convergence but also for repurposing of the model behavior (backdooring). We show that 20% of corrupt workers are sufficient to degrade a CIFAR10 model accuracy by 50%, as well as to introduce backdoors into MNIST and CIFAR10 models without hurting their accuracy"
                },
                {
                    "title": "SentiNet: Detecting Localized Universal Attacks Against Deep Learning Systems",
                    "abstract": "SentiNet is a novel detection framework for localized universal attacks on neural networks. These attacks restrict adversarial noise to contiguous portions of an image and are reusable with different images-constraints that prove useful for generating physically-realizable attacks. Unlike most other works on adversarial detection, SentiNet does not require training a model or preknowledge of an attack prior to detection. Our approach is appealing due to the large number of possible mechanisms and attack-vectors that an attack-specific defense would have to consider. By leveraging the neural network's susceptibility to attacks and by using techniques from model interpretability and object detection as detection mechanisms, SentiNet turns a weakness of a model into a strength. We demonstrate the effectiveness of SentiNet on three different attacks-i.e., data poisoning attacks, trojaned networks, and adversarial patches (including physically realizable attacks)-and show that our defense is able to achieve very competitive performance metrics for all three threats. Finally, we show that SentiNet is robust against strong adaptive adversaries, who build adversarial patches that specifically target the components of SentiNet's architecture."
                },
                {
                    "title": "Analyzing Federated Learning through an Adversarial Lens",
                    "abstract": "Federated learning distributes model training among a multitude of agents, who, guided by privacy concerns, perform training using their local data but share only model parameter updates, for iterative aggregation at the server. In this work, we explore the threat of model poisoning attacks on federated learning initiated by a single, non-colluding malicious agent where the adversarial objective is to cause the model to misclassify a set of chosen inputs with high confidence. We explore a number of strategies to carry out this attack, starting with simple boosting of the malicious agent's update to overcome the effects of other agents' updates. To increase attack stealth, we propose an alternating minimization strategy, which alternately optimizes for the training loss and the adversarial objective. We follow up by using parameter estimation for the benign agents' updates to improve on attack success. Finally, we use a suite of interpretability techniques to generate visual explanations of model decisions for both benign and malicious models and show that the explanations are nearly visually indistinguishable. Our results indicate that even a highly constrained adversary can carry out model poisoning attacks while simultaneously maintaining stealth, thus highlighting the vulnerability of the federated learning setting and the need to develop effective defense strategies."
                },
                {
                    "title": "How To Backdoor Federated Learning",
                    "abstract": "Federated learning enables thousands of participants to construct a deep learning model without sharing their private training data with each other. For example, multiple smartphones can jointly train a next-word predictor for keyboards without revealing what individual users type. We demonstrate that any participant in federated learning can introduce hidden backdoor functionality into the joint global model, e.g., to ensure that an image classifier assigns an attacker-chosen label to images with certain features, or that a word predictor completes certain sentences with an attacker-chosen word. \nWe design and evaluate a new model-poisoning methodology based on model replacement. An attacker selected in a single round of federated learning can cause the global model to immediately reach 100% accuracy on the backdoor task. We evaluate the attack under different assumptions for the standard federated-learning tasks and show that it greatly outperforms data poisoning. Our generic constrain-and-scale technique also evades anomaly detection-based defenses by incorporating the evasion into the attacker's loss function during training."
                },
                {
                    "title": "Zeno: Distributed Stochastic Gradient Descent with Suspicion-based Fault-tolerance",
                    "abstract": "We present Zeno, a technique to make distributed machine learning, particularly Stochastic Gradient Descent (SGD), tolerant to an arbitrary number of faulty workers. Zeno generalizes previous results that assumed a majority of non-faulty nodes; we need assume only one non-faulty worker. Our key idea is to suspect workers that are potentially defective. Since this is likely to lead to false positives, we use a ranking-based preference mechanism. We prove the convergence of SGD for non-convex problems under these scenarios. Experimental results show that Zeno outperforms existing approaches."
                },
                {
                    "title": "Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networks",
                    "abstract": "Data poisoning is an attack on machine learning models wherein the attacker adds examples to the training set to manipulate the behavior of the model at test time. This paper explores poisoning attacks on neural nets. The proposed attacks use \"clean-labels\"; they don't require the attacker to have any control over the labeling of training data. They are also targeted; they control the behavior of the classifier on a $\\textit{specific}$ test instance without degrading overall classifier performance. For example, an attacker could add a seemingly innocuous image (that is properly labeled) to a training set for a face recognition engine, and control the identity of a chosen person at test time. Because the attacker does not need to control the labeling function, poisons could be entered into the training set simply by leaving them on the web and waiting for them to be scraped by a data collection bot. \nWe present an optimization-based method for crafting poisons, and show that just one single poison image can control classifier behavior when transfer learning is used. For full end-to-end training, we present a \"watermarking\" strategy that makes poisoning reliable using multiple ($\\approx$50) poisoned training instances. We demonstrate our method by generating poisoned frog images from the CIFAR dataset and using them to manipulate image classifiers."
                },
                {
                    "title": "Learning to Reweight Examples for Robust Deep Learning",
                    "abstract": "Deep neural networks have been shown to be very powerful modeling tools for many supervised learning tasks involving complex input patterns. However, they can also easily overfit to training set biases and label noises. In addition to various regularizers, example reweighting algorithms are popular solutions to these problems, but they require careful tuning of additional hyperparameters, such as example mining schedules and regularization hyperparameters. In contrast to past reweighting methods, which typically consist of functions of the cost value of each example, in this work we propose a novel meta-learning algorithm that learns to assign weights to training examples based on their gradient directions. To determine the example weights, our method performs a meta gradient descent step on the current mini-batch example weights (which are initialized from zero) to minimize the loss on a clean unbiased validation set. Our proposed method can be easily implemented on any type of deep network, does not require any additional hyperparameter tuning, and achieves impressive performance on class imbalance and corrupted label problems where only a small amount of clean validation data is available."
                },
                {
                    "title": "Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates",
                    "abstract": "In large-scale distributed learning, security issues have become increasingly important. Particularly in a decentralized environment, some computing units may behave abnormally, or even exhibit Byzantine failures---arbitrary and potentially adversarial behavior. In this paper, we develop distributed learning algorithms that are provably robust against such failures, with a focus on achieving optimal statistical performance. A main result of this work is a sharp analysis of two robust distributed gradient descent algorithms based on median and trimmed mean operations, respectively. We prove statistical error rates for three kinds of population loss functions: strongly convex, non-strongly convex, and smooth non-convex. In particular, these algorithms are shown to achieve order-optimal statistical error rates for strongly convex losses. To achieve better communication efficiency, we further propose a median-based distributed algorithm that is provably robust, and uses only one communication round. For strongly convex quadratic loss, we show that this algorithm achieves the same optimal error rate as the robust distributed gradient descent algorithms."
                },
                {
                    "title": "The Hidden Vulnerability of Distributed Learning in Byzantium",
                    "abstract": "While machine learning is going through an era of celebrated success, concerns have been raised about the vulnerability of its backbone: stochastic gradient descent (SGD). Recent approaches have been proposed to ensure the robustness of distributed SGD against adversarial (Byzantine) workers sending poisoned gradients during the training phase. Some of these approaches have been proven Byzantine-resilient: they ensure the convergence of SGD despite the presence of a minority of adversarial workers. \nWe show in this paper that convergence is not enough. In high dimension $d \\gg 1$, an adver\\-sary can build on the loss function's non-convexity to make SGD converge to ineffective models. More precisely, we bring to light that existing Byzantine-resilient schemes leave a margin of poisoning of $\\Omega\\left(f(d)\\right)$, where $f(d)$ increases at least like $\\sqrt{d~}$. Based on this leeway, we build a simple attack, and experimentally show its strong to utmost effectivity on CIFAR-10 and MNIST. \nWe introduce Bulyan, and prove it significantly reduces the attackers leeway to a narrow $O( \\frac{1}{\\sqrt{d~}})$ bound. We empirically show that Bulyan does not suffer the fragility of existing aggregation rules and, at a reasonable cost in terms of required batch size, achieves convergence as if only non-Byzantine gradients had been used to update the model."
                },
                {
                    "title": "Distributed Statistical Machine Learning in Adversarial Settings: Byzantine Gradient Descent",
                    "abstract": "We consider the distributed statistical learning problem over decentralized systems that are prone to adversarial attacks. This setup arises in many practical applications, including Google's Federated Learning. Formally, we focus on a decentralized system that consists of a parameter server and m working machines; each working machine keeps N/m data samples, where N is the total number of samples. In each iteration, up to q of the m working machines suffer Byzantine faults -- a faulty machine in the given iteration behaves arbitrarily badly against the system and has complete knowledge of the system. Additionally, the sets of faulty machines may be different across iterations. Our goal is to design robust algorithms such that the system can learn the underlying true parameter, which is of dimension d, despite the interruption of the Byzantine attacks. In this paper, based on the geometric median of means of the gradients, we propose a simple variant of the classical gradient descent method. We show that our method can tolerate q Byzantine failures up to 2(1+\u03b5)q \u0142e m for an arbitrarily small but fixed constant \u03b5>0. The parameter estimate converges in O(\u0142og N) rounds with an estimation error on the order of max \u221adq/N, ~\u221ad/N , which is larger than the minimax-optimal error rate \u221ad/N in the centralized and failure-free setting by at most a factor of \u221aq . The total computational complexity of our algorithm is of O((Nd/m) log N) at each working machine and O(md + kd log 3 N) at the central server, and the total communication cost is of O(m d log N). We further provide an application of our general results to the linear regression problem. A key challenge arises in the above problem is that Byzantine failures create arbitrary and unspecified dependency among the iterations and the aggregated gradients. To handle this issue in the analysis, we prove that the aggregated gradient, as a function of model parameter, converges uniformly to the true gradient function."
                },
                {
                    "title": "Machine Learning with Adversaries: Byzantine Tolerant Gradient Descent",
                    "abstract": "We study the resilience to Byzantine failures of distributed implementations of Stochastic Gradient Descent (SGD). So far, distributed machine learning frameworks have largely ignored the possibility of failures, especially arbitrary (i.e., Byzantine) ones. Causes of failures include software bugs, network asynchrony, biases in local datasets, as well as attackers trying to compromise the entire system. Assuming a set of n workers, up to f being Byzantine, we ask how resilient can SGD be, without limiting the dimension, nor the size of the parameter space. We first show that no gradient aggregation rule based on a linear combination of the vectors proposed by the workers (i.e, current approaches) tolerates a single Byzantine failure. We then formulate a resilience property of the aggregation rule capturing the basic requirements to guarantee convergence despite f Byzantine workers. We propose Krum, an aggregation rule that satisfies our resilience property, which we argue is the first provably Byzantine-resilient algorithm for distributed SGD. We also report on experimental evaluations of Krum."
                },
                {
                    "title": "Counterfactual Explanations Without Opening the Black Box: Automated Decisions and the GDPR",
                    "abstract": "There has been much discussion of the right to explanation in the EU General Data Protection Regulation, and its existence, merits, and disadvantages. Implementing a right to explanation that opens the black box of algorithmic decision-making faces major legal and technical barriers. Explaining the functionality of complex algorithmic decision-making systems and their rationale in specific cases is a technically challenging problem. Some explanations may offer little meaningful information to data subjects, raising questions around their value. Explanations of automated decisions need not hinge on the general public understanding how algorithmic systems function. Even though such interpretability is of great importance and should be pursued, explanations can, in principle, be offered without opening the black box. Looking at explanations as a means to help a data subject act rather than merely understand, one could gauge the scope and content of explanations according to the specific goal or action they are intended to support. From the perspective of individuals affected by automated decision-making, we propose three aims for explanations: (1) to inform and help the individual understand why a particular decision was reached, (2) to provide grounds to contest the decision if the outcome is undesired, and (3) to understand what would need to change in order to receive a desired result in the future, based on the current decision-making model. We assess how each of these goals finds support in the GDPR. We suggest data controllers should offer a particular type of explanation, unconditional counterfactual explanations, to support these three aims. These counterfactual explanations describe the smallest change to the world that can be made to obtain a desirable outcome, or to arrive at the closest possible world, without needing to explain the internal logic of the system."
                },
                {
                    "title": "Communication-Efficient Learning of Deep Networks from Decentralized Data",
                    "abstract": "Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data center and training there using conventional approaches. We advocate an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. \nWe present a practical method for the federated learning of deep networks based on iterative model averaging, and conduct an extensive empirical evaluation, considering five different model architectures and four datasets. These experiments demonstrate the approach is robust to the unbalanced and non-IID data distributions that are a defining characteristic of this setting. Communication costs are the principal constraint, and we show a reduction in required communication rounds by 10-100x as compared to synchronized stochastic gradient descent."
                },
                {
                    "title": "Illuminating search spaces by mapping elites",
                    "abstract": "Many fields use search algorithms, which automatically explore a search space to find high-performing solutions: chemists search through the space of molecules to discover new drugs; engineers search for stronger, cheaper, safer designs, scientists search for models that best explain data, etc. The goal of search algorithms has traditionally been to return the single highest-performing solution in a search space. Here we describe a new, fundamentally different type of algorithm that is more useful because it provides a holistic view of how high-performing solutions are distributed throughout a search space. It creates a map of high-performing solutions at each point in a space defined by dimensions of variation that a user gets to choose. This Multi-dimensional Archive of Phenotypic Elites (MAP-Elites) algorithm illuminates search spaces, allowing researchers to understand how interesting attributes of solutions combine to affect performance, either positively or, equally of interest, negatively. For example, a drug company may wish to understand how performance changes as the size of molecules and their cost-to-produce vary. MAP-Elites produces a large diversity of high-performing, yet qualitatively different solutions, which can be more helpful than a single, high-performing solution. Interestingly, because MAP-Elites explores more of the search space, it also tends to find a better overall solution than state-of-the-art search algorithms. We demonstrate the benefits of this new algorithm in three different problem domains ranging from producing modular neural networks to designing simulated and real soft robots. Because MAP- Elites (1) illuminates the relationship between performance and dimensions of interest in solutions, (2) returns a set of high-performing, yet diverse solutions, and (3) improves finding a single, best solution, it will advance science and engineering."
                },
                {
                    "title": "The MNIST Database of Handwritten Digit Images for Machine Learning Research [Best of the Web]",
                    "abstract": "In this issue, \u201cBest of the Web\u201d presents the modified National Institute of Standards and Technology (MNIST) resources, consisting of a collection of handwritten digit images used extensively in optical character recognition and machine learning research."
                },
                {
                    "title": "Scikit-learn: Machine Learning in Python",
                    "abstract": "Scikit-learn is a Python module integrating a wide range of state-of-the-art machine learning algorithms for medium-scale supervised and unsupervised problems. This package focuses on bringing machine learning to non-specialists using a general-purpose high-level language. Emphasis is put on ease of use, performance, documentation, and API consistency. It has minimal dependencies and is distributed under the simplified BSD license, encouraging its use in both academic and commercial settings. Source code, binaries, and documentation can be downloaded from http://scikit-learn.sourceforge.net."
                },
                {
                    "title": "Matplotlib: A 2D Graphics Environment",
                    "abstract": "Matplotlib is a 2D graphics package used for Python for application development, interactive scripting,and publication-quality image generation across user interfaces and operating systems"
                },
                {
                    "title": "Attacks against Federated Learning Defense Systems and their Mitigation",
                    "abstract": "The susceptibility of federated learning (FL) to attacks from untrustworthy endpoints has led to the design of several defense systems. FL defense systems enhance the federated optimization algorithm using anomaly detection, scaling the updates from endpoints depending on their anomalous behavior. However, the defense systems themselves may be exploited by the endpoints with more sophisticated attacks. First, this paper proposes three categories of attacks and shows that they can e\ufb00ectively deceive some well-known FL defense systems. In the \ufb01rst two categories, referred to as on-o\ufb00 attacks, the adversary toggles between being honest and engaging in attacks. We analyse two such on-o\ufb00 attacks, label \ufb02ipping and free riding, and show their impact against existing FL defense systems. As a third category, we propose attacks based on \u201cgood mouthing\u201d and \u201cbad mouthing\u201d, to boost or diminish in\ufb02uence of the victim endpoints on the global model. Secondly, we propose a new federated optimization algorithm, Viceroy, that can successfully mitigate all the proposed attacks. The proposed attacks and the mitigation strategy have been tested on a number of di\ufb00erent experiments establishing their e\ufb00ectiveness in comparison with other contemporary methods. The proposed algorithm has also been made available as open source. Finally, in the appendices, we provide an induction proof for the on-o\ufb00 model poisoning attack, and the proof of convergence and adversarial tolerance for the new federated optimization algorithm."
                },
                {
                    "title": "Adaptive Client Clustering for Efficient Federated Learning over Non-IID and Imbalanced Data",
                    "abstract": "\u2014Federated learning (FL) is an emerging distributed and privacy-preserving machine learning framework. However, the performance of traditional FL methods is seriously impaired by the real-world data, which appear to be non-independent and identically distributed (non-IID). The recent clustered federated learning (CFL) methods eliminate the impact of non-IID data by grouping clients with similar data distribution into the same cluster. Unfortunately, existing CFL methods heavily rely on the pre-setting of the cluster number, failing to achieve adaptive client clustering. Even worse, we experimentally observe that imbalanced data across clients largely degrade their correctness of client clustering. In this paper, we present a novel CFL method without manual intervention, named AutoCFL, which can eliminate both effects of non-IID and imbalanced data simultaneously. To deal with imbalanced data, the local training adjustment strategy adaptively adjusts the number of local training epochs for each client. To further improve the clustering correctness and adaptability, the weighted voting-based client clustering strategy automatically groups each client into an appropriate cluster. Extensive experiments are conducted to evaluate the design of AutoCFL with three popular datasets under various data settings. Experimental results demonstrate that AutoCFL outperforms the state-of-the-art methods under non-IID and imbalanced data settings, e.g. , on average improving the model accuracy by 9 . 24% when compared to the standard FL method, i.e. , FedAvg , while significantly reducing communication costs by 4 . 67 \u00d7 in an adaptive client clustering manner."
                },
                {
                    "title": "Manipulating the Byzantine: Optimizing Model Poisoning Attacks and Defenses for Federated Learning",
                    "abstract": "\u2014Federated learning (FL) enables many data owners (e.g., mobile devices) to train a joint ML model (e.g., a next-word prediction classi\ufb01er) without the need of sharing their private training data. However, FL is known to be susceptible to poisoning attacks by malicious participants (e.g., adversary-owned mobile devices) who aim at hampering the accuracy of the jointly trained model through sending malicious inputs during the federated training process. In this paper, we present a generic framework for model poisoning attacks on FL. We show that our framework leads to poisoning attacks that substantially outperform state-of-the-art model poisoning attacks by large margins. For instance, our attacks result in 1 . 5 \u00d7 to 60 \u00d7 higher reductions in the accuracy of FL models compared to previously discovered poisoning attacks. Our work demonstrates that existing Byzantine-robust FL algorithms are signi\ufb01cantly more susceptible to model poisoning than previously thought. Motivated by this, we design a defense against FL poisoning, called divide-and-conquer (DnC). We demonstrate that DnC outperforms all existing Byzantine-robust FL algorithms in defeating model poisoning attacks, speci\ufb01cally, it is 2 . 5 \u00d7 to 12 \u00d7 more resilient in our experiments with different datasets and models."
                },
                {
                    "title": "Seaborn: Statistical Data Visualization",
                    "abstract": "seaborn is a library for making statistical graphics in Python. It provides a high-level interface to matplotlib and integrates closely with pandas data structures. Functions in the seaborn library expose a declarative, dataset-oriented API that makes it easy to translate questions about data into graphics that can answer them. When given a dataset and a specification of the plot to make, seaborn automatically maps the data values to visual attributes such as color, size, or style, internally computes statistical transformations, and decorates the plot with informative axis labels and a legend. Many seaborn functions can generate figures with multiple panels that elicit comparisons between conditional subsets of data or across different pairings of variables in a dataset. seaborn is designed to be useful throughout the lifecycle of a scientific project. By producing complete graphics from a single function call with minimal arguments, seaborn facilitates rapid prototyping and exploratory data analysis. And by offering extensive options for customization, along with exposing the underlying matplotlib objects, it can be used to create polished, publication-quality figures."
                },
                {
                    "title": "The Limitations of Federated Learning in Sybil Settings",
                    "abstract": "Federated learning over distributed multi-party data is an emerging paradigm that iteratively aggregates updates from a group of devices to train a globally shared model. Relying on a set of devices, however, opens up the door for sybil attacks: malicious devices may be controlled by a single adversary who directs these devices to attack the system. We consider the susceptibility of federated learning to sybil attacks and propose a taxonomy of sybil objectives and strategies in this setting. We describe a new DoS attack that we term training inflation and present several ways to carry out this attack. We then evaluate recent distributed ML fault tolerance proposals and show that these are insufficient to mitigate several sybil-based attacks. Finally, we introduce a defense against targeted sybil-based poisoning called FoolsGold, which identifies sybils based on the diversity of client updates. We show that FoolsGold exceeds state of the art approaches when countering several types of poisoning attacks. Our work is open source and is available online: https://github.com/DistributedML/FoolsGold"
                },
                {
                    "title": "Data Structures for Statistical Computing in Python",
                    "abstract": "In this paper we are concerned with the practical issues of working with data sets common to finance, statistics, and other related fields. pandas is a new library which aims to facilitate working with these data sets and to provide a set of fundamental building blocks for implementing statistical models. We will discuss specific design issues encountered in the course of developing pandas with relevant examples and some comparisons with the R language. We conclude by discussing possible future directions for statistical computing and data analysis using Python."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.DC"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we effectively visualize, explain, and understand the dynamics of client behaviors in Federated Learning systems to enhance model performance and trust?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the growing need for transparency and trust in Federated Learning (FL) systems, particularly in sensitive applications where data privacy is paramount. By providing insights into client behaviors and their impact on model performance, this research could lead to more robust and reliable FL systems, fostering wider adoption in real-world applications. Furthermore, it could advance knowledge in the fields of explainable AI and distributed learning, paving the way for future research that explores more sophisticated methods for anomaly detection and client behavior analysis.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the inherent complexity of FL systems, where multiple clients with potentially diverse and biased data contribute to a global model. Naive approaches may fail because they do not account for the interactions between clients or the nuances of their decision-making processes. Additionally, technical obstacles include the need for effective visualization techniques that can handle high-dimensional data and the integration of counterfactual explanations in a distributed setting. Theoretical challenges also arise in accurately modeling and interpreting the behavior of clients in a way that is both meaningful and actionable.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on either predictive performance or decision-making processes in isolation, without integrating these perspectives in the context of FL systems. Existing solutions lack the capability to track and visualize client behaviors comprehensively, which has limited their effectiveness in addressing anomalies. Barriers such as the complexity of client interactions, the need for privacy-preserving techniques, and the absence of robust aggregation mechanisms have hindered progress. Our approach differs by introducing Federated Behavioural Planes (FBPs) that combine these aspects, providing a holistic view of client behavior and enabling the development of Federated Behavioural Shields for improved model robustness.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the creation of two behavioral planes: the Error Behavioural Plane for visualizing predictive diversity and the Counterfactuals Behavioural Plane for understanding decision-making diversity among clients. We will utilize a dataset representative of typical FL scenarios and employ metrics such as clustering effectiveness and model performance improvements to evaluate our approach. The expected outcomes"
            }
        },
        "author_data": {
            "56d9ae9a-95c5-4b66-943b-f7f2b2c6fe4e": {
                "pk": "56d9ae9a-95c5-4b66-943b-f7f2b2c6fe4e",
                "name": "Dario Fenoglio",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "3955a5b8-d149-4775-aa58-c87d254f5be7": {
                "pk": "3955a5b8-d149-4775-aa58-c87d254f5be7",
                "name": "Gabriele Dominici",
                "collaborators": [
                    "Pietro Barbiero",
                    "Martin Gjoreski",
                    "Marc Langheinrich",
                    "Francesco Giannini",
                    "Giuseppe Marra",
                    "Marc Langhenirich",
                    "Lucie Charlotte Magister",
                    "Pietro Li\u00f2",
                    "Nikola Simidjievski",
                    "Mateo Espinosa Zarlenga"
                ],
                "domain": [
                    "Interpretable AI",
                    "Multimodal Learning",
                    "Causal Inference",
                    "Explainable AI"
                ],
                "publications": [
                    {
                        "title": "AnyCBMs: How to Turn Any Black Box into a Concept Bottleneck Model",
                        "abstract": "Interpretable deep learning aims at developing neural architectures whose decision-making processes could be understood by their users. Among these techniqes, Concept Bottleneck Models enhance the interpretability of neural networks by integrating a layer of human-understandable concepts. These models, however, necessitate training a new model from the beginning, consuming significant resources and failing to utilize already trained large models. To address this issue, we introduce \"AnyCBM\", a method that transforms any existing trained model into a Concept Bottleneck Model with minimal impact on computational resources. We provide both theoretical and experimental insights showing the effectiveness of AnyCBMs in terms of classification performances and effectivenss of concept-based interventions on downstream tasks."
                    },
                    {
                        "title": "SHARCS: Shared Concept Space for Explainable Multimodal Learning",
                        "abstract": "Multimodal learning is an essential paradigm for addressing complex real-world problems, where individual data modalities are typically insufficient to accurately solve a given modelling task. While various deep learning approaches have successfully addressed these challenges, their reasoning process is often opaque; limiting the capabilities for a principled explainable cross-modal analysis and any domain-expert intervention. In this paper, we introduce SHARCS (SHARed Concept Space) -- a novel concept-based approach for explainable multimodal learning. SHARCS learns and maps interpretable concepts from different heterogeneous modalities into a single unified concept-manifold, which leads to an intuitive projection of semantically similar cross-modal concepts. We demonstrate that such an approach can lead to inherently explainable task predictions while also improving downstream predictive performance. Moreover, we show that SHARCS can operate and significantly outperform other approaches in practically significant scenarios, such as retrieval of missing modalities and cross-modal explanations. Our approach is model-agnostic and easily applicable to different types (and number) of modalities, thus advancing the development of effective, interpretable, and trustworthy multimodal approaches."
                    },
                    {
                        "title": "Counterfactual Concept Bottleneck Models",
                        "abstract": "Current deep learning models are not designed to simultaneously address three fundamental questions: predict class labels to solve a given classification task (the \"What?\"), simulate changes in the situation to evaluate how this impacts class predictions (the \"How?\"), and imagine how the scenario should change to result in different class predictions (the \"Why not?\"). The inability to answer these questions represents a crucial gap in deploying reliable AI agents, calibrating human trust, and improving human-machine interaction. To bridge this gap, we introduce CounterFactual Concept Bottleneck Models (CF-CBMs), a class of models designed to efficiently address the above queries all at once without the need to run post-hoc searches. Our experimental results demonstrate that CF-CBMs: achieve classification accuracy comparable to black-box models and existing CBMs (\"What?\"), rely on fewer important concepts leading to simpler explanations (\"How?\"), and produce interpretable, concept-based counterfactuals (\"Why not?\"). Additionally, we show that training the counterfactual generator jointly with the CBM leads to two key improvements: (i) it alters the model's decision-making process, making the model rely on fewer important concepts (leading to simpler explanations), and (ii) it significantly increases the causal effect of concept interventions on class predictions, making the model more responsive to these changes."
                    },
                    {
                        "title": "Causal Concept Graph Models: Beyond Causal Opacity in Deep Learning",
                        "abstract": "Causal opacity denotes the difficulty in understanding the \"hidden\" causal structure underlying the decisions of deep neural network (DNN) models. This leads to the inability to rely on and verify state-of-the-art DNN-based systems, especially in high-stakes scenarios. For this reason, circumventing causal opacity in DNNs represents a key open challenge at the intersection of deep learning, interpretability, and causality. This work addresses this gap by introducing Causal Concept Graph Models (Causal CGMs), a class of interpretable models whose decision-making process is causally transparent by design. Our experiments show that Causal CGMs can: (i) match the generalisation performance of causally opaque models, (ii) enable human-in-the-loop corrections to mispredicted intermediate reasoning steps, boosting not just downstream accuracy after corrections but also the reliability of the explanations provided for specific instances, and (iii) support the analysis of interventional and counterfactual scenarios, thereby improving the model's causal interpretability and supporting the effective verification of its reliability and fairness."
                    },
                    {
                        "title": "Evaluating Explanations Through LLMs: Beyond Traditional User Studies",
                        "abstract": "As AI becomes fundamental in sectors like healthcare, explainable AI (XAI) tools are essential for trust and transparency. However, traditional user studies used to evaluate these tools are often costly, time consuming, and difficult to scale. In this paper, we explore the use of Large Language Models (LLMs) to replicate human participants to help streamline XAI evaluation. We reproduce a user study comparing counterfactual and causal explanations, replicating human participants with seven LLMs under various settings. Our results show that (i) LLMs can replicate most conclusions from the original study, (ii) different LLMs yield varying levels of alignment in the results, and (iii) experimental factors such as LLM memory and output variability affect alignment with human responses. These initial findings suggest that LLMs could provide a scalable and cost-effective way to simplify qualitative XAI evaluation."
                    }
                ]
            },
            "5e7cf2ba-393f-4608-af38-bd47353324eb": {
                "pk": "5e7cf2ba-393f-4608-af38-bd47353324eb",
                "name": "Pietro Barbiero",
                "collaborators": [
                    "Gabriele Ciravegna",
                    "Francesco Giannini",
                    "Pietro Li\u00f3",
                    "Pietro Lio",
                    "Pietro Li\u00f2",
                    "Lucie Charlotte Magister",
                    "Dobrik Georgiev",
                    "Alberto Tonda",
                    "Stefano Fioravanti",
                    "Dmitry Kazhdan"
                ],
                "domain": [
                    "Computational Medicine",
                    "Explainable AI",
                    "Graph Neural Network",
                    "Multimodal Learning"
                ],
                "publications": [
                    {
                        "title": "The Computational Patient has Diabetes and a COVID",
                        "abstract": "Medicine is moving from a curative discipline to a preventative discipline relying on personalised and precise treatment plans. The complex and multi level pathophysiological patterns of most diseases require a systemic medicine approach and are challenging current medical therapies. On the other hand, computational medicine is a vibrant interdisciplinary field that could help move from an organ-centered approach to a process-oriented approach. The ideal computational patient would require an international interdisciplinary effort, of larger scientific and technological interdisciplinarity than the Human Genome Project. When deployed, such a patient would have a profound impact on how healthcare is delivered to patients. Here we present a computational patient model that integrates, refines and extends recent mechanistic or phenomenological models of cardiovascular, RAS and diabetic processes. Our aim is twofold: analyse the modularity and composability of the model-building blocks of the computational patient and to study the dynamical properties of well-being and disease states in a broader functional context. We present results from a number of experiments among which we characterise the dynamic impact of COVID-19 and type-2 diabetes (T2D) on cardiovascular and inflammation conditions. We tested these experiments under different exercise, meal and drug regimens. We report results showing the striking importance of transient dynamical responses to acute state conditions and we provide guidelines for system design principles for the inter-relationship between modules and components in systemic medicine. Finally this initial computational Patient can be used as a toolbox for further modifications and extensions."
                    },
                    {
                        "title": "Graph representation forecasting of patient's medical conditions: towards a digital twin",
                        "abstract": "Objective: Modern medicine needs to shift from a wait and react, curative discipline to a preventative, interdisciplinary science aiming at providing personalised, systemic and precise treatment plans to patients. The aim of this work is to present how the integration of machine learning approaches with mechanistic computational modelling could yield a reliable infrastructure to run probabilistic simulations where the entire organism is considered as a whole. Methods: We propose a general framework that composes advanced AI approaches and integrates mathematical modelling in order to provide a panoramic view over current and future physiological conditions. The proposed architecture is based on a graph neural network (GNNs) forecasting clinically relevant endpoints (such as blood pressure) and a generative adversarial network (GANs) providing a proof of concept of transcriptomic integrability. Results: We show the results of the investigation of pathological effects of overexpression of ACE2 across different signalling pathways in multiple tissues on cardiovascular functions. We provide a proof of concept of integrating a large set of composable clinical models using molecular data to drive local and global clinical parameters and derive future trajectories representing the evolution of the physiological state of the patient. Significance: We argue that the graph representation of a computational patient has potential to solve important technological challenges in integrating multiscale computational modelling with AI. We believe that this work represents a step forward towards a healthcare digital twin."
                    },
                    {
                        "title": "PyTorch, Explain! A Python library for Logic Explained Networks",
                        "abstract": "\"PyTorch, Explain!\" is a Python module integrating a variety of state-of-the-art approaches to provide logic explanations from neural networks. This package focuses on bringing these methods to non-specialists. It has minimal dependencies and it is distributed under the Apache 2.0 licence allowing both academic and commercial use. Source code and documentation can be downloaded from the github repository: https://github.com/pietrobarbiero/pytorch_explain."
                    },
                    {
                        "title": "Categorical Foundations of Explainable AI: A Unifying Theory",
                        "abstract": "Explainable AI (XAI) aims to address the human need for safe and reliable AI systems. However, numerous surveys emphasize the absence of a sound mathematical formalization of key XAI notions -- remarkably including the term \"explanation\" which still lacks a precise definition. To bridge this gap, this paper presents the first mathematically rigorous definitions of key XAI notions and processes, using the well-funded formalism of Category theory. We show that our categorical framework allows to: (i) model existing learning schemes and architectures, (ii) formally define the term \"explanation\", (iii) establish a theoretical basis for XAI taxonomies, and (iv) analyze commonly overlooked aspects of explaining methods. As a consequence, our categorical framework promotes the ethical and secure deployment of AI technologies as it represents a significant step towards a sound theoretical foundation of explainable AI."
                    },
                    {
                        "title": "Algorithmic Concept-based Explainable Reasoning",
                        "abstract": "Recent research on graph neural network (GNN) models successfully applied GNNs to classical graph algorithms and combinatorial optimisation problems. This has numerous benefits, such as allowing applications of algorithms when preconditions are not satisfied, or reusing learned models when sufficient training data is not available or can't be generated. Unfortunately, a key hindrance of these approaches is their lack of explainability, since GNNs are black-box models that cannot be interpreted directly. In this work, we address this limitation by applying existing work on concept-based explanations to GNN models. We introduce concept-bottleneck GNNs, which rely on a modification to the GNN readout mechanism. Using three case studies we demonstrate that: (i) our proposed model is capable of accurately learning concepts and extracting propositional formulas based on the learned concepts for each target class; (ii) our concept-based GNN models achieve comparative performance with state-of-the-art models; (iii) we can derive global graph concepts, without explicitly providing any supervision on graph-level concepts."
                    },
                    {
                        "title": "Topological Gradient-based Competitive Learning",
                        "abstract": "Topological learning is a wide research area aiming at uncovering the mutual spatial relationships between the elements of a set. Some of the most common and oldest approaches involve the use of unsupervised competitive neural networks. However, these methods are not based on gradient optimization which has been proven to provide striking results in feature extraction also in unsupervised learning. Unfortunately, by focusing mostly on algorithmic efficiency and accuracy, deep clustering techniques are composed of overly complex feature extractors, while using trivial algorithms in their top layer. The aim of this work is to present a novel comprehensive theory aspiring at bridging competitive learning with gradient-based learning, thus allowing the use of extremely powerful deep neural networks for feature extraction and projection combined with the remarkable flexibility and expressiveness of competitive learning. In this paper we fully demonstrate the theoretical equivalence of two novel gradient-based competitive layers. Preliminary experiments show how the dual approach, trained on the transpose of the input matrix i.e. $X^T$, lead to faster convergence rate and higher training accuracy both in low and high-dimensional scenarios."
                    },
                    {
                        "title": "Entropy-based Logic Explanations of Neural Networks",
                        "abstract": "Explainable artificial intelligence has rapidly emerged since lawmakers have started requiring interpretable models for safety-critical domains. Concept-based neural networks have arisen as explainable-by-design methods as they leverage human-understandable symbols (i.e. concepts) to predict class memberships. However, most of these approaches focus on the identification of the most relevant concepts but do not provide concise, formal explanations of how such concepts are leveraged by the classifier to make predictions. In this paper, we propose a novel end-to-end differentiable approach enabling the extraction of logic explanations from neural networks using the formalism of First-Order Logic. The method relies on an entropy-based criterion which automatically identifies the most relevant concepts. We consider four different case studies to demonstrate that: (i) this entropy-based criterion enables the distillation of concise logic explanations in safety-critical domains from clinical data to computer vision; (ii) the proposed approach outperforms state-of-the-art white-box models in terms of classification accuracy and matches black box performances."
                    },
                    {
                        "title": "Logic Explained Networks",
                        "abstract": "The large and still increasing popularity of deep learning clashes with a major limit of neural network architectures, that consists in their lack of capability in providing human-understandable motivations of their decisions. In situations in which the machine is expected to support the decision of human experts, providing a comprehensible explanation is a feature of crucial importance. The language used to communicate the explanations must be formal enough to be implementable in a machine and friendly enough to be understandable by a wide audience. In this paper, we propose a general approach to Explainable Artificial Intelligence in the case of neural architectures, showing how a mindful design of the networks leads to a family of interpretable deep learning models called Logic Explained Networks (LENs). LENs only require their inputs to be human-understandable predicates, and they provide explanations in terms of simple First-Order Logic (FOL) formulas involving such predicates. LENs are general enough to cover a large number of scenarios. Amongst them, we consider the case in which LENs are directly used as special classifiers with the capability of being explainable, or when they act as additional networks with the role of creating the conditions for making a black-box classifier explainable by FOL formulas. Despite supervised learning problems are mostly emphasized, we also show that LENs can learn and provide explanations in unsupervised learning settings. Experimental results on several datasets and tasks show that LENs may yield better classifications than established white-box models, such as decision trees and Bayesian rule lists, while providing more compact and meaningful explanations."
                    },
                    {
                        "title": "Global Explainability of GNNs via Logic Combination of Learned Concepts",
                        "abstract": "While instance-level explanation of GNN is a well-studied problem with plenty of approaches being developed, providing a global explanation for the behaviour of a GNN is much less explored, despite its potential in interpretability and debugging. Existing solutions either simply list local explanations for a given class, or generate a synthetic prototypical graph with maximal score for a given class, completely missing any combinatorial aspect that the GNN could have learned. In this work, we propose GLGExplainer (Global Logic-based GNN Explainer), the first Global Explainer capable of generating explanations as arbitrary Boolean combinations of learned graphical concepts. GLGExplainer is a fully differentiable architecture that takes local explanations as inputs and combines them into a logic formula over graphical concepts, represented as clusters of local explanations. Contrary to existing solutions, GLGExplainer provides accurate and human-interpretable global explanations that are perfectly aligned with ground-truth explanations (on synthetic data) or match existing domain knowledge (on real-world data). Extracted formulas are faithful to the model predictions, to the point of providing insights into some occasionally incorrect rules learned by the model, making GLGExplainer a promising diagnostic tool for learned GNNs."
                    },
                    {
                        "title": "Gradient-based Competitive Learning: Theory",
                        "abstract": "Deep learning has been widely used for supervised learning and classification/regression problems. Recently, a novel area of research has applied this paradigm to unsupervised tasks; indeed, a gradient-based approach extracts, efficiently and autonomously, the relevant features for handling input data. However, state-of-the-art techniques focus mostly on algorithmic efficiency and accuracy rather than mimic the input manifold. On the contrary, competitive learning is a powerful tool for replicating the input distribution topology. This paper introduces a novel perspective in this area by combining these two techniques: unsupervised gradient-based and competitive learning. The theory is based on the intuition that neural networks are able to learn topological structures by working directly on the transpose of the input matrix. At this purpose, the vanilla competitive layer and its dual are presented. The former is just an adaptation of a standard competitive layer for deep clustering, while the latter is trained on the transposed matrix. Their equivalence is extensively proven both theoretically and experimentally. However, the dual layer is better suited for handling very high-dimensional datasets. The proposed approach has a great potential as it can be generalized to a vast selection of topological learning tasks, such as non-stationary and hierarchical clustering; furthermore, it can also be integrated within more complex architectures such as autoencoders and generative adversarial networks."
                    },
                    {
                        "title": "Interpretable Graph Networks Formulate Universal Algebra Conjectures",
                        "abstract": "The rise of Artificial Intelligence (AI) recently empowered researchers to investigate hard mathematical problems which eluded traditional approaches for decades. Yet, the use of AI in Universal Algebra (UA) -- one of the fields laying the foundations of modern mathematics -- is still completely unexplored. This work proposes the first use of AI to investigate UA's conjectures with an equivalent equational and topological characterization. While topological representations would enable the analysis of such properties using graph neural networks, the limited transparency and brittle explainability of these models hinder their straightforward use to empirically validate existing conjectures or to formulate new ones. To bridge these gaps, we propose a general algorithm generating AI-ready datasets based on UA's conjectures, and introduce a novel neural layer to build fully interpretable graph networks. The results of our experiments demonstrate that interpretable graph networks: (i) enhance interpretability without sacrificing task accuracy, (ii) strongly generalize when predicting universal algebra's properties, (iii) generate simple explanations that empirically validate existing conjectures, and (iv) identify subgraphs suggesting the formulation of novel conjectures."
                    },
                    {
                        "title": "Relational Concept Based Models",
                        "abstract": "The design of interpretable deep learning models working in relational domains poses an open challenge: interpretable deep learning methods, such as Concept-Based Models (CBMs), are not designed to solve relational problems, while relational models are not as interpretable as CBMs. To address this problem, we propose Relational Concept-Based Models, a family of relational deep learning methods providing interpretable task predictions. Our experiments, ranging from image classification to link prediction in knowledge graphs, show that relational CBMs (i) match generalization performance of existing relational black-boxes (as opposed to non-relational CBMs), (ii) support the generation of quantified concept-based explanations, (iii) effectively respond to test-time interventions, and (iv) withstand demanding settings including out-of-distribution scenarios, limited training data regimes, and scarce concept supervisions."
                    },
                    {
                        "title": "Counterfactual Explanations for Clustering Models",
                        "abstract": "Clustering algorithms rely on complex optimisation processes that may be difficult to comprehend, especially for individuals who lack technical expertise. While many explainable artificial intelligence techniques exist for supervised machine learning, unsupervised learning -- and clustering in particular -- has been largely neglected. To complicate matters further, the notion of a ``true'' cluster is inherently challenging to define. These facets of unsupervised learning and its explainability make it difficult to foster trust in such methods and curtail their adoption. To address these challenges, we propose a new, model-agnostic technique for explaining clustering algorithms with counterfactual statements. Our approach relies on a novel soft-scoring method that captures the spatial information utilised by clustering models. It builds upon a state-of-the-art Bayesian counterfactual generator for supervised learning to deliver high-quality explanations. We evaluate its performance on five datasets and two clustering algorithms, and demonstrate that introducing soft scores to guide counterfactual search significantly improves the results."
                    },
                    {
                        "title": "Extending Logic Explained Networks to Text Classification",
                        "abstract": "Recently, Logic Explained Networks (LENs) have been proposed as explainable-by-design neural models providing logic explanations for their predictions. However, these models have only been applied to vision and tabular data, and they mostly favour the generation of global explanations, while local ones tend to be noisy and verbose. For these reasons, we propose LENp, improving local explanations by perturbing input words, and we test it on text classification. Our results show that (i) LENp provides better local explanations than LIME in terms of sensitivity and faithfulness, and (ii) logic explanations are more useful and user-friendly than feature scoring provided by LIME as attested by a human survey."
                    },
                    {
                        "title": "GCI: A (G)raph (C)oncept (I)nterpretation Framework",
                        "abstract": "Explainable AI (XAI) underwent a recent surge in research on concept extraction, focusing on extracting human-interpretable concepts from Deep Neural Networks. An important challenge facing concept extraction approaches is the difficulty of interpreting and evaluating discovered concepts, especially for complex tasks such as molecular property prediction. We address this challenge by presenting GCI: a (G)raph (C)oncept (I)nterpretation framework, used for quantitatively measuring alignment between concepts discovered from Graph Neural Networks (GNNs) and their corresponding human interpretations. GCI encodes concept interpretations as functions, which can be used to quantitatively measure the alignment between a given interpretation and concept definition. We demonstrate four applications of GCI: (i) quantitatively evaluating concept extractors, (ii) measuring alignment between concept extractors and human interpretations, (iii) measuring the completeness of interpretations with respect to an end task and (iv) a practical application of GCI to molecular property prediction, in which we demonstrate how to use chemical functional groups to explain GNNs trained on molecular property prediction tasks, and implement interpretations with a 0.76 AUCROC completeness score."
                    },
                    {
                        "title": "SHARCS: Shared Concept Space for Explainable Multimodal Learning",
                        "abstract": "Multimodal learning is an essential paradigm for addressing complex real-world problems, where individual data modalities are typically insufficient to accurately solve a given modelling task. While various deep learning approaches have successfully addressed these challenges, their reasoning process is often opaque; limiting the capabilities for a principled explainable cross-modal analysis and any domain-expert intervention. In this paper, we introduce SHARCS (SHARed Concept Space) -- a novel concept-based approach for explainable multimodal learning. SHARCS learns and maps interpretable concepts from different heterogeneous modalities into a single unified concept-manifold, which leads to an intuitive projection of semantically similar cross-modal concepts. We demonstrate that such an approach can lead to inherently explainable task predictions while also improving downstream predictive performance. Moreover, we show that SHARCS can operate and significantly outperform other approaches in practically significant scenarios, such as retrieval of missing modalities and cross-modal explanations. Our approach is model-agnostic and easily applicable to different types (and number) of modalities, thus advancing the development of effective, interpretable, and trustworthy multimodal approaches."
                    },
                    {
                        "title": "From Charts to Atlas: Merging Latent Spaces into One",
                        "abstract": "Models trained on semantically related datasets and tasks exhibit comparable inter-sample relations within their latent spaces. We investigate in this study the aggregation of such latent spaces to create a unified space encompassing the combined information. To this end, we introduce Relative Latent Space Aggregation, a two-step approach that first renders the spaces comparable using relative representations, and then aggregates them via a simple mean. We carefully divide a classification problem into a series of learning tasks under three different settings: sharing samples, classes, or neither. We then train a model on each task and aggregate the resulting latent spaces. We compare the aggregated space with that derived from an end-to-end model trained over all tasks and show that the two spaces are similar. We then observe that the aggregated space is better suited for classification, and empirically demonstrate that it is due to the unique imprints left by task-specific embedders within the representations. We finally test our framework in scenarios where no shared region exists and show that it can still be used to merge the spaces, albeit with diminished benefits over naive merging."
                    },
                    {
                        "title": "Uncovering Coresets for Classification With Multi-Objective Evolutionary Algorithms",
                        "abstract": "A coreset is a subset of the training set, using which a machine learning algorithm obtains performances similar to what it would deliver if trained over the whole original data. Coreset discovery is an active and open line of research as it allows improving training speed for the algorithms and may help human understanding the results. Building on previous works, a novel approach is presented: candidate corsets are iteratively optimized, adding and removing samples. As there is an obvious trade-off between limiting training size and quality of the results, a multi-objective evolutionary algorithm is used to minimize simultaneously the number of points in the set and the classification error. Experimental results on non-trivial benchmarks show that the proposed approach is able to deliver results that allow a classifier to obtain lower error and better ability of generalizing on unseen data than state-of-the-art coreset discovery techniques."
                    },
                    {
                        "title": "Modeling Generalization in Machine Learning: A Methodological and Computational Study",
                        "abstract": "As machine learning becomes more and more available to the general public, theoretical questions are turning into pressing practical issues. Possibly, one of the most relevant concerns is the assessment of our confidence in trusting machine learning predictions. In many real-world cases, it is of utmost importance to estimate the capabilities of a machine learning algorithm to generalize, i.e., to provide accurate predictions on unseen data, depending on the characteristics of the target problem. In this work, we perform a meta-analysis of 109 publicly-available classification data sets, modeling machine learning generalization as a function of a variety of data set characteristics, ranging from number of samples to intrinsic dimensionality, from class-wise feature skewness to $F1$ evaluated on test samples falling outside the convex hull of the training set. Experimental results demonstrate the relevance of using the concept of the convex hull of the training data in assessing machine learning generalization, by emphasizing the difference between interpolated and extrapolated predictions. Besides several predictable correlations, we observe unexpectedly weak associations between the generalization ability of machine learning models and all metrics related to dimensionality, thus challenging the common assumption that the \\textit{curse of dimensionality} might impair generalization in machine learning."
                    },
                    {
                        "title": "Global Concept-Based Interpretability for Graph Neural Networks via Neuron Analysis",
                        "abstract": "Graph neural networks (GNNs) are highly effective on a variety of graph-related tasks; however, they lack interpretability and transparency. Current explainability approaches are typically local and treat GNNs as black-boxes. They do not look inside the model, inhibiting human trust in the model and explanations. Motivated by the ability of neurons to detect high-level semantic concepts in vision models, we perform a novel analysis on the behaviour of individual GNN neurons to answer questions about GNN interpretability, and propose new metrics for evaluating the interpretability of GNN neurons. We propose a novel approach for producing global explanations for GNNs using neuron-level concepts to enable practitioners to have a high-level view of the model. Specifically, (i) to the best of our knowledge, this is the first work which shows that GNN neurons act as concept detectors and have strong alignment with concepts formulated as logical compositions of node degree and neighbourhood properties; (ii) we quantitatively assess the importance of detected concepts, and identify a trade-off between training duration and neuron-level interpretability; (iii) we demonstrate that our global explainability approach has advantages over the current state-of-the-art -- we can disentangle the explanation into individual interpretable concepts backed by logical descriptions, which reduces potential for bias and improves user-friendliness."
                    }
                ]
            },
            "6bae0e0c-ac4f-483c-b8a9-8e28fe5ffe9c": {
                "pk": "6bae0e0c-ac4f-483c-b8a9-8e28fe5ffe9c",
                "name": "Alberto Tonda",
                "collaborators": [
                    "Pietro Barbiero",
                    "Giovanni Squillero",
                    "Francesco Giannini",
                    "Soumita Das",
                    "Bijita Singha",
                    "Anupam Biswas",
                    "Stefano Fioravanti",
                    "Pietro Lio",
                    "Elena Di Lavore",
                    "Gabriele Ciravegna"
                ],
                "domain": [
                    "Coreset Discovery",
                    "Explainable AI",
                    "Community Detection",
                    "Deep Learning"
                ],
                "publications": [
                    {
                        "title": "Uncovering Coresets for Classification With Multi-Objective Evolutionary Algorithms",
                        "abstract": "A coreset is a subset of the training set, using which a machine learning algorithm obtains performances similar to what it would deliver if trained over the whole original data. Coreset discovery is an active and open line of research as it allows improving training speed for the algorithms and may help human understanding the results. Building on previous works, a novel approach is presented: candidate corsets are iteratively optimized, adding and removing samples. As there is an obvious trade-off between limiting training size and quality of the results, a multi-objective evolutionary algorithm is used to minimize simultaneously the number of points in the set and the classification error. Experimental results on non-trivial benchmarks show that the proposed approach is able to deliver results that allow a classifier to obtain lower error and better ability of generalizing on unseen data than state-of-the-art coreset discovery techniques."
                    },
                    {
                        "title": "Modeling Generalization in Machine Learning: A Methodological and Computational Study",
                        "abstract": "As machine learning becomes more and more available to the general public, theoretical questions are turning into pressing practical issues. Possibly, one of the most relevant concerns is the assessment of our confidence in trusting machine learning predictions. In many real-world cases, it is of utmost importance to estimate the capabilities of a machine learning algorithm to generalize, i.e., to provide accurate predictions on unseen data, depending on the characteristics of the target problem. In this work, we perform a meta-analysis of 109 publicly-available classification data sets, modeling machine learning generalization as a function of a variety of data set characteristics, ranging from number of samples to intrinsic dimensionality, from class-wise feature skewness to $F1$ evaluated on test samples falling outside the convex hull of the training set. Experimental results demonstrate the relevance of using the concept of the convex hull of the training data in assessing machine learning generalization, by emphasizing the difference between interpolated and extrapolated predictions. Besides several predictable correlations, we observe unexpectedly weak associations between the generalization ability of machine learning models and all metrics related to dimensionality, thus challenging the common assumption that the \\textit{curse of dimensionality} might impair generalization in machine learning."
                    },
                    {
                        "title": "Direct Comparative Analysis of Nature-inspired Optimization Algorithms on Community Detection Problem in Social Networks",
                        "abstract": "Nature-inspired optimization Algorithms (NIOAs) are nowadays a popular choice for community detection in social networks. Community detection problem in social network is treated as optimization problem, where the objective is to either maximize the connection within the community or minimize connections between the communities. To apply NIOAs, either of the two, or both objectives are explored. Since NIOAs mostly exploit randomness in their strategies, it is necessary to analyze their performance for specific applications. In this paper, NIOAs are analyzed on the community detection problem. A direct comparison approach is followed to perform pairwise comparison of NIOAs. The performance is measured in terms of five scores designed based on prasatul matrix and also with average isolability. Three widely used real-world social networks and four NIOAs are considered for analyzing the quality of communities generated by NIOAs."
                    },
                    {
                        "title": "Categorical Foundations of Explainable AI: A Unifying Theory",
                        "abstract": "Explainable AI (XAI) aims to address the human need for safe and reliable AI systems. However, numerous surveys emphasize the absence of a sound mathematical formalization of key XAI notions -- remarkably including the term \"explanation\" which still lacks a precise definition. To bridge this gap, this paper presents the first mathematically rigorous definitions of key XAI notions and processes, using the well-funded formalism of Category theory. We show that our categorical framework allows to: (i) model existing learning schemes and architectures, (ii) formally define the term \"explanation\", (iii) establish a theoretical basis for XAI taxonomies, and (iv) analyze commonly overlooked aspects of explaining methods. As a consequence, our categorical framework promotes the ethical and secure deployment of AI technologies as it represents a significant step towards a sound theoretical foundation of explainable AI."
                    },
                    {
                        "title": "Interpretable Neural-Symbolic Concept Reasoning",
                        "abstract": "Deep learning methods are highly accurate, yet their opaque decision process prevents them from earning full human trust. Concept-based models aim to address this issue by learning tasks based on a set of human-understandable concepts. However, state-of-the-art concept-based models rely on high-dimensional concept embedding representations which lack a clear semantic meaning, thus questioning the interpretability of their decision process. To overcome this limitation, we propose the Deep Concept Reasoner (DCR), the first interpretable concept-based model that builds upon concept embeddings. In DCR, neural networks do not make task predictions directly, but they build syntactic rule structures using concept embeddings. DCR then executes these rules on meaningful concept truth degrees to provide a final interpretable and semantically-consistent prediction in a differentiable manner. Our experiments show that DCR: (i) improves up to +25% w.r.t. state-of-the-art interpretable concept-based models on challenging benchmarks (ii) discovers meaningful logic rules matching known ground truths even in the absence of concept supervision during training, and (iii), facilitates the generation of counterfactual examples providing the learnt rules as guidance."
                    }
                ]
            },
            "1a5503d3-14a6-4df7-9f2c-49d475f85780": {
                "pk": "1a5503d3-14a6-4df7-9f2c-49d475f85780",
                "name": "Martin Gjoreski",
                "collaborators": [
                    "Marc Langheinrich",
                    "Gabriele Dominici",
                    "Pietro Barbiero",
                    "Francesco Giannini",
                    "Giuseppe Marra",
                    "Marc Langhenirich",
                    "Aurora Spagnol",
                    "Kacper Sokol",
                    "Fatima Ezzeddine",
                    "Mirna Saad"
                ],
                "domain": [
                    "Interpretable Machine Learning",
                    "Explainable AI",
                    "Causal Inference",
                    "Anomaly Detection"
                ],
                "publications": [
                    {
                        "title": "AnyCBMs: How to Turn Any Black Box into a Concept Bottleneck Model",
                        "abstract": "Interpretable deep learning aims at developing neural architectures whose decision-making processes could be understood by their users. Among these techniqes, Concept Bottleneck Models enhance the interpretability of neural networks by integrating a layer of human-understandable concepts. These models, however, necessitate training a new model from the beginning, consuming significant resources and failing to utilize already trained large models. To address this issue, we introduce \"AnyCBM\", a method that transforms any existing trained model into a Concept Bottleneck Model with minimal impact on computational resources. We provide both theoretical and experimental insights showing the effectiveness of AnyCBMs in terms of classification performances and effectivenss of concept-based interventions on downstream tasks."
                    },
                    {
                        "title": "Counterfactual Explanations for Clustering Models",
                        "abstract": "Clustering algorithms rely on complex optimisation processes that may be difficult to comprehend, especially for individuals who lack technical expertise. While many explainable artificial intelligence techniques exist for supervised machine learning, unsupervised learning -- and clustering in particular -- has been largely neglected. To complicate matters further, the notion of a ``true'' cluster is inherently challenging to define. These facets of unsupervised learning and its explainability make it difficult to foster trust in such methods and curtail their adoption. To address these challenges, we propose a new, model-agnostic technique for explaining clustering algorithms with counterfactual statements. Our approach relies on a novel soft-scoring method that captures the spatial information utilised by clustering models. It builds upon a state-of-the-art Bayesian counterfactual generator for supervised learning to deliver high-quality explanations. We evaluate its performance on five datasets and two clustering algorithms, and demonstrate that introducing soft scores to guide counterfactual search significantly improves the results."
                    },
                    {
                        "title": "Differential Privacy for Anomaly Detection: Analyzing the Trade-off Between Privacy and Explainability",
                        "abstract": "Anomaly detection (AD), also referred to as outlier detection, is a statistical process aimed at identifying observations within a dataset that significantly deviate from the expected pattern of the majority of the data. Such a process finds wide application in various fields, such as finance and healthcare. While the primary objective of AD is to yield high detection accuracy, the requirements of explainability and privacy are also paramount. The first ensures the transparency of the AD process, while the second guarantees that no sensitive information is leaked to untrusted parties. In this work, we exploit the trade-off of applying Explainable AI (XAI) through SHapley Additive exPlanations (SHAP) and differential privacy (DP). We perform AD with different models and on various datasets, and we thoroughly evaluate the cost of privacy in terms of decreased accuracy and explainability. Our results show that the enforcement of privacy through DP has a significant impact on detection accuracy and explainability, which depends on both the dataset and the considered AD model. We further show that the visual interpretation of explanations is also influenced by the choice of the AD algorithm."
                    },
                    {
                        "title": "Causal Concept Graph Models: Beyond Causal Opacity in Deep Learning",
                        "abstract": "Causal opacity denotes the difficulty in understanding the \"hidden\" causal structure underlying the decisions of deep neural network (DNN) models. This leads to the inability to rely on and verify state-of-the-art DNN-based systems, especially in high-stakes scenarios. For this reason, circumventing causal opacity in DNNs represents a key open challenge at the intersection of deep learning, interpretability, and causality. This work addresses this gap by introducing Causal Concept Graph Models (Causal CGMs), a class of interpretable models whose decision-making process is causally transparent by design. Our experiments show that Causal CGMs can: (i) match the generalisation performance of causally opaque models, (ii) enable human-in-the-loop corrections to mispredicted intermediate reasoning steps, boosting not just downstream accuracy after corrections but also the reliability of the explanations provided for specific instances, and (iii) support the analysis of interventional and counterfactual scenarios, thereby improving the model's causal interpretability and supporting the effective verification of its reliability and fairness."
                    },
                    {
                        "title": "Evaluating Explanations Through LLMs: Beyond Traditional User Studies",
                        "abstract": "As AI becomes fundamental in sectors like healthcare, explainable AI (XAI) tools are essential for trust and transparency. However, traditional user studies used to evaluate these tools are often costly, time consuming, and difficult to scale. In this paper, we explore the use of Large Language Models (LLMs) to replicate human participants to help streamline XAI evaluation. We reproduce a user study comparing counterfactual and causal explanations, replicating human participants with seven LLMs under various settings. Our results show that (i) LLMs can replicate most conclusions from the original study, (ii) different LLMs yield varying levels of alignment in the results, and (iii) experimental factors such as LLM memory and output variability affect alignment with human responses. These initial findings suggest that LLMs could provide a scalable and cost-effective way to simplify qualitative XAI evaluation."
                    },
                    {
                        "title": "Counterfactual Concept Bottleneck Models",
                        "abstract": "Current deep learning models are not designed to simultaneously address three fundamental questions: predict class labels to solve a given classification task (the \"What?\"), simulate changes in the situation to evaluate how this impacts class predictions (the \"How?\"), and imagine how the scenario should change to result in different class predictions (the \"Why not?\"). The inability to answer these questions represents a crucial gap in deploying reliable AI agents, calibrating human trust, and improving human-machine interaction. To bridge this gap, we introduce CounterFactual Concept Bottleneck Models (CF-CBMs), a class of models designed to efficiently address the above queries all at once without the need to run post-hoc searches. Our experimental results demonstrate that CF-CBMs: achieve classification accuracy comparable to black-box models and existing CBMs (\"What?\"), rely on fewer important concepts leading to simpler explanations (\"How?\"), and produce interpretable, concept-based counterfactuals (\"Why not?\"). Additionally, we show that training the counterfactual generator jointly with the CBM leads to two key improvements: (i) it alters the model's decision-making process, making the model rely on fewer important concepts (leading to simpler explanations), and (ii) it significantly increases the causal effect of concept interventions on class predictions, making the model more responsive to these changes."
                    }
                ]
            },
            "dc7f65c5-5de2-4f2c-9efb-3090f81f794c": {
                "pk": "dc7f65c5-5de2-4f2c-9efb-3090f81f794c",
                "name": "Marc Langheinrich",
                "collaborators": [
                    "Martin Gjoreski",
                    "Pietro Barbiero",
                    "Gabriele Dominici",
                    "Giuseppe Marra",
                    "Kathrin Bednar",
                    "Sarah Spiekermann",
                    "Aurora Spagnol",
                    "Kacper Sokol",
                    "Francesco Giannini",
                    "Fatima Ezzeddine"
                ],
                "domain": [
                    "Explainable AI",
                    "Causal Inference",
                    "Anomaly Detection",
                    "Privacy Protection"
                ],
                "publications": [
                    {
                        "title": "Engineering Privacy by Design: Are engineers ready to live up to the challenge?",
                        "abstract": "Organizations struggle to comply with legal requirements as well as customers calls for better data protection. On the implementation level, incorporation of privacy protections in products and services depends on the commitment of the engineers who design them. We interviewed six senior engineers, who work for globally leading IT corporations and research institutions to investigate their motivation and ability to comply with privacy regulations. Our findings point to a lack of perceived responsibility, control, autonomy, and frustrations with interactions with the legal world. While we increasingly call on engineers to go beyond functional requirements and be responsive to human values in our increasingly technological society, we may be facing the dilemma of asking engineers to live up to a challenge they are currently not ready to embrace."
                    },
                    {
                        "title": "Counterfactual Explanations for Clustering Models",
                        "abstract": "Clustering algorithms rely on complex optimisation processes that may be difficult to comprehend, especially for individuals who lack technical expertise. While many explainable artificial intelligence techniques exist for supervised machine learning, unsupervised learning -- and clustering in particular -- has been largely neglected. To complicate matters further, the notion of a ``true'' cluster is inherently challenging to define. These facets of unsupervised learning and its explainability make it difficult to foster trust in such methods and curtail their adoption. To address these challenges, we propose a new, model-agnostic technique for explaining clustering algorithms with counterfactual statements. Our approach relies on a novel soft-scoring method that captures the spatial information utilised by clustering models. It builds upon a state-of-the-art Bayesian counterfactual generator for supervised learning to deliver high-quality explanations. We evaluate its performance on five datasets and two clustering algorithms, and demonstrate that introducing soft scores to guide counterfactual search significantly improves the results."
                    },
                    {
                        "title": "Counterfactual Concept Bottleneck Models",
                        "abstract": "Current deep learning models are not designed to simultaneously address three fundamental questions: predict class labels to solve a given classification task (the \"What?\"), simulate changes in the situation to evaluate how this impacts class predictions (the \"How?\"), and imagine how the scenario should change to result in different class predictions (the \"Why not?\"). The inability to answer these questions represents a crucial gap in deploying reliable AI agents, calibrating human trust, and improving human-machine interaction. To bridge this gap, we introduce CounterFactual Concept Bottleneck Models (CF-CBMs), a class of models designed to efficiently address the above queries all at once without the need to run post-hoc searches. Our experimental results demonstrate that CF-CBMs: achieve classification accuracy comparable to black-box models and existing CBMs (\"What?\"), rely on fewer important concepts leading to simpler explanations (\"How?\"), and produce interpretable, concept-based counterfactuals (\"Why not?\"). Additionally, we show that training the counterfactual generator jointly with the CBM leads to two key improvements: (i) it alters the model's decision-making process, making the model rely on fewer important concepts (leading to simpler explanations), and (ii) it significantly increases the causal effect of concept interventions on class predictions, making the model more responsive to these changes."
                    },
                    {
                        "title": "Differential Privacy for Anomaly Detection: Analyzing the Trade-off Between Privacy and Explainability",
                        "abstract": "Anomaly detection (AD), also referred to as outlier detection, is a statistical process aimed at identifying observations within a dataset that significantly deviate from the expected pattern of the majority of the data. Such a process finds wide application in various fields, such as finance and healthcare. While the primary objective of AD is to yield high detection accuracy, the requirements of explainability and privacy are also paramount. The first ensures the transparency of the AD process, while the second guarantees that no sensitive information is leaked to untrusted parties. In this work, we exploit the trade-off of applying Explainable AI (XAI) through SHapley Additive exPlanations (SHAP) and differential privacy (DP). We perform AD with different models and on various datasets, and we thoroughly evaluate the cost of privacy in terms of decreased accuracy and explainability. Our results show that the enforcement of privacy through DP has a significant impact on detection accuracy and explainability, which depends on both the dataset and the considered AD model. We further show that the visual interpretation of explanations is also influenced by the choice of the AD algorithm."
                    },
                    {
                        "title": "Causal Concept Graph Models: Beyond Causal Opacity in Deep Learning",
                        "abstract": "Causal opacity denotes the difficulty in understanding the \"hidden\" causal structure underlying the decisions of deep neural network (DNN) models. This leads to the inability to rely on and verify state-of-the-art DNN-based systems, especially in high-stakes scenarios. For this reason, circumventing causal opacity in DNNs represents a key open challenge at the intersection of deep learning, interpretability, and causality. This work addresses this gap by introducing Causal Concept Graph Models (Causal CGMs), a class of interpretable models whose decision-making process is causally transparent by design. Our experiments show that Causal CGMs can: (i) match the generalisation performance of causally opaque models, (ii) enable human-in-the-loop corrections to mispredicted intermediate reasoning steps, boosting not just downstream accuracy after corrections but also the reliability of the explanations provided for specific instances, and (iii) support the analysis of interventional and counterfactual scenarios, thereby improving the model's causal interpretability and supporting the effective verification of its reliability and fairness."
                    },
                    {
                        "title": "Evaluating Explanations Through LLMs: Beyond Traditional User Studies",
                        "abstract": "As AI becomes fundamental in sectors like healthcare, explainable AI (XAI) tools are essential for trust and transparency. However, traditional user studies used to evaluate these tools are often costly, time consuming, and difficult to scale. In this paper, we explore the use of Large Language Models (LLMs) to replicate human participants to help streamline XAI evaluation. We reproduce a user study comparing counterfactual and causal explanations, replicating human participants with seven LLMs under various settings. Our results show that (i) LLMs can replicate most conclusions from the original study, (ii) different LLMs yield varying levels of alignment in the results, and (iii) experimental factors such as LLM memory and output variability affect alignment with human responses. These initial findings suggest that LLMs could provide a scalable and cost-effective way to simplify qualitative XAI evaluation."
                    }
                ]
            }
        }
    },
    "2305.19358": {
        "paper_data": {
            "title": "Stable Anisotropic Regularization",
            "url": "http://arxiv.org/abs/2305.19358v3",
            "arxiv_id": "2305.19358",
            "authors": [
                "William Rudman",
                "Carsten Eickhoff"
            ],
            "abstract": "Given the success of Large Language Models (LLMs), there has been considerable interest in studying the properties of model activations. The literature overwhelmingly agrees that LLM representations are dominated by a few \"outlier dimensions\" with exceedingly high variance and magnitude. Several studies in Natural Language Processing (NLP) have sought to mitigate the impact of such outlier dimensions and force LLMs to be isotropic (i.e., have uniform variance across all dimensions in embedding space). Isotropy is thought to be a desirable property for LLMs that improves model performance and more closely aligns textual representations with human intuition. However, many of the claims regarding isotropy in NLP have been based on the average cosine similarity of embeddings, which has recently been shown to be a flawed measure of isotropy. In this paper, we propose I-STAR: IsoScore*-based STable Anisotropic Regularization, a novel regularization method that can be used to increase or decrease levels of isotropy in embedding space during training. I-STAR uses IsoScore*, the first accurate measure of isotropy that is both differentiable and stable on mini-batch computations. In contrast to several previous works, we find that decreasing isotropy in contextualized embeddings improves performance on the majority of tasks and models considered in this paper.",
            "introduction": "   1 Introduction  Several previous works have investigated the role of isotropy in Large Language Model (LLM) representations (Rudman et\u00a0al., 2022). A distribution is isotropic if the variance of the data is uniform and the data dimensions are uncorrelated. In practice, a distribution is isotropic when its covariance matrix is proportional to the identity matrix. Studies have found that representations from LLMs, such as BERT or GPT-2, lack the property of isotropy and that contextualized word embeddings are dominated by a few \u201crogue\u201d or \u201coutlier\u201d dimensions (Timkey and van Schijndel, 2021; Kovaleva et\u00a0al., 2021). Several previous works have argued that anisotropy, i.e., the lack of isotropy, is detrimental to LLM embeddings as it 1) forces representations to occupy a \u201cnarrow cone\u201d in space (Ethayarajh, 2019; Cai et\u00a0al., 2021); 2) obscures linguistic information, thereby limiting the expressive power of the embeddings (Gao et\u00a0al., 2019; Zhang et\u00a0al., 2020; Mickus et\u00a0al., 2019), and; 3) hinders performance on a variety of downstream tasks (Kovaleva et\u00a0al., 2021; Bi\u015b et\u00a0al., 2021; Timkey and van Schijndel, 2021). However, some recent works have challenged previously held conceptions about isotropy, arguing that current methods of measuring isotropy are fundamentally flawed (Rudman et\u00a0al., 2022; Rajaee and Pilehvar, 2021a). To address these concerns, Rudman et\u00a0al. (2022) propose IsoScore, an accurate and robust method for measuring isotropy based on the covariance matrix of a distribution. Although IsoScore is an effective method for measuring isotropy, we demonstrate that IsoScore is neither differentiable nor stable when the number of points in a given sample is small. Therefore, IsoScore cannot serve as an effective model regularizer.   Figure 1: Forward pass of our I-STAR loss function. Let xlsubscript\ud835\udc65\ud835\udc59x_{l}italic_x start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT be the token embeddings in a mini-batch at layer l\u2208{1,2,\u2026,n}\ud835\udc5912\u2026\ud835\udc5bl\\in\\{1,2,...,n\\}italic_l \u2208 { 1 , 2 , \u2026 , italic_n }, let X~=\u22c3l=1nxl~\ud835\udc4bsuperscriptsubscript\ud835\udc591\ud835\udc5bsubscript\ud835\udc65\ud835\udc59\\tilde{X}=\\bigcup_{l=1}^{n}x_{l}over~ start_ARG italic_X end_ARG = \u22c3 start_POSTSUBSCRIPT italic_l = 1 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_n end_POSTSUPERSCRIPT italic_x start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT, let \u03a3Sisubscript\u03a3subscript\ud835\udc46\ud835\udc56\\Sigma_{S_{i}}roman_\u03a3 start_POSTSUBSCRIPT italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT end_POSTSUBSCRIPT be the shrinkage covariance matrix for epoch i\ud835\udc56iitalic_i and let \u03b6\u2208(0,1)\ud835\udf0101\\zeta\\in(0,1)italic_\u03b6 \u2208 ( 0 , 1 ) be the shrinkage parameter. I-STAR loss is a weighted sum between cross-entropy loss, LC\u2062Esubscript\ud835\udc3f\ud835\udc36\ud835\udc38L_{CE}italic_L start_POSTSUBSCRIPT italic_C italic_E end_POSTSUBSCRIPT, and IsoScore\u22c6\u2062(X~,\u03b6,\u03a3Si)superscriptIsoScore\u22c6~\ud835\udc4b\ud835\udf01subscript\u03a3subscript\ud835\udc46\ud835\udc56\\text{IsoScore}^{\\star}(\\tilde{X},\\zeta,\\Sigma_{S_{i}})IsoScore start_POSTSUPERSCRIPT \u22c6 end_POSTSUPERSCRIPT ( over~ start_ARG italic_X end_ARG , italic_\u03b6 , roman_\u03a3 start_POSTSUBSCRIPT italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT end_POSTSUBSCRIPT ) where \u03bb\ud835\udf06\\lambdaitalic_\u03bb is the tuning-parameter. Negative values of \u03bb\ud835\udf06\\lambdaitalic_\u03bb correspond to decreasing isotropy in representations, and positive values of \u03bb\ud835\udf06\\lambdaitalic_\u03bb encourage isotropy.   Given the recent criticism of methods for measuring isotropy, a reassessment of previously accepted theories of isotropy in LLMs is needed. This paper aims to determine the relationship between isotropy and model performance on various language models and fine-tuning tasks. We first propose IsoScore\u22c6, a method for measuring isotropy that is 1) fully differentiable, 2) incorporates classical techniques for covariance estimation to create stable isotropy estimates for mini-batch data, and 3) approximates IsoScore for large sample sizes. We then use IsoScore\u22c6 to develop I-STAR: IsoScore\u22c6-based STable Anisotropic Regularization. I-STAR is a flexible way to adjust isotropy in model representations during training or fine-tuning. In contrast to works that use \u201cflawed\u201d measures of isotropy, we find that",
            "references": [
                {
                    "title": "How Does Fine-tuning Affect the Geometry of Embedding Space: A Case Study on Isotropy",
                    "abstract": "It is widely accepted that fine-tuning pre-trained language models usually brings about performance improvements in downstream tasks. However, there are limited studies on the reasons behind this effectiveness, particularly from the viewpoint of structural changes in the embedding space. Trying to fill this gap, in this paper, we analyze the extent to which the isotropy of the embedding space changes after fine-tuning. We demonstrate that, even though isotropy is a desirable geometrical property, fine-tuning does not necessarily result in isotropy enhancements. Moreover, local structures in pre-trained contextual word representations (CWRs), such as those encoding token types or frequency, undergo a massive change during fine-tuning. Our experiments show dramatic growth in the number of elongated directions in the embedding space, which, in contrast to pre-trained CWRs, carry the essential linguistic knowledge in the fine-tuned embedding space, making existing isotropy enhancement methods ineffective."
                },
                {
                    "title": "All Bark and No Bite: Rogue Dimensions in Transformer Language Models Obscure Representational Quality",
                    "abstract": "Similarity measures are a vital tool for understanding how language models represent and process language. Standard representational similarity measures such as cosine similarity and Euclidean distance have been successfully used in static word embedding models to understand how words cluster in semantic space. Recently, these measures have been applied to embeddings from contextualized models such as BERT and GPT-2. In this work, we call into question the informativity of such measures for contextualized language models. We find that a small number of rogue dimensions, often just 1-3, dominate these measures. Moreover, we find a striking mismatch between the dimensions that dominate similarity measures and those which are important to the behavior of the model. We show that simple postprocessing techniques such as standardization are able to correct for rogue dimensions and reveal underlying representational quality. We argue that accounting for rogue dimensions is essential for any similarity-based analysis of contextual language models."
                },
                {
                    "title": "IsoScore: Measuring the Uniformity of Embedding Space Utilization",
                    "abstract": "The recent success of distributed word representations has led to an increased interest in analyzing the properties of their spatial distribution. Several studies have suggested that contextualized word embedding models do not isotropically project tokens into vector space. However, current methods designed to measure isotropy, such as average random cosine similarity and the partition score, have not been thoroughly analyzed and are not appropriate for measuring isotropy. We propose IsoScore: a novel tool that quantifies the degree to which a point cloud uniformly utilizes the ambient vector space. Using rigorously designed tests, we demonstrate that IsoScore is the only tool available in the literature that accurately measures how uniformly distributed variance is across dimensions in vector space. Additionally, we use IsoScore to challenge a number of recent conclusions in the NLP literature that have been derived using brittle metrics of isotropy. We caution future studies from using existing tools to measure isotropy in contextualized embedding space as resulting conclusions will be misleading or altogether inaccurate."
                },
                {
                    "title": "A Cluster-based Approach for Improving Isotropy in Contextual Embedding Space",
                    "abstract": "The representation degeneration problem in Contextual Word Representations (CWRs) hurts the expressiveness of the embedding space by forming an anisotropic cone where even unrelated words have excessively positive correlations. Existing techniques for tackling this issue require a learning process to re-train models with additional objectives and mostly employ a global assessment to study isotropy. Our quantitative analysis over isotropy shows that a local assessment could be more accurate due to the clustered structure of CWRs. Based on this observation, we propose a local cluster-based method to address the degeneration issue in contextual embedding spaces. We show that in clusters including punctuations and stop words, local dominant directions encode structural information, removing which can improve CWRs performance on semantic tasks. Moreover, we find that tense information in verb representations dominates sense semantics. We show that removing dominant directions of verb representations can transform the space to better suit semantic applications. Our experiments demonstrate that the proposed cluster-based method can mitigate the degeneration problem on multiple tasks."
                },
                {
                    "title": "Too Much in Common: Shifting of Embeddings in Transformer Language Models and its Implications",
                    "abstract": "The success of language models based on the Transformer architecture appears to be inconsistent with observed anisotropic properties of representations learned by such models. We resolve this by showing, contrary to previous studies, that the representations do not occupy a narrow cone, but rather drift in common directions. At any training step, all of the embeddings except for the ground-truth target embedding are updated with gradient in the same direction. Compounded over the training set, the embeddings drift and share common components, manifested in their shape in all the models we have empirically tested. Our experiments show that isotropy can be restored using a simple transformation."
                },
                {
                    "title": "BERT Busters: Outlier Dimensions that Disrupt Transformers",
                    "abstract": "Multiple studies have shown that Transformers are remarkably robust to pruning. Contrary to this received wisdom, we demonstrate that pre-trained Transformer encoders are surprisingly fragile to the removal of a very small number of features in the layer outputs (<0.0001% of model weights). In case of BERT and other pre-trained encoder Transformers, the affected component is the scaling factors and biases in the LayerNorm. The outliers are high-magnitude normalization parameters that emerge early in pre-training and show up consistently in the same dimensional position throughout the model. We show that disabling them significantly degrades both the MLM loss and the downstream task performance. This effect is observed across several BERT-family models and other popular pre-trained Transformer architectures, including BART, XLNet and ELECTRA; we also show a similar effect in GPT-2."
                },
                {
                    "title": "Low Anisotropy Sense Retrofitting (LASeR) : Towards Isotropic and Sense Enriched Representations",
                    "abstract": "Contextual word representation models have shown massive improvements on a multitude of NLP tasks, yet their word sense disambiguation capabilities remain poorly explained. To address this gap, we assess whether contextual word representations extracted from deep pretrained language models create distinguishable representations for different senses of a given word. We analyze the representation geometry and find that most layers of deep pretrained language models create highly anisotropic representations, pointing towards the existence of representation degeneration problem in contextual word representations. After accounting for anisotropy, our study further reveals that there is variability in sense learning capabilities across different language models. Finally, we propose LASeR, a \u2018Low Anisotropy Sense Retrofitting\u2019 approach that renders off-the-shelf representations isotropic and semantically more meaningful, resolving the representation degeneration problem as a post-processing step, and conducting sense-enrichment of contextualized representations extracted from deep neural language models."
                },
                {
                    "title": "Effect of Post-processing on Contextualized Word Representations",
                    "abstract": "Post-processing of static embedding has been shown to improve their performance on both lexical and sequence-level tasks. However, post-processing for contextualized embeddings is an under-studied problem. In this work, we question the usefulness of post-processing for contextualized embeddings obtained from different layers of pre-trained language models. More specifically, we standardize individual neuron activations using z-score, min-max normalization, and by removing top principal components using the all-but-the-top method. Additionally, we apply unit length normalization to word representations. On a diverse set of pre-trained models, we show that post-processing unwraps vital information present in the representations for both lexical tasks (such as word similarity and analogy) and sequence classification tasks. Our findings raise interesting points in relation to the research studies that use contextualized representations, and suggest z-score normalization as an essential step to consider when using them in an application."
                },
                {
                    "title": "Revisiting Representation Degeneration Problem in Language Modeling",
                    "abstract": "Weight tying is now a common setting in many language generation tasks such as language modeling and machine translation. However, a recent study reveals that there is a potential flaw in weight tying. They find that the learned word embeddings are likely to degenerate and lie in a narrow cone when training a language model. They call it the representation degeneration problem and propose a cosine regularization to solve it. Nevertheless, we prove that the cosine regularization is insufficient to solve the problem, as the degeneration is still likely to happen under certain conditions. In this paper, we revisit the representation degeneration problem and theoretically analyze the limitations of the previously proposed solution. Afterward, we propose an alternative regularization method called Laplacian regularization to tackle the problem. Experiments on language modeling demonstrate the effectiveness of the proposed Laplacian regularization."
                },
                {
                    "title": "IsoBN: Fine-Tuning BERT with Isotropic Batch Normalization",
                    "abstract": "Fine-tuning pre-trained language models (PTLMs), such as BERT and its better variant RoBERTa, has been a common practice for advancing performance in natural language understanding (NLU) tasks. Recent advance in representation learning shows that isotropic (i.e., unit-variance and uncorrelated) embeddings can significantly improve performance on downstream tasks with faster convergence and better generalization. The isotropy of the pre-trained embeddings in PTLMs, however, is relatively under-explored. In this paper, we analyze the isotropy of the pre-trained [CLS] embeddings of PTLMs with straightforward visualization, and point out two major issues: high variance in their standard deviation, and high correlation between different dimensions. We also propose a new network regularization method, isotropic batch normalization (IsoBN) to address the issues, towards learning more isotropic representations in fine-tuning by dynamically penalizing dominating principal components. This simple yet effective fine-tuning method yields about 1.0 absolute increment on the average of seven NLU tasks."
                },
                {
                    "title": "Embedding Compression with Isotropic Iterative Quantization",
                    "abstract": "Continuous representation of words is a standard component in deep learning-based NLP models. However, representing a large vocabulary requires significant memory, which can cause problems, particularly on resource-constrained platforms. Therefore, in this paper we propose an isotropic iterative quantization (IIQ) approach for compressing embedding vectors into binary ones, leveraging the iterative quantization technique well established for image retrieval, while satisfying the desired isotropic property of PMI based models. Experiments with pre-trained embeddings (i.e., GloVe and HDC) demonstrate a more than thirty-fold compression ratio with comparable and sometimes even improved performance over the original real-valued embedding vectors."
                },
                {
                    "title": "What do you mean, BERT? Assessing BERT as a Distributional Semantics Model",
                    "abstract": "Contextualized word embeddings, i.e. vector representations for words in context, are naturally seen as an extension of previous noncontextual distributional semantic models. In this work, we focus on BERT, a deep neural network that produces contextualized embeddings and has set the state-of-the-art in several semantic tasks, and study the semantic coherence of its embedding space. While showing a tendency towards coherence, BERT does not fully live up to the natural expectations for a semantic vector space. In particular, we find that the position of the sentence in which a word occurs, while having no meaning correlates, leaves a noticeable trace on the word embeddings and disturbs similarity relationships."
                },
                {
                    "title": "DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter",
                    "abstract": "As Transfer Learning from large-scale pre-trained models becomes more prevalent in Natural Language Processing (NLP), operating these large models in on-the-edge and/or under constrained computational training or inference budgets remains challenging. In this work, we propose a method to pre-train a smaller general-purpose language representation model, called DistilBERT, which can then be fine-tuned with good performances on a wide range of tasks like its larger counterparts. While most prior work investigated the use of distillation for building task-specific models, we leverage knowledge distillation during the pre-training phase and show that it is possible to reduce the size of a BERT model by 40%, while retaining 97% of its language understanding capabilities and being 60% faster. To leverage the inductive biases learned by larger models during pre-training, we introduce a triple loss combining language modeling, distillation and cosine-distance losses. Our smaller, faster and lighter model is cheaper to pre-train and we demonstrate its capabilities for on-device computations in a proof-of-concept experiment and a comparative on-device study."
                },
                {
                    "title": "ALBERT: A Lite BERT for Self-supervised Learning of Language Representations",
                    "abstract": "Increasing model size when pretraining natural language representations often results in improved performance on downstream tasks. However, at some point further model increases become harder due to GPU/TPU memory limitations and longer training times. To address these problems, we present two parameter-reduction techniques to lower memory consumption and increase the training speed of BERT. Comprehensive empirical evidence shows that our proposed methods lead to models that scale much better compared to the original BERT. We also use a self-supervised loss that focuses on modeling inter-sentence coherence, and show it consistently helps downstream tasks with multi-sentence inputs. As a result, our best model establishes new state-of-the-art results on the GLUE, RACE, and \\squad benchmarks while having fewer parameters compared to BERT-large. The code and the pretrained models are available at this https URL."
                },
                {
                    "title": "How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings",
                    "abstract": "Replacing static word embeddings with contextualized word representations has yielded significant improvements on many NLP tasks. However, just how contextual are the contextualized representations produced by models such as ELMo and BERT? Are there infinitely many context-specific representations for each word, or are words essentially assigned one of a finite number of word-sense representations? For one, we find that the contextualized representations of all words are not isotropic in any layer of the contextualizing model. While representations of the same word in different contexts still have a greater cosine similarity than those of two different words, this self-similarity is much lower in upper layers. This suggests that upper layers of contextualizing models produce more context-specific representations, much like how upper layers of LSTMs produce more task-specific representations. In all layers of ELMo, BERT, and GPT-2, on average, less than 5% of the variance in a word\u2019s contextualized representations can be explained by a static embedding for that word, providing some justification for the success of contextualized representations."
                },
                {
                    "title": "Representation Degeneration Problem in Training Natural Language Generation Models",
                    "abstract": "We study an interesting problem in training neural network-based models for natural language generation tasks, which we call the \\emph{representation degeneration problem}. We observe that when training a model for natural language generation tasks through likelihood maximization with the weight tying trick, especially with big training datasets, most of the learnt word embeddings tend to degenerate and be distributed into a narrow cone, which largely limits the representation power of word embeddings. We analyze the conditions and causes of this problem and propose a novel regularization method to address it. Experiments on language modeling and machine translation show that our method can largely mitigate the representation degeneration problem and achieve better performance than baseline algorithms."
                },
                {
                    "title": "Dimensionality compression and expansion in Deep Neural Networks",
                    "abstract": "Datasets such as images, text, or movies are embedded in high-dimensional spaces. However, in important cases such as images of objects, the statistical structure in the data constrains samples to a manifold of dramatically lower dimensionality. Learning to identify and extract task-relevant variables from this embedded manifold is crucial when dealing with high-dimensional problems. We find that neural networks are often very effective at solving this task and investigate why. To this end, we apply state-of-the-art techniques for intrinsic dimensionality estimation to show that neural networks learn low-dimensional manifolds in two phases: first, dimensionality expansion driven by feature generation in initial layers, and second, dimensionality compression driven by the selection of task-relevant features in later layers. We model noise generated by Stochastic Gradient Descent and show how this noise balances the dimensionality of neural representations by inducing an effective regularization term in the loss. We highlight the important relationship between low-dimensional compressed representations and generalization properties of the network. Our work contributes by shedding light on the success of deep neural networks in disentangling data in high-dimensional space while achieving good generalization. Furthermore, it invites new learning strategies focused on optimizing measurable geometric properties of learned representations, beginning with their intrinsic dimensionality."
                },
                {
                    "title": "Intrinsic dimension of data representations in deep neural networks",
                    "abstract": "Deep neural networks progressively transform their inputs across multiple processing layers. What are the geometrical properties of the representations learned by these networks? Here we study the intrinsic dimensionality (ID) of data-representations, i.e. the minimal number of parameters needed to describe a representation. We find that, in a trained network, the ID is orders of magnitude smaller than the number of units in each layer. Across layers, the ID first increases and then progressively decreases in the final layers. Remarkably, the ID of the last hidden layer predicts classification accuracy on the test set. These results can neither be found by linear dimensionality estimates (e.g., with principal component analysis), nor in representations that had been artificially linearized. They are neither found in untrained networks, nor in networks that are trained on randomized labels. This suggests that neural networks that can generalize are those that transform the data into low-dimensional, but not necessarily flat manifolds."
                },
                {
                    "title": "Neural Network Acceptability Judgments",
                    "abstract": "Abstract This paper investigates the ability of artificial neural networks to judge the grammatical acceptability of a sentence, with the goal of testing their linguistic competence. We introduce the Corpus of Linguistic Acceptability (CoLA), a set of 10,657 English sentences labeled as grammatical or ungrammatical from published linguistics literature. As baselines, we train several recurrent neural network models on acceptability classification, and find that our models outperform unsupervised models by Lau et al. (2016) on CoLA. Error-analysis on specific grammatical phenomena reveals that both Lau et al.\u2019s models and ours learn systematic generalizations like subject-verb-object order. However, all models we test perform far below human level on a wide range of grammatical constructions."
                },
                {
                    "title": "Decorrelated Batch Normalization",
                    "abstract": "Batch Normalization (BN) is capable of accelerating the training of deep models by centering and scaling activations within mini-batches. In this work, we propose Decorrelated Batch Normalization (DBN), which not just centers and scales activations but whitens them. We explore multiple whitening techniques, and find that PCA whitening causes a problem we call stochastic axis swapping, which is detrimental to learning. We show that ZCA whitening does not suffer from this problem, permitting successful learning. DBN retains the desirable qualities of BN and further improves BN's optimization efficiency and generalization ability. We design comprehensive experiments to show that DBN can improve the performance of BN on multilayer perceptrons and convolutional neural networks. Furthermore, we consistently improve the accuracy of residual networks on CIFAR-10, CIFAR-100, and ImageNet."
                },
                {
                    "title": "GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding",
                    "abstract": "Human ability to understand language is general, flexible, and robust. In contrast, most NLU models above the word level are designed for a specific task and struggle with out-of-domain data. If we aspire to develop models with understanding beyond the detection of superficial correspondences between inputs and outputs, then it is critical to develop a unified model that can execute a range of linguistic tasks across different domains. To facilitate research in this direction, we present the General Language Understanding Evaluation (GLUE, gluebenchmark.com): a benchmark of nine diverse NLU tasks, an auxiliary dataset for probing models for understanding of specific linguistic phenomena, and an online platform for evaluating and comparing models. For some benchmark tasks, training data is plentiful, but for others it is limited or does not match the genre of the test set. GLUE thus favors models that can represent linguistic knowledge in a way that facilitates sample-efficient learning and effective knowledge-transfer across tasks. While none of the datasets in GLUE were created from scratch for the benchmark, four of them feature privately-held test data, which is used to ensure that the benchmark is used fairly. We evaluate baselines that use ELMo (Peters et al., 2018), a powerful transfer learning technique, as well as state-of-the-art sentence representation models. The best models still achieve fairly low absolute scores. Analysis with our diagnostic dataset yields similarly weak performance over all phenomena tested, with some exceptions."
                },
                {
                    "title": "The Anisotropic Noise in Stochastic Gradient Descent: Its Behavior of Escaping from Sharp Minima and Regularization Effects",
                    "abstract": "Understanding the behavior of stochastic gradient descent (SGD) in the context of deep neural networks has raised lots of concerns recently. Along this line, we study a general form of gradient based optimization dynamics with unbiased noise, which unifies SGD and standard Langevin dynamics. Through investigating this general optimization dynamics, we analyze the behavior of SGD on escaping from minima and its regularization effects. A novel indicator is derived to characterize the efficiency of escaping from minima through measuring the alignment of noise covariance and the curvature of loss function. Based on this indicator, two conditions are established to show which type of noise structure is superior to isotropic noise in term of escaping efficiency. We further show that the anisotropic noise in SGD satisfies the two conditions, and thus helps to escape from sharp and poor minima effectively, towards more stable and flat minima that typically generalize well. We systematically design various experiments to verify the benefits of the anisotropic noise, compared with full gradient descent plus isotropic diffusion (i.e. Langevin dynamics)."
                },
                {
                    "title": "Classification and Geometry of General Perceptual Manifolds",
                    "abstract": "Perceptual manifolds arise when a neural population responds to an ensemble of sensory signals associated with different physical features (e.g., orientation, pose, scale, location, and intensity) of the same perceptual object. Object recognition and discrimination requires classifying the manifolds in a manner that is insensitive to variability within a manifold. How neuronal systems give rise to invariant object classification and recognition is a fundamental problem in brain theory as well as in machine learning. Here we study the ability of a readout network to classify objects from their perceptual manifold representations. We develop a statistical mechanical theory for the linear classification of manifolds with arbitrary geometry revealing a remarkable relation to the mathematics of conic decomposition. Novel geometrical measures of manifold radius and manifold dimension are introduced which can explain the classification capacity for manifolds of various geometries. The general theory is demonstrated on a number of representative manifolds, including L2 ellipsoids prototypical of strictly convex manifolds, L1 balls representing polytopes consisting of finite sample points, and orientation manifolds which arise from neurons tuned to respond to a continuous angle variable, such as object orientation. The effects of label sparsity on the classification capacity of manifolds are elucidated, revealing a scaling relation between label sparsity and manifold radius. Theoretical predictions are corroborated by numerical simulations using recently developed algorithms to compute maximum margin solutions for manifold dichotomies. Our theory and its extensions provide a powerful and rich framework for applying statistical mechanics of linear classification to data arising from neuronal responses to object stimuli, as well as to artificial deep networks trained for object recognition tasks."
                },
                {
                    "title": "SemEval-2017 Task 1: Semantic Textual Similarity Multilingual and Crosslingual Focused Evaluation",
                    "abstract": "Semantic Textual Similarity (STS) measures the meaning similarity of sentences. Applications include machine translation (MT), summarization, generation, question answering (QA), short answer grading, semantic search, dialog and conversational systems. The STS shared task is a venue for assessing the current state-of-the-art. The 2017 task focuses on multilingual and cross-lingual pairs with one sub-track exploring MT quality estimation (MTQE) data. The task obtained strong participation from 31 teams, with 17 participating in all language tracks. We summarize performance and review a selection of well performing methods. Analysis highlights common errors, providing insight into the limitations of existing models. To support ongoing work on semantic representations, the STS Benchmark is introduced as a new shared training and evaluation set carefully selected from the corpus of English STS shared task data (2012-2017)."
                },
                {
                    "title": "All-but-the-Top: Simple and Effective Postprocessing for Word Representations",
                    "abstract": "Real-valued word representations have transformed NLP applications; popular examples are word2vec and GloVe, recognized for their ability to capture linguistic regularities. In this paper, we demonstrate a {\\em very simple}, and yet counter-intuitive, postprocessing technique -- eliminate the common mean vector and a few top dominating directions from the word vectors -- that renders off-the-shelf representations {\\em even stronger}. The postprocessing is empirically validated on a variety of lexical-level intrinsic tasks (word similarity, concept categorization, word analogy) and sentence-level tasks (semantic textural similarity and { text classification}) on multiple datasets and with a variety of representation methods and hyperparameter choices in multiple languages; in each case, the processed representations are consistently better than the original ones."
                },
                {
                    "title": "SQuAD: 100,000+ Questions for Machine Comprehension of Text",
                    "abstract": "We present the Stanford Question Answering Dataset (SQuAD), a new reading comprehension dataset consisting of 100,000+ questions posed by crowdworkers on a set of Wikipedia articles, where the answer to each question is a segment of text from the corresponding reading passage. We analyze the dataset to understand the types of reasoning required to answer the questions, leaning heavily on dependency and constituency trees. We build a strong logistic regression model, which achieves an F1 score of 51.0%, a significant improvement over a simple baseline (20%). However, human performance (86.8%) is much higher, indicating that the dataset presents a good challenge problem for future research. \nThe dataset is freely available at this https URL"
                },
                {
                    "title": "Random Walks on Context Spaces: Towards an Explanation of the Mysteries of Semantic Word Embeddings",
                    "abstract": "The papers of Mikolov et al. 2013 as well as subsequent works have led to dramatic progress in solving word analogy tasks using semantic word embeddings. This leverages linear structure that is often found in the word embeddings, which is surprising since the training method is usually nonlinear. There were attempts ---notably by Levy and Goldberg and Pennington et al.--- to explain how this linear structure arises. The current paper points out the gaps in these explanations and provides a more complete explanation using a loglinear generative model for the corpus that directly models the latent semantic structure in words. The novel methodological twist is that instead of trying to fit the best model parameters to the data, a rigorous mathematical analysis is performed using the model priors to arrive at a simple closed form expression that approximately relates co-occurrence statistics and word embeddings. This expression closely corresponds to ---and a bit simpler than--- the existing training methods, and leads to good solutions to analogy tasks. Empirical support is provided also for the validity of the modeling assumptions. \nThis methodology of letting some mathematical analysis substitute for some of the computational difficulty may be useful in other settings with generative models."
                },
                {
                    "title": "Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank",
                    "abstract": "Semantic word spaces have been very useful but cannot express the meaning of longer phrases in a principled way. Further progress towards understanding compositionality in tasks such as sentiment detection requires richer supervised training and evaluation resources and more powerful models of composition. To remedy this, we introduce a Sentiment Treebank. It includes fine grained sentiment labels for 215,154 phrases in the parse trees of 11,855 sentences and presents new challenges for sentiment compositionality. To address them, we introduce the Recursive Neural Tensor Network. When trained on the new treebank, this model outperforms all previous methods on several metrics. It pushes the state of the art in single sentence positive/negative classification from 80% up to 85.4%. The accuracy of predicting fine-grained sentiment labels for all phrases reaches 80.7%, an improvement of 9.7% over bag of features baselines. Lastly, it is the only model that can accurately capture the effects of negation and its scope at various tree levels for both positive and negative phrases."
                },
                {
                    "title": "R1-PCA: rotational invariant L1-norm principal component analysis for robust subspace factorization",
                    "abstract": "Principal component analysis (PCA) minimizes the sum of squared errors (L2-norm) and is sensitive to the presence of outliers. We propose a rotational invariant L1-norm PCA (R1-PCA). R1-PCA is similar to PCA in that (1) it has a unique global solution, (2) the solution are principal eigenvectors of a robust covariance matrix (re-weighted to soften the effects of outliers), (3) the solution is rotational invariant. These properties are not shared by the L1-norm PCA. A new subspace iteration algorithm is given to compute R1-PCA efficiently. Experiments on several real-life datasets show R1-PCA can effectively handle outliers. We extend R1-norm to K-means clustering and show that L1-norm K-means leads to poor results while R1-K-means outperforms standard K-means."
                },
                {
                    "title": "Regularized Discriminant Analysis",
                    "abstract": "Abstract Linear and quadratic discriminant analysis are considered in the small-sample, high-dimensional setting. Alternatives to the usual maximum likelihood (plug-in) estimates for the covariance matrices are proposed. These alternatives are characterized by two parameters, the values of which are customized to individual situations by jointly minimizing a sample-based estimate of future misclassification risk. Computationally fast implementations are presented, and the efficacy of the approach is examined through simulation studies and application to data. These studies indicate that in many circumstances dramatic gains in classification accuracy can be achieved."
                },
                {
                    "title": "Automatically Constructing a Corpus of Sentential Paraphrases",
                    "abstract": "An obstacle to research in automatic paraphrase identification and generation is the lack of large-scale, publiclyavailable labeled corpora of sentential paraphrases. This paper describes the creation of the recently-released Microsoft Research Paraphrase Corpus, which contains 5801 sentence pairs, each hand-labeled with a binary judgment as to whether the pair constitutes a paraphrase. The corpus was created using heuristic extraction techniques in conjunction with an SVM-based classifier to select likely sentence-level paraphrases from a large corpus of topicclustered news data. These pairs were then submitted to human judges, who confirmed that 67% were in fact semantically equivalent. In addition to describing the corpus itself, we explore a number of issues that arose in defining guidelines for the human raters."
                },
                {
                    "title": "The PASCAL Recognising Textual Entailment Challenge",
                    "abstract": "This paper describes the PASCAL Network of Excellence first Recognising Textual Entailment (RTE-1) Challenge benchmark 1 . The RTE task is defined as recognizing, given two text fragments, whether the meaning of one text can be inferred (entailed) from the other. This application-independent task is suggested as capturing major inferences about the variability of semantic expression which are commonly needed across multiple applications. The Challenge has raised noticeable attention in the research community, attracting 17 submissions from diverse groups, suggesting the generic relevance of the task."
                }
            ],
            "categories": [
                "cs.CL",
                "cs.AI",
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does isotropy in Large Language Model (LLM) representations affect model performance on various language tasks, and how can we effectively measure and regularize isotropy during training?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the ongoing debate about the role of isotropy in LLMs, which has implications for the understanding of representation quality in neural networks. By providing a robust method for measuring isotropy and a regularization technique, this research could lead to improved model performance across various tasks, enhancing the expressive power of embeddings. Furthermore, it could inspire future research into representation learning and regularization techniques, ultimately leading to more effective and interpretable language models.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the complexities of accurately measuring isotropy in high-dimensional spaces, particularly with small sample sizes where existing methods like IsoScore are neither differentiable nor stable. Naive approaches may fail because they do not account for the nuances of covariance estimation or the need for differentiability in optimization processes. Additionally, the presence of outlier dimensions in LLM representations complicates the task of achieving isotropy, as it requires sophisticated techniques to mitigate their influence.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the inadequacies of existing isotropy measurement methods, which have been criticized for their flaws. These limitations have created barriers to understanding the true impact of isotropy on model performance. Additionally, the lack of a differentiable and stable method for isotropy measurement has hindered the development of effective regularization techniques. Our approach differs by introducing IsoScore\u22c6, which overcomes these limitations and provides a more reliable framework for assessing and adjusting isotropy in LLMs.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves the development of IsoScore\u22c6, a differentiable method for measuring isotropy that incorporates classical covariance estimation techniques to ensure stability with mini-batch data. We will utilize this method to create I-STAR, a regularization technique that adjusts isotropy during training. The expected outcomes include improved isotropy in LLM representations, leading to enhanced performance on downstream language tasks, as well as a clearer understanding of the relationship between isotropy and model effectiveness. The evaluation will be based on performance metrics relevant"
            }
        },
        "author_data": {
            "7f5d488a-269f-4c0d-a98b-f4270ddaf266": {
                "pk": "7f5d488a-269f-4c0d-a98b-f4270ddaf266",
                "name": "William Rudman",
                "collaborators": [
                    "Carsten Eickhoff",
                    "Jack Merullo",
                    "L. Mercurio",
                    "William Jurayj",
                    "Michal Golovanevsky",
                    "Vedant Palit",
                    "Ritambhara Singh",
                    "Catherine Chen",
                    "Isaac Sears",
                    "Augusto Garcia-Agundez"
                ],
                "domain": [
                    "Vision-Language Models",
                    "Interpretability",
                    "Deep Learning",
                    "Cardiac Abnormalities"
                ],
                "publications": [
                    {
                        "title": "What Do VLMs NOTICE? A Mechanistic Interpretability Pipeline for Noise-free Text-Image Corruption and Evaluation",
                        "abstract": "Vision-Language Models (VLMs) have gained community-spanning prominence due to their ability to integrate visual and textual inputs to perform complex tasks. Despite their success, the internal decision-making processes of these models remain opaque, posing challenges in high-stakes applications. To address this, we introduce NOTICE, the first Noise-free Text-Image Corruption and Evaluation pipeline for mechanistic interpretability in VLMs. NOTICE incorporates a Semantic Minimal Pairs (SMP) framework for image corruption and Symmetric Token Replacement (STR) for text. This approach enables semantically meaningful causal mediation analysis for both modalities, providing a robust method for analyzing multimodal integration within models like BLIP. Our experiments on the SVO-Probes, MIT-States, and Facial Expression Recognition datasets reveal crucial insights into VLM decision-making, identifying the significant role of middle-layer cross-attention heads. Further, we uncover a set of ``universal cross-attention heads'' that consistently contribute across tasks and modalities, each performing distinct functions such as implicit image segmentation, object inhibition, and outlier inhibition. This work paves the way for more transparent and interpretable multimodal systems."
                    },
                    {
                        "title": "Outlier Dimensions Encode Task-Specific Knowledge",
                        "abstract": "Representations from large language models (LLMs) are known to be dominated by a small subset of dimensions with exceedingly high variance. Previous works have argued that although ablating these outlier dimensions in LLM representations hurts downstream performance, outlier dimensions are detrimental to the representational quality of embeddings. In this study, we investigate how fine-tuning impacts outlier dimensions and show that 1) outlier dimensions that occur in pre-training persist in fine-tuned models and 2) a single outlier dimension can complete downstream tasks with a minimal error rate. Our results suggest that outlier dimensions can encode crucial task-specific knowledge and that the value of a representation in a single outlier dimension drives downstream model decisions."
                    },
                    {
                        "title": "Leveraging Unlabeled Electroencephalographic Data to Predict Neurological Recovery for Comatose Patients Following Cardiac Arrest",
                        "abstract": "In response to the 2023 George B. Moody PhysioNet Challenge, we propose an automated, unsupervised pretraining approach to boost the performance of models that predict neurologic outcomes after cardiac arrest. Our team, (BrownBAI), developed a model architecture consisting of three parts: a pre-processor to convert raw electroencephalograms (EEGs) into two-dimensional spectrograms, a three-layer convolutional neural network (CNN) encoder for unsupervised pre-training, and a time series transformer (TST) model. We trained the CNN encoder on unlabeled five-minute EEG samples from the Temple University EEG Corpus (TUEG), which included more than 20x the patients available in the PhysioNet competition training dataset. We then incorporated the pre-trained encoder into the TST as a base layer and trained the composite model as a classifier on EEGs from the 2023 PhysioNet Challenge dataset. Our team was not able to submit an official competition entry and was therefore not scored on the test set. However, in a side-by-side comparison on the competition training dataset, our model performed better with a pretrained (competition score 0.351), rather than randomly initialized (competition score 0.211) CNN encoder layer. These results show the potential benefits of leveraging unlabeled data to boost task-specific performance of predictive EEG models."
                    },
                    {
                        "title": "ACQuA: Anomaly Classification with Quasi-Attractors",
                        "abstract": "In recent years, deep learning has redefined algorithms for detecting cardiac abnormalities. However, many state of the art algorithms still rely on calculating handcrafted features from a given heart signal that are then fed into shallow 1D convolutional networks or transformer architectures. We propose ACQuA (Anomaly Classification with Quasi Attractors), a task agnostic algorithm that can be used in a wide variety of cardiac settings, from classifying cardiac arrhythmias from ECG signals to detecting heart murmurs from PCG signals. Using theorems from dynamical analysis and topological data analysis, we create informative attractor images that 1) are human distinguishable and 2) can be used to train small, off the shelf deep neural networks for anomaly classification. In the George B. Moody 2022 Challenge, we receive an official score of 0.433 (263/305) for murmur classification and a score of 12616 (208/305) for outcome classification. Additionally, we evaluate our model on the CinC 2017 Challenge data that tasks practitioners to classify cardiac arrhythmias from ECG signals. On the CinC 2017 Challenge data, we improve upon the winning F1 scores by approximately 14% on the hidden validation data."
                    },
                    {
                        "title": "ACQuA: Anomaly Classification with Quasi-Attractors",
                        "abstract": "In recent years, deep learning has redefined algorithms for detecting cardiac abnormalities. However, many state of the art algorithms still rely on calculating handcrafted features from a given heart signal that are then fed into shallow 1D convolutional networks or transformer architectures. We propose ACQuA (Anomaly Classification with Quasi Attractors), a task agnostic algorithm that can be used in a wide variety of cardiac settings, from classifying cardiac arrhythmias from ECG signals to detecting heart murmurs from PCG signals. Using theorems from dynamical analysis and topological data analysis, we create informative attractor images that 1) are human distinguishable and 2) can be used to train neural networks for anomaly classification. In the George B. Moody 2022 Challenge, our team, BrownBAI, received an official score of 0.406 (38/40) for murmur classification and a score of 16773 (39/40) for outcome classification. Additionally, we evaluate our model on the CinC 2017 Challenge data that tasks practitioners to classify cardiac arrhythmias from ECG signals. On the CinC 2017 Challenge data, we improve upon the winning F1 scores by approximately 14% on the hidden validation data."
                    },
                    {
                        "title": "Garden-Path Traversal within GPT-2",
                        "abstract": "In recent years, massive language models consisting exclusively of transformer decoders, led by the GPT-x family, have become increasingly popular. While studies have examined the behavior of these models, they tend to only focus on the output of the language model, avoiding analyzing their internal states despite such analyses being popular tools used within BERTology to study transformer encoders. We present a collection of methods for analyzing GPT-2\u2019s hidden states, and use the model\u2019s navigation of garden path sentences as a case study to demonstrate the utility of studying this model\u2019s behavior beyond its output alone. To support this analysis, we introduce a novel dataset consisting of 3 different types of garden path sentences, along with scripts to manipulate them. We \ufb01nd that measuring Manhattan distances and cosine similarities between hidden states shows that GPT-2 navigates these sentences more intuitively than conventional methods that predict from the model\u2019s output alone."
                    },
                    {
                        "title": "Garden Path Traversal in GPT-2",
                        "abstract": "In recent years, large-scale transformer decoders such as the GPT-x family of models have become increasingly popular. Studies examining the behavior of these models tend to focus only on the output of the language modeling head and avoid analysis of the internal states of the transformer decoder. In this study, we present a collection of methods to analyze the hidden states of GPT-2 and use the model\u2019s navigation of garden path sentences as a case study. To enable this, we compile the largest currently available dataset of garden path sentences. We show that Manhattan distances and cosine similarities provide more reliable insights compared to established surprisal methods that analyze next-token probabilities computed by a language modeling head. Using these methods, we find that negating tokens have minimal impacts on the model\u2019s representations for unambiguous forms of sentences with ambiguity solely over what the object of a verb is, but have a more substantial impact of representations for unambiguous sentences whose ambiguity would stem from the voice of a verb. Further, we find that analyzing the decoder model\u2019s hidden states reveals periods of ambiguity that might conclude in a garden path effect but happen not to, whereas surprisal analyses routinely miss this detail."
                    }
                ]
            },
            "83c57cca-6362-4a7d-8ced-7d91e0225bfc": {
                "pk": "83c57cca-6362-4a7d-8ced-7d91e0225bfc",
                "name": "Carsten Eickhoff",
                "collaborators": [
                    "Navid Rekabsaz",
                    "George Zerveas",
                    "Jack Merullo",
                    "Daniel Cohen",
                    "William Rudman",
                    "Ellie Pavlick",
                    "L. Mercurio",
                    "Seyed Ali Bahrainian",
                    "Oleg Lesota",
                    "Bhaskar Mitra"
                ],
                "domain": [
                    "Natural Language Processing",
                    "Machine Learning",
                    "Multimodal Learning",
                    "Retrieval Models"
                ],
                "publications": [
                    {
                        "title": "Enhancing the Ranking Context of Dense Retrieval through Reciprocal Nearest Neighbors",
                        "abstract": "Sparse annotation poses persistent challenges to training dense retrieval models; for example, it distorts the training signal when unlabeled relevant documents are used spuriously as negatives in contrastive learning. To alleviate this problem, we introduce evidence-based label smoothing, a novel, computationally efficient method that prevents penalizing the model for assigning high relevance to false negatives. To compute the target relevance distribution over candidate documents within the ranking context of a given query, we assign a non-zero relevance probability to those candidates most similar to the ground truth based on the degree of their similarity to the ground-truth document(s). To estimate relevance we leverage an improved similarity metric based on reciprocal nearest neighbors, which can also be used independently to rerank candidates in post-processing. Through extensive experiments on two large-scale ad hoc text retrieval datasets, we demonstrate that reciprocal nearest neighbors can improve the ranking effectiveness of dense retrieval models, both when used for label smoothing, as well as for reranking. This indicates that by considering relationships between documents and queries beyond simple geometric distance we can effectively enhance the ranking context."
                    },
                    {
                        "title": "Enhancing the Ranking Context of Dense Retrieval Methods through Reciprocal Nearest Neighbors",
                        "abstract": "Sparse annotation poses persistent challenges to training dense retrieval models, such as the problem of false negatives, i.e. unlabeled relevant documents that are spuriously used as negatives in contrastive learning, distorting the training signal. To alleviate this problem, we introduce evidence-based label smoothing, a computationally efficient method that prevents penalizing the model for assigning high relevance to false negatives. To compute the target relevance distribution over candidate documents within the ranking context of a given query, candidates most similar to the ground truth are assigned a non-zero relevance probability based on the degree of their similarity to the ground-truth document(s). As a relevance estimate we leverage an improved similarity metric based on reciprocal nearest neighbors, which can also be used independently to rerank candidates in post-processing. Through extensive experiments on two large-scale ad hoc text retrieval datasets we demonstrate that both methods can improve the ranking effectiveness of dense retrieval models."
                    },
                    {
                        "title": "One-Versus-Others Attention: Scalable Multimodal Integration",
                        "abstract": "Multimodal learning models have become increasingly important as they surpass single-modality approaches on diverse tasks ranging from question-answering to autonomous driving. Despite the importance of multimodal learning, existing efforts focus on NLP applications, where the number of modalities is typically less than four (audio, video, text, images). However, data inputs in other domains, such as the medical field, may include X-rays, PET scans, MRIs, genetic screening, clinical notes, and more, creating a need for both efficient and accurate information fusion. Many state-of-the-art models rely on pairwise cross-modal attention, which does not scale well for applications with more than three modalities. For $n$ modalities, computing attention will result in $n \\choose 2$ operations, potentially requiring considerable amounts of computational resources. To address this, we propose a new domain-neutral attention mechanism, One-Versus-Others (OvO) attention, that scales linearly with the number of modalities and requires only $n$ attention operations, thus offering a significant reduction in computational complexity compared to existing cross-modal attention algorithms. Using three diverse real-world datasets as well as an additional simulation experiment, we show that our method improves performance compared to popular fusion techniques while decreasing computation costs."
                    },
                    {
                        "title": "Neural Summarization of Electronic Health Records",
                        "abstract": "Hospital discharge documentation is among the most essential, yet time-consuming documents written by medical practitioners. The objective of this study was to automatically generate hospital discharge summaries using neural network summarization models. We studied various data preparation and neural network training techniques that generate discharge summaries. Using nursing notes and discharge summaries from the MIMIC-III dataset, we studied the viability of the automatic generation of various sections of a discharge summary using four state-of-the-art neural network summarization models (BART, T5, Longformer and FLAN-T5). Our experiments indicated that training environments including nursing notes as the source, and discrete sections of the discharge summary as the target output (e.g.\"History of Present Illness\") improve language model efficiency and text quality. According to our findings, the fine-tuned BART model improved its ROUGE F1 score by 43.6% against its standard off-the-shelf version. We also found that fine-tuning the baseline BART model with other setups caused different degrees of improvement (up to 80% relative improvement). We also observed that a fine-tuned T5 generally achieves higher ROUGE F1 scores than other fine-tuned models and a fine-tuned FLAN-T5 achieves the highest ROUGE score overall, i.e., 45.6. For majority of the fine-tuned language models, summarizing discharge summary report sections separately outperformed the summarization the entire report quantitatively. On the other hand, fine-tuning language models that were previously instruction fine-tuned showed better performance in summarizing entire reports. This study concludes that a focused dataset designed for the automatic generation of discharge summaries by a language model can produce coherent Discharge Summary sections."
                    },
                    {
                        "title": "Language Models Implement Simple Word2Vec-style Vector Arithmetic",
                        "abstract": "A primary criticism towards language models (LMs) is their inscrutability. This paper presents evidence that, despite their size and complexity, LMs sometimes exploit a simple vector arithmetic style mechanism to solve some relational tasks using regularities encoded in the hidden space of the model (e.g., Poland:Warsaw::China:Beijing). We investigate a range of language model sizes (from 124M parameters to 176B parameters) in an in-context learning setting, and find that for a variety of tasks (involving capital cities, uppercasing, and past-tensing) a key part of the mechanism reduces to a simple additive update typically applied by the feedforward (FFN) networks. We further show that this mechanism is specific to tasks that require retrieval from pretraining memory, rather than retrieval from local context. Our results contribute to a growing body of work on the interpretability of LMs, and offer reason to be optimistic that, despite the massive and non-linear nature of the models, the strategies they ultimately use to solve tasks can sometimes reduce to familiar and even intuitive algorithms."
                    },
                    {
                        "title": "SIMSUM: Document-level Text Simplification via Simultaneous Summarization",
                        "abstract": "Document-level text simplification is a specific type of simplification which involves simplifying documents consisting of several sentences by rewriting them into fewer or more sentences. In this paper, we propose a new two-stage framework SIMSUM for automated document-level text simplification. Our model is designed with explicit summarization and simplification models and guides the generation using the main keywords of a source text.In order to evaluate our new model, we use two existing benchmark datasets for simplification, namely D-Wikipedia and Wiki-Doc. We compare our model\u2019s performance with state of the art and show that SIMSUM achieves top results on the D-Wikipedia dataset SARI (+1.20), D-SARI (+1.64), and FKGL (-0.35) scores, improving over the best baseline models. In order to evaluate the quality of the generated text, we analyze the outputs from different models qualitatively and demonstrate the merit of our new model. Our code and datasets are available."
                    },
                    {
                        "title": "Parameter-efficient Modularised Bias Mitigation via AdapterFusion",
                        "abstract": "Large pre-trained language models contain societal biases and carry along these biases to downstream tasks. Current in-processing bias mitigation approaches (like adversarial training) impose debiasing by updating a model\u2019s parameters, effectively transferring the model to a new, irreversible debiased state. In this work, we propose a novel approach to develop stand-alone debiasing functionalities separate from the model, which can be integrated into the model on-demand, while keeping the core model untouched. Drawing from the concept of AdapterFusion in multi-task learning, we introduce DAM (Debiasing with Adapter Modules) \u2013 a debiasing approach to first encapsulate arbitrary bias mitigation functionalities into separate adapters, and then add them to the model on-demand in order to deliver fairness qualities. We conduct a large set of experiments on three classification tasks with gender, race, and age as protected attributes. Our results show that DAM improves or maintains the effectiveness of bias mitigation, avoids catastrophic forgetting in a multi-attribute scenario, and maintains on-par task performance, while granting parameter-efficiency and easy switching between the original and debiased models."
                    },
                    {
                        "title": "Unsupervised Multivariate Time-Series Transformers for Seizure Identification on EEG",
                        "abstract": "Epilepsy is one of the most common neurological disorders, typically observed via seizure episodes. Epileptic seizures are commonly monitored through electroencephalogram (EEG) recordings due to their routine and low expense collection. The stochastic nature of EEG makes seizure identification via manual inspections performed by highly-trained experts a tedious endeavor, motivating the use of automated identification. The literature on automated identification focuses mostly on supervised learning methods requiring expert labels of EEG segments that contain seizures, which are difficult to obtain. Motivated by these observations, we pose seizure identification as an unsupervised anomaly detection problem. To this end, we employ the first unsupervised transformer-based model for seizure identification on raw EEG. We train an autoencoder involving a transformer encoder via an unsupervised loss function, incorporating a novel masking strategy uniquely designed for multivariate time-series data such as EEG. Training employs EEG recordings that do not contain any seizures, while seizures are identified with respect to reconstruction errors at inference time. We evaluate our method on three publicly available benchmark EEG datasets for distinguishing seizure vs. non-seizure windows. Our method leads to significantly better seizure identification performance than supervised learning counterparts, by up to 16% recall, 9% accuracy, and 9% Area under the Receiver Operating Characteristics Curve (AUC), establishing particular benefits on highly imbalanced data. Through accurate seizure identification, our method could facilitate widely accessible and early detection of epilepsy development, without needing expensive label collection or manual feature extraction."
                    },
                    {
                        "title": "Wasserstein Adversarial Learning based Temporal Knowledge Graph Embedding",
                        "abstract": "Research on knowledge graph embedding (KGE) has emerged as an active field in which most existing KGE approaches mainly focus on static structural data and ignore the influence of temporal variation involved in time-aware triples. In order to deal with this issue, several temporal knowledge graph embedding (TKGE) approaches have been proposed to integrate temporal and structural information in recent years. However, these methods only employ a uniformly random sampling to construct negative facts. As a consequence, the corrupted samples are often too simplistic for training an effective model. In this paper, we propose a new temporal knowledge graph embedding framework by introducing adversarial learning to further refine the performance of traditional TKGE models. In our framework, a generator is utilized to construct high-quality plausible quadruples and a discriminator learns to obtain the embeddings of entities and relations based on both positive and negative samples. Meanwhile, we also apply a Gumbel-Softmax relaxation and the Wasserstein distance to prevent vanishing gradient problems on discrete data; an inherent flaw in traditional generative adversarial networks. Through comprehensive experimentation on temporal datasets, the results indicate that our proposed framework can attain significant improvements based on benchmark models and also demonstrate the effectiveness and applicability of our framework."
                    },
                    {
                        "title": "Inconsistent Ranking Assumptions in Medical Search and Their Downstream Consequences",
                        "abstract": "Given a query, neural retrieval models predict point estimates of relevance for each document; however, a significant drawback of relying solely on point estimates is that they contain no indication of the model's confidence in its predictions. Despite this lack of information, downstream methods such as reranking, cutoff prediction, and none-of-the-above classification are still able to learn effective functions to accomplish their respective tasks. Unfortunately, these downstream methods can suffer poor performance when the initial ranking model loses confidence in its score predictions. This becomes increasingly important in high-stakes settings, such as medical searches that can influence health decision making. Recent work has resolved this lack of information by introducing Bayesian uncertainty to capture the possible distribution of a document score. This paper presents the use of this uncertainty information as an indicator of how well downstream methods will function over a ranklist. We highlight a significant bias against certain disease-related queries within the posterior distribution of a neural model, and show that this bias in a model's predictive distribution propagates to downstream methods. Finally, we introduce a multi-distribution uncertainty metric, confidence decay, as a valid way of partially identifying these failure cases in an offline setting without the need of any user feedback."
                    },
                    {
                        "title": "Pretraining on Interactions for Learning Grounded Affordance Representations",
                        "abstract": "Lexical semantics and cognitive science point to affordances (i.e. the actions that objects support) as critical for understanding and representing nouns and verbs. However, study of these semantic features has not yet been integrated with the ?foundation? models that currently dominate language representation research. We hypothesize that predictive modeling of object state over time will result in representations that encode object affordance information ?for free?. We train a neural network to predict objects? trajectories in a simulated interaction and show that our network?s latent representations differentiate between both observed and unobserved affordances. We find that models trained using 3D simulations outperform conventional 2D computer vision models trained on a similar task, and, on initial inspection, that differences between concepts correspond to expected features (e.g., roll entails rotation) . Our results suggest a way in which modern deep learning approaches to grounded language learning can be integrated with traditional formal semantic notions of lexical representations."
                    },
                    {
                        "title": "ACQuA: Anomaly Classification with Quasi-Attractors",
                        "abstract": "In recent years, deep learning has redefined algorithms for detecting cardiac abnormalities. However, many state of the art algorithms still rely on calculating handcrafted features from a given heart signal that are then fed into shallow 1D convolutional networks or transformer architectures. We propose ACQuA (Anomaly Classification with Quasi Attractors), a task agnostic algorithm that can be used in a wide variety of cardiac settings, from classifying cardiac arrhythmias from ECG signals to detecting heart murmurs from PCG signals. Using theorems from dynamical analysis and topological data analysis, we create informative attractor images that 1) are human distinguishable and 2) can be used to train small, off the shelf deep neural networks for anomaly classification. In the George B. Moody 2022 Challenge, we receive an official score of 0.433 (263/305) for murmur classification and a score of 12616 (208/305) for outcome classification. Additionally, we evaluate our model on the CinC 2017 Challenge data that tasks practitioners to classify cardiac arrhythmias from ECG signals. On the CinC 2017 Challenge data, we improve upon the winning F1 scores by approximately 14% on the hidden validation data."
                    },
                    {
                        "title": "ACQuA: Anomaly Classification with Quasi-Attractors",
                        "abstract": "In recent years, deep learning has redefined algorithms for detecting cardiac abnormalities. However, many state of the art algorithms still rely on calculating handcrafted features from a given heart signal that are then fed into shallow 1D convolutional networks or transformer architectures. We propose ACQuA (Anomaly Classification with Quasi Attractors), a task agnostic algorithm that can be used in a wide variety of cardiac settings, from classifying cardiac arrhythmias from ECG signals to detecting heart murmurs from PCG signals. Using theorems from dynamical analysis and topological data analysis, we create informative attractor images that 1) are human distinguishable and 2) can be used to train neural networks for anomaly classification. In the George B. Moody 2022 Challenge, our team, BrownBAI, received an official score of 0.406 (38/40) for murmur classification and a score of 16773 (39/40) for outcome classification. Additionally, we evaluate our model on the CinC 2017 Challenge data that tasks practitioners to classify cardiac arrhythmias from ECG signals. On the CinC 2017 Challenge data, we improve upon the winning F1 scores by approximately 14% on the hidden validation data."
                    },
                    {
                        "title": "Garden-Path Traversal within GPT-2",
                        "abstract": "In recent years, massive language models consisting exclusively of transformer decoders, led by the GPT-x family, have become increasingly popular. While studies have examined the behavior of these models, they tend to only focus on the output of the language model, avoiding analyzing their internal states despite such analyses being popular tools used within BERTology to study transformer encoders. We present a collection of methods for analyzing GPT-2\u2019s hidden states, and use the model\u2019s navigation of garden path sentences as a case study to demonstrate the utility of studying this model\u2019s behavior beyond its output alone. To support this analysis, we introduce a novel dataset consisting of 3 different types of garden path sentences, along with scripts to manipulate them. We \ufb01nd that measuring Manhattan distances and cosine similarities between hidden states shows that GPT-2 navigates these sentences more intuitively than conventional methods that predict from the model\u2019s output alone."
                    },
                    {
                        "title": "Garden Path Traversal in GPT-2",
                        "abstract": "In recent years, large-scale transformer decoders such as the GPT-x family of models have become increasingly popular. Studies examining the behavior of these models tend to focus only on the output of the language modeling head and avoid analysis of the internal states of the transformer decoder. In this study, we present a collection of methods to analyze the hidden states of GPT-2 and use the model\u2019s navigation of garden path sentences as a case study. To enable this, we compile the largest currently available dataset of garden path sentences. We show that Manhattan distances and cosine similarities provide more reliable insights compared to established surprisal methods that analyze next-token probabilities computed by a language modeling head. Using these methods, we find that negating tokens have minimal impacts on the model\u2019s representations for unambiguous forms of sentences with ambiguity solely over what the object of a verb is, but have a more substantial impact of representations for unambiguous sentences whose ambiguity would stem from the voice of a verb. Further, we find that analyzing the decoder model\u2019s hidden states reveals periods of ambiguity that might conclude in a garden path effect but happen not to, whereas surprisal analyses routinely miss this detail."
                    },
                    {
                        "title": "Linearly Mapping from Image to Text Space",
                        "abstract": "The extent to which text-only language models (LMs) learn to represent features of the non-linguistic world is an open question. Prior work has shown that pretrained LMs can be taught to caption images when a vision model's parameters are optimized to encode images in the language space. We test a stronger hypothesis: that the conceptual representations learned by frozen text-only models and vision-only models are similar enough that this can be achieved with a linear map. We show that the image representations from vision models can be transferred as continuous prompts to frozen LMs by training only a single linear projection. Using these to prompt the LM achieves competitive performance on captioning and visual question answering tasks compared to models that tune both the image encoder and text decoder (such as the MAGMA model). We compare three image encoders with increasing amounts of linguistic supervision seen during pretraining: BEIT (no linguistic information), NF-ResNET (lexical category information), and CLIP (full natural language descriptions). We find that all three encoders perform equally well at transferring visual property information to the language model (e.g., whether an animal is large or small), but that image encoders pretrained with linguistic supervision more saliently encode category information (e.g., distinguishing hippo vs. elephant) and thus perform significantly better on benchmark language-and-vision tasks. Our results indicate that LMs encode conceptual information structurally similarly to vision-based models, even those that are solely trained on images. Code is available here: https://github.com/jmerullo/limber"
                    },
                    {
                        "title": "Not All Relevance Scores are Equal: Efficient Uncertainty and Calibration Modeling for Deep Retrieval Models",
                        "abstract": "In any ranking system, the retrieval model outputs a single score for a document based on its belief on how relevant it is to a given search query. While retrieval models have continued to improve with the introduction of increasingly complex architectures, few works have investigated a retrieval model's belief in the score beyond the scope of a single value. We argue that capturing the model's uncertainty with respect to its own scoring of a document is a critical aspect of retrieval that allows for greater use of current models across new document distributions, collections, or even improving effectiveness for down-stream tasks. In this paper, we address this problem via an efficient Bayesian framework for retrieval models which captures the model's belief in the relevance score through a stochastic process while adding only negligible computational overhead. We evaluate this belief via a ranking based calibration metric showing that our approximate Bayesian framework significantly improves a retrieval model's ranking effectiveness through a risk aware reranking as well as its confidence calibration. Lastly, we demonstrate that this additional uncertainty information is actionable and reliable on down-stream tasks represented via cutoff prediction."
                    },
                    {
                        "title": "IsoScore: Measuring the Uniformity of Embedding Space Utilization",
                        "abstract": "The recent success of distributed word representations has led to an increased interest in analyzing the properties of their spatial distribution. Several studies have suggested that contextualized word embedding models do not isotropically project tokens into vector space. However, current methods designed to measure isotropy, such as average random cosine similarity and the partition score, have not been thoroughly analyzed and are not appropriate for measuring isotropy. We propose IsoScore: a novel tool that quantifies the degree to which a point cloud uniformly utilizes the ambient vector space. Using rigorously designed tests, we demonstrate that IsoScore is the only tool available in the literature that accurately measures how uniformly distributed variance is across dimensions in vector space. Additionally, we use IsoScore to challenge a number of recent conclusions in the NLP literature that have been derived using brittle metrics of isotropy. We caution future studies from using existing tools to measure isotropy in contextualized embedding space as resulting conclusions will be misleading or altogether inaccurate."
                    },
                    {
                        "title": "CODER: An efficient framework for improving retrieval through COntextualized Document Embedding Reranking",
                        "abstract": "We present a framework for improving the performance of a wide class of retrieval models at minimal computational cost. It utilizes pre-computed document representations extracted by a base dense retrieval method and involves training a model to jointly score a large set of retrieved candidate documents for each query, while potentially transforming on the \ufb02y the representation of each document in the context of the other candidates as well as the query itself. When scoring a document representation based on its similarity to a query, the model is thus aware of the representation of its \u201cpeer\u201d documents. We show that our approach leads to substantial improvement in retrieval performance over scoring candidate documents in isolation from one another, as in a pair-wise training setting. Crucially, unlike term-interaction rerankers based on BERT-like encoders, it incurs a negligible computational overhead on top of any \ufb01rst-stage method at run time, allowing it to be easily combined with any state-of-the-art dense retrieval method. Finally, concurrently considering a set of candidate documents for a given query enables additional valuable capabilities in retrieval, such as score calibration and mitigating societal biases in ranking."
                    }
                ]
            }
        }
    },
    "2305.01377": {
        "paper_data": {
            "title": "Random Function Descent",
            "url": "http://arxiv.org/abs/2305.01377v3",
            "arxiv_id": "2305.01377",
            "authors": [
                "Felix Benning",
                "Leif D\u00f6ring"
            ],
            "abstract": "Classical worst-case optimization theory neither explains the success of optimization in machine learning, nor does it help with step size selection. In this paper we demonstrate the viability and advantages of replacing the classical 'convex function' framework with a 'random function' framework. With complexity $\\mathcal{O}(n^3d^3)$, where $n$ is the number of steps and $d$ the number of dimensions, Bayesian optimization with gradients has not been viable in large dimension so far. By bridging the gap between Bayesian optimization (i.e. random function optimization theory) and classical optimization we establish viability. Specifically, we use a 'stochastic Taylor approximation' to rediscover gradient descent, which is scalable in high dimension due to $\\mathcal{O}(nd)$ complexity. This rediscovery yields a specific step size schedule we call Random Function Descent (RFD). The advantage of this random function framework is that RFD is scale invariant and that it provides a theoretical foundation for common step size heuristics such as gradient clipping and gradual learning rate warmup.",
            "introduction": " Introduction Cost function minimization is one of the most fundamental mathematical problems in machine learning. Gradient-based methods like RMSProp [22], AdaDelta [57] and in particular Adam [29] replace this sum with an exponential mean estimate, i.e. in the case of Adam the decay rate \u03b22is used to get an exponential moving average vt=\u03b22vt\u22121+ (1\u2212\u03b22) diag( \u2207J(wt)\u2207J(wt)T) =\u03b22vt\u22121+ (1\u2212\u03b22)(\u2207J(wt))2. They then take the expectation E[vt] =Eh (1\u2212\u03b22)tX k=1\u03b2t\u2212k 2\u2207J(wk)2i =Eh (1\u2212\u03b22)tX k=1\u03b2t\u2212k 2\u2207J(wk)2i = (1\u2212\u03b2t 2)E[\u2207J(xt)2]|{z} \u221ddiag(\u03a3) So\u02c6vt=vt/(1\u2212\u03b2t 2)in the Adam optimizer is essentially an estimator for diag(\u03a3) . It is noteworthy, that Kingma and Ba [29] already used the expectation symbol. This is despite the fact, that they did not yet model the optimization objective Jas a random function. We can not yet explain why they then use the square root of their estimate diag(\u03a3)\u22121/2instead of diag(\u03a3)\u22121itself. This might have something to do with the fact that the estimation of Gthappens online and the J(wk)are therefore highly correlated. Another reason might be that the inverse of an estimator has different properties than the estimator itself. Finally, the fact that only the diagonal is used might also be the reason, if the preconditioner diag(\u03a3)\u22121/2is simply better when we restrict ourselves to diagonal matrices. 35E.1.1 Proof of Extension E.2 Since the application of scale invariance provides no intuition, we provide a proof which retraces some of the steps of the original proof. Recall, that for an isotropic random function gwe have the stochastic Taylor approximation E[g(w\u2212d)|g(x),\u2207g(x)] =\u00b5+C\u0000\u2225d\u22252 2\u0001 C(0)(g(w)\u2212\u00b5) +C\u2032\u0000\u2225d\u22252 2\u0001 C\u2032(0)\u27e8d,\u2207g(w)\u27e9 This implies for a random function with geometric anisotropy J(w) =g(Aw)that E[J(w\u2212d)|J(w),\u2207J(w)] =E[g(A(w\u2212d))|g(Aw),\u2207g(Aw)] =\u00b5+C\u0000\u2225Ad\u22252 2\u0001 C(0)(g(Aw)\u2212\u00b5)\u2212C\u2032\u0000\u2225Ad\u22252 2\u0001 C\u2032(0)\u27e8Ad,\u2207g(Aw)\u27e9 =\u00b5+C\u0000\u2225d\u22252 \u03a3 2\u0001 C(0)(J(w)\u2212\u00b5)\u2212C\u2032\u0000\u2225d\u22252 \u03a3 2\u0001 C\u2032(0)\u27e8d, AT\u2207g(Aw)|{z} =\u2207J(w)\u27e9 with\u03a3 := ATA. As in the original proof, we now optimize over the direction first, while keeping the step size constant, although we now fix the step size with regard to the norm \u2225 \u00b7 \u2225 \u03a3(which basically means that we still do the optimization in the isotropic space). Note that max d\u27e8d,\u2207J(x)\u27e9s.t.\u2225d\u2225\u03a3=\u03b7 is equivalent to max d\u27e8d,\u03a3\u22121\u2207J(x)\u27e9\u03a3s.t.\u2225d\u2225\u03a3=\u03b7 which is solved by \u00b1\u03b7\u03a3\u22121\u2207J(x) \u2225\u03a3\u22121\u2207J(x)\u2225\u03a3 The remainder of the proof is exactly the same as in the original. E.2 Conservative RFD In the first paragraph of Section 2 we motivated the relation between RFD and classical optimization with the observation, that gradient descent is the minimizer of a regularized first order Taylor approximation 1 L\u2207J(w) = argmin dT[J(w\u2212d)|J(w),\u2207J(w)] +L 2\u2225w\u22252. This regularized Taylor approximation is in fact an upper bound on our function under the L- smoothness assumption [38], i.e. J(w\u2212d)\u2264T[J(w\u2212d)|J(w),\u2207J(w)] +L 2\u2225d\u22252 An improvement on of this upper bound compared to J(x)therefore guarantees an improvement of the loss. This guarantee was lost with the conditional expectation (on purpose, as we wanted to consider the average case). Losing this guarantee also makes convergence proofs more difficult as they typically make use of this improvement. In view of the confidence intervals of Figure 1, it is natural to ask for a similar upper bound in the random setting, where this can only be the top of an confidence interval. This is provided in the following theorem Lemma E.3 (An\u03b3-upper bound) .We have P\u0010 J(w\u2212d)\u2264E[J(w\u2212d)|J(w),\u2207J(w)] +\u03c1\u03b3(\u2225d\u22252)\u0011 \u2265\u03b3 for\u03c1\u03b3(\u03b72) := \u03a6\u22121(\u03b3)\u03c3(\u03b72)with \u03c32(\u03b72) :=C(0)\u2212C\u0000\u03b72 2\u00012 C(0)\u2212C\u2032\u0000\u03b72 2\u00012 \u2212C\u2032(0)\u03b72 where \u03a6is the cumulative distribution function (cdf) of the standard normal distribution. 36Proof. Note that the conditional variance is with the usual argument about the covariance matrices (cf. the proof of Thoerem 4.2) using Lemma G.2 and an application of Theorem G.1 given by \u03c32(\u2225w\u22252) := Var[ J(w\u2212d)|J(w),\u2207J(w)] =C(0)\u2212C\u0000\u2225d\u22252 2\u00012 C(0)\u2212C\u2032\u0000\u2225d\u22252 2\u00012 \u2212C\u2032(0)\u2225d\u22252. Since the conditional distribution is normal (by Theorem G.1), we have J(w\u2212d)\u2212E[J(w\u2212d)|J(w),\u2207J(w)] \u03c3(\u2225w\u22252)\u223c N(0,1). But this implies the claim P\u0010 J(w\u2212d)\u2264E[J(w\u2212d)|J(w),\u2207J(w)] +\u03c1\u03b3(\u2225d\u22252)\u0011 =P\u0010J(w\u2212d)\u2212E[J(w\u2212d)|J(w),\u2207J(w)] \u03c3(\u2225w\u22252)\u2264\u03a6\u22121(\u03b3)\u0011 = \u03a6(\u03a6\u22121(\u03b3)) =\u03b3. To avoid the Gaussian assumption, one could apply the Markov",
            "references": [
                {
                    "title": "Gradient Span Algorithms Make Predictable Progress in High Dimension",
                    "abstract": "We prove that all 'gradient span algorithms' have asymptotically deterministic behavior on scaled Gaussian random functions as the dimension tends to infinity. In particular, this result explains the counterintuitive phenomenon that different training runs of many large machine learning models result in approximately equal cost curves despite random initialization on a complicated non-convex landscape. The distributional assumption of (non-stationary) isotropic Gaussian random functions we use is sufficiently general to serve as realistic model for machine learning training but also encompass spin glasses and random quadratic functions."
                },
                {
                    "title": "Handbook of Convergence Theorems for (Stochastic) Gradient Methods",
                    "abstract": "This is a handbook of simple proofs of the convergence of gradient and stochastic gradient descent type methods. We consider functions that are Lipschitz, smooth, convex, strongly convex, and/or Polyak-{\\L}ojasiewicz functions. Our focus is on ``good proofs'' that are also simple. Each section can be consulted separately. We start with proofs of gradient descent, then on stochastic variants, including minibatching and momentum. Then move on to nonsmooth problems with the subgradient method, the proximal gradient descent and their stochastic variants. Our focus is on global convergence rates and complexity rates. Some slightly less common proofs found here include that of SGD (Stochastic gradient descent) with a proximal step, with momentum, and with mini-batching without replacement."
                },
                {
                    "title": "Learning-Rate-Free Learning by D-Adaptation",
                    "abstract": "D-Adaptation is an approach to automatically setting the learning rate which asymptotically achieves the optimal rate of convergence for minimizing convex Lipschitz functions, with no back-tracking or line searches, and no additional function value or gradient evaluations per step. Our approach is the first hyper-parameter free method for this class without additional multiplicative log factors in the convergence rate. We present extensive experiments for SGD and Adam variants of our method, where the method automatically matches hand-tuned learning rates across more than a dozen diverse machine learning problems, including large-scale vision and language problems. An open-source implementation is available."
                },
                {
                    "title": "Only Tails Matter: Average-Case Universality and Robustness in the Convex Regime",
                    "abstract": "The recently developed average-case analysis of optimization methods allows a more fine-grained and representative convergence analysis than usual worst-case results. In exchange, this analysis requires a more precise hypothesis over the data generating process, namely assuming knowledge of the expected spectral distribution (ESD) of the random matrix associated with the problem. This work shows that the concentration of eigenvalues near the edges of the ESD determines a problem's asymptotic average complexity. This a priori information on this concentration is a more grounded assumption than complete knowledge of the ESD. This approximate concentration is effectively a middle ground between the coarseness of the worst-case scenario convergence and the restrictive previous average-case analysis. We also introduce the Generalized Chebyshev method, asymptotically optimal under a hypothesis on this concentration and globally optimal when the ESD follows a Beta distribution. We compare its performance to classical optimization algorithms, such as gradient descent or Nesterov's scheme, and we show that, in the average-case context, Nesterov's method is universally nearly optimal asymptotically."
                },
                {
                    "title": "Tight Lipschitz Hardness for optimizing Mean Field Spin Glasses",
                    "abstract": "We study the problem of algorithmically optimizing the Hamiltonian of a spherical or Ising mean field spin glass. The maximum asymptotic value OPT of this random function is characterized by a variational principle known as the Parisi formula, proved first by Talagrand and in more generality by Panchenko. Recently developed approximate message passing algorithms efficiently optimize these functions up to a value ALG given by an extended Parisi formula, which minimizes over a larger space of functional order parameters. These two objectives are equal for spin glasses exhibiting a no overlap gap property. However, ALG can be strictly smaller than OPT, and no efficient algorithm producing a value exceeding ALG is known. We prove that when all interactions have even degree, no algorithm satisfying an overlap concentration property can produce an objective larger than ALG with non-negligible probability. This property holds for all algorithms with suitably Lipschitz dependence on the random disorder coefficients of the objective. It encompasses natural formulations of gradient descent, approximate message passing, and Langevin dynamics run for bounded time and in particular includes the algorithms achieving ALG mentioned above. To prove this result, we substantially generalize the overlap gap property framework introduced by Gamarnik and Sudan to arbitrary ultrametric forbidden structures of solutions."
                },
                {
                    "title": "Scaling Gaussian Processes with Derivative Information Using Variational Inference",
                    "abstract": "Gaussian processes with derivative information are useful in many settings where derivative information is available, including numerous Bayesian optimization and regression tasks that arise in the natural sciences. Incorporating derivative observations, however, comes with a dominating $O(N^3D^3)$ computational cost when training on $N$ points in $D$ input dimensions. This is intractable for even moderately sized problems. While recent work has addressed this intractability in the low-$D$ setting, the high-$N$, high-$D$ setting is still unexplored and of great value, particularly as machine learning problems increasingly become high dimensional. In this paper, we introduce methods to achieve fully scalable Gaussian process regression with derivatives using variational inference. Analogous to the use of inducing values to sparsify the labels of a training set, we introduce the concept of inducing directional derivatives to sparsify the partial derivative information of a training set. This enables us to construct a variational posterior that incorporates derivative information but whose size depends neither on the full dataset size $N$ nor the full dimensionality $D$. We demonstrate the full scalability of our approach on a variety of tasks, ranging from a high dimensional stellarator fusion regression task to training graph convolutional neural networks on Pubmed using Bayesian optimization. Surprisingly, we find that our approach can improve regression performance even in settings where only label data is available."
                },
                {
                    "title": "A Universal Law of Robustness via Isoperimetry",
                    "abstract": "Classically, data interpolation with a parametrized model class is possible as long as the number of parameters is larger than the number of equations to be satisfied. A puzzling phenomenon in deep learning is that models are trained with many more parameters than what this classical theory would suggest. We propose a partial theoretical explanation for this phenomenon. We prove that for a broad class of data distributions and model classes, overparametrization is necessary if one wants to interpolate the data smoothly. Namely we show that smooth interpolation requires d times more parameters than mere interpolation, where d is the ambient data dimension. We prove this universal law of robustness for any smoothly parametrized function class with polynomial size weights, and any covariate distribution verifying isoperimetry (or a mixture thereof). In the case of two-layer neural networks and Gaussian covariates, this law was conjectured in prior work by Bubeck, Li, and Nagaraj. We also give an interpretation of our result as an improved generalization bound for model classes consisting of smooth functions."
                },
                {
                    "title": "High-Dimensional Gaussian Process Inference with Derivatives",
                    "abstract": "Although it is widely known that Gaussian processes can be conditioned on observations of the gradient, this functionality is of limited use due to the prohibitive computational cost of $\\mathcal{O}(N^3 D^3)$ in data points $N$ and dimension $D$. The dilemma of gradient observations is that a single one of them comes at the same cost as $D$ independent function evaluations, so the latter are often preferred. Careful scrutiny reveals, however, that derivative observations give rise to highly structured kernel Gram matrices for very general classes of kernels (inter alia, stationary kernels). We show that in the low-data regime $N<D$, the Gram matrix can be decomposed in a manner that reduces the cost of inference to $\\mathcal{O}(N^2D + (N^2)^3)$ (i.e., linear in the number of dimensions) and, in special cases, to $\\mathcal{O}(N^2D + N^3)$. This reduction in complexity opens up new use-cases for inference with gradients especially in the high-dimensional regime, where the information-to-cost ratio of gradient observations significantly increases. We demonstrate this potential in a variety of tasks relevant for machine learning, such as optimization and Hamiltonian Monte Carlo with predictive gradients."
                },
                {
                    "title": "An Ensemble of Simple Convolutional Neural Network Models for MNIST Digit Recognition",
                    "abstract": "We report that a very high accuracy on the MNIST test set can be achieved by using simple convolutional neural network (CNN) models. We use three different models with 3x3, 5x5, and 7x7 kernel size in the convolution layers. Each model consists of a set of convolution layers followed by a single fully connected layer. Every convolution layer uses batch normalization and ReLU activation, and pooling is not used. Rotation and translation is used to augment training data, which is frequently used in most image classification tasks. A majority voting using the three models independently trained on the training data set can achieve up to 99.87% accuracy on the test set, which is one of the state-of-the-art results. A two-layer ensemble, a heterogeneous ensemble of three homogeneous ensemble networks, can achieve up to 99.91% test accuracy. The results can be reproduced by using the code at: this https URL"
                },
                {
                    "title": "Descending through a Crowded Valley - Benchmarking Deep Learning Optimizers",
                    "abstract": "Choosing the optimizer is among the most crucial decisions of deep learning engineers, and it is not an easy one. The growing literature now lists literally hundreds of optimization methods. In the absence of clear theoretical guidance and conclusive empirical evidence, the decision is often done according to personal anecdotes. In this work, we aim to replace these anecdotes, if not with evidence, then at least with heuristics. To do so, we perform an extensive, standardized benchmark of more than a dozen particularly popular deep learning optimizers while giving a concise overview of the wide range of possible choices. Analyzing almost 35 000 individual runs, we contribute the following three points: Optimizer performance varies greatly across tasks. We observe that evaluating multiple optimizers with default parameters works approximately as well as tuning the hyperparameters of a single, fixed optimizer. While we can not identify an individual optimization method clearly dominating across all tested tasks, we identify a significantly reduced subset of specific algorithms and parameter choices that generally provided competitive results in our experiments. This subset includes popular favorites and some less well-known contenders. We have open-sourced all our experimental results, making it available to use as well-tuned baselines when evaluating novel optimization methods and therefore reducing the necessary computational efforts."
                },
                {
                    "title": "Universality for the Conjugate Gradient and MINRES Algorithms on Sample Covariance Matrices",
                    "abstract": "We present a probabilistic analysis of two Krylov subspace methods for solving linear systems. We prove a central limit theorem for norms of the residual vectors that are produced by the conjugate gradient and MINRES algorithms when applied to a wide class of sample covariance matrices satisfying some standard moment conditions. The proof involves establishing a four\u2010moment theorem for the so\u2010called spectral measure, implying, in particular, universality for the matrix produced by the Lanczos iteration. The central limit theorem then implies an almost\u2010deterministic iteration count for the iterative methods in question. \u00a9 2022 Wiley Periodicals LLC."
                },
                {
                    "title": "Optimal Randomized First-Order Methods for Least-Squares Problems",
                    "abstract": "We provide an exact analysis of a class of randomized algorithms for solving overdetermined least-squares problems. We consider first-order methods, where the gradients are pre-conditioned by an approximation of the Hessian, based on a subspace embedding of the data matrix. This class of algorithms encompasses several randomized methods among the fastest solvers for least-squares problems. We focus on two classical embeddings, namely, Gaussian projections and subsampled randomized Hadamard transforms (SRHT). Our key technical innovation is the derivation of the limiting spectral density of SRHT embeddings. Leveraging this novel result, we derive the family of normalized orthogonal polynomials of the SRHT density and we find the optimal pre-conditioned first-order method along with its rate of convergence. Our analysis of Gaussian embeddings proceeds similarly, and leverages classical random matrix theory results. In particular, we show that for a given sketch size, SRHT embeddings exhibits a faster rate of convergence than Gaussian embeddings. Then, we propose a new algorithm by optimizing the computational complexity over the choice of the sketching dimension. To our knowledge, our resulting algorithm yields the best known complexity for solving least-squares problems with no condition number dependence."
                },
                {
                    "title": "Optimization of mean-field spin glasses",
                    "abstract": "Mean-field spin glasses are families of random energy functions (Hamiltonians) on high-dimensional product spaces. In this paper we consider the case of Ising mixed $p$-spin models, namely Hamiltonians $H_N:\\Sigma_N\\to {\\mathbb R}$ on the Hamming hypercube $\\Sigma_N = \\{\\pm 1\\}^N$, which are defined by the property that $\\{H_N({\\boldsymbol \\sigma})\\}_{{\\boldsymbol \\sigma}\\in \\Sigma_N}$ is a centered Gaussian process with covariance ${\\mathbb E}\\{H_N({\\boldsymbol \\sigma}_1)H_N({\\boldsymbol \\sigma}_2)\\}$ depending only on the scalar product $\\langle {\\boldsymbol \\sigma}_1,{\\boldsymbol \\sigma}_2\\rangle$. The asymptotic value of the optimum $\\max_{{\\boldsymbol \\sigma}\\in \\Sigma_N}H_N({\\boldsymbol \\sigma})$ was characterized in terms of a variational principle known as the Parisi formula, first proved by Talagrand and, in a more general setting, by Panchenko. The structure of superlevel sets is extremely rich and has been studied by a number of authors. Here we ask whether a near optimal configuration ${\\boldsymbol \\sigma}$ can be computed in polynomial time. We develop a message passing algorithm whose complexity per-iteration is of the same order as the complexity of evaluating the gradient of $H_N$, and characterize the typical energy value it achieves. When the $p$-spin model $H_N$ satisfies a certain no-overlap gap assumption, for any $\\varepsilon>0$, the algorithm outputs ${\\boldsymbol \\sigma}\\in\\Sigma_N$ such that $H_N({\\boldsymbol \\sigma})\\ge (1-\\varepsilon)\\max_{{\\boldsymbol \\sigma}'} H_N({\\boldsymbol \\sigma}')$, with high probability. The number of iterations is bounded in $N$ and depends uniquely on $\\varepsilon$. More generally, regardless of whether the no-overlap gap assumption holds, the energy achieved is given by an extended variational principle, which generalizes the Parisi formula."
                },
                {
                    "title": "Which Algorithmic Choices Matter at Which Batch Sizes? Insights From a Noisy Quadratic Model",
                    "abstract": "Increasing the batch size is a popular way to speed up neural network training, but beyond some critical batch size, larger batch sizes yield diminishing returns. In this work, we study how the critical batch size changes based on properties of the optimization algorithm, including acceleration and preconditioning, through two different lenses: large scale experiments, and analysis of a simple noisy quadratic model (NQM). We experimentally demonstrate that optimization algorithms that employ preconditioning, specifically Adam and K-FAC, result in much larger critical batch sizes than stochastic gradient descent with momentum. We also demonstrate that the NQM captures many of the essential features of real neural network training, despite being drastically simpler to work with. The NQM predicts our results with preconditioned optimizers, previous results with accelerated gradient descent, and other results around optimal learning rates and large batch training, making it a useful tool to generate testable predictions about neural network optimization."
                },
                {
                    "title": "The conjugate gradient algorithm on well-conditioned Wishart matrices is almost deteriministic",
                    "abstract": "<p>We prove that the number of iterations required to solve a random positive definite linear system with the conjugate gradient algorithm is almost deterministic for large matrices. We treat the case of Wishart matrices <inline-formula content-type=\"math/mathml\">\n<mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\" alttext=\"upper W equals upper X upper X Superscript asterisk\">\n  <mml:semantics>\n    <mml:mrow>\n      <mml:mi>W</mml:mi>\n      <mml:mo>=</mml:mo>\n      <mml:mi>X</mml:mi>\n      <mml:msup>\n        <mml:mi>X</mml:mi>\n        <mml:mo>\u2217<!-- \u2217 --></mml:mo>\n      </mml:msup>\n    </mml:mrow>\n    <mml:annotation encoding=\"application/x-tex\">W = XX^*</mml:annotation>\n  </mml:semantics>\n</mml:math>\n</inline-formula> where <inline-formula content-type=\"math/mathml\">\n<mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\" alttext=\"upper X\">\n  <mml:semantics>\n    <mml:mi>X</mml:mi>\n    <mml:annotation encoding=\"application/x-tex\">X</mml:annotation>\n  </mml:semantics>\n</mml:math>\n</inline-formula> is <inline-formula content-type=\"math/mathml\">\n<mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\" alttext=\"n times m\">\n  <mml:semantics>\n    <mml:mrow>\n      <mml:mi>n</mml:mi>\n      <mml:mo>\u00d7<!-- \u00d7 --></mml:mo>\n      <mml:mi>m</mml:mi>\n    </mml:mrow>\n    <mml:annotation encoding=\"application/x-tex\">n \\times m</mml:annotation>\n  </mml:semantics>\n</mml:math>\n</inline-formula> and <inline-formula content-type=\"math/mathml\">\n<mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\" alttext=\"n slash m tilde d\">\n  <mml:semantics>\n    <mml:mrow>\n      <mml:mi>n</mml:mi>\n      <mml:mrow class=\"MJX-TeXAtom-ORD\">\n        <mml:mo>/</mml:mo>\n      </mml:mrow>\n      <mml:mi>m</mml:mi>\n      <mml:mo>\u223c<!-- \u223c --></mml:mo>\n      <mml:mi>d</mml:mi>\n    </mml:mrow>\n    <mml:annotation encoding=\"application/x-tex\">n/m \\sim d</mml:annotation>\n  </mml:semantics>\n</mml:math>\n</inline-formula> for <inline-formula content-type=\"math/mathml\">\n<mml:math xmlns:mml=\"http://www.w3.org/1998/Math/MathML\" alttext=\"0 greater-than d greater-than 1\">\n  <mml:semantics>\n    <mml:mrow>\n      <mml:mn>0</mml:mn>\n      <mml:mo>></mml:mo>\n      <mml:mi>d</mml:mi>\n      <mml:mo>></mml:mo>\n      <mml:mn>1</mml:mn>\n    </mml:mrow>\n    <mml:annotation encoding=\"application/x-tex\">0 > d > 1</mml:annotation>\n  </mml:semantics>\n</mml:math>\n</inline-formula>. Precisely, we prove that for most choices of error tolerance, as the matrix increases in size, the probability that the iteration count deviates from an explicit deterministic value tends to zero. In addition, for a fixed iteration count, we show that the norm of the error vector and the norm of the residual converge exponentially fast in probability, converge in mean, and converge almost surely.</p>"
                },
                {
                    "title": "Optimization of the Sherrington-Kirkpatrick Hamiltonian",
                    "abstract": "Let A be a symmetric random matrix with independent and identically distributed Gaussian entries above the diagonal. We consider the problem of maximizing the quadratic form associated to A over binary vectors. In the language of statistical physics, this amounts to finding the ground state of the Sherrington-Kirkpatrick model of spin glasses. The asymptotic value of this optimization problem was characterized by Parisi via a celebrated variational principle, subsequently proved by Talagrand. We give an algorithm that, for any \u03b5 > 0, outputs a feasible solution whose value is at least (1 \u2013 \u03b5) of the optimum, with probability converging to one as the dimension n of the matrix diverges. The algorithm's time complexity is of order n^2. It is a message-passing algorithm, but the specific structure of its update rules is new. As a side result, we prove that, at (low) non-zero temperature, the algorithm constructs approximate solutions of the Thouless-Anderson-Palmer equations."
                },
                {
                    "title": "Following the Ground States of Full\u2010RSB Spherical Spin Glasses",
                    "abstract": "We focus on spherical spin glasses whose Parisi distribution has support of the form [0, q]. For such models we construct paths from the origin to the sphere that consistently remain close to the ground\u2010state energy on the sphere of corresponding radius. The construction uses a greedy strategy, which always follows a direction corresponding to the most negative eigenvalues of the Hessian of the Hamiltonian. For finite mixtures \u03be(x) it provides an algorithm of time complexity O(Ndeg(\u03be)) to find w.h.p. points with the ground\u2010state energy, up to a small error."
                },
                {
                    "title": "Fashionable Modelling with Flux",
                    "abstract": "Machine learning as a discipline has seen an incredible surge of interest in recent years due in large part to a perfect storm of new theory, superior tooling, renewed interest in its capabilities. We present in this paper a framework named Flux that shows how further refinement of the core ideas of machine learning, built upon the foundation of the Julia programming language, can yield an environment that is simple, easily modifiable, and performant. We detail the fundamental principles of Flux as a framework for differentiable programming, give examples of models that are implemented within Flux to display many of the language and framework-level features that contribute to its ease of use and high productivity, display internal compiler techniques used to enable the acceleration and performance that lies at the heart of Flux, and finally give an overview of the larger ecosystem that Flux fits inside of."
                },
                {
                    "title": "Interpolation of Spatial Data",
                    "abstract": "In the course of the study, we have disclosed methods for correcting and analyzing spatial data recorded in a vector format. The latter is best suited for spatial analysis of discrete objects. However, in the case when the spatial variable is represented as a field of scalar or vector magnitudes (for example, spatial concentration distribution of concentrations of heavy metals in soils or the velocity field of groundwater movement). Convenient ways of data recording is a raster format. This approach is most often used for phenomena of processes that are characterized by significant anisotropy. However, the characteristic feature of the inverse distance method is the fact that the interpolated value at the measured point is equal to the measured value."
                },
                {
                    "title": "Backtracking gradient descent method for general $C^1$ functions, with applications to Deep Learning",
                    "abstract": "While Standard gradient descent is one very popular optimisation method, its convergence cannot be proven beyond the class of functions whose gradient is globally Lipschitz continuous. As such, it is not actually applicable to realistic applications such as Deep Neural Networks. In this paper, we prove that its backtracking variant behaves very nicely, in particular convergence can be shown for all Morse functions. The main theoretical result of this paper is as follows. \nTheorem. Let $f:\\mathbb{R}^k\\rightarrow \\mathbb{R}$ be a $C^1$ function, and $\\{z_n\\}$ a sequence constructed from the Backtracking gradient descent algorithm. (1) Either $\\lim _{n\\rightarrow\\infty}||z_n||=\\infty$ or $\\lim _{n\\rightarrow\\infty}||z_{n+1}-z_n||=0$. (2) Assume that $f$ has at most countably many critical points. Then either $\\lim _{n\\rightarrow\\infty}||z_n||=\\infty$ or $\\{z_n\\}$ converges to a critical point of $f$. (3) More generally, assume that all connected components of the set of critical points of $f$ are compact. Then either $\\lim _{n\\rightarrow\\infty}||z_n||=\\infty$ or $\\{z_n\\}$ is bounded. Moreover, in the latter case the set of cluster points of $\\{z_n\\}$ is connected. \nSome generalised versions of this result, including an inexact version, are included. Another result in this paper concerns the problem of saddle points. We then present a heuristic argument to explain why Standard gradient descent method works so well, and modifications of the backtracking versions of GD, MMT and NAG. Experiments with datasets CIFAR10 and CIFAR100 on various popular architectures verify the heuristic argument also for the mini-batch practice and show that our new algorithms, while automatically fine tuning learning rates, perform better than current state-of-the-art methods such as MMT, NAG, Adagrad, Adadelta, RMSProp, Adam and Adamax."
                },
                {
                    "title": "Springer Optimization and Its Applications",
                    "abstract": "In the book, several generalized convex functions and generalized monotone functions are introduced, and various properties and relations are established. These generalized convexities and generalized monotonicities are then applied to the study of optimality conditions and duality theory in optimization. The book consists of three parts. In Part I, the concepts of semistrictly preinvex functions and generalized preinvex functions are introduced and their properties are given. A preinvex function is characterized by an intermediate-point preinvexity condition. Some properties of semistrictly preinvex functions and semipreinvex functions are discussed. In particular, the relationship between a semistrictly preinvex function and a preinvex function is investigated. It is shown that a function is semistrictly preinvex if and only if it satis\ufb01es a strict invexity inequality for any two points with distinct function values. It is worth noting that these characterizations reveal various interesting relationships among prequasiinvex, semistrictly prequasiinvex, and strictly prequasiinvex functions. These relationships are useful in the study of optimization problems."
                },
                {
                    "title": "A disciplined approach to neural network hyper-parameters: Part 1 - learning rate, batch size, momentum, and weight decay",
                    "abstract": "Although deep learning has produced dazzling successes for applications of image, speech, and video processing in the past few years, most trainings are with suboptimal hyper-parameters, requiring unnecessarily long training times. Setting the hyper-parameters remains a black art that requires years of experience to acquire. This report proposes several efficient ways to set the hyper-parameters that significantly reduce training time and improves performance. Specifically, this report shows how to examine the training validation/test loss function for subtle clues of underfitting and overfitting and suggests guidelines for moving toward the optimal balance point. Then it discusses how to increase/decrease the learning rate/momentum to speed up training. Our experiments show that it is crucial to balance every manner of regularization for each dataset and architecture. Weight decay is used as a sample regularizer to show how its optimal value is tightly coupled with the learning rates and momentums. Files to help replicate the results reported here are available."
                },
                {
                    "title": "On the Convergence of Adam and Beyond",
                    "abstract": "Several recently proposed stochastic optimization methods that have been successfully used in training deep networks such as RMSProp, Adam, Adadelta, Nadam are based on using gradient updates scaled by square roots of exponential moving averages of squared past gradients. In many applications, e.g. learning with large output spaces, it has been empirically observed that these algorithms fail to converge to an optimal solution (or a critical point in nonconvex settings). We show that one cause for such failures is the exponential moving average used in the algorithms. We provide an explicit example of a simple convex optimization setting where Adam does not converge to the optimal solution, and describe the precise problems with the previous analysis of Adam algorithm. Our analysis suggests that the convergence issues can be fixed by endowing such algorithms with `long-term memory' of past gradients, and propose new variants of the Adam algorithm which not only fix the convergence issues but often also lead to improved empirical performance."
                },
                {
                    "title": "Deep Neural Networks as Gaussian Processes",
                    "abstract": "It has long been known that a single-layer fully-connected neural network with an i.i.d. prior over its parameters is equivalent to a Gaussian process (GP), in the limit of infinite network width. This correspondence enables exact Bayesian inference for infinite width neural networks on regression tasks by means of evaluating the corresponding GP. Recently, kernel functions which mimic multi-layer random neural networks have been developed, but only outside of a Bayesian framework. As such, previous work has not identified that these kernels can be used as covariance functions for GPs and allow fully Bayesian prediction with a deep neural network. \nIn this work, we derive the exact equivalence between infinitely wide deep networks and GPs. We further develop a computationally efficient pipeline to compute the covariance function for these GPs. We then use the resulting GPs to perform Bayesian inference for wide deep neural networks on MNIST and CIFAR-10. We observe that trained neural network accuracy approaches that of the corresponding GP with increasing layer width, and that the GP uncertainty is strongly correlated with trained network prediction error. We further find that test performance increases as finite-width trained networks are made wider and more similar to a GP, and thus that GP predictions typically outperform those of finite-width networks. Finally we connect the performance of these GPs to the recent theory of signal propagation in random neural networks."
                },
                {
                    "title": "Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms",
                    "abstract": "We present Fashion-MNIST, a new dataset comprising of 28x28 grayscale images of 70,000 fashion products from 10 categories, with 7,000 images per category. The training set has 60,000 images and the test set has 10,000 images. Fashion-MNIST is intended to serve as a direct drop-in replacement for the original MNIST dataset for benchmarking machine learning algorithms, as it shares the same image size, data format and the structure of training and testing splits. The dataset is freely available at this https URL"
                },
                {
                    "title": "Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour",
                    "abstract": "Deep learning thrives with large neural networks and large datasets. However, larger networks and larger datasets result in longer training times that impede research and development progress. Distributed synchronous SGD offers a potential solution to this problem by dividing SGD minibatches over a pool of parallel workers. Yet to make this scheme efficient, the per-worker workload must be large, which implies nontrivial growth in the SGD minibatch size. In this paper, we empirically show that on the ImageNet dataset large minibatches cause optimization difficulties, but when these are addressed the trained networks exhibit good generalization. Specifically, we show no loss of accuracy when training with large minibatch sizes up to 8192 images. To achieve this result, we adopt a hyper-parameter-free linear scaling rule for adjusting learning rates as a function of minibatch size and develop a new warmup scheme that overcomes optimization challenges early in training. With these simple techniques, our Caffe2-based system trains ResNet-50 with a minibatch size of 8192 on 256 GPUs in one hour, while matching small minibatch accuracy. Using commodity hardware, our implementation achieves ~90% scaling efficiency when moving from 8 to 256 GPUs. Our findings enable training visual recognition models on internet-scale data with high efficiency."
                },
                {
                    "title": "The Marginal Value of Adaptive Gradient Methods in Machine Learning",
                    "abstract": "Adaptive optimization methods, which perform local optimization with a metric constructed from the history of iterates, are becoming increasingly popular for training deep neural networks. Examples include AdaGrad, RMSProp, and Adam. We show that for simple overparameterized problems, adaptive methods often find drastically different solutions than gradient descent (GD) or stochastic gradient descent (SGD). We construct an illustrative binary classification problem where the data is linearly separable, GD and SGD achieve zero test error, and AdaGrad, Adam, and RMSProp attain test errors arbitrarily close to half. We additionally study the empirical generalization capability of adaptive methods on several state-of-the-art deep learning models. We observe that the solutions found by adaptive methods generalize worse (often significantly worse) than SGD, even when these solutions have better training performance. These results suggest that practitioners should reconsider the use of adaptive methods to train neural networks."
                },
                {
                    "title": "Bayesian Optimization with Gradients",
                    "abstract": "Bayesian optimization has been successful at global optimization of expensive-to-evaluate multimodal objective functions. However, unlike most optimization methods, Bayesian optimization typically does not use derivative information. In this paper we show how Bayesian optimization can exploit derivative information to decrease the number of objective function evaluations required for good performance. In particular, we develop a novel Bayesian optimization algorithm, the derivative-enabled knowledge-gradient (dKG), for which we show one-step Bayes-optimality, asymptotic consistency, and greater one-step value of information than is possible in the derivative-free setting. Our procedure accommodates noisy and incomplete derivative information, comes in both sequential and batch forms, and can optionally reduce the computational cost of inference through automatically selected retention of a single directional derivative. We also compute the d-KG acquisition function and its gradient using a novel fast discretization-free technique. We show d-KG provides state-of-the-art performance compared to a wide range of optimization procedures with and without gradients, on benchmarks including logistic regression, deep learning, kernel learning, and k-nearest neighbors."
                },
                {
                    "title": "Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks",
                    "abstract": "We present weight normalization: a reparameterization of the weight vectors in a neural network that decouples the length of those weight vectors from their direction. By reparameterizing the weights in this way we improve the conditioning of the optimization problem and we speed up convergence of stochastic gradient descent. Our reparameterization is inspired by batch normalization but does not introduce any dependencies between the examples in a minibatch. This means that our method can also be applied successfully to recurrent models such as LSTMs and to noise-sensitive applications such as deep reinforcement learning or generative models, for which batch normalization is less well suited. Although our method is much simpler, it still provides much of the speed-up of full batch normalization. In addition, the computational overhead of our method is lower, permitting more optimization steps to be taken in the same amount of time. We demonstrate the usefulness of our method on applications in supervised image recognition, generative modelling, and deep reinforcement learning."
                },
                {
                    "title": "Cyclical Learning Rates for Training Neural Networks",
                    "abstract": "It is known that the learning rate is the most important hyper-parameter to tune for training deep neural networks. This paper describes a new method for setting the learning rate, named cyclical learning rates, which practically eliminates the need to experimentally find the best values and schedule for the global learning rates. Instead of monotonically decreasing the learning rate, this method lets the learning rate cyclically vary between reasonable boundary values. Training with cyclical learning rates instead of fixed values achieves improved classification accuracy without a need to tune and often in fewer iterations. This paper also describes a simple way to estimate \"reasonable bounds\" \u2013 linearly increasing the learning rate of the network for a few epochs. In addition, cyclical learning rates are demonstrated on the CIFAR-10 and CIFAR-100 datasets with ResNets, Stochastic Depth networks, and DenseNets, and the ImageNet dataset with the AlexNet and GoogLeNet architectures. These are practical tools for everyone who trains neural networks."
                },
                {
                    "title": "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift",
                    "abstract": "Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.82% top-5 test error, exceeding the accuracy of human raters."
                },
                {
                    "title": "Adam: A Method for Stochastic Optimization",
                    "abstract": "We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
                },
                {
                    "title": "Julia: A Fresh Approach to Numerical Computing",
                    "abstract": "Bridging cultures that have often been distant, Julia combines expertise from the diverse fields of computer science and computational science to create a new approach to numerical computing. Julia is designed to be easy and fast and questions notions generally held to be \u201claws of nature\" by practitioners of numerical computing: \\beginlist \\item High-level dynamic programs have to be slow. \\item One must prototype in one language and then rewrite in another language for speed or deployment. \\item There are parts of a system appropriate for the programmer, and other parts that are best left untouched as they have been built by the experts. \\endlist We introduce the Julia programming language and its design---a dance between specialization and abstraction. Specialization allows for custom treatment. Multiple dispatch, a technique from computer science, picks the right algorithm for the right circumstance. Abstraction, which is what good computation is really about, recognizes what remains the same after dif..."
                },
                {
                    "title": "Analysis and Design of Optimization Algorithms via Integral Quadratic Constraints",
                    "abstract": "This manuscript develops a new framework to analyze and design iterative optimization algorithms built on the notion of Integral Quadratic Constraints (IQC) from robust control theory. IQCs provide sufficient conditions for the stability of complicated interconnected systems, and these conditions can be checked by semidefinite programming. We discuss how to adapt IQC theory to study optimization algorithms, proving new inequalities about convex functions and providing a version of IQC theory adapted for use by optimization researchers. Using these inequalities, we derive numerical upper bounds on convergence rates for the gradient method, the heavy-ball method, Nesterov's accelerated method, and related variants by solving small, simple semidefinite programming problems. We also briefly show how these techniques can be used to search for optimization algorithms with desired performance characteristics, establishing a new methodology for algorithm design."
                },
                {
                    "title": "Identifying and attacking the saddle point problem in high-dimensional non-convex optimization",
                    "abstract": "A central challenge to many fields of science and engineering involves minimizing non-convex error functions over continuous, high dimensional spaces. Gradient descent or quasi-Newton methods are almost ubiquitously used to perform such minimizations, and it is often thought that a main source of difficulty for these local methods to find the global minimum is the proliferation of local minima with much higher error than the global minimum. Here we argue, based on results from statistical physics, random matrix theory, neural network theory, and empirical evidence, that a deeper and more profound difficulty originates from the proliferation of saddle points, not local minima, especially in high dimensional problems of practical interest. Such saddle points are surrounded by high error plateaus that can dramatically slow down learning, and give the illusory impression of the existence of a local minimum. Motivated by these arguments, we propose a new approach to second-order optimization, the saddle-free Newton method, that can rapidly escape high dimensional saddle points, unlike gradient descent and quasi-Newton methods. We apply this algorithm to deep or recurrent neural network training, and provide numerical evidence for its superior optimization performance."
                },
                {
                    "title": "High-Dimensional Random Fields and Random Matrix Theory",
                    "abstract": "Our goal is to discuss in detail the calculation of the mean number of stationary points and minima for random isotropic Gaussian fields on a sphere as well as for stationary Gaussian random fields in a background parabolic confinement. After developing the general formalism based on the high-dimensional Kac-Rice formulae we combine it with the Random Matrix Theory (RMT) techniques to perform analysis of the random energy landscape of $p-$spin spherical spinglasses and a related glass model, both displaying a zero-temperature one-step replica symmetry breaking glass transition as a function of control parameters (e.g. a magnetic field or curvature of the confining potential). A particular emphasis of the presented analysis is on understanding in detail the picture of \"topology trivialization\" (in the sense of drastic reduction of the number of stationary points) of the landscape which takes place in the vicinity of the zero-temperature glass transition in both models. We will reveal the important role of the GOE \"edge scaling\" spectral region and the Tracy-Widom distribution of the maximal eigenvalue of GOE matrices for providing an accurate quantitative description of the universal features of the topology trivialization scenario."
                },
                {
                    "title": "Bayesian Optimization in a Billion Dimensions via Random Embeddings",
                    "abstract": "Bayesian optimization techniques have been successfully applied to robotics, planning, sensor placement, recommendation, advertising, intelligent user interfaces and automatic algorithm configuration. Despite these successes, the approach is restricted to problems of moderate dimension, and several workshops on Bayesian optimization have identified its scaling to high-dimensions as one of the holy grails of the field. In this paper, we introduce a novel random embedding idea to attack this problem. The resulting Random EMbedding Bayesian Optimization (REMBO) algorithm is very simple, has important invariance properties, and applies to domains with both categorical and continuous variables. We present a thorough theoretical analysis of REMBO. Empirical results confirm that REMBO can effectively solve problems with billions of dimensions, provided the intrinsic dimensionality is low. They also show that REMBO achieves state-of-the-art performance in optimizing the 47 discrete parameters of a popular mixed integer linear programming solver."
                },
                {
                    "title": "ADADELTA: An Adaptive Learning Rate Method",
                    "abstract": "We present a novel per-dimension learning rate method for gradient descent called ADADELTA. The method dynamically adapts over time using only first order information and has minimal computational overhead beyond vanilla stochastic gradient descent. The method requires no manual tuning of a learning rate and appears robust to noisy gradient information, different model architecture choices, various data modalities and selection of hyperparameters. We show promising results compared to other methods on the MNIST digit classification task using a single machine and on a large scale voice dataset in a distributed cluster environment."
                },
                {
                    "title": "On the difficulty of training recurrent neural networks",
                    "abstract": "There are two widely known issues with properly training recurrent neural networks, the vanishing and the exploding gradient problems detailed in Bengio et al. (1994). In this paper we attempt to improve the understanding of the underlying issues by exploring these problems from an analytical, a geometric and a dynamical systems perspective. Our analysis is used to justify a simple yet effective solution. We propose a gradient norm clipping strategy to deal with exploding gradients and a soft constraint for the vanishing gradients problem. We validate empirically our hypothesis and proposed solutions in the experimental section."
                },
                {
                    "title": "Adaptive Subgradient Methods for Online Learning and Stochastic Optimization",
                    "abstract": "We present a new family of subgradient methods that dynamically incorporate knowledge of the geometry of the data observed in earlier iterations to perform more informative gradient-based learning. Metaphorically, the adaptation allows us to find needles in haystacks in the form of very predictive but rarely seen features. Our paradigm stems from recent advances in stochastic optimization and online learning which employ proximal functions to control the gradient steps of the algorithm. We describe and analyze an apparatus for adaptively modifying the proximal function, which significantly simplifies setting a learning rate and results in regret guarantees that are provably as good as the best proximal function that can be chosen in hindsight. We give several efficient algorithms for empirical risk minimization problems with common and important regularization functions and domain constraints. We experimentally study our theoretical analysis and show that adaptive subgradient methods outperform state-of-the-art, yet non-adaptive, subgradient algorithms."
                },
                {
                    "title": "Understanding the difficulty of training deep feedforward neural networks",
                    "abstract": "Whereas before 2006 it appears that deep multilayer neural networks were not successfully trained, since then several algorithms have been shown to successfully train them, with experimental results showing the superiority of deeper vs less deep architectures. All these experimental results were obtained with new initialization or training mechanisms. Our objective here is to understand better why standard gradient descent from random initialization is doing so poorly with deep neural networks, to better understand these recent relative successes and help design better algorithms in the future. We first observe the influence of the non-linear activations functions. We find that the logistic sigmoid activation is unsuited for deep networks with random initialization because of its mean value, which can drive especially the top hidden layer into saturation. Surprisingly, we find that saturated units can move out of saturation by themselves, albeit slowly, and explaining the plateaus sometimes seen when training neural networks. We find that a new non-linearity that saturates less can often be beneficial. Finally, we study how activations and gradients vary across layers and during training, with the idea that training may be more difficult when the singular values of the Jacobian associated with each layer are far from 1. Based on these considerations, we propose a new initialization scheme that brings substantially faster convergence. 1 Deep Neural Networks Deep learning methods aim at learning feature hierarchies with features from higher levels of the hierarchy formed by the composition of lower level features. They include Appearing in Proceedings of the 13 International Conference on Artificial Intelligence and Statistics (AISTATS) 2010, Chia Laguna Resort, Sardinia, Italy. Volume 9 of JMLR: WC Weston et al., 2008). Much attention has recently been devoted to them (see (Bengio, 2009) for a review), because of their theoretical appeal, inspiration from biology and human cognition, and because of empirical success in vision (Ranzato et al., 2007; Larochelle et al., 2007; Vincent et al., 2008) and natural language processing (NLP) (Collobert & Weston, 2008; Mnih & Hinton, 2009). Theoretical results reviewed and discussed by Bengio (2009), suggest that in order to learn the kind of complicated functions that can represent high-level abstractions (e.g. in vision, language, and other AI-level tasks), one may need deep architectures. Most of the recent experimental results with deep architecture are obtained with models that can be turned into deep supervised neural networks, but with initialization or training schemes different from the classical feedforward neural networks (Rumelhart et al., 1986). Why are these new algorithms working so much better than the standard random initialization and gradient-based optimization of a supervised training criterion? Part of the answer may be found in recent analyses of the effect of unsupervised pretraining (Erhan et al., 2009), showing that it acts as a regularizer that initializes the parameters in a \u201cbetter\u201d basin of attraction of the optimization procedure, corresponding to an apparent local minimum associated with better generalization. But earlier work (Bengio et al., 2007) had shown that even a purely supervised but greedy layer-wise procedure would give better results. So here instead of focusing on what unsupervised pre-training or semi-supervised criteria bring to deep architectures, we focus on analyzing what may be going wrong with good old (but deep) multilayer neural networks. Our analysis is driven by investigative experiments to monitor activations (watching for saturation of hidden units) and gradients, across layers and across training iterations. We also evaluate the effects on these of choices of activation function (with the idea that it might affect saturation) and initialization procedure (since unsupervised pretraining is a particular form of initialization and it has a drastic impact)."
                },
                {
                    "title": "Kernel Methods for Deep Learning",
                    "abstract": "We introduce a new family of positive-definite kernel functions that mimic the computation in large, multilayer neural nets. These kernel functions can be used in shallow architectures, such as support vector machines (SVMs), or in deep kernel-based architectures that we call multilayer kernel machines (MKMs). We evaluate SVMs and MKMs with these kernel functions on problems designed to illustrate the advantages of deep architectures. On several problems, we obtain better results than previous, leading benchmarks from both SVMs with Gaussian kernels as well as deep belief nets."
                },
                {
                    "title": "Robert:",
                    "abstract": "\u4f8b1 \u60a3\u800519\u5c81,\u56e0\u7ecf\u671f\u53f3\u4e0b\u8179\u75db4\u5e74\u3001\u52a0\u91cd2 d\u4e8e2002\u5e742\u670824\u65e5\u5165\u9662.\u60a3\u800513\u5c81\u6708\u7ecf\u521d\u6f6e,\u6708\u7ecf\u4e0d\u89c4\u5f8b,\u5143\u660e\u663e\u75db\u7ecf.4\u5e74\u524d\u51fa\u73b0\u7ecf\u671f\u53f3\u4e0b\u8179\u5267\u75db,\u5728\u5f53\u5730\u533b\u9662\u8bca\u4e3a\"\u9611\u5c3e\u708e\"\u884c\u9611\u5c3e\u5207\u9664\u672f,\u672f\u540e\u8179\u75db\u672a\u7f13\u89e3.2 d\u524d\u75db\u7ecf\u660e\u663e\u52a0\u91cd,\u9042\u4eba\u9662\u6cbb\u7597.\u8eab\u4f53\u68c0\u67e5:\u4e00\u822c\u60c5\u51b5\u826f\u597d,\u53f3\u4e0b\u8179\u53ef\u89c1\u4e006 cm\u957f\u624b\u672f\u7622\u75d5\u3002"
                },
                {
                    "title": "Extreme value statistics of eigenvalues of Gaussian random matrices.",
                    "abstract": "We compute exact asymptotic results for the probability of the occurrence of large deviations of the largest (smallest) eigenvalue of random matrices belonging to the Gaussian orthogonal, unitary, and symplectic ensembles. In particular, we show that the probability that all the eigenvalues of an (NxN) random matrix are positive (negative) decreases for large N as approximately exp [-beta theta(0)N2] where the Dyson index beta characterizes the ensemble and the exponent theta(0)=(ln 3)/4=0.274653... is universal. We compute the probability that the eigenvalues lie in the interval [zeta1,zeta2] which allows us to calculate the joint probability distribution of the minimum and the maximum eigenvalue. As a by-product, we also obtain exactly the average density of states in Gaussian ensembles whose eigenvalues are restricted to lie in the interval [zeta1,zeta2] , thus generalizing the celebrated Wigner semi-circle law to these restricted ensembles. It is found that the density of states generically exhibits an inverse square-root singularity at the location of the barriers. These results are confirmed by numerical simulations. Some of the results presented in detail here were announced in a previous paper [D. S. Dean and S. N. Majumdar, Phys. Rev. Lett. 97, 160201 (2006)]."
                },
                {
                    "title": "Statistics of critical points of Gaussian fields on large-dimensional spaces.",
                    "abstract": "We calculate the average number of critical points of a Gaussian field on a high-dimensional space as a function of their energy and their index. Our results give a complete picture of the organization of critical points and are of relevance to glassy and disordered systems and landscape scenarios coming from the anthropic approach to string theory."
                },
                {
                    "title": "Gaussian Processes For Machine Learning",
                    "abstract": "Gaussian processes (GPs) are natural generalisations of multivariate Gaussian random variables to infinite (countably or continuous) index sets. GPs have been applied in a large number of fields to a diverse range of ends, and very many deep theoretical analyses of various properties are available. This paper gives an introduction to Gaussian processes on a fairly elementary level with special emphasis on characteristics relevant in machine learning. It draws explicit connections to branches such as spline smoothing models and support vector machines in which similar ideas have been investigated. Gaussian process models are routinely used to solve hard machine learning problems. They are attractive because of their flexible non-parametric nature and computational simplicity. Treated within a Bayesian framework, very powerful statistical methods can be implemented which offer valid estimates of uncertainties in our predictions and generic model selection procedures cast as nonlinear optimization problems. Their main drawback of heavy computational scaling has recently been alleviated by the introduction of generic sparse approximations.13,78,31 The mathematical literature on GPs is large and often uses deep concepts which are not required to fully understand most machine learning applications. In this tutorial paper, we aim to present characteristics of GPs relevant to machine learning and to show up precise connections to other \"kernel machines\" popular in the community. Our focus is on a simple presentation, but references to more detailed sources are provided."
                },
                {
                    "title": "Bayesian Classification With Gaussian Processes",
                    "abstract": "We consider the problem of assigning an input vector to one of m classes by predicting P(c|x) for c=1,...,m. For a two-class problem, the probability of class one given x is estimated by /spl sigma/(y(x)), where /spl sigma/(y)=1/(1+e/sup -y/). A Gaussian process prior is placed on y(x), and is combined with the training data to obtain predictions for new x points. We provide a Bayesian treatment, integrating over uncertainty in y and in the parameters that control the Gaussian process prior the necessary integration over y is carried out using Laplace's approximation. The method is generalized to multiclass problems (m>2) using the softmax function. We demonstrate the effectiveness of the method on a number of datasets."
                },
                {
                    "title": "Computation with Infinite Neural Networks",
                    "abstract": "For neural networks with a wide class of weight priors, it can be shown that in the limit of an infinite number of hidden units, the prior over functions tends to a gaussian process. In this article, analytic forms are derived for the covariance function of the gaussian processes corresponding to networks with sigmoidal and gaussian hidden units. This allows predictions to be made efficiently using networks with an infinite number of hidden units and shows, somewhat paradoxically, that it may be easier to carry out Bayesian prediction with infinite networks rather than finite ones."
                },
                {
                    "title": "Computing with Infinite Networks",
                    "abstract": "For neural networks with a wide class of weight-priors, it can be shown that in the limit of an infinite number of hidden units the prior over functions tends to a Gaussian process. In this paper analytic forms are derived for the covariance function of the Gaussian processes corresponding to networks with sigmoidal and Gaussian hidden units. This allows predictions to be made efficiently using networks with an infinite number of hidden units, and shows that, somewhat paradoxically, it may be easier to compute with infinite networks than finite ones."
                },
                {
                    "title": "Fundamentals of statistical signal processing: estimation theory",
                    "abstract": "Minimum variance unbiased estimation Cramer-Rao lower bound linear models general minimum variance unbiased estimation best linear unbiased estimators maximum likelihood estimation least squares method of moments the Bayesian philosophy general Bayesian estimators linear Bayesian estimators Kalman filters summary of estimators extension for complex data and parameters."
                },
                {
                    "title": "Affine Invariant Convergence Theorems for Newton\u2019s Method and Extensions to Related Methods",
                    "abstract": "For a given starting point, the sequence of Newton iterates is well known to be invariant under affine transformation of the operator equation to be solved. This property, however, is not sufficiently reflected in most convergence theorems that are presently in common use. For this reason, certain affine invariant convergence theorems are given in this paper. The new theorems can be understood as refined versions of the Newton\u2013Mysovskii theorem, the Newton\u2013Kantorovitch theorem (including optimal error bounds) and a convergence theorem for approximate Newton processes. In addition, the concept of affine invariance associated with Newton\u2019s method is extended to S-invariance associated with related iterative methods where S is some set of linear transformations not changing the iterates. As an application of these considerations, a new convergence theorem for a class of generalized Gauss\u2013Newton methods is given. This theorem is S-invariant with an S containing all unitary transformations. Unlike previous com..."
                },
                {
                    "title": "The intrinsic random functions and their applications",
                    "abstract": "The intrinsic random functions (IRF) are a particular case of the Guelfand generalized processes with stationary increments. They constitute a much wider class than the stationary RF, and are used in practical applications for representing non-stationary phenomena. The most important topics are: existence of a generalized covariance (GC) for which statistical inference is possible from a unique realization; theory of the best linear intrinsic estimator (BLIE) used for contouring and estimating problems; the turning bands method for simulating IRF; and the models with polynomial GC, for which statistical inference may be performed by automatic procedures."
                },
                {
                    "title": "Differentiation Under the Integral Sign",
                    "abstract": "(1973). Differentiation Under the Integral Sign. The American Mathematical Monthly: Vol. 80, No. 6, pp. 615-627."
                },
                {
                    "title": "Information Theory and Statistical Mechanics",
                    "abstract": "Treatment of the predictive aspect of statistical mechanics as a form of statistical inference is extended to the density-matrix formalism and applied to a discussion of the relation between irreversibility and information loss. A principle of \"statistical complementarity\" is pointed out, according to which the empirically verifiable probabilities of statistical mechanics necessarily correspond to incomplete predictions. A preliminary discussion is given of the second law of thermodynamics and of a certain class of irreversible processes, in an approximation equivalent to that of the semiclassical theory of radiation."
                },
                {
                    "title": "A parametric model bridging between bounded and unbounded variograms",
                    "abstract": "A simple variogram model with two parameters is presented that includes the power variogram for fractional Brownian motion, a modified De Wijsian model, the generalized Cauchy model and the multiquadric model. One parameter controls the sample path roughness of the process. The other parameter allows for a smooth transition between bounded and unbounded variograms, that is, between stationary and intrinsically stationary processes in a Gaussian framework, or between mixing and non\u2010ergodic Brown\u2013Resnick processes when modeling spatial extremes. Copyright \u00a9 2017 John Wiley & Sons, Ltd."
                },
                {
                    "title": "Taking the Human Out of the Loop: A Review of Bayesian Optimization",
                    "abstract": "Big Data applications are typically associated with systems involving large numbers of users, massive complex software systems, and large-scale heterogeneous computing and storage architectures. The construction of such systems involves many distributed design choices. The end products (e.g., recommendation systems, medical analysis tools, real-time game engines, speech recognizers) thus involve many tunable configuration parameters. These parameters are often specified and hard-coded into the software by various developers or teams. If optimized jointly, these parameters can result in significant improvements. Bayesian optimization is a powerful tool for the joint optimization of design choices that is gaining great popularity in recent years. It promises greater automation so as to increase both product quality and human productivity. This review paper introduces Bayesian optimization, highlights some of its methodological aspects, and showcases a wide range of applications."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Practical bayesian optimization",
                    "abstract": "Global optimization of non-convex functions over real vector spaces is a problem of widespread theoretical and practical interest. In the past fifty years, research in global optimization has produced many important approaches including Lipschitz optimization, simulated annealing, homotopy methods, genetic algorithms, and Bayesian response-surface methods. This work examines the last of these approaches. The Bayesian response-surface approach to global optimization maintains a posterior model of the function being optimized by combining a prior over functions with accumulating function evaluations. The model is then used to compute which point the method should acquire next in its search for the optimum of the function. Bayesian methods can be some of the most efficient approaches to optimization in terms of the number of function evaluations required, but they have significant drawbacks: Current approaches are needlessly data-inefficient, approximations to the Bayes-optimal acquisition criterion are poorly studied, and current approaches do not take advantage of the small-scale properties of differentiable functions near local optima. This work addresses each of these problems to make Bayesian methods more widely applicable."
                },
                {
                    "title": "The mnist database of handwritten digits",
                    "abstract": "Disclosed is an improved articulated bar flail having shearing edges for efficiently shredding materials. An improved shredder cylinder is disclosed with a plurality of these flails circumferentially spaced and pivotally attached to the periphery of a rotatable shaft. Also disclosed is an improved shredder apparatus which has a pair of these shredder cylinders mounted to rotate about spaced parallel axes which cooperates with a conveyer apparatus which has a pair of inclined converging conveyer belts with one of the belts mounted to move with respect to the other belt to allow the transport of articles of various sizes therethrough."
                },
                {
                    "title": "Collecting Spatial Data: Optimum Design of Experiments for Random Fields",
                    "abstract": "Fundamentals of Spatial Statistics.- Fundamentals of Experimental Design.- Exploratory Designs.- Designs for Spatial Trend Estimation.- Design and Dependence.- Multipurpose Designs."
                }
            ],
            "categories": [
                "math.OC",
                "cs.LG",
                "stat.ML"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we improve the efficiency and effectiveness of gradient-based optimization methods in machine learning by better modeling the cost function as a random function?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem could significantly enhance the performance of machine learning algorithms by providing more robust optimization techniques. Improved optimization methods can lead to faster convergence, better generalization, and the ability to handle more complex models. This research could pave the way for future studies that explore advanced optimization strategies, potentially leading to practical applications in various fields such as deep learning, reinforcement learning, and beyond.\n\n### [Question 3] - Why is it hard?\nThe challenges in addressing this problem stem from the inherent complexities of modeling cost functions as random functions, which introduces uncertainties and variabilities that traditional optimization methods do not account for. Naive approaches may fail because they do not consider the stochastic nature of the optimization landscape, leading to suboptimal convergence. Additionally, technical obstacles include the need for accurate estimation of gradients and the handling of high-dimensional spaces, which complicate the optimization process.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has primarily focused on deterministic optimization methods, overlooking the stochastic characteristics of cost functions. Existing solutions often lack the necessary theoretical foundations to model these functions effectively. Barriers such as limited computational resources and the complexity of integrating randomness into optimization frameworks have hindered progress. Our approach differs by explicitly incorporating the stochastic nature of the cost function into the optimization process, providing a more comprehensive framework for analysis and improvement.\n\n### [Question 5] - What are the key components of my approach and results?\nOur proposed methodology involves developing a new optimization algorithm that models the cost function as a random function, utilizing techniques such as stochastic Taylor approximations and regularized upper bounds. We will use benchmark datasets to evaluate the performance of our algorithm against traditional methods, measuring outcomes based on convergence speed and accuracy. The expected results include demonstrating improved optimization efficiency and robustness, as well as providing theoretical insights into the behavior of gradient-based methods in stochastic settings."
            }
        },
        "author_data": {
            "9e8e1263-df84-45f7-8d15-fb1a889f28a2": {
                "pk": "9e8e1263-df84-45f7-8d15-fb1a889f28a2",
                "name": "Felix Benning",
                "collaborators": [
                    "Leif D\u00f6ring"
                ],
                "domain": [
                    "Machine Learning",
                    "Optimization",
                    "Stochastic Methods",
                    "High Dimensional Analysis"
                ],
                "publications": [
                    {
                        "title": "High Dimensional Optimization through the Lens of Machine Learning",
                        "abstract": "This thesis reviews numerical optimization methods with machine learning problems in mind. Since machine learning models are highly parametrized, we focus on methods suited for high dimensional optimization. We build intuition on quadratic models to figure out which methods are suited for non-convex optimization, and develop convergence proofs on convex functions for this selection of methods. With this theoretical foundation for stochastic gradient descent and momentum methods, we try to explain why the methods used commonly in the machine learning field are so successful. Besides explaining successful heuristics, the last chapter also provides a less extensive review of more theoretical methods, which are not quite as popular in practice. So in some sense this work attempts to answer the question: Why are the default Tensorflow optimizers included in the defaults?"
                    },
                    {
                        "title": "Gradient Span Algorithms Make Predictable Progress in High Dimension",
                        "abstract": "We prove that all 'gradient span algorithms' have asymptotically deterministic behavior on scaled Gaussian random functions as the dimension tends to infinity. In particular, this result explains the counterintuitive phenomenon that different training runs of many large machine learning models result in approximately equal cost curves despite random initialization on a complicated non-convex landscape.   The distributional assumption of (non-stationary) isotropic Gaussian random functions we use is sufficiently general to serve as realistic model for machine learning training but also encompass spin glasses and random quadratic functions."
                    }
                ]
            },
            "865ff30b-1025-416f-be22-0d65a13fb70f": {
                "pk": "865ff30b-1025-416f-be22-0d65a13fb70f",
                "name": "Leif D\u00f6ring",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2406.14544": {
        "paper_data": {
            "title": "Prism: A Framework for Decoupling and Assessing the Capabilities of VLMs",
            "url": "http://arxiv.org/abs/2406.14544v1",
            "arxiv_id": "2406.14544",
            "authors": [
                "Yuxuan Qiao",
                "Haodong Duan",
                "Xinyu Fang",
                "Junming Yang",
                "Lin Chen",
                "Songyang Zhang",
                "Jiaqi Wang",
                "Dahua Lin",
                "Kai Chen"
            ],
            "abstract": "Vision Language Models (VLMs) demonstrate remarkable proficiency in addressing a wide array of visual questions, which requires strong perception and reasoning faculties. Assessing these two competencies independently is crucial for model refinement, despite the inherent difficulty due to the intertwined nature of seeing and reasoning in existing VLMs. To tackle this issue, we present Prism, an innovative framework designed to disentangle the perception and reasoning processes involved in visual question solving. Prism comprises two distinct stages: a perception stage that utilizes a VLM to extract and articulate visual information in textual form, and a reasoning stage that formulates responses based on the extracted visual information using a Large Language Model (LLM). This modular design enables the systematic comparison and assessment of both proprietary and open-source VLM for their perception and reasoning strengths. Our analytical framework provides several valuable insights, underscoring Prism's potential as a cost-effective solution for vision-language tasks. By combining a streamlined VLM focused on perception with a powerful LLM tailored for reasoning, Prism achieves superior results in general vision-language tasks while substantially cutting down on training and operational expenses. Quantitative evaluations show that Prism, when configured with a vanilla 2B LLaVA and freely accessible GPT-3.5, delivers performance on par with VLMs $10 \\times$ larger on the rigorous multimodal benchmark MMStar. The project is released at: https://github.com/SparksJoe/Prism.",
            "introduction": "   1 Introduction  With the rapid development of Large Language Models (LLMs) [2022chatgpt, touvron2023llama, touvron2023llama2, 2023internlm, qwen, du2022glm, Claude3], Vision Language Models (VLMs) [OpenAI2023GPT4TR, team2023gemini, bai2023qwen, liu2023visual, dai2023instructblip, internlmxcomposer2] have also experienced significant advancements. As an end-to-end approach, VLMs trained on large-scale multimodal data [liu2023visual, chen2023sharegpt4v, schuhmann2021laion, changpinyo2021conceptual] exhibit superior performance on a variety of tasks. These tasks range from basic ones such as object localization\u00a0[lin2014microsoft] and optical character recognition\u00a0[mishra2019ocr, liu2023hidden] to more complex challenges like document or diagram comprehension\u00a0[kembhavi2016diagram, mathew2021docvqa, masry2022chartqa] and solving geometric problems\u00a0[lu2023mathvista]. For VLMs to solve a general visual question, two essential capabilities are required: 1) Perception: extracting necessary information from the image; 2) Reasoning: generating answers based on the extracted information and contextual understanding. Limitations in either capability can impede the overall performance of a VLM.   A systematic evaluation of the perception and reasoning capabilities is crucial to provide valuable insights for future model optimization. However, seeing and reasoning are mostly entangled in existing VLMs. Proprietary VLMs, such as GPT-4v are opaque systems that only provide the final answer to visual questions. Meanwhile, open-source VLMs\u00a0[liu2023visual] commonly utilize vision encoders to extract visual embeddings, which are often difficult to interpret, from the image and employ adapted LLMs to generate answers based on these visual and linguistic embeddings. In light of this challenge, we introduce Prism, a framework designed to disentangle the perception and reasoning processes of any given VLM. It can serve as a proxy for assessing the real perception capabilities of VLMs.   Prism decomposes the process of solving general visual questions into two distinct stages: 1) a perception stage that concentrates on extracting visual information from the image using VLMs and articulating this information in textual form; and 2) a reasoning stage that utilizes LLMs to answer the question based on the extracted visual information. Prism facilitates the analysis of VLMs\u2019 capabilities through two approaches. To assess the true perception capabilities, one can use a constant LLM in the reasoning stage while testing various VLMs in the perception stage. Conversely, by keeping the VLM fixed and varying the LLM in the reasoning stage, one can determine whether a VLM\u2019s performance is limited by its reasoning capabilities. Through this analysis, we have uncovered several key insights: 1) Proprietary VLMs, such as GPT-4o and GPT-4v, take the lead in the perception capabilities competition; 2) For open-source VLMs, perception capabilities remain relatively consistent regardless of the language model\u2019s size; and 3) The overall performance of open-source VLMs, particularly those with smaller-scale language models like 7B variants, is often constrained by the limited reasoning capabilities.   Figure 1: Prism Framework Architecture. Prism framework takes image-query pairs as input. An instruction (can be query-agnostic or query-aware) and the image are first fed into the VLM to extract visual information. Then, an LLM is used to generate the answer based on the reformatted query which combines the original question and visual information in textual form.    Beyond its role as an evaluation framework, Prism also excels as an efficient general Vision-Language Model (VLM). Building upon findings 2 and 3, we posit that integrating a small-scale VLM as a visual captioner with a powerful LLM as a reasoning engine offers a promising and efficient strategy for general vision-language tasks. By concentrating on visual information extraction, a lightweight VLM can achieve decent performance on par with much larger VLMs. When paired with",
            "references": [
                {
                    "title": "How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suites",
                    "abstract": "In this report, we introduce InternVL 1.5, an open-source multimodal large language model (MLLM) to bridge the capability gap between open-source and proprietary commercial models in multimodal understanding. We introduce three simple improvements: (1) Strong Vision Encoder: we explored a continuous learning strategy for the large-scale vision foundation model -- InternViT-6B, boosting its visual understanding capabilities, and making it can be transferred and reused in different LLMs. (2) Dynamic High-Resolution: we divide images into tiles ranging from 1 to 40 of 448$\\times$448 pixels according to the aspect ratio and resolution of the input images, which supports up to 4K resolution input. (3) High-Quality Bilingual Dataset: we carefully collected a high-quality bilingual dataset that covers common scenes, document images, and annotated them with English and Chinese question-answer pairs, significantly enhancing performance in OCR- and Chinese-related tasks. We evaluate InternVL 1.5 through a series of benchmarks and comparative studies. Compared to both open-source and proprietary models, InternVL 1.5 shows competitive performance, achieving state-of-the-art results in 8 of 18 benchmarks. Code has been released at https://github.com/OpenGVLab/InternVL."
                },
                {
                    "title": "BLINK: Multimodal Large Language Models Can See but Not Perceive",
                    "abstract": "We introduce Blink, a new benchmark for multimodal language models (LLMs) that focuses on core visual perception abilities not found in other evaluations. Most of the Blink tasks can be solved by humans\"within a blink\"(e.g., relative depth estimation, visual correspondence, forensics detection, and multi-view reasoning). However, we find these perception-demanding tasks cast significant challenges for current multimodal LLMs because they resist mediation through natural language. Blink reformats 14 classic computer vision tasks into 3,807 multiple-choice questions, paired with single or multiple images and visual prompting. While humans get 95.70% accuracy on average, Blink is surprisingly challenging for existing multimodal LLMs: even the best-performing GPT-4V and Gemini achieve accuracies of 51.26% and 45.72%, only 13.17% and 7.63% higher than random guessing, indicating that such perception abilities have not\"emerged\"yet in recent multimodal LLMs. Our analysis also highlights that specialist CV models could solve these problems much better, suggesting potential pathways for future improvements. We believe Blink will stimulate the community to help multimodal LLMs catch up with human-level visual perception."
                },
                {
                    "title": "Reka Core, Flash, and Edge: A Series of Powerful Multimodal Language Models",
                    "abstract": "We introduce Reka Core, Flash, and Edge, a series of powerful multimodal language models trained from scratch by Reka. Reka models are able to process and reason with text, images, video, and audio inputs. This technical report discusses details of training some of these models and provides comprehensive evaluation results. We show that Reka Edge and Reka Flash are not only state-of-the-art but also outperform many much larger models, delivering outsized values for their respective compute class. Meanwhile, our most capable and largest model, Reka Core, approaches the best frontier models on both automatic evaluations and blind human evaluations. On image question answering benchmarks (e.g. MMMU, VQAv2), Core performs competitively to GPT4-V. Meanwhile, on multimodal chat, Core ranks as the second most preferred model under a blind third-party human evaluation setup, outperforming other models such as Claude 3 Opus. On text benchmarks, Core not only performs competitively to other frontier models on a set of well-established benchmarks (e.g. MMLU, GSM8K) but also outperforms GPT4-0613 on human evaluation. On video question answering (Perception-Test), Core outperforms Gemini Ultra. Models are shipped in production at http://chat.reka.ai . A showcase of non cherry picked qualitative examples can also be found at http://showcase.reka.ai ."
                },
                {
                    "title": "Are We on the Right Way for Evaluating Large Vision-Language Models?",
                    "abstract": "Large vision-language models (LVLMs) have recently achieved rapid progress, sparking numerous studies to evaluate their multi-modal capabilities. However, we dig into current evaluation works and identify two primary issues: 1) Visual content is unnecessary for many samples. The answers can be directly inferred from the questions and options, or the world knowledge embedded in LLMs. This phenomenon is prevalent across current benchmarks. For instance, GeminiPro achieves 42.9% on the MMMU benchmark without any visual input, and outperforms the random choice baseline across six benchmarks over 24% on average. 2) Unintentional data leakage exists in LLM and LVLM training. LLM and LVLM could still answer some visual-necessary questions without visual content, indicating the memorizing of these samples within large-scale training data. For example, Sphinx-X-MoE gets 43.6% on MMMU without accessing images, surpassing its LLM backbone with 17.9%. Both problems lead to misjudgments of actual multi-modal gains and potentially misguide the study of LVLM. To this end, we present MMStar, an elite vision-indispensable multi-modal benchmark comprising 1,500 samples meticulously selected by humans. MMStar benchmarks 6 core capabilities and 18 detailed axes, aiming to evaluate LVLMs' multi-modal capacities with carefully balanced and purified samples. These samples are first roughly selected from current benchmarks with an automated pipeline, human review is then involved to ensure each curated sample exhibits visual dependency, minimal data leakage, and requires advanced multi-modal capabilities. Moreover, two metrics are developed to measure data leakage and actual performance gain in multi-modal training. We evaluate 16 leading LVLMs on MMStar to assess their multi-modal capabilities, and on 7 benchmarks with the proposed metrics to investigate their data leakage and actual multi-modal gain."
                },
                {
                    "title": "DeepSeek-VL: Towards Real-World Vision-Language Understanding",
                    "abstract": "We present DeepSeek-VL, an open-source Vision-Language (VL) Model designed for real-world vision and language understanding applications. Our approach is structured around three key dimensions: We strive to ensure our data is diverse, scalable, and extensively covers real-world scenarios including web screenshots, PDFs, OCR, charts, and knowledge-based content, aiming for a comprehensive representation of practical contexts. Further, we create a use case taxonomy from real user scenarios and construct an instruction tuning dataset accordingly. The fine-tuning with this dataset substantially improves the model's user experience in practical applications. Considering efficiency and the demands of most real-world scenarios, DeepSeek-VL incorporates a hybrid vision encoder that efficiently processes high-resolution images (1024 x 1024), while maintaining a relatively low computational overhead. This design choice ensures the model's ability to capture critical semantic and detailed information across various visual tasks. We posit that a proficient Vision-Language Model should, foremost, possess strong language abilities. To ensure the preservation of LLM capabilities during pretraining, we investigate an effective VL pretraining strategy by integrating LLM training from the beginning and carefully managing the competitive dynamics observed between vision and language modalities. The DeepSeek-VL family (both 1.3B and 7B models) showcases superior user experiences as a vision-language chatbot in real-world applications, achieving state-of-the-art or competitive performance across a wide range of visual-language benchmarks at the same model size while maintaining robust performance on language-centric benchmarks. We have made both 1.3B and 7B models publicly accessible to foster innovations based on this foundation model."
                },
                {
                    "title": "How Far Are We from Intelligent Visual Deductive Reasoning?",
                    "abstract": "Vision-Language Models (VLMs) have recently demonstrated incredible strides on diverse vision language tasks. We dig into vision-based deductive reasoning, a more sophisticated but less explored realm, and find previously unexposed blindspots in the current SOTA VLMs. Specifically, we leverage Raven's Progressive Matrices (RPMs), to assess VLMs' abilities to perform multi-hop relational and deductive reasoning relying solely on visual clues. We perform comprehensive evaluations of several popular VLMs employing standard strategies such as in-context learning, self-consistency, and Chain-of-thoughts (CoT) on three diverse datasets, including the Mensa IQ test, IntelligenceTest, and RAVEN. The results reveal that despite the impressive capabilities of LLMs in text-based reasoning, we are still far from achieving comparable proficiency in visual deductive reasoning. We found that certain standard strategies that are effective when applied to LLMs do not seamlessly translate to the challenges presented by visual reasoning tasks. A detailed analysis reveals that VLMs struggle to solve these tasks mainly because they are unable to perceive and comprehend multiple, confounding abstract patterns in RPM examples."
                },
                {
                    "title": "InternLM-XComposer2: Mastering Free-form Text-Image Composition and Comprehension in Vision-Language Large Model",
                    "abstract": "We introduce InternLM-XComposer2, a cutting-edge vision-language model excelling in free-form text-image composition and comprehension. This model goes beyond conventional vision-language understanding, adeptly crafting interleaved text-image content from diverse inputs like outlines, detailed textual specifications, and reference images, enabling highly customizable content creation. InternLM-XComposer2 proposes a Partial LoRA (PLoRA) approach that applies additional LoRA parameters exclusively to image tokens to preserve the integrity of pre-trained language knowledge, striking a balance between precise vision understanding and text composition with literary talent. Experimental results demonstrate the superiority of InternLM-XComposer2 based on InternLM2-7B in producing high-quality long-text multi-modal content and its exceptional vision-language understanding performance across various benchmarks, where it not only significantly outperforms existing multimodal models but also matches or even surpasses GPT-4V and Gemini Pro in certain assessments. This highlights its remarkable proficiency in the realm of multimodal understanding. The InternLM-XComposer2 model series with 7B parameters are publicly available at https://github.com/InternLM/InternLM-XComposer."
                },
                {
                    "title": "Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs",
                    "abstract": "Is vision good enough for language? Recent advancements in multimodal models primarily stem from the powerful reasoning abilities of large language models (LLMs). However, the visual component typically depends only on the instance-level contrastive language-image pre-training (CLIP). Our research reveals that the visual capabilities in recent MultiModal LLMs (MLLMs) still exhibit systematic shortcomings. To understand the roots of these errors, we explore the gap between the visual embedding space of CLIP and vision-only self-supervised learning. We identify \u201cCLIP-blind pairs\u201d- images that CLIP perceives as similar despite their clear visual differences. With these pairs, we construct the Multimodal Visual Patterns (MMVP) benchmark. MMVP exposes areas where state-of-the-art systems, including GPT-4V, struggle with straightforward questions across nine basic visual patterns, often providing incorrect answers and hallucinated explanations. We further evaluate various CLIP-based vision-and-language models and found a notable correlation between visual patterns that challenge CLIP models and those problematic for multimodal LLMs. As an initial effort to address these issues, we propose a Mixture of Features (MoF) approach, demonstrating that integrating vision self-supervised learning features with MLLMs can significantly enhance their visual grounding capabilities. Together, our research suggests visual representation learning remains an open challenge, and accurate visual grounding is crucial for future successful multimodal systems."
                },
                {
                    "title": "Intern VL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks",
                    "abstract": "The exponential growth of large language models (LLMs) has opened up numerous possibilities for multi-modal AGI systems. However, the progress in vision and vision-language foundation models, which are also critical elements of multi-modal AGI, has not kept pace with LLMs. In this work, we design a large-scale vision-language foun-dation model (Intern VL), which scales up the vision foun-dation model to 6 billion parameters and progressively aligns it with the LLM, using web-scale image-text data from various sources. This model can be broadly applied to and achieve state-of-the-art performance on 32 generic visual-linguistic benchmarks including visual perception tasks such as image-level or pixel-level recognition, vision-language tasks such as zero-shot image/video classification, zero-shot image/video-text retrieval, and link with LLMs to create multi-modal dialogue systems. It has powerful visual capabilities and can be a good alternative to the ViT-22B. We hope that our research could contribute to the development of multi-modal large models."
                },
                {
                    "title": "Generative Multimodal Models are In-Context Learners",
                    "abstract": "The human ability to easily solve multimodal tasks in context (i.e., with only a few demonstrations or simple instructions), is what current multimodal systems have largely struggled to imitate. In this work, we demonstrate that the task-agnostic in-context learning capabilities of large multimodal models can be significantly enhanced by effective scaling-up. We introduce Emu2, a generative multimodal model with 37 billion parameters, trained on large-scale multimodal sequences with a unified autoregressive objective. Emu2 exhibits strong multimodal in-context learning abilities, even emerging to solve tasks that require on-the-fly reasoning, such as visual prompting and object-grounded generation. The model sets a new record on multiple multimodal understanding tasks in few-shot settings. When instruction-tuned to follow specific instructions, Emu2 further achieves new state-of-the-art on challenging tasks such as question answering benchmarks for large multimodal models and open-ended subject-driven generation. These achievements demonstrate that Emu2 can serve as a base model and general-purpose interface for a wide range of multimodal tasks. Code and models are publicly available to facilitate future research."
                },
                {
                    "title": "MMMU: A Massive Multi-Discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI",
                    "abstract": "We introduce MMMU: a new benchmark designed to evaluate multimodal models on massive multi-discipline tasks demanding college-level subject knowledge and deliberate reasoning. MMMU includes 11.5K meticulously collected multimodal questions from college exams, quizzes, and text-books, covering six core disciplines: Art & Design, Busi-ness, Science, Health & Medicine, Humanities & Social Science, and Tech & Engineering. These questions span 30 subjects and 183 subfields, comprising 30 highly het-erogeneous image types, such as charts, diagrams, maps, tables, music sheets, and chemical structures. Unlike existing benchmarks, MMMU focuses on advanced perception and reasoning with domain-specific knowledge, challenging models to perform tasks akin to those faced by experts. The evaluation of 28 open-source LMMs as well as the propri-etary GPT-4V(ision) and Gemini highlights the substantial challenges posed by MMMU. Even the advanced GPT-4V and Gemini Ultra only achieve accuracies of 56% and 59% respectively, indicating significant room for improvement. We believe MMMU will stimulate the community to build next-generation multimodal foundation models towards expert artificial general intelligence."
                },
                {
                    "title": "ShareGPT4V: Improving Large Multi-Modal Models with Better Captions",
                    "abstract": "In the realm of large multi-modal models (LMMs), efficient modality alignment is crucial yet often constrained by the scarcity of high-quality image-text data. To address this bottleneck, we introduce the ShareGPT4V dataset, a pioneering large-scale resource featuring 1.2 million highly descriptive captions, which surpasses existing datasets in diversity and information content, covering world knowledge, object properties, spatial relationships, and aesthetic evaluations. Specifically, ShareGPT4V originates from a curated 100K high-quality captions collected from advanced GPT4-Vision and has been expanded to 1.2M with a superb caption model trained on this subset. ShareGPT4V first demonstrates its effectiveness for the Supervised Fine-Tuning (SFT) phase, by substituting an equivalent quantity of detailed captions in existing SFT datasets with a subset of our high-quality captions, significantly enhancing the LMMs like LLaVA-7B, LLaVA-1.5-13B, and Qwen-VL-Chat-7B on the MME and MMBench benchmarks, with respective gains of 222.8/22.0/22.3 and 2.7/1.3/1.5. We further incorporate ShareGPT4V data into both the pre-training and SFT phases, obtaining ShareGPT4V-7B, a superior LMM based on a simple architecture that has remarkable performance across a majority of the multi-modal benchmarks. This project is available at https://ShareGPT4V.github.io to serve as a pivotal resource for advancing the LMMs community."
                },
                {
                    "title": "InfiMM-Eval: Complex Open-Ended Reasoning Evaluation For Multi-Modal Large Language Models",
                    "abstract": "Multi-modal Large Language Models (MLLMs) are increasingly prominent in the field of artificial intelligence. These models not only excel in traditional vision-language tasks but also demonstrate impressive performance in contemporary multi-modal benchmarks. Although many of these benchmarks attempt to holistically evaluate MLLMs, they typically concentrate on basic reasoning tasks, often yielding only simple yes/no or multi-choice responses. These methods naturally lead to confusion and difficulties in conclusively determining the reasoning capabilities of MLLMs. To mitigate this issue, we manually curate a benchmark dataset specifically designed for MLLMs, with a focus on complex reasoning tasks. Our benchmark comprises three key reasoning categories: deductive, abductive, and analogical reasoning. The queries in our dataset are intentionally constructed to engage the reasoning capabilities of MLLMs in the process of generating answers. For a fair comparison across various MLLMs, we incorporate intermediate reasoning steps into our evaluation criteria. In instances where an MLLM is unable to produce a definitive answer, its reasoning ability is evaluated by requesting intermediate reasoning steps. If these steps align with our manual annotations, appropriate scores are assigned. This evaluation scheme resembles methods commonly used in human assessments, such as exams or assignments, and represents what we consider a more effective assessment technique compared with existing benchmarks. We evaluate a selection of representative MLLMs using this rigorously developed open-ended multi-step elaborate reasoning benchmark, designed to challenge and accurately measure their reasoning capabilities. The code and data will be released at https://infimm.github.io/InfiMM-Eval/"
                },
                {
                    "title": "Improved Baselines with Visual Instruction Tuning",
                    "abstract": "Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this paper, we present the first systematic study to investigate the design choices of LMMs in a controlled setting under the LLaVA framework. We show that the fully-connected vision-language connector in LLaVA is surprisingly power-ful and data-efficient. With simple modifications to LLa VA, namely, using CLIP- ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~ 1 day on a single 8-AI00 node. Furthermore, we present some early exploration of open problems in LMMs, including scaling to higher resolution inputs, compositional capabilities, and model hallucination, etc. We hope this makes state-of-the-art LMM research more accessible. Code and model will be publicly available."
                },
                {
                    "title": "MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts",
                    "abstract": "Large Language Models (LLMs) and Large Multimodal Models (LMMs) exhibit impressive problem-solving skills in many tasks and domains, but their ability in mathematical reasoning in visual contexts has not been systematically studied. To bridge this gap, we present MathVista, a benchmark designed to combine challenges from diverse mathematical and visual tasks. It consists of 6,141 examples, derived from 28 existing multimodal datasets involving mathematics and 3 newly created datasets (i.e., IQTest, FunctionQA, and PaperQA). Completing these tasks requires fine-grained, deep visual understanding and compositional reasoning, which all state-of-the-art foundation models find challenging. With MathVista, we have conducted a comprehensive, quantitative evaluation of 12 prominent foundation models. The best-performing GPT-4V model achieves an overall accuracy of 49.9%, substantially outperforming Bard, the second-best performer, by 15.1%. Our in-depth analysis reveals that the superiority of GPT-4V is mainly attributed to its enhanced visual perception and mathematical reasoning. However, GPT-4V still falls short of human performance by 10.4%, as it often struggles to understand complex figures and perform rigorous reasoning. This significant gap underscores the critical role that MathVista will play in the development of general-purpose AI agents capable of tackling mathematically intensive and visually rich real-world tasks. We further explore the new ability of self-verification, the application of self-consistency, and the interactive chatbot capabilities of GPT-4V, highlighting its promising potential for future research. The project is available at https://mathvista.github.io/."
                },
                {
                    "title": "Qwen Technical Report",
                    "abstract": "Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural language processing tasks that were previously thought to be exclusive to humans. In this work, we introduce Qwen, the first installment of our large language model series. Qwen is a comprehensive language model series that encompasses distinct models with varying parameter counts. It includes Qwen, the base pretrained language models, and Qwen-Chat, the chat models finetuned with human alignment techniques. The base language models consistently demonstrate superior performance across a multitude of downstream tasks, and the chat models, particularly those trained using Reinforcement Learning from Human Feedback (RLHF), are highly competitive. The chat models possess advanced tool-use and planning capabilities for creating agent applications, showcasing impressive performance even when compared to bigger models on complex tasks like utilizing a code interpreter. Furthermore, we have developed coding-specialized models, Code-Qwen and Code-Qwen-Chat, as well as mathematics-focused models, Math-Qwen-Chat, which are built upon base language models. These models demonstrate significantly improved performance in comparison with open-source models, and slightly fall behind the proprietary models."
                },
                {
                    "title": "SEED-Bench: Benchmarking Multimodal LLMs with Generative Comprehension",
                    "abstract": "Based on powerful Large Language Models (LLMs), recent generative Multimodal Large Language Models (MLLMs) have gained prominence as a pivotal research area, exhibiting remarkable capability for both comprehension and generation. In this work, we address the evaluation of generative comprehension in MLLMs as a preliminary step towards a comprehensive assessment of generative models, by introducing a benchmark named SEED-Bench. SEED-Bench consists of 19K multiple choice questions with accurate human annotations (x 6 larger than existing benchmarks), which spans 12 evaluation dimensions including the comprehension of both the image and video modality. We develop an advanced pipeline for generating multiple-choice questions that target specific evaluation dimensions, integrating both automatic filtering and manual verification processes. Multiple-choice questions with groundtruth options derived from human annotation enables an objective and efficient assessment of model performance, eliminating the need for human or GPT intervention during evaluation. We further evaluate the performance of 18 models across all 12 dimensions, covering both the spatial and temporal understanding. By revealing the limitations of existing MLLMs through evaluation results, we aim for SEED-Bench to provide insights for motivating future research. We will launch and consistently maintain a leaderboard to provide a platform for the community to assess and investigate model capability."
                },
                {
                    "title": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
                    "abstract": "In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs."
                },
                {
                    "title": "MMBench: Is Your Multi-modal Model an All-around Player?",
                    "abstract": "Large vision-language models (VLMs) have recently achieved remarkable progress, exhibiting impressive multimodal perception and reasoning abilities. However, effectively evaluating these large VLMs remains a major challenge, hindering future development in this domain. Traditional benchmarks like VQAv2 or COCO Caption provide quantitative performance measurements but lack fine-grained ability assessment and robust evaluation metrics. Meanwhile, subjective benchmarks, such as OwlEval, offer comprehensive evaluations of a model's abilities by incorporating human labor, which is not scalable and may display significant bias. In response to these challenges, we propose MMBench, a bilingual benchmark for assessing the multi-modal capabilities of VLMs. MMBench methodically develops a comprehensive evaluation pipeline, primarily comprised of the following key features: 1. MMBench is meticulously curated with well-designed quality control schemes, surpassing existing similar benchmarks in terms of the number and variety of evaluation questions and abilities; 2. MMBench introduces a rigorous CircularEval strategy and incorporates large language models to convert free-form predictions into pre-defined choices, which helps to yield accurate evaluation results for models with limited instruction-following capabilities. 3. MMBench incorporates multiple-choice questions in both English and Chinese versions, enabling an apples-to-apples comparison of VLMs' performance under a bilingual context. To summarize, MMBench is a systematically designed objective benchmark for a robust and holistic evaluation of vision-language models. We hope MMBench will assist the research community in better evaluating their models and facilitate future progress in this area. The evalutation code of MMBench has been integrated into VLMEvalKit: https://github.com/open-compass/VLMEvalKit."
                },
                {
                    "title": "Kosmos-2: Grounding Multimodal Large Language Models to the World",
                    "abstract": "We introduce Kosmos-2, a Multimodal Large Language Model (MLLM), enabling new capabilities of perceiving object descriptions (e.g., bounding boxes) and grounding text to the visual world. Specifically, we represent refer expressions as links in Markdown, i.e., ``[text span](bounding boxes)'', where object descriptions are sequences of location tokens. Together with multimodal corpora, we construct large-scale data of grounded image-text pairs (called GrIT) to train the model. In addition to the existing capabilities of MLLMs (e.g., perceiving general modalities, following instructions, and performing in-context learning), Kosmos-2 integrates the grounding capability into downstream applications. We evaluate Kosmos-2 on a wide range of tasks, including (i) multimodal grounding, such as referring expression comprehension, and phrase grounding, (ii) multimodal referring, such as referring expression generation, (iii) perception-language tasks, and (iv) language understanding and generation. This work lays out the foundation for the development of Embodiment AI and sheds light on the big convergence of language, multimodal perception, action, and world modeling, which is a key step toward artificial general intelligence. Code and pretrained models are available at https://aka.ms/kosmos-2."
                },
                {
                    "title": "MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models",
                    "abstract": "Multimodal Large Language Model (MLLM) relies on the powerful LLM to perform multimodal tasks, showing amazing emergent abilities in recent studies, such as writing poems based on an image. However, it is difficult for these case studies to fully reflect the performance of MLLM, lacking a comprehensive evaluation. In this paper, we fill in this blank, presenting the first comprehensive MLLM Evaluation benchmark MME. It measures both perception and cognition abilities on a total of 14 subtasks. In order to avoid data leakage that may arise from direct use of public datasets for evaluation, the annotations of instruction-answer pairs are all manually designed. The concise instruction design allows us to fairly compare MLLMs, instead of struggling in prompt engineering. Besides, with such an instruction, we can also easily carry out quantitative statistics. A total of 30 advanced MLLMs are comprehensively evaluated on our MME, which not only suggests that existing MLLMs still have a large room for improvement, but also reveals the potential directions for the subsequent model optimization. The data application manner and online leaderboards are released at https://github.com/BradyFU/Awesome-Multimodal-Large-Language-Models/tree/Evaluation."
                },
                {
                    "title": "LVLM-eHub: A Comprehensive Evaluation Benchmark for Large Vision-Language Models",
                    "abstract": "Large Vision-Language Models (LVLMs) have recently played a dominant role in multimodal vision-language learning. Despite the great success, it lacks a holistic evaluation of their efficacy. This paper presents a comprehensive evaluation of publicly available large multimodal models by building a LVLM evaluation Hub (LVLM-eHub). Our LVLM-eHub consists of $8$ representative LVLMs such as InstructBLIP and MiniGPT-4, which are thoroughly evaluated by a quantitative capability evaluation and an online arena platform. The former evaluates $6$ categories of multimodal capabilities of LVLMs such as visual question answering and embodied artificial intelligence on $47$ standard text-related visual benchmarks, while the latter provides the user-level evaluation of LVLMs in an open-world question-answering scenario. The study reveals several innovative findings. First, instruction-tuned LVLM with massive in-domain data such as InstructBLIP heavily overfits many existing tasks, generalizing poorly in the open-world scenario. Second, instruction-tuned LVLM with moderate instruction-following data may result in object hallucination issues (i.e., generate objects that are inconsistent with target images in the descriptions). It either makes the current evaluation metric such as CIDEr for image captioning ineffective or generates wrong answers. Third, employing a multi-turn reasoning evaluation framework can mitigate the issue of object hallucination, shedding light on developing an effective pipeline for LVLM evaluation. The findings provide a foundational framework for the conception and assessment of innovative strategies aimed at enhancing zero-shot multimodal techniques. Our LVLM-eHub will be available at https://github.com/OpenGVLab/Multi-Modality-Arena"
                },
                {
                    "title": "Evaluating Object Hallucination in Large Vision-Language Models",
                    "abstract": "Inspired by the superior language abilities of large language models (LLM), large vision-language models (LVLM) have been recently explored by integrating powerful LLMs for improving the performance on complex multimodal tasks. Despite the promising progress on LVLMs, we find that LVLMs suffer from the hallucination problem, i.e. they tend to generate objects that are inconsistent with the target images in the descriptions. To investigate it, this work presents the first systematic study on object hallucination of LVLMs. We conduct the evaluation experiments on several representative LVLMs, and show that they mostly suffer from severe object hallucination issue. We further discuss that the visual instructions may influence the hallucination, and find that: objects that frequently occur in the visual instructions or co-occur with the image objects, are obviously prone to be hallucinated by LVLMs. Besides, we find that existing evaluation methods might be affected by the input instructions and generation styles of LVLMs. Thus, we further design an improved evaluation method for object hallucination by proposing a polling-based query method called POPE. Experiment results demonstrate that our POPE can evaluate the object hallucination in a more stable and flexible way. Our codes and data are publicly available at https://github.com/RUCAIBox/POPE."
                },
                {
                    "title": "On the Hidden Mystery of OCR in Large Multimodal Models",
                    "abstract": "Large models have recently played a dominant role in natural language processing and multimodal vision-language learning. However, their effectiveness in text-related visual tasks remains relatively unexplored. In this paper, we conducted a comprehensive evaluation of Large Multimodal Models, such as GPT4V and Gemini, in various text-related visual tasks including Text Recognition, Scene Text-Centric Visual Question Answering (VQA), Document-Oriented VQA, Key Information Extraction (KIE), and Handwritten Mathematical Expression Recognition (HMER). To facilitate the assessment of Optical Character Recognition (OCR) capabilities in Large Multimodal Models, we propose OCRBench, a comprehensive evaluation benchmark. OCRBench contains 29 datasets, making it the most comprehensive OCR evaluation benchmark available. Furthermore, our study reveals both the strengths and weaknesses of these models, particularly in handling multilingual text, handwritten text, non-semantic text, and mathematical expression recognition. Most importantly, the baseline results presented in this study could provide a foundational framework for the conception and assessment of innovative strategies targeted at enhancing zero-shot multimodal techniques. The evaluation pipeline and benchmark are available at https://github.com/Yuliang-Liu/MultimodalOCR."
                },
                {
                    "title": "InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning",
                    "abstract": "Large-scale pre-training and instruction tuning have been successful at creating general-purpose language models with broad competence. However, building general-purpose vision-language models is challenging due to the rich input distributions and task diversity resulting from the additional visual input. Although vision-language pretraining has been widely studied, vision-language instruction tuning remains under-explored. In this paper, we conduct a systematic and comprehensive study on vision-language instruction tuning based on the pretrained BLIP-2 models. We gather 26 publicly available datasets, covering a wide variety of tasks and capabilities, and transform them into instruction tuning format. Additionally, we introduce an instruction-aware Query Transformer, which extracts informative features tailored to the given instruction. Trained on 13 held-in datasets, InstructBLIP attains state-of-the-art zero-shot performance across all 13 held-out datasets, substantially outperforming BLIP-2 and larger Flamingo models. Our models also lead to state-of-the-art performance when finetuned on individual downstream tasks (e.g., 90.7% accuracy on ScienceQA questions with image contexts). Furthermore, we qualitatively demonstrate the advantages of InstructBLIP over concurrent multimodal models. All InstructBLIP models are open-sourced at https://github.com/salesforce/LAVIS/tree/main/projects/instructblip."
                },
                {
                    "title": "mPLUG-Owl: Modularization Empowers Large Language Models with Multimodality",
                    "abstract": "Large language models (LLMs) have demonstrated impressive zero-shot abilities on a variety of open-ended tasks, while recent research has also explored the use of LLMs for multi-modal generation. In this study, we introduce mPLUG-Owl, a novel training paradigm that equips LLMs with multi-modal abilities through modularized learning of foundation LLM, a visual knowledge module, and a visual abstractor module. This approach can support multiple modalities and facilitate diverse unimodal and multimodal abilities through modality collaboration. The training paradigm of mPLUG-Owl involves a two-stage method for aligning image and text, which learns visual knowledge with the assistance of LLM while maintaining and even improving the generation abilities of LLM. In the first stage, the visual knowledge module and abstractor module are trained with a frozen LLM module to align the image and text. In the second stage, language-only and multi-modal supervised datasets are used to jointly fine-tune a low-rank adaption (LoRA) module on LLM and the abstractor module by freezing the visual knowledge module. We carefully build a visually-related instruction evaluation set OwlEval. Experimental results show that our model outperforms existing multi-modal models, demonstrating mPLUG-Owl's impressive instruction and visual understanding ability, multi-turn conversation ability, and knowledge reasoning ability. Besides, we observe some unexpected and exciting abilities such as multi-image correlation and scene text understanding, which makes it possible to leverage it for harder real scenarios, such as vision-only document comprehension. Our code, pre-trained model, instruction-tuned models, and evaluation set are available at https://github.com/X-PLUG/mPLUG-Owl. The online demo is available at https://www.modelscope.cn/studios/damo/mPLUG-Owl."
                },
                {
                    "title": "Visual Instruction Tuning",
                    "abstract": "Instruction tuning large language models (LLMs) using machine-generated instruction-following data has improved zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. In this paper, we present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and LLM for general-purpose visual and language understanding.Our early experiments show that LLaVA demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model and code base publicly available."
                },
                {
                    "title": "Sigmoid Loss for Language Image Pre-Training",
                    "abstract": "We propose a simple pairwise sigmoid loss for imagetext pre-training. Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization. The sigmoid loss simultaneously allows further scaling up the batch size, while also performing better at smaller batch sizes. With only four TPUv4 chips, we can train a Base CLIP model at 4k batch size and a Large LiT model at 20k batch size, the latter achieves 84.5% ImageNet zero-shot accuracy in two days. This disentanglement of the batch size from the loss further allows us to study the impact of examples vs pairs and negative to positive ratio. Finally, we push the batch size to the extreme, up to one million, and find that the benefits of growing batch size quickly diminish, with a more reasonable batch size of 32k being sufficient. We hope our research motivates further explorations in improving the quality and efficiency of language-image pre-training."
                },
                {
                    "title": "EVA-CLIP: Improved Training Techniques for CLIP at Scale",
                    "abstract": "Contrastive language-image pre-training, CLIP for short, has gained increasing attention for its potential in various scenarios. In this paper, we propose EVA-CLIP, a series of models that significantly improve the efficiency and effectiveness of CLIP training. Our approach incorporates new techniques for representation learning, optimization, and augmentation, enabling EVA-CLIP to achieve superior performance compared to previous CLIP models with the same number of parameters but significantly smaller training costs. Notably, our largest 5.0B-parameter EVA-02-CLIP-E/14+ with only 9 billion seen samples achieves 82.0 zero-shot top-1 accuracy on ImageNet-1K val. A smaller EVA-02-CLIP-L/14+ with only 430 million parameters and 6 billion seen samples achieves 80.4 zero-shot top-1 accuracy on ImageNet-1K val. To facilitate open access and open research, we release the complete suite of EVA-CLIP to the community at https://github.com/baaivision/EVA/tree/master/EVA-CLIP."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "LLaMA: Open and Efficient Foundation Language Models",
                    "abstract": "We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla-70B and PaLM-540B. We release all our models to the research community."
                },
                {
                    "title": "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models",
                    "abstract": "The cost of vision-and-language pre-training has become increasingly prohibitive due to end-to-end training of large-scale models. This paper proposes BLIP-2, a generic and efficient pre-training strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models. BLIP-2 bridges the modality gap with a lightweight Querying Transformer, which is pre-trained in two stages. The first stage bootstraps vision-language representation learning from a frozen image encoder. The second stage bootstraps vision-to-language generative learning from a frozen language model. BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods. For example, our model outperforms Flamingo80B by 8.7% on zero-shot VQAv2 with 54x fewer trainable parameters. We also demonstrate the model's emerging capabilities of zero-shot image-to-text generation that can follow natural language instructions."
                },
                {
                    "title": "Visual Spatial Reasoning",
                    "abstract": "Spatial relations are a basic part of human cognition. However, they are expressed in natural language in a variety of ways, and previous work has suggested that current vision-and-language models (VLMs) struggle to capture relational information. In this paper, we present Visual Spatial Reasoning (VSR), a dataset containing more than 10k natural text-image pairs with 66 types of spatial relations in English (e.g., under, in front of, facing). While using a seemingly simple annotation format, we show how the dataset includes challenging linguistic phenomena, such as varying reference frames. We demonstrate a large gap between human and model performance: The human ceiling is above 95%, while state-of-the-art models only achieve around 70%. We observe that VLMs\u2019 by-relation performances have little correlation with the number of training examples and the tested models are in general incapable of recognising relations concerning the orientations of objects.1"
                },
                {
                    "title": "Flamingo: a Visual Language Model for Few-Shot Learning",
                    "abstract": "Building models that can be rapidly adapted to novel tasks using only a handful of annotated examples is an open challenge for multimodal machine learning research. We introduce Flamingo, a family of Visual Language Models (VLM) with this ability. We propose key architectural innovations to: (i) bridge powerful pretrained vision-only and language-only models, (ii) handle sequences of arbitrarily interleaved visual and textual data, and (iii) seamlessly ingest images or videos as inputs. Thanks to their flexibility, Flamingo models can be trained on large-scale multimodal web corpora containing arbitrarily interleaved text and images, which is key to endow them with in-context few-shot learning capabilities. We perform a thorough evaluation of our models, exploring and measuring their ability to rapidly adapt to a variety of image and video tasks. These include open-ended tasks such as visual question-answering, where the model is prompted with a question which it has to answer; captioning tasks, which evaluate the ability to describe a scene or an event; and close-ended tasks such as multiple-choice visual question-answering. For tasks lying anywhere on this spectrum, a single Flamingo model can achieve a new state of the art with few-shot learning, simply by prompting the model with task-specific examples. On numerous benchmarks, Flamingo outperforms models fine-tuned on thousands of times more task-specific data."
                },
                {
                    "title": "ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning",
                    "abstract": "Charts are very popular for analyzing data. When exploring charts, people often ask a variety of complex reasoning questions that involve several logical and arithmetic operations. They also commonly refer to visual features of a chart in their questions. However, most existing datasets do not focus on such complex reasoning questions as their questions are template-based and answers come from a fixed-vocabulary. In this work, we present a large-scale benchmark covering 9.6K human-written questions as well as 23.1K questions generated from human-written chart summaries. To address the unique challenges in our benchmark involving visual and logical reasoning over charts, we present two transformer-based models that combine visual features and the data table of the chart in a unified way to answer questions. While our models achieve the state-of-the-art results on the previous datasets as well as on our benchmark, the evaluation also reveals several challenges in answering complex reasoning questions."
                },
                {
                    "title": "Chain of Thought Prompting Elicits Reasoning in Large Language Models",
                    "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier."
                },
                {
                    "title": "LAION-400M: Open Dataset of CLIP-Filtered 400 Million Image-Text Pairs",
                    "abstract": "Multi-modal language-vision models trained on hundreds of millions of image-text pairs (e.g. CLIP, DALL-E) gained a recent surge, showing remarkable capability to perform zero- or few-shot learning and transfer even in absence of per-sample labels on target image data. Despite this trend, to date there has been no publicly available datasets of sufficient scale for training such models from scratch. To address this issue, in a community effort we build and release for public LAION-400M, a dataset with CLIP-filtered 400 million image-text pairs, their CLIP embeddings and kNN indices that allow efficient similarity search."
                },
                {
                    "title": "GLM: General Language Model Pretraining with Autoregressive Blank Infilling",
                    "abstract": "There have been various types of pretraining architectures including autoencoding models (e.g., BERT), autoregressive models (e.g., GPT), and encoder-decoder models (e.g., T5). However, none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding (NLU), unconditional generation, and conditional generation. We propose a General Language Model (GLM) based on autoregressive blank infilling to address this challenge. GLM improves blank filling pretraining by adding 2D positional encodings and allowing an arbitrary order to predict spans, which results in performance gains over BERT and T5 on NLU tasks. Meanwhile, GLM can be pretrained for different types of tasks by varying the number and lengths of blanks. On a wide range of tasks across NLU, conditional and unconditional generation, GLM outperforms BERT, T5, and GPT given the same model sizes and data, and achieves the best performance from a single pretrained model with 1.25\u00d7 parameters of BERT Large , demonstrating its generalizability to different downstream tasks."
                },
                {
                    "title": "Learning Transferable Visual Models From Natural Language Supervision",
                    "abstract": "State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. We release our code and pre-trained model weights at https://github.com/OpenAI/CLIP."
                },
                {
                    "title": "Conceptual 12M: Pushing Web-Scale Image-Text Pre-Training To Recognize Long-Tail Visual Concepts",
                    "abstract": "The availability of large-scale image captioning and visual question answering datasets has contributed significantly to recent successes in vision-and-language pretraining. However, these datasets are often collected with overrestrictive requirements inherited from their original target tasks (e.g., image caption generation), which limit the resulting dataset scale and diversity. We take a step further in pushing the limits of vision-and-language pretraining data by relaxing the data collection pipeline used in Conceptual Captions 3M (CC3M) [54] and introduce the Conceptual 12M (CC12M), a dataset with 12 million image-text pairs specifically meant to be used for visionand-language pre-training. We perform an analysis of this dataset and benchmark its effectiveness against CC3M on multiple downstream tasks with an emphasis on long-tail visual recognition. Our results clearly illustrate the benefit of scaling up pre-training data for vision-and-language tasks, as indicated by the new state-of-the-art results on both the nocaps and Conceptual Captions benchmarks.1"
                },
                {
                    "title": "DocVQA: A Dataset for VQA on Document Images",
                    "abstract": "We present a new dataset for Visual Question Answering (VQA) on document images called DocVQA. The dataset consists of 50,000 questions defined on 12,000+ document images. Detailed analysis of the dataset in comparison with similar datasets for VQA and reading comprehension is presented. We report several baseline results by adopting existing VQA and reading comprehension models. Although the existing models perform reasonably well on certain types of questions, there is large performance gap compared to human performance (94.36% accuracy). The models need to improve specifically on questions where understanding structure of the document is crucial. The dataset, code and leaderboard are available at docvqa.org"
                },
                {
                    "title": "OCR-VQA: Visual Question Answering by Reading Text in Images",
                    "abstract": "The problem of answering questions about an image is popularly known as visual question answering (or VQA in short). It is a well-established problem in computer vision. However, none of the VQA methods currently utilize the text often present in the image. These \"texts in images\" provide additional useful cues and facilitate better understanding of the visual content. In this paper, we introduce a novel task of visual question answering by reading text in images, i.e., by optical character recognition or OCR. We refer to this problem as OCR-VQA. To facilitate a systematic way of studying this new problem, we introduce a large-scale dataset, namely OCRVQA-200K. This dataset comprises of 207,572 images of book covers and contains more than 1 million question-answer pairs about these images. We judiciously combine well-established techniques from OCR and VQA domains to present a novel baseline for OCR-VQA-200K. The experimental results and rigorous analysis demonstrate various challenges present in this dataset leaving ample scope for the future research. We are optimistic that this new task along with compiled dataset will open-up many exciting research avenues both for the document image analysis and the VQA communities."
                },
                {
                    "title": "OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge",
                    "abstract": "Visual Question Answering (VQA) in its ideal form lets us study reasoning in the joint space of vision and language and serves as a proxy for the AI task of scene understanding. However, most VQA benchmarks to date are focused on questions such as simple counting, visual attributes, and object detection that do not require reasoning or knowledge beyond what is in the image. In this paper, we address the task of knowledge-based visual question answering and provide a benchmark, called OK-VQA, where the image content is not sufficient to answer the questions, encouraging methods that rely on external knowledge resources. Our new dataset includes more than 14,000 questions that require external knowledge to answer. We show that the performance of the state-of-the-art VQA models degrades drastically in this new setting. Our analysis shows that our knowledge-based VQA task is diverse, difficult, and large compared to previous knowledge-based VQA datasets. We hope that this dataset enables researchers to open up new avenues for research in this domain."
                },
                {
                    "title": "Towards VQA Models That Can Read",
                    "abstract": "Studies have shown that a dominant class of questions asked by visually impaired users on images of their surroundings involves reading text in the image. But today\u2019s VQA models can not read! Our paper takes a first step towards addressing this problem. First, we introduce a new \u201cTextVQA\u201d dataset to facilitate progress on this important problem. Existing datasets either have a small proportion of questions about text (e.g., the VQA dataset) or are too small (e.g., the VizWiz dataset). TextVQA contains 45,336 questions on 28,408 images that require reasoning about text to answer. Second, we introduce a novel model architecture that reads text in the image, reasons about it in the context of the image and the question, and predicts an answer which might be a deduction based on the text and the image or composed of the strings found in the image. Consequently, we call our approach Look, Read, Reason & Answer (LoRRA). We show that LoRRA outperforms existing state-of-the-art VQA models on our TextVQA dataset. We find that the gap between human performance and machine performance is significantly larger on TextVQA than on VQA 2.0, suggesting that TextVQA is well-suited to benchmark progress along directions complementary to VQA 2.0."
                },
                {
                    "title": "GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering",
                    "abstract": "We introduce GQA, a new dataset for real-world visual reasoning and compositional question answering, seeking to address key shortcomings of previous VQA datasets. We have developed a strong and robust question engine that leverages Visual Genome scene graph structures to create 22M diverse reasoning questions, which all come with functional programs that represent their semantics. We use the programs to gain tight control over the answer distribution and present a new tunable smoothing technique to mitigate question biases. Accompanying the dataset is a suite of new metrics that evaluate essential qualities such as consistency, grounding and plausibility. A careful analysis is performed for baselines as well as state-of-the-art models, providing fine-grained results for different question types and topologies. Whereas a blind LSTM obtains a mere 42.1%, and strong VQA models achieve 54.1%, human performance tops at 89.3%, offering ample opportunity for new research to explore. We hope GQA will provide an enabling resource for the next generation of models with enhanced robustness, improved consistency, and deeper semantic understanding of vision and language."
                },
                {
                    "title": "Qwen-VL: A Frontier Large Vision-Language Model with Versatile Abilities",
                    "abstract": "We introduce the Qwen-VL series, a set of large-scale vision-language models designed to perceive and understand both text and images. Comprising Qwen-VL and Qwen-VL-Chat, these models exhibit remarkable performance in tasks like image captioning, question answering, visual localization, and flexible interaction. The evaluation covers a wide range of tasks including zero-shot captioning, visual or document visual question answering, and grounding. We demonstrate the Qwen-VL outperforms existing Large Vision Language Models (LVLMs). We present their architecture, training, capabilities, and performance, highlighting their contributions to advancing multimodal artificial intelligence. Code, demo and models are available at https://github.com/QwenLM/Qwen-VL ."
                },
                {
                    "title": "HallusionBench: You See What You Think? Or You Think What You See? An Image-Context Reasoning Benchmark Challenging for GPT-4V(ision), LLaVA-1.5, and Other Multi-modality Models",
                    "abstract": "Large language models (LLMs), after being aligned with vision models and integrated into vision-language models (VLMs), can bring impressive improvement in image reasoning tasks. This was shown by the recently released GPT-4V(ison), LLaVA-1.5, etc. However, the strong language prior in these SOTA LVLMs can be a double-edged sword: they may ignore the image context and solely rely on the (even contradictory) language prior for reasoning. In contrast, the vision modules in VLMs are weaker than LLMs and may result in misleading visual representations, which are then translated to confident mistakes by LLMs. To study these two types of VLM mistakes, i.e., language hallucination and visual illusion , we curated \u201cH ALLUSION B ENCH 1 ,\u201d an image-context reasoning benchmark that is still challenging to even GPT-4V and LLaVA-1.5. We provide a detailed analysis of examples in H ALLUSION B ENCH , which sheds novel insights on the illusion or hallucination of VLMs and how to improve them in the future. The benchmark and codebase will be released at https://github.com/tianyi-lab/HallusionBench."
                },
                {
                    "title": "nocaps: novel object captioning at scale",
                    "abstract": "Image captioning models have achieved impressive results on datasets containing limited visual concepts and large amounts of paired image-caption training data. However, if these models are to ever function in the wild, a much larger variety of visual concepts must be learned, ideally from less supervision. To encourage the development of image captioning models that can learn visual concepts from alternative data sources, such as object detection datasets, we present the first large-scale benchmark for this task. Dubbed \u2018nocaps\u2019, for novel object captioning at scale, our benchmark consists of 166,100 human-generated captions describing 15,100 images from the Open Images validation and test sets. The associated training data consists of COCO image-caption pairs, plus Open Images image-level labels and object bounding boxes. Since Open Images contains many more classes than COCO, nearly 400 object classes seen in test images have no or very few associated training captions (hence, nocaps). We extend existing novel object captioning models to establish strong baselines for this benchmark and provide analysis to guide future work."
                },
                {
                    "title": "The Claude 3 Model Family: Opus, Sonnet, Haiku",
                    "abstract": "We introduce Claude 3, a new family of large multimodal models \u2013 Claude 3 Opus , our most capable offering, Claude 3 Sonnet , which provides a combination of skills and speed, and Claude 3 Haiku , our fastest and least expensive model. All new models have vision capabilities that enable them to process and analyze image data. The Claude 3 family demonstrates strong performance across benchmark evaluations and sets a new standard on measures of reasoning, math, and coding. Claude 3 Opus achieves state-of-the-art results on evaluations like GPQA [1], MMLU [2], MMMU [3] and many more. Claude 3 Haiku performs as well or better than Claude 2 [4] on most pure-text tasks, while Sonnet and Opus significantly outperform it. Additionally, these models exhibit improved fluency in non-English languages, making them more versatile for a global audience. In this report, we provide an in-depth analysis of our evaluations, focusing on core capabilities, safety, societal impacts, and the catastrophic risk assessments we committed to in our Responsible Scaling Policy [5]."
                }
            ],
            "categories": [
                "cs.CV",
                "cs.CL"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "### [Question 1] - What is the problem?\nHow can we effectively disentangle the perception and reasoning processes in Vision Language Models (VLMs) to assess their individual capabilities?\n\n### [Question 2] - Why is it interesting and important?\nSolving this problem is crucial for the research community as it provides a systematic evaluation framework that can lead to a deeper understanding of VLMs' strengths and weaknesses. By disentangling perception and reasoning, researchers can optimize models more effectively, leading to advancements in multimodal AI applications. This could enhance the development of more capable VLMs, improve their interpretability, and foster innovations in areas such as computer vision, natural language processing, and human-computer interaction.\n\n### [Question 3] - Why is it hard?\nThe challenge lies in the entangled nature of perception and reasoning in existing VLMs, making it difficult to isolate and evaluate each capability independently. Naive approaches may fail because they do not account for the complex interactions between visual information extraction and reasoning processes. Technical obstacles include the need for interpretable visual embeddings and the difficulty in designing experiments that can accurately measure the contributions of each component without interference.\n\n### [Question 4] - Why hasn't it been solved before?\nPrevious research has often treated perception and reasoning as a single, integrated process, leading to a lack of clarity in evaluating their individual contributions. Existing solutions are limited by their proprietary nature or the opacity of their architectures, which prevents a thorough understanding of their inner workings. Our approach differs by introducing the Prism framework, which explicitly separates these processes, allowing for targeted assessments and insights that were previously unattainable.\n\n### [Question 5] - What are the key components of my approach and results?\nThe proposed methodology involves the Prism framework, which consists of two stages: a perception stage using various VLMs to extract visual information and a reasoning stage utilizing a constant LLM to generate answers. The evaluation will be conducted using a diverse dataset of image-query pairs, with performance metrics focusing on accuracy and interpretability of the extracted visual information. Expected outcomes include a clearer understanding of the perception and reasoning capabilities of different VLMs, as well as the identification of optimal configurations for improved performance in general vision-language tasks."
            }
        },
        "author_data": {
            "5630869e-94bd-4e42-9c84-67f10a757451": {
                "pk": "5630869e-94bd-4e42-9c84-67f10a757451",
                "name": "Yuxuan Qiao",
                "collaborators": [
                    "Yuntao Zhu",
                    "Xiaoping Yang"
                ],
                "domain": [
                    "Medical Imaging",
                    "Image Segmentation",
                    "Deep Learning",
                    "Vascular Analysis"
                ],
                "publications": [
                    {
                        "title": "Morphology Edge Attention Network and Optimal Geometric Matching Connection model for vascular segmentation",
                        "abstract": "There are many unsolved problems in vascular image segmentation, including vascular structural connectivity, scarce branches and missing small vessels. Obtaining vessels that preserve their correct topological structures is currently a crucial research issue, as it provides an overall view of one vascular system. In order to preserve the topology and accuracy of vessel segmentation, we proposed a novel Morphology Edge Attention Network (MEA-Net) for the segmentation of vessel-like structures, and an Optimal Geometric Matching Connection (OGMC) model to connect the broken vessel segments. The MEA-Net has an edge attention module that improves the segmentation of edges and small objects by morphology operation extracting boundary voxels on multi-scale. The OGMC model uses the concept of curve touching from differential geometry to filter out fragmented vessel endpoints, and then employs minimal surfaces to determine the optimal connection order between blood vessels. Finally, we calculate the geodesic to repair missing vessels under a given Riemannian metric. Our method achieves superior or competitive results compared to state-of-the-art methods on four datasets of 3D vascular segmentation tasks, both effectively reducing vessel broken and increasing vessel branch richness, yielding blood vessels with a more precise topological structure."
                    }
                ]
            },
            "b3025726-b7a0-499d-b947-6625c1028f08": {
                "pk": "b3025726-b7a0-499d-b947-6625c1028f08",
                "name": "Haodong Duan",
                "collaborators": [
                    "Dahua Lin",
                    "Kai Chen",
                    "Jiaqi Wang",
                    "Yue Zhao",
                    "Yuanjun Xiong",
                    "Wentao Liu",
                    "Songyang Zhang",
                    "Xinyu Fang",
                    "Nanxuan Zhao",
                    "Bingzhe Wu"
                ],
                "domain": [
                    "Skeleton-based Action Recognition",
                    "Graph Neural Network",
                    "Self-Supervised Learning",
                    "Video Understanding"
                ],
                "publications": [
                    {
                        "title": "PYSKL: Towards Good Practices for Skeleton Action Recognition",
                        "abstract": "We present PYSKL: an open-source toolbox for skeleton-based action recognition based on PyTorch. The toolbox supports a wide variety of skeleton action recognition algorithms, including approaches based on GCN and CNN. In contrast to existing open-source skeleton action recognition projects that include only one or two algorithms, PYSKL implements six different algorithms under a unified framework with both the latest and original good practices to ease the comparison of efficacy and efficiency. We also provide an original GCN-based skeleton action recognition model named ST-GCN++, which achieves competitive recognition performance without any complicated attention schemes, serving as a strong baseline. Meanwhile, PYSKL supports the training and testing of nine skeleton-based action recognition benchmarks and achieves state-of-the-art recognition performance on eight of them. To facilitate future research on skeleton action recognition, we also provide a large number of trained models and detailed benchmark results to give some insights. PYSKL is released at https://github.com/kennymckormick/pyskl and is actively maintained. We will update this report when we add new features or benchmarks. The current version corresponds to PYSKL v0.2."
                    },
                    {
                        "title": "DG-STGCN: Dynamic Spatial-Temporal Modeling for Skeleton-based Action Recognition",
                        "abstract": "Graph convolution networks (GCN) have been widely used in skeleton-based action recognition. We note that existing GCN-based approaches primarily rely on prescribed graphical structures (ie., a manually defined topology of skeleton joints), which limits their flexibility to capture complicated correlations between joints. To move beyond this limitation, we propose a new framework for skeleton-based action recognition, namely Dynamic Group Spatio-Temporal GCN (DG-STGCN). It consists of two modules, DG-GCN and DG-TCN, respectively, for spatial and temporal modeling. In particular, DG-GCN uses learned affinity matrices to capture dynamic graphical structures instead of relying on a prescribed one, while DG-TCN performs group-wise temporal convolutions with varying receptive fields and incorporates a dynamic joint-skeleton fusion module for adaptive multi-level temporal modeling. On a wide range of benchmarks, including NTURGB+D, Kinetics-Skeleton, BABEL, and Toyota SmartHome, DG-STGCN consistently outperforms state-of-the-art methods, often by a notable margin."
                    },
                    {
                        "title": "TransRank: Self-supervised Video Representation Learning via Ranking-based Transformation Recognition",
                        "abstract": "Recognizing transformation types applied to a video clip (RecogTrans) is a long-established paradigm for self-supervised video representation learning, which achieves much inferior performance compared to instance discrimination approaches (InstDisc) in recent works. However, based on a thorough comparison of representative RecogTrans and InstDisc methods, we observe the great potential of RecogTrans on both semantic-related and temporal-related downstream tasks. Based on hard-label classification, existing RecogTrans approaches suffer from noisy supervision signals in pre-training. To mitigate this problem, we developed TransRank, a unified framework for recognizing Transformations in a Ranking formulation. TransRank provides accurate supervision signals by recognizing transformations relatively, consistently outperforming the classification-based formulation. Meanwhile, the unified framework can be instantiated with an arbitrary set of temporal or spatial transformations, demonstrating good generality. With a ranking-based formulation and several empirical practices, we achieve competitive performance on video retrieval and action recognition. Under the same setting, TransRank surpasses the previous state-of-the-art method by 6.4% on UCF101 and 8.3% on HMDB51 for action recognition (Top1 Acc); improves video retrieval on UCF101 by 20.4% (R@1). The promising results validate that RecogTrans is still a worth exploring paradigm for video self-supervised learning. Codes will be released at https://github.com/kennymckormick/TransRank."
                    },
                    {
                        "title": "SRPGAN: Perceptual Generative Adversarial Network for Single Image Super Resolution",
                        "abstract": "Single image super resolution (SISR) is to reconstruct a high resolution image from a single low resolution image. The SISR task has been a very attractive research topic over the last two decades. In recent years, convolutional neural network (CNN) based models have achieved great performance on SISR task. Despite the breakthroughs achieved by using CNN models, there are still some problems remaining unsolved, such as how to recover high frequency details of high resolution images. Previous CNN based models always use a pixel wise loss, such as l2 loss. Although the high resolution images constructed by these models have high peak signal-to-noise ratio (PSNR), they often tend to be blurry and lack high-frequency details, especially at a large scaling factor. In this paper, we build a super resolution perceptual generative adversarial network (SRPGAN) framework for SISR tasks. In the framework, we propose a robust perceptual loss based on the discriminator of the built SRPGAN model. We use the Charbonnier loss function to build the content loss and combine it with the proposed perceptual loss and the adversarial loss. Compared with other state-of-the-art methods, our method has demonstrated great ability to construct images with sharp edges and rich details. We also evaluate our method on different benchmarks and compare it with previous CNN based methods. The results show that our method can achieve much higher structural similarity index (SSIM) scores on most of the benchmarks than the previous state-of-art methods."
                    },
                    {
                        "title": "Revisiting Skeleton-based Action Recognition",
                        "abstract": "Human skeleton, as a compact representation of human action, has received increasing attention in recent years. Many skeleton-based action recognition methods adopt graph convolutional networks (GCN) to extract features on top of human skeletons. Despite the positive results shown in previous works, GCN-based methods are subject to limitations in robustness, interoperability, and scalability. In this work, we propose PoseC3D, a new approach to skeleton-based action recognition, which relies on a 3D heatmap stack instead of a graph sequence as the base representation of human skeletons. Compared to GCN-based methods, PoseC3D is more effective in learning spatiotemporal features, more robust against pose estimation noises, and generalizes better in cross-dataset settings. Also, PoseC3D can handle multiple-person scenarios without additional computation cost, and its features can be easily integrated with other modalities at early fusion stages, which provides a great design space to further boost the performance. On four challenging datasets, PoseC3D consistently obtains superior performance, when used alone on skeletons and in combination with the RGB modality."
                    },
                    {
                        "title": "Mitigating Representation Bias in Action Recognition: Algorithms and Benchmarks",
                        "abstract": "Deep learning models have achieved excellent recognition results on large-scale video benchmarks. However, they perform poorly when applied to videos with rare scenes or objects, primarily due to the bias of existing video datasets. We tackle this problem from two different angles: algorithm and dataset. From the perspective of algorithms, we propose Spatial-aware Multi-Aspect Debiasing (SMAD), which incorporates both explicit debiasing with multi-aspect adversarial training and implicit debiasing with the spatial actionness reweighting module, to learn a more generic representation invariant to non-action aspects. To neutralize the intrinsic dataset bias, we propose OmniDebias to leverage web data for joint training selectively, which can achieve higher performance with far fewer web data. To verify the effectiveness, we establish evaluation protocols and perform extensive experiments on both re-distributed splits of existing datasets and a new evaluation dataset focusing on the action with rare scenes. We also show that the debiased representation can generalize better when transferred to other datasets and tasks."
                    },
                    {
                        "title": "Omni-sourced Webly-supervised Learning for Video Recognition",
                        "abstract": "We introduce OmniSource, a novel framework for leveraging web data to train video recognition models. OmniSource overcomes the barriers between data formats, such as images, short videos, and long untrimmed videos for webly-supervised learning. First, data samples with multiple formats, curated by task-specific data collection and automatically filtered by a teacher model, are transformed into a unified form. Then a joint-training strategy is proposed to deal with the domain gaps between multiple data sources and formats in webly-supervised learning. Several good practices, including data balancing, resampling, and cross-dataset mixup are adopted in joint training. Experiments show that by utilizing data from multiple sources and formats, OmniSource is more data-efficient in training. With only 3.5M images and 800K minutes videos crawled from the internet without human labeling (less than 2% of prior works), our models learned with OmniSource improve Top-1 accuracy of 2D- and 3D-ConvNet baseline models by 3.0% and 3.9%, respectively, on the Kinetics-400 benchmark. With OmniSource, we establish new records with different pretraining strategies for video recognition. Our best models achieve 80.4%, 80.5%, and 83.6 Top-1 accuracies on the Kinetics-400 benchmark respectively for training-from-scratch, ImageNet pre-training and IG-65M pre-training."
                    },
                    {
                        "title": "Ada-LEval: Evaluating long-context LLMs with length-adaptable benchmarks",
                        "abstract": "Recently, the large language model (LLM) community has shown increasing interest in enhancing LLMs' capability to handle extremely long documents. As various long-text techniques and model architectures emerge, the precise and detailed evaluation of models' long-text capabilities has become increasingly important. Existing long-text evaluation benchmarks, such as L-Eval and LongBench, construct long-text test sets based on open-source datasets, focusing mainly on QA and summarization tasks. These datasets include test samples of varying lengths (from 2k to 32k+) entangled together, making it challenging to assess model capabilities across different length ranges. Moreover, they do not cover the ultralong settings (100k+ tokens) that the latest LLMs claim to achieve. In this paper, we introduce Ada-LEval, a length-adaptable benchmark for evaluating the long-context understanding of LLMs. Ada-LEval includes two challenging subsets, TSort and BestAnswer, which enable a more reliable evaluation of LLMs' long context capabilities. These benchmarks support intricate manipulation of the length of test cases, and can easily produce text samples up to 128k tokens. We evaluate 4 state-of-the-art closed-source API models and 6 open-source models with Ada-LEval. The evaluation results demonstrate the limitations of current LLMs, especially in ultra-long-context settings. Our code is available at https://github.com/open-compass/Ada-LEval."
                    },
                    {
                        "title": "OCSampler: Compressing Videos to One Clip with Single-step Sampling",
                        "abstract": "In this paper, we propose a framework named OCSampler to explore a compact yet effective video representation with one short clip for efficient video recognition. Recent works prefer to formulate frame sampling as a sequential decision task by selecting frames one by one according to their importance, while we present a new paradigm of learning instance-specific video condensation policies to select informative frames for representing the entire video only in a single step. Our basic motivation is that the efficient video recognition task lies in processing a whole sequence at once rather than picking up frames sequentially. Accordingly, these policies are derived from a light-weighted skim network together with a simple yet effective policy network within one step. Moreover, we extend the proposed method with a frame number budget, enabling the framework to produce correct predictions in high confidence with as few frames as possible. Experiments on four benchmarks, i.e., ActivityNet, Mini-Kinetics, FCVID, Mini-Sports1M, demonstrate the effectiveness of our OCSampler over previous methods in terms of accuracy, theoretical computational expense, actual inference speed. We also evaluate its generalization power across different classifiers, sampled frames, and search spaces. Especially, we achieve 76.9% mAP and 21.7 GFLOPs on ActivityNet with an impressive throughput: 123.9 Videos/s on a single TITAN Xp GPU."
                    },
                    {
                        "title": "Self-supervised Action Representation Learning from Partial Spatio-Temporal Skeleton Sequences",
                        "abstract": "Self-supervised learning has demonstrated remarkable capability in representation learning for skeleton-based action recognition. Existing methods mainly focus on applying global data augmentation to generate different views of the skeleton sequence for contrastive learning. However, due to the rich action clues in the skeleton sequences, existing methods may only take a global perspective to learn to discriminate different skeletons without thoroughly leveraging the local relationship between different skeleton joints and video frames, which is essential for real-world applications. In this work, we propose a Partial Spatio-Temporal Learning (PSTL) framework to exploit the local relationship from a partial skeleton sequences built by a unique spatio-temporal masking strategy. Specifically, we construct a negative-sample-free triplet steam structure that is composed of an anchor stream without any masking, a spatial masking stream with Central Spatial Masking (CSM), and a temporal masking stream with Motion Attention Temporal Masking (MATM). The feature cross-correlation matrix is measured between the anchor stream and the other two masking streams, respectively. (1) Central Spatial Masking discards selected joints from the feature calculation process, where the joints with a higher degree of centrality have a higher possibility of being selected. (2) Motion Attention Temporal Masking leverages the motion of action and remove frames that move faster with a higher possibility. Our method achieves SOTA performance on NTU-60, NTU-120 and PKU-MMD under various downstream tasks. A practical evaluation is performed where some skeleton joints are lost in downstream tasks. In contrast to previous methods that suffer from large performance drops, our PSTL can still achieve remarkable results, validating the robustness of our method. Code: https://github.com/YujieOuO/PSTL.git."
                    },
                    {
                        "title": "TRB: A Novel Triplet Representation for Understanding 2D Human Body",
                        "abstract": "Human pose and shape are two important components of 2D human body. However, how to efficiently represent both of them in images is still an open question. In this paper, we propose the Triplet Representation for Body (TRB) -- a compact 2D human body representation, with skeleton keypoints capturing human pose information and contour keypoints containing human shape information. TRB not only preserves the flexibility of skeleton keypoint representation, but also contains rich pose and human shape information. Therefore, it promises broader application areas, such as human shape editing and conditional image generation. We further introduce the challenging problem of TRB estimation, where joint learning of human pose and shape is required. We construct several large-scale TRB estimation datasets, based on popular 2D pose datasets: LSP, MPII, COCO. To effectively solve TRB estimation, we propose a two-branch network (TRB-net) with three novel techniques, namely X-structure (Xs), Directional Convolution (DC) and Pairwise Mapping (PM), to enforce multi-level message passing for joint feature learning. We evaluate our proposed TRB-net and several leading approaches on our proposed TRB datasets, and demonstrate the superiority of our method through extensive evaluations."
                    },
                    {
                        "title": "SkeleTR: Towrads Skeleton-based Action Recognition in the Wild",
                        "abstract": "We present SkeleTR, a new framework for skeleton-based action recognition. In contrast to prior work, which focuses mainly on controlled environments, we target more general scenarios that typically involve a variable number of people and various forms of interaction between people. SkeleTR works with a two-stage paradigm. It first models the intra-person skeleton dynamics for each skeleton sequence with graph convolutions, and then uses stacked Transformer encoders to capture person interactions that are important for action recognition in general scenarios. To mitigate the negative impact of inaccurate skeleton associations, SkeleTR takes relative short skeleton sequences as input and increases the number of sequences. As a unified solution, SkeleTR can be directly applied to multiple skeleton-based action tasks, including video-level action classification, instance-level action detection, and group-level activity recognition. It also enables transfer learning and joint training across different action tasks and datasets, which result in performance improvement. When evaluated on various skeleton-based action recognition benchmarks, SkeleTR achieves the state-of-the-art performance."
                    },
                    {
                        "title": "ProSA: Assessing and Understanding the Prompt Sensitivity of LLMs",
                        "abstract": "Large language models (LLMs) have demonstrated impressive capabilities across various tasks, but their performance is highly sensitive to the prompts utilized. This variability poses challenges for accurate assessment and user satisfaction. Current research frequently overlooks instance-level prompt variations and their implications on subjective evaluations. To address these shortcomings, we introduce ProSA, a framework designed to evaluate and comprehend prompt sensitivity in LLMs. ProSA incorporates a novel sensitivity metric, PromptSensiScore, and leverages decoding confidence to elucidate underlying mechanisms. Our extensive study, spanning multiple tasks, uncovers that prompt sensitivity fluctuates across datasets and models, with larger models exhibiting enhanced robustness. We observe that few-shot examples can alleviate this sensitivity issue, and subjective evaluations are also susceptible to prompt sensitivities, particularly in complex, reasoning-oriented tasks. Furthermore, our findings indicate that higher model confidence correlates with increased prompt robustness. We believe this work will serve as a helpful tool in studying prompt sensitivity of LLMs. The project is released at: https://github.com/open-compass/ProSA ."
                    },
                    {
                        "title": "MMBench-Video: A Long-Form Multi-Shot Benchmark for Holistic Video Understanding",
                        "abstract": "The advent of large vision-language models (LVLMs) has spurred research into their applications in multi-modal contexts, particularly in video understanding. Traditional VideoQA benchmarks, despite providing quantitative metrics, often fail to encompass the full spectrum of video content and inadequately assess models' temporal comprehension. To address these limitations, we introduce MMBench-Video, a quantitative benchmark designed to rigorously evaluate LVLMs' proficiency in video understanding. MMBench-Video incorporates lengthy videos from YouTube and employs free-form questions, mirroring practical use cases. The benchmark is meticulously crafted to probe the models' temporal reasoning skills, with all questions human-annotated according to a carefully constructed ability taxonomy. We employ GPT-4 for automated assessment, demonstrating superior accuracy and robustness over earlier LLM-based evaluations. Utilizing MMBench-Video, we have conducted comprehensive evaluations that include both proprietary and open-source LVLMs for images and videos. MMBench-Video stands as a valuable resource for the research community, facilitating improved evaluation of LVLMs and catalyzing progress in the field of video understanding. The evalutation code of MMBench-Video will be integrated into VLMEvalKit: https://github.com/open-compass/VLMEvalKit."
                    }
                ]
            },
            "63aa347f-07f2-48c1-840d-715f27fb8ba8": {
                "pk": "63aa347f-07f2-48c1-840d-715f27fb8ba8",
                "name": "Xinyu Fang",
                "collaborators": [
                    "Yujia Gu",
                    "Zihan Peng",
                    "Yinan Peng",
                    "Steven J. Miller",
                    "Maxwell Sun",
                    "Amanda Verga",
                    "Xueyuan Deng",
                    "Haofeng Li"
                ],
                "domain": [
                    "Statistical Modeling",
                    "Machine Learning",
                    "Art Generation",
                    "Style Transfer"
                ],
                "publications": [
                    {
                        "title": "Generalized Continuous and Discrete Stick Fragmentation and Benford's Law",
                        "abstract": "Inspired by the basic stick fragmentation model proposed by Becker et al. in arXiv:1309.5603v4, we consider three new versions of such fragmentation models, namely, continuous with random number of parts, continuous with probabilistic stopping, and discrete with congruence stopping conditions. In all of these situations, we state and prove precise conditions for the ending stick lengths to obey Benford's law when taking the appropriate limits. We introduce the aggregated limit, necessary to guarantee convergence to Benford's law in the latter two models. We also show that resulting stick lengths are non-Benford when our conditions are not met. Moreover, we give a sufficient condition for a distribution to satisfy the Mellin transform condition introduced in arXiv:0805.4226v2, yielding a large family of examples."
                    },
                    {
                        "title": "A New Chinese Landscape Paintings Generation Model based on Stable Diffusion using DreamBooth",
                        "abstract": "This study mainly introduces a method combining the Stable Diffusion Model (SDM) and Parameter-Efficient Fine-Tuning method for generating Chinese Landscape Paintings. This training process is accelerated by combining LoRA with pre-trained SDM and DreamBooth with pre-trained SDM, respectively. On the Chinese Landscape Paintings Internet dataset used in this paper, this study finds that SDM combined with DreamBooth exhibits superior performance, outperforming other models, including the generic pre-trained SDM and LoRA-based fine-tuning SDM. The SDM combined with DreamBooth achieves a FID of 12.75 on the dataset and outperforms all other models in terms of expert evaluation, highlighting the model's versatility in the field of Chinese Landscape Paintings given the unique identifier, high fidelity and high quality. This study illustrates the potential of specialised fine-tuning method to improve the performance of SDM on domain-specific tasks, particularly in the domain of Landscape Paintings."
                    },
                    {
                        "title": "JieHua Paintings Style Feature Extracting Model using Stable Diffusion with ControlNet",
                        "abstract": "This study proposes a novel approach to extract stylistic features of Jiehua: the utilization of the Fine-tuned Stable Diffusion Model with ControlNet (FSDMC) to refine depiction techniques from artists' Jiehua. The training data for FSDMC is based on the opensource Jiehua artist's work collected from the Internet, which were subsequently manually constructed in the format of (Original Image, Canny Edge Features, Text Prompt). By employing the optimal hyperparameters identified in this paper, it was observed FSDMC outperforms CycleGAN, another mainstream style transfer model. FSDMC achieves FID of 3.27 on the dataset and also surpasses CycleGAN in terms of expert evaluation. This not only demonstrates the model's high effectiveness in extracting Jiehua's style features, but also preserves the original pre-trained semantic information. The findings of this study suggest that the application of FSDMC with appropriate hyperparameters can enhance the efficacy of the Stable Diffusion Model in the field of traditional art style migration tasks, particularly within the context of Jiehua."
                    }
                ]
            },
            "9c304789-32b1-4789-bba3-58f9551a5b24": {
                "pk": "9c304789-32b1-4789-bba3-58f9551a5b24",
                "name": "Junming Yang",
                "collaborators": [
                    "Xingguo Chen",
                    "Shengyuan Wang",
                    "Bolei Zhang",
                    "Qi Wang",
                    "Yunbo Wang",
                    "Xin Jin",
                    "Wenjun Zeng",
                    "Xiaokang Yang",
                    "Yaoqi Li",
                    "Xuanyu Chen"
                ],
                "domain": [
                    "Reinforcement Learning",
                    "Adversarial Learning",
                    "Multimodal Models",
                    "Stock Prediction"
                ],
                "publications": [
                    {
                        "title": "Model-based Offline Policy Optimization with Adversarial Network",
                        "abstract": "Model-based offline reinforcement learning (RL), which builds a supervised transition model with logging dataset to avoid costly interactions with the online environment, has been a promising approach for offline policy optimization. As the discrepancy between the logging data and online environment may result in a distributional shift problem, many prior works have studied how to build robust transition models conservatively and estimate the model uncertainty accurately. However, the over-conservatism can limit the exploration of the agent, and the uncertainty estimates may be unreliable. In this work, we propose a novel Model-based Offline policy optimization framework with Adversarial Network (MOAN). The key idea is to use adversarial learning to build a transition model with better generalization, where an adversary is introduced to distinguish between in-distribution and out-of-distribution samples. Moreover, the adversary can naturally provide a quantification of the model's uncertainty with theoretical guarantees. Extensive experiments showed that our approach outperforms existing state-of-the-art baselines on widely studied offline RL benchmarks. It can also generate diverse in-distribution samples, and quantify the uncertainty more accurately."
                    },
                    {
                        "title": "Making Offline RL Online: Collaborative World Models for Offline Visual Reinforcement Learning",
                        "abstract": "Training offline reinforcement learning (RL) models using visual inputs poses two significant challenges, i.e., the overfitting problem in representation learning and the overestimation bias for expected future rewards. Recent work has attempted to alleviate the overestimation bias by encouraging conservative behaviors. This paper, in contrast, tries to build more flexible constraints for value estimation without impeding the exploration of potential advantages. The key idea is to leverage off-the-shelf RL simulators, which can be easily interacted with in an online manner, as the \"test bed\" for offline policies. To enable effective online-to-offline knowledge transfer, we introduce CoWorld, a model-based RL approach that mitigates cross-domain discrepancies in state and reward spaces. Experimental results demonstrate the effectiveness of CoWorld, outperforming existing RL approaches by large margins."
                    },
                    {
                        "title": "Deep Learning for Stock Selection Based on High Frequency Price-Volume Data",
                        "abstract": "Training a practical and effective model for stock selection has been a greatly concerned problem in the field of artificial intelligence. Even though some of the models from previous works have achieved good performance in the U.S. market by using low-frequency data and features, training a suitable model with high-frequency stock data is still a problem worth exploring. Based on the high-frequency price data of the past several days, we construct two separate models-Convolution Neural Network and Long Short-Term Memory-which can predict the expected return rate of stocks on the current day, and select the stocks with the highest expected yield at the opening to maximize the total return. In our CNN model, we propose improvements on the CNNpred model presented by E. Hoseinzade and S. Haratizadeh in their paper which deals with low-frequency features. Such improvements enable our CNN model to exploit the convolution layer's ability to extract high-level factors and avoid excessive loss of original information at the same time. Our LSTM model utilizes Recurrent Neural Network'advantages in handling time series data. Despite considerable transaction fees due to the daily changes of our stock position, annualized net rate of return is 62.27% for our CNN model, and 50.31% for our LSTM model."
                    },
                    {
                        "title": "CoCoT: Contrastive Chain-of-Thought Prompting for Large Multimodal Models with Multiple Image Inputs",
                        "abstract": "When exploring the development of Artificial General Intelligence (AGI), a critical task for these models involves interpreting and processing information from multiple image inputs. However, Large Multimodal Models (LMMs) encounter two issues in such scenarios: (1) a lack of fine-grained perception, and (2) a tendency to blend information across multiple images. We first extensively investigate the capability of LMMs to perceive fine-grained visual details when dealing with multiple input images. The research focuses on two aspects: first, image-to-image matching (to evaluate whether LMMs can effectively reason and pair relevant images), and second, multi-image-to-text matching (to assess whether LMMs can accurately capture and summarize detailed image information). We conduct evaluations on a range of both open-source and closed-source large models, including GPT-4V, Gemini, OpenFlamingo, and MMICL. To enhance model performance, we further develop a Contrastive Chain-of-Thought (CoCoT) prompting approach based on multi-input multimodal models. This method requires LMMs to compare the similarities and differences among multiple image inputs, and then guide the models to answer detailed questions about multi-image inputs based on the identified similarities and differences. Our experimental results showcase CoCoT's proficiency in enhancing the multi-image comprehension capabilities of large multimodal models."
                    }
                ]
            },
            "f064f019-93fb-429b-902a-4937071f6c20": {
                "pk": "f064f019-93fb-429b-902a-4937071f6c20",
                "name": "Lin Chen",
                "collaborators": [
                    "Qingtao Chen",
                    "Yutong Jiang",
                    "Lin Zhang",
                    "Kaifeng Bu"
                ],
                "domain": [
                    "Quantum Physics",
                    "Algebraic Geometry",
                    "Representation Theory",
                    "Mathematical Physics"
                ],
                "publications": [
                    {
                        "title": "Symmetrized cut-join equation of Marino-Vafa formula",
                        "abstract": "In this note, we symmetrized the cut-join equation from the proof of Marino-Vafa formula by applying a transcendental changing of variable. One can derive more recursion formulas of Hodge integrals out of this polynomial equations. We also give some applications."
                    },
                    {
                        "title": "Bouchard-Klemm-Marino-Pasquetti Conjecture for $\\mathbb{C}^3$",
                        "abstract": "In this paper, we give a proof of the Bouchard-Klemm-Marino-Pasquetti conjecture for a framed vertex, by using the symmetrized Cut-Join Equation developed in a previous paper."
                    },
                    {
                        "title": "Length of separable states and symmetrical informationally complete (SIC) POVM",
                        "abstract": "This short note reviews the notion and fundamental properties of SIC-POVM and its connection with the length of separable states. We also review the t-design."
                    },
                    {
                        "title": "Framed Deformation of Galois Representation",
                        "abstract": "We studied framed deformations of two dimensional Galois representation of which the residue representation restrict to decomposition groups are scalars, and established a modular lifting theorem for certain cases. We then proved a family version of the result, and used it to determine the structure of deformation rings over characteristic zero fields. These can be applied to the study of exceptional zero of p-adic L-function."
                    },
                    {
                        "title": "Deligne--Lusztig duality on the moduli stack of bundles",
                        "abstract": "Let $Bun_G(X)$ be the moduli stack of $G$-torsors on a smooth projective curve $X$ for a reductive group $G$. We prove a conjecture made by Drinfeld-Wang and Gaitsgory on the Deligne-Lusztig duality for D-modules on $Bun_G(X)$. This conjecture relates Drinfeld-Gaitsgory's pseudo-identity functors to the enhanced Eisenstein series and geometric constant term functors on $DMod(Bun_G(X))$. We also prove a \"second adjointness\" result for these enhanced functors."
                    },
                    {
                        "title": "Nearby cycles on Drinfeld-Gaitsgory-Vinberg Interpolation Grassmannian and long intertwining functor",
                        "abstract": "Let $G$ be a reductive group and $U,U^-$ be the unipotent radicals of a pair of opposite parabolic subgroups $P,P^-$. We prove that the DG-categories of $U(\\!(t)\\!)$-equivariant and $U^-(\\!(t)\\!)$-equivariant D-modules on the affine Grassmannian $Gr_G$ are canonically dual to each other. We show that the unit object witnessing this duality is given by nearby cycles on the Drinfeld-Gaitsgory-Vinberg interpolation Grassmannian defined in arXiv:1805.07721. We study various properties of the mentioned nearby cycles, in particular compare them with the nearby cycles studied in arXiv:1411.4206 and arXiv:1607.00586. We also generalize our results to the Beilinson-Drinfeld Grassmannian $Gr_{G,X^I}$ and to the affine flag variety $Fl_G$."
                    },
                    {
                        "title": "Orthogonal Quantum Group Invariants of Links",
                        "abstract": "We study the Chern-Simons partition function of orthogonal quantum group invariants, and propose a new orthogonal Labastida-Mari\\~{n}o-Ooguri-Vafa conjecture as well as degree conjecture for free energy associated to the orthogonal Chern-Simons partition function. We prove the degree conjecture and some interesting cases of orthogonal LMOV conjecture. In particular, We provide a formula of colored Kauffman polynomials for torus knots and links, and applied this formula to verify certain case of the conjecture at roots of unity except $1$. We also derive formulas of Lickorish-Millett type for Kauffman polynomials and relate all these to the orthogonal LMOV conjecture."
                    },
                    {
                        "title": "Nonlocality via multiqubit orthogonal product bases",
                        "abstract": "We investigate the quantum nonlocality via the discrimination on two, three and four-qubit orthogonal product bases (OPBs). We show that every two-qubit, and some three and four-qubit OPBs can be locally distinguished. It turns out that the remaining three and four-qubit OPBs cannot be locally distinguished, and thus they show the quantum nonlocality without entanglement. We also distinguish them by merging some systems using quantum teleportation with assisted Bell states."
                    },
                    {
                        "title": "Fidelity between a bipartite state and another one undergoing local unitary dynamics",
                        "abstract": "The fidelity and local unitary transformation are two widely useful notions in quantum physics. We study two constrained optimization problems in terms of the maximal and minimal fidelity between two bipartite quantum states undergoing local unitary dynamics. The problems are related to the geometric measure of entanglement and the distillability problem. We show that the problems can be reduced to semi-definite programming optimization problems. We give close-form formulae of the fidelity when the two states are both pure states, or a pure product state and the Werner state. We explain from the point of view of local unitary actions that why the entanglement in Werner states is hard to accessible. For general mixed states, we give upper and lower bounds of the fidelity using tools such as affine fidelity, channels and relative entropy from information theory. We also investigate the power of local unitaries and quantification for the commutativity of quantum states, and the equivalence of the two optimization problems."
                    }
                ]
            },
            "6c8d9b48-af19-4fe6-9597-2b3762ab9fd2": {
                "pk": "6c8d9b48-af19-4fe6-9597-2b3762ab9fd2",
                "name": "Songyang Zhang",
                "collaborators": [
                    "Zhi Ding",
                    "Shuguang Cui",
                    "Jiebo Luo",
                    "Xuming He",
                    "Weimin Lei",
                    "Qinwen Deng",
                    "Changpeng Yang",
                    "Han Zhang",
                    "Wei Zhang",
                    "Yunchong Guan"
                ],
                "domain": [
                    "Congestion Control",
                    "Graph Signal Processing",
                    "Video Analysis",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Congestion Control for RTP Media: a Comparison on Simulated Environment",
                        "abstract": "To develop low latency congestion control algorithm for real time taffic has been gained attention recently. RTP Media Congestion Avoidance Techniques (RMCAT) working group was initiated for standard defination. There are three algorithms under this group, Network Assisted Dynamic Adaptation (NADA) proposed by Cisco, Google Congestion Control (GCC) proposed by Google and Self-Clocked Rate Adaptation for Multimedia(SCReAM) proposed by Ericsson. This paper compares and analyses the performance of these algorithms on simulation environment. Results show GCC has well fairness property and performs well in lossy link but slow convergence in dynamic link, NADA stabilizes its rate quickly but suffers \"late-comer\" effect, SCReAM has the lowest queue occupation but also lower link capacity utilization."
                    },
                    {
                        "title": "An Evaluation of BBR and its variants",
                        "abstract": "The congestion control algorithm bring such importance that it avoids the network link into severe congestion and guarantees network normal operation. Since The loss based algorithms introduce high transmission delay, to design new algorithm simultaneously achieving high throughout and low buffer occupation is a new working direction. The bottleneck bandwidth and round trip time (BBR) belongs such kind, and it has drawn much attention since its release. There are other algorithms modified from BBR to gain better performance. And the implementation of BBR v2.0 is released recently. We implement a framework to compare the performance of these algorithms in simulated environment."
                    },
                    {
                        "title": "An Online Learning Based Path Selection for Multipath Video Telephony Service in Overlay",
                        "abstract": "Even real time video telephony services have been pervasively applied, providing satisfactory quality of experience to users is still a challenge task especially in wireless networks. Multipath transmission is a promising solution to improve video quality by aggregating bandwidth. In existing multipath transmission solutions, sender concurrently splits traffic on default routing paths and has no flexibility to select paths. When default paths fall into severe congestion and the available bandwidth decreases, sender has to decrease video quality by reducing resolution or encoding bitrate. Deploying relay servers in current infrastructure to form overlay network provides path diversity. An online learning approach based on multi-armed bandits is applied for path selection to harvest maximum profit. Further, a congestion control algorithm adapted from BBR is implemented to probe bandwidth and to avoid link congestion. To maintain throughput stability and fairness, a smaller probe up gain value is used and the cycle length in bandwidth probe phase is randomized. To reduce delay, the inflight packets will be reduced to match with the estimated bandwidth delay product in the probe down phase. Experiments are conducted to verify the effectiveness the proposed solution to improve throughput and quality in video communication service."
                    },
                    {
                        "title": "LearningCC: An online learning approach for congestion control",
                        "abstract": "Recently, much effort has been devoted by researchers from both academia and industry to develop novel congestion control methods. LearningCC is presented in this letter, in which the congestion control problem is solved by reinforce learning approach. Instead of adjusting the congestion window with fixed policy, there are serval options for an endpoint to choose. To predict the best option is a hard task. Each option is mapped as an arm of a bandit machine. The endpoint can learn to determine the optimal choice through trial and error method. Experiments are performed on ns3 platform to verify the effectiveness of LearningCC by comparing with other benchmark algorithms. Results indicate it can achieve lower transmission delay than loss based algorithms. Especially, we found LearningCC makes significant improvement in link suffering from random loss."
                    },
                    {
                        "title": "An Optimized BBR for Multipath Real Time Video Streaming",
                        "abstract": "The multipath transmission scheme can work as an effective way to provide better quality of experiments to end users. Two key research points in the multipath real time video transmission context are congestion control and packet scheduling. As Utility maximization theory shows, to provide better satisfaction to end users is to provide higher throughput and lower transmission delay. The congestion control is responsible to converge to the maximum available bandwidth and avoid leading the network into congestion. A delay response BBR (Delay-BBR) algorithm optimized for real time video transmission is proposed, and the main idea is to reduce sending rate when the link delay has exceeded certain threshold to let the intermediate routers drain the occupied buffer. It can achieve better transmission delay and lower packet loss rate compared with QUIC-BBR and WebRTC-BBR by experiment. And a packet scheduling algorithm induced from Utility maximization theory works on top of the congestion control algorithm is tested and achieves lower frame delivery delay further compared with benchmark algorithms."
                    },
                    {
                        "title": "Cross: A Delay Based Congestion Control Method for RTP Media",
                        "abstract": "After more than a decade of development, real time communication (RTC) for video telephony has made significantly progress. However, emerging high-quality RTC applications with high definition and high frame rate requires sufficient bandwidth. The default congestion control mechanism specifically tuned for video telephony leaves plenty of room for optimization under high-rate scenarios. It is necessary to develop new rate control solutions to utilize bandwidth efficiently and to provide better experience for such services. A delay-based congestion control method called Cross is proposed, which regulates rate based on queue load with a multiplicative increase and multiplicative decrease fashion. A simulation module is developed to validate the effectiveness of these congestion control algorithms for RTC services. The module is released with the hope to provide convenience for RTC research community. Simulation results demonstrate that Cross can achieve low queuing delay and maintain high channel utilization under random loss environments. Online deployment shows that Cross can reduce the video freezing ratio by up to 58.45\\% on average when compared with a benchmark algorithm."
                    },
                    {
                        "title": "From Spectrum Wavelet to Vertex Propagation: Graph Convolutional Networks Based on Taylor Approximation",
                        "abstract": "Graph convolutional networks (GCN) have been recently utilized to extract the underlying structures of datasets with some labeled data and high-dimensional features. Existing GCNs mostly rely on a first-order Chebyshev approximation of graph wavelet-kernels. Such a generic propagation model does not always suit the various datasets and their features. This work revisits the fundamentals of graph wavelet and explores the utility of signal propagation in the vertex domain to approximate the spectral wavelet-kernels. We first derive the conditions for representing the graph wavelet-kernels via vertex propagation. We next propose alternative propagation models for GCN layers based on Taylor expansions. We further analyze the choices of detailed graph representations for TGCNs. Experiments on citation networks, multimedia datasets and synthetic graphs demonstrate the advantage of Taylor-based GCN (TGCN) in the node classification problems over the traditional GCN methods."
                    },
                    {
                        "title": "Shared Bottleneck Detecction Based on Trend Line Regression for Multipath Transmission",
                        "abstract": "The current deployed multipath congestion control algorithms couple all the subflows together to avoid bandwidth occupation aggressiveness if the subflows of multipath transmission protocol share common bottleneck with single path TCP. The coupled congestion control algorithms can guarantee well fairness property in common bottleneck but result in rate increase conservativeness in none-sharing bottleneck situation. Thus, the throughput of multipath session can be further improved when combing with effective shared bottleneck detection mechanism. This paper proposes a delay trend line regression method to detect if flows share common bottleneck. Deduced from TCP fluid model, the packet round trip delay signal shows linear increase property during the queue building up process of the narrowest link and the delay trend line slopes of two flows are in close proximity if they traverse the same bottleneck link. The proposed method is implemented on multipath QUIC golang codebase and extensive simulations are performed to validate its effectiveness in detecting out flows traversing common bottleneck. If the subflows are detected out via a common bottleneck, the sender would perform coupled congestion control algorithm and perform congestion control seperately on flow level in none sharing bottleneck case. Results show a multipath session with two subflows can obtain 74\\% gain on average in throughput compared with single path connection when Linked Increases Algorithm (LIA) is in combination with trend line regession shared bottlenck detection algorithm in none shared bottleneck, and show well fairness property in common bottleneck scenarios."
                    },
                    {
                        "title": "Content-based Analysis of the Cultural Differences between TikTok and Douyin",
                        "abstract": "Short-form video social media shifts away from the traditional media paradigm by telling the audience a dynamic story to attract their attention. In particular, different combinations of everyday objects can be employed to represent a unique scene that is both interesting and understandable. Offered by the same company, TikTok and Douyin are popular examples of such new media that has become popular in recent years, while being tailored for different markets (e.g. the United States and China). The hypothesis that they express cultural differences together with media fashion and social idiosyncrasy is the primary target of our research. To that end, we first employ the Faster Regional Convolutional Neural Network (Faster R-CNN) pre-trained with the Microsoft Common Objects in COntext (MS-COCO) dataset to perform object detection. Based on a suite of objects detected from videos, we perform statistical analysis including label statistics, label similarity, and label-person distribution. We further use the Two-Stream Inflated 3D ConvNet (I3D) pre-trained with the Kinetics dataset to categorize and analyze human actions. By comparing the distributional results of TikTok and Douyin, we uncover a wealth of similarity and contrast between the two closely related video social media platforms along the content dimensions of object quantity, object categories, and human action categories."
                    },
                    {
                        "title": "Introducing Hypergraph Signal Processing: Theoretical Foundation and Practical Applications",
                        "abstract": "Signal processing over graphs has recently attracted significant attentions for dealing with structured data. Normal graphs, however, only model pairwise relationships between nodes and are not effective in representing and capturing some high-order relationships of data samples, which are common in many applications such as Internet of Things (IoT). In this work, we propose a new framework of hypergraph signal processing (HGSP) based on tensor representation to generalize the traditional graph signal processing (GSP) to tackle high-order interactions. We introduce the core concepts of HGSP and define the hypergraph Fourier space. We then study the spectrum properties of hypergraph Fourier transform and explain its connection to mainstream digital signal processing. We derive the novel hypergraph sampling theory and present the fundamentals of hypergraph filter design based on the tensor framework. We present HGSP-based methods for several signal processing and data analysis applications. Our experimental results demonstrate significant performance improvement using our HGSP framework over some traditional signal processing solutions."
                    },
                    {
                        "title": "Hypergraph Spectral Analysis and Processing in 3D Point Cloud",
                        "abstract": "Along with increasingly popular virtual reality applications, the three-dimensional (3D) point cloud has become a fundamental data structure to characterize 3D objects and surroundings. To process 3D point clouds efficiently, a suitable model for the underlying structure and outlier noises is always critical. In this work, we propose a hypergraph-based new point cloud model that is amenable to efficient analysis and processing. We introduce tensor-based methods to estimate hypergraph spectrum components and frequency coefficients of point clouds in both ideal and noisy settings. We establish an analytical connection between hypergraph frequencies and structural features. We further evaluate the efficacy of hypergraph spectrum estimation in two common point cloud applications of sampling and denoising for which also we elaborate specific hypergraph filter design and spectral properties. The empirical performance demonstrates the strength of hypergraph signal processing as a tool in 3D point clouds and the underlying properties."
                    },
                    {
                        "title": "Point Cloud Segmentation based on Hypergraph Spectral Clustering",
                        "abstract": "Hypergraph spectral analysis has emerged as an effective tool processing complex data structures in data analysis. The surface of a three-dimensional (3D) point cloud and the multilateral relationship among their points can be naturally captured by the high-dimensional hyperedges. This work investigates the power of hypergraph spectral analysis in unsupervised segmentation of 3D point clouds. We estimate and order the hypergraph spectrum from observed point cloud coordinates. By trimming the redundancy from the estimated hypergraph spectral space based on spectral component strengths, we develop a clustering-based segmentation method. We apply the proposed method to various point clouds, and analyze their respective spectral properties. Our experimental results demonstrate the effectiveness and efficiency of the proposed segmentation method."
                    },
                    {
                        "title": "Mi YouTube es Su YouTube? Analyzing the Cultures using YouTube Thumbnails of Popular Videos",
                        "abstract": "YouTube, a world-famous video sharing website, maintains a list of the top trending videos on the platform. Due to its huge amount of users, it enables researchers to understand people's preference by analyzing the trending videos. Trending videos vary from country to country. By analyzing such differences and changes, we can tell how users' preferences differ over locations. Previous work focuses on analyzing such culture preferences from videos' metadata, while the culture information hidden within the visual content has not been discovered. In this study, we explore culture preferences among countries using the thumbnails of YouTube trending videos. We first process the thumbnail images of the videos using object detectors. The collected object information is then used for various statistical analysis. In particular, we examine the data from three perspectives: geographical locations, video genres and users' reactions. Experimental results indicate that the users from similar cultures shares interests in watching similar videos on YouTube. Our study demonstrates that discovering the culture preference through the thumbnails can be an effective mechanism for video social media analysis."
                    },
                    {
                        "title": "Exploiting Temporal Relationships in Video Moment Localization with Natural Language",
                        "abstract": "We address the problem of video moment localization with natural language, i.e. localizing a video segment described by a natural language sentence. While most prior work focuses on grounding the query as a whole, temporal dependencies and reasoning between events within the text are not fully considered. In this paper, we propose a novel Temporal Compositional Modular Network (TCMN) where a tree attention network first automatically decomposes a sentence into three descriptions with respect to the main event, context event and temporal signal. Two modules are then utilized to measure the visual similarity and location similarity between each segment and the decomposed descriptions. Moreover, since the main event and context event may rely on different modalities (RGB or optical flow), we use late fusion to form an ensemble of four models, where each model is independently trained by one combination of the visual input. Experiments show that our model outperforms the state-of-the-art methods on the TEMPO dataset."
                    },
                    {
                        "title": "LatentGNN: Learning Efficient Non-local Relations for Visual Recognition",
                        "abstract": "Capturing long-range dependencies in feature representations is crucial for many visual recognition tasks. Despite recent successes of deep convolutional networks, it remains challenging to model non-local context relations between visual features. A promising strategy is to model the feature context by a fully-connected graph neural network (GNN), which augments traditional convolutional features with an estimated non-local context representation. However, most GNN-based approaches require computing a dense graph affinity matrix and hence have difficulty in scaling up to tackle complex real-world visual problems. In this work, we propose an efficient and yet flexible non-local relation representation based on a novel class of graph neural networks. Our key idea is to introduce a latent space to reduce the complexity of graph, which allows us to use a low-rank representation for the graph affinity matrix and to achieve a linear complexity in computation. Extensive experimental evaluations on three major visual recognition tasks show that our method outperforms the prior works with a large margin while maintaining a low computation cost."
                    },
                    {
                        "title": "Learning Implicit Temporal Alignment for Few-shot Video Classification",
                        "abstract": "Few-shot video classification aims to learn new video categories with only a few labeled examples, alleviating the burden of costly annotation in real-world applications. However, it is particularly challenging to learn a class-invariant spatial-temporal representation in such a setting. To address this, we propose a novel matching-based few-shot learning strategy for video sequences in this work. Our main idea is to introduce an implicit temporal alignment for a video pair, capable of estimating the similarity between them in an accurate and robust manner. Moreover, we design an effective context encoding module to incorporate spatial and feature channel context, resulting in better modeling of intra-class variations. To train our model, we develop a multi-task loss for learning video matching, leading to video features with better generalization. Extensive experimental results on two challenging benchmarks, show that our method outperforms the prior arts with a sizable margin on SomethingSomething-V2 and competitive results on Kinetics."
                    },
                    {
                        "title": "Image Processing via Multilayer Graph Spectra",
                        "abstract": "Graph signal processing (GSP) has become an important tool in image processing because of its ability to reveal underlying data structures. Many real-life multimedia datasets, however, exhibit heterogeneous structures across frames. Multilayer graphs (MLG), instead of traditional single-layer graphs, provide better representation of these datasets such as videos and hyperspectral images. To generalize GSP to multilayer graph models and develop multilayer analysis for image processing, this work introduces a tensor-based framework of multilayer graph signal processing (M-GSP) and present useful M-GSP tools for image processing. We then present guidelines for applying M-GSP in image processing and introduce several applications, including RGB image compression, edge detection and hyperspectral image segmentation. Successful experimental results demonstrate the efficacy and promising futures of M-GSP in image processing."
                    },
                    {
                        "title": "Learning 2D Temporal Adjacent Networks for Moment Localization with Natural Language",
                        "abstract": "We address the problem of retrieving a specific moment from an untrimmed video by a query sentence. This is a challenging problem because a target moment may take place in relations to other temporal moments in the untrimmed video. Existing methods cannot tackle this challenge well since they consider temporal moments individually and neglect the temporal dependencies. In this paper, we model the temporal relations between video moments by a two-dimensional map, where one dimension indicates the starting time of a moment and the other indicates the end time. This 2D temporal map can cover diverse video moments with different lengths, while representing their adjacent relations. Based on the 2D map, we propose a Temporal Adjacent Network (2D-TAN), a single-shot framework for moment localization. It is capable of encoding the adjacent temporal relation, while learning discriminative features for matching video moments with referring expressions. We evaluate the proposed 2D-TAN on three challenging benchmarks, i.e., Charades-STA, ActivityNet Captions, and TACoS, where our 2D-TAN outperforms the state-of-the-art."
                    },
                    {
                        "title": "Part-aware Prototype Network for Few-shot Semantic Segmentation",
                        "abstract": "Few-shot semantic segmentation aims to learn to segment new object classes with only a few annotated examples, which has a wide range of real-world applications. Most existing methods either focus on the restrictive setting of one-way few-shot segmentation or suffer from incomplete coverage of object regions. In this paper, we propose a novel few-shot semantic segmentation framework based on the prototype representation. Our key idea is to decompose the holistic class representation into a set of part-aware prototypes, capable of capturing diverse and fine-grained object features. In addition, we propose to leverage unlabeled data to enrich our part-aware prototypes, resulting in better modeling of intra-class variations of semantic objects. We develop a novel graph neural network model to generate and enhance the proposed part-aware prototypes based on labeled and unlabeled images. Extensive experimental evaluations on two benchmarks show that our method outperforms the prior art with a sizable margin."
                    },
                    {
                        "title": "Hyperspectral Image Segmentation based on Graph Processing over Multilayer Networks",
                        "abstract": "Hyperspectral imaging is an important sensing technology with broad applications and impact in areas including environmental science, weather, and geo/space exploration. One important task of hyperspectral image (HSI) processing is the extraction of spectral-spatial features. Leveraging on the recent-developed graph signal processing over multilayer networks (M-GSP), this work proposes several approaches to HSI segmentation based on M-GSP feature extraction. To capture joint spectral-spatial information, we first customize a tensor-based multilayer network (MLN) model for HSI, and define a MLN singular space for feature extraction. We then develop an unsupervised HSI segmentation method by utilizing MLN spectral clustering. Regrouping HSI pixels via MLN-based clustering, we further propose a semi-supervised HSI classification based on multi-resolution fusions of superpixels. Our experimental results demonstrate the strength of M-GSP in HSI processing and spectral-spatial information extraction."
                    }
                ]
            },
            "2d062c26-9f61-421e-bca7-cfadcb697e9c": {
                "pk": "2d062c26-9f61-421e-bca7-cfadcb697e9c",
                "name": "Jiaqi Wang",
                "collaborators": [
                    "Xi Li",
                    "Chen Wu",
                    "Shuzhen Yang",
                    "Daqing Yang",
                    "Dongyan Fu",
                    "Yubing Dong",
                    "Hang Xie",
                    "Yihong Wu",
                    "Yong Ao",
                    "Wenming Zou"
                ],
                "domain": [
                    "Federated Learning",
                    "Privacy",
                    "Machine Learning",
                    "Graph Theory"
                ],
                "publications": [
                    {
                        "title": "An In-depth Review of Privacy Concerns Raised by the COVID-19 Pandemic",
                        "abstract": "COVID-19 has hugely changed our lives, work, and interactions with people. With more and more online activities, people are easily exposed to privacy threats. In this paper, we explore how users self-disclose on social media and privacy concerns raised from these behaviors. Based on recent news, techniques, and research, we indicate three increasing privacy threats caused by the COVID-19 pandemic. After that, we provide a systematic analysis of potential privacy issues related to the COVID pandemic. Furthermore, we propose a series of research directions about online user self-disclosure and privacy issues for future work as well as possible solutions."
                    },
                    {
                        "title": "Stochastic maximum principle for recursive optimal control problems with varying terminal time",
                        "abstract": "This paper introduces a new recursive stochastic optimal control problem driven by a forward-backward stochastic differential equations (FBSDEs), where the ter?minal time varies according to the constraints of the state of the forward equation. This new optimal control problem can be used to describe the investment portfolio problems with the varying investment period. Based on novel \\r{ho}-moving variational and adjoint equations, we establish the stochastic maximum principle for this optimal control problem including the classical optimal control problem as a particular case. Furthermore, we propose an example to verify our main results."
                    },
                    {
                        "title": "Position Paper: Assessing Robustness, Privacy, and Fairness in Federated Learning Integrated with Foundation Models",
                        "abstract": "Federated Learning (FL), while a breakthrough in decentralized machine learning, contends with significant challenges such as limited data availability and the variability of computational resources, which can stifle the performance and scalability of the models. The integration of Foundation Models (FMs) into FL presents a compelling solution to these issues, with the potential to enhance data richness and reduce computational demands through pre-training and data augmentation. However, this incorporation introduces novel issues in terms of robustness, privacy, and fairness, which have not been sufficiently addressed in the existing research. We make a preliminary investigation into this field by systematically evaluating the implications of FM-FL integration across these dimensions. We analyze the trade-offs involved, uncover the threats and issues introduced by this integration, and propose a set of criteria and strategies for navigating these challenges. Furthermore, we identify potential research directions for advancing this field, laying a foundation for future development in creating reliable, secure, and equitable FL systems."
                    },
                    {
                        "title": "Queue layouts and nonrepetitive colouring of planar graphs and powers of trees",
                        "abstract": "Dujmovi\\'{c}, Joret, Micek, Morin, Ueckerdt and Wood recently in [Planar graphs have bounded queue-number, Journal of the ACM, Volume 67, Issue 4, Article No.: 22, August 2020] showed some attractive graph product structure theorems for planar graphs. By using the product structure, they proved that planar graphs have bounded queue-number $48$; in [Planar graphs have bounded nonrepetitive chromatic number, Advances in Combinatorics, 5, 11 pp, 2020], the authors proved that planar graphs have bounded nonrepetitive chromatic number $768$.   In this paper, still by using some product structure theorem, we improve the upper bound of queue-number of planar graphs to $27$ and the non-repetitive chromatic number to $320$. We also study powers of trees. We show a graph product structure theorem of the $k$-th power $T^k$ of tree $T$, then use it giving an upper bound of the nonrepetitive~chromatic~number of $T^k$. We also give an asymptotically tight upper bound of the queue-number of $T^k$."
                    },
                    {
                        "title": "A systematical study of the nucleon form factors with the pion cloud effect",
                        "abstract": "The electromagnetic and gravitational form factors of the nucleon are studied simultaneously using a covariant quark-diquark approach, and the pion cloud effect on the form factors is explicitly discussed. In this study, the electromagnetic form factors are first calculated to determine the parameters of our approach. Then, the gravitational form factors of the nucleon are evaluated with the same parameters and the cloud effect is addressed. The mechanical properties, including mass radii, energy densities, and spin distributions, are shown and discussed."
                    },
                    {
                        "title": "Unveiling Backdoor Risks Brought by Foundation Models in Heterogeneous Federated Learning",
                        "abstract": "The foundation models (FMs) have been used to generate synthetic public datasets for the heterogeneous federated learning (HFL) problem where each client uses a unique model architecture. However, the vulnerabilities of integrating FMs, especially against backdoor attacks, are not well-explored in the HFL contexts. In this paper, we introduce a novel backdoor attack mechanism for HFL that circumvents the need for client compromise or ongoing participation in the FL process. This method plants and transfers the backdoor through a generated synthetic public dataset, which could help evade existing backdoor defenses in FL by presenting normal client behaviors. Empirical experiments across different HFL configurations and benchmark datasets demonstrate the effectiveness of our attack compared to traditional client-based attacks. Our findings reveal significant security risks in developing robust FM-assisted HFL systems. This research contributes to enhancing the safety and integrity of FL systems, highlighting the need for advanced security measures in the era of FMs."
                    },
                    {
                        "title": "Vulnerabilities of Foundation Model Integrated Federated Learning Under Adversarial Threats",
                        "abstract": "Federated Learning (FL) addresses critical issues in machine learning related to data privacy and security, yet suffering from data insufficiency and imbalance under certain circumstances. The emergence of foundation models (FMs) offers potential solutions to the limitations of existing FL frameworks, e.g., by generating synthetic data for model initialization. However, due to the inherent safety concerns of FMs, integrating FMs into FL could introduce new risks, which remains largely unexplored. To address this gap, we conduct the first investigation on the vulnerability of FM integrated FL (FM-FL) under adversarial threats. Based on a unified framework of FM-FL, we introduce a novel attack strategy that exploits safety issues of FM to compromise FL client models. Through extensive experiments with well-known models and benchmark datasets in both image and text domains, we reveal the high susceptibility of the FM-FL to this new threat under various FL configurations. Furthermore, we find that existing FL defense strategies offer limited protection against this novel attack approach. This research highlights the critical need for enhanced security measures in FL in the era of FMs."
                    },
                    {
                        "title": "Angular Position Sensor Based on Anisotropic Magnetoresistive and Anomalous Nernst Effect",
                        "abstract": "Magnetic position sensors find extensive applications in various industrial sectors and consumer products. However, measuring angles in the full range of 0{\\deg} to 360{\\deg} in a wide field range using a single magnetic sensor remains a challenge. Here, we propose a magnetic position sensor based on a single Wheatstone bridge structure made from a single ferromagnetic layer. By measuring the anisotropic magnetoresistance (AMR) signal from the bridge and two sets of anomalous Nernst effect (ANE) signals from the transverse ports on two perpendicular Wheatstone bridge arms concurrently, we show that it is possible to achieve 0{\\deg} to 360{\\deg} angle detection using a single bridge sensor. The combined use of AMR and ANE signals allows to achieve a mean angle error in the range of 0.51{\\deg} to 1.05{\\deg} within a field range of 100 Oe to 10,000 Oe."
                    },
                    {
                        "title": "On the existence and regularity of vector solutions for quasilinear systems with linear coupling",
                        "abstract": "We study a class of linearly coupled system of quasilinear equations. Under some assumptions on the nonlinear terms, we establish some results about the existence and regularity of vector solutions for the p-Laplacian systems by using variational methods. In particular, we get two pairs of nontrivial solutions. We also study their different asymptotic behavior of solutions as the coupling parameter tends to zero."
                    },
                    {
                        "title": "Hierarchical Attention Generative Adversarial Networks for Cross-domain Sentiment Classification",
                        "abstract": "Cross-domain sentiment classification (CDSC) is an importance task in domain adaptation and sentiment classification. Due to the domain discrepancy, a sentiment classifier trained on source domain data may not works well on target domain data. In recent years, many researchers have used deep neural network models for cross-domain sentiment classification task, many of which use Gradient Reversal Layer (GRL) to design an adversarial network structure to train a domain-shared sentiment classifier. Different from those methods, we proposed Hierarchical Attention Generative Adversarial Networks (HAGAN) which alternately trains a generator and a discriminator in order to produce a document representation which is sentiment-distinguishable but domain-indistinguishable. Besides, the HAGAN model applies Bidirectional Gated Recurrent Unit (Bi-GRU) to encode the contextual information of a word and a sentence into the document representation. In addition, the HAGAN model use hierarchical attention mechanism to optimize the document representation and automatically capture the pivots and non-pivots. The experiments on Amazon review dataset show the effectiveness of HAGAN."
                    }
                ]
            },
            "cea8e7e3-2690-46b0-b5fb-214cf71dc57a": {
                "pk": "cea8e7e3-2690-46b0-b5fb-214cf71dc57a",
                "name": "Dahua Lin",
                "collaborators": [
                    "Bo Dai",
                    "Yuanjun Xiong",
                    "Zhirong Wu",
                    "Sijie Yan",
                    "Sanja Fidler",
                    "Xiaoou Tang",
                    "Junjie Yan",
                    "Jianqiao Wangni",
                    "Zhiyu Min",
                    "Zhizhong Li"
                ],
                "domain": [
                    "Computer Vision",
                    "Image Captioning",
                    "Deep Learning",
                    "Graph Neural Network"
                ],
                "publications": [
                    {
                        "title": "Contrastive Learning for Image Captioning",
                        "abstract": "Image captioning, a popular topic in computer vision, has achieved substantial progress in recent years. However, the distinctiveness of natural descriptions is often overlooked in previous work. It is closely related to the quality of captions, as distinctive captions are more likely to describe images with their unique aspects. In this work, we propose a new learning method, Contrastive Learning (CL), for image captioning. Specifically, via two constraints formulated on top of a reference model, the proposed method can encourage distinctiveness, while maintaining the overall quality of the generated captions. We tested our method on two challenging datasets, where it improves the baseline model by significant margins. We also showed in our studies that the proposed method is generic and can be used for models with various structures."
                    },
                    {
                        "title": "Learning Sparse Visual Representations with Leaky Capped Norm Regularizers",
                        "abstract": "Sparsity inducing regularization is an important part for learning over-complete visual representations. Despite the popularity of $\\ell_1$ regularization, in this paper, we investigate the usage of non-convex regularizations in this problem. Our contribution consists of three parts. First, we propose the leaky capped norm regularization (LCNR), which allows model weights below a certain threshold to be regularized more strongly as opposed to those above, therefore imposes strong sparsity and only introduces controllable estimation bias. We propose a majorization-minimization algorithm to optimize the joint objective function. Second, our study over monocular 3D shape recovery and neural networks with LCNR outperforms $\\ell_1$ and other non-convex regularizations, achieving state-of-the-art performance and faster convergence. Third, we prove a theoretical global convergence speed on the 3D recovery problem. To the best of our knowledge, this is the first convergence analysis of the 3D recovery problem."
                    },
                    {
                        "title": "Probabilistic Ensemble of Collaborative Filters",
                        "abstract": "Collaborative filtering is an important technique for recommendation. Whereas it has been repeatedly shown to be effective in previous work, its performance remains unsatisfactory in many real-world applications, especially those where the items or users are highly diverse. In this paper, we explore an ensemble-based framework to enhance the capability of a recommender in handling diverse data. Specifically, we formulate a probabilistic model which integrates the items, the users, as well as the associations between them into a generative process. On top of this formulation, we further derive a progressive algorithm to construct an ensemble of collaborative filters. In each iteration, a new filter is derived from re-weighted entries and incorporated into the ensemble. It is noteworthy that while the algorithmic procedure of our algorithm is apparently similar to boosting, it is derived from an essentially different formulation and thus differs in several key technical aspects. We tested the proposed method on three large datasets, and observed substantial improvement over the state of the art, including L2Boost, an effective method based on boosting."
                    },
                    {
                        "title": "Integrating Specialized Classifiers Based on Continuous Time Markov Chain",
                        "abstract": "Specialized classifiers, namely those dedicated to a subset of classes, are often adopted in real-world recognition systems. However, integrating such classifiers is nontrivial. Existing methods, e.g. weighted average, usually implicitly assume that all constituents of an ensemble cover the same set of classes. Such methods can produce misleading predictions when used to combine specialized classifiers. This work explores a novel approach. Instead of combining predictions from individual classifiers directly, it first decomposes the predictions into sets of pairwise preferences, treating them as transition channels between classes, and thereon constructs a continuous-time Markov chain, and use the equilibrium distribution of this chain as the final prediction. This way allows us to form a coherent picture over all specialized predictions. On large public datasets, the proposed method obtains considerable improvement compared to mainstream ensemble methods, especially when the classifier coverage is highly unbalanced."
                    },
                    {
                        "title": "Scalable Estimation of Dirichlet Process Mixture Models on Distributed Data",
                        "abstract": "We consider the estimation of Dirichlet Process Mixture Models (DPMMs) in distributed environments, where data are distributed across multiple computing nodes. A key advantage of Bayesian nonparametric models such as DPMMs is that they allow new components to be introduced on the fly as needed. This, however, posts an important challenge to distributed estimation -- how to handle new components efficiently and consistently. To tackle this problem, we propose a new estimation method, which allows new components to be created locally in individual computing nodes. Components corresponding to the same cluster will be identified and merged via a probabilistic consolidation scheme. In this way, we can maintain the consistency of estimation with very low communication cost. Experiments on large real-world data sets show that the proposed method can achieve high scalability in distributed and asynchronous environments without compromising the mixing performance."
                    },
                    {
                        "title": "Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition",
                        "abstract": "Dynamics of human body skeletons convey significant information for human action recognition. Conventional approaches for modeling skeletons usually rely on hand-crafted parts or traversal rules, thus resulting in limited expressive power and difficulties of generalization. In this work, we propose a novel model of dynamic skeletons called Spatial-Temporal Graph Convolutional Networks (ST-GCN), which moves beyond the limitations of previous methods by automatically learning both the spatial and temporal patterns from data. This formulation not only leads to greater expressive power but also stronger generalization capability. On two large datasets, Kinetics and NTU-RGBD, it achieves substantial improvements over mainstream methods."
                    },
                    {
                        "title": "Motion Guided 3D Pose Estimation from Videos",
                        "abstract": "We propose a new loss function, called motion loss, for the problem of monocular 3D Human pose estimation from 2D pose. In computing motion loss, a simple yet effective representation for keypoint motion, called pairwise motion encoding, is introduced. We design a new graph convolutional network architecture, U-shaped GCN (UGCN). It captures both short-term and long-term motion information to fully leverage the additional supervision from the motion loss. We experiment training UGCN with the motion loss on two large scale benchmarks: Human3.6M and MPI-INF-3DHP. Our model surpasses other state-of-the-art models by a large margin. It also demonstrates strong capacity in producing smooth 3D sequences and recovering keypoint motion."
                    },
                    {
                        "title": "MINI: Mining Implicit Novel Instances for Few-Shot Object Detection",
                        "abstract": "Learning from a few training samples is a desirable ability of an object detector, inspiring the explorations of Few-Shot Object Detection (FSOD). Most existing approaches employ a pretrain-transfer paradigm. The model is first pre-trained on base classes with abundant data and then transferred to novel classes with a few annotated samples. Despite the substantial progress, the FSOD performance is still far behind satisfactory. During pre-training, due to the co-occurrence between base and novel classes, the model is learned to treat the co-occurred novel classes as backgrounds. During transferring, given scarce samples of novel classes, the model suffers from learning discriminative features to distinguish novel instances from backgrounds and base classes. To overcome the obstacles, we propose a novel framework, Mining Implicit Novel Instances (MINI), to mine the implicit novel instances as auxiliary training samples, which widely exist in abundant base data but are not annotated. MINI comprises an offline mining mechanism and an online mining mechanism. The offline mining mechanism leverages a self-supervised discriminative model to collaboratively mine implicit novel instances with a trained FSOD network. Taking the mined novel instances as auxiliary training samples, the online mining mechanism takes a teacher-student framework to simultaneously update the FSOD network and the mined implicit novel instances on the fly. Extensive experiments on PASCAL VOC and MS-COCO datasets show MINI achieves new state-of-the-art performance on any shot and split. The significant performance improvements demonstrate the superiority of our method."
                    },
                    {
                        "title": "Generating Multi-Sentence Lingual Descriptions of Indoor Scenes",
                        "abstract": "This paper proposes a novel framework for generating lingual descriptions of indoor scenes. Whereas substantial efforts have been made to tackle this problem, previous approaches focusing primarily on generating a single sentence for each image, which is not sufficient for describing complex scenes. We attempt to go beyond this, by generating coherent descriptions with multiple sentences. Our approach is distinguished from conventional ones in several aspects: (1) a 3D visual parsing system that jointly infers objects, attributes, and relations; (2) a generative grammar learned automatically from training text; and (3) a text generation algorithm that takes into account the coherence among sentences. Experiments on the augmented NYU-v2 dataset show that our framework can generate natural descriptions with substantially higher ROGUE scores compared to those produced by the baseline."
                    },
                    {
                        "title": "Adjustable Bounded Rectifiers: Towards Deep Binary Representations",
                        "abstract": "Binary representation is desirable for its memory efficiency, computation speed and robustness. In this paper, we propose adjustable bounded rectifiers to learn binary representations for deep neural networks. While hard constraining representations across layers to be binary makes training unreasonably difficult, we softly encourage activations to diverge from real values to binary by approximating step functions. Our final representation is completely binary. We test our approach on MNIST, CIFAR10, and ILSVRC2012 dataset, and systematically study the training dynamics of the binarization process. Our approach can binarize the last layer representation without loss of performance and binarize all the layers with reasonably small degradations. The memory space that it saves may allow more sophisticated models to be deployed, thus compensating the loss. To the best of our knowledge, this is the first work to report results on current deep network architectures using complete binary middle representations. Given the learned representations, we find that the firing or inhibition of a binary neuron is usually associated with a meaningful interpretation across different classes. This suggests that the semantic structure of a neural network may be manifested through a guided binarization process."
                    },
                    {
                        "title": "Deep Markov Random Field for Image Modeling",
                        "abstract": "Markov Random Fields (MRFs), a formulation widely used in generative image modeling, have long been plagued by the lack of expressive power. This issue is primarily due to the fact that conventional MRFs formulations tend to use simplistic factors to capture local patterns. In this paper, we move beyond such limitations, and propose a novel MRF model that uses fully-connected neurons to express the complex interactions among pixels. Through theoretical analysis, we reveal an inherent connection between this model and recurrent neural networks, and thereon derive an approximated feed-forward network that couples multiple RNNs along opposite directions. This formulation combines the expressive power of deep neural networks and the cyclic dependency structure of MRF in a unified model, bringing the modeling capability to a new level. The feed-forward approximation also allows it to be efficiently learned from data. Experimental results on a variety of low-level vision tasks show notable improvement over state-of-the-arts."
                    },
                    {
                        "title": "Detecting Visual Relationships with Deep Relational Networks",
                        "abstract": "Relationships among objects play a crucial role in image understanding. Despite the great success of deep learning techniques in recognizing individual objects, reasoning about the relationships among objects remains a challenging task. Previous methods often treat this as a classification problem, considering each type of relationship (e.g. \"ride\") or each distinct visual phrase (e.g. \"person-ride-horse\") as a category. Such approaches are faced with significant difficulties caused by the high diversity of visual appearance for each kind of relationships or the large number of distinct visual phrases. We propose an integrated framework to tackle this problem. At the heart of this framework is the Deep Relational Network, a novel formulation designed specifically for exploiting the statistical dependencies between objects and their relationships. On two large datasets, the proposed method achieves substantial improvement over state-of-the-art."
                    },
                    {
                        "title": "Peephole: Predicting Network Performance Before Training",
                        "abstract": "The quest for performant networks has been a significant force that drives the advancements of deep learning in recent years. While rewarding, improving network design has never been an easy journey. The large design space combined with the tremendous cost required for network training poses a major obstacle to this endeavor. In this work, we propose a new approach to this problem, namely, predicting the performance of a network before training, based on its architecture. Specifically, we develop a unified way to encode individual layers into vectors and bring them together to form an integrated description via LSTM. Taking advantage of the recurrent network's strong expressive power, this method can reliably predict the performances of various network architectures. Our empirical studies showed that it not only achieved accurate predictions but also produced consistent rankings across datasets -- a key desideratum in performance prediction."
                    },
                    {
                        "title": "Accelerated Training for Massive Classification via Dynamic Class Selection",
                        "abstract": "Massive classification, a classification task defined over a vast number of classes (hundreds of thousands or even millions), has become an essential part of many real-world systems, such as face recognition. Existing methods, including the deep networks that achieved remarkable success in recent years, were mostly devised for problems with a moderate number of classes. They would meet with substantial difficulties, e.g. excessive memory demand and computational cost, when applied to massive problems. We present a new method to tackle this problem. This method can efficiently and accurately identify a small number of \"active classes\" for each mini-batch, based on a set of dynamic class hierarchies constructed on the fly. We also develop an adaptive allocation scheme thereon, which leads to a better tradeoff between performance and cost. On several large-scale benchmarks, our method significantly reduces the training cost and memory demand, while maintaining competitive performance."
                    },
                    {
                        "title": "Unsupervised Feature Learning via Non-Parametric Instance-level Discrimination",
                        "abstract": "Neural net classifiers trained on data with annotated class labels can also capture apparent visual similarity among categories without being directed to do so. We study whether this observation can be extended beyond the conventional domain of supervised learning: Can we learn a good feature representation that captures apparent similarity among instances, instead of classes, by merely asking the feature to be discriminative of individual instances? We formulate this intuition as a non-parametric classification problem at the instance-level, and use noise-contrastive estimation to tackle the computational challenges imposed by the large number of instance classes. Our experimental results demonstrate that, under unsupervised learning settings, our method surpasses the state-of-the-art on ImageNet classification by a large margin. Our method is also remarkable for consistently improving test performance with more training data and better network architectures. By fine-tuning the learned feature, we further obtain competitive results for semi-supervised learning and object detection tasks. Our non-parametric model is highly compact: With 128 features per image, our method requires only 600MB storage for a million images, enabling fast nearest neighbour retrieval at the run time."
                    },
                    {
                        "title": "A Neural Compositional Paradigm for Image Captioning",
                        "abstract": "Mainstream captioning models often follow a sequential structure to generate captions, leading to issues such as introduction of irrelevant semantics, lack of diversity in the generated captions, and inadequate generalization performance. In this paper, we present an alternative paradigm for image captioning, which factorizes the captioning procedure into two stages: (1) extracting an explicit semantic representation from the given image; and (2) constructing the caption based on a recursive compositional procedure in a bottom-up manner. Compared to conventional ones, our paradigm better preserves the semantic content through an explicit factorization of semantics and syntax. By using the compositional generation procedure, caption construction follows a recursive structure, which naturally fits the properties of human language. Moreover, the proposed compositional procedure requires less data to train, generalizes better, and yields more diverse captions."
                    },
                    {
                        "title": "Recursive Visual Sound Separation Using Minus-Plus Net",
                        "abstract": "Sounds provide rich semantics, complementary to visual data, for many tasks. However, in practice, sounds from multiple sources are often mixed together. In this paper we propose a novel framework, referred to as MinusPlus Network (MP-Net), for the task of visual sound separation. MP-Net separates sounds recursively in the order of average energy, removing the separated sound from the mixture at the end of each prediction, until the mixture becomes empty or contains only noise. In this way, MP-Net could be applied to sound mixtures with arbitrary numbers and types of sounds. Moreover, while MP-Net keeps removing sounds with large energy from the mixture, sounds with small energy could emerge and become clearer, so that the separation is more accurate. Compared to previous methods, MP-Net obtains state-of-the-art results on two large scale datasets, across mixtures with different types and numbers of sounds."
                    },
                    {
                        "title": "Low-Latency Video Semantic Segmentation",
                        "abstract": "Recent years have seen remarkable progress in semantic segmentation. Yet, it remains a challenging task to apply segmentation techniques to video-based applications. Specifically, the high throughput of video streams, the sheer cost of running fully convolutional networks, together with the low-latency requirements in many real-world applications, e.g. autonomous driving, present a significant challenge to the design of the video segmentation framework. To tackle this combined challenge, we develop a framework for video semantic segmentation, which incorporates two novel components: (1) a feature propagation module that adaptively fuses features over time via spatially variant convolution, thus reducing the cost of per-frame computation; and (2) an adaptive scheduler that dynamically allocate computation based on accuracy prediction. Both components work together to ensure low latency while maintaining high segmentation quality. On both Cityscapes and CamVid, the proposed framework obtained competitive performance compared to the state of the art, while substantially reducing the latency, from 360 ms to 119 ms."
                    },
                    {
                        "title": "Unifying Identification and Context Learning for Person Recognition",
                        "abstract": "Despite the great success of face recognition techniques, recognizing persons under unconstrained settings remains challenging. Issues like profile views, unfavorable lighting, and occlusions can cause substantial difficulties. Previous works have attempted to tackle this problem by exploiting the context, e.g. clothes and social relations. While showing promising improvement, they are usually limited in two important aspects, relying on simple heuristics to combine different cues and separating the construction of context from people identities. In this work, we aim to move beyond such limitations and propose a new framework to leverage context for person recognition. In particular, we propose a Region Attention Network, which is learned to adaptively combine visual cues with instance-dependent weights. We also develop a unified formulation, where the social contexts are learned along with the reasoning of people identities. These models substantially improve the robustness when working with the complex contextual relations in unconstrained environments. On two large datasets, PIPA and Cast In Movies (CIM), a new dataset proposed in this work, our method consistently achieves state-of-the-art performance under multiple evaluation policies."
                    },
                    {
                        "title": "Rethinking the Form of Latent States in Image Captioning",
                        "abstract": "RNNs and their variants have been widely adopted for image captioning. In RNNs, the production of a caption is driven by a sequence of latent states. Existing captioning models usually represent latent states as vectors, taking this practice for granted. We rethink this choice and study an alternative formulation, namely using two-dimensional maps to encode latent states. This is motivated by the curiosity about a question: how the spatial structures in the latent states affect the resultant captions? Our study on MSCOCO and Flickr30k leads to two significant observations. First, the formulation with 2D states is generally more effective in captioning, consistently achieving higher performance with comparable parameter sizes. Second, 2D states preserve spatial locality. Taking advantage of this, we visually reveal the internal dynamics in the process of caption generation, as well as the connections between input visual domain and output linguistic domain."
                    }
                ]
            },
            "51f3b1f4-e3a1-432e-a64d-fac8b5d16dc4": {
                "pk": "51f3b1f4-e3a1-432e-a64d-fac8b5d16dc4",
                "name": "Kai Chen",
                "collaborators": [
                    "Kai Niu",
                    "Jiaru Lin",
                    "Jia-Ru Lin",
                    "Kai Liu",
                    "Ruohui Wang",
                    "Jianfei Gao",
                    "Wenbing Tao",
                    "Mathias Seuret",
                    "Yuqian Zhang",
                    "Pavan Hosur"
                ],
                "domain": [
                    "Polar Codes",
                    "Neural Networks",
                    "Communication Systems",
                    "Semi-Supervised Learning"
                ],
                "publications": [
                    {
                        "title": "Chains of Unusual Excellent Local Rings",
                        "abstract": "Let (T,M) be a complete local domain containing the integers. Let p1 \\subseteq p2 \\subseteq ... \\subseteq pn be a chain of nonmaximal prime ideals T such that T_pn is a regular local ring. We construct a chain of excellent local domains An \\subseteq A1 such that for each i, the completion of Ai is T, the generic formal fiber of Ai is local with maximal ideal pi, and if I is a nonzero ideal of Ai then Ai/I is complete. Consequently, if in addition T is a UFD, then we can construct a chain of excellent local UFDs satisfying the same conditions."
                    },
                    {
                        "title": "Polar Coded Modulation with Optimal Constellation Labeling",
                        "abstract": "A practical $2^m$-ary polar coded modulation (PCM) scheme with optimal constellation labeling is proposed. To efficiently find the optimal labeling rule, the search space is reduced by exploiting the symmetry properties of the channels. Simulation results show that the proposed PCM scheme can outperform the bit-interleaved turbo coded modulation scheme used in the WCDMA (Wideband Code Division Multiple Access) mobile communication systems by up to 1.5dB."
                    },
                    {
                        "title": "A Hybrid ARQ Scheme Based on Polar Codes",
                        "abstract": "A hybrid automatic repeat request (HARQ) scheme based on a novel class of rate-compatible polar (\\mbox{RCP}) codes are proposed. The RCP codes are constructed by performing punctures and repetitions on the conventional polar codes. Simulation results over binary-input additive white Gaussian noise channels (BAWGNCs) show that, using a low-complexity successive cancellation (SC) decoder, the proposed HARQ scheme performs as well as the existing schemes based on turbo codes and low-density parity-check (LDPC) codes. The proposed transmission scheme is only about 1.0-1.5dB away from the channel capacity with the information block length of 1024 bits."
                    },
                    {
                        "title": "Improved Successive Cancellation Decoding of Polar Codes",
                        "abstract": "As improved versions of successive cancellation (SC) decoding algorithm, successive cancellation list (SCL) decoding and successive cancellation stack (SCS) decoding are used to improve the finite-length performance of polar codes. Unified descriptions of SC, SCL and SCS decoding algorithms are given as path searching procedures on the code tree of polar codes. Combining the ideas of SCL and SCS, a new decoding algorithm named successive cancellation hybrid (SCH) is proposed, which can achieve a better trade-off between computational complexity and space complexity. Further, to reduce the complexity, a pruning technique is proposed to avoid unnecessary path searching operations. Performance and complexity analysis based on simulations show that, with proper configurations, all the three improved successive cancellation (ISC) decoding algorithms can have a performance very close to that of maximum-likelihood (ML) decoding with acceptable complexity. Moreover, with the help of the proposed pruning technique, the complexities of ISC decoders can be very close to that of SC decoder in the moderate and high signal-to-noise ratio (SNR) regime."
                    },
                    {
                        "title": "Space-Time Polar Coded Modulation",
                        "abstract": "The polar codes are proven to be capacity-achieving and are shown to have equivalent or even better finite-length performance than the turbo/LDPC codes under some improved decoding algorithms over the additive white Gaussian noise (AWGN) channels. Polar coding is based on the so-called channel polarization phenomenon induced by a transform over the underlying binary-input channel. The channel polarization is found to be universal in many signal processing problems and has been applied to the coded modulation schemes. In this paper, the channel polarization is further extended to the multiple antenna transmission following a multilevel coding principle. The multiple-input multile-output (MIMO) channel under quadrature amplitude modulation (QAM) are transformed into a series of synthesized binary-input channels under a three-stage channel transform. Based on this generalized channel polarization, the proposed space-time polar coded modulation (STPCM) scheme allows a joint optimization of the binary polar coding, modulation and MIMO transmission. In addition, a practical solution of polar code construction over the fading channels is also provided, where the fading channels are approximated by an AWGN channel which shares the same capacity with the original. The simulations over the MIMO channel with uncorrelated Rayleigh fast fading show that the proposed STPCM scheme can outperform the bit-interleaved turbo coded scheme in all the simulated cases, where the latter is adopted in many existing communication systems."
                    },
                    {
                        "title": "Low-Complexity Sphere Decoding of Polar Codes based on Optimum Path Metric",
                        "abstract": "Sphere decoding (SD) of polar codes is an efficient method to achieve the error performance of maximum likelihood (ML) decoding. But the complexity of the conventional sphere decoder is still high, where the candidates in a target sphere are enumerated and the radius is decreased gradually until no available candidate is in the sphere. In order to reduce the complexity of SD, a stack SD (SSD) algorithm with an efficient enumeration is proposed in this paper. Based on a novel path metric, SSD can effectively narrow the search range when enumerating the candidates within a sphere. The proposed metric follows an exact ML rule and takes the full usage of the whole received sequence. Furthermore, another very simple metric is provided as an approximation of the ML metric in the high signal-to-noise ratio regime. For short polar codes, simulation results over the additive white Gaussian noise channels show that the complexity of SSD based on the proposed metrics is up to 100 times lower than that of the conventional SD."
                    },
                    {
                        "title": "Differentiable Model Scaling using Differentiable Topk",
                        "abstract": "Over the past few years, as large language models have ushered in an era of intelligence emergence, there has been an intensified focus on scaling networks. Currently, many network architectures are designed manually, often resulting in sub-optimal configurations. Although Neural Architecture Search (NAS) methods have been proposed to automate this process, they suffer from low search efficiency. This study introduces Differentiable Model Scaling (DMS), increasing the efficiency for searching optimal width and depth in networks. DMS can model both width and depth in a direct and fully differentiable way, making it easy to optimize. We have evaluated our DMS across diverse tasks, ranging from vision tasks to NLP tasks and various network architectures, including CNNs and Transformers. Results consistently indicate that our DMS can find improved structures and outperforms state-of-the-art NAS methods. Specifically, for image classification on ImageNet, our DMS improves the top-1 accuracy of EfficientNet-B0 and Deit-Tiny by 1.4% and 0.6%, respectively, and outperforms the state-of-the-art zero-shot NAS method, ZiCo, by 1.3% while requiring only 0.4 GPU days for searching. For object detection on COCO, DMS improves the mAP of Yolo-v8-n by 2.0%. For language modeling, our pruned Llama-7B outperforms the prior method with lower perplexity and higher zero-shot classification accuracy. We will release our code in the future."
                    },
                    {
                        "title": "Once for All: a Two-flow Convolutional Neural Network for Visual Tracking",
                        "abstract": "One of the main challenges of visual object tracking comes from the arbitrary appearance of objects. Most existing algorithms try to resolve this problem as an object-specific task, i.e., the model is trained to regenerate or classify a specific object. As a result, the model need to be initialized and retrained for different objects. In this paper, we propose a more generic approach utilizing a novel two-flow convolutional neural network (named YCNN). The YCNN takes two inputs (one is object image patch, the other is search image patch), then outputs a response map which predicts how likely the object appears in a specific location. Unlike those object-specific approach, the YCNN is trained to measure the similarity between two image patches. Thus it will not be confined to any specific object. Furthermore the network can be end-to-end trained to extract both shallow and deep convolutional features which are dedicated for visual tracking. And once properly trained, the YCNN can be applied to track all kinds of objects without further training and updating. Benefiting from the once-for-all model, our algorithm is able to run at a very high speed of 45 frames-per-second. The experiments on 51 sequences also show that our algorithm achieves an outstanding performance."
                    },
                    {
                        "title": "Convolutional Neural Networks for Page Segmentation of Historical Document Images",
                        "abstract": "This paper presents a Convolutional Neural Network (CNN) based page segmentation method for handwritten historical document images. We consider page segmentation as a pixel labeling problem, i.e., each pixel is classified as one of the predefined classes. Traditional methods in this area rely on carefully hand-crafted features or large amounts of prior knowledge. In contrast, we propose to learn features from raw image pixels using a CNN. While many researchers focus on developing deep CNN architectures to solve different problems, we train a simple CNN with only one convolution layer. We show that the simple architecture achieves competitive results against other deep architectures on different public datasets. Experiments also demonstrate the effectiveness and superiority of the proposed method compared to previous methods."
                    },
                    {
                        "title": "Enhancing efficiency and robustness in high-dimensional linear regression with additional unlabeled data",
                        "abstract": "In semi-supervised learning, the prevailing understanding suggests that observing additional unlabeled samples improves estimation accuracy for linear parameters only in the case of model misspecification. This paper challenges this notion, demonstrating its inaccuracy in high dimensions. Initially focusing on a dense scenario, we introduce robust semi-supervised estimators for the regression coefficient without relying on sparse structures in the population slope. Even when the true underlying model is linear, we show that leveraging information from large-scale unlabeled data improves both estimation accuracy and inference robustness. Moreover, we propose semi-supervised methods with further enhanced efficiency in scenarios with a sparse linear slope. Diverging from the standard semi-supervised literature, we also allow for covariate shift. The performance of the proposed methods is illustrated through extensive numerical studies, including simulations and a real-data application to the AIDS Clinical Trials Group Protocol 175 (ACTG175)."
                    },
                    {
                        "title": "Surface Luttinger surfaces and Luttinger-Lifshitz transitions in topological band structures",
                        "abstract": "The standard paradigm of topological phases posits that two phases with the same symmetries are necessarily separated by a bulk phase transition, while breaking the symmetry unlocks a path in parameter space that allows the phases to be connected adiabatically. Moreover, if the symmetry is broken only on the boundary, topological surface states are generically gapped and single-particle surface properties are expected to be blind to distinction between the two phases. In this work, we prove this last expectation incorrect. We first reveal that the single-particle surface Green's function contains zeros or ``Luttinger surfaces'' that respect the same bulk-boundary correspondence as the well-known topological surface states. Remarkably, the Luttinger surfaces survive symmetry-breaking perturbations that destroy the surface states. Thus, a bulk topological phase transition in the presence of surface symmetry breaking causes a reconstruction of Luttinger surfaces on the surface, which we refer to as a surface Luttinger-Lifshitz transition. At non-zero temperatures, the Luttinger surfaces contribute negatively to an effective surface specific heat, and a Luttinger-Lifshitz transition manifests as a discontinuity in this specific heat."
                    },
                    {
                        "title": "Polar Coded HARQ Scheme with Chase Combining",
                        "abstract": "A hybrid automatic repeat request scheme with Chase combing (HARQ-CC) of polar codes is proposed. The existing analysis tools of the underlying rate-compatible punctured polar (RCPP) codes for additive white Gaussian noise (AWGN) channels are extended to Rayleigh fading channels. Then, an approximation bound of the throughput efficiency for the polar coded HARQ-CC scheme is derived. Utilizing this bound, the parameter configurations of the proposed scheme can be optimized. Simulation results show that, the proposed HARQ-CC scheme under a low-complexity SC decoding is only about $1.0$dB away from the existing schemes with incremental redundancy (\\mbox{HARQ-IR}). Compared with the polar coded \\mbox{HARQ-IR} scheme, the proposed HARQ-CC scheme requires less retransmissions and has the advantage of good compatibility to other communication techniques."
                    }
                ]
            }
        }
    },
    "2305.11512": {
        "paper_data": {
            "title": "Enriching Disentanglement: From Logical Definitions to Quantitative Metrics",
            "url": "http://arxiv.org/abs/2305.11512v2",
            "arxiv_id": "2305.11512",
            "authors": [
                "Yivan Zhang",
                "Masashi Sugiyama"
            ],
            "abstract": "Disentangling the explanatory factors in complex data is a promising approach for generalizable and data-efficient representation learning. While a variety of quantitative metrics for learning and evaluating disentangled representations have been proposed, it remains unclear what properties these metrics truly quantify. In this work, we establish a theoretical connection between logical definitions of disentanglement and quantitative metrics using topos theory and enriched category theory. We introduce a systematic approach for converting a first-order predicate into a real-valued quantity by replacing (i) equality with a strict premetric, (ii) the Heyting algebra of binary truth values with a quantale of continuous values, and (iii) quantifiers with aggregators. The metrics induced by logical definitions have strong theoretical guarantees, and some of them are easily differentiable and can be used as learning objectives directly. Finally, we empirically demonstrate the effectiveness of the proposed metrics by isolating different aspects of disentangled representations.",
            "introduction": "   1 Introduction  In supervised learning, we use a real-valued cost function \u2113:Y\u00d7Y\u2192\u211d\u22650:\u2113\u2192\ud835\udc4c\ud835\udc4csubscript\u211dabsent0\\ell:Y\\times Y\\to\\mathbb{R}_{\\geq 0}roman_\u2113 : italic_Y \u00d7 italic_Y \u2192 blackboard_R start_POSTSUBSCRIPT \u2265 0 end_POSTSUBSCRIPT to measure how close an output f\u2062(x)\ud835\udc53\ud835\udc65f(x)italic_f ( italic_x ) of a function f:X\u2192Y:\ud835\udc53\u2192\ud835\udc4b\ud835\udc4cf:X\\to Yitalic_f : italic_X \u2192 italic_Y is to a target label y\ud835\udc66yitalic_y, i.e., \u2113\u2062(f\u2062(x),y)\u2113\ud835\udc53\ud835\udc65\ud835\udc66\\ell(f(x),y)roman_\u2113 ( italic_f ( italic_x ) , italic_y ), to quantify the cost of inaccurate prediction. Then, we can use the total cost over a collection of input-output pairs to measure the performance of this function. From a functional perspective, this construction induces a quantitative metric L:[X,Y]\u00d7[X,Y]\u2192\u211d\u22650:\ud835\udc3f\u2192\ud835\udc4b\ud835\udc4c\ud835\udc4b\ud835\udc4csubscript\u211dabsent0L:[X,Y]\\times[X,Y]\\to\\mathbb{R}_{\\geq 0}italic_L : [ italic_X , italic_Y ] \u00d7 [ italic_X , italic_Y ] \u2192 blackboard_R start_POSTSUBSCRIPT \u2265 0 end_POSTSUBSCRIPT between functions:111[X,Y]\ud835\udc4b\ud835\udc4c[X,Y][ italic_X , italic_Y ] denotes the set of all functions from a set X\ud835\udc4bXitalic_X to a set Y\ud835\udc4cYitalic_Y.    L\u2062(f,g):=\u2211x\u2208X\u2113\u2062(f\u2062(x),g\u2062(x)),assign\ud835\udc3f\ud835\udc53\ud835\udc54subscript\ud835\udc65\ud835\udc4b\u2113\ud835\udc53\ud835\udc65\ud835\udc54\ud835\udc65L(f,g)\\vcentcolon=\\sum_{x\\in X}\\ell(f(x),g(x)),italic_L ( italic_f , italic_g ) := \u2211 start_POSTSUBSCRIPT italic_x \u2208 italic_X end_POSTSUBSCRIPT roman_\u2113 ( italic_f ( italic_x ) , italic_g ( italic_x ) ) ,  (1)   where g:X\u2192Y:\ud835\udc54\u2192\ud835\udc4b\ud835\udc4cg:X\\to Yitalic_g : italic_X \u2192 italic_Y is a \u201cground-truth function\u201d that maps each input x\ud835\udc65xitalic_x to its target label y\ud835\udc66yitalic_y. This metric can be used as both learning objective and evaluation metric for the learning model f:X\u2192Y:\ud835\udc53\u2192\ud835\udc4b\ud835\udc4cf:X\\to Yitalic_f : italic_X \u2192 italic_Y. What does L\u2062(f,g)\ud835\udc3f\ud835\udc53\ud835\udc54L(f,g)italic_L ( italic_f , italic_g ) quantify? It quantifies the extent to which two functions f\ud835\udc53fitalic_f and g\ud835\udc54gitalic_g are equal:    (f=[X,Y]g):=\u2200x\u2208X.(f\u2062(x)=Yg\u2062(x)).formulae-sequenceassignsubscript\ud835\udc4b\ud835\udc4c\ud835\udc53\ud835\udc54for-all\ud835\udc65\ud835\udc4bsubscript\ud835\udc4c\ud835\udc53\ud835\udc65\ud835\udc54\ud835\udc65(f=_{[X,Y]}g)\\vcentcolon=\\forall x\\in X.\\;(f(x)=_{Y}g(x)).( italic_f = start_POSTSUBSCRIPT [ italic_X , italic_Y ] end_POSTSUBSCRIPT italic_g ) := \u2200 italic_x \u2208 italic_X . ( italic_f ( italic_x ) = start_POSTSUBSCRIPT italic_Y end_POSTSUBSCRIPT italic_g ( italic_x ) ) .  (2)   Considering the equality as a predicate, we can observe a parallel between (i) =Y:Y\u00d7Y\u2192{\u22a5,\u22a4}=_{Y}:Y\\times Y\\to\\{\\bot,\\top\\}= start_POSTSUBSCRIPT italic_Y end_POSTSUBSCRIPT : italic_Y \u00d7 italic_Y \u2192 { \u22a5 , \u22a4 } and \u2113:Y\u00d7Y\u2192\u211d\u22650:\u2113\u2192\ud835\udc4c\ud835\udc4csubscript\u211dabsent0\\ell:Y\\times Y\\to\\mathbb{R}_{\\geq 0}roman_\u2113 : italic_Y \u00d7 italic_Y \u2192 blackboard_R start_POSTSUBSCRIPT \u2265 0 end_POSTSUBSCRIPT, (ii) \u2200x\u2208Xfor-all\ud835\udc65\ud835\udc4b\\forall x\\in X\u2200 italic_x \u2208 italic_X and \u2211x\u2208Xsubscript\ud835\udc65\ud835\udc4b\\sum_{x\\in X}\u2211 start_POSTSUBSCRIPT italic_x \u2208 italic_X end_POSTSUBSCRIPT, and (iii) =[X,Y]:[X,Y]\u00d7[X,Y]\u2192{\u22a5,\u22a4}=_{[X,Y]}:[X,Y]\\times[X,Y]\\to\\{\\bot,\\top\\}= start_POSTSUBSCRIPT [ italic_X , italic_Y ] end_POSTSUBSCRIPT : [ italic_X , italic_Y ] \u00d7 [ italic_X , italic_Y ] \u2192 { \u22a5 , \u22a4 } and L:[X,Y]\u00d7[X,Y]\u2192\u211d\u22650:\ud835\udc3f\u2192\ud835\udc4b\ud835\udc4c\ud835\udc4b\ud835\udc4csubscript\u211dabsent0L:[X,Y]\\times[X,Y]\\to\\mathbb{R}_{\\geq 0}italic_L : [ italic_X , italic_Y ] \u00d7 [ italic_X , italic_Y ] \u2192 blackboard_R start_POSTSUBSCRIPT \u2265 0 end_POSTSUBSCRIPT. We can prove that L\u2062(f,g)=0\ud835\udc3f\ud835\udc53\ud835\udc540L(f,g)=0italic_L ( italic_f , italic_g ) = 0 if and only if f=[X,Y]gsubscript\ud835\udc4b\ud835\udc4c\ud835\udc53\ud835\udc54f=_{[X,Y]}gitalic_f = start_POSTSUBSCRIPT [ italic_X , italic_Y ] end_POSTSUBSCRIPT italic_g, and if the function f\ud835\udc53fitalic_f is restricted to a subset of all functions [X,Y]\ud835\udc4b\ud835\udc4c[X,Y][ italic_X , italic_Y ], we can minimize L\u2062(f,g)\ud835\udc3f\ud835\udc53\ud835\udc54L(f,g)italic_L ( italic_f , italic_g ), possibly by gradient descent, to find an optimal function f\u2217:X\u2192Y:superscript\ud835\udc53\u2192\ud835\udc4b\ud835\udc4cf^{*}:X\\to Yitalic_f start_POSTSUPERSCRIPT \u2217 end_POSTSUPERSCRIPT : italic_X \u2192 italic_Y that is closest to the ground truth g\ud835\udc54gitalic_g.   We would like to ask: can we measure and optimize other properties like this?    1.1 Can we measure injectivity?  As the learning problem becomes more complex, measuring the quality of a learning model becomes a non-trivial task. The performance cannot always be measured by how close a learning model is to a fixed ground truth.   For example, in representation learning (Bengio et\u00a0al., 2013), we may want a function to preserve informative factors in data (Eastwood and Williams, 2018). This criterion",
            "references": [
                {
                    "title": "Position: Categorical Deep Learning is an Algebraic Theory of All Architectures",
                    "abstract": "We present our position on the elusive quest for a general-purpose framework for specifying and studying deep learning architectures. Our opinion is that the key attempts made so far lack a coherent bridge between specifying constraints which models must satisfy and specifying their implementations. Focusing on building a such a bridge, we propose to apply category theory -- precisely, the universal algebra of monads valued in a 2-category of parametric maps -- as a single theory elegantly subsuming both of these flavours of neural network design. To defend our position, we show how this theory recovers constraints induced by geometric deep learning, as well as implementations of many architectures drawn from the diverse landscape of neural networks, such as RNNs. We also illustrate how the theory naturally encodes many standard constructs in computer science and automata theory."
                },
                {
                    "title": "Ordered semirings and subadditive morphisms",
                    "abstract": "An ordered semiring is a commutative semiring equipped with a compatible preorder. Ordered semirings generalise both distributive lattices and commutative rings, and provide a convenient framework to unify certain aspects of lattice theory and ring theory. The ideals of an ordered semiring $A$ form a commutative integral quantale $\\mathrm{Idl}(A)$, and similarly, the radical ideals of $A$ form a (spatial) frame $\\mathrm{Rad}(A)$. We characterise $\\mathrm{Idl}$ and $\\mathrm{Rad}$ as the left adjoints of the (non-full) inclusion functors from the categories of commutative integral quantales and of frames, respectively, to that of ordered semirings and subadditive morphisms between them. The (sober) topological space $\\mathrm{pt}(\\mathrm{Rad}(A))$ corresponding to $\\mathrm{Rad}(A)$ is homeomorphic to the space $\\mathrm{Spec}(A)$ of prime ideals of $A$."
                },
                {
                    "title": "Going Beyond Neural Network Feature Similarity: The Network Feature Complexity and Its Interpretation Using Category Theory",
                    "abstract": "The behavior of neural networks still remains opaque, and a recently widely noted phenomenon is that networks often achieve similar performance when initialized with different random parameters. This phenomenon has attracted significant attention in measuring the similarity between features learned by distinct networks. However, feature similarity could be vague in describing the same feature since equivalent features hardly exist. In this paper, we expand the concept of equivalent feature and provide the definition of what we call functionally equivalent features. These features produce equivalent output under certain transformations. Using this definition, we aim to derive a more intrinsic metric for the so-called feature complexity regarding the redundancy of features learned by a neural network at each layer. We offer a formal interpretation of our approach through the lens of category theory, a well-developed area in mathematics. To quantify the feature complexity, we further propose an efficient algorithm named Iterative Feature Merging. Our experimental results validate our ideas and theories from various perspectives. We empirically demonstrate that the functionally equivalence widely exists among different features learned by the same neural network and we could reduce the number of parameters of the network without affecting the performance.The IFM shows great potential as a data-agnostic model prune method. We have also drawn several interesting empirical findings regarding the defined feature complexity."
                },
                {
                    "title": "Byzantine Machine Learning: A Primer",
                    "abstract": "The problem of Byzantine resilience in distributed machine learning, a.k.a. Byzantine machine learning, consists of designing distributed algorithms that can train an accurate model despite the presence of Byzantine nodes\u2014that is, nodes with corrupt data or machines that can misbehave arbitrarily. By now, many solutions to this important problem have been proposed, most of which build upon the classical stochastic gradient descent scheme. Yet, the literature lacks a unified structure of this emerging field. Consequently, the general understanding on the principles of Byzantine machine learning remains poor. This article addresses this issue by presenting a primer on Byzantine machine learning. In particular, we introduce three pillars of Byzantine machine learning, namely the concepts of breakdown point, robustness, and gradient complexity, to curate the efficacy of a solution. The introduced systematization enables us to (i) bring forth the merits and limitations of the state-of-the-art solutions, and (ii) pave a clear path for future advancements in this field."
                },
                {
                    "title": "Geometric Algebra Transformer",
                    "abstract": "Problems involving geometric data arise in physics, chemistry, robotics, computer vision, and many other fields. Such data can take numerous forms, for instance points, direction vectors, translations, or rotations, but to date there is no single architecture that can be applied to such a wide variety of geometric types while respecting their symmetries. In this paper we introduce the Geometric Algebra Transformer (GATr), a general-purpose architecture for geometric data. GATr represents inputs, outputs, and hidden states in the projective geometric (or Clifford) algebra, which offers an efficient 16-dimensional vector-space representation of common geometric objects as well as operators acting on them. GATr is equivariant with respect to E(3), the symmetry group of 3D Euclidean space. As a Transformer, GATr is versatile, efficient, and scalable. We demonstrate GATr in problems from n-body modeling to wall-shear-stress estimation on large arterial meshes to robotic motion planning. GATr consistently outperforms both non-geometric and equivariant baselines in terms of error, data efficiency, and scalability."
                },
                {
                    "title": "A Category-theoretical Meta-analysis of Definitions of Disentanglement",
                    "abstract": "Disentangling the factors of variation in data is a fundamental concept in machine learning and has been studied in various ways by different researchers, leading to a multitude of definitions. Despite the numerous empirical studies, more theoretical research is needed to fully understand the defining properties of disentanglement and how different definitions relate to each other. This paper presents a meta-analysis of existing definitions of disentanglement, using category theory as a unifying and rigorous framework. We propose that the concepts of the cartesian and monoidal products should serve as the core of disentanglement. With these core concepts, we show the similarities and crucial differences in dealing with (i) functions, (ii) equivariant maps, (iii) relations, and (iv) stochastic maps. Overall, our meta-analysis deepens our understanding of disentanglement and its various formulations and can help researchers navigate different definitions and choose the most appropriate one for their specific context."
                },
                {
                    "title": "Categorification of Group Equivariant Neural Networks",
                    "abstract": "We present a novel application of category theory for deep learning. We show how category theory can be used to understand and work with the linear layer functions of group equivariant neural networks whose layers are some tensor power space of $\\mathbb{R}^{n}$ for the groups $S_n$, $O(n)$, $Sp(n)$, and $SO(n)$. By using category theoretic constructions, we build a richer structure that is not seen in the original formulation of these neural networks, leading to new insights. In particular, we outline the development of an algorithm for quickly computing the result of a vector that is passed through an equivariant, linear layer for each group in question. The success of our approach suggests that category theory could be beneficial for other areas of deep learning."
                },
                {
                    "title": "Correcting Flaws in Common Disentanglement Metrics",
                    "abstract": "Recent years have seen growing interest in learning disentangled representations, in which distinct features, such as size or shape, are represented by distinct neurons. Quantifying the extent to which a given representation is disentangled is not straightforward; multiple metrics have been proposed. In this paper, we identify two failings of existing metrics, which mean they can assign a high score to a model which is still entangled, and we propose two new metrics, which redress these problems. We then consider the task of compositional generalization. Unlike prior works, we treat this as a classification problem, which allows us to use it to measure the disentanglement ability of the encoder, without depending on the decoder. We show that performance on this task is (a) generally quite poor, (b) correlated with most disentanglement metrics, and (c) most strongly correlated with our newly proposed metrics."
                },
                {
                    "title": "Propositional Logics for the Lawvere Quantale",
                    "abstract": "Lawvere showed that generalised metric spaces are categories enriched over\n$[0, \\infty]$, the quantale of the positive extended reals. The statement of\nenrichment is a quantitative analogue of being a preorder. Towards seeking a\nlogic for quantitative metric reasoning, we investigate three\n$[0,\\infty]$-valued propositional logics over the Lawvere quantale. The basic\nlogical connectives shared by all three logics are those that can be\ninterpreted in any quantale, viz finite conjunctions and disjunctions, tensor\n(addition for the Lawvere quantale) and linear implication (here a truncated\nsubtraction); to these we add, in turn, the constant $1$ to express integer\nvalues, and scalar multiplication by a non-negative real to express general\naffine combinations. Quantitative equational logic can be interpreted in the\nthird logic if we allow inference systems instead of axiomatic systems. For\neach of these logics we develop a natural deduction system which we prove to be\ndecidably complete w.r.t. the quantale-valued semantics. The heart of the\ncompleteness proof makes use of the Motzkin transposition theorem. Consistency\nis also decidable; the proof makes use of Fourier-Motzkin elimination of linear\ninequalities. Strong completeness does not hold in general, even (as is known)\nfor theories over finitely-many propositional variables; indeed even an\napproximate form of strong completeness in the sense of Pavelka or Ben Yaacov\n-- provability up to arbitrary precision -- does not hold. However, we can show\nit for theories axiomatized by a (not necessarily finite) set of judgements in\nnormal form over a finite set of propositional variables when we restrict to\nmodels that do not map variables to $\\infty$; the proof uses Hurwicz's general\nform of the Farkas' Lemma."
                },
                {
                    "title": "Equivariant Architectures for Learning in Deep Weight Spaces",
                    "abstract": "Designing machine learning architectures for processing neural networks in their raw weight matrix form is a newly introduced research direction. Unfortunately, the unique symmetry structure of deep weight spaces makes this design very challenging. If successful, such architectures would be capable of performing a wide range of intriguing tasks, from adapting a pre-trained network to a new domain to editing objects represented as functions (INRs or NeRFs). As a first step towards this goal, we present here a novel network architecture for learning in deep weight spaces. It takes as input a concatenation of weights and biases of a pre-trained MLP and processes it using a composition of layers that are equivariant to the natural permutation symmetry of the MLP's weights: Changing the order of neurons in intermediate layers of the MLP does not affect the function it represents. We provide a full characterization of all affine equivariant and invariant layers for these symmetries and show how these layers can be implemented using three basic operations: pooling, broadcasting, and fully connected layers applied to the input in an appropriate manner. We demonstrate the effectiveness of our architecture and its advantages over natural baselines in a variety of learning tasks."
                },
                {
                    "title": "Markov Categories and Entropy",
                    "abstract": "Markov categories are a novel framework to describe and treat problems in probability and information theory. In this work we combine the categorical formalism with the traditional quantitative notions of entropy, mutual information, and data processing inequalities. We show that several quantitative aspects of information theory can be captured by an enriched version of Markov categories, where the spaces of morphisms are equipped with a divergence or even a metric. Following standard practices of information theory, we get measures of mutual information by quantifying, with a chosen divergence, how far a joint source is from displaying independence of its components. More strikingly, Markov categories give a notion of determinism for sources and channels, and we can define entropy exactly by quantifying how far a source or channel is from being deterministic. This recovers Shannon and R\u00e9nyi entropies, as well as the Gini-Simpson index used in ecology to quantify diversity, and it can be used to give a conceptual definition of generalized entropy. No previous knowledge of category theory is assumed."
                },
                {
                    "title": "On the power of foundation models",
                    "abstract": "With infinitely many high-quality data points, infinite computational power, an infinitely large foundation model with a perfect training algorithm and guaranteed zero generalization error on the pretext task, can the model be used for everything? This question cannot be answered by the existing theory of representation, optimization or generalization, because the issues they mainly investigate are assumed to be nonexistent here. In this paper, we show that category theory provides powerful machinery to answer this question. We have proved three results. The first one limits the power of prompt-based learning, saying that the model can solve a downstream task with prompts if and only if the task is representable. The second one says fine tuning does not have this limit, as a foundation model with the minimum required power (up to symmetry) can theoretically solve downstream tasks for the category defined by pretext task, with fine tuning and enough resources. Our final result can be seen as a new type of generalization theorem, showing that the foundation model can generate unseen objects from the target category (e.g., images) using the structural information from the source category (e.g., texts). Along the way, we provide a categorical framework for supervised and self-supervised learning, which might be of independent interest."
                },
                {
                    "title": "Disentanglement of Correlated Factors via Hausdorff Factorized Support",
                    "abstract": "A grand goal in deep learning research is to learn representations capable of generalizing across distribution shifts. Disentanglement is one promising direction aimed at aligning a model's representation with the underlying factors generating the data (e.g. color or background). Existing disentanglement methods, however, rely on an often unrealistic assumption: that factors are statistically independent. In reality, factors (like object color and shape) are correlated. To address this limitation, we consider the use of a relaxed disentanglement criterion -- the Hausdorff Factorized Support (HFS) criterion -- that encourages only pairwise factorized \\emph{support}, rather than a factorial distribution, by minimizing a Hausdorff distance. This allows for arbitrary distributions of the factors over their support, including correlations between them. We show that the use of HFS consistently facilitates disentanglement and recovery of ground-truth factors across a variety of correlation settings and benchmarks, even under severe training correlations and correlation shifts, with in parts over $+60\\%$ in relative improvement over existing disentanglement methods. In addition, we find that leveraging HFS for representation learning can even facilitate transfer to downstream tasks such as classification under distribution shifts. We hope our original approach and positive empirical results inspire further progress on the open problem of robust generalization. Code available at https://github.com/facebookresearch/disentangling-correlated-factors."
                },
                {
                    "title": "Unsupervised Learning of Equivariant Structure from Sequences",
                    "abstract": "In this study, we present meta-sequential prediction (MSP), an unsupervised framework to learn the symmetry from the time sequence of length at least three. Our method leverages the stationary property (e.g. constant velocity, constant acceleration) of the time sequence to learn the underlying equivariant structure of the dataset by simply training the encoder-decoder model to be able to predict the future observations. We will demonstrate that, with our framework, the hidden disentangled structure of the dataset naturally emerges as a by-product by applying simultaneous block-diagonalization to the transition operators in the latent space, the procedure which is commonly used in representation theory to decompose the feature-space based on the type of response to group actions. We will showcase our method from both empirical and theoretical perspectives. Our result suggests that finding a simple structured relation and learning a model with extrapolation capability are two sides of the same coin. The code is available at https://github.com/takerum/meta_sequential_prediction."
                },
                {
                    "title": "Compositional Generalization in Unsupervised Compositional Representation Learning: A Study on Disentanglement and Emergent Language",
                    "abstract": "Deep learning models struggle with compositional generalization, i.e. the ability to recognize or generate novel combinations of observed elementary concepts. In hopes of enabling compositional generalization, various unsupervised learning algorithms have been proposed with inductive biases that aim to induce compositional structure in learned representations (e.g. disentangled representation and emergent language learning). In this work, we evaluate these unsupervised learning algorithms in terms of how well they enable compositional generalization. Specifically, our evaluation protocol focuses on whether or not it is easy to train a simple model on top of the learned representation that generalizes to new combinations of compositional factors. We systematically study three unsupervised representation learning algorithms - $\\beta$-VAE, $\\beta$-TCVAE, and emergent language (EL) autoencoders - on two datasets that allow directly testing compositional generalization. We find that directly using the bottleneck representation with simple models and few labels may lead to worse generalization than using representations from layers before or after the learned representation itself. In addition, we find that the previously proposed metrics for evaluating the levels of compositionality are not correlated with actual compositional generalization in our framework. Surprisingly, we find that increasing pressure to produce a disentangled representation produces representations with worse generalization, while representations from EL models show strong compositional generalization. Taken together, our results shed new light on the compositional generalization behavior of different unsupervised learning algorithms with a new setting to rigorously test this behavior, and suggest the potential benefits of delevoping EL learning algorithms for more generalizable representations."
                },
                {
                    "title": "DCI-ES: An Extended Disentanglement Framework with Connections to Identifiability",
                    "abstract": "In representation learning, a common approach is to seek representations which disentangle the underlying factors of variation. Eastwood&Williams (2018) proposed three metrics for quantifying the quality of such disentangled representations: disentanglement (D), completeness (C) and informativeness (I). In this work, we first connect this DCI framework to two common notions of linear and nonlinear identifiability, thereby establishing a formal link between disentanglement and the closely-related field of independent component analysis. We then propose an extended DCI-ES framework with two new measures of representation quality - explicitness (E) and size (S) - and point out how D and C can be computed for black-box predictors. Our main idea is that the functional capacity required to use a representation is an important but thus-far neglected aspect of representation quality, which we quantify using explicitness or ease-of-use (E). We illustrate the relevance of our extensions on the MPI3D and Cars3D datasets."
                },
                {
                    "title": "Homomorphism Autoencoder - Learning Group Structured Representations from Observed Transitions",
                    "abstract": "How can agents learn internal models that veridically represent interactions with the real world is a largely open question. As machine learning is moving towards representations containing not just observational but also interventional knowledge, we study this problem using tools from representation learning and group theory. We propose methods enabling an agent acting upon the world to learn internal representations of sensory information that are consistent with actions that modify it. We use an autoencoder equipped with a group representation acting on its latent space, trained using an equivariance-derived loss in order to enforce a suitable homomorphism property on the group representation. In contrast to existing work, our approach does not require prior knowledge of the group and does not restrict the set of actions the agent can perform. We motivate our method theoretically, and show empirically that it can learn a group representation of the actions, thereby capturing the structure of the set of transformations applied to the environment. We further show that this allows agents to predict the effect of sequences of future actions with improved accuracy."
                },
                {
                    "title": "Universal approximation property of invertible neural networks",
                    "abstract": "Invertible neural networks (INNs) are neural network architectures with invertibility by design. Thanks to their invertibility and the tractability of Jacobian, INNs have various machine learning applications such as probabilistic modeling, generative modeling, and representation learning. However, their attractive properties often come at the cost of restricting the layer designs, which poses a question on their representation power: can we use these models to approximate sufficiently diverse functions? To answer this question, we have developed a general theoretical framework to investigate the representation power of INNs, building on a structure theorem of differential geometry. The framework simplifies the approximation problem of diffeomorphisms, which enables us to show the universal approximation properties of INNs. We apply the framework to two representative classes of INNs, namely Coupling-Flow-based INNs (CF-INNs) and Neural Ordinary Differential Equations (NODEs), and elucidate their high representation power despite the restrictions on their architectures."
                },
                {
                    "title": "From Statistical to Causal Learning",
                    "abstract": "We describe basic ideas underlying research to build and understand artificially intelligent systems: from symbolic approaches via statistical learning to interventional models relying on concepts of causality. Some of the hard open problems of machine learning and AI are intrinsically related to causality, and progress may require advances in our understanding of how to model and infer causality from data."
                },
                {
                    "title": "Weakly supervised causal representation learning",
                    "abstract": "Learning high-level causal representations together with a causal model from unstructured low-level data such as pixels is impossible from observational data alone. We prove under mild assumptions that this representation is however identifiable in a weakly supervised setting. This involves a dataset with paired samples before and after random, unknown interventions, but no further labels. We then introduce implicit latent causal models, variational autoencoders that represent causal variables and causal structure without having to optimize an explicit discrete graph structure. On simple image data, including a novel dataset of simulated robotic manipulation, we demonstrate that such models can reliably identify the causal structure and disentangle causal variables."
                },
                {
                    "title": "Graph Neural Networks are Dynamic Programmers",
                    "abstract": "Recent advances in neural algorithmic reasoning with graph neural networks (GNNs) are propped up by the notion of algorithmic alignment. Broadly, a neural network will be better at learning to execute a reasoning task (in terms of sample complexity) if its individual components align well with the target algorithm. Specifically, GNNs are claimed to align with dynamic programming (DP), a general problem-solving strategy which expresses many polynomial-time algorithms. However, has this alignment truly been demonstrated and theoretically quantified? Here we show, using methods from category theory and abstract algebra, that there exists an intricate connection between GNNs and DP, going well beyond the initial observations over individual algorithms such as Bellman-Ford. Exposing this connection, we easily verify several prior findings in the literature, produce better-grounded GNN architectures for edge-centric tasks, and demonstrate empirical results on the CLRS algorithmic reasoning benchmark. We hope our exposition will serve as a foundation for building stronger algorithmically aligned GNNs."
                },
                {
                    "title": "BASIC CONCEPTS OF ENRICHED CATEGORY THEORY",
                    "abstract": "Although numerous contributions from divers authors, over the past fifteen years or so, have brought enriched category theory to a developed state, there is still no connected account of the theory, or even of a substantial part of it. As the applications of the theory continue to expand - some recent examples are given below - the lack of such an account is the more acutely felt. The present book is designed to supply the want in part, by giving a fairly complete treatment of the limited area to which the title refers. The basic concepts of category theory certainly include the notion of functor-category, of limit and colimit, of Kan extension, and of density; with their applications to completions, perhaps including those relative completions given by categories of algebras for limit-defined theories. If we read 'V-category' for 'category' here, this is essentially the list of our chapter-headings below, after the first chapter introducing V-categories.  In fact our scope is wider than this might suggest; for what we give is also a selfcontained account of basic category theory as described above, assuming as prior knowledge only the most elementary categorical concepts, and treating the ordinary and enriched cases together from Chapter 3 on."
                },
                {
                    "title": "Multi-Agent MDP Homomorphic Networks",
                    "abstract": "This paper introduces Multi-Agent MDP Homomorphic Networks, a class of networks that allows distributed execution using only local information, yet is able to share experience between global symmetries in the joint state-action space of cooperative multi-agent systems. In cooperative multi-agent systems, complex symmetries arise between different configurations of the agents and their local observations. For example, consider a group of agents navigating: rotating the state globally results in a permutation of the optimal joint policy. Existing work on symmetries in single agent reinforcement learning can only be generalized to the fully centralized setting, because such approaches rely on the global symmetry in the full state-action spaces, and these can result in correspondences across agents. To encode such symmetries while still allowing distributed execution we propose a factorization that decomposes global symmetries into local transformations. Our proposed factorization allows for distributing the computation that enforces global symmetries over local agents and local interactions. We introduce a multi-agent equivariant policy network based on this factorization. We show empirically on symmetric multi-agent problems that globally symmetric distributable policies improve data efficiency compared to non-equivariant baselines."
                },
                {
                    "title": "Disentanglement Analysis with Partial Information Decomposition",
                    "abstract": "We propose a framework to analyze how multivariate representations disentangle ground-truth generative factors. A quantitative analysis of disentanglement has been based on metrics designed to compare how one variable explains each generative factor. Current metrics, however, may fail to detect entanglement that involves more than two variables, e.g., representations that duplicate and rotate generative factors in high dimensional spaces. In this work, we establish a framework to analyze information sharing in a multivariate representation with Partial Information Decomposition and propose a new disentanglement metric. This framework enables us to understand disentanglement in terms of uniqueness, redundancy, and synergy. We develop an experimental protocol to assess how increasingly entangled representations are evaluated with each metric and confirm that the proposed metric correctly responds to entanglement. Through experiments on variational autoencoders, we find that models with similar disentanglement scores have a variety of characteristics in entanglement, for each of which a distinct strategy may be required to obtain a disentangled representation."
                },
                {
                    "title": "An Invitation to Applied Category Theory. Seven Sketches in Compositionality. By Brendan Fong and David I. Spivak. Cambridge University Press, 2019. Paperback, pp. 348. Price GBP 37.99. ISBN 9781108711821.",
                    "abstract": "An invitation to applied category theory seven sketches. p d f an invitation to applied category theory seven. sevensketchesinpositionality. an invitation to applied category theory seven sketches. an invitation to applied category theory seven sketches. what is applied category theory applied category theory. seven sketches in positionality an invitation to. brendan fong. seven sketches in positionality an invitation to. introduction applied category theory. seven sketches in positionality an invitation to. pdf notes on category theory with examples from basic. seven sketches in positionality an invitation to. seven sketches in positionality an invitation reddit. an invitation to applied category theory seven sketches. seven sketches in positionality an invitation to. customer reviews an invitation to applied. seven sketches in positionality an invitation to. applied category theory. seven sketches in positionality mathematics. applied category theory munich m\u00fcnchen germany meetup. lecture videos and readings applied category theory. an invitation to applied category theory mathematical. an invitation to applied category theory by brendan fong. pdf seven sketches in positionality an invitation to. category theory haskellwiki. an invitation to applied category theory seven sketches. is there a category theory course on a mooc quora. chapter 1 generative effects orders and adjunctions. david spivak. syllabus applied category theory mathematics mit. p d f an invitation to applied category theory seven. seven sketches in positionality an invitation to. an invitation to applied category theory seven sketches. chapter 5 signal flow graphs props presentations and. seven sketches in positionality an invitation to. seven sketches in positionality an invitation reddit. seven sketches in positionality an invitation to. an invitation to applied category theory seven sketches. chapter 7 logic of behavior sheaves toposes languages. an invitation to applied category theory seven sketches. 1803 05316 seven sketches in positionality an. sevensketchesinpositionality xah lee. applied category theory course in the azimuth project. an invitation to applied category theory seven sketches"
                },
                {
                    "title": "Category Theory in Machine Learning",
                    "abstract": "Over the past two decades machine learning has permeated almost every realm of technology. At the same time, many researchers have begun using category theory as a unifying language, facilitating communication between different scientific disciplines. It is therefore unsurprising that there is a burgeoning interest in applying category theory to machine learning. We aim to document the motivations, goals and common themes across these applications. We touch on gradient-based learning, probability, and equivariant learning."
                },
                {
                    "title": "Learning disentangled representations via product manifold projection",
                    "abstract": "We propose a novel approach to disentangle the generative factors of variation underlying a given set of observations. Our method builds upon the idea that the (unknown) low-dimensional manifold underlying the data space can be explicitly modeled as a product of submanifolds. This definition of disentanglement gives rise to a novel weakly-supervised algorithm for recovering the unknown explanatory factors behind the data. At training time, our algorithm only requires pairs of non i.i.d. data samples whose elements share at least one, possibly multidimensional, generative factor of variation. We require no knowledge on the nature of these transformations, and do not make any limiting assumption on the properties of each subspace. Our approach is easy to implement, and can be successfully applied to different kinds of data (from images to 3D surfaces) undergoing arbitrary transformations. In addition to standard synthetic benchmarks, we showcase our method in challenging real-world applications, where we compare favorably with the state of the art."
                },
                {
                    "title": "Towards Building A Group-based Unsupervised Representation Disentanglement Framework",
                    "abstract": "Disentangled representation learning is one of the major goals of deep learning, and is a key step for achieving explainable and generalizable models. A well-defined theoretical guarantee still lacks for the VAE-based unsupervised methods, which are a set of popular methods to achieve unsupervised disentanglement. The Group Theory based definition of representation disentanglement mathematically connects the data transformations to the representations using the formalism of group. In this paper, built on the group-based definition and inspired by the n-th dihedral group, we first propose a theoretical framework towards achieving unsupervised representation disentanglement. We then propose a model, based on existing VAE-based methods, to tackle the unsupervised learning problem of the framework. In the theoretical framework, we prove three sufficient conditions on model, group structure, and data respectively in an effort to achieve, in an unsupervised way, disentangled representation per group-based definition. With the first two of the conditions satisfied and a necessary condition derived for the third one, we offer additional constraints, from the perspective of the group-based definition, for the existing VAE-based models. Experimentally, we train 1800 models covering the most prominent VAE-based methods on five datasets to verify the effectiveness of our theoretical framework. Compared to the original VAE-based methods, these Groupified VAEs consistently achieve better mean performance with smaller variances."
                },
                {
                    "title": "Measuring Disentanglement: A Review of Metrics",
                    "abstract": "Learning to disentangle and represent factors of variation in data is an important problem in artificial intelligence. While many advances have been made to learn these representations, it is still unclear how to quantify disentanglement. While several metrics exist, little is known on their implicit assumptions, what they truly measure, and their limits. In consequence, it is difficult to interpret results when comparing different representations. In this work, we survey supervised disentanglement metrics and thoroughly analyze them. We propose a new taxonomy in which all metrics fall into one of the three families: intervention-based, predictor-based, and information-based. We conduct extensive experiments in which we isolate properties of disentangled representations, allowing stratified comparison along several axes. From our experiment results and analysis, we provide insights on relations between disentangled representation properties. Finally, we share guidelines on how to measure disentanglement."
                },
                {
                    "title": "Quantifying and Learning Linear Symmetry-Based Disentanglement",
                    "abstract": "The definition of Linear Symmetry-Based Disentanglement (LSBD) formalizes the notion of linearly disentangled representations, but there is currently no metric to quantify LSBD. Such a metric is crucial to evaluate LSBD methods and to compare to previous understandings of disentanglement. We propose $\\mathcal{D}_\\mathrm{LSBD}$, a mathematically sound metric to quantify LSBD, and provide a practical implementation for $\\mathrm{SO}(2)$ groups. Furthermore, from this metric we derive LSBD-VAE, a semi-supervised method to learn LSBD representations. We demonstrate the utility of our metric by showing that (1) common VAE-based disentanglement methods don't learn LSBD representations, (2) LSBD-VAE as well as other recent methods can learn LSBD representations, needing only limited supervision on transformations, and (3) various desirable properties expressed by existing disentanglement metrics are also achieved by LSBD representations."
                },
                {
                    "title": "A Survey on Contrastive Self-supervised Learning",
                    "abstract": "Self-supervised learning has gained popularity because of its ability to avoid the cost of annotating large-scale datasets. It is capable of adopting self-defined pseudolabels as supervision and use the learned representations for several downstream tasks. Specifically, contrastive learning has recently become a dominant component in self-supervised learning for computer vision, natural language processing (NLP), and other domains. It aims at embedding augmented versions of the same sample close to each other while trying to push away embeddings from different samples. This paper provides an extensive review of self-supervised methods that follow the contrastive approach. The work explains commonly used pretext tasks in a contrastive learning setup, followed by different architectures that have been proposed so far. Next, we present a performance comparison of different methods for multiple downstream tasks such as image classification, object detection, and action recognition. Finally, we conclude with the limitations of the current methods and the need for further techniques and future directions to make meaningful progress."
                },
                {
                    "title": "On the Transfer of Disentangled Representations in Realistic Settings",
                    "abstract": "Learning meaningful representations that disentangle the underlying structure of the data generating process is considered to be of key importance in machine learning. While disentangled representations were found to be useful for diverse tasks such as abstract reasoning and fair classification, their scalability and real-world impact remain questionable. We introduce a new high-resolution dataset with 1M simulated images and over 1,800 annotated real-world images of the same robotic setup. In contrast to previous work, this new dataset exhibits correlations, a complex underlying structure, and allows to evaluate transfer to unseen simulated and real-world settings where the encoder i) remains in distribution or ii) is out of distribution. We propose new architectures in order to scale disentangled representation learning to realistic high-resolution settings and conduct a large-scale empirical study of disentangled representations on this dataset. We observe that disentanglement is a good predictor for out-of-distribution (OOD) task performance."
                },
                {
                    "title": "Product Manifold Learning",
                    "abstract": "We consider problems of dimensionality reduction and learning data representations for continuous spaces with two or more independent degrees of freedom. Such problems occur, for example, when observing shapes with several components that move independently. Mathematically, if the parameter space of each continuous independent motion is a manifold, then their combination is known as a product manifold. In this paper, we present a new paradigm for non-linear independent component analysis called manifold factorization. Our factorization algorithm is based on spectral graph methods for manifold learning and the separability of the Laplacian operator on product spaces. Recovering the factors of a manifold yields meaningful lower-dimensional representations and provides a new way to focus on particular aspects of the data space while ignoring others. We demonstrate the potential use of our method for an important and challenging problem in structural biology: mapping the motions of proteins and other large molecules using cryo-electron microscopy datasets."
                },
                {
                    "title": "Linear Disentangled Representations and Unsupervised Action Estimation",
                    "abstract": "Disentangled representation learning has seen a surge in interest over recent times, generally focusing on new models which optimise one of many disparate disentanglement metrics. Symmetry Based Disentangled Representation learning introduced a robust mathematical framework that defined precisely what is meant by a \"linear disentangled representation\". This framework determined that such representations would depend on a particular decomposition of the symmetry group acting on the data, showing that actions would manifest through irreducible group representations acting on independent representational subspaces. Caselles-Dupre et al [2019] subsequently proposed the first model to induce and demonstrate a linear disentangled representation in a VAE model. In this work we empirically show that linear disentangled representations are not generally present in standard VAE models and that they instead require altering the loss landscape to induce them. We proceed to show that such representations are a desirable property with regard to classical disentanglement metrics. Finally we propose a method to induce irreducible representations which forgoes the need for labelled action sequences, as was required by prior work. We explore a number of properties of this method, including the ability to learn from action sequences without knowledge of intermediate states and robustness under visual noise. We also demonstrate that it can successfully learn 4 independent symmetries directly from pixels."
                },
                {
                    "title": "Natural Graph Networks",
                    "abstract": "Conventional neural message passing algorithms are invariant under permutation of the messages and hence forget how the information flows through the network. Studying the local symmetries of graphs, we propose a more general algorithm that uses different kernels on different edges, making the network equivariant to local and global graph isomorphisms and hence more expressive. Using elementary category theory, we formalize many distinct equivariant neural networks as natural networks, and show that their kernels are 'just' a natural transformation between two functors. We give one practical instantiation of a natural network on graphs which uses a equivariant message network parameterization, yielding good performance on several benchmarks."
                },
                {
                    "title": "Disentangling by Subspace Diffusion",
                    "abstract": "We present a novel nonparametric algorithm for symmetry-based disentangling of data manifolds, the Geometric Manifold Component Estimator (GEOMANCER). GEOMANCER provides a partial answer to the question posed by Higgins et al. (2018): is it possible to learn how to factorize a Lie group solely from observations of the orbit of an object it acts on? We show that fully unsupervised factorization of a data manifold is possible *if* the true metric of the manifold is known and each factor manifold has nontrivial holonomy -- for example, rotation in 3D. Our algorithm works by estimating the subspaces that are invariant under random walk diffusion, giving an approximation to the de Rham decomposition from differential geometry. We demonstrate the efficacy of GEOMANCER on several complex synthetic manifolds. Our work reduces the question of whether unsupervised disentangling is possible to the question of whether unsupervised metric learning is possible, providing a unifying insight into the geometric nature of representation learning."
                },
                {
                    "title": "On Disentangled Representations Learned from Correlated Data",
                    "abstract": "The focus of disentanglement approaches has been on identifying independent factors of variation in data. However, the causal variables underlying real-world observations are often not statistically independent. In this work, we bridge the gap to real-world scenarios by analyzing the behavior of the most prominent disentanglement approaches on correlated data in a large-scale empirical study (including 4260 models). We show and quantify that systematically induced correlations in the dataset are being learned and reflected in the latent representations, which has implications for downstream applications of disentanglement such as fairness. We also demonstrate how to resolve these latent correlations, either using weak supervision during training or by post-hoc correcting a pre-trained model with a small number of labels."
                },
                {
                    "title": "Evaluating the Disentanglement of Deep Generative Models through Manifold Topology",
                    "abstract": "Learning disentangled representations is regarded as a fundamental task for improving the generalization, robustness, and interpretability of generative models. However, measuring disentanglement has been challenging and inconsistent, often dependent on an ad-hoc external model or specific to a certain dataset. To address this, we present a method for quantifying disentanglement that only uses the generative model, by measuring the topological similarity of conditional submanifolds in the learned representation. This method showcases both unsupervised and supervised variants. To illustrate the effectiveness and applicability of our method, we empirically evaluate several state-of-the-art models across multiple datasets. We find that our method ranks models similarly to existing methods."
                },
                {
                    "title": "Equivariant Flows: exact likelihood generative learning for symmetric densities",
                    "abstract": "Normalizing flows are exact-likelihood generative neural networks which approximately transform samples from a simple prior distribution to samples of the probability distribution of interest. Recent work showed that such generative models can be utilized in statistical mechanics to sample equilibrium states of many-body systems in physics and chemistry. To scale and generalize these results, it is essential that the natural symmetries in the probability density - in physics defined by the invariances of the target potential - are built into the flow. We provide a theoretical sufficient criterion showing that the distribution generated by equivariant normalizing flows is invariant with respect to these symmetries by design. Furthermore, we propose building blocks for flows which preserve symmetries which are usually found in physical/chemical many-body particle systems. Using benchmark systems motivated from molecular physics, we demonstrate that those symmetry preserving flows can provide better generalization capabilities and sampling efficiency."
                },
                {
                    "title": "Calibrated Surrogate Losses for Adversarially Robust Classification",
                    "abstract": "Adversarially robust classification seeks a classifier that is insensitive to adversarial perturbations of test patterns. This problem is often formulated via a minimax objective, where the target loss is the worst-case value of the 0-1 loss subject to a bound on the size of perturbation. Recent work has proposed convex surrogates for the adversarial 0-1 loss, in an effort to make optimization more tractable. In this work, we consider the question of which surrogate losses are calibrated with respect to the adversarial 0-1 loss, meaning that minimization of the former implies minimization of the latter. We show that no convex surrogate loss is calibrated with respect to the adversarial 0-1 loss when restricted to the class of linear models. We further introduce a class of nonconvex losses and offer necessary and sufficient conditions for losses in this class to be calibrated."
                },
                {
                    "title": "Progressive Learning and Disentanglement of Hierarchical Representations",
                    "abstract": "Learning rich representation from data is an important task for deep generative models such as variational auto-encoder (VAE). However, by extracting high-level abstractions in the bottom-up inference process, the goal of preserving all factors of variations for top-down generation is compromised. Motivated by the concept of \"starting small\", we present a strategy to progressively learn independent hierarchical representations from high- to low-levels of abstractions. The model starts with learning the most abstract representation, and then progressively grow the network architecture to introduce new representations at different levels of abstraction. We quantitatively demonstrate the ability of the presented model to improve disentanglement in comparison to existing works on two benchmark data sets using three disentanglement metrics, including a new metric we proposed to complement the previously-presented metric of mutual information gap. We further present both qualitative and quantitative evidence on how the progression of learning improves disentangling of hierarchical representations. By drawing on the respective advantage of hierarchical representation learning and progressive learning, this is to our knowledge the first attempt to improve disentanglement by progressively growing the capacity of VAE to learn hierarchical representations."
                },
                {
                    "title": "Weakly-Supervised Disentanglement Without Compromises",
                    "abstract": "Intelligent agents should be able to learn useful representations by observing changes in their environment. We model such observations as pairs of non-i.i.d. images sharing at least one of the underlying factors of variation. First, we theoretically show that only knowing how many factors have changed, but not which ones, is sufficient to learn disentangled representations. Second, we provide practical algorithms that learn disentangled representations from pairs of images without requiring annotation of groups, individual factors, or the number of factors that have changed. Third, we perform a large-scale empirical study and show that such pairs of observations are sufficient to reliably learn disentangled representations on several benchmark data sets. Finally, we evaluate our learned representations and find that they are simultaneously useful on a diverse suite of tasks, including generalization under covariate shifts, fairness, and abstract reasoning. Overall, our results demonstrate that weak supervision enables learning of useful disentangled representations in realistic scenarios."
                },
                {
                    "title": "Robust Aggregation for Federated Learning",
                    "abstract": "We present a novel approach to federated learning that endows its aggregation process with greater robustness to potential poisoning of local data or model parameters of participating devices. The proposed approach, Robust Federated Aggregation (RFA), relies on the aggregation of updates using the geometric median, which can be computed efficiently using a Weiszfeld-type algorithm. RFA is agnostic to the level of corruption and aggregates model updates without revealing each device\u2019s individual contribution. We establish the convergence of the robust federated learning algorithm for the stochastic learning of additive models with least squares. We also offer two variants of RFA: a faster one with one-step robust aggregation, and another one with on-device personalization. We present experimental results with additive models and deep networks for three tasks in computer vision and natural language processing. The experiments show that RFA is competitive with the classical aggregation when the level of corruption is low, while demonstrating greater robustness under high corruption."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "Weakly Supervised Disentanglement with Guarantees",
                    "abstract": "Learning disentangled representations that correspond to factors of variation in real-world data is critical to interpretable and human-controllable machine learning. Recently, concerns about the viability of learning disentangled representations in a purely unsupervised manner has spurred a shift toward the incorporation of weak supervision. However, there is currently no formalism that identifies when and how weak supervision will guarantee disentanglement. To address this issue, we provide a theoretical framework to assist in analyzing the disentanglement guarantees (or lack thereof) conferred by weak supervision when coupled with learning algorithms based on distribution matching. We empirically verify the guarantees and limitations of several weak supervision methods (restricted labeling, match-pairing, and rank-pairing), demonstrating the predictive power and usefulness of our theoretical framework."
                },
                {
                    "title": "Theory and Evaluation Metrics for Learning Disentangled Representations",
                    "abstract": "We make two theoretical contributions to disentanglement learning by (a) defining precise semantics of disentangled representations, and (b) establishing robust metrics for evaluation. First, we characterize the concept \"disentangled representations\" used in supervised and unsupervised methods along three dimensions-informativeness, separability and interpretability - which can be expressed and quantified explicitly using information-theoretic constructs. This helps explain the behaviors of several well-known disentanglement learning models. We then propose robust metrics for measuring informativeness, separability and interpretability. Through a comprehensive suite of experiments, we show that our metrics correctly characterize the representations learned by different methods and are consistent with qualitative (visual) results. Thus, the metrics allow disentanglement learning methods to be compared on a fair ground. We also empirically uncovered new interesting properties of VAE-based methods and interpreted them with our formulation. These findings are promising and hopefully will encourage the design of more theoretically driven models for learning disentangled representations."
                },
                {
                    "title": "A Group-Theoretic Framework for Data Augmentation",
                    "abstract": "Data augmentation is a widely used trick when training deep neural networks: in addition to the original data, properly transformed data are also added to the training set. However, to the best of our knowledge, a clear mathematical framework to explain the performance benefits of data augmentation is not available. \nIn this paper, we develop such a theoretical framework. We show data augmentation is equivalent to an averaging operation over the orbits of a certain group that keeps the data distribution approximately invariant. We prove that it leads to variance reduction. We study empirical risk minimization, and the examples of exponential families, linear regression, and certain two-layer neural networks. We also discuss how data augmentation could be used in problems with symmetry where other approaches are prevalent, such as in cryo-electron microscopy (cryo-EM)."
                },
                {
                    "title": "On the Transfer of Inductive Bias from Simulation to the Real World: a New Disentanglement Dataset",
                    "abstract": "Learning meaningful and compact representations with disentangled semantic aspects is considered to be of key importance in representation learning. Since real-world data is notoriously costly to collect, many recent state-of-the-art disentanglement models have heavily relied on synthetic toy data-sets. In this paper, we propose a novel data-set which consists of over one million images of physical 3D objects with seven factors of variation, such as object color, shape, size and position. In order to be able to control all the factors of variation precisely, we built an experimental platform where the objects are being moved by a robotic arm. In addition, we provide two more datasets which consist of simulations of the experimental setup. These datasets provide for the first time the possibility to systematically investigate how well different disentanglement methods perform on real data in comparison to simulation, and how simulated data can be leveraged to build better representations of the real world. We provide a first experimental study of these questions and our results indicate that learned models transfer poorly, but that model and hyperparameter selection is an effective means of transferring information to the real world."
                },
                {
                    "title": "On the Fairness of Disentangled Representations",
                    "abstract": "Recently there has been a significant interest in learning disentangled representations, as they promise increased interpretability, generalization to unseen scenarios and faster learning on downstream tasks. In this paper, we investigate the usefulness of different notions of disentanglement for improving the fairness of downstream prediction tasks based on representations. We consider the setting where the goal is to predict a target variable based on the learned representation of high-dimensional observations (such as images) that depend on both the target variable and an \\emph{unobserved} sensitive variable. We show that in this setting both the optimal and empirical predictions can be unfair, even if the target variable and the sensitive variable are independent. Analyzing the representations of more than \\num{12600} trained state-of-the-art disentangled models, we observe that several disentanglement scores are consistently correlated with increased fairness, suggesting that disentanglement may be a useful property to encourage fairness when sensitive variables are not observed."
                },
                {
                    "title": "Calibrated Surrogate Maximization of Linear-fractional Utility in Binary Classification",
                    "abstract": "Complex classification performance metrics such as the F${}_\\beta$-measure and Jaccard index are often used, in order to handle class-imbalanced cases such as information retrieval and image segmentation. These performance metrics are not decomposable, that is, they cannot be expressed in a per-example manner, which hinders a straightforward application of M-estimation widely used in supervised learning. In this paper, we consider linear-fractional metrics, which are a family of classification performance metrics that encompasses many standard ones such as the F${}_\\beta$-measure and Jaccard index, and propose methods to directly maximize performances under those metrics. A clue to tackle their direct optimization is a calibrated surrogate utility, which is a tractable lower bound of the true utility function representing a given metric. We characterize sufficient conditions which make the surrogate maximization coincide with the maximization of the true utility. Simulation results on benchmark datasets validate the effectiveness of our calibrated surrogate maximization especially if the sample sizes are extremely small."
                },
                {
                    "title": "Symmetry-Based Disentangled Representation Learning requires Interaction with Environments",
                    "abstract": "Finding a generally accepted formal definition of a disentangled representation in the context of an agent behaving in an environment is an important challenge towards the construction of data-efficient autonomous agents. Higgins et al. recently proposed Symmetry-Based Disentangled Representation Learning, a definition based on a characterization of symmetries in the environment using group theory. We build on their work and make observations, theoretical and empirical, that lead us to argue that Symmetry-Based Disentangled Representation Learning cannot only be based on static observations: agents should interact with the environment to discover its symmetries. Our experiments can be reproduced in Colab and the code is available on GitHub."
                },
                {
                    "title": "On the Calibration of Multiclass Classification with Rejection",
                    "abstract": "We investigate the problem of multiclass classification with rejection, where a classifier can choose not to make a prediction to avoid critical misclassification. First, we consider an approach based on simultaneous training of a classifier and a rejector, which achieves the state-of-the-art performance in the binary case. We analyze this approach for the multiclass case and derive a general condition for calibration to the Bayes-optimal solution, which suggests that calibration is hard to achieve by general loss functions unlike the binary case. Next, we consider another traditional approach based on confidence scores, in which the existing work focuses on a specific class of losses. We propose rejection criteria for more general losses for this approach and guarantee calibration to the Bayes-optimal solution. Finally, we conduct experiments to validate the relevance of our theoretical findings."
                },
                {
                    "title": "Towards a Definition of Disentangled Representations",
                    "abstract": "How can intelligent agents solve a diverse set of tasks in a data-efficient manner? The disentangled representation learning approach posits that such an agent would benefit from separating out (disentangling) the underlying structure of the world into disjoint parts of its representation. However, there is no generally agreed-upon definition of disentangling, not least because it is unclear how to formalise the notion of world structure beyond toy datasets with a known ground truth generative process. Here we propose that a principled solution to characterising disentangled representations can be found by focusing on the transformation properties of the world. In particular, we suggest that those transformations that change only some properties of the underlying world state, while leaving all other properties invariant, are what gives exploitable structure to any kind of data. Similar ideas have already been successfully applied in physics, where the study of symmetry transformations has revolutionised the understanding of the world structure. By connecting symmetry transformations to vector representations using the formalism of group and representation theory we arrive at the first formal definition of disentangled representations. Our new definition is in agreement with many of the current intuitions about disentangling, while also providing principled resolutions to a number of previous points of contention. While this work focuses on formally defining disentangling - as opposed to solving the learning problem - we believe that the shift in perspective to studying data transformations can stimulate the development of better representation learning algorithms."
                },
                {
                    "title": "Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations",
                    "abstract": "The key idea behind the unsupervised learning of disentangled representations is that real-world data is generated by a few explanatory factors of variation which can be recovered by unsupervised learning algorithms. In this paper, we provide a sober look at recent progress in the field and challenge some common assumptions. We first theoretically show that the unsupervised learning of disentangled representations is fundamentally impossible without inductive biases on both the models and the data. Then, we train more than 12000 models covering most prominent methods and evaluation metrics in a reproducible large-scale experimental study on seven different data sets. We observe that while the different methods successfully enforce properties ``encouraged'' by the corresponding losses, well-disentangled models seemingly cannot be identified without supervision. Furthermore, increased disentanglement does not seem to lead to a decreased sample complexity of learning for downstream tasks. Our results suggest that future work on disentanglement learning should be explicit about the role of inductive biases and (implicit) supervision, investigate concrete benefits of enforcing disentanglement of the learned representations, and consider a reproducible experimental setup covering several data sets."
                },
                {
                    "title": "Invertible Residual Networks",
                    "abstract": "We show that standard ResNet architectures can be made invertible, allowing the same model to be used for classification, density estimation, and generation. Typically, enforcing invertibility requires partitioning dimensions or restricting network architectures. In contrast, our approach only requires adding a simple normalization step during training, already available in standard frameworks. Invertible ResNets define a generative model which can be trained by maximum likelihood on unlabeled data. To compute likelihoods, we introduce a tractable approximation to the Jacobian log-determinant of a residual block. Our empirical evaluation shows that invertible ResNets perform competitively with both state-of-the-art image classifiers and flow-based generative models, something that has not been previously achieved with a single architecture."
                },
                {
                    "title": "Robustly Disentangled Causal Mechanisms: Validating Deep Representations for Interventional Robustness",
                    "abstract": "The ability to learn disentangled representations that split underlying sources of variation in high dimensional, unstructured data is important for data efficient and robust use of neural networks. While various approaches aiming towards this goal have been proposed in recent times, a commonly accepted definition and validation procedure is missing. We provide a causal perspective on representation learning which covers disentanglement and domain shift robustness as special cases. Our causal framework allows us to introduce a new metric for the quantitative evaluation of deep latent variable models. We show how this metric can be estimated from labeled observational data and further provide an efficient estimation algorithm that scales linearly in the dataset size."
                },
                {
                    "title": "Invariant and Equivariant Graph Networks",
                    "abstract": "Invariant and equivariant networks have been successfully used for learning images, sets, point clouds, and graphs. A basic challenge in developing such networks is finding the maximal collection of invariant and equivariant linear layers. Although this question is answered for the first three examples (for popular transformations, at-least), a full characterization of invariant and equivariant linear layers for graphs is not known. \nIn this paper we provide a characterization of all permutation invariant and equivariant linear layers for (hyper-)graph data, and show that their dimension, in case of edge-value graph data, is 2 and 15, respectively. More generally, for graph data defined on k-tuples of nodes, the dimension is the k-th and 2k-th Bell numbers. Orthogonal bases for the layers are computed, including generalization to multi-graph data. The constant number of basis elements and their characteristics allow successfully applying the networks to different size graphs. From the theoretical point of view, our results generalize and unify recent advancement in equivariant deep learning. In particular, we show that our model is capable of approximating any message passing neural network \nApplying these new linear layers in a simple deep neural network framework is shown to achieve comparable results to state-of-the-art and to have better expressivity than previous invariant and equivariant bases."
                },
                {
                    "title": "Disentangling by Factorising",
                    "abstract": "We define and address the problem of unsupervised learning of disentangled representations on data generated from independent factors of variation. We propose FactorVAE, a method that disentangles by encouraging the distribution of representations to be factorial and hence independent across the dimensions. We show that it improves upon $\\beta$-VAE by providing a better trade-off between disentanglement and reconstruction quality. Moreover, we highlight the problems of a commonly used disentanglement metric and introduce a new metric that does not suffer from them."
                },
                {
                    "title": "A Framework for the Quantitative Evaluation of Disentangled Representations",
                    "abstract": "Recent AI research has emphasised the importance of learning disentangled representations of the explanatory factors behind data. Despite the growing interest in models which can learn such representations, visual inspection remains the standard evaluation metric. While various desiderata have been implied in recent definitions, it is currently unclear what exactly makes one disentangled representation better than another. In this work we propose a framework for the quantitative evaluation of disentangled representations when the ground-truth latent structure is available. Three criteria are explicitly defined and quantified to elucidate the quality of learnt representations and thus compare models on an equal basis. To illustrate the appropriateness of the framework, we employ it to compare quantitatively the representations learned by recent state-of-the-art models."
                },
                {
                    "title": "Isolating Sources of Disentanglement in Variational Autoencoders",
                    "abstract": "We decompose the evidence lower bound to show the existence of a term measuring the total correlation between latent variables. We use this to motivate our $\\beta$-TCVAE (Total Correlation Variational Autoencoder), a refinement of the state-of-the-art $\\beta$-VAE objective for learning disentangled representations, requiring no additional hyperparameters during training. We further propose a principled classifier-free measure of disentanglement called the mutual information gap (MIG). We perform extensive quantitative and qualitative experiments, in both restricted and non-restricted settings, and show a strong relation between total correlation and disentanglement, when the latent variables model is trained using our framework."
                },
                {
                    "title": "Learning Deep Disentangled Embeddings with the F-Statistic Loss",
                    "abstract": "Deep-embedding methods aim to discover representations of a domain that make explicit the domain's class structure. Disentangling methods aim to make explicit compositional or factorial structure. We combine these two active but independent lines of research and propose a new paradigm for discovering disentangled representations of class structure; these representations reveal the underlying factors that jointly determine class. We propose and evaluate a novel loss function based on the $F$ statistic, which describes the separation of two or more distributions. By ensuring that distinct classes are well separated on a subset of embedding dimensions, we obtain embeddings that are useful for few-shot learning. By not requiring separation on all dimensions, we encourage the discovery of disentangled representations. Our embedding procedure matches or beats state-of-the-art procedures on deep embeddings, as evaluated by performance on recall@$k$ and few-shot learning tasks. To evaluate alternative approaches on disentangling, we formalize two key properties of a disentangled representation: modularity and explicitness. By these criteria, our procedure yields disentangled representations, whereas traditional procedures fail. The goal of our work is to obtain more interpretable, manipulable, and generalizable deep representations of concepts and categories."
                },
                {
                    "title": "Quantales and Hyperstructures: Monads, Mo' Problems",
                    "abstract": "We present a theory of lattice-enriched semirings, called quantic semirings, which generalize both quantales and powersets of hyperrings. Using these structures, we show how to recover the spectrum of a Krasner hyperring (and in particular, a commutative ring with unity) via universal constructions, and generalize the spectrum to a new class of hyperstructures, hypersemirings. (These include hyperstructures currently studied under the name \"semihyperrings\", but we have weakened the distributivity axioms.) \nMuch of the work consists of background material on closure systems, suplattices, quantales, and hyperoperations, some of which is new. In particular, we define the category of covered semigroups, show their close relationship with quantales, and construct their spectra by exploiting the construction of a universal quotient frame by Rosenthal. \nWe extend these results to hypersemigroups, demonstrating various folkloric correspondences between hyperstructures and lattice-enriched structures on the powerset. Building on this, we proceed to define quantic semirings, and show that they are the lattice-enriched counterparts of hypersemirings. To a quantic semiring, we show how to define a universal quotient quantale, which we call the quantic spectrum, and using this, we show how to obtain the spectrum of a hypersemiring as a topological space in a canonical fashion. \nFinally, we we conclude with some applications of the theory to the ordered blueprints of Lorscheid."
                },
                {
                    "title": "Deep Sets",
                    "abstract": "In this paper, we study the problem of designing objective functions for machine learning problems defined on finite \\emph{sets}. In contrast to traditional objective functions defined for machine learning problems operating on finite dimensional vectors, the new objective functions we propose are operating on finite sets and are invariant to permutations. Such problems are widespread, ranging from estimation of population statistics \\citep{poczos13aistats}, via anomaly detection in piezometer data of embankment dams \\citep{Jung15Exploration}, to cosmology \\citep{Ntampaka16Dynamical,Ravanbakhsh16ICML1}. Our main theorem characterizes the permutation invariant objective functions and provides a family of functions to which any permutation invariant objective function must belong. This family of functions has a special structure which enables us to design a deep network architecture that can operate on sets and which can be deployed on a variety of scenarios including both unsupervised and supervised learning tasks. We demonstrate the applicability of our method on population statistic estimation, point cloud classification, set expansion, and image tagging."
                },
                {
                    "title": "Who Invented the Gromov-Hausdorff Distance?",
                    "abstract": "One of the most beautiful notions of metric geometry is the Gromov-Hausdorff distance which measures the difference between two metric spaces. To define the distance, let us isometrically embed these spaces into various metric spaces and measure the Hausdorff distance between their images. The best matching corresponds to the least Hausdorff distance. The idea to compare metric spaces in such a way was described in M.Gromov publications dating back to 1981. It was shown that this distance being restricted to isometry classes of compact metric spaces forms a metric which is now called Gromov-Hausdorff metric. However, whether M.Gromov was the first who introduced this metric? It turns out that 6 years before these Gromov's works, in 1975, another mathematician, namely, David Edwards published a paper in which he defined this metric in another way. Also, Edwards found and proved the basic properties of the distance. It seems unfair that almost no one mentions this Edwards's paper including well-known specialists in metric geometry. The main goal of the present publication is to fill this gap."
                },
                {
                    "title": "beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework",
                    "abstract": "an"
                },
                {
                    "title": "Input Convex Neural Networks",
                    "abstract": "This paper presents the input convex neural network architecture. These are scalar-valued (potentially deep) neural networks with constraints on the network parameters such that the output of the network is a convex function of (some of) the inputs. The networks allow for efficient inference via optimization over some inputs to the network given others, and can be applied to settings including structured prediction, data imputation, reinforcement learning, and others. In this paper we lay the basic groundwork for these models, proposing methods for inference, optimization and learning, and analyze their representational power. We show that many existing neural network architectures can be made input-convex with a minor modification, and develop specialized optimization algorithms tailored to this setting. Finally, we highlight the performance of the methods on multi-label prediction, image completion, and reinforcement learning problems, where we show improvement over the existing state of the art in many cases."
                },
                {
                    "title": "Geometric median in nearly linear time",
                    "abstract": "In this paper we provide faster algorithms for solving the geometric median problem: given n points in d compute a point that minimizes the sum of Euclidean distances to the points. This is one of the oldest non-trivial problems in computational geometry yet despite a long history of research the previous fastest running times for computing a (1+\u0454)-approximate geometric median were O(d\u00b7 n4/3\u0454\u22128/3) by Chin et. al, \u00d5(dexp\u0454\u22124log\u0454\u22121) by Badoiu et. al, O(nd+poly(d,\u0454\u22121)) by Feldman and Langberg, and the polynomial running time of O((nd)O(1)log1/\u0454) by Parrilo and Sturmfels and Xue and Ye. In this paper we show how to compute such an approximate geometric median in time O(ndlog3n/\u0454) and O(d\u0454\u22122). While our O(d\u0454\u22122) is a fairly straightforward application of stochastic subgradient descent, our O(ndlog3n/\u0454) time algorithm is a novel long step interior point method. We start with a simple O((nd)O(1)log1/\u0454) time interior point method and show how to improve it, ultimately building an algorithm that is quite non-standard from the perspective of interior point literature. Our result is one of few cases of outperforming standard interior point theory. Furthermore, it is the only case we know of where interior point methods yield a nearly linear time algorithm for a canonical optimization problem that traditionally requires superlinear time."
                },
                {
                    "title": "Group Equivariant Convolutional Networks",
                    "abstract": "We introduce Group equivariant Convolutional Neural Networks (G-CNNs), a natural generalization of convolutional neural networks that reduces sample complexity by exploiting symmetries. G-CNNs use G-convolutions, a new type of layer that enjoys a substantially higher degree of weight sharing than regular convolution layers. G-convolutions increase the expressive capacity of the network without increasing the number of parameters. Group convolution layers are easy to use and can be implemented with negligible computational overhead for discrete groups generated by translations, reflections and rotations. G-CNNs achieve state of the art results on CI- FAR10 and rotated MNIST."
                },
                {
                    "title": "Deep Visual Analogy-Making",
                    "abstract": "In addition to identifying the content within a single image, relating images and generating related images are critical tasks for image understanding. Recently, deep convolutional networks have yielded breakthroughs in predicting image labels, annotations and captions, but have only just begun to be used for generating high-quality images. In this paper we develop a novel deep network trained end-to-end to perform visual analogy making, which is the task of transforming a query image according to an example pair of related images. Solving this problem requires both accurately recognizing a visual relationship and generating a transformed query image accordingly. Inspired by recent advances in language modeling, we propose to solve visual analogies by learning to map images to a neural embedding in which analogical reasoning is simple, such as by vector subtraction and addition. In experiments, our model effectively models visual analogies on several datasets: 2D shapes, animated video game sprites, and 3D car models."
                },
                {
                    "title": "Variational Inference with Normalizing Flows",
                    "abstract": "The choice of approximate posterior distribution is one of the core problems in variational inference. Most applications of variational inference employ simple families of posterior approximations in order to allow for efficient inference, focusing on mean-field or other simple structured approximations. This restriction has a significant impact on the quality of inferences made using variational methods. We introduce a new approach for specifying flexible, arbitrarily complex and scalable approximate posterior distributions. Our approximations are distributions constructed through a normalizing flow, whereby a simple initial density is transformed into a more complex one by applying a sequence of invertible transformations until a desired level of complexity is attained. We use this view of normalizing flows to develop categories of finite and infinitesimal flows and provide a unified view of approaches for constructing rich posterior approximations. We demonstrate that the theoretical advantages of having posteriors that better match the true posterior, combined with the scalability of amortized variational approaches, provides a clear improvement in performance and applicability of variational inference."
                },
                {
                    "title": "Transformation Properties of Learned Visual Representations",
                    "abstract": "When a three-dimensional object moves relative to an observer, a change occurs on the observer's image plane and in the visual representation computed by a learned model. Starting with the idea that a good visual representation is one that transforms linearly under scene motions, we show, using the theory of group representations, that any such representation is equivalent to a combination of the elementary irreducible representations. We derive a striking relationship between irreducibility and the statistical dependency structure of the representation, by showing that under restricted conditions, irreducible representations are decorrelated. Under partial observability, as induced by the perspective projection of a scene onto the image plane, the motion group does not have a linear action on the space of images, so that it becomes necessary to perform inference over a latent representation that does transform linearly. This idea is demonstrated in a model of rotating NORB objects that employs a latent representation of the non-commutative 3D rotation group SO(3)."
                },
                {
                    "title": "Learning the Irreducible Representations of Commutative Lie Groups",
                    "abstract": "We present a new probabilistic model of compact commutative Lie groups that produces invariant-equivariant and disentangled representations of data. To define the notion of disentangling, we borrow a fundamental principle from physics that is used to derive the elementary particles of a system from its symmetries. Our model employs a newfound Bayesian conjugacy relation that enables fully tractable probabilistic inference over compact commutative Lie groups - a class that includes the groups that describe the rotation and cyclic translation of images. We train the model on pairs of transformed image patches, and show that the learned invariant representation is highly effective for classification."
                },
                {
                    "title": "Geometric median and robust estimation in Banach spaces",
                    "abstract": "In many real-world applications, collected data are contaminated by noise with heavy-tailed distribution and might contain outliers of large magnitude. In this situation, it is necessary to apply methods which produce reliable outcomes even if the input contains corrupted measurements. We describe a general method which allows one to obtain estimators with tight concentration around the true parameter of interest taking values in a Banach space. Suggested construction relies on the fact that the geometric median of a collection of independent \"weakly concentrated\" estimators satisfies a much stronger deviation bound than each individual element in the collection. Our approach is illustrated through several examples, including sparse linear regression and low-rank matrix recovery problems."
                },
                {
                    "title": "Representation Learning: A Review and New Perspectives",
                    "abstract": "The success of machine learning algorithms generally depends on data representation, and we hypothesize that this is because different representations can entangle and hide more or less the different explanatory factors of variation behind the data. Although specific domain knowledge can be used to help design representations, learning with generic priors can also be used, and the quest for AI is motivating the design of more powerful representation-learning algorithms implementing such priors. This paper reviews recent work in the area of unsupervised feature learning and deep learning, covering advances in probabilistic models, autoencoders, manifold learning, and deep networks. This motivates longer term unanswered questions about the appropriate objectives for learning good representations, for computing representations (i.e., inference), and the geometrical connections between representation learning, density estimation, and manifold learning."
                },
                {
                    "title": "Scikit-learn: Machine Learning in Python",
                    "abstract": "Scikit-learn is a Python module integrating a wide range of state-of-the-art machine learning algorithms for medium-scale supervised and unsupervised problems. This package focuses on bringing machine learning to non-specialists using a general-purpose high-level language. Emphasis is put on ease of use, performance, documentation, and API consistency. It has minimal dependencies and is distributed under the simplified BSD license, encouraging its use in both academic and commercial settings. Source code, binaries, and documentation can be downloaded from http://scikit-learn.sourceforge.net."
                },
                {
                    "title": "An informal introduction to topos theory",
                    "abstract": "This short expository text is for readers who are confident in basic category theory but know little or nothing about toposes. It is based on some impromptu talks given to a small group of category theorists."
                },
                {
                    "title": "Monotone and Partially Monotone Neural Networks",
                    "abstract": "In many classification and prediction problems it is known that the response variable depends on certain explanatory variables. Monotone neural networks can be used as powerful tools to build monotone models with better accuracy and lower variance compared to ordinary nonmonotone models. Monotonicity is usually obtained by putting constraints on the parameters of the network. In this paper, we will clarify some of the theoretical results on monotone neural networks with positive weights, issues that are sometimes misunderstood in the neural network literature. Furthermore, we will generalize some of the results obtained by Sill for the so-called min-max networks to the case of partially monotone problems. The method is illustrated in practical case studies."
                },
                {
                    "title": "Measuring Invariances in Deep Networks",
                    "abstract": "For many pattern recognition tasks, the ideal input feature would be invariant to multiple confounding properties (such as illumination and viewing angle, in computer vision applications). Recently, deep architectures trained in an unsupervised manner have been proposed as an automatic method for extracting useful features. However, it is difficult to evaluate the learned features by any means other than using them in a classifier. In this paper, we propose a number of empirical tests that directly measure the degree to which these learned features are invariant to different input transformations. We find that stacked autoencoders learn modestly increasingly invariant features with depth when trained on natural images. We find that convolutional deep belief networks learn substantially more invariant features in each layer. These results further justify the use of \"deep\" vs. \"shallower\" representations, but suggest that mechanisms beyond merely stacking one autoencoder on top of another may be important for achieving invariance. Our evaluation metrics can also be used to evaluate future work in deep learning, and thus help the development of future algorithms."
                },
                {
                    "title": "Probabilistic Graphical Models: Principles and Techniques - Adaptive Computation and Machine Learning",
                    "abstract": "Most tasks require a person or an automated system to reasonto reach conclusions based on available information. The framework of probabilistic graphical models, presented in this book, provides a general approach for this task. The approach is model-based, allowing interpretable models to be constructed and then manipulated by reasoning algorithms. These models can also be learned automatically from data, allowing the approach to be used in cases where manually constructing a model is difficult or even impossible. Because uncertainty is an inescapable aspect of most real-world applications, the book focuses on probabilistic models, which make the uncertainty explicit and provide models that are more faithful to reality. Probabilistic Graphical Models discusses a variety of models, spanning Bayesian networks, undirected Markov networks, discrete and continuous models, and extensions to deal with dynamical systems and relational data. For each class of models, the text describes the three fundamental cornerstones: representation, inference, and learning, presenting both basic concepts and advanced techniques. Finally, the book considers the use of the proposed framework for causal reasoning and decision making under uncertainty. The main text in each chapter provides the detailed technical development of the key ideas. Most chapters also include boxes with additional material: skill boxes, which describe techniques; case study boxes, which discuss empirical cases related to the approach described in the text, including applications in computer vision, robotics, natural language understanding, and computational biology; and concept boxes, which present significant concepts drawn from the material in the chapter. Instructors (and readers) can group chapters in various combinations, from core topics to more technically advanced material, to suit their particular needs. Adaptive Computation and Machine Learning series"
                },
                {
                    "title": "An Introduction to Many-Valued and Fuzzy Logic: Semantics, Algebras, and Derivation Systems",
                    "abstract": "Preface 1. Introduction 2. Review of classical propositional logic 3. Review of classical first-order logic 4. Alternative semantics for truth-values and truth-functions 5. Three-valued propositional logics: semantics 6. Derivation systems for three-valued propositional logic 7. Three-valued first-order logics: semantics 8. Derivation systems for three-valued first-order logics 9. Alternative semantics for three-valued systems 10. The principle of charity reconsidered and a new problem of the fringe 11. Fuzzy propositional logics: semantics 12. Fuzzy algebras 13. Derivational systems for fuzzy propositional logics 14. Fuzzy first-order logics: semantics 15. Derivation systems for fuzzy first-order logics 16. Extensions of fuzziness 17. Fuzzy membership functions."
                },
                {
                    "title": "An Investigation of the Laws of Thought: On which are founded the mathematical theories of logic and probabilities",
                    "abstract": "1. Nature and design of this work 2. Signs and their laws 3. Derivation of the laws 4. Division of propositions 5. Principles of symbolic reasoning 6. Of interpretation 7. Of elimination 8. Of reduction 9. Methods of abbreviation 10. Conditions of a perfect method 11. Of secondary propositions 12. Methods in secondary propositions 13. Clarke and Spinoza 14. Examples of analysis 15. Of the Aristotelian logic 16. Of the theory of probabilities 17. General method in probabilities 18. Elementary illustrations 19. Of statistical conditions 20. Problems on causes 21. Probability of judgements 22. Constitution of the intellect."
                },
                {
                    "title": "Integrating constraints and metric learning in semi-supervised clustering",
                    "abstract": "Semi-supervised clustering employs a small amount of labeled data to aid unsupervised learning. Previous work in the area has utilized supervised data in one of two approaches: 1) constraint-based methods that guide the clustering algorithm towards a better grouping of the data, and 2) distance-function learning methods that adapt the underlying similarity metric used by the clustering algorithm. This paper provides new methods for the two approaches as well as presents a new semi-supervised clustering algorithm that integrates both of these techniques in a uniform, principled framework. Experimental results demonstrate that the unified approach produces better clusters than both individual approaches as well as previously proposed semi-supervised clustering algorithms."
                },
                {
                    "title": "Sets for Mathematics",
                    "abstract": "Foreword 1. Abstract sets and mappings 2. Sums, monomorphisms and parts 3. Finite inverse limits 4. Colimits, epimorphisms and the axiom of choice 5. Mapping sets and exponentials 6. Summary of the axioms and an example of variable sets 7. Consequences and uses of exponentials 8. More on power sets 9. Introduction to variable sets 10. Models of additional variation Appendices Bibliography."
                },
                {
                    "title": "Monotonic Networks",
                    "abstract": "Monotonicity is a constraint which arises in many application domains. We present a machine learning model, the monotonic network, for which monotonicity can be enforced exactly, i.e., by virtue of functional form. A straightforward method for implementing and training a monotonic network is described. Monotonic networks are proven to be universal approximators of continuous, differentiable monotonic functions. We apply monotonic networks to a real-world task in corporate bond rating prediction and compare them to other approaches."
                },
                {
                    "title": "Taking categories seriously",
                    "abstract": "The theory of categories originated [1] with the need to guide complicated calculations involving passage to the limit in the study of the qualitative leap from spaces to homotopical/homological objects."
                }
            ],
            "categories": [
                "cs.LG",
                "math.CT",
                "math.LO"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nCan we measure and optimize properties such as injectivity in machine learning models beyond just their closeness to a fixed ground truth?\n\n**[Question 2] - Why is it interesting and important?**  \nAddressing this problem is crucial for advancing the field of machine learning, particularly in areas like representation learning where the quality of models cannot solely be evaluated by their accuracy against a ground truth. By developing methods to measure and optimize properties like injectivity, we can enhance our understanding of model behavior, leading to more robust and interpretable models. This could pave the way for practical applications in various domains, such as healthcare and finance, where understanding the underlying factors in data is essential.\n\n**[Question 3] - Why is it hard?**  \nThe challenge lies in the complexity of defining and quantifying properties like injectivity in a meaningful way. Naive approaches may fail because they often rely on direct comparisons to ground truth labels, which may not exist or be insufficient for evaluating model quality in more complex scenarios. Additionally, the theoretical underpinnings of these properties are not well-established, making it difficult to create metrics that are both mathematically sound and practically applicable. Overcoming these obstacles requires innovative methodologies and a deep understanding of both the mathematical and practical aspects of machine learning.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on accuracy and error metrics related to ground truth comparisons, neglecting the need for alternative quality measures like injectivity. This gap exists partly due to a lack of theoretical frameworks that adequately capture the nuances of model properties beyond accuracy. Additionally, existing solutions may not have considered the implications of these properties on model performance and interpretability. Our approach aims to fill this gap by proposing new metrics and methodologies that extend beyond traditional evaluation methods.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves developing a framework to measure injectivity and other relevant properties in machine learning models. We will utilize a diverse set of datasets to evaluate our metrics, focusing on representation learning tasks. The evaluation will be based on metrics that quantify the degree of injectivity and other properties, alongside traditional performance metrics. We expect our results to demonstrate that optimizing for these properties can lead to improved model performance and interpretability, ultimately contributing to more effective machine learning applications."
            }
        },
        "author_data": {
            "7d7f76f3-7279-4a99-bfc8-686f037b3fdc": {
                "pk": "7d7f76f3-7279-4a99-bfc8-686f037b3fdc",
                "name": "Yivan Zhang",
                "collaborators": [
                    "Masashi Sugiyama",
                    "Nontawat Charoenphakdee",
                    "Gang Niu",
                    "Zhenguo Wu",
                    "Jindong Wang",
                    "Xing Xie",
                    "Zhenghang Cui"
                ],
                "domain": [
                    "Weakly Supervised Learning",
                    "Label Noise",
                    "Classification",
                    "Disentanglement"
                ],
                "publications": [
                    {
                        "title": "Approximating Instance-Dependent Noise via Instance-Confidence Embedding",
                        "abstract": "Label noise in multiclass classification is a major obstacle to the deployment of learning systems. However, unlike the widely used class-conditional noise (CCN) assumption that the noisy label is independent of the input feature given the true label, label noise in real-world datasets can be aleatory and heavily dependent on individual instances. In this work, we investigate the instance-dependent noise (IDN) model and propose an efficient approximation of IDN to capture the instance-specific label corruption. Concretely, noting the fact that most columns of the IDN transition matrix have only limited influence on the class-posterior estimation, we propose a variational approximation that uses a single-scalar confidence parameter. To cope with the situation where the mapping from the instance to its confidence value could vary significantly for two adjacent instances, we suggest using instance embedding that assigns a trainable parameter to each instance. The resulting instance-confidence embedding (ICE) method not only performs well under label noise but also can effectively detect ambiguous or mislabeled instances. We validate its utility on various image and text classification tasks."
                    },
                    {
                        "title": "A Category-theoretical Meta-analysis of Definitions of Disentanglement",
                        "abstract": "Disentangling the factors of variation in data is a fundamental concept in machine learning and has been studied in various ways by different researchers, leading to a multitude of definitions. Despite the numerous empirical studies, more theoretical research is needed to fully understand the defining properties of disentanglement and how different definitions relate to each other. This paper presents a meta-analysis of existing definitions of disentanglement, using category theory as a unifying and rigorous framework. We propose that the concepts of the cartesian and monoidal products should serve as the core of disentanglement. With these core concepts, we show the similarities and crucial differences in dealing with (i) functions, (ii) equivariant maps, (iii) relations, and (iv) stochastic maps. Overall, our meta-analysis deepens our understanding of disentanglement and its various formulations and can help researchers navigate different definitions and choose the most appropriate one for their specific context."
                    },
                    {
                        "title": "Learning Noise Transition Matrix from Only Noisy Labels via Total Variation Regularization",
                        "abstract": "Many weakly supervised classification methods employ a noise transition matrix to capture the class-conditional label corruption. To estimate the transition matrix from noisy data, existing methods often need to estimate the noisy class-posterior, which could be unreliable due to the overconfidence of neural networks. In this work, we propose a theoretically grounded method that can estimate the noise transition matrix and learn a classifier simultaneously, without relying on the error-prone noisy class-posterior estimation. Concretely, inspired by the characteristics of the stochastic label corruption process, we propose total variation regularization, which encourages the predicted probabilities to be more distinguishable from each other. Under mild assumptions, the proposed method yields a consistent estimator of the transition matrix. We show the effectiveness of the proposed method through experiments on benchmark and real-world datasets."
                    },
                    {
                        "title": "Learning from Indirect Observations",
                        "abstract": "Weakly-supervised learning is a paradigm for alleviating the scarcity of labeled data by leveraging lower-quality but larger-scale supervision signals. While existing work mainly focuses on utilizing a certain type of weak supervision, we present a probabilistic framework, learning from indirect observations, for learning from a wide range of weak supervision in real-world problems, e.g., noisy labels, complementary labels and coarse-grained labels. We propose a general method based on the maximum likelihood principle, which has desirable theoretical properties and can be straightforwardly implemented for deep neural networks. Concretely, a discriminative model for the true target is used for modeling the indirect observation, which is a random variable entirely depending on the true target stochastically or deterministically. Then, maximizing the likelihood given indirect observations leads to an estimator of the true target implicitly. Comprehensive experiments for two novel problem settings --- learning from multiclass label proportions and learning from coarse-grained labels, illustrate practical usefulness of our method and demonstrate how to integrate various sources of weak supervision."
                    },
                    {
                        "title": "Learning from Aggregate Observations",
                        "abstract": "We study the problem of learning from aggregate observations where supervision signals are given to sets of instances instead of individual instances, while the goal is still to predict labels of unseen individuals. A well-known example is multiple instance learning (MIL). In this paper, we extend MIL beyond binary classification to other problems such as multiclass classification and regression. We present a general probabilistic framework that accommodates a variety of aggregate observations, e.g., pairwise similarity/triplet comparison for classification and mean/difference/rank observation for regression. Simple maximum likelihood solutions can be applied to various differentiable models such as deep neural networks and gradient boosting machines. Moreover, we develop the concept of consistency up to an equivalence relation to characterize our estimator and show that it has nice convergence properties under mild assumptions. Experiments on three problem settings -- classification via triplet comparison and regression via mean/rank observation indicate the effectiveness of the proposed method."
                    },
                    {
                        "title": "Equivariant Disentangled Transformation for Domain Generalization under Combination Shift",
                        "abstract": "Machine learning systems may encounter unexpected problems when the data distribution changes in the deployment environment. A major reason is that certain combinations of domains and labels are not observed during training but appear in the test environment. Although various invariance-based algorithms can be applied, we find that the performance gain is often marginal. To formally analyze this issue, we provide a unique algebraic formulation of the combination shift problem based on the concepts of homomorphism, equivariance, and a refined definition of disentanglement. The algebraic requirements naturally derive a simple yet effective method, referred to as equivariant disentangled transformation (EDT), which augments the data based on the algebraic structures of labels and makes the transformation satisfy the equivariance and disentanglement requirements. Experimental results demonstrate that invariance may be insufficient, and it is important to exploit the equivariance structure in the combination shift problem."
                    },
                    {
                        "title": "Classification with Rejection Based on Cost-sensitive Classification",
                        "abstract": "The goal of classification with rejection is to avoid risky misclassification in error-critical applications such as medical diagnosis and product inspection. In this paper, based on the relationship between classification with rejection and cost-sensitive classification, we propose a novel method of classification with rejection by learning an ensemble of cost-sensitive classifiers, which satisfies all the following properties: (i) it can avoid estimating class-posterior probabilities, resulting in improved classification accuracy, (ii) it allows a flexible choice of losses including non-convex ones, (iii) it does not require complicated modifications when using different losses, (iv) it is applicable to both binary and multiclass cases, and (v) it is theoretically justifiable for any classification-calibrated loss. Experimental results demonstrate the usefulness of our proposed approach in clean-labeled, noisy-labeled, and positive-unlabeled classification."
                    }
                ]
            },
            "e81001b3-e675-4ca4-8ed5-517d83bee97e": {
                "pk": "e81001b3-e675-4ca4-8ed5-517d83bee97e",
                "name": "Masashi Sugiyama",
                "collaborators": [
                    "Shintaro Nakamura",
                    "Paavo Parmas",
                    "Yivan Zhang",
                    "Ryota Tomioka",
                    "Hao Zhang",
                    "Makoto Yamada",
                    "Taiji Suzuki",
                    "Florian Yger",
                    "Takayuki Osa",
                    "Nontawat Charoenphakdee"
                ],
                "domain": [
                    "Machine Learning",
                    "Reinforcement Learning",
                    "Signal Processing",
                    "Metric Learning"
                ],
                "publications": [
                    {
                        "title": "Dual Augmented Lagrangian Method for Efficient Sparse Reconstruction",
                        "abstract": "We propose an efficient algorithm for sparse signal reconstruction problems. The proposed algorithm is an augmented Lagrangian method based on the dual sparse reconstruction problem. It is efficient when the number of unknown variables is much larger than the number of observations because of the dual formulation. Moreover, the primal variable is explicitly updated and the sparsity in the solution is exploited. Numerical comparison with the state-of-the-art algorithms shows that the proposed algorithm is favorable when the design matrix is poorly conditioned or dense and very large."
                    },
                    {
                        "title": "Task Selection for Bandit-Based Task Assignment in Heterogeneous Crowdsourcing",
                        "abstract": "Task selection (picking an appropriate labeling task) and worker selection (assigning the labeling task to a suitable worker) are two major challenges in task assignment for crowdsourcing. Recently, worker selection has been successfully addressed by the bandit-based task assignment (BBTA) method, while task selection has not been thoroughly investigated yet. In this paper, we experimentally compare several task selection strategies borrowed from active learning literature, and show that the least confidence strategy significantly improves the performance of task assignment in crowdsourcing."
                    },
                    {
                        "title": "Cross-Domain Object Matching with Model Selection",
                        "abstract": "The goal of cross-domain object matching (CDOM) is to find correspondence between two sets of objects in different domains in an unsupervised way. Photo album summarization is a typical application of CDOM, where photos are automatically aligned into a designed frame expressed in the Cartesian coordinate system. CDOM is usually formulated as finding a mapping from objects in one domain (photos) to objects in the other domain (frame) so that the pairwise dependency is maximized. A state-of-the-art CDOM method employs a kernel-based dependency measure, but it has a drawback that the kernel parameter needs to be determined manually. In this paper, we propose alternative CDOM methods that can naturally address the model selection problem. Through experiments on image matching, unpaired voice conversion, and photo album summarization tasks, the effectiveness of the proposed methods is demonstrated."
                    },
                    {
                        "title": "Fast learning rate of multiple kernel learning: Trade-off between sparsity and smoothness",
                        "abstract": "We investigate the learning rate of multiple kernel learning (MKL) with $\\ell_1$ and elastic-net regularizations. The elastic-net regularization is a composition of an $\\ell_1$-regularizer for inducing the sparsity and an $\\ell_2$-regularizer for controlling the smoothness. We focus on a sparse setting where the total number of kernels is large, but the number of nonzero components of the ground truth is relatively small, and show sharper convergence rates than the learning rates have ever shown for both $\\ell_1$ and elastic-net regularizations. Our analysis reveals some relations between the choice of a regularization function and the performance. If the ground truth is smooth, we show a faster convergence rate for the elastic-net regularization with less conditions than $\\ell_1$-regularization; otherwise, a faster convergence rate for the $\\ell_1$-regularization is shown."
                    },
                    {
                        "title": "Supervised LogEuclidean Metric Learning for Symmetric Positive Definite Matrices",
                        "abstract": "Metric learning has been shown to be highly effective to improve the performance of nearest neighbor classification. In this paper, we address the problem of metric learning for Symmetric Positive Definite (SPD) matrices such as covariance matrices, which arise in many real-world applications. Naively using standard Mahalanobis metric learning methods under the Euclidean geometry for SPD matrices is not appropriate, because the difference of SPD matrices can be a non-SPD matrix and thus the obtained solution can be uninterpretable. To cope with this problem, we propose to use a properly parameterized LogEuclidean distance and optimize the metric with respect to kernel-target alignment, which is a supervised criterion for kernel learning. Then the resulting non-trivial optimization problem is solved by utilizing the Riemannian geometry. Finally, we experimentally demonstrate the usefulness of our LogEuclidean metric learning algorithm on real-world classification tasks for EEG signals and texture patches."
                    },
                    {
                        "title": "Hierarchical Policy Search via Return-Weighted Density Estimation",
                        "abstract": "Learning an optimal policy from a multi-modal reward function is a challenging problem in reinforcement learning (RL). Hierarchical RL (HRL) tackles this problem by learning a hierarchical policy, where multiple option policies are in charge of different strategies corresponding to modes of a reward function and a gating policy selects the best option for a given context. Although HRL has been demonstrated to be promising, current state-of-the-art methods cannot still perform well in complex real-world problems due to the difficulty of identifying modes of the reward function. In this paper, we propose a novel method called hierarchical policy search via return-weighted density estimation (HPSDE), which can efficiently identify the modes through density estimation with return-weighted importance sampling. Our proposed method finds option policies corresponding to the modes of the return function and automatically determines the number and the location of option policies, which significantly reduces the burden of hyper-parameters tuning. Through experiments, we demonstrate that the proposed HPSDE successfully learns option policies corresponding to modes of the return function and that it can be successfully applied to a challenging motion planning problem of a redundant robotic manipulator."
                    },
                    {
                        "title": "Positive-Unlabeled Classification under Class Prior Shift and Asymmetric Error",
                        "abstract": "Bottlenecks of binary classification from positive and unlabeled data (PU classification) are the requirements that given unlabeled patterns are drawn from the test marginal distribution, and the penalty of the false positive error is identical to the false negative error. However, such requirements are often not fulfilled in practice. In this paper, we generalize PU classification to the class prior shift and asymmetric error scenarios. Based on the analysis of the Bayes optimal classifier, we show that given a test class prior, PU classification under class prior shift is equivalent to PU classification with asymmetric error. Then, we propose two different frameworks to handle these problems, namely, a risk minimization framework and density ratio estimation framework. Finally, we demonstrate the effectiveness of the proposed frameworks and compare both frameworks through experiments using benchmark datasets."
                    },
                    {
                        "title": "Calibrated Surrogate Maximization of Linear-fractional Utility in Binary Classification",
                        "abstract": "Complex classification performance metrics such as the F${}_\\beta$-measure and Jaccard index are often used, in order to handle class-imbalanced cases such as information retrieval and image segmentation. These performance metrics are not decomposable, that is, they cannot be expressed in a per-example manner, which hinders a straightforward application of M-estimation widely used in supervised learning. In this paper, we consider linear-fractional metrics, which are a family of classification performance metrics that encompasses many standard ones such as the F${}_\\beta$-measure and Jaccard index, and propose methods to directly maximize performances under those metrics. A clue to tackle their direct optimization is a calibrated surrogate utility, which is a tractable lower bound of the true utility function representing a given metric. We characterize sufficient conditions which make the surrogate maximization coincide with the maximization of the true utility. Simulation results on benchmark datasets validate the effectiveness of our calibrated surrogate maximization especially if the sample sizes are extremely small."
                    },
                    {
                        "title": "A unified view of likelihood ratio and reparameterization gradients and an optimal importance sampling scheme",
                        "abstract": "Reparameterization (RP) and likelihood ratio (LR) gradient estimators are used throughout machine and reinforcement learning; however, they are usually explained as simple mathematical tricks without providing any insight into their nature. We use a first principles approach to explain LR and RP, and show a connection between the two via the divergence theorem. The theory motivated us to derive optimal importance sampling schemes to reduce LR gradient variance. Our newly derived distributions have analytic probability densities and can be directly sampled from. The improvement for Gaussian target distributions was modest, but for other distributions such as a Beta distribution, our method could lead to arbitrarily large improvements, and was crucial to obtain competitive performance in evolution strategies experiments."
                    },
                    {
                        "title": "A unified view of likelihood ratio and reparameterization gradients",
                        "abstract": "Reparameterization (RP) and likelihood ratio (LR) gradient estimators are used to estimate gradients of expectations throughout machine learning and reinforcement learning; however, they are usually explained as simple mathematical tricks, with no insight into their nature. We use a first principles approach to explain that LR and RP are alternative methods of keeping track of the movement of probability mass, and the two are connected via the divergence theorem. Moreover, we show that the space of all possible estimators combining LR and RP can be completely parameterized by a flow field $u(x)$ and an importance sampling distribution $q(x)$. We prove that there cannot exist a single-sample estimator of this type outside our characterized space, thus, clarifying where we should be searching for better Monte Carlo gradient estimators."
                    },
                    {
                        "title": "Positive-Unlabeled Classification under Class-Prior Shift: A Prior-invariant Approach Based on Density Ratio Estimation",
                        "abstract": "Learning from positive and unlabeled (PU) data is an important problem in various applications. Most of the recent approaches for PU classification assume that the class-prior (the ratio of positive samples) in the training unlabeled dataset is identical to that of the test data, which does not hold in many practical cases. In addition, we usually do not know the class-priors of the training and test data, thus we have no clue on how to train a classifier without them. To address these problems, we propose a novel PU classification method based on density ratio estimation. A notable advantage of our proposed method is that it does not require the class-priors in the training phase; class-prior shift is incorporated only in the test phase. We theoretically justify our proposed method and experimentally demonstrate its effectiveness."
                    },
                    {
                        "title": "Source-free Domain Adaptation via Distributional Alignment by Matching Batch Normalization Statistics",
                        "abstract": "In this paper, we propose a novel domain adaptation method for the source-free setting. In this setting, we cannot access source data during adaptation, while unlabeled target data and a model pretrained with source data are given. Due to lack of source data, we cannot directly match the data distributions between domains unlike typical domain adaptation algorithms. To cope with this problem, we propose utilizing batch normalization statistics stored in the pretrained model to approximate the distribution of unobserved source data. Specifically, we fix the classifier part of the model during adaptation and only fine-tune the remaining feature encoder part so that batch normalization statistics of the features extracted by the encoder match those stored in the fixed classifier. Additionally, we also maximize the mutual information between the features and the classifier's outputs to further boost the classification performance. Experimental results with several benchmark datasets show that our method achieves competitive performance with state-of-the-art domain adaptation methods even though it does not require access to source data."
                    },
                    {
                        "title": "Incorporating Causal Graphical Prior Knowledge into Predictive Modeling via Simple Data Augmentation",
                        "abstract": "Causal graphs (CGs) are compact representations of the knowledge of the data generating processes behind the data distributions. When a CG is available, e.g., from the domain knowledge, we can infer the conditional independence (CI) relations that should hold in the data distribution. However, it is not straightforward how to incorporate this knowledge into predictive modeling. In this work, we propose a model-agnostic data augmentation method that allows us to exploit the prior knowledge of the CI encoded in a CG for supervised machine learning. We theoretically justify the proposed method by providing an excess risk bound indicating that the proposed method suppresses overfitting by reducing the apparent complexity of the predictor hypothesis class. Using real-world data with CGs provided by domain experts, we experimentally show that the proposed method is effective in improving the prediction accuracy, especially in the small-data regime."
                    },
                    {
                        "title": "Approximating Instance-Dependent Noise via Instance-Confidence Embedding",
                        "abstract": "Label noise in multiclass classification is a major obstacle to the deployment of learning systems. However, unlike the widely used class-conditional noise (CCN) assumption that the noisy label is independent of the input feature given the true label, label noise in real-world datasets can be aleatory and heavily dependent on individual instances. In this work, we investigate the instance-dependent noise (IDN) model and propose an efficient approximation of IDN to capture the instance-specific label corruption. Concretely, noting the fact that most columns of the IDN transition matrix have only limited influence on the class-posterior estimation, we propose a variational approximation that uses a single-scalar confidence parameter. To cope with the situation where the mapping from the instance to its confidence value could vary significantly for two adjacent instances, we suggest using instance embedding that assigns a trainable parameter to each instance. The resulting instance-confidence embedding (ICE) method not only performs well under label noise but also can effectively detect ambiguous or mislabeled instances. We validate its utility on various image and text classification tasks."
                    },
                    {
                        "title": "Audio Signal Enhancement with Learning from Positive and Unlabelled Data",
                        "abstract": "Supervised learning is a mainstream approach to audio signal enhancement (SE) and requires parallel training data consisting of both noisy signals and the corresponding clean signals. Such data can only be synthesised and are mismatched with real data, which can result in poor performance on real data. Moreover, clean signals may be inaccessible in certain scenarios, which renders this conventional approach infeasible. Here we explore SE using non-parallel training data consisting of noisy signals and noise, which can be easily recorded. We define the positive (P) and the negative (N) classes as signal inactivity and activity, respectively. We observe that the spectrogram patches of noise clips can be used as P data and those of noisy signal clips as unlabelled data. Thus, learning from positive and unlabelled data enables a convolutional neural network to learn to classify each spectrogram patch as P or N to enable SE."
                    },
                    {
                        "title": "A Category-theoretical Meta-analysis of Definitions of Disentanglement",
                        "abstract": "Disentangling the factors of variation in data is a fundamental concept in machine learning and has been studied in various ways by different researchers, leading to a multitude of definitions. Despite the numerous empirical studies, more theoretical research is needed to fully understand the defining properties of disentanglement and how different definitions relate to each other. This paper presents a meta-analysis of existing definitions of disentanglement, using category theory as a unifying and rigorous framework. We propose that the concepts of the cartesian and monoidal products should serve as the core of disentanglement. With these core concepts, we show the similarities and crucial differences in dealing with (i) functions, (ii) equivariant maps, (iii) relations, and (iv) stochastic maps. Overall, our meta-analysis deepens our understanding of disentanglement and its various formulations and can help researchers navigate different definitions and choose the most appropriate one for their specific context."
                    },
                    {
                        "title": "An Optimal Algorithm for the Real-Valued Combinatorial Pure Exploration of Multi-Armed Bandit",
                        "abstract": "We study the real-valued combinatorial pure exploration problem in the stochastic multi-armed bandit (R-CPE-MAB). We study the case where the size of the action set is polynomial with respect to the number of arms. In such a case, the R-CPE-MAB can be seen as a special case of the so-called transductive linear bandits. Existing methods in the R-CPE-MAB and transductive linear bandits have a gap of problem-dependent constant terms and logarithmic terms between the upper and lower bounds of the sample complexity, respectively. We close these gaps by proposing an algorithm named the combinatorial gap-based exploration (CombGapE) algorithm, whose sample complexity upper bound matches the lower bound. Finally, we numerically show that the CombGapE algorithm outperforms existing methods significantly."
                    },
                    {
                        "title": "Thompson Sampling for Real-Valued Combinatorial Pure Exploration of Multi-Armed Bandit",
                        "abstract": "We study the real-valued combinatorial pure exploration of the multi-armed bandit (R-CPE-MAB) problem. In R-CPE-MAB, a player is given $d$ stochastic arms, and the reward of each arm $s\\in\\{1, \\ldots, d\\}$ follows an unknown distribution with mean $\\mu_s$. In each time step, a player pulls a single arm and observes its reward. The player's goal is to identify the optimal \\emph{action} $\\boldsymbol{\\pi}^{*} = \\argmax_{\\boldsymbol{\\pi} \\in \\mathcal{A}} \\boldsymbol{\\mu}^{\\top}\\boldsymbol{\\pi}$ from a finite-sized real-valued \\emph{action set} $\\mathcal{A}\\subset \\mathbb{R}^{d}$ with as few arm pulls as possible. Previous methods in the R-CPE-MAB assume that the size of the action set $\\mathcal{A}$ is polynomial in $d$. We introduce an algorithm named the Generalized Thompson Sampling Explore (GenTS-Explore) algorithm, which is the first algorithm that can work even when the size of the action set is exponentially large in $d$. We also introduce a novel problem-dependent sample complexity lower bound of the R-CPE-MAB problem, and show that the GenTS-Explore algorithm achieves the optimal sample complexity up to a problem-dependent constant factor."
                    },
                    {
                        "title": "Fixed-Budget Real-Valued Combinatorial Pure Exploration of Multi-Armed Bandit",
                        "abstract": "We study the real-valued combinatorial pure exploration of the multi-armed bandit in the fixed-budget setting. We first introduce the Combinatorial Successive Asign (CSA) algorithm, which is the first algorithm that can identify the best action even when the size of the action class is exponentially large with respect to the number of arms. We show that the upper bound of the probability of error of the CSA algorithm matches a lower bound up to a logarithmic factor in the exponent. Then, we introduce another algorithm named the Minimax Combinatorial Successive Accepts and Rejects (Minimax-CombSAR) algorithm for the case where the size of the action class is polynomial, and show that it is optimal, which matches a lower bound. Finally, we experimentally compare the algorithms with previous methods and show that our algorithm performs better."
                    }
                ]
            }
        }
    },
    "2405.12489": {
        "paper_data": {
            "title": "Exploring and Exploiting the Asymmetric Valley of Deep Neural Networks",
            "url": "http://arxiv.org/abs/2405.12489v4",
            "arxiv_id": "2405.12489",
            "authors": [
                "Xin-Chun Li",
                "Jin-Lin Tang",
                "Bo Zhang",
                "Lan Li",
                "De-Chuan Zhan"
            ],
            "abstract": "Exploring the loss landscape offers insights into the inherent principles of deep neural networks (DNNs). Recent work suggests an additional asymmetry of the valley beyond the flat and sharp ones, yet without thoroughly examining its causes or implications. Our study methodically explores the factors affecting the symmetry of DNN valleys, encompassing (1) the dataset, network architecture, initialization, and hyperparameters that influence the convergence point; and (2) the magnitude and direction of the noise for 1D visualization. Our major observation shows that the {\\it degree of sign consistency} between the noise and the convergence point is a critical indicator of valley symmetry. Theoretical insights from the aspects of ReLU activation and softmax function could explain the interesting phenomenon. Our discovery propels novel understanding and applications in the scenario of Model Fusion: (1) the efficacy of interpolating separate models significantly correlates with their sign consistency ratio, and (2) imposing sign alignment during federated learning emerges as an innovative approach for model parameter alignment.",
            "introduction": "   1 Introduction  The massive number of parameters and complex structure of deep neural networks (DNNs) have catalyzed extensive research to mine their underlying mechanics\u00a0[54, 33, 41]. Visualizing and exploring the loss surfaces of DNNs is the most intuitive way\u00a0[42, 18], which has ignited many interesting findings, such as the monotonic linear interpolation\u00a0[22, 19, 64] and linear mode connectivity\u00a0[22, 14, 21, 15, 63, 2]. Loss landscape visualization has also been applied to show the optimization trajectory\u00a0[52, 32, 34], understand the effectiveness of Batch Normalization\u00a0[31, 60], BERT\u00a0[11, 23], deep ensemble\u00a0[17, 29, 21], and so on.   Perturbation analysis around the local minima of DNNs\u00a0[7, 61], i.e., the shape of the valleys they reside in, is a very popular research topic. The concept of flat minima was originally proposed by\u00a0[27], who defines the size of the connected region around the minima where the loss remains relatively unchanged as flatness. Subsequent studies debate whether the flat or sharp minima could reflect the generalization ability\u00a0[37, 42, 35, 56, 16, 12, 40, 3]. The previous works constrain the valley shape to be symmetric, while recent work points out that not all DNN valleys are flat or sharp, and there also exist asymmetric valleys\u00a0[24], which has not been systematically studied as far as we know.   This paper in-depth analyzes the factors that may affect the valley symmetry of DNNs. Previous work\u2019s analysis of valley shape primarily utilizes the 1D interpolation of \u03b8f+\u03bb\u2062\u03f5subscript\ud835\udf03\ud835\udc53\ud835\udf06italic-\u03f5\\theta_{f}+\\lambda\\epsilonitalic_\u03b8 start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT + italic_\u03bb italic_\u03f5, where \u03b8fsubscript\ud835\udf03\ud835\udc53\\theta_{f}italic_\u03b8 start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT represents the minima solution and \u03f5italic-\u03f5\\epsilonitalic_\u03f5 denotes a random noise. As shown in Fig.\u00a02, we believe that the valley symmetry depends both on the convergence solution and noise, with each of them being influenced by some factors. The most significant innovation in our research is considering the effect of noise direction on valley visualization, as previous work has simply taken the Gaussian noise.   Specifically, we start by comprehensively and carefully determining the visualization method to plot the valley shape, which could influence the conclusion significantly\u00a0[30, 42, 40]. We finally take the Norm-Scaled (NS) visualization method\u00a0[30] that normalizes the noise \u03f5italic-\u03f5\\epsilonitalic_\u03f5 and further scales it to \u2016\u03b8f\u2016normsubscript\ud835\udf03\ud835\udc53||\\theta_{f}||| | italic_\u03b8 start_POSTSUBSCRIPT italic_f end_POSTSUBSCRIPT | | for better determining the plot range of \u03bb\ud835\udf06\\lambdaitalic_\u03bb, where the direction of raw noise is not changed. Then, after investigating 7 common noise directions and 6 special ones, we conclude: the degree of sign consistency between the noise and the convergence solution should be a determining factor for asymmetry. This phenomenon is basically insensitive to the utilized datasets. Next, we focus on the impact of the network architecture with or without Batch Normalization (BN)\u00a0[31], indicating that the BN initialization also impacts the valley symmetry. Finally, different hyperparameters lead to solutions with various valley widths but show no asymmetry consistently.   Aside from empirical observations, theoretical insights for our interesting findings are provided. We first declare that adding sign-consistent noise to parameters may have a larger probability of keeping activating the neurons or keeping the overwhelming score in classification tasks. Then, we show that the trace of the Hessian matrix along the sign-consistent direction is smaller, implying a flatter region. The",
            "references": [
                {
                    "title": "Visualizing, Rethinking, and Mining the Loss Landscape of Deep Neural Networks",
                    "abstract": "The loss landscape of deep neural networks (DNNs) is commonly considered complex and wildly fluctuated. However, an interesting observation is that the loss surfaces plotted along Gaussian noise directions are almost v-basin ones with the perturbed model lying on the basin. This motivates us to rethink whether the 1D or 2D subspace could cover more complex local geometry structures, and how to mine the corresponding perturbation directions. This paper systematically and gradually categorizes the 1D curves from simple to complex, including v-basin, v-side, w-basin, w-peak, and vvv-basin curves. Notably, the latter two types are already hard to obtain via the intuitive construction of specific perturbation directions, and we need to propose proper mining algorithms to plot the corresponding 1D curves. Combining these 1D directions, various types of 2D surfaces are visualized such as the saddle surfaces and the bottom of a bottle of wine that are only shown by demo functions in previous works. Finally, we propose theoretical insights from the lens of the Hessian matrix to explain the observed several interesting phenomena."
                },
                {
                    "title": "MAP: Model Aggregation and Personalization in Federated Learning With Incomplete Classes",
                    "abstract": "In some real-world applications, data samples are usually distributed on local devices, where federated learning (FL) techniques are proposed to coordinate decentralized clients without directly sharing users\u2019 private data. FL commonly follows the parameter server architecture and contains multiple personalization and aggregation procedures. The natural data heterogeneity across clients, i.e., Non-I.I.D. data, challenges both the aggregation and personalization goals in FL. In this paper, we focus on a special kind of Non-I.I.D. scene where clients own incomplete classes, i.e., each client can only access a partial set of the whole class set. The server aims to aggregate a complete classification model that could generalize to all classes, while the clients are inclined to improve the performance of distinguishing their observed classes. For better model aggregation, we point out that the standard softmax will encounter several problems caused by missing classes and propose \u201crestricted softmax\u201d as an alternative. For better model personalization, we point out that the hard-won personalized models are not well exploited and propose \u201cinherited private model\u201d to store the personalization experience. Our proposed algorithm named MAP could simultaneously achieve the aggregation and personalization goals in FL. Abundant experimental studies verify the superiorities of our algorithm."
                },
                {
                    "title": "Deep Model Fusion: A Survey",
                    "abstract": "Deep model fusion/merging is an emerging technique that merges the parameters or predictions of multiple deep learning models into a single one. It combines the abilities of different models to make up for the biases and errors of a single model to achieve better performance. However, deep model fusion on large-scale deep learning models (e.g., LLMs and foundation models) faces several challenges, including high computational cost, high-dimensional parameter space, interference between different heterogeneous models, etc. Although model fusion has attracted widespread attention due to its potential to solve complex real-world tasks, there is still a lack of complete and detailed survey research on this technique. Accordingly, in order to understand the model fusion method better and promote its development, we present a comprehensive survey to summarize the recent progress. Specifically, we categorize existing deep model fusion methods as four-fold: (1)\"Mode connectivity\", which connects the solutions in weight space via a path of non-increasing loss, in order to obtain better initialization for model fusion; (2)\"Alignment\"matches units between neural networks to create better conditions for fusion; (3)\"Weight average\", a classical model fusion method, averages the weights of multiple models to obtain more accurate results closer to the optimal solution; (4)\"Ensemble learning\"combines the outputs of diverse models, which is a foundational technique for improving the accuracy and robustness of the final model. In addition, we analyze the challenges faced by deep model fusion and propose possible research directions for model fusion in the future. Our review is helpful in deeply understanding the correlation between different model fusion methods and practical application methods, which can enlighten the research in the field of deep model fusion."
                },
                {
                    "title": "TIES-Merging: Resolving Interference When Merging Models",
                    "abstract": "Transfer learning - i.e., further fine-tuning a pre-trained model on a downstream task - can confer significant advantages, including improved downstream performance, faster convergence, and better sample efficiency. These advantages have led to a proliferation of task-specific fine-tuned models, which typically can only perform a single task and do not benefit from one another. Recently, model merging techniques have emerged as a solution to combine multiple task-specific models into a single multitask model without performing additional training. However, existing merging methods often ignore the interference between parameters of different models, resulting in large performance drops when merging multiple models. In this paper, we demonstrate that prior merging techniques inadvertently lose valuable information due to two major sources of interference: (a) interference due to redundant parameter values and (b) disagreement on the sign of a given parameter's values across models. To address this, we propose our method, TRIM, ELECT SIGN&MERGE (TIES-Merging), which introduces three novel steps when merging models: (1) resetting parameters that only changed a small amount during fine-tuning, (2) resolving sign conflicts, and (3) merging only the parameters that are in alignment with the final agreed-upon sign. We find that TIES-Merging outperforms several existing methods in diverse settings covering a range of modalities, domains, number of tasks, model sizes, architectures, and fine-tuning settings. We further analyze the impact of different types of interference on model parameters, and highlight the importance of resolving sign interference. Our code is available at https://github.com/prateeky2806/ties-merging"
                },
                {
                    "title": "A modern look at the relationship between sharpness and generalization",
                    "abstract": "Sharpness of minima is a promising quantity that can correlate with generalization in deep networks and, when optimized during training, can improve generalization. However, standard sharpness is not invariant under reparametrizations of neural networks, and, to fix this, reparametrization-invariant sharpness definitions have been proposed, most prominently adaptive sharpness (Kwon et al., 2021). But does it really capture generalization in modern practical settings? We comprehensively explore this question in a detailed study of various definitions of adaptive sharpness in settings ranging from training from scratch on ImageNet and CIFAR-10 to fine-tuning CLIP on ImageNet and BERT on MNLI. We focus mostly on transformers for which little is known in terms of sharpness despite their widespread usage. Overall, we observe that sharpness does not correlate well with generalization but rather with some training parameters like the learning rate that can be positively or negatively correlated with generalization depending on the setup. Interestingly, in multiple cases, we observe a consistent negative correlation of sharpness with out-of-distribution error implying that sharper minima can generalize better. Finally, we illustrate on a simple model that the right sharpness measure is highly data-dependent, and that we do not understand well this aspect for realistic data distributions. The code of our experiments is available at https://github.com/tml-epfl/sharpness-vs-generalization."
                },
                {
                    "title": "Asymmetric Temperature Scaling Makes Larger Networks Teach Well Again",
                    "abstract": "Knowledge Distillation (KD) aims at transferring the knowledge of a well-performed neural network (the {\\it teacher}) to a weaker one (the {\\it student}). A peculiar phenomenon is that a more accurate model doesn't necessarily teach better, and temperature adjustment can neither alleviate the mismatched capacity. To explain this, we decompose the efficacy of KD into three parts: {\\it correct guidance}, {\\it smooth regularization}, and {\\it class discriminability}. The last term describes the distinctness of {\\it wrong class probabilities} that the teacher provides in KD. Complex teachers tend to be over-confident and traditional temperature scaling limits the efficacy of {\\it class discriminability}, resulting in less discriminative wrong class probabilities. Therefore, we propose {\\it Asymmetric Temperature Scaling (ATS)}, which separately applies a higher/lower temperature to the correct/wrong class. ATS enlarges the variance of wrong class probabilities in the teacher's label and makes the students grasp the absolute affinities of wrong classes to the target class as discriminative as possible. Both theoretical analysis and extensive experimental results demonstrate the effectiveness of ATS. The demo developed in Mindspore is available at https://gitee.com/lxcnju/ats-mindspore and will be available at https://gitee.com/mindspore/models/tree/master/research/cv/ats."
                },
                {
                    "title": "Git Re-Basin: Merging Models modulo Permutation Symmetries",
                    "abstract": "The success of deep learning is due in large part to our ability to solve certain massive non-convex optimization problems with relative ease. Though non-convex optimization is NP-hard, simple algorithms -- often variants of stochastic gradient descent -- exhibit surprising effectiveness in fitting large neural networks in practice. We argue that neural network loss landscapes often contain (nearly) a single basin after accounting for all possible permutation symmetries of hidden units a la Entezari et al. 2021. We introduce three algorithms to permute the units of one model to bring them into alignment with a reference model in order to merge the two models in weight space. This transformation produces a functionally equivalent set of weights that lie in an approximately convex basin near the reference model. Experimentally, we demonstrate the single basin phenomenon across a variety of model architectures and datasets, including the first (to our knowledge) demonstration of zero-barrier linear mode connectivity between independently trained ResNet models on CIFAR-10. Additionally, we identify intriguing phenomena relating model width and training time to mode connectivity. Finally, we discuss shortcomings of the linear mode connectivity hypothesis, including a counterexample to the single basin theory."
                },
                {
                    "title": "Federated Learning with Position-Aware Neurons",
                    "abstract": "Federated Learning (FL) fuses collaborative models from local nodes without centralizing users' data. The permutation invariance property of neural networks and the non-i.i.d. data across clients make the locally updated parameters imprecisely aligned, disabling the coordinate-based parameter averaging. Traditional neurons do not explicitly consider position information. Hence, we propose Position-Aware Neurons (PANs) as an alternative, fusing position-related values (i.e., position encodings) into neuron outputs. PANs couple themselves to their positions and minimize the possibility of dislocation, even updating on heterogeneous data. We turn on/off PANs to disable/enable the permutation invariance property of neural networks. PANs are tightly coupled with positions when applied to FL, making parameters across clients pre-aligned and facilitating coordinate-based parameter averaging. PANs are algorithm-agnostic and could universally improve existing FL algorithms. Furthermore, \u201cFL with PANs\u201d is simple to implement and computationally friendly."
                },
                {
                    "title": "Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time",
                    "abstract": "The conventional recipe for maximizing model accuracy is to (1) train multiple models with various hyperparameters and (2) pick the individual model which performs best on a held-out validation set, discarding the remainder. In this paper, we revisit the second step of this procedure in the context of fine-tuning large pre-trained models, where fine-tuned models often appear to lie in a single low error basin. We show that averaging the weights of multiple models fine-tuned with different hyperparameter configurations often improves accuracy and robustness. Unlike a conventional ensemble, we may average many models without incurring any additional inference or memory costs -- we call the results\"model soups.\"When fine-tuning large pre-trained models such as CLIP, ALIGN, and a ViT-G pre-trained on JFT, our soup recipe provides significant improvements over the best model in a hyperparameter sweep on ImageNet. The resulting ViT-G model, which attains 90.94% top-1 accuracy on ImageNet, achieved a new state of the art. Furthermore, we show that the model soup approach extends to multiple image classification and natural language processing tasks, improves out-of-distribution performance, and improves zero-shot performance on new downstream tasks. Finally, we analytically relate the performance similarity of weight-averaging and logit-ensembling to flatness of the loss and confidence of the predictions, and validate this relation empirically. Code is available at https://github.com/mlfoundations/model-soups."
                },
                {
                    "title": "Penalizing Gradient Norm for Efficiently Improving Generalization in Deep Learning",
                    "abstract": "How to train deep neural networks (DNNs) to generalize well is a central concern in deep learning, especially for severely overparameterized networks nowadays. In this paper, we propose an effective method to improve the model generalization by additionally penalizing the gradient norm of loss function during optimization. We demonstrate that confining the gradient norm of loss function could help lead the optimizers towards finding flat minima. We leverage the first-order approximation to efficiently implement the corresponding gradient to fit well in the gradient descent framework. In our experiments, we confirm that when using our methods, generalization performance of various models could be improved on different datasets. Also, we show that the recent sharpness-aware minimization method (Foret et al., 2021) is a special, but not the best, case of our method, where the best case of our method could give new state-of-art performance on these tasks. Code is available at {https://github.com/zhaoyang-0204/gnp}."
                },
                {
                    "title": "When Do Flat Minima Optimizers Work?",
                    "abstract": "Recently, flat-minima optimizers, which seek to find parameters in low-loss neighborhoods, have been shown to improve a neural network's generalization performance over stochastic and adaptive gradient-based optimizers. Two methods have received significant attention due to their scalability: 1. Stochastic Weight Averaging (SWA), and 2. Sharpness-Aware Minimization (SAM). However, there has been limited investigation into their properties and no systematic benchmarking of them across different domains. We fill this gap here by comparing the loss surfaces of the models trained with each method and through broad benchmarking across computer vision, natural language processing, and graph representation learning tasks. We discover several surprising findings from these results, which we hope will help researchers further improve deep learning optimizers, and practitioners identify the right optimizer for their problem."
                },
                {
                    "title": "Federated Learning Based on Dynamic Regularization",
                    "abstract": "We propose a novel federated learning method for distributively training neural network models, where the server orchestrates cooperation between a subset of randomly chosen devices in each round. We view Federated Learning problem primarily from a communication perspective and allow more device level computations to save transmission costs. We point out a fundamental dilemma, in that the minima of the local-device level empirical loss are inconsistent with those of the global empirical loss. Different from recent prior works, that either attempt inexact minimization or utilize devices for parallelizing gradient computation, we propose a dynamic regularizer for each device at each round, so that in the limit the global and device solutions are aligned. We demonstrate both through empirical results on real and synthetic data as well as analytical results that our scheme leads to efficient training, in both convex and non-convex settings, while being fully agnostic to device heterogeneity and robust to large number of devices, partial participation and unbalanced data."
                },
                {
                    "title": "The Role of Permutation Invariance in Linear Mode Connectivity of Neural Networks",
                    "abstract": "In this paper, we conjecture that if the permutation invariance of neural networks is taken into account, SGD solutions will likely have no barrier in the linear interpolation between them. Although it is a bold conjecture, we show how extensive empirical attempts fall short of refuting it. We further provide a preliminary theoretical result to support our conjecture. Our conjecture has implications for lottery ticket hypothesis, distributed training, and ensemble methods."
                },
                {
                    "title": "Fed2: Feature-Aligned Federated Learning",
                    "abstract": "Federated learning learns from scattered data by fusing collaborative models from local nodes. However, conventional coordinate-based model averaging by FedAvg ignored the random information encoded per parameter and may suffer from structural feature misalignment. In this work, we propose Fed2, a feature-aligned federated learning framework to resolve this issue by establishing a firm structure-feature alignment across the collaborative models. Fed2 is composed of two major designs: First, we design a feature-oriented model structure adaptation method to ensure explicit feature allocation in different neural network structures. Applying the structure adaptation to collaborative models, matchable structures with similar feature information can be initialized at the very early training stage. During the federated learning process, we then propose a feature paired averaging scheme to guarantee aligned feature distribution and maintain no feature fusion conflicts under either IID or non-IID scenarios. Eventually, Fed2 could effectively enhance the federated learning convergence performance under extensive homo- and heterogeneous settings, providing excellent convergence speed, accuracy, and computation/communication efficiency."
                },
                {
                    "title": "What can linear interpolation of neural network loss landscapes tell us?",
                    "abstract": "Studying neural network loss landscapes provides insights into the nature of the underlying optimization problems. Unfortunately, loss landscapes are notoriously difficult to visualize in a human-comprehensible fashion. One common way to address this problem is to plot linear slices of the landscape, for example from the initial state of the network to the final state after optimization. On the basis of this analysis, prior work has drawn broader conclusions about the difficulty of the optimization problem. In this paper, we put inferences of this kind to the test, systematically evaluating how linear interpolation and final performance vary when altering the data, choice of initialization, and other optimizer and architecture design choices. Further, we use linear interpolation to study the role played by individual layers and substructures of the network. We find that certain layers are more sensitive to the choice of initialization, but that the shape of the linear path is not indicative of the changes in test accuracy of the model. Our results cast doubt on the broader intuition that the presence or absence of barriers when interpolating necessarily relates to the success of optimization."
                },
                {
                    "title": "Model-Contrastive Federated Learning",
                    "abstract": "Federated learning enables multiple parties to collaboratively train a machine learning model without communicating their local data. A key challenge in federated learning is to handle the heterogeneity of local data distribution across parties. Although many studies have been proposed to address this challenge, we find that they fail to achieve high performance in image datasets with deep learning models. In this paper, we propose MOON: model-contrastive federated learning. MOON is a simple and effective federated learning framework. The key idea of MOON is to utilize the similarity between model representations to correct the local training of individual parties, i.e., conducting contrastive learning in model-level. Our extensive experiments show that MOON significantly outperforms the other state-of-the-art federated learning algorithms on various image classification tasks."
                },
                {
                    "title": "Gradient Projection Memory for Continual Learning",
                    "abstract": "The ability to learn continually without forgetting the past tasks is a desired attribute for artificial learning systems. Existing approaches to enable such learning in artificial neural networks usually rely on network growth, importance based weight update or replay of old data from the memory. In contrast, we propose a novel approach where a neural network learns new tasks by taking gradient steps in the orthogonal direction to the gradient subspaces deemed important for the past tasks. We find the bases of these subspaces by analyzing network representations (activations) after learning each task with Singular Value Decomposition (SVD) in a single shot manner and store them in the memory as Gradient Projection Memory (GPM). With qualitative and quantitative analyses, we show that such orthogonal gradient descent induces minimum to no interference with the past tasks, thereby mitigates forgetting. We evaluate our algorithm on diverse image classification datasets with short and long sequences of tasks and report better or on-par performance compared to the state-of-the-art approaches."
                },
                {
                    "title": "ASAM: Adaptive Sharpness-Aware Minimization for Scale-Invariant Learning of Deep Neural Networks",
                    "abstract": "Recently, learning algorithms motivated from sharpness of loss surface as an effective measure of generalization gap have shown state-of-the-art performances. Nevertheless, sharpness defined in a rigid region with a fixed radius, has a drawback in sensitivity to parameter re-scaling which leaves the loss unaffected, leading to weakening of the connection between sharpness and generalization gap. In this paper, we introduce the concept of adaptive sharpness which is scale-invariant and propose the corresponding generalization bound. We suggest a novel learning method, adaptive sharpness-aware minimization (ASAM), utilizing the proposed generalization bound. Experimental results in various benchmark datasets show that ASAM contributes to significant improvement of model generalization performance."
                },
                {
                    "title": "Revisiting \"Qualitatively Characterizing Neural Network Optimization Problems\"",
                    "abstract": "We revisit and extend the experiments of Goodfellow et al. (2014), who showed that - for then state-of-the-art networks - \"the objective function has a simple, approximately convex shape\" along the linear path between initialization and the trained weights. We do not find this to be the case for modern networks on CIFAR-10 and ImageNet. Instead, although loss is roughly monotonically non-increasing along this path, it remains high until close to the optimum. In addition, training quickly becomes linearly separated from the optimum by loss barriers. We conclude that, although Goodfellow et al.'s findings describe the \"relatively easy to optimize\" MNIST setting, behavior is qualitatively different in modern settings."
                },
                {
                    "title": "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale",
                    "abstract": "While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train."
                },
                {
                    "title": "Sharpness-Aware Minimization for Efficiently Improving Generalization",
                    "abstract": "In today's heavily overparameterized models, the value of the training loss provides few guarantees on model generalization ability. Indeed, optimizing only the training loss value, as is commonly done, can easily lead to suboptimal model quality. Motivated by the connection between geometry of the loss landscape and generalization---including a generalization bound that we prove here---we introduce a novel, effective procedure for instead simultaneously minimizing loss value and loss sharpness. In particular, our procedure, Sharpness-Aware Minimization (SAM), seeks parameters that lie in neighborhoods having uniformly low loss; this formulation results in a min-max optimization problem on which gradient descent can be performed efficiently. We present empirical results showing that SAM improves model generalization across a variety of benchmark datasets (e.g., CIFAR-{10, 100}, ImageNet, finetuning tasks) and models, yielding novel state-of-the-art performance for several. Additionally, we find that SAM natively provides robustness to label noise on par with that provided by state-of-the-art procedures that specifically target learning with noisy labels."
                },
                {
                    "title": "What is being transferred in transfer learning?",
                    "abstract": "One desired capability for machines is the ability to transfer their knowledge of one domain to another where data is (usually) scarce. Despite ample adaptation of transfer learning in various deep learning applications, we yet do not understand what enables a successful transfer and which part of the network is responsible for that. In this paper, we provide new tools and analyses to address these fundamental questions. Through a series of analyses on transferring to block-shuffled images, we separate the effect of feature reuse from learning low-level statistics of data and show that some benefit of transfer learning comes from the latter. We present that when training from pre-trained weights, the model stays in the same basin in the loss landscape and different instances of such model are similar in feature space and close in parameter space."
                },
                {
                    "title": "Ensemble Distillation for Robust Model Fusion in Federated Learning",
                    "abstract": "Federated Learning (FL) is a machine learning setting where many devices collaboratively train a machine learning model while keeping the training data decentralized. In most of the current training schemes the central model is refined by averaging the parameters of the server model and the updated parameters from the client side. However, directly averaging model parameters is only possible if all models have the same structure and size, which could be a restrictive constraint in many scenarios. \nIn this work we investigate more powerful and more flexible aggregation schemes for FL. Specifically, we propose ensemble distillation for model fusion, i.e. training the central classifier through unlabeled data on the outputs of the models from the clients. This knowledge distillation technique mitigates privacy risk and cost to the same extent as the baseline FL algorithms, but allows flexible aggregation over heterogeneous client models that can differ e.g. in size, numerical precision or structure. We show in extensive empirical experiments on various CV/NLP datasets (CIFAR-10/100, ImageNet, AG News, SST2) and settings (heterogeneous models/data) that the server model can be trained much faster, requiring fewer communication rounds than any existing FL technique so far."
                },
                {
                    "title": "The Break-Even Point on Optimization Trajectories of Deep Neural Networks",
                    "abstract": "The early phase of training of deep neural networks is critical for their final performance. In this work, we study how the hyperparameters of stochastic gradient descent (SGD) used in the early phase of training affect the rest of the optimization trajectory. We argue for the existence of the \"break-even\" point on this trajectory, beyond which the curvature of the loss surface and noise in the gradient are implicitly regularized by SGD. In particular, we demonstrate on multiple classification tasks that using a large learning rate in the initial phase of training reduces the variance of the gradient, and improves the conditioning of the covariance of gradients. These effects are beneficial from the optimization perspective and become visible after the break-even point. Complementing prior work, we also show that using a low learning rate results in bad conditioning of the loss surface even for a neural network with batch normalization layers. In short, our work shows that key properties of the loss surface are strongly influenced by SGD in the early phase of training. We argue that studying the impact of the identified effects on generalization is a promising future direction."
                },
                {
                    "title": "Federated Learning with Matched Averaging",
                    "abstract": "Federated learning allows edge devices to collaboratively learn a shared model while keeping the training data on device, decoupling the ability to do model training from the need to store the data in the cloud. We propose Federated matched averaging (FedMA) algorithm designed for federated learning of modern neural network architectures e.g. convolutional neural networks (CNNs) and LSTMs. FedMA constructs the shared global model in a layer-wise manner by matching and averaging hidden elements (i.e. channels for convolution layers; hidden states for LSTM; neurons for fully connected layers) with similar feature extraction signatures. Our experiments indicate that FedMA outperforms popular state-of-the-art federated learning algorithms on deep CNN and LSTM architectures trained on real world datasets, while improving the communication efficiency."
                },
                {
                    "title": "Linear Mode Connectivity and the Lottery Ticket Hypothesis",
                    "abstract": "We study whether a neural network optimizes to the same, linearly connected minimum under different samples of SGD noise (e.g., random data order and augmentation). We find that standard vision models become stable to SGD noise in this way early in training. From then on, the outcome of optimization is determined to a linearly connected region. We use this technique to study iterative magnitude pruning (IMP), the procedure used by work on the lottery ticket hypothesis to identify subnetworks that could have trained in isolation to full accuracy. We find that these subnetworks only reach full accuracy when they are stable to SGD noise, which either occurs at initialization for small-scale settings (MNIST) or early in training for large-scale settings (ResNet-50 and Inception-v3 on ImageNet)."
                },
                {
                    "title": "Fantastic Generalization Measures and Where to Find Them",
                    "abstract": "Generalization of deep networks has been of great interest in recent years, resulting in a number of theoretically and empirically motivated complexity measures. However, most papers proposing such measures study only a small set of models, leaving open the question of whether the conclusion drawn from those experiments would remain valid in other settings. We present the first large scale study of generalization in deep networks. We investigate more then 40 complexity measures taken from both theoretical bounds and empirical studies. We train over 10,000 convolutional networks by systematically varying commonly used hyperparameters. Hoping to uncover potentially causal relationships between each measure and generalization, we analyze carefully controlled experiments and show surprising failures of some measures as well as promising measures for further research."
                },
                {
                    "title": "SCAFFOLD: Stochastic Controlled Averaging for Federated Learning",
                    "abstract": "Federated Averaging (FedAvg) has emerged as the algorithm of choice for federated learning due to its simplicity and low communication cost. However, in spite of recent research efforts, its performance is not fully understood. We obtain tight convergence rates for FedAvg and prove that it suffers from `client-drift' when the data is heterogeneous (non-iid), resulting in unstable and slow convergence. \nAs a solution, we propose a new algorithm (SCAFFOLD) which uses control variates (variance reduction) to correct for the `client-drift' in its local updates. We prove that SCAFFOLD requires significantly fewer communication rounds and is not affected by data heterogeneity or client sampling. Further, we show that (for quadratics) SCAFFOLD can take advantage of similarity in the client's data yielding even faster convergence. The latter is the first result to quantify the usefulness of local-steps in distributed optimization."
                },
                {
                    "title": "Model Fusion via Optimal Transport",
                    "abstract": "Combining different models is a widely used paradigm in machine learning applications. While the most common approach is to form an ensemble of models and average their individual predictions, this approach is often rendered infeasible by given resource constraints in terms of memory and computation, which grow linearly with the number of models. We present a layer-wise model fusion algorithm for neural networks that utilizes optimal transport to (soft-) align neurons across the models before averaging their associated parameters. We show that this can successfully yield \"one-shot\" knowledge transfer (i.e, without requiring any retraining) between neural networks trained on heterogeneous non-i.i.d. data. In both i.i.d. and non-i.i.d. settings, we illustrate that our approach significantly outperforms vanilla averaging, as well as how it can serve as an efficient replacement for the ensemble with moderate fine-tuning, for standard convolutional networks (like VGG11), residual networks (like ResNet18), and multi-layer perceptrons on CIFAR10 and MNIST. Finally, our approach also provides a principled way to combine the parameters of neural networks with different widths, and we explore its application for model compression. Code is available under the following link https://github.com/modelfusion."
                },
                {
                    "title": "The Non-IID Data Quagmire of Decentralized Machine Learning",
                    "abstract": "Many large-scale machine learning (ML) applications need to perform decentralized learning over datasets generated at different devices and locations. Such datasets pose a significant challenge to decentralized learning because their different contexts result in significant data distribution skew across devices/locations. In this paper, we take a step toward better understanding this challenge by presenting a detailed experimental study of decentralized DNN training on a common type of data skew: skewed distribution of data labels across devices/locations. Our study shows that: (i) skewed data labels are a fundamental and pervasive problem for decentralized learning, causing significant accuracy loss across many ML applications, DNN models, training datasets, and decentralized learning algorithms; (ii) the problem is particularly challenging for DNN models with batch normalization; and (iii) the degree of data skew is a key determinant of the difficulty of the problem. Based on these findings, we present SkewScout, a system-level approach that adapts the communication frequency of decentralized learning algorithms to the (skew-induced) accuracy loss between data partitions. We also show that group normalization can recover much of the accuracy loss of batch normalization."
                },
                {
                    "title": "Deep Ensembles: A Loss Landscape Perspective",
                    "abstract": "Deep ensembles have been empirically shown to be a promising approach for improving accuracy, uncertainty and out-of-distribution robustness of deep learning models. While deep ensembles were theoretically motivated by the bootstrap, non-bootstrap ensembles trained with just random initialization also perform well in practice, which suggests that there could be other explanations for why deep ensembles work well. Bayesian neural networks, which learn distributions over the parameters of the network, are theoretically well-motivated by Bayesian principles, but do not perform as well as deep ensembles in practice, particularly under dataset shift. One possible explanation for this gap between theory and practice is that popular scalable variational Bayesian methods tend to focus on a single mode, whereas deep ensembles tend to explore diverse modes in function space. We investigate this hypothesis by building on recent work on understanding the loss landscape of neural networks and adding our own exploration to measure the similarity of functions in the space of predictions. Our results show that random initializations explore entirely different modes, while functions along an optimization trajectory or sampled from the subspace thereof cluster within a single mode predictions-wise, while often deviating significantly in the weight space. Developing the concept of the diversity--accuracy plane, we show that the decorrelation power of random initializations is unmatched by popular subspace sampling methods. Finally, we evaluate the relative effects of ensembling, subspace based methods and ensembles of subspace based methods, and the experimental results validate our hypothesis."
                },
                {
                    "title": "Visualizing and Understanding the Effectiveness of BERT",
                    "abstract": "Language model pre-training, such as BERT, has achieved remarkable results in many NLP tasks. However, it is unclear why the pre-training-then-fine-tuning paradigm can improve performance and generalization capability across different tasks. In this paper, we propose to visualize loss landscapes and optimization trajectories of fine-tuning BERT on specific datasets. First, we find that pre-training reaches a good initial point across downstream tasks, which leads to wider optima and easier optimization compared with training from scratch. We also demonstrate that the fine-tuning procedure is robust to overfitting, even though BERT is highly over-parameterized for downstream tasks. Second, the visualization results indicate that fine-tuning BERT tends to generalize better because of the flat and wide optima, and the consistency between the training loss surface and the generalization error surface. Third, the lower layers of BERT are more invariant during fine-tuning, which suggests that the layers that are close to input learn more transferable representations of language."
                },
                {
                    "title": "Large Scale Structure of Neural Network Loss Landscapes",
                    "abstract": "There are many surprising and perhaps counter-intuitive properties of optimization of deep neural networks. We propose and experimentally verify a unified phenomenological model of the loss landscape that incorporates many of them. High dimensionality plays a key role in our model. Our core idea is to model the loss landscape as a set of high dimensional \\emph{wedges} that together form a large-scale, inter-connected structure and towards which optimization is drawn. We first show that hyperparameter choices such as learning rate, network width and $L_2$ regularization, affect the path optimizer takes through the landscape in a similar ways, influencing the large scale curvature of the regions the optimizer explores. Finally, we predict and demonstrate new counter-intuitive properties of the loss-landscape. We show an existence of low loss subspaces connecting a set (not only a pair) of solutions, and verify it experimentally. Finally, we analyze recently popular ensembling techniques for deep networks in the light of our model."
                },
                {
                    "title": "Wide neural networks of any depth evolve as linear models under gradient descent",
                    "abstract": "A longstanding goal in deep learning research has been to precisely characterize training and generalization. However, the often complex loss landscapes of neural networks (NNs) have made a theory of learning dynamics elusive. In this work, we show that for wide NNs the learning dynamics simplify considerably and that, in the infinite width limit, they are governed by a linear model obtained from the first-order Taylor expansion of the network around its initial parameters. Furthermore, mirroring the correspondence between wide Bayesian NNs and Gaussian processes (GPs), gradient-based training of wide NNs with a squared loss produces test set predictions drawn from a GP with a particular compositional kernel. While these theoretical results are only exact in the infinite width limit, we nevertheless find excellent empirical agreement between the predictions of the original network and those of the linearized version even for finite practically-sized networks. This agreement is robust across different architectures, optimization methods, and loss functions."
                },
                {
                    "title": "Asymmetric Valleys: Beyond Sharp and Flat Local Minima",
                    "abstract": "Despite the non-convex nature of their loss functions, deep neural networks are known to generalize well when optimized with stochastic gradient descent (SGD). Recent work conjectures that SGD with proper configuration is able to find wide and flat local minima, which have been proposed to be associated with good generalization performance. In this paper, we observe that local minima of modern deep networks are more than being flat or sharp. Specifically, at a local minimum there exist many asymmetric directions such that the loss increases abruptly along one side, and slowly along the opposite side--we formally define such minima as asymmetric valleys. Under mild assumptions, we prove that for asymmetric valleys, a solution biased towards the flat side generalizes better than the exact minimizer. Further, we show that simply averaging the weights along the SGD trajectory gives rise to such biased solutions implicitly. This provides a theoretical explanation for the intriguing phenomenon observed by Izmailov et al. (2018). In addition, we empirically find that batch normalization (BN) appears to be a major cause for asymmetric valleys."
                },
                {
                    "title": "Sensitivity Analysis of Deep Neural Networks",
                    "abstract": "Deep neural networks (DNNs) have achieved superior performance in various prediction tasks, but can be very vulnerable to adversarial examples or perturbations. Therefore, it is crucial to measure the sensitivity of DNNs to various forms of perturbations in real applications. We introduce a novel perturbation manifold and its associated influence measure to quantify the effects of various perturbations on DNN classifiers. Such perturbations include various external and internal perturbations to input samples and network parameters. The proposed measure is motivated by information geometry and provides desirable invariance properties. We demonstrate that our influence measure is useful for four model building tasks: detecting potential \u2018outliers\u2019, analyzing the sensitivity of model architectures, comparing network sensitivity between training and test sets, and locating vulnerable areas. Experiments show reasonably good performance of the proposed measure for the popular DNN models ResNet50 and DenseNet121 on CIFAR10 and MNIST datasets."
                },
                {
                    "title": "Federated Optimization in Heterogeneous Networks",
                    "abstract": "Federated Learning is a distributed learning paradigm with two key challenges that differentiate it from traditional distributed optimization: (1) significant variability in terms of the systems characteristics on each device in the network (systems heterogeneity), and (2) non-identically distributed data across the network (statistical heterogeneity). In this work, we introduce a framework, FedProx, to tackle heterogeneity in federated networks. FedProx can be viewed as a generalization and re-parametrization of FedAvg, the current state-of-the-art method for federated learning. While this re-parameterization makes only minor modifications to the method itself, these modifications have important ramifications both in theory and in practice. Theoretically, we provide convergence guarantees for our framework when learning over data from non-identical distributions (statistical heterogeneity), and while adhering to device-level systems constraints by allowing each participating device to perform a variable amount of work (systems heterogeneity). Practically, we demonstrate that FedProx allows for more robust convergence than FedAvg across a suite of realistic federated datasets. In particular, in highly heterogeneous settings, FedProx demonstrates significantly more stable and accurate convergence behavior relative to FedAvg---improving absolute test accuracy by 22% on average."
                },
                {
                    "title": "Neural Tangent Kernel: Convergence and Generalization in Neural Networks",
                    "abstract": "At initialization, artificial neural networks (ANNs) are equivalent to Gaussian processes in the infinite-width limit, thus connecting them to kernel methods. We prove that the evolution of an ANN during training can also be described by a kernel: during gradient descent on the parameters of an ANN, the network function (which maps input vectors to output vectors) follows the so-called kernel gradient associated with a new object, which we call the Neural Tangent Kernel (NTK). This kernel is central to describe the generalization features of ANNs. While the NTK is random at initialization and varies during training, in the infinite-width limit it converges to an explicit limiting kernel and stays constant during training. This makes it possible to study the training of ANNs in function space instead of parameter space. Convergence of the training can then be related to the positive-definiteness of the limiting NTK. We then focus on the setting of least-squares regression and show that in the infinite-width limit, the network function follows a linear differential equation during training. The convergence is fastest along the largest kernel principal components of the input data with respect to the NTK, hence suggesting a theoretical motivation for early stopping. Finally we study the NTK numerically, observe its behavior for wide networks, and compare it to the infinite-width limit."
                },
                {
                    "title": "Federated Learning with Non-IID Data",
                    "abstract": "Federated learning enables resource-constrained edge compute devices, such as mobile phones and IoT devices, to learn a shared model for prediction, while keeping the training data local. This decentralized approach to train models provides privacy, security, regulatory and economic benefits. In this work, we focus on the statistical challenge of federated learning when local data is non-IID. We first show that the accuracy of federated learning reduces significantly, by up to 55% for neural networks trained for highly skewed non-IID data, where each client device trains only on a single class of data. We further show that this accuracy reduction can be explained by the weight divergence, which can be quantified by the earth mover's distance (EMD) between the distribution over classes on each device and the population distribution. As a solution, we propose a strategy to improve training on non-IID data by creating a small subset of data which is globally shared between all the edge devices. Experiments show that accuracy can be increased by 30% for the CIFAR-10 dataset with only 5% globally shared data."
                },
                {
                    "title": "How Does Batch Normalization Help Optimization? (No, It Is Not About Internal Covariate Shift)",
                    "abstract": "Batch Normalization (BatchNorm) is a widely adopted technique that enables faster and more stable training of deep neural networks (DNNs). Despite its pervasiveness, the exact reasons for BatchNorm's effectiveness are still poorly understood. The popular belief is that this effectiveness stems from controlling the change of the layers' input distributions during training to reduce the so-called \"internal covariate shift\". In this work, we demonstrate that such distributional stability of layer inputs has little to do with the success of BatchNorm. Instead, we uncover a more fundamental impact of BatchNorm on the training process: it makes the optimization landscape significantly smoother. This smoothness induces a more predictive and stable behavior of the gradients, allowing for faster training. These findings bring us closer to a true understanding of our DNN training toolkit."
                },
                {
                    "title": "Averaging Weights Leads to Wider Optima and Better Generalization",
                    "abstract": "Deep neural networks are typically trained by optimizing a loss function with an SGD variant, in conjunction with a decaying learning rate, until convergence. We show that simple averaging of multiple points along the trajectory of SGD, with a cyclical or constant learning rate, leads to better generalization than conventional training. We also show that this Stochastic Weight Averaging (SWA) procedure finds much flatter solutions than SGD, and approximates the recent Fast Geometric Ensembling (FGE) approach with a single model. Using SWA we achieve notable improvement in test accuracy over conventional SGD training on a range of state-of-the-art residual networks, PyramidNets, DenseNets, and Shake-Shake networks on CIFAR-10, CIFAR-100, and ImageNet. In short, SWA is extremely easy to implement, improves generalization, and has almost no computational overhead."
                },
                {
                    "title": "Essentially No Barriers in Neural Network Energy Landscape",
                    "abstract": "Training neural networks involves finding minima of a high-dimensional non-convex loss function. Knowledge of the structure of this energy landscape is sparse. Relaxing from linear interpolations, we construct continuous paths between minima of recent neural network architectures on CIFAR10 and CIFAR100. Surprisingly, the paths are essentially flat in both the training and test landscapes. This implies that neural networks have enough capacity for structural changes, or that these changes are small between minima. Also, each minimum has at least one vanishing Hessian eigenvalue in addition to those resulting from trivial invariance."
                },
                {
                    "title": "Loss Surfaces, Mode Connectivity, and Fast Ensembling of DNNs",
                    "abstract": "The loss functions of deep neural networks are complex and their geometric properties are not well understood. We show that the optima of these complex loss functions are in fact connected by simple curves over which training and test accuracy are nearly constant. We introduce a training procedure to discover these high-accuracy pathways between modes. Inspired by this new geometric insight, we also propose a new ensembling method entitled Fast Geometric Ensembling (FGE). Using FGE we can train high-performing ensembles in the time required to train a single model. We achieve improved performance compared to the recent state-of-the-art Snapshot Ensembles, on CIFAR-10, CIFAR-100, and ImageNet."
                },
                {
                    "title": "Visualizing the Loss Landscape of Neural Nets",
                    "abstract": "Neural network training relies on our ability to find \"good\" minimizers of highly non-convex loss functions. It is well-known that certain network architecture designs (e.g., skip connections) produce loss functions that train easier, and well-chosen training parameters (batch size, learning rate, optimizer) produce minimizers that generalize better. However, the reasons for these differences, and their effects on the underlying loss landscape, are not well understood. In this paper, we explore the structure of neural loss functions, and the effect of loss landscapes on generalization, using a range of visualization methods. First, we introduce a simple \"filter normalization\" method that helps us visualize loss function curvature and make meaningful side-by-side comparisons between loss functions. Then, using a variety of visualizations, we explore how network architecture affects the loss landscape, and how training parameters affect the shape of minimizers."
                },
                {
                    "title": "Exploring Generalization in Deep Learning",
                    "abstract": "With a goal of understanding what drives generalization in deep networks, we consider several recently suggested explanations, including norm-based control, sharpness and robustness. We study how these measures can ensure generalization, highlighting the importance of scale normalization, and making a connection between sharpness and PAC-Bayes theory. We then investigate how well the measures explain different observed phenomena."
                },
                {
                    "title": "Snapshot Ensembles: Train 1, get M for free",
                    "abstract": "Ensembles of neural networks are known to be much more robust and accurate than individual networks. However, training multiple deep networks for model averaging is computationally expensive. In this paper, we propose a method to obtain the seemingly contradictory goal of ensembling multiple neural networks at no additional training cost. We achieve this goal by training a single neural network, converging to several local minima along its optimization path and saving the model parameters. To obtain repeated rapid convergence, we leverage recent work on cyclic learning rate schedules. The resulting technique, which we refer to as Snapshot Ensembling, is simple, yet surprisingly effective. We show in a series of experiments that our approach is compatible with diverse network architectures and learning tasks. It consistently yields lower error rates than state-of-the-art single models at no additional training cost, and compares favorably with traditional network ensembles. On CIFAR-10 and CIFAR-100 our DenseNet Snapshot Ensembles obtain error rates of 3.4% and 17.4% respectively."
                },
                {
                    "title": "On the Robustness of Convolutional Neural Networks to Internal Architecture and Weight Perturbations",
                    "abstract": "Deep convolutional neural networks are generally regarded as robust function approximators. So far, this intuition is based on perturbations to external stimuli such as the images to be classified. Here we explore the robustness of convolutional neural networks to perturbations to the internal weights and architecture of the network itself. We show that convolutional networks are surprisingly robust to a number of internal perturbations in the higher convolutional layers but the bottom convolutional layers are much more fragile. For instance, Alexnet shows less than a 30% decrease in classification performance when randomly removing over 70% of weight connections in the top convolutional or dense layers but performance is almost at chance with the same perturbation in the first convolutional layer. Finally, we suggest further investigations which could continue to inform the robustness of convolutional networks to internal perturbations."
                },
                {
                    "title": "Sharp Minima Can Generalize For Deep Nets",
                    "abstract": "Despite their overwhelming capacity to overfit, deep learning architectures tend to generalize relatively well to unseen data, allowing them to be deployed in practice. However, explaining why this is the case is still an open area of research. One standing hypothesis that is gaining popularity, e.g. Hochreiter & Schmidhuber (1997); Keskar et al. (2017), is that the flatness of minima of the loss function found by stochastic gradient based methods results in good generalization. This paper argues that most notions of flatness are problematic for deep models and can not be directly applied to explain generalization. Specifically, when focusing on deep networks with rectifier units, we can exploit the particular geometry of parameter space induced by the inherent symmetries that these architectures exhibit to build equivalent models corresponding to arbitrarily sharper minima. Furthermore, if we allow to reparametrize a function, the geometry of its parameters can change drastically without affecting its generalization properties."
                },
                {
                    "title": "An empirical analysis of the optimization of deep network loss surfaces",
                    "abstract": "The success of deep neural networks hinges on our ability to accurately and efficiently optimize high-dimensional, non-convex functions. In this paper, we empirically investigate the loss functions of state-of-the-art networks, and how commonly-used stochastic gradient descent variants optimize these loss functions. To do this, we visualize the loss function by projecting them down to low-dimensional spaces chosen based on the convergence points of different optimization algorithms. Our observations suggest that optimization algorithms encounter and choose different descent directions at many saddle points to find different final weights. Based on consistency we observe across re-runs of the same stochastic optimization algorithm, we hypothesize that each optimization algorithm makes characteristic choices at these saddle points."
                },
                {
                    "title": "Aggregated Residual Transformations for Deep Neural Networks",
                    "abstract": "We present a simple, highly modularized network architecture for image classification. Our network is constructed by repeating a building block that aggregates a set of transformations with the same topology. Our simple design results in a homogeneous, multi-branch architecture that has only a few hyper-parameters to set. This strategy exposes a new dimension, which we call cardinality (the size of the set of transformations), as an essential factor in addition to the dimensions of depth and width. On the ImageNet-1K dataset, we empirically show that even under the restricted condition of maintaining complexity, increasing cardinality is able to improve classification accuracy. Moreover, increasing cardinality is more effective than going deeper or wider when we increase the capacity. Our models, named ResNeXt, are the foundations of our entry to the ILSVRC 2016 classification task in which we secured 2nd place. We further investigate ResNeXt on an ImageNet-5K set and the COCO detection set, also showing better results than its ResNet counterpart. The code and models are publicly available online."
                },
                {
                    "title": "Entropy-SGD: biasing gradient descent into wide valleys",
                    "abstract": "This paper proposes a new optimization algorithm called Entropy-SGD for training deep neural networks that is motivated by the local geometry of the energy landscape. Local extrema with low generalization error have a large proportion of almost-zero eigenvalues in the Hessian with very few positive or negative eigenvalues. We leverage upon this observation to construct a local-entropy-based objective function that favors well-generalizable solutions lying in large flat regions of the energy landscape, while avoiding poorly-generalizable solutions located in the sharp valleys. Conceptually, our algorithm resembles two nested loops of SGD where we use Langevin dynamics in the inner loop to compute the gradient of the local entropy before each update of the weights. We show that the new objective has a smoother energy landscape and show improved generalization over SGD using uniform stability, under certain assumptions. Our experiments on convolutional and recurrent networks demonstrate that Entropy-SGD compares favorably to state-of-the-art techniques in terms of generalization error and training time."
                },
                {
                    "title": "On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima",
                    "abstract": "The stochastic gradient descent (SGD) method and its variants are algorithms of choice for many Deep Learning tasks. These methods operate in a small-batch regime wherein a fraction of the training data, say $32$-$512$ data points, is sampled to compute an approximation to the gradient. It has been observed in practice that when using a larger batch there is a degradation in the quality of the model, as measured by its ability to generalize. We investigate the cause for this generalization drop in the large-batch regime and present numerical evidence that supports the view that large-batch methods tend to converge to sharp minimizers of the training and testing functions - and as is well known, sharp minima lead to poorer generalization. In contrast, small-batch methods consistently converge to flat minimizers, and our experiments support a commonly held view that this is due to the inherent noise in the gradient estimation. We discuss several strategies to attempt to help large-batch methods eliminate this generalization gap."
                },
                {
                    "title": "Communication-Efficient Learning of Deep Networks from Decentralized Data",
                    "abstract": "Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data center and training there using conventional approaches. We advocate an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. \nWe present a practical method for the federated learning of deep networks based on iterative model averaging, and conduct an extensive empirical evaluation, considering five different model architectures and four datasets. These experiments demonstrate the approach is robust to the unbalanced and non-IID data distributions that are a defining characteristic of this setting. Communication costs are the principal constraint, and we show a reduction in required communication rounds by 10-100x as compared to synchronized stochastic gradient descent."
                },
                {
                    "title": "Deep Residual Learning for Image Recognition",
                    "abstract": "Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8\u00d7 deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation."
                },
                {
                    "title": "A method for finding similarity between multi-layer perceptrons by Forward Bipartite Alignment",
                    "abstract": "We present Forward Bipartite Alignment (FBA), a method that aligns the topological structures of two neural networks. Neural networks are considered to be a black box, because neural networks have a complex model surface determined by their weights that combine attributes non-linearly. Two networks that make similar predictions on training data may still generalize differently. FBA enables a diversity of applications, including visualization and canonicalization of neural networks, ensembles, and cross-over between unrelated neural networks in evolutionary optimization. We describe the FBA algorithm, and describe implementations for three applications: genetic algorithms, visualization, and ensembles. We demonstrate FBA's usefulness by comparing a bag of neural networks to a bag of FBA-aligned neural networks. We also show that aligning, and then combining two neural networks has no appreciable loss in accuracy which means that Forward Bipartite Alignment aligns neural networks in a meaningful way."
                },
                {
                    "title": "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift",
                    "abstract": "Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.82% top-5 test error, exceeding the accuracy of human raters."
                },
                {
                    "title": "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification",
                    "abstract": "Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on the learnable activation and advanced initialization, we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66% [33]). To our knowledge, our result is the first to surpass the reported human-level performance (5.1%, [26]) on this dataset."
                },
                {
                    "title": "Qualitatively characterizing neural network optimization problems",
                    "abstract": "Training neural networks involves solving large-scale non-convex optimization problems. This task has long been believed to be extremely difficult, with fear of local minima and other obstacles motivating a variety of schemes to improve optimization, such as unsupervised pretraining. However, modern neural networks are able to achieve negligible training error on complex tasks, using only direct training with stochastic gradient descent. We introduce a simple analysis technique to look for evidence that such networks are overcoming local optima. We find that, in fact, on a straight path from initialization to solution, a variety of state of the art neural networks never encounter any significant obstacles."
                },
                {
                    "title": "Very Deep Convolutional Networks for Large-Scale Image Recognition",
                    "abstract": "In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision."
                },
                {
                    "title": "On the Number of Linear Regions of Deep Neural Networks",
                    "abstract": "We study the complexity of functions computable by deep feedforward neural networks with piecewise linear activations in terms of the symmetries and the number of linear regions that they have. Deep networks are able to sequentially map portions of each layer's input-space to the same output. In this way, deep models compute functions that react equally to complicated patterns of different inputs. The compositional structure of these functions enables them to re-use pieces of computation exponentially often in terms of the network's depth. This paper investigates the complexity of such compositional maps and contributes new theoretical results regarding the advantage of depth for neural networks with piecewise linear activation functions. In particular, our analysis is not specific to a single family of models, and as an example, we employ it for rectifier and maxout networks. We improve complexity bounds from pre-existing work and investigate the behavior of units in higher layers."
                },
                {
                    "title": "ImageNet classification with deep convolutional neural networks",
                    "abstract": "We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry."
                },
                {
                    "title": "ImageNet: A large-scale hierarchical image database",
                    "abstract": "The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called \u201cImageNet\u201d, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond."
                },
                {
                    "title": "Automated Flower Classification over a Large Number of Classes",
                    "abstract": "We investigate to what extent combinations of features can improve classification performance on a large dataset of similar classes. To this end we introduce a 103 class flower dataset. We compute four different features for the flowers, each describing different aspects, namely the local shape/texture, the shape of the boundary, the overall spatial distribution of petals, and the colour. We combine the features using a multiple kernel framework with a SVM classifier. The weights for each class are learnt using the method of Varma and Ray, which has achieved state of the art performance on other large dataset, such as Caltech 101/256. Our dataset has a similar challenge in the number of classes, but with the added difficulty of large between class similarity and small within class similarity. Results show that learning the optimum kernel combination of multiple features vastly improves the performance, from 55.1% for the best single feature to 72.8% for the combination of all features."
                },
                {
                    "title": "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding",
                    "abstract": "We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5 (7.7 point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement)."
                },
                {
                    "title": "Visualizing Deep Network Training Trajectories with PCA",
                    "abstract": "Neural network training is a form of numerical optimization in a high-dimensional parameter space. Such optimization processes are rarely visualized because of the dif\ufb01culty of representing high-dimensional dynamics in a visually in-telligible way. We present a simple method for rendering the optimization trajectories of deep networks with low-dimensional plots, using linear projections obtained by principal component analysis. We show that such plots reveal visually distinctive properties of the training processes, and outline opportunities for future investigation."
                },
                {
                    "title": "Parameter Server for Distributed Machine Learning",
                    "abstract": "We propose a parameter server framework to solve distributed machine learning problems. Both data and workload are distributed into client nodes, while server nodes maintain globally shared parameters, which are represented as sparse vectors and matrices. The framework manages asynchronous data communications between clients and servers. Flexible consistency models, elastic scalability and fault tolerance are supported by this framework. We present algorithms and theoretical analysis for challenging nonconvex and nonsmooth problems. To demonstrate the scalability of the proposed framework, we show experimental results on real data with billions of parameters."
                },
                {
                    "title": "Reading Digits in Natural Images with Unsupervised Feature Learning",
                    "abstract": "Detecting and reading text from natural images is a hard computer vision task that is central to a variety of emerging applications. Related problems like document character recognition have been widely studied by computer vision and machine learning researchers and are virtually solved for practical applications like reading handwritten digits. Reliably recognizing characters in more complex scenes like photographs, however, is far more difficult: the best existing methods lag well behind human performance on the same tasks. In this paper we attack the problem of recognizing digits in a real application using unsupervised feature learning methods: reading house numbers from street level photos. To this end, we introduce a new benchmark dataset for research use containing over 600,000 labeled digits cropped from Street View images. We then demonstrate the difficulty of recognizing these digits when the problem is approached with hand-designed features. Finally, we employ variants of two recently proposed unsupervised feature learning methods and find that they are convincingly superior on our benchmarks."
                },
                {
                    "title": "Learning Multiple Layers of Features from Tiny Images",
                    "abstract": "Groups at MIT and NYU have collected a dataset of millions of tiny colour images from the web. It is, in principle, an excellent dataset for unsupervised training of deep generative models, but previous researchers who have tried this have found it dicult to learn a good set of lters from the images. We show how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex. Using a novel parallelization algorithm to distribute the work among multiple machines connected on a network, we show how training such a model can be done in reasonable time. A second problematic aspect of the tiny images dataset is that there are no reliable class labels which makes it hard to use for object recognition experiments. We created two sets of reliable labels. The CIFAR-10 set has 6000 examples of each of 10 classes and the CIFAR-100 set has 600 examples of each of 100 non-overlapping classes. Using these labels, we show that object recognition is signicantly improved by pre-training a layer of features on a large set of unlabeled tiny images."
                },
                {
                    "title": "Flat Minima",
                    "abstract": "We present a new algorithm for finding low-complexity neural networks with high generalization capability. The algorithm searches for a flat minimum of the error function. A flat minimum is a large connected region in weight space where the error remains approximately constant. An MDL-based, Bayesian argument suggests that flat minima correspond to simple networks and low expected overfitting. The argument is based on a Gibbs algorithm variant and a novel way of splitting generalization error into underfitting and overfitting error. Unlike many previous approaches, ours does not require gaussian assumptions and does not depend on a good weight prior. Instead we have a prior over input output functions, thus taking into account net architecture and training set. Although our algorithm requires the computation of second-order derivatives, it has backpropagation's order of complexity. Automatically, it effectively prunes units, weights, and input lines. Various experiments with feedforward and recurrent nets are described. In an application to stock market prediction, flat minimum search outperforms conventional backprop, weight decay, and optimal brain surgeon/optimal brain damage."
                }
            ],
            "categories": [
                "cs.LG",
                "cs.AI"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow does the direction of noise influence the symmetry of loss valleys in deep neural networks?\n\n---\n\n**[Question 2] - Why is it interesting and important?**  \nUnderstanding the symmetry of loss valleys in deep neural networks (DNNs) is crucial for advancing our knowledge of their optimization landscape and generalization capabilities. By addressing this problem, we can provide insights into the conditions that lead to better model performance and robustness. This research could influence future studies on optimization techniques, model architecture design, and the interpretation of DNN behavior, ultimately leading to more effective machine learning applications across various domains.\n\n---\n\n**[Question 3] - Why is it hard?**  \nThe complexity of this problem arises from the intricate nature of DNNs and the multifaceted interactions between convergence solutions and noise. Naive approaches that only consider Gaussian noise may overlook critical factors affecting valley symmetry. Additionally, the challenge lies in accurately visualizing and analyzing the loss landscape, as well as understanding how various network architectures and hyperparameters contribute to valley characteristics. Overcoming these technical and theoretical obstacles requires a nuanced approach to noise direction and its effects on the optimization process.\n\n---\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has primarily focused on symmetric valley shapes and has not systematically explored the impact of asymmetric valleys or the role of noise direction. Limitations in visualization methods and a lack of comprehensive analysis of noise effects have hindered progress in this area. Our approach differs by introducing a detailed examination of various noise directions and their relationship with convergence solutions, providing a more holistic understanding of valley symmetry that previous studies have not addressed.\n\n---\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology involves using the Norm-Scaled (NS) visualization method to analyze the valley shapes of DNNs, considering both convergence solutions and noise direction. We will investigate seven common noise directions and six special ones, focusing on the degree of sign consistency between noise and convergence solutions as a key factor for valley asymmetry. The expected outcomes include empirical observations of valley characteristics across different datasets and network architectures, as well as theoretical insights into the implications of sign-consistent noise on neuron activation and the Hessian matrix's trace, which may indicate flatter regions in the loss landscape."
            }
        },
        "author_data": {
            "842fc6ff-fe21-459b-b76d-692780f89338": {
                "pk": "842fc6ff-fe21-459b-b76d-692780f89338",
                "name": "Xin-Chun Li",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "ea6a0ca6-1d18-4dcd-a54e-80d4f9370642": {
                "pk": "ea6a0ca6-1d18-4dcd-a54e-80d4f9370642",
                "name": "Jin-Lin Tang",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            },
            "01a623fe-405d-44b6-a7a4-e706d0273671": {
                "pk": "01a623fe-405d-44b6-a7a4-e706d0273671",
                "name": "Bo Zhang",
                "collaborators": [
                    "Haiwen Zhang",
                    "Xin Wang",
                    "Bo Yang",
                    "Bo Xiong",
                    "Jinyu Yang",
                    "Bing Zhang",
                    "Ruming Zhang",
                    "Jiayao Zhang"
                ],
                "domain": [
                    "Inverse Problems",
                    "Scattering Theory",
                    "Machine Learning",
                    "Statistical Matching"
                ],
                "publications": [
                    {
                        "title": "Efficient algorithms for building representative matched pairs with enhanced generalizability",
                        "abstract": "Many recent efforts center on assessing the ability of real-world evidence (RWE) generated from non-randomized, observational data to produce results compatible with those from randomized controlled trials (RCTs). One noticeable endeavor is the RCT DUPLICATE initiative (Franklin et al., 2020, 2021). To better reconcile findings from an observational study and an RCT, or two observational studies based on different databases, it is desirable to eliminate differences between study populations. We outline an efficient, network-flow-based statistical matching algorithm that designs well-matched pairs from observational data that resemble the covariate distributions of a target population, for instance, the target-RCT-eligible population in the RCT DUPLICATE initiative studies or a generic population of scientific interest. We demonstrate the usefulness of the method by revisiting the inconsistency regarding a cardioprotective effect of the hormone replacement therapy (HRT) in the Women's Health Initiative (WHI) clinical trial and corresponding observational study. We found that the discrepancy between the trial and observational study persisted in a design that adjusted for study populations' cardiovascular risk profile, but seemed to disappear in a study design that further adjusted for the HRT initiation age and previous estrogen-plus-progestin use. The proposed method is integrated into the R package match2C."
                    },
                    {
                        "title": "Outer-space branch-and-bound algorithm for generalized linear multiplicative programs",
                        "abstract": "This paper introduces a new global optimization algorithm for solving the generalized linear multiplicative problem (GLMP). The algorithm starts by introducing $\\bar{p}$ new variables and applying a logarithmic transformation to convert the problem into an equivalent problem (EP). By using the strong duality of linear program, a new convex relaxation subproblem is formulated to obtain the lower bounds for the optimal value of EP. This relaxation subproblem, combined with a simplicial branching process, forms the foundation of a simplicial branch-and-bound algorithm that can globally solve the problem. The paper also includes an analysis of the theoretical convergence and computational complexity of the algorithm. Additionally, numerical experiments are conducted to demonstrate the effectiveness of the proposed algorithm in various test instances."
                    },
                    {
                        "title": "Path-dependent correlations in dynamically tuned Ising models and its short-time behavior: application of Magnus expansion",
                        "abstract": "We study the buildup of antiferromagnetic (AF) correlation in the dynamically tuned Ising models which are realized by the Rydberg atomic system. In short-time scale, we apply Magnus expansion (ME) to derive the high-order analytic expression of the connected correlation functions and compare it with exactly numerical results for the different lattice geometries, e.g., 1D chain, $2 \\times n$ lattice, and $n \\times n$ lattice. It is shown that the high-order expansion is required to describe accurately the buildup of AF correlation in the quench dynamics. Moreover, through a 2D square lattice, we find that the magnitude of AF correlation for the same Manhattan distance is proportional to the number of the shortest paths in a sufficiently long time until long and distinct paths are involved significantly with the buildup of the correlation. Finally, we propose an applicable experimental setup to realize our findings."
                    },
                    {
                        "title": "A Newton method for simultaneous reconstruction of an interface and a buried obstacle from far-field data",
                        "abstract": "This paper is concerned with the inverse problem of scattering of time-harmonic acoustic waves from a penetrable and buried obstacles. By introducing a related transmission scattering problem, a Newton iteration method is proposed to simultaneously reconstruct both the penetrable interface and the buried obstacle inside from far-field data. A main feature of our method is that we do not need to know the type of boundary conditions on the buried obstacle. In particular, the boundary condition on the buried obstacle can also be determined simultaneously by the method. Finally, numerical examples using multi-frequency data are carried out to illustrate the effectiveness of our method."
                    },
                    {
                        "title": "Artificial Intelligence in Intelligent Tutoring Robots: A Systematic Review and Design Guidelines",
                        "abstract": "This study provides a systematic review of the recent advances in designing the intelligent tutoring robot (ITR), and summarises the status quo of applying artificial intelligence (AI) techniques. We first analyse the environment of the ITR and propose a relationship model for describing interactions of ITR with the students, the social milieu and the curriculum. Then, we transform the relationship model into the perception-planning-action model for exploring what AI techniques are suitable to be applied in the ITR. This article provides insights on promoting human-robot teaching-learning process and AI-assisted educational techniques, illustrating the design guidelines and future research perspectives in intelligent tutoring robots."
                    },
                    {
                        "title": "A novel integral equation for scattering by locally rough surfaces and application to the inverse problem",
                        "abstract": "This paper is concerned with the direct and inverse acoustic or electromagnetic scattering problems by a locally perturbed, perfectly reflecting, infinite plane (which is called a locally rough surface in this paper). We propose a novel integral equation formulation for the direct scattering problem which is defined on a bounded curve (consisting of a bounded part of the infinite plane containing the local perturbation and the lower part of a circle) with two corners. This novel integral equation can be solved efficiently by using the Nystrom method with a graded mesh introduced previously by Kress and is capable of dealing with large wavenumber cases. For the inverse problem, we propose a Newton iteration method to reconstruct the local perturbation of the plane from multiple frequency far-field data, based on the novel integral equation formulation. Numerical examples are carried out to demonstrate that our reconstruction method is stable and accurate even for the case of multiple-scale profiles."
                    },
                    {
                        "title": "Gamma-Ray Burst Prompt Emission Light Curves and Power Density Spectra in the ICMART Model",
                        "abstract": "In this paper, we simulate the prompt emission light curves of gamma-ray bursts (GRBs) within the framework of the Internal-Collision-induced MAgnetic Reconnection and Turbulence (ICMART) model. This model applies to GRBs with a moderately-high magnetization parameter $\\sigma$ in the emission region. We show that this model can produce highly variable light curves with both fast and slow components. The rapid variability is caused by many locally Doppler-boosted mini-emitters due to turbulent magnetic reconnection in a moderately-high-$\\sigma$ flow. The run-away growth and subsequent depletion of these mini-emitters as a function time define a broad slow component for each ICMART event. A GRB light curve is usually composed of multiple ICMART events that are fundamentally driven by the erratic GRB central engine activity. Allowing variations of the model parameters, one is able to reproduce a variety of light curves and the power density spectra as observed."
                    },
                    {
                        "title": "Imaging of locally rough surfaces from intensity-only far-field or near-field data",
                        "abstract": "This paper is concerned with a nonlinear imaging problem, which aims to reconstruct a locally perturbed, perfectly reflecting, infinite plane from intensity-only (or phaseless) far-field or near-field data. A recursive Newton iteration algorithm in frequencies is developed to reconstruct the locally rough surface from multi-frequency intensity-only far-field or near-field data, where the fast integral equation solver developed in [39] is used to solve the direct scattering problem in each iteration. For the case with far-field data, a main feature of our work is that the incident field is taken as a superposition of two plane waves with different directions rather than one plane wave, so the location and shape of the local perturbation of the infinite plane can be reconstructed simultaneously from intensity-only far-field data with multiple wave numbers. This is different from previous work on inverse scattering from phaseless far-field data, where only the shape reconstruction was considered due to the translation invariance property of the phaseless far-field pattern corresponding to one plane wave as the incident field. Finally, numerical examples are carried out to demonstrate that our reconstruction algorithm is stable and accurate even for the case of multiple-scale profiles."
                    },
                    {
                        "title": "Recovering scattering obstacles by multi-frequency phaseless far-field data",
                        "abstract": "It is well known that the modulus of the far-field pattern (or phaseless far-field pattern) is invariant under translations of the scattering obstacle if only one plane wave is used as the incident field, so the shape but not the location of the obstacle can be recovered from the phaseless far-field data. In this paper, it is proved that the translation invariance property of the phaseless far-field pattern can be broken if superpositions of two plane waves are used as the incident fields for all wave numbers in a finite interval. Based on this, a recursive Newton-type iteration algorithm in frequencies is then developed to recover both the location and the shape of the obstacle simultaneously from multi-frequency phaseless far-field data. Numerical examples are also carried out to illustrate the validity of the approach and the effectiveness of the inversion algorithm."
                    },
                    {
                        "title": "An FFT-based algorithm for efficient computation of Green's functions for the Helmholtz and Maxwell's equations in periodic domains",
                        "abstract": "The integral equation method is widely used in numerical simulations of 2D/3D acoustic and electromagnetic scattering problems, which needs a large number of values of the Green's functions. A significant topic is the scattering problems in periodic domains, where the corresponding Green's functions are quasi-periodic. The quasi-periodic Green's functions are defined by series that converge too slowly to be used for calculations. Many mathematicians have developed several efficient numerical methods to calculate quasi-periodic Green's functions. In this paper, we will propose a new FFT-based fast algorithm to compute the 2D/3D quasi-periodic Green's functions for both the Helmholtz equations and Maxwell's equations. The convergence results and error estimates are also investigated in this paper. Further, the numerical examples are given to show that, when a large number of values are needed, the new algorithm is very competitive."
                    },
                    {
                        "title": "Some Reflections on Drawing Causal Inference using Textual Data: Parallels Between Human Subjects and Organized Texts",
                        "abstract": "We examine the role of textual data as study units when conducting causal inference by drawing parallels between human subjects and organized texts. %in human population research. We elaborate on key causal concepts and principles, and expose some ambiguity and sometimes fallacies. To facilitate better framing a causal query, we discuss two strategies: (i) shifting from immutable traits to perceptions of them, and (ii) shifting from some abstract concept/property to its constituent parts, i.e., adopting a constructivist perspective of an abstract concept. We hope this article would raise the awareness of the importance of articulating and clarifying fundamental concepts before delving into developing methodologies when drawing causal inference using textual data."
                    },
                    {
                        "title": "Fast imaging of scattering obstacles from phaseless far-field measurements at a fixed frequency",
                        "abstract": "This paper is concerned with the inverse obstacle scattering problem with phaseless far-field data at a fixed frequency. The main difficulty of this problem is the so-called translation invariance property of the modulus of the far-field pattern or the phaseless far-field pattern generated by one plane wave as the incident field, which means that the location of the obstacle can not be recovered from such phaseless far-field data at a fixed frequency. It was recently proved in our previous work \\cite{XZZ18} that the obstacle can be uniquely determined by the phaseless far-field patterns generated by infinitely many sets of superpositions of two plane waves with different directions at a fixed frequency if the obstacle is a priori known to be a sound-soft or an impedance obstacle with real-valued impedance function. The purpose of this paper is to develop a direct imaging algorithm to reconstruct the location and shape of the obstacle from the phaseless far-field data corresponding to infinitely many sets of superpositions of two plane waves with a fixed frequency as the incident fields. Our imaging algorithm only involves the calculation of the products of the measurement data with two exponential functions at each sampling point and is thus fast and easy to implement. Further, the proposed imaging algorithm does not need to know the type of boundary conditions on the obstacle in advance and is capable to reconstruct multiple obstacles with different boundary conditions. Numerical experiments are also carried out to illustrate that our imaging method is stable, accurate and robust to noise."
                    },
                    {
                        "title": "An approximate factorization method for inverse acoustic scattering with phaseless total-field data",
                        "abstract": "This paper is concerned with the inverse acoustic scattering problem with phaseless total-field data at a fixed frequency. An approximate factorization method is developed to numerically reconstruct both the location and shape of the unknown scatterer from the phaseless total-field data generated by incident plane waves at a fixed frequency and measured on the circle $\\partial B_R$ with a sufficiently large radius $R$. The theoretical analysis of our method is based on the asymptotic property in the operator norm from $H^{1/2}(\\mathbb{S}^1)$ to $H^{-1/2}(\\mathbb{S}^1)$ of the phaseless total-field operator defined in terms of the phaseless total-field data measured on $\\partial B_R$ with large enough $R$, where $H^s(\\mathbb{S}^1)$ is a Sobolev space on the unit circle $\\mathbb{S}^1$ for real number $s$, together with the factorization of a modified far-field operator. The asymptotic property of the phaseless total-field operator is also established in this paper with the theory of oscillatory integrals. The unknown scatterer can be either an impenetrable obstacle of sound-soft, sound-hard or impedance type or an inhomogeneous medium with a compact support, and the proposed inversion algorithm does not need to know the boundary condition of the unknown obstacle in advance. Numerical examples are also carried out to demonstrate the effectiveness of our inversion method. To the best of our knowledge, it is the first attempt to develop a factorization type method for inverse scattering problems with phaseless data."
                    }
                ]
            },
            "66061ddc-2d74-40a8-a1b7-0450cd2a0c92": {
                "pk": "66061ddc-2d74-40a8-a1b7-0450cd2a0c92",
                "name": "Lan Li",
                "collaborators": [
                    "Jinpeng Lv",
                    "Eric Cockayne"
                ],
                "domain": [
                    "Semantic Search",
                    "Large Language Models",
                    "Information Retrieval",
                    "Density Functional Theory"
                ],
                "publications": [
                    {
                        "title": "Unlocking Insights: Semantic Search in Jupyter Notebooks",
                        "abstract": "Semantic search, a process aimed at delivering highly relevant search results by comprehending the searcher's intent and the contextual meaning of terms within a searchable dataspace, plays a pivotal role in information retrieval. In this paper, we investigate the application of large language models to enhance semantic search capabilities, specifically tailored for the domain of Jupyter Notebooks. Our objective is to retrieve generated outputs, such as figures or tables, associated functions and methods, and other pertinent information.   We demonstrate a semantic search framework that achieves a comprehensive semantic understanding of the entire notebook's contents, enabling it to effectively handle various types of user queries. Key components of this framework include:   1). A data preprocessor is designed to handle diverse types of cells within Jupyter Notebooks, encompassing both markdown and code cells. 2). An innovative methodology is devised to address token size limitations that arise with code-type cells. We implement a finer-grained approach to data input, transitioning from the cell level to the function level, effectively resolving these issues."
                    },
                    {
                        "title": "First-Principles Studies of the Atomic, Electronic, and Magnetic Structure of a-MnO2 (Cryptomelane)",
                        "abstract": "Density functional theory calculations are used to investigate a-MnO2, a structure containing a framework of corner and edge sharing MnO6 octahedra with tunnels in between. Placing K+ ions into the tunnels stabilizes a-MnO2 with respect to the rutile-structure b-MnO2 phase, in agreement with experiment. The computed magnetic structure has antiferromagnetic (ferromagnetic) Mn-Mn interactions between corner-sharing (edge-sharing) octahedra. Pure a-MnO2 is found to be a semiconductor with an indirect band gap of 1.3 eV. Water and related hydrides (OH-; H3O+) can also be accommodated in the tunnels; the equilibrium K-O distance increases with increasing oxygen hydride charge."
                    }
                ]
            },
            "c957af3f-7108-432b-944e-d6a099b6b6d0": {
                "pk": "c957af3f-7108-432b-944e-d6a099b6b6d0",
                "name": "De-Chuan Zhan",
                "collaborators": [],
                "domain": [
                    "Graph Neural Network",
                    "Machine Learning",
                    "Multi-task Learning",
                    "AutoML"
                ]
            }
        }
    },
    "2406.05346": {
        "paper_data": {
            "title": "ProG: A Graph Prompt Learning Benchmark",
            "url": "http://arxiv.org/abs/2406.05346v2",
            "arxiv_id": "2406.05346",
            "authors": [
                "Chenyi Zi",
                "Haihong Zhao",
                "Xiangguo Sun",
                "Yiqing Lin",
                "Hong Cheng",
                "Jia Li"
            ],
            "abstract": "Artificial general intelligence on graphs has shown significant advancements across various applications, yet the traditional 'Pre-train & Fine-tune' paradigm faces inefficiencies and negative transfer issues, particularly in complex and few-shot settings. Graph prompt learning emerges as a promising alternative, leveraging lightweight prompts to manipulate data and fill the task gap by reformulating downstream tasks to the pretext. However, several critical challenges still remain: how to unify diverse graph prompt models, how to evaluate the quality of graph prompts, and to improve their usability for practical comparisons and selection. In response to these challenges, we introduce the first comprehensive benchmark for graph prompt learning. Our benchmark integrates SIX pre-training methods and FIVE state-of-the-art graph prompt techniques, evaluated across FIFTEEN diverse datasets to assess performance, flexibility, and efficiency. We also present 'ProG', an easy-to-use open-source library that streamlines the execution of various graph prompt models, facilitating objective evaluations. Additionally, we propose a unified framework that categorizes existing graph prompt methods into two main approaches: prompts as graphs and prompts as tokens. This framework enhances the applicability and comparison of graph prompt techniques. The code is available at: https://github.com/sheldonresearch/ProG.",
            "introduction": "   1 Introduction  Recently, artificial general intelligence (AGI) on graphs has emerged as a new trend in various applications like drug design\u00a0[53, 33], protein prediction\u00a0[9], social analysis\u00a0[44, 58], etc. To achieve this vision, one key question is how to learn useful knowledge from non-linear data (like graphs) and how to apply it to various downstream tasks or domains. Classical approaches mostly leverage the \u2018Pre-train & Fine-tune\u2019 paradigm, which first designs some pretext with easily access data as the pre-training task for the graph neural network model, and then adjusts partial or entire model parameters to fit new downstream tasks. Although much progress has been achieved, they are still not effective and efficient enough. For example, adjusting the pre-trained model will become very time-consuming with the increase in model complexity [24, 23, 43, 22]. In addition, a natural gap between these pretexts and downstream tasks makes the task transferring very hard, and sometimes may even cause negative transfer\u00a0[51]. These problems are particularly serious in few-shot settings.   To further alleviate the problems above, Graph Prompt Learning [45] has attracted more attention recently. As shown in Figure 1, graph prompts seek to manipulate downstream data by inserting an additional small prompt graph and then reformulating the downstream task to the pre-training task without changing the pre-trained Graph Neural Network (GNN) model. Since a graph prompt is usually lightweight, tuning this prompt is more efficient than the large backbone model. Compared with graph \u2018pre-training & fine-tuning\u2019, which can be seen as a model-level retraining skill, graph \u2018pre-training & prompting\u2019 is more customized for data-level operations, which means its potential applications might go beyond task transferring in the graph intelligence area like data quality evaluation, multi-domain alignment, learnable data augmentation, etc. Recently, many graph prompt models have been proposed and presented significant performance in graph learning areas [42, 41, 28, 7, 60, 4], indicating their huge potential towards more general graph intelligence. Despite the enthusiasm around graph prompts in research communities, three significant challenges impede further exploration:   Figure 1: An overview of our benchmark.     \u2776  How can we unify diverse graph prompt models given their varying methodologies? The fragmented landscape of graph prompts hinders systematic research advancement. A unified framework is essential for integrating these diverse methods, creating a cohesive taxonomy to better understand current research and support future studies.    \u2777  How do we evaluate the quality of graph prompts? Assessing the efficiency, power, and flexibility of graph prompts is crucial for understanding their impact and limitations. Currently, there is no standardized benchmark for fair and comprehensive comparison, as existing works have inconsistent experimental setups, varying tasks, metrics, pre-training strategies, and data-splitting techniques, which obstruct a comprehensive understanding of the current state of research.    \u2778  How can we make graph prompt approaches more user-friendly for practical comparison and selection? Despite numerous proposed methods, the lack of an easy-to-use toolkit for creating graph prompts limits their potential applications. The implementation details of existing works vary significantly in programming frameworks, tricks, and running requirements. Developing a standardized library for various graph prompt learning approaches is urgently needed to facilitate broader exploration and application.      In this work, we wish to push",
            "references": [
                {
                    "title": "Effective Fault Scenario Identification for Communication Networks via Knowledge-Enhanced Graph Neural Networks",
                    "abstract": "Fault Scenario Identification (FSI) is a challenging task that aims to automatically identify the fault types in communication networks from massive alarms to guarantee effective fault recoveries. Existing methods are developed based on rules, which are not accurate enough due to the mismatching issue. In this paper, we propose an effective method named Knowledge-Enhanced Graph Neural Network (KE-GNN), the main idea of which is to integrate the advantages of both the rules and GNN. This work is the first work that employs GNN and rules to tackle the FSI task. Specifically, we encode knowledge using propositional logic and map them into a knowledge space. Then, we elaborately design a teacher-student scheme to minimize the distance between the knowledge embedding and the prediction of GNN, integrating knowledge and enhancing the GNN. To validate the performance of the proposed method, we collected and labeled three real-world 5G fault scenario datasets. Extensive evaluation conducted on these datasets indicates that our method achieves the best performance compared with other representative methods, improving the accuracy by up to 8.10%. Furthermore, the proposed method achieves the best performance against a small dataset setting and can be effectively applied to a new carrier site with a different topology structure."
                },
                {
                    "title": "Protein Multimer Structure Prediction via Prompt Learning",
                    "abstract": "Understanding the 3D structures of protein multimers is crucial, as they play a vital role in regulating various cellular processes. It has been empirically confirmed that the multimer structure prediction~(MSP) can be well handled in a step-wise assembly fashion using provided dimer structures and predicted protein-protein interactions~(PPIs). However, due to the biological gap in the formation of dimers and larger multimers, directly applying PPI prediction techniques can often cause a \\textit{poor generalization} to the MSP task. To address this challenge, we aim to extend the PPI knowledge to multimers of different scales~(i.e., chain numbers). Specifically, we propose \\textbf{\\textsc{PromptMSP}}, a pre-training and \\textbf{Prompt} tuning framework for \\textbf{M}ultimer \\textbf{S}tructure \\textbf{P}rediction. First, we tailor the source and target tasks for effective PPI knowledge learning and efficient inference, respectively. We design PPI-inspired prompt learning to narrow the gaps of two task formats and generalize the PPI knowledge to multimers of different scales. We provide a meta-learning strategy to learn a reliable initialization of the prompt model, enabling our prompting framework to effectively adapt to limited data for large-scale multimers. Empirically, we achieve both significant accuracy (RMSD and TM-Score) and efficiency improvements compared to advanced MSP models. The code, data and checkpoints are released at \\url{https://github.com/zqgao22/PromptMSP}."
                },
                {
                    "title": "Advanced Drug Interaction Event Prediction",
                    "abstract": "Predicting drug-drug interaction adverse events, so-called DDI events, is increasingly valuable as it facilitates the study of mechanisms underlying drug use or adverse reactions. Existing models often neglect the distinctive characteristics of individual event classes when integrating multi-source features, which contributes to systematic unfairness when dealing with highly imbalanced event samples. Moreover, the limited capacity of these models to abstract the unique attributes of each event subclass considerably hampers their application in predicting rare drug-drug interaction events with a limited sample size. Reducing dataset bias and abstracting event subclass characteristics are two unresolved challenges. Recently, prompt tuning with frozen pre-trained graph models, namely\"pre-train, prompt, fine-tune\"strategy, has demonstrated impressive performance in few-shot tasks. Motivated by this, we propose an advanced method as a solution to address these aforementioned challenges. Specifically, our proposed approach entails a hierarchical pre-training task that aims to capture crucial aspects of drug molecular structure and intermolecular interactions while effectively mitigating implicit dataset bias within the node embeddings. Furthermore, we construct a prototypical graph by strategically sampling data from distinct event types and design subgraph prompts utilizing pre-trained node features. Through comprehensive benchmark experiments, we validate the efficacy of our subgraph prompts in accurately representing event classes and achieve exemplary results in both overall and subclass prediction tasks."
                },
                {
                    "title": "All in One and One for All: A Simple yet Effective Method towards Cross-domain Graph Pretraining",
                    "abstract": "Large Language Models (LLMs) have revolutionized the fields of computer vision (CV) and natural language processing (NLP). One of the most notable advancements of LLMs is that a single model is trained on vast and diverse datasets spanning multiple domains -- a paradigm we term `All in One'. This methodology empowers LLMs with super generalization capabilities, facilitating an encompassing comprehension of varied data distributions. Leveraging these capabilities, a single LLM demonstrates remarkable versatility across a variety of domains -- a paradigm we term `One for All'. However, applying this idea to the graph field remains a formidable challenge, with cross-domain pretraining often resulting in negative transfer. This issue is particularly important in few-shot learning scenarios, where the paucity of training data necessitates the incorporation of external knowledge sources. In response to this challenge, we propose a novel approach called Graph COordinators for PrEtraining (GCOPE), that harnesses the underlying commonalities across diverse graph datasets to enhance few-shot learning. Our novel methodology involves a unification framework that amalgamates disparate graph datasets during the pretraining phase to distill and transfer meaningful knowledge to target tasks. Extensive experiments across multiple graph datasets demonstrate the superior efficacy of our approach. By successfully leveraging the synergistic potential of multiple graph datasets for pretraining, our work stands as a pioneering contribution to the realm of graph foundational model."
                },
                {
                    "title": "Prompt Learning on Temporal Interaction Graphs",
                    "abstract": "Temporal Interaction Graphs (TIGs) are widely utilized to represent real-world systems. To facilitate representation learning on TIGs, researchers have proposed a series of TIG models. However, these models are still facing two tough gaps between the pre-training and downstream predictions in their ``pre-train, predict'' training paradigm. First, the temporal discrepancy between the pre-training and inference data severely undermines the models' applicability in distant future predictions on the dynamically evolving data. Second, the semantic divergence between pretext and downstream tasks hinders their practical applications, as they struggle to align with their learning and prediction capabilities across application scenarios. Recently, the ``pre-train, prompt'' paradigm has emerged as a lightweight mechanism for model generalization. Applying this paradigm is a potential solution to solve the aforementioned challenges. However, the adaptation of this paradigm to TIGs is not straightforward. The application of prompting in static graph contexts falls short in temporal settings due to a lack of consideration for time-sensitive dynamics and a deficiency in expressive power. To address this issue, we introduce Temporal Interaction Graph Prompting (TIGPrompt), a versatile framework that seamlessly integrates with TIG models, bridging both the temporal and semantic gaps. In detail, we propose a temporal prompt generator to offer temporally-aware prompts for different tasks. These prompts stand out for their minimalistic design, relying solely on the tuning of the prompt generator with very little supervision data. To cater to varying computational resource demands, we propose an extended ``pre-train, prompt-based fine-tune'' paradigm, offering greater flexibility. Through extensive experiments, the TIGPrompt demonstrates the SOTA performance and remarkable efficiency advantages."
                },
                {
                    "title": "GraphPro: Graph Pre-training and Prompt Learning for Recommendation",
                    "abstract": "GNN-based recommendation systems have been successful in capturing complex user-item interactions using multi-hop message passing. However, these methods often struggle to handle the dynamic nature of user-item interactions, making it challenging to adapt to changes in user preferences and new data distributions. This limits their scalability and performance in real-world dynamic scenarios. In our study, we propose a framework called GraphPro that combines dynamic graph pre-training with prompt learning in an efficient way. This unique approach allows GNNs to effectively capture both long-term user preferences and short-term behavior changes, resulting in accurate and up-to-date recommendations. To address the issue of changing user preferences, we integrate a temporal prompt mechanism and a graph-structural prompt learning mechanism into the pre-trained GNN architecture. The temporal prompt mechanism incorporates time-related information into user-item interactions, enabling the model to naturally incorporate temporal dynamics. The graph-structural prompt learning mechanism allows the model to apply pre-trained insights to new behavior dynamics without the need for continuous retraining. We also introduce a dynamic evaluation framework for recommendations that better reflects real-world scenarios and reduces the offline-online discrepancy. Through comprehensive experiments, including deployment in a large-scale industrial scenario, we demonstrate the seamless scalability of GraphPro with various leading recommenders. Our results highlight the superiority of GraphPro in terms of effectiveness, robustness, and efficiency. We release the model implementation at the link: https://github.com/HKUDS/GraphPro."
                },
                {
                    "title": "Graph Prompt Learning: A Comprehensive Survey and Beyond",
                    "abstract": "Artificial General Intelligence (AGI) has revolutionized numerous fields, yet its integration with graph data, a cornerstone in our interconnected world, remains nascent. This paper presents a pioneering survey on the emerging domain of graph prompts in AGI, addressing key challenges and opportunities in harnessing graph data for AGI applications. Despite substantial advancements in AGI across natural language processing and computer vision, the application to graph data is relatively underexplored. This survey critically evaluates the current landscape of AGI in handling graph data, highlighting the distinct challenges in cross-modality, cross-domain, and cross-task applications specific to graphs. Our work is the first to propose a unified framework for understanding graph prompt learning, offering clarity on prompt tokens, token structures, and insertion patterns in the graph domain. We delve into the intrinsic properties of graph prompts, exploring their flexibility, expressiveness, and interplay with existing graph models. A comprehensive taxonomy categorizes over 100 works in this field, aligning them with pre-training tasks across node-level, edge-level, and graph-level objectives. Additionally, we present, ProG, a Python library, and an accompanying website, to support and advance research in graph prompting. The survey culminates in a discussion of current challenges and future directions, offering a roadmap for research in graph prompting within AGI. Through this comprehensive analysis, we aim to catalyze further exploration and practical applications of AGI in graph data, underlining its potential to reshape AGI fields and beyond. ProG and the website can be accessed by \\url{https://github.com/WxxShirley/Awesome-Graph-Prompt}, and \\url{https://github.com/sheldonresearch/ProG}, respectively."
                },
                {
                    "title": "ULTRA-DP: Unifying Graph Pre-training with Multi-task Graph Dual Prompt",
                    "abstract": "Recent research has demonstrated the efficacy of pre-training graph neural networks (GNNs) to capture the transferable graph semantics and enhance the performance of various downstream tasks. However, the semantic knowledge learned from pretext tasks might be unrelated to the downstream task, leading to a semantic gap that limits the application of graph pre-training. To reduce this gap, traditional approaches propose hybrid pre-training to combine various pretext tasks together in a multi-task learning fashion and learn multi-grained knowledge, which, however, cannot distinguish tasks and results in some transferable task-specific knowledge distortion by each other. Moreover, most GNNs cannot distinguish nodes located in different parts of the graph, making them fail to learn position-specific knowledge and lead to suboptimal performance. In this work, inspired by the prompt-based tuning in natural language processing, we propose a unified framework for graph hybrid pre-training which injects the task identification and position identification into GNNs through a prompt mechanism, namely multi-task graph dual prompt (ULTRA-DP). Based on this framework, we propose a prompt-based transferability test to find the most relevant pretext task in order to reduce the semantic gap. To implement the hybrid pre-training tasks, beyond the classical edge prediction task (node-node level), we further propose a novel pre-training paradigm based on a group of $k$-nearest neighbors (node-group level). The combination of them across different scales is able to comprehensively express more structural semantics and derive richer multi-grained knowledge. Extensive experiments show that our proposed ULTRA-DP can significantly enhance the performance of hybrid pre-training methods and show the generalizability to other pre-training tasks and backbone architectures."
                },
                {
                    "title": "GSLB: The Graph Structure Learning Benchmark",
                    "abstract": "Graph Structure Learning (GSL) has recently garnered considerable attention due to its ability to optimize both the parameters of Graph Neural Networks (GNNs) and the computation graph structure simultaneously. Despite the proliferation of GSL methods developed in recent years, there is no standard experimental setting or fair comparison for performance evaluation, which creates a great obstacle to understanding the progress in this field. To fill this gap, we systematically analyze the performance of GSL in different scenarios and develop a comprehensive Graph Structure Learning Benchmark (GSLB) curated from 20 diverse graph datasets and 16 distinct GSL algorithms. Specifically, GSLB systematically investigates the characteristics of GSL in terms of three dimensions: effectiveness, robustness, and complexity. We comprehensively evaluate state-of-the-art GSL algorithms in node- and graph-level tasks, and analyze their performance in robust learning and model complexity. Further, to facilitate reproducible research, we have developed an easy-to-use library for training, evaluating, and visualizing different GSL methods. Empirical results of our extensive experiments demonstrate the ability of GSL and reveal its potential benefits on various downstream tasks, offering insights and opportunities for future research. The code of GSLB is available at: https://github.com/GSL-Benchmark/GSLB."
                },
                {
                    "title": "Structure Learning Via Meta-Hyperedge for Dynamic Rumor Detection",
                    "abstract": "Online social networks have greatly facilitated our lives but have also propagated the spreading of rumours. Traditional works mostly find rumors from content, but content can be strategically manipulated to evade such detection, making these methods brittle. To improve the accuracy and robustness of rumor detection, we propose to integrate and exploit the content, propagation structure, and temporal relations because information in the networks always spreads dynamically with significant structures. In this paper, we propose a novel rumor detection framework in online temporal networks via structure learning. Specifically, to exploit the propagation structure, we propose a novel hyperedge walking strategy on a meta-hyperedge graph to learn the representations of sub-structures in the networks. Then a hyperedge expansion method is proposed to generate more global structural features. The expanded hyperedges are more hierarchical, making the learned structural embeddings more expressive. To make full use of content, we design a hypergraph learning model using hyperedge expansion to fuse node content with structural features and generate comprehensive representations for the entire graph. To exploit temporal relations, we design a masked temporal attention unit for learning the evolving patterns of the network. Extensive evaluations with six state-of-the-art baselines on two real-world datasets demonstrate the superiority of our solution."
                },
                {
                    "title": "Can Large Language Models Empower Molecular Property Prediction?",
                    "abstract": "Molecular property prediction has gained significant attention due to its transformative potential in multiple scientific disciplines. Conventionally, a molecule graph can be represented either as a graph-structured data or a SMILES text. Recently, the rapid development of Large Language Models (LLMs) has revolutionized the field of NLP. Although it is natural to utilize LLMs to assist in understanding molecules represented by SMILES, the exploration of how LLMs will impact molecular property prediction is still in its early stage. In this work, we advance towards this objective through two perspectives: zero/few-shot molecular classification, and using the new explanations generated by LLMs as representations of molecules. To be specific, we first prompt LLMs to do in-context molecular classification and evaluate their performance. After that, we employ LLMs to generate semantically enriched explanations for the original SMILES and then leverage that to fine-tune a small-scale LM model for multiple downstream tasks. The experimental results highlight the superiority of text explanations as molecular representations across multiple benchmark datasets, and confirm the immense potential of LLMs in molecular property prediction tasks. Codes are available at \\url{https://github.com/ChnQ/LLM4Mol}."
                },
                {
                    "title": "All in One: Multi-Task Prompting for Graph Neural Networks",
                    "abstract": "Recently, \"pre-training and fine-tuning'' has been adopted as a standard workflow for many graph tasks since it can take general graph knowledge to relieve the lack of graph annotations from each application. However, graph tasks with node level, edge level, and graph level are far diversified, making the pre-training pretext often incompatible with these multiple tasks. This gap may even cause a \"negative transfer'' to the specific application, leading to poor results. Inspired by the prompt learning in natural language processing (NLP), which has presented significant effectiveness in leveraging prior knowledge for various NLP tasks, we study the prompting topic for graphs with the motivation of filling the gap between pre-trained models and various graph tasks. In this paper, we propose a novel multi-task prompting method for graph models. Specifically, we first unify the format of graph prompts and language prompts with the prompt token, token structure, and inserting pattern. In this way, the prompting idea from NLP can be seamlessly introduced to the graph area. Then, to further narrow the gap between various graph tasks and state-of-the-art pre-training strategies, we further study the task space of various graph applications and reformulate downstream problems to the graph-level task. Afterward, we introduce meta-learning to efficiently learn a better initialization for the multi-task prompt of graphs so that our prompting framework can be more reliable and general for different tasks. We conduct extensive experiments, results from which demonstrate the superiority of our method."
                },
                {
                    "title": "Virtual Node Tuning for Few-shot Node Classification",
                    "abstract": "Few-shot Node Classification (FSNC) is a challenge in graph representation learning where only a few labeled nodes per class are available for training. To tackle this issue, meta-learning has been proposed to transfer structural knowledge from base classes with abundant labels to target novel classes. However, existing solutions become ineffective or inapplicable when base classes have no or limited labeled nodes. To address this challenge, we propose an innovative method dubbed Virtual Node Tuning (VNT). Our approach utilizes a pretrained graph transformer as the encoder and injects virtual nodes as soft prompts in the embedding space, which can be optimized with few-shot labels in novel classes to modulate node embeddings for each specific FSNC task. A unique feature of VNT is that, by incorporating a Graph-based Pseudo Prompt Evolution (GPPE) module, VNT-GPPE can handle scenarios with sparse labels in base classes. Experimental results on four datasets demonstrate the superiority of the proposed approach in addressing FSNC with unlabeled or sparsely labeled base classes, outperforming existing state-of-the-art methods and even fully supervised baselines."
                },
                {
                    "title": "PRODIGY: Enabling In-context Learning Over Graphs",
                    "abstract": "In-context learning is the ability of a pretrained model to adapt to novel and diverse downstream tasks by conditioning on prompt examples, without optimizing any parameters. While large language models have demonstrated this ability, how in-context learning could be performed over graphs is unexplored. In this paper, we develop \\textbf{Pr}etraining \\textbf{O}ver \\textbf{D}iverse \\textbf{I}n-Context \\textbf{G}raph S\\textbf{y}stems (PRODIGY), the first pretraining framework that enables in-context learning over graphs. The key idea of our framework is to formulate in-context learning over graphs with a novel \\emph{prompt graph} representation, which connects prompt examples and queries. We then propose a graph neural network architecture over the prompt graph and a corresponding family of in-context pretraining objectives. With PRODIGY, the pretrained model can directly perform novel downstream classification tasks on unseen graphs via in-context learning. We provide empirical evidence of the effectiveness of our framework by showcasing its strong in-context learning performance on tasks involving citation networks and knowledge graphs. Our approach outperforms the in-context learning accuracy of contrastive pretraining baselines with hard-coded adaptation by 18\\% on average across all setups. Moreover, it also outperforms standard finetuning with limited data by 33\\% on average with in-context learning."
                },
                {
                    "title": "GPT-4 Technical Report",
                    "abstract": "We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4."
                },
                {
                    "title": "SGL-PT: A Strong Graph Learner with Graph Prompt Tuning",
                    "abstract": "Recently, much exertion has been paid to design graph self-supervised methods to obtain generalized pre-trained models, and adapt pre-trained models onto downstream tasks through fine-tuning. However, there exists an inherent gap between pretext and downstream graph tasks, which insufficiently exerts the ability of pre-trained models and even leads to negative transfer. Meanwhile, prompt tuning has seen emerging success in natural language processing by aligning pre-training and fine-tuning with consistent training objectives. In this paper, we identify the challenges for graph prompt tuning: The first is the lack of a strong and universal pre-training task across sundry pre-training methods in graph domain. The second challenge lies in the difficulty of designing a consistent training objective for both pre-training and downstream tasks. To overcome above obstacles, we propose a novel framework named SGL-PT which follows the learning strategy ``Pre-train, Prompt, and Predict''. Specifically, we raise a strong and universal pre-training task coined as SGL that acquires the complementary merits of generative and contrastive self-supervised graph learning. And aiming for graph classification task, we unify pre-training and fine-tuning by designing a novel verbalizer-free prompting function, which reformulates the downstream task in a similar format as pretext task. Empirical results show that our method surpasses other baselines under unsupervised setting, and our prompt tuning method can greatly facilitate models on biological datasets over fine-tuning methods."
                },
                {
                    "title": "GraphPrompt: Unifying Pre-Training and Downstream Tasks for Graph Neural Networks",
                    "abstract": "Graphs can model complex relationships between objects, enabling a myriad of Web applications such as online page/article classification and social recommendation. While graph neural networks (GNNs) have emerged as a powerful tool for graph representation learning, in an end-to-end supervised setting, their performance heavily relies on a large amount of task-specific supervision. To reduce labeling requirement, the \u201cpre-train, fine-tune\u201d and \u201cpre-train, prompt\u201d paradigms have become increasingly common. In particular, prompting is a popular alternative to fine-tuning in natural language processing, which is designed to narrow the gap between pre-training and downstream objectives in a task-specific manner. However, existing study of prompting on graphs is still limited, lacking a universal treatment to appeal to different downstream tasks. In this paper, we propose GraphPrompt, a novel pre-training and prompting framework on graphs. GraphPrompt not only unifies pre-training and downstream tasks into a common task template, but also employs a learnable prompt to assist a downstream task in locating the most relevant knowledge from the pre-trained model in a task-specific manner. Finally, we conduct extensive experiments on five public datasets to evaluate and analyze GraphPrompt."
                },
                {
                    "title": "Self-Supervised Hypergraph Representation Learning for Sociological Analysis",
                    "abstract": "Modern sociology has profoundly uncovered many convincing social criteria for behavioral analysis. Unfortunately, many of them are too subjective to be measured and very challenging to be presented in online social networks (OSNs) for the large data volume and complicated environments to be explored. On the other hand, data mining techniques can better find data patterns but many of them leave behind unnatural understanding to humans. Although there are some works trying to integrate social observations for specific tasks, they are still hard to be applied to more general cases. In this paper, we propose a fundamental methodology to support the further fusion of data mining techniques and sociological behavioral criteria. Our highlights are three-fold: First, we propose an effective hypergraph awareness and a fast line graph construction framework. The hypergraph can more profoundly indicate the interactions between individuals and their environments because each edge in the hypergraph (a.k.a hyperedge) contains more than two nodes, which is perfect to describe social. A line graph treats each social environment as a super node with the underlying influence between different environments. In this way, we go beyond traditional pair-wise relations and explore richer patterns under various sociological criteria; Second, we propose a novel hypergraph-based neural network to learn social influence flowing from users to users, users to environments, environment to users, and environments to environments. The neural network can be learned via a task-free method, making our model very flexible to support various data mining tasks and sociological analysis; Third, we propose both qualitative and quantitive solutions to effectively evaluate the most common sociological criteria like social conformity, social equivalence, environmental evolving and social polarization. Our extensive experiments show that our framework can better support both data mining tasks for online user behaviors and sociological analysis."
                },
                {
                    "title": "Universal Prompt Tuning for Graph Neural Networks",
                    "abstract": "In recent years, prompt tuning has sparked a research surge in adapting pre-trained models. Unlike the unified pre-training strategy employed in the language field, the graph field exhibits diverse pre-training strategies, posing challenges in designing appropriate prompt-based tuning methods for graph neural networks. While some pioneering work has devised specialized prompting functions for models that employ edge prediction as their pre-training tasks, these methods are limited to specific pre-trained GNN models and lack broader applicability. In this paper, we introduce a universal prompt-based tuning method called Graph Prompt Feature (GPF) for pre-trained GNN models under any pre-training strategy. GPF operates on the input graph's feature space and can theoretically achieve an equivalent effect to any form of prompting function. Consequently, we no longer need to illustrate the prompting function corresponding to each pre-training strategy explicitly. Instead, we employ GPF to obtain the prompted graph for the downstream task in an adaptive manner. We provide rigorous derivations to demonstrate the universality of GPF and make guarantee of its effectiveness. The experimental results under various pre-training strategies indicate that our method performs better than fine-tuning, with an average improvement of about 1.4% in full-shot scenarios and about 3.2% in few-shot scenarios. Moreover, our method significantly outperforms existing specialized prompt-based tuning methods when applied to models utilizing the pre-training strategy they specialize in. These numerous advantages position our method as a compelling alternative to fine-tuning for downstream adaptations."
                },
                {
                    "title": "GPPT: Graph Pre-training and Prompt Tuning to Generalize Graph Neural Networks",
                    "abstract": "Despite the promising representation learning of graph neural networks (GNNs), the supervised training of GNNs notoriously requires large amounts of labeled data from each application. An effective solution is to apply the transfer learning in graph: using easily accessible information to pre-train GNNs, and fine-tuning them to optimize the downstream task with only a few labels. Recently, many efforts have been paid to design the self-supervised pretext tasks, and encode the universal graph knowledge among the various applications. However, they rarely notice the inherent training objective gap between the pretext and downstream tasks. This significant gap often requires costly fine-tuning for adapting the pre-trained model to downstream problem, which prevents the efficient elicitation of pre-trained knowledge and then results in poor results. Even worse, the naive pre-training strategy usually deteriorates the downstream task, and damages the reliability of transfer learning in graph data. To bridge the task gap, we propose a novel transfer learning paradigm to generalize GNNs, namely graph pre-training and prompt tuning (GPPT). Specifically, we first adopt the masked edge prediction, the most simplest and popular pretext task, to pre-train GNNs. Based on the pre-trained model, we propose the graph prompting function to modify the standalone node into a token pair, and reformulate the downstream node classification looking the same as edge prediction. The token pair is consisted of candidate label class and node entity. Therefore, the pre-trained GNNs could be applied without tedious fine-tuning to evaluate the linking probability of token pair, and produce the node classification decision. The extensive experiments on eight benchmark datasets demonstrate the superiority of GPPT, delivering an average improvement of 4.29% in few-shot graph analysis and accelerating the model convergence up to 4.32X. The code is available in: https://github.com/MingChen-Sun/GPPT."
                },
                {
                    "title": "Wiener Graph Deconvolutional Network Improves Graph Self-Supervised Learning",
                    "abstract": "Graph self-supervised learning (SSL) has been vastly employed to learn representations from unlabeled graphs. Existing methods can be roughly divided into predictive learning and contrastive learning, where the latter one attracts more research attention with better empirical performance. We argue that, however, predictive models weaponed with powerful decoder could achieve comparable or even better representation power than contrastive models. In this work, we propose a Wiener Graph Deconvolutional Network (WGDN), an augmentation-adaptive decoder empowered by graph wiener filter to perform information reconstruction. Theoretical analysis proves the superior reconstruction ability of graph wiener filter. Extensive experimental results on various datasets demonstrate the effectiveness of our approach."
                },
                {
                    "title": "GraphMAE: Self-Supervised Masked Graph Autoencoders",
                    "abstract": "Self-supervised learning (SSL) has been extensively explored in recent years. Particularly, generative SSL has seen emerging success in natural language processing and other fields, such as the wide adoption of BERT and GPT. Despite this, contrastive learning---which heavily relies on structural data augmentation and complicated training strategies---has been the dominant approach in graph SSL, while the progress of generative SSL on graphs, especially graph autoencoders (GAEs), has thus far not reached the potential as promised in other fields. In this paper, we identify and examine the issues that negatively impact the development of GAEs, including their reconstruction objective, training robustness, and error metric. We present a masked graph autoencoder GraphMAE (code is publicly available at https://github.com/THUDM/GraphMAE) that mitigates these issues for generative self-supervised graph learning. Instead of reconstructing structures, we propose to focus on feature reconstruction with both a masking strategy and scaled cosine error that benefit the robust training of GraphMAE. We conduct extensive experiments on 21 public datasets for three different graph learning tasks. The results manifest that GraphMAE---a simple graph autoencoder with our careful designs---can consistently generate outperformance over both contrastive and generative state-of-the-art baselines. This study provides an understanding of graph autoencoders and demonstrates the potential of generative self-supervised learning on graphs."
                },
                {
                    "title": "FAITH: Few-Shot Graph Classification with Hierarchical Task Graphs",
                    "abstract": "Few-shot graph classification aims at predicting classes for graphs, given limited labeled graphs for each class. To tackle the bottleneck of label scarcity, recent works propose to incorporate few-shot learning frameworks for fast adaptations to graph classes with limited labeled graphs. Specifically, these works propose to accumulate meta-knowledge across diverse meta-training tasks, and then generalize such meta-knowledge to the target task with a disjoint label set. However, existing methods generally ignore task correlations among meta-training tasks while treating them independently. Nevertheless, such task correlations can advance the model generalization to the target task for better classification performance. On the other hand, it remains non-trivial to utilize task correlations due to the complex components in a large number of meta-training tasks. To deal with this, we propose a novel few-shot learning framework FAITH that captures task correlations via constructing a hierarchical task graph at different granularities. Then we further design a loss-based sampling strategy to select tasks with more correlated classes. Moreover, a task-specific classifier is proposed to utilize the learned task correlations for few-shot classification. Extensive experiments on four prevalent few-shot graph classification datasets demonstrate the superiority of FAITH over other state-of-the-art baselines."
                },
                {
                    "title": "Pre-training graph neural networks for link prediction in biomedical networks",
                    "abstract": "MOTIVATION\nGraphs or networks are widely utilized to model the interactions between different entities (e.g., proteins, drugs, etc) for biomedical applications. Predicting potential interactions/links in biomedical networks is important for understanding the pathological mechanisms of various complex human diseases, as well as screening compound targets for drug discovery. Graph neural networks (GNNs) have been utilized for link prediction in various biomedical networks, which rely on the node features extracted from different data sources, e.g., sequence, structure and network data. However, it is challenging to effectively integrate these data sources and automatically extract features for different link prediction tasks.\n\n\nRESULTS\nIn this paper, we propose a novel Pre-Training Graph Neural Networks based framework named PT-GNN to integrate different data sources for link prediction in biomedical networks. First, we design expressive deep learning methods (e.g., convolutional neural network (CNN) and graph convolutional network (GCN)) to learn features for individual nodes from sequence and structure data. Second, we further propose a GCN-based encoder to effectively refine the node features by modelling the dependencies among nodes in the network. Third, the node features are pre-trained based on graph reconstruction tasks. The pre-trained features can be used for model initialization in downstream tasks. Extensive experiments have been conducted on two critical link prediction tasks, i.e., synthetic lethality (SL) prediction and drug-target interaction (DTI) prediction. Experimental results demonstrate PT-GNN outperforms the state-of-the-art methods for SL prediction and DTI prediction. In addition, the pre-trained features benefit improving the performance and reduce the training time of existing models.\n\n\nAVAILABILITY\nPython codes and dataset are available at: https://github.com/longyahui/PT-GNN."
                },
                {
                    "title": "Self-Supervised Representation Learning via Latent Graph Prediction",
                    "abstract": "Self-supervised learning (SSL) of graph neural networks is emerging as a promising way of leveraging unlabeled data. Currently, most methods are based on contrastive learning adapted from the image domain, which requires view generation and a sufficient number of negative samples. In contrast, existing predictive models do not require negative sampling, but lack theoretical guidance on the design of pretext training tasks. In this work, we propose the LaGraph, a theoretically grounded predictive SSL framework based on latent graph prediction. Learning objectives of LaGraph are derived as self-supervised upper bounds to objectives for predicting unobserved latent graphs. In addition to its improved performance, LaGraph provides explanations for recent successes of predictive models that include invariance-based objectives. We provide theoretical analysis comparing LaGraph to related methods in different domains. Our experimental results demonstrate the superiority of LaGraph in performance and the robustness to decreasing of training sample size on both graph-level and node-level tasks."
                },
                {
                    "title": "SimGRACE: A Simple Framework for Graph Contrastive Learning without Data Augmentation",
                    "abstract": "Graph contrastive learning (GCL) has emerged as a dominant technique for graph representation learning which maximizes the mutual information between paired graph augmentations that share the same semantics. Unfortunately, it is difficult to preserve semantics well during augmentations in view of the diverse nature of graph data. Currently, data augmentations in GCL broadly fall into three unsatisfactory ways. First, the augmentations can be manually picked per dataset by trial-and-errors. Second, the augmentations can be selected via cumbersome search. Third, the augmentations can be obtained with expensive domain knowledge as guidance. All of these limit the efficiency and more general applicability of existing GCL methods. To circumvent these crucial issues, we propose a Simple framework for GRAph Contrastive lEarning, SimGRACE for brevity, which does not require data augmentations. Specifically, we take original graph as input and GNN model with its perturbed version as two encoders to obtain two correlated views for contrast. SimGRACE is inspired by the observation that graph data can preserve their semantics well during encoder perturbations while not requiring manual trial-and-errors, cumbersome search or expensive domain knowledge for augmentations selection. Also, we explain why SimGRACE can succeed. Furthermore, we devise adversarial training scheme, dubbed AT-SimGRACE, to enhance the robustness of graph contrastive learning and theoretically explain the reasons. Albeit simple, we show that SimGRACE can yield competitive or better performance compared with state-of-the-art methods in terms of generalizability, transferability and robustness, while enjoying unprecedented degree of flexibility and efficiency. The code is available at: https://github.com/junxia97/SimGRACE."
                },
                {
                    "title": "InfoGCL: Information-Aware Graph Contrastive Learning",
                    "abstract": "Various graph contrastive learning models have been proposed to improve the performance of learning tasks on graph datasets in recent years. While effective and prevalent, these models are usually carefully customized. In particular, although all recent researches create two contrastive views, they differ greatly in view augmentations, architectures, and objectives. It remains an open question how to build your graph contrastive learning model from scratch for particular graph learning tasks and datasets. In this work, we aim to fill this gap by studying how graph information is transformed and transferred during the contrastive learning process and proposing an information-aware graph contrastive learning framework called InfoGCL. The key point of this framework is to follow the Information Bottleneck principle to reduce the mutual information between contrastive parts while keeping task-relevant information intact at both the levels of the individual module and the entire framework so that the information loss during graph representation learning can be minimized. We show for the first time that all recent graph contrastive learning methods can be unified by our framework. We empirically validate our theoretical analysis on both node and graph classification benchmark datasets, and demonstrate that our algorithm significantly outperforms the state-of-the-arts."
                },
                {
                    "title": "Contrastive Pre-Training of GNNs on Heterogeneous Graphs",
                    "abstract": "While graph neural networks (GNNs) emerge as the state-of-the-art representation learning methods on graphs, they often require a large amount of labeled data to achieve satisfactory performance, which is often expensive or unavailable. To relieve the label scarcity issue, some pre-training strategies have been devised for GNNs, to learn transferable knowledge from the universal structural properties of the graph. However, existing pre-training strategies are only designed for homogeneous graphs, in which each node and edge belongs to the same type. In contrast, a heterogeneous graph embodies rich semantics, as multiple types of nodes interact with each other via different kinds of edges, which are neglected by existing strategies. In this paper, we propose a novel Contrastive Pre-Training strategy of GNNs on Heterogeneous Graphs (CPT-HG), to capture both the semantic and structural properties in a self-supervised manner. Specifically, we design semantic-aware pre-training tasks at both the relation- and subgraph-levels, and further enhance their representativeness by employing contrastive learning. We conduct extensive experiments on three real-world heterogeneous graphs, and promising results demonstrate the superior ability of our CPT-HG to transfer knowledge to various downstream tasks via pre-training."
                },
                {
                    "title": "AFEC: Active Forgetting of Negative Transfer in Continual Learning",
                    "abstract": "Continual learning aims to learn a sequence of tasks from dynamic data distributions. Without accessing to the old training samples, knowledge transfer from the old tasks to each new task is difficult to determine, which might be either positive or negative. If the old knowledge interferes with the learning of a new task, i.e., the forward knowledge transfer is negative, then precisely remembering the old tasks will further aggravate the interference, thus decreasing the performance of continual learning. By contrast, biological neural networks can actively forget the old knowledge that conflicts with the learning of a new experience, through regulating the learning-triggered synaptic expansion and synaptic convergence. Inspired by the biological active forgetting, we propose to actively forget the old knowledge that limits the learning of new tasks to benefit continual learning. Under the framework of Bayesian continual learning, we develop a novel approach named Active Forgetting with synaptic Expansion-Convergence (AFEC). Our method dynamically expands parameters to learn each new task and then selectively combines them, which is formally consistent with the underlying mechanism of biological active forgetting. We extensively evaluate AFEC on a variety of continual learning benchmarks, including CIFAR-10 regression tasks, visual classification tasks and Atari reinforcement tasks, where AFEC effectively improves the learning of new tasks and achieves the state-of-the-art performance in a plug-and-play way."
                },
                {
                    "title": "Automated Anomaly Detection via Curiosity-Guided Search and Self-Imitation Learning",
                    "abstract": "Anomaly detection is an important data mining task with numerous applications, such as intrusion detection, credit card fraud detection, and video surveillance. However, given a specific complicated task with complicated data, the process of building an effective deep learning-based system for anomaly detection still highly relies on human expertise and laboring trials. Also, while neural architecture search (NAS) has shown its promise in discovering effective deep architectures in various domains, such as image classification, object detection, and semantic segmentation, contemporary NAS methods are not suitable for anomaly detection due to the lack of intrinsic search space, unstable search process, and low sample efficiency. To bridge the gap, in this article, we propose AutoAD, an automated anomaly detection framework, which aims to search for an optimal neural network model within a predefined search space. Specifically, we first design a curiosity-guided search strategy to overcome the curse of local optimality. A controller, which acts as a search agent, is encouraged to take actions to maximize the information gain about the controller\u2019s internal belief. We further introduce an experience replay mechanism based on self-imitation learning to improve the sample efficiency. Experimental results on various real-world benchmark datasets demonstrate that the deep model identified by AutoAD achieves the best performance, comparing with existing handcrafted models and traditional search methods."
                },
                {
                    "title": "Training Graph Neural Networks with 1000 Layers",
                    "abstract": "Deep graph neural networks (GNNs) have achieved excellent results on various tasks on increasingly large graph datasets with millions of nodes and edges. However, memory complexity has become a major obstacle when training deep GNNs for practical applications due to the immense number of nodes, edges, and intermediate activations. To improve the scalability of GNNs, prior works propose smart graph sampling or partitioning strategies to train GNNs with a smaller set of nodes or sub-graphs. In this work, we study reversible connections, group convolutions, weight tying, and equilibrium models to advance the memory and parameter efficiency of GNNs. We find that reversible connections in combination with deep network architectures enable the training of overparameterized GNNs that significantly outperform existing methods on multiple datasets. Our models RevGNN-Deep (1001 layers with 80 channels each) and RevGNN-Wide (448 layers with 224 channels each) were both trained on a single commodity GPU and achieve an ROC-AUC of $87.74 \\pm 0.13$ and $88.24 \\pm 0.15$ on the ogbn-proteins dataset. To the best of our knowledge, RevGNN-Deep is the deepest GNN in the literature by one order of magnitude. Please visit our project website https://www.deepgcns.org/arch/gnn1000 for more information."
                },
                {
                    "title": "Multi-Scale Contrastive Siamese Networks for Self-Supervised Graph Representation Learning",
                    "abstract": "Graph representation learning plays a vital role in processing graph-structured data. However, prior arts on graph representation learning heavily rely on labeling information. To overcome this problem, inspired by the recent success of graph contrastive learning and Siamese networks in visual representation learning, we propose a novel self-supervised approach in this paper to learn node representations by enhancing Siamese self-distillation with multi-scale contrastive learning. Specifically, we first generate two augmented views from the input graph based on local and global perspectives. Then, we employ two objectives called cross-view and cross-network contrastiveness to maximize the agreement between node representations across different views and networks. To demonstrate the effectiveness of our approach, we perform empirical experiments on five real-world datasets. Our method not only achieves new state-of-the-art results but also surpasses some semi-supervised counterparts by large margins. Code is made available at https://github.com/GRAND-Lab/MERIT"
                },
                {
                    "title": "Node Similarity Preserving Graph Convolutional Networks",
                    "abstract": "Graph Neural Networks (GNNs) have achieved tremendous success in various real-world applications due to their strong ability in graph representation learning. GNNs explore the graph structure and node features by aggregating and transforming information within node neighborhoods. However, through theoretical and empirical analysis, we reveal that the aggregation process of GNNs tends to destroy node similarity in the original feature space. There are many scenarios where node similarity plays a crucial role. Thus, it has motivated the proposed framework SimP-GCN that can effectively and efficiently preserve node similarity while exploiting graph structure. Specifically, to balance information from graph structure and node features, we propose a feature similarity preserving aggregation which adaptively integrates graph structure and node features. Furthermore, we employ self-supervised learning to explicitly capture the complex feature similarity and dissimilarity relations between nodes. We validate the effectiveness of SimP-GCN on seven benchmark datasets including three assortative and four disassorative graphs. The results demonstrate that SimP-GCN outperforms representative baselines. Further probe shows various advantages of the proposed framework. The implementation of SimP-GCN is available at https://github.com/ChandlerBang/SimP-GCN."
                },
                {
                    "title": "Graph Contrastive Learning with Adaptive Augmentation",
                    "abstract": "Recently, contrastive learning (CL) has emerged as a successful method for unsupervised graph representation learning. Most graph CL methods first perform stochastic augmentation on the input graph to obtain two graph views and maximize the agreement of representations in the two views. Despite the prosperous development of graph CL methods, the design of graph augmentation schemes\u2014a crucial component in CL\u2014remains rarely explored. We argue that the data augmentation schemes should preserve intrinsic structures and attributes of graphs, which will force the model to learn representations that are insensitive to perturbation on unimportant nodes and edges. However, most existing methods adopt uniform data augmentation schemes, like uniformly dropping edges and uniformly shuffling features, leading to suboptimal performance. In this paper, we propose a novel graph contrastive representation learning method with adaptive augmentation that incorporates various priors for topological and semantic aspects of the graph. Specifically, on the topology level, we design augmentation schemes based on node centrality measures to highlight important connective structures. On the node attribute level, we corrupt node features by adding more noise to unimportant node features, to enforce the model to recognize underlying semantic information. We perform extensive experiments of node classification on a variety of real-world datasets. Experimental results demonstrate that our proposed method consistently outperforms existing state-of-the-art baselines and even surpasses some supervised counterparts, which validates the effectiveness of the proposed contrastive framework with adaptive augmentation."
                },
                {
                    "title": "Graph Contrastive Learning with Augmentations",
                    "abstract": "Generalizable, transferrable, and robust representation learning on graph-structured data remains a challenge for current graph neural networks (GNNs). Unlike what has been developed for convolutional neural networks (CNNs) for image data, self-supervised learning and pre-training are less explored for GNNs. In this paper, we propose a graph contrastive learning (GraphCL) framework for learning unsupervised representations of graph data. We first design four types of graph augmentations to incorporate various priors. We then systematically study the impact of various combinations of graph augmentations on multiple datasets, in four different settings: semi-supervised, unsupervised, and transfer learning as well as adversarial attacks. The results show that, even without tuning augmentation extents nor using sophisticated GNN architectures, our GraphCL framework can produce graph representations of similar or better generalizability, transferrability, and robustness compared to state-of-the-art methods. We also investigate the impact of parameterized graph augmentation extents and patterns, and observe further performance gains in preliminary experiments. Our codes are available at https://github.com/Shen-Lab/GraphCL."
                },
                {
                    "title": "Masked Label Prediction: Unified Massage Passing Model for Semi-Supervised Classification",
                    "abstract": "Graph neural network (GNN) and label propagation algorithm (LPA) are both message passing algorithms, which have achieved superior performance in semi-supervised classification. GNN performs feature propagation by a neural network to make predictions, while LPA uses label propagation across graph adjacency matrix to get results. However, there is still no effective way to directly combine these two kinds of algorithms. To address this issue, we propose a novel Unified Message Passaging Model (UniMP) that can incorporate feature and label propagation at both training and inference time. First, UniMP adopts a Graph Transformer network, taking feature embedding and label embedding as input information for propagation. Second, to train the network without overfitting in self-loop input label information, UniMP introduces a masked label prediction strategy, in which some percentage of input label information are masked at random, and then predicted. UniMP conceptually unifies feature propagation and label propagation and is empirically powerful. It obtains new state-of-the-art semi-supervised classification results in Open Graph Benchmark (OGB)."
                },
                {
                    "title": "Self-Supervised Graph Transformer on Large-Scale Molecular Data",
                    "abstract": "How to obtain informative representations of molecules is a crucial prerequisite in AI-driven drug design and discovery. Recent researches abstract molecules as graphs and employ Graph Neural Networks (GNNs) for molecular representation learning. Nevertheless, two issues impede the usage of GNNs in real scenarios: (1) insufficient labeled molecules for supervised training; (2) poor generalization capability to new-synthesized molecules. To address them both, we propose a novel framework, GROVER, which stands for Graph Representation frOm self-superVised mEssage passing tRansformer. With carefully designed self-supervised tasks in node-, edge- and graph-level, GROVER can learn rich structural and semantic information of molecules from enormous unlabelled molecular data. Rather, to encode such complex information, GROVER integrates Message Passing Networks into the Transformer-style architecture to deliver a class of more expressive encoders of molecules. The flexibility of GROVER allows it to be trained efficiently on large-scale molecular dataset without requiring any supervision, thus being immunized to the two issues mentioned above. We pre-train GROVER with 100 million parameters on 10 million unlabelled molecules -- the biggest GNN and the largest training dataset in molecular representation learning. We then leverage the pre-trained GROVER for molecular property prediction followed by task-specific fine-tuning, where we observe a huge improvement (more than 6% on average) from current state-of-the-art methods on 11 challenging benchmarks. The insights we gained are that well-designed self-supervision losses and largely-expressive pre-trained models enjoy the significant potential on performance boosting."
                },
                {
                    "title": "Open Graph Benchmark: Datasets for Machine Learning on Graphs",
                    "abstract": "We present the Open Graph Benchmark (OGB), a diverse set of challenging and realistic benchmark datasets to facilitate scalable, robust, and reproducible graph machine learning (ML) research. OGB datasets are large-scale (up to 100+ million nodes and 1+ billion edges), encompass multiple important graph ML tasks, and cover a diverse range of domains, ranging from social and information networks to biological networks, molecular graphs, source code ASTs, and knowledge graphs. For each dataset, we provide a unified evaluation protocol using meaningful application-specific data splits and evaluation metrics. In addition to building the datasets, we also perform extensive benchmark experiments for each dataset. Our experiments suggest that OGB datasets present significant challenges of scalability to large-scale graphs and out-of-distribution generalization under realistic data splits, indicating fruitful opportunities for future research. Finally, OGB provides an automated end-to-end graph ML pipeline that simplifies and standardizes the process of graph data loading, experimental setup, and model evaluation. OGB will be regularly updated and welcomes inputs from the community. OGB datasets as well as data loaders, evaluation scripts, baseline code, and leaderboards are publicly available at this https URL ."
                },
                {
                    "title": "Geom-GCN: Geometric Graph Convolutional Networks",
                    "abstract": "Message-passing neural networks (MPNNs) have been successfully applied in a wide variety of applications in the real world. However, two fundamental weaknesses of MPNNs' aggregators limit their ability to represent graph-structured data: losing the structural information of nodes in neighborhoods and lacking the ability to capture long-range dependencies in disassortative graphs. Few studies have noticed the weaknesses from different perspectives. From the observations on classical neural network and network geometry, we propose a novel geometric aggregation scheme for graph neural networks to overcome the two weaknesses. The behind basic idea is the aggregation on a graph can benefit from a continuous space underlying the graph. The proposed aggregation scheme is permutation-invariant and consists of three modules, node embedding, structural neighborhood, and bi-level aggregation. We also present an implementation of the scheme in graph convolutional networks, termed Geom-GCN, to perform transductive learning on graphs. Experimental results show the proposed Geom-GCN achieved state-of-the-art performance on a wide range of open datasets of graphs."
                },
                {
                    "title": "Microsoft Academic Graph: When experts are not enough",
                    "abstract": "Abstract An ongoing project explores the extent to which artificial intelligence (AI), specifically in the areas of natural language processing and semantic reasoning, can be exploited to facilitate the studies of science by deploying software agents equipped with natural language understanding capabilities to read scholarly publications on the web. The knowledge extracted by these AI agents is organized into a heterogeneous graph, called Microsoft Academic Graph (MAG), where the nodes and the edges represent the entities engaging in scholarly communications and the relationships among them, respectively. The frequently updated data set and a few software tools central to the underlying AI components are distributed under an open data license for research and commercial applications. This paper describes the design, schema, and technical and business motivations behind MAG and elaborates how MAG can be used in analytics, search, and recommendation scenarios. How AI plays an important role in avoiding various biases and human induced errors in other data sets and how the technologies can be further improved in the future are also discussed."
                },
                {
                    "title": "PyTorch: An Imperative Style, High-Performance Deep Learning Library",
                    "abstract": "Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks."
                },
                {
                    "title": "DeepGCNs: Making GCNs Go as Deep as CNNs",
                    "abstract": "Convolutional neural networks (CNNs) have been very successful at solving a variety of computer vision tasks such as object classification and detection, semantic segmentation, activity understanding, to name just a few. One key enabling factor for their great performance has been the ability to train very deep networks. Despite their huge success in many tasks, CNNs do not work well with non-euclidean data, which is prevalent in many real-world applications. Graph Convolutional Networks (GCNs) offer an alternative that allows for non-Eucledian data input to a neural network. While GCNs already achieve encouraging results, they are currently limited to architectures with a relatively small number of layers, primarily due to vanishing gradients during training. This work transfers concepts such as residual/dense connections and dilated convolutions from CNNs to GCNs in order to successfully train very deep GCNs. We show the benefit of using deep GCNs (with as many as 112 layers) experimentally across various datasets and tasks. Specifically, we achieve very promising performance in part segmentation and semantic segmentation on point clouds and in node classification of protein functions across biological protein-protein interaction (PPI) graphs. We believe that the insights in this work will open avenues for future research on GCNs and their application to further tasks not explored in this paper. The source code for this work is available at https://github.com/lightaime/deep_gcns_torch and https://github.com/lightaime/deep_gcns for PyTorch and TensorFlow implementations respectively."
                },
                {
                    "title": "InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation Learning via Mutual Information Maximization",
                    "abstract": "This paper studies learning the representations of whole graphs in both unsupervised and semi-supervised scenarios. Graph-level representations are critical in a variety of real-world applications such as predicting the properties of molecules and community analysis in social networks. Traditional graph kernel based methods are simple, yet effective for obtaining fixed-length representations for graphs but they suffer from poor generalization due to hand-crafted designs. There are also some recent methods based on language models (e.g. graph2vec) but they tend to only consider certain substructures (e.g. subtrees) as graph representatives. Inspired by recent progress of unsupervised representation learning, in this paper we proposed a novel method called InfoGraph for learning graph-level representations. We maximize the mutual information between the graph-level representation and the representations of substructures of different scales (e.g., nodes, edges, triangles). By doing so, the graph-level representations encode aspects of the data that are shared across different scales of substructures. Furthermore, we further propose InfoGraph*, an extension of InfoGraph for semi-supervised scenarios. InfoGraph* maximizes the mutual information between unsupervised graph representations learned by InfoGraph and the representations learned by existing supervised methods. As a result, the supervised encoder learns from unlabeled data while preserving the latent semantic space favored by the current supervised task. Experimental results on the tasks of graph classification and molecular property prediction show that InfoGraph is superior to state-of-the-art baselines and InfoGraph* can achieve performance competitive with state-of-the-art semi-supervised models."
                },
                {
                    "title": "Strategies for Pre-training Graph Neural Networks",
                    "abstract": "Many applications of machine learning require a model to make accurate pre-dictions on test examples that are distributionally different from training ones, while task-specific labels are scarce during training. An effective approach to this challenge is to pre-train a model on related tasks where data is abundant, and then fine-tune it on a downstream task of interest. While pre-training has been effective in many language and vision domains, it remains an open question how to effectively use pre-training on graph datasets. In this paper, we develop a new strategy and self-supervised methods for pre-training Graph Neural Networks (GNNs). The key to the success of our strategy is to pre-train an expressive GNN at the level of individual nodes as well as entire graphs so that the GNN can learn useful local and global representations simultaneously. We systematically study pre-training on multiple graph classification datasets. We find that naive strategies, which pre-train GNNs at the level of either entire graphs or individual nodes, give limited improvement and can even lead to negative transfer on many downstream tasks. In contrast, our strategy avoids negative transfer and improves generalization significantly across downstream tasks, leading up to 9.4% absolute improvements in ROC-AUC over non-pre-trained models and achieving state-of-the-art performance for molecular property prediction and protein function prediction."
                },
                {
                    "title": "DeepGCNs: Can GCNs Go As Deep As CNNs?",
                    "abstract": "Convolutional Neural Networks (CNNs) achieve impressive performance in a wide variety of fields. Their success benefited from a massive boost when very deep CNN models were able to be reliably trained. Despite their merits, CNNs fail to properly address problems with non-Euclidean data. To overcome this challenge, Graph Convolutional Networks (GCNs) build graphs to represent non-Euclidean data, borrow concepts from CNNs, and apply them in training. GCNs show promising results, but they are usually limited to very shallow models due to the vanishing gradient problem. As a result, most state-of-the-art GCN models are no deeper than 3 or 4 layers. In this work, we present new ways to successfully train very deep GCNs. We do this by borrowing concepts from CNNs, specifically residual/dense connections and dilated convolutions, and adapting them to GCN architectures. Extensive experiments show the positive effect of these deep GCN frameworks. Finally, we use these new concepts to build a very deep 56-layer GCN, and show how it significantly boosts performance (+3.7% mIoU over state-of-the-art) in the task of point cloud semantic segmentation. We believe that the community can greatly benefit from this work, as it opens up many opportunities for advancing GCN-based research."
                },
                {
                    "title": "Fast Graph Representation Learning with PyTorch Geometric",
                    "abstract": "We introduce PyTorch Geometric, a library for deep learning on irregularly structured input data such as graphs, point clouds and manifolds, built upon PyTorch. In addition to general graph data structures and processing methods, it contains a variety of recently published methods from the domains of relational learning and 3D data processing. PyTorch Geometric achieves high data throughput by leveraging sparse GPU acceleration, by providing dedicated CUDA kernels and by introducing efficient mini-batch handling for input examples of different size. In this work, we present the library in detail and perform a comprehensive comparative study of the implemented methods in homogeneous evaluation scenarios."
                },
                {
                    "title": "Deep Graph Infomax",
                    "abstract": "We present Deep Graph Infomax (DGI), a general approach for learning node representations within graph-structured data in an unsupervised manner. DGI relies on maximizing mutual information between patch representations and corresponding high-level summaries of graphs---both derived using established graph convolutional network architectures. The learnt patch representations summarize subgraphs centered around nodes of interest, and can thus be reused for downstream node-wise learning tasks. In contrast to most prior approaches to unsupervised learning with GCNs, DGI does not rely on random walk objectives, and is readily applicable to both transductive and inductive learning setups. We demonstrate competitive performance on a variety of node classification benchmarks, which at times even exceeds the performance of supervised learning."
                },
                {
                    "title": "Graph Attention Networks",
                    "abstract": "We present graph attention networks (GATs), novel neural network architectures that operate on graph-structured data, leveraging masked self-attentional layers to address the shortcomings of prior methods based on graph convolutions or their approximations. By stacking layers in which nodes are able to attend over their neighborhoods' features, we enable (implicitly) specifying different weights to different nodes in a neighborhood, without requiring any kind of costly matrix operation (such as inversion) or depending on knowing the graph structure upfront. In this way, we address several key challenges of spectral-based graph neural networks simultaneously, and make our model readily applicable to inductive as well as transductive problems. Our GAT models have achieved or matched state-of-the-art results across four established transductive and inductive graph benchmarks: the Cora, Citeseer and Pubmed citation network datasets, as well as a protein-protein interaction dataset (wherein test graphs remain unseen during training)."
                },
                {
                    "title": "Inductive Representation Learning on Large Graphs",
                    "abstract": "Low-dimensional embeddings of nodes in large graphs have proved extremely useful in a variety of prediction tasks, from content recommendation to identifying protein functions. However, most existing approaches require that all nodes in the graph are present during training of the embeddings; these previous approaches are inherently transductive and do not naturally generalize to unseen nodes. Here we present GraphSAGE, a general, inductive framework that leverages node feature information (e.g., text attributes) to efficiently generate node embeddings for previously unseen data. Instead of training individual embeddings for each node, we learn a function that generates embeddings by sampling and aggregating features from a node's local neighborhood. Our algorithm outperforms strong baselines on three inductive node-classification benchmarks: we classify the category of unseen nodes in evolving information graphs based on citation and Reddit post data, and we show that our algorithm generalizes to completely unseen graphs using a multi-graph dataset of protein-protein interactions."
                },
                {
                    "title": "Semi-Supervised Classification with Graph Convolutional Networks",
                    "abstract": "We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin."
                },
                {
                    "title": "Deep Graph Kernels",
                    "abstract": "In this paper, we present Deep Graph Kernels, a unified framework to learn latent representations of sub-structures for graphs, inspired by latest advancements in language modeling and deep learning. Our framework leverages the dependency information between sub-structures by learning their latent representations. We demonstrate instances of our framework on three popular graph kernels, namely Graphlet kernels, Weisfeiler-Lehman subtree kernels, and Shortest-Path graph kernels. Our experiments on several benchmark datasets show that Deep Graph Kernels achieve significant improvements in classification accuracy over state-of-the-art graph kernels."
                },
                {
                    "title": "The Network Data Repository with Interactive Graph Analytics and Visualization",
                    "abstract": "\n \n NetworkRepository (NR) is the first interactive data repository with a web-based platform for visual interactive analytics. Unlike other data repositories (e.g., UCI ML Data Repository, and SNAP), the network data repository (networkrepository.com) allows users to not only download, but to interactively analyze and visualize such data using our web-based interactive graph analytics platform. Users can in real-time analyze, visualize, compare, and explore data along many different dimensions. The aim of NR is to make it easy to discover key insights into the data extremely fast with little effort while also providing a medium for users to share data, visualizations, and insights. Other key factors that differentiate NR from the current data repositories is the number of graph datasets, their size, and variety. While other data repositories are static, they also lack a means for users to collaboratively discuss a particular dataset, corrections, or challenges with using the data for certain applications. In contrast, NR incorporates many social and collaborative aspects that facilitate scientific research, e.g., users can discuss each graph, post observations, and visualizations.\n \n"
                },
                {
                    "title": "Subgraph Matching Kernels for Attributed Graphs",
                    "abstract": "We propose graph kernels based on subgraph matchings, i.e. structure-preserving bijections between subgraphs. While recently proposed kernels based on common subgraphs (Wale et al., 2008; Shervashidze et al., 2009) in general can not be applied to attributed graphs, our approach allows to rate mappings of subgraphs by a exible scoring scheme comparing vertex and edge attributes by kernels. We show that subgraph matching kernels generalize several known kernels. To compute the kernel we propose a graph-theoretical algorithm inspired by a classical relation between common subgraphs of two graphs and cliques in their product graph observed by Levi (1973). Encouraging experimental results on a classification task of real-world graphs are presented."
                },
                {
                    "title": "Social influence analysis in large-scale networks",
                    "abstract": "In large social networks, nodes (users, entities) are influenced by others for various reasons. For example, the colleagues have strong influence on one's work, while the friends have strong influence on one's daily life. How to differentiate the social influences from different angles(topics)? How to quantify the strength of those social influences? How to estimate the model on real large networks?\n To address these fundamental questions, we propose Topical Affinity Propagation (TAP) to model the topic-level social influence on large networks. In particular, TAP can take results of any topic modeling and the existing network structure to perform topic-level influence propagation. With the help of the influence analysis, we present several important applications on real data sets such as 1) what are the representative nodes on a given topic? 2) how to identify the social influences of neighboring nodes on a particular node?\n To scale to real large networks, TAP is designed with efficient distributed learning algorithms that is implemented and tested under the Map-Reduce framework. We further present the common characteristics of distributed learning algorithms for Map-Reduce. Finally, we demonstrate the effectiveness and efficiency of TAP on real large data sets."
                },
                {
                    "title": "Collective Classification in Network Data",
                    "abstract": "Many real-world applications produce networked data such as the world-wide web (hypertext documents connected via hyperlinks), social networks (for example, people connected by friendship links), communication networks (computers connected via communication links) and biological networks (for example, protein interaction networks). A recent focus in machine learning research has been to extend traditional machine learning classification techniques to classify nodes in such networks. In this article, we provide a brief introduction to this area of research and how it has progressed during the past decade. We introduce four of the most widely used inference algorithms for classifying networked data and empirically compare them on both synthetic and real-world data."
                },
                {
                    "title": "AdapterGNN: Efficient Delta Tuning Improves Generalization Ability in Graph Neural Networks",
                    "abstract": "Fine-tuning pre-trained models has recently yielded remarkable performance gains in graph neural networks (GNNs). In addition to pre-training techniques, inspired by the latest work in the natural language \ufb01elds, more recent work has shifted towards applying effective \ufb01ne-tuning approaches, such as parameter-ef\ufb01cient tuning (delta tuning). However, given the substantial differences between GNNs and transformer-based models, applying such approaches directly to GNNs proved to be less effective. In this paper, we present a comprehensive comparison of delta tuning techniques for GNNs and propose a novel delta tuning method speci\ufb01cally designed for GNNs, called AdapterGNN. AdapterGNN preserves the knowledge of the large pre-trained model and leverages highly expressive adapters for GNNs, which can adapt to downstream tasks effectively with only a few parameters, while also improving the model\u2019s generalization ability on the downstream tasks. Extensive experiments show that AdapterGNN achieves higher evaluation performance (outperforming full \ufb01ne-tuning by 1.4% and 5.5% in the chemistry and biology domains respectively, with only 5% of its parameters tuned) and lower generalization gaps compared to full \ufb01ne-tuning. Moreover, we empirically show that a larger GNN model can have a worse generalization ability, which differs from the trend observed in large language models. We have also provided a theoretical justi\ufb01cation for delta tuning can improve the generalization ability of GNNs by applying generalization bounds."
                },
                {
                    "title": "Query-driven Active Surveying for Collective Classification",
                    "abstract": "In network classification problems such as those found in intelligence gathering, public health, and viral marketing, one is often only interested in inferring the labels of a subset of the nodes. We refer to this subset as the query set, and define the problem as query-driven collective classification. We study this problem in a practical active learning framework, in which the learning algorithm can survey non-query nodes to obtain their labels and network structure. We derive a surveying strategy aimed toward optimal inference on the query set. Considering both feature and structural smoothness, concepts that we formally define, we develop an algorithm which adaptively selects survey nodes by estimating which form of smoothness is most appropriate. We evaluate our algorithm on several network datasets and demonstrate its improvements over standard active learning methods."
                },
                {
                    "title": "Protein function prediction via graph kernels",
                    "abstract": "MOTIVATION\nComputational approaches to protein function prediction infer protein function by finding proteins with similar sequence, structure, surface clefts, chemical properties, amino acid motifs, interaction partners or phylogenetic profiles. We present a new approach that combines sequential, structural and chemical information into one graph model of proteins. We predict functional class membership of enzymes and non-enzymes using graph kernels and support vector machine classification on these protein graphs.\n\n\nRESULTS\nOur graph model, derivable from protein sequence and structure only, is competitive with vector models that require additional protein information, such as the size of surface pockets. If we include this extra information into our graph model, our classifier yields significantly higher accuracy levels than the vector models. Hyperkernels allow us to select and to optimally combine the most relevant node attributes in our protein graphs. We have laid the foundation for a protein function prediction system that integrates protein information from various sources efficiently and effectively.\n\n\nAVAILABILITY\nMore information available via www.dbs.ifi.lmu.de/Mitarbeiter/borgwardt.html."
                },
                {
                    "title": "To transfer or not to transfer",
                    "abstract": "With transfer learning, one set of tasks is used to bias learning and improve performance on another task. However, transfer learning may actually hinder performance if the tasks are too dissimilar. As described in this paper, one challenge for transfer learning research is to develop approaches that detect and avoid negative transfer using very little data from the target task."
                }
            ],
            "categories": [
                "cs.LG"
            ],
            "evaluation_format": {
                "model": "gpt-4o-mini",
                "content": "**[Question 1] - What is the problem?**  \nHow can we unify diverse graph prompt models, evaluate their quality, and make them more user-friendly for practical comparison and selection in the context of graph neural networks?\n\n**[Question 2] - Why is it interesting and important?**  \nSolving this problem is crucial for the research community as it addresses the fragmented landscape of graph prompt methodologies, enabling systematic advancement in graph learning. A unified framework will facilitate better understanding and comparison of existing methods, leading to improved performance in various applications such as drug design and social analysis. By establishing standardized evaluation metrics and user-friendly toolkits, future research can build upon a solid foundation, fostering innovation and practical applications in graph intelligence.\n\n**[Question 3] - Why is it hard?**  \nThe challenges in solving this problem stem from the diversity of existing graph prompt models, which complicates the creation of a unified framework. Additionally, the lack of standardized benchmarks for evaluating graph prompts makes it difficult to assess their effectiveness and limitations. Naive approaches may fail because they do not account for the varying methodologies and experimental setups, leading to inconsistent results. Furthermore, the technical complexity of developing a user-friendly toolkit that accommodates different programming frameworks and implementation details poses a significant obstacle.\n\n**[Question 4] - Why hasn't it been solved before?**  \nPrevious research has been limited by the lack of a cohesive taxonomy for graph prompt models, which has hindered systematic exploration and comparison. Existing solutions often focus on specific methodologies without addressing the broader landscape, leading to gaps in understanding. Barriers such as inconsistent experimental setups and varying metrics have prevented comprehensive evaluations. Our approach differs by aiming to create a unified framework, standardized benchmarks, and a user-friendly toolkit, which collectively address these limitations and facilitate further research.\n\n**[Question 5] - What are the key components of my approach and results?**  \nOur proposed methodology includes developing a unified framework for graph prompt models, establishing standardized evaluation metrics, and creating a user-friendly toolkit for practical implementation. We will utilize a diverse set of datasets relevant to graph learning tasks and employ metrics that assess efficiency, power, and flexibility of graph prompts. The expected outcomes include a comprehensive understanding of graph prompt methodologies, improved performance in downstream tasks, and a widely accessible toolkit that encourages broader exploration and application of graph prompt learning techniques."
            }
        },
        "author_data": {
            "a6ab882b-b5d8-4fc1-9b4a-f408fa29055c": {
                "pk": "a6ab882b-b5d8-4fc1-9b4a-f408fa29055c",
                "name": "Chenyi Zi",
                "collaborators": [
                    "Chen Zhang",
                    "Yan Zhou",
                    "Jia Li",
                    "Haihong Zhao",
                    "Yang Liu",
                    "Ziqi Gao",
                    "Tao Feng",
                    "Jiaxuan You"
                ],
                "domain": [
                    "Anomaly Detection",
                    "Knowledge Integration",
                    "Protein Structure Prediction",
                    "Generative Adversarial Networks"
                ],
                "publications": [
                    {
                        "title": "Weakly Supervised Anomaly Detection via Knowledge-Data Alignment",
                        "abstract": "Anomaly detection (AD) plays a pivotal role in numerous web-based applications, including malware detection, anti-money laundering, device failure detection, and network fault analysis. Most methods, which rely on unsupervised learning, are hard to reach satisfactory detection accuracy due to the lack of labels. Weakly Supervised Anomaly Detection (WSAD) has been introduced with a limited number of labeled anomaly samples to enhance model performance. Nevertheless, it is still challenging for models, trained on an inadequate amount of labeled data, to generalize to unseen anomalies. In this paper, we introduce a novel framework Knowledge-Data Alignment (KDAlign) to integrate rule knowledge, typically summarized by human experts, to supplement the limited labeled data. Specifically, we transpose these rules into the knowledge space and subsequently recast the incorporation of knowledge as the alignment of knowledge and data. To facilitate this alignment, we employ the Optimal Transport (OT) technique. We then incorporate the OT distance as an additional loss term to the original objective function of WSAD methodologies. Comprehensive experimental results on five real-world datasets demonstrate that our proposed KDAlign framework markedly surpasses its state-of-the-art counterparts, achieving superior performance across various anomaly types."
                    },
                    {
                        "title": "Deep Reinforcement Learning for Modelling Protein Complexes",
                        "abstract": "AlphaFold can be used for both single-chain and multi-chain protein structure prediction, while the latter becomes extremely challenging as the number of chains increases. In this work, by taking each chain as a node and assembly actions as edges, we show that an acyclic undirected connected graph can be used to predict the structure of multi-chain protein complexes (a.k.a., protein complex modelling, PCM). However, there are still two challenges: 1) The huge combinatorial optimization space of $N^{N-2}$ ($N$ is the number of chains) for the PCM problem can easily lead to high computational cost. 2) The scales of protein complexes exhibit distribution shift due to variance in chain numbers, which calls for the generalization in modelling complexes of various scales. To address these challenges, we propose GAPN, a Generative Adversarial Policy Network powered by domain-specific rewards and adversarial loss through policy gradient for automatic PCM prediction. Specifically, GAPN learns to efficiently search through the immense assembly space and optimize the direct docking reward through policy gradient. Importantly, we design an adversarial reward function to enhance the receptive field of our model. In this way, GAPN will simultaneously focus on a specific batch of complexes and the global assembly rules learned from complexes with varied chain numbers. Empirically, we have achieved both significant accuracy (measured by RMSD and TM-Score) and efficiency improvements compared to leading PCM softwares."
                    }
                ]
            },
            "253e3394-c4d5-4a9b-944d-935a5e4b94dc": {
                "pk": "253e3394-c4d5-4a9b-944d-935a5e4b94dc",
                "name": "Haihong Zhao",
                "collaborators": [
                    "Jia Li",
                    "Yang Liu",
                    "Chenyi Zi",
                    "Chen Zhang",
                    "Yan Zhou",
                    "Aochuan Chen",
                    "Xiangguo Sun",
                    "Hong Cheng",
                    "Jiashun Cheng",
                    "Tingyang Xu"
                ],
                "domain": [
                    "Anomaly Detection",
                    "Graph Neural Network",
                    "Few-Shot Learning",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "Weakly Supervised Anomaly Detection via Knowledge-Data Alignment",
                        "abstract": "Anomaly detection (AD) plays a pivotal role in numerous web-based applications, including malware detection, anti-money laundering, device failure detection, and network fault analysis. Most methods, which rely on unsupervised learning, are hard to reach satisfactory detection accuracy due to the lack of labels. Weakly Supervised Anomaly Detection (WSAD) has been introduced with a limited number of labeled anomaly samples to enhance model performance. Nevertheless, it is still challenging for models, trained on an inadequate amount of labeled data, to generalize to unseen anomalies. In this paper, we introduce a novel framework Knowledge-Data Alignment (KDAlign) to integrate rule knowledge, typically summarized by human experts, to supplement the limited labeled data. Specifically, we transpose these rules into the knowledge space and subsequently recast the incorporation of knowledge as the alignment of knowledge and data. To facilitate this alignment, we employ the Optimal Transport (OT) technique. We then incorporate the OT distance as an additional loss term to the original objective function of WSAD methodologies. Comprehensive experimental results on five real-world datasets demonstrate that our proposed KDAlign framework markedly surpasses its state-of-the-art counterparts, achieving superior performance across various anomaly types."
                    },
                    {
                        "title": "All in One and One for All: A Simple yet Effective Method towards Cross-domain Graph Pretraining",
                        "abstract": "Large Language Models (LLMs) have revolutionized the fields of computer vision (CV) and natural language processing (NLP). One of the most notable advancements of LLMs is that a single model is trained on vast and diverse datasets spanning multiple domains -- a paradigm we term `All in One'. This methodology empowers LLMs with super generalization capabilities, facilitating an encompassing comprehension of varied data distributions. Leveraging these capabilities, a single LLM demonstrates remarkable versatility across a variety of domains -- a paradigm we term `One for All'. However, applying this idea to the graph field remains a formidable challenge, with cross-domain pretraining often resulting in negative transfer. This issue is particularly important in few-shot learning scenarios, where the paucity of training data necessitates the incorporation of external knowledge sources. In response to this challenge, we propose a novel approach called Graph COordinators for PrEtraining (GCOPE), that harnesses the underlying commonalities across diverse graph datasets to enhance few-shot learning. Our novel methodology involves a unification framework that amalgamates disparate graph datasets during the pretraining phase to distill and transfer meaningful knowledge to target tasks. Extensive experiments across multiple graph datasets demonstrate the superior efficacy of our approach. By successfully leveraging the synergistic potential of multiple graph datasets for pretraining, our work stands as a pioneering contribution to the realm of graph foundational model."
                    },
                    {
                        "title": "SEGNO: Generalizing Equivariant Graph Neural Networks with Physical Inductive Biases",
                        "abstract": "Graph Neural Networks (GNNs) with equivariant properties have emerged as powerful tools for modeling complex dynamics of multi-object physical systems. However, their generalization ability is limited by the inadequate consideration of physical inductive biases: (1) Existing studies overlook the continuity of transitions among system states, opting to employ several discrete transformation layers to learn the direct mapping between two adjacent states; (2) Most models only account for first-order velocity information, despite the fact that many physical systems are governed by second-order motion laws. To incorporate these inductive biases, we propose the Second-order Equivariant Graph Neural Ordinary Differential Equation (SEGNO). Specifically, we show how the second-order continuity can be incorporated into GNNs while maintaining the equivariant property. Furthermore, we offer theoretical insights into SEGNO, highlighting that it can learn a unique trajectory between adjacent states, which is crucial for model generalization. Additionally, we prove that the discrepancy between this learned trajectory of SEGNO and the true trajectory is bounded. Extensive experiments on complex dynamical systems including molecular dynamics and motion capture demonstrate that our model yields a significant improvement over the state-of-the-art baselines."
                    }
                ]
            },
            "77b1c44f-30f1-4ce8-b5ae-3df40f6e97ea": {
                "pk": "77b1c44f-30f1-4ce8-b5ae-3df40f6e97ea",
                "name": "Xiangguo Sun",
                "collaborators": [
                    "Hong Cheng",
                    "Hongxu Chen",
                    "Guandong Xu",
                    "Jia Li",
                    "Sixiao Zhang",
                    "Yicong Li",
                    "Hongzhi Yin",
                    "Bo Liu",
                    "Jihong Guan",
                    "Yun Xiong"
                ],
                "domain": [
                    "Graph Neural Network",
                    "Personality Analysis",
                    "Explainable AI",
                    "Large Language Models"
                ],
                "publications": [
                    {
                        "title": "When LLM Meets Hypergraph: A Sociological Analysis on Personality via Online Social Networks",
                        "abstract": "Individual personalities significantly influence our perceptions, decisions, and social interactions, which is particularly crucial for gaining insights into human behavior patterns in online social network analysis. Many psychological studies have observed that personalities are strongly reflected in their social behaviors and social environments. In light of these problems, this paper proposes a sociological analysis framework for one's personality in an environment-based view instead of individual-level data mining. Specifically, to comprehensively understand an individual's behavior from low-quality records, we leverage the powerful associative ability of LLMs by designing an effective prompt. In this way, LLMs can integrate various scattered information with their external knowledge to generate higher-quality profiles, which can significantly improve the personality analysis performance. To explore the interactive mechanism behind the users and their online environments, we design an effective hypergraph neural network where the hypergraph nodes are users and the hyperedges in the hypergraph are social environments. We offer a useful dataset with user profile data, personality traits, and several detected environments from the real-world social platform. To the best of our knowledge, this is the first network-based dataset containing both hypergraph structure and social information, which could push forward future research in this area further. By employing the framework on this dataset, we can effectively capture the nuances of individual personalities and their online behaviors, leading to a deeper understanding of human interactions in the digital world."
                    },
                    {
                        "title": "Does Graph Prompt Work? A Data Operation Perspective with Theoretical Analysis",
                        "abstract": "In recent years, graph prompting has emerged as a promising research direction, enabling the learning of additional tokens or subgraphs appended to the original graphs without requiring retraining of pre-trained graph models across various applications. This novel paradigm, shifting from the traditional pretraining and finetuning to pretraining and prompting has shown significant empirical success in simulating graph data operations, with applications ranging from recommendation systems to biological networks and graph transferring. However, despite its potential, the theoretical underpinnings of graph prompting remain underexplored, raising critical questions about its fundamental effectiveness. The lack of rigorous theoretical proof of why and how much it works is more like a dark cloud over the graph prompt area to go further. To fill this gap, this paper introduces a theoretical framework that rigorously analyzes graph prompting from a data operation perspective. Our contributions are threefold: First, we provide a formal guarantee theorem, demonstrating graph prompts capacity to approximate graph transformation operators, effectively linking upstream and downstream tasks. Second, we derive upper bounds on the error of these data operations by graph prompts for a single graph and extend this discussion to batches of graphs, which are common in graph model training. Third, we analyze the distribution of data operation errors, extending our theoretical findings from linear graph models (e.g., GCN) to non-linear graph models (e.g., GAT). Extensive experiments support our theoretical results and confirm the practical implications of these guarantees."
                    },
                    {
                        "title": "Temporal Meta-path Guided Explainable Recommendation",
                        "abstract": "This paper utilizes well-designed item-item path modelling between consecutive items with attention mechanisms to sequentially model dynamic user-item evolutions on dynamic knowledge graph for explainable recommendations. Compared with existing works that use heavy recurrent neural networks to model temporal information, we propose simple but effective neural networks to capture user historical item features and path-based context to characterise next purchased item. Extensive evaluations of TMER on three real-world benchmark datasets show state-of-the-art performance compared against recent strong baselines."
                    },
                    {
                        "title": "All in One: Multi-Task Prompting for Graph Neural Networks (Extended Abstract)",
                        "abstract": "This paper is an extended abstract of our original work published in KDD23, where we won the best research paper award (Xiangguo Sun, Hong Cheng, Jia Li, Bo Liu, and Jihong Guan. All in one: Multi-task prompting for graph neural networks. KDD 23) The paper introduces a novel approach to bridging the gap between pre-trained graph models and the diverse tasks they're applied to, inspired by the success of prompt learning in NLP. Recognizing the challenge of aligning pre-trained models with varied graph tasks (node level, edge level, and graph level), which can lead to negative transfer and poor performance, we propose a multi-task prompting method for graphs. This method involves unifying graph and language prompt formats, enabling NLP's prompting strategies to be adapted for graph tasks. By analyzing the task space of graph applications, we reformulate problems to fit graph-level tasks and apply meta-learning to improve prompt initialization for multiple tasks. Experiments show our method's effectiveness in enhancing model performance across different graph tasks.   Beyond the original work, in this extended abstract, we further discuss the graph prompt from a bigger picture and provide some of the latest work toward this area."
                    },
                    {
                        "title": "All in One: Multi-task Prompting for Graph Neural Networks",
                        "abstract": "Recently, ''pre-training and fine-tuning'' has been adopted as a standard workflow for many graph tasks since it can take general graph knowledge to relieve the lack of graph annotations from each application. However, graph tasks with node level, edge level, and graph level are far diversified, making the pre-training pretext often incompatible with these multiple tasks. This gap may even cause a ''negative transfer'' to the specific application, leading to poor results. Inspired by the prompt learning in natural language processing (NLP), which has presented significant effectiveness in leveraging prior knowledge for various NLP tasks, we study the prompting topic for graphs with the motivation of filling the gap between pre-trained models and various graph tasks. In this paper, we propose a novel multi-task prompting method for graph models. Specifically, we first unify the format of graph prompts and language prompts with the prompt token, token structure, and inserting pattern. In this way, the prompting idea from NLP can be seamlessly introduced to the graph area. Then, to further narrow the gap between various graph tasks and state-of-the-art pre-training strategies, we further study the task space of various graph applications and reformulate downstream problems to the graph-level task. Afterward, we introduce meta-learning to efficiently learn a better initialization for the multi-task prompt of graphs so that our prompting framework can be more reliable and general for different tasks. We conduct extensive experiments, results from which demonstrate the superiority of our method."
                    },
                    {
                        "title": "Advanced Drug Interaction Event Prediction",
                        "abstract": "Predicting drug-drug interaction adverse events, so-called DDI events, is increasingly valuable as it facilitates the study of mechanisms underlying drug use or adverse reactions. Existing models often neglect the distinctive characteristics of individual event classes when integrating multi-source features, which contributes to systematic unfairness when dealing with highly imbalanced event samples. Moreover, the limited capacity of these models to abstract the unique attributes of each event subclass considerably hampers their application in predicting rare drug-drug interaction events with a limited sample size. Reducing dataset bias and abstracting event subclass characteristics are two unresolved challenges. Recently, prompt tuning with frozen pre-trained graph models, namely \"pre-train, prompt, fine-tune\" strategy, has demonstrated impressive performance in few-shot tasks. Motivated by this, we propose an advanced method as a solution to address these aforementioned challenges. Specifically, our proposed approach entails a hierarchical pre-training task that aims to capture crucial aspects of drug molecular structure and intermolecular interactions while effectively mitigating implicit dataset bias within the node embeddings. Furthermore, we construct a prototypical graph by strategically sampling data from distinct event types and design subgraph prompts utilizing pre-trained node features. Through comprehensive benchmark experiments, we validate the efficacy of our subgraph prompts in accurately representing event classes and achieve exemplary results in both overall and subclass prediction tasks."
                    },
                    {
                        "title": "Unsupervised Graph Poisoning Attack via Contrastive Loss Back-propagation",
                        "abstract": "Graph contrastive learning is the state-of-the-art unsupervised graph representation learning framework and has shown comparable performance with supervised approaches. However, evaluating whether the graph contrastive learning is robust to adversarial attacks is still an open problem because most existing graph adversarial attacks are supervised models, which means they heavily rely on labels and can only be used to evaluate the graph contrastive learning in a specific scenario. For unsupervised graph representation methods such as graph contrastive learning, it is difficult to acquire labels in real-world scenarios, making traditional supervised graph attack methods difficult to be applied to test their robustness. In this paper, we propose a novel unsupervised gradient-based adversarial attack that does not rely on labels for graph contrastive learning. We compute the gradients of the adjacency matrices of the two views and flip the edges with gradient ascent to maximize the contrastive loss. In this way, we can fully use multiple views generated by the graph contrastive learning models and pick the most informative edges without knowing their labels, and therefore can promisingly support our model adapted to more kinds of downstream tasks. Extensive experiments show that our attack outperforms unsupervised baseline attacks and has comparable performance with supervised attacks in multiple downstream tasks including node classification and link prediction. We further show that our attack can be transferred to other graph representation models as well."
                    },
                    {
                        "title": "Attention Is Not the Only Choice: Counterfactual Reasoning for Path-Based Explainable Recommendation",
                        "abstract": "Compared with only pursuing recommendation accuracy, the explainability of a recommendation model has drawn more attention in recent years. Many graph-based recommendations resort to informative paths with the attention mechanism for the explanation. Unfortunately, these attention weights are intentionally designed for model accuracy but not explainability. Recently, some researchers have started to question attention-based explainability because the attention weights are unstable for different reproductions, and they may not always align with human intuition. Inspired by the counterfactual reasoning from causality learning theory, we propose a novel explainable framework targeting path-based recommendations, wherein the explainable weights of paths are learned to replace attention weights. Specifically, we design two counterfactual reasoning algorithms from both path representation and path topological structure perspectives. Moreover, unlike traditional case studies, we also propose a package of explainability evaluation solutions with both qualitative and quantitative methods. We conduct extensive experiments on three real-world datasets, the results of which further demonstrate the effectiveness and reliability of our method."
                    },
                    {
                        "title": "Generating Counterfactual Hard Negative Samples for Graph Contrastive Learning",
                        "abstract": "Graph contrastive learning has emerged as a powerful tool for unsupervised graph representation learning. The key to the success of graph contrastive learning is to acquire high-quality positive and negative samples as contrasting pairs for the purpose of learning underlying structural semantics of the input graph. Recent works usually sample negative samples from the same training batch with the positive samples, or from an external irrelevant graph. However, a significant limitation lies in such strategies, which is the unavoidable problem of sampling false negative samples. In this paper, we propose a novel method to utilize \\textbf{C}ounterfactual mechanism to generate artificial hard negative samples for \\textbf{G}raph \\textbf{C}ontrastive learning, namely \\textbf{CGC}, which has a different perspective compared to those sampling-based strategies. We utilize counterfactual mechanism to produce hard negative samples, which ensures that the generated samples are similar to, but have labels that different from the positive sample. The proposed method achieves satisfying results on several datasets compared to some traditional unsupervised graph learning methods and some SOTA graph contrastive learning methods. We also conduct some supplementary experiments to give an extensive illustration of the proposed method, including the performances of CGC with different hard negative samples and evaluations for hard negative samples generated with different similarity measurements."
                    },
                    {
                        "title": "Self-supervised Hypergraph Representation Learning for Sociological Analysis",
                        "abstract": "Modern sociology has profoundly uncovered many convincing social criteria for behavioural analysis. Unfortunately, many of them are too subjective to be measured and presented in online social networks. On the other hand, data mining techniques can better find data patterns but many of them leave behind unnatural understanding. In this paper, we propose a fundamental methodology to support the further fusion of data mining techniques and sociological behavioral criteria. Our highlights are three-fold: First, we propose an effective hypergraph awareness and a fast line graph construction framework. The hypergraph can more profoundly indicate the interactions between individuals and their environments because each edge in the hypergraph (a.k.a hyperedge) contains more than two nodes, which is perfect to describe social environments. A line graph treats each social environment as a super node with the underlying influence between different environments. In this way, we go beyond traditional pair-wise relations and explore richer patterns under various sociological criteria; Second, we propose a novel hypergraph-based neural network to learn social influence flowing from users to users, users to environments, environment to users, and environments to environments. The neural network can be learned via a task-free method, making our model very flexible to support various data mining tasks and sociological analysis; Third, we propose both qualitative and quantitive solutions to effectively evaluate the most common sociological criteria like social conformity, social equivalence, environmental evolving and social polarization. Our extensive experiments show that our framework can better support both data mining tasks for online user behaviours and sociological analysis."
                    },
                    {
                        "title": "Counter-Empirical Attacking based on Adversarial Reinforcement Learning for Time-Relevant Scoring System",
                        "abstract": "Scoring systems are commonly seen for platforms in the era of big data. From credit scoring systems in financial services to membership scores in E-commerce shopping platforms, platform managers use such systems to guide users towards the encouraged activity pattern, and manage resources more effectively and more efficiently thereby. To establish such scoring systems, several \"empirical criteria\" are firstly determined, followed by dedicated top-down design for each factor of the score, which usually requires enormous effort to adjust and tune the scoring function in the new application scenario. What's worse, many fresh projects usually have no ground-truth or any experience to evaluate a reasonable scoring system, making the designing even harder. To reduce the effort of manual adjustment of the scoring function in every new scoring system, we innovatively study the scoring system from the preset empirical criteria without any ground truth, and propose a novel framework to improve the system from scratch. In this paper, we propose a \"counter-empirical attacking\" mechanism that can generate \"attacking\" behavior traces and try to break the empirical rules of the scoring system. Then an adversarial \"enhancer\" is applied to evaluate the scoring system and find the improvement strategy. By training the adversarial learning problem, a proper scoring function can be learned to be robust to the attacking activity traces that are trying to violate the empirical criteria. Extensive experiments have been conducted on two scoring systems including a shared computing resource platform and a financial credit system. The experimental results have validated the effectiveness of our proposed framework."
                    },
                    {
                        "title": "Graph Prompt Learning: A Comprehensive Survey and Beyond",
                        "abstract": "Artificial General Intelligence (AGI) has revolutionized numerous fields, yet its integration with graph data, a cornerstone in our interconnected world, remains nascent. This paper presents a pioneering survey on the emerging domain of graph prompts in AGI, addressing key challenges and opportunities in harnessing graph data for AGI applications. Despite substantial advancements in AGI across natural language processing and computer vision, the application to graph data is relatively underexplored. This survey critically evaluates the current landscape of AGI in handling graph data, highlighting the distinct challenges in cross-modality, cross-domain, and cross-task applications specific to graphs. Our work is the first to propose a unified framework for understanding graph prompt learning, offering clarity on prompt tokens, token structures, and insertion patterns in the graph domain. We delve into the intrinsic properties of graph prompts, exploring their flexibility, expressiveness, and interplay with existing graph models. A comprehensive taxonomy categorizes over 100 works in this field, aligning them with pre-training tasks across node-level, edge-level, and graph-level objectives. Additionally, we present, ProG, a Python library, and an accompanying website, to support and advance research in graph prompting. The survey culminates in a discussion of current challenges and future directions, offering a roadmap for research in graph prompting within AGI. Through this comprehensive analysis, we aim to catalyze further exploration and practical applications of AGI in graph data, underlining its potential to reshape AGI fields and beyond. ProG and the website can be accessed by \\url{https://github.com/WxxShirley/Awesome-Graph-Prompt}, and \\url{https://github.com/sheldonresearch/ProG}, respectively."
                    },
                    {
                        "title": "Multi-level Graph Convolutional Networks for Cross-platform Anchor Link Prediction",
                        "abstract": "Cross-platform account matching plays a significant role in social network analytics, and is beneficial for a wide range of applications. However, existing methods either heavily rely on high-quality user generated content (including user profiles) or suffer from data insufficiency problem if only focusing on network topology, which brings researchers into an insoluble dilemma of model selection. In this paper, to address this problem, we propose a novel framework that considers multi-level graph convolutions on both local network structure and hypergraph structure in a unified manner. The proposed method overcomes data insufficiency problem of existing work and does not necessarily rely on user demographic information. Moreover, to adapt the proposed method to be capable of handling large-scale social networks, we propose a two-phase space reconciliation mechanism to align the embedding spaces in both network partitioning based parallel training and account matching across different social networks. Extensive experiments have been conducted on two large-scale real-life social networks. The experimental results demonstrate that the proposed method outperforms the state-of-the-art models with a big margin."
                    },
                    {
                        "title": "Graph Masked Autoencoders with Transformers",
                        "abstract": "Recently, transformers have shown promising performance in learning graph representations. However, there are still some challenges when applying transformers to real-world scenarios due to the fact that deep transformers are hard to train from scratch and the quadratic memory consumption w.r.t. the number of nodes. In this paper, we propose Graph Masked Autoencoders (GMAEs), a self-supervised transformer-based model for learning graph representations. To address the above two challenges, we adopt the masking mechanism and the asymmetric encoder-decoder design. Specifically, GMAE takes partially masked graphs as input, and reconstructs the features of the masked nodes. The encoder and decoder are asymmetric, where the encoder is a deep transformer and the decoder is a shallow transformer. The masking mechanism and the asymmetric design make GMAE a memory-efficient model compared with conventional transformers. We show that, when serving as a conventional self-supervised graph representation model, GMAE achieves state-of-the-art performance on both the graph classification task and the node classification task under common downstream evaluation protocols. We also show that, compared with training in an end-to-end manner from scratch, we can achieve comparable performance after pre-training and fine-tuning using GMAE while simplifying the training process."
                    },
                    {
                        "title": "All in One and One for All: A Simple yet Effective Method towards Cross-domain Graph Pretraining",
                        "abstract": "Large Language Models (LLMs) have revolutionized the fields of computer vision (CV) and natural language processing (NLP). One of the most notable advancements of LLMs is that a single model is trained on vast and diverse datasets spanning multiple domains -- a paradigm we term `All in One'. This methodology empowers LLMs with super generalization capabilities, facilitating an encompassing comprehension of varied data distributions. Leveraging these capabilities, a single LLM demonstrates remarkable versatility across a variety of domains -- a paradigm we term `One for All'. However, applying this idea to the graph field remains a formidable challenge, with cross-domain pretraining often resulting in negative transfer. This issue is particularly important in few-shot learning scenarios, where the paucity of training data necessitates the incorporation of external knowledge sources. In response to this challenge, we propose a novel approach called Graph COordinators for PrEtraining (GCOPE), that harnesses the underlying commonalities across diverse graph datasets to enhance few-shot learning. Our novel methodology involves a unification framework that amalgamates disparate graph datasets during the pretraining phase to distill and transfer meaningful knowledge to target tasks. Extensive experiments across multiple graph datasets demonstrate the superior efficacy of our approach. By successfully leveraging the synergistic potential of multiple graph datasets for pretraining, our work stands as a pioneering contribution to the realm of graph foundational model."
                    }
                ]
            },
            "39b062f3-6e9c-4e57-af33-c9c926472638": {
                "pk": "39b062f3-6e9c-4e57-af33-c9c926472638",
                "name": "Yiqing Lin",
                "collaborators": [
                    "Kun Xu",
                    "Junjian Yang",
                    "Lingqi Gu",
                    "Zihao Gu",
                    "Xinpeng Li",
                    "Xuepeng Bai",
                    "Xiangyun Lin",
                    "Abdoulaye Soumana Hima",
                    "Falei Wang",
                    "Hui Zhao"
                ],
                "domain": [
                    "Stochastic Differential Equations",
                    "Backward Stochastic Differential Equations",
                    "Utility Maximization",
                    "Financial Mathematics"
                ],
                "publications": [
                    {
                        "title": "A New Result for Second Order BSDEs with Quadratic Growth and its Applications",
                        "abstract": "In this paper, we study a class of second order backward stochastic differential equations (2BSDEs) with quadratic growth in coefficients. We first establish solvability for such 2BSDEs and then give their applications to robust utility maximization problems."
                    },
                    {
                        "title": "Stochastic differential equations driven by $G$-Brownian motion with reflecting boundary conditions",
                        "abstract": "In this paper, we introduce the idea of stochastic integrals with respect to an increasing process in the $G$-framework and extend $G$-It\\^o's formula. Moreover, we study the solvability of the scalar valued stochastic differential equations driven by $G$-Brownian motion with reflecting boundary conditions (RGSDEs)."
                    },
                    {
                        "title": "Generalized Wasserstein distance and weak convergence of sublinear expectations",
                        "abstract": "In this paper, we define the generalized Wasserstein distance for sets of Borel probability measures and demonstrate that the weak convergence of sublinear expectations can be characterized by means of this distance."
                    },
                    {
                        "title": "On the existence and uniqueness of solutions to stochastic differential equations driven by G-Brownian motion with integral-Lipschitz coefficients",
                        "abstract": "In this paper, we study the existence and uniqueness of solutions to stochastic differential equations driven by G-Brownian motion (GSDEs) with integral-Lipschitz conditions on their coefficients."
                    },
                    {
                        "title": "Mean-field reflected BSDEs driven by a marked point process",
                        "abstract": "In this paper, we study a class of mean-field reflected backward stochastic differential equations (MFRBSDEs) driven by a marked point process. Based on a g-expectation representation lemma, we give the existence and uniqueness of MFRBSDEs driven by a marked point process under Lipschitz generator conditions. Besides, the well-posedness of this kind of BSDEs with exponential growth generator and unbounded terminal is also provided by $\\theta$-method."
                    },
                    {
                        "title": "Particle systems for mean reflected BSDEs with jumps",
                        "abstract": "In this paper, we study the mean reflected backward stochastic differential equations with jump (BSDEJs). We extend the work of Briand and Hibon on the propagation of chaos for mean reflected BSDEs \\cite{briand2021particles} to the jump framework. Besides, we study the reflections for the particle system and obtain the rate of of convergence of the particle system towards the deterministic flat solution to the mean reflected BSDEJ."
                    },
                    {
                        "title": "Utility maximization problem with random endowment and transaction costs: when wealth may become negative",
                        "abstract": "In this paper we study the problem of maximizing expected utility from the terminal wealth with proportional transaction costs and random endowment. In the context of the existence of consistent price systems, we consider the duality between the primal utility maximization problem and the dual one, which is set up on the domain of finitely additive measures. In particular, we prove duality results for utility functions supporting possibly negative values. Moreover, we construct the shadow market by the dual optimal process and consider the utility based pricing for random endowment."
                    },
                    {
                        "title": "Mean-field reflected BSDEs with jumps",
                        "abstract": "In this paper, we study a class of mean-field reflected backward stochastic differential equations (MF-RBSDEs) driven by a marked point process and also analyze MF-RBSDEs driven by a poission process. Based on a g-expectation representation lemma, we give the existence and uniqueness of the particle system of MF-RBSDEs driven by a marked point process under Lipschitz generator conditions and obtain a convergence of this system. We also establish the well-posedness of the MF-RBSDEs driven by a poission process and the rate of of convergence of the corresponding particle system towards the solution to the MF-RBSDEs driven by a poission process under bounded terminal, bounded obstacle conditions."
                    },
                    {
                        "title": "Non-concave utility maximization problem with transaction costs and a given consistent price system",
                        "abstract": "We investigate expected utility maximization problems from the terminal liquidation value in continuous time in markets with transaction costs and one fixed consistent price system, where a non-concave utility function is defined on the positive half real line. The sufficient conditions are given by the convex conjugate of the utility function, then the existence of the optimizer is proved by a maximizing sequence. Finally, we show that the value function of the envelope of the utility function and the concave envelope of the value function are coincide."
                    },
                    {
                        "title": "Lyapunov-type conditions and stochastic differential equations driven by $G$-Brownian motion",
                        "abstract": "This paper studies the solvability and the stability of stochastic differential equations driven by G-Brownian motion (GSDEs). In particular, the existence and uniqueness of the solution for locally Lipschitz GSDEs is obtained by localization methods, also the stability of such GSDEs are discussed with Lyapunov-type conditions."
                    },
                    {
                        "title": "On the dual problem of utility maximization in incomplete markets",
                        "abstract": "In this paper, we study the dual problem of the expected utility maximization in incomplete markets with bounded random endowment. We start with the problem formulated in the paper of Cvitani\\'{c}-Schachermayer-Wang (2001) and prove the following statement: in the Brownian framework, the countably additive part $Q^r$ of the dual optimizer $Q\\in (L^\\infty)^*$ obtained in that paper can be represented by the terminal value of a supermartingale deflator $Y$ defined in the paper of Kramkov-Schachermayer (1999), which is a local martingale."
                    },
                    {
                        "title": "On the existence of shadow prices for optimal investment with random endowment",
                        "abstract": "In this paper, we consider a num\\'eraire-based utility maximization problem under constant proportional transaction costs and random endowment. Assuming that the agent cannot short sell assets and is endowed with a strictly positive contingent claim, a primal optimizer of this utility maximization problem exists. Moreover, we observe that the original market with transaction costs can be replaced by a frictionless shadow market that yields the same optimality. On the other hand, we present an example to show that in some case when these constraints are relaxed, the existence of shadow prices is still warranted."
                    },
                    {
                        "title": "Reflected stochastic differential equations driven by $G$-Brownian motion in non-convex domains",
                        "abstract": "In this paper, we first review the penalization method for solving deterministic Skorokhod problems in non-convex domains and establish estimates for problems with $\\alpha$-H\\\"older continuous functions. With the help of these results obtained previously for deterministic problems, we pathwisely define the reflected $G$-Brownian motion and prove its existence and uniqueness in a Banach space. Finally, multi-dimensional reflected stochastic differential equations driven by $G$-Brownian motion are investigated via a fixed-point argument."
                    },
                    {
                        "title": "Exponential growth BSDE driven by a marked point process",
                        "abstract": "In this study, we investigate the well-posedness of exponential growth backward stochastic differential equations (BSDEs) driven by a marked point process (MPP) under unbounded terminal conditions. Our analysis utilizes a fixed-point argument, the $\\theta$-method, and an approximation procedure. Additionally, we establish the solvability of mean-reflected exponential growth BSDEs driven by the MPP using the $\\theta$-method."
                    },
                    {
                        "title": "Mean-reflected $G$-BSDEs with multi-variate constraints",
                        "abstract": "In this paper, we study the multi-dimensional reflected backward stochastic differential equation driven by $G$-Brownian motion ($G$-BSDE) with a multi-variate constraint on the $G$-expectation of its solution. The generators are diagonally dependent on $Z$ and on all $Y$-components. We obtain the existence and uniqueness result via a fixed-point argumentation."
                    },
                    {
                        "title": "A note on utility maximization with transaction costs and random endoment: num\u00e9raire-based model and convex duality",
                        "abstract": "In this note, we study the utility maximization problem on the terminal wealth under proportional transaction costs and bounded random endowment. In particular, we restrict ourselves to the num\\'eraire-based model and work with utility functions only supporting R+. Under the assumption of existence of consistent price systems and natural regularity conditions, standard convex duality results are established. Precisely, we first enlarge the dual domain from the collection of martingale densities associated with consistent price systems to a set of finitely additive measures; then the dual formulation of the utility maximization problem can be regarded as an extension of the paper of Cvitani\\'c-Schachermayer-Wang (2001) to the context under proportional transaction costs."
                    },
                    {
                        "title": "Utility maximization problem under transaction costs: optimal dual processes and stability",
                        "abstract": "This paper discusses the num\\'eraire-based utility maximization problem in markets with proportional transaction costs. In particular, the investor is required to liquidate all her position in stock at the terminal time. We first observe the stability of the primal and dual value functions as well as the convergence of the primal and dual optimizers when perturbations occur on the utility function and on the physical probability. We then study the properties of the optimal dual process (ODP), that is, a process from the dual domain that induces the optimality of the dual problem. When the market is driven by a continuous process, we construct the ODP for the problem in the limiting market by a sequence of ODPs corresponding to the problems with small misspecificated parameters. Moreover, we prove that this limiting ODP defines a shadow price."
                    },
                    {
                        "title": "Second order backward SDE with random terminal time",
                        "abstract": "Backward stochastic differential equations extend the martingale representation theorem to the nonlinear setting. This can be seen as path-dependent counterpart of the extension from the heat equation to fully nonlinear parabolic equations in the Markov setting. This paper extends such a nonlinear representation to the context where the random variable of interest is measurable with respect to the information at a finite stopping time. We provide a complete wellposedness theory which covers the semilinear case (backward SDE), the semilinear case with obstacle (reflected backward SDE), and the fully nonlinear case (second order backward SDE)."
                    },
                    {
                        "title": "Mean reflected BSDE driven by a marked point process and application in insurance risk management",
                        "abstract": "This paper aims to solve a super-hedging problem along with insurance re-payment under running risk management constraints. The initial endowment for the super-heding problem is characterized by a class of mean reflected backward stochastic differential equation driven by a marked point process (MPP) and a Brownian motion. By Lipschitz assumptions on the generators and proper integrability on the terminal value, we give the well-posedness of this kind of BSDEs by combining a representation theorem with the fixed point argument."
                    },
                    {
                        "title": "Reflected BSDE driven by a marked point process with a convex/concave generator",
                        "abstract": "In this paper, a class of reflected backward stochastic differential equations (RBSDE) driven by a marked point process (MPP) with a convex/concave generator is studied. Based on fixed point argument, $\\theta$-method and truncation technique, the well-posedness of this kind of RBSDE with unbounded terminal condition and obstacle is investigated. Besides, we present an application on the pricing of American options via utility maximization, which is solved by constructing an RBSDE with a convex generator."
                    }
                ]
            },
            "6f936a13-19db-441e-be4c-6f935522ccc5": {
                "pk": "6f936a13-19db-441e-be4c-6f935522ccc5",
                "name": "Hong Cheng",
                "collaborators": [
                    "Xinyue Cheng",
                    "Pengsheng Wu",
                    "Rebekah Leamons",
                    "Ahmad Al Shami",
                    "Fanhua Shang",
                    "James Cheng"
                ],
                "domain": [
                    "Geometry",
                    "Topological Data Analysis",
                    "Tensor Decomposition",
                    "Machine Learning"
                ],
                "publications": [
                    {
                        "title": "The characterizations on a class of weakly weighted Einstein-Finsler metrics",
                        "abstract": "In this paper, we study the weakly weighted Einstein-Finsler metrics. First, we show that weakly weighted Einstein-Kropina metrics must be of isotropic S-curvature with respect to the Busemann-Hausdorff volume form under a certain condition about the weight constants. Then we characterize weakly weighted Einstein-Kropina metrics completely via their navigation expressions or via $\\alpha$ and $\\beta$ respectively."
                    },
                    {
                        "title": "Modified Topological Image Preprocessing for Skin Lesion Classifications",
                        "abstract": "This paper proposes a modified Topological Data Analysis model for skin images preprocessing and enhancements. The skin lesion dataset HAM10000 used with the intention of identifying the important objects in relevant regions of the images. In order to evaluate both the original dataset and the preprocessed dataset, Deep Convolutional Neural Network and Vision Transformer models were utilized to train both models. After training, the experimental results demonstrate that the images preprocessed using the Modified Topological Data Analysis consistently perform better."
                    },
                    {
                        "title": "Regularized Orthogonal Tensor Decompositions for Multi-Relational Learning",
                        "abstract": "Multi-relational learning has received lots of attention from researchers in various research communities. Most existing methods either suffer from superlinear per-iteration cost, or are sensitive to the given ranks. To address both issues, we propose a scalable core tensor trace norm Regularized Orthogonal Iteration Decomposition (ROID) method for full or incomplete tensor analytics, which can be generalized as a graph Laplacian regularized version by using auxiliary information or a sparse higher-order orthogonal iteration (SHOOI) version. We first induce the equivalence relation of the Schatten p-norm (0<p<\\infty) of a low multi-linear rank tensor and its core tensor. Then we achieve a much smaller matrix trace norm minimization problem. Finally, we develop two efficient augmented Lagrange multiplier algorithms to solve our problems with convergence guarantees. Extensive experiments using both real and synthetic datasets, even though with only a few observations, verified both the efficiency and effectiveness of our methods."
                    }
                ]
            },
            "8e9269f1-cb76-4034-9b11-41a73e0f27e6": {
                "pk": "8e9269f1-cb76-4034-9b11-41a73e0f27e6",
                "name": "Jia Li",
                "collaborators": [
                    "Huaibo Huang",
                    "Xiaofei Jia",
                    "Ran He",
                    "Xiang Li"
                ],
                "domain": [
                    "Causal Inference",
                    "Multiple Zeta Functions",
                    "Image Restoration",
                    "Representation Learning"
                ],
                "publications": [
                    {
                        "title": "A continuous version of multiple zeta values with double variables",
                        "abstract": "In this paper we define a continuous version of multiple zeta functions with double variables. They can be analytically continued to meromorphic functions on $\\mathbb{C}^r$ with only simple poles at some special hyperplanes. The evaluations of these functions at positive integers (continuous multiple zeta values) satisfy the first shuffle product and the second shuffle product. We proved that the dimension of the $\\mathbb{Q}-$linear spaces generated by continuous multiple zeta values with given weight are finite. By using a theorem of C.Glanois, we proved that continuous multiple zeta values include all cyclotomic multiple zeta values of level 2. We will give a detail analysis about the two different shuffle products. Furthermore, we will discuss the extension of the two different products, we proved a theorem about comparing the two different shuffle product, this is an analogy of Ihara-Kaneko-Zagier's comparison theorem in the case of continuous multiple zeta values. As an application, we will give a new method to proof some Ramanujan's identities. Finally, we will provide some conjectures."
                    },
                    {
                        "title": "Universal Face Restoration With Memorized Modulation",
                        "abstract": "Blind face restoration (BFR) is a challenging problem because of the uncertainty of the degradation patterns. This paper proposes a Restoration with Memorized Modulation (RMM) framework for universal BFR in diverse degraded scenes and heterogeneous domains. We apply random noise as well as unsupervised wavelet memory to adaptively modulate the face-enhancement generator, considering attentional denormalization in both layer and instance levels. Specifically, in the training stage, the low-level spatial feature embedding, the wavelet memory embedding obtained by wavelet transformation of the high-resolution image, as well as the disentangled high-level noise embeddings are integrated, with the guidance of attentional maps generated from layer normalization, instance normalization, and the original feature map. These three embeddings are respectively associated with the spatial content, high-frequency texture details, and a learnable universal prior against other blind image degradation patterns. We store the spatial feature of the low-resolution image and the corresponding wavelet style code as key and value in the memory unit, respectively. In the test stage, the wavelet memory value whose corresponding spatial key is the most matching with that of the inferred image is retrieved to modulate the generator. Moreover, the universal prior learned from the random noise has been memorized by training the modulation network. Experimental results show the superiority of the proposed method compared with the state-of-the-art methods, and a good generalization in the wild."
                    },
                    {
                        "title": "Relation-First Modeling Paradigm for Causal Representation Learning toward the Development of AGI",
                        "abstract": "The traditional i.i.d.-based learning paradigm faces inherent challenges in addressing causal relationships, which has become increasingly evident with the rise of applications in causal representation learning.   Our understanding of causality naturally requires a perspective as the creator rather than observer, as the ``what...if'' questions only hold within the possible world we conceive. The traditional perspective limits capturing dynamic causal outcomes and leads to compensatory efforts such as the reliance on hidden confounders. This paper lays the groundwork for the new perspective, which enables the \\emph{relation-first} modeling paradigm for causality. Also, it introduces the Relation-Indexed Representation Learning (RIRL) as a practical implementation, supported by experiments that validate its efficacy."
                    }
                ]
            }
        }
    }
}